Linux-2.6.12-rc2v2.6.12-rc2

Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!

1 files changed, 2324 insertions, 0 deletions

diff --git a/fs/ntfs/aops.c b/fs/ntfs/aops.c
new file mode 100644
index 00000000000..45d56e41ed9
--- /dev/null
+++ b/fs/ntfs/aops.c
@@ -0,0 +1,2324 @@
+/**
+ * aops.c - 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
+ *
+ * This program/include file is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as published
+ * by the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program/include file is distributed in the hope that it will be
+ * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program (in the main directory of the Linux-NTFS
+ * distribution in the file COPYING); if not, write to the Free Software
+ * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/errno.h>
+#include <linux/mm.h>
+#include <linux/pagemap.h>
+#include <linux/swap.h>
+#include <linux/buffer_head.h>
+#include <linux/writeback.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)
+{
+	static DEFINE_SPINLOCK(page_uptodate_lock);
+	unsigned long flags;
+	struct buffer_head *tmp;
+	struct page *page;
+	ntfs_inode *ni;
+	int page_uptodate = 1;
+
+	page = bh->b_page;
+	ni = NTFS_I(page->mapping->host);
+
+	if (likely(uptodate)) {
+		s64 file_ofs;
+
+		set_buffer_uptodate(bh);
+
+		file_ofs = ((s64)page->index << PAGE_CACHE_SHIFT) +
+				bh_offset(bh);
+		/* Check for the current buffer head overflowing. */
+		if (file_ofs + bh->b_size > ni->initialized_size) {
+			char *addr;
+			int ofs = 0;
+
+			if (file_ofs < ni->initialized_size)
+				ofs = ni->initialized_size - file_ofs;
+			addr = kmap_atomic(page, KM_BIO_SRC_IRQ);
+			memset(addr + bh_offset(bh) + ofs, 0, bh->b_size - ofs);
+			flush_dcache_page(page);
+			kunmap_atomic(addr, KM_BIO_SRC_IRQ);
+		}
+	} else {
+		clear_buffer_uptodate(bh);
+		ntfs_error(ni->vol->sb, "Buffer I/O error, logical block %llu.",
+				(unsigned long long)bh->b_blocknr);
+		SetPageError(page);
+	}
+	spin_lock_irqsave(&page_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(&page_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 {
+		char *addr;
+		unsigned int i, recs;
+		u32 rec_size;
+
+		rec_size = ni->itype.index.block_size;
+		recs = PAGE_CACHE_SIZE / rec_size;
+		/* Should have been verified before we got here... */
+		BUG_ON(!recs);
+		addr = kmap_atomic(page, KM_BIO_SRC_IRQ);
+		for (i = 0; i < recs; i++)
+			post_read_mst_fixup((NTFS_RECORD*)(addr +
+					i * rec_size), rec_size);
+		flush_dcache_page(page);
+		kunmap_atomic(addr, KM_BIO_SRC_IRQ);
+		if (likely(!PageError(page) && page_uptodate))
+			SetPageUptodate(page);
+	}
+	unlock_page(page);
+	return;
+still_busy:
+	spin_unlock_irqrestore(&page_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)
+{
+	VCN vcn;
+	LCN lcn;
+	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 int blocksize, vcn_ofs;
+	int i, nr;
+	unsigned char blocksize_bits;
+
+	ni = NTFS_I(page->mapping->host);
+	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_bits = VFS_I(ni)->i_blkbits;
+	blocksize = 1 << blocksize_bits;
+
+	if (!page_has_buffers(page))
+		create_empty_buffers(page, blocksize, 0);
+	bh = head = page_buffers(page);
+	if (unlikely(!bh)) {
+		unlock_page(page);
+		return -ENOMEM;
+	}
+
+	iblock = (s64)page->index << (PAGE_CACHE_SHIFT - blocksize_bits);
+	lblock = (ni->allocated_size + blocksize - 1) >> blocksize_bits;
+	zblock = (ni->initialized_size + blocksize - 1) >> blocksize_bits;
+
+	/* Loop through all the buffers in the page. */
+	rl = NULL;
+	nr = i = 0;
+	do {
+		u8 *kaddr;
+
+		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) {
+				int err;
+				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;
+				lcn = err;
+			}
+			/* Hard error, zero out region. */
+			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 %lli).", ni->mft_no,
+					ni->type, (unsigned long long)vcn,
+					vcn_ofs, is_retry ? " even after "
+					"retrying" : "", (long long)lcn);
+		}
+		/*
+		 * 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:
+		kaddr = kmap_atomic(page, KM_USER0);
+		memset(kaddr + i * blocksize, 0, blocksize);
+		flush_dcache_page(page);
+		kunmap_atomic(kaddr, KM_USER0);
+		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(READ, 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;
+	ntfs_inode *ni, *base_ni;
+	u8 *kaddr;
+	ntfs_attr_search_ctx *ctx;
+	MFT_RECORD *mrec;
+	u32 attr_len;
+	int err = 0;
+
+	BUG_ON(!PageLocked(page));
+	/*
+	 * 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(page->mapping->host);
+
+	/* NInoNonResident() == NInoIndexAllocPresent() */
+	if (NInoNonResident(ni)) {
+		/*
+		 * Only unnamed $DATA attributes can be compressed or
+		 * encrypted.
+		 */
+		if (ni->type == AT_DATA && !ni->name_len) {
+			/* If file is encrypted, deny access, just like NT4. */
+			if (NInoEncrypted(ni)) {
+				err = -EACCES;
+				goto err_out;
+			}
+			/* Compressed data streams are handled in compress.c. */
+			if (NInoCompressed(ni))
+				return ntfs_read_compressed_block(page);
+		}
+		/* Normal 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.  We can also do this if the file size is 0.
+	 */
+	if (unlikely(page->index > 0 || !i_size_read(VFS_I(ni)))) {
+		kaddr = kmap_atomic(page, KM_USER0);
+		memset(kaddr, 0, PAGE_CACHE_SIZE);
+		flush_dcache_page(page);
+		kunmap_atomic(kaddr, KM_USER0);
+		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;
+	}
+	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);
+	i_size = i_size_read(VFS_I(ni));
+	if (unlikely(attr_len > i_size))
+		attr_len = i_size;
+	kaddr = kmap_atomic(page, KM_USER0);
+	/* Copy the data to the page. */
+	memcpy(kaddr, (u8*)ctx->attr +
+			le16_to_cpu(ctx->attr->data.resident.value_offset),
+			attr_len);
+	/* Zero the remainder of the page. */
+	memset(kaddr + attr_len, 0, PAGE_CACHE_SIZE - attr_len);
+	flush_dcache_page(page);
+	kunmap_atomic(kaddr, KM_USER0);
+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;
+	sector_t block, dblock, iblock;
+	struct inode *vi;
+	ntfs_inode *ni;
+	ntfs_volume *vol;
+	runlist_element *rl;
+	struct buffer_head *bh, *head;
+	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_bits = vi->i_blkbits;
+	blocksize = 1 << blocksize_bits;
+
+	if (!page_has_buffers(page)) {
+		BUG_ON(!PageUptodate(page));
+		create_empty_buffers(page, blocksize,
+				(1 << BH_Uptodate) | (1 << BH_Dirty));
+	}
+	bh = head = page_buffers(page);
+	if (unlikely(!bh)) {
+		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
+		 * buffer's dirty state as-is.
+		 */
+		redirty_page_for_writepage(wbc, page);
+		unlock_page(page);
+		return 0;
+	}
+
+	/* NOTE: Different naming scheme to ntfs_read_block()! */
+
+	/* The first block in the page. */
+	block = (s64)page->index << (PAGE_CACHE_SHIFT - blocksize_bits);
+
+	/* The first out of bounds block for the data size. */
+	dblock = (vi->i_size + blocksize - 1) >> blocksize_bits;
+
+	/* The last (fully or partially) initialized block. */
+	iblock = ni->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) &&
+				(ni->initialized_size < vi->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();
+				// page_cache_release()
+				// 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) {
+			// TODO: Instantiate the hole.
+			// clear_buffer_new(bh);
+			// unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
+			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;
+			lcn = err;
+		}
+		/* 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, offset 0x%x "
+				"because its location on disk could not be "
+				"determined%s (error code %lli).", ni->mft_no,
+				ni->type, (unsigned long long)vcn,
+				vcn_ofs, is_retry ? " even after "
+				"retrying" : "", (long long)lcn);
+		if (!err)
+			err = -EIO;
+		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() isn't 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 {
+		get_bh(bh);
+		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. */
+	unlock_page(page);
+
+	/*
+	 * Submit the prepared buffers for i/o. Note the page is unlocked,
+	 * and the async write i/o completion handler can end_page_writeback()
+	 * at any time after the *first* submit_bh(). So the buffers can then
+	 * disappear...
+	 */
+	need_end_writeback = TRUE;
+	do {
+		struct buffer_head *next = bh->b_this_page;
+		if (buffer_async_write(bh)) {
+			submit_bh(WRITE, bh);
+			need_end_writeback = FALSE;
+		}
+		put_bh(bh);
+		bh = next;
+	} while (bh != head);
+
+	/* 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 char bh_size_bits = vi->i_blkbits;
+	unsigned int bh_size = 1 << bh_size_bits;
+	unsigned int rec_size = ni->itype.index.block_size;
+	ntfs_inode *locked_nis[PAGE_CACHE_SIZE / rec_size];
+	struct buffer_head *bh, *head, *tbh, *rec_start_bh;
+	int max_bhs = PAGE_CACHE_SIZE / bh_size;
+	struct buffer_head *bhs[max_bhs];
+	runlist_element *rl;
+	int i, nr_locked_nis, nr_recs, nr_bhs, bhs_per_rec, err, err2;
+	unsigned rec_size_bits;
+	BOOL sync, is_mft, page_is_dirty, rec_is_dirty;
+
+	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)));
+	BUG_ON(!max_bhs);
+
+	/* Were we called for sync purposes? */
+	sync = (wbc->sync_mode == WB_SYNC_ALL);
+
+	/* Make sure we have mapped buffers. */
+	BUG_ON(!page_has_buffers(page));
+	bh = head = page_buffers(page);
+	BUG_ON(!bh);
+
+	rec_size_bits = ni->itype.index.block_size_bits;
+	BUG_ON(!(PAGE_CACHE_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_CACHE_SHIFT - bh_size_bits);
+
+	/* The first out of bounds block for the data size. */
+	dblock = (vi->i_size + 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);
+				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;
+
+			/* 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;
+				/* 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).", (s64)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_CACHE_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 (unlikely(test_set_buffer_locked(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(WRITE, 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));