Back to home page

OSCL-LXR

 
 

    


0001 // SPDX-License-Identifier: GPL-2.0
0002 /*
0003  *
0004  * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
0005  *
0006  */
0007 
0008 #include <linux/blkdev.h>
0009 #include <linux/fs.h>
0010 #include <linux/random.h>
0011 #include <linux/slab.h>
0012 
0013 #include "debug.h"
0014 #include "ntfs.h"
0015 #include "ntfs_fs.h"
0016 
0017 /*
0018  * LOG FILE structs
0019  */
0020 
0021 // clang-format off
0022 
0023 #define MaxLogFileSize     0x100000000ull
0024 #define DefaultLogPageSize 4096
0025 #define MinLogRecordPages  0x30
0026 
0027 struct RESTART_HDR {
0028     struct NTFS_RECORD_HEADER rhdr; // 'RSTR'
0029     __le32 sys_page_size; // 0x10: Page size of the system which initialized the log.
0030     __le32 page_size;     // 0x14: Log page size used for this log file.
0031     __le16 ra_off;        // 0x18:
0032     __le16 minor_ver;     // 0x1A:
0033     __le16 major_ver;     // 0x1C:
0034     __le16 fixups[];
0035 };
0036 
0037 #define LFS_NO_CLIENT 0xffff
0038 #define LFS_NO_CLIENT_LE cpu_to_le16(0xffff)
0039 
0040 struct CLIENT_REC {
0041     __le64 oldest_lsn;
0042     __le64 restart_lsn; // 0x08:
0043     __le16 prev_client; // 0x10:
0044     __le16 next_client; // 0x12:
0045     __le16 seq_num;     // 0x14:
0046     u8 align[6];        // 0x16:
0047     __le32 name_bytes;  // 0x1C: In bytes.
0048     __le16 name[32];    // 0x20: Name of client.
0049 };
0050 
0051 static_assert(sizeof(struct CLIENT_REC) == 0x60);
0052 
0053 /* Two copies of these will exist at the beginning of the log file */
0054 struct RESTART_AREA {
0055     __le64 current_lsn;    // 0x00: Current logical end of log file.
0056     __le16 log_clients;    // 0x08: Maximum number of clients.
0057     __le16 client_idx[2];  // 0x0A: Free/use index into the client record arrays.
0058     __le16 flags;          // 0x0E: See RESTART_SINGLE_PAGE_IO.
0059     __le32 seq_num_bits;   // 0x10: The number of bits in sequence number.
0060     __le16 ra_len;         // 0x14:
0061     __le16 client_off;     // 0x16:
0062     __le64 l_size;         // 0x18: Usable log file size.
0063     __le32 last_lsn_data_len; // 0x20:
0064     __le16 rec_hdr_len;    // 0x24: Log page data offset.
0065     __le16 data_off;       // 0x26: Log page data length.
0066     __le32 open_log_count; // 0x28:
0067     __le32 align[5];       // 0x2C:
0068     struct CLIENT_REC clients[]; // 0x40:
0069 };
0070 
0071 struct LOG_REC_HDR {
0072     __le16 redo_op;      // 0x00:  NTFS_LOG_OPERATION
0073     __le16 undo_op;      // 0x02:  NTFS_LOG_OPERATION
0074     __le16 redo_off;     // 0x04:  Offset to Redo record.
0075     __le16 redo_len;     // 0x06:  Redo length.
0076     __le16 undo_off;     // 0x08:  Offset to Undo record.
0077     __le16 undo_len;     // 0x0A:  Undo length.
0078     __le16 target_attr;  // 0x0C:
0079     __le16 lcns_follow;  // 0x0E:
0080     __le16 record_off;   // 0x10:
0081     __le16 attr_off;     // 0x12:
0082     __le16 cluster_off;  // 0x14:
0083     __le16 reserved;     // 0x16:
0084     __le64 target_vcn;   // 0x18:
0085     __le64 page_lcns[];  // 0x20:
0086 };
0087 
0088 static_assert(sizeof(struct LOG_REC_HDR) == 0x20);
0089 
0090 #define RESTART_ENTRY_ALLOCATED    0xFFFFFFFF
0091 #define RESTART_ENTRY_ALLOCATED_LE cpu_to_le32(0xFFFFFFFF)
0092 
0093 struct RESTART_TABLE {
0094     __le16 size;       // 0x00: In bytes
0095     __le16 used;       // 0x02: Entries
0096     __le16 total;      // 0x04: Entries
0097     __le16 res[3];     // 0x06:
0098     __le32 free_goal;  // 0x0C:
0099     __le32 first_free; // 0x10:
0100     __le32 last_free;  // 0x14:
0101 
0102 };
0103 
0104 static_assert(sizeof(struct RESTART_TABLE) == 0x18);
0105 
0106 struct ATTR_NAME_ENTRY {
0107     __le16 off; // Offset in the Open attribute Table.
0108     __le16 name_bytes;
0109     __le16 name[];
0110 };
0111 
0112 struct OPEN_ATTR_ENRTY {
0113     __le32 next;            // 0x00: RESTART_ENTRY_ALLOCATED if allocated
0114     __le32 bytes_per_index; // 0x04:
0115     enum ATTR_TYPE type;    // 0x08:
0116     u8 is_dirty_pages;      // 0x0C:
0117     u8 is_attr_name;        // 0x0B: Faked field to manage 'ptr'
0118     u8 name_len;            // 0x0C: Faked field to manage 'ptr'
0119     u8 res;
0120     struct MFT_REF ref;     // 0x10: File Reference of file containing attribute
0121     __le64 open_record_lsn; // 0x18:
0122     void *ptr;              // 0x20:
0123 };
0124 
0125 /* 32 bit version of 'struct OPEN_ATTR_ENRTY' */
0126 struct OPEN_ATTR_ENRTY_32 {
0127     __le32 next;            // 0x00: RESTART_ENTRY_ALLOCATED if allocated
0128     __le32 ptr;             // 0x04:
0129     struct MFT_REF ref;     // 0x08:
0130     __le64 open_record_lsn; // 0x10:
0131     u8 is_dirty_pages;      // 0x18:
0132     u8 is_attr_name;        // 0x19:
0133     u8 res1[2];
0134     enum ATTR_TYPE type;    // 0x1C:
0135     u8 name_len;            // 0x20: In wchar
0136     u8 res2[3];
0137     __le32 AttributeName;   // 0x24:
0138     __le32 bytes_per_index; // 0x28:
0139 };
0140 
0141 #define SIZEOF_OPENATTRIBUTEENTRY0 0x2c
0142 // static_assert( 0x2C == sizeof(struct OPEN_ATTR_ENRTY_32) );
0143 static_assert(sizeof(struct OPEN_ATTR_ENRTY) < SIZEOF_OPENATTRIBUTEENTRY0);
0144 
0145 /*
0146  * One entry exists in the Dirty Pages Table for each page which is dirty at
0147  * the time the Restart Area is written.
0148  */
0149 struct DIR_PAGE_ENTRY {
0150     __le32 next;         // 0x00: RESTART_ENTRY_ALLOCATED if allocated
0151     __le32 target_attr;  // 0x04: Index into the Open attribute Table
0152     __le32 transfer_len; // 0x08:
0153     __le32 lcns_follow;  // 0x0C:
0154     __le64 vcn;          // 0x10: Vcn of dirty page
0155     __le64 oldest_lsn;   // 0x18:
0156     __le64 page_lcns[];  // 0x20:
0157 };
0158 
0159 static_assert(sizeof(struct DIR_PAGE_ENTRY) == 0x20);
0160 
0161 /* 32 bit version of 'struct DIR_PAGE_ENTRY' */
0162 struct DIR_PAGE_ENTRY_32 {
0163     __le32 next;        // 0x00: RESTART_ENTRY_ALLOCATED if allocated
0164     __le32 target_attr; // 0x04: Index into the Open attribute Table
0165     __le32 transfer_len;    // 0x08:
0166     __le32 lcns_follow; // 0x0C:
0167     __le32 reserved;    // 0x10:
0168     __le32 vcn_low;     // 0x14: Vcn of dirty page
0169     __le32 vcn_hi;      // 0x18: Vcn of dirty page
0170     __le32 oldest_lsn_low;  // 0x1C:
0171     __le32 oldest_lsn_hi;   // 0x1C:
0172     __le32 page_lcns_low;   // 0x24:
0173     __le32 page_lcns_hi;    // 0x24:
0174 };
0175 
0176 static_assert(offsetof(struct DIR_PAGE_ENTRY_32, vcn_low) == 0x14);
0177 static_assert(sizeof(struct DIR_PAGE_ENTRY_32) == 0x2c);
0178 
0179 enum transact_state {
0180     TransactionUninitialized = 0,
0181     TransactionActive,
0182     TransactionPrepared,
0183     TransactionCommitted
0184 };
0185 
0186 struct TRANSACTION_ENTRY {
0187     __le32 next;          // 0x00: RESTART_ENTRY_ALLOCATED if allocated
0188     u8 transact_state;    // 0x04:
0189     u8 reserved[3];       // 0x05:
0190     __le64 first_lsn;     // 0x08:
0191     __le64 prev_lsn;      // 0x10:
0192     __le64 undo_next_lsn; // 0x18:
0193     __le32 undo_records;  // 0x20: Number of undo log records pending abort
0194     __le32 undo_len;      // 0x24: Total undo size
0195 };
0196 
0197 static_assert(sizeof(struct TRANSACTION_ENTRY) == 0x28);
0198 
0199 struct NTFS_RESTART {
0200     __le32 major_ver;             // 0x00:
0201     __le32 minor_ver;             // 0x04:
0202     __le64 check_point_start;     // 0x08:
0203     __le64 open_attr_table_lsn;   // 0x10:
0204     __le64 attr_names_lsn;        // 0x18:
0205     __le64 dirty_pages_table_lsn; // 0x20:
0206     __le64 transact_table_lsn;    // 0x28:
0207     __le32 open_attr_len;         // 0x30: In bytes
0208     __le32 attr_names_len;        // 0x34: In bytes
0209     __le32 dirty_pages_len;       // 0x38: In bytes
0210     __le32 transact_table_len;    // 0x3C: In bytes
0211 };
0212 
0213 static_assert(sizeof(struct NTFS_RESTART) == 0x40);
0214 
0215 struct NEW_ATTRIBUTE_SIZES {
0216     __le64 alloc_size;
0217     __le64 valid_size;
0218     __le64 data_size;
0219     __le64 total_size;
0220 };
0221 
0222 struct BITMAP_RANGE {
0223     __le32 bitmap_off;
0224     __le32 bits;
0225 };
0226 
0227 struct LCN_RANGE {
0228     __le64 lcn;
0229     __le64 len;
0230 };
0231 
0232 /* The following type defines the different log record types. */
0233 #define LfsClientRecord  cpu_to_le32(1)
0234 #define LfsClientRestart cpu_to_le32(2)
0235 
0236 /* This is used to uniquely identify a client for a particular log file. */
0237 struct CLIENT_ID {
0238     __le16 seq_num;
0239     __le16 client_idx;
0240 };
0241 
0242 /* This is the header that begins every Log Record in the log file. */
0243 struct LFS_RECORD_HDR {
0244     __le64 this_lsn;        // 0x00:
0245     __le64 client_prev_lsn;     // 0x08:
0246     __le64 client_undo_next_lsn;    // 0x10:
0247     __le32 client_data_len;     // 0x18:
0248     struct CLIENT_ID client;    // 0x1C: Owner of this log record.
0249     __le32 record_type;     // 0x20: LfsClientRecord or LfsClientRestart.
0250     __le32 transact_id;     // 0x24:
0251     __le16 flags;           // 0x28: LOG_RECORD_MULTI_PAGE
0252     u8 align[6];            // 0x2A:
0253 };
0254 
0255 #define LOG_RECORD_MULTI_PAGE cpu_to_le16(1)
0256 
0257 static_assert(sizeof(struct LFS_RECORD_HDR) == 0x30);
0258 
0259 struct LFS_RECORD {
0260     __le16 next_record_off; // 0x00: Offset of the free space in the page,
0261     u8 align[6];        // 0x02:
0262     __le64 last_end_lsn;    // 0x08: lsn for the last log record which ends on the page,
0263 };
0264 
0265 static_assert(sizeof(struct LFS_RECORD) == 0x10);
0266 
0267 struct RECORD_PAGE_HDR {
0268     struct NTFS_RECORD_HEADER rhdr; // 'RCRD'
0269     __le32 rflags;          // 0x10: See LOG_PAGE_LOG_RECORD_END
0270     __le16 page_count;      // 0x14:
0271     __le16 page_pos;        // 0x16:
0272     struct LFS_RECORD record_hdr;   // 0x18:
0273     __le16 fixups[10];      // 0x28:
0274     __le32 file_off;        // 0x3c: Used when major version >= 2
0275 };
0276 
0277 // clang-format on
0278 
0279 // Page contains the end of a log record.
0280 #define LOG_PAGE_LOG_RECORD_END cpu_to_le32(0x00000001)
0281 
0282 static inline bool is_log_record_end(const struct RECORD_PAGE_HDR *hdr)
0283 {
0284     return hdr->rflags & LOG_PAGE_LOG_RECORD_END;
0285 }
0286 
0287 static_assert(offsetof(struct RECORD_PAGE_HDR, file_off) == 0x3c);
0288 
0289 /*
0290  * END of NTFS LOG structures
0291  */
0292 
0293 /* Define some tuning parameters to keep the restart tables a reasonable size. */
0294 #define INITIAL_NUMBER_TRANSACTIONS 5
0295 
0296 enum NTFS_LOG_OPERATION {
0297 
0298     Noop = 0x00,
0299     CompensationLogRecord = 0x01,
0300     InitializeFileRecordSegment = 0x02,
0301     DeallocateFileRecordSegment = 0x03,
0302     WriteEndOfFileRecordSegment = 0x04,
0303     CreateAttribute = 0x05,
0304     DeleteAttribute = 0x06,
0305     UpdateResidentValue = 0x07,
0306     UpdateNonresidentValue = 0x08,
0307     UpdateMappingPairs = 0x09,
0308     DeleteDirtyClusters = 0x0A,
0309     SetNewAttributeSizes = 0x0B,
0310     AddIndexEntryRoot = 0x0C,
0311     DeleteIndexEntryRoot = 0x0D,
0312     AddIndexEntryAllocation = 0x0E,
0313     DeleteIndexEntryAllocation = 0x0F,
0314     WriteEndOfIndexBuffer = 0x10,
0315     SetIndexEntryVcnRoot = 0x11,
0316     SetIndexEntryVcnAllocation = 0x12,
0317     UpdateFileNameRoot = 0x13,
0318     UpdateFileNameAllocation = 0x14,
0319     SetBitsInNonresidentBitMap = 0x15,
0320     ClearBitsInNonresidentBitMap = 0x16,
0321     HotFix = 0x17,
0322     EndTopLevelAction = 0x18,
0323     PrepareTransaction = 0x19,
0324     CommitTransaction = 0x1A,
0325     ForgetTransaction = 0x1B,
0326     OpenNonresidentAttribute = 0x1C,
0327     OpenAttributeTableDump = 0x1D,
0328     AttributeNamesDump = 0x1E,
0329     DirtyPageTableDump = 0x1F,
0330     TransactionTableDump = 0x20,
0331     UpdateRecordDataRoot = 0x21,
0332     UpdateRecordDataAllocation = 0x22,
0333 
0334     UpdateRelativeDataInIndex =
0335         0x23, // NtOfsRestartUpdateRelativeDataInIndex
0336     UpdateRelativeDataInIndex2 = 0x24,
0337     ZeroEndOfFileRecord = 0x25,
0338 };
0339 
0340 /*
0341  * Array for log records which require a target attribute.
0342  * A true indicates that the corresponding restart operation
0343  * requires a target attribute.
0344  */
0345 static const u8 AttributeRequired[] = {
0346     0xFC, 0xFB, 0xFF, 0x10, 0x06,
0347 };
0348 
0349 static inline bool is_target_required(u16 op)
0350 {
0351     bool ret = op <= UpdateRecordDataAllocation &&
0352            (AttributeRequired[op >> 3] >> (op & 7) & 1);
0353     return ret;
0354 }
0355 
0356 static inline bool can_skip_action(enum NTFS_LOG_OPERATION op)
0357 {
0358     switch (op) {
0359     case Noop:
0360     case DeleteDirtyClusters:
0361     case HotFix:
0362     case EndTopLevelAction:
0363     case PrepareTransaction:
0364     case CommitTransaction:
0365     case ForgetTransaction:
0366     case CompensationLogRecord:
0367     case OpenNonresidentAttribute:
0368     case OpenAttributeTableDump:
0369     case AttributeNamesDump:
0370     case DirtyPageTableDump:
0371     case TransactionTableDump:
0372         return true;
0373     default:
0374         return false;
0375     }
0376 }
0377 
0378 enum { lcb_ctx_undo_next, lcb_ctx_prev, lcb_ctx_next };
0379 
0380 /* Bytes per restart table. */
0381 static inline u32 bytes_per_rt(const struct RESTART_TABLE *rt)
0382 {
0383     return le16_to_cpu(rt->used) * le16_to_cpu(rt->size) +
0384            sizeof(struct RESTART_TABLE);
0385 }
0386 
0387 /* Log record length. */
0388 static inline u32 lrh_length(const struct LOG_REC_HDR *lr)
0389 {
0390     u16 t16 = le16_to_cpu(lr->lcns_follow);
0391 
0392     return struct_size(lr, page_lcns, max_t(u16, 1, t16));
0393 }
0394 
0395 struct lcb {
0396     struct LFS_RECORD_HDR *lrh; // Log record header of the current lsn.
0397     struct LOG_REC_HDR *log_rec;
0398     u32 ctx_mode; // lcb_ctx_undo_next/lcb_ctx_prev/lcb_ctx_next
0399     struct CLIENT_ID client;
0400     bool alloc; // If true the we should deallocate 'log_rec'.
0401 };
0402 
0403 static void lcb_put(struct lcb *lcb)
0404 {
0405     if (lcb->alloc)
0406         kfree(lcb->log_rec);
0407     kfree(lcb->lrh);
0408     kfree(lcb);
0409 }
0410 
0411 /* Find the oldest lsn from active clients. */
0412 static inline void oldest_client_lsn(const struct CLIENT_REC *ca,
0413                      __le16 next_client, u64 *oldest_lsn)
0414 {
0415     while (next_client != LFS_NO_CLIENT_LE) {
0416         const struct CLIENT_REC *cr = ca + le16_to_cpu(next_client);
0417         u64 lsn = le64_to_cpu(cr->oldest_lsn);
0418 
0419         /* Ignore this block if it's oldest lsn is 0. */
0420         if (lsn && lsn < *oldest_lsn)
0421             *oldest_lsn = lsn;
0422 
0423         next_client = cr->next_client;
0424     }
0425 }
0426 
0427 static inline bool is_rst_page_hdr_valid(u32 file_off,
0428                      const struct RESTART_HDR *rhdr)
0429 {
0430     u32 sys_page = le32_to_cpu(rhdr->sys_page_size);
0431     u32 page_size = le32_to_cpu(rhdr->page_size);
0432     u32 end_usa;
0433     u16 ro;
0434 
0435     if (sys_page < SECTOR_SIZE || page_size < SECTOR_SIZE ||
0436         sys_page & (sys_page - 1) || page_size & (page_size - 1)) {
0437         return false;
0438     }
0439 
0440     /* Check that if the file offset isn't 0, it is the system page size. */
0441     if (file_off && file_off != sys_page)
0442         return false;
0443 
0444     /* Check support version 1.1+. */
0445     if (le16_to_cpu(rhdr->major_ver) <= 1 && !rhdr->minor_ver)
0446         return false;
0447 
0448     if (le16_to_cpu(rhdr->major_ver) > 2)
0449         return false;
0450 
0451     ro = le16_to_cpu(rhdr->ra_off);
0452     if (!IS_ALIGNED(ro, 8) || ro > sys_page)
0453         return false;
0454 
0455     end_usa = ((sys_page >> SECTOR_SHIFT) + 1) * sizeof(short);
0456     end_usa += le16_to_cpu(rhdr->rhdr.fix_off);
0457 
0458     if (ro < end_usa)
0459         return false;
0460 
0461     return true;
0462 }
0463 
0464 static inline bool is_rst_area_valid(const struct RESTART_HDR *rhdr)
0465 {
0466     const struct RESTART_AREA *ra;
0467     u16 cl, fl, ul;
0468     u32 off, l_size, file_dat_bits, file_size_round;
0469     u16 ro = le16_to_cpu(rhdr->ra_off);
0470     u32 sys_page = le32_to_cpu(rhdr->sys_page_size);
0471 
0472     if (ro + offsetof(struct RESTART_AREA, l_size) >
0473         SECTOR_SIZE - sizeof(short))
0474         return false;
0475 
0476     ra = Add2Ptr(rhdr, ro);
0477     cl = le16_to_cpu(ra->log_clients);
0478 
0479     if (cl > 1)
0480         return false;
0481 
0482     off = le16_to_cpu(ra->client_off);
0483 
0484     if (!IS_ALIGNED(off, 8) || ro + off > SECTOR_SIZE - sizeof(short))
0485         return false;
0486 
0487     off += cl * sizeof(struct CLIENT_REC);
0488 
0489     if (off > sys_page)
0490         return false;
0491 
0492     /*
0493      * Check the restart length field and whether the entire
0494      * restart area is contained that length.
0495      */
0496     if (le16_to_cpu(rhdr->ra_off) + le16_to_cpu(ra->ra_len) > sys_page ||
0497         off > le16_to_cpu(ra->ra_len)) {
0498         return false;
0499     }
0500 
0501     /*
0502      * As a final check make sure that the use list and the free list
0503      * are either empty or point to a valid client.
0504      */
0505     fl = le16_to_cpu(ra->client_idx[0]);
0506     ul = le16_to_cpu(ra->client_idx[1]);
0507     if ((fl != LFS_NO_CLIENT && fl >= cl) ||
0508         (ul != LFS_NO_CLIENT && ul >= cl))
0509         return false;
0510 
0511     /* Make sure the sequence number bits match the log file size. */
0512     l_size = le64_to_cpu(ra->l_size);
0513 
0514     file_dat_bits = sizeof(u64) * 8 - le32_to_cpu(ra->seq_num_bits);
0515     file_size_round = 1u << (file_dat_bits + 3);
0516     if (file_size_round != l_size &&
0517         (file_size_round < l_size || (file_size_round / 2) > l_size)) {
0518         return false;
0519     }
0520 
0521     /* The log page data offset and record header length must be quad-aligned. */
0522     if (!IS_ALIGNED(le16_to_cpu(ra->data_off), 8) ||
0523         !IS_ALIGNED(le16_to_cpu(ra->rec_hdr_len), 8))
0524         return false;
0525 
0526     return true;
0527 }
0528 
0529 static inline bool is_client_area_valid(const struct RESTART_HDR *rhdr,
0530                     bool usa_error)
0531 {
0532     u16 ro = le16_to_cpu(rhdr->ra_off);
0533     const struct RESTART_AREA *ra = Add2Ptr(rhdr, ro);
0534     u16 ra_len = le16_to_cpu(ra->ra_len);
0535     const struct CLIENT_REC *ca;
0536     u32 i;
0537 
0538     if (usa_error && ra_len + ro > SECTOR_SIZE - sizeof(short))
0539         return false;
0540 
0541     /* Find the start of the client array. */
0542     ca = Add2Ptr(ra, le16_to_cpu(ra->client_off));
0543 
0544     /*
0545      * Start with the free list.
0546      * Check that all the clients are valid and that there isn't a cycle.
0547      * Do the in-use list on the second pass.
0548      */
0549     for (i = 0; i < 2; i++) {
0550         u16 client_idx = le16_to_cpu(ra->client_idx[i]);
0551         bool first_client = true;
0552         u16 clients = le16_to_cpu(ra->log_clients);
0553 
0554         while (client_idx != LFS_NO_CLIENT) {
0555             const struct CLIENT_REC *cr;
0556 
0557             if (!clients ||
0558                 client_idx >= le16_to_cpu(ra->log_clients))
0559                 return false;
0560 
0561             clients -= 1;
0562             cr = ca + client_idx;
0563 
0564             client_idx = le16_to_cpu(cr->next_client);
0565 
0566             if (first_client) {
0567                 first_client = false;
0568                 if (cr->prev_client != LFS_NO_CLIENT_LE)
0569                     return false;
0570             }
0571         }
0572     }
0573 
0574     return true;
0575 }
0576 
0577 /*
0578  * remove_client
0579  *
0580  * Remove a client record from a client record list an restart area.
0581  */
0582 static inline void remove_client(struct CLIENT_REC *ca,
0583                  const struct CLIENT_REC *cr, __le16 *head)
0584 {
0585     if (cr->prev_client == LFS_NO_CLIENT_LE)
0586         *head = cr->next_client;
0587     else
0588         ca[le16_to_cpu(cr->prev_client)].next_client = cr->next_client;
0589 
0590     if (cr->next_client != LFS_NO_CLIENT_LE)
0591         ca[le16_to_cpu(cr->next_client)].prev_client = cr->prev_client;
0592 }
0593 
0594 /*
0595  * add_client - Add a client record to the start of a list.
0596  */
0597 static inline void add_client(struct CLIENT_REC *ca, u16 index, __le16 *head)
0598 {
0599     struct CLIENT_REC *cr = ca + index;
0600 
0601     cr->prev_client = LFS_NO_CLIENT_LE;
0602     cr->next_client = *head;
0603 
0604     if (*head != LFS_NO_CLIENT_LE)
0605         ca[le16_to_cpu(*head)].prev_client = cpu_to_le16(index);
0606 
0607     *head = cpu_to_le16(index);
0608 }
0609 
0610 static inline void *enum_rstbl(struct RESTART_TABLE *t, void *c)
0611 {
0612     __le32 *e;
0613     u32 bprt;
0614     u16 rsize = t ? le16_to_cpu(t->size) : 0;
0615 
0616     if (!c) {
0617         if (!t || !t->total)
0618             return NULL;
0619         e = Add2Ptr(t, sizeof(struct RESTART_TABLE));
0620     } else {
0621         e = Add2Ptr(c, rsize);
0622     }
0623 
0624     /* Loop until we hit the first one allocated, or the end of the list. */
0625     for (bprt = bytes_per_rt(t); PtrOffset(t, e) < bprt;
0626          e = Add2Ptr(e, rsize)) {
0627         if (*e == RESTART_ENTRY_ALLOCATED_LE)
0628             return e;
0629     }
0630     return NULL;
0631 }
0632 
0633 /*
0634  * find_dp - Search for a @vcn in Dirty Page Table.
0635  */
0636 static inline struct DIR_PAGE_ENTRY *find_dp(struct RESTART_TABLE *dptbl,
0637                          u32 target_attr, u64 vcn)
0638 {
0639     __le32 ta = cpu_to_le32(target_attr);
0640     struct DIR_PAGE_ENTRY *dp = NULL;
0641 
0642     while ((dp = enum_rstbl(dptbl, dp))) {
0643         u64 dp_vcn = le64_to_cpu(dp->vcn);
0644 
0645         if (dp->target_attr == ta && vcn >= dp_vcn &&
0646             vcn < dp_vcn + le32_to_cpu(dp->lcns_follow)) {
0647             return dp;
0648         }
0649     }
0650     return NULL;
0651 }
0652 
0653 static inline u32 norm_file_page(u32 page_size, u32 *l_size, bool use_default)
0654 {
0655     if (use_default)
0656         page_size = DefaultLogPageSize;
0657 
0658     /* Round the file size down to a system page boundary. */
0659     *l_size &= ~(page_size - 1);
0660 
0661     /* File should contain at least 2 restart pages and MinLogRecordPages pages. */
0662     if (*l_size < (MinLogRecordPages + 2) * page_size)
0663         return 0;
0664 
0665     return page_size;
0666 }
0667 
0668 static bool check_log_rec(const struct LOG_REC_HDR *lr, u32 bytes, u32 tr,
0669               u32 bytes_per_attr_entry)
0670 {
0671     u16 t16;
0672 
0673     if (bytes < sizeof(struct LOG_REC_HDR))
0674         return false;
0675     if (!tr)
0676         return false;
0677 
0678     if ((tr - sizeof(struct RESTART_TABLE)) %
0679         sizeof(struct TRANSACTION_ENTRY))
0680         return false;
0681 
0682     if (le16_to_cpu(lr->redo_off) & 7)
0683         return false;
0684 
0685     if (le16_to_cpu(lr->undo_off) & 7)
0686         return false;
0687 
0688     if (lr->target_attr)
0689         goto check_lcns;
0690 
0691     if (is_target_required(le16_to_cpu(lr->redo_op)))
0692         return false;
0693 
0694     if (is_target_required(le16_to_cpu(lr->undo_op)))
0695         return false;
0696 
0697 check_lcns:
0698     if (!lr->lcns_follow)
0699         goto check_length;
0700 
0701     t16 = le16_to_cpu(lr->target_attr);
0702     if ((t16 - sizeof(struct RESTART_TABLE)) % bytes_per_attr_entry)
0703         return false;
0704 
0705 check_length:
0706     if (bytes < lrh_length(lr))
0707         return false;
0708 
0709     return true;
0710 }
0711 
0712 static bool check_rstbl(const struct RESTART_TABLE *rt, size_t bytes)
0713 {
0714     u32 ts;
0715     u32 i, off;
0716     u16 rsize = le16_to_cpu(rt->size);
0717     u16 ne = le16_to_cpu(rt->used);
0718     u32 ff = le32_to_cpu(rt->first_free);
0719     u32 lf = le32_to_cpu(rt->last_free);
0720 
0721     ts = rsize * ne + sizeof(struct RESTART_TABLE);
0722 
0723     if (!rsize || rsize > bytes ||
0724         rsize + sizeof(struct RESTART_TABLE) > bytes || bytes < ts ||
0725         le16_to_cpu(rt->total) > ne || ff > ts || lf > ts ||
0726         (ff && ff < sizeof(struct RESTART_TABLE)) ||
0727         (lf && lf < sizeof(struct RESTART_TABLE))) {
0728         return false;
0729     }
0730 
0731     /*
0732      * Verify each entry is either allocated or points
0733      * to a valid offset the table.
0734      */
0735     for (i = 0; i < ne; i++) {
0736         off = le32_to_cpu(*(__le32 *)Add2Ptr(
0737             rt, i * rsize + sizeof(struct RESTART_TABLE)));
0738 
0739         if (off != RESTART_ENTRY_ALLOCATED && off &&
0740             (off < sizeof(struct RESTART_TABLE) ||
0741              ((off - sizeof(struct RESTART_TABLE)) % rsize))) {
0742             return false;
0743         }
0744     }
0745 
0746     /*
0747      * Walk through the list headed by the first entry to make
0748      * sure none of the entries are currently being used.
0749      */
0750     for (off = ff; off;) {
0751         if (off == RESTART_ENTRY_ALLOCATED)
0752             return false;
0753 
0754         off = le32_to_cpu(*(__le32 *)Add2Ptr(rt, off));
0755     }
0756 
0757     return true;
0758 }
0759 
0760 /*
0761  * free_rsttbl_idx - Free a previously allocated index a Restart Table.
0762  */
0763 static inline void free_rsttbl_idx(struct RESTART_TABLE *rt, u32 off)
0764 {
0765     __le32 *e;
0766     u32 lf = le32_to_cpu(rt->last_free);
0767     __le32 off_le = cpu_to_le32(off);
0768 
0769     e = Add2Ptr(rt, off);
0770 
0771     if (off < le32_to_cpu(rt->free_goal)) {
0772         *e = rt->first_free;
0773         rt->first_free = off_le;
0774         if (!lf)
0775             rt->last_free = off_le;
0776     } else {
0777         if (lf)
0778             *(__le32 *)Add2Ptr(rt, lf) = off_le;
0779         else
0780             rt->first_free = off_le;
0781 
0782         rt->last_free = off_le;
0783         *e = 0;
0784     }
0785 
0786     le16_sub_cpu(&rt->total, 1);
0787 }
0788 
0789 static inline struct RESTART_TABLE *init_rsttbl(u16 esize, u16 used)
0790 {
0791     __le32 *e, *last_free;
0792     u32 off;
0793     u32 bytes = esize * used + sizeof(struct RESTART_TABLE);
0794     u32 lf = sizeof(struct RESTART_TABLE) + (used - 1) * esize;
0795     struct RESTART_TABLE *t = kzalloc(bytes, GFP_NOFS);
0796 
0797     if (!t)
0798         return NULL;
0799 
0800     t->size = cpu_to_le16(esize);
0801     t->used = cpu_to_le16(used);
0802     t->free_goal = cpu_to_le32(~0u);
0803     t->first_free = cpu_to_le32(sizeof(struct RESTART_TABLE));
0804     t->last_free = cpu_to_le32(lf);
0805 
0806     e = (__le32 *)(t + 1);
0807     last_free = Add2Ptr(t, lf);
0808 
0809     for (off = sizeof(struct RESTART_TABLE) + esize; e < last_free;
0810          e = Add2Ptr(e, esize), off += esize) {
0811         *e = cpu_to_le32(off);
0812     }
0813     return t;
0814 }
0815 
0816 static inline struct RESTART_TABLE *extend_rsttbl(struct RESTART_TABLE *tbl,
0817                           u32 add, u32 free_goal)
0818 {
0819     u16 esize = le16_to_cpu(tbl->size);
0820     __le32 osize = cpu_to_le32(bytes_per_rt(tbl));
0821     u32 used = le16_to_cpu(tbl->used);
0822     struct RESTART_TABLE *rt;
0823 
0824     rt = init_rsttbl(esize, used + add);
0825     if (!rt)
0826         return NULL;
0827 
0828     memcpy(rt + 1, tbl + 1, esize * used);
0829 
0830     rt->free_goal = free_goal == ~0u
0831                 ? cpu_to_le32(~0u)
0832                 : cpu_to_le32(sizeof(struct RESTART_TABLE) +
0833                           free_goal * esize);
0834 
0835     if (tbl->first_free) {
0836         rt->first_free = tbl->first_free;
0837         *(__le32 *)Add2Ptr(rt, le32_to_cpu(tbl->last_free)) = osize;
0838     } else {
0839         rt->first_free = osize;
0840     }
0841 
0842     rt->total = tbl->total;
0843 
0844     kfree(tbl);
0845     return rt;
0846 }
0847 
0848 /*
0849  * alloc_rsttbl_idx
0850  *
0851  * Allocate an index from within a previously initialized Restart Table.
0852  */
0853 static inline void *alloc_rsttbl_idx(struct RESTART_TABLE **tbl)
0854 {
0855     u32 off;
0856     __le32 *e;
0857     struct RESTART_TABLE *t = *tbl;
0858 
0859     if (!t->first_free) {
0860         *tbl = t = extend_rsttbl(t, 16, ~0u);
0861         if (!t)
0862             return NULL;
0863     }
0864 
0865     off = le32_to_cpu(t->first_free);
0866 
0867     /* Dequeue this entry and zero it. */
0868     e = Add2Ptr(t, off);
0869 
0870     t->first_free = *e;
0871 
0872     memset(e, 0, le16_to_cpu(t->size));
0873 
0874     *e = RESTART_ENTRY_ALLOCATED_LE;
0875 
0876     /* If list is going empty, then we fix the last_free as well. */
0877     if (!t->first_free)
0878         t->last_free = 0;
0879 
0880     le16_add_cpu(&t->total, 1);
0881 
0882     return Add2Ptr(t, off);
0883 }
0884 
0885 /*
0886  * alloc_rsttbl_from_idx
0887  *
0888  * Allocate a specific index from within a previously initialized Restart Table.
0889  */
0890 static inline void *alloc_rsttbl_from_idx(struct RESTART_TABLE **tbl, u32 vbo)
0891 {
0892     u32 off;
0893     __le32 *e;
0894     struct RESTART_TABLE *rt = *tbl;
0895     u32 bytes = bytes_per_rt(rt);
0896     u16 esize = le16_to_cpu(rt->size);
0897 
0898     /* If the entry is not the table, we will have to extend the table. */
0899     if (vbo >= bytes) {
0900         /*
0901          * Extend the size by computing the number of entries between
0902          * the existing size and the desired index and adding 1 to that.
0903          */
0904         u32 bytes2idx = vbo - bytes;
0905 
0906         /*
0907          * There should always be an integral number of entries
0908          * being added. Now extend the table.
0909          */
0910         *tbl = rt = extend_rsttbl(rt, bytes2idx / esize + 1, bytes);
0911         if (!rt)
0912             return NULL;
0913     }
0914 
0915     /* See if the entry is already allocated, and just return if it is. */
0916     e = Add2Ptr(rt, vbo);
0917 
0918     if (*e == RESTART_ENTRY_ALLOCATED_LE)
0919         return e;
0920 
0921     /*
0922      * Walk through the table, looking for the entry we're
0923      * interested and the previous entry.
0924      */
0925     off = le32_to_cpu(rt->first_free);
0926     e = Add2Ptr(rt, off);
0927 
0928     if (off == vbo) {
0929         /* this is a match */
0930         rt->first_free = *e;
0931         goto skip_looking;
0932     }
0933 
0934     /*
0935      * Need to walk through the list looking for the predecessor
0936      * of our entry.
0937      */
0938     for (;;) {
0939         /* Remember the entry just found */
0940         u32 last_off = off;
0941         __le32 *last_e = e;
0942 
0943         /* Should never run of entries. */
0944 
0945         /* Lookup up the next entry the list. */
0946         off = le32_to_cpu(*last_e);
0947         e = Add2Ptr(rt, off);
0948 
0949         /* If this is our match we are done. */
0950         if (off == vbo) {
0951             *last_e = *e;
0952 
0953             /*
0954              * If this was the last entry, we update that
0955              * table as well.
0956              */
0957             if (le32_to_cpu(rt->last_free) == off)
0958                 rt->last_free = cpu_to_le32(last_off);
0959             break;
0960         }
0961     }
0962 
0963 skip_looking:
0964     /* If the list is now empty, we fix the last_free as well. */
0965     if (!rt->first_free)
0966         rt->last_free = 0;
0967 
0968     /* Zero this entry. */
0969     memset(e, 0, esize);
0970     *e = RESTART_ENTRY_ALLOCATED_LE;
0971 
0972     le16_add_cpu(&rt->total, 1);
0973 
0974     return e;
0975 }
0976 
0977 #define RESTART_SINGLE_PAGE_IO cpu_to_le16(0x0001)
0978 
0979 #define NTFSLOG_WRAPPED 0x00000001
0980 #define NTFSLOG_MULTIPLE_PAGE_IO 0x00000002
0981 #define NTFSLOG_NO_LAST_LSN 0x00000004
0982 #define NTFSLOG_REUSE_TAIL 0x00000010
0983 #define NTFSLOG_NO_OLDEST_LSN 0x00000020
0984 
0985 /* Helper struct to work with NTFS $LogFile. */
0986 struct ntfs_log {
0987     struct ntfs_inode *ni;
0988 
0989     u32 l_size;
0990     u32 sys_page_size;
0991     u32 sys_page_mask;
0992     u32 page_size;
0993     u32 page_mask; // page_size - 1
0994     u8 page_bits;
0995     struct RECORD_PAGE_HDR *one_page_buf;
0996 
0997     struct RESTART_TABLE *open_attr_tbl;
0998     u32 transaction_id;
0999     u32 clst_per_page;
1000 
1001     u32 first_page;
1002     u32 next_page;
1003     u32 ra_off;
1004     u32 data_off;
1005     u32 restart_size;
1006     u32 data_size;
1007     u16 record_header_len;
1008     u64 seq_num;
1009     u32 seq_num_bits;
1010     u32 file_data_bits;
1011     u32 seq_num_mask; /* (1 << file_data_bits) - 1 */
1012 
1013     struct RESTART_AREA *ra; /* In-memory image of the next restart area. */
1014     u32 ra_size; /* The usable size of the restart area. */
1015 
1016     /*
1017      * If true, then the in-memory restart area is to be written
1018      * to the first position on the disk.
1019      */
1020     bool init_ra;
1021     bool set_dirty; /* True if we need to set dirty flag. */
1022 
1023     u64 oldest_lsn;
1024 
1025     u32 oldest_lsn_off;
1026     u64 last_lsn;
1027 
1028     u32 total_avail;
1029     u32 total_avail_pages;
1030     u32 total_undo_commit;
1031     u32 max_current_avail;
1032     u32 current_avail;
1033     u32 reserved;
1034 
1035     short major_ver;
1036     short minor_ver;
1037 
1038     u32 l_flags; /* See NTFSLOG_XXX */
1039     u32 current_openlog_count; /* On-disk value for open_log_count. */
1040 
1041     struct CLIENT_ID client_id;
1042     u32 client_undo_commit;
1043 };
1044 
1045 static inline u32 lsn_to_vbo(struct ntfs_log *log, const u64 lsn)
1046 {
1047     u32 vbo = (lsn << log->seq_num_bits) >> (log->seq_num_bits - 3);
1048 
1049     return vbo;
1050 }
1051 
1052 /* Compute the offset in the log file of the next log page. */
1053 static inline u32 next_page_off(struct ntfs_log *log, u32 off)
1054 {
1055     off = (off & ~log->sys_page_mask) + log->page_size;
1056     return off >= log->l_size ? log->first_page : off;
1057 }
1058 
1059 static inline u32 lsn_to_page_off(struct ntfs_log *log, u64 lsn)
1060 {
1061     return (((u32)lsn) << 3) & log->page_mask;
1062 }
1063 
1064 static inline u64 vbo_to_lsn(struct ntfs_log *log, u32 off, u64 Seq)
1065 {
1066     return (off >> 3) + (Seq << log->file_data_bits);
1067 }
1068 
1069 static inline bool is_lsn_in_file(struct ntfs_log *log, u64 lsn)
1070 {
1071     return lsn >= log->oldest_lsn &&
1072            lsn <= le64_to_cpu(log->ra->current_lsn);
1073 }
1074 
1075 static inline u32 hdr_file_off(struct ntfs_log *log,
1076                    struct RECORD_PAGE_HDR *hdr)
1077 {
1078     if (log->major_ver < 2)
1079         return le64_to_cpu(hdr->rhdr.lsn);
1080 
1081     return le32_to_cpu(hdr->file_off);
1082 }
1083 
1084 static inline u64 base_lsn(struct ntfs_log *log,
1085                const struct RECORD_PAGE_HDR *hdr, u64 lsn)
1086 {
1087     u64 h_lsn = le64_to_cpu(hdr->rhdr.lsn);
1088     u64 ret = (((h_lsn >> log->file_data_bits) +
1089             (lsn < (lsn_to_vbo(log, h_lsn) & ~log->page_mask) ? 1 : 0))
1090            << log->file_data_bits) +
1091           ((((is_log_record_end(hdr) &&
1092               h_lsn <= le64_to_cpu(hdr->record_hdr.last_end_lsn))
1093                  ? le16_to_cpu(hdr->record_hdr.next_record_off)
1094                  : log->page_size) +
1095             lsn) >>
1096            3);
1097 
1098     return ret;
1099 }
1100 
1101 static inline bool verify_client_lsn(struct ntfs_log *log,
1102                      const struct CLIENT_REC *client, u64 lsn)
1103 {
1104     return lsn >= le64_to_cpu(client->oldest_lsn) &&
1105            lsn <= le64_to_cpu(log->ra->current_lsn) && lsn;
1106 }
1107 
1108 struct restart_info {
1109     u64 last_lsn;
1110     struct RESTART_HDR *r_page;
1111     u32 vbo;
1112     bool chkdsk_was_run;
1113     bool valid_page;
1114     bool initialized;
1115     bool restart;
1116 };
1117 
1118 static int read_log_page(struct ntfs_log *log, u32 vbo,
1119              struct RECORD_PAGE_HDR **buffer, bool *usa_error)
1120 {
1121     int err = 0;
1122     u32 page_idx = vbo >> log->page_bits;
1123     u32 page_off = vbo & log->page_mask;
1124     u32 bytes = log->page_size - page_off;
1125     void *to_free = NULL;
1126     u32 page_vbo = page_idx << log->page_bits;
1127     struct RECORD_PAGE_HDR *page_buf;
1128     struct ntfs_inode *ni = log->ni;
1129     bool bBAAD;
1130 
1131     if (vbo >= log->l_size)
1132         return -EINVAL;
1133 
1134     if (!*buffer) {
1135         to_free = kmalloc(bytes, GFP_NOFS);
1136         if (!to_free)
1137             return -ENOMEM;
1138         *buffer = to_free;
1139     }
1140 
1141     page_buf = page_off ? log->one_page_buf : *buffer;
1142 
1143     err = ntfs_read_run_nb(ni->mi.sbi, &ni->file.run, page_vbo, page_buf,
1144                    log->page_size, NULL);
1145     if (err)
1146         goto out;
1147 
1148     if (page_buf->rhdr.sign != NTFS_FFFF_SIGNATURE)
1149         ntfs_fix_post_read(&page_buf->rhdr, PAGE_SIZE, false);
1150 
1151     if (page_buf != *buffer)
1152         memcpy(*buffer, Add2Ptr(page_buf, page_off), bytes);
1153 
1154     bBAAD = page_buf->rhdr.sign == NTFS_BAAD_SIGNATURE;
1155 
1156     if (usa_error)
1157         *usa_error = bBAAD;
1158     /* Check that the update sequence array for this page is valid */
1159     /* If we don't allow errors, raise an error status */
1160     else if (bBAAD)
1161         err = -EINVAL;
1162 
1163 out:
1164     if (err && to_free) {
1165         kfree(to_free);
1166         *buffer = NULL;
1167     }
1168 
1169     return err;
1170 }
1171 
1172 /*
1173  * log_read_rst
1174  *
1175  * It walks through 512 blocks of the file looking for a valid
1176  * restart page header. It will stop the first time we find a
1177  * valid page header.
1178  */
1179 static int log_read_rst(struct ntfs_log *log, u32 l_size, bool first,
1180             struct restart_info *info)
1181 {
1182     u32 skip, vbo;
1183     struct RESTART_HDR *r_page = kmalloc(DefaultLogPageSize, GFP_NOFS);
1184 
1185     if (!r_page)
1186         return -ENOMEM;
1187 
1188     /* Determine which restart area we are looking for. */
1189     if (first) {
1190         vbo = 0;
1191         skip = 512;
1192     } else {
1193         vbo = 512;
1194         skip = 0;
1195     }
1196 
1197     /* Loop continuously until we succeed. */
1198     for (; vbo < l_size; vbo = 2 * vbo + skip, skip = 0) {
1199         bool usa_error;
1200         u32 sys_page_size;
1201         bool brst, bchk;
1202         struct RESTART_AREA *ra;
1203 
1204         /* Read a page header at the current offset. */
1205         if (read_log_page(log, vbo, (struct RECORD_PAGE_HDR **)&r_page,
1206                   &usa_error)) {
1207             /* Ignore any errors. */
1208             continue;
1209         }
1210 
1211         /* Exit if the signature is a log record page. */
1212         if (r_page->rhdr.sign == NTFS_RCRD_SIGNATURE) {
1213             info->initialized = true;
1214             break;
1215         }
1216 
1217         brst = r_page->rhdr.sign == NTFS_RSTR_SIGNATURE;
1218         bchk = r_page->rhdr.sign == NTFS_CHKD_SIGNATURE;
1219 
1220         if (!bchk && !brst) {
1221             if (r_page->rhdr.sign != NTFS_FFFF_SIGNATURE) {
1222                 /*
1223                  * Remember if the signature does not
1224                  * indicate uninitialized file.
1225                  */
1226                 info->initialized = true;
1227             }
1228             continue;
1229         }
1230 
1231         ra = NULL;
1232         info->valid_page = false;
1233         info->initialized = true;
1234         info->vbo = vbo;
1235 
1236         /* Let's check the restart area if this is a valid page. */
1237         if (!is_rst_page_hdr_valid(vbo, r_page))
1238             goto check_result;
1239         ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
1240 
1241         if (!is_rst_area_valid(r_page))
1242             goto check_result;
1243 
1244         /*
1245          * We have a valid restart page header and restart area.
1246          * If chkdsk was run or we have no clients then we have
1247          * no more checking to do.
1248          */
1249         if (bchk || ra->client_idx[1] == LFS_NO_CLIENT_LE) {
1250             info->valid_page = true;
1251             goto check_result;
1252         }
1253 
1254         /* Read the entire restart area. */
1255         sys_page_size = le32_to_cpu(r_page->sys_page_size);
1256         if (DefaultLogPageSize != sys_page_size) {
1257             kfree(r_page);
1258             r_page = kzalloc(sys_page_size, GFP_NOFS);
1259             if (!r_page)
1260                 return -ENOMEM;
1261 
1262             if (read_log_page(log, vbo,
1263                       (struct RECORD_PAGE_HDR **)&r_page,
1264                       &usa_error)) {
1265                 /* Ignore any errors. */
1266                 kfree(r_page);
1267                 r_page = NULL;
1268                 continue;
1269             }
1270         }
1271 
1272         if (is_client_area_valid(r_page, usa_error)) {
1273             info->valid_page = true;
1274             ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
1275         }
1276 
1277 check_result:
1278         /*
1279          * If chkdsk was run then update the caller's
1280          * values and return.
1281          */
1282         if (r_page->rhdr.sign == NTFS_CHKD_SIGNATURE) {
1283             info->chkdsk_was_run = true;
1284             info->last_lsn = le64_to_cpu(r_page->rhdr.lsn);
1285             info->restart = true;
1286             info->r_page = r_page;
1287             return 0;
1288         }
1289 
1290         /*
1291          * If we have a valid page then copy the values
1292          * we need from it.
1293          */
1294         if (info->valid_page) {
1295             info->last_lsn = le64_to_cpu(ra->current_lsn);
1296             info->restart = true;
1297             info->r_page = r_page;
1298             return 0;
1299         }
1300     }
1301 
1302     kfree(r_page);
1303 
1304     return 0;
1305 }
1306 
1307 /*
1308  * Ilog_init_pg_hdr - Init @log from restart page header.
1309  */
1310 static void log_init_pg_hdr(struct ntfs_log *log, u32 sys_page_size,
1311                 u32 page_size, u16 major_ver, u16 minor_ver)
1312 {
1313     log->sys_page_size = sys_page_size;
1314     log->sys_page_mask = sys_page_size - 1;
1315     log->page_size = page_size;
1316     log->page_mask = page_size - 1;
1317     log->page_bits = blksize_bits(page_size);
1318 
1319     log->clst_per_page = log->page_size >> log->ni->mi.sbi->cluster_bits;
1320     if (!log->clst_per_page)
1321         log->clst_per_page = 1;
1322 
1323     log->first_page = major_ver >= 2
1324                   ? 0x22 * page_size
1325                   : ((sys_page_size << 1) + (page_size << 1));
1326     log->major_ver = major_ver;
1327     log->minor_ver = minor_ver;
1328 }
1329 
1330 /*
1331  * log_create - Init @log in cases when we don't have a restart area to use.
1332  */
1333 static void log_create(struct ntfs_log *log, u32 l_size, const u64 last_lsn,
1334                u32 open_log_count, bool wrapped, bool use_multi_page)
1335 {
1336     log->l_size = l_size;
1337     /* All file offsets must be quadword aligned. */
1338     log->file_data_bits = blksize_bits(l_size) - 3;
1339     log->seq_num_mask = (8 << log->file_data_bits) - 1;
1340     log->seq_num_bits = sizeof(u64) * 8 - log->file_data_bits;
1341     log->seq_num = (last_lsn >> log->file_data_bits) + 2;
1342     log->next_page = log->first_page;
1343     log->oldest_lsn = log->seq_num << log->file_data_bits;
1344     log->oldest_lsn_off = 0;
1345     log->last_lsn = log->oldest_lsn;
1346 
1347     log->l_flags |= NTFSLOG_NO_LAST_LSN | NTFSLOG_NO_OLDEST_LSN;
1348 
1349     /* Set the correct flags for the I/O and indicate if we have wrapped. */
1350     if (wrapped)
1351         log->l_flags |= NTFSLOG_WRAPPED;
1352 
1353     if (use_multi_page)
1354         log->l_flags |= NTFSLOG_MULTIPLE_PAGE_IO;
1355 
1356     /* Compute the log page values. */
1357     log->data_off = ALIGN(
1358         offsetof(struct RECORD_PAGE_HDR, fixups) +
1359             sizeof(short) * ((log->page_size >> SECTOR_SHIFT) + 1),
1360         8);
1361     log->data_size = log->page_size - log->data_off;
1362     log->record_header_len = sizeof(struct LFS_RECORD_HDR);
1363 
1364     /* Remember the different page sizes for reservation. */
1365     log->reserved = log->data_size - log->record_header_len;
1366 
1367     /* Compute the restart page values. */
1368     log->ra_off = ALIGN(
1369         offsetof(struct RESTART_HDR, fixups) +
1370             sizeof(short) *
1371                 ((log->sys_page_size >> SECTOR_SHIFT) + 1),
1372         8);
1373     log->restart_size = log->sys_page_size - log->ra_off;
1374     log->ra_size = struct_size(log->ra, clients, 1);
1375     log->current_openlog_count = open_log_count;
1376 
1377     /*
1378      * The total available log file space is the number of
1379      * log file pages times the space available on each page.
1380      */
1381     log->total_avail_pages = log->l_size - log->first_page;
1382     log->total_avail = log->total_avail_pages >> log->page_bits;
1383 
1384     /*
1385      * We assume that we can't use the end of the page less than
1386      * the file record size.
1387      * Then we won't need to reserve more than the caller asks for.
1388      */
1389     log->max_current_avail = log->total_avail * log->reserved;
1390     log->total_avail = log->total_avail * log->data_size;
1391     log->current_avail = log->max_current_avail;
1392 }
1393 
1394 /*
1395  * log_create_ra - Fill a restart area from the values stored in @log.
1396  */
1397 static struct RESTART_AREA *log_create_ra(struct ntfs_log *log)
1398 {
1399     struct CLIENT_REC *cr;
1400     struct RESTART_AREA *ra = kzalloc(log->restart_size, GFP_NOFS);
1401 
1402     if (!ra)
1403         return NULL;
1404 
1405     ra->current_lsn = cpu_to_le64(log->last_lsn);
1406     ra->log_clients = cpu_to_le16(1);
1407     ra->client_idx[1] = LFS_NO_CLIENT_LE;
1408     if (log->l_flags & NTFSLOG_MULTIPLE_PAGE_IO)
1409         ra->flags = RESTART_SINGLE_PAGE_IO;
1410     ra->seq_num_bits = cpu_to_le32(log->seq_num_bits);
1411     ra->ra_len = cpu_to_le16(log->ra_size);
1412     ra->client_off = cpu_to_le16(offsetof(struct RESTART_AREA, clients));
1413     ra->l_size = cpu_to_le64(log->l_size);
1414     ra->rec_hdr_len = cpu_to_le16(log->record_header_len);
1415     ra->data_off = cpu_to_le16(log->data_off);
1416     ra->open_log_count = cpu_to_le32(log->current_openlog_count + 1);
1417 
1418     cr = ra->clients;
1419 
1420     cr->prev_client = LFS_NO_CLIENT_LE;
1421     cr->next_client = LFS_NO_CLIENT_LE;
1422 
1423     return ra;
1424 }
1425 
1426 static u32 final_log_off(struct ntfs_log *log, u64 lsn, u32 data_len)
1427 {
1428     u32 base_vbo = lsn << 3;
1429     u32 final_log_off = (base_vbo & log->seq_num_mask) & ~log->page_mask;
1430     u32 page_off = base_vbo & log->page_mask;
1431     u32 tail = log->page_size - page_off;
1432 
1433     page_off -= 1;
1434 
1435     /* Add the length of the header. */
1436     data_len += log->record_header_len;
1437 
1438     /*
1439      * If this lsn is contained this log page we are done.
1440      * Otherwise we need to walk through several log pages.
1441      */
1442     if (data_len > tail) {
1443         data_len -= tail;
1444         tail = log->data_size;
1445         page_off = log->data_off - 1;
1446 
1447         for (;;) {
1448             final_log_off = next_page_off(log, final_log_off);
1449 
1450             /*
1451              * We are done if the remaining bytes
1452              * fit on this page.
1453              */
1454             if (data_len <= tail)
1455                 break;
1456             data_len -= tail;
1457         }
1458     }
1459 
1460     /*
1461      * We add the remaining bytes to our starting position on this page
1462      * and then add that value to the file offset of this log page.
1463      */
1464     return final_log_off + data_len + page_off;
1465 }
1466 
1467 static int next_log_lsn(struct ntfs_log *log, const struct LFS_RECORD_HDR *rh,
1468             u64 *lsn)
1469 {
1470     int err;
1471     u64 this_lsn = le64_to_cpu(rh->this_lsn);
1472     u32 vbo = lsn_to_vbo(log, this_lsn);
1473     u32 end =
1474         final_log_off(log, this_lsn, le32_to_cpu(rh->client_data_len));
1475     u32 hdr_off = end & ~log->sys_page_mask;
1476     u64 seq = this_lsn >> log->file_data_bits;
1477     struct RECORD_PAGE_HDR *page = NULL;
1478 
1479     /* Remember if we wrapped. */
1480     if (end <= vbo)
1481         seq += 1;
1482 
1483     /* Log page header for this page. */
1484     err = read_log_page(log, hdr_off, &page, NULL);
1485     if (err)
1486         return err;
1487 
1488     /*
1489      * If the lsn we were given was not the last lsn on this page,
1490      * then the starting offset for the next lsn is on a quad word
1491      * boundary following the last file offset for the current lsn.
1492      * Otherwise the file offset is the start of the data on the next page.
1493      */
1494     if (this_lsn == le64_to_cpu(page->rhdr.lsn)) {
1495         /* If we wrapped, we need to increment the sequence number. */
1496         hdr_off = next_page_off(log, hdr_off);
1497         if (hdr_off == log->first_page)
1498             seq += 1;
1499 
1500         vbo = hdr_off + log->data_off;
1501     } else {
1502         vbo = ALIGN(end, 8);
1503     }
1504 
1505     /* Compute the lsn based on the file offset and the sequence count. */
1506     *lsn = vbo_to_lsn(log, vbo, seq);
1507 
1508     /*
1509      * If this lsn is within the legal range for the file, we return true.
1510      * Otherwise false indicates that there are no more lsn's.
1511      */
1512     if (!is_lsn_in_file(log, *lsn))
1513         *lsn = 0;
1514 
1515     kfree(page);
1516 
1517     return 0;
1518 }
1519 
1520 /*
1521  * current_log_avail - Calculate the number of bytes available for log records.
1522  */
1523 static u32 current_log_avail(struct ntfs_log *log)
1524 {
1525     u32 oldest_off, next_free_off, free_bytes;
1526 
1527     if (log->l_flags & NTFSLOG_NO_LAST_LSN) {
1528         /* The entire file is available. */
1529         return log->max_current_avail;
1530     }
1531 
1532     /*
1533      * If there is a last lsn the restart area then we know that we will
1534      * have to compute the free range.
1535      * If there is no oldest lsn then start at the first page of the file.
1536      */
1537     oldest_off = (log->l_flags & NTFSLOG_NO_OLDEST_LSN)
1538                  ? log->first_page
1539                  : (log->oldest_lsn_off & ~log->sys_page_mask);
1540 
1541     /*
1542      * We will use the next log page offset to compute the next free page.
1543      * If we are going to reuse this page go to the next page.
1544      * If we are at the first page then use the end of the file.
1545      */
1546     next_free_off = (log->l_flags & NTFSLOG_REUSE_TAIL)
1547                 ? log->next_page + log->page_size
1548                 : log->next_page == log->first_page
1549                       ? log->l_size
1550                       : log->next_page;
1551 
1552     /* If the two offsets are the same then there is no available space. */
1553     if (oldest_off == next_free_off)
1554         return 0;
1555     /*
1556      * If the free offset follows the oldest offset then subtract
1557      * this range from the total available pages.
1558      */
1559     free_bytes =
1560         oldest_off < next_free_off
1561             ? log->total_avail_pages - (next_free_off - oldest_off)
1562             : oldest_off - next_free_off;
1563 
1564     free_bytes >>= log->page_bits;
1565     return free_bytes * log->reserved;
1566 }
1567 
1568 static bool check_subseq_log_page(struct ntfs_log *log,
1569                   const struct RECORD_PAGE_HDR *rp, u32 vbo,
1570                   u64 seq)
1571 {
1572     u64 lsn_seq;
1573     const struct NTFS_RECORD_HEADER *rhdr = &rp->rhdr;
1574     u64 lsn = le64_to_cpu(rhdr->lsn);
1575 
1576     if (rhdr->sign == NTFS_FFFF_SIGNATURE || !rhdr->sign)
1577         return false;
1578 
1579     /*
1580      * If the last lsn on the page occurs was written after the page
1581      * that caused the original error then we have a fatal error.
1582      */
1583     lsn_seq = lsn >> log->file_data_bits;
1584 
1585     /*
1586      * If the sequence number for the lsn the page is equal or greater
1587      * than lsn we expect, then this is a subsequent write.
1588      */
1589     return lsn_seq >= seq ||
1590            (lsn_seq == seq - 1 && log->first_page == vbo &&
1591         vbo != (lsn_to_vbo(log, lsn) & ~log->page_mask));
1592 }
1593 
1594 /*
1595  * last_log_lsn
1596  *
1597  * Walks through the log pages for a file, searching for the
1598  * last log page written to the file.
1599  */
1600 static int last_log_lsn(struct ntfs_log *log)
1601 {
1602     int err;
1603     bool usa_error = false;
1604     bool replace_page = false;
1605     bool reuse_page = log->l_flags & NTFSLOG_REUSE_TAIL;
1606     bool wrapped_file, wrapped;
1607 
1608     u32 page_cnt = 1, page_pos = 1;
1609     u32 page_off = 0, page_off1 = 0, saved_off = 0;
1610     u32 final_off, second_off, final_off_prev = 0, second_off_prev = 0;
1611     u32 first_file_off = 0, second_file_off = 0;
1612     u32 part_io_count = 0;
1613     u32 tails = 0;
1614     u32 this_off, curpage_off, nextpage_off, remain_pages;
1615 
1616     u64 expected_seq, seq_base = 0, lsn_base = 0;
1617     u64 best_lsn, best_lsn1, best_lsn2;
1618     u64 lsn_cur, lsn1, lsn2;
1619     u64 last_ok_lsn = reuse_page ? log->last_lsn : 0;
1620 
1621     u16 cur_pos, best_page_pos;
1622 
1623     struct RECORD_PAGE_HDR *page = NULL;
1624     struct RECORD_PAGE_HDR *tst_page = NULL;
1625     struct RECORD_PAGE_HDR *first_tail = NULL;
1626     struct RECORD_PAGE_HDR *second_tail = NULL;
1627     struct RECORD_PAGE_HDR *tail_page = NULL;
1628     struct RECORD_PAGE_HDR *second_tail_prev = NULL;
1629     struct RECORD_PAGE_HDR *first_tail_prev = NULL;
1630     struct RECORD_PAGE_HDR *page_bufs = NULL;
1631     struct RECORD_PAGE_HDR *best_page;
1632 
1633     if (log->major_ver >= 2) {
1634         final_off = 0x02 * log->page_size;
1635         second_off = 0x12 * log->page_size;
1636 
1637         // 0x10 == 0x12 - 0x2
1638         page_bufs = kmalloc(log->page_size * 0x10, GFP_NOFS);
1639         if (!page_bufs)
1640             return -ENOMEM;
1641     } else {
1642         second_off = log->first_page - log->page_size;
1643         final_off = second_off - log->page_size;
1644     }
1645 
1646 next_tail:
1647     /* Read second tail page (at pos 3/0x12000). */
1648     if (read_log_page(log, second_off, &second_tail, &usa_error) ||
1649         usa_error || second_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
1650         kfree(second_tail);
1651         second_tail = NULL;
1652         second_file_off = 0;
1653         lsn2 = 0;
1654     } else {
1655         second_file_off = hdr_file_off(log, second_tail);
1656         lsn2 = le64_to_cpu(second_tail->record_hdr.last_end_lsn);
1657     }
1658 
1659     /* Read first tail page (at pos 2/0x2000). */
1660     if (read_log_page(log, final_off, &first_tail, &usa_error) ||
1661         usa_error || first_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
1662         kfree(first_tail);
1663         first_tail = NULL;
1664         first_file_off = 0;
1665         lsn1 = 0;
1666     } else {
1667         first_file_off = hdr_file_off(log, first_tail);
1668         lsn1 = le64_to_cpu(first_tail->record_hdr.last_end_lsn);
1669     }
1670 
1671     if (log->major_ver < 2) {
1672         int best_page;
1673 
1674         first_tail_prev = first_tail;
1675         final_off_prev = first_file_off;
1676         second_tail_prev = second_tail;
1677         second_off_prev = second_file_off;
1678         tails = 1;
1679 
1680         if (!first_tail && !second_tail)
1681             goto tail_read;
1682 
1683         if (first_tail && second_tail)
1684             best_page = lsn1 < lsn2 ? 1 : 0;
1685         else if (first_tail)
1686             best_page = 0;
1687         else
1688             best_page = 1;
1689 
1690         page_off = best_page ? second_file_off : first_file_off;
1691         seq_base = (best_page ? lsn2 : lsn1) >> log->file_data_bits;
1692         goto tail_read;
1693     }
1694 
1695     best_lsn1 = first_tail ? base_lsn(log, first_tail, first_file_off) : 0;
1696     best_lsn2 =
1697         second_tail ? base_lsn(log, second_tail, second_file_off) : 0;
1698 
1699     if (first_tail && second_tail) {
1700         if (best_lsn1 > best_lsn2) {
1701             best_lsn = best_lsn1;
1702             best_page = first_tail;
1703             this_off = first_file_off;
1704         } else {
1705             best_lsn = best_lsn2;
1706             best_page = second_tail;
1707             this_off = second_file_off;
1708         }
1709     } else if (first_tail) {
1710         best_lsn = best_lsn1;
1711         best_page = first_tail;
1712         this_off = first_file_off;
1713     } else if (second_tail) {
1714         best_lsn = best_lsn2;
1715         best_page = second_tail;
1716         this_off = second_file_off;
1717     } else {
1718         goto tail_read;
1719     }
1720 
1721     best_page_pos = le16_to_cpu(best_page->page_pos);
1722 
1723     if (!tails) {
1724         if (best_page_pos == page_pos) {
1725             seq_base = best_lsn >> log->file_data_bits;
1726             saved_off = page_off = le32_to_cpu(best_page->file_off);
1727             lsn_base = best_lsn;
1728 
1729             memmove(page_bufs, best_page, log->page_size);
1730 
1731             page_cnt = le16_to_cpu(best_page->page_count);
1732             if (page_cnt > 1)
1733                 page_pos += 1;
1734 
1735             tails = 1;
1736         }
1737     } else if (seq_base == (best_lsn >> log->file_data_bits) &&
1738            saved_off + log->page_size == this_off &&
1739            lsn_base < best_lsn &&
1740            (page_pos != page_cnt || best_page_pos == page_pos ||
1741             best_page_pos == 1) &&
1742            (page_pos >= page_cnt || best_page_pos == page_pos)) {
1743         u16 bppc = le16_to_cpu(best_page->page_count);
1744 
1745         saved_off += log->page_size;
1746         lsn_base = best_lsn;
1747 
1748         memmove(Add2Ptr(page_bufs, tails * log->page_size), best_page,
1749             log->page_size);
1750 
1751         tails += 1;
1752 
1753         if (best_page_pos != bppc) {
1754             page_cnt = bppc;
1755             page_pos = best_page_pos;
1756 
1757             if (page_cnt > 1)
1758                 page_pos += 1;
1759         } else {
1760             page_pos = page_cnt = 1;
1761         }
1762     } else {
1763         kfree(first_tail);
1764         kfree(second_tail);
1765         goto tail_read;
1766     }
1767 
1768     kfree(first_tail_prev);
1769     first_tail_prev = first_tail;
1770     final_off_prev = first_file_off;
1771     first_tail = NULL;
1772 
1773     kfree(second_tail_prev);
1774     second_tail_prev = second_tail;
1775     second_off_prev = second_file_off;
1776     second_tail = NULL;
1777 
1778     final_off += log->page_size;
1779     second_off += log->page_size;
1780 
1781     if (tails < 0x10)
1782         goto next_tail;
1783 tail_read:
1784     first_tail = first_tail_prev;
1785     final_off = final_off_prev;
1786 
1787     second_tail = second_tail_prev;
1788     second_off = second_off_prev;
1789 
1790     page_cnt = page_pos = 1;
1791 
1792     curpage_off = seq_base == log->seq_num ? min(log->next_page, page_off)
1793                            : log->next_page;
1794 
1795     wrapped_file =
1796         curpage_off == log->first_page &&
1797         !(log->l_flags & (NTFSLOG_NO_LAST_LSN | NTFSLOG_REUSE_TAIL));
1798 
1799     expected_seq = wrapped_file ? (log->seq_num + 1) : log->seq_num;
1800 
1801     nextpage_off = curpage_off;
1802 
1803 next_page:
1804     tail_page = NULL;
1805     /* Read the next log page. */
1806     err = read_log_page(log, curpage_off, &page, &usa_error);
1807 
1808     /* Compute the next log page offset the file. */
1809     nextpage_off = next_page_off(log, curpage_off);
1810     wrapped = nextpage_off == log->first_page;
1811 
1812     if (tails > 1) {
1813         struct RECORD_PAGE_HDR *cur_page =
1814             Add2Ptr(page_bufs, curpage_off - page_off);
1815 
1816         if (curpage_off == saved_off) {
1817             tail_page = cur_page;
1818             goto use_tail_page;
1819         }
1820 
1821         if (page_off > curpage_off || curpage_off >= saved_off)
1822             goto use_tail_page;
1823 
1824         if (page_off1)
1825             goto use_cur_page;
1826 
1827         if (!err && !usa_error &&
1828             page->rhdr.sign == NTFS_RCRD_SIGNATURE &&
1829             cur_page->rhdr.lsn == page->rhdr.lsn &&
1830             cur_page->record_hdr.next_record_off ==
1831                 page->record_hdr.next_record_off &&
1832             ((page_pos == page_cnt &&
1833               le16_to_cpu(page->page_pos) == 1) ||
1834              (page_pos != page_cnt &&
1835               le16_to_cpu(page->page_pos) == page_pos + 1 &&
1836               le16_to_cpu(page->page_count) == page_cnt))) {
1837             cur_page = NULL;
1838             goto use_tail_page;
1839         }
1840 
1841         page_off1 = page_off;
1842 
1843 use_cur_page:
1844 
1845         lsn_cur = le64_to_cpu(cur_page->rhdr.lsn);
1846 
1847         if (last_ok_lsn !=
1848                 le64_to_cpu(cur_page->record_hdr.last_end_lsn) &&
1849             ((lsn_cur >> log->file_data_bits) +
1850              ((curpage_off <
1851                (lsn_to_vbo(log, lsn_cur) & ~log->page_mask))
1852                   ? 1
1853                   : 0)) != expected_seq) {
1854             goto check_tail;
1855         }
1856 
1857         if (!is_log_record_end(cur_page)) {
1858             tail_page = NULL;
1859             last_ok_lsn = lsn_cur;
1860             goto next_page_1;
1861         }
1862 
1863         log->seq_num = expected_seq;
1864         log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
1865         log->last_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
1866         log->ra->current_lsn = cur_page->record_hdr.last_end_lsn;
1867 
1868         if (log->record_header_len <=
1869             log->page_size -
1870                 le16_to_cpu(cur_page->record_hdr.next_record_off)) {
1871             log->l_flags |= NTFSLOG_REUSE_TAIL;
1872             log->next_page = curpage_off;
1873         } else {
1874             log->l_flags &= ~NTFSLOG_REUSE_TAIL;
1875             log->next_page = nextpage_off;
1876         }
1877 
1878         if (wrapped_file)
1879             log->l_flags |= NTFSLOG_WRAPPED;
1880 
1881         last_ok_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
1882         goto next_page_1;
1883     }
1884 
1885     /*
1886      * If we are at the expected first page of a transfer check to see
1887      * if either tail copy is at this offset.
1888      * If this page is the last page of a transfer, check if we wrote
1889      * a subsequent tail copy.
1890      */
1891     if (page_cnt == page_pos || page_cnt == page_pos + 1) {
1892         /*
1893          * Check if the offset matches either the first or second
1894          * tail copy. It is possible it will match both.
1895          */
1896         if (curpage_off == final_off)
1897             tail_page = first_tail;
1898 
1899         /*
1900          * If we already matched on the first page then
1901          * check the ending lsn's.
1902          */
1903         if (curpage_off == second_off) {
1904             if (!tail_page ||
1905                 (second_tail &&
1906                  le64_to_cpu(second_tail->record_hdr.last_end_lsn) >
1907                      le64_to_cpu(first_tail->record_hdr
1908                              .last_end_lsn))) {
1909                 tail_page = second_tail;
1910             }
1911         }
1912     }
1913 
1914 use_tail_page:
1915     if (tail_page) {
1916         /* We have a candidate for a tail copy. */
1917         lsn_cur = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
1918 
1919         if (last_ok_lsn < lsn_cur) {
1920             /*
1921              * If the sequence number is not expected,
1922              * then don't use the tail copy.
1923              */
1924             if (expected_seq != (lsn_cur >> log->file_data_bits))
1925                 tail_page = NULL;
1926         } else if (last_ok_lsn > lsn_cur) {
1927             /*
1928              * If the last lsn is greater than the one on
1929              * this page then forget this tail.
1930              */
1931             tail_page = NULL;
1932         }
1933     }
1934 
1935     /*
1936      *If we have an error on the current page,
1937      * we will break of this loop.
1938      */
1939     if (err || usa_error)
1940         goto check_tail;
1941 
1942     /*
1943      * Done if the last lsn on this page doesn't match the previous known
1944      * last lsn or the sequence number is not expected.
1945      */
1946     lsn_cur = le64_to_cpu(page->rhdr.lsn);
1947     if (last_ok_lsn != lsn_cur &&
1948         expected_seq != (lsn_cur >> log->file_data_bits)) {
1949         goto check_tail;
1950     }
1951 
1952     /*
1953      * Check that the page position and page count values are correct.
1954      * If this is the first page of a transfer the position must be 1
1955      * and the count will be unknown.
1956      */
1957     if (page_cnt == page_pos) {
1958         if (page->page_pos != cpu_to_le16(1) &&
1959             (!reuse_page || page->page_pos != page->page_count)) {
1960             /*
1961              * If the current page is the first page we are
1962              * looking at and we are reusing this page then
1963              * it can be either the first or last page of a
1964              * transfer. Otherwise it can only be the first.
1965              */
1966             goto check_tail;
1967         }
1968     } else if (le16_to_cpu(page->page_count) != page_cnt ||
1969            le16_to_cpu(page->page_pos) != page_pos + 1) {
1970         /*
1971          * The page position better be 1 more than the last page
1972          * position and the page count better match.
1973          */
1974         goto check_tail;
1975     }
1976 
1977     /*
1978      * We have a valid page the file and may have a valid page
1979      * the tail copy area.
1980      * If the tail page was written after the page the file then
1981      * break of the loop.
1982      */
1983     if (tail_page &&
1984         le64_to_cpu(tail_page->record_hdr.last_end_lsn) > lsn_cur) {
1985         /* Remember if we will replace the page. */
1986         replace_page = true;
1987         goto check_tail;
1988     }
1989 
1990     tail_page = NULL;
1991 
1992     if (is_log_record_end(page)) {
1993         /*
1994          * Since we have read this page we know the sequence number
1995          * is the same as our expected value.
1996          */
1997         log->seq_num = expected_seq;
1998         log->last_lsn = le64_to_cpu(page->record_hdr.last_end_lsn);
1999         log->ra->current_lsn = page->record_hdr.last_end_lsn;
2000         log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
2001 
2002         /*
2003          * If there is room on this page for another header then
2004          * remember we want to reuse the page.
2005          */
2006         if (log->record_header_len <=
2007             log->page_size -
2008                 le16_to_cpu(page->record_hdr.next_record_off)) {
2009             log->l_flags |= NTFSLOG_REUSE_TAIL;
2010             log->next_page = curpage_off;
2011         } else {
2012             log->l_flags &= ~NTFSLOG_REUSE_TAIL;
2013             log->next_page = nextpage_off;
2014         }
2015 
2016         /* Remember if we wrapped the log file. */
2017         if (wrapped_file)
2018             log->l_flags |= NTFSLOG_WRAPPED;
2019     }
2020 
2021     /*
2022      * Remember the last page count and position.
2023      * Also remember the last known lsn.
2024      */
2025     page_cnt = le16_to_cpu(page->page_count);
2026     page_pos = le16_to_cpu(page->page_pos);
2027     last_ok_lsn = le64_to_cpu(page->rhdr.lsn);
2028 
2029 next_page_1:
2030 
2031     if (wrapped) {
2032         expected_seq += 1;
2033         wrapped_file = 1;
2034     }
2035 
2036     curpage_off = nextpage_off;
2037     kfree(page);
2038     page = NULL;
2039     reuse_page = 0;
2040     goto next_page;
2041 
2042 check_tail:
2043     if (tail_page) {
2044         log->seq_num = expected_seq;
2045         log->last_lsn = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
2046         log->ra->current_lsn = tail_page->record_hdr.last_end_lsn;
2047         log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
2048 
2049         if (log->page_size -
2050                 le16_to_cpu(
2051                     tail_page->record_hdr.next_record_off) >=
2052             log->record_header_len) {
2053             log->l_flags |= NTFSLOG_REUSE_TAIL;
2054             log->next_page = curpage_off;
2055         } else {
2056             log->l_flags &= ~NTFSLOG_REUSE_TAIL;
2057             log->next_page = nextpage_off;
2058         }
2059 
2060         if (wrapped)
2061             log->l_flags |= NTFSLOG_WRAPPED;
2062     }
2063 
2064     /* Remember that the partial IO will start at the next page. */
2065     second_off = nextpage_off;
2066 
2067     /*
2068      * If the next page is the first page of the file then update
2069      * the sequence number for log records which begon the next page.
2070      */
2071     if (wrapped)
2072         expected_seq += 1;
2073 
2074     /*
2075      * If we have a tail copy or are performing single page I/O we can
2076      * immediately look at the next page.
2077      */
2078     if (replace_page || (log->ra->flags & RESTART_SINGLE_PAGE_IO)) {
2079         page_cnt = 2;
2080         page_pos = 1;
2081         goto check_valid;
2082     }
2083 
2084     if (page_pos != page_cnt)
2085         goto check_valid;
2086     /*
2087      * If the next page causes us to wrap to the beginning of the log
2088      * file then we know which page to check next.
2089      */
2090     if (wrapped) {
2091         page_cnt = 2;
2092         page_pos = 1;
2093         goto check_valid;
2094     }
2095 
2096     cur_pos = 2;
2097 
2098 next_test_page:
2099     kfree(tst_page);
2100     tst_page = NULL;
2101 
2102     /* Walk through the file, reading log pages. */
2103     err = read_log_page(log, nextpage_off, &tst_page, &usa_error);
2104 
2105     /*
2106      * If we get a USA error then assume that we correctly found
2107      * the end of the original transfer.
2108      */
2109     if (usa_error)
2110         goto file_is_valid;
2111 
2112     /*
2113      * If we were able to read the page, we examine it to see if it
2114      * is the same or different Io block.
2115      */
2116     if (err)
2117         goto next_test_page_1;
2118 
2119     if (le16_to_cpu(tst_page->page_pos) == cur_pos &&
2120         check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
2121         page_cnt = le16_to_cpu(tst_page->page_count) + 1;
2122         page_pos = le16_to_cpu(tst_page->page_pos);
2123         goto check_valid;
2124     } else {
2125         goto file_is_valid;
2126     }
2127 
2128 next_test_page_1:
2129 
2130     nextpage_off = next_page_off(log, curpage_off);
2131     wrapped = nextpage_off == log->first_page;
2132 
2133     if (wrapped) {
2134         expected_seq += 1;
2135         page_cnt = 2;
2136         page_pos = 1;
2137     }
2138 
2139     cur_pos += 1;
2140     part_io_count += 1;
2141     if (!wrapped)
2142         goto next_test_page;
2143 
2144 check_valid:
2145     /* Skip over the remaining pages this transfer. */
2146     remain_pages = page_cnt - page_pos - 1;
2147     part_io_count += remain_pages;
2148 
2149     while (remain_pages--) {
2150         nextpage_off = next_page_off(log, curpage_off);
2151         wrapped = nextpage_off == log->first_page;
2152 
2153         if (wrapped)
2154             expected_seq += 1;
2155     }
2156 
2157     /* Call our routine to check this log page. */
2158     kfree(tst_page);
2159     tst_page = NULL;
2160 
2161     err = read_log_page(log, nextpage_off, &tst_page, &usa_error);
2162     if (!err && !usa_error &&
2163         check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
2164         err = -EINVAL;
2165         goto out;
2166     }
2167 
2168 file_is_valid:
2169 
2170     /* We have a valid file. */
2171     if (page_off1 || tail_page) {
2172         struct RECORD_PAGE_HDR *tmp_page;
2173 
2174         if (sb_rdonly(log->ni->mi.sbi->sb)) {
2175             err = -EROFS;
2176             goto out;
2177         }
2178 
2179         if (page_off1) {
2180             tmp_page = Add2Ptr(page_bufs, page_off1 - page_off);
2181             tails -= (page_off1 - page_off) / log->page_size;
2182             if (!tail_page)
2183                 tails -= 1;
2184         } else {
2185             tmp_page = tail_page;
2186             tails = 1;
2187         }
2188 
2189         while (tails--) {
2190             u64 off = hdr_file_off(log, tmp_page);
2191 
2192             if (!page) {
2193                 page = kmalloc(log->page_size, GFP_NOFS);
2194                 if (!page)
2195                     return -ENOMEM;
2196             }
2197 
2198             /*
2199              * Correct page and copy the data from this page
2200              * into it and flush it to disk.
2201              */
2202             memcpy(page, tmp_page, log->page_size);
2203 
2204             /* Fill last flushed lsn value flush the page. */
2205             if (log->major_ver < 2)
2206                 page->rhdr.lsn = page->record_hdr.last_end_lsn;
2207             else
2208                 page->file_off = 0;
2209 
2210             page->page_pos = page->page_count = cpu_to_le16(1);
2211 
2212             ntfs_fix_pre_write(&page->rhdr, log->page_size);
2213 
2214             err = ntfs_sb_write_run(log->ni->mi.sbi,
2215                         &log->ni->file.run, off, page,
2216                         log->page_size, 0);
2217 
2218             if (err)
2219                 goto out;
2220 
2221             if (part_io_count && second_off == off) {
2222                 second_off += log->page_size;
2223                 part_io_count -= 1;
2224             }
2225 
2226             tmp_page = Add2Ptr(tmp_page, log->page_size);
2227         }
2228     }
2229 
2230     if (part_io_count) {
2231         if (sb_rdonly(log->ni->mi.sbi->sb)) {
2232             err = -EROFS;
2233             goto out;
2234         }
2235     }
2236 
2237 out:
2238     kfree(second_tail);
2239     kfree(first_tail);
2240     kfree(page);
2241     kfree(tst_page);
2242     kfree(page_bufs);
2243 
2244     return err;
2245 }
2246 
2247 /*
2248  * read_log_rec_buf - Copy a log record from the file to a buffer.
2249  *
2250  * The log record may span several log pages and may even wrap the file.
2251  */
2252 static int read_log_rec_buf(struct ntfs_log *log,
2253                 const struct LFS_RECORD_HDR *rh, void *buffer)
2254 {
2255     int err;
2256     struct RECORD_PAGE_HDR *ph = NULL;
2257     u64 lsn = le64_to_cpu(rh->this_lsn);
2258     u32 vbo = lsn_to_vbo(log, lsn) & ~log->page_mask;
2259     u32 off = lsn_to_page_off(log, lsn) + log->record_header_len;
2260     u32 data_len = le32_to_cpu(rh->client_data_len);
2261 
2262     /*
2263      * While there are more bytes to transfer,
2264      * we continue to attempt to perform the read.
2265      */
2266     for (;;) {
2267         bool usa_error;
2268         u32 tail = log->page_size - off;
2269 
2270         if (tail >= data_len)
2271             tail = data_len;
2272 
2273         data_len -= tail;
2274 
2275         err = read_log_page(log, vbo, &ph, &usa_error);
2276         if (err)
2277             goto out;
2278 
2279         /*
2280          * The last lsn on this page better be greater or equal
2281          * to the lsn we are copying.
2282          */
2283         if (lsn > le64_to_cpu(ph->rhdr.lsn)) {
2284             err = -EINVAL;
2285             goto out;
2286         }
2287 
2288         memcpy(buffer, Add2Ptr(ph, off), tail);
2289 
2290         /* If there are no more bytes to transfer, we exit the loop. */
2291         if (!data_len) {
2292             if (!is_log_record_end(ph) ||
2293                 lsn > le64_to_cpu(ph->record_hdr.last_end_lsn)) {
2294                 err = -EINVAL;
2295                 goto out;
2296             }
2297             break;
2298         }
2299 
2300         if (ph->rhdr.lsn == ph->record_hdr.last_end_lsn ||
2301             lsn > le64_to_cpu(ph->rhdr.lsn)) {
2302             err = -EINVAL;
2303             goto out;
2304         }
2305 
2306         vbo = next_page_off(log, vbo);
2307         off = log->data_off;
2308 
2309         /*
2310          * Adjust our pointer the user's buffer to transfer
2311          * the next block to.
2312          */
2313         buffer = Add2Ptr(buffer, tail);
2314     }
2315 
2316 out:
2317     kfree(ph);
2318     return err;
2319 }
2320 
2321 static int read_rst_area(struct ntfs_log *log, struct NTFS_RESTART **rst_,
2322              u64 *lsn)
2323 {
2324     int err;
2325     struct LFS_RECORD_HDR *rh = NULL;
2326     const struct CLIENT_REC *cr =
2327         Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
2328     u64 lsnr, lsnc = le64_to_cpu(cr->restart_lsn);
2329     u32 len;
2330     struct NTFS_RESTART *rst;
2331 
2332     *lsn = 0;
2333     *rst_ = NULL;
2334 
2335     /* If the client doesn't have a restart area, go ahead and exit now. */
2336     if (!lsnc)
2337         return 0;
2338 
2339     err = read_log_page(log, lsn_to_vbo(log, lsnc),
2340                 (struct RECORD_PAGE_HDR **)&rh, NULL);
2341     if (err)
2342         return err;
2343 
2344     rst = NULL;
2345     lsnr = le64_to_cpu(rh->this_lsn);
2346 
2347     if (lsnc != lsnr) {
2348         /* If the lsn values don't match, then the disk is corrupt. */
2349         err = -EINVAL;
2350         goto out;
2351     }
2352 
2353     *lsn = lsnr;
2354     len = le32_to_cpu(rh->client_data_len);
2355 
2356     if (!len) {
2357         err = 0;
2358         goto out;
2359     }
2360 
2361     if (len < sizeof(struct NTFS_RESTART)) {
2362         err = -EINVAL;
2363         goto out;
2364     }
2365 
2366     rst = kmalloc(len, GFP_NOFS);
2367     if (!rst) {
2368         err = -ENOMEM;
2369         goto out;
2370     }
2371 
2372     /* Copy the data into the 'rst' buffer. */
2373     err = read_log_rec_buf(log, rh, rst);
2374     if (err)
2375         goto out;
2376 
2377     *rst_ = rst;
2378     rst = NULL;
2379 
2380 out:
2381     kfree(rh);
2382     kfree(rst);
2383 
2384     return err;
2385 }
2386 
2387 static int find_log_rec(struct ntfs_log *log, u64 lsn, struct lcb *lcb)
2388 {
2389     int err;
2390     struct LFS_RECORD_HDR *rh = lcb->lrh;
2391     u32 rec_len, len;
2392 
2393     /* Read the record header for this lsn. */
2394     if (!rh) {
2395         err = read_log_page(log, lsn_to_vbo(log, lsn),
2396                     (struct RECORD_PAGE_HDR **)&rh, NULL);
2397 
2398         lcb->lrh = rh;
2399         if (err)
2400             return err;
2401     }
2402 
2403     /*
2404      * If the lsn the log record doesn't match the desired
2405      * lsn then the disk is corrupt.
2406      */
2407     if (lsn != le64_to_cpu(rh->this_lsn))
2408         return -EINVAL;
2409 
2410     len = le32_to_cpu(rh->client_data_len);
2411 
2412     /*
2413      * Check that the length field isn't greater than the total
2414      * available space the log file.
2415      */
2416     rec_len = len + log->record_header_len;
2417     if (rec_len >= log->total_avail)
2418         return -EINVAL;
2419 
2420     /*
2421      * If the entire log record is on this log page,
2422      * put a pointer to the log record the context block.
2423      */
2424     if (rh->flags & LOG_RECORD_MULTI_PAGE) {
2425         void *lr = kmalloc(len, GFP_NOFS);
2426 
2427         if (!lr)
2428             return -ENOMEM;
2429 
2430         lcb->log_rec = lr;
2431         lcb->alloc = true;
2432 
2433         /* Copy the data into the buffer returned. */
2434         err = read_log_rec_buf(log, rh, lr);
2435         if (err)
2436             return err;
2437     } else {
2438         /* If beyond the end of the current page -> an error. */
2439         u32 page_off = lsn_to_page_off(log, lsn);
2440 
2441         if (page_off + len + log->record_header_len > log->page_size)
2442             return -EINVAL;
2443 
2444         lcb->log_rec = Add2Ptr(rh, sizeof(struct LFS_RECORD_HDR));
2445         lcb->alloc = false;
2446     }
2447 
2448     return 0;
2449 }
2450 
2451 /*
2452  * read_log_rec_lcb - Init the query operation.
2453  */
2454 static int read_log_rec_lcb(struct ntfs_log *log, u64 lsn, u32 ctx_mode,
2455                 struct lcb **lcb_)
2456 {
2457     int err;
2458     const struct CLIENT_REC *cr;
2459     struct lcb *lcb;
2460 
2461     switch (ctx_mode) {
2462     case lcb_ctx_undo_next:
2463     case lcb_ctx_prev:
2464     case lcb_ctx_next:
2465         break;
2466     default:
2467         return -EINVAL;
2468     }
2469 
2470     /* Check that the given lsn is the legal range for this client. */
2471     cr = Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
2472 
2473     if (!verify_client_lsn(log, cr, lsn))
2474         return -EINVAL;
2475 
2476     lcb = kzalloc(sizeof(struct lcb), GFP_NOFS);
2477     if (!lcb)
2478         return -ENOMEM;
2479     lcb->client = log->client_id;
2480     lcb->ctx_mode = ctx_mode;
2481 
2482     /* Find the log record indicated by the given lsn. */
2483     err = find_log_rec(log, lsn, lcb);
2484     if (err)
2485         goto out;
2486 
2487     *lcb_ = lcb;
2488     return 0;
2489 
2490 out:
2491     lcb_put(lcb);
2492     *lcb_ = NULL;
2493     return err;
2494 }
2495 
2496 /*
2497  * find_client_next_lsn
2498  *
2499  * Attempt to find the next lsn to return to a client based on the context mode.
2500  */
2501 static int find_client_next_lsn(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
2502 {
2503     int err;
2504     u64 next_lsn;
2505     struct LFS_RECORD_HDR *hdr;
2506 
2507     hdr = lcb->lrh;
2508     *lsn = 0;
2509 
2510     if (lcb_ctx_next != lcb->ctx_mode)
2511         goto check_undo_next;
2512 
2513     /* Loop as long as another lsn can be found. */
2514     for (;;) {
2515         u64 current_lsn;
2516 
2517         err = next_log_lsn(log, hdr, &current_lsn);
2518         if (err)
2519             goto out;
2520 
2521         if (!current_lsn)
2522             break;
2523 
2524         if (hdr != lcb->lrh)
2525             kfree(hdr);
2526 
2527         hdr = NULL;
2528         err = read_log_page(log, lsn_to_vbo(log, current_lsn),
2529                     (struct RECORD_PAGE_HDR **)&hdr, NULL);
2530         if (err)
2531             goto out;
2532 
2533         if (memcmp(&hdr->client, &lcb->client,
2534                sizeof(struct CLIENT_ID))) {
2535             /*err = -EINVAL; */
2536         } else if (LfsClientRecord == hdr->record_type) {
2537             kfree(lcb->lrh);
2538             lcb->lrh = hdr;
2539             *lsn = current_lsn;
2540             return 0;
2541         }
2542     }
2543 
2544 out:
2545     if (hdr != lcb->lrh)
2546         kfree(hdr);
2547     return err;
2548 
2549 check_undo_next:
2550     if (lcb_ctx_undo_next == lcb->ctx_mode)
2551         next_lsn = le64_to_cpu(hdr->client_undo_next_lsn);
2552     else if (lcb_ctx_prev == lcb->ctx_mode)
2553         next_lsn = le64_to_cpu(hdr->client_prev_lsn);
2554     else
2555         return 0;
2556 
2557     if (!next_lsn)
2558         return 0;
2559 
2560     if (!verify_client_lsn(
2561             log, Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off)),
2562             next_lsn))
2563         return 0;
2564 
2565     hdr = NULL;
2566     err = read_log_page(log, lsn_to_vbo(log, next_lsn),
2567                 (struct RECORD_PAGE_HDR **)&hdr, NULL);
2568     if (err)
2569         return err;
2570     kfree(lcb->lrh);
2571     lcb->lrh = hdr;
2572 
2573     *lsn = next_lsn;
2574 
2575     return 0;
2576 }
2577 
2578 static int read_next_log_rec(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
2579 {
2580     int err;
2581 
2582     err = find_client_next_lsn(log, lcb, lsn);
2583     if (err)
2584         return err;
2585 
2586     if (!*lsn)
2587         return 0;
2588 
2589     if (lcb->alloc)
2590         kfree(lcb->log_rec);
2591 
2592     lcb->log_rec = NULL;
2593     lcb->alloc = false;
2594     kfree(lcb->lrh);
2595     lcb->lrh = NULL;
2596 
2597     return find_log_rec(log, *lsn, lcb);
2598 }
2599 
2600 static inline bool check_index_header(const struct INDEX_HDR *hdr, size_t bytes)
2601 {
2602     __le16 mask;
2603     u32 min_de, de_off, used, total;
2604     const struct NTFS_DE *e;
2605 
2606     if (hdr_has_subnode(hdr)) {
2607         min_de = sizeof(struct NTFS_DE) + sizeof(u64);
2608         mask = NTFS_IE_HAS_SUBNODES;
2609     } else {
2610         min_de = sizeof(struct NTFS_DE);
2611         mask = 0;
2612     }
2613 
2614     de_off = le32_to_cpu(hdr->de_off);
2615     used = le32_to_cpu(hdr->used);
2616     total = le32_to_cpu(hdr->total);
2617 
2618     if (de_off > bytes - min_de || used > bytes || total > bytes ||
2619         de_off + min_de > used || used > total) {
2620         return false;
2621     }
2622 
2623     e = Add2Ptr(hdr, de_off);
2624     for (;;) {
2625         u16 esize = le16_to_cpu(e->size);
2626         struct NTFS_DE *next = Add2Ptr(e, esize);
2627 
2628         if (esize < min_de || PtrOffset(hdr, next) > used ||
2629             (e->flags & NTFS_IE_HAS_SUBNODES) != mask) {
2630             return false;
2631         }
2632 
2633         if (de_is_last(e))
2634             break;
2635 
2636         e = next;
2637     }
2638 
2639     return true;
2640 }
2641 
2642 static inline bool check_index_buffer(const struct INDEX_BUFFER *ib, u32 bytes)
2643 {
2644     u16 fo;
2645     const struct NTFS_RECORD_HEADER *r = &ib->rhdr;
2646 
2647     if (r->sign != NTFS_INDX_SIGNATURE)
2648         return false;
2649 
2650     fo = (SECTOR_SIZE - ((bytes >> SECTOR_SHIFT) + 1) * sizeof(short));
2651 
2652     if (le16_to_cpu(r->fix_off) > fo)
2653         return false;
2654 
2655     if ((le16_to_cpu(r->fix_num) - 1) * SECTOR_SIZE != bytes)
2656         return false;
2657 
2658     return check_index_header(&ib->ihdr,
2659                   bytes - offsetof(struct INDEX_BUFFER, ihdr));
2660 }
2661 
2662 static inline bool check_index_root(const struct ATTRIB *attr,
2663                     struct ntfs_sb_info *sbi)
2664 {
2665     bool ret;
2666     const struct INDEX_ROOT *root = resident_data(attr);
2667     u8 index_bits = le32_to_cpu(root->index_block_size) >= sbi->cluster_size
2668                 ? sbi->cluster_bits
2669                 : SECTOR_SHIFT;
2670     u8 block_clst = root->index_block_clst;
2671 
2672     if (le32_to_cpu(attr->res.data_size) < sizeof(struct INDEX_ROOT) ||
2673         (root->type != ATTR_NAME && root->type != ATTR_ZERO) ||
2674         (root->type == ATTR_NAME &&
2675          root->rule != NTFS_COLLATION_TYPE_FILENAME) ||
2676         (le32_to_cpu(root->index_block_size) !=
2677          (block_clst << index_bits)) ||
2678         (block_clst != 1 && block_clst != 2 && block_clst != 4 &&
2679          block_clst != 8 && block_clst != 0x10 && block_clst != 0x20 &&
2680          block_clst != 0x40 && block_clst != 0x80)) {
2681         return false;
2682     }
2683 
2684     ret = check_index_header(&root->ihdr,
2685                  le32_to_cpu(attr->res.data_size) -
2686                      offsetof(struct INDEX_ROOT, ihdr));
2687     return ret;
2688 }
2689 
2690 static inline bool check_attr(const struct MFT_REC *rec,
2691                   const struct ATTRIB *attr,
2692                   struct ntfs_sb_info *sbi)
2693 {
2694     u32 asize = le32_to_cpu(attr->size);
2695     u32 rsize = 0;
2696     u64 dsize, svcn, evcn;
2697     u16 run_off;
2698 
2699     /* Check the fixed part of the attribute record header. */
2700     if (asize >= sbi->record_size ||
2701         asize + PtrOffset(rec, attr) >= sbi->record_size ||
2702         (attr->name_len &&
2703          le16_to_cpu(attr->name_off) + attr->name_len * sizeof(short) >
2704              asize)) {
2705         return false;
2706     }
2707 
2708     /* Check the attribute fields. */
2709     switch (attr->non_res) {
2710     case 0:
2711         rsize = le32_to_cpu(attr->res.data_size);
2712         if (rsize >= asize ||
2713             le16_to_cpu(attr->res.data_off) + rsize > asize) {
2714             return false;
2715         }
2716         break;
2717 
2718     case 1:
2719         dsize = le64_to_cpu(attr->nres.data_size);
2720         svcn = le64_to_cpu(attr->nres.svcn);
2721         evcn = le64_to_cpu(attr->nres.evcn);
2722         run_off = le16_to_cpu(attr->nres.run_off);
2723 
2724         if (svcn > evcn + 1 || run_off >= asize ||
2725             le64_to_cpu(attr->nres.valid_size) > dsize ||
2726             dsize > le64_to_cpu(attr->nres.alloc_size)) {
2727             return false;
2728         }
2729 
2730         if (run_unpack(NULL, sbi, 0, svcn, evcn, svcn,
2731                    Add2Ptr(attr, run_off), asize - run_off) < 0) {
2732             return false;
2733         }
2734 
2735         return true;
2736 
2737     default:
2738         return false;
2739     }
2740 
2741     switch (attr->type) {
2742     case ATTR_NAME:
2743         if (fname_full_size(Add2Ptr(
2744                 attr, le16_to_cpu(attr->res.data_off))) > asize) {
2745             return false;
2746         }
2747         break;
2748 
2749     case ATTR_ROOT:
2750         return check_index_root(attr, sbi);
2751 
2752     case ATTR_STD:
2753         if (rsize < sizeof(struct ATTR_STD_INFO5) &&
2754             rsize != sizeof(struct ATTR_STD_INFO)) {
2755             return false;
2756         }
2757         break;
2758 
2759     case ATTR_LIST:
2760     case ATTR_ID:
2761     case ATTR_SECURE:
2762     case ATTR_LABEL:
2763     case ATTR_VOL_INFO:
2764     case ATTR_DATA:
2765     case ATTR_ALLOC:
2766     case ATTR_BITMAP:
2767     case ATTR_REPARSE:
2768     case ATTR_EA_INFO:
2769     case ATTR_EA:
2770     case ATTR_PROPERTYSET:
2771     case ATTR_LOGGED_UTILITY_STREAM:
2772         break;
2773 
2774     default:
2775         return false;
2776     }
2777 
2778     return true;
2779 }
2780 
2781 static inline bool check_file_record(const struct MFT_REC *rec,
2782                      const struct MFT_REC *rec2,
2783                      struct ntfs_sb_info *sbi)
2784 {
2785     const struct ATTRIB *attr;
2786     u16 fo = le16_to_cpu(rec->rhdr.fix_off);
2787     u16 fn = le16_to_cpu(rec->rhdr.fix_num);
2788     u16 ao = le16_to_cpu(rec->attr_off);
2789     u32 rs = sbi->record_size;
2790 
2791     /* Check the file record header for consistency. */
2792     if (rec->rhdr.sign != NTFS_FILE_SIGNATURE ||
2793         fo > (SECTOR_SIZE - ((rs >> SECTOR_SHIFT) + 1) * sizeof(short)) ||
2794         (fn - 1) * SECTOR_SIZE != rs || ao < MFTRECORD_FIXUP_OFFSET_1 ||
2795         ao > sbi->record_size - SIZEOF_RESIDENT || !is_rec_inuse(rec) ||
2796         le32_to_cpu(rec->total) != rs) {
2797         return false;
2798     }
2799 
2800     /* Loop to check all of the attributes. */
2801     for (attr = Add2Ptr(rec, ao); attr->type != ATTR_END;
2802          attr = Add2Ptr(attr, le32_to_cpu(attr->size))) {
2803         if (check_attr(rec, attr, sbi))
2804             continue;
2805         return false;
2806     }
2807 
2808     return true;
2809 }
2810 
2811 static inline int check_lsn(const struct NTFS_RECORD_HEADER *hdr,
2812                 const u64 *rlsn)
2813 {
2814     u64 lsn;
2815 
2816     if (!rlsn)
2817         return true;
2818 
2819     lsn = le64_to_cpu(hdr->lsn);
2820 
2821     if (hdr->sign == NTFS_HOLE_SIGNATURE)
2822         return false;
2823 
2824     if (*rlsn > lsn)
2825         return true;
2826 
2827     return false;
2828 }
2829 
2830 static inline bool check_if_attr(const struct MFT_REC *rec,
2831                  const struct LOG_REC_HDR *lrh)
2832 {
2833     u16 ro = le16_to_cpu(lrh->record_off);
2834     u16 o = le16_to_cpu(rec->attr_off);
2835     const struct ATTRIB *attr = Add2Ptr(rec, o);
2836 
2837     while (o < ro) {
2838         u32 asize;
2839 
2840         if (attr->type == ATTR_END)
2841             break;
2842 
2843         asize = le32_to_cpu(attr->size);
2844         if (!asize)
2845             break;
2846 
2847         o += asize;
2848         attr = Add2Ptr(attr, asize);
2849     }
2850 
2851     return o == ro;
2852 }
2853 
2854 static inline bool check_if_index_root(const struct MFT_REC *rec,
2855                        const struct LOG_REC_HDR *lrh)
2856 {
2857     u16 ro = le16_to_cpu(lrh->record_off);
2858     u16 o = le16_to_cpu(rec->attr_off);
2859     const struct ATTRIB *attr = Add2Ptr(rec, o);
2860 
2861     while (o < ro) {
2862         u32 asize;
2863 
2864         if (attr->type == ATTR_END)
2865             break;
2866 
2867         asize = le32_to_cpu(attr->size);
2868         if (!asize)
2869             break;
2870 
2871         o += asize;
2872         attr = Add2Ptr(attr, asize);
2873     }
2874 
2875     return o == ro && attr->type == ATTR_ROOT;
2876 }
2877 
2878 static inline bool check_if_root_index(const struct ATTRIB *attr,
2879                        const struct INDEX_HDR *hdr,
2880                        const struct LOG_REC_HDR *lrh)
2881 {
2882     u16 ao = le16_to_cpu(lrh->attr_off);
2883     u32 de_off = le32_to_cpu(hdr->de_off);
2884     u32 o = PtrOffset(attr, hdr) + de_off;
2885     const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
2886     u32 asize = le32_to_cpu(attr->size);
2887 
2888     while (o < ao) {
2889         u16 esize;
2890 
2891         if (o >= asize)
2892             break;
2893 
2894         esize = le16_to_cpu(e->size);
2895         if (!esize)
2896             break;
2897 
2898         o += esize;
2899         e = Add2Ptr(e, esize);
2900     }
2901 
2902     return o == ao;
2903 }
2904 
2905 static inline bool check_if_alloc_index(const struct INDEX_HDR *hdr,
2906                     u32 attr_off)
2907 {
2908     u32 de_off = le32_to_cpu(hdr->de_off);
2909     u32 o = offsetof(struct INDEX_BUFFER, ihdr) + de_off;
2910     const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
2911     u32 used = le32_to_cpu(hdr->used);
2912 
2913     while (o < attr_off) {
2914         u16 esize;
2915 
2916         if (de_off >= used)
2917             break;
2918 
2919         esize = le16_to_cpu(e->size);
2920         if (!esize)
2921             break;
2922 
2923         o += esize;
2924         de_off += esize;
2925         e = Add2Ptr(e, esize);
2926     }
2927 
2928     return o == attr_off;
2929 }
2930 
2931 static inline void change_attr_size(struct MFT_REC *rec, struct ATTRIB *attr,
2932                     u32 nsize)
2933 {
2934     u32 asize = le32_to_cpu(attr->size);
2935     int dsize = nsize - asize;
2936     u8 *next = Add2Ptr(attr, asize);
2937     u32 used = le32_to_cpu(rec->used);
2938 
2939     memmove(Add2Ptr(attr, nsize), next, used - PtrOffset(rec, next));
2940 
2941     rec->used = cpu_to_le32(used + dsize);
2942     attr->size = cpu_to_le32(nsize);
2943 }
2944 
2945 struct OpenAttr {
2946     struct ATTRIB *attr;
2947     struct runs_tree *run1;
2948     struct runs_tree run0;
2949     struct ntfs_inode *ni;
2950     // CLST rno;
2951 };
2952 
2953 /*
2954  * cmp_type_and_name
2955  *
2956  * Return: 0 if 'attr' has the same type and name.
2957  */
2958 static inline int cmp_type_and_name(const struct ATTRIB *a1,
2959                     const struct ATTRIB *a2)
2960 {
2961     return a1->type != a2->type || a1->name_len != a2->name_len ||
2962            (a1->name_len && memcmp(attr_name(a1), attr_name(a2),
2963                        a1->name_len * sizeof(short)));
2964 }
2965 
2966 static struct OpenAttr *find_loaded_attr(struct ntfs_log *log,
2967                      const struct ATTRIB *attr, CLST rno)
2968 {
2969     struct OPEN_ATTR_ENRTY *oe = NULL;
2970 
2971     while ((oe = enum_rstbl(log->open_attr_tbl, oe))) {
2972         struct OpenAttr *op_attr;
2973 
2974         if (ino_get(&oe->ref) != rno)
2975             continue;
2976 
2977         op_attr = (struct OpenAttr *)oe->ptr;
2978         if (!cmp_type_and_name(op_attr->attr, attr))
2979             return op_attr;
2980     }
2981     return NULL;
2982 }
2983 
2984 static struct ATTRIB *attr_create_nonres_log(struct ntfs_sb_info *sbi,
2985                          enum ATTR_TYPE type, u64 size,
2986                          const u16 *name, size_t name_len,
2987                          __le16 flags)
2988 {
2989     struct ATTRIB *attr;
2990     u32 name_size = ALIGN(name_len * sizeof(short), 8);
2991     bool is_ext = flags & (ATTR_FLAG_COMPRESSED | ATTR_FLAG_SPARSED);
2992     u32 asize = name_size +
2993             (is_ext ? SIZEOF_NONRESIDENT_EX : SIZEOF_NONRESIDENT);
2994 
2995     attr = kzalloc(asize, GFP_NOFS);
2996     if (!attr)
2997         return NULL;
2998 
2999     attr->type = type;
3000     attr->size = cpu_to_le32(asize);
3001     attr->flags = flags;
3002     attr->non_res = 1;
3003     attr->name_len = name_len;
3004 
3005     attr->nres.evcn = cpu_to_le64((u64)bytes_to_cluster(sbi, size) - 1);
3006     attr->nres.alloc_size = cpu_to_le64(ntfs_up_cluster(sbi, size));
3007     attr->nres.data_size = cpu_to_le64(size);
3008     attr->nres.valid_size = attr->nres.data_size;
3009     if (is_ext) {
3010         attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
3011         if (is_attr_compressed(attr))
3012             attr->nres.c_unit = COMPRESSION_UNIT;
3013 
3014         attr->nres.run_off =
3015             cpu_to_le16(SIZEOF_NONRESIDENT_EX + name_size);
3016         memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT_EX), name,
3017                name_len * sizeof(short));
3018     } else {
3019         attr->name_off = SIZEOF_NONRESIDENT_LE;
3020         attr->nres.run_off =
3021             cpu_to_le16(SIZEOF_NONRESIDENT + name_size);
3022         memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT), name,
3023                name_len * sizeof(short));
3024     }
3025 
3026     return attr;
3027 }
3028 
3029 /*
3030  * do_action - Common routine for the Redo and Undo Passes.
3031  * @rlsn: If it is NULL then undo.
3032  */
3033 static int do_action(struct ntfs_log *log, struct OPEN_ATTR_ENRTY *oe,
3034              const struct LOG_REC_HDR *lrh, u32 op, void *data,
3035              u32 dlen, u32 rec_len, const u64 *rlsn)
3036 {
3037     int err = 0;
3038     struct ntfs_sb_info *sbi = log->ni->mi.sbi;
3039     struct inode *inode = NULL, *inode_parent;
3040     struct mft_inode *mi = NULL, *mi2_child = NULL;
3041     CLST rno = 0, rno_base = 0;
3042     struct INDEX_BUFFER *ib = NULL;
3043     struct MFT_REC *rec = NULL;
3044     struct ATTRIB *attr = NULL, *attr2;
3045     struct INDEX_HDR *hdr;
3046     struct INDEX_ROOT *root;
3047     struct NTFS_DE *e, *e1, *e2;
3048     struct NEW_ATTRIBUTE_SIZES *new_sz;
3049     struct ATTR_FILE_NAME *fname;
3050     struct OpenAttr *oa, *oa2;
3051     u32 nsize, t32, asize, used, esize, bmp_off, bmp_bits;
3052     u16 id, id2;
3053     u32 record_size = sbi->record_size;
3054     u64 t64;
3055     u16 roff = le16_to_cpu(lrh->record_off);
3056     u16 aoff = le16_to_cpu(lrh->attr_off);
3057     u64 lco = 0;
3058     u64 cbo = (u64)le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
3059     u64 tvo = le64_to_cpu(lrh->target_vcn) << sbi->cluster_bits;
3060     u64 vbo = cbo + tvo;
3061     void *buffer_le = NULL;
3062     u32 bytes = 0;
3063     bool a_dirty = false;
3064     u16 data_off;
3065 
3066     oa = oe->ptr;
3067 
3068     /* Big switch to prepare. */
3069     switch (op) {
3070     /* ============================================================
3071      * Process MFT records, as described by the current log record.
3072      * ============================================================
3073      */
3074     case InitializeFileRecordSegment:
3075     case DeallocateFileRecordSegment:
3076     case WriteEndOfFileRecordSegment:
3077     case CreateAttribute:
3078     case DeleteAttribute:
3079     case UpdateResidentValue:
3080     case UpdateMappingPairs:
3081     case SetNewAttributeSizes:
3082     case AddIndexEntryRoot:
3083     case DeleteIndexEntryRoot:
3084     case SetIndexEntryVcnRoot:
3085     case UpdateFileNameRoot:
3086     case UpdateRecordDataRoot:
3087     case ZeroEndOfFileRecord:
3088         rno = vbo >> sbi->record_bits;
3089         inode = ilookup(sbi->sb, rno);
3090         if (inode) {
3091             mi = &ntfs_i(inode)->mi;
3092         } else if (op == InitializeFileRecordSegment) {
3093             mi = kzalloc(sizeof(struct mft_inode), GFP_NOFS);
3094             if (!mi)
3095                 return -ENOMEM;
3096             err = mi_format_new(mi, sbi, rno, 0, false);
3097             if (err)
3098                 goto out;
3099         } else {
3100             /* Read from disk. */
3101             err = mi_get(sbi, rno, &mi);
3102             if (err)
3103                 return err;
3104         }
3105         rec = mi->mrec;
3106 
3107         if (op == DeallocateFileRecordSegment)
3108             goto skip_load_parent;
3109 
3110         if (InitializeFileRecordSegment != op) {
3111             if (rec->rhdr.sign == NTFS_BAAD_SIGNATURE)
3112                 goto dirty_vol;
3113             if (!check_lsn(&rec->rhdr, rlsn))
3114                 goto out;
3115             if (!check_file_record(rec, NULL, sbi))
3116                 goto dirty_vol;
3117             attr = Add2Ptr(rec, roff);
3118         }
3119 
3120         if (is_rec_base(rec) || InitializeFileRecordSegment == op) {
3121             rno_base = rno;
3122             goto skip_load_parent;
3123         }
3124 
3125         rno_base = ino_get(&rec->parent_ref);
3126         inode_parent = ntfs_iget5(sbi->sb, &rec->parent_ref, NULL);
3127         if (IS_ERR(inode_parent))
3128             goto skip_load_parent;
3129 
3130         if (is_bad_inode(inode_parent)) {
3131             iput(inode_parent);
3132             goto skip_load_parent;
3133         }
3134 
3135         if (ni_load_mi_ex(ntfs_i(inode_parent), rno, &mi2_child)) {
3136             iput(inode_parent);
3137         } else {
3138             if (mi2_child->mrec != mi->mrec)
3139                 memcpy(mi2_child->mrec, mi->mrec,
3140                        sbi->record_size);
3141 
3142             if (inode)
3143                 iput(inode);
3144             else if (mi)
3145                 mi_put(mi);
3146 
3147             inode = inode_parent;
3148             mi = mi2_child;
3149             rec = mi2_child->mrec;
3150             attr = Add2Ptr(rec, roff);
3151         }
3152 
3153 skip_load_parent:
3154         inode_parent = NULL;
3155         break;
3156 
3157     /*
3158      * Process attributes, as described by the current log record.
3159      */
3160     case UpdateNonresidentValue:
3161     case AddIndexEntryAllocation:
3162     case DeleteIndexEntryAllocation:
3163     case WriteEndOfIndexBuffer:
3164     case SetIndexEntryVcnAllocation:
3165     case UpdateFileNameAllocation:
3166     case SetBitsInNonresidentBitMap:
3167     case ClearBitsInNonresidentBitMap:
3168     case UpdateRecordDataAllocation:
3169         attr = oa->attr;
3170         bytes = UpdateNonresidentValue == op ? dlen : 0;
3171         lco = (u64)le16_to_cpu(lrh->lcns_follow) << sbi->cluster_bits;
3172 
3173         if (attr->type == ATTR_ALLOC) {
3174             t32 = le32_to_cpu(oe->bytes_per_index);
3175             if (bytes < t32)
3176                 bytes = t32;
3177         }
3178 
3179         if (!bytes)
3180             bytes = lco - cbo;
3181 
3182         bytes += roff;
3183         if (attr->type == ATTR_ALLOC)
3184             bytes = (bytes + 511) & ~511; // align
3185 
3186         buffer_le = kmalloc(bytes, GFP_NOFS);
3187         if (!buffer_le)
3188             return -ENOMEM;
3189 
3190         err = ntfs_read_run_nb(sbi, oa->run1, vbo, buffer_le, bytes,
3191                        NULL);
3192         if (err)
3193             goto out;
3194 
3195         if (attr->type == ATTR_ALLOC && *(int *)buffer_le)
3196             ntfs_fix_post_read(buffer_le, bytes, false);
3197         break;
3198 
3199     default:
3200         WARN_ON(1);
3201     }
3202 
3203     /* Big switch to do operation. */
3204     switch (op) {
3205     case InitializeFileRecordSegment:
3206         if (roff + dlen > record_size)
3207             goto dirty_vol;
3208 
3209         memcpy(Add2Ptr(rec, roff), data, dlen);
3210         mi->dirty = true;
3211         break;
3212 
3213     case DeallocateFileRecordSegment:
3214         clear_rec_inuse(rec);
3215         le16_add_cpu(&rec->seq, 1);
3216         mi->dirty = true;
3217         break;
3218 
3219     case WriteEndOfFileRecordSegment:
3220         attr2 = (struct ATTRIB *)data;
3221         if (!check_if_attr(rec, lrh) || roff + dlen > record_size)
3222             goto dirty_vol;
3223 
3224         memmove(attr, attr2, dlen);
3225         rec->used = cpu_to_le32(ALIGN(roff + dlen, 8));
3226 
3227         mi->dirty = true;
3228         break;
3229 
3230     case CreateAttribute:
3231         attr2 = (struct ATTRIB *)data;
3232         asize = le32_to_cpu(attr2->size);
3233         used = le32_to_cpu(rec->used);
3234 
3235         if (!check_if_attr(rec, lrh) || dlen < SIZEOF_RESIDENT ||
3236             !IS_ALIGNED(asize, 8) ||
3237             Add2Ptr(attr2, asize) > Add2Ptr(lrh, rec_len) ||
3238             dlen > record_size - used) {
3239             goto dirty_vol;
3240         }
3241 
3242         memmove(Add2Ptr(attr, asize), attr, used - roff);
3243         memcpy(attr, attr2, asize);
3244 
3245         rec->used = cpu_to_le32(used + asize);
3246         id = le16_to_cpu(rec->next_attr_id);
3247         id2 = le16_to_cpu(attr2->id);
3248         if (id <= id2)
3249             rec->next_attr_id = cpu_to_le16(id2 + 1);
3250         if (is_attr_indexed(attr))
3251             le16_add_cpu(&rec->hard_links, 1);
3252 
3253         oa2 = find_loaded_attr(log, attr, rno_base);
3254         if (oa2) {
3255             void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3256                        GFP_NOFS);
3257             if (p2) {
3258                 // run_close(oa2->run1);
3259                 kfree(oa2->attr);
3260                 oa2->attr = p2;
3261             }
3262         }
3263 
3264         mi->dirty = true;
3265         break;
3266 
3267     case DeleteAttribute:
3268         asize = le32_to_cpu(attr->size);
3269         used = le32_to_cpu(rec->used);
3270 
3271         if (!check_if_attr(rec, lrh))
3272             goto dirty_vol;
3273 
3274         rec->used = cpu_to_le32(used - asize);
3275         if (is_attr_indexed(attr))
3276             le16_add_cpu(&rec->hard_links, -1);
3277 
3278         memmove(attr, Add2Ptr(attr, asize), used - asize - roff);
3279 
3280         mi->dirty = true;
3281         break;
3282 
3283     case UpdateResidentValue:
3284         nsize = aoff + dlen;
3285 
3286         if (!check_if_attr(rec, lrh))
3287             goto dirty_vol;
3288 
3289         asize = le32_to_cpu(attr->size);
3290         used = le32_to_cpu(rec->used);
3291 
3292         if (lrh->redo_len == lrh->undo_len) {
3293             if (nsize > asize)
3294                 goto dirty_vol;
3295             goto move_data;
3296         }
3297 
3298         if (nsize > asize && nsize - asize > record_size - used)
3299             goto dirty_vol;
3300 
3301         nsize = ALIGN(nsize, 8);
3302         data_off = le16_to_cpu(attr->res.data_off);
3303 
3304         if (nsize < asize) {
3305             memmove(Add2Ptr(attr, aoff), data, dlen);
3306             data = NULL; // To skip below memmove().
3307         }
3308 
3309         memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
3310             used - le16_to_cpu(lrh->record_off) - asize);
3311 
3312         rec->used = cpu_to_le32(used + nsize - asize);
3313         attr->size = cpu_to_le32(nsize);
3314         attr->res.data_size = cpu_to_le32(aoff + dlen - data_off);
3315 
3316 move_data:
3317         if (data)
3318             memmove(Add2Ptr(attr, aoff), data, dlen);
3319 
3320         oa2 = find_loaded_attr(log, attr, rno_base);
3321         if (oa2) {
3322             void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3323                        GFP_NOFS);
3324             if (p2) {
3325                 // run_close(&oa2->run0);
3326                 oa2->run1 = &oa2->run0;
3327                 kfree(oa2->attr);
3328                 oa2->attr = p2;
3329             }
3330         }
3331 
3332         mi->dirty = true;
3333         break;
3334 
3335     case UpdateMappingPairs:
3336         nsize = aoff + dlen;
3337         asize = le32_to_cpu(attr->size);
3338         used = le32_to_cpu(rec->used);
3339 
3340         if (!check_if_attr(rec, lrh) || !attr->non_res ||
3341             aoff < le16_to_cpu(attr->nres.run_off) || aoff > asize ||
3342             (nsize > asize && nsize - asize > record_size - used)) {
3343             goto dirty_vol;
3344         }
3345 
3346         nsize = ALIGN(nsize, 8);
3347 
3348         memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
3349             used - le16_to_cpu(lrh->record_off) - asize);
3350         rec->used = cpu_to_le32(used + nsize - asize);
3351         attr->size = cpu_to_le32(nsize);
3352         memmove(Add2Ptr(attr, aoff), data, dlen);
3353 
3354         if (run_get_highest_vcn(le64_to_cpu(attr->nres.svcn),
3355                     attr_run(attr), &t64)) {
3356             goto dirty_vol;
3357         }
3358 
3359         attr->nres.evcn = cpu_to_le64(t64);
3360         oa2 = find_loaded_attr(log, attr, rno_base);
3361         if (oa2 && oa2->attr->non_res)
3362             oa2->attr->nres.evcn = attr->nres.evcn;
3363 
3364         mi->dirty = true;
3365         break;
3366 
3367     case SetNewAttributeSizes:
3368         new_sz = data;
3369         if (!check_if_attr(rec, lrh) || !attr->non_res)
3370             goto dirty_vol;
3371 
3372         attr->nres.alloc_size = new_sz->alloc_size;
3373         attr->nres.data_size = new_sz->data_size;
3374         attr->nres.valid_size = new_sz->valid_size;
3375 
3376         if (dlen >= sizeof(struct NEW_ATTRIBUTE_SIZES))
3377             attr->nres.total_size = new_sz->total_size;
3378 
3379         oa2 = find_loaded_attr(log, attr, rno_base);
3380         if (oa2) {
3381             void *p2 = kmemdup(attr, le32_to_cpu(attr->size),
3382                        GFP_NOFS);
3383             if (p2) {
3384                 kfree(oa2->attr);
3385                 oa2->attr = p2;
3386             }
3387         }
3388         mi->dirty = true;
3389         break;
3390 
3391     case AddIndexEntryRoot:
3392         e = (struct NTFS_DE *)data;
3393         esize = le16_to_cpu(e->size);
3394         root = resident_data(attr);
3395         hdr = &root->ihdr;
3396         used = le32_to_cpu(hdr->used);
3397 
3398         if (!check_if_index_root(rec, lrh) ||
3399             !check_if_root_index(attr, hdr, lrh) ||
3400             Add2Ptr(data, esize) > Add2Ptr(lrh, rec_len) ||
3401             esize > le32_to_cpu(rec->total) - le32_to_cpu(rec->used)) {
3402             goto dirty_vol;
3403         }
3404 
3405         e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3406 
3407         change_attr_size(rec, attr, le32_to_cpu(attr->size) + esize);
3408 
3409         memmove(Add2Ptr(e1, esize), e1,
3410             PtrOffset(e1, Add2Ptr(hdr, used)));
3411         memmove(e1, e, esize);
3412 
3413         le32_add_cpu(&attr->res.data_size, esize);
3414         hdr->used = cpu_to_le32(used + esize);
3415         le32_add_cpu(&hdr->total, esize);
3416 
3417         mi->dirty = true;
3418         break;
3419 
3420     case DeleteIndexEntryRoot:
3421         root = resident_data(attr);
3422         hdr = &root->ihdr;
3423         used = le32_to_cpu(hdr->used);
3424 
3425         if (!check_if_index_root(rec, lrh) ||
3426             !check_if_root_index(attr, hdr, lrh)) {
3427             goto dirty_vol;
3428         }
3429 
3430         e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3431         esize = le16_to_cpu(e1->size);
3432         e2 = Add2Ptr(e1, esize);
3433 
3434         memmove(e1, e2, PtrOffset(e2, Add2Ptr(hdr, used)));
3435 
3436         le32_sub_cpu(&attr->res.data_size, esize);
3437         hdr->used = cpu_to_le32(used - esize);
3438         le32_sub_cpu(&hdr->total, esize);
3439 
3440         change_attr_size(rec, attr, le32_to_cpu(attr->size) - esize);
3441 
3442         mi->dirty = true;
3443         break;
3444 
3445     case SetIndexEntryVcnRoot:
3446         root = resident_data(attr);
3447         hdr = &root->ihdr;
3448 
3449         if (!check_if_index_root(rec, lrh) ||
3450             !check_if_root_index(attr, hdr, lrh)) {
3451             goto dirty_vol;
3452         }
3453 
3454         e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3455 
3456         de_set_vbn_le(e, *(__le64 *)data);
3457         mi->dirty = true;
3458         break;
3459 
3460     case UpdateFileNameRoot:
3461         root = resident_data(attr);
3462         hdr = &root->ihdr;
3463 
3464         if (!check_if_index_root(rec, lrh) ||
3465             !check_if_root_index(attr, hdr, lrh)) {
3466             goto dirty_vol;
3467         }
3468 
3469         e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3470         fname = (struct ATTR_FILE_NAME *)(e + 1);
3471         memmove(&fname->dup, data, sizeof(fname->dup)); //
3472         mi->dirty = true;
3473         break;
3474 
3475     case UpdateRecordDataRoot:
3476         root = resident_data(attr);
3477         hdr = &root->ihdr;
3478 
3479         if (!check_if_index_root(rec, lrh) ||
3480             !check_if_root_index(attr, hdr, lrh)) {
3481             goto dirty_vol;
3482         }
3483 
3484         e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
3485 
3486         memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
3487 
3488         mi->dirty = true;
3489         break;
3490 
3491     case ZeroEndOfFileRecord:
3492         if (roff + dlen > record_size)
3493             goto dirty_vol;
3494 
3495         memset(attr, 0, dlen);
3496         mi->dirty = true;
3497         break;
3498 
3499     case UpdateNonresidentValue:
3500         if (lco < cbo + roff + dlen)
3501             goto dirty_vol;
3502 
3503         memcpy(Add2Ptr(buffer_le, roff), data, dlen);
3504 
3505         a_dirty = true;
3506         if (attr->type == ATTR_ALLOC)
3507             ntfs_fix_pre_write(buffer_le, bytes);
3508         break;
3509 
3510     case AddIndexEntryAllocation:
3511         ib = Add2Ptr(buffer_le, roff);
3512         hdr = &ib->ihdr;
3513         e = data;
3514         esize = le16_to_cpu(e->size);
3515         e1 = Add2Ptr(ib, aoff);
3516 
3517         if (is_baad(&ib->rhdr))
3518             goto dirty_vol;
3519         if (!check_lsn(&ib->rhdr, rlsn))
3520             goto out;
3521 
3522         used = le32_to_cpu(hdr->used);
3523 
3524         if (!check_index_buffer(ib, bytes) ||
3525             !check_if_alloc_index(hdr, aoff) ||
3526             Add2Ptr(e, esize) > Add2Ptr(lrh, rec_len) ||
3527             used + esize > le32_to_cpu(hdr->total)) {
3528             goto dirty_vol;
3529         }
3530 
3531         memmove(Add2Ptr(e1, esize), e1,
3532             PtrOffset(e1, Add2Ptr(hdr, used)));
3533         memcpy(e1, e, esize);
3534 
3535         hdr->used = cpu_to_le32(used + esize);
3536 
3537         a_dirty = true;
3538 
3539         ntfs_fix_pre_write(&ib->rhdr, bytes);
3540         break;
3541 
3542     case DeleteIndexEntryAllocation:
3543         ib = Add2Ptr(buffer_le, roff);
3544         hdr = &ib->ihdr;
3545         e = Add2Ptr(ib, aoff);
3546         esize = le16_to_cpu(e->size);
3547 
3548         if (is_baad(&ib->rhdr))
3549             goto dirty_vol;
3550         if (!check_lsn(&ib->rhdr, rlsn))
3551             goto out;
3552 
3553         if (!check_index_buffer(ib, bytes) ||
3554             !check_if_alloc_index(hdr, aoff)) {
3555             goto dirty_vol;
3556         }
3557 
3558         e1 = Add2Ptr(e, esize);
3559         nsize = esize;
3560         used = le32_to_cpu(hdr->used);
3561 
3562         memmove(e, e1, PtrOffset(e1, Add2Ptr(hdr, used)));
3563 
3564         hdr->used = cpu_to_le32(used - nsize);
3565 
3566         a_dirty = true;
3567 
3568         ntfs_fix_pre_write(&ib->rhdr, bytes);
3569         break;
3570 
3571     case WriteEndOfIndexBuffer:
3572         ib = Add2Ptr(buffer_le, roff);
3573         hdr = &ib->ihdr;
3574         e = Add2Ptr(ib, aoff);
3575 
3576         if (is_baad(&ib->rhdr))
3577             goto dirty_vol;
3578         if (!check_lsn(&ib->rhdr, rlsn))
3579             goto out;
3580         if (!check_index_buffer(ib, bytes) ||
3581             !check_if_alloc_index(hdr, aoff) ||
3582             aoff + dlen > offsetof(struct INDEX_BUFFER, ihdr) +
3583                       le32_to_cpu(hdr->total)) {
3584             goto dirty_vol;
3585         }
3586 
3587         hdr->used = cpu_to_le32(dlen + PtrOffset(hdr, e));
3588         memmove(e, data, dlen);
3589 
3590         a_dirty = true;
3591         ntfs_fix_pre_write(&ib->rhdr, bytes);
3592         break;
3593 
3594     case SetIndexEntryVcnAllocation:
3595         ib = Add2Ptr(buffer_le, roff);
3596         hdr = &ib->ihdr;
3597         e = Add2Ptr(ib, aoff);
3598 
3599         if (is_baad(&ib->rhdr))
3600             goto dirty_vol;
3601 
3602         if (!check_lsn(&ib->rhdr, rlsn))
3603             goto out;
3604         if (!check_index_buffer(ib, bytes) ||
3605             !check_if_alloc_index(hdr, aoff)) {
3606             goto dirty_vol;
3607         }
3608 
3609         de_set_vbn_le(e, *(__le64 *)data);
3610 
3611         a_dirty = true;
3612         ntfs_fix_pre_write(&ib->rhdr, bytes);
3613         break;
3614 
3615     case UpdateFileNameAllocation:
3616         ib = Add2Ptr(buffer_le, roff);
3617         hdr = &ib->ihdr;
3618         e = Add2Ptr(ib, aoff);
3619 
3620         if (is_baad(&ib->rhdr))
3621             goto dirty_vol;
3622 
3623         if (!check_lsn(&ib->rhdr, rlsn))
3624             goto out;
3625         if (!check_index_buffer(ib, bytes) ||
3626             !check_if_alloc_index(hdr, aoff)) {
3627             goto dirty_vol;
3628         }
3629 
3630         fname = (struct ATTR_FILE_NAME *)(e + 1);
3631         memmove(&fname->dup, data, sizeof(fname->dup));
3632 
3633         a_dirty = true;
3634         ntfs_fix_pre_write(&ib->rhdr, bytes);
3635         break;
3636 
3637     case SetBitsInNonresidentBitMap:
3638         bmp_off =
3639             le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
3640         bmp_bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
3641 
3642         if (cbo + (bmp_off + 7) / 8 > lco ||
3643             cbo + ((bmp_off + bmp_bits + 7) / 8) > lco) {
3644             goto dirty_vol;
3645         }
3646 
3647         __bitmap_set(Add2Ptr(buffer_le, roff), bmp_off, bmp_bits);
3648         a_dirty = true;
3649         break;
3650 
3651     case ClearBitsInNonresidentBitMap:
3652         bmp_off =
3653             le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
3654         bmp_bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
3655 
3656         if (cbo + (bmp_off + 7) / 8 > lco ||
3657             cbo + ((bmp_off + bmp_bits + 7) / 8) > lco) {
3658             goto dirty_vol;
3659         }
3660 
3661         __bitmap_clear(Add2Ptr(buffer_le, roff), bmp_off, bmp_bits);
3662         a_dirty = true;
3663         break;
3664 
3665     case UpdateRecordDataAllocation:
3666         ib = Add2Ptr(buffer_le, roff);
3667         hdr = &ib->ihdr;
3668         e = Add2Ptr(ib, aoff);
3669 
3670         if (is_baad(&ib->rhdr))
3671             goto dirty_vol;
3672 
3673         if (!check_lsn(&ib->rhdr, rlsn))
3674             goto out;
3675         if (!check_index_buffer(ib, bytes) ||
3676             !check_if_alloc_index(hdr, aoff)) {
3677             goto dirty_vol;
3678         }
3679 
3680         memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
3681 
3682         a_dirty = true;
3683         ntfs_fix_pre_write(&ib->rhdr, bytes);
3684         break;
3685 
3686     default:
3687         WARN_ON(1);
3688     }
3689 
3690     if (rlsn) {
3691         __le64 t64 = cpu_to_le64(*rlsn);
3692 
3693         if (rec)
3694             rec->rhdr.lsn = t64;
3695         if (ib)
3696             ib->rhdr.lsn = t64;
3697     }
3698 
3699     if (mi && mi->dirty) {
3700         err = mi_write(mi, 0);
3701         if (err)
3702             goto out;
3703     }
3704 
3705     if (a_dirty) {
3706         attr = oa->attr;
3707         err = ntfs_sb_write_run(sbi, oa->run1, vbo, buffer_le, bytes, 0);
3708         if (err)
3709             goto out;
3710     }
3711 
3712 out:
3713 
3714     if (inode)
3715         iput(inode);
3716     else if (mi != mi2_child)
3717         mi_put(mi);
3718 
3719     kfree(buffer_le);
3720 
3721     return err;
3722 
3723 dirty_vol:
3724     log->set_dirty = true;
3725     goto out;
3726 }
3727 
3728 /*
3729  * log_replay - Replays log and empties it.
3730  *
3731  * This function is called during mount operation.
3732  * It replays log and empties it.
3733  * Initialized is set false if logfile contains '-1'.
3734  */
3735 int log_replay(struct ntfs_inode *ni, bool *initialized)
3736 {
3737     int err;
3738     struct ntfs_sb_info *sbi = ni->mi.sbi;
3739     struct ntfs_log *log;
3740 
3741     struct restart_info rst_info, rst_info2;
3742     u64 rec_lsn, ra_lsn, checkpt_lsn = 0, rlsn = 0;
3743     struct ATTR_NAME_ENTRY *attr_names = NULL;
3744     struct ATTR_NAME_ENTRY *ane;
3745     struct RESTART_TABLE *dptbl = NULL;
3746     struct RESTART_TABLE *trtbl = NULL;
3747     const struct RESTART_TABLE *rt;
3748     struct RESTART_TABLE *oatbl = NULL;
3749     struct inode *inode;
3750     struct OpenAttr *oa;
3751     struct ntfs_inode *ni_oe;
3752     struct ATTRIB *attr = NULL;
3753     u64 size, vcn, undo_next_lsn;
3754     CLST rno, lcn, lcn0, len0, clen;
3755     void *data;
3756     struct NTFS_RESTART *rst = NULL;
3757     struct lcb *lcb = NULL;
3758     struct OPEN_ATTR_ENRTY *oe;
3759     struct TRANSACTION_ENTRY *tr;
3760     struct DIR_PAGE_ENTRY *dp;
3761     u32 i, bytes_per_attr_entry;
3762     u32 l_size = ni->vfs_inode.i_size;
3763     u32 orig_file_size = l_size;
3764     u32 page_size, vbo, tail, off, dlen;
3765     u32 saved_len, rec_len, transact_id;
3766     bool use_second_page;
3767     struct RESTART_AREA *ra2, *ra = NULL;
3768     struct CLIENT_REC *ca, *cr;
3769     __le16 client;
3770     struct RESTART_HDR *rh;
3771     const struct LFS_RECORD_HDR *frh;
3772     const struct LOG_REC_HDR *lrh;
3773     bool is_mapped;
3774     bool is_ro = sb_rdonly(sbi->sb);
3775     u64 t64;
3776     u16 t16;
3777     u32 t32;
3778 
3779     /* Get the size of page. NOTE: To replay we can use default page. */
3780 #if PAGE_SIZE >= DefaultLogPageSize && PAGE_SIZE <= DefaultLogPageSize * 2
3781     page_size = norm_file_page(PAGE_SIZE, &l_size, true);
3782 #else
3783     page_size = norm_file_page(PAGE_SIZE, &l_size, false);
3784 #endif
3785     if (!page_size)
3786         return -EINVAL;
3787 
3788     log = kzalloc(sizeof(struct ntfs_log), GFP_NOFS);
3789     if (!log)
3790         return -ENOMEM;
3791 
3792     memset(&rst_info, 0, sizeof(struct restart_info));
3793 
3794     log->ni = ni;
3795     log->l_size = l_size;
3796     log->one_page_buf = kmalloc(page_size, GFP_NOFS);
3797     if (!log->one_page_buf) {
3798         err = -ENOMEM;
3799         goto out;
3800     }
3801 
3802     log->page_size = page_size;
3803     log->page_mask = page_size - 1;
3804     log->page_bits = blksize_bits(page_size);
3805 
3806     /* Look for a restart area on the disk. */
3807     err = log_read_rst(log, l_size, true, &rst_info);
3808     if (err)
3809         goto out;
3810 
3811     /* remember 'initialized' */
3812     *initialized = rst_info.initialized;
3813 
3814     if (!rst_info.restart) {
3815         if (rst_info.initialized) {
3816             /* No restart area but the file is not initialized. */
3817             err = -EINVAL;
3818             goto out;
3819         }
3820 
3821         log_init_pg_hdr(log, page_size, page_size, 1, 1);
3822         log_create(log, l_size, 0, get_random_int(), false, false);
3823 
3824         log->ra = ra;
3825 
3826         ra = log_create_ra(log);
3827         if (!ra) {
3828             err = -ENOMEM;
3829             goto out;
3830         }
3831         log->ra = ra;
3832         log->init_ra = true;
3833 
3834         goto process_log;
3835     }
3836 
3837     /*
3838      * If the restart offset above wasn't zero then we won't
3839      * look for a second restart.
3840      */
3841     if (rst_info.vbo)
3842         goto check_restart_area;
3843 
3844     memset(&rst_info2, 0, sizeof(struct restart_info));
3845     err = log_read_rst(log, l_size, false, &rst_info2);
3846     if (err)
3847         goto out;
3848 
3849     /* Determine which restart area to use. */
3850     if (!rst_info2.restart || rst_info2.last_lsn <= rst_info.last_lsn)
3851         goto use_first_page;
3852 
3853     use_second_page = true;
3854 
3855     if (rst_info.chkdsk_was_run && page_size != rst_info.vbo) {
3856         struct RECORD_PAGE_HDR *sp = NULL;
3857         bool usa_error;
3858 
3859         if (!read_log_page(log, page_size, &sp, &usa_error) &&
3860             sp->rhdr.sign == NTFS_CHKD_SIGNATURE) {
3861             use_second_page = false;
3862         }
3863         kfree(sp);
3864     }
3865 
3866     if (use_second_page) {
3867         kfree(rst_info.r_page);
3868         memcpy(&rst_info, &rst_info2, sizeof(struct restart_info));
3869         rst_info2.r_page = NULL;
3870     }
3871 
3872 use_first_page:
3873     kfree(rst_info2.r_page);
3874 
3875 check_restart_area:
3876     /*
3877      * If the restart area is at offset 0, we want
3878      * to write the second restart area first.
3879      */
3880     log->init_ra = !!rst_info.vbo;
3881 
3882     /* If we have a valid page then grab a pointer to the restart area. */
3883     ra2 = rst_info.valid_page
3884               ? Add2Ptr(rst_info.r_page,
3885                 le16_to_cpu(rst_info.r_page->ra_off))
3886               : NULL;
3887 
3888     if (rst_info.chkdsk_was_run ||
3889         (ra2 && ra2->client_idx[1] == LFS_NO_CLIENT_LE)) {
3890         bool wrapped = false;
3891         bool use_multi_page = false;
3892         u32 open_log_count;
3893 
3894         /* Do some checks based on whether we have a valid log page. */
3895         if (!rst_info.valid_page) {
3896             open_log_count = get_random_int();
3897             goto init_log_instance;
3898         }
3899         open_log_count = le32_to_cpu(ra2->open_log_count);
3900 
3901         /*
3902          * If the restart page size isn't changing then we want to
3903          * check how much work we need to do.
3904          */
3905         if (page_size != le32_to_cpu(rst_info.r_page->sys_page_size))
3906             goto init_log_instance;
3907 
3908 init_log_instance:
3909         log_init_pg_hdr(log, page_size, page_size, 1, 1);
3910 
3911         log_create(log, l_size, rst_info.last_lsn, open_log_count,
3912                wrapped, use_multi_page);
3913 
3914         ra = log_create_ra(log);
3915         if (!ra) {
3916             err = -ENOMEM;
3917             goto out;
3918         }
3919         log->ra = ra;
3920 
3921         /* Put the restart areas and initialize
3922          * the log file as required.
3923          */
3924         goto process_log;
3925     }
3926 
3927     if (!ra2) {
3928         err = -EINVAL;
3929         goto out;
3930     }
3931 
3932     /*
3933      * If the log page or the system page sizes have changed, we can't
3934      * use the log file. We must use the system page size instead of the
3935      * default size if there is not a clean shutdown.
3936      */
3937     t32 = le32_to_cpu(rst_info.r_page->sys_page_size);
3938     if (page_size != t32) {
3939         l_size = orig_file_size;
3940         page_size =
3941             norm_file_page(t32, &l_size, t32 == DefaultLogPageSize);
3942     }
3943 
3944     if (page_size != t32 ||
3945         page_size != le32_to_cpu(rst_info.r_page->page_size)) {
3946         err = -EINVAL;
3947         goto out;
3948     }
3949 
3950     /* If the file size has shrunk then we won't mount it. */
3951     if (l_size < le64_to_cpu(ra2->l_size)) {
3952         err = -EINVAL;
3953         goto out;
3954     }
3955 
3956     log_init_pg_hdr(log, page_size, page_size,
3957             le16_to_cpu(rst_info.r_page->major_ver),
3958             le16_to_cpu(rst_info.r_page->minor_ver));
3959 
3960     log->l_size = le64_to_cpu(ra2->l_size);
3961     log->seq_num_bits = le32_to_cpu(ra2->seq_num_bits);
3962     log->file_data_bits = sizeof(u64) * 8 - log->seq_num_bits;
3963     log->seq_num_mask = (8 << log->file_data_bits) - 1;
3964     log->last_lsn = le64_to_cpu(ra2->current_lsn);
3965     log->seq_num = log->last_lsn >> log->file_data_bits;
3966     log->ra_off = le16_to_cpu(rst_info.r_page->ra_off);
3967     log->restart_size = log->sys_page_size - log->ra_off;
3968     log->record_header_len = le16_to_cpu(ra2->rec_hdr_len);
3969     log->ra_size = le16_to_cpu(ra2->ra_len);
3970     log->data_off = le16_to_cpu(ra2->data_off);
3971     log->data_size = log->page_size - log->data_off;
3972     log->reserved = log->data_size - log->record_header_len;
3973 
3974     vbo = lsn_to_vbo(log, log->last_lsn);
3975 
3976     if (vbo < log->first_page) {
3977         /* This is a pseudo lsn. */
3978         log->l_flags |= NTFSLOG_NO_LAST_LSN;
3979         log->next_page = log->first_page;
3980         goto find_oldest;
3981     }
3982 
3983     /* Find the end of this log record. */
3984     off = final_log_off(log, log->last_lsn,
3985                 le32_to_cpu(ra2->last_lsn_data_len));
3986 
3987     /* If we wrapped the file then increment the sequence number. */
3988     if (off <= vbo) {
3989         log->seq_num += 1;
3990         log->l_flags |= NTFSLOG_WRAPPED;
3991     }
3992 
3993     /* Now compute the next log page to use. */
3994     vbo &= ~log->sys_page_mask;
3995     tail = log->page_size - (off & log->page_mask) - 1;
3996 
3997     /*
3998      *If we can fit another log record on the page,
3999      * move back a page the log file.
4000      */
4001     if (tail >= log->record_header_len) {
4002         log->l_flags |= NTFSLOG_REUSE_TAIL;
4003         log->next_page = vbo;
4004     } else {
4005         log->next_page = next_page_off(log, vbo);
4006     }
4007 
4008 find_oldest:
4009     /*
4010      * Find the oldest client lsn. Use the last
4011      * flushed lsn as a starting point.
4012      */
4013     log->oldest_lsn = log->last_lsn;
4014     oldest_client_lsn(Add2Ptr(ra2, le16_to_cpu(ra2->client_off)),
4015               ra2->client_idx[1], &log->oldest_lsn);
4016     log->oldest_lsn_off = lsn_to_vbo(log, log->oldest_lsn);
4017 
4018     if (log->oldest_lsn_off < log->first_page)
4019         log->l_flags |= NTFSLOG_NO_OLDEST_LSN;
4020 
4021     if (!(ra2->flags & RESTART_SINGLE_PAGE_IO))
4022         log->l_flags |= NTFSLOG_WRAPPED | NTFSLOG_MULTIPLE_PAGE_IO;
4023 
4024     log->current_openlog_count = le32_to_cpu(ra2->open_log_count);
4025     log->total_avail_pages = log->l_size - log->first_page;
4026     log->total_avail = log->total_avail_pages >> log->page_bits;
4027     log->max_current_avail = log->total_avail * log->reserved;
4028     log->total_avail = log->total_avail * log->data_size;
4029 
4030     log->current_avail = current_log_avail(log);
4031 
4032     ra = kzalloc(log->restart_size, GFP_NOFS);
4033     if (!ra) {
4034         err = -ENOMEM;
4035         goto out;
4036     }
4037     log->ra = ra;
4038 
4039     t16 = le16_to_cpu(ra2->client_off);
4040     if (t16 == offsetof(struct RESTART_AREA, clients)) {
4041         memcpy(ra, ra2, log->ra_size);
4042     } else {
4043         memcpy(ra, ra2, offsetof(struct RESTART_AREA, clients));
4044         memcpy(ra->clients, Add2Ptr(ra2, t16),
4045                le16_to_cpu(ra2->ra_len) - t16);
4046 
4047         log->current_openlog_count = get_random_int();
4048         ra->open_log_count = cpu_to_le32(log->current_openlog_count);
4049         log->ra_size = offsetof(struct RESTART_AREA, clients) +
4050                    sizeof(struct CLIENT_REC);
4051         ra->client_off =
4052             cpu_to_le16(offsetof(struct RESTART_AREA, clients));
4053         ra->ra_len = cpu_to_le16(log->ra_size);
4054     }
4055 
4056     le32_add_cpu(&ra->open_log_count, 1);
4057 
4058     /* Now we need to walk through looking for the last lsn. */
4059     err = last_log_lsn(log);
4060     if (err)
4061         goto out;
4062 
4063     log->current_avail = current_log_avail(log);
4064 
4065     /* Remember which restart area to write first. */
4066     log->init_ra = rst_info.vbo;
4067 
4068 process_log:
4069     /* 1.0, 1.1, 2.0 log->major_ver/minor_ver - short values. */
4070     switch ((log->major_ver << 16) + log->minor_ver) {
4071     case 0x10000:
4072     case 0x10001:
4073     case 0x20000:
4074         break;
4075     default:
4076         ntfs_warn(sbi->sb, "\x24LogFile version %d.%d is not supported",
4077               log->major_ver, log->minor_ver);
4078         err = -EOPNOTSUPP;
4079         log->set_dirty = true;
4080         goto out;
4081     }
4082 
4083     /* One client "NTFS" per logfile. */
4084     ca = Add2Ptr(ra, le16_to_cpu(ra->client_off));
4085 
4086     for (client = ra->client_idx[1];; client = cr->next_client) {
4087         if (client == LFS_NO_CLIENT_LE) {
4088             /* Insert "NTFS" client LogFile. */
4089             client = ra->client_idx[0];
4090             if (client == LFS_NO_CLIENT_LE) {
4091                 err = -EINVAL;
4092                 goto out;
4093             }
4094 
4095             t16 = le16_to_cpu(client);
4096             cr = ca + t16;
4097 
4098             remove_client(ca, cr, &ra->client_idx[0]);
4099 
4100             cr->restart_lsn = 0;
4101             cr->oldest_lsn = cpu_to_le64(log->oldest_lsn);
4102             cr->name_bytes = cpu_to_le32(8);
4103             cr->name[0] = cpu_to_le16('N');
4104             cr->name[1] = cpu_to_le16('T');
4105             cr->name[2] = cpu_to_le16('F');
4106             cr->name[3] = cpu_to_le16('S');
4107 
4108             add_client(ca, t16, &ra->client_idx[1]);
4109             break;
4110         }
4111 
4112         cr = ca + le16_to_cpu(client);
4113 
4114         if (cpu_to_le32(8) == cr->name_bytes &&
4115             cpu_to_le16('N') == cr->name[0] &&
4116             cpu_to_le16('T') == cr->name[1] &&
4117             cpu_to_le16('F') == cr->name[2] &&
4118             cpu_to_le16('S') == cr->name[3])
4119             break;
4120     }
4121 
4122     /* Update the client handle with the client block information. */
4123     log->client_id.seq_num = cr->seq_num;
4124     log->client_id.client_idx = client;
4125 
4126     err = read_rst_area(log, &rst, &ra_lsn);
4127     if (err)
4128         goto out;
4129 
4130     if (!rst)
4131         goto out;
4132 
4133     bytes_per_attr_entry = !rst->major_ver ? 0x2C : 0x28;
4134 
4135     checkpt_lsn = le64_to_cpu(rst->check_point_start);
4136     if (!checkpt_lsn)
4137         checkpt_lsn = ra_lsn;
4138 
4139     /* Allocate and Read the Transaction Table. */
4140     if (!rst->transact_table_len)
4141         goto check_dirty_page_table;
4142 
4143     t64 = le64_to_cpu(rst->transact_table_lsn);
4144     err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4145     if (err)
4146         goto out;
4147 
4148     lrh = lcb->log_rec;
4149     frh = lcb->lrh;
4150     rec_len = le32_to_cpu(frh->client_data_len);
4151 
4152     if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4153                bytes_per_attr_entry)) {
4154         err = -EINVAL;
4155         goto out;
4156     }
4157 
4158     t16 = le16_to_cpu(lrh->redo_off);
4159 
4160     rt = Add2Ptr(lrh, t16);
4161     t32 = rec_len - t16;
4162 
4163     /* Now check that this is a valid restart table. */
4164     if (!check_rstbl(rt, t32)) {
4165         err = -EINVAL;
4166         goto out;
4167     }
4168 
4169     trtbl = kmemdup(rt, t32, GFP_NOFS);
4170     if (!trtbl) {
4171         err = -ENOMEM;
4172         goto out;
4173     }
4174 
4175     lcb_put(lcb);
4176     lcb = NULL;
4177 
4178 check_dirty_page_table:
4179     /* The next record back should be the Dirty Pages Table. */
4180     if (!rst->dirty_pages_len)
4181         goto check_attribute_names;
4182 
4183     t64 = le64_to_cpu(rst->dirty_pages_table_lsn);
4184     err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4185     if (err)
4186         goto out;
4187 
4188     lrh = lcb->log_rec;
4189     frh = lcb->lrh;
4190     rec_len = le32_to_cpu(frh->client_data_len);
4191 
4192     if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4193                bytes_per_attr_entry)) {
4194         err = -EINVAL;
4195         goto out;
4196     }
4197 
4198     t16 = le16_to_cpu(lrh->redo_off);
4199 
4200     rt = Add2Ptr(lrh, t16);
4201     t32 = rec_len - t16;
4202 
4203     /* Now check that this is a valid restart table. */
4204     if (!check_rstbl(rt, t32)) {
4205         err = -EINVAL;
4206         goto out;
4207     }
4208 
4209     dptbl = kmemdup(rt, t32, GFP_NOFS);
4210     if (!dptbl) {
4211         err = -ENOMEM;
4212         goto out;
4213     }
4214 
4215     /* Convert Ra version '0' into version '1'. */
4216     if (rst->major_ver)
4217         goto end_conv_1;
4218 
4219     dp = NULL;
4220     while ((dp = enum_rstbl(dptbl, dp))) {
4221         struct DIR_PAGE_ENTRY_32 *dp0 = (struct DIR_PAGE_ENTRY_32 *)dp;
4222         // NOTE: Danger. Check for of boundary.
4223         memmove(&dp->vcn, &dp0->vcn_low,
4224             2 * sizeof(u64) +
4225                 le32_to_cpu(dp->lcns_follow) * sizeof(u64));
4226     }
4227 
4228 end_conv_1:
4229     lcb_put(lcb);
4230     lcb = NULL;
4231 
4232     /*
4233      * Go through the table and remove the duplicates,
4234      * remembering the oldest lsn values.
4235      */
4236     if (sbi->cluster_size <= log->page_size)
4237         goto trace_dp_table;
4238 
4239     dp = NULL;
4240     while ((dp = enum_rstbl(dptbl, dp))) {
4241         struct DIR_PAGE_ENTRY *next = dp;
4242 
4243         while ((next = enum_rstbl(dptbl, next))) {
4244             if (next->target_attr == dp->target_attr &&
4245                 next->vcn == dp->vcn) {
4246                 if (le64_to_cpu(next->oldest_lsn) <
4247                     le64_to_cpu(dp->oldest_lsn)) {
4248                     dp->oldest_lsn = next->oldest_lsn;
4249                 }
4250 
4251                 free_rsttbl_idx(dptbl, PtrOffset(dptbl, next));
4252             }
4253         }
4254     }
4255 trace_dp_table:
4256 check_attribute_names:
4257     /* The next record should be the Attribute Names. */
4258     if (!rst->attr_names_len)
4259         goto check_attr_table;
4260 
4261     t64 = le64_to_cpu(rst->attr_names_lsn);
4262     err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4263     if (err)
4264         goto out;
4265 
4266     lrh = lcb->log_rec;
4267     frh = lcb->lrh;
4268     rec_len = le32_to_cpu(frh->client_data_len);
4269 
4270     if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4271                bytes_per_attr_entry)) {
4272         err = -EINVAL;
4273         goto out;
4274     }
4275 
4276     t32 = lrh_length(lrh);
4277     rec_len -= t32;
4278 
4279     attr_names = kmemdup(Add2Ptr(lrh, t32), rec_len, GFP_NOFS);
4280 
4281     lcb_put(lcb);
4282     lcb = NULL;
4283 
4284 check_attr_table:
4285     /* The next record should be the attribute Table. */
4286     if (!rst->open_attr_len)
4287         goto check_attribute_names2;
4288 
4289     t64 = le64_to_cpu(rst->open_attr_table_lsn);
4290     err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
4291     if (err)
4292         goto out;
4293 
4294     lrh = lcb->log_rec;
4295     frh = lcb->lrh;
4296     rec_len = le32_to_cpu(frh->client_data_len);
4297 
4298     if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
4299                bytes_per_attr_entry)) {
4300         err = -EINVAL;
4301         goto out;
4302     }
4303 
4304     t16 = le16_to_cpu(lrh->redo_off);
4305 
4306     rt = Add2Ptr(lrh, t16);
4307     t32 = rec_len - t16;
4308 
4309     if (!check_rstbl(rt, t32)) {
4310         err = -EINVAL;
4311         goto out;
4312     }
4313 
4314     oatbl = kmemdup(rt, t32, GFP_NOFS);
4315     if (!oatbl) {
4316         err = -ENOMEM;
4317         goto out;
4318     }
4319 
4320     log->open_attr_tbl = oatbl;
4321 
4322     /* Clear all of the Attr pointers. */
4323     oe = NULL;
4324     while ((oe = enum_rstbl(oatbl, oe))) {
4325         if (!rst->major_ver) {
4326             struct OPEN_ATTR_ENRTY_32 oe0;
4327 
4328             /* Really 'oe' points to OPEN_ATTR_ENRTY_32. */
4329             memcpy(&oe0, oe, SIZEOF_OPENATTRIBUTEENTRY0);
4330 
4331             oe->bytes_per_index = oe0.bytes_per_index;
4332             oe->type = oe0.type;
4333             oe->is_dirty_pages = oe0.is_dirty_pages;
4334             oe->name_len = 0;
4335             oe->ref = oe0.ref;
4336             oe->open_record_lsn = oe0.open_record_lsn;
4337         }
4338 
4339         oe->is_attr_name = 0;
4340         oe->ptr = NULL;
4341     }
4342 
4343     lcb_put(lcb);
4344     lcb = NULL;
4345 
4346 check_attribute_names2:
4347     if (!rst->attr_names_len)
4348         goto trace_attribute_table;
4349 
4350     ane = attr_names;
4351     if (!oatbl)
4352         goto trace_attribute_table;
4353     while (ane->off) {
4354         /* TODO: Clear table on exit! */
4355         oe = Add2Ptr(oatbl, le16_to_cpu(ane->off));
4356         t16 = le16_to_cpu(ane->name_bytes);
4357         oe->name_len = t16 / sizeof(short);
4358         oe->ptr = ane->name;
4359         oe->is_attr_name = 2;
4360         ane = Add2Ptr(ane, sizeof(struct ATTR_NAME_ENTRY) + t16);
4361     }
4362 
4363 trace_attribute_table:
4364     /*
4365      * If the checkpt_lsn is zero, then this is a freshly
4366      * formatted disk and we have no work to do.
4367      */
4368     if (!checkpt_lsn) {
4369         err = 0;
4370         goto out;
4371     }
4372 
4373     if (!oatbl) {
4374         oatbl = init_rsttbl(bytes_per_attr_entry, 8);
4375         if (!oatbl) {
4376             err = -ENOMEM;
4377             goto out;
4378         }
4379     }
4380 
4381     log->open_attr_tbl = oatbl;
4382 
4383     /* Start the analysis pass from the Checkpoint lsn. */
4384     rec_lsn = checkpt_lsn;
4385 
4386     /* Read the first lsn. */
4387     err = read_log_rec_lcb(log, checkpt_lsn, lcb_ctx_next, &lcb);
4388     if (err)
4389         goto out;
4390 
4391     /* Loop to read all subsequent records to the end of the log file. */
4392 next_log_record_analyze:
4393     err = read_next_log_rec(log, lcb, &rec_lsn);
4394     if (err)
4395         goto out;
4396 
4397     if (!rec_lsn)
4398         goto end_log_records_enumerate;
4399 
4400     frh = lcb->lrh;
4401     transact_id = le32_to_cpu(frh->transact_id);
4402     rec_len = le32_to_cpu(frh->client_data_len);
4403     lrh = lcb->log_rec;
4404 
4405     if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
4406         err = -EINVAL;
4407         goto out;
4408     }
4409 
4410     /*
4411      * The first lsn after the previous lsn remembered
4412      * the checkpoint is the first candidate for the rlsn.
4413      */
4414     if (!rlsn)
4415         rlsn = rec_lsn;
4416 
4417     if (LfsClientRecord != frh->record_type)
4418         goto next_log_record_analyze;
4419 
4420     /*
4421      * Now update the Transaction Table for this transaction. If there
4422      * is no entry present or it is unallocated we allocate the entry.
4423      */
4424     if (!trtbl) {
4425         trtbl = init_rsttbl(sizeof(struct TRANSACTION_ENTRY),
4426                     INITIAL_NUMBER_TRANSACTIONS);
4427         if (!trtbl) {
4428             err = -ENOMEM;
4429             goto out;
4430         }
4431     }
4432 
4433     tr = Add2Ptr(trtbl, transact_id);
4434 
4435     if (transact_id >= bytes_per_rt(trtbl) ||
4436         tr->next != RESTART_ENTRY_ALLOCATED_LE) {
4437         tr = alloc_rsttbl_from_idx(&trtbl, transact_id);
4438         if (!tr) {
4439             err = -ENOMEM;
4440             goto out;
4441         }
4442         tr->transact_state = TransactionActive;
4443         tr->first_lsn = cpu_to_le64(rec_lsn);
4444     }
4445 
4446     tr->prev_lsn = tr->undo_next_lsn = cpu_to_le64(rec_lsn);
4447 
4448     /*
4449      * If this is a compensation log record, then change
4450      * the undo_next_lsn to be the undo_next_lsn of this record.
4451      */
4452     if (lrh->undo_op == cpu_to_le16(CompensationLogRecord))
4453         tr->undo_next_lsn = frh->client_undo_next_lsn;
4454 
4455     /* Dispatch to handle log record depending on type. */
4456     switch (le16_to_cpu(lrh->redo_op)) {
4457     case InitializeFileRecordSegment:
4458     case DeallocateFileRecordSegment:
4459     case WriteEndOfFileRecordSegment:
4460     case CreateAttribute:
4461     case DeleteAttribute:
4462     case UpdateResidentValue:
4463     case UpdateNonresidentValue:
4464     case UpdateMappingPairs:
4465     case SetNewAttributeSizes:
4466     case AddIndexEntryRoot:
4467     case DeleteIndexEntryRoot:
4468     case AddIndexEntryAllocation:
4469     case DeleteIndexEntryAllocation:
4470     case WriteEndOfIndexBuffer:
4471     case SetIndexEntryVcnRoot:
4472     case SetIndexEntryVcnAllocation:
4473     case UpdateFileNameRoot:
4474     case UpdateFileNameAllocation:
4475     case SetBitsInNonresidentBitMap:
4476     case ClearBitsInNonresidentBitMap:
4477     case UpdateRecordDataRoot:
4478     case UpdateRecordDataAllocation:
4479     case ZeroEndOfFileRecord:
4480         t16 = le16_to_cpu(lrh->target_attr);
4481         t64 = le64_to_cpu(lrh->target_vcn);
4482         dp = find_dp(dptbl, t16, t64);
4483 
4484         if (dp)
4485             goto copy_lcns;
4486 
4487         /*
4488          * Calculate the number of clusters per page the system
4489          * which wrote the checkpoint, possibly creating the table.
4490          */
4491         if (dptbl) {
4492             t32 = (le16_to_cpu(dptbl->size) -
4493                    sizeof(struct DIR_PAGE_ENTRY)) /
4494                   sizeof(u64);
4495         } else {
4496             t32 = log->clst_per_page;
4497             kfree(dptbl);
4498             dptbl = init_rsttbl(struct_size(dp, page_lcns, t32),
4499                         32);
4500             if (!dptbl) {
4501                 err = -ENOMEM;
4502                 goto out;
4503             }
4504         }
4505 
4506         dp = alloc_rsttbl_idx(&dptbl);
4507         if (!dp) {
4508             err = -ENOMEM;
4509             goto out;
4510         }
4511         dp->target_attr = cpu_to_le32(t16);
4512         dp->transfer_len = cpu_to_le32(t32 << sbi->cluster_bits);
4513         dp->lcns_follow = cpu_to_le32(t32);
4514         dp->vcn = cpu_to_le64(t64 & ~((u64)t32 - 1));
4515         dp->oldest_lsn = cpu_to_le64(rec_lsn);
4516 
4517 copy_lcns:
4518         /*
4519          * Copy the Lcns from the log record into the Dirty Page Entry.
4520          * TODO: For different page size support, must somehow make
4521          * whole routine a loop, case Lcns do not fit below.
4522          */
4523         t16 = le16_to_cpu(lrh->lcns_follow);
4524         for (i = 0; i < t16; i++) {
4525             size_t j = (size_t)(le64_to_cpu(lrh->target_vcn) -
4526                         le64_to_cpu(dp->vcn));
4527             dp->page_lcns[j + i] = lrh->page_lcns[i];
4528         }
4529 
4530         goto next_log_record_analyze;
4531 
4532     case DeleteDirtyClusters: {
4533         u32 range_count =
4534             le16_to_cpu(lrh->redo_len) / sizeof(struct LCN_RANGE);
4535         const struct LCN_RANGE *r =
4536             Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
4537 
4538         /* Loop through all of the Lcn ranges this log record. */
4539         for (i = 0; i < range_count; i++, r++) {
4540             u64 lcn0 = le64_to_cpu(r->lcn);
4541             u64 lcn_e = lcn0 + le64_to_cpu(r->len) - 1;
4542 
4543             dp = NULL;
4544             while ((dp = enum_rstbl(dptbl, dp))) {
4545                 u32 j;
4546 
4547                 t32 = le32_to_cpu(dp->lcns_follow);
4548                 for (j = 0; j < t32; j++) {
4549                     t64 = le64_to_cpu(dp->page_lcns[j]);
4550                     if (t64 >= lcn0 && t64 <= lcn_e)
4551                         dp->page_lcns[j] = 0;
4552                 }
4553             }
4554         }
4555         goto next_log_record_analyze;
4556         ;
4557     }
4558 
4559     case OpenNonresidentAttribute:
4560         t16 = le16_to_cpu(lrh->target_attr);
4561         if (t16 >= bytes_per_rt(oatbl)) {
4562             /*
4563              * Compute how big the table needs to be.
4564              * Add 10 extra entries for some cushion.
4565              */
4566             u32 new_e = t16 / le16_to_cpu(oatbl->size);
4567 
4568             new_e += 10 - le16_to_cpu(oatbl->used);
4569 
4570             oatbl = extend_rsttbl(oatbl, new_e, ~0u);
4571             log->open_attr_tbl = oatbl;
4572             if (!oatbl) {
4573                 err = -ENOMEM;
4574                 goto out;
4575             }
4576         }
4577 
4578         /* Point to the entry being opened. */
4579         oe = alloc_rsttbl_from_idx(&oatbl, t16);
4580         log->open_attr_tbl = oatbl;
4581         if (!oe) {
4582             err = -ENOMEM;
4583             goto out;
4584         }
4585 
4586         /* Initialize this entry from the log record. */
4587         t16 = le16_to_cpu(lrh->redo_off);
4588         if (!rst->major_ver) {
4589             /* Convert version '0' into version '1'. */
4590             struct OPEN_ATTR_ENRTY_32 *oe0 = Add2Ptr(lrh, t16);
4591 
4592             oe->bytes_per_index = oe0->bytes_per_index;
4593             oe->type = oe0->type;
4594             oe->is_dirty_pages = oe0->is_dirty_pages;
4595             oe->name_len = 0; //oe0.name_len;
4596             oe->ref = oe0->ref;
4597             oe->open_record_lsn = oe0->open_record_lsn;
4598         } else {
4599             memcpy(oe, Add2Ptr(lrh, t16), bytes_per_attr_entry);
4600         }
4601 
4602         t16 = le16_to_cpu(lrh->undo_len);
4603         if (t16) {
4604             oe->ptr = kmalloc(t16, GFP_NOFS);
4605             if (!oe->ptr) {
4606                 err = -ENOMEM;
4607                 goto out;
4608             }
4609             oe->name_len = t16 / sizeof(short);
4610             memcpy(oe->ptr,
4611                    Add2Ptr(lrh, le16_to_cpu(lrh->undo_off)), t16);
4612             oe->is_attr_name = 1;
4613         } else {
4614             oe->ptr = NULL;
4615             oe->is_attr_name = 0;
4616         }
4617 
4618         goto next_log_record_analyze;
4619 
4620     case HotFix:
4621         t16 = le16_to_cpu(lrh->target_attr);
4622         t64 = le64_to_cpu(lrh->target_vcn);
4623         dp = find_dp(dptbl, t16, t64);
4624         if (dp) {
4625             size_t j = le64_to_cpu(lrh->target_vcn) -
4626                    le64_to_cpu(dp->vcn);
4627             if (dp->page_lcns[j])
4628                 dp->page_lcns[j] = lrh->page_lcns[0];
4629         }
4630         goto next_log_record_analyze;
4631 
4632     case EndTopLevelAction:
4633         tr = Add2Ptr(trtbl, transact_id);
4634         tr->prev_lsn = cpu_to_le64(rec_lsn);
4635         tr->undo_next_lsn = frh->client_undo_next_lsn;
4636         goto next_log_record_analyze;
4637 
4638     case PrepareTransaction:
4639         tr = Add2Ptr(trtbl, transact_id);
4640         tr->transact_state = TransactionPrepared;
4641         goto next_log_record_analyze;
4642 
4643     case CommitTransaction:
4644         tr = Add2Ptr(trtbl, transact_id);
4645         tr->transact_state = TransactionCommitted;
4646         goto next_log_record_analyze;
4647 
4648     case ForgetTransaction:
4649         free_rsttbl_idx(trtbl, transact_id);
4650         goto next_log_record_analyze;
4651 
4652     case Noop:
4653     case OpenAttributeTableDump:
4654     case AttributeNamesDump:
4655     case DirtyPageTableDump:
4656     case TransactionTableDump:
4657         /* The following cases require no action the Analysis Pass. */
4658         goto next_log_record_analyze;
4659 
4660     default:
4661         /*
4662          * All codes will be explicitly handled.
4663          * If we see a code we do not expect, then we are trouble.
4664          */
4665         goto next_log_record_analyze;
4666     }
4667 
4668 end_log_records_enumerate:
4669     lcb_put(lcb);
4670     lcb = NULL;
4671 
4672     /*
4673      * Scan the Dirty Page Table and Transaction Table for
4674      * the lowest lsn, and return it as the Redo lsn.
4675      */
4676     dp = NULL;
4677     while ((dp = enum_rstbl(dptbl, dp))) {
4678         t64 = le64_to_cpu(dp->oldest_lsn);
4679         if (t64 && t64 < rlsn)
4680             rlsn = t64;
4681     }
4682 
4683     tr = NULL;
4684     while ((tr = enum_rstbl(trtbl, tr))) {
4685         t64 = le64_to_cpu(tr->first_lsn);
4686         if (t64 && t64 < rlsn)
4687             rlsn = t64;
4688     }
4689 
4690     /*
4691      * Only proceed if the Dirty Page Table or Transaction
4692      * table are not empty.
4693      */
4694     if ((!dptbl || !dptbl->total) && (!trtbl || !trtbl->total))
4695         goto end_reply;
4696 
4697     sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
4698     if (is_ro)
4699         goto out;
4700 
4701     /* Reopen all of the attributes with dirty pages. */
4702     oe = NULL;
4703 next_open_attribute:
4704 
4705     oe = enum_rstbl(oatbl, oe);
4706     if (!oe) {
4707         err = 0;
4708         dp = NULL;
4709         goto next_dirty_page;
4710     }
4711 
4712     oa = kzalloc(sizeof(struct OpenAttr), GFP_NOFS);
4713     if (!oa) {
4714         err = -ENOMEM;
4715         goto out;
4716     }
4717 
4718     inode = ntfs_iget5(sbi->sb, &oe->ref, NULL);
4719     if (IS_ERR(inode))
4720         goto fake_attr;
4721 
4722     if (is_bad_inode(inode)) {
4723         iput(inode);
4724 fake_attr:
4725         if (oa->ni) {
4726             iput(&oa->ni->vfs_inode);
4727             oa->ni = NULL;
4728         }
4729 
4730         attr = attr_create_nonres_log(sbi, oe->type, 0, oe->ptr,
4731                           oe->name_len, 0);
4732         if (!attr) {
4733             kfree(oa);
4734             err = -ENOMEM;
4735             goto out;
4736         }
4737         oa->attr = attr;
4738         oa->run1 = &oa->run0;
4739         goto final_oe;
4740     }
4741 
4742     ni_oe = ntfs_i(inode);
4743     oa->ni = ni_oe;
4744 
4745     attr = ni_find_attr(ni_oe, NULL, NULL, oe->type, oe->ptr, oe->name_len,
4746                 NULL, NULL);
4747 
4748     if (!attr)
4749         goto fake_attr;
4750 
4751     t32 = le32_to_cpu(attr->size);
4752     oa->attr = kmemdup(attr, t32, GFP_NOFS);
4753     if (!oa->attr)
4754         goto fake_attr;
4755 
4756     if (!S_ISDIR(inode->i_mode)) {
4757         if (attr->type == ATTR_DATA && !attr->name_len) {
4758             oa->run1 = &ni_oe->file.run;
4759             goto final_oe;
4760         }
4761     } else {
4762         if (attr->type == ATTR_ALLOC &&
4763             attr->name_len == ARRAY_SIZE(I30_NAME) &&
4764             !memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME))) {
4765             oa->run1 = &ni_oe->dir.alloc_run;
4766             goto final_oe;
4767         }
4768     }
4769 
4770     if (attr->non_res) {
4771         u16 roff = le16_to_cpu(attr->nres.run_off);
4772         CLST svcn = le64_to_cpu(attr->nres.svcn);
4773 
4774         err = run_unpack(&oa->run0, sbi, inode->i_ino, svcn,
4775                  le64_to_cpu(attr->nres.evcn), svcn,
4776                  Add2Ptr(attr, roff), t32 - roff);
4777         if (err < 0) {
4778             kfree(oa->attr);
4779             oa->attr = NULL;
4780             goto fake_attr;
4781         }
4782         err = 0;
4783     }
4784     oa->run1 = &oa->run0;
4785     attr = oa->attr;
4786 
4787 final_oe:
4788     if (oe->is_attr_name == 1)
4789         kfree(oe->ptr);
4790     oe->is_attr_name = 0;
4791     oe->ptr = oa;
4792     oe->name_len = attr->name_len;
4793 
4794     goto next_open_attribute;
4795 
4796     /*
4797      * Now loop through the dirty page table to extract all of the Vcn/Lcn.
4798      * Mapping that we have, and insert it into the appropriate run.
4799      */
4800 next_dirty_page:
4801     dp = enum_rstbl(dptbl, dp);
4802     if (!dp)
4803         goto do_redo_1;
4804 
4805     oe = Add2Ptr(oatbl, le32_to_cpu(dp->target_attr));
4806 
4807     if (oe->next != RESTART_ENTRY_ALLOCATED_LE)
4808         goto next_dirty_page;
4809 
4810     oa = oe->ptr;
4811     if (!oa)
4812         goto next_dirty_page;
4813 
4814     i = -1;
4815 next_dirty_page_vcn:
4816     i += 1;
4817     if (i >= le32_to_cpu(dp->lcns_follow))
4818         goto next_dirty_page;
4819 
4820     vcn = le64_to_cpu(dp->vcn) + i;
4821     size = (vcn + 1) << sbi->cluster_bits;
4822 
4823     if (!dp->page_lcns[i])
4824         goto next_dirty_page_vcn;
4825 
4826     rno = ino_get(&oe->ref);
4827     if (rno <= MFT_REC_MIRR &&
4828         size < (MFT_REC_VOL + 1) * sbi->record_size &&
4829         oe->type == ATTR_DATA) {
4830         goto next_dirty_page_vcn;
4831     }
4832 
4833     lcn = le64_to_cpu(dp->page_lcns[i]);
4834 
4835     if ((!run_lookup_entry(oa->run1, vcn, &lcn0, &len0, NULL) ||
4836          lcn0 != lcn) &&
4837         !run_add_entry(oa->run1, vcn, lcn, 1, false)) {
4838         err = -ENOMEM;
4839         goto out;
4840     }
4841     attr = oa->attr;
4842     t64 = le64_to_cpu(attr->nres.alloc_size);
4843     if (size > t64) {
4844         attr->nres.valid_size = attr->nres.data_size =
4845             attr->nres.alloc_size = cpu_to_le64(size);
4846     }
4847     goto next_dirty_page_vcn;
4848 
4849 do_redo_1:
4850     /*
4851      * Perform the Redo Pass, to restore all of the dirty pages to the same
4852      * contents that they had immediately before the crash. If the dirty
4853      * page table is empty, then we can skip the entire Redo Pass.
4854      */
4855     if (!dptbl || !dptbl->total)
4856         goto do_undo_action;
4857 
4858     rec_lsn = rlsn;
4859 
4860     /*
4861      * Read the record at the Redo lsn, before falling
4862      * into common code to handle each record.
4863      */
4864     err = read_log_rec_lcb(log, rlsn, lcb_ctx_next, &lcb);
4865     if (err)
4866         goto out;
4867 
4868     /*
4869      * Now loop to read all of our log records forwards, until
4870      * we hit the end of the file, cleaning up at the end.
4871      */
4872 do_action_next:
4873     frh = lcb->lrh;
4874 
4875     if (LfsClientRecord != frh->record_type)
4876         goto read_next_log_do_action;
4877 
4878     transact_id = le32_to_cpu(frh->transact_id);
4879     rec_len = le32_to_cpu(frh->client_data_len);
4880     lrh = lcb->log_rec;
4881 
4882     if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
4883         err = -EINVAL;
4884         goto out;
4885     }
4886 
4887     /* Ignore log records that do not update pages. */
4888     if (lrh->lcns_follow)
4889         goto find_dirty_page;
4890 
4891     goto read_next_log_do_action;
4892 
4893 find_dirty_page:
4894     t16 = le16_to_cpu(lrh->target_attr);
4895     t64 = le64_to_cpu(lrh->target_vcn);
4896     dp = find_dp(dptbl, t16, t64);
4897 
4898     if (!dp)
4899         goto read_next_log_do_action;
4900 
4901     if (rec_lsn < le64_to_cpu(dp->oldest_lsn))
4902         goto read_next_log_do_action;
4903 
4904     t16 = le16_to_cpu(lrh->target_attr);
4905     if (t16 >= bytes_per_rt(oatbl)) {
4906         err = -EINVAL;
4907         goto out;
4908     }
4909 
4910     oe = Add2Ptr(oatbl, t16);
4911 
4912     if (oe->next != RESTART_ENTRY_ALLOCATED_LE) {
4913         err = -EINVAL;
4914         goto out;
4915     }
4916 
4917     oa = oe->ptr;
4918 
4919     if (!oa) {
4920         err = -EINVAL;
4921         goto out;
4922     }
4923     attr = oa->attr;
4924 
4925     vcn = le64_to_cpu(lrh->target_vcn);
4926 
4927     if (!run_lookup_entry(oa->run1, vcn, &lcn, NULL, NULL) ||
4928         lcn == SPARSE_LCN) {
4929         goto read_next_log_do_action;
4930     }
4931 
4932     /* Point to the Redo data and get its length. */
4933     data = Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
4934     dlen = le16_to_cpu(lrh->redo_len);
4935 
4936     /* Shorten length by any Lcns which were deleted. */
4937     saved_len = dlen;
4938 
4939     for (i = le16_to_cpu(lrh->lcns_follow); i; i--) {
4940         size_t j;
4941         u32 alen, voff;
4942 
4943         voff = le16_to_cpu(lrh->record_off) +
4944                le16_to_cpu(lrh->attr_off);
4945         voff += le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
4946 
4947         /* If the Vcn question is allocated, we can just get out. */
4948         j = le64_to_cpu(lrh->target_vcn) - le64_to_cpu(dp->vcn);
4949         if (dp->page_lcns[j + i - 1])
4950             break;
4951 
4952         if (!saved_len)
4953             saved_len = 1;
4954 
4955         /*
4956          * Calculate the allocated space left relative to the
4957          * log record Vcn, after removing this unallocated Vcn.
4958          */
4959         alen = (i - 1) << sbi->cluster_bits;
4960 
4961         /*
4962          * If the update described this log record goes beyond
4963          * the allocated space, then we will have to reduce the length.
4964          */
4965         if (voff >= alen)
4966             dlen = 0;
4967         else if (voff + dlen > alen)
4968             dlen = alen - voff;
4969     }
4970 
4971     /*
4972      * If the resulting dlen from above is now zero,
4973      * we can skip this log record.
4974      */
4975     if (!dlen && saved_len)
4976         goto read_next_log_do_action;
4977 
4978     t16 = le16_to_cpu(lrh->redo_op);
4979     if (can_skip_action(t16))
4980         goto read_next_log_do_action;
4981 
4982     /* Apply the Redo operation a common routine. */
4983     err = do_action(log, oe, lrh, t16, data, dlen, rec_len, &rec_lsn);
4984     if (err)
4985         goto out;
4986 
4987     /* Keep reading and looping back until end of file. */
4988 read_next_log_do_action:
4989     err = read_next_log_rec(log, lcb, &rec_lsn);
4990     if (!err && rec_lsn)
4991         goto do_action_next;
4992 
4993     lcb_put(lcb);
4994     lcb = NULL;
4995 
4996 do_undo_action:
4997     /* Scan Transaction Table. */
4998     tr = NULL;
4999 transaction_table_next:
5000     tr = enum_rstbl(trtbl, tr);
5001     if (!tr)
5002         goto undo_action_done;
5003 
5004     if (TransactionActive != tr->transact_state || !tr->undo_next_lsn) {
5005         free_rsttbl_idx(trtbl, PtrOffset(trtbl, tr));
5006         goto transaction_table_next;
5007     }
5008 
5009     log->transaction_id = PtrOffset(trtbl, tr);
5010     undo_next_lsn = le64_to_cpu(tr->undo_next_lsn);
5011 
5012     /*
5013      * We only have to do anything if the transaction has
5014      * something its undo_next_lsn field.
5015      */
5016     if (!undo_next_lsn)
5017         goto commit_undo;
5018 
5019     /* Read the first record to be undone by this transaction. */
5020     err = read_log_rec_lcb(log, undo_next_lsn, lcb_ctx_undo_next, &lcb);
5021     if (err)
5022         goto out;
5023 
5024     /*
5025      * Now loop to read all of our log records forwards,
5026      * until we hit the end of the file, cleaning up at the end.
5027      */
5028 undo_action_next:
5029 
5030     lrh = lcb->log_rec;
5031     frh = lcb->lrh;
5032     transact_id = le32_to_cpu(frh->transact_id);
5033     rec_len = le32_to_cpu(frh->client_data_len);
5034 
5035     if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
5036         err = -EINVAL;
5037         goto out;
5038     }
5039 
5040     if (lrh->undo_op == cpu_to_le16(Noop))
5041         goto read_next_log_undo_action;
5042 
5043     oe = Add2Ptr(oatbl, le16_to_cpu(lrh->target_attr));
5044     oa = oe->ptr;
5045 
5046     t16 = le16_to_cpu(lrh->lcns_follow);
5047     if (!t16)
5048         goto add_allocated_vcns;
5049 
5050     is_mapped = run_lookup_entry(oa->run1, le64_to_cpu(lrh->target_vcn),
5051                      &lcn, &clen, NULL);
5052 
5053     /*
5054      * If the mapping isn't already the table or the  mapping
5055      * corresponds to a hole the mapping, we need to make sure
5056      * there is no partial page already memory.
5057      */
5058     if (is_mapped && lcn != SPARSE_LCN && clen >= t16)
5059         goto add_allocated_vcns;
5060 
5061     vcn = le64_to_cpu(lrh->target_vcn);
5062     vcn &= ~(u64)(log->clst_per_page - 1);
5063 
5064 add_allocated_vcns:
5065     for (i = 0, vcn = le64_to_cpu(lrh->target_vcn),
5066         size = (vcn + 1) << sbi->cluster_bits;
5067          i < t16; i++, vcn += 1, size += sbi->cluster_size) {
5068         attr = oa->attr;
5069         if (!attr->non_res) {
5070             if (size > le32_to_cpu(attr->res.data_size))
5071                 attr->res.data_size = cpu_to_le32(size);
5072         } else {
5073             if (size > le64_to_cpu(attr->nres.data_size))
5074                 attr->nres.valid_size = attr->nres.data_size =
5075                     attr->nres.alloc_size =
5076                         cpu_to_le64(size);
5077         }
5078     }
5079 
5080     t16 = le16_to_cpu(lrh->undo_op);
5081     if (can_skip_action(t16))
5082         goto read_next_log_undo_action;
5083 
5084     /* Point to the Redo data and get its length. */
5085     data = Add2Ptr(lrh, le16_to_cpu(lrh->undo_off));
5086     dlen = le16_to_cpu(lrh->undo_len);
5087 
5088     /* It is time to apply the undo action. */
5089     err = do_action(log, oe, lrh, t16, data, dlen, rec_len, NULL);
5090 
5091 read_next_log_undo_action:
5092     /*
5093      * Keep reading and looping back until we have read the
5094      * last record for this transaction.
5095      */
5096     err = read_next_log_rec(log, lcb, &rec_lsn);
5097     if (err)
5098         goto out;
5099 
5100     if (rec_lsn)
5101         goto undo_action_next;
5102 
5103     lcb_put(lcb);
5104     lcb = NULL;
5105 
5106 commit_undo:
5107     free_rsttbl_idx(trtbl, log->transaction_id);
5108 
5109     log->transaction_id = 0;
5110 
5111     goto transaction_table_next;
5112 
5113 undo_action_done:
5114 
5115     ntfs_update_mftmirr(sbi, 0);
5116 
5117     sbi->flags &= ~NTFS_FLAGS_NEED_REPLAY;
5118 
5119 end_reply:
5120 
5121     err = 0;
5122     if (is_ro)
5123         goto out;
5124 
5125     rh = kzalloc(log->page_size, GFP_NOFS);
5126     if (!rh) {
5127         err = -ENOMEM;
5128         goto out;
5129     }
5130 
5131     rh->rhdr.sign = NTFS_RSTR_SIGNATURE;
5132     rh->rhdr.fix_off = cpu_to_le16(offsetof(struct RESTART_HDR, fixups));
5133     t16 = (log->page_size >> SECTOR_SHIFT) + 1;
5134     rh->rhdr.fix_num = cpu_to_le16(t16);
5135     rh->sys_page_size = cpu_to_le32(log->page_size);
5136     rh->page_size = cpu_to_le32(log->page_size);
5137 
5138     t16 = ALIGN(offsetof(struct RESTART_HDR, fixups) + sizeof(short) * t16,
5139             8);
5140     rh->ra_off = cpu_to_le16(t16);
5141     rh->minor_ver = cpu_to_le16(1); // 0x1A:
5142     rh->major_ver = cpu_to_le16(1); // 0x1C:
5143 
5144     ra2 = Add2Ptr(rh, t16);
5145     memcpy(ra2, ra, sizeof(struct RESTART_AREA));
5146 
5147     ra2->client_idx[0] = 0;
5148     ra2->client_idx[1] = LFS_NO_CLIENT_LE;
5149     ra2->flags = cpu_to_le16(2);
5150 
5151     le32_add_cpu(&ra2->open_log_count, 1);
5152 
5153     ntfs_fix_pre_write(&rh->rhdr, log->page_size);
5154 
5155     err = ntfs_sb_write_run(sbi, &ni->file.run, 0, rh, log->page_size, 0);
5156     if (!err)
5157         err = ntfs_sb_write_run(sbi, &log->ni->file.run, log->page_size,
5158                     rh, log->page_size, 0);
5159 
5160     kfree(rh);
5161     if (err)
5162         goto out;
5163 
5164 out:
5165     kfree(rst);
5166     if (lcb)
5167         lcb_put(lcb);
5168 
5169     /*
5170      * Scan the Open Attribute Table to close all of
5171      * the open attributes.
5172      */
5173     oe = NULL;
5174     while ((oe = enum_rstbl(oatbl, oe))) {
5175         rno = ino_get(&oe->ref);
5176 
5177         if (oe->is_attr_name == 1) {
5178             kfree(oe->ptr);
5179             oe->ptr = NULL;
5180             continue;
5181         }
5182 
5183         if (oe->is_attr_name)
5184             continue;
5185 
5186         oa = oe->ptr;
5187         if (!oa)
5188             continue;
5189 
5190         run_close(&oa->run0);
5191         kfree(oa->attr);
5192         if (oa->ni)
5193             iput(&oa->ni->vfs_inode);
5194         kfree(oa);
5195     }
5196 
5197     kfree(trtbl);
5198     kfree(oatbl);
5199     kfree(dptbl);
5200     kfree(attr_names);
5201     kfree(rst_info.r_page);
5202 
5203     kfree(ra);
5204     kfree(log->one_page_buf);
5205 
5206     if (err)
5207         sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
5208 
5209     if (err == -EROFS)
5210         err = 0;
5211     else if (log->set_dirty)
5212         ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
5213 
5214     kfree(log);
5215 
5216     return err;
5217 }