Adding upstream version 5.10.209.upstream/5.10.209

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

46 files changed, 28871 insertions, 0 deletions

diff --git a/fs/ntfs/Kconfig b/fs/ntfs/Kconfig
new file mode 100644
index 000000000..1667a7e59
--- /dev/null
+++ b/fs/ntfs/Kconfig
@@ -0,0 +1,79 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config NTFS_FS
+	tristate "NTFS file system support"
+	select NLS
+	help
+	  NTFS is the file system of Microsoft Windows NT, 2000, XP and 2003.
+
+	  Saying Y or M here enables read support.  There is partial, but
+	  safe, write support available.  For write support you must also
+	  say Y to "NTFS write support" below.
+
+	  There are also a number of user-space tools available, called
+	  ntfsprogs.  These include ntfsundelete and ntfsresize, that work
+	  without NTFS support enabled in the kernel.
+
+	  This is a rewrite from scratch of Linux NTFS support and replaced
+	  the old NTFS code starting with Linux 2.5.11.  A backport to
+	  the Linux 2.4 kernel series is separately available as a patch
+	  from the project web site.
+
+	  For more information see <file:Documentation/filesystems/ntfs.rst>
+	  and <http://www.linux-ntfs.org/>.
+
+	  To compile this file system support as a module, choose M here: the
+	  module will be called ntfs.
+
+	  If you are not using Windows NT, 2000, XP or 2003 in addition to
+	  Linux on your computer it is safe to say N.
+
+config NTFS_DEBUG
+	bool "NTFS debugging support"
+	depends on NTFS_FS
+	help
+	  If you are experiencing any problems with the NTFS file system, say
+	  Y here.  This will result in additional consistency checks to be
+	  performed by the driver as well as additional debugging messages to
+	  be written to the system log.  Note that debugging messages are
+	  disabled by default.  To enable them, supply the option debug_msgs=1
+	  at the kernel command line when booting the kernel or as an option
+	  to insmod when loading the ntfs module.  Once the driver is active,
+	  you can enable debugging messages by doing (as root):
+	  echo 1 > /proc/sys/fs/ntfs-debug
+	  Replacing the "1" with "0" would disable debug messages.
+
+	  If you leave debugging messages disabled, this results in little
+	  overhead, but enabling debug messages results in very significant
+	  slowdown of the system.
+
+	  When reporting bugs, please try to have available a full dump of
+	  debugging messages while the misbehaviour was occurring.
+
+config NTFS_RW
+	bool "NTFS write support"
+	depends on NTFS_FS
+	help
+	  This enables the partial, but safe, write support in the NTFS driver.
+
+	  The only supported operation is overwriting existing files, without
+	  changing the file length.  No file or directory creation, deletion or
+	  renaming is possible.  Note only non-resident files can be written to
+	  so you may find that some very small files (<500 bytes or so) cannot
+	  be written to.
+
+	  While we cannot guarantee that it will not damage any data, we have
+	  so far not received a single report where the driver would have
+	  damaged someones data so we assume it is perfectly safe to use.
+
+	  Note:  While write support is safe in this version (a rewrite from
+	  scratch of the NTFS support), it should be noted that the old NTFS
+	  write support, included in Linux 2.5.10 and before (since 1997),
+	  is not safe.
+
+	  This is currently useful with TopologiLinux.  TopologiLinux is run
+	  on top of any DOS/Microsoft Windows system without partitioning your
+	  hard disk.  Unlike other Linux distributions TopologiLinux does not
+	  need its own partition.  For more information see
+	  <http://topologi-linux.sourceforge.net/>
+
+	  It is perfectly safe to say N here.
diff --git a/fs/ntfs/Makefile b/fs/ntfs/Makefile
new file mode 100644
index 000000000..3e736572e
--- /dev/null
+++ b/fs/ntfs/Makefile
@@ -0,0 +1,15 @@
+# SPDX-License-Identifier: GPL-2.0
+# Rules for making the NTFS driver.
+
+obj-$(CONFIG_NTFS_FS) += ntfs.o
+
+ntfs-y := aops.o attrib.o collate.o compress.o debug.o dir.o file.o \
+	  index.o inode.o mft.o mst.o namei.o runlist.o super.o sysctl.o \
+	  unistr.o upcase.o
+
+ntfs-$(CONFIG_NTFS_RW) += bitmap.o lcnalloc.o logfile.o quota.o usnjrnl.o
+
+ccflags-y := -DNTFS_VERSION=\"2.1.32\"
+ccflags-$(CONFIG_NTFS_DEBUG)	+= -DDEBUG
+ccflags-$(CONFIG_NTFS_RW)	+= -DNTFS_RW
+
diff --git a/fs/ntfs/aops.c b/fs/ntfs/aops.c
new file mode 100644
index 000000000..bb0a43860
--- /dev/null
+++ b/fs/ntfs/aops.c
@@ -0,0 +1,1760 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/**
+ * aops.c - NTFS kernel address space operations and page cache handling.
+ *
+ * Copyright (c) 2001-2014 Anton Altaparmakov and Tuxera Inc.
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#include <linux/errno.h>
+#include <linux/fs.h>
+#include <linux/gfp.h>
+#include <linux/mm.h>
+#include <linux/pagemap.h>
+#include <linux/swap.h>
+#include <linux/buffer_head.h>
+#include <linux/writeback.h>
+#include <linux/bit_spinlock.h>
+#include <linux/bio.h>
+
+#include "aops.h"
+#include "attrib.h"
+#include "debug.h"
+#include "inode.h"
+#include "mft.h"
+#include "runlist.h"
+#include "types.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_end_buffer_async_read - async io completion for reading attributes
+ * @bh:		buffer head on which io is completed
+ * @uptodate:	whether @bh is now uptodate or not
+ *
+ * Asynchronous I/O completion handler for reading pages belonging to the
+ * attribute address space of an inode.  The inodes can either be files or
+ * directories or they can be fake inodes describing some attribute.
+ *
+ * If NInoMstProtected(), perform the post read mst fixups when all IO on the
+ * page has been completed and mark the page uptodate or set the error bit on
+ * the page.  To determine the size of the records that need fixing up, we
+ * cheat a little bit by setting the index_block_size in ntfs_inode to the ntfs
+ * record size, and index_block_size_bits, to the log(base 2) of the ntfs
+ * record size.
+ */
+static void ntfs_end_buffer_async_read(struct buffer_head *bh, int uptodate)
+{
+	unsigned long flags;
+	struct buffer_head *first, *tmp;
+	struct page *page;
+	struct inode *vi;
+	ntfs_inode *ni;
+	int page_uptodate = 1;
+
+	page = bh->b_page;
+	vi = page->mapping->host;
+	ni = NTFS_I(vi);
+
+	if (likely(uptodate)) {
+		loff_t i_size;
+		s64 file_ofs, init_size;
+
+		set_buffer_uptodate(bh);
+
+		file_ofs = ((s64)page->index << PAGE_SHIFT) +
+				bh_offset(bh);
+		read_lock_irqsave(&ni->size_lock, flags);
+		init_size = ni->initialized_size;
+		i_size = i_size_read(vi);
+		read_unlock_irqrestore(&ni->size_lock, flags);
+		if (unlikely(init_size > i_size)) {
+			/* Race with shrinking truncate. */
+			init_size = i_size;
+		}
+		/* Check for the current buffer head overflowing. */
+		if (unlikely(file_ofs + bh->b_size > init_size)) {
+			int ofs;
+			void *kaddr;
+
+			ofs = 0;
+			if (file_ofs < init_size)
+				ofs = init_size - file_ofs;
+			kaddr = kmap_atomic(page);
+			memset(kaddr + bh_offset(bh) + ofs, 0,
+					bh->b_size - ofs);
+			flush_dcache_page(page);
+			kunmap_atomic(kaddr);
+		}
+	} else {
+		clear_buffer_uptodate(bh);
+		SetPageError(page);
+		ntfs_error(ni->vol->sb, "Buffer I/O error, logical block "
+				"0x%llx.", (unsigned long long)bh->b_blocknr);
+	}
+	first = page_buffers(page);
+	spin_lock_irqsave(&first->b_uptodate_lock, flags);
+	clear_buffer_async_read(bh);
+	unlock_buffer(bh);
+	tmp = bh;
+	do {
+		if (!buffer_uptodate(tmp))
+			page_uptodate = 0;
+		if (buffer_async_read(tmp)) {
+			if (likely(buffer_locked(tmp)))
+				goto still_busy;
+			/* Async buffers must be locked. */
+			BUG();
+		}
+		tmp = tmp->b_this_page;
+	} while (tmp != bh);
+	spin_unlock_irqrestore(&first->b_uptodate_lock, flags);
+	/*
+	 * If none of the buffers had errors then we can set the page uptodate,
+	 * but we first have to perform the post read mst fixups, if the
+	 * attribute is mst protected, i.e. if NInoMstProteced(ni) is true.
+	 * Note we ignore fixup errors as those are detected when
+	 * map_mft_record() is called which gives us per record granularity
+	 * rather than per page granularity.
+	 */
+	if (!NInoMstProtected(ni)) {
+		if (likely(page_uptodate && !PageError(page)))
+			SetPageUptodate(page);
+	} else {
+		u8 *kaddr;
+		unsigned int i, recs;
+		u32 rec_size;
+
+		rec_size = ni->itype.index.block_size;
+		recs = PAGE_SIZE / rec_size;
+		/* Should have been verified before we got here... */
+		BUG_ON(!recs);
+		kaddr = kmap_atomic(page);
+		for (i = 0; i < recs; i++)
+			post_read_mst_fixup((NTFS_RECORD*)(kaddr +
+					i * rec_size), rec_size);
+		kunmap_atomic(kaddr);
+		flush_dcache_page(page);
+		if (likely(page_uptodate && !PageError(page)))
+			SetPageUptodate(page);
+	}
+	unlock_page(page);
+	return;
+still_busy:
+	spin_unlock_irqrestore(&first->b_uptodate_lock, flags);
+	return;
+}
+
+/**
+ * ntfs_read_block - fill a @page of an address space with data
+ * @page:	page cache page to fill with data
+ *
+ * Fill the page @page of the address space belonging to the @page->host inode.
+ * We read each buffer asynchronously and when all buffers are read in, our io
+ * completion handler ntfs_end_buffer_read_async(), if required, automatically
+ * applies the mst fixups to the page before finally marking it uptodate and
+ * unlocking it.
+ *
+ * We only enforce allocated_size limit because i_size is checked for in
+ * generic_file_read().
+ *
+ * Return 0 on success and -errno on error.
+ *
+ * Contains an adapted version of fs/buffer.c::block_read_full_page().
+ */
+static int ntfs_read_block(struct page *page)
+{
+	loff_t i_size;
+	VCN vcn;
+	LCN lcn;
+	s64 init_size;
+	struct inode *vi;
+	ntfs_inode *ni;
+	ntfs_volume *vol;
+	runlist_element *rl;
+	struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
+	sector_t iblock, lblock, zblock;
+	unsigned long flags;
+	unsigned int blocksize, vcn_ofs;
+	int i, nr;
+	unsigned char blocksize_bits;
+
+	vi = page->mapping->host;
+	ni = NTFS_I(vi);
+	vol = ni->vol;
+
+	/* $MFT/$DATA must have its complete runlist in memory at all times. */
+	BUG_ON(!ni->runlist.rl && !ni->mft_no && !NInoAttr(ni));
+
+	blocksize = vol->sb->s_blocksize;
+	blocksize_bits = vol->sb->s_blocksize_bits;
+
+	if (!page_has_buffers(page)) {
+		create_empty_buffers(page, blocksize, 0);
+		if (unlikely(!page_has_buffers(page))) {
+			unlock_page(page);
+			return -ENOMEM;
+		}
+	}
+	bh = head = page_buffers(page);
+	BUG_ON(!bh);
+
+	/*
+	 * We may be racing with truncate.  To avoid some of the problems we
+	 * now take a snapshot of the various sizes and use those for the whole
+	 * of the function.  In case of an extending truncate it just means we
+	 * may leave some buffers unmapped which are now allocated.  This is
+	 * not a problem since these buffers will just get mapped when a write
+	 * occurs.  In case of a shrinking truncate, we will detect this later
+	 * on due to the runlist being incomplete and if the page is being
+	 * fully truncated, truncate will throw it away as soon as we unlock
+	 * it so no need to worry what we do with it.
+	 */
+	iblock = (s64)page->index << (PAGE_SHIFT - blocksize_bits);
+	read_lock_irqsave(&ni->size_lock, flags);
+	lblock = (ni->allocated_size + blocksize - 1) >> blocksize_bits;
+	init_size = ni->initialized_size;
+	i_size = i_size_read(vi);
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	if (unlikely(init_size > i_size)) {
+		/* Race with shrinking truncate. */
+		init_size = i_size;
+	}
+	zblock = (init_size + blocksize - 1) >> blocksize_bits;
+
+	/* Loop through all the buffers in the page. */
+	rl = NULL;
+	nr = i = 0;
+	do {
+		int err = 0;
+
+		if (unlikely(buffer_uptodate(bh)))
+			continue;
+		if (unlikely(buffer_mapped(bh))) {
+			arr[nr++] = bh;
+			continue;
+		}
+		bh->b_bdev = vol->sb->s_bdev;
+		/* Is the block within the allowed limits? */
+		if (iblock < lblock) {
+			bool is_retry = false;
+
+			/* Convert iblock into corresponding vcn and offset. */
+			vcn = (VCN)iblock << blocksize_bits >>
+					vol->cluster_size_bits;
+			vcn_ofs = ((VCN)iblock << blocksize_bits) &
+					vol->cluster_size_mask;
+			if (!rl) {
+lock_retry_remap:
+				down_read(&ni->runlist.lock);
+				rl = ni->runlist.rl;
+			}
+			if (likely(rl != NULL)) {
+				/* Seek to element containing target vcn. */
+				while (rl->length && rl[1].vcn <= vcn)
+					rl++;
+				lcn = ntfs_rl_vcn_to_lcn(rl, vcn);
+			} else
+				lcn = LCN_RL_NOT_MAPPED;
+			/* Successful remap. */
+			if (lcn >= 0) {
+				/* Setup buffer head to correct block. */
+				bh->b_blocknr = ((lcn << vol->cluster_size_bits)
+						+ vcn_ofs) >> blocksize_bits;
+				set_buffer_mapped(bh);
+				/* Only read initialized data blocks. */
+				if (iblock < zblock) {
+					arr[nr++] = bh;
+					continue;
+				}
+				/* Fully non-initialized data block, zero it. */
+				goto handle_zblock;
+			}
+			/* It is a hole, need to zero it. */
+			if (lcn == LCN_HOLE)
+				goto handle_hole;
+			/* If first try and runlist unmapped, map and retry. */
+			if (!is_retry && lcn == LCN_RL_NOT_MAPPED) {
+				is_retry = true;
+				/*
+				 * Attempt to map runlist, dropping lock for
+				 * the duration.
+				 */
+				up_read(&ni->runlist.lock);
+				err = ntfs_map_runlist(ni, vcn);
+				if (likely(!err))
+					goto lock_retry_remap;
+				rl = NULL;
+			} else if (!rl)
+				up_read(&ni->runlist.lock);
+			/*
+			 * If buffer is outside the runlist, treat it as a
+			 * hole.  This can happen due to concurrent truncate
+			 * for example.
+			 */
+			if (err == -ENOENT || lcn == LCN_ENOENT) {
+				err = 0;
+				goto handle_hole;
+			}
+			/* Hard error, zero out region. */
+			if (!err)
+				err = -EIO;
+			bh->b_blocknr = -1;
+			SetPageError(page);
+			ntfs_error(vol->sb, "Failed to read from inode 0x%lx, "
+					"attribute type 0x%x, vcn 0x%llx, "
+					"offset 0x%x because its location on "
+					"disk could not be determined%s "
+					"(error code %i).", ni->mft_no,
+					ni->type, (unsigned long long)vcn,
+					vcn_ofs, is_retry ? " even after "
+					"retrying" : "", err);
+		}
+		/*
+		 * Either iblock was outside lblock limits or
+		 * ntfs_rl_vcn_to_lcn() returned error.  Just zero that portion
+		 * of the page and set the buffer uptodate.
+		 */
+handle_hole:
+		bh->b_blocknr = -1UL;
+		clear_buffer_mapped(bh);
+handle_zblock:
+		zero_user(page, i * blocksize, blocksize);
+		if (likely(!err))
+			set_buffer_uptodate(bh);
+	} while (i++, iblock++, (bh = bh->b_this_page) != head);
+
+	/* Release the lock if we took it. */
+	if (rl)
+		up_read(&ni->runlist.lock);
+
+	/* Check we have at least one buffer ready for i/o. */
+	if (nr) {
+		struct buffer_head *tbh;
+
+		/* Lock the buffers. */
+		for (i = 0; i < nr; i++) {
+			tbh = arr[i];
+			lock_buffer(tbh);
+			tbh->b_end_io = ntfs_end_buffer_async_read;
+			set_buffer_async_read(tbh);
+		}
+		/* Finally, start i/o on the buffers. */
+		for (i = 0; i < nr; i++) {
+			tbh = arr[i];
+			if (likely(!buffer_uptodate(tbh)))
+				submit_bh(REQ_OP_READ, 0, tbh);
+			else
+				ntfs_end_buffer_async_read(tbh, 1);
+		}
+		return 0;
+	}
+	/* No i/o was scheduled on any of the buffers. */
+	if (likely(!PageError(page)))
+		SetPageUptodate(page);
+	else /* Signal synchronous i/o error. */
+		nr = -EIO;
+	unlock_page(page);
+	return nr;
+}
+
+/**
+ * ntfs_readpage - fill a @page of a @file with data from the device
+ * @file:	open file to which the page @page belongs or NULL
+ * @page:	page cache page to fill with data
+ *
+ * For non-resident attributes, ntfs_readpage() fills the @page of the open
+ * file @file by calling the ntfs version of the generic block_read_full_page()
+ * function, ntfs_read_block(), which in turn creates and reads in the buffers
+ * associated with the page asynchronously.
+ *
+ * For resident attributes, OTOH, ntfs_readpage() fills @page by copying the
+ * data from the mft record (which at this stage is most likely in memory) and
+ * fills the remainder with zeroes. Thus, in this case, I/O is synchronous, as
+ * even if the mft record is not cached at this point in time, we need to wait
+ * for it to be read in before we can do the copy.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static int ntfs_readpage(struct file *file, struct page *page)
+{
+	loff_t i_size;
+	struct inode *vi;
+	ntfs_inode *ni, *base_ni;
+	u8 *addr;
+	ntfs_attr_search_ctx *ctx;
+	MFT_RECORD *mrec;
+	unsigned long flags;
+	u32 attr_len;
+	int err = 0;
+
+retry_readpage:
+	BUG_ON(!PageLocked(page));
+	vi = page->mapping->host;
+	i_size = i_size_read(vi);
+	/* Is the page fully outside i_size? (truncate in progress) */
+	if (unlikely(page->index >= (i_size + PAGE_SIZE - 1) >>
+			PAGE_SHIFT)) {
+		zero_user(page, 0, PAGE_SIZE);
+		ntfs_debug("Read outside i_size - truncated?");
+		goto done;
+	}
+	/*
+	 * This can potentially happen because we clear PageUptodate() during
+	 * ntfs_writepage() of MstProtected() attributes.
+	 */
+	if (PageUptodate(page)) {
+		unlock_page(page);
+		return 0;
+	}
+	ni = NTFS_I(vi);
+	/*
+	 * Only $DATA attributes can be encrypted and only unnamed $DATA
+	 * attributes can be compressed.  Index root can have the flags set but
+	 * this means to create compressed/encrypted files, not that the
+	 * attribute is compressed/encrypted.  Note we need to check for
+	 * AT_INDEX_ALLOCATION since this is the type of both directory and
+	 * index inodes.
+	 */
+	if (ni->type != AT_INDEX_ALLOCATION) {
+		/* If attribute is encrypted, deny access, just like NT4. */
+		if (NInoEncrypted(ni)) {
+			BUG_ON(ni->type != AT_DATA);
+			err = -EACCES;
+			goto err_out;
+		}
+		/* Compressed data streams are handled in compress.c. */
+		if (NInoNonResident(ni) && NInoCompressed(ni)) {
+			BUG_ON(ni->type != AT_DATA);
+			BUG_ON(ni->name_len);
+			return ntfs_read_compressed_block(page);
+		}
+	}
+	/* NInoNonResident() == NInoIndexAllocPresent() */
+	if (NInoNonResident(ni)) {
+		/* Normal, non-resident data stream. */
+		return ntfs_read_block(page);
+	}
+	/*
+	 * Attribute is resident, implying it is not compressed or encrypted.
+	 * This also means the attribute is smaller than an mft record and
+	 * hence smaller than a page, so can simply zero out any pages with
+	 * index above 0.  Note the attribute can actually be marked compressed
+	 * but if it is resident the actual data is not compressed so we are
+	 * ok to ignore the compressed flag here.
+	 */
+	if (unlikely(page->index > 0)) {
+		zero_user(page, 0, PAGE_SIZE);
+		goto done;
+	}
+	if (!NInoAttr(ni))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	/* Map, pin, and lock the mft record. */
+	mrec = map_mft_record(base_ni);
+	if (IS_ERR(mrec)) {
+		err = PTR_ERR(mrec);
+		goto err_out;
+	}
+	/*
+	 * If a parallel write made the attribute non-resident, drop the mft
+	 * record and retry the readpage.
+	 */
+	if (unlikely(NInoNonResident(ni))) {
+		unmap_mft_record(base_ni);
+		goto retry_readpage;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, mrec);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto unm_err_out;
+	}
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err))
+		goto put_unm_err_out;
+	attr_len = le32_to_cpu(ctx->attr->data.resident.value_length);
+	read_lock_irqsave(&ni->size_lock, flags);
+	if (unlikely(attr_len > ni->initialized_size))
+		attr_len = ni->initialized_size;
+	i_size = i_size_read(vi);
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	if (unlikely(attr_len > i_size)) {
+		/* Race with shrinking truncate. */
+		attr_len = i_size;
+	}
+	addr = kmap_atomic(page);
+	/* Copy the data to the page. */
+	memcpy(addr, (u8*)ctx->attr +
+			le16_to_cpu(ctx->attr->data.resident.value_offset),
+			attr_len);
+	/* Zero the remainder of the page. */
+	memset(addr + attr_len, 0, PAGE_SIZE - attr_len);
+	flush_dcache_page(page);
+	kunmap_atomic(addr);
+put_unm_err_out:
+	ntfs_attr_put_search_ctx(ctx);
+unm_err_out:
+	unmap_mft_record(base_ni);
+done:
+	SetPageUptodate(page);
+err_out:
+	unlock_page(page);
+	return err;
+}
+
+#ifdef NTFS_RW
+
+/**
+ * ntfs_write_block - write a @page to the backing store
+ * @page:	page cache page to write out
+ * @wbc:	writeback control structure
+ *
+ * This function is for writing pages belonging to non-resident, non-mst
+ * protected attributes to their backing store.
+ *
+ * For a page with buffers, map and write the dirty buffers asynchronously
+ * under page writeback. For a page without buffers, create buffers for the
+ * page, then proceed as above.
+ *
+ * If a page doesn't have buffers the page dirty state is definitive. If a page
+ * does have buffers, the page dirty state is just a hint, and the buffer dirty
+ * state is definitive. (A hint which has rules: dirty buffers against a clean
+ * page is illegal. Other combinations are legal and need to be handled. In
+ * particular a dirty page containing clean buffers for example.)
+ *
+ * Return 0 on success and -errno on error.
+ *
+ * Based on ntfs_read_block() and __block_write_full_page().
+ */
+static int ntfs_write_block(struct page *page, struct writeback_control *wbc)
+{
+	VCN vcn;
+	LCN lcn;
+	s64 initialized_size;
+	loff_t i_size;
+	sector_t block, dblock, iblock;
+	struct inode *vi;
+	ntfs_inode *ni;
+	ntfs_volume *vol;
+	runlist_element *rl;
+	struct buffer_head *bh, *head;
+	unsigned long flags;
+	unsigned int blocksize, vcn_ofs;
+	int err;
+	bool need_end_writeback;
+	unsigned char blocksize_bits;
+
+	vi = page->mapping->host;
+	ni = NTFS_I(vi);
+	vol = ni->vol;
+
+	ntfs_debug("Entering for inode 0x%lx, attribute type 0x%x, page index "
+			"0x%lx.", ni->mft_no, ni->type, page->index);
+
+	BUG_ON(!NInoNonResident(ni));
+	BUG_ON(NInoMstProtected(ni));
+	blocksize = vol->sb->s_blocksize;
+	blocksize_bits = vol->sb->s_blocksize_bits;
+	if (!page_has_buffers(page)) {
+		BUG_ON(!PageUptodate(page));
+		create_empty_buffers(page, blocksize,
+				(1 << BH_Uptodate) | (1 << BH_Dirty));
+		if (unlikely(!page_has_buffers(page))) {
+			ntfs_warning(vol->sb, "Error allocating page "
+					"buffers.  Redirtying page so we try "
+					"again later.");
+			/*
+			 * Put the page back on mapping->dirty_pages, but leave
+			 * its buffers' dirty state as-is.
+			 */
+			redirty_page_for_writepage(wbc, page);
+			unlock_page(page);
+			return 0;
+		}
+	}
+	bh = head = page_buffers(page);
+	BUG_ON(!bh);
+
+	/* NOTE: Different naming scheme to ntfs_read_block()! */
+
+	/* The first block in the page. */
+	block = (s64)page->index << (PAGE_SHIFT - blocksize_bits);
+
+	read_lock_irqsave(&ni->size_lock, flags);
+	i_size = i_size_read(vi);
+	initialized_size = ni->initialized_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+
+	/* The first out of bounds block for the data size. */
+	dblock = (i_size + blocksize - 1) >> blocksize_bits;
+
+	/* The last (fully or partially) initialized block. */
+	iblock = initialized_size >> blocksize_bits;
+
+	/*
+	 * Be very careful.  We have no exclusion from __set_page_dirty_buffers
+	 * here, and the (potentially unmapped) buffers may become dirty at
+	 * any time.  If a buffer becomes dirty here after we've inspected it
+	 * then we just miss that fact, and the page stays dirty.
+	 *
+	 * Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
+	 * handle that here by just cleaning them.
+	 */
+
+	/*
+	 * Loop through all the buffers in the page, mapping all the dirty
+	 * buffers to disk addresses and handling any aliases from the
+	 * underlying block device's mapping.
+	 */
+	rl = NULL;
+	err = 0;
+	do {
+		bool is_retry = false;
+
+		if (unlikely(block >= dblock)) {
+			/*
+			 * Mapped buffers outside i_size will occur, because
+			 * this page can be outside i_size when there is a
+			 * truncate in progress. The contents of such buffers
+			 * were zeroed by ntfs_writepage().
+			 *
+			 * FIXME: What about the small race window where
+			 * ntfs_writepage() has not done any clearing because
+			 * the page was within i_size but before we get here,
+			 * vmtruncate() modifies i_size?
+			 */
+			clear_buffer_dirty(bh);
+			set_buffer_uptodate(bh);
+			continue;
+		}
+
+		/* Clean buffers are not written out, so no need to map them. */
+		if (!buffer_dirty(bh))
+			continue;
+
+		/* Make sure we have enough initialized size. */
+		if (unlikely((block >= iblock) &&
+				(initialized_size < i_size))) {
+			/*
+			 * If this page is fully outside initialized size, zero
+			 * out all pages between the current initialized size
+			 * and the current page. Just use ntfs_readpage() to do
+			 * the zeroing transparently.
+			 */
+			if (block > iblock) {
+				// TODO:
+				// For each page do:
+				// - read_cache_page()
+				// Again for each page do:
+				// - wait_on_page_locked()
+				// - Check (PageUptodate(page) &&
+				//			!PageError(page))
+				// Update initialized size in the attribute and
+				// in the inode.
+				// Again, for each page do:
+				//	__set_page_dirty_buffers();
+				// put_page()
+				// We don't need to wait on the writes.
+				// Update iblock.
+			}
+			/*
+			 * The current page straddles initialized size. Zero
+			 * all non-uptodate buffers and set them uptodate (and
+			 * dirty?). Note, there aren't any non-uptodate buffers
+			 * if the page is uptodate.
+			 * FIXME: For an uptodate page, the buffers may need to
+			 * be written out because they were not initialized on
+			 * disk before.
+			 */
+			if (!PageUptodate(page)) {
+				// TODO:
+				// Zero any non-uptodate buffers up to i_size.
+				// Set them uptodate and dirty.
+			}
+			// TODO:
+			// Update initialized size in the attribute and in the
+			// inode (up to i_size).
+			// Update iblock.
+			// FIXME: This is inefficient. Try to batch the two
+			// size changes to happen in one go.
+			ntfs_error(vol->sb, "Writing beyond initialized size "
+					"is not supported yet. Sorry.");
+			err = -EOPNOTSUPP;
+			break;
+			// Do NOT set_buffer_new() BUT DO clear buffer range
+			// outside write request range.
+			// set_buffer_uptodate() on complete buffers as well as
+			// set_buffer_dirty().
+		}
+
+		/* No need to map buffers that are already mapped. */
+		if (buffer_mapped(bh))
+			continue;
+
+		/* Unmapped, dirty buffer. Need to map it. */
+		bh->b_bdev = vol->sb->s_bdev;
+
+		/* Convert block into corresponding vcn and offset. */
+		vcn = (VCN)block << blocksize_bits;
+		vcn_ofs = vcn & vol->cluster_size_mask;
+		vcn >>= vol->cluster_size_bits;
+		if (!rl) {
+lock_retry_remap:
+			down_read(&ni->runlist.lock);
+			rl = ni->runlist.rl;
+		}
+		if (likely(rl != NULL)) {
+			/* Seek to element containing target vcn. */
+			while (rl->length && rl[1].vcn <= vcn)
+				rl++;
+			lcn = ntfs_rl_vcn_to_lcn(rl, vcn);
+		} else
+			lcn = LCN_RL_NOT_MAPPED;
+		/* Successful remap. */
+		if (lcn >= 0) {
+			/* Setup buffer head to point to correct block. */
+			bh->b_blocknr = ((lcn << vol->cluster_size_bits) +
+					vcn_ofs) >> blocksize_bits;
+			set_buffer_mapped(bh);
+			continue;
+		}
+		/* It is a hole, need to instantiate it. */
+		if (lcn == LCN_HOLE) {
+			u8 *kaddr;
+			unsigned long *bpos, *bend;
+
+			/* Check if the buffer is zero. */
+			kaddr = kmap_atomic(page);
+			bpos = (unsigned long *)(kaddr + bh_offset(bh));
+			bend = (unsigned long *)((u8*)bpos + blocksize);
+			do {
+				if (unlikely(*bpos))
+					break;
+			} while (likely(++bpos < bend));
+			kunmap_atomic(kaddr);
+			if (bpos == bend) {
+				/*
+				 * Buffer is zero and sparse, no need to write
+				 * it.
+				 */
+				bh->b_blocknr = -1;
+				clear_buffer_dirty(bh);
+				continue;
+			}
+			// TODO: Instantiate the hole.
+			// clear_buffer_new(bh);
+			// clean_bdev_bh_alias(bh);
+			ntfs_error(vol->sb, "Writing into sparse regions is "
+					"not supported yet. Sorry.");
+			err = -EOPNOTSUPP;
+			break;
+		}
+		/* If first try and runlist unmapped, map and retry. */
+		if (!is_retry && lcn == LCN_RL_NOT_MAPPED) {
+			is_retry = true;
+			/*
+			 * Attempt to map runlist, dropping lock for
+			 * the duration.
+			 */
+			up_read(&ni->runlist.lock);
+			err = ntfs_map_runlist(ni, vcn);
+			if (likely(!err))
+				goto lock_retry_remap;
+			rl = NULL;
+		} else if (!rl)
+			up_read(&ni->runlist.lock);
+		/*
+		 * If buffer is outside the runlist, truncate has cut it out
+		 * of the runlist.  Just clean and clear the buffer and set it
+		 * uptodate so it can get discarded by the VM.
+		 */
+		if (err == -ENOENT || lcn == LCN_ENOENT) {
+			bh->b_blocknr = -1;
+			clear_buffer_dirty(bh);
+			zero_user(page, bh_offset(bh), blocksize);
+			set_buffer_uptodate(bh);
+			err = 0;
+			continue;
+		}
+		/* Failed to map the buffer, even after retrying. */
+		if (!err)
+			err = -EIO;
+		bh->b_blocknr = -1;
+		ntfs_error(vol->sb, "Failed to write to inode 0x%lx, "
+				"attribute type 0x%x, vcn 0x%llx, offset 0x%x "
+				"because its location on disk could not be "
+				"determined%s (error code %i).", ni->mft_no,
+				ni->type, (unsigned long long)vcn,
+				vcn_ofs, is_retry ? " even after "
+				"retrying" : "", err);
+		break;
+	} while (block++, (bh = bh->b_this_page) != head);
+
+	/* Release the lock if we took it. */
+	if (rl)
+		up_read(&ni->runlist.lock);
+
+	/* For the error case, need to reset bh to the beginning. */
+	bh = head;
+
+	/* Just an optimization, so ->readpage() is not called later. */
+	if (unlikely(!PageUptodate(page))) {
+		int uptodate = 1;
+		do {
+			if (!buffer_uptodate(bh)) {
+				uptodate = 0;
+				bh = head;
+				break;
+			}
+		} while ((bh = bh->b_this_page) != head);
+		if (uptodate)
+			SetPageUptodate(page);
+	}
+
+	/* Setup all mapped, dirty buffers for async write i/o. */
+	do {
+		if (buffer_mapped(bh) && buffer_dirty(bh)) {
+			lock_buffer(bh);
+			if (test_clear_buffer_dirty(bh)) {
+				BUG_ON(!buffer_uptodate(bh));
+				mark_buffer_async_write(bh);
+			} else
+				unlock_buffer(bh);
+		} else if (unlikely(err)) {
+			/*
+			 * For the error case. The buffer may have been set
+			 * dirty during attachment to a dirty page.
+			 */
+			if (err != -ENOMEM)
+				clear_buffer_dirty(bh);
+		}
+	} while ((bh = bh->b_this_page) != head);
+
+	if (unlikely(err)) {
+		// TODO: Remove the -EOPNOTSUPP check later on...
+		if (unlikely(err == -EOPNOTSUPP))
+			err = 0;
+		else if (err == -ENOMEM) {
+			ntfs_warning(vol->sb, "Error allocating memory. "
+					"Redirtying page so we try again "
+					"later.");
+			/*
+			 * Put the page back on mapping->dirty_pages, but
+			 * leave its buffer's dirty state as-is.
+			 */
+			redirty_page_for_writepage(wbc, page);
+			err = 0;
+		} else
+			SetPageError(page);
+	}
+
+	BUG_ON(PageWriteback(page));
+	set_page_writeback(page);	/* Keeps try_to_free_buffers() away. */
+
+	/* Submit the prepared buffers for i/o. */
+	need_end_writeback = true;
+	do {
+		struct buffer_head *next = bh->b_this_page;
+		if (buffer_async_write(bh)) {
+			submit_bh(REQ_OP_WRITE, 0, bh);
+			need_end_writeback = false;
+		}
+		bh = next;
+	} while (bh != head);
+	unlock_page(page);
+
+	/* If no i/o was started, need to end_page_writeback(). */
+	if (unlikely(need_end_writeback))
+		end_page_writeback(page);
+
+	ntfs_debug("Done.");
+	return err;
+}
+
+/**
+ * ntfs_write_mst_block - write a @page to the backing store
+ * @page:	page cache page to write out
+ * @wbc:	writeback control structure
+ *
+ * This function is for writing pages belonging to non-resident, mst protected
+ * attributes to their backing store.  The only supported attributes are index
+ * allocation and $MFT/$DATA.  Both directory inodes and index inodes are
+ * supported for the index allocation case.
+ *
+ * The page must remain locked for the duration of the write because we apply
+ * the mst fixups, write, and then undo the fixups, so if we were to unlock the
+ * page before undoing the fixups, any other user of the page will see the
+ * page contents as corrupt.
+ *
+ * We clear the page uptodate flag for the duration of the function to ensure
+ * exclusion for the $MFT/$DATA case against someone mapping an mft record we
+ * are about to apply the mst fixups to.
+ *
+ * Return 0 on success and -errno on error.
+ *
+ * Based on ntfs_write_block(), ntfs_mft_writepage(), and
+ * write_mft_record_nolock().
+ */
+static int ntfs_write_mst_block(struct page *page,
+		struct writeback_control *wbc)
+{
+	sector_t block, dblock, rec_block;
+	struct inode *vi = page->mapping->host;
+	ntfs_inode *ni = NTFS_I(vi);
+	ntfs_volume *vol = ni->vol;
+	u8 *kaddr;
+	unsigned int rec_size = ni->itype.index.block_size;
+	ntfs_inode *locked_nis[PAGE_SIZE / NTFS_BLOCK_SIZE];
+	struct buffer_head *bh, *head, *tbh, *rec_start_bh;
+	struct buffer_head *bhs[MAX_BUF_PER_PAGE];
+	runlist_element *rl;
+	int i, nr_locked_nis, nr_recs, nr_bhs, max_bhs, bhs_per_rec, err, err2;
+	unsigned bh_size, rec_size_bits;
+	bool sync, is_mft, page_is_dirty, rec_is_dirty;
+	unsigned char bh_size_bits;
+
+	if (WARN_ON(rec_size < NTFS_BLOCK_SIZE))
+		return -EINVAL;
+
+	ntfs_debug("Entering for inode 0x%lx, attribute type 0x%x, page index "
+			"0x%lx.", vi->i_ino, ni->type, page->index);
+	BUG_ON(!NInoNonResident(ni));
+	BUG_ON(!NInoMstProtected(ni));
+	is_mft = (S_ISREG(vi->i_mode) && !vi->i_ino);
+	/*
+	 * NOTE: ntfs_write_mst_block() would be called for $MFTMirr if a page
+	 * in its page cache were to be marked dirty.  However this should
+	 * never happen with the current driver and considering we do not
+	 * handle this case here we do want to BUG(), at least for now.
+	 */
+	BUG_ON(!(is_mft || S_ISDIR(vi->i_mode) ||
+			(NInoAttr(ni) && ni->type == AT_INDEX_ALLOCATION)));
+	bh_size = vol->sb->s_blocksize;
+	bh_size_bits = vol->sb->s_blocksize_bits;
+	max_bhs = PAGE_SIZE / bh_size;
+	BUG_ON(!max_bhs);
+	BUG_ON(max_bhs > MAX_BUF_PER_PAGE);
+
+	/* Were we called for sync purposes? */
+	sync = (wbc->sync_mode == WB_SYNC_ALL);
+
+	/* Make sure we have mapped buffers. */
+	bh = head = page_buffers(page);
+	BUG_ON(!bh);
+
+	rec_size_bits = ni->itype.index.block_size_bits;
+	BUG_ON(!(PAGE_SIZE >> rec_size_bits));
+	bhs_per_rec = rec_size >> bh_size_bits;
+	BUG_ON(!bhs_per_rec);
+
+	/* The first block in the page. */
+	rec_block = block = (sector_t)page->index <<
+			(PAGE_SHIFT - bh_size_bits);
+
+	/* The first out of bounds block for the data size. */
+	dblock = (i_size_read(vi) + bh_size - 1) >> bh_size_bits;
+
+	rl = NULL;
+	err = err2 = nr_bhs = nr_recs = nr_locked_nis = 0;
+	page_is_dirty = rec_is_dirty = false;
+	rec_start_bh = NULL;
+	do {
+		bool is_retry = false;
+
+		if (likely(block < rec_block)) {
+			if (unlikely(block >= dblock)) {
+				clear_buffer_dirty(bh);
+				set_buffer_uptodate(bh);
+				continue;
+			}
+			/*
+			 * This block is not the first one in the record.  We
+			 * ignore the buffer's dirty state because we could
+			 * have raced with a parallel mark_ntfs_record_dirty().
+			 */
+			if (!rec_is_dirty)
+				continue;
+			if (unlikely(err2)) {
+				if (err2 != -ENOMEM)
+					clear_buffer_dirty(bh);
+				continue;
+			}
+		} else /* if (block == rec_block) */ {
+			BUG_ON(block > rec_block);
+			/* This block is the first one in the record. */
+			rec_block += bhs_per_rec;
+			err2 = 0;
+			if (unlikely(block >= dblock)) {
+				clear_buffer_dirty(bh);
+				continue;
+			}
+			if (!buffer_dirty(bh)) {
+				/* Clean records are not written out. */
+				rec_is_dirty = false;
+				continue;
+			}
+			rec_is_dirty = true;
+			rec_start_bh = bh;
+		}
+		/* Need to map the buffer if it is not mapped already. */
+		if (unlikely(!buffer_mapped(bh))) {
+			VCN vcn;
+			LCN lcn;
+			unsigned int vcn_ofs;
+
+			bh->b_bdev = vol->sb->s_bdev;
+			/* Obtain the vcn and offset of the current block. */
+			vcn = (VCN)block << bh_size_bits;
+			vcn_ofs = vcn & vol->cluster_size_mask;
+			vcn >>= vol->cluster_size_bits;
+			if (!rl) {
+lock_retry_remap:
+				down_read(&ni->runlist.lock);
+				rl = ni->runlist.rl;
+			}
+			if (likely(rl != NULL)) {
+				/* Seek to element containing target vcn. */
+				while (rl->length && rl[1].vcn <= vcn)
+					rl++;
+				lcn = ntfs_rl_vcn_to_lcn(rl, vcn);
+			} else
+				lcn = LCN_RL_NOT_MAPPED;
+			/* Successful remap. */
+			if (likely(lcn >= 0)) {
+				/* Setup buffer head to correct block. */
+				bh->b_blocknr = ((lcn <<
+						vol->cluster_size_bits) +
+						vcn_ofs) >> bh_size_bits;
+				set_buffer_mapped(bh);
+			} else {
+				/*
+				 * Remap failed.  Retry to map the runlist once
+				 * unless we are working on $MFT which always
+				 * has the whole of its runlist in memory.
+				 */
+				if (!is_mft && !is_retry &&
+						lcn == LCN_RL_NOT_MAPPED) {
+					is_retry = true;
+					/*
+					 * Attempt to map runlist, dropping
+					 * lock for the duration.
+					 */
+					up_read(&ni->runlist.lock);
+					err2 = ntfs_map_runlist(ni, vcn);
+					if (likely(!err2))
+						goto lock_retry_remap;
+					if (err2 == -ENOMEM)
+						page_is_dirty = true;
+					lcn = err2;
+				} else {
+					err2 = -EIO;
+					if (!rl)
+						up_read(&ni->runlist.lock);
+				}
+				/* Hard error.  Abort writing this record. */
+				if (!err || err == -ENOMEM)
+					err = err2;
+				bh->b_blocknr = -1;
+				ntfs_error(vol->sb, "Cannot write ntfs record "
+						"0x%llx (inode 0x%lx, "
+						"attribute type 0x%x) because "
+						"its location on disk could "
+						"not be determined (error "
+						"code %lli).",
+						(long long)block <<
+						bh_size_bits >>
+						vol->mft_record_size_bits,
+						ni->mft_no, ni->type,
+						(long long)lcn);
+				/*
+				 * If this is not the first buffer, remove the
+				 * buffers in this record from the list of
+				 * buffers to write and clear their dirty bit
+				 * if not error -ENOMEM.
+				 */
+				if (rec_start_bh != bh) {
+					while (bhs[--nr_bhs] != rec_start_bh)
+						;
+					if (err2 != -ENOMEM) {
+						do {
+							clear_buffer_dirty(
+								rec_start_bh);
+						} while ((rec_start_bh =
+								rec_start_bh->
+								b_this_page) !=
+								bh);
+					}
+				}
+				continue;
+			}
+		}
+		BUG_ON(!buffer_uptodate(bh));
+		BUG_ON(nr_bhs >= max_bhs);
+		bhs[nr_bhs++] = bh;
+	} while (block++, (bh = bh->b_this_page) != head);
+	if (unlikely(rl))
+		up_read(&ni->runlist.lock);
+	/* If there were no dirty buffers, we are done. */
+	if (!nr_bhs)
+		goto done;
+	/* Map the page so we can access its contents. */
+	kaddr = kmap(page);
+	/* Clear the page uptodate flag whilst the mst fixups are applied. */
+	BUG_ON(!PageUptodate(page));
+	ClearPageUptodate(page);
+	for (i = 0; i < nr_bhs; i++) {
+		unsigned int ofs;
+
+		/* Skip buffers which are not at the beginning of records. */
+		if (i % bhs_per_rec)
+			continue;
+		tbh = bhs[i];
+		ofs = bh_offset(tbh);
+		if (is_mft) {
+			ntfs_inode *tni;
+			unsigned long mft_no;
+
+			/* Get the mft record number. */
+			mft_no = (((s64)page->index << PAGE_SHIFT) + ofs)
+					>> rec_size_bits;
+			/* Check whether to write this mft record. */
+			tni = NULL;
+			if (!ntfs_may_write_mft_record(vol, mft_no,
+					(MFT_RECORD*)(kaddr + ofs), &tni)) {
+				/*
+				 * The record should not be written.  This
+				 * means we need to redirty the page before
+				 * returning.
+				 */
+				page_is_dirty = true;
+				/*
+				 * Remove the buffers in this mft record from
+				 * the list of buffers to write.
+				 */
+				do {
+					bhs[i] = NULL;
+				} while (++i % bhs_per_rec);
+				continue;
+			}
+			/*
+			 * The record should be written.  If a locked ntfs
+			 * inode was returned, add it to the array of locked
+			 * ntfs inodes.
+			 */
+			if (tni)
+				locked_nis[nr_locked_nis++] = tni;
+		}
+		/* Apply the mst protection fixups. */
+		err2 = pre_write_mst_fixup((NTFS_RECORD*)(kaddr + ofs),
+				rec_size);
+		if (unlikely(err2)) {
+			if (!err || err == -ENOMEM)
+				err = -EIO;
+			ntfs_error(vol->sb, "Failed to apply mst fixups "
+					"(inode 0x%lx, attribute type 0x%x, "
+					"page index 0x%lx, page offset 0x%x)!"
+					"  Unmount and run chkdsk.", vi->i_ino,
+					ni->type, page->index, ofs);
+			/*
+			 * Mark all the buffers in this record clean as we do
+			 * not want to write corrupt data to disk.
+			 */
+			do {
+				clear_buffer_dirty(bhs[i]);
+				bhs[i] = NULL;
+			} while (++i % bhs_per_rec);
+			continue;
+		}
+		nr_recs++;
+	}
+	/* If no records are to be written out, we are done. */
+	if (!nr_recs)
+		goto unm_done;
+	flush_dcache_page(page);
+	/* Lock buffers and start synchronous write i/o on them. */
+	for (i = 0; i < nr_bhs; i++) {
+		tbh = bhs[i];
+		if (!tbh)
+			continue;
+		if (!trylock_buffer(tbh))
+			BUG();
+		/* The buffer dirty state is now irrelevant, just clean it. */
+		clear_buffer_dirty(tbh);
+		BUG_ON(!buffer_uptodate(tbh));
+		BUG_ON(!buffer_mapped(tbh));
+		get_bh(tbh);
+		tbh->b_end_io = end_buffer_write_sync;
+		submit_bh(REQ_OP_WRITE, 0, tbh);
+	}
+	/* Synchronize the mft mirror now if not @sync. */
+	if (is_mft && !sync)
+		goto do_mirror;
+do_wait:
+	/* Wait on i/o completion of buffers. */
+	for (i = 0; i < nr_bhs; i++) {
+		tbh = bhs[i];
+		if (!tbh)
+			continue;
+		wait_on_buffer(tbh);
+		if (unlikely(!buffer_uptodate(tbh))) {
+			ntfs_error(vol->sb, "I/O error while writing ntfs "
+					"record buffer (inode 0x%lx, "
+					"attribute type 0x%x, page index "
+					"0x%lx, page offset 0x%lx)!  Unmount "
+					"and run chkdsk.", vi->i_ino, ni->type,
+					page->index, bh_offset(tbh));
+			if (!err || err == -ENOMEM)
+				err = -EIO;
+			/*
+			 * Set the buffer uptodate so the page and buffer
+			 * states do not become out of sync.
+			 */
+			set_buffer_uptodate(tbh);
+		}
+	}
+	/* If @sync, now synchronize the mft mirror. */
+	if (is_mft && sync) {
+do_mirror:
+		for (i = 0; i < nr_bhs; i++) {
+			unsigned long mft_no;
+			unsigned int ofs;
+
+			/*
+			 * Skip buffers which are not at the beginning of
+			 * records.
+			 */
+			if (i % bhs_per_rec)
+				continue;
+			tbh = bhs[i];
+			/* Skip removed buffers (and hence records). */
+			if (!tbh)
+				continue;
+			ofs = bh_offset(tbh);
+			/* Get the mft record number. */
+			mft_no = (((s64)page->index << PAGE_SHIFT) + ofs)
+					>> rec_size_bits;
+			if (mft_no < vol->mftmirr_size)
+				ntfs_sync_mft_mirror(vol, mft_no,
+						(MFT_RECORD*)(kaddr + ofs),
+						sync);
+		}
+		if (!sync)
+			goto do_wait;
+	}
+	/* Remove the mst protection fixups again. */
+	for (i = 0; i < nr_bhs; i++) {
+		if (!(i % bhs_per_rec)) {
+			tbh = bhs[i];
+			if (!tbh)
+				continue;
+			post_write_mst_fixup((NTFS_RECORD*)(kaddr +
+					bh_offset(tbh)));
+		}
+	}
+	flush_dcache_page(page);
+unm_done:
+	/* Unlock any locked inodes. */
+	while (nr_locked_nis-- > 0) {
+		ntfs_inode *tni, *base_tni;
+		
+		tni = locked_nis[nr_locked_nis];
+		/* Get the base inode. */
+		mutex_lock(&tni->extent_lock);
+		if (tni->nr_extents >= 0)
+			base_tni = tni;
+		else {
+			base_tni = tni->ext.base_ntfs_ino;
+			BUG_ON(!base_tni);
+		}
+		mutex_unlock(&tni->extent_lock);
+		ntfs_debug("Unlocking %s inode 0x%lx.",
+				tni == base_tni ? "base" : "extent",
+				tni->mft_no);
+		mutex_unlock(&tni->mrec_lock);
+		atomic_dec(&tni->count);
+		iput(VFS_I(base_tni));
+	}
+	SetPageUptodate(page);
+	kunmap(page);
+done:
+	if (unlikely(err && err != -ENOMEM)) {
+		/*
+		 * Set page error if there is only one ntfs record in the page.
+		 * Otherwise we would loose per-record granularity.
+		 */
+		if (ni->itype.index.block_size == PAGE_SIZE)
+			SetPageError(page);
+		NVolSetErrors(vol);
+	}
+	if (page_is_dirty) {
+		ntfs_debug("Page still contains one or more dirty ntfs "
+				"records.  Redirtying the page starting at "
+				"record 0x%lx.", page->index <<
+				(PAGE_SHIFT - rec_size_bits));
+		redirty_page_for_writepage(wbc, page);
+		unlock_page(page);
+	} else {
+		/*
+		 * Keep the VM happy.  This must be done otherwise the
+		 * radix-tree tag PAGECACHE_TAG_DIRTY remains set even though
+		 * the page is clean.
+		 */
+		BUG_ON(PageWriteback(page));
+		set_page_writeback(page);
+		unlock_page(page);
+		end_page_writeback(page);
+	}
+	if (likely(!err))
+		ntfs_debug("Done.");
+	return err;
+}
+
+/**
+ * ntfs_writepage - write a @page to the backing store
+ * @page:	page cache page to write out
+ * @wbc:	writeback control structure
+ *
+ * This is called from the VM when it wants to have a dirty ntfs page cache
+ * page cleaned.  The VM has already locked the page and marked it clean.
+ *
+ * For non-resident attributes, ntfs_writepage() writes the @page by calling
+ * the ntfs version of the generic block_write_full_page() function,
+ * ntfs_write_block(), which in turn if necessary creates and writes the
+ * buffers associated with the page asynchronously.
+ *
+ * For resident attributes, OTOH, ntfs_writepage() writes the @page by copying
+ * the data to the mft record (which at this stage is most likely in memory).
+ * The mft record is then marked dirty and written out asynchronously via the
+ * vfs inode dirty code path for the inode the mft record belongs to or via the
+ * vm page dirty code path for the page the mft record is in.
+ *
+ * Based on ntfs_readpage() and fs/buffer.c::block_write_full_page().
+ *
+ * Return 0 on success and -errno on error.
+ */
+static int ntfs_writepage(struct page *page, struct writeback_control *wbc)
+{
+	loff_t i_size;
+	struct inode *vi = page->mapping->host;
+	ntfs_inode *base_ni = NULL, *ni = NTFS_I(vi);
+	char *addr;
+	ntfs_attr_search_ctx *ctx = NULL;
+	MFT_RECORD *m = NULL;
+	u32 attr_len;
+	int err;
+
+retry_writepage:
+	BUG_ON(!PageLocked(page));
+	i_size = i_size_read(vi);
+	/* Is the page fully outside i_size? (truncate in progress) */
+	if (unlikely(page->index >= (i_size + PAGE_SIZE - 1) >>
+			PAGE_SHIFT)) {
+		/*
+		 * The page may have dirty, unmapped buffers.  Make them
+		 * freeable here, so the page does not leak.
+		 */
+		block_invalidatepage(page, 0, PAGE_SIZE);
+		unlock_page(page);
+		ntfs_debug("Write outside i_size - truncated?");
+		return 0;
+	}
+	/*
+	 * Only $DATA attributes can be encrypted and only unnamed $DATA
+	 * attributes can be compressed.  Index root can have the flags set but
+	 * this means to create compressed/encrypted files, not that the
+	 * attribute is compressed/encrypted.  Note we need to check for
+	 * AT_INDEX_ALLOCATION since this is the type of both directory and
+	 * index inodes.
+	 */
+	if (ni->type != AT_INDEX_ALLOCATION) {
+		/* If file is encrypted, deny access, just like NT4. */
+		if (NInoEncrypted(ni)) {
+			unlock_page(page);
+			BUG_ON(ni->type != AT_DATA);
+			ntfs_debug("Denying write access to encrypted file.");
+			return -EACCES;
+		}
+		/* Compressed data streams are handled in compress.c. */
+		if (NInoNonResident(ni) && NInoCompressed(ni)) {
+			BUG_ON(ni->type != AT_DATA);
+			BUG_ON(ni->name_len);
+			// TODO: Implement and replace this with
+			// return ntfs_write_compressed_block(page);
+			unlock_page(page);
+			ntfs_error(vi->i_sb, "Writing to compressed files is "
+					"not supported yet.  Sorry.");
+			return -EOPNOTSUPP;
+		}
+		// TODO: Implement and remove this check.
+		if (NInoNonResident(ni) && NInoSparse(ni)) {
+			unlock_page(page);
+			ntfs_error(vi->i_sb, "Writing to sparse files is not "
+					"supported yet.  Sorry.");
+			return -EOPNOTSUPP;
+		}
+	}
+	/* NInoNonResident() == NInoIndexAllocPresent() */
+	if (NInoNonResident(ni)) {
+		/* We have to zero every time due to mmap-at-end-of-file. */
+		if (page->index >= (i_size >> PAGE_SHIFT)) {
+			/* The page straddles i_size. */
+			unsigned int ofs = i_size & ~PAGE_MASK;
+			zero_user_segment(page, ofs, PAGE_SIZE);
+		}
+		/* Handle mst protected attributes. */
+		if (NInoMstProtected(ni))
+			return ntfs_write_mst_block(page, wbc);
+		/* Normal, non-resident data stream. */
+		return ntfs_write_block(page, wbc);
+	}
+	/*
+	 * Attribute is resident, implying it is not compressed, encrypted, or
+	 * mst protected.  This also means the attribute is smaller than an mft
+	 * record and hence smaller than a page, so can simply return error on
+	 * any pages with index above 0.  Note the attribute can actually be
+	 * marked compressed but if it is resident the actual data is not
+	 * compressed so we are ok to ignore the compressed flag here.
+	 */
+	BUG_ON(page_has_buffers(page));
+	BUG_ON(!PageUptodate(page));
+	if (unlikely(page->index > 0)) {
+		ntfs_error(vi->i_sb, "BUG()! page->index (0x%lx) > 0.  "
+				"Aborting write.", page->index);
+		BUG_ON(PageWriteback(page));
+		set_page_writeback(page);
+		unlock_page(page);
+		end_page_writeback(page);
+		return -EIO;
+	}
+	if (!NInoAttr(ni))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	/* Map, pin, and lock the mft record. */
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		m = NULL;
+		ctx = NULL;
+		goto err_out;
+	}
+	/*
+	 * If a parallel write made the attribute non-resident, drop the mft
+	 * record and retry the writepage.
+	 */
+	if (unlikely(NInoNonResident(ni))) {
+		unmap_mft_record(base_ni);
+		goto retry_writepage;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err))
+		goto err_out;
+	/*
+	 * Keep the VM happy.  This must be done otherwise the radix-tree tag
+	 * PAGECACHE_TAG_DIRTY remains set even though the page is clean.
+	 */
+	BUG_ON(PageWriteback(page));
+	set_page_writeback(page);
+	unlock_page(page);
+	attr_len = le32_to_cpu(ctx->attr->data.resident.value_length);
+	i_size = i_size_read(vi);
+	if (unlikely(attr_len > i_size)) {
+		/* Race with shrinking truncate or a failed truncate. */
+		attr_len = i_size;
+		/*
+		 * If the truncate failed, fix it up now.  If a concurrent
+		 * truncate, we do its job, so it does not have to do anything.
+		 */
+		err = ntfs_resident_attr_value_resize(ctx->mrec, ctx->attr,
+				attr_len);
+		/* Shrinking cannot fail. */
+		BUG_ON(err);
+	}
+	addr = kmap_atomic(page);
+	/* Copy the data from the page to the mft record. */
+	memcpy((u8*)ctx->attr +
+			le16_to_cpu(ctx->attr->data.resident.value_offset),
+			addr, attr_len);
+	/* Zero out of bounds area in the page cache page. */
+	memset(addr + attr_len, 0, PAGE_SIZE - attr_len);
+	kunmap_atomic(addr);
+	flush_dcache_page(page);
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	/* We are done with the page. */
+	end_page_writeback(page);
+	/* Finally, mark the mft record dirty, so it gets written back. */
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+	return 0;
+err_out:
+	if (err == -ENOMEM) {
+		ntfs_warning(vi->i_sb, "Error allocating memory. Redirtying "
+				"page so we try again later.");
+		/*
+		 * Put the page back on mapping->dirty_pages, but leave its
+		 * buffers' dirty state as-is.
+		 */
+		redirty_page_for_writepage(wbc, page);
+		err = 0;
+	} else {
+		ntfs_error(vi->i_sb, "Resident attribute write failed with "
+				"error %i.", err);
+		SetPageError(page);
+		NVolSetErrors(ni->vol);
+	}
+	unlock_page(page);
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(base_ni);
+	return err;
+}
+
+#endif	/* NTFS_RW */
+
+/**
+ * ntfs_bmap - map logical file block to physical device block
+ * @mapping:	address space mapping to which the block to be mapped belongs
+ * @block:	logical block to map to its physical device block
+ *
+ * For regular, non-resident files (i.e. not compressed and not encrypted), map
+ * the logical @block belonging to the file described by the address space
+ * mapping @mapping to its physical device block.
+ *
+ * The size of the block is equal to the @s_blocksize field of the super block
+ * of the mounted file system which is guaranteed to be smaller than or equal
+ * to the cluster size thus the block is guaranteed to fit entirely inside the
+ * cluster which means we do not need to care how many contiguous bytes are
+ * available after the beginning of the block.
+ *
+ * Return the physical device block if the mapping succeeded or 0 if the block
+ * is sparse or there was an error.
+ *
+ * Note: This is a problem if someone tries to run bmap() on $Boot system file
+ * as that really is in block zero but there is nothing we can do.  bmap() is
+ * just broken in that respect (just like it cannot distinguish sparse from
+ * not available or error).
+ */
+static sector_t ntfs_bmap(struct address_space *mapping, sector_t block)
+{
+	s64 ofs, size;
+	loff_t i_size;
+	LCN lcn;
+	unsigned long blocksize, flags;
+	ntfs_inode *ni = NTFS_I(mapping->host);
+	ntfs_volume *vol = ni->vol;
+	unsigned delta;
+	unsigned char blocksize_bits, cluster_size_shift;
+
+	ntfs_debug("Entering for mft_no 0x%lx, logical block 0x%llx.",
+			ni->mft_no, (unsigned long long)block);
+	if (ni->type != AT_DATA || !NInoNonResident(ni) || NInoEncrypted(ni)) {
+		ntfs_error(vol->sb, "BMAP does not make sense for %s "
+				"attributes, returning 0.",
+				(ni->type != AT_DATA) ? "non-data" :
+				(!NInoNonResident(ni) ? "resident" :
+				"encrypted"));
+		return 0;
+	}
+	/* None of these can happen. */
+	BUG_ON(NInoCompressed(ni));
+	BUG_ON(NInoMstProtected(ni));
+	blocksize = vol->sb->s_blocksize;
+	blocksize_bits = vol->sb->s_blocksize_bits;
+	ofs = (s64)block << blocksize_bits;
+	read_lock_irqsave(&ni->size_lock, flags);
+	size = ni->initialized_size;
+	i_size = i_size_read(VFS_I(ni));
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	/*
+	 * If the offset is outside the initialized size or the block straddles
+	 * the initialized size then pretend it is a hole unless the
+	 * initialized size equals the file size.
+	 */
+	if (unlikely(ofs >= size || (ofs + blocksize > size && size < i_size)))
+		goto hole;
+	cluster_size_shift = vol->cluster_size_bits;
+	down_read(&ni->runlist.lock);
+	lcn = ntfs_attr_vcn_to_lcn_nolock(ni, ofs >> cluster_size_shift, false);
+	up_read(&ni->runlist.lock);
+	if (unlikely(lcn < LCN_HOLE)) {
+		/*
+		 * Step down to an integer to avoid gcc doing a long long
+		 * comparision in the switch when we know @lcn is between
+		 * LCN_HOLE and LCN_EIO (i.e. -1 to -5).
+		 *
+		 * Otherwise older gcc (at least on some architectures) will
+		 * try to use __cmpdi2() which is of course not available in
+		 * the kernel.
+		 */
+		switch ((int)lcn) {
+		case LCN_ENOENT:
+			/*
+			 * If the offset is out of bounds then pretend it is a
+			 * hole.
+			 */
+			goto hole;
+		case LCN_ENOMEM:
+			ntfs_error(vol->sb, "Not enough memory to complete "
+					"mapping for inode 0x%lx.  "
+					"Returning 0.", ni->mft_no);
+			break;
+		default:
+			ntfs_error(vol->sb, "Failed to complete mapping for "
+					"inode 0x%lx.  Run chkdsk.  "
+					"Returning 0.", ni->mft_no);
+			break;
+		}
+		return 0;
+	}
+	if (lcn < 0) {
+		/* It is a hole. */
+hole:
+		ntfs_debug("Done (returning hole).");
+		return 0;
+	}
+	/*
+	 * The block is really allocated and fullfils all our criteria.
+	 * Convert the cluster to units of block size and return the result.
+	 */
+	delta = ofs & vol->cluster_size_mask;
+	if (unlikely(sizeof(block) < sizeof(lcn))) {
+		block = lcn = ((lcn << cluster_size_shift) + delta) >>
+				blocksize_bits;
+		/* If the block number was truncated return 0. */
+		if (unlikely(block != lcn)) {
+			ntfs_error(vol->sb, "Physical block 0x%llx is too "
+					"large to be returned, returning 0.",
+					(long long)lcn);
+			return 0;
+		}
+	} else
+		block = ((lcn << cluster_size_shift) + delta) >>
+				blocksize_bits;
+	ntfs_debug("Done (returning block 0x%llx).", (unsigned long long)lcn);
+	return block;
+}
+
+/**
+ * ntfs_normal_aops - address space operations for normal inodes and attributes
+ *
+ * Note these are not used for compressed or mst protected inodes and
+ * attributes.
+ */
+const struct address_space_operations ntfs_normal_aops = {
+	.readpage	= ntfs_readpage,
+#ifdef NTFS_RW
+	.writepage	= ntfs_writepage,
+	.set_page_dirty	= __set_page_dirty_buffers,
+#endif /* NTFS_RW */
+	.bmap		= ntfs_bmap,
+	.migratepage	= buffer_migrate_page,
+	.is_partially_uptodate = block_is_partially_uptodate,
+	.error_remove_page = generic_error_remove_page,
+};
+
+/**
+ * ntfs_compressed_aops - address space operations for compressed inodes
+ */
+const struct address_space_operations ntfs_compressed_aops = {
+	.readpage	= ntfs_readpage,
+#ifdef NTFS_RW
+	.writepage	= ntfs_writepage,
+	.set_page_dirty	= __set_page_dirty_buffers,
+#endif /* NTFS_RW */
+	.migratepage	= buffer_migrate_page,
+	.is_partially_uptodate = block_is_partially_uptodate,
+	.error_remove_page = generic_error_remove_page,
+};
+
+/**
+ * ntfs_mst_aops - general address space operations for mst protecteed inodes
+ *		   and attributes
+ */
+const struct address_space_operations ntfs_mst_aops = {
+	.readpage	= ntfs_readpage,	/* Fill page with data. */
+#ifdef NTFS_RW
+	.writepage	= ntfs_writepage,	/* Write dirty page to disk. */
+	.set_page_dirty	= __set_page_dirty_nobuffers,	/* Set the page dirty
+						   without touching the buffers
+						   belonging to the page. */
+#endif /* NTFS_RW */
+	.migratepage	= buffer_migrate_page,
+	.is_partially_uptodate	= block_is_partially_uptodate,
+	.error_remove_page = generic_error_remove_page,
+};
+
+#ifdef NTFS_RW
+
+/**
+ * mark_ntfs_record_dirty - mark an ntfs record dirty
+ * @page:	page containing the ntfs record to mark dirty
+ * @ofs:	byte offset within @page at which the ntfs record begins
+ *
+ * Set the buffers and the page in which the ntfs record is located dirty.
+ *
+ * The latter also marks the vfs inode the ntfs record belongs to dirty
+ * (I_DIRTY_PAGES only).
+ *
+ * If the page does not have buffers, we create them and set them uptodate.
+ * The page may not be locked which is why we need to handle the buffers under
+ * the mapping->private_lock.  Once the buffers are marked dirty we no longer
+ * need the lock since try_to_free_buffers() does not free dirty buffers.
+ */
+void mark_ntfs_record_dirty(struct page *page, const unsigned int ofs) {
+	struct address_space *mapping = page->mapping;
+	ntfs_inode *ni = NTFS_I(mapping->host);
+	struct buffer_head *bh, *head, *buffers_to_free = NULL;
+	unsigned int end, bh_size, bh_ofs;
+
+	BUG_ON(!PageUptodate(page));
+	end = ofs + ni->itype.index.block_size;
+	bh_size = VFS_I(ni)->i_sb->s_blocksize;
+	spin_lock(&mapping->private_lock);
+	if (unlikely(!page_has_buffers(page))) {
+		spin_unlock(&mapping->private_lock);
+		bh = head = alloc_page_buffers(page, bh_size, true);
+		spin_lock(&mapping->private_lock);
+		if (likely(!page_has_buffers(page))) {
+			struct buffer_head *tail;
+
+			do {
+				set_buffer_uptodate(bh);
+				tail = bh;
+				bh = bh->b_this_page;
+			} while (bh);
+			tail->b_this_page = head;
+			attach_page_private(page, head);
+		} else
+			buffers_to_free = bh;
+	}
+	bh = head = page_buffers(page);
+	BUG_ON(!bh);
+	do {
+		bh_ofs = bh_offset(bh);
+		if (bh_ofs + bh_size <= ofs)
+			continue;
+		if (unlikely(bh_ofs >= end))
+			break;
+		set_buffer_dirty(bh);
+	} while ((bh = bh->b_this_page) != head);
+	spin_unlock(&mapping->private_lock);
+	__set_page_dirty_nobuffers(page);
+	if (unlikely(buffers_to_free)) {
+		do {
+			bh = buffers_to_free->b_this_page;
+			free_buffer_head(buffers_to_free);
+			buffers_to_free = bh;
+		} while (buffers_to_free);
+	}
+}
+
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/aops.h b/fs/ntfs/aops.h
new file mode 100644
index 000000000..f0962d46b
--- /dev/null
+++ b/fs/ntfs/aops.h
@@ -0,0 +1,93 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/**
+ * aops.h - Defines for NTFS kernel address space operations and page cache
+ *	    handling.  Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2004 Anton Altaparmakov
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#ifndef _LINUX_NTFS_AOPS_H
+#define _LINUX_NTFS_AOPS_H
+
+#include <linux/mm.h>
+#include <linux/highmem.h>
+#include <linux/pagemap.h>
+#include <linux/fs.h>
+
+#include "inode.h"
+
+/**
+ * ntfs_unmap_page - release a page that was mapped using ntfs_map_page()
+ * @page:	the page to release
+ *
+ * Unpin, unmap and release a page that was obtained from ntfs_map_page().
+ */
+static inline void ntfs_unmap_page(struct page *page)
+{
+	kunmap(page);
+	put_page(page);
+}
+
+/**
+ * ntfs_map_page - map a page into accessible memory, reading it if necessary
+ * @mapping:	address space for which to obtain the page
+ * @index:	index into the page cache for @mapping of the page to map
+ *
+ * Read a page from the page cache of the address space @mapping at position
+ * @index, where @index is in units of PAGE_SIZE, and not in bytes.
+ *
+ * If the page is not in memory it is loaded from disk first using the readpage
+ * method defined in the address space operations of @mapping and the page is
+ * added to the page cache of @mapping in the process.
+ *
+ * If the page belongs to an mst protected attribute and it is marked as such
+ * in its ntfs inode (NInoMstProtected()) the mst fixups are applied but no
+ * error checking is performed.  This means the caller has to verify whether
+ * the ntfs record(s) contained in the page are valid or not using one of the
+ * ntfs_is_XXXX_record{,p}() macros, where XXXX is the record type you are
+ * expecting to see.  (For details of the macros, see fs/ntfs/layout.h.)
+ *
+ * If the page is in high memory it is mapped into memory directly addressible
+ * by the kernel.
+ *
+ * Finally the page count is incremented, thus pinning the page into place.
+ *
+ * The above means that page_address(page) can be used on all pages obtained
+ * with ntfs_map_page() to get the kernel virtual address of the page.
+ *
+ * When finished with the page, the caller has to call ntfs_unmap_page() to
+ * unpin, unmap and release the page.
+ *
+ * Note this does not grant exclusive access. If such is desired, the caller
+ * must provide it independently of the ntfs_{un}map_page() calls by using
+ * a {rw_}semaphore or other means of serialization. A spin lock cannot be
+ * used as ntfs_map_page() can block.
+ *
+ * The unlocked and uptodate page is returned on success or an encoded error
+ * on failure. Caller has to test for error using the IS_ERR() macro on the
+ * return value. If that evaluates to 'true', the negative error code can be
+ * obtained using PTR_ERR() on the return value of ntfs_map_page().
+ */
+static inline struct page *ntfs_map_page(struct address_space *mapping,
+		unsigned long index)
+{
+	struct page *page = read_mapping_page(mapping, index, NULL);
+
+	if (!IS_ERR(page)) {
+		kmap(page);
+		if (!PageError(page))
+			return page;
+		ntfs_unmap_page(page);
+		return ERR_PTR(-EIO);
+	}
+	return page;
+}
+
+#ifdef NTFS_RW
+
+extern void mark_ntfs_record_dirty(struct page *page, const unsigned int ofs);
+
+#endif /* NTFS_RW */
+
+#endif /* _LINUX_NTFS_AOPS_H */
diff --git a/fs/ntfs/attrib.c b/fs/ntfs/attrib.c
new file mode 100644
index 000000000..c0881d39d
--- /dev/null
+++ b/fs/ntfs/attrib.c
@@ -0,0 +1,2624 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/**
+ * attrib.c - NTFS attribute operations.  Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2012 Anton Altaparmakov and Tuxera Inc.
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#include <linux/buffer_head.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/swap.h>
+#include <linux/writeback.h>
+
+#include "attrib.h"
+#include "debug.h"
+#include "layout.h"
+#include "lcnalloc.h"
+#include "malloc.h"
+#include "mft.h"
+#include "ntfs.h"
+#include "types.h"
+
+/**
+ * ntfs_map_runlist_nolock - map (a part of) a runlist of an ntfs inode
+ * @ni:		ntfs inode for which to map (part of) a runlist
+ * @vcn:	map runlist part containing this vcn
+ * @ctx:	active attribute search context if present or NULL if not
+ *
+ * Map the part of a runlist containing the @vcn of the ntfs inode @ni.
+ *
+ * If @ctx is specified, it is an active search context of @ni and its base mft
+ * record.  This is needed when ntfs_map_runlist_nolock() encounters unmapped
+ * runlist fragments and allows their mapping.  If you do not have the mft
+ * record mapped, you can specify @ctx as NULL and ntfs_map_runlist_nolock()
+ * will perform the necessary mapping and unmapping.
+ *
+ * Note, ntfs_map_runlist_nolock() saves the state of @ctx on entry and
+ * restores it before returning.  Thus, @ctx will be left pointing to the same
+ * attribute on return as on entry.  However, the actual pointers in @ctx may
+ * point to different memory locations on return, so you must remember to reset
+ * any cached pointers from the @ctx, i.e. after the call to
+ * ntfs_map_runlist_nolock(), you will probably want to do:
+ *	m = ctx->mrec;
+ *	a = ctx->attr;
+ * Assuming you cache ctx->attr in a variable @a of type ATTR_RECORD * and that
+ * you cache ctx->mrec in a variable @m of type MFT_RECORD *.
+ *
+ * Return 0 on success and -errno on error.  There is one special error code
+ * which is not an error as such.  This is -ENOENT.  It means that @vcn is out
+ * of bounds of the runlist.
+ *
+ * Note the runlist can be NULL after this function returns if @vcn is zero and
+ * the attribute has zero allocated size, i.e. there simply is no runlist.
+ *
+ * WARNING: If @ctx is supplied, regardless of whether success or failure is
+ *	    returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
+ *	    is no longer valid, i.e. you need to either call
+ *	    ntfs_attr_reinit_search_ctx() or ntfs_attr_put_search_ctx() on it.
+ *	    In that case PTR_ERR(@ctx->mrec) will give you the error code for
+ *	    why the mapping of the old inode failed.
+ *
+ * Locking: - The runlist described by @ni must be locked for writing on entry
+ *	      and is locked on return.  Note the runlist will be modified.
+ *	    - If @ctx is NULL, the base mft record of @ni must not be mapped on
+ *	      entry and it will be left unmapped on return.
+ *	    - If @ctx is not NULL, the base mft record must be mapped on entry
+ *	      and it will be left mapped on return.
+ */
+int ntfs_map_runlist_nolock(ntfs_inode *ni, VCN vcn, ntfs_attr_search_ctx *ctx)
+{
+	VCN end_vcn;
+	unsigned long flags;
+	ntfs_inode *base_ni;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	runlist_element *rl;
+	struct page *put_this_page = NULL;
+	int err = 0;
+	bool ctx_is_temporary, ctx_needs_reset;
+	ntfs_attr_search_ctx old_ctx = { NULL, };
+
+	ntfs_debug("Mapping runlist part containing vcn 0x%llx.",
+			(unsigned long long)vcn);
+	if (!NInoAttr(ni))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	if (!ctx) {
+		ctx_is_temporary = ctx_needs_reset = true;
+		m = map_mft_record(base_ni);
+		if (IS_ERR(m))
+			return PTR_ERR(m);
+		ctx = ntfs_attr_get_search_ctx(base_ni, m);
+		if (unlikely(!ctx)) {
+			err = -ENOMEM;
+			goto err_out;
+		}
+	} else {
+		VCN allocated_size_vcn;
+
+		BUG_ON(IS_ERR(ctx->mrec));
+		a = ctx->attr;
+		BUG_ON(!a->non_resident);
+		ctx_is_temporary = false;
+		end_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
+		read_lock_irqsave(&ni->size_lock, flags);
+		allocated_size_vcn = ni->allocated_size >>
+				ni->vol->cluster_size_bits;
+		read_unlock_irqrestore(&ni->size_lock, flags);
+		if (!a->data.non_resident.lowest_vcn && end_vcn <= 0)
+			end_vcn = allocated_size_vcn - 1;
+		/*
+		 * If we already have the attribute extent containing @vcn in
+		 * @ctx, no need to look it up again.  We slightly cheat in
+		 * that if vcn exceeds the allocated size, we will refuse to
+		 * map the runlist below, so there is definitely no need to get
+		 * the right attribute extent.
+		 */
+		if (vcn >= allocated_size_vcn || (a->type == ni->type &&
+				a->name_length == ni->name_len &&
+				!memcmp((u8*)a + le16_to_cpu(a->name_offset),
+				ni->name, ni->name_len) &&
+				sle64_to_cpu(a->data.non_resident.lowest_vcn)
+				<= vcn && end_vcn >= vcn))
+			ctx_needs_reset = false;
+		else {
+			/* Save the old search context. */
+			old_ctx = *ctx;
+			/*
+			 * If the currently mapped (extent) inode is not the
+			 * base inode we will unmap it when we reinitialize the
+			 * search context which means we need to get a
+			 * reference to the page containing the mapped mft
+			 * record so we do not accidentally drop changes to the
+			 * mft record when it has not been marked dirty yet.
+			 */
+			if (old_ctx.base_ntfs_ino && old_ctx.ntfs_ino !=
+					old_ctx.base_ntfs_ino) {
+				put_this_page = old_ctx.ntfs_ino->page;
+				get_page(put_this_page);
+			}
+			/*
+			 * Reinitialize the search context so we can lookup the
+			 * needed attribute extent.
+			 */
+			ntfs_attr_reinit_search_ctx(ctx);
+			ctx_needs_reset = true;
+		}
+	}
+	if (ctx_needs_reset) {
+		err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+				CASE_SENSITIVE, vcn, NULL, 0, ctx);
+		if (unlikely(err)) {
+			if (err == -ENOENT)
+				err = -EIO;
+			goto err_out;
+		}
+		BUG_ON(!ctx->attr->non_resident);
+	}
+	a = ctx->attr;
+	/*
+	 * Only decompress the mapping pairs if @vcn is inside it.  Otherwise
+	 * we get into problems when we try to map an out of bounds vcn because
+	 * we then try to map the already mapped runlist fragment and
+	 * ntfs_mapping_pairs_decompress() fails.
+	 */
+	end_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn) + 1;
+	if (unlikely(vcn && vcn >= end_vcn)) {
+		err = -ENOENT;
+		goto err_out;
+	}
+	rl = ntfs_mapping_pairs_decompress(ni->vol, a, ni->runlist.rl);
+	if (IS_ERR(rl))
+		err = PTR_ERR(rl);
+	else
+		ni->runlist.rl = rl;
+err_out:
+	if (ctx_is_temporary) {
+		if (likely(ctx))
+			ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(base_ni);
+	} else if (ctx_needs_reset) {
+		/*
+		 * If there is no attribute list, restoring the search context
+		 * is accomplished simply by copying the saved context back over
+		 * the caller supplied context.  If there is an attribute list,
+		 * things are more complicated as we need to deal with mapping
+		 * of mft records and resulting potential changes in pointers.
+		 */
+		if (NInoAttrList(base_ni)) {
+			/*
+			 * If the currently mapped (extent) inode is not the
+			 * one we had before, we need to unmap it and map the
+			 * old one.
+			 */
+			if (ctx->ntfs_ino != old_ctx.ntfs_ino) {
+				/*
+				 * If the currently mapped inode is not the
+				 * base inode, unmap it.
+				 */
+				if (ctx->base_ntfs_ino && ctx->ntfs_ino !=
+						ctx->base_ntfs_ino) {
+					unmap_extent_mft_record(ctx->ntfs_ino);
+					ctx->mrec = ctx->base_mrec;
+					BUG_ON(!ctx->mrec);
+				}
+				/*
+				 * If the old mapped inode is not the base
+				 * inode, map it.
+				 */
+				if (old_ctx.base_ntfs_ino &&
+						old_ctx.ntfs_ino !=
+						old_ctx.base_ntfs_ino) {
+retry_map:
+					ctx->mrec = map_mft_record(
+							old_ctx.ntfs_ino);
+					/*
+					 * Something bad has happened.  If out
+					 * of memory retry till it succeeds.
+					 * Any other errors are fatal and we
+					 * return the error code in ctx->mrec.
+					 * Let the caller deal with it...  We
+					 * just need to fudge things so the
+					 * caller can reinit and/or put the
+					 * search context safely.
+					 */
+					if (IS_ERR(ctx->mrec)) {
+						if (PTR_ERR(ctx->mrec) ==
+								-ENOMEM) {
+							schedule();
+							goto retry_map;
+						} else
+							old_ctx.ntfs_ino =
+								old_ctx.
+								base_ntfs_ino;
+					}
+				}
+			}
+			/* Update the changed pointers in the saved context. */
+			if (ctx->mrec != old_ctx.mrec) {
+				if (!IS_ERR(ctx->mrec))
+					old_ctx.attr = (ATTR_RECORD*)(
+							(u8*)ctx->mrec +
+							((u8*)old_ctx.attr -
+							(u8*)old_ctx.mrec));
+				old_ctx.mrec = ctx->mrec;
+			}
+		}
+		/* Restore the search context to the saved one. */
+		*ctx = old_ctx;
+		/*
+		 * We drop the reference on the page we took earlier.  In the
+		 * case that IS_ERR(ctx->mrec) is true this means we might lose
+		 * some changes to the mft record that had been made between
+		 * the last time it was marked dirty/written out and now.  This
+		 * at this stage is not a problem as the mapping error is fatal
+		 * enough that the mft record cannot be written out anyway and
+		 * the caller is very likely to shutdown the whole inode
+		 * immediately and mark the volume dirty for chkdsk to pick up
+		 * the pieces anyway.
+		 */
+		if (put_this_page)
+			put_page(put_this_page);
+	}
+	return err;
+}
+
+/**
+ * ntfs_map_runlist - map (a part of) a runlist of an ntfs inode
+ * @ni:		ntfs inode for which to map (part of) a runlist
+ * @vcn:	map runlist part containing this vcn
+ *
+ * Map the part of a runlist containing the @vcn of the ntfs inode @ni.
+ *
+ * Return 0 on success and -errno on error.  There is one special error code
+ * which is not an error as such.  This is -ENOENT.  It means that @vcn is out
+ * of bounds of the runlist.
+ *
+ * Locking: - The runlist must be unlocked on entry and is unlocked on return.
+ *	    - This function takes the runlist lock for writing and may modify
+ *	      the runlist.
+ */
+int ntfs_map_runlist(ntfs_inode *ni, VCN vcn)
+{
+	int err = 0;
+
+	down_write(&ni->runlist.lock);
+	/* Make sure someone else didn't do the work while we were sleeping. */
+	if (likely(ntfs_rl_vcn_to_lcn(ni->runlist.rl, vcn) <=
+			LCN_RL_NOT_MAPPED))
+		err = ntfs_map_runlist_nolock(ni, vcn, NULL);
+	up_write(&ni->runlist.lock);
+	return err;
+}
+
+/**
+ * ntfs_attr_vcn_to_lcn_nolock - convert a vcn into a lcn given an ntfs inode
+ * @ni:			ntfs inode of the attribute whose runlist to search
+ * @vcn:		vcn to convert
+ * @write_locked:	true if the runlist is locked for writing
+ *
+ * Find the virtual cluster number @vcn in the runlist of the ntfs attribute
+ * described by the ntfs inode @ni and return the corresponding logical cluster
+ * number (lcn).
+ *
+ * If the @vcn is not mapped yet, the attempt is made to map the attribute
+ * extent containing the @vcn and the vcn to lcn conversion is retried.
+ *
+ * If @write_locked is true the caller has locked the runlist for writing and
+ * if false for reading.
+ *
+ * Since lcns must be >= 0, we use negative return codes with special meaning:
+ *
+ * Return code	Meaning / Description
+ * ==========================================
+ *  LCN_HOLE	Hole / not allocated on disk.
+ *  LCN_ENOENT	There is no such vcn in the runlist, i.e. @vcn is out of bounds.
+ *  LCN_ENOMEM	Not enough memory to map runlist.
+ *  LCN_EIO	Critical error (runlist/file is corrupt, i/o error, etc).
+ *
+ * Locking: - The runlist must be locked on entry and is left locked on return.
+ *	    - If @write_locked is 'false', i.e. the runlist is locked for reading,
+ *	      the lock may be dropped inside the function so you cannot rely on
+ *	      the runlist still being the same when this function returns.
+ */
+LCN ntfs_attr_vcn_to_lcn_nolock(ntfs_inode *ni, const VCN vcn,
+		const bool write_locked)
+{
+	LCN lcn;
+	unsigned long flags;
+	bool is_retry = false;
+
+	BUG_ON(!ni);
+	ntfs_debug("Entering for i_ino 0x%lx, vcn 0x%llx, %s_locked.",
+			ni->mft_no, (unsigned long long)vcn,
+			write_locked ? "write" : "read");
+	BUG_ON(!NInoNonResident(ni));
+	BUG_ON(vcn < 0);
+	if (!ni->runlist.rl) {
+		read_lock_irqsave(&ni->size_lock, flags);
+		if (!ni->allocated_size) {
+			read_unlock_irqrestore(&ni->size_lock, flags);
+			return LCN_ENOENT;
+		}
+		read_unlock_irqrestore(&ni->size_lock, flags);
+	}
+retry_remap:
+	/* Convert vcn to lcn.  If that fails map the runlist and retry once. */
+	lcn = ntfs_rl_vcn_to_lcn(ni->runlist.rl, vcn);
+	if (likely(lcn >= LCN_HOLE)) {
+		ntfs_debug("Done, lcn 0x%llx.", (long long)lcn);
+		return lcn;
+	}
+	if (lcn != LCN_RL_NOT_MAPPED) {
+		if (lcn != LCN_ENOENT)
+			lcn = LCN_EIO;
+	} else if (!is_retry) {
+		int err;
+
+		if (!write_locked) {
+			up_read(&ni->runlist.lock);
+			down_write(&ni->runlist.lock);
+			if (unlikely(ntfs_rl_vcn_to_lcn(ni->runlist.rl, vcn) !=
+					LCN_RL_NOT_MAPPED)) {
+				up_write(&ni->runlist.lock);
+				down_read(&ni->runlist.lock);
+				goto retry_remap;
+			}
+		}
+		err = ntfs_map_runlist_nolock(ni, vcn, NULL);
+		if (!write_locked) {
+			up_write(&ni->runlist.lock);
+			down_read(&ni->runlist.lock);
+		}
+		if (likely(!err)) {
+			is_retry = true;
+			goto retry_remap;
+		}
+		if (err == -ENOENT)
+			lcn = LCN_ENOENT;
+		else if (err == -ENOMEM)
+			lcn = LCN_ENOMEM;
+		else
+			lcn = LCN_EIO;
+	}
+	if (lcn != LCN_ENOENT)
+		ntfs_error(ni->vol->sb, "Failed with error code %lli.",
+				(long long)lcn);
+	return lcn;
+}
+
+/**
+ * ntfs_attr_find_vcn_nolock - find a vcn in the runlist of an ntfs inode
+ * @ni:		ntfs inode describing the runlist to search
+ * @vcn:	vcn to find
+ * @ctx:	active attribute search context if present or NULL if not
+ *
+ * Find the virtual cluster number @vcn in the runlist described by the ntfs
+ * inode @ni and return the address of the runlist element containing the @vcn.
+ *
+ * If the @vcn is not mapped yet, the attempt is made to map the attribute
+ * extent containing the @vcn and the vcn to lcn conversion is retried.
+ *
+ * If @ctx is specified, it is an active search context of @ni and its base mft
+ * record.  This is needed when ntfs_attr_find_vcn_nolock() encounters unmapped
+ * runlist fragments and allows their mapping.  If you do not have the mft
+ * record mapped, you can specify @ctx as NULL and ntfs_attr_find_vcn_nolock()
+ * will perform the necessary mapping and unmapping.
+ *
+ * Note, ntfs_attr_find_vcn_nolock() saves the state of @ctx on entry and
+ * restores it before returning.  Thus, @ctx will be left pointing to the same
+ * attribute on return as on entry.  However, the actual pointers in @ctx may
+ * point to different memory locations on return, so you must remember to reset
+ * any cached pointers from the @ctx, i.e. after the call to
+ * ntfs_attr_find_vcn_nolock(), you will probably want to do:
+ *	m = ctx->mrec;
+ *	a = ctx->attr;
+ * Assuming you cache ctx->attr in a variable @a of type ATTR_RECORD * and that
+ * you cache ctx->mrec in a variable @m of type MFT_RECORD *.
+ * Note you need to distinguish between the lcn of the returned runlist element
+ * being >= 0 and LCN_HOLE.  In the later case you have to return zeroes on
+ * read and allocate clusters on write.
+ *
+ * Return the runlist element containing the @vcn on success and
+ * ERR_PTR(-errno) on error.  You need to test the return value with IS_ERR()
+ * to decide if the return is success or failure and PTR_ERR() to get to the
+ * error code if IS_ERR() is true.
+ *
+ * The possible error return codes are:
+ *	-ENOENT - No such vcn in the runlist, i.e. @vcn is out of bounds.
+ *	-ENOMEM - Not enough memory to map runlist.
+ *	-EIO	- Critical error (runlist/file is corrupt, i/o error, etc).
+ *
+ * WARNING: If @ctx is supplied, regardless of whether success or failure is
+ *	    returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
+ *	    is no longer valid, i.e. you need to either call
+ *	    ntfs_attr_reinit_search_ctx() or ntfs_attr_put_search_ctx() on it.
+ *	    In that case PTR_ERR(@ctx->mrec) will give you the error code for
+ *	    why the mapping of the old inode failed.
+ *
+ * Locking: - The runlist described by @ni must be locked for writing on entry
+ *	      and is locked on return.  Note the runlist may be modified when
+ *	      needed runlist fragments need to be mapped.
+ *	    - If @ctx is NULL, the base mft record of @ni must not be mapped on
+ *	      entry and it will be left unmapped on return.
+ *	    - If @ctx is not NULL, the base mft record must be mapped on entry
+ *	      and it will be left mapped on return.
+ */
+runlist_element *ntfs_attr_find_vcn_nolock(ntfs_inode *ni, const VCN vcn,
+		ntfs_attr_search_ctx *ctx)
+{
+	unsigned long flags;
+	runlist_element *rl;
+	int err = 0;
+	bool is_retry = false;
+
+	BUG_ON(!ni);
+	ntfs_debug("Entering for i_ino 0x%lx, vcn 0x%llx, with%s ctx.",
+			ni->mft_no, (unsigned long long)vcn, ctx ? "" : "out");
+	BUG_ON(!NInoNonResident(ni));
+	BUG_ON(vcn < 0);
+	if (!ni->runlist.rl) {
+		read_lock_irqsave(&ni->size_lock, flags);
+		if (!ni->allocated_size) {
+			read_unlock_irqrestore(&ni->size_lock, flags);
+			return ERR_PTR(-ENOENT);
+		}
+		read_unlock_irqrestore(&ni->size_lock, flags);
+	}
+retry_remap:
+	rl = ni->runlist.rl;
+	if (likely(rl && vcn >= rl[0].vcn)) {
+		while (likely(rl->length)) {
+			if (unlikely(vcn < rl[1].vcn)) {
+				if (likely(rl->lcn >= LCN_HOLE)) {
+					ntfs_debug("Done.");
+					return rl;
+				}
+				break;
+			}
+			rl++;
+		}
+		if (likely(rl->lcn != LCN_RL_NOT_MAPPED)) {
+			if (likely(rl->lcn == LCN_ENOENT))
+				err = -ENOENT;
+			else
+				err = -EIO;
+		}
+	}
+	if (!err && !is_retry) {
+		/*
+		 * If the search context is invalid we cannot map the unmapped
+		 * region.
+		 */
+		if (IS_ERR(ctx->mrec))
+			err = PTR_ERR(ctx->mrec);
+		else {
+			/*
+			 * The @vcn is in an unmapped region, map the runlist
+			 * and retry.
+			 */
+			err = ntfs_map_runlist_nolock(ni, vcn, ctx);
+			if (likely(!err)) {
+				is_retry = true;
+				goto retry_remap;
+			}
+		}
+		if (err == -EINVAL)
+			err = -EIO;
+	} else if (!err)
+		err = -EIO;
+	if (err != -ENOENT)
+		ntfs_error(ni->vol->sb, "Failed with error code %i.", err);
+	return ERR_PTR(err);
+}
+
+/**
+ * ntfs_attr_find - find (next) attribute in mft record
+ * @type:	attribute type to find
+ * @name:	attribute name to find (optional, i.e. NULL means don't care)
+ * @name_len:	attribute name length (only needed if @name present)
+ * @ic:		IGNORE_CASE or CASE_SENSITIVE (ignored if @name not present)
+ * @val:	attribute value to find (optional, resident attributes only)
+ * @val_len:	attribute value length
+ * @ctx:	search context with mft record and attribute to search from
+ *
+ * You should not need to call this function directly.  Use ntfs_attr_lookup()
+ * instead.
+ *
+ * ntfs_attr_find() takes a search context @ctx as parameter and searches the
+ * mft record specified by @ctx->mrec, beginning at @ctx->attr, for an
+ * attribute of @type, optionally @name and @val.
+ *
+ * If the attribute is found, ntfs_attr_find() returns 0 and @ctx->attr will
+ * point to the found attribute.
+ *
+ * If the attribute is not found, ntfs_attr_find() returns -ENOENT and
+ * @ctx->attr will point to the attribute before which the attribute being
+ * searched for would need to be inserted if such an action were to be desired.
+ *
+ * On actual error, ntfs_attr_find() returns -EIO.  In this case @ctx->attr is
+ * undefined and in particular do not rely on it not changing.
+ *
+ * If @ctx->is_first is 'true', the search begins with @ctx->attr itself.  If it
+ * is 'false', the search begins after @ctx->attr.
+ *
+ * If @ic is IGNORE_CASE, the @name comparisson is not case sensitive and
+ * @ctx->ntfs_ino must be set to the ntfs inode to which the mft record
+ * @ctx->mrec belongs.  This is so we can get at the ntfs volume and hence at
+ * the upcase table.  If @ic is CASE_SENSITIVE, the comparison is case
+ * sensitive.  When @name is present, @name_len is the @name length in Unicode
+ * characters.
+ *
+ * If @name is not present (NULL), we assume that the unnamed attribute is
+ * being searched for.
+ *
+ * Finally, the resident attribute value @val is looked for, if present.  If
+ * @val is not present (NULL), @val_len is ignored.
+ *
+ * ntfs_attr_find() only searches the specified mft record and it ignores the
+ * presence of an attribute list attribute (unless it is the one being searched
+ * for, obviously).  If you need to take attribute lists into consideration,
+ * use ntfs_attr_lookup() instead (see below).  This also means that you cannot
+ * use ntfs_attr_find() to search for extent records of non-resident
+ * attributes, as extents with lowest_vcn != 0 are usually described by the
+ * attribute list attribute only. - Note that it is possible that the first
+ * extent is only in the attribute list while the last extent is in the base
+ * mft record, so do not rely on being able to find the first extent in the
+ * base mft record.
+ *
+ * Warning: Never use @val when looking for attribute types which can be
+ *	    non-resident as this most likely will result in a crash!
+ */
+static int ntfs_attr_find(const ATTR_TYPE type, const ntfschar *name,
+		const u32 name_len, const IGNORE_CASE_BOOL ic,
+		const u8 *val, const u32 val_len, ntfs_attr_search_ctx *ctx)
+{
+	ATTR_RECORD *a;
+	ntfs_volume *vol = ctx->ntfs_ino->vol;
+	ntfschar *upcase = vol->upcase;
+	u32 upcase_len = vol->upcase_len;
+
+	/*
+	 * Iterate over attributes in mft record starting at @ctx->attr, or the
+	 * attribute following that, if @ctx->is_first is 'true'.
+	 */
+	if (ctx->is_first) {
+		a = ctx->attr;
+		ctx->is_first = false;
+	} else
+		a = (ATTR_RECORD*)((u8*)ctx->attr +
+				le32_to_cpu(ctx->attr->length));
+	for (;;	a = (ATTR_RECORD*)((u8*)a + le32_to_cpu(a->length))) {
+		u8 *mrec_end = (u8 *)ctx->mrec +
+		               le32_to_cpu(ctx->mrec->bytes_allocated);
+		u8 *name_end;
+
+		/* check whether ATTR_RECORD wrap */
+		if ((u8 *)a < (u8 *)ctx->mrec)
+			break;
+
+		/* check whether Attribute Record Header is within bounds */
+		if ((u8 *)a > mrec_end ||
+		    (u8 *)a + sizeof(ATTR_RECORD) > mrec_end)
+			break;
+
+		/* check whether ATTR_RECORD's name is within bounds */
+		name_end = (u8 *)a + le16_to_cpu(a->name_offset) +
+			   a->name_length * sizeof(ntfschar);
+		if (name_end > mrec_end)
+			break;
+
+		ctx->attr = a;
+		if (unlikely(le32_to_cpu(a->type) > le32_to_cpu(type) ||
+				a->type == AT_END))
+			return -ENOENT;
+		if (unlikely(!a->length))
+			break;
+
+		/* check whether ATTR_RECORD's length wrap */
+		if ((u8 *)a + le32_to_cpu(a->length) < (u8 *)a)
+			break;
+		/* check whether ATTR_RECORD's length is within bounds */
+		if ((u8 *)a + le32_to_cpu(a->length) > mrec_end)
+			break;
+
+		if (a->type != type)
+			continue;
+		/*
+		 * If @name is present, compare the two names.  If @name is
+		 * missing, assume we want an unnamed attribute.
+		 */
+		if (!name) {
+			/* The search failed if the found attribute is named. */
+			if (a->name_length)
+				return -ENOENT;
+		} else if (!ntfs_are_names_equal(name, name_len,
+			    (ntfschar*)((u8*)a + le16_to_cpu(a->name_offset)),
+			    a->name_length, ic, upcase, upcase_len)) {
+			register int rc;
+
+			rc = ntfs_collate_names(name, name_len,
+					(ntfschar*)((u8*)a +
+					le16_to_cpu(a->name_offset)),
+					a->name_length, 1, IGNORE_CASE,
+					upcase, upcase_len);
+			/*
+			 * If @name collates before a->name, there is no
+			 * matching attribute.
+			 */
+			if (rc == -1)
+				return -ENOENT;
+			/* If the strings are not equal, continue search. */
+			if (rc)
+				continue;
+			rc = ntfs_collate_names(name, name_len,
+					(ntfschar*)((u8*)a +
+					le16_to_cpu(a->name_offset)),
+					a->name_length, 1, CASE_SENSITIVE,
+					upcase, upcase_len);
+			if (rc == -1)
+				return -ENOENT;
+			if (rc)
+				continue;
+		}
+		/*
+		 * The names match or @name not present and attribute is
+		 * unnamed.  If no @val specified, we have found the attribute
+		 * and are done.
+		 */
+		if (!val)
+			return 0;
+		/* @val is present; compare values. */
+		else {
+			register int rc;
+
+			rc = memcmp(val, (u8*)a + le16_to_cpu(
+					a->data.resident.value_offset),
+					min_t(u32, val_len, le32_to_cpu(
+					a->data.resident.value_length)));
+			/*
+			 * If @val collates before the current attribute's
+			 * value, there is no matching attribute.
+			 */
+			if (!rc) {
+				register u32 avl;
+
+				avl = le32_to_cpu(
+						a->data.resident.value_length);
+				if (val_len == avl)
+					return 0;
+				if (val_len < avl)
+					return -ENOENT;
+			} else if (rc < 0)
+				return -ENOENT;
+		}
+	}
+	ntfs_error(vol->sb, "Inode is corrupt.  Run chkdsk.");
+	NVolSetErrors(vol);
+	return -EIO;
+}
+
+/**
+ * load_attribute_list - load an attribute list into memory
+ * @vol:		ntfs volume from which to read
+ * @runlist:		runlist of the attribute list
+ * @al_start:		destination buffer
+ * @size:		size of the destination buffer in bytes
+ * @initialized_size:	initialized size of the attribute list
+ *
+ * Walk the runlist @runlist and load all clusters from it copying them into
+ * the linear buffer @al. The maximum number of bytes copied to @al is @size
+ * bytes. Note, @size does not need to be a multiple of the cluster size. If
+ * @initialized_size is less than @size, the region in @al between
+ * @initialized_size and @size will be zeroed and not read from disk.
+ *
+ * Return 0 on success or -errno on error.
+ */
+int load_attribute_list(ntfs_volume *vol, runlist *runlist, u8 *al_start,
+		const s64 size, const s64 initialized_size)
+{
+	LCN lcn;
+	u8 *al = al_start;
+	u8 *al_end = al + initialized_size;
+	runlist_element *rl;
+	struct buffer_head *bh;
+	struct super_block *sb;
+	unsigned long block_size;
+	unsigned long block, max_block;
+	int err = 0;
+	unsigned char block_size_bits;
+
+	ntfs_debug("Entering.");
+	if (!vol || !runlist || !al || size <= 0 || initialized_size < 0 ||
+			initialized_size > size)
+		return -EINVAL;
+	if (!initialized_size) {
+		memset(al, 0, size);
+		return 0;
+	}
+	sb = vol->sb;
+	block_size = sb->s_blocksize;
+	block_size_bits = sb->s_blocksize_bits;
+	down_read(&runlist->lock);
+	rl = runlist->rl;
+	if (!rl) {
+		ntfs_error(sb, "Cannot read attribute list since runlist is "
+				"missing.");
+		goto err_out;	
+	}
+	/* Read all clusters specified by the runlist one run at a time. */
+	while (rl->length) {
+		lcn = ntfs_rl_vcn_to_lcn(rl, rl->vcn);
+		ntfs_debug("Reading vcn = 0x%llx, lcn = 0x%llx.",
+				(unsigned long long)rl->vcn,
+				(unsigned long long)lcn);
+		/* The attribute list cannot be sparse. */
+		if (lcn < 0) {
+			ntfs_error(sb, "ntfs_rl_vcn_to_lcn() failed.  Cannot "
+					"read attribute list.");
+			goto err_out;
+		}
+		block = lcn << vol->cluster_size_bits >> block_size_bits;
+		/* Read the run from device in chunks of block_size bytes. */
+		max_block = block + (rl->length << vol->cluster_size_bits >>
+				block_size_bits);
+		ntfs_debug("max_block = 0x%lx.", max_block);
+		do {
+			ntfs_debug("Reading block = 0x%lx.", block);
+			bh = sb_bread(sb, block);
+			if (!bh) {
+				ntfs_error(sb, "sb_bread() failed. Cannot "
+						"read attribute list.");
+				goto err_out;
+			}
+			if (al + block_size >= al_end)
+				goto do_final;
+			memcpy(al, bh->b_data, block_size);
+			brelse(bh);
+			al += block_size;
+		} while (++block < max_block);
+		rl++;
+	}
+	if (initialized_size < size) {
+initialize:
+		memset(al_start + initialized_size, 0, size - initialized_size);
+	}
+done:
+	up_read(&runlist->lock);
+	return err;
+do_final:
+	if (al < al_end) {
+		/*
+		 * Partial block.
+		 *
+		 * Note: The attribute list can be smaller than its allocation
+		 * by multiple clusters.  This has been encountered by at least
+		 * two people running Windows XP, thus we cannot do any
+		 * truncation sanity checking here. (AIA)
+		 */
+		memcpy(al, bh->b_data, al_end - al);
+		brelse(bh);
+		if (initialized_size < size)
+			goto initialize;
+		goto done;
+	}
+	brelse(bh);
+	/* Real overflow! */
+	ntfs_error(sb, "Attribute list buffer overflow. Read attribute list "
+			"is truncated.");
+err_out:
+	err = -EIO;
+	goto done;
+}
+
+/**
+ * ntfs_external_attr_find - find an attribute in the attribute list of an inode
+ * @type:	attribute type to find
+ * @name:	attribute name to find (optional, i.e. NULL means don't care)
+ * @name_len:	attribute name length (only needed if @name present)
+ * @ic:		IGNORE_CASE or CASE_SENSITIVE (ignored if @name not present)
+ * @lowest_vcn:	lowest vcn to find (optional, non-resident attributes only)
+ * @val:	attribute value to find (optional, resident attributes only)
+ * @val_len:	attribute value length
+ * @ctx:	search context with mft record and attribute to search from
+ *
+ * You should not need to call this function directly.  Use ntfs_attr_lookup()
+ * instead.
+ *
+ * Find an attribute by searching the attribute list for the corresponding
+ * attribute list entry.  Having found the entry, map the mft record if the
+ * attribute is in a different mft record/inode, ntfs_attr_find() the attribute
+ * in there and return it.
+ *
+ * On first search @ctx->ntfs_ino must be the base mft record and @ctx must
+ * have been obtained from a call to ntfs_attr_get_search_ctx().  On subsequent
+ * calls @ctx->ntfs_ino can be any extent inode, too (@ctx->base_ntfs_ino is
+ * then the base inode).
+ *
+ * After finishing with the attribute/mft record you need to call
+ * ntfs_attr_put_search_ctx() to cleanup the search context (unmapping any
+ * mapped inodes, etc).
+ *
+ * If the attribute is found, ntfs_external_attr_find() returns 0 and
+ * @ctx->attr will point to the found attribute.  @ctx->mrec will point to the
+ * mft record in which @ctx->attr is located and @ctx->al_entry will point to
+ * the attribute list entry for the attribute.
+ *
+ * If the attribute is not found, ntfs_external_attr_find() returns -ENOENT and
+ * @ctx->attr will point to the attribute in the base mft record before which
+ * the attribute being searched for would need to be inserted if such an action
+ * were to be desired.  @ctx->mrec will point to the mft record in which
+ * @ctx->attr is located and @ctx->al_entry will point to the attribute list
+ * entry of the attribute before which the attribute being searched for would
+ * need to be inserted if such an action were to be desired.
+ *
+ * Thus to insert the not found attribute, one wants to add the attribute to
+ * @ctx->mrec (the base mft record) and if there is not enough space, the
+ * attribute should be placed in a newly allocated extent mft record.  The
+ * attribute list entry for the inserted attribute should be inserted in the
+ * attribute list attribute at @ctx->al_entry.
+ *
+ * On actual error, ntfs_external_attr_find() returns -EIO.  In this case
+ * @ctx->attr is undefined and in particular do not rely on it not changing.
+ */
+static int ntfs_external_attr_find(const ATTR_TYPE type,
+		const ntfschar *name, const u32 name_len,
+		const IGNORE_CASE_BOOL ic, const VCN lowest_vcn,
+		const u8 *val, const u32 val_len, ntfs_attr_search_ctx *ctx)
+{
+	ntfs_inode *base_ni, *ni;
+	ntfs_volume *vol;
+	ATTR_LIST_ENTRY *al_entry, *next_al_entry;
+	u8 *al_start, *al_end;
+	ATTR_RECORD *a;
+	ntfschar *al_name;
+	u32 al_name_len;
+	int err = 0;
+	static const char *es = " Unmount and run chkdsk.";
+
+	ni = ctx->ntfs_ino;
+	base_ni = ctx->base_ntfs_ino;
+	ntfs_debug("Entering for inode 0x%lx, type 0x%x.", ni->mft_no, type);
+	if (!base_ni) {
+		/* First call happens with the base mft record. */
+		base_ni = ctx->base_ntfs_ino = ctx->ntfs_ino;
+		ctx->base_mrec = ctx->mrec;
+	}
+	if (ni == base_ni)
+		ctx->base_attr = ctx->attr;
+	if (type == AT_END)
+		goto not_found;
+	vol = base_ni->vol;
+	al_start = base_ni->attr_list;
+	al_end = al_start + base_ni->attr_list_size;
+	if (!ctx->al_entry)
+		ctx->al_entry = (ATTR_LIST_ENTRY*)al_start;
+	/*
+	 * Iterate over entries in attribute list starting at @ctx->al_entry,
+	 * or the entry following that, if @ctx->is_first is 'true'.
+	 */
+	if (ctx->is_first) {
+		al_entry = ctx->al_entry;
+		ctx->is_first = false;
+	} else
+		al_entry = (ATTR_LIST_ENTRY*)((u8*)ctx->al_entry +
+				le16_to_cpu(ctx->al_entry->length));
+	for (;; al_entry = next_al_entry) {
+		/* Out of bounds check. */
+		if ((u8*)al_entry < base_ni->attr_list ||
+				(u8*)al_entry > al_end)
+			break;	/* Inode is corrupt. */
+		ctx->al_entry = al_entry;
+		/* Catch the end of the attribute list. */
+		if ((u8*)al_entry == al_end)
+			goto not_found;
+		if (!al_entry->length)
+			break;
+		if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
+				le16_to_cpu(al_entry->length) > al_end)
+			break;
+		next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
+				le16_to_cpu(al_entry->length));
+		if (le32_to_cpu(al_entry->type) > le32_to_cpu(type))
+			goto not_found;
+		if (type != al_entry->type)
+			continue;
+		/*
+		 * If @name is present, compare the two names.  If @name is
+		 * missing, assume we want an unnamed attribute.
+		 */
+		al_name_len = al_entry->name_length;
+		al_name = (ntfschar*)((u8*)al_entry + al_entry->name_offset);
+		if (!name) {
+			if (al_name_len)
+				goto not_found;
+		} else if (!ntfs_are_names_equal(al_name, al_name_len, name,
+				name_len, ic, vol->upcase, vol->upcase_len)) {
+			register int rc;
+
+			rc = ntfs_collate_names(name, name_len, al_name,
+					al_name_len, 1, IGNORE_CASE,
+					vol->upcase, vol->upcase_len);
+			/*
+			 * If @name collates before al_name, there is no
+			 * matching attribute.
+			 */
+			if (rc == -1)
+				goto not_found;
+			/* If the strings are not equal, continue search. */
+			if (rc)
+				continue;
+			/*
+			 * FIXME: Reverse engineering showed 0, IGNORE_CASE but
+			 * that is inconsistent with ntfs_attr_find().  The
+			 * subsequent rc checks were also different.  Perhaps I
+			 * made a mistake in one of the two.  Need to recheck
+			 * which is correct or at least see what is going on...
+			 * (AIA)
+			 */
+			rc = ntfs_collate_names(name, name_len, al_name,
+					al_name_len, 1, CASE_SENSITIVE,
+					vol->upcase, vol->upcase_len);
+			if (rc == -1)
+				goto not_found;
+			if (rc)
+				continue;
+		}
+		/*
+		 * The names match or @name not present and attribute is
+		 * unnamed.  Now check @lowest_vcn.  Continue search if the
+		 * next attribute list entry still fits @lowest_vcn.  Otherwise
+		 * we have reached the right one or the search has failed.
+		 */
+		if (lowest_vcn && (u8*)next_al_entry >= al_start	    &&
+				(u8*)next_al_entry + 6 < al_end		    &&
+				(u8*)next_al_entry + le16_to_cpu(
+					next_al_entry->length) <= al_end    &&
+				sle64_to_cpu(next_al_entry->lowest_vcn) <=
+					lowest_vcn			    &&
+				next_al_entry->type == al_entry->type	    &&
+				next_al_entry->name_length == al_name_len   &&
+				ntfs_are_names_equal((ntfschar*)((u8*)
+					next_al_entry +
+					next_al_entry->name_offset),
+					next_al_entry->name_length,
+					al_name, al_name_len, CASE_SENSITIVE,
+					vol->upcase, vol->upcase_len))
+			continue;
+		if (MREF_LE(al_entry->mft_reference) == ni->mft_no) {
+			if (MSEQNO_LE(al_entry->mft_reference) != ni->seq_no) {
+				ntfs_error(vol->sb, "Found stale mft "
+						"reference in attribute list "
+						"of base inode 0x%lx.%s",
+						base_ni->mft_no, es);
+				err = -EIO;
+				break;
+			}
+		} else { /* Mft references do not match. */
+			/* If there is a mapped record unmap it first. */
+			if (ni != base_ni)
+				unmap_extent_mft_record(ni);
+			/* Do we want the base record back? */
+			if (MREF_LE(al_entry->mft_reference) ==
+					base_ni->mft_no) {
+				ni = ctx->ntfs_ino = base_ni;
+				ctx->mrec = ctx->base_mrec;
+			} else {
+				/* We want an extent record. */
+				ctx->mrec = map_extent_mft_record(base_ni,
+						le64_to_cpu(
+						al_entry->mft_reference), &ni);
+				if (IS_ERR(ctx->mrec)) {
+					ntfs_error(vol->sb, "Failed to map "
+							"extent mft record "
+							"0x%lx of base inode "
+							"0x%lx.%s",
+							MREF_LE(al_entry->
+							mft_reference),
+							base_ni->mft_no, es);
+					err = PTR_ERR(ctx->mrec);
+					if (err == -ENOENT)
+						err = -EIO;
+					/* Cause @ctx to be sanitized below. */
+					ni = NULL;
+					break;
+				}
+				ctx->ntfs_ino = ni;
+			}
+			ctx->attr = (ATTR_RECORD*)((u8*)ctx->mrec +
+					le16_to_cpu(ctx->mrec->attrs_offset));
+		}
+		/*
+		 * ctx->vfs_ino, ctx->mrec, and ctx->attr now point to the
+		 * mft record containing the attribute represented by the
+		 * current al_entry.
+		 */
+		/*
+		 * We could call into ntfs_attr_find() to find the right
+		 * attribute in this mft record but this would be less
+		 * efficient and not quite accurate as ntfs_attr_find() ignores
+		 * the attribute instance numbers for example which become
+		 * important when one plays with attribute lists.  Also,
+		 * because a proper match has been found in the attribute list
+		 * entry above, the comparison can now be optimized.  So it is
+		 * worth re-implementing a simplified ntfs_attr_find() here.
+		 */
+		a = ctx->attr;
+		/*
+		 * Use a manual loop so we can still use break and continue
+		 * with the same meanings as above.
+		 */
+do_next_attr_loop:
+		if ((u8*)a < (u8*)ctx->mrec || (u8*)a > (u8*)ctx->mrec +
+				le32_to_cpu(ctx->mrec->bytes_allocated))
+			break;
+		if (a->type == AT_END)
+			break;
+		if (!a->length)
+			break;
+		if (al_entry->instance != a->instance)
+			goto do_next_attr;
+		/*
+		 * If the type and/or the name are mismatched between the
+		 * attribute list entry and the attribute record, there is
+		 * corruption so we break and return error EIO.
+		 */
+		if (al_entry->type != a->type)
+			break;
+		if (!ntfs_are_names_equal((ntfschar*)((u8*)a +
+				le16_to_cpu(a->name_offset)), a->name_length,
+				al_name, al_name_len, CASE_SENSITIVE,
+				vol->upcase, vol->upcase_len))
+			break;
+		ctx->attr = a;
+		/*
+		 * If no @val specified or @val specified and it matches, we
+		 * have found it!
+		 */
+		if (!val || (!a->non_resident && le32_to_cpu(
+				a->data.resident.value_length) == val_len &&
+				!memcmp((u8*)a +
+				le16_to_cpu(a->data.resident.value_offset),
+				val, val_len))) {
+			ntfs_debug("Done, found.");
+			return 0;
+		}
+do_next_attr:
+		/* Proceed to the next attribute in the current mft record. */
+		a = (ATTR_RECORD*)((u8*)a + le32_to_cpu(a->length));
+		goto do_next_attr_loop;
+	}
+	if (!err) {
+		ntfs_error(vol->sb, "Base inode 0x%lx contains corrupt "
+				"attribute list attribute.%s", base_ni->mft_no,
+				es);
+		err = -EIO;
+	}
+	if (ni != base_ni) {
+		if (ni)
+			unmap_extent_mft_record(ni);
+		ctx->ntfs_ino = base_ni;
+		ctx->mrec = ctx->base_mrec;
+		ctx->attr = ctx->base_attr;
+	}
+	if (err != -ENOMEM)
+		NVolSetErrors(vol);
+	return err;
+not_found:
+	/*
+	 * If we were looking for AT_END, we reset the search context @ctx and
+	 * use ntfs_attr_find() to seek to the end of the base mft record.
+	 */
+	if (type == AT_END) {
+		ntfs_attr_reinit_search_ctx(ctx);
+		return ntfs_attr_find(AT_END, name, name_len, ic, val, val_len,
+				ctx);
+	}
+	/*
+	 * The attribute was not found.  Before we return, we want to ensure
+	 * @ctx->mrec and @ctx->attr indicate the position at which the
+	 * attribute should be inserted in the base mft record.  Since we also
+	 * want to preserve @ctx->al_entry we cannot reinitialize the search
+	 * context using ntfs_attr_reinit_search_ctx() as this would set
+	 * @ctx->al_entry to NULL.  Thus we do the necessary bits manually (see
+	 * ntfs_attr_init_search_ctx() below).  Note, we _only_ preserve
+	 * @ctx->al_entry as the remaining fields (base_*) are identical to
+	 * their non base_ counterparts and we cannot set @ctx->base_attr
+	 * correctly yet as we do not know what @ctx->attr will be set to by
+	 * the call to ntfs_attr_find() below.
+	 */
+	if (ni != base_ni)
+		unmap_extent_mft_record(ni);
+	ctx->mrec = ctx->base_mrec;
+	ctx->attr = (ATTR_RECORD*)((u8*)ctx->mrec +
+			le16_to_cpu(ctx->mrec->attrs_offset));
+	ctx->is_first = true;
+	ctx->ntfs_ino = base_ni;
+	ctx->base_ntfs_ino = NULL;
+	ctx->base_mrec = NULL;
+	ctx->base_attr = NULL;
+	/*
+	 * In case there are multiple matches in the base mft record, need to
+	 * keep enumerating until we get an attribute not found response (or
+	 * another error), otherwise we would keep returning the same attribute
+	 * over and over again and all programs using us for enumeration would
+	 * lock up in a tight loop.
+	 */
+	do {
+		err = ntfs_attr_find(type, name, name_len, ic, val, val_len,
+				ctx);
+	} while (!err);
+	ntfs_debug("Done, not found.");
+	return err;
+}
+
+/**
+ * ntfs_attr_lookup - find an attribute in an ntfs inode
+ * @type:	attribute type to find
+ * @name:	attribute name to find (optional, i.e. NULL means don't care)
+ * @name_len:	attribute name length (only needed if @name present)
+ * @ic:		IGNORE_CASE or CASE_SENSITIVE (ignored if @name not present)
+ * @lowest_vcn:	lowest vcn to find (optional, non-resident attributes only)
+ * @val:	attribute value to find (optional, resident attributes only)
+ * @val_len:	attribute value length
+ * @ctx:	search context with mft record and attribute to search from
+ *
+ * Find an attribute in an ntfs inode.  On first search @ctx->ntfs_ino must
+ * be the base mft record and @ctx must have been obtained from a call to
+ * ntfs_attr_get_search_ctx().
+ *
+ * This function transparently handles attribute lists and @ctx is used to
+ * continue searches where they were left off at.
+ *
+ * After finishing with the attribute/mft record you need to call
+ * ntfs_attr_put_search_ctx() to cleanup the search context (unmapping any
+ * mapped inodes, etc).
+ *
+ * Return 0 if the search was successful and -errno if not.
+ *
+ * When 0, @ctx->attr is the found attribute and it is in mft record
+ * @ctx->mrec.  If an attribute list attribute is present, @ctx->al_entry is
+ * the attribute list entry of the found attribute.
+ *
+ * When -ENOENT, @ctx->attr is the attribute which collates just after the
+ * attribute being searched for, i.e. if one wants to add the attribute to the
+ * mft record this is the correct place to insert it into.  If an attribute
+ * list attribute is present, @ctx->al_entry is the attribute list entry which
+ * collates just after the attribute list entry of the attribute being searched
+ * for, i.e. if one wants to add the attribute to the mft record this is the
+ * correct place to insert its attribute list entry into.
+ *
+ * When -errno != -ENOENT, an error occurred during the lookup.  @ctx->attr is
+ * then undefined and in particular you should not rely on it not changing.
+ */
+int ntfs_attr_lookup(const ATTR_TYPE type, const ntfschar *name,
+		const u32 name_len, const IGNORE_CASE_BOOL ic,
+		const VCN lowest_vcn, const u8 *val, const u32 val_len,
+		ntfs_attr_search_ctx *ctx)
+{
+	ntfs_inode *base_ni;
+
+	ntfs_debug("Entering.");
+	BUG_ON(IS_ERR(ctx->mrec));
+	if (ctx->base_ntfs_ino)
+		base_ni = ctx->base_ntfs_ino;
+	else
+		base_ni = ctx->ntfs_ino;
+	/* Sanity check, just for debugging really. */
+	BUG_ON(!base_ni);
+	if (!NInoAttrList(base_ni) || type == AT_ATTRIBUTE_LIST)
+		return ntfs_attr_find(type, name, name_len, ic, val, val_len,
+				ctx);
+	return ntfs_external_attr_find(type, name, name_len, ic, lowest_vcn,
+			val, val_len, ctx);
+}
+
+/**
+ * ntfs_attr_init_search_ctx - initialize an attribute search context
+ * @ctx:	attribute search context to initialize
+ * @ni:		ntfs inode with which to initialize the search context
+ * @mrec:	mft record with which to initialize the search context
+ *
+ * Initialize the attribute search context @ctx with @ni and @mrec.
+ */
+static inline void ntfs_attr_init_search_ctx(ntfs_attr_search_ctx *ctx,
+		ntfs_inode *ni, MFT_RECORD *mrec)
+{
+	*ctx = (ntfs_attr_search_ctx) {
+		.mrec = mrec,
+		/* Sanity checks are performed elsewhere. */
+		.attr = (ATTR_RECORD*)((u8*)mrec +
+				le16_to_cpu(mrec->attrs_offset)),
+		.is_first = true,
+		.ntfs_ino = ni,
+	};
+}
+
+/**
+ * ntfs_attr_reinit_search_ctx - reinitialize an attribute search context
+ * @ctx:	attribute search context to reinitialize
+ *
+ * Reinitialize the attribute search context @ctx, unmapping an associated
+ * extent mft record if present, and initialize the search context again.
+ *
+ * This is used when a search for a new attribute is being started to reset
+ * the search context to the beginning.
+ */
+void ntfs_attr_reinit_search_ctx(ntfs_attr_search_ctx *ctx)
+{
+	if (likely(!ctx->base_ntfs_ino)) {
+		/* No attribute list. */
+		ctx->is_first = true;
+		/* Sanity checks are performed elsewhere. */
+		ctx->attr = (ATTR_RECORD*)((u8*)ctx->mrec +
+				le16_to_cpu(ctx->mrec->attrs_offset));
+		/*
+		 * This needs resetting due to ntfs_external_attr_find() which
+		 * can leave it set despite having zeroed ctx->base_ntfs_ino.
+		 */
+		ctx->al_entry = NULL;
+		return;
+	} /* Attribute list. */
+	if (ctx->ntfs_ino != ctx->base_ntfs_ino)
+		unmap_extent_mft_record(ctx->ntfs_ino);
+	ntfs_attr_init_search_ctx(ctx, ctx->base_ntfs_ino, ctx->base_mrec);
+	return;
+}
+
+/**
+ * ntfs_attr_get_search_ctx - allocate/initialize a new attribute search context
+ * @ni:		ntfs inode with which to initialize the search context
+ * @mrec:	mft record with which to initialize the search context
+ *
+ * Allocate a new attribute search context, initialize it with @ni and @mrec,
+ * and return it. Return NULL if allocation failed.
+ */
+ntfs_attr_search_ctx *ntfs_attr_get_search_ctx(ntfs_inode *ni, MFT_RECORD *mrec)
+{
+	ntfs_attr_search_ctx *ctx;
+
+	ctx = kmem_cache_alloc(ntfs_attr_ctx_cache, GFP_NOFS);
+	if (ctx)
+		ntfs_attr_init_search_ctx(ctx, ni, mrec);
+	return ctx;
+}
+
+/**
+ * ntfs_attr_put_search_ctx - release an attribute search context
+ * @ctx:	attribute search context to free
+ *
+ * Release the attribute search context @ctx, unmapping an associated extent
+ * mft record if present.
+ */
+void ntfs_attr_put_search_ctx(ntfs_attr_search_ctx *ctx)
+{
+	if (ctx->base_ntfs_ino && ctx->ntfs_ino != ctx->base_ntfs_ino)
+		unmap_extent_mft_record(ctx->ntfs_ino);
+	kmem_cache_free(ntfs_attr_ctx_cache, ctx);
+	return;
+}
+
+#ifdef NTFS_RW
+
+/**
+ * ntfs_attr_find_in_attrdef - find an attribute in the $AttrDef system file
+ * @vol:	ntfs volume to which the attribute belongs
+ * @type:	attribute type which to find
+ *
+ * Search for the attribute definition record corresponding to the attribute
+ * @type in the $AttrDef system file.
+ *
+ * Return the attribute type definition record if found and NULL if not found.
+ */
+static ATTR_DEF *ntfs_attr_find_in_attrdef(const ntfs_volume *vol,
+		const ATTR_TYPE type)
+{
+	ATTR_DEF *ad;
+
+	BUG_ON(!vol->attrdef);
+	BUG_ON(!type);
+	for (ad = vol->attrdef; (u8*)ad - (u8*)vol->attrdef <
+			vol->attrdef_size && ad->type; ++ad) {
+		/* We have not found it yet, carry on searching. */
+		if (likely(le32_to_cpu(ad->type) < le32_to_cpu(type)))
+			continue;
+		/* We found the attribute; return it. */
+		if (likely(ad->type == type))
+			return ad;
+		/* We have gone too far already.  No point in continuing. */
+		break;
+	}
+	/* Attribute not found. */
+	ntfs_debug("Attribute type 0x%x not found in $AttrDef.",
+			le32_to_cpu(type));
+	return NULL;
+}
+
+/**
+ * ntfs_attr_size_bounds_check - check a size of an attribute type for validity
+ * @vol:	ntfs volume to which the attribute belongs
+ * @type:	attribute type which to check
+ * @size:	size which to check
+ *
+ * Check whether the @size in bytes is valid for an attribute of @type on the
+ * ntfs volume @vol.  This information is obtained from $AttrDef system file.
+ *
+ * Return 0 if valid, -ERANGE if not valid, or -ENOENT if the attribute is not
+ * listed in $AttrDef.
+ */
+int ntfs_attr_size_bounds_check(const ntfs_volume *vol, const ATTR_TYPE type,
+		const s64 size)
+{
+	ATTR_DEF *ad;
+
+	BUG_ON(size < 0);
+	/*
+	 * $ATTRIBUTE_LIST has a maximum size of 256kiB, but this is not
+	 * listed in $AttrDef.
+	 */
+	if (unlikely(type == AT_ATTRIBUTE_LIST && size > 256 * 1024))
+		return -ERANGE;
+	/* Get the $AttrDef entry for the attribute @type. */
+	ad = ntfs_attr_find_in_attrdef(vol, type);
+	if (unlikely(!ad))
+		return -ENOENT;
+	/* Do the bounds check. */
+	if (((sle64_to_cpu(ad->min_size) > 0) &&
+			size < sle64_to_cpu(ad->min_size)) ||
+			((sle64_to_cpu(ad->max_size) > 0) && size >
+			sle64_to_cpu(ad->max_size)))
+		return -ERANGE;
+	return 0;
+}
+
+/**
+ * ntfs_attr_can_be_non_resident - check if an attribute can be non-resident
+ * @vol:	ntfs volume to which the attribute belongs
+ * @type:	attribute type which to check
+ *
+ * Check whether the attribute of @type on the ntfs volume @vol is allowed to
+ * be non-resident.  This information is obtained from $AttrDef system file.
+ *
+ * Return 0 if the attribute is allowed to be non-resident, -EPERM if not, and
+ * -ENOENT if the attribute is not listed in $AttrDef.
+ */
+int ntfs_attr_can_be_non_resident(const ntfs_volume *vol, const ATTR_TYPE type)
+{
+	ATTR_DEF *ad;
+
+	/* Find the attribute definition record in $AttrDef. */
+	ad = ntfs_attr_find_in_attrdef(vol, type);
+	if (unlikely(!ad))
+		return -ENOENT;
+	/* Check the flags and return the result. */
+	if (ad->flags & ATTR_DEF_RESIDENT)
+		return -EPERM;
+	return 0;
+}
+
+/**
+ * ntfs_attr_can_be_resident - check if an attribute can be resident
+ * @vol:	ntfs volume to which the attribute belongs
+ * @type:	attribute type which to check
+ *
+ * Check whether the attribute of @type on the ntfs volume @vol is allowed to
+ * be resident.  This information is derived from our ntfs knowledge and may
+ * not be completely accurate, especially when user defined attributes are
+ * present.  Basically we allow everything to be resident except for index
+ * allocation and $EA attributes.
+ *
+ * Return 0 if the attribute is allowed to be non-resident and -EPERM if not.
+ *
+ * Warning: In the system file $MFT the attribute $Bitmap must be non-resident
+ *	    otherwise windows will not boot (blue screen of death)!  We cannot
+ *	    check for this here as we do not know which inode's $Bitmap is
+ *	    being asked about so the caller needs to special case this.
+ */
+int ntfs_attr_can_be_resident(const ntfs_volume *vol, const ATTR_TYPE type)
+{
+	if (type == AT_INDEX_ALLOCATION)
+		return -EPERM;
+	return 0;
+}
+
+/**
+ * ntfs_attr_record_resize - resize an attribute record
+ * @m:		mft record containing attribute record
+ * @a:		attribute record to resize
+ * @new_size:	new size in bytes to which to resize the attribute record @a
+ *
+ * Resize the attribute record @a, i.e. the resident part of the attribute, in
+ * the mft record @m to @new_size bytes.
+ *
+ * Return 0 on success and -errno on error.  The following error codes are
+ * defined:
+ *	-ENOSPC	- Not enough space in the mft record @m to perform the resize.
+ *
+ * Note: On error, no modifications have been performed whatsoever.
+ *
+ * Warning: If you make a record smaller without having copied all the data you
+ *	    are interested in the data may be overwritten.
+ */
+int ntfs_attr_record_resize(MFT_RECORD *m, ATTR_RECORD *a, u32 new_size)
+{
+	ntfs_debug("Entering for new_size %u.", new_size);
+	/* Align to 8 bytes if it is not already done. */
+	if (new_size & 7)
+		new_size = (new_size + 7) & ~7;
+	/* If the actual attribute length has changed, move things around. */
+	if (new_size != le32_to_cpu(a->length)) {
+		u32 new_muse = le32_to_cpu(m->bytes_in_use) -
+				le32_to_cpu(a->length) + new_size;
+		/* Not enough space in this mft record. */
+		if (new_muse > le32_to_cpu(m->bytes_allocated))
+			return -ENOSPC;
+		/* Move attributes following @a to their new location. */
+		memmove((u8*)a + new_size, (u8*)a + le32_to_cpu(a->length),
+				le32_to_cpu(m->bytes_in_use) - ((u8*)a -
+				(u8*)m) - le32_to_cpu(a->length));
+		/* Adjust @m to reflect the change in used space. */
+		m->bytes_in_use = cpu_to_le32(new_muse);
+		/* Adjust @a to reflect the new size. */
+		if (new_size >= offsetof(ATTR_REC, length) + sizeof(a->length))
+			a->length = cpu_to_le32(new_size);
+	}
+	return 0;
+}
+
+/**
+ * ntfs_resident_attr_value_resize - resize the value of a resident attribute
+ * @m:		mft record containing attribute record
+ * @a:		attribute record whose value to resize
+ * @new_size:	new size in bytes to which to resize the attribute value of @a
+ *
+ * Resize the value of the attribute @a in the mft record @m to @new_size bytes.
+ * If the value is made bigger, the newly allocated space is cleared.
+ *
+ * Return 0 on success and -errno on error.  The following error codes are
+ * defined:
+ *	-ENOSPC	- Not enough space in the mft record @m to perform the resize.
+ *
+ * Note: On error, no modifications have been performed whatsoever.
+ *
+ * Warning: If you make a record smaller without having copied all the data you
+ *	    are interested in the data may be overwritten.
+ */
+int ntfs_resident_attr_value_resize(MFT_RECORD *m, ATTR_RECORD *a,
+		const u32 new_size)
+{
+	u32 old_size;
+
+	/* Resize the resident part of the attribute record. */
+	if (ntfs_attr_record_resize(m, a,
+			le16_to_cpu(a->data.resident.value_offset) + new_size))
+		return -ENOSPC;
+	/*
+	 * The resize succeeded!  If we made the attribute value bigger, clear
+	 * the area between the old size and @new_size.
+	 */
+	old_size = le32_to_cpu(a->data.resident.value_length);
+	if (new_size > old_size)
+		memset((u8*)a + le16_to_cpu(a->data.resident.value_offset) +
+				old_size, 0, new_size - old_size);
+	/* Finally update the length of the attribute value. */
+	a->data.resident.value_length = cpu_to_le32(new_size);
+	return 0;
+}
+
+/**
+ * ntfs_attr_make_non_resident - convert a resident to a non-resident attribute
+ * @ni:		ntfs inode describing the attribute to convert
+ * @data_size:	size of the resident data to copy to the non-resident attribute
+ *
+ * Convert the resident ntfs attribute described by the ntfs inode @ni to a
+ * non-resident one.
+ *
+ * @data_size must be equal to the attribute value size.  This is needed since
+ * we need to know the size before we can map the mft record and our callers
+ * always know it.  The reason we cannot simply read the size from the vfs
+ * inode i_size is that this is not necessarily uptodate.  This happens when
+ * ntfs_attr_make_non_resident() is called in the ->truncate call path(s).
+ *
+ * Return 0 on success and -errno on error.  The following error return codes
+ * are defined:
+ *	-EPERM	- The attribute is not allowed to be non-resident.
+ *	-ENOMEM	- Not enough memory.
+ *	-ENOSPC	- Not enough disk space.
+ *	-EINVAL	- Attribute not defined on the volume.
+ *	-EIO	- I/o error or other error.
+ * Note that -ENOSPC is also returned in the case that there is not enough
+ * space in the mft record to do the conversion.  This can happen when the mft
+ * record is already very full.  The caller is responsible for trying to make
+ * space in the mft record and trying again.  FIXME: Do we need a separate
+ * error return code for this kind of -ENOSPC or is it always worth trying
+ * again in case the attribute may then fit in a resident state so no need to
+ * make it non-resident at all?  Ho-hum...  (AIA)
+ *
+ * NOTE to self: No changes in the attribute list are required to move from
+ *		 a resident to a non-resident attribute.
+ *
+ * Locking: - The caller must hold i_mutex on the inode.
+ */
+int ntfs_attr_make_non_resident(ntfs_inode *ni, const u32 data_size)
+{
+	s64 new_size;
+	struct inode *vi = VFS_I(ni);
+	ntfs_volume *vol = ni->vol;
+	ntfs_inode *base_ni;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	ntfs_attr_search_ctx *ctx;
+	struct page *page;
+	runlist_element *rl;
+	u8 *kaddr;
+	unsigned long flags;
+	int mp_size, mp_ofs, name_ofs, arec_size, err, err2;
+	u32 attr_size;
+	u8 old_res_attr_flags;
+
+	/* Check that the attribute is allowed to be non-resident. */
+	err = ntfs_attr_can_be_non_resident(vol, ni->type);
+	if (unlikely(err)) {
+		if (err == -EPERM)
+			ntfs_debug("Attribute is not allowed to be "
+					"non-resident.");
+		else
+			ntfs_debug("Attribute not defined on the NTFS "
+					"volume!");
+		return err;
+	}
+	/*
+	 * FIXME: Compressed and encrypted attributes are not supported when
+	 * writing and we should never have gotten here for them.
+	 */
+	BUG_ON(NInoCompressed(ni));
+	BUG_ON(NInoEncrypted(ni));
+	/*
+	 * The size needs to be aligned to a cluster boundary for allocation
+	 * purposes.
+	 */
+	new_size = (data_size + vol->cluster_size - 1) &
+			~(vol->cluster_size - 1);
+	if (new_size > 0) {
+		/*
+		 * Will need the page later and since the page lock nests
+		 * outside all ntfs locks, we need to get the page now.
+		 */
+		page = find_or_create_page(vi->i_mapping, 0,
+				mapping_gfp_mask(vi->i_mapping));
+		if (unlikely(!page))
+			return -ENOMEM;
+		/* Start by allocating clusters to hold the attribute value. */
+		rl = ntfs_cluster_alloc(vol, 0, new_size >>
+				vol->cluster_size_bits, -1, DATA_ZONE, true);
+		if (IS_ERR(rl)) {
+			err = PTR_ERR(rl);
+			ntfs_debug("Failed to allocate cluster%s, error code "
+					"%i.", (new_size >>
+					vol->cluster_size_bits) > 1 ? "s" : "",
+					err);
+			goto page_err_out;
+		}
+	} else {
+		rl = NULL;
+		page = NULL;
+	}
+	/* Determine the size of the mapping pairs array. */
+	mp_size = ntfs_get_size_for_mapping_pairs(vol, rl, 0, -1);
+	if (unlikely(mp_size < 0)) {
+		err = mp_size;
+		ntfs_debug("Failed to get size for mapping pairs array, error "
+				"code %i.", err);
+		goto rl_err_out;
+	}
+	down_write(&ni->runlist.lock);
+	if (!NInoAttr(ni))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		m = NULL;
+		ctx = NULL;
+		goto err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT)
+			err = -EIO;
+		goto err_out;
+	}
+	m = ctx->mrec;
+	a = ctx->attr;
+	BUG_ON(NInoNonResident(ni));
+	BUG_ON(a->non_resident);
+	/*
+	 * Calculate new offsets for the name and the mapping pairs array.
+	 */
+	if (NInoSparse(ni) || NInoCompressed(ni))
+		name_ofs = (offsetof(ATTR_REC,
+				data.non_resident.compressed_size) +
+				sizeof(a->data.non_resident.compressed_size) +
+				7) & ~7;
+	else
+		name_ofs = (offsetof(ATTR_REC,
+				data.non_resident.compressed_size) + 7) & ~7;
+	mp_ofs = (name_ofs + a->name_length * sizeof(ntfschar) + 7) & ~7;
+	/*
+	 * Determine the size of the resident part of the now non-resident
+	 * attribute record.
+	 */
+	arec_size = (mp_ofs + mp_size + 7) & ~7;
+	/*
+	 * If the page is not uptodate bring it uptodate by copying from the
+	 * attribute value.
+	 */
+	attr_size = le32_to_cpu(a->data.resident.value_length);
+	BUG_ON(attr_size != data_size);
+	if (page && !PageUptodate(page)) {
+		kaddr = kmap_atomic(page);
+		memcpy(kaddr, (u8*)a +
+				le16_to_cpu(a->data.resident.value_offset),
+				attr_size);
+		memset(kaddr + attr_size, 0, PAGE_SIZE - attr_size);
+		kunmap_atomic(kaddr);
+		flush_dcache_page(page);
+		SetPageUptodate(page);
+	}
+	/* Backup the attribute flag. */
+	old_res_attr_flags = a->data.resident.flags;
+	/* Resize the resident part of the attribute record. */
+	err = ntfs_attr_record_resize(m, a, arec_size);
+	if (unlikely(err))
+		goto err_out;
+	/*
+	 * Convert the resident part of the attribute record to describe a
+	 * non-resident attribute.
+	 */
+	a->non_resident = 1;
+	/* Move the attribute name if it exists and update the offset. */
+	if (a->name_length)
+		memmove((u8*)a + name_ofs, (u8*)a + le16_to_cpu(a->name_offset),
+				a->name_length * sizeof(ntfschar));
+	a->name_offset = cpu_to_le16(name_ofs);
+	/* Setup the fields specific to non-resident attributes. */
+	a->data.non_resident.lowest_vcn = 0;
+	a->data.non_resident.highest_vcn = cpu_to_sle64((new_size - 1) >>
+			vol->cluster_size_bits);
+	a->data.non_resident.mapping_pairs_offset = cpu_to_le16(mp_ofs);
+	memset(&a->data.non_resident.reserved, 0,
+			sizeof(a->data.non_resident.reserved));
+	a->data.non_resident.allocated_size = cpu_to_sle64(new_size);
+	a->data.non_resident.data_size =
+			a->data.non_resident.initialized_size =
+			cpu_to_sle64(attr_size);
+	if (NInoSparse(ni) || NInoCompressed(ni)) {
+		a->data.non_resident.compression_unit = 0;
+		if (NInoCompressed(ni) || vol->major_ver < 3)
+			a->data.non_resident.compression_unit = 4;
+		a->data.non_resident.compressed_size =
+				a->data.non_resident.allocated_size;
+	} else
+		a->data.non_resident.compression_unit = 0;
+	/* Generate the mapping pairs array into the attribute record. */
+	err = ntfs_mapping_pairs_build(vol, (u8*)a + mp_ofs,
+			arec_size - mp_ofs, rl, 0, -1, NULL);
+	if (unlikely(err)) {
+		ntfs_debug("Failed to build mapping pairs, error code %i.",
+				err);
+		goto undo_err_out;
+	}
+	/* Setup the in-memory attribute structure to be non-resident. */
+	ni->runlist.rl = rl;
+	write_lock_irqsave(&ni->size_lock, flags);
+	ni->allocated_size = new_size;
+	if (NInoSparse(ni) || NInoCompressed(ni)) {
+		ni->itype.compressed.size = ni->allocated_size;
+		if (a->data.non_resident.compression_unit) {
+			ni->itype.compressed.block_size = 1U << (a->data.
+					non_resident.compression_unit +
+					vol->cluster_size_bits);
+			ni->itype.compressed.block_size_bits =
+					ffs(ni->itype.compressed.block_size) -
+					1;
+			ni->itype.compressed.block_clusters = 1U <<
+					a->data.non_resident.compression_unit;
+		} else {
+			ni->itype.compressed.block_size = 0;
+			ni->itype.compressed.block_size_bits = 0;
+			ni->itype.compressed.block_clusters = 0;
+		}
+		vi->i_blocks = ni->itype.compressed.size >> 9;
+	} else
+		vi->i_blocks = ni->allocated_size >> 9;
+	write_unlock_irqrestore(&ni->size_lock, flags);
+	/*
+	 * This needs to be last since the address space operations ->readpage
+	 * and ->writepage can run concurrently with us as they are not
+	 * serialized on i_mutex.  Note, we are not allowed to fail once we flip
+	 * this switch, which is another reason to do this last.
+	 */
+	NInoSetNonResident(ni);
+	/* Mark the mft record dirty, so it gets written back. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+	up_write(&ni->runlist.lock);
+	if (page) {
+		set_page_dirty(page);
+		unlock_page(page);
+		put_page(page);
+	}
+	ntfs_debug("Done.");
+	return 0;
+undo_err_out:
+	/* Convert the attribute back into a resident attribute. */
+	a->non_resident = 0;
+	/* Move the attribute name if it exists and update the offset. */
+	name_ofs = (offsetof(ATTR_RECORD, data.resident.reserved) +
+			sizeof(a->data.resident.reserved) + 7) & ~7;
+	if (a->name_length)
+		memmove((u8*)a + name_ofs, (u8*)a + le16_to_cpu(a->name_offset),
+				a->name_length * sizeof(ntfschar));
+	mp_ofs = (name_ofs + a->name_length * sizeof(ntfschar) + 7) & ~7;
+	a->name_offset = cpu_to_le16(name_ofs);
+	arec_size = (mp_ofs + attr_size + 7) & ~7;
+	/* Resize the resident part of the attribute record. */
+	err2 = ntfs_attr_record_resize(m, a, arec_size);
+	if (unlikely(err2)) {
+		/*
+		 * This cannot happen (well if memory corruption is at work it
+		 * could happen in theory), but deal with it as well as we can.
+		 * If the old size is too small, truncate the attribute,
+		 * otherwise simply give it a larger allocated size.
+		 * FIXME: Should check whether chkdsk complains when the
+		 * allocated size is much bigger than the resident value size.
+		 */
+		arec_size = le32_to_cpu(a->length);
+		if ((mp_ofs + attr_size) > arec_size) {
+			err2 = attr_size;
+			attr_size = arec_size - mp_ofs;
+			ntfs_error(vol->sb, "Failed to undo partial resident "
+					"to non-resident attribute "
+					"conversion.  Truncating inode 0x%lx, "
+					"attribute type 0x%x from %i bytes to "
+					"%i bytes to maintain metadata "
+					"consistency.  THIS MEANS YOU ARE "
+					"LOSING %i BYTES DATA FROM THIS %s.",
+					vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type),
+					err2, attr_size, err2 - attr_size,
+					((ni->type == AT_DATA) &&
+					!ni->name_len) ? "FILE": "ATTRIBUTE");
+			write_lock_irqsave(&ni->size_lock, flags);
+			ni->initialized_size = attr_size;
+			i_size_write(vi, attr_size);
+			write_unlock_irqrestore(&ni->size_lock, flags);
+		}
+	}
+	/* Setup the fields specific to resident attributes. */
+	a->data.resident.value_length = cpu_to_le32(attr_size);
+	a->data.resident.value_offset = cpu_to_le16(mp_ofs);
+	a->data.resident.flags = old_res_attr_flags;
+	memset(&a->data.resident.reserved, 0,
+			sizeof(a->data.resident.reserved));
+	/* Copy the data from the page back to the attribute value. */
+	if (page) {
+		kaddr = kmap_atomic(page);
+		memcpy((u8*)a + mp_ofs, kaddr, attr_size);
+		kunmap_atomic(kaddr);
+	}
+	/* Setup the allocated size in the ntfs inode in case it changed. */
+	write_lock_irqsave(&ni->size_lock, flags);
+	ni->allocated_size = arec_size - mp_ofs;
+	write_unlock_irqrestore(&ni->size_lock, flags);
+	/* Mark the mft record dirty, so it gets written back. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+err_out:
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(base_ni);
+	ni->runlist.rl = NULL;
+	up_write(&ni->runlist.lock);
+rl_err_out:
+	if (rl) {
+		if (ntfs_cluster_free_from_rl(vol, rl) < 0) {
+			ntfs_error(vol->sb, "Failed to release allocated "
+					"cluster(s) in error code path.  Run "
+					"chkdsk to recover the lost "
+					"cluster(s).");
+			NVolSetErrors(vol);
+		}
+		ntfs_free(rl);
+page_err_out:
+		unlock_page(page);
+		put_page(page);
+	}
+	if (err == -EINVAL)
+		err = -EIO;
+	return err;
+}
+
+/**
+ * ntfs_attr_extend_allocation - extend the allocated space of an attribute
+ * @ni:			ntfs inode of the attribute whose allocation to extend
+ * @new_alloc_size:	new size in bytes to which to extend the allocation to
+ * @new_data_size:	new size in bytes to which to extend the data to
+ * @data_start:		beginning of region which is required to be non-sparse
+ *
+ * Extend the allocated space of an attribute described by the ntfs inode @ni
+ * to @new_alloc_size bytes.  If @data_start is -1, the whole extension may be
+ * implemented as a hole in the file (as long as both the volume and the ntfs
+ * inode @ni have sparse support enabled).  If @data_start is >= 0, then the
+ * region between the old allocated size and @data_start - 1 may be made sparse
+ * but the regions between @data_start and @new_alloc_size must be backed by
+ * actual clusters.
+ *
+ * If @new_data_size is -1, it is ignored.  If it is >= 0, then the data size
+ * of the attribute is extended to @new_data_size.  Note that the i_size of the
+ * vfs inode is not updated.  Only the data size in the base attribute record
+ * is updated.  The caller has to update i_size separately if this is required.
+ * WARNING: It is a BUG() for @new_data_size to be smaller than the old data
+ * size as well as for @new_data_size to be greater than @new_alloc_size.
+ *
+ * For resident attributes this involves resizing the attribute record and if
+ * necessary moving it and/or other attributes into extent mft records and/or
+ * converting the attribute to a non-resident attribute which in turn involves
+ * extending the allocation of a non-resident attribute as described below.
+ *
+ * For non-resident attributes this involves allocating clusters in the data
+ * zone on the volume (except for regions that are being made sparse) and
+ * extending the run list to describe the allocated clusters as well as
+ * updating the mapping pairs array of the attribute.  This in turn involves
+ * resizing the attribute record and if necessary moving it and/or other
+ * attributes into extent mft records and/or splitting the attribute record
+ * into multiple extent attribute records.
+ *
+ * Also, the attribute list attribute is updated if present and in some of the
+ * above cases (the ones where extent mft records/attributes come into play),
+ * an attribute list attribute is created if not already present.
+ *
+ * Return the new allocated size on success and -errno on error.  In the case
+ * that an error is encountered but a partial extension at least up to
+ * @data_start (if present) is possible, the allocation is partially extended
+ * and this is returned.  This means the caller must check the returned size to
+ * determine if the extension was partial.  If @data_start is -1 then partial
+ * allocations are not performed.
+ *
+ * WARNING: Do not call ntfs_attr_extend_allocation() for $MFT/$DATA.
+ *
+ * Locking: This function takes the runlist lock of @ni for writing as well as
+ * locking the mft record of the base ntfs inode.  These locks are maintained
+ * throughout execution of the function.  These locks are required so that the
+ * attribute can be resized safely and so that it can for example be converted
+ * from resident to non-resident safely.
+ *
+ * TODO: At present attribute list attribute handling is not implemented.
+ *
+ * TODO: At present it is not safe to call this function for anything other
+ * than the $DATA attribute(s) of an uncompressed and unencrypted file.
+ */
+s64 ntfs_attr_extend_allocation(ntfs_inode *ni, s64 new_alloc_size,
+		const s64 new_data_size, const s64 data_start)
+{
+	VCN vcn;
+	s64 ll, allocated_size, start = data_start;
+	struct inode *vi = VFS_I(ni);
+	ntfs_volume *vol = ni->vol;
+	ntfs_inode *base_ni;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	ntfs_attr_search_ctx *ctx;
+	runlist_element *rl, *rl2;
+	unsigned long flags;
+	int err, mp_size;
+	u32 attr_len = 0; /* Silence stupid gcc warning. */
+	bool mp_rebuilt;
+
+#ifdef DEBUG
+	read_lock_irqsave(&ni->size_lock, flags);
+	allocated_size = ni->allocated_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	ntfs_debug("Entering for i_ino 0x%lx, attribute type 0x%x, "
+			"old_allocated_size 0x%llx, "
+			"new_allocated_size 0x%llx, new_data_size 0x%llx, "
+			"data_start 0x%llx.", vi->i_ino,
+			(unsigned)le32_to_cpu(ni->type),
+			(unsigned long long)allocated_size,
+			(unsigned long long)new_alloc_size,
+			(unsigned long long)new_data_size,
+			(unsigned long long)start);
+#endif
+retry_extend:
+	/*
+	 * For non-resident attributes, @start and @new_size need to be aligned
+	 * to cluster boundaries for allocation purposes.
+	 */
+	if (NInoNonResident(ni)) {
+		if (start > 0)
+			start &= ~(s64)vol->cluster_size_mask;
+		new_alloc_size = (new_alloc_size + vol->cluster_size - 1) &
+				~(s64)vol->cluster_size_mask;
+	}
+	BUG_ON(new_data_size >= 0 && new_data_size > new_alloc_size);
+	/* Check if new size is allowed in $AttrDef. */
+	err = ntfs_attr_size_bounds_check(vol, ni->type, new_alloc_size);
+	if (unlikely(err)) {
+		/* Only emit errors when the write will fail completely. */
+		read_lock_irqsave(&ni->size_lock, flags);
+		allocated_size = ni->allocated_size;
+		read_unlock_irqrestore(&ni->size_lock, flags);
+		if (start < 0 || start >= allocated_size) {
+			if (err == -ERANGE) {
+				ntfs_error(vol->sb, "Cannot extend allocation "
+						"of inode 0x%lx, attribute "
+						"type 0x%x, because the new "
+						"allocation would exceed the "
+						"maximum allowed size for "
+						"this attribute type.",
+						vi->i_ino, (unsigned)
+						le32_to_cpu(ni->type));
+			} else {
+				ntfs_error(vol->sb, "Cannot extend allocation "
+						"of inode 0x%lx, attribute "
+						"type 0x%x, because this "
+						"attribute type is not "
+						"defined on the NTFS volume.  "
+						"Possible corruption!  You "
+						"should run chkdsk!",
+						vi->i_ino, (unsigned)
+						le32_to_cpu(ni->type));
+			}
+		}
+		/* Translate error code to be POSIX conformant for write(2). */
+		if (err == -ERANGE)
+			err = -EFBIG;
+		else
+			err = -EIO;
+		return err;
+	}
+	if (!NInoAttr(ni))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	/*
+	 * We will be modifying both the runlist (if non-resident) and the mft
+	 * record so lock them both down.
+	 */
+	down_write(&ni->runlist.lock);
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		m = NULL;
+		ctx = NULL;
+		goto err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	read_lock_irqsave(&ni->size_lock, flags);
+	allocated_size = ni->allocated_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	/*
+	 * If non-resident, seek to the last extent.  If resident, there is
+	 * only one extent, so seek to that.
+	 */
+	vcn = NInoNonResident(ni) ? allocated_size >> vol->cluster_size_bits :
+			0;
+	/*
+	 * Abort if someone did the work whilst we waited for the locks.  If we
+	 * just converted the attribute from resident to non-resident it is
+	 * likely that exactly this has happened already.  We cannot quite
+	 * abort if we need to update the data size.
+	 */
+	if (unlikely(new_alloc_size <= allocated_size)) {
+		ntfs_debug("Allocated size already exceeds requested size.");
+		new_alloc_size = allocated_size;
+		if (new_data_size < 0)
+			goto done;
+		/*
+		 * We want the first attribute extent so that we can update the
+		 * data size.
+		 */
+		vcn = 0;
+	}
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, vcn, NULL, 0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT)
+			err = -EIO;
+		goto err_out;
+	}
+	m = ctx->mrec;
+	a = ctx->attr;
+	/* Use goto to reduce indentation. */
+	if (a->non_resident)
+		goto do_non_resident_extend;
+	BUG_ON(NInoNonResident(ni));
+	/* The total length of the attribute value. */
+	attr_len = le32_to_cpu(a->data.resident.value_length);
+	/*
+	 * Extend the attribute record to be able to store the new attribute
+	 * size.  ntfs_attr_record_resize() will not do anything if the size is
+	 * not changing.
+	 */
+	if (new_alloc_size < vol->mft_record_size &&
+			!ntfs_attr_record_resize(m, a,
+			le16_to_cpu(a->data.resident.value_offset) +
+			new_alloc_size)) {
+		/* The resize succeeded! */
+		write_lock_irqsave(&ni->size_lock, flags);
+		ni->allocated_size = le32_to_cpu(a->length) -
+				le16_to_cpu(a->data.resident.value_offset);
+		write_unlock_irqrestore(&ni->size_lock, flags);
+		if (new_data_size >= 0) {
+			BUG_ON(new_data_size < attr_len);
+			a->data.resident.value_length =
+					cpu_to_le32((u32)new_data_size);
+		}
+		goto flush_done;
+	}
+	/*
+	 * We have to drop all the locks so we can call
+	 * ntfs_attr_make_non_resident().  This could be optimised by try-
+	 * locking the first page cache page and only if that fails dropping
+	 * the locks, locking the page, and redoing all the locking and
+	 * lookups.  While this would be a huge optimisation, it is not worth
+	 * it as this is definitely a slow code path.
+	 */
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+	up_write(&ni->runlist.lock);
+	/*
+	 * Not enough space in the mft record, try to make the attribute
+	 * non-resident and if successful restart the extension process.
+	 */
+	err = ntfs_attr_make_non_resident(ni, attr_len);
+	if (likely(!err))
+		goto retry_extend;
+	/*
+	 * Could not make non-resident.  If this is due to this not being
+	 * permitted for this attribute type or there not being enough space,
+	 * try to make other attributes non-resident.  Otherwise fail.
+	 */
+	if (unlikely(err != -EPERM && err != -ENOSPC)) {
+		/* Only emit errors when the write will fail completely. */
+		read_lock_irqsave(&ni->size_lock, flags);
+		allocated_size = ni->allocated_size;
+		read_unlock_irqrestore(&ni->size_lock, flags);
+		if (start < 0 || start >= allocated_size)
+			ntfs_error(vol->sb, "Cannot extend allocation of "
+					"inode 0x%lx, attribute type 0x%x, "
+					"because the conversion from resident "
+					"to non-resident attribute failed "
+					"with error code %i.", vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type), err);
+		if (err != -ENOMEM)
+			err = -EIO;
+		goto conv_err_out;
+	}
+	/* TODO: Not implemented from here, abort. */
+	read_lock_irqsave(&ni->size_lock, flags);
+	allocated_size = ni->allocated_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	if (start < 0 || start >= allocated_size) {
+		if (err == -ENOSPC)
+			ntfs_error(vol->sb, "Not enough space in the mft "
+					"record/on disk for the non-resident "
+					"attribute value.  This case is not "
+					"implemented yet.");
+		else /* if (err == -EPERM) */
+			ntfs_error(vol->sb, "This attribute type may not be "
+					"non-resident.  This case is not "
+					"implemented yet.");
+	}
+	err = -EOPNOTSUPP;
+	goto conv_err_out;
+#if 0
+	// TODO: Attempt to make other attributes non-resident.
+	if (!err)
+		goto do_resident_extend;
+	/*
+	 * Both the attribute list attribute and the standard information
+	 * attribute must remain in the base inode.  Thus, if this is one of
+	 * these attributes, we have to try to move other attributes out into
+	 * extent mft records instead.
+	 */
+	if (ni->type == AT_ATTRIBUTE_LIST ||
+			ni->type == AT_STANDARD_INFORMATION) {
+		// TODO: Attempt to move other attributes into extent mft
+		// records.
+		err = -EOPNOTSUPP;
+		if (!err)
+			goto do_resident_extend;
+		goto err_out;
+	}
+	// TODO: Attempt to move this attribute to an extent mft record, but
+	// only if it is not already the only attribute in an mft record in
+	// which case there would be nothing to gain.
+	err = -EOPNOTSUPP;
+	if (!err)
+		goto do_resident_extend;
+	/* There is nothing we can do to make enough space. )-: */
+	goto err_out;
+#endif
+do_non_resident_extend:
+	BUG_ON(!NInoNonResident(ni));
+	if (new_alloc_size == allocated_size) {
+		BUG_ON(vcn);
+		goto alloc_done;
+	}
+	/*
+	 * If the data starts after the end of the old allocation, this is a
+	 * $DATA attribute and sparse attributes are enabled on the volume and
+	 * for this inode, then create a sparse region between the old
+	 * allocated size and the start of the data.  Otherwise simply proceed
+	 * with filling the whole space between the old allocated size and the
+	 * new allocated size with clusters.
+	 */
+	if ((start >= 0 && start <= allocated_size) || ni->type != AT_DATA ||
+			!NVolSparseEnabled(vol) || NInoSparseDisabled(ni))
+		goto skip_sparse;
+	// TODO: This is not implemented yet.  We just fill in with real
+	// clusters for now...
+	ntfs_debug("Inserting holes is not-implemented yet.  Falling back to "
+			"allocating real clusters instead.");
+skip_sparse:
+	rl = ni->runlist.rl;
+	if (likely(rl)) {
+		/* Seek to the end of the runlist. */
+		while (rl->length)
+			rl++;
+	}
+	/* If this attribute extent is not mapped, map it now. */
+	if (unlikely(!rl || rl->lcn == LCN_RL_NOT_MAPPED ||
+			(rl->lcn == LCN_ENOENT && rl > ni->runlist.rl &&
+			(rl-1)->lcn == LCN_RL_NOT_MAPPED))) {
+		if (!rl && !allocated_size)
+			goto first_alloc;
+		rl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
+		if (IS_ERR(rl)) {
+			err = PTR_ERR(rl);
+			if (start < 0 || start >= allocated_size)
+				ntfs_error(vol->sb, "Cannot extend allocation "
+						"of inode 0x%lx, attribute "
+						"type 0x%x, because the "
+						"mapping of a runlist "
+						"fragment failed with error "
+						"code %i.", vi->i_ino,
+						(unsigned)le32_to_cpu(ni->type),
+						err);
+			if (err != -ENOMEM)
+				err = -EIO;
+			goto err_out;
+		}
+		ni->runlist.rl = rl;
+		/* Seek to the end of the runlist. */
+		while (rl->length)
+			rl++;
+	}
+	/*
+	 * We now know the runlist of the last extent is mapped and @rl is at
+	 * the end of the runlist.  We want to begin allocating clusters
+	 * starting at the last allocated cluster to reduce fragmentation.  If
+	 * there are no valid LCNs in the attribute we let the cluster
+	 * allocator choose the starting cluster.
+	 */
+	/* If the last LCN is a hole or simillar seek back to last real LCN. */
+	while (rl->lcn < 0 && rl > ni->runlist.rl)
+		rl--;
+first_alloc:
+	// FIXME: Need to implement partial allocations so at least part of the
+	// write can be performed when start >= 0.  (Needed for POSIX write(2)
+	// conformance.)
+	rl2 = ntfs_cluster_alloc(vol, allocated_size >> vol->cluster_size_bits,
+			(new_alloc_size - allocated_size) >>
+			vol->cluster_size_bits, (rl && (rl->lcn >= 0)) ?
+			rl->lcn + rl->length : -1, DATA_ZONE, true);
+	if (IS_ERR(rl2)) {
+		err = PTR_ERR(rl2);
+		if (start < 0 || start >= allocated_size)
+			ntfs_error(vol->sb, "Cannot extend allocation of "
+					"inode 0x%lx, attribute type 0x%x, "
+					"because the allocation of clusters "
+					"failed with error code %i.", vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type), err);
+		if (err != -ENOMEM && err != -ENOSPC)
+			err = -EIO;
+		goto err_out;
+	}
+	rl = ntfs_runlists_merge(ni->runlist.rl, rl2);
+	if (IS_ERR(rl)) {
+		err = PTR_ERR(rl);
+		if (start < 0 || start >= allocated_size)
+			ntfs_error(vol->sb, "Cannot extend allocation of "
+					"inode 0x%lx, attribute type 0x%x, "
+					"because the runlist merge failed "
+					"with error code %i.", vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type), err);
+		if (err != -ENOMEM)
+			err = -EIO;
+		if (ntfs_cluster_free_from_rl(vol, rl2)) {
+			ntfs_error(vol->sb, "Failed to release allocated "
+					"cluster(s) in error code path.  Run "
+					"chkdsk to recover the lost "
+					"cluster(s).");
+			NVolSetErrors(vol);
+		}
+		ntfs_free(rl2);
+		goto err_out;
+	}
+	ni->runlist.rl = rl;
+	ntfs_debug("Allocated 0x%llx clusters.", (long long)(new_alloc_size -
+			allocated_size) >> vol->cluster_size_bits);
+	/* Find the runlist element with which the attribute extent starts. */
+	ll = sle64_to_cpu(a->data.non_resident.lowest_vcn);
+	rl2 = ntfs_rl_find_vcn_nolock(rl, ll);
+	BUG_ON(!rl2);
+	BUG_ON(!rl2->length);
+	BUG_ON(rl2->lcn < LCN_HOLE);
+	mp_rebuilt = false;
+	/* Get the size for the new mapping pairs array for this extent. */
+	mp_size = ntfs_get_size_for_mapping_pairs(vol, rl2, ll, -1);
+	if (unlikely(mp_size <= 0)) {
+		err = mp_size;
+		if (start < 0 || start >= allocated_size)
+			ntfs_error(vol->sb, "Cannot extend allocation of "
+					"inode 0x%lx, attribute type 0x%x, "
+					"because determining the size for the "
+					"mapping pairs failed with error code "
+					"%i.", vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type), err);
+		err = -EIO;
+		goto undo_alloc;
+	}
+	/* Extend the attribute record to fit the bigger mapping pairs array. */
+	attr_len = le32_to_cpu(a->length);
+	err = ntfs_attr_record_resize(m, a, mp_size +
+			le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
+	if (unlikely(err)) {
+		BUG_ON(err != -ENOSPC);
+		// TODO: Deal with this by moving this extent to a new mft
+		// record or by starting a new extent in a new mft record,
+		// possibly by extending this extent partially and filling it
+		// and creating a new extent for the remainder, or by making
+		// other attributes non-resident and/or by moving other
+		// attributes out of this mft record.
+		if (start < 0 || start >= allocated_size)
+			ntfs_error(vol->sb, "Not enough space in the mft "
+					"record for the extended attribute "
+					"record.  This case is not "
+					"implemented yet.");
+		err = -EOPNOTSUPP;
+		goto undo_alloc;
+	}
+	mp_rebuilt = true;
+	/* Generate the mapping pairs array directly into the attr record. */
+	err = ntfs_mapping_pairs_build(vol, (u8*)a +
+			le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
+			mp_size, rl2, ll, -1, NULL);
+	if (unlikely(err)) {
+		if (start < 0 || start >= allocated_size)
+			ntfs_error(vol->sb, "Cannot extend allocation of "
+					"inode 0x%lx, attribute type 0x%x, "
+					"because building the mapping pairs "
+					"failed with error code %i.", vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type), err);
+		err = -EIO;
+		goto undo_alloc;
+	}
+	/* Update the highest_vcn. */
+	a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
+			vol->cluster_size_bits) - 1);
+	/*
+	 * We now have extended the allocated size of the attribute.  Reflect
+	 * this in the ntfs_inode structure and the attribute record.
+	 */
+	if (a->data.non_resident.lowest_vcn) {
+		/*
+		 * We are not in the first attribute extent, switch to it, but
+		 * first ensure the changes will make it to disk later.
+		 */
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		mark_mft_record_dirty(ctx->ntfs_ino);
+		ntfs_attr_reinit_search_ctx(ctx);
+		err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+				CASE_SENSITIVE, 0, NULL, 0, ctx);
+		if (unlikely(err))
+			goto restore_undo_alloc;
+		/* @m is not used any more so no need to set it. */
+		a = ctx->attr;
+	}
+	write_lock_irqsave(&ni->size_lock, flags);
+	ni->allocated_size = new_alloc_size;
+	a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
+	/*
+	 * FIXME: This would fail if @ni is a directory, $MFT, or an index,
+	 * since those can have sparse/compressed set.  For example can be
+	 * set compressed even though it is not compressed itself and in that
+	 * case the bit means that files are to be created compressed in the
+	 * directory...  At present this is ok as this code is only called for
+	 * regular files, and only for their $DATA attribute(s).
+	 * FIXME: The calculation is wrong if we created a hole above.  For now
+	 * it does not matter as we never create holes.
+	 */
+	if (NInoSparse(ni) || NInoCompressed(ni)) {
+		ni->itype.compressed.size += new_alloc_size - allocated_size;
+		a->data.non_resident.compressed_size =
+				cpu_to_sle64(ni->itype.compressed.size);
+		vi->i_blocks = ni->itype.compressed.size >> 9;
+	} else
+		vi->i_blocks = new_alloc_size >> 9;
+	write_unlock_irqrestore(&ni->size_lock, flags);
+alloc_done:
+	if (new_data_size >= 0) {
+		BUG_ON(new_data_size <
+				sle64_to_cpu(a->data.non_resident.data_size));
+		a->data.non_resident.data_size = cpu_to_sle64(new_data_size);
+	}
+flush_done:
+	/* Ensure the changes make it to disk. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+done:
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+	up_write(&ni->runlist.lock);
+	ntfs_debug("Done, new_allocated_size 0x%llx.",
+			(unsigned long long)new_alloc_size);
+	return new_alloc_size;
+restore_undo_alloc:
+	if (start < 0 || start >= allocated_size)
+		ntfs_error(vol->sb, "Cannot complete extension of allocation "
+				"of inode 0x%lx, attribute type 0x%x, because "
+				"lookup of first attribute extent failed with "
+				"error code %i.", vi->i_ino,
+				(unsigned)le32_to_cpu(ni->type), err);
+	if (err == -ENOENT)
+		err = -EIO;
+	ntfs_attr_reinit_search_ctx(ctx);
+	if (ntfs_attr_lookup(ni->type, ni->name, ni->name_len, CASE_SENSITIVE,
+			allocated_size >> vol->cluster_size_bits, NULL, 0,
+			ctx)) {
+		ntfs_error(vol->sb, "Failed to find last attribute extent of "
+				"attribute in error code path.  Run chkdsk to "
+				"recover.");
+		write_lock_irqsave(&ni->size_lock, flags);
+		ni->allocated_size = new_alloc_size;
+		/*
+		 * FIXME: This would fail if @ni is a directory...  See above.
+		 * FIXME: The calculation is wrong if we created a hole above.
+		 * For now it does not matter as we never create holes.
+		 */
+		if (NInoSparse(ni) || NInoCompressed(ni)) {
+			ni->itype.compressed.size += new_alloc_size -
+					allocated_size;
+			vi->i_blocks = ni->itype.compressed.size >> 9;
+		} else
+			vi->i_blocks = new_alloc_size >> 9;
+		write_unlock_irqrestore(&ni->size_lock, flags);
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(base_ni);
+		up_write(&ni->runlist.lock);
+		/*
+		 * The only thing that is now wrong is the allocated size of the
+		 * base attribute extent which chkdsk should be able to fix.
+		 */
+		NVolSetErrors(vol);
+		return err;
+	}
+	ctx->attr->data.non_resident.highest_vcn = cpu_to_sle64(
+			(allocated_size >> vol->cluster_size_bits) - 1);
+undo_alloc:
+	ll = allocated_size >> vol->cluster_size_bits;
+	if (ntfs_cluster_free(ni, ll, -1, ctx) < 0) {
+		ntfs_error(vol->sb, "Failed to release allocated cluster(s) "
+				"in error code path.  Run chkdsk to recover "
+				"the lost cluster(s).");
+		NVolSetErrors(vol);
+	}
+	m = ctx->mrec;
+	a = ctx->attr;
+	/*
+	 * If the runlist truncation fails and/or the search context is no
+	 * longer valid, we cannot resize the attribute record or build the
+	 * mapping pairs array thus we mark the inode bad so that no access to
+	 * the freed clusters can happen.
+	 */
+	if (ntfs_rl_truncate_nolock(vol, &ni->runlist, ll) || IS_ERR(m)) {
+		ntfs_error(vol->sb, "Failed to %s in error code path.  Run "
+				"chkdsk to recover.", IS_ERR(m) ?
+				"restore attribute search context" :
+				"truncate attribute runlist");
+		NVolSetErrors(vol);
+	} else if (mp_rebuilt) {
+		if (ntfs_attr_record_resize(m, a, attr_len)) {
+			ntfs_error(vol->sb, "Failed to restore attribute "
+					"record in error code path.  Run "
+					"chkdsk to recover.");
+			NVolSetErrors(vol);
+		} else /* if (success) */ {
+			if (ntfs_mapping_pairs_build(vol, (u8*)a + le16_to_cpu(
+					a->data.non_resident.
+					mapping_pairs_offset), attr_len -
+					le16_to_cpu(a->data.non_resident.
+					mapping_pairs_offset), rl2, ll, -1,
+					NULL)) {
+				ntfs_error(vol->sb, "Failed to restore "
+						"mapping pairs array in error "
+						"code path.  Run chkdsk to "
+						"recover.");
+				NVolSetErrors(vol);
+			}
+			flush_dcache_mft_record_page(ctx->ntfs_ino);
+			mark_mft_record_dirty(ctx->ntfs_ino);
+		}
+	}
+err_out:
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(base_ni);
+	up_write(&ni->runlist.lock);
+conv_err_out:
+	ntfs_debug("Failed.  Returning error code %i.", err);
+	return err;
+}
+
+/**
+ * ntfs_attr_set - fill (a part of) an attribute with a byte
+ * @ni:		ntfs inode describing the attribute to fill
+ * @ofs:	offset inside the attribute at which to start to fill
+ * @cnt:	number of bytes to fill
+ * @val:	the unsigned 8-bit value with which to fill the attribute
+ *
+ * Fill @cnt bytes of the attribute described by the ntfs inode @ni starting at
+ * byte offset @ofs inside the attribute with the constant byte @val.
+ *
+ * This function is effectively like memset() applied to an ntfs attribute.
+ * Note thie function actually only operates on the page cache pages belonging
+ * to the ntfs attribute and it marks them dirty after doing the memset().
+ * Thus it relies on the vm dirty page write code paths to cause the modified
+ * pages to be written to the mft record/disk.
+ *
+ * Return 0 on success and -errno on error.  An error code of -ESPIPE means
+ * that @ofs + @cnt were outside the end of the attribute and no write was
+ * performed.
+ */
+int ntfs_attr_set(ntfs_inode *ni, const s64 ofs, const s64 cnt, const u8 val)
+{
+	ntfs_volume *vol = ni->vol;
+	struct address_space *mapping;
+	struct page *page;
+	u8 *kaddr;
+	pgoff_t idx, end;
+	unsigned start_ofs, end_ofs, size;
+
+	ntfs_debug("Entering for ofs 0x%llx, cnt 0x%llx, val 0x%hx.",
+			(long long)ofs, (long long)cnt, val);
+	BUG_ON(ofs < 0);
+	BUG_ON(cnt < 0);
+	if (!cnt)
+		goto done;
+	/*
+	 * FIXME: Compressed and encrypted attributes are not supported when
+	 * writing and we should never have gotten here for them.
+	 */
+	BUG_ON(NInoCompressed(ni));
+	BUG_ON(NInoEncrypted(ni));
+	mapping = VFS_I(ni)->i_mapping;
+	/* Work out the starting index and page offset. */
+	idx = ofs >> PAGE_SHIFT;
+	start_ofs = ofs & ~PAGE_MASK;
+	/* Work out the ending index and page offset. */
+	end = ofs + cnt;
+	end_ofs = end & ~PAGE_MASK;
+	/* If the end is outside the inode size return -ESPIPE. */
+	if (unlikely(end > i_size_read(VFS_I(ni)))) {
+		ntfs_error(vol->sb, "Request exceeds end of attribute.");
+		return -ESPIPE;
+	}
+	end >>= PAGE_SHIFT;
+	/* If there is a first partial page, need to do it the slow way. */
+	if (start_ofs) {
+		page = read_mapping_page(mapping, idx, NULL);
+		if (IS_ERR(page)) {
+			ntfs_error(vol->sb, "Failed to read first partial "
+					"page (error, index 0x%lx).", idx);
+			return PTR_ERR(page);
+		}
+		/*
+		 * If the last page is the same as the first page, need to
+		 * limit the write to the end offset.
+		 */
+		size = PAGE_SIZE;
+		if (idx == end)
+			size = end_ofs;
+		kaddr = kmap_atomic(page);
+		memset(kaddr + start_ofs, val, size - start_ofs);
+		flush_dcache_page(page);
+		kunmap_atomic(kaddr);
+		set_page_dirty(page);
+		put_page(page);
+		balance_dirty_pages_ratelimited(mapping);
+		cond_resched();
+		if (idx == end)
+			goto done;
+		idx++;
+	}
+	/* Do the whole pages the fast way. */
+	for (; idx < end; idx++) {
+		/* Find or create the current page.  (The page is locked.) */
+		page = grab_cache_page(mapping, idx);
+		if (unlikely(!page)) {
+			ntfs_error(vol->sb, "Insufficient memory to grab "
+					"page (index 0x%lx).", idx);
+			return -ENOMEM;
+		}
+		kaddr = kmap_atomic(page);
+		memset(kaddr, val, PAGE_SIZE);
+		flush_dcache_page(page);
+		kunmap_atomic(kaddr);
+		/*
+		 * If the page has buffers, mark them uptodate since buffer
+		 * state and not page state is definitive in 2.6 kernels.
+		 */
+		if (page_has_buffers(page)) {
+			struct buffer_head *bh, *head;
+
+			bh = head = page_buffers(page);
+			do {
+				set_buffer_uptodate(bh);
+			} while ((bh = bh->b_this_page) != head);
+		}
+		/* Now that buffers are uptodate, set the page uptodate, too. */
+		SetPageUptodate(page);
+		/*
+		 * Set the page and all its buffers dirty and mark the inode
+		 * dirty, too.  The VM will write the page later on.
+		 */
+		set_page_dirty(page);
+		/* Finally unlock and release the page. */
+		unlock_page(page);
+		put_page(page);
+		balance_dirty_pages_ratelimited(mapping);
+		cond_resched();
+	}
+	/* If there is a last partial page, need to do it the slow way. */
+	if (end_ofs) {
+		page = read_mapping_page(mapping, idx, NULL);
+		if (IS_ERR(page)) {
+			ntfs_error(vol->sb, "Failed to read last partial page "
+					"(error, index 0x%lx).", idx);
+			return PTR_ERR(page);
+		}
+		kaddr = kmap_atomic(page);
+		memset(kaddr, val, end_ofs);
+		flush_dcache_page(page);
+		kunmap_atomic(kaddr);
+		set_page_dirty(page);
+		put_page(page);
+		balance_dirty_pages_ratelimited(mapping);
+		cond_resched();
+	}
+done:
+	ntfs_debug("Done.");
+	return 0;
+}
+
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/attrib.h b/fs/ntfs/attrib.h
new file mode 100644
index 000000000..fe0890d3d
--- /dev/null
+++ b/fs/ntfs/attrib.h
@@ -0,0 +1,102 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * attrib.h - Defines for attribute handling in NTFS Linux kernel driver.
+ *	      Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2005 Anton Altaparmakov
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#ifndef _LINUX_NTFS_ATTRIB_H
+#define _LINUX_NTFS_ATTRIB_H
+
+#include "endian.h"
+#include "types.h"
+#include "layout.h"
+#include "inode.h"
+#include "runlist.h"
+#include "volume.h"
+
+/**
+ * ntfs_attr_search_ctx - used in attribute search functions
+ * @mrec:	buffer containing mft record to search
+ * @attr:	attribute record in @mrec where to begin/continue search
+ * @is_first:	if true ntfs_attr_lookup() begins search with @attr, else after
+ *
+ * Structure must be initialized to zero before the first call to one of the
+ * attribute search functions. Initialize @mrec to point to the mft record to
+ * search, and @attr to point to the first attribute within @mrec (not necessary
+ * if calling the _first() functions), and set @is_first to 'true' (not necessary
+ * if calling the _first() functions).
+ *
+ * If @is_first is 'true', the search begins with @attr. If @is_first is 'false',
+ * the search begins after @attr. This is so that, after the first call to one
+ * of the search attribute functions, we can call the function again, without
+ * any modification of the search context, to automagically get the next
+ * matching attribute.
+ */
+typedef struct {
+	MFT_RECORD *mrec;
+	ATTR_RECORD *attr;
+	bool is_first;
+	ntfs_inode *ntfs_ino;
+	ATTR_LIST_ENTRY *al_entry;
+	ntfs_inode *base_ntfs_ino;
+	MFT_RECORD *base_mrec;
+	ATTR_RECORD *base_attr;
+} ntfs_attr_search_ctx;
+
+extern int ntfs_map_runlist_nolock(ntfs_inode *ni, VCN vcn,
+		ntfs_attr_search_ctx *ctx);
+extern int ntfs_map_runlist(ntfs_inode *ni, VCN vcn);
+
+extern LCN ntfs_attr_vcn_to_lcn_nolock(ntfs_inode *ni, const VCN vcn,
+		const bool write_locked);
+
+extern runlist_element *ntfs_attr_find_vcn_nolock(ntfs_inode *ni,
+		const VCN vcn, ntfs_attr_search_ctx *ctx);
+
+int ntfs_attr_lookup(const ATTR_TYPE type, const ntfschar *name,
+		const u32 name_len, const IGNORE_CASE_BOOL ic,
+		const VCN lowest_vcn, const u8 *val, const u32 val_len,
+		ntfs_attr_search_ctx *ctx);
+
+extern int load_attribute_list(ntfs_volume *vol, runlist *rl, u8 *al_start,
+		const s64 size, const s64 initialized_size);
+
+static inline s64 ntfs_attr_size(const ATTR_RECORD *a)
+{
+	if (!a->non_resident)
+		return (s64)le32_to_cpu(a->data.resident.value_length);
+	return sle64_to_cpu(a->data.non_resident.data_size);
+}
+
+extern void ntfs_attr_reinit_search_ctx(ntfs_attr_search_ctx *ctx);
+extern ntfs_attr_search_ctx *ntfs_attr_get_search_ctx(ntfs_inode *ni,
+		MFT_RECORD *mrec);
+extern void ntfs_attr_put_search_ctx(ntfs_attr_search_ctx *ctx);
+
+#ifdef NTFS_RW
+
+extern int ntfs_attr_size_bounds_check(const ntfs_volume *vol,
+		const ATTR_TYPE type, const s64 size);
+extern int ntfs_attr_can_be_non_resident(const ntfs_volume *vol,
+		const ATTR_TYPE type);
+extern int ntfs_attr_can_be_resident(const ntfs_volume *vol,
+		const ATTR_TYPE type);
+
+extern int ntfs_attr_record_resize(MFT_RECORD *m, ATTR_RECORD *a, u32 new_size);
+extern int ntfs_resident_attr_value_resize(MFT_RECORD *m, ATTR_RECORD *a,
+		const u32 new_size);
+
+extern int ntfs_attr_make_non_resident(ntfs_inode *ni, const u32 data_size);
+
+extern s64 ntfs_attr_extend_allocation(ntfs_inode *ni, s64 new_alloc_size,
+		const s64 new_data_size, const s64 data_start);
+
+extern int ntfs_attr_set(ntfs_inode *ni, const s64 ofs, const s64 cnt,
+		const u8 val);
+
+#endif /* NTFS_RW */
+
+#endif /* _LINUX_NTFS_ATTRIB_H */
diff --git a/fs/ntfs/bitmap.c b/fs/ntfs/bitmap.c
new file mode 100644
index 000000000..0675b2400
--- /dev/null
+++ b/fs/ntfs/bitmap.c
@@ -0,0 +1,179 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * bitmap.c - NTFS kernel bitmap handling.  Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2004-2005 Anton Altaparmakov
+ */
+
+#ifdef NTFS_RW
+
+#include <linux/pagemap.h>
+
+#include "bitmap.h"
+#include "debug.h"
+#include "aops.h"
+#include "ntfs.h"
+
+/**
+ * __ntfs_bitmap_set_bits_in_run - set a run of bits in a bitmap to a value
+ * @vi:			vfs inode describing the bitmap
+ * @start_bit:		first bit to set
+ * @count:		number of bits to set
+ * @value:		value to set the bits to (i.e. 0 or 1)
+ * @is_rollback:	if 'true' this is a rollback operation
+ *
+ * Set @count bits starting at bit @start_bit in the bitmap described by the
+ * vfs inode @vi to @value, where @value is either 0 or 1.
+ *
+ * @is_rollback should always be 'false', it is for internal use to rollback
+ * errors.  You probably want to use ntfs_bitmap_set_bits_in_run() instead.
+ *
+ * Return 0 on success and -errno on error.
+ */
+int __ntfs_bitmap_set_bits_in_run(struct inode *vi, const s64 start_bit,
+		const s64 count, const u8 value, const bool is_rollback)
+{
+	s64 cnt = count;
+	pgoff_t index, end_index;
+	struct address_space *mapping;
+	struct page *page;
+	u8 *kaddr;
+	int pos, len;
+	u8 bit;
+
+	BUG_ON(!vi);
+	ntfs_debug("Entering for i_ino 0x%lx, start_bit 0x%llx, count 0x%llx, "
+			"value %u.%s", vi->i_ino, (unsigned long long)start_bit,
+			(unsigned long long)cnt, (unsigned int)value,
+			is_rollback ? " (rollback)" : "");
+	BUG_ON(start_bit < 0);
+	BUG_ON(cnt < 0);
+	BUG_ON(value > 1);
+	/*
+	 * Calculate the indices for the pages containing the first and last
+	 * bits, i.e. @start_bit and @start_bit + @cnt - 1, respectively.
+	 */
+	index = start_bit >> (3 + PAGE_SHIFT);
+	end_index = (start_bit + cnt - 1) >> (3 + PAGE_SHIFT);
+
+	/* Get the page containing the first bit (@start_bit). */
+	mapping = vi->i_mapping;
+	page = ntfs_map_page(mapping, index);
+	if (IS_ERR(page)) {
+		if (!is_rollback)
+			ntfs_error(vi->i_sb, "Failed to map first page (error "
+					"%li), aborting.", PTR_ERR(page));
+		return PTR_ERR(page);
+	}
+	kaddr = page_address(page);
+
+	/* Set @pos to the position of the byte containing @start_bit. */
+	pos = (start_bit >> 3) & ~PAGE_MASK;
+
+	/* Calculate the position of @start_bit in the first byte. */
+	bit = start_bit & 7;
+
+	/* If the first byte is partial, modify the appropriate bits in it. */
+	if (bit) {
+		u8 *byte = kaddr + pos;
+		while ((bit & 7) && cnt) {
+			cnt--;
+			if (value)
+				*byte |= 1 << bit++;
+			else
+				*byte &= ~(1 << bit++);
+		}
+		/* If we are done, unmap the page and return success. */
+		if (!cnt)
+			goto done;
+
+		/* Update @pos to the new position. */
+		pos++;
+	}
+	/*
+	 * Depending on @value, modify all remaining whole bytes in the page up
+	 * to @cnt.
+	 */
+	len = min_t(s64, cnt >> 3, PAGE_SIZE - pos);
+	memset(kaddr + pos, value ? 0xff : 0, len);
+	cnt -= len << 3;
+
+	/* Update @len to point to the first not-done byte in the page. */
+	if (cnt < 8)
+		len += pos;
+
+	/* If we are not in the last page, deal with all subsequent pages. */
+	while (index < end_index) {
+		BUG_ON(cnt <= 0);
+
+		/* Update @index and get the next page. */
+		flush_dcache_page(page);
+		set_page_dirty(page);
+		ntfs_unmap_page(page);
+		page = ntfs_map_page(mapping, ++index);
+		if (IS_ERR(page))
+			goto rollback;
+		kaddr = page_address(page);
+		/*
+		 * Depending on @value, modify all remaining whole bytes in the
+		 * page up to @cnt.
+		 */
+		len = min_t(s64, cnt >> 3, PAGE_SIZE);
+		memset(kaddr, value ? 0xff : 0, len);
+		cnt -= len << 3;
+	}
+	/*
+	 * The currently mapped page is the last one.  If the last byte is
+	 * partial, modify the appropriate bits in it.  Note, @len is the
+	 * position of the last byte inside the page.
+	 */
+	if (cnt) {
+		u8 *byte;
+
+		BUG_ON(cnt > 7);
+
+		bit = cnt;
+		byte = kaddr + len;
+		while (bit--) {
+			if (value)
+				*byte |= 1 << bit;
+			else
+				*byte &= ~(1 << bit);
+		}
+	}
+done:
+	/* We are done.  Unmap the page and return success. */
+	flush_dcache_page(page);
+	set_page_dirty(page);
+	ntfs_unmap_page(page);
+	ntfs_debug("Done.");
+	return 0;
+rollback:
+	/*
+	 * Current state:
+	 *	- no pages are mapped
+	 *	- @count - @cnt is the number of bits that have been modified
+	 */
+	if (is_rollback)
+		return PTR_ERR(page);
+	if (count != cnt)
+		pos = __ntfs_bitmap_set_bits_in_run(vi, start_bit, count - cnt,
+				value ? 0 : 1, true);
+	else
+		pos = 0;
+	if (!pos) {
+		/* Rollback was successful. */
+		ntfs_error(vi->i_sb, "Failed to map subsequent page (error "
+				"%li), aborting.", PTR_ERR(page));
+	} else {
+		/* Rollback failed. */
+		ntfs_error(vi->i_sb, "Failed to map subsequent page (error "
+				"%li) and rollback failed (error %i).  "
+				"Aborting and leaving inconsistent metadata.  "
+				"Unmount and run chkdsk.", PTR_ERR(page), pos);
+		NVolSetErrors(NTFS_SB(vi->i_sb));
+	}
+	return PTR_ERR(page);
+}
+
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/bitmap.h b/fs/ntfs/bitmap.h
new file mode 100644
index 000000000..9dd2224ca
--- /dev/null
+++ b/fs/ntfs/bitmap.h
@@ -0,0 +1,104 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * bitmap.h - Defines for NTFS kernel bitmap handling.  Part of the Linux-NTFS
+ *	      project.
+ *
+ * Copyright (c) 2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_BITMAP_H
+#define _LINUX_NTFS_BITMAP_H
+
+#ifdef NTFS_RW
+
+#include <linux/fs.h>
+
+#include "types.h"
+
+extern int __ntfs_bitmap_set_bits_in_run(struct inode *vi, const s64 start_bit,
+		const s64 count, const u8 value, const bool is_rollback);
+
+/**
+ * ntfs_bitmap_set_bits_in_run - set a run of bits in a bitmap to a value
+ * @vi:			vfs inode describing the bitmap
+ * @start_bit:		first bit to set
+ * @count:		number of bits to set
+ * @value:		value to set the bits to (i.e. 0 or 1)
+ *
+ * Set @count bits starting at bit @start_bit in the bitmap described by the
+ * vfs inode @vi to @value, where @value is either 0 or 1.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static inline int ntfs_bitmap_set_bits_in_run(struct inode *vi,
+		const s64 start_bit, const s64 count, const u8 value)
+{
+	return __ntfs_bitmap_set_bits_in_run(vi, start_bit, count, value,
+			false);
+}
+
+/**
+ * ntfs_bitmap_set_run - set a run of bits in a bitmap
+ * @vi:		vfs inode describing the bitmap
+ * @start_bit:	first bit to set
+ * @count:	number of bits to set
+ *
+ * Set @count bits starting at bit @start_bit in the bitmap described by the
+ * vfs inode @vi.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static inline int ntfs_bitmap_set_run(struct inode *vi, const s64 start_bit,
+		const s64 count)
+{
+	return ntfs_bitmap_set_bits_in_run(vi, start_bit, count, 1);
+}
+
+/**
+ * ntfs_bitmap_clear_run - clear a run of bits in a bitmap
+ * @vi:		vfs inode describing the bitmap
+ * @start_bit:	first bit to clear
+ * @count:	number of bits to clear
+ *
+ * Clear @count bits starting at bit @start_bit in the bitmap described by the
+ * vfs inode @vi.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static inline int ntfs_bitmap_clear_run(struct inode *vi, const s64 start_bit,
+		const s64 count)
+{
+	return ntfs_bitmap_set_bits_in_run(vi, start_bit, count, 0);
+}
+
+/**
+ * ntfs_bitmap_set_bit - set a bit in a bitmap
+ * @vi:		vfs inode describing the bitmap
+ * @bit:	bit to set
+ *
+ * Set bit @bit in the bitmap described by the vfs inode @vi.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static inline int ntfs_bitmap_set_bit(struct inode *vi, const s64 bit)
+{
+	return ntfs_bitmap_set_run(vi, bit, 1);
+}
+
+/**
+ * ntfs_bitmap_clear_bit - clear a bit in a bitmap
+ * @vi:		vfs inode describing the bitmap
+ * @bit:	bit to clear
+ *
+ * Clear bit @bit in the bitmap described by the vfs inode @vi.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static inline int ntfs_bitmap_clear_bit(struct inode *vi, const s64 bit)
+{
+	return ntfs_bitmap_clear_run(vi, bit, 1);
+}
+
+#endif /* NTFS_RW */
+
+#endif /* defined _LINUX_NTFS_BITMAP_H */
diff --git a/fs/ntfs/collate.c b/fs/ntfs/collate.c
new file mode 100644
index 000000000..3ab6ec96a
--- /dev/null
+++ b/fs/ntfs/collate.c
@@ -0,0 +1,110 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * collate.c - NTFS kernel collation handling.  Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2004 Anton Altaparmakov
+ */
+
+#include "collate.h"
+#include "debug.h"
+#include "ntfs.h"
+
+static int ntfs_collate_binary(ntfs_volume *vol,
+		const void *data1, const int data1_len,
+		const void *data2, const int data2_len)
+{
+	int rc;
+
+	ntfs_debug("Entering.");
+	rc = memcmp(data1, data2, min(data1_len, data2_len));
+	if (!rc && (data1_len != data2_len)) {
+		if (data1_len < data2_len)
+			rc = -1;
+		else
+			rc = 1;
+	}
+	ntfs_debug("Done, returning %i", rc);
+	return rc;
+}
+
+static int ntfs_collate_ntofs_ulong(ntfs_volume *vol,
+		const void *data1, const int data1_len,
+		const void *data2, const int data2_len)
+{
+	int rc;
+	u32 d1, d2;
+
+	ntfs_debug("Entering.");
+	// FIXME:  We don't really want to bug here.
+	BUG_ON(data1_len != data2_len);
+	BUG_ON(data1_len != 4);
+	d1 = le32_to_cpup(data1);
+	d2 = le32_to_cpup(data2);
+	if (d1 < d2)
+		rc = -1;
+	else {
+		if (d1 == d2)
+			rc = 0;
+		else
+			rc = 1;
+	}
+	ntfs_debug("Done, returning %i", rc);
+	return rc;
+}
+
+typedef int (*ntfs_collate_func_t)(ntfs_volume *, const void *, const int,
+		const void *, const int);
+
+static ntfs_collate_func_t ntfs_do_collate0x0[3] = {
+	ntfs_collate_binary,
+	NULL/*ntfs_collate_file_name*/,
+	NULL/*ntfs_collate_unicode_string*/,
+};
+
+static ntfs_collate_func_t ntfs_do_collate0x1[4] = {
+	ntfs_collate_ntofs_ulong,
+	NULL/*ntfs_collate_ntofs_sid*/,
+	NULL/*ntfs_collate_ntofs_security_hash*/,
+	NULL/*ntfs_collate_ntofs_ulongs*/,
+};
+
+/**
+ * ntfs_collate - collate two data items using a specified collation rule
+ * @vol:	ntfs volume to which the data items belong
+ * @cr:		collation rule to use when comparing the items
+ * @data1:	first data item to collate
+ * @data1_len:	length in bytes of @data1
+ * @data2:	second data item to collate
+ * @data2_len:	length in bytes of @data2
+ *
+ * Collate the two data items @data1 and @data2 using the collation rule @cr
+ * and return -1, 0, ir 1 if @data1 is found, respectively, to collate before,
+ * to match, or to collate after @data2.
+ *
+ * For speed we use the collation rule @cr as an index into two tables of
+ * function pointers to call the appropriate collation function.
+ */
+int ntfs_collate(ntfs_volume *vol, COLLATION_RULE cr,
+		const void *data1, const int data1_len,
+		const void *data2, const int data2_len) {
+	int i;
+
+	ntfs_debug("Entering.");
+	/*
+	 * FIXME:  At the moment we only support COLLATION_BINARY and
+	 * COLLATION_NTOFS_ULONG, so we BUG() for everything else for now.
+	 */
+	BUG_ON(cr != COLLATION_BINARY && cr != COLLATION_NTOFS_ULONG);
+	i = le32_to_cpu(cr);
+	BUG_ON(i < 0);
+	if (i <= 0x02)
+		return ntfs_do_collate0x0[i](vol, data1, data1_len,
+				data2, data2_len);
+	BUG_ON(i < 0x10);
+	i -= 0x10;
+	if (likely(i <= 3))
+		return ntfs_do_collate0x1[i](vol, data1, data1_len,
+				data2, data2_len);
+	BUG();
+	return 0;
+}
diff --git a/fs/ntfs/collate.h b/fs/ntfs/collate.h
new file mode 100644
index 000000000..f2255619b
--- /dev/null
+++ b/fs/ntfs/collate.h
@@ -0,0 +1,36 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * collate.h - Defines for NTFS kernel collation handling.  Part of the
+ *	       Linux-NTFS project.
+ *
+ * Copyright (c) 2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_COLLATE_H
+#define _LINUX_NTFS_COLLATE_H
+
+#include "types.h"
+#include "volume.h"
+
+static inline bool ntfs_is_collation_rule_supported(COLLATION_RULE cr) {
+	int i;
+
+	/*
+	 * FIXME:  At the moment we only support COLLATION_BINARY and
+	 * COLLATION_NTOFS_ULONG, so we return false for everything else for
+	 * now.
+	 */
+	if (unlikely(cr != COLLATION_BINARY && cr != COLLATION_NTOFS_ULONG))
+		return false;
+	i = le32_to_cpu(cr);
+	if (likely(((i >= 0) && (i <= 0x02)) ||
+			((i >= 0x10) && (i <= 0x13))))
+		return true;
+	return false;
+}
+
+extern int ntfs_collate(ntfs_volume *vol, COLLATION_RULE cr,
+		const void *data1, const int data1_len,
+		const void *data2, const int data2_len);
+
+#endif /* _LINUX_NTFS_COLLATE_H */
diff --git a/fs/ntfs/compress.c b/fs/ntfs/compress.c
new file mode 100644
index 000000000..d2f9d6a0e
--- /dev/null
+++ b/fs/ntfs/compress.c
@@ -0,0 +1,950 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/**
+ * compress.c - NTFS kernel compressed attributes handling.
+ *		Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2004 Anton Altaparmakov
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#include <linux/fs.h>
+#include <linux/buffer_head.h>
+#include <linux/blkdev.h>
+#include <linux/vmalloc.h>
+#include <linux/slab.h>
+
+#include "attrib.h"
+#include "inode.h"
+#include "debug.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_compression_constants - enum of constants used in the compression code
+ */
+typedef enum {
+	/* Token types and access mask. */
+	NTFS_SYMBOL_TOKEN	=	0,
+	NTFS_PHRASE_TOKEN	=	1,
+	NTFS_TOKEN_MASK		=	1,
+
+	/* Compression sub-block constants. */
+	NTFS_SB_SIZE_MASK	=	0x0fff,
+	NTFS_SB_SIZE		=	0x1000,
+	NTFS_SB_IS_COMPRESSED	=	0x8000,
+
+	/*
+	 * The maximum compression block size is by definition 16 * the cluster
+	 * size, with the maximum supported cluster size being 4kiB. Thus the
+	 * maximum compression buffer size is 64kiB, so we use this when
+	 * initializing the compression buffer.
+	 */
+	NTFS_MAX_CB_SIZE	= 64 * 1024,
+} ntfs_compression_constants;
+
+/**
+ * ntfs_compression_buffer - one buffer for the decompression engine
+ */
+static u8 *ntfs_compression_buffer;
+
+/**
+ * ntfs_cb_lock - spinlock which protects ntfs_compression_buffer
+ */
+static DEFINE_SPINLOCK(ntfs_cb_lock);
+
+/**
+ * allocate_compression_buffers - allocate the decompression buffers
+ *
+ * Caller has to hold the ntfs_lock mutex.
+ *
+ * Return 0 on success or -ENOMEM if the allocations failed.
+ */
+int allocate_compression_buffers(void)
+{
+	BUG_ON(ntfs_compression_buffer);
+
+	ntfs_compression_buffer = vmalloc(NTFS_MAX_CB_SIZE);
+	if (!ntfs_compression_buffer)
+		return -ENOMEM;
+	return 0;
+}
+
+/**
+ * free_compression_buffers - free the decompression buffers
+ *
+ * Caller has to hold the ntfs_lock mutex.
+ */
+void free_compression_buffers(void)
+{
+	BUG_ON(!ntfs_compression_buffer);
+	vfree(ntfs_compression_buffer);
+	ntfs_compression_buffer = NULL;
+}
+
+/**
+ * zero_partial_compressed_page - zero out of bounds compressed page region
+ */
+static void zero_partial_compressed_page(struct page *page,
+		const s64 initialized_size)
+{
+	u8 *kp = page_address(page);
+	unsigned int kp_ofs;
+
+	ntfs_debug("Zeroing page region outside initialized size.");
+	if (((s64)page->index << PAGE_SHIFT) >= initialized_size) {
+		clear_page(kp);
+		return;
+	}
+	kp_ofs = initialized_size & ~PAGE_MASK;
+	memset(kp + kp_ofs, 0, PAGE_SIZE - kp_ofs);
+	return;
+}
+
+/**
+ * handle_bounds_compressed_page - test for&handle out of bounds compressed page
+ */
+static inline void handle_bounds_compressed_page(struct page *page,
+		const loff_t i_size, const s64 initialized_size)
+{
+	if ((page->index >= (initialized_size >> PAGE_SHIFT)) &&
+			(initialized_size < i_size))
+		zero_partial_compressed_page(page, initialized_size);
+	return;
+}
+
+/**
+ * ntfs_decompress - decompress a compression block into an array of pages
+ * @dest_pages:		destination array of pages
+ * @completed_pages:	scratch space to track completed pages
+ * @dest_index:		current index into @dest_pages (IN/OUT)
+ * @dest_ofs:		current offset within @dest_pages[@dest_index] (IN/OUT)
+ * @dest_max_index:	maximum index into @dest_pages (IN)
+ * @dest_max_ofs:	maximum offset within @dest_pages[@dest_max_index] (IN)
+ * @xpage:		the target page (-1 if none) (IN)
+ * @xpage_done:		set to 1 if xpage was completed successfully (IN/OUT)
+ * @cb_start:		compression block to decompress (IN)
+ * @cb_size:		size of compression block @cb_start in bytes (IN)
+ * @i_size:		file size when we started the read (IN)
+ * @initialized_size:	initialized file size when we started the read (IN)
+ *
+ * The caller must have disabled preemption. ntfs_decompress() reenables it when
+ * the critical section is finished.
+ *
+ * This decompresses the compression block @cb_start into the array of
+ * destination pages @dest_pages starting at index @dest_index into @dest_pages
+ * and at offset @dest_pos into the page @dest_pages[@dest_index].
+ *
+ * When the page @dest_pages[@xpage] is completed, @xpage_done is set to 1.
+ * If xpage is -1 or @xpage has not been completed, @xpage_done is not modified.
+ *
+ * @cb_start is a pointer to the compression block which needs decompressing
+ * and @cb_size is the size of @cb_start in bytes (8-64kiB).
+ *
+ * Return 0 if success or -EOVERFLOW on error in the compressed stream.
+ * @xpage_done indicates whether the target page (@dest_pages[@xpage]) was
+ * completed during the decompression of the compression block (@cb_start).
+ *
+ * Warning: This function *REQUIRES* PAGE_SIZE >= 4096 or it will blow up
+ * unpredicatbly! You have been warned!
+ *
+ * Note to hackers: This function may not sleep until it has finished accessing
+ * the compression block @cb_start as it is a per-CPU buffer.
+ */
+static int ntfs_decompress(struct page *dest_pages[], int completed_pages[],
+		int *dest_index, int *dest_ofs, const int dest_max_index,
+		const int dest_max_ofs, const int xpage, char *xpage_done,
+		u8 *const cb_start, const u32 cb_size, const loff_t i_size,
+		const s64 initialized_size)
+{
+	/*
+	 * Pointers into the compressed data, i.e. the compression block (cb),
+	 * and the therein contained sub-blocks (sb).
+	 */
+	u8 *cb_end = cb_start + cb_size; /* End of cb. */
+	u8 *cb = cb_start;	/* Current position in cb. */
+	u8 *cb_sb_start = cb;	/* Beginning of the current sb in the cb. */
+	u8 *cb_sb_end;		/* End of current sb / beginning of next sb. */
+
+	/* Variables for uncompressed data / destination. */
+	struct page *dp;	/* Current destination page being worked on. */
+	u8 *dp_addr;		/* Current pointer into dp. */
+	u8 *dp_sb_start;	/* Start of current sub-block in dp. */
+	u8 *dp_sb_end;		/* End of current sb in dp (dp_sb_start +
+				   NTFS_SB_SIZE). */
+	u16 do_sb_start;	/* @dest_ofs when starting this sub-block. */
+	u16 do_sb_end;		/* @dest_ofs of end of this sb (do_sb_start +
+				   NTFS_SB_SIZE). */
+
+	/* Variables for tag and token parsing. */
+	u8 tag;			/* Current tag. */
+	int token;		/* Loop counter for the eight tokens in tag. */
+	int nr_completed_pages = 0;
+
+	/* Default error code. */
+	int err = -EOVERFLOW;
+
+	ntfs_debug("Entering, cb_size = 0x%x.", cb_size);
+do_next_sb:
+	ntfs_debug("Beginning sub-block at offset = 0x%zx in the cb.",
+			cb - cb_start);
+	/*
+	 * Have we reached the end of the compression block or the end of the
+	 * decompressed data?  The latter can happen for example if the current
+	 * position in the compression block is one byte before its end so the
+	 * first two checks do not detect it.
+	 */
+	if (cb == cb_end || !le16_to_cpup((le16*)cb) ||
+			(*dest_index == dest_max_index &&
+			*dest_ofs == dest_max_ofs)) {
+		int i;
+
+		ntfs_debug("Completed. Returning success (0).");
+		err = 0;
+return_error:
+		/* We can sleep from now on, so we drop lock. */
+		spin_unlock(&ntfs_cb_lock);
+		/* Second stage: finalize completed pages. */
+		if (nr_completed_pages > 0) {
+			for (i = 0; i < nr_completed_pages; i++) {
+				int di = completed_pages[i];
+
+				dp = dest_pages[di];
+				/*
+				 * If we are outside the initialized size, zero
+				 * the out of bounds page range.
+				 */
+				handle_bounds_compressed_page(dp, i_size,
+						initialized_size);
+				flush_dcache_page(dp);
+				kunmap(dp);
+				SetPageUptodate(dp);
+				unlock_page(dp);
+				if (di == xpage)
+					*xpage_done = 1;
+				else
+					put_page(dp);
+				dest_pages[di] = NULL;
+			}
+		}
+		return err;
+	}
+
+	/* Setup offsets for the current sub-block destination. */
+	do_sb_start = *dest_ofs;
+	do_sb_end = do_sb_start + NTFS_SB_SIZE;
+
+	/* Check that we are still within allowed boundaries. */
+	if (*dest_index == dest_max_index && do_sb_end > dest_max_ofs)
+		goto return_overflow;
+
+	/* Does the minimum size of a compressed sb overflow valid range? */
+	if (cb + 6 > cb_end)
+		goto return_overflow;
+
+	/* Setup the current sub-block source pointers and validate range. */
+	cb_sb_start = cb;
+	cb_sb_end = cb_sb_start + (le16_to_cpup((le16*)cb) & NTFS_SB_SIZE_MASK)
+			+ 3;
+	if (cb_sb_end > cb_end)
+		goto return_overflow;
+
+	/* Get the current destination page. */
+	dp = dest_pages[*dest_index];
+	if (!dp) {
+		/* No page present. Skip decompression of this sub-block. */
+		cb = cb_sb_end;
+
+		/* Advance destination position to next sub-block. */
+		*dest_ofs = (*dest_ofs + NTFS_SB_SIZE) & ~PAGE_MASK;
+		if (!*dest_ofs && (++*dest_index > dest_max_index))
+			goto return_overflow;
+		goto do_next_sb;
+	}
+
+	/* We have a valid destination page. Setup the destination pointers. */
+	dp_addr = (u8*)page_address(dp) + do_sb_start;
+
+	/* Now, we are ready to process the current sub-block (sb). */
+	if (!(le16_to_cpup((le16*)cb) & NTFS_SB_IS_COMPRESSED)) {
+		ntfs_debug("Found uncompressed sub-block.");
+		/* This sb is not compressed, just copy it into destination. */
+
+		/* Advance source position to first data byte. */
+		cb += 2;
+
+		/* An uncompressed sb must be full size. */
+		if (cb_sb_end - cb != NTFS_SB_SIZE)
+			goto return_overflow;
+
+		/* Copy the block and advance the source position. */
+		memcpy(dp_addr, cb, NTFS_SB_SIZE);
+		cb += NTFS_SB_SIZE;
+
+		/* Advance destination position to next sub-block. */
+		*dest_ofs += NTFS_SB_SIZE;
+		if (!(*dest_ofs &= ~PAGE_MASK)) {
+finalize_page:
+			/*
+			 * First stage: add current page index to array of
+			 * completed pages.
+			 */
+			completed_pages[nr_completed_pages++] = *dest_index;
+			if (++*dest_index > dest_max_index)
+				goto return_overflow;
+		}
+		goto do_next_sb;
+	}
+	ntfs_debug("Found compressed sub-block.");
+	/* This sb is compressed, decompress it into destination. */
+
+	/* Setup destination pointers. */
+	dp_sb_start = dp_addr;
+	dp_sb_end = dp_sb_start + NTFS_SB_SIZE;
+
+	/* Forward to the first tag in the sub-block. */
+	cb += 2;
+do_next_tag:
+	if (cb == cb_sb_end) {
+		/* Check if the decompressed sub-block was not full-length. */
+		if (dp_addr < dp_sb_end) {
+			int nr_bytes = do_sb_end - *dest_ofs;
+
+			ntfs_debug("Filling incomplete sub-block with "
+					"zeroes.");
+			/* Zero remainder and update destination position. */
+			memset(dp_addr, 0, nr_bytes);
+			*dest_ofs += nr_bytes;
+		}
+		/* We have finished the current sub-block. */
+		if (!(*dest_ofs &= ~PAGE_MASK))
+			goto finalize_page;
+		goto do_next_sb;
+	}
+
+	/* Check we are still in range. */
+	if (cb > cb_sb_end || dp_addr > dp_sb_end)
+		goto return_overflow;
+
+	/* Get the next tag and advance to first token. */
+	tag = *cb++;
+
+	/* Parse the eight tokens described by the tag. */
+	for (token = 0; token < 8; token++, tag >>= 1) {
+		u16 lg, pt, length, max_non_overlap;
+		register u16 i;
+		u8 *dp_back_addr;
+
+		/* Check if we are done / still in range. */
+		if (cb >= cb_sb_end || dp_addr > dp_sb_end)
+			break;
+
+		/* Determine token type and parse appropriately.*/
+		if ((tag & NTFS_TOKEN_MASK) == NTFS_SYMBOL_TOKEN) {
+			/*
+			 * We have a symbol token, copy the symbol across, and
+			 * advance the source and destination positions.
+			 */
+			*dp_addr++ = *cb++;
+			++*dest_ofs;
+
+			/* Continue with the next token. */
+			continue;
+		}
+
+		/*
+		 * We have a phrase token. Make sure it is not the first tag in
+		 * the sb as this is illegal and would confuse the code below.
+		 */
+		if (dp_addr == dp_sb_start)
+			goto return_overflow;
+
+		/*
+		 * Determine the number of bytes to go back (p) and the number
+		 * of bytes to copy (l). We use an optimized algorithm in which
+		 * we first calculate log2(current destination position in sb),
+		 * which allows determination of l and p in O(1) rather than
+		 * O(n). We just need an arch-optimized log2() function now.
+		 */
+		lg = 0;
+		for (i = *dest_ofs - do_sb_start - 1; i >= 0x10; i >>= 1)
+			lg++;
+
+		/* Get the phrase token into i. */
+		pt = le16_to_cpup((le16*)cb);
+
+		/*
+		 * Calculate starting position of the byte sequence in
+		 * the destination using the fact that p = (pt >> (12 - lg)) + 1
+		 * and make sure we don't go too far back.
+		 */
+		dp_back_addr = dp_addr - (pt >> (12 - lg)) - 1;
+		if (dp_back_addr < dp_sb_start)
+			goto return_overflow;
+
+		/* Now calculate the length of the byte sequence. */
+		length = (pt & (0xfff >> lg)) + 3;
+
+		/* Advance destination position and verify it is in range. */
+		*dest_ofs += length;
+		if (*dest_ofs > do_sb_end)
+			goto return_overflow;
+
+		/* The number of non-overlapping bytes. */
+		max_non_overlap = dp_addr - dp_back_addr;
+
+		if (length <= max_non_overlap) {
+			/* The byte sequence doesn't overlap, just copy it. */
+			memcpy(dp_addr, dp_back_addr, length);
+
+			/* Advance destination pointer. */
+			dp_addr += length;
+		} else {
+			/*
+			 * The byte sequence does overlap, copy non-overlapping
+			 * part and then do a slow byte by byte copy for the
+			 * overlapping part. Also, advance the destination
+			 * pointer.
+			 */
+			memcpy(dp_addr, dp_back_addr, max_non_overlap);
+			dp_addr += max_non_overlap;
+			dp_back_addr += max_non_overlap;
+			length -= max_non_overlap;
+			while (length--)
+				*dp_addr++ = *dp_back_addr++;
+		}
+
+		/* Advance source position and continue with the next token. */
+		cb += 2;
+	}
+
+	/* No tokens left in the current tag. Continue with the next tag. */
+	goto do_next_tag;
+
+return_overflow:
+	ntfs_error(NULL, "Failed. Returning -EOVERFLOW.");
+	goto return_error;
+}
+
+/**
+ * ntfs_read_compressed_block - read a compressed block into the page cache
+ * @page:	locked page in the compression block(s) we need to read
+ *
+ * When we are called the page has already been verified to be locked and the
+ * attribute is known to be non-resident, not encrypted, but compressed.
+ *
+ * 1. Determine which compression block(s) @page is in.
+ * 2. Get hold of all pages corresponding to this/these compression block(s).
+ * 3. Read the (first) compression block.
+ * 4. Decompress it into the corresponding pages.
+ * 5. Throw the compressed data away and proceed to 3. for the next compression
+ *    block or return success if no more compression blocks left.
+ *
+ * Warning: We have to be careful what we do about existing pages. They might
+ * have been written to so that we would lose data if we were to just overwrite
+ * them with the out-of-date uncompressed data.
+ *
+ * FIXME: For PAGE_SIZE > cb_size we are not doing the Right Thing(TM) at
+ * the end of the file I think. We need to detect this case and zero the out
+ * of bounds remainder of the page in question and mark it as handled. At the
+ * moment we would just return -EIO on such a page. This bug will only become
+ * apparent if pages are above 8kiB and the NTFS volume only uses 512 byte
+ * clusters so is probably not going to be seen by anyone. Still this should
+ * be fixed. (AIA)
+ *
+ * FIXME: Again for PAGE_SIZE > cb_size we are screwing up both in
+ * handling sparse and compressed cbs. (AIA)
+ *
+ * FIXME: At the moment we don't do any zeroing out in the case that
+ * initialized_size is less than data_size. This should be safe because of the
+ * nature of the compression algorithm used. Just in case we check and output
+ * an error message in read inode if the two sizes are not equal for a
+ * compressed file. (AIA)
+ */
+int ntfs_read_compressed_block(struct page *page)
+{
+	loff_t i_size;
+	s64 initialized_size;
+	struct address_space *mapping = page->mapping;
+	ntfs_inode *ni = NTFS_I(mapping->host);
+	ntfs_volume *vol = ni->vol;
+	struct super_block *sb = vol->sb;
+	runlist_element *rl;
+	unsigned long flags, block_size = sb->s_blocksize;
+	unsigned char block_size_bits = sb->s_blocksize_bits;
+	u8 *cb, *cb_pos, *cb_end;
+	struct buffer_head **bhs;
+	unsigned long offset, index = page->index;
+	u32 cb_size = ni->itype.compressed.block_size;
+	u64 cb_size_mask = cb_size - 1UL;
+	VCN vcn;
+	LCN lcn;
+	/* The first wanted vcn (minimum alignment is PAGE_SIZE). */
+	VCN start_vcn = (((s64)index << PAGE_SHIFT) & ~cb_size_mask) >>
+			vol->cluster_size_bits;
+	/*
+	 * The first vcn after the last wanted vcn (minimum alignment is again
+	 * PAGE_SIZE.
+	 */
+	VCN end_vcn = ((((s64)(index + 1UL) << PAGE_SHIFT) + cb_size - 1)
+			& ~cb_size_mask) >> vol->cluster_size_bits;
+	/* Number of compression blocks (cbs) in the wanted vcn range. */
+	unsigned int nr_cbs = (end_vcn - start_vcn) << vol->cluster_size_bits
+			>> ni->itype.compressed.block_size_bits;
+	/*
+	 * Number of pages required to store the uncompressed data from all
+	 * compression blocks (cbs) overlapping @page. Due to alignment
+	 * guarantees of start_vcn and end_vcn, no need to round up here.
+	 */
+	unsigned int nr_pages = (end_vcn - start_vcn) <<
+			vol->cluster_size_bits >> PAGE_SHIFT;
+	unsigned int xpage, max_page, cur_page, cur_ofs, i;
+	unsigned int cb_clusters, cb_max_ofs;
+	int block, max_block, cb_max_page, bhs_size, nr_bhs, err = 0;
+	struct page **pages;
+	int *completed_pages;
+	unsigned char xpage_done = 0;
+
+	ntfs_debug("Entering, page->index = 0x%lx, cb_size = 0x%x, nr_pages = "
+			"%i.", index, cb_size, nr_pages);
+	/*
+	 * Bad things happen if we get here for anything that is not an
+	 * unnamed $DATA attribute.
+	 */
+	BUG_ON(ni->type != AT_DATA);
+	BUG_ON(ni->name_len);
+
+	pages = kmalloc_array(nr_pages, sizeof(struct page *), GFP_NOFS);
+	completed_pages = kmalloc_array(nr_pages + 1, sizeof(int), GFP_NOFS);
+
+	/* Allocate memory to store the buffer heads we need. */
+	bhs_size = cb_size / block_size * sizeof(struct buffer_head *);
+	bhs = kmalloc(bhs_size, GFP_NOFS);
+
+	if (unlikely(!pages || !bhs || !completed_pages)) {
+		kfree(bhs);
+		kfree(pages);
+		kfree(completed_pages);
+		unlock_page(page);
+		ntfs_error(vol->sb, "Failed to allocate internal buffers.");
+		return -ENOMEM;
+	}
+
+	/*
+	 * We have already been given one page, this is the one we must do.
+	 * Once again, the alignment guarantees keep it simple.
+	 */
+	offset = start_vcn << vol->cluster_size_bits >> PAGE_SHIFT;
+	xpage = index - offset;
+	pages[xpage] = page;
+	/*
+	 * The remaining pages need to be allocated and inserted into the page
+	 * cache, alignment guarantees keep all the below much simpler. (-8
+	 */
+	read_lock_irqsave(&ni->size_lock, flags);
+	i_size = i_size_read(VFS_I(ni));
+	initialized_size = ni->initialized_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	max_page = ((i_size + PAGE_SIZE - 1) >> PAGE_SHIFT) -
+			offset;
+	/* Is the page fully outside i_size? (truncate in progress) */
+	if (xpage >= max_page) {
+		kfree(bhs);
+		kfree(pages);
+		kfree(completed_pages);
+		zero_user(page, 0, PAGE_SIZE);
+		ntfs_debug("Compressed read outside i_size - truncated?");
+		SetPageUptodate(page);
+		unlock_page(page);
+		return 0;
+	}
+	if (nr_pages < max_page)
+		max_page = nr_pages;
+	for (i = 0; i < max_page; i++, offset++) {
+		if (i != xpage)
+			pages[i] = grab_cache_page_nowait(mapping, offset);
+		page = pages[i];
+		if (page) {
+			/*
+			 * We only (re)read the page if it isn't already read
+			 * in and/or dirty or we would be losing data or at
+			 * least wasting our time.
+			 */
+			if (!PageDirty(page) && (!PageUptodate(page) ||
+					PageError(page))) {
+				ClearPageError(page);
+				kmap(page);
+				continue;
+			}
+			unlock_page(page);
+			put_page(page);
+			pages[i] = NULL;
+		}
+	}
+
+	/*
+	 * We have the runlist, and all the destination pages we need to fill.
+	 * Now read the first compression block.
+	 */
+	cur_page = 0;
+	cur_ofs = 0;
+	cb_clusters = ni->itype.compressed.block_clusters;
+do_next_cb:
+	nr_cbs--;
+	nr_bhs = 0;
+
+	/* Read all cb buffer heads one cluster at a time. */
+	rl = NULL;
+	for (vcn = start_vcn, start_vcn += cb_clusters; vcn < start_vcn;
+			vcn++) {
+		bool is_retry = false;
+
+		if (!rl) {
+lock_retry_remap:
+			down_read(&ni->runlist.lock);
+			rl = ni->runlist.rl;
+		}
+		if (likely(rl != NULL)) {
+			/* Seek to element containing target vcn. */
+			while (rl->length && rl[1].vcn <= vcn)
+				rl++;
+			lcn = ntfs_rl_vcn_to_lcn(rl, vcn);
+		} else
+			lcn = LCN_RL_NOT_MAPPED;
+		ntfs_debug("Reading vcn = 0x%llx, lcn = 0x%llx.",
+				(unsigned long long)vcn,
+				(unsigned long long)lcn);
+		if (lcn < 0) {
+			/*
+			 * When we reach the first sparse cluster we have
+			 * finished with the cb.
+			 */
+			if (lcn == LCN_HOLE)
+				break;
+			if (is_retry || lcn != LCN_RL_NOT_MAPPED)
+				goto rl_err;
+			is_retry = true;
+			/*
+			 * Attempt to map runlist, dropping lock for the
+			 * duration.
+			 */
+			up_read(&ni->runlist.lock);
+			if (!ntfs_map_runlist(ni, vcn))
+				goto lock_retry_remap;
+			goto map_rl_err;
+		}
+		block = lcn << vol->cluster_size_bits >> block_size_bits;
+		/* Read the lcn from device in chunks of block_size bytes. */
+		max_block = block + (vol->cluster_size >> block_size_bits);
+		do {
+			ntfs_debug("block = 0x%x.", block);
+			if (unlikely(!(bhs[nr_bhs] = sb_getblk(sb, block))))
+				goto getblk_err;
+			nr_bhs++;
+		} while (++block < max_block);
+	}
+
+	/* Release the lock if we took it. */
+	if (rl)
+		up_read(&ni->runlist.lock);
+
+	/* Setup and initiate io on all buffer heads. */
+	for (i = 0; i < nr_bhs; i++) {
+		struct buffer_head *tbh = bhs[i];
+
+		if (!trylock_buffer(tbh))
+			continue;
+		if (unlikely(buffer_uptodate(tbh))) {
+			unlock_buffer(tbh);
+			continue;
+		}
+		get_bh(tbh);
+		tbh->b_end_io = end_buffer_read_sync;
+		submit_bh(REQ_OP_READ, 0, tbh);
+	}
+
+	/* Wait for io completion on all buffer heads. */
+	for (i = 0; i < nr_bhs; i++) {
+		struct buffer_head *tbh = bhs[i];
+
+		if (buffer_uptodate(tbh))
+			continue;
+		wait_on_buffer(tbh);
+		/*
+		 * We need an optimization barrier here, otherwise we start
+		 * hitting the below fixup code when accessing a loopback
+		 * mounted ntfs partition. This indicates either there is a
+		 * race condition in the loop driver or, more likely, gcc
+		 * overoptimises the code without the barrier and it doesn't
+		 * do the Right Thing(TM).
+		 */
+		barrier();
+		if (unlikely(!buffer_uptodate(tbh))) {
+			ntfs_warning(vol->sb, "Buffer is unlocked but not "
+					"uptodate! Unplugging the disk queue "
+					"and rescheduling.");
+			get_bh(tbh);
+			io_schedule();
+			put_bh(tbh);
+			if (unlikely(!buffer_uptodate(tbh)))
+				goto read_err;
+			ntfs_warning(vol->sb, "Buffer is now uptodate. Good.");
+		}
+	}
+
+	/*
+	 * Get the compression buffer. We must not sleep any more
+	 * until we are finished with it.
+	 */
+	spin_lock(&ntfs_cb_lock);
+	cb = ntfs_compression_buffer;
+
+	BUG_ON(!cb);
+
+	cb_pos = cb;
+	cb_end = cb + cb_size;
+
+	/* Copy the buffer heads into the contiguous buffer. */
+	for (i = 0; i < nr_bhs; i++) {
+		memcpy(cb_pos, bhs[i]->b_data, block_size);
+		cb_pos += block_size;
+	}
+
+	/* Just a precaution. */
+	if (cb_pos + 2 <= cb + cb_size)
+		*(u16*)cb_pos = 0;
+
+	/* Reset cb_pos back to the beginning. */
+	cb_pos = cb;
+
+	/* We now have both source (if present) and destination. */
+	ntfs_debug("Successfully read the compression block.");
+
+	/* The last page and maximum offset within it for the current cb. */
+	cb_max_page = (cur_page << PAGE_SHIFT) + cur_ofs + cb_size;
+	cb_max_ofs = cb_max_page & ~PAGE_MASK;
+	cb_max_page >>= PAGE_SHIFT;
+
+	/* Catch end of file inside a compression block. */
+	if (cb_max_page > max_page)
+		cb_max_page = max_page;
+
+	if (vcn == start_vcn - cb_clusters) {
+		/* Sparse cb, zero out page range overlapping the cb. */
+		ntfs_debug("Found sparse compression block.");
+		/* We can sleep from now on, so we drop lock. */
+		spin_unlock(&ntfs_cb_lock);
+		if (cb_max_ofs)
+			cb_max_page--;
+		for (; cur_page < cb_max_page; cur_page++) {
+			page = pages[cur_page];
+			if (page) {
+				if (likely(!cur_ofs))
+					clear_page(page_address(page));
+				else
+					memset(page_address(page) + cur_ofs, 0,
+							PAGE_SIZE -
+							cur_ofs);
+				flush_dcache_page(page);
+				kunmap(page);
+				SetPageUptodate(page);
+				unlock_page(page);
+				if (cur_page == xpage)
+					xpage_done = 1;
+				else
+					put_page(page);
+				pages[cur_page] = NULL;
+			}
+			cb_pos += PAGE_SIZE - cur_ofs;
+			cur_ofs = 0;
+			if (cb_pos >= cb_end)
+				break;
+		}
+		/* If we have a partial final page, deal with it now. */
+		if (cb_max_ofs && cb_pos < cb_end) {
+			page = pages[cur_page];
+			if (page)
+				memset(page_address(page) + cur_ofs, 0,
+						cb_max_ofs - cur_ofs);
+			/*
+			 * No need to update cb_pos at this stage:
+			 *	cb_pos += cb_max_ofs - cur_ofs;
+			 */
+			cur_ofs = cb_max_ofs;
+		}
+	} else if (vcn == start_vcn) {
+		/* We can't sleep so we need two stages. */
+		unsigned int cur2_page = cur_page;
+		unsigned int cur_ofs2 = cur_ofs;
+		u8 *cb_pos2 = cb_pos;
+
+		ntfs_debug("Found uncompressed compression block.");
+		/* Uncompressed cb, copy it to the destination pages. */
+		/*
+		 * TODO: As a big optimization, we could detect this case
+		 * before we read all the pages and use block_read_full_page()
+		 * on all full pages instead (we still have to treat partial
+		 * pages especially but at least we are getting rid of the
+		 * synchronous io for the majority of pages.
+		 * Or if we choose not to do the read-ahead/-behind stuff, we
+		 * could just return block_read_full_page(pages[xpage]) as long
+		 * as PAGE_SIZE <= cb_size.
+		 */
+		if (cb_max_ofs)
+			cb_max_page--;
+		/* First stage: copy data into destination pages. */
+		for (; cur_page < cb_max_page; cur_page++) {
+			page = pages[cur_page];
+			if (page)
+				memcpy(page_address(page) + cur_ofs, cb_pos,
+						PAGE_SIZE - cur_ofs);
+			cb_pos += PAGE_SIZE - cur_ofs;
+			cur_ofs = 0;
+			if (cb_pos >= cb_end)
+				break;
+		}
+		/* If we have a partial final page, deal with it now. */
+		if (cb_max_ofs && cb_pos < cb_end) {
+			page = pages[cur_page];
+			if (page)
+				memcpy(page_address(page) + cur_ofs, cb_pos,
+						cb_max_ofs - cur_ofs);
+			cb_pos += cb_max_ofs - cur_ofs;
+			cur_ofs = cb_max_ofs;
+		}
+		/* We can sleep from now on, so drop lock. */
+		spin_unlock(&ntfs_cb_lock);
+		/* Second stage: finalize pages. */
+		for (; cur2_page < cb_max_page; cur2_page++) {
+			page = pages[cur2_page];
+			if (page) {
+				/*
+				 * If we are outside the initialized size, zero
+				 * the out of bounds page range.
+				 */
+				handle_bounds_compressed_page(page, i_size,
+						initialized_size);
+				flush_dcache_page(page);
+				kunmap(page);
+				SetPageUptodate(page);
+				unlock_page(page);
+				if (cur2_page == xpage)
+					xpage_done = 1;
+				else
+					put_page(page);
+				pages[cur2_page] = NULL;
+			}
+			cb_pos2 += PAGE_SIZE - cur_ofs2;
+			cur_ofs2 = 0;
+			if (cb_pos2 >= cb_end)
+				break;
+		}
+	} else {
+		/* Compressed cb, decompress it into the destination page(s). */
+		unsigned int prev_cur_page = cur_page;
+
+		ntfs_debug("Found compressed compression block.");
+		err = ntfs_decompress(pages, completed_pages, &cur_page,
+				&cur_ofs, cb_max_page, cb_max_ofs, xpage,
+				&xpage_done, cb_pos, cb_size - (cb_pos - cb),
+				i_size, initialized_size);
+		/*
+		 * We can sleep from now on, lock already dropped by
+		 * ntfs_decompress().
+		 */
+		if (err) {
+			ntfs_error(vol->sb, "ntfs_decompress() failed in inode "
+					"0x%lx with error code %i. Skipping "
+					"this compression block.",
+					ni->mft_no, -err);
+			/* Release the unfinished pages. */
+			for (; prev_cur_page < cur_page; prev_cur_page++) {
+				page = pages[prev_cur_page];
+				if (page) {
+					flush_dcache_page(page);
+					kunmap(page);
+					unlock_page(page);
+					if (prev_cur_page != xpage)
+						put_page(page);
+					pages[prev_cur_page] = NULL;
+				}
+			}
+		}
+	}
+
+	/* Release the buffer heads. */
+	for (i = 0; i < nr_bhs; i++)
+		brelse(bhs[i]);
+
+	/* Do we have more work to do? */
+	if (nr_cbs)
+		goto do_next_cb;
+
+	/* We no longer need the list of buffer heads. */
+	kfree(bhs);
+
+	/* Clean up if we have any pages left. Should never happen. */
+	for (cur_page = 0; cur_page < max_page; cur_page++) {
+		page = pages[cur_page];
+		if (page) {
+			ntfs_error(vol->sb, "Still have pages left! "
+					"Terminating them with extreme "
+					"prejudice.  Inode 0x%lx, page index "
+					"0x%lx.", ni->mft_no, page->index);
+			flush_dcache_page(page);
+			kunmap(page);
+			unlock_page(page);
+			if (cur_page != xpage)
+				put_page(page);
+			pages[cur_page] = NULL;
+		}
+	}
+
+	/* We no longer need the list of pages. */
+	kfree(pages);
+	kfree(completed_pages);
+
+	/* If we have completed the requested page, we return success. */
+	if (likely(xpage_done))
+		return 0;
+
+	ntfs_debug("Failed. Returning error code %s.", err == -EOVERFLOW ?
+			"EOVERFLOW" : (!err ? "EIO" : "unknown error"));
+	return err < 0 ? err : -EIO;
+
+read_err:
+	ntfs_error(vol->sb, "IO error while reading compressed data.");
+	/* Release the buffer heads. */
+	for (i = 0; i < nr_bhs; i++)
+		brelse(bhs[i]);
+	goto err_out;
+
+map_rl_err:
+	ntfs_error(vol->sb, "ntfs_map_runlist() failed. Cannot read "
+			"compression block.");
+	goto err_out;
+
+rl_err:
+	up_read(&ni->runlist.lock);
+	ntfs_error(vol->sb, "ntfs_rl_vcn_to_lcn() failed. Cannot read "
+			"compression block.");
+	goto err_out;
+
+getblk_err:
+	up_read(&ni->runlist.lock);
+	ntfs_error(vol->sb, "getblk() failed. Cannot read compression block.");
+
+err_out:
+	kfree(bhs);
+	for (i = cur_page; i < max_page; i++) {
+		page = pages[i];
+		if (page) {
+			flush_dcache_page(page);
+			kunmap(page);
+			unlock_page(page);
+			if (i != xpage)
+				put_page(page);
+		}
+	}
+	kfree(pages);
+	kfree(completed_pages);
+	return -EIO;
+}
diff --git a/fs/ntfs/debug.c b/fs/ntfs/debug.c
new file mode 100644
index 000000000..a3c1c5656
--- /dev/null
+++ b/fs/ntfs/debug.c
@@ -0,0 +1,159 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * debug.c - NTFS kernel debug support. Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2004 Anton Altaparmakov
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#include "debug.h"
+
+/**
+ * __ntfs_warning - output a warning to the syslog
+ * @function:	name of function outputting the warning
+ * @sb:		super block of mounted ntfs filesystem
+ * @fmt:	warning string containing format specifications
+ * @...:	a variable number of arguments specified in @fmt
+ *
+ * Outputs a warning to the syslog for the mounted ntfs filesystem described
+ * by @sb.
+ *
+ * @fmt and the corresponding @... is printf style format string containing
+ * the warning string and the corresponding format arguments, respectively.
+ *
+ * @function is the name of the function from which __ntfs_warning is being
+ * called.
+ *
+ * Note, you should be using debug.h::ntfs_warning(@sb, @fmt, @...) instead
+ * as this provides the @function parameter automatically.
+ */
+void __ntfs_warning(const char *function, const struct super_block *sb,
+		const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list args;
+	int flen = 0;
+
+#ifndef DEBUG
+	if (!printk_ratelimit())
+		return;
+#endif
+	if (function)
+		flen = strlen(function);
+	va_start(args, fmt);
+	vaf.fmt = fmt;
+	vaf.va = &args;
+	if (sb)
+		pr_warn("(device %s): %s(): %pV\n",
+			sb->s_id, flen ? function : "", &vaf);
+	else
+		pr_warn("%s(): %pV\n", flen ? function : "", &vaf);
+	va_end(args);
+}
+
+/**
+ * __ntfs_error - output an error to the syslog
+ * @function:	name of function outputting the error
+ * @sb:		super block of mounted ntfs filesystem
+ * @fmt:	error string containing format specifications
+ * @...:	a variable number of arguments specified in @fmt
+ *
+ * Outputs an error to the syslog for the mounted ntfs filesystem described
+ * by @sb.
+ *
+ * @fmt and the corresponding @... is printf style format string containing
+ * the error string and the corresponding format arguments, respectively.
+ *
+ * @function is the name of the function from which __ntfs_error is being
+ * called.
+ *
+ * Note, you should be using debug.h::ntfs_error(@sb, @fmt, @...) instead
+ * as this provides the @function parameter automatically.
+ */
+void __ntfs_error(const char *function, const struct super_block *sb,
+		const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list args;
+	int flen = 0;
+
+#ifndef DEBUG
+	if (!printk_ratelimit())
+		return;
+#endif
+	if (function)
+		flen = strlen(function);
+	va_start(args, fmt);
+	vaf.fmt = fmt;
+	vaf.va = &args;
+	if (sb)
+		pr_err("(device %s): %s(): %pV\n",
+		       sb->s_id, flen ? function : "", &vaf);
+	else
+		pr_err("%s(): %pV\n", flen ? function : "", &vaf);
+	va_end(args);
+}
+
+#ifdef DEBUG
+
+/* If 1, output debug messages, and if 0, don't. */
+int debug_msgs = 0;
+
+void __ntfs_debug(const char *file, int line, const char *function,
+		const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list args;
+	int flen = 0;
+
+	if (!debug_msgs)
+		return;
+	if (function)
+		flen = strlen(function);
+	va_start(args, fmt);
+	vaf.fmt = fmt;
+	vaf.va = &args;
+	pr_debug("(%s, %d): %s(): %pV", file, line, flen ? function : "", &vaf);
+	va_end(args);
+}
+
+/* Dump a runlist. Caller has to provide synchronisation for @rl. */
+void ntfs_debug_dump_runlist(const runlist_element *rl)
+{
+	int i;
+	const char *lcn_str[5] = { "LCN_HOLE         ", "LCN_RL_NOT_MAPPED",
+				   "LCN_ENOENT       ", "LCN_unknown      " };
+
+	if (!debug_msgs)
+		return;
+	pr_debug("Dumping runlist (values in hex):\n");
+	if (!rl) {
+		pr_debug("Run list not present.\n");
+		return;
+	}
+	pr_debug("VCN              LCN               Run length\n");
+	for (i = 0; ; i++) {
+		LCN lcn = (rl + i)->lcn;
+
+		if (lcn < (LCN)0) {
+			int index = -lcn - 1;
+
+			if (index > -LCN_ENOENT - 1)
+				index = 3;
+			pr_debug("%-16Lx %s %-16Lx%s\n",
+					(long long)(rl + i)->vcn, lcn_str[index],
+					(long long)(rl + i)->length,
+					(rl + i)->length ? "" :
+						" (runlist end)");
+		} else
+			pr_debug("%-16Lx %-16Lx  %-16Lx%s\n",
+					(long long)(rl + i)->vcn,
+					(long long)(rl + i)->lcn,
+					(long long)(rl + i)->length,
+					(rl + i)->length ? "" :
+						" (runlist end)");
+		if (!(rl + i)->length)
+			break;
+	}
+}
+
+#endif
diff --git a/fs/ntfs/debug.h b/fs/ntfs/debug.h
new file mode 100644
index 000000000..6fdef388f
--- /dev/null
+++ b/fs/ntfs/debug.h
@@ -0,0 +1,57 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * debug.h - NTFS kernel debug support. Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_DEBUG_H
+#define _LINUX_NTFS_DEBUG_H
+
+#include <linux/fs.h>
+
+#include "runlist.h"
+
+#ifdef DEBUG
+
+extern int debug_msgs;
+
+extern __printf(4, 5)
+void __ntfs_debug(const char *file, int line, const char *function,
+		  const char *format, ...);
+/**
+ * ntfs_debug - write a debug level message to syslog
+ * @f:		a printf format string containing the message
+ * @...:	the variables to substitute into @f
+ *
+ * ntfs_debug() writes a DEBUG level message to the syslog but only if the
+ * driver was compiled with -DDEBUG. Otherwise, the call turns into a NOP.
+ */
+#define ntfs_debug(f, a...)						\
+	__ntfs_debug(__FILE__, __LINE__, __func__, f, ##a)
+
+extern void ntfs_debug_dump_runlist(const runlist_element *rl);
+
+#else	/* !DEBUG */
+
+#define ntfs_debug(fmt, ...)						\
+do {									\
+	if (0)								\
+		no_printk(fmt, ##__VA_ARGS__);				\
+} while (0)
+
+#define ntfs_debug_dump_runlist(rl)	do {} while (0)
+
+#endif	/* !DEBUG */
+
+extern  __printf(3, 4)
+void __ntfs_warning(const char *function, const struct super_block *sb,
+		    const char *fmt, ...);
+#define ntfs_warning(sb, f, a...)	__ntfs_warning(__func__, sb, f, ##a)
+
+extern  __printf(3, 4)
+void __ntfs_error(const char *function, const struct super_block *sb,
+		  const char *fmt, ...);
+#define ntfs_error(sb, f, a...)		__ntfs_error(__func__, sb, f, ##a)
+
+#endif /* _LINUX_NTFS_DEBUG_H */
diff --git a/fs/ntfs/dir.c b/fs/ntfs/dir.c
new file mode 100644
index 000000000..cd96083a1
--- /dev/null
+++ b/fs/ntfs/dir.c
@@ -0,0 +1,1538 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/**
+ * dir.c - NTFS kernel directory operations. Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2007 Anton Altaparmakov
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#include <linux/buffer_head.h>
+#include <linux/slab.h>
+#include <linux/blkdev.h>
+
+#include "dir.h"
+#include "aops.h"
+#include "attrib.h"
+#include "mft.h"
+#include "debug.h"
+#include "ntfs.h"
+
+/**
+ * The little endian Unicode string $I30 as a global constant.
+ */
+ntfschar I30[5] = { cpu_to_le16('$'), cpu_to_le16('I'),
+		cpu_to_le16('3'),	cpu_to_le16('0'), 0 };
+
+/**
+ * ntfs_lookup_inode_by_name - find an inode in a directory given its name
+ * @dir_ni:	ntfs inode of the directory in which to search for the name
+ * @uname:	Unicode name for which to search in the directory
+ * @uname_len:	length of the name @uname in Unicode characters
+ * @res:	return the found file name if necessary (see below)
+ *
+ * Look for an inode with name @uname in the directory with inode @dir_ni.
+ * ntfs_lookup_inode_by_name() walks the contents of the directory looking for
+ * the Unicode name. If the name is found in the directory, the corresponding
+ * inode number (>= 0) is returned as a mft reference in cpu format, i.e. it
+ * is a 64-bit number containing the sequence number.
+ *
+ * On error, a negative value is returned corresponding to the error code. In
+ * particular if the inode is not found -ENOENT is returned. Note that you
+ * can't just check the return value for being negative, you have to check the
+ * inode number for being negative which you can extract using MREC(return
+ * value).
+ *
+ * Note, @uname_len does not include the (optional) terminating NULL character.
+ *
+ * Note, we look for a case sensitive match first but we also look for a case
+ * insensitive match at the same time. If we find a case insensitive match, we
+ * save that for the case that we don't find an exact match, where we return
+ * the case insensitive match and setup @res (which we allocate!) with the mft
+ * reference, the file name type, length and with a copy of the little endian
+ * Unicode file name itself. If we match a file name which is in the DOS name
+ * space, we only return the mft reference and file name type in @res.
+ * ntfs_lookup() then uses this to find the long file name in the inode itself.
+ * This is to avoid polluting the dcache with short file names. We want them to
+ * work but we don't care for how quickly one can access them. This also fixes
+ * the dcache aliasing issues.
+ *
+ * Locking:  - Caller must hold i_mutex on the directory.
+ *	     - Each page cache page in the index allocation mapping must be
+ *	       locked whilst being accessed otherwise we may find a corrupt
+ *	       page due to it being under ->writepage at the moment which
+ *	       applies the mst protection fixups before writing out and then
+ *	       removes them again after the write is complete after which it 
+ *	       unlocks the page.
+ */
+MFT_REF ntfs_lookup_inode_by_name(ntfs_inode *dir_ni, const ntfschar *uname,
+		const int uname_len, ntfs_name **res)
+{
+	ntfs_volume *vol = dir_ni->vol;
+	struct super_block *sb = vol->sb;
+	MFT_RECORD *m;
+	INDEX_ROOT *ir;
+	INDEX_ENTRY *ie;
+	INDEX_ALLOCATION *ia;
+	u8 *index_end;
+	u64 mref;
+	ntfs_attr_search_ctx *ctx;
+	int err, rc;
+	VCN vcn, old_vcn;
+	struct address_space *ia_mapping;
+	struct page *page;
+	u8 *kaddr;
+	ntfs_name *name = NULL;
+
+	BUG_ON(!S_ISDIR(VFS_I(dir_ni)->i_mode));
+	BUG_ON(NInoAttr(dir_ni));
+	/* Get hold of the mft record for the directory. */
+	m = map_mft_record(dir_ni);
+	if (IS_ERR(m)) {
+		ntfs_error(sb, "map_mft_record() failed with error code %ld.",
+				-PTR_ERR(m));
+		return ERR_MREF(PTR_ERR(m));
+	}
+	ctx = ntfs_attr_get_search_ctx(dir_ni, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	/* Find the index root attribute in the mft record. */
+	err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE, 0, NULL,
+			0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT) {
+			ntfs_error(sb, "Index root attribute missing in "
+					"directory inode 0x%lx.",
+					dir_ni->mft_no);
+			err = -EIO;
+		}
+		goto err_out;
+	}
+	/* Get to the index root value (it's been verified in read_inode). */
+	ir = (INDEX_ROOT*)((u8*)ctx->attr +
+			le16_to_cpu(ctx->attr->data.resident.value_offset));
+	index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
+	/* The first index entry. */
+	ie = (INDEX_ENTRY*)((u8*)&ir->index +
+			le32_to_cpu(ir->index.entries_offset));
+	/*
+	 * Loop until we exceed valid memory (corruption case) or until we
+	 * reach the last entry.
+	 */
+	for (;; ie = (INDEX_ENTRY*)((u8*)ie + le16_to_cpu(ie->length))) {
+		/* Bounds checks. */
+		if ((u8*)ie < (u8*)ctx->mrec || (u8*)ie +
+				sizeof(INDEX_ENTRY_HEADER) > index_end ||
+				(u8*)ie + le16_to_cpu(ie->key_length) >
+				index_end)
+			goto dir_err_out;
+		/*
+		 * The last entry cannot contain a name. 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;
+		/*
+		 * We perform a case sensitive comparison and if that matches
+		 * we are done and return the mft reference of the inode (i.e.
+		 * the inode number together with the sequence number for
+		 * consistency checking). We convert it to cpu format before
+		 * returning.
+		 */
+		if (ntfs_are_names_equal(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length,
+				CASE_SENSITIVE, vol->upcase, vol->upcase_len)) {
+found_it:
+			/*
+			 * We have a perfect match, so we don't need to care
+			 * about having matched imperfectly before, so we can
+			 * free name and set *res to NULL.
+			 * However, if the perfect match is a short file name,
+			 * we need to signal this through *res, so that
+			 * ntfs_lookup() can fix dcache aliasing issues.
+			 * As an optimization we just reuse an existing
+			 * allocation of *res.
+			 */
+			if (ie->key.file_name.file_name_type == FILE_NAME_DOS) {
+				if (!name) {
+					name = kmalloc(sizeof(ntfs_name),
+							GFP_NOFS);
+					if (!name) {
+						err = -ENOMEM;
+						goto err_out;
+					}
+				}
+				name->mref = le64_to_cpu(
+						ie->data.dir.indexed_file);
+				name->type = FILE_NAME_DOS;
+				name->len = 0;
+				*res = name;
+			} else {
+				kfree(name);
+				*res = NULL;
+			}
+			mref = le64_to_cpu(ie->data.dir.indexed_file);
+			ntfs_attr_put_search_ctx(ctx);
+			unmap_mft_record(dir_ni);
+			return mref;
+		}
+		/*
+		 * For a case insensitive mount, we also perform a case
+		 * insensitive comparison (provided the file name is not in the
+		 * POSIX namespace). If the comparison matches, and the name is
+		 * in the WIN32 namespace, we cache the filename in *res so
+		 * that the caller, ntfs_lookup(), can work on it. If the
+		 * comparison matches, and the name is in the DOS namespace, we
+		 * only cache the mft reference and the file name type (we set
+		 * the name length to zero for simplicity).
+		 */
+		if (!NVolCaseSensitive(vol) &&
+				ie->key.file_name.file_name_type &&
+				ntfs_are_names_equal(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length,
+				IGNORE_CASE, vol->upcase, vol->upcase_len)) {
+			int name_size = sizeof(ntfs_name);
+			u8 type = ie->key.file_name.file_name_type;
+			u8 len = ie->key.file_name.file_name_length;
+
+			/* Only one case insensitive matching name allowed. */
+			if (name) {
+				ntfs_error(sb, "Found already allocated name "
+						"in phase 1. Please run chkdsk "
+						"and if that doesn't find any "
+						"errors please report you saw "
+						"this message to "
+						"linux-ntfs-dev@lists."
+						"sourceforge.net.");
+				goto dir_err_out;
+			}
+
+			if (type != FILE_NAME_DOS)
+				name_size += len * sizeof(ntfschar);
+			name = kmalloc(name_size, GFP_NOFS);
+			if (!name) {
+				err = -ENOMEM;
+				goto err_out;
+			}
+			name->mref = le64_to_cpu(ie->data.dir.indexed_file);
+			name->type = type;
+			if (type != FILE_NAME_DOS) {
+				name->len = len;
+				memcpy(name->name, ie->key.file_name.file_name,
+						len * sizeof(ntfschar));
+			} else
+				name->len = 0;
+			*res = name;
+		}
+		/*
+		 * Not a perfect match, need to do full blown collation so we
+		 * know which way in the B+tree we have to go.
+		 */
+		rc = ntfs_collate_names(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, 1,
+				IGNORE_CASE, vol->upcase, vol->upcase_len);
+		/*
+		 * If uname collates before the name of the current entry, there
+		 * is definitely no such name in this index but we might need to
+		 * descend into the B+tree so we just break out of the loop.
+		 */
+		if (rc == -1)
+			break;
+		/* The names are not equal, continue the search. */
+		if (rc)
+			continue;
+		/*
+		 * Names match with case insensitive comparison, now try the
+		 * case sensitive comparison, which is required for proper
+		 * collation.
+		 */
+		rc = ntfs_collate_names(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, 1,
+				CASE_SENSITIVE, vol->upcase, vol->upcase_len);
+		if (rc == -1)
+			break;
+		if (rc)
+			continue;
+		/*
+		 * Perfect match, this will never happen as the
+		 * ntfs_are_names_equal() call will have gotten a match but we
+		 * still treat it correctly.
+		 */
+		goto found_it;
+	}
+	/*
+	 * We have finished with this index without success. Check for the
+	 * presence of a child node and if not present return -ENOENT, unless
+	 * we have got a matching name cached in name in which case return the
+	 * mft reference associated with it.
+	 */
+	if (!(ie->flags & INDEX_ENTRY_NODE)) {
+		if (name) {
+			ntfs_attr_put_search_ctx(ctx);
+			unmap_mft_record(dir_ni);
+			return name->mref;
+		}
+		ntfs_debug("Entry not found.");
+		err = -ENOENT;
+		goto err_out;
+	} /* Child node present, descend into it. */
+	/* Consistency check: Verify that an index allocation exists. */
+	if (!NInoIndexAllocPresent(dir_ni)) {
+		ntfs_error(sb, "No index allocation attribute but index entry "
+				"requires one. Directory inode 0x%lx is "
+				"corrupt or driver bug.", dir_ni->mft_no);
+		goto err_out;
+	}
+	/* Get the starting vcn of the index_block holding the child node. */
+	vcn = sle64_to_cpup((sle64*)((u8*)ie + le16_to_cpu(ie->length) - 8));
+	ia_mapping = VFS_I(dir_ni)->i_mapping;
+	/*
+	 * We are done with the index root and the mft record. Release them,
+	 * otherwise we deadlock with ntfs_map_page().
+	 */
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(dir_ni);
+	m = NULL;
+	ctx = NULL;
+descend_into_child_node:
+	/*
+	 * Convert vcn to index into the index allocation attribute in units
+	 * of PAGE_SIZE and map the page cache page, reading it from
+	 * disk if necessary.
+	 */
+	page = ntfs_map_page(ia_mapping, vcn <<
+			dir_ni->itype.index.vcn_size_bits >> PAGE_SHIFT);
+	if (IS_ERR(page)) {
+		ntfs_error(sb, "Failed to map directory index page, error %ld.",
+				-PTR_ERR(page));
+		err = PTR_ERR(page);
+		goto err_out;
+	}
+	lock_page(page);
+	kaddr = (u8*)page_address(page);
+fast_descend_into_child_node:
+	/* Get to the index allocation block. */
+	ia = (INDEX_ALLOCATION*)(kaddr + ((vcn <<
+			dir_ni->itype.index.vcn_size_bits) & ~PAGE_MASK));
+	/* Bounds checks. */
+	if ((u8*)ia < kaddr || (u8*)ia > kaddr + PAGE_SIZE) {
+		ntfs_error(sb, "Out of bounds check failed. Corrupt directory "
+				"inode 0x%lx or driver bug.", dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	/* Catch multi sector transfer fixup errors. */
+	if (unlikely(!ntfs_is_indx_record(ia->magic))) {
+		ntfs_error(sb, "Directory index record with vcn 0x%llx is "
+				"corrupt.  Corrupt inode 0x%lx.  Run chkdsk.",
+				(unsigned long long)vcn, dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	if (sle64_to_cpu(ia->index_block_vcn) != vcn) {
+		ntfs_error(sb, "Actual VCN (0x%llx) of index buffer is "
+				"different from expected VCN (0x%llx). "
+				"Directory inode 0x%lx is corrupt or driver "
+				"bug.", (unsigned long long)
+				sle64_to_cpu(ia->index_block_vcn),
+				(unsigned long long)vcn, dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	if (le32_to_cpu(ia->index.allocated_size) + 0x18 !=
+			dir_ni->itype.index.block_size) {
+		ntfs_error(sb, "Index buffer (VCN 0x%llx) of directory inode "
+				"0x%lx has a size (%u) differing from the "
+				"directory specified size (%u). Directory "
+				"inode is corrupt or driver bug.",
+				(unsigned long long)vcn, dir_ni->mft_no,
+				le32_to_cpu(ia->index.allocated_size) + 0x18,
+				dir_ni->itype.index.block_size);
+		goto unm_err_out;
+	}
+	index_end = (u8*)ia + dir_ni->itype.index.block_size;
+	if (index_end > kaddr + PAGE_SIZE) {
+		ntfs_error(sb, "Index buffer (VCN 0x%llx) of directory inode "
+				"0x%lx crosses page boundary. Impossible! "
+				"Cannot access! This is probably a bug in the "
+				"driver.", (unsigned long long)vcn,
+				dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	index_end = (u8*)&ia->index + le32_to_cpu(ia->index.index_length);
+	if (index_end > (u8*)ia + dir_ni->itype.index.block_size) {
+		ntfs_error(sb, "Size of index buffer (VCN 0x%llx) of directory "
+				"inode 0x%lx exceeds maximum size.",
+				(unsigned long long)vcn, dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	/* The first index entry. */
+	ie = (INDEX_ENTRY*)((u8*)&ia->index +
+			le32_to_cpu(ia->index.entries_offset));
+	/*
+	 * Iterate similar to above big loop but applied to index buffer, thus
+	 * loop until we exceed valid memory (corruption case) or until we
+	 * reach the last entry.
+	 */
+	for (;; ie = (INDEX_ENTRY*)((u8*)ie + le16_to_cpu(ie->length))) {
+		/* Bounds check. */
+		if ((u8*)ie < (u8*)ia || (u8*)ie +
+				sizeof(INDEX_ENTRY_HEADER) > index_end ||
+				(u8*)ie + le16_to_cpu(ie->key_length) >
+				index_end) {
+			ntfs_error(sb, "Index entry out of bounds in "
+					"directory inode 0x%lx.",
+					dir_ni->mft_no);
+			goto unm_err_out;
+		}
+		/*
+		 * The last entry cannot contain a name. 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;
+		/*
+		 * We perform a case sensitive comparison and if that matches
+		 * we are done and return the mft reference of the inode (i.e.
+		 * the inode number together with the sequence number for
+		 * consistency checking). We convert it to cpu format before
+		 * returning.
+		 */
+		if (ntfs_are_names_equal(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length,
+				CASE_SENSITIVE, vol->upcase, vol->upcase_len)) {
+found_it2:
+			/*
+			 * We have a perfect match, so we don't need to care
+			 * about having matched imperfectly before, so we can
+			 * free name and set *res to NULL.
+			 * However, if the perfect match is a short file name,
+			 * we need to signal this through *res, so that
+			 * ntfs_lookup() can fix dcache aliasing issues.
+			 * As an optimization we just reuse an existing
+			 * allocation of *res.
+			 */
+			if (ie->key.file_name.file_name_type == FILE_NAME_DOS) {
+				if (!name) {
+					name = kmalloc(sizeof(ntfs_name),
+							GFP_NOFS);
+					if (!name) {
+						err = -ENOMEM;
+						goto unm_err_out;
+					}
+				}
+				name->mref = le64_to_cpu(
+						ie->data.dir.indexed_file);
+				name->type = FILE_NAME_DOS;
+				name->len = 0;
+				*res = name;
+			} else {
+				kfree(name);
+				*res = NULL;
+			}
+			mref = le64_to_cpu(ie->data.dir.indexed_file);
+			unlock_page(page);
+			ntfs_unmap_page(page);
+			return mref;
+		}
+		/*
+		 * For a case insensitive mount, we also perform a case
+		 * insensitive comparison (provided the file name is not in the
+		 * POSIX namespace). If the comparison matches, and the name is
+		 * in the WIN32 namespace, we cache the filename in *res so
+		 * that the caller, ntfs_lookup(), can work on it. If the
+		 * comparison matches, and the name is in the DOS namespace, we
+		 * only cache the mft reference and the file name type (we set
+		 * the name length to zero for simplicity).
+		 */
+		if (!NVolCaseSensitive(vol) &&
+				ie->key.file_name.file_name_type &&
+				ntfs_are_names_equal(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length,
+				IGNORE_CASE, vol->upcase, vol->upcase_len)) {
+			int name_size = sizeof(ntfs_name);
+			u8 type = ie->key.file_name.file_name_type;
+			u8 len = ie->key.file_name.file_name_length;
+
+			/* Only one case insensitive matching name allowed. */
+			if (name) {
+				ntfs_error(sb, "Found already allocated name "
+						"in phase 2. Please run chkdsk "
+						"and if that doesn't find any "
+						"errors please report you saw "
+						"this message to "
+						"linux-ntfs-dev@lists."
+						"sourceforge.net.");
+				unlock_page(page);
+				ntfs_unmap_page(page);
+				goto dir_err_out;
+			}
+
+			if (type != FILE_NAME_DOS)
+				name_size += len * sizeof(ntfschar);
+			name = kmalloc(name_size, GFP_NOFS);
+			if (!name) {
+				err = -ENOMEM;
+				goto unm_err_out;
+			}
+			name->mref = le64_to_cpu(ie->data.dir.indexed_file);
+			name->type = type;
+			if (type != FILE_NAME_DOS) {
+				name->len = len;
+				memcpy(name->name, ie->key.file_name.file_name,
+						len * sizeof(ntfschar));
+			} else
+				name->len = 0;
+			*res = name;
+		}
+		/*
+		 * Not a perfect match, need to do full blown collation so we
+		 * know which way in the B+tree we have to go.
+		 */
+		rc = ntfs_collate_names(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, 1,
+				IGNORE_CASE, vol->upcase, vol->upcase_len);
+		/*
+		 * If uname collates before the name of the current entry, there
+		 * is definitely no such name in this index but we might need to
+		 * descend into the B+tree so we just break out of the loop.
+		 */
+		if (rc == -1)
+			break;
+		/* The names are not equal, continue the search. */
+		if (rc)
+			continue;
+		/*
+		 * Names match with case insensitive comparison, now try the
+		 * case sensitive comparison, which is required for proper
+		 * collation.
+		 */
+		rc = ntfs_collate_names(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, 1,
+				CASE_SENSITIVE, vol->upcase, vol->upcase_len);
+		if (rc == -1)
+			break;
+		if (rc)
+			continue;
+		/*
+		 * Perfect match, this will never happen as the
+		 * ntfs_are_names_equal() call will have gotten a match but we
+		 * still treat it correctly.
+		 */
+		goto found_it2;
+	}
+	/*
+	 * We have finished with this index buffer without success. Check for
+	 * the presence of a child node.
+	 */
+	if (ie->flags & INDEX_ENTRY_NODE) {
+		if ((ia->index.flags & NODE_MASK) == LEAF_NODE) {
+			ntfs_error(sb, "Index entry with child node found in "
+					"a leaf node in directory inode 0x%lx.",
+					dir_ni->mft_no);
+			goto unm_err_out;
+		}
+		/* Child node present, descend into it. */
+		old_vcn = vcn;
+		vcn = sle64_to_cpup((sle64*)((u8*)ie +
+				le16_to_cpu(ie->length) - 8));
+		if (vcn >= 0) {
+			/* If vcn is in the same page cache page as old_vcn we
+			 * recycle the mapped page. */
+			if (old_vcn << vol->cluster_size_bits >>
+					PAGE_SHIFT == vcn <<
+					vol->cluster_size_bits >>
+					PAGE_SHIFT)
+				goto fast_descend_into_child_node;
+			unlock_page(page);
+			ntfs_unmap_page(page);
+			goto descend_into_child_node;
+		}
+		ntfs_error(sb, "Negative child node vcn in directory inode "
+				"0x%lx.", dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	/*
+	 * No child node present, return -ENOENT, unless we have got a matching
+	 * name cached in name in which case return the mft reference
+	 * associated with it.
+	 */
+	if (name) {
+		unlock_page(page);
+		ntfs_unmap_page(page);
+		return name->mref;
+	}
+	ntfs_debug("Entry not found.");
+	err = -ENOENT;
+unm_err_out:
+	unlock_page(page);
+	ntfs_unmap_page(page);
+err_out:
+	if (!err)
+		err = -EIO;
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(dir_ni);
+	if (name) {
+		kfree(name);
+		*res = NULL;
+	}
+	return ERR_MREF(err);
+dir_err_out:
+	ntfs_error(sb, "Corrupt directory.  Aborting lookup.");
+	goto err_out;
+}
+
+#if 0
+
+// TODO: (AIA)
+// The algorithm embedded in this code will be required for the time when we
+// want to support adding of entries to directories, where we require correct
+// collation of file names in order not to cause corruption of the filesystem.
+
+/**
+ * ntfs_lookup_inode_by_name - find an inode in a directory given its name
+ * @dir_ni:	ntfs inode of the directory in which to search for the name
+ * @uname:	Unicode name for which to search in the directory
+ * @uname_len:	length of the name @uname in Unicode characters
+ *
+ * Look for an inode with name @uname in the directory with inode @dir_ni.
+ * ntfs_lookup_inode_by_name() walks the contents of the directory looking for
+ * the Unicode name. If the name is found in the directory, the corresponding
+ * inode number (>= 0) is returned as a mft reference in cpu format, i.e. it
+ * is a 64-bit number containing the sequence number.
+ *
+ * On error, a negative value is returned corresponding to the error code. In
+ * particular if the inode is not found -ENOENT is returned. Note that you
+ * can't just check the return value for being negative, you have to check the
+ * inode number for being negative which you can extract using MREC(return
+ * value).
+ *
+ * Note, @uname_len does not include the (optional) terminating NULL character.
+ */
+u64 ntfs_lookup_inode_by_name(ntfs_inode *dir_ni, const ntfschar *uname,
+		const int uname_len)
+{
+	ntfs_volume *vol = dir_ni->vol;
+	struct super_block *sb = vol->sb;
+	MFT_RECORD *m;
+	INDEX_ROOT *ir;
+	INDEX_ENTRY *ie;
+	INDEX_ALLOCATION *ia;
+	u8 *index_end;
+	u64 mref;
+	ntfs_attr_search_ctx *ctx;
+	int err, rc;
+	IGNORE_CASE_BOOL ic;
+	VCN vcn, old_vcn;
+	struct address_space *ia_mapping;
+	struct page *page;
+	u8 *kaddr;
+
+	/* Get hold of the mft record for the directory. */
+	m = map_mft_record(dir_ni);
+	if (IS_ERR(m)) {
+		ntfs_error(sb, "map_mft_record() failed with error code %ld.",
+				-PTR_ERR(m));
+		return ERR_MREF(PTR_ERR(m));
+	}
+	ctx = ntfs_attr_get_search_ctx(dir_ni, m);
+	if (!ctx) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	/* Find the index root attribute in the mft record. */
+	err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE, 0, NULL,
+			0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT) {
+			ntfs_error(sb, "Index root attribute missing in "
+					"directory inode 0x%lx.",
+					dir_ni->mft_no);
+			err = -EIO;
+		}
+		goto err_out;
+	}
+	/* Get to the index root value (it's been verified in read_inode). */
+	ir = (INDEX_ROOT*)((u8*)ctx->attr +
+			le16_to_cpu(ctx->attr->data.resident.value_offset));
+	index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
+	/* The first index entry. */
+	ie = (INDEX_ENTRY*)((u8*)&ir->index +
+			le32_to_cpu(ir->index.entries_offset));
+	/*
+	 * Loop until we exceed valid memory (corruption case) or until we
+	 * reach the last entry.
+	 */
+	for (;; ie = (INDEX_ENTRY*)((u8*)ie + le16_to_cpu(ie->length))) {
+		/* Bounds checks. */
+		if ((u8*)ie < (u8*)ctx->mrec || (u8*)ie +
+				sizeof(INDEX_ENTRY_HEADER) > index_end ||
+				(u8*)ie + le16_to_cpu(ie->key_length) >
+				index_end)
+			goto dir_err_out;
+		/*
+		 * The last entry cannot contain a name. 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 the current entry has a name type of POSIX, the name is
+		 * case sensitive and not otherwise. This has the effect of us
+		 * not being able to access any POSIX file names which collate
+		 * after the non-POSIX one when they only differ in case, but
+		 * anyone doing screwy stuff like that deserves to burn in
+		 * hell... Doing that kind of stuff on NT4 actually causes
+		 * corruption on the partition even when using SP6a and Linux
+		 * is not involved at all.
+		 */
+		ic = ie->key.file_name.file_name_type ? IGNORE_CASE :
+				CASE_SENSITIVE;
+		/*
+		 * If the names match perfectly, we are done and return the
+		 * mft reference of the inode (i.e. the inode number together
+		 * with the sequence number for consistency checking. We
+		 * convert it to cpu format before returning.
+		 */
+		if (ntfs_are_names_equal(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, ic,
+				vol->upcase, vol->upcase_len)) {
+found_it:
+			mref = le64_to_cpu(ie->data.dir.indexed_file);
+			ntfs_attr_put_search_ctx(ctx);
+			unmap_mft_record(dir_ni);
+			return mref;
+		}
+		/*
+		 * Not a perfect match, need to do full blown collation so we
+		 * know which way in the B+tree we have to go.
+		 */
+		rc = ntfs_collate_names(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, 1,
+				IGNORE_CASE, vol->upcase, vol->upcase_len);
+		/*
+		 * If uname collates before the name of the current entry, there
+		 * is definitely no such name in this index but we might need to
+		 * descend into the B+tree so we just break out of the loop.
+		 */
+		if (rc == -1)
+			break;
+		/* The names are not equal, continue the search. */
+		if (rc)
+			continue;
+		/*
+		 * Names match with case insensitive comparison, now try the
+		 * case sensitive comparison, which is required for proper
+		 * collation.
+		 */
+		rc = ntfs_collate_names(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, 1,
+				CASE_SENSITIVE, vol->upcase, vol->upcase_len);
+		if (rc == -1)
+			break;
+		if (rc)
+			continue;
+		/*
+		 * Perfect match, this will never happen as the
+		 * ntfs_are_names_equal() call will have gotten a match but we
+		 * still treat it correctly.
+		 */
+		goto found_it;
+	}
+	/*
+	 * We have finished with this index without success. Check for the
+	 * presence of a child node.
+	 */
+	if (!(ie->flags & INDEX_ENTRY_NODE)) {
+		/* No child node, return -ENOENT. */
+		err = -ENOENT;
+		goto err_out;
+	} /* Child node present, descend into it. */
+	/* Consistency check: Verify that an index allocation exists. */
+	if (!NInoIndexAllocPresent(dir_ni)) {
+		ntfs_error(sb, "No index allocation attribute but index entry "
+				"requires one. Directory inode 0x%lx is "
+				"corrupt or driver bug.", dir_ni->mft_no);
+		goto err_out;
+	}
+	/* Get the starting vcn of the index_block holding the child node. */
+	vcn = sle64_to_cpup((u8*)ie + le16_to_cpu(ie->length) - 8);
+	ia_mapping = VFS_I(dir_ni)->i_mapping;
+	/*
+	 * We are done with the index root and the mft record. Release them,
+	 * otherwise we deadlock with ntfs_map_page().
+	 */
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(dir_ni);
+	m = NULL;
+	ctx = NULL;
+descend_into_child_node:
+	/*
+	 * Convert vcn to index into the index allocation attribute in units
+	 * of PAGE_SIZE and map the page cache page, reading it from
+	 * disk if necessary.
+	 */
+	page = ntfs_map_page(ia_mapping, vcn <<
+			dir_ni->itype.index.vcn_size_bits >> PAGE_SHIFT);
+	if (IS_ERR(page)) {
+		ntfs_error(sb, "Failed to map directory index page, error %ld.",
+				-PTR_ERR(page));
+		err = PTR_ERR(page);
+		goto err_out;
+	}
+	lock_page(page);
+	kaddr = (u8*)page_address(page);
+fast_descend_into_child_node:
+	/* Get to the index allocation block. */
+	ia = (INDEX_ALLOCATION*)(kaddr + ((vcn <<
+			dir_ni->itype.index.vcn_size_bits) & ~PAGE_MASK));
+	/* Bounds checks. */
+	if ((u8*)ia < kaddr || (u8*)ia > kaddr + PAGE_SIZE) {
+		ntfs_error(sb, "Out of bounds check failed. Corrupt directory "
+				"inode 0x%lx or driver bug.", dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	/* Catch multi sector transfer fixup errors. */
+	if (unlikely(!ntfs_is_indx_record(ia->magic))) {
+		ntfs_error(sb, "Directory index record with vcn 0x%llx is "
+				"corrupt.  Corrupt inode 0x%lx.  Run chkdsk.",
+				(unsigned long long)vcn, dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	if (sle64_to_cpu(ia->index_block_vcn) != vcn) {
+		ntfs_error(sb, "Actual VCN (0x%llx) of index buffer is "
+				"different from expected VCN (0x%llx). "
+				"Directory inode 0x%lx is corrupt or driver "
+				"bug.", (unsigned long long)
+				sle64_to_cpu(ia->index_block_vcn),
+				(unsigned long long)vcn, dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	if (le32_to_cpu(ia->index.allocated_size) + 0x18 !=
+			dir_ni->itype.index.block_size) {
+		ntfs_error(sb, "Index buffer (VCN 0x%llx) of directory inode "
+				"0x%lx has a size (%u) differing from the "
+				"directory specified size (%u). Directory "
+				"inode is corrupt or driver bug.",
+				(unsigned long long)vcn, dir_ni->mft_no,
+				le32_to_cpu(ia->index.allocated_size) + 0x18,
+				dir_ni->itype.index.block_size);
+		goto unm_err_out;
+	}
+	index_end = (u8*)ia + dir_ni->itype.index.block_size;
+	if (index_end > kaddr + PAGE_SIZE) {
+		ntfs_error(sb, "Index buffer (VCN 0x%llx) of directory inode "
+				"0x%lx crosses page boundary. Impossible! "
+				"Cannot access! This is probably a bug in the "
+				"driver.", (unsigned long long)vcn,
+				dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	index_end = (u8*)&ia->index + le32_to_cpu(ia->index.index_length);
+	if (index_end > (u8*)ia + dir_ni->itype.index.block_size) {
+		ntfs_error(sb, "Size of index buffer (VCN 0x%llx) of directory "
+				"inode 0x%lx exceeds maximum size.",
+				(unsigned long long)vcn, dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	/* The first index entry. */
+	ie = (INDEX_ENTRY*)((u8*)&ia->index +
+			le32_to_cpu(ia->index.entries_offset));
+	/*
+	 * Iterate similar to above big loop but applied to index buffer, thus
+	 * loop until we exceed valid memory (corruption case) or until we
+	 * reach the last entry.
+	 */
+	for (;; ie = (INDEX_ENTRY*)((u8*)ie + le16_to_cpu(ie->length))) {
+		/* Bounds check. */
+		if ((u8*)ie < (u8*)ia || (u8*)ie +
+				sizeof(INDEX_ENTRY_HEADER) > index_end ||
+				(u8*)ie + le16_to_cpu(ie->key_length) >
+				index_end) {
+			ntfs_error(sb, "Index entry out of bounds in "
+					"directory inode 0x%lx.",
+					dir_ni->mft_no);
+			goto unm_err_out;
+		}
+		/*
+		 * The last entry cannot contain a name. 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 the current entry has a name type of POSIX, the name is
+		 * case sensitive and not otherwise. This has the effect of us
+		 * not being able to access any POSIX file names which collate
+		 * after the non-POSIX one when they only differ in case, but
+		 * anyone doing screwy stuff like that deserves to burn in
+		 * hell... Doing that kind of stuff on NT4 actually causes
+		 * corruption on the partition even when using SP6a and Linux
+		 * is not involved at all.
+		 */
+		ic = ie->key.file_name.file_name_type ? IGNORE_CASE :
+				CASE_SENSITIVE;
+		/*
+		 * If the names match perfectly, we are done and return the
+		 * mft reference of the inode (i.e. the inode number together
+		 * with the sequence number for consistency checking. We
+		 * convert it to cpu format before returning.
+		 */
+		if (ntfs_are_names_equal(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, ic,
+				vol->upcase, vol->upcase_len)) {
+found_it2:
+			mref = le64_to_cpu(ie->data.dir.indexed_file);
+			unlock_page(page);
+			ntfs_unmap_page(page);
+			return mref;
+		}
+		/*
+		 * Not a perfect match, need to do full blown collation so we
+		 * know which way in the B+tree we have to go.
+		 */
+		rc = ntfs_collate_names(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, 1,
+				IGNORE_CASE, vol->upcase, vol->upcase_len);
+		/*
+		 * If uname collates before the name of the current entry, there
+		 * is definitely no such name in this index but we might need to
+		 * descend into the B+tree so we just break out of the loop.
+		 */
+		if (rc == -1)
+			break;
+		/* The names are not equal, continue the search. */
+		if (rc)
+			continue;
+		/*
+		 * Names match with case insensitive comparison, now try the
+		 * case sensitive comparison, which is required for proper
+		 * collation.
+		 */
+		rc = ntfs_collate_names(uname, uname_len,
+				(ntfschar*)&ie->key.file_name.file_name,
+				ie->key.file_name.file_name_length, 1,
+				CASE_SENSITIVE, vol->upcase, vol->upcase_len);
+		if (rc == -1)
+			break;
+		if (rc)
+			continue;
+		/*
+		 * Perfect match, this will never happen as the
+		 * ntfs_are_names_equal() call will have gotten a match but we
+		 * still treat it correctly.
+		 */
+		goto found_it2;
+	}
+	/*
+	 * We have finished with this index buffer without success. Check for
+	 * the presence of a child node.
+	 */
+	if (ie->flags & INDEX_ENTRY_NODE) {
+		if ((ia->index.flags & NODE_MASK) == LEAF_NODE) {
+			ntfs_error(sb, "Index entry with child node found in "
+					"a leaf node in directory inode 0x%lx.",
+					dir_ni->mft_no);
+			goto unm_err_out;
+		}
+		/* Child node present, descend into it. */
+		old_vcn = vcn;
+		vcn = sle64_to_cpup((u8*)ie + le16_to_cpu(ie->length) - 8);
+		if (vcn >= 0) {
+			/* If vcn is in the same page cache page as old_vcn we
+			 * recycle the mapped page. */
+			if (old_vcn << vol->cluster_size_bits >>
+					PAGE_SHIFT == vcn <<
+					vol->cluster_size_bits >>
+					PAGE_SHIFT)
+				goto fast_descend_into_child_node;
+			unlock_page(page);
+			ntfs_unmap_page(page);
+			goto descend_into_child_node;
+		}
+		ntfs_error(sb, "Negative child node vcn in directory inode "
+				"0x%lx.", dir_ni->mft_no);
+		goto unm_err_out;
+	}
+	/* No child node, return -ENOENT. */
+	ntfs_debug("Entry not found.");
+	err = -ENOENT;
+unm_err_out:
+	unlock_page(page);
+	ntfs_unmap_page(page);
+err_out:
+	if (!err)
+		err = -EIO;
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(dir_ni);
+	return ERR_MREF(err);
+dir_err_out:
+	ntfs_error(sb, "Corrupt directory. Aborting lookup.");
+	goto err_out;
+}
+
+#endif
+
+/**
+ * ntfs_filldir - ntfs specific filldir method
+ * @vol:	current ntfs volume
+ * @ndir:	ntfs inode of current directory
+ * @ia_page:	page in which the index allocation buffer @ie is in resides
+ * @ie:		current index entry
+ * @name:	buffer to use for the converted name
+ * @actor:	what to feed the entries to
+ *
+ * Convert the Unicode @name to the loaded NLS and pass it to the @filldir
+ * callback.
+ *
+ * If @ia_page is not NULL it is the locked page containing the index
+ * allocation block containing the index entry @ie.
+ *
+ * Note, we drop (and then reacquire) the page lock on @ia_page across the
+ * @filldir() call otherwise we would deadlock with NFSd when it calls ->lookup
+ * since ntfs_lookup() will lock the same page.  As an optimization, we do not
+ * retake the lock if we are returning a non-zero value as ntfs_readdir()
+ * would need to drop the lock immediately anyway.
+ */
+static inline int ntfs_filldir(ntfs_volume *vol,
+		ntfs_inode *ndir, struct page *ia_page, INDEX_ENTRY *ie,
+		u8 *name, struct dir_context *actor)
+{
+	unsigned long mref;
+	int name_len;
+	unsigned dt_type;
+	FILE_NAME_TYPE_FLAGS name_type;
+
+	name_type = ie->key.file_name.file_name_type;
+	if (name_type == FILE_NAME_DOS) {
+		ntfs_debug("Skipping DOS name space entry.");
+		return 0;
+	}
+	if (MREF_LE(ie->data.dir.indexed_file) == FILE_root) {
+		ntfs_debug("Skipping root directory self reference entry.");
+		return 0;
+	}
+	if (MREF_LE(ie->data.dir.indexed_file) < FILE_first_user &&
+			!NVolShowSystemFiles(vol)) {
+		ntfs_debug("Skipping system file.");
+		return 0;
+	}
+	name_len = ntfs_ucstonls(vol, (ntfschar*)&ie->key.file_name.file_name,
+			ie->key.file_name.file_name_length, &name,
+			NTFS_MAX_NAME_LEN * NLS_MAX_CHARSET_SIZE + 1);
+	if (name_len <= 0) {
+		ntfs_warning(vol->sb, "Skipping unrepresentable inode 0x%llx.",
+				(long long)MREF_LE(ie->data.dir.indexed_file));
+		return 0;
+	}
+	if (ie->key.file_name.file_attributes &
+			FILE_ATTR_DUP_FILE_NAME_INDEX_PRESENT)
+		dt_type = DT_DIR;
+	else
+		dt_type = DT_REG;
+	mref = MREF_LE(ie->data.dir.indexed_file);
+	/*
+	 * Drop the page lock otherwise we deadlock with NFS when it calls
+	 * ->lookup since ntfs_lookup() will lock the same page.
+	 */
+	if (ia_page)
+		unlock_page(ia_page);
+	ntfs_debug("Calling filldir for %s with len %i, fpos 0x%llx, inode "
+			"0x%lx, DT_%s.", name, name_len, actor->pos, mref,
+			dt_type == DT_DIR ? "DIR" : "REG");
+	if (!dir_emit(actor, name, name_len, mref, dt_type))
+		return 1;
+	/* Relock the page but not if we are aborting ->readdir. */
+	if (ia_page)
+		lock_page(ia_page);
+	return 0;
+}
+
+/*
+ * We use the same basic approach as the old NTFS driver, i.e. we parse the
+ * index root entries and then the index allocation entries that are marked
+ * as in use in the index bitmap.
+ *
+ * While this will return the names in random order this doesn't matter for
+ * ->readdir but OTOH results in a faster ->readdir.
+ *
+ * VFS calls ->readdir without BKL but with i_mutex held. This protects the VFS
+ * parts (e.g. ->f_pos and ->i_size, and it also protects against directory
+ * modifications).
+ *
+ * Locking:  - Caller must hold i_mutex on the directory.
+ *	     - Each page cache page in the index allocation mapping must be
+ *	       locked whilst being accessed otherwise we may find a corrupt
+ *	       page due to it being under ->writepage at the moment which
+ *	       applies the mst protection fixups before writing out and then
+ *	       removes them again after the write is complete after which it 
+ *	       unlocks the page.
+ */
+static int ntfs_readdir(struct file *file, struct dir_context *actor)
+{
+	s64 ia_pos, ia_start, prev_ia_pos, bmp_pos;
+	loff_t i_size;
+	struct inode *bmp_vi, *vdir = file_inode(file);
+	struct super_block *sb = vdir->i_sb;
+	ntfs_inode *ndir = NTFS_I(vdir);
+	ntfs_volume *vol = NTFS_SB(sb);
+	MFT_RECORD *m;
+	INDEX_ROOT *ir = NULL;
+	INDEX_ENTRY *ie;
+	INDEX_ALLOCATION *ia;
+	u8 *name = NULL;
+	int rc, err, ir_pos, cur_bmp_pos;
+	struct address_space *ia_mapping, *bmp_mapping;
+	struct page *bmp_page = NULL, *ia_page = NULL;
+	u8 *kaddr, *bmp, *index_end;
+	ntfs_attr_search_ctx *ctx;
+
+	ntfs_debug("Entering for inode 0x%lx, fpos 0x%llx.",
+			vdir->i_ino, actor->pos);
+	rc = err = 0;
+	/* Are we at end of dir yet? */
+	i_size = i_size_read(vdir);
+	if (actor->pos >= i_size + vol->mft_record_size)
+		return 0;
+	/* Emulate . and .. for all directories. */
+	if (!dir_emit_dots(file, actor))
+		return 0;
+	m = NULL;
+	ctx = NULL;
+	/*
+	 * Allocate a buffer to store the current name being processed
+	 * converted to format determined by current NLS.
+	 */
+	name = kmalloc(NTFS_MAX_NAME_LEN * NLS_MAX_CHARSET_SIZE + 1, GFP_NOFS);
+	if (unlikely(!name)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	/* Are we jumping straight into the index allocation attribute? */
+	if (actor->pos >= vol->mft_record_size)
+		goto skip_index_root;
+	/* Get hold of the mft record for the directory. */
+	m = map_mft_record(ndir);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		m = NULL;
+		goto err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(ndir, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	/* Get the offset into the index root attribute. */
+	ir_pos = (s64)actor->pos;
+	/* Find the index root attribute in the mft record. */
+	err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE, 0, NULL,
+			0, ctx);
+	if (unlikely(err)) {
+		ntfs_error(sb, "Index root attribute missing in directory "
+				"inode 0x%lx.", vdir->i_ino);
+		goto err_out;
+	}
+	/*
+	 * Copy the index root attribute value to a buffer so that we can put
+	 * the search context and unmap the mft record before calling the
+	 * filldir() callback.  We need to do this because of NFSd which calls
+	 * ->lookup() from its filldir callback() and this causes NTFS to
+	 * deadlock as ntfs_lookup() maps the mft record of the directory and
+	 * we have got it mapped here already.  The only solution is for us to
+	 * unmap the mft record here so that a call to ntfs_lookup() is able to
+	 * map the mft record without deadlocking.
+	 */
+	rc = le32_to_cpu(ctx->attr->data.resident.value_length);
+	ir = kmalloc(rc, GFP_NOFS);
+	if (unlikely(!ir)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	/* Copy the index root value (it has been verified in read_inode). */
+	memcpy(ir, (u8*)ctx->attr +
+			le16_to_cpu(ctx->attr->data.resident.value_offset), rc);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(ndir);
+	ctx = NULL;
+	m = NULL;
+	index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
+	/* The first index entry. */
+	ie = (INDEX_ENTRY*)((u8*)&ir->index +
+			le32_to_cpu(ir->index.entries_offset));
+	/*
+	 * Loop until we exceed valid memory (corruption case) or until we
+	 * reach the last entry or until filldir tells us it has had enough
+	 * or signals an error (both covered by the rc test).
+	 */
+	for (;; ie = (INDEX_ENTRY*)((u8*)ie + le16_to_cpu(ie->length))) {
+		ntfs_debug("In index root, offset 0x%zx.", (u8*)ie - (u8*)ir);
+		/* Bounds checks. */
+		if (unlikely((u8*)ie < (u8*)ir || (u8*)ie +
+				sizeof(INDEX_ENTRY_HEADER) > index_end ||
+				(u8*)ie + le16_to_cpu(ie->key_length) >
+				index_end))
+			goto err_out;
+		/* The last entry cannot contain a name. */
+		if (ie->flags & INDEX_ENTRY_END)
+			break;
+		/* Skip index root entry if continuing previous readdir. */
+		if (ir_pos > (u8*)ie - (u8*)ir)
+			continue;
+		/* Advance the position even if going to skip the entry. */
+		actor->pos = (u8*)ie - (u8*)ir;
+		/* Submit the name to the filldir callback. */
+		rc = ntfs_filldir(vol, ndir, NULL, ie, name, actor);
+		if (rc) {
+			kfree(ir);
+			goto abort;
+		}
+	}
+	/* We are done with the index root and can free the buffer. */
+	kfree(ir);
+	ir = NULL;
+	/* If there is no index allocation attribute we are finished. */
+	if (!NInoIndexAllocPresent(ndir))
+		goto EOD;
+	/* Advance fpos to the beginning of the index allocation. */
+	actor->pos = vol->mft_record_size;
+skip_index_root:
+	kaddr = NULL;
+	prev_ia_pos = -1LL;
+	/* Get the offset into the index allocation attribute. */
+	ia_pos = (s64)actor->pos - vol->mft_record_size;
+	ia_mapping = vdir->i_mapping;
+	ntfs_debug("Inode 0x%lx, getting index bitmap.", vdir->i_ino);
+	bmp_vi = ntfs_attr_iget(vdir, AT_BITMAP, I30, 4);
+	if (IS_ERR(bmp_vi)) {
+		ntfs_error(sb, "Failed to get bitmap attribute.");
+		err = PTR_ERR(bmp_vi);
+		goto err_out;
+	}
+	bmp_mapping = bmp_vi->i_mapping;
+	/* Get the starting bitmap bit position and sanity check it. */
+	bmp_pos = ia_pos >> ndir->itype.index.block_size_bits;
+	if (unlikely(bmp_pos >> 3 >= i_size_read(bmp_vi))) {
+		ntfs_error(sb, "Current index allocation position exceeds "
+				"index bitmap size.");
+		goto iput_err_out;
+	}
+	/* Get the starting bit position in the current bitmap page. */
+	cur_bmp_pos = bmp_pos & ((PAGE_SIZE * 8) - 1);
+	bmp_pos &= ~(u64)((PAGE_SIZE * 8) - 1);
+get_next_bmp_page:
+	ntfs_debug("Reading bitmap with page index 0x%llx, bit ofs 0x%llx",
+			(unsigned long long)bmp_pos >> (3 + PAGE_SHIFT),
+			(unsigned long long)bmp_pos &
+			(unsigned long long)((PAGE_SIZE * 8) - 1));
+	bmp_page = ntfs_map_page(bmp_mapping,
+			bmp_pos >> (3 + PAGE_SHIFT));
+	if (IS_ERR(bmp_page)) {
+		ntfs_error(sb, "Reading index bitmap failed.");
+		err = PTR_ERR(bmp_page);
+		bmp_page = NULL;
+		goto iput_err_out;
+	}
+	bmp = (u8*)page_address(bmp_page);
+	/* Find next index block in use. */
+	while (!(bmp[cur_bmp_pos >> 3] & (1 << (cur_bmp_pos & 7)))) {
+find_next_index_buffer:
+		cur_bmp_pos++;
+		/*
+		 * If we have reached the end of the bitmap page, get the next
+		 * page, and put away the old one.
+		 */
+		if (unlikely((cur_bmp_pos >> 3) >= PAGE_SIZE)) {
+			ntfs_unmap_page(bmp_page);
+			bmp_pos += PAGE_SIZE * 8;
+			cur_bmp_pos = 0;
+			goto get_next_bmp_page;
+		}
+		/* If we have reached the end of the bitmap, we are done. */
+		if (unlikely(((bmp_pos + cur_bmp_pos) >> 3) >= i_size))
+			goto unm_EOD;
+		ia_pos = (bmp_pos + cur_bmp_pos) <<
+				ndir->itype.index.block_size_bits;
+	}
+	ntfs_debug("Handling index buffer 0x%llx.",
+			(unsigned long long)bmp_pos + cur_bmp_pos);
+	/* If the current index buffer is in the same page we reuse the page. */
+	if ((prev_ia_pos & (s64)PAGE_MASK) !=
+			(ia_pos & (s64)PAGE_MASK)) {
+		prev_ia_pos = ia_pos;
+		if (likely(ia_page != NULL)) {
+			unlock_page(ia_page);
+			ntfs_unmap_page(ia_page);
+		}
+		/*
+		 * Map the page cache page containing the current ia_pos,
+		 * reading it from disk if necessary.
+		 */
+		ia_page = ntfs_map_page(ia_mapping, ia_pos >> PAGE_SHIFT);
+		if (IS_ERR(ia_page)) {
+			ntfs_error(sb, "Reading index allocation data failed.");
+			err = PTR_ERR(ia_page);
+			ia_page = NULL;
+			goto err_out;
+		}
+		lock_page(ia_page);
+		kaddr = (u8*)page_address(ia_page);
+	}
+	/* Get the current index buffer. */
+	ia = (INDEX_ALLOCATION*)(kaddr + (ia_pos & ~PAGE_MASK &
+					  ~(s64)(ndir->itype.index.block_size - 1)));
+	/* Bounds checks. */
+	if (unlikely((u8*)ia < kaddr || (u8*)ia > kaddr + PAGE_SIZE)) {
+		ntfs_error(sb, "Out of bounds check failed. Corrupt directory "
+				"inode 0x%lx or driver bug.", vdir->i_ino);
+		goto err_out;
+	}
+	/* Catch multi sector transfer fixup errors. */
+	if (unlikely(!ntfs_is_indx_record(ia->magic))) {
+		ntfs_error(sb, "Directory index record with vcn 0x%llx is "
+				"corrupt.  Corrupt inode 0x%lx.  Run chkdsk.",
+				(unsigned long long)ia_pos >>
+				ndir->itype.index.vcn_size_bits, vdir->i_ino);
+		goto err_out;
+	}
+	if (unlikely(sle64_to_cpu(ia->index_block_vcn) != (ia_pos &
+			~(s64)(ndir->itype.index.block_size - 1)) >>
+			ndir->itype.index.vcn_size_bits)) {
+		ntfs_error(sb, "Actual VCN (0x%llx) of index buffer is "
+				"different from expected VCN (0x%llx). "
+				"Directory inode 0x%lx is corrupt or driver "
+				"bug. ", (unsigned long long)
+				sle64_to_cpu(ia->index_block_vcn),
+				(unsigned long long)ia_pos >>
+				ndir->itype.index.vcn_size_bits, vdir->i_ino);
+		goto err_out;
+	}
+	if (unlikely(le32_to_cpu(ia->index.allocated_size) + 0x18 !=
+			ndir->itype.index.block_size)) {
+		ntfs_error(sb, "Index buffer (VCN 0x%llx) of directory inode "
+				"0x%lx has a size (%u) differing from the "
+				"directory specified size (%u). Directory "
+				"inode is corrupt or driver bug.",
+				(unsigned long long)ia_pos >>
+				ndir->itype.index.vcn_size_bits, vdir->i_ino,
+				le32_to_cpu(ia->index.allocated_size) + 0x18,
+				ndir->itype.index.block_size);
+		goto err_out;
+	}
+	index_end = (u8*)ia + ndir->itype.index.block_size;
+	if (unlikely(index_end > kaddr + PAGE_SIZE)) {
+		ntfs_error(sb, "Index buffer (VCN 0x%llx) of directory inode "
+				"0x%lx crosses page boundary. Impossible! "
+				"Cannot access! This is probably a bug in the "
+				"driver.", (unsigned long long)ia_pos >>
+				ndir->itype.index.vcn_size_bits, vdir->i_ino);
+		goto err_out;
+	}
+	ia_start = ia_pos & ~(s64)(ndir->itype.index.block_size - 1);
+	index_end = (u8*)&ia->index + le32_to_cpu(ia->index.index_length);
+	if (unlikely(index_end > (u8*)ia + ndir->itype.index.block_size)) {
+		ntfs_error(sb, "Size of index buffer (VCN 0x%llx) of directory "
+				"inode 0x%lx exceeds maximum size.",
+				(unsigned long long)ia_pos >>
+				ndir->itype.index.vcn_size_bits, vdir->i_ino);
+		goto err_out;
+	}
+	/* The first index entry in this index buffer. */
+	ie = (INDEX_ENTRY*)((u8*)&ia->index +
+			le32_to_cpu(ia->index.entries_offset));
+	/*
+	 * Loop until we exceed valid memory (corruption case) or until we
+	 * reach the last entry or until filldir tells us it has had enough
+	 * or signals an error (both covered by the rc test).
+	 */
+	for (;; ie = (INDEX_ENTRY*)((u8*)ie + le16_to_cpu(ie->length))) {
+		ntfs_debug("In index allocation, offset 0x%llx.",
+				(unsigned long long)ia_start +
+				(unsigned long long)((u8*)ie - (u8*)ia));
+		/* Bounds checks. */
+		if (unlikely((u8*)ie < (u8*)ia || (u8*)ie +
+				sizeof(INDEX_ENTRY_HEADER) > index_end ||
+				(u8*)ie + le16_to_cpu(ie->key_length) >
+				index_end))
+			goto err_out;
+		/* The last entry cannot contain a name. */
+		if (ie->flags & INDEX_ENTRY_END)
+			break;
+		/* Skip index block entry if continuing previous readdir. */
+		if (ia_pos - ia_start > (u8*)ie - (u8*)ia)
+			continue;
+		/* Advance the position even if going to skip the entry. */
+		actor->pos = (u8*)ie - (u8*)ia +
+				(sle64_to_cpu(ia->index_block_vcn) <<
+				ndir->itype.index.vcn_size_bits) +
+				vol->mft_record_size;
+		/*
+		 * Submit the name to the @filldir callback.  Note,
+		 * ntfs_filldir() drops the lock on @ia_page but it retakes it
+		 * before returning, unless a non-zero value is returned in
+		 * which case the page is left unlocked.
+		 */
+		rc = ntfs_filldir(vol, ndir, ia_page, ie, name, actor);
+		if (rc) {
+			/* @ia_page is already unlocked in this case. */
+			ntfs_unmap_page(ia_page);
+			ntfs_unmap_page(bmp_page);
+			iput(bmp_vi);
+			goto abort;
+		}
+	}
+	goto find_next_index_buffer;
+unm_EOD:
+	if (ia_page) {
+		unlock_page(ia_page);
+		ntfs_unmap_page(ia_page);
+	}
+	ntfs_unmap_page(bmp_page);
+	iput(bmp_vi);
+EOD:
+	/* We are finished, set fpos to EOD. */
+	actor->pos = i_size + vol->mft_record_size;
+abort:
+	kfree(name);
+	return 0;
+err_out:
+	if (bmp_page) {
+		ntfs_unmap_page(bmp_page);
+iput_err_out:
+		iput(bmp_vi);
+	}
+	if (ia_page) {
+		unlock_page(ia_page);
+		ntfs_unmap_page(ia_page);
+	}
+	kfree(ir);
+	kfree(name);
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(ndir);
+	if (!err)
+		err = -EIO;
+	ntfs_debug("Failed. Returning error code %i.", -err);
+	return err;
+}
+
+/**
+ * ntfs_dir_open - called when an inode is about to be opened
+ * @vi:		inode to be opened
+ * @filp:	file structure describing the inode
+ *
+ * Limit directory size to the page cache limit on architectures where unsigned
+ * long is 32-bits. This is the most we can do for now without overflowing the
+ * page cache page index. Doing it this way means we don't run into problems
+ * because of existing too large directories. It would be better to allow the
+ * user to read the accessible part of the directory but I doubt very much
+ * anyone is going to hit this check on a 32-bit architecture, so there is no
+ * point in adding the extra complexity required to support this.
+ *
+ * On 64-bit architectures, the check is hopefully optimized away by the
+ * compiler.
+ */
+static int ntfs_dir_open(struct inode *vi, struct file *filp)
+{
+	if (sizeof(unsigned long) < 8) {
+		if (i_size_read(vi) > MAX_LFS_FILESIZE)
+			return -EFBIG;
+	}
+	return 0;
+}
+
+#ifdef NTFS_RW
+
+/**
+ * ntfs_dir_fsync - sync a directory to disk
+ * @filp:	directory to be synced
+ * @dentry:	dentry describing the directory to sync
+ * @datasync:	if non-zero only flush user data and not metadata
+ *
+ * Data integrity sync of a directory to disk.  Used for fsync, fdatasync, and
+ * msync system calls.  This function is based on file.c::ntfs_file_fsync().
+ *
+ * Write the mft record and all associated extent mft records as well as the
+ * $INDEX_ALLOCATION and $BITMAP attributes and then sync the block device.
+ *
+ * If @datasync is true, we do not wait on the inode(s) to be written out
+ * but we always wait on the page cache pages to be written out.
+ *
+ * Note: In the past @filp could be NULL so we ignore it as we don't need it
+ * anyway.
+ *
+ * Locking: Caller must hold i_mutex on the inode.
+ *
+ * TODO: We should probably also write all attribute/index inodes associated
+ * with this inode but since we have no simple way of getting to them we ignore
+ * this problem for now.  We do write the $BITMAP attribute if it is present
+ * which is the important one for a directory so things are not too bad.
+ */
+static int ntfs_dir_fsync(struct file *filp, loff_t start, loff_t end,
+			  int datasync)
+{
+	struct inode *bmp_vi, *vi = filp->f_mapping->host;
+	int err, ret;
+	ntfs_attr na;
+
+	ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
+
+	err = file_write_and_wait_range(filp, start, end);
+	if (err)
+		return err;
+	inode_lock(vi);
+
+	BUG_ON(!S_ISDIR(vi->i_mode));
+	/* If the bitmap attribute inode is in memory sync it, too. */
+	na.mft_no = vi->i_ino;
+	na.type = AT_BITMAP;
+	na.name = I30;
+	na.name_len = 4;
+	bmp_vi = ilookup5(vi->i_sb, vi->i_ino, ntfs_test_inode, &na);
+	if (bmp_vi) {
+ 		write_inode_now(bmp_vi, !datasync);
+		iput(bmp_vi);
+	}
+	ret = __ntfs_write_inode(vi, 1);
+	write_inode_now(vi, !datasync);
+	err = sync_blockdev(vi->i_sb->s_bdev);
+	if (unlikely(err && !ret))
+		ret = err;
+	if (likely(!ret))
+		ntfs_debug("Done.");
+	else
+		ntfs_warning(vi->i_sb, "Failed to f%ssync inode 0x%lx.  Error "
+				"%u.", datasync ? "data" : "", vi->i_ino, -ret);
+	inode_unlock(vi);
+	return ret;
+}
+
+#endif /* NTFS_RW */
+
+const struct file_operations ntfs_dir_ops = {
+	.llseek		= generic_file_llseek,	/* Seek inside directory. */
+	.read		= generic_read_dir,	/* Return -EISDIR. */
+	.iterate	= ntfs_readdir,		/* Read directory contents. */
+#ifdef NTFS_RW
+	.fsync		= ntfs_dir_fsync,	/* Sync a directory to disk. */
+#endif /* NTFS_RW */
+	/*.ioctl	= ,*/			/* Perform function on the
+						   mounted filesystem. */
+	.open		= ntfs_dir_open,	/* Open directory. */
+};
diff --git a/fs/ntfs/dir.h b/fs/ntfs/dir.h
new file mode 100644
index 000000000..0e326753d
--- /dev/null
+++ b/fs/ntfs/dir.h
@@ -0,0 +1,34 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * dir.h - Defines for directory handling in NTFS Linux kernel driver. Part of
+ *	   the Linux-NTFS project.
+ *
+ * Copyright (c) 2002-2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_DIR_H
+#define _LINUX_NTFS_DIR_H
+
+#include "layout.h"
+#include "inode.h"
+#include "types.h"
+
+/*
+ * ntfs_name is used to return the file name to the caller of
+ * ntfs_lookup_inode_by_name() in order for the caller (namei.c::ntfs_lookup())
+ * to be able to deal with dcache aliasing issues.
+ */
+typedef struct {
+	MFT_REF mref;
+	FILE_NAME_TYPE_FLAGS type;
+	u8 len;
+	ntfschar name[0];
+} __attribute__ ((__packed__)) ntfs_name;
+
+/* The little endian Unicode string $I30 as a global constant. */
+extern ntfschar I30[5];
+
+extern MFT_REF ntfs_lookup_inode_by_name(ntfs_inode *dir_ni,
+		const ntfschar *uname, const int uname_len, ntfs_name **res);
+
+#endif /* _LINUX_NTFS_FS_DIR_H */
diff --git a/fs/ntfs/endian.h b/fs/ntfs/endian.h
new file mode 100644
index 000000000..f30c139bf
--- /dev/null
+++ b/fs/ntfs/endian.h
@@ -0,0 +1,79 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * endian.h - Defines for endianness handling in NTFS Linux kernel driver.
+ *	      Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_ENDIAN_H
+#define _LINUX_NTFS_ENDIAN_H
+
+#include <asm/byteorder.h>
+#include "types.h"
+
+/*
+ * Signed endianness conversion functions.
+ */
+
+static inline s16 sle16_to_cpu(sle16 x)
+{
+	return le16_to_cpu((__force le16)x);
+}
+
+static inline s32 sle32_to_cpu(sle32 x)
+{
+	return le32_to_cpu((__force le32)x);
+}
+
+static inline s64 sle64_to_cpu(sle64 x)
+{
+	return le64_to_cpu((__force le64)x);
+}
+
+static inline s16 sle16_to_cpup(sle16 *x)
+{
+	return le16_to_cpu(*(__force le16*)x);
+}
+
+static inline s32 sle32_to_cpup(sle32 *x)
+{
+	return le32_to_cpu(*(__force le32*)x);
+}
+
+static inline s64 sle64_to_cpup(sle64 *x)
+{
+	return le64_to_cpu(*(__force le64*)x);
+}
+
+static inline sle16 cpu_to_sle16(s16 x)
+{
+	return (__force sle16)cpu_to_le16(x);
+}
+
+static inline sle32 cpu_to_sle32(s32 x)
+{
+	return (__force sle32)cpu_to_le32(x);
+}
+
+static inline sle64 cpu_to_sle64(s64 x)
+{
+	return (__force sle64)cpu_to_le64(x);
+}
+
+static inline sle16 cpu_to_sle16p(s16 *x)
+{
+	return (__force sle16)cpu_to_le16(*x);
+}
+
+static inline sle32 cpu_to_sle32p(s32 *x)
+{
+	return (__force sle32)cpu_to_le32(*x);
+}
+
+static inline sle64 cpu_to_sle64p(s64 *x)
+{
+	return (__force sle64)cpu_to_le64(*x);
+}
+
+#endif /* _LINUX_NTFS_ENDIAN_H */
diff --git a/fs/ntfs/file.c b/fs/ntfs/file.c
new file mode 100644
index 000000000..f42967b73
--- /dev/null
+++ b/fs/ntfs/file.c
@@ -0,0 +1,2026 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * file.c - NTFS kernel file operations.  Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2015 Anton Altaparmakov and Tuxera Inc.
+ */
+
+#include <linux/backing-dev.h>
+#include <linux/buffer_head.h>
+#include <linux/gfp.h>
+#include <linux/pagemap.h>
+#include <linux/pagevec.h>
+#include <linux/sched/signal.h>
+#include <linux/swap.h>
+#include <linux/uio.h>
+#include <linux/writeback.h>
+
+#include <asm/page.h>
+#include <linux/uaccess.h>
+
+#include "attrib.h"
+#include "bitmap.h"
+#include "inode.h"
+#include "debug.h"
+#include "lcnalloc.h"
+#include "malloc.h"
+#include "mft.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_file_open - called when an inode is about to be opened
+ * @vi:		inode to be opened
+ * @filp:	file structure describing the inode
+ *
+ * Limit file size to the page cache limit on architectures where unsigned long
+ * is 32-bits. This is the most we can do for now without overflowing the page
+ * cache page index. Doing it this way means we don't run into problems because
+ * of existing too large files. It would be better to allow the user to read
+ * the beginning of the file but I doubt very much anyone is going to hit this
+ * check on a 32-bit architecture, so there is no point in adding the extra
+ * complexity required to support this.
+ *
+ * On 64-bit architectures, the check is hopefully optimized away by the
+ * compiler.
+ *
+ * After the check passes, just call generic_file_open() to do its work.
+ */
+static int ntfs_file_open(struct inode *vi, struct file *filp)
+{
+	if (sizeof(unsigned long) < 8) {
+		if (i_size_read(vi) > MAX_LFS_FILESIZE)
+			return -EOVERFLOW;
+	}
+	return generic_file_open(vi, filp);
+}
+
+#ifdef NTFS_RW
+
+/**
+ * ntfs_attr_extend_initialized - extend the initialized size of an attribute
+ * @ni:			ntfs inode of the attribute to extend
+ * @new_init_size:	requested new initialized size in bytes
+ *
+ * Extend the initialized size of an attribute described by the ntfs inode @ni
+ * to @new_init_size bytes.  This involves zeroing any non-sparse space between
+ * the old initialized size and @new_init_size both in the page cache and on
+ * disk (if relevant complete pages are already uptodate in the page cache then
+ * these are simply marked dirty).
+ *
+ * As a side-effect, the file size (vfs inode->i_size) may be incremented as,
+ * in the resident attribute case, it is tied to the initialized size and, in
+ * the non-resident attribute case, it may not fall below the initialized size.
+ *
+ * Note that if the attribute is resident, we do not need to touch the page
+ * cache at all.  This is because if the page cache page is not uptodate we
+ * bring it uptodate later, when doing the write to the mft record since we
+ * then already have the page mapped.  And if the page is uptodate, the
+ * non-initialized region will already have been zeroed when the page was
+ * brought uptodate and the region may in fact already have been overwritten
+ * with new data via mmap() based writes, so we cannot just zero it.  And since
+ * POSIX specifies that the behaviour of resizing a file whilst it is mmap()ped
+ * is unspecified, we choose not to do zeroing and thus we do not need to touch
+ * the page at all.  For a more detailed explanation see ntfs_truncate() in
+ * fs/ntfs/inode.c.
+ *
+ * Return 0 on success and -errno on error.  In the case that an error is
+ * encountered it is possible that the initialized size will already have been
+ * incremented some way towards @new_init_size but it is guaranteed that if
+ * this is the case, the necessary zeroing will also have happened and that all
+ * metadata is self-consistent.
+ *
+ * Locking: i_mutex on the vfs inode corrseponsind to the ntfs inode @ni must be
+ *	    held by the caller.
+ */
+static int ntfs_attr_extend_initialized(ntfs_inode *ni, const s64 new_init_size)
+{
+	s64 old_init_size;
+	loff_t old_i_size;
+	pgoff_t index, end_index;
+	unsigned long flags;
+	struct inode *vi = VFS_I(ni);
+	ntfs_inode *base_ni;
+	MFT_RECORD *m = NULL;
+	ATTR_RECORD *a;
+	ntfs_attr_search_ctx *ctx = NULL;
+	struct address_space *mapping;
+	struct page *page = NULL;
+	u8 *kattr;
+	int err;
+	u32 attr_len;
+
+	read_lock_irqsave(&ni->size_lock, flags);
+	old_init_size = ni->initialized_size;
+	old_i_size = i_size_read(vi);
+	BUG_ON(new_init_size > ni->allocated_size);
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	ntfs_debug("Entering for i_ino 0x%lx, attribute type 0x%x, "
+			"old_initialized_size 0x%llx, "
+			"new_initialized_size 0x%llx, i_size 0x%llx.",
+			vi->i_ino, (unsigned)le32_to_cpu(ni->type),
+			(unsigned long long)old_init_size,
+			(unsigned long long)new_init_size, old_i_size);
+	if (!NInoAttr(ni))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	/* Use goto to reduce indentation and we need the label below anyway. */
+	if (NInoNonResident(ni))
+		goto do_non_resident_extend;
+	BUG_ON(old_init_size != old_i_size);
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		m = NULL;
+		goto err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT)
+			err = -EIO;
+		goto err_out;
+	}
+	m = ctx->mrec;
+	a = ctx->attr;
+	BUG_ON(a->non_resident);
+	/* The total length of the attribute value. */
+	attr_len = le32_to_cpu(a->data.resident.value_length);
+	BUG_ON(old_i_size != (loff_t)attr_len);
+	/*
+	 * Do the zeroing in the mft record and update the attribute size in
+	 * the mft record.
+	 */
+	kattr = (u8*)a + le16_to_cpu(a->data.resident.value_offset);
+	memset(kattr + attr_len, 0, new_init_size - attr_len);
+	a->data.resident.value_length = cpu_to_le32((u32)new_init_size);
+	/* Finally, update the sizes in the vfs and ntfs inodes. */
+	write_lock_irqsave(&ni->size_lock, flags);
+	i_size_write(vi, new_init_size);
+	ni->initialized_size = new_init_size;
+	write_unlock_irqrestore(&ni->size_lock, flags);
+	goto done;
+do_non_resident_extend:
+	/*
+	 * If the new initialized size @new_init_size exceeds the current file
+	 * size (vfs inode->i_size), we need to extend the file size to the
+	 * new initialized size.
+	 */
+	if (new_init_size > old_i_size) {
+		m = map_mft_record(base_ni);
+		if (IS_ERR(m)) {
+			err = PTR_ERR(m);
+			m = NULL;
+			goto err_out;
+		}
+		ctx = ntfs_attr_get_search_ctx(base_ni, m);
+		if (unlikely(!ctx)) {
+			err = -ENOMEM;
+			goto err_out;
+		}
+		err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+				CASE_SENSITIVE, 0, NULL, 0, ctx);
+		if (unlikely(err)) {
+			if (err == -ENOENT)
+				err = -EIO;
+			goto err_out;
+		}
+		m = ctx->mrec;
+		a = ctx->attr;
+		BUG_ON(!a->non_resident);
+		BUG_ON(old_i_size != (loff_t)
+				sle64_to_cpu(a->data.non_resident.data_size));
+		a->data.non_resident.data_size = cpu_to_sle64(new_init_size);
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		mark_mft_record_dirty(ctx->ntfs_ino);
+		/* Update the file size in the vfs inode. */
+		i_size_write(vi, new_init_size);
+		ntfs_attr_put_search_ctx(ctx);
+		ctx = NULL;
+		unmap_mft_record(base_ni);
+		m = NULL;
+	}
+	mapping = vi->i_mapping;
+	index = old_init_size >> PAGE_SHIFT;
+	end_index = (new_init_size + PAGE_SIZE - 1) >> PAGE_SHIFT;
+	do {
+		/*
+		 * Read the page.  If the page is not present, this will zero
+		 * the uninitialized regions for us.
+		 */
+		page = read_mapping_page(mapping, index, NULL);
+		if (IS_ERR(page)) {
+			err = PTR_ERR(page);
+			goto init_err_out;
+		}
+		if (unlikely(PageError(page))) {
+			put_page(page);
+			err = -EIO;
+			goto init_err_out;
+		}
+		/*
+		 * Update the initialized size in the ntfs inode.  This is
+		 * enough to make ntfs_writepage() work.
+		 */
+		write_lock_irqsave(&ni->size_lock, flags);
+		ni->initialized_size = (s64)(index + 1) << PAGE_SHIFT;
+		if (ni->initialized_size > new_init_size)
+			ni->initialized_size = new_init_size;
+		write_unlock_irqrestore(&ni->size_lock, flags);
+		/* Set the page dirty so it gets written out. */
+		set_page_dirty(page);
+		put_page(page);
+		/*
+		 * Play nice with the vm and the rest of the system.  This is
+		 * very much needed as we can potentially be modifying the
+		 * initialised size from a very small value to a really huge
+		 * value, e.g.
+		 *	f = open(somefile, O_TRUNC);
+		 *	truncate(f, 10GiB);
+		 *	seek(f, 10GiB);
+		 *	write(f, 1);
+		 * And this would mean we would be marking dirty hundreds of
+		 * thousands of pages or as in the above example more than
+		 * two and a half million pages!
+		 *
+		 * TODO: For sparse pages could optimize this workload by using
+		 * the FsMisc / MiscFs page bit as a "PageIsSparse" bit.  This
+		 * would be set in readpage for sparse pages and here we would
+		 * not need to mark dirty any pages which have this bit set.
+		 * The only caveat is that we have to clear the bit everywhere
+		 * where we allocate any clusters that lie in the page or that
+		 * contain the page.
+		 *
+		 * TODO: An even greater optimization would be for us to only
+		 * call readpage() on pages which are not in sparse regions as
+		 * determined from the runlist.  This would greatly reduce the
+		 * number of pages we read and make dirty in the case of sparse
+		 * files.
+		 */
+		balance_dirty_pages_ratelimited(mapping);
+		cond_resched();
+	} while (++index < end_index);
+	read_lock_irqsave(&ni->size_lock, flags);
+	BUG_ON(ni->initialized_size != new_init_size);
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	/* Now bring in sync the initialized_size in the mft record. */
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		m = NULL;
+		goto init_err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto init_err_out;
+	}
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT)
+			err = -EIO;
+		goto init_err_out;
+	}
+	m = ctx->mrec;
+	a = ctx->attr;
+	BUG_ON(!a->non_resident);
+	a->data.non_resident.initialized_size = cpu_to_sle64(new_init_size);
+done:
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(base_ni);
+	ntfs_debug("Done, initialized_size 0x%llx, i_size 0x%llx.",
+			(unsigned long long)new_init_size, i_size_read(vi));
+	return 0;
+init_err_out:
+	write_lock_irqsave(&ni->size_lock, flags);
+	ni->initialized_size = old_init_size;
+	write_unlock_irqrestore(&ni->size_lock, flags);
+err_out:
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(base_ni);
+	ntfs_debug("Failed.  Returning error code %i.", err);
+	return err;
+}
+
+static ssize_t ntfs_prepare_file_for_write(struct kiocb *iocb,
+		struct iov_iter *from)
+{
+	loff_t pos;
+	s64 end, ll;
+	ssize_t err;
+	unsigned long flags;
+	struct file *file = iocb->ki_filp;
+	struct inode *vi = file_inode(file);
+	ntfs_inode *base_ni, *ni = NTFS_I(vi);
+	ntfs_volume *vol = ni->vol;
+
+	ntfs_debug("Entering for i_ino 0x%lx, attribute type 0x%x, pos "
+			"0x%llx, count 0x%zx.", vi->i_ino,
+			(unsigned)le32_to_cpu(ni->type),
+			(unsigned long long)iocb->ki_pos,
+			iov_iter_count(from));
+	err = generic_write_checks(iocb, from);
+	if (unlikely(err <= 0))
+		goto out;
+	/*
+	 * All checks have passed.  Before we start doing any writing we want
+	 * to abort any totally illegal writes.
+	 */
+	BUG_ON(NInoMstProtected(ni));
+	BUG_ON(ni->type != AT_DATA);
+	/* If file is encrypted, deny access, just like NT4. */
+	if (NInoEncrypted(ni)) {
+		/* Only $DATA attributes can be encrypted. */
+		/*
+		 * Reminder for later: Encrypted files are _always_
+		 * non-resident so that the content can always be encrypted.
+		 */
+		ntfs_debug("Denying write access to encrypted file.");
+		err = -EACCES;
+		goto out;
+	}
+	if (NInoCompressed(ni)) {
+		/* Only unnamed $DATA attribute can be compressed. */
+		BUG_ON(ni->name_len);
+		/*
+		 * Reminder for later: If resident, the data is not actually
+		 * compressed.  Only on the switch to non-resident does
+		 * compression kick in.  This is in contrast to encrypted files
+		 * (see above).
+		 */
+		ntfs_error(vi->i_sb, "Writing to compressed files is not "
+				"implemented yet.  Sorry.");
+		err = -EOPNOTSUPP;
+		goto out;
+	}
+	base_ni = ni;
+	if (NInoAttr(ni))
+		base_ni = ni->ext.base_ntfs_ino;
+	err = file_remove_privs(file);
+	if (unlikely(err))
+		goto out;
+	/*
+	 * Our ->update_time method always succeeds thus file_update_time()
+	 * cannot fail either so there is no need to check the return code.
+	 */
+	file_update_time(file);
+	pos = iocb->ki_pos;
+	/* The first byte after the last cluster being written to. */
+	end = (pos + iov_iter_count(from) + vol->cluster_size_mask) &
+			~(u64)vol->cluster_size_mask;
+	/*
+	 * If the write goes beyond the allocated size, extend the allocation
+	 * to cover the whole of the write, rounded up to the nearest cluster.
+	 */
+	read_lock_irqsave(&ni->size_lock, flags);
+	ll = ni->allocated_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	if (end > ll) {
+		/*
+		 * Extend the allocation without changing the data size.
+		 *
+		 * Note we ensure the allocation is big enough to at least
+		 * write some data but we do not require the allocation to be
+		 * complete, i.e. it may be partial.
+		 */
+		ll = ntfs_attr_extend_allocation(ni, end, -1, pos);
+		if (likely(ll >= 0)) {
+			BUG_ON(pos >= ll);
+			/* If the extension was partial truncate the write. */
+			if (end > ll) {
+				ntfs_debug("Truncating write to inode 0x%lx, "
+						"attribute type 0x%x, because "
+						"the allocation was only "
+						"partially extended.",
+						vi->i_ino, (unsigned)
+						le32_to_cpu(ni->type));
+				iov_iter_truncate(from, ll - pos);
+			}
+		} else {
+			err = ll;
+			read_lock_irqsave(&ni->size_lock, flags);
+			ll = ni->allocated_size;
+			read_unlock_irqrestore(&ni->size_lock, flags);
+			/* Perform a partial write if possible or fail. */
+			if (pos < ll) {
+				ntfs_debug("Truncating write to inode 0x%lx "
+						"attribute type 0x%x, because "
+						"extending the allocation "
+						"failed (error %d).",
+						vi->i_ino, (unsigned)
+						le32_to_cpu(ni->type),
+						(int)-err);
+				iov_iter_truncate(from, ll - pos);
+			} else {
+				if (err != -ENOSPC)
+					ntfs_error(vi->i_sb, "Cannot perform "
+							"write to inode "
+							"0x%lx, attribute "
+							"type 0x%x, because "
+							"extending the "
+							"allocation failed "
+							"(error %ld).",
+							vi->i_ino, (unsigned)
+							le32_to_cpu(ni->type),
+							(long)-err);
+				else
+					ntfs_debug("Cannot perform write to "
+							"inode 0x%lx, "
+							"attribute type 0x%x, "
+							"because there is not "
+							"space left.",
+							vi->i_ino, (unsigned)
+							le32_to_cpu(ni->type));
+				goto out;
+			}
+		}
+	}
+	/*
+	 * If the write starts beyond the initialized size, extend it up to the
+	 * beginning of the write and initialize all non-sparse space between
+	 * the old initialized size and the new one.  This automatically also
+	 * increments the vfs inode->i_size to keep it above or equal to the
+	 * initialized_size.
+	 */
+	read_lock_irqsave(&ni->size_lock, flags);
+	ll = ni->initialized_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	if (pos > ll) {
+		/*
+		 * Wait for ongoing direct i/o to complete before proceeding.
+		 * New direct i/o cannot start as we hold i_mutex.
+		 */
+		inode_dio_wait(vi);
+		err = ntfs_attr_extend_initialized(ni, pos);
+		if (unlikely(err < 0))
+			ntfs_error(vi->i_sb, "Cannot perform write to inode "
+					"0x%lx, attribute type 0x%x, because "
+					"extending the initialized size "
+					"failed (error %d).", vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type),
+					(int)-err);
+	}
+out:
+	return err;
+}
+
+/**
+ * __ntfs_grab_cache_pages - obtain a number of locked pages
+ * @mapping:	address space mapping from which to obtain page cache pages
+ * @index:	starting index in @mapping at which to begin obtaining pages
+ * @nr_pages:	number of page cache pages to obtain
+ * @pages:	array of pages in which to return the obtained page cache pages
+ * @cached_page: allocated but as yet unused page
+ *
+ * Obtain @nr_pages locked page cache pages from the mapping @mapping and
+ * starting at index @index.
+ *
+ * If a page is newly created, add it to lru list
+ *
+ * Note, the page locks are obtained in ascending page index order.
+ */
+static inline int __ntfs_grab_cache_pages(struct address_space *mapping,
+		pgoff_t index, const unsigned nr_pages, struct page **pages,
+		struct page **cached_page)
+{
+	int err, nr;
+
+	BUG_ON(!nr_pages);
+	err = nr = 0;
+	do {
+		pages[nr] = find_get_page_flags(mapping, index, FGP_LOCK |
+				FGP_ACCESSED);
+		if (!pages[nr]) {
+			if (!*cached_page) {
+				*cached_page = page_cache_alloc(mapping);
+				if (unlikely(!*cached_page)) {
+					err = -ENOMEM;
+					goto err_out;
+				}
+			}
+			err = add_to_page_cache_lru(*cached_page, mapping,
+				   index,
+				   mapping_gfp_constraint(mapping, GFP_KERNEL));
+			if (unlikely(err)) {
+				if (err == -EEXIST)
+					continue;
+				goto err_out;
+			}
+			pages[nr] = *cached_page;
+			*cached_page = NULL;
+		}
+		index++;
+		nr++;
+	} while (nr < nr_pages);
+out:
+	return err;
+err_out:
+	while (nr > 0) {
+		unlock_page(pages[--nr]);
+		put_page(pages[nr]);
+	}
+	goto out;
+}
+
+static inline int ntfs_submit_bh_for_read(struct buffer_head *bh)
+{
+	lock_buffer(bh);
+	get_bh(bh);
+	bh->b_end_io = end_buffer_read_sync;
+	return submit_bh(REQ_OP_READ, 0, bh);
+}
+
+/**
+ * ntfs_prepare_pages_for_non_resident_write - prepare pages for receiving data
+ * @pages:	array of destination pages
+ * @nr_pages:	number of pages in @pages
+ * @pos:	byte position in file at which the write begins
+ * @bytes:	number of bytes to be written
+ *
+ * This is called for non-resident attributes from ntfs_file_buffered_write()
+ * with i_mutex held on the inode (@pages[0]->mapping->host).  There are
+ * @nr_pages pages in @pages which are locked but not kmap()ped.  The source
+ * data has not yet been copied into the @pages.
+ * 
+ * Need to fill any holes with actual clusters, allocate buffers if necessary,
+ * ensure all the buffers are mapped, and bring uptodate any buffers that are
+ * only partially being written to.
+ *
+ * If @nr_pages is greater than one, we are guaranteed that the cluster size is
+ * greater than PAGE_SIZE, that all pages in @pages are entirely inside
+ * the same cluster and that they are the entirety of that cluster, and that
+ * the cluster is sparse, i.e. we need to allocate a cluster to fill the hole.
+ *
+ * i_size is not to be modified yet.
+ *
+ * Return 0 on success or -errno on error.
+ */
+static int ntfs_prepare_pages_for_non_resident_write(struct page **pages,
+		unsigned nr_pages, s64 pos, size_t bytes)
+{
+	VCN vcn, highest_vcn = 0, cpos, cend, bh_cpos, bh_cend;
+	LCN lcn;
+	s64 bh_pos, vcn_len, end, initialized_size;
+	sector_t lcn_block;
+	struct page *page;
+	struct inode *vi;
+	ntfs_inode *ni, *base_ni = NULL;
+	ntfs_volume *vol;
+	runlist_element *rl, *rl2;
+	struct buffer_head *bh, *head, *wait[2], **wait_bh = wait;
+	ntfs_attr_search_ctx *ctx = NULL;
+	MFT_RECORD *m = NULL;
+	ATTR_RECORD *a = NULL;
+	unsigned long flags;
+	u32 attr_rec_len = 0;
+	unsigned blocksize, u;
+	int err, mp_size;
+	bool rl_write_locked, was_hole, is_retry;
+	unsigned char blocksize_bits;
+	struct {
+		u8 runlist_merged:1;
+		u8 mft_attr_mapped:1;
+		u8 mp_rebuilt:1;
+		u8 attr_switched:1;
+	} status = { 0, 0, 0, 0 };
+
+	BUG_ON(!nr_pages);
+	BUG_ON(!pages);
+	BUG_ON(!*pages);
+	vi = pages[0]->mapping->host;
+	ni = NTFS_I(vi);
+	vol = ni->vol;
+	ntfs_debug("Entering for inode 0x%lx, attribute type 0x%x, start page "
+			"index 0x%lx, nr_pages 0x%x, pos 0x%llx, bytes 0x%zx.",
+			vi->i_ino, ni->type, pages[0]->index, nr_pages,
+			(long long)pos, bytes);
+	blocksize = vol->sb->s_blocksize;
+	blocksize_bits = vol->sb->s_blocksize_bits;
+	u = 0;
+	do {
+		page = pages[u];
+		BUG_ON(!page);
+		/*
+		 * create_empty_buffers() will create uptodate/dirty buffers if
+		 * the page is uptodate/dirty.
+		 */
+		if (!page_has_buffers(page)) {
+			create_empty_buffers(page, blocksize, 0);
+			if (unlikely(!page_has_buffers(page)))
+				return -ENOMEM;
+		}
+	} while (++u < nr_pages);
+	rl_write_locked = false;
+	rl = NULL;
+	err = 0;
+	vcn = lcn = -1;
+	vcn_len = 0;
+	lcn_block = -1;
+	was_hole = false;
+	cpos = pos >> vol->cluster_size_bits;
+	end = pos + bytes;
+	cend = (end + vol->cluster_size - 1) >> vol->cluster_size_bits;
+	/*
+	 * Loop over each page and for each page over each buffer.  Use goto to
+	 * reduce indentation.
+	 */
+	u = 0;
+do_next_page:
+	page = pages[u];
+	bh_pos = (s64)page->index << PAGE_SHIFT;
+	bh = head = page_buffers(page);
+	do {
+		VCN cdelta;
+		s64 bh_end;
+		unsigned bh_cofs;
+
+		/* Clear buffer_new on all buffers to reinitialise state. */
+		if (buffer_new(bh))
+			clear_buffer_new(bh);
+		bh_end = bh_pos + blocksize;
+		bh_cpos = bh_pos >> vol->cluster_size_bits;
+		bh_cofs = bh_pos & vol->cluster_size_mask;
+		if (buffer_mapped(bh)) {
+			/*
+			 * The buffer is already mapped.  If it is uptodate,
+			 * ignore it.
+			 */
+			if (buffer_uptodate(bh))
+				continue;
+			/*
+			 * The buffer is not uptodate.  If the page is uptodate
+			 * set the buffer uptodate and otherwise ignore it.
+			 */
+			if (PageUptodate(page)) {
+				set_buffer_uptodate(bh);
+				continue;
+			}
+			/*
+			 * Neither the page nor the buffer are uptodate.  If
+			 * the buffer is only partially being written to, we
+			 * need to read it in before the write, i.e. now.
+			 */
+			if ((bh_pos < pos && bh_end > pos) ||
+					(bh_pos < end && bh_end > end)) {
+				/*
+				 * If the buffer is fully or partially within
+				 * the initialized size, do an actual read.
+				 * Otherwise, simply zero the buffer.
+				 */
+				read_lock_irqsave(&ni->size_lock, flags);
+				initialized_size = ni->initialized_size;
+				read_unlock_irqrestore(&ni->size_lock, flags);
+				if (bh_pos < initialized_size) {
+					ntfs_submit_bh_for_read(bh);
+					*wait_bh++ = bh;
+				} else {
+					zero_user(page, bh_offset(bh),
+							blocksize);
+					set_buffer_uptodate(bh);
+				}
+			}
+			continue;
+		}
+		/* Unmapped buffer.  Need to map it. */
+		bh->b_bdev = vol->sb->s_bdev;
+		/*
+		 * If the current buffer is in the same clusters as the map
+		 * cache, there is no need to check the runlist again.  The
+		 * map cache is made up of @vcn, which is the first cached file
+		 * cluster, @vcn_len which is the number of cached file
+		 * clusters, @lcn is the device cluster corresponding to @vcn,
+		 * and @lcn_block is the block number corresponding to @lcn.
+		 */
+		cdelta = bh_cpos - vcn;
+		if (likely(!cdelta || (cdelta > 0 && cdelta < vcn_len))) {
+map_buffer_cached:
+			BUG_ON(lcn < 0);
+			bh->b_blocknr = lcn_block +
+					(cdelta << (vol->cluster_size_bits -
+					blocksize_bits)) +
+					(bh_cofs >> blocksize_bits);
+			set_buffer_mapped(bh);
+			/*
+			 * If the page is uptodate so is the buffer.  If the
+			 * buffer is fully outside the write, we ignore it if
+			 * it was already allocated and we mark it dirty so it
+			 * gets written out if we allocated it.  On the other
+			 * hand, if we allocated the buffer but we are not
+			 * marking it dirty we set buffer_new so we can do
+			 * error recovery.
+			 */
+			if (PageUptodate(page)) {
+				if (!buffer_uptodate(bh))
+					set_buffer_uptodate(bh);
+				if (unlikely(was_hole)) {
+					/* We allocated the buffer. */
+					clean_bdev_bh_alias(bh);
+					if (bh_end <= pos || bh_pos >= end)
+						mark_buffer_dirty(bh);
+					else
+						set_buffer_new(bh);
+				}
+				continue;
+			}
+			/* Page is _not_ uptodate. */
+			if (likely(!was_hole)) {
+				/*
+				 * Buffer was already allocated.  If it is not
+				 * uptodate and is only partially being written
+				 * to, we need to read it in before the write,
+				 * i.e. now.
+				 */
+				if (!buffer_uptodate(bh) && bh_pos < end &&
+						bh_end > pos &&
+						(bh_pos < pos ||
+						bh_end > end)) {
+					/*
+					 * If the buffer is fully or partially
+					 * within the initialized size, do an
+					 * actual read.  Otherwise, simply zero
+					 * the buffer.
+					 */
+					read_lock_irqsave(&ni->size_lock,
+							flags);
+					initialized_size = ni->initialized_size;
+					read_unlock_irqrestore(&ni->size_lock,
+							flags);
+					if (bh_pos < initialized_size) {
+						ntfs_submit_bh_for_read(bh);
+						*wait_bh++ = bh;
+					} else {
+						zero_user(page, bh_offset(bh),
+								blocksize);
+						set_buffer_uptodate(bh);
+					}
+				}
+				continue;
+			}
+			/* We allocated the buffer. */
+			clean_bdev_bh_alias(bh);
+			/*
+			 * If the buffer is fully outside the write, zero it,
+			 * set it uptodate, and mark it dirty so it gets
+			 * written out.  If it is partially being written to,
+			 * zero region surrounding the write but leave it to
+			 * commit write to do anything else.  Finally, if the
+			 * buffer is fully being overwritten, do nothing.
+			 */
+			if (bh_end <= pos || bh_pos >= end) {
+				if (!buffer_uptodate(bh)) {
+					zero_user(page, bh_offset(bh),
+							blocksize);
+					set_buffer_uptodate(bh);
+				}
+				mark_buffer_dirty(bh);
+				continue;
+			}
+			set_buffer_new(bh);
+			if (!buffer_uptodate(bh) &&
+					(bh_pos < pos || bh_end > end)) {
+				u8 *kaddr;
+				unsigned pofs;
+					
+				kaddr = kmap_atomic(page);
+				if (bh_pos < pos) {
+					pofs = bh_pos & ~PAGE_MASK;
+					memset(kaddr + pofs, 0, pos - bh_pos);
+				}
+				if (bh_end > end) {
+					pofs = end & ~PAGE_MASK;
+					memset(kaddr + pofs, 0, bh_end - end);
+				}
+				kunmap_atomic(kaddr);
+				flush_dcache_page(page);
+			}
+			continue;
+		}
+		/*
+		 * Slow path: this is the first buffer in the cluster.  If it
+		 * is outside allocated size and is not uptodate, zero it and
+		 * set it uptodate.
+		 */
+		read_lock_irqsave(&ni->size_lock, flags);
+		initialized_size = ni->allocated_size;
+		read_unlock_irqrestore(&ni->size_lock, flags);
+		if (bh_pos > initialized_size) {
+			if (PageUptodate(page)) {
+				if (!buffer_uptodate(bh))
+					set_buffer_uptodate(bh);
+			} else if (!buffer_uptodate(bh)) {
+				zero_user(page, bh_offset(bh), blocksize);
+				set_buffer_uptodate(bh);
+			}
+			continue;
+		}
+		is_retry = false;
+		if (!rl) {
+			down_read(&ni->runlist.lock);
+retry_remap:
+			rl = ni->runlist.rl;
+		}
+		if (likely(rl != NULL)) {
+			/* Seek to element containing target cluster. */
+			while (rl->length && rl[1].vcn <= bh_cpos)
+				rl++;
+			lcn = ntfs_rl_vcn_to_lcn(rl, bh_cpos);
+			if (likely(lcn >= 0)) {
+				/*
+				 * Successful remap, setup the map cache and
+				 * use that to deal with the buffer.
+				 */
+				was_hole = false;
+				vcn = bh_cpos;
+				vcn_len = rl[1].vcn - vcn;
+				lcn_block = lcn << (vol->cluster_size_bits -
+						blocksize_bits);
+				cdelta = 0;
+				/*
+				 * If the number of remaining clusters touched
+				 * by the write is smaller or equal to the
+				 * number of cached clusters, unlock the
+				 * runlist as the map cache will be used from
+				 * now on.
+				 */
+				if (likely(vcn + vcn_len >= cend)) {
+					if (rl_write_locked) {
+						up_write(&ni->runlist.lock);
+						rl_write_locked = false;
+					} else
+						up_read(&ni->runlist.lock);
+					rl = NULL;
+				}
+				goto map_buffer_cached;
+			}
+		} else
+			lcn = LCN_RL_NOT_MAPPED;
+		/*
+		 * If it is not a hole and not out of bounds, the runlist is
+		 * probably unmapped so try to map it now.
+		 */
+		if (unlikely(lcn != LCN_HOLE && lcn != LCN_ENOENT)) {
+			if (likely(!is_retry && lcn == LCN_RL_NOT_MAPPED)) {
+				/* Attempt to map runlist. */
+				if (!rl_write_locked) {
+					/*
+					 * We need the runlist locked for
+					 * writing, so if it is locked for
+					 * reading relock it now and retry in
+					 * case it changed whilst we dropped
+					 * the lock.
+					 */
+					up_read(&ni->runlist.lock);
+					down_write(&ni->runlist.lock);
+					rl_write_locked = true;
+					goto retry_remap;
+				}
+				err = ntfs_map_runlist_nolock(ni, bh_cpos,
+						NULL);
+				if (likely(!err)) {
+					is_retry = true;
+					goto retry_remap;
+				}
+				/*
+				 * If @vcn is out of bounds, pretend @lcn is
+				 * LCN_ENOENT.  As long as the buffer is out
+				 * of bounds this will work fine.
+				 */
+				if (err == -ENOENT) {
+					lcn = LCN_ENOENT;
+					err = 0;
+					goto rl_not_mapped_enoent;
+				}
+			} else
+				err = -EIO;
+			/* Failed to map the buffer, even after retrying. */
+			bh->b_blocknr = -1;
+			ntfs_error(vol->sb, "Failed to write to inode 0x%lx, "
+					"attribute type 0x%x, vcn 0x%llx, "
+					"vcn offset 0x%x, because its "
+					"location on disk could not be "
+					"determined%s (error code %i).",
+					ni->mft_no, ni->type,
+					(unsigned long long)bh_cpos,
+					(unsigned)bh_pos &
+					vol->cluster_size_mask,
+					is_retry ? " even after retrying" : "",
+					err);
+			break;
+		}
+rl_not_mapped_enoent:
+		/*
+		 * The buffer is in a hole or out of bounds.  We need to fill
+		 * the hole, unless the buffer is in a cluster which is not
+		 * touched by the write, in which case we just leave the buffer
+		 * unmapped.  This can only happen when the cluster size is
+		 * less than the page cache size.
+		 */
+		if (unlikely(vol->cluster_size < PAGE_SIZE)) {
+			bh_cend = (bh_end + vol->cluster_size - 1) >>
+					vol->cluster_size_bits;
+			if ((bh_cend <= cpos || bh_cpos >= cend)) {
+				bh->b_blocknr = -1;
+				/*
+				 * If the buffer is uptodate we skip it.  If it
+				 * is not but the page is uptodate, we can set
+				 * the buffer uptodate.  If the page is not
+				 * uptodate, we can clear the buffer and set it
+				 * uptodate.  Whether this is worthwhile is
+				 * debatable and this could be removed.
+				 */
+				if (PageUptodate(page)) {
+					if (!buffer_uptodate(bh))
+						set_buffer_uptodate(bh);
+				} else if (!buffer_uptodate(bh)) {
+					zero_user(page, bh_offset(bh),
+						blocksize);
+					set_buffer_uptodate(bh);
+				}
+				continue;
+			}
+		}
+		/*
+		 * Out of bounds buffer is invalid if it was not really out of
+		 * bounds.
+		 */
+		BUG_ON(lcn != LCN_HOLE);
+		/*
+		 * We need the runlist locked for writing, so if it is locked
+		 * for reading relock it now and retry in case it changed
+		 * whilst we dropped the lock.
+		 */
+		BUG_ON(!rl);
+		if (!rl_write_locked) {
+			up_read(&ni->runlist.lock);
+			down_write(&ni->runlist.lock);
+			rl_write_locked = true;
+			goto retry_remap;
+		}
+		/* Find the previous last allocated cluster. */
+		BUG_ON(rl->lcn != LCN_HOLE);
+		lcn = -1;
+		rl2 = rl;
+		while (--rl2 >= ni->runlist.rl) {
+			if (rl2->lcn >= 0) {
+				lcn = rl2->lcn + rl2->length;
+				break;
+			}
+		}
+		rl2 = ntfs_cluster_alloc(vol, bh_cpos, 1, lcn, DATA_ZONE,
+				false);
+		if (IS_ERR(rl2)) {
+			err = PTR_ERR(rl2);
+			ntfs_debug("Failed to allocate cluster, error code %i.",
+					err);
+			break;
+		}
+		lcn = rl2->lcn;
+		rl = ntfs_runlists_merge(ni->runlist.rl, rl2);
+		if (IS_ERR(rl)) {
+			err = PTR_ERR(rl);
+			if (err != -ENOMEM)
+				err = -EIO;
+			if (ntfs_cluster_free_from_rl(vol, rl2)) {
+				ntfs_error(vol->sb, "Failed to release "
+						"allocated cluster in error "
+						"code path.  Run chkdsk to "
+						"recover the lost cluster.");
+				NVolSetErrors(vol);
+			}
+			ntfs_free(rl2);
+			break;
+		}
+		ni->runlist.rl = rl;
+		status.runlist_merged = 1;
+		ntfs_debug("Allocated cluster, lcn 0x%llx.",
+				(unsigned long long)lcn);
+		/* Map and lock the mft record and get the attribute record. */
+		if (!NInoAttr(ni))
+			base_ni = ni;
+		else
+			base_ni = ni->ext.base_ntfs_ino;
+		m = map_mft_record(base_ni);
+		if (IS_ERR(m)) {
+			err = PTR_ERR(m);
+			break;
+		}
+		ctx = ntfs_attr_get_search_ctx(base_ni, m);
+		if (unlikely(!ctx)) {
+			err = -ENOMEM;
+			unmap_mft_record(base_ni);
+			break;
+		}
+		status.mft_attr_mapped = 1;
+		err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+				CASE_SENSITIVE, bh_cpos, NULL, 0, ctx);
+		if (unlikely(err)) {
+			if (err == -ENOENT)
+				err = -EIO;
+			break;
+		}
+		m = ctx->mrec;
+		a = ctx->attr;
+		/*
+		 * Find the runlist element with which the attribute extent
+		 * starts.  Note, we cannot use the _attr_ version because we
+		 * have mapped the mft record.  That is ok because we know the
+		 * runlist fragment must be mapped already to have ever gotten
+		 * here, so we can just use the _rl_ version.
+		 */
+		vcn = sle64_to_cpu(a->data.non_resident.lowest_vcn);
+		rl2 = ntfs_rl_find_vcn_nolock(rl, vcn);
+		BUG_ON(!rl2);
+		BUG_ON(!rl2->length);
+		BUG_ON(rl2->lcn < LCN_HOLE);
+		highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
+		/*
+		 * If @highest_vcn is zero, calculate the real highest_vcn
+		 * (which can really be zero).
+		 */
+		if (!highest_vcn)
+			highest_vcn = (sle64_to_cpu(
+					a->data.non_resident.allocated_size) >>
+					vol->cluster_size_bits) - 1;
+		/*
+		 * Determine the size of the mapping pairs array for the new
+		 * extent, i.e. the old extent with the hole filled.
+		 */
+		mp_size = ntfs_get_size_for_mapping_pairs(vol, rl2, vcn,
+				highest_vcn);
+		if (unlikely(mp_size <= 0)) {
+			if (!(err = mp_size))
+				err = -EIO;
+			ntfs_debug("Failed to get size for mapping pairs "
+					"array, error code %i.", err);
+			break;
+		}
+		/*
+		 * Resize the attribute record to fit the new mapping pairs
+		 * array.
+		 */
+		attr_rec_len = le32_to_cpu(a->length);
+		err = ntfs_attr_record_resize(m, a, mp_size + le16_to_cpu(
+				a->data.non_resident.mapping_pairs_offset));
+		if (unlikely(err)) {
+			BUG_ON(err != -ENOSPC);
+			// TODO: Deal with this by using the current attribute
+			// and fill it with as much of the mapping pairs
+			// array as possible.  Then loop over each attribute
+			// extent rewriting the mapping pairs arrays as we go
+			// along and if when we reach the end we have not
+			// enough space, try to resize the last attribute
+			// extent and if even that fails, add a new attribute
+			// extent.
+			// We could also try to resize at each step in the hope
+			// that we will not need to rewrite every single extent.
+			// Note, we may need to decompress some extents to fill
+			// the runlist as we are walking the extents...
+			ntfs_error(vol->sb, "Not enough space in the mft "
+					"record for the extended attribute "
+					"record.  This case is not "
+					"implemented yet.");
+			err = -EOPNOTSUPP;
+			break ;
+		}
+		status.mp_rebuilt = 1;
+		/*
+		 * Generate the mapping pairs array directly into the attribute
+		 * record.
+		 */
+		err = ntfs_mapping_pairs_build(vol, (u8*)a + le16_to_cpu(
+				a->data.non_resident.mapping_pairs_offset),
+				mp_size, rl2, vcn, highest_vcn, NULL);
+		if (unlikely(err)) {
+			ntfs_error(vol->sb, "Cannot fill hole in inode 0x%lx, "
+					"attribute type 0x%x, because building "
+					"the mapping pairs failed with error "
+					"code %i.", vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type), err);
+			err = -EIO;
+			break;
+		}
+		/* Update the highest_vcn but only if it was not set. */
+		if (unlikely(!a->data.non_resident.highest_vcn))
+			a->data.non_resident.highest_vcn =
+					cpu_to_sle64(highest_vcn);
+		/*
+		 * If the attribute is sparse/compressed, update the compressed
+		 * size in the ntfs_inode structure and the attribute record.
+		 */
+		if (likely(NInoSparse(ni) || NInoCompressed(ni))) {
+			/*
+			 * If we are not in the first attribute extent, switch
+			 * to it, but first ensure the changes will make it to
+			 * disk later.
+			 */
+			if (a->data.non_resident.lowest_vcn) {
+				flush_dcache_mft_record_page(ctx->ntfs_ino);
+				mark_mft_record_dirty(ctx->ntfs_ino);
+				ntfs_attr_reinit_search_ctx(ctx);
+				err = ntfs_attr_lookup(ni->type, ni->name,
+						ni->name_len, CASE_SENSITIVE,
+						0, NULL, 0, ctx);
+				if (unlikely(err)) {
+					status.attr_switched = 1;
+					break;
+				}
+				/* @m is not used any more so do not set it. */
+				a = ctx->attr;
+			}
+			write_lock_irqsave(&ni->size_lock, flags);
+			ni->itype.compressed.size += vol->cluster_size;
+			a->data.non_resident.compressed_size =
+					cpu_to_sle64(ni->itype.compressed.size);
+			write_unlock_irqrestore(&ni->size_lock, flags);
+		}
+		/* Ensure the changes make it to disk. */
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		mark_mft_record_dirty(ctx->ntfs_ino);
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(base_ni);
+		/* Successfully filled the hole. */
+		status.runlist_merged = 0;
+		status.mft_attr_mapped = 0;
+		status.mp_rebuilt = 0;
+		/* Setup the map cache and use that to deal with the buffer. */
+		was_hole = true;
+		vcn = bh_cpos;
+		vcn_len = 1;
+		lcn_block = lcn << (vol->cluster_size_bits - blocksize_bits);
+		cdelta = 0;
+		/*
+		 * If the number of remaining clusters in the @pages is smaller
+		 * or equal to the number of cached clusters, unlock the
+		 * runlist as the map cache will be used from now on.
+		 */
+		if (likely(vcn + vcn_len >= cend)) {
+			up_write(&ni->runlist.lock);
+			rl_write_locked = false;
+			rl = NULL;
+		}
+		goto map_buffer_cached;
+	} while (bh_pos += blocksize, (bh = bh->b_this_page) != head);
+	/* If there are no errors, do the next page. */
+	if (likely(!err && ++u < nr_pages))
+		goto do_next_page;
+	/* If there are no errors, release the runlist lock if we took it. */
+	if (likely(!err)) {
+		if (unlikely(rl_write_locked)) {
+			up_write(&ni->runlist.lock);
+			rl_write_locked = false;
+		} else if (unlikely(rl))
+			up_read(&ni->runlist.lock);
+		rl = NULL;
+	}
+	/* If we issued read requests, let them complete. */
+	read_lock_irqsave(&ni->size_lock, flags);
+	initialized_size = ni->initialized_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	while (wait_bh > wait) {
+		bh = *--wait_bh;
+		wait_on_buffer(bh);
+		if (likely(buffer_uptodate(bh))) {
+			page = bh->b_page;
+			bh_pos = ((s64)page->index << PAGE_SHIFT) +
+					bh_offset(bh);
+			/*
+			 * If the buffer overflows the initialized size, need
+			 * to zero the overflowing region.
+			 */
+			if (unlikely(bh_pos + blocksize > initialized_size)) {
+				int ofs = 0;
+
+				if (likely(bh_pos < initialized_size))
+					ofs = initialized_size - bh_pos;
+				zero_user_segment(page, bh_offset(bh) + ofs,
+						blocksize);
+			}
+		} else /* if (unlikely(!buffer_uptodate(bh))) */
+			err = -EIO;
+	}
+	if (likely(!err)) {
+		/* Clear buffer_new on all buffers. */
+		u = 0;
+		do {
+			bh = head = page_buffers(pages[u]);
+			do {
+				if (buffer_new(bh))
+					clear_buffer_new(bh);
+			} while ((bh = bh->b_this_page) != head);
+		} while (++u < nr_pages);
+		ntfs_debug("Done.");
+		return err;
+	}
+	if (status.attr_switched) {
+		/* Get back to the attribute extent we modified. */
+		ntfs_attr_reinit_search_ctx(ctx);
+		if (ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+				CASE_SENSITIVE, bh_cpos, NULL, 0, ctx)) {
+			ntfs_error(vol->sb, "Failed to find required "
+					"attribute extent of attribute in "
+					"error code path.  Run chkdsk to "
+					"recover.");
+			write_lock_irqsave(&ni->size_lock, flags);
+			ni->itype.compressed.size += vol->cluster_size;
+			write_unlock_irqrestore(&ni->size_lock, flags);
+			flush_dcache_mft_record_page(ctx->ntfs_ino);
+			mark_mft_record_dirty(ctx->ntfs_ino);
+			/*
+			 * The only thing that is now wrong is the compressed
+			 * size of the base attribute extent which chkdsk
+			 * should be able to fix.
+			 */
+			NVolSetErrors(vol);
+		} else {
+			m = ctx->mrec;
+			a = ctx->attr;
+			status.attr_switched = 0;
+		}
+	}
+	/*
+	 * If the runlist has been modified, need to restore it by punching a
+	 * hole into it and we then need to deallocate the on-disk cluster as
+	 * well.  Note, we only modify the runlist if we are able to generate a
+	 * new mapping pairs array, i.e. only when the mapped attribute extent
+	 * is not switched.
+	 */
+	if (status.runlist_merged && !status.attr_switched) {
+		BUG_ON(!rl_write_locked);
+		/* Make the file cluster we allocated sparse in the runlist. */
+		if (ntfs_rl_punch_nolock(vol, &ni->runlist, bh_cpos, 1)) {
+			ntfs_error(vol->sb, "Failed to punch hole into "
+					"attribute runlist in error code "
+					"path.  Run chkdsk to recover the "
+					"lost cluster.");
+			NVolSetErrors(vol);
+		} else /* if (success) */ {
+			status.runlist_merged = 0;
+			/*
+			 * Deallocate the on-disk cluster we allocated but only
+			 * if we succeeded in punching its vcn out of the
+			 * runlist.
+			 */
+			down_write(&vol->lcnbmp_lock);
+			if (ntfs_bitmap_clear_bit(vol->lcnbmp_ino, lcn)) {
+				ntfs_error(vol->sb, "Failed to release "
+						"allocated cluster in error "
+						"code path.  Run chkdsk to "
+						"recover the lost cluster.");
+				NVolSetErrors(vol);
+			}
+			up_write(&vol->lcnbmp_lock);
+		}
+	}
+	/*
+	 * Resize the attribute record to its old size and rebuild the mapping
+	 * pairs array.  Note, we only can do this if the runlist has been
+	 * restored to its old state which also implies that the mapped
+	 * attribute extent is not switched.
+	 */
+	if (status.mp_rebuilt && !status.runlist_merged) {
+		if (ntfs_attr_record_resize(m, a, attr_rec_len)) {
+			ntfs_error(vol->sb, "Failed to restore attribute "
+					"record in error code path.  Run "
+					"chkdsk to recover.");
+			NVolSetErrors(vol);
+		} else /* if (success) */ {
+			if (ntfs_mapping_pairs_build(vol, (u8*)a +
+					le16_to_cpu(a->data.non_resident.
+					mapping_pairs_offset), attr_rec_len -
+					le16_to_cpu(a->data.non_resident.
+					mapping_pairs_offset), ni->runlist.rl,
+					vcn, highest_vcn, NULL)) {
+				ntfs_error(vol->sb, "Failed to restore "
+						"mapping pairs array in error "
+						"code path.  Run chkdsk to "
+						"recover.");
+				NVolSetErrors(vol);
+			}
+			flush_dcache_mft_record_page(ctx->ntfs_ino);
+			mark_mft_record_dirty(ctx->ntfs_ino);
+		}
+	}
+	/* Release the mft record and the attribute. */
+	if (status.mft_attr_mapped) {
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(base_ni);
+	}
+	/* Release the runlist lock. */
+	if (rl_write_locked)
+		up_write(&ni->runlist.lock);
+	else if (rl)
+		up_read(&ni->runlist.lock);
+	/*
+	 * Zero out any newly allocated blocks to avoid exposing stale data.
+	 * If BH_New is set, we know that the block was newly allocated above
+	 * and that it has not been fully zeroed and marked dirty yet.
+	 */
+	nr_pages = u;
+	u = 0;
+	end = bh_cpos << vol->cluster_size_bits;
+	do {
+		page = pages[u];
+		bh = head = page_buffers(page);
+		do {
+			if (u == nr_pages &&
+					((s64)page->index << PAGE_SHIFT) +
+					bh_offset(bh) >= end)
+				break;
+			if (!buffer_new(bh))
+				continue;
+			clear_buffer_new(bh);
+			if (!buffer_uptodate(bh)) {
+				if (PageUptodate(page))
+					set_buffer_uptodate(bh);
+				else {
+					zero_user(page, bh_offset(bh),
+							blocksize);
+					set_buffer_uptodate(bh);
+				}
+			}
+			mark_buffer_dirty(bh);
+		} while ((bh = bh->b_this_page) != head);
+	} while (++u <= nr_pages);
+	ntfs_error(vol->sb, "Failed.  Returning error code %i.", err);
+	return err;
+}
+
+static inline void ntfs_flush_dcache_pages(struct page **pages,
+		unsigned nr_pages)
+{
+	BUG_ON(!nr_pages);
+	/*
+	 * Warning: Do not do the decrement at the same time as the call to
+	 * flush_dcache_page() because it is a NULL macro on i386 and hence the
+	 * decrement never happens so the loop never terminates.
+	 */
+	do {
+		--nr_pages;
+		flush_dcache_page(pages[nr_pages]);
+	} while (nr_pages > 0);
+}
+
+/**
+ * ntfs_commit_pages_after_non_resident_write - commit the received data
+ * @pages:	array of destination pages
+ * @nr_pages:	number of pages in @pages
+ * @pos:	byte position in file at which the write begins
+ * @bytes:	number of bytes to be written
+ *
+ * See description of ntfs_commit_pages_after_write(), below.
+ */
+static inline int ntfs_commit_pages_after_non_resident_write(
+		struct page **pages, const unsigned nr_pages,
+		s64 pos, size_t bytes)
+{
+	s64 end, initialized_size;
+	struct inode *vi;
+	ntfs_inode *ni, *base_ni;
+	struct buffer_head *bh, *head;
+	ntfs_attr_search_ctx *ctx;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	unsigned long flags;
+	unsigned blocksize, u;
+	int err;
+
+	vi = pages[0]->mapping->host;
+	ni = NTFS_I(vi);
+	blocksize = vi->i_sb->s_blocksize;
+	end = pos + bytes;
+	u = 0;
+	do {
+		s64 bh_pos;
+		struct page *page;
+		bool partial;
+
+		page = pages[u];
+		bh_pos = (s64)page->index << PAGE_SHIFT;
+		bh = head = page_buffers(page);
+		partial = false;
+		do {
+			s64 bh_end;
+
+			bh_end = bh_pos + blocksize;
+			if (bh_end <= pos || bh_pos >= end) {
+				if (!buffer_uptodate(bh))
+					partial = true;
+			} else {
+				set_buffer_uptodate(bh);
+				mark_buffer_dirty(bh);
+			}
+		} while (bh_pos += blocksize, (bh = bh->b_this_page) != head);
+		/*
+		 * If all buffers are now uptodate but the page is not, set the
+		 * page uptodate.
+		 */
+		if (!partial && !PageUptodate(page))
+			SetPageUptodate(page);
+	} while (++u < nr_pages);
+	/*
+	 * Finally, if we do not need to update initialized_size or i_size we
+	 * are finished.
+	 */
+	read_lock_irqsave(&ni->size_lock, flags);
+	initialized_size = ni->initialized_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	if (end <= initialized_size) {
+		ntfs_debug("Done.");
+		return 0;
+	}
+	/*
+	 * Update initialized_size/i_size as appropriate, both in the inode and
+	 * the mft record.
+	 */
+	if (!NInoAttr(ni))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	/* Map, pin, and lock the mft record. */
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		m = NULL;
+		ctx = NULL;
+		goto err_out;
+	}
+	BUG_ON(!NInoNonResident(ni));
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT)
+			err = -EIO;
+		goto err_out;
+	}
+	a = ctx->attr;
+	BUG_ON(!a->non_resident);
+	write_lock_irqsave(&ni->size_lock, flags);
+	BUG_ON(end > ni->allocated_size);
+	ni->initialized_size = end;
+	a->data.non_resident.initialized_size = cpu_to_sle64(end);
+	if (end > i_size_read(vi)) {
+		i_size_write(vi, end);
+		a->data.non_resident.data_size =
+				a->data.non_resident.initialized_size;
+	}
+	write_unlock_irqrestore(&ni->size_lock, flags);
+	/* Mark the mft record dirty, so it gets written back. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+	ntfs_debug("Done.");
+	return 0;
+err_out:
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(base_ni);
+	ntfs_error(vi->i_sb, "Failed to update initialized_size/i_size (error "
+			"code %i).", err);
+	if (err != -ENOMEM)
+		NVolSetErrors(ni->vol);
+	return err;
+}
+
+/**
+ * ntfs_commit_pages_after_write - commit the received data
+ * @pages:	array of destination pages
+ * @nr_pages:	number of pages in @pages
+ * @pos:	byte position in file at which the write begins
+ * @bytes:	number of bytes to be written
+ *
+ * This is called from ntfs_file_buffered_write() with i_mutex held on the inode
+ * (@pages[0]->mapping->host).  There are @nr_pages pages in @pages which are
+ * locked but not kmap()ped.  The source data has already been copied into the
+ * @page.  ntfs_prepare_pages_for_non_resident_write() has been called before
+ * the data was copied (for non-resident attributes only) and it returned
+ * success.
+ *
+ * Need to set uptodate and mark dirty all buffers within the boundary of the
+ * write.  If all buffers in a page are uptodate we set the page uptodate, too.
+ *
+ * Setting the buffers dirty ensures that they get written out later when
+ * ntfs_writepage() is invoked by the VM.
+ *
+ * Finally, we need to update i_size and initialized_size as appropriate both
+ * in the inode and the mft record.
+ *
+ * This is modelled after fs/buffer.c::generic_commit_write(), which marks
+ * buffers uptodate and dirty, sets the page uptodate if all buffers in the
+ * page are uptodate, and updates i_size if the end of io is beyond i_size.  In
+ * that case, it also marks the inode dirty.
+ *
+ * If things have gone as outlined in
+ * ntfs_prepare_pages_for_non_resident_write(), we do not need to do any page
+ * content modifications here for non-resident attributes.  For resident
+ * attributes we need to do the uptodate bringing here which we combine with
+ * the copying into the mft record which means we save one atomic kmap.
+ *
+ * Return 0 on success or -errno on error.
+ */
+static int ntfs_commit_pages_after_write(struct page **pages,
+		const unsigned nr_pages, s64 pos, size_t bytes)
+{
+	s64 end, initialized_size;
+	loff_t i_size;
+	struct inode *vi;
+	ntfs_inode *ni, *base_ni;
+	struct page *page;
+	ntfs_attr_search_ctx *ctx;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	char *kattr, *kaddr;
+	unsigned long flags;
+	u32 attr_len;
+	int err;
+
+	BUG_ON(!nr_pages);
+	BUG_ON(!pages);
+	page = pages[0];
+	BUG_ON(!page);
+	vi = page->mapping->host;
+	ni = NTFS_I(vi);
+	ntfs_debug("Entering for inode 0x%lx, attribute type 0x%x, start page "
+			"index 0x%lx, nr_pages 0x%x, pos 0x%llx, bytes 0x%zx.",
+			vi->i_ino, ni->type, page->index, nr_pages,
+			(long long)pos, bytes);
+	if (NInoNonResident(ni))
+		return ntfs_commit_pages_after_non_resident_write(pages,
+				nr_pages, pos, bytes);
+	BUG_ON(nr_pages > 1);
+	/*
+	 * Attribute is resident, implying it is not compressed, encrypted, or
+	 * sparse.
+	 */
+	if (!NInoAttr(ni))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	BUG_ON(NInoNonResident(ni));
+	/* Map, pin, and lock the mft record. */
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		m = NULL;
+		ctx = NULL;
+		goto err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT)
+			err = -EIO;
+		goto err_out;
+	}
+	a = ctx->attr;
+	BUG_ON(a->non_resident);
+	/* The total length of the attribute value. */
+	attr_len = le32_to_cpu(a->data.resident.value_length);
+	i_size = i_size_read(vi);
+	BUG_ON(attr_len != i_size);
+	BUG_ON(pos > attr_len);
+	end = pos + bytes;
+	BUG_ON(end > le32_to_cpu(a->length) -
+			le16_to_cpu(a->data.resident.value_offset));
+	kattr = (u8*)a + le16_to_cpu(a->data.resident.value_offset);
+	kaddr = kmap_atomic(page);
+	/* Copy the received data from the page to the mft record. */
+	memcpy(kattr + pos, kaddr + pos, bytes);
+	/* Update the attribute length if necessary. */
+	if (end > attr_len) {
+		attr_len = end;
+		a->data.resident.value_length = cpu_to_le32(attr_len);
+	}
+	/*
+	 * If the page is not uptodate, bring the out of bounds area(s)
+	 * uptodate by copying data from the mft record to the page.
+	 */
+	if (!PageUptodate(page)) {
+		if (pos > 0)
+			memcpy(kaddr, kattr, pos);
+		if (end < attr_len)
+			memcpy(kaddr + end, kattr + end, attr_len - end);
+		/* Zero the region outside the end of the attribute value. */
+		memset(kaddr + attr_len, 0, PAGE_SIZE - attr_len);
+		flush_dcache_page(page);
+		SetPageUptodate(page);
+	}
+	kunmap_atomic(kaddr);
+	/* Update initialized_size/i_size if necessary. */
+	read_lock_irqsave(&ni->size_lock, flags);
+	initialized_size = ni->initialized_size;
+	BUG_ON(end > ni->allocated_size);
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	BUG_ON(initialized_size != i_size);
+	if (end > initialized_size) {
+		write_lock_irqsave(&ni->size_lock, flags);
+		ni->initialized_size = end;
+		i_size_write(vi, end);
+		write_unlock_irqrestore(&ni->size_lock, flags);
+	}
+	/* Mark the mft record dirty, so it gets written back. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+	ntfs_debug("Done.");
+	return 0;
+err_out:
+	if (err == -ENOMEM) {
+		ntfs_warning(vi->i_sb, "Error allocating memory required to "
+				"commit the write.");
+		if (PageUptodate(page)) {
+			ntfs_warning(vi->i_sb, "Page is uptodate, setting "
+					"dirty so the write will be retried "
+					"later on by the VM.");
+			/*
+			 * Put the page on mapping->dirty_pages, but leave its
+			 * buffers' dirty state as-is.
+			 */
+			__set_page_dirty_nobuffers(page);
+			err = 0;
+		} else
+			ntfs_error(vi->i_sb, "Page is not uptodate.  Written "
+					"data has been lost.");
+	} else {
+		ntfs_error(vi->i_sb, "Resident attribute commit write failed "
+				"with error %i.", err);
+		NVolSetErrors(ni->vol);
+	}
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(base_ni);
+	return err;
+}
+
+/*
+ * Copy as much as we can into the pages and return the number of bytes which
+ * were successfully copied.  If a fault is encountered then clear the pages
+ * out to (ofs + bytes) and return the number of bytes which were copied.
+ */
+static size_t ntfs_copy_from_user_iter(struct page **pages, unsigned nr_pages,
+		unsigned ofs, struct iov_iter *i, size_t bytes)
+{
+	struct page **last_page = pages + nr_pages;
+	size_t total = 0;
+	struct iov_iter data = *i;
+	unsigned len, copied;
+
+	do {
+		len = PAGE_SIZE - ofs;
+		if (len > bytes)
+			len = bytes;
+		copied = iov_iter_copy_from_user_atomic(*pages, &data, ofs,
+				len);
+		total += copied;
+		bytes -= copied;
+		if (!bytes)
+			break;
+		iov_iter_advance(&data, copied);
+		if (copied < len)
+			goto err;
+		ofs = 0;
+	} while (++pages < last_page);
+out:
+	return total;
+err:
+	/* Zero the rest of the target like __copy_from_user(). */
+	len = PAGE_SIZE - copied;
+	do {
+		if (len > bytes)
+			len = bytes;
+		zero_user(*pages, copied, len);
+		bytes -= len;
+		copied = 0;
+		len = PAGE_SIZE;
+	} while (++pages < last_page);
+	goto out;
+}
+
+/**
+ * ntfs_perform_write - perform buffered write to a file
+ * @file:	file to write to
+ * @i:		iov_iter with data to write
+ * @pos:	byte offset in file at which to begin writing to
+ */
+static ssize_t ntfs_perform_write(struct file *file, struct iov_iter *i,
+		loff_t pos)
+{
+	struct address_space *mapping = file->f_mapping;
+	struct inode *vi = mapping->host;
+	ntfs_inode *ni = NTFS_I(vi);
+	ntfs_volume *vol = ni->vol;
+	struct page *pages[NTFS_MAX_PAGES_PER_CLUSTER];
+	struct page *cached_page = NULL;
+	VCN last_vcn;
+	LCN lcn;
+	size_t bytes;
+	ssize_t status, written = 0;
+	unsigned nr_pages;
+
+	ntfs_debug("Entering for i_ino 0x%lx, attribute type 0x%x, pos "
+			"0x%llx, count 0x%lx.", vi->i_ino,
+			(unsigned)le32_to_cpu(ni->type),
+			(unsigned long long)pos,
+			(unsigned long)iov_iter_count(i));
+	/*
+	 * If a previous ntfs_truncate() failed, repeat it and abort if it
+	 * fails again.
+	 */
+	if (unlikely(NInoTruncateFailed(ni))) {
+		int err;
+
+		inode_dio_wait(vi);
+		err = ntfs_truncate(vi);
+		if (err || NInoTruncateFailed(ni)) {
+			if (!err)
+				err = -EIO;
+			ntfs_error(vol->sb, "Cannot perform write to inode "
+					"0x%lx, attribute type 0x%x, because "
+					"ntfs_truncate() failed (error code "
+					"%i).", vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type), err);
+			return err;
+		}
+	}
+	/*
+	 * Determine the number of pages per cluster for non-resident
+	 * attributes.
+	 */
+	nr_pages = 1;
+	if (vol->cluster_size > PAGE_SIZE && NInoNonResident(ni))
+		nr_pages = vol->cluster_size >> PAGE_SHIFT;
+	last_vcn = -1;
+	do {
+		VCN vcn;
+		pgoff_t idx, start_idx;
+		unsigned ofs, do_pages, u;
+		size_t copied;
+
+		start_idx = idx = pos >> PAGE_SHIFT;
+		ofs = pos & ~PAGE_MASK;
+		bytes = PAGE_SIZE - ofs;
+		do_pages = 1;
+		if (nr_pages > 1) {
+			vcn = pos >> vol->cluster_size_bits;
+			if (vcn != last_vcn) {
+				last_vcn = vcn;
+				/*
+				 * Get the lcn of the vcn the write is in.  If
+				 * it is a hole, need to lock down all pages in
+				 * the cluster.
+				 */
+				down_read(&ni->runlist.lock);
+				lcn = ntfs_attr_vcn_to_lcn_nolock(ni, pos >>
+						vol->cluster_size_bits, false);
+				up_read(&ni->runlist.lock);
+				if (unlikely(lcn < LCN_HOLE)) {
+					if (lcn == LCN_ENOMEM)
+						status = -ENOMEM;
+					else {
+						status = -EIO;
+						ntfs_error(vol->sb, "Cannot "
+							"perform write to "
+							"inode 0x%lx, "
+							"attribute type 0x%x, "
+							"because the attribute "
+							"is corrupt.",
+							vi->i_ino, (unsigned)
+							le32_to_cpu(ni->type));
+					}
+					break;
+				}
+				if (lcn == LCN_HOLE) {
+					start_idx = (pos & ~(s64)
+							vol->cluster_size_mask)
+							>> PAGE_SHIFT;
+					bytes = vol->cluster_size - (pos &
+							vol->cluster_size_mask);
+					do_pages = nr_pages;
+				}
+			}
+		}
+		if (bytes > iov_iter_count(i))
+			bytes = iov_iter_count(i);
+again:
+		/*
+		 * Bring in the user page(s) that we will copy from _first_.
+		 * Otherwise there is a nasty deadlock on copying from the same
+		 * page(s) as we are writing to, without it/them being marked
+		 * up-to-date.  Note, at present there is nothing to stop the
+		 * pages being swapped out between us bringing them into memory
+		 * and doing the actual copying.
+		 */
+		if (unlikely(iov_iter_fault_in_readable(i, bytes))) {
+			status = -EFAULT;
+			break;
+		}
+		/* Get and lock @do_pages starting at index @start_idx. */
+		status = __ntfs_grab_cache_pages(mapping, start_idx, do_pages,
+				pages, &cached_page);
+		if (unlikely(status))
+			break;
+		/*
+		 * For non-resident attributes, we need to fill any holes with
+		 * actual clusters and ensure all bufferes are mapped.  We also
+		 * need to bring uptodate any buffers that are only partially
+		 * being written to.
+		 */
+		if (NInoNonResident(ni)) {
+			status = ntfs_prepare_pages_for_non_resident_write(
+					pages, do_pages, pos, bytes);
+			if (unlikely(status)) {
+				do {
+					unlock_page(pages[--do_pages]);
+					put_page(pages[do_pages]);
+				} while (do_pages);
+				break;
+			}
+		}
+		u = (pos >> PAGE_SHIFT) - pages[0]->index;
+		copied = ntfs_copy_from_user_iter(pages + u, do_pages - u, ofs,
+					i, bytes);
+		ntfs_flush_dcache_pages(pages + u, do_pages - u);
+		status = 0;
+		if (likely(copied == bytes)) {
+			status = ntfs_commit_pages_after_write(pages, do_pages,
+					pos, bytes);
+			if (!status)
+				status = bytes;
+		}
+		do {
+			unlock_page(pages[--do_pages]);
+			put_page(pages[do_pages]);
+		} while (do_pages);
+		if (unlikely(status < 0))
+			break;
+		copied = status;
+		cond_resched();
+		if (unlikely(!copied)) {
+			size_t sc;
+
+			/*
+			 * We failed to copy anything.  Fall back to single
+			 * segment length write.
+			 *
+			 * This is needed to avoid possible livelock in the
+			 * case that all segments in the iov cannot be copied
+			 * at once without a pagefault.
+			 */
+			sc = iov_iter_single_seg_count(i);
+			if (bytes > sc)
+				bytes = sc;
+			goto again;
+		}
+		iov_iter_advance(i, copied);
+		pos += copied;
+		written += copied;
+		balance_dirty_pages_ratelimited(mapping);
+		if (fatal_signal_pending(current)) {
+			status = -EINTR;
+			break;
+		}
+	} while (iov_iter_count(i));
+	if (cached_page)
+		put_page(cached_page);
+	ntfs_debug("Done.  Returning %s (written 0x%lx, status %li).",
+			written ? "written" : "status", (unsigned long)written,
+			(long)status);
+	return written ? written : status;
+}
+
+/**
+ * ntfs_file_write_iter - simple wrapper for ntfs_file_write_iter_nolock()
+ * @iocb:	IO state structure
+ * @from:	iov_iter with data to write
+ *
+ * Basically the same as generic_file_write_iter() except that it ends up
+ * up calling ntfs_perform_write() instead of generic_perform_write() and that
+ * O_DIRECT is not implemented.
+ */
+static ssize_t ntfs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
+{
+	struct file *file = iocb->ki_filp;
+	struct inode *vi = file_inode(file);
+	ssize_t written = 0;
+	ssize_t err;
+
+	inode_lock(vi);
+	/* We can write back this queue in page reclaim. */
+	current->backing_dev_info = inode_to_bdi(vi);
+	err = ntfs_prepare_file_for_write(iocb, from);
+	if (iov_iter_count(from) && !err)
+		written = ntfs_perform_write(file, from, iocb->ki_pos);
+	current->backing_dev_info = NULL;
+	inode_unlock(vi);
+	iocb->ki_pos += written;
+	if (likely(written > 0))
+		written = generic_write_sync(iocb, written);
+	return written ? written : err;
+}
+
+/**
+ * ntfs_file_fsync - sync a file to disk
+ * @filp:	file to be synced
+ * @datasync:	if non-zero only flush user data and not metadata
+ *
+ * Data integrity sync of a file to disk.  Used for fsync, fdatasync, and msync
+ * system calls.  This function is inspired by fs/buffer.c::file_fsync().
+ *
+ * If @datasync is false, write the mft record and all associated extent mft
+ * records as well as the $DATA attribute and then sync the block device.
+ *
+ * If @datasync is true and the attribute is non-resident, we skip the writing
+ * of the mft record and all associated extent mft records (this might still
+ * happen due to the write_inode_now() call).
+ *
+ * Also, if @datasync is true, we do not wait on the inode to be written out
+ * but we always wait on the page cache pages to be written out.
+ *
+ * Locking: Caller must hold i_mutex on the inode.
+ *
+ * TODO: We should probably also write all attribute/index inodes associated
+ * with this inode but since we have no simple way of getting to them we ignore
+ * this problem for now.
+ */
+static int ntfs_file_fsync(struct file *filp, loff_t start, loff_t end,
+			   int datasync)
+{
+	struct inode *vi = filp->f_mapping->host;
+	int err, ret = 0;
+
+	ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
+
+	err = file_write_and_wait_range(filp, start, end);
+	if (err)
+		return err;
+	inode_lock(vi);
+
+	BUG_ON(S_ISDIR(vi->i_mode));
+	if (!datasync || !NInoNonResident(NTFS_I(vi)))
+		ret = __ntfs_write_inode(vi, 1);
+	write_inode_now(vi, !datasync);
+	/*
+	 * NOTE: If we were to use mapping->private_list (see ext2 and
+	 * fs/buffer.c) for dirty blocks then we could optimize the below to be
+	 * sync_mapping_buffers(vi->i_mapping).
+	 */
+	err = sync_blockdev(vi->i_sb->s_bdev);
+	if (unlikely(err && !ret))
+		ret = err;
+	if (likely(!ret))
+		ntfs_debug("Done.");
+	else
+		ntfs_warning(vi->i_sb, "Failed to f%ssync inode 0x%lx.  Error "
+				"%u.", datasync ? "data" : "", vi->i_ino, -ret);
+	inode_unlock(vi);
+	return ret;
+}
+
+#endif /* NTFS_RW */
+
+const struct file_operations ntfs_file_ops = {
+	.llseek		= generic_file_llseek,
+	.read_iter	= generic_file_read_iter,
+#ifdef NTFS_RW
+	.write_iter	= ntfs_file_write_iter,
+	.fsync		= ntfs_file_fsync,
+#endif /* NTFS_RW */
+	.mmap		= generic_file_mmap,
+	.open		= ntfs_file_open,
+	.splice_read	= generic_file_splice_read,
+};
+
+const struct inode_operations ntfs_file_inode_ops = {
+#ifdef NTFS_RW
+	.setattr	= ntfs_setattr,
+#endif /* NTFS_RW */
+};
+
+const struct file_operations ntfs_empty_file_ops = {};
+
+const struct inode_operations ntfs_empty_inode_ops = {};
diff --git a/fs/ntfs/index.c b/fs/ntfs/index.c
new file mode 100644
index 000000000..d46c2c03a
--- /dev/null
+++ b/fs/ntfs/index.c
@@ -0,0 +1,440 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * index.c - NTFS kernel index handling.  Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2004-2005 Anton Altaparmakov
+ */
+
+#include <linux/slab.h>
+
+#include "aops.h"
+#include "collate.h"
+#include "debug.h"
+#include "index.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_index_ctx_get - allocate and initialize a new index context
+ * @idx_ni:	ntfs index inode with which to initialize the context
+ *
+ * Allocate a new index context, initialize it with @idx_ni and return it.
+ * Return NULL if allocation failed.
+ *
+ * Locking:  Caller must hold i_mutex on the index inode.
+ */
+ntfs_index_context *ntfs_index_ctx_get(ntfs_inode *idx_ni)
+{
+	ntfs_index_context *ictx;
+
+	ictx = kmem_cache_alloc(ntfs_index_ctx_cache, GFP_NOFS);
+	if (ictx)
+		*ictx = (ntfs_index_context){ .idx_ni = idx_ni };
+	return ictx;
+}
+
+/**
+ * ntfs_index_ctx_put - release an index context
+ * @ictx:	index context to free
+ *
+ * Release the index context @ictx, releasing all associated resources.
+ *
+ * Locking:  Caller must hold i_mutex on the index inode.
+ */
+void ntfs_index_ctx_put(ntfs_index_context *ictx)
+{
+	if (ictx->entry) {
+		if (ictx->is_in_root) {
+			if (ictx->actx)
+				ntfs_attr_put_search_ctx(ictx->actx);
+			if (ictx->base_ni)
+				unmap_mft_record(ictx->base_ni);
+		} else {
+			struct page *page = ictx->page;
+			if (page) {
+				BUG_ON(!PageLocked(page));
+				unlock_page(page);
+				ntfs_unmap_page(page);
+			}
+		}
+	}
+	kmem_cache_free(ntfs_index_ctx_cache, ictx);
+	return;
+}
+
+/**
+ * ntfs_index_lookup - find a key in an index and return its index entry
+ * @key:	[IN] key for which to search in the index
+ * @key_len:	[IN] length of @key in bytes
+ * @ictx:	[IN/OUT] context describing the index and the returned entry
+ *
+ * Before calling ntfs_index_lookup(), @ictx must have been obtained from a
+ * call to ntfs_index_ctx_get().
+ *
+ * Look for the @key in the index specified by the index lookup context @ictx.
+ * ntfs_index_lookup() walks the contents of the index looking for the @key.
+ *
+ * If the @key is found in the index, 0 is returned and @ictx is setup to
+ * describe the index entry containing the matching @key.  @ictx->entry is the
+ * index entry and @ictx->data and @ictx->data_len are the index entry data and
+ * its length in bytes, respectively.
+ *
+ * If the @key is not found in the index, -ENOENT is returned and @ictx is
+ * setup to describe the index entry whose key collates immediately after the
+ * search @key, i.e. this is the position in the index at which an index entry
+ * with a key of @key would need to be inserted.
+ *
+ * If an error occurs return the negative error code and @ictx is left
+ * untouched.
+ *
+ * When finished with the entry and its data, call ntfs_index_ctx_put() to free
+ * the context and other associated resources.
+ *
+ * If the index entry was modified, call flush_dcache_index_entry_page()
+ * immediately after the modification and either ntfs_index_entry_mark_dirty()
+ * or ntfs_index_entry_write() before the call to ntfs_index_ctx_put() to
+ * ensure that the changes are written to disk.
+ *
+ * Locking:  - Caller must hold i_mutex on the index inode.
+ *	     - Each page cache page in the index allocation mapping must be
+ *	       locked whilst being accessed otherwise we may find a corrupt
+ *	       page due to it being under ->writepage at the moment which
+ *	       applies the mst protection fixups before writing out and then
+ *	       removes them again after the write is complete after which it 
+ *	       unlocks the page.
+ */
+int ntfs_index_lookup(const void *key, const int key_len,
+		ntfs_index_context *ictx)
+{
+	VCN vcn, old_vcn;
+	ntfs_inode *idx_ni = ictx->idx_ni;
+	ntfs_volume *vol = idx_ni->vol;
+	struct super_block *sb = vol->sb;
+	ntfs_inode *base_ni = idx_ni->ext.base_ntfs_ino;
+	MFT_RECORD *m;
+	INDEX_ROOT *ir;
+	INDEX_ENTRY *ie;
+	INDEX_ALLOCATION *ia;
+	u8 *index_end, *kaddr;
+	ntfs_attr_search_ctx *actx;
+	struct address_space *ia_mapping;
+	struct page *page;
+	int rc, err = 0;
+
+	ntfs_debug("Entering.");
+	BUG_ON(!NInoAttr(idx_ni));
+	BUG_ON(idx_ni->type != AT_INDEX_ALLOCATION);
+	BUG_ON(idx_ni->nr_extents != -1);
+	BUG_ON(!base_ni);
+	BUG_ON(!key);
+	BUG_ON(key_len <= 0);
+	if (!ntfs_is_collation_rule_supported(
+			idx_ni->itype.index.collation_rule)) {
+		ntfs_error(sb, "Index uses unsupported collation rule 0x%x.  "
+				"Aborting lookup.", le32_to_cpu(
+				idx_ni->itype.index.collation_rule));
+		return -EOPNOTSUPP;
+	}
+	/* Get hold of the mft record for the index inode. */
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		ntfs_error(sb, "map_mft_record() failed with error code %ld.",
+				-PTR_ERR(m));
+		return PTR_ERR(m);
+	}
+	actx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (unlikely(!actx)) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	/* Find the index root attribute in the mft record. */
+	err = ntfs_attr_lookup(AT_INDEX_ROOT, idx_ni->name, idx_ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, actx);
+	if (unlikely(err)) {
+		if (err == -ENOENT) {
+			ntfs_error(sb, "Index root attribute missing in inode "
+					"0x%lx.", idx_ni->mft_no);
+			err = -EIO;
+		}
+		goto err_out;
+	}
+	/* Get to the index root value (it has been verified in read_inode). */
+	ir = (INDEX_ROOT*)((u8*)actx->attr +
+			le16_to_cpu(actx->attr->data.resident.value_offset));
+	index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
+	/* The first index entry. */
+	ie = (INDEX_ENTRY*)((u8*)&ir->index +
+			le32_to_cpu(ir->index.entries_offset));
+	/*
+	 * Loop until we exceed valid memory (corruption case) or until we
+	 * reach the last entry.
+	 */
+	for (;; ie = (INDEX_ENTRY*)((u8*)ie + le16_to_cpu(ie->length))) {
+		/* Bounds checks. */
+		if ((u8*)ie < (u8*)actx->mrec || (u8*)ie +
+				sizeof(INDEX_ENTRY_HEADER) > index_end ||
+				(u8*)ie + le16_to_cpu(ie->length) > index_end)
+			goto idx_err_out;
+		/*
+		 * The 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;
+		/* Further bounds checks. */
+		if ((u32)sizeof(INDEX_ENTRY_HEADER) +
+				le16_to_cpu(ie->key_length) >
+				le16_to_cpu(ie->data.vi.data_offset) ||
+				(u32)le16_to_cpu(ie->data.vi.data_offset) +
+				le16_to_cpu(ie->data.vi.data_length) >
+				le16_to_cpu(ie->length))
+			goto idx_err_out;
+		/* If the keys match perfectly, we setup @ictx and return 0. */
+		if ((key_len == le16_to_cpu(ie->key_length)) && !memcmp(key,
+				&ie->key, key_len)) {
+ir_done:
+			ictx->is_in_root = true;
+			ictx->ir = ir;
+			ictx->actx = actx;
+			ictx->base_ni = base_ni;
+			ictx->ia = NULL;
+			ictx->page = NULL;
+done:
+			ictx->entry = ie;
+			ictx->data = (u8*)ie +
+					le16_to_cpu(ie->data.vi.data_offset);
+			ictx->data_len = le16_to_cpu(ie->data.vi.data_length);
+			ntfs_debug("Done.");
+			return err;
+		}
+		/*
+		 * Not a perfect match, need to do full blown collation so we
+		 * know which way in the B+tree we have to go.
+		 */
+		rc = ntfs_collate(vol, idx_ni->itype.index.collation_rule, key,
+				key_len, &ie->key, le16_to_cpu(ie->key_length));
+		/*
+		 * If @key collates before the key of the current entry, there
+		 * is definitely no such key in this index but we might need to
+		 * descend into the B+tree so we just break out of the loop.
+		 */
+		if (rc == -1)
+			break;
+		/*
+		 * A match should never happen as the memcmp() call should have
+		 * cought it, but we still treat it correctly.
+		 */
+		if (!rc)
+			goto ir_done;
+		/* The keys are not equal, continue the search. */
+	}
+	/*
+	 * We have finished with this index without success.  Check for the
+	 * presence of a child node and if not present setup @ictx and return
+	 * -ENOENT.
+	 */
+	if (!(ie->flags & INDEX_ENTRY_NODE)) {
+		ntfs_debug("Entry not found.");
+		err = -ENOENT;
+		goto ir_done;
+	} /* Child node present, descend into it. */
+	/* Consistency check: Verify that an index allocation exists. */
+	if (!NInoIndexAllocPresent(idx_ni)) {
+		ntfs_error(sb, "No index allocation attribute but index entry "
+				"requires one.  Inode 0x%lx is corrupt or "
+				"driver bug.", idx_ni->mft_no);
+		goto err_out;
+	}
+	/* Get the starting vcn of the index_block holding the child node. */
+	vcn = sle64_to_cpup((sle64*)((u8*)ie + le16_to_cpu(ie->length) - 8));
+	ia_mapping = VFS_I(idx_ni)->i_mapping;
+	/*
+	 * We are done with the index root and the mft record.  Release them,
+	 * otherwise we deadlock with ntfs_map_page().
+	 */
+	ntfs_attr_put_search_ctx(actx);
+	unmap_mft_record(base_ni);
+	m = NULL;
+	actx = NULL;
+descend_into_child_node:
+	/*
+	 * Convert vcn to index into the index allocation attribute in units
+	 * of PAGE_SIZE and map the page cache page, reading it from
+	 * disk if necessary.
+	 */
+	page = ntfs_map_page(ia_mapping, vcn <<
+			idx_ni->itype.index.vcn_size_bits >> PAGE_SHIFT);
+	if (IS_ERR(page)) {
+		ntfs_error(sb, "Failed to map index page, error %ld.",
+				-PTR_ERR(page));
+		err = PTR_ERR(page);
+		goto err_out;
+	}
+	lock_page(page);
+	kaddr = (u8*)page_address(page);
+fast_descend_into_child_node:
+	/* Get to the index allocation block. */
+	ia = (INDEX_ALLOCATION*)(kaddr + ((vcn <<
+			idx_ni->itype.index.vcn_size_bits) & ~PAGE_MASK));
+	/* Bounds checks. */
+	if ((u8*)ia < kaddr || (u8*)ia > kaddr + PAGE_SIZE) {
+		ntfs_error(sb, "Out of bounds check failed.  Corrupt inode "
+				"0x%lx or driver bug.", idx_ni->mft_no);
+		goto unm_err_out;
+	}
+	/* Catch multi sector transfer fixup errors. */
+	if (unlikely(!ntfs_is_indx_record(ia->magic))) {
+		ntfs_error(sb, "Index record with vcn 0x%llx is corrupt.  "
+				"Corrupt inode 0x%lx.  Run chkdsk.",
+				(long long)vcn, idx_ni->mft_no);
+		goto unm_err_out;
+	}
+	if (sle64_to_cpu(ia->index_block_vcn) != vcn) {
+		ntfs_error(sb, "Actual VCN (0x%llx) of index buffer is "
+				"different from expected VCN (0x%llx).  Inode "
+				"0x%lx is corrupt or driver bug.",
+				(unsigned long long)
+				sle64_to_cpu(ia->index_block_vcn),
+				(unsigned long long)vcn, idx_ni->mft_no);
+		goto unm_err_out;
+	}
+	if (le32_to_cpu(ia->index.allocated_size) + 0x18 !=
+			idx_ni->itype.index.block_size) {
+		ntfs_error(sb, "Index buffer (VCN 0x%llx) of inode 0x%lx has "
+				"a size (%u) differing from the index "
+				"specified size (%u).  Inode is corrupt or "
+				"driver bug.", (unsigned long long)vcn,
+				idx_ni->mft_no,
+				le32_to_cpu(ia->index.allocated_size) + 0x18,
+				idx_ni->itype.index.block_size);
+		goto unm_err_out;
+	}
+	index_end = (u8*)ia + idx_ni->itype.index.block_size;
+	if (index_end > kaddr + PAGE_SIZE) {
+		ntfs_error(sb, "Index buffer (VCN 0x%llx) of inode 0x%lx "
+				"crosses page boundary.  Impossible!  Cannot "
+				"access!  This is probably a bug in the "
+				"driver.", (unsigned long long)vcn,
+				idx_ni->mft_no);
+		goto unm_err_out;
+	}
+	index_end = (u8*)&ia->index + le32_to_cpu(ia->index.index_length);
+	if (index_end > (u8*)ia + idx_ni->itype.index.block_size) {
+		ntfs_error(sb, "Size of index buffer (VCN 0x%llx) of inode "
+				"0x%lx exceeds maximum size.",
+				(unsigned long long)vcn, idx_ni->mft_no);
+		goto unm_err_out;
+	}
+	/* The first index entry. */
+	ie = (INDEX_ENTRY*)((u8*)&ia->index +
+			le32_to_cpu(ia->index.entries_offset));
+	/*
+	 * Iterate similar to above big loop but applied to index buffer, thus
+	 * loop until we exceed valid memory (corruption case) or until we
+	 * reach the last entry.
+	 */
+	for (;; ie = (INDEX_ENTRY*)((u8*)ie + le16_to_cpu(ie->length))) {
+		/* Bounds checks. */
+		if ((u8*)ie < (u8*)ia || (u8*)ie +
+				sizeof(INDEX_ENTRY_HEADER) > index_end ||
+				(u8*)ie + le16_to_cpu(ie->length) > index_end) {
+			ntfs_error(sb, "Index entry out of bounds in inode "
+					"0x%lx.", idx_ni->mft_no);
+			goto unm_err_out;
+		}
+		/*
+		 * The 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;
+		/* Further bounds checks. */
+		if ((u32)sizeof(INDEX_ENTRY_HEADER) +
+				le16_to_cpu(ie->key_length) >
+				le16_to_cpu(ie->data.vi.data_offset) ||
+				(u32)le16_to_cpu(ie->data.vi.data_offset) +
+				le16_to_cpu(ie->data.vi.data_length) >
+				le16_to_cpu(ie->length)) {
+			ntfs_error(sb, "Index entry out of bounds in inode "
+					"0x%lx.", idx_ni->mft_no);
+			goto unm_err_out;
+		}
+		/* If the keys match perfectly, we setup @ictx and return 0. */
+		if ((key_len == le16_to_cpu(ie->key_length)) && !memcmp(key,
+				&ie->key, key_len)) {
+ia_done:
+			ictx->is_in_root = false;
+			ictx->actx = NULL;
+			ictx->base_ni = NULL;
+			ictx->ia = ia;
+			ictx->page = page;
+			goto done;
+		}
+		/*
+		 * Not a perfect match, need to do full blown collation so we
+		 * know which way in the B+tree we have to go.
+		 */
+		rc = ntfs_collate(vol, idx_ni->itype.index.collation_rule, key,
+				key_len, &ie->key, le16_to_cpu(ie->key_length));
+		/*
+		 * If @key collates before the key of the current entry, there
+		 * is definitely no such key in this index but we might need to
+		 * descend into the B+tree so we just break out of the loop.
+		 */
+		if (rc == -1)
+			break;
+		/*
+		 * A match should never happen as the memcmp() call should have
+		 * cought it, but we still treat it correctly.
+		 */
+		if (!rc)
+			goto ia_done;
+		/* The keys are not equal, continue the search. */
+	}
+	/*
+	 * We have finished with this index buffer without success.  Check for
+	 * the presence of a child node and if not present return -ENOENT.
+	 */
+	if (!(ie->flags & INDEX_ENTRY_NODE)) {
+		ntfs_debug("Entry not found.");
+		err = -ENOENT;
+		goto ia_done;
+	}
+	if ((ia->index.flags & NODE_MASK) == LEAF_NODE) {
+		ntfs_error(sb, "Index entry with child node found in a leaf "
+				"node in inode 0x%lx.", idx_ni->mft_no);
+		goto unm_err_out;
+	}
+	/* Child node present, descend into it. */
+	old_vcn = vcn;
+	vcn = sle64_to_cpup((sle64*)((u8*)ie + le16_to_cpu(ie->length) - 8));
+	if (vcn >= 0) {
+		/*
+		 * If vcn is in the same page cache page as old_vcn we recycle
+		 * the mapped page.
+		 */
+		if (old_vcn << vol->cluster_size_bits >>
+				PAGE_SHIFT == vcn <<
+				vol->cluster_size_bits >>
+				PAGE_SHIFT)
+			goto fast_descend_into_child_node;
+		unlock_page(page);
+		ntfs_unmap_page(page);
+		goto descend_into_child_node;
+	}
+	ntfs_error(sb, "Negative child node vcn in inode 0x%lx.",
+			idx_ni->mft_no);
+unm_err_out:
+	unlock_page(page);
+	ntfs_unmap_page(page);
+err_out:
+	if (!err)
+		err = -EIO;
+	if (actx)
+		ntfs_attr_put_search_ctx(actx);
+	if (m)
+		unmap_mft_record(base_ni);
+	return err;
+idx_err_out:
+	ntfs_error(sb, "Corrupt index.  Aborting lookup.");
+	goto err_out;
+}
diff --git a/fs/ntfs/index.h b/fs/ntfs/index.h
new file mode 100644
index 000000000..bb3c3ae55
--- /dev/null
+++ b/fs/ntfs/index.h
@@ -0,0 +1,134 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * index.h - Defines for NTFS kernel index handling.  Part of the Linux-NTFS
+ *	     project.
+ *
+ * Copyright (c) 2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_INDEX_H
+#define _LINUX_NTFS_INDEX_H
+
+#include <linux/fs.h>
+
+#include "types.h"
+#include "layout.h"
+#include "inode.h"
+#include "attrib.h"
+#include "mft.h"
+#include "aops.h"
+
+/**
+ * @idx_ni:	index inode containing the @entry described by this context
+ * @entry:	index entry (points into @ir or @ia)
+ * @data:	index entry data (points into @entry)
+ * @data_len:	length in bytes of @data
+ * @is_in_root:	'true' if @entry is in @ir and 'false' if it is in @ia
+ * @ir:		index root if @is_in_root and NULL otherwise
+ * @actx:	attribute search context if @is_in_root and NULL otherwise
+ * @base_ni:	base inode if @is_in_root and NULL otherwise
+ * @ia:		index block if @is_in_root is 'false' and NULL otherwise
+ * @page:	page if @is_in_root is 'false' and NULL otherwise
+ *
+ * @idx_ni is the index inode this context belongs to.
+ *
+ * @entry is the index entry described by this context.  @data and @data_len
+ * are the index entry data and its length in bytes, respectively.  @data
+ * simply points into @entry.  This is probably what the user is interested in.
+ *
+ * If @is_in_root is 'true', @entry is in the index root attribute @ir described
+ * by the attribute search context @actx and the base inode @base_ni.  @ia and
+ * @page are NULL in this case.
+ *
+ * If @is_in_root is 'false', @entry is in the index allocation attribute and @ia
+ * and @page point to the index allocation block and the mapped, locked page it
+ * is in, respectively.  @ir, @actx and @base_ni are NULL in this case.
+ *
+ * To obtain a context call ntfs_index_ctx_get().
+ *
+ * We use this context to allow ntfs_index_lookup() to return the found index
+ * @entry and its @data without having to allocate a buffer and copy the @entry
+ * and/or its @data into it.
+ *
+ * When finished with the @entry and its @data, call ntfs_index_ctx_put() to
+ * free the context and other associated resources.
+ *
+ * If the index entry was modified, call flush_dcache_index_entry_page()
+ * immediately after the modification and either ntfs_index_entry_mark_dirty()
+ * or ntfs_index_entry_write() before the call to ntfs_index_ctx_put() to
+ * ensure that the changes are written to disk.
+ */
+typedef struct {
+	ntfs_inode *idx_ni;
+	INDEX_ENTRY *entry;
+	void *data;
+	u16 data_len;
+	bool is_in_root;
+	INDEX_ROOT *ir;
+	ntfs_attr_search_ctx *actx;
+	ntfs_inode *base_ni;
+	INDEX_ALLOCATION *ia;
+	struct page *page;
+} ntfs_index_context;
+
+extern ntfs_index_context *ntfs_index_ctx_get(ntfs_inode *idx_ni);
+extern void ntfs_index_ctx_put(ntfs_index_context *ictx);
+
+extern int ntfs_index_lookup(const void *key, const int key_len,
+		ntfs_index_context *ictx);
+
+#ifdef NTFS_RW
+
+/**
+ * ntfs_index_entry_flush_dcache_page - flush_dcache_page() for index entries
+ * @ictx:	ntfs index context describing the index entry
+ *
+ * Call flush_dcache_page() for the page in which an index entry resides.
+ *
+ * This must be called every time an index entry is modified, just after the
+ * modification.
+ *
+ * If the index entry is in the index root attribute, simply flush the page
+ * containing the mft record containing the index root attribute.
+ *
+ * If the index entry is in an index block belonging to the index allocation
+ * attribute, simply flush the page cache page containing the index block.
+ */
+static inline void ntfs_index_entry_flush_dcache_page(ntfs_index_context *ictx)
+{
+	if (ictx->is_in_root)
+		flush_dcache_mft_record_page(ictx->actx->ntfs_ino);
+	else
+		flush_dcache_page(ictx->page);
+}
+
+/**
+ * ntfs_index_entry_mark_dirty - mark an index entry dirty
+ * @ictx:	ntfs index context describing the index entry
+ *
+ * Mark the index entry described by the index entry context @ictx dirty.
+ *
+ * If the index entry is in the index root attribute, simply mark the mft
+ * record containing the index root attribute dirty.  This ensures the mft
+ * record, and hence the index root attribute, will be written out to disk
+ * later.
+ *
+ * If the index entry is in an index block belonging to the index allocation
+ * attribute, mark the buffers belonging to the index record as well as the
+ * page cache page the index block is in dirty.  This automatically marks the
+ * VFS inode of the ntfs index inode to which the index entry belongs dirty,
+ * too (I_DIRTY_PAGES) and this in turn ensures the page buffers, and hence the
+ * dirty index block, will be written out to disk later.
+ */
+static inline void ntfs_index_entry_mark_dirty(ntfs_index_context *ictx)
+{
+	if (ictx->is_in_root)
+		mark_mft_record_dirty(ictx->actx->ntfs_ino);
+	else
+		mark_ntfs_record_dirty(ictx->page,
+				(u8*)ictx->ia - (u8*)page_address(ictx->page));
+}
+
+#endif /* NTFS_RW */
+
+#endif /* _LINUX_NTFS_INDEX_H */
diff --git a/fs/ntfs/inode.c b/fs/ntfs/inode.c
new file mode 100644
index 000000000..645c4b1b2
--- /dev/null
+++ b/fs/ntfs/inode.c
@@ -0,0 +1,3100 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/**
+ * inode.c - NTFS kernel inode handling.
+ *
+ * Copyright (c) 2001-2014 Anton Altaparmakov and Tuxera Inc.
+ */
+
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/mount.h>
+#include <linux/mutex.h>
+#include <linux/pagemap.h>
+#include <linux/quotaops.h>
+#include <linux/slab.h>
+#include <linux/log2.h>
+
+#include "aops.h"
+#include "attrib.h"
+#include "bitmap.h"
+#include "dir.h"
+#include "debug.h"
+#include "inode.h"
+#include "lcnalloc.h"
+#include "malloc.h"
+#include "mft.h"
+#include "time.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_test_inode - compare two (possibly fake) inodes for equality
+ * @vi:		vfs inode which to test
+ * @data:	data which is being tested with
+ *
+ * Compare the ntfs attribute embedded in the ntfs specific part of the vfs
+ * inode @vi for equality with the ntfs attribute @data.
+ *
+ * If searching for the normal file/directory inode, set @na->type to AT_UNUSED.
+ * @na->name and @na->name_len are then ignored.
+ *
+ * Return 1 if the attributes match and 0 if not.
+ *
+ * NOTE: This function runs with the inode_hash_lock spin lock held so it is not
+ * allowed to sleep.
+ */
+int ntfs_test_inode(struct inode *vi, void *data)
+{
+	ntfs_attr *na = (ntfs_attr *)data;
+	ntfs_inode *ni;
+
+	if (vi->i_ino != na->mft_no)
+		return 0;
+	ni = NTFS_I(vi);
+	/* If !NInoAttr(ni), @vi is a normal file or directory inode. */
+	if (likely(!NInoAttr(ni))) {
+		/* If not looking for a normal inode this is a mismatch. */
+		if (unlikely(na->type != AT_UNUSED))
+			return 0;
+	} else {
+		/* A fake inode describing an attribute. */
+		if (ni->type != na->type)
+			return 0;
+		if (ni->name_len != na->name_len)
+			return 0;
+		if (na->name_len && memcmp(ni->name, na->name,
+				na->name_len * sizeof(ntfschar)))
+			return 0;
+	}
+	/* Match! */
+	return 1;
+}
+
+/**
+ * ntfs_init_locked_inode - initialize an inode
+ * @vi:		vfs inode to initialize
+ * @data:	data which to initialize @vi to
+ *
+ * Initialize the vfs inode @vi with the values from the ntfs attribute @data in
+ * order to enable ntfs_test_inode() to do its work.
+ *
+ * If initializing the normal file/directory inode, set @na->type to AT_UNUSED.
+ * In that case, @na->name and @na->name_len should be set to NULL and 0,
+ * respectively. Although that is not strictly necessary as
+ * ntfs_read_locked_inode() will fill them in later.
+ *
+ * Return 0 on success and -errno on error.
+ *
+ * NOTE: This function runs with the inode->i_lock spin lock held so it is not
+ * allowed to sleep. (Hence the GFP_ATOMIC allocation.)
+ */
+static int ntfs_init_locked_inode(struct inode *vi, void *data)
+{
+	ntfs_attr *na = (ntfs_attr *)data;
+	ntfs_inode *ni = NTFS_I(vi);
+
+	vi->i_ino = na->mft_no;
+
+	ni->type = na->type;
+	if (na->type == AT_INDEX_ALLOCATION)
+		NInoSetMstProtected(ni);
+
+	ni->name = na->name;
+	ni->name_len = na->name_len;
+
+	/* If initializing a normal inode, we are done. */
+	if (likely(na->type == AT_UNUSED)) {
+		BUG_ON(na->name);
+		BUG_ON(na->name_len);
+		return 0;
+	}
+
+	/* It is a fake inode. */
+	NInoSetAttr(ni);
+
+	/*
+	 * We have I30 global constant as an optimization as it is the name
+	 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC
+	 * allocation but that is ok. And most attributes are unnamed anyway,
+	 * thus the fraction of named attributes with name != I30 is actually
+	 * absolutely tiny.
+	 */
+	if (na->name_len && na->name != I30) {
+		unsigned int i;
+
+		BUG_ON(!na->name);
+		i = na->name_len * sizeof(ntfschar);
+		ni->name = kmalloc(i + sizeof(ntfschar), GFP_ATOMIC);
+		if (!ni->name)
+			return -ENOMEM;
+		memcpy(ni->name, na->name, i);
+		ni->name[na->name_len] = 0;
+	}
+	return 0;
+}
+
+static int ntfs_read_locked_inode(struct inode *vi);
+static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi);
+static int ntfs_read_locked_index_inode(struct inode *base_vi,
+		struct inode *vi);
+
+/**
+ * ntfs_iget - obtain a struct inode corresponding to a specific normal inode
+ * @sb:		super block of mounted volume
+ * @mft_no:	mft record number / inode number to obtain
+ *
+ * Obtain the struct inode corresponding to a specific normal inode (i.e. a
+ * file or directory).
+ *
+ * If the inode is in the cache, it is just returned with an increased
+ * reference count. Otherwise, a new struct inode is allocated and initialized,
+ * and finally ntfs_read_locked_inode() is called to read in the inode and
+ * fill in the remainder of the inode structure.
+ *
+ * Return the struct inode on success. Check the return value with IS_ERR() and
+ * if true, the function failed and the error code is obtained from PTR_ERR().
+ */
+struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no)
+{
+	struct inode *vi;
+	int err;
+	ntfs_attr na;
+
+	na.mft_no = mft_no;
+	na.type = AT_UNUSED;
+	na.name = NULL;
+	na.name_len = 0;
+
+	vi = iget5_locked(sb, mft_no, ntfs_test_inode,
+			ntfs_init_locked_inode, &na);
+	if (unlikely(!vi))
+		return ERR_PTR(-ENOMEM);
+
+	err = 0;
+
+	/* If this is a freshly allocated inode, need to read it now. */
+	if (vi->i_state & I_NEW) {
+		err = ntfs_read_locked_inode(vi);
+		unlock_new_inode(vi);
+	}
+	/*
+	 * There is no point in keeping bad inodes around if the failure was
+	 * due to ENOMEM. We want to be able to retry again later.
+	 */
+	if (unlikely(err == -ENOMEM)) {
+		iput(vi);
+		vi = ERR_PTR(err);
+	}
+	return vi;
+}
+
+/**
+ * ntfs_attr_iget - obtain a struct inode corresponding to an attribute
+ * @base_vi:	vfs base inode containing the attribute
+ * @type:	attribute type
+ * @name:	Unicode name of the attribute (NULL if unnamed)
+ * @name_len:	length of @name in Unicode characters (0 if unnamed)
+ *
+ * Obtain the (fake) struct inode corresponding to the attribute specified by
+ * @type, @name, and @name_len, which is present in the base mft record
+ * specified by the vfs inode @base_vi.
+ *
+ * If the attribute inode is in the cache, it is just returned with an
+ * increased reference count. Otherwise, a new struct inode is allocated and
+ * initialized, and finally ntfs_read_locked_attr_inode() is called to read the
+ * attribute and fill in the inode structure.
+ *
+ * Note, for index allocation attributes, you need to use ntfs_index_iget()
+ * instead of ntfs_attr_iget() as working with indices is a lot more complex.
+ *
+ * Return the struct inode of the attribute inode on success. Check the return
+ * value with IS_ERR() and if true, the function failed and the error code is
+ * obtained from PTR_ERR().
+ */
+struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
+		ntfschar *name, u32 name_len)
+{
+	struct inode *vi;
+	int err;
+	ntfs_attr na;
+
+	/* Make sure no one calls ntfs_attr_iget() for indices. */
+	BUG_ON(type == AT_INDEX_ALLOCATION);
+
+	na.mft_no = base_vi->i_ino;
+	na.type = type;
+	na.name = name;
+	na.name_len = name_len;
+
+	vi = iget5_locked(base_vi->i_sb, na.mft_no, ntfs_test_inode,
+			ntfs_init_locked_inode, &na);
+	if (unlikely(!vi))
+		return ERR_PTR(-ENOMEM);
+
+	err = 0;
+
+	/* If this is a freshly allocated inode, need to read it now. */
+	if (vi->i_state & I_NEW) {
+		err = ntfs_read_locked_attr_inode(base_vi, vi);
+		unlock_new_inode(vi);
+	}
+	/*
+	 * There is no point in keeping bad attribute inodes around. This also
+	 * simplifies things in that we never need to check for bad attribute
+	 * inodes elsewhere.
+	 */
+	if (unlikely(err)) {
+		iput(vi);
+		vi = ERR_PTR(err);
+	}
+	return vi;
+}
+
+/**
+ * ntfs_index_iget - obtain a struct inode corresponding to an index
+ * @base_vi:	vfs base inode containing the index related attributes
+ * @name:	Unicode name of the index
+ * @name_len:	length of @name in Unicode characters
+ *
+ * Obtain the (fake) struct inode corresponding to the index specified by @name
+ * and @name_len, which is present in the base mft record specified by the vfs
+ * inode @base_vi.
+ *
+ * If the index inode is in the cache, it is just returned with an increased
+ * reference count.  Otherwise, a new struct inode is allocated and
+ * initialized, and finally ntfs_read_locked_index_inode() is called to read
+ * the index related attributes and fill in the inode structure.
+ *
+ * Return the struct inode of the index inode on success. Check the return
+ * value with IS_ERR() and if true, the function failed and the error code is
+ * obtained from PTR_ERR().
+ */
+struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
+		u32 name_len)
+{
+	struct inode *vi;
+	int err;
+	ntfs_attr na;
+
+	na.mft_no = base_vi->i_ino;
+	na.type = AT_INDEX_ALLOCATION;
+	na.name = name;
+	na.name_len = name_len;
+
+	vi = iget5_locked(base_vi->i_sb, na.mft_no, ntfs_test_inode,
+			ntfs_init_locked_inode, &na);
+	if (unlikely(!vi))
+		return ERR_PTR(-ENOMEM);
+
+	err = 0;
+
+	/* If this is a freshly allocated inode, need to read it now. */
+	if (vi->i_state & I_NEW) {
+		err = ntfs_read_locked_index_inode(base_vi, vi);
+		unlock_new_inode(vi);
+	}
+	/*
+	 * There is no point in keeping bad index inodes around.  This also
+	 * simplifies things in that we never need to check for bad index
+	 * inodes elsewhere.
+	 */
+	if (unlikely(err)) {
+		iput(vi);
+		vi = ERR_PTR(err);
+	}
+	return vi;
+}
+
+struct inode *ntfs_alloc_big_inode(struct super_block *sb)
+{
+	ntfs_inode *ni;
+
+	ntfs_debug("Entering.");
+	ni = kmem_cache_alloc(ntfs_big_inode_cache, GFP_NOFS);
+	if (likely(ni != NULL)) {
+		ni->state = 0;
+		return VFS_I(ni);
+	}
+	ntfs_error(sb, "Allocation of NTFS big inode structure failed.");
+	return NULL;
+}
+
+void ntfs_free_big_inode(struct inode *inode)
+{
+	kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode));
+}
+
+static inline ntfs_inode *ntfs_alloc_extent_inode(void)
+{
+	ntfs_inode *ni;
+
+	ntfs_debug("Entering.");
+	ni = kmem_cache_alloc(ntfs_inode_cache, GFP_NOFS);
+	if (likely(ni != NULL)) {
+		ni->state = 0;
+		return ni;
+	}
+	ntfs_error(NULL, "Allocation of NTFS inode structure failed.");
+	return NULL;
+}
+
+static void ntfs_destroy_extent_inode(ntfs_inode *ni)
+{
+	ntfs_debug("Entering.");
+	BUG_ON(ni->page);
+	if (!atomic_dec_and_test(&ni->count))
+		BUG();
+	kmem_cache_free(ntfs_inode_cache, ni);
+}
+
+/*
+ * The attribute runlist lock has separate locking rules from the
+ * normal runlist lock, so split the two lock-classes:
+ */
+static struct lock_class_key attr_list_rl_lock_class;
+
+/**
+ * __ntfs_init_inode - initialize ntfs specific part of an inode
+ * @sb:		super block of mounted volume
+ * @ni:		freshly allocated ntfs inode which to initialize
+ *
+ * Initialize an ntfs inode to defaults.
+ *
+ * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left
+ * untouched. Make sure to initialize them elsewhere.
+ *
+ * Return zero on success and -ENOMEM on error.
+ */
+void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni)
+{
+	ntfs_debug("Entering.");
+	rwlock_init(&ni->size_lock);
+	ni->initialized_size = ni->allocated_size = 0;
+	ni->seq_no = 0;
+	atomic_set(&ni->count, 1);
+	ni->vol = NTFS_SB(sb);
+	ntfs_init_runlist(&ni->runlist);
+	mutex_init(&ni->mrec_lock);
+	ni->page = NULL;
+	ni->page_ofs = 0;
+	ni->attr_list_size = 0;
+	ni->attr_list = NULL;
+	ntfs_init_runlist(&ni->attr_list_rl);
+	lockdep_set_class(&ni->attr_list_rl.lock,
+				&attr_list_rl_lock_class);
+	ni->itype.index.block_size = 0;
+	ni->itype.index.vcn_size = 0;
+	ni->itype.index.collation_rule = 0;
+	ni->itype.index.block_size_bits = 0;
+	ni->itype.index.vcn_size_bits = 0;
+	mutex_init(&ni->extent_lock);
+	ni->nr_extents = 0;
+	ni->ext.base_ntfs_ino = NULL;
+}
+
+/*
+ * Extent inodes get MFT-mapped in a nested way, while the base inode
+ * is still mapped. Teach this nesting to the lock validator by creating
+ * a separate class for nested inode's mrec_lock's:
+ */
+static struct lock_class_key extent_inode_mrec_lock_key;
+
+inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
+		unsigned long mft_no)
+{
+	ntfs_inode *ni = ntfs_alloc_extent_inode();
+
+	ntfs_debug("Entering.");
+	if (likely(ni != NULL)) {
+		__ntfs_init_inode(sb, ni);
+		lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key);
+		ni->mft_no = mft_no;
+		ni->type = AT_UNUSED;
+		ni->name = NULL;
+		ni->name_len = 0;
+	}
+	return ni;
+}
+
+/**
+ * ntfs_is_extended_system_file - check if a file is in the $Extend directory
+ * @ctx:	initialized attribute search context
+ *
+ * Search all file name attributes in the inode described by the attribute
+ * search context @ctx and check if any of the names are in the $Extend system
+ * directory.
+ *
+ * Return values:
+ *	   1: file is in $Extend directory
+ *	   0: file is not in $Extend directory
+ *    -errno: failed to determine if the file is in the $Extend directory
+ */
+static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx)
+{
+	int nr_links, err;
+
+	/* Restart search. */
+	ntfs_attr_reinit_search_ctx(ctx);
+
+	/* Get number of hard links. */
+	nr_links = le16_to_cpu(ctx->mrec->link_count);
+
+	/* Loop through all hard links. */
+	while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0,
+			ctx))) {
+		FILE_NAME_ATTR *file_name_attr;
+		ATTR_RECORD *attr = ctx->attr;
+		u8 *p, *p2;
+
+		nr_links--;
+		/*
+		 * Maximum sanity checking as we are called on an inode that
+		 * we suspect might be corrupt.
+		 */
+		p = (u8*)attr + le32_to_cpu(attr->length);
+		if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec +
+				le32_to_cpu(ctx->mrec->bytes_in_use)) {
+err_corrupt_attr:
+			ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name "
+					"attribute. You should run chkdsk.");
+			return -EIO;
+		}
+		if (attr->non_resident) {
+			ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file "
+					"name. You should run chkdsk.");
+			return -EIO;
+		}
+		if (attr->flags) {
+			ntfs_error(ctx->ntfs_ino->vol->sb, "File name with "
+					"invalid flags. You should run "
+					"chkdsk.");
+			return -EIO;
+		}
+		if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) {
+			ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file "
+					"name. You should run chkdsk.");
+			return -EIO;
+		}
+		file_name_attr = (FILE_NAME_ATTR*)((u8*)attr +
+				le16_to_cpu(attr->data.resident.value_offset));
+		p2 = (u8 *)file_name_attr + le32_to_cpu(attr->data.resident.value_length);
+		if (p2 < (u8*)attr || p2 > p)
+			goto err_corrupt_attr;
+		/* This attribute is ok, but is it in the $Extend directory? */
+		if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend)
+			return 1;	/* YES, it's an extended system file. */
+	}
+	if (unlikely(err != -ENOENT))
+		return err;
+	if (unlikely(nr_links)) {
+		ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count "
+				"doesn't match number of name attributes. You "
+				"should run chkdsk.");
+		return -EIO;
+	}
+	return 0;	/* NO, it is not an extended system file. */
+}
+
+/**
+ * ntfs_read_locked_inode - read an inode from its device
+ * @vi:		inode to read
+ *
+ * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode
+ * described by @vi into memory from the device.
+ *
+ * The only fields in @vi that we need to/can look at when the function is
+ * called are i_sb, pointing to the mounted device's super block, and i_ino,
+ * the number of the inode to load.
+ *
+ * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino
+ * for reading and sets up the necessary @vi fields as well as initializing
+ * the ntfs inode.
+ *
+ * Q: What locks are held when the function is called?
+ * A: i_state has I_NEW set, hence the inode is locked, also
+ *    i_count is set to 1, so it is not going to go away
+ *    i_flags is set to 0 and we have no business touching it.  Only an ioctl()
+ *    is allowed to write to them. We should of course be honouring them but
+ *    we need to do that using the IS_* macros defined in include/linux/fs.h.
+ *    In any case ntfs_read_locked_inode() has nothing to do with i_flags.
+ *
+ * Return 0 on success and -errno on error.  In the error case, the inode will
+ * have had make_bad_inode() executed on it.
+ */
+static int ntfs_read_locked_inode(struct inode *vi)
+{
+	ntfs_volume *vol = NTFS_SB(vi->i_sb);
+	ntfs_inode *ni;
+	struct inode *bvi;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	STANDARD_INFORMATION *si;
+	ntfs_attr_search_ctx *ctx;
+	int err = 0;
+
+	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
+
+	/* Setup the generic vfs inode parts now. */
+	vi->i_uid = vol->uid;
+	vi->i_gid = vol->gid;
+	vi->i_mode = 0;
+
+	/*
+	 * Initialize the ntfs specific part of @vi special casing
+	 * FILE_MFT which we need to do at mount time.
+	 */
+	if (vi->i_ino != FILE_MFT)
+		ntfs_init_big_inode(vi);
+	ni = NTFS_I(vi);
+
+	m = map_mft_record(ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		goto err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(ni, m);
+	if (!ctx) {
+		err = -ENOMEM;
+		goto unm_err_out;
+	}
+
+	if (!(m->flags & MFT_RECORD_IN_USE)) {
+		ntfs_error(vi->i_sb, "Inode is not in use!");
+		goto unm_err_out;
+	}
+	if (m->base_mft_record) {
+		ntfs_error(vi->i_sb, "Inode is an extent inode!");
+		goto unm_err_out;
+	}
+
+	/* Transfer information from mft record into vfs and ntfs inodes. */
+	vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
+
+	/*
+	 * FIXME: Keep in mind that link_count is two for files which have both
+	 * a long file name and a short file name as separate entries, so if
+	 * we are hiding short file names this will be too high. Either we need
+	 * to account for the short file names by subtracting them or we need
+	 * to make sure we delete files even though i_nlink is not zero which
+	 * might be tricky due to vfs interactions. Need to think about this
+	 * some more when implementing the unlink command.
+	 */
+	set_nlink(vi, le16_to_cpu(m->link_count));
+	/*
+	 * FIXME: Reparse points can have the directory bit set even though
+	 * they would be S_IFLNK. Need to deal with this further below when we
+	 * implement reparse points / symbolic links but it will do for now.
+	 * Also if not a directory, it could be something else, rather than
+	 * a regular file. But again, will do for now.
+	 */
+	/* Everyone gets all permissions. */
+	vi->i_mode |= S_IRWXUGO;
+	/* If read-only, no one gets write permissions. */
+	if (IS_RDONLY(vi))
+		vi->i_mode &= ~S_IWUGO;
+	if (m->flags & MFT_RECORD_IS_DIRECTORY) {
+		vi->i_mode |= S_IFDIR;
+		/*
+		 * Apply the directory permissions mask set in the mount
+		 * options.
+		 */
+		vi->i_mode &= ~vol->dmask;
+		/* Things break without this kludge! */
+		if (vi->i_nlink > 1)
+			set_nlink(vi, 1);
+	} else {
+		vi->i_mode |= S_IFREG;
+		/* Apply the file permissions mask set in the mount options. */
+		vi->i_mode &= ~vol->fmask;
+	}
+	/*
+	 * Find the standard information attribute in the mft record. At this
+	 * stage we haven't setup the attribute list stuff yet, so this could
+	 * in fact fail if the standard information is in an extent record, but
+	 * I don't think this actually ever happens.
+	 */
+	err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0,
+			ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT) {
+			/*
+			 * TODO: We should be performing a hot fix here (if the
+			 * recover mount option is set) by creating a new
+			 * attribute.
+			 */
+			ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute "
+					"is missing.");
+		}
+		goto unm_err_out;
+	}
+	a = ctx->attr;
+	/* Get the standard information attribute value. */
+	if ((u8 *)a + le16_to_cpu(a->data.resident.value_offset)
+			+ le32_to_cpu(a->data.resident.value_length) >
+			(u8 *)ctx->mrec + vol->mft_record_size) {
+		ntfs_error(vi->i_sb, "Corrupt standard information attribute in inode.");
+		goto unm_err_out;
+	}
+	si = (STANDARD_INFORMATION*)((u8*)a +
+			le16_to_cpu(a->data.resident.value_offset));
+
+	/* Transfer information from the standard information into vi. */
+	/*
+	 * Note: The i_?times do not quite map perfectly onto the NTFS times,
+	 * but they are close enough, and in the end it doesn't really matter
+	 * that much...
+	 */
+	/*
+	 * mtime is the last change of the data within the file. Not changed
+	 * when only metadata is changed, e.g. a rename doesn't affect mtime.
+	 */
+	vi->i_mtime = ntfs2utc(si->last_data_change_time);
+	/*
+	 * ctime is the last change of the metadata of the file. This obviously
+	 * always changes, when mtime is changed. ctime can be changed on its
+	 * own, mtime is then not changed, e.g. when a file is renamed.
+	 */
+	vi->i_ctime = ntfs2utc(si->last_mft_change_time);
+	/*
+	 * Last access to the data within the file. Not changed during a rename
+	 * for example but changed whenever the file is written to.
+	 */
+	vi->i_atime = ntfs2utc(si->last_access_time);
+
+	/* Find the attribute list attribute if present. */
+	ntfs_attr_reinit_search_ctx(ctx);
+	err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
+	if (err) {
+		if (unlikely(err != -ENOENT)) {
+			ntfs_error(vi->i_sb, "Failed to lookup attribute list "
+					"attribute.");
+			goto unm_err_out;
+		}
+	} else /* if (!err) */ {
+		if (vi->i_ino == FILE_MFT)
+			goto skip_attr_list_load;
+		ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino);
+		NInoSetAttrList(ni);
+		a = ctx->attr;
+		if (a->flags & ATTR_COMPRESSION_MASK) {
+			ntfs_error(vi->i_sb, "Attribute list attribute is "
+					"compressed.");
+			goto unm_err_out;
+		}
+		if (a->flags & ATTR_IS_ENCRYPTED ||
+				a->flags & ATTR_IS_SPARSE) {
+			if (a->non_resident) {
+				ntfs_error(vi->i_sb, "Non-resident attribute "
+						"list attribute is encrypted/"
+						"sparse.");
+				goto unm_err_out;
+			}
+			ntfs_warning(vi->i_sb, "Resident attribute list "
+					"attribute in inode 0x%lx is marked "
+					"encrypted/sparse which is not true.  "
+					"However, Windows allows this and "
+					"chkdsk does not detect or correct it "
+					"so we will just ignore the invalid "
+					"flags and pretend they are not set.",
+					vi->i_ino);
+		}
+		/* Now allocate memory for the attribute list. */
+		ni->attr_list_size = (u32)ntfs_attr_size(a);
+		ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
+		if (!ni->attr_list) {
+			ntfs_error(vi->i_sb, "Not enough memory to allocate "
+					"buffer for attribute list.");
+			err = -ENOMEM;
+			goto unm_err_out;
+		}
+		if (a->non_resident) {
+			NInoSetAttrListNonResident(ni);
+			if (a->data.non_resident.lowest_vcn) {
+				ntfs_error(vi->i_sb, "Attribute list has non "
+						"zero lowest_vcn.");
+				goto unm_err_out;
+			}
+			/*
+			 * Setup the runlist. No need for locking as we have
+			 * exclusive access to the inode at this time.
+			 */
+			ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
+					a, NULL);
+			if (IS_ERR(ni->attr_list_rl.rl)) {
+				err = PTR_ERR(ni->attr_list_rl.rl);
+				ni->attr_list_rl.rl = NULL;
+				ntfs_error(vi->i_sb, "Mapping pairs "
+						"decompression failed.");
+				goto unm_err_out;
+			}
+			/* Now load the attribute list. */
+			if ((err = load_attribute_list(vol, &ni->attr_list_rl,
+					ni->attr_list, ni->attr_list_size,
+					sle64_to_cpu(a->data.non_resident.
+					initialized_size)))) {
+				ntfs_error(vi->i_sb, "Failed to load "
+						"attribute list attribute.");
+				goto unm_err_out;
+			}
+		} else /* if (!a->non_resident) */ {
+			if ((u8*)a + le16_to_cpu(a->data.resident.value_offset)
+					+ le32_to_cpu(
+					a->data.resident.value_length) >
+					(u8*)ctx->mrec + vol->mft_record_size) {
+				ntfs_error(vi->i_sb, "Corrupt attribute list "
+						"in inode.");
+				goto unm_err_out;
+			}
+			/* Now copy the attribute list. */
+			memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
+					a->data.resident.value_offset),
+					le32_to_cpu(
+					a->data.resident.value_length));
+		}
+	}
+skip_attr_list_load:
+	/*
+	 * If an attribute list is present we now have the attribute list value
+	 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes.
+	 */
+	if (S_ISDIR(vi->i_mode)) {
+		loff_t bvi_size;
+		ntfs_inode *bni;
+		INDEX_ROOT *ir;
+		u8 *ir_end, *index_end;
+
+		/* It is a directory, find index root attribute. */
+		ntfs_attr_reinit_search_ctx(ctx);
+		err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE,
+				0, NULL, 0, ctx);
+		if (unlikely(err)) {
+			if (err == -ENOENT) {
+				// FIXME: File is corrupt! Hot-fix with empty
+				// index root attribute if recovery option is
+				// set.
+				ntfs_error(vi->i_sb, "$INDEX_ROOT attribute "
+						"is missing.");
+			}
+			goto unm_err_out;
+		}
+		a = ctx->attr;
+		/* Set up the state. */
+		if (unlikely(a->non_resident)) {
+			ntfs_error(vol->sb, "$INDEX_ROOT attribute is not "
+					"resident.");
+			goto unm_err_out;
+		}
+		/* Ensure the attribute name is placed before the value. */
+		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
+				le16_to_cpu(a->data.resident.value_offset)))) {
+			ntfs_error(vol->sb, "$INDEX_ROOT attribute name is "
+					"placed after the attribute value.");
+			goto unm_err_out;
+		}
+		/*
+		 * Compressed/encrypted index root just means that the newly
+		 * created files in that directory should be created compressed/
+		 * encrypted. However index root cannot be both compressed and
+		 * encrypted.
+		 */
+		if (a->flags & ATTR_COMPRESSION_MASK)
+			NInoSetCompressed(ni);
+		if (a->flags & ATTR_IS_ENCRYPTED) {
+			if (a->flags & ATTR_COMPRESSION_MASK) {
+				ntfs_error(vi->i_sb, "Found encrypted and "
+						"compressed attribute.");
+				goto unm_err_out;
+			}
+			NInoSetEncrypted(ni);
+		}
+		if (a->flags & ATTR_IS_SPARSE)
+			NInoSetSparse(ni);
+		ir = (INDEX_ROOT*)((u8*)a +
+				le16_to_cpu(a->data.resident.value_offset));
+		ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
+		if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
+			ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
+					"corrupt.");
+			goto unm_err_out;
+		}
+		index_end = (u8*)&ir->index +
+				le32_to_cpu(ir->index.index_length);
+		if (index_end > ir_end) {
+			ntfs_error(vi->i_sb, "Directory index is corrupt.");
+			goto unm_err_out;
+		}
+		if (ir->type != AT_FILE_NAME) {
+			ntfs_error(vi->i_sb, "Indexed attribute is not "
+					"$FILE_NAME.");
+			goto unm_err_out;
+		}
+		if (ir->collation_rule != COLLATION_FILE_NAME) {
+			ntfs_error(vi->i_sb, "Index collation rule is not "
+					"COLLATION_FILE_NAME.");
+			goto unm_err_out;
+		}
+		ni->itype.index.collation_rule = ir->collation_rule;
+		ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
+		if (ni->itype.index.block_size &
+				(ni->itype.index.block_size - 1)) {
+			ntfs_error(vi->i_sb, "Index block size (%u) is not a "
+					"power of two.",
+					ni->itype.index.block_size);
+			goto unm_err_out;
+		}
+		if (ni->itype.index.block_size > PAGE_SIZE) {
+			ntfs_error(vi->i_sb, "Index block size (%u) > "
+					"PAGE_SIZE (%ld) is not "
+					"supported.  Sorry.",
+					ni->itype.index.block_size,
+					PAGE_SIZE);
+			err = -EOPNOTSUPP;
+			goto unm_err_out;
+		}
+		if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
+			ntfs_error(vi->i_sb, "Index block size (%u) < "
+					"NTFS_BLOCK_SIZE (%i) is not "
+					"supported.  Sorry.",
+					ni->itype.index.block_size,
+					NTFS_BLOCK_SIZE);
+			err = -EOPNOTSUPP;
+			goto unm_err_out;
+		}
+		ni->itype.index.block_size_bits =
+				ffs(ni->itype.index.block_size) - 1;
+		/* Determine the size of a vcn in the directory index. */
+		if (vol->cluster_size <= ni->itype.index.block_size) {
+			ni->itype.index.vcn_size = vol->cluster_size;
+			ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
+		} else {
+			ni->itype.index.vcn_size = vol->sector_size;
+			ni->itype.index.vcn_size_bits = vol->sector_size_bits;
+		}
+
+		/* Setup the index allocation attribute, even if not present. */
+		NInoSetMstProtected(ni);
+		ni->type = AT_INDEX_ALLOCATION;
+		ni->name = I30;
+		ni->name_len = 4;
+
+		if (!(ir->index.flags & LARGE_INDEX)) {
+			/* No index allocation. */
+			vi->i_size = ni->initialized_size =
+					ni->allocated_size = 0;
+			/* We are done with the mft record, so we release it. */
+			ntfs_attr_put_search_ctx(ctx);
+			unmap_mft_record(ni);
+			m = NULL;
+			ctx = NULL;
+			goto skip_large_dir_stuff;
+		} /* LARGE_INDEX: Index allocation present. Setup state. */
+		NInoSetIndexAllocPresent(ni);
+		/* Find index allocation attribute. */
+		ntfs_attr_reinit_search_ctx(ctx);
+		err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4,
+				CASE_SENSITIVE, 0, NULL, 0, ctx);
+		if (unlikely(err)) {
+			if (err == -ENOENT)
+				ntfs_error(vi->i_sb, "$INDEX_ALLOCATION "
+						"attribute is not present but "
+						"$INDEX_ROOT indicated it is.");
+			else
+				ntfs_error(vi->i_sb, "Failed to lookup "
+						"$INDEX_ALLOCATION "
+						"attribute.");
+			goto unm_err_out;
+		}
+		a = ctx->attr;
+		if (!a->non_resident) {
+			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
+					"is resident.");
+			goto unm_err_out;
+		}
+		/*
+		 * Ensure the attribute name is placed before the mapping pairs
+		 * array.
+		 */
+		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
+				le16_to_cpu(
+				a->data.non_resident.mapping_pairs_offset)))) {
+			ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name "
+					"is placed after the mapping pairs "
+					"array.");
+			goto unm_err_out;
+		}
+		if (a->flags & ATTR_IS_ENCRYPTED) {
+			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
+					"is encrypted.");
+			goto unm_err_out;
+		}
+		if (a->flags & ATTR_IS_SPARSE) {
+			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
+					"is sparse.");
+			goto unm_err_out;
+		}
+		if (a->flags & ATTR_COMPRESSION_MASK) {
+			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute "
+					"is compressed.");
+			goto unm_err_out;
+		}
+		if (a->data.non_resident.lowest_vcn) {
+			ntfs_error(vi->i_sb, "First extent of "
+					"$INDEX_ALLOCATION attribute has non "
+					"zero lowest_vcn.");
+			goto unm_err_out;
+		}
+		vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
+		ni->initialized_size = sle64_to_cpu(
+				a->data.non_resident.initialized_size);
+		ni->allocated_size = sle64_to_cpu(
+				a->data.non_resident.allocated_size);
+		/*
+		 * We are done with the mft record, so we release it. Otherwise
+		 * we would deadlock in ntfs_attr_iget().
+		 */
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(ni);
+		m = NULL;
+		ctx = NULL;
+		/* Get the index bitmap attribute inode. */
+		bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4);
+		if (IS_ERR(bvi)) {
+			ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
+			err = PTR_ERR(bvi);
+			goto unm_err_out;
+		}
+		bni = NTFS_I(bvi);
+		if (NInoCompressed(bni) || NInoEncrypted(bni) ||
+				NInoSparse(bni)) {
+			ntfs_error(vi->i_sb, "$BITMAP attribute is compressed "
+					"and/or encrypted and/or sparse.");
+			goto iput_unm_err_out;
+		}
+		/* Consistency check bitmap size vs. index allocation size. */
+		bvi_size = i_size_read(bvi);
+		if ((bvi_size << 3) < (vi->i_size >>
+				ni->itype.index.block_size_bits)) {
+			ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) "
+					"for index allocation (0x%llx).",
+					bvi_size << 3, vi->i_size);
+			goto iput_unm_err_out;
+		}
+		/* No longer need the bitmap attribute inode. */
+		iput(bvi);
+skip_large_dir_stuff:
+		/* Setup the operations for this inode. */
+		vi->i_op = &ntfs_dir_inode_ops;
+		vi->i_fop = &ntfs_dir_ops;
+		vi->i_mapping->a_ops = &ntfs_mst_aops;
+	} else {
+		/* It is a file. */
+		ntfs_attr_reinit_search_ctx(ctx);
+
+		/* Setup the data attribute, even if not present. */
+		ni->type = AT_DATA;
+		ni->name = NULL;
+		ni->name_len = 0;
+
+		/* Find first extent of the unnamed data attribute. */
+		err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx);
+		if (unlikely(err)) {
+			vi->i_size = ni->initialized_size =
+					ni->allocated_size = 0;
+			if (err != -ENOENT) {
+				ntfs_error(vi->i_sb, "Failed to lookup $DATA "
+						"attribute.");
+				goto unm_err_out;
+			}
+			/*
+			 * FILE_Secure does not have an unnamed $DATA
+			 * attribute, so we special case it here.
+			 */
+			if (vi->i_ino == FILE_Secure)
+				goto no_data_attr_special_case;
+			/*
+			 * Most if not all the system files in the $Extend
+			 * system directory do not have unnamed data
+			 * attributes so we need to check if the parent
+			 * directory of the file is FILE_Extend and if it is
+			 * ignore this error. To do this we need to get the
+			 * name of this inode from the mft record as the name
+			 * contains the back reference to the parent directory.
+			 */
+			if (ntfs_is_extended_system_file(ctx) > 0)
+				goto no_data_attr_special_case;
+			// FIXME: File is corrupt! Hot-fix with empty data
+			// attribute if recovery option is set.
+			ntfs_error(vi->i_sb, "$DATA attribute is missing.");
+			goto unm_err_out;
+		}
+		a = ctx->attr;
+		/* Setup the state. */
+		if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
+			if (a->flags & ATTR_COMPRESSION_MASK) {
+				NInoSetCompressed(ni);
+				if (vol->cluster_size > 4096) {
+					ntfs_error(vi->i_sb, "Found "
+							"compressed data but "
+							"compression is "
+							"disabled due to "
+							"cluster size (%i) > "
+							"4kiB.",
+							vol->cluster_size);
+					goto unm_err_out;
+				}
+				if ((a->flags & ATTR_COMPRESSION_MASK)
+						!= ATTR_IS_COMPRESSED) {
+					ntfs_error(vi->i_sb, "Found unknown "
+							"compression method "
+							"or corrupt file.");
+					goto unm_err_out;
+				}
+			}
+			if (a->flags & ATTR_IS_SPARSE)
+				NInoSetSparse(ni);
+		}
+		if (a->flags & ATTR_IS_ENCRYPTED) {
+			if (NInoCompressed(ni)) {
+				ntfs_error(vi->i_sb, "Found encrypted and "
+						"compressed data.");
+				goto unm_err_out;
+			}
+			NInoSetEncrypted(ni);
+		}
+		if (a->non_resident) {
+			NInoSetNonResident(ni);
+			if (NInoCompressed(ni) || NInoSparse(ni)) {
+				if (NInoCompressed(ni) && a->data.non_resident.
+						compression_unit != 4) {
+					ntfs_error(vi->i_sb, "Found "
+							"non-standard "
+							"compression unit (%u "
+							"instead of 4).  "
+							"Cannot handle this.",
+							a->data.non_resident.
+							compression_unit);
+					err = -EOPNOTSUPP;
+					goto unm_err_out;
+				}
+				if (a->data.non_resident.compression_unit) {
+					ni->itype.compressed.block_size = 1U <<
+							(a->data.non_resident.
+							compression_unit +
+							vol->cluster_size_bits);
+					ni->itype.compressed.block_size_bits =
+							ffs(ni->itype.
+							compressed.
+							block_size) - 1;
+					ni->itype.compressed.block_clusters =
+							1U << a->data.
+							non_resident.
+							compression_unit;
+				} else {
+					ni->itype.compressed.block_size = 0;
+					ni->itype.compressed.block_size_bits =
+							0;
+					ni->itype.compressed.block_clusters =
+							0;
+				}
+				ni->itype.compressed.size = sle64_to_cpu(
+						a->data.non_resident.
+						compressed_size);
+			}
+			if (a->data.non_resident.lowest_vcn) {
+				ntfs_error(vi->i_sb, "First extent of $DATA "
+						"attribute has non zero "
+						"lowest_vcn.");
+				goto unm_err_out;
+			}
+			vi->i_size = sle64_to_cpu(
+					a->data.non_resident.data_size);
+			ni->initialized_size = sle64_to_cpu(
+					a->data.non_resident.initialized_size);
+			ni->allocated_size = sle64_to_cpu(
+					a->data.non_resident.allocated_size);
+		} else { /* Resident attribute. */
+			vi->i_size = ni->initialized_size = le32_to_cpu(
+					a->data.resident.value_length);
+			ni->allocated_size = le32_to_cpu(a->length) -
+					le16_to_cpu(
+					a->data.resident.value_offset);
+			if (vi->i_size > ni->allocated_size) {
+				ntfs_error(vi->i_sb, "Resident data attribute "
+						"is corrupt (size exceeds "
+						"allocation).");
+				goto unm_err_out;
+			}
+		}
+no_data_attr_special_case:
+		/* We are done with the mft record, so we release it. */
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(ni);
+		m = NULL;
+		ctx = NULL;
+		/* Setup the operations for this inode. */
+		vi->i_op = &ntfs_file_inode_ops;
+		vi->i_fop = &ntfs_file_ops;
+		vi->i_mapping->a_ops = &ntfs_normal_aops;
+		if (NInoMstProtected(ni))
+			vi->i_mapping->a_ops = &ntfs_mst_aops;
+		else if (NInoCompressed(ni))
+			vi->i_mapping->a_ops = &ntfs_compressed_aops;
+	}
+	/*
+	 * The number of 512-byte blocks used on disk (for stat). This is in so
+	 * far inaccurate as it doesn't account for any named streams or other
+	 * special non-resident attributes, but that is how Windows works, too,
+	 * so we are at least consistent with Windows, if not entirely
+	 * consistent with the Linux Way. Doing it the Linux Way would cause a
+	 * significant slowdown as it would involve iterating over all
+	 * attributes in the mft record and adding the allocated/compressed
+	 * sizes of all non-resident attributes present to give us the Linux
+	 * correct size that should go into i_blocks (after division by 512).
+	 */
+	if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni)))
+		vi->i_blocks = ni->itype.compressed.size >> 9;
+	else
+		vi->i_blocks = ni->allocated_size >> 9;
+	ntfs_debug("Done.");
+	return 0;
+iput_unm_err_out:
+	iput(bvi);
+unm_err_out:
+	if (!err)
+		err = -EIO;
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(ni);
+err_out:
+	ntfs_error(vol->sb, "Failed with error code %i.  Marking corrupt "
+			"inode 0x%lx as bad.  Run chkdsk.", err, vi->i_ino);
+	make_bad_inode(vi);
+	if (err != -EOPNOTSUPP && err != -ENOMEM)
+		NVolSetErrors(vol);
+	return err;
+}
+
+/**
+ * ntfs_read_locked_attr_inode - read an attribute inode from its base inode
+ * @base_vi:	base inode
+ * @vi:		attribute inode to read
+ *
+ * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the
+ * attribute inode described by @vi into memory from the base mft record
+ * described by @base_ni.
+ *
+ * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for
+ * reading and looks up the attribute described by @vi before setting up the
+ * necessary fields in @vi as well as initializing the ntfs inode.
+ *
+ * Q: What locks are held when the function is called?
+ * A: i_state has I_NEW set, hence the inode is locked, also
+ *    i_count is set to 1, so it is not going to go away
+ *
+ * Return 0 on success and -errno on error.  In the error case, the inode will
+ * have had make_bad_inode() executed on it.
+ *
+ * Note this cannot be called for AT_INDEX_ALLOCATION.
+ */
+static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi)
+{
+	ntfs_volume *vol = NTFS_SB(vi->i_sb);
+	ntfs_inode *ni, *base_ni;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	ntfs_attr_search_ctx *ctx;
+	int err = 0;
+
+	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
+
+	ntfs_init_big_inode(vi);
+
+	ni	= NTFS_I(vi);
+	base_ni = NTFS_I(base_vi);
+
+	/* Just mirror the values from the base inode. */
+	vi->i_uid	= base_vi->i_uid;
+	vi->i_gid	= base_vi->i_gid;
+	set_nlink(vi, base_vi->i_nlink);
+	vi->i_mtime	= base_vi->i_mtime;
+	vi->i_ctime	= base_vi->i_ctime;
+	vi->i_atime	= base_vi->i_atime;
+	vi->i_generation = ni->seq_no = base_ni->seq_no;
+
+	/* Set inode type to zero but preserve permissions. */
+	vi->i_mode	= base_vi->i_mode & ~S_IFMT;
+
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		goto err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (!ctx) {
+		err = -ENOMEM;
+		goto unm_err_out;
+	}
+	/* Find the attribute. */
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err))
+		goto unm_err_out;
+	a = ctx->attr;
+	if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) {
+		if (a->flags & ATTR_COMPRESSION_MASK) {
+			NInoSetCompressed(ni);
+			if ((ni->type != AT_DATA) || (ni->type == AT_DATA &&
+					ni->name_len)) {
+				ntfs_error(vi->i_sb, "Found compressed "
+						"non-data or named data "
+						"attribute.  Please report "
+						"you saw this message to "
+						"linux-ntfs-dev@lists."
+						"sourceforge.net");
+				goto unm_err_out;
+			}
+			if (vol->cluster_size > 4096) {
+				ntfs_error(vi->i_sb, "Found compressed "
+						"attribute but compression is "
+						"disabled due to cluster size "
+						"(%i) > 4kiB.",
+						vol->cluster_size);
+				goto unm_err_out;
+			}
+			if ((a->flags & ATTR_COMPRESSION_MASK) !=
+					ATTR_IS_COMPRESSED) {
+				ntfs_error(vi->i_sb, "Found unknown "
+						"compression method.");
+				goto unm_err_out;
+			}
+		}
+		/*
+		 * The compressed/sparse flag set in an index root just means
+		 * to compress all files.
+		 */
+		if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
+			ntfs_error(vi->i_sb, "Found mst protected attribute "
+					"but the attribute is %s.  Please "
+					"report you saw this message to "
+					"linux-ntfs-dev@lists.sourceforge.net",
+					NInoCompressed(ni) ? "compressed" :
+					"sparse");
+			goto unm_err_out;
+		}
+		if (a->flags & ATTR_IS_SPARSE)
+			NInoSetSparse(ni);
+	}
+	if (a->flags & ATTR_IS_ENCRYPTED) {
+		if (NInoCompressed(ni)) {
+			ntfs_error(vi->i_sb, "Found encrypted and compressed "
+					"data.");
+			goto unm_err_out;
+		}
+		/*
+		 * The encryption flag set in an index root just means to
+		 * encrypt all files.
+		 */
+		if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) {
+			ntfs_error(vi->i_sb, "Found mst protected attribute "
+					"but the attribute is encrypted.  "
+					"Please report you saw this message "
+					"to linux-ntfs-dev@lists.sourceforge."
+					"net");
+			goto unm_err_out;
+		}
+		if (ni->type != AT_DATA) {
+			ntfs_error(vi->i_sb, "Found encrypted non-data "
+					"attribute.");
+			goto unm_err_out;
+		}
+		NInoSetEncrypted(ni);
+	}
+	if (!a->non_resident) {
+		/* Ensure the attribute name is placed before the value. */
+		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
+				le16_to_cpu(a->data.resident.value_offset)))) {
+			ntfs_error(vol->sb, "Attribute name is placed after "
+					"the attribute value.");
+			goto unm_err_out;
+		}
+		if (NInoMstProtected(ni)) {
+			ntfs_error(vi->i_sb, "Found mst protected attribute "
+					"but the attribute is resident.  "
+					"Please report you saw this message to "
+					"linux-ntfs-dev@lists.sourceforge.net");
+			goto unm_err_out;
+		}
+		vi->i_size = ni->initialized_size = le32_to_cpu(
+				a->data.resident.value_length);
+		ni->allocated_size = le32_to_cpu(a->length) -
+				le16_to_cpu(a->data.resident.value_offset);
+		if (vi->i_size > ni->allocated_size) {
+			ntfs_error(vi->i_sb, "Resident attribute is corrupt "
+					"(size exceeds allocation).");
+			goto unm_err_out;
+		}
+	} else {
+		NInoSetNonResident(ni);
+		/*
+		 * Ensure the attribute name is placed before the mapping pairs
+		 * array.
+		 */
+		if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
+				le16_to_cpu(
+				a->data.non_resident.mapping_pairs_offset)))) {
+			ntfs_error(vol->sb, "Attribute name is placed after "
+					"the mapping pairs array.");
+			goto unm_err_out;
+		}
+		if (NInoCompressed(ni) || NInoSparse(ni)) {
+			if (NInoCompressed(ni) && a->data.non_resident.
+					compression_unit != 4) {
+				ntfs_error(vi->i_sb, "Found non-standard "
+						"compression unit (%u instead "
+						"of 4).  Cannot handle this.",
+						a->data.non_resident.
+						compression_unit);
+				err = -EOPNOTSUPP;
+				goto unm_err_out;
+			}
+			if (a->data.non_resident.compression_unit) {
+				ni->itype.compressed.block_size = 1U <<
+						(a->data.non_resident.
+						compression_unit +
+						vol->cluster_size_bits);
+				ni->itype.compressed.block_size_bits =
+						ffs(ni->itype.compressed.
+						block_size) - 1;
+				ni->itype.compressed.block_clusters = 1U <<
+						a->data.non_resident.
+						compression_unit;
+			} else {
+				ni->itype.compressed.block_size = 0;
+				ni->itype.compressed.block_size_bits = 0;
+				ni->itype.compressed.block_clusters = 0;
+			}
+			ni->itype.compressed.size = sle64_to_cpu(
+					a->data.non_resident.compressed_size);
+		}
+		if (a->data.non_resident.lowest_vcn) {
+			ntfs_error(vi->i_sb, "First extent of attribute has "
+					"non-zero lowest_vcn.");
+			goto unm_err_out;
+		}
+		vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
+		ni->initialized_size = sle64_to_cpu(
+				a->data.non_resident.initialized_size);
+		ni->allocated_size = sle64_to_cpu(
+				a->data.non_resident.allocated_size);
+	}
+	vi->i_mapping->a_ops = &ntfs_normal_aops;
+	if (NInoMstProtected(ni))
+		vi->i_mapping->a_ops = &ntfs_mst_aops;
+	else if (NInoCompressed(ni))
+		vi->i_mapping->a_ops = &ntfs_compressed_aops;
+	if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT)
+		vi->i_blocks = ni->itype.compressed.size >> 9;
+	else
+		vi->i_blocks = ni->allocated_size >> 9;
+	/*
+	 * Make sure the base inode does not go away and attach it to the
+	 * attribute inode.
+	 */
+	igrab(base_vi);
+	ni->ext.base_ntfs_ino = base_ni;
+	ni->nr_extents = -1;
+
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+
+	ntfs_debug("Done.");
+	return 0;
+
+unm_err_out:
+	if (!err)
+		err = -EIO;
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+err_out:
+	ntfs_error(vol->sb, "Failed with error code %i while reading attribute "
+			"inode (mft_no 0x%lx, type 0x%x, name_len %i).  "
+			"Marking corrupt inode and base inode 0x%lx as bad.  "
+			"Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len,
+			base_vi->i_ino);
+	make_bad_inode(vi);
+	if (err != -ENOMEM)
+		NVolSetErrors(vol);
+	return err;
+}
+
+/**
+ * ntfs_read_locked_index_inode - read an index inode from its base inode
+ * @base_vi:	base inode
+ * @vi:		index inode to read
+ *
+ * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the
+ * index inode described by @vi into memory from the base mft record described
+ * by @base_ni.
+ *
+ * ntfs_read_locked_index_inode() maps, pins and locks the base inode for
+ * reading and looks up the attributes relating to the index described by @vi
+ * before setting up the necessary fields in @vi as well as initializing the
+ * ntfs inode.
+ *
+ * Note, index inodes are essentially attribute inodes (NInoAttr() is true)
+ * with the attribute type set to AT_INDEX_ALLOCATION.  Apart from that, they
+ * are setup like directory inodes since directories are a special case of
+ * indices ao they need to be treated in much the same way.  Most importantly,
+ * for small indices the index allocation attribute might not actually exist.
+ * However, the index root attribute always exists but this does not need to
+ * have an inode associated with it and this is why we define a new inode type
+ * index.  Also, like for directories, we need to have an attribute inode for
+ * the bitmap attribute corresponding to the index allocation attribute and we
+ * can store this in the appropriate field of the inode, just like we do for
+ * normal directory inodes.
+ *
+ * Q: What locks are held when the function is called?
+ * A: i_state has I_NEW set, hence the inode is locked, also
+ *    i_count is set to 1, so it is not going to go away
+ *
+ * Return 0 on success and -errno on error.  In the error case, the inode will
+ * have had make_bad_inode() executed on it.
+ */
+static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi)
+{
+	loff_t bvi_size;
+	ntfs_volume *vol = NTFS_SB(vi->i_sb);
+	ntfs_inode *ni, *base_ni, *bni;
+	struct inode *bvi;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	ntfs_attr_search_ctx *ctx;
+	INDEX_ROOT *ir;
+	u8 *ir_end, *index_end;
+	int err = 0;
+
+	ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino);
+	ntfs_init_big_inode(vi);
+	ni	= NTFS_I(vi);
+	base_ni = NTFS_I(base_vi);
+	/* Just mirror the values from the base inode. */
+	vi->i_uid	= base_vi->i_uid;
+	vi->i_gid	= base_vi->i_gid;
+	set_nlink(vi, base_vi->i_nlink);
+	vi->i_mtime	= base_vi->i_mtime;
+	vi->i_ctime	= base_vi->i_ctime;
+	vi->i_atime	= base_vi->i_atime;
+	vi->i_generation = ni->seq_no = base_ni->seq_no;
+	/* Set inode type to zero but preserve permissions. */
+	vi->i_mode	= base_vi->i_mode & ~S_IFMT;
+	/* Map the mft record for the base inode. */
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		goto err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (!ctx) {
+		err = -ENOMEM;
+		goto unm_err_out;
+	}
+	/* Find the index root attribute. */
+	err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT)
+			ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is "
+					"missing.");
+		goto unm_err_out;
+	}
+	a = ctx->attr;
+	/* Set up the state. */
+	if (unlikely(a->non_resident)) {
+		ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident.");
+		goto unm_err_out;
+	}
+	/* Ensure the attribute name is placed before the value. */
+	if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
+			le16_to_cpu(a->data.resident.value_offset)))) {
+		ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed "
+				"after the attribute value.");
+		goto unm_err_out;
+	}
+	/*
+	 * Compressed/encrypted/sparse index root is not allowed, except for
+	 * directories of course but those are not dealt with here.
+	 */
+	if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED |
+			ATTR_IS_SPARSE)) {
+		ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index "
+				"root attribute.");
+		goto unm_err_out;
+	}
+	ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset));
+	ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length);
+	if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) {
+		ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt.");
+		goto unm_err_out;
+	}
+	index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length);
+	if (index_end > ir_end) {
+		ntfs_error(vi->i_sb, "Index is corrupt.");
+		goto unm_err_out;
+	}
+	if (ir->type) {
+		ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).",
+				le32_to_cpu(ir->type));
+		goto unm_err_out;
+	}
+	ni->itype.index.collation_rule = ir->collation_rule;
+	ntfs_debug("Index collation rule is 0x%x.",
+			le32_to_cpu(ir->collation_rule));
+	ni->itype.index.block_size = le32_to_cpu(ir->index_block_size);
+	if (!is_power_of_2(ni->itype.index.block_size)) {
+		ntfs_error(vi->i_sb, "Index block size (%u) is not a power of "
+				"two.", ni->itype.index.block_size);
+		goto unm_err_out;
+	}
+	if (ni->itype.index.block_size > PAGE_SIZE) {
+		ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_SIZE "
+				"(%ld) is not supported.  Sorry.",
+				ni->itype.index.block_size, PAGE_SIZE);
+		err = -EOPNOTSUPP;
+		goto unm_err_out;
+	}
+	if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) {
+		ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE "
+				"(%i) is not supported.  Sorry.",
+				ni->itype.index.block_size, NTFS_BLOCK_SIZE);
+		err = -EOPNOTSUPP;
+		goto unm_err_out;
+	}
+	ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1;
+	/* Determine the size of a vcn in the index. */
+	if (vol->cluster_size <= ni->itype.index.block_size) {
+		ni->itype.index.vcn_size = vol->cluster_size;
+		ni->itype.index.vcn_size_bits = vol->cluster_size_bits;
+	} else {
+		ni->itype.index.vcn_size = vol->sector_size;
+		ni->itype.index.vcn_size_bits = vol->sector_size_bits;
+	}
+	/* Check for presence of index allocation attribute. */
+	if (!(ir->index.flags & LARGE_INDEX)) {
+		/* No index allocation. */
+		vi->i_size = ni->initialized_size = ni->allocated_size = 0;
+		/* We are done with the mft record, so we release it. */
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(base_ni);
+		m = NULL;
+		ctx = NULL;
+		goto skip_large_index_stuff;
+	} /* LARGE_INDEX:  Index allocation present.  Setup state. */
+	NInoSetIndexAllocPresent(ni);
+	/* Find index allocation attribute. */
+	ntfs_attr_reinit_search_ctx(ctx);
+	err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT)
+			ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
+					"not present but $INDEX_ROOT "
+					"indicated it is.");
+		else
+			ntfs_error(vi->i_sb, "Failed to lookup "
+					"$INDEX_ALLOCATION attribute.");
+		goto unm_err_out;
+	}
+	a = ctx->attr;
+	if (!a->non_resident) {
+		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
+				"resident.");
+		goto unm_err_out;
+	}
+	/*
+	 * Ensure the attribute name is placed before the mapping pairs array.
+	 */
+	if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >=
+			le16_to_cpu(
+			a->data.non_resident.mapping_pairs_offset)))) {
+		ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is "
+				"placed after the mapping pairs array.");
+		goto unm_err_out;
+	}
+	if (a->flags & ATTR_IS_ENCRYPTED) {
+		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
+				"encrypted.");
+		goto unm_err_out;
+	}
+	if (a->flags & ATTR_IS_SPARSE) {
+		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse.");
+		goto unm_err_out;
+	}
+	if (a->flags & ATTR_COMPRESSION_MASK) {
+		ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is "
+				"compressed.");
+		goto unm_err_out;
+	}
+	if (a->data.non_resident.lowest_vcn) {
+		ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION "
+				"attribute has non zero lowest_vcn.");
+		goto unm_err_out;
+	}
+	vi->i_size = sle64_to_cpu(a->data.non_resident.data_size);
+	ni->initialized_size = sle64_to_cpu(
+			a->data.non_resident.initialized_size);
+	ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size);
+	/*
+	 * We are done with the mft record, so we release it.  Otherwise
+	 * we would deadlock in ntfs_attr_iget().
+	 */
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+	m = NULL;
+	ctx = NULL;
+	/* Get the index bitmap attribute inode. */
+	bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len);
+	if (IS_ERR(bvi)) {
+		ntfs_error(vi->i_sb, "Failed to get bitmap attribute.");
+		err = PTR_ERR(bvi);
+		goto unm_err_out;
+	}
+	bni = NTFS_I(bvi);
+	if (NInoCompressed(bni) || NInoEncrypted(bni) ||
+			NInoSparse(bni)) {
+		ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or "
+				"encrypted and/or sparse.");
+		goto iput_unm_err_out;
+	}
+	/* Consistency check bitmap size vs. index allocation size. */
+	bvi_size = i_size_read(bvi);
+	if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) {
+		ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for "
+				"index allocation (0x%llx).", bvi_size << 3,
+				vi->i_size);
+		goto iput_unm_err_out;
+	}
+	iput(bvi);
+skip_large_index_stuff:
+	/* Setup the operations for this index inode. */
+	vi->i_mapping->a_ops = &ntfs_mst_aops;
+	vi->i_blocks = ni->allocated_size >> 9;
+	/*
+	 * Make sure the base inode doesn't go away and attach it to the
+	 * index inode.
+	 */
+	igrab(base_vi);
+	ni->ext.base_ntfs_ino = base_ni;
+	ni->nr_extents = -1;
+
+	ntfs_debug("Done.");
+	return 0;
+iput_unm_err_out:
+	iput(bvi);
+unm_err_out:
+	if (!err)
+		err = -EIO;
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(base_ni);
+err_out:
+	ntfs_error(vi->i_sb, "Failed with error code %i while reading index "
+			"inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino,
+			ni->name_len);
+	make_bad_inode(vi);
+	if (err != -EOPNOTSUPP && err != -ENOMEM)
+		NVolSetErrors(vol);
+	return err;
+}
+
+/*
+ * The MFT inode has special locking, so teach the lock validator
+ * about this by splitting off the locking rules of the MFT from
+ * the locking rules of other inodes. The MFT inode can never be
+ * accessed from the VFS side (or even internally), only by the
+ * map_mft functions.
+ */
+static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key;
+
+/**
+ * ntfs_read_inode_mount - special read_inode for mount time use only
+ * @vi:		inode to read
+ *
+ * Read inode FILE_MFT at mount time, only called with super_block lock
+ * held from within the read_super() code path.
+ *
+ * This function exists because when it is called the page cache for $MFT/$DATA
+ * is not initialized and hence we cannot get at the contents of mft records
+ * by calling map_mft_record*().
+ *
+ * Further it needs to cope with the circular references problem, i.e. cannot
+ * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because
+ * we do not know where the other extent mft records are yet and again, because
+ * we cannot call map_mft_record*() yet.  Obviously this applies only when an
+ * attribute list is actually present in $MFT inode.
+ *
+ * We solve these problems by starting with the $DATA attribute before anything
+ * else and iterating using ntfs_attr_lookup($DATA) over all extents.  As each
+ * extent is found, we ntfs_mapping_pairs_decompress() including the implied
+ * ntfs_runlists_merge().  Each step of the iteration necessarily provides
+ * sufficient information for the next step to complete.
+ *
+ * This should work but there are two possible pit falls (see inline comments
+ * below), but only time will tell if they are real pits or just smoke...
+ */
+int ntfs_read_inode_mount(struct inode *vi)
+{
+	VCN next_vcn, last_vcn, highest_vcn;
+	s64 block;
+	struct super_block *sb = vi->i_sb;
+	ntfs_volume *vol = NTFS_SB(sb);
+	struct buffer_head *bh;
+	ntfs_inode *ni;
+	MFT_RECORD *m = NULL;
+	ATTR_RECORD *a;
+	ntfs_attr_search_ctx *ctx;
+	unsigned int i, nr_blocks;
+	int err;
+
+	ntfs_debug("Entering.");
+
+	/* Initialize the ntfs specific part of @vi. */
+	ntfs_init_big_inode(vi);
+
+	ni = NTFS_I(vi);
+
+	/* Setup the data attribute. It is special as it is mst protected. */
+	NInoSetNonResident(ni);
+	NInoSetMstProtected(ni);
+	NInoSetSparseDisabled(ni);
+	ni->type = AT_DATA;
+	ni->name = NULL;
+	ni->name_len = 0;
+	/*
+	 * This sets up our little cheat allowing us to reuse the async read io
+	 * completion handler for directories.
+	 */
+	ni->itype.index.block_size = vol->mft_record_size;
+	ni->itype.index.block_size_bits = vol->mft_record_size_bits;
+
+	/* Very important! Needed to be able to call map_mft_record*(). */
+	vol->mft_ino = vi;
+
+	/* Allocate enough memory to read the first mft record. */
+	if (vol->mft_record_size > 64 * 1024) {
+		ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).",
+				vol->mft_record_size);
+		goto err_out;
+	}
+	i = vol->mft_record_size;
+	if (i < sb->s_blocksize)
+		i = sb->s_blocksize;
+	m = (MFT_RECORD*)ntfs_malloc_nofs(i);
+	if (!m) {
+		ntfs_error(sb, "Failed to allocate buffer for $MFT record 0.");
+		goto err_out;
+	}
+
+	/* Determine the first block of the $MFT/$DATA attribute. */
+	block = vol->mft_lcn << vol->cluster_size_bits >>
+			sb->s_blocksize_bits;
+	nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits;
+	if (!nr_blocks)
+		nr_blocks = 1;
+
+	/* Load $MFT/$DATA's first mft record. */
+	for (i = 0; i < nr_blocks; i++) {
+		bh = sb_bread(sb, block++);
+		if (!bh) {
+			ntfs_error(sb, "Device read failed.");
+			goto err_out;
+		}
+		memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data,
+				sb->s_blocksize);
+		brelse(bh);
+	}
+
+	if (le32_to_cpu(m->bytes_allocated) != vol->mft_record_size) {
+		ntfs_error(sb, "Incorrect mft record size %u in superblock, should be %u.",
+				le32_to_cpu(m->bytes_allocated), vol->mft_record_size);
+		goto err_out;
+	}
+
+	/* Apply the mst fixups. */
+	if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) {
+		/* FIXME: Try to use the $MFTMirr now. */
+		ntfs_error(sb, "MST fixup failed. $MFT is corrupt.");
+		goto err_out;
+	}
+
+	/* Sanity check offset to the first attribute */
+	if (le16_to_cpu(m->attrs_offset) >= le32_to_cpu(m->bytes_allocated)) {
+		ntfs_error(sb, "Incorrect mft offset to the first attribute %u in superblock.",
+			       le16_to_cpu(m->attrs_offset));
+		goto err_out;
+	}
+
+	/* Need this to sanity check attribute list references to $MFT. */
+	vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
+
+	/* Provides readpage() for map_mft_record(). */
+	vi->i_mapping->a_ops = &ntfs_mst_aops;
+
+	ctx = ntfs_attr_get_search_ctx(ni, m);
+	if (!ctx) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+
+	/* Find the attribute list attribute if present. */
+	err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx);
+	if (err) {
+		if (unlikely(err != -ENOENT)) {
+			ntfs_error(sb, "Failed to lookup attribute list "
+					"attribute. You should run chkdsk.");
+			goto put_err_out;
+		}
+	} else /* if (!err) */ {
+		ATTR_LIST_ENTRY *al_entry, *next_al_entry;
+		u8 *al_end;
+		static const char *es = "  Not allowed.  $MFT is corrupt.  "
+				"You should run chkdsk.";
+
+		ntfs_debug("Attribute list attribute found in $MFT.");
+		NInoSetAttrList(ni);
+		a = ctx->attr;
+		if (a->flags & ATTR_COMPRESSION_MASK) {
+			ntfs_error(sb, "Attribute list attribute is "
+					"compressed.%s", es);
+			goto put_err_out;
+		}
+		if (a->flags & ATTR_IS_ENCRYPTED ||
+				a->flags & ATTR_IS_SPARSE) {
+			if (a->non_resident) {
+				ntfs_error(sb, "Non-resident attribute list "
+						"attribute is encrypted/"
+						"sparse.%s", es);
+				goto put_err_out;
+			}
+			ntfs_warning(sb, "Resident attribute list attribute "
+					"in $MFT system file is marked "
+					"encrypted/sparse which is not true.  "
+					"However, Windows allows this and "
+					"chkdsk does not detect or correct it "
+					"so we will just ignore the invalid "
+					"flags and pretend they are not set.");
+		}
+		/* Now allocate memory for the attribute list. */
+		ni->attr_list_size = (u32)ntfs_attr_size(a);
+		if (!ni->attr_list_size) {
+			ntfs_error(sb, "Attr_list_size is zero");
+			goto put_err_out;
+		}
+		ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size);
+		if (!ni->attr_list) {
+			ntfs_error(sb, "Not enough memory to allocate buffer "
+					"for attribute list.");
+			goto put_err_out;
+		}
+		if (a->non_resident) {
+			NInoSetAttrListNonResident(ni);
+			if (a->data.non_resident.lowest_vcn) {
+				ntfs_error(sb, "Attribute list has non zero "
+						"lowest_vcn. $MFT is corrupt. "
+						"You should run chkdsk.");
+				goto put_err_out;
+			}
+			/* Setup the runlist. */
+			ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol,
+					a, NULL);
+			if (IS_ERR(ni->attr_list_rl.rl)) {
+				err = PTR_ERR(ni->attr_list_rl.rl);
+				ni->attr_list_rl.rl = NULL;
+				ntfs_error(sb, "Mapping pairs decompression "
+						"failed with error code %i.",
+						-err);
+				goto put_err_out;
+			}
+			/* Now load the attribute list. */
+			if ((err = load_attribute_list(vol, &ni->attr_list_rl,
+					ni->attr_list, ni->attr_list_size,
+					sle64_to_cpu(a->data.
+					non_resident.initialized_size)))) {
+				ntfs_error(sb, "Failed to load attribute list "
+						"attribute with error code %i.",
+						-err);
+				goto put_err_out;
+			}
+		} else /* if (!ctx.attr->non_resident) */ {
+			if ((u8*)a + le16_to_cpu(
+					a->data.resident.value_offset) +
+					le32_to_cpu(
+					a->data.resident.value_length) >
+					(u8*)ctx->mrec + vol->mft_record_size) {
+				ntfs_error(sb, "Corrupt attribute list "
+						"attribute.");
+				goto put_err_out;
+			}
+			/* Now copy the attribute list. */
+			memcpy(ni->attr_list, (u8*)a + le16_to_cpu(
+					a->data.resident.value_offset),
+					le32_to_cpu(
+					a->data.resident.value_length));
+		}
+		/* The attribute list is now setup in memory. */
+		/*
+		 * FIXME: I don't know if this case is actually possible.
+		 * According to logic it is not possible but I have seen too
+		 * many weird things in MS software to rely on logic... Thus we
+		 * perform a manual search and make sure the first $MFT/$DATA
+		 * extent is in the base inode. If it is not we abort with an
+		 * error and if we ever see a report of this error we will need
+		 * to do some magic in order to have the necessary mft record
+		 * loaded and in the right place in the page cache. But
+		 * hopefully logic will prevail and this never happens...
+		 */
+		al_entry = (ATTR_LIST_ENTRY*)ni->attr_list;
+		al_end = (u8*)al_entry + ni->attr_list_size;
+		for (;; al_entry = next_al_entry) {
+			/* Out of bounds check. */
+			if ((u8*)al_entry < ni->attr_list ||
+					(u8*)al_entry > al_end)
+				goto em_put_err_out;
+			/* Catch the end of the attribute list. */
+			if ((u8*)al_entry == al_end)
+				goto em_put_err_out;
+			if (!al_entry->length)
+				goto em_put_err_out;
+			if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
+					le16_to_cpu(al_entry->length) > al_end)
+				goto em_put_err_out;
+			next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
+					le16_to_cpu(al_entry->length));
+			if (le32_to_cpu(al_entry->type) > le32_to_cpu(AT_DATA))
+				goto em_put_err_out;
+			if (AT_DATA != al_entry->type)
+				continue;
+			/* We want an unnamed attribute. */
+			if (al_entry->name_length)
+				goto em_put_err_out;
+			/* Want the first entry, i.e. lowest_vcn == 0. */
+			if (al_entry->lowest_vcn)
+				goto em_put_err_out;
+			/* First entry has to be in the base mft record. */
+			if (MREF_LE(al_entry->mft_reference) != vi->i_ino) {
+				/* MFT references do not match, logic fails. */
+				ntfs_error(sb, "BUG: The first $DATA extent "
+						"of $MFT is not in the base "
+						"mft record. Please report "
+						"you saw this message to "
+						"linux-ntfs-dev@lists."
+						"sourceforge.net");
+				goto put_err_out;
+			} else {
+				/* Sequence numbers must match. */
+				if (MSEQNO_LE(al_entry->mft_reference) !=
+						ni->seq_no)
+					goto em_put_err_out;
+				/* Got it. All is ok. We can stop now. */
+				break;
+			}
+		}
+	}
+
+	ntfs_attr_reinit_search_ctx(ctx);
+
+	/* Now load all attribute extents. */
+	a = NULL;
+	next_vcn = last_vcn = highest_vcn = 0;
+	while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0,
+			ctx))) {
+		runlist_element *nrl;
+
+		/* Cache the current attribute. */
+		a = ctx->attr;
+		/* $MFT must be non-resident. */
+		if (!a->non_resident) {
+			ntfs_error(sb, "$MFT must be non-resident but a "
+					"resident extent was found. $MFT is "
+					"corrupt. Run chkdsk.");
+			goto put_err_out;
+		}
+		/* $MFT must be uncompressed and unencrypted. */
+		if (a->flags & ATTR_COMPRESSION_MASK ||
+				a->flags & ATTR_IS_ENCRYPTED ||
+				a->flags & ATTR_IS_SPARSE) {
+			ntfs_error(sb, "$MFT must be uncompressed, "
+					"non-sparse, and unencrypted but a "
+					"compressed/sparse/encrypted extent "
+					"was found. $MFT is corrupt. Run "
+					"chkdsk.");
+			goto put_err_out;
+		}
+		/*
+		 * Decompress the mapping pairs array of this extent and merge
+		 * the result into the existing runlist. No need for locking
+		 * as we have exclusive access to the inode at this time and we
+		 * are a mount in progress task, too.
+		 */
+		nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
+		if (IS_ERR(nrl)) {
+			ntfs_error(sb, "ntfs_mapping_pairs_decompress() "
+					"failed with error code %ld.  $MFT is "
+					"corrupt.", PTR_ERR(nrl));
+			goto put_err_out;
+		}
+		ni->runlist.rl = nrl;
+
+		/* Are we in the first extent? */
+		if (!next_vcn) {
+			if (a->data.non_resident.lowest_vcn) {
+				ntfs_error(sb, "First extent of $DATA "
+						"attribute has non zero "
+						"lowest_vcn. $MFT is corrupt. "
+						"You should run chkdsk.");
+				goto put_err_out;
+			}
+			/* Get the last vcn in the $DATA attribute. */
+			last_vcn = sle64_to_cpu(
+					a->data.non_resident.allocated_size)
+					>> vol->cluster_size_bits;
+			/* Fill in the inode size. */
+			vi->i_size = sle64_to_cpu(
+					a->data.non_resident.data_size);
+			ni->initialized_size = sle64_to_cpu(
+					a->data.non_resident.initialized_size);
+			ni->allocated_size = sle64_to_cpu(
+					a->data.non_resident.allocated_size);
+			/*
+			 * Verify the number of mft records does not exceed
+			 * 2^32 - 1.
+			 */
+			if ((vi->i_size >> vol->mft_record_size_bits) >=
+					(1ULL << 32)) {
+				ntfs_error(sb, "$MFT is too big! Aborting.");
+				goto put_err_out;
+			}
+			/*
+			 * We have got the first extent of the runlist for
+			 * $MFT which means it is now relatively safe to call
+			 * the normal ntfs_read_inode() function.
+			 * Complete reading the inode, this will actually
+			 * re-read the mft record for $MFT, this time entering
+			 * it into the page cache with which we complete the
+			 * kick start of the volume. It should be safe to do
+			 * this now as the first extent of $MFT/$DATA is
+			 * already known and we would hope that we don't need
+			 * further extents in order to find the other
+			 * attributes belonging to $MFT. Only time will tell if
+			 * this is really the case. If not we will have to play
+			 * magic at this point, possibly duplicating a lot of
+			 * ntfs_read_inode() at this point. We will need to
+			 * ensure we do enough of its work to be able to call
+			 * ntfs_read_inode() on extents of $MFT/$DATA. But lets
+			 * hope this never happens...
+			 */
+			ntfs_read_locked_inode(vi);
+			if (is_bad_inode(vi)) {
+				ntfs_error(sb, "ntfs_read_inode() of $MFT "
+						"failed. BUG or corrupt $MFT. "
+						"Run chkdsk and if no errors "
+						"are found, please report you "
+						"saw this message to "
+						"linux-ntfs-dev@lists."
+						"sourceforge.net");
+				ntfs_attr_put_search_ctx(ctx);
+				/* Revert to the safe super operations. */
+				ntfs_free(m);
+				return -1;
+			}
+			/*
+			 * Re-initialize some specifics about $MFT's inode as
+			 * ntfs_read_inode() will have set up the default ones.
+			 */
+			/* Set uid and gid to root. */
+			vi->i_uid = GLOBAL_ROOT_UID;
+			vi->i_gid = GLOBAL_ROOT_GID;
+			/* Regular file. No access for anyone. */
+			vi->i_mode = S_IFREG;
+			/* No VFS initiated operations allowed for $MFT. */
+			vi->i_op = &ntfs_empty_inode_ops;
+			vi->i_fop = &ntfs_empty_file_ops;
+		}
+
+		/* Get the lowest vcn for the next extent. */
+		highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
+		next_vcn = highest_vcn + 1;
+
+		/* Only one extent or error, which we catch below. */
+		if (next_vcn <= 0)
+			break;
+
+		/* Avoid endless loops due to corruption. */
+		if (next_vcn < sle64_to_cpu(
+				a->data.non_resident.lowest_vcn)) {
+			ntfs_error(sb, "$MFT has corrupt attribute list "
+					"attribute. Run chkdsk.");
+			goto put_err_out;
+		}
+	}
+	if (err != -ENOENT) {
+		ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. "
+				"$MFT is corrupt. Run chkdsk.");
+		goto put_err_out;
+	}
+	if (!a) {
+		ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is "
+				"corrupt. Run chkdsk.");
+		goto put_err_out;
+	}
+	if (highest_vcn && highest_vcn != last_vcn - 1) {
+		ntfs_error(sb, "Failed to load the complete runlist for "
+				"$MFT/$DATA. Driver bug or corrupt $MFT. "
+				"Run chkdsk.");
+		ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx",
+				(unsigned long long)highest_vcn,
+				(unsigned long long)last_vcn - 1);
+		goto put_err_out;
+	}
+	ntfs_attr_put_search_ctx(ctx);
+	ntfs_debug("Done.");
+	ntfs_free(m);
+
+	/*
+	 * Split the locking rules of the MFT inode from the
+	 * locking rules of other inodes:
+	 */
+	lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key);
+	lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key);
+
+	return 0;
+
+em_put_err_out:
+	ntfs_error(sb, "Couldn't find first extent of $DATA attribute in "
+			"attribute list. $MFT is corrupt. Run chkdsk.");
+put_err_out:
+	ntfs_attr_put_search_ctx(ctx);
+err_out:
+	ntfs_error(sb, "Failed. Marking inode as bad.");
+	make_bad_inode(vi);
+	ntfs_free(m);
+	return -1;
+}
+
+static void __ntfs_clear_inode(ntfs_inode *ni)
+{
+	/* Free all alocated memory. */
+	down_write(&ni->runlist.lock);
+	if (ni->runlist.rl) {
+		ntfs_free(ni->runlist.rl);
+		ni->runlist.rl = NULL;
+	}
+	up_write(&ni->runlist.lock);
+
+	if (ni->attr_list) {
+		ntfs_free(ni->attr_list);
+		ni->attr_list = NULL;
+	}
+
+	down_write(&ni->attr_list_rl.lock);
+	if (ni->attr_list_rl.rl) {
+		ntfs_free(ni->attr_list_rl.rl);
+		ni->attr_list_rl.rl = NULL;
+	}
+	up_write(&ni->attr_list_rl.lock);
+
+	if (ni->name_len && ni->name != I30) {
+		/* Catch bugs... */
+		BUG_ON(!ni->name);
+		kfree(ni->name);
+	}
+}
+
+void ntfs_clear_extent_inode(ntfs_inode *ni)
+{
+	ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
+
+	BUG_ON(NInoAttr(ni));
+	BUG_ON(ni->nr_extents != -1);
+
+#ifdef NTFS_RW
+	if (NInoDirty(ni)) {
+		if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino)))
+			ntfs_error(ni->vol->sb, "Clearing dirty extent inode!  "
+					"Losing data!  This is a BUG!!!");
+		// FIXME:  Do something!!!
+	}
+#endif /* NTFS_RW */
+
+	__ntfs_clear_inode(ni);
+
+	/* Bye, bye... */
+	ntfs_destroy_extent_inode(ni);
+}
+
+/**
+ * ntfs_evict_big_inode - clean up the ntfs specific part of an inode
+ * @vi:		vfs inode pending annihilation
+ *
+ * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode()
+ * is called, which deallocates all memory belonging to the NTFS specific part
+ * of the inode and returns.
+ *
+ * If the MFT record is dirty, we commit it before doing anything else.
+ */
+void ntfs_evict_big_inode(struct inode *vi)
+{
+	ntfs_inode *ni = NTFS_I(vi);
+
+	truncate_inode_pages_final(&vi->i_data);
+	clear_inode(vi);
+
+#ifdef NTFS_RW
+	if (NInoDirty(ni)) {
+		bool was_bad = (is_bad_inode(vi));
+
+		/* Committing the inode also commits all extent inodes. */
+		ntfs_commit_inode(vi);
+
+		if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) {
+			ntfs_error(vi->i_sb, "Failed to commit dirty inode "
+					"0x%lx.  Losing data!", vi->i_ino);
+			// FIXME:  Do something!!!
+		}
+	}
+#endif /* NTFS_RW */
+
+	/* No need to lock at this stage as no one else has a reference. */
+	if (ni->nr_extents > 0) {
+		int i;
+
+		for (i = 0; i < ni->nr_extents; i++)
+			ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]);
+		kfree(ni->ext.extent_ntfs_inos);
+	}
+
+	__ntfs_clear_inode(ni);
+
+	if (NInoAttr(ni)) {
+		/* Release the base inode if we are holding it. */
+		if (ni->nr_extents == -1) {
+			iput(VFS_I(ni->ext.base_ntfs_ino));
+			ni->nr_extents = 0;
+			ni->ext.base_ntfs_ino = NULL;
+		}
+	}
+	BUG_ON(ni->page);
+	if (!atomic_dec_and_test(&ni->count))
+		BUG();
+	return;
+}
+
+/**
+ * ntfs_show_options - show mount options in /proc/mounts
+ * @sf:		seq_file in which to write our mount options
+ * @root:	root of the mounted tree whose mount options to display
+ *
+ * Called by the VFS once for each mounted ntfs volume when someone reads
+ * /proc/mounts in order to display the NTFS specific mount options of each
+ * mount. The mount options of fs specified by @root are written to the seq file
+ * @sf and success is returned.
+ */
+int ntfs_show_options(struct seq_file *sf, struct dentry *root)
+{
+	ntfs_volume *vol = NTFS_SB(root->d_sb);
+	int i;
+
+	seq_printf(sf, ",uid=%i", from_kuid_munged(&init_user_ns, vol->uid));
+	seq_printf(sf, ",gid=%i", from_kgid_munged(&init_user_ns, vol->gid));
+	if (vol->fmask == vol->dmask)
+		seq_printf(sf, ",umask=0%o", vol->fmask);
+	else {
+		seq_printf(sf, ",fmask=0%o", vol->fmask);
+		seq_printf(sf, ",dmask=0%o", vol->dmask);
+	}
+	seq_printf(sf, ",nls=%s", vol->nls_map->charset);
+	if (NVolCaseSensitive(vol))
+		seq_printf(sf, ",case_sensitive");
+	if (NVolShowSystemFiles(vol))
+		seq_printf(sf, ",show_sys_files");
+	if (!NVolSparseEnabled(vol))
+		seq_printf(sf, ",disable_sparse");
+	for (i = 0; on_errors_arr[i].val; i++) {
+		if (on_errors_arr[i].val & vol->on_errors)
+			seq_printf(sf, ",errors=%s", on_errors_arr[i].str);
+	}
+	seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier);
+	return 0;
+}
+
+#ifdef NTFS_RW
+
+static const char *es = "  Leaving inconsistent metadata.  Unmount and run "
+		"chkdsk.";
+
+/**
+ * ntfs_truncate - called when the i_size of an ntfs inode is changed
+ * @vi:		inode for which the i_size was changed
+ *
+ * We only support i_size changes for normal files at present, i.e. not
+ * compressed and not encrypted.  This is enforced in ntfs_setattr(), see
+ * below.
+ *
+ * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and
+ * that the change is allowed.
+ *
+ * This implies for us that @vi is a file inode rather than a directory, index,
+ * or attribute inode as well as that @vi is a base inode.
+ *
+ * Returns 0 on success or -errno on error.
+ *
+ * Called with ->i_mutex held.
+ */
+int ntfs_truncate(struct inode *vi)
+{
+	s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size;
+	VCN highest_vcn;
+	unsigned long flags;
+	ntfs_inode *base_ni, *ni = NTFS_I(vi);
+	ntfs_volume *vol = ni->vol;
+	ntfs_attr_search_ctx *ctx;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	const char *te = "  Leaving file length out of sync with i_size.";
+	int err, mp_size, size_change, alloc_change;
+	u32 attr_len;
+
+	ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
+	BUG_ON(NInoAttr(ni));
+	BUG_ON(S_ISDIR(vi->i_mode));
+	BUG_ON(NInoMstProtected(ni));
+	BUG_ON(ni->nr_extents < 0);
+retry_truncate:
+	/*
+	 * Lock the runlist for writing and map the mft record to ensure it is
+	 * safe to mess with the attribute runlist and sizes.
+	 */
+	down_write(&ni->runlist.lock);
+	if (!NInoAttr(ni))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	m = map_mft_record(base_ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx "
+				"(error code %d).%s", vi->i_ino, err, te);
+		ctx = NULL;
+		m = NULL;
+		goto old_bad_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(base_ni, m);
+	if (unlikely(!ctx)) {
+		ntfs_error(vi->i_sb, "Failed to allocate a search context for "
+				"inode 0x%lx (not enough memory).%s",
+				vi->i_ino, te);
+		err = -ENOMEM;
+		goto old_bad_out;
+	}
+	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		if (err == -ENOENT) {
+			ntfs_error(vi->i_sb, "Open attribute is missing from "
+					"mft record.  Inode 0x%lx is corrupt.  "
+					"Run chkdsk.%s", vi->i_ino, te);
+			err = -EIO;
+		} else
+			ntfs_error(vi->i_sb, "Failed to lookup attribute in "
+					"inode 0x%lx (error code %d).%s",
+					vi->i_ino, err, te);
+		goto old_bad_out;
+	}
+	m = ctx->mrec;
+	a = ctx->attr;
+	/*
+	 * The i_size of the vfs inode is the new size for the attribute value.
+	 */
+	new_size = i_size_read(vi);
+	/* The current size of the attribute value is the old size. */
+	old_size = ntfs_attr_size(a);
+	/* Calculate the new allocated size. */
+	if (NInoNonResident(ni))
+		new_alloc_size = (new_size + vol->cluster_size - 1) &
+				~(s64)vol->cluster_size_mask;
+	else
+		new_alloc_size = (new_size + 7) & ~7;
+	/* The current allocated size is the old allocated size. */
+	read_lock_irqsave(&ni->size_lock, flags);
+	old_alloc_size = ni->allocated_size;
+	read_unlock_irqrestore(&ni->size_lock, flags);
+	/*
+	 * The change in the file size.  This will be 0 if no change, >0 if the
+	 * size is growing, and <0 if the size is shrinking.
+	 */
+	size_change = -1;
+	if (new_size - old_size >= 0) {
+		size_change = 1;
+		if (new_size == old_size)
+			size_change = 0;
+	}
+	/* As above for the allocated size. */
+	alloc_change = -1;
+	if (new_alloc_size - old_alloc_size >= 0) {
+		alloc_change = 1;
+		if (new_alloc_size == old_alloc_size)
+			alloc_change = 0;
+	}
+	/*
+	 * If neither the size nor the allocation are being changed there is
+	 * nothing to do.
+	 */
+	if (!size_change && !alloc_change)
+		goto unm_done;
+	/* If the size is changing, check if new size is allowed in $AttrDef. */
+	if (size_change) {
+		err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
+		if (unlikely(err)) {
+			if (err == -ERANGE) {
+				ntfs_error(vol->sb, "Truncate would cause the "
+						"inode 0x%lx to %simum size "
+						"for its attribute type "
+						"(0x%x).  Aborting truncate.",
+						vi->i_ino,
+						new_size > old_size ? "exceed "
+						"the max" : "go under the min",
+						le32_to_cpu(ni->type));
+				err = -EFBIG;
+			} else {
+				ntfs_error(vol->sb, "Inode 0x%lx has unknown "
+						"attribute type 0x%x.  "
+						"Aborting truncate.",
+						vi->i_ino,
+						le32_to_cpu(ni->type));
+				err = -EIO;
+			}
+			/* Reset the vfs inode size to the old size. */
+			i_size_write(vi, old_size);
+			goto err_out;
+		}
+	}
+	if (NInoCompressed(ni) || NInoEncrypted(ni)) {
+		ntfs_warning(vi->i_sb, "Changes in inode size are not "
+				"supported yet for %s files, ignoring.",
+				NInoCompressed(ni) ? "compressed" :
+				"encrypted");
+		err = -EOPNOTSUPP;
+		goto bad_out;
+	}
+	if (a->non_resident)
+		goto do_non_resident_truncate;
+	BUG_ON(NInoNonResident(ni));
+	/* Resize the attribute record to best fit the new attribute size. */
+	if (new_size < vol->mft_record_size &&
+			!ntfs_resident_attr_value_resize(m, a, new_size)) {
+		/* The resize succeeded! */
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		mark_mft_record_dirty(ctx->ntfs_ino);
+		write_lock_irqsave(&ni->size_lock, flags);
+		/* Update the sizes in the ntfs inode and all is done. */
+		ni->allocated_size = le32_to_cpu(a->length) -
+				le16_to_cpu(a->data.resident.value_offset);
+		/*
+		 * Note ntfs_resident_attr_value_resize() has already done any
+		 * necessary data clearing in the attribute record.  When the
+		 * file is being shrunk vmtruncate() will already have cleared
+		 * the top part of the last partial page, i.e. since this is
+		 * the resident case this is the page with index 0.  However,
+		 * when the file is being expanded, the page cache page data
+		 * between the old data_size, i.e. old_size, and the new_size
+		 * has not been zeroed.  Fortunately, we do not need to zero it
+		 * either since on one hand it will either already be zero due
+		 * to both readpage and writepage clearing partial page data
+		 * beyond i_size in which case there is nothing to do or in the
+		 * case of the file being mmap()ped at the same time, POSIX
+		 * specifies that the behaviour is unspecified thus we do not
+		 * have to do anything.  This means that in our implementation
+		 * in the rare case that the file is mmap()ped and a write
+		 * occurred into the mmap()ped region just beyond the file size
+		 * and writepage has not yet been called to write out the page
+		 * (which would clear the area beyond the file size) and we now
+		 * extend the file size to incorporate this dirty region
+		 * outside the file size, a write of the page would result in
+		 * this data being written to disk instead of being cleared.
+		 * Given both POSIX and the Linux mmap(2) man page specify that
+		 * this corner case is undefined, we choose to leave it like
+		 * that as this is much simpler for us as we cannot lock the
+		 * relevant page now since we are holding too many ntfs locks
+		 * which would result in a lock reversal deadlock.
+		 */
+		ni->initialized_size = new_size;
+		write_unlock_irqrestore(&ni->size_lock, flags);
+		goto unm_done;
+	}
+	/* If the above resize failed, this must be an attribute extension. */
+	BUG_ON(size_change < 0);
+	/*
+	 * We have to drop all the locks so we can call
+	 * ntfs_attr_make_non_resident().  This could be optimised by try-
+	 * locking the first page cache page and only if that fails dropping
+	 * the locks, locking the page, and redoing all the locking and
+	 * lookups.  While this would be a huge optimisation, it is not worth
+	 * it as this is definitely a slow code path as it only ever can happen
+	 * once for any given file.
+	 */
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+	up_write(&ni->runlist.lock);
+	/*
+	 * Not enough space in the mft record, try to make the attribute
+	 * non-resident and if successful restart the truncation process.
+	 */
+	err = ntfs_attr_make_non_resident(ni, old_size);
+	if (likely(!err))
+		goto retry_truncate;
+	/*
+	 * Could not make non-resident.  If this is due to this not being
+	 * permitted for this attribute type or there not being enough space,
+	 * try to make other attributes non-resident.  Otherwise fail.
+	 */
+	if (unlikely(err != -EPERM && err != -ENOSPC)) {
+		ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute "
+				"type 0x%x, because the conversion from "
+				"resident to non-resident attribute failed "
+				"with error code %i.", vi->i_ino,
+				(unsigned)le32_to_cpu(ni->type), err);
+		if (err != -ENOMEM)
+			err = -EIO;
+		goto conv_err_out;
+	}
+	/* TODO: Not implemented from here, abort. */
+	if (err == -ENOSPC)
+		ntfs_error(vol->sb, "Not enough space in the mft record/on "
+				"disk for the non-resident attribute value.  "
+				"This case is not implemented yet.");
+	else /* if (err == -EPERM) */
+		ntfs_error(vol->sb, "This attribute type may not be "
+				"non-resident.  This case is not implemented "
+				"yet.");
+	err = -EOPNOTSUPP;
+	goto conv_err_out;
+#if 0
+	// TODO: Attempt to make other attributes non-resident.
+	if (!err)
+		goto do_resident_extend;
+	/*
+	 * Both the attribute list attribute and the standard information
+	 * attribute must remain in the base inode.  Thus, if this is one of
+	 * these attributes, we have to try to move other attributes out into
+	 * extent mft records instead.
+	 */
+	if (ni->type == AT_ATTRIBUTE_LIST ||
+			ni->type == AT_STANDARD_INFORMATION) {
+		// TODO: Attempt to move other attributes into extent mft
+		// records.
+		err = -EOPNOTSUPP;
+		if (!err)
+			goto do_resident_extend;
+		goto err_out;
+	}
+	// TODO: Attempt to move this attribute to an extent mft record, but
+	// only if it is not already the only attribute in an mft record in
+	// which case there would be nothing to gain.
+	err = -EOPNOTSUPP;
+	if (!err)
+		goto do_resident_extend;
+	/* There is nothing we can do to make enough space. )-: */
+	goto err_out;
+#endif
+do_non_resident_truncate:
+	BUG_ON(!NInoNonResident(ni));
+	if (alloc_change < 0) {
+		highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
+		if (highest_vcn > 0 &&
+				old_alloc_size >> vol->cluster_size_bits >
+				highest_vcn + 1) {
+			/*
+			 * This attribute has multiple extents.  Not yet
+			 * supported.
+			 */
+			ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, "
+					"attribute type 0x%x, because the "
+					"attribute is highly fragmented (it "
+					"consists of multiple extents) and "
+					"this case is not implemented yet.",
+					vi->i_ino,
+					(unsigned)le32_to_cpu(ni->type));
+			err = -EOPNOTSUPP;
+			goto bad_out;
+		}
+	}
+	/*
+	 * If the size is shrinking, need to reduce the initialized_size and
+	 * the data_size before reducing the allocation.
+	 */
+	if (size_change < 0) {
+		/*
+		 * Make the valid size smaller (i_size is already up-to-date).
+		 */
+		write_lock_irqsave(&ni->size_lock, flags);
+		if (new_size < ni->initialized_size) {
+			ni->initialized_size = new_size;
+			a->data.non_resident.initialized_size =
+					cpu_to_sle64(new_size);
+		}
+		a->data.non_resident.data_size = cpu_to_sle64(new_size);
+		write_unlock_irqrestore(&ni->size_lock, flags);
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		mark_mft_record_dirty(ctx->ntfs_ino);
+		/* If the allocated size is not changing, we are done. */
+		if (!alloc_change)
+			goto unm_done;
+		/*
+		 * If the size is shrinking it makes no sense for the
+		 * allocation to be growing.
+		 */
+		BUG_ON(alloc_change > 0);
+	} else /* if (size_change >= 0) */ {
+		/*
+		 * The file size is growing or staying the same but the
+		 * allocation can be shrinking, growing or staying the same.
+		 */
+		if (alloc_change > 0) {
+			/*
+			 * We need to extend the allocation and possibly update
+			 * the data size.  If we are updating the data size,
+			 * since we are not touching the initialized_size we do
+			 * not need to worry about the actual data on disk.
+			 * And as far as the page cache is concerned, there
+			 * will be no pages beyond the old data size and any
+			 * partial region in the last page between the old and
+			 * new data size (or the end of the page if the new
+			 * data size is outside the page) does not need to be
+			 * modified as explained above for the resident
+			 * attribute truncate case.  To do this, we simply drop
+			 * the locks we hold and leave all the work to our
+			 * friendly helper ntfs_attr_extend_allocation().
+			 */
+			ntfs_attr_put_search_ctx(ctx);
+			unmap_mft_record(base_ni);
+			up_write(&ni->runlist.lock);
+			err = ntfs_attr_extend_allocation(ni, new_size,
+					size_change > 0 ? new_size : -1, -1);
+			/*
+			 * ntfs_attr_extend_allocation() will have done error
+			 * output already.
+			 */
+			goto done;
+		}
+		if (!alloc_change)
+			goto alloc_done;
+	}
+	/* alloc_change < 0 */
+	/* Free the clusters. */
+	nr_freed = ntfs_cluster_free(ni, new_alloc_size >>
+			vol->cluster_size_bits, -1, ctx);
+	m = ctx->mrec;
+	a = ctx->attr;
+	if (unlikely(nr_freed < 0)) {
+		ntfs_error(vol->sb, "Failed to release cluster(s) (error code "
+				"%lli).  Unmount and run chkdsk to recover "
+				"the lost cluster(s).", (long long)nr_freed);
+		NVolSetErrors(vol);
+		nr_freed = 0;
+	}
+	/* Truncate the runlist. */
+	err = ntfs_rl_truncate_nolock(vol, &ni->runlist,
+			new_alloc_size >> vol->cluster_size_bits);
+	/*
+	 * If the runlist truncation failed and/or the search context is no
+	 * longer valid, we cannot resize the attribute record or build the
+	 * mapping pairs array thus we mark the inode bad so that no access to
+	 * the freed clusters can happen.
+	 */
+	if (unlikely(err || IS_ERR(m))) {
+		ntfs_error(vol->sb, "Failed to %s (error code %li).%s",
+				IS_ERR(m) ?
+				"restore attribute search context" :
+				"truncate attribute runlist",
+				IS_ERR(m) ? PTR_ERR(m) : err, es);
+		err = -EIO;
+		goto bad_out;
+	}
+	/* Get the size for the shrunk mapping pairs array for the runlist. */
+	mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1);
+	if (unlikely(mp_size <= 0)) {
+		ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
+				"attribute type 0x%x, because determining the "
+				"size for the mapping pairs failed with error "
+				"code %i.%s", vi->i_ino,
+				(unsigned)le32_to_cpu(ni->type), mp_size, es);
+		err = -EIO;
+		goto bad_out;
+	}
+	/*
+	 * Shrink the attribute record for the new mapping pairs array.  Note,
+	 * this cannot fail since we are making the attribute smaller thus by
+	 * definition there is enough space to do so.
+	 */
+	attr_len = le32_to_cpu(a->length);
+	err = ntfs_attr_record_resize(m, a, mp_size +
+			le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
+	BUG_ON(err);
+	/*
+	 * Generate the mapping pairs array directly into the attribute record.
+	 */
+	err = ntfs_mapping_pairs_build(vol, (u8*)a +
+			le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
+			mp_size, ni->runlist.rl, 0, -1, NULL);
+	if (unlikely(err)) {
+		ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, "
+				"attribute type 0x%x, because building the "
+				"mapping pairs failed with error code %i.%s",
+				vi->i_ino, (unsigned)le32_to_cpu(ni->type),
+				err, es);
+		err = -EIO;
+		goto bad_out;
+	}
+	/* Update the allocated/compressed size as well as the highest vcn. */
+	a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
+			vol->cluster_size_bits) - 1);
+	write_lock_irqsave(&ni->size_lock, flags);
+	ni->allocated_size = new_alloc_size;
+	a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
+	if (NInoSparse(ni) || NInoCompressed(ni)) {
+		if (nr_freed) {
+			ni->itype.compressed.size -= nr_freed <<
+					vol->cluster_size_bits;
+			BUG_ON(ni->itype.compressed.size < 0);
+			a->data.non_resident.compressed_size = cpu_to_sle64(
+					ni->itype.compressed.size);
+			vi->i_blocks = ni->itype.compressed.size >> 9;
+		}
+	} else
+		vi->i_blocks = new_alloc_size >> 9;
+	write_unlock_irqrestore(&ni->size_lock, flags);
+	/*
+	 * We have shrunk the allocation.  If this is a shrinking truncate we
+	 * have already dealt with the initialized_size and the data_size above
+	 * and we are done.  If the truncate is only changing the allocation
+	 * and not the data_size, we are also done.  If this is an extending
+	 * truncate, need to extend the data_size now which is ensured by the
+	 * fact that @size_change is positive.
+	 */
+alloc_done:
+	/*
+	 * If the size is growing, need to update it now.  If it is shrinking,
+	 * we have already updated it above (before the allocation change).
+	 */
+	if (size_change > 0)
+		a->data.non_resident.data_size = cpu_to_sle64(new_size);
+	/* Ensure the modified mft record is written out. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+unm_done:
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(base_ni);
+	up_write(&ni->runlist.lock);
+done:
+	/* Update the mtime and ctime on the base inode. */
+	/* normally ->truncate shouldn't update ctime or mtime,
+	 * but ntfs did before so it got a copy & paste version
+	 * of file_update_time.  one day someone should fix this
+	 * for real.
+	 */
+	if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) {
+		struct timespec64 now = current_time(VFS_I(base_ni));
+		int sync_it = 0;
+
+		if (!timespec64_equal(&VFS_I(base_ni)->i_mtime, &now) ||
+		    !timespec64_equal(&VFS_I(base_ni)->i_ctime, &now))
+			sync_it = 1;
+		VFS_I(base_ni)->i_mtime = now;
+		VFS_I(base_ni)->i_ctime = now;
+
+		if (sync_it)
+			mark_inode_dirty_sync(VFS_I(base_ni));
+	}
+
+	if (likely(!err)) {
+		NInoClearTruncateFailed(ni);
+		ntfs_debug("Done.");
+	}
+	return err;
+old_bad_out:
+	old_size = -1;
+bad_out:
+	if (err != -ENOMEM && err != -EOPNOTSUPP)
+		NVolSetErrors(vol);
+	if (err != -EOPNOTSUPP)
+		NInoSetTruncateFailed(ni);
+	else if (old_size >= 0)
+		i_size_write(vi, old_size);
+err_out:
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(base_ni);
+	up_write(&ni->runlist.lock);
+out:
+	ntfs_debug("Failed.  Returning error code %i.", err);
+	return err;
+conv_err_out:
+	if (err != -ENOMEM && err != -EOPNOTSUPP)
+		NVolSetErrors(vol);
+	if (err != -EOPNOTSUPP)
+		NInoSetTruncateFailed(ni);
+	else
+		i_size_write(vi, old_size);
+	goto out;
+}
+
+/**
+ * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value
+ * @vi:		inode for which the i_size was changed
+ *
+ * Wrapper for ntfs_truncate() that has no return value.
+ *
+ * See ntfs_truncate() description above for details.
+ */
+#ifdef NTFS_RW
+void ntfs_truncate_vfs(struct inode *vi) {
+	ntfs_truncate(vi);
+}
+#endif
+
+/**
+ * ntfs_setattr - called from notify_change() when an attribute is being changed
+ * @dentry:	dentry whose attributes to change
+ * @attr:	structure describing the attributes and the changes
+ *
+ * We have to trap VFS attempts to truncate the file described by @dentry as
+ * soon as possible, because we do not implement changes in i_size yet.  So we
+ * abort all i_size changes here.
+ *
+ * We also abort all changes of user, group, and mode as we do not implement
+ * the NTFS ACLs yet.
+ *
+ * Called with ->i_mutex held.
+ */
+int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
+{
+	struct inode *vi = d_inode(dentry);
+	int err;
+	unsigned int ia_valid = attr->ia_valid;
+
+	err = setattr_prepare(dentry, attr);
+	if (err)
+		goto out;
+	/* We do not support NTFS ACLs yet. */
+	if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) {
+		ntfs_warning(vi->i_sb, "Changes in user/group/mode are not "
+				"supported yet, ignoring.");
+		err = -EOPNOTSUPP;
+		goto out;
+	}
+	if (ia_valid & ATTR_SIZE) {
+		if (attr->ia_size != i_size_read(vi)) {
+			ntfs_inode *ni = NTFS_I(vi);
+			/*
+			 * FIXME: For now we do not support resizing of
+			 * compressed or encrypted files yet.
+			 */
+			if (NInoCompressed(ni) || NInoEncrypted(ni)) {
+				ntfs_warning(vi->i_sb, "Changes in inode size "
+						"are not supported yet for "
+						"%s files, ignoring.",
+						NInoCompressed(ni) ?
+						"compressed" : "encrypted");
+				err = -EOPNOTSUPP;
+			} else {
+				truncate_setsize(vi, attr->ia_size);
+				ntfs_truncate_vfs(vi);
+			}
+			if (err || ia_valid == ATTR_SIZE)
+				goto out;
+		} else {
+			/*
+			 * We skipped the truncate but must still update
+			 * timestamps.
+			 */
+			ia_valid |= ATTR_MTIME | ATTR_CTIME;
+		}
+	}
+	if (ia_valid & ATTR_ATIME)
+		vi->i_atime = attr->ia_atime;
+	if (ia_valid & ATTR_MTIME)
+		vi->i_mtime = attr->ia_mtime;
+	if (ia_valid & ATTR_CTIME)
+		vi->i_ctime = attr->ia_ctime;
+	mark_inode_dirty(vi);
+out:
+	return err;
+}
+
+/**
+ * ntfs_write_inode - write out a dirty inode
+ * @vi:		inode to write out
+ * @sync:	if true, write out synchronously
+ *
+ * Write out a dirty inode to disk including any extent inodes if present.
+ *
+ * If @sync is true, commit the inode to disk and wait for io completion.  This
+ * is done using write_mft_record().
+ *
+ * If @sync is false, just schedule the write to happen but do not wait for i/o
+ * completion.  In 2.6 kernels, scheduling usually happens just by virtue of
+ * marking the page (and in this case mft record) dirty but we do not implement
+ * this yet as write_mft_record() largely ignores the @sync parameter and
+ * always performs synchronous writes.
+ *
+ * Return 0 on success and -errno on error.
+ */
+int __ntfs_write_inode(struct inode *vi, int sync)
+{
+	sle64 nt;
+	ntfs_inode *ni = NTFS_I(vi);
+	ntfs_attr_search_ctx *ctx;
+	MFT_RECORD *m;
+	STANDARD_INFORMATION *si;
+	int err = 0;
+	bool modified = false;
+
+	ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "",
+			vi->i_ino);
+	/*
+	 * Dirty attribute inodes are written via their real inodes so just
+	 * clean them here.  Access time updates are taken care off when the
+	 * real inode is written.
+	 */
+	if (NInoAttr(ni)) {
+		NInoClearDirty(ni);
+		ntfs_debug("Done.");
+		return 0;
+	}
+	/* Map, pin, and lock the mft record belonging to the inode. */
+	m = map_mft_record(ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		goto err_out;
+	}
+	/* Update the access times in the standard information attribute. */
+	ctx = ntfs_attr_get_search_ctx(ni, m);
+	if (unlikely(!ctx)) {
+		err = -ENOMEM;
+		goto unm_err_out;
+	}
+	err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		ntfs_attr_put_search_ctx(ctx);
+		goto unm_err_out;
+	}
+	si = (STANDARD_INFORMATION*)((u8*)ctx->attr +
+			le16_to_cpu(ctx->attr->data.resident.value_offset));
+	/* Update the access times if they have changed. */
+	nt = utc2ntfs(vi->i_mtime);
+	if (si->last_data_change_time != nt) {
+		ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, "
+				"new = 0x%llx", vi->i_ino, (long long)
+				sle64_to_cpu(si->last_data_change_time),
+				(long long)sle64_to_cpu(nt));
+		si->last_data_change_time = nt;
+		modified = true;
+	}
+	nt = utc2ntfs(vi->i_ctime);
+	if (si->last_mft_change_time != nt) {
+		ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, "
+				"new = 0x%llx", vi->i_ino, (long long)
+				sle64_to_cpu(si->last_mft_change_time),
+				(long long)sle64_to_cpu(nt));
+		si->last_mft_change_time = nt;
+		modified = true;
+	}
+	nt = utc2ntfs(vi->i_atime);
+	if (si->last_access_time != nt) {
+		ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, "
+				"new = 0x%llx", vi->i_ino,
+				(long long)sle64_to_cpu(si->last_access_time),
+				(long long)sle64_to_cpu(nt));
+		si->last_access_time = nt;
+		modified = true;
+	}
+	/*
+	 * If we just modified the standard information attribute we need to
+	 * mark the mft record it is in dirty.  We do this manually so that
+	 * mark_inode_dirty() is not called which would redirty the inode and
+	 * hence result in an infinite loop of trying to write the inode.
+	 * There is no need to mark the base inode nor the base mft record
+	 * dirty, since we are going to write this mft record below in any case
+	 * and the base mft record may actually not have been modified so it
+	 * might not need to be written out.
+	 * NOTE: It is not a problem when the inode for $MFT itself is being
+	 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES
+	 * on the $MFT inode and hence ntfs_write_inode() will not be
+	 * re-invoked because of it which in turn is ok since the dirtied mft
+	 * record will be cleaned and written out to disk below, i.e. before
+	 * this function returns.
+	 */
+	if (modified) {
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		if (!NInoTestSetDirty(ctx->ntfs_ino))
+			mark_ntfs_record_dirty(ctx->ntfs_ino->page,
+					ctx->ntfs_ino->page_ofs);
+	}
+	ntfs_attr_put_search_ctx(ctx);
+	/* Now the access times are updated, write the base mft record. */
+	if (NInoDirty(ni))
+		err = write_mft_record(ni, m, sync);
+	/* Write all attached extent mft records. */
+	mutex_lock(&ni->extent_lock);
+	if (ni->nr_extents > 0) {
+		ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos;
+		int i;
+
+		ntfs_debug("Writing %i extent inodes.", ni->nr_extents);
+		for (i = 0; i < ni->nr_extents; i++) {
+			ntfs_inode *tni = extent_nis[i];
+
+			if (NInoDirty(tni)) {
+				MFT_RECORD *tm = map_mft_record(tni);
+				int ret;
+
+				if (IS_ERR(tm)) {
+					if (!err || err == -ENOMEM)
+						err = PTR_ERR(tm);
+					continue;
+				}
+				ret = write_mft_record(tni, tm, sync);
+				unmap_mft_record(tni);
+				if (unlikely(ret)) {
+					if (!err || err == -ENOMEM)
+						err = ret;
+				}
+			}
+		}
+	}
+	mutex_unlock(&ni->extent_lock);
+	unmap_mft_record(ni);
+	if (unlikely(err))
+		goto err_out;
+	ntfs_debug("Done.");
+	return 0;
+unm_err_out:
+	unmap_mft_record(ni);
+err_out:
+	if (err == -ENOMEM) {
+		ntfs_warning(vi->i_sb, "Not enough memory to write inode.  "
+				"Marking the inode dirty again, so the VFS "
+				"retries later.");
+		mark_inode_dirty(vi);
+	} else {
+		ntfs_error(vi->i_sb, "Failed (error %i):  Run chkdsk.", -err);
+		NVolSetErrors(ni->vol);
+	}
+	return err;
+}
+
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/inode.h b/fs/ntfs/inode.h
new file mode 100644
index 000000000..363e4e820
--- /dev/null
+++ b/fs/ntfs/inode.h
@@ -0,0 +1,309 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * inode.h - Defines for inode structures NTFS Linux kernel driver. Part of
+ *	     the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2007 Anton Altaparmakov
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#ifndef _LINUX_NTFS_INODE_H
+#define _LINUX_NTFS_INODE_H
+
+#include <linux/atomic.h>
+
+#include <linux/fs.h>
+#include <linux/list.h>
+#include <linux/mm.h>
+#include <linux/mutex.h>
+#include <linux/seq_file.h>
+
+#include "layout.h"
+#include "volume.h"
+#include "types.h"
+#include "runlist.h"
+#include "debug.h"
+
+typedef struct _ntfs_inode ntfs_inode;
+
+/*
+ * The NTFS in-memory inode structure. It is just used as an extension to the
+ * fields already provided in the VFS inode.
+ */
+struct _ntfs_inode {
+	rwlock_t size_lock;	/* Lock serializing access to inode sizes. */
+	s64 initialized_size;	/* Copy from the attribute record. */
+	s64 allocated_size;	/* Copy from the attribute record. */
+	unsigned long state;	/* NTFS specific flags describing this inode.
+				   See ntfs_inode_state_bits below. */
+	unsigned long mft_no;	/* Number of the mft record / inode. */
+	u16 seq_no;		/* Sequence number of the mft record. */
+	atomic_t count;		/* Inode reference count for book keeping. */
+	ntfs_volume *vol;	/* Pointer to the ntfs volume of this inode. */
+	/*
+	 * If NInoAttr() is true, the below fields describe the attribute which
+	 * this fake inode belongs to. The actual inode of this attribute is
+	 * pointed to by base_ntfs_ino and nr_extents is always set to -1 (see
+	 * below). For real inodes, we also set the type (AT_DATA for files and
+	 * AT_INDEX_ALLOCATION for directories), with the name = NULL and
+	 * name_len = 0 for files and name = I30 (global constant) and
+	 * name_len = 4 for directories.
+	 */
+	ATTR_TYPE type;	/* Attribute type of this fake inode. */
+	ntfschar *name;		/* Attribute name of this fake inode. */
+	u32 name_len;		/* Attribute name length of this fake inode. */
+	runlist runlist;	/* If state has the NI_NonResident bit set,
+				   the runlist of the unnamed data attribute
+				   (if a file) or of the index allocation
+				   attribute (directory) or of the attribute
+				   described by the fake inode (if NInoAttr()).
+				   If runlist.rl is NULL, the runlist has not
+				   been read in yet or has been unmapped. If
+				   NI_NonResident is clear, the attribute is
+				   resident (file and fake inode) or there is
+				   no $I30 index allocation attribute
+				   (small directory). In the latter case
+				   runlist.rl is always NULL.*/
+	/*
+	 * The following fields are only valid for real inodes and extent
+	 * inodes.
+	 */
+	struct mutex mrec_lock;	/* Lock for serializing access to the
+				   mft record belonging to this inode. */
+	struct page *page;	/* The page containing the mft record of the
+				   inode. This should only be touched by the
+				   (un)map_mft_record*() functions. */
+	int page_ofs;		/* Offset into the page at which the mft record
+				   begins. This should only be touched by the
+				   (un)map_mft_record*() functions. */
+	/*
+	 * Attribute list support (only for use by the attribute lookup
+	 * functions). Setup during read_inode for all inodes with attribute
+	 * lists. Only valid if NI_AttrList is set in state, and attr_list_rl is
+	 * further only valid if NI_AttrListNonResident is set.
+	 */
+	u32 attr_list_size;	/* Length of attribute list value in bytes. */
+	u8 *attr_list;		/* Attribute list value itself. */
+	runlist attr_list_rl;	/* Run list for the attribute list value. */
+	union {
+		struct { /* It is a directory, $MFT, or an index inode. */
+			u32 block_size;		/* Size of an index block. */
+			u32 vcn_size;		/* Size of a vcn in this
+						   index. */
+			COLLATION_RULE collation_rule; /* The collation rule
+						   for the index. */
+			u8 block_size_bits; 	/* Log2 of the above. */
+			u8 vcn_size_bits;	/* Log2 of the above. */
+		} index;
+		struct { /* It is a compressed/sparse file/attribute inode. */
+			s64 size;		/* Copy of compressed_size from
+						   $DATA. */
+			u32 block_size;		/* Size of a compression block
+						   (cb). */
+			u8 block_size_bits;	/* Log2 of the size of a cb. */
+			u8 block_clusters;	/* Number of clusters per cb. */
+		} compressed;
+	} itype;
+	struct mutex extent_lock;	/* Lock for accessing/modifying the
+					   below . */
+	s32 nr_extents;	/* For a base mft record, the number of attached extent
+			   inodes (0 if none), for extent records and for fake
+			   inodes describing an attribute this is -1. */
+	union {		/* This union is only used if nr_extents != 0. */
+		ntfs_inode **extent_ntfs_inos;	/* For nr_extents > 0, array of
+						   the ntfs inodes of the extent
+						   mft records belonging to
+						   this base inode which have
+						   been loaded. */
+		ntfs_inode *base_ntfs_ino;	/* For nr_extents == -1, the
+						   ntfs inode of the base mft
+						   record. For fake inodes, the
+						   real (base) inode to which
+						   the attribute belongs. */
+	} ext;
+};
+
+/*
+ * Defined bits for the state field in the ntfs_inode structure.
+ * (f) = files only, (d) = directories only, (a) = attributes/fake inodes only
+ */
+typedef enum {
+	NI_Dirty,		/* 1: Mft record needs to be written to disk. */
+	NI_AttrList,		/* 1: Mft record contains an attribute list. */
+	NI_AttrListNonResident,	/* 1: Attribute list is non-resident. Implies
+				      NI_AttrList is set. */
+
+	NI_Attr,		/* 1: Fake inode for attribute i/o.
+				   0: Real inode or extent inode. */
+
+	NI_MstProtected,	/* 1: Attribute is protected by MST fixups.
+				   0: Attribute is not protected by fixups. */
+	NI_NonResident,		/* 1: Unnamed data attr is non-resident (f).
+				   1: Attribute is non-resident (a). */
+	NI_IndexAllocPresent = NI_NonResident,	/* 1: $I30 index alloc attr is
+						   present (d). */
+	NI_Compressed,		/* 1: Unnamed data attr is compressed (f).
+				   1: Create compressed files by default (d).
+				   1: Attribute is compressed (a). */
+	NI_Encrypted,		/* 1: Unnamed data attr is encrypted (f).
+				   1: Create encrypted files by default (d).
+				   1: Attribute is encrypted (a). */
+	NI_Sparse,		/* 1: Unnamed data attr is sparse (f).
+				   1: Create sparse files by default (d).
+				   1: Attribute is sparse (a). */
+	NI_SparseDisabled,	/* 1: May not create sparse regions. */
+	NI_TruncateFailed,	/* 1: Last ntfs_truncate() call failed. */
+} ntfs_inode_state_bits;
+
+/*
+ * NOTE: We should be adding dirty mft records to a list somewhere and they
+ * should be independent of the (ntfs/vfs) inode structure so that an inode can
+ * be removed but the record can be left dirty for syncing later.
+ */
+
+/*
+ * Macro tricks to expand the NInoFoo(), NInoSetFoo(), and NInoClearFoo()
+ * functions.
+ */
+#define NINO_FNS(flag)					\
+static inline int NIno##flag(ntfs_inode *ni)		\
+{							\
+	return test_bit(NI_##flag, &(ni)->state);	\
+}							\
+static inline void NInoSet##flag(ntfs_inode *ni)	\
+{							\
+	set_bit(NI_##flag, &(ni)->state);		\
+}							\
+static inline void NInoClear##flag(ntfs_inode *ni)	\
+{							\
+	clear_bit(NI_##flag, &(ni)->state);		\
+}
+
+/*
+ * As above for NInoTestSetFoo() and NInoTestClearFoo().
+ */
+#define TAS_NINO_FNS(flag)					\
+static inline int NInoTestSet##flag(ntfs_inode *ni)		\
+{								\
+	return test_and_set_bit(NI_##flag, &(ni)->state);	\
+}								\
+static inline int NInoTestClear##flag(ntfs_inode *ni)		\
+{								\
+	return test_and_clear_bit(NI_##flag, &(ni)->state);	\
+}
+
+/* Emit the ntfs inode bitops functions. */
+NINO_FNS(Dirty)
+TAS_NINO_FNS(Dirty)
+NINO_FNS(AttrList)
+NINO_FNS(AttrListNonResident)
+NINO_FNS(Attr)
+NINO_FNS(MstProtected)
+NINO_FNS(NonResident)
+NINO_FNS(IndexAllocPresent)
+NINO_FNS(Compressed)
+NINO_FNS(Encrypted)
+NINO_FNS(Sparse)
+NINO_FNS(SparseDisabled)
+NINO_FNS(TruncateFailed)
+
+/*
+ * The full structure containing a ntfs_inode and a vfs struct inode. Used for
+ * all real and fake inodes but not for extent inodes which lack the vfs struct
+ * inode.
+ */
+typedef struct {
+	ntfs_inode ntfs_inode;
+	struct inode vfs_inode;		/* The vfs inode structure. */
+} big_ntfs_inode;
+
+/**
+ * NTFS_I - return the ntfs inode given a vfs inode
+ * @inode:	VFS inode
+ *
+ * NTFS_I() returns the ntfs inode associated with the VFS @inode.
+ */
+static inline ntfs_inode *NTFS_I(struct inode *inode)
+{
+	return (ntfs_inode *)container_of(inode, big_ntfs_inode, vfs_inode);
+}
+
+static inline struct inode *VFS_I(ntfs_inode *ni)
+{
+	return &((big_ntfs_inode *)ni)->vfs_inode;
+}
+
+/**
+ * ntfs_attr - ntfs in memory attribute structure
+ * @mft_no:	mft record number of the base mft record of this attribute
+ * @name:	Unicode name of the attribute (NULL if unnamed)
+ * @name_len:	length of @name in Unicode characters (0 if unnamed)
+ * @type:	attribute type (see layout.h)
+ *
+ * This structure exists only to provide a small structure for the
+ * ntfs_{attr_}iget()/ntfs_test_inode()/ntfs_init_locked_inode() mechanism.
+ *
+ * NOTE: Elements are ordered by size to make the structure as compact as
+ * possible on all architectures.
+ */
+typedef struct {
+	unsigned long mft_no;
+	ntfschar *name;
+	u32 name_len;
+	ATTR_TYPE type;
+} ntfs_attr;
+
+extern int ntfs_test_inode(struct inode *vi, void *data);
+
+extern struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no);
+extern struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type,
+		ntfschar *name, u32 name_len);
+extern struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name,
+		u32 name_len);
+
+extern struct inode *ntfs_alloc_big_inode(struct super_block *sb);
+extern void ntfs_free_big_inode(struct inode *inode);
+extern void ntfs_evict_big_inode(struct inode *vi);
+
+extern void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni);
+
+static inline void ntfs_init_big_inode(struct inode *vi)
+{
+	ntfs_inode *ni = NTFS_I(vi);
+
+	ntfs_debug("Entering.");
+	__ntfs_init_inode(vi->i_sb, ni);
+	ni->mft_no = vi->i_ino;
+}
+
+extern ntfs_inode *ntfs_new_extent_inode(struct super_block *sb,
+		unsigned long mft_no);
+extern void ntfs_clear_extent_inode(ntfs_inode *ni);
+
+extern int ntfs_read_inode_mount(struct inode *vi);
+
+extern int ntfs_show_options(struct seq_file *sf, struct dentry *root);
+
+#ifdef NTFS_RW
+
+extern int ntfs_truncate(struct inode *vi);
+extern void ntfs_truncate_vfs(struct inode *vi);
+
+extern int ntfs_setattr(struct dentry *dentry, struct iattr *attr);
+
+extern int __ntfs_write_inode(struct inode *vi, int sync);
+
+static inline void ntfs_commit_inode(struct inode *vi)
+{
+	if (!is_bad_inode(vi))
+		__ntfs_write_inode(vi, 1);
+	return;
+}
+
+#else
+
+static inline void ntfs_truncate_vfs(struct inode *vi) {}
+
+#endif /* NTFS_RW */
+
+#endif /* _LINUX_NTFS_INODE_H */
diff --git a/fs/ntfs/layout.h b/fs/ntfs/layout.h
new file mode 100644
index 000000000..85422761f
--- /dev/null
+++ b/fs/ntfs/layout.h
@@ -0,0 +1,2421 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * layout.h - All NTFS associated on-disk structures. Part of the Linux-NTFS
+ *	      project.
+ *
+ * Copyright (c) 2001-2005 Anton Altaparmakov
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#ifndef _LINUX_NTFS_LAYOUT_H
+#define _LINUX_NTFS_LAYOUT_H
+
+#include <linux/types.h>
+#include <linux/bitops.h>
+#include <linux/list.h>
+#include <asm/byteorder.h>
+
+#include "types.h"
+
+/* The NTFS oem_id "NTFS    " */
+#define magicNTFS	cpu_to_le64(0x202020205346544eULL)
+
+/*
+ * Location of bootsector on partition:
+ *	The standard NTFS_BOOT_SECTOR is on sector 0 of the partition.
+ *	On NT4 and above there is one backup copy of the boot sector to
+ *	be found on the last sector of the partition (not normally accessible
+ *	from within Windows as the bootsector contained number of sectors
+ *	value is one less than the actual value!).
+ *	On versions of NT 3.51 and earlier, the backup copy was located at
+ *	number of sectors/2 (integer divide), i.e. in the middle of the volume.
+ */
+
+/*
+ * BIOS parameter block (bpb) structure.
+ */
+typedef struct {
+	le16 bytes_per_sector;		/* Size of a sector in bytes. */
+	u8  sectors_per_cluster;	/* Size of a cluster in sectors. */
+	le16 reserved_sectors;		/* zero */
+	u8  fats;			/* zero */
+	le16 root_entries;		/* zero */
+	le16 sectors;			/* zero */
+	u8  media_type;			/* 0xf8 = hard disk */
+	le16 sectors_per_fat;		/* zero */
+	le16 sectors_per_track;		/* irrelevant */
+	le16 heads;			/* irrelevant */
+	le32 hidden_sectors;		/* zero */
+	le32 large_sectors;		/* zero */
+} __attribute__ ((__packed__)) BIOS_PARAMETER_BLOCK;
+
+/*
+ * NTFS boot sector structure.
+ */
+typedef struct {
+	u8  jump[3];			/* Irrelevant (jump to boot up code).*/
+	le64 oem_id;			/* Magic "NTFS    ". */
+	BIOS_PARAMETER_BLOCK bpb;	/* See BIOS_PARAMETER_BLOCK. */
+	u8  unused[4];			/* zero, NTFS diskedit.exe states that
+					   this is actually:
+						__u8 physical_drive;	// 0x80
+						__u8 current_head;	// zero
+						__u8 extended_boot_signature;
+									// 0x80
+						__u8 unused;		// zero
+					 */
+/*0x28*/sle64 number_of_sectors;	/* Number of sectors in volume. Gives
+					   maximum volume size of 2^63 sectors.
+					   Assuming standard sector size of 512
+					   bytes, the maximum byte size is
+					   approx. 4.7x10^21 bytes. (-; */
+	sle64 mft_lcn;			/* Cluster location of mft data. */
+	sle64 mftmirr_lcn;		/* Cluster location of copy of mft. */
+	s8  clusters_per_mft_record;	/* Mft record size in clusters. */
+	u8  reserved0[3];		/* zero */
+	s8  clusters_per_index_record;	/* Index block size in clusters. */
+	u8  reserved1[3];		/* zero */
+	le64 volume_serial_number;	/* Irrelevant (serial number). */
+	le32 checksum;			/* Boot sector checksum. */
+/*0x54*/u8  bootstrap[426];		/* Irrelevant (boot up code). */
+	le16 end_of_sector_marker;	/* End of bootsector magic. Always is
+					   0xaa55 in little endian. */
+/* sizeof() = 512 (0x200) bytes */
+} __attribute__ ((__packed__)) NTFS_BOOT_SECTOR;
+
+/*
+ * Magic identifiers present at the beginning of all ntfs record containing
+ * records (like mft records for example).
+ */
+enum {
+	/* Found in $MFT/$DATA. */
+	magic_FILE = cpu_to_le32(0x454c4946), /* Mft entry. */
+	magic_INDX = cpu_to_le32(0x58444e49), /* Index buffer. */
+	magic_HOLE = cpu_to_le32(0x454c4f48), /* ? (NTFS 3.0+?) */
+
+	/* Found in $LogFile/$DATA. */
+	magic_RSTR = cpu_to_le32(0x52545352), /* Restart page. */
+	magic_RCRD = cpu_to_le32(0x44524352), /* Log record page. */
+
+	/* Found in $LogFile/$DATA.  (May be found in $MFT/$DATA, also?) */
+	magic_CHKD = cpu_to_le32(0x444b4843), /* Modified by chkdsk. */
+
+	/* Found in all ntfs record containing records. */
+	magic_BAAD = cpu_to_le32(0x44414142), /* Failed multi sector
+						       transfer was detected. */
+	/*
+	 * Found in $LogFile/$DATA when a page is full of 0xff bytes and is
+	 * thus not initialized.  Page must be initialized before using it.
+	 */
+	magic_empty = cpu_to_le32(0xffffffff) /* Record is empty. */
+};
+
+typedef le32 NTFS_RECORD_TYPE;
+
+/*
+ * Generic magic comparison macros. Finally found a use for the ## preprocessor
+ * operator! (-8
+ */
+
+static inline bool __ntfs_is_magic(le32 x, NTFS_RECORD_TYPE r)
+{
+	return (x == r);
+}
+#define ntfs_is_magic(x, m)	__ntfs_is_magic(x, magic_##m)
+
+static inline bool __ntfs_is_magicp(le32 *p, NTFS_RECORD_TYPE r)
+{
+	return (*p == r);
+}
+#define ntfs_is_magicp(p, m)	__ntfs_is_magicp(p, magic_##m)
+
+/*
+ * Specialised magic comparison macros for the NTFS_RECORD_TYPEs defined above.
+ */
+#define ntfs_is_file_record(x)		( ntfs_is_magic (x, FILE) )
+#define ntfs_is_file_recordp(p)		( ntfs_is_magicp(p, FILE) )
+#define ntfs_is_mft_record(x)		( ntfs_is_file_record (x) )
+#define ntfs_is_mft_recordp(p)		( ntfs_is_file_recordp(p) )
+#define ntfs_is_indx_record(x)		( ntfs_is_magic (x, INDX) )
+#define ntfs_is_indx_recordp(p)		( ntfs_is_magicp(p, INDX) )
+#define ntfs_is_hole_record(x)		( ntfs_is_magic (x, HOLE) )
+#define ntfs_is_hole_recordp(p)		( ntfs_is_magicp(p, HOLE) )
+
+#define ntfs_is_rstr_record(x)		( ntfs_is_magic (x, RSTR) )
+#define ntfs_is_rstr_recordp(p)		( ntfs_is_magicp(p, RSTR) )
+#define ntfs_is_rcrd_record(x)		( ntfs_is_magic (x, RCRD) )
+#define ntfs_is_rcrd_recordp(p)		( ntfs_is_magicp(p, RCRD) )
+
+#define ntfs_is_chkd_record(x)		( ntfs_is_magic (x, CHKD) )
+#define ntfs_is_chkd_recordp(p)		( ntfs_is_magicp(p, CHKD) )
+
+#define ntfs_is_baad_record(x)		( ntfs_is_magic (x, BAAD) )
+#define ntfs_is_baad_recordp(p)		( ntfs_is_magicp(p, BAAD) )
+
+#define ntfs_is_empty_record(x)		( ntfs_is_magic (x, empty) )
+#define ntfs_is_empty_recordp(p)	( ntfs_is_magicp(p, empty) )
+
+/*
+ * The Update Sequence Array (usa) is an array of the le16 values which belong
+ * to the end of each sector protected by the update sequence record in which
+ * this array is contained. Note that the first entry is the Update Sequence
+ * Number (usn), a cyclic counter of how many times the protected record has
+ * been written to disk. The values 0 and -1 (ie. 0xffff) are not used. All
+ * last le16's of each sector have to be equal to the usn (during reading) or
+ * are set to it (during writing). If they are not, an incomplete multi sector
+ * transfer has occurred when the data was written.
+ * The maximum size for the update sequence array is fixed to:
+ *	maximum size = usa_ofs + (usa_count * 2) = 510 bytes
+ * The 510 bytes comes from the fact that the last le16 in the array has to
+ * (obviously) finish before the last le16 of the first 512-byte sector.
+ * This formula can be used as a consistency check in that usa_ofs +
+ * (usa_count * 2) has to be less than or equal to 510.
+ */
+typedef struct {
+	NTFS_RECORD_TYPE magic;	/* A four-byte magic identifying the record
+				   type and/or status. */
+	le16 usa_ofs;		/* Offset to the Update Sequence Array (usa)
+				   from the start of the ntfs record. */
+	le16 usa_count;		/* Number of le16 sized entries in the usa
+				   including the Update Sequence Number (usn),
+				   thus the number of fixups is the usa_count
+				   minus 1. */
+} __attribute__ ((__packed__)) NTFS_RECORD;
+
+/*
+ * System files mft record numbers. All these files are always marked as used
+ * in the bitmap attribute of the mft; presumably in order to avoid accidental
+ * allocation for random other mft records. Also, the sequence number for each
+ * of the system files is always equal to their mft record number and it is
+ * never modified.
+ */
+typedef enum {
+	FILE_MFT       = 0,	/* Master file table (mft). Data attribute
+				   contains the entries and bitmap attribute
+				   records which ones are in use (bit==1). */
+	FILE_MFTMirr   = 1,	/* Mft mirror: copy of first four mft records
+				   in data attribute. If cluster size > 4kiB,
+				   copy of first N mft records, with
+					N = cluster_size / mft_record_size. */
+	FILE_LogFile   = 2,	/* Journalling log in data attribute. */
+	FILE_Volume    = 3,	/* Volume name attribute and volume information
+				   attribute (flags and ntfs version). Windows
+				   refers to this file as volume DASD (Direct
+				   Access Storage Device). */
+	FILE_AttrDef   = 4,	/* Array of attribute definitions in data
+				   attribute. */
+	FILE_root      = 5,	/* Root directory. */
+	FILE_Bitmap    = 6,	/* Allocation bitmap of all clusters (lcns) in
+				   data attribute. */
+	FILE_Boot      = 7,	/* Boot sector (always at cluster 0) in data
+				   attribute. */
+	FILE_BadClus   = 8,	/* Contains all bad clusters in the non-resident
+				   data attribute. */
+	FILE_Secure    = 9,	/* Shared security descriptors in data attribute
+				   and two indexes into the descriptors.
+				   Appeared in Windows 2000. Before that, this
+				   file was named $Quota but was unused. */
+	FILE_UpCase    = 10,	/* Uppercase equivalents of all 65536 Unicode
+				   characters in data attribute. */
+	FILE_Extend    = 11,	/* Directory containing other system files (eg.
+				   $ObjId, $Quota, $Reparse and $UsnJrnl). This
+				   is new to NTFS3.0. */
+	FILE_reserved12 = 12,	/* Reserved for future use (records 12-15). */
+	FILE_reserved13 = 13,
+	FILE_reserved14 = 14,
+	FILE_reserved15 = 15,
+	FILE_first_user = 16,	/* First user file, used as test limit for
+				   whether to allow opening a file or not. */
+} NTFS_SYSTEM_FILES;
+
+/*
+ * These are the so far known MFT_RECORD_* flags (16-bit) which contain
+ * information about the mft record in which they are present.
+ */
+enum {
+	MFT_RECORD_IN_USE	= cpu_to_le16(0x0001),
+	MFT_RECORD_IS_DIRECTORY = cpu_to_le16(0x0002),
+} __attribute__ ((__packed__));
+
+typedef le16 MFT_RECORD_FLAGS;
+
+/*
+ * mft references (aka file references or file record segment references) are
+ * used whenever a structure needs to refer to a record in the mft.
+ *
+ * A reference consists of a 48-bit index into the mft and a 16-bit sequence
+ * number used to detect stale references.
+ *
+ * For error reporting purposes we treat the 48-bit index as a signed quantity.
+ *
+ * The sequence number is a circular counter (skipping 0) describing how many
+ * times the referenced mft record has been (re)used. This has to match the
+ * sequence number of the mft record being referenced, otherwise the reference
+ * is considered stale and removed (FIXME: only ntfsck or the driver itself?).
+ *
+ * If the sequence number is zero it is assumed that no sequence number
+ * consistency checking should be performed.
+ *
+ * FIXME: Since inodes are 32-bit as of now, the driver needs to always check
+ * for high_part being 0 and if not either BUG(), cause a panic() or handle
+ * the situation in some other way. This shouldn't be a problem as a volume has
+ * to become HUGE in order to need more than 32-bits worth of mft records.
+ * Assuming the standard mft record size of 1kb only the records (never mind
+ * the non-resident attributes, etc.) would require 4Tb of space on their own
+ * for the first 32 bits worth of records. This is only if some strange person
+ * doesn't decide to foul play and make the mft sparse which would be a really
+ * horrible thing to do as it would trash our current driver implementation. )-:
+ * Do I hear screams "we want 64-bit inodes!" ?!? (-;
+ *
+ * FIXME: The mft zone is defined as the first 12% of the volume. This space is
+ * reserved so that the mft can grow contiguously and hence doesn't become
+ * fragmented. Volume free space includes the empty part of the mft zone and
+ * when the volume's free 88% are used up, the mft zone is shrunk by a factor
+ * of 2, thus making more space available for more files/data. This process is
+ * repeated every time there is no more free space except for the mft zone until
+ * there really is no more free space.
+ */
+
+/*
+ * Typedef the MFT_REF as a 64-bit value for easier handling.
+ * Also define two unpacking macros to get to the reference (MREF) and
+ * sequence number (MSEQNO) respectively.
+ * The _LE versions are to be applied on little endian MFT_REFs.
+ * Note: The _LE versions will return a CPU endian formatted value!
+ */
+#define MFT_REF_MASK_CPU 0x0000ffffffffffffULL
+#define MFT_REF_MASK_LE cpu_to_le64(MFT_REF_MASK_CPU)
+
+typedef u64 MFT_REF;
+typedef le64 leMFT_REF;
+
+#define MK_MREF(m, s)	((MFT_REF)(((MFT_REF)(s) << 48) |		\
+					((MFT_REF)(m) & MFT_REF_MASK_CPU)))
+#define MK_LE_MREF(m, s) cpu_to_le64(MK_MREF(m, s))
+
+#define MREF(x)		((unsigned long)((x) & MFT_REF_MASK_CPU))
+#define MSEQNO(x)	((u16)(((x) >> 48) & 0xffff))
+#define MREF_LE(x)	((unsigned long)(le64_to_cpu(x) & MFT_REF_MASK_CPU))
+#define MSEQNO_LE(x)	((u16)((le64_to_cpu(x) >> 48) & 0xffff))
+
+#define IS_ERR_MREF(x)	(((x) & 0x0000800000000000ULL) ? true : false)
+#define ERR_MREF(x)	((u64)((s64)(x)))
+#define MREF_ERR(x)	((int)((s64)(x)))
+
+/*
+ * The mft record header present at the beginning of every record in the mft.
+ * This is followed by a sequence of variable length attribute records which
+ * is terminated by an attribute of type AT_END which is a truncated attribute
+ * in that it only consists of the attribute type code AT_END and none of the
+ * other members of the attribute structure are present.
+ */
+typedef struct {
+/*Ofs*/
+/*  0	NTFS_RECORD; -- Unfolded here as gcc doesn't like unnamed structs. */
+	NTFS_RECORD_TYPE magic;	/* Usually the magic is "FILE". */
+	le16 usa_ofs;		/* See NTFS_RECORD definition above. */
+	le16 usa_count;		/* See NTFS_RECORD definition above. */
+
+/*  8*/	le64 lsn;		/* $LogFile sequence number for this record.
+				   Changed every time the record is modified. */
+/* 16*/	le16 sequence_number;	/* Number of times this mft record has been
+				   reused. (See description for MFT_REF
+				   above.) NOTE: The increment (skipping zero)
+				   is done when the file is deleted. NOTE: If
+				   this is zero it is left zero. */
+/* 18*/	le16 link_count;	/* Number of hard links, i.e. the number of
+				   directory entries referencing this record.
+				   NOTE: Only used in mft base records.
+				   NOTE: When deleting a directory entry we
+				   check the link_count and if it is 1 we
+				   delete the file. Otherwise we delete the
+				   FILE_NAME_ATTR being referenced by the
+				   directory entry from the mft record and
+				   decrement the link_count.
+				   FIXME: Careful with Win32 + DOS names! */
+/* 20*/	le16 attrs_offset;	/* Byte offset to the first attribute in this
+				   mft record from the start of the mft record.
+				   NOTE: Must be aligned to 8-byte boundary. */
+/* 22*/	MFT_RECORD_FLAGS flags;	/* Bit array of MFT_RECORD_FLAGS. When a file
+				   is deleted, the MFT_RECORD_IN_USE flag is
+				   set to zero. */
+/* 24*/	le32 bytes_in_use;	/* Number of bytes used in this mft record.
+				   NOTE: Must be aligned to 8-byte boundary. */
+/* 28*/	le32 bytes_allocated;	/* Number of bytes allocated for this mft
+				   record. This should be equal to the mft
+				   record size. */
+/* 32*/	leMFT_REF base_mft_record;/* This is zero for base mft records.
+				   When it is not zero it is a mft reference
+				   pointing to the base mft record to which
+				   this record belongs (this is then used to
+				   locate the attribute list attribute present
+				   in the base record which describes this
+				   extension record and hence might need
+				   modification when the extension record
+				   itself is modified, also locating the
+				   attribute list also means finding the other
+				   potential extents, belonging to the non-base
+				   mft record). */
+/* 40*/	le16 next_attr_instance;/* The instance number that will be assigned to
+				   the next attribute added to this mft record.
+				   NOTE: Incremented each time after it is used.
+				   NOTE: Every time the mft record is reused
+				   this number is set to zero.  NOTE: The first
+				   instance number is always 0. */
+/* The below fields are specific to NTFS 3.1+ (Windows XP and above): */
+/* 42*/ le16 reserved;		/* Reserved/alignment. */
+/* 44*/ le32 mft_record_number;	/* Number of this mft record. */
+/* sizeof() = 48 bytes */
+/*
+ * When (re)using the mft record, we place the update sequence array at this
+ * offset, i.e. before we start with the attributes.  This also makes sense,
+ * otherwise we could run into problems with the update sequence array
+ * containing in itself the last two bytes of a sector which would mean that
+ * multi sector transfer protection wouldn't work.  As you can't protect data
+ * by overwriting it since you then can't get it back...
+ * When reading we obviously use the data from the ntfs record header.
+ */
+} __attribute__ ((__packed__)) MFT_RECORD;
+
+/* This is the version without the NTFS 3.1+ specific fields. */
+typedef struct {
+/*Ofs*/
+/*  0	NTFS_RECORD; -- Unfolded here as gcc doesn't like unnamed structs. */
+	NTFS_RECORD_TYPE magic;	/* Usually the magic is "FILE". */
+	le16 usa_ofs;		/* See NTFS_RECORD definition above. */
+	le16 usa_count;		/* See NTFS_RECORD definition above. */
+
+/*  8*/	le64 lsn;		/* $LogFile sequence number for this record.
+				   Changed every time the record is modified. */
+/* 16*/	le16 sequence_number;	/* Number of times this mft record has been
+				   reused. (See description for MFT_REF
+				   above.) NOTE: The increment (skipping zero)
+				   is done when the file is deleted. NOTE: If
+				   this is zero it is left zero. */
+/* 18*/	le16 link_count;	/* Number of hard links, i.e. the number of
+				   directory entries referencing this record.
+				   NOTE: Only used in mft base records.
+				   NOTE: When deleting a directory entry we
+				   check the link_count and if it is 1 we
+				   delete the file. Otherwise we delete the
+				   FILE_NAME_ATTR being referenced by the
+				   directory entry from the mft record and
+				   decrement the link_count.
+				   FIXME: Careful with Win32 + DOS names! */
+/* 20*/	le16 attrs_offset;	/* Byte offset to the first attribute in this
+				   mft record from the start of the mft record.
+				   NOTE: Must be aligned to 8-byte boundary. */
+/* 22*/	MFT_RECORD_FLAGS flags;	/* Bit array of MFT_RECORD_FLAGS. When a file
+				   is deleted, the MFT_RECORD_IN_USE flag is
+				   set to zero. */
+/* 24*/	le32 bytes_in_use;	/* Number of bytes used in this mft record.
+				   NOTE: Must be aligned to 8-byte boundary. */
+/* 28*/	le32 bytes_allocated;	/* Number of bytes allocated for this mft
+				   record. This should be equal to the mft
+				   record size. */
+/* 32*/	leMFT_REF base_mft_record;/* This is zero for base mft records.
+				   When it is not zero it is a mft reference
+				   pointing to the base mft record to which
+				   this record belongs (this is then used to
+				   locate the attribute list attribute present
+				   in the base record which describes this
+				   extension record and hence might need
+				   modification when the extension record
+				   itself is modified, also locating the
+				   attribute list also means finding the other
+				   potential extents, belonging to the non-base
+				   mft record). */
+/* 40*/	le16 next_attr_instance;/* The instance number that will be assigned to
+				   the next attribute added to this mft record.
+				   NOTE: Incremented each time after it is used.
+				   NOTE: Every time the mft record is reused
+				   this number is set to zero.  NOTE: The first
+				   instance number is always 0. */
+/* sizeof() = 42 bytes */
+/*
+ * When (re)using the mft record, we place the update sequence array at this
+ * offset, i.e. before we start with the attributes.  This also makes sense,
+ * otherwise we could run into problems with the update sequence array
+ * containing in itself the last two bytes of a sector which would mean that
+ * multi sector transfer protection wouldn't work.  As you can't protect data
+ * by overwriting it since you then can't get it back...
+ * When reading we obviously use the data from the ntfs record header.
+ */
+} __attribute__ ((__packed__)) MFT_RECORD_OLD;
+
+/*
+ * System defined attributes (32-bit).  Each attribute type has a corresponding
+ * attribute name (Unicode string of maximum 64 character length) as described
+ * by the attribute definitions present in the data attribute of the $AttrDef
+ * system file.  On NTFS 3.0 volumes the names are just as the types are named
+ * in the below defines exchanging AT_ for the dollar sign ($).  If that is not
+ * a revealing choice of symbol I do not know what is... (-;
+ */
+enum {
+	AT_UNUSED			= cpu_to_le32(         0),
+	AT_STANDARD_INFORMATION		= cpu_to_le32(      0x10),
+	AT_ATTRIBUTE_LIST		= cpu_to_le32(      0x20),
+	AT_FILE_NAME			= cpu_to_le32(      0x30),
+	AT_OBJECT_ID			= cpu_to_le32(      0x40),
+	AT_SECURITY_DESCRIPTOR		= cpu_to_le32(      0x50),
+	AT_VOLUME_NAME			= cpu_to_le32(      0x60),
+	AT_VOLUME_INFORMATION		= cpu_to_le32(      0x70),
+	AT_DATA				= cpu_to_le32(      0x80),
+	AT_INDEX_ROOT			= cpu_to_le32(      0x90),
+	AT_INDEX_ALLOCATION		= cpu_to_le32(      0xa0),
+	AT_BITMAP			= cpu_to_le32(      0xb0),
+	AT_REPARSE_POINT		= cpu_to_le32(      0xc0),
+	AT_EA_INFORMATION		= cpu_to_le32(      0xd0),
+	AT_EA				= cpu_to_le32(      0xe0),
+	AT_PROPERTY_SET			= cpu_to_le32(      0xf0),
+	AT_LOGGED_UTILITY_STREAM	= cpu_to_le32(     0x100),
+	AT_FIRST_USER_DEFINED_ATTRIBUTE	= cpu_to_le32(    0x1000),
+	AT_END				= cpu_to_le32(0xffffffff)
+};
+
+typedef le32 ATTR_TYPE;
+
+/*
+ * The collation rules for sorting views/indexes/etc (32-bit).
+ *
+ * COLLATION_BINARY - Collate by binary compare where the first byte is most
+ *	significant.
+ * COLLATION_UNICODE_STRING - Collate Unicode strings by comparing their binary
+ *	Unicode values, except that when a character can be uppercased, the
+ *	upper case value collates before the lower case one.
+ * COLLATION_FILE_NAME - Collate file names as Unicode strings. The collation
+ *	is done very much like COLLATION_UNICODE_STRING. In fact I have no idea
+ *	what the difference is. Perhaps the difference is that file names
+ *	would treat some special characters in an odd way (see
+ *	unistr.c::ntfs_collate_names() and unistr.c::legal_ansi_char_array[]
+ *	for what I mean but COLLATION_UNICODE_STRING would not give any special
+ *	treatment to any characters at all, but this is speculation.
+ * COLLATION_NTOFS_ULONG - Sorting is done according to ascending le32 key
+ *	values. E.g. used for $SII index in FILE_Secure, which sorts by
+ *	security_id (le32).
+ * COLLATION_NTOFS_SID - Sorting is done according to ascending SID values.
+ *	E.g. used for $O index in FILE_Extend/$Quota.
+ * COLLATION_NTOFS_SECURITY_HASH - Sorting is done first by ascending hash
+ *	values and second by ascending security_id values. E.g. used for $SDH
+ *	index in FILE_Secure.
+ * COLLATION_NTOFS_ULONGS - Sorting is done according to a sequence of ascending
+ *	le32 key values. E.g. used for $O index in FILE_Extend/$ObjId, which
+ *	sorts by object_id (16-byte), by splitting up the object_id in four
+ *	le32 values and using them as individual keys. E.g. take the following
+ *	two security_ids, stored as follows on disk:
+ *		1st: a1 61 65 b7 65 7b d4 11 9e 3d 00 e0 81 10 42 59
+ *		2nd: 38 14 37 d2 d2 f3 d4 11 a5 21 c8 6b 79 b1 97 45
+ *	To compare them, they are split into four le32 values each, like so:
+ *		1st: 0xb76561a1 0x11d47b65 0xe0003d9e 0x59421081
+ *		2nd: 0xd2371438 0x11d4f3d2 0x6bc821a5 0x4597b179
+ *	Now, it is apparent why the 2nd object_id collates after the 1st: the
+ *	first le32 value of the 1st object_id is less than the first le32 of
+ *	the 2nd object_id. If the first le32 values of both object_ids were
+ *	equal then the second le32 values would be compared, etc.
+ */
+enum {
+	COLLATION_BINARY		= cpu_to_le32(0x00),
+	COLLATION_FILE_NAME		= cpu_to_le32(0x01),
+	COLLATION_UNICODE_STRING	= cpu_to_le32(0x02),
+	COLLATION_NTOFS_ULONG		= cpu_to_le32(0x10),
+	COLLATION_NTOFS_SID		= cpu_to_le32(0x11),
+	COLLATION_NTOFS_SECURITY_HASH	= cpu_to_le32(0x12),
+	COLLATION_NTOFS_ULONGS		= cpu_to_le32(0x13),
+};
+
+typedef le32 COLLATION_RULE;
+
+/*
+ * The flags (32-bit) describing attribute properties in the attribute
+ * definition structure.  FIXME: This information is based on Regis's
+ * information and, according to him, it is not certain and probably
+ * incomplete.  The INDEXABLE flag is fairly certainly correct as only the file
+ * name attribute has this flag set and this is the only attribute indexed in
+ * NT4.
+ */
+enum {
+	ATTR_DEF_INDEXABLE	= cpu_to_le32(0x02), /* Attribute can be
+					indexed. */
+	ATTR_DEF_MULTIPLE	= cpu_to_le32(0x04), /* Attribute type
+					can be present multiple times in the
+					mft records of an inode. */
+	ATTR_DEF_NOT_ZERO	= cpu_to_le32(0x08), /* Attribute value
+					must contain at least one non-zero
+					byte. */
+	ATTR_DEF_INDEXED_UNIQUE	= cpu_to_le32(0x10), /* Attribute must be
+					indexed and the attribute value must be
+					unique for the attribute type in all of
+					the mft records of an inode. */
+	ATTR_DEF_NAMED_UNIQUE	= cpu_to_le32(0x20), /* Attribute must be
+					named and the name must be unique for
+					the attribute type in all of the mft
+					records of an inode. */
+	ATTR_DEF_RESIDENT	= cpu_to_le32(0x40), /* Attribute must be
+					resident. */
+	ATTR_DEF_ALWAYS_LOG	= cpu_to_le32(0x80), /* Always log
+					modifications to this attribute,
+					regardless of whether it is resident or
+					non-resident.  Without this, only log
+					modifications if the attribute is
+					resident. */
+};
+
+typedef le32 ATTR_DEF_FLAGS;
+
+/*
+ * The data attribute of FILE_AttrDef contains a sequence of attribute
+ * definitions for the NTFS volume. With this, it is supposed to be safe for an
+ * older NTFS driver to mount a volume containing a newer NTFS version without
+ * damaging it (that's the theory. In practice it's: not damaging it too much).
+ * Entries are sorted by attribute type. The flags describe whether the
+ * attribute can be resident/non-resident and possibly other things, but the
+ * actual bits are unknown.
+ */
+typedef struct {
+/*hex ofs*/
+/*  0*/	ntfschar name[0x40];		/* Unicode name of the attribute. Zero
+					   terminated. */
+/* 80*/	ATTR_TYPE type;			/* Type of the attribute. */
+/* 84*/	le32 display_rule;		/* Default display rule.
+					   FIXME: What does it mean? (AIA) */
+/* 88*/ COLLATION_RULE collation_rule;	/* Default collation rule. */
+/* 8c*/	ATTR_DEF_FLAGS flags;		/* Flags describing the attribute. */
+/* 90*/	sle64 min_size;			/* Optional minimum attribute size. */
+/* 98*/	sle64 max_size;			/* Maximum size of attribute. */
+/* sizeof() = 0xa0 or 160 bytes */
+} __attribute__ ((__packed__)) ATTR_DEF;
+
+/*
+ * Attribute flags (16-bit).
+ */
+enum {
+	ATTR_IS_COMPRESSED    = cpu_to_le16(0x0001),
+	ATTR_COMPRESSION_MASK = cpu_to_le16(0x00ff), /* Compression method
+							      mask.  Also, first
+							      illegal value. */
+	ATTR_IS_ENCRYPTED     = cpu_to_le16(0x4000),
+	ATTR_IS_SPARSE	      = cpu_to_le16(0x8000),
+} __attribute__ ((__packed__));
+
+typedef le16 ATTR_FLAGS;
+
+/*
+ * Attribute compression.
+ *
+ * Only the data attribute is ever compressed in the current ntfs driver in
+ * Windows. Further, compression is only applied when the data attribute is
+ * non-resident. Finally, to use compression, the maximum allowed cluster size
+ * on a volume is 4kib.
+ *
+ * The compression method is based on independently compressing blocks of X
+ * clusters, where X is determined from the compression_unit value found in the
+ * non-resident attribute record header (more precisely: X = 2^compression_unit
+ * clusters). On Windows NT/2k, X always is 16 clusters (compression_unit = 4).
+ *
+ * There are three different cases of how a compression block of X clusters
+ * can be stored:
+ *
+ *   1) The data in the block is all zero (a sparse block):
+ *	  This is stored as a sparse block in the runlist, i.e. the runlist
+ *	  entry has length = X and lcn = -1. The mapping pairs array actually
+ *	  uses a delta_lcn value length of 0, i.e. delta_lcn is not present at
+ *	  all, which is then interpreted by the driver as lcn = -1.
+ *	  NOTE: Even uncompressed files can be sparse on NTFS 3.0 volumes, then
+ *	  the same principles apply as above, except that the length is not
+ *	  restricted to being any particular value.
+ *
+ *   2) The data in the block is not compressed:
+ *	  This happens when compression doesn't reduce the size of the block
+ *	  in clusters. I.e. if compression has a small effect so that the
+ *	  compressed data still occupies X clusters, then the uncompressed data
+ *	  is stored in the block.
+ *	  This case is recognised by the fact that the runlist entry has
+ *	  length = X and lcn >= 0. The mapping pairs array stores this as
+ *	  normal with a run length of X and some specific delta_lcn, i.e.
+ *	  delta_lcn has to be present.
+ *
+ *   3) The data in the block is compressed:
+ *	  The common case. This case is recognised by the fact that the run
+ *	  list entry has length L < X and lcn >= 0. The mapping pairs array
+ *	  stores this as normal with a run length of X and some specific
+ *	  delta_lcn, i.e. delta_lcn has to be present. This runlist entry is
+ *	  immediately followed by a sparse entry with length = X - L and
+ *	  lcn = -1. The latter entry is to make up the vcn counting to the
+ *	  full compression block size X.
+ *
+ * In fact, life is more complicated because adjacent entries of the same type
+ * can be coalesced. This means that one has to keep track of the number of
+ * clusters handled and work on a basis of X clusters at a time being one
+ * block. An example: if length L > X this means that this particular runlist
+ * entry contains a block of length X and part of one or more blocks of length
+ * L - X. Another example: if length L < X, this does not necessarily mean that
+ * the block is compressed as it might be that the lcn changes inside the block
+ * and hence the following runlist entry describes the continuation of the
+ * potentially compressed block. The block would be compressed if the
+ * following runlist entry describes at least X - L sparse clusters, thus
+ * making up the compression block length as described in point 3 above. (Of
+ * course, there can be several runlist entries with small lengths so that the
+ * sparse entry does not follow the first data containing entry with
+ * length < X.)
+ *
+ * NOTE: At the end of the compressed attribute value, there most likely is not
+ * just the right amount of data to make up a compression block, thus this data
+ * is not even attempted to be compressed. It is just stored as is, unless
+ * the number of clusters it occupies is reduced when compressed in which case
+ * it is stored as a compressed compression block, complete with sparse
+ * clusters at the end.
+ */
+
+/*
+ * Flags of resident attributes (8-bit).
+ */
+enum {
+	RESIDENT_ATTR_IS_INDEXED = 0x01, /* Attribute is referenced in an index
+					    (has implications for deleting and
+					    modifying the attribute). */
+} __attribute__ ((__packed__));
+
+typedef u8 RESIDENT_ATTR_FLAGS;
+
+/*
+ * Attribute record header. Always aligned to 8-byte boundary.
+ */
+typedef struct {
+/*Ofs*/
+/*  0*/	ATTR_TYPE type;		/* The (32-bit) type of the attribute. */
+/*  4*/	le32 length;		/* Byte size of the resident part of the
+				   attribute (aligned to 8-byte boundary).
+				   Used to get to the next attribute. */
+/*  8*/	u8 non_resident;	/* If 0, attribute is resident.
+				   If 1, attribute is non-resident. */
+/*  9*/	u8 name_length;		/* Unicode character size of name of attribute.
+				   0 if unnamed. */
+/* 10*/	le16 name_offset;	/* If name_length != 0, the byte offset to the
+				   beginning of the name from the attribute
+				   record. Note that the name is stored as a
+				   Unicode string. When creating, place offset
+				   just at the end of the record header. Then,
+				   follow with attribute value or mapping pairs
+				   array, resident and non-resident attributes
+				   respectively, aligning to an 8-byte
+				   boundary. */
+/* 12*/	ATTR_FLAGS flags;	/* Flags describing the attribute. */
+/* 14*/	le16 instance;		/* The instance of this attribute record. This
+				   number is unique within this mft record (see
+				   MFT_RECORD/next_attribute_instance notes in
+				   in mft.h for more details). */
+/* 16*/	union {
+		/* Resident attributes. */
+		struct {
+/* 16 */		le32 value_length;/* Byte size of attribute value. */
+/* 20 */		le16 value_offset;/* Byte offset of the attribute
+					     value from the start of the
+					     attribute record. When creating,
+					     align to 8-byte boundary if we
+					     have a name present as this might
+					     not have a length of a multiple
+					     of 8-bytes. */
+/* 22 */		RESIDENT_ATTR_FLAGS flags; /* See above. */
+/* 23 */		s8 reserved;	  /* Reserved/alignment to 8-byte
+					     boundary. */
+		} __attribute__ ((__packed__)) resident;
+		/* Non-resident attributes. */
+		struct {
+/* 16*/			leVCN lowest_vcn;/* Lowest valid virtual cluster number
+				for this portion of the attribute value or
+				0 if this is the only extent (usually the
+				case). - Only when an attribute list is used
+				does lowest_vcn != 0 ever occur. */
+/* 24*/			leVCN highest_vcn;/* Highest valid vcn of this extent of
+				the attribute value. - Usually there is only one
+				portion, so this usually equals the attribute
+				value size in clusters minus 1. Can be -1 for
+				zero length files. Can be 0 for "single extent"
+				attributes. */
+/* 32*/			le16 mapping_pairs_offset; /* Byte offset from the
+				beginning of the structure to the mapping pairs
+				array which contains the mappings between the
+				vcns and the logical cluster numbers (lcns).
+				When creating, place this at the end of this
+				record header aligned to 8-byte boundary. */
+/* 34*/			u8 compression_unit; /* The compression unit expressed
+				as the log to the base 2 of the number of
+				clusters in a compression unit.  0 means not
+				compressed.  (This effectively limits the
+				compression unit size to be a power of two
+				clusters.)  WinNT4 only uses a value of 4.
+				Sparse files have this set to 0 on XPSP2. */
+/* 35*/			u8 reserved[5];		/* Align to 8-byte boundary. */
+/* The sizes below are only used when lowest_vcn is zero, as otherwise it would
+   be difficult to keep them up-to-date.*/
+/* 40*/			sle64 allocated_size;	/* Byte size of disk space
+				allocated to hold the attribute value. Always
+				is a multiple of the cluster size. When a file
+				is compressed, this field is a multiple of the
+				compression block size (2^compression_unit) and
+				it represents the logically allocated space
+				rather than the actual on disk usage. For this
+				use the compressed_size (see below). */
+/* 48*/			sle64 data_size;	/* Byte size of the attribute
+				value. Can be larger than allocated_size if
+				attribute value is compressed or sparse. */
+/* 56*/			sle64 initialized_size;	/* Byte size of initialized
+				portion of the attribute value. Usually equals
+				data_size. */
+/* sizeof(uncompressed attr) = 64*/
+/* 64*/			sle64 compressed_size;	/* Byte size of the attribute
+				value after compression.  Only present when
+				compressed or sparse.  Always is a multiple of
+				the cluster size.  Represents the actual amount
+				of disk space being used on the disk. */
+/* sizeof(compressed attr) = 72*/
+		} __attribute__ ((__packed__)) non_resident;
+	} __attribute__ ((__packed__)) data;
+} __attribute__ ((__packed__)) ATTR_RECORD;
+
+typedef ATTR_RECORD ATTR_REC;
+
+/*
+ * File attribute flags (32-bit) appearing in the file_attributes fields of the
+ * STANDARD_INFORMATION attribute of MFT_RECORDs and the FILENAME_ATTR
+ * attributes of MFT_RECORDs and directory index entries.
+ *
+ * All of the below flags appear in the directory index entries but only some
+ * appear in the STANDARD_INFORMATION attribute whilst only some others appear
+ * in the FILENAME_ATTR attribute of MFT_RECORDs.  Unless otherwise stated the
+ * flags appear in all of the above.
+ */
+enum {
+	FILE_ATTR_READONLY		= cpu_to_le32(0x00000001),
+	FILE_ATTR_HIDDEN		= cpu_to_le32(0x00000002),
+	FILE_ATTR_SYSTEM		= cpu_to_le32(0x00000004),
+	/* Old DOS volid. Unused in NT.	= cpu_to_le32(0x00000008), */
+
+	FILE_ATTR_DIRECTORY		= cpu_to_le32(0x00000010),
+	/* Note, FILE_ATTR_DIRECTORY is not considered valid in NT.  It is
+	   reserved for the DOS SUBDIRECTORY flag. */
+	FILE_ATTR_ARCHIVE		= cpu_to_le32(0x00000020),
+	FILE_ATTR_DEVICE		= cpu_to_le32(0x00000040),
+	FILE_ATTR_NORMAL		= cpu_to_le32(0x00000080),
+
+	FILE_ATTR_TEMPORARY		= cpu_to_le32(0x00000100),
+	FILE_ATTR_SPARSE_FILE		= cpu_to_le32(0x00000200),
+	FILE_ATTR_REPARSE_POINT		= cpu_to_le32(0x00000400),
+	FILE_ATTR_COMPRESSED		= cpu_to_le32(0x00000800),
+
+	FILE_ATTR_OFFLINE		= cpu_to_le32(0x00001000),
+	FILE_ATTR_NOT_CONTENT_INDEXED	= cpu_to_le32(0x00002000),
+	FILE_ATTR_ENCRYPTED		= cpu_to_le32(0x00004000),
+
+	FILE_ATTR_VALID_FLAGS		= cpu_to_le32(0x00007fb7),
+	/* Note, FILE_ATTR_VALID_FLAGS masks out the old DOS VolId and the
+	   FILE_ATTR_DEVICE and preserves everything else.  This mask is used
+	   to obtain all flags that are valid for reading. */
+	FILE_ATTR_VALID_SET_FLAGS	= cpu_to_le32(0x000031a7),
+	/* Note, FILE_ATTR_VALID_SET_FLAGS masks out the old DOS VolId, the
+	   F_A_DEVICE, F_A_DIRECTORY, F_A_SPARSE_FILE, F_A_REPARSE_POINT,
+	   F_A_COMPRESSED, and F_A_ENCRYPTED and preserves the rest.  This mask
+	   is used to obtain all flags that are valid for setting. */
+	/*
+	 * The flag FILE_ATTR_DUP_FILENAME_INDEX_PRESENT is present in all
+	 * FILENAME_ATTR attributes but not in the STANDARD_INFORMATION
+	 * attribute of an mft record.
+	 */
+	FILE_ATTR_DUP_FILE_NAME_INDEX_PRESENT	= cpu_to_le32(0x10000000),
+	/* Note, this is a copy of the corresponding bit from the mft record,
+	   telling us whether this is a directory or not, i.e. whether it has
+	   an index root attribute or not. */
+	FILE_ATTR_DUP_VIEW_INDEX_PRESENT	= cpu_to_le32(0x20000000),
+	/* Note, this is a copy of the corresponding bit from the mft record,
+	   telling us whether this file has a view index present (eg. object id
+	   index, quota index, one of the security indexes or the encrypting
+	   filesystem related indexes). */
+};
+
+typedef le32 FILE_ATTR_FLAGS;
+
+/*
+ * NOTE on times in NTFS: All times are in MS standard time format, i.e. they
+ * are the number of 100-nanosecond intervals since 1st January 1601, 00:00:00
+ * universal coordinated time (UTC). (In Linux time starts 1st January 1970,
+ * 00:00:00 UTC and is stored as the number of 1-second intervals since then.)
+ */
+
+/*
+ * Attribute: Standard information (0x10).
+ *
+ * NOTE: Always resident.
+ * NOTE: Present in all base file records on a volume.
+ * NOTE: There is conflicting information about the meaning of each of the time
+ *	 fields but the meaning as defined below has been verified to be
+ *	 correct by practical experimentation on Windows NT4 SP6a and is hence
+ *	 assumed to be the one and only correct interpretation.
+ */
+typedef struct {
+/*Ofs*/
+/*  0*/	sle64 creation_time;		/* Time file was created. Updated when
+					   a filename is changed(?). */
+/*  8*/	sle64 last_data_change_time;	/* Time the data attribute was last
+					   modified. */
+/* 16*/	sle64 last_mft_change_time;	/* Time this mft record was last
+					   modified. */
+/* 24*/	sle64 last_access_time;		/* Approximate time when the file was
+					   last accessed (obviously this is not
+					   updated on read-only volumes). In
+					   Windows this is only updated when
+					   accessed if some time delta has
+					   passed since the last update. Also,
+					   last access time updates can be
+					   disabled altogether for speed. */
+/* 32*/	FILE_ATTR_FLAGS file_attributes; /* Flags describing the file. */
+/* 36*/	union {
+	/* NTFS 1.2 */
+		struct {
+		/* 36*/	u8 reserved12[12];	/* Reserved/alignment to 8-byte
+						   boundary. */
+		} __attribute__ ((__packed__)) v1;
+	/* sizeof() = 48 bytes */
+	/* NTFS 3.x */
+		struct {
+/*
+ * If a volume has been upgraded from a previous NTFS version, then these
+ * fields are present only if the file has been accessed since the upgrade.
+ * Recognize the difference by comparing the length of the resident attribute
+ * value. If it is 48, then the following fields are missing. If it is 72 then
+ * the fields are present. Maybe just check like this:
+ *	if (resident.ValueLength < sizeof(STANDARD_INFORMATION)) {
+ *		Assume NTFS 1.2- format.
+ *		If (volume version is 3.x)
+ *			Upgrade attribute to NTFS 3.x format.
+ *		else
+ *			Use NTFS 1.2- format for access.
+ *	} else
+ *		Use NTFS 3.x format for access.
+ * Only problem is that it might be legal to set the length of the value to
+ * arbitrarily large values thus spoiling this check. - But chkdsk probably
+ * views that as a corruption, assuming that it behaves like this for all
+ * attributes.
+ */
+		/* 36*/	le32 maximum_versions;	/* Maximum allowed versions for
+				file. Zero if version numbering is disabled. */
+		/* 40*/	le32 version_number;	/* This file's version (if any).
+				Set to zero if maximum_versions is zero. */
+		/* 44*/	le32 class_id;		/* Class id from bidirectional
+				class id index (?). */
+		/* 48*/	le32 owner_id;		/* Owner_id of the user owning
+				the file. Translate via $Q index in FILE_Extend
+				/$Quota to the quota control entry for the user
+				owning the file. Zero if quotas are disabled. */
+		/* 52*/	le32 security_id;	/* Security_id for the file.
+				Translate via $SII index and $SDS data stream
+				in FILE_Secure to the security descriptor. */
+		/* 56*/	le64 quota_charged;	/* Byte size of the charge to
+				the quota for all streams of the file. Note: Is
+				zero if quotas are disabled. */
+		/* 64*/	leUSN usn;		/* Last update sequence number
+				of the file.  This is a direct index into the
+				transaction log file ($UsnJrnl).  It is zero if
+				the usn journal is disabled or this file has
+				not been subject to logging yet.  See usnjrnl.h
+				for details. */
+		} __attribute__ ((__packed__)) v3;
+	/* sizeof() = 72 bytes (NTFS 3.x) */
+	} __attribute__ ((__packed__)) ver;
+} __attribute__ ((__packed__)) STANDARD_INFORMATION;
+
+/*
+ * Attribute: Attribute list (0x20).
+ *
+ * - Can be either resident or non-resident.
+ * - Value consists of a sequence of variable length, 8-byte aligned,
+ * ATTR_LIST_ENTRY records.
+ * - The list is not terminated by anything at all! The only way to know when
+ * the end is reached is to keep track of the current offset and compare it to
+ * the attribute value size.
+ * - The attribute list attribute contains one entry for each attribute of
+ * the file in which the list is located, except for the list attribute
+ * itself. The list is sorted: first by attribute type, second by attribute
+ * name (if present), third by instance number. The extents of one
+ * non-resident attribute (if present) immediately follow after the initial
+ * extent. They are ordered by lowest_vcn and have their instace set to zero.
+ * It is not allowed to have two attributes with all sorting keys equal.
+ * - Further restrictions:
+ *	- If not resident, the vcn to lcn mapping array has to fit inside the
+ *	  base mft record.
+ *	- The attribute list attribute value has a maximum size of 256kb. This
+ *	  is imposed by the Windows cache manager.
+ * - Attribute lists are only used when the attributes of mft record do not
+ * fit inside the mft record despite all attributes (that can be made
+ * non-resident) having been made non-resident. This can happen e.g. when:
+ *	- File has a large number of hard links (lots of file name
+ *	  attributes present).
+ *	- The mapping pairs array of some non-resident attribute becomes so
+ *	  large due to fragmentation that it overflows the mft record.
+ *	- The security descriptor is very complex (not applicable to
+ *	  NTFS 3.0 volumes).
+ *	- There are many named streams.
+ */
+typedef struct {
+/*Ofs*/
+/*  0*/	ATTR_TYPE type;		/* Type of referenced attribute. */
+/*  4*/	le16 length;		/* Byte size of this entry (8-byte aligned). */
+/*  6*/	u8 name_length;		/* Size in Unicode chars of the name of the
+				   attribute or 0 if unnamed. */
+/*  7*/	u8 name_offset;		/* Byte offset to beginning of attribute name
+				   (always set this to where the name would
+				   start even if unnamed). */
+/*  8*/	leVCN lowest_vcn;	/* Lowest virtual cluster number of this portion
+				   of the attribute value. This is usually 0. It
+				   is non-zero for the case where one attribute
+				   does not fit into one mft record and thus
+				   several mft records are allocated to hold
+				   this attribute. In the latter case, each mft
+				   record holds one extent of the attribute and
+				   there is one attribute list entry for each
+				   extent. NOTE: This is DEFINITELY a signed
+				   value! The windows driver uses cmp, followed
+				   by jg when comparing this, thus it treats it
+				   as signed. */
+/* 16*/	leMFT_REF mft_reference;/* The reference of the mft record holding
+				   the ATTR_RECORD for this portion of the
+				   attribute value. */
+/* 24*/	le16 instance;		/* If lowest_vcn = 0, the instance of the
+				   attribute being referenced; otherwise 0. */
+/* 26*/	ntfschar name[0];	/* Use when creating only. When reading use
+				   name_offset to determine the location of the
+				   name. */
+/* sizeof() = 26 + (attribute_name_length * 2) bytes */
+} __attribute__ ((__packed__)) ATTR_LIST_ENTRY;
+
+/*
+ * The maximum allowed length for a file name.
+ */
+#define MAXIMUM_FILE_NAME_LENGTH	255
+
+/*
+ * Possible namespaces for filenames in ntfs (8-bit).
+ */
+enum {
+	FILE_NAME_POSIX		= 0x00,
+	/* This is the largest namespace. It is case sensitive and allows all
+	   Unicode characters except for: '\0' and '/'.  Beware that in
+	   WinNT/2k/2003 by default files which eg have the same name except
+	   for their case will not be distinguished by the standard utilities
+	   and thus a "del filename" will delete both "filename" and "fileName"
+	   without warning.  However if for example Services For Unix (SFU) are
+	   installed and the case sensitive option was enabled at installation
+	   time, then you can create/access/delete such files.
+	   Note that even SFU places restrictions on the filenames beyond the
+	   '\0' and '/' and in particular the following set of characters is
+	   not allowed: '"', '/', '<', '>', '\'.  All other characters,
+	   including the ones no allowed in WIN32 namespace are allowed.
+	   Tested with SFU 3.5 (this is now free) running on Windows XP. */
+	FILE_NAME_WIN32		= 0x01,
+	/* The standard WinNT/2k NTFS long filenames. Case insensitive.  All
+	   Unicode chars except: '\0', '"', '*', '/', ':', '<', '>', '?', '\',
+	   and '|'.  Further, names cannot end with a '.' or a space. */
+	FILE_NAME_DOS		= 0x02,
+	/* The standard DOS filenames (8.3 format). Uppercase only.  All 8-bit
+	   characters greater space, except: '"', '*', '+', ',', '/', ':', ';',
+	   '<', '=', '>', '?', and '\'. */
+	FILE_NAME_WIN32_AND_DOS	= 0x03,
+	/* 3 means that both the Win32 and the DOS filenames are identical and
+	   hence have been saved in this single filename record. */
+} __attribute__ ((__packed__));
+
+typedef u8 FILE_NAME_TYPE_FLAGS;
+
+/*
+ * Attribute: Filename (0x30).
+ *
+ * NOTE: Always resident.
+ * NOTE: All fields, except the parent_directory, are only updated when the
+ *	 filename is changed. Until then, they just become out of sync with
+ *	 reality and the more up to date values are present in the standard
+ *	 information attribute.
+ * NOTE: There is conflicting information about the meaning of each of the time
+ *	 fields but the meaning as defined below has been verified to be
+ *	 correct by practical experimentation on Windows NT4 SP6a and is hence
+ *	 assumed to be the one and only correct interpretation.
+ */
+typedef struct {
+/*hex ofs*/
+/*  0*/	leMFT_REF parent_directory;	/* Directory this filename is
+					   referenced from. */
+/*  8*/	sle64 creation_time;		/* Time file was created. */
+/* 10*/	sle64 last_data_change_time;	/* Time the data attribute was last
+					   modified. */
+/* 18*/	sle64 last_mft_change_time;	/* Time this mft record was last
+					   modified. */
+/* 20*/	sle64 last_access_time;		/* Time this mft record was last
+					   accessed. */
+/* 28*/	sle64 allocated_size;		/* Byte size of on-disk allocated space
+					   for the unnamed data attribute.  So
+					   for normal $DATA, this is the
+					   allocated_size from the unnamed
+					   $DATA attribute and for compressed
+					   and/or sparse $DATA, this is the
+					   compressed_size from the unnamed
+					   $DATA attribute.  For a directory or
+					   other inode without an unnamed $DATA
+					   attribute, this is always 0.  NOTE:
+					   This is a multiple of the cluster
+					   size. */
+/* 30*/	sle64 data_size;		/* Byte size of actual data in unnamed
+					   data attribute.  For a directory or
+					   other inode without an unnamed $DATA
+					   attribute, this is always 0. */
+/* 38*/	FILE_ATTR_FLAGS file_attributes;	/* Flags describing the file. */
+/* 3c*/	union {
+	/* 3c*/	struct {
+		/* 3c*/	le16 packed_ea_size;	/* Size of the buffer needed to
+						   pack the extended attributes
+						   (EAs), if such are present.*/
+		/* 3e*/	le16 reserved;		/* Reserved for alignment. */
+		} __attribute__ ((__packed__)) ea;
+	/* 3c*/	struct {
+		/* 3c*/	le32 reparse_point_tag;	/* Type of reparse point,
+						   present only in reparse
+						   points and only if there are
+						   no EAs. */
+		} __attribute__ ((__packed__)) rp;
+	} __attribute__ ((__packed__)) type;
+/* 40*/	u8 file_name_length;			/* Length of file name in
+						   (Unicode) characters. */
+/* 41*/	FILE_NAME_TYPE_FLAGS file_name_type;	/* Namespace of the file name.*/
+/* 42*/	ntfschar file_name[0];			/* File name in Unicode. */
+} __attribute__ ((__packed__)) FILE_NAME_ATTR;
+
+/*
+ * GUID structures store globally unique identifiers (GUID). A GUID is a
+ * 128-bit value consisting of one group of eight hexadecimal digits, followed
+ * by three groups of four hexadecimal digits each, followed by one group of
+ * twelve hexadecimal digits. GUIDs are Microsoft's implementation of the
+ * distributed computing environment (DCE) universally unique identifier (UUID).
+ * Example of a GUID:
+ *	1F010768-5A73-BC91-0010A52216A7
+ */
+typedef struct {
+	le32 data1;	/* The first eight hexadecimal digits of the GUID. */
+	le16 data2;	/* The first group of four hexadecimal digits. */
+	le16 data3;	/* The second group of four hexadecimal digits. */
+	u8 data4[8];	/* The first two bytes are the third group of four
+			   hexadecimal digits. The remaining six bytes are the
+			   final 12 hexadecimal digits. */
+} __attribute__ ((__packed__)) GUID;
+
+/*
+ * FILE_Extend/$ObjId contains an index named $O. This index contains all
+ * object_ids present on the volume as the index keys and the corresponding
+ * mft_record numbers as the index entry data parts. The data part (defined
+ * below) also contains three other object_ids:
+ *	birth_volume_id - object_id of FILE_Volume on which the file was first
+ *			  created. Optional (i.e. can be zero).
+ *	birth_object_id - object_id of file when it was first created. Usually
+ *			  equals the object_id. Optional (i.e. can be zero).
+ *	domain_id	- Reserved (always zero).
+ */
+typedef struct {
+	leMFT_REF mft_reference;/* Mft record containing the object_id in
+				   the index entry key. */
+	union {
+		struct {
+			GUID birth_volume_id;
+			GUID birth_object_id;
+			GUID domain_id;
+		} __attribute__ ((__packed__)) origin;
+		u8 extended_info[48];
+	} __attribute__ ((__packed__)) opt;
+} __attribute__ ((__packed__)) OBJ_ID_INDEX_DATA;
+
+/*
+ * Attribute: Object id (NTFS 3.0+) (0x40).
+ *
+ * NOTE: Always resident.
+ */
+typedef struct {
+	GUID object_id;				/* Unique id assigned to the
+						   file.*/
+	/* The following fields are optional. The attribute value size is 16
+	   bytes, i.e. sizeof(GUID), if these are not present at all. Note,
+	   the entries can be present but one or more (or all) can be zero
+	   meaning that that particular value(s) is(are) not defined. */
+	union {
+		struct {
+			GUID birth_volume_id;	/* Unique id of volume on which
+						   the file was first created.*/
+			GUID birth_object_id;	/* Unique id of file when it was
+						   first created. */
+			GUID domain_id;		/* Reserved, zero. */
+		} __attribute__ ((__packed__)) origin;
+		u8 extended_info[48];
+	} __attribute__ ((__packed__)) opt;
+} __attribute__ ((__packed__)) OBJECT_ID_ATTR;
+
+/*
+ * The pre-defined IDENTIFIER_AUTHORITIES used as SID_IDENTIFIER_AUTHORITY in
+ * the SID structure (see below).
+ */
+//typedef enum {					/* SID string prefix. */
+//	SECURITY_NULL_SID_AUTHORITY	= {0, 0, 0, 0, 0, 0},	/* S-1-0 */
+//	SECURITY_WORLD_SID_AUTHORITY	= {0, 0, 0, 0, 0, 1},	/* S-1-1 */
+//	SECURITY_LOCAL_SID_AUTHORITY	= {0, 0, 0, 0, 0, 2},	/* S-1-2 */
+//	SECURITY_CREATOR_SID_AUTHORITY	= {0, 0, 0, 0, 0, 3},	/* S-1-3 */
+//	SECURITY_NON_UNIQUE_AUTHORITY	= {0, 0, 0, 0, 0, 4},	/* S-1-4 */
+//	SECURITY_NT_SID_AUTHORITY	= {0, 0, 0, 0, 0, 5},	/* S-1-5 */
+//} IDENTIFIER_AUTHORITIES;
+
+/*
+ * These relative identifiers (RIDs) are used with the above identifier
+ * authorities to make up universal well-known SIDs.
+ *
+ * Note: The relative identifier (RID) refers to the portion of a SID, which
+ * identifies a user or group in relation to the authority that issued the SID.
+ * For example, the universal well-known SID Creator Owner ID (S-1-3-0) is
+ * made up of the identifier authority SECURITY_CREATOR_SID_AUTHORITY (3) and
+ * the relative identifier SECURITY_CREATOR_OWNER_RID (0).
+ */
+typedef enum {					/* Identifier authority. */
+	SECURITY_NULL_RID		  = 0,	/* S-1-0 */
+	SECURITY_WORLD_RID		  = 0,	/* S-1-1 */
+	SECURITY_LOCAL_RID		  = 0,	/* S-1-2 */
+
+	SECURITY_CREATOR_OWNER_RID	  = 0,	/* S-1-3 */
+	SECURITY_CREATOR_GROUP_RID	  = 1,	/* S-1-3 */
+
+	SECURITY_CREATOR_OWNER_SERVER_RID = 2,	/* S-1-3 */
+	SECURITY_CREATOR_GROUP_SERVER_RID = 3,	/* S-1-3 */
+
+	SECURITY_DIALUP_RID		  = 1,
+	SECURITY_NETWORK_RID		  = 2,
+	SECURITY_BATCH_RID		  = 3,
+	SECURITY_INTERACTIVE_RID	  = 4,
+	SECURITY_SERVICE_RID		  = 6,
+	SECURITY_ANONYMOUS_LOGON_RID	  = 7,
+	SECURITY_PROXY_RID		  = 8,
+	SECURITY_ENTERPRISE_CONTROLLERS_RID=9,
+	SECURITY_SERVER_LOGON_RID	  = 9,
+	SECURITY_PRINCIPAL_SELF_RID	  = 0xa,
+	SECURITY_AUTHENTICATED_USER_RID	  = 0xb,
+	SECURITY_RESTRICTED_CODE_RID	  = 0xc,
+	SECURITY_TERMINAL_SERVER_RID	  = 0xd,
+
+	SECURITY_LOGON_IDS_RID		  = 5,
+	SECURITY_LOGON_IDS_RID_COUNT	  = 3,
+
+	SECURITY_LOCAL_SYSTEM_RID	  = 0x12,
+
+	SECURITY_NT_NON_UNIQUE		  = 0x15,
+
+	SECURITY_BUILTIN_DOMAIN_RID	  = 0x20,
+
+	/*
+	 * Well-known domain relative sub-authority values (RIDs).
+	 */
+
+	/* Users. */
+	DOMAIN_USER_RID_ADMIN		  = 0x1f4,
+	DOMAIN_USER_RID_GUEST		  = 0x1f5,
+	DOMAIN_USER_RID_KRBTGT		  = 0x1f6,
+
+	/* Groups. */
+	DOMAIN_GROUP_RID_ADMINS		  = 0x200,
+	DOMAIN_GROUP_RID_USERS		  = 0x201,
+	DOMAIN_GROUP_RID_GUESTS		  = 0x202,
+	DOMAIN_GROUP_RID_COMPUTERS	  = 0x203,
+	DOMAIN_GROUP_RID_CONTROLLERS	  = 0x204,
+	DOMAIN_GROUP_RID_CERT_ADMINS	  = 0x205,
+	DOMAIN_GROUP_RID_SCHEMA_ADMINS	  = 0x206,
+	DOMAIN_GROUP_RID_ENTERPRISE_ADMINS= 0x207,
+	DOMAIN_GROUP_RID_POLICY_ADMINS	  = 0x208,
+
+	/* Aliases. */
+	DOMAIN_ALIAS_RID_ADMINS		  = 0x220,
+	DOMAIN_ALIAS_RID_USERS		  = 0x221,
+	DOMAIN_ALIAS_RID_GUESTS		  = 0x222,
+	DOMAIN_ALIAS_RID_POWER_USERS	  = 0x223,
+
+	DOMAIN_ALIAS_RID_ACCOUNT_OPS	  = 0x224,
+	DOMAIN_ALIAS_RID_SYSTEM_OPS	  = 0x225,
+	DOMAIN_ALIAS_RID_PRINT_OPS	  = 0x226,
+	DOMAIN_ALIAS_RID_BACKUP_OPS	  = 0x227,
+
+	DOMAIN_ALIAS_RID_REPLICATOR	  = 0x228,
+	DOMAIN_ALIAS_RID_RAS_SERVERS	  = 0x229,
+	DOMAIN_ALIAS_RID_PREW2KCOMPACCESS = 0x22a,
+} RELATIVE_IDENTIFIERS;
+
+/*
+ * The universal well-known SIDs:
+ *
+ *	NULL_SID			S-1-0-0
+ *	WORLD_SID			S-1-1-0
+ *	LOCAL_SID			S-1-2-0
+ *	CREATOR_OWNER_SID		S-1-3-0
+ *	CREATOR_GROUP_SID		S-1-3-1
+ *	CREATOR_OWNER_SERVER_SID	S-1-3-2
+ *	CREATOR_GROUP_SERVER_SID	S-1-3-3
+ *
+ *	(Non-unique IDs)		S-1-4
+ *
+ * NT well-known SIDs:
+ *
+ *	NT_AUTHORITY_SID	S-1-5
+ *	DIALUP_SID		S-1-5-1
+ *
+ *	NETWORD_SID		S-1-5-2
+ *	BATCH_SID		S-1-5-3
+ *	INTERACTIVE_SID		S-1-5-4
+ *	SERVICE_SID		S-1-5-6
+ *	ANONYMOUS_LOGON_SID	S-1-5-7		(aka null logon session)
+ *	PROXY_SID		S-1-5-8
+ *	SERVER_LOGON_SID	S-1-5-9		(aka domain controller account)
+ *	SELF_SID		S-1-5-10	(self RID)
+ *	AUTHENTICATED_USER_SID	S-1-5-11
+ *	RESTRICTED_CODE_SID	S-1-5-12	(running restricted code)
+ *	TERMINAL_SERVER_SID	S-1-5-13	(running on terminal server)
+ *
+ *	(Logon IDs)		S-1-5-5-X-Y
+ *
+ *	(NT non-unique IDs)	S-1-5-0x15-...
+ *
+ *	(Built-in domain)	S-1-5-0x20
+ */
+
+/*
+ * The SID_IDENTIFIER_AUTHORITY is a 48-bit value used in the SID structure.
+ *
+ * NOTE: This is stored as a big endian number, hence the high_part comes
+ * before the low_part.
+ */
+typedef union {
+	struct {
+		u16 high_part;	/* High 16-bits. */
+		u32 low_part;	/* Low 32-bits. */
+	} __attribute__ ((__packed__)) parts;
+	u8 value[6];		/* Value as individual bytes. */
+} __attribute__ ((__packed__)) SID_IDENTIFIER_AUTHORITY;
+
+/*
+ * The SID structure is a variable-length structure used to uniquely identify
+ * users or groups. SID stands for security identifier.
+ *
+ * The standard textual representation of the SID is of the form:
+ *	S-R-I-S-S...
+ * Where:
+ *    - The first "S" is the literal character 'S' identifying the following
+ *	digits as a SID.
+ *    - R is the revision level of the SID expressed as a sequence of digits
+ *	either in decimal or hexadecimal (if the later, prefixed by "0x").
+ *    - I is the 48-bit identifier_authority, expressed as digits as R above.
+ *    - S... is one or more sub_authority values, expressed as digits as above.
+ *
+ * Example SID; the domain-relative SID of the local Administrators group on
+ * Windows NT/2k:
+ *	S-1-5-32-544
+ * This translates to a SID with:
+ *	revision = 1,
+ *	sub_authority_count = 2,
+ *	identifier_authority = {0,0,0,0,0,5},	// SECURITY_NT_AUTHORITY
+ *	sub_authority[0] = 32,			// SECURITY_BUILTIN_DOMAIN_RID
+ *	sub_authority[1] = 544			// DOMAIN_ALIAS_RID_ADMINS
+ */
+typedef struct {
+	u8 revision;
+	u8 sub_authority_count;
+	SID_IDENTIFIER_AUTHORITY identifier_authority;
+	le32 sub_authority[1];		/* At least one sub_authority. */
+} __attribute__ ((__packed__)) SID;
+
+/*
+ * Current constants for SIDs.
+ */
+typedef enum {
+	SID_REVISION			=  1,	/* Current revision level. */
+	SID_MAX_SUB_AUTHORITIES		= 15,	/* Maximum number of those. */
+	SID_RECOMMENDED_SUB_AUTHORITIES	=  1,	/* Will change to around 6 in
+						   a future revision. */
+} SID_CONSTANTS;
+
+/*
+ * The predefined ACE types (8-bit, see below).
+ */
+enum {
+	ACCESS_MIN_MS_ACE_TYPE		= 0,
+	ACCESS_ALLOWED_ACE_TYPE		= 0,
+	ACCESS_DENIED_ACE_TYPE		= 1,
+	SYSTEM_AUDIT_ACE_TYPE		= 2,
+	SYSTEM_ALARM_ACE_TYPE		= 3, /* Not implemented as of Win2k. */
+	ACCESS_MAX_MS_V2_ACE_TYPE	= 3,
+
+	ACCESS_ALLOWED_COMPOUND_ACE_TYPE= 4,
+	ACCESS_MAX_MS_V3_ACE_TYPE	= 4,
+
+	/* The following are Win2k only. */
+	ACCESS_MIN_MS_OBJECT_ACE_TYPE	= 5,
+	ACCESS_ALLOWED_OBJECT_ACE_TYPE	= 5,
+	ACCESS_DENIED_OBJECT_ACE_TYPE	= 6,
+	SYSTEM_AUDIT_OBJECT_ACE_TYPE	= 7,
+	SYSTEM_ALARM_OBJECT_ACE_TYPE	= 8,
+	ACCESS_MAX_MS_OBJECT_ACE_TYPE	= 8,
+
+	ACCESS_MAX_MS_V4_ACE_TYPE	= 8,
+
+	/* This one is for WinNT/2k. */
+	ACCESS_MAX_MS_ACE_TYPE		= 8,
+} __attribute__ ((__packed__));
+
+typedef u8 ACE_TYPES;
+
+/*
+ * The ACE flags (8-bit) for audit and inheritance (see below).
+ *
+ * SUCCESSFUL_ACCESS_ACE_FLAG is only used with system audit and alarm ACE
+ * types to indicate that a message is generated (in Windows!) for successful
+ * accesses.
+ *
+ * FAILED_ACCESS_ACE_FLAG is only used with system audit and alarm ACE types
+ * to indicate that a message is generated (in Windows!) for failed accesses.
+ */
+enum {
+	/* The inheritance flags. */
+	OBJECT_INHERIT_ACE		= 0x01,
+	CONTAINER_INHERIT_ACE		= 0x02,
+	NO_PROPAGATE_INHERIT_ACE	= 0x04,
+	INHERIT_ONLY_ACE		= 0x08,
+	INHERITED_ACE			= 0x10,	/* Win2k only. */
+	VALID_INHERIT_FLAGS		= 0x1f,
+
+	/* The audit flags. */
+	SUCCESSFUL_ACCESS_ACE_FLAG	= 0x40,
+	FAILED_ACCESS_ACE_FLAG		= 0x80,
+} __attribute__ ((__packed__));
+
+typedef u8 ACE_FLAGS;
+
+/*
+ * An ACE is an access-control entry in an access-control list (ACL).
+ * An ACE defines access to an object for a specific user or group or defines
+ * the types of access that generate system-administration messages or alarms
+ * for a specific user or group. The user or group is identified by a security
+ * identifier (SID).
+ *
+ * Each ACE starts with an ACE_HEADER structure (aligned on 4-byte boundary),
+ * which specifies the type and size of the ACE. The format of the subsequent
+ * data depends on the ACE type.
+ */
+typedef struct {
+/*Ofs*/
+/*  0*/	ACE_TYPES type;		/* Type of the ACE. */
+/*  1*/	ACE_FLAGS flags;	/* Flags describing the ACE. */
+/*  2*/	le16 size;		/* Size in bytes of the ACE. */
+} __attribute__ ((__packed__)) ACE_HEADER;
+
+/*
+ * The access mask (32-bit). Defines the access rights.
+ *
+ * The specific rights (bits 0 to 15).  These depend on the type of the object
+ * being secured by the ACE.
+ */
+enum {
+	/* Specific rights for files and directories are as follows: */
+
+	/* Right to read data from the file. (FILE) */
+	FILE_READ_DATA			= cpu_to_le32(0x00000001),
+	/* Right to list contents of a directory. (DIRECTORY) */
+	FILE_LIST_DIRECTORY		= cpu_to_le32(0x00000001),
+
+	/* Right to write data to the file. (FILE) */
+	FILE_WRITE_DATA			= cpu_to_le32(0x00000002),
+	/* Right to create a file in the directory. (DIRECTORY) */
+	FILE_ADD_FILE			= cpu_to_le32(0x00000002),
+
+	/* Right to append data to the file. (FILE) */
+	FILE_APPEND_DATA		= cpu_to_le32(0x00000004),
+	/* Right to create a subdirectory. (DIRECTORY) */
+	FILE_ADD_SUBDIRECTORY		= cpu_to_le32(0x00000004),
+
+	/* Right to read extended attributes. (FILE/DIRECTORY) */
+	FILE_READ_EA			= cpu_to_le32(0x00000008),
+
+	/* Right to write extended attributes. (FILE/DIRECTORY) */
+	FILE_WRITE_EA			= cpu_to_le32(0x00000010),
+
+	/* Right to execute a file. (FILE) */
+	FILE_EXECUTE			= cpu_to_le32(0x00000020),
+	/* Right to traverse the directory. (DIRECTORY) */
+	FILE_TRAVERSE			= cpu_to_le32(0x00000020),
+
+	/*
+	 * Right to delete a directory and all the files it contains (its
+	 * children), even if the files are read-only. (DIRECTORY)
+	 */
+	FILE_DELETE_CHILD		= cpu_to_le32(0x00000040),
+
+	/* Right to read file attributes. (FILE/DIRECTORY) */
+	FILE_READ_ATTRIBUTES		= cpu_to_le32(0x00000080),
+
+	/* Right to change file attributes. (FILE/DIRECTORY) */
+	FILE_WRITE_ATTRIBUTES		= cpu_to_le32(0x00000100),
+
+	/*
+	 * The standard rights (bits 16 to 23).  These are independent of the
+	 * type of object being secured.
+	 */
+
+	/* Right to delete the object. */
+	DELETE				= cpu_to_le32(0x00010000),
+
+	/*
+	 * Right to read the information in the object's security descriptor,
+	 * not including the information in the SACL, i.e. right to read the
+	 * security descriptor and owner.
+	 */
+	READ_CONTROL			= cpu_to_le32(0x00020000),
+
+	/* Right to modify the DACL in the object's security descriptor. */
+	WRITE_DAC			= cpu_to_le32(0x00040000),
+
+	/* Right to change the owner in the object's security descriptor. */
+	WRITE_OWNER			= cpu_to_le32(0x00080000),
+
+	/*
+	 * Right to use the object for synchronization.  Enables a process to
+	 * wait until the object is in the signalled state.  Some object types
+	 * do not support this access right.
+	 */
+	SYNCHRONIZE			= cpu_to_le32(0x00100000),
+
+	/*
+	 * The following STANDARD_RIGHTS_* are combinations of the above for
+	 * convenience and are defined by the Win32 API.
+	 */
+
+	/* These are currently defined to READ_CONTROL. */
+	STANDARD_RIGHTS_READ		= cpu_to_le32(0x00020000),
+	STANDARD_RIGHTS_WRITE		= cpu_to_le32(0x00020000),
+	STANDARD_RIGHTS_EXECUTE		= cpu_to_le32(0x00020000),
+
+	/* Combines DELETE, READ_CONTROL, WRITE_DAC, and WRITE_OWNER access. */
+	STANDARD_RIGHTS_REQUIRED	= cpu_to_le32(0x000f0000),
+
+	/*
+	 * Combines DELETE, READ_CONTROL, WRITE_DAC, WRITE_OWNER, and
+	 * SYNCHRONIZE access.
+	 */
+	STANDARD_RIGHTS_ALL		= cpu_to_le32(0x001f0000),
+
+	/*
+	 * The access system ACL and maximum allowed access types (bits 24 to
+	 * 25, bits 26 to 27 are reserved).
+	 */
+	ACCESS_SYSTEM_SECURITY		= cpu_to_le32(0x01000000),
+	MAXIMUM_ALLOWED			= cpu_to_le32(0x02000000),
+
+	/*
+	 * The generic rights (bits 28 to 31).  These map onto the standard and
+	 * specific rights.
+	 */
+
+	/* Read, write, and execute access. */
+	GENERIC_ALL			= cpu_to_le32(0x10000000),
+
+	/* Execute access. */
+	GENERIC_EXECUTE			= cpu_to_le32(0x20000000),
+
+	/*
+	 * Write access.  For files, this maps onto:
+	 *	FILE_APPEND_DATA | FILE_WRITE_ATTRIBUTES | FILE_WRITE_DATA |
+	 *	FILE_WRITE_EA | STANDARD_RIGHTS_WRITE | SYNCHRONIZE
+	 * For directories, the mapping has the same numerical value.  See
+	 * above for the descriptions of the rights granted.
+	 */
+	GENERIC_WRITE			= cpu_to_le32(0x40000000),
+
+	/*
+	 * Read access.  For files, this maps onto:
+	 *	FILE_READ_ATTRIBUTES | FILE_READ_DATA | FILE_READ_EA |
+	 *	STANDARD_RIGHTS_READ | SYNCHRONIZE
+	 * For directories, the mapping has the same numberical value.  See
+	 * above for the descriptions of the rights granted.
+	 */
+	GENERIC_READ			= cpu_to_le32(0x80000000),
+};
+
+typedef le32 ACCESS_MASK;
+
+/*
+ * The generic mapping array. Used to denote the mapping of each generic
+ * access right to a specific access mask.
+ *
+ * FIXME: What exactly is this and what is it for? (AIA)
+ */
+typedef struct {
+	ACCESS_MASK generic_read;
+	ACCESS_MASK generic_write;
+	ACCESS_MASK generic_execute;
+	ACCESS_MASK generic_all;
+} __attribute__ ((__packed__)) GENERIC_MAPPING;
+
+/*
+ * The predefined ACE type structures are as defined below.
+ */
+
+/*
+ * ACCESS_ALLOWED_ACE, ACCESS_DENIED_ACE, SYSTEM_AUDIT_ACE, SYSTEM_ALARM_ACE
+ */
+typedef struct {
+/*  0	ACE_HEADER; -- Unfolded here as gcc doesn't like unnamed structs. */
+	ACE_TYPES type;		/* Type of the ACE. */
+	ACE_FLAGS flags;	/* Flags describing the ACE. */
+	le16 size;		/* Size in bytes of the ACE. */
+/*  4*/	ACCESS_MASK mask;	/* Access mask associated with the ACE. */
+
+/*  8*/	SID sid;		/* The SID associated with the ACE. */
+} __attribute__ ((__packed__)) ACCESS_ALLOWED_ACE, ACCESS_DENIED_ACE,
+			       SYSTEM_AUDIT_ACE, SYSTEM_ALARM_ACE;
+
+/*
+ * The object ACE flags (32-bit).
+ */
+enum {
+	ACE_OBJECT_TYPE_PRESENT			= cpu_to_le32(1),
+	ACE_INHERITED_OBJECT_TYPE_PRESENT	= cpu_to_le32(2),
+};
+
+typedef le32 OBJECT_ACE_FLAGS;
+
+typedef struct {
+/*  0	ACE_HEADER; -- Unfolded here as gcc doesn't like unnamed structs. */
+	ACE_TYPES type;		/* Type of the ACE. */
+	ACE_FLAGS flags;	/* Flags describing the ACE. */
+	le16 size;		/* Size in bytes of the ACE. */
+/*  4*/	ACCESS_MASK mask;	/* Access mask associated with the ACE. */
+
+/*  8*/	OBJECT_ACE_FLAGS object_flags;	/* Flags describing the object ACE. */
+/* 12*/	GUID object_type;
+/* 28*/	GUID inherited_object_type;
+
+/* 44*/	SID sid;		/* The SID associated with the ACE. */
+} __attribute__ ((__packed__)) ACCESS_ALLOWED_OBJECT_ACE,
+			       ACCESS_DENIED_OBJECT_ACE,
+			       SYSTEM_AUDIT_OBJECT_ACE,
+			       SYSTEM_ALARM_OBJECT_ACE;
+
+/*
+ * An ACL is an access-control list (ACL).
+ * An ACL starts with an ACL header structure, which specifies the size of
+ * the ACL and the number of ACEs it contains. The ACL header is followed by
+ * zero or more access control entries (ACEs). The ACL as well as each ACE
+ * are aligned on 4-byte boundaries.
+ */
+typedef struct {
+	u8 revision;	/* Revision of this ACL. */
+	u8 alignment1;
+	le16 size;	/* Allocated space in bytes for ACL. Includes this
+			   header, the ACEs and the remaining free space. */
+	le16 ace_count;	/* Number of ACEs in the ACL. */
+	le16 alignment2;
+/* sizeof() = 8 bytes */
+} __attribute__ ((__packed__)) ACL;
+
+/*
+ * Current constants for ACLs.
+ */
+typedef enum {
+	/* Current revision. */
+	ACL_REVISION		= 2,
+	ACL_REVISION_DS		= 4,
+
+	/* History of revisions. */
+	ACL_REVISION1		= 1,
+	MIN_ACL_REVISION	= 2,
+	ACL_REVISION2		= 2,
+	ACL_REVISION3		= 3,
+	ACL_REVISION4		= 4,
+	MAX_ACL_REVISION	= 4,
+} ACL_CONSTANTS;
+
+/*
+ * The security descriptor control flags (16-bit).
+ *
+ * SE_OWNER_DEFAULTED - This boolean flag, when set, indicates that the SID
+ *	pointed to by the Owner field was provided by a defaulting mechanism
+ *	rather than explicitly provided by the original provider of the
+ *	security descriptor.  This may affect the treatment of the SID with
+ *	respect to inheritance of an owner.
+ *
+ * SE_GROUP_DEFAULTED - This boolean flag, when set, indicates that the SID in
+ *	the Group field was provided by a defaulting mechanism rather than
+ *	explicitly provided by the original provider of the security
+ *	descriptor.  This may affect the treatment of the SID with respect to
+ *	inheritance of a primary group.
+ *
+ * SE_DACL_PRESENT - This boolean flag, when set, indicates that the security
+ *	descriptor contains a discretionary ACL.  If this flag is set and the
+ *	Dacl field of the SECURITY_DESCRIPTOR is null, then a null ACL is
+ *	explicitly being specified.
+ *
+ * SE_DACL_DEFAULTED - This boolean flag, when set, indicates that the ACL
+ *	pointed to by the Dacl field was provided by a defaulting mechanism
+ *	rather than explicitly provided by the original provider of the
+ *	security descriptor.  This may affect the treatment of the ACL with
+ *	respect to inheritance of an ACL.  This flag is ignored if the
+ *	DaclPresent flag is not set.
+ *
+ * SE_SACL_PRESENT - This boolean flag, when set,  indicates that the security
+ *	descriptor contains a system ACL pointed to by the Sacl field.  If this
+ *	flag is set and the Sacl field of the SECURITY_DESCRIPTOR is null, then
+ *	an empty (but present) ACL is being specified.
+ *
+ * SE_SACL_DEFAULTED - This boolean flag, when set, indicates that the ACL
+ *	pointed to by the Sacl field was provided by a defaulting mechanism
+ *	rather than explicitly provided by the original provider of the
+ *	security descriptor.  This may affect the treatment of the ACL with
+ *	respect to inheritance of an ACL.  This flag is ignored if the
+ *	SaclPresent flag is not set.
+ *
+ * SE_SELF_RELATIVE - This boolean flag, when set, indicates that the security
+ *	descriptor is in self-relative form.  In this form, all fields of the
+ *	security descriptor are contiguous in memory and all pointer fields are
+ *	expressed as offsets from the beginning of the security descriptor.
+ */
+enum {
+	SE_OWNER_DEFAULTED		= cpu_to_le16(0x0001),
+	SE_GROUP_DEFAULTED		= cpu_to_le16(0x0002),
+	SE_DACL_PRESENT			= cpu_to_le16(0x0004),
+	SE_DACL_DEFAULTED		= cpu_to_le16(0x0008),
+
+	SE_SACL_PRESENT			= cpu_to_le16(0x0010),
+	SE_SACL_DEFAULTED		= cpu_to_le16(0x0020),
+
+	SE_DACL_AUTO_INHERIT_REQ	= cpu_to_le16(0x0100),
+	SE_SACL_AUTO_INHERIT_REQ	= cpu_to_le16(0x0200),
+	SE_DACL_AUTO_INHERITED		= cpu_to_le16(0x0400),
+	SE_SACL_AUTO_INHERITED		= cpu_to_le16(0x0800),
+
+	SE_DACL_PROTECTED		= cpu_to_le16(0x1000),
+	SE_SACL_PROTECTED		= cpu_to_le16(0x2000),
+	SE_RM_CONTROL_VALID		= cpu_to_le16(0x4000),
+	SE_SELF_RELATIVE		= cpu_to_le16(0x8000)
+} __attribute__ ((__packed__));
+
+typedef le16 SECURITY_DESCRIPTOR_CONTROL;
+
+/*
+ * Self-relative security descriptor. Contains the owner and group SIDs as well
+ * as the sacl and dacl ACLs inside the security descriptor itself.
+ */
+typedef struct {
+	u8 revision;	/* Revision level of the security descriptor. */
+	u8 alignment;
+	SECURITY_DESCRIPTOR_CONTROL control; /* Flags qualifying the type of
+			   the descriptor as well as the following fields. */
+	le32 owner;	/* Byte offset to a SID representing an object's
+			   owner. If this is NULL, no owner SID is present in
+			   the descriptor. */
+	le32 group;	/* Byte offset to a SID representing an object's
+			   primary group. If this is NULL, no primary group
+			   SID is present in the descriptor. */
+	le32 sacl;	/* Byte offset to a system ACL. Only valid, if
+			   SE_SACL_PRESENT is set in the control field. If
+			   SE_SACL_PRESENT is set but sacl is NULL, a NULL ACL
+			   is specified. */
+	le32 dacl;	/* Byte offset to a discretionary ACL. Only valid, if
+			   SE_DACL_PRESENT is set in the control field. If
+			   SE_DACL_PRESENT is set but dacl is NULL, a NULL ACL
+			   (unconditionally granting access) is specified. */
+/* sizeof() = 0x14 bytes */
+} __attribute__ ((__packed__)) SECURITY_DESCRIPTOR_RELATIVE;
+
+/*
+ * Absolute security descriptor. Does not contain the owner and group SIDs, nor
+ * the sacl and dacl ACLs inside the security descriptor. Instead, it contains
+ * pointers to these structures in memory. Obviously, absolute security
+ * descriptors are only useful for in memory representations of security
+ * descriptors. On disk, a self-relative security descriptor is used.
+ */
+typedef struct {
+	u8 revision;	/* Revision level of the security descriptor. */
+	u8 alignment;
+	SECURITY_DESCRIPTOR_CONTROL control;	/* Flags qualifying the type of
+			   the descriptor as well as the following fields. */
+	SID *owner;	/* Points to a SID representing an object's owner. If
+			   this is NULL, no owner SID is present in the
+			   descriptor. */
+	SID *group;	/* Points to a SID representing an object's primary
+			   group. If this is NULL, no primary group SID is
+			   present in the descriptor. */
+	ACL *sacl;	/* Points to a system ACL. Only valid, if
+			   SE_SACL_PRESENT is set in the control field. If
+			   SE_SACL_PRESENT is set but sacl is NULL, a NULL ACL
+			   is specified. */
+	ACL *dacl;	/* Points to a discretionary ACL. Only valid, if
+			   SE_DACL_PRESENT is set in the control field. If
+			   SE_DACL_PRESENT is set but dacl is NULL, a NULL ACL
+			   (unconditionally granting access) is specified. */
+} __attribute__ ((__packed__)) SECURITY_DESCRIPTOR;
+
+/*
+ * Current constants for security descriptors.
+ */
+typedef enum {
+	/* Current revision. */
+	SECURITY_DESCRIPTOR_REVISION	= 1,
+	SECURITY_DESCRIPTOR_REVISION1	= 1,
+
+	/* The sizes of both the absolute and relative security descriptors is
+	   the same as pointers, at least on ia32 architecture are 32-bit. */
+	SECURITY_DESCRIPTOR_MIN_LENGTH	= sizeof(SECURITY_DESCRIPTOR),
+} SECURITY_DESCRIPTOR_CONSTANTS;
+
+/*
+ * Attribute: Security descriptor (0x50). A standard self-relative security
+ * descriptor.
+ *
+ * NOTE: Can be resident or non-resident.
+ * NOTE: Not used in NTFS 3.0+, as security descriptors are stored centrally
+ * in FILE_Secure and the correct descriptor is found using the security_id
+ * from the standard information attribute.
+ */
+typedef SECURITY_DESCRIPTOR_RELATIVE SECURITY_DESCRIPTOR_ATTR;
+
+/*
+ * On NTFS 3.0+, all security descriptors are stored in FILE_Secure. Only one
+ * referenced instance of each unique security descriptor is stored.
+ *
+ * FILE_Secure contains no unnamed data attribute, i.e. it has zero length. It
+ * does, however, contain two indexes ($SDH and $SII) as well as a named data
+ * stream ($SDS).
+ *
+ * Every unique security descriptor is assigned a unique security identifier
+ * (security_id, not to be confused with a SID). The security_id is unique for
+ * the NTFS volume and is used as an index into the $SII index, which maps
+ * security_ids to the security descriptor's storage location within the $SDS
+ * data attribute. The $SII index is sorted by ascending security_id.
+ *
+ * A simple hash is computed from each security descriptor. This hash is used
+ * as an index into the $SDH index, which maps security descriptor hashes to
+ * the security descriptor's storage location within the $SDS data attribute.
+ * The $SDH index is sorted by security descriptor hash and is stored in a B+
+ * tree. When searching $SDH (with the intent of determining whether or not a
+ * new security descriptor is already present in the $SDS data stream), if a
+ * matching hash is found, but the security descriptors do not match, the
+ * search in the $SDH index is continued, searching for a next matching hash.
+ *
+ * When a precise match is found, the security_id coresponding to the security
+ * descriptor in the $SDS attribute is read from the found $SDH index entry and
+ * is stored in the $STANDARD_INFORMATION attribute of the file/directory to
+ * which the security descriptor is being applied. The $STANDARD_INFORMATION
+ * attribute is present in all base mft records (i.e. in all files and
+ * directories).
+ *
+ * If a match is not found, the security descriptor is assigned a new unique
+ * security_id and is added to the $SDS data attribute. Then, entries
+ * referencing the this security descriptor in the $SDS data attribute are
+ * added to the $SDH and $SII indexes.
+ *
+ * Note: Entries are never deleted from FILE_Secure, even if nothing
+ * references an entry any more.
+ */
+
+/*
+ * This header precedes each security descriptor in the $SDS data stream.
+ * This is also the index entry data part of both the $SII and $SDH indexes.
+ */
+typedef struct {
+	le32 hash;	  /* Hash of the security descriptor. */
+	le32 security_id; /* The security_id assigned to the descriptor. */
+	le64 offset;	  /* Byte offset of this entry in the $SDS stream. */
+	le32 length;	  /* Size in bytes of this entry in $SDS stream. */
+} __attribute__ ((__packed__)) SECURITY_DESCRIPTOR_HEADER;
+
+/*
+ * The $SDS data stream contains the security descriptors, aligned on 16-byte
+ * boundaries, sorted by security_id in a B+ tree. Security descriptors cannot
+ * cross 256kib boundaries (this restriction is imposed by the Windows cache
+ * manager). Each security descriptor is contained in a SDS_ENTRY structure.
+ * Also, each security descriptor is stored twice in the $SDS stream with a
+ * fixed offset of 0x40000 bytes (256kib, the Windows cache manager's max size)
+ * between them; i.e. if a SDS_ENTRY specifies an offset of 0x51d0, then the
+ * the first copy of the security descriptor will be at offset 0x51d0 in the
+ * $SDS data stream and the second copy will be at offset 0x451d0.
+ */
+typedef struct {
+/*Ofs*/
+/*  0	SECURITY_DESCRIPTOR_HEADER; -- Unfolded here as gcc doesn't like
+				       unnamed structs. */
+	le32 hash;	  /* Hash of the security descriptor. */
+	le32 security_id; /* The security_id assigned to the descriptor. */
+	le64 offset;	  /* Byte offset of this entry in the $SDS stream. */
+	le32 length;	  /* Size in bytes of this entry in $SDS stream. */
+/* 20*/	SECURITY_DESCRIPTOR_RELATIVE sid; /* The self-relative security
+					     descriptor. */
+} __attribute__ ((__packed__)) SDS_ENTRY;
+
+/*
+ * The index entry key used in the $SII index. The collation type is
+ * COLLATION_NTOFS_ULONG.
+ */
+typedef struct {
+	le32 security_id; /* The security_id assigned to the descriptor. */
+} __attribute__ ((__packed__)) SII_INDEX_KEY;
+
+/*
+ * The index entry key used in the $SDH index. The keys are sorted first by
+ * hash and then by security_id. The collation rule is
+ * COLLATION_NTOFS_SECURITY_HASH.
+ */
+typedef struct {
+	le32 hash;	  /* Hash of the security descriptor. */
+	le32 security_id; /* The security_id assigned to the descriptor. */
+} __attribute__ ((__packed__)) SDH_INDEX_KEY;
+
+/*
+ * Attribute: Volume name (0x60).
+ *
+ * NOTE: Always resident.
+ * NOTE: Present only in FILE_Volume.
+ */
+typedef struct {
+	ntfschar name[0];	/* The name of the volume in Unicode. */
+} __attribute__ ((__packed__)) VOLUME_NAME;
+
+/*
+ * Possible flags for the volume (16-bit).
+ */
+enum {
+	VOLUME_IS_DIRTY			= cpu_to_le16(0x0001),
+	VOLUME_RESIZE_LOG_FILE		= cpu_to_le16(0x0002),
+	VOLUME_UPGRADE_ON_MOUNT		= cpu_to_le16(0x0004),
+	VOLUME_MOUNTED_ON_NT4		= cpu_to_le16(0x0008),
+
+	VOLUME_DELETE_USN_UNDERWAY	= cpu_to_le16(0x0010),
+	VOLUME_REPAIR_OBJECT_ID		= cpu_to_le16(0x0020),
+
+	VOLUME_CHKDSK_UNDERWAY		= cpu_to_le16(0x4000),
+	VOLUME_MODIFIED_BY_CHKDSK	= cpu_to_le16(0x8000),
+
+	VOLUME_FLAGS_MASK		= cpu_to_le16(0xc03f),
+
+	/* To make our life easier when checking if we must mount read-only. */
+	VOLUME_MUST_MOUNT_RO_MASK	= cpu_to_le16(0xc027),
+} __attribute__ ((__packed__));
+
+typedef le16 VOLUME_FLAGS;
+
+/*
+ * Attribute: Volume information (0x70).
+ *
+ * NOTE: Always resident.
+ * NOTE: Present only in FILE_Volume.
+ * NOTE: Windows 2000 uses NTFS 3.0 while Windows NT4 service pack 6a uses
+ *	 NTFS 1.2. I haven't personally seen other values yet.
+ */
+typedef struct {
+	le64 reserved;		/* Not used (yet?). */
+	u8 major_ver;		/* Major version of the ntfs format. */
+	u8 minor_ver;		/* Minor version of the ntfs format. */
+	VOLUME_FLAGS flags;	/* Bit array of VOLUME_* flags. */
+} __attribute__ ((__packed__)) VOLUME_INFORMATION;
+
+/*
+ * Attribute: Data attribute (0x80).
+ *
+ * NOTE: Can be resident or non-resident.
+ *
+ * Data contents of a file (i.e. the unnamed stream) or of a named stream.
+ */
+typedef struct {
+	u8 data[0];		/* The file's data contents. */
+} __attribute__ ((__packed__)) DATA_ATTR;
+
+/*
+ * Index header flags (8-bit).
+ */
+enum {
+	/*
+	 * When index header is in an index root attribute:
+	 */
+	SMALL_INDEX = 0, /* The index is small enough to fit inside the index
+			    root attribute and there is no index allocation
+			    attribute present. */
+	LARGE_INDEX = 1, /* The index is too large to fit in the index root
+			    attribute and/or an index allocation attribute is
+			    present. */
+	/*
+	 * When index header is in an index block, i.e. is part of index
+	 * allocation attribute:
+	 */
+	LEAF_NODE  = 0, /* This is a leaf node, i.e. there are no more nodes
+			   branching off it. */
+	INDEX_NODE = 1, /* This node indexes other nodes, i.e. it is not a leaf
+			   node. */
+	NODE_MASK  = 1, /* Mask for accessing the *_NODE bits. */
+} __attribute__ ((__packed__));
+
+typedef u8 INDEX_HEADER_FLAGS;
+
+/*
+ * This is the header for indexes, describing the INDEX_ENTRY records, which
+ * follow the INDEX_HEADER. Together the index header and the index entries
+ * make up a complete index.
+ *
+ * IMPORTANT NOTE: The offset, length and size structure members are counted
+ * relative to the start of the index header structure and not relative to the
+ * start of the index root or index allocation structures themselves.
+ */
+typedef struct {
+	le32 entries_offset;		/* Byte offset to first INDEX_ENTRY
+					   aligned to 8-byte boundary. */
+	le32 index_length;		/* Data size of the index in bytes,
+					   i.e. bytes used from allocated
+					   size, aligned to 8-byte boundary. */
+	le32 allocated_size;		/* Byte size of this index (block),
+					   multiple of 8 bytes. */
+	/* NOTE: For the index root attribute, the above two numbers are always
+	   equal, as the attribute is resident and it is resized as needed. In
+	   the case of the index allocation attribute the attribute is not
+	   resident and hence the allocated_size is a fixed value and must
+	   equal the index_block_size specified by the INDEX_ROOT attribute
+	   corresponding to the INDEX_ALLOCATION attribute this INDEX_BLOCK
+	   belongs to. */
+	INDEX_HEADER_FLAGS flags;	/* Bit field of INDEX_HEADER_FLAGS. */
+	u8 reserved[3];			/* Reserved/align to 8-byte boundary. */
+} __attribute__ ((__packed__)) INDEX_HEADER;
+
+/*
+ * Attribute: Index root (0x90).
+ *
+ * NOTE: Always resident.
+ *
+ * This is followed by a sequence of index entries (INDEX_ENTRY structures)
+ * as described by the index header.
+ *
+ * When a directory is small enough to fit inside the index root then this
+ * is the only attribute describing the directory. When the directory is too
+ * large to fit in the index root, on the other hand, two additional attributes
+ * are present: an index allocation attribute, containing sub-nodes of the B+
+ * directory tree (see below), and a bitmap attribute, describing which virtual
+ * cluster numbers (vcns) in the index allocation attribute are in use by an
+ * index block.
+ *
+ * NOTE: The root directory (FILE_root) contains an entry for itself. Other
+ * directories do not contain entries for themselves, though.
+ */
+typedef struct {
+	ATTR_TYPE type;			/* Type of the indexed attribute. Is
+					   $FILE_NAME for directories, zero
+					   for view indexes. No other values
+					   allowed. */
+	COLLATION_RULE collation_rule;	/* Collation rule used to sort the
+					   index entries. If type is $FILE_NAME,
+					   this must be COLLATION_FILE_NAME. */
+	le32 index_block_size;		/* Size of each index block in bytes (in
+					   the index allocation attribute). */
+	u8 clusters_per_index_block;	/* Cluster size of each index block (in
+					   the index allocation attribute), when
+					   an index block is >= than a cluster,
+					   otherwise this will be the log of
+					   the size (like how the encoding of
+					   the mft record size and the index
+					   record size found in the boot sector
+					   work). Has to be a power of 2. */
+	u8 reserved[3];			/* Reserved/align to 8-byte boundary. */
+	INDEX_HEADER index;		/* Index header describing the
+					   following index entries. */
+} __attribute__ ((__packed__)) INDEX_ROOT;
+
+/*
+ * Attribute: Index allocation (0xa0).
+ *
+ * NOTE: Always non-resident (doesn't make sense to be resident anyway!).
+ *
+ * This is an array of index blocks. Each index block starts with an
+ * INDEX_BLOCK structure containing an index header, followed by a sequence of
+ * index entries (INDEX_ENTRY structures), as described by the INDEX_HEADER.
+ */
+typedef struct {
+/*  0	NTFS_RECORD; -- Unfolded here as gcc doesn't like unnamed structs. */
+	NTFS_RECORD_TYPE magic;	/* Magic is "INDX". */
+	le16 usa_ofs;		/* See NTFS_RECORD definition. */
+	le16 usa_count;		/* See NTFS_RECORD definition. */
+
+/*  8*/	sle64 lsn;		/* $LogFile sequence number of the last
+				   modification of this index block. */
+/* 16*/	leVCN index_block_vcn;	/* Virtual cluster number of the index block.
+				   If the cluster_size on the volume is <= the
+				   index_block_size of the directory,
+				   index_block_vcn counts in units of clusters,
+				   and in units of sectors otherwise. */
+/* 24*/	INDEX_HEADER index;	/* Describes the following index entries. */
+/* sizeof()= 40 (0x28) bytes */
+/*
+ * When creating the index block, we place the update sequence array at this
+ * offset, i.e. before we start with the index entries. This also makes sense,
+ * otherwise we could run into problems with the update sequence array
+ * containing in itself the last two bytes of a sector which would mean that
+ * multi sector transfer protection wouldn't work. As you can't protect data
+ * by overwriting it since you then can't get it back...
+ * When reading use the data from the ntfs record header.
+ */
+} __attribute__ ((__packed__)) INDEX_BLOCK;
+
+typedef INDEX_BLOCK INDEX_ALLOCATION;
+
+/*
+ * The system file FILE_Extend/$Reparse contains an index named $R listing
+ * all reparse points on the volume. The index entry keys are as defined
+ * below. Note, that there is no index data associated with the index entries.
+ *
+ * The index entries are sorted by the index key file_id. The collation rule is
+ * COLLATION_NTOFS_ULONGS. FIXME: Verify whether the reparse_tag is not the
+ * primary key / is not a key at all. (AIA)
+ */
+typedef struct {
+	le32 reparse_tag;	/* Reparse point type (inc. flags). */
+	leMFT_REF file_id;	/* Mft record of the file containing the
+				   reparse point attribute. */
+} __attribute__ ((__packed__)) REPARSE_INDEX_KEY;
+
+/*
+ * Quota flags (32-bit).
+ *
+ * The user quota flags.  Names explain meaning.
+ */
+enum {
+	QUOTA_FLAG_DEFAULT_LIMITS	= cpu_to_le32(0x00000001),
+	QUOTA_FLAG_LIMIT_REACHED	= cpu_to_le32(0x00000002),
+	QUOTA_FLAG_ID_DELETED		= cpu_to_le32(0x00000004),
+
+	QUOTA_FLAG_USER_MASK		= cpu_to_le32(0x00000007),
+	/* This is a bit mask for the user quota flags. */
+
+	/*
+	 * These flags are only present in the quota defaults index entry, i.e.
+	 * in the entry where owner_id = QUOTA_DEFAULTS_ID.
+	 */
+	QUOTA_FLAG_TRACKING_ENABLED	= cpu_to_le32(0x00000010),
+	QUOTA_FLAG_ENFORCEMENT_ENABLED	= cpu_to_le32(0x00000020),
+	QUOTA_FLAG_TRACKING_REQUESTED	= cpu_to_le32(0x00000040),
+	QUOTA_FLAG_LOG_THRESHOLD	= cpu_to_le32(0x00000080),
+
+	QUOTA_FLAG_LOG_LIMIT		= cpu_to_le32(0x00000100),
+	QUOTA_FLAG_OUT_OF_DATE		= cpu_to_le32(0x00000200),
+	QUOTA_FLAG_CORRUPT		= cpu_to_le32(0x00000400),
+	QUOTA_FLAG_PENDING_DELETES	= cpu_to_le32(0x00000800),
+};
+
+typedef le32 QUOTA_FLAGS;
+
+/*
+ * The system file FILE_Extend/$Quota contains two indexes $O and $Q. Quotas
+ * are on a per volume and per user basis.
+ *
+ * The $Q index contains one entry for each existing user_id on the volume. The
+ * index key is the user_id of the user/group owning this quota control entry,
+ * i.e. the key is the owner_id. The user_id of the owner of a file, i.e. the
+ * owner_id, is found in the standard information attribute. The collation rule
+ * for $Q is COLLATION_NTOFS_ULONG.
+ *
+ * The $O index contains one entry for each user/group who has been assigned
+ * a quota on that volume. The index key holds the SID of the user_id the
+ * entry belongs to, i.e. the owner_id. The collation rule for $O is
+ * COLLATION_NTOFS_SID.
+ *
+ * The $O index entry data is the user_id of the user corresponding to the SID.
+ * This user_id is used as an index into $Q to find the quota control entry
+ * associated with the SID.
+ *
+ * The $Q index entry data is the quota control entry and is defined below.
+ */
+typedef struct {
+	le32 version;		/* Currently equals 2. */
+	QUOTA_FLAGS flags;	/* Flags describing this quota entry. */
+	le64 bytes_used;	/* How many bytes of the quota are in use. */
+	sle64 change_time;	/* Last time this quota entry was changed. */
+	sle64 threshold;	/* Soft quota (-1 if not limited). */
+	sle64 limit;		/* Hard quota (-1 if not limited). */
+	sle64 exceeded_time;	/* How long the soft quota has been exceeded. */
+	SID sid;		/* The SID of the user/object associated with
+				   this quota entry.  Equals zero for the quota
+				   defaults entry (and in fact on a WinXP
+				   volume, it is not present at all). */
+} __attribute__ ((__packed__)) QUOTA_CONTROL_ENTRY;
+
+/*
+ * Predefined owner_id values (32-bit).
+ */
+enum {
+	QUOTA_INVALID_ID	= cpu_to_le32(0x00000000),
+	QUOTA_DEFAULTS_ID	= cpu_to_le32(0x00000001),
+	QUOTA_FIRST_USER_ID	= cpu_to_le32(0x00000100),
+};
+
+/*
+ * Current constants for quota control entries.
+ */
+typedef enum {
+	/* Current version. */
+	QUOTA_VERSION	= 2,
+} QUOTA_CONTROL_ENTRY_CONSTANTS;
+
+/*
+ * Index entry flags (16-bit).
+ */
+enum {
+	INDEX_ENTRY_NODE = cpu_to_le16(1), /* This entry contains a
+			sub-node, i.e. a reference to an index block in form of
+			a virtual cluster number (see below). */
+	INDEX_ENTRY_END  = cpu_to_le16(2), /* This signifies the last
+			entry in an index block.  The index entry does not
+			represent a file but it can point to a sub-node. */
+
+	INDEX_ENTRY_SPACE_FILLER = cpu_to_le16(0xffff), /* gcc: Force
+			enum bit width to 16-bit. */
+} __attribute__ ((__packed__));
+
+typedef le16 INDEX_ENTRY_FLAGS;
+
+/*
+ * This the index entry header (see below).
+ */
+typedef struct {
+/*  0*/	union {
+		struct { /* Only valid when INDEX_ENTRY_END is not set. */
+			leMFT_REF indexed_file;	/* The mft reference of the file
+						   described by this index
+						   entry. Used for directory
+						   indexes. */
+		} __attribute__ ((__packed__)) dir;
+		struct { /* Used for views/indexes to find the entry's data. */
+			le16 data_offset;	/* Data byte offset from this
+						   INDEX_ENTRY. Follows the
+						   index key. */
+			le16 data_length;	/* Data length in bytes. */
+			le32 reservedV;		/* Reserved (zero). */
+		} __attribute__ ((__packed__)) vi;
+	} __attribute__ ((__packed__)) data;
+/*  8*/	le16 length;		 /* Byte size of this index entry, multiple of
+				    8-bytes. */
+/* 10*/	le16 key_length;	 /* Byte size of the key value, which is in the
+				    index entry. It follows field reserved. Not
+				    multiple of 8-bytes. */
+/* 12*/	INDEX_ENTRY_FLAGS flags; /* Bit field of INDEX_ENTRY_* flags. */
+/* 14*/	le16 reserved;		 /* Reserved/align to 8-byte boundary. */
+/* sizeof() = 16 bytes */
+} __attribute__ ((__packed__)) INDEX_ENTRY_HEADER;
+
+/*
+ * This is an index entry. A sequence of such entries follows each INDEX_HEADER
+ * structure. Together they make up a complete index. The index follows either
+ * an index root attribute or an index allocation attribute.
+ *
+ * NOTE: Before NTFS 3.0 only filename attributes were indexed.
+ */
+typedef struct {
+/*Ofs*/
+/*  0	INDEX_ENTRY_HEADER; -- Unfolded here as gcc dislikes unnamed structs. */
+	union {
+		struct { /* Only valid when INDEX_ENTRY_END is not set. */
+			leMFT_REF indexed_file;	/* The mft reference of the file
+						   described by this index
+						   entry. Used for directory
+						   indexes. */
+		} __attribute__ ((__packed__)) dir;
+		struct { /* Used for views/indexes to find the entry's data. */
+			le16 data_offset;	/* Data byte offset from this
+						   INDEX_ENTRY. Follows the
+						   index key. */
+			le16 data_length;	/* Data length in bytes. */
+			le32 reservedV;		/* Reserved (zero). */
+		} __attribute__ ((__packed__)) vi;
+	} __attribute__ ((__packed__)) data;
+	le16 length;		 /* Byte size of this index entry, multiple of
+				    8-bytes. */
+	le16 key_length;	 /* Byte size of the key value, which is in the
+				    index entry. It follows field reserved. Not
+				    multiple of 8-bytes. */
+	INDEX_ENTRY_FLAGS flags; /* Bit field of INDEX_ENTRY_* flags. */
+	le16 reserved;		 /* Reserved/align to 8-byte boundary. */
+
+/* 16*/	union {		/* The key of the indexed attribute. NOTE: Only present
+			   if INDEX_ENTRY_END bit in flags is not set. NOTE: On
+			   NTFS versions before 3.0 the only valid key is the
+			   FILE_NAME_ATTR. On NTFS 3.0+ the following
+			   additional index keys are defined: */
+		FILE_NAME_ATTR file_name;/* $I30 index in directories. */
+		SII_INDEX_KEY sii;	/* $SII index in $Secure. */
+		SDH_INDEX_KEY sdh;	/* $SDH index in $Secure. */
+		GUID object_id;		/* $O index in FILE_Extend/$ObjId: The
+					   object_id of the mft record found in
+					   the data part of the index. */
+		REPARSE_INDEX_KEY reparse;	/* $R index in
+						   FILE_Extend/$Reparse. */
+		SID sid;		/* $O index in FILE_Extend/$Quota:
+					   SID of the owner of the user_id. */
+		le32 owner_id;		/* $Q index in FILE_Extend/$Quota:
+					   user_id of the owner of the quota
+					   control entry in the data part of
+					   the index. */
+	} __attribute__ ((__packed__)) key;
+	/* The (optional) index data is inserted here when creating. */
+	// leVCN vcn;	/* If INDEX_ENTRY_NODE bit in flags is set, the last
+	//		   eight bytes of this index entry contain the virtual
+	//		   cluster number of the index block that holds the
+	//		   entries immediately preceding the current entry (the
+	//		   vcn references the corresponding cluster in the data
+	//		   of the non-resident index allocation attribute). If
+	//		   the key_length is zero, then the vcn immediately
+	//		   follows the INDEX_ENTRY_HEADER. Regardless of
+	//		   key_length, the address of the 8-byte boundary
+	//		   aligned vcn of INDEX_ENTRY{_HEADER} *ie is given by
+	//		   (char*)ie + le16_to_cpu(ie*)->length) - sizeof(VCN),
+	//		   where sizeof(VCN) can be hardcoded as 8 if wanted. */
+} __attribute__ ((__packed__)) INDEX_ENTRY;
+
+/*
+ * Attribute: Bitmap (0xb0).
+ *
+ * Contains an array of bits (aka a bitfield).
+ *
+ * When used in conjunction with the index allocation attribute, each bit
+ * corresponds to one index block within the index allocation attribute. Thus
+ * the number of bits in the bitmap * index block size / cluster size is the
+ * number of clusters in the index allocation attribute.
+ */
+typedef struct {
+	u8 bitmap[0];			/* Array of bits. */
+} __attribute__ ((__packed__)) BITMAP_ATTR;
+
+/*
+ * The reparse point tag defines the type of the reparse point. It also
+ * includes several flags, which further describe the reparse point.
+ *
+ * The reparse point tag is an unsigned 32-bit value divided in three parts:
+ *
+ * 1. The least significant 16 bits (i.e. bits 0 to 15) specifiy the type of
+ *    the reparse point.
+ * 2. The 13 bits after this (i.e. bits 16 to 28) are reserved for future use.
+ * 3. The most significant three bits are flags describing the reparse point.
+ *    They are defined as follows:
+ *	bit 29: Name surrogate bit. If set, the filename is an alias for
+ *		another object in the system.
+ *	bit 30: High-latency bit. If set, accessing the first byte of data will
+ *		be slow. (E.g. the data is stored on a tape drive.)
+ *	bit 31: Microsoft bit. If set, the tag is owned by Microsoft. User
+ *		defined tags have to use zero here.
+ *
+ * These are the predefined reparse point tags:
+ */
+enum {
+	IO_REPARSE_TAG_IS_ALIAS		= cpu_to_le32(0x20000000),
+	IO_REPARSE_TAG_IS_HIGH_LATENCY	= cpu_to_le32(0x40000000),
+	IO_REPARSE_TAG_IS_MICROSOFT	= cpu_to_le32(0x80000000),
+
+	IO_REPARSE_TAG_RESERVED_ZERO	= cpu_to_le32(0x00000000),
+	IO_REPARSE_TAG_RESERVED_ONE	= cpu_to_le32(0x00000001),
+	IO_REPARSE_TAG_RESERVED_RANGE	= cpu_to_le32(0x00000001),
+
+	IO_REPARSE_TAG_NSS		= cpu_to_le32(0x68000005),
+	IO_REPARSE_TAG_NSS_RECOVER	= cpu_to_le32(0x68000006),
+	IO_REPARSE_TAG_SIS		= cpu_to_le32(0x68000007),
+	IO_REPARSE_TAG_DFS		= cpu_to_le32(0x68000008),
+
+	IO_REPARSE_TAG_MOUNT_POINT	= cpu_to_le32(0x88000003),
+
+	IO_REPARSE_TAG_HSM		= cpu_to_le32(0xa8000004),
+
+	IO_REPARSE_TAG_SYMBOLIC_LINK	= cpu_to_le32(0xe8000000),
+
+	IO_REPARSE_TAG_VALID_VALUES	= cpu_to_le32(0xe000ffff),
+};
+
+/*
+ * Attribute: Reparse point (0xc0).
+ *
+ * NOTE: Can be resident or non-resident.
+ */
+typedef struct {
+	le32 reparse_tag;		/* Reparse point type (inc. flags). */
+	le16 reparse_data_length;	/* Byte size of reparse data. */
+	le16 reserved;			/* Align to 8-byte boundary. */
+	u8 reparse_data[0];		/* Meaning depends on reparse_tag. */
+} __attribute__ ((__packed__)) REPARSE_POINT;
+
+/*
+ * Attribute: Extended attribute (EA) information (0xd0).
+ *
+ * NOTE: Always resident. (Is this true???)
+ */
+typedef struct {
+	le16 ea_length;		/* Byte size of the packed extended
+				   attributes. */
+	le16 need_ea_count;	/* The number of extended attributes which have
+				   the NEED_EA bit set. */
+	le32 ea_query_length;	/* Byte size of the buffer required to query
+				   the extended attributes when calling
+				   ZwQueryEaFile() in Windows NT/2k. I.e. the
+				   byte size of the unpacked extended
+				   attributes. */
+} __attribute__ ((__packed__)) EA_INFORMATION;
+
+/*
+ * Extended attribute flags (8-bit).
+ */
+enum {
+	NEED_EA	= 0x80		/* If set the file to which the EA belongs
+				   cannot be interpreted without understanding
+				   the associates extended attributes. */
+} __attribute__ ((__packed__));
+
+typedef u8 EA_FLAGS;
+
+/*
+ * Attribute: Extended attribute (EA) (0xe0).
+ *
+ * NOTE: Can be resident or non-resident.
+ *
+ * Like the attribute list and the index buffer list, the EA attribute value is
+ * a sequence of EA_ATTR variable length records.
+ */
+typedef struct {
+	le32 next_entry_offset;	/* Offset to the next EA_ATTR. */
+	EA_FLAGS flags;		/* Flags describing the EA. */
+	u8 ea_name_length;	/* Length of the name of the EA in bytes
+				   excluding the '\0' byte terminator. */
+	le16 ea_value_length;	/* Byte size of the EA's value. */
+	u8 ea_name[0];		/* Name of the EA.  Note this is ASCII, not
+				   Unicode and it is zero terminated. */
+	u8 ea_value[0];		/* The value of the EA.  Immediately follows
+				   the name. */
+} __attribute__ ((__packed__)) EA_ATTR;
+
+/*
+ * Attribute: Property set (0xf0).
+ *
+ * Intended to support Native Structure Storage (NSS) - a feature removed from
+ * NTFS 3.0 during beta testing.
+ */
+typedef struct {
+	/* Irrelevant as feature unused. */
+} __attribute__ ((__packed__)) PROPERTY_SET;
+
+/*
+ * Attribute: Logged utility stream (0x100).
+ *
+ * NOTE: Can be resident or non-resident.
+ *
+ * Operations on this attribute are logged to the journal ($LogFile) like
+ * normal metadata changes.
+ *
+ * Used by the Encrypting File System (EFS). All encrypted files have this
+ * attribute with the name $EFS.
+ */
+typedef struct {
+	/* Can be anything the creator chooses. */
+	/* EFS uses it as follows: */
+	// FIXME: Type this info, verifying it along the way. (AIA)
+} __attribute__ ((__packed__)) LOGGED_UTILITY_STREAM, EFS_ATTR;
+
+#endif /* _LINUX_NTFS_LAYOUT_H */
diff --git a/fs/ntfs/lcnalloc.c b/fs/ntfs/lcnalloc.c
new file mode 100644
index 000000000..eda9972e6
--- /dev/null
+++ b/fs/ntfs/lcnalloc.c
@@ -0,0 +1,1000 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * lcnalloc.c - Cluster (de)allocation code.  Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2004-2005 Anton Altaparmakov
+ */
+
+#ifdef NTFS_RW
+
+#include <linux/pagemap.h>
+
+#include "lcnalloc.h"
+#include "debug.h"
+#include "bitmap.h"
+#include "inode.h"
+#include "volume.h"
+#include "attrib.h"
+#include "malloc.h"
+#include "aops.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_cluster_free_from_rl_nolock - free clusters from runlist
+ * @vol:	mounted ntfs volume on which to free the clusters
+ * @rl:		runlist describing the clusters to free
+ *
+ * Free all the clusters described by the runlist @rl on the volume @vol.  In
+ * the case of an error being returned, at least some of the clusters were not
+ * freed.
+ *
+ * Return 0 on success and -errno on error.
+ *
+ * Locking: - The volume lcn bitmap must be locked for writing on entry and is
+ *	      left locked on return.
+ */
+int ntfs_cluster_free_from_rl_nolock(ntfs_volume *vol,
+		const runlist_element *rl)
+{
+	struct inode *lcnbmp_vi = vol->lcnbmp_ino;
+	int ret = 0;
+
+	ntfs_debug("Entering.");
+	if (!rl)
+		return 0;
+	for (; rl->length; rl++) {
+		int err;
+
+		if (rl->lcn < 0)
+			continue;
+		err = ntfs_bitmap_clear_run(lcnbmp_vi, rl->lcn, rl->length);
+		if (unlikely(err && (!ret || ret == -ENOMEM) && ret != err))
+			ret = err;
+	}
+	ntfs_debug("Done.");
+	return ret;
+}
+
+/**
+ * ntfs_cluster_alloc - allocate clusters on an ntfs volume
+ * @vol:	mounted ntfs volume on which to allocate the clusters
+ * @start_vcn:	vcn to use for the first allocated cluster
+ * @count:	number of clusters to allocate
+ * @start_lcn:	starting lcn at which to allocate the clusters (or -1 if none)
+ * @zone:	zone from which to allocate the clusters
+ * @is_extension:	if 'true', this is an attribute extension
+ *
+ * Allocate @count clusters preferably starting at cluster @start_lcn or at the
+ * current allocator position if @start_lcn is -1, on the mounted ntfs volume
+ * @vol. @zone is either DATA_ZONE for allocation of normal clusters or
+ * MFT_ZONE for allocation of clusters for the master file table, i.e. the
+ * $MFT/$DATA attribute.
+ *
+ * @start_vcn specifies the vcn of the first allocated cluster.  This makes
+ * merging the resulting runlist with the old runlist easier.
+ *
+ * If @is_extension is 'true', the caller is allocating clusters to extend an
+ * attribute and if it is 'false', the caller is allocating clusters to fill a
+ * hole in an attribute.  Practically the difference is that if @is_extension
+ * is 'true' the returned runlist will be terminated with LCN_ENOENT and if
+ * @is_extension is 'false' the runlist will be terminated with
+ * LCN_RL_NOT_MAPPED.
+ *
+ * You need to check the return value with IS_ERR().  If this is false, the
+ * function was successful and the return value is a runlist describing the
+ * allocated cluster(s).  If IS_ERR() is true, the function failed and
+ * PTR_ERR() gives you the error code.
+ *
+ * Notes on the allocation algorithm
+ * =================================
+ *
+ * There are two data zones.  First is the area between the end of the mft zone
+ * and the end of the volume, and second is the area between the start of the
+ * volume and the start of the mft zone.  On unmodified/standard NTFS 1.x
+ * volumes, the second data zone does not exist due to the mft zone being
+ * expanded to cover the start of the volume in order to reserve space for the
+ * mft bitmap attribute.
+ *
+ * This is not the prettiest function but the complexity stems from the need of
+ * implementing the mft vs data zoned approach and from the fact that we have
+ * access to the lcn bitmap in portions of up to 8192 bytes at a time, so we
+ * need to cope with crossing over boundaries of two buffers.  Further, the
+ * fact that the allocator allows for caller supplied hints as to the location
+ * of where allocation should begin and the fact that the allocator keeps track
+ * of where in the data zones the next natural allocation should occur,
+ * contribute to the complexity of the function.  But it should all be
+ * worthwhile, because this allocator should: 1) be a full implementation of
+ * the MFT zone approach used by Windows NT, 2) cause reduction in
+ * fragmentation, and 3) be speedy in allocations (the code is not optimized
+ * for speed, but the algorithm is, so further speed improvements are probably
+ * possible).
+ *
+ * FIXME: We should be monitoring cluster allocation and increment the MFT zone
+ * size dynamically but this is something for the future.  We will just cause
+ * heavier fragmentation by not doing it and I am not even sure Windows would
+ * grow the MFT zone dynamically, so it might even be correct not to do this.
+ * The overhead in doing dynamic MFT zone expansion would be very large and
+ * unlikely worth the effort. (AIA)
+ *
+ * TODO: I have added in double the required zone position pointer wrap around
+ * logic which can be optimized to having only one of the two logic sets.
+ * However, having the double logic will work fine, but if we have only one of
+ * the sets and we get it wrong somewhere, then we get into trouble, so
+ * removing the duplicate logic requires _very_ careful consideration of _all_
+ * possible code paths.  So at least for now, I am leaving the double logic -
+ * better safe than sorry... (AIA)
+ *
+ * Locking: - The volume lcn bitmap must be unlocked on entry and is unlocked
+ *	      on return.
+ *	    - This function takes the volume lcn bitmap lock for writing and
+ *	      modifies the bitmap contents.
+ */
+runlist_element *ntfs_cluster_alloc(ntfs_volume *vol, const VCN start_vcn,
+		const s64 count, const LCN start_lcn,
+		const NTFS_CLUSTER_ALLOCATION_ZONES zone,
+		const bool is_extension)
+{
+	LCN zone_start, zone_end, bmp_pos, bmp_initial_pos, last_read_pos, lcn;
+	LCN prev_lcn = 0, prev_run_len = 0, mft_zone_size;
+	s64 clusters;
+	loff_t i_size;
+	struct inode *lcnbmp_vi;
+	runlist_element *rl = NULL;
+	struct address_space *mapping;
+	struct page *page = NULL;
+	u8 *buf, *byte;
+	int err = 0, rlpos, rlsize, buf_size;
+	u8 pass, done_zones, search_zone, need_writeback = 0, bit;
+
+	ntfs_debug("Entering for start_vcn 0x%llx, count 0x%llx, start_lcn "
+			"0x%llx, zone %s_ZONE.", (unsigned long long)start_vcn,
+			(unsigned long long)count,
+			(unsigned long long)start_lcn,
+			zone == MFT_ZONE ? "MFT" : "DATA");
+	BUG_ON(!vol);
+	lcnbmp_vi = vol->lcnbmp_ino;
+	BUG_ON(!lcnbmp_vi);
+	BUG_ON(start_vcn < 0);
+	BUG_ON(count < 0);
+	BUG_ON(start_lcn < -1);
+	BUG_ON(zone < FIRST_ZONE);
+	BUG_ON(zone > LAST_ZONE);
+
+	/* Return NULL if @count is zero. */
+	if (!count)
+		return NULL;
+	/* Take the lcnbmp lock for writing. */
+	down_write(&vol->lcnbmp_lock);
+	/*
+	 * If no specific @start_lcn was requested, use the current data zone
+	 * position, otherwise use the requested @start_lcn but make sure it
+	 * lies outside the mft zone.  Also set done_zones to 0 (no zones done)
+	 * and pass depending on whether we are starting inside a zone (1) or
+	 * at the beginning of a zone (2).  If requesting from the MFT_ZONE,
+	 * we either start at the current position within the mft zone or at
+	 * the specified position.  If the latter is out of bounds then we start
+	 * at the beginning of the MFT_ZONE.
+	 */
+	done_zones = 0;
+	pass = 1;
+	/*
+	 * zone_start and zone_end are the current search range.  search_zone
+	 * is 1 for mft zone, 2 for data zone 1 (end of mft zone till end of
+	 * volume) and 4 for data zone 2 (start of volume till start of mft
+	 * zone).
+	 */
+	zone_start = start_lcn;
+	if (zone_start < 0) {
+		if (zone == DATA_ZONE)
+			zone_start = vol->data1_zone_pos;
+		else
+			zone_start = vol->mft_zone_pos;
+		if (!zone_start) {
+			/*
+			 * Zone starts at beginning of volume which means a
+			 * single pass is sufficient.
+			 */
+			pass = 2;
+		}
+	} else if (zone == DATA_ZONE && zone_start >= vol->mft_zone_start &&
+			zone_start < vol->mft_zone_end) {
+		zone_start = vol->mft_zone_end;
+		/*
+		 * Starting at beginning of data1_zone which means a single
+		 * pass in this zone is sufficient.
+		 */
+		pass = 2;
+	} else if (zone == MFT_ZONE && (zone_start < vol->mft_zone_start ||
+			zone_start >= vol->mft_zone_end)) {
+		zone_start = vol->mft_lcn;
+		if (!vol->mft_zone_end)
+			zone_start = 0;
+		/*
+		 * Starting at beginning of volume which means a single pass
+		 * is sufficient.
+		 */
+		pass = 2;
+	}
+	if (zone == MFT_ZONE) {
+		zone_end = vol->mft_zone_end;
+		search_zone = 1;
+	} else /* if (zone == DATA_ZONE) */ {
+		/* Skip searching the mft zone. */
+		done_zones |= 1;
+		if (zone_start >= vol->mft_zone_end) {
+			zone_end = vol->nr_clusters;
+			search_zone = 2;
+		} else {
+			zone_end = vol->mft_zone_start;
+			search_zone = 4;
+		}
+	}
+	/*
+	 * bmp_pos is the current bit position inside the bitmap.  We use
+	 * bmp_initial_pos to determine whether or not to do a zone switch.
+	 */
+	bmp_pos = bmp_initial_pos = zone_start;
+
+	/* Loop until all clusters are allocated, i.e. clusters == 0. */
+	clusters = count;
+	rlpos = rlsize = 0;
+	mapping = lcnbmp_vi->i_mapping;
+	i_size = i_size_read(lcnbmp_vi);
+	while (1) {
+		ntfs_debug("Start of outer while loop: done_zones 0x%x, "
+				"search_zone %i, pass %i, zone_start 0x%llx, "
+				"zone_end 0x%llx, bmp_initial_pos 0x%llx, "
+				"bmp_pos 0x%llx, rlpos %i, rlsize %i.",
+				done_zones, search_zone, pass,
+				(unsigned long long)zone_start,
+				(unsigned long long)zone_end,
+				(unsigned long long)bmp_initial_pos,
+				(unsigned long long)bmp_pos, rlpos, rlsize);
+		/* Loop until we run out of free clusters. */
+		last_read_pos = bmp_pos >> 3;
+		ntfs_debug("last_read_pos 0x%llx.",
+				(unsigned long long)last_read_pos);
+		if (last_read_pos > i_size) {
+			ntfs_debug("End of attribute reached.  "
+					"Skipping to zone_pass_done.");
+			goto zone_pass_done;
+		}
+		if (likely(page)) {
+			if (need_writeback) {
+				ntfs_debug("Marking page dirty.");
+				flush_dcache_page(page);
+				set_page_dirty(page);
+				need_writeback = 0;
+			}
+			ntfs_unmap_page(page);
+		}
+		page = ntfs_map_page(mapping, last_read_pos >>
+				PAGE_SHIFT);
+		if (IS_ERR(page)) {
+			err = PTR_ERR(page);
+			ntfs_error(vol->sb, "Failed to map page.");
+			goto out;
+		}
+		buf_size = last_read_pos & ~PAGE_MASK;
+		buf = page_address(page) + buf_size;
+		buf_size = PAGE_SIZE - buf_size;
+		if (unlikely(last_read_pos + buf_size > i_size))
+			buf_size = i_size - last_read_pos;
+		buf_size <<= 3;
+		lcn = bmp_pos & 7;
+		bmp_pos &= ~(LCN)7;
+		ntfs_debug("Before inner while loop: buf_size %i, lcn 0x%llx, "
+				"bmp_pos 0x%llx, need_writeback %i.", buf_size,
+				(unsigned long long)lcn,
+				(unsigned long long)bmp_pos, need_writeback);
+		while (lcn < buf_size && lcn + bmp_pos < zone_end) {
+			byte = buf + (lcn >> 3);
+			ntfs_debug("In inner while loop: buf_size %i, "
+					"lcn 0x%llx, bmp_pos 0x%llx, "
+					"need_writeback %i, byte ofs 0x%x, "
+					"*byte 0x%x.", buf_size,
+					(unsigned long long)lcn,
+					(unsigned long long)bmp_pos,
+					need_writeback,
+					(unsigned int)(lcn >> 3),
+					(unsigned int)*byte);
+			/* Skip full bytes. */
+			if (*byte == 0xff) {
+				lcn = (lcn + 8) & ~(LCN)7;
+				ntfs_debug("Continuing while loop 1.");
+				continue;
+			}
+			bit = 1 << (lcn & 7);
+			ntfs_debug("bit 0x%x.", bit);
+			/* If the bit is already set, go onto the next one. */
+			if (*byte & bit) {
+				lcn++;
+				ntfs_debug("Continuing while loop 2.");
+				continue;
+			}
+			/*
+			 * Allocate more memory if needed, including space for
+			 * the terminator element.
+			 * ntfs_malloc_nofs() operates on whole pages only.
+			 */
+			if ((rlpos + 2) * sizeof(*rl) > rlsize) {
+				runlist_element *rl2;
+
+				ntfs_debug("Reallocating memory.");
+				if (!rl)
+					ntfs_debug("First free bit is at LCN "
+							"0x%llx.",
+							(unsigned long long)
+							(lcn + bmp_pos));
+				rl2 = ntfs_malloc_nofs(rlsize + (int)PAGE_SIZE);
+				if (unlikely(!rl2)) {
+					err = -ENOMEM;
+					ntfs_error(vol->sb, "Failed to "
+							"allocate memory.");
+					goto out;
+				}
+				memcpy(rl2, rl, rlsize);
+				ntfs_free(rl);
+				rl = rl2;
+				rlsize += PAGE_SIZE;
+				ntfs_debug("Reallocated memory, rlsize 0x%x.",
+						rlsize);
+			}
+			/* Allocate the bitmap bit. */
+			*byte |= bit;
+			/* We need to write this bitmap page to disk. */
+			need_writeback = 1;
+			ntfs_debug("*byte 0x%x, need_writeback is set.",
+					(unsigned int)*byte);
+			/*
+			 * Coalesce with previous run if adjacent LCNs.
+			 * Otherwise, append a new run.
+			 */
+			ntfs_debug("Adding run (lcn 0x%llx, len 0x%llx), "
+					"prev_lcn 0x%llx, lcn 0x%llx, "
+					"bmp_pos 0x%llx, prev_run_len 0x%llx, "
+					"rlpos %i.",
+					(unsigned long long)(lcn + bmp_pos),
+					1ULL, (unsigned long long)prev_lcn,
+					(unsigned long long)lcn,
+					(unsigned long long)bmp_pos,
+					(unsigned long long)prev_run_len,
+					rlpos);
+			if (prev_lcn == lcn + bmp_pos - prev_run_len && rlpos) {
+				ntfs_debug("Coalescing to run (lcn 0x%llx, "
+						"len 0x%llx).",
+						(unsigned long long)
+						rl[rlpos - 1].lcn,
+						(unsigned long long)
+						rl[rlpos - 1].length);
+				rl[rlpos - 1].length = ++prev_run_len;
+				ntfs_debug("Run now (lcn 0x%llx, len 0x%llx), "
+						"prev_run_len 0x%llx.",
+						(unsigned long long)
+						rl[rlpos - 1].lcn,
+						(unsigned long long)
+						rl[rlpos - 1].length,
+						(unsigned long long)
+						prev_run_len);
+			} else {
+				if (likely(rlpos)) {
+					ntfs_debug("Adding new run, (previous "
+							"run lcn 0x%llx, "
+							"len 0x%llx).",
+							(unsigned long long)
+							rl[rlpos - 1].lcn,
+							(unsigned long long)
+							rl[rlpos - 1].length);
+					rl[rlpos].vcn = rl[rlpos - 1].vcn +
+							prev_run_len;
+				} else {
+					ntfs_debug("Adding new run, is first "
+							"run.");
+					rl[rlpos].vcn = start_vcn;
+				}
+				rl[rlpos].lcn = prev_lcn = lcn + bmp_pos;
+				rl[rlpos].length = prev_run_len = 1;
+				rlpos++;
+			}
+			/* Done? */
+			if (!--clusters) {
+				LCN tc;
+				/*
+				 * Update the current zone position.  Positions
+				 * of already scanned zones have been updated
+				 * during the respective zone switches.
+				 */
+				tc = lcn + bmp_pos + 1;
+				ntfs_debug("Done. Updating current zone "
+						"position, tc 0x%llx, "
+						"search_zone %i.",
+						(unsigned long long)tc,
+						search_zone);
+				switch (search_zone) {
+				case 1:
+					ntfs_debug("Before checks, "
+							"vol->mft_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->mft_zone_pos);
+					if (tc >= vol->mft_zone_end) {
+						vol->mft_zone_pos =
+								vol->mft_lcn;
+						if (!vol->mft_zone_end)
+							vol->mft_zone_pos = 0;
+					} else if ((bmp_initial_pos >=
+							vol->mft_zone_pos ||
+							tc > vol->mft_zone_pos)
+							&& tc >= vol->mft_lcn)
+						vol->mft_zone_pos = tc;
+					ntfs_debug("After checks, "
+							"vol->mft_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->mft_zone_pos);
+					break;
+				case 2:
+					ntfs_debug("Before checks, "
+							"vol->data1_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->data1_zone_pos);
+					if (tc >= vol->nr_clusters)
+						vol->data1_zone_pos =
+							     vol->mft_zone_end;
+					else if ((bmp_initial_pos >=
+						    vol->data1_zone_pos ||
+						    tc > vol->data1_zone_pos)
+						    && tc >= vol->mft_zone_end)
+						vol->data1_zone_pos = tc;
+					ntfs_debug("After checks, "
+							"vol->data1_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->data1_zone_pos);
+					break;
+				case 4:
+					ntfs_debug("Before checks, "
+							"vol->data2_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->data2_zone_pos);
+					if (tc >= vol->mft_zone_start)
+						vol->data2_zone_pos = 0;
+					else if (bmp_initial_pos >=
+						      vol->data2_zone_pos ||
+						      tc > vol->data2_zone_pos)
+						vol->data2_zone_pos = tc;
+					ntfs_debug("After checks, "
+							"vol->data2_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->data2_zone_pos);
+					break;
+				default:
+					BUG();
+				}
+				ntfs_debug("Finished.  Going to out.");
+				goto out;
+			}
+			lcn++;
+		}
+		bmp_pos += buf_size;
+		ntfs_debug("After inner while loop: buf_size 0x%x, lcn "
+				"0x%llx, bmp_pos 0x%llx, need_writeback %i.",
+				buf_size, (unsigned long long)lcn,
+				(unsigned long long)bmp_pos, need_writeback);
+		if (bmp_pos < zone_end) {
+			ntfs_debug("Continuing outer while loop, "
+					"bmp_pos 0x%llx, zone_end 0x%llx.",
+					(unsigned long long)bmp_pos,
+					(unsigned long long)zone_end);
+			continue;
+		}
+zone_pass_done:	/* Finished with the current zone pass. */
+		ntfs_debug("At zone_pass_done, pass %i.", pass);
+		if (pass == 1) {
+			/*
+			 * Now do pass 2, scanning the first part of the zone
+			 * we omitted in pass 1.
+			 */
+			pass = 2;
+			zone_end = zone_start;
+			switch (search_zone) {
+			case 1: /* mft_zone */
+				zone_start = vol->mft_zone_start;
+				break;
+			case 2: /* data1_zone */
+				zone_start = vol->mft_zone_end;
+				break;
+			case 4: /* data2_zone */
+				zone_start = 0;
+				break;
+			default:
+				BUG();
+			}
+			/* Sanity check. */
+			if (zone_end < zone_start)
+				zone_end = zone_start;
+			bmp_pos = zone_start;
+			ntfs_debug("Continuing outer while loop, pass 2, "
+					"zone_start 0x%llx, zone_end 0x%llx, "
+					"bmp_pos 0x%llx.",
+					(unsigned long long)zone_start,
+					(unsigned long long)zone_end,
+					(unsigned long long)bmp_pos);
+			continue;
+		} /* pass == 2 */
+done_zones_check:
+		ntfs_debug("At done_zones_check, search_zone %i, done_zones "
+				"before 0x%x, done_zones after 0x%x.",
+				search_zone, done_zones,
+				done_zones | search_zone);
+		done_zones |= search_zone;
+		if (done_zones < 7) {
+			ntfs_debug("Switching zone.");
+			/* Now switch to the next zone we haven't done yet. */
+			pass = 1;
+			switch (search_zone) {
+			case 1:
+				ntfs_debug("Switching from mft zone to data1 "
+						"zone.");
+				/* Update mft zone position. */
+				if (rlpos) {
+					LCN tc;
+
+					ntfs_debug("Before checks, "
+							"vol->mft_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->mft_zone_pos);
+					tc = rl[rlpos - 1].lcn +
+							rl[rlpos - 1].length;
+					if (tc >= vol->mft_zone_end) {
+						vol->mft_zone_pos =
+								vol->mft_lcn;
+						if (!vol->mft_zone_end)
+							vol->mft_zone_pos = 0;
+					} else if ((bmp_initial_pos >=
+							vol->mft_zone_pos ||
+							tc > vol->mft_zone_pos)
+							&& tc >= vol->mft_lcn)
+						vol->mft_zone_pos = tc;
+					ntfs_debug("After checks, "
+							"vol->mft_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->mft_zone_pos);
+				}
+				/* Switch from mft zone to data1 zone. */
+switch_to_data1_zone:		search_zone = 2;
+				zone_start = bmp_initial_pos =
+						vol->data1_zone_pos;
+				zone_end = vol->nr_clusters;
+				if (zone_start == vol->mft_zone_end)
+					pass = 2;
+				if (zone_start >= zone_end) {
+					vol->data1_zone_pos = zone_start =
+							vol->mft_zone_end;
+					pass = 2;
+				}
+				break;
+			case 2:
+				ntfs_debug("Switching from data1 zone to "
+						"data2 zone.");
+				/* Update data1 zone position. */
+				if (rlpos) {
+					LCN tc;
+
+					ntfs_debug("Before checks, "
+							"vol->data1_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->data1_zone_pos);
+					tc = rl[rlpos - 1].lcn +
+							rl[rlpos - 1].length;
+					if (tc >= vol->nr_clusters)
+						vol->data1_zone_pos =
+							     vol->mft_zone_end;
+					else if ((bmp_initial_pos >=
+						    vol->data1_zone_pos ||
+						    tc > vol->data1_zone_pos)
+						    && tc >= vol->mft_zone_end)
+						vol->data1_zone_pos = tc;
+					ntfs_debug("After checks, "
+							"vol->data1_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->data1_zone_pos);
+				}
+				/* Switch from data1 zone to data2 zone. */
+				search_zone = 4;
+				zone_start = bmp_initial_pos =
+						vol->data2_zone_pos;
+				zone_end = vol->mft_zone_start;
+				if (!zone_start)
+					pass = 2;
+				if (zone_start >= zone_end) {
+					vol->data2_zone_pos = zone_start =
+							bmp_initial_pos = 0;
+					pass = 2;
+				}
+				break;
+			case 4:
+				ntfs_debug("Switching from data2 zone to "
+						"data1 zone.");
+				/* Update data2 zone position. */
+				if (rlpos) {
+					LCN tc;
+
+					ntfs_debug("Before checks, "
+							"vol->data2_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->data2_zone_pos);
+					tc = rl[rlpos - 1].lcn +
+							rl[rlpos - 1].length;
+					if (tc >= vol->mft_zone_start)
+						vol->data2_zone_pos = 0;
+					else if (bmp_initial_pos >=
+						      vol->data2_zone_pos ||
+						      tc > vol->data2_zone_pos)
+						vol->data2_zone_pos = tc;
+					ntfs_debug("After checks, "
+							"vol->data2_zone_pos "
+							"0x%llx.",
+							(unsigned long long)
+							vol->data2_zone_pos);
+				}
+				/* Switch from data2 zone to data1 zone. */
+				goto switch_to_data1_zone;
+			default:
+				BUG();
+			}
+			ntfs_debug("After zone switch, search_zone %i, "
+					"pass %i, bmp_initial_pos 0x%llx, "
+					"zone_start 0x%llx, zone_end 0x%llx.",
+					search_zone, pass,
+					(unsigned long long)bmp_initial_pos,
+					(unsigned long long)zone_start,
+					(unsigned long long)zone_end);
+			bmp_pos = zone_start;
+			if (zone_start == zone_end) {
+				ntfs_debug("Empty zone, going to "
+						"done_zones_check.");
+				/* Empty zone. Don't bother searching it. */
+				goto done_zones_check;
+			}
+			ntfs_debug("Continuing outer while loop.");
+			continue;
+		} /* done_zones == 7 */
+		ntfs_debug("All zones are finished.");
+		/*
+		 * All zones are finished!  If DATA_ZONE, shrink mft zone.  If
+		 * MFT_ZONE, we have really run out of space.
+		 */
+		mft_zone_size = vol->mft_zone_end - vol->mft_zone_start;
+		ntfs_debug("vol->mft_zone_start 0x%llx, vol->mft_zone_end "
+				"0x%llx, mft_zone_size 0x%llx.",
+				(unsigned long long)vol->mft_zone_start,
+				(unsigned long long)vol->mft_zone_end,
+				(unsigned long long)mft_zone_size);
+		if (zone == MFT_ZONE || mft_zone_size <= 0) {
+			ntfs_debug("No free clusters left, going to out.");
+			/* Really no more space left on device. */
+			err = -ENOSPC;
+			goto out;
+		} /* zone == DATA_ZONE && mft_zone_size > 0 */
+		ntfs_debug("Shrinking mft zone.");
+		zone_end = vol->mft_zone_end;
+		mft_zone_size >>= 1;
+		if (mft_zone_size > 0)
+			vol->mft_zone_end = vol->mft_zone_start + mft_zone_size;
+		else /* mft zone and data2 zone no longer exist. */
+			vol->data2_zone_pos = vol->mft_zone_start =
+					vol->mft_zone_end = 0;
+		if (vol->mft_zone_pos >= vol->mft_zone_end) {
+			vol->mft_zone_pos = vol->mft_lcn;
+			if (!vol->mft_zone_end)
+				vol->mft_zone_pos = 0;
+		}
+		bmp_pos = zone_start = bmp_initial_pos =
+				vol->data1_zone_pos = vol->mft_zone_end;
+		search_zone = 2;
+		pass = 2;
+		done_zones &= ~2;
+		ntfs_debug("After shrinking mft zone, mft_zone_size 0x%llx, "
+				"vol->mft_zone_start 0x%llx, "
+				"vol->mft_zone_end 0x%llx, "
+				"vol->mft_zone_pos 0x%llx, search_zone 2, "
+				"pass 2, dones_zones 0x%x, zone_start 0x%llx, "
+				"zone_end 0x%llx, vol->data1_zone_pos 0x%llx, "
+				"continuing outer while loop.",
+				(unsigned long long)mft_zone_size,
+				(unsigned long long)vol->mft_zone_start,
+				(unsigned long long)vol->mft_zone_end,
+				(unsigned long long)vol->mft_zone_pos,
+				done_zones, (unsigned long long)zone_start,
+				(unsigned long long)zone_end,
+				(unsigned long long)vol->data1_zone_pos);
+	}
+	ntfs_debug("After outer while loop.");
+out:
+	ntfs_debug("At out.");
+	/* Add runlist terminator element. */
+	if (likely(rl)) {
+		rl[rlpos].vcn = rl[rlpos - 1].vcn + rl[rlpos - 1].length;
+		rl[rlpos].lcn = is_extension ? LCN_ENOENT : LCN_RL_NOT_MAPPED;
+		rl[rlpos].length = 0;
+	}
+	if (likely(page && !IS_ERR(page))) {
+		if (need_writeback) {
+			ntfs_debug("Marking page dirty.");
+			flush_dcache_page(page);
+			set_page_dirty(page);
+			need_writeback = 0;
+		}
+		ntfs_unmap_page(page);
+	}
+	if (likely(!err)) {
+		up_write(&vol->lcnbmp_lock);
+		ntfs_debug("Done.");
+		return rl;
+	}
+	ntfs_error(vol->sb, "Failed to allocate clusters, aborting "
+			"(error %i).", err);
+	if (rl) {
+		int err2;
+
+		if (err == -ENOSPC)
+			ntfs_debug("Not enough space to complete allocation, "
+					"err -ENOSPC, first free lcn 0x%llx, "
+					"could allocate up to 0x%llx "
+					"clusters.",
+					(unsigned long long)rl[0].lcn,
+					(unsigned long long)(count - clusters));
+		/* Deallocate all allocated clusters. */
+		ntfs_debug("Attempting rollback...");
+		err2 = ntfs_cluster_free_from_rl_nolock(vol, rl);
+		if (err2) {
+			ntfs_error(vol->sb, "Failed to rollback (error %i).  "
+					"Leaving inconsistent metadata!  "
+					"Unmount and run chkdsk.", err2);
+			NVolSetErrors(vol);
+		}
+		/* Free the runlist. */
+		ntfs_free(rl);
+	} else if (err == -ENOSPC)
+		ntfs_debug("No space left at all, err = -ENOSPC, first free "
+				"lcn = 0x%llx.",
+				(long long)vol->data1_zone_pos);
+	up_write(&vol->lcnbmp_lock);
+	return ERR_PTR(err);
+}
+
+/**
+ * __ntfs_cluster_free - free clusters on an ntfs volume
+ * @ni:		ntfs inode whose runlist describes the clusters to free
+ * @start_vcn:	vcn in the runlist of @ni at which to start freeing clusters
+ * @count:	number of clusters to free or -1 for all clusters
+ * @ctx:	active attribute search context if present or NULL if not
+ * @is_rollback:	true if this is a rollback operation
+ *
+ * Free @count clusters starting at the cluster @start_vcn in the runlist
+ * described by the vfs inode @ni.
+ *
+ * If @count is -1, all clusters from @start_vcn to the end of the runlist are
+ * deallocated.  Thus, to completely free all clusters in a runlist, use
+ * @start_vcn = 0 and @count = -1.
+ *
+ * If @ctx is specified, it is an active search context of @ni and its base mft
+ * record.  This is needed when __ntfs_cluster_free() encounters unmapped
+ * runlist fragments and allows their mapping.  If you do not have the mft
+ * record mapped, you can specify @ctx as NULL and __ntfs_cluster_free() will
+ * perform the necessary mapping and unmapping.
+ *
+ * Note, __ntfs_cluster_free() saves the state of @ctx on entry and restores it
+ * before returning.  Thus, @ctx will be left pointing to the same attribute on
+ * return as on entry.  However, the actual pointers in @ctx may point to
+ * different memory locations on return, so you must remember to reset any
+ * cached pointers from the @ctx, i.e. after the call to __ntfs_cluster_free(),
+ * you will probably want to do:
+ *	m = ctx->mrec;
+ *	a = ctx->attr;
+ * Assuming you cache ctx->attr in a variable @a of type ATTR_RECORD * and that
+ * you cache ctx->mrec in a variable @m of type MFT_RECORD *.
+ *
+ * @is_rollback should always be 'false', it is for internal use to rollback
+ * errors.  You probably want to use ntfs_cluster_free() instead.
+ *
+ * Note, __ntfs_cluster_free() does not modify the runlist, so you have to
+ * remove from the runlist or mark sparse the freed runs later.
+ *
+ * Return the number of deallocated clusters (not counting sparse ones) on
+ * success and -errno on error.
+ *
+ * WARNING: If @ctx is supplied, regardless of whether success or failure is
+ *	    returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
+ *	    is no longer valid, i.e. you need to either call
+ *	    ntfs_attr_reinit_search_ctx() or ntfs_attr_put_search_ctx() on it.
+ *	    In that case PTR_ERR(@ctx->mrec) will give you the error code for
+ *	    why the mapping of the old inode failed.
+ *
+ * Locking: - The runlist described by @ni must be locked for writing on entry
+ *	      and is locked on return.  Note the runlist may be modified when
+ *	      needed runlist fragments need to be mapped.
+ *	    - The volume lcn bitmap must be unlocked on entry and is unlocked
+ *	      on return.
+ *	    - This function takes the volume lcn bitmap lock for writing and
+ *	      modifies the bitmap contents.
+ *	    - If @ctx is NULL, the base mft record of @ni must not be mapped on
+ *	      entry and it will be left unmapped on return.
+ *	    - If @ctx is not NULL, the base mft record must be mapped on entry
+ *	      and it will be left mapped on return.
+ */
+s64 __ntfs_cluster_free(ntfs_inode *ni, const VCN start_vcn, s64 count,
+		ntfs_attr_search_ctx *ctx, const bool is_rollback)
+{
+	s64 delta, to_free, total_freed, real_freed;
+	ntfs_volume *vol;
+	struct inode *lcnbmp_vi;
+	runlist_element *rl;
+	int err;
+
+	BUG_ON(!ni);
+	ntfs_debug("Entering for i_ino 0x%lx, start_vcn 0x%llx, count "
+			"0x%llx.%s", ni->mft_no, (unsigned long long)start_vcn,
+			(unsigned long long)count,
+			is_rollback ? " (rollback)" : "");
+	vol = ni->vol;
+	lcnbmp_vi = vol->lcnbmp_ino;
+	BUG_ON(!lcnbmp_vi);
+	BUG_ON(start_vcn < 0);
+	BUG_ON(count < -1);
+	/*
+	 * Lock the lcn bitmap for writing but only if not rolling back.  We
+	 * must hold the lock all the way including through rollback otherwise
+	 * rollback is not possible because once we have cleared a bit and
+	 * dropped the lock, anyone could have set the bit again, thus
+	 * allocating the cluster for another use.
+	 */
+	if (likely(!is_rollback))
+		down_write(&vol->lcnbmp_lock);
+
+	total_freed = real_freed = 0;
+
+	rl = ntfs_attr_find_vcn_nolock(ni, start_vcn, ctx);
+	if (IS_ERR(rl)) {
+		if (!is_rollback)
+			ntfs_error(vol->sb, "Failed to find first runlist "
+					"element (error %li), aborting.",
+					PTR_ERR(rl));
+		err = PTR_ERR(rl);
+		goto err_out;
+	}
+	if (unlikely(rl->lcn < LCN_HOLE)) {
+		if (!is_rollback)
+			ntfs_error(vol->sb, "First runlist element has "
+					"invalid lcn, aborting.");
+		err = -EIO;
+		goto err_out;
+	}
+	/* Find the starting cluster inside the run that needs freeing. */
+	delta = start_vcn - rl->vcn;
+
+	/* The number of clusters in this run that need freeing. */
+	to_free = rl->length - delta;
+	if (count >= 0 && to_free > count)
+		to_free = count;
+
+	if (likely(rl->lcn >= 0)) {
+		/* Do the actual freeing of the clusters in this run. */
+		err = ntfs_bitmap_set_bits_in_run(lcnbmp_vi, rl->lcn + delta,
+				to_free, likely(!is_rollback) ? 0 : 1);
+		if (unlikely(err)) {
+			if (!is_rollback)
+				ntfs_error(vol->sb, "Failed to clear first run "
+						"(error %i), aborting.", err);
+			goto err_out;
+		}
+		/* We have freed @to_free real clusters. */
+		real_freed = to_free;
+	};
+	/* Go to the next run and adjust the number of clusters left to free. */
+	++rl;
+	if (count >= 0)
+		count -= to_free;
+
+	/* Keep track of the total "freed" clusters, including sparse ones. */
+	total_freed = to_free;
+	/*
+	 * Loop over the remaining runs, using @count as a capping value, and
+	 * free them.
+	 */
+	for (; rl->length && count != 0; ++rl) {
+		if (unlikely(rl->lcn < LCN_HOLE)) {
+			VCN vcn;
+
+			/* Attempt to map runlist. */
+			vcn = rl->vcn;
+			rl = ntfs_attr_find_vcn_nolock(ni, vcn, ctx);
+			if (IS_ERR(rl)) {
+				err = PTR_ERR(rl);
+				if (!is_rollback)
+					ntfs_error(vol->sb, "Failed to map "
+							"runlist fragment or "
+							"failed to find "
+							"subsequent runlist "
+							"element.");
+				goto err_out;
+			}
+			if (unlikely(rl->lcn < LCN_HOLE)) {
+				if (!is_rollback)
+					ntfs_error(vol->sb, "Runlist element "
+							"has invalid lcn "
+							"(0x%llx).",
+							(unsigned long long)
+							rl->lcn);
+				err = -EIO;
+				goto err_out;
+			}
+		}
+		/* The number of clusters in this run that need freeing. */
+		to_free = rl->length;
+		if (count >= 0 && to_free > count)
+			to_free = count;
+
+		if (likely(rl->lcn >= 0)) {
+			/* Do the actual freeing of the clusters in the run. */
+			err = ntfs_bitmap_set_bits_in_run(lcnbmp_vi, rl->lcn,
+					to_free, likely(!is_rollback) ? 0 : 1);
+			if (unlikely(err)) {
+				if (!is_rollback)
+					ntfs_error(vol->sb, "Failed to clear "
+							"subsequent run.");
+				goto err_out;
+			}
+			/* We have freed @to_free real clusters. */
+			real_freed += to_free;
+		}
+		/* Adjust the number of clusters left to free. */
+		if (count >= 0)
+			count -= to_free;
+	
+		/* Update the total done clusters. */
+		total_freed += to_free;
+	}
+	if (likely(!is_rollback))
+		up_write(&vol->lcnbmp_lock);
+
+	BUG_ON(count > 0);
+
+	/* We are done.  Return the number of actually freed clusters. */
+	ntfs_debug("Done.");
+	return real_freed;
+err_out:
+	if (is_rollback)
+		return err;
+	/* If no real clusters were freed, no need to rollback. */
+	if (!real_freed) {
+		up_write(&vol->lcnbmp_lock);
+		return err;
+	}
+	/*
+	 * Attempt to rollback and if that succeeds just return the error code.
+	 * If rollback fails, set the volume errors flag, emit an error
+	 * message, and return the error code.
+	 */
+	delta = __ntfs_cluster_free(ni, start_vcn, total_freed, ctx, true);
+	if (delta < 0) {
+		ntfs_error(vol->sb, "Failed to rollback (error %i).  Leaving "
+				"inconsistent metadata!  Unmount and run "
+				"chkdsk.", (int)delta);
+		NVolSetErrors(vol);
+	}
+	up_write(&vol->lcnbmp_lock);
+	ntfs_error(vol->sb, "Aborting (error %i).", err);
+	return err;
+}
+
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/lcnalloc.h b/fs/ntfs/lcnalloc.h
new file mode 100644
index 000000000..1589a6d84
--- /dev/null
+++ b/fs/ntfs/lcnalloc.h
@@ -0,0 +1,131 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * lcnalloc.h - Exports for NTFS kernel cluster (de)allocation.  Part of the
+ *		Linux-NTFS project.
+ *
+ * Copyright (c) 2004-2005 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_LCNALLOC_H
+#define _LINUX_NTFS_LCNALLOC_H
+
+#ifdef NTFS_RW
+
+#include <linux/fs.h>
+
+#include "attrib.h"
+#include "types.h"
+#include "inode.h"
+#include "runlist.h"
+#include "volume.h"
+
+typedef enum {
+	FIRST_ZONE	= 0,	/* For sanity checking. */
+	MFT_ZONE	= 0,	/* Allocate from $MFT zone. */
+	DATA_ZONE	= 1,	/* Allocate from $DATA zone. */
+	LAST_ZONE	= 1,	/* For sanity checking. */
+} NTFS_CLUSTER_ALLOCATION_ZONES;
+
+extern runlist_element *ntfs_cluster_alloc(ntfs_volume *vol,
+		const VCN start_vcn, const s64 count, const LCN start_lcn,
+		const NTFS_CLUSTER_ALLOCATION_ZONES zone,
+		const bool is_extension);
+
+extern s64 __ntfs_cluster_free(ntfs_inode *ni, const VCN start_vcn,
+		s64 count, ntfs_attr_search_ctx *ctx, const bool is_rollback);
+
+/**
+ * ntfs_cluster_free - free clusters on an ntfs volume
+ * @ni:		ntfs inode whose runlist describes the clusters to free
+ * @start_vcn:	vcn in the runlist of @ni at which to start freeing clusters
+ * @count:	number of clusters to free or -1 for all clusters
+ * @ctx:	active attribute search context if present or NULL if not
+ *
+ * Free @count clusters starting at the cluster @start_vcn in the runlist
+ * described by the ntfs inode @ni.
+ *
+ * If @count is -1, all clusters from @start_vcn to the end of the runlist are
+ * deallocated.  Thus, to completely free all clusters in a runlist, use
+ * @start_vcn = 0 and @count = -1.
+ *
+ * If @ctx is specified, it is an active search context of @ni and its base mft
+ * record.  This is needed when ntfs_cluster_free() encounters unmapped runlist
+ * fragments and allows their mapping.  If you do not have the mft record
+ * mapped, you can specify @ctx as NULL and ntfs_cluster_free() will perform
+ * the necessary mapping and unmapping.
+ *
+ * Note, ntfs_cluster_free() saves the state of @ctx on entry and restores it
+ * before returning.  Thus, @ctx will be left pointing to the same attribute on
+ * return as on entry.  However, the actual pointers in @ctx may point to
+ * different memory locations on return, so you must remember to reset any
+ * cached pointers from the @ctx, i.e. after the call to ntfs_cluster_free(),
+ * you will probably want to do:
+ *	m = ctx->mrec;
+ *	a = ctx->attr;
+ * Assuming you cache ctx->attr in a variable @a of type ATTR_RECORD * and that
+ * you cache ctx->mrec in a variable @m of type MFT_RECORD *.
+ *
+ * Note, ntfs_cluster_free() does not modify the runlist, so you have to remove
+ * from the runlist or mark sparse the freed runs later.
+ *
+ * Return the number of deallocated clusters (not counting sparse ones) on
+ * success and -errno on error.
+ *
+ * WARNING: If @ctx is supplied, regardless of whether success or failure is
+ *	    returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
+ *	    is no longer valid, i.e. you need to either call
+ *	    ntfs_attr_reinit_search_ctx() or ntfs_attr_put_search_ctx() on it.
+ *	    In that case PTR_ERR(@ctx->mrec) will give you the error code for
+ *	    why the mapping of the old inode failed.
+ *
+ * Locking: - The runlist described by @ni must be locked for writing on entry
+ *	      and is locked on return.  Note the runlist may be modified when
+ *	      needed runlist fragments need to be mapped.
+ *	    - The volume lcn bitmap must be unlocked on entry and is unlocked
+ *	      on return.
+ *	    - This function takes the volume lcn bitmap lock for writing and
+ *	      modifies the bitmap contents.
+ *	    - If @ctx is NULL, the base mft record of @ni must not be mapped on
+ *	      entry and it will be left unmapped on return.
+ *	    - If @ctx is not NULL, the base mft record must be mapped on entry
+ *	      and it will be left mapped on return.
+ */
+static inline s64 ntfs_cluster_free(ntfs_inode *ni, const VCN start_vcn,
+		s64 count, ntfs_attr_search_ctx *ctx)
+{
+	return __ntfs_cluster_free(ni, start_vcn, count, ctx, false);
+}
+
+extern int ntfs_cluster_free_from_rl_nolock(ntfs_volume *vol,
+		const runlist_element *rl);
+
+/**
+ * ntfs_cluster_free_from_rl - free clusters from runlist
+ * @vol:	mounted ntfs volume on which to free the clusters
+ * @rl:		runlist describing the clusters to free
+ *
+ * Free all the clusters described by the runlist @rl on the volume @vol.  In
+ * the case of an error being returned, at least some of the clusters were not
+ * freed.
+ *
+ * Return 0 on success and -errno on error.
+ *
+ * Locking: - This function takes the volume lcn bitmap lock for writing and
+ *	      modifies the bitmap contents.
+ *	    - The caller must have locked the runlist @rl for reading or
+ *	      writing.
+ */
+static inline int ntfs_cluster_free_from_rl(ntfs_volume *vol,
+		const runlist_element *rl)
+{
+	int ret;
+
+	down_write(&vol->lcnbmp_lock);
+	ret = ntfs_cluster_free_from_rl_nolock(vol, rl);
+	up_write(&vol->lcnbmp_lock);
+	return ret;
+}
+
+#endif /* NTFS_RW */
+
+#endif /* defined _LINUX_NTFS_LCNALLOC_H */
diff --git a/fs/ntfs/logfile.c b/fs/ntfs/logfile.c
new file mode 100644
index 000000000..a0c40f1be
--- /dev/null
+++ b/fs/ntfs/logfile.c
@@ -0,0 +1,850 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * logfile.c - NTFS kernel journal handling. Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2002-2007 Anton Altaparmakov
+ */
+
+#ifdef NTFS_RW
+
+#include <linux/types.h>
+#include <linux/fs.h>
+#include <linux/highmem.h>
+#include <linux/buffer_head.h>
+#include <linux/bitops.h>
+#include <linux/log2.h>
+#include <linux/bio.h>
+
+#include "attrib.h"
+#include "aops.h"
+#include "debug.h"
+#include "logfile.h"
+#include "malloc.h"
+#include "volume.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_check_restart_page_header - check the page header for consistency
+ * @vi:		$LogFile inode to which the restart page header belongs
+ * @rp:		restart page header to check
+ * @pos:	position in @vi at which the restart page header resides
+ *
+ * Check the restart page header @rp for consistency and return 'true' if it is
+ * consistent and 'false' otherwise.
+ *
+ * This function only needs NTFS_BLOCK_SIZE bytes in @rp, i.e. it does not
+ * require the full restart page.
+ */
+static bool ntfs_check_restart_page_header(struct inode *vi,
+		RESTART_PAGE_HEADER *rp, s64 pos)
+{
+	u32 logfile_system_page_size, logfile_log_page_size;
+	u16 ra_ofs, usa_count, usa_ofs, usa_end = 0;
+	bool have_usa = true;
+
+	ntfs_debug("Entering.");
+	/*
+	 * If the system or log page sizes are smaller than the ntfs block size
+	 * or either is not a power of 2 we cannot handle this log file.
+	 */
+	logfile_system_page_size = le32_to_cpu(rp->system_page_size);
+	logfile_log_page_size = le32_to_cpu(rp->log_page_size);
+	if (logfile_system_page_size < NTFS_BLOCK_SIZE ||
+			logfile_log_page_size < NTFS_BLOCK_SIZE ||
+			logfile_system_page_size &
+			(logfile_system_page_size - 1) ||
+			!is_power_of_2(logfile_log_page_size)) {
+		ntfs_error(vi->i_sb, "$LogFile uses unsupported page size.");
+		return false;
+	}
+	/*
+	 * We must be either at !pos (1st restart page) or at pos = system page
+	 * size (2nd restart page).
+	 */
+	if (pos && pos != logfile_system_page_size) {
+		ntfs_error(vi->i_sb, "Found restart area in incorrect "
+				"position in $LogFile.");
+		return false;
+	}
+	/* We only know how to handle version 1.1. */
+	if (sle16_to_cpu(rp->major_ver) != 1 ||
+			sle16_to_cpu(rp->minor_ver) != 1) {
+		ntfs_error(vi->i_sb, "$LogFile version %i.%i is not "
+				"supported.  (This driver supports version "
+				"1.1 only.)", (int)sle16_to_cpu(rp->major_ver),
+				(int)sle16_to_cpu(rp->minor_ver));
+		return false;
+	}
+	/*
+	 * If chkdsk has been run the restart page may not be protected by an
+	 * update sequence array.
+	 */
+	if (ntfs_is_chkd_record(rp->magic) && !le16_to_cpu(rp->usa_count)) {
+		have_usa = false;
+		goto skip_usa_checks;
+	}
+	/* Verify the size of the update sequence array. */
+	usa_count = 1 + (logfile_system_page_size >> NTFS_BLOCK_SIZE_BITS);
+	if (usa_count != le16_to_cpu(rp->usa_count)) {
+		ntfs_error(vi->i_sb, "$LogFile restart page specifies "
+				"inconsistent update sequence array count.");
+		return false;
+	}
+	/* Verify the position of the update sequence array. */
+	usa_ofs = le16_to_cpu(rp->usa_ofs);
+	usa_end = usa_ofs + usa_count * sizeof(u16);
+	if (usa_ofs < sizeof(RESTART_PAGE_HEADER) ||
+			usa_end > NTFS_BLOCK_SIZE - sizeof(u16)) {
+		ntfs_error(vi->i_sb, "$LogFile restart page specifies "
+				"inconsistent update sequence array offset.");
+		return false;
+	}
+skip_usa_checks:
+	/*
+	 * Verify the position of the restart area.  It must be:
+	 *	- aligned to 8-byte boundary,
+	 *	- after the update sequence array, and
+	 *	- within the system page size.
+	 */
+	ra_ofs = le16_to_cpu(rp->restart_area_offset);
+	if (ra_ofs & 7 || (have_usa ? ra_ofs < usa_end :
+			ra_ofs < sizeof(RESTART_PAGE_HEADER)) ||
+			ra_ofs > logfile_system_page_size) {
+		ntfs_error(vi->i_sb, "$LogFile restart page specifies "
+				"inconsistent restart area offset.");
+		return false;
+	}
+	/*
+	 * Only restart pages modified by chkdsk are allowed to have chkdsk_lsn
+	 * set.
+	 */
+	if (!ntfs_is_chkd_record(rp->magic) && sle64_to_cpu(rp->chkdsk_lsn)) {
+		ntfs_error(vi->i_sb, "$LogFile restart page is not modified "
+				"by chkdsk but a chkdsk LSN is specified.");
+		return false;
+	}
+	ntfs_debug("Done.");
+	return true;
+}
+
+/**
+ * ntfs_check_restart_area - check the restart area for consistency
+ * @vi:		$LogFile inode to which the restart page belongs
+ * @rp:		restart page whose restart area to check
+ *
+ * Check the restart area of the restart page @rp for consistency and return
+ * 'true' if it is consistent and 'false' otherwise.
+ *
+ * This function assumes that the restart page header has already been
+ * consistency checked.
+ *
+ * This function only needs NTFS_BLOCK_SIZE bytes in @rp, i.e. it does not
+ * require the full restart page.
+ */
+static bool ntfs_check_restart_area(struct inode *vi, RESTART_PAGE_HEADER *rp)
+{
+	u64 file_size;
+	RESTART_AREA *ra;
+	u16 ra_ofs, ra_len, ca_ofs;
+	u8 fs_bits;
+
+	ntfs_debug("Entering.");
+	ra_ofs = le16_to_cpu(rp->restart_area_offset);
+	ra = (RESTART_AREA*)((u8*)rp + ra_ofs);
+	/*
+	 * Everything before ra->file_size must be before the first word
+	 * protected by an update sequence number.  This ensures that it is
+	 * safe to access ra->client_array_offset.
+	 */
+	if (ra_ofs + offsetof(RESTART_AREA, file_size) >
+			NTFS_BLOCK_SIZE - sizeof(u16)) {
+		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
+				"inconsistent file offset.");
+		return false;
+	}
+	/*
+	 * Now that we can access ra->client_array_offset, make sure everything
+	 * up to the log client array is before the first word protected by an
+	 * update sequence number.  This ensures we can access all of the
+	 * restart area elements safely.  Also, the client array offset must be
+	 * aligned to an 8-byte boundary.
+	 */
+	ca_ofs = le16_to_cpu(ra->client_array_offset);
+	if (((ca_ofs + 7) & ~7) != ca_ofs ||
+			ra_ofs + ca_ofs > NTFS_BLOCK_SIZE - sizeof(u16)) {
+		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
+				"inconsistent client array offset.");
+		return false;
+	}
+	/*
+	 * The restart area must end within the system page size both when
+	 * calculated manually and as specified by ra->restart_area_length.
+	 * Also, the calculated length must not exceed the specified length.
+	 */
+	ra_len = ca_ofs + le16_to_cpu(ra->log_clients) *
+			sizeof(LOG_CLIENT_RECORD);
+	if (ra_ofs + ra_len > le32_to_cpu(rp->system_page_size) ||
+			ra_ofs + le16_to_cpu(ra->restart_area_length) >
+			le32_to_cpu(rp->system_page_size) ||
+			ra_len > le16_to_cpu(ra->restart_area_length)) {
+		ntfs_error(vi->i_sb, "$LogFile restart area is out of bounds "
+				"of the system page size specified by the "
+				"restart page header and/or the specified "
+				"restart area length is inconsistent.");
+		return false;
+	}
+	/*
+	 * The ra->client_free_list and ra->client_in_use_list must be either
+	 * LOGFILE_NO_CLIENT or less than ra->log_clients or they are
+	 * overflowing the client array.
+	 */
+	if ((ra->client_free_list != LOGFILE_NO_CLIENT &&
+			le16_to_cpu(ra->client_free_list) >=
+			le16_to_cpu(ra->log_clients)) ||
+			(ra->client_in_use_list != LOGFILE_NO_CLIENT &&
+			le16_to_cpu(ra->client_in_use_list) >=
+			le16_to_cpu(ra->log_clients))) {
+		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
+				"overflowing client free and/or in use lists.");
+		return false;
+	}
+	/*
+	 * Check ra->seq_number_bits against ra->file_size for consistency.
+	 * We cannot just use ffs() because the file size is not a power of 2.
+	 */
+	file_size = (u64)sle64_to_cpu(ra->file_size);
+	fs_bits = 0;
+	while (file_size) {
+		file_size >>= 1;
+		fs_bits++;
+	}
+	if (le32_to_cpu(ra->seq_number_bits) != 67 - fs_bits) {
+		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
+				"inconsistent sequence number bits.");
+		return false;
+	}
+	/* The log record header length must be a multiple of 8. */
+	if (((le16_to_cpu(ra->log_record_header_length) + 7) & ~7) !=
+			le16_to_cpu(ra->log_record_header_length)) {
+		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
+				"inconsistent log record header length.");
+		return false;
+	}
+	/* Dito for the log page data offset. */
+	if (((le16_to_cpu(ra->log_page_data_offset) + 7) & ~7) !=
+			le16_to_cpu(ra->log_page_data_offset)) {
+		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
+				"inconsistent log page data offset.");
+		return false;
+	}
+	ntfs_debug("Done.");
+	return true;
+}
+
+/**
+ * ntfs_check_log_client_array - check the log client array for consistency
+ * @vi:		$LogFile inode to which the restart page belongs
+ * @rp:		restart page whose log client array to check
+ *
+ * Check the log client array of the restart page @rp for consistency and
+ * return 'true' if it is consistent and 'false' otherwise.
+ *
+ * This function assumes that the restart page header and the restart area have
+ * already been consistency checked.
+ *
+ * Unlike ntfs_check_restart_page_header() and ntfs_check_restart_area(), this
+ * function needs @rp->system_page_size bytes in @rp, i.e. it requires the full
+ * restart page and the page must be multi sector transfer deprotected.
+ */
+static bool ntfs_check_log_client_array(struct inode *vi,
+		RESTART_PAGE_HEADER *rp)
+{
+	RESTART_AREA *ra;
+	LOG_CLIENT_RECORD *ca, *cr;
+	u16 nr_clients, idx;
+	bool in_free_list, idx_is_first;
+
+	ntfs_debug("Entering.");
+	ra = (RESTART_AREA*)((u8*)rp + le16_to_cpu(rp->restart_area_offset));
+	ca = (LOG_CLIENT_RECORD*)((u8*)ra +
+			le16_to_cpu(ra->client_array_offset));
+	/*
+	 * Check the ra->client_free_list first and then check the
+	 * ra->client_in_use_list.  Check each of the log client records in
+	 * each of the lists and check that the array does not overflow the
+	 * ra->log_clients value.  Also keep track of the number of records
+	 * visited as there cannot be more than ra->log_clients records and
+	 * that way we detect eventual loops in within a list.
+	 */
+	nr_clients = le16_to_cpu(ra->log_clients);
+	idx = le16_to_cpu(ra->client_free_list);
+	in_free_list = true;
+check_list:
+	for (idx_is_first = true; idx != LOGFILE_NO_CLIENT_CPU; nr_clients--,
+			idx = le16_to_cpu(cr->next_client)) {
+		if (!nr_clients || idx >= le16_to_cpu(ra->log_clients))
+			goto err_out;
+		/* Set @cr to the current log client record. */
+		cr = ca + idx;
+		/* The first log client record must not have a prev_client. */
+		if (idx_is_first) {
+			if (cr->prev_client != LOGFILE_NO_CLIENT)
+				goto err_out;
+			idx_is_first = false;
+		}
+	}
+	/* Switch to and check the in use list if we just did the free list. */
+	if (in_free_list) {
+		in_free_list = false;
+		idx = le16_to_cpu(ra->client_in_use_list);
+		goto check_list;
+	}
+	ntfs_debug("Done.");
+	return true;
+err_out:
+	ntfs_error(vi->i_sb, "$LogFile log client array is corrupt.");
+	return false;
+}
+
+/**
+ * ntfs_check_and_load_restart_page - check the restart page for consistency
+ * @vi:		$LogFile inode to which the restart page belongs
+ * @rp:		restart page to check
+ * @pos:	position in @vi at which the restart page resides
+ * @wrp:	[OUT] copy of the multi sector transfer deprotected restart page
+ * @lsn:	[OUT] set to the current logfile lsn on success
+ *
+ * Check the restart page @rp for consistency and return 0 if it is consistent
+ * and -errno otherwise.  The restart page may have been modified by chkdsk in
+ * which case its magic is CHKD instead of RSTR.
+ *
+ * This function only needs NTFS_BLOCK_SIZE bytes in @rp, i.e. it does not
+ * require the full restart page.
+ *
+ * If @wrp is not NULL, on success, *@wrp will point to a buffer containing a
+ * copy of the complete multi sector transfer deprotected page.  On failure,
+ * *@wrp is undefined.
+ *
+ * Simillarly, if @lsn is not NULL, on success *@lsn will be set to the current
+ * logfile lsn according to this restart page.  On failure, *@lsn is undefined.
+ *
+ * The following error codes are defined:
+ *	-EINVAL	- The restart page is inconsistent.
+ *	-ENOMEM	- Not enough memory to load the restart page.
+ *	-EIO	- Failed to reading from $LogFile.
+ */
+static int ntfs_check_and_load_restart_page(struct inode *vi,
+		RESTART_PAGE_HEADER *rp, s64 pos, RESTART_PAGE_HEADER **wrp,
+		LSN *lsn)
+{
+	RESTART_AREA *ra;
+	RESTART_PAGE_HEADER *trp;
+	int size, err;
+
+	ntfs_debug("Entering.");
+	/* Check the restart page header for consistency. */
+	if (!ntfs_check_restart_page_header(vi, rp, pos)) {
+		/* Error output already done inside the function. */
+		return -EINVAL;
+	}
+	/* Check the restart area for consistency. */
+	if (!ntfs_check_restart_area(vi, rp)) {
+		/* Error output already done inside the function. */
+		return -EINVAL;
+	}
+	ra = (RESTART_AREA*)((u8*)rp + le16_to_cpu(rp->restart_area_offset));
+	/*
+	 * Allocate a buffer to store the whole restart page so we can multi
+	 * sector transfer deprotect it.
+	 */
+	trp = ntfs_malloc_nofs(le32_to_cpu(rp->system_page_size));
+	if (!trp) {
+		ntfs_error(vi->i_sb, "Failed to allocate memory for $LogFile "
+				"restart page buffer.");
+		return -ENOMEM;
+	}
+	/*
+	 * Read the whole of the restart page into the buffer.  If it fits
+	 * completely inside @rp, just copy it from there.  Otherwise map all
+	 * the required pages and copy the data from them.
+	 */
+	size = PAGE_SIZE - (pos & ~PAGE_MASK);
+	if (size >= le32_to_cpu(rp->system_page_size)) {
+		memcpy(trp, rp, le32_to_cpu(rp->system_page_size));
+	} else {
+		pgoff_t idx;
+		struct page *page;
+		int have_read, to_read;
+
+		/* First copy what we already have in @rp. */
+		memcpy(trp, rp, size);
+		/* Copy the remaining data one page at a time. */
+		have_read = size;
+		to_read = le32_to_cpu(rp->system_page_size) - size;
+		idx = (pos + size) >> PAGE_SHIFT;
+		BUG_ON((pos + size) & ~PAGE_MASK);
+		do {
+			page = ntfs_map_page(vi->i_mapping, idx);
+			if (IS_ERR(page)) {
+				ntfs_error(vi->i_sb, "Error mapping $LogFile "
+						"page (index %lu).", idx);
+				err = PTR_ERR(page);
+				if (err != -EIO && err != -ENOMEM)
+					err = -EIO;
+				goto err_out;
+			}
+			size = min_t(int, to_read, PAGE_SIZE);
+			memcpy((u8*)trp + have_read, page_address(page), size);
+			ntfs_unmap_page(page);
+			have_read += size;
+			to_read -= size;
+			idx++;
+		} while (to_read > 0);
+	}
+	/*
+	 * Perform the multi sector transfer deprotection on the buffer if the
+	 * restart page is protected.
+	 */
+	if ((!ntfs_is_chkd_record(trp->magic) || le16_to_cpu(trp->usa_count))
+			&& post_read_mst_fixup((NTFS_RECORD*)trp,
+			le32_to_cpu(rp->system_page_size))) {
+		/*
+		 * A multi sector tranfer error was detected.  We only need to
+		 * abort if the restart page contents exceed the multi sector
+		 * transfer fixup of the first sector.
+		 */
+		if (le16_to_cpu(rp->restart_area_offset) +
+				le16_to_cpu(ra->restart_area_length) >
+				NTFS_BLOCK_SIZE - sizeof(u16)) {
+			ntfs_error(vi->i_sb, "Multi sector transfer error "
+					"detected in $LogFile restart page.");
+			err = -EINVAL;
+			goto err_out;
+		}
+	}
+	/*
+	 * If the restart page is modified by chkdsk or there are no active
+	 * logfile clients, the logfile is consistent.  Otherwise, need to
+	 * check the log client records for consistency, too.
+	 */
+	err = 0;
+	if (ntfs_is_rstr_record(rp->magic) &&
+			ra->client_in_use_list != LOGFILE_NO_CLIENT) {
+		if (!ntfs_check_log_client_array(vi, trp)) {
+			err = -EINVAL;
+			goto err_out;
+		}
+	}
+	if (lsn) {
+		if (ntfs_is_rstr_record(rp->magic))
+			*lsn = sle64_to_cpu(ra->current_lsn);
+		else /* if (ntfs_is_chkd_record(rp->magic)) */
+			*lsn = sle64_to_cpu(rp->chkdsk_lsn);
+	}
+	ntfs_debug("Done.");
+	if (wrp)
+		*wrp = trp;
+	else {
+err_out:
+		ntfs_free(trp);
+	}
+	return err;
+}
+
+/**
+ * ntfs_check_logfile - check the journal for consistency
+ * @log_vi:	struct inode of loaded journal $LogFile to check
+ * @rp:		[OUT] on success this is a copy of the current restart page
+ *
+ * Check the $LogFile journal for consistency and return 'true' if it is
+ * consistent and 'false' if not.  On success, the current restart page is
+ * returned in *@rp.  Caller must call ntfs_free(*@rp) when finished with it.
+ *
+ * At present we only check the two restart pages and ignore the log record
+ * pages.
+ *
+ * Note that the MstProtected flag is not set on the $LogFile inode and hence
+ * when reading pages they are not deprotected.  This is because we do not know
+ * if the $LogFile was created on a system with a different page size to ours
+ * yet and mst deprotection would fail if our page size is smaller.
+ */
+bool ntfs_check_logfile(struct inode *log_vi, RESTART_PAGE_HEADER **rp)
+{
+	s64 size, pos;
+	LSN rstr1_lsn, rstr2_lsn;
+	ntfs_volume *vol = NTFS_SB(log_vi->i_sb);
+	struct address_space *mapping = log_vi->i_mapping;
+	struct page *page = NULL;
+	u8 *kaddr = NULL;
+	RESTART_PAGE_HEADER *rstr1_ph = NULL;
+	RESTART_PAGE_HEADER *rstr2_ph = NULL;
+	int log_page_size, log_page_mask, err;
+	bool logfile_is_empty = true;
+	u8 log_page_bits;
+
+	ntfs_debug("Entering.");
+	/* An empty $LogFile must have been clean before it got emptied. */
+	if (NVolLogFileEmpty(vol))
+		goto is_empty;
+	size = i_size_read(log_vi);
+	/* Make sure the file doesn't exceed the maximum allowed size. */
+	if (size > MaxLogFileSize)
+		size = MaxLogFileSize;
+	/*
+	 * Truncate size to a multiple of the page cache size or the default
+	 * log page size if the page cache size is between the default log page
+	 * log page size if the page cache size is between the default log page
+	 * size and twice that.
+	 */
+	if (PAGE_SIZE >= DefaultLogPageSize && PAGE_SIZE <=
+			DefaultLogPageSize * 2)
+		log_page_size = DefaultLogPageSize;
+	else
+		log_page_size = PAGE_SIZE;
+	log_page_mask = log_page_size - 1;
+	/*
+	 * Use ntfs_ffs() instead of ffs() to enable the compiler to
+	 * optimize log_page_size and log_page_bits into constants.
+	 */
+	log_page_bits = ntfs_ffs(log_page_size) - 1;
+	size &= ~(s64)(log_page_size - 1);
+	/*
+	 * Ensure the log file is big enough to store at least the two restart
+	 * pages and the minimum number of log record pages.
+	 */
+	if (size < log_page_size * 2 || (size - log_page_size * 2) >>
+			log_page_bits < MinLogRecordPages) {
+		ntfs_error(vol->sb, "$LogFile is too small.");
+		return false;
+	}
+	/*
+	 * Read through the file looking for a restart page.  Since the restart
+	 * page header is at the beginning of a page we only need to search at
+	 * what could be the beginning of a page (for each page size) rather
+	 * than scanning the whole file byte by byte.  If all potential places
+	 * contain empty and uninitialzed records, the log file can be assumed
+	 * to be empty.
+	 */
+	for (pos = 0; pos < size; pos <<= 1) {
+		pgoff_t idx = pos >> PAGE_SHIFT;
+		if (!page || page->index != idx) {
+			if (page)
+				ntfs_unmap_page(page);
+			page = ntfs_map_page(mapping, idx);
+			if (IS_ERR(page)) {
+				ntfs_error(vol->sb, "Error mapping $LogFile "
+						"page (index %lu).", idx);
+				goto err_out;
+			}
+		}
+		kaddr = (u8*)page_address(page) + (pos & ~PAGE_MASK);
+		/*
+		 * A non-empty block means the logfile is not empty while an
+		 * empty block after a non-empty block has been encountered
+		 * means we are done.
+		 */
+		if (!ntfs_is_empty_recordp((le32*)kaddr))
+			logfile_is_empty = false;
+		else if (!logfile_is_empty)
+			break;
+		/*
+		 * A log record page means there cannot be a restart page after
+		 * this so no need to continue searching.
+		 */
+		if (ntfs_is_rcrd_recordp((le32*)kaddr))
+			break;
+		/* If not a (modified by chkdsk) restart page, continue. */
+		if (!ntfs_is_rstr_recordp((le32*)kaddr) &&
+				!ntfs_is_chkd_recordp((le32*)kaddr)) {
+			if (!pos)
+				pos = NTFS_BLOCK_SIZE >> 1;
+			continue;
+		}
+		/*
+		 * Check the (modified by chkdsk) restart page for consistency
+		 * and get a copy of the complete multi sector transfer
+		 * deprotected restart page.
+		 */
+		err = ntfs_check_and_load_restart_page(log_vi,
+				(RESTART_PAGE_HEADER*)kaddr, pos,
+				!rstr1_ph ? &rstr1_ph : &rstr2_ph,
+				!rstr1_ph ? &rstr1_lsn : &rstr2_lsn);
+		if (!err) {
+			/*
+			 * If we have now found the first (modified by chkdsk)
+			 * restart page, continue looking for the second one.
+			 */
+			if (!pos) {
+				pos = NTFS_BLOCK_SIZE >> 1;
+				continue;
+			}
+			/*
+			 * We have now found the second (modified by chkdsk)
+			 * restart page, so we can stop looking.
+			 */
+			break;
+		}
+		/*
+		 * Error output already done inside the function.  Note, we do
+		 * not abort if the restart page was invalid as we might still
+		 * find a valid one further in the file.
+		 */
+		if (err != -EINVAL) {
+			ntfs_unmap_page(page);
+			goto err_out;
+		}
+		/* Continue looking. */
+		if (!pos)
+			pos = NTFS_BLOCK_SIZE >> 1;
+	}
+	if (page)
+		ntfs_unmap_page(page);
+	if (logfile_is_empty) {
+		NVolSetLogFileEmpty(vol);
+is_empty:
+		ntfs_debug("Done.  ($LogFile is empty.)");
+		return true;
+	}
+	if (!rstr1_ph) {
+		BUG_ON(rstr2_ph);
+		ntfs_error(vol->sb, "Did not find any restart pages in "
+				"$LogFile and it was not empty.");
+		return false;
+	}
+	/* If both restart pages were found, use the more recent one. */
+	if (rstr2_ph) {
+		/*
+		 * If the second restart area is more recent, switch to it.
+		 * Otherwise just throw it away.
+		 */
+		if (rstr2_lsn > rstr1_lsn) {
+			ntfs_debug("Using second restart page as it is more "
+					"recent.");
+			ntfs_free(rstr1_ph);
+			rstr1_ph = rstr2_ph;
+			/* rstr1_lsn = rstr2_lsn; */
+		} else {
+			ntfs_debug("Using first restart page as it is more "
+					"recent.");
+			ntfs_free(rstr2_ph);
+		}
+		rstr2_ph = NULL;
+	}
+	/* All consistency checks passed. */
+	if (rp)
+		*rp = rstr1_ph;
+	else
+		ntfs_free(rstr1_ph);
+	ntfs_debug("Done.");
+	return true;
+err_out:
+	if (rstr1_ph)
+		ntfs_free(rstr1_ph);
+	return false;
+}
+
+/**
+ * ntfs_is_logfile_clean - check in the journal if the volume is clean
+ * @log_vi:	struct inode of loaded journal $LogFile to check
+ * @rp:		copy of the current restart page
+ *
+ * Analyze the $LogFile journal and return 'true' if it indicates the volume was
+ * shutdown cleanly and 'false' if not.
+ *
+ * At present we only look at the two restart pages and ignore the log record
+ * pages.  This is a little bit crude in that there will be a very small number
+ * of cases where we think that a volume is dirty when in fact it is clean.
+ * This should only affect volumes that have not been shutdown cleanly but did
+ * not have any pending, non-check-pointed i/o, i.e. they were completely idle
+ * at least for the five seconds preceding the unclean shutdown.
+ *
+ * This function assumes that the $LogFile journal has already been consistency
+ * checked by a call to ntfs_check_logfile() and in particular if the $LogFile
+ * is empty this function requires that NVolLogFileEmpty() is true otherwise an
+ * empty volume will be reported as dirty.
+ */
+bool ntfs_is_logfile_clean(struct inode *log_vi, const RESTART_PAGE_HEADER *rp)
+{
+	ntfs_volume *vol = NTFS_SB(log_vi->i_sb);
+	RESTART_AREA *ra;
+
+	ntfs_debug("Entering.");
+	/* An empty $LogFile must have been clean before it got emptied. */
+	if (NVolLogFileEmpty(vol)) {
+		ntfs_debug("Done.  ($LogFile is empty.)");
+		return true;
+	}
+	BUG_ON(!rp);
+	if (!ntfs_is_rstr_record(rp->magic) &&
+			!ntfs_is_chkd_record(rp->magic)) {
+		ntfs_error(vol->sb, "Restart page buffer is invalid.  This is "
+				"probably a bug in that the $LogFile should "
+				"have been consistency checked before calling "
+				"this function.");
+		return false;
+	}
+	ra = (RESTART_AREA*)((u8*)rp + le16_to_cpu(rp->restart_area_offset));
+	/*
+	 * If the $LogFile has active clients, i.e. it is open, and we do not
+	 * have the RESTART_VOLUME_IS_CLEAN bit set in the restart area flags,
+	 * we assume there was an unclean shutdown.
+	 */
+	if (ra->client_in_use_list != LOGFILE_NO_CLIENT &&
+			!(ra->flags & RESTART_VOLUME_IS_CLEAN)) {
+		ntfs_debug("Done.  $LogFile indicates a dirty shutdown.");
+		return false;
+	}
+	/* $LogFile indicates a clean shutdown. */
+	ntfs_debug("Done.  $LogFile indicates a clean shutdown.");
+	return true;
+}
+
+/**
+ * ntfs_empty_logfile - empty the contents of the $LogFile journal
+ * @log_vi:	struct inode of loaded journal $LogFile to empty
+ *
+ * Empty the contents of the $LogFile journal @log_vi and return 'true' on
+ * success and 'false' on error.
+ *
+ * This function assumes that the $LogFile journal has already been consistency
+ * checked by a call to ntfs_check_logfile() and that ntfs_is_logfile_clean()
+ * has been used to ensure that the $LogFile is clean.
+ */
+bool ntfs_empty_logfile(struct inode *log_vi)
+{
+	VCN vcn, end_vcn;
+	ntfs_inode *log_ni = NTFS_I(log_vi);
+	ntfs_volume *vol = log_ni->vol;
+	struct super_block *sb = vol->sb;
+	runlist_element *rl;
+	unsigned long flags;
+	unsigned block_size, block_size_bits;
+	int err;
+	bool should_wait = true;
+
+	ntfs_debug("Entering.");
+	if (NVolLogFileEmpty(vol)) {
+		ntfs_debug("Done.");
+		return true;
+	}
+	/*
+	 * We cannot use ntfs_attr_set() because we may be still in the middle
+	 * of a mount operation.  Thus we do the emptying by hand by first
+	 * zapping the page cache pages for the $LogFile/$DATA attribute and
+	 * then emptying each of the buffers in each of the clusters specified
+	 * by the runlist by hand.
+	 */
+	block_size = sb->s_blocksize;
+	block_size_bits = sb->s_blocksize_bits;
+	vcn = 0;
+	read_lock_irqsave(&log_ni->size_lock, flags);
+	end_vcn = (log_ni->initialized_size + vol->cluster_size_mask) >>
+			vol->cluster_size_bits;
+	read_unlock_irqrestore(&log_ni->size_lock, flags);
+	truncate_inode_pages(log_vi->i_mapping, 0);
+	down_write(&log_ni->runlist.lock);
+	rl = log_ni->runlist.rl;
+	if (unlikely(!rl || vcn < rl->vcn || !rl->length)) {
+map_vcn:
+		err = ntfs_map_runlist_nolock(log_ni, vcn, NULL);
+		if (err) {
+			ntfs_error(sb, "Failed to map runlist fragment (error "
+					"%d).", -err);
+			goto err;
+		}
+		rl = log_ni->runlist.rl;
+		BUG_ON(!rl || vcn < rl->vcn || !rl->length);
+	}
+	/* Seek to the runlist element containing @vcn. */
+	while (rl->length && vcn >= rl[1].vcn)
+		rl++;
+	do {
+		LCN lcn;
+		sector_t block, end_block;
+		s64 len;
+
+		/*
+		 * If this run is not mapped map it now and start again as the
+		 * runlist will have been updated.
+		 */
+		lcn = rl->lcn;
+		if (unlikely(lcn == LCN_RL_NOT_MAPPED)) {
+			vcn = rl->vcn;
+			goto map_vcn;
+		}
+		/* If this run is not valid abort with an error. */
+		if (unlikely(!rl->length || lcn < LCN_HOLE))
+			goto rl_err;
+		/* Skip holes. */
+		if (lcn == LCN_HOLE)
+			continue;
+		block = lcn << vol->cluster_size_bits >> block_size_bits;
+		len = rl->length;
+		if (rl[1].vcn > end_vcn)
+			len = end_vcn - rl->vcn;
+		end_block = (lcn + len) << vol->cluster_size_bits >>
+				block_size_bits;
+		/* Iterate over the blocks in the run and empty them. */
+		do {
+			struct buffer_head *bh;
+
+			/* Obtain the buffer, possibly not uptodate. */
+			bh = sb_getblk(sb, block);
+			BUG_ON(!bh);
+			/* Setup buffer i/o submission. */
+			lock_buffer(bh);
+			bh->b_end_io = end_buffer_write_sync;
+			get_bh(bh);
+			/* Set the entire contents of the buffer to 0xff. */
+			memset(bh->b_data, -1, block_size);
+			if (!buffer_uptodate(bh))
+				set_buffer_uptodate(bh);
+			if (buffer_dirty(bh))
+				clear_buffer_dirty(bh);
+			/*
+			 * Submit the buffer and wait for i/o to complete but
+			 * only for the first buffer so we do not miss really
+			 * serious i/o errors.  Once the first buffer has
+			 * completed ignore errors afterwards as we can assume
+			 * that if one buffer worked all of them will work.
+			 */
+			submit_bh(REQ_OP_WRITE, 0, bh);
+			if (should_wait) {
+				should_wait = false;
+				wait_on_buffer(bh);
+				if (unlikely(!buffer_uptodate(bh)))
+					goto io_err;
+			}
+			brelse(bh);
+		} while (++block < end_block);
+	} while ((++rl)->vcn < end_vcn);
+	up_write(&log_ni->runlist.lock);
+	/*
+	 * Zap the pages again just in case any got instantiated whilst we were
+	 * emptying the blocks by hand.  FIXME: We may not have completed
+	 * writing to all the buffer heads yet so this may happen too early.
+	 * We really should use a kernel thread to do the emptying
+	 * asynchronously and then we can also set the volume dirty and output
+	 * an error message if emptying should fail.
+	 */
+	truncate_inode_pages(log_vi->i_mapping, 0);
+	/* Set the flag so we do not have to do it again on remount. */
+	NVolSetLogFileEmpty(vol);
+	ntfs_debug("Done.");
+	return true;
+io_err:
+	ntfs_error(sb, "Failed to write buffer.  Unmount and run chkdsk.");
+	goto dirty_err;
+rl_err:
+	ntfs_error(sb, "Runlist is corrupt.  Unmount and run chkdsk.");
+dirty_err:
+	NVolSetErrors(vol);
+	err = -EIO;
+err:
+	up_write(&log_ni->runlist.lock);
+	ntfs_error(sb, "Failed to fill $LogFile with 0xff bytes (error %d).",
+			-err);
+	return false;
+}
+
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/logfile.h b/fs/ntfs/logfile.h
new file mode 100644
index 000000000..429d4909c
--- /dev/null
+++ b/fs/ntfs/logfile.h
@@ -0,0 +1,295 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * logfile.h - Defines for NTFS kernel journal ($LogFile) handling.  Part of
+ *	       the Linux-NTFS project.
+ *
+ * Copyright (c) 2000-2005 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_LOGFILE_H
+#define _LINUX_NTFS_LOGFILE_H
+
+#ifdef NTFS_RW
+
+#include <linux/fs.h>
+
+#include "types.h"
+#include "endian.h"
+#include "layout.h"
+
+/*
+ * Journal ($LogFile) organization:
+ *
+ * Two restart areas present in the first two pages (restart pages, one restart
+ * area in each page).  When the volume is dismounted they should be identical,
+ * except for the update sequence array which usually has a different update
+ * sequence number.
+ *
+ * These are followed by log records organized in pages headed by a log record
+ * header going up to log file size.  Not all pages contain log records when a
+ * volume is first formatted, but as the volume ages, all records will be used.
+ * When the log file fills up, the records at the beginning are purged (by
+ * modifying the oldest_lsn to a higher value presumably) and writing begins
+ * at the beginning of the file.  Effectively, the log file is viewed as a
+ * circular entity.
+ *
+ * NOTE: Windows NT, 2000, and XP all use log file version 1.1 but they accept
+ * versions <= 1.x, including 0.-1.  (Yes, that is a minus one in there!)  We
+ * probably only want to support 1.1 as this seems to be the current version
+ * and we don't know how that differs from the older versions.  The only
+ * exception is if the journal is clean as marked by the two restart pages
+ * then it doesn't matter whether we are on an earlier version.  We can just
+ * reinitialize the logfile and start again with version 1.1.
+ */
+
+/* Some $LogFile related constants. */
+#define MaxLogFileSize		0x100000000ULL
+#define DefaultLogPageSize	4096
+#define MinLogRecordPages	48
+
+/*
+ * Log file restart page header (begins the restart area).
+ */
+typedef struct {
+/*Ofs*/
+/*  0	NTFS_RECORD; -- Unfolded here as gcc doesn't like unnamed structs. */
+/*  0*/	NTFS_RECORD_TYPE magic;	/* The magic is "RSTR". */
+/*  4*/	le16 usa_ofs;		/* See NTFS_RECORD definition in layout.h.
+				   When creating, set this to be immediately
+				   after this header structure (without any
+				   alignment). */
+/*  6*/	le16 usa_count;		/* See NTFS_RECORD definition in layout.h. */
+
+/*  8*/	leLSN chkdsk_lsn;	/* The last log file sequence number found by
+				   chkdsk.  Only used when the magic is changed
+				   to "CHKD".  Otherwise this is zero. */
+/* 16*/	le32 system_page_size;	/* Byte size of system pages when the log file
+				   was created, has to be >= 512 and a power of
+				   2.  Use this to calculate the required size
+				   of the usa (usa_count) and add it to usa_ofs.
+				   Then verify that the result is less than the
+				   value of the restart_area_offset. */
+/* 20*/	le32 log_page_size;	/* Byte size of log file pages, has to be >=
+				   512 and a power of 2.  The default is 4096
+				   and is used when the system page size is
+				   between 4096 and 8192.  Otherwise this is
+				   set to the system page size instead. */
+/* 24*/	le16 restart_area_offset;/* Byte offset from the start of this header to
+				   the RESTART_AREA.  Value has to be aligned
+				   to 8-byte boundary.  When creating, set this
+				   to be after the usa. */
+/* 26*/	sle16 minor_ver;	/* Log file minor version.  Only check if major
+				   version is 1. */
+/* 28*/	sle16 major_ver;	/* Log file major version.  We only support
+				   version 1.1. */
+/* sizeof() = 30 (0x1e) bytes */
+} __attribute__ ((__packed__)) RESTART_PAGE_HEADER;
+
+/*
+ * Constant for the log client indices meaning that there are no client records
+ * in this particular client array.  Also inside the client records themselves,
+ * this means that there are no client records preceding or following this one.
+ */
+#define LOGFILE_NO_CLIENT	cpu_to_le16(0xffff)
+#define LOGFILE_NO_CLIENT_CPU	0xffff
+
+/*
+ * These are the so far known RESTART_AREA_* flags (16-bit) which contain
+ * information about the log file in which they are present.
+ */
+enum {
+	RESTART_VOLUME_IS_CLEAN	= cpu_to_le16(0x0002),
+	RESTART_SPACE_FILLER	= cpu_to_le16(0xffff), /* gcc: Force enum bit width to 16. */
+} __attribute__ ((__packed__));
+
+typedef le16 RESTART_AREA_FLAGS;
+
+/*
+ * Log file restart area record.  The offset of this record is found by adding
+ * the offset of the RESTART_PAGE_HEADER to the restart_area_offset value found
+ * in it.  See notes at restart_area_offset above.
+ */
+typedef struct {
+/*Ofs*/
+/*  0*/	leLSN current_lsn;	/* The current, i.e. last LSN inside the log
+				   when the restart area was last written.
+				   This happens often but what is the interval?
+				   Is it just fixed time or is it every time a
+				   check point is written or somethine else?
+				   On create set to 0. */
+/*  8*/	le16 log_clients;	/* Number of log client records in the array of
+				   log client records which follows this
+				   restart area.  Must be 1.  */
+/* 10*/	le16 client_free_list;	/* The index of the first free log client record
+				   in the array of log client records.
+				   LOGFILE_NO_CLIENT means that there are no
+				   free log client records in the array.
+				   If != LOGFILE_NO_CLIENT, check that
+				   log_clients > client_free_list.  On Win2k
+				   and presumably earlier, on a clean volume
+				   this is != LOGFILE_NO_CLIENT, and it should
+				   be 0, i.e. the first (and only) client
+				   record is free and thus the logfile is
+				   closed and hence clean.  A dirty volume
+				   would have left the logfile open and hence
+				   this would be LOGFILE_NO_CLIENT.  On WinXP
+				   and presumably later, the logfile is always
+				   open, even on clean shutdown so this should
+				   always be LOGFILE_NO_CLIENT. */
+/* 12*/	le16 client_in_use_list;/* The index of the first in-use log client
+				   record in the array of log client records.
+				   LOGFILE_NO_CLIENT means that there are no
+				   in-use log client records in the array.  If
+				   != LOGFILE_NO_CLIENT check that log_clients
+				   > client_in_use_list.  On Win2k and
+				   presumably earlier, on a clean volume this
+				   is LOGFILE_NO_CLIENT, i.e. there are no
+				   client records in use and thus the logfile
+				   is closed and hence clean.  A dirty volume
+				   would have left the logfile open and hence
+				   this would be != LOGFILE_NO_CLIENT, and it
+				   should be 0, i.e. the first (and only)
+				   client record is in use.  On WinXP and
+				   presumably later, the logfile is always
+				   open, even on clean shutdown so this should
+				   always be 0. */
+/* 14*/	RESTART_AREA_FLAGS flags;/* Flags modifying LFS behaviour.  On Win2k
+				   and presumably earlier this is always 0.  On
+				   WinXP and presumably later, if the logfile
+				   was shutdown cleanly, the second bit,
+				   RESTART_VOLUME_IS_CLEAN, is set.  This bit
+				   is cleared when the volume is mounted by
+				   WinXP and set when the volume is dismounted,
+				   thus if the logfile is dirty, this bit is
+				   clear.  Thus we don't need to check the
+				   Windows version to determine if the logfile
+				   is clean.  Instead if the logfile is closed,
+				   we know it must be clean.  If it is open and
+				   this bit is set, we also know it must be
+				   clean.  If on the other hand the logfile is
+				   open and this bit is clear, we can be almost
+				   certain that the logfile is dirty. */
+/* 16*/	le32 seq_number_bits;	/* How many bits to use for the sequence
+				   number.  This is calculated as 67 - the
+				   number of bits required to store the logfile
+				   size in bytes and this can be used in with
+				   the specified file_size as a consistency
+				   check. */
+/* 20*/	le16 restart_area_length;/* Length of the restart area including the
+				   client array.  Following checks required if
+				   version matches.  Otherwise, skip them.
+				   restart_area_offset + restart_area_length
+				   has to be <= system_page_size.  Also,
+				   restart_area_length has to be >=
+				   client_array_offset + (log_clients *
+				   sizeof(log client record)). */
+/* 22*/	le16 client_array_offset;/* Offset from the start of this record to
+				   the first log client record if versions are
+				   matched.  When creating, set this to be
+				   after this restart area structure, aligned
+				   to 8-bytes boundary.  If the versions do not
+				   match, this is ignored and the offset is
+				   assumed to be (sizeof(RESTART_AREA) + 7) &
+				   ~7, i.e. rounded up to first 8-byte
+				   boundary.  Either way, client_array_offset
+				   has to be aligned to an 8-byte boundary.
+				   Also, restart_area_offset +
+				   client_array_offset has to be <= 510.
+				   Finally, client_array_offset + (log_clients
+				   * sizeof(log client record)) has to be <=
+				   system_page_size.  On Win2k and presumably
+				   earlier, this is 0x30, i.e. immediately
+				   following this record.  On WinXP and
+				   presumably later, this is 0x40, i.e. there
+				   are 16 extra bytes between this record and
+				   the client array.  This probably means that
+				   the RESTART_AREA record is actually bigger
+				   in WinXP and later. */
+/* 24*/	sle64 file_size;	/* Usable byte size of the log file.  If the
+				   restart_area_offset + the offset of the
+				   file_size are > 510 then corruption has
+				   occurred.  This is the very first check when
+				   starting with the restart_area as if it
+				   fails it means that some of the above values
+				   will be corrupted by the multi sector
+				   transfer protection.  The file_size has to
+				   be rounded down to be a multiple of the
+				   log_page_size in the RESTART_PAGE_HEADER and
+				   then it has to be at least big enough to
+				   store the two restart pages and 48 (0x30)
+				   log record pages. */
+/* 32*/	le32 last_lsn_data_length;/* Length of data of last LSN, not including
+				   the log record header.  On create set to
+				   0. */
+/* 36*/	le16 log_record_header_length;/* Byte size of the log record header.
+				   If the version matches then check that the
+				   value of log_record_header_length is a
+				   multiple of 8, i.e.
+				   (log_record_header_length + 7) & ~7 ==
+				   log_record_header_length.  When creating set
+				   it to sizeof(LOG_RECORD_HEADER), aligned to
+				   8 bytes. */
+/* 38*/	le16 log_page_data_offset;/* Offset to the start of data in a log record
+				   page.  Must be a multiple of 8.  On create
+				   set it to immediately after the update
+				   sequence array of the log record page. */
+/* 40*/	le32 restart_log_open_count;/* A counter that gets incremented every
+				   time the logfile is restarted which happens
+				   at mount time when the logfile is opened.
+				   When creating set to a random value.  Win2k
+				   sets it to the low 32 bits of the current
+				   system time in NTFS format (see time.h). */
+/* 44*/	le32 reserved;		/* Reserved/alignment to 8-byte boundary. */
+/* sizeof() = 48 (0x30) bytes */
+} __attribute__ ((__packed__)) RESTART_AREA;
+
+/*
+ * Log client record.  The offset of this record is found by adding the offset
+ * of the RESTART_AREA to the client_array_offset value found in it.
+ */
+typedef struct {
+/*Ofs*/
+/*  0*/	leLSN oldest_lsn;	/* Oldest LSN needed by this client.  On create
+				   set to 0. */
+/*  8*/	leLSN client_restart_lsn;/* LSN at which this client needs to restart
+				   the volume, i.e. the current position within
+				   the log file.  At present, if clean this
+				   should = current_lsn in restart area but it
+				   probably also = current_lsn when dirty most
+				   of the time.  At create set to 0. */
+/* 16*/	le16 prev_client;	/* The offset to the previous log client record
+				   in the array of log client records.
+				   LOGFILE_NO_CLIENT means there is no previous
+				   client record, i.e. this is the first one.
+				   This is always LOGFILE_NO_CLIENT. */
+/* 18*/	le16 next_client;	/* The offset to the next log client record in
+				   the array of log client records.
+				   LOGFILE_NO_CLIENT means there are no next
+				   client records, i.e. this is the last one.
+				   This is always LOGFILE_NO_CLIENT. */
+/* 20*/	le16 seq_number;	/* On Win2k and presumably earlier, this is set
+				   to zero every time the logfile is restarted
+				   and it is incremented when the logfile is
+				   closed at dismount time.  Thus it is 0 when
+				   dirty and 1 when clean.  On WinXP and
+				   presumably later, this is always 0. */
+/* 22*/	u8 reserved[6];		/* Reserved/alignment. */
+/* 28*/	le32 client_name_length;/* Length of client name in bytes.  Should
+				   always be 8. */
+/* 32*/	ntfschar client_name[64];/* Name of the client in Unicode.  Should
+				   always be "NTFS" with the remaining bytes
+				   set to 0. */
+/* sizeof() = 160 (0xa0) bytes */
+} __attribute__ ((__packed__)) LOG_CLIENT_RECORD;
+
+extern bool ntfs_check_logfile(struct inode *log_vi,
+		RESTART_PAGE_HEADER **rp);
+
+extern bool ntfs_is_logfile_clean(struct inode *log_vi,
+		const RESTART_PAGE_HEADER *rp);
+
+extern bool ntfs_empty_logfile(struct inode *log_vi);
+
+#endif /* NTFS_RW */
+
+#endif /* _LINUX_NTFS_LOGFILE_H */
diff --git a/fs/ntfs/malloc.h b/fs/ntfs/malloc.h
new file mode 100644
index 000000000..706842573
--- /dev/null
+++ b/fs/ntfs/malloc.h
@@ -0,0 +1,77 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * malloc.h - NTFS kernel memory handling. Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2005 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_MALLOC_H
+#define _LINUX_NTFS_MALLOC_H
+
+#include <linux/vmalloc.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+
+/**
+ * __ntfs_malloc - allocate memory in multiples of pages
+ * @size:	number of bytes to allocate
+ * @gfp_mask:	extra flags for the allocator
+ *
+ * Internal function.  You probably want ntfs_malloc_nofs()...
+ *
+ * Allocates @size bytes of memory, rounded up to multiples of PAGE_SIZE and
+ * returns a pointer to the allocated memory.
+ *
+ * If there was insufficient memory to complete the request, return NULL.
+ * Depending on @gfp_mask the allocation may be guaranteed to succeed.
+ */
+static inline void *__ntfs_malloc(unsigned long size, gfp_t gfp_mask)
+{
+	if (likely(size <= PAGE_SIZE)) {
+		BUG_ON(!size);
+		/* kmalloc() has per-CPU caches so is faster for now. */
+		return kmalloc(PAGE_SIZE, gfp_mask & ~__GFP_HIGHMEM);
+		/* return (void *)__get_free_page(gfp_mask); */
+	}
+	if (likely((size >> PAGE_SHIFT) < totalram_pages()))
+		return __vmalloc(size, gfp_mask);
+	return NULL;
+}
+
+/**
+ * ntfs_malloc_nofs - allocate memory in multiples of pages
+ * @size:	number of bytes to allocate
+ *
+ * Allocates @size bytes of memory, rounded up to multiples of PAGE_SIZE and
+ * returns a pointer to the allocated memory.
+ *
+ * If there was insufficient memory to complete the request, return NULL.
+ */
+static inline void *ntfs_malloc_nofs(unsigned long size)
+{
+	return __ntfs_malloc(size, GFP_NOFS | __GFP_HIGHMEM);
+}
+
+/**
+ * ntfs_malloc_nofs_nofail - allocate memory in multiples of pages
+ * @size:	number of bytes to allocate
+ *
+ * Allocates @size bytes of memory, rounded up to multiples of PAGE_SIZE and
+ * returns a pointer to the allocated memory.
+ *
+ * This function guarantees that the allocation will succeed.  It will sleep
+ * for as long as it takes to complete the allocation.
+ *
+ * If there was insufficient memory to complete the request, return NULL.
+ */
+static inline void *ntfs_malloc_nofs_nofail(unsigned long size)
+{
+	return __ntfs_malloc(size, GFP_NOFS | __GFP_HIGHMEM | __GFP_NOFAIL);
+}
+
+static inline void ntfs_free(void *addr)
+{
+	kvfree(addr);
+}
+
+#endif /* _LINUX_NTFS_MALLOC_H */
diff --git a/fs/ntfs/mft.c b/fs/ntfs/mft.c
new file mode 100644
index 000000000..0d62cd5bb
--- /dev/null
+++ b/fs/ntfs/mft.c
@@ -0,0 +1,2906 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/**
+ * mft.c - NTFS kernel mft record operations. Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2012 Anton Altaparmakov and Tuxera Inc.
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#include <linux/buffer_head.h>
+#include <linux/slab.h>
+#include <linux/swap.h>
+#include <linux/bio.h>
+
+#include "attrib.h"
+#include "aops.h"
+#include "bitmap.h"
+#include "debug.h"
+#include "dir.h"
+#include "lcnalloc.h"
+#include "malloc.h"
+#include "mft.h"
+#include "ntfs.h"
+
+#define MAX_BHS	(PAGE_SIZE / NTFS_BLOCK_SIZE)
+
+/**
+ * map_mft_record_page - map the page in which a specific mft record resides
+ * @ni:		ntfs inode whose mft record page to map
+ *
+ * This maps the page in which the mft record of the ntfs inode @ni is situated
+ * and returns a pointer to the mft record within the mapped page.
+ *
+ * Return value needs to be checked with IS_ERR() and if that is true PTR_ERR()
+ * contains the negative error code returned.
+ */
+static inline MFT_RECORD *map_mft_record_page(ntfs_inode *ni)
+{
+	loff_t i_size;
+	ntfs_volume *vol = ni->vol;
+	struct inode *mft_vi = vol->mft_ino;
+	struct page *page;
+	unsigned long index, end_index;
+	unsigned ofs;
+
+	BUG_ON(ni->page);
+	/*
+	 * The index into the page cache and the offset within the page cache
+	 * page of the wanted mft record. FIXME: We need to check for
+	 * overflowing the unsigned long, but I don't think we would ever get
+	 * here if the volume was that big...
+	 */
+	index = (u64)ni->mft_no << vol->mft_record_size_bits >>
+			PAGE_SHIFT;
+	ofs = (ni->mft_no << vol->mft_record_size_bits) & ~PAGE_MASK;
+
+	i_size = i_size_read(mft_vi);
+	/* The maximum valid index into the page cache for $MFT's data. */
+	end_index = i_size >> PAGE_SHIFT;
+
+	/* If the wanted index is out of bounds the mft record doesn't exist. */
+	if (unlikely(index >= end_index)) {
+		if (index > end_index || (i_size & ~PAGE_MASK) < ofs +
+				vol->mft_record_size) {
+			page = ERR_PTR(-ENOENT);
+			ntfs_error(vol->sb, "Attempt to read mft record 0x%lx, "
+					"which is beyond the end of the mft.  "
+					"This is probably a bug in the ntfs "
+					"driver.", ni->mft_no);
+			goto err_out;
+		}
+	}
+	/* Read, map, and pin the page. */
+	page = ntfs_map_page(mft_vi->i_mapping, index);
+	if (!IS_ERR(page)) {
+		/* Catch multi sector transfer fixup errors. */
+		if (likely(ntfs_is_mft_recordp((le32*)(page_address(page) +
+				ofs)))) {
+			ni->page = page;
+			ni->page_ofs = ofs;
+			return page_address(page) + ofs;
+		}
+		ntfs_error(vol->sb, "Mft record 0x%lx is corrupt.  "
+				"Run chkdsk.", ni->mft_no);
+		ntfs_unmap_page(page);
+		page = ERR_PTR(-EIO);
+		NVolSetErrors(vol);
+	}
+err_out:
+	ni->page = NULL;
+	ni->page_ofs = 0;
+	return (void*)page;
+}
+
+/**
+ * map_mft_record - map, pin and lock an mft record
+ * @ni:		ntfs inode whose MFT record to map
+ *
+ * First, take the mrec_lock mutex.  We might now be sleeping, while waiting
+ * for the mutex if it was already locked by someone else.
+ *
+ * The page of the record is mapped using map_mft_record_page() before being
+ * returned to the caller.
+ *
+ * This in turn uses ntfs_map_page() to get the page containing the wanted mft
+ * record (it in turn calls read_cache_page() which reads it in from disk if
+ * necessary, increments the use count on the page so that it cannot disappear
+ * under us and returns a reference to the page cache page).
+ *
+ * If read_cache_page() invokes ntfs_readpage() to load the page from disk, it
+ * sets PG_locked and clears PG_uptodate on the page. Once I/O has completed
+ * and the post-read mst fixups on each mft record in the page have been
+ * performed, the page gets PG_uptodate set and PG_locked cleared (this is done
+ * in our asynchronous I/O completion handler end_buffer_read_mft_async()).
+ * ntfs_map_page() waits for PG_locked to become clear and checks if
+ * PG_uptodate is set and returns an error code if not. This provides
+ * sufficient protection against races when reading/using the page.
+ *
+ * However there is the write mapping to think about. Doing the above described
+ * checking here will be fine, because when initiating the write we will set
+ * PG_locked and clear PG_uptodate making sure nobody is touching the page
+ * contents. Doing the locking this way means that the commit to disk code in
+ * the page cache code paths is automatically sufficiently locked with us as
+ * we will not touch a page that has been locked or is not uptodate. The only
+ * locking problem then is them locking the page while we are accessing it.
+ *
+ * So that code will end up having to own the mrec_lock of all mft
+ * records/inodes present in the page before I/O can proceed. In that case we
+ * wouldn't need to bother with PG_locked and PG_uptodate as nobody will be
+ * accessing anything without owning the mrec_lock mutex.  But we do need to
+ * use them because of the read_cache_page() invocation and the code becomes so
+ * much simpler this way that it is well worth it.
+ *
+ * The mft record is now ours and we return a pointer to it. You need to check
+ * the returned pointer with IS_ERR() and if that is true, PTR_ERR() will return
+ * the error code.
+ *
+ * NOTE: Caller is responsible for setting the mft record dirty before calling
+ * unmap_mft_record(). This is obviously only necessary if the caller really
+ * modified the mft record...
+ * Q: Do we want to recycle one of the VFS inode state bits instead?
+ * A: No, the inode ones mean we want to change the mft record, not we want to
+ * write it out.
+ */
+MFT_RECORD *map_mft_record(ntfs_inode *ni)
+{
+	MFT_RECORD *m;
+
+	ntfs_debug("Entering for mft_no 0x%lx.", ni->mft_no);
+
+	/* Make sure the ntfs inode doesn't go away. */
+	atomic_inc(&ni->count);
+
+	/* Serialize access to this mft record. */
+	mutex_lock(&ni->mrec_lock);
+
+	m = map_mft_record_page(ni);
+	if (!IS_ERR(m))
+		return m;
+
+	mutex_unlock(&ni->mrec_lock);
+	atomic_dec(&ni->count);
+	ntfs_error(ni->vol->sb, "Failed with error code %lu.", -PTR_ERR(m));
+	return m;
+}
+
+/**
+ * unmap_mft_record_page - unmap the page in which a specific mft record resides
+ * @ni:		ntfs inode whose mft record page to unmap
+ *
+ * This unmaps the page in which the mft record of the ntfs inode @ni is
+ * situated and returns. This is a NOOP if highmem is not configured.
+ *
+ * The unmap happens via ntfs_unmap_page() which in turn decrements the use
+ * count on the page thus releasing it from the pinned state.
+ *
+ * We do not actually unmap the page from memory of course, as that will be
+ * done by the page cache code itself when memory pressure increases or
+ * whatever.
+ */
+static inline void unmap_mft_record_page(ntfs_inode *ni)
+{
+	BUG_ON(!ni->page);
+
+	// TODO: If dirty, blah...
+	ntfs_unmap_page(ni->page);
+	ni->page = NULL;
+	ni->page_ofs = 0;
+	return;
+}
+
+/**
+ * unmap_mft_record - release a mapped mft record
+ * @ni:		ntfs inode whose MFT record to unmap
+ *
+ * We release the page mapping and the mrec_lock mutex which unmaps the mft
+ * record and releases it for others to get hold of. We also release the ntfs
+ * inode by decrementing the ntfs inode reference count.
+ *
+ * NOTE: If caller has modified the mft record, it is imperative to set the mft
+ * record dirty BEFORE calling unmap_mft_record().
+ */
+void unmap_mft_record(ntfs_inode *ni)
+{
+	struct page *page = ni->page;
+
+	BUG_ON(!page);
+
+	ntfs_debug("Entering for mft_no 0x%lx.", ni->mft_no);
+
+	unmap_mft_record_page(ni);
+	mutex_unlock(&ni->mrec_lock);
+	atomic_dec(&ni->count);
+	/*
+	 * If pure ntfs_inode, i.e. no vfs inode attached, we leave it to
+	 * ntfs_clear_extent_inode() in the extent inode case, and to the
+	 * caller in the non-extent, yet pure ntfs inode case, to do the actual
+	 * tear down of all structures and freeing of all allocated memory.
+	 */
+	return;
+}
+
+/**
+ * map_extent_mft_record - load an extent inode and attach it to its base
+ * @base_ni:	base ntfs inode
+ * @mref:	mft reference of the extent inode to load
+ * @ntfs_ino:	on successful return, pointer to the ntfs_inode structure
+ *
+ * Load the extent mft record @mref and attach it to its base inode @base_ni.
+ * Return the mapped extent mft record if IS_ERR(result) is false.  Otherwise
+ * PTR_ERR(result) gives the negative error code.
+ *
+ * On successful return, @ntfs_ino contains a pointer to the ntfs_inode
+ * structure of the mapped extent inode.
+ */
+MFT_RECORD *map_extent_mft_record(ntfs_inode *base_ni, MFT_REF mref,
+		ntfs_inode **ntfs_ino)
+{
+	MFT_RECORD *m;
+	ntfs_inode *ni = NULL;
+	ntfs_inode **extent_nis = NULL;
+	int i;
+	unsigned long mft_no = MREF(mref);
+	u16 seq_no = MSEQNO(mref);
+	bool destroy_ni = false;
+
+	ntfs_debug("Mapping extent mft record 0x%lx (base mft record 0x%lx).",
+			mft_no, base_ni->mft_no);
+	/* Make sure the base ntfs inode doesn't go away. */
+	atomic_inc(&base_ni->count);
+	/*
+	 * Check if this extent inode has already been added to the base inode,
+	 * in which case just return it. If not found, add it to the base
+	 * inode before returning it.
+	 */
+	mutex_lock(&base_ni->extent_lock);
+	if (base_ni->nr_extents > 0) {
+		extent_nis = base_ni->ext.extent_ntfs_inos;
+		for (i = 0; i < base_ni->nr_extents; i++) {
+			if (mft_no != extent_nis[i]->mft_no)
+				continue;
+			ni = extent_nis[i];
+			/* Make sure the ntfs inode doesn't go away. */
+			atomic_inc(&ni->count);
+			break;
+		}
+	}
+	if (likely(ni != NULL)) {
+		mutex_unlock(&base_ni->extent_lock);
+		atomic_dec(&base_ni->count);
+		/* We found the record; just have to map and return it. */
+		m = map_mft_record(ni);
+		/* map_mft_record() has incremented this on success. */
+		atomic_dec(&ni->count);
+		if (!IS_ERR(m)) {
+			/* Verify the sequence number. */
+			if (likely(le16_to_cpu(m->sequence_number) == seq_no)) {
+				ntfs_debug("Done 1.");
+				*ntfs_ino = ni;
+				return m;
+			}
+			unmap_mft_record(ni);
+			ntfs_error(base_ni->vol->sb, "Found stale extent mft "
+					"reference! Corrupt filesystem. "
+					"Run chkdsk.");
+			return ERR_PTR(-EIO);
+		}
+map_err_out:
+		ntfs_error(base_ni->vol->sb, "Failed to map extent "
+				"mft record, error code %ld.", -PTR_ERR(m));
+		return m;
+	}
+	/* Record wasn't there. Get a new ntfs inode and initialize it. */
+	ni = ntfs_new_extent_inode(base_ni->vol->sb, mft_no);
+	if (unlikely(!ni)) {
+		mutex_unlock(&base_ni->extent_lock);
+		atomic_dec(&base_ni->count);
+		return ERR_PTR(-ENOMEM);
+	}
+	ni->vol = base_ni->vol;
+	ni->seq_no = seq_no;
+	ni->nr_extents = -1;
+	ni->ext.base_ntfs_ino = base_ni;
+	/* Now map the record. */
+	m = map_mft_record(ni);
+	if (IS_ERR(m)) {
+		mutex_unlock(&base_ni->extent_lock);
+		atomic_dec(&base_ni->count);
+		ntfs_clear_extent_inode(ni);
+		goto map_err_out;
+	}
+	/* Verify the sequence number if it is present. */
+	if (seq_no && (le16_to_cpu(m->sequence_number) != seq_no)) {
+		ntfs_error(base_ni->vol->sb, "Found stale extent mft "
+				"reference! Corrupt filesystem. Run chkdsk.");
+		destroy_ni = true;
+		m = ERR_PTR(-EIO);
+		goto unm_err_out;
+	}
+	/* Attach extent inode to base inode, reallocating memory if needed. */
+	if (!(base_ni->nr_extents & 3)) {
+		ntfs_inode **tmp;
+		int new_size = (base_ni->nr_extents + 4) * sizeof(ntfs_inode *);
+
+		tmp = kmalloc(new_size, GFP_NOFS);
+		if (unlikely(!tmp)) {
+			ntfs_error(base_ni->vol->sb, "Failed to allocate "
+					"internal buffer.");
+			destroy_ni = true;
+			m = ERR_PTR(-ENOMEM);
+			goto unm_err_out;
+		}
+		if (base_ni->nr_extents) {
+			BUG_ON(!base_ni->ext.extent_ntfs_inos);
+			memcpy(tmp, base_ni->ext.extent_ntfs_inos, new_size -
+					4 * sizeof(ntfs_inode *));
+			kfree(base_ni->ext.extent_ntfs_inos);
+		}
+		base_ni->ext.extent_ntfs_inos = tmp;
+	}
+	base_ni->ext.extent_ntfs_inos[base_ni->nr_extents++] = ni;
+	mutex_unlock(&base_ni->extent_lock);
+	atomic_dec(&base_ni->count);
+	ntfs_debug("Done 2.");
+	*ntfs_ino = ni;
+	return m;
+unm_err_out:
+	unmap_mft_record(ni);
+	mutex_unlock(&base_ni->extent_lock);
+	atomic_dec(&base_ni->count);
+	/*
+	 * If the extent inode was not attached to the base inode we need to
+	 * release it or we will leak memory.
+	 */
+	if (destroy_ni)
+		ntfs_clear_extent_inode(ni);
+	return m;
+}
+
+#ifdef NTFS_RW
+
+/**
+ * __mark_mft_record_dirty - set the mft record and the page containing it dirty
+ * @ni:		ntfs inode describing the mapped mft record
+ *
+ * Internal function.  Users should call mark_mft_record_dirty() instead.
+ *
+ * Set the mapped (extent) mft record of the (base or extent) ntfs inode @ni,
+ * as well as the page containing the mft record, dirty.  Also, mark the base
+ * vfs inode dirty.  This ensures that any changes to the mft record are
+ * written out to disk.
+ *
+ * NOTE:  We only set I_DIRTY_DATASYNC (and not I_DIRTY_PAGES)
+ * on the base vfs inode, because even though file data may have been modified,
+ * it is dirty in the inode meta data rather than the data page cache of the
+ * inode, and thus there are no data pages that need writing out.  Therefore, a
+ * full mark_inode_dirty() is overkill.  A mark_inode_dirty_sync(), on the
+ * other hand, is not sufficient, because ->write_inode needs to be called even
+ * in case of fdatasync. This needs to happen or the file data would not
+ * necessarily hit the device synchronously, even though the vfs inode has the
+ * O_SYNC flag set.  Also, I_DIRTY_DATASYNC simply "feels" better than just
+ * I_DIRTY_SYNC, since the file data has not actually hit the block device yet,
+ * which is not what I_DIRTY_SYNC on its own would suggest.
+ */
+void __mark_mft_record_dirty(ntfs_inode *ni)
+{
+	ntfs_inode *base_ni;
+
+	ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
+	BUG_ON(NInoAttr(ni));
+	mark_ntfs_record_dirty(ni->page, ni->page_ofs);
+	/* Determine the base vfs inode and mark it dirty, too. */
+	mutex_lock(&ni->extent_lock);
+	if (likely(ni->nr_extents >= 0))
+		base_ni = ni;
+	else
+		base_ni = ni->ext.base_ntfs_ino;
+	mutex_unlock(&ni->extent_lock);
+	__mark_inode_dirty(VFS_I(base_ni), I_DIRTY_DATASYNC);
+}
+
+static const char *ntfs_please_email = "Please email "
+		"linux-ntfs-dev@lists.sourceforge.net and say that you saw "
+		"this message.  Thank you.";
+
+/**
+ * ntfs_sync_mft_mirror_umount - synchronise an mft record to the mft mirror
+ * @vol:	ntfs volume on which the mft record to synchronize resides
+ * @mft_no:	mft record number of mft record to synchronize
+ * @m:		mapped, mst protected (extent) mft record to synchronize
+ *
+ * Write the mapped, mst protected (extent) mft record @m with mft record
+ * number @mft_no to the mft mirror ($MFTMirr) of the ntfs volume @vol,
+ * bypassing the page cache and the $MFTMirr inode itself.
+ *
+ * This function is only for use at umount time when the mft mirror inode has
+ * already been disposed off.  We BUG() if we are called while the mft mirror
+ * inode is still attached to the volume.
+ *
+ * On success return 0.  On error return -errno.
+ *
+ * NOTE:  This function is not implemented yet as I am not convinced it can
+ * actually be triggered considering the sequence of commits we do in super.c::
+ * ntfs_put_super().  But just in case we provide this place holder as the
+ * alternative would be either to BUG() or to get a NULL pointer dereference
+ * and Oops.
+ */
+static int ntfs_sync_mft_mirror_umount(ntfs_volume *vol,
+		const unsigned long mft_no, MFT_RECORD *m)
+{
+	BUG_ON(vol->mftmirr_ino);
+	ntfs_error(vol->sb, "Umount time mft mirror syncing is not "
+			"implemented yet.  %s", ntfs_please_email);
+	return -EOPNOTSUPP;
+}
+
+/**
+ * ntfs_sync_mft_mirror - synchronize an mft record to the mft mirror
+ * @vol:	ntfs volume on which the mft record to synchronize resides
+ * @mft_no:	mft record number of mft record to synchronize
+ * @m:		mapped, mst protected (extent) mft record to synchronize
+ * @sync:	if true, wait for i/o completion
+ *
+ * Write the mapped, mst protected (extent) mft record @m with mft record
+ * number @mft_no to the mft mirror ($MFTMirr) of the ntfs volume @vol.
+ *
+ * On success return 0.  On error return -errno and set the volume errors flag
+ * in the ntfs volume @vol.
+ *
+ * NOTE:  We always perform synchronous i/o and ignore the @sync parameter.
+ *
+ * TODO:  If @sync is false, want to do truly asynchronous i/o, i.e. just
+ * schedule i/o via ->writepage or do it via kntfsd or whatever.
+ */
+int ntfs_sync_mft_mirror(ntfs_volume *vol, const unsigned long mft_no,
+		MFT_RECORD *m, int sync)
+{
+	struct page *page;
+	unsigned int blocksize = vol->sb->s_blocksize;
+	int max_bhs = vol->mft_record_size / blocksize;
+	struct buffer_head *bhs[MAX_BHS];
+	struct buffer_head *bh, *head;
+	u8 *kmirr;
+	runlist_element *rl;
+	unsigned int block_start, block_end, m_start, m_end, page_ofs;
+	int i_bhs, nr_bhs, err = 0;
+	unsigned char blocksize_bits = vol->sb->s_blocksize_bits;
+
+	ntfs_debug("Entering for inode 0x%lx.", mft_no);
+	BUG_ON(!max_bhs);
+	if (WARN_ON(max_bhs > MAX_BHS))
+		return -EINVAL;
+	if (unlikely(!vol->mftmirr_ino)) {
+		/* This could happen during umount... */
+		err = ntfs_sync_mft_mirror_umount(vol, mft_no, m);
+		if (likely(!err))
+			return err;
+		goto err_out;
+	}
+	/* Get the page containing the mirror copy of the mft record @m. */
+	page = ntfs_map_page(vol->mftmirr_ino->i_mapping, mft_no >>
+			(PAGE_SHIFT - vol->mft_record_size_bits));
+	if (IS_ERR(page)) {
+		ntfs_error(vol->sb, "Failed to map mft mirror page.");
+		err = PTR_ERR(page);
+		goto err_out;
+	}
+	lock_page(page);
+	BUG_ON(!PageUptodate(page));
+	ClearPageUptodate(page);
+	/* Offset of the mft mirror record inside the page. */
+	page_ofs = (mft_no << vol->mft_record_size_bits) & ~PAGE_MASK;
+	/* The address in the page of the mirror copy of the mft record @m. */
+	kmirr = page_address(page) + page_ofs;
+	/* Copy the mst protected mft record to the mirror. */
+	memcpy(kmirr, m, vol->mft_record_size);
+	/* Create uptodate buffers if not present. */
+	if (unlikely(!page_has_buffers(page))) {
+		struct buffer_head *tail;
+
+		bh = head = alloc_page_buffers(page, blocksize, true);
+		do {
+			set_buffer_uptodate(bh);
+			tail = bh;
+			bh = bh->b_this_page;
+		} while (bh);
+		tail->b_this_page = head;
+		attach_page_private(page, head);
+	}
+	bh = head = page_buffers(page);
+	BUG_ON(!bh);
+	rl = NULL;
+	nr_bhs = 0;
+	block_start = 0;
+	m_start = kmirr - (u8*)page_address(page);
+	m_end = m_start + vol->mft_record_size;
+	do {
+		block_end = block_start + blocksize;
+		/* If the buffer is outside the mft record, skip it. */
+		if (block_end <= m_start)
+			continue;
+		if (unlikely(block_start >= m_end))
+			break;
+		/* Need to map the buffer if it is not mapped already. */
+		if (unlikely(!buffer_mapped(bh))) {
+			VCN vcn;
+			LCN lcn;
+			unsigned int vcn_ofs;
+
+			bh->b_bdev = vol->sb->s_bdev;
+			/* Obtain the vcn and offset of the current block. */
+			vcn = ((VCN)mft_no << vol->mft_record_size_bits) +
+					(block_start - m_start);
+			vcn_ofs = vcn & vol->cluster_size_mask;
+			vcn >>= vol->cluster_size_bits;
+			if (!rl) {
+				down_read(&NTFS_I(vol->mftmirr_ino)->
+						runlist.lock);
+				rl = NTFS_I(vol->mftmirr_ino)->runlist.rl;
+				/*
+				 * $MFTMirr always has the whole of its runlist
+				 * in memory.
+				 */
+				BUG_ON(!rl);
+			}
+			/* Seek to element containing target vcn. */
+			while (rl->length && rl[1].vcn <= vcn)
+				rl++;
+			lcn = ntfs_rl_vcn_to_lcn(rl, vcn);
+			/* For $MFTMirr, only lcn >= 0 is a successful remap. */
+			if (likely(lcn >= 0)) {
+				/* Setup buffer head to correct block. */
+				bh->b_blocknr = ((lcn <<
+						vol->cluster_size_bits) +
+						vcn_ofs) >> blocksize_bits;
+				set_buffer_mapped(bh);
+			} else {
+				bh->b_blocknr = -1;
+				ntfs_error(vol->sb, "Cannot write mft mirror "
+						"record 0x%lx because its "
+						"location on disk could not "
+						"be determined (error code "
+						"%lli).", mft_no,
+						(long long)lcn);
+				err = -EIO;
+			}
+		}
+		BUG_ON(!buffer_uptodate(bh));
+		BUG_ON(!nr_bhs && (m_start != block_start));
+		BUG_ON(nr_bhs >= max_bhs);
+		bhs[nr_bhs++] = bh;
+		BUG_ON((nr_bhs >= max_bhs) && (m_end != block_end));
+	} while (block_start = block_end, (bh = bh->b_this_page) != head);
+	if (unlikely(rl))
+		up_read(&NTFS_I(vol->mftmirr_ino)->runlist.lock);
+	if (likely(!err)) {
+		/* Lock buffers and start synchronous write i/o on them. */
+		for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++) {
+			struct buffer_head *tbh = bhs[i_bhs];
+
+			if (!trylock_buffer(tbh))
+				BUG();
+			BUG_ON(!buffer_uptodate(tbh));
+			clear_buffer_dirty(tbh);
+			get_bh(tbh);
+			tbh->b_end_io = end_buffer_write_sync;
+			submit_bh(REQ_OP_WRITE, 0, tbh);
+		}
+		/* Wait on i/o completion of buffers. */
+		for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++) {
+			struct buffer_head *tbh = bhs[i_bhs];
+
+			wait_on_buffer(tbh);
+			if (unlikely(!buffer_uptodate(tbh))) {
+				err = -EIO;
+				/*
+				 * Set the buffer uptodate so the page and
+				 * buffer states do not become out of sync.
+				 */
+				set_buffer_uptodate(tbh);
+			}
+		}
+	} else /* if (unlikely(err)) */ {
+		/* Clean the buffers. */
+		for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++)
+			clear_buffer_dirty(bhs[i_bhs]);
+	}
+	/* Current state: all buffers are clean, unlocked, and uptodate. */
+	/* Remove the mst protection fixups again. */
+	post_write_mst_fixup((NTFS_RECORD*)kmirr);
+	flush_dcache_page(page);
+	SetPageUptodate(page);
+	unlock_page(page);
+	ntfs_unmap_page(page);
+	if (likely(!err)) {
+		ntfs_debug("Done.");
+	} else {
+		ntfs_error(vol->sb, "I/O error while writing mft mirror "
+				"record 0x%lx!", mft_no);
+err_out:
+		ntfs_error(vol->sb, "Failed to synchronize $MFTMirr (error "
+				"code %i).  Volume will be left marked dirty "
+				"on umount.  Run ntfsfix on the partition "
+				"after umounting to correct this.", -err);
+		NVolSetErrors(vol);
+	}
+	return err;
+}
+
+/**
+ * write_mft_record_nolock - write out a mapped (extent) mft record
+ * @ni:		ntfs inode describing the mapped (extent) mft record
+ * @m:		mapped (extent) mft record to write
+ * @sync:	if true, wait for i/o completion
+ *
+ * Write the mapped (extent) mft record @m described by the (regular or extent)
+ * ntfs inode @ni to backing store.  If the mft record @m has a counterpart in
+ * the mft mirror, that is also updated.
+ *
+ * We only write the mft record if the ntfs inode @ni is dirty and the first
+ * buffer belonging to its mft record is dirty, too.  We ignore the dirty state
+ * of subsequent buffers because we could have raced with
+ * fs/ntfs/aops.c::mark_ntfs_record_dirty().
+ *
+ * On success, clean the mft record and return 0.  On error, leave the mft
+ * record dirty and return -errno.
+ *
+ * NOTE:  We always perform synchronous i/o and ignore the @sync parameter.
+ * However, if the mft record has a counterpart in the mft mirror and @sync is
+ * true, we write the mft record, wait for i/o completion, and only then write
+ * the mft mirror copy.  This ensures that if the system crashes either the mft
+ * or the mft mirror will contain a self-consistent mft record @m.  If @sync is
+ * false on the other hand, we start i/o on both and then wait for completion
+ * on them.  This provides a speedup but no longer guarantees that you will end
+ * up with a self-consistent mft record in the case of a crash but if you asked
+ * for asynchronous writing you probably do not care about that anyway.
+ *
+ * TODO:  If @sync is false, want to do truly asynchronous i/o, i.e. just
+ * schedule i/o via ->writepage or do it via kntfsd or whatever.
+ */
+int write_mft_record_nolock(ntfs_inode *ni, MFT_RECORD *m, int sync)
+{
+	ntfs_volume *vol = ni->vol;
+	struct page *page = ni->page;
+	unsigned int blocksize = vol->sb->s_blocksize;
+	unsigned char blocksize_bits = vol->sb->s_blocksize_bits;
+	int max_bhs = vol->mft_record_size / blocksize;
+	struct buffer_head *bhs[MAX_BHS];
+	struct buffer_head *bh, *head;
+	runlist_element *rl;
+	unsigned int block_start, block_end, m_start, m_end;
+	int i_bhs, nr_bhs, err = 0;
+
+	ntfs_debug("Entering for inode 0x%lx.", ni->mft_no);
+	BUG_ON(NInoAttr(ni));
+	BUG_ON(!max_bhs);
+	BUG_ON(!PageLocked(page));
+	if (WARN_ON(max_bhs > MAX_BHS)) {
+		err = -EINVAL;
+		goto err_out;
+	}
+	/*
+	 * If the ntfs_inode is clean no need to do anything.  If it is dirty,
+	 * mark it as clean now so that it can be redirtied later on if needed.
+	 * There is no danger of races since the caller is holding the locks
+	 * for the mft record @m and the page it is in.
+	 */
+	if (!NInoTestClearDirty(ni))
+		goto done;
+	bh = head = page_buffers(page);
+	BUG_ON(!bh);
+	rl = NULL;
+	nr_bhs = 0;
+	block_start = 0;
+	m_start = ni->page_ofs;
+	m_end = m_start + vol->mft_record_size;
+	do {
+		block_end = block_start + blocksize;
+		/* If the buffer is outside the mft record, skip it. */
+		if (block_end <= m_start)
+			continue;
+		if (unlikely(block_start >= m_end))
+			break;
+		/*
+		 * If this block is not the first one in the record, we ignore
+		 * the buffer's dirty state because we could have raced with a
+		 * parallel mark_ntfs_record_dirty().
+		 */
+		if (block_start == m_start) {
+			/* This block is the first one in the record. */
+			if (!buffer_dirty(bh)) {
+				BUG_ON(nr_bhs);
+				/* Clean records are not written out. */
+				break;
+			}
+		}
+		/* Need to map the buffer if it is not mapped already. */
+		if (unlikely(!buffer_mapped(bh))) {
+			VCN vcn;
+			LCN lcn;
+			unsigned int vcn_ofs;
+
+			bh->b_bdev = vol->sb->s_bdev;
+			/* Obtain the vcn and offset of the current block. */
+			vcn = ((VCN)ni->mft_no << vol->mft_record_size_bits) +
+					(block_start - m_start);
+			vcn_ofs = vcn & vol->cluster_size_mask;
+			vcn >>= vol->cluster_size_bits;
+			if (!rl) {
+				down_read(&NTFS_I(vol->mft_ino)->runlist.lock);
+				rl = NTFS_I(vol->mft_ino)->runlist.rl;
+				BUG_ON(!rl);
+			}
+			/* Seek to element containing target vcn. */
+			while (rl->length && rl[1].vcn <= vcn)
+				rl++;
+			lcn = ntfs_rl_vcn_to_lcn(rl, vcn);
+			/* For $MFT, only lcn >= 0 is a successful remap. */
+			if (likely(lcn >= 0)) {
+				/* Setup buffer head to correct block. */
+				bh->b_blocknr = ((lcn <<
+						vol->cluster_size_bits) +
+						vcn_ofs) >> blocksize_bits;
+				set_buffer_mapped(bh);
+			} else {
+				bh->b_blocknr = -1;
+				ntfs_error(vol->sb, "Cannot write mft record "
+						"0x%lx because its location "
+						"on disk could not be "
+						"determined (error code %lli).",
+						ni->mft_no, (long long)lcn);
+				err = -EIO;
+			}
+		}
+		BUG_ON(!buffer_uptodate(bh));
+		BUG_ON(!nr_bhs && (m_start != block_start));
+		BUG_ON(nr_bhs >= max_bhs);
+		bhs[nr_bhs++] = bh;
+		BUG_ON((nr_bhs >= max_bhs) && (m_end != block_end));
+	} while (block_start = block_end, (bh = bh->b_this_page) != head);
+	if (unlikely(rl))
+		up_read(&NTFS_I(vol->mft_ino)->runlist.lock);
+	if (!nr_bhs)
+		goto done;
+	if (unlikely(err))
+		goto cleanup_out;
+	/* Apply the mst protection fixups. */
+	err = pre_write_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size);
+	if (err) {
+		ntfs_error(vol->sb, "Failed to apply mst fixups!");
+		goto cleanup_out;
+	}
+	flush_dcache_mft_record_page(ni);
+	/* Lock buffers and start synchronous write i/o on them. */
+	for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++) {
+		struct buffer_head *tbh = bhs[i_bhs];
+
+		if (!trylock_buffer(tbh))
+			BUG();
+		BUG_ON(!buffer_uptodate(tbh));
+		clear_buffer_dirty(tbh);
+		get_bh(tbh);
+		tbh->b_end_io = end_buffer_write_sync;
+		submit_bh(REQ_OP_WRITE, 0, tbh);
+	}
+	/* Synchronize the mft mirror now if not @sync. */
+	if (!sync && ni->mft_no < vol->mftmirr_size)
+		ntfs_sync_mft_mirror(vol, ni->mft_no, m, sync);
+	/* Wait on i/o completion of buffers. */
+	for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++) {
+		struct buffer_head *tbh = bhs[i_bhs];
+
+		wait_on_buffer(tbh);
+		if (unlikely(!buffer_uptodate(tbh))) {
+			err = -EIO;
+			/*
+			 * Set the buffer uptodate so the page and buffer
+			 * states do not become out of sync.
+			 */
+			if (PageUptodate(page))
+				set_buffer_uptodate(tbh);
+		}
+	}
+	/* If @sync, now synchronize the mft mirror. */
+	if (sync && ni->mft_no < vol->mftmirr_size)
+		ntfs_sync_mft_mirror(vol, ni->mft_no, m, sync);
+	/* Remove the mst protection fixups again. */
+	post_write_mst_fixup((NTFS_RECORD*)m);
+	flush_dcache_mft_record_page(ni);
+	if (unlikely(err)) {
+		/* I/O error during writing.  This is really bad! */
+		ntfs_error(vol->sb, "I/O error while writing mft record "
+				"0x%lx!  Marking base inode as bad.  You "
+				"should unmount the volume and run chkdsk.",
+				ni->mft_no);
+		goto err_out;
+	}
+done:
+	ntfs_debug("Done.");
+	return 0;
+cleanup_out:
+	/* Clean the buffers. */
+	for (i_bhs = 0; i_bhs < nr_bhs; i_bhs++)
+		clear_buffer_dirty(bhs[i_bhs]);
+err_out:
+	/*
+	 * Current state: all buffers are clean, unlocked, and uptodate.
+	 * The caller should mark the base inode as bad so that no more i/o
+	 * happens.  ->clear_inode() will still be invoked so all extent inodes
+	 * and other allocated memory will be freed.
+	 */
+	if (err == -ENOMEM) {
+		ntfs_error(vol->sb, "Not enough memory to write mft record.  "
+				"Redirtying so the write is retried later.");
+		mark_mft_record_dirty(ni);
+		err = 0;
+	} else
+		NVolSetErrors(vol);
+	return err;
+}
+
+/**
+ * ntfs_may_write_mft_record - check if an mft record may be written out
+ * @vol:	[IN]  ntfs volume on which the mft record to check resides
+ * @mft_no:	[IN]  mft record number of the mft record to check
+ * @m:		[IN]  mapped mft record to check
+ * @locked_ni:	[OUT] caller has to unlock this ntfs inode if one is returned
+ *
+ * Check if the mapped (base or extent) mft record @m with mft record number
+ * @mft_no belonging to the ntfs volume @vol may be written out.  If necessary
+ * and possible the ntfs inode of the mft record is locked and the base vfs
+ * inode is pinned.  The locked ntfs inode is then returned in @locked_ni.  The
+ * caller is responsible for unlocking the ntfs inode and unpinning the base
+ * vfs inode.
+ *
+ * Return 'true' if the mft record may be written out and 'false' if not.
+ *
+ * The caller has locked the page and cleared the uptodate flag on it which
+ * means that we can safely write out any dirty mft records that do not have
+ * their inodes in icache as determined by ilookup5() as anyone
+ * opening/creating such an inode would block when attempting to map the mft
+ * record in read_cache_page() until we are finished with the write out.
+ *
+ * Here is a description of the tests we perform:
+ *
+ * If the inode is found in icache we know the mft record must be a base mft
+ * record.  If it is dirty, we do not write it and return 'false' as the vfs
+ * inode write paths will result in the access times being updated which would
+ * cause the base mft record to be redirtied and written out again.  (We know
+ * the access time update will modify the base mft record because Windows
+ * chkdsk complains if the standard information attribute is not in the base
+ * mft record.)
+ *
+ * If the inode is in icache and not dirty, we attempt to lock the mft record
+ * and if we find the lock was already taken, it is not safe to write the mft
+ * record and we return 'false'.
+ *
+ * If we manage to obtain the lock we have exclusive access to the mft record,
+ * which also allows us safe writeout of the mft record.  We then set
+ * @locked_ni to the locked ntfs inode and return 'true'.
+ *
+ * Note we cannot just lock the mft record and sleep while waiting for the lock
+ * because this would deadlock due to lock reversal (normally the mft record is
+ * locked before the page is locked but we already have the page locked here
+ * when we try to lock the mft record).
+ *
+ * If the inode is not in icache we need to perform further checks.
+ *
+ * If the mft record is not a FILE record or it is a base mft record, we can
+ * safely write it and return 'true'.
+ *
+ * We now know the mft record is an extent mft record.  We check if the inode
+ * corresponding to its base mft record is in icache and obtain a reference to
+ * it if it is.  If it is not, we can safely write it and return 'true'.
+ *
+ * We now have the base inode for the extent mft record.  We check if it has an
+ * ntfs inode for the extent mft record attached and if not it is safe to write
+ * the extent mft record and we return 'true'.
+ *
+ * The ntfs inode for the extent mft record is attached to the base inode so we
+ * attempt to lock the extent mft record and if we find the lock was already
+ * taken, it is not safe to write the extent mft record and we return 'false'.
+ *
+ * If we manage to obtain the lock we have exclusive access to the extent mft
+ * record, which also allows us safe writeout of the extent mft record.  We
+ * set the ntfs inode of the extent mft record clean and then set @locked_ni to
+ * the now locked ntfs inode and return 'true'.
+ *
+ * Note, the reason for actually writing dirty mft records here and not just
+ * relying on the vfs inode dirty code paths is that we can have mft records
+ * modified without them ever having actual inodes in memory.  Also we can have
+ * dirty mft records with clean ntfs inodes in memory.  None of the described
+ * cases would result in the dirty mft records being written out if we only
+ * relied on the vfs inode dirty code paths.  And these cases can really occur
+ * during allocation of new mft records and in particular when the
+ * initialized_size of the $MFT/$DATA attribute is extended and the new space
+ * is initialized using ntfs_mft_record_format().  The clean inode can then
+ * appear if the mft record is reused for a new inode before it got written
+ * out.
+ */
+bool ntfs_may_write_mft_record(ntfs_volume *vol, const unsigned long mft_no,
+		const MFT_RECORD *m, ntfs_inode **locked_ni)
+{
+	struct super_block *sb = vol->sb;
+	struct inode *mft_vi = vol->mft_ino;
+	struct inode *vi;
+	ntfs_inode *ni, *eni, **extent_nis;
+	int i;
+	ntfs_attr na;
+
+	ntfs_debug("Entering for inode 0x%lx.", mft_no);
+	/*
+	 * Normally we do not return a locked inode so set @locked_ni to NULL.
+	 */
+	BUG_ON(!locked_ni);
+	*locked_ni = NULL;
+	/*
+	 * Check if the inode corresponding to this mft record is in the VFS
+	 * inode cache and obtain a reference to it if it is.
+	 */
+	ntfs_debug("Looking for inode 0x%lx in icache.", mft_no);
+	na.mft_no = mft_no;
+	na.name = NULL;
+	na.name_len = 0;
+	na.type = AT_UNUSED;
+	/*
+	 * Optimize inode 0, i.e. $MFT itself, since we have it in memory and
+	 * we get here for it rather often.
+	 */
+	if (!mft_no) {
+		/* Balance the below iput(). */
+		vi = igrab(mft_vi);
+		BUG_ON(vi != mft_vi);
+	} else {
+		/*
+		 * Have to use ilookup5_nowait() since ilookup5() waits for the
+		 * inode lock which causes ntfs to deadlock when a concurrent
+		 * inode write via the inode dirty code paths and the page
+		 * dirty code path of the inode dirty code path when writing
+		 * $MFT occurs.
+		 */
+		vi = ilookup5_nowait(sb, mft_no, ntfs_test_inode, &na);
+	}
+	if (vi) {
+		ntfs_debug("Base inode 0x%lx is in icache.", mft_no);
+		/* The inode is in icache. */
+		ni = NTFS_I(vi);
+		/* Take a reference to the ntfs inode. */
+		atomic_inc(&ni->count);
+		/* If the inode is dirty, do not write this record. */
+		if (NInoDirty(ni)) {
+			ntfs_debug("Inode 0x%lx is dirty, do not write it.",
+					mft_no);
+			atomic_dec(&ni->count);
+			iput(vi);
+			return false;
+		}
+		ntfs_debug("Inode 0x%lx is not dirty.", mft_no);
+		/* The inode is not dirty, try to take the mft record lock. */
+		if (unlikely(!mutex_trylock(&ni->mrec_lock))) {
+			ntfs_debug("Mft record 0x%lx is already locked, do "
+					"not write it.", mft_no);
+			atomic_dec(&ni->count);
+			iput(vi);
+			return false;
+		}
+		ntfs_debug("Managed to lock mft record 0x%lx, write it.",
+				mft_no);
+		/*
+		 * The write has to occur while we hold the mft record lock so
+		 * return the locked ntfs inode.
+		 */
+		*locked_ni = ni;
+		return true;
+	}
+	ntfs_debug("Inode 0x%lx is not in icache.", mft_no);
+	/* The inode is not in icache. */
+	/* Write the record if it is not a mft record (type "FILE"). */
+	if (!ntfs_is_mft_record(m->magic)) {
+		ntfs_debug("Mft record 0x%lx is not a FILE record, write it.",
+				mft_no);
+		return true;
+	}
+	/* Write the mft record if it is a base inode. */
+	if (!m->base_mft_record) {
+		ntfs_debug("Mft record 0x%lx is a base record, write it.",
+				mft_no);
+		return true;
+	}
+	/*
+	 * This is an extent mft record.  Check if the inode corresponding to
+	 * its base mft record is in icache and obtain a reference to it if it
+	 * is.
+	 */
+	na.mft_no = MREF_LE(m->base_mft_record);
+	ntfs_debug("Mft record 0x%lx is an extent record.  Looking for base "
+			"inode 0x%lx in icache.", mft_no, na.mft_no);
+	if (!na.mft_no) {
+		/* Balance the below iput(). */
+		vi = igrab(mft_vi);
+		BUG_ON(vi != mft_vi);
+	} else
+		vi = ilookup5_nowait(sb, na.mft_no, ntfs_test_inode,
+				&na);
+	if (!vi) {
+		/*
+		 * The base inode is not in icache, write this extent mft
+		 * record.
+		 */
+		ntfs_debug("Base inode 0x%lx is not in icache, write the "
+				"extent record.", na.mft_no);
+		return true;
+	}
+	ntfs_debug("Base inode 0x%lx is in icache.", na.mft_no);
+	/*
+	 * The base inode is in icache.  Check if it has the extent inode
+	 * corresponding to this extent mft record attached.
+	 */
+	ni = NTFS_I(vi);
+	mutex_lock(&ni->extent_lock);
+	if (ni->nr_extents <= 0) {
+		/*
+		 * The base inode has no attached extent inodes, write this
+		 * extent mft record.
+		 */
+		mutex_unlock(&ni->extent_lock);
+		iput(vi);
+		ntfs_debug("Base inode 0x%lx has no attached extent inodes, "
+				"write the extent record.", na.mft_no);
+		return true;
+	}
+	/* Iterate over the attached extent inodes. */
+	extent_nis = ni->ext.extent_ntfs_inos;
+	for (eni = NULL, i = 0; i < ni->nr_extents; ++i) {
+		if (mft_no == extent_nis[i]->mft_no) {
+			/*
+			 * Found the extent inode corresponding to this extent
+			 * mft record.
+			 */
+			eni = extent_nis[i];
+			break;
+		}
+	}
+	/*
+	 * If the extent inode was not attached to the base inode, write this
+	 * extent mft record.
+	 */
+	if (!eni) {
+		mutex_unlock(&ni->extent_lock);
+		iput(vi);
+		ntfs_debug("Extent inode 0x%lx is not attached to its base "
+				"inode 0x%lx, write the extent record.",
+				mft_no, na.mft_no);
+		return true;
+	}
+	ntfs_debug("Extent inode 0x%lx is attached to its base inode 0x%lx.",
+			mft_no, na.mft_no);
+	/* Take a reference to the extent ntfs inode. */
+	atomic_inc(&eni->count);
+	mutex_unlock(&ni->extent_lock);
+	/*
+	 * Found the extent inode coresponding to this extent mft record.
+	 * Try to take the mft record lock.
+	 */
+	if (unlikely(!mutex_trylock(&eni->mrec_lock))) {
+		atomic_dec(&eni->count);
+		iput(vi);
+		ntfs_debug("Extent mft record 0x%lx is already locked, do "
+				"not write it.", mft_no);
+		return false;
+	}
+	ntfs_debug("Managed to lock extent mft record 0x%lx, write it.",
+			mft_no);
+	if (NInoTestClearDirty(eni))
+		ntfs_debug("Extent inode 0x%lx is dirty, marking it clean.",
+				mft_no);
+	/*
+	 * The write has to occur while we hold the mft record lock so return
+	 * the locked extent ntfs inode.
+	 */
+	*locked_ni = eni;
+	return true;
+}
+
+static const char *es = "  Leaving inconsistent metadata.  Unmount and run "
+		"chkdsk.";
+
+/**
+ * ntfs_mft_bitmap_find_and_alloc_free_rec_nolock - see name
+ * @vol:	volume on which to search for a free mft record
+ * @base_ni:	open base inode if allocating an extent mft record or NULL
+ *
+ * Search for a free mft record in the mft bitmap attribute on the ntfs volume
+ * @vol.
+ *
+ * If @base_ni is NULL start the search at the default allocator position.
+ *
+ * If @base_ni is not NULL start the search at the mft record after the base
+ * mft record @base_ni.
+ *
+ * Return the free mft record on success and -errno on error.  An error code of
+ * -ENOSPC means that there are no free mft records in the currently
+ * initialized mft bitmap.
+ *
+ * Locking: Caller must hold vol->mftbmp_lock for writing.
+ */
+static int ntfs_mft_bitmap_find_and_alloc_free_rec_nolock(ntfs_volume *vol,
+		ntfs_inode *base_ni)
+{
+	s64 pass_end, ll, data_pos, pass_start, ofs, bit;
+	unsigned long flags;
+	struct address_space *mftbmp_mapping;
+	u8 *buf, *byte;
+	struct page *page;
+	unsigned int page_ofs, size;
+	u8 pass, b;
+
+	ntfs_debug("Searching for free mft record in the currently "
+			"initialized mft bitmap.");
+	mftbmp_mapping = vol->mftbmp_ino->i_mapping;
+	/*
+	 * Set the end of the pass making sure we do not overflow the mft
+	 * bitmap.
+	 */
+	read_lock_irqsave(&NTFS_I(vol->mft_ino)->size_lock, flags);
+	pass_end = NTFS_I(vol->mft_ino)->allocated_size >>
+			vol->mft_record_size_bits;
+	read_unlock_irqrestore(&NTFS_I(vol->mft_ino)->size_lock, flags);
+	read_lock_irqsave(&NTFS_I(vol->mftbmp_ino)->size_lock, flags);
+	ll = NTFS_I(vol->mftbmp_ino)->initialized_size << 3;
+	read_unlock_irqrestore(&NTFS_I(vol->mftbmp_ino)->size_lock, flags);
+	if (pass_end > ll)
+		pass_end = ll;
+	pass = 1;
+	if (!base_ni)
+		data_pos = vol->mft_data_pos;
+	else
+		data_pos = base_ni->mft_no + 1;
+	if (data_pos < 24)
+		data_pos = 24;
+	if (data_pos >= pass_end) {
+		data_pos = 24;
+		pass = 2;
+		/* This happens on a freshly formatted volume. */
+		if (data_pos >= pass_end)
+			return -ENOSPC;
+	}
+	pass_start = data_pos;
+	ntfs_debug("Starting bitmap search: pass %u, pass_start 0x%llx, "
+			"pass_end 0x%llx, data_pos 0x%llx.", pass,
+			(long long)pass_start, (long long)pass_end,
+			(long long)data_pos);
+	/* Loop until a free mft record is found. */
+	for (; pass <= 2;) {
+		/* Cap size to pass_end. */
+		ofs = data_pos >> 3;
+		page_ofs = ofs & ~PAGE_MASK;
+		size = PAGE_SIZE - page_ofs;
+		ll = ((pass_end + 7) >> 3) - ofs;
+		if (size > ll)
+			size = ll;
+		size <<= 3;
+		/*
+		 * If we are still within the active pass, search the next page
+		 * for a zero bit.
+		 */
+		if (size) {
+			page = ntfs_map_page(mftbmp_mapping,
+					ofs >> PAGE_SHIFT);
+			if (IS_ERR(page)) {
+				ntfs_error(vol->sb, "Failed to read mft "
+						"bitmap, aborting.");
+				return PTR_ERR(page);
+			}
+			buf = (u8*)page_address(page) + page_ofs;
+			bit = data_pos & 7;
+			data_pos &= ~7ull;
+			ntfs_debug("Before inner for loop: size 0x%x, "
+					"data_pos 0x%llx, bit 0x%llx", size,
+					(long long)data_pos, (long long)bit);
+			for (; bit < size && data_pos + bit < pass_end;
+					bit &= ~7ull, bit += 8) {
+				byte = buf + (bit >> 3);
+				if (*byte == 0xff)
+					continue;
+				b = ffz((unsigned long)*byte);
+				if (b < 8 && b >= (bit & 7)) {
+					ll = data_pos + (bit & ~7ull) + b;
+					if (unlikely(ll > (1ll << 32))) {
+						ntfs_unmap_page(page);
+						return -ENOSPC;
+					}
+					*byte |= 1 << b;
+					flush_dcache_page(page);
+					set_page_dirty(page);
+					ntfs_unmap_page(page);
+					ntfs_debug("Done.  (Found and "
+							"allocated mft record "
+							"0x%llx.)",
+							(long long)ll);
+					return ll;
+				}
+			}
+			ntfs_debug("After inner for loop: size 0x%x, "
+					"data_pos 0x%llx, bit 0x%llx", size,
+					(long long)data_pos, (long long)bit);
+			data_pos += size;
+			ntfs_unmap_page(page);
+			/*
+			 * If the end of the pass has not been reached yet,
+			 * continue searching the mft bitmap for a zero bit.
+			 */
+			if (data_pos < pass_end)
+				continue;
+		}
+		/* Do the next pass. */
+		if (++pass == 2) {
+			/*
+			 * Starting the second pass, in which we scan the first
+			 * part of the zone which we omitted earlier.
+			 */
+			pass_end = pass_start;
+			data_pos = pass_start = 24;
+			ntfs_debug("pass %i, pass_start 0x%llx, pass_end "
+					"0x%llx.", pass, (long long)pass_start,
+					(long long)pass_end);
+			if (data_pos >= pass_end)
+				break;
+		}
+	}
+	/* No free mft records in currently initialized mft bitmap. */
+	ntfs_debug("Done.  (No free mft records left in currently initialized "
+			"mft bitmap.)");
+	return -ENOSPC;
+}
+
+/**
+ * ntfs_mft_bitmap_extend_allocation_nolock - extend mft bitmap by a cluster
+ * @vol:	volume on which to extend the mft bitmap attribute
+ *
+ * Extend the mft bitmap attribute on the ntfs volume @vol by one cluster.
+ *
+ * Note: Only changes allocated_size, i.e. does not touch initialized_size or
+ * data_size.
+ *
+ * Return 0 on success and -errno on error.
+ *
+ * Locking: - Caller must hold vol->mftbmp_lock for writing.
+ *	    - This function takes NTFS_I(vol->mftbmp_ino)->runlist.lock for
+ *	      writing and releases it before returning.
+ *	    - This function takes vol->lcnbmp_lock for writing and releases it
+ *	      before returning.
+ */
+static int ntfs_mft_bitmap_extend_allocation_nolock(ntfs_volume *vol)
+{
+	LCN lcn;
+	s64 ll;
+	unsigned long flags;
+	struct page *page;
+	ntfs_inode *mft_ni, *mftbmp_ni;
+	runlist_element *rl, *rl2 = NULL;
+	ntfs_attr_search_ctx *ctx = NULL;
+	MFT_RECORD *mrec;
+	ATTR_RECORD *a = NULL;
+	int ret, mp_size;
+	u32 old_alen = 0;
+	u8 *b, tb;
+	struct {
+		u8 added_cluster:1;
+		u8 added_run:1;
+		u8 mp_rebuilt:1;
+	} status = { 0, 0, 0 };
+
+	ntfs_debug("Extending mft bitmap allocation.");
+	mft_ni = NTFS_I(vol->mft_ino);
+	mftbmp_ni = NTFS_I(vol->mftbmp_ino);
+	/*
+	 * Determine the last lcn of the mft bitmap.  The allocated size of the
+	 * mft bitmap cannot be zero so we are ok to do this.
+	 */
+	down_write(&mftbmp_ni->runlist.lock);
+	read_lock_irqsave(&mftbmp_ni->size_lock, flags);
+	ll = mftbmp_ni->allocated_size;
+	read_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+	rl = ntfs_attr_find_vcn_nolock(mftbmp_ni,
+			(ll - 1) >> vol->cluster_size_bits, NULL);
+	if (IS_ERR(rl) || unlikely(!rl->length || rl->lcn < 0)) {
+		up_write(&mftbmp_ni->runlist.lock);
+		ntfs_error(vol->sb, "Failed to determine last allocated "
+				"cluster of mft bitmap attribute.");
+		if (!IS_ERR(rl))
+			ret = -EIO;
+		else
+			ret = PTR_ERR(rl);
+		return ret;
+	}
+	lcn = rl->lcn + rl->length;
+	ntfs_debug("Last lcn of mft bitmap attribute is 0x%llx.",
+			(long long)lcn);
+	/*
+	 * Attempt to get the cluster following the last allocated cluster by
+	 * hand as it may be in the MFT zone so the allocator would not give it
+	 * to us.
+	 */
+	ll = lcn >> 3;
+	page = ntfs_map_page(vol->lcnbmp_ino->i_mapping,
+			ll >> PAGE_SHIFT);
+	if (IS_ERR(page)) {
+		up_write(&mftbmp_ni->runlist.lock);
+		ntfs_error(vol->sb, "Failed to read from lcn bitmap.");
+		return PTR_ERR(page);
+	}
+	b = (u8*)page_address(page) + (ll & ~PAGE_MASK);
+	tb = 1 << (lcn & 7ull);
+	down_write(&vol->lcnbmp_lock);
+	if (*b != 0xff && !(*b & tb)) {
+		/* Next cluster is free, allocate it. */
+		*b |= tb;
+		flush_dcache_page(page);
+		set_page_dirty(page);
+		up_write(&vol->lcnbmp_lock);
+		ntfs_unmap_page(page);
+		/* Update the mft bitmap runlist. */
+		rl->length++;
+		rl[1].vcn++;
+		status.added_cluster = 1;
+		ntfs_debug("Appending one cluster to mft bitmap.");
+	} else {
+		up_write(&vol->lcnbmp_lock);
+		ntfs_unmap_page(page);
+		/* Allocate a cluster from the DATA_ZONE. */
+		rl2 = ntfs_cluster_alloc(vol, rl[1].vcn, 1, lcn, DATA_ZONE,
+				true);
+		if (IS_ERR(rl2)) {
+			up_write(&mftbmp_ni->runlist.lock);
+			ntfs_error(vol->sb, "Failed to allocate a cluster for "
+					"the mft bitmap.");
+			return PTR_ERR(rl2);
+		}
+		rl = ntfs_runlists_merge(mftbmp_ni->runlist.rl, rl2);
+		if (IS_ERR(rl)) {
+			up_write(&mftbmp_ni->runlist.lock);
+			ntfs_error(vol->sb, "Failed to merge runlists for mft "
+					"bitmap.");
+			if (ntfs_cluster_free_from_rl(vol, rl2)) {
+				ntfs_error(vol->sb, "Failed to deallocate "
+						"allocated cluster.%s", es);
+				NVolSetErrors(vol);
+			}
+			ntfs_free(rl2);
+			return PTR_ERR(rl);
+		}
+		mftbmp_ni->runlist.rl = rl;
+		status.added_run = 1;
+		ntfs_debug("Adding one run to mft bitmap.");
+		/* Find the last run in the new runlist. */
+		for (; rl[1].length; rl++)
+			;
+	}
+	/*
+	 * Update the attribute record as well.  Note: @rl is the last
+	 * (non-terminator) runlist element of mft bitmap.
+	 */
+	mrec = map_mft_record(mft_ni);
+	if (IS_ERR(mrec)) {
+		ntfs_error(vol->sb, "Failed to map mft record.");
+		ret = PTR_ERR(mrec);
+		goto undo_alloc;
+	}
+	ctx = ntfs_attr_get_search_ctx(mft_ni, mrec);
+	if (unlikely(!ctx)) {
+		ntfs_error(vol->sb, "Failed to get search context.");
+		ret = -ENOMEM;
+		goto undo_alloc;
+	}
+	ret = ntfs_attr_lookup(mftbmp_ni->type, mftbmp_ni->name,
+			mftbmp_ni->name_len, CASE_SENSITIVE, rl[1].vcn, NULL,
+			0, ctx);
+	if (unlikely(ret)) {
+		ntfs_error(vol->sb, "Failed to find last attribute extent of "
+				"mft bitmap attribute.");
+		if (ret == -ENOENT)
+			ret = -EIO;
+		goto undo_alloc;
+	}
+	a = ctx->attr;
+	ll = sle64_to_cpu(a->data.non_resident.lowest_vcn);
+	/* Search back for the previous last allocated cluster of mft bitmap. */
+	for (rl2 = rl; rl2 > mftbmp_ni->runlist.rl; rl2--) {
+		if (ll >= rl2->vcn)
+			break;
+	}
+	BUG_ON(ll < rl2->vcn);
+	BUG_ON(ll >= rl2->vcn + rl2->length);
+	/* Get the size for the new mapping pairs array for this extent. */
+	mp_size = ntfs_get_size_for_mapping_pairs(vol, rl2, ll, -1);
+	if (unlikely(mp_size <= 0)) {
+		ntfs_error(vol->sb, "Get size for mapping pairs failed for "
+				"mft bitmap attribute extent.");
+		ret = mp_size;
+		if (!ret)
+			ret = -EIO;
+		goto undo_alloc;
+	}
+	/* Expand the attribute record if necessary. */
+	old_alen = le32_to_cpu(a->length);
+	ret = ntfs_attr_record_resize(ctx->mrec, a, mp_size +
+			le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
+	if (unlikely(ret)) {
+		if (ret != -ENOSPC) {
+			ntfs_error(vol->sb, "Failed to resize attribute "
+					"record for mft bitmap attribute.");
+			goto undo_alloc;
+		}
+		// TODO: Deal with this by moving this extent to a new mft
+		// record or by starting a new extent in a new mft record or by
+		// moving other attributes out of this mft record.
+		// Note: It will need to be a special mft record and if none of
+		// those are available it gets rather complicated...
+		ntfs_error(vol->sb, "Not enough space in this mft record to "
+				"accommodate extended mft bitmap attribute "
+				"extent.  Cannot handle this yet.");
+		ret = -EOPNOTSUPP;
+		goto undo_alloc;
+	}
+	status.mp_rebuilt = 1;
+	/* Generate the mapping pairs array directly into the attr record. */
+	ret = ntfs_mapping_pairs_build(vol, (u8*)a +
+			le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
+			mp_size, rl2, ll, -1, NULL);
+	if (unlikely(ret)) {
+		ntfs_error(vol->sb, "Failed to build mapping pairs array for "
+				"mft bitmap attribute.");
+		goto undo_alloc;
+	}
+	/* Update the highest_vcn. */
+	a->data.non_resident.highest_vcn = cpu_to_sle64(rl[1].vcn - 1);
+	/*
+	 * We now have extended the mft bitmap allocated_size by one cluster.
+	 * Reflect this in the ntfs_inode structure and the attribute record.
+	 */
+	if (a->data.non_resident.lowest_vcn) {
+		/*
+		 * We are not in the first attribute extent, switch to it, but
+		 * first ensure the changes will make it to disk later.
+		 */
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		mark_mft_record_dirty(ctx->ntfs_ino);
+		ntfs_attr_reinit_search_ctx(ctx);
+		ret = ntfs_attr_lookup(mftbmp_ni->type, mftbmp_ni->name,
+				mftbmp_ni->name_len, CASE_SENSITIVE, 0, NULL,
+				0, ctx);
+		if (unlikely(ret)) {
+			ntfs_error(vol->sb, "Failed to find first attribute "
+					"extent of mft bitmap attribute.");
+			goto restore_undo_alloc;
+		}
+		a = ctx->attr;
+	}
+	write_lock_irqsave(&mftbmp_ni->size_lock, flags);
+	mftbmp_ni->allocated_size += vol->cluster_size;
+	a->data.non_resident.allocated_size =
+			cpu_to_sle64(mftbmp_ni->allocated_size);
+	write_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+	/* Ensure the changes make it to disk. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(mft_ni);
+	up_write(&mftbmp_ni->runlist.lock);
+	ntfs_debug("Done.");
+	return 0;
+restore_undo_alloc:
+	ntfs_attr_reinit_search_ctx(ctx);
+	if (ntfs_attr_lookup(mftbmp_ni->type, mftbmp_ni->name,
+			mftbmp_ni->name_len, CASE_SENSITIVE, rl[1].vcn, NULL,
+			0, ctx)) {
+		ntfs_error(vol->sb, "Failed to find last attribute extent of "
+				"mft bitmap attribute.%s", es);
+		write_lock_irqsave(&mftbmp_ni->size_lock, flags);
+		mftbmp_ni->allocated_size += vol->cluster_size;
+		write_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(mft_ni);
+		up_write(&mftbmp_ni->runlist.lock);
+		/*
+		 * The only thing that is now wrong is ->allocated_size of the
+		 * base attribute extent which chkdsk should be able to fix.
+		 */
+		NVolSetErrors(vol);
+		return ret;
+	}
+	a = ctx->attr;
+	a->data.non_resident.highest_vcn = cpu_to_sle64(rl[1].vcn - 2);
+undo_alloc:
+	if (status.added_cluster) {
+		/* Truncate the last run in the runlist by one cluster. */
+		rl->length--;
+		rl[1].vcn--;
+	} else if (status.added_run) {
+		lcn = rl->lcn;
+		/* Remove the last run from the runlist. */
+		rl->lcn = rl[1].lcn;
+		rl->length = 0;
+	}
+	/* Deallocate the cluster. */
+	down_write(&vol->lcnbmp_lock);
+	if (ntfs_bitmap_clear_bit(vol->lcnbmp_ino, lcn)) {
+		ntfs_error(vol->sb, "Failed to free allocated cluster.%s", es);
+		NVolSetErrors(vol);
+	}
+	up_write(&vol->lcnbmp_lock);
+	if (status.mp_rebuilt) {
+		if (ntfs_mapping_pairs_build(vol, (u8*)a + le16_to_cpu(
+				a->data.non_resident.mapping_pairs_offset),
+				old_alen - le16_to_cpu(
+				a->data.non_resident.mapping_pairs_offset),
+				rl2, ll, -1, NULL)) {
+			ntfs_error(vol->sb, "Failed to restore mapping pairs "
+					"array.%s", es);
+			NVolSetErrors(vol);
+		}
+		if (ntfs_attr_record_resize(ctx->mrec, a, old_alen)) {
+			ntfs_error(vol->sb, "Failed to restore attribute "
+					"record.%s", es);
+			NVolSetErrors(vol);
+		}
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		mark_mft_record_dirty(ctx->ntfs_ino);
+	}
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (!IS_ERR(mrec))
+		unmap_mft_record(mft_ni);
+	up_write(&mftbmp_ni->runlist.lock);
+	return ret;
+}
+
+/**
+ * ntfs_mft_bitmap_extend_initialized_nolock - extend mftbmp initialized data
+ * @vol:	volume on which to extend the mft bitmap attribute
+ *
+ * Extend the initialized portion of the mft bitmap attribute on the ntfs
+ * volume @vol by 8 bytes.
+ *
+ * Note:  Only changes initialized_size and data_size, i.e. requires that
+ * allocated_size is big enough to fit the new initialized_size.
+ *
+ * Return 0 on success and -error on error.
+ *
+ * Locking: Caller must hold vol->mftbmp_lock for writing.
+ */
+static int ntfs_mft_bitmap_extend_initialized_nolock(ntfs_volume *vol)
+{
+	s64 old_data_size, old_initialized_size;
+	unsigned long flags;
+	struct inode *mftbmp_vi;
+	ntfs_inode *mft_ni, *mftbmp_ni;
+	ntfs_attr_search_ctx *ctx;
+	MFT_RECORD *mrec;
+	ATTR_RECORD *a;
+	int ret;
+
+	ntfs_debug("Extending mft bitmap initiailized (and data) size.");
+	mft_ni = NTFS_I(vol->mft_ino);
+	mftbmp_vi = vol->mftbmp_ino;
+	mftbmp_ni = NTFS_I(mftbmp_vi);
+	/* Get the attribute record. */
+	mrec = map_mft_record(mft_ni);
+	if (IS_ERR(mrec)) {
+		ntfs_error(vol->sb, "Failed to map mft record.");
+		return PTR_ERR(mrec);
+	}
+	ctx = ntfs_attr_get_search_ctx(mft_ni, mrec);
+	if (unlikely(!ctx)) {
+		ntfs_error(vol->sb, "Failed to get search context.");
+		ret = -ENOMEM;
+		goto unm_err_out;
+	}
+	ret = ntfs_attr_lookup(mftbmp_ni->type, mftbmp_ni->name,
+			mftbmp_ni->name_len, CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(ret)) {
+		ntfs_error(vol->sb, "Failed to find first attribute extent of "
+				"mft bitmap attribute.");
+		if (ret == -ENOENT)
+			ret = -EIO;
+		goto put_err_out;
+	}
+	a = ctx->attr;
+	write_lock_irqsave(&mftbmp_ni->size_lock, flags);
+	old_data_size = i_size_read(mftbmp_vi);
+	old_initialized_size = mftbmp_ni->initialized_size;
+	/*
+	 * We can simply update the initialized_size before filling the space
+	 * with zeroes because the caller is holding the mft bitmap lock for
+	 * writing which ensures that no one else is trying to access the data.
+	 */
+	mftbmp_ni->initialized_size += 8;
+	a->data.non_resident.initialized_size =
+			cpu_to_sle64(mftbmp_ni->initialized_size);
+	if (mftbmp_ni->initialized_size > old_data_size) {
+		i_size_write(mftbmp_vi, mftbmp_ni->initialized_size);
+		a->data.non_resident.data_size =
+				cpu_to_sle64(mftbmp_ni->initialized_size);
+	}
+	write_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+	/* Ensure the changes make it to disk. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(mft_ni);
+	/* Initialize the mft bitmap attribute value with zeroes. */
+	ret = ntfs_attr_set(mftbmp_ni, old_initialized_size, 8, 0);
+	if (likely(!ret)) {
+		ntfs_debug("Done.  (Wrote eight initialized bytes to mft "
+				"bitmap.");
+		return 0;
+	}
+	ntfs_error(vol->sb, "Failed to write to mft bitmap.");
+	/* Try to recover from the error. */
+	mrec = map_mft_record(mft_ni);
+	if (IS_ERR(mrec)) {
+		ntfs_error(vol->sb, "Failed to map mft record.%s", es);
+		NVolSetErrors(vol);
+		return ret;
+	}
+	ctx = ntfs_attr_get_search_ctx(mft_ni, mrec);
+	if (unlikely(!ctx)) {
+		ntfs_error(vol->sb, "Failed to get search context.%s", es);
+		NVolSetErrors(vol);
+		goto unm_err_out;
+	}
+	if (ntfs_attr_lookup(mftbmp_ni->type, mftbmp_ni->name,
+			mftbmp_ni->name_len, CASE_SENSITIVE, 0, NULL, 0, ctx)) {
+		ntfs_error(vol->sb, "Failed to find first attribute extent of "
+				"mft bitmap attribute.%s", es);
+		NVolSetErrors(vol);
+put_err_out:
+		ntfs_attr_put_search_ctx(ctx);
+unm_err_out:
+		unmap_mft_record(mft_ni);
+		goto err_out;
+	}
+	a = ctx->attr;
+	write_lock_irqsave(&mftbmp_ni->size_lock, flags);
+	mftbmp_ni->initialized_size = old_initialized_size;
+	a->data.non_resident.initialized_size =
+			cpu_to_sle64(old_initialized_size);
+	if (i_size_read(mftbmp_vi) != old_data_size) {
+		i_size_write(mftbmp_vi, old_data_size);
+		a->data.non_resident.data_size = cpu_to_sle64(old_data_size);
+	}
+	write_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(mft_ni);
+#ifdef DEBUG
+	read_lock_irqsave(&mftbmp_ni->size_lock, flags);
+	ntfs_debug("Restored status of mftbmp: allocated_size 0x%llx, "
+			"data_size 0x%llx, initialized_size 0x%llx.",
+			(long long)mftbmp_ni->allocated_size,
+			(long long)i_size_read(mftbmp_vi),
+			(long long)mftbmp_ni->initialized_size);
+	read_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+#endif /* DEBUG */
+err_out:
+	return ret;
+}
+
+/**
+ * ntfs_mft_data_extend_allocation_nolock - extend mft data attribute
+ * @vol:	volume on which to extend the mft data attribute
+ *
+ * Extend the mft data attribute on the ntfs volume @vol by 16 mft records
+ * worth of clusters or if not enough space for this by one mft record worth
+ * of clusters.
+ *
+ * Note:  Only changes allocated_size, i.e. does not touch initialized_size or
+ * data_size.
+ *
+ * Return 0 on success and -errno on error.
+ *
+ * Locking: - Caller must hold vol->mftbmp_lock for writing.
+ *	    - This function takes NTFS_I(vol->mft_ino)->runlist.lock for
+ *	      writing and releases it before returning.
+ *	    - This function calls functions which take vol->lcnbmp_lock for
+ *	      writing and release it before returning.
+ */
+static int ntfs_mft_data_extend_allocation_nolock(ntfs_volume *vol)
+{
+	LCN lcn;
+	VCN old_last_vcn;
+	s64 min_nr, nr, ll;
+	unsigned long flags;
+	ntfs_inode *mft_ni;
+	runlist_element *rl, *rl2;
+	ntfs_attr_search_ctx *ctx = NULL;
+	MFT_RECORD *mrec;
+	ATTR_RECORD *a = NULL;
+	int ret, mp_size;
+	u32 old_alen = 0;
+	bool mp_rebuilt = false;
+
+	ntfs_debug("Extending mft data allocation.");
+	mft_ni = NTFS_I(vol->mft_ino);
+	/*
+	 * Determine the preferred allocation location, i.e. the last lcn of
+	 * the mft data attribute.  The allocated size of the mft data
+	 * attribute cannot be zero so we are ok to do this.
+	 */
+	down_write(&mft_ni->runlist.lock);
+	read_lock_irqsave(&mft_ni->size_lock, flags);
+	ll = mft_ni->allocated_size;
+	read_unlock_irqrestore(&mft_ni->size_lock, flags);
+	rl = ntfs_attr_find_vcn_nolock(mft_ni,
+			(ll - 1) >> vol->cluster_size_bits, NULL);
+	if (IS_ERR(rl) || unlikely(!rl->length || rl->lcn < 0)) {
+		up_write(&mft_ni->runlist.lock);
+		ntfs_error(vol->sb, "Failed to determine last allocated "
+				"cluster of mft data attribute.");
+		if (!IS_ERR(rl))
+			ret = -EIO;
+		else
+			ret = PTR_ERR(rl);
+		return ret;
+	}
+	lcn = rl->lcn + rl->length;
+	ntfs_debug("Last lcn of mft data attribute is 0x%llx.", (long long)lcn);
+	/* Minimum allocation is one mft record worth of clusters. */
+	min_nr = vol->mft_record_size >> vol->cluster_size_bits;
+	if (!min_nr)
+		min_nr = 1;
+	/* Want to allocate 16 mft records worth of clusters. */
+	nr = vol->mft_record_size << 4 >> vol->cluster_size_bits;
+	if (!nr)
+		nr = min_nr;
+	/* Ensure we do not go above 2^32-1 mft records. */
+	read_lock_irqsave(&mft_ni->size_lock, flags);
+	ll = mft_ni->allocated_size;
+	read_unlock_irqrestore(&mft_ni->size_lock, flags);
+	if (unlikely((ll + (nr << vol->cluster_size_bits)) >>
+			vol->mft_record_size_bits >= (1ll << 32))) {
+		nr = min_nr;
+		if (unlikely((ll + (nr << vol->cluster_size_bits)) >>
+				vol->mft_record_size_bits >= (1ll << 32))) {
+			ntfs_warning(vol->sb, "Cannot allocate mft record "
+					"because the maximum number of inodes "
+					"(2^32) has already been reached.");
+			up_write(&mft_ni->runlist.lock);
+			return -ENOSPC;
+		}
+	}
+	ntfs_debug("Trying mft data allocation with %s cluster count %lli.",
+			nr > min_nr ? "default" : "minimal", (long long)nr);
+	old_last_vcn = rl[1].vcn;
+	do {
+		rl2 = ntfs_cluster_alloc(vol, old_last_vcn, nr, lcn, MFT_ZONE,
+				true);
+		if (!IS_ERR(rl2))
+			break;
+		if (PTR_ERR(rl2) != -ENOSPC || nr == min_nr) {
+			ntfs_error(vol->sb, "Failed to allocate the minimal "
+					"number of clusters (%lli) for the "
+					"mft data attribute.", (long long)nr);
+			up_write(&mft_ni->runlist.lock);
+			return PTR_ERR(rl2);
+		}
+		/*
+		 * There is not enough space to do the allocation, but there
+		 * might be enough space to do a minimal allocation so try that
+		 * before failing.
+		 */
+		nr = min_nr;
+		ntfs_debug("Retrying mft data allocation with minimal cluster "
+				"count %lli.", (long long)nr);
+	} while (1);
+	rl = ntfs_runlists_merge(mft_ni->runlist.rl, rl2);
+	if (IS_ERR(rl)) {
+		up_write(&mft_ni->runlist.lock);
+		ntfs_error(vol->sb, "Failed to merge runlists for mft data "
+				"attribute.");
+		if (ntfs_cluster_free_from_rl(vol, rl2)) {
+			ntfs_error(vol->sb, "Failed to deallocate clusters "
+					"from the mft data attribute.%s", es);
+			NVolSetErrors(vol);
+		}
+		ntfs_free(rl2);
+		return PTR_ERR(rl);
+	}
+	mft_ni->runlist.rl = rl;
+	ntfs_debug("Allocated %lli clusters.", (long long)nr);
+	/* Find the last run in the new runlist. */
+	for (; rl[1].length; rl++)
+		;
+	/* Update the attribute record as well. */
+	mrec = map_mft_record(mft_ni);
+	if (IS_ERR(mrec)) {
+		ntfs_error(vol->sb, "Failed to map mft record.");
+		ret = PTR_ERR(mrec);
+		goto undo_alloc;
+	}
+	ctx = ntfs_attr_get_search_ctx(mft_ni, mrec);
+	if (unlikely(!ctx)) {
+		ntfs_error(vol->sb, "Failed to get search context.");
+		ret = -ENOMEM;
+		goto undo_alloc;
+	}
+	ret = ntfs_attr_lookup(mft_ni->type, mft_ni->name, mft_ni->name_len,
+			CASE_SENSITIVE, rl[1].vcn, NULL, 0, ctx);
+	if (unlikely(ret)) {
+		ntfs_error(vol->sb, "Failed to find last attribute extent of "
+				"mft data attribute.");
+		if (ret == -ENOENT)
+			ret = -EIO;
+		goto undo_alloc;
+	}
+	a = ctx->attr;
+	ll = sle64_to_cpu(a->data.non_resident.lowest_vcn);
+	/* Search back for the previous last allocated cluster of mft bitmap. */
+	for (rl2 = rl; rl2 > mft_ni->runlist.rl; rl2--) {
+		if (ll >= rl2->vcn)
+			break;
+	}
+	BUG_ON(ll < rl2->vcn);
+	BUG_ON(ll >= rl2->vcn + rl2->length);
+	/* Get the size for the new mapping pairs array for this extent. */
+	mp_size = ntfs_get_size_for_mapping_pairs(vol, rl2, ll, -1);
+	if (unlikely(mp_size <= 0)) {
+		ntfs_error(vol->sb, "Get size for mapping pairs failed for "
+				"mft data attribute extent.");
+		ret = mp_size;
+		if (!ret)
+			ret = -EIO;
+		goto undo_alloc;
+	}
+	/* Expand the attribute record if necessary. */
+	old_alen = le32_to_cpu(a->length);
+	ret = ntfs_attr_record_resize(ctx->mrec, a, mp_size +
+			le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
+	if (unlikely(ret)) {
+		if (ret != -ENOSPC) {
+			ntfs_error(vol->sb, "Failed to resize attribute "
+					"record for mft data attribute.");
+			goto undo_alloc;
+		}
+		// TODO: Deal with this by moving this extent to a new mft
+		// record or by starting a new extent in a new mft record or by
+		// moving other attributes out of this mft record.
+		// Note: Use the special reserved mft records and ensure that
+		// this extent is not required to find the mft record in
+		// question.  If no free special records left we would need to
+		// move an existing record away, insert ours in its place, and
+		// then place the moved record into the newly allocated space
+		// and we would then need to update all references to this mft
+		// record appropriately.  This is rather complicated...
+		ntfs_error(vol->sb, "Not enough space in this mft record to "
+				"accommodate extended mft data attribute "
+				"extent.  Cannot handle this yet.");
+		ret = -EOPNOTSUPP;
+		goto undo_alloc;
+	}
+	mp_rebuilt = true;
+	/* Generate the mapping pairs array directly into the attr record. */
+	ret = ntfs_mapping_pairs_build(vol, (u8*)a +
+			le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
+			mp_size, rl2, ll, -1, NULL);
+	if (unlikely(ret)) {
+		ntfs_error(vol->sb, "Failed to build mapping pairs array of "
+				"mft data attribute.");
+		goto undo_alloc;
+	}
+	/* Update the highest_vcn. */
+	a->data.non_resident.highest_vcn = cpu_to_sle64(rl[1].vcn - 1);
+	/*
+	 * We now have extended the mft data allocated_size by nr clusters.
+	 * Reflect this in the ntfs_inode structure and the attribute record.
+	 * @rl is the last (non-terminator) runlist element of mft data
+	 * attribute.
+	 */
+	if (a->data.non_resident.lowest_vcn) {
+		/*
+		 * We are not in the first attribute extent, switch to it, but
+		 * first ensure the changes will make it to disk later.
+		 */
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		mark_mft_record_dirty(ctx->ntfs_ino);
+		ntfs_attr_reinit_search_ctx(ctx);
+		ret = ntfs_attr_lookup(mft_ni->type, mft_ni->name,
+				mft_ni->name_len, CASE_SENSITIVE, 0, NULL, 0,
+				ctx);
+		if (unlikely(ret)) {
+			ntfs_error(vol->sb, "Failed to find first attribute "
+					"extent of mft data attribute.");
+			goto restore_undo_alloc;
+		}
+		a = ctx->attr;
+	}
+	write_lock_irqsave(&mft_ni->size_lock, flags);
+	mft_ni->allocated_size += nr << vol->cluster_size_bits;
+	a->data.non_resident.allocated_size =
+			cpu_to_sle64(mft_ni->allocated_size);
+	write_unlock_irqrestore(&mft_ni->size_lock, flags);
+	/* Ensure the changes make it to disk. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(mft_ni);
+	up_write(&mft_ni->runlist.lock);
+	ntfs_debug("Done.");
+	return 0;
+restore_undo_alloc:
+	ntfs_attr_reinit_search_ctx(ctx);
+	if (ntfs_attr_lookup(mft_ni->type, mft_ni->name, mft_ni->name_len,
+			CASE_SENSITIVE, rl[1].vcn, NULL, 0, ctx)) {
+		ntfs_error(vol->sb, "Failed to find last attribute extent of "
+				"mft data attribute.%s", es);
+		write_lock_irqsave(&mft_ni->size_lock, flags);
+		mft_ni->allocated_size += nr << vol->cluster_size_bits;
+		write_unlock_irqrestore(&mft_ni->size_lock, flags);
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(mft_ni);
+		up_write(&mft_ni->runlist.lock);
+		/*
+		 * The only thing that is now wrong is ->allocated_size of the
+		 * base attribute extent which chkdsk should be able to fix.
+		 */
+		NVolSetErrors(vol);
+		return ret;
+	}
+	ctx->attr->data.non_resident.highest_vcn =
+			cpu_to_sle64(old_last_vcn - 1);
+undo_alloc:
+	if (ntfs_cluster_free(mft_ni, old_last_vcn, -1, ctx) < 0) {
+		ntfs_error(vol->sb, "Failed to free clusters from mft data "
+				"attribute.%s", es);
+		NVolSetErrors(vol);
+	}
+	a = ctx->attr;
+	if (ntfs_rl_truncate_nolock(vol, &mft_ni->runlist, old_last_vcn)) {
+		ntfs_error(vol->sb, "Failed to truncate mft data attribute "
+				"runlist.%s", es);
+		NVolSetErrors(vol);
+	}
+	if (mp_rebuilt && !IS_ERR(ctx->mrec)) {
+		if (ntfs_mapping_pairs_build(vol, (u8*)a + le16_to_cpu(
+				a->data.non_resident.mapping_pairs_offset),
+				old_alen - le16_to_cpu(
+				a->data.non_resident.mapping_pairs_offset),
+				rl2, ll, -1, NULL)) {
+			ntfs_error(vol->sb, "Failed to restore mapping pairs "
+					"array.%s", es);
+			NVolSetErrors(vol);
+		}
+		if (ntfs_attr_record_resize(ctx->mrec, a, old_alen)) {
+			ntfs_error(vol->sb, "Failed to restore attribute "
+					"record.%s", es);
+			NVolSetErrors(vol);
+		}
+		flush_dcache_mft_record_page(ctx->ntfs_ino);
+		mark_mft_record_dirty(ctx->ntfs_ino);
+	} else if (IS_ERR(ctx->mrec)) {
+		ntfs_error(vol->sb, "Failed to restore attribute search "
+				"context.%s", es);
+		NVolSetErrors(vol);
+	}
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (!IS_ERR(mrec))
+		unmap_mft_record(mft_ni);
+	up_write(&mft_ni->runlist.lock);
+	return ret;
+}
+
+/**
+ * ntfs_mft_record_layout - layout an mft record into a memory buffer
+ * @vol:	volume to which the mft record will belong
+ * @mft_no:	mft reference specifying the mft record number
+ * @m:		destination buffer of size >= @vol->mft_record_size bytes
+ *
+ * Layout an empty, unused mft record with the mft record number @mft_no into
+ * the buffer @m.  The volume @vol is needed because the mft record structure
+ * was modified in NTFS 3.1 so we need to know which volume version this mft
+ * record will be used on.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static int ntfs_mft_record_layout(const ntfs_volume *vol, const s64 mft_no,
+		MFT_RECORD *m)
+{
+	ATTR_RECORD *a;
+
+	ntfs_debug("Entering for mft record 0x%llx.", (long long)mft_no);
+	if (mft_no >= (1ll << 32)) {
+		ntfs_error(vol->sb, "Mft record number 0x%llx exceeds "
+				"maximum of 2^32.", (long long)mft_no);
+		return -ERANGE;
+	}
+	/* Start by clearing the whole mft record to gives us a clean slate. */
+	memset(m, 0, vol->mft_record_size);
+	/* Aligned to 2-byte boundary. */
+	if (vol->major_ver < 3 || (vol->major_ver == 3 && !vol->minor_ver))
+		m->usa_ofs = cpu_to_le16((sizeof(MFT_RECORD_OLD) + 1) & ~1);
+	else {
+		m->usa_ofs = cpu_to_le16((sizeof(MFT_RECORD) + 1) & ~1);
+		/*
+		 * Set the NTFS 3.1+ specific fields while we know that the
+		 * volume version is 3.1+.
+		 */
+		m->reserved = 0;
+		m->mft_record_number = cpu_to_le32((u32)mft_no);
+	}
+	m->magic = magic_FILE;
+	if (vol->mft_record_size >= NTFS_BLOCK_SIZE)
+		m->usa_count = cpu_to_le16(vol->mft_record_size /
+				NTFS_BLOCK_SIZE + 1);
+	else {
+		m->usa_count = cpu_to_le16(1);
+		ntfs_warning(vol->sb, "Sector size is bigger than mft record "
+				"size.  Setting usa_count to 1.  If chkdsk "
+				"reports this as corruption, please email "
+				"linux-ntfs-dev@lists.sourceforge.net stating "
+				"that you saw this message and that the "
+				"modified filesystem created was corrupt.  "
+				"Thank you.");
+	}
+	/* Set the update sequence number to 1. */
+	*(le16*)((u8*)m + le16_to_cpu(m->usa_ofs)) = cpu_to_le16(1);
+	m->lsn = 0;
+	m->sequence_number = cpu_to_le16(1);
+	m->link_count = 0;
+	/*
+	 * Place the attributes straight after the update sequence array,
+	 * aligned to 8-byte boundary.
+	 */
+	m->attrs_offset = cpu_to_le16((le16_to_cpu(m->usa_ofs) +
+			(le16_to_cpu(m->usa_count) << 1) + 7) & ~7);
+	m->flags = 0;
+	/*
+	 * Using attrs_offset plus eight bytes (for the termination attribute).
+	 * attrs_offset is already aligned to 8-byte boundary, so no need to
+	 * align again.
+	 */
+	m->bytes_in_use = cpu_to_le32(le16_to_cpu(m->attrs_offset) + 8);
+	m->bytes_allocated = cpu_to_le32(vol->mft_record_size);
+	m->base_mft_record = 0;
+	m->next_attr_instance = 0;
+	/* Add the termination attribute. */
+	a = (ATTR_RECORD*)((u8*)m + le16_to_cpu(m->attrs_offset));
+	a->type = AT_END;
+	a->length = 0;
+	ntfs_debug("Done.");
+	return 0;
+}
+
+/**
+ * ntfs_mft_record_format - format an mft record on an ntfs volume
+ * @vol:	volume on which to format the mft record
+ * @mft_no:	mft record number to format
+ *
+ * Format the mft record @mft_no in $MFT/$DATA, i.e. lay out an empty, unused
+ * mft record into the appropriate place of the mft data attribute.  This is
+ * used when extending the mft data attribute.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static int ntfs_mft_record_format(const ntfs_volume *vol, const s64 mft_no)
+{
+	loff_t i_size;
+	struct inode *mft_vi = vol->mft_ino;
+	struct page *page;
+	MFT_RECORD *m;
+	pgoff_t index, end_index;
+	unsigned int ofs;
+	int err;
+
+	ntfs_debug("Entering for mft record 0x%llx.", (long long)mft_no);
+	/*
+	 * The index into the page cache and the offset within the page cache
+	 * page of the wanted mft record.
+	 */
+	index = mft_no << vol->mft_record_size_bits >> PAGE_SHIFT;
+	ofs = (mft_no << vol->mft_record_size_bits) & ~PAGE_MASK;
+	/* The maximum valid index into the page cache for $MFT's data. */
+	i_size = i_size_read(mft_vi);
+	end_index = i_size >> PAGE_SHIFT;
+	if (unlikely(index >= end_index)) {
+		if (unlikely(index > end_index || ofs + vol->mft_record_size >=
+				(i_size & ~PAGE_MASK))) {
+			ntfs_error(vol->sb, "Tried to format non-existing mft "
+					"record 0x%llx.", (long long)mft_no);
+			return -ENOENT;
+		}
+	}
+	/* Read, map, and pin the page containing the mft record. */
+	page = ntfs_map_page(mft_vi->i_mapping, index);
+	if (IS_ERR(page)) {
+		ntfs_error(vol->sb, "Failed to map page containing mft record "
+				"to format 0x%llx.", (long long)mft_no);
+		return PTR_ERR(page);
+	}
+	lock_page(page);
+	BUG_ON(!PageUptodate(page));
+	ClearPageUptodate(page);
+	m = (MFT_RECORD*)((u8*)page_address(page) + ofs);
+	err = ntfs_mft_record_layout(vol, mft_no, m);
+	if (unlikely(err)) {
+		ntfs_error(vol->sb, "Failed to layout mft record 0x%llx.",
+				(long long)mft_no);
+		SetPageUptodate(page);
+		unlock_page(page);
+		ntfs_unmap_page(page);
+		return err;
+	}
+	flush_dcache_page(page);
+	SetPageUptodate(page);
+	unlock_page(page);
+	/*
+	 * Make sure the mft record is written out to disk.  We could use
+	 * ilookup5() to check if an inode is in icache and so on but this is
+	 * unnecessary as ntfs_writepage() will write the dirty record anyway.
+	 */
+	mark_ntfs_record_dirty(page, ofs);
+	ntfs_unmap_page(page);
+	ntfs_debug("Done.");
+	return 0;
+}
+
+/**
+ * ntfs_mft_record_alloc - allocate an mft record on an ntfs volume
+ * @vol:	[IN]  volume on which to allocate the mft record
+ * @mode:	[IN]  mode if want a file or directory, i.e. base inode or 0
+ * @base_ni:	[IN]  open base inode if allocating an extent mft record or NULL
+ * @mrec:	[OUT] on successful return this is the mapped mft record
+ *
+ * Allocate an mft record in $MFT/$DATA of an open ntfs volume @vol.
+ *
+ * If @base_ni is NULL make the mft record a base mft record, i.e. a file or
+ * direvctory inode, and allocate it at the default allocator position.  In
+ * this case @mode is the file mode as given to us by the caller.  We in
+ * particular use @mode to distinguish whether a file or a directory is being
+ * created (S_IFDIR(mode) and S_IFREG(mode), respectively).
+ *
+ * If @base_ni is not NULL make the allocated mft record an extent record,
+ * allocate it starting at the mft record after the base mft record and attach
+ * the allocated and opened ntfs inode to the base inode @base_ni.  In this
+ * case @mode must be 0 as it is meaningless for extent inodes.
+ *
+ * You need to check the return value with IS_ERR().  If false, the function
+ * was successful and the return value is the now opened ntfs inode of the
+ * allocated mft record.  *@mrec is then set to the allocated, mapped, pinned,
+ * and locked mft record.  If IS_ERR() is true, the function failed and the
+ * error code is obtained from PTR_ERR(return value).  *@mrec is undefined in
+ * this case.
+ *
+ * Allocation strategy:
+ *
+ * To find a free mft record, we scan the mft bitmap for a zero bit.  To
+ * optimize this we start scanning at the place specified by @base_ni or if
+ * @base_ni is NULL we start where we last stopped and we perform wrap around
+ * when we reach the end.  Note, we do not try to allocate mft records below
+ * number 24 because numbers 0 to 15 are the defined system files anyway and 16
+ * to 24 are special in that they are used for storing extension mft records
+ * for the $DATA attribute of $MFT.  This is required to avoid the possibility
+ * of creating a runlist with a circular dependency which once written to disk
+ * can never be read in again.  Windows will only use records 16 to 24 for
+ * normal files if the volume is completely out of space.  We never use them
+ * which means that when the volume is really out of space we cannot create any
+ * more files while Windows can still create up to 8 small files.  We can start
+ * doing this at some later time, it does not matter much for now.
+ *
+ * When scanning the mft bitmap, we only search up to the last allocated mft
+ * record.  If there are no free records left in the range 24 to number of
+ * allocated mft records, then we extend the $MFT/$DATA attribute in order to
+ * create free mft records.  We extend the allocated size of $MFT/$DATA by 16
+ * records at a time or one cluster, if cluster size is above 16kiB.  If there
+ * is not sufficient space to do this, we try to extend by a single mft record
+ * or one cluster, if cluster size is above the mft record size.
+ *
+ * No matter how many mft records we allocate, we initialize only the first
+ * allocated mft record, incrementing mft data size and initialized size
+ * accordingly, open an ntfs_inode for it and return it to the caller, unless
+ * there are less than 24 mft records, in which case we allocate and initialize
+ * mft records until we reach record 24 which we consider as the first free mft
+ * record for use by normal files.
+ *
+ * If during any stage we overflow the initialized data in the mft bitmap, we
+ * extend the initialized size (and data size) by 8 bytes, allocating another
+ * cluster if required.  The bitmap data size has to be at least equal to the
+ * number of mft records in the mft, but it can be bigger, in which case the
+ * superflous bits are padded with zeroes.
+ *
+ * Thus, when we return successfully (IS_ERR() is false), we will have:
+ *	- initialized / extended the mft bitmap if necessary,
+ *	- initialized / extended the mft data if necessary,
+ *	- set the bit corresponding to the mft record being allocated in the
+ *	  mft bitmap,
+ *	- opened an ntfs_inode for the allocated mft record, and we will have
+ *	- returned the ntfs_inode as well as the allocated mapped, pinned, and
+ *	  locked mft record.
+ *
+ * On error, the volume will be left in a consistent state and no record will
+ * be allocated.  If rolling back a partial operation fails, we may leave some
+ * inconsistent metadata in which case we set NVolErrors() so the volume is
+ * left dirty when unmounted.
+ *
+ * Note, this function cannot make use of most of the normal functions, like
+ * for example for attribute resizing, etc, because when the run list overflows
+ * the base mft record and an attribute list is used, it is very important that
+ * the extension mft records used to store the $DATA attribute of $MFT can be
+ * reached without having to read the information contained inside them, as
+ * this would make it impossible to find them in the first place after the
+ * volume is unmounted.  $MFT/$BITMAP probably does not need to follow this
+ * rule because the bitmap is not essential for finding the mft records, but on
+ * the other hand, handling the bitmap in this special way would make life
+ * easier because otherwise there might be circular invocations of functions
+ * when reading the bitmap.
+ */
+ntfs_inode *ntfs_mft_record_alloc(ntfs_volume *vol, const int mode,
+		ntfs_inode *base_ni, MFT_RECORD **mrec)
+{
+	s64 ll, bit, old_data_initialized, old_data_size;
+	unsigned long flags;
+	struct inode *vi;
+	struct page *page;
+	ntfs_inode *mft_ni, *mftbmp_ni, *ni;
+	ntfs_attr_search_ctx *ctx;
+	MFT_RECORD *m;
+	ATTR_RECORD *a;
+	pgoff_t index;
+	unsigned int ofs;
+	int err;
+	le16 seq_no, usn;
+	bool record_formatted = false;
+
+	if (base_ni) {
+		ntfs_debug("Entering (allocating an extent mft record for "
+				"base mft record 0x%llx).",
+				(long long)base_ni->mft_no);
+		/* @mode and @base_ni are mutually exclusive. */
+		BUG_ON(mode);
+	} else
+		ntfs_debug("Entering (allocating a base mft record).");
+	if (mode) {
+		/* @mode and @base_ni are mutually exclusive. */
+		BUG_ON(base_ni);
+		/* We only support creation of normal files and directories. */
+		if (!S_ISREG(mode) && !S_ISDIR(mode))
+			return ERR_PTR(-EOPNOTSUPP);
+	}
+	BUG_ON(!mrec);
+	mft_ni = NTFS_I(vol->mft_ino);
+	mftbmp_ni = NTFS_I(vol->mftbmp_ino);
+	down_write(&vol->mftbmp_lock);
+	bit = ntfs_mft_bitmap_find_and_alloc_free_rec_nolock(vol, base_ni);
+	if (bit >= 0) {
+		ntfs_debug("Found and allocated free record (#1), bit 0x%llx.",
+				(long long)bit);
+		goto have_alloc_rec;
+	}
+	if (bit != -ENOSPC) {
+		up_write(&vol->mftbmp_lock);
+		return ERR_PTR(bit);
+	}
+	/*
+	 * No free mft records left.  If the mft bitmap already covers more
+	 * than the currently used mft records, the next records are all free,
+	 * so we can simply allocate the first unused mft record.
+	 * Note: We also have to make sure that the mft bitmap at least covers
+	 * the first 24 mft records as they are special and whilst they may not
+	 * be in use, we do not allocate from them.
+	 */
+	read_lock_irqsave(&mft_ni->size_lock, flags);
+	ll = mft_ni->initialized_size >> vol->mft_record_size_bits;
+	read_unlock_irqrestore(&mft_ni->size_lock, flags);
+	read_lock_irqsave(&mftbmp_ni->size_lock, flags);
+	old_data_initialized = mftbmp_ni->initialized_size;
+	read_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+	if (old_data_initialized << 3 > ll && old_data_initialized > 3) {
+		bit = ll;
+		if (bit < 24)
+			bit = 24;
+		if (unlikely(bit >= (1ll << 32)))
+			goto max_err_out;
+		ntfs_debug("Found free record (#2), bit 0x%llx.",
+				(long long)bit);
+		goto found_free_rec;
+	}
+	/*
+	 * The mft bitmap needs to be expanded until it covers the first unused
+	 * mft record that we can allocate.
+	 * Note: The smallest mft record we allocate is mft record 24.
+	 */
+	bit = old_data_initialized << 3;
+	if (unlikely(bit >= (1ll << 32)))
+		goto max_err_out;
+	read_lock_irqsave(&mftbmp_ni->size_lock, flags);
+	old_data_size = mftbmp_ni->allocated_size;
+	ntfs_debug("Status of mftbmp before extension: allocated_size 0x%llx, "
+			"data_size 0x%llx, initialized_size 0x%llx.",
+			(long long)old_data_size,
+			(long long)i_size_read(vol->mftbmp_ino),
+			(long long)old_data_initialized);
+	read_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+	if (old_data_initialized + 8 > old_data_size) {
+		/* Need to extend bitmap by one more cluster. */
+		ntfs_debug("mftbmp: initialized_size + 8 > allocated_size.");
+		err = ntfs_mft_bitmap_extend_allocation_nolock(vol);
+		if (unlikely(err)) {
+			up_write(&vol->mftbmp_lock);
+			goto err_out;
+		}
+#ifdef DEBUG
+		read_lock_irqsave(&mftbmp_ni->size_lock, flags);
+		ntfs_debug("Status of mftbmp after allocation extension: "
+				"allocated_size 0x%llx, data_size 0x%llx, "
+				"initialized_size 0x%llx.",
+				(long long)mftbmp_ni->allocated_size,
+				(long long)i_size_read(vol->mftbmp_ino),
+				(long long)mftbmp_ni->initialized_size);
+		read_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+#endif /* DEBUG */
+	}
+	/*
+	 * We now have sufficient allocated space, extend the initialized_size
+	 * as well as the data_size if necessary and fill the new space with
+	 * zeroes.
+	 */
+	err = ntfs_mft_bitmap_extend_initialized_nolock(vol);
+	if (unlikely(err)) {
+		up_write(&vol->mftbmp_lock);
+		goto err_out;
+	}
+#ifdef DEBUG
+	read_lock_irqsave(&mftbmp_ni->size_lock, flags);
+	ntfs_debug("Status of mftbmp after initialized extension: "
+			"allocated_size 0x%llx, data_size 0x%llx, "
+			"initialized_size 0x%llx.",
+			(long long)mftbmp_ni->allocated_size,
+			(long long)i_size_read(vol->mftbmp_ino),
+			(long long)mftbmp_ni->initialized_size);
+	read_unlock_irqrestore(&mftbmp_ni->size_lock, flags);
+#endif /* DEBUG */
+	ntfs_debug("Found free record (#3), bit 0x%llx.", (long long)bit);
+found_free_rec:
+	/* @bit is the found free mft record, allocate it in the mft bitmap. */
+	ntfs_debug("At found_free_rec.");
+	err = ntfs_bitmap_set_bit(vol->mftbmp_ino, bit);
+	if (unlikely(err)) {
+		ntfs_error(vol->sb, "Failed to allocate bit in mft bitmap.");
+		up_write(&vol->mftbmp_lock);
+		goto err_out;
+	}
+	ntfs_debug("Set bit 0x%llx in mft bitmap.", (long long)bit);
+have_alloc_rec:
+	/*
+	 * The mft bitmap is now uptodate.  Deal with mft data attribute now.
+	 * Note, we keep hold of the mft bitmap lock for writing until all
+	 * modifications to the mft data attribute are complete, too, as they
+	 * will impact decisions for mft bitmap and mft record allocation done
+	 * by a parallel allocation and if the lock is not maintained a
+	 * parallel allocation could allocate the same mft record as this one.
+	 */
+	ll = (bit + 1) << vol->mft_record_size_bits;
+	read_lock_irqsave(&mft_ni->size_lock, flags);
+	old_data_initialized = mft_ni->initialized_size;
+	read_unlock_irqrestore(&mft_ni->size_lock, flags);
+	if (ll <= old_data_initialized) {
+		ntfs_debug("Allocated mft record already initialized.");
+		goto mft_rec_already_initialized;
+	}
+	ntfs_debug("Initializing allocated mft record.");
+	/*
+	 * The mft record is outside the initialized data.  Extend the mft data
+	 * attribute until it covers the allocated record.  The loop is only
+	 * actually traversed more than once when a freshly formatted volume is
+	 * first written to so it optimizes away nicely in the common case.
+	 */
+	read_lock_irqsave(&mft_ni->size_lock, flags);
+	ntfs_debug("Status of mft data before extension: "
+			"allocated_size 0x%llx, data_size 0x%llx, "
+			"initialized_size 0x%llx.",
+			(long long)mft_ni->allocated_size,
+			(long long)i_size_read(vol->mft_ino),
+			(long long)mft_ni->initialized_size);
+	while (ll > mft_ni->allocated_size) {
+		read_unlock_irqrestore(&mft_ni->size_lock, flags);
+		err = ntfs_mft_data_extend_allocation_nolock(vol);
+		if (unlikely(err)) {
+			ntfs_error(vol->sb, "Failed to extend mft data "
+					"allocation.");
+			goto undo_mftbmp_alloc_nolock;
+		}
+		read_lock_irqsave(&mft_ni->size_lock, flags);
+		ntfs_debug("Status of mft data after allocation extension: "
+				"allocated_size 0x%llx, data_size 0x%llx, "
+				"initialized_size 0x%llx.",
+				(long long)mft_ni->allocated_size,
+				(long long)i_size_read(vol->mft_ino),
+				(long long)mft_ni->initialized_size);
+	}
+	read_unlock_irqrestore(&mft_ni->size_lock, flags);
+	/*
+	 * Extend mft data initialized size (and data size of course) to reach
+	 * the allocated mft record, formatting the mft records allong the way.
+	 * Note: We only modify the ntfs_inode structure as that is all that is
+	 * needed by ntfs_mft_record_format().  We will update the attribute
+	 * record itself in one fell swoop later on.
+	 */
+	write_lock_irqsave(&mft_ni->size_lock, flags);
+	old_data_initialized = mft_ni->initialized_size;
+	old_data_size = vol->mft_ino->i_size;
+	while (ll > mft_ni->initialized_size) {
+		s64 new_initialized_size, mft_no;
+		
+		new_initialized_size = mft_ni->initialized_size +
+				vol->mft_record_size;
+		mft_no = mft_ni->initialized_size >> vol->mft_record_size_bits;
+		if (new_initialized_size > i_size_read(vol->mft_ino))
+			i_size_write(vol->mft_ino, new_initialized_size);
+		write_unlock_irqrestore(&mft_ni->size_lock, flags);
+		ntfs_debug("Initializing mft record 0x%llx.",
+				(long long)mft_no);
+		err = ntfs_mft_record_format(vol, mft_no);
+		if (unlikely(err)) {
+			ntfs_error(vol->sb, "Failed to format mft record.");
+			goto undo_data_init;
+		}
+		write_lock_irqsave(&mft_ni->size_lock, flags);
+		mft_ni->initialized_size = new_initialized_size;
+	}
+	write_unlock_irqrestore(&mft_ni->size_lock, flags);
+	record_formatted = true;
+	/* Update the mft data attribute record to reflect the new sizes. */
+	m = map_mft_record(mft_ni);
+	if (IS_ERR(m)) {
+		ntfs_error(vol->sb, "Failed to map mft record.");
+		err = PTR_ERR(m);
+		goto undo_data_init;
+	}
+	ctx = ntfs_attr_get_search_ctx(mft_ni, m);
+	if (unlikely(!ctx)) {
+		ntfs_error(vol->sb, "Failed to get search context.");
+		err = -ENOMEM;
+		unmap_mft_record(mft_ni);
+		goto undo_data_init;
+	}
+	err = ntfs_attr_lookup(mft_ni->type, mft_ni->name, mft_ni->name_len,
+			CASE_SENSITIVE, 0, NULL, 0, ctx);
+	if (unlikely(err)) {
+		ntfs_error(vol->sb, "Failed to find first attribute extent of "
+				"mft data attribute.");
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(mft_ni);
+		goto undo_data_init;
+	}
+	a = ctx->attr;
+	read_lock_irqsave(&mft_ni->size_lock, flags);
+	a->data.non_resident.initialized_size =
+			cpu_to_sle64(mft_ni->initialized_size);
+	a->data.non_resident.data_size =
+			cpu_to_sle64(i_size_read(vol->mft_ino));
+	read_unlock_irqrestore(&mft_ni->size_lock, flags);
+	/* Ensure the changes make it to disk. */
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(mft_ni);
+	read_lock_irqsave(&mft_ni->size_lock, flags);
+	ntfs_debug("Status of mft data after mft record initialization: "
+			"allocated_size 0x%llx, data_size 0x%llx, "
+			"initialized_size 0x%llx.",
+			(long long)mft_ni->allocated_size,
+			(long long)i_size_read(vol->mft_ino),
+			(long long)mft_ni->initialized_size);
+	BUG_ON(i_size_read(vol->mft_ino) > mft_ni->allocated_size);
+	BUG_ON(mft_ni->initialized_size > i_size_read(vol->mft_ino));
+	read_unlock_irqrestore(&mft_ni->size_lock, flags);
+mft_rec_already_initialized:
+	/*
+	 * We can finally drop the mft bitmap lock as the mft data attribute
+	 * has been fully updated.  The only disparity left is that the
+	 * allocated mft record still needs to be marked as in use to match the
+	 * set bit in the mft bitmap but this is actually not a problem since
+	 * this mft record is not referenced from anywhere yet and the fact
+	 * that it is allocated in the mft bitmap means that no-one will try to
+	 * allocate it either.
+	 */
+	up_write(&vol->mftbmp_lock);
+	/*
+	 * We now have allocated and initialized the mft record.  Calculate the
+	 * index of and the offset within the page cache page the record is in.
+	 */
+	index = bit << vol->mft_record_size_bits >> PAGE_SHIFT;
+	ofs = (bit << vol->mft_record_size_bits) & ~PAGE_MASK;
+	/* Read, map, and pin the page containing the mft record. */
+	page = ntfs_map_page(vol->mft_ino->i_mapping, index);
+	if (IS_ERR(page)) {
+		ntfs_error(vol->sb, "Failed to map page containing allocated "
+				"mft record 0x%llx.", (long long)bit);
+		err = PTR_ERR(page);
+		goto undo_mftbmp_alloc;
+	}
+	lock_page(page);
+	BUG_ON(!PageUptodate(page));
+	ClearPageUptodate(page);
+	m = (MFT_RECORD*)((u8*)page_address(page) + ofs);
+	/* If we just formatted the mft record no need to do it again. */
+	if (!record_formatted) {
+		/* Sanity check that the mft record is really not in use. */
+		if (ntfs_is_file_record(m->magic) &&
+				(m->flags & MFT_RECORD_IN_USE)) {
+			ntfs_error(vol->sb, "Mft record 0x%llx was marked "
+					"free in mft bitmap but is marked "
+					"used itself.  Corrupt filesystem.  "
+					"Unmount and run chkdsk.",
+					(long long)bit);
+			err = -EIO;
+			SetPageUptodate(page);
+			unlock_page(page);
+			ntfs_unmap_page(page);
+			NVolSetErrors(vol);
+			goto undo_mftbmp_alloc;
+		}
+		/*
+		 * We need to (re-)format the mft record, preserving the
+		 * sequence number if it is not zero as well as the update
+		 * sequence number if it is not zero or -1 (0xffff).  This
+		 * means we do not need to care whether or not something went
+		 * wrong with the previous mft record.
+		 */
+		seq_no = m->sequence_number;
+		usn = *(le16*)((u8*)m + le16_to_cpu(m->usa_ofs));
+		err = ntfs_mft_record_layout(vol, bit, m);
+		if (unlikely(err)) {
+			ntfs_error(vol->sb, "Failed to layout allocated mft "
+					"record 0x%llx.", (long long)bit);
+			SetPageUptodate(page);
+			unlock_page(page);
+			ntfs_unmap_page(page);
+			goto undo_mftbmp_alloc;
+		}
+		if (seq_no)
+			m->sequence_number = seq_no;
+		if (usn && le16_to_cpu(usn) != 0xffff)
+			*(le16*)((u8*)m + le16_to_cpu(m->usa_ofs)) = usn;
+	}
+	/* Set the mft record itself in use. */
+	m->flags |= MFT_RECORD_IN_USE;
+	if (S_ISDIR(mode))
+		m->flags |= MFT_RECORD_IS_DIRECTORY;
+	flush_dcache_page(page);
+	SetPageUptodate(page);
+	if (base_ni) {
+		MFT_RECORD *m_tmp;
+
+		/*
+		 * Setup the base mft record in the extent mft record.  This
+		 * completes initialization of the allocated extent mft record
+		 * and we can simply use it with map_extent_mft_record().
+		 */
+		m->base_mft_record = MK_LE_MREF(base_ni->mft_no,
+				base_ni->seq_no);
+		/*
+		 * Allocate an extent inode structure for the new mft record,
+		 * attach it to the base inode @base_ni and map, pin, and lock
+		 * its, i.e. the allocated, mft record.
+		 */
+		m_tmp = map_extent_mft_record(base_ni, bit, &ni);
+		if (IS_ERR(m_tmp)) {
+			ntfs_error(vol->sb, "Failed to map allocated extent "
+					"mft record 0x%llx.", (long long)bit);
+			err = PTR_ERR(m_tmp);
+			/* Set the mft record itself not in use. */
+			m->flags &= cpu_to_le16(
+					~le16_to_cpu(MFT_RECORD_IN_USE));
+			flush_dcache_page(page);
+			/* Make sure the mft record is written out to disk. */
+			mark_ntfs_record_dirty(page, ofs);
+			unlock_page(page);
+			ntfs_unmap_page(page);
+			goto undo_mftbmp_alloc;
+		}
+		BUG_ON(m != m_tmp);
+		/*
+		 * Make sure the allocated mft record is written out to disk.
+		 * No need to set the inode dirty because the caller is going
+		 * to do that anyway after finishing with the new extent mft
+		 * record (e.g. at a minimum a new attribute will be added to
+		 * the mft record.
+		 */
+		mark_ntfs_record_dirty(page, ofs);
+		unlock_page(page);
+		/*
+		 * Need to unmap the page since map_extent_mft_record() mapped
+		 * it as well so we have it mapped twice at the moment.
+		 */
+		ntfs_unmap_page(page);
+	} else {
+		/*
+		 * Allocate a new VFS inode and set it up.  NOTE: @vi->i_nlink
+		 * is set to 1 but the mft record->link_count is 0.  The caller
+		 * needs to bear this in mind.
+		 */
+		vi = new_inode(vol->sb);
+		if (unlikely(!vi)) {
+			err = -ENOMEM;
+			/* Set the mft record itself not in use. */
+			m->flags &= cpu_to_le16(
+					~le16_to_cpu(MFT_RECORD_IN_USE));
+			flush_dcache_page(page);
+			/* Make sure the mft record is written out to disk. */
+			mark_ntfs_record_dirty(page, ofs);
+			unlock_page(page);
+			ntfs_unmap_page(page);
+			goto undo_mftbmp_alloc;
+		}
+		vi->i_ino = bit;
+
+		/* The owner and group come from the ntfs volume. */
+		vi->i_uid = vol->uid;
+		vi->i_gid = vol->gid;
+
+		/* Initialize the ntfs specific part of @vi. */
+		ntfs_init_big_inode(vi);
+		ni = NTFS_I(vi);
+		/*
+		 * Set the appropriate mode, attribute type, and name.  For
+		 * directories, also setup the index values to the defaults.
+		 */
+		if (S_ISDIR(mode)) {
+			vi->i_mode = S_IFDIR | S_IRWXUGO;
+			vi->i_mode &= ~vol->dmask;
+
+			NInoSetMstProtected(ni);
+			ni->type = AT_INDEX_ALLOCATION;
+			ni->name = I30;
+			ni->name_len = 4;
+
+			ni->itype.index.block_size = 4096;
+			ni->itype.index.block_size_bits = ntfs_ffs(4096) - 1;
+			ni->itype.index.collation_rule = COLLATION_FILE_NAME;
+			if (vol->cluster_size <= ni->itype.index.block_size) {
+				ni->itype.index.vcn_size = vol->cluster_size;
+				ni->itype.index.vcn_size_bits =
+						vol->cluster_size_bits;
+			} else {
+				ni->itype.index.vcn_size = vol->sector_size;
+				ni->itype.index.vcn_size_bits =
+						vol->sector_size_bits;
+			}
+		} else {
+			vi->i_mode = S_IFREG | S_IRWXUGO;
+			vi->i_mode &= ~vol->fmask;
+
+			ni->type = AT_DATA;
+			ni->name = NULL;
+			ni->name_len = 0;
+		}
+		if (IS_RDONLY(vi))
+			vi->i_mode &= ~S_IWUGO;
+
+		/* Set the inode times to the current time. */
+		vi->i_atime = vi->i_mtime = vi->i_ctime =
+			current_time(vi);
+		/*
+		 * Set the file size to 0, the ntfs inode sizes are set to 0 by
+		 * the call to ntfs_init_big_inode() below.
+		 */
+		vi->i_size = 0;
+		vi->i_blocks = 0;
+
+		/* Set the sequence number. */
+		vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number);
+		/*
+		 * Manually map, pin, and lock the mft record as we already
+		 * have its page mapped and it is very easy to do.
+		 */
+		atomic_inc(&ni->count);
+		mutex_lock(&ni->mrec_lock);
+		ni->page = page;
+		ni->page_ofs = ofs;
+		/*
+		 * Make sure the allocated mft record is written out to disk.
+		 * NOTE: We do not set the ntfs inode dirty because this would
+		 * fail in ntfs_write_inode() because the inode does not have a
+		 * standard information attribute yet.  Also, there is no need
+		 * to set the inode dirty because the caller is going to do
+		 * that anyway after finishing with the new mft record (e.g. at
+		 * a minimum some new attributes will be added to the mft
+		 * record.
+		 */
+		mark_ntfs_record_dirty(page, ofs);
+		unlock_page(page);
+
+		/* Add the inode to the inode hash for the superblock. */
+		insert_inode_hash(vi);
+
+		/* Update the default mft allocation position. */
+		vol->mft_data_pos = bit + 1;
+	}
+	/*
+	 * Return the opened, allocated inode of the allocated mft record as
+	 * well as the mapped, pinned, and locked mft record.
+	 */
+	ntfs_debug("Returning opened, allocated %sinode 0x%llx.",
+			base_ni ? "extent " : "", (long long)bit);
+	*mrec = m;
+	return ni;
+undo_data_init:
+	write_lock_irqsave(&mft_ni->size_lock, flags);
+	mft_ni->initialized_size = old_data_initialized;
+	i_size_write(vol->mft_ino, old_data_size);
+	write_unlock_irqrestore(&mft_ni->size_lock, flags);
+	goto undo_mftbmp_alloc_nolock;
+undo_mftbmp_alloc:
+	down_write(&vol->mftbmp_lock);
+undo_mftbmp_alloc_nolock:
+	if (ntfs_bitmap_clear_bit(vol->mftbmp_ino, bit)) {
+		ntfs_error(vol->sb, "Failed to clear bit in mft bitmap.%s", es);
+		NVolSetErrors(vol);
+	}
+	up_write(&vol->mftbmp_lock);
+err_out:
+	return ERR_PTR(err);
+max_err_out:
+	ntfs_warning(vol->sb, "Cannot allocate mft record because the maximum "
+			"number of inodes (2^32) has already been reached.");
+	up_write(&vol->mftbmp_lock);
+	return ERR_PTR(-ENOSPC);
+}
+
+/**
+ * ntfs_extent_mft_record_free - free an extent mft record on an ntfs volume
+ * @ni:		ntfs inode of the mapped extent mft record to free
+ * @m:		mapped extent mft record of the ntfs inode @ni
+ *
+ * Free the mapped extent mft record @m of the extent ntfs inode @ni.
+ *
+ * Note that this function unmaps the mft record and closes and destroys @ni
+ * internally and hence you cannot use either @ni nor @m any more after this
+ * function returns success.
+ *
+ * On success return 0 and on error return -errno.  @ni and @m are still valid
+ * in this case and have not been freed.
+ *
+ * For some errors an error message is displayed and the success code 0 is
+ * returned and the volume is then left dirty on umount.  This makes sense in
+ * case we could not rollback the changes that were already done since the
+ * caller no longer wants to reference this mft record so it does not matter to
+ * the caller if something is wrong with it as long as it is properly detached
+ * from the base inode.
+ */
+int ntfs_extent_mft_record_free(ntfs_inode *ni, MFT_RECORD *m)
+{
+	unsigned long mft_no = ni->mft_no;
+	ntfs_volume *vol = ni->vol;
+	ntfs_inode *base_ni;
+	ntfs_inode **extent_nis;
+	int i, err;
+	le16 old_seq_no;
+	u16 seq_no;
+	
+	BUG_ON(NInoAttr(ni));
+	BUG_ON(ni->nr_extents != -1);
+
+	mutex_lock(&ni->extent_lock);
+	base_ni = ni->ext.base_ntfs_ino;
+	mutex_unlock(&ni->extent_lock);
+
+	BUG_ON(base_ni->nr_extents <= 0);
+
+	ntfs_debug("Entering for extent inode 0x%lx, base inode 0x%lx.\n",
+			mft_no, base_ni->mft_no);
+
+	mutex_lock(&base_ni->extent_lock);
+
+	/* Make sure we are holding the only reference to the extent inode. */
+	if (atomic_read(&ni->count) > 2) {
+		ntfs_error(vol->sb, "Tried to free busy extent inode 0x%lx, "
+				"not freeing.", base_ni->mft_no);
+		mutex_unlock(&base_ni->extent_lock);
+		return -EBUSY;
+	}
+
+	/* Dissociate the ntfs inode from the base inode. */
+	extent_nis = base_ni->ext.extent_ntfs_inos;
+	err = -ENOENT;
+	for (i = 0; i < base_ni->nr_extents; i++) {
+		if (ni != extent_nis[i])
+			continue;
+		extent_nis += i;
+		base_ni->nr_extents--;
+		memmove(extent_nis, extent_nis + 1, (base_ni->nr_extents - i) *
+				sizeof(ntfs_inode*));
+		err = 0;
+		break;
+	}
+
+	mutex_unlock(&base_ni->extent_lock);
+
+	if (unlikely(err)) {
+		ntfs_error(vol->sb, "Extent inode 0x%lx is not attached to "
+				"its base inode 0x%lx.", mft_no,
+				base_ni->mft_no);
+		BUG();
+	}
+
+	/*
+	 * The extent inode is no longer attached to the base inode so no one
+	 * can get a reference to it any more.
+	 */
+
+	/* Mark the mft record as not in use. */
+	m->flags &= ~MFT_RECORD_IN_USE;
+
+	/* Increment the sequence number, skipping zero, if it is not zero. */
+	old_seq_no = m->sequence_number;
+	seq_no = le16_to_cpu(old_seq_no);
+	if (seq_no == 0xffff)
+		seq_no = 1;
+	else if (seq_no)
+		seq_no++;
+	m->sequence_number = cpu_to_le16(seq_no);
+
+	/*
+	 * Set the ntfs inode dirty and write it out.  We do not need to worry
+	 * about the base inode here since whatever caused the extent mft
+	 * record to be freed is guaranteed to do it already.
+	 */
+	NInoSetDirty(ni);
+	err = write_mft_record(ni, m, 0);
+	if (unlikely(err)) {
+		ntfs_error(vol->sb, "Failed to write mft record 0x%lx, not "
+				"freeing.", mft_no);
+		goto rollback;
+	}
+rollback_error:
+	/* Unmap and throw away the now freed extent inode. */
+	unmap_extent_mft_record(ni);
+	ntfs_clear_extent_inode(ni);
+
+	/* Clear the bit in the $MFT/$BITMAP corresponding to this record. */
+	down_write(&vol->mftbmp_lock);
+	err = ntfs_bitmap_clear_bit(vol->mftbmp_ino, mft_no);
+	up_write(&vol->mftbmp_lock);
+	if (unlikely(err)) {
+		/*
+		 * The extent inode is gone but we failed to deallocate it in
+		 * the mft bitmap.  Just emit a warning and leave the volume
+		 * dirty on umount.
+		 */
+		ntfs_error(vol->sb, "Failed to clear bit in mft bitmap.%s", es);
+		NVolSetErrors(vol);
+	}
+	return 0;
+rollback:
+	/* Rollback what we did... */
+	mutex_lock(&base_ni->extent_lock);
+	extent_nis = base_ni->ext.extent_ntfs_inos;
+	if (!(base_ni->nr_extents & 3)) {
+		int new_size = (base_ni->nr_extents + 4) * sizeof(ntfs_inode*);
+
+		extent_nis = kmalloc(new_size, GFP_NOFS);
+		if (unlikely(!extent_nis)) {
+			ntfs_error(vol->sb, "Failed to allocate internal "
+					"buffer during rollback.%s", es);
+			mutex_unlock(&base_ni->extent_lock);
+			NVolSetErrors(vol);
+			goto rollback_error;
+		}
+		if (base_ni->nr_extents) {
+			BUG_ON(!base_ni->ext.extent_ntfs_inos);
+			memcpy(extent_nis, base_ni->ext.extent_ntfs_inos,
+					new_size - 4 * sizeof(ntfs_inode*));
+			kfree(base_ni->ext.extent_ntfs_inos);
+		}
+		base_ni->ext.extent_ntfs_inos = extent_nis;
+	}
+	m->flags |= MFT_RECORD_IN_USE;
+	m->sequence_number = old_seq_no;
+	extent_nis[base_ni->nr_extents++] = ni;
+	mutex_unlock(&base_ni->extent_lock);
+	mark_mft_record_dirty(ni);
+	return err;
+}
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/mft.h b/fs/ntfs/mft.h
new file mode 100644
index 000000000..17bfefc30
--- /dev/null
+++ b/fs/ntfs/mft.h
@@ -0,0 +1,110 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * mft.h - Defines for mft record handling in NTFS Linux kernel driver.
+ *	   Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_MFT_H
+#define _LINUX_NTFS_MFT_H
+
+#include <linux/fs.h>
+#include <linux/highmem.h>
+#include <linux/pagemap.h>
+
+#include "inode.h"
+
+extern MFT_RECORD *map_mft_record(ntfs_inode *ni);
+extern void unmap_mft_record(ntfs_inode *ni);
+
+extern MFT_RECORD *map_extent_mft_record(ntfs_inode *base_ni, MFT_REF mref,
+		ntfs_inode **ntfs_ino);
+
+static inline void unmap_extent_mft_record(ntfs_inode *ni)
+{
+	unmap_mft_record(ni);
+	return;
+}
+
+#ifdef NTFS_RW
+
+/**
+ * flush_dcache_mft_record_page - flush_dcache_page() for mft records
+ * @ni:		ntfs inode structure of mft record
+ *
+ * Call flush_dcache_page() for the page in which an mft record resides.
+ *
+ * This must be called every time an mft record is modified, just after the
+ * modification.
+ */
+static inline void flush_dcache_mft_record_page(ntfs_inode *ni)
+{
+	flush_dcache_page(ni->page);
+}
+
+extern void __mark_mft_record_dirty(ntfs_inode *ni);
+
+/**
+ * mark_mft_record_dirty - set the mft record and the page containing it dirty
+ * @ni:		ntfs inode describing the mapped mft record
+ *
+ * Set the mapped (extent) mft record of the (base or extent) ntfs inode @ni,
+ * as well as the page containing the mft record, dirty.  Also, mark the base
+ * vfs inode dirty.  This ensures that any changes to the mft record are
+ * written out to disk.
+ *
+ * NOTE:  Do not do anything if the mft record is already marked dirty.
+ */
+static inline void mark_mft_record_dirty(ntfs_inode *ni)
+{
+	if (!NInoTestSetDirty(ni))
+		__mark_mft_record_dirty(ni);
+}
+
+extern int ntfs_sync_mft_mirror(ntfs_volume *vol, const unsigned long mft_no,
+		MFT_RECORD *m, int sync);
+
+extern int write_mft_record_nolock(ntfs_inode *ni, MFT_RECORD *m, int sync);
+
+/**
+ * write_mft_record - write out a mapped (extent) mft record
+ * @ni:		ntfs inode describing the mapped (extent) mft record
+ * @m:		mapped (extent) mft record to write
+ * @sync:	if true, wait for i/o completion
+ *
+ * This is just a wrapper for write_mft_record_nolock() (see mft.c), which
+ * locks the page for the duration of the write.  This ensures that there are
+ * no race conditions between writing the mft record via the dirty inode code
+ * paths and via the page cache write back code paths or between writing
+ * neighbouring mft records residing in the same page.
+ *
+ * Locking the page also serializes us against ->readpage() if the page is not
+ * uptodate.
+ *
+ * On success, clean the mft record and return 0.  On error, leave the mft
+ * record dirty and return -errno.
+ */
+static inline int write_mft_record(ntfs_inode *ni, MFT_RECORD *m, int sync)
+{
+	struct page *page = ni->page;
+	int err;
+
+	BUG_ON(!page);
+	lock_page(page);
+	err = write_mft_record_nolock(ni, m, sync);
+	unlock_page(page);
+	return err;
+}
+
+extern bool ntfs_may_write_mft_record(ntfs_volume *vol,
+		const unsigned long mft_no, const MFT_RECORD *m,
+		ntfs_inode **locked_ni);
+
+extern ntfs_inode *ntfs_mft_record_alloc(ntfs_volume *vol, const int mode,
+		ntfs_inode *base_ni, MFT_RECORD **mrec);
+extern int ntfs_extent_mft_record_free(ntfs_inode *ni, MFT_RECORD *m);
+
+#endif /* NTFS_RW */
+
+#endif /* _LINUX_NTFS_MFT_H */
diff --git a/fs/ntfs/mst.c b/fs/ntfs/mst.c
new file mode 100644
index 000000000..16b3c884a
--- /dev/null
+++ b/fs/ntfs/mst.c
@@ -0,0 +1,189 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * mst.c - NTFS multi sector transfer protection handling code. Part of the
+ *	   Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2004 Anton Altaparmakov
+ */
+
+#include "ntfs.h"
+
+/**
+ * post_read_mst_fixup - deprotect multi sector transfer protected data
+ * @b:		pointer to the data to deprotect
+ * @size:	size in bytes of @b
+ *
+ * Perform the necessary post read multi sector transfer fixup and detect the
+ * presence of incomplete multi sector transfers. - In that case, overwrite the
+ * magic of the ntfs record header being processed with "BAAD" (in memory only!)
+ * and abort processing.
+ *
+ * Return 0 on success and -EINVAL on error ("BAAD" magic will be present).
+ *
+ * NOTE: We consider the absence / invalidity of an update sequence array to
+ * mean that the structure is not protected at all and hence doesn't need to
+ * be fixed up. Thus, we return success and not failure in this case. This is
+ * in contrast to pre_write_mst_fixup(), see below.
+ */
+int post_read_mst_fixup(NTFS_RECORD *b, const u32 size)
+{
+	u16 usa_ofs, usa_count, usn;
+	u16 *usa_pos, *data_pos;
+
+	/* Setup the variables. */
+	usa_ofs = le16_to_cpu(b->usa_ofs);
+	/* Decrement usa_count to get number of fixups. */
+	usa_count = le16_to_cpu(b->usa_count) - 1;
+	/* Size and alignment checks. */
+	if ( size & (NTFS_BLOCK_SIZE - 1)	||
+	     usa_ofs & 1			||
+	     usa_ofs + (usa_count * 2) > size	||
+	     (size >> NTFS_BLOCK_SIZE_BITS) != usa_count)
+		return 0;
+	/* Position of usn in update sequence array. */
+	usa_pos = (u16*)b + usa_ofs/sizeof(u16);
+	/*
+	 * The update sequence number which has to be equal to each of the
+	 * u16 values before they are fixed up. Note no need to care for
+	 * endianness since we are comparing and moving data for on disk
+	 * structures which means the data is consistent. - If it is
+	 * consistenty the wrong endianness it doesn't make any difference.
+	 */
+	usn = *usa_pos;
+	/*
+	 * Position in protected data of first u16 that needs fixing up.
+	 */
+	data_pos = (u16*)b + NTFS_BLOCK_SIZE/sizeof(u16) - 1;
+	/*
+	 * Check for incomplete multi sector transfer(s).
+	 */
+	while (usa_count--) {
+		if (*data_pos != usn) {
+			/*
+			 * Incomplete multi sector transfer detected! )-:
+			 * Set the magic to "BAAD" and return failure.
+			 * Note that magic_BAAD is already converted to le32.
+			 */
+			b->magic = magic_BAAD;
+			return -EINVAL;
+		}
+		data_pos += NTFS_BLOCK_SIZE/sizeof(u16);
+	}
+	/* Re-setup the variables. */
+	usa_count = le16_to_cpu(b->usa_count) - 1;
+	data_pos = (u16*)b + NTFS_BLOCK_SIZE/sizeof(u16) - 1;
+	/* Fixup all sectors. */
+	while (usa_count--) {
+		/*
+		 * Increment position in usa and restore original data from
+		 * the usa into the data buffer.
+		 */
+		*data_pos = *(++usa_pos);
+		/* Increment position in data as well. */
+		data_pos += NTFS_BLOCK_SIZE/sizeof(u16);
+	}
+	return 0;
+}
+
+/**
+ * pre_write_mst_fixup - apply multi sector transfer protection
+ * @b:		pointer to the data to protect
+ * @size:	size in bytes of @b
+ *
+ * Perform the necessary pre write multi sector transfer fixup on the data
+ * pointer to by @b of @size.
+ *
+ * Return 0 if fixup applied (success) or -EINVAL if no fixup was performed
+ * (assumed not needed). This is in contrast to post_read_mst_fixup() above.
+ *
+ * NOTE: We consider the absence / invalidity of an update sequence array to
+ * mean that the structure is not subject to protection and hence doesn't need
+ * to be fixed up. This means that you have to create a valid update sequence
+ * array header in the ntfs record before calling this function, otherwise it
+ * will fail (the header needs to contain the position of the update sequence
+ * array together with the number of elements in the array). You also need to
+ * initialise the update sequence number before calling this function
+ * otherwise a random word will be used (whatever was in the record at that
+ * position at that time).
+ */
+int pre_write_mst_fixup(NTFS_RECORD *b, const u32 size)
+{
+	le16 *usa_pos, *data_pos;
+	u16 usa_ofs, usa_count, usn;
+	le16 le_usn;
+
+	/* Sanity check + only fixup if it makes sense. */
+	if (!b || ntfs_is_baad_record(b->magic) ||
+			ntfs_is_hole_record(b->magic))
+		return -EINVAL;
+	/* Setup the variables. */
+	usa_ofs = le16_to_cpu(b->usa_ofs);
+	/* Decrement usa_count to get number of fixups. */
+	usa_count = le16_to_cpu(b->usa_count) - 1;
+	/* Size and alignment checks. */
+	if ( size & (NTFS_BLOCK_SIZE - 1)	||
+	     usa_ofs & 1			||
+	     usa_ofs + (usa_count * 2) > size	||
+	     (size >> NTFS_BLOCK_SIZE_BITS) != usa_count)
+		return -EINVAL;
+	/* Position of usn in update sequence array. */
+	usa_pos = (le16*)((u8*)b + usa_ofs);
+	/*
+	 * Cyclically increment the update sequence number
+	 * (skipping 0 and -1, i.e. 0xffff).
+	 */
+	usn = le16_to_cpup(usa_pos) + 1;
+	if (usn == 0xffff || !usn)
+		usn = 1;
+	le_usn = cpu_to_le16(usn);
+	*usa_pos = le_usn;
+	/* Position in data of first u16 that needs fixing up. */
+	data_pos = (le16*)b + NTFS_BLOCK_SIZE/sizeof(le16) - 1;
+	/* Fixup all sectors. */
+	while (usa_count--) {
+		/*
+		 * Increment the position in the usa and save the
+		 * original data from the data buffer into the usa.
+		 */
+		*(++usa_pos) = *data_pos;
+		/* Apply fixup to data. */
+		*data_pos = le_usn;
+		/* Increment position in data as well. */
+		data_pos += NTFS_BLOCK_SIZE/sizeof(le16);
+	}
+	return 0;
+}
+
+/**
+ * post_write_mst_fixup - fast deprotect multi sector transfer protected data
+ * @b:		pointer to the data to deprotect
+ *
+ * Perform the necessary post write multi sector transfer fixup, not checking
+ * for any errors, because we assume we have just used pre_write_mst_fixup(),
+ * thus the data will be fine or we would never have gotten here.
+ */
+void post_write_mst_fixup(NTFS_RECORD *b)
+{
+	le16 *usa_pos, *data_pos;
+
+	u16 usa_ofs = le16_to_cpu(b->usa_ofs);
+	u16 usa_count = le16_to_cpu(b->usa_count) - 1;
+
+	/* Position of usn in update sequence array. */
+	usa_pos = (le16*)b + usa_ofs/sizeof(le16);
+
+	/* Position in protected data of first u16 that needs fixing up. */
+	data_pos = (le16*)b + NTFS_BLOCK_SIZE/sizeof(le16) - 1;
+
+	/* Fixup all sectors. */
+	while (usa_count--) {
+		/*
+		 * Increment position in usa and restore original data from
+		 * the usa into the data buffer.
+		 */
+		*data_pos = *(++usa_pos);
+
+		/* Increment position in data as well. */
+		data_pos += NTFS_BLOCK_SIZE/sizeof(le16);
+	}
+}
diff --git a/fs/ntfs/namei.c b/fs/ntfs/namei.c
new file mode 100644
index 000000000..4e6a44bc6
--- /dev/null
+++ b/fs/ntfs/namei.c
@@ -0,0 +1,391 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * namei.c - NTFS kernel directory inode operations. Part of the Linux-NTFS
+ *	     project.
+ *
+ * Copyright (c) 2001-2006 Anton Altaparmakov
+ */
+
+#include <linux/dcache.h>
+#include <linux/exportfs.h>
+#include <linux/security.h>
+#include <linux/slab.h>
+
+#include "attrib.h"
+#include "debug.h"
+#include "dir.h"
+#include "mft.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_lookup - find the inode represented by a dentry in a directory inode
+ * @dir_ino:	directory inode in which to look for the inode
+ * @dent:	dentry representing the inode to look for
+ * @flags:	lookup flags
+ *
+ * In short, ntfs_lookup() looks for the inode represented by the dentry @dent
+ * in the directory inode @dir_ino and if found attaches the inode to the
+ * dentry @dent.
+ *
+ * In more detail, the dentry @dent specifies which inode to look for by
+ * supplying the name of the inode in @dent->d_name.name. ntfs_lookup()
+ * converts the name to Unicode and walks the contents of the directory inode
+ * @dir_ino looking for the converted Unicode name. If the name is found in the
+ * directory, the corresponding inode is loaded by calling ntfs_iget() on its
+ * inode number and the inode is associated with the dentry @dent via a call to
+ * d_splice_alias().
+ *
+ * If the name is not found in the directory, a NULL inode is inserted into the
+ * dentry @dent via a call to d_add(). The dentry is then termed a negative
+ * dentry.
+ *
+ * Only if an actual error occurs, do we return an error via ERR_PTR().
+ *
+ * In order to handle the case insensitivity issues of NTFS with regards to the
+ * dcache and the dcache requiring only one dentry per directory, we deal with
+ * dentry aliases that only differ in case in ->ntfs_lookup() while maintaining
+ * a case sensitive dcache. This means that we get the full benefit of dcache
+ * speed when the file/directory is looked up with the same case as returned by
+ * ->ntfs_readdir() but that a lookup for any other case (or for the short file
+ * name) will not find anything in dcache and will enter ->ntfs_lookup()
+ * instead, where we search the directory for a fully matching file name
+ * (including case) and if that is not found, we search for a file name that
+ * matches with different case and if that has non-POSIX semantics we return
+ * that. We actually do only one search (case sensitive) and keep tabs on
+ * whether we have found a case insensitive match in the process.
+ *
+ * To simplify matters for us, we do not treat the short vs long filenames as
+ * two hard links but instead if the lookup matches a short filename, we
+ * return the dentry for the corresponding long filename instead.
+ *
+ * There are three cases we need to distinguish here:
+ *
+ * 1) @dent perfectly matches (i.e. including case) a directory entry with a
+ *    file name in the WIN32 or POSIX namespaces. In this case
+ *    ntfs_lookup_inode_by_name() will return with name set to NULL and we
+ *    just d_splice_alias() @dent.
+ * 2) @dent matches (not including case) a directory entry with a file name in
+ *    the WIN32 namespace. In this case ntfs_lookup_inode_by_name() will return
+ *    with name set to point to a kmalloc()ed ntfs_name structure containing
+ *    the properly cased little endian Unicode name. We convert the name to the
+ *    current NLS code page, search if a dentry with this name already exists
+ *    and if so return that instead of @dent.  At this point things are
+ *    complicated by the possibility of 'disconnected' dentries due to NFS
+ *    which we deal with appropriately (see the code comments).  The VFS will
+ *    then destroy the old @dent and use the one we returned.  If a dentry is
+ *    not found, we allocate a new one, d_splice_alias() it, and return it as
+ *    above.
+ * 3) @dent matches either perfectly or not (i.e. we don't care about case) a
+ *    directory entry with a file name in the DOS namespace. In this case
+ *    ntfs_lookup_inode_by_name() will return with name set to point to a
+ *    kmalloc()ed ntfs_name structure containing the mft reference (cpu endian)
+ *    of the inode. We use the mft reference to read the inode and to find the
+ *    file name in the WIN32 namespace corresponding to the matched short file
+ *    name. We then convert the name to the current NLS code page, and proceed
+ *    searching for a dentry with this name, etc, as in case 2), above.
+ *
+ * Locking: Caller must hold i_mutex on the directory.
+ */
+static struct dentry *ntfs_lookup(struct inode *dir_ino, struct dentry *dent,
+		unsigned int flags)
+{
+	ntfs_volume *vol = NTFS_SB(dir_ino->i_sb);
+	struct inode *dent_inode;
+	ntfschar *uname;
+	ntfs_name *name = NULL;
+	MFT_REF mref;
+	unsigned long dent_ino;
+	int uname_len;
+
+	ntfs_debug("Looking up %pd in directory inode 0x%lx.",
+			dent, dir_ino->i_ino);
+	/* Convert the name of the dentry to Unicode. */
+	uname_len = ntfs_nlstoucs(vol, dent->d_name.name, dent->d_name.len,
+			&uname);
+	if (uname_len < 0) {
+		if (uname_len != -ENAMETOOLONG)
+			ntfs_error(vol->sb, "Failed to convert name to "
+					"Unicode.");
+		return ERR_PTR(uname_len);
+	}
+	mref = ntfs_lookup_inode_by_name(NTFS_I(dir_ino), uname, uname_len,
+			&name);
+	kmem_cache_free(ntfs_name_cache, uname);
+	if (!IS_ERR_MREF(mref)) {
+		dent_ino = MREF(mref);
+		ntfs_debug("Found inode 0x%lx. Calling ntfs_iget.", dent_ino);
+		dent_inode = ntfs_iget(vol->sb, dent_ino);
+		if (!IS_ERR(dent_inode)) {
+			/* Consistency check. */
+			if (is_bad_inode(dent_inode) || MSEQNO(mref) ==
+					NTFS_I(dent_inode)->seq_no ||
+					dent_ino == FILE_MFT) {
+				/* Perfect WIN32/POSIX match. -- Case 1. */
+				if (!name) {
+					ntfs_debug("Done.  (Case 1.)");
+					return d_splice_alias(dent_inode, dent);
+				}
+				/*
+				 * We are too indented.  Handle imperfect
+				 * matches and short file names further below.
+				 */
+				goto handle_name;
+			}
+			ntfs_error(vol->sb, "Found stale reference to inode "
+					"0x%lx (reference sequence number = "
+					"0x%x, inode sequence number = 0x%x), "
+					"returning -EIO. Run chkdsk.",
+					dent_ino, MSEQNO(mref),
+					NTFS_I(dent_inode)->seq_no);
+			iput(dent_inode);
+			dent_inode = ERR_PTR(-EIO);
+		} else
+			ntfs_error(vol->sb, "ntfs_iget(0x%lx) failed with "
+					"error code %li.", dent_ino,
+					PTR_ERR(dent_inode));
+		kfree(name);
+		/* Return the error code. */
+		return ERR_CAST(dent_inode);
+	}
+	/* It is guaranteed that @name is no longer allocated at this point. */
+	if (MREF_ERR(mref) == -ENOENT) {
+		ntfs_debug("Entry was not found, adding negative dentry.");
+		/* The dcache will handle negative entries. */
+		d_add(dent, NULL);
+		ntfs_debug("Done.");
+		return NULL;
+	}
+	ntfs_error(vol->sb, "ntfs_lookup_ino_by_name() failed with error "
+			"code %i.", -MREF_ERR(mref));
+	return ERR_PTR(MREF_ERR(mref));
+	// TODO: Consider moving this lot to a separate function! (AIA)
+handle_name:
+   {
+	MFT_RECORD *m;
+	ntfs_attr_search_ctx *ctx;
+	ntfs_inode *ni = NTFS_I(dent_inode);
+	int err;
+	struct qstr nls_name;
+
+	nls_name.name = NULL;
+	if (name->type != FILE_NAME_DOS) {			/* Case 2. */
+		ntfs_debug("Case 2.");
+		nls_name.len = (unsigned)ntfs_ucstonls(vol,
+				(ntfschar*)&name->name, name->len,
+				(unsigned char**)&nls_name.name, 0);
+		kfree(name);
+	} else /* if (name->type == FILE_NAME_DOS) */ {		/* Case 3. */
+		FILE_NAME_ATTR *fn;
+
+		ntfs_debug("Case 3.");
+		kfree(name);
+
+		/* Find the WIN32 name corresponding to the matched DOS name. */
+		ni = NTFS_I(dent_inode);
+		m = map_mft_record(ni);
+		if (IS_ERR(m)) {
+			err = PTR_ERR(m);
+			m = NULL;
+			ctx = NULL;
+			goto err_out;
+		}
+		ctx = ntfs_attr_get_search_ctx(ni, m);
+		if (unlikely(!ctx)) {
+			err = -ENOMEM;
+			goto err_out;
+		}
+		do {
+			ATTR_RECORD *a;
+			u32 val_len;
+
+			err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0,
+					NULL, 0, ctx);
+			if (unlikely(err)) {
+				ntfs_error(vol->sb, "Inode corrupt: No WIN32 "
+						"namespace counterpart to DOS "
+						"file name. Run chkdsk.");
+				if (err == -ENOENT)
+					err = -EIO;
+				goto err_out;
+			}
+			/* Consistency checks. */
+			a = ctx->attr;
+			if (a->non_resident || a->flags)
+				goto eio_err_out;
+			val_len = le32_to_cpu(a->data.resident.value_length);
+			if (le16_to_cpu(a->data.resident.value_offset) +
+					val_len > le32_to_cpu(a->length))
+				goto eio_err_out;
+			fn = (FILE_NAME_ATTR*)((u8*)ctx->attr + le16_to_cpu(
+					ctx->attr->data.resident.value_offset));
+			if ((u32)(fn->file_name_length * sizeof(ntfschar) +
+					sizeof(FILE_NAME_ATTR)) > val_len)
+				goto eio_err_out;
+		} while (fn->file_name_type != FILE_NAME_WIN32);
+
+		/* Convert the found WIN32 name to current NLS code page. */
+		nls_name.len = (unsigned)ntfs_ucstonls(vol,
+				(ntfschar*)&fn->file_name, fn->file_name_length,
+				(unsigned char**)&nls_name.name, 0);
+
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(ni);
+	}
+	m = NULL;
+	ctx = NULL;
+
+	/* Check if a conversion error occurred. */
+	if ((signed)nls_name.len < 0) {
+		err = (signed)nls_name.len;
+		goto err_out;
+	}
+	nls_name.hash = full_name_hash(dent, nls_name.name, nls_name.len);
+
+	dent = d_add_ci(dent, dent_inode, &nls_name);
+	kfree(nls_name.name);
+	return dent;
+
+eio_err_out:
+	ntfs_error(vol->sb, "Illegal file name attribute. Run chkdsk.");
+	err = -EIO;
+err_out:
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	if (m)
+		unmap_mft_record(ni);
+	iput(dent_inode);
+	ntfs_error(vol->sb, "Failed, returning error code %i.", err);
+	return ERR_PTR(err);
+   }
+}
+
+/**
+ * Inode operations for directories.
+ */
+const struct inode_operations ntfs_dir_inode_ops = {
+	.lookup	= ntfs_lookup,	/* VFS: Lookup directory. */
+};
+
+/**
+ * ntfs_get_parent - find the dentry of the parent of a given directory dentry
+ * @child_dent:		dentry of the directory whose parent directory to find
+ *
+ * Find the dentry for the parent directory of the directory specified by the
+ * dentry @child_dent.  This function is called from
+ * fs/exportfs/expfs.c::find_exported_dentry() which in turn is called from the
+ * default ->decode_fh() which is export_decode_fh() in the same file.
+ *
+ * The code is based on the ext3 ->get_parent() implementation found in
+ * fs/ext3/namei.c::ext3_get_parent().
+ *
+ * Note: ntfs_get_parent() is called with @d_inode(child_dent)->i_mutex down.
+ *
+ * Return the dentry of the parent directory on success or the error code on
+ * error (IS_ERR() is true).
+ */
+static struct dentry *ntfs_get_parent(struct dentry *child_dent)
+{
+	struct inode *vi = d_inode(child_dent);
+	ntfs_inode *ni = NTFS_I(vi);
+	MFT_RECORD *mrec;
+	ntfs_attr_search_ctx *ctx;
+	ATTR_RECORD *attr;
+	FILE_NAME_ATTR *fn;
+	unsigned long parent_ino;
+	int err;
+
+	ntfs_debug("Entering for inode 0x%lx.", vi->i_ino);
+	/* Get the mft record of the inode belonging to the child dentry. */
+	mrec = map_mft_record(ni);
+	if (IS_ERR(mrec))
+		return ERR_CAST(mrec);
+	/* Find the first file name attribute in the mft record. */
+	ctx = ntfs_attr_get_search_ctx(ni, mrec);
+	if (unlikely(!ctx)) {
+		unmap_mft_record(ni);
+		return ERR_PTR(-ENOMEM);
+	}
+try_next:
+	err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, CASE_SENSITIVE, 0, NULL,
+			0, ctx);
+	if (unlikely(err)) {
+		ntfs_attr_put_search_ctx(ctx);
+		unmap_mft_record(ni);
+		if (err == -ENOENT)
+			ntfs_error(vi->i_sb, "Inode 0x%lx does not have a "
+					"file name attribute.  Run chkdsk.",
+					vi->i_ino);
+		return ERR_PTR(err);
+	}
+	attr = ctx->attr;
+	if (unlikely(attr->non_resident))
+		goto try_next;
+	fn = (FILE_NAME_ATTR *)((u8 *)attr +
+			le16_to_cpu(attr->data.resident.value_offset));
+	if (unlikely((u8 *)fn + le32_to_cpu(attr->data.resident.value_length) >
+			(u8*)attr + le32_to_cpu(attr->length)))
+		goto try_next;
+	/* Get the inode number of the parent directory. */
+	parent_ino = MREF_LE(fn->parent_directory);
+	/* Release the search context and the mft record of the child. */
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(ni);
+
+	return d_obtain_alias(ntfs_iget(vi->i_sb, parent_ino));
+}
+
+static struct inode *ntfs_nfs_get_inode(struct super_block *sb,
+		u64 ino, u32 generation)
+{
+	struct inode *inode;
+
+	inode = ntfs_iget(sb, ino);
+	if (!IS_ERR(inode)) {
+		if (is_bad_inode(inode) || inode->i_generation != generation) {
+			iput(inode);
+			inode = ERR_PTR(-ESTALE);
+		}
+	}
+
+	return inode;
+}
+
+static struct dentry *ntfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
+		int fh_len, int fh_type)
+{
+	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
+				    ntfs_nfs_get_inode);
+}
+
+static struct dentry *ntfs_fh_to_parent(struct super_block *sb, struct fid *fid,
+		int fh_len, int fh_type)
+{
+	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
+				    ntfs_nfs_get_inode);
+}
+
+/**
+ * Export operations allowing NFS exporting of mounted NTFS partitions.
+ *
+ * We use the default ->encode_fh() for now.  Note that they
+ * use 32 bits to store the inode number which is an unsigned long so on 64-bit
+ * architectures is usually 64 bits so it would all fail horribly on huge
+ * volumes.  I guess we need to define our own encode and decode fh functions
+ * that store 64-bit inode numbers at some point but for now we will ignore the
+ * problem...
+ *
+ * We also use the default ->get_name() helper (used by ->decode_fh() via
+ * fs/exportfs/expfs.c::find_exported_dentry()) as that is completely fs
+ * independent.
+ *
+ * The default ->get_parent() just returns -EACCES so we have to provide our
+ * own and the default ->get_dentry() is incompatible with NTFS due to not
+ * allowing the inode number 0 which is used in NTFS for the system file $MFT
+ * and due to using iget() whereas NTFS needs ntfs_iget().
+ */
+const struct export_operations ntfs_export_ops = {
+	.get_parent	= ntfs_get_parent,	/* Find the parent of a given
+						   directory. */
+	.fh_to_dentry	= ntfs_fh_to_dentry,
+	.fh_to_parent	= ntfs_fh_to_parent,
+};
diff --git a/fs/ntfs/ntfs.h b/fs/ntfs/ntfs.h
new file mode 100644
index 000000000..e81376ea9
--- /dev/null
+++ b/fs/ntfs/ntfs.h
@@ -0,0 +1,150 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * ntfs.h - Defines for NTFS Linux kernel driver.
+ *
+ * Copyright (c) 2001-2014 Anton Altaparmakov and Tuxera Inc.
+ * Copyright (C) 2002 Richard Russon
+ */
+
+#ifndef _LINUX_NTFS_H
+#define _LINUX_NTFS_H
+
+#include <linux/stddef.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/compiler.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+#include <linux/smp.h>
+#include <linux/pagemap.h>
+
+#include "types.h"
+#include "volume.h"
+#include "layout.h"
+
+typedef enum {
+	NTFS_BLOCK_SIZE		= 512,
+	NTFS_BLOCK_SIZE_BITS	= 9,
+	NTFS_SB_MAGIC		= 0x5346544e,	/* 'NTFS' */
+	NTFS_MAX_NAME_LEN	= 255,
+	NTFS_MAX_ATTR_NAME_LEN	= 255,
+	NTFS_MAX_CLUSTER_SIZE	= 64 * 1024,	/* 64kiB */
+	NTFS_MAX_PAGES_PER_CLUSTER = NTFS_MAX_CLUSTER_SIZE / PAGE_SIZE,
+} NTFS_CONSTANTS;
+
+/* Global variables. */
+
+/* Slab caches (from super.c). */
+extern struct kmem_cache *ntfs_name_cache;
+extern struct kmem_cache *ntfs_inode_cache;
+extern struct kmem_cache *ntfs_big_inode_cache;
+extern struct kmem_cache *ntfs_attr_ctx_cache;
+extern struct kmem_cache *ntfs_index_ctx_cache;
+
+/* The various operations structs defined throughout the driver files. */
+extern const struct address_space_operations ntfs_normal_aops;
+extern const struct address_space_operations ntfs_compressed_aops;
+extern const struct address_space_operations ntfs_mst_aops;
+
+extern const struct  file_operations ntfs_file_ops;
+extern const struct inode_operations ntfs_file_inode_ops;
+
+extern const struct  file_operations ntfs_dir_ops;
+extern const struct inode_operations ntfs_dir_inode_ops;
+
+extern const struct  file_operations ntfs_empty_file_ops;
+extern const struct inode_operations ntfs_empty_inode_ops;
+
+extern const struct export_operations ntfs_export_ops;
+
+/**
+ * NTFS_SB - return the ntfs volume given a vfs super block
+ * @sb:		VFS super block
+ *
+ * NTFS_SB() returns the ntfs volume associated with the VFS super block @sb.
+ */
+static inline ntfs_volume *NTFS_SB(struct super_block *sb)
+{
+	return sb->s_fs_info;
+}
+
+/* Declarations of functions and global variables. */
+
+/* From fs/ntfs/compress.c */
+extern int ntfs_read_compressed_block(struct page *page);
+extern int allocate_compression_buffers(void);
+extern void free_compression_buffers(void);
+
+/* From fs/ntfs/super.c */
+#define default_upcase_len 0x10000
+extern struct mutex ntfs_lock;
+
+typedef struct {
+	int val;
+	char *str;
+} option_t;
+extern const option_t on_errors_arr[];
+
+/* From fs/ntfs/mst.c */
+extern int post_read_mst_fixup(NTFS_RECORD *b, const u32 size);
+extern int pre_write_mst_fixup(NTFS_RECORD *b, const u32 size);
+extern void post_write_mst_fixup(NTFS_RECORD *b);
+
+/* From fs/ntfs/unistr.c */
+extern bool ntfs_are_names_equal(const ntfschar *s1, size_t s1_len,
+		const ntfschar *s2, size_t s2_len,
+		const IGNORE_CASE_BOOL ic,
+		const ntfschar *upcase, const u32 upcase_size);
+extern int ntfs_collate_names(const ntfschar *name1, const u32 name1_len,
+		const ntfschar *name2, const u32 name2_len,
+		const int err_val, const IGNORE_CASE_BOOL ic,
+		const ntfschar *upcase, const u32 upcase_len);
+extern int ntfs_ucsncmp(const ntfschar *s1, const ntfschar *s2, size_t n);
+extern int ntfs_ucsncasecmp(const ntfschar *s1, const ntfschar *s2, size_t n,
+		const ntfschar *upcase, const u32 upcase_size);
+extern void ntfs_upcase_name(ntfschar *name, u32 name_len,
+		const ntfschar *upcase, const u32 upcase_len);
+extern void ntfs_file_upcase_value(FILE_NAME_ATTR *file_name_attr,
+		const ntfschar *upcase, const u32 upcase_len);
+extern int ntfs_file_compare_values(FILE_NAME_ATTR *file_name_attr1,
+		FILE_NAME_ATTR *file_name_attr2,
+		const int err_val, const IGNORE_CASE_BOOL ic,
+		const ntfschar *upcase, const u32 upcase_len);
+extern int ntfs_nlstoucs(const ntfs_volume *vol, const char *ins,
+		const int ins_len, ntfschar **outs);
+extern int ntfs_ucstonls(const ntfs_volume *vol, const ntfschar *ins,
+		const int ins_len, unsigned char **outs, int outs_len);
+
+/* From fs/ntfs/upcase.c */
+extern ntfschar *generate_default_upcase(void);
+
+static inline int ntfs_ffs(int x)
+{
+	int r = 1;
+
+	if (!x)
+		return 0;
+	if (!(x & 0xffff)) {
+		x >>= 16;
+		r += 16;
+	}
+	if (!(x & 0xff)) {
+		x >>= 8;
+		r += 8;
+	}
+	if (!(x & 0xf)) {
+		x >>= 4;
+		r += 4;
+	}
+	if (!(x & 3)) {
+		x >>= 2;
+		r += 2;
+	}
+	if (!(x & 1)) {
+		x >>= 1;
+		r += 1;
+	}
+	return r;
+}
+
+#endif /* _LINUX_NTFS_H */
diff --git a/fs/ntfs/quota.c b/fs/ntfs/quota.c
new file mode 100644
index 000000000..916048022
--- /dev/null
+++ b/fs/ntfs/quota.c
@@ -0,0 +1,103 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * quota.c - NTFS kernel quota ($Quota) handling.  Part of the Linux-NTFS
+ *	     project.
+ *
+ * Copyright (c) 2004 Anton Altaparmakov
+ */
+
+#ifdef NTFS_RW
+
+#include "index.h"
+#include "quota.h"
+#include "debug.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_mark_quotas_out_of_date - mark the quotas out of date on an ntfs volume
+ * @vol:	ntfs volume on which to mark the quotas out of date
+ *
+ * Mark the quotas out of date on the ntfs volume @vol and return 'true' on
+ * success and 'false' on error.
+ */
+bool ntfs_mark_quotas_out_of_date(ntfs_volume *vol)
+{
+	ntfs_index_context *ictx;
+	QUOTA_CONTROL_ENTRY *qce;
+	const le32 qid = QUOTA_DEFAULTS_ID;
+	int err;
+
+	ntfs_debug("Entering.");
+	if (NVolQuotaOutOfDate(vol))
+		goto done;
+	if (!vol->quota_ino || !vol->quota_q_ino) {
+		ntfs_error(vol->sb, "Quota inodes are not open.");
+		return false;
+	}
+	inode_lock(vol->quota_q_ino);
+	ictx = ntfs_index_ctx_get(NTFS_I(vol->quota_q_ino));
+	if (!ictx) {
+		ntfs_error(vol->sb, "Failed to get index context.");
+		goto err_out;
+	}
+	err = ntfs_index_lookup(&qid, sizeof(qid), ictx);
+	if (err) {
+		if (err == -ENOENT)
+			ntfs_error(vol->sb, "Quota defaults entry is not "
+					"present.");
+		else
+			ntfs_error(vol->sb, "Lookup of quota defaults entry "
+					"failed.");
+		goto err_out;
+	}
+	if (ictx->data_len < offsetof(QUOTA_CONTROL_ENTRY, sid)) {
+		ntfs_error(vol->sb, "Quota defaults entry size is invalid.  "
+				"Run chkdsk.");
+		goto err_out;
+	}
+	qce = (QUOTA_CONTROL_ENTRY*)ictx->data;
+	if (le32_to_cpu(qce->version) != QUOTA_VERSION) {
+		ntfs_error(vol->sb, "Quota defaults entry version 0x%x is not "
+				"supported.", le32_to_cpu(qce->version));
+		goto err_out;
+	}
+	ntfs_debug("Quota defaults flags = 0x%x.", le32_to_cpu(qce->flags));
+	/* If quotas are already marked out of date, no need to do anything. */
+	if (qce->flags & QUOTA_FLAG_OUT_OF_DATE)
+		goto set_done;
+	/*
+	 * If quota tracking is neither requested, nor enabled and there are no
+	 * pending deletes, no need to mark the quotas out of date.
+	 */
+	if (!(qce->flags & (QUOTA_FLAG_TRACKING_ENABLED |
+			QUOTA_FLAG_TRACKING_REQUESTED |
+			QUOTA_FLAG_PENDING_DELETES)))
+		goto set_done;
+	/*
+	 * Set the QUOTA_FLAG_OUT_OF_DATE bit thus marking quotas out of date.
+	 * This is verified on WinXP to be sufficient to cause windows to
+	 * rescan the volume on boot and update all quota entries.
+	 */
+	qce->flags |= QUOTA_FLAG_OUT_OF_DATE;
+	/* Ensure the modified flags are written to disk. */
+	ntfs_index_entry_flush_dcache_page(ictx);
+	ntfs_index_entry_mark_dirty(ictx);
+set_done:
+	ntfs_index_ctx_put(ictx);
+	inode_unlock(vol->quota_q_ino);
+	/*
+	 * We set the flag so we do not try to mark the quotas out of date
+	 * again on remount.
+	 */
+	NVolSetQuotaOutOfDate(vol);
+done:
+	ntfs_debug("Done.");
+	return true;
+err_out:
+	if (ictx)
+		ntfs_index_ctx_put(ictx);
+	inode_unlock(vol->quota_q_ino);
+	return false;
+}
+
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/quota.h b/fs/ntfs/quota.h
new file mode 100644
index 000000000..fe3132a3d
--- /dev/null
+++ b/fs/ntfs/quota.h
@@ -0,0 +1,21 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * quota.h - Defines for NTFS kernel quota ($Quota) handling.  Part of the
+ *	     Linux-NTFS project.
+ *
+ * Copyright (c) 2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_QUOTA_H
+#define _LINUX_NTFS_QUOTA_H
+
+#ifdef NTFS_RW
+
+#include "types.h"
+#include "volume.h"
+
+extern bool ntfs_mark_quotas_out_of_date(ntfs_volume *vol);
+
+#endif /* NTFS_RW */
+
+#endif /* _LINUX_NTFS_QUOTA_H */
diff --git a/fs/ntfs/runlist.c b/fs/ntfs/runlist.c
new file mode 100644
index 000000000..97932fb51
--- /dev/null
+++ b/fs/ntfs/runlist.c
@@ -0,0 +1,1893 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/**
+ * runlist.c - NTFS runlist handling code.  Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2007 Anton Altaparmakov
+ * Copyright (c) 2002-2005 Richard Russon
+ */
+
+#include "debug.h"
+#include "dir.h"
+#include "endian.h"
+#include "malloc.h"
+#include "ntfs.h"
+
+/**
+ * ntfs_rl_mm - runlist memmove
+ *
+ * It is up to the caller to serialize access to the runlist @base.
+ */
+static inline void ntfs_rl_mm(runlist_element *base, int dst, int src,
+		int size)
+{
+	if (likely((dst != src) && (size > 0)))
+		memmove(base + dst, base + src, size * sizeof(*base));
+}
+
+/**
+ * ntfs_rl_mc - runlist memory copy
+ *
+ * It is up to the caller to serialize access to the runlists @dstbase and
+ * @srcbase.
+ */
+static inline void ntfs_rl_mc(runlist_element *dstbase, int dst,
+		runlist_element *srcbase, int src, int size)
+{
+	if (likely(size > 0))
+		memcpy(dstbase + dst, srcbase + src, size * sizeof(*dstbase));
+}
+
+/**
+ * ntfs_rl_realloc - Reallocate memory for runlists
+ * @rl:		original runlist
+ * @old_size:	number of runlist elements in the original runlist @rl
+ * @new_size:	number of runlist elements we need space for
+ *
+ * As the runlists grow, more memory will be required.  To prevent the
+ * kernel having to allocate and reallocate large numbers of small bits of
+ * memory, this function returns an entire page of memory.
+ *
+ * It is up to the caller to serialize access to the runlist @rl.
+ *
+ * N.B.  If the new allocation doesn't require a different number of pages in
+ *       memory, the function will return the original pointer.
+ *
+ * On success, return a pointer to the newly allocated, or recycled, memory.
+ * On error, return -errno. The following error codes are defined:
+ *	-ENOMEM	- Not enough memory to allocate runlist array.
+ *	-EINVAL	- Invalid parameters were passed in.
+ */
+static inline runlist_element *ntfs_rl_realloc(runlist_element *rl,
+		int old_size, int new_size)
+{
+	runlist_element *new_rl;
+
+	old_size = PAGE_ALIGN(old_size * sizeof(*rl));
+	new_size = PAGE_ALIGN(new_size * sizeof(*rl));
+	if (old_size == new_size)
+		return rl;
+
+	new_rl = ntfs_malloc_nofs(new_size);
+	if (unlikely(!new_rl))
+		return ERR_PTR(-ENOMEM);
+
+	if (likely(rl != NULL)) {
+		if (unlikely(old_size > new_size))
+			old_size = new_size;
+		memcpy(new_rl, rl, old_size);
+		ntfs_free(rl);
+	}
+	return new_rl;
+}
+
+/**
+ * ntfs_rl_realloc_nofail - Reallocate memory for runlists
+ * @rl:		original runlist
+ * @old_size:	number of runlist elements in the original runlist @rl
+ * @new_size:	number of runlist elements we need space for
+ *
+ * As the runlists grow, more memory will be required.  To prevent the
+ * kernel having to allocate and reallocate large numbers of small bits of
+ * memory, this function returns an entire page of memory.
+ *
+ * This function guarantees that the allocation will succeed.  It will sleep
+ * for as long as it takes to complete the allocation.
+ *
+ * It is up to the caller to serialize access to the runlist @rl.
+ *
+ * N.B.  If the new allocation doesn't require a different number of pages in
+ *       memory, the function will return the original pointer.
+ *
+ * On success, return a pointer to the newly allocated, or recycled, memory.
+ * On error, return -errno. The following error codes are defined:
+ *	-ENOMEM	- Not enough memory to allocate runlist array.
+ *	-EINVAL	- Invalid parameters were passed in.
+ */
+static inline runlist_element *ntfs_rl_realloc_nofail(runlist_element *rl,
+		int old_size, int new_size)
+{
+	runlist_element *new_rl;
+
+	old_size = PAGE_ALIGN(old_size * sizeof(*rl));
+	new_size = PAGE_ALIGN(new_size * sizeof(*rl));
+	if (old_size == new_size)
+		return rl;
+
+	new_rl = ntfs_malloc_nofs_nofail(new_size);
+	BUG_ON(!new_rl);
+
+	if (likely(rl != NULL)) {
+		if (unlikely(old_size > new_size))
+			old_size = new_size;
+		memcpy(new_rl, rl, old_size);
+		ntfs_free(rl);
+	}
+	return new_rl;
+}
+
+/**
+ * ntfs_are_rl_mergeable - test if two runlists can be joined together
+ * @dst:	original runlist
+ * @src:	new runlist to test for mergeability with @dst
+ *
+ * Test if two runlists can be joined together. For this, their VCNs and LCNs
+ * must be adjacent.
+ *
+ * It is up to the caller to serialize access to the runlists @dst and @src.
+ *
+ * Return: true   Success, the runlists can be merged.
+ *	   false  Failure, the runlists cannot be merged.
+ */
+static inline bool ntfs_are_rl_mergeable(runlist_element *dst,
+		runlist_element *src)
+{
+	BUG_ON(!dst);
+	BUG_ON(!src);
+
+	/* We can merge unmapped regions even if they are misaligned. */
+	if ((dst->lcn == LCN_RL_NOT_MAPPED) && (src->lcn == LCN_RL_NOT_MAPPED))
+		return true;
+	/* If the runs are misaligned, we cannot merge them. */
+	if ((dst->vcn + dst->length) != src->vcn)
+		return false;
+	/* If both runs are non-sparse and contiguous, we can merge them. */
+	if ((dst->lcn >= 0) && (src->lcn >= 0) &&
+			((dst->lcn + dst->length) == src->lcn))
+		return true;
+	/* If we are merging two holes, we can merge them. */
+	if ((dst->lcn == LCN_HOLE) && (src->lcn == LCN_HOLE))
+		return true;
+	/* Cannot merge. */
+	return false;
+}
+
+/**
+ * __ntfs_rl_merge - merge two runlists without testing if they can be merged
+ * @dst:	original, destination runlist
+ * @src:	new runlist to merge with @dst
+ *
+ * Merge the two runlists, writing into the destination runlist @dst. The
+ * caller must make sure the runlists can be merged or this will corrupt the
+ * destination runlist.
+ *
+ * It is up to the caller to serialize access to the runlists @dst and @src.
+ */
+static inline void __ntfs_rl_merge(runlist_element *dst, runlist_element *src)
+{
+	dst->length += src->length;
+}
+
+/**
+ * ntfs_rl_append - append a runlist after a given element
+ * @dst:	original runlist to be worked on
+ * @dsize:	number of elements in @dst (including end marker)
+ * @src:	runlist to be inserted into @dst
+ * @ssize:	number of elements in @src (excluding end marker)
+ * @loc:	append the new runlist @src after this element in @dst
+ *
+ * Append the runlist @src after element @loc in @dst.  Merge the right end of
+ * the new runlist, if necessary. Adjust the size of the hole before the
+ * appended runlist.
+ *
+ * It is up to the caller to serialize access to the runlists @dst and @src.
+ *
+ * On success, return a pointer to the new, combined, runlist. Note, both
+ * runlists @dst and @src are deallocated before returning so you cannot use
+ * the pointers for anything any more. (Strictly speaking the returned runlist
+ * may be the same as @dst but this is irrelevant.)
+ *
+ * On error, return -errno. Both runlists are left unmodified. The following
+ * error codes are defined:
+ *	-ENOMEM	- Not enough memory to allocate runlist array.
+ *	-EINVAL	- Invalid parameters were passed in.
+ */
+static inline runlist_element *ntfs_rl_append(runlist_element *dst,
+		int dsize, runlist_element *src, int ssize, int loc)
+{
+	bool right = false;	/* Right end of @src needs merging. */
+	int marker;		/* End of the inserted runs. */
+
+	BUG_ON(!dst);
+	BUG_ON(!src);
+
+	/* First, check if the right hand end needs merging. */
+	if ((loc + 1) < dsize)
+		right = ntfs_are_rl_mergeable(src + ssize - 1, dst + loc + 1);
+
+	/* Space required: @dst size + @src size, less one if we merged. */
+	dst = ntfs_rl_realloc(dst, dsize, dsize + ssize - right);
+	if (IS_ERR(dst))
+		return dst;
+	/*
+	 * We are guaranteed to succeed from here so can start modifying the
+	 * original runlists.
+	 */
+
+	/* First, merge the right hand end, if necessary. */
+	if (right)
+		__ntfs_rl_merge(src + ssize - 1, dst + loc + 1);
+
+	/* First run after the @src runs that have been inserted. */
+	marker = loc + ssize + 1;
+
+	/* Move the tail of @dst out of the way, then copy in @src. */
+	ntfs_rl_mm(dst, marker, loc + 1 + right, dsize - (loc + 1 + right));
+	ntfs_rl_mc(dst, loc + 1, src, 0, ssize);
+
+	/* Adjust the size of the preceding hole. */
+	dst[loc].length = dst[loc + 1].vcn - dst[loc].vcn;
+
+	/* We may have changed the length of the file, so fix the end marker */
+	if (dst[marker].lcn == LCN_ENOENT)
+		dst[marker].vcn = dst[marker - 1].vcn + dst[marker - 1].length;
+
+	return dst;
+}
+
+/**
+ * ntfs_rl_insert - insert a runlist into another
+ * @dst:	original runlist to be worked on
+ * @dsize:	number of elements in @dst (including end marker)
+ * @src:	new runlist to be inserted
+ * @ssize:	number of elements in @src (excluding end marker)
+ * @loc:	insert the new runlist @src before this element in @dst
+ *
+ * Insert the runlist @src before element @loc in the runlist @dst. Merge the
+ * left end of the new runlist, if necessary. Adjust the size of the hole
+ * after the inserted runlist.
+ *
+ * It is up to the caller to serialize access to the runlists @dst and @src.
+ *
+ * On success, return a pointer to the new, combined, runlist. Note, both
+ * runlists @dst and @src are deallocated before returning so you cannot use
+ * the pointers for anything any more. (Strictly speaking the returned runlist
+ * may be the same as @dst but this is irrelevant.)
+ *
+ * On error, return -errno. Both runlists are left unmodified. The following
+ * error codes are defined:
+ *	-ENOMEM	- Not enough memory to allocate runlist array.
+ *	-EINVAL	- Invalid parameters were passed in.
+ */
+static inline runlist_element *ntfs_rl_insert(runlist_element *dst,
+		int dsize, runlist_element *src, int ssize, int loc)
+{
+	bool left = false;	/* Left end of @src needs merging. */
+	bool disc = false;	/* Discontinuity between @dst and @src. */
+	int marker;		/* End of the inserted runs. */
+
+	BUG_ON(!dst);
+	BUG_ON(!src);
+
+	/*
+	 * disc => Discontinuity between the end of @dst and the start of @src.
+	 *	   This means we might need to insert a "not mapped" run.
+	 */
+	if (loc == 0)
+		disc = (src[0].vcn > 0);
+	else {
+		s64 merged_length;
+
+		left = ntfs_are_rl_mergeable(dst + loc - 1, src);
+
+		merged_length = dst[loc - 1].length;
+		if (left)
+			merged_length += src->length;
+
+		disc = (src[0].vcn > dst[loc - 1].vcn + merged_length);
+	}
+	/*
+	 * Space required: @dst size + @src size, less one if we merged, plus
+	 * one if there was a discontinuity.
+	 */
+	dst = ntfs_rl_realloc(dst, dsize, dsize + ssize - left + disc);
+	if (IS_ERR(dst))
+		return dst;
+	/*
+	 * We are guaranteed to succeed from here so can start modifying the
+	 * original runlist.
+	 */
+	if (left)
+		__ntfs_rl_merge(dst + loc - 1, src);
+	/*
+	 * First run after the @src runs that have been inserted.
+	 * Nominally,  @marker equals @loc + @ssize, i.e. location + number of
+	 * runs in @src.  However, if @left, then the first run in @src has
+	 * been merged with one in @dst.  And if @disc, then @dst and @src do
+	 * not meet and we need an extra run to fill the gap.
+	 */
+	marker = loc + ssize - left + disc;
+
+	/* Move the tail of @dst out of the way, then copy in @src. */
+	ntfs_rl_mm(dst, marker, loc, dsize - loc);
+	ntfs_rl_mc(dst, loc + disc, src, left, ssize - left);
+
+	/* Adjust the VCN of the first run after the insertion... */
+	dst[marker].vcn = dst[marker - 1].vcn + dst[marker - 1].length;
+	/* ... and the length. */
+	if (dst[marker].lcn == LCN_HOLE || dst[marker].lcn == LCN_RL_NOT_MAPPED)
+		dst[marker].length = dst[marker + 1].vcn - dst[marker].vcn;
+
+	/* Writing beyond the end of the file and there is a discontinuity. */
+	if (disc) {
+		if (loc > 0) {
+			dst[loc].vcn = dst[loc - 1].vcn + dst[loc - 1].length;
+			dst[loc].length = dst[loc + 1].vcn - dst[loc].vcn;
+		} else {
+			dst[loc].vcn = 0;
+			dst[loc].length = dst[loc + 1].vcn;
+		}
+		dst[loc].lcn = LCN_RL_NOT_MAPPED;
+	}
+	return dst;
+}
+
+/**
+ * ntfs_rl_replace - overwrite a runlist element with another runlist
+ * @dst:	original runlist to be worked on
+ * @dsize:	number of elements in @dst (including end marker)
+ * @src:	new runlist to be inserted
+ * @ssize:	number of elements in @src (excluding end marker)
+ * @loc:	index in runlist @dst to overwrite with @src
+ *
+ * Replace the runlist element @dst at @loc with @src. Merge the left and
+ * right ends of the inserted runlist, if necessary.
+ *
+ * It is up to the caller to serialize access to the runlists @dst and @src.
+ *
+ * On success, return a pointer to the new, combined, runlist. Note, both
+ * runlists @dst and @src are deallocated before returning so you cannot use
+ * the pointers for anything any more. (Strictly speaking the returned runlist
+ * may be the same as @dst but this is irrelevant.)
+ *
+ * On error, return -errno. Both runlists are left unmodified. The following
+ * error codes are defined:
+ *	-ENOMEM	- Not enough memory to allocate runlist array.
+ *	-EINVAL	- Invalid parameters were passed in.
+ */
+static inline runlist_element *ntfs_rl_replace(runlist_element *dst,
+		int dsize, runlist_element *src, int ssize, int loc)
+{
+	signed delta;
+	bool left = false;	/* Left end of @src needs merging. */
+	bool right = false;	/* Right end of @src needs merging. */
+	int tail;		/* Start of tail of @dst. */
+	int marker;		/* End of the inserted runs. */
+
+	BUG_ON(!dst);
+	BUG_ON(!src);
+
+	/* First, see if the left and right ends need merging. */
+	if ((loc + 1) < dsize)
+		right = ntfs_are_rl_mergeable(src + ssize - 1, dst + loc + 1);
+	if (loc > 0)
+		left = ntfs_are_rl_mergeable(dst + loc - 1, src);
+	/*
+	 * Allocate some space.  We will need less if the left, right, or both
+	 * ends get merged.  The -1 accounts for the run being replaced.
+	 */
+	delta = ssize - 1 - left - right;
+	if (delta > 0) {
+		dst = ntfs_rl_realloc(dst, dsize, dsize + delta);
+		if (IS_ERR(dst))
+			return dst;
+	}
+	/*
+	 * We are guaranteed to succeed from here so can start modifying the
+	 * original runlists.
+	 */
+
+	/* First, merge the left and right ends, if necessary. */
+	if (right)
+		__ntfs_rl_merge(src + ssize - 1, dst + loc + 1);
+	if (left)
+		__ntfs_rl_merge(dst + loc - 1, src);
+	/*
+	 * Offset of the tail of @dst.  This needs to be moved out of the way
+	 * to make space for the runs to be copied from @src, i.e. the first
+	 * run of the tail of @dst.
+	 * Nominally, @tail equals @loc + 1, i.e. location, skipping the
+	 * replaced run.  However, if @right, then one of @dst's runs is
+	 * already merged into @src.
+	 */
+	tail = loc + right + 1;
+	/*
+	 * First run after the @src runs that have been inserted, i.e. where
+	 * the tail of @dst needs to be moved to.
+	 * Nominally, @marker equals @loc + @ssize, i.e. location + number of
+	 * runs in @src.  However, if @left, then the first run in @src has
+	 * been merged with one in @dst.
+	 */
+	marker = loc + ssize - left;
+
+	/* Move the tail of @dst out of the way, then copy in @src. */
+	ntfs_rl_mm(dst, marker, tail, dsize - tail);
+	ntfs_rl_mc(dst, loc, src, left, ssize - left);
+
+	/* We may have changed the length of the file, so fix the end marker. */
+	if (dsize - tail > 0 && dst[marker].lcn == LCN_ENOENT)
+		dst[marker].vcn = dst[marker - 1].vcn + dst[marker - 1].length;
+	return dst;
+}
+
+/**
+ * ntfs_rl_split - insert a runlist into the centre of a hole
+ * @dst:	original runlist to be worked on
+ * @dsize:	number of elements in @dst (including end marker)
+ * @src:	new runlist to be inserted
+ * @ssize:	number of elements in @src (excluding end marker)
+ * @loc:	index in runlist @dst at which to split and insert @src
+ *
+ * Split the runlist @dst at @loc into two and insert @new in between the two
+ * fragments. No merging of runlists is necessary. Adjust the size of the
+ * holes either side.
+ *
+ * It is up to the caller to serialize access to the runlists @dst and @src.
+ *
+ * On success, return a pointer to the new, combined, runlist. Note, both
+ * runlists @dst and @src are deallocated before returning so you cannot use
+ * the pointers for anything any more. (Strictly speaking the returned runlist
+ * may be the same as @dst but this is irrelevant.)
+ *
+ * On error, return -errno. Both runlists are left unmodified. The following
+ * error codes are defined:
+ *	-ENOMEM	- Not enough memory to allocate runlist array.
+ *	-EINVAL	- Invalid parameters were passed in.
+ */
+static inline runlist_element *ntfs_rl_split(runlist_element *dst, int dsize,
+		runlist_element *src, int ssize, int loc)
+{
+	BUG_ON(!dst);
+	BUG_ON(!src);
+
+	/* Space required: @dst size + @src size + one new hole. */
+	dst = ntfs_rl_realloc(dst, dsize, dsize + ssize + 1);
+	if (IS_ERR(dst))
+		return dst;
+	/*
+	 * We are guaranteed to succeed from here so can start modifying the
+	 * original runlists.
+	 */
+
+	/* Move the tail of @dst out of the way, then copy in @src. */
+	ntfs_rl_mm(dst, loc + 1 + ssize, loc, dsize - loc);
+	ntfs_rl_mc(dst, loc + 1, src, 0, ssize);
+
+	/* Adjust the size of the holes either size of @src. */
+	dst[loc].length		= dst[loc+1].vcn       - dst[loc].vcn;
+	dst[loc+ssize+1].vcn    = dst[loc+ssize].vcn   + dst[loc+ssize].length;
+	dst[loc+ssize+1].length = dst[loc+ssize+2].vcn - dst[loc+ssize+1].vcn;
+
+	return dst;
+}
+
+/**
+ * ntfs_runlists_merge - merge two runlists into one
+ * @drl:	original runlist to be worked on
+ * @srl:	new runlist to be merged into @drl
+ *
+ * First we sanity check the two runlists @srl and @drl to make sure that they
+ * are sensible and can be merged. The runlist @srl must be either after the
+ * runlist @drl or completely within a hole (or unmapped region) in @drl.
+ *
+ * It is up to the caller to serialize access to the runlists @drl and @srl.
+ *
+ * Merging of runlists is necessary in two cases:
+ *   1. When attribute lists are used and a further extent is being mapped.
+ *   2. When new clusters are allocated to fill a hole or extend a file.
+ *
+ * There are four possible ways @srl can be merged. It can:
+ *	- be inserted at the beginning of a hole,
+ *	- split the hole in two and be inserted between the two fragments,
+ *	- be appended at the end of a hole, or it can
+ *	- replace the whole hole.
+ * It can also be appended to the end of the runlist, which is just a variant
+ * of the insert case.
+ *
+ * On success, return a pointer to the new, combined, runlist. Note, both
+ * runlists @drl and @srl are deallocated before returning so you cannot use
+ * the pointers for anything any more. (Strictly speaking the returned runlist
+ * may be the same as @dst but this is irrelevant.)
+ *
+ * On error, return -errno. Both runlists are left unmodified. The following
+ * error codes are defined:
+ *	-ENOMEM	- Not enough memory to allocate runlist array.
+ *	-EINVAL	- Invalid parameters were passed in.
+ *	-ERANGE	- The runlists overlap and cannot be merged.
+ */
+runlist_element *ntfs_runlists_merge(runlist_element *drl,
+		runlist_element *srl)
+{
+	int di, si;		/* Current index into @[ds]rl. */
+	int sstart;		/* First index with lcn > LCN_RL_NOT_MAPPED. */
+	int dins;		/* Index into @drl at which to insert @srl. */
+	int dend, send;		/* Last index into @[ds]rl. */
+	int dfinal, sfinal;	/* The last index into @[ds]rl with
+				   lcn >= LCN_HOLE. */
+	int marker = 0;
+	VCN marker_vcn = 0;
+
+#ifdef DEBUG
+	ntfs_debug("dst:");
+	ntfs_debug_dump_runlist(drl);
+	ntfs_debug("src:");
+	ntfs_debug_dump_runlist(srl);
+#endif
+
+	/* Check for silly calling... */
+	if (unlikely(!srl))
+		return drl;
+	if (IS_ERR(srl) || IS_ERR(drl))
+		return ERR_PTR(-EINVAL);
+
+	/* Check for the case where the first mapping is being done now. */
+	if (unlikely(!drl)) {
+		drl = srl;
+		/* Complete the source runlist if necessary. */
+		if (unlikely(drl[0].vcn)) {
+			/* Scan to the end of the source runlist. */
+			for (dend = 0; likely(drl[dend].length); dend++)
+				;
+			dend++;
+			drl = ntfs_rl_realloc(drl, dend, dend + 1);
+			if (IS_ERR(drl))
+				return drl;
+			/* Insert start element at the front of the runlist. */
+			ntfs_rl_mm(drl, 1, 0, dend);
+			drl[0].vcn = 0;
+			drl[0].lcn = LCN_RL_NOT_MAPPED;
+			drl[0].length = drl[1].vcn;
+		}
+		goto finished;
+	}
+
+	si = di = 0;
+
+	/* Skip any unmapped start element(s) in the source runlist. */
+	while (srl[si].length && srl[si].lcn < LCN_HOLE)
+		si++;
+
+	/* Can't have an entirely unmapped source runlist. */
+	BUG_ON(!srl[si].length);
+
+	/* Record the starting points. */
+	sstart = si;
+
+	/*
+	 * Skip forward in @drl until we reach the position where @srl needs to
+	 * be inserted. If we reach the end of @drl, @srl just needs to be
+	 * appended to @drl.
+	 */
+	for (; drl[di].length; di++) {
+		if (drl[di].vcn + drl[di].length > srl[sstart].vcn)
+			break;
+	}
+	dins = di;
+
+	/* Sanity check for illegal overlaps. */
+	if ((drl[di].vcn == srl[si].vcn) && (drl[di].lcn >= 0) &&
+			(srl[si].lcn >= 0)) {
+		ntfs_error(NULL, "Run lists overlap. Cannot merge!");
+		return ERR_PTR(-ERANGE);
+	}
+
+	/* Scan to the end of both runlists in order to know their sizes. */
+	for (send = si; srl[send].length; send++)
+		;
+	for (dend = di; drl[dend].length; dend++)
+		;
+
+	if (srl[send].lcn == LCN_ENOENT)
+		marker_vcn = srl[marker = send].vcn;
+
+	/* Scan to the last element with lcn >= LCN_HOLE. */
+	for (sfinal = send; sfinal >= 0 && srl[sfinal].lcn < LCN_HOLE; sfinal--)
+		;
+	for (dfinal = dend; dfinal >= 0 && drl[dfinal].lcn < LCN_HOLE; dfinal--)
+		;
+
+	{
+	bool start;
+	bool finish;
+	int ds = dend + 1;		/* Number of elements in drl & srl */
+	int ss = sfinal - sstart + 1;
+
+	start  = ((drl[dins].lcn <  LCN_RL_NOT_MAPPED) ||    /* End of file   */
+		  (drl[dins].vcn == srl[sstart].vcn));	     /* Start of hole */
+	finish = ((drl[dins].lcn >= LCN_RL_NOT_MAPPED) &&    /* End of file   */
+		 ((drl[dins].vcn + drl[dins].length) <=      /* End of hole   */
+		  (srl[send - 1].vcn + srl[send - 1].length)));
+
+	/* Or we will lose an end marker. */
+	if (finish && !drl[dins].length)
+		ss++;
+	if (marker && (drl[dins].vcn + drl[dins].length > srl[send - 1].vcn))
+		finish = false;
+#if 0
+	ntfs_debug("dfinal = %i, dend = %i", dfinal, dend);
+	ntfs_debug("sstart = %i, sfinal = %i, send = %i", sstart, sfinal, send);
+	ntfs_debug("start = %i, finish = %i", start, finish);
+	ntfs_debug("ds = %i, ss = %i, dins = %i", ds, ss, dins);
+#endif
+	if (start) {
+		if (finish)
+			drl = ntfs_rl_replace(drl, ds, srl + sstart, ss, dins);
+		else
+			drl = ntfs_rl_insert(drl, ds, srl + sstart, ss, dins);
+	} else {
+		if (finish)
+			drl = ntfs_rl_append(drl, ds, srl + sstart, ss, dins);
+		else
+			drl = ntfs_rl_split(drl, ds, srl + sstart, ss, dins);
+	}
+	if (IS_ERR(drl)) {
+		ntfs_error(NULL, "Merge failed.");
+		return drl;
+	}
+	ntfs_free(srl);
+	if (marker) {
+		ntfs_debug("Triggering marker code.");
+		for (ds = dend; drl[ds].length; ds++)
+			;
+		/* We only need to care if @srl ended after @drl. */
+		if (drl[ds].vcn <= marker_vcn) {
+			int slots = 0;
+
+			if (drl[ds].vcn == marker_vcn) {
+				ntfs_debug("Old marker = 0x%llx, replacing "
+						"with LCN_ENOENT.",
+						(unsigned long long)
+						drl[ds].lcn);
+				drl[ds].lcn = LCN_ENOENT;
+				goto finished;
+			}
+			/*
+			 * We need to create an unmapped runlist element in
+			 * @drl or extend an existing one before adding the
+			 * ENOENT terminator.
+			 */
+			if (drl[ds].lcn == LCN_ENOENT) {
+				ds--;
+				slots = 1;
+			}
+			if (drl[ds].lcn != LCN_RL_NOT_MAPPED) {
+				/* Add an unmapped runlist element. */
+				if (!slots) {
+					drl = ntfs_rl_realloc_nofail(drl, ds,
+							ds + 2);
+					slots = 2;
+				}
+				ds++;
+				/* Need to set vcn if it isn't set already. */
+				if (slots != 1)
+					drl[ds].vcn = drl[ds - 1].vcn +
+							drl[ds - 1].length;
+				drl[ds].lcn = LCN_RL_NOT_MAPPED;
+				/* We now used up a slot. */
+				slots--;
+			}
+			drl[ds].length = marker_vcn - drl[ds].vcn;
+			/* Finally add the ENOENT terminator. */
+			ds++;
+			if (!slots)
+				drl = ntfs_rl_realloc_nofail(drl, ds, ds + 1);
+			drl[ds].vcn = marker_vcn;
+			drl[ds].lcn = LCN_ENOENT;
+			drl[ds].length = (s64)0;
+		}
+	}
+	}
+
+finished:
+	/* The merge was completed successfully. */
+	ntfs_debug("Merged runlist:");
+	ntfs_debug_dump_runlist(drl);
+	return drl;
+}
+
+/**
+ * ntfs_mapping_pairs_decompress - convert mapping pairs array to runlist
+ * @vol:	ntfs volume on which the attribute resides
+ * @attr:	attribute record whose mapping pairs array to decompress
+ * @old_rl:	optional runlist in which to insert @attr's runlist
+ *
+ * It is up to the caller to serialize access to the runlist @old_rl.
+ *
+ * Decompress the attribute @attr's mapping pairs array into a runlist. On
+ * success, return the decompressed runlist.
+ *
+ * If @old_rl is not NULL, decompressed runlist is inserted into the
+ * appropriate place in @old_rl and the resultant, combined runlist is
+ * returned. The original @old_rl is deallocated.
+ *
+ * On error, return -errno. @old_rl is left unmodified in that case.
+ *
+ * The following error codes are defined:
+ *	-ENOMEM	- Not enough memory to allocate runlist array.
+ *	-EIO	- Corrupt runlist.
+ *	-EINVAL	- Invalid parameters were passed in.
+ *	-ERANGE	- The two runlists overlap.
+ *
+ * FIXME: For now we take the conceptionally simplest approach of creating the
+ * new runlist disregarding the already existing one and then splicing the
+ * two into one, if that is possible (we check for overlap and discard the new
+ * runlist if overlap present before returning ERR_PTR(-ERANGE)).
+ */
+runlist_element *ntfs_mapping_pairs_decompress(const ntfs_volume *vol,
+		const ATTR_RECORD *attr, runlist_element *old_rl)
+{
+	VCN vcn;		/* Current vcn. */
+	LCN lcn;		/* Current lcn. */
+	s64 deltaxcn;		/* Change in [vl]cn. */
+	runlist_element *rl;	/* The output runlist. */
+	u8 *buf;		/* Current position in mapping pairs array. */
+	u8 *attr_end;		/* End of attribute. */
+	int rlsize;		/* Size of runlist buffer. */
+	u16 rlpos;		/* Current runlist position in units of
+				   runlist_elements. */
+	u8 b;			/* Current byte offset in buf. */
+
+#ifdef DEBUG
+	/* Make sure attr exists and is non-resident. */
+	if (!attr || !attr->non_resident || sle64_to_cpu(
+			attr->data.non_resident.lowest_vcn) < (VCN)0) {
+		ntfs_error(vol->sb, "Invalid arguments.");
+		return ERR_PTR(-EINVAL);
+	}
+#endif
+	/* Start at vcn = lowest_vcn and lcn 0. */
+	vcn = sle64_to_cpu(attr->data.non_resident.lowest_vcn);
+	lcn = 0;
+	/* Get start of the mapping pairs array. */
+	buf = (u8*)attr + le16_to_cpu(
+			attr->data.non_resident.mapping_pairs_offset);
+	attr_end = (u8*)attr + le32_to_cpu(attr->length);
+	if (unlikely(buf < (u8*)attr || buf > attr_end)) {
+		ntfs_error(vol->sb, "Corrupt attribute.");
+		return ERR_PTR(-EIO);
+	}
+	/* If the mapping pairs array is valid but empty, nothing to do. */
+	if (!vcn && !*buf)
+		return old_rl;
+	/* Current position in runlist array. */
+	rlpos = 0;
+	/* Allocate first page and set current runlist size to one page. */
+	rl = ntfs_malloc_nofs(rlsize = PAGE_SIZE);
+	if (unlikely(!rl))
+		return ERR_PTR(-ENOMEM);
+	/* Insert unmapped starting element if necessary. */
+	if (vcn) {
+		rl->vcn = 0;
+		rl->lcn = LCN_RL_NOT_MAPPED;
+		rl->length = vcn;
+		rlpos++;
+	}
+	while (buf < attr_end && *buf) {
+		/*
+		 * Allocate more memory if needed, including space for the
+		 * not-mapped and terminator elements. ntfs_malloc_nofs()
+		 * operates on whole pages only.
+		 */
+		if (((rlpos + 3) * sizeof(*old_rl)) > rlsize) {
+			runlist_element *rl2;
+
+			rl2 = ntfs_malloc_nofs(rlsize + (int)PAGE_SIZE);
+			if (unlikely(!rl2)) {
+				ntfs_free(rl);
+				return ERR_PTR(-ENOMEM);
+			}
+			memcpy(rl2, rl, rlsize);
+			ntfs_free(rl);
+			rl = rl2;
+			rlsize += PAGE_SIZE;
+		}
+		/* Enter the current vcn into the current runlist element. */
+		rl[rlpos].vcn = vcn;
+		/*
+		 * Get the change in vcn, i.e. the run length in clusters.
+		 * Doing it this way ensures that we signextend negative values.
+		 * A negative run length doesn't make any sense, but hey, I
+		 * didn't make up the NTFS specs and Windows NT4 treats the run
+		 * length as a signed value so that's how it is...
+		 */
+		b = *buf & 0xf;
+		if (b) {
+			if (unlikely(buf + b > attr_end))
+				goto io_error;
+			for (deltaxcn = (s8)buf[b--]; b; b--)
+				deltaxcn = (deltaxcn << 8) + buf[b];
+		} else { /* The length entry is compulsory. */
+			ntfs_error(vol->sb, "Missing length entry in mapping "
+					"pairs array.");
+			deltaxcn = (s64)-1;
+		}
+		/*
+		 * Assume a negative length to indicate data corruption and
+		 * hence clean-up and return NULL.
+		 */
+		if (unlikely(deltaxcn < 0)) {
+			ntfs_error(vol->sb, "Invalid length in mapping pairs "
+					"array.");
+			goto err_out;
+		}
+		/*
+		 * Enter the current run length into the current runlist
+		 * element.
+		 */
+		rl[rlpos].length = deltaxcn;
+		/* Increment the current vcn by the current run length. */
+		vcn += deltaxcn;
+		/*
+		 * There might be no lcn change at all, as is the case for
+		 * sparse clusters on NTFS 3.0+, in which case we set the lcn
+		 * to LCN_HOLE.
+		 */
+		if (!(*buf & 0xf0))
+			rl[rlpos].lcn = LCN_HOLE;
+		else {
+			/* Get the lcn change which really can be negative. */
+			u8 b2 = *buf & 0xf;
+			b = b2 + ((*buf >> 4) & 0xf);
+			if (buf + b > attr_end)
+				goto io_error;
+			for (deltaxcn = (s8)buf[b--]; b > b2; b--)
+				deltaxcn = (deltaxcn << 8) + buf[b];
+			/* Change the current lcn to its new value. */
+			lcn += deltaxcn;
+#ifdef DEBUG
+			/*
+			 * On NTFS 1.2-, apparently can have lcn == -1 to
+			 * indicate a hole. But we haven't verified ourselves
+			 * whether it is really the lcn or the deltaxcn that is
+			 * -1. So if either is found give us a message so we
+			 * can investigate it further!
+			 */
+			if (vol->major_ver < 3) {
+				if (unlikely(deltaxcn == (LCN)-1))
+					ntfs_error(vol->sb, "lcn delta == -1");
+				if (unlikely(lcn == (LCN)-1))
+					ntfs_error(vol->sb, "lcn == -1");
+			}
+#endif
+			/* Check lcn is not below -1. */
+			if (unlikely(lcn < (LCN)-1)) {
+				ntfs_error(vol->sb, "Invalid LCN < -1 in "
+						"mapping pairs array.");
+				goto err_out;
+			}
+			/* Enter the current lcn into the runlist element. */
+			rl[rlpos].lcn = lcn;
+		}
+		/* Get to the next runlist element. */
+		rlpos++;
+		/* Increment the buffer position to the next mapping pair. */
+		buf += (*buf & 0xf) + ((*buf >> 4) & 0xf) + 1;
+	}
+	if (unlikely(buf >= attr_end))
+		goto io_error;
+	/*
+	 * If there is a highest_vcn specified, it must be equal to the final
+	 * vcn in the runlist - 1, or something has gone badly wrong.
+	 */
+	deltaxcn = sle64_to_cpu(attr->data.non_resident.highest_vcn);
+	if (unlikely(deltaxcn && vcn - 1 != deltaxcn)) {
+mpa_err:
+		ntfs_error(vol->sb, "Corrupt mapping pairs array in "
+				"non-resident attribute.");
+		goto err_out;
+	}
+	/* Setup not mapped runlist element if this is the base extent. */
+	if (!attr->data.non_resident.lowest_vcn) {
+		VCN max_cluster;
+
+		max_cluster = ((sle64_to_cpu(
+				attr->data.non_resident.allocated_size) +
+				vol->cluster_size - 1) >>
+				vol->cluster_size_bits) - 1;
+		/*
+		 * A highest_vcn of zero means this is a single extent
+		 * attribute so simply terminate the runlist with LCN_ENOENT).
+		 */
+		if (deltaxcn) {
+			/*
+			 * If there is a difference between the highest_vcn and
+			 * the highest cluster, the runlist is either corrupt
+			 * or, more likely, there are more extents following
+			 * this one.
+			 */
+			if (deltaxcn < max_cluster) {
+				ntfs_debug("More extents to follow; deltaxcn "
+						"= 0x%llx, max_cluster = "
+						"0x%llx",
+						(unsigned long long)deltaxcn,
+						(unsigned long long)
+						max_cluster);
+				rl[rlpos].vcn = vcn;
+				vcn += rl[rlpos].length = max_cluster -
+						deltaxcn;
+				rl[rlpos].lcn = LCN_RL_NOT_MAPPED;
+				rlpos++;
+			} else if (unlikely(deltaxcn > max_cluster)) {
+				ntfs_error(vol->sb, "Corrupt attribute.  "
+						"deltaxcn = 0x%llx, "
+						"max_cluster = 0x%llx",
+						(unsigned long long)deltaxcn,
+						(unsigned long long)
+						max_cluster);
+				goto mpa_err;
+			}
+		}
+		rl[rlpos].lcn = LCN_ENOENT;
+	} else /* Not the base extent. There may be more extents to follow. */
+		rl[rlpos].lcn = LCN_RL_NOT_MAPPED;
+
+	/* Setup terminating runlist element. */
+	rl[rlpos].vcn = vcn;
+	rl[rlpos].length = (s64)0;
+	/* If no existing runlist was specified, we are done. */
+	if (!old_rl) {
+		ntfs_debug("Mapping pairs array successfully decompressed:");
+		ntfs_debug_dump_runlist(rl);
+		return rl;
+	}
+	/* Now combine the new and old runlists checking for overlaps. */
+	old_rl = ntfs_runlists_merge(old_rl, rl);
+	if (!IS_ERR(old_rl))
+		return old_rl;
+	ntfs_free(rl);
+	ntfs_error(vol->sb, "Failed to merge runlists.");
+	return old_rl;
+io_error:
+	ntfs_error(vol->sb, "Corrupt attribute.");
+err_out:
+	ntfs_free(rl);
+	return ERR_PTR(-EIO);
+}
+
+/**
+ * ntfs_rl_vcn_to_lcn - convert a vcn into a lcn given a runlist
+ * @rl:		runlist to use for conversion
+ * @vcn:	vcn to convert
+ *
+ * Convert the virtual cluster number @vcn of an attribute into a logical
+ * cluster number (lcn) of a device using the runlist @rl to map vcns to their
+ * corresponding lcns.
+ *
+ * It is up to the caller to serialize access to the runlist @rl.
+ *
+ * Since lcns must be >= 0, we use negative return codes with special meaning:
+ *
+ * Return code		Meaning / Description
+ * ==================================================
+ *  LCN_HOLE		Hole / not allocated on disk.
+ *  LCN_RL_NOT_MAPPED	This is part of the runlist which has not been
+ *			inserted into the runlist yet.
+ *  LCN_ENOENT		There is no such vcn in the attribute.
+ *
+ * Locking: - The caller must have locked the runlist (for reading or writing).
+ *	    - This function does not touch the lock, nor does it modify the
+ *	      runlist.
+ */
+LCN ntfs_rl_vcn_to_lcn(const runlist_element *rl, const VCN vcn)
+{
+	int i;
+
+	BUG_ON(vcn < 0);
+	/*
+	 * If rl is NULL, assume that we have found an unmapped runlist. The
+	 * caller can then attempt to map it and fail appropriately if
+	 * necessary.
+	 */
+	if (unlikely(!rl))
+		return LCN_RL_NOT_MAPPED;
+
+	/* Catch out of lower bounds vcn. */
+	if (unlikely(vcn < rl[0].vcn))
+		return LCN_ENOENT;
+
+	for (i = 0; likely(rl[i].length); i++) {
+		if (unlikely(vcn < rl[i+1].vcn)) {
+			if (likely(rl[i].lcn >= (LCN)0))
+				return rl[i].lcn + (vcn - rl[i].vcn);
+			return rl[i].lcn;
+		}
+	}
+	/*
+	 * The terminator element is setup to the correct value, i.e. one of
+	 * LCN_HOLE, LCN_RL_NOT_MAPPED, or LCN_ENOENT.
+	 */
+	if (likely(rl[i].lcn < (LCN)0))
+		return rl[i].lcn;
+	/* Just in case... We could replace this with BUG() some day. */
+	return LCN_ENOENT;
+}
+
+#ifdef NTFS_RW
+
+/**
+ * ntfs_rl_find_vcn_nolock - find a vcn in a runlist
+ * @rl:		runlist to search
+ * @vcn:	vcn to find
+ *
+ * Find the virtual cluster number @vcn in the runlist @rl and return the
+ * address of the runlist element containing the @vcn on success.
+ *
+ * Return NULL if @rl is NULL or @vcn is in an unmapped part/out of bounds of
+ * the runlist.
+ *
+ * Locking: The runlist must be locked on entry.
+ */
+runlist_element *ntfs_rl_find_vcn_nolock(runlist_element *rl, const VCN vcn)
+{
+	BUG_ON(vcn < 0);
+	if (unlikely(!rl || vcn < rl[0].vcn))
+		return NULL;
+	while (likely(rl->length)) {
+		if (unlikely(vcn < rl[1].vcn)) {
+			if (likely(rl->lcn >= LCN_HOLE))
+				return rl;
+			return NULL;
+		}
+		rl++;
+	}
+	if (likely(rl->lcn == LCN_ENOENT))
+		return rl;
+	return NULL;
+}
+
+/**
+ * ntfs_get_nr_significant_bytes - get number of bytes needed to store a number
+ * @n:		number for which to get the number of bytes for
+ *
+ * Return the number of bytes required to store @n unambiguously as
+ * a signed number.
+ *
+ * This is used in the context of the mapping pairs array to determine how
+ * many bytes will be needed in the array to store a given logical cluster
+ * number (lcn) or a specific run length.
+ *
+ * Return the number of bytes written.  This function cannot fail.
+ */
+static inline int ntfs_get_nr_significant_bytes(const s64 n)
+{
+	s64 l = n;
+	int i;
+	s8 j;
+
+	i = 0;
+	do {
+		l >>= 8;
+		i++;
+	} while (l != 0 && l != -1);
+	j = (n >> 8 * (i - 1)) & 0xff;
+	/* If the sign bit is wrong, we need an extra byte. */
+	if ((n < 0 && j >= 0) || (n > 0 && j < 0))
+		i++;
+	return i;
+}
+
+/**
+ * ntfs_get_size_for_mapping_pairs - get bytes needed for mapping pairs array
+ * @vol:	ntfs volume (needed for the ntfs version)
+ * @rl:		locked runlist to determine the size of the mapping pairs of
+ * @first_vcn:	first vcn which to include in the mapping pairs array
+ * @last_vcn:	last vcn which to include in the mapping pairs array
+ *
+ * Walk the locked runlist @rl and calculate the size in bytes of the mapping
+ * pairs array corresponding to the runlist @rl, starting at vcn @first_vcn and
+ * finishing with vcn @last_vcn.
+ *
+ * A @last_vcn of -1 means end of runlist and in that case the size of the
+ * mapping pairs array corresponding to the runlist starting at vcn @first_vcn
+ * and finishing at the end of the runlist is determined.
+ *
+ * This for example allows us to allocate a buffer of the right size when
+ * building the mapping pairs array.
+ *
+ * If @rl is NULL, just return 1 (for the single terminator byte).
+ *
+ * Return the calculated size in bytes on success.  On error, return -errno.
+ * The following error codes are defined:
+ *	-EINVAL	- Run list contains unmapped elements.  Make sure to only pass
+ *		  fully mapped runlists to this function.
+ *	-EIO	- The runlist is corrupt.
+ *
+ * Locking: @rl must be locked on entry (either for reading or writing), it
+ *	    remains locked throughout, and is left locked upon return.
+ */
+int ntfs_get_size_for_mapping_pairs(const ntfs_volume *vol,
+		const runlist_element *rl, const VCN first_vcn,
+		const VCN last_vcn)
+{
+	LCN prev_lcn;
+	int rls;
+	bool the_end = false;
+
+	BUG_ON(first_vcn < 0);
+	BUG_ON(last_vcn < -1);
+	BUG_ON(last_vcn >= 0 && first_vcn > last_vcn);
+	if (!rl) {
+		BUG_ON(first_vcn);
+		BUG_ON(last_vcn > 0);
+		return 1;
+	}
+	/* Skip to runlist element containing @first_vcn. */
+	while (rl->length && first_vcn >= rl[1].vcn)
+		rl++;
+	if (unlikely((!rl->length && first_vcn > rl->vcn) ||
+			first_vcn < rl->vcn))
+		return -EINVAL;
+	prev_lcn = 0;
+	/* Always need the termining zero byte. */
+	rls = 1;
+	/* Do the first partial run if present. */
+	if (first_vcn > rl->vcn) {
+		s64 delta, length = rl->length;
+
+		/* We know rl->length != 0 already. */
+		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
+			goto err_out;
+		/*
+		 * If @stop_vcn is given and finishes inside this run, cap the
+		 * run length.
+		 */
+		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
+			s64 s1 = last_vcn + 1;
+			if (unlikely(rl[1].vcn > s1))
+				length = s1 - rl->vcn;
+			the_end = true;
+		}
+		delta = first_vcn - rl->vcn;
+		/* Header byte + length. */
+		rls += 1 + ntfs_get_nr_significant_bytes(length - delta);
+		/*
+		 * If the logical cluster number (lcn) denotes a hole and we
+		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
+		 * zero space.  On earlier NTFS versions we just store the lcn.
+		 * Note: this assumes that on NTFS 1.2-, holes are stored with
+		 * an lcn of -1 and not a delta_lcn of -1 (unless both are -1).
+		 */
+		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
+			prev_lcn = rl->lcn;
+			if (likely(rl->lcn >= 0))
+				prev_lcn += delta;
+			/* Change in lcn. */
+			rls += ntfs_get_nr_significant_bytes(prev_lcn);
+		}
+		/* Go to next runlist element. */
+		rl++;
+	}
+	/* Do the full runs. */
+	for (; rl->length && !the_end; rl++) {
+		s64 length = rl->length;
+
+		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
+			goto err_out;
+		/*
+		 * If @stop_vcn is given and finishes inside this run, cap the
+		 * run length.
+		 */
+		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
+			s64 s1 = last_vcn + 1;
+			if (unlikely(rl[1].vcn > s1))
+				length = s1 - rl->vcn;
+			the_end = true;
+		}
+		/* Header byte + length. */
+		rls += 1 + ntfs_get_nr_significant_bytes(length);
+		/*
+		 * If the logical cluster number (lcn) denotes a hole and we
+		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
+		 * zero space.  On earlier NTFS versions we just store the lcn.
+		 * Note: this assumes that on NTFS 1.2-, holes are stored with
+		 * an lcn of -1 and not a delta_lcn of -1 (unless both are -1).
+		 */
+		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
+			/* Change in lcn. */
+			rls += ntfs_get_nr_significant_bytes(rl->lcn -
+					prev_lcn);
+			prev_lcn = rl->lcn;
+		}
+	}
+	return rls;
+err_out:
+	if (rl->lcn == LCN_RL_NOT_MAPPED)
+		rls = -EINVAL;
+	else
+		rls = -EIO;
+	return rls;
+}
+
+/**
+ * ntfs_write_significant_bytes - write the significant bytes of a number
+ * @dst:	destination buffer to write to
+ * @dst_max:	pointer to last byte of destination buffer for bounds checking
+ * @n:		number whose significant bytes to write
+ *
+ * Store in @dst, the minimum bytes of the number @n which are required to
+ * identify @n unambiguously as a signed number, taking care not to exceed
+ * @dest_max, the maximum position within @dst to which we are allowed to
+ * write.
+ *
+ * This is used when building the mapping pairs array of a runlist to compress
+ * a given logical cluster number (lcn) or a specific run length to the minimum
+ * size possible.
+ *
+ * Return the number of bytes written on success.  On error, i.e. the
+ * destination buffer @dst is too small, return -ENOSPC.
+ */
+static inline int ntfs_write_significant_bytes(s8 *dst, const s8 *dst_max,
+		const s64 n)
+{
+	s64 l = n;
+	int i;
+	s8 j;
+
+	i = 0;
+	do {
+		if (unlikely(dst > dst_max))
+			goto err_out;
+		*dst++ = l & 0xffll;
+		l >>= 8;
+		i++;
+	} while (l != 0 && l != -1);
+	j = (n >> 8 * (i - 1)) & 0xff;
+	/* If the sign bit is wrong, we need an extra byte. */
+	if (n < 0 && j >= 0) {
+		if (unlikely(dst > dst_max))
+			goto err_out;
+		i++;
+		*dst = (s8)-1;
+	} else if (n > 0 && j < 0) {
+		if (unlikely(dst > dst_max))
+			goto err_out;
+		i++;
+		*dst = (s8)0;
+	}
+	return i;
+err_out:
+	return -ENOSPC;
+}
+
+/**
+ * ntfs_mapping_pairs_build - build the mapping pairs array from a runlist
+ * @vol:	ntfs volume (needed for the ntfs version)
+ * @dst:	destination buffer to which to write the mapping pairs array
+ * @dst_len:	size of destination buffer @dst in bytes
+ * @rl:		locked runlist for which to build the mapping pairs array
+ * @first_vcn:	first vcn which to include in the mapping pairs array
+ * @last_vcn:	last vcn which to include in the mapping pairs array
+ * @stop_vcn:	first vcn outside destination buffer on success or -ENOSPC
+ *
+ * Create the mapping pairs array from the locked runlist @rl, starting at vcn
+ * @first_vcn and finishing with vcn @last_vcn and save the array in @dst.
+ * @dst_len is the size of @dst in bytes and it should be at least equal to the
+ * value obtained by calling ntfs_get_size_for_mapping_pairs().
+ *
+ * A @last_vcn of -1 means end of runlist and in that case the mapping pairs
+ * array corresponding to the runlist starting at vcn @first_vcn and finishing
+ * at the end of the runlist is created.
+ *
+ * If @rl is NULL, just write a single terminator byte to @dst.
+ *
+ * On success or -ENOSPC error, if @stop_vcn is not NULL, *@stop_vcn is set to
+ * the first vcn outside the destination buffer.  Note that on error, @dst has
+ * been filled with all the mapping pairs that will fit, thus it can be treated
+ * as partial success, in that a new attribute extent needs to be created or
+ * the next extent has to be used and the mapping pairs build has to be
+ * continued with @first_vcn set to *@stop_vcn.
+ *
+ * Return 0 on success and -errno on error.  The following error codes are
+ * defined:
+ *	-EINVAL	- Run list contains unmapped elements.  Make sure to only pass
+ *		  fully mapped runlists to this function.
+ *	-EIO	- The runlist is corrupt.
+ *	-ENOSPC	- The destination buffer is too small.
+ *
+ * Locking: @rl must be locked on entry (either for reading or writing), it
+ *	    remains locked throughout, and is left locked upon return.
+ */
+int ntfs_mapping_pairs_build(const ntfs_volume *vol, s8 *dst,
+		const int dst_len, const runlist_element *rl,
+		const VCN first_vcn, const VCN last_vcn, VCN *const stop_vcn)
+{
+	LCN prev_lcn;
+	s8 *dst_max, *dst_next;
+	int err = -ENOSPC;
+	bool the_end = false;
+	s8 len_len, lcn_len;
+
+	BUG_ON(first_vcn < 0);
+	BUG_ON(last_vcn < -1);
+	BUG_ON(last_vcn >= 0 && first_vcn > last_vcn);
+	BUG_ON(dst_len < 1);
+	if (!rl) {
+		BUG_ON(first_vcn);
+		BUG_ON(last_vcn > 0);
+		if (stop_vcn)
+			*stop_vcn = 0;
+		/* Terminator byte. */
+		*dst = 0;
+		return 0;
+	}
+	/* Skip to runlist element containing @first_vcn. */
+	while (rl->length && first_vcn >= rl[1].vcn)
+		rl++;
+	if (unlikely((!rl->length && first_vcn > rl->vcn) ||
+			first_vcn < rl->vcn))
+		return -EINVAL;
+	/*
+	 * @dst_max is used for bounds checking in
+	 * ntfs_write_significant_bytes().
+	 */
+	dst_max = dst + dst_len - 1;
+	prev_lcn = 0;
+	/* Do the first partial run if present. */
+	if (first_vcn > rl->vcn) {
+		s64 delta, length = rl->length;
+
+		/* We know rl->length != 0 already. */
+		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
+			goto err_out;
+		/*
+		 * If @stop_vcn is given and finishes inside this run, cap the
+		 * run length.
+		 */
+		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
+			s64 s1 = last_vcn + 1;
+			if (unlikely(rl[1].vcn > s1))
+				length = s1 - rl->vcn;
+			the_end = true;
+		}
+		delta = first_vcn - rl->vcn;
+		/* Write length. */
+		len_len = ntfs_write_significant_bytes(dst + 1, dst_max,
+				length - delta);
+		if (unlikely(len_len < 0))
+			goto size_err;
+		/*
+		 * If the logical cluster number (lcn) denotes a hole and we
+		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
+		 * zero space.  On earlier NTFS versions we just write the lcn
+		 * change.  FIXME: Do we need to write the lcn change or just
+		 * the lcn in that case?  Not sure as I have never seen this
+		 * case on NT4. - We assume that we just need to write the lcn
+		 * change until someone tells us otherwise... (AIA)
+		 */
+		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
+			prev_lcn = rl->lcn;
+			if (likely(rl->lcn >= 0))
+				prev_lcn += delta;
+			/* Write change in lcn. */
+			lcn_len = ntfs_write_significant_bytes(dst + 1 +
+					len_len, dst_max, prev_lcn);
+			if (unlikely(lcn_len < 0))
+				goto size_err;
+		} else
+			lcn_len = 0;
+		dst_next = dst + len_len + lcn_len + 1;
+		if (unlikely(dst_next > dst_max))
+			goto size_err;
+		/* Update header byte. */
+		*dst = lcn_len << 4 | len_len;
+		/* Position at next mapping pairs array element. */
+		dst = dst_next;
+		/* Go to next runlist element. */
+		rl++;
+	}
+	/* Do the full runs. */
+	for (; rl->length && !the_end; rl++) {
+		s64 length = rl->length;
+
+		if (unlikely(length < 0 || rl->lcn < LCN_HOLE))
+			goto err_out;
+		/*
+		 * If @stop_vcn is given and finishes inside this run, cap the
+		 * run length.
+		 */
+		if (unlikely(last_vcn >= 0 && rl[1].vcn > last_vcn)) {
+			s64 s1 = last_vcn + 1;
+			if (unlikely(rl[1].vcn > s1))
+				length = s1 - rl->vcn;
+			the_end = true;
+		}
+		/* Write length. */
+		len_len = ntfs_write_significant_bytes(dst + 1, dst_max,
+				length);
+		if (unlikely(len_len < 0))
+			goto size_err;
+		/*
+		 * If the logical cluster number (lcn) denotes a hole and we
+		 * are on NTFS 3.0+, we don't store it at all, i.e. we need
+		 * zero space.  On earlier NTFS versions we just write the lcn
+		 * change.  FIXME: Do we need to write the lcn change or just
+		 * the lcn in that case?  Not sure as I have never seen this
+		 * case on NT4. - We assume that we just need to write the lcn
+		 * change until someone tells us otherwise... (AIA)
+		 */
+		if (likely(rl->lcn >= 0 || vol->major_ver < 3)) {
+			/* Write change in lcn. */
+			lcn_len = ntfs_write_significant_bytes(dst + 1 +
+					len_len, dst_max, rl->lcn - prev_lcn);
+			if (unlikely(lcn_len < 0))
+				goto size_err;
+			prev_lcn = rl->lcn;
+		} else
+			lcn_len = 0;
+		dst_next = dst + len_len + lcn_len + 1;
+		if (unlikely(dst_next > dst_max))
+			goto size_err;
+		/* Update header byte. */
+		*dst = lcn_len << 4 | len_len;
+		/* Position at next mapping pairs array element. */
+		dst = dst_next;
+	}
+	/* Success. */
+	err = 0;
+size_err:
+	/* Set stop vcn. */
+	if (stop_vcn)
+		*stop_vcn = rl->vcn;
+	/* Add terminator byte. */
+	*dst = 0;
+	return err;
+err_out:
+	if (rl->lcn == LCN_RL_NOT_MAPPED)
+		err = -EINVAL;
+	else
+		err = -EIO;
+	return err;
+}
+
+/**
+ * ntfs_rl_truncate_nolock - truncate a runlist starting at a specified vcn
+ * @vol:	ntfs volume (needed for error output)
+ * @runlist:	runlist to truncate
+ * @new_length:	the new length of the runlist in VCNs
+ *
+ * Truncate the runlist described by @runlist as well as the memory buffer
+ * holding the runlist elements to a length of @new_length VCNs.
+ *
+ * If @new_length lies within the runlist, the runlist elements with VCNs of
+ * @new_length and above are discarded.  As a special case if @new_length is
+ * zero, the runlist is discarded and set to NULL.
+ *
+ * If @new_length lies beyond the runlist, a sparse runlist element is added to
+ * the end of the runlist @runlist or if the last runlist element is a sparse
+ * one already, this is extended.
+ *
+ * Note, no checking is done for unmapped runlist elements.  It is assumed that
+ * the caller has mapped any elements that need to be mapped already.
+ *
+ * Return 0 on success and -errno on error.
+ *
+ * Locking: The caller must hold @runlist->lock for writing.
+ */
+int ntfs_rl_truncate_nolock(const ntfs_volume *vol, runlist *const runlist,
+		const s64 new_length)
+{
+	runlist_element *rl;
+	int old_size;
+
+	ntfs_debug("Entering for new_length 0x%llx.", (long long)new_length);
+	BUG_ON(!runlist);
+	BUG_ON(new_length < 0);
+	rl = runlist->rl;
+	if (!new_length) {
+		ntfs_debug("Freeing runlist.");
+		runlist->rl = NULL;
+		if (rl)
+			ntfs_free(rl);
+		return 0;
+	}
+	if (unlikely(!rl)) {
+		/*
+		 * Create a runlist consisting of a sparse runlist element of
+		 * length @new_length followed by a terminator runlist element.
+		 */
+		rl = ntfs_malloc_nofs(PAGE_SIZE);
+		if (unlikely(!rl)) {
+			ntfs_error(vol->sb, "Not enough memory to allocate "
+					"runlist element buffer.");
+			return -ENOMEM;
+		}
+		runlist->rl = rl;
+		rl[1].length = rl->vcn = 0;
+		rl->lcn = LCN_HOLE;
+		rl[1].vcn = rl->length = new_length;
+		rl[1].lcn = LCN_ENOENT;
+		return 0;
+	}
+	BUG_ON(new_length < rl->vcn);
+	/* Find @new_length in the runlist. */
+	while (likely(rl->length && new_length >= rl[1].vcn))
+		rl++;
+	/*
+	 * If not at the end of the runlist we need to shrink it.
+	 * If at the end of the runlist we need to expand it.
+	 */
+	if (rl->length) {
+		runlist_element *trl;
+		bool is_end;
+
+		ntfs_debug("Shrinking runlist.");
+		/* Determine the runlist size. */
+		trl = rl + 1;
+		while (likely(trl->length))
+			trl++;
+		old_size = trl - runlist->rl + 1;
+		/* Truncate the run. */
+		rl->length = new_length - rl->vcn;
+		/*
+		 * If a run was partially truncated, make the following runlist
+		 * element a terminator.
+		 */
+		is_end = false;
+		if (rl->length) {
+			rl++;
+			if (!rl->length)
+				is_end = true;
+			rl->vcn = new_length;
+			rl->length = 0;
+		}
+		rl->lcn = LCN_ENOENT;
+		/* Reallocate memory if necessary. */
+		if (!is_end) {
+			int new_size = rl - runlist->rl + 1;
+			rl = ntfs_rl_realloc(runlist->rl, old_size, new_size);
+			if (IS_ERR(rl))
+				ntfs_warning(vol->sb, "Failed to shrink "
+						"runlist buffer.  This just "
+						"wastes a bit of memory "
+						"temporarily so we ignore it "
+						"and return success.");
+			else
+				runlist->rl = rl;
+		}
+	} else if (likely(/* !rl->length && */ new_length > rl->vcn)) {
+		ntfs_debug("Expanding runlist.");
+		/*
+		 * If there is a previous runlist element and it is a sparse
+		 * one, extend it.  Otherwise need to add a new, sparse runlist
+		 * element.
+		 */
+		if ((rl > runlist->rl) && ((rl - 1)->lcn == LCN_HOLE))
+			(rl - 1)->length = new_length - (rl - 1)->vcn;
+		else {
+			/* Determine the runlist size. */
+			old_size = rl - runlist->rl + 1;
+			/* Reallocate memory if necessary. */
+			rl = ntfs_rl_realloc(runlist->rl, old_size,
+					old_size + 1);
+			if (IS_ERR(rl)) {
+				ntfs_error(vol->sb, "Failed to expand runlist "
+						"buffer, aborting.");
+				return PTR_ERR(rl);
+			}
+			runlist->rl = rl;
+			/*
+			 * Set @rl to the same runlist element in the new
+			 * runlist as before in the old runlist.
+			 */
+			rl += old_size - 1;
+			/* Add a new, sparse runlist element. */
+			rl->lcn = LCN_HOLE;
+			rl->length = new_length - rl->vcn;
+			/* Add a new terminator runlist element. */
+			rl++;
+			rl->length = 0;
+		}
+		rl->vcn = new_length;
+		rl->lcn = LCN_ENOENT;
+	} else /* if (unlikely(!rl->length && new_length == rl->vcn)) */ {
+		/* Runlist already has same size as requested. */
+		rl->lcn = LCN_ENOENT;
+	}
+	ntfs_debug("Done.");
+	return 0;
+}
+
+/**
+ * ntfs_rl_punch_nolock - punch a hole into a runlist
+ * @vol:	ntfs volume (needed for error output)
+ * @runlist:	runlist to punch a hole into
+ * @start:	starting VCN of the hole to be created
+ * @length:	size of the hole to be created in units of clusters
+ *
+ * Punch a hole into the runlist @runlist starting at VCN @start and of size
+ * @length clusters.
+ *
+ * Return 0 on success and -errno on error, in which case @runlist has not been
+ * modified.
+ *
+ * If @start and/or @start + @length are outside the runlist return error code
+ * -ENOENT.
+ *
+ * If the runlist contains unmapped or error elements between @start and @start
+ * + @length return error code -EINVAL.
+ *
+ * Locking: The caller must hold @runlist->lock for writing.
+ */
+int ntfs_rl_punch_nolock(const ntfs_volume *vol, runlist *const runlist,
+		const VCN start, const s64 length)
+{
+	const VCN end = start + length;
+	s64 delta;
+	runlist_element *rl, *rl_end, *rl_real_end, *trl;
+	int old_size;
+	bool lcn_fixup = false;
+
+	ntfs_debug("Entering for start 0x%llx, length 0x%llx.",
+			(long long)start, (long long)length);
+	BUG_ON(!runlist);
+	BUG_ON(start < 0);
+	BUG_ON(length < 0);
+	BUG_ON(end < 0);
+	rl = runlist->rl;
+	if (unlikely(!rl)) {
+		if (likely(!start && !length))
+			return 0;
+		return -ENOENT;
+	}
+	/* Find @start in the runlist. */
+	while (likely(rl->length && start >= rl[1].vcn))
+		rl++;
+	rl_end = rl;
+	/* Find @end in the runlist. */
+	while (likely(rl_end->length && end >= rl_end[1].vcn)) {
+		/* Verify there are no unmapped or error elements. */
+		if (unlikely(rl_end->lcn < LCN_HOLE))
+			return -EINVAL;
+		rl_end++;
+	}
+	/* Check the last element. */
+	if (unlikely(rl_end->length && rl_end->lcn < LCN_HOLE))
+		return -EINVAL;
+	/* This covers @start being out of bounds, too. */
+	if (!rl_end->length && end > rl_end->vcn)
+		return -ENOENT;
+	if (!length)
+		return 0;
+	if (!rl->length)
+		return -ENOENT;
+	rl_real_end = rl_end;
+	/* Determine the runlist size. */
+	while (likely(rl_real_end->length))
+		rl_real_end++;
+	old_size = rl_real_end - runlist->rl + 1;
+	/* If @start is in a hole simply extend the hole. */
+	if (rl->lcn == LCN_HOLE) {
+		/*
+		 * If both @start and @end are in the same sparse run, we are
+		 * done.
+		 */
+		if (end <= rl[1].vcn) {
+			ntfs_debug("Done (requested hole is already sparse).");
+			return 0;
+		}
+extend_hole:
+		/* Extend the hole. */
+		rl->length = end - rl->vcn;
+		/* If @end is in a hole, merge it with the current one. */
+		if (rl_end->lcn == LCN_HOLE) {
+			rl_end++;
+			rl->length = rl_end->vcn - rl->vcn;
+		}
+		/* We have done the hole.  Now deal with the remaining tail. */
+		rl++;
+		/* Cut out all runlist elements up to @end. */
+		if (rl < rl_end)
+			memmove(rl, rl_end, (rl_real_end - rl_end + 1) *
+					sizeof(*rl));
+		/* Adjust the beginning of the tail if necessary. */
+		if (end > rl->vcn) {
+			delta = end - rl->vcn;
+			rl->vcn = end;
+			rl->length -= delta;
+			/* Only adjust the lcn if it is real. */
+			if (rl->lcn >= 0)
+				rl->lcn += delta;
+		}
+shrink_allocation:
+		/* Reallocate memory if the allocation changed. */
+		if (rl < rl_end) {
+			rl = ntfs_rl_realloc(runlist->rl, old_size,
+					old_size - (rl_end - rl));
+			if (IS_ERR(rl))
+				ntfs_warning(vol->sb, "Failed to shrink "
+						"runlist buffer.  This just "
+						"wastes a bit of memory "
+						"temporarily so we ignore it "
+						"and return success.");
+			else
+				runlist->rl = rl;
+		}
+		ntfs_debug("Done (extend hole).");
+		return 0;
+	}
+	/*
+	 * If @start is at the beginning of a run things are easier as there is
+	 * no need to split the first run.
+	 */
+	if (start == rl->vcn) {
+		/*
+		 * @start is at the beginning of a run.
+		 *
+		 * If the previous run is sparse, extend its hole.
+		 *
+		 * If @end is not in the same run, switch the run to be sparse
+		 * and extend the newly created hole.
+		 *
+		 * Thus both of these cases reduce the problem to the above
+		 * case of "@start is in a hole".
+		 */
+		if (rl > runlist->rl && (rl - 1)->lcn == LCN_HOLE) {
+			rl--;
+			goto extend_hole;
+		}
+		if (end >= rl[1].vcn) {
+			rl->lcn = LCN_HOLE;
+			goto extend_hole;
+		}
+		/*
+		 * The final case is when @end is in the same run as @start.
+		 * For this need to split the run into two.  One run for the
+		 * sparse region between the beginning of the old run, i.e.
+		 * @start, and @end and one for the remaining non-sparse
+		 * region, i.e. between @end and the end of the old run.
+		 */
+		trl = ntfs_rl_realloc(runlist->rl, old_size, old_size + 1);
+		if (IS_ERR(trl))
+			goto enomem_out;
+		old_size++;
+		if (runlist->rl != trl) {
+			rl = trl + (rl - runlist->rl);
+			rl_end = trl + (rl_end - runlist->rl);
+			rl_real_end = trl + (rl_real_end - runlist->rl);
+			runlist->rl = trl;
+		}
+split_end:
+		/* Shift all the runs up by one. */
+		memmove(rl + 1, rl, (rl_real_end - rl + 1) * sizeof(*rl));
+		/* Finally, setup the two split runs. */
+		rl->lcn = LCN_HOLE;
+		rl->length = length;
+		rl++;
+		rl->vcn += length;
+		/* Only adjust the lcn if it is real. */
+		if (rl->lcn >= 0 || lcn_fixup)
+			rl->lcn += length;
+		rl->length -= length;
+		ntfs_debug("Done (split one).");
+		return 0;
+	}
+	/*
+	 * @start is neither in a hole nor at the beginning of a run.
+	 *
+	 * If @end is in a hole, things are easier as simply truncating the run
+	 * @start is in to end at @start - 1, deleting all runs after that up
+	 * to @end, and finally extending the beginning of the run @end is in
+	 * to be @start is all that is needed.
+	 */
+	if (rl_end->lcn == LCN_HOLE) {
+		/* Truncate the run containing @start. */
+		rl->length = start - rl->vcn;
+		rl++;
+		/* Cut out all runlist elements up to @end. */
+		if (rl < rl_end)
+			memmove(rl, rl_end, (rl_real_end - rl_end + 1) *
+					sizeof(*rl));
+		/* Extend the beginning of the run @end is in to be @start. */
+		rl->vcn = start;
+		rl->length = rl[1].vcn - start;
+		goto shrink_allocation;
+	}
+	/* 
+	 * If @end is not in a hole there are still two cases to distinguish.
+	 * Either @end is or is not in the same run as @start.
+	 *
+	 * The second case is easier as it can be reduced to an already solved
+	 * problem by truncating the run @start is in to end at @start - 1.
+	 * Then, if @end is in the next run need to split the run into a sparse
+	 * run followed by a non-sparse run (already covered above) and if @end
+	 * is not in the next run switching it to be sparse, again reduces the
+	 * problem to the already covered case of "@start is in a hole".
+	 */
+	if (end >= rl[1].vcn) {
+		/*
+		 * If @end is not in the next run, reduce the problem to the
+		 * case of "@start is in a hole".
+		 */
+		if (rl[1].length && end >= rl[2].vcn) {
+			/* Truncate the run containing @start. */
+			rl->length = start - rl->vcn;
+			rl++;
+			rl->vcn = start;
+			rl->lcn = LCN_HOLE;
+			goto extend_hole;
+		}
+		trl = ntfs_rl_realloc(runlist->rl, old_size, old_size + 1);
+		if (IS_ERR(trl))
+			goto enomem_out;
+		old_size++;
+		if (runlist->rl != trl) {
+			rl = trl + (rl - runlist->rl);
+			rl_end = trl + (rl_end - runlist->rl);
+			rl_real_end = trl + (rl_real_end - runlist->rl);
+			runlist->rl = trl;
+		}
+		/* Truncate the run containing @start. */
+		rl->length = start - rl->vcn;
+		rl++;
+		/*
+		 * @end is in the next run, reduce the problem to the case
+		 * where "@start is at the beginning of a run and @end is in
+		 * the same run as @start".
+		 */
+		delta = rl->vcn - start;
+		rl->vcn = start;
+		if (rl->lcn >= 0) {
+			rl->lcn -= delta;
+			/* Need this in case the lcn just became negative. */
+			lcn_fixup = true;
+		}
+		rl->length += delta;
+		goto split_end;
+	}
+	/*
+	 * The first case from above, i.e. @end is in the same run as @start.
+	 * We need to split the run into three.  One run for the non-sparse
+	 * region between the beginning of the old run and @start, one for the
+	 * sparse region between @start and @end, and one for the remaining
+	 * non-sparse region, i.e. between @end and the end of the old run.
+	 */
+	trl = ntfs_rl_realloc(runlist->rl, old_size, old_size + 2);
+	if (IS_ERR(trl))
+		goto enomem_out;
+	old_size += 2;
+	if (runlist->rl != trl) {
+		rl = trl + (rl - runlist->rl);
+		rl_end = trl + (rl_end - runlist->rl);
+		rl_real_end = trl + (rl_real_end - runlist->rl);
+		runlist->rl = trl;
+	}
+	/* Shift all the runs up by two. */
+	memmove(rl + 2, rl, (rl_real_end - rl + 1) * sizeof(*rl));
+	/* Finally, setup the three split runs. */
+	rl->length = start - rl->vcn;
+	rl++;
+	rl->vcn = start;
+	rl->lcn = LCN_HOLE;
+	rl->length = length;
+	rl++;
+	delta = end - rl->vcn;
+	rl->vcn = end;
+	rl->lcn += delta;
+	rl->length -= delta;
+	ntfs_debug("Done (split both).");
+	return 0;
+enomem_out:
+	ntfs_error(vol->sb, "Not enough memory to extend runlist buffer.");
+	return -ENOMEM;
+}
+
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/runlist.h b/fs/ntfs/runlist.h
new file mode 100644
index 000000000..38de0a375
--- /dev/null
+++ b/fs/ntfs/runlist.h
@@ -0,0 +1,88 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * runlist.h - Defines for runlist handling in NTFS Linux kernel driver.
+ *	       Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2005 Anton Altaparmakov
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#ifndef _LINUX_NTFS_RUNLIST_H
+#define _LINUX_NTFS_RUNLIST_H
+
+#include "types.h"
+#include "layout.h"
+#include "volume.h"
+
+/**
+ * runlist_element - in memory vcn to lcn mapping array element
+ * @vcn:	starting vcn of the current array element
+ * @lcn:	starting lcn of the current array element
+ * @length:	length in clusters of the current array element
+ *
+ * The last vcn (in fact the last vcn + 1) is reached when length == 0.
+ *
+ * When lcn == -1 this means that the count vcns starting at vcn are not
+ * physically allocated (i.e. this is a hole / data is sparse).
+ */
+typedef struct {	/* In memory vcn to lcn mapping structure element. */
+	VCN vcn;	/* vcn = Starting virtual cluster number. */
+	LCN lcn;	/* lcn = Starting logical cluster number. */
+	s64 length;	/* Run length in clusters. */
+} runlist_element;
+
+/**
+ * runlist - in memory vcn to lcn mapping array including a read/write lock
+ * @rl:		pointer to an array of runlist elements
+ * @lock:	read/write spinlock for serializing access to @rl
+ *
+ */
+typedef struct {
+	runlist_element *rl;
+	struct rw_semaphore lock;
+} runlist;
+
+static inline void ntfs_init_runlist(runlist *rl)
+{
+	rl->rl = NULL;
+	init_rwsem(&rl->lock);
+}
+
+typedef enum {
+	LCN_HOLE		= -1,	/* Keep this as highest value or die! */
+	LCN_RL_NOT_MAPPED	= -2,
+	LCN_ENOENT		= -3,
+	LCN_ENOMEM		= -4,
+	LCN_EIO			= -5,
+} LCN_SPECIAL_VALUES;
+
+extern runlist_element *ntfs_runlists_merge(runlist_element *drl,
+		runlist_element *srl);
+
+extern runlist_element *ntfs_mapping_pairs_decompress(const ntfs_volume *vol,
+		const ATTR_RECORD *attr, runlist_element *old_rl);
+
+extern LCN ntfs_rl_vcn_to_lcn(const runlist_element *rl, const VCN vcn);
+
+#ifdef NTFS_RW
+
+extern runlist_element *ntfs_rl_find_vcn_nolock(runlist_element *rl,
+		const VCN vcn);
+
+extern int ntfs_get_size_for_mapping_pairs(const ntfs_volume *vol,
+		const runlist_element *rl, const VCN first_vcn,
+		const VCN last_vcn);
+
+extern int ntfs_mapping_pairs_build(const ntfs_volume *vol, s8 *dst,
+		const int dst_len, const runlist_element *rl,
+		const VCN first_vcn, const VCN last_vcn, VCN *const stop_vcn);
+
+extern int ntfs_rl_truncate_nolock(const ntfs_volume *vol,
+		runlist *const runlist, const s64 new_length);
+
+int ntfs_rl_punch_nolock(const ntfs_volume *vol, runlist *const runlist,
+		const VCN start, const s64 length);
+
+#endif /* NTFS_RW */
+
+#endif /* _LINUX_NTFS_RUNLIST_H */
diff --git a/fs/ntfs/super.c b/fs/ntfs/super.c
new file mode 100644
index 000000000..7f69422d5
--- /dev/null
+++ b/fs/ntfs/super.c
@@ -0,0 +1,3196 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * super.c - NTFS kernel super block handling. Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2012 Anton Altaparmakov and Tuxera Inc.
+ * Copyright (c) 2001,2002 Richard Russon
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/stddef.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/spinlock.h>
+#include <linux/blkdev.h>	/* For bdev_logical_block_size(). */
+#include <linux/backing-dev.h>
+#include <linux/buffer_head.h>
+#include <linux/vfs.h>
+#include <linux/moduleparam.h>
+#include <linux/bitmap.h>
+
+#include "sysctl.h"
+#include "logfile.h"
+#include "quota.h"
+#include "usnjrnl.h"
+#include "dir.h"
+#include "debug.h"
+#include "index.h"
+#include "inode.h"
+#include "aops.h"
+#include "layout.h"
+#include "malloc.h"
+#include "ntfs.h"
+
+/* Number of mounted filesystems which have compression enabled. */
+static unsigned long ntfs_nr_compression_users;
+
+/* A global default upcase table and a corresponding reference count. */
+static ntfschar *default_upcase;
+static unsigned long ntfs_nr_upcase_users;
+
+/* Error constants/strings used in inode.c::ntfs_show_options(). */
+typedef enum {
+	/* One of these must be present, default is ON_ERRORS_CONTINUE. */
+	ON_ERRORS_PANIC			= 0x01,
+	ON_ERRORS_REMOUNT_RO		= 0x02,
+	ON_ERRORS_CONTINUE		= 0x04,
+	/* Optional, can be combined with any of the above. */
+	ON_ERRORS_RECOVER		= 0x10,
+} ON_ERRORS_ACTIONS;
+
+const option_t on_errors_arr[] = {
+	{ ON_ERRORS_PANIC,	"panic" },
+	{ ON_ERRORS_REMOUNT_RO,	"remount-ro", },
+	{ ON_ERRORS_CONTINUE,	"continue", },
+	{ ON_ERRORS_RECOVER,	"recover" },
+	{ 0,			NULL }
+};
+
+/**
+ * simple_getbool -
+ *
+ * Copied from old ntfs driver (which copied from vfat driver).
+ */
+static int simple_getbool(char *s, bool *setval)
+{
+	if (s) {
+		if (!strcmp(s, "1") || !strcmp(s, "yes") || !strcmp(s, "true"))
+			*setval = true;
+		else if (!strcmp(s, "0") || !strcmp(s, "no") ||
+							!strcmp(s, "false"))
+			*setval = false;
+		else
+			return 0;
+	} else
+		*setval = true;
+	return 1;
+}
+
+/**
+ * parse_options - parse the (re)mount options
+ * @vol:	ntfs volume
+ * @opt:	string containing the (re)mount options
+ *
+ * Parse the recognized options in @opt for the ntfs volume described by @vol.
+ */
+static bool parse_options(ntfs_volume *vol, char *opt)
+{
+	char *p, *v, *ov;
+	static char *utf8 = "utf8";
+	int errors = 0, sloppy = 0;
+	kuid_t uid = INVALID_UID;
+	kgid_t gid = INVALID_GID;
+	umode_t fmask = (umode_t)-1, dmask = (umode_t)-1;
+	int mft_zone_multiplier = -1, on_errors = -1;
+	int show_sys_files = -1, case_sensitive = -1, disable_sparse = -1;
+	struct nls_table *nls_map = NULL, *old_nls;
+
+	/* I am lazy... (-8 */
+#define NTFS_GETOPT_WITH_DEFAULT(option, variable, default_value)	\
+	if (!strcmp(p, option)) {					\
+		if (!v || !*v)						\
+			variable = default_value;			\
+		else {							\
+			variable = simple_strtoul(ov = v, &v, 0);	\
+			if (*v)						\
+				goto needs_val;				\
+		}							\
+	}
+#define NTFS_GETOPT(option, variable)					\
+	if (!strcmp(p, option)) {					\
+		if (!v || !*v)						\
+			goto needs_arg;					\
+		variable = simple_strtoul(ov = v, &v, 0);		\
+		if (*v)							\
+			goto needs_val;					\
+	}
+#define NTFS_GETOPT_UID(option, variable)				\
+	if (!strcmp(p, option)) {					\
+		uid_t uid_value;					\
+		if (!v || !*v)						\
+			goto needs_arg;					\
+		uid_value = simple_strtoul(ov = v, &v, 0);		\
+		if (*v)							\
+			goto needs_val;					\
+		variable = make_kuid(current_user_ns(), uid_value);	\
+		if (!uid_valid(variable))				\
+			goto needs_val;					\
+	}
+#define NTFS_GETOPT_GID(option, variable)				\
+	if (!strcmp(p, option)) {					\
+		gid_t gid_value;					\
+		if (!v || !*v)						\
+			goto needs_arg;					\
+		gid_value = simple_strtoul(ov = v, &v, 0);		\
+		if (*v)							\
+			goto needs_val;					\
+		variable = make_kgid(current_user_ns(), gid_value);	\
+		if (!gid_valid(variable))				\
+			goto needs_val;					\
+	}
+#define NTFS_GETOPT_OCTAL(option, variable)				\
+	if (!strcmp(p, option)) {					\
+		if (!v || !*v)						\
+			goto needs_arg;					\
+		variable = simple_strtoul(ov = v, &v, 8);		\
+		if (*v)							\
+			goto needs_val;					\
+	}
+#define NTFS_GETOPT_BOOL(option, variable)				\
+	if (!strcmp(p, option)) {					\
+		bool val;						\
+		if (!simple_getbool(v, &val))				\
+			goto needs_bool;				\
+		variable = val;						\
+	}
+#define NTFS_GETOPT_OPTIONS_ARRAY(option, variable, opt_array)		\
+	if (!strcmp(p, option)) {					\
+		int _i;							\
+		if (!v || !*v)						\
+			goto needs_arg;					\
+		ov = v;							\
+		if (variable == -1)					\
+			variable = 0;					\
+		for (_i = 0; opt_array[_i].str && *opt_array[_i].str; _i++) \
+			if (!strcmp(opt_array[_i].str, v)) {		\
+				variable |= opt_array[_i].val;		\
+				break;					\
+			}						\
+		if (!opt_array[_i].str || !*opt_array[_i].str)		\
+			goto needs_val;					\
+	}
+	if (!opt || !*opt)
+		goto no_mount_options;
+	ntfs_debug("Entering with mount options string: %s", opt);
+	while ((p = strsep(&opt, ","))) {
+		if ((v = strchr(p, '=')))
+			*v++ = 0;
+		NTFS_GETOPT_UID("uid", uid)
+		else NTFS_GETOPT_GID("gid", gid)
+		else NTFS_GETOPT_OCTAL("umask", fmask = dmask)
+		else NTFS_GETOPT_OCTAL("fmask", fmask)
+		else NTFS_GETOPT_OCTAL("dmask", dmask)
+		else NTFS_GETOPT("mft_zone_multiplier", mft_zone_multiplier)
+		else NTFS_GETOPT_WITH_DEFAULT("sloppy", sloppy, true)
+		else NTFS_GETOPT_BOOL("show_sys_files", show_sys_files)
+		else NTFS_GETOPT_BOOL("case_sensitive", case_sensitive)
+		else NTFS_GETOPT_BOOL("disable_sparse", disable_sparse)
+		else NTFS_GETOPT_OPTIONS_ARRAY("errors", on_errors,
+				on_errors_arr)
+		else if (!strcmp(p, "posix") || !strcmp(p, "show_inodes"))
+			ntfs_warning(vol->sb, "Ignoring obsolete option %s.",
+					p);
+		else if (!strcmp(p, "nls") || !strcmp(p, "iocharset")) {
+			if (!strcmp(p, "iocharset"))
+				ntfs_warning(vol->sb, "Option iocharset is "
+						"deprecated. Please use "
+						"option nls=<charsetname> in "
+						"the future.");
+			if (!v || !*v)
+				goto needs_arg;
+use_utf8:
+			old_nls = nls_map;
+			nls_map = load_nls(v);
+			if (!nls_map) {
+				if (!old_nls) {
+					ntfs_error(vol->sb, "NLS character set "
+							"%s not found.", v);
+					return false;
+				}
+				ntfs_error(vol->sb, "NLS character set %s not "
+						"found. Using previous one %s.",
+						v, old_nls->charset);
+				nls_map = old_nls;
+			} else /* nls_map */ {
+				unload_nls(old_nls);
+			}
+		} else if (!strcmp(p, "utf8")) {
+			bool val = false;
+			ntfs_warning(vol->sb, "Option utf8 is no longer "
+				   "supported, using option nls=utf8. Please "
+				   "use option nls=utf8 in the future and "
+				   "make sure utf8 is compiled either as a "
+				   "module or into the kernel.");
+			if (!v || !*v)
+				val = true;
+			else if (!simple_getbool(v, &val))
+				goto needs_bool;
+			if (val) {
+				v = utf8;
+				goto use_utf8;
+			}
+		} else {
+			ntfs_error(vol->sb, "Unrecognized mount option %s.", p);
+			if (errors < INT_MAX)
+				errors++;
+		}
+#undef NTFS_GETOPT_OPTIONS_ARRAY
+#undef NTFS_GETOPT_BOOL
+#undef NTFS_GETOPT
+#undef NTFS_GETOPT_WITH_DEFAULT
+	}
+no_mount_options:
+	if (errors && !sloppy)
+		return false;
+	if (sloppy)
+		ntfs_warning(vol->sb, "Sloppy option given. Ignoring "
+				"unrecognized mount option(s) and continuing.");
+	/* Keep this first! */
+	if (on_errors != -1) {
+		if (!on_errors) {
+			ntfs_error(vol->sb, "Invalid errors option argument "
+					"or bug in options parser.");
+			return false;
+		}
+	}
+	if (nls_map) {
+		if (vol->nls_map && vol->nls_map != nls_map) {
+			ntfs_error(vol->sb, "Cannot change NLS character set "
+					"on remount.");
+			return false;
+		} /* else (!vol->nls_map) */
+		ntfs_debug("Using NLS character set %s.", nls_map->charset);
+		vol->nls_map = nls_map;
+	} else /* (!nls_map) */ {
+		if (!vol->nls_map) {
+			vol->nls_map = load_nls_default();
+			if (!vol->nls_map) {
+				ntfs_error(vol->sb, "Failed to load default "
+						"NLS character set.");
+				return false;
+			}
+			ntfs_debug("Using default NLS character set (%s).",
+					vol->nls_map->charset);
+		}
+	}
+	if (mft_zone_multiplier != -1) {
+		if (vol->mft_zone_multiplier && vol->mft_zone_multiplier !=
+				mft_zone_multiplier) {
+			ntfs_error(vol->sb, "Cannot change mft_zone_multiplier "
+					"on remount.");
+			return false;
+		}
+		if (mft_zone_multiplier < 1 || mft_zone_multiplier > 4) {
+			ntfs_error(vol->sb, "Invalid mft_zone_multiplier. "
+					"Using default value, i.e. 1.");
+			mft_zone_multiplier = 1;
+		}
+		vol->mft_zone_multiplier = mft_zone_multiplier;
+	}
+	if (!vol->mft_zone_multiplier)
+		vol->mft_zone_multiplier = 1;
+	if (on_errors != -1)
+		vol->on_errors = on_errors;
+	if (!vol->on_errors || vol->on_errors == ON_ERRORS_RECOVER)
+		vol->on_errors |= ON_ERRORS_CONTINUE;
+	if (uid_valid(uid))
+		vol->uid = uid;
+	if (gid_valid(gid))
+		vol->gid = gid;
+	if (fmask != (umode_t)-1)
+		vol->fmask = fmask;
+	if (dmask != (umode_t)-1)
+		vol->dmask = dmask;
+	if (show_sys_files != -1) {
+		if (show_sys_files)
+			NVolSetShowSystemFiles(vol);
+		else
+			NVolClearShowSystemFiles(vol);
+	}
+	if (case_sensitive != -1) {
+		if (case_sensitive)
+			NVolSetCaseSensitive(vol);
+		else
+			NVolClearCaseSensitive(vol);
+	}
+	if (disable_sparse != -1) {
+		if (disable_sparse)
+			NVolClearSparseEnabled(vol);
+		else {
+			if (!NVolSparseEnabled(vol) &&
+					vol->major_ver && vol->major_ver < 3)
+				ntfs_warning(vol->sb, "Not enabling sparse "
+						"support due to NTFS volume "
+						"version %i.%i (need at least "
+						"version 3.0).", vol->major_ver,
+						vol->minor_ver);
+			else
+				NVolSetSparseEnabled(vol);
+		}
+	}
+	return true;
+needs_arg:
+	ntfs_error(vol->sb, "The %s option requires an argument.", p);
+	return false;
+needs_bool:
+	ntfs_error(vol->sb, "The %s option requires a boolean argument.", p);
+	return false;
+needs_val:
+	ntfs_error(vol->sb, "Invalid %s option argument: %s", p, ov);
+	return false;
+}
+
+#ifdef NTFS_RW
+
+/**
+ * ntfs_write_volume_flags - write new flags to the volume information flags
+ * @vol:	ntfs volume on which to modify the flags
+ * @flags:	new flags value for the volume information flags
+ *
+ * Internal function.  You probably want to use ntfs_{set,clear}_volume_flags()
+ * instead (see below).
+ *
+ * Replace the volume information flags on the volume @vol with the value
+ * supplied in @flags.  Note, this overwrites the volume information flags, so
+ * make sure to combine the flags you want to modify with the old flags and use
+ * the result when calling ntfs_write_volume_flags().
+ *
+ * Return 0 on success and -errno on error.
+ */
+static int ntfs_write_volume_flags(ntfs_volume *vol, const VOLUME_FLAGS flags)
+{
+	ntfs_inode *ni = NTFS_I(vol->vol_ino);
+	MFT_RECORD *m;
+	VOLUME_INFORMATION *vi;
+	ntfs_attr_search_ctx *ctx;
+	int err;
+
+	ntfs_debug("Entering, old flags = 0x%x, new flags = 0x%x.",
+			le16_to_cpu(vol->vol_flags), le16_to_cpu(flags));
+	if (vol->vol_flags == flags)
+		goto done;
+	BUG_ON(!ni);
+	m = map_mft_record(ni);
+	if (IS_ERR(m)) {
+		err = PTR_ERR(m);
+		goto err_out;
+	}
+	ctx = ntfs_attr_get_search_ctx(ni, m);
+	if (!ctx) {
+		err = -ENOMEM;
+		goto put_unm_err_out;
+	}
+	err = ntfs_attr_lookup(AT_VOLUME_INFORMATION, NULL, 0, 0, 0, NULL, 0,
+			ctx);
+	if (err)
+		goto put_unm_err_out;
+	vi = (VOLUME_INFORMATION*)((u8*)ctx->attr +
+			le16_to_cpu(ctx->attr->data.resident.value_offset));
+	vol->vol_flags = vi->flags = flags;
+	flush_dcache_mft_record_page(ctx->ntfs_ino);
+	mark_mft_record_dirty(ctx->ntfs_ino);
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(ni);
+done:
+	ntfs_debug("Done.");
+	return 0;
+put_unm_err_out:
+	if (ctx)
+		ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(ni);
+err_out:
+	ntfs_error(vol->sb, "Failed with error code %i.", -err);
+	return err;
+}
+
+/**
+ * ntfs_set_volume_flags - set bits in the volume information flags
+ * @vol:	ntfs volume on which to modify the flags
+ * @flags:	flags to set on the volume
+ *
+ * Set the bits in @flags in the volume information flags on the volume @vol.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static inline int ntfs_set_volume_flags(ntfs_volume *vol, VOLUME_FLAGS flags)
+{
+	flags &= VOLUME_FLAGS_MASK;
+	return ntfs_write_volume_flags(vol, vol->vol_flags | flags);
+}
+
+/**
+ * ntfs_clear_volume_flags - clear bits in the volume information flags
+ * @vol:	ntfs volume on which to modify the flags
+ * @flags:	flags to clear on the volume
+ *
+ * Clear the bits in @flags in the volume information flags on the volume @vol.
+ *
+ * Return 0 on success and -errno on error.
+ */
+static inline int ntfs_clear_volume_flags(ntfs_volume *vol, VOLUME_FLAGS flags)
+{
+	flags &= VOLUME_FLAGS_MASK;
+	flags = vol->vol_flags & cpu_to_le16(~le16_to_cpu(flags));
+	return ntfs_write_volume_flags(vol, flags);
+}
+
+#endif /* NTFS_RW */
+
+/**
+ * ntfs_remount - change the mount options of a mounted ntfs filesystem
+ * @sb:		superblock of mounted ntfs filesystem
+ * @flags:	remount flags
+ * @opt:	remount options string
+ *
+ * Change the mount options of an already mounted ntfs filesystem.
+ *
+ * NOTE:  The VFS sets the @sb->s_flags remount flags to @flags after
+ * ntfs_remount() returns successfully (i.e. returns 0).  Otherwise,
+ * @sb->s_flags are not changed.
+ */
+static int ntfs_remount(struct super_block *sb, int *flags, char *opt)
+{
+	ntfs_volume *vol = NTFS_SB(sb);
+
+	ntfs_debug("Entering with remount options string: %s", opt);
+
+	sync_filesystem(sb);
+
+#ifndef NTFS_RW
+	/* For read-only compiled driver, enforce read-only flag. */
+	*flags |= SB_RDONLY;
+#else /* NTFS_RW */
+	/*
+	 * For the read-write compiled driver, if we are remounting read-write,
+	 * make sure there are no volume errors and that no unsupported volume
+	 * flags are set.  Also, empty the logfile journal as it would become
+	 * stale as soon as something is written to the volume and mark the
+	 * volume dirty so that chkdsk is run if the volume is not umounted
+	 * cleanly.  Finally, mark the quotas out of date so Windows rescans
+	 * the volume on boot and updates them.
+	 *
+	 * When remounting read-only, mark the volume clean if no volume errors
+	 * have occurred.
+	 */
+	if (sb_rdonly(sb) && !(*flags & SB_RDONLY)) {
+		static const char *es = ".  Cannot remount read-write.";
+
+		/* Remounting read-write. */
+		if (NVolErrors(vol)) {
+			ntfs_error(sb, "Volume has errors and is read-only%s",
+					es);
+			return -EROFS;
+		}
+		if (vol->vol_flags & VOLUME_IS_DIRTY) {
+			ntfs_error(sb, "Volume is dirty and read-only%s", es);
+			return -EROFS;
+		}
+		if (vol->vol_flags & VOLUME_MODIFIED_BY_CHKDSK) {
+			ntfs_error(sb, "Volume has been modified by chkdsk "
+					"and is read-only%s", es);
+			return -EROFS;
+		}
+		if (vol->vol_flags & VOLUME_MUST_MOUNT_RO_MASK) {
+			ntfs_error(sb, "Volume has unsupported flags set "
+					"(0x%x) and is read-only%s",
+					(unsigned)le16_to_cpu(vol->vol_flags),
+					es);
+			return -EROFS;
+		}
+		if (ntfs_set_volume_flags(vol, VOLUME_IS_DIRTY)) {
+			ntfs_error(sb, "Failed to set dirty bit in volume "
+					"information flags%s", es);
+			return -EROFS;
+		}
+#if 0
+		// TODO: Enable this code once we start modifying anything that
+		//	 is different between NTFS 1.2 and 3.x...
+		/* Set NT4 compatibility flag on newer NTFS version volumes. */
+		if ((vol->major_ver > 1)) {
+			if (ntfs_set_volume_flags(vol, VOLUME_MOUNTED_ON_NT4)) {
+				ntfs_error(sb, "Failed to set NT4 "
+						"compatibility flag%s", es);
+				NVolSetErrors(vol);
+				return -EROFS;
+			}
+		}
+#endif
+		if (!ntfs_empty_logfile(vol->logfile_ino)) {
+			ntfs_error(sb, "Failed to empty journal $LogFile%s",
+					es);
+			NVolSetErrors(vol);
+			return -EROFS;
+		}
+		if (!ntfs_mark_quotas_out_of_date(vol)) {
+			ntfs_error(sb, "Failed to mark quotas out of date%s",
+					es);
+			NVolSetErrors(vol);
+			return -EROFS;
+		}
+		if (!ntfs_stamp_usnjrnl(vol)) {
+			ntfs_error(sb, "Failed to stamp transaction log "
+					"($UsnJrnl)%s", es);
+			NVolSetErrors(vol);
+			return -EROFS;
+		}
+	} else if (!sb_rdonly(sb) && (*flags & SB_RDONLY)) {
+		/* Remounting read-only. */
+		if (!NVolErrors(vol)) {
+			if (ntfs_clear_volume_flags(vol, VOLUME_IS_DIRTY))
+				ntfs_warning(sb, "Failed to clear dirty bit "
+						"in volume information "
+						"flags.  Run chkdsk.");
+		}
+	}
+#endif /* NTFS_RW */
+
+	// TODO: Deal with *flags.
+
+	if (!parse_options(vol, opt))
+		return -EINVAL;
+
+	ntfs_debug("Done.");
+	return 0;
+}
+
+/**
+ * is_boot_sector_ntfs - check whether a boot sector is a valid NTFS boot sector
+ * @sb:		Super block of the device to which @b belongs.
+ * @b:		Boot sector of device @sb to check.
+ * @silent:	If 'true', all output will be silenced.
+ *
+ * is_boot_sector_ntfs() checks whether the boot sector @b is a valid NTFS boot
+ * sector. Returns 'true' if it is valid and 'false' if not.
+ *
+ * @sb is only needed for warning/error output, i.e. it can be NULL when silent
+ * is 'true'.
+ */
+static bool is_boot_sector_ntfs(const struct super_block *sb,
+		const NTFS_BOOT_SECTOR *b, const bool silent)
+{
+	/*
+	 * Check that checksum == sum of u32 values from b to the checksum
+	 * field.  If checksum is zero, no checking is done.  We will work when
+	 * the checksum test fails, since some utilities update the boot sector
+	 * ignoring the checksum which leaves the checksum out-of-date.  We
+	 * report a warning if this is the case.
+	 */
+	if ((void*)b < (void*)&b->checksum && b->checksum && !silent) {
+		le32 *u;
+		u32 i;
+
+		for (i = 0, u = (le32*)b; u < (le32*)(&b->checksum); ++u)
+			i += le32_to_cpup(u);
+		if (le32_to_cpu(b->checksum) != i)
+			ntfs_warning(sb, "Invalid boot sector checksum.");
+	}
+	/* Check OEMidentifier is "NTFS    " */
+	if (b->oem_id != magicNTFS)
+		goto not_ntfs;
+	/* Check bytes per sector value is between 256 and 4096. */
+	if (le16_to_cpu(b->bpb.bytes_per_sector) < 0x100 ||
+			le16_to_cpu(b->bpb.bytes_per_sector) > 0x1000)
+		goto not_ntfs;
+	/* Check sectors per cluster value is valid. */
+	switch (b->bpb.sectors_per_cluster) {
+	case 1: case 2: case 4: case 8: case 16: case 32: case 64: case 128:
+		break;
+	default:
+		goto not_ntfs;
+	}
+	/* Check the cluster size is not above the maximum (64kiB). */
+	if ((u32)le16_to_cpu(b->bpb.bytes_per_sector) *
+			b->bpb.sectors_per_cluster > NTFS_MAX_CLUSTER_SIZE)
+		goto not_ntfs;
+	/* Check reserved/unused fields are really zero. */
+	if (le16_to_cpu(b->bpb.reserved_sectors) ||
+			le16_to_cpu(b->bpb.root_entries) ||
+			le16_to_cpu(b->bpb.sectors) ||
+			le16_to_cpu(b->bpb.sectors_per_fat) ||
+			le32_to_cpu(b->bpb.large_sectors) || b->bpb.fats)
+		goto not_ntfs;
+	/* Check clusters per file mft record value is valid. */
+	if ((u8)b->clusters_per_mft_record < 0xe1 ||
+			(u8)b->clusters_per_mft_record > 0xf7)
+		switch (b->clusters_per_mft_record) {
+		case 1: case 2: case 4: case 8: case 16: case 32: case 64:
+			break;
+		default:
+			goto not_ntfs;
+		}
+	/* Check clusters per index block value is valid. */
+	if ((u8)b->clusters_per_index_record < 0xe1 ||
+			(u8)b->clusters_per_index_record > 0xf7)
+		switch (b->clusters_per_index_record) {
+		case 1: case 2: case 4: case 8: case 16: case 32: case 64:
+			break;
+		default:
+			goto not_ntfs;
+		}
+	/*
+	 * Check for valid end of sector marker. We will work without it, but
+	 * many BIOSes will refuse to boot from a bootsector if the magic is
+	 * incorrect, so we emit a warning.
+	 */
+	if (!silent && b->end_of_sector_marker != cpu_to_le16(0xaa55))
+		ntfs_warning(sb, "Invalid end of sector marker.");
+	return true;
+not_ntfs:
+	return false;
+}
+
+/**
+ * read_ntfs_boot_sector - read the NTFS boot sector of a device
+ * @sb:		super block of device to read the boot sector from
+ * @silent:	if true, suppress all output
+ *
+ * Reads the boot sector from the device and validates it. If that fails, tries
+ * to read the backup boot sector, first from the end of the device a-la NT4 and
+ * later and then from the middle of the device a-la NT3.51 and before.
+ *
+ * If a valid boot sector is found but it is not the primary boot sector, we
+ * repair the primary boot sector silently (unless the device is read-only or
+ * the primary boot sector is not accessible).
+ *
+ * NOTE: To call this function, @sb must have the fields s_dev, the ntfs super
+ * block (u.ntfs_sb), nr_blocks and the device flags (s_flags) initialized
+ * to their respective values.
+ *
+ * Return the unlocked buffer head containing the boot sector or NULL on error.
+ */
+static struct buffer_head *read_ntfs_boot_sector(struct super_block *sb,
+		const int silent)
+{
+	const char *read_err_str = "Unable to read %s boot sector.";
+	struct buffer_head *bh_primary, *bh_backup;
+	sector_t nr_blocks = NTFS_SB(sb)->nr_blocks;
+
+	/* Try to read primary boot sector. */
+	if ((bh_primary = sb_bread(sb, 0))) {
+		if (is_boot_sector_ntfs(sb, (NTFS_BOOT_SECTOR*)
+				bh_primary->b_data, silent))
+			return bh_primary;
+		if (!silent)
+			ntfs_error(sb, "Primary boot sector is invalid.");
+	} else if (!silent)
+		ntfs_error(sb, read_err_str, "primary");
+	if (!(NTFS_SB(sb)->on_errors & ON_ERRORS_RECOVER)) {
+		if (bh_primary)
+			brelse(bh_primary);
+		if (!silent)
+			ntfs_error(sb, "Mount option errors=recover not used. "
+					"Aborting without trying to recover.");
+		return NULL;
+	}
+	/* Try to read NT4+ backup boot sector. */
+	if ((bh_backup = sb_bread(sb, nr_blocks - 1))) {
+		if (is_boot_sector_ntfs(sb, (NTFS_BOOT_SECTOR*)
+				bh_backup->b_data, silent))
+			goto hotfix_primary_boot_sector;
+		brelse(bh_backup);
+	} else if (!silent)
+		ntfs_error(sb, read_err_str, "backup");
+	/* Try to read NT3.51- backup boot sector. */
+	if ((bh_backup = sb_bread(sb, nr_blocks >> 1))) {
+		if (is_boot_sector_ntfs(sb, (NTFS_BOOT_SECTOR*)
+				bh_backup->b_data, silent))
+			goto hotfix_primary_boot_sector;
+		if (!silent)
+			ntfs_error(sb, "Could not find a valid backup boot "
+					"sector.");
+		brelse(bh_backup);
+	} else if (!silent)
+		ntfs_error(sb, read_err_str, "backup");
+	/* We failed. Cleanup and return. */
+	if (bh_primary)
+		brelse(bh_primary);
+	return NULL;
+hotfix_primary_boot_sector:
+	if (bh_primary) {
+		/*
+		 * If we managed to read sector zero and the volume is not
+		 * read-only, copy the found, valid backup boot sector to the
+		 * primary boot sector.  Note we only copy the actual boot
+		 * sector structure, not the actual whole device sector as that
+		 * may be bigger and would potentially damage the $Boot system
+		 * file (FIXME: Would be nice to know if the backup boot sector
+		 * on a large sector device contains the whole boot loader or
+		 * just the first 512 bytes).
+		 */
+		if (!sb_rdonly(sb)) {
+			ntfs_warning(sb, "Hot-fix: Recovering invalid primary "
+					"boot sector from backup copy.");
+			memcpy(bh_primary->b_data, bh_backup->b_data,
+					NTFS_BLOCK_SIZE);
+			mark_buffer_dirty(bh_primary);
+			sync_dirty_buffer(bh_primary);
+			if (buffer_uptodate(bh_primary)) {
+				brelse(bh_backup);
+				return bh_primary;
+			}
+			ntfs_error(sb, "Hot-fix: Device write error while "
+					"recovering primary boot sector.");
+		} else {
+			ntfs_warning(sb, "Hot-fix: Recovery of primary boot "
+					"sector failed: Read-only mount.");
+		}
+		brelse(bh_primary);
+	}
+	ntfs_warning(sb, "Using backup boot sector.");
+	return bh_backup;
+}
+
+/**
+ * parse_ntfs_boot_sector - parse the boot sector and store the data in @vol
+ * @vol:	volume structure to initialise with data from boot sector
+ * @b:		boot sector to parse
+ *
+ * Parse the ntfs boot sector @b and store all imporant information therein in
+ * the ntfs super block @vol.  Return 'true' on success and 'false' on error.
+ */
+static bool parse_ntfs_boot_sector(ntfs_volume *vol, const NTFS_BOOT_SECTOR *b)
+{
+	unsigned int sectors_per_cluster_bits, nr_hidden_sects;
+	int clusters_per_mft_record, clusters_per_index_record;
+	s64 ll;
+
+	vol->sector_size = le16_to_cpu(b->bpb.bytes_per_sector);
+	vol->sector_size_bits = ffs(vol->sector_size) - 1;
+	ntfs_debug("vol->sector_size = %i (0x%x)", vol->sector_size,
+			vol->sector_size);
+	ntfs_debug("vol->sector_size_bits = %i (0x%x)", vol->sector_size_bits,
+			vol->sector_size_bits);
+	if (vol->sector_size < vol->sb->s_blocksize) {
+		ntfs_error(vol->sb, "Sector size (%i) is smaller than the "
+				"device block size (%lu).  This is not "
+				"supported.  Sorry.", vol->sector_size,
+				vol->sb->s_blocksize);
+		return false;
+	}
+	ntfs_debug("sectors_per_cluster = 0x%x", b->bpb.sectors_per_cluster);
+	sectors_per_cluster_bits = ffs(b->bpb.sectors_per_cluster) - 1;
+	ntfs_debug("sectors_per_cluster_bits = 0x%x",
+			sectors_per_cluster_bits);
+	nr_hidden_sects = le32_to_cpu(b->bpb.hidden_sectors);
+	ntfs_debug("number of hidden sectors = 0x%x", nr_hidden_sects);
+	vol->cluster_size = vol->sector_size << sectors_per_cluster_bits;
+	vol->cluster_size_mask = vol->cluster_size - 1;
+	vol->cluster_size_bits = ffs(vol->cluster_size) - 1;
+	ntfs_debug("vol->cluster_size = %i (0x%x)", vol->cluster_size,
+			vol->cluster_size);
+	ntfs_debug("vol->cluster_size_mask = 0x%x", vol->cluster_size_mask);
+	ntfs_debug("vol->cluster_size_bits = %i", vol->cluster_size_bits);
+	if (vol->cluster_size < vol->sector_size) {
+		ntfs_error(vol->sb, "Cluster size (%i) is smaller than the "
+				"sector size (%i).  This is not supported.  "
+				"Sorry.", vol->cluster_size, vol->sector_size);
+		return false;
+	}
+	clusters_per_mft_record = b->clusters_per_mft_record;
+	ntfs_debug("clusters_per_mft_record = %i (0x%x)",
+			clusters_per_mft_record, clusters_per_mft_record);
+	if (clusters_per_mft_record > 0)
+		vol->mft_record_size = vol->cluster_size <<
+				(ffs(clusters_per_mft_record) - 1);
+	else
+		/*
+		 * When mft_record_size < cluster_size, clusters_per_mft_record
+		 * = -log2(mft_record_size) bytes. mft_record_size normaly is
+		 * 1024 bytes, which is encoded as 0xF6 (-10 in decimal).
+		 */
+		vol->mft_record_size = 1 << -clusters_per_mft_record;
+	vol->mft_record_size_mask = vol->mft_record_size - 1;
+	vol->mft_record_size_bits = ffs(vol->mft_record_size) - 1;
+	ntfs_debug("vol->mft_record_size = %i (0x%x)", vol->mft_record_size,
+			vol->mft_record_size);
+	ntfs_debug("vol->mft_record_size_mask = 0x%x",
+			vol->mft_record_size_mask);
+	ntfs_debug("vol->mft_record_size_bits = %i (0x%x)",
+			vol->mft_record_size_bits, vol->mft_record_size_bits);
+	/*
+	 * We cannot support mft record sizes above the PAGE_SIZE since
+	 * we store $MFT/$DATA, the table of mft records in the page cache.
+	 */
+	if (vol->mft_record_size > PAGE_SIZE) {
+		ntfs_error(vol->sb, "Mft record size (%i) exceeds the "
+				"PAGE_SIZE on your system (%lu).  "
+				"This is not supported.  Sorry.",
+				vol->mft_record_size, PAGE_SIZE);
+		return false;
+	}
+	/* We cannot support mft record sizes below the sector size. */
+	if (vol->mft_record_size < vol->sector_size) {
+		ntfs_error(vol->sb, "Mft record size (%i) is smaller than the "
+				"sector size (%i).  This is not supported.  "
+				"Sorry.", vol->mft_record_size,
+				vol->sector_size);
+		return false;
+	}
+	clusters_per_index_record = b->clusters_per_index_record;
+	ntfs_debug("clusters_per_index_record = %i (0x%x)",
+			clusters_per_index_record, clusters_per_index_record);
+	if (clusters_per_index_record > 0)
+		vol->index_record_size = vol->cluster_size <<
+				(ffs(clusters_per_index_record) - 1);
+	else
+		/*
+		 * When index_record_size < cluster_size,
+		 * clusters_per_index_record = -log2(index_record_size) bytes.
+		 * index_record_size normaly equals 4096 bytes, which is
+		 * encoded as 0xF4 (-12 in decimal).
+		 */
+		vol->index_record_size = 1 << -clusters_per_index_record;
+	vol->index_record_size_mask = vol->index_record_size - 1;
+	vol->index_record_size_bits = ffs(vol->index_record_size) - 1;
+	ntfs_debug("vol->index_record_size = %i (0x%x)",
+			vol->index_record_size, vol->index_record_size);
+	ntfs_debug("vol->index_record_size_mask = 0x%x",
+			vol->index_record_size_mask);
+	ntfs_debug("vol->index_record_size_bits = %i (0x%x)",
+			vol->index_record_size_bits,
+			vol->index_record_size_bits);
+	/* We cannot support index record sizes below the sector size. */
+	if (vol->index_record_size < vol->sector_size) {
+		ntfs_error(vol->sb, "Index record size (%i) is smaller than "
+				"the sector size (%i).  This is not "
+				"supported.  Sorry.", vol->index_record_size,
+				vol->sector_size);
+		return false;
+	}
+	/*
+	 * Get the size of the volume in clusters and check for 64-bit-ness.
+	 * Windows currently only uses 32 bits to save the clusters so we do
+	 * the same as it is much faster on 32-bit CPUs.
+	 */
+	ll = sle64_to_cpu(b->number_of_sectors) >> sectors_per_cluster_bits;
+	if ((u64)ll >= 1ULL << 32) {
+		ntfs_error(vol->sb, "Cannot handle 64-bit clusters.  Sorry.");
+		return false;
+	}
+	vol->nr_clusters = ll;
+	ntfs_debug("vol->nr_clusters = 0x%llx", (long long)vol->nr_clusters);
+	/*
+	 * On an architecture where unsigned long is 32-bits, we restrict the
+	 * volume size to 2TiB (2^41). On a 64-bit architecture, the compiler
+	 * will hopefully optimize the whole check away.
+	 */
+	if (sizeof(unsigned long) < 8) {
+		if ((ll << vol->cluster_size_bits) >= (1ULL << 41)) {
+			ntfs_error(vol->sb, "Volume size (%lluTiB) is too "
+					"large for this architecture.  "
+					"Maximum supported is 2TiB.  Sorry.",
+					(unsigned long long)ll >> (40 -
+					vol->cluster_size_bits));
+			return false;
+		}
+	}
+	ll = sle64_to_cpu(b->mft_lcn);
+	if (ll >= vol->nr_clusters) {
+		ntfs_error(vol->sb, "MFT LCN (%lli, 0x%llx) is beyond end of "
+				"volume.  Weird.", (unsigned long long)ll,
+				(unsigned long long)ll);
+		return false;
+	}
+	vol->mft_lcn = ll;
+	ntfs_debug("vol->mft_lcn = 0x%llx", (long long)vol->mft_lcn);
+	ll = sle64_to_cpu(b->mftmirr_lcn);
+	if (ll >= vol->nr_clusters) {
+		ntfs_error(vol->sb, "MFTMirr LCN (%lli, 0x%llx) is beyond end "
+				"of volume.  Weird.", (unsigned long long)ll,
+				(unsigned long long)ll);
+		return false;
+	}
+	vol->mftmirr_lcn = ll;
+	ntfs_debug("vol->mftmirr_lcn = 0x%llx", (long long)vol->mftmirr_lcn);
+#ifdef NTFS_RW
+	/*
+	 * Work out the size of the mft mirror in number of mft records. If the
+	 * cluster size is less than or equal to the size taken by four mft
+	 * records, the mft mirror stores the first four mft records. If the
+	 * cluster size is bigger than the size taken by four mft records, the
+	 * mft mirror contains as many mft records as will fit into one
+	 * cluster.
+	 */
+	if (vol->cluster_size <= (4 << vol->mft_record_size_bits))
+		vol->mftmirr_size = 4;
+	else
+		vol->mftmirr_size = vol->cluster_size >>
+				vol->mft_record_size_bits;
+	ntfs_debug("vol->mftmirr_size = %i", vol->mftmirr_size);
+#endif /* NTFS_RW */
+	vol->serial_no = le64_to_cpu(b->volume_serial_number);
+	ntfs_debug("vol->serial_no = 0x%llx",
+			(unsigned long long)vol->serial_no);
+	return true;
+}
+
+/**
+ * ntfs_setup_allocators - initialize the cluster and mft allocators
+ * @vol:	volume structure for which to setup the allocators
+ *
+ * Setup the cluster (lcn) and mft allocators to the starting values.
+ */
+static void ntfs_setup_allocators(ntfs_volume *vol)
+{
+#ifdef NTFS_RW
+	LCN mft_zone_size, mft_lcn;
+#endif /* NTFS_RW */
+
+	ntfs_debug("vol->mft_zone_multiplier = 0x%x",
+			vol->mft_zone_multiplier);
+#ifdef NTFS_RW
+	/* Determine the size of the MFT zone. */
+	mft_zone_size = vol->nr_clusters;
+	switch (vol->mft_zone_multiplier) {  /* % of volume size in clusters */
+	case 4:
+		mft_zone_size >>= 1;			/* 50%   */
+		break;
+	case 3:
+		mft_zone_size = (mft_zone_size +
+				(mft_zone_size >> 1)) >> 2;	/* 37.5% */
+		break;
+	case 2:
+		mft_zone_size >>= 2;			/* 25%   */
+		break;
+	/* case 1: */
+	default:
+		mft_zone_size >>= 3;			/* 12.5% */
+		break;
+	}
+	/* Setup the mft zone. */
+	vol->mft_zone_start = vol->mft_zone_pos = vol->mft_lcn;
+	ntfs_debug("vol->mft_zone_pos = 0x%llx",
+			(unsigned long long)vol->mft_zone_pos);
+	/*
+	 * Calculate the mft_lcn for an unmodified NTFS volume (see mkntfs
+	 * source) and if the actual mft_lcn is in the expected place or even
+	 * further to the front of the volume, extend the mft_zone to cover the
+	 * beginning of the volume as well.  This is in order to protect the
+	 * area reserved for the mft bitmap as well within the mft_zone itself.
+	 * On non-standard volumes we do not protect it as the overhead would
+	 * be higher than the speed increase we would get by doing it.
+	 */
+	mft_lcn = (8192 + 2 * vol->cluster_size - 1) / vol->cluster_size;
+	if (mft_lcn * vol->cluster_size < 16 * 1024)
+		mft_lcn = (16 * 1024 + vol->cluster_size - 1) /
+				vol->cluster_size;
+	if (vol->mft_zone_start <= mft_lcn)
+		vol->mft_zone_start = 0;
+	ntfs_debug("vol->mft_zone_start = 0x%llx",
+			(unsigned long long)vol->mft_zone_start);
+	/*
+	 * Need to cap the mft zone on non-standard volumes so that it does
+	 * not point outside the boundaries of the volume.  We do this by
+	 * halving the zone size until we are inside the volume.
+	 */
+	vol->mft_zone_end = vol->mft_lcn + mft_zone_size;
+	while (vol->mft_zone_end >= vol->nr_clusters) {
+		mft_zone_size >>= 1;
+		vol->mft_zone_end = vol->mft_lcn + mft_zone_size;
+	}
+	ntfs_debug("vol->mft_zone_end = 0x%llx",
+			(unsigned long long)vol->mft_zone_end);
+	/*
+	 * Set the current position within each data zone to the start of the
+	 * respective zone.
+	 */
+	vol->data1_zone_pos = vol->mft_zone_end;
+	ntfs_debug("vol->data1_zone_pos = 0x%llx",
+			(unsigned long long)vol->data1_zone_pos);
+	vol->data2_zone_pos = 0;
+	ntfs_debug("vol->data2_zone_pos = 0x%llx",
+			(unsigned long long)vol->data2_zone_pos);
+
+	/* Set the mft data allocation position to mft record 24. */
+	vol->mft_data_pos = 24;
+	ntfs_debug("vol->mft_data_pos = 0x%llx",
+			(unsigned long long)vol->mft_data_pos);
+#endif /* NTFS_RW */
+}
+
+#ifdef NTFS_RW
+
+/**
+ * load_and_init_mft_mirror - load and setup the mft mirror inode for a volume
+ * @vol:	ntfs super block describing device whose mft mirror to load
+ *
+ * Return 'true' on success or 'false' on error.
+ */
+static bool load_and_init_mft_mirror(ntfs_volume *vol)
+{
+	struct inode *tmp_ino;
+	ntfs_inode *tmp_ni;
+
+	ntfs_debug("Entering.");
+	/* Get mft mirror inode. */
+	tmp_ino = ntfs_iget(vol->sb, FILE_MFTMirr);
+	if (IS_ERR(tmp_ino) || is_bad_inode(tmp_ino)) {
+		if (!IS_ERR(tmp_ino))
+			iput(tmp_ino);
+		/* Caller will display error message. */
+		return false;
+	}
+	/*
+	 * Re-initialize some specifics about $MFTMirr's inode as
+	 * ntfs_read_inode() will have set up the default ones.
+	 */
+	/* Set uid and gid to root. */
+	tmp_ino->i_uid = GLOBAL_ROOT_UID;
+	tmp_ino->i_gid = GLOBAL_ROOT_GID;
+	/* Regular file.  No access for anyone. */
+	tmp_ino->i_mode = S_IFREG;
+	/* No VFS initiated operations allowed for $MFTMirr. */
+	tmp_ino->i_op = &ntfs_empty_inode_ops;
+	tmp_ino->i_fop = &ntfs_empty_file_ops;
+	/* Put in our special address space operations. */
+	tmp_ino->i_mapping->a_ops = &ntfs_mst_aops;
+	tmp_ni = NTFS_I(tmp_ino);
+	/* The $MFTMirr, like the $MFT is multi sector transfer protected. */
+	NInoSetMstProtected(tmp_ni);
+	NInoSetSparseDisabled(tmp_ni);
+	/*
+	 * Set up our little cheat allowing us to reuse the async read io
+	 * completion handler for directories.
+	 */
+	tmp_ni->itype.index.block_size = vol->mft_record_size;
+	tmp_ni->itype.index.block_size_bits = vol->mft_record_size_bits;
+	vol->mftmirr_ino = tmp_ino;
+	ntfs_debug("Done.");
+	return true;
+}
+
+/**
+ * check_mft_mirror - compare contents of the mft mirror with the mft
+ * @vol:	ntfs super block describing device whose mft mirror to check
+ *
+ * Return 'true' on success or 'false' on error.
+ *
+ * Note, this function also results in the mft mirror runlist being completely
+ * mapped into memory.  The mft mirror write code requires this and will BUG()
+ * should it find an unmapped runlist element.
+ */
+static bool check_mft_mirror(ntfs_volume *vol)
+{
+	struct super_block *sb = vol->sb;
+	ntfs_inode *mirr_ni;
+	struct page *mft_page, *mirr_page;
+	u8 *kmft, *kmirr;
+	runlist_element *rl, rl2[2];
+	pgoff_t index;
+	int mrecs_per_page, i;
+
+	ntfs_debug("Entering.");
+	/* Compare contents of $MFT and $MFTMirr. */
+	mrecs_per_page = PAGE_SIZE / vol->mft_record_size;
+	BUG_ON(!mrecs_per_page);
+	BUG_ON(!vol->mftmirr_size);
+	mft_page = mirr_page = NULL;
+	kmft = kmirr = NULL;
+	index = i = 0;
+	do {
+		u32 bytes;
+
+		/* Switch pages if necessary. */
+		if (!(i % mrecs_per_page)) {
+			if (index) {
+				ntfs_unmap_page(mft_page);
+				ntfs_unmap_page(mirr_page);
+			}
+			/* Get the $MFT page. */
+			mft_page = ntfs_map_page(vol->mft_ino->i_mapping,
+					index);
+			if (IS_ERR(mft_page)) {
+				ntfs_error(sb, "Failed to read $MFT.");
+				return false;
+			}
+			kmft = page_address(mft_page);
+			/* Get the $MFTMirr page. */
+			mirr_page = ntfs_map_page(vol->mftmirr_ino->i_mapping,
+					index);
+			if (IS_ERR(mirr_page)) {
+				ntfs_error(sb, "Failed to read $MFTMirr.");
+				goto mft_unmap_out;
+			}
+			kmirr = page_address(mirr_page);
+			++index;
+		}
+		/* Do not check the record if it is not in use. */
+		if (((MFT_RECORD*)kmft)->flags & MFT_RECORD_IN_USE) {
+			/* Make sure the record is ok. */
+			if (ntfs_is_baad_recordp((le32*)kmft)) {
+				ntfs_error(sb, "Incomplete multi sector "
+						"transfer detected in mft "
+						"record %i.", i);
+mm_unmap_out:
+				ntfs_unmap_page(mirr_page);
+mft_unmap_out:
+				ntfs_unmap_page(mft_page);
+				return false;
+			}
+		}
+		/* Do not check the mirror record if it is not in use. */
+		if (((MFT_RECORD*)kmirr)->flags & MFT_RECORD_IN_USE) {
+			if (ntfs_is_baad_recordp((le32*)kmirr)) {
+				ntfs_error(sb, "Incomplete multi sector "
+						"transfer detected in mft "
+						"mirror record %i.", i);
+				goto mm_unmap_out;
+			}
+		}
+		/* Get the amount of data in the current record. */
+		bytes = le32_to_cpu(((MFT_RECORD*)kmft)->bytes_in_use);
+		if (bytes < sizeof(MFT_RECORD_OLD) ||
+				bytes > vol->mft_record_size ||
+				ntfs_is_baad_recordp((le32*)kmft)) {
+			bytes = le32_to_cpu(((MFT_RECORD*)kmirr)->bytes_in_use);
+			if (bytes < sizeof(MFT_RECORD_OLD) ||
+					bytes > vol->mft_record_size ||
+					ntfs_is_baad_recordp((le32*)kmirr))
+				bytes = vol->mft_record_size;
+		}
+		/* Compare the two records. */
+		if (memcmp(kmft, kmirr, bytes)) {
+			ntfs_error(sb, "$MFT and $MFTMirr (record %i) do not "
+					"match.  Run ntfsfix or chkdsk.", i);
+			goto mm_unmap_out;
+		}
+		kmft += vol->mft_record_size;
+		kmirr += vol->mft_record_size;
+	} while (++i < vol->mftmirr_size);
+	/* Release the last pages. */
+	ntfs_unmap_page(mft_page);
+	ntfs_unmap_page(mirr_page);
+
+	/* Construct the mft mirror runlist by hand. */
+	rl2[0].vcn = 0;
+	rl2[0].lcn = vol->mftmirr_lcn;
+	rl2[0].length = (vol->mftmirr_size * vol->mft_record_size +
+			vol->cluster_size - 1) / vol->cluster_size;
+	rl2[1].vcn = rl2[0].length;
+	rl2[1].lcn = LCN_ENOENT;
+	rl2[1].length = 0;
+	/*
+	 * Because we have just read all of the mft mirror, we know we have
+	 * mapped the full runlist for it.
+	 */
+	mirr_ni = NTFS_I(vol->mftmirr_ino);
+	down_read(&mirr_ni->runlist.lock);
+	rl = mirr_ni->runlist.rl;
+	/* Compare the two runlists.  They must be identical. */
+	i = 0;
+	do {
+		if (rl2[i].vcn != rl[i].vcn || rl2[i].lcn != rl[i].lcn ||
+				rl2[i].length != rl[i].length) {
+			ntfs_error(sb, "$MFTMirr location mismatch.  "
+					"Run chkdsk.");
+			up_read(&mirr_ni->runlist.lock);
+			return false;
+		}
+	} while (rl2[i++].length);
+	up_read(&mirr_ni->runlist.lock);
+	ntfs_debug("Done.");
+	return true;
+}
+
+/**
+ * load_and_check_logfile - load and check the logfile inode for a volume
+ * @vol:	ntfs super block describing device whose logfile to load
+ *
+ * Return 'true' on success or 'false' on error.
+ */
+static bool load_and_check_logfile(ntfs_volume *vol,
+		RESTART_PAGE_HEADER **rp)
+{
+	struct inode *tmp_ino;
+
+	ntfs_debug("Entering.");
+	tmp_ino = ntfs_iget(vol->sb, FILE_LogFile);
+	if (IS_ERR(tmp_ino) || is_bad_inode(tmp_ino)) {
+		if (!IS_ERR(tmp_ino))
+			iput(tmp_ino);
+		/* Caller will display error message. */
+		return false;
+	}
+	if (!ntfs_check_logfile(tmp_ino, rp)) {
+		iput(tmp_ino);
+		/* ntfs_check_logfile() will have displayed error output. */
+		return false;
+	}
+	NInoSetSparseDisabled(NTFS_I(tmp_ino));
+	vol->logfile_ino = tmp_ino;
+	ntfs_debug("Done.");
+	return true;
+}
+
+#define NTFS_HIBERFIL_HEADER_SIZE	4096
+
+/**
+ * check_windows_hibernation_status - check if Windows is suspended on a volume
+ * @vol:	ntfs super block of device to check
+ *
+ * Check if Windows is hibernated on the ntfs volume @vol.  This is done by
+ * looking for the file hiberfil.sys in the root directory of the volume.  If
+ * the file is not present Windows is definitely not suspended.
+ *
+ * If hiberfil.sys exists and is less than 4kiB in size it means Windows is
+ * definitely suspended (this volume is not the system volume).  Caveat:  on a
+ * system with many volumes it is possible that the < 4kiB check is bogus but
+ * for now this should do fine.
+ *
+ * If hiberfil.sys exists and is larger than 4kiB in size, we need to read the
+ * hiberfil header (which is the first 4kiB).  If this begins with "hibr",
+ * Windows is definitely suspended.  If it is completely full of zeroes,
+ * Windows is definitely not hibernated.  Any other case is treated as if
+ * Windows is suspended.  This caters for the above mentioned caveat of a
+ * system with many volumes where no "hibr" magic would be present and there is
+ * no zero header.
+ *
+ * Return 0 if Windows is not hibernated on the volume, >0 if Windows is
+ * hibernated on the volume, and -errno on error.
+ */
+static int check_windows_hibernation_status(ntfs_volume *vol)
+{
+	MFT_REF mref;
+	struct inode *vi;
+	struct page *page;
+	u32 *kaddr, *kend;
+	ntfs_name *name = NULL;
+	int ret = 1;
+	static const ntfschar hiberfil[13] = { cpu_to_le16('h'),
+			cpu_to_le16('i'), cpu_to_le16('b'),
+			cpu_to_le16('e'), cpu_to_le16('r'),
+			cpu_to_le16('f'), cpu_to_le16('i'),
+			cpu_to_le16('l'), cpu_to_le16('.'),
+			cpu_to_le16('s'), cpu_to_le16('y'),
+			cpu_to_le16('s'), 0 };
+
+	ntfs_debug("Entering.");
+	/*
+	 * Find the inode number for the hibernation file by looking up the
+	 * filename hiberfil.sys in the root directory.
+	 */
+	inode_lock(vol->root_ino);
+	mref = ntfs_lookup_inode_by_name(NTFS_I(vol->root_ino), hiberfil, 12,
+			&name);
+	inode_unlock(vol->root_ino);
+	if (IS_ERR_MREF(mref)) {
+		ret = MREF_ERR(mref);
+		/* If the file does not exist, Windows is not hibernated. */
+		if (ret == -ENOENT) {
+			ntfs_debug("hiberfil.sys not present.  Windows is not "
+					"hibernated on the volume.");
+			return 0;
+		}
+		/* A real error occurred. */
+		ntfs_error(vol->sb, "Failed to find inode number for "
+				"hiberfil.sys.");
+		return ret;
+	}
+	/* We do not care for the type of match that was found. */
+	kfree(name);
+	/* Get the inode. */
+	vi = ntfs_iget(vol->sb, MREF(mref));
+	if (IS_ERR(vi) || is_bad_inode(vi)) {
+		if (!IS_ERR(vi))
+			iput(vi);
+		ntfs_error(vol->sb, "Failed to load hiberfil.sys.");
+		return IS_ERR(vi) ? PTR_ERR(vi) : -EIO;
+	}
+	if (unlikely(i_size_read(vi) < NTFS_HIBERFIL_HEADER_SIZE)) {
+		ntfs_debug("hiberfil.sys is smaller than 4kiB (0x%llx).  "
+				"Windows is hibernated on the volume.  This "
+				"is not the system volume.", i_size_read(vi));
+		goto iput_out;
+	}
+	page = ntfs_map_page(vi->i_mapping, 0);
+	if (IS_ERR(page)) {
+		ntfs_error(vol->sb, "Failed to read from hiberfil.sys.");
+		ret = PTR_ERR(page);
+		goto iput_out;
+	}
+	kaddr = (u32*)page_address(page);
+	if (*(le32*)kaddr == cpu_to_le32(0x72626968)/*'hibr'*/) {
+		ntfs_debug("Magic \"hibr\" found in hiberfil.sys.  Windows is "
+				"hibernated on the volume.  This is the "
+				"system volume.");
+		goto unm_iput_out;
+	}
+	kend = kaddr + NTFS_HIBERFIL_HEADER_SIZE/sizeof(*kaddr);
+	do {
+		if (unlikely(*kaddr)) {
+			ntfs_debug("hiberfil.sys is larger than 4kiB "
+					"(0x%llx), does not contain the "
+					"\"hibr\" magic, and does not have a "
+					"zero header.  Windows is hibernated "
+					"on the volume.  This is not the "
+					"system volume.", i_size_read(vi));
+			goto unm_iput_out;
+		}
+	} while (++kaddr < kend);
+	ntfs_debug("hiberfil.sys contains a zero header.  Windows is not "
+			"hibernated on the volume.  This is the system "
+			"volume.");
+	ret = 0;
+unm_iput_out:
+	ntfs_unmap_page(page);
+iput_out:
+	iput(vi);
+	return ret;
+}
+
+/**
+ * load_and_init_quota - load and setup the quota file for a volume if present
+ * @vol:	ntfs super block describing device whose quota file to load
+ *
+ * Return 'true' on success or 'false' on error.  If $Quota is not present, we
+ * leave vol->quota_ino as NULL and return success.
+ */
+static bool load_and_init_quota(ntfs_volume *vol)
+{
+	MFT_REF mref;
+	struct inode *tmp_ino;
+	ntfs_name *name = NULL;
+	static const ntfschar Quota[7] = { cpu_to_le16('$'),
+			cpu_to_le16('Q'), cpu_to_le16('u'),
+			cpu_to_le16('o'), cpu_to_le16('t'),
+			cpu_to_le16('a'), 0 };
+	static ntfschar Q[3] = { cpu_to_le16('$'),
+			cpu_to_le16('Q'), 0 };
+
+	ntfs_debug("Entering.");
+	/*
+	 * Find the inode number for the quota file by looking up the filename
+	 * $Quota in the extended system files directory $Extend.
+	 */
+	inode_lock(vol->extend_ino);
+	mref = ntfs_lookup_inode_by_name(NTFS_I(vol->extend_ino), Quota, 6,
+			&name);
+	inode_unlock(vol->extend_ino);
+	if (IS_ERR_MREF(mref)) {
+		/*
+		 * If the file does not exist, quotas are disabled and have
+		 * never been enabled on this volume, just return success.
+		 */
+		if (MREF_ERR(mref) == -ENOENT) {
+			ntfs_debug("$Quota not present.  Volume does not have "
+					"quotas enabled.");
+			/*
+			 * No need to try to set quotas out of date if they are
+			 * not enabled.
+			 */
+			NVolSetQuotaOutOfDate(vol);
+			return true;
+		}
+		/* A real error occurred. */
+		ntfs_error(vol->sb, "Failed to find inode number for $Quota.");
+		return false;
+	}
+	/* We do not care for the type of match that was found. */
+	kfree(name);
+	/* Get the inode. */
+	tmp_ino = ntfs_iget(vol->sb, MREF(mref));
+	if (IS_ERR(tmp_ino) || is_bad_inode(tmp_ino)) {
+		if (!IS_ERR(tmp_ino))
+			iput(tmp_ino);
+		ntfs_error(vol->sb, "Failed to load $Quota.");
+		return false;
+	}
+	vol->quota_ino = tmp_ino;
+	/* Get the $Q index allocation attribute. */
+	tmp_ino = ntfs_index_iget(vol->quota_ino, Q, 2);
+	if (IS_ERR(tmp_ino)) {
+		ntfs_error(vol->sb, "Failed to load $Quota/$Q index.");
+		return false;
+	}
+	vol->quota_q_ino = tmp_ino;
+	ntfs_debug("Done.");
+	return true;
+}
+
+/**
+ * load_and_init_usnjrnl - load and setup the transaction log if present
+ * @vol:	ntfs super block describing device whose usnjrnl file to load
+ *
+ * Return 'true' on success or 'false' on error.
+ *
+ * If $UsnJrnl is not present or in the process of being disabled, we set
+ * NVolUsnJrnlStamped() and return success.
+ *
+ * If the $UsnJrnl $DATA/$J attribute has a size equal to the lowest valid usn,
+ * i.e. transaction logging has only just been enabled or the journal has been
+ * stamped and nothing has been logged since, we also set NVolUsnJrnlStamped()
+ * and return success.
+ */
+static bool load_and_init_usnjrnl(ntfs_volume *vol)
+{
+	MFT_REF mref;
+	struct inode *tmp_ino;
+	ntfs_inode *tmp_ni;
+	struct page *page;
+	ntfs_name *name = NULL;
+	USN_HEADER *uh;
+	static const ntfschar UsnJrnl[9] = { cpu_to_le16('$'),
+			cpu_to_le16('U'), cpu_to_le16('s'),
+			cpu_to_le16('n'), cpu_to_le16('J'),
+			cpu_to_le16('r'), cpu_to_le16('n'),
+			cpu_to_le16('l'), 0 };
+	static ntfschar Max[5] = { cpu_to_le16('$'),
+			cpu_to_le16('M'), cpu_to_le16('a'),
+			cpu_to_le16('x'), 0 };
+	static ntfschar J[3] = { cpu_to_le16('$'),
+			cpu_to_le16('J'), 0 };
+
+	ntfs_debug("Entering.");
+	/*
+	 * Find the inode number for the transaction log file by looking up the
+	 * filename $UsnJrnl in the extended system files directory $Extend.
+	 */
+	inode_lock(vol->extend_ino);
+	mref = ntfs_lookup_inode_by_name(NTFS_I(vol->extend_ino), UsnJrnl, 8,
+			&name);
+	inode_unlock(vol->extend_ino);
+	if (IS_ERR_MREF(mref)) {
+		/*
+		 * If the file does not exist, transaction logging is disabled,
+		 * just return success.
+		 */
+		if (MREF_ERR(mref) == -ENOENT) {
+			ntfs_debug("$UsnJrnl not present.  Volume does not "
+					"have transaction logging enabled.");
+not_enabled:
+			/*
+			 * No need to try to stamp the transaction log if
+			 * transaction logging is not enabled.
+			 */
+			NVolSetUsnJrnlStamped(vol);
+			return true;
+		}
+		/* A real error occurred. */
+		ntfs_error(vol->sb, "Failed to find inode number for "
+				"$UsnJrnl.");
+		return false;
+	}
+	/* We do not care for the type of match that was found. */
+	kfree(name);
+	/* Get the inode. */
+	tmp_ino = ntfs_iget(vol->sb, MREF(mref));
+	if (IS_ERR(tmp_ino) || unlikely(is_bad_inode(tmp_ino))) {
+		if (!IS_ERR(tmp_ino))
+			iput(tmp_ino);
+		ntfs_error(vol->sb, "Failed to load $UsnJrnl.");
+		return false;
+	}
+	vol->usnjrnl_ino = tmp_ino;
+	/*
+	 * If the transaction log is in the process of being deleted, we can
+	 * ignore it.
+	 */
+	if (unlikely(vol->vol_flags & VOLUME_DELETE_USN_UNDERWAY)) {
+		ntfs_debug("$UsnJrnl in the process of being disabled.  "
+				"Volume does not have transaction logging "
+				"enabled.");
+		goto not_enabled;
+	}
+	/* Get the $DATA/$Max attribute. */
+	tmp_ino = ntfs_attr_iget(vol->usnjrnl_ino, AT_DATA, Max, 4);
+	if (IS_ERR(tmp_ino)) {
+		ntfs_error(vol->sb, "Failed to load $UsnJrnl/$DATA/$Max "
+				"attribute.");
+		return false;
+	}
+	vol->usnjrnl_max_ino = tmp_ino;
+	if (unlikely(i_size_read(tmp_ino) < sizeof(USN_HEADER))) {
+		ntfs_error(vol->sb, "Found corrupt $UsnJrnl/$DATA/$Max "
+				"attribute (size is 0x%llx but should be at "
+				"least 0x%zx bytes).", i_size_read(tmp_ino),
+				sizeof(USN_HEADER));
+		return false;
+	}
+	/* Get the $DATA/$J attribute. */
+	tmp_ino = ntfs_attr_iget(vol->usnjrnl_ino, AT_DATA, J, 2);
+	if (IS_ERR(tmp_ino)) {
+		ntfs_error(vol->sb, "Failed to load $UsnJrnl/$DATA/$J "
+				"attribute.");
+		return false;
+	}
+	vol->usnjrnl_j_ino = tmp_ino;
+	/* Verify $J is non-resident and sparse. */
+	tmp_ni = NTFS_I(vol->usnjrnl_j_ino);
+	if (unlikely(!NInoNonResident(tmp_ni) || !NInoSparse(tmp_ni))) {
+		ntfs_error(vol->sb, "$UsnJrnl/$DATA/$J attribute is resident "
+				"and/or not sparse.");
+		return false;
+	}
+	/* Read the USN_HEADER from $DATA/$Max. */
+	page = ntfs_map_page(vol->usnjrnl_max_ino->i_mapping, 0);
+	if (IS_ERR(page)) {
+		ntfs_error(vol->sb, "Failed to read from $UsnJrnl/$DATA/$Max "
+				"attribute.");
+		return false;
+	}
+	uh = (USN_HEADER*)page_address(page);
+	/* Sanity check the $Max. */
+	if (unlikely(sle64_to_cpu(uh->allocation_delta) >
+			sle64_to_cpu(uh->maximum_size))) {
+		ntfs_error(vol->sb, "Allocation delta (0x%llx) exceeds "
+				"maximum size (0x%llx).  $UsnJrnl is corrupt.",
+				(long long)sle64_to_cpu(uh->allocation_delta),
+				(long long)sle64_to_cpu(uh->maximum_size));
+		ntfs_unmap_page(page);
+		return false;
+	}
+	/*
+	 * If the transaction log has been stamped and nothing has been written
+	 * to it since, we do not need to stamp it.
+	 */
+	if (unlikely(sle64_to_cpu(uh->lowest_valid_usn) >=
+			i_size_read(vol->usnjrnl_j_ino))) {
+		if (likely(sle64_to_cpu(uh->lowest_valid_usn) ==
+				i_size_read(vol->usnjrnl_j_ino))) {
+			ntfs_unmap_page(page);
+			ntfs_debug("$UsnJrnl is enabled but nothing has been "
+					"logged since it was last stamped.  "
+					"Treating this as if the volume does "
+					"not have transaction logging "
+					"enabled.");
+			goto not_enabled;
+		}
+		ntfs_error(vol->sb, "$UsnJrnl has lowest valid usn (0x%llx) "
+				"which is out of bounds (0x%llx).  $UsnJrnl "
+				"is corrupt.",
+				(long long)sle64_to_cpu(uh->lowest_valid_usn),
+				i_size_read(vol->usnjrnl_j_ino));
+		ntfs_unmap_page(page);
+		return false;
+	}
+	ntfs_unmap_page(page);
+	ntfs_debug("Done.");
+	return true;
+}
+
+/**
+ * load_and_init_attrdef - load the attribute definitions table for a volume
+ * @vol:	ntfs super block describing device whose attrdef to load
+ *
+ * Return 'true' on success or 'false' on error.
+ */
+static bool load_and_init_attrdef(ntfs_volume *vol)
+{
+	loff_t i_size;
+	struct super_block *sb = vol->sb;
+	struct inode *ino;
+	struct page *page;
+	pgoff_t index, max_index;
+	unsigned int size;
+
+	ntfs_debug("Entering.");
+	/* Read attrdef table and setup vol->attrdef and vol->attrdef_size. */
+	ino = ntfs_iget(sb, FILE_AttrDef);
+	if (IS_ERR(ino) || is_bad_inode(ino)) {
+		if (!IS_ERR(ino))
+			iput(ino);
+		goto failed;
+	}
+	NInoSetSparseDisabled(NTFS_I(ino));
+	/* The size of FILE_AttrDef must be above 0 and fit inside 31 bits. */
+	i_size = i_size_read(ino);
+	if (i_size <= 0 || i_size > 0x7fffffff)
+		goto iput_failed;
+	vol->attrdef = (ATTR_DEF*)ntfs_malloc_nofs(i_size);
+	if (!vol->attrdef)
+		goto iput_failed;
+	index = 0;
+	max_index = i_size >> PAGE_SHIFT;
+	size = PAGE_SIZE;
+	while (index < max_index) {
+		/* Read the attrdef table and copy it into the linear buffer. */
+read_partial_attrdef_page:
+		page = ntfs_map_page(ino->i_mapping, index);
+		if (IS_ERR(page))
+			goto free_iput_failed;
+		memcpy((u8*)vol->attrdef + (index++ << PAGE_SHIFT),
+				page_address(page), size);
+		ntfs_unmap_page(page);
+	};
+	if (size == PAGE_SIZE) {
+		size = i_size & ~PAGE_MASK;
+		if (size)
+			goto read_partial_attrdef_page;
+	}
+	vol->attrdef_size = i_size;
+	ntfs_debug("Read %llu bytes from $AttrDef.", i_size);
+	iput(ino);
+	return true;
+free_iput_failed:
+	ntfs_free(vol->attrdef);
+	vol->attrdef = NULL;
+iput_failed:
+	iput(ino);
+failed:
+	ntfs_error(sb, "Failed to initialize attribute definition table.");
+	return false;
+}
+
+#endif /* NTFS_RW */
+
+/**
+ * load_and_init_upcase - load the upcase table for an ntfs volume
+ * @vol:	ntfs super block describing device whose upcase to load
+ *
+ * Return 'true' on success or 'false' on error.
+ */
+static bool load_and_init_upcase(ntfs_volume *vol)
+{
+	loff_t i_size;
+	struct super_block *sb = vol->sb;
+	struct inode *ino;
+	struct page *page;
+	pgoff_t index, max_index;
+	unsigned int size;
+	int i, max;
+
+	ntfs_debug("Entering.");
+	/* Read upcase table and setup vol->upcase and vol->upcase_len. */
+	ino = ntfs_iget(sb, FILE_UpCase);
+	if (IS_ERR(ino) || is_bad_inode(ino)) {
+		if (!IS_ERR(ino))
+			iput(ino);
+		goto upcase_failed;
+	}
+	/*
+	 * The upcase size must not be above 64k Unicode characters, must not
+	 * be zero and must be a multiple of sizeof(ntfschar).
+	 */
+	i_size = i_size_read(ino);
+	if (!i_size || i_size & (sizeof(ntfschar) - 1) ||
+			i_size > 64ULL * 1024 * sizeof(ntfschar))
+		goto iput_upcase_failed;
+	vol->upcase = (ntfschar*)ntfs_malloc_nofs(i_size);
+	if (!vol->upcase)
+		goto iput_upcase_failed;
+	index = 0;
+	max_index = i_size >> PAGE_SHIFT;
+	size = PAGE_SIZE;
+	while (index < max_index) {
+		/* Read the upcase table and copy it into the linear buffer. */
+read_partial_upcase_page:
+		page = ntfs_map_page(ino->i_mapping, index);
+		if (IS_ERR(page))
+			goto iput_upcase_failed;
+		memcpy((char*)vol->upcase + (index++ << PAGE_SHIFT),
+				page_address(page), size);
+		ntfs_unmap_page(page);
+	};
+	if (size == PAGE_SIZE) {
+		size = i_size & ~PAGE_MASK;
+		if (size)
+			goto read_partial_upcase_page;
+	}
+	vol->upcase_len = i_size >> UCHAR_T_SIZE_BITS;
+	ntfs_debug("Read %llu bytes from $UpCase (expected %zu bytes).",
+			i_size, 64 * 1024 * sizeof(ntfschar));
+	iput(ino);
+	mutex_lock(&ntfs_lock);
+	if (!default_upcase) {
+		ntfs_debug("Using volume specified $UpCase since default is "
+				"not present.");
+		mutex_unlock(&ntfs_lock);
+		return true;
+	}
+	max = default_upcase_len;
+	if (max > vol->upcase_len)
+		max = vol->upcase_len;
+	for (i = 0; i < max; i++)
+		if (vol->upcase[i] != default_upcase[i])
+			break;
+	if (i == max) {
+		ntfs_free(vol->upcase);
+		vol->upcase = default_upcase;
+		vol->upcase_len = max;
+		ntfs_nr_upcase_users++;
+		mutex_unlock(&ntfs_lock);
+		ntfs_debug("Volume specified $UpCase matches default. Using "
+				"default.");
+		return true;
+	}
+	mutex_unlock(&ntfs_lock);
+	ntfs_debug("Using volume specified $UpCase since it does not match "
+			"the default.");
+	return true;
+iput_upcase_failed:
+	iput(ino);
+	ntfs_free(vol->upcase);
+	vol->upcase = NULL;
+upcase_failed:
+	mutex_lock(&ntfs_lock);
+	if (default_upcase) {
+		vol->upcase = default_upcase;
+		vol->upcase_len = default_upcase_len;
+		ntfs_nr_upcase_users++;
+		mutex_unlock(&ntfs_lock);
+		ntfs_error(sb, "Failed to load $UpCase from the volume. Using "
+				"default.");
+		return true;
+	}
+	mutex_unlock(&ntfs_lock);
+	ntfs_error(sb, "Failed to initialize upcase table.");
+	return false;
+}
+
+/*
+ * The lcn and mft bitmap inodes are NTFS-internal inodes with
+ * their own special locking rules:
+ */
+static struct lock_class_key
+	lcnbmp_runlist_lock_key, lcnbmp_mrec_lock_key,
+	mftbmp_runlist_lock_key, mftbmp_mrec_lock_key;
+
+/**
+ * load_system_files - open the system files using normal functions
+ * @vol:	ntfs super block describing device whose system files to load
+ *
+ * Open the system files with normal access functions and complete setting up
+ * the ntfs super block @vol.
+ *
+ * Return 'true' on success or 'false' on error.
+ */
+static bool load_system_files(ntfs_volume *vol)
+{
+	struct super_block *sb = vol->sb;
+	MFT_RECORD *m;
+	VOLUME_INFORMATION *vi;
+	ntfs_attr_search_ctx *ctx;
+#ifdef NTFS_RW
+	RESTART_PAGE_HEADER *rp;
+	int err;
+#endif /* NTFS_RW */
+
+	ntfs_debug("Entering.");
+#ifdef NTFS_RW
+	/* Get mft mirror inode compare the contents of $MFT and $MFTMirr. */
+	if (!load_and_init_mft_mirror(vol) || !check_mft_mirror(vol)) {
+		static const char *es1 = "Failed to load $MFTMirr";
+		static const char *es2 = "$MFTMirr does not match $MFT";
+		static const char *es3 = ".  Run ntfsfix and/or chkdsk.";
+
+		/* If a read-write mount, convert it to a read-only mount. */
+		if (!sb_rdonly(sb)) {
+			if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+					ON_ERRORS_CONTINUE))) {
+				ntfs_error(sb, "%s and neither on_errors="
+						"continue nor on_errors="
+						"remount-ro was specified%s",
+						!vol->mftmirr_ino ? es1 : es2,
+						es3);
+				goto iput_mirr_err_out;
+			}
+			sb->s_flags |= SB_RDONLY;
+			ntfs_error(sb, "%s.  Mounting read-only%s",
+					!vol->mftmirr_ino ? es1 : es2, es3);
+		} else
+			ntfs_warning(sb, "%s.  Will not be able to remount "
+					"read-write%s",
+					!vol->mftmirr_ino ? es1 : es2, es3);
+		/* This will prevent a read-write remount. */
+		NVolSetErrors(vol);
+	}
+#endif /* NTFS_RW */
+	/* Get mft bitmap attribute inode. */
+	vol->mftbmp_ino = ntfs_attr_iget(vol->mft_ino, AT_BITMAP, NULL, 0);
+	if (IS_ERR(vol->mftbmp_ino)) {
+		ntfs_error(sb, "Failed to load $MFT/$BITMAP attribute.");
+		goto iput_mirr_err_out;
+	}
+	lockdep_set_class(&NTFS_I(vol->mftbmp_ino)->runlist.lock,
+			   &mftbmp_runlist_lock_key);
+	lockdep_set_class(&NTFS_I(vol->mftbmp_ino)->mrec_lock,
+			   &mftbmp_mrec_lock_key);
+	/* Read upcase table and setup @vol->upcase and @vol->upcase_len. */
+	if (!load_and_init_upcase(vol))
+		goto iput_mftbmp_err_out;
+#ifdef NTFS_RW
+	/*
+	 * Read attribute definitions table and setup @vol->attrdef and
+	 * @vol->attrdef_size.
+	 */
+	if (!load_and_init_attrdef(vol))
+		goto iput_upcase_err_out;
+#endif /* NTFS_RW */
+	/*
+	 * Get the cluster allocation bitmap inode and verify the size, no
+	 * need for any locking at this stage as we are already running
+	 * exclusively as we are mount in progress task.
+	 */
+	vol->lcnbmp_ino = ntfs_iget(sb, FILE_Bitmap);
+	if (IS_ERR(vol->lcnbmp_ino) || is_bad_inode(vol->lcnbmp_ino)) {
+		if (!IS_ERR(vol->lcnbmp_ino))
+			iput(vol->lcnbmp_ino);
+		goto bitmap_failed;
+	}
+	lockdep_set_class(&NTFS_I(vol->lcnbmp_ino)->runlist.lock,
+			   &lcnbmp_runlist_lock_key);
+	lockdep_set_class(&NTFS_I(vol->lcnbmp_ino)->mrec_lock,
+			   &lcnbmp_mrec_lock_key);
+
+	NInoSetSparseDisabled(NTFS_I(vol->lcnbmp_ino));
+	if ((vol->nr_clusters + 7) >> 3 > i_size_read(vol->lcnbmp_ino)) {
+		iput(vol->lcnbmp_ino);
+bitmap_failed:
+		ntfs_error(sb, "Failed to load $Bitmap.");
+		goto iput_attrdef_err_out;
+	}
+	/*
+	 * Get the volume inode and setup our cache of the volume flags and
+	 * version.
+	 */
+	vol->vol_ino = ntfs_iget(sb, FILE_Volume);
+	if (IS_ERR(vol->vol_ino) || is_bad_inode(vol->vol_ino)) {
+		if (!IS_ERR(vol->vol_ino))
+			iput(vol->vol_ino);
+volume_failed:
+		ntfs_error(sb, "Failed to load $Volume.");
+		goto iput_lcnbmp_err_out;
+	}
+	m = map_mft_record(NTFS_I(vol->vol_ino));
+	if (IS_ERR(m)) {
+iput_volume_failed:
+		iput(vol->vol_ino);
+		goto volume_failed;
+	}
+	if (!(ctx = ntfs_attr_get_search_ctx(NTFS_I(vol->vol_ino), m))) {
+		ntfs_error(sb, "Failed to get attribute search context.");
+		goto get_ctx_vol_failed;
+	}
+	if (ntfs_attr_lookup(AT_VOLUME_INFORMATION, NULL, 0, 0, 0, NULL, 0,
+			ctx) || ctx->attr->non_resident || ctx->attr->flags) {
+err_put_vol:
+		ntfs_attr_put_search_ctx(ctx);
+get_ctx_vol_failed:
+		unmap_mft_record(NTFS_I(vol->vol_ino));
+		goto iput_volume_failed;
+	}
+	vi = (VOLUME_INFORMATION*)((char*)ctx->attr +
+			le16_to_cpu(ctx->attr->data.resident.value_offset));
+	/* Some bounds checks. */
+	if ((u8*)vi < (u8*)ctx->attr || (u8*)vi +
+			le32_to_cpu(ctx->attr->data.resident.value_length) >
+			(u8*)ctx->attr + le32_to_cpu(ctx->attr->length))
+		goto err_put_vol;
+	/* Copy the volume flags and version to the ntfs_volume structure. */
+	vol->vol_flags = vi->flags;
+	vol->major_ver = vi->major_ver;
+	vol->minor_ver = vi->minor_ver;
+	ntfs_attr_put_search_ctx(ctx);
+	unmap_mft_record(NTFS_I(vol->vol_ino));
+	pr_info("volume version %i.%i.\n", vol->major_ver,
+			vol->minor_ver);
+	if (vol->major_ver < 3 && NVolSparseEnabled(vol)) {
+		ntfs_warning(vol->sb, "Disabling sparse support due to NTFS "
+				"volume version %i.%i (need at least version "
+				"3.0).", vol->major_ver, vol->minor_ver);
+		NVolClearSparseEnabled(vol);
+	}
+#ifdef NTFS_RW
+	/* Make sure that no unsupported volume flags are set. */
+	if (vol->vol_flags & VOLUME_MUST_MOUNT_RO_MASK) {
+		static const char *es1a = "Volume is dirty";
+		static const char *es1b = "Volume has been modified by chkdsk";
+		static const char *es1c = "Volume has unsupported flags set";
+		static const char *es2a = ".  Run chkdsk and mount in Windows.";
+		static const char *es2b = ".  Mount in Windows.";
+		const char *es1, *es2;
+
+		es2 = es2a;
+		if (vol->vol_flags & VOLUME_IS_DIRTY)
+			es1 = es1a;
+		else if (vol->vol_flags & VOLUME_MODIFIED_BY_CHKDSK) {
+			es1 = es1b;
+			es2 = es2b;
+		} else {
+			es1 = es1c;
+			ntfs_warning(sb, "Unsupported volume flags 0x%x "
+					"encountered.",
+					(unsigned)le16_to_cpu(vol->vol_flags));
+		}
+		/* If a read-write mount, convert it to a read-only mount. */
+		if (!sb_rdonly(sb)) {
+			if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+					ON_ERRORS_CONTINUE))) {
+				ntfs_error(sb, "%s and neither on_errors="
+						"continue nor on_errors="
+						"remount-ro was specified%s",
+						es1, es2);
+				goto iput_vol_err_out;
+			}
+			sb->s_flags |= SB_RDONLY;
+			ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		} else
+			ntfs_warning(sb, "%s.  Will not be able to remount "
+					"read-write%s", es1, es2);
+		/*
+		 * Do not set NVolErrors() because ntfs_remount() re-checks the
+		 * flags which we need to do in case any flags have changed.
+		 */
+	}
+	/*
+	 * Get the inode for the logfile, check it and determine if the volume
+	 * was shutdown cleanly.
+	 */
+	rp = NULL;
+	if (!load_and_check_logfile(vol, &rp) ||
+			!ntfs_is_logfile_clean(vol->logfile_ino, rp)) {
+		static const char *es1a = "Failed to load $LogFile";
+		static const char *es1b = "$LogFile is not clean";
+		static const char *es2 = ".  Mount in Windows.";
+		const char *es1;
+
+		es1 = !vol->logfile_ino ? es1a : es1b;
+		/* If a read-write mount, convert it to a read-only mount. */
+		if (!sb_rdonly(sb)) {
+			if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+					ON_ERRORS_CONTINUE))) {
+				ntfs_error(sb, "%s and neither on_errors="
+						"continue nor on_errors="
+						"remount-ro was specified%s",
+						es1, es2);
+				if (vol->logfile_ino) {
+					BUG_ON(!rp);
+					ntfs_free(rp);
+				}
+				goto iput_logfile_err_out;
+			}
+			sb->s_flags |= SB_RDONLY;
+			ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		} else
+			ntfs_warning(sb, "%s.  Will not be able to remount "
+					"read-write%s", es1, es2);
+		/* This will prevent a read-write remount. */
+		NVolSetErrors(vol);
+	}
+	ntfs_free(rp);
+#endif /* NTFS_RW */
+	/* Get the root directory inode so we can do path lookups. */
+	vol->root_ino = ntfs_iget(sb, FILE_root);
+	if (IS_ERR(vol->root_ino) || is_bad_inode(vol->root_ino)) {
+		if (!IS_ERR(vol->root_ino))
+			iput(vol->root_ino);
+		ntfs_error(sb, "Failed to load root directory.");
+		goto iput_logfile_err_out;
+	}
+#ifdef NTFS_RW
+	/*
+	 * Check if Windows is suspended to disk on the target volume.  If it
+	 * is hibernated, we must not write *anything* to the disk so set
+	 * NVolErrors() without setting the dirty volume flag and mount
+	 * read-only.  This will prevent read-write remounting and it will also
+	 * prevent all writes.
+	 */
+	err = check_windows_hibernation_status(vol);
+	if (unlikely(err)) {
+		static const char *es1a = "Failed to determine if Windows is "
+				"hibernated";
+		static const char *es1b = "Windows is hibernated";
+		static const char *es2 = ".  Run chkdsk.";
+		const char *es1;
+
+		es1 = err < 0 ? es1a : es1b;
+		/* If a read-write mount, convert it to a read-only mount. */
+		if (!sb_rdonly(sb)) {
+			if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+					ON_ERRORS_CONTINUE))) {
+				ntfs_error(sb, "%s and neither on_errors="
+						"continue nor on_errors="
+						"remount-ro was specified%s",
+						es1, es2);
+				goto iput_root_err_out;
+			}
+			sb->s_flags |= SB_RDONLY;
+			ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		} else
+			ntfs_warning(sb, "%s.  Will not be able to remount "
+					"read-write%s", es1, es2);
+		/* This will prevent a read-write remount. */
+		NVolSetErrors(vol);
+	}
+	/* If (still) a read-write mount, mark the volume dirty. */
+	if (!sb_rdonly(sb) && ntfs_set_volume_flags(vol, VOLUME_IS_DIRTY)) {
+		static const char *es1 = "Failed to set dirty bit in volume "
+				"information flags";
+		static const char *es2 = ".  Run chkdsk.";
+
+		/* Convert to a read-only mount. */
+		if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+				ON_ERRORS_CONTINUE))) {
+			ntfs_error(sb, "%s and neither on_errors=continue nor "
+					"on_errors=remount-ro was specified%s",
+					es1, es2);
+			goto iput_root_err_out;
+		}
+		ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		sb->s_flags |= SB_RDONLY;
+		/*
+		 * Do not set NVolErrors() because ntfs_remount() might manage
+		 * to set the dirty flag in which case all would be well.
+		 */
+	}
+#if 0
+	// TODO: Enable this code once we start modifying anything that is
+	//	 different between NTFS 1.2 and 3.x...
+	/*
+	 * If (still) a read-write mount, set the NT4 compatibility flag on
+	 * newer NTFS version volumes.
+	 */
+	if (!(sb->s_flags & SB_RDONLY) && (vol->major_ver > 1) &&
+			ntfs_set_volume_flags(vol, VOLUME_MOUNTED_ON_NT4)) {
+		static const char *es1 = "Failed to set NT4 compatibility flag";
+		static const char *es2 = ".  Run chkdsk.";
+
+		/* Convert to a read-only mount. */
+		if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+				ON_ERRORS_CONTINUE))) {
+			ntfs_error(sb, "%s and neither on_errors=continue nor "
+					"on_errors=remount-ro was specified%s",
+					es1, es2);
+			goto iput_root_err_out;
+		}
+		ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		sb->s_flags |= SB_RDONLY;
+		NVolSetErrors(vol);
+	}
+#endif
+	/* If (still) a read-write mount, empty the logfile. */
+	if (!sb_rdonly(sb) && !ntfs_empty_logfile(vol->logfile_ino)) {
+		static const char *es1 = "Failed to empty $LogFile";
+		static const char *es2 = ".  Mount in Windows.";
+
+		/* Convert to a read-only mount. */
+		if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+				ON_ERRORS_CONTINUE))) {
+			ntfs_error(sb, "%s and neither on_errors=continue nor "
+					"on_errors=remount-ro was specified%s",
+					es1, es2);
+			goto iput_root_err_out;
+		}
+		ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		sb->s_flags |= SB_RDONLY;
+		NVolSetErrors(vol);
+	}
+#endif /* NTFS_RW */
+	/* If on NTFS versions before 3.0, we are done. */
+	if (unlikely(vol->major_ver < 3))
+		return true;
+	/* NTFS 3.0+ specific initialization. */
+	/* Get the security descriptors inode. */
+	vol->secure_ino = ntfs_iget(sb, FILE_Secure);
+	if (IS_ERR(vol->secure_ino) || is_bad_inode(vol->secure_ino)) {
+		if (!IS_ERR(vol->secure_ino))
+			iput(vol->secure_ino);
+		ntfs_error(sb, "Failed to load $Secure.");
+		goto iput_root_err_out;
+	}
+	// TODO: Initialize security.
+	/* Get the extended system files' directory inode. */
+	vol->extend_ino = ntfs_iget(sb, FILE_Extend);
+	if (IS_ERR(vol->extend_ino) || is_bad_inode(vol->extend_ino) ||
+	    !S_ISDIR(vol->extend_ino->i_mode)) {
+		if (!IS_ERR(vol->extend_ino))
+			iput(vol->extend_ino);
+		ntfs_error(sb, "Failed to load $Extend.");
+		goto iput_sec_err_out;
+	}
+#ifdef NTFS_RW
+	/* Find the quota file, load it if present, and set it up. */
+	if (!load_and_init_quota(vol)) {
+		static const char *es1 = "Failed to load $Quota";
+		static const char *es2 = ".  Run chkdsk.";
+
+		/* If a read-write mount, convert it to a read-only mount. */
+		if (!sb_rdonly(sb)) {
+			if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+					ON_ERRORS_CONTINUE))) {
+				ntfs_error(sb, "%s and neither on_errors="
+						"continue nor on_errors="
+						"remount-ro was specified%s",
+						es1, es2);
+				goto iput_quota_err_out;
+			}
+			sb->s_flags |= SB_RDONLY;
+			ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		} else
+			ntfs_warning(sb, "%s.  Will not be able to remount "
+					"read-write%s", es1, es2);
+		/* This will prevent a read-write remount. */
+		NVolSetErrors(vol);
+	}
+	/* If (still) a read-write mount, mark the quotas out of date. */
+	if (!sb_rdonly(sb) && !ntfs_mark_quotas_out_of_date(vol)) {
+		static const char *es1 = "Failed to mark quotas out of date";
+		static const char *es2 = ".  Run chkdsk.";
+
+		/* Convert to a read-only mount. */
+		if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+				ON_ERRORS_CONTINUE))) {
+			ntfs_error(sb, "%s and neither on_errors=continue nor "
+					"on_errors=remount-ro was specified%s",
+					es1, es2);
+			goto iput_quota_err_out;
+		}
+		ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		sb->s_flags |= SB_RDONLY;
+		NVolSetErrors(vol);
+	}
+	/*
+	 * Find the transaction log file ($UsnJrnl), load it if present, check
+	 * it, and set it up.
+	 */
+	if (!load_and_init_usnjrnl(vol)) {
+		static const char *es1 = "Failed to load $UsnJrnl";
+		static const char *es2 = ".  Run chkdsk.";
+
+		/* If a read-write mount, convert it to a read-only mount. */
+		if (!sb_rdonly(sb)) {
+			if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+					ON_ERRORS_CONTINUE))) {
+				ntfs_error(sb, "%s and neither on_errors="
+						"continue nor on_errors="
+						"remount-ro was specified%s",
+						es1, es2);
+				goto iput_usnjrnl_err_out;
+			}
+			sb->s_flags |= SB_RDONLY;
+			ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		} else
+			ntfs_warning(sb, "%s.  Will not be able to remount "
+					"read-write%s", es1, es2);
+		/* This will prevent a read-write remount. */
+		NVolSetErrors(vol);
+	}
+	/* If (still) a read-write mount, stamp the transaction log. */
+	if (!sb_rdonly(sb) && !ntfs_stamp_usnjrnl(vol)) {
+		static const char *es1 = "Failed to stamp transaction log "
+				"($UsnJrnl)";
+		static const char *es2 = ".  Run chkdsk.";
+
+		/* Convert to a read-only mount. */
+		if (!(vol->on_errors & (ON_ERRORS_REMOUNT_RO |
+				ON_ERRORS_CONTINUE))) {
+			ntfs_error(sb, "%s and neither on_errors=continue nor "
+					"on_errors=remount-ro was specified%s",
+					es1, es2);
+			goto iput_usnjrnl_err_out;
+		}
+		ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
+		sb->s_flags |= SB_RDONLY;
+		NVolSetErrors(vol);
+	}
+#endif /* NTFS_RW */
+	return true;
+#ifdef NTFS_RW
+iput_usnjrnl_err_out:
+	iput(vol->usnjrnl_j_ino);
+	iput(vol->usnjrnl_max_ino);
+	iput(vol->usnjrnl_ino);
+iput_quota_err_out:
+	iput(vol->quota_q_ino);
+	iput(vol->quota_ino);
+	iput(vol->extend_ino);
+#endif /* NTFS_RW */
+iput_sec_err_out:
+	iput(vol->secure_ino);
+iput_root_err_out:
+	iput(vol->root_ino);
+iput_logfile_err_out:
+#ifdef NTFS_RW
+	iput(vol->logfile_ino);
+iput_vol_err_out:
+#endif /* NTFS_RW */
+	iput(vol->vol_ino);
+iput_lcnbmp_err_out:
+	iput(vol->lcnbmp_ino);
+iput_attrdef_err_out:
+	vol->attrdef_size = 0;
+	if (vol->attrdef) {
+		ntfs_free(vol->attrdef);
+		vol->attrdef = NULL;
+	}
+#ifdef NTFS_RW
+iput_upcase_err_out:
+#endif /* NTFS_RW */
+	vol->upcase_len = 0;
+	mutex_lock(&ntfs_lock);
+	if (vol->upcase == default_upcase) {
+		ntfs_nr_upcase_users--;
+		vol->upcase = NULL;
+	}
+	mutex_unlock(&ntfs_lock);
+	if (vol->upcase) {
+		ntfs_free(vol->upcase);
+		vol->upcase = NULL;
+	}
+iput_mftbmp_err_out:
+	iput(vol->mftbmp_ino);
+iput_mirr_err_out:
+#ifdef NTFS_RW
+	iput(vol->mftmirr_ino);
+#endif /* NTFS_RW */
+	return false;
+}
+
+/**
+ * ntfs_put_super - called by the vfs to unmount a volume
+ * @sb:		vfs superblock of volume to unmount
+ *
+ * ntfs_put_super() is called by the VFS (from fs/super.c::do_umount()) when
+ * the volume is being unmounted (umount system call has been invoked) and it
+ * releases all inodes and memory belonging to the NTFS specific part of the
+ * super block.
+ */
+static void ntfs_put_super(struct super_block *sb)
+{
+	ntfs_volume *vol = NTFS_SB(sb);
+
+	ntfs_debug("Entering.");
+
+#ifdef NTFS_RW
+	/*
+	 * Commit all inodes while they are still open in case some of them
+	 * cause others to be dirtied.
+	 */
+	ntfs_commit_inode(vol->vol_ino);
+
+	/* NTFS 3.0+ specific. */
+	if (vol->major_ver >= 3) {
+		if (vol->usnjrnl_j_ino)
+			ntfs_commit_inode(vol->usnjrnl_j_ino);
+		if (vol->usnjrnl_max_ino)
+			ntfs_commit_inode(vol->usnjrnl_max_ino);
+		if (vol->usnjrnl_ino)
+			ntfs_commit_inode(vol->usnjrnl_ino);
+		if (vol->quota_q_ino)
+			ntfs_commit_inode(vol->quota_q_ino);
+		if (vol->quota_ino)
+			ntfs_commit_inode(vol->quota_ino);
+		if (vol->extend_ino)
+			ntfs_commit_inode(vol->extend_ino);
+		if (vol->secure_ino)
+			ntfs_commit_inode(vol->secure_ino);
+	}
+
+	ntfs_commit_inode(vol->root_ino);
+
+	down_write(&vol->lcnbmp_lock);
+	ntfs_commit_inode(vol->lcnbmp_ino);
+	up_write(&vol->lcnbmp_lock);
+
+	down_write(&vol->mftbmp_lock);
+	ntfs_commit_inode(vol->mftbmp_ino);
+	up_write(&vol->mftbmp_lock);
+
+	if (vol->logfile_ino)
+		ntfs_commit_inode(vol->logfile_ino);
+
+	if (vol->mftmirr_ino)
+		ntfs_commit_inode(vol->mftmirr_ino);
+	ntfs_commit_inode(vol->mft_ino);
+
+	/*
+	 * If a read-write mount and no volume errors have occurred, mark the
+	 * volume clean.  Also, re-commit all affected inodes.
+	 */
+	if (!sb_rdonly(sb)) {
+		if (!NVolErrors(vol)) {
+			if (ntfs_clear_volume_flags(vol, VOLUME_IS_DIRTY))
+				ntfs_warning(sb, "Failed to clear dirty bit "
+						"in volume information "
+						"flags.  Run chkdsk.");
+			ntfs_commit_inode(vol->vol_ino);
+			ntfs_commit_inode(vol->root_ino);
+			if (vol->mftmirr_ino)
+				ntfs_commit_inode(vol->mftmirr_ino);
+			ntfs_commit_inode(vol->mft_ino);
+		} else {
+			ntfs_warning(sb, "Volume has errors.  Leaving volume "
+					"marked dirty.  Run chkdsk.");
+		}
+	}
+#endif /* NTFS_RW */
+
+	iput(vol->vol_ino);
+	vol->vol_ino = NULL;
+
+	/* NTFS 3.0+ specific clean up. */
+	if (vol->major_ver >= 3) {
+#ifdef NTFS_RW
+		if (vol->usnjrnl_j_ino) {
+			iput(vol->usnjrnl_j_ino);
+			vol->usnjrnl_j_ino = NULL;
+		}
+		if (vol->usnjrnl_max_ino) {
+			iput(vol->usnjrnl_max_ino);
+			vol->usnjrnl_max_ino = NULL;
+		}
+		if (vol->usnjrnl_ino) {
+			iput(vol->usnjrnl_ino);
+			vol->usnjrnl_ino = NULL;
+		}
+		if (vol->quota_q_ino) {
+			iput(vol->quota_q_ino);
+			vol->quota_q_ino = NULL;
+		}
+		if (vol->quota_ino) {
+			iput(vol->quota_ino);
+			vol->quota_ino = NULL;
+		}
+#endif /* NTFS_RW */
+		if (vol->extend_ino) {
+			iput(vol->extend_ino);
+			vol->extend_ino = NULL;
+		}
+		if (vol->secure_ino) {
+			iput(vol->secure_ino);
+			vol->secure_ino = NULL;
+		}
+	}
+
+	iput(vol->root_ino);
+	vol->root_ino = NULL;
+
+	down_write(&vol->lcnbmp_lock);
+	iput(vol->lcnbmp_ino);
+	vol->lcnbmp_ino = NULL;
+	up_write(&vol->lcnbmp_lock);
+
+	down_write(&vol->mftbmp_lock);
+	iput(vol->mftbmp_ino);
+	vol->mftbmp_ino = NULL;
+	up_write(&vol->mftbmp_lock);
+
+#ifdef NTFS_RW
+	if (vol->logfile_ino) {
+		iput(vol->logfile_ino);
+		vol->logfile_ino = NULL;
+	}
+	if (vol->mftmirr_ino) {
+		/* Re-commit the mft mirror and mft just in case. */
+		ntfs_commit_inode(vol->mftmirr_ino);
+		ntfs_commit_inode(vol->mft_ino);
+		iput(vol->mftmirr_ino);
+		vol->mftmirr_ino = NULL;
+	}
+	/*
+	 * We should have no dirty inodes left, due to
+	 * mft.c::ntfs_mft_writepage() cleaning all the dirty pages as
+	 * the underlying mft records are written out and cleaned.
+	 */
+	ntfs_commit_inode(vol->mft_ino);
+	write_inode_now(vol->mft_ino, 1);
+#endif /* NTFS_RW */
+
+	iput(vol->mft_ino);
+	vol->mft_ino = NULL;
+
+	/* Throw away the table of attribute definitions. */
+	vol->attrdef_size = 0;
+	if (vol->attrdef) {
+		ntfs_free(vol->attrdef);
+		vol->attrdef = NULL;
+	}
+	vol->upcase_len = 0;
+	/*
+	 * Destroy the global default upcase table if necessary.  Also decrease
+	 * the number of upcase users if we are a user.
+	 */
+	mutex_lock(&ntfs_lock);
+	if (vol->upcase == default_upcase) {
+		ntfs_nr_upcase_users--;
+		vol->upcase = NULL;
+	}
+	if (!ntfs_nr_upcase_users && default_upcase) {
+		ntfs_free(default_upcase);
+		default_upcase = NULL;
+	}
+	if (vol->cluster_size <= 4096 && !--ntfs_nr_compression_users)
+		free_compression_buffers();
+	mutex_unlock(&ntfs_lock);
+	if (vol->upcase) {
+		ntfs_free(vol->upcase);
+		vol->upcase = NULL;
+	}
+
+	unload_nls(vol->nls_map);
+
+	sb->s_fs_info = NULL;
+	kfree(vol);
+}
+
+/**
+ * get_nr_free_clusters - return the number of free clusters on a volume
+ * @vol:	ntfs volume for which to obtain free cluster count
+ *
+ * Calculate the number of free clusters on the mounted NTFS volume @vol. We
+ * actually calculate the number of clusters in use instead because this
+ * allows us to not care about partial pages as these will be just zero filled
+ * and hence not be counted as allocated clusters.
+ *
+ * The only particularity is that clusters beyond the end of the logical ntfs
+ * volume will be marked as allocated to prevent errors which means we have to
+ * discount those at the end. This is important as the cluster bitmap always
+ * has a size in multiples of 8 bytes, i.e. up to 63 clusters could be outside
+ * the logical volume and marked in use when they are not as they do not exist.
+ *
+ * If any pages cannot be read we assume all clusters in the erroring pages are
+ * in use. This means we return an underestimate on errors which is better than
+ * an overestimate.
+ */
+static s64 get_nr_free_clusters(ntfs_volume *vol)
+{
+	s64 nr_free = vol->nr_clusters;
+	struct address_space *mapping = vol->lcnbmp_ino->i_mapping;
+	struct page *page;
+	pgoff_t index, max_index;
+
+	ntfs_debug("Entering.");
+	/* Serialize accesses to the cluster bitmap. */
+	down_read(&vol->lcnbmp_lock);
+	/*
+	 * Convert the number of bits into bytes rounded up, then convert into
+	 * multiples of PAGE_SIZE, rounding up so that if we have one
+	 * full and one partial page max_index = 2.
+	 */
+	max_index = (((vol->nr_clusters + 7) >> 3) + PAGE_SIZE - 1) >>
+			PAGE_SHIFT;
+	/* Use multiples of 4 bytes, thus max_size is PAGE_SIZE / 4. */
+	ntfs_debug("Reading $Bitmap, max_index = 0x%lx, max_size = 0x%lx.",
+			max_index, PAGE_SIZE / 4);
+	for (index = 0; index < max_index; index++) {
+		unsigned long *kaddr;
+
+		/*
+		 * Read the page from page cache, getting it from backing store
+		 * if necessary, and increment the use count.
+		 */
+		page = read_mapping_page(mapping, index, NULL);
+		/* Ignore pages which errored synchronously. */
+		if (IS_ERR(page)) {
+			ntfs_debug("read_mapping_page() error. Skipping "
+					"page (index 0x%lx).", index);
+			nr_free -= PAGE_SIZE * 8;
+			continue;
+		}
+		kaddr = kmap_atomic(page);
+		/*
+		 * Subtract the number of set bits. If this
+		 * is the last page and it is partial we don't really care as
+		 * it just means we do a little extra work but it won't affect
+		 * the result as all out of range bytes are set to zero by
+		 * ntfs_readpage().
+		 */
+		nr_free -= bitmap_weight(kaddr,
+					PAGE_SIZE * BITS_PER_BYTE);
+		kunmap_atomic(kaddr);
+		put_page(page);
+	}
+	ntfs_debug("Finished reading $Bitmap, last index = 0x%lx.", index - 1);
+	/*
+	 * Fixup for eventual bits outside logical ntfs volume (see function
+	 * description above).
+	 */
+	if (vol->nr_clusters & 63)
+		nr_free += 64 - (vol->nr_clusters & 63);
+	up_read(&vol->lcnbmp_lock);
+	/* If errors occurred we may well have gone below zero, fix this. */
+	if (nr_free < 0)
+		nr_free = 0;
+	ntfs_debug("Exiting.");
+	return nr_free;
+}
+
+/**
+ * __get_nr_free_mft_records - return the number of free inodes on a volume
+ * @vol:	ntfs volume for which to obtain free inode count
+ * @nr_free:	number of mft records in filesystem
+ * @max_index:	maximum number of pages containing set bits
+ *
+ * Calculate the number of free mft records (inodes) on the mounted NTFS
+ * volume @vol. We actually calculate the number of mft records in use instead
+ * because this allows us to not care about partial pages as these will be just
+ * zero filled and hence not be counted as allocated mft record.
+ *
+ * If any pages cannot be read we assume all mft records in the erroring pages
+ * are in use. This means we return an underestimate on errors which is better
+ * than an overestimate.
+ *
+ * NOTE: Caller must hold mftbmp_lock rw_semaphore for reading or writing.
+ */
+static unsigned long __get_nr_free_mft_records(ntfs_volume *vol,
+		s64 nr_free, const pgoff_t max_index)
+{
+	struct address_space *mapping = vol->mftbmp_ino->i_mapping;
+	struct page *page;
+	pgoff_t index;
+
+	ntfs_debug("Entering.");
+	/* Use multiples of 4 bytes, thus max_size is PAGE_SIZE / 4. */
+	ntfs_debug("Reading $MFT/$BITMAP, max_index = 0x%lx, max_size = "
+			"0x%lx.", max_index, PAGE_SIZE / 4);
+	for (index = 0; index < max_index; index++) {
+		unsigned long *kaddr;
+
+		/*
+		 * Read the page from page cache, getting it from backing store
+		 * if necessary, and increment the use count.
+		 */
+		page = read_mapping_page(mapping, index, NULL);
+		/* Ignore pages which errored synchronously. */
+		if (IS_ERR(page)) {
+			ntfs_debug("read_mapping_page() error. Skipping "
+					"page (index 0x%lx).", index);
+			nr_free -= PAGE_SIZE * 8;
+			continue;
+		}
+		kaddr = kmap_atomic(page);
+		/*
+		 * Subtract the number of set bits. If this
+		 * is the last page and it is partial we don't really care as
+		 * it just means we do a little extra work but it won't affect
+		 * the result as all out of range bytes are set to zero by
+		 * ntfs_readpage().
+		 */
+		nr_free -= bitmap_weight(kaddr,
+					PAGE_SIZE * BITS_PER_BYTE);
+		kunmap_atomic(kaddr);
+		put_page(page);
+	}
+	ntfs_debug("Finished reading $MFT/$BITMAP, last index = 0x%lx.",
+			index - 1);
+	/* If errors occurred we may well have gone below zero, fix this. */
+	if (nr_free < 0)
+		nr_free = 0;
+	ntfs_debug("Exiting.");
+	return nr_free;
+}
+
+/**
+ * ntfs_statfs - return information about mounted NTFS volume
+ * @dentry:	dentry from mounted volume
+ * @sfs:	statfs structure in which to return the information
+ *
+ * Return information about the mounted NTFS volume @dentry in the statfs structure
+ * pointed to by @sfs (this is initialized with zeros before ntfs_statfs is
+ * called). We interpret the values to be correct of the moment in time at
+ * which we are called. Most values are variable otherwise and this isn't just
+ * the free values but the totals as well. For example we can increase the
+ * total number of file nodes if we run out and we can keep doing this until
+ * there is no more space on the volume left at all.
+ *
+ * Called from vfs_statfs which is used to handle the statfs, fstatfs, and
+ * ustat system calls.
+ *
+ * Return 0 on success or -errno on error.
+ */
+static int ntfs_statfs(struct dentry *dentry, struct kstatfs *sfs)
+{
+	struct super_block *sb = dentry->d_sb;
+	s64 size;
+	ntfs_volume *vol = NTFS_SB(sb);
+	ntfs_inode *mft_ni = NTFS_I(vol->mft_ino);
+	pgoff_t max_index;
+	unsigned long flags;
+
+	ntfs_debug("Entering.");
+	/* Type of filesystem. */
+	sfs->f_type   = NTFS_SB_MAGIC;
+	/* Optimal transfer block size. */
+	sfs->f_bsize  = PAGE_SIZE;
+	/*
+	 * Total data blocks in filesystem in units of f_bsize and since
+	 * inodes are also stored in data blocs ($MFT is a file) this is just
+	 * the total clusters.
+	 */
+	sfs->f_blocks = vol->nr_clusters << vol->cluster_size_bits >>
+				PAGE_SHIFT;
+	/* Free data blocks in filesystem in units of f_bsize. */
+	size	      = get_nr_free_clusters(vol) << vol->cluster_size_bits >>
+				PAGE_SHIFT;
+	if (size < 0LL)
+		size = 0LL;
+	/* Free blocks avail to non-superuser, same as above on NTFS. */
+	sfs->f_bavail = sfs->f_bfree = size;
+	/* Serialize accesses to the inode bitmap. */
+	down_read(&vol->mftbmp_lock);
+	read_lock_irqsave(&mft_ni->size_lock, flags);
+	size = i_size_read(vol->mft_ino) >> vol->mft_record_size_bits;
+	/*
+	 * Convert the maximum number of set bits into bytes rounded up, then
+	 * convert into multiples of PAGE_SIZE, rounding up so that if we
+	 * have one full and one partial page max_index = 2.
+	 */
+	max_index = ((((mft_ni->initialized_size >> vol->mft_record_size_bits)
+			+ 7) >> 3) + PAGE_SIZE - 1) >> PAGE_SHIFT;
+	read_unlock_irqrestore(&mft_ni->size_lock, flags);
+	/* Number of inodes in filesystem (at this point in time). */
+	sfs->f_files = size;
+	/* Free inodes in fs (based on current total count). */
+	sfs->f_ffree = __get_nr_free_mft_records(vol, size, max_index);
+	up_read(&vol->mftbmp_lock);
+	/*
+	 * File system id. This is extremely *nix flavour dependent and even
+	 * within Linux itself all fs do their own thing. I interpret this to
+	 * mean a unique id associated with the mounted fs and not the id
+	 * associated with the filesystem driver, the latter is already given
+	 * by the filesystem type in sfs->f_type. Thus we use the 64-bit
+	 * volume serial number splitting it into two 32-bit parts. We enter
+	 * the least significant 32-bits in f_fsid[0] and the most significant
+	 * 32-bits in f_fsid[1].
+	 */
+	sfs->f_fsid = u64_to_fsid(vol->serial_no);
+	/* Maximum length of filenames. */
+	sfs->f_namelen	   = NTFS_MAX_NAME_LEN;
+	return 0;
+}
+
+#ifdef NTFS_RW
+static int ntfs_write_inode(struct inode *vi, struct writeback_control *wbc)
+{
+	return __ntfs_write_inode(vi, wbc->sync_mode == WB_SYNC_ALL);
+}
+#endif
+
+/**
+ * The complete super operations.
+ */
+static const struct super_operations ntfs_sops = {
+	.alloc_inode	= ntfs_alloc_big_inode,	  /* VFS: Allocate new inode. */
+	.free_inode	= ntfs_free_big_inode, /* VFS: Deallocate inode. */
+#ifdef NTFS_RW
+	.write_inode	= ntfs_write_inode,	/* VFS: Write dirty inode to
+						   disk. */
+#endif /* NTFS_RW */
+	.put_super	= ntfs_put_super,	/* Syscall: umount. */
+	.statfs		= ntfs_statfs,		/* Syscall: statfs */
+	.remount_fs	= ntfs_remount,		/* Syscall: mount -o remount. */
+	.evict_inode	= ntfs_evict_big_inode,	/* VFS: Called when an inode is
+						   removed from memory. */
+	.show_options	= ntfs_show_options,	/* Show mount options in
+						   proc. */
+};
+
+/**
+ * ntfs_fill_super - mount an ntfs filesystem
+ * @sb:		super block of ntfs filesystem to mount
+ * @opt:	string containing the mount options
+ * @silent:	silence error output
+ *
+ * ntfs_fill_super() is called by the VFS to mount the device described by @sb
+ * with the mount otions in @data with the NTFS filesystem.
+ *
+ * If @silent is true, remain silent even if errors are detected. This is used
+ * during bootup, when the kernel tries to mount the root filesystem with all
+ * registered filesystems one after the other until one succeeds. This implies
+ * that all filesystems except the correct one will quite correctly and
+ * expectedly return an error, but nobody wants to see error messages when in
+ * fact this is what is supposed to happen.
+ *
+ * NOTE: @sb->s_flags contains the mount options flags.
+ */
+static int ntfs_fill_super(struct super_block *sb, void *opt, const int silent)
+{
+	ntfs_volume *vol;
+	struct buffer_head *bh;
+	struct inode *tmp_ino;
+	int blocksize, result;
+
+	/*
+	 * We do a pretty difficult piece of bootstrap by reading the
+	 * MFT (and other metadata) from disk into memory. We'll only
+	 * release this metadata during umount, so the locking patterns
+	 * observed during bootstrap do not count. So turn off the
+	 * observation of locking patterns (strictly for this context
+	 * only) while mounting NTFS. [The validator is still active
+	 * otherwise, even for this context: it will for example record
+	 * lock class registrations.]
+	 */
+	lockdep_off();
+	ntfs_debug("Entering.");
+#ifndef NTFS_RW
+	sb->s_flags |= SB_RDONLY;
+#endif /* ! NTFS_RW */
+	/* Allocate a new ntfs_volume and place it in sb->s_fs_info. */
+	sb->s_fs_info = kmalloc(sizeof(ntfs_volume), GFP_NOFS);
+	vol = NTFS_SB(sb);
+	if (!vol) {
+		if (!silent)
+			ntfs_error(sb, "Allocation of NTFS volume structure "
+					"failed. Aborting mount...");
+		lockdep_on();
+		return -ENOMEM;
+	}
+	/* Initialize ntfs_volume structure. */
+	*vol = (ntfs_volume) {
+		.sb = sb,
+		/*
+		 * Default is group and other don't have any access to files or
+		 * directories while owner has full access. Further, files by
+		 * default are not executable but directories are of course
+		 * browseable.
+		 */
+		.fmask = 0177,
+		.dmask = 0077,
+	};
+	init_rwsem(&vol->mftbmp_lock);
+	init_rwsem(&vol->lcnbmp_lock);
+
+	/* By default, enable sparse support. */
+	NVolSetSparseEnabled(vol);
+
+	/* Important to get the mount options dealt with now. */
+	if (!parse_options(vol, (char*)opt))
+		goto err_out_now;
+
+	/* We support sector sizes up to the PAGE_SIZE. */
+	if (bdev_logical_block_size(sb->s_bdev) > PAGE_SIZE) {
+		if (!silent)
+			ntfs_error(sb, "Device has unsupported sector size "
+					"(%i).  The maximum supported sector "
+					"size on this architecture is %lu "
+					"bytes.",
+					bdev_logical_block_size(sb->s_bdev),
+					PAGE_SIZE);
+		goto err_out_now;
+	}
+	/*
+	 * Setup the device access block size to NTFS_BLOCK_SIZE or the hard
+	 * sector size, whichever is bigger.
+	 */
+	blocksize = sb_min_blocksize(sb, NTFS_BLOCK_SIZE);
+	if (blocksize < NTFS_BLOCK_SIZE) {
+		if (!silent)
+			ntfs_error(sb, "Unable to set device block size.");
+		goto err_out_now;
+	}
+	BUG_ON(blocksize != sb->s_blocksize);
+	ntfs_debug("Set device block size to %i bytes (block size bits %i).",
+			blocksize, sb->s_blocksize_bits);
+	/* Determine the size of the device in units of block_size bytes. */
+	if (!i_size_read(sb->s_bdev->bd_inode)) {
+		if (!silent)
+			ntfs_error(sb, "Unable to determine device size.");
+		goto err_out_now;
+	}
+	vol->nr_blocks = i_size_read(sb->s_bdev->bd_inode) >>
+			sb->s_blocksize_bits;
+	/* Read the boot sector and return unlocked buffer head to it. */
+	if (!(bh = read_ntfs_boot_sector(sb, silent))) {
+		if (!silent)
+			ntfs_error(sb, "Not an NTFS volume.");
+		goto err_out_now;
+	}
+	/*
+	 * Extract the data from the boot sector and setup the ntfs volume
+	 * using it.
+	 */
+	result = parse_ntfs_boot_sector(vol, (NTFS_BOOT_SECTOR*)bh->b_data);
+	brelse(bh);
+	if (!result) {
+		if (!silent)
+			ntfs_error(sb, "Unsupported NTFS filesystem.");
+		goto err_out_now;
+	}
+	/*
+	 * If the boot sector indicates a sector size bigger than the current
+	 * device block size, switch the device block size to the sector size.
+	 * TODO: It may be possible to support this case even when the set
+	 * below fails, we would just be breaking up the i/o for each sector
+	 * into multiple blocks for i/o purposes but otherwise it should just
+	 * work.  However it is safer to leave disabled until someone hits this
+	 * error message and then we can get them to try it without the setting
+	 * so we know for sure that it works.
+	 */
+	if (vol->sector_size > blocksize) {
+		blocksize = sb_set_blocksize(sb, vol->sector_size);
+		if (blocksize != vol->sector_size) {
+			if (!silent)
+				ntfs_error(sb, "Unable to set device block "
+						"size to sector size (%i).",
+						vol->sector_size);
+			goto err_out_now;
+		}
+		BUG_ON(blocksize != sb->s_blocksize);
+		vol->nr_blocks = i_size_read(sb->s_bdev->bd_inode) >>
+				sb->s_blocksize_bits;
+		ntfs_debug("Changed device block size to %i bytes (block size "
+				"bits %i) to match volume sector size.",
+				blocksize, sb->s_blocksize_bits);
+	}
+	/* Initialize the cluster and mft allocators. */
+	ntfs_setup_allocators(vol);
+	/* Setup remaining fields in the super block. */
+	sb->s_magic = NTFS_SB_MAGIC;
+	/*
+	 * Ntfs allows 63 bits for the file size, i.e. correct would be:
+	 *	sb->s_maxbytes = ~0ULL >> 1;
+	 * But the kernel uses a long as the page cache page index which on
+	 * 32-bit architectures is only 32-bits. MAX_LFS_FILESIZE is kernel
+	 * defined to the maximum the page cache page index can cope with
+	 * without overflowing the index or to 2^63 - 1, whichever is smaller.
+	 */
+	sb->s_maxbytes = MAX_LFS_FILESIZE;
+	/* Ntfs measures time in 100ns intervals. */
+	sb->s_time_gran = 100;
+	/*
+	 * Now load the metadata required for the page cache and our address
+	 * space operations to function. We do this by setting up a specialised
+	 * read_inode method and then just calling the normal iget() to obtain
+	 * the inode for $MFT which is sufficient to allow our normal inode
+	 * operations and associated address space operations to function.
+	 */
+	sb->s_op = &ntfs_sops;
+	tmp_ino = new_inode(sb);
+	if (!tmp_ino) {
+		if (!silent)
+			ntfs_error(sb, "Failed to load essential metadata.");
+		goto err_out_now;
+	}
+	tmp_ino->i_ino = FILE_MFT;
+	insert_inode_hash(tmp_ino);
+	if (ntfs_read_inode_mount(tmp_ino) < 0) {
+		if (!silent)
+			ntfs_error(sb, "Failed to load essential metadata.");
+		goto iput_tmp_ino_err_out_now;
+	}
+	mutex_lock(&ntfs_lock);
+	/*
+	 * The current mount is a compression user if the cluster size is
+	 * less than or equal 4kiB.
+	 */
+	if (vol->cluster_size <= 4096 && !ntfs_nr_compression_users++) {
+		result = allocate_compression_buffers();
+		if (result) {
+			ntfs_error(NULL, "Failed to allocate buffers "
+					"for compression engine.");
+			ntfs_nr_compression_users--;
+			mutex_unlock(&ntfs_lock);
+			goto iput_tmp_ino_err_out_now;
+		}
+	}
+	/*
+	 * Generate the global default upcase table if necessary.  Also
+	 * temporarily increment the number of upcase users to avoid race
+	 * conditions with concurrent (u)mounts.
+	 */
+	if (!default_upcase)
+		default_upcase = generate_default_upcase();
+	ntfs_nr_upcase_users++;
+	mutex_unlock(&ntfs_lock);
+	/*
+	 * From now on, ignore @silent parameter. If we fail below this line,
+	 * it will be due to a corrupt fs or a system error, so we report it.
+	 */
+	/*
+	 * Open the system files with normal access functions and complete
+	 * setting up the ntfs super block.
+	 */
+	if (!load_system_files(vol)) {
+		ntfs_error(sb, "Failed to load system files.");
+		goto unl_upcase_iput_tmp_ino_err_out_now;
+	}
+
+	/* We grab a reference, simulating an ntfs_iget(). */
+	ihold(vol->root_ino);
+	if ((sb->s_root = d_make_root(vol->root_ino))) {
+		ntfs_debug("Exiting, status successful.");
+		/* Release the default upcase if it has no users. */
+		mutex_lock(&ntfs_lock);
+		if (!--ntfs_nr_upcase_users && default_upcase) {
+			ntfs_free(default_upcase);
+			default_upcase = NULL;
+		}
+		mutex_unlock(&ntfs_lock);
+		sb->s_export_op = &ntfs_export_ops;
+		lockdep_on();
+		return 0;
+	}
+	ntfs_error(sb, "Failed to allocate root directory.");
+	/* Clean up after the successful load_system_files() call from above. */
+	// TODO: Use ntfs_put_super() instead of repeating all this code...
+	// FIXME: Should mark the volume clean as the error is most likely
+	// 	  -ENOMEM.
+	iput(vol->vol_ino);
+	vol->vol_ino = NULL;
+	/* NTFS 3.0+ specific clean up. */
+	if (vol->major_ver >= 3) {
+#ifdef NTFS_RW
+		if (vol->usnjrnl_j_ino) {
+			iput(vol->usnjrnl_j_ino);
+			vol->usnjrnl_j_ino = NULL;
+		}
+		if (vol->usnjrnl_max_ino) {
+			iput(vol->usnjrnl_max_ino);
+			vol->usnjrnl_max_ino = NULL;
+		}
+		if (vol->usnjrnl_ino) {
+			iput(vol->usnjrnl_ino);
+			vol->usnjrnl_ino = NULL;
+		}
+		if (vol->quota_q_ino) {
+			iput(vol->quota_q_ino);
+			vol->quota_q_ino = NULL;
+		}
+		if (vol->quota_ino) {
+			iput(vol->quota_ino);
+			vol->quota_ino = NULL;
+		}
+#endif /* NTFS_RW */
+		if (vol->extend_ino) {
+			iput(vol->extend_ino);
+			vol->extend_ino = NULL;
+		}
+		if (vol->secure_ino) {
+			iput(vol->secure_ino);
+			vol->secure_ino = NULL;
+		}
+	}
+	iput(vol->root_ino);
+	vol->root_ino = NULL;
+	iput(vol->lcnbmp_ino);
+	vol->lcnbmp_ino = NULL;
+	iput(vol->mftbmp_ino);
+	vol->mftbmp_ino = NULL;
+#ifdef NTFS_RW
+	if (vol->logfile_ino) {
+		iput(vol->logfile_ino);
+		vol->logfile_ino = NULL;
+	}
+	if (vol->mftmirr_ino) {
+		iput(vol->mftmirr_ino);
+		vol->mftmirr_ino = NULL;
+	}
+#endif /* NTFS_RW */
+	/* Throw away the table of attribute definitions. */
+	vol->attrdef_size = 0;
+	if (vol->attrdef) {
+		ntfs_free(vol->attrdef);
+		vol->attrdef = NULL;
+	}
+	vol->upcase_len = 0;
+	mutex_lock(&ntfs_lock);
+	if (vol->upcase == default_upcase) {
+		ntfs_nr_upcase_users--;
+		vol->upcase = NULL;
+	}
+	mutex_unlock(&ntfs_lock);
+	if (vol->upcase) {
+		ntfs_free(vol->upcase);
+		vol->upcase = NULL;
+	}
+	if (vol->nls_map) {
+		unload_nls(vol->nls_map);
+		vol->nls_map = NULL;
+	}
+	/* Error exit code path. */
+unl_upcase_iput_tmp_ino_err_out_now:
+	/*
+	 * Decrease the number of upcase users and destroy the global default
+	 * upcase table if necessary.
+	 */
+	mutex_lock(&ntfs_lock);
+	if (!--ntfs_nr_upcase_users && default_upcase) {
+		ntfs_free(default_upcase);
+		default_upcase = NULL;
+	}
+	if (vol->cluster_size <= 4096 && !--ntfs_nr_compression_users)
+		free_compression_buffers();
+	mutex_unlock(&ntfs_lock);
+iput_tmp_ino_err_out_now:
+	iput(tmp_ino);
+	if (vol->mft_ino && vol->mft_ino != tmp_ino)
+		iput(vol->mft_ino);
+	vol->mft_ino = NULL;
+	/* Errors at this stage are irrelevant. */
+err_out_now:
+	sb->s_fs_info = NULL;
+	kfree(vol);
+	ntfs_debug("Failed, returning -EINVAL.");
+	lockdep_on();
+	return -EINVAL;
+}
+
+/*
+ * This is a slab cache to optimize allocations and deallocations of Unicode
+ * strings of the maximum length allowed by NTFS, which is NTFS_MAX_NAME_LEN
+ * (255) Unicode characters + a terminating NULL Unicode character.
+ */
+struct kmem_cache *ntfs_name_cache;
+
+/* Slab caches for efficient allocation/deallocation of inodes. */
+struct kmem_cache *ntfs_inode_cache;
+struct kmem_cache *ntfs_big_inode_cache;
+
+/* Init once constructor for the inode slab cache. */
+static void ntfs_big_inode_init_once(void *foo)
+{
+	ntfs_inode *ni = (ntfs_inode *)foo;
+
+	inode_init_once(VFS_I(ni));
+}
+
+/*
+ * Slab caches to optimize allocations and deallocations of attribute search
+ * contexts and index contexts, respectively.
+ */
+struct kmem_cache *ntfs_attr_ctx_cache;
+struct kmem_cache *ntfs_index_ctx_cache;
+
+/* Driver wide mutex. */
+DEFINE_MUTEX(ntfs_lock);
+
+static struct dentry *ntfs_mount(struct file_system_type *fs_type,
+	int flags, const char *dev_name, void *data)
+{
+	return mount_bdev(fs_type, flags, dev_name, data, ntfs_fill_super);
+}
+
+static struct file_system_type ntfs_fs_type = {
+	.owner		= THIS_MODULE,
+	.name		= "ntfs",
+	.mount		= ntfs_mount,
+	.kill_sb	= kill_block_super,
+	.fs_flags	= FS_REQUIRES_DEV,
+};
+MODULE_ALIAS_FS("ntfs");
+
+/* Stable names for the slab caches. */
+static const char ntfs_index_ctx_cache_name[] = "ntfs_index_ctx_cache";
+static const char ntfs_attr_ctx_cache_name[] = "ntfs_attr_ctx_cache";
+static const char ntfs_name_cache_name[] = "ntfs_name_cache";
+static const char ntfs_inode_cache_name[] = "ntfs_inode_cache";
+static const char ntfs_big_inode_cache_name[] = "ntfs_big_inode_cache";
+
+static int __init init_ntfs_fs(void)
+{
+	int err = 0;
+
+	/* This may be ugly but it results in pretty output so who cares. (-8 */
+	pr_info("driver " NTFS_VERSION " [Flags: R/"
+#ifdef NTFS_RW
+			"W"
+#else
+			"O"
+#endif
+#ifdef DEBUG
+			" DEBUG"
+#endif
+#ifdef MODULE
+			" MODULE"
+#endif
+			"].\n");
+
+	ntfs_debug("Debug messages are enabled.");
+
+	ntfs_index_ctx_cache = kmem_cache_create(ntfs_index_ctx_cache_name,
+			sizeof(ntfs_index_context), 0 /* offset */,
+			SLAB_HWCACHE_ALIGN, NULL /* ctor */);
+	if (!ntfs_index_ctx_cache) {
+		pr_crit("Failed to create %s!\n", ntfs_index_ctx_cache_name);
+		goto ictx_err_out;
+	}
+	ntfs_attr_ctx_cache = kmem_cache_create(ntfs_attr_ctx_cache_name,
+			sizeof(ntfs_attr_search_ctx), 0 /* offset */,
+			SLAB_HWCACHE_ALIGN, NULL /* ctor */);
+	if (!ntfs_attr_ctx_cache) {
+		pr_crit("NTFS: Failed to create %s!\n",
+			ntfs_attr_ctx_cache_name);
+		goto actx_err_out;
+	}
+
+	ntfs_name_cache = kmem_cache_create(ntfs_name_cache_name,
+			(NTFS_MAX_NAME_LEN+1) * sizeof(ntfschar), 0,
+			SLAB_HWCACHE_ALIGN, NULL);
+	if (!ntfs_name_cache) {
+		pr_crit("Failed to create %s!\n", ntfs_name_cache_name);
+		goto name_err_out;
+	}
+
+	ntfs_inode_cache = kmem_cache_create(ntfs_inode_cache_name,
+			sizeof(ntfs_inode), 0,
+			SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD, NULL);
+	if (!ntfs_inode_cache) {
+		pr_crit("Failed to create %s!\n", ntfs_inode_cache_name);
+		goto inode_err_out;
+	}
+
+	ntfs_big_inode_cache = kmem_cache_create(ntfs_big_inode_cache_name,
+			sizeof(big_ntfs_inode), 0,
+			SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
+			SLAB_ACCOUNT, ntfs_big_inode_init_once);
+	if (!ntfs_big_inode_cache) {
+		pr_crit("Failed to create %s!\n", ntfs_big_inode_cache_name);
+		goto big_inode_err_out;
+	}
+
+	/* Register the ntfs sysctls. */
+	err = ntfs_sysctl(1);
+	if (err) {
+		pr_crit("Failed to register NTFS sysctls!\n");
+		goto sysctl_err_out;
+	}
+
+	err = register_filesystem(&ntfs_fs_type);
+	if (!err) {
+		ntfs_debug("NTFS driver registered successfully.");
+		return 0; /* Success! */
+	}
+	pr_crit("Failed to register NTFS filesystem driver!\n");
+
+	/* Unregister the ntfs sysctls. */
+	ntfs_sysctl(0);
+sysctl_err_out:
+	kmem_cache_destroy(ntfs_big_inode_cache);
+big_inode_err_out:
+	kmem_cache_destroy(ntfs_inode_cache);
+inode_err_out:
+	kmem_cache_destroy(ntfs_name_cache);
+name_err_out:
+	kmem_cache_destroy(ntfs_attr_ctx_cache);
+actx_err_out:
+	kmem_cache_destroy(ntfs_index_ctx_cache);
+ictx_err_out:
+	if (!err) {
+		pr_crit("Aborting NTFS filesystem driver registration...\n");
+		err = -ENOMEM;
+	}
+	return err;
+}
+
+static void __exit exit_ntfs_fs(void)
+{
+	ntfs_debug("Unregistering NTFS driver.");
+
+	unregister_filesystem(&ntfs_fs_type);
+
+	/*
+	 * Make sure all delayed rcu free inodes are flushed before we
+	 * destroy cache.
+	 */
+	rcu_barrier();
+	kmem_cache_destroy(ntfs_big_inode_cache);
+	kmem_cache_destroy(ntfs_inode_cache);
+	kmem_cache_destroy(ntfs_name_cache);
+	kmem_cache_destroy(ntfs_attr_ctx_cache);
+	kmem_cache_destroy(ntfs_index_ctx_cache);
+	/* Unregister the ntfs sysctls. */
+	ntfs_sysctl(0);
+}
+
+MODULE_AUTHOR("Anton Altaparmakov <anton@tuxera.com>");
+MODULE_DESCRIPTION("NTFS 1.2/3.x driver - Copyright (c) 2001-2014 Anton Altaparmakov and Tuxera Inc.");
+MODULE_VERSION(NTFS_VERSION);
+MODULE_LICENSE("GPL");
+#ifdef DEBUG
+module_param(debug_msgs, bint, 0);
+MODULE_PARM_DESC(debug_msgs, "Enable debug messages.");
+#endif
+
+module_init(init_ntfs_fs)
+module_exit(exit_ntfs_fs)
diff --git a/fs/ntfs/sysctl.c b/fs/ntfs/sysctl.c
new file mode 100644
index 000000000..a030d00af
--- /dev/null
+++ b/fs/ntfs/sysctl.c
@@ -0,0 +1,69 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * sysctl.c - Code for sysctl handling in NTFS Linux kernel driver. Part of
+ *	      the Linux-NTFS project. Adapted from the old NTFS driver,
+ *	      Copyright (C) 1997 Martin von Löwis, Régis Duchesne
+ *
+ * Copyright (c) 2002-2005 Anton Altaparmakov
+ */
+
+#ifdef DEBUG
+
+#include <linux/module.h>
+
+#ifdef CONFIG_SYSCTL
+
+#include <linux/proc_fs.h>
+#include <linux/sysctl.h>
+
+#include "sysctl.h"
+#include "debug.h"
+
+/* Definition of the ntfs sysctl. */
+static struct ctl_table ntfs_sysctls[] = {
+	{
+		.procname	= "ntfs-debug",
+		.data		= &debug_msgs,		/* Data pointer and size. */
+		.maxlen		= sizeof(debug_msgs),
+		.mode		= 0644,			/* Mode, proc handler. */
+		.proc_handler	= proc_dointvec
+	},
+	{}
+};
+
+/* Define the parent directory /proc/sys/fs. */
+static struct ctl_table sysctls_root[] = {
+	{
+		.procname	= "fs",
+		.mode		= 0555,
+		.child		= ntfs_sysctls
+	},
+	{}
+};
+
+/* Storage for the sysctls header. */
+static struct ctl_table_header *sysctls_root_table;
+
+/**
+ * ntfs_sysctl - add or remove the debug sysctl
+ * @add:	add (1) or remove (0) the sysctl
+ *
+ * Add or remove the debug sysctl. Return 0 on success or -errno on error.
+ */
+int ntfs_sysctl(int add)
+{
+	if (add) {
+		BUG_ON(sysctls_root_table);
+		sysctls_root_table = register_sysctl_table(sysctls_root);
+		if (!sysctls_root_table)
+			return -ENOMEM;
+	} else {
+		BUG_ON(!sysctls_root_table);
+		unregister_sysctl_table(sysctls_root_table);
+		sysctls_root_table = NULL;
+	}
+	return 0;
+}
+
+#endif /* CONFIG_SYSCTL */
+#endif /* DEBUG */
diff --git a/fs/ntfs/sysctl.h b/fs/ntfs/sysctl.h
new file mode 100644
index 000000000..96bb2299d
--- /dev/null
+++ b/fs/ntfs/sysctl.h
@@ -0,0 +1,27 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * sysctl.h - Defines for sysctl handling in NTFS Linux kernel driver. Part of
+ *	      the Linux-NTFS project. Adapted from the old NTFS driver,
+ *	      Copyright (C) 1997 Martin von Löwis, Régis Duchesne
+ *
+ * Copyright (c) 2002-2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_SYSCTL_H
+#define _LINUX_NTFS_SYSCTL_H
+
+
+#if defined(DEBUG) && defined(CONFIG_SYSCTL)
+
+extern int ntfs_sysctl(int add);
+
+#else
+
+/* Just return success. */
+static inline int ntfs_sysctl(int add)
+{
+	return 0;
+}
+
+#endif /* DEBUG && CONFIG_SYSCTL */
+#endif /* _LINUX_NTFS_SYSCTL_H */
diff --git a/fs/ntfs/time.h b/fs/ntfs/time.h
new file mode 100644
index 000000000..6b6326130
--- /dev/null
+++ b/fs/ntfs/time.h
@@ -0,0 +1,89 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * time.h - NTFS time conversion functions.  Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2005 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_TIME_H
+#define _LINUX_NTFS_TIME_H
+
+#include <linux/time.h>		/* For current_kernel_time(). */
+#include <asm/div64.h>		/* For do_div(). */
+
+#include "endian.h"
+
+#define NTFS_TIME_OFFSET ((s64)(369 * 365 + 89) * 24 * 3600 * 10000000)
+
+/**
+ * utc2ntfs - convert Linux UTC time to NTFS time
+ * @ts:		Linux UTC time to convert to NTFS time
+ *
+ * Convert the Linux UTC time @ts to its corresponding NTFS time and return
+ * that in little endian format.
+ *
+ * Linux stores time in a struct timespec64 consisting of a time64_t tv_sec
+ * and a long tv_nsec where tv_sec is the number of 1-second intervals since
+ * 1st January 1970, 00:00:00 UTC and tv_nsec is the number of 1-nano-second
+ * intervals since the value of tv_sec.
+ *
+ * NTFS uses Microsoft's standard time format which is stored in a s64 and is
+ * measured as the number of 100-nano-second intervals since 1st January 1601,
+ * 00:00:00 UTC.
+ */
+static inline sle64 utc2ntfs(const struct timespec64 ts)
+{
+	/*
+	 * Convert the seconds to 100ns intervals, add the nano-seconds
+	 * converted to 100ns intervals, and then add the NTFS time offset.
+	 */
+	return cpu_to_sle64((s64)ts.tv_sec * 10000000 + ts.tv_nsec / 100 +
+			NTFS_TIME_OFFSET);
+}
+
+/**
+ * get_current_ntfs_time - get the current time in little endian NTFS format
+ *
+ * Get the current time from the Linux kernel, convert it to its corresponding
+ * NTFS time and return that in little endian format.
+ */
+static inline sle64 get_current_ntfs_time(void)
+{
+	struct timespec64 ts;
+
+	ktime_get_coarse_real_ts64(&ts);
+	return utc2ntfs(ts);
+}
+
+/**
+ * ntfs2utc - convert NTFS time to Linux time
+ * @time:	NTFS time (little endian) to convert to Linux UTC
+ *
+ * Convert the little endian NTFS time @time to its corresponding Linux UTC
+ * time and return that in cpu format.
+ *
+ * Linux stores time in a struct timespec64 consisting of a time64_t tv_sec
+ * and a long tv_nsec where tv_sec is the number of 1-second intervals since
+ * 1st January 1970, 00:00:00 UTC and tv_nsec is the number of 1-nano-second
+ * intervals since the value of tv_sec.
+ *
+ * NTFS uses Microsoft's standard time format which is stored in a s64 and is
+ * measured as the number of 100 nano-second intervals since 1st January 1601,
+ * 00:00:00 UTC.
+ */
+static inline struct timespec64 ntfs2utc(const sle64 time)
+{
+	struct timespec64 ts;
+
+	/* Subtract the NTFS time offset. */
+	u64 t = (u64)(sle64_to_cpu(time) - NTFS_TIME_OFFSET);
+	/*
+	 * Convert the time to 1-second intervals and the remainder to
+	 * 1-nano-second intervals.
+	 */
+	ts.tv_nsec = do_div(t, 10000000) * 100;
+	ts.tv_sec = t;
+	return ts;
+}
+
+#endif /* _LINUX_NTFS_TIME_H */
diff --git a/fs/ntfs/types.h b/fs/ntfs/types.h
new file mode 100644
index 000000000..9a47859e7
--- /dev/null
+++ b/fs/ntfs/types.h
@@ -0,0 +1,55 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * types.h - Defines for NTFS Linux kernel driver specific types.
+ *	     Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2005 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_TYPES_H
+#define _LINUX_NTFS_TYPES_H
+
+#include <linux/types.h>
+
+typedef __le16 le16;
+typedef __le32 le32;
+typedef __le64 le64;
+typedef __u16 __bitwise sle16;
+typedef __u32 __bitwise sle32;
+typedef __u64 __bitwise sle64;
+
+/* 2-byte Unicode character type. */
+typedef le16 ntfschar;
+#define UCHAR_T_SIZE_BITS 1
+
+/*
+ * Clusters are signed 64-bit values on NTFS volumes. We define two types, LCN
+ * and VCN, to allow for type checking and better code readability.
+ */
+typedef s64 VCN;
+typedef sle64 leVCN;
+typedef s64 LCN;
+typedef sle64 leLCN;
+
+/*
+ * The NTFS journal $LogFile uses log sequence numbers which are signed 64-bit
+ * values.  We define our own type LSN, to allow for type checking and better
+ * code readability.
+ */
+typedef s64 LSN;
+typedef sle64 leLSN;
+
+/*
+ * The NTFS transaction log $UsnJrnl uses usn which are signed 64-bit values.
+ * We define our own type USN, to allow for type checking and better code
+ * readability.
+ */
+typedef s64 USN;
+typedef sle64 leUSN;
+
+typedef enum {
+	CASE_SENSITIVE = 0,
+	IGNORE_CASE = 1,
+} IGNORE_CASE_BOOL;
+
+#endif /* _LINUX_NTFS_TYPES_H */
diff --git a/fs/ntfs/unistr.c b/fs/ntfs/unistr.c
new file mode 100644
index 000000000..a6b6c64f1
--- /dev/null
+++ b/fs/ntfs/unistr.c
@@ -0,0 +1,384 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * unistr.c - NTFS Unicode string handling. Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2006 Anton Altaparmakov
+ */
+
+#include <linux/slab.h>
+
+#include "types.h"
+#include "debug.h"
+#include "ntfs.h"
+
+/*
+ * IMPORTANT
+ * =========
+ *
+ * All these routines assume that the Unicode characters are in little endian
+ * encoding inside the strings!!!
+ */
+
+/*
+ * This is used by the name collation functions to quickly determine what
+ * characters are (in)valid.
+ */
+static const u8 legal_ansi_char_array[0x40] = {
+	0x00, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10,
+	0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10,
+
+	0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10,
+	0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10,
+
+	0x17, 0x07, 0x18, 0x17, 0x17, 0x17, 0x17, 0x17,
+	0x17, 0x17, 0x18, 0x16, 0x16, 0x17, 0x07, 0x00,
+
+	0x17, 0x17, 0x17, 0x17, 0x17, 0x17, 0x17, 0x17,
+	0x17, 0x17, 0x04, 0x16, 0x18, 0x16, 0x18, 0x18,
+};
+
+/**
+ * ntfs_are_names_equal - compare two Unicode names for equality
+ * @s1:			name to compare to @s2
+ * @s1_len:		length in Unicode characters of @s1
+ * @s2:			name to compare to @s1
+ * @s2_len:		length in Unicode characters of @s2
+ * @ic:			ignore case bool
+ * @upcase:		upcase table (only if @ic == IGNORE_CASE)
+ * @upcase_size:	length in Unicode characters of @upcase (if present)
+ *
+ * Compare the names @s1 and @s2 and return 'true' (1) if the names are
+ * identical, or 'false' (0) if they are not identical. If @ic is IGNORE_CASE,
+ * the @upcase table is used to performa a case insensitive comparison.
+ */
+bool ntfs_are_names_equal(const ntfschar *s1, size_t s1_len,
+		const ntfschar *s2, size_t s2_len, const IGNORE_CASE_BOOL ic,
+		const ntfschar *upcase, const u32 upcase_size)
+{
+	if (s1_len != s2_len)
+		return false;
+	if (ic == CASE_SENSITIVE)
+		return !ntfs_ucsncmp(s1, s2, s1_len);
+	return !ntfs_ucsncasecmp(s1, s2, s1_len, upcase, upcase_size);
+}
+
+/**
+ * ntfs_collate_names - collate two Unicode names
+ * @name1:	first Unicode name to compare
+ * @name2:	second Unicode name to compare
+ * @err_val:	if @name1 contains an invalid character return this value
+ * @ic:		either CASE_SENSITIVE or IGNORE_CASE
+ * @upcase:	upcase table (ignored if @ic is CASE_SENSITIVE)
+ * @upcase_len:	upcase table size (ignored if @ic is CASE_SENSITIVE)
+ *
+ * ntfs_collate_names collates two Unicode names and returns:
+ *
+ *  -1 if the first name collates before the second one,
+ *   0 if the names match,
+ *   1 if the second name collates before the first one, or
+ * @err_val if an invalid character is found in @name1 during the comparison.
+ *
+ * The following characters are considered invalid: '"', '*', '<', '>' and '?'.
+ */
+int ntfs_collate_names(const ntfschar *name1, const u32 name1_len,
+		const ntfschar *name2, const u32 name2_len,
+		const int err_val, const IGNORE_CASE_BOOL ic,
+		const ntfschar *upcase, const u32 upcase_len)
+{
+	u32 cnt, min_len;
+	u16 c1, c2;
+
+	min_len = name1_len;
+	if (name1_len > name2_len)
+		min_len = name2_len;
+	for (cnt = 0; cnt < min_len; ++cnt) {
+		c1 = le16_to_cpu(*name1++);
+		c2 = le16_to_cpu(*name2++);
+		if (ic) {
+			if (c1 < upcase_len)
+				c1 = le16_to_cpu(upcase[c1]);
+			if (c2 < upcase_len)
+				c2 = le16_to_cpu(upcase[c2]);
+		}
+		if (c1 < 64 && legal_ansi_char_array[c1] & 8)
+			return err_val;
+		if (c1 < c2)
+			return -1;
+		if (c1 > c2)
+			return 1;
+	}
+	if (name1_len < name2_len)
+		return -1;
+	if (name1_len == name2_len)
+		return 0;
+	/* name1_len > name2_len */
+	c1 = le16_to_cpu(*name1);
+	if (c1 < 64 && legal_ansi_char_array[c1] & 8)
+		return err_val;
+	return 1;
+}
+
+/**
+ * ntfs_ucsncmp - compare two little endian Unicode strings
+ * @s1:		first string
+ * @s2:		second string
+ * @n:		maximum unicode characters to compare
+ *
+ * Compare the first @n characters of the Unicode strings @s1 and @s2,
+ * The strings in little endian format and appropriate le16_to_cpu()
+ * conversion is performed on non-little endian machines.
+ *
+ * The function returns an integer less than, equal to, or greater than zero
+ * if @s1 (or the first @n Unicode characters thereof) is found, respectively,
+ * to be less than, to match, or be greater than @s2.
+ */
+int ntfs_ucsncmp(const ntfschar *s1, const ntfschar *s2, size_t n)
+{
+	u16 c1, c2;
+	size_t i;
+
+	for (i = 0; i < n; ++i) {
+		c1 = le16_to_cpu(s1[i]);
+		c2 = le16_to_cpu(s2[i]);
+		if (c1 < c2)
+			return -1;
+		if (c1 > c2)
+			return 1;
+		if (!c1)
+			break;
+	}
+	return 0;
+}
+
+/**
+ * ntfs_ucsncasecmp - compare two little endian Unicode strings, ignoring case
+ * @s1:			first string
+ * @s2:			second string
+ * @n:			maximum unicode characters to compare
+ * @upcase:		upcase table
+ * @upcase_size:	upcase table size in Unicode characters
+ *
+ * Compare the first @n characters of the Unicode strings @s1 and @s2,
+ * ignoring case. The strings in little endian format and appropriate
+ * le16_to_cpu() conversion is performed on non-little endian machines.
+ *
+ * Each character is uppercased using the @upcase table before the comparison.
+ *
+ * The function returns an integer less than, equal to, or greater than zero
+ * if @s1 (or the first @n Unicode characters thereof) is found, respectively,
+ * to be less than, to match, or be greater than @s2.
+ */
+int ntfs_ucsncasecmp(const ntfschar *s1, const ntfschar *s2, size_t n,
+		const ntfschar *upcase, const u32 upcase_size)
+{
+	size_t i;
+	u16 c1, c2;
+
+	for (i = 0; i < n; ++i) {
+		if ((c1 = le16_to_cpu(s1[i])) < upcase_size)
+			c1 = le16_to_cpu(upcase[c1]);
+		if ((c2 = le16_to_cpu(s2[i])) < upcase_size)
+			c2 = le16_to_cpu(upcase[c2]);
+		if (c1 < c2)
+			return -1;
+		if (c1 > c2)
+			return 1;
+		if (!c1)
+			break;
+	}
+	return 0;
+}
+
+void ntfs_upcase_name(ntfschar *name, u32 name_len, const ntfschar *upcase,
+		const u32 upcase_len)
+{
+	u32 i;
+	u16 u;
+
+	for (i = 0; i < name_len; i++)
+		if ((u = le16_to_cpu(name[i])) < upcase_len)
+			name[i] = upcase[u];
+}
+
+void ntfs_file_upcase_value(FILE_NAME_ATTR *file_name_attr,
+		const ntfschar *upcase, const u32 upcase_len)
+{
+	ntfs_upcase_name((ntfschar*)&file_name_attr->file_name,
+			file_name_attr->file_name_length, upcase, upcase_len);
+}
+
+int ntfs_file_compare_values(FILE_NAME_ATTR *file_name_attr1,
+		FILE_NAME_ATTR *file_name_attr2,
+		const int err_val, const IGNORE_CASE_BOOL ic,
+		const ntfschar *upcase, const u32 upcase_len)
+{
+	return ntfs_collate_names((ntfschar*)&file_name_attr1->file_name,
+			file_name_attr1->file_name_length,
+			(ntfschar*)&file_name_attr2->file_name,
+			file_name_attr2->file_name_length,
+			err_val, ic, upcase, upcase_len);
+}
+
+/**
+ * ntfs_nlstoucs - convert NLS string to little endian Unicode string
+ * @vol:	ntfs volume which we are working with
+ * @ins:	input NLS string buffer
+ * @ins_len:	length of input string in bytes
+ * @outs:	on return contains the allocated output Unicode string buffer
+ *
+ * Convert the input string @ins, which is in whatever format the loaded NLS
+ * map dictates, into a little endian, 2-byte Unicode string.
+ *
+ * This function allocates the string and the caller is responsible for
+ * calling kmem_cache_free(ntfs_name_cache, *@outs); when finished with it.
+ *
+ * On success the function returns the number of Unicode characters written to
+ * the output string *@outs (>= 0), not counting the terminating Unicode NULL
+ * character. *@outs is set to the allocated output string buffer.
+ *
+ * On error, a negative number corresponding to the error code is returned. In
+ * that case the output string is not allocated. Both *@outs and *@outs_len
+ * are then undefined.
+ *
+ * This might look a bit odd due to fast path optimization...
+ */
+int ntfs_nlstoucs(const ntfs_volume *vol, const char *ins,
+		const int ins_len, ntfschar **outs)
+{
+	struct nls_table *nls = vol->nls_map;
+	ntfschar *ucs;
+	wchar_t wc;
+	int i, o, wc_len;
+
+	/* We do not trust outside sources. */
+	if (likely(ins)) {
+		ucs = kmem_cache_alloc(ntfs_name_cache, GFP_NOFS);
+		if (likely(ucs)) {
+			for (i = o = 0; i < ins_len; i += wc_len) {
+				wc_len = nls->char2uni(ins + i, ins_len - i,
+						&wc);
+				if (likely(wc_len >= 0 &&
+						o < NTFS_MAX_NAME_LEN)) {
+					if (likely(wc)) {
+						ucs[o++] = cpu_to_le16(wc);
+						continue;
+					} /* else if (!wc) */
+					break;
+				} /* else if (wc_len < 0 ||
+						o >= NTFS_MAX_NAME_LEN) */
+				goto name_err;
+			}
+			ucs[o] = 0;
+			*outs = ucs;
+			return o;
+		} /* else if (!ucs) */
+		ntfs_error(vol->sb, "Failed to allocate buffer for converted "
+				"name from ntfs_name_cache.");
+		return -ENOMEM;
+	} /* else if (!ins) */
+	ntfs_error(vol->sb, "Received NULL pointer.");
+	return -EINVAL;
+name_err:
+	kmem_cache_free(ntfs_name_cache, ucs);
+	if (wc_len < 0) {
+		ntfs_error(vol->sb, "Name using character set %s contains "
+				"characters that cannot be converted to "
+				"Unicode.", nls->charset);
+		i = -EILSEQ;
+	} else /* if (o >= NTFS_MAX_NAME_LEN) */ {
+		ntfs_error(vol->sb, "Name is too long (maximum length for a "
+				"name on NTFS is %d Unicode characters.",
+				NTFS_MAX_NAME_LEN);
+		i = -ENAMETOOLONG;
+	}
+	return i;
+}
+
+/**
+ * ntfs_ucstonls - convert little endian Unicode string to NLS string
+ * @vol:	ntfs volume which we are working with
+ * @ins:	input Unicode string buffer
+ * @ins_len:	length of input string in Unicode characters
+ * @outs:	on return contains the (allocated) output NLS string buffer
+ * @outs_len:	length of output string buffer in bytes
+ *
+ * Convert the input little endian, 2-byte Unicode string @ins, of length
+ * @ins_len into the string format dictated by the loaded NLS.
+ *
+ * If *@outs is NULL, this function allocates the string and the caller is
+ * responsible for calling kfree(*@outs); when finished with it. In this case
+ * @outs_len is ignored and can be 0.
+ *
+ * On success the function returns the number of bytes written to the output
+ * string *@outs (>= 0), not counting the terminating NULL byte. If the output
+ * string buffer was allocated, *@outs is set to it.
+ *
+ * On error, a negative number corresponding to the error code is returned. In
+ * that case the output string is not allocated. The contents of *@outs are
+ * then undefined.
+ *
+ * This might look a bit odd due to fast path optimization...
+ */
+int ntfs_ucstonls(const ntfs_volume *vol, const ntfschar *ins,
+		const int ins_len, unsigned char **outs, int outs_len)
+{
+	struct nls_table *nls = vol->nls_map;
+	unsigned char *ns;
+	int i, o, ns_len, wc;
+
+	/* We don't trust outside sources. */
+	if (ins) {
+		ns = *outs;
+		ns_len = outs_len;
+		if (ns && !ns_len) {
+			wc = -ENAMETOOLONG;
+			goto conversion_err;
+		}
+		if (!ns) {
+			ns_len = ins_len * NLS_MAX_CHARSET_SIZE;
+			ns = kmalloc(ns_len + 1, GFP_NOFS);
+			if (!ns)
+				goto mem_err_out;
+		}
+		for (i = o = 0; i < ins_len; i++) {
+retry:			wc = nls->uni2char(le16_to_cpu(ins[i]), ns + o,
+					ns_len - o);
+			if (wc > 0) {
+				o += wc;
+				continue;
+			} else if (!wc)
+				break;
+			else if (wc == -ENAMETOOLONG && ns != *outs) {
+				unsigned char *tc;
+				/* Grow in multiples of 64 bytes. */
+				tc = kmalloc((ns_len + 64) &
+						~63, GFP_NOFS);
+				if (tc) {
+					memcpy(tc, ns, ns_len);
+					ns_len = ((ns_len + 64) & ~63) - 1;
+					kfree(ns);
+					ns = tc;
+					goto retry;
+				} /* No memory so goto conversion_error; */
+			} /* wc < 0, real error. */
+			goto conversion_err;
+		}
+		ns[o] = 0;
+		*outs = ns;
+		return o;
+	} /* else (!ins) */
+	ntfs_error(vol->sb, "Received NULL pointer.");
+	return -EINVAL;
+conversion_err:
+	ntfs_error(vol->sb, "Unicode name contains characters that cannot be "
+			"converted to character set %s.  You might want to "
+			"try to use the mount option nls=utf8.", nls->charset);
+	if (ns != *outs)
+		kfree(ns);
+	if (wc != -ENAMETOOLONG)
+		wc = -EILSEQ;
+	return wc;
+mem_err_out:
+	ntfs_error(vol->sb, "Failed to allocate name!");
+	return -ENOMEM;
+}
diff --git a/fs/ntfs/upcase.c b/fs/ntfs/upcase.c
new file mode 100644
index 000000000..4ebe84a78
--- /dev/null
+++ b/fs/ntfs/upcase.c
@@ -0,0 +1,73 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * upcase.c - Generate the full NTFS Unicode upcase table in little endian.
+ *	      Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001 Richard Russon <ntfs@flatcap.org>
+ * Copyright (c) 2001-2006 Anton Altaparmakov
+ */
+
+#include "malloc.h"
+#include "ntfs.h"
+
+ntfschar *generate_default_upcase(void)
+{
+	static const int uc_run_table[][3] = { /* Start, End, Add */
+	{0x0061, 0x007B,  -32}, {0x0451, 0x045D, -80}, {0x1F70, 0x1F72,  74},
+	{0x00E0, 0x00F7,  -32}, {0x045E, 0x0460, -80}, {0x1F72, 0x1F76,  86},
+	{0x00F8, 0x00FF,  -32}, {0x0561, 0x0587, -48}, {0x1F76, 0x1F78, 100},
+	{0x0256, 0x0258, -205}, {0x1F00, 0x1F08,   8}, {0x1F78, 0x1F7A, 128},
+	{0x028A, 0x028C, -217}, {0x1F10, 0x1F16,   8}, {0x1F7A, 0x1F7C, 112},
+	{0x03AC, 0x03AD,  -38}, {0x1F20, 0x1F28,   8}, {0x1F7C, 0x1F7E, 126},
+	{0x03AD, 0x03B0,  -37}, {0x1F30, 0x1F38,   8}, {0x1FB0, 0x1FB2,   8},
+	{0x03B1, 0x03C2,  -32}, {0x1F40, 0x1F46,   8}, {0x1FD0, 0x1FD2,   8},
+	{0x03C2, 0x03C3,  -31}, {0x1F51, 0x1F52,   8}, {0x1FE0, 0x1FE2,   8},
+	{0x03C3, 0x03CC,  -32}, {0x1F53, 0x1F54,   8}, {0x1FE5, 0x1FE6,   7},
+	{0x03CC, 0x03CD,  -64}, {0x1F55, 0x1F56,   8}, {0x2170, 0x2180, -16},
+	{0x03CD, 0x03CF,  -63}, {0x1F57, 0x1F58,   8}, {0x24D0, 0x24EA, -26},
+	{0x0430, 0x0450,  -32}, {0x1F60, 0x1F68,   8}, {0xFF41, 0xFF5B, -32},
+	{0}
+	};
+
+	static const int uc_dup_table[][2] = { /* Start, End */
+	{0x0100, 0x012F}, {0x01A0, 0x01A6}, {0x03E2, 0x03EF}, {0x04CB, 0x04CC},
+	{0x0132, 0x0137}, {0x01B3, 0x01B7}, {0x0460, 0x0481}, {0x04D0, 0x04EB},
+	{0x0139, 0x0149}, {0x01CD, 0x01DD}, {0x0490, 0x04BF}, {0x04EE, 0x04F5},
+	{0x014A, 0x0178}, {0x01DE, 0x01EF}, {0x04BF, 0x04BF}, {0x04F8, 0x04F9},
+	{0x0179, 0x017E}, {0x01F4, 0x01F5}, {0x04C1, 0x04C4}, {0x1E00, 0x1E95},
+	{0x018B, 0x018B}, {0x01FA, 0x0218}, {0x04C7, 0x04C8}, {0x1EA0, 0x1EF9},
+	{0}
+	};
+
+	static const int uc_word_table[][2] = { /* Offset, Value */
+	{0x00FF, 0x0178}, {0x01AD, 0x01AC}, {0x01F3, 0x01F1}, {0x0269, 0x0196},
+	{0x0183, 0x0182}, {0x01B0, 0x01AF}, {0x0253, 0x0181}, {0x026F, 0x019C},
+	{0x0185, 0x0184}, {0x01B9, 0x01B8}, {0x0254, 0x0186}, {0x0272, 0x019D},
+	{0x0188, 0x0187}, {0x01BD, 0x01BC}, {0x0259, 0x018F}, {0x0275, 0x019F},
+	{0x018C, 0x018B}, {0x01C6, 0x01C4}, {0x025B, 0x0190}, {0x0283, 0x01A9},
+	{0x0192, 0x0191}, {0x01C9, 0x01C7}, {0x0260, 0x0193}, {0x0288, 0x01AE},
+	{0x0199, 0x0198}, {0x01CC, 0x01CA}, {0x0263, 0x0194}, {0x0292, 0x01B7},
+	{0x01A8, 0x01A7}, {0x01DD, 0x018E}, {0x0268, 0x0197},
+	{0}
+	};
+
+	int i, r;
+	ntfschar *uc;
+
+	uc = ntfs_malloc_nofs(default_upcase_len * sizeof(ntfschar));
+	if (!uc)
+		return uc;
+	memset(uc, 0, default_upcase_len * sizeof(ntfschar));
+	/* Generate the little endian Unicode upcase table used by ntfs. */
+	for (i = 0; i < default_upcase_len; i++)
+		uc[i] = cpu_to_le16(i);
+	for (r = 0; uc_run_table[r][0]; r++)
+		for (i = uc_run_table[r][0]; i < uc_run_table[r][1]; i++)
+			le16_add_cpu(&uc[i], uc_run_table[r][2]);
+	for (r = 0; uc_dup_table[r][0]; r++)
+		for (i = uc_dup_table[r][0]; i < uc_dup_table[r][1]; i += 2)
+			le16_add_cpu(&uc[i + 1], -1);
+	for (r = 0; uc_word_table[r][0]; r++)
+		uc[uc_word_table[r][0]] = cpu_to_le16(uc_word_table[r][1]);
+	return uc;
+}
diff --git a/fs/ntfs/usnjrnl.c b/fs/ntfs/usnjrnl.c
new file mode 100644
index 000000000..9097a0b4e
--- /dev/null
+++ b/fs/ntfs/usnjrnl.c
@@ -0,0 +1,70 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * usnjrnl.h - NTFS kernel transaction log ($UsnJrnl) handling.  Part of the
+ *	       Linux-NTFS project.
+ *
+ * Copyright (c) 2005 Anton Altaparmakov
+ */
+
+#ifdef NTFS_RW
+
+#include <linux/fs.h>
+#include <linux/highmem.h>
+#include <linux/mm.h>
+
+#include "aops.h"
+#include "debug.h"
+#include "endian.h"
+#include "time.h"
+#include "types.h"
+#include "usnjrnl.h"
+#include "volume.h"
+
+/**
+ * ntfs_stamp_usnjrnl - stamp the transaction log ($UsnJrnl) on an ntfs volume
+ * @vol:	ntfs volume on which to stamp the transaction log
+ *
+ * Stamp the transaction log ($UsnJrnl) on the ntfs volume @vol and return
+ * 'true' on success and 'false' on error.
+ *
+ * This function assumes that the transaction log has already been loaded and
+ * consistency checked by a call to fs/ntfs/super.c::load_and_init_usnjrnl().
+ */
+bool ntfs_stamp_usnjrnl(ntfs_volume *vol)
+{
+	ntfs_debug("Entering.");
+	if (likely(!NVolUsnJrnlStamped(vol))) {
+		sle64 stamp;
+		struct page *page;
+		USN_HEADER *uh;
+
+		page = ntfs_map_page(vol->usnjrnl_max_ino->i_mapping, 0);
+		if (IS_ERR(page)) {
+			ntfs_error(vol->sb, "Failed to read from "
+					"$UsnJrnl/$DATA/$Max attribute.");
+			return false;
+		}
+		uh = (USN_HEADER*)page_address(page);
+		stamp = get_current_ntfs_time();
+		ntfs_debug("Stamping transaction log ($UsnJrnl): old "
+				"journal_id 0x%llx, old lowest_valid_usn "
+				"0x%llx, new journal_id 0x%llx, new "
+				"lowest_valid_usn 0x%llx.",
+				(long long)sle64_to_cpu(uh->journal_id),
+				(long long)sle64_to_cpu(uh->lowest_valid_usn),
+				(long long)sle64_to_cpu(stamp),
+				i_size_read(vol->usnjrnl_j_ino));
+		uh->lowest_valid_usn =
+				cpu_to_sle64(i_size_read(vol->usnjrnl_j_ino));
+		uh->journal_id = stamp;
+		flush_dcache_page(page);
+		set_page_dirty(page);
+		ntfs_unmap_page(page);
+		/* Set the flag so we do not have to do it again on remount. */
+		NVolSetUsnJrnlStamped(vol);
+	}
+	ntfs_debug("Done.");
+	return true;
+}
+
+#endif /* NTFS_RW */
diff --git a/fs/ntfs/usnjrnl.h b/fs/ntfs/usnjrnl.h
new file mode 100644
index 000000000..85f531b59
--- /dev/null
+++ b/fs/ntfs/usnjrnl.h
@@ -0,0 +1,191 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * usnjrnl.h - Defines for NTFS kernel transaction log ($UsnJrnl) handling.
+ *	       Part of the Linux-NTFS project.
+ *
+ * Copyright (c) 2005 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_USNJRNL_H
+#define _LINUX_NTFS_USNJRNL_H
+
+#ifdef NTFS_RW
+
+#include "types.h"
+#include "endian.h"
+#include "layout.h"
+#include "volume.h"
+
+/*
+ * Transaction log ($UsnJrnl) organization:
+ *
+ * The transaction log records whenever a file is modified in any way.  So for
+ * example it will record that file "blah" was written to at a particular time
+ * but not what was written.  If will record that a file was deleted or
+ * created, that a file was truncated, etc.  See below for all the reason
+ * codes used.
+ *
+ * The transaction log is in the $Extend directory which is in the root
+ * directory of each volume.  If it is not present it means transaction
+ * logging is disabled.  If it is present it means transaction logging is
+ * either enabled or in the process of being disabled in which case we can
+ * ignore it as it will go away as soon as Windows gets its hands on it.
+ *
+ * To determine whether the transaction logging is enabled or in the process
+ * of being disabled, need to check the volume flags in the
+ * $VOLUME_INFORMATION attribute in the $Volume system file (which is present
+ * in the root directory and has a fixed mft record number, see layout.h).
+ * If the flag VOLUME_DELETE_USN_UNDERWAY is set it means the transaction log
+ * is in the process of being disabled and if this flag is clear it means the
+ * transaction log is enabled.
+ *
+ * The transaction log consists of two parts; the $DATA/$Max attribute as well
+ * as the $DATA/$J attribute.  $Max is a header describing the transaction
+ * log whilst $J is the transaction log data itself as a sequence of variable
+ * sized USN_RECORDs (see below for all the structures).
+ *
+ * We do not care about transaction logging at this point in time but we still
+ * need to let windows know that the transaction log is out of date.  To do
+ * this we need to stamp the transaction log.  This involves setting the
+ * lowest_valid_usn field in the $DATA/$Max attribute to the usn to be used
+ * for the next added USN_RECORD to the $DATA/$J attribute as well as
+ * generating a new journal_id in $DATA/$Max.
+ *
+ * The journal_id is as of the current version (2.0) of the transaction log
+ * simply the 64-bit timestamp of when the journal was either created or last
+ * stamped.
+ *
+ * To determine the next usn there are two ways.  The first is to parse
+ * $DATA/$J and to find the last USN_RECORD in it and to add its record_length
+ * to its usn (which is the byte offset in the $DATA/$J attribute).  The
+ * second is simply to take the data size of the attribute.  Since the usns
+ * are simply byte offsets into $DATA/$J, this is exactly the next usn.  For
+ * obvious reasons we use the second method as it is much simpler and faster.
+ *
+ * As an aside, note that to actually disable the transaction log, one would
+ * need to set the VOLUME_DELETE_USN_UNDERWAY flag (see above), then go
+ * through all the mft records on the volume and set the usn field in their
+ * $STANDARD_INFORMATION attribute to zero.  Once that is done, one would need
+ * to delete the transaction log file, i.e. \$Extent\$UsnJrnl, and finally,
+ * one would need to clear the VOLUME_DELETE_USN_UNDERWAY flag.
+ *
+ * Note that if a volume is unmounted whilst the transaction log is being
+ * disabled, the process will continue the next time the volume is mounted.
+ * This is why we can safely mount read-write when we see a transaction log
+ * in the process of being deleted.
+ */
+
+/* Some $UsnJrnl related constants. */
+#define UsnJrnlMajorVer		2
+#define UsnJrnlMinorVer		0
+
+/*
+ * $DATA/$Max attribute.  This is (always?) resident and has a fixed size of
+ * 32 bytes.  It contains the header describing the transaction log.
+ */
+typedef struct {
+/*Ofs*/
+/*   0*/sle64 maximum_size;	/* The maximum on-disk size of the $DATA/$J
+				   attribute. */
+/*   8*/sle64 allocation_delta;	/* Number of bytes by which to increase the
+				   size of the $DATA/$J attribute. */
+/*0x10*/sle64 journal_id;	/* Current id of the transaction log. */
+/*0x18*/leUSN lowest_valid_usn;	/* Lowest valid usn in $DATA/$J for the
+				   current journal_id. */
+/* sizeof() = 32 (0x20) bytes */
+} __attribute__ ((__packed__)) USN_HEADER;
+
+/*
+ * Reason flags (32-bit).  Cumulative flags describing the change(s) to the
+ * file since it was last opened.  I think the names speak for themselves but
+ * if you disagree check out the descriptions in the Linux NTFS project NTFS
+ * documentation: http://www.linux-ntfs.org/
+ */
+enum {
+	USN_REASON_DATA_OVERWRITE	= cpu_to_le32(0x00000001),
+	USN_REASON_DATA_EXTEND		= cpu_to_le32(0x00000002),
+	USN_REASON_DATA_TRUNCATION	= cpu_to_le32(0x00000004),
+	USN_REASON_NAMED_DATA_OVERWRITE	= cpu_to_le32(0x00000010),
+	USN_REASON_NAMED_DATA_EXTEND	= cpu_to_le32(0x00000020),
+	USN_REASON_NAMED_DATA_TRUNCATION= cpu_to_le32(0x00000040),
+	USN_REASON_FILE_CREATE		= cpu_to_le32(0x00000100),
+	USN_REASON_FILE_DELETE		= cpu_to_le32(0x00000200),
+	USN_REASON_EA_CHANGE		= cpu_to_le32(0x00000400),
+	USN_REASON_SECURITY_CHANGE	= cpu_to_le32(0x00000800),
+	USN_REASON_RENAME_OLD_NAME	= cpu_to_le32(0x00001000),
+	USN_REASON_RENAME_NEW_NAME	= cpu_to_le32(0x00002000),
+	USN_REASON_INDEXABLE_CHANGE	= cpu_to_le32(0x00004000),
+	USN_REASON_BASIC_INFO_CHANGE	= cpu_to_le32(0x00008000),
+	USN_REASON_HARD_LINK_CHANGE	= cpu_to_le32(0x00010000),
+	USN_REASON_COMPRESSION_CHANGE	= cpu_to_le32(0x00020000),
+	USN_REASON_ENCRYPTION_CHANGE	= cpu_to_le32(0x00040000),
+	USN_REASON_OBJECT_ID_CHANGE	= cpu_to_le32(0x00080000),
+	USN_REASON_REPARSE_POINT_CHANGE	= cpu_to_le32(0x00100000),
+	USN_REASON_STREAM_CHANGE	= cpu_to_le32(0x00200000),
+	USN_REASON_CLOSE		= cpu_to_le32(0x80000000),
+};
+
+typedef le32 USN_REASON_FLAGS;
+
+/*
+ * Source info flags (32-bit).  Information about the source of the change(s)
+ * to the file.  For detailed descriptions of what these mean, see the Linux
+ * NTFS project NTFS documentation:
+ *	http://www.linux-ntfs.org/
+ */
+enum {
+	USN_SOURCE_DATA_MANAGEMENT	  = cpu_to_le32(0x00000001),
+	USN_SOURCE_AUXILIARY_DATA	  = cpu_to_le32(0x00000002),
+	USN_SOURCE_REPLICATION_MANAGEMENT = cpu_to_le32(0x00000004),
+};
+
+typedef le32 USN_SOURCE_INFO_FLAGS;
+
+/*
+ * $DATA/$J attribute.  This is always non-resident, is marked as sparse, and
+ * is of variabled size.  It consists of a sequence of variable size
+ * USN_RECORDS.  The minimum allocated_size is allocation_delta as
+ * specified in $DATA/$Max.  When the maximum_size specified in $DATA/$Max is
+ * exceeded by more than allocation_delta bytes, allocation_delta bytes are
+ * allocated and appended to the $DATA/$J attribute and an equal number of
+ * bytes at the beginning of the attribute are freed and made sparse.  Note the
+ * making sparse only happens at volume checkpoints and hence the actual
+ * $DATA/$J size can exceed maximum_size + allocation_delta temporarily.
+ */
+typedef struct {
+/*Ofs*/
+/*   0*/le32 length;		/* Byte size of this record (8-byte
+				   aligned). */
+/*   4*/le16 major_ver;		/* Major version of the transaction log used
+				   for this record. */
+/*   6*/le16 minor_ver;		/* Minor version of the transaction log used
+				   for this record. */
+/*   8*/leMFT_REF mft_reference;/* The mft reference of the file (or
+				   directory) described by this record. */
+/*0x10*/leMFT_REF parent_directory;/* The mft reference of the parent
+				   directory of the file described by this
+				   record. */
+/*0x18*/leUSN usn;		/* The usn of this record.  Equals the offset
+				   within the $DATA/$J attribute. */
+/*0x20*/sle64 time;		/* Time when this record was created. */
+/*0x28*/USN_REASON_FLAGS reason;/* Reason flags (see above). */
+/*0x2c*/USN_SOURCE_INFO_FLAGS source_info;/* Source info flags (see above). */
+/*0x30*/le32 security_id;	/* File security_id copied from
+				   $STANDARD_INFORMATION. */
+/*0x34*/FILE_ATTR_FLAGS file_attributes;	/* File attributes copied from
+				   $STANDARD_INFORMATION or $FILE_NAME (not
+				   sure which). */
+/*0x38*/le16 file_name_size;	/* Size of the file name in bytes. */
+/*0x3a*/le16 file_name_offset;	/* Offset to the file name in bytes from the
+				   start of this record. */
+/*0x3c*/ntfschar file_name[0];	/* Use when creating only.  When reading use
+				   file_name_offset to determine the location
+				   of the name. */
+/* sizeof() = 60 (0x3c) bytes */
+} __attribute__ ((__packed__)) USN_RECORD;
+
+extern bool ntfs_stamp_usnjrnl(ntfs_volume *vol);
+
+#endif /* NTFS_RW */
+
+#endif /* _LINUX_NTFS_USNJRNL_H */
diff --git a/fs/ntfs/volume.h b/fs/ntfs/volume.h
new file mode 100644
index 000000000..930a9ae8a
--- /dev/null
+++ b/fs/ntfs/volume.h
@@ -0,0 +1,164 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * volume.h - Defines for volume structures in NTFS Linux kernel driver. Part
+ *	      of the Linux-NTFS project.
+ *
+ * Copyright (c) 2001-2006 Anton Altaparmakov
+ * Copyright (c) 2002 Richard Russon
+ */
+
+#ifndef _LINUX_NTFS_VOLUME_H
+#define _LINUX_NTFS_VOLUME_H
+
+#include <linux/rwsem.h>
+#include <linux/uidgid.h>
+
+#include "types.h"
+#include "layout.h"
+
+/*
+ * The NTFS in memory super block structure.
+ */
+typedef struct {
+	/*
+	 * FIXME: Reorder to have commonly used together element within the
+	 * same cache line, aiming at a cache line size of 32 bytes. Aim for
+	 * 64 bytes for less commonly used together elements. Put most commonly
+	 * used elements to front of structure. Obviously do this only when the
+	 * structure has stabilized... (AIA)
+	 */
+	/* Device specifics. */
+	struct super_block *sb;		/* Pointer back to the super_block. */
+	LCN nr_blocks;			/* Number of sb->s_blocksize bytes
+					   sized blocks on the device. */
+	/* Configuration provided by user at mount time. */
+	unsigned long flags;		/* Miscellaneous flags, see below. */
+	kuid_t uid;			/* uid that files will be mounted as. */
+	kgid_t gid;			/* gid that files will be mounted as. */
+	umode_t fmask;			/* The mask for file permissions. */
+	umode_t dmask;			/* The mask for directory
+					   permissions. */
+	u8 mft_zone_multiplier;		/* Initial mft zone multiplier. */
+	u8 on_errors;			/* What to do on filesystem errors. */
+	/* NTFS bootsector provided information. */
+	u16 sector_size;		/* in bytes */
+	u8 sector_size_bits;		/* log2(sector_size) */
+	u32 cluster_size;		/* in bytes */
+	u32 cluster_size_mask;		/* cluster_size - 1 */
+	u8 cluster_size_bits;		/* log2(cluster_size) */
+	u32 mft_record_size;		/* in bytes */
+	u32 mft_record_size_mask;	/* mft_record_size - 1 */
+	u8 mft_record_size_bits;	/* log2(mft_record_size) */
+	u32 index_record_size;		/* in bytes */
+	u32 index_record_size_mask;	/* index_record_size - 1 */
+	u8 index_record_size_bits;	/* log2(index_record_size) */
+	LCN nr_clusters;		/* Volume size in clusters == number of
+					   bits in lcn bitmap. */
+	LCN mft_lcn;			/* Cluster location of mft data. */
+	LCN mftmirr_lcn;		/* Cluster location of copy of mft. */
+	u64 serial_no;			/* The volume serial number. */
+	/* Mount specific NTFS information. */
+	u32 upcase_len;			/* Number of entries in upcase[]. */
+	ntfschar *upcase;		/* The upcase table. */
+
+	s32 attrdef_size;		/* Size of the attribute definition
+					   table in bytes. */
+	ATTR_DEF *attrdef;		/* Table of attribute definitions.
+					   Obtained from FILE_AttrDef. */
+
+#ifdef NTFS_RW
+	/* Variables used by the cluster and mft allocators. */
+	s64 mft_data_pos;		/* Mft record number at which to
+					   allocate the next mft record. */
+	LCN mft_zone_start;		/* First cluster of the mft zone. */
+	LCN mft_zone_end;		/* First cluster beyond the mft zone. */
+	LCN mft_zone_pos;		/* Current position in the mft zone. */
+	LCN data1_zone_pos;		/* Current position in the first data
+					   zone. */
+	LCN data2_zone_pos;		/* Current position in the second data
+					   zone. */
+#endif /* NTFS_RW */
+
+	struct inode *mft_ino;		/* The VFS inode of $MFT. */
+
+	struct inode *mftbmp_ino;	/* Attribute inode for $MFT/$BITMAP. */
+	struct rw_semaphore mftbmp_lock; /* Lock for serializing accesses to the
+					    mft record bitmap ($MFT/$BITMAP). */
+#ifdef NTFS_RW
+	struct inode *mftmirr_ino;	/* The VFS inode of $MFTMirr. */
+	int mftmirr_size;		/* Size of mft mirror in mft records. */
+
+	struct inode *logfile_ino;	/* The VFS inode of $LogFile. */
+#endif /* NTFS_RW */
+
+	struct inode *lcnbmp_ino;	/* The VFS inode of $Bitmap. */
+	struct rw_semaphore lcnbmp_lock; /* Lock for serializing accesses to the
+					    cluster bitmap ($Bitmap/$DATA). */
+
+	struct inode *vol_ino;		/* The VFS inode of $Volume. */
+	VOLUME_FLAGS vol_flags;		/* Volume flags. */
+	u8 major_ver;			/* Ntfs major version of volume. */
+	u8 minor_ver;			/* Ntfs minor version of volume. */
+
+	struct inode *root_ino;		/* The VFS inode of the root
+					   directory. */
+	struct inode *secure_ino;	/* The VFS inode of $Secure (NTFS3.0+
+					   only, otherwise NULL). */
+	struct inode *extend_ino;	/* The VFS inode of $Extend (NTFS3.0+
+					   only, otherwise NULL). */
+#ifdef NTFS_RW
+	/* $Quota stuff is NTFS3.0+ specific.  Unused/NULL otherwise. */
+	struct inode *quota_ino;	/* The VFS inode of $Quota. */
+	struct inode *quota_q_ino;	/* Attribute inode for $Quota/$Q. */
+	/* $UsnJrnl stuff is NTFS3.0+ specific.  Unused/NULL otherwise. */
+	struct inode *usnjrnl_ino;	/* The VFS inode of $UsnJrnl. */
+	struct inode *usnjrnl_max_ino;	/* Attribute inode for $UsnJrnl/$Max. */
+	struct inode *usnjrnl_j_ino;	/* Attribute inode for $UsnJrnl/$J. */
+#endif /* NTFS_RW */
+	struct nls_table *nls_map;
+} ntfs_volume;
+
+/*
+ * Defined bits for the flags field in the ntfs_volume structure.
+ */
+typedef enum {
+	NV_Errors,		/* 1: Volume has errors, prevent remount rw. */
+	NV_ShowSystemFiles,	/* 1: Return system files in ntfs_readdir(). */
+	NV_CaseSensitive,	/* 1: Treat file names as case sensitive and
+				      create filenames in the POSIX namespace.
+				      Otherwise be case insensitive but still
+				      create file names in POSIX namespace. */
+	NV_LogFileEmpty,	/* 1: $LogFile journal is empty. */
+	NV_QuotaOutOfDate,	/* 1: $Quota is out of date. */
+	NV_UsnJrnlStamped,	/* 1: $UsnJrnl has been stamped. */
+	NV_SparseEnabled,	/* 1: May create sparse files. */
+} ntfs_volume_flags;
+
+/*
+ * Macro tricks to expand the NVolFoo(), NVolSetFoo(), and NVolClearFoo()
+ * functions.
+ */
+#define DEFINE_NVOL_BIT_OPS(flag)					\
+static inline int NVol##flag(ntfs_volume *vol)		\
+{							\
+	return test_bit(NV_##flag, &(vol)->flags);	\
+}							\
+static inline void NVolSet##flag(ntfs_volume *vol)	\
+{							\
+	set_bit(NV_##flag, &(vol)->flags);		\
+}							\
+static inline void NVolClear##flag(ntfs_volume *vol)	\
+{							\
+	clear_bit(NV_##flag, &(vol)->flags);		\
+}
+
+/* Emit the ntfs volume bitops functions. */
+DEFINE_NVOL_BIT_OPS(Errors)
+DEFINE_NVOL_BIT_OPS(ShowSystemFiles)
+DEFINE_NVOL_BIT_OPS(CaseSensitive)
+DEFINE_NVOL_BIT_OPS(LogFileEmpty)
+DEFINE_NVOL_BIT_OPS(QuotaOutOfDate)
+DEFINE_NVOL_BIT_OPS(UsnJrnlStamped)
+DEFINE_NVOL_BIT_OPS(SparseEnabled)
+
+#endif /* _LINUX_NTFS_VOLUME_H */