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