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