xref: /openbmc/linux/fs/ntfs/logfile.c (revision fadbafc1)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * logfile.c - NTFS kernel journal handling. Part of the Linux-NTFS project.
4  *
5  * Copyright (c) 2002-2007 Anton Altaparmakov
6  */
7 
8 #ifdef NTFS_RW
9 
10 #include <linux/types.h>
11 #include <linux/fs.h>
12 #include <linux/highmem.h>
13 #include <linux/buffer_head.h>
14 #include <linux/bitops.h>
15 #include <linux/log2.h>
16 #include <linux/bio.h>
17 
18 #include "attrib.h"
19 #include "aops.h"
20 #include "debug.h"
21 #include "logfile.h"
22 #include "malloc.h"
23 #include "volume.h"
24 #include "ntfs.h"
25 
26 /**
27  * ntfs_check_restart_page_header - check the page header for consistency
28  * @vi:		$LogFile inode to which the restart page header belongs
29  * @rp:		restart page header to check
30  * @pos:	position in @vi at which the restart page header resides
31  *
32  * Check the restart page header @rp for consistency and return 'true' if it is
33  * consistent and 'false' otherwise.
34  *
35  * This function only needs NTFS_BLOCK_SIZE bytes in @rp, i.e. it does not
36  * require the full restart page.
37  */
38 static bool ntfs_check_restart_page_header(struct inode *vi,
39 		RESTART_PAGE_HEADER *rp, s64 pos)
40 {
41 	u32 logfile_system_page_size, logfile_log_page_size;
42 	u16 ra_ofs, usa_count, usa_ofs, usa_end = 0;
43 	bool have_usa = true;
44 
45 	ntfs_debug("Entering.");
46 	/*
47 	 * If the system or log page sizes are smaller than the ntfs block size
48 	 * or either is not a power of 2 we cannot handle this log file.
49 	 */
50 	logfile_system_page_size = le32_to_cpu(rp->system_page_size);
51 	logfile_log_page_size = le32_to_cpu(rp->log_page_size);
52 	if (logfile_system_page_size < NTFS_BLOCK_SIZE ||
53 			logfile_log_page_size < NTFS_BLOCK_SIZE ||
54 			logfile_system_page_size &
55 			(logfile_system_page_size - 1) ||
56 			!is_power_of_2(logfile_log_page_size)) {
57 		ntfs_error(vi->i_sb, "$LogFile uses unsupported page size.");
58 		return false;
59 	}
60 	/*
61 	 * We must be either at !pos (1st restart page) or at pos = system page
62 	 * size (2nd restart page).
63 	 */
64 	if (pos && pos != logfile_system_page_size) {
65 		ntfs_error(vi->i_sb, "Found restart area in incorrect "
66 				"position in $LogFile.");
67 		return false;
68 	}
69 	/* We only know how to handle version 1.1. */
70 	if (sle16_to_cpu(rp->major_ver) != 1 ||
71 			sle16_to_cpu(rp->minor_ver) != 1) {
72 		ntfs_error(vi->i_sb, "$LogFile version %i.%i is not "
73 				"supported.  (This driver supports version "
74 				"1.1 only.)", (int)sle16_to_cpu(rp->major_ver),
75 				(int)sle16_to_cpu(rp->minor_ver));
76 		return false;
77 	}
78 	/*
79 	 * If chkdsk has been run the restart page may not be protected by an
80 	 * update sequence array.
81 	 */
82 	if (ntfs_is_chkd_record(rp->magic) && !le16_to_cpu(rp->usa_count)) {
83 		have_usa = false;
84 		goto skip_usa_checks;
85 	}
86 	/* Verify the size of the update sequence array. */
87 	usa_count = 1 + (logfile_system_page_size >> NTFS_BLOCK_SIZE_BITS);
88 	if (usa_count != le16_to_cpu(rp->usa_count)) {
89 		ntfs_error(vi->i_sb, "$LogFile restart page specifies "
90 				"inconsistent update sequence array count.");
91 		return false;
92 	}
93 	/* Verify the position of the update sequence array. */
94 	usa_ofs = le16_to_cpu(rp->usa_ofs);
95 	usa_end = usa_ofs + usa_count * sizeof(u16);
96 	if (usa_ofs < sizeof(RESTART_PAGE_HEADER) ||
97 			usa_end > NTFS_BLOCK_SIZE - sizeof(u16)) {
98 		ntfs_error(vi->i_sb, "$LogFile restart page specifies "
99 				"inconsistent update sequence array offset.");
100 		return false;
101 	}
102 skip_usa_checks:
103 	/*
104 	 * Verify the position of the restart area.  It must be:
105 	 *	- aligned to 8-byte boundary,
106 	 *	- after the update sequence array, and
107 	 *	- within the system page size.
108 	 */
109 	ra_ofs = le16_to_cpu(rp->restart_area_offset);
110 	if (ra_ofs & 7 || (have_usa ? ra_ofs < usa_end :
111 			ra_ofs < sizeof(RESTART_PAGE_HEADER)) ||
112 			ra_ofs > logfile_system_page_size) {
113 		ntfs_error(vi->i_sb, "$LogFile restart page specifies "
114 				"inconsistent restart area offset.");
115 		return false;
116 	}
117 	/*
118 	 * Only restart pages modified by chkdsk are allowed to have chkdsk_lsn
119 	 * set.
120 	 */
121 	if (!ntfs_is_chkd_record(rp->magic) && sle64_to_cpu(rp->chkdsk_lsn)) {
122 		ntfs_error(vi->i_sb, "$LogFile restart page is not modified "
123 				"by chkdsk but a chkdsk LSN is specified.");
124 		return false;
125 	}
126 	ntfs_debug("Done.");
127 	return true;
128 }
129 
130 /**
131  * ntfs_check_restart_area - check the restart area for consistency
132  * @vi:		$LogFile inode to which the restart page belongs
133  * @rp:		restart page whose restart area to check
134  *
135  * Check the restart area of the restart page @rp for consistency and return
136  * 'true' if it is consistent and 'false' otherwise.
137  *
138  * This function assumes that the restart page header has already been
139  * consistency checked.
140  *
141  * This function only needs NTFS_BLOCK_SIZE bytes in @rp, i.e. it does not
142  * require the full restart page.
143  */
144 static bool ntfs_check_restart_area(struct inode *vi, RESTART_PAGE_HEADER *rp)
145 {
146 	u64 file_size;
147 	RESTART_AREA *ra;
148 	u16 ra_ofs, ra_len, ca_ofs;
149 	u8 fs_bits;
150 
151 	ntfs_debug("Entering.");
152 	ra_ofs = le16_to_cpu(rp->restart_area_offset);
153 	ra = (RESTART_AREA*)((u8*)rp + ra_ofs);
154 	/*
155 	 * Everything before ra->file_size must be before the first word
156 	 * protected by an update sequence number.  This ensures that it is
157 	 * safe to access ra->client_array_offset.
158 	 */
159 	if (ra_ofs + offsetof(RESTART_AREA, file_size) >
160 			NTFS_BLOCK_SIZE - sizeof(u16)) {
161 		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
162 				"inconsistent file offset.");
163 		return false;
164 	}
165 	/*
166 	 * Now that we can access ra->client_array_offset, make sure everything
167 	 * up to the log client array is before the first word protected by an
168 	 * update sequence number.  This ensures we can access all of the
169 	 * restart area elements safely.  Also, the client array offset must be
170 	 * aligned to an 8-byte boundary.
171 	 */
172 	ca_ofs = le16_to_cpu(ra->client_array_offset);
173 	if (((ca_ofs + 7) & ~7) != ca_ofs ||
174 			ra_ofs + ca_ofs > NTFS_BLOCK_SIZE - sizeof(u16)) {
175 		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
176 				"inconsistent client array offset.");
177 		return false;
178 	}
179 	/*
180 	 * The restart area must end within the system page size both when
181 	 * calculated manually and as specified by ra->restart_area_length.
182 	 * Also, the calculated length must not exceed the specified length.
183 	 */
184 	ra_len = ca_ofs + le16_to_cpu(ra->log_clients) *
185 			sizeof(LOG_CLIENT_RECORD);
186 	if (ra_ofs + ra_len > le32_to_cpu(rp->system_page_size) ||
187 			ra_ofs + le16_to_cpu(ra->restart_area_length) >
188 			le32_to_cpu(rp->system_page_size) ||
189 			ra_len > le16_to_cpu(ra->restart_area_length)) {
190 		ntfs_error(vi->i_sb, "$LogFile restart area is out of bounds "
191 				"of the system page size specified by the "
192 				"restart page header and/or the specified "
193 				"restart area length is inconsistent.");
194 		return false;
195 	}
196 	/*
197 	 * The ra->client_free_list and ra->client_in_use_list must be either
198 	 * LOGFILE_NO_CLIENT or less than ra->log_clients or they are
199 	 * overflowing the client array.
200 	 */
201 	if ((ra->client_free_list != LOGFILE_NO_CLIENT &&
202 			le16_to_cpu(ra->client_free_list) >=
203 			le16_to_cpu(ra->log_clients)) ||
204 			(ra->client_in_use_list != LOGFILE_NO_CLIENT &&
205 			le16_to_cpu(ra->client_in_use_list) >=
206 			le16_to_cpu(ra->log_clients))) {
207 		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
208 				"overflowing client free and/or in use lists.");
209 		return false;
210 	}
211 	/*
212 	 * Check ra->seq_number_bits against ra->file_size for consistency.
213 	 * We cannot just use ffs() because the file size is not a power of 2.
214 	 */
215 	file_size = (u64)sle64_to_cpu(ra->file_size);
216 	fs_bits = 0;
217 	while (file_size) {
218 		file_size >>= 1;
219 		fs_bits++;
220 	}
221 	if (le32_to_cpu(ra->seq_number_bits) != 67 - fs_bits) {
222 		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
223 				"inconsistent sequence number bits.");
224 		return false;
225 	}
226 	/* The log record header length must be a multiple of 8. */
227 	if (((le16_to_cpu(ra->log_record_header_length) + 7) & ~7) !=
228 			le16_to_cpu(ra->log_record_header_length)) {
229 		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
230 				"inconsistent log record header length.");
231 		return false;
232 	}
233 	/* Dito for the log page data offset. */
234 	if (((le16_to_cpu(ra->log_page_data_offset) + 7) & ~7) !=
235 			le16_to_cpu(ra->log_page_data_offset)) {
236 		ntfs_error(vi->i_sb, "$LogFile restart area specifies "
237 				"inconsistent log page data offset.");
238 		return false;
239 	}
240 	ntfs_debug("Done.");
241 	return true;
242 }
243 
244 /**
245  * ntfs_check_log_client_array - check the log client array for consistency
246  * @vi:		$LogFile inode to which the restart page belongs
247  * @rp:		restart page whose log client array to check
248  *
249  * Check the log client array of the restart page @rp for consistency and
250  * return 'true' if it is consistent and 'false' otherwise.
251  *
252  * This function assumes that the restart page header and the restart area have
253  * already been consistency checked.
254  *
255  * Unlike ntfs_check_restart_page_header() and ntfs_check_restart_area(), this
256  * function needs @rp->system_page_size bytes in @rp, i.e. it requires the full
257  * restart page and the page must be multi sector transfer deprotected.
258  */
259 static bool ntfs_check_log_client_array(struct inode *vi,
260 		RESTART_PAGE_HEADER *rp)
261 {
262 	RESTART_AREA *ra;
263 	LOG_CLIENT_RECORD *ca, *cr;
264 	u16 nr_clients, idx;
265 	bool in_free_list, idx_is_first;
266 
267 	ntfs_debug("Entering.");
268 	ra = (RESTART_AREA*)((u8*)rp + le16_to_cpu(rp->restart_area_offset));
269 	ca = (LOG_CLIENT_RECORD*)((u8*)ra +
270 			le16_to_cpu(ra->client_array_offset));
271 	/*
272 	 * Check the ra->client_free_list first and then check the
273 	 * ra->client_in_use_list.  Check each of the log client records in
274 	 * each of the lists and check that the array does not overflow the
275 	 * ra->log_clients value.  Also keep track of the number of records
276 	 * visited as there cannot be more than ra->log_clients records and
277 	 * that way we detect eventual loops in within a list.
278 	 */
279 	nr_clients = le16_to_cpu(ra->log_clients);
280 	idx = le16_to_cpu(ra->client_free_list);
281 	in_free_list = true;
282 check_list:
283 	for (idx_is_first = true; idx != LOGFILE_NO_CLIENT_CPU; nr_clients--,
284 			idx = le16_to_cpu(cr->next_client)) {
285 		if (!nr_clients || idx >= le16_to_cpu(ra->log_clients))
286 			goto err_out;
287 		/* Set @cr to the current log client record. */
288 		cr = ca + idx;
289 		/* The first log client record must not have a prev_client. */
290 		if (idx_is_first) {
291 			if (cr->prev_client != LOGFILE_NO_CLIENT)
292 				goto err_out;
293 			idx_is_first = false;
294 		}
295 	}
296 	/* Switch to and check the in use list if we just did the free list. */
297 	if (in_free_list) {
298 		in_free_list = false;
299 		idx = le16_to_cpu(ra->client_in_use_list);
300 		goto check_list;
301 	}
302 	ntfs_debug("Done.");
303 	return true;
304 err_out:
305 	ntfs_error(vi->i_sb, "$LogFile log client array is corrupt.");
306 	return false;
307 }
308 
309 /**
310  * ntfs_check_and_load_restart_page - check the restart page for consistency
311  * @vi:		$LogFile inode to which the restart page belongs
312  * @rp:		restart page to check
313  * @pos:	position in @vi at which the restart page resides
314  * @wrp:	[OUT] copy of the multi sector transfer deprotected restart page
315  * @lsn:	[OUT] set to the current logfile lsn on success
316  *
317  * Check the restart page @rp for consistency and return 0 if it is consistent
318  * and -errno otherwise.  The restart page may have been modified by chkdsk in
319  * which case its magic is CHKD instead of RSTR.
320  *
321  * This function only needs NTFS_BLOCK_SIZE bytes in @rp, i.e. it does not
322  * require the full restart page.
323  *
324  * If @wrp is not NULL, on success, *@wrp will point to a buffer containing a
325  * copy of the complete multi sector transfer deprotected page.  On failure,
326  * *@wrp is undefined.
327  *
328  * Simillarly, if @lsn is not NULL, on success *@lsn will be set to the current
329  * logfile lsn according to this restart page.  On failure, *@lsn is undefined.
330  *
331  * The following error codes are defined:
332  *	-EINVAL	- The restart page is inconsistent.
333  *	-ENOMEM	- Not enough memory to load the restart page.
334  *	-EIO	- Failed to reading from $LogFile.
335  */
336 static int ntfs_check_and_load_restart_page(struct inode *vi,
337 		RESTART_PAGE_HEADER *rp, s64 pos, RESTART_PAGE_HEADER **wrp,
338 		LSN *lsn)
339 {
340 	RESTART_AREA *ra;
341 	RESTART_PAGE_HEADER *trp;
342 	int size, err;
343 
344 	ntfs_debug("Entering.");
345 	/* Check the restart page header for consistency. */
346 	if (!ntfs_check_restart_page_header(vi, rp, pos)) {
347 		/* Error output already done inside the function. */
348 		return -EINVAL;
349 	}
350 	/* Check the restart area for consistency. */
351 	if (!ntfs_check_restart_area(vi, rp)) {
352 		/* Error output already done inside the function. */
353 		return -EINVAL;
354 	}
355 	ra = (RESTART_AREA*)((u8*)rp + le16_to_cpu(rp->restart_area_offset));
356 	/*
357 	 * Allocate a buffer to store the whole restart page so we can multi
358 	 * sector transfer deprotect it.
359 	 */
360 	trp = ntfs_malloc_nofs(le32_to_cpu(rp->system_page_size));
361 	if (!trp) {
362 		ntfs_error(vi->i_sb, "Failed to allocate memory for $LogFile "
363 				"restart page buffer.");
364 		return -ENOMEM;
365 	}
366 	/*
367 	 * Read the whole of the restart page into the buffer.  If it fits
368 	 * completely inside @rp, just copy it from there.  Otherwise map all
369 	 * the required pages and copy the data from them.
370 	 */
371 	size = PAGE_SIZE - (pos & ~PAGE_MASK);
372 	if (size >= le32_to_cpu(rp->system_page_size)) {
373 		memcpy(trp, rp, le32_to_cpu(rp->system_page_size));
374 	} else {
375 		pgoff_t idx;
376 		struct page *page;
377 		int have_read, to_read;
378 
379 		/* First copy what we already have in @rp. */
380 		memcpy(trp, rp, size);
381 		/* Copy the remaining data one page at a time. */
382 		have_read = size;
383 		to_read = le32_to_cpu(rp->system_page_size) - size;
384 		idx = (pos + size) >> PAGE_SHIFT;
385 		BUG_ON((pos + size) & ~PAGE_MASK);
386 		do {
387 			page = ntfs_map_page(vi->i_mapping, idx);
388 			if (IS_ERR(page)) {
389 				ntfs_error(vi->i_sb, "Error mapping $LogFile "
390 						"page (index %lu).", idx);
391 				err = PTR_ERR(page);
392 				if (err != -EIO && err != -ENOMEM)
393 					err = -EIO;
394 				goto err_out;
395 			}
396 			size = min_t(int, to_read, PAGE_SIZE);
397 			memcpy((u8*)trp + have_read, page_address(page), size);
398 			ntfs_unmap_page(page);
399 			have_read += size;
400 			to_read -= size;
401 			idx++;
402 		} while (to_read > 0);
403 	}
404 	/*
405 	 * Perform the multi sector transfer deprotection on the buffer if the
406 	 * restart page is protected.
407 	 */
408 	if ((!ntfs_is_chkd_record(trp->magic) || le16_to_cpu(trp->usa_count))
409 			&& post_read_mst_fixup((NTFS_RECORD*)trp,
410 			le32_to_cpu(rp->system_page_size))) {
411 		/*
412 		 * A multi sector tranfer error was detected.  We only need to
413 		 * abort if the restart page contents exceed the multi sector
414 		 * transfer fixup of the first sector.
415 		 */
416 		if (le16_to_cpu(rp->restart_area_offset) +
417 				le16_to_cpu(ra->restart_area_length) >
418 				NTFS_BLOCK_SIZE - sizeof(u16)) {
419 			ntfs_error(vi->i_sb, "Multi sector transfer error "
420 					"detected in $LogFile restart page.");
421 			err = -EINVAL;
422 			goto err_out;
423 		}
424 	}
425 	/*
426 	 * If the restart page is modified by chkdsk or there are no active
427 	 * logfile clients, the logfile is consistent.  Otherwise, need to
428 	 * check the log client records for consistency, too.
429 	 */
430 	err = 0;
431 	if (ntfs_is_rstr_record(rp->magic) &&
432 			ra->client_in_use_list != LOGFILE_NO_CLIENT) {
433 		if (!ntfs_check_log_client_array(vi, trp)) {
434 			err = -EINVAL;
435 			goto err_out;
436 		}
437 	}
438 	if (lsn) {
439 		if (ntfs_is_rstr_record(rp->magic))
440 			*lsn = sle64_to_cpu(ra->current_lsn);
441 		else /* if (ntfs_is_chkd_record(rp->magic)) */
442 			*lsn = sle64_to_cpu(rp->chkdsk_lsn);
443 	}
444 	ntfs_debug("Done.");
445 	if (wrp)
446 		*wrp = trp;
447 	else {
448 err_out:
449 		ntfs_free(trp);
450 	}
451 	return err;
452 }
453 
454 /**
455  * ntfs_check_logfile - check the journal for consistency
456  * @log_vi:	struct inode of loaded journal $LogFile to check
457  * @rp:		[OUT] on success this is a copy of the current restart page
458  *
459  * Check the $LogFile journal for consistency and return 'true' if it is
460  * consistent and 'false' if not.  On success, the current restart page is
461  * returned in *@rp.  Caller must call ntfs_free(*@rp) when finished with it.
462  *
463  * At present we only check the two restart pages and ignore the log record
464  * pages.
465  *
466  * Note that the MstProtected flag is not set on the $LogFile inode and hence
467  * when reading pages they are not deprotected.  This is because we do not know
468  * if the $LogFile was created on a system with a different page size to ours
469  * yet and mst deprotection would fail if our page size is smaller.
470  */
471 bool ntfs_check_logfile(struct inode *log_vi, RESTART_PAGE_HEADER **rp)
472 {
473 	s64 size, pos;
474 	LSN rstr1_lsn, rstr2_lsn;
475 	ntfs_volume *vol = NTFS_SB(log_vi->i_sb);
476 	struct address_space *mapping = log_vi->i_mapping;
477 	struct page *page = NULL;
478 	u8 *kaddr = NULL;
479 	RESTART_PAGE_HEADER *rstr1_ph = NULL;
480 	RESTART_PAGE_HEADER *rstr2_ph = NULL;
481 	int log_page_size, err;
482 	bool logfile_is_empty = true;
483 	u8 log_page_bits;
484 
485 	ntfs_debug("Entering.");
486 	/* An empty $LogFile must have been clean before it got emptied. */
487 	if (NVolLogFileEmpty(vol))
488 		goto is_empty;
489 	size = i_size_read(log_vi);
490 	/* Make sure the file doesn't exceed the maximum allowed size. */
491 	if (size > MaxLogFileSize)
492 		size = MaxLogFileSize;
493 	/*
494 	 * Truncate size to a multiple of the page cache size or the default
495 	 * log page size if the page cache size is between the default log page
496 	 * log page size if the page cache size is between the default log page
497 	 * size and twice that.
498 	 */
499 	if (PAGE_SIZE >= DefaultLogPageSize && PAGE_SIZE <=
500 			DefaultLogPageSize * 2)
501 		log_page_size = DefaultLogPageSize;
502 	else
503 		log_page_size = PAGE_SIZE;
504 	/*
505 	 * Use ntfs_ffs() instead of ffs() to enable the compiler to
506 	 * optimize log_page_size and log_page_bits into constants.
507 	 */
508 	log_page_bits = ntfs_ffs(log_page_size) - 1;
509 	size &= ~(s64)(log_page_size - 1);
510 	/*
511 	 * Ensure the log file is big enough to store at least the two restart
512 	 * pages and the minimum number of log record pages.
513 	 */
514 	if (size < log_page_size * 2 || (size - log_page_size * 2) >>
515 			log_page_bits < MinLogRecordPages) {
516 		ntfs_error(vol->sb, "$LogFile is too small.");
517 		return false;
518 	}
519 	/*
520 	 * Read through the file looking for a restart page.  Since the restart
521 	 * page header is at the beginning of a page we only need to search at
522 	 * what could be the beginning of a page (for each page size) rather
523 	 * than scanning the whole file byte by byte.  If all potential places
524 	 * contain empty and uninitialzed records, the log file can be assumed
525 	 * to be empty.
526 	 */
527 	for (pos = 0; pos < size; pos <<= 1) {
528 		pgoff_t idx = pos >> PAGE_SHIFT;
529 		if (!page || page->index != idx) {
530 			if (page)
531 				ntfs_unmap_page(page);
532 			page = ntfs_map_page(mapping, idx);
533 			if (IS_ERR(page)) {
534 				ntfs_error(vol->sb, "Error mapping $LogFile "
535 						"page (index %lu).", idx);
536 				goto err_out;
537 			}
538 		}
539 		kaddr = (u8*)page_address(page) + (pos & ~PAGE_MASK);
540 		/*
541 		 * A non-empty block means the logfile is not empty while an
542 		 * empty block after a non-empty block has been encountered
543 		 * means we are done.
544 		 */
545 		if (!ntfs_is_empty_recordp((le32*)kaddr))
546 			logfile_is_empty = false;
547 		else if (!logfile_is_empty)
548 			break;
549 		/*
550 		 * A log record page means there cannot be a restart page after
551 		 * this so no need to continue searching.
552 		 */
553 		if (ntfs_is_rcrd_recordp((le32*)kaddr))
554 			break;
555 		/* If not a (modified by chkdsk) restart page, continue. */
556 		if (!ntfs_is_rstr_recordp((le32*)kaddr) &&
557 				!ntfs_is_chkd_recordp((le32*)kaddr)) {
558 			if (!pos)
559 				pos = NTFS_BLOCK_SIZE >> 1;
560 			continue;
561 		}
562 		/*
563 		 * Check the (modified by chkdsk) restart page for consistency
564 		 * and get a copy of the complete multi sector transfer
565 		 * deprotected restart page.
566 		 */
567 		err = ntfs_check_and_load_restart_page(log_vi,
568 				(RESTART_PAGE_HEADER*)kaddr, pos,
569 				!rstr1_ph ? &rstr1_ph : &rstr2_ph,
570 				!rstr1_ph ? &rstr1_lsn : &rstr2_lsn);
571 		if (!err) {
572 			/*
573 			 * If we have now found the first (modified by chkdsk)
574 			 * restart page, continue looking for the second one.
575 			 */
576 			if (!pos) {
577 				pos = NTFS_BLOCK_SIZE >> 1;
578 				continue;
579 			}
580 			/*
581 			 * We have now found the second (modified by chkdsk)
582 			 * restart page, so we can stop looking.
583 			 */
584 			break;
585 		}
586 		/*
587 		 * Error output already done inside the function.  Note, we do
588 		 * not abort if the restart page was invalid as we might still
589 		 * find a valid one further in the file.
590 		 */
591 		if (err != -EINVAL) {
592 			ntfs_unmap_page(page);
593 			goto err_out;
594 		}
595 		/* Continue looking. */
596 		if (!pos)
597 			pos = NTFS_BLOCK_SIZE >> 1;
598 	}
599 	if (page)
600 		ntfs_unmap_page(page);
601 	if (logfile_is_empty) {
602 		NVolSetLogFileEmpty(vol);
603 is_empty:
604 		ntfs_debug("Done.  ($LogFile is empty.)");
605 		return true;
606 	}
607 	if (!rstr1_ph) {
608 		BUG_ON(rstr2_ph);
609 		ntfs_error(vol->sb, "Did not find any restart pages in "
610 				"$LogFile and it was not empty.");
611 		return false;
612 	}
613 	/* If both restart pages were found, use the more recent one. */
614 	if (rstr2_ph) {
615 		/*
616 		 * If the second restart area is more recent, switch to it.
617 		 * Otherwise just throw it away.
618 		 */
619 		if (rstr2_lsn > rstr1_lsn) {
620 			ntfs_debug("Using second restart page as it is more "
621 					"recent.");
622 			ntfs_free(rstr1_ph);
623 			rstr1_ph = rstr2_ph;
624 			/* rstr1_lsn = rstr2_lsn; */
625 		} else {
626 			ntfs_debug("Using first restart page as it is more "
627 					"recent.");
628 			ntfs_free(rstr2_ph);
629 		}
630 		rstr2_ph = NULL;
631 	}
632 	/* All consistency checks passed. */
633 	if (rp)
634 		*rp = rstr1_ph;
635 	else
636 		ntfs_free(rstr1_ph);
637 	ntfs_debug("Done.");
638 	return true;
639 err_out:
640 	if (rstr1_ph)
641 		ntfs_free(rstr1_ph);
642 	return false;
643 }
644 
645 /**
646  * ntfs_is_logfile_clean - check in the journal if the volume is clean
647  * @log_vi:	struct inode of loaded journal $LogFile to check
648  * @rp:		copy of the current restart page
649  *
650  * Analyze the $LogFile journal and return 'true' if it indicates the volume was
651  * shutdown cleanly and 'false' if not.
652  *
653  * At present we only look at the two restart pages and ignore the log record
654  * pages.  This is a little bit crude in that there will be a very small number
655  * of cases where we think that a volume is dirty when in fact it is clean.
656  * This should only affect volumes that have not been shutdown cleanly but did
657  * not have any pending, non-check-pointed i/o, i.e. they were completely idle
658  * at least for the five seconds preceding the unclean shutdown.
659  *
660  * This function assumes that the $LogFile journal has already been consistency
661  * checked by a call to ntfs_check_logfile() and in particular if the $LogFile
662  * is empty this function requires that NVolLogFileEmpty() is true otherwise an
663  * empty volume will be reported as dirty.
664  */
665 bool ntfs_is_logfile_clean(struct inode *log_vi, const RESTART_PAGE_HEADER *rp)
666 {
667 	ntfs_volume *vol = NTFS_SB(log_vi->i_sb);
668 	RESTART_AREA *ra;
669 
670 	ntfs_debug("Entering.");
671 	/* An empty $LogFile must have been clean before it got emptied. */
672 	if (NVolLogFileEmpty(vol)) {
673 		ntfs_debug("Done.  ($LogFile is empty.)");
674 		return true;
675 	}
676 	BUG_ON(!rp);
677 	if (!ntfs_is_rstr_record(rp->magic) &&
678 			!ntfs_is_chkd_record(rp->magic)) {
679 		ntfs_error(vol->sb, "Restart page buffer is invalid.  This is "
680 				"probably a bug in that the $LogFile should "
681 				"have been consistency checked before calling "
682 				"this function.");
683 		return false;
684 	}
685 	ra = (RESTART_AREA*)((u8*)rp + le16_to_cpu(rp->restart_area_offset));
686 	/*
687 	 * If the $LogFile has active clients, i.e. it is open, and we do not
688 	 * have the RESTART_VOLUME_IS_CLEAN bit set in the restart area flags,
689 	 * we assume there was an unclean shutdown.
690 	 */
691 	if (ra->client_in_use_list != LOGFILE_NO_CLIENT &&
692 			!(ra->flags & RESTART_VOLUME_IS_CLEAN)) {
693 		ntfs_debug("Done.  $LogFile indicates a dirty shutdown.");
694 		return false;
695 	}
696 	/* $LogFile indicates a clean shutdown. */
697 	ntfs_debug("Done.  $LogFile indicates a clean shutdown.");
698 	return true;
699 }
700 
701 /**
702  * ntfs_empty_logfile - empty the contents of the $LogFile journal
703  * @log_vi:	struct inode of loaded journal $LogFile to empty
704  *
705  * Empty the contents of the $LogFile journal @log_vi and return 'true' on
706  * success and 'false' on error.
707  *
708  * This function assumes that the $LogFile journal has already been consistency
709  * checked by a call to ntfs_check_logfile() and that ntfs_is_logfile_clean()
710  * has been used to ensure that the $LogFile is clean.
711  */
712 bool ntfs_empty_logfile(struct inode *log_vi)
713 {
714 	VCN vcn, end_vcn;
715 	ntfs_inode *log_ni = NTFS_I(log_vi);
716 	ntfs_volume *vol = log_ni->vol;
717 	struct super_block *sb = vol->sb;
718 	runlist_element *rl;
719 	unsigned long flags;
720 	unsigned block_size, block_size_bits;
721 	int err;
722 	bool should_wait = true;
723 
724 	ntfs_debug("Entering.");
725 	if (NVolLogFileEmpty(vol)) {
726 		ntfs_debug("Done.");
727 		return true;
728 	}
729 	/*
730 	 * We cannot use ntfs_attr_set() because we may be still in the middle
731 	 * of a mount operation.  Thus we do the emptying by hand by first
732 	 * zapping the page cache pages for the $LogFile/$DATA attribute and
733 	 * then emptying each of the buffers in each of the clusters specified
734 	 * by the runlist by hand.
735 	 */
736 	block_size = sb->s_blocksize;
737 	block_size_bits = sb->s_blocksize_bits;
738 	vcn = 0;
739 	read_lock_irqsave(&log_ni->size_lock, flags);
740 	end_vcn = (log_ni->initialized_size + vol->cluster_size_mask) >>
741 			vol->cluster_size_bits;
742 	read_unlock_irqrestore(&log_ni->size_lock, flags);
743 	truncate_inode_pages(log_vi->i_mapping, 0);
744 	down_write(&log_ni->runlist.lock);
745 	rl = log_ni->runlist.rl;
746 	if (unlikely(!rl || vcn < rl->vcn || !rl->length)) {
747 map_vcn:
748 		err = ntfs_map_runlist_nolock(log_ni, vcn, NULL);
749 		if (err) {
750 			ntfs_error(sb, "Failed to map runlist fragment (error "
751 					"%d).", -err);
752 			goto err;
753 		}
754 		rl = log_ni->runlist.rl;
755 		BUG_ON(!rl || vcn < rl->vcn || !rl->length);
756 	}
757 	/* Seek to the runlist element containing @vcn. */
758 	while (rl->length && vcn >= rl[1].vcn)
759 		rl++;
760 	do {
761 		LCN lcn;
762 		sector_t block, end_block;
763 		s64 len;
764 
765 		/*
766 		 * If this run is not mapped map it now and start again as the
767 		 * runlist will have been updated.
768 		 */
769 		lcn = rl->lcn;
770 		if (unlikely(lcn == LCN_RL_NOT_MAPPED)) {
771 			vcn = rl->vcn;
772 			goto map_vcn;
773 		}
774 		/* If this run is not valid abort with an error. */
775 		if (unlikely(!rl->length || lcn < LCN_HOLE))
776 			goto rl_err;
777 		/* Skip holes. */
778 		if (lcn == LCN_HOLE)
779 			continue;
780 		block = lcn << vol->cluster_size_bits >> block_size_bits;
781 		len = rl->length;
782 		if (rl[1].vcn > end_vcn)
783 			len = end_vcn - rl->vcn;
784 		end_block = (lcn + len) << vol->cluster_size_bits >>
785 				block_size_bits;
786 		/* Iterate over the blocks in the run and empty them. */
787 		do {
788 			struct buffer_head *bh;
789 
790 			/* Obtain the buffer, possibly not uptodate. */
791 			bh = sb_getblk(sb, block);
792 			BUG_ON(!bh);
793 			/* Setup buffer i/o submission. */
794 			lock_buffer(bh);
795 			bh->b_end_io = end_buffer_write_sync;
796 			get_bh(bh);
797 			/* Set the entire contents of the buffer to 0xff. */
798 			memset(bh->b_data, -1, block_size);
799 			if (!buffer_uptodate(bh))
800 				set_buffer_uptodate(bh);
801 			if (buffer_dirty(bh))
802 				clear_buffer_dirty(bh);
803 			/*
804 			 * Submit the buffer and wait for i/o to complete but
805 			 * only for the first buffer so we do not miss really
806 			 * serious i/o errors.  Once the first buffer has
807 			 * completed ignore errors afterwards as we can assume
808 			 * that if one buffer worked all of them will work.
809 			 */
810 			submit_bh(REQ_OP_WRITE, bh);
811 			if (should_wait) {
812 				should_wait = false;
813 				wait_on_buffer(bh);
814 				if (unlikely(!buffer_uptodate(bh)))
815 					goto io_err;
816 			}
817 			brelse(bh);
818 		} while (++block < end_block);
819 	} while ((++rl)->vcn < end_vcn);
820 	up_write(&log_ni->runlist.lock);
821 	/*
822 	 * Zap the pages again just in case any got instantiated whilst we were
823 	 * emptying the blocks by hand.  FIXME: We may not have completed
824 	 * writing to all the buffer heads yet so this may happen too early.
825 	 * We really should use a kernel thread to do the emptying
826 	 * asynchronously and then we can also set the volume dirty and output
827 	 * an error message if emptying should fail.
828 	 */
829 	truncate_inode_pages(log_vi->i_mapping, 0);
830 	/* Set the flag so we do not have to do it again on remount. */
831 	NVolSetLogFileEmpty(vol);
832 	ntfs_debug("Done.");
833 	return true;
834 io_err:
835 	ntfs_error(sb, "Failed to write buffer.  Unmount and run chkdsk.");
836 	goto dirty_err;
837 rl_err:
838 	ntfs_error(sb, "Runlist is corrupt.  Unmount and run chkdsk.");
839 dirty_err:
840 	NVolSetErrors(vol);
841 	err = -EIO;
842 err:
843 	up_write(&log_ni->runlist.lock);
844 	ntfs_error(sb, "Failed to fill $LogFile with 0xff bytes (error %d).",
845 			-err);
846 	return false;
847 }
848 
849 #endif /* NTFS_RW */
850