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(bytes, 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 = kmalloc(DefaultLogPageSize, GFP_NOFS);
1184
1185         if (!r_page)
1186                 return -ENOMEM;
1187
1188         memset(info, 0, sizeof(struct restart_info));
1189
1190         /* Determine which restart area we are looking for. */
1191         if (first) {
1192                 vbo = 0;
1193                 skip = 512;
1194         } else {
1195                 vbo = 512;
1196                 skip = 0;
1197         }
1198
1199         /* Loop continuously until we succeed. */
1200         for (; vbo < l_size; vbo = 2 * vbo + skip, skip = 0) {
1201                 bool usa_error;
1202                 u32 sys_page_size;
1203                 bool brst, bchk;
1204                 struct RESTART_AREA *ra;
1205
1206                 /* Read a page header at the current offset. */
1207                 if (read_log_page(log, vbo, (struct RECORD_PAGE_HDR **)&r_page,
1208                                   &usa_error)) {
1209                         /* Ignore any errors. */
1210                         continue;
1211                 }
1212
1213                 /* Exit if the signature is a log record page. */
1214                 if (r_page->rhdr.sign == NTFS_RCRD_SIGNATURE) {
1215                         info->initialized = true;
1216                         break;
1217                 }
1218
1219                 brst = r_page->rhdr.sign == NTFS_RSTR_SIGNATURE;
1220                 bchk = r_page->rhdr.sign == NTFS_CHKD_SIGNATURE;
1221
1222                 if (!bchk && !brst) {
1223                         if (r_page->rhdr.sign != NTFS_FFFF_SIGNATURE) {
1224                                 /*
1225                                  * Remember if the signature does not
1226                                  * indicate uninitialized file.
1227                                  */
1228                                 info->initialized = true;
1229                         }
1230                         continue;
1231                 }
1232
1233                 ra = NULL;
1234                 info->valid_page = false;
1235                 info->initialized = true;
1236                 info->vbo = vbo;
1237
1238                 /* Let's check the restart area if this is a valid page. */
1239                 if (!is_rst_page_hdr_valid(vbo, r_page))
1240                         goto check_result;
1241                 ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
1242
1243                 if (!is_rst_area_valid(r_page))
1244                         goto check_result;
1245
1246                 /*
1247                  * We have a valid restart page header and restart area.
1248                  * If chkdsk was run or we have no clients then we have
1249                  * no more checking to do.
1250                  */
1251                 if (bchk || ra->client_idx[1] == LFS_NO_CLIENT_LE) {
1252                         info->valid_page = true;
1253                         goto check_result;
1254                 }
1255
1256                 /* Read the entire restart area. */
1257                 sys_page_size = le32_to_cpu(r_page->sys_page_size);
1258                 if (DefaultLogPageSize != sys_page_size) {
1259                         kfree(r_page);
1260                         r_page = kzalloc(sys_page_size, GFP_NOFS);
1261                         if (!r_page)
1262                                 return -ENOMEM;
1263
1264                         if (read_log_page(log, vbo,
1265                                           (struct RECORD_PAGE_HDR **)&r_page,
1266                                           &usa_error)) {
1267                                 /* Ignore any errors. */
1268                                 kfree(r_page);
1269                                 r_page = NULL;
1270                                 continue;
1271                         }
1272                 }
1273
1274                 if (is_client_area_valid(r_page, usa_error)) {
1275                         info->valid_page = true;
1276                         ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
1277                 }
1278
1279 check_result:
1280                 /*
1281                  * If chkdsk was run then update the caller's
1282                  * values and return.
1283                  */
1284                 if (r_page->rhdr.sign == NTFS_CHKD_SIGNATURE) {
1285                         info->chkdsk_was_run = true;
1286                         info->last_lsn = le64_to_cpu(r_page->rhdr.lsn);
1287                         info->restart = true;
1288                         info->r_page = r_page;
1289                         return 0;
1290                 }
1291
1292                 /*
1293                  * If we have a valid page then copy the values
1294                  * we need from it.
1295                  */
1296                 if (info->valid_page) {
1297                         info->last_lsn = le64_to_cpu(ra->current_lsn);
1298                         info->restart = true;
1299                         info->r_page = r_page;
1300                         return 0;
1301                 }
1302         }
1303
1304         kfree(r_page);
1305
1306         return 0;
1307 }
1308
1309 /*
1310  * Ilog_init_pg_hdr - Init @log from restart page header.
1311  */
1312 static void log_init_pg_hdr(struct ntfs_log *log, u32 sys_page_size,
1313                             u32 page_size, u16 major_ver, u16 minor_ver)
1314 {
1315         log->sys_page_size = sys_page_size;
1316         log->sys_page_mask = sys_page_size - 1;
1317         log->page_size = page_size;
1318         log->page_mask = page_size - 1;
1319         log->page_bits = blksize_bits(page_size);
1320
1321         log->clst_per_page = log->page_size >> log->ni->mi.sbi->cluster_bits;
1322         if (!log->clst_per_page)
1323                 log->clst_per_page = 1;
1324
1325         log->first_page = major_ver >= 2
1326                                   ? 0x22 * page_size
1327                                   : ((sys_page_size << 1) + (page_size << 1));
1328         log->major_ver = major_ver;
1329         log->minor_ver = minor_ver;
1330 }
1331
1332 /*
1333  * log_create - Init @log in cases when we don't have a restart area to use.
1334  */
1335 static void log_create(struct ntfs_log *log, u32 l_size, const u64 last_lsn,
1336                        u32 open_log_count, bool wrapped, bool use_multi_page)
1337 {
1338         log->l_size = l_size;
1339         /* All file offsets must be quadword aligned. */
1340         log->file_data_bits = blksize_bits(l_size) - 3;
1341         log->seq_num_mask = (8 << log->file_data_bits) - 1;
1342         log->seq_num_bits = sizeof(u64) * 8 - log->file_data_bits;
1343         log->seq_num = (last_lsn >> log->file_data_bits) + 2;
1344         log->next_page = log->first_page;
1345         log->oldest_lsn = log->seq_num << log->file_data_bits;
1346         log->oldest_lsn_off = 0;
1347         log->last_lsn = log->oldest_lsn;
1348
1349         log->l_flags |= NTFSLOG_NO_LAST_LSN | NTFSLOG_NO_OLDEST_LSN;
1350
1351         /* Set the correct flags for the I/O and indicate if we have wrapped. */
1352         if (wrapped)
1353                 log->l_flags |= NTFSLOG_WRAPPED;
1354
1355         if (use_multi_page)
1356                 log->l_flags |= NTFSLOG_MULTIPLE_PAGE_IO;
1357
1358         /* Compute the log page values. */
1359         log->data_off = ALIGN(
1360                 offsetof(struct RECORD_PAGE_HDR, fixups) +
1361                         sizeof(short) * ((log->page_size >> SECTOR_SHIFT) + 1),
1362                 8);
1363         log->data_size = log->page_size - log->data_off;
1364         log->record_header_len = sizeof(struct LFS_RECORD_HDR);
1365
1366         /* Remember the different page sizes for reservation. */
1367         log->reserved = log->data_size - log->record_header_len;
1368
1369         /* Compute the restart page values. */
1370         log->ra_off = ALIGN(
1371                 offsetof(struct RESTART_HDR, fixups) +
1372                         sizeof(short) *
1373                                 ((log->sys_page_size >> SECTOR_SHIFT) + 1),
1374                 8);
1375         log->restart_size = log->sys_page_size - log->ra_off;
1376         log->ra_size = struct_size(log->ra, clients, 1);
1377         log->current_openlog_count = open_log_count;
1378
1379         /*
1380          * The total available log file space is the number of
1381          * log file pages times the space available on each page.
1382          */
1383         log->total_avail_pages = log->l_size - log->first_page;
1384         log->total_avail = log->total_avail_pages >> log->page_bits;
1385
1386         /*
1387          * We assume that we can't use the end of the page less than
1388          * the file record size.
1389          * Then we won't need to reserve more than the caller asks for.
1390          */
1391         log->max_current_avail = log->total_avail * log->reserved;
1392         log->total_avail = log->total_avail * log->data_size;
1393         log->current_avail = log->max_current_avail;
1394 }
1395
1396 /*
1397  * log_create_ra - Fill a restart area from the values stored in @log.
1398  */
1399 static struct RESTART_AREA *log_create_ra(struct ntfs_log *log)
1400 {
1401         struct CLIENT_REC *cr;
1402         struct RESTART_AREA *ra = kzalloc(log->restart_size, GFP_NOFS);
1403
1404         if (!ra)
1405                 return NULL;
1406
1407         ra->current_lsn = cpu_to_le64(log->last_lsn);
1408         ra->log_clients = cpu_to_le16(1);
1409         ra->client_idx[1] = LFS_NO_CLIENT_LE;
1410         if (log->l_flags & NTFSLOG_MULTIPLE_PAGE_IO)
1411                 ra->flags = RESTART_SINGLE_PAGE_IO;
1412         ra->seq_num_bits = cpu_to_le32(log->seq_num_bits);
1413         ra->ra_len = cpu_to_le16(log->ra_size);
1414         ra->client_off = cpu_to_le16(offsetof(struct RESTART_AREA, clients));
1415         ra->l_size = cpu_to_le64(log->l_size);
1416         ra->rec_hdr_len = cpu_to_le16(log->record_header_len);
1417         ra->data_off = cpu_to_le16(log->data_off);
1418         ra->open_log_count = cpu_to_le32(log->current_openlog_count + 1);
1419
1420         cr = ra->clients;
1421
1422         cr->prev_client = LFS_NO_CLIENT_LE;
1423         cr->next_client = LFS_NO_CLIENT_LE;
1424
1425         return ra;
1426 }
1427
1428 static u32 final_log_off(struct ntfs_log *log, u64 lsn, u32 data_len)
1429 {
1430         u32 base_vbo = lsn << 3;
1431         u32 final_log_off = (base_vbo & log->seq_num_mask) & ~log->page_mask;
1432         u32 page_off = base_vbo & log->page_mask;
1433         u32 tail = log->page_size - page_off;
1434
1435         page_off -= 1;
1436
1437         /* Add the length of the header. */
1438         data_len += log->record_header_len;
1439
1440         /*
1441          * If this lsn is contained this log page we are done.
1442          * Otherwise we need to walk through several log pages.
1443          */
1444         if (data_len > tail) {
1445                 data_len -= tail;
1446                 tail = log->data_size;
1447                 page_off = log->data_off - 1;
1448
1449                 for (;;) {
1450                         final_log_off = next_page_off(log, final_log_off);
1451
1452                         /*
1453                          * We are done if the remaining bytes
1454                          * fit on this page.
1455                          */
1456                         if (data_len <= tail)
1457                                 break;
1458                         data_len -= tail;
1459                 }
1460         }
1461
1462         /*
1463          * We add the remaining bytes to our starting position on this page
1464          * and then add that value to the file offset of this log page.
1465          */
1466         return final_log_off + data_len + page_off;
1467 }
1468
1469 static int next_log_lsn(struct ntfs_log *log, const struct LFS_RECORD_HDR *rh,
1470                         u64 *lsn)
1471 {
1472         int err;
1473         u64 this_lsn = le64_to_cpu(rh->this_lsn);
1474         u32 vbo = lsn_to_vbo(log, this_lsn);
1475         u32 end =
1476                 final_log_off(log, this_lsn, le32_to_cpu(rh->client_data_len));
1477         u32 hdr_off = end & ~log->sys_page_mask;
1478         u64 seq = this_lsn >> log->file_data_bits;
1479         struct RECORD_PAGE_HDR *page = NULL;
1480
1481         /* Remember if we wrapped. */
1482         if (end <= vbo)
1483                 seq += 1;
1484
1485         /* Log page header for this page. */
1486         err = read_log_page(log, hdr_off, &page, NULL);
1487         if (err)
1488                 return err;
1489
1490         /*
1491          * If the lsn we were given was not the last lsn on this page,
1492          * then the starting offset for the next lsn is on a quad word
1493          * boundary following the last file offset for the current lsn.
1494          * Otherwise the file offset is the start of the data on the next page.
1495          */
1496         if (this_lsn == le64_to_cpu(page->rhdr.lsn)) {
1497                 /* If we wrapped, we need to increment the sequence number. */
1498                 hdr_off = next_page_off(log, hdr_off);
1499                 if (hdr_off == log->first_page)
1500                         seq += 1;
1501
1502                 vbo = hdr_off + log->data_off;
1503         } else {
1504                 vbo = ALIGN(end, 8);
1505         }
1506
1507         /* Compute the lsn based on the file offset and the sequence count. */
1508         *lsn = vbo_to_lsn(log, vbo, seq);
1509
1510         /*
1511          * If this lsn is within the legal range for the file, we return true.
1512          * Otherwise false indicates that there are no more lsn's.
1513          */
1514         if (!is_lsn_in_file(log, *lsn))
1515                 *lsn = 0;
1516
1517         kfree(page);
1518
1519         return 0;
1520 }
1521
1522 /*
1523  * current_log_avail - Calculate the number of bytes available for log records.
1524  */
1525 static u32 current_log_avail(struct ntfs_log *log)
1526 {
1527         u32 oldest_off, next_free_off, free_bytes;
1528
1529         if (log->l_flags & NTFSLOG_NO_LAST_LSN) {
1530                 /* The entire file is available. */
1531                 return log->max_current_avail;
1532         }
1533
1534         /*
1535          * If there is a last lsn the restart area then we know that we will
1536          * have to compute the free range.
1537          * If there is no oldest lsn then start at the first page of the file.
1538          */
1539         oldest_off = (log->l_flags & NTFSLOG_NO_OLDEST_LSN)
1540                              ? log->first_page
1541                              : (log->oldest_lsn_off & ~log->sys_page_mask);
1542
1543         /*
1544          * We will use the next log page offset to compute the next free page.
1545          * If we are going to reuse this page go to the next page.
1546          * If we are at the first page then use the end of the file.
1547          */
1548         next_free_off = (log->l_flags & NTFSLOG_REUSE_TAIL)
1549                                 ? log->next_page + log->page_size
1550                                 : log->next_page == log->first_page
1551                                           ? log->l_size
1552                                           : log->next_page;
1553
1554         /* If the two offsets are the same then there is no available space. */
1555         if (oldest_off == next_free_off)
1556                 return 0;
1557         /*
1558          * If the free offset follows the oldest offset then subtract
1559          * this range from the total available pages.
1560          */
1561         free_bytes =
1562                 oldest_off < next_free_off
1563                         ? log->total_avail_pages - (next_free_off - oldest_off)
1564                         : oldest_off - next_free_off;
1565
1566         free_bytes >>= log->page_bits;
1567         return free_bytes * log->reserved;
1568 }
1569
1570 static bool check_subseq_log_page(struct ntfs_log *log,
1571                                   const struct RECORD_PAGE_HDR *rp, u32 vbo,
1572                                   u64 seq)
1573 {
1574         u64 lsn_seq;
1575         const struct NTFS_RECORD_HEADER *rhdr = &rp->rhdr;
1576         u64 lsn = le64_to_cpu(rhdr->lsn);
1577
1578         if (rhdr->sign == NTFS_FFFF_SIGNATURE || !rhdr->sign)
1579                 return false;
1580
1581         /*
1582          * If the last lsn on the page occurs was written after the page
1583          * that caused the original error then we have a fatal error.
1584          */
1585         lsn_seq = lsn >> log->file_data_bits;
1586
1587         /*
1588          * If the sequence number for the lsn the page is equal or greater
1589          * than lsn we expect, then this is a subsequent write.
1590          */
1591         return lsn_seq >= seq ||
1592                (lsn_seq == seq - 1 && log->first_page == vbo &&
1593                 vbo != (lsn_to_vbo(log, lsn) & ~log->page_mask));
1594 }
1595
1596 /*
1597  * last_log_lsn
1598  *
1599  * Walks through the log pages for a file, searching for the
1600  * last log page written to the file.
1601  */
1602 static int last_log_lsn(struct ntfs_log *log)
1603 {
1604         int err;
1605         bool usa_error = false;
1606         bool replace_page = false;
1607         bool reuse_page = log->l_flags & NTFSLOG_REUSE_TAIL;
1608         bool wrapped_file, wrapped;
1609
1610         u32 page_cnt = 1, page_pos = 1;
1611         u32 page_off = 0, page_off1 = 0, saved_off = 0;
1612         u32 final_off, second_off, final_off_prev = 0, second_off_prev = 0;
1613         u32 first_file_off = 0, second_file_off = 0;
1614         u32 part_io_count = 0;
1615         u32 tails = 0;
1616         u32 this_off, curpage_off, nextpage_off, remain_pages;
1617
1618         u64 expected_seq, seq_base = 0, lsn_base = 0;
1619         u64 best_lsn, best_lsn1, best_lsn2;
1620         u64 lsn_cur, lsn1, lsn2;
1621         u64 last_ok_lsn = reuse_page ? log->last_lsn : 0;
1622
1623         u16 cur_pos, best_page_pos;
1624
1625         struct RECORD_PAGE_HDR *page = NULL;
1626         struct RECORD_PAGE_HDR *tst_page = NULL;
1627         struct RECORD_PAGE_HDR *first_tail = NULL;
1628         struct RECORD_PAGE_HDR *second_tail = NULL;
1629         struct RECORD_PAGE_HDR *tail_page = NULL;
1630         struct RECORD_PAGE_HDR *second_tail_prev = NULL;
1631         struct RECORD_PAGE_HDR *first_tail_prev = NULL;
1632         struct RECORD_PAGE_HDR *page_bufs = NULL;
1633         struct RECORD_PAGE_HDR *best_page;
1634
1635         if (log->major_ver >= 2) {
1636                 final_off = 0x02 * log->page_size;
1637                 second_off = 0x12 * log->page_size;
1638
1639                 // 0x10 == 0x12 - 0x2
1640                 page_bufs = kmalloc(log->page_size * 0x10, GFP_NOFS);
1641                 if (!page_bufs)
1642                         return -ENOMEM;
1643         } else {
1644                 second_off = log->first_page - log->page_size;
1645                 final_off = second_off - log->page_size;
1646         }
1647
1648 next_tail:
1649         /* Read second tail page (at pos 3/0x12000). */
1650         if (read_log_page(log, second_off, &second_tail, &usa_error) ||
1651             usa_error || second_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
1652                 kfree(second_tail);
1653                 second_tail = NULL;
1654                 second_file_off = 0;
1655                 lsn2 = 0;
1656         } else {
1657                 second_file_off = hdr_file_off(log, second_tail);
1658                 lsn2 = le64_to_cpu(second_tail->record_hdr.last_end_lsn);
1659         }
1660
1661         /* Read first tail page (at pos 2/0x2000). */
1662         if (read_log_page(log, final_off, &first_tail, &usa_error) ||
1663             usa_error || first_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
1664                 kfree(first_tail);
1665                 first_tail = NULL;
1666                 first_file_off = 0;
1667                 lsn1 = 0;
1668         } else {
1669                 first_file_off = hdr_file_off(log, first_tail);
1670                 lsn1 = le64_to_cpu(first_tail->record_hdr.last_end_lsn);
1671         }
1672
1673         if (log->major_ver < 2) {
1674                 int best_page;
1675
1676                 first_tail_prev = first_tail;
1677                 final_off_prev = first_file_off;
1678                 second_tail_prev = second_tail;
1679                 second_off_prev = second_file_off;
1680                 tails = 1;
1681
1682                 if (!first_tail && !second_tail)
1683                         goto tail_read;
1684
1685                 if (first_tail && second_tail)
1686                         best_page = lsn1 < lsn2 ? 1 : 0;
1687                 else if (first_tail)
1688                         best_page = 0;
1689                 else
1690                         best_page = 1;
1691
1692                 page_off = best_page ? second_file_off : first_file_off;
1693                 seq_base = (best_page ? lsn2 : lsn1) >> log->file_data_bits;
1694                 goto tail_read;
1695         }
1696
1697         best_lsn1 = first_tail ? base_lsn(log, first_tail, first_file_off) : 0;
1698         best_lsn2 =
1699                 second_tail ? base_lsn(log, second_tail, second_file_off) : 0;
1700
1701         if (first_tail && second_tail) {
1702                 if (best_lsn1 > best_lsn2) {
1703                         best_lsn = best_lsn1;
1704                         best_page = first_tail;
1705                         this_off = first_file_off;
1706                 } else {
1707                         best_lsn = best_lsn2;
1708                         best_page = second_tail;
1709                         this_off = second_file_off;
1710                 }
1711         } else if (first_tail) {
1712                 best_lsn = best_lsn1;
1713                 best_page = first_tail;
1714                 this_off = first_file_off;
1715         } else if (second_tail) {
1716                 best_lsn = best_lsn2;
1717                 best_page = second_tail;
1718                 this_off = second_file_off;
1719         } else {
1720                 goto tail_read;
1721         }
1722
1723         best_page_pos = le16_to_cpu(best_page->page_pos);
1724
1725         if (!tails) {
1726                 if (best_page_pos == page_pos) {
1727                         seq_base = best_lsn >> log->file_data_bits;
1728                         saved_off = page_off = le32_to_cpu(best_page->file_off);
1729                         lsn_base = best_lsn;
1730
1731                         memmove(page_bufs, best_page, log->page_size);
1732
1733                         page_cnt = le16_to_cpu(best_page->page_count);
1734                         if (page_cnt > 1)
1735                                 page_pos += 1;
1736
1737                         tails = 1;
1738                 }
1739         } else if (seq_base == (best_lsn >> log->file_data_bits) &&
1740                    saved_off + log->page_size == this_off &&
1741                    lsn_base < best_lsn &&
1742                    (page_pos != page_cnt || best_page_pos == page_pos ||
1743                     best_page_pos == 1) &&
1744                    (page_pos >= page_cnt || best_page_pos == page_pos)) {
1745                 u16 bppc = le16_to_cpu(best_page->page_count);
1746
1747                 saved_off += log->page_size;
1748                 lsn_base = best_lsn;
1749
1750                 memmove(Add2Ptr(page_bufs, tails * log->page_size), best_page,
1751                         log->page_size);
1752
1753                 tails += 1;
1754
1755                 if (best_page_pos != bppc) {
1756                         page_cnt = bppc;
1757                         page_pos = best_page_pos;
1758
1759                         if (page_cnt > 1)
1760                                 page_pos += 1;
1761                 } else {
1762                         page_pos = page_cnt = 1;
1763                 }
1764         } else {
1765                 kfree(first_tail);
1766                 kfree(second_tail);
1767                 goto tail_read;
1768         }
1769
1770         kfree(first_tail_prev);
1771         first_tail_prev = first_tail;
1772         final_off_prev = first_file_off;
1773         first_tail = NULL;
1774
1775         kfree(second_tail_prev);
1776         second_tail_prev = second_tail;
1777         second_off_prev = second_file_off;
1778         second_tail = NULL;
1779
1780         final_off += log->page_size;
1781         second_off += log->page_size;
1782
1783         if (tails < 0x10)
1784                 goto next_tail;
1785 tail_read:
1786         first_tail = first_tail_prev;
1787         final_off = final_off_prev;
1788
1789         second_tail = second_tail_prev;
1790         second_off = second_off_prev;
1791
1792         page_cnt = page_pos = 1;
1793
1794         curpage_off = seq_base == log->seq_num ? min(log->next_page, page_off)
1795                                                : log->next_page;
1796
1797         wrapped_file =
1798                 curpage_off == log->first_page &&
1799                 !(log->l_flags & (NTFSLOG_NO_LAST_LSN | NTFSLOG_REUSE_TAIL));
1800
1801         expected_seq = wrapped_file ? (log->seq_num + 1) : log->seq_num;
1802
1803         nextpage_off = curpage_off;
1804
1805 next_page:
1806         tail_page = NULL;
1807         /* Read the next log page. */
1808         err = read_log_page(log, curpage_off, &page, &usa_error);
1809
1810         /* Compute the next log page offset the file. */
1811         nextpage_off = next_page_off(log, curpage_off);
1812         wrapped = nextpage_off == log->first_page;
1813
1814         if (tails > 1) {
1815                 struct RECORD_PAGE_HDR *cur_page =
1816                         Add2Ptr(page_bufs, curpage_off - page_off);
1817
1818                 if (curpage_off == saved_off) {
1819                         tail_page = cur_page;
1820                         goto use_tail_page;
1821                 }
1822
1823                 if (page_off > curpage_off || curpage_off >= saved_off)
1824                         goto use_tail_page;
1825
1826                 if (page_off1)
1827                         goto use_cur_page;
1828
1829                 if (!err && !usa_error &&
1830                     page->rhdr.sign == NTFS_RCRD_SIGNATURE &&
1831                     cur_page->rhdr.lsn == page->rhdr.lsn &&
1832                     cur_page->record_hdr.next_record_off ==
1833                             page->record_hdr.next_record_off &&
1834                     ((page_pos == page_cnt &&
1835                       le16_to_cpu(page->page_pos) == 1) ||
1836                      (page_pos != page_cnt &&
1837                       le16_to_cpu(page->page_pos) == page_pos + 1 &&
1838                       le16_to_cpu(page->page_count) == page_cnt))) {
1839                         cur_page = NULL;
1840                         goto use_tail_page;
1841                 }
1842
1843                 page_off1 = page_off;
1844
1845 use_cur_page:
1846
1847                 lsn_cur = le64_to_cpu(cur_page->rhdr.lsn);
1848
1849                 if (last_ok_lsn !=
1850                             le64_to_cpu(cur_page->record_hdr.last_end_lsn) &&
1851                     ((lsn_cur >> log->file_data_bits) +
1852                      ((curpage_off <
1853                        (lsn_to_vbo(log, lsn_cur) & ~log->page_mask))
1854                               ? 1
1855                               : 0)) != expected_seq) {
1856                         goto check_tail;
1857                 }
1858
1859                 if (!is_log_record_end(cur_page)) {
1860                         tail_page = NULL;
1861                         last_ok_lsn = lsn_cur;
1862                         goto next_page_1;
1863                 }
1864
1865                 log->seq_num = expected_seq;
1866                 log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
1867                 log->last_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
1868                 log->ra->current_lsn = cur_page->record_hdr.last_end_lsn;
1869
1870                 if (log->record_header_len <=
1871                     log->page_size -
1872                             le16_to_cpu(cur_page->record_hdr.next_record_off)) {
1873                         log->l_flags |= NTFSLOG_REUSE_TAIL;
1874                         log->next_page = curpage_off;
1875                 } else {
1876                         log->l_flags &= ~NTFSLOG_REUSE_TAIL;
1877                         log->next_page = nextpage_off;
1878                 }
1879
1880                 if (wrapped_file)
1881                         log->l_flags |= NTFSLOG_WRAPPED;
1882
1883                 last_ok_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
1884                 goto next_page_1;
1885         }
1886
1887         /*
1888          * If we are at the expected first page of a transfer check to see
1889          * if either tail copy is at this offset.
1890          * If this page is the last page of a transfer, check if we wrote
1891          * a subsequent tail copy.
1892          */
1893         if (page_cnt == page_pos || page_cnt == page_pos + 1) {
1894                 /*
1895                  * Check if the offset matches either the first or second
1896                  * tail copy. It is possible it will match both.
1897                  */
1898                 if (curpage_off == final_off)
1899                         tail_page = first_tail;
1900
1901                 /*
1902                  * If we already matched on the first page then
1903                  * check the ending lsn's.
1904                  */
1905                 if (curpage_off == second_off) {
1906                         if (!tail_page ||
1907                             (second_tail &&
1908                              le64_to_cpu(second_tail->record_hdr.last_end_lsn) >
1909                                      le64_to_cpu(first_tail->record_hdr
1910                                                          .last_end_lsn))) {
1911                                 tail_page = second_tail;
1912                         }
1913                 }
1914         }
1915
1916 use_tail_page:
1917         if (tail_page) {
1918                 /* We have a candidate for a tail copy. */
1919                 lsn_cur = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
1920
1921                 if (last_ok_lsn < lsn_cur) {
1922                         /*
1923                          * If the sequence number is not expected,
1924                          * then don't use the tail copy.
1925                          */
1926                         if (expected_seq != (lsn_cur >> log->file_data_bits))
1927                                 tail_page = NULL;
1928                 } else if (last_ok_lsn > lsn_cur) {
1929                         /*
1930                          * If the last lsn is greater than the one on
1931                          * this page then forget this tail.
1932                          */
1933                         tail_page = NULL;
1934                 }
1935         }
1936
1937         /*
1938          *If we have an error on the current page,
1939          * we will break of this loop.
1940          */
1941         if (err || usa_error)
1942                 goto check_tail;
1943
1944         /*
1945          * Done if the last lsn on this page doesn't match the previous known
1946          * last lsn or the sequence number is not expected.
1947          */
1948         lsn_cur = le64_to_cpu(page->rhdr.lsn);
1949         if (last_ok_lsn != lsn_cur &&
1950             expected_seq != (lsn_cur >> log->file_data_bits)) {
1951                 goto check_tail;
1952         }
1953
1954         /*
1955          * Check that the page position and page count values are correct.
1956          * If this is the first page of a transfer the position must be 1
1957          * and the count will be unknown.
1958          */
1959         if (page_cnt == page_pos) {
1960                 if (page->page_pos != cpu_to_le16(1) &&
1961                     (!reuse_page || page->page_pos != page->page_count)) {
1962                         /*
1963                          * If the current page is the first page we are
1964                          * looking at and we are reusing this page then
1965                          * it can be either the first or last page of a
1966                          * transfer. Otherwise it can only be the first.
1967                          */
1968                         goto check_tail;
1969                 }
1970         } else if (le16_to_cpu(page->page_count) != page_cnt ||
1971                    le16_to_cpu(page->page_pos) != page_pos + 1) {
1972                 /*
1973                  * The page position better be 1 more than the last page
1974                  * position and the page count better match.
1975                  */
1976                 goto check_tail;
1977         }
1978
1979         /*
1980          * We have a valid page the file and may have a valid page
1981          * the tail copy area.
1982          * If the tail page was written after the page the file then
1983          * break of the loop.
1984          */
1985         if (tail_page &&
1986             le64_to_cpu(tail_page->record_hdr.last_end_lsn) > lsn_cur) {
1987                 /* Remember if we will replace the page. */
1988                 replace_page = true;
1989                 goto check_tail;
1990         }
1991
1992         tail_page = NULL;
1993
1994         if (is_log_record_end(page)) {
1995                 /*
1996                  * Since we have read this page we know the sequence number
1997                  * is the same as our expected value.
1998                  */
1999                 log->seq_num = expected_seq;
2000                 log->last_lsn = le64_to_cpu(page->record_hdr.last_end_lsn);
2001                 log->ra->current_lsn = page->record_hdr.last_end_lsn;
2002                 log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
2003
2004                 /*
2005                  * If there is room on this page for another header then
2006                  * remember we want to reuse the page.
2007                  */
2008                 if (log->record_header_len <=
2009                     log->page_size -
2010                             le16_to_cpu(page->record_hdr.next_record_off)) {
2011                         log->l_flags |= NTFSLOG_REUSE_TAIL;
2012                         log->next_page = curpage_off;
2013                 } else {
2014                         log->l_flags &= ~NTFSLOG_REUSE_TAIL;
2015                         log->next_page = nextpage_off;
2016                 }
2017
2018                 /* Remember if we wrapped the log file. */
2019                 if (wrapped_file)
2020                         log->l_flags |= NTFSLOG_WRAPPED;
2021         }
2022
2023         /*
2024          * Remember the last page count and position.
2025          * Also remember the last known lsn.
2026          */
2027         page_cnt = le16_to_cpu(page->page_count);
2028         page_pos = le16_to_cpu(page->page_pos);
2029         last_ok_lsn = le64_to_cpu(page->rhdr.lsn);
2030
2031 next_page_1:
2032
2033         if (wrapped) {
2034                 expected_seq += 1;
2035                 wrapped_file = 1;
2036         }
2037
2038         curpage_off = nextpage_off;
2039         kfree(page);
2040         page = NULL;
2041         reuse_page = 0;
2042         goto next_page;
2043
2044 check_tail:
2045         if (tail_page) {
2046                 log->seq_num = expected_seq;
2047                 log->last_lsn = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
2048                 log->ra->current_lsn = tail_page->record_hdr.last_end_lsn;
2049                 log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
2050
2051                 if (log->page_size -
2052                             le16_to_cpu(
2053                                     tail_page->record_hdr.next_record_off) >=
2054                     log->record_header_len) {
2055                         log->l_flags |= NTFSLOG_REUSE_TAIL;
2056                         log->next_page = curpage_off;
2057                 } else {
2058                         log->l_flags &= ~NTFSLOG_REUSE_TAIL;
2059                         log->next_page = nextpage_off;
2060                 }
2061
2062                 if (wrapped)
2063                         log->l_flags |= NTFSLOG_WRAPPED;
2064         }
2065
2066         /* Remember that the partial IO will start at the next page. */
2067         second_off = nextpage_off;
2068
2069         /*
2070          * If the next page is the first page of the file then update
2071          * the sequence number for log records which begon the next page.
2072          */
2073         if (wrapped)
2074                 expected_seq += 1;
2075
2076         /*
2077          * If we have a tail copy or are performing single page I/O we can
2078          * immediately look at the next page.
2079          */
2080         if (replace_page || (log->ra->flags & RESTART_SINGLE_PAGE_IO)) {
2081                 page_cnt = 2;
2082                 page_pos = 1;
2083                 goto check_valid;
2084         }
2085
2086         if (page_pos != page_cnt)
2087                 goto check_valid;
2088         /*
2089          * If the next page causes us to wrap to the beginning of the log
2090          * file then we know which page to check next.
2091          */
2092         if (wrapped) {
2093                 page_cnt = 2;
2094                 page_pos = 1;
2095                 goto check_valid;
2096         }
2097
2098         cur_pos = 2;
2099
2100 next_test_page:
2101         kfree(tst_page);
2102         tst_page = NULL;
2103
2104         /* Walk through the file, reading log pages. */
2105         err = read_log_page(log, nextpage_off, &tst_page, &usa_error);
2106
2107         /*
2108          * If we get a USA error then assume that we correctly found
2109          * the end of the original transfer.
2110          */
2111         if (usa_error)
2112                 goto file_is_valid;
2113
2114         /*
2115          * If we were able to read the page, we examine it to see if it
2116          * is the same or different Io block.
2117          */
2118         if (err)
2119                 goto next_test_page_1;
2120
2121         if (le16_to_cpu(tst_page->page_pos) == cur_pos &&
2122             check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
2123                 page_cnt = le16_to_cpu(tst_page->page_count) + 1;
2124                 page_pos = le16_to_cpu(tst_page->page_pos);
2125                 goto check_valid;
2126         } else {
2127                 goto file_is_valid;
2128         }
2129
2130 next_test_page_1:
2131
2132         nextpage_off = next_page_off(log, curpage_off);
2133         wrapped = nextpage_off == log->first_page;
2134
2135         if (wrapped) {
2136                 expected_seq += 1;
2137                 page_cnt = 2;
2138                 page_pos = 1;
2139         }
2140
2141         cur_pos += 1;
2142         part_io_count += 1;
2143         if (!wrapped)
2144                 goto next_test_page;
2145
2146 check_valid:
2147         /* Skip over the remaining pages this transfer. */
2148         remain_pages = page_cnt - page_pos - 1;
2149         part_io_count += remain_pages;
2150
2151         while (remain_pages--) {
2152                 nextpage_off = next_page_off(log, curpage_off);
2153                 wrapped = nextpage_off == log->first_page;
2154
2155                 if (wrapped)
2156                         expected_seq += 1;
2157         }
2158
2159         /* Call our routine to check this log page. */
2160         kfree(tst_page);
2161         tst_page = NULL;
2162
2163         err = read_log_page(log, nextpage_off, &tst_page, &usa_error);
2164         if (!err && !usa_error &&
2165             check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
2166                 err = -EINVAL;
2167                 goto out;
2168         }
2169
2170 file_is_valid:
2171
2172         /* We have a valid file. */
2173         if (page_off1 || tail_page) {
2174                 struct RECORD_PAGE_HDR *tmp_page;
2175
2176                 if (sb_rdonly(log->ni->mi.sbi->sb)) {
2177                         err = -EROFS;
2178                         goto out;
2179                 }
2180
2181                 if (page_off1) {
2182                         tmp_page = Add2Ptr(page_bufs, page_off1 - page_off);
2183                         tails -= (page_off1 - page_off) / log->page_size;
2184                         if (!tail_page)
2185                                 tails -= 1;
2186                 } else {
2187                         tmp_page = tail_page;
2188                         tails = 1;
2189                 }
2190
2191                 while (tails--) {
2192                         u64 off = hdr_file_off(log, tmp_page);
2193
2194                         if (!page) {
2195                                 page = kmalloc(log->page_size, GFP_NOFS);
2196                                 if (!page)
2197                                         return -ENOMEM;
2198                         }
2199
2200                         /*
2201                          * Correct page and copy the data from this page
2202                          * into it and flush it to disk.
2203                          */
2204                         memcpy(page, tmp_page, log->page_size);
2205
2206                         /* Fill last flushed lsn value flush the page. */
2207                         if (log->major_ver < 2)
2208                                 page->rhdr.lsn = page->record_hdr.last_end_lsn;
2209                         else
2210                                 page->file_off = 0;
2211
2212                         page->page_pos = page->page_count = cpu_to_le16(1);
2213
2214                         ntfs_fix_pre_write(&page->rhdr, log->page_size);
2215
2216                         err = ntfs_sb_write_run(log->ni->mi.sbi,
2217                                                 &log->ni->file.run, off, page,
2218                                                 log->page_size, 0);
2219
2220                         if (err)
2221                                 goto out;
2222
2223                         if (part_io_count && second_off == off) {
2224                                 second_off += log->page_size;
2225                                 part_io_count -= 1;
2226                         }
2227
2228                         tmp_page = Add2Ptr(tmp_page, log->page_size);
2229                 }
2230         }
2231
2232         if (part_io_count) {
2233                 if (sb_rdonly(log->ni->mi.sbi->sb)) {
2234                         err = -EROFS;
2235                         goto out;
2236                 }
2237         }
2238
2239 out:
2240         kfree(second_tail);
2241         kfree(first_tail);
2242         kfree(page);
2243         kfree(tst_page);
2244         kfree(page_bufs);
2245
2246         return err;
2247 }
2248
2249 /*
2250  * read_log_rec_buf - Copy a log record from the file to a buffer.
2251  *
2252  * The log record may span several log pages and may even wrap the file.
2253  */
2254 static int read_log_rec_buf(struct ntfs_log *log,
2255                             const struct LFS_RECORD_HDR *rh, void *buffer)
2256 {
2257         int err;
2258         struct RECORD_PAGE_HDR *ph = NULL;
2259         u64 lsn = le64_to_cpu(rh->this_lsn);
2260         u32 vbo = lsn_to_vbo(log, lsn) & ~log->page_mask;
2261         u32 off = lsn_to_page_off(log, lsn) + log->record_header_len;
2262         u32 data_len = le32_to_cpu(rh->client_data_len);
2263
2264         /*
2265          * While there are more bytes to transfer,
2266          * we continue to attempt to perform the read.
2267          */
2268         for (;;) {
2269                 bool usa_error;
2270                 u32 tail = log->page_size - off;
2271
2272                 if (tail >= data_len)
2273                         tail = data_len;
2274
2275                 data_len -= tail;
2276
2277                 err = read_log_page(log, vbo, &ph, &usa_error);
2278                 if (err)
2279                         goto out;
2280
2281                 /*
2282                  * The last lsn on this page better be greater or equal
2283                  * to the lsn we are copying.
2284                  */
2285                 if (lsn > le64_to_cpu(ph->rhdr.lsn)) {
2286                         err = -EINVAL;
2287                         goto out;
2288                 }
2289
2290                 memcpy(buffer, Add2Ptr(ph, off), tail);
2291
2292                 /* If there are no more bytes to transfer, we exit the loop. */
2293                 if (!data_len) {
2294                         if (!is_log_record_end(ph) ||
2295                             lsn > le64_to_cpu(ph->record_hdr.last_end_lsn)) {
2296                                 err = -EINVAL;
2297                                 goto out;
2298                         }
2299                         break;
2300                 }
2301
2302                 if (ph->rhdr.lsn == ph->record_hdr.last_end_lsn ||
2303                     lsn > le64_to_cpu(ph->rhdr.lsn)) {
2304                         err = -EINVAL;
2305                         goto out;
2306                 }
2307
2308                 vbo = next_page_off(log, vbo);
2309                 off = log->data_off;
2310
2311                 /*
2312                  * Adjust our pointer the user's buffer to transfer
2313                  * the next block to.
2314                  */
2315                 buffer = Add2Ptr(buffer, tail);
2316         }
2317
2318 out:
2319         kfree(ph);
2320         return err;
2321 }
2322
2323 static int read_rst_area(struct ntfs_log *log, struct NTFS_RESTART **rst_,
2324                          u64 *lsn)
2325 {
2326         int err;
2327         struct LFS_RECORD_HDR *rh = NULL;
2328         const struct CLIENT_REC *cr =
2329                 Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
2330         u64 lsnr, lsnc = le64_to_cpu(cr->restart_lsn);
2331         u32 len;
2332         struct NTFS_RESTART *rst;
2333
2334         *lsn = 0;
2335         *rst_ = NULL;
2336
2337         /* If the client doesn't have a restart area, go ahead and exit now. */
2338         if (!lsnc)
2339                 return 0;
2340
2341         err = read_log_page(log, lsn_to_vbo(log, lsnc),
2342                             (struct RECORD_PAGE_HDR **)&rh, NULL);
2343         if (err)
2344                 return err;
2345
2346         rst = NULL;
2347         lsnr = le64_to_cpu(rh->this_lsn);
2348
2349         if (lsnc != lsnr) {
2350                 /* If the lsn values don't match, then the disk is corrupt. */
2351                 err = -EINVAL;
2352                 goto out;
2353         }
2354
2355         *lsn = lsnr;
2356         len = le32_to_cpu(rh->client_data_len);
2357
2358         if (!len) {
2359                 err = 0;
2360                 goto out;
2361         }
2362
2363         if (len < sizeof(struct NTFS_RESTART)) {
2364                 err = -EINVAL;
2365                 goto out;
2366         }
2367
2368         rst = kmalloc(len, GFP_NOFS);
2369         if (!rst) {
2370                 err = -ENOMEM;
2371                 goto out;
2372         }
2373
2374         /* Copy the data into the 'rst' buffer. */
2375         err = read_log_rec_buf(log, rh, rst);
2376         if (err)
2377                 goto out;
2378
2379         *rst_ = rst;
2380         rst = NULL;
2381
2382 out:
2383         kfree(rh);
2384         kfree(rst);
2385
2386         return err;
2387 }
2388
2389 static int find_log_rec(struct ntfs_log *log, u64 lsn, struct lcb *lcb)
2390 {
2391         int err;
2392         struct LFS_RECORD_HDR *rh = lcb->lrh;
2393         u32 rec_len, len;
2394
2395         /* Read the record header for this lsn. */
2396         if (!rh) {
2397                 err = read_log_page(log, lsn_to_vbo(log, lsn),
2398                                     (struct RECORD_PAGE_HDR **)&rh, NULL);
2399
2400                 lcb->lrh = rh;
2401                 if (err)
2402                         return err;
2403         }
2404
2405         /*
2406          * If the lsn the log record doesn't match the desired
2407          * lsn then the disk is corrupt.
2408          */
2409         if (lsn != le64_to_cpu(rh->this_lsn))
2410                 return -EINVAL;
2411
2412         len = le32_to_cpu(rh->client_data_len);
2413
2414         /*
2415          * Check that the length field isn't greater than the total
2416          * available space the log file.
2417          */
2418         rec_len = len + log->record_header_len;
2419         if (rec_len >= log->total_avail)
2420                 return -EINVAL;
2421
2422         /*
2423          * If the entire log record is on this log page,
2424          * put a pointer to the log record the context block.
2425          */
2426         if (rh->flags & LOG_RECORD_MULTI_PAGE) {
2427                 void *lr = kmalloc(len, GFP_NOFS);
2428
2429                 if (!lr)
2430                         return -ENOMEM;
2431
2432                 lcb->log_rec = lr;
2433                 lcb->alloc = true;
2434
2435                 /* Copy the data into the buffer returned. */
2436                 err = read_log_rec_buf(log, rh, lr);
2437                 if (err)
2438                         return err;
2439         } else {
2440                 /* If beyond the end of the current page -> an error. */
2441                 u32 page_off = lsn_to_page_off(log, lsn);
2442
2443                 if (page_off + len + log->record_header_len > log->page_size)
2444                         return -EINVAL;
2445
2446                 lcb->log_rec = Add2Ptr(rh, sizeof(struct LFS_RECORD_HDR));
2447                 lcb->alloc = false;
2448         }
2449
2450         return 0;
2451 }
2452
2453 /*
2454  * read_log_rec_lcb - Init the query operation.
2455  */
2456 static int read_log_rec_lcb(struct ntfs_log *log, u64 lsn, u32 ctx_mode,
2457                             struct lcb **lcb_)
2458 {
2459         int err;
2460         const struct CLIENT_REC *cr;
2461         struct lcb *lcb;
2462
2463         switch (ctx_mode) {
2464         case lcb_ctx_undo_next:
2465         case lcb_ctx_prev:
2466         case lcb_ctx_next:
2467                 break;
2468         default:
2469                 return -EINVAL;
2470         }
2471
2472         /* Check that the given lsn is the legal range for this client. */
2473         cr = Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
2474
2475         if (!verify_client_lsn(log, cr, lsn))
2476                 return -EINVAL;
2477
2478         lcb = kzalloc(sizeof(struct lcb), GFP_NOFS);
2479         if (!lcb)
2480                 return -ENOMEM;
2481         lcb->client = log->client_id;
2482         lcb->ctx_mode = ctx_mode;
2483
2484         /* Find the log record indicated by the given lsn. */
2485         err = find_log_rec(log, lsn, lcb);
2486         if (err)
2487                 goto out;
2488
2489         *lcb_ = lcb;
2490         return 0;
2491
2492 out:
2493         lcb_put(lcb);
2494         *lcb_ = NULL;
2495         return err;
2496 }
2497
2498 /*
2499  * find_client_next_lsn
2500  *
2501  * Attempt to find the next lsn to return to a client based on the context mode.
2502  */
2503 static int find_client_next_lsn(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
2504 {
2505         int err;
2506         u64 next_lsn;
2507         struct LFS_RECORD_HDR *hdr;
2508
2509         hdr = lcb->lrh;
2510         *lsn = 0;
2511
2512         if (lcb_ctx_next != lcb->ctx_mode)
2513                 goto check_undo_next;
2514
2515         /* Loop as long as another lsn can be found. */
2516         for (;;) {
2517                 u64 current_lsn;
2518
2519                 err = next_log_lsn(log, hdr, &current_lsn);
2520                 if (err)
2521                         goto out;
2522
2523                 if (!current_lsn)
2524                         break;
2525
2526                 if (hdr != lcb->lrh)
2527                         kfree(hdr);
2528
2529                 hdr = NULL;
2530                 err = read_log_page(log, lsn_to_vbo(log, current_lsn),
2531                                     (struct RECORD_PAGE_HDR **)&hdr, NULL);
2532                 if (err)
2533                         goto out;
2534
2535                 if (memcmp(&hdr->client, &lcb->client,
2536                            sizeof(struct CLIENT_ID))) {
2537                         /*err = -EINVAL; */
2538                 } else if (LfsClientRecord == hdr->record_type) {
2539                         kfree(lcb->lrh);
2540                         lcb->lrh = hdr;
2541                         *lsn = current_lsn;
2542                         return 0;
2543                 }
2544         }
2545
2546 out:
2547         if (hdr != lcb->lrh)
2548                 kfree(hdr);
2549         return err;
2550
2551 check_undo_next:
2552         if (lcb_ctx_undo_next == lcb->ctx_mode)
2553                 next_lsn = le64_to_cpu(hdr->client_undo_next_lsn);
2554         else if (lcb_ctx_prev == lcb->ctx_mode)
2555                 next_lsn = le64_to_cpu(hdr->client_prev_lsn);
2556         else
2557                 return 0;
2558
2559         if (!next_lsn)
2560                 return 0;
2561
2562         if (!verify_client_lsn(
2563                     log, Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off)),
2564                     next_lsn))
2565                 return 0;
2566
2567         hdr = NULL;
2568         err = read_log_page(log, lsn_to_vbo(log, next_lsn),
2569                             (struct RECORD_PAGE_HDR **)&hdr, NULL);
2570         if (err)
2571                 return err;
2572         kfree(lcb->lrh);
2573         lcb->lrh = hdr;
2574
2575         *lsn = next_lsn;
2576
2577         return 0;
2578 }
2579
2580 static int read_next_log_rec(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
2581 {
2582         int err;
2583
2584         err = find_client_next_lsn(log, lcb, lsn);
2585         if (err)
2586                 return err;
2587
2588         if (!*lsn)
2589                 return 0;
2590
2591         if (lcb->alloc)
2592                 kfree(lcb->log_rec);
2593
2594         lcb->log_rec = NULL;
2595         lcb->alloc = false;
2596         kfree(lcb->lrh);
2597         lcb->lrh = NULL;
2598
2599         return find_log_rec(log, *lsn, lcb);
2600 }
2601
2602 static inline bool check_index_header(const struct INDEX_HDR *hdr, size_t bytes)
2603 {
2604         __le16 mask;
2605         u32 min_de, de_off, used, total;
2606         const struct NTFS_DE *e;
2607
2608         if (hdr_has_subnode(hdr)) {
2609                 min_de = sizeof(struct NTFS_DE) + sizeof(u64);
2610                 mask = NTFS_IE_HAS_SUBNODES;
2611         } else {
2612                 min_de = sizeof(struct NTFS_DE);
2613                 mask = 0;
2614         }
2615
2616         de_off = le32_to_cpu(hdr->de_off);
2617         used = le32_to_cpu(hdr->used);
2618         total = le32_to_cpu(hdr->total);
2619
2620         if (de_off > bytes - min_de || used > bytes || total > bytes ||
2621             de_off + min_de > used || used > total) {
2622                 return false;
2623         }
2624
2625         e = Add2Ptr(hdr, de_off);
2626         for (;;) {
2627                 u16 esize = le16_to_cpu(e->size);
2628                 struct NTFS_DE *next = Add2Ptr(e, esize);
2629
2630                 if (esize < min_de || PtrOffset(hdr, next) > used ||
2631                     (e->flags & NTFS_IE_HAS_SUBNODES) != mask) {
2632                         return false;
2633                 }
2634
2635                 if (de_is_last(e))
2636                         break;
2637
2638                 e = next;
2639         }
2640
2641         return true;
2642 }
2643
2644 static inline bool check_index_buffer(const struct INDEX_BUFFER *ib, u32 bytes)
2645 {
2646         u16 fo;
2647         const struct NTFS_RECORD_HEADER *r = &ib->rhdr;
2648
2649         if (r->sign != NTFS_INDX_SIGNATURE)
2650                 return false;
2651
2652         fo = (SECTOR_SIZE - ((bytes >> SECTOR_SHIFT) + 1) * sizeof(short));
2653
2654         if (le16_to_cpu(r->fix_off) > fo)
2655                 return false;
2656
2657         if ((le16_to_cpu(r->fix_num) - 1) * SECTOR_SIZE != bytes)
2658                 return false;
2659
2660         return check_index_header(&ib->ihdr,
2661                                   bytes - offsetof(struct INDEX_BUFFER, ihdr));
2662 }
2663
2664 static inline bool check_index_root(const struct ATTRIB *attr,
2665                                     struct ntfs_sb_info *sbi)
2666 {
2667         bool ret;
2668         const struct INDEX_ROOT *root = resident_data(attr);
2669         u8 index_bits = le32_to_cpu(root->index_block_size) >= sbi->cluster_size
2670                                 ? sbi->cluster_bits
2671                                 : SECTOR_SHIFT;
2672         u8 block_clst = root->index_block_clst;
2673
2674         if (le32_to_cpu(attr->res.data_size) < sizeof(struct INDEX_ROOT) ||
2675             (root->type != ATTR_NAME && root->type != ATTR_ZERO) ||
2676             (root->type == ATTR_NAME &&
2677              root->rule != NTFS_COLLATION_TYPE_FILENAME) ||
2678             (le32_to_cpu(root->index_block_size) !=
2679              (block_clst << index_bits)) ||
2680             (block_clst != 1 && block_clst != 2 && block_clst != 4 &&
2681              block_clst != 8 && block_clst != 0x10 && block_clst != 0x20 &&
2682              block_clst != 0x40 && block_clst != 0x80)) {
2683                 return false;
2684         }
2685
2686         ret = check_index_header(&root->ihdr,
2687                                  le32_to_cpu(attr->res.data_size) -
2688                                          offsetof(struct INDEX_ROOT, ihdr));
2689         return ret;
2690 }
2691
2692 static inline bool check_attr(const struct MFT_REC *rec,
2693                               const struct ATTRIB *attr,
2694                               struct ntfs_sb_info *sbi)
2695 {
2696         u32 asize = le32_to_cpu(attr->size);
2697         u32 rsize = 0;
2698         u64 dsize, svcn, evcn;
2699         u16 run_off;
2700
2701         /* Check the fixed part of the attribute record header. */
2702         if (asize >= sbi->record_size ||
2703             asize + PtrOffset(rec, attr) >= sbi->record_size ||
2704             (attr->name_len &&
2705              le16_to_cpu(attr->name_off) + attr->name_len * sizeof(short) >
2706                      asize)) {
2707                 return false;
2708         }
2709
2710         /* Check the attribute fields. */
2711         switch (attr->non_res) {
2712         case 0:
2713                 rsize = le32_to_cpu(attr->res.data_size);
2714                 if (rsize >= asize ||
2715                     le16_to_cpu(attr->res.data_off) + rsize > asize) {
2716                         return false;
2717                 }
2718                 break;
2719
2720         case 1:
2721                 dsize = le64_to_cpu(attr->nres.data_size);
2722                 svcn = le64_to_cpu(attr->nres.svcn);
2723                 evcn = le64_to_cpu(attr->nres.evcn);
2724                 run_off = le16_to_cpu(attr->nres.run_off);
2725
2726                 if (svcn > evcn + 1 || run_off >= asize ||
2727                     le64_to_cpu(attr->nres.valid_size) > dsize ||
2728                     dsize > le64_to_cpu(attr->nres.alloc_size)) {
2729                         return false;
2730                 }
2731
2732                 if (run_unpack(NULL, sbi, 0, svcn, evcn, svcn,
2733                                Add2Ptr(attr, run_off), asize - run_off) < 0) {
2734                         return false;
2735                 }
2736
2737                 return true;
2738
2739         default:
2740                 return false;
2741         }
2742
2743         switch (attr->type) {
2744         case ATTR_NAME:
2745                 if (fname_full_size(Add2Ptr(
2746                             attr, le16_to_cpu(attr->res.data_off))) > asize) {
2747                         return false;
2748                 }
2749                 break;
2750
2751         case ATTR_ROOT:
2752                 return check_index_root(attr, sbi);
2753
2754         case ATTR_STD:
2755                 if (rsize < sizeof(struct ATTR_STD_INFO5) &&
2756                     rsize != sizeof(struct ATTR_STD_INFO)) {
2757                         return false;
2758                 }
2759                 break;
2760
2761         case ATTR_LIST:
2762         case ATTR_ID:
2763         case ATTR_SECURE:
2764         case ATTR_LABEL:
2765         case ATTR_VOL_INFO:
2766         case ATTR_DATA:
2767         case ATTR_ALLOC:
2768         case ATTR_BITMAP:
2769         case ATTR_REPARSE:
2770         case ATTR_EA_INFO:
2771         case ATTR_EA:
2772         case ATTR_PROPERTYSET:
2773         case ATTR_LOGGED_UTILITY_STREAM:
2774                 break;
2775
2776         default:
2777                 return false;
2778         }
2779
2780         return true;
2781 }
2782
2783 static inline bool check_file_record(const struct MFT_REC *rec,
2784                                      const struct MFT_REC *rec2,
2785                                      struct ntfs_sb_info *sbi)
2786 {
2787         const struct ATTRIB *attr;
2788         u16 fo = le16_to_cpu(rec->rhdr.fix_off);
2789         u16 fn = le16_to_cpu(rec->rhdr.fix_num);
2790         u16 ao = le16_to_cpu(rec->attr_off);
2791         u32 rs = sbi->record_size;
2792
2793         /* Check the file record header for consistency. */
2794         if (rec->rhdr.sign != NTFS_FILE_SIGNATURE ||
2795             fo > (SECTOR_SIZE - ((rs >> SECTOR_SHIFT) + 1) * sizeof(short)) ||
2796             (fn - 1) * SECTOR_SIZE != rs || ao < MFTRECORD_FIXUP_OFFSET_1 ||
2797             ao > sbi->record_size - SIZEOF_RESIDENT || !is_rec_inuse(rec) ||
2798             le32_to_cpu(rec->total) != rs) {
2799                 return false;
2800         }
2801
2802         /* Loop to check all of the attributes. */
2803         for (attr = Add2Ptr(rec, ao); attr->type != ATTR_END;
2804              attr = Add2Ptr(attr, le32_to_cpu(attr->size))) {
2805                 if (check_attr(rec, attr, sbi))
2806                         continue;
2807                 return false;
2808         }
2809
2810         return true;
2811 }
2812
2813 static inline int check_lsn(const struct NTFS_RECORD_HEADER *hdr,
2814                             const u64 *rlsn)
2815 {
2816         u64 lsn;
2817
2818         if (!rlsn)
2819                 return true;
2820
2821         lsn = le64_to_cpu(hdr->lsn);
2822
2823         if (hdr->sign == NTFS_HOLE_SIGNATURE)
2824                 return false;
2825
2826         if (*rlsn > lsn)
2827                 return true;
2828
2829         return false;
2830 }
2831
2832 static inline bool check_if_attr(const struct MFT_REC *rec,
2833                                  const struct LOG_REC_HDR *lrh)
2834 {
2835         u16 ro = le16_to_cpu(lrh->record_off);
2836         u16 o = le16_to_cpu(rec->attr_off);
2837         const struct ATTRIB *attr = Add2Ptr(rec, o);
2838
2839         while (o < ro) {
2840                 u32 asize;
2841
2842                 if (attr->type == ATTR_END)
2843                         break;
2844
2845                 asize = le32_to_cpu(attr->size);
2846                 if (!asize)
2847                         break;
2848
2849                 o += asize;
2850                 attr = Add2Ptr(attr, asize);
2851         }
2852
2853         return o == ro;
2854 }
2855
2856 static inline bool check_if_index_root(const struct MFT_REC *rec,
2857                                        const struct LOG_REC_HDR *lrh)
2858 {
2859         u16 ro = le16_to_cpu(lrh->record_off);
2860         u16 o = le16_to_cpu(rec->attr_off);
2861         const struct ATTRIB *attr = Add2Ptr(rec, o);
2862
2863         while (o < ro) {
2864                 u32 asize;
2865
2866                 if (attr->type == ATTR_END)
2867                         break;
2868
2869                 asize = le32_to_cpu(attr->size);
2870                 if (!asize)
2871                         break;
2872
2873                 o += asize;
2874                 attr = Add2Ptr(attr, asize);
2875         }
2876
2877         return o == ro && attr->type == ATTR_ROOT;
2878 }
2879
2880 static inline bool check_if_root_index(const struct ATTRIB *attr,
2881                                        const struct INDEX_HDR *hdr,
2882                                        const struct LOG_REC_HDR *lrh)
2883 {
2884         u16 ao = le16_to_cpu(lrh->attr_off);
2885         u32 de_off = le32_to_cpu(hdr->de_off);
2886         u32 o = PtrOffset(attr, hdr) + de_off;
2887         const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
2888         u32 asize = le32_to_cpu(attr->size);
2889
2890         while (o < ao) {
2891                 u16 esize;
2892
2893                 if (o >= asize)
2894                         break;
2895
2896                 esize = le16_to_cpu(e->size);
2897                 if (!esize)
2898                         break;
2899
2900                 o += esize;
2901                 e = Add2Ptr(e, esize);
2902         }
2903
2904         return o == ao;
2905 }
2906
2907 static inline bool check_if_alloc_index(const struct INDEX_HDR *hdr,
2908                                         u32 attr_off)
2909 {
2910         u32 de_off = le32_to_cpu(hdr->de_off);
2911         u32 o = offsetof(struct INDEX_BUFFER, ihdr) + de_off;
2912         const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
2913         u32 used = le32_to_cpu(hdr->used);
2914
2915         while (o < attr_off) {
2916                 u16 esize;
2917
2918                 if (de_off >= used)
2919                         break;
2920
2921                 esize = le16_to_cpu(e->size);
2922                 if (!esize)
2923                         break;
2924
2925                 o += esize;
2926                 de_off += esize;
2927                 e = Add2Ptr(e, esize);
2928         }
2929
2930         return o == attr_off;
2931 }
2932
2933 static inline void change_attr_size(struct MFT_REC *rec, struct ATTRIB *attr,
2934                                     u32 nsize)
2935 {
2936         u32 asize = le32_to_cpu(attr->size);
2937         int dsize = nsize - asize;
2938         u8 *next = Add2Ptr(attr, asize);
2939         u32 used = le32_to_cpu(rec->used);
2940
2941         memmove(Add2Ptr(attr, nsize), next, used - PtrOffset(rec, next));
2942
2943         rec->used = cpu_to_le32(used + dsize);
2944         attr->size = cpu_to_le32(nsize);
2945 }
2946
2947 struct OpenAttr {
2948         struct ATTRIB *attr;
2949         struct runs_tree *run1;
2950         struct runs_tree run0;
2951         struct ntfs_inode *ni;
2952         // CLST rno;
2953 };
2954
2955 /*
2956  * cmp_type_and_name
2957  *
2958  * Return: 0 if 'attr' has the same type and name.
2959  */
2960 static inline int cmp_type_and_name(const struct ATTRIB *a1,
2961                                     const struct ATTRIB *a2)
2962 {
2963         return a1->type != a2->type || a1->name_len != a2->name_len ||
2964                (a1->name_len && memcmp(attr_name(a1), attr_name(a2),
2965                                        a1->name_len * sizeof(short)));
2966 }
2967
2968 static struct OpenAttr *find_loaded_attr(struct ntfs_log *log,
2969                                          const struct ATTRIB *attr, CLST rno)
2970 {
2971         struct OPEN_ATTR_ENRTY *oe = NULL;
2972
2973         while ((oe = enum_rstbl(log->open_attr_tbl, oe))) {
2974                 struct OpenAttr *op_attr;
2975
2976                 if (ino_get(&oe->ref) != rno)
2977                         continue;
2978
2979                 op_attr = (struct OpenAttr *)oe->ptr;
2980                 if (!cmp_type_and_name(op_attr->attr, attr))
2981                         return op_attr;
2982         }
2983         return NULL;
2984 }
2985
2986 static struct ATTRIB *attr_create_nonres_log(struct ntfs_sb_info *sbi,
2987                                              enum ATTR_TYPE type, u64 size,
2988                                              const u16 *name, size_t name_len,
2989                                              __le16 flags)
2990 {
2991         struct ATTRIB *attr;
2992         u32 name_size = ALIGN(name_len * sizeof(short), 8);
2993         bool is_ext = flags & (ATTR_FLAG_COMPRESSED | ATTR_FLAG_SPARSED);
2994         u32 asize = name_size +
2995                     (is_ext ? SIZEOF_NONRESIDENT_EX : SIZEOF_NONRESIDENT);
2996
2997         attr = kzalloc(asize, GFP_NOFS);
2998         if (!attr)
2999                 return NULL;
3000
3001         attr->type = type;
3002         attr->size = cpu_to_le32(asize);
3003         attr->flags = flags;
3004         attr->non_res = 1;
3005         attr->name_len = name_len;
3006
3007         attr->nres.evcn = cpu_to_le64((u64)bytes_to_cluster(sbi, size) - 1);
3008         attr->nres.alloc_size = cpu_to_le64(ntfs_up_cluster(sbi, size));
3009         attr->nres.data_size = cpu_to_le64(size);
3010         attr->nres.valid_size = attr->nres.data_size;
3011         if (is_ext) {
3012                 attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
3013                 if (is_attr_compressed(attr))
3014                         attr->nres.c_unit = COMPRESSION_UNIT;
3015
3016                 attr->nres.run_off =
3017                         cpu_to_le16(SIZEOF_NONRESIDENT_EX + name_size);
3018                 memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT_EX), name,
3019                        name_len * sizeof(short));
3020         } else {
3021                 attr->name_off = SIZEOF_NONRESIDENT_LE;
3022                 attr->nres.run_off =
3023                         cpu_to_le16(SIZEOF_NONRESIDENT + name_size);
3024                 memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT), name,
3025                        name_len * sizeof(short));
3026         }
3027
3028         return attr;
3029 }
3030
3031 /*
3032  * do_action - Common routine for the Redo and Undo Passes.
3033  * @rlsn: If it is NULL then undo.
3034  */
3035 static int do_action(struct ntfs_log *log, struct OPEN_ATTR_ENRTY *oe,
3036                      const struct LOG_REC_HDR *lrh, u32 op, void *data,
3037                      u32 dlen, u32 rec_len, const u64 *rlsn)
3038 {
3039         int err = 0;
3040         struct ntfs_sb_info *sbi = log->ni->mi.sbi;
3041         struct inode *inode = NULL, *inode_parent;
3042         struct mft_inode *mi = NULL, *mi2_child = NULL;
3043         CLST rno = 0, rno_base = 0;
3044         struct INDEX_BUFFER *ib = NULL;
3045         struct MFT_REC *rec = NULL;
3046         struct ATTRIB *attr = NULL, *attr2;
3047         struct INDEX_HDR *hdr;
3048         struct INDEX_ROOT *root;
3049         struct NTFS_DE *e, *e1, *e2;
3050         struct NEW_ATTRIBUTE_SIZES *new_sz;
3051         struct ATTR_FILE_NAME *fname;
3052         struct OpenAttr *oa, *oa2;
3053         u32 nsize, t32, asize, used, esize, bmp_off, bmp_bits;
3054         u16 id, id2;
3055         u32 record_size = sbi->record_size;
3056         u64 t64;
3057         u16 roff = le16_to_cpu(lrh->record_off);
3058         u16 aoff = le16_to_cpu(lrh->attr_off);
3059         u64 lco = 0;
3060         u64 cbo = (u64)le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
3061         u64 tvo = le64_to_cpu(lrh->target_vcn) << sbi->cluster_bits;
3062         u64 vbo = cbo + tvo;
3063         void *buffer_le = NULL;
3064         u32 bytes = 0;
3065         bool a_dirty = false;
3066         u16 data_off;
3067
3068         oa = oe->ptr;
3069
3070         /* Big switch to prepare. */
3071         switch (op) {
3072         /* ============================================================
3073          * Process MFT records, as described by the current log record.
3074          * ============================================================
3075          */
3076         case InitializeFileRecordSegment:
3077         case DeallocateFileRecordSegment:
3078         case WriteEndOfFileRecordSegment:
3079         case CreateAttribute:
3080         case DeleteAttribute:
3081         case UpdateResidentValue:
3082         case UpdateMappingPairs:
3083         case SetNewAttributeSizes:
3084         case AddIndexEntryRoot:
3085         case DeleteIndexEntryRoot:
3086         case SetIndexEntryVcnRoot:
3087         case UpdateFileNameRoot:
3088         case UpdateRecordDataRoot:
3089         case ZeroEndOfFileRecord:
3090                 rno = vbo >> sbi->record_bits;
3091                 inode = ilookup(sbi->sb, rno);
3092                 if (inode) {
3093                         mi = &ntfs_i(inode)->mi;
3094                 } else if (op == InitializeFileRecordSegment) {
3095                         mi = kzalloc(sizeof(struct mft_inode), GFP_NOFS);
3096                         if (!mi)
3097                                 return -ENOMEM;
3098                         err = mi_format_new(mi, sbi, rno, 0, false);
3099                         if (err)
3100                                 goto out;
3101                 } else {
3102                         /* Read from disk. */
3103                         err = mi_get(sbi, rno, &mi);
3104                         if (err)
3105                                 return err;
3106                 }
3107                 rec = mi->mrec;
3108
3109                 if (op == DeallocateFileRecordSegment)
3110                         goto skip_load_parent;
3111
3112                 if (InitializeFileRecordSegment != op) {
3113                         if (rec->rhdr.sign == NTFS_BAAD_SIGNATURE)
3114                                 goto dirty_vol;
3115                         if (!check_lsn(&rec->rhdr, rlsn))
3116                                 goto out;
3117                         if (!check_file_record(rec, NULL, sbi))
3118                                 goto dirty_vol;
3119                         attr = Add2Ptr(rec, roff);
3120                 }
3121
3122                 if (is_rec_base(rec) || InitializeFileRecordSegment == op) {
3123                         rno_base = rno;
3124                         goto skip_load_parent;
3125                 }
3126
3127                 rno_base = ino_get(&rec->parent_ref);
3128                 inode_parent = ntfs_iget5(sbi->sb, &rec->parent_ref, NULL);
3129                 if (IS_ERR(inode_parent))
3130                         goto skip_load_parent;
3131
3132                 if (is_bad_inode(inode_parent)) {
3133                         iput(inode_parent);
3134                         goto skip_load_parent;
3135                 }
3136
3137                 if (ni_load_mi_ex(ntfs_i(inode_parent), rno, &mi2_child)) {
3138                         iput(inode_parent);
3139                 } else {
3140                         if (mi2_child->mrec != mi->mrec)
3141                                 memcpy(mi2_child->mrec, mi->mrec,
3142                                        sbi->record_size);
3143
3144                         if (inode)
3145                                 iput(inode);
3146                         else if (mi)
3147                                 mi_put(mi);
3148
3149                         inode = inode_parent;
3150                         mi = mi2_child;
3151                         rec = mi2_child->mrec;
3152                         attr = Add2Ptr(rec, roff);
3153                 }
3154
3155 skip_load_parent:
3156                 inode_parent = NULL;
3157                 break;
3158
3159         /*
3160          * Process attributes, as described by the current log record.
3161          */
3162         case UpdateNonresidentValue:
3163         case AddIndexEntryAllocation:
3164         case DeleteIndexEntryAllocation:
3165         case WriteEndOfIndexBuffer:
3166         case SetIndexEntryVcnAllocation:
3167         case UpdateFileNameAllocation:
3168         case SetBitsInNonresidentBitMap:
3169         case ClearBitsInNonresidentBitMap:
3170         case UpdateRecordDataAllocation:
3171                 attr = oa->attr;
3172                 bytes = UpdateNonresidentValue == op ? dlen : 0;
3173                 lco = (u64)le16_to_cpu(lrh->lcns_follow) << sbi->cluster_bits;
3174
3175                 if (attr->type == ATTR_ALLOC) {
3176                         t32 = le32_to_cpu(oe->bytes_per_index);
3177                         if (bytes < t32)
3178                                 bytes = t32;
3179                 }
3180
3181                 if (!bytes)
3182                         bytes = lco - cbo;
3183
3184                 bytes += roff;
3185                 if (attr->type == ATTR_ALLOC)
3186                         bytes = (bytes + 511) & ~511; // align
3187
3188                 buffer_le = kmalloc(bytes, GFP_NOFS);
3189                 if (!buffer_le)
3190                         return -ENOMEM;
3191
3192                 err = ntfs_read_run_nb(sbi, oa->run1, vbo, buffer_le, bytes,
3193                                        NULL);
3194                 if (err)
3195                         goto out;
3196
3197                 if (attr->type == ATTR_ALLOC && *(int *)buffer_le)
3198                         ntfs_fix_post_read(buffer_le, bytes, false);
3199                 break;
3200
3201         default:
3202                 WARN_ON(1);
3203         }
3204
3205         /* Big switch to do operation. */
3206         switch (op) {
3207         case InitializeFileRecordSegment:
3208                 if (roff + dlen > record_size)
3209                         goto dirty_vol;
3210
3211                 memcpy(Add2Ptr(rec, roff), data, dlen);
3212                 mi->dirty = true;
3213                 break;
3214
3215         case DeallocateFileRecordSegment:
3216                 clear_rec_inuse(rec);
3217                 le16_add_cpu(&rec->seq, 1);
3218                 mi->dirty = true;
3219                 break;
3220
3221         case WriteEndOfFileRecordSegment:
3222                 attr2 = (struct ATTRIB *)data;
3223                 if (!check_if_attr(rec, lrh) || roff + dlen > record_size)
3224                         goto dirty_vol;
3225
3226                 memmove(attr, attr2, dlen);
3227                 rec->used = cpu_to_le32(ALIGN(roff + dlen, 8));
3228
3229                 mi->dirty = true;
3230                 break;
3231
3232         case CreateAttribute:
3233                 attr2 = (struct ATTRIB *)data;
3234                 asize = le32_to_cpu(attr2->size);
3235                 used = le32_to_cpu(rec->used);
3236
3237                 if (!check_if_attr(rec, lrh) || dlen < SIZEOF_RESIDENT ||
3238                     !IS_ALIGNED(asize, 8) ||
3239                     Add2Ptr(attr2, asize) > Add2Ptr(lrh, rec_len) ||
3240                     dlen > record_size - used) {
3241                         goto dirty_vol;
3242                 }
3243
3244                 memmove(Add2Ptr(attr, asize), attr, used - roff);
3245                 memcpy(attr, attr2, asize);
3246
3247                 rec->used = cpu_to_le32(used + asize);
3248                 id = le16_to_cpu(rec->next_attr_id);
3249                 id2 = le16_to_cpu(attr2->id);
3250                 if (id <= id2)
3251                         rec->next_attr_id = cpu_to_le16(id2 + 1);
3252                 if (is_attr_indexed(attr))
3253                         le16_add_cpu(&rec->hard_links, 1);
3254
3255                 oa2 = find_loaded_attr(log, attr, rno_base);
3256                 if (oa2) {
3257                         void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3258                                            GFP_NOFS);
3259                         if (p2) {
3260                                 // run_close(oa2->run1);
3261                                 kfree(oa2->attr);
3262                                 oa2->attr = p2;
3263                         }
3264                 }
3265
3266                 mi->dirty = true;
3267                 break;
3268
3269         case DeleteAttribute:
3270                 asize = le32_to_cpu(attr->size);
3271                 used = le32_to_cpu(rec->used);
3272
3273                 if (!check_if_attr(rec, lrh))
3274                         goto dirty_vol;
3275
3276                 rec->used = cpu_to_le32(used - asize);
3277                 if (is_attr_indexed(attr))
3278                         le16_add_cpu(&rec->hard_links, -1);
3279
3280                 memmove(attr, Add2Ptr(attr, asize), used - asize - roff);
3281
3282                 mi->dirty = true;
3283                 break;
3284
3285         case UpdateResidentValue:
3286                 nsize = aoff + dlen;
3287
3288                 if (!check_if_attr(rec, lrh))
3289                         goto dirty_vol;
3290
3291                 asize = le32_to_cpu(attr->size);
3292                 used = le32_to_cpu(rec->used);
3293
3294                 if (lrh->redo_len == lrh->undo_len) {
3295                         if (nsize > asize)
3296                                 goto dirty_vol;
3297                         goto move_data;
3298                 }
3299
3300                 if (nsize > asize && nsize - asize > record_size - used)
3301                         goto dirty_vol;
3302
3303                 nsize = ALIGN(nsize, 8);
3304                 data_off = le16_to_cpu(attr->res.data_off);
3305
3306                 if (nsize < asize) {
3307                         memmove(Add2Ptr(attr, aoff), data, dlen);
3308                         data = NULL; // To skip below memmove().
3309                 }
3310
3311                 memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
3312                         used - le16_to_cpu(lrh->record_off) - asize);
3313
3314                 rec->used = cpu_to_le32(used + nsize - asize);
3315                 attr->size = cpu_to_le32(nsize);
3316                 attr->res.data_size = cpu_to_le32(aoff + dlen - data_off);
3317
3318 move_data:
3319                 if (data)
3320                         memmove(Add2Ptr(attr, aoff), data, dlen);
3321
3322                 oa2 = find_loaded_attr(log, attr, rno_base);
3323                 if (oa2) {
3324                         void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3325                                            GFP_NOFS);
3326                         if (p2) {
3327                                 // run_close(&oa2->run0);
3328                                 oa2->run1 = &oa2->run0;
3329                                 kfree(oa2->attr);
3330                                 oa2->attr = p2;
3331                         }
3332                 }
3333
3334                 mi->dirty = true;
3335                 break;
3336
3337         case UpdateMappingPairs:
3338                 nsize = aoff + dlen;
3339                 asize = le32_to_cpu(attr->size);
3340                 used = le32_to_cpu(rec->used);
3341
3342                 if (!check_if_attr(rec, lrh) || !attr->non_res ||
3343                     aoff < le16_to_cpu(attr->nres.run_off) || aoff > asize ||
3344                     (nsize > asize && nsize - asize > record_size - used)) {
3345                         goto dirty_vol;
3346                 }
3347
3348                 nsize = ALIGN(nsize, 8);
3349
3350                 memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
3351                         used - le16_to_cpu(lrh->record_off) - asize);
3352                 rec->used = cpu_to_le32(used + nsize - asize);
3353                 attr->size = cpu_to_le32(nsize);
3354                 memmove(Add2Ptr(attr, aoff), data, dlen);
3355
3356                 if (run_get_highest_vcn(le64_to_cpu(attr->nres.svcn),
3357                                         attr_run(attr), &t64)) {
3358                         goto dirty_vol;
3359                 }
3360
3361                 attr->nres.evcn = cpu_to_le64(t64);
3362                 oa2 = find_loaded_attr(log, attr, rno_base);
3363                 if (oa2 && oa2->attr->non_res)
3364                         oa2->attr->nres.evcn = attr->nres.evcn;
3365
3366                 mi->dirty = true;
3367                 break;
3368
3369         case SetNewAttributeSizes:
3370                 new_sz = data;
3371                 if (!check_if_attr(rec, lrh) || !attr->non_res)
3372                         goto dirty_vol;
3373
3374                 attr->nres.alloc_size = new_sz->alloc_size;
3375                 attr->nres.data_size = new_sz->data_size;
3376                 attr->nres.valid_size = new_sz->valid_size;
3377
3378                 if (dlen >= sizeof(struct NEW_ATTRIBUTE_SIZES))
3379                         attr->nres.total_size = new_sz->total_size;
3380
3381                 oa2 = find_loaded_attr(log, attr, rno_base);
3382                 if (oa2) {
3383                         void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3384                                            GFP_NOFS);
3385                         if (p2) {
3386                                 kfree(oa2->attr);
3387                                 oa2->attr = p2;
3388                         }
3389                 }
3390                 mi->dirty = true;
3391                 break;
3392
3393         case AddIndexEntryRoot:
3394                 e = (struct NTFS_DE *)data;
3395                 esize = le16_to_cpu(e->size);
3396                 root = resident_data(attr);
3397                 hdr = &root->ihdr;
3398                 used = le32_to_cpu(hdr->used);
3399
3400                 if (!check_if_index_root(rec, lrh) ||
3401                     !check_if_root_index(attr, hdr, lrh) ||
3402                     Add2Ptr(data, esize) > Add2Ptr(lrh, rec_len) ||
3403                     esize > le32_to_cpu(rec->total) - le32_to_cpu(rec->used)) {
3404                         goto dirty_vol;
3405                 }
3406
3407                 e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3408
3409                 change_attr_size(rec, attr, le32_to_cpu(attr->size) + esize);
3410
3411                 memmove(Add2Ptr(e1, esize), e1,
3412                         PtrOffset(e1, Add2Ptr(hdr, used)));
3413                 memmove(e1, e, esize);
3414
3415                 le32_add_cpu(&attr->res.data_size, esize);
3416                 hdr->used = cpu_to_le32(used + esize);
3417                 le32_add_cpu(&hdr->total, esize);
3418
3419                 mi->dirty = true;
3420                 break;
3421
3422         case DeleteIndexEntryRoot:
3423                 root = resident_data(attr);
3424                 hdr = &root->ihdr;
3425                 used = le32_to_cpu(hdr->used);
3426
3427                 if (!check_if_index_root(rec, lrh) ||
3428                     !check_if_root_index(attr, hdr, lrh)) {
3429                         goto dirty_vol;
3430                 }
3431
3432                 e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3433                 esize = le16_to_cpu(e1->size);
3434                 e2 = Add2Ptr(e1, esize);
3435
3436                 memmove(e1, e2, PtrOffset(e2, Add2Ptr(hdr, used)));
3437
3438                 le32_sub_cpu(&attr->res.data_size, esize);
3439                 hdr->used = cpu_to_le32(used - esize);
3440                 le32_sub_cpu(&hdr->total, esize);
3441
3442                 change_attr_size(rec, attr, le32_to_cpu(attr->size) - esize);
3443
3444                 mi->dirty = true;
3445                 break;
3446
3447         case SetIndexEntryVcnRoot:
3448                 root = resident_data(attr);
3449                 hdr = &root->ihdr;
3450
3451                 if (!check_if_index_root(rec, lrh) ||
3452                     !check_if_root_index(attr, hdr, lrh)) {
3453                         goto dirty_vol;
3454                 }
3455
3456                 e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3457
3458                 de_set_vbn_le(e, *(__le64 *)data);
3459                 mi->dirty = true;
3460                 break;
3461
3462         case UpdateFileNameRoot:
3463                 root = resident_data(attr);
3464                 hdr = &root->ihdr;
3465
3466                 if (!check_if_index_root(rec, lrh) ||
3467                     !check_if_root_index(attr, hdr, lrh)) {
3468                         goto dirty_vol;
3469                 }
3470
3471                 e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3472                 fname = (struct ATTR_FILE_NAME *)(e + 1);
3473                 memmove(&fname->dup, data, sizeof(fname->dup)); //
3474                 mi->dirty = true;
3475                 break;
3476
3477         case UpdateRecordDataRoot:
3478                 root = resident_data(attr);
3479                 hdr = &root->ihdr;
3480
3481                 if (!check_if_index_root(rec, lrh) ||
3482                     !check_if_root_index(attr, hdr, lrh)) {
3483                         goto dirty_vol;
3484                 }
3485
3486                 e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3487
3488                 memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
3489
3490                 mi->dirty = true;
3491                 break;
3492
3493         case ZeroEndOfFileRecord:
3494                 if (roff + dlen > record_size)
3495                         goto dirty_vol;
3496
3497                 memset(attr, 0, dlen);
3498                 mi->dirty = true;
3499                 break;
3500
3501         case UpdateNonresidentValue:
3502                 if (lco < cbo + roff + dlen)
3503                         goto dirty_vol;
3504
3505                 memcpy(Add2Ptr(buffer_le, roff), data, dlen);
3506
3507                 a_dirty = true;
3508                 if (attr->type == ATTR_ALLOC)
3509                         ntfs_fix_pre_write(buffer_le, bytes);
3510                 break;
3511
3512         case AddIndexEntryAllocation:
3513                 ib = Add2Ptr(buffer_le, roff);
3514                 hdr = &ib->ihdr;
3515                 e = data;
3516                 esize = le16_to_cpu(e->size);
3517                 e1 = Add2Ptr(ib, aoff);
3518
3519                 if (is_baad(&ib->rhdr))
3520                         goto dirty_vol;
3521                 if (!check_lsn(&ib->rhdr, rlsn))
3522                         goto out;
3523
3524                 used = le32_to_cpu(hdr->used);
3525
3526                 if (!check_index_buffer(ib, bytes) ||
3527                     !check_if_alloc_index(hdr, aoff) ||
3528                     Add2Ptr(e, esize) > Add2Ptr(lrh, rec_len) ||
3529                     used + esize > le32_to_cpu(hdr->total)) {
3530                         goto dirty_vol;
3531                 }
3532
3533                 memmove(Add2Ptr(e1, esize), e1,
3534                         PtrOffset(e1, Add2Ptr(hdr, used)));
3535                 memcpy(e1, e, esize);
3536
3537                 hdr->used = cpu_to_le32(used + esize);
3538
3539                 a_dirty = true;
3540
3541                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3542                 break;
3543
3544         case DeleteIndexEntryAllocation:
3545                 ib = Add2Ptr(buffer_le, roff);
3546                 hdr = &ib->ihdr;
3547                 e = Add2Ptr(ib, aoff);
3548                 esize = le16_to_cpu(e->size);
3549
3550                 if (is_baad(&ib->rhdr))
3551                         goto dirty_vol;
3552                 if (!check_lsn(&ib->rhdr, rlsn))
3553                         goto out;
3554
3555                 if (!check_index_buffer(ib, bytes) ||
3556                     !check_if_alloc_index(hdr, aoff)) {
3557                         goto dirty_vol;
3558                 }
3559
3560                 e1 = Add2Ptr(e, esize);
3561                 nsize = esize;
3562                 used = le32_to_cpu(hdr->used);
3563
3564                 memmove(e, e1, PtrOffset(e1, Add2Ptr(hdr, used)));
3565
3566                 hdr->used = cpu_to_le32(used - nsize);
3567
3568                 a_dirty = true;
3569
3570                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3571                 break;
3572
3573         case WriteEndOfIndexBuffer:
3574                 ib = Add2Ptr(buffer_le, roff);
3575                 hdr = &ib->ihdr;
3576                 e = Add2Ptr(ib, aoff);
3577
3578                 if (is_baad(&ib->rhdr))
3579                         goto dirty_vol;
3580                 if (!check_lsn(&ib->rhdr, rlsn))
3581                         goto out;
3582                 if (!check_index_buffer(ib, bytes) ||
3583                     !check_if_alloc_index(hdr, aoff) ||
3584                     aoff + dlen > offsetof(struct INDEX_BUFFER, ihdr) +
3585                                           le32_to_cpu(hdr->total)) {
3586                         goto dirty_vol;
3587                 }
3588
3589                 hdr->used = cpu_to_le32(dlen + PtrOffset(hdr, e));
3590                 memmove(e, data, dlen);
3591
3592                 a_dirty = true;
3593                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3594                 break;
3595
3596         case SetIndexEntryVcnAllocation:
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                 de_set_vbn_le(e, *(__le64 *)data);
3612
3613                 a_dirty = true;
3614                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3615                 break;
3616
3617         case UpdateFileNameAllocation:
3618                 ib = Add2Ptr(buffer_le, roff);
3619                 hdr = &ib->ihdr;
3620                 e = Add2Ptr(ib, aoff);
3621
3622                 if (is_baad(&ib->rhdr))
3623                         goto dirty_vol;
3624
3625                 if (!check_lsn(&ib->rhdr, rlsn))
3626                         goto out;
3627                 if (!check_index_buffer(ib, bytes) ||
3628                     !check_if_alloc_index(hdr, aoff)) {
3629                         goto dirty_vol;
3630                 }
3631
3632                 fname = (struct ATTR_FILE_NAME *)(e + 1);
3633                 memmove(&fname->dup, data, sizeof(fname->dup));
3634
3635                 a_dirty = true;
3636                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3637                 break;
3638
3639         case SetBitsInNonresidentBitMap:
3640                 bmp_off =
3641                         le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
3642                 bmp_bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
3643
3644                 if (cbo + (bmp_off + 7) / 8 > lco ||
3645                     cbo + ((bmp_off + bmp_bits + 7) / 8) > lco) {
3646                         goto dirty_vol;
3647                 }
3648
3649                 __bitmap_set(Add2Ptr(buffer_le, roff), bmp_off, bmp_bits);
3650                 a_dirty = true;
3651                 break;
3652
3653         case ClearBitsInNonresidentBitMap:
3654                 bmp_off =
3655                         le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
3656                 bmp_bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
3657
3658                 if (cbo + (bmp_off + 7) / 8 > lco ||
3659                     cbo + ((bmp_off + bmp_bits + 7) / 8) > lco) {
3660                         goto dirty_vol;
3661                 }
3662
3663                 __bitmap_clear(Add2Ptr(buffer_le, roff), bmp_off, bmp_bits);
3664                 a_dirty = true;
3665                 break;
3666
3667         case UpdateRecordDataAllocation:
3668                 ib = Add2Ptr(buffer_le, roff);
3669                 hdr = &ib->ihdr;
3670                 e = Add2Ptr(ib, aoff);
3671
3672                 if (is_baad(&ib->rhdr))
3673                         goto dirty_vol;
3674
3675                 if (!check_lsn(&ib->rhdr, rlsn))
3676                         goto out;
3677                 if (!check_index_buffer(ib, bytes) ||
3678                     !check_if_alloc_index(hdr, aoff)) {
3679                         goto dirty_vol;
3680                 }
3681
3682                 memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
3683
3684                 a_dirty = true;
3685                 ntfs_fix_pre_write(&ib->rhdr, bytes);
3686                 break;
3687
3688         default:
3689                 WARN_ON(1);
3690         }
3691
3692         if (rlsn) {
3693                 __le64 t64 = cpu_to_le64(*rlsn);
3694
3695                 if (rec)
3696                         rec->rhdr.lsn = t64;
3697                 if (ib)
3698                         ib->rhdr.lsn = t64;
3699         }
3700
3701         if (mi && mi->dirty) {
3702                 err = mi_write(mi, 0);
3703                 if (err)
3704                         goto out;
3705         }
3706
3707         if (a_dirty) {
3708                 attr = oa->attr;
3709                 err = ntfs_sb_write_run(sbi, oa->run1, vbo, buffer_le, bytes, 0);
3710                 if (err)
3711                         goto out;
3712         }
3713
3714 out:
3715
3716         if (inode)
3717                 iput(inode);
3718         else if (mi != mi2_child)
3719                 mi_put(mi);
3720
3721         kfree(buffer_le);
3722
3723         return err;
3724
3725 dirty_vol:
3726         log->set_dirty = true;
3727         goto out;
3728 }
3729
3730 /*
3731  * log_replay - Replays log and empties it.
3732  *
3733  * This function is called during mount operation.
3734  * It replays log and empties it.
3735  * Initialized is set false if logfile contains '-1'.
3736  */
3737 int log_replay(struct ntfs_inode *ni, bool *initialized)
3738 {
3739         int err;
3740         struct ntfs_sb_info *sbi = ni->mi.sbi;
3741         struct ntfs_log *log;
3742
3743         struct restart_info rst_info, rst_info2;
3744         u64 rec_lsn, ra_lsn, checkpt_lsn = 0, rlsn = 0;
3745         struct ATTR_NAME_ENTRY *attr_names = NULL;
3746         struct ATTR_NAME_ENTRY *ane;
3747         struct RESTART_TABLE *dptbl = NULL;
3748         struct RESTART_TABLE *trtbl = NULL;
3749         const struct RESTART_TABLE *rt;
3750         struct RESTART_TABLE *oatbl = NULL;
3751         struct inode *inode;
3752         struct OpenAttr *oa;
3753         struct ntfs_inode *ni_oe;
3754         struct ATTRIB *attr = NULL;
3755         u64 size, vcn, undo_next_lsn;
3756         CLST rno, lcn, lcn0, len0, clen;
3757         void *data;
3758         struct NTFS_RESTART *rst = NULL;
3759         struct lcb *lcb = NULL;
3760         struct OPEN_ATTR_ENRTY *oe;
3761         struct TRANSACTION_ENTRY *tr;
3762         struct DIR_PAGE_ENTRY *dp;
3763         u32 i, bytes_per_attr_entry;
3764         u32 l_size = ni->vfs_inode.i_size;
3765         u32 orig_file_size = l_size;
3766         u32 page_size, vbo, tail, off, dlen;
3767         u32 saved_len, rec_len, transact_id;
3768         bool use_second_page;
3769         struct RESTART_AREA *ra2, *ra = NULL;
3770         struct CLIENT_REC *ca, *cr;
3771         __le16 client;
3772         struct RESTART_HDR *rh;
3773         const struct LFS_RECORD_HDR *frh;
3774         const struct LOG_REC_HDR *lrh;
3775         bool is_mapped;
3776         bool is_ro = sb_rdonly(sbi->sb);
3777         u64 t64;
3778         u16 t16;
3779         u32 t32;
3780
3781         /* Get the size of page. NOTE: To replay we can use default page. */
3782 #if PAGE_SIZE >= DefaultLogPageSize && PAGE_SIZE <= DefaultLogPageSize * 2
3783         page_size = norm_file_page(PAGE_SIZE, &l_size, true);
3784 #else
3785         page_size = norm_file_page(PAGE_SIZE, &l_size, false);
3786 #endif
3787         if (!page_size)
3788                 return -EINVAL;
3789
3790         log = kzalloc(sizeof(struct ntfs_log), GFP_NOFS);
3791         if (!log)
3792                 return -ENOMEM;
3793
3794         log->ni = ni;
3795         log->l_size = l_size;
3796         log->one_page_buf = kmalloc(page_size, GFP_NOFS);
3797
3798         if (!log->one_page_buf) {
3799                 err = -ENOMEM;
3800                 goto out;
3801         }
3802
3803         log->page_size = page_size;
3804         log->page_mask = page_size - 1;
3805         log->page_bits = blksize_bits(page_size);
3806
3807         /* Look for a restart area on the disk. */
3808         err = log_read_rst(log, l_size, true, &rst_info);
3809         if (err)
3810                 goto out;
3811
3812         /* remember 'initialized' */
3813         *initialized = rst_info.initialized;
3814
3815         if (!rst_info.restart) {
3816                 if (rst_info.initialized) {
3817                         /* No restart area but the file is not initialized. */
3818                         err = -EINVAL;
3819                         goto out;
3820                 }
3821
3822                 log_init_pg_hdr(log, page_size, page_size, 1, 1);
3823                 log_create(log, l_size, 0, get_random_int(), false, false);
3824
3825                 log->ra = ra;
3826
3827                 ra = log_create_ra(log);
3828                 if (!ra) {
3829                         err = -ENOMEM;
3830                         goto out;
3831                 }
3832                 log->ra = ra;
3833                 log->init_ra = true;
3834
3835                 goto process_log;
3836         }
3837
3838         /*
3839          * If the restart offset above wasn't zero then we won't
3840          * look for a second restart.
3841          */
3842         if (rst_info.vbo)
3843                 goto check_restart_area;
3844
3845         err = log_read_rst(log, l_size, false, &rst_info2);
3846
3847         /* Determine which restart area to use. */
3848         if (!rst_info2.restart || rst_info2.last_lsn <= rst_info.last_lsn)
3849                 goto use_first_page;
3850
3851         use_second_page = true;
3852
3853         if (rst_info.chkdsk_was_run && page_size != rst_info.vbo) {
3854                 struct RECORD_PAGE_HDR *sp = NULL;
3855                 bool usa_error;
3856
3857                 if (!read_log_page(log, page_size, &sp, &usa_error) &&
3858                     sp->rhdr.sign == NTFS_CHKD_SIGNATURE) {
3859                         use_second_page = false;
3860                 }
3861                 kfree(sp);
3862         }
3863
3864         if (use_second_page) {
3865                 kfree(rst_info.r_page);
3866                 memcpy(&rst_info, &rst_info2, sizeof(struct restart_info));
3867                 rst_info2.r_page = NULL;
3868         }
3869
3870 use_first_page:
3871         kfree(rst_info2.r_page);
3872
3873 check_restart_area:
3874         /*
3875          * If the restart area is at offset 0, we want
3876          * to write the second restart area first.
3877          */
3878         log->init_ra = !!rst_info.vbo;
3879
3880         /* If we have a valid page then grab a pointer to the restart area. */
3881         ra2 = rst_info.valid_page
3882                       ? Add2Ptr(rst_info.r_page,
3883                                 le16_to_cpu(rst_info.r_page->ra_off))
3884                       : NULL;
3885
3886         if (rst_info.chkdsk_was_run ||
3887             (ra2 && ra2->client_idx[1] == LFS_NO_CLIENT_LE)) {
3888                 bool wrapped = false;
3889                 bool use_multi_page = false;
3890                 u32 open_log_count;
3891
3892                 /* Do some checks based on whether we have a valid log page. */
3893                 if (!rst_info.valid_page) {
3894                         open_log_count = get_random_int();
3895                         goto init_log_instance;
3896                 }
3897                 open_log_count = le32_to_cpu(ra2->open_log_count);
3898
3899                 /*
3900                  * If the restart page size isn't changing then we want to
3901                  * check how much work we need to do.
3902                  */
3903                 if (page_size != le32_to_cpu(rst_info.r_page->sys_page_size))
3904                         goto init_log_instance;
3905
3906 init_log_instance:
3907                 log_init_pg_hdr(log, page_size, page_size, 1, 1);
3908
3909                 log_create(log, l_size, rst_info.last_lsn, open_log_count,
3910                            wrapped, use_multi_page);
3911
3912                 ra = log_create_ra(log);
3913                 if (!ra) {
3914                         err = -ENOMEM;
3915                         goto out;
3916                 }
3917                 log->ra = ra;
3918
3919                 /* Put the restart areas and initialize
3920                  * the log file as required.
3921                  */
3922                 goto process_log;
3923         }
3924
3925         if (!ra2) {
3926                 err = -EINVAL;
3927                 goto out;
3928         }
3929
3930         /*
3931          * If the log page or the system page sizes have changed, we can't
3932          * use the log file. We must use the system page size instead of the
3933          * default size if there is not a clean shutdown.
3934          */
3935         t32 = le32_to_cpu(rst_info.r_page->sys_page_size);
3936         if (page_size != t32) {
3937                 l_size = orig_file_size;
3938                 page_size =
3939                         norm_file_page(t32, &l_size, t32 == DefaultLogPageSize);
3940         }
3941
3942         if (page_size != t32 ||
3943             page_size != le32_to_cpu(rst_info.r_page->page_size)) {
3944                 err = -EINVAL;
3945                 goto out;
3946         }
3947
3948         /* If the file size has shrunk then we won't mount it. */
3949         if (l_size < le64_to_cpu(ra2->l_size)) {
3950                 err = -EINVAL;
3951                 goto out;
3952         }
3953
3954         log_init_pg_hdr(log, page_size, page_size,
3955                         le16_to_cpu(rst_info.r_page->major_ver),
3956                         le16_to_cpu(rst_info.r_page->minor_ver));
3957
3958         log->l_size = le64_to_cpu(ra2->l_size);
3959         log->seq_num_bits = le32_to_cpu(ra2->seq_num_bits);
3960         log->file_data_bits = sizeof(u64) * 8 - log->seq_num_bits;
3961         log->seq_num_mask = (8 << log->file_data_bits) - 1;
3962         log->last_lsn = le64_to_cpu(ra2->current_lsn);
3963         log->seq_num = log->last_lsn >> log->file_data_bits;
3964         log->ra_off = le16_to_cpu(rst_info.r_page->ra_off);
3965         log->restart_size = log->sys_page_size - log->ra_off;
3966         log->record_header_len = le16_to_cpu(ra2->rec_hdr_len);
3967         log->ra_size = le16_to_cpu(ra2->ra_len);
3968         log->data_off = le16_to_cpu(ra2->data_off);
3969         log->data_size = log->page_size - log->data_off;
3970         log->reserved = log->data_size - log->record_header_len;
3971
3972         vbo = lsn_to_vbo(log, log->last_lsn);
3973
3974         if (vbo < log->first_page) {
3975                 /* This is a pseudo lsn. */
3976                 log->l_flags |= NTFSLOG_NO_LAST_LSN;
3977                 log->next_page = log->first_page;
3978                 goto find_oldest;
3979         }
3980
3981         /* Find the end of this log record. */
3982         off = final_log_off(log, log->last_lsn,
3983                             le32_to_cpu(ra2->last_lsn_data_len));
3984
3985         /* If we wrapped the file then increment the sequence number. */
3986         if (off <= vbo) {
3987                 log->seq_num += 1;
3988                 log->l_flags |= NTFSLOG_WRAPPED;
3989         }
3990
3991         /* Now compute the next log page to use. */
3992         vbo &= ~log->sys_page_mask;
3993         tail = log->page_size - (off & log->page_mask) - 1;
3994
3995         /*
3996          *If we can fit another log record on the page,
3997          * move back a page the log file.
3998          */
3999         if (tail >= log->record_header_len) {
4000                 log->l_flags |= NTFSLOG_REUSE_TAIL;
4001                 log->next_page = vbo;
4002         } else {
4003                 log->next_page = next_page_off(log, vbo);
4004         }
4005
4006 find_oldest:
4007         /*
4008          * Find the oldest client lsn. Use the last
4009          * flushed lsn as a starting point.
4010          */
4011         log->oldest_lsn = log->last_lsn;
4012         oldest_client_lsn(Add2Ptr(ra2, le16_to_cpu(ra2->client_off)),
4013                           ra2->client_idx[1], &log->oldest_lsn);
4014         log->oldest_lsn_off = lsn_to_vbo(log, log->oldest_lsn);
4015
4016         if (log->oldest_lsn_off < log->first_page)
4017                 log->l_flags |= NTFSLOG_NO_OLDEST_LSN;
4018
4019         if (!(ra2->flags & RESTART_SINGLE_PAGE_IO))
4020                 log->l_flags |= NTFSLOG_WRAPPED | NTFSLOG_MULTIPLE_PAGE_IO;
4021
4022         log->current_openlog_count = le32_to_cpu(ra2->open_log_count);
4023         log->total_avail_pages = log->l_size - log->first_page;
4024         log->total_avail = log->total_avail_pages >> log->page_bits;
4025         log->max_current_avail = log->total_avail * log->reserved;
4026         log->total_avail = log->total_avail * log->data_size;
4027
4028         log->current_avail = current_log_avail(log);
4029
4030         ra = kzalloc(log->restart_size, GFP_NOFS);
4031         if (!ra) {
4032                 err = -ENOMEM;
4033                 goto out;
4034         }
4035         log->ra = ra;
4036
4037         t16 = le16_to_cpu(ra2->client_off);
4038         if (t16 == offsetof(struct RESTART_AREA, clients)) {
4039                 memcpy(ra, ra2, log->ra_size);
4040         } else {
4041                 memcpy(ra, ra2, offsetof(struct RESTART_AREA, clients));
4042                 memcpy(ra->clients, Add2Ptr(ra2, t16),
4043                        le16_to_cpu(ra2->ra_len) - t16);
4044
4045                 log->current_openlog_count = get_random_int();
4046                 ra->open_log_count = cpu_to_le32(log->current_openlog_count);
4047                 log->ra_size = offsetof(struct RESTART_AREA, clients) +
4048                                sizeof(struct CLIENT_REC);
4049                 ra->client_off =
4050                         cpu_to_le16(offsetof(struct RESTART_AREA, clients));
4051                 ra->ra_len = cpu_to_le16(log->ra_size);
4052         }
4053
4054         le32_add_cpu(&ra->open_log_count, 1);
4055
4056         /* Now we need to walk through looking for the last lsn. */
4057         err = last_log_lsn(log);
4058         if (err)
4059                 goto out;
4060
4061         log->current_avail = current_log_avail(log);
4062
4063         /* Remember which restart area to write first. */
4064         log->init_ra = rst_info.vbo;
4065
4066 process_log:
4067         /* 1.0, 1.1, 2.0 log->major_ver/minor_ver - short values. */
4068         switch ((log->major_ver << 16) + log->minor_ver) {
4069         case 0x10000:
4070         case 0x10001:
4071         case 0x20000:
4072                 break;
4073         default:
4074                 ntfs_warn(sbi->sb, "\x24LogFile version %d.%d is not supported",
4075                           log->major_ver, log->minor_ver);
4076                 err = -EOPNOTSUPP;
4077                 log->set_dirty = true;
4078                 goto out;
4079         }
4080
4081         /* One client "NTFS" per logfile. */
4082         ca = Add2Ptr(ra, le16_to_cpu(ra->client_off));
4083
4084         for (client = ra->client_idx[1];; client = cr->next_client) {
4085                 if (client == LFS_NO_CLIENT_LE) {
4086                         /* Insert "NTFS" client LogFile. */
4087                         client = ra->client_idx[0];
4088                         if (client == LFS_NO_CLIENT_LE)
4089                                 return -EINVAL;
4090
4091                         t16 = le16_to_cpu(client);
4092                         cr = ca + t16;
4093
4094                         remove_client(ca, cr, &ra->client_idx[0]);
4095
4096                         cr->restart_lsn = 0;
4097                         cr->oldest_lsn = cpu_to_le64(log->oldest_lsn);
4098                         cr->name_bytes = cpu_to_le32(8);
4099                         cr->name[0] = cpu_to_le16('N');
4100                         cr->name[1] = cpu_to_le16('T');
4101                         cr->name[2] = cpu_to_le16('F');
4102                         cr->name[3] = cpu_to_le16('S');
4103
4104                         add_client(ca, t16, &ra->client_idx[1]);
4105                         break;
4106                 }
4107
4108                 cr = ca + le16_to_cpu(client);
4109
4110                 if (cpu_to_le32(8) == cr->name_bytes &&
4111                     cpu_to_le16('N') == cr->name[0] &&
4112                     cpu_to_le16('T') == cr->name[1] &&
4113                     cpu_to_le16('F') == cr->name[2] &&
4114                     cpu_to_le16('S') == cr->name[3])
4115                         break;
4116         }
4117
4118         /* Update the client handle with the client block information. */
4119         log->client_id.seq_num = cr->seq_num;
4120         log->client_id.client_idx = client;
4121
4122         err = read_rst_area(log, &rst, &ra_lsn);
4123         if (err)
4124                 goto out;
4125
4126         if (!rst)
4127                 goto out;
4128
4129         bytes_per_attr_entry = !rst->major_ver ? 0x2C : 0x28;
4130
4131         checkpt_lsn = le64_to_cpu(rst->check_point_start);
4132         if (!checkpt_lsn)
4133                 checkpt_lsn = ra_lsn;
4134
4135         /* Allocate and Read the Transaction Table. */
4136         if (!rst->transact_table_len)
4137                 goto check_dirty_page_table;
4138
4139         t64 = le64_to_cpu(rst->transact_table_lsn);
4140         err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4141         if (err)
4142                 goto out;
4143
4144         lrh = lcb->log_rec;
4145         frh = lcb->lrh;
4146         rec_len = le32_to_cpu(frh->client_data_len);
4147
4148         if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4149                            bytes_per_attr_entry)) {
4150                 err = -EINVAL;
4151                 goto out;
4152         }
4153
4154         t16 = le16_to_cpu(lrh->redo_off);
4155
4156         rt = Add2Ptr(lrh, t16);
4157         t32 = rec_len - t16;
4158
4159         /* Now check that this is a valid restart table. */
4160         if (!check_rstbl(rt, t32)) {
4161                 err = -EINVAL;
4162                 goto out;
4163         }
4164
4165         trtbl = kmemdup(rt, t32, GFP_NOFS);
4166         if (!trtbl) {
4167                 err = -ENOMEM;
4168                 goto out;
4169         }
4170
4171         lcb_put(lcb);
4172         lcb = NULL;
4173
4174 check_dirty_page_table:
4175         /* The next record back should be the Dirty Pages Table. */
4176         if (!rst->dirty_pages_len)
4177                 goto check_attribute_names;
4178
4179         t64 = le64_to_cpu(rst->dirty_pages_table_lsn);
4180         err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4181         if (err)
4182                 goto out;
4183
4184         lrh = lcb->log_rec;
4185         frh = lcb->lrh;
4186         rec_len = le32_to_cpu(frh->client_data_len);
4187
4188         if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4189                            bytes_per_attr_entry)) {
4190                 err = -EINVAL;
4191                 goto out;
4192         }
4193
4194         t16 = le16_to_cpu(lrh->redo_off);
4195
4196         rt = Add2Ptr(lrh, t16);
4197         t32 = rec_len - t16;
4198
4199         /* Now check that this is a valid restart table. */
4200         if (!check_rstbl(rt, t32)) {
4201                 err = -EINVAL;
4202                 goto out;
4203         }
4204
4205         dptbl = kmemdup(rt, t32, GFP_NOFS);
4206         if (!dptbl) {
4207                 err = -ENOMEM;
4208                 goto out;
4209         }
4210
4211         /* Convert Ra version '0' into version '1'. */
4212         if (rst->major_ver)
4213                 goto end_conv_1;
4214
4215         dp = NULL;
4216         while ((dp = enum_rstbl(dptbl, dp))) {
4217                 struct DIR_PAGE_ENTRY_32 *dp0 = (struct DIR_PAGE_ENTRY_32 *)dp;
4218                 // NOTE: Danger. Check for of boundary.
4219                 memmove(&dp->vcn, &dp0->vcn_low,
4220                         2 * sizeof(u64) +
4221                                 le32_to_cpu(dp->lcns_follow) * sizeof(u64));
4222         }
4223
4224 end_conv_1:
4225         lcb_put(lcb);
4226         lcb = NULL;
4227
4228         /*
4229          * Go through the table and remove the duplicates,
4230          * remembering the oldest lsn values.
4231          */
4232         if (sbi->cluster_size <= log->page_size)
4233                 goto trace_dp_table;
4234
4235         dp = NULL;
4236         while ((dp = enum_rstbl(dptbl, dp))) {
4237                 struct DIR_PAGE_ENTRY *next = dp;
4238
4239                 while ((next = enum_rstbl(dptbl, next))) {
4240                         if (next->target_attr == dp->target_attr &&
4241                             next->vcn == dp->vcn) {
4242                                 if (le64_to_cpu(next->oldest_lsn) <
4243                                     le64_to_cpu(dp->oldest_lsn)) {
4244                                         dp->oldest_lsn = next->oldest_lsn;
4245                                 }
4246
4247                                 free_rsttbl_idx(dptbl, PtrOffset(dptbl, next));
4248                         }
4249                 }
4250         }
4251 trace_dp_table:
4252 check_attribute_names:
4253         /* The next record should be the Attribute Names. */
4254         if (!rst->attr_names_len)
4255                 goto check_attr_table;
4256
4257         t64 = le64_to_cpu(rst->attr_names_lsn);
4258         err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4259         if (err)
4260                 goto out;
4261
4262         lrh = lcb->log_rec;
4263         frh = lcb->lrh;
4264         rec_len = le32_to_cpu(frh->client_data_len);
4265
4266         if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4267                            bytes_per_attr_entry)) {
4268                 err = -EINVAL;
4269                 goto out;
4270         }
4271
4272         t32 = lrh_length(lrh);
4273         rec_len -= t32;
4274
4275         attr_names = kmemdup(Add2Ptr(lrh, t32), rec_len, GFP_NOFS);
4276
4277         lcb_put(lcb);
4278         lcb = NULL;
4279
4280 check_attr_table:
4281         /* The next record should be the attribute Table. */
4282         if (!rst->open_attr_len)
4283                 goto check_attribute_names2;
4284
4285         t64 = le64_to_cpu(rst->open_attr_table_lsn);
4286         err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4287         if (err)
4288                 goto out;
4289
4290         lrh = lcb->log_rec;
4291         frh = lcb->lrh;
4292         rec_len = le32_to_cpu(frh->client_data_len);
4293
4294         if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4295                            bytes_per_attr_entry)) {
4296                 err = -EINVAL;
4297                 goto out;
4298         }
4299
4300         t16 = le16_to_cpu(lrh->redo_off);
4301
4302         rt = Add2Ptr(lrh, t16);
4303         t32 = rec_len - t16;
4304
4305         if (!check_rstbl(rt, t32)) {
4306                 err = -EINVAL;
4307                 goto out;
4308         }
4309
4310         oatbl = kmemdup(rt, t32, GFP_NOFS);
4311         if (!oatbl) {
4312                 err = -ENOMEM;
4313                 goto out;
4314         }
4315
4316         log->open_attr_tbl = oatbl;
4317
4318         /* Clear all of the Attr pointers. */
4319         oe = NULL;
4320         while ((oe = enum_rstbl(oatbl, oe))) {
4321                 if (!rst->major_ver) {
4322                         struct OPEN_ATTR_ENRTY_32 oe0;
4323
4324                         /* Really 'oe' points to OPEN_ATTR_ENRTY_32. */
4325                         memcpy(&oe0, oe, SIZEOF_OPENATTRIBUTEENTRY0);
4326
4327                         oe->bytes_per_index = oe0.bytes_per_index;
4328                         oe->type = oe0.type;
4329                         oe->is_dirty_pages = oe0.is_dirty_pages;
4330                         oe->name_len = 0;
4331                         oe->ref = oe0.ref;
4332                         oe->open_record_lsn = oe0.open_record_lsn;
4333                 }
4334
4335                 oe->is_attr_name = 0;
4336                 oe->ptr = NULL;
4337         }
4338
4339         lcb_put(lcb);
4340         lcb = NULL;
4341
4342 check_attribute_names2:
4343         if (!rst->attr_names_len)
4344                 goto trace_attribute_table;
4345
4346         ane = attr_names;
4347         if (!oatbl)
4348                 goto trace_attribute_table;
4349         while (ane->off) {
4350                 /* TODO: Clear table on exit! */
4351                 oe = Add2Ptr(oatbl, le16_to_cpu(ane->off));
4352                 t16 = le16_to_cpu(ane->name_bytes);
4353                 oe->name_len = t16 / sizeof(short);
4354                 oe->ptr = ane->name;
4355                 oe->is_attr_name = 2;
4356                 ane = Add2Ptr(ane, sizeof(struct ATTR_NAME_ENTRY) + t16);
4357         }
4358
4359 trace_attribute_table:
4360         /*
4361          * If the checkpt_lsn is zero, then this is a freshly
4362          * formatted disk and we have no work to do.
4363          */
4364         if (!checkpt_lsn) {
4365                 err = 0;
4366                 goto out;
4367         }
4368
4369         if (!oatbl) {
4370                 oatbl = init_rsttbl(bytes_per_attr_entry, 8);
4371                 if (!oatbl) {
4372                         err = -ENOMEM;
4373                         goto out;
4374                 }
4375         }
4376
4377         log->open_attr_tbl = oatbl;
4378
4379         /* Start the analysis pass from the Checkpoint lsn. */
4380         rec_lsn = checkpt_lsn;
4381
4382         /* Read the first lsn. */
4383         err = read_log_rec_lcb(log, checkpt_lsn, lcb_ctx_next, &lcb);
4384         if (err)
4385                 goto out;
4386
4387         /* Loop to read all subsequent records to the end of the log file. */
4388 next_log_record_analyze:
4389         err = read_next_log_rec(log, lcb, &rec_lsn);
4390         if (err)
4391                 goto out;
4392
4393         if (!rec_lsn)
4394                 goto end_log_records_enumerate;
4395
4396         frh = lcb->lrh;
4397         transact_id = le32_to_cpu(frh->transact_id);
4398         rec_len = le32_to_cpu(frh->client_data_len);
4399         lrh = lcb->log_rec;
4400
4401         if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
4402                 err = -EINVAL;
4403                 goto out;
4404         }
4405
4406         /*
4407          * The first lsn after the previous lsn remembered
4408          * the checkpoint is the first candidate for the rlsn.
4409          */
4410         if (!rlsn)
4411                 rlsn = rec_lsn;
4412
4413         if (LfsClientRecord != frh->record_type)
4414                 goto next_log_record_analyze;
4415
4416         /*
4417          * Now update the Transaction Table for this transaction. If there
4418          * is no entry present or it is unallocated we allocate the entry.
4419          */
4420         if (!trtbl) {
4421                 trtbl = init_rsttbl(sizeof(struct TRANSACTION_ENTRY),
4422                                     INITIAL_NUMBER_TRANSACTIONS);
4423                 if (!trtbl) {
4424                         err = -ENOMEM;
4425                         goto out;
4426                 }
4427         }
4428
4429         tr = Add2Ptr(trtbl, transact_id);
4430
4431         if (transact_id >= bytes_per_rt(trtbl) ||
4432             tr->next != RESTART_ENTRY_ALLOCATED_LE) {
4433                 tr = alloc_rsttbl_from_idx(&trtbl, transact_id);
4434                 if (!tr) {
4435                         err = -ENOMEM;
4436                         goto out;
4437                 }
4438                 tr->transact_state = TransactionActive;
4439                 tr->first_lsn = cpu_to_le64(rec_lsn);
4440         }
4441
4442         tr->prev_lsn = tr->undo_next_lsn = cpu_to_le64(rec_lsn);
4443
4444         /*
4445          * If this is a compensation log record, then change
4446          * the undo_next_lsn to be the undo_next_lsn of this record.
4447          */
4448         if (lrh->undo_op == cpu_to_le16(CompensationLogRecord))
4449                 tr->undo_next_lsn = frh->client_undo_next_lsn;
4450
4451         /* Dispatch to handle log record depending on type. */
4452         switch (le16_to_cpu(lrh->redo_op)) {
4453         case InitializeFileRecordSegment:
4454         case DeallocateFileRecordSegment:
4455         case WriteEndOfFileRecordSegment:
4456         case CreateAttribute:
4457         case DeleteAttribute:
4458         case UpdateResidentValue:
4459         case UpdateNonresidentValue:
4460         case UpdateMappingPairs:
4461         case SetNewAttributeSizes:
4462         case AddIndexEntryRoot:
4463         case DeleteIndexEntryRoot:
4464         case AddIndexEntryAllocation:
4465         case DeleteIndexEntryAllocation:
4466         case WriteEndOfIndexBuffer:
4467         case SetIndexEntryVcnRoot:
4468         case SetIndexEntryVcnAllocation:
4469         case UpdateFileNameRoot:
4470         case UpdateFileNameAllocation:
4471         case SetBitsInNonresidentBitMap:
4472         case ClearBitsInNonresidentBitMap:
4473         case UpdateRecordDataRoot:
4474         case UpdateRecordDataAllocation:
4475         case ZeroEndOfFileRecord:
4476                 t16 = le16_to_cpu(lrh->target_attr);
4477                 t64 = le64_to_cpu(lrh->target_vcn);
4478                 dp = find_dp(dptbl, t16, t64);
4479
4480                 if (dp)
4481                         goto copy_lcns;
4482
4483                 /*
4484                  * Calculate the number of clusters per page the system
4485                  * which wrote the checkpoint, possibly creating the table.
4486                  */
4487                 if (dptbl) {
4488                         t32 = (le16_to_cpu(dptbl->size) -
4489                                sizeof(struct DIR_PAGE_ENTRY)) /
4490                               sizeof(u64);
4491                 } else {
4492                         t32 = log->clst_per_page;
4493                         kfree(dptbl);
4494                         dptbl = init_rsttbl(struct_size(dp, page_lcns, t32),
4495                                             32);
4496                         if (!dptbl) {
4497                                 err = -ENOMEM;
4498                                 goto out;
4499                         }
4500                 }
4501
4502                 dp = alloc_rsttbl_idx(&dptbl);
4503                 if (!dp) {
4504                         err = -ENOMEM;
4505                         goto out;
4506                 }
4507                 dp->target_attr = cpu_to_le32(t16);
4508                 dp->transfer_len = cpu_to_le32(t32 << sbi->cluster_bits);
4509                 dp->lcns_follow = cpu_to_le32(t32);
4510                 dp->vcn = cpu_to_le64(t64 & ~((u64)t32 - 1));
4511                 dp->oldest_lsn = cpu_to_le64(rec_lsn);
4512
4513 copy_lcns:
4514                 /*
4515                  * Copy the Lcns from the log record into the Dirty Page Entry.
4516                  * TODO: For different page size support, must somehow make
4517                  * whole routine a loop, case Lcns do not fit below.
4518                  */
4519                 t16 = le16_to_cpu(lrh->lcns_follow);
4520                 for (i = 0; i < t16; i++) {
4521                         size_t j = (size_t)(le64_to_cpu(lrh->target_vcn) -
4522                                             le64_to_cpu(dp->vcn));
4523                         dp->page_lcns[j + i] = lrh->page_lcns[i];
4524                 }
4525
4526                 goto next_log_record_analyze;
4527
4528         case DeleteDirtyClusters: {
4529                 u32 range_count =
4530                         le16_to_cpu(lrh->redo_len) / sizeof(struct LCN_RANGE);
4531                 const struct LCN_RANGE *r =
4532                         Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
4533
4534                 /* Loop through all of the Lcn ranges this log record. */
4535                 for (i = 0; i < range_count; i++, r++) {
4536                         u64 lcn0 = le64_to_cpu(r->lcn);
4537                         u64 lcn_e = lcn0 + le64_to_cpu(r->len) - 1;
4538
4539                         dp = NULL;
4540                         while ((dp = enum_rstbl(dptbl, dp))) {
4541                                 u32 j;
4542
4543                                 t32 = le32_to_cpu(dp->lcns_follow);
4544                                 for (j = 0; j < t32; j++) {
4545                                         t64 = le64_to_cpu(dp->page_lcns[j]);
4546                                         if (t64 >= lcn0 && t64 <= lcn_e)
4547                                                 dp->page_lcns[j] = 0;
4548                                 }
4549                         }
4550                 }
4551                 goto next_log_record_analyze;
4552                 ;
4553         }
4554
4555         case OpenNonresidentAttribute:
4556                 t16 = le16_to_cpu(lrh->target_attr);
4557                 if (t16 >= bytes_per_rt(oatbl)) {
4558                         /*
4559                          * Compute how big the table needs to be.
4560                          * Add 10 extra entries for some cushion.
4561                          */
4562                         u32 new_e = t16 / le16_to_cpu(oatbl->size);
4563
4564                         new_e += 10 - le16_to_cpu(oatbl->used);
4565
4566                         oatbl = extend_rsttbl(oatbl, new_e, ~0u);
4567                         log->open_attr_tbl = oatbl;
4568                         if (!oatbl) {
4569                                 err = -ENOMEM;
4570                                 goto out;
4571                         }
4572                 }
4573
4574                 /* Point to the entry being opened. */
4575                 oe = alloc_rsttbl_from_idx(&oatbl, t16);
4576                 log->open_attr_tbl = oatbl;
4577                 if (!oe) {
4578                         err = -ENOMEM;
4579                         goto out;
4580                 }
4581
4582                 /* Initialize this entry from the log record. */
4583                 t16 = le16_to_cpu(lrh->redo_off);
4584                 if (!rst->major_ver) {
4585                         /* Convert version '0' into version '1'. */
4586                         struct OPEN_ATTR_ENRTY_32 *oe0 = Add2Ptr(lrh, t16);
4587
4588                         oe->bytes_per_index = oe0->bytes_per_index;
4589                         oe->type = oe0->type;
4590                         oe->is_dirty_pages = oe0->is_dirty_pages;
4591                         oe->name_len = 0; //oe0.name_len;
4592                         oe->ref = oe0->ref;
4593                         oe->open_record_lsn = oe0->open_record_lsn;
4594                 } else {
4595                         memcpy(oe, Add2Ptr(lrh, t16), bytes_per_attr_entry);
4596                 }
4597
4598                 t16 = le16_to_cpu(lrh->undo_len);
4599                 if (t16) {
4600                         oe->ptr = kmalloc(t16, GFP_NOFS);
4601                         if (!oe->ptr) {
4602                                 err = -ENOMEM;
4603                                 goto out;
4604                         }
4605                         oe->name_len = t16 / sizeof(short);
4606                         memcpy(oe->ptr,
4607                                Add2Ptr(lrh, le16_to_cpu(lrh->undo_off)), t16);
4608                         oe->is_attr_name = 1;
4609                 } else {
4610                         oe->ptr = NULL;
4611                         oe->is_attr_name = 0;
4612                 }
4613
4614                 goto next_log_record_analyze;
4615
4616         case HotFix:
4617                 t16 = le16_to_cpu(lrh->target_attr);
4618                 t64 = le64_to_cpu(lrh->target_vcn);
4619                 dp = find_dp(dptbl, t16, t64);
4620                 if (dp) {
4621                         size_t j = le64_to_cpu(lrh->target_vcn) -
4622                                    le64_to_cpu(dp->vcn);
4623                         if (dp->page_lcns[j])
4624                                 dp->page_lcns[j] = lrh->page_lcns[0];
4625                 }
4626                 goto next_log_record_analyze;
4627
4628         case EndTopLevelAction:
4629                 tr = Add2Ptr(trtbl, transact_id);
4630                 tr->prev_lsn = cpu_to_le64(rec_lsn);
4631                 tr->undo_next_lsn = frh->client_undo_next_lsn;
4632                 goto next_log_record_analyze;
4633
4634         case PrepareTransaction:
4635                 tr = Add2Ptr(trtbl, transact_id);
4636                 tr->transact_state = TransactionPrepared;
4637                 goto next_log_record_analyze;
4638
4639         case CommitTransaction:
4640                 tr = Add2Ptr(trtbl, transact_id);
4641                 tr->transact_state = TransactionCommitted;
4642                 goto next_log_record_analyze;
4643
4644         case ForgetTransaction:
4645                 free_rsttbl_idx(trtbl, transact_id);
4646                 goto next_log_record_analyze;
4647
4648         case Noop:
4649         case OpenAttributeTableDump:
4650         case AttributeNamesDump:
4651         case DirtyPageTableDump:
4652         case TransactionTableDump:
4653                 /* The following cases require no action the Analysis Pass. */
4654                 goto next_log_record_analyze;
4655
4656         default:
4657                 /*
4658                  * All codes will be explicitly handled.
4659                  * If we see a code we do not expect, then we are trouble.
4660                  */
4661                 goto next_log_record_analyze;
4662         }
4663
4664 end_log_records_enumerate:
4665         lcb_put(lcb);
4666         lcb = NULL;
4667
4668         /*
4669          * Scan the Dirty Page Table and Transaction Table for
4670          * the lowest lsn, and return it as the Redo lsn.
4671          */
4672         dp = NULL;
4673         while ((dp = enum_rstbl(dptbl, dp))) {
4674                 t64 = le64_to_cpu(dp->oldest_lsn);
4675                 if (t64 && t64 < rlsn)
4676                         rlsn = t64;
4677         }
4678
4679         tr = NULL;
4680         while ((tr = enum_rstbl(trtbl, tr))) {
4681                 t64 = le64_to_cpu(tr->first_lsn);
4682                 if (t64 && t64 < rlsn)
4683                         rlsn = t64;
4684         }
4685
4686         /*
4687          * Only proceed if the Dirty Page Table or Transaction
4688          * table are not empty.
4689          */
4690         if ((!dptbl || !dptbl->total) && (!trtbl || !trtbl->total))
4691                 goto end_reply;
4692
4693         sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
4694         if (is_ro)
4695                 goto out;
4696
4697         /* Reopen all of the attributes with dirty pages. */
4698         oe = NULL;
4699 next_open_attribute:
4700
4701         oe = enum_rstbl(oatbl, oe);
4702         if (!oe) {
4703                 err = 0;
4704                 dp = NULL;
4705                 goto next_dirty_page;
4706         }
4707
4708         oa = kzalloc(sizeof(struct OpenAttr), GFP_NOFS);
4709         if (!oa) {
4710                 err = -ENOMEM;
4711                 goto out;
4712         }
4713
4714         inode = ntfs_iget5(sbi->sb, &oe->ref, NULL);
4715         if (IS_ERR(inode))
4716                 goto fake_attr;
4717
4718         if (is_bad_inode(inode)) {
4719                 iput(inode);
4720 fake_attr:
4721                 if (oa->ni) {
4722                         iput(&oa->ni->vfs_inode);
4723                         oa->ni = NULL;
4724                 }
4725
4726                 attr = attr_create_nonres_log(sbi, oe->type, 0, oe->ptr,
4727                                               oe->name_len, 0);
4728                 if (!attr) {
4729                         kfree(oa);
4730                         err = -ENOMEM;
4731                         goto out;
4732                 }
4733                 oa->attr = attr;
4734                 oa->run1 = &oa->run0;
4735                 goto final_oe;
4736         }
4737
4738         ni_oe = ntfs_i(inode);
4739         oa->ni = ni_oe;
4740
4741         attr = ni_find_attr(ni_oe, NULL, NULL, oe->type, oe->ptr, oe->name_len,
4742                             NULL, NULL);
4743
4744         if (!attr)
4745                 goto fake_attr;
4746
4747         t32 = le32_to_cpu(attr->size);
4748         oa->attr = kmemdup(attr, t32, GFP_NOFS);
4749         if (!oa->attr)
4750                 goto fake_attr;
4751
4752         if (!S_ISDIR(inode->i_mode)) {
4753                 if (attr->type == ATTR_DATA && !attr->name_len) {
4754                         oa->run1 = &ni_oe->file.run;
4755                         goto final_oe;
4756                 }
4757         } else {
4758                 if (attr->type == ATTR_ALLOC &&
4759                     attr->name_len == ARRAY_SIZE(I30_NAME) &&
4760                     !memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME))) {
4761                         oa->run1 = &ni_oe->dir.alloc_run;
4762                         goto final_oe;
4763                 }
4764         }
4765
4766         if (attr->non_res) {
4767                 u16 roff = le16_to_cpu(attr->nres.run_off);
4768                 CLST svcn = le64_to_cpu(attr->nres.svcn);
4769
4770                 err = run_unpack(&oa->run0, sbi, inode->i_ino, svcn,
4771                                  le64_to_cpu(attr->nres.evcn), svcn,
4772                                  Add2Ptr(attr, roff), t32 - roff);
4773                 if (err < 0) {
4774                         kfree(oa->attr);
4775                         oa->attr = NULL;
4776                         goto fake_attr;
4777                 }
4778                 err = 0;
4779         }
4780         oa->run1 = &oa->run0;
4781         attr = oa->attr;
4782
4783 final_oe:
4784         if (oe->is_attr_name == 1)
4785                 kfree(oe->ptr);
4786         oe->is_attr_name = 0;
4787         oe->ptr = oa;
4788         oe->name_len = attr->name_len;
4789
4790         goto next_open_attribute;
4791
4792         /*
4793          * Now loop through the dirty page table to extract all of the Vcn/Lcn.
4794          * Mapping that we have, and insert it into the appropriate run.
4795          */
4796 next_dirty_page:
4797         dp = enum_rstbl(dptbl, dp);
4798         if (!dp)
4799                 goto do_redo_1;
4800
4801         oe = Add2Ptr(oatbl, le32_to_cpu(dp->target_attr));
4802
4803         if (oe->next != RESTART_ENTRY_ALLOCATED_LE)
4804                 goto next_dirty_page;
4805
4806         oa = oe->ptr;
4807         if (!oa)
4808                 goto next_dirty_page;
4809
4810         i = -1;
4811 next_dirty_page_vcn:
4812         i += 1;
4813         if (i >= le32_to_cpu(dp->lcns_follow))
4814                 goto next_dirty_page;
4815
4816         vcn = le64_to_cpu(dp->vcn) + i;
4817         size = (vcn + 1) << sbi->cluster_bits;
4818
4819         if (!dp->page_lcns[i])
4820                 goto next_dirty_page_vcn;
4821
4822         rno = ino_get(&oe->ref);
4823         if (rno <= MFT_REC_MIRR &&
4824             size < (MFT_REC_VOL + 1) * sbi->record_size &&
4825             oe->type == ATTR_DATA) {
4826                 goto next_dirty_page_vcn;
4827         }
4828
4829         lcn = le64_to_cpu(dp->page_lcns[i]);
4830
4831         if ((!run_lookup_entry(oa->run1, vcn, &lcn0, &len0, NULL) ||
4832              lcn0 != lcn) &&
4833             !run_add_entry(oa->run1, vcn, lcn, 1, false)) {
4834                 err = -ENOMEM;
4835                 goto out;
4836         }
4837         attr = oa->attr;
4838         t64 = le64_to_cpu(attr->nres.alloc_size);
4839         if (size > t64) {
4840                 attr->nres.valid_size = attr->nres.data_size =
4841                         attr->nres.alloc_size = cpu_to_le64(size);
4842         }
4843         goto next_dirty_page_vcn;
4844
4845 do_redo_1:
4846         /*
4847          * Perform the Redo Pass, to restore all of the dirty pages to the same
4848          * contents that they had immediately before the crash. If the dirty
4849          * page table is empty, then we can skip the entire Redo Pass.
4850          */
4851         if (!dptbl || !dptbl->total)
4852                 goto do_undo_action;
4853
4854         rec_lsn = rlsn;
4855
4856         /*
4857          * Read the record at the Redo lsn, before falling
4858          * into common code to handle each record.
4859          */
4860         err = read_log_rec_lcb(log, rlsn, lcb_ctx_next, &lcb);
4861         if (err)
4862                 goto out;
4863
4864         /*
4865          * Now loop to read all of our log records forwards, until
4866          * we hit the end of the file, cleaning up at the end.
4867          */
4868 do_action_next:
4869         frh = lcb->lrh;
4870
4871         if (LfsClientRecord != frh->record_type)
4872                 goto read_next_log_do_action;
4873
4874         transact_id = le32_to_cpu(frh->transact_id);
4875         rec_len = le32_to_cpu(frh->client_data_len);
4876         lrh = lcb->log_rec;
4877
4878         if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
4879                 err = -EINVAL;
4880                 goto out;
4881         }
4882
4883         /* Ignore log records that do not update pages. */
4884         if (lrh->lcns_follow)
4885                 goto find_dirty_page;
4886
4887         goto read_next_log_do_action;
4888
4889 find_dirty_page:
4890         t16 = le16_to_cpu(lrh->target_attr);
4891         t64 = le64_to_cpu(lrh->target_vcn);
4892         dp = find_dp(dptbl, t16, t64);
4893
4894         if (!dp)
4895                 goto read_next_log_do_action;
4896
4897         if (rec_lsn < le64_to_cpu(dp->oldest_lsn))
4898                 goto read_next_log_do_action;
4899
4900         t16 = le16_to_cpu(lrh->target_attr);
4901         if (t16 >= bytes_per_rt(oatbl)) {
4902                 err = -EINVAL;
4903                 goto out;
4904         }
4905
4906         oe = Add2Ptr(oatbl, t16);
4907
4908         if (oe->next != RESTART_ENTRY_ALLOCATED_LE) {
4909                 err = -EINVAL;
4910                 goto out;
4911         }
4912
4913         oa = oe->ptr;
4914
4915         if (!oa) {
4916                 err = -EINVAL;
4917                 goto out;
4918         }
4919         attr = oa->attr;
4920
4921         vcn = le64_to_cpu(lrh->target_vcn);
4922
4923         if (!run_lookup_entry(oa->run1, vcn, &lcn, NULL, NULL) ||
4924             lcn == SPARSE_LCN) {
4925                 goto read_next_log_do_action;
4926         }
4927
4928         /* Point to the Redo data and get its length. */
4929         data = Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
4930         dlen = le16_to_cpu(lrh->redo_len);
4931
4932         /* Shorten length by any Lcns which were deleted. */
4933         saved_len = dlen;
4934
4935         for (i = le16_to_cpu(lrh->lcns_follow); i; i--) {
4936                 size_t j;
4937                 u32 alen, voff;
4938
4939                 voff = le16_to_cpu(lrh->record_off) +
4940                        le16_to_cpu(lrh->attr_off);
4941                 voff += le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
4942
4943                 /* If the Vcn question is allocated, we can just get out. */
4944                 j = le64_to_cpu(lrh->target_vcn) - le64_to_cpu(dp->vcn);
4945                 if (dp->page_lcns[j + i - 1])
4946                         break;
4947
4948                 if (!saved_len)
4949                         saved_len = 1;
4950
4951                 /*
4952                  * Calculate the allocated space left relative to the
4953                  * log record Vcn, after removing this unallocated Vcn.
4954                  */
4955                 alen = (i - 1) << sbi->cluster_bits;
4956
4957                 /*
4958                  * If the update described this log record goes beyond
4959                  * the allocated space, then we will have to reduce the length.
4960                  */
4961                 if (voff >= alen)
4962                         dlen = 0;
4963                 else if (voff + dlen > alen)
4964                         dlen = alen - voff;
4965         }
4966
4967         /*
4968          * If the resulting dlen from above is now zero,
4969          * we can skip this log record.
4970          */
4971         if (!dlen && saved_len)
4972                 goto read_next_log_do_action;
4973
4974         t16 = le16_to_cpu(lrh->redo_op);
4975         if (can_skip_action(t16))
4976                 goto read_next_log_do_action;
4977
4978         /* Apply the Redo operation a common routine. */
4979         err = do_action(log, oe, lrh, t16, data, dlen, rec_len, &rec_lsn);
4980         if (err)
4981                 goto out;
4982
4983         /* Keep reading and looping back until end of file. */
4984 read_next_log_do_action:
4985         err = read_next_log_rec(log, lcb, &rec_lsn);
4986         if (!err && rec_lsn)
4987                 goto do_action_next;
4988
4989         lcb_put(lcb);
4990         lcb = NULL;
4991
4992 do_undo_action:
4993         /* Scan Transaction Table. */
4994         tr = NULL;
4995 transaction_table_next:
4996         tr = enum_rstbl(trtbl, tr);
4997         if (!tr)
4998                 goto undo_action_done;
4999
5000         if (TransactionActive != tr->transact_state || !tr->undo_next_lsn) {
5001                 free_rsttbl_idx(trtbl, PtrOffset(trtbl, tr));
5002                 goto transaction_table_next;
5003         }
5004
5005         log->transaction_id = PtrOffset(trtbl, tr);
5006         undo_next_lsn = le64_to_cpu(tr->undo_next_lsn);
5007
5008         /*
5009          * We only have to do anything if the transaction has
5010          * something its undo_next_lsn field.
5011          */
5012         if (!undo_next_lsn)
5013                 goto commit_undo;
5014
5015         /* Read the first record to be undone by this transaction. */
5016         err = read_log_rec_lcb(log, undo_next_lsn, lcb_ctx_undo_next, &lcb);
5017         if (err)
5018                 goto out;
5019
5020         /*
5021          * Now loop to read all of our log records forwards,
5022          * until we hit the end of the file, cleaning up at the end.
5023          */
5024 undo_action_next:
5025
5026         lrh = lcb->log_rec;
5027         frh = lcb->lrh;
5028         transact_id = le32_to_cpu(frh->transact_id);
5029         rec_len = le32_to_cpu(frh->client_data_len);
5030
5031         if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
5032                 err = -EINVAL;
5033                 goto out;
5034         }
5035
5036         if (lrh->undo_op == cpu_to_le16(Noop))
5037                 goto read_next_log_undo_action;
5038
5039         oe = Add2Ptr(oatbl, le16_to_cpu(lrh->target_attr));
5040         oa = oe->ptr;
5041
5042         t16 = le16_to_cpu(lrh->lcns_follow);
5043         if (!t16)
5044                 goto add_allocated_vcns;
5045
5046         is_mapped = run_lookup_entry(oa->run1, le64_to_cpu(lrh->target_vcn),
5047                                      &lcn, &clen, NULL);
5048
5049         /*
5050          * If the mapping isn't already the table or the  mapping
5051          * corresponds to a hole the mapping, we need to make sure
5052          * there is no partial page already memory.
5053          */
5054         if (is_mapped && lcn != SPARSE_LCN && clen >= t16)
5055                 goto add_allocated_vcns;
5056
5057         vcn = le64_to_cpu(lrh->target_vcn);
5058         vcn &= ~(log->clst_per_page - 1);
5059
5060 add_allocated_vcns:
5061         for (i = 0, vcn = le64_to_cpu(lrh->target_vcn),
5062             size = (vcn + 1) << sbi->cluster_bits;
5063              i < t16; i++, vcn += 1, size += sbi->cluster_size) {
5064                 attr = oa->attr;
5065                 if (!attr->non_res) {
5066                         if (size > le32_to_cpu(attr->res.data_size))
5067                                 attr->res.data_size = cpu_to_le32(size);
5068                 } else {
5069                         if (size > le64_to_cpu(attr->nres.data_size))
5070                                 attr->nres.valid_size = attr->nres.data_size =
5071                                         attr->nres.alloc_size =
5072                                                 cpu_to_le64(size);
5073                 }
5074         }
5075
5076         t16 = le16_to_cpu(lrh->undo_op);
5077         if (can_skip_action(t16))
5078                 goto read_next_log_undo_action;
5079
5080         /* Point to the Redo data and get its length. */
5081         data = Add2Ptr(lrh, le16_to_cpu(lrh->undo_off));
5082         dlen = le16_to_cpu(lrh->undo_len);
5083
5084         /* It is time to apply the undo action. */
5085         err = do_action(log, oe, lrh, t16, data, dlen, rec_len, NULL);
5086
5087 read_next_log_undo_action:
5088         /*
5089          * Keep reading and looping back until we have read the
5090          * last record for this transaction.
5091          */
5092         err = read_next_log_rec(log, lcb, &rec_lsn);
5093         if (err)
5094                 goto out;
5095
5096         if (rec_lsn)
5097                 goto undo_action_next;
5098
5099         lcb_put(lcb);
5100         lcb = NULL;
5101
5102 commit_undo:
5103         free_rsttbl_idx(trtbl, log->transaction_id);
5104
5105         log->transaction_id = 0;
5106
5107         goto transaction_table_next;
5108
5109 undo_action_done:
5110
5111         ntfs_update_mftmirr(sbi, 0);
5112
5113         sbi->flags &= ~NTFS_FLAGS_NEED_REPLAY;
5114
5115 end_reply:
5116
5117         err = 0;
5118         if (is_ro)
5119                 goto out;
5120
5121         rh = kzalloc(log->page_size, GFP_NOFS);
5122         if (!rh) {
5123                 err = -ENOMEM;
5124                 goto out;
5125         }
5126
5127         rh->rhdr.sign = NTFS_RSTR_SIGNATURE;
5128         rh->rhdr.fix_off = cpu_to_le16(offsetof(struct RESTART_HDR, fixups));
5129         t16 = (log->page_size >> SECTOR_SHIFT) + 1;
5130         rh->rhdr.fix_num = cpu_to_le16(t16);
5131         rh->sys_page_size = cpu_to_le32(log->page_size);
5132         rh->page_size = cpu_to_le32(log->page_size);
5133
5134         t16 = ALIGN(offsetof(struct RESTART_HDR, fixups) + sizeof(short) * t16,
5135                     8);
5136         rh->ra_off = cpu_to_le16(t16);
5137         rh->minor_ver = cpu_to_le16(1); // 0x1A:
5138         rh->major_ver = cpu_to_le16(1); // 0x1C:
5139
5140         ra2 = Add2Ptr(rh, t16);
5141         memcpy(ra2, ra, sizeof(struct RESTART_AREA));
5142
5143         ra2->client_idx[0] = 0;
5144         ra2->client_idx[1] = LFS_NO_CLIENT_LE;
5145         ra2->flags = cpu_to_le16(2);
5146
5147         le32_add_cpu(&ra2->open_log_count, 1);
5148
5149         ntfs_fix_pre_write(&rh->rhdr, log->page_size);
5150
5151         err = ntfs_sb_write_run(sbi, &ni->file.run, 0, rh, log->page_size, 0);
5152         if (!err)
5153                 err = ntfs_sb_write_run(sbi, &log->ni->file.run, log->page_size,
5154                                         rh, log->page_size, 0);
5155
5156         kfree(rh);
5157         if (err)
5158                 goto out;
5159
5160 out:
5161         kfree(rst);
5162         if (lcb)
5163                 lcb_put(lcb);
5164
5165         /*
5166          * Scan the Open Attribute Table to close all of
5167          * the open attributes.
5168          */
5169         oe = NULL;
5170         while ((oe = enum_rstbl(oatbl, oe))) {
5171                 rno = ino_get(&oe->ref);
5172
5173                 if (oe->is_attr_name == 1) {
5174                         kfree(oe->ptr);
5175                         oe->ptr = NULL;
5176                         continue;
5177                 }
5178
5179                 if (oe->is_attr_name)
5180                         continue;
5181
5182                 oa = oe->ptr;
5183                 if (!oa)
5184                         continue;
5185
5186                 run_close(&oa->run0);
5187                 kfree(oa->attr);
5188                 if (oa->ni)
5189                         iput(&oa->ni->vfs_inode);
5190                 kfree(oa);
5191         }
5192
5193         kfree(trtbl);
5194         kfree(oatbl);
5195         kfree(dptbl);
5196         kfree(attr_names);
5197         kfree(rst_info.r_page);
5198
5199         kfree(ra);
5200         kfree(log->one_page_buf);
5201
5202         if (err)
5203                 sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
5204
5205         if (err == -EROFS)
5206                 err = 0;
5207         else if (log->set_dirty)
5208                 ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
5209
5210         kfree(log);
5211
5212         return err;
5213 }