aops.c 48.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4
/**
 * aops.c - NTFS kernel address space operations and page cache handling.
 *	    Part of the Linux-NTFS project.
 *
5
 * Copyright (c) 2001-2007 Anton Altaparmakov
L
Linus Torvalds 已提交
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
 * 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>
25
#include <linux/fs.h>
26
#include <linux/gfp.h>
L
Linus Torvalds 已提交
27 28 29 30 31
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/swap.h>
#include <linux/buffer_head.h>
#include <linux/writeback.h>
A
Andrew Morton 已提交
32
#include <linux/bit_spinlock.h>
L
Linus Torvalds 已提交
33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61

#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;
62
	struct buffer_head *first, *tmp;
L
Linus Torvalds 已提交
63
	struct page *page;
64
	struct inode *vi;
L
Linus Torvalds 已提交
65 66 67 68
	ntfs_inode *ni;
	int page_uptodate = 1;

	page = bh->b_page;
69 70
	vi = page->mapping->host;
	ni = NTFS_I(vi);
L
Linus Torvalds 已提交
71 72

	if (likely(uptodate)) {
73 74
		loff_t i_size;
		s64 file_ofs, init_size;
L
Linus Torvalds 已提交
75 76 77 78 79

		set_buffer_uptodate(bh);

		file_ofs = ((s64)page->index << PAGE_CACHE_SHIFT) +
				bh_offset(bh);
80
		read_lock_irqsave(&ni->size_lock, flags);
81 82
		init_size = ni->initialized_size;
		i_size = i_size_read(vi);
83
		read_unlock_irqrestore(&ni->size_lock, flags);
84 85 86 87
		if (unlikely(init_size > i_size)) {
			/* Race with shrinking truncate. */
			init_size = i_size;
		}
L
Linus Torvalds 已提交
88
		/* Check for the current buffer head overflowing. */
89 90
		if (unlikely(file_ofs + bh->b_size > init_size)) {
			int ofs;
91
			void *kaddr;
92 93 94 95

			ofs = 0;
			if (file_ofs < init_size)
				ofs = init_size - file_ofs;
96
			local_irq_save(flags);
97
			kaddr = kmap_atomic(page);
98 99 100
			memset(kaddr + bh_offset(bh) + ofs, 0,
					bh->b_size - ofs);
			flush_dcache_page(page);
101
			kunmap_atomic(kaddr);
102
			local_irq_restore(flags);
L
Linus Torvalds 已提交
103 104 105
		}
	} else {
		clear_buffer_uptodate(bh);
106
		SetPageError(page);
107 108
		ntfs_error(ni->vol->sb, "Buffer I/O error, logical block "
				"0x%llx.", (unsigned long long)bh->b_blocknr);
L
Linus Torvalds 已提交
109
	}
110 111 112
	first = page_buffers(page);
	local_irq_save(flags);
	bit_spin_lock(BH_Uptodate_Lock, &first->b_state);
L
Linus Torvalds 已提交
113 114 115 116 117 118 119 120 121 122 123 124 125 126
	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);
127 128
	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
	local_irq_restore(flags);
L
Linus Torvalds 已提交
129 130 131 132 133 134 135 136 137 138 139 140
	/*
	 * 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 {
141
		u8 *kaddr;
L
Linus Torvalds 已提交
142 143 144 145 146 147 148
		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);
149
		local_irq_save(flags);
150
		kaddr = kmap_atomic(page);
L
Linus Torvalds 已提交
151
		for (i = 0; i < recs; i++)
152
			post_read_mst_fixup((NTFS_RECORD*)(kaddr +
L
Linus Torvalds 已提交
153
					i * rec_size), rec_size);
154
		kunmap_atomic(kaddr);
155
		local_irq_restore(flags);
L
Linus Torvalds 已提交
156
		flush_dcache_page(page);
157
		if (likely(page_uptodate && !PageError(page)))
L
Linus Torvalds 已提交
158 159 160 161 162
			SetPageUptodate(page);
	}
	unlock_page(page);
	return;
still_busy:
163 164
	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
	local_irq_restore(flags);
L
Linus Torvalds 已提交
165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186
	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)
{
187
	loff_t i_size;
L
Linus Torvalds 已提交
188 189
	VCN vcn;
	LCN lcn;
190 191
	s64 init_size;
	struct inode *vi;
L
Linus Torvalds 已提交
192 193 194 195 196
	ntfs_inode *ni;
	ntfs_volume *vol;
	runlist_element *rl;
	struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
	sector_t iblock, lblock, zblock;
197
	unsigned long flags;
L
Linus Torvalds 已提交
198 199 200 201
	unsigned int blocksize, vcn_ofs;
	int i, nr;
	unsigned char blocksize_bits;

202 203
	vi = page->mapping->host;
	ni = NTFS_I(vi);
L
Linus Torvalds 已提交
204 205 206 207 208
	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));

209 210
	blocksize = vol->sb->s_blocksize;
	blocksize_bits = vol->sb->s_blocksize_bits;
L
Linus Torvalds 已提交
211

212
	if (!page_has_buffers(page)) {
L
Linus Torvalds 已提交
213
		create_empty_buffers(page, blocksize, 0);
214 215 216 217
		if (unlikely(!page_has_buffers(page))) {
			unlock_page(page);
			return -ENOMEM;
		}
L
Linus Torvalds 已提交
218
	}
219 220
	bh = head = page_buffers(page);
	BUG_ON(!bh);
L
Linus Torvalds 已提交
221

222 223 224 225 226 227 228 229 230 231 232
	/*
	 * 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.
	 */
L
Linus Torvalds 已提交
233
	iblock = (s64)page->index << (PAGE_CACHE_SHIFT - blocksize_bits);
234
	read_lock_irqsave(&ni->size_lock, flags);
L
Linus Torvalds 已提交
235
	lblock = (ni->allocated_size + blocksize - 1) >> blocksize_bits;
236 237
	init_size = ni->initialized_size;
	i_size = i_size_read(vi);
238
	read_unlock_irqrestore(&ni->size_lock, flags);
239 240 241 242 243
	if (unlikely(init_size > i_size)) {
		/* Race with shrinking truncate. */
		init_size = i_size;
	}
	zblock = (init_size + blocksize - 1) >> blocksize_bits;
L
Linus Torvalds 已提交
244 245 246 247 248

	/* Loop through all the buffers in the page. */
	rl = NULL;
	nr = i = 0;
	do {
N
Nate Diller 已提交
249
		int err = 0;
L
Linus Torvalds 已提交
250 251 252 253 254 255 256 257 258 259

		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) {
260
			bool is_retry = false;
L
Linus Torvalds 已提交
261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297

			/* 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) {
298
				is_retry = true;
L
Linus Torvalds 已提交
299 300 301 302 303 304 305 306 307
				/*
				 * 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;
308 309
			} else if (!rl)
				up_read(&ni->runlist.lock);
310 311 312 313 314 315 316 317 318
			/*
			 * 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;
			}
L
Linus Torvalds 已提交
319
			/* Hard error, zero out region. */
320 321
			if (!err)
				err = -EIO;
L
Linus Torvalds 已提交
322 323 324 325 326 327
			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 "
328
					"(error code %i).", ni->mft_no,
L
Linus Torvalds 已提交
329 330
					ni->type, (unsigned long long)vcn,
					vcn_ofs, is_retry ? " even after "
331
					"retrying" : "", err);
L
Linus Torvalds 已提交
332 333 334 335 336 337 338 339 340 341
		}
		/*
		 * 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:
342
		zero_user(page, i * blocksize, blocksize);
343 344
		if (likely(!err))
			set_buffer_uptodate(bh);
L
Linus Torvalds 已提交
345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
	} 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)
{
401 402
	loff_t i_size;
	struct inode *vi;
L
Linus Torvalds 已提交
403
	ntfs_inode *ni, *base_ni;
404
	u8 *addr;
L
Linus Torvalds 已提交
405 406
	ntfs_attr_search_ctx *ctx;
	MFT_RECORD *mrec;
407
	unsigned long flags;
L
Linus Torvalds 已提交
408 409 410
	u32 attr_len;
	int err = 0;

411
retry_readpage:
L
Linus Torvalds 已提交
412
	BUG_ON(!PageLocked(page));
A
Anton Altaparmakov 已提交
413 414 415 416 417
	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_CACHE_SIZE - 1) >>
			PAGE_CACHE_SHIFT)) {
418
		zero_user(page, 0, PAGE_CACHE_SIZE);
A
Anton Altaparmakov 已提交
419 420 421
		ntfs_debug("Read outside i_size - truncated?");
		goto done;
	}
L
Linus Torvalds 已提交
422 423 424 425 426 427 428 429
	/*
	 * This can potentially happen because we clear PageUptodate() during
	 * ntfs_writepage() of MstProtected() attributes.
	 */
	if (PageUptodate(page)) {
		unlock_page(page);
		return 0;
	}
430
	ni = NTFS_I(vi);
431 432 433 434
	/*
	 * 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
435 436 437
	 * attribute is compressed/encrypted.  Note we need to check for
	 * AT_INDEX_ALLOCATION since this is the type of both directory and
	 * index inodes.
438
	 */
439
	if (ni->type != AT_INDEX_ALLOCATION) {
440 441 442 443 444 445 446 447 448 449 450 451 452
		/* 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);
		}
	}
L
Linus Torvalds 已提交
453 454
	/* NInoNonResident() == NInoIndexAllocPresent() */
	if (NInoNonResident(ni)) {
455
		/* Normal, non-resident data stream. */
L
Linus Torvalds 已提交
456 457 458 459 460 461
		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
462 463 464
	 * 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.
L
Linus Torvalds 已提交
465
	 */
466
	if (unlikely(page->index > 0)) {
467
		zero_user(page, 0, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
468 469 470 471 472 473 474 475 476 477 478 479
		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;
	}
480 481 482 483 484 485 486 487
	/*
	 * 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;
	}
L
Linus Torvalds 已提交
488 489 490 491 492 493 494 495 496 497
	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);
498 499 500
	read_lock_irqsave(&ni->size_lock, flags);
	if (unlikely(attr_len > ni->initialized_size))
		attr_len = ni->initialized_size;
501
	i_size = i_size_read(vi);
502
	read_unlock_irqrestore(&ni->size_lock, flags);
503 504 505 506
	if (unlikely(attr_len > i_size)) {
		/* Race with shrinking truncate. */
		attr_len = i_size;
	}
507
	addr = kmap_atomic(page);
L
Linus Torvalds 已提交
508
	/* Copy the data to the page. */
509
	memcpy(addr, (u8*)ctx->attr +
L
Linus Torvalds 已提交
510 511 512
			le16_to_cpu(ctx->attr->data.resident.value_offset),
			attr_len);
	/* Zero the remainder of the page. */
513
	memset(addr + attr_len, 0, PAGE_CACHE_SIZE - attr_len);
L
Linus Torvalds 已提交
514
	flush_dcache_page(page);
515
	kunmap_atomic(addr);
L
Linus Torvalds 已提交
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
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;
555 556
	s64 initialized_size;
	loff_t i_size;
L
Linus Torvalds 已提交
557 558 559 560 561 562
	sector_t block, dblock, iblock;
	struct inode *vi;
	ntfs_inode *ni;
	ntfs_volume *vol;
	runlist_element *rl;
	struct buffer_head *bh, *head;
563
	unsigned long flags;
L
Linus Torvalds 已提交
564 565
	unsigned int blocksize, vcn_ofs;
	int err;
566
	bool need_end_writeback;
L
Linus Torvalds 已提交
567 568 569 570 571 572 573 574 575 576 577
	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));
578 579
	blocksize = vol->sb->s_blocksize;
	blocksize_bits = vol->sb->s_blocksize_bits;
L
Linus Torvalds 已提交
580 581 582 583
	if (!page_has_buffers(page)) {
		BUG_ON(!PageUptodate(page));
		create_empty_buffers(page, blocksize,
				(1 << BH_Uptodate) | (1 << BH_Dirty));
584 585 586 587 588 589 590 591 592 593 594 595
		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;
		}
L
Linus Torvalds 已提交
596 597
	}
	bh = head = page_buffers(page);
598
	BUG_ON(!bh);
L
Linus Torvalds 已提交
599 600 601 602 603 604

	/* NOTE: Different naming scheme to ntfs_read_block()! */

	/* The first block in the page. */
	block = (s64)page->index << (PAGE_CACHE_SHIFT - blocksize_bits);

605 606 607 608 609
	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);

L
Linus Torvalds 已提交
610
	/* The first out of bounds block for the data size. */
611
	dblock = (i_size + blocksize - 1) >> blocksize_bits;
L
Linus Torvalds 已提交
612 613

	/* The last (fully or partially) initialized block. */
614
	iblock = initialized_size >> blocksize_bits;
L
Linus Torvalds 已提交
615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633

	/*
	 * 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 {
634
		bool is_retry = false;
L
Linus Torvalds 已提交
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658

		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) &&
659
				(initialized_size < i_size))) {
L
Linus Torvalds 已提交
660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744
			/*
			 * 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) {
745 746 747 748
			u8 *kaddr;
			unsigned long *bpos, *bend;

			/* Check if the buffer is zero. */
749
			kaddr = kmap_atomic(page);
750 751 752 753 754 755
			bpos = (unsigned long *)(kaddr + bh_offset(bh));
			bend = (unsigned long *)((u8*)bpos + blocksize);
			do {
				if (unlikely(*bpos))
					break;
			} while (likely(++bpos < bend));
756
			kunmap_atomic(kaddr);
757 758 759 760 761 762 763 764 765
			if (bpos == bend) {
				/*
				 * Buffer is zero and sparse, no need to write
				 * it.
				 */
				bh->b_blocknr = -1;
				clear_buffer_dirty(bh);
				continue;
			}
L
Linus Torvalds 已提交
766 767 768 769 770 771 772 773 774 775
			// 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) {
776
			is_retry = true;
L
Linus Torvalds 已提交
777 778 779 780 781 782 783 784 785
			/*
			 * 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;
786 787
		} else if (!rl)
			up_read(&ni->runlist.lock);
788 789 790 791 792 793 794 795
		/*
		 * 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);
796
			zero_user(page, bh_offset(bh), blocksize);
797 798 799 800
			set_buffer_uptodate(bh);
			err = 0;
			continue;
		}
L
Linus Torvalds 已提交
801
		/* Failed to map the buffer, even after retrying. */
802 803
		if (!err)
			err = -EIO;
L
Linus Torvalds 已提交
804 805 806 807
		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 "
808
				"determined%s (error code %i).", ni->mft_no,
L
Linus Torvalds 已提交
809 810
				ni->type, (unsigned long long)vcn,
				vcn_ofs, is_retry ? " even after "
811
				"retrying" : "", err);
L
Linus Torvalds 已提交
812 813 814 815 816 817 818 819 820 821
		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;

822
	/* Just an optimization, so ->readpage() is not called later. */
L
Linus Torvalds 已提交
823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
	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. */

876
	/* Submit the prepared buffers for i/o. */
877
	need_end_writeback = true;
L
Linus Torvalds 已提交
878 879 880 881
	do {
		struct buffer_head *next = bh->b_this_page;
		if (buffer_async_write(bh)) {
			submit_bh(WRITE, bh);
882
			need_end_writeback = false;
L
Linus Torvalds 已提交
883 884 885
		}
		bh = next;
	} while (bh != head);
886
	unlock_page(page);
L
Linus Torvalds 已提交
887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930

	/* 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_CACHE_SIZE / rec_size];
	struct buffer_head *bh, *head, *tbh, *rec_start_bh;
931
	struct buffer_head *bhs[MAX_BUF_PER_PAGE];
L
Linus Torvalds 已提交
932
	runlist_element *rl;
933 934
	int i, nr_locked_nis, nr_recs, nr_bhs, max_bhs, bhs_per_rec, err, err2;
	unsigned bh_size, rec_size_bits;
935
	bool sync, is_mft, page_is_dirty, rec_is_dirty;
936
	unsigned char bh_size_bits;
L
Linus Torvalds 已提交
937 938 939 940 941 942 943 944 945 946 947 948 949 950

	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)));
951 952
	bh_size = vol->sb->s_blocksize;
	bh_size_bits = vol->sb->s_blocksize_bits;
953
	max_bhs = PAGE_CACHE_SIZE / bh_size;
L
Linus Torvalds 已提交
954
	BUG_ON(!max_bhs);
955
	BUG_ON(max_bhs > MAX_BUF_PER_PAGE);
L
Linus Torvalds 已提交
956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973

	/* 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_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. */
974
	dblock = (i_size_read(vi) + bh_size - 1) >> bh_size_bits;
L
Linus Torvalds 已提交
975 976 977

	rl = NULL;
	err = err2 = nr_bhs = nr_recs = nr_locked_nis = 0;
978
	page_is_dirty = rec_is_dirty = false;
L
Linus Torvalds 已提交
979 980
	rec_start_bh = NULL;
	do {
981
		bool is_retry = false;
L
Linus Torvalds 已提交
982 983 984 985

		if (likely(block < rec_block)) {
			if (unlikely(block >= dblock)) {
				clear_buffer_dirty(bh);
986
				set_buffer_uptodate(bh);
L
Linus Torvalds 已提交
987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
				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. */
1012
				rec_is_dirty = false;
L
Linus Torvalds 已提交
1013 1014
				continue;
			}
1015
			rec_is_dirty = true;
L
Linus Torvalds 已提交
1016 1017 1018 1019 1020 1021 1022 1023
			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;

1024
			bh->b_bdev = vol->sb->s_bdev;
L
Linus Torvalds 已提交
1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055
			/* 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) {
1056
					is_retry = true;
L
Linus Torvalds 已提交
1057 1058 1059 1060 1061 1062 1063 1064 1065
					/*
					 * 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)
1066
						page_is_dirty = true;
L
Linus Torvalds 已提交
1067
					lcn = err2;
1068
				} else {
L
Linus Torvalds 已提交
1069
					err2 = -EIO;
1070 1071 1072
					if (!rl)
						up_read(&ni->runlist.lock);
				}
L
Linus Torvalds 已提交
1073 1074 1075 1076 1077 1078 1079 1080 1081
				/* 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 "
1082 1083
						"code %lli).",
						(long long)block <<
L
Linus Torvalds 已提交
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
						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.
				 */
1148
				page_is_dirty = true;
L
Linus Torvalds 已提交
1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
				/*
				 * 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;
N
Nick Piggin 已提交
1198
		if (!trylock_buffer(tbh))
L
Linus Torvalds 已提交
1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
			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));
			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_CACHE_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. */
1281
		mutex_lock(&tni->extent_lock);
L
Linus Torvalds 已提交
1282 1283 1284 1285 1286 1287
		if (tni->nr_extents >= 0)
			base_tni = tni;
		else {
			base_tni = tni->ext.base_ntfs_ino;
			BUG_ON(!base_tni);
		}
1288
		mutex_unlock(&tni->extent_lock);
L
Linus Torvalds 已提交
1289 1290 1291
		ntfs_debug("Unlocking %s inode 0x%lx.",
				tni == base_tni ? "base" : "extent",
				tni->mft_no);
1292
		mutex_unlock(&tni->mrec_lock);
L
Linus Torvalds 已提交
1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356
		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_CACHE_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_CACHE_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;
1357 1358
	struct inode *vi = page->mapping->host;
	ntfs_inode *base_ni = NULL, *ni = NTFS_I(vi);
1359
	char *addr;
1360 1361
	ntfs_attr_search_ctx *ctx = NULL;
	MFT_RECORD *m = NULL;
L
Linus Torvalds 已提交
1362 1363 1364
	u32 attr_len;
	int err;

1365
retry_writepage:
L
Linus Torvalds 已提交
1366 1367 1368 1369 1370 1371 1372 1373 1374
	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_CACHE_SIZE - 1) >>
			PAGE_CACHE_SHIFT)) {
		/*
		 * The page may have dirty, unmapped buffers.  Make them
		 * freeable here, so the page does not leak.
		 */
1375
		block_invalidatepage(page, 0, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
1376 1377 1378 1379
		unlock_page(page);
		ntfs_debug("Write outside i_size - truncated?");
		return 0;
	}
1380 1381 1382 1383
	/*
	 * 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
1384 1385 1386
	 * attribute is compressed/encrypted.  Note we need to check for
	 * AT_INDEX_ALLOCATION since this is the type of both directory and
	 * index inodes.
1387
	 */
1388
	if (ni->type != AT_INDEX_ALLOCATION) {
1389 1390 1391 1392
		/* If file is encrypted, deny access, just like NT4. */
		if (NInoEncrypted(ni)) {
			unlock_page(page);
			BUG_ON(ni->type != AT_DATA);
1393
			ntfs_debug("Denying write access to encrypted file.");
1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414
			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;
		}
	}
L
Linus Torvalds 已提交
1415 1416 1417 1418 1419 1420
	/* NInoNonResident() == NInoIndexAllocPresent() */
	if (NInoNonResident(ni)) {
		/* We have to zero every time due to mmap-at-end-of-file. */
		if (page->index >= (i_size >> PAGE_CACHE_SHIFT)) {
			/* The page straddles i_size. */
			unsigned int ofs = i_size & ~PAGE_CACHE_MASK;
1421
			zero_user_segment(page, ofs, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
1422 1423 1424 1425
		}
		/* Handle mst protected attributes. */
		if (NInoMstProtected(ni))
			return ntfs_write_mst_block(page, wbc);
1426
		/* Normal, non-resident data stream. */
L
Linus Torvalds 已提交
1427 1428 1429
		return ntfs_write_block(page, wbc);
	}
	/*
1430 1431 1432 1433 1434 1435
	 * 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.
L
Linus Torvalds 已提交
1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459
	 */
	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;
	}
1460 1461 1462 1463 1464 1465 1466 1467
	/*
	 * 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;
	}
L
Linus Torvalds 已提交
1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484
	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);
1485
	i_size = i_size_read(vi);
L
Linus Torvalds 已提交
1486
	if (unlikely(attr_len > i_size)) {
1487
		/* Race with shrinking truncate or a failed truncate. */
L
Linus Torvalds 已提交
1488
		attr_len = i_size;
1489 1490 1491 1492 1493 1494 1495 1496
		/*
		 * 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);
L
Linus Torvalds 已提交
1497
	}
1498
	addr = kmap_atomic(page);
L
Linus Torvalds 已提交
1499 1500 1501
	/* Copy the data from the page to the mft record. */
	memcpy((u8*)ctx->attr +
			le16_to_cpu(ctx->attr->data.resident.value_offset),
1502
			addr, attr_len);
L
Linus Torvalds 已提交
1503
	/* Zero out of bounds area in the page cache page. */
1504
	memset(addr + attr_len, 0, PAGE_CACHE_SIZE - attr_len);
1505
	kunmap_atomic(addr);
1506
	flush_dcache_page(page);
1507
	flush_dcache_mft_record_page(ctx->ntfs_ino);
1508
	/* We are done with the page. */
L
Linus Torvalds 已提交
1509
	end_page_writeback(page);
1510
	/* Finally, mark the mft record dirty, so it gets written back. */
L
Linus Torvalds 已提交
1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526
	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 "
1527
				"error %i.", err);
L
Linus Torvalds 已提交
1528
		SetPageError(page);
1529
		NVolSetErrors(ni->vol);
L
Linus Torvalds 已提交
1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
	}
	unlock_page(page);
	if (ctx)
		ntfs_attr_put_search_ctx(ctx);
	if (m)
		unmap_mft_record(base_ni);
	return err;
}

#endif	/* NTFS_RW */

/**
 * ntfs_aops - general address space operations for inodes and attributes
 */
1544
const struct address_space_operations ntfs_aops = {
L
Linus Torvalds 已提交
1545 1546 1547 1548
	.readpage	= ntfs_readpage,	/* Fill page with data. */
#ifdef NTFS_RW
	.writepage	= ntfs_writepage,	/* Write dirty page to disk. */
#endif /* NTFS_RW */
1549 1550 1551
	.migratepage	= buffer_migrate_page,	/* Move a page cache page from
						   one physical page to an
						   other. */
1552
	.error_remove_page = generic_error_remove_page,
L
Linus Torvalds 已提交
1553 1554 1555 1556 1557 1558
};

/**
 * ntfs_mst_aops - general address space operations for mst protecteed inodes
 *		   and attributes
 */
1559
const struct address_space_operations ntfs_mst_aops = {
L
Linus Torvalds 已提交
1560 1561 1562 1563 1564 1565 1566
	.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 */
1567 1568 1569
	.migratepage	= buffer_migrate_page,	/* Move a page cache page from
						   one physical page to an
						   other. */
1570
	.error_remove_page = generic_error_remove_page,
L
Linus Torvalds 已提交
1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597
};

#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;
1598
	bh_size = VFS_I(ni)->i_sb->s_blocksize;
L
Linus Torvalds 已提交
1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
	spin_lock(&mapping->private_lock);
	if (unlikely(!page_has_buffers(page))) {
		spin_unlock(&mapping->private_lock);
		bh = head = alloc_page_buffers(page, bh_size, 1);
		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_buffers(page, head);
		} else
			buffers_to_free = bh;
	}
	bh = head = page_buffers(page);
1618
	BUG_ON(!bh);
L
Linus Torvalds 已提交
1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638
	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 */