1 // SPDX-License-Identifier: GPL-2.0
4 * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
8 #include <linux/blkdev.h>
10 #include <linux/random.h>
11 #include <linux/slab.h>
23 #define MaxLogFileSize 0x100000000ull
24 #define DefaultLogPageSize 4096
25 #define MinLogRecordPages 0x30
28 struct NTFS_RECORD_HEADER rhdr; // 'RSTR'
29 __le32 sys_page_size; // 0x10: Page size of the system which initialized the log.
30 __le32 page_size; // 0x14: Log page size used for this log file.
31 __le16 ra_off; // 0x18:
32 __le16 minor_ver; // 0x1A:
33 __le16 major_ver; // 0x1C:
37 #define LFS_NO_CLIENT 0xffff
38 #define LFS_NO_CLIENT_LE cpu_to_le16(0xffff)
42 __le64 restart_lsn; // 0x08:
43 __le16 prev_client; // 0x10:
44 __le16 next_client; // 0x12:
45 __le16 seq_num; // 0x14:
47 __le32 name_bytes; // 0x1C: In bytes.
48 __le16 name[32]; // 0x20: Name of client.
51 static_assert(sizeof(struct CLIENT_REC) == 0x60);
53 /* Two copies of these will exist at the beginning of the log file */
55 __le64 current_lsn; // 0x00: Current logical end of log file.
56 __le16 log_clients; // 0x08: Maximum number of clients.
57 __le16 client_idx[2]; // 0x0A: Free/use index into the client record arrays.
58 __le16 flags; // 0x0E: See RESTART_SINGLE_PAGE_IO.
59 __le32 seq_num_bits; // 0x10: The number of bits in sequence number.
60 __le16 ra_len; // 0x14:
61 __le16 client_off; // 0x16:
62 __le64 l_size; // 0x18: Usable log file size.
63 __le32 last_lsn_data_len; // 0x20:
64 __le16 rec_hdr_len; // 0x24: Log page data offset.
65 __le16 data_off; // 0x26: Log page data length.
66 __le32 open_log_count; // 0x28:
67 __le32 align[5]; // 0x2C:
68 struct CLIENT_REC clients[]; // 0x40:
72 __le16 redo_op; // 0x00: NTFS_LOG_OPERATION
73 __le16 undo_op; // 0x02: NTFS_LOG_OPERATION
74 __le16 redo_off; // 0x04: Offset to Redo record.
75 __le16 redo_len; // 0x06: Redo length.
76 __le16 undo_off; // 0x08: Offset to Undo record.
77 __le16 undo_len; // 0x0A: Undo length.
78 __le16 target_attr; // 0x0C:
79 __le16 lcns_follow; // 0x0E:
80 __le16 record_off; // 0x10:
81 __le16 attr_off; // 0x12:
82 __le16 cluster_off; // 0x14:
83 __le16 reserved; // 0x16:
84 __le64 target_vcn; // 0x18:
85 __le64 page_lcns[]; // 0x20:
88 static_assert(sizeof(struct LOG_REC_HDR) == 0x20);
90 #define RESTART_ENTRY_ALLOCATED 0xFFFFFFFF
91 #define RESTART_ENTRY_ALLOCATED_LE cpu_to_le32(0xFFFFFFFF)
93 struct RESTART_TABLE {
94 __le16 size; // 0x00: In bytes
95 __le16 used; // 0x02: Entries
96 __le16 total; // 0x04: Entries
97 __le16 res[3]; // 0x06:
98 __le32 free_goal; // 0x0C:
99 __le32 first_free; // 0x10:
100 __le32 last_free; // 0x14:
104 static_assert(sizeof(struct RESTART_TABLE) == 0x18);
106 struct ATTR_NAME_ENTRY {
107 __le16 off; // Offset in the Open attribute Table.
112 struct OPEN_ATTR_ENRTY {
113 __le32 next; // 0x00: RESTART_ENTRY_ALLOCATED if allocated
114 __le32 bytes_per_index; // 0x04:
115 enum ATTR_TYPE type; // 0x08:
116 u8 is_dirty_pages; // 0x0C:
117 u8 is_attr_name; // 0x0B: Faked field to manage 'ptr'
118 u8 name_len; // 0x0C: Faked field to manage 'ptr'
120 struct MFT_REF ref; // 0x10: File Reference of file containing attribute
121 __le64 open_record_lsn; // 0x18:
125 /* 32 bit version of 'struct OPEN_ATTR_ENRTY' */
126 struct OPEN_ATTR_ENRTY_32 {
127 __le32 next; // 0x00: RESTART_ENTRY_ALLOCATED if allocated
129 struct MFT_REF ref; // 0x08:
130 __le64 open_record_lsn; // 0x10:
131 u8 is_dirty_pages; // 0x18:
132 u8 is_attr_name; // 0x19:
134 enum ATTR_TYPE type; // 0x1C:
135 u8 name_len; // 0x20: In wchar
137 __le32 AttributeName; // 0x24:
138 __le32 bytes_per_index; // 0x28:
141 #define SIZEOF_OPENATTRIBUTEENTRY0 0x2c
142 // static_assert( 0x2C == sizeof(struct OPEN_ATTR_ENRTY_32) );
143 static_assert(sizeof(struct OPEN_ATTR_ENRTY) < SIZEOF_OPENATTRIBUTEENTRY0);
146 * One entry exists in the Dirty Pages Table for each page which is dirty at
147 * the time the Restart Area is written.
149 struct DIR_PAGE_ENTRY {
150 __le32 next; // 0x00: RESTART_ENTRY_ALLOCATED if allocated
151 __le32 target_attr; // 0x04: Index into the Open attribute Table
152 __le32 transfer_len; // 0x08:
153 __le32 lcns_follow; // 0x0C:
154 __le64 vcn; // 0x10: Vcn of dirty page
155 __le64 oldest_lsn; // 0x18:
156 __le64 page_lcns[]; // 0x20:
159 static_assert(sizeof(struct DIR_PAGE_ENTRY) == 0x20);
161 /* 32 bit version of 'struct DIR_PAGE_ENTRY' */
162 struct DIR_PAGE_ENTRY_32 {
163 __le32 next; // 0x00: RESTART_ENTRY_ALLOCATED if allocated
164 __le32 target_attr; // 0x04: Index into the Open attribute Table
165 __le32 transfer_len; // 0x08:
166 __le32 lcns_follow; // 0x0C:
167 __le32 reserved; // 0x10:
168 __le32 vcn_low; // 0x14: Vcn of dirty page
169 __le32 vcn_hi; // 0x18: Vcn of dirty page
170 __le32 oldest_lsn_low; // 0x1C:
171 __le32 oldest_lsn_hi; // 0x1C:
172 __le32 page_lcns_low; // 0x24:
173 __le32 page_lcns_hi; // 0x24:
176 static_assert(offsetof(struct DIR_PAGE_ENTRY_32, vcn_low) == 0x14);
177 static_assert(sizeof(struct DIR_PAGE_ENTRY_32) == 0x2c);
179 enum transact_state {
180 TransactionUninitialized = 0,
186 struct TRANSACTION_ENTRY {
187 __le32 next; // 0x00: RESTART_ENTRY_ALLOCATED if allocated
188 u8 transact_state; // 0x04:
189 u8 reserved[3]; // 0x05:
190 __le64 first_lsn; // 0x08:
191 __le64 prev_lsn; // 0x10:
192 __le64 undo_next_lsn; // 0x18:
193 __le32 undo_records; // 0x20: Number of undo log records pending abort
194 __le32 undo_len; // 0x24: Total undo size
197 static_assert(sizeof(struct TRANSACTION_ENTRY) == 0x28);
199 struct NTFS_RESTART {
200 __le32 major_ver; // 0x00:
201 __le32 minor_ver; // 0x04:
202 __le64 check_point_start; // 0x08:
203 __le64 open_attr_table_lsn; // 0x10:
204 __le64 attr_names_lsn; // 0x18:
205 __le64 dirty_pages_table_lsn; // 0x20:
206 __le64 transact_table_lsn; // 0x28:
207 __le32 open_attr_len; // 0x30: In bytes
208 __le32 attr_names_len; // 0x34: In bytes
209 __le32 dirty_pages_len; // 0x38: In bytes
210 __le32 transact_table_len; // 0x3C: In bytes
213 static_assert(sizeof(struct NTFS_RESTART) == 0x40);
215 struct NEW_ATTRIBUTE_SIZES {
222 struct BITMAP_RANGE {
232 /* The following type defines the different log record types. */
233 #define LfsClientRecord cpu_to_le32(1)
234 #define LfsClientRestart cpu_to_le32(2)
236 /* This is used to uniquely identify a client for a particular log file. */
242 /* This is the header that begins every Log Record in the log file. */
243 struct LFS_RECORD_HDR {
244 __le64 this_lsn; // 0x00:
245 __le64 client_prev_lsn; // 0x08:
246 __le64 client_undo_next_lsn; // 0x10:
247 __le32 client_data_len; // 0x18:
248 struct CLIENT_ID client; // 0x1C: Owner of this log record.
249 __le32 record_type; // 0x20: LfsClientRecord or LfsClientRestart.
250 __le32 transact_id; // 0x24:
251 __le16 flags; // 0x28: LOG_RECORD_MULTI_PAGE
252 u8 align[6]; // 0x2A:
255 #define LOG_RECORD_MULTI_PAGE cpu_to_le16(1)
257 static_assert(sizeof(struct LFS_RECORD_HDR) == 0x30);
260 __le16 next_record_off; // 0x00: Offset of the free space in the page,
261 u8 align[6]; // 0x02:
262 __le64 last_end_lsn; // 0x08: lsn for the last log record which ends on the page,
265 static_assert(sizeof(struct LFS_RECORD) == 0x10);
267 struct RECORD_PAGE_HDR {
268 struct NTFS_RECORD_HEADER rhdr; // 'RCRD'
269 __le32 rflags; // 0x10: See LOG_PAGE_LOG_RECORD_END
270 __le16 page_count; // 0x14:
271 __le16 page_pos; // 0x16:
272 struct LFS_RECORD record_hdr; // 0x18:
273 __le16 fixups[10]; // 0x28:
274 __le32 file_off; // 0x3c: Used when major version >= 2
279 // Page contains the end of a log record.
280 #define LOG_PAGE_LOG_RECORD_END cpu_to_le32(0x00000001)
282 static inline bool is_log_record_end(const struct RECORD_PAGE_HDR *hdr)
284 return hdr->rflags & LOG_PAGE_LOG_RECORD_END;
287 static_assert(offsetof(struct RECORD_PAGE_HDR, file_off) == 0x3c);
290 * END of NTFS LOG structures
293 /* Define some tuning parameters to keep the restart tables a reasonable size. */
294 #define INITIAL_NUMBER_TRANSACTIONS 5
296 enum NTFS_LOG_OPERATION {
299 CompensationLogRecord = 0x01,
300 InitializeFileRecordSegment = 0x02,
301 DeallocateFileRecordSegment = 0x03,
302 WriteEndOfFileRecordSegment = 0x04,
303 CreateAttribute = 0x05,
304 DeleteAttribute = 0x06,
305 UpdateResidentValue = 0x07,
306 UpdateNonresidentValue = 0x08,
307 UpdateMappingPairs = 0x09,
308 DeleteDirtyClusters = 0x0A,
309 SetNewAttributeSizes = 0x0B,
310 AddIndexEntryRoot = 0x0C,
311 DeleteIndexEntryRoot = 0x0D,
312 AddIndexEntryAllocation = 0x0E,
313 DeleteIndexEntryAllocation = 0x0F,
314 WriteEndOfIndexBuffer = 0x10,
315 SetIndexEntryVcnRoot = 0x11,
316 SetIndexEntryVcnAllocation = 0x12,
317 UpdateFileNameRoot = 0x13,
318 UpdateFileNameAllocation = 0x14,
319 SetBitsInNonresidentBitMap = 0x15,
320 ClearBitsInNonresidentBitMap = 0x16,
322 EndTopLevelAction = 0x18,
323 PrepareTransaction = 0x19,
324 CommitTransaction = 0x1A,
325 ForgetTransaction = 0x1B,
326 OpenNonresidentAttribute = 0x1C,
327 OpenAttributeTableDump = 0x1D,
328 AttributeNamesDump = 0x1E,
329 DirtyPageTableDump = 0x1F,
330 TransactionTableDump = 0x20,
331 UpdateRecordDataRoot = 0x21,
332 UpdateRecordDataAllocation = 0x22,
334 UpdateRelativeDataInIndex =
335 0x23, // NtOfsRestartUpdateRelativeDataInIndex
336 UpdateRelativeDataInIndex2 = 0x24,
337 ZeroEndOfFileRecord = 0x25,
341 * Array for log records which require a target attribute.
342 * A true indicates that the corresponding restart operation
343 * requires a target attribute.
345 static const u8 AttributeRequired[] = {
346 0xFC, 0xFB, 0xFF, 0x10, 0x06,
349 static inline bool is_target_required(u16 op)
351 bool ret = op <= UpdateRecordDataAllocation &&
352 (AttributeRequired[op >> 3] >> (op & 7) & 1);
356 static inline bool can_skip_action(enum NTFS_LOG_OPERATION op)
360 case DeleteDirtyClusters:
362 case EndTopLevelAction:
363 case PrepareTransaction:
364 case CommitTransaction:
365 case ForgetTransaction:
366 case CompensationLogRecord:
367 case OpenNonresidentAttribute:
368 case OpenAttributeTableDump:
369 case AttributeNamesDump:
370 case DirtyPageTableDump:
371 case TransactionTableDump:
378 enum { lcb_ctx_undo_next, lcb_ctx_prev, lcb_ctx_next };
380 /* Bytes per restart table. */
381 static inline u32 bytes_per_rt(const struct RESTART_TABLE *rt)
383 return le16_to_cpu(rt->used) * le16_to_cpu(rt->size) +
384 sizeof(struct RESTART_TABLE);
387 /* Log record length. */
388 static inline u32 lrh_length(const struct LOG_REC_HDR *lr)
390 u16 t16 = le16_to_cpu(lr->lcns_follow);
392 return struct_size(lr, page_lcns, max_t(u16, 1, t16));
396 struct LFS_RECORD_HDR *lrh; // Log record header of the current lsn.
397 struct LOG_REC_HDR *log_rec;
398 u32 ctx_mode; // lcb_ctx_undo_next/lcb_ctx_prev/lcb_ctx_next
399 struct CLIENT_ID client;
400 bool alloc; // If true the we should deallocate 'log_rec'.
403 static void lcb_put(struct lcb *lcb)
411 /* Find the oldest lsn from active clients. */
412 static inline void oldest_client_lsn(const struct CLIENT_REC *ca,
413 __le16 next_client, u64 *oldest_lsn)
415 while (next_client != LFS_NO_CLIENT_LE) {
416 const struct CLIENT_REC *cr = ca + le16_to_cpu(next_client);
417 u64 lsn = le64_to_cpu(cr->oldest_lsn);
419 /* Ignore this block if it's oldest lsn is 0. */
420 if (lsn && lsn < *oldest_lsn)
423 next_client = cr->next_client;
427 static inline bool is_rst_page_hdr_valid(u32 file_off,
428 const struct RESTART_HDR *rhdr)
430 u32 sys_page = le32_to_cpu(rhdr->sys_page_size);
431 u32 page_size = le32_to_cpu(rhdr->page_size);
435 if (sys_page < SECTOR_SIZE || page_size < SECTOR_SIZE ||
436 sys_page & (sys_page - 1) || page_size & (page_size - 1)) {
440 /* Check that if the file offset isn't 0, it is the system page size. */
441 if (file_off && file_off != sys_page)
444 /* Check support version 1.1+. */
445 if (le16_to_cpu(rhdr->major_ver) <= 1 && !rhdr->minor_ver)
448 if (le16_to_cpu(rhdr->major_ver) > 2)
451 ro = le16_to_cpu(rhdr->ra_off);
452 if (!IS_ALIGNED(ro, 8) || ro > sys_page)
455 end_usa = ((sys_page >> SECTOR_SHIFT) + 1) * sizeof(short);
456 end_usa += le16_to_cpu(rhdr->rhdr.fix_off);
464 static inline bool is_rst_area_valid(const struct RESTART_HDR *rhdr)
466 const struct RESTART_AREA *ra;
468 u32 off, l_size, seq_bits;
469 u16 ro = le16_to_cpu(rhdr->ra_off);
470 u32 sys_page = le32_to_cpu(rhdr->sys_page_size);
472 if (ro + offsetof(struct RESTART_AREA, l_size) >
473 SECTOR_SIZE - sizeof(short))
476 ra = Add2Ptr(rhdr, ro);
477 cl = le16_to_cpu(ra->log_clients);
482 off = le16_to_cpu(ra->client_off);
484 if (!IS_ALIGNED(off, 8) || ro + off > SECTOR_SIZE - sizeof(short))
487 off += cl * sizeof(struct CLIENT_REC);
493 * Check the restart length field and whether the entire
494 * restart area is contained that length.
496 if (le16_to_cpu(rhdr->ra_off) + le16_to_cpu(ra->ra_len) > sys_page ||
497 off > le16_to_cpu(ra->ra_len)) {
502 * As a final check make sure that the use list and the free list
503 * are either empty or point to a valid client.
505 fl = le16_to_cpu(ra->client_idx[0]);
506 ul = le16_to_cpu(ra->client_idx[1]);
507 if ((fl != LFS_NO_CLIENT && fl >= cl) ||
508 (ul != LFS_NO_CLIENT && ul >= cl))
511 /* Make sure the sequence number bits match the log file size. */
512 l_size = le64_to_cpu(ra->l_size);
514 seq_bits = sizeof(u64) * 8 + 3;
520 if (seq_bits != ra->seq_num_bits)
523 /* The log page data offset and record header length must be quad-aligned. */
524 if (!IS_ALIGNED(le16_to_cpu(ra->data_off), 8) ||
525 !IS_ALIGNED(le16_to_cpu(ra->rec_hdr_len), 8))
531 static inline bool is_client_area_valid(const struct RESTART_HDR *rhdr,
534 u16 ro = le16_to_cpu(rhdr->ra_off);
535 const struct RESTART_AREA *ra = Add2Ptr(rhdr, ro);
536 u16 ra_len = le16_to_cpu(ra->ra_len);
537 const struct CLIENT_REC *ca;
540 if (usa_error && ra_len + ro > SECTOR_SIZE - sizeof(short))
543 /* Find the start of the client array. */
544 ca = Add2Ptr(ra, le16_to_cpu(ra->client_off));
547 * Start with the free list.
548 * Check that all the clients are valid and that there isn't a cycle.
549 * Do the in-use list on the second pass.
551 for (i = 0; i < 2; i++) {
552 u16 client_idx = le16_to_cpu(ra->client_idx[i]);
553 bool first_client = true;
554 u16 clients = le16_to_cpu(ra->log_clients);
556 while (client_idx != LFS_NO_CLIENT) {
557 const struct CLIENT_REC *cr;
560 client_idx >= le16_to_cpu(ra->log_clients))
564 cr = ca + client_idx;
566 client_idx = le16_to_cpu(cr->next_client);
569 first_client = false;
570 if (cr->prev_client != LFS_NO_CLIENT_LE)
582 * Remove a client record from a client record list an restart area.
584 static inline void remove_client(struct CLIENT_REC *ca,
585 const struct CLIENT_REC *cr, __le16 *head)
587 if (cr->prev_client == LFS_NO_CLIENT_LE)
588 *head = cr->next_client;
590 ca[le16_to_cpu(cr->prev_client)].next_client = cr->next_client;
592 if (cr->next_client != LFS_NO_CLIENT_LE)
593 ca[le16_to_cpu(cr->next_client)].prev_client = cr->prev_client;
597 * add_client - Add a client record to the start of a list.
599 static inline void add_client(struct CLIENT_REC *ca, u16 index, __le16 *head)
601 struct CLIENT_REC *cr = ca + index;
603 cr->prev_client = LFS_NO_CLIENT_LE;
604 cr->next_client = *head;
606 if (*head != LFS_NO_CLIENT_LE)
607 ca[le16_to_cpu(*head)].prev_client = cpu_to_le16(index);
609 *head = cpu_to_le16(index);
612 static inline void *enum_rstbl(struct RESTART_TABLE *t, void *c)
616 u16 rsize = t ? le16_to_cpu(t->size) : 0;
621 e = Add2Ptr(t, sizeof(struct RESTART_TABLE));
623 e = Add2Ptr(c, rsize);
626 /* Loop until we hit the first one allocated, or the end of the list. */
627 for (bprt = bytes_per_rt(t); PtrOffset(t, e) < bprt;
628 e = Add2Ptr(e, rsize)) {
629 if (*e == RESTART_ENTRY_ALLOCATED_LE)
636 * find_dp - Search for a @vcn in Dirty Page Table.
638 static inline struct DIR_PAGE_ENTRY *find_dp(struct RESTART_TABLE *dptbl,
639 u32 target_attr, u64 vcn)
641 __le32 ta = cpu_to_le32(target_attr);
642 struct DIR_PAGE_ENTRY *dp = NULL;
644 while ((dp = enum_rstbl(dptbl, dp))) {
645 u64 dp_vcn = le64_to_cpu(dp->vcn);
647 if (dp->target_attr == ta && vcn >= dp_vcn &&
648 vcn < dp_vcn + le32_to_cpu(dp->lcns_follow)) {
655 static inline u32 norm_file_page(u32 page_size, u32 *l_size, bool use_default)
658 page_size = DefaultLogPageSize;
660 /* Round the file size down to a system page boundary. */
661 *l_size &= ~(page_size - 1);
663 /* File should contain at least 2 restart pages and MinLogRecordPages pages. */
664 if (*l_size < (MinLogRecordPages + 2) * page_size)
670 static bool check_log_rec(const struct LOG_REC_HDR *lr, u32 bytes, u32 tr,
671 u32 bytes_per_attr_entry)
675 if (bytes < sizeof(struct LOG_REC_HDR))
680 if ((tr - sizeof(struct RESTART_TABLE)) %
681 sizeof(struct TRANSACTION_ENTRY))
684 if (le16_to_cpu(lr->redo_off) & 7)
687 if (le16_to_cpu(lr->undo_off) & 7)
693 if (is_target_required(le16_to_cpu(lr->redo_op)))
696 if (is_target_required(le16_to_cpu(lr->undo_op)))
700 if (!lr->lcns_follow)
703 t16 = le16_to_cpu(lr->target_attr);
704 if ((t16 - sizeof(struct RESTART_TABLE)) % bytes_per_attr_entry)
708 if (bytes < lrh_length(lr))
714 static bool check_rstbl(const struct RESTART_TABLE *rt, size_t bytes)
718 u16 rsize = le16_to_cpu(rt->size);
719 u16 ne = le16_to_cpu(rt->used);
720 u32 ff = le32_to_cpu(rt->first_free);
721 u32 lf = le32_to_cpu(rt->last_free);
723 ts = rsize * ne + sizeof(struct RESTART_TABLE);
725 if (!rsize || rsize > bytes ||
726 rsize + sizeof(struct RESTART_TABLE) > bytes || bytes < ts ||
727 le16_to_cpu(rt->total) > ne || ff > ts || lf > ts ||
728 (ff && ff < sizeof(struct RESTART_TABLE)) ||
729 (lf && lf < sizeof(struct RESTART_TABLE))) {
734 * Verify each entry is either allocated or points
735 * to a valid offset the table.
737 for (i = 0; i < ne; i++) {
738 off = le32_to_cpu(*(__le32 *)Add2Ptr(
739 rt, i * rsize + sizeof(struct RESTART_TABLE)));
741 if (off != RESTART_ENTRY_ALLOCATED && off &&
742 (off < sizeof(struct RESTART_TABLE) ||
743 ((off - sizeof(struct RESTART_TABLE)) % rsize))) {
749 * Walk through the list headed by the first entry to make
750 * sure none of the entries are currently being used.
752 for (off = ff; off;) {
753 if (off == RESTART_ENTRY_ALLOCATED)
756 off = le32_to_cpu(*(__le32 *)Add2Ptr(rt, off));
763 * free_rsttbl_idx - Free a previously allocated index a Restart Table.
765 static inline void free_rsttbl_idx(struct RESTART_TABLE *rt, u32 off)
768 u32 lf = le32_to_cpu(rt->last_free);
769 __le32 off_le = cpu_to_le32(off);
771 e = Add2Ptr(rt, off);
773 if (off < le32_to_cpu(rt->free_goal)) {
775 rt->first_free = off_le;
777 rt->last_free = off_le;
780 *(__le32 *)Add2Ptr(rt, lf) = off_le;
782 rt->first_free = off_le;
784 rt->last_free = off_le;
788 le16_sub_cpu(&rt->total, 1);
791 static inline struct RESTART_TABLE *init_rsttbl(u16 esize, u16 used)
793 __le32 *e, *last_free;
795 u32 bytes = esize * used + sizeof(struct RESTART_TABLE);
796 u32 lf = sizeof(struct RESTART_TABLE) + (used - 1) * esize;
797 struct RESTART_TABLE *t = kzalloc(bytes, GFP_NOFS);
802 t->size = cpu_to_le16(esize);
803 t->used = cpu_to_le16(used);
804 t->free_goal = cpu_to_le32(~0u);
805 t->first_free = cpu_to_le32(sizeof(struct RESTART_TABLE));
806 t->last_free = cpu_to_le32(lf);
808 e = (__le32 *)(t + 1);
809 last_free = Add2Ptr(t, lf);
811 for (off = sizeof(struct RESTART_TABLE) + esize; e < last_free;
812 e = Add2Ptr(e, esize), off += esize) {
813 *e = cpu_to_le32(off);
818 static inline struct RESTART_TABLE *extend_rsttbl(struct RESTART_TABLE *tbl,
819 u32 add, u32 free_goal)
821 u16 esize = le16_to_cpu(tbl->size);
822 __le32 osize = cpu_to_le32(bytes_per_rt(tbl));
823 u32 used = le16_to_cpu(tbl->used);
824 struct RESTART_TABLE *rt;
826 rt = init_rsttbl(esize, used + add);
830 memcpy(rt + 1, tbl + 1, esize * used);
832 rt->free_goal = free_goal == ~0u
834 : cpu_to_le32(sizeof(struct RESTART_TABLE) +
837 if (tbl->first_free) {
838 rt->first_free = tbl->first_free;
839 *(__le32 *)Add2Ptr(rt, le32_to_cpu(tbl->last_free)) = osize;
841 rt->first_free = osize;
844 rt->total = tbl->total;
853 * Allocate an index from within a previously initialized Restart Table.
855 static inline void *alloc_rsttbl_idx(struct RESTART_TABLE **tbl)
859 struct RESTART_TABLE *t = *tbl;
861 if (!t->first_free) {
862 *tbl = t = extend_rsttbl(t, 16, ~0u);
867 off = le32_to_cpu(t->first_free);
869 /* Dequeue this entry and zero it. */
874 memset(e, 0, le16_to_cpu(t->size));
876 *e = RESTART_ENTRY_ALLOCATED_LE;
878 /* If list is going empty, then we fix the last_free as well. */
882 le16_add_cpu(&t->total, 1);
884 return Add2Ptr(t, off);
888 * alloc_rsttbl_from_idx
890 * Allocate a specific index from within a previously initialized Restart Table.
892 static inline void *alloc_rsttbl_from_idx(struct RESTART_TABLE **tbl, u32 vbo)
896 struct RESTART_TABLE *rt = *tbl;
897 u32 bytes = bytes_per_rt(rt);
898 u16 esize = le16_to_cpu(rt->size);
900 /* If the entry is not the table, we will have to extend the table. */
903 * Extend the size by computing the number of entries between
904 * the existing size and the desired index and adding 1 to that.
906 u32 bytes2idx = vbo - bytes;
909 * There should always be an integral number of entries
910 * being added. Now extend the table.
912 *tbl = rt = extend_rsttbl(rt, bytes2idx / esize + 1, bytes);
917 /* See if the entry is already allocated, and just return if it is. */
918 e = Add2Ptr(rt, vbo);
920 if (*e == RESTART_ENTRY_ALLOCATED_LE)
924 * Walk through the table, looking for the entry we're
925 * interested and the previous entry.
927 off = le32_to_cpu(rt->first_free);
928 e = Add2Ptr(rt, off);
931 /* this is a match */
937 * Need to walk through the list looking for the predecessor
941 /* Remember the entry just found */
945 /* Should never run of entries. */
947 /* Lookup up the next entry the list. */
948 off = le32_to_cpu(*last_e);
949 e = Add2Ptr(rt, off);
951 /* If this is our match we are done. */
956 * If this was the last entry, we update that
959 if (le32_to_cpu(rt->last_free) == off)
960 rt->last_free = cpu_to_le32(last_off);
966 /* If the list is now empty, we fix the last_free as well. */
970 /* Zero this entry. */
972 *e = RESTART_ENTRY_ALLOCATED_LE;
974 le16_add_cpu(&rt->total, 1);
979 #define RESTART_SINGLE_PAGE_IO cpu_to_le16(0x0001)
981 #define NTFSLOG_WRAPPED 0x00000001
982 #define NTFSLOG_MULTIPLE_PAGE_IO 0x00000002
983 #define NTFSLOG_NO_LAST_LSN 0x00000004
984 #define NTFSLOG_REUSE_TAIL 0x00000010
985 #define NTFSLOG_NO_OLDEST_LSN 0x00000020
987 /* Helper struct to work with NTFS $LogFile. */
989 struct ntfs_inode *ni;
995 u32 page_mask; // page_size - 1
997 struct RECORD_PAGE_HDR *one_page_buf;
999 struct RESTART_TABLE *open_attr_tbl;
1009 u16 record_header_len;
1013 u32 seq_num_mask; /* (1 << file_data_bits) - 1 */
1015 struct RESTART_AREA *ra; /* In-memory image of the next restart area. */
1016 u32 ra_size; /* The usable size of the restart area. */
1019 * If true, then the in-memory restart area is to be written
1020 * to the first position on the disk.
1023 bool set_dirty; /* True if we need to set dirty flag. */
1031 u32 total_avail_pages;
1032 u32 total_undo_commit;
1033 u32 max_current_avail;
1040 u32 l_flags; /* See NTFSLOG_XXX */
1041 u32 current_openlog_count; /* On-disk value for open_log_count. */
1043 struct CLIENT_ID client_id;
1044 u32 client_undo_commit;
1047 static inline u32 lsn_to_vbo(struct ntfs_log *log, const u64 lsn)
1049 u32 vbo = (lsn << log->seq_num_bits) >> (log->seq_num_bits - 3);
1054 /* Compute the offset in the log file of the next log page. */
1055 static inline u32 next_page_off(struct ntfs_log *log, u32 off)
1057 off = (off & ~log->sys_page_mask) + log->page_size;
1058 return off >= log->l_size ? log->first_page : off;
1061 static inline u32 lsn_to_page_off(struct ntfs_log *log, u64 lsn)
1063 return (((u32)lsn) << 3) & log->page_mask;
1066 static inline u64 vbo_to_lsn(struct ntfs_log *log, u32 off, u64 Seq)
1068 return (off >> 3) + (Seq << log->file_data_bits);
1071 static inline bool is_lsn_in_file(struct ntfs_log *log, u64 lsn)
1073 return lsn >= log->oldest_lsn &&
1074 lsn <= le64_to_cpu(log->ra->current_lsn);
1077 static inline u32 hdr_file_off(struct ntfs_log *log,
1078 struct RECORD_PAGE_HDR *hdr)
1080 if (log->major_ver < 2)
1081 return le64_to_cpu(hdr->rhdr.lsn);
1083 return le32_to_cpu(hdr->file_off);
1086 static inline u64 base_lsn(struct ntfs_log *log,
1087 const struct RECORD_PAGE_HDR *hdr, u64 lsn)
1089 u64 h_lsn = le64_to_cpu(hdr->rhdr.lsn);
1090 u64 ret = (((h_lsn >> log->file_data_bits) +
1091 (lsn < (lsn_to_vbo(log, h_lsn) & ~log->page_mask) ? 1 : 0))
1092 << log->file_data_bits) +
1093 ((((is_log_record_end(hdr) &&
1094 h_lsn <= le64_to_cpu(hdr->record_hdr.last_end_lsn))
1095 ? le16_to_cpu(hdr->record_hdr.next_record_off)
1103 static inline bool verify_client_lsn(struct ntfs_log *log,
1104 const struct CLIENT_REC *client, u64 lsn)
1106 return lsn >= le64_to_cpu(client->oldest_lsn) &&
1107 lsn <= le64_to_cpu(log->ra->current_lsn) && lsn;
1110 struct restart_info {
1112 struct RESTART_HDR *r_page;
1114 bool chkdsk_was_run;
1120 static int read_log_page(struct ntfs_log *log, u32 vbo,
1121 struct RECORD_PAGE_HDR **buffer, bool *usa_error)
1124 u32 page_idx = vbo >> log->page_bits;
1125 u32 page_off = vbo & log->page_mask;
1126 u32 bytes = log->page_size - page_off;
1127 void *to_free = NULL;
1128 u32 page_vbo = page_idx << log->page_bits;
1129 struct RECORD_PAGE_HDR *page_buf;
1130 struct ntfs_inode *ni = log->ni;
1133 if (vbo >= log->l_size)
1137 to_free = kmalloc(log->page_size, GFP_NOFS);
1143 page_buf = page_off ? log->one_page_buf : *buffer;
1145 err = ntfs_read_run_nb(ni->mi.sbi, &ni->file.run, page_vbo, page_buf,
1146 log->page_size, NULL);
1150 if (page_buf->rhdr.sign != NTFS_FFFF_SIGNATURE)
1151 ntfs_fix_post_read(&page_buf->rhdr, PAGE_SIZE, false);
1153 if (page_buf != *buffer)
1154 memcpy(*buffer, Add2Ptr(page_buf, page_off), bytes);
1156 bBAAD = page_buf->rhdr.sign == NTFS_BAAD_SIGNATURE;
1160 /* Check that the update sequence array for this page is valid */
1161 /* If we don't allow errors, raise an error status */
1166 if (err && to_free) {
1177 * It walks through 512 blocks of the file looking for a valid
1178 * restart page header. It will stop the first time we find a
1179 * valid page header.
1181 static int log_read_rst(struct ntfs_log *log, u32 l_size, bool first,
1182 struct restart_info *info)
1185 struct RESTART_HDR *r_page = NULL;
1187 /* Determine which restart area we are looking for. */
1196 /* Loop continuously until we succeed. */
1197 for (; vbo < l_size; vbo = 2 * vbo + skip, skip = 0) {
1200 struct RESTART_AREA *ra;
1202 /* Read a page header at the current offset. */
1203 if (read_log_page(log, vbo, (struct RECORD_PAGE_HDR **)&r_page,
1205 /* Ignore any errors. */
1209 /* Exit if the signature is a log record page. */
1210 if (r_page->rhdr.sign == NTFS_RCRD_SIGNATURE) {
1211 info->initialized = true;
1215 brst = r_page->rhdr.sign == NTFS_RSTR_SIGNATURE;
1216 bchk = r_page->rhdr.sign == NTFS_CHKD_SIGNATURE;
1218 if (!bchk && !brst) {
1219 if (r_page->rhdr.sign != NTFS_FFFF_SIGNATURE) {
1221 * Remember if the signature does not
1222 * indicate uninitialized file.
1224 info->initialized = true;
1230 info->valid_page = false;
1231 info->initialized = true;
1234 /* Let's check the restart area if this is a valid page. */
1235 if (!is_rst_page_hdr_valid(vbo, r_page))
1237 ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
1239 if (!is_rst_area_valid(r_page))
1243 * We have a valid restart page header and restart area.
1244 * If chkdsk was run or we have no clients then we have
1245 * no more checking to do.
1247 if (bchk || ra->client_idx[1] == LFS_NO_CLIENT_LE) {
1248 info->valid_page = true;
1252 if (is_client_area_valid(r_page, usa_error)) {
1253 info->valid_page = true;
1254 ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
1259 * If chkdsk was run then update the caller's
1260 * values and return.
1262 if (r_page->rhdr.sign == NTFS_CHKD_SIGNATURE) {
1263 info->chkdsk_was_run = true;
1264 info->last_lsn = le64_to_cpu(r_page->rhdr.lsn);
1265 info->restart = true;
1266 info->r_page = r_page;
1271 * If we have a valid page then copy the values
1274 if (info->valid_page) {
1275 info->last_lsn = le64_to_cpu(ra->current_lsn);
1276 info->restart = true;
1277 info->r_page = r_page;
1288 * Ilog_init_pg_hdr - Init @log from restart page header.
1290 static void log_init_pg_hdr(struct ntfs_log *log, u32 sys_page_size,
1291 u32 page_size, u16 major_ver, u16 minor_ver)
1293 log->sys_page_size = sys_page_size;
1294 log->sys_page_mask = sys_page_size - 1;
1295 log->page_size = page_size;
1296 log->page_mask = page_size - 1;
1297 log->page_bits = blksize_bits(page_size);
1299 log->clst_per_page = log->page_size >> log->ni->mi.sbi->cluster_bits;
1300 if (!log->clst_per_page)
1301 log->clst_per_page = 1;
1303 log->first_page = major_ver >= 2
1305 : ((sys_page_size << 1) + (page_size << 1));
1306 log->major_ver = major_ver;
1307 log->minor_ver = minor_ver;
1311 * log_create - Init @log in cases when we don't have a restart area to use.
1313 static void log_create(struct ntfs_log *log, u32 l_size, const u64 last_lsn,
1314 u32 open_log_count, bool wrapped, bool use_multi_page)
1316 log->l_size = l_size;
1317 /* All file offsets must be quadword aligned. */
1318 log->file_data_bits = blksize_bits(l_size) - 3;
1319 log->seq_num_mask = (8 << log->file_data_bits) - 1;
1320 log->seq_num_bits = sizeof(u64) * 8 - log->file_data_bits;
1321 log->seq_num = (last_lsn >> log->file_data_bits) + 2;
1322 log->next_page = log->first_page;
1323 log->oldest_lsn = log->seq_num << log->file_data_bits;
1324 log->oldest_lsn_off = 0;
1325 log->last_lsn = log->oldest_lsn;
1327 log->l_flags |= NTFSLOG_NO_LAST_LSN | NTFSLOG_NO_OLDEST_LSN;
1329 /* Set the correct flags for the I/O and indicate if we have wrapped. */
1331 log->l_flags |= NTFSLOG_WRAPPED;
1334 log->l_flags |= NTFSLOG_MULTIPLE_PAGE_IO;
1336 /* Compute the log page values. */
1337 log->data_off = ALIGN(
1338 offsetof(struct RECORD_PAGE_HDR, fixups) +
1339 sizeof(short) * ((log->page_size >> SECTOR_SHIFT) + 1),
1341 log->data_size = log->page_size - log->data_off;
1342 log->record_header_len = sizeof(struct LFS_RECORD_HDR);
1344 /* Remember the different page sizes for reservation. */
1345 log->reserved = log->data_size - log->record_header_len;
1347 /* Compute the restart page values. */
1348 log->ra_off = ALIGN(
1349 offsetof(struct RESTART_HDR, fixups) +
1351 ((log->sys_page_size >> SECTOR_SHIFT) + 1),
1353 log->restart_size = log->sys_page_size - log->ra_off;
1354 log->ra_size = struct_size(log->ra, clients, 1);
1355 log->current_openlog_count = open_log_count;
1358 * The total available log file space is the number of
1359 * log file pages times the space available on each page.
1361 log->total_avail_pages = log->l_size - log->first_page;
1362 log->total_avail = log->total_avail_pages >> log->page_bits;
1365 * We assume that we can't use the end of the page less than
1366 * the file record size.
1367 * Then we won't need to reserve more than the caller asks for.
1369 log->max_current_avail = log->total_avail * log->reserved;
1370 log->total_avail = log->total_avail * log->data_size;
1371 log->current_avail = log->max_current_avail;
1375 * log_create_ra - Fill a restart area from the values stored in @log.
1377 static struct RESTART_AREA *log_create_ra(struct ntfs_log *log)
1379 struct CLIENT_REC *cr;
1380 struct RESTART_AREA *ra = kzalloc(log->restart_size, GFP_NOFS);
1385 ra->current_lsn = cpu_to_le64(log->last_lsn);
1386 ra->log_clients = cpu_to_le16(1);
1387 ra->client_idx[1] = LFS_NO_CLIENT_LE;
1388 if (log->l_flags & NTFSLOG_MULTIPLE_PAGE_IO)
1389 ra->flags = RESTART_SINGLE_PAGE_IO;
1390 ra->seq_num_bits = cpu_to_le32(log->seq_num_bits);
1391 ra->ra_len = cpu_to_le16(log->ra_size);
1392 ra->client_off = cpu_to_le16(offsetof(struct RESTART_AREA, clients));
1393 ra->l_size = cpu_to_le64(log->l_size);
1394 ra->rec_hdr_len = cpu_to_le16(log->record_header_len);
1395 ra->data_off = cpu_to_le16(log->data_off);
1396 ra->open_log_count = cpu_to_le32(log->current_openlog_count + 1);
1400 cr->prev_client = LFS_NO_CLIENT_LE;
1401 cr->next_client = LFS_NO_CLIENT_LE;
1406 static u32 final_log_off(struct ntfs_log *log, u64 lsn, u32 data_len)
1408 u32 base_vbo = lsn << 3;
1409 u32 final_log_off = (base_vbo & log->seq_num_mask) & ~log->page_mask;
1410 u32 page_off = base_vbo & log->page_mask;
1411 u32 tail = log->page_size - page_off;
1415 /* Add the length of the header. */
1416 data_len += log->record_header_len;
1419 * If this lsn is contained this log page we are done.
1420 * Otherwise we need to walk through several log pages.
1422 if (data_len > tail) {
1424 tail = log->data_size;
1425 page_off = log->data_off - 1;
1428 final_log_off = next_page_off(log, final_log_off);
1431 * We are done if the remaining bytes
1434 if (data_len <= tail)
1441 * We add the remaining bytes to our starting position on this page
1442 * and then add that value to the file offset of this log page.
1444 return final_log_off + data_len + page_off;
1447 static int next_log_lsn(struct ntfs_log *log, const struct LFS_RECORD_HDR *rh,
1451 u64 this_lsn = le64_to_cpu(rh->this_lsn);
1452 u32 vbo = lsn_to_vbo(log, this_lsn);
1454 final_log_off(log, this_lsn, le32_to_cpu(rh->client_data_len));
1455 u32 hdr_off = end & ~log->sys_page_mask;
1456 u64 seq = this_lsn >> log->file_data_bits;
1457 struct RECORD_PAGE_HDR *page = NULL;
1459 /* Remember if we wrapped. */
1463 /* Log page header for this page. */
1464 err = read_log_page(log, hdr_off, &page, NULL);
1469 * If the lsn we were given was not the last lsn on this page,
1470 * then the starting offset for the next lsn is on a quad word
1471 * boundary following the last file offset for the current lsn.
1472 * Otherwise the file offset is the start of the data on the next page.
1474 if (this_lsn == le64_to_cpu(page->rhdr.lsn)) {
1475 /* If we wrapped, we need to increment the sequence number. */
1476 hdr_off = next_page_off(log, hdr_off);
1477 if (hdr_off == log->first_page)
1480 vbo = hdr_off + log->data_off;
1482 vbo = ALIGN(end, 8);
1485 /* Compute the lsn based on the file offset and the sequence count. */
1486 *lsn = vbo_to_lsn(log, vbo, seq);
1489 * If this lsn is within the legal range for the file, we return true.
1490 * Otherwise false indicates that there are no more lsn's.
1492 if (!is_lsn_in_file(log, *lsn))
1501 * current_log_avail - Calculate the number of bytes available for log records.
1503 static u32 current_log_avail(struct ntfs_log *log)
1505 u32 oldest_off, next_free_off, free_bytes;
1507 if (log->l_flags & NTFSLOG_NO_LAST_LSN) {
1508 /* The entire file is available. */
1509 return log->max_current_avail;
1513 * If there is a last lsn the restart area then we know that we will
1514 * have to compute the free range.
1515 * If there is no oldest lsn then start at the first page of the file.
1517 oldest_off = (log->l_flags & NTFSLOG_NO_OLDEST_LSN)
1519 : (log->oldest_lsn_off & ~log->sys_page_mask);
1522 * We will use the next log page offset to compute the next free page.
1523 * If we are going to reuse this page go to the next page.
1524 * If we are at the first page then use the end of the file.
1526 next_free_off = (log->l_flags & NTFSLOG_REUSE_TAIL)
1527 ? log->next_page + log->page_size
1528 : log->next_page == log->first_page
1532 /* If the two offsets are the same then there is no available space. */
1533 if (oldest_off == next_free_off)
1536 * If the free offset follows the oldest offset then subtract
1537 * this range from the total available pages.
1540 oldest_off < next_free_off
1541 ? log->total_avail_pages - (next_free_off - oldest_off)
1542 : oldest_off - next_free_off;
1544 free_bytes >>= log->page_bits;
1545 return free_bytes * log->reserved;
1548 static bool check_subseq_log_page(struct ntfs_log *log,
1549 const struct RECORD_PAGE_HDR *rp, u32 vbo,
1553 const struct NTFS_RECORD_HEADER *rhdr = &rp->rhdr;
1554 u64 lsn = le64_to_cpu(rhdr->lsn);
1556 if (rhdr->sign == NTFS_FFFF_SIGNATURE || !rhdr->sign)
1560 * If the last lsn on the page occurs was written after the page
1561 * that caused the original error then we have a fatal error.
1563 lsn_seq = lsn >> log->file_data_bits;
1566 * If the sequence number for the lsn the page is equal or greater
1567 * than lsn we expect, then this is a subsequent write.
1569 return lsn_seq >= seq ||
1570 (lsn_seq == seq - 1 && log->first_page == vbo &&
1571 vbo != (lsn_to_vbo(log, lsn) & ~log->page_mask));
1577 * Walks through the log pages for a file, searching for the
1578 * last log page written to the file.
1580 static int last_log_lsn(struct ntfs_log *log)
1583 bool usa_error = false;
1584 bool replace_page = false;
1585 bool reuse_page = log->l_flags & NTFSLOG_REUSE_TAIL;
1586 bool wrapped_file, wrapped;
1588 u32 page_cnt = 1, page_pos = 1;
1589 u32 page_off = 0, page_off1 = 0, saved_off = 0;
1590 u32 final_off, second_off, final_off_prev = 0, second_off_prev = 0;
1591 u32 first_file_off = 0, second_file_off = 0;
1592 u32 part_io_count = 0;
1594 u32 this_off, curpage_off, nextpage_off, remain_pages;
1596 u64 expected_seq, seq_base = 0, lsn_base = 0;
1597 u64 best_lsn, best_lsn1, best_lsn2;
1598 u64 lsn_cur, lsn1, lsn2;
1599 u64 last_ok_lsn = reuse_page ? log->last_lsn : 0;
1601 u16 cur_pos, best_page_pos;
1603 struct RECORD_PAGE_HDR *page = NULL;
1604 struct RECORD_PAGE_HDR *tst_page = NULL;
1605 struct RECORD_PAGE_HDR *first_tail = NULL;
1606 struct RECORD_PAGE_HDR *second_tail = NULL;
1607 struct RECORD_PAGE_HDR *tail_page = NULL;
1608 struct RECORD_PAGE_HDR *second_tail_prev = NULL;
1609 struct RECORD_PAGE_HDR *first_tail_prev = NULL;
1610 struct RECORD_PAGE_HDR *page_bufs = NULL;
1611 struct RECORD_PAGE_HDR *best_page;
1613 if (log->major_ver >= 2) {
1614 final_off = 0x02 * log->page_size;
1615 second_off = 0x12 * log->page_size;
1617 // 0x10 == 0x12 - 0x2
1618 page_bufs = kmalloc(log->page_size * 0x10, GFP_NOFS);
1622 second_off = log->first_page - log->page_size;
1623 final_off = second_off - log->page_size;
1627 /* Read second tail page (at pos 3/0x12000). */
1628 if (read_log_page(log, second_off, &second_tail, &usa_error) ||
1629 usa_error || second_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
1632 second_file_off = 0;
1635 second_file_off = hdr_file_off(log, second_tail);
1636 lsn2 = le64_to_cpu(second_tail->record_hdr.last_end_lsn);
1639 /* Read first tail page (at pos 2/0x2000). */
1640 if (read_log_page(log, final_off, &first_tail, &usa_error) ||
1641 usa_error || first_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
1647 first_file_off = hdr_file_off(log, first_tail);
1648 lsn1 = le64_to_cpu(first_tail->record_hdr.last_end_lsn);
1651 if (log->major_ver < 2) {
1654 first_tail_prev = first_tail;
1655 final_off_prev = first_file_off;
1656 second_tail_prev = second_tail;
1657 second_off_prev = second_file_off;
1660 if (!first_tail && !second_tail)
1663 if (first_tail && second_tail)
1664 best_page = lsn1 < lsn2 ? 1 : 0;
1665 else if (first_tail)
1670 page_off = best_page ? second_file_off : first_file_off;
1671 seq_base = (best_page ? lsn2 : lsn1) >> log->file_data_bits;
1675 best_lsn1 = first_tail ? base_lsn(log, first_tail, first_file_off) : 0;
1677 second_tail ? base_lsn(log, second_tail, second_file_off) : 0;
1679 if (first_tail && second_tail) {
1680 if (best_lsn1 > best_lsn2) {
1681 best_lsn = best_lsn1;
1682 best_page = first_tail;
1683 this_off = first_file_off;
1685 best_lsn = best_lsn2;
1686 best_page = second_tail;
1687 this_off = second_file_off;
1689 } else if (first_tail) {
1690 best_lsn = best_lsn1;
1691 best_page = first_tail;
1692 this_off = first_file_off;
1693 } else if (second_tail) {
1694 best_lsn = best_lsn2;
1695 best_page = second_tail;
1696 this_off = second_file_off;
1701 best_page_pos = le16_to_cpu(best_page->page_pos);
1704 if (best_page_pos == page_pos) {
1705 seq_base = best_lsn >> log->file_data_bits;
1706 saved_off = page_off = le32_to_cpu(best_page->file_off);
1707 lsn_base = best_lsn;
1709 memmove(page_bufs, best_page, log->page_size);
1711 page_cnt = le16_to_cpu(best_page->page_count);
1717 } else if (seq_base == (best_lsn >> log->file_data_bits) &&
1718 saved_off + log->page_size == this_off &&
1719 lsn_base < best_lsn &&
1720 (page_pos != page_cnt || best_page_pos == page_pos ||
1721 best_page_pos == 1) &&
1722 (page_pos >= page_cnt || best_page_pos == page_pos)) {
1723 u16 bppc = le16_to_cpu(best_page->page_count);
1725 saved_off += log->page_size;
1726 lsn_base = best_lsn;
1728 memmove(Add2Ptr(page_bufs, tails * log->page_size), best_page,
1733 if (best_page_pos != bppc) {
1735 page_pos = best_page_pos;
1740 page_pos = page_cnt = 1;
1748 kfree(first_tail_prev);
1749 first_tail_prev = first_tail;
1750 final_off_prev = first_file_off;
1753 kfree(second_tail_prev);
1754 second_tail_prev = second_tail;
1755 second_off_prev = second_file_off;
1758 final_off += log->page_size;
1759 second_off += log->page_size;
1764 first_tail = first_tail_prev;
1765 final_off = final_off_prev;
1767 second_tail = second_tail_prev;
1768 second_off = second_off_prev;
1770 page_cnt = page_pos = 1;
1772 curpage_off = seq_base == log->seq_num ? min(log->next_page, page_off)
1776 curpage_off == log->first_page &&
1777 !(log->l_flags & (NTFSLOG_NO_LAST_LSN | NTFSLOG_REUSE_TAIL));
1779 expected_seq = wrapped_file ? (log->seq_num + 1) : log->seq_num;
1781 nextpage_off = curpage_off;
1785 /* Read the next log page. */
1786 err = read_log_page(log, curpage_off, &page, &usa_error);
1788 /* Compute the next log page offset the file. */
1789 nextpage_off = next_page_off(log, curpage_off);
1790 wrapped = nextpage_off == log->first_page;
1793 struct RECORD_PAGE_HDR *cur_page =
1794 Add2Ptr(page_bufs, curpage_off - page_off);
1796 if (curpage_off == saved_off) {
1797 tail_page = cur_page;
1801 if (page_off > curpage_off || curpage_off >= saved_off)
1807 if (!err && !usa_error &&
1808 page->rhdr.sign == NTFS_RCRD_SIGNATURE &&
1809 cur_page->rhdr.lsn == page->rhdr.lsn &&
1810 cur_page->record_hdr.next_record_off ==
1811 page->record_hdr.next_record_off &&
1812 ((page_pos == page_cnt &&
1813 le16_to_cpu(page->page_pos) == 1) ||
1814 (page_pos != page_cnt &&
1815 le16_to_cpu(page->page_pos) == page_pos + 1 &&
1816 le16_to_cpu(page->page_count) == page_cnt))) {
1821 page_off1 = page_off;
1825 lsn_cur = le64_to_cpu(cur_page->rhdr.lsn);
1828 le64_to_cpu(cur_page->record_hdr.last_end_lsn) &&
1829 ((lsn_cur >> log->file_data_bits) +
1831 (lsn_to_vbo(log, lsn_cur) & ~log->page_mask))
1833 : 0)) != expected_seq) {
1837 if (!is_log_record_end(cur_page)) {
1839 last_ok_lsn = lsn_cur;
1843 log->seq_num = expected_seq;
1844 log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
1845 log->last_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
1846 log->ra->current_lsn = cur_page->record_hdr.last_end_lsn;
1848 if (log->record_header_len <=
1850 le16_to_cpu(cur_page->record_hdr.next_record_off)) {
1851 log->l_flags |= NTFSLOG_REUSE_TAIL;
1852 log->next_page = curpage_off;
1854 log->l_flags &= ~NTFSLOG_REUSE_TAIL;
1855 log->next_page = nextpage_off;
1859 log->l_flags |= NTFSLOG_WRAPPED;
1861 last_ok_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
1866 * If we are at the expected first page of a transfer check to see
1867 * if either tail copy is at this offset.
1868 * If this page is the last page of a transfer, check if we wrote
1869 * a subsequent tail copy.
1871 if (page_cnt == page_pos || page_cnt == page_pos + 1) {
1873 * Check if the offset matches either the first or second
1874 * tail copy. It is possible it will match both.
1876 if (curpage_off == final_off)
1877 tail_page = first_tail;
1880 * If we already matched on the first page then
1881 * check the ending lsn's.
1883 if (curpage_off == second_off) {
1886 le64_to_cpu(second_tail->record_hdr.last_end_lsn) >
1887 le64_to_cpu(first_tail->record_hdr
1889 tail_page = second_tail;
1896 /* We have a candidate for a tail copy. */
1897 lsn_cur = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
1899 if (last_ok_lsn < lsn_cur) {
1901 * If the sequence number is not expected,
1902 * then don't use the tail copy.
1904 if (expected_seq != (lsn_cur >> log->file_data_bits))
1906 } else if (last_ok_lsn > lsn_cur) {
1908 * If the last lsn is greater than the one on
1909 * this page then forget this tail.
1916 *If we have an error on the current page,
1917 * we will break of this loop.
1919 if (err || usa_error)
1923 * Done if the last lsn on this page doesn't match the previous known
1924 * last lsn or the sequence number is not expected.
1926 lsn_cur = le64_to_cpu(page->rhdr.lsn);
1927 if (last_ok_lsn != lsn_cur &&
1928 expected_seq != (lsn_cur >> log->file_data_bits)) {
1933 * Check that the page position and page count values are correct.
1934 * If this is the first page of a transfer the position must be 1
1935 * and the count will be unknown.
1937 if (page_cnt == page_pos) {
1938 if (page->page_pos != cpu_to_le16(1) &&
1939 (!reuse_page || page->page_pos != page->page_count)) {
1941 * If the current page is the first page we are
1942 * looking at and we are reusing this page then
1943 * it can be either the first or last page of a
1944 * transfer. Otherwise it can only be the first.
1948 } else if (le16_to_cpu(page->page_count) != page_cnt ||
1949 le16_to_cpu(page->page_pos) != page_pos + 1) {
1951 * The page position better be 1 more than the last page
1952 * position and the page count better match.
1958 * We have a valid page the file and may have a valid page
1959 * the tail copy area.
1960 * If the tail page was written after the page the file then
1961 * break of the loop.
1964 le64_to_cpu(tail_page->record_hdr.last_end_lsn) > lsn_cur) {
1965 /* Remember if we will replace the page. */
1966 replace_page = true;
1972 if (is_log_record_end(page)) {
1974 * Since we have read this page we know the sequence number
1975 * is the same as our expected value.
1977 log->seq_num = expected_seq;
1978 log->last_lsn = le64_to_cpu(page->record_hdr.last_end_lsn);
1979 log->ra->current_lsn = page->record_hdr.last_end_lsn;
1980 log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
1983 * If there is room on this page for another header then
1984 * remember we want to reuse the page.
1986 if (log->record_header_len <=
1988 le16_to_cpu(page->record_hdr.next_record_off)) {
1989 log->l_flags |= NTFSLOG_REUSE_TAIL;
1990 log->next_page = curpage_off;
1992 log->l_flags &= ~NTFSLOG_REUSE_TAIL;
1993 log->next_page = nextpage_off;
1996 /* Remember if we wrapped the log file. */
1998 log->l_flags |= NTFSLOG_WRAPPED;
2002 * Remember the last page count and position.
2003 * Also remember the last known lsn.
2005 page_cnt = le16_to_cpu(page->page_count);
2006 page_pos = le16_to_cpu(page->page_pos);
2007 last_ok_lsn = le64_to_cpu(page->rhdr.lsn);
2016 curpage_off = nextpage_off;
2024 log->seq_num = expected_seq;
2025 log->last_lsn = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
2026 log->ra->current_lsn = tail_page->record_hdr.last_end_lsn;
2027 log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
2029 if (log->page_size -
2031 tail_page->record_hdr.next_record_off) >=
2032 log->record_header_len) {
2033 log->l_flags |= NTFSLOG_REUSE_TAIL;
2034 log->next_page = curpage_off;
2036 log->l_flags &= ~NTFSLOG_REUSE_TAIL;
2037 log->next_page = nextpage_off;
2041 log->l_flags |= NTFSLOG_WRAPPED;
2044 /* Remember that the partial IO will start at the next page. */
2045 second_off = nextpage_off;
2048 * If the next page is the first page of the file then update
2049 * the sequence number for log records which begon the next page.
2055 * If we have a tail copy or are performing single page I/O we can
2056 * immediately look at the next page.
2058 if (replace_page || (log->ra->flags & RESTART_SINGLE_PAGE_IO)) {
2064 if (page_pos != page_cnt)
2067 * If the next page causes us to wrap to the beginning of the log
2068 * file then we know which page to check next.
2082 /* Walk through the file, reading log pages. */
2083 err = read_log_page(log, nextpage_off, &tst_page, &usa_error);
2086 * If we get a USA error then assume that we correctly found
2087 * the end of the original transfer.
2093 * If we were able to read the page, we examine it to see if it
2094 * is the same or different Io block.
2097 goto next_test_page_1;
2099 if (le16_to_cpu(tst_page->page_pos) == cur_pos &&
2100 check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
2101 page_cnt = le16_to_cpu(tst_page->page_count) + 1;
2102 page_pos = le16_to_cpu(tst_page->page_pos);
2110 nextpage_off = next_page_off(log, curpage_off);
2111 wrapped = nextpage_off == log->first_page;
2122 goto next_test_page;
2125 /* Skip over the remaining pages this transfer. */
2126 remain_pages = page_cnt - page_pos - 1;
2127 part_io_count += remain_pages;
2129 while (remain_pages--) {
2130 nextpage_off = next_page_off(log, curpage_off);
2131 wrapped = nextpage_off == log->first_page;
2137 /* Call our routine to check this log page. */
2141 err = read_log_page(log, nextpage_off, &tst_page, &usa_error);
2142 if (!err && !usa_error &&
2143 check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
2150 /* We have a valid file. */
2151 if (page_off1 || tail_page) {
2152 struct RECORD_PAGE_HDR *tmp_page;
2154 if (sb_rdonly(log->ni->mi.sbi->sb)) {
2160 tmp_page = Add2Ptr(page_bufs, page_off1 - page_off);
2161 tails -= (page_off1 - page_off) / log->page_size;
2165 tmp_page = tail_page;
2170 u64 off = hdr_file_off(log, tmp_page);
2173 page = kmalloc(log->page_size, GFP_NOFS);
2181 * Correct page and copy the data from this page
2182 * into it and flush it to disk.
2184 memcpy(page, tmp_page, log->page_size);
2186 /* Fill last flushed lsn value flush the page. */
2187 if (log->major_ver < 2)
2188 page->rhdr.lsn = page->record_hdr.last_end_lsn;
2192 page->page_pos = page->page_count = cpu_to_le16(1);
2194 ntfs_fix_pre_write(&page->rhdr, log->page_size);
2196 err = ntfs_sb_write_run(log->ni->mi.sbi,
2197 &log->ni->file.run, off, page,
2203 if (part_io_count && second_off == off) {
2204 second_off += log->page_size;
2208 tmp_page = Add2Ptr(tmp_page, log->page_size);
2212 if (part_io_count) {
2213 if (sb_rdonly(log->ni->mi.sbi->sb)) {
2230 * read_log_rec_buf - Copy a log record from the file to a buffer.
2232 * The log record may span several log pages and may even wrap the file.
2234 static int read_log_rec_buf(struct ntfs_log *log,
2235 const struct LFS_RECORD_HDR *rh, void *buffer)
2238 struct RECORD_PAGE_HDR *ph = NULL;
2239 u64 lsn = le64_to_cpu(rh->this_lsn);
2240 u32 vbo = lsn_to_vbo(log, lsn) & ~log->page_mask;
2241 u32 off = lsn_to_page_off(log, lsn) + log->record_header_len;
2242 u32 data_len = le32_to_cpu(rh->client_data_len);
2245 * While there are more bytes to transfer,
2246 * we continue to attempt to perform the read.
2250 u32 tail = log->page_size - off;
2252 if (tail >= data_len)
2257 err = read_log_page(log, vbo, &ph, &usa_error);
2262 * The last lsn on this page better be greater or equal
2263 * to the lsn we are copying.
2265 if (lsn > le64_to_cpu(ph->rhdr.lsn)) {
2270 memcpy(buffer, Add2Ptr(ph, off), tail);
2272 /* If there are no more bytes to transfer, we exit the loop. */
2274 if (!is_log_record_end(ph) ||
2275 lsn > le64_to_cpu(ph->record_hdr.last_end_lsn)) {
2282 if (ph->rhdr.lsn == ph->record_hdr.last_end_lsn ||
2283 lsn > le64_to_cpu(ph->rhdr.lsn)) {
2288 vbo = next_page_off(log, vbo);
2289 off = log->data_off;
2292 * Adjust our pointer the user's buffer to transfer
2293 * the next block to.
2295 buffer = Add2Ptr(buffer, tail);
2303 static int read_rst_area(struct ntfs_log *log, struct NTFS_RESTART **rst_,
2307 struct LFS_RECORD_HDR *rh = NULL;
2308 const struct CLIENT_REC *cr =
2309 Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
2310 u64 lsnr, lsnc = le64_to_cpu(cr->restart_lsn);
2312 struct NTFS_RESTART *rst;
2317 /* If the client doesn't have a restart area, go ahead and exit now. */
2321 err = read_log_page(log, lsn_to_vbo(log, lsnc),
2322 (struct RECORD_PAGE_HDR **)&rh, NULL);
2327 lsnr = le64_to_cpu(rh->this_lsn);
2330 /* If the lsn values don't match, then the disk is corrupt. */
2336 len = le32_to_cpu(rh->client_data_len);
2343 if (len < sizeof(struct NTFS_RESTART)) {
2348 rst = kmalloc(len, GFP_NOFS);
2354 /* Copy the data into the 'rst' buffer. */
2355 err = read_log_rec_buf(log, rh, rst);
2369 static int find_log_rec(struct ntfs_log *log, u64 lsn, struct lcb *lcb)
2372 struct LFS_RECORD_HDR *rh = lcb->lrh;
2375 /* Read the record header for this lsn. */
2377 err = read_log_page(log, lsn_to_vbo(log, lsn),
2378 (struct RECORD_PAGE_HDR **)&rh, NULL);
2386 * If the lsn the log record doesn't match the desired
2387 * lsn then the disk is corrupt.
2389 if (lsn != le64_to_cpu(rh->this_lsn))
2392 len = le32_to_cpu(rh->client_data_len);
2395 * Check that the length field isn't greater than the total
2396 * available space the log file.
2398 rec_len = len + log->record_header_len;
2399 if (rec_len >= log->total_avail)
2403 * If the entire log record is on this log page,
2404 * put a pointer to the log record the context block.
2406 if (rh->flags & LOG_RECORD_MULTI_PAGE) {
2407 void *lr = kmalloc(len, GFP_NOFS);
2415 /* Copy the data into the buffer returned. */
2416 err = read_log_rec_buf(log, rh, lr);
2420 /* If beyond the end of the current page -> an error. */
2421 u32 page_off = lsn_to_page_off(log, lsn);
2423 if (page_off + len + log->record_header_len > log->page_size)
2426 lcb->log_rec = Add2Ptr(rh, sizeof(struct LFS_RECORD_HDR));
2434 * read_log_rec_lcb - Init the query operation.
2436 static int read_log_rec_lcb(struct ntfs_log *log, u64 lsn, u32 ctx_mode,
2440 const struct CLIENT_REC *cr;
2444 case lcb_ctx_undo_next:
2452 /* Check that the given lsn is the legal range for this client. */
2453 cr = Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
2455 if (!verify_client_lsn(log, cr, lsn))
2458 lcb = kzalloc(sizeof(struct lcb), GFP_NOFS);
2461 lcb->client = log->client_id;
2462 lcb->ctx_mode = ctx_mode;
2464 /* Find the log record indicated by the given lsn. */
2465 err = find_log_rec(log, lsn, lcb);
2479 * find_client_next_lsn
2481 * Attempt to find the next lsn to return to a client based on the context mode.
2483 static int find_client_next_lsn(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
2487 struct LFS_RECORD_HDR *hdr;
2492 if (lcb_ctx_next != lcb->ctx_mode)
2493 goto check_undo_next;
2495 /* Loop as long as another lsn can be found. */
2499 err = next_log_lsn(log, hdr, ¤t_lsn);
2506 if (hdr != lcb->lrh)
2510 err = read_log_page(log, lsn_to_vbo(log, current_lsn),
2511 (struct RECORD_PAGE_HDR **)&hdr, NULL);
2515 if (memcmp(&hdr->client, &lcb->client,
2516 sizeof(struct CLIENT_ID))) {
2518 } else if (LfsClientRecord == hdr->record_type) {
2527 if (hdr != lcb->lrh)
2532 if (lcb_ctx_undo_next == lcb->ctx_mode)
2533 next_lsn = le64_to_cpu(hdr->client_undo_next_lsn);
2534 else if (lcb_ctx_prev == lcb->ctx_mode)
2535 next_lsn = le64_to_cpu(hdr->client_prev_lsn);
2542 if (!verify_client_lsn(
2543 log, Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off)),
2548 err = read_log_page(log, lsn_to_vbo(log, next_lsn),
2549 (struct RECORD_PAGE_HDR **)&hdr, NULL);
2560 static int read_next_log_rec(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
2564 err = find_client_next_lsn(log, lcb, lsn);
2572 kfree(lcb->log_rec);
2574 lcb->log_rec = NULL;
2579 return find_log_rec(log, *lsn, lcb);
2582 bool check_index_header(const struct INDEX_HDR *hdr, size_t bytes)
2585 u32 min_de, de_off, used, total;
2586 const struct NTFS_DE *e;
2588 if (hdr_has_subnode(hdr)) {
2589 min_de = sizeof(struct NTFS_DE) + sizeof(u64);
2590 mask = NTFS_IE_HAS_SUBNODES;
2592 min_de = sizeof(struct NTFS_DE);
2596 de_off = le32_to_cpu(hdr->de_off);
2597 used = le32_to_cpu(hdr->used);
2598 total = le32_to_cpu(hdr->total);
2600 if (de_off > bytes - min_de || used > bytes || total > bytes ||
2601 de_off + min_de > used || used > total) {
2605 e = Add2Ptr(hdr, de_off);
2607 u16 esize = le16_to_cpu(e->size);
2608 struct NTFS_DE *next = Add2Ptr(e, esize);
2610 if (esize < min_de || PtrOffset(hdr, next) > used ||
2611 (e->flags & NTFS_IE_HAS_SUBNODES) != mask) {
2624 static inline bool check_index_buffer(const struct INDEX_BUFFER *ib, u32 bytes)
2627 const struct NTFS_RECORD_HEADER *r = &ib->rhdr;
2629 if (r->sign != NTFS_INDX_SIGNATURE)
2632 fo = (SECTOR_SIZE - ((bytes >> SECTOR_SHIFT) + 1) * sizeof(short));
2634 if (le16_to_cpu(r->fix_off) > fo)
2637 if ((le16_to_cpu(r->fix_num) - 1) * SECTOR_SIZE != bytes)
2640 return check_index_header(&ib->ihdr,
2641 bytes - offsetof(struct INDEX_BUFFER, ihdr));
2644 static inline bool check_index_root(const struct ATTRIB *attr,
2645 struct ntfs_sb_info *sbi)
2648 const struct INDEX_ROOT *root = resident_data(attr);
2649 u8 index_bits = le32_to_cpu(root->index_block_size) >= sbi->cluster_size
2652 u8 block_clst = root->index_block_clst;
2654 if (le32_to_cpu(attr->res.data_size) < sizeof(struct INDEX_ROOT) ||
2655 (root->type != ATTR_NAME && root->type != ATTR_ZERO) ||
2656 (root->type == ATTR_NAME &&
2657 root->rule != NTFS_COLLATION_TYPE_FILENAME) ||
2658 (le32_to_cpu(root->index_block_size) !=
2659 (block_clst << index_bits)) ||
2660 (block_clst != 1 && block_clst != 2 && block_clst != 4 &&
2661 block_clst != 8 && block_clst != 0x10 && block_clst != 0x20 &&
2662 block_clst != 0x40 && block_clst != 0x80)) {
2666 ret = check_index_header(&root->ihdr,
2667 le32_to_cpu(attr->res.data_size) -
2668 offsetof(struct INDEX_ROOT, ihdr));
2672 static inline bool check_attr(const struct MFT_REC *rec,
2673 const struct ATTRIB *attr,
2674 struct ntfs_sb_info *sbi)
2676 u32 asize = le32_to_cpu(attr->size);
2678 u64 dsize, svcn, evcn;
2681 /* Check the fixed part of the attribute record header. */
2682 if (asize >= sbi->record_size ||
2683 asize + PtrOffset(rec, attr) >= sbi->record_size ||
2685 le16_to_cpu(attr->name_off) + attr->name_len * sizeof(short) >
2690 /* Check the attribute fields. */
2691 switch (attr->non_res) {
2693 rsize = le32_to_cpu(attr->res.data_size);
2694 if (rsize >= asize ||
2695 le16_to_cpu(attr->res.data_off) + rsize > asize) {
2701 dsize = le64_to_cpu(attr->nres.data_size);
2702 svcn = le64_to_cpu(attr->nres.svcn);
2703 evcn = le64_to_cpu(attr->nres.evcn);
2704 run_off = le16_to_cpu(attr->nres.run_off);
2706 if (svcn > evcn + 1 || run_off >= asize ||
2707 le64_to_cpu(attr->nres.valid_size) > dsize ||
2708 dsize > le64_to_cpu(attr->nres.alloc_size)) {
2712 if (run_off > asize)
2715 if (run_unpack(NULL, sbi, 0, svcn, evcn, svcn,
2716 Add2Ptr(attr, run_off), asize - run_off) < 0) {
2726 switch (attr->type) {
2728 if (fname_full_size(Add2Ptr(
2729 attr, le16_to_cpu(attr->res.data_off))) > asize) {
2735 return check_index_root(attr, sbi);
2738 if (rsize < sizeof(struct ATTR_STD_INFO5) &&
2739 rsize != sizeof(struct ATTR_STD_INFO)) {
2755 case ATTR_PROPERTYSET:
2756 case ATTR_LOGGED_UTILITY_STREAM:
2766 static inline bool check_file_record(const struct MFT_REC *rec,
2767 const struct MFT_REC *rec2,
2768 struct ntfs_sb_info *sbi)
2770 const struct ATTRIB *attr;
2771 u16 fo = le16_to_cpu(rec->rhdr.fix_off);
2772 u16 fn = le16_to_cpu(rec->rhdr.fix_num);
2773 u16 ao = le16_to_cpu(rec->attr_off);
2774 u32 rs = sbi->record_size;
2776 /* Check the file record header for consistency. */
2777 if (rec->rhdr.sign != NTFS_FILE_SIGNATURE ||
2778 fo > (SECTOR_SIZE - ((rs >> SECTOR_SHIFT) + 1) * sizeof(short)) ||
2779 (fn - 1) * SECTOR_SIZE != rs || ao < MFTRECORD_FIXUP_OFFSET_1 ||
2780 ao > sbi->record_size - SIZEOF_RESIDENT || !is_rec_inuse(rec) ||
2781 le32_to_cpu(rec->total) != rs) {
2785 /* Loop to check all of the attributes. */
2786 for (attr = Add2Ptr(rec, ao); attr->type != ATTR_END;
2787 attr = Add2Ptr(attr, le32_to_cpu(attr->size))) {
2788 if (check_attr(rec, attr, sbi))
2796 static inline int check_lsn(const struct NTFS_RECORD_HEADER *hdr,
2804 lsn = le64_to_cpu(hdr->lsn);
2806 if (hdr->sign == NTFS_HOLE_SIGNATURE)
2815 static inline bool check_if_attr(const struct MFT_REC *rec,
2816 const struct LOG_REC_HDR *lrh)
2818 u16 ro = le16_to_cpu(lrh->record_off);
2819 u16 o = le16_to_cpu(rec->attr_off);
2820 const struct ATTRIB *attr = Add2Ptr(rec, o);
2825 if (attr->type == ATTR_END)
2828 asize = le32_to_cpu(attr->size);
2833 attr = Add2Ptr(attr, asize);
2839 static inline bool check_if_index_root(const struct MFT_REC *rec,
2840 const struct LOG_REC_HDR *lrh)
2842 u16 ro = le16_to_cpu(lrh->record_off);
2843 u16 o = le16_to_cpu(rec->attr_off);
2844 const struct ATTRIB *attr = Add2Ptr(rec, o);
2849 if (attr->type == ATTR_END)
2852 asize = le32_to_cpu(attr->size);
2857 attr = Add2Ptr(attr, asize);
2860 return o == ro && attr->type == ATTR_ROOT;
2863 static inline bool check_if_root_index(const struct ATTRIB *attr,
2864 const struct INDEX_HDR *hdr,
2865 const struct LOG_REC_HDR *lrh)
2867 u16 ao = le16_to_cpu(lrh->attr_off);
2868 u32 de_off = le32_to_cpu(hdr->de_off);
2869 u32 o = PtrOffset(attr, hdr) + de_off;
2870 const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
2871 u32 asize = le32_to_cpu(attr->size);
2879 esize = le16_to_cpu(e->size);
2884 e = Add2Ptr(e, esize);
2890 static inline bool check_if_alloc_index(const struct INDEX_HDR *hdr,
2893 u32 de_off = le32_to_cpu(hdr->de_off);
2894 u32 o = offsetof(struct INDEX_BUFFER, ihdr) + de_off;
2895 const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
2896 u32 used = le32_to_cpu(hdr->used);
2898 while (o < attr_off) {
2904 esize = le16_to_cpu(e->size);
2910 e = Add2Ptr(e, esize);
2913 return o == attr_off;
2916 static inline void change_attr_size(struct MFT_REC *rec, struct ATTRIB *attr,
2919 u32 asize = le32_to_cpu(attr->size);
2920 int dsize = nsize - asize;
2921 u8 *next = Add2Ptr(attr, asize);
2922 u32 used = le32_to_cpu(rec->used);
2924 memmove(Add2Ptr(attr, nsize), next, used - PtrOffset(rec, next));
2926 rec->used = cpu_to_le32(used + dsize);
2927 attr->size = cpu_to_le32(nsize);
2931 struct ATTRIB *attr;
2932 struct runs_tree *run1;
2933 struct runs_tree run0;
2934 struct ntfs_inode *ni;
2941 * Return: 0 if 'attr' has the same type and name.
2943 static inline int cmp_type_and_name(const struct ATTRIB *a1,
2944 const struct ATTRIB *a2)
2946 return a1->type != a2->type || a1->name_len != a2->name_len ||
2947 (a1->name_len && memcmp(attr_name(a1), attr_name(a2),
2948 a1->name_len * sizeof(short)));
2951 static struct OpenAttr *find_loaded_attr(struct ntfs_log *log,
2952 const struct ATTRIB *attr, CLST rno)
2954 struct OPEN_ATTR_ENRTY *oe = NULL;
2956 while ((oe = enum_rstbl(log->open_attr_tbl, oe))) {
2957 struct OpenAttr *op_attr;
2959 if (ino_get(&oe->ref) != rno)
2962 op_attr = (struct OpenAttr *)oe->ptr;
2963 if (!cmp_type_and_name(op_attr->attr, attr))
2969 static struct ATTRIB *attr_create_nonres_log(struct ntfs_sb_info *sbi,
2970 enum ATTR_TYPE type, u64 size,
2971 const u16 *name, size_t name_len,
2974 struct ATTRIB *attr;
2975 u32 name_size = ALIGN(name_len * sizeof(short), 8);
2976 bool is_ext = flags & (ATTR_FLAG_COMPRESSED | ATTR_FLAG_SPARSED);
2977 u32 asize = name_size +
2978 (is_ext ? SIZEOF_NONRESIDENT_EX : SIZEOF_NONRESIDENT);
2980 attr = kzalloc(asize, GFP_NOFS);
2985 attr->size = cpu_to_le32(asize);
2986 attr->flags = flags;
2988 attr->name_len = name_len;
2990 attr->nres.evcn = cpu_to_le64((u64)bytes_to_cluster(sbi, size) - 1);
2991 attr->nres.alloc_size = cpu_to_le64(ntfs_up_cluster(sbi, size));
2992 attr->nres.data_size = cpu_to_le64(size);
2993 attr->nres.valid_size = attr->nres.data_size;
2995 attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
2996 if (is_attr_compressed(attr))
2997 attr->nres.c_unit = COMPRESSION_UNIT;
2999 attr->nres.run_off =
3000 cpu_to_le16(SIZEOF_NONRESIDENT_EX + name_size);
3001 memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT_EX), name,
3002 name_len * sizeof(short));
3004 attr->name_off = SIZEOF_NONRESIDENT_LE;
3005 attr->nres.run_off =
3006 cpu_to_le16(SIZEOF_NONRESIDENT + name_size);
3007 memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT), name,
3008 name_len * sizeof(short));
3015 * do_action - Common routine for the Redo and Undo Passes.
3016 * @rlsn: If it is NULL then undo.
3018 static int do_action(struct ntfs_log *log, struct OPEN_ATTR_ENRTY *oe,
3019 const struct LOG_REC_HDR *lrh, u32 op, void *data,
3020 u32 dlen, u32 rec_len, const u64 *rlsn)
3023 struct ntfs_sb_info *sbi = log->ni->mi.sbi;
3024 struct inode *inode = NULL, *inode_parent;
3025 struct mft_inode *mi = NULL, *mi2_child = NULL;
3026 CLST rno = 0, rno_base = 0;
3027 struct INDEX_BUFFER *ib = NULL;
3028 struct MFT_REC *rec = NULL;
3029 struct ATTRIB *attr = NULL, *attr2;
3030 struct INDEX_HDR *hdr;
3031 struct INDEX_ROOT *root;
3032 struct NTFS_DE *e, *e1, *e2;
3033 struct NEW_ATTRIBUTE_SIZES *new_sz;
3034 struct ATTR_FILE_NAME *fname;
3035 struct OpenAttr *oa, *oa2;
3036 u32 nsize, t32, asize, used, esize, bmp_off, bmp_bits;
3038 u32 record_size = sbi->record_size;
3040 u16 roff = le16_to_cpu(lrh->record_off);
3041 u16 aoff = le16_to_cpu(lrh->attr_off);
3043 u64 cbo = (u64)le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
3044 u64 tvo = le64_to_cpu(lrh->target_vcn) << sbi->cluster_bits;
3045 u64 vbo = cbo + tvo;
3046 void *buffer_le = NULL;
3048 bool a_dirty = false;
3053 /* Big switch to prepare. */
3055 /* ============================================================
3056 * Process MFT records, as described by the current log record.
3057 * ============================================================
3059 case InitializeFileRecordSegment:
3060 case DeallocateFileRecordSegment:
3061 case WriteEndOfFileRecordSegment:
3062 case CreateAttribute:
3063 case DeleteAttribute:
3064 case UpdateResidentValue:
3065 case UpdateMappingPairs:
3066 case SetNewAttributeSizes:
3067 case AddIndexEntryRoot:
3068 case DeleteIndexEntryRoot:
3069 case SetIndexEntryVcnRoot:
3070 case UpdateFileNameRoot:
3071 case UpdateRecordDataRoot:
3072 case ZeroEndOfFileRecord:
3073 rno = vbo >> sbi->record_bits;
3074 inode = ilookup(sbi->sb, rno);
3076 mi = &ntfs_i(inode)->mi;
3077 } else if (op == InitializeFileRecordSegment) {
3078 mi = kzalloc(sizeof(struct mft_inode), GFP_NOFS);
3081 err = mi_format_new(mi, sbi, rno, 0, false);
3085 /* Read from disk. */
3086 err = mi_get(sbi, rno, &mi);
3092 if (op == DeallocateFileRecordSegment)
3093 goto skip_load_parent;
3095 if (InitializeFileRecordSegment != op) {
3096 if (rec->rhdr.sign == NTFS_BAAD_SIGNATURE)
3098 if (!check_lsn(&rec->rhdr, rlsn))
3100 if (!check_file_record(rec, NULL, sbi))
3102 attr = Add2Ptr(rec, roff);
3105 if (is_rec_base(rec) || InitializeFileRecordSegment == op) {
3107 goto skip_load_parent;
3110 rno_base = ino_get(&rec->parent_ref);
3111 inode_parent = ntfs_iget5(sbi->sb, &rec->parent_ref, NULL);
3112 if (IS_ERR(inode_parent))
3113 goto skip_load_parent;
3115 if (is_bad_inode(inode_parent)) {
3117 goto skip_load_parent;
3120 if (ni_load_mi_ex(ntfs_i(inode_parent), rno, &mi2_child)) {
3123 if (mi2_child->mrec != mi->mrec)
3124 memcpy(mi2_child->mrec, mi->mrec,
3132 inode = inode_parent;
3134 rec = mi2_child->mrec;
3135 attr = Add2Ptr(rec, roff);
3139 inode_parent = NULL;
3143 * Process attributes, as described by the current log record.
3145 case UpdateNonresidentValue:
3146 case AddIndexEntryAllocation:
3147 case DeleteIndexEntryAllocation:
3148 case WriteEndOfIndexBuffer:
3149 case SetIndexEntryVcnAllocation:
3150 case UpdateFileNameAllocation:
3151 case SetBitsInNonresidentBitMap:
3152 case ClearBitsInNonresidentBitMap:
3153 case UpdateRecordDataAllocation:
3155 bytes = UpdateNonresidentValue == op ? dlen : 0;
3156 lco = (u64)le16_to_cpu(lrh->lcns_follow) << sbi->cluster_bits;
3158 if (attr->type == ATTR_ALLOC) {
3159 t32 = le32_to_cpu(oe->bytes_per_index);
3168 if (attr->type == ATTR_ALLOC)
3169 bytes = (bytes + 511) & ~511; // align
3171 buffer_le = kmalloc(bytes, GFP_NOFS);
3175 err = ntfs_read_run_nb(sbi, oa->run1, vbo, buffer_le, bytes,
3180 if (attr->type == ATTR_ALLOC && *(int *)buffer_le)
3181 ntfs_fix_post_read(buffer_le, bytes, false);
3188 /* Big switch to do operation. */
3190 case InitializeFileRecordSegment:
3191 if (roff + dlen > record_size)
3194 memcpy(Add2Ptr(rec, roff), data, dlen);
3198 case DeallocateFileRecordSegment:
3199 clear_rec_inuse(rec);
3200 le16_add_cpu(&rec->seq, 1);
3204 case WriteEndOfFileRecordSegment:
3205 attr2 = (struct ATTRIB *)data;
3206 if (!check_if_attr(rec, lrh) || roff + dlen > record_size)
3209 memmove(attr, attr2, dlen);
3210 rec->used = cpu_to_le32(ALIGN(roff + dlen, 8));
3215 case CreateAttribute:
3216 attr2 = (struct ATTRIB *)data;
3217 asize = le32_to_cpu(attr2->size);
3218 used = le32_to_cpu(rec->used);
3220 if (!check_if_attr(rec, lrh) || dlen < SIZEOF_RESIDENT ||
3221 !IS_ALIGNED(asize, 8) ||
3222 Add2Ptr(attr2, asize) > Add2Ptr(lrh, rec_len) ||
3223 dlen > record_size - used) {
3227 memmove(Add2Ptr(attr, asize), attr, used - roff);
3228 memcpy(attr, attr2, asize);
3230 rec->used = cpu_to_le32(used + asize);
3231 id = le16_to_cpu(rec->next_attr_id);
3232 id2 = le16_to_cpu(attr2->id);
3234 rec->next_attr_id = cpu_to_le16(id2 + 1);
3235 if (is_attr_indexed(attr))
3236 le16_add_cpu(&rec->hard_links, 1);
3238 oa2 = find_loaded_attr(log, attr, rno_base);
3240 void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3243 // run_close(oa2->run1);
3252 case DeleteAttribute:
3253 asize = le32_to_cpu(attr->size);
3254 used = le32_to_cpu(rec->used);
3256 if (!check_if_attr(rec, lrh))
3259 rec->used = cpu_to_le32(used - asize);
3260 if (is_attr_indexed(attr))
3261 le16_add_cpu(&rec->hard_links, -1);
3263 memmove(attr, Add2Ptr(attr, asize), used - asize - roff);
3268 case UpdateResidentValue:
3269 nsize = aoff + dlen;
3271 if (!check_if_attr(rec, lrh))
3274 asize = le32_to_cpu(attr->size);
3275 used = le32_to_cpu(rec->used);
3277 if (lrh->redo_len == lrh->undo_len) {
3283 if (nsize > asize && nsize - asize > record_size - used)
3286 nsize = ALIGN(nsize, 8);
3287 data_off = le16_to_cpu(attr->res.data_off);
3289 if (nsize < asize) {
3290 memmove(Add2Ptr(attr, aoff), data, dlen);
3291 data = NULL; // To skip below memmove().
3294 memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
3295 used - le16_to_cpu(lrh->record_off) - asize);
3297 rec->used = cpu_to_le32(used + nsize - asize);
3298 attr->size = cpu_to_le32(nsize);
3299 attr->res.data_size = cpu_to_le32(aoff + dlen - data_off);
3303 memmove(Add2Ptr(attr, aoff), data, dlen);
3305 oa2 = find_loaded_attr(log, attr, rno_base);
3307 void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3310 // run_close(&oa2->run0);
3311 oa2->run1 = &oa2->run0;
3320 case UpdateMappingPairs:
3321 nsize = aoff + dlen;
3322 asize = le32_to_cpu(attr->size);
3323 used = le32_to_cpu(rec->used);
3325 if (!check_if_attr(rec, lrh) || !attr->non_res ||
3326 aoff < le16_to_cpu(attr->nres.run_off) || aoff > asize ||
3327 (nsize > asize && nsize - asize > record_size - used)) {
3331 nsize = ALIGN(nsize, 8);
3333 memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
3334 used - le16_to_cpu(lrh->record_off) - asize);
3335 rec->used = cpu_to_le32(used + nsize - asize);
3336 attr->size = cpu_to_le32(nsize);
3337 memmove(Add2Ptr(attr, aoff), data, dlen);
3339 if (run_get_highest_vcn(le64_to_cpu(attr->nres.svcn),
3340 attr_run(attr), &t64)) {
3344 attr->nres.evcn = cpu_to_le64(t64);
3345 oa2 = find_loaded_attr(log, attr, rno_base);
3346 if (oa2 && oa2->attr->non_res)
3347 oa2->attr->nres.evcn = attr->nres.evcn;
3352 case SetNewAttributeSizes:
3354 if (!check_if_attr(rec, lrh) || !attr->non_res)
3357 attr->nres.alloc_size = new_sz->alloc_size;
3358 attr->nres.data_size = new_sz->data_size;
3359 attr->nres.valid_size = new_sz->valid_size;
3361 if (dlen >= sizeof(struct NEW_ATTRIBUTE_SIZES))
3362 attr->nres.total_size = new_sz->total_size;
3364 oa2 = find_loaded_attr(log, attr, rno_base);
3366 void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3376 case AddIndexEntryRoot:
3377 e = (struct NTFS_DE *)data;
3378 esize = le16_to_cpu(e->size);
3379 root = resident_data(attr);
3381 used = le32_to_cpu(hdr->used);
3383 if (!check_if_index_root(rec, lrh) ||
3384 !check_if_root_index(attr, hdr, lrh) ||
3385 Add2Ptr(data, esize) > Add2Ptr(lrh, rec_len) ||
3386 esize > le32_to_cpu(rec->total) - le32_to_cpu(rec->used)) {
3390 e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3392 change_attr_size(rec, attr, le32_to_cpu(attr->size) + esize);
3394 memmove(Add2Ptr(e1, esize), e1,
3395 PtrOffset(e1, Add2Ptr(hdr, used)));
3396 memmove(e1, e, esize);
3398 le32_add_cpu(&attr->res.data_size, esize);
3399 hdr->used = cpu_to_le32(used + esize);
3400 le32_add_cpu(&hdr->total, esize);
3405 case DeleteIndexEntryRoot:
3406 root = resident_data(attr);
3408 used = le32_to_cpu(hdr->used);
3410 if (!check_if_index_root(rec, lrh) ||
3411 !check_if_root_index(attr, hdr, lrh)) {
3415 e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3416 esize = le16_to_cpu(e1->size);
3417 e2 = Add2Ptr(e1, esize);
3419 memmove(e1, e2, PtrOffset(e2, Add2Ptr(hdr, used)));
3421 le32_sub_cpu(&attr->res.data_size, esize);
3422 hdr->used = cpu_to_le32(used - esize);
3423 le32_sub_cpu(&hdr->total, esize);
3425 change_attr_size(rec, attr, le32_to_cpu(attr->size) - esize);
3430 case SetIndexEntryVcnRoot:
3431 root = resident_data(attr);
3434 if (!check_if_index_root(rec, lrh) ||
3435 !check_if_root_index(attr, hdr, lrh)) {
3439 e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3441 de_set_vbn_le(e, *(__le64 *)data);
3445 case UpdateFileNameRoot:
3446 root = resident_data(attr);
3449 if (!check_if_index_root(rec, lrh) ||
3450 !check_if_root_index(attr, hdr, lrh)) {
3454 e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3455 fname = (struct ATTR_FILE_NAME *)(e + 1);
3456 memmove(&fname->dup, data, sizeof(fname->dup)); //
3460 case UpdateRecordDataRoot:
3461 root = resident_data(attr);
3464 if (!check_if_index_root(rec, lrh) ||
3465 !check_if_root_index(attr, hdr, lrh)) {
3469 e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3471 memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
3476 case ZeroEndOfFileRecord:
3477 if (roff + dlen > record_size)
3480 memset(attr, 0, dlen);
3484 case UpdateNonresidentValue:
3485 if (lco < cbo + roff + dlen)
3488 memcpy(Add2Ptr(buffer_le, roff), data, dlen);
3491 if (attr->type == ATTR_ALLOC)
3492 ntfs_fix_pre_write(buffer_le, bytes);
3495 case AddIndexEntryAllocation:
3496 ib = Add2Ptr(buffer_le, roff);
3499 esize = le16_to_cpu(e->size);
3500 e1 = Add2Ptr(ib, aoff);
3502 if (is_baad(&ib->rhdr))
3504 if (!check_lsn(&ib->rhdr, rlsn))
3507 used = le32_to_cpu(hdr->used);
3509 if (!check_index_buffer(ib, bytes) ||
3510 !check_if_alloc_index(hdr, aoff) ||
3511 Add2Ptr(e, esize) > Add2Ptr(lrh, rec_len) ||
3512 used + esize > le32_to_cpu(hdr->total)) {
3516 memmove(Add2Ptr(e1, esize), e1,
3517 PtrOffset(e1, Add2Ptr(hdr, used)));
3518 memcpy(e1, e, esize);
3520 hdr->used = cpu_to_le32(used + esize);
3524 ntfs_fix_pre_write(&ib->rhdr, bytes);
3527 case DeleteIndexEntryAllocation:
3528 ib = Add2Ptr(buffer_le, roff);
3530 e = Add2Ptr(ib, aoff);
3531 esize = le16_to_cpu(e->size);
3533 if (is_baad(&ib->rhdr))
3535 if (!check_lsn(&ib->rhdr, rlsn))
3538 if (!check_index_buffer(ib, bytes) ||
3539 !check_if_alloc_index(hdr, aoff)) {
3543 e1 = Add2Ptr(e, esize);
3545 used = le32_to_cpu(hdr->used);
3547 memmove(e, e1, PtrOffset(e1, Add2Ptr(hdr, used)));
3549 hdr->used = cpu_to_le32(used - nsize);
3553 ntfs_fix_pre_write(&ib->rhdr, bytes);
3556 case WriteEndOfIndexBuffer:
3557 ib = Add2Ptr(buffer_le, roff);
3559 e = Add2Ptr(ib, aoff);
3561 if (is_baad(&ib->rhdr))
3563 if (!check_lsn(&ib->rhdr, rlsn))
3565 if (!check_index_buffer(ib, bytes) ||
3566 !check_if_alloc_index(hdr, aoff) ||
3567 aoff + dlen > offsetof(struct INDEX_BUFFER, ihdr) +
3568 le32_to_cpu(hdr->total)) {
3572 hdr->used = cpu_to_le32(dlen + PtrOffset(hdr, e));
3573 memmove(e, data, dlen);
3576 ntfs_fix_pre_write(&ib->rhdr, bytes);
3579 case SetIndexEntryVcnAllocation:
3580 ib = Add2Ptr(buffer_le, roff);
3582 e = Add2Ptr(ib, aoff);
3584 if (is_baad(&ib->rhdr))
3587 if (!check_lsn(&ib->rhdr, rlsn))
3589 if (!check_index_buffer(ib, bytes) ||
3590 !check_if_alloc_index(hdr, aoff)) {
3594 de_set_vbn_le(e, *(__le64 *)data);
3597 ntfs_fix_pre_write(&ib->rhdr, bytes);
3600 case UpdateFileNameAllocation:
3601 ib = Add2Ptr(buffer_le, roff);
3603 e = Add2Ptr(ib, aoff);
3605 if (is_baad(&ib->rhdr))
3608 if (!check_lsn(&ib->rhdr, rlsn))
3610 if (!check_index_buffer(ib, bytes) ||
3611 !check_if_alloc_index(hdr, aoff)) {
3615 fname = (struct ATTR_FILE_NAME *)(e + 1);
3616 memmove(&fname->dup, data, sizeof(fname->dup));
3619 ntfs_fix_pre_write(&ib->rhdr, bytes);
3622 case SetBitsInNonresidentBitMap:
3624 le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
3625 bmp_bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
3627 if (cbo + (bmp_off + 7) / 8 > lco ||
3628 cbo + ((bmp_off + bmp_bits + 7) / 8) > lco) {
3632 __bitmap_set(Add2Ptr(buffer_le, roff), bmp_off, bmp_bits);
3636 case ClearBitsInNonresidentBitMap:
3638 le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
3639 bmp_bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
3641 if (cbo + (bmp_off + 7) / 8 > lco ||
3642 cbo + ((bmp_off + bmp_bits + 7) / 8) > lco) {
3646 __bitmap_clear(Add2Ptr(buffer_le, roff), bmp_off, bmp_bits);
3650 case UpdateRecordDataAllocation:
3651 ib = Add2Ptr(buffer_le, roff);
3653 e = Add2Ptr(ib, aoff);
3655 if (is_baad(&ib->rhdr))
3658 if (!check_lsn(&ib->rhdr, rlsn))
3660 if (!check_index_buffer(ib, bytes) ||
3661 !check_if_alloc_index(hdr, aoff)) {
3665 memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
3668 ntfs_fix_pre_write(&ib->rhdr, bytes);
3676 __le64 t64 = cpu_to_le64(*rlsn);
3679 rec->rhdr.lsn = t64;
3684 if (mi && mi->dirty) {
3685 err = mi_write(mi, 0);
3692 err = ntfs_sb_write_run(sbi, oa->run1, vbo, buffer_le, bytes, 0);
3701 else if (mi != mi2_child)
3709 log->set_dirty = true;
3714 * log_replay - Replays log and empties it.
3716 * This function is called during mount operation.
3717 * It replays log and empties it.
3718 * Initialized is set false if logfile contains '-1'.
3720 int log_replay(struct ntfs_inode *ni, bool *initialized)
3723 struct ntfs_sb_info *sbi = ni->mi.sbi;
3724 struct ntfs_log *log;
3726 struct restart_info rst_info, rst_info2;
3727 u64 rec_lsn, ra_lsn, checkpt_lsn = 0, rlsn = 0;
3728 struct ATTR_NAME_ENTRY *attr_names = NULL;
3729 struct ATTR_NAME_ENTRY *ane;
3730 struct RESTART_TABLE *dptbl = NULL;
3731 struct RESTART_TABLE *trtbl = NULL;
3732 const struct RESTART_TABLE *rt;
3733 struct RESTART_TABLE *oatbl = NULL;
3734 struct inode *inode;
3735 struct OpenAttr *oa;
3736 struct ntfs_inode *ni_oe;
3737 struct ATTRIB *attr = NULL;
3738 u64 size, vcn, undo_next_lsn;
3739 CLST rno, lcn, lcn0, len0, clen;
3741 struct NTFS_RESTART *rst = NULL;
3742 struct lcb *lcb = NULL;
3743 struct OPEN_ATTR_ENRTY *oe;
3744 struct TRANSACTION_ENTRY *tr;
3745 struct DIR_PAGE_ENTRY *dp;
3746 u32 i, bytes_per_attr_entry;
3747 u32 l_size = ni->vfs_inode.i_size;
3748 u32 orig_file_size = l_size;
3749 u32 page_size, vbo, tail, off, dlen;
3750 u32 saved_len, rec_len, transact_id;
3751 bool use_second_page;
3752 struct RESTART_AREA *ra2, *ra = NULL;
3753 struct CLIENT_REC *ca, *cr;
3755 struct RESTART_HDR *rh;
3756 const struct LFS_RECORD_HDR *frh;
3757 const struct LOG_REC_HDR *lrh;
3759 bool is_ro = sb_rdonly(sbi->sb);
3764 /* Get the size of page. NOTE: To replay we can use default page. */
3765 #if PAGE_SIZE >= DefaultLogPageSize && PAGE_SIZE <= DefaultLogPageSize * 2
3766 page_size = norm_file_page(PAGE_SIZE, &l_size, true);
3768 page_size = norm_file_page(PAGE_SIZE, &l_size, false);
3773 log = kzalloc(sizeof(struct ntfs_log), GFP_NOFS);
3777 memset(&rst_info, 0, sizeof(struct restart_info));
3780 log->l_size = l_size;
3781 log->one_page_buf = kmalloc(page_size, GFP_NOFS);
3782 if (!log->one_page_buf) {
3787 log->page_size = page_size;
3788 log->page_mask = page_size - 1;
3789 log->page_bits = blksize_bits(page_size);
3791 /* Look for a restart area on the disk. */
3792 err = log_read_rst(log, l_size, true, &rst_info);
3796 /* remember 'initialized' */
3797 *initialized = rst_info.initialized;
3799 if (!rst_info.restart) {
3800 if (rst_info.initialized) {
3801 /* No restart area but the file is not initialized. */
3806 log_init_pg_hdr(log, page_size, page_size, 1, 1);
3807 log_create(log, l_size, 0, get_random_u32(), false, false);
3811 ra = log_create_ra(log);
3817 log->init_ra = true;
3823 * If the restart offset above wasn't zero then we won't
3824 * look for a second restart.
3827 goto check_restart_area;
3829 memset(&rst_info2, 0, sizeof(struct restart_info));
3830 err = log_read_rst(log, l_size, false, &rst_info2);
3834 /* Determine which restart area to use. */
3835 if (!rst_info2.restart || rst_info2.last_lsn <= rst_info.last_lsn)
3836 goto use_first_page;
3838 use_second_page = true;
3840 if (rst_info.chkdsk_was_run && page_size != rst_info.vbo) {
3841 struct RECORD_PAGE_HDR *sp = NULL;
3844 if (!read_log_page(log, page_size, &sp, &usa_error) &&
3845 sp->rhdr.sign == NTFS_CHKD_SIGNATURE) {
3846 use_second_page = false;
3851 if (use_second_page) {
3852 kfree(rst_info.r_page);
3853 memcpy(&rst_info, &rst_info2, sizeof(struct restart_info));
3854 rst_info2.r_page = NULL;
3858 kfree(rst_info2.r_page);
3862 * If the restart area is at offset 0, we want
3863 * to write the second restart area first.
3865 log->init_ra = !!rst_info.vbo;
3867 /* If we have a valid page then grab a pointer to the restart area. */
3868 ra2 = rst_info.valid_page
3869 ? Add2Ptr(rst_info.r_page,
3870 le16_to_cpu(rst_info.r_page->ra_off))
3873 if (rst_info.chkdsk_was_run ||
3874 (ra2 && ra2->client_idx[1] == LFS_NO_CLIENT_LE)) {
3875 bool wrapped = false;
3876 bool use_multi_page = false;
3879 /* Do some checks based on whether we have a valid log page. */
3880 if (!rst_info.valid_page) {
3881 open_log_count = get_random_u32();
3882 goto init_log_instance;
3884 open_log_count = le32_to_cpu(ra2->open_log_count);
3887 * If the restart page size isn't changing then we want to
3888 * check how much work we need to do.
3890 if (page_size != le32_to_cpu(rst_info.r_page->sys_page_size))
3891 goto init_log_instance;
3894 log_init_pg_hdr(log, page_size, page_size, 1, 1);
3896 log_create(log, l_size, rst_info.last_lsn, open_log_count,
3897 wrapped, use_multi_page);
3899 ra = log_create_ra(log);
3906 /* Put the restart areas and initialize
3907 * the log file as required.
3918 * If the log page or the system page sizes have changed, we can't
3919 * use the log file. We must use the system page size instead of the
3920 * default size if there is not a clean shutdown.
3922 t32 = le32_to_cpu(rst_info.r_page->sys_page_size);
3923 if (page_size != t32) {
3924 l_size = orig_file_size;
3926 norm_file_page(t32, &l_size, t32 == DefaultLogPageSize);
3929 if (page_size != t32 ||
3930 page_size != le32_to_cpu(rst_info.r_page->page_size)) {
3935 /* If the file size has shrunk then we won't mount it. */
3936 if (l_size < le64_to_cpu(ra2->l_size)) {
3941 log_init_pg_hdr(log, page_size, page_size,
3942 le16_to_cpu(rst_info.r_page->major_ver),
3943 le16_to_cpu(rst_info.r_page->minor_ver));
3945 log->l_size = le64_to_cpu(ra2->l_size);
3946 log->seq_num_bits = le32_to_cpu(ra2->seq_num_bits);
3947 log->file_data_bits = sizeof(u64) * 8 - log->seq_num_bits;
3948 log->seq_num_mask = (8 << log->file_data_bits) - 1;
3949 log->last_lsn = le64_to_cpu(ra2->current_lsn);
3950 log->seq_num = log->last_lsn >> log->file_data_bits;
3951 log->ra_off = le16_to_cpu(rst_info.r_page->ra_off);
3952 log->restart_size = log->sys_page_size - log->ra_off;
3953 log->record_header_len = le16_to_cpu(ra2->rec_hdr_len);
3954 log->ra_size = le16_to_cpu(ra2->ra_len);
3955 log->data_off = le16_to_cpu(ra2->data_off);
3956 log->data_size = log->page_size - log->data_off;
3957 log->reserved = log->data_size - log->record_header_len;
3959 vbo = lsn_to_vbo(log, log->last_lsn);
3961 if (vbo < log->first_page) {
3962 /* This is a pseudo lsn. */
3963 log->l_flags |= NTFSLOG_NO_LAST_LSN;
3964 log->next_page = log->first_page;
3968 /* Find the end of this log record. */
3969 off = final_log_off(log, log->last_lsn,
3970 le32_to_cpu(ra2->last_lsn_data_len));
3972 /* If we wrapped the file then increment the sequence number. */
3975 log->l_flags |= NTFSLOG_WRAPPED;
3978 /* Now compute the next log page to use. */
3979 vbo &= ~log->sys_page_mask;
3980 tail = log->page_size - (off & log->page_mask) - 1;
3983 *If we can fit another log record on the page,
3984 * move back a page the log file.
3986 if (tail >= log->record_header_len) {
3987 log->l_flags |= NTFSLOG_REUSE_TAIL;
3988 log->next_page = vbo;
3990 log->next_page = next_page_off(log, vbo);
3995 * Find the oldest client lsn. Use the last
3996 * flushed lsn as a starting point.
3998 log->oldest_lsn = log->last_lsn;
3999 oldest_client_lsn(Add2Ptr(ra2, le16_to_cpu(ra2->client_off)),
4000 ra2->client_idx[1], &log->oldest_lsn);
4001 log->oldest_lsn_off = lsn_to_vbo(log, log->oldest_lsn);
4003 if (log->oldest_lsn_off < log->first_page)
4004 log->l_flags |= NTFSLOG_NO_OLDEST_LSN;
4006 if (!(ra2->flags & RESTART_SINGLE_PAGE_IO))
4007 log->l_flags |= NTFSLOG_WRAPPED | NTFSLOG_MULTIPLE_PAGE_IO;
4009 log->current_openlog_count = le32_to_cpu(ra2->open_log_count);
4010 log->total_avail_pages = log->l_size - log->first_page;
4011 log->total_avail = log->total_avail_pages >> log->page_bits;
4012 log->max_current_avail = log->total_avail * log->reserved;
4013 log->total_avail = log->total_avail * log->data_size;
4015 log->current_avail = current_log_avail(log);
4017 ra = kzalloc(log->restart_size, GFP_NOFS);
4024 t16 = le16_to_cpu(ra2->client_off);
4025 if (t16 == offsetof(struct RESTART_AREA, clients)) {
4026 memcpy(ra, ra2, log->ra_size);
4028 memcpy(ra, ra2, offsetof(struct RESTART_AREA, clients));
4029 memcpy(ra->clients, Add2Ptr(ra2, t16),
4030 le16_to_cpu(ra2->ra_len) - t16);
4032 log->current_openlog_count = get_random_u32();
4033 ra->open_log_count = cpu_to_le32(log->current_openlog_count);
4034 log->ra_size = offsetof(struct RESTART_AREA, clients) +
4035 sizeof(struct CLIENT_REC);
4037 cpu_to_le16(offsetof(struct RESTART_AREA, clients));
4038 ra->ra_len = cpu_to_le16(log->ra_size);
4041 le32_add_cpu(&ra->open_log_count, 1);
4043 /* Now we need to walk through looking for the last lsn. */
4044 err = last_log_lsn(log);
4048 log->current_avail = current_log_avail(log);
4050 /* Remember which restart area to write first. */
4051 log->init_ra = rst_info.vbo;
4054 /* 1.0, 1.1, 2.0 log->major_ver/minor_ver - short values. */
4055 switch ((log->major_ver << 16) + log->minor_ver) {
4061 ntfs_warn(sbi->sb, "\x24LogFile version %d.%d is not supported",
4062 log->major_ver, log->minor_ver);
4064 log->set_dirty = true;
4068 /* One client "NTFS" per logfile. */
4069 ca = Add2Ptr(ra, le16_to_cpu(ra->client_off));
4071 for (client = ra->client_idx[1];; client = cr->next_client) {
4072 if (client == LFS_NO_CLIENT_LE) {
4073 /* Insert "NTFS" client LogFile. */
4074 client = ra->client_idx[0];
4075 if (client == LFS_NO_CLIENT_LE) {
4080 t16 = le16_to_cpu(client);
4083 remove_client(ca, cr, &ra->client_idx[0]);
4085 cr->restart_lsn = 0;
4086 cr->oldest_lsn = cpu_to_le64(log->oldest_lsn);
4087 cr->name_bytes = cpu_to_le32(8);
4088 cr->name[0] = cpu_to_le16('N');
4089 cr->name[1] = cpu_to_le16('T');
4090 cr->name[2] = cpu_to_le16('F');
4091 cr->name[3] = cpu_to_le16('S');
4093 add_client(ca, t16, &ra->client_idx[1]);
4097 cr = ca + le16_to_cpu(client);
4099 if (cpu_to_le32(8) == cr->name_bytes &&
4100 cpu_to_le16('N') == cr->name[0] &&
4101 cpu_to_le16('T') == cr->name[1] &&
4102 cpu_to_le16('F') == cr->name[2] &&
4103 cpu_to_le16('S') == cr->name[3])
4107 /* Update the client handle with the client block information. */
4108 log->client_id.seq_num = cr->seq_num;
4109 log->client_id.client_idx = client;
4111 err = read_rst_area(log, &rst, &ra_lsn);
4118 bytes_per_attr_entry = !rst->major_ver ? 0x2C : 0x28;
4120 checkpt_lsn = le64_to_cpu(rst->check_point_start);
4122 checkpt_lsn = ra_lsn;
4124 /* Allocate and Read the Transaction Table. */
4125 if (!rst->transact_table_len)
4126 goto check_dirty_page_table;
4128 t64 = le64_to_cpu(rst->transact_table_lsn);
4129 err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4135 rec_len = le32_to_cpu(frh->client_data_len);
4137 if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4138 bytes_per_attr_entry)) {
4143 t16 = le16_to_cpu(lrh->redo_off);
4145 rt = Add2Ptr(lrh, t16);
4146 t32 = rec_len - t16;
4148 /* Now check that this is a valid restart table. */
4149 if (!check_rstbl(rt, t32)) {
4154 trtbl = kmemdup(rt, t32, GFP_NOFS);
4163 check_dirty_page_table:
4164 /* The next record back should be the Dirty Pages Table. */
4165 if (!rst->dirty_pages_len)
4166 goto check_attribute_names;
4168 t64 = le64_to_cpu(rst->dirty_pages_table_lsn);
4169 err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4175 rec_len = le32_to_cpu(frh->client_data_len);
4177 if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4178 bytes_per_attr_entry)) {
4183 t16 = le16_to_cpu(lrh->redo_off);
4185 rt = Add2Ptr(lrh, t16);
4186 t32 = rec_len - t16;
4188 /* Now check that this is a valid restart table. */
4189 if (!check_rstbl(rt, t32)) {
4194 dptbl = kmemdup(rt, t32, GFP_NOFS);
4200 /* Convert Ra version '0' into version '1'. */
4205 while ((dp = enum_rstbl(dptbl, dp))) {
4206 struct DIR_PAGE_ENTRY_32 *dp0 = (struct DIR_PAGE_ENTRY_32 *)dp;
4207 // NOTE: Danger. Check for of boundary.
4208 memmove(&dp->vcn, &dp0->vcn_low,
4210 le32_to_cpu(dp->lcns_follow) * sizeof(u64));
4218 * Go through the table and remove the duplicates,
4219 * remembering the oldest lsn values.
4221 if (sbi->cluster_size <= log->page_size)
4222 goto trace_dp_table;
4225 while ((dp = enum_rstbl(dptbl, dp))) {
4226 struct DIR_PAGE_ENTRY *next = dp;
4228 while ((next = enum_rstbl(dptbl, next))) {
4229 if (next->target_attr == dp->target_attr &&
4230 next->vcn == dp->vcn) {
4231 if (le64_to_cpu(next->oldest_lsn) <
4232 le64_to_cpu(dp->oldest_lsn)) {
4233 dp->oldest_lsn = next->oldest_lsn;
4236 free_rsttbl_idx(dptbl, PtrOffset(dptbl, next));
4241 check_attribute_names:
4242 /* The next record should be the Attribute Names. */
4243 if (!rst->attr_names_len)
4244 goto check_attr_table;
4246 t64 = le64_to_cpu(rst->attr_names_lsn);
4247 err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4253 rec_len = le32_to_cpu(frh->client_data_len);
4255 if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4256 bytes_per_attr_entry)) {
4261 t32 = lrh_length(lrh);
4264 attr_names = kmemdup(Add2Ptr(lrh, t32), rec_len, GFP_NOFS);
4274 /* The next record should be the attribute Table. */
4275 if (!rst->open_attr_len)
4276 goto check_attribute_names2;
4278 t64 = le64_to_cpu(rst->open_attr_table_lsn);
4279 err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4285 rec_len = le32_to_cpu(frh->client_data_len);
4287 if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4288 bytes_per_attr_entry)) {
4293 t16 = le16_to_cpu(lrh->redo_off);
4295 rt = Add2Ptr(lrh, t16);
4296 t32 = rec_len - t16;
4298 if (!check_rstbl(rt, t32)) {
4303 oatbl = kmemdup(rt, t32, GFP_NOFS);
4309 log->open_attr_tbl = oatbl;
4311 /* Clear all of the Attr pointers. */
4313 while ((oe = enum_rstbl(oatbl, oe))) {
4314 if (!rst->major_ver) {
4315 struct OPEN_ATTR_ENRTY_32 oe0;
4317 /* Really 'oe' points to OPEN_ATTR_ENRTY_32. */
4318 memcpy(&oe0, oe, SIZEOF_OPENATTRIBUTEENTRY0);
4320 oe->bytes_per_index = oe0.bytes_per_index;
4321 oe->type = oe0.type;
4322 oe->is_dirty_pages = oe0.is_dirty_pages;
4325 oe->open_record_lsn = oe0.open_record_lsn;
4328 oe->is_attr_name = 0;
4335 check_attribute_names2:
4336 if (!rst->attr_names_len)
4337 goto trace_attribute_table;
4341 goto trace_attribute_table;
4343 /* TODO: Clear table on exit! */
4344 oe = Add2Ptr(oatbl, le16_to_cpu(ane->off));
4345 t16 = le16_to_cpu(ane->name_bytes);
4346 oe->name_len = t16 / sizeof(short);
4347 oe->ptr = ane->name;
4348 oe->is_attr_name = 2;
4349 ane = Add2Ptr(ane, sizeof(struct ATTR_NAME_ENTRY) + t16);
4352 trace_attribute_table:
4354 * If the checkpt_lsn is zero, then this is a freshly
4355 * formatted disk and we have no work to do.
4363 oatbl = init_rsttbl(bytes_per_attr_entry, 8);
4370 log->open_attr_tbl = oatbl;
4372 /* Start the analysis pass from the Checkpoint lsn. */
4373 rec_lsn = checkpt_lsn;
4375 /* Read the first lsn. */
4376 err = read_log_rec_lcb(log, checkpt_lsn, lcb_ctx_next, &lcb);
4380 /* Loop to read all subsequent records to the end of the log file. */
4381 next_log_record_analyze:
4382 err = read_next_log_rec(log, lcb, &rec_lsn);
4387 goto end_log_records_enumerate;
4390 transact_id = le32_to_cpu(frh->transact_id);
4391 rec_len = le32_to_cpu(frh->client_data_len);
4394 if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
4400 * The first lsn after the previous lsn remembered
4401 * the checkpoint is the first candidate for the rlsn.
4406 if (LfsClientRecord != frh->record_type)
4407 goto next_log_record_analyze;
4410 * Now update the Transaction Table for this transaction. If there
4411 * is no entry present or it is unallocated we allocate the entry.
4414 trtbl = init_rsttbl(sizeof(struct TRANSACTION_ENTRY),
4415 INITIAL_NUMBER_TRANSACTIONS);
4422 tr = Add2Ptr(trtbl, transact_id);
4424 if (transact_id >= bytes_per_rt(trtbl) ||
4425 tr->next != RESTART_ENTRY_ALLOCATED_LE) {
4426 tr = alloc_rsttbl_from_idx(&trtbl, transact_id);
4431 tr->transact_state = TransactionActive;
4432 tr->first_lsn = cpu_to_le64(rec_lsn);
4435 tr->prev_lsn = tr->undo_next_lsn = cpu_to_le64(rec_lsn);
4438 * If this is a compensation log record, then change
4439 * the undo_next_lsn to be the undo_next_lsn of this record.
4441 if (lrh->undo_op == cpu_to_le16(CompensationLogRecord))
4442 tr->undo_next_lsn = frh->client_undo_next_lsn;
4444 /* Dispatch to handle log record depending on type. */
4445 switch (le16_to_cpu(lrh->redo_op)) {
4446 case InitializeFileRecordSegment:
4447 case DeallocateFileRecordSegment:
4448 case WriteEndOfFileRecordSegment:
4449 case CreateAttribute:
4450 case DeleteAttribute:
4451 case UpdateResidentValue:
4452 case UpdateNonresidentValue:
4453 case UpdateMappingPairs:
4454 case SetNewAttributeSizes:
4455 case AddIndexEntryRoot:
4456 case DeleteIndexEntryRoot:
4457 case AddIndexEntryAllocation:
4458 case DeleteIndexEntryAllocation:
4459 case WriteEndOfIndexBuffer:
4460 case SetIndexEntryVcnRoot:
4461 case SetIndexEntryVcnAllocation:
4462 case UpdateFileNameRoot:
4463 case UpdateFileNameAllocation:
4464 case SetBitsInNonresidentBitMap:
4465 case ClearBitsInNonresidentBitMap:
4466 case UpdateRecordDataRoot:
4467 case UpdateRecordDataAllocation:
4468 case ZeroEndOfFileRecord:
4469 t16 = le16_to_cpu(lrh->target_attr);
4470 t64 = le64_to_cpu(lrh->target_vcn);
4471 dp = find_dp(dptbl, t16, t64);
4477 * Calculate the number of clusters per page the system
4478 * which wrote the checkpoint, possibly creating the table.
4481 t32 = (le16_to_cpu(dptbl->size) -
4482 sizeof(struct DIR_PAGE_ENTRY)) /
4485 t32 = log->clst_per_page;
4487 dptbl = init_rsttbl(struct_size(dp, page_lcns, t32),
4495 dp = alloc_rsttbl_idx(&dptbl);
4500 dp->target_attr = cpu_to_le32(t16);
4501 dp->transfer_len = cpu_to_le32(t32 << sbi->cluster_bits);
4502 dp->lcns_follow = cpu_to_le32(t32);
4503 dp->vcn = cpu_to_le64(t64 & ~((u64)t32 - 1));
4504 dp->oldest_lsn = cpu_to_le64(rec_lsn);
4508 * Copy the Lcns from the log record into the Dirty Page Entry.
4509 * TODO: For different page size support, must somehow make
4510 * whole routine a loop, case Lcns do not fit below.
4512 t16 = le16_to_cpu(lrh->lcns_follow);
4513 for (i = 0; i < t16; i++) {
4514 size_t j = (size_t)(le64_to_cpu(lrh->target_vcn) -
4515 le64_to_cpu(dp->vcn));
4516 dp->page_lcns[j + i] = lrh->page_lcns[i];
4519 goto next_log_record_analyze;
4521 case DeleteDirtyClusters: {
4523 le16_to_cpu(lrh->redo_len) / sizeof(struct LCN_RANGE);
4524 const struct LCN_RANGE *r =
4525 Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
4527 /* Loop through all of the Lcn ranges this log record. */
4528 for (i = 0; i < range_count; i++, r++) {
4529 u64 lcn0 = le64_to_cpu(r->lcn);
4530 u64 lcn_e = lcn0 + le64_to_cpu(r->len) - 1;
4533 while ((dp = enum_rstbl(dptbl, dp))) {
4536 t32 = le32_to_cpu(dp->lcns_follow);
4537 for (j = 0; j < t32; j++) {
4538 t64 = le64_to_cpu(dp->page_lcns[j]);
4539 if (t64 >= lcn0 && t64 <= lcn_e)
4540 dp->page_lcns[j] = 0;
4544 goto next_log_record_analyze;
4548 case OpenNonresidentAttribute:
4549 t16 = le16_to_cpu(lrh->target_attr);
4550 if (t16 >= bytes_per_rt(oatbl)) {
4552 * Compute how big the table needs to be.
4553 * Add 10 extra entries for some cushion.
4555 u32 new_e = t16 / le16_to_cpu(oatbl->size);
4557 new_e += 10 - le16_to_cpu(oatbl->used);
4559 oatbl = extend_rsttbl(oatbl, new_e, ~0u);
4560 log->open_attr_tbl = oatbl;
4567 /* Point to the entry being opened. */
4568 oe = alloc_rsttbl_from_idx(&oatbl, t16);
4569 log->open_attr_tbl = oatbl;
4575 /* Initialize this entry from the log record. */
4576 t16 = le16_to_cpu(lrh->redo_off);
4577 if (!rst->major_ver) {
4578 /* Convert version '0' into version '1'. */
4579 struct OPEN_ATTR_ENRTY_32 *oe0 = Add2Ptr(lrh, t16);
4581 oe->bytes_per_index = oe0->bytes_per_index;
4582 oe->type = oe0->type;
4583 oe->is_dirty_pages = oe0->is_dirty_pages;
4584 oe->name_len = 0; //oe0.name_len;
4586 oe->open_record_lsn = oe0->open_record_lsn;
4588 memcpy(oe, Add2Ptr(lrh, t16), bytes_per_attr_entry);
4591 t16 = le16_to_cpu(lrh->undo_len);
4593 oe->ptr = kmalloc(t16, GFP_NOFS);
4598 oe->name_len = t16 / sizeof(short);
4600 Add2Ptr(lrh, le16_to_cpu(lrh->undo_off)), t16);
4601 oe->is_attr_name = 1;
4604 oe->is_attr_name = 0;
4607 goto next_log_record_analyze;
4610 t16 = le16_to_cpu(lrh->target_attr);
4611 t64 = le64_to_cpu(lrh->target_vcn);
4612 dp = find_dp(dptbl, t16, t64);
4614 size_t j = le64_to_cpu(lrh->target_vcn) -
4615 le64_to_cpu(dp->vcn);
4616 if (dp->page_lcns[j])
4617 dp->page_lcns[j] = lrh->page_lcns[0];
4619 goto next_log_record_analyze;
4621 case EndTopLevelAction:
4622 tr = Add2Ptr(trtbl, transact_id);
4623 tr->prev_lsn = cpu_to_le64(rec_lsn);
4624 tr->undo_next_lsn = frh->client_undo_next_lsn;
4625 goto next_log_record_analyze;
4627 case PrepareTransaction:
4628 tr = Add2Ptr(trtbl, transact_id);
4629 tr->transact_state = TransactionPrepared;
4630 goto next_log_record_analyze;
4632 case CommitTransaction:
4633 tr = Add2Ptr(trtbl, transact_id);
4634 tr->transact_state = TransactionCommitted;
4635 goto next_log_record_analyze;
4637 case ForgetTransaction:
4638 free_rsttbl_idx(trtbl, transact_id);
4639 goto next_log_record_analyze;
4642 case OpenAttributeTableDump:
4643 case AttributeNamesDump:
4644 case DirtyPageTableDump:
4645 case TransactionTableDump:
4646 /* The following cases require no action the Analysis Pass. */
4647 goto next_log_record_analyze;
4651 * All codes will be explicitly handled.
4652 * If we see a code we do not expect, then we are trouble.
4654 goto next_log_record_analyze;
4657 end_log_records_enumerate:
4662 * Scan the Dirty Page Table and Transaction Table for
4663 * the lowest lsn, and return it as the Redo lsn.
4666 while ((dp = enum_rstbl(dptbl, dp))) {
4667 t64 = le64_to_cpu(dp->oldest_lsn);
4668 if (t64 && t64 < rlsn)
4673 while ((tr = enum_rstbl(trtbl, tr))) {
4674 t64 = le64_to_cpu(tr->first_lsn);
4675 if (t64 && t64 < rlsn)
4680 * Only proceed if the Dirty Page Table or Transaction
4681 * table are not empty.
4683 if ((!dptbl || !dptbl->total) && (!trtbl || !trtbl->total))
4686 sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
4690 /* Reopen all of the attributes with dirty pages. */
4692 next_open_attribute:
4694 oe = enum_rstbl(oatbl, oe);
4698 goto next_dirty_page;
4701 oa = kzalloc(sizeof(struct OpenAttr), GFP_NOFS);
4707 inode = ntfs_iget5(sbi->sb, &oe->ref, NULL);
4711 if (is_bad_inode(inode)) {
4715 iput(&oa->ni->vfs_inode);
4719 attr = attr_create_nonres_log(sbi, oe->type, 0, oe->ptr,
4727 oa->run1 = &oa->run0;
4731 ni_oe = ntfs_i(inode);
4734 attr = ni_find_attr(ni_oe, NULL, NULL, oe->type, oe->ptr, oe->name_len,
4740 t32 = le32_to_cpu(attr->size);
4741 oa->attr = kmemdup(attr, t32, GFP_NOFS);
4745 if (!S_ISDIR(inode->i_mode)) {
4746 if (attr->type == ATTR_DATA && !attr->name_len) {
4747 oa->run1 = &ni_oe->file.run;
4751 if (attr->type == ATTR_ALLOC &&
4752 attr->name_len == ARRAY_SIZE(I30_NAME) &&
4753 !memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME))) {
4754 oa->run1 = &ni_oe->dir.alloc_run;
4759 if (attr->non_res) {
4760 u16 roff = le16_to_cpu(attr->nres.run_off);
4761 CLST svcn = le64_to_cpu(attr->nres.svcn);
4769 err = run_unpack(&oa->run0, sbi, inode->i_ino, svcn,
4770 le64_to_cpu(attr->nres.evcn), svcn,
4771 Add2Ptr(attr, roff), t32 - roff);
4779 oa->run1 = &oa->run0;
4783 if (oe->is_attr_name == 1)
4785 oe->is_attr_name = 0;
4787 oe->name_len = attr->name_len;
4789 goto next_open_attribute;
4792 * Now loop through the dirty page table to extract all of the Vcn/Lcn.
4793 * Mapping that we have, and insert it into the appropriate run.
4796 dp = enum_rstbl(dptbl, dp);
4800 oe = Add2Ptr(oatbl, le32_to_cpu(dp->target_attr));
4802 if (oe->next != RESTART_ENTRY_ALLOCATED_LE)
4803 goto next_dirty_page;
4807 goto next_dirty_page;
4810 next_dirty_page_vcn:
4812 if (i >= le32_to_cpu(dp->lcns_follow))
4813 goto next_dirty_page;
4815 vcn = le64_to_cpu(dp->vcn) + i;
4816 size = (vcn + 1) << sbi->cluster_bits;
4818 if (!dp->page_lcns[i])
4819 goto next_dirty_page_vcn;
4821 rno = ino_get(&oe->ref);
4822 if (rno <= MFT_REC_MIRR &&
4823 size < (MFT_REC_VOL + 1) * sbi->record_size &&
4824 oe->type == ATTR_DATA) {
4825 goto next_dirty_page_vcn;
4828 lcn = le64_to_cpu(dp->page_lcns[i]);
4830 if ((!run_lookup_entry(oa->run1, vcn, &lcn0, &len0, NULL) ||
4832 !run_add_entry(oa->run1, vcn, lcn, 1, false)) {
4837 t64 = le64_to_cpu(attr->nres.alloc_size);
4839 attr->nres.valid_size = attr->nres.data_size =
4840 attr->nres.alloc_size = cpu_to_le64(size);
4842 goto next_dirty_page_vcn;
4846 * Perform the Redo Pass, to restore all of the dirty pages to the same
4847 * contents that they had immediately before the crash. If the dirty
4848 * page table is empty, then we can skip the entire Redo Pass.
4850 if (!dptbl || !dptbl->total)
4851 goto do_undo_action;
4856 * Read the record at the Redo lsn, before falling
4857 * into common code to handle each record.
4859 err = read_log_rec_lcb(log, rlsn, lcb_ctx_next, &lcb);
4864 * Now loop to read all of our log records forwards, until
4865 * we hit the end of the file, cleaning up at the end.
4870 if (LfsClientRecord != frh->record_type)
4871 goto read_next_log_do_action;
4873 transact_id = le32_to_cpu(frh->transact_id);
4874 rec_len = le32_to_cpu(frh->client_data_len);
4877 if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
4882 /* Ignore log records that do not update pages. */
4883 if (lrh->lcns_follow)
4884 goto find_dirty_page;
4886 goto read_next_log_do_action;
4889 t16 = le16_to_cpu(lrh->target_attr);
4890 t64 = le64_to_cpu(lrh->target_vcn);
4891 dp = find_dp(dptbl, t16, t64);
4894 goto read_next_log_do_action;
4896 if (rec_lsn < le64_to_cpu(dp->oldest_lsn))
4897 goto read_next_log_do_action;
4899 t16 = le16_to_cpu(lrh->target_attr);
4900 if (t16 >= bytes_per_rt(oatbl)) {
4905 oe = Add2Ptr(oatbl, t16);
4907 if (oe->next != RESTART_ENTRY_ALLOCATED_LE) {
4920 vcn = le64_to_cpu(lrh->target_vcn);
4922 if (!run_lookup_entry(oa->run1, vcn, &lcn, NULL, NULL) ||
4923 lcn == SPARSE_LCN) {
4924 goto read_next_log_do_action;
4927 /* Point to the Redo data and get its length. */
4928 data = Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
4929 dlen = le16_to_cpu(lrh->redo_len);
4931 /* Shorten length by any Lcns which were deleted. */
4934 for (i = le16_to_cpu(lrh->lcns_follow); i; i--) {
4938 voff = le16_to_cpu(lrh->record_off) +
4939 le16_to_cpu(lrh->attr_off);
4940 voff += le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
4942 /* If the Vcn question is allocated, we can just get out. */
4943 j = le64_to_cpu(lrh->target_vcn) - le64_to_cpu(dp->vcn);
4944 if (dp->page_lcns[j + i - 1])
4951 * Calculate the allocated space left relative to the
4952 * log record Vcn, after removing this unallocated Vcn.
4954 alen = (i - 1) << sbi->cluster_bits;
4957 * If the update described this log record goes beyond
4958 * the allocated space, then we will have to reduce the length.
4962 else if (voff + dlen > alen)
4967 * If the resulting dlen from above is now zero,
4968 * we can skip this log record.
4970 if (!dlen && saved_len)
4971 goto read_next_log_do_action;
4973 t16 = le16_to_cpu(lrh->redo_op);
4974 if (can_skip_action(t16))
4975 goto read_next_log_do_action;
4977 /* Apply the Redo operation a common routine. */
4978 err = do_action(log, oe, lrh, t16, data, dlen, rec_len, &rec_lsn);
4982 /* Keep reading and looping back until end of file. */
4983 read_next_log_do_action:
4984 err = read_next_log_rec(log, lcb, &rec_lsn);
4985 if (!err && rec_lsn)
4986 goto do_action_next;
4992 /* Scan Transaction Table. */
4994 transaction_table_next:
4995 tr = enum_rstbl(trtbl, tr);
4997 goto undo_action_done;
4999 if (TransactionActive != tr->transact_state || !tr->undo_next_lsn) {
5000 free_rsttbl_idx(trtbl, PtrOffset(trtbl, tr));
5001 goto transaction_table_next;
5004 log->transaction_id = PtrOffset(trtbl, tr);
5005 undo_next_lsn = le64_to_cpu(tr->undo_next_lsn);
5008 * We only have to do anything if the transaction has
5009 * something its undo_next_lsn field.
5014 /* Read the first record to be undone by this transaction. */
5015 err = read_log_rec_lcb(log, undo_next_lsn, lcb_ctx_undo_next, &lcb);
5020 * Now loop to read all of our log records forwards,
5021 * until we hit the end of the file, cleaning up at the end.
5027 transact_id = le32_to_cpu(frh->transact_id);
5028 rec_len = le32_to_cpu(frh->client_data_len);
5030 if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
5035 if (lrh->undo_op == cpu_to_le16(Noop))
5036 goto read_next_log_undo_action;
5038 oe = Add2Ptr(oatbl, le16_to_cpu(lrh->target_attr));
5041 t16 = le16_to_cpu(lrh->lcns_follow);
5043 goto add_allocated_vcns;
5045 is_mapped = run_lookup_entry(oa->run1, le64_to_cpu(lrh->target_vcn),
5049 * If the mapping isn't already the table or the mapping
5050 * corresponds to a hole the mapping, we need to make sure
5051 * there is no partial page already memory.
5053 if (is_mapped && lcn != SPARSE_LCN && clen >= t16)
5054 goto add_allocated_vcns;
5056 vcn = le64_to_cpu(lrh->target_vcn);
5057 vcn &= ~(u64)(log->clst_per_page - 1);
5060 for (i = 0, vcn = le64_to_cpu(lrh->target_vcn),
5061 size = (vcn + 1) << sbi->cluster_bits;
5062 i < t16; i++, vcn += 1, size += sbi->cluster_size) {
5064 if (!attr->non_res) {
5065 if (size > le32_to_cpu(attr->res.data_size))
5066 attr->res.data_size = cpu_to_le32(size);
5068 if (size > le64_to_cpu(attr->nres.data_size))
5069 attr->nres.valid_size = attr->nres.data_size =
5070 attr->nres.alloc_size =
5075 t16 = le16_to_cpu(lrh->undo_op);
5076 if (can_skip_action(t16))
5077 goto read_next_log_undo_action;
5079 /* Point to the Redo data and get its length. */
5080 data = Add2Ptr(lrh, le16_to_cpu(lrh->undo_off));
5081 dlen = le16_to_cpu(lrh->undo_len);
5083 /* It is time to apply the undo action. */
5084 err = do_action(log, oe, lrh, t16, data, dlen, rec_len, NULL);
5086 read_next_log_undo_action:
5088 * Keep reading and looping back until we have read the
5089 * last record for this transaction.
5091 err = read_next_log_rec(log, lcb, &rec_lsn);
5096 goto undo_action_next;
5102 free_rsttbl_idx(trtbl, log->transaction_id);
5104 log->transaction_id = 0;
5106 goto transaction_table_next;
5110 ntfs_update_mftmirr(sbi, 0);
5112 sbi->flags &= ~NTFS_FLAGS_NEED_REPLAY;
5120 rh = kzalloc(log->page_size, GFP_NOFS);
5126 rh->rhdr.sign = NTFS_RSTR_SIGNATURE;
5127 rh->rhdr.fix_off = cpu_to_le16(offsetof(struct RESTART_HDR, fixups));
5128 t16 = (log->page_size >> SECTOR_SHIFT) + 1;
5129 rh->rhdr.fix_num = cpu_to_le16(t16);
5130 rh->sys_page_size = cpu_to_le32(log->page_size);
5131 rh->page_size = cpu_to_le32(log->page_size);
5133 t16 = ALIGN(offsetof(struct RESTART_HDR, fixups) + sizeof(short) * t16,
5135 rh->ra_off = cpu_to_le16(t16);
5136 rh->minor_ver = cpu_to_le16(1); // 0x1A:
5137 rh->major_ver = cpu_to_le16(1); // 0x1C:
5139 ra2 = Add2Ptr(rh, t16);
5140 memcpy(ra2, ra, sizeof(struct RESTART_AREA));
5142 ra2->client_idx[0] = 0;
5143 ra2->client_idx[1] = LFS_NO_CLIENT_LE;
5144 ra2->flags = cpu_to_le16(2);
5146 le32_add_cpu(&ra2->open_log_count, 1);
5148 ntfs_fix_pre_write(&rh->rhdr, log->page_size);
5150 err = ntfs_sb_write_run(sbi, &ni->file.run, 0, rh, log->page_size, 0);
5152 err = ntfs_sb_write_run(sbi, &log->ni->file.run, log->page_size,
5153 rh, log->page_size, 0);
5165 * Scan the Open Attribute Table to close all of
5166 * the open attributes.
5169 while ((oe = enum_rstbl(oatbl, oe))) {
5170 rno = ino_get(&oe->ref);
5172 if (oe->is_attr_name == 1) {
5178 if (oe->is_attr_name)
5185 run_close(&oa->run0);
5188 iput(&oa->ni->vfs_inode);
5196 kfree(rst_info.r_page);
5199 kfree(log->one_page_buf);
5202 sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
5206 else if (log->set_dirty)
5207 ntfs_set_state(sbi, NTFS_DIRTY_ERROR);