fs/ntfs3/fslog.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  *
   4  * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
   5  *
   6  */
   7
   8 #include <linux/blkdev.h>
   9 #include <linux/fs.h>
  10 #include <linux/random.h>
  11 #include <linux/slab.h>
  12
  13 #include "debug.h"
  14 #include "ntfs.h"
  15 #include "ntfs_fs.h"
  16
  17 /*
  18  * LOG FILE structs
  19  */
  20
  21 // clang-format off
  22
  23 #define MaxLogFileSize     0x100000000ull
  24 #define DefaultLogPageSize 4096
  25 #define MinLogRecordPages  0x30
  26
  27 struct RESTART_HDR {
  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:
  34         __le16 fixups[];
  35 };
  36
  37 #define LFS_NO_CLIENT 0xffff
  38 #define LFS_NO_CLIENT_LE cpu_to_le16(0xffff)
  39
  40 struct CLIENT_REC {
  41         __le64 oldest_lsn;
  42         __le64 restart_lsn; // 0x08:
  43         __le16 prev_client; // 0x10:
  44         __le16 next_client; // 0x12:
  45         __le16 seq_num;     // 0x14:
  46         u8 align[6];        // 0x16:
  47         __le32 name_bytes;  // 0x1C: In bytes.
  48         __le16 name[32];    // 0x20: Name of client.
  49 };
  50
  51 static_assert(sizeof(struct CLIENT_REC) == 0x60);
  52
  53 /* Two copies of these will exist at the beginning of the log file */
  54 struct RESTART_AREA {
  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:
  69 };
  70
  71 struct LOG_REC_HDR {
  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:
  86 };
  87
  88 static_assert(sizeof(struct LOG_REC_HDR) == 0x20);
  89
  90 #define RESTART_ENTRY_ALLOCATED    0xFFFFFFFF
  91 #define RESTART_ENTRY_ALLOCATED_LE cpu_to_le32(0xFFFFFFFF)
  92
  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:
 101
 102 };
 103
 104 static_assert(sizeof(struct RESTART_TABLE) == 0x18);
 105
 106 struct ATTR_NAME_ENTRY {
 107         __le16 off; // Offset in the Open attribute Table.
 108         __le16 name_bytes;
 109         __le16 name[];
 110 };
 111
 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'
 119         u8 res;
 120         struct MFT_REF ref;     // 0x10: File Reference of file containing attribute
 121         __le64 open_record_lsn; // 0x18:
 122         void *ptr;              // 0x20:
 123 };
 124
 125 /* 32 bit version of 'struct OPEN_ATTR_ENRTY' */
 126 struct OPEN_ATTR_ENRTY_32 {
 127         __le32 next;            // 0x00: RESTART_ENTRY_ALLOCATED if allocated
 128         __le32 ptr;             // 0x04:
 129         struct MFT_REF ref;     // 0x08:
 130         __le64 open_record_lsn; // 0x10:
 131         u8 is_dirty_pages;      // 0x18:
 132         u8 is_attr_name;        // 0x19:
 133         u8 res1[2];
 134         enum ATTR_TYPE type;    // 0x1C:
 135         u8 name_len;            // 0x20: In wchar
 136         u8 res2[3];
 137         __le32 AttributeName;   // 0x24:
 138         __le32 bytes_per_index; // 0x28:
 139 };
 140
 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);
 144
 145 /*
 146  * One entry exists in the Dirty Pages Table for each page which is dirty at
 147  * the time the Restart Area is written.
 148  */
 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:
 157 };
 158
 159 static_assert(sizeof(struct DIR_PAGE_ENTRY) == 0x20);
 160
 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:
 174 };
 175
 176 static_assert(offsetof(struct DIR_PAGE_ENTRY_32, vcn_low) == 0x14);
 177 static_assert(sizeof(struct DIR_PAGE_ENTRY_32) == 0x2c);
 178
 179 enum transact_state {
 180         TransactionUninitialized = 0,
 181         TransactionActive,
 182         TransactionPrepared,
 183         TransactionCommitted
 184 };
 185
 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
 195 };
 196
 197 static_assert(sizeof(struct TRANSACTION_ENTRY) == 0x28);
 198
 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
 211 };
 212
 213 static_assert(sizeof(struct NTFS_RESTART) == 0x40);
 214
 215 struct NEW_ATTRIBUTE_SIZES {
 216         __le64 alloc_size;
 217         __le64 valid_size;
 218         __le64 data_size;
 219         __le64 total_size;
 220 };
 221
 222 struct BITMAP_RANGE {
 223         __le32 bitmap_off;
 224         __le32 bits;
 225 };
 226
 227 struct LCN_RANGE {
 228         __le64 lcn;
 229         __le64 len;
 230 };
 231
 232 /* The following type defines the different log record types. */
 233 #define LfsClientRecord  cpu_to_le32(1)
 234 #define LfsClientRestart cpu_to_le32(2)
 235
 236 /* This is used to uniquely identify a client for a particular log file. */
 237 struct CLIENT_ID {
 238         __le16 seq_num;
 239         __le16 client_idx;
 240 };
 241
 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:
 253 };
 254
 255 #define LOG_RECORD_MULTI_PAGE cpu_to_le16(1)
 256
 257 static_assert(sizeof(struct LFS_RECORD_HDR) == 0x30);
 258
 259 struct LFS_RECORD {
 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,
 263 };
 264
 265 static_assert(sizeof(struct LFS_RECORD) == 0x10);
 266
 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
 275 };
 276
 277 // clang-format on
 278
 279 // Page contains the end of a log record.
 280 #define LOG_PAGE_LOG_RECORD_END cpu_to_le32(0x00000001)
 281
 282 static inline bool is_log_record_end(const struct RECORD_PAGE_HDR *hdr)
 283 {
 284         return hdr->rflags & LOG_PAGE_LOG_RECORD_END;
 285 }
 286
 287 static_assert(offsetof(struct RECORD_PAGE_HDR, file_off) == 0x3c);
 288
 289 /*
 290  * END of NTFS LOG structures
 291  */
 292
 293 /* Define some tuning parameters to keep the restart tables a reasonable size. */
 294 #define INITIAL_NUMBER_TRANSACTIONS 5
 295
 296 enum NTFS_LOG_OPERATION {
 297
 298         Noop = 0x00,
 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,
 321         HotFix = 0x17,
 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,
 333
 334         UpdateRelativeDataInIndex =
 335                 0x23, // NtOfsRestartUpdateRelativeDataInIndex
 336         UpdateRelativeDataInIndex2 = 0x24,
 337         ZeroEndOfFileRecord = 0x25,
 338 };
 339
 340 /*
 341  * Array for log records which require a target attribute.
 342  * A true indicates that the corresponding restart operation
 343  * requires a target attribute.
 344  */
 345 static const u8 AttributeRequired[] = {
 346         0xFC, 0xFB, 0xFF, 0x10, 0x06,
 347 };
 348
 349 static inline bool is_target_required(u16 op)
 350 {
 351         bool ret = op <= UpdateRecordDataAllocation &&
 352                    (AttributeRequired[op >> 3] >> (op & 7) & 1);
 353         return ret;
 354 }
 355
 356 static inline bool can_skip_action(enum NTFS_LOG_OPERATION op)
 357 {
 358         switch (op) {
 359         case Noop:
 360         case DeleteDirtyClusters:
 361         case HotFix:
 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:
 372                 return true;
 373         default:
 374                 return false;
 375         }
 376 }
 377
 378 enum { lcb_ctx_undo_next, lcb_ctx_prev, lcb_ctx_next };
 379
 380 /* Bytes per restart table. */
 381 static inline u32 bytes_per_rt(const struct RESTART_TABLE *rt)
 382 {
 383         return le16_to_cpu(rt->used) * le16_to_cpu(rt->size) +
 384                sizeof(struct RESTART_TABLE);
 385 }
 386
 387 /* Log record length. */
 388 static inline u32 lrh_length(const struct LOG_REC_HDR *lr)
 389 {
 390         u16 t16 = le16_to_cpu(lr->lcns_follow);
 391
 392         return struct_size(lr, page_lcns, max_t(u16, 1, t16));
 393 }
 394
 395 struct lcb {
 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'.
 401 };
 402
 403 static void lcb_put(struct lcb *lcb)
 404 {
 405         if (lcb->alloc)
 406                 kfree(lcb->log_rec);
 407         kfree(lcb->lrh);
 408         kfree(lcb);
 409 }
 410
 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)
 414 {
 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);
 418
 419                 /* Ignore this block if it's oldest lsn is 0. */
 420                 if (lsn && lsn < *oldest_lsn)
 421                         *oldest_lsn = lsn;
 422
 423                 next_client = cr->next_client;
 424         }
 425 }
 426
 427 static inline bool is_rst_page_hdr_valid(u32 file_off,
 428                                          const struct RESTART_HDR *rhdr)
 429 {
 430         u32 sys_page = le32_to_cpu(rhdr->sys_page_size);
 431         u32 page_size = le32_to_cpu(rhdr->page_size);
 432         u32 end_usa;
 433         u16 ro;
 434
 435         if (sys_page < SECTOR_SIZE || page_size < SECTOR_SIZE ||
 436             sys_page & (sys_page - 1) || page_size & (page_size - 1)) {
 437                 return false;
 438         }
 439
 440         /* Check that if the file offset isn't 0, it is the system page size. */
 441         if (file_off && file_off != sys_page)
 442                 return false;
 443
 444         /* Check support version 1.1+. */
 445         if (le16_to_cpu(rhdr->major_ver) <= 1 && !rhdr->minor_ver)
 446                 return false;
 447
 448         if (le16_to_cpu(rhdr->major_ver) > 2)
 449                 return false;
 450
 451         ro = le16_to_cpu(rhdr->ra_off);
 452         if (!IS_ALIGNED(ro, 8) || ro > sys_page)
 453                 return false;
 454
 455         end_usa = ((sys_page >> SECTOR_SHIFT) + 1) * sizeof(short);
 456         end_usa += le16_to_cpu(rhdr->rhdr.fix_off);
 457
 458         if (ro < end_usa)
 459                 return false;
 460
 461         return true;
 462 }
 463
 464 static inline bool is_rst_area_valid(const struct RESTART_HDR *rhdr)
 465 {
 466         const struct RESTART_AREA *ra;
 467         u16 cl, fl, ul;
 468         u32 off, l_size, file_dat_bits, file_size_round;
 469         u16 ro = le16_to_cpu(rhdr->ra_off);
 470         u32 sys_page = le32_to_cpu(rhdr->sys_page_size);
 471
 472         if (ro + offsetof(struct RESTART_AREA, l_size) >
 473             SECTOR_SIZE - sizeof(short))
 474                 return false;
 475
 476         ra = Add2Ptr(rhdr, ro);
 477         cl = le16_to_cpu(ra->log_clients);
 478
 479         if (cl > 1)
 480                 return false;
 481
 482         off = le16_to_cpu(ra->client_off);
 483
 484         if (!IS_ALIGNED(off, 8) || ro + off > SECTOR_SIZE - sizeof(short))
 485                 return false;
 486
 487         off += cl * sizeof(struct CLIENT_REC);
 488
 489         if (off > sys_page)
 490                 return false;
 491
 492         /*
 493          * Check the restart length field and whether the entire
 494          * restart area is contained that length.
 495          */
 496         if (le16_to_cpu(rhdr->ra_off) + le16_to_cpu(ra->ra_len) > sys_page ||
 497             off > le16_to_cpu(ra->ra_len)) {
 498                 return false;
 499         }
 500
 501         /*
 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.
 504          */
 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))
 509                 return false;
 510
 511         /* Make sure the sequence number bits match the log file size. */
 512         l_size = le64_to_cpu(ra->l_size);
 513
 514         file_dat_bits = sizeof(u64) * 8 - le32_to_cpu(ra->seq_num_bits);
 515         file_size_round = 1u << (file_dat_bits + 3);
 516         if (file_size_round != l_size &&
 517             (file_size_round < l_size || (file_size_round / 2) > l_size)) {
 518                 return false;
 519         }
 520
 521         /* The log page data offset and record header length must be quad-aligned. */
 522         if (!IS_ALIGNED(le16_to_cpu(ra->data_off), 8) ||
 523             !IS_ALIGNED(le16_to_cpu(ra->rec_hdr_len), 8))
 524                 return false;
 525
 526         return true;
 527 }
 528
 529 static inline bool is_client_area_valid(const struct RESTART_HDR *rhdr,
 530                                         bool usa_error)
 531 {
 532         u16 ro = le16_to_cpu(rhdr->ra_off);
 533         const struct RESTART_AREA *ra = Add2Ptr(rhdr, ro);
 534         u16 ra_len = le16_to_cpu(ra->ra_len);
 535         const struct CLIENT_REC *ca;
 536         u32 i;
 537
 538         if (usa_error && ra_len + ro > SECTOR_SIZE - sizeof(short))
 539                 return false;
 540
 541         /* Find the start of the client array. */
 542         ca = Add2Ptr(ra, le16_to_cpu(ra->client_off));
 543
 544         /*
 545          * Start with the free list.
 546          * Check that all the clients are valid and that there isn't a cycle.
 547          * Do the in-use list on the second pass.
 548          */
 549         for (i = 0; i < 2; i++) {
 550                 u16 client_idx = le16_to_cpu(ra->client_idx[i]);
 551                 bool first_client = true;
 552                 u16 clients = le16_to_cpu(ra->log_clients);
 553
 554                 while (client_idx != LFS_NO_CLIENT) {
 555                         const struct CLIENT_REC *cr;
 556
 557                         if (!clients ||
 558                             client_idx >= le16_to_cpu(ra->log_clients))
 559                                 return false;
 560
 561                         clients -= 1;
 562                         cr = ca + client_idx;
 563
 564                         client_idx = le16_to_cpu(cr->next_client);
 565
 566                         if (first_client) {
 567                                 first_client = false;
 568                                 if (cr->prev_client != LFS_NO_CLIENT_LE)
 569                                         return false;
 570                         }
 571                 }
 572         }
 573
 574         return true;
 575 }
 576
 577 /*
 578  * remove_client
 579  *
 580  * Remove a client record from a client record list an restart area.
 581  */
 582 static inline void remove_client(struct CLIENT_REC *ca,
 583                                  const struct CLIENT_REC *cr, __le16 *head)
 584 {
 585         if (cr->prev_client == LFS_NO_CLIENT_LE)
 586                 *head = cr->next_client;
 587         else
 588                 ca[le16_to_cpu(cr->prev_client)].next_client = cr->next_client;
 589
 590         if (cr->next_client != LFS_NO_CLIENT_LE)
 591                 ca[le16_to_cpu(cr->next_client)].prev_client = cr->prev_client;
 592 }
 593
 594 /*
 595  * add_client - Add a client record to the start of a list.
 596  */
 597 static inline void add_client(struct CLIENT_REC *ca, u16 index, __le16 *head)
 598 {
 599         struct CLIENT_REC *cr = ca + index;
 600
 601         cr->prev_client = LFS_NO_CLIENT_LE;
 602         cr->next_client = *head;
 603
 604         if (*head != LFS_NO_CLIENT_LE)
 605                 ca[le16_to_cpu(*head)].prev_client = cpu_to_le16(index);
 606
 607         *head = cpu_to_le16(index);
 608 }
 609
 610 static inline void *enum_rstbl(struct RESTART_TABLE *t, void *c)
 611 {
 612         __le32 *e;
 613         u32 bprt;
 614         u16 rsize = t ? le16_to_cpu(t->size) : 0;
 615
 616         if (!c) {
 617                 if (!t || !t->total)
 618                         return NULL;
 619                 e = Add2Ptr(t, sizeof(struct RESTART_TABLE));
 620         } else {
 621                 e = Add2Ptr(c, rsize);
 622         }
 623
 624         /* Loop until we hit the first one allocated, or the end of the list. */
 625         for (bprt = bytes_per_rt(t); PtrOffset(t, e) < bprt;
 626              e = Add2Ptr(e, rsize)) {
 627                 if (*e == RESTART_ENTRY_ALLOCATED_LE)
 628                         return e;
 629         }
 630         return NULL;
 631 }
 632
 633 /*
 634  * find_dp - Search for a @vcn in Dirty Page Table.
 635  */
 636 static inline struct DIR_PAGE_ENTRY *find_dp(struct RESTART_TABLE *dptbl,
 637                                              u32 target_attr, u64 vcn)
 638 {
 639         __le32 ta = cpu_to_le32(target_attr);
 640         struct DIR_PAGE_ENTRY *dp = NULL;
 641
 642         while ((dp = enum_rstbl(dptbl, dp))) {
 643                 u64 dp_vcn = le64_to_cpu(dp->vcn);
 644
 645                 if (dp->target_attr == ta && vcn >= dp_vcn &&
 646                     vcn < dp_vcn + le32_to_cpu(dp->lcns_follow)) {
 647                         return dp;
 648                 }
 649         }
 650         return NULL;
 651 }
 652
 653 static inline u32 norm_file_page(u32 page_size, u32 *l_size, bool use_default)
 654 {
 655         if (use_default)
 656                 page_size = DefaultLogPageSize;
 657
 658         /* Round the file size down to a system page boundary. */
 659         *l_size &= ~(page_size - 1);
 660
 661         /* File should contain at least 2 restart pages and MinLogRecordPages pages. */
 662         if (*l_size < (MinLogRecordPages + 2) * page_size)
 663                 return 0;
 664
 665         return page_size;
 666 }
 667
 668 static bool check_log_rec(const struct LOG_REC_HDR *lr, u32 bytes, u32 tr,
 669                           u32 bytes_per_attr_entry)
 670 {
 671         u16 t16;
 672
 673         if (bytes < sizeof(struct LOG_REC_HDR))
 674                 return false;
 675         if (!tr)
 676                 return false;
 677
 678         if ((tr - sizeof(struct RESTART_TABLE)) %
 679             sizeof(struct TRANSACTION_ENTRY))
 680                 return false;
 681
 682         if (le16_to_cpu(lr->redo_off) & 7)
 683                 return false;
 684
 685         if (le16_to_cpu(lr->undo_off) & 7)
 686                 return false;
 687
 688         if (lr->target_attr)
 689                 goto check_lcns;
 690
 691         if (is_target_required(le16_to_cpu(lr->redo_op)))
 692                 return false;
 693
 694         if (is_target_required(le16_to_cpu(lr->undo_op)))
 695                 return false;
 696
 697 check_lcns:
 698         if (!lr->lcns_follow)
 699                 goto check_length;
 700
 701         t16 = le16_to_cpu(lr->target_attr);
 702         if ((t16 - sizeof(struct RESTART_TABLE)) % bytes_per_attr_entry)
 703                 return false;
 704
 705 check_length:
 706         if (bytes < lrh_length(lr))
 707                 return false;
 708
 709         return true;
 710 }
 711
 712 static bool check_rstbl(const struct RESTART_TABLE *rt, size_t bytes)
 713 {
 714         u32 ts;
 715         u32 i, off;
 716         u16 rsize = le16_to_cpu(rt->size);
 717         u16 ne = le16_to_cpu(rt->used);
 718         u32 ff = le32_to_cpu(rt->first_free);
 719         u32 lf = le32_to_cpu(rt->last_free);
 720
 721         ts = rsize * ne + sizeof(struct RESTART_TABLE);
 722
 723         if (!rsize || rsize > bytes ||
 724             rsize + sizeof(struct RESTART_TABLE) > bytes || bytes < ts ||
 725             le16_to_cpu(rt->total) > ne || ff > ts || lf > ts ||
 726             (ff && ff < sizeof(struct RESTART_TABLE)) ||
 727             (lf && lf < sizeof(struct RESTART_TABLE))) {
 728                 return false;
 729         }
 730
 731         /*
 732          * Verify each entry is either allocated or points
 733          * to a valid offset the table.
 734          */
 735         for (i = 0; i < ne; i++) {
 736                 off = le32_to_cpu(*(__le32 *)Add2Ptr(
 737                         rt, i * rsize + sizeof(struct RESTART_TABLE)));
 738
 739                 if (off != RESTART_ENTRY_ALLOCATED && off &&
 740                     (off < sizeof(struct RESTART_TABLE) ||
 741                      ((off - sizeof(struct RESTART_TABLE)) % rsize))) {
 742                         return false;
 743                 }
 744         }
 745
 746         /*
 747          * Walk through the list headed by the first entry to make
 748          * sure none of the entries are currently being used.
 749          */
 750         for (off = ff; off;) {
 751                 if (off == RESTART_ENTRY_ALLOCATED)
 752                         return false;
 753
 754                 off = le32_to_cpu(*(__le32 *)Add2Ptr(rt, off));
 755         }
 756
 757         return true;
 758 }
 759
 760 /*
 761  * free_rsttbl_idx - Free a previously allocated index a Restart Table.
 762  */
 763 static inline void free_rsttbl_idx(struct RESTART_TABLE *rt, u32 off)
 764 {
 765         __le32 *e;
 766         u32 lf = le32_to_cpu(rt->last_free);
 767         __le32 off_le = cpu_to_le32(off);
 768
 769         e = Add2Ptr(rt, off);
 770
 771         if (off < le32_to_cpu(rt->free_goal)) {
 772                 *e = rt->first_free;
 773                 rt->first_free = off_le;
 774                 if (!lf)
 775                         rt->last_free = off_le;
 776         } else {
 777                 if (lf)
 778                         *(__le32 *)Add2Ptr(rt, lf) = off_le;
 779                 else
 780                         rt->first_free = off_le;
 781
 782                 rt->last_free = off_le;
 783                 *e = 0;
 784         }
 785
 786         le16_sub_cpu(&rt->total, 1);
 787 }
 788
 789 static inline struct RESTART_TABLE *init_rsttbl(u16 esize, u16 used)
 790 {
 791         __le32 *e, *last_free;
 792         u32 off;
 793         u32 bytes = esize * used + sizeof(struct RESTART_TABLE);
 794         u32 lf = sizeof(struct RESTART_TABLE) + (used - 1) * esize;
 795         struct RESTART_TABLE *t = kzalloc(bytes, GFP_NOFS);
 796
 797         if (!t)
 798                 return NULL;
 799
 800         t->size = cpu_to_le16(esize);
 801         t->used = cpu_to_le16(used);
 802         t->free_goal = cpu_to_le32(~0u);
 803         t->first_free = cpu_to_le32(sizeof(struct RESTART_TABLE));
 804         t->last_free = cpu_to_le32(lf);
 805
 806         e = (__le32 *)(t + 1);
 807         last_free = Add2Ptr(t, lf);
 808
 809         for (off = sizeof(struct RESTART_TABLE) + esize; e < last_free;
 810              e = Add2Ptr(e, esize), off += esize) {
 811                 *e = cpu_to_le32(off);
 812         }
 813         return t;
 814 }
 815
 816 static inline struct RESTART_TABLE *extend_rsttbl(struct RESTART_TABLE *tbl,
 817                                                   u32 add, u32 free_goal)
 818 {
 819         u16 esize = le16_to_cpu(tbl->size);
 820         __le32 osize = cpu_to_le32(bytes_per_rt(tbl));
 821         u32 used = le16_to_cpu(tbl->used);
 822         struct RESTART_TABLE *rt;
 823
 824         rt = init_rsttbl(esize, used + add);
 825         if (!rt)
 826                 return NULL;
 827
 828         memcpy(rt + 1, tbl + 1, esize * used);
 829
 830         rt->free_goal = free_goal == ~0u ?
 831                                 cpu_to_le32(~0u) :
 832                                 cpu_to_le32(sizeof(struct RESTART_TABLE) +
 833                                             free_goal * esize);
 834
 835         if (tbl->first_free) {
 836                 rt->first_free = tbl->first_free;
 837                 *(__le32 *)Add2Ptr(rt, le32_to_cpu(tbl->last_free)) = osize;
 838         } else {
 839                 rt->first_free = osize;
 840         }
 841
 842         rt->total = tbl->total;
 843
 844         kfree(tbl);
 845         return rt;
 846 }
 847
 848 /*
 849  * alloc_rsttbl_idx
 850  *
 851  * Allocate an index from within a previously initialized Restart Table.
 852  */
 853 static inline void *alloc_rsttbl_idx(struct RESTART_TABLE **tbl)
 854 {
 855         u32 off;
 856         __le32 *e;
 857         struct RESTART_TABLE *t = *tbl;
 858
 859         if (!t->first_free) {
 860                 *tbl = t = extend_rsttbl(t, 16, ~0u);
 861                 if (!t)
 862                         return NULL;
 863         }
 864
 865         off = le32_to_cpu(t->first_free);
 866
 867         /* Dequeue this entry and zero it. */
 868         e = Add2Ptr(t, off);
 869
 870         t->first_free = *e;
 871
 872         memset(e, 0, le16_to_cpu(t->size));
 873
 874         *e = RESTART_ENTRY_ALLOCATED_LE;
 875
 876         /* If list is going empty, then we fix the last_free as well. */
 877         if (!t->first_free)
 878                 t->last_free = 0;
 879
 880         le16_add_cpu(&t->total, 1);
 881
 882         return Add2Ptr(t, off);
 883 }
 884
 885 /*
 886  * alloc_rsttbl_from_idx
 887  *
 888  * Allocate a specific index from within a previously initialized Restart Table.
 889  */
 890 static inline void *alloc_rsttbl_from_idx(struct RESTART_TABLE **tbl, u32 vbo)
 891 {
 892         u32 off;
 893         __le32 *e;
 894         struct RESTART_TABLE *rt = *tbl;
 895         u32 bytes = bytes_per_rt(rt);
 896         u16 esize = le16_to_cpu(rt->size);
 897
 898         /* If the entry is not the table, we will have to extend the table. */
 899         if (vbo >= bytes) {
 900                 /*
 901                  * Extend the size by computing the number of entries between
 902                  * the existing size and the desired index and adding 1 to that.
 903                  */
 904                 u32 bytes2idx = vbo - bytes;
 905
 906                 /*
 907                  * There should always be an integral number of entries
 908                  * being added. Now extend the table.
 909                  */
 910                 *tbl = rt = extend_rsttbl(rt, bytes2idx / esize + 1, bytes);
 911                 if (!rt)
 912                         return NULL;
 913         }
 914
 915         /* See if the entry is already allocated, and just return if it is. */
 916         e = Add2Ptr(rt, vbo);
 917
 918         if (*e == RESTART_ENTRY_ALLOCATED_LE)
 919                 return e;
 920
 921         /*
 922          * Walk through the table, looking for the entry we're
 923          * interested and the previous entry.
 924          */
 925         off = le32_to_cpu(rt->first_free);
 926         e = Add2Ptr(rt, off);
 927
 928         if (off == vbo) {
 929                 /* this is a match */
 930                 rt->first_free = *e;
 931                 goto skip_looking;
 932         }
 933
 934         /*
 935          * Need to walk through the list looking for the predecessor
 936          * of our entry.
 937          */
 938         for (;;) {
 939                 /* Remember the entry just found */
 940                 u32 last_off = off;
 941                 __le32 *last_e = e;
 942
 943                 /* Should never run of entries. */
 944
 945                 /* Lookup up the next entry the list. */
 946                 off = le32_to_cpu(*last_e);
 947                 e = Add2Ptr(rt, off);
 948
 949                 /* If this is our match we are done. */
 950                 if (off == vbo) {
 951                         *last_e = *e;
 952
 953                         /*
 954                          * If this was the last entry, we update that
 955                          * table as well.
 956                          */
 957                         if (le32_to_cpu(rt->last_free) == off)
 958                                 rt->last_free = cpu_to_le32(last_off);
 959                         break;
 960                 }
 961         }
 962
 963 skip_looking:
 964         /* If the list is now empty, we fix the last_free as well. */
 965         if (!rt->first_free)
 966                 rt->last_free = 0;
 967
 968         /* Zero this entry. */
 969         memset(e, 0, esize);
 970         *e = RESTART_ENTRY_ALLOCATED_LE;
 971
 972         le16_add_cpu(&rt->total, 1);
 973
 974         return e;
 975 }
 976
 977 #define RESTART_SINGLE_PAGE_IO cpu_to_le16(0x0001)
 978
 979 #define NTFSLOG_WRAPPED 0x00000001
 980 #define NTFSLOG_MULTIPLE_PAGE_IO 0x00000002
 981 #define NTFSLOG_NO_LAST_LSN 0x00000004
 982 #define NTFSLOG_REUSE_TAIL 0x00000010
 983 #define NTFSLOG_NO_OLDEST_LSN 0x00000020
 984
 985 /* Helper struct to work with NTFS $LogFile. */
 986 struct ntfs_log {
 987         struct ntfs_inode *ni;
 988
 989         u32 l_size;
 990         u32 sys_page_size;
 991         u32 sys_page_mask;
 992         u32 page_size;
 993         u32 page_mask; // page_size - 1
 994         u8 page_bits;
 995         struct RECORD_PAGE_HDR *one_page_buf;
 996
 997         struct RESTART_TABLE *open_attr_tbl;
 998         u32 transaction_id;
 999         u32 clst_per_page;
1000
1001         u32 first_page;
1002         u32 next_page;
1003         u32 ra_off;
1004         u32 data_off;
1005         u32 restart_size;
1006         u32 data_size;
1007         u16 record_header_len;
1008         u64 seq_num;
1009         u32 seq_num_bits;
1010         u32 file_data_bits;
1011         u32 seq_num_mask; /* (1 << file_data_bits) - 1 */
1012
1013         struct RESTART_AREA *ra; /* In-memory image of the next restart area. */
1014         u32 ra_size; /* The usable size of the restart area. */
1015
1016         /*
1017          * If true, then the in-memory restart area is to be written
1018          * to the first position on the disk.
1019          */
1020         bool init_ra;
1021         bool set_dirty; /* True if we need to set dirty flag. */
1022
1023         u64 oldest_lsn;
1024
1025         u32 oldest_lsn_off;
1026         u64 last_lsn;
1027
1028         u32 total_avail;
1029         u32 total_avail_pages;
1030         u32 total_undo_commit;
1031         u32 max_current_avail;
1032         u32 current_avail;
1033         u32 reserved;
1034
1035         short major_ver;
1036         short minor_ver;
1037
1038         u32 l_flags; /* See NTFSLOG_XXX */
1039         u32 current_openlog_count; /* On-disk value for open_log_count. */
1040
1041         struct CLIENT_ID client_id;
1042         u32 client_undo_commit;
1043 };
1044
1045 static inline u32 lsn_to_vbo(struct ntfs_log *log, const u64 lsn)
1046 {
1047         u32 vbo = (lsn << log->seq_num_bits) >> (log->seq_num_bits - 3);
1048
1049         return vbo;
1050 }
1051
1052 /* Compute the offset in the log file of the next log page. */
1053 static inline u32 next_page_off(struct ntfs_log *log, u32 off)
1054 {
1055         off = (off & ~log->sys_page_mask) + log->page_size;
1056         return off >= log->l_size ? log->first_page : off;
1057 }
1058
1059 static inline u32 lsn_to_page_off(struct ntfs_log *log, u64 lsn)
1060 {
1061         return (((u32)lsn) << 3) & log->page_mask;
1062 }
1063
1064 static inline u64 vbo_to_lsn(struct ntfs_log *log, u32 off, u64 Seq)
1065 {
1066         return (off >> 3) + (Seq << log->file_data_bits);
1067 }
1068
1069 static inline bool is_lsn_in_file(struct ntfs_log *log, u64 lsn)
1070 {
1071         return lsn >= log->oldest_lsn &&
1072                lsn <= le64_to_cpu(log->ra->current_lsn);
1073 }
1074
1075 static inline u32 hdr_file_off(struct ntfs_log *log,
1076                                struct RECORD_PAGE_HDR *hdr)
1077 {
1078         if (log->major_ver < 2)
1079                 return le64_to_cpu(hdr->rhdr.lsn);
1080
1081         return le32_to_cpu(hdr->file_off);
1082 }
1083
1084 static inline u64 base_lsn(struct ntfs_log *log,
1085                            const struct RECORD_PAGE_HDR *hdr, u64 lsn)
1086 {
1087         u64 h_lsn = le64_to_cpu(hdr->rhdr.lsn);
1088         u64 ret = (((h_lsn >> log->file_data_bits) +
1089                     (lsn < (lsn_to_vbo(log, h_lsn) & ~log->page_mask) ? 1 : 0))
1090                    << log->file_data_bits) +
1091                   ((((is_log_record_end(hdr) &&
1092                       h_lsn <= le64_to_cpu(hdr->record_hdr.last_end_lsn)) ?
1093                              le16_to_cpu(hdr->record_hdr.next_record_off) :
1094                              log->page_size) +
1095                     lsn) >>
1096                    3);
1097
1098         return ret;
1099 }
1100
1101 static inline bool verify_client_lsn(struct ntfs_log *log,
1102                                      const struct CLIENT_REC *client, u64 lsn)
1103 {
1104         return lsn >= le64_to_cpu(client->oldest_lsn) &&
1105                lsn <= le64_to_cpu(log->ra->current_lsn) && lsn;
1106 }
1107
1108 struct restart_info {
1109         u64 last_lsn;
1110         struct RESTART_HDR *r_page;
1111         u32 vbo;
1112         bool chkdsk_was_run;
1113         bool valid_page;
1114         bool initialized;
1115         bool restart;
1116 };
1117
1118 static int read_log_page(struct ntfs_log *log, u32 vbo,
1119                          struct RECORD_PAGE_HDR **buffer, bool *usa_error)
1120 {
1121         int err = 0;
1122         u32 page_idx = vbo >> log->page_bits;
1123         u32 page_off = vbo & log->page_mask;
1124         u32 bytes = log->page_size - page_off;
1125         void *to_free = NULL;
1126         u32 page_vbo = page_idx << log->page_bits;
1127         struct RECORD_PAGE_HDR *page_buf;
1128         struct ntfs_inode *ni = log->ni;
1129         bool bBAAD;
1130
1131         if (vbo >= log->l_size)
1132                 return -EINVAL;
1133
1134         if (!*buffer) {
1135                 to_free = kmalloc(log->page_size, GFP_NOFS);
1136                 if (!to_free)
1137                         return -ENOMEM;
1138                 *buffer = to_free;
1139         }
1140
1141         page_buf = page_off ? log->one_page_buf : *buffer;
1142
1143         err = ntfs_read_run_nb(ni->mi.sbi, &ni->file.run, page_vbo, page_buf,
1144                                log->page_size, NULL);
1145         if (err)
1146                 goto out;
1147
1148         if (page_buf->rhdr.sign != NTFS_FFFF_SIGNATURE)
1149                 ntfs_fix_post_read(&page_buf->rhdr, PAGE_SIZE, false);
1150
1151         if (page_buf != *buffer)
1152                 memcpy(*buffer, Add2Ptr(page_buf, page_off), bytes);
1153
1154         bBAAD = page_buf->rhdr.sign == NTFS_BAAD_SIGNATURE;
1155
1156         if (usa_error)
1157                 *usa_error = bBAAD;
1158         /* Check that the update sequence array for this page is valid */
1159         /* If we don't allow errors, raise an error status */
1160         else if (bBAAD)
1161                 err = -EINVAL;
1162
1163 out:
1164         if (err && to_free) {
1165                 kfree(to_free);
1166                 *buffer = NULL;
1167         }
1168
1169         return err;
1170 }
1171
1172 /*
1173  * log_read_rst
1174  *
1175  * It walks through 512 blocks of the file looking for a valid
1176  * restart page header. It will stop the first time we find a
1177  * valid page header.
1178  */
1179 static int log_read_rst(struct ntfs_log *log, u32 l_size, bool first,
1180                         struct restart_info *info)
1181 {
1182         u32 skip, vbo;
1183         struct RESTART_HDR *r_page = NULL;
1184
1185         /* Determine which restart area we are looking for. */
1186         if (first) {
1187                 vbo = 0;
1188                 skip = 512;
1189         } else {
1190                 vbo = 512;
1191                 skip = 0;
1192         }
1193
1194         /* Loop continuously until we succeed. */
1195         for (; vbo < l_size; vbo = 2 * vbo + skip, skip = 0) {
1196                 bool usa_error;
1197                 bool brst, bchk;
1198                 struct RESTART_AREA *ra;
1199
1200                 /* Read a page header at the current offset. */
1201                 if (read_log_page(log, vbo, (struct RECORD_PAGE_HDR **)&r_page,
1202                                   &usa_error)) {
1203                         /* Ignore any errors. */
1204                         continue;
1205                 }
1206
1207                 /* Exit if the signature is a log record page. */
1208                 if (r_page->rhdr.sign == NTFS_RCRD_SIGNATURE) {
1209                         info->initialized = true;
1210                         break;
1211                 }
1212
1213                 brst = r_page->rhdr.sign == NTFS_RSTR_SIGNATURE;
1214                 bchk = r_page->rhdr.sign == NTFS_CHKD_SIGNATURE;
1215
1216                 if (!bchk && !brst) {
1217                         if (r_page->rhdr.sign != NTFS_FFFF_SIGNATURE) {
1218                                 /*
1219                                  * Remember if the signature does not
1220                                  * indicate uninitialized file.
1221                                  */
1222                                 info->initialized = true;
1223                         }
1224                         continue;
1225                 }
1226
1227                 ra = NULL;
1228                 info->valid_page = false;
1229                 info->initialized = true;
1230                 info->vbo = vbo;
1231
1232                 /* Let's check the restart area if this is a valid page. */
1233                 if (!is_rst_page_hdr_valid(vbo, r_page))
1234                         goto check_result;
1235                 ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
1236
1237                 if (!is_rst_area_valid(r_page))
1238                         goto check_result;
1239
1240                 /*
1241                  * We have a valid restart page header and restart area.
1242                  * If chkdsk was run or we have no clients then we have
1243                  * no more checking to do.
1244                  */
1245                 if (bchk || ra->client_idx[1] == LFS_NO_CLIENT_LE) {
1246                         info->valid_page = true;
1247                         goto check_result;
1248                 }
1249
1250                 if (is_client_area_valid(r_page, usa_error)) {
1251                         info->valid_page = true;
1252                         ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
1253                 }
1254
1255 check_result:
1256                 /*
1257                  * If chkdsk was run then update the caller's
1258                  * values and return.
1259                  */
1260                 if (r_page->rhdr.sign == NTFS_CHKD_SIGNATURE) {
1261                         info->chkdsk_was_run = true;
1262                         info->last_lsn = le64_to_cpu(r_page->rhdr.lsn);
1263                         info->restart = true;
1264                         info->r_page = r_page;
1265                         return 0;
1266                 }
1267
1268                 /*
1269                  * If we have a valid page then copy the values
1270                  * we need from it.
1271                  */
1272                 if (info->valid_page) {
1273                         info->last_lsn = le64_to_cpu(ra->current_lsn);
1274                         info->restart = true;
1275                         info->r_page = r_page;
1276                         return 0;
1277                 }
1278         }
1279
1280         kfree(r_page);
1281
1282         return 0;
1283 }
1284
1285 /*
1286  * Ilog_init_pg_hdr - Init @log from restart page header.
1287  */
1288 static void log_init_pg_hdr(struct ntfs_log *log, u32 sys_page_size,
1289                             u32 page_size, u16 major_ver, u16 minor_ver)
1290 {
1291         log->sys_page_size = sys_page_size;
1292         log->sys_page_mask = sys_page_size - 1;
1293         log->page_size = page_size;
1294         log->page_mask = page_size - 1;
1295         log->page_bits = blksize_bits(page_size);
1296
1297         log->clst_per_page = log->page_size >> log->ni->mi.sbi->cluster_bits;
1298         if (!log->clst_per_page)
1299                 log->clst_per_page = 1;
1300
1301         log->first_page = major_ver >= 2 ?
1302                                   0x22 * page_size :
1303                                   ((sys_page_size << 1) + (page_size << 1));
1304         log->major_ver = major_ver;
1305         log->minor_ver = minor_ver;
1306 }
1307
1308 /*
1309  * log_create - Init @log in cases when we don't have a restart area to use.
1310  */
1311 static void log_create(struct ntfs_log *log, u32 l_size, const u64 last_lsn,
1312                        u32 open_log_count, bool wrapped, bool use_multi_page)
1313 {
1314         log->l_size = l_size;
1315         /* All file offsets must be quadword aligned. */
1316         log->file_data_bits = blksize_bits(l_size) - 3;
1317         log->seq_num_mask = (8 << log->file_data_bits) - 1;
1318         log->seq_num_bits = sizeof(u64) * 8 - log->file_data_bits;
1319         log->seq_num = (last_lsn >> log->file_data_bits) + 2;
1320         log->next_page = log->first_page;
1321         log->oldest_lsn = log->seq_num << log->file_data_bits;
1322         log->oldest_lsn_off = 0;
1323         log->last_lsn = log->oldest_lsn;
1324
1325         log->l_flags |= NTFSLOG_NO_LAST_LSN | NTFSLOG_NO_OLDEST_LSN;
1326
1327         /* Set the correct flags for the I/O and indicate if we have wrapped. */
1328         if (wrapped)
1329                 log->l_flags |= NTFSLOG_WRAPPED;
1330
1331         if (use_multi_page)
1332                 log->l_flags |= NTFSLOG_MULTIPLE_PAGE_IO;
1333
1334         /* Compute the log page values. */
1335         log->data_off = ALIGN(
1336                 offsetof(struct RECORD_PAGE_HDR, fixups) +
1337                         sizeof(short) * ((log->page_size >> SECTOR_SHIFT) + 1),
1338                 8);
1339         log->data_size = log->page_size - log->data_off;
1340         log->record_header_len = sizeof(struct LFS_RECORD_HDR);
1341
1342         /* Remember the different page sizes for reservation. */
1343         log->reserved = log->data_size - log->record_header_len;
1344
1345         /* Compute the restart page values. */
1346         log->ra_off = ALIGN(
1347                 offsetof(struct RESTART_HDR, fixups) +
1348                         sizeof(short) *
1349                                 ((log->sys_page_size >> SECTOR_SHIFT) + 1),
1350                 8);
1351         log->restart_size = log->sys_page_size - log->ra_off;
1352         log->ra_size = struct_size(log->ra, clients, 1);
1353         log->current_openlog_count = open_log_count;
1354
1355         /*
1356          * The total available log file space is the number of
1357          * log file pages times the space available on each page.
1358          */
1359         log->total_avail_pages = log->l_size - log->first_page;
1360         log->total_avail = log->total_avail_pages >> log->page_bits;
1361
1362         /*
1363          * We assume that we can't use the end of the page less than
1364          * the file record size.
1365          * Then we won't need to reserve more than the caller asks for.
1366          */
1367         log->max_current_avail = log->total_avail * log->reserved;
1368         log->total_avail = log->total_avail * log->data_size;
1369         log->current_avail = log->max_current_avail;
1370 }
1371
1372 /*
1373  * log_create_ra - Fill a restart area from the values stored in @log.
1374  */
1375 static struct RESTART_AREA *log_create_ra(struct ntfs_log *log)
1376 {
1377         struct CLIENT_REC *cr;
1378         struct RESTART_AREA *ra = kzalloc(log->restart_size, GFP_NOFS);
1379
1380         if (!ra)
1381                 return NULL;
1382
1383         ra->current_lsn = cpu_to_le64(log->last_lsn);
1384         ra->log_clients = cpu_to_le16(1);
1385         ra->client_idx[1] = LFS_NO_CLIENT_LE;
1386         if (log->l_flags & NTFSLOG_MULTIPLE_PAGE_IO)
1387                 ra->flags = RESTART_SINGLE_PAGE_IO;
1388         ra->seq_num_bits = cpu_to_le32(log->seq_num_bits);
1389         ra->ra_len = cpu_to_le16(log->ra_size);
1390         ra->client_off = cpu_to_le16(offsetof(struct RESTART_AREA, clients));
1391         ra->l_size = cpu_to_le64(log->l_size);
1392         ra->rec_hdr_len = cpu_to_le16(log->record_header_len);
1393         ra->data_off = cpu_to_le16(log->data_off);
1394         ra->open_log_count = cpu_to_le32(log->current_openlog_count + 1);
1395
1396         cr = ra->clients;
1397
1398         cr->prev_client = LFS_NO_CLIENT_LE;
1399         cr->next_client = LFS_NO_CLIENT_LE;
1400
1401         return ra;
1402 }
1403
1404 static u32 final_log_off(struct ntfs_log *log, u64 lsn, u32 data_len)
1405 {
1406         u32 base_vbo = lsn << 3;
1407         u32 final_log_off = (base_vbo & log->seq_num_mask) & ~log->page_mask;
1408         u32 page_off = base_vbo & log->page_mask;
1409         u32 tail = log->page_size - page_off;
1410
1411         page_off -= 1;
1412
1413         /* Add the length of the header. */
1414         data_len += log->record_header_len;
1415
1416         /*
1417          * If this lsn is contained this log page we are done.
1418          * Otherwise we need to walk through several log pages.
1419          */
1420         if (data_len > tail) {
1421                 data_len -= tail;
1422                 tail = log->data_size;
1423                 page_off = log->data_off - 1;
1424
1425                 for (;;) {
1426                         final_log_off = next_page_off(log, final_log_off);
1427
1428                         /*
1429                          * We are done if the remaining bytes
1430                          * fit on this page.
1431                          */
1432                         if (data_len <= tail)
1433                                 break;
1434                         data_len -= tail;
1435                 }
1436         }
1437
1438         /*
1439          * We add the remaining bytes to our starting position on this page
1440          * and then add that value to the file offset of this log page.
1441          */
1442         return final_log_off + data_len + page_off;
1443 }
1444
1445 static int next_log_lsn(struct ntfs_log *log, const struct LFS_RECORD_HDR *rh,
1446                         u64 *lsn)
1447 {
1448         int err;
1449         u64 this_lsn = le64_to_cpu(rh->this_lsn);
1450         u32 vbo = lsn_to_vbo(log, this_lsn);
1451         u32 end =
1452                 final_log_off(log, this_lsn, le32_to_cpu(rh->client_data_len));
1453         u32 hdr_off = end & ~log->sys_page_mask;
1454         u64 seq = this_lsn >> log->file_data_bits;
1455         struct RECORD_PAGE_HDR *page = NULL;
1456
1457         /* Remember if we wrapped. */
1458         if (end <= vbo)
1459                 seq += 1;
1460
1461         /* Log page header for this page. */
1462         err = read_log_page(log, hdr_off, &page, NULL);
1463         if (err)
1464                 return err;
1465
1466         /*
1467          * If the lsn we were given was not the last lsn on this page,
1468          * then the starting offset for the next lsn is on a quad word
1469          * boundary following the last file offset for the current lsn.
1470          * Otherwise the file offset is the start of the data on the next page.
1471          */
1472         if (this_lsn == le64_to_cpu(page->rhdr.lsn)) {
1473                 /* If we wrapped, we need to increment the sequence number. */
1474                 hdr_off = next_page_off(log, hdr_off);
1475                 if (hdr_off == log->first_page)
1476                         seq += 1;
1477
1478                 vbo = hdr_off + log->data_off;
1479         } else {
1480                 vbo = ALIGN(end, 8);
1481         }
1482
1483         /* Compute the lsn based on the file offset and the sequence count. */
1484         *lsn = vbo_to_lsn(log, vbo, seq);
1485
1486         /*
1487          * If this lsn is within the legal range for the file, we return true.
1488          * Otherwise false indicates that there are no more lsn's.
1489          */
1490         if (!is_lsn_in_file(log, *lsn))
1491                 *lsn = 0;
1492
1493         kfree(page);
1494
1495         return 0;
1496 }
1497
1498 /*
1499  * current_log_avail - Calculate the number of bytes available for log records.
1500  */
1501 static u32 current_log_avail(struct ntfs_log *log)
1502 {
1503         u32 oldest_off, next_free_off, free_bytes;
1504
1505         if (log->l_flags & NTFSLOG_NO_LAST_LSN) {
1506                 /* The entire file is available. */
1507                 return log->max_current_avail;
1508         }
1509
1510         /*
1511          * If there is a last lsn the restart area then we know that we will
1512          * have to compute the free range.
1513          * If there is no oldest lsn then start at the first page of the file.
1514          */
1515         oldest_off = (log->l_flags & NTFSLOG_NO_OLDEST_LSN) ?
1516                              log->first_page :
1517                              (log->oldest_lsn_off & ~log->sys_page_mask);
1518
1519         /*
1520          * We will use the next log page offset to compute the next free page.
1521          * If we are going to reuse this page go to the next page.
1522          * If we are at the first page then use the end of the file.
1523          */
1524         next_free_off = (log->l_flags & NTFSLOG_REUSE_TAIL) ?
1525                                 log->next_page + log->page_size :
1526                         log->next_page == log->first_page ? log->l_size :
1527                                                             log->next_page;
1528
1529         /* If the two offsets are the same then there is no available space. */
1530         if (oldest_off == next_free_off)
1531                 return 0;
1532         /*
1533          * If the free offset follows the oldest offset then subtract
1534          * this range from the total available pages.
1535          */
1536         free_bytes =
1537                 oldest_off < next_free_off ?
1538                         log->total_avail_pages - (next_free_off - oldest_off) :
1539                         oldest_off - next_free_off;
1540
1541         free_bytes >>= log->page_bits;
1542         return free_bytes * log->reserved;
1543 }
1544
1545 static bool check_subseq_log_page(struct ntfs_log *log,
1546                                   const struct RECORD_PAGE_HDR *rp, u32 vbo,
1547                                   u64 seq)
1548 {
1549         u64 lsn_seq;
1550         const struct NTFS_RECORD_HEADER *rhdr = &rp->rhdr;
1551         u64 lsn = le64_to_cpu(rhdr->lsn);
1552
1553         if (rhdr->sign == NTFS_FFFF_SIGNATURE || !rhdr->sign)
1554                 return false;
1555
1556         /*
1557          * If the last lsn on the page occurs was written after the page
1558          * that caused the original error then we have a fatal error.
1559          */
1560         lsn_seq = lsn >> log->file_data_bits;
1561
1562         /*
1563          * If the sequence number for the lsn the page is equal or greater
1564          * than lsn we expect, then this is a subsequent write.
1565          */
1566         return lsn_seq >= seq ||
1567                (lsn_seq == seq - 1 && log->first_page == vbo &&
1568                 vbo != (lsn_to_vbo(log, lsn) & ~log->page_mask));
1569 }
1570
1571 /*
1572  * last_log_lsn
1573  *
1574  * Walks through the log pages for a file, searching for the
1575  * last log page written to the file.
1576  */
1577 static int last_log_lsn(struct ntfs_log *log)
1578 {
1579         int err;
1580         bool usa_error = false;
1581         bool replace_page = false;
1582         bool reuse_page = log->l_flags & NTFSLOG_REUSE_TAIL;
1583         bool wrapped_file, wrapped;
1584
1585         u32 page_cnt = 1, page_pos = 1;
1586         u32 page_off = 0, page_off1 = 0, saved_off = 0;
1587         u32 final_off, second_off, final_off_prev = 0, second_off_prev = 0;
1588         u32 first_file_off = 0, second_file_off = 0;
1589         u32 part_io_count = 0;
1590         u32 tails = 0;
1591         u32 this_off, curpage_off, nextpage_off, remain_pages;
1592
1593         u64 expected_seq, seq_base = 0, lsn_base = 0;
1594         u64 best_lsn, best_lsn1, best_lsn2;
1595         u64 lsn_cur, lsn1, lsn2;
1596         u64 last_ok_lsn = reuse_page ? log->last_lsn : 0;
1597
1598         u16 cur_pos, best_page_pos;
1599
1600         struct RECORD_PAGE_HDR *page = NULL;
1601         struct RECORD_PAGE_HDR *tst_page = NULL;
1602         struct RECORD_PAGE_HDR *first_tail = NULL;
1603         struct RECORD_PAGE_HDR *second_tail = NULL;
1604         struct RECORD_PAGE_HDR *tail_page = NULL;
1605         struct RECORD_PAGE_HDR *second_tail_prev = NULL;
1606         struct RECORD_PAGE_HDR *first_tail_prev = NULL;
1607         struct RECORD_PAGE_HDR *page_bufs = NULL;
1608         struct RECORD_PAGE_HDR *best_page;
1609
1610         if (log->major_ver >= 2) {
1611                 final_off = 0x02 * log->page_size;
1612                 second_off = 0x12 * log->page_size;
1613
1614                 // 0x10 == 0x12 - 0x2
1615                 page_bufs = kmalloc(log->page_size * 0x10, GFP_NOFS);
1616                 if (!page_bufs)
1617                         return -ENOMEM;
1618         } else {
1619                 second_off = log->first_page - log->page_size;
1620                 final_off = second_off - log->page_size;
1621         }
1622
1623 next_tail:
1624         /* Read second tail page (at pos 3/0x12000). */
1625         if (read_log_page(log, second_off, &second_tail, &usa_error) ||
1626             usa_error || second_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
1627                 kfree(second_tail);
1628                 second_tail = NULL;
1629                 second_file_off = 0;
1630                 lsn2 = 0;
1631         } else {
1632                 second_file_off = hdr_file_off(log, second_tail);
1633                 lsn2 = le64_to_cpu(second_tail->record_hdr.last_end_lsn);
1634         }
1635
1636         /* Read first tail page (at pos 2/0x2000). */
1637         if (read_log_page(log, final_off, &first_tail, &usa_error) ||
1638             usa_error || first_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
1639                 kfree(first_tail);
1640                 first_tail = NULL;
1641                 first_file_off = 0;
1642                 lsn1 = 0;
1643         } else {
1644                 first_file_off = hdr_file_off(log, first_tail);
1645                 lsn1 = le64_to_cpu(first_tail->record_hdr.last_end_lsn);
1646         }
1647
1648         if (log->major_ver < 2) {
1649                 int best_page;
1650
1651                 first_tail_prev = first_tail;
1652                 final_off_prev = first_file_off;
1653                 second_tail_prev = second_tail;
1654                 second_off_prev = second_file_off;
1655                 tails = 1;
1656
1657                 if (!first_tail && !second_tail)
1658                         goto tail_read;
1659
1660                 if (first_tail && second_tail)
1661                         best_page = lsn1 < lsn2 ? 1 : 0;
1662                 else if (first_tail)
1663                         best_page = 0;
1664                 else
1665                         best_page = 1;
1666
1667                 page_off = best_page ? second_file_off : first_file_off;
1668                 seq_base = (best_page ? lsn2 : lsn1) >> log->file_data_bits;
1669                 goto tail_read;
1670         }
1671
1672         best_lsn1 = first_tail ? base_lsn(log, first_tail, first_file_off) : 0;
1673         best_lsn2 = second_tail ? base_lsn(log, second_tail, second_file_off) :
1674                                   0;
1675
1676         if (first_tail && second_tail) {
1677                 if (best_lsn1 > best_lsn2) {
1678                         best_lsn = best_lsn1;
1679                         best_page = first_tail;
1680                         this_off = first_file_off;
1681                 } else {
1682                         best_lsn = best_lsn2;
1683                         best_page = second_tail;
1684                         this_off = second_file_off;
1685                 }
1686         } else if (first_tail) {
1687                 best_lsn = best_lsn1;
1688                 best_page = first_tail;
1689                 this_off = first_file_off;
1690         } else if (second_tail) {
1691                 best_lsn = best_lsn2;
1692                 best_page = second_tail;
1693                 this_off = second_file_off;
1694         } else {
1695                 goto tail_read;
1696         }
1697
1698         best_page_pos = le16_to_cpu(best_page->page_pos);
1699
1700         if (!tails) {
1701                 if (best_page_pos == page_pos) {
1702                         seq_base = best_lsn >> log->file_data_bits;
1703                         saved_off = page_off = le32_to_cpu(best_page->file_off);
1704                         lsn_base = best_lsn;
1705
1706                         memmove(page_bufs, best_page, log->page_size);
1707
1708                         page_cnt = le16_to_cpu(best_page->page_count);
1709                         if (page_cnt > 1)
1710                                 page_pos += 1;
1711
1712                         tails = 1;
1713                 }
1714         } else if (seq_base == (best_lsn >> log->file_data_bits) &&
1715                    saved_off + log->page_size == this_off &&
1716                    lsn_base < best_lsn &&
1717                    (page_pos != page_cnt || best_page_pos == page_pos ||
1718                     best_page_pos == 1) &&
1719                    (page_pos >= page_cnt || best_page_pos == page_pos)) {
1720                 u16 bppc = le16_to_cpu(best_page->page_count);
1721
1722                 saved_off += log->page_size;
1723                 lsn_base = best_lsn;
1724
1725                 memmove(Add2Ptr(page_bufs, tails * log->page_size), best_page,
1726                         log->page_size);
1727
1728                 tails += 1;
1729
1730                 if (best_page_pos != bppc) {
1731                         page_cnt = bppc;
1732                         page_pos = best_page_pos;
1733
1734                         if (page_cnt > 1)
1735                                 page_pos += 1;
1736                 } else {
1737                         page_pos = page_cnt = 1;
1738                 }
1739         } else {
1740                 kfree(first_tail);
1741                 kfree(second_tail);
1742                 goto tail_read;
1743         }
1744
1745         kfree(first_tail_prev);
1746         first_tail_prev = first_tail;
1747         final_off_prev = first_file_off;
1748         first_tail = NULL;
1749
1750         kfree(second_tail_prev);
1751         second_tail_prev = second_tail;
1752         second_off_prev = second_file_off;
1753         second_tail = NULL;
1754
1755         final_off += log->page_size;
1756         second_off += log->page_size;
1757
1758         if (tails < 0x10)
1759                 goto next_tail;
1760 tail_read:
1761         first_tail = first_tail_prev;
1762         final_off = final_off_prev;
1763
1764         second_tail = second_tail_prev;
1765         second_off = second_off_prev;
1766
1767         page_cnt = page_pos = 1;
1768
1769         curpage_off = seq_base == log->seq_num ? min(log->next_page, page_off) :
1770                                                  log->next_page;
1771
1772         wrapped_file =
1773                 curpage_off == log->first_page &&
1774                 !(log->l_flags & (NTFSLOG_NO_LAST_LSN | NTFSLOG_REUSE_TAIL));
1775
1776         expected_seq = wrapped_file ? (log->seq_num + 1) : log->seq_num;
1777
1778         nextpage_off = curpage_off;
1779
1780 next_page:
1781         tail_page = NULL;
1782         /* Read the next log page. */
1783         err = read_log_page(log, curpage_off, &page, &usa_error);
1784
1785         /* Compute the next log page offset the file. */
1786         nextpage_off = next_page_off(log, curpage_off);
1787         wrapped = nextpage_off == log->first_page;
1788
1789         if (tails > 1) {
1790                 struct RECORD_PAGE_HDR *cur_page =
1791                         Add2Ptr(page_bufs, curpage_off - page_off);
1792
1793                 if (curpage_off == saved_off) {
1794                         tail_page = cur_page;
1795                         goto use_tail_page;
1796                 }
1797
1798                 if (page_off > curpage_off || curpage_off >= saved_off)
1799                         goto use_tail_page;
1800
1801                 if (page_off1)
1802                         goto use_cur_page;
1803
1804                 if (!err && !usa_error &&
1805                     page->rhdr.sign == NTFS_RCRD_SIGNATURE &&
1806                     cur_page->rhdr.lsn == page->rhdr.lsn &&
1807                     cur_page->record_hdr.next_record_off ==
1808                             page->record_hdr.next_record_off &&
1809                     ((page_pos == page_cnt &&
1810                       le16_to_cpu(page->page_pos) == 1) ||
1811                      (page_pos != page_cnt &&
1812                       le16_to_cpu(page->page_pos) == page_pos + 1 &&
1813                       le16_to_cpu(page->page_count) == page_cnt))) {
1814                         cur_page = NULL;
1815                         goto use_tail_page;
1816                 }
1817
1818                 page_off1 = page_off;
1819
1820 use_cur_page:
1821
1822                 lsn_cur = le64_to_cpu(cur_page->rhdr.lsn);
1823
1824                 if (last_ok_lsn !=
1825                             le64_to_cpu(cur_page->record_hdr.last_end_lsn) &&
1826                     ((lsn_cur >> log->file_data_bits) +
1827                      ((curpage_off <
1828                        (lsn_to_vbo(log, lsn_cur) & ~log->page_mask)) ?
1829                               1 :
1830                               0)) != expected_seq) {
1831                         goto check_tail;
1832                 }
1833
1834                 if (!is_log_record_end(cur_page)) {
1835                         tail_page = NULL;
1836                         last_ok_lsn = lsn_cur;
1837                         goto next_page_1;
1838                 }
1839
1840                 log->seq_num = expected_seq;
1841                 log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
1842                 log->last_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
1843                 log->ra->current_lsn = cur_page->record_hdr.last_end_lsn;
1844
1845                 if (log->record_header_len <=
1846                     log->page_size -
1847                             le16_to_cpu(cur_page->record_hdr.next_record_off)) {
1848                         log->l_flags |= NTFSLOG_REUSE_TAIL;
1849                         log->next_page = curpage_off;
1850                 } else {
1851                         log->l_flags &= ~NTFSLOG_REUSE_TAIL;
1852                         log->next_page = nextpage_off;
1853                 }
1854
1855                 if (wrapped_file)
1856                         log->l_flags |= NTFSLOG_WRAPPED;
1857
1858                 last_ok_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
1859                 goto next_page_1;
1860         }
1861
1862         /*
1863          * If we are at the expected first page of a transfer check to see
1864          * if either tail copy is at this offset.
1865          * If this page is the last page of a transfer, check if we wrote
1866          * a subsequent tail copy.
1867          */
1868         if (page_cnt == page_pos || page_cnt == page_pos + 1) {
1869                 /*
1870                  * Check if the offset matches either the first or second
1871                  * tail copy. It is possible it will match both.
1872                  */
1873                 if (curpage_off == final_off)
1874                         tail_page = first_tail;
1875
1876                 /*
1877                  * If we already matched on the first page then
1878                  * check the ending lsn's.
1879                  */
1880                 if (curpage_off == second_off) {
1881                         if (!tail_page ||
1882                             (second_tail &&
1883                              le64_to_cpu(second_tail->record_hdr.last_end_lsn) >
1884                                      le64_to_cpu(first_tail->record_hdr
1885                                                          .last_end_lsn))) {
1886                                 tail_page = second_tail;
1887                         }
1888                 }
1889         }
1890
1891 use_tail_page:
1892         if (tail_page) {
1893                 /* We have a candidate for a tail copy. */
1894                 lsn_cur = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
1895
1896                 if (last_ok_lsn < lsn_cur) {
1897                         /*
1898                          * If the sequence number is not expected,
1899                          * then don't use the tail copy.
1900                          */
1901                         if (expected_seq != (lsn_cur >> log->file_data_bits))
1902                                 tail_page = NULL;
1903                 } else if (last_ok_lsn > lsn_cur) {
1904                         /*
1905                          * If the last lsn is greater than the one on
1906                          * this page then forget this tail.
1907                          */
1908                         tail_page = NULL;
1909                 }
1910         }
1911
1912         /*
1913          *If we have an error on the current page,
1914          * we will break of this loop.
1915          */
1916         if (err || usa_error)
1917                 goto check_tail;
1918
1919         /*
1920          * Done if the last lsn on this page doesn't match the previous known
1921          * last lsn or the sequence number is not expected.
1922          */
1923         lsn_cur = le64_to_cpu(page->rhdr.lsn);
1924         if (last_ok_lsn != lsn_cur &&
1925             expected_seq != (lsn_cur >> log->file_data_bits)) {
1926                 goto check_tail;
1927         }
1928
1929         /*
1930          * Check that the page position and page count values are correct.
1931          * If this is the first page of a transfer the position must be 1
1932          * and the count will be unknown.
1933          */
1934         if (page_cnt == page_pos) {
1935                 if (page->page_pos != cpu_to_le16(1) &&
1936                     (!reuse_page || page->page_pos != page->page_count)) {
1937                         /*
1938                          * If the current page is the first page we are
1939                          * looking at and we are reusing this page then
1940                          * it can be either the first or last page of a
1941                          * transfer. Otherwise it can only be the first.
1942                          */
1943                         goto check_tail;
1944                 }
1945         } else if (le16_to_cpu(page->page_count) != page_cnt ||
1946                    le16_to_cpu(page->page_pos) != page_pos + 1) {
1947                 /*
1948                  * The page position better be 1 more than the last page
1949                  * position and the page count better match.
1950                  */
1951                 goto check_tail;
1952         }
1953
1954         /*
1955          * We have a valid page the file and may have a valid page
1956          * the tail copy area.
1957          * If the tail page was written after the page the file then
1958          * break of the loop.
1959          */
1960         if (tail_page &&
1961             le64_to_cpu(tail_page->record_hdr.last_end_lsn) > lsn_cur) {
1962                 /* Remember if we will replace the page. */
1963                 replace_page = true;
1964                 goto check_tail;
1965         }
1966
1967         tail_page = NULL;
1968
1969         if (is_log_record_end(page)) {
1970                 /*
1971                  * Since we have read this page we know the sequence number
1972                  * is the same as our expected value.
1973                  */
1974                 log->seq_num = expected_seq;
1975                 log->last_lsn = le64_to_cpu(page->record_hdr.last_end_lsn);
1976                 log->ra->current_lsn = page->record_hdr.last_end_lsn;
1977                 log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
1978
1979                 /*
1980                  * If there is room on this page for another header then
1981                  * remember we want to reuse the page.
1982                  */
1983                 if (log->record_header_len <=
1984                     log->page_size -
1985                             le16_to_cpu(page->record_hdr.next_record_off)) {
1986                         log->l_flags |= NTFSLOG_REUSE_TAIL;
1987                         log->next_page = curpage_off;
1988                 } else {
1989                         log->l_flags &= ~NTFSLOG_REUSE_TAIL;
1990                         log->next_page = nextpage_off;
1991                 }
1992
1993                 /* Remember if we wrapped the log file. */
1994                 if (wrapped_file)
1995                         log->l_flags |= NTFSLOG_WRAPPED;
1996         }
1997
1998         /*
1999          * Remember the last page count and position.
2000          * Also remember the last known lsn.
2001          */
2002         page_cnt = le16_to_cpu(page->page_count);
2003         page_pos = le16_to_cpu(page->page_pos);
2004         last_ok_lsn = le64_to_cpu(page->rhdr.lsn);
2005
2006 next_page_1:
2007
2008         if (wrapped) {
2009                 expected_seq += 1;
2010                 wrapped_file = 1;
2011         }
2012
2013         curpage_off = nextpage_off;
2014         kfree(page);
2015         page = NULL;
2016         reuse_page = 0;
2017         goto next_page;
2018
2019 check_tail:
2020         if (tail_page) {
2021                 log->seq_num = expected_seq;
2022                 log->last_lsn = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
2023                 log->ra->current_lsn = tail_page->record_hdr.last_end_lsn;
2024                 log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
2025
2026                 if (log->page_size -
2027                             le16_to_cpu(
2028                                     tail_page->record_hdr.next_record_off) >=
2029                     log->record_header_len) {
2030                         log->l_flags |= NTFSLOG_REUSE_TAIL;
2031                         log->next_page = curpage_off;
2032                 } else {
2033                         log->l_flags &= ~NTFSLOG_REUSE_TAIL;
2034                         log->next_page = nextpage_off;
2035                 }
2036
2037                 if (wrapped)
2038                         log->l_flags |= NTFSLOG_WRAPPED;
2039         }
2040
2041         /* Remember that the partial IO will start at the next page. */
2042         second_off = nextpage_off;
2043
2044         /*
2045          * If the next page is the first page of the file then update
2046          * the sequence number for log records which begon the next page.
2047          */
2048         if (wrapped)
2049                 expected_seq += 1;
2050
2051         /*
2052          * If we have a tail copy or are performing single page I/O we can
2053          * immediately look at the next page.
2054          */
2055         if (replace_page || (log->ra->flags & RESTART_SINGLE_PAGE_IO)) {
2056                 page_cnt = 2;
2057                 page_pos = 1;
2058                 goto check_valid;
2059         }
2060
2061         if (page_pos != page_cnt)
2062                 goto check_valid;
2063         /*
2064          * If the next page causes us to wrap to the beginning of the log
2065          * file then we know which page to check next.
2066          */
2067         if (wrapped) {
2068                 page_cnt = 2;
2069                 page_pos = 1;
2070                 goto check_valid;
2071         }
2072
2073         cur_pos = 2;
2074
2075 next_test_page:
2076         kfree(tst_page);
2077         tst_page = NULL;
2078
2079         /* Walk through the file, reading log pages. */
2080         err = read_log_page(log, nextpage_off, &tst_page, &usa_error);
2081
2082         /*
2083          * If we get a USA error then assume that we correctly found
2084          * the end of the original transfer.
2085          */
2086         if (usa_error)
2087                 goto file_is_valid;
2088
2089         /*
2090          * If we were able to read the page, we examine it to see if it
2091          * is the same or different Io block.
2092          */
2093         if (err)
2094                 goto next_test_page_1;
2095
2096         if (le16_to_cpu(tst_page->page_pos) == cur_pos &&
2097             check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
2098                 page_cnt = le16_to_cpu(tst_page->page_count) + 1;
2099                 page_pos = le16_to_cpu(tst_page->page_pos);
2100                 goto check_valid;
2101         } else {
2102                 goto file_is_valid;
2103         }
2104
2105 next_test_page_1:
2106
2107         nextpage_off = next_page_off(log, curpage_off);
2108         wrapped = nextpage_off == log->first_page;
2109
2110         if (wrapped) {
2111                 expected_seq += 1;
2112                 page_cnt = 2;
2113                 page_pos = 1;
2114         }
2115
2116         cur_pos += 1;
2117         part_io_count += 1;
2118         if (!wrapped)
2119                 goto next_test_page;
2120
2121 check_valid:
2122         /* Skip over the remaining pages this transfer. */
2123         remain_pages = page_cnt - page_pos - 1;
2124         part_io_count += remain_pages;
2125
2126         while (remain_pages--) {
2127                 nextpage_off = next_page_off(log, curpage_off);
2128                 wrapped = nextpage_off == log->first_page;
2129
2130                 if (wrapped)
2131                         expected_seq += 1;
2132         }
2133
2134         /* Call our routine to check this log page. */
2135         kfree(tst_page);
2136         tst_page = NULL;
2137
2138         err = read_log_page(log, nextpage_off, &tst_page, &usa_error);
2139         if (!err && !usa_error &&
2140             check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
2141                 err = -EINVAL;
2142                 goto out;
2143         }
2144
2145 file_is_valid:
2146
2147         /* We have a valid file. */
2148         if (page_off1 || tail_page) {
2149                 struct RECORD_PAGE_HDR *tmp_page;
2150
2151                 if (sb_rdonly(log->ni->mi.sbi->sb)) {
2152                         err = -EROFS;
2153                         goto out;
2154                 }
2155
2156                 if (page_off1) {
2157                         tmp_page = Add2Ptr(page_bufs, page_off1 - page_off);
2158                         tails -= (page_off1 - page_off) / log->page_size;
2159                         if (!tail_page)
2160                                 tails -= 1;
2161                 } else {
2162                         tmp_page = tail_page;
2163                         tails = 1;
2164                 }
2165
2166                 while (tails--) {
2167                         u64 off = hdr_file_off(log, tmp_page);
2168
2169                         if (!page) {
2170                                 page = kmalloc(log->page_size, GFP_NOFS);
2171                                 if (!page)
2172                                         return -ENOMEM;
2173                         }
2174
2175                         /*
2176                          * Correct page and copy the data from this page
2177                          * into it and flush it to disk.
2178                          */
2179                         memcpy(page, tmp_page, log->page_size);
2180
2181                         /* Fill last flushed lsn value flush the page. */
2182                         if (log->major_ver < 2)
2183                                 page->rhdr.lsn = page->record_hdr.last_end_lsn;
2184                         else
2185                                 page->file_off = 0;
2186
2187                         page->page_pos = page->page_count = cpu_to_le16(1);
2188
2189                         ntfs_fix_pre_write(&page->rhdr, log->page_size);
2190
2191                         err = ntfs_sb_write_run(log->ni->mi.sbi,
2192                                                 &log->ni->file.run, off, page,
2193                                                 log->page_size, 0);
2194
2195                         if (err)
2196                                 goto out;
2197
2198                         if (part_io_count && second_off == off) {
2199                                 second_off += log->page_size;
2200                                 part_io_count -= 1;
2201                         }
2202
2203                         tmp_page = Add2Ptr(tmp_page, log->page_size);
2204                 }
2205         }
2206
2207         if (part_io_count) {
2208                 if (sb_rdonly(log->ni->mi.sbi->sb)) {
2209                         err = -EROFS;
2210                         goto out;
2211                 }
2212         }
2213
2214 out:
2215         kfree(second_tail);
2216         kfree(first_tail);
2217         kfree(page);
2218         kfree(tst_page);
2219         kfree(page_bufs);
2220
2221         return err;
2222 }
2223
2224 /*
2225  * read_log_rec_buf - Copy a log record from the file to a buffer.
2226  *
2227  * The log record may span several log pages and may even wrap the file.
2228  */
2229 static int read_log_rec_buf(struct ntfs_log *log,
2230                             const struct LFS_RECORD_HDR *rh, void *buffer)
2231 {
2232         int err;
2233         struct RECORD_PAGE_HDR *ph = NULL;
2234         u64 lsn = le64_to_cpu(rh->this_lsn);
2235         u32 vbo = lsn_to_vbo(log, lsn) & ~log->page_mask;
2236         u32 off = lsn_to_page_off(log, lsn) + log->record_header_len;
2237         u32 data_len = le32_to_cpu(rh->client_data_len);
2238
2239         /*
2240          * While there are more bytes to transfer,
2241          * we continue to attempt to perform the read.
2242          */
2243         for (;;) {
2244                 bool usa_error;
2245                 u32 tail = log->page_size - off;
2246
2247                 if (tail >= data_len)
2248                         tail = data_len;
2249
2250                 data_len -= tail;
2251
2252                 err = read_log_page(log, vbo, &ph, &usa_error);
2253                 if (err)
2254                         goto out;
2255
2256                 /*
2257                  * The last lsn on this page better be greater or equal
2258                  * to the lsn we are copying.
2259                  */
2260                 if (lsn > le64_to_cpu(ph->rhdr.lsn)) {
2261                         err = -EINVAL;
2262                         goto out;
2263                 }
2264
2265                 memcpy(buffer, Add2Ptr(ph, off), tail);
2266
2267                 /* If there are no more bytes to transfer, we exit the loop. */
2268                 if (!data_len) {
2269                         if (!is_log_record_end(ph) ||
2270                             lsn > le64_to_cpu(ph->record_hdr.last_end_lsn)) {
2271                                 err = -EINVAL;
2272                                 goto out;
2273                         }
2274                         break;
2275                 }
2276
2277                 if (ph->rhdr.lsn == ph->record_hdr.last_end_lsn ||
2278                     lsn > le64_to_cpu(ph->rhdr.lsn)) {
2279                         err = -EINVAL;
2280                         goto out;
2281                 }
2282
2283                 vbo = next_page_off(log, vbo);
2284                 off = log->data_off;
2285
2286                 /*
2287                  * Adjust our pointer the user's buffer to transfer
2288                  * the next block to.
2289                  */
2290                 buffer = Add2Ptr(buffer, tail);
2291         }
2292
2293 out:
2294         kfree(ph);
2295         return err;
2296 }
2297
2298 static int read_rst_area(struct ntfs_log *log, struct NTFS_RESTART **rst_,
2299                          u64 *lsn)
2300 {
2301         int err;
2302         struct LFS_RECORD_HDR *rh = NULL;
2303         const struct CLIENT_REC *cr =
2304                 Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
2305         u64 lsnr, lsnc = le64_to_cpu(cr->restart_lsn);
2306         u32 len;
2307         struct NTFS_RESTART *rst;
2308
2309         *lsn = 0;
2310         *rst_ = NULL;
2311
2312         /* If the client doesn't have a restart area, go ahead and exit now. */
2313         if (!lsnc)
2314                 return 0;
2315
2316         err = read_log_page(log, lsn_to_vbo(log, lsnc),
2317                             (struct RECORD_PAGE_HDR **)&rh, NULL);
2318         if (err)
2319                 return err;
2320
2321         rst = NULL;
2322         lsnr = le64_to_cpu(rh->this_lsn);
2323
2324         if (lsnc != lsnr) {
2325                 /* If the lsn values don't match, then the disk is corrupt. */
2326                 err = -EINVAL;
2327                 goto out;
2328         }
2329
2330         *lsn = lsnr;
2331         len = le32_to_cpu(rh->client_data_len);
2332
2333         if (!len) {
2334                 err = 0;
2335                 goto out;
2336         }
2337
2338         if (len < sizeof(struct NTFS_RESTART)) {
2339                 err = -EINVAL;
2340                 goto out;
2341         }
2342
2343         rst = kmalloc(len, GFP_NOFS);
2344         if (!rst) {
2345                 err = -ENOMEM;
2346                 goto out;
2347         }
2348
2349         /* Copy the data into the 'rst' buffer. */
2350         err = read_log_rec_buf(log, rh, rst);
2351         if (err)
2352                 goto out;
2353
2354         *rst_ = rst;
2355         rst = NULL;
2356
2357 out:
2358         kfree(rh);
2359         kfree(rst);
2360
2361         return err;
2362 }
2363
2364 static int find_log_rec(struct ntfs_log *log, u64 lsn, struct lcb *lcb)
2365 {
2366         int err;
2367         struct LFS_RECORD_HDR *rh = lcb->lrh;
2368         u32 rec_len, len;
2369
2370         /* Read the record header for this lsn. */
2371         if (!rh) {
2372                 err = read_log_page(log, lsn_to_vbo(log, lsn),
2373                                     (struct RECORD_PAGE_HDR **)&rh, NULL);
2374
2375                 lcb->lrh = rh;
2376                 if (err)
2377                         return err;
2378         }
2379
2380         /*
2381          * If the lsn the log record doesn't match the desired
2382          * lsn then the disk is corrupt.
2383          */
2384         if (lsn != le64_to_cpu(rh->this_lsn))
2385                 return -EINVAL;
2386
2387         len = le32_to_cpu(rh->client_data_len);
2388
2389         /*
2390          * Check that the length field isn't greater than the total
2391          * available space the log file.
2392          */
2393         rec_len = len + log->record_header_len;
2394         if (rec_len >= log->total_avail)
2395                 return -EINVAL;
2396
2397         /*
2398          * If the entire log record is on this log page,
2399          * put a pointer to the log record the context block.
2400          */
2401         if (rh->flags & LOG_RECORD_MULTI_PAGE) {
2402                 void *lr = kmalloc(len, GFP_NOFS);
2403
2404                 if (!lr)
2405                         return -ENOMEM;
2406
2407                 lcb->log_rec = lr;
2408                 lcb->alloc = true;
2409
2410                 /* Copy the data into the buffer returned. */
2411                 err = read_log_rec_buf(log, rh, lr);
2412                 if (err)
2413                         return err;
2414         } else {
2415                 /* If beyond the end of the current page -> an error. */
2416                 u32 page_off = lsn_to_page_off(log, lsn);
2417
2418                 if (page_off + len + log->record_header_len > log->page_size)
2419                         return -EINVAL;
2420
2421                 lcb->log_rec = Add2Ptr(rh, sizeof(struct LFS_RECORD_HDR));
2422                 lcb->alloc = false;
2423         }
2424
2425         return 0;
2426 }
2427
2428 /*
2429  * read_log_rec_lcb - Init the query operation.
2430  */
2431 static int read_log_rec_lcb(struct ntfs_log *log, u64 lsn, u32 ctx_mode,
2432                             struct lcb **lcb_)
2433 {
2434         int err;
2435         const struct CLIENT_REC *cr;
2436         struct lcb *lcb;
2437
2438         switch (ctx_mode) {
2439         case lcb_ctx_undo_next:
2440         case lcb_ctx_prev:
2441         case lcb_ctx_next:
2442                 break;
2443         default:
2444                 return -EINVAL;
2445         }
2446
2447         /* Check that the given lsn is the legal range for this client. */
2448         cr = Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
2449
2450         if (!verify_client_lsn(log, cr, lsn))
2451                 return -EINVAL;
2452
2453         lcb = kzalloc(sizeof(struct lcb), GFP_NOFS);
2454         if (!lcb)
2455                 return -ENOMEM;
2456         lcb->client = log->client_id;
2457         lcb->ctx_mode = ctx_mode;
2458
2459         /* Find the log record indicated by the given lsn. */
2460         err = find_log_rec(log, lsn, lcb);
2461         if (err)
2462                 goto out;
2463
2464         *lcb_ = lcb;
2465         return 0;
2466
2467 out:
2468         lcb_put(lcb);
2469         *lcb_ = NULL;
2470         return err;
2471 }
2472
2473 /*
2474  * find_client_next_lsn
2475  *
2476  * Attempt to find the next lsn to return to a client based on the context mode.
2477  */
2478 static int find_client_next_lsn(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
2479 {
2480         int err;
2481         u64 next_lsn;
2482         struct LFS_RECORD_HDR *hdr;
2483
2484         hdr = lcb->lrh;
2485         *lsn = 0;
2486
2487         if (lcb_ctx_next != lcb->ctx_mode)
2488                 goto check_undo_next;
2489
2490         /* Loop as long as another lsn can be found. */
2491         for (;;) {
2492                 u64 current_lsn;
2493
2494                 err = next_log_lsn(log, hdr, &current_lsn);
2495                 if (err)
2496                         goto out;
2497
2498                 if (!current_lsn)
2499                         break;
2500
2501                 if (hdr != lcb->lrh)
2502                         kfree(hdr);
2503
2504                 hdr = NULL;
2505                 err = read_log_page(log, lsn_to_vbo(log, current_lsn),
2506                                     (struct RECORD_PAGE_HDR **)&hdr, NULL);
2507                 if (err)
2508                         goto out;
2509
2510                 if (memcmp(&hdr->client, &lcb->client,
2511                            sizeof(struct CLIENT_ID))) {
2512                         /*err = -EINVAL; */
2513                 } else if (LfsClientRecord == hdr->record_type) {
2514                         kfree(lcb->lrh);
2515                         lcb->lrh = hdr;
2516                         *lsn = current_lsn;
2517                         return 0;
2518                 }
2519         }
2520
2521 out:
2522         if (hdr != lcb->lrh)
2523                 kfree(hdr);
2524         return err;
2525
2526 check_undo_next:
2527         if (lcb_ctx_undo_next == lcb->ctx_mode)
2528                 next_lsn = le64_to_cpu(hdr->client_undo_next_lsn);
2529         else if (lcb_ctx_prev == lcb->ctx_mode)
2530                 next_lsn = le64_to_cpu(hdr->client_prev_lsn);
2531         else
2532                 return 0;
2533
2534         if (!next_lsn)
2535                 return 0;
2536
2537         if (!verify_client_lsn(
2538                     log, Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off)),
2539                     next_lsn))
2540                 return 0;
2541
2542         hdr = NULL;
2543         err = read_log_page(log, lsn_to_vbo(log, next_lsn),
2544                             (struct RECORD_PAGE_HDR **)&hdr, NULL);
2545         if (err)
2546                 return err;
2547         kfree(lcb->lrh);
2548         lcb->lrh = hdr;
2549
2550         *lsn = next_lsn;
2551
2552         return 0;
2553 }
2554
2555 static int read_next_log_rec(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
2556 {
2557         int err;
2558
2559         err = find_client_next_lsn(log, lcb, lsn);
2560         if (err)
2561                 return err;
2562
2563         if (!*lsn)
2564                 return 0;
2565
2566         if (lcb->alloc)
2567                 kfree(lcb->log_rec);
2568
2569         lcb->log_rec = NULL;
2570         lcb->alloc = false;
2571         kfree(lcb->lrh);
2572         lcb->lrh = NULL;
2573
2574         return find_log_rec(log, *lsn, lcb);
2575 }
2576
2577 bool check_index_header(const struct INDEX_HDR *hdr, size_t bytes)
2578 {
2579         __le16 mask;
2580         u32 min_de, de_off, used, total;
2581         const struct NTFS_DE *e;
2582
2583         if (hdr_has_subnode(hdr)) {
2584                 min_de = sizeof(struct NTFS_DE) + sizeof(u64);
2585                 mask = NTFS_IE_HAS_SUBNODES;
2586         } else {
2587                 min_de = sizeof(struct NTFS_DE);
2588                 mask = 0;
2589         }
2590
2591         de_off = le32_to_cpu(hdr->de_off);
2592         used = le32_to_cpu(hdr->used);
2593         total = le32_to_cpu(hdr->total);
2594
2595         if (de_off > bytes - min_de || used > bytes || total > bytes ||
2596             de_off + min_de > used || used > total) {
2597                 return false;
2598         }
2599
2600         e = Add2Ptr(hdr, de_off);
2601         for (;;) {
2602                 u16 esize = le16_to_cpu(e->size);
2603                 struct NTFS_DE *next = Add2Ptr(e, esize);
2604
2605                 if (esize < min_de || PtrOffset(hdr, next) > used ||
2606                     (e->flags & NTFS_IE_HAS_SUBNODES) != mask) {
2607                         return false;
2608                 }
2609
2610                 if (de_is_last(e))
2611                         break;
2612
2613                 e = next;
2614         }
2615
2616         return true;
2617 }
2618
2619 static inline bool check_index_buffer(const struct INDEX_BUFFER *ib, u32 bytes)
2620 {
2621         u16 fo;
2622         const struct NTFS_RECORD_HEADER *r = &ib->rhdr;
2623
2624         if (r->sign != NTFS_INDX_SIGNATURE)
2625                 return false;
2626
2627         fo = (SECTOR_SIZE - ((bytes >> SECTOR_SHIFT) + 1) * sizeof(short));
2628
2629         if (le16_to_cpu(r->fix_off) > fo)
2630                 return false;
2631
2632         if ((le16_to_cpu(r->fix_num) - 1) * SECTOR_SIZE != bytes)
2633                 return false;
2634
2635         return check_index_header(&ib->ihdr,
2636                                   bytes - offsetof(struct INDEX_BUFFER, ihdr));
2637 }
2638
2639 static inline bool check_index_root(const struct ATTRIB *attr,
2640                                     struct ntfs_sb_info *sbi)
2641 {
2642         bool ret;
2643         const struct INDEX_ROOT *root = resident_data(attr);
2644         u8 index_bits = le32_to_cpu(root->index_block_size) >=
2645                                         sbi->cluster_size ?
2646                                 sbi->cluster_bits :
2647                                 SECTOR_SHIFT;
2648         u8 block_clst = root->index_block_clst;
2649
2650         if (le32_to_cpu(attr->res.data_size) < sizeof(struct INDEX_ROOT) ||
2651             (root->type != ATTR_NAME && root->type != ATTR_ZERO) ||
2652             (root->type == ATTR_NAME &&
2653              root->rule != NTFS_COLLATION_TYPE_FILENAME) ||
2654             (le32_to_cpu(root->index_block_size) !=
2655              (block_clst << index_bits)) ||
2656             (block_clst != 1 && block_clst != 2 && block_clst != 4 &&
2657              block_clst != 8 && block_clst != 0x10 && block_clst != 0x20 &&
2658              block_clst != 0x40 && block_clst != 0x80)) {
2659                 return false;
2660         }
2661
2662         ret = check_index_header(&root->ihdr,
2663                                  le32_to_cpu(attr->res.data_size) -
2664                                          offsetof(struct INDEX_ROOT, ihdr));
2665         return ret;
2666 }
2667
2668 static inline bool check_attr(const struct MFT_REC *rec,
2669                               const struct ATTRIB *attr,
2670                               struct ntfs_sb_info *sbi)
2671 {
2672         u32 asize = le32_to_cpu(attr->size);
2673         u32 rsize = 0;
2674         u64 dsize, svcn, evcn;
2675         u16 run_off;
2676
2677         /* Check the fixed part of the attribute record header. */
2678         if (asize >= sbi->record_size ||
2679             asize + PtrOffset(rec, attr) >= sbi->record_size ||
2680             (attr->name_len &&
2681              le16_to_cpu(attr->name_off) + attr->name_len * sizeof(short) >
2682                      asize)) {
2683                 return false;
2684         }
2685
2686         /* Check the attribute fields. */
2687         switch (attr->non_res) {
2688         case 0:
2689                 rsize = le32_to_cpu(attr->res.data_size);
2690                 if (rsize >= asize ||
2691                     le16_to_cpu(attr->res.data_off) + rsize > asize) {
2692                         return false;
2693                 }
2694                 break;
2695
2696         case 1:
2697                 dsize = le64_to_cpu(attr->nres.data_size);
2698                 svcn = le64_to_cpu(attr->nres.svcn);
2699                 evcn = le64_to_cpu(attr->nres.evcn);
2700                 run_off = le16_to_cpu(attr->nres.run_off);
2701
2702                 if (svcn > evcn + 1 || run_off >= asize ||
2703                     le64_to_cpu(attr->nres.valid_size) > dsize ||
2704                     dsize > le64_to_cpu(attr->nres.alloc_size)) {
2705                         return false;
2706                 }
2707
2708                 if (run_off > asize)
2709                         return false;
2710
2711                 if (run_unpack(NULL, sbi, 0, svcn, evcn, svcn,
2712                                Add2Ptr(attr, run_off), asize - run_off) < 0) {
2713                         return false;
2714                 }
2715
2716                 return true;
2717
2718         default:
2719                 return false;
2720         }
2721
2722         switch (attr->type) {
2723         case ATTR_NAME:
2724                 if (fname_full_size(Add2Ptr(
2725                             attr, le16_to_cpu(attr->res.data_off))) > asize) {
2726                         return false;
2727                 }
2728                 break;
2729
2730         case ATTR_ROOT:
2731                 return check_index_root(attr, sbi);
2732
2733         case ATTR_STD:
2734                 if (rsize < sizeof(struct ATTR_STD_INFO5) &&
2735                     rsize != sizeof(struct ATTR_STD_INFO)) {
2736                         return false;
2737                 }
2738                 break;
2739
2740         case ATTR_LIST:
2741         case ATTR_ID:
2742         case ATTR_SECURE:
2743         case ATTR_LABEL:
2744         case ATTR_VOL_INFO:
2745         case ATTR_DATA:
2746         case ATTR_ALLOC:
2747         case ATTR_BITMAP:
2748         case ATTR_REPARSE:
2749         case ATTR_EA_INFO:
2750         case ATTR_EA:
2751         case ATTR_PROPERTYSET:
2752         case ATTR_LOGGED_UTILITY_STREAM:
2753                 break;
2754
2755         default:
2756                 return false;
2757         }
2758
2759         return true;
2760 }
2761
2762 static inline bool check_file_record(const struct MFT_REC *rec,
2763                                      const struct MFT_REC *rec2,
2764                                      struct ntfs_sb_info *sbi)
2765 {
2766         const struct ATTRIB *attr;
2767         u16 fo = le16_to_cpu(rec->rhdr.fix_off);
2768         u16 fn = le16_to_cpu(rec->rhdr.fix_num);
2769         u16 ao = le16_to_cpu(rec->attr_off);
2770         u32 rs = sbi->record_size;
2771
2772         /* Check the file record header for consistency. */
2773         if (rec->rhdr.sign != NTFS_FILE_SIGNATURE ||
2774             fo > (SECTOR_SIZE - ((rs >> SECTOR_SHIFT) + 1) * sizeof(short)) ||
2775             (fn - 1) * SECTOR_SIZE != rs || ao < MFTRECORD_FIXUP_OFFSET_1 ||
2776             ao > sbi->record_size - SIZEOF_RESIDENT || !is_rec_inuse(rec) ||
2777             le32_to_cpu(rec->total) != rs) {
2778                 return false;
2779         }
2780
2781         /* Loop to check all of the attributes. */
2782         for (attr = Add2Ptr(rec, ao); attr->type != ATTR_END;
2783              attr = Add2Ptr(attr, le32_to_cpu(attr->size))) {
2784                 if (check_attr(rec, attr, sbi))
2785                         continue;
2786                 return false;
2787         }
2788
2789         return true;
2790 }
2791
2792 static inline int check_lsn(const struct NTFS_RECORD_HEADER *hdr,
2793                             const u64 *rlsn)
2794 {
2795         u64 lsn;
2796
2797         if (!rlsn)
2798                 return true;
2799
2800         lsn = le64_to_cpu(hdr->lsn);
2801
2802         if (hdr->sign == NTFS_HOLE_SIGNATURE)
2803                 return false;
2804
2805         if (*rlsn > lsn)
2806                 return true;
2807
2808         return false;
2809 }
2810
2811 static inline bool check_if_attr(const struct MFT_REC *rec,
2812                                  const struct LOG_REC_HDR *lrh)
2813 {
2814         u16 ro = le16_to_cpu(lrh->record_off);
2815         u16 o = le16_to_cpu(rec->attr_off);
2816         const struct ATTRIB *attr = Add2Ptr(rec, o);
2817
2818         while (o < ro) {
2819                 u32 asize;
2820
2821                 if (attr->type == ATTR_END)
2822                         break;
2823
2824                 asize = le32_to_cpu(attr->size);
2825                 if (!asize)
2826                         break;
2827
2828                 o += asize;
2829                 attr = Add2Ptr(attr, asize);
2830         }
2831
2832         return o == ro;
2833 }
2834
2835 static inline bool check_if_index_root(const struct MFT_REC *rec,
2836                                        const struct LOG_REC_HDR *lrh)
2837 {
2838         u16 ro = le16_to_cpu(lrh->record_off);
2839         u16 o = le16_to_cpu(rec->attr_off);
2840         const struct ATTRIB *attr = Add2Ptr(rec, o);
2841
2842         while (o < ro) {
2843                 u32 asize;
2844
2845                 if (attr->type == ATTR_END)
2846                         break;
2847
2848                 asize = le32_to_cpu(attr->size);
2849                 if (!asize)
2850                         break;
2851
2852                 o += asize;
2853                 attr = Add2Ptr(attr, asize);
2854         }
2855
2856         return o == ro && attr->type == ATTR_ROOT;
2857 }
2858
2859 static inline bool check_if_root_index(const struct ATTRIB *attr,
2860                                        const struct INDEX_HDR *hdr,
2861                                        const struct LOG_REC_HDR *lrh)
2862 {
2863         u16 ao = le16_to_cpu(lrh->attr_off);
2864         u32 de_off = le32_to_cpu(hdr->de_off);
2865         u32 o = PtrOffset(attr, hdr) + de_off;
2866         const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
2867         u32 asize = le32_to_cpu(attr->size);
2868
2869         while (o < ao) {
2870                 u16 esize;
2871
2872                 if (o >= asize)
2873                         break;
2874
2875                 esize = le16_to_cpu(e->size);
2876                 if (!esize)
2877                         break;
2878
2879                 o += esize;
2880                 e = Add2Ptr(e, esize);
2881         }
2882
2883         return o == ao;
2884 }
2885
2886 static inline bool check_if_alloc_index(const struct INDEX_HDR *hdr,
2887                                         u32 attr_off)
2888 {
2889         u32 de_off = le32_to_cpu(hdr->de_off);
2890         u32 o = offsetof(struct INDEX_BUFFER, ihdr) + de_off;
2891         const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
2892         u32 used = le32_to_cpu(hdr->used);
2893
2894         while (o < attr_off) {
2895                 u16 esize;
2896
2897                 if (de_off >= used)
2898                         break;
2899
2900                 esize = le16_to_cpu(e->size);
2901                 if (!esize)
2902                         break;
2903
2904                 o += esize;
2905                 de_off += esize;
2906                 e = Add2Ptr(e, esize);
2907         }
2908
2909         return o == attr_off;
2910 }
2911
2912 static inline void change_attr_size(struct MFT_REC *rec, struct ATTRIB *attr,
2913                                     u32 nsize)
2914 {
2915         u32 asize = le32_to_cpu(attr->size);
2916         int dsize = nsize - asize;
2917         u8 *next = Add2Ptr(attr, asize);
2918         u32 used = le32_to_cpu(rec->used);
2919
2920         memmove(Add2Ptr(attr, nsize), next, used - PtrOffset(rec, next));
2921
2922         rec->used = cpu_to_le32(used + dsize);
2923         attr->size = cpu_to_le32(nsize);
2924 }
2925
2926 struct OpenAttr {
2927         struct ATTRIB *attr;
2928         struct runs_tree *run1;
2929         struct runs_tree run0;
2930         struct ntfs_inode *ni;
2931         // CLST rno;
2932 };
2933
2934 /*
2935  * cmp_type_and_name
2936  *
2937  * Return: 0 if 'attr' has the same type and name.
2938  */
2939 static inline int cmp_type_and_name(const struct ATTRIB *a1,
2940                                     const struct ATTRIB *a2)
2941 {
2942         return a1->type != a2->type || a1->name_len != a2->name_len ||
2943                (a1->name_len && memcmp(attr_name(a1), attr_name(a2),
2944                                        a1->name_len * sizeof(short)));
2945 }
2946
2947 static struct OpenAttr *find_loaded_attr(struct ntfs_log *log,
2948                                          const struct ATTRIB *attr, CLST rno)
2949 {
2950         struct OPEN_ATTR_ENRTY *oe = NULL;
2951
2952         while ((oe = enum_rstbl(log->open_attr_tbl, oe))) {
2953                 struct OpenAttr *op_attr;
2954
2955                 if (ino_get(&oe->ref) != rno)
2956                         continue;
2957
2958                 op_attr = (struct OpenAttr *)oe->ptr;
2959                 if (!cmp_type_and_name(op_attr->attr, attr))
2960                         return op_attr;
2961         }
2962         return NULL;
2963 }
2964
2965 static struct ATTRIB *attr_create_nonres_log(struct ntfs_sb_info *sbi,
2966                                              enum ATTR_TYPE type, u64 size,
2967                                              const u16 *name, size_t name_len,
2968                                              __le16 flags)
2969 {
2970         struct ATTRIB *attr;
2971         u32 name_size = ALIGN(name_len * sizeof(short), 8);
2972         bool is_ext = flags & (ATTR_FLAG_COMPRESSED | ATTR_FLAG_SPARSED);
2973         u32 asize = name_size +
2974                     (is_ext ? SIZEOF_NONRESIDENT_EX : SIZEOF_NONRESIDENT);
2975
2976         attr = kzalloc(asize, GFP_NOFS);
2977         if (!attr)
2978                 return NULL;
2979
2980         attr->type = type;
2981         attr->size = cpu_to_le32(asize);
2982         attr->flags = flags;
2983         attr->non_res = 1;
2984         attr->name_len = name_len;
2985
2986         attr->nres.evcn = cpu_to_le64((u64)bytes_to_cluster(sbi, size) - 1);
2987         attr->nres.alloc_size = cpu_to_le64(ntfs_up_cluster(sbi, size));
2988         attr->nres.data_size = cpu_to_le64(size);
2989         attr->nres.valid_size = attr->nres.data_size;
2990         if (is_ext) {
2991                 attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
2992                 if (is_attr_compressed(attr))
2993                         attr->nres.c_unit = COMPRESSION_UNIT;
2994
2995                 attr->nres.run_off =
2996                         cpu_to_le16(SIZEOF_NONRESIDENT_EX + name_size);
2997                 memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT_EX), name,
2998                        name_len * sizeof(short));
2999         } else {
3000                 attr->name_off = SIZEOF_NONRESIDENT_LE;
3001                 attr->nres.run_off =
3002                         cpu_to_le16(SIZEOF_NONRESIDENT + name_size);
3003                 memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT), name,
3004                        name_len * sizeof(short));
3005         }
3006
3007         return attr;
3008 }
3009
3010 /*
3011  * do_action - Common routine for the Redo and Undo Passes.
3012  * @rlsn: If it is NULL then undo.
3013  */
3014 static int do_action(struct ntfs_log *log, struct OPEN_ATTR_ENRTY *oe,
3015                      const struct LOG_REC_HDR *lrh, u32 op, void *data,
3016                      u32 dlen, u32 rec_len, const u64 *rlsn)
3017 {
3018         int err = 0;
3019         struct ntfs_sb_info *sbi = log->ni->mi.sbi;
3020         struct inode *inode = NULL, *inode_parent;
3021         struct mft_inode *mi = NULL, *mi2_child = NULL;
3022         CLST rno = 0, rno_base = 0;
3023         struct INDEX_BUFFER *ib = NULL;
3024         struct MFT_REC *rec = NULL;
3025         struct ATTRIB *attr = NULL, *attr2;
3026         struct INDEX_HDR *hdr;
3027         struct INDEX_ROOT *root;
3028         struct NTFS_DE *e, *e1, *e2;
3029         struct NEW_ATTRIBUTE_SIZES *new_sz;
3030         struct ATTR_FILE_NAME *fname;
3031         struct OpenAttr *oa, *oa2;
3032         u32 nsize, t32, asize, used, esize, off, bits;
3033         u16 id, id2;
3034         u32 record_size = sbi->record_size;
3035         u64 t64;
3036         u16 roff = le16_to_cpu(lrh->record_off);
3037         u16 aoff = le16_to_cpu(lrh->attr_off);
3038         u64 lco = 0;
3039         u64 cbo = (u64)le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
3040         u64 tvo = le64_to_cpu(lrh->target_vcn) << sbi->cluster_bits;
3041         u64 vbo = cbo + tvo;
3042         void *buffer_le = NULL;
3043         u32 bytes = 0;
3044         bool a_dirty = false;
3045         u16 data_off;
3046
3047         oa = oe->ptr;
3048
3049         /* Big switch to prepare. */
3050         switch (op) {
3051         /* ============================================================
3052          * Process MFT records, as described by the current log record.
3053          * ============================================================
3054          */
3055         case InitializeFileRecordSegment:
3056         case DeallocateFileRecordSegment:
3057         case WriteEndOfFileRecordSegment:
3058         case CreateAttribute:
3059         case DeleteAttribute:
3060         case UpdateResidentValue:
3061         case UpdateMappingPairs:
3062         case SetNewAttributeSizes:
3063         case AddIndexEntryRoot:
3064         case DeleteIndexEntryRoot:
3065         case SetIndexEntryVcnRoot:
3066         case UpdateFileNameRoot:
3067         case UpdateRecordDataRoot:
3068         case ZeroEndOfFileRecord:
3069                 rno = vbo >> sbi->record_bits;
3070                 inode = ilookup(sbi->sb, rno);
3071                 if (inode) {
3072                         mi = &ntfs_i(inode)->mi;
3073                 } else if (op == InitializeFileRecordSegment) {
3074                         mi = kzalloc(sizeof(struct mft_inode), GFP_NOFS);
3075                         if (!mi)
3076                                 return -ENOMEM;
3077                         err = mi_format_new(mi, sbi, rno, 0, false);
3078                         if (err)
3079                                 goto out;
3080                 } else {
3081                         /* Read from disk. */
3082                         err = mi_get(sbi, rno, &mi);
3083                         if (err)
3084                                 return err;
3085                 }
3086                 rec = mi->mrec;
3087
3088                 if (op == DeallocateFileRecordSegment)
3089                         goto skip_load_parent;
3090
3091                 if (InitializeFileRecordSegment != op) {
3092                         if (rec->rhdr.sign == NTFS_BAAD_SIGNATURE)
3093                                 goto dirty_vol;
3094                         if (!check_lsn(&rec->rhdr, rlsn))
3095                                 goto out;
3096                         if (!check_file_record(rec, NULL, sbi))
3097                                 goto dirty_vol;
3098                         attr = Add2Ptr(rec, roff);
3099                 }
3100
3101                 if (is_rec_base(rec) || InitializeFileRecordSegment == op) {
3102                         rno_base = rno;
3103                         goto skip_load_parent;
3104                 }
3105
3106                 rno_base = ino_get(&rec->parent_ref);
3107                 inode_parent = ntfs_iget5(sbi->sb, &rec->parent_ref, NULL);
3108                 if (IS_ERR(inode_parent))
3109                         goto skip_load_parent;
3110
3111                 if (is_bad_inode(inode_parent)) {
3112                         iput(inode_parent);
3113                         goto skip_load_parent;
3114                 }
3115
3116                 if (ni_load_mi_ex(ntfs_i(inode_parent), rno, &mi2_child)) {
3117                         iput(inode_parent);
3118                 } else {
3119                         if (mi2_child->mrec != mi->mrec)
3120                                 memcpy(mi2_child->mrec, mi->mrec,
3121                                        sbi->record_size);
3122
3123                         if (inode)
3124                                 iput(inode);
3125                         else if (mi)
3126                                 mi_put(mi);
3127
3128                         inode = inode_parent;
3129                         mi = mi2_child;
3130                         rec = mi2_child->mrec;
3131                         attr = Add2Ptr(rec, roff);
3132                 }
3133
3134 skip_load_parent:
3135                 inode_parent = NULL;
3136                 break;
3137
3138         /*
3139          * Process attributes, as described by the current log record.
3140          */
3141         case UpdateNonresidentValue:
3142         case AddIndexEntryAllocation:
3143         case DeleteIndexEntryAllocation:
3144         case WriteEndOfIndexBuffer:
3145         case SetIndexEntryVcnAllocation:
3146         case UpdateFileNameAllocation:
3147         case SetBitsInNonresidentBitMap:
3148         case ClearBitsInNonresidentBitMap:
3149         case UpdateRecordDataAllocation:
3150                 attr = oa->attr;
3151                 bytes = UpdateNonresidentValue == op ? dlen : 0;
3152                 lco = (u64)le16_to_cpu(lrh->lcns_follow) << sbi->cluster_bits;
3153
3154                 if (attr->type == ATTR_ALLOC) {
3155                         t32 = le32_to_cpu(oe->bytes_per_index);
3156                         if (bytes < t32)
3157                                 bytes = t32;
3158                 }
3159
3160                 if (!bytes)
3161                         bytes = lco - cbo;
3162
3163                 bytes += roff;
3164                 if (attr->type == ATTR_ALLOC)
3165                         bytes = (bytes + 511) & ~511; // align
3166
3167                 buffer_le = kmalloc(bytes, GFP_NOFS);
3168                 if (!buffer_le)
3169                         return -ENOMEM;
3170
3171                 err = ntfs_read_run_nb(sbi, oa->run1, vbo, buffer_le, bytes,
3172                                        NULL);
3173                 if (err)
3174                         goto out;
3175
3176                 if (attr->type == ATTR_ALLOC && *(int *)buffer_le)
3177                         ntfs_fix_post_read(buffer_le, bytes, false);
3178                 break;
3179
3180         default:
3181                 WARN_ON(1);
3182         }
3183
3184         /* Big switch to do operation. */
3185         switch (op) {
3186         case InitializeFileRecordSegment:
3187                 if (roff + dlen > record_size)
3188                         goto dirty_vol;
3189
3190                 memcpy(Add2Ptr(rec, roff), data, dlen);
3191                 mi->dirty = true;
3192                 break;
3193
3194         case DeallocateFileRecordSegment:
3195                 clear_rec_inuse(rec);
3196                 le16_add_cpu(&rec->seq, 1);
3197                 mi->dirty = true;
3198                 break;
3199
3200         case WriteEndOfFileRecordSegment:
3201                 attr2 = (struct ATTRIB *)data;
3202                 if (!check_if_attr(rec, lrh) || roff + dlen > record_size)
3203                         goto dirty_vol;
3204
3205                 memmove(attr, attr2, dlen);
3206                 rec->used = cpu_to_le32(ALIGN(roff + dlen, 8));
3207
3208                 mi->dirty = true;
3209                 break;
3210
3211         case CreateAttribute:
3212                 attr2 = (struct ATTRIB *)data;
3213                 asize = le32_to_cpu(attr2->size);
3214                 used = le32_to_cpu(rec->used);
3215
3216                 if (!check_if_attr(rec, lrh) || dlen < SIZEOF_RESIDENT ||
3217                     !IS_ALIGNED(asize, 8) ||
3218                     Add2Ptr(attr2, asize) > Add2Ptr(lrh, rec_len) ||
3219                     dlen > record_size - used) {
3220                         goto dirty_vol;
3221                 }
3222
3223                 memmove(Add2Ptr(attr, asize), attr, used - roff);
3224                 memcpy(attr, attr2, asize);
3225
3226                 rec->used = cpu_to_le32(used + asize);
3227                 id = le16_to_cpu(rec->next_attr_id);
3228                 id2 = le16_to_cpu(attr2->id);
3229                 if (id <= id2)
3230                         rec->next_attr_id = cpu_to_le16(id2 + 1);
3231                 if (is_attr_indexed(attr))
3232                         le16_add_cpu(&rec->hard_links, 1);
3233
3234                 oa2 = find_loaded_attr(log, attr, rno_base);
3235                 if (oa2) {
3236                         void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3237                                            GFP_NOFS);
3238                         if (p2) {
3239                                 // run_close(oa2->run1);
3240                                 kfree(oa2->attr);
3241                                 oa2->attr = p2;
3242                         }
3243                 }
3244
3245                 mi->dirty = true;
3246                 break;
3247
3248         case DeleteAttribute:
3249                 asize = le32_to_cpu(attr->size);
3250                 used = le32_to_cpu(rec->used);
3251
3252                 if (!check_if_attr(rec, lrh))
3253                         goto dirty_vol;
3254
3255                 rec->used = cpu_to_le32(used - asize);
3256                 if (is_attr_indexed(attr))
3257                         le16_add_cpu(&rec->hard_links, -1);
3258
3259                 memmove(attr, Add2Ptr(attr, asize), used - asize - roff);
3260
3261                 mi->dirty = true;
3262                 break;
3263
3264         case UpdateResidentValue:
3265                 nsize = aoff + dlen;
3266
3267                 if (!check_if_attr(rec, lrh))
3268                         goto dirty_vol;
3269
3270                 asize = le32_to_cpu(attr->size);
3271                 used = le32_to_cpu(rec->used);
3272
3273                 if (lrh->redo_len == lrh->undo_len) {
3274                         if (nsize > asize)
3275                                 goto dirty_vol;
3276                         goto move_data;
3277                 }
3278
3279                 if (nsize > asize && nsize - asize > record_size - used)
3280                         goto dirty_vol;
3281
3282                 nsize = ALIGN(nsize, 8);
3283                 data_off = le16_to_cpu(attr->res.data_off);
3284
3285                 if (nsize < asize) {
3286                         memmove(Add2Ptr(attr, aoff), data, dlen);
3287                         data = NULL; // To skip below memmove().
3288                 }
3289
3290                 memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
3291                         used - le16_to_cpu(lrh->record_off) - asize);
3292
3293                 rec->used = cpu_to_le32(used + nsize - asize);
3294                 attr->size = cpu_to_le32(nsize);
3295                 attr->res.data_size = cpu_to_le32(aoff + dlen - data_off);
3296
3297 move_data:
3298                 if (data)
3299                         memmove(Add2Ptr(attr, aoff), data, dlen);
3300
3301                 oa2 = find_loaded_attr(log, attr, rno_base);
3302                 if (oa2) {
3303                         void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3304                                            GFP_NOFS);
3305                         if (p2) {
3306                                 // run_close(&oa2->run0);
3307                                 oa2->run1 = &oa2->run0;
3308                                 kfree(oa2->attr);
3309                                 oa2->attr = p2;
3310                         }
3311                 }
3312
3313                 mi->dirty = true;
3314                 break;
3315
3316         case UpdateMappingPairs:
3317                 nsize = aoff + dlen;
3318                 asize = le32_to_cpu(attr->size);
3319                 used = le32_to_cpu(rec->used);
3320
3321                 if (!check_if_attr(rec, lrh) || !attr->non_res ||
3322                     aoff < le16_to_cpu(attr->nres.run_off) || aoff > asize ||
3323                     (nsize > asize && nsize - asize > record_size - used)) {
3324                         goto dirty_vol;
3325                 }
3326
3327                 nsize = ALIGN(nsize, 8);
3328
3329                 memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
3330                         used - le16_to_cpu(lrh->record_off) - asize);
3331                 rec->used = cpu_to_le32(used + nsize - asize);
3332                 attr->size = cpu_to_le32(nsize);
3333                 memmove(Add2Ptr(attr, aoff), data, dlen);
3334
3335                 if (run_get_highest_vcn(le64_to_cpu(attr->nres.svcn),
3336                                         attr_run(attr), &t64)) {
3337                         goto dirty_vol;
3338                 }
3339
3340                 attr->nres.evcn = cpu_to_le64(t64);
3341                 oa2 = find_loaded_attr(log, attr, rno_base);
3342                 if (oa2 && oa2->attr->non_res)
3343                         oa2->attr->nres.evcn = attr->nres.evcn;
3344
3345                 mi->dirty = true;
3346                 break;
3347
3348         case SetNewAttributeSizes:
3349                 new_sz = data;
3350                 if (!check_if_attr(rec, lrh) || !attr->non_res)
3351                         goto dirty_vol;
3352
3353                 attr->nres.alloc_size = new_sz->alloc_size;
3354                 attr->nres.data_size = new_sz->data_size;
3355                 attr->nres.valid_size = new_sz->valid_size;
3356
3357                 if (dlen >= sizeof(struct NEW_ATTRIBUTE_SIZES))
3358                         attr->nres.total_size = new_sz->total_size;
3359
3360                 oa2 = find_loaded_attr(log, attr, rno_base);
3361                 if (oa2) {
3362                         void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3363                                            GFP_NOFS);
3364                         if (p2) {
3365                                 kfree(oa2->attr);
3366                                 oa2->attr = p2;
3367                         }
3368                 }
3369                 mi->dirty = true;
3370                 break;
3371
3372         case AddIndexEntryRoot:
3373                 e = (struct NTFS_DE *)data;
3374                 esize = le16_to_cpu(e->size);
3375                 root = resident_data(attr);
3376                 hdr = &root->ihdr;
3377                 used = le32_to_cpu(hdr->used);
3378
3379                 if (!check_if_index_root(rec, lrh) ||
3380                     !check_if_root_index(attr, hdr, lrh) ||
3381                     Add2Ptr(data, esize) > Add2Ptr(lrh, rec_len) ||
3382                     esize > le32_to_cpu(rec->total) - le32_to_cpu(rec->used)) {
3383                         goto dirty_vol;
3384                 }
3385
3386                 e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3387
3388                 change_attr_size(rec, attr, le32_to_cpu(attr->size) + esize);
3389
3390                 memmove(Add2Ptr(e1, esize), e1,
3391                         PtrOffset(e1, Add2Ptr(hdr, used)));
3392                 memmove(e1, e, esize);
3393
3394                 le32_add_cpu(&attr->res.data_size, esize);
3395                 hdr->used = cpu_to_le32(used + esize);
3396                 le32_add_cpu(&hdr->total, esize);
3397
3398                 mi->dirty = true;
3399                 break;
3400
3401         case DeleteIndexEntryRoot:
3402                 root = resident_data(attr);
3403                 hdr = &root->ihdr;
3404                 used = le32_to_cpu(hdr->used);
3405
3406                 if (!check_if_index_root(rec, lrh) ||
3407                     !check_if_root_index(attr, hdr, lrh)) {
3408                         goto dirty_vol;
3409                 }
3410
3411                 e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3412                 esize = le16_to_cpu(e1->size);
3413                 e2 = Add2Ptr(e1, esize);
3414
3415                 memmove(e1, e2, PtrOffset(e2, Add2Ptr(hdr, used)));
3416
3417                 le32_sub_cpu(&attr->res.data_size, esize);
3418                 hdr->used = cpu_to_le32(used - esize);
3419                 le32_sub_cpu(&hdr->total, esize);
3420
3421                 change_attr_size(rec, attr, le32_to_cpu(attr->size) - esize);
3422
3423                 mi->dirty = true;
3424                 break;
3425
3426         case SetIndexEntryVcnRoot:
3427                 root = resident_data(attr);
3428                 hdr = &root->ihdr;
3429
3430                 if (!check_if_index_root(rec, lrh) ||
3431                     !check_if_root_index(attr, hdr, lrh)) {
3432                         goto dirty_vol;
3433                 }
3434
3435                 e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3436
3437                 de_set_vbn_le(e, *(__le64 *)data);
3438                 mi->dirty = true;
3439                 break;
3440
3441         case UpdateFileNameRoot:
3442                 root = resident_data(attr);
3443                 hdr = &root->ihdr;
3444
3445                 if (!check_if_index_root(rec, lrh) ||
3446                     !check_if_root_index(attr, hdr, lrh)) {
3447                         goto dirty_vol;
3448                 }
3449
3450                 e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3451                 fname = (struct ATTR_FILE_NAME *)(e + 1);
3452                 memmove(&fname->dup, data, sizeof(fname->dup)); //
3453                 mi->dirty = true;
3454                 break;
3455
3456         case UpdateRecordDataRoot:
3457                 root = resident_data(attr);
3458                 hdr = &root->ihdr;
3459
3460                 if (!check_if_index_root(rec, lrh) ||
3461                     !check_if_root_index(attr, hdr, lrh)) {
3462                         goto dirty_vol;
3463                 }
3464
3465                 e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3466
3467                 memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
3468
3469                 mi->dirty = true;
3470                 break;
3471
3472         case ZeroEndOfFileRecord:
3473                 if (roff + dlen > record_size)
3474                         goto dirty_vol;
3475
3476                 memset(attr, 0, dlen);
3477                 mi->dirty = true;
3478                 break;
3479
3480         case UpdateNonresidentValue:
3481                 if (lco < cbo + roff + dlen)
3482                         goto dirty_vol;
3483
3484                 memcpy(Add2Ptr(buffer_le, roff), data, dlen);
3485
3486                 a_dirty = true;
3487                 if (attr->type == ATTR_ALLOC)
3488                         ntfs_fix_pre_write(buffer_le, bytes);
3489                 break;
3490
3491         case AddIndexEntryAllocation:
3492                 ib = Add2Ptr(buffer_le, roff);
3493                 hdr = &ib->ihdr;
3494                 e = data;
3495                 esize = le16_to_cpu(e->size);
3496                 e1 = Add2Ptr(ib, aoff);
3497
3498                 if (is_baad(&ib->rhdr))
3499                         goto dirty_vol;
3500                 if (!check_lsn(&ib->rhdr, rlsn))
3501                         goto out;
3502
3503                 used = le32_to_cpu(hdr->used);
3504
3505                 if (!check_index_buffer(ib, bytes) ||
3506                     !check_if_alloc_index(hdr, aoff) ||
3507                     Add2Ptr(e, esize) > Add2Ptr(lrh, rec_len) ||
3508                     used + esize > le32_to_cpu(hdr->total)) {
3509                         goto dirty_vol;
3510                 }
3511
3512                 memmove(Add2Ptr(e1, esize), e1,
3513                         PtrOffset(e1, Add2Ptr(hdr, used)));
3514                 memcpy(e1, e, esize);
3515
3516                 hdr->used = cpu_to_le32(used + esize);
3517
3518                 a_dirty = true;
3519
3520                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3521                 break;
3522
3523         case DeleteIndexEntryAllocation:
3524                 ib = Add2Ptr(buffer_le, roff);
3525                 hdr = &ib->ihdr;
3526                 e = Add2Ptr(ib, aoff);
3527                 esize = le16_to_cpu(e->size);
3528
3529                 if (is_baad(&ib->rhdr))
3530                         goto dirty_vol;
3531                 if (!check_lsn(&ib->rhdr, rlsn))
3532                         goto out;
3533
3534                 if (!check_index_buffer(ib, bytes) ||
3535                     !check_if_alloc_index(hdr, aoff)) {
3536                         goto dirty_vol;
3537                 }
3538
3539                 e1 = Add2Ptr(e, esize);
3540                 nsize = esize;
3541                 used = le32_to_cpu(hdr->used);
3542
3543                 memmove(e, e1, PtrOffset(e1, Add2Ptr(hdr, used)));
3544
3545                 hdr->used = cpu_to_le32(used - nsize);
3546
3547                 a_dirty = true;
3548
3549                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3550                 break;
3551
3552         case WriteEndOfIndexBuffer:
3553                 ib = Add2Ptr(buffer_le, roff);
3554                 hdr = &ib->ihdr;
3555                 e = Add2Ptr(ib, aoff);
3556
3557                 if (is_baad(&ib->rhdr))
3558                         goto dirty_vol;
3559                 if (!check_lsn(&ib->rhdr, rlsn))
3560                         goto out;
3561                 if (!check_index_buffer(ib, bytes) ||
3562                     !check_if_alloc_index(hdr, aoff) ||
3563                     aoff + dlen > offsetof(struct INDEX_BUFFER, ihdr) +
3564                                           le32_to_cpu(hdr->total)) {
3565                         goto dirty_vol;
3566                 }
3567
3568                 hdr->used = cpu_to_le32(dlen + PtrOffset(hdr, e));
3569                 memmove(e, data, dlen);
3570
3571                 a_dirty = true;
3572                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3573                 break;
3574
3575         case SetIndexEntryVcnAllocation:
3576                 ib = Add2Ptr(buffer_le, roff);
3577                 hdr = &ib->ihdr;
3578                 e = Add2Ptr(ib, aoff);
3579
3580                 if (is_baad(&ib->rhdr))
3581                         goto dirty_vol;
3582
3583                 if (!check_lsn(&ib->rhdr, rlsn))
3584                         goto out;
3585                 if (!check_index_buffer(ib, bytes) ||
3586                     !check_if_alloc_index(hdr, aoff)) {
3587                         goto dirty_vol;
3588                 }
3589
3590                 de_set_vbn_le(e, *(__le64 *)data);
3591
3592                 a_dirty = true;
3593                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3594                 break;
3595
3596         case UpdateFileNameAllocation:
3597                 ib = Add2Ptr(buffer_le, roff);
3598                 hdr = &ib->ihdr;
3599                 e = Add2Ptr(ib, aoff);
3600
3601                 if (is_baad(&ib->rhdr))
3602                         goto dirty_vol;
3603
3604                 if (!check_lsn(&ib->rhdr, rlsn))
3605                         goto out;
3606                 if (!check_index_buffer(ib, bytes) ||
3607                     !check_if_alloc_index(hdr, aoff)) {
3608                         goto dirty_vol;
3609                 }
3610
3611                 fname = (struct ATTR_FILE_NAME *)(e + 1);
3612                 memmove(&fname->dup, data, sizeof(fname->dup));
3613
3614                 a_dirty = true;
3615                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3616                 break;
3617
3618         case SetBitsInNonresidentBitMap:
3619                 off = le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
3620                 bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
3621
3622                 if (cbo + (off + 7) / 8 > lco ||
3623                     cbo + ((off + bits + 7) / 8) > lco) {
3624                         goto dirty_vol;
3625                 }
3626
3627                 ntfs_bitmap_set_le(Add2Ptr(buffer_le, roff), off, bits);
3628                 a_dirty = true;
3629                 break;
3630
3631         case ClearBitsInNonresidentBitMap:
3632                 off = le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
3633                 bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
3634
3635                 if (cbo + (off + 7) / 8 > lco ||
3636                     cbo + ((off + bits + 7) / 8) > lco) {
3637                         goto dirty_vol;
3638                 }
3639
3640                 ntfs_bitmap_clear_le(Add2Ptr(buffer_le, roff), off, bits);
3641                 a_dirty = true;
3642                 break;
3643
3644         case UpdateRecordDataAllocation:
3645                 ib = Add2Ptr(buffer_le, roff);
3646                 hdr = &ib->ihdr;
3647                 e = Add2Ptr(ib, aoff);
3648
3649                 if (is_baad(&ib->rhdr))
3650                         goto dirty_vol;
3651
3652                 if (!check_lsn(&ib->rhdr, rlsn))
3653                         goto out;
3654                 if (!check_index_buffer(ib, bytes) ||
3655                     !check_if_alloc_index(hdr, aoff)) {
3656                         goto dirty_vol;
3657                 }
3658
3659                 memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
3660
3661                 a_dirty = true;
3662                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3663                 break;
3664
3665         default:
3666                 WARN_ON(1);
3667         }
3668
3669         if (rlsn) {
3670                 __le64 t64 = cpu_to_le64(*rlsn);
3671
3672                 if (rec)
3673                         rec->rhdr.lsn = t64;
3674                 if (ib)
3675                         ib->rhdr.lsn = t64;
3676         }
3677
3678         if (mi && mi->dirty) {
3679                 err = mi_write(mi, 0);
3680                 if (err)
3681                         goto out;
3682         }
3683
3684         if (a_dirty) {
3685                 attr = oa->attr;
3686                 err = ntfs_sb_write_run(sbi, oa->run1, vbo, buffer_le, bytes,
3687                                         0);
3688                 if (err)
3689                         goto out;
3690         }
3691
3692 out:
3693
3694         if (inode)
3695                 iput(inode);
3696         else if (mi != mi2_child)
3697                 mi_put(mi);
3698
3699         kfree(buffer_le);
3700
3701         return err;
3702
3703 dirty_vol:
3704         log->set_dirty = true;
3705         goto out;
3706 }
3707
3708 /*
3709  * log_replay - Replays log and empties it.
3710  *
3711  * This function is called during mount operation.
3712  * It replays log and empties it.
3713  * Initialized is set false if logfile contains '-1'.
3714  */
3715 int log_replay(struct ntfs_inode *ni, bool *initialized)
3716 {
3717         int err;
3718         struct ntfs_sb_info *sbi = ni->mi.sbi;
3719         struct ntfs_log *log;
3720
3721         struct restart_info rst_info, rst_info2;
3722         u64 rec_lsn, ra_lsn, checkpt_lsn = 0, rlsn = 0;
3723         struct ATTR_NAME_ENTRY *attr_names = NULL;
3724         struct ATTR_NAME_ENTRY *ane;
3725         struct RESTART_TABLE *dptbl = NULL;
3726         struct RESTART_TABLE *trtbl = NULL;
3727         const struct RESTART_TABLE *rt;
3728         struct RESTART_TABLE *oatbl = NULL;
3729         struct inode *inode;
3730         struct OpenAttr *oa;
3731         struct ntfs_inode *ni_oe;
3732         struct ATTRIB *attr = NULL;
3733         u64 size, vcn, undo_next_lsn;
3734         CLST rno, lcn, lcn0, len0, clen;
3735         void *data;
3736         struct NTFS_RESTART *rst = NULL;
3737         struct lcb *lcb = NULL;
3738         struct OPEN_ATTR_ENRTY *oe;
3739         struct TRANSACTION_ENTRY *tr;
3740         struct DIR_PAGE_ENTRY *dp;
3741         u32 i, bytes_per_attr_entry;
3742         u32 l_size = ni->vfs_inode.i_size;
3743         u32 orig_file_size = l_size;
3744         u32 page_size, vbo, tail, off, dlen;
3745         u32 saved_len, rec_len, transact_id;
3746         bool use_second_page;
3747         struct RESTART_AREA *ra2, *ra = NULL;
3748         struct CLIENT_REC *ca, *cr;
3749         __le16 client;
3750         struct RESTART_HDR *rh;
3751         const struct LFS_RECORD_HDR *frh;
3752         const struct LOG_REC_HDR *lrh;
3753         bool is_mapped;
3754         bool is_ro = sb_rdonly(sbi->sb);
3755         u64 t64;
3756         u16 t16;
3757         u32 t32;
3758
3759         /* Get the size of page. NOTE: To replay we can use default page. */
3760 #if PAGE_SIZE >= DefaultLogPageSize && PAGE_SIZE <= DefaultLogPageSize * 2
3761         page_size = norm_file_page(PAGE_SIZE, &l_size, true);
3762 #else
3763         page_size = norm_file_page(PAGE_SIZE, &l_size, false);
3764 #endif
3765         if (!page_size)
3766                 return -EINVAL;
3767
3768         log = kzalloc(sizeof(struct ntfs_log), GFP_NOFS);
3769         if (!log)
3770                 return -ENOMEM;
3771
3772         log->ni = ni;
3773         log->l_size = l_size;
3774         log->one_page_buf = kmalloc(page_size, GFP_NOFS);
3775
3776         if (!log->one_page_buf) {
3777                 err = -ENOMEM;
3778                 goto out;
3779         }
3780
3781         log->page_size = page_size;
3782         log->page_mask = page_size - 1;
3783         log->page_bits = blksize_bits(page_size);
3784
3785         /* Look for a restart area on the disk. */
3786         memset(&rst_info, 0, sizeof(struct restart_info));
3787         err = log_read_rst(log, l_size, true, &rst_info);
3788         if (err)
3789                 goto out;
3790
3791         /* remember 'initialized' */
3792         *initialized = rst_info.initialized;
3793
3794         if (!rst_info.restart) {
3795                 if (rst_info.initialized) {
3796                         /* No restart area but the file is not initialized. */
3797                         err = -EINVAL;
3798                         goto out;
3799                 }
3800
3801                 log_init_pg_hdr(log, page_size, page_size, 1, 1);
3802                 log_create(log, l_size, 0, get_random_u32(), false, false);
3803
3804                 log->ra = ra;
3805
3806                 ra = log_create_ra(log);
3807                 if (!ra) {
3808                         err = -ENOMEM;
3809                         goto out;
3810                 }
3811                 log->ra = ra;
3812                 log->init_ra = true;
3813
3814                 goto process_log;
3815         }
3816
3817         /*
3818          * If the restart offset above wasn't zero then we won't
3819          * look for a second restart.
3820          */
3821         if (rst_info.vbo)
3822                 goto check_restart_area;
3823
3824         memset(&rst_info2, 0, sizeof(struct restart_info));
3825         err = log_read_rst(log, l_size, false, &rst_info2);
3826         if (err)
3827                 goto out;
3828
3829         /* Determine which restart area to use. */
3830         if (!rst_info2.restart || rst_info2.last_lsn <= rst_info.last_lsn)
3831                 goto use_first_page;
3832
3833         use_second_page = true;
3834
3835         if (rst_info.chkdsk_was_run && page_size != rst_info.vbo) {
3836                 struct RECORD_PAGE_HDR *sp = NULL;
3837                 bool usa_error;
3838
3839                 if (!read_log_page(log, page_size, &sp, &usa_error) &&
3840                     sp->rhdr.sign == NTFS_CHKD_SIGNATURE) {
3841                         use_second_page = false;
3842                 }
3843                 kfree(sp);
3844         }
3845
3846         if (use_second_page) {
3847                 kfree(rst_info.r_page);
3848                 memcpy(&rst_info, &rst_info2, sizeof(struct restart_info));
3849                 rst_info2.r_page = NULL;
3850         }
3851
3852 use_first_page:
3853         kfree(rst_info2.r_page);
3854
3855 check_restart_area:
3856         /*
3857          * If the restart area is at offset 0, we want
3858          * to write the second restart area first.
3859          */
3860         log->init_ra = !!rst_info.vbo;
3861
3862         /* If we have a valid page then grab a pointer to the restart area. */
3863         ra2 = rst_info.valid_page ?
3864                       Add2Ptr(rst_info.r_page,
3865                               le16_to_cpu(rst_info.r_page->ra_off)) :
3866                       NULL;
3867
3868         if (rst_info.chkdsk_was_run ||
3869             (ra2 && ra2->client_idx[1] == LFS_NO_CLIENT_LE)) {
3870                 bool wrapped = false;
3871                 bool use_multi_page = false;
3872                 u32 open_log_count;
3873
3874                 /* Do some checks based on whether we have a valid log page. */
3875                 if (!rst_info.valid_page) {
3876                         open_log_count = get_random_u32();
3877                         goto init_log_instance;
3878                 }
3879                 open_log_count = le32_to_cpu(ra2->open_log_count);
3880
3881                 /*
3882                  * If the restart page size isn't changing then we want to
3883                  * check how much work we need to do.
3884                  */
3885                 if (page_size != le32_to_cpu(rst_info.r_page->sys_page_size))
3886                         goto init_log_instance;
3887
3888 init_log_instance:
3889                 log_init_pg_hdr(log, page_size, page_size, 1, 1);
3890
3891                 log_create(log, l_size, rst_info.last_lsn, open_log_count,
3892                            wrapped, use_multi_page);
3893
3894                 ra = log_create_ra(log);
3895                 if (!ra) {
3896                         err = -ENOMEM;
3897                         goto out;
3898                 }
3899                 log->ra = ra;
3900
3901                 /* Put the restart areas and initialize
3902                  * the log file as required.
3903                  */
3904                 goto process_log;
3905         }
3906
3907         if (!ra2) {
3908                 err = -EINVAL;
3909                 goto out;
3910         }
3911
3912         /*
3913          * If the log page or the system page sizes have changed, we can't
3914          * use the log file. We must use the system page size instead of the
3915          * default size if there is not a clean shutdown.
3916          */
3917         t32 = le32_to_cpu(rst_info.r_page->sys_page_size);
3918         if (page_size != t32) {
3919                 l_size = orig_file_size;
3920                 page_size =
3921                         norm_file_page(t32, &l_size, t32 == DefaultLogPageSize);
3922         }
3923
3924         if (page_size != t32 ||
3925             page_size != le32_to_cpu(rst_info.r_page->page_size)) {
3926                 err = -EINVAL;
3927                 goto out;
3928         }
3929
3930         /* If the file size has shrunk then we won't mount it. */
3931         if (l_size < le64_to_cpu(ra2->l_size)) {
3932                 err = -EINVAL;
3933                 goto out;
3934         }
3935
3936         log_init_pg_hdr(log, page_size, page_size,
3937                         le16_to_cpu(rst_info.r_page->major_ver),
3938                         le16_to_cpu(rst_info.r_page->minor_ver));
3939
3940         log->l_size = le64_to_cpu(ra2->l_size);
3941         log->seq_num_bits = le32_to_cpu(ra2->seq_num_bits);
3942         log->file_data_bits = sizeof(u64) * 8 - log->seq_num_bits;
3943         log->seq_num_mask = (8 << log->file_data_bits) - 1;
3944         log->last_lsn = le64_to_cpu(ra2->current_lsn);
3945         log->seq_num = log->last_lsn >> log->file_data_bits;
3946         log->ra_off = le16_to_cpu(rst_info.r_page->ra_off);
3947         log->restart_size = log->sys_page_size - log->ra_off;
3948         log->record_header_len = le16_to_cpu(ra2->rec_hdr_len);
3949         log->ra_size = le16_to_cpu(ra2->ra_len);
3950         log->data_off = le16_to_cpu(ra2->data_off);
3951         log->data_size = log->page_size - log->data_off;
3952         log->reserved = log->data_size - log->record_header_len;
3953
3954         vbo = lsn_to_vbo(log, log->last_lsn);
3955
3956         if (vbo < log->first_page) {
3957                 /* This is a pseudo lsn. */
3958                 log->l_flags |= NTFSLOG_NO_LAST_LSN;
3959                 log->next_page = log->first_page;
3960                 goto find_oldest;
3961         }
3962
3963         /* Find the end of this log record. */
3964         off = final_log_off(log, log->last_lsn,
3965                             le32_to_cpu(ra2->last_lsn_data_len));
3966
3967         /* If we wrapped the file then increment the sequence number. */
3968         if (off <= vbo) {
3969                 log->seq_num += 1;
3970                 log->l_flags |= NTFSLOG_WRAPPED;
3971         }
3972
3973         /* Now compute the next log page to use. */
3974         vbo &= ~log->sys_page_mask;
3975         tail = log->page_size - (off & log->page_mask) - 1;
3976
3977         /*
3978          *If we can fit another log record on the page,
3979          * move back a page the log file.
3980          */
3981         if (tail >= log->record_header_len) {
3982                 log->l_flags |= NTFSLOG_REUSE_TAIL;
3983                 log->next_page = vbo;
3984         } else {
3985                 log->next_page = next_page_off(log, vbo);
3986         }
3987
3988 find_oldest:
3989         /*
3990          * Find the oldest client lsn. Use the last
3991          * flushed lsn as a starting point.
3992          */
3993         log->oldest_lsn = log->last_lsn;
3994         oldest_client_lsn(Add2Ptr(ra2, le16_to_cpu(ra2->client_off)),
3995                           ra2->client_idx[1], &log->oldest_lsn);
3996         log->oldest_lsn_off = lsn_to_vbo(log, log->oldest_lsn);
3997
3998         if (log->oldest_lsn_off < log->first_page)
3999                 log->l_flags |= NTFSLOG_NO_OLDEST_LSN;
4000
4001         if (!(ra2->flags & RESTART_SINGLE_PAGE_IO))
4002                 log->l_flags |= NTFSLOG_WRAPPED | NTFSLOG_MULTIPLE_PAGE_IO;
4003
4004         log->current_openlog_count = le32_to_cpu(ra2->open_log_count);
4005         log->total_avail_pages = log->l_size - log->first_page;
4006         log->total_avail = log->total_avail_pages >> log->page_bits;
4007         log->max_current_avail = log->total_avail * log->reserved;
4008         log->total_avail = log->total_avail * log->data_size;
4009
4010         log->current_avail = current_log_avail(log);
4011
4012         ra = kzalloc(log->restart_size, GFP_NOFS);
4013         if (!ra) {
4014                 err = -ENOMEM;
4015                 goto out;
4016         }
4017         log->ra = ra;
4018
4019         t16 = le16_to_cpu(ra2->client_off);
4020         if (t16 == offsetof(struct RESTART_AREA, clients)) {
4021                 memcpy(ra, ra2, log->ra_size);
4022         } else {
4023                 memcpy(ra, ra2, offsetof(struct RESTART_AREA, clients));
4024                 memcpy(ra->clients, Add2Ptr(ra2, t16),
4025                        le16_to_cpu(ra2->ra_len) - t16);
4026
4027                 log->current_openlog_count = get_random_u32();
4028                 ra->open_log_count = cpu_to_le32(log->current_openlog_count);
4029                 log->ra_size = offsetof(struct RESTART_AREA, clients) +
4030                                sizeof(struct CLIENT_REC);
4031                 ra->client_off =
4032                         cpu_to_le16(offsetof(struct RESTART_AREA, clients));
4033                 ra->ra_len = cpu_to_le16(log->ra_size);
4034         }
4035
4036         le32_add_cpu(&ra->open_log_count, 1);
4037
4038         /* Now we need to walk through looking for the last lsn. */
4039         err = last_log_lsn(log);
4040         if (err)
4041                 goto out;
4042
4043         log->current_avail = current_log_avail(log);
4044
4045         /* Remember which restart area to write first. */
4046         log->init_ra = rst_info.vbo;
4047
4048 process_log:
4049         /* 1.0, 1.1, 2.0 log->major_ver/minor_ver - short values. */
4050         switch ((log->major_ver << 16) + log->minor_ver) {
4051         case 0x10000:
4052         case 0x10001:
4053         case 0x20000:
4054                 break;
4055         default:
4056                 ntfs_warn(sbi->sb, "\x24LogFile version %d.%d is not supported",
4057                           log->major_ver, log->minor_ver);
4058                 err = -EOPNOTSUPP;
4059                 log->set_dirty = true;
4060                 goto out;
4061         }
4062
4063         /* One client "NTFS" per logfile. */
4064         ca = Add2Ptr(ra, le16_to_cpu(ra->client_off));
4065
4066         for (client = ra->client_idx[1];; client = cr->next_client) {
4067                 if (client == LFS_NO_CLIENT_LE) {
4068                         /* Insert "NTFS" client LogFile. */
4069                         client = ra->client_idx[0];
4070                         if (client == LFS_NO_CLIENT_LE) {
4071                                 err = -EINVAL;
4072                                 goto out;
4073                         }
4074
4075                         t16 = le16_to_cpu(client);
4076                         cr = ca + t16;
4077
4078                         remove_client(ca, cr, &ra->client_idx[0]);
4079
4080                         cr->restart_lsn = 0;
4081                         cr->oldest_lsn = cpu_to_le64(log->oldest_lsn);
4082                         cr->name_bytes = cpu_to_le32(8);
4083                         cr->name[0] = cpu_to_le16('N');
4084                         cr->name[1] = cpu_to_le16('T');
4085                         cr->name[2] = cpu_to_le16('F');
4086                         cr->name[3] = cpu_to_le16('S');
4087
4088                         add_client(ca, t16, &ra->client_idx[1]);
4089                         break;
4090                 }
4091
4092                 cr = ca + le16_to_cpu(client);
4093
4094                 if (cpu_to_le32(8) == cr->name_bytes &&
4095                     cpu_to_le16('N') == cr->name[0] &&
4096                     cpu_to_le16('T') == cr->name[1] &&
4097                     cpu_to_le16('F') == cr->name[2] &&
4098                     cpu_to_le16('S') == cr->name[3])
4099                         break;
4100         }
4101
4102         /* Update the client handle with the client block information. */
4103         log->client_id.seq_num = cr->seq_num;
4104         log->client_id.client_idx = client;
4105
4106         err = read_rst_area(log, &rst, &ra_lsn);
4107         if (err)
4108                 goto out;
4109
4110         if (!rst)
4111                 goto out;
4112
4113         bytes_per_attr_entry = !rst->major_ver ? 0x2C : 0x28;
4114
4115         checkpt_lsn = le64_to_cpu(rst->check_point_start);
4116         if (!checkpt_lsn)
4117                 checkpt_lsn = ra_lsn;
4118
4119         /* Allocate and Read the Transaction Table. */
4120         if (!rst->transact_table_len)
4121                 goto check_dirty_page_table;
4122
4123         t64 = le64_to_cpu(rst->transact_table_lsn);
4124         err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4125         if (err)
4126                 goto out;
4127
4128         lrh = lcb->log_rec;
4129         frh = lcb->lrh;
4130         rec_len = le32_to_cpu(frh->client_data_len);
4131
4132         if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4133                            bytes_per_attr_entry)) {
4134                 err = -EINVAL;
4135                 goto out;
4136         }
4137
4138         t16 = le16_to_cpu(lrh->redo_off);
4139
4140         rt = Add2Ptr(lrh, t16);
4141         t32 = rec_len - t16;
4142
4143         /* Now check that this is a valid restart table. */
4144         if (!check_rstbl(rt, t32)) {
4145                 err = -EINVAL;
4146                 goto out;
4147         }
4148
4149         trtbl = kmemdup(rt, t32, GFP_NOFS);
4150         if (!trtbl) {
4151                 err = -ENOMEM;
4152                 goto out;
4153         }
4154
4155         lcb_put(lcb);
4156         lcb = NULL;
4157
4158 check_dirty_page_table:
4159         /* The next record back should be the Dirty Pages Table. */
4160         if (!rst->dirty_pages_len)
4161                 goto check_attribute_names;
4162
4163         t64 = le64_to_cpu(rst->dirty_pages_table_lsn);
4164         err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4165         if (err)
4166                 goto out;
4167
4168         lrh = lcb->log_rec;
4169         frh = lcb->lrh;
4170         rec_len = le32_to_cpu(frh->client_data_len);
4171
4172         if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4173                            bytes_per_attr_entry)) {
4174                 err = -EINVAL;
4175                 goto out;
4176         }
4177
4178         t16 = le16_to_cpu(lrh->redo_off);
4179
4180         rt = Add2Ptr(lrh, t16);
4181         t32 = rec_len - t16;
4182
4183         /* Now check that this is a valid restart table. */
4184         if (!check_rstbl(rt, t32)) {
4185                 err = -EINVAL;
4186                 goto out;
4187         }
4188
4189         dptbl = kmemdup(rt, t32, GFP_NOFS);
4190         if (!dptbl) {
4191                 err = -ENOMEM;
4192                 goto out;
4193         }
4194
4195         /* Convert Ra version '0' into version '1'. */
4196         if (rst->major_ver)
4197                 goto end_conv_1;
4198
4199         dp = NULL;
4200         while ((dp = enum_rstbl(dptbl, dp))) {
4201                 struct DIR_PAGE_ENTRY_32 *dp0 = (struct DIR_PAGE_ENTRY_32 *)dp;
4202                 // NOTE: Danger. Check for of boundary.
4203                 memmove(&dp->vcn, &dp0->vcn_low,
4204                         2 * sizeof(u64) +
4205                                 le32_to_cpu(dp->lcns_follow) * sizeof(u64));
4206         }
4207
4208 end_conv_1:
4209         lcb_put(lcb);
4210         lcb = NULL;
4211
4212         /*
4213          * Go through the table and remove the duplicates,
4214          * remembering the oldest lsn values.
4215          */
4216         if (sbi->cluster_size <= log->page_size)
4217                 goto trace_dp_table;
4218
4219         dp = NULL;
4220         while ((dp = enum_rstbl(dptbl, dp))) {
4221                 struct DIR_PAGE_ENTRY *next = dp;
4222
4223                 while ((next = enum_rstbl(dptbl, next))) {
4224                         if (next->target_attr == dp->target_attr &&
4225                             next->vcn == dp->vcn) {
4226                                 if (le64_to_cpu(next->oldest_lsn) <
4227                                     le64_to_cpu(dp->oldest_lsn)) {
4228                                         dp->oldest_lsn = next->oldest_lsn;
4229                                 }
4230
4231                                 free_rsttbl_idx(dptbl, PtrOffset(dptbl, next));
4232                         }
4233                 }
4234         }
4235 trace_dp_table:
4236 check_attribute_names:
4237         /* The next record should be the Attribute Names. */
4238         if (!rst->attr_names_len)
4239                 goto check_attr_table;
4240
4241         t64 = le64_to_cpu(rst->attr_names_lsn);
4242         err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4243         if (err)
4244                 goto out;
4245
4246         lrh = lcb->log_rec;
4247         frh = lcb->lrh;
4248         rec_len = le32_to_cpu(frh->client_data_len);
4249
4250         if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4251                            bytes_per_attr_entry)) {
4252                 err = -EINVAL;
4253                 goto out;
4254         }
4255
4256         t32 = lrh_length(lrh);
4257         rec_len -= t32;
4258
4259         attr_names = kmemdup(Add2Ptr(lrh, t32), rec_len, GFP_NOFS);
4260         if (!attr_names) {
4261                 err = -ENOMEM;
4262                 goto out;
4263         }
4264
4265         lcb_put(lcb);
4266         lcb = NULL;
4267
4268 check_attr_table:
4269         /* The next record should be the attribute Table. */
4270         if (!rst->open_attr_len)
4271                 goto check_attribute_names2;
4272
4273         t64 = le64_to_cpu(rst->open_attr_table_lsn);
4274         err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4275         if (err)
4276                 goto out;
4277
4278         lrh = lcb->log_rec;
4279         frh = lcb->lrh;
4280         rec_len = le32_to_cpu(frh->client_data_len);
4281
4282         if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4283                            bytes_per_attr_entry)) {
4284                 err = -EINVAL;
4285                 goto out;
4286         }
4287
4288         t16 = le16_to_cpu(lrh->redo_off);
4289
4290         rt = Add2Ptr(lrh, t16);
4291         t32 = rec_len - t16;
4292
4293         if (!check_rstbl(rt, t32)) {
4294                 err = -EINVAL;
4295                 goto out;
4296         }
4297
4298         oatbl = kmemdup(rt, t32, GFP_NOFS);
4299         if (!oatbl) {
4300                 err = -ENOMEM;
4301                 goto out;
4302         }
4303
4304         log->open_attr_tbl = oatbl;
4305
4306         /* Clear all of the Attr pointers. */
4307         oe = NULL;
4308         while ((oe = enum_rstbl(oatbl, oe))) {
4309                 if (!rst->major_ver) {
4310                         struct OPEN_ATTR_ENRTY_32 oe0;
4311
4312                         /* Really 'oe' points to OPEN_ATTR_ENRTY_32. */
4313                         memcpy(&oe0, oe, SIZEOF_OPENATTRIBUTEENTRY0);
4314
4315                         oe->bytes_per_index = oe0.bytes_per_index;
4316                         oe->type = oe0.type;
4317                         oe->is_dirty_pages = oe0.is_dirty_pages;
4318                         oe->name_len = 0;
4319                         oe->ref = oe0.ref;
4320                         oe->open_record_lsn = oe0.open_record_lsn;
4321                 }
4322
4323                 oe->is_attr_name = 0;
4324                 oe->ptr = NULL;
4325         }
4326
4327         lcb_put(lcb);
4328         lcb = NULL;
4329
4330 check_attribute_names2:
4331         if (!rst->attr_names_len)
4332                 goto trace_attribute_table;
4333
4334         ane = attr_names;
4335         if (!oatbl)
4336                 goto trace_attribute_table;
4337         while (ane->off) {
4338                 /* TODO: Clear table on exit! */
4339                 oe = Add2Ptr(oatbl, le16_to_cpu(ane->off));
4340                 t16 = le16_to_cpu(ane->name_bytes);
4341                 oe->name_len = t16 / sizeof(short);
4342                 oe->ptr = ane->name;
4343                 oe->is_attr_name = 2;
4344                 ane = Add2Ptr(ane, sizeof(struct ATTR_NAME_ENTRY) + t16);
4345         }
4346
4347 trace_attribute_table:
4348         /*
4349          * If the checkpt_lsn is zero, then this is a freshly
4350          * formatted disk and we have no work to do.
4351          */
4352         if (!checkpt_lsn) {
4353                 err = 0;
4354                 goto out;
4355         }
4356
4357         if (!oatbl) {
4358                 oatbl = init_rsttbl(bytes_per_attr_entry, 8);
4359                 if (!oatbl) {
4360                         err = -ENOMEM;
4361                         goto out;
4362                 }
4363         }
4364
4365         log->open_attr_tbl = oatbl;
4366
4367         /* Start the analysis pass from the Checkpoint lsn. */
4368         rec_lsn = checkpt_lsn;
4369
4370         /* Read the first lsn. */
4371         err = read_log_rec_lcb(log, checkpt_lsn, lcb_ctx_next, &lcb);
4372         if (err)
4373                 goto out;
4374
4375         /* Loop to read all subsequent records to the end of the log file. */
4376 next_log_record_analyze:
4377         err = read_next_log_rec(log, lcb, &rec_lsn);
4378         if (err)
4379                 goto out;
4380
4381         if (!rec_lsn)
4382                 goto end_log_records_enumerate;
4383
4384         frh = lcb->lrh;
4385         transact_id = le32_to_cpu(frh->transact_id);
4386         rec_len = le32_to_cpu(frh->client_data_len);
4387         lrh = lcb->log_rec;
4388
4389         if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
4390                 err = -EINVAL;
4391                 goto out;
4392         }
4393
4394         /*
4395          * The first lsn after the previous lsn remembered
4396          * the checkpoint is the first candidate for the rlsn.
4397          */
4398         if (!rlsn)
4399                 rlsn = rec_lsn;
4400
4401         if (LfsClientRecord != frh->record_type)
4402                 goto next_log_record_analyze;
4403
4404         /*
4405          * Now update the Transaction Table for this transaction. If there
4406          * is no entry present or it is unallocated we allocate the entry.
4407          */
4408         if (!trtbl) {
4409                 trtbl = init_rsttbl(sizeof(struct TRANSACTION_ENTRY),
4410                                     INITIAL_NUMBER_TRANSACTIONS);
4411                 if (!trtbl) {
4412                         err = -ENOMEM;
4413                         goto out;
4414                 }
4415         }
4416
4417         tr = Add2Ptr(trtbl, transact_id);
4418
4419         if (transact_id >= bytes_per_rt(trtbl) ||
4420             tr->next != RESTART_ENTRY_ALLOCATED_LE) {
4421                 tr = alloc_rsttbl_from_idx(&trtbl, transact_id);
4422                 if (!tr) {
4423                         err = -ENOMEM;
4424                         goto out;
4425                 }
4426                 tr->transact_state = TransactionActive;
4427                 tr->first_lsn = cpu_to_le64(rec_lsn);
4428         }
4429
4430         tr->prev_lsn = tr->undo_next_lsn = cpu_to_le64(rec_lsn);
4431
4432         /*
4433          * If this is a compensation log record, then change
4434          * the undo_next_lsn to be the undo_next_lsn of this record.
4435          */
4436         if (lrh->undo_op == cpu_to_le16(CompensationLogRecord))
4437                 tr->undo_next_lsn = frh->client_undo_next_lsn;
4438
4439         /* Dispatch to handle log record depending on type. */
4440         switch (le16_to_cpu(lrh->redo_op)) {
4441         case InitializeFileRecordSegment:
4442         case DeallocateFileRecordSegment:
4443         case WriteEndOfFileRecordSegment:
4444         case CreateAttribute:
4445         case DeleteAttribute:
4446         case UpdateResidentValue:
4447         case UpdateNonresidentValue:
4448         case UpdateMappingPairs:
4449         case SetNewAttributeSizes:
4450         case AddIndexEntryRoot:
4451         case DeleteIndexEntryRoot:
4452         case AddIndexEntryAllocation:
4453         case DeleteIndexEntryAllocation:
4454         case WriteEndOfIndexBuffer:
4455         case SetIndexEntryVcnRoot:
4456         case SetIndexEntryVcnAllocation:
4457         case UpdateFileNameRoot:
4458         case UpdateFileNameAllocation:
4459         case SetBitsInNonresidentBitMap:
4460         case ClearBitsInNonresidentBitMap:
4461         case UpdateRecordDataRoot:
4462         case UpdateRecordDataAllocation:
4463         case ZeroEndOfFileRecord:
4464                 t16 = le16_to_cpu(lrh->target_attr);
4465                 t64 = le64_to_cpu(lrh->target_vcn);
4466                 dp = find_dp(dptbl, t16, t64);
4467
4468                 if (dp)
4469                         goto copy_lcns;
4470
4471                 /*
4472                  * Calculate the number of clusters per page the system
4473                  * which wrote the checkpoint, possibly creating the table.
4474                  */
4475                 if (dptbl) {
4476                         t32 = (le16_to_cpu(dptbl->size) -
4477                                sizeof(struct DIR_PAGE_ENTRY)) /
4478                               sizeof(u64);
4479                 } else {
4480                         t32 = log->clst_per_page;
4481                         kfree(dptbl);
4482                         dptbl = init_rsttbl(struct_size(dp, page_lcns, t32),
4483                                             32);
4484                         if (!dptbl) {
4485                                 err = -ENOMEM;
4486                                 goto out;
4487                         }
4488                 }
4489
4490                 dp = alloc_rsttbl_idx(&dptbl);
4491                 if (!dp) {
4492                         err = -ENOMEM;
4493                         goto out;
4494                 }
4495                 dp->target_attr = cpu_to_le32(t16);
4496                 dp->transfer_len = cpu_to_le32(t32 << sbi->cluster_bits);
4497                 dp->lcns_follow = cpu_to_le32(t32);
4498                 dp->vcn = cpu_to_le64(t64 & ~((u64)t32 - 1));
4499                 dp->oldest_lsn = cpu_to_le64(rec_lsn);
4500
4501 copy_lcns:
4502                 /*
4503                  * Copy the Lcns from the log record into the Dirty Page Entry.
4504                  * TODO: For different page size support, must somehow make
4505                  * whole routine a loop, case Lcns do not fit below.
4506                  */
4507                 t16 = le16_to_cpu(lrh->lcns_follow);
4508                 for (i = 0; i < t16; i++) {
4509                         size_t j = (size_t)(le64_to_cpu(lrh->target_vcn) -
4510                                             le64_to_cpu(dp->vcn));
4511                         dp->page_lcns[j + i] = lrh->page_lcns[i];
4512                 }
4513
4514                 goto next_log_record_analyze;
4515
4516         case DeleteDirtyClusters: {
4517                 u32 range_count =
4518                         le16_to_cpu(lrh->redo_len) / sizeof(struct LCN_RANGE);
4519                 const struct LCN_RANGE *r =
4520                         Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
4521
4522                 /* Loop through all of the Lcn ranges this log record. */
4523                 for (i = 0; i < range_count; i++, r++) {
4524                         u64 lcn0 = le64_to_cpu(r->lcn);
4525                         u64 lcn_e = lcn0 + le64_to_cpu(r->len) - 1;
4526
4527                         dp = NULL;
4528                         while ((dp = enum_rstbl(dptbl, dp))) {
4529                                 u32 j;
4530
4531                                 t32 = le32_to_cpu(dp->lcns_follow);
4532                                 for (j = 0; j < t32; j++) {
4533                                         t64 = le64_to_cpu(dp->page_lcns[j]);
4534                                         if (t64 >= lcn0 && t64 <= lcn_e)
4535                                                 dp->page_lcns[j] = 0;
4536                                 }
4537                         }
4538                 }
4539                 goto next_log_record_analyze;
4540                 ;
4541         }
4542
4543         case OpenNonresidentAttribute:
4544                 t16 = le16_to_cpu(lrh->target_attr);
4545                 if (t16 >= bytes_per_rt(oatbl)) {
4546                         /*
4547                          * Compute how big the table needs to be.
4548                          * Add 10 extra entries for some cushion.
4549                          */
4550                         u32 new_e = t16 / le16_to_cpu(oatbl->size);
4551
4552                         new_e += 10 - le16_to_cpu(oatbl->used);
4553
4554                         oatbl = extend_rsttbl(oatbl, new_e, ~0u);
4555                         log->open_attr_tbl = oatbl;
4556                         if (!oatbl) {
4557                                 err = -ENOMEM;
4558                                 goto out;
4559                         }
4560                 }
4561
4562                 /* Point to the entry being opened. */
4563                 oe = alloc_rsttbl_from_idx(&oatbl, t16);
4564                 log->open_attr_tbl = oatbl;
4565                 if (!oe) {
4566                         err = -ENOMEM;
4567                         goto out;
4568                 }
4569
4570                 /* Initialize this entry from the log record. */
4571                 t16 = le16_to_cpu(lrh->redo_off);
4572                 if (!rst->major_ver) {
4573                         /* Convert version '0' into version '1'. */
4574                         struct OPEN_ATTR_ENRTY_32 *oe0 = Add2Ptr(lrh, t16);
4575
4576                         oe->bytes_per_index = oe0->bytes_per_index;
4577                         oe->type = oe0->type;
4578                         oe->is_dirty_pages = oe0->is_dirty_pages;
4579                         oe->name_len = 0; //oe0.name_len;
4580                         oe->ref = oe0->ref;
4581                         oe->open_record_lsn = oe0->open_record_lsn;
4582                 } else {
4583                         memcpy(oe, Add2Ptr(lrh, t16), bytes_per_attr_entry);
4584                 }
4585
4586                 t16 = le16_to_cpu(lrh->undo_len);
4587                 if (t16) {
4588                         oe->ptr = kmalloc(t16, GFP_NOFS);
4589                         if (!oe->ptr) {
4590                                 err = -ENOMEM;
4591                                 goto out;
4592                         }
4593                         oe->name_len = t16 / sizeof(short);
4594                         memcpy(oe->ptr,
4595                                Add2Ptr(lrh, le16_to_cpu(lrh->undo_off)), t16);
4596                         oe->is_attr_name = 1;
4597                 } else {
4598                         oe->ptr = NULL;
4599                         oe->is_attr_name = 0;
4600                 }
4601
4602                 goto next_log_record_analyze;
4603
4604         case HotFix:
4605                 t16 = le16_to_cpu(lrh->target_attr);
4606                 t64 = le64_to_cpu(lrh->target_vcn);
4607                 dp = find_dp(dptbl, t16, t64);
4608                 if (dp) {
4609                         size_t j = le64_to_cpu(lrh->target_vcn) -
4610                                    le64_to_cpu(dp->vcn);
4611                         if (dp->page_lcns[j])
4612                                 dp->page_lcns[j] = lrh->page_lcns[0];
4613                 }
4614                 goto next_log_record_analyze;
4615
4616         case EndTopLevelAction:
4617                 tr = Add2Ptr(trtbl, transact_id);
4618                 tr->prev_lsn = cpu_to_le64(rec_lsn);
4619                 tr->undo_next_lsn = frh->client_undo_next_lsn;
4620                 goto next_log_record_analyze;
4621
4622         case PrepareTransaction:
4623                 tr = Add2Ptr(trtbl, transact_id);
4624                 tr->transact_state = TransactionPrepared;
4625                 goto next_log_record_analyze;
4626
4627         case CommitTransaction:
4628                 tr = Add2Ptr(trtbl, transact_id);
4629                 tr->transact_state = TransactionCommitted;
4630                 goto next_log_record_analyze;
4631
4632         case ForgetTransaction:
4633                 free_rsttbl_idx(trtbl, transact_id);
4634                 goto next_log_record_analyze;
4635
4636         case Noop:
4637         case OpenAttributeTableDump:
4638         case AttributeNamesDump:
4639         case DirtyPageTableDump:
4640         case TransactionTableDump:
4641                 /* The following cases require no action the Analysis Pass. */
4642                 goto next_log_record_analyze;
4643
4644         default:
4645                 /*
4646                  * All codes will be explicitly handled.
4647                  * If we see a code we do not expect, then we are trouble.
4648                  */
4649                 goto next_log_record_analyze;
4650         }
4651
4652 end_log_records_enumerate:
4653         lcb_put(lcb);
4654         lcb = NULL;
4655
4656         /*
4657          * Scan the Dirty Page Table and Transaction Table for
4658          * the lowest lsn, and return it as the Redo lsn.
4659          */
4660         dp = NULL;
4661         while ((dp = enum_rstbl(dptbl, dp))) {
4662                 t64 = le64_to_cpu(dp->oldest_lsn);
4663                 if (t64 && t64 < rlsn)
4664                         rlsn = t64;
4665         }
4666
4667         tr = NULL;
4668         while ((tr = enum_rstbl(trtbl, tr))) {
4669                 t64 = le64_to_cpu(tr->first_lsn);
4670                 if (t64 && t64 < rlsn)
4671                         rlsn = t64;
4672         }
4673
4674         /*
4675          * Only proceed if the Dirty Page Table or Transaction
4676          * table are not empty.
4677          */
4678         if ((!dptbl || !dptbl->total) && (!trtbl || !trtbl->total))
4679                 goto end_reply;
4680
4681         sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
4682         if (is_ro)
4683                 goto out;
4684
4685         /* Reopen all of the attributes with dirty pages. */
4686         oe = NULL;
4687 next_open_attribute:
4688
4689         oe = enum_rstbl(oatbl, oe);
4690         if (!oe) {
4691                 err = 0;
4692                 dp = NULL;
4693                 goto next_dirty_page;
4694         }
4695
4696         oa = kzalloc(sizeof(struct OpenAttr), GFP_NOFS);
4697         if (!oa) {
4698                 err = -ENOMEM;
4699                 goto out;
4700         }
4701
4702         inode = ntfs_iget5(sbi->sb, &oe->ref, NULL);
4703         if (IS_ERR(inode))
4704                 goto fake_attr;
4705
4706         if (is_bad_inode(inode)) {
4707                 iput(inode);
4708 fake_attr:
4709                 if (oa->ni) {
4710                         iput(&oa->ni->vfs_inode);
4711                         oa->ni = NULL;
4712                 }
4713
4714                 attr = attr_create_nonres_log(sbi, oe->type, 0, oe->ptr,
4715                                               oe->name_len, 0);
4716                 if (!attr) {
4717                         kfree(oa);
4718                         err = -ENOMEM;
4719                         goto out;
4720                 }
4721                 oa->attr = attr;
4722                 oa->run1 = &oa->run0;
4723                 goto final_oe;
4724         }
4725
4726         ni_oe = ntfs_i(inode);
4727         oa->ni = ni_oe;
4728
4729         attr = ni_find_attr(ni_oe, NULL, NULL, oe->type, oe->ptr, oe->name_len,
4730                             NULL, NULL);
4731
4732         if (!attr)
4733                 goto fake_attr;
4734
4735         t32 = le32_to_cpu(attr->size);
4736         oa->attr = kmemdup(attr, t32, GFP_NOFS);
4737         if (!oa->attr)
4738                 goto fake_attr;
4739
4740         if (!S_ISDIR(inode->i_mode)) {
4741                 if (attr->type == ATTR_DATA && !attr->name_len) {
4742                         oa->run1 = &ni_oe->file.run;
4743                         goto final_oe;
4744                 }
4745         } else {
4746                 if (attr->type == ATTR_ALLOC &&
4747                     attr->name_len == ARRAY_SIZE(I30_NAME) &&
4748                     !memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME))) {
4749                         oa->run1 = &ni_oe->dir.alloc_run;
4750                         goto final_oe;
4751                 }
4752         }
4753
4754         if (attr->non_res) {
4755                 u16 roff = le16_to_cpu(attr->nres.run_off);
4756                 CLST svcn = le64_to_cpu(attr->nres.svcn);
4757
4758                 if (roff > t32) {
4759                         kfree(oa->attr);
4760                         oa->attr = NULL;
4761                         goto fake_attr;
4762                 }
4763
4764                 err = run_unpack(&oa->run0, sbi, inode->i_ino, svcn,
4765                                  le64_to_cpu(attr->nres.evcn), svcn,
4766                                  Add2Ptr(attr, roff), t32 - roff);
4767                 if (err < 0) {
4768                         kfree(oa->attr);
4769                         oa->attr = NULL;
4770                         goto fake_attr;
4771                 }
4772                 err = 0;
4773         }
4774         oa->run1 = &oa->run0;
4775         attr = oa->attr;
4776
4777 final_oe:
4778         if (oe->is_attr_name == 1)
4779                 kfree(oe->ptr);
4780         oe->is_attr_name = 0;
4781         oe->ptr = oa;
4782         oe->name_len = attr->name_len;
4783
4784         goto next_open_attribute;
4785
4786         /*
4787          * Now loop through the dirty page table to extract all of the Vcn/Lcn.
4788          * Mapping that we have, and insert it into the appropriate run.
4789          */
4790 next_dirty_page:
4791         dp = enum_rstbl(dptbl, dp);
4792         if (!dp)
4793                 goto do_redo_1;
4794
4795         oe = Add2Ptr(oatbl, le32_to_cpu(dp->target_attr));
4796
4797         if (oe->next != RESTART_ENTRY_ALLOCATED_LE)
4798                 goto next_dirty_page;
4799
4800         oa = oe->ptr;
4801         if (!oa)
4802                 goto next_dirty_page;
4803
4804         i = -1;
4805 next_dirty_page_vcn:
4806         i += 1;
4807         if (i >= le32_to_cpu(dp->lcns_follow))
4808                 goto next_dirty_page;
4809
4810         vcn = le64_to_cpu(dp->vcn) + i;
4811         size = (vcn + 1) << sbi->cluster_bits;
4812
4813         if (!dp->page_lcns[i])
4814                 goto next_dirty_page_vcn;
4815
4816         rno = ino_get(&oe->ref);
4817         if (rno <= MFT_REC_MIRR &&
4818             size < (MFT_REC_VOL + 1) * sbi->record_size &&
4819             oe->type == ATTR_DATA) {
4820                 goto next_dirty_page_vcn;
4821         }
4822
4823         lcn = le64_to_cpu(dp->page_lcns[i]);
4824
4825         if ((!run_lookup_entry(oa->run1, vcn, &lcn0, &len0, NULL) ||
4826              lcn0 != lcn) &&
4827             !run_add_entry(oa->run1, vcn, lcn, 1, false)) {
4828                 err = -ENOMEM;
4829                 goto out;
4830         }
4831         attr = oa->attr;
4832         if (size > le64_to_cpu(attr->nres.alloc_size)) {
4833                 attr->nres.valid_size = attr->nres.data_size =
4834                         attr->nres.alloc_size = cpu_to_le64(size);
4835         }
4836         goto next_dirty_page_vcn;
4837
4838 do_redo_1:
4839         /*
4840          * Perform the Redo Pass, to restore all of the dirty pages to the same
4841          * contents that they had immediately before the crash. If the dirty
4842          * page table is empty, then we can skip the entire Redo Pass.
4843          */
4844         if (!dptbl || !dptbl->total)
4845                 goto do_undo_action;
4846
4847         rec_lsn = rlsn;
4848
4849         /*
4850          * Read the record at the Redo lsn, before falling
4851          * into common code to handle each record.
4852          */
4853         err = read_log_rec_lcb(log, rlsn, lcb_ctx_next, &lcb);
4854         if (err)
4855                 goto out;
4856
4857         /*
4858          * Now loop to read all of our log records forwards, until
4859          * we hit the end of the file, cleaning up at the end.
4860          */
4861 do_action_next:
4862         frh = lcb->lrh;
4863
4864         if (LfsClientRecord != frh->record_type)
4865                 goto read_next_log_do_action;
4866
4867         transact_id = le32_to_cpu(frh->transact_id);
4868         rec_len = le32_to_cpu(frh->client_data_len);
4869         lrh = lcb->log_rec;
4870
4871         if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
4872                 err = -EINVAL;
4873                 goto out;
4874         }
4875
4876         /* Ignore log records that do not update pages. */
4877         if (lrh->lcns_follow)
4878                 goto find_dirty_page;
4879
4880         goto read_next_log_do_action;
4881
4882 find_dirty_page:
4883         t16 = le16_to_cpu(lrh->target_attr);
4884         t64 = le64_to_cpu(lrh->target_vcn);
4885         dp = find_dp(dptbl, t16, t64);
4886
4887         if (!dp)
4888                 goto read_next_log_do_action;
4889
4890         if (rec_lsn < le64_to_cpu(dp->oldest_lsn))
4891                 goto read_next_log_do_action;
4892
4893         t16 = le16_to_cpu(lrh->target_attr);
4894         if (t16 >= bytes_per_rt(oatbl)) {
4895                 err = -EINVAL;
4896                 goto out;
4897         }
4898
4899         oe = Add2Ptr(oatbl, t16);
4900
4901         if (oe->next != RESTART_ENTRY_ALLOCATED_LE) {
4902                 err = -EINVAL;
4903                 goto out;
4904         }
4905
4906         oa = oe->ptr;
4907
4908         if (!oa) {
4909                 err = -EINVAL;
4910                 goto out;
4911         }
4912         attr = oa->attr;
4913
4914         vcn = le64_to_cpu(lrh->target_vcn);
4915
4916         if (!run_lookup_entry(oa->run1, vcn, &lcn, NULL, NULL) ||
4917             lcn == SPARSE_LCN) {
4918                 goto read_next_log_do_action;
4919         }
4920
4921         /* Point to the Redo data and get its length. */
4922         data = Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
4923         dlen = le16_to_cpu(lrh->redo_len);
4924
4925         /* Shorten length by any Lcns which were deleted. */
4926         saved_len = dlen;
4927
4928         for (i = le16_to_cpu(lrh->lcns_follow); i; i--) {
4929                 size_t j;
4930                 u32 alen, voff;
4931
4932                 voff = le16_to_cpu(lrh->record_off) +
4933                        le16_to_cpu(lrh->attr_off);
4934                 voff += le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
4935
4936                 /* If the Vcn question is allocated, we can just get out. */
4937                 j = le64_to_cpu(lrh->target_vcn) - le64_to_cpu(dp->vcn);
4938                 if (dp->page_lcns[j + i - 1])
4939                         break;
4940
4941                 if (!saved_len)
4942                         saved_len = 1;
4943
4944                 /*
4945                  * Calculate the allocated space left relative to the
4946                  * log record Vcn, after removing this unallocated Vcn.
4947                  */
4948                 alen = (i - 1) << sbi->cluster_bits;
4949
4950                 /*
4951                  * If the update described this log record goes beyond
4952                  * the allocated space, then we will have to reduce the length.
4953                  */
4954                 if (voff >= alen)
4955                         dlen = 0;
4956                 else if (voff + dlen > alen)
4957                         dlen = alen - voff;
4958         }
4959
4960         /*
4961          * If the resulting dlen from above is now zero,
4962          * we can skip this log record.
4963          */
4964         if (!dlen && saved_len)
4965                 goto read_next_log_do_action;
4966
4967         t16 = le16_to_cpu(lrh->redo_op);
4968         if (can_skip_action(t16))
4969                 goto read_next_log_do_action;
4970
4971         /* Apply the Redo operation a common routine. */
4972         err = do_action(log, oe, lrh, t16, data, dlen, rec_len, &rec_lsn);
4973         if (err)
4974                 goto out;
4975
4976         /* Keep reading and looping back until end of file. */
4977 read_next_log_do_action:
4978         err = read_next_log_rec(log, lcb, &rec_lsn);
4979         if (!err && rec_lsn)
4980                 goto do_action_next;
4981
4982         lcb_put(lcb);
4983         lcb = NULL;
4984
4985 do_undo_action:
4986         /* Scan Transaction Table. */
4987         tr = NULL;
4988 transaction_table_next:
4989         tr = enum_rstbl(trtbl, tr);
4990         if (!tr)
4991                 goto undo_action_done;
4992
4993         if (TransactionActive != tr->transact_state || !tr->undo_next_lsn) {
4994                 free_rsttbl_idx(trtbl, PtrOffset(trtbl, tr));
4995                 goto transaction_table_next;
4996         }
4997
4998         log->transaction_id = PtrOffset(trtbl, tr);
4999         undo_next_lsn = le64_to_cpu(tr->undo_next_lsn);
5000
5001         /*
5002          * We only have to do anything if the transaction has
5003          * something its undo_next_lsn field.
5004          */
5005         if (!undo_next_lsn)
5006                 goto commit_undo;
5007
5008         /* Read the first record to be undone by this transaction. */
5009         err = read_log_rec_lcb(log, undo_next_lsn, lcb_ctx_undo_next, &lcb);
5010         if (err)
5011                 goto out;
5012
5013         /*
5014          * Now loop to read all of our log records forwards,
5015          * until we hit the end of the file, cleaning up at the end.
5016          */
5017 undo_action_next:
5018
5019         lrh = lcb->log_rec;
5020         frh = lcb->lrh;
5021         transact_id = le32_to_cpu(frh->transact_id);
5022         rec_len = le32_to_cpu(frh->client_data_len);
5023
5024         if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
5025                 err = -EINVAL;
5026                 goto out;
5027         }
5028
5029         if (lrh->undo_op == cpu_to_le16(Noop))
5030                 goto read_next_log_undo_action;
5031
5032         oe = Add2Ptr(oatbl, le16_to_cpu(lrh->target_attr));
5033         oa = oe->ptr;
5034
5035         t16 = le16_to_cpu(lrh->lcns_follow);
5036         if (!t16)
5037                 goto add_allocated_vcns;
5038
5039         is_mapped = run_lookup_entry(oa->run1, le64_to_cpu(lrh->target_vcn),
5040                                      &lcn, &clen, NULL);
5041
5042         /*
5043          * If the mapping isn't already the table or the  mapping
5044          * corresponds to a hole the mapping, we need to make sure
5045          * there is no partial page already memory.
5046          */
5047         if (is_mapped && lcn != SPARSE_LCN && clen >= t16)
5048                 goto add_allocated_vcns;
5049
5050         vcn = le64_to_cpu(lrh->target_vcn);
5051         vcn &= ~(u64)(log->clst_per_page - 1);
5052
5053 add_allocated_vcns:
5054         for (i = 0, vcn = le64_to_cpu(lrh->target_vcn),
5055             size = (vcn + 1) << sbi->cluster_bits;
5056              i < t16; i++, vcn += 1, size += sbi->cluster_size) {
5057                 attr = oa->attr;
5058                 if (!attr->non_res) {
5059                         if (size > le32_to_cpu(attr->res.data_size))
5060                                 attr->res.data_size = cpu_to_le32(size);
5061                 } else {
5062                         if (size > le64_to_cpu(attr->nres.data_size))
5063                                 attr->nres.valid_size = attr->nres.data_size =
5064                                         attr->nres.alloc_size =
5065                                                 cpu_to_le64(size);
5066                 }
5067         }
5068
5069         t16 = le16_to_cpu(lrh->undo_op);
5070         if (can_skip_action(t16))
5071                 goto read_next_log_undo_action;
5072
5073         /* Point to the Redo data and get its length. */
5074         data = Add2Ptr(lrh, le16_to_cpu(lrh->undo_off));
5075         dlen = le16_to_cpu(lrh->undo_len);
5076
5077         /* It is time to apply the undo action. */
5078         err = do_action(log, oe, lrh, t16, data, dlen, rec_len, NULL);
5079
5080 read_next_log_undo_action:
5081         /*
5082          * Keep reading and looping back until we have read the
5083          * last record for this transaction.
5084          */
5085         err = read_next_log_rec(log, lcb, &rec_lsn);
5086         if (err)
5087                 goto out;
5088
5089         if (rec_lsn)
5090                 goto undo_action_next;
5091
5092         lcb_put(lcb);
5093         lcb = NULL;
5094
5095 commit_undo:
5096         free_rsttbl_idx(trtbl, log->transaction_id);
5097
5098         log->transaction_id = 0;
5099
5100         goto transaction_table_next;
5101
5102 undo_action_done:
5103
5104         ntfs_update_mftmirr(sbi, 0);
5105
5106         sbi->flags &= ~NTFS_FLAGS_NEED_REPLAY;
5107
5108 end_reply:
5109
5110         err = 0;
5111         if (is_ro)
5112                 goto out;
5113
5114         rh = kzalloc(log->page_size, GFP_NOFS);
5115         if (!rh) {
5116                 err = -ENOMEM;
5117                 goto out;
5118         }
5119
5120         rh->rhdr.sign = NTFS_RSTR_SIGNATURE;
5121         rh->rhdr.fix_off = cpu_to_le16(offsetof(struct RESTART_HDR, fixups));
5122         t16 = (log->page_size >> SECTOR_SHIFT) + 1;
5123         rh->rhdr.fix_num = cpu_to_le16(t16);
5124         rh->sys_page_size = cpu_to_le32(log->page_size);
5125         rh->page_size = cpu_to_le32(log->page_size);
5126
5127         t16 = ALIGN(offsetof(struct RESTART_HDR, fixups) + sizeof(short) * t16,
5128                     8);
5129         rh->ra_off = cpu_to_le16(t16);
5130         rh->minor_ver = cpu_to_le16(1); // 0x1A:
5131         rh->major_ver = cpu_to_le16(1); // 0x1C:
5132
5133         ra2 = Add2Ptr(rh, t16);
5134         memcpy(ra2, ra, sizeof(struct RESTART_AREA));
5135
5136         ra2->client_idx[0] = 0;
5137         ra2->client_idx[1] = LFS_NO_CLIENT_LE;
5138         ra2->flags = cpu_to_le16(2);
5139
5140         le32_add_cpu(&ra2->open_log_count, 1);
5141
5142         ntfs_fix_pre_write(&rh->rhdr, log->page_size);
5143
5144         err = ntfs_sb_write_run(sbi, &ni->file.run, 0, rh, log->page_size, 0);
5145         if (!err)
5146                 err = ntfs_sb_write_run(sbi, &log->ni->file.run, log->page_size,
5147                                         rh, log->page_size, 0);
5148
5149         kfree(rh);
5150         if (err)
5151                 goto out;
5152
5153 out:
5154         kfree(rst);
5155         if (lcb)
5156                 lcb_put(lcb);
5157
5158         /*
5159          * Scan the Open Attribute Table to close all of
5160          * the open attributes.
5161          */
5162         oe = NULL;
5163         while ((oe = enum_rstbl(oatbl, oe))) {
5164                 rno = ino_get(&oe->ref);
5165
5166                 if (oe->is_attr_name == 1) {
5167                         kfree(oe->ptr);
5168                         oe->ptr = NULL;
5169                         continue;
5170                 }
5171
5172                 if (oe->is_attr_name)
5173                         continue;
5174
5175                 oa = oe->ptr;
5176                 if (!oa)
5177                         continue;
5178
5179                 run_close(&oa->run0);
5180                 kfree(oa->attr);
5181                 if (oa->ni)
5182                         iput(&oa->ni->vfs_inode);
5183                 kfree(oa);
5184         }
5185
5186         kfree(trtbl);
5187         kfree(oatbl);
5188         kfree(dptbl);
5189         kfree(attr_names);
5190         kfree(rst_info.r_page);
5191
5192         kfree(ra);
5193         kfree(log->one_page_buf);
5194
5195         if (err)
5196                 sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
5197
5198         if (err == -EROFS)
5199                 err = 0;
5200         else if (log->set_dirty)
5201                 ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
5202
5203         kfree(log);
5204
5205         return err;
5206 }