fsck/mount.c

   1 /**
   2  * mount.c
   3  *
   4  * Copyright (c) 2013 Samsung Electronics Co., Ltd.
   5  *             http://www.samsung.com/
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  */
  11 #include "fsck.h"
  12 #include "node.h"
  13 #include "xattr.h"
  14 #include "quota.h"
  15 #include <locale.h>
  16 #include <stdbool.h>
  17 #include <time.h>
  18 #ifdef HAVE_LINUX_POSIX_ACL_H
  19 #include <linux/posix_acl.h>
  20 #endif
  21 #ifdef HAVE_SYS_ACL_H
  22 #include <sys/acl.h>
  23 #endif
  24 #ifdef HAVE_UUID_UUID_H
  25 #include <uuid/uuid.h>
  26 #endif
  27
  28 #ifndef ACL_UNDEFINED_TAG
  29 #define ACL_UNDEFINED_TAG       (0x00)
  30 #define ACL_USER_OBJ            (0x01)
  31 #define ACL_USER                (0x02)
  32 #define ACL_GROUP_OBJ           (0x04)
  33 #define ACL_GROUP               (0x08)
  34 #define ACL_MASK                (0x10)
  35 #define ACL_OTHER               (0x20)
  36 #endif
  37
  38 #ifdef HAVE_LINUX_BLKZONED_H
  39
  40 static int get_device_idx(struct f2fs_sb_info *sbi, uint32_t segno)
  41 {
  42         block_t seg_start_blkaddr;
  43         int i;
  44
  45         seg_start_blkaddr = SM_I(sbi)->main_blkaddr +
  46                                 segno * DEFAULT_BLOCKS_PER_SEGMENT;
  47         for (i = 0; i < c.ndevs; i++)
  48                 if (c.devices[i].start_blkaddr <= seg_start_blkaddr &&
  49                         c.devices[i].end_blkaddr > seg_start_blkaddr)
  50                         return i;
  51         return 0;
  52 }
  53
  54 static int get_zone_idx_from_dev(struct f2fs_sb_info *sbi,
  55                                         uint32_t segno, uint32_t dev_idx)
  56 {
  57         block_t seg_start_blkaddr = START_BLOCK(sbi, segno);
  58
  59         return (seg_start_blkaddr - c.devices[dev_idx].start_blkaddr) >>
  60                         log_base_2(sbi->segs_per_sec * sbi->blocks_per_seg);
  61 }
  62
  63 bool is_usable_seg(struct f2fs_sb_info *sbi, unsigned int segno)
  64 {
  65         unsigned int secno = segno / sbi->segs_per_sec;
  66         block_t seg_start = START_BLOCK(sbi, segno);
  67         block_t blocks_per_sec = sbi->blocks_per_seg * sbi->segs_per_sec;
  68         unsigned int dev_idx = get_device_idx(sbi, segno);
  69         unsigned int zone_idx = get_zone_idx_from_dev(sbi, segno, dev_idx);
  70         unsigned int sec_off = SM_I(sbi)->main_blkaddr >>
  71                                                 log_base_2(blocks_per_sec);
  72
  73         if (zone_idx < c.devices[dev_idx].nr_rnd_zones)
  74                 return true;
  75
  76         if (c.devices[dev_idx].zoned_model != F2FS_ZONED_HM)
  77                 return true;
  78
  79         return seg_start < ((sec_off + secno) * blocks_per_sec) +
  80                                 c.devices[dev_idx].zone_cap_blocks[zone_idx];
  81 }
  82
  83 unsigned int get_usable_seg_count(struct f2fs_sb_info *sbi)
  84 {
  85         unsigned int i, usable_seg_count = 0;
  86
  87         for (i = 0; i < MAIN_SEGS(sbi); i++)
  88                 if (is_usable_seg(sbi, i))
  89                         usable_seg_count++;
  90
  91         return usable_seg_count;
  92 }
  93
  94 #else
  95
  96 bool is_usable_seg(struct f2fs_sb_info *UNUSED(sbi), unsigned int UNUSED(segno))
  97 {
  98         return true;
  99 }
 100
 101 unsigned int get_usable_seg_count(struct f2fs_sb_info *sbi)
 102 {
 103         return MAIN_SEGS(sbi);
 104 }
 105
 106 #endif
 107
 108 u32 get_free_segments(struct f2fs_sb_info *sbi)
 109 {
 110         u32 i, free_segs = 0;
 111
 112         for (i = 0; i < MAIN_SEGS(sbi); i++) {
 113                 struct seg_entry *se = get_seg_entry(sbi, i);
 114
 115                 if (se->valid_blocks == 0x0 && !IS_CUR_SEGNO(sbi, i) &&
 116                                                         is_usable_seg(sbi, i))
 117                         free_segs++;
 118         }
 119         return free_segs;
 120 }
 121
 122 void update_free_segments(struct f2fs_sb_info *sbi)
 123 {
 124         char *progress = "-*|*-";
 125         static int i = 0;
 126
 127         if (c.dbg_lv)
 128                 return;
 129
 130         MSG(0, "\r [ %c ] Free segments: 0x%x", progress[i % 5], get_free_segments(sbi));
 131         fflush(stdout);
 132         i++;
 133 }
 134
 135 #if defined(HAVE_LINUX_POSIX_ACL_H) || defined(HAVE_SYS_ACL_H)
 136 static void print_acl(const u8 *value, int size)
 137 {
 138         const struct f2fs_acl_header *hdr = (struct f2fs_acl_header *)value;
 139         const struct f2fs_acl_entry *entry = (struct f2fs_acl_entry *)(hdr + 1);
 140         const u8 *end = value + size;
 141         int i, count;
 142
 143         if (hdr->a_version != cpu_to_le32(F2FS_ACL_VERSION)) {
 144                 MSG(0, "Invalid ACL version [0x%x : 0x%x]\n",
 145                                 le32_to_cpu(hdr->a_version), F2FS_ACL_VERSION);
 146                 return;
 147         }
 148
 149         count = f2fs_acl_count(size);
 150         if (count <= 0) {
 151                 MSG(0, "Invalid ACL value size %d\n", size);
 152                 return;
 153         }
 154
 155         for (i = 0; i < count; i++) {
 156                 if ((u8 *)entry > end) {
 157                         MSG(0, "Invalid ACL entries count %d\n", count);
 158                         return;
 159                 }
 160
 161                 switch (le16_to_cpu(entry->e_tag)) {
 162                 case ACL_USER_OBJ:
 163                 case ACL_GROUP_OBJ:
 164                 case ACL_MASK:
 165                 case ACL_OTHER:
 166                         MSG(0, "tag:0x%x perm:0x%x\n",
 167                                         le16_to_cpu(entry->e_tag),
 168                                         le16_to_cpu(entry->e_perm));
 169                         entry = (struct f2fs_acl_entry *)((char *)entry +
 170                                         sizeof(struct f2fs_acl_entry_short));
 171                         break;
 172                 case ACL_USER:
 173                         MSG(0, "tag:0x%x perm:0x%x uid:%u\n",
 174                                         le16_to_cpu(entry->e_tag),
 175                                         le16_to_cpu(entry->e_perm),
 176                                         le32_to_cpu(entry->e_id));
 177                         entry = (struct f2fs_acl_entry *)((char *)entry +
 178                                         sizeof(struct f2fs_acl_entry));
 179                         break;
 180                 case ACL_GROUP:
 181                         MSG(0, "tag:0x%x perm:0x%x gid:%u\n",
 182                                         le16_to_cpu(entry->e_tag),
 183                                         le16_to_cpu(entry->e_perm),
 184                                         le32_to_cpu(entry->e_id));
 185                         entry = (struct f2fs_acl_entry *)((char *)entry +
 186                                         sizeof(struct f2fs_acl_entry));
 187                         break;
 188                 default:
 189                         MSG(0, "Unknown ACL tag 0x%x\n",
 190                                         le16_to_cpu(entry->e_tag));
 191                         return;
 192                 }
 193         }
 194 }
 195 #endif /* HAVE_LINUX_POSIX_ACL_H || HAVE_SYS_ACL_H */
 196
 197 static void print_xattr_entry(const struct f2fs_xattr_entry *ent)
 198 {
 199         const u8 *value = (const u8 *)&ent->e_name[ent->e_name_len];
 200         const int size = le16_to_cpu(ent->e_value_size);
 201         const struct fscrypt_context *ctx;
 202         int i;
 203
 204         MSG(0, "\nxattr: e_name_index:%d e_name:", ent->e_name_index);
 205         for (i = 0; i < ent->e_name_len; i++)
 206                 MSG(0, "%c", ent->e_name[i]);
 207         MSG(0, " e_name_len:%d e_value_size:%d e_value:\n",
 208                         ent->e_name_len, size);
 209
 210         switch (ent->e_name_index) {
 211 #if defined(HAVE_LINUX_POSIX_ACL_H) || defined(HAVE_SYS_ACL_H)
 212         case F2FS_XATTR_INDEX_POSIX_ACL_ACCESS:
 213         case F2FS_XATTR_INDEX_POSIX_ACL_DEFAULT:
 214                 print_acl(value, size);
 215                 return;
 216 #endif
 217         case F2FS_XATTR_INDEX_ENCRYPTION:
 218                 ctx = (const struct fscrypt_context *)value;
 219                 if (size != sizeof(*ctx) ||
 220                     ctx->format != FS_ENCRYPTION_CONTEXT_FORMAT_V1)
 221                         break;
 222                 MSG(0, "format: %d\n", ctx->format);
 223                 MSG(0, "contents_encryption_mode: 0x%x\n", ctx->contents_encryption_mode);
 224                 MSG(0, "filenames_encryption_mode: 0x%x\n", ctx->filenames_encryption_mode);
 225                 MSG(0, "flags: 0x%x\n", ctx->flags);
 226                 MSG(0, "master_key_descriptor: ");
 227                 for (i = 0; i < FS_KEY_DESCRIPTOR_SIZE; i++)
 228                         MSG(0, "%02X", ctx->master_key_descriptor[i]);
 229                 MSG(0, "\nnonce: ");
 230                 for (i = 0; i < FS_KEY_DERIVATION_NONCE_SIZE; i++)
 231                         MSG(0, "%02X", ctx->nonce[i]);
 232                 MSG(0, "\n");
 233                 return;
 234         }
 235         for (i = 0; i < size; i++)
 236                 MSG(0, "%02X", value[i]);
 237         MSG(0, "\n");
 238 }
 239
 240 void print_inode_info(struct f2fs_sb_info *sbi,
 241                         struct f2fs_node *node, int name)
 242 {
 243         struct f2fs_inode *inode = &node->i;
 244         void *xattr_addr;
 245         struct f2fs_xattr_entry *ent;
 246         char en[F2FS_PRINT_NAMELEN];
 247         unsigned int i = 0;
 248         u32 namelen = le32_to_cpu(inode->i_namelen);
 249         int enc_name = file_enc_name(inode);
 250         int ofs = get_extra_isize(node);
 251
 252         pretty_print_filename(inode->i_name, namelen, en, enc_name);
 253         if (name && en[0]) {
 254                 MSG(0, " - File name         : %s%s\n", en,
 255                                 enc_name ? " <encrypted>" : "");
 256                 setlocale(LC_ALL, "");
 257                 MSG(0, " - File size         : %'" PRIu64 " (bytes)\n",
 258                                 le64_to_cpu(inode->i_size));
 259                 return;
 260         }
 261
 262         DISP_u32(inode, i_mode);
 263         DISP_u32(inode, i_advise);
 264         DISP_u32(inode, i_uid);
 265         DISP_u32(inode, i_gid);
 266         DISP_u32(inode, i_links);
 267         DISP_u64(inode, i_size);
 268         DISP_u64(inode, i_blocks);
 269
 270         DISP_u64(inode, i_atime);
 271         DISP_u32(inode, i_atime_nsec);
 272         DISP_u64(inode, i_ctime);
 273         DISP_u32(inode, i_ctime_nsec);
 274         DISP_u64(inode, i_mtime);
 275         DISP_u32(inode, i_mtime_nsec);
 276
 277         DISP_u32(inode, i_generation);
 278         DISP_u32(inode, i_current_depth);
 279         DISP_u32(inode, i_xattr_nid);
 280         DISP_u32(inode, i_flags);
 281         DISP_u32(inode, i_inline);
 282         DISP_u32(inode, i_pino);
 283         DISP_u32(inode, i_dir_level);
 284
 285         if (en[0]) {
 286                 DISP_u32(inode, i_namelen);
 287                 printf("%-30s\t\t[%s]\n", "i_name", en);
 288         }
 289
 290         printf("i_ext: fofs:%x blkaddr:%x len:%x\n",
 291                         le32_to_cpu(inode->i_ext.fofs),
 292                         le32_to_cpu(inode->i_ext.blk_addr),
 293                         le32_to_cpu(inode->i_ext.len));
 294
 295         if (c.feature & cpu_to_le32(F2FS_FEATURE_EXTRA_ATTR)) {
 296                 DISP_u16(inode, i_extra_isize);
 297                 if (c.feature & cpu_to_le32(F2FS_FEATURE_FLEXIBLE_INLINE_XATTR))
 298                         DISP_u16(inode, i_inline_xattr_size);
 299                 if (c.feature & cpu_to_le32(F2FS_FEATURE_PRJQUOTA))
 300                         DISP_u32(inode, i_projid);
 301                 if (c.feature & cpu_to_le32(F2FS_FEATURE_INODE_CHKSUM))
 302                         DISP_u32(inode, i_inode_checksum);
 303                 if (c.feature & cpu_to_le32(F2FS_FEATURE_INODE_CRTIME)) {
 304                         DISP_u64(inode, i_crtime);
 305                         DISP_u32(inode, i_crtime_nsec);
 306                 }
 307                 if (c.feature & cpu_to_le32(F2FS_FEATURE_COMPRESSION)) {
 308                         DISP_u64(inode, i_compr_blocks);
 309                         DISP_u8(inode, i_compress_algorithm);
 310                         DISP_u8(inode, i_log_cluster_size);
 311                         DISP_u16(inode, i_compress_flag);
 312                 }
 313         }
 314
 315         for (i = 0; i < ADDRS_PER_INODE(inode); i++) {
 316                 block_t blkaddr;
 317                 char *flag = "";
 318
 319                 if (i + ofs >= DEF_ADDRS_PER_INODE)
 320                         break;
 321
 322                 blkaddr = le32_to_cpu(inode->i_addr[i + ofs]);
 323
 324                 if (blkaddr == 0x0)
 325                         continue;
 326                 if (blkaddr == COMPRESS_ADDR)
 327                         flag = "cluster flag";
 328                 else if (blkaddr == NEW_ADDR)
 329                         flag = "reserved flag";
 330                 printf("i_addr[0x%x] %-16s\t\t[0x%8x : %u]\n", i + ofs, flag,
 331                                 blkaddr, blkaddr);
 332         }
 333
 334         DISP_u32(inode, i_nid[0]);      /* direct */
 335         DISP_u32(inode, i_nid[1]);      /* direct */
 336         DISP_u32(inode, i_nid[2]);      /* indirect */
 337         DISP_u32(inode, i_nid[3]);      /* indirect */
 338         DISP_u32(inode, i_nid[4]);      /* double indirect */
 339
 340         xattr_addr = read_all_xattrs(sbi, node);
 341         if (xattr_addr) {
 342                 list_for_each_xattr(ent, xattr_addr) {
 343                         print_xattr_entry(ent);
 344                 }
 345                 free(xattr_addr);
 346         }
 347
 348         printf("\n");
 349 }
 350
 351 void print_node_info(struct f2fs_sb_info *sbi,
 352                         struct f2fs_node *node_block, int verbose)
 353 {
 354         nid_t ino = le32_to_cpu(node_block->footer.ino);
 355         nid_t nid = le32_to_cpu(node_block->footer.nid);
 356         /* Is this inode? */
 357         if (ino == nid) {
 358                 DBG(verbose, "Node ID [0x%x:%u] is inode\n", nid, nid);
 359                 print_inode_info(sbi, node_block, verbose);
 360         } else {
 361                 int i;
 362                 u32 *dump_blk = (u32 *)node_block;
 363                 DBG(verbose,
 364                         "Node ID [0x%x:%u] is direct node or indirect node.\n",
 365                                                                 nid, nid);
 366                 for (i = 0; i < DEF_ADDRS_PER_BLOCK; i++)
 367                         MSG(verbose, "[%d]\t\t\t[0x%8x : %d]\n",
 368                                                 i, dump_blk[i], dump_blk[i]);
 369         }
 370 }
 371
 372 void print_extention_list(struct f2fs_super_block *sb, int cold)
 373 {
 374         int start, end, i;
 375
 376         if (cold) {
 377                 DISP_u32(sb, extension_count);
 378
 379                 start = 0;
 380                 end = le32_to_cpu(sb->extension_count);
 381         } else {
 382                 DISP_u8(sb, hot_ext_count);
 383
 384                 start = le32_to_cpu(sb->extension_count);
 385                 end = start + sb->hot_ext_count;
 386         }
 387
 388         printf("%s file extentsions\n", cold ? "cold" : "hot");
 389
 390         for (i = start; i < end; i++) {
 391                 if (c.layout) {
 392                         printf("%-30s %-8.8s\n", "extension_list",
 393                                                 sb->extension_list[i]);
 394                 } else {
 395                         if (i % 4 == 0)
 396                                 printf("%-30s\t\t[", "");
 397
 398                         printf("%-8.8s", sb->extension_list[i]);
 399
 400                         if (i % 4 == 4 - 1 || i == end - start - 1)
 401                                 printf("]\n");
 402                 }
 403         }
 404 }
 405
 406 static void DISP_label(const char *name)
 407 {
 408         char buffer[MAX_VOLUME_NAME];
 409
 410         utf16_to_utf8(buffer, name, MAX_VOLUME_NAME, MAX_VOLUME_NAME);
 411         if (c.layout)
 412                 printf("%-30s %s\n", "Filesystem volume name:", buffer);
 413         else
 414                 printf("%-30s" "\t\t[%s]\n", "volum_name", buffer);
 415 }
 416
 417 void print_sb_debug_info(struct f2fs_super_block *sb);
 418 void print_raw_sb_info(struct f2fs_super_block *sb)
 419 {
 420 #ifdef HAVE_LIBUUID
 421         char uuid[40];
 422         char encrypt_pw_salt[40];
 423 #endif
 424
 425         if (c.layout)
 426                 goto printout;
 427         if (!c.dbg_lv)
 428                 return;
 429
 430         printf("\n");
 431         printf("+--------------------------------------------------------+\n");
 432         printf("| Super block                                            |\n");
 433         printf("+--------------------------------------------------------+\n");
 434 printout:
 435         DISP_u32(sb, magic);
 436         DISP_u32(sb, major_ver);
 437
 438         DISP_u32(sb, minor_ver);
 439         DISP_u32(sb, log_sectorsize);
 440         DISP_u32(sb, log_sectors_per_block);
 441
 442         DISP_u32(sb, log_blocksize);
 443         DISP_u32(sb, log_blocks_per_seg);
 444         DISP_u32(sb, segs_per_sec);
 445         DISP_u32(sb, secs_per_zone);
 446         DISP_u32(sb, checksum_offset);
 447         DISP_u64(sb, block_count);
 448
 449         DISP_u32(sb, section_count);
 450         DISP_u32(sb, segment_count);
 451         DISP_u32(sb, segment_count_ckpt);
 452         DISP_u32(sb, segment_count_sit);
 453         DISP_u32(sb, segment_count_nat);
 454
 455         DISP_u32(sb, segment_count_ssa);
 456         DISP_u32(sb, segment_count_main);
 457         DISP_u32(sb, segment0_blkaddr);
 458
 459         DISP_u32(sb, cp_blkaddr);
 460         DISP_u32(sb, sit_blkaddr);
 461         DISP_u32(sb, nat_blkaddr);
 462         DISP_u32(sb, ssa_blkaddr);
 463         DISP_u32(sb, main_blkaddr);
 464
 465         DISP_u32(sb, root_ino);
 466         DISP_u32(sb, node_ino);
 467         DISP_u32(sb, meta_ino);
 468
 469 #ifdef HAVE_LIBUUID
 470         uuid_unparse(sb->uuid, uuid);
 471         DISP_raw_str("%-.36s", uuid);
 472 #endif
 473
 474         DISP_label((const char *)sb->volume_name);
 475
 476         print_extention_list(sb, 1);
 477         print_extention_list(sb, 0);
 478
 479         DISP_u32(sb, cp_payload);
 480
 481         DISP_str("%-.252s", sb, version);
 482         DISP_str("%-.252s", sb, init_version);
 483
 484         DISP_u32(sb, feature);
 485         DISP_u8(sb, encryption_level);
 486
 487 #ifdef HAVE_LIBUUID
 488         uuid_unparse(sb->encrypt_pw_salt, encrypt_pw_salt);
 489         DISP_raw_str("%-.36s", encrypt_pw_salt);
 490 #endif
 491
 492         DISP_u32(sb, qf_ino[USRQUOTA]);
 493         DISP_u32(sb, qf_ino[GRPQUOTA]);
 494         DISP_u32(sb, qf_ino[PRJQUOTA]);
 495
 496         DISP_u16(sb, s_encoding);
 497         DISP_u32(sb, crc);
 498
 499         print_sb_debug_info(sb);
 500
 501         printf("\n");
 502 }
 503
 504 void print_ckpt_info(struct f2fs_sb_info *sbi)
 505 {
 506         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
 507
 508         if (c.layout)
 509                 goto printout;
 510         if (!c.dbg_lv)
 511                 return;
 512
 513         printf("\n");
 514         printf("+--------------------------------------------------------+\n");
 515         printf("| Checkpoint                                             |\n");
 516         printf("+--------------------------------------------------------+\n");
 517 printout:
 518         DISP_u64(cp, checkpoint_ver);
 519         DISP_u64(cp, user_block_count);
 520         DISP_u64(cp, valid_block_count);
 521         DISP_u32(cp, rsvd_segment_count);
 522         DISP_u32(cp, overprov_segment_count);
 523         DISP_u32(cp, free_segment_count);
 524
 525         DISP_u32(cp, alloc_type[CURSEG_HOT_NODE]);
 526         DISP_u32(cp, alloc_type[CURSEG_WARM_NODE]);
 527         DISP_u32(cp, alloc_type[CURSEG_COLD_NODE]);
 528         DISP_u32(cp, cur_node_segno[0]);
 529         DISP_u32(cp, cur_node_segno[1]);
 530         DISP_u32(cp, cur_node_segno[2]);
 531
 532         DISP_u32(cp, cur_node_blkoff[0]);
 533         DISP_u32(cp, cur_node_blkoff[1]);
 534         DISP_u32(cp, cur_node_blkoff[2]);
 535
 536
 537         DISP_u32(cp, alloc_type[CURSEG_HOT_DATA]);
 538         DISP_u32(cp, alloc_type[CURSEG_WARM_DATA]);
 539         DISP_u32(cp, alloc_type[CURSEG_COLD_DATA]);
 540         DISP_u32(cp, cur_data_segno[0]);
 541         DISP_u32(cp, cur_data_segno[1]);
 542         DISP_u32(cp, cur_data_segno[2]);
 543
 544         DISP_u32(cp, cur_data_blkoff[0]);
 545         DISP_u32(cp, cur_data_blkoff[1]);
 546         DISP_u32(cp, cur_data_blkoff[2]);
 547
 548         DISP_u32(cp, ckpt_flags);
 549         DISP_u32(cp, cp_pack_total_block_count);
 550         DISP_u32(cp, cp_pack_start_sum);
 551         DISP_u32(cp, valid_node_count);
 552         DISP_u32(cp, valid_inode_count);
 553         DISP_u32(cp, next_free_nid);
 554         DISP_u32(cp, sit_ver_bitmap_bytesize);
 555         DISP_u32(cp, nat_ver_bitmap_bytesize);
 556         DISP_u32(cp, checksum_offset);
 557         DISP_u64(cp, elapsed_time);
 558
 559         DISP_u32(cp, sit_nat_version_bitmap[0]);
 560         printf("\n\n");
 561 }
 562
 563 void print_cp_state(u32 flag)
 564 {
 565         if (c.show_file_map)
 566                 return;
 567
 568         MSG(0, "Info: checkpoint state = %x : ", flag);
 569         if (flag & CP_QUOTA_NEED_FSCK_FLAG)
 570                 MSG(0, "%s", " quota_need_fsck");
 571         if (flag & CP_LARGE_NAT_BITMAP_FLAG)
 572                 MSG(0, "%s", " large_nat_bitmap");
 573         if (flag & CP_NOCRC_RECOVERY_FLAG)
 574                 MSG(0, "%s", " allow_nocrc");
 575         if (flag & CP_TRIMMED_FLAG)
 576                 MSG(0, "%s", " trimmed");
 577         if (flag & CP_NAT_BITS_FLAG)
 578                 MSG(0, "%s", " nat_bits");
 579         if (flag & CP_CRC_RECOVERY_FLAG)
 580                 MSG(0, "%s", " crc");
 581         if (flag & CP_FASTBOOT_FLAG)
 582                 MSG(0, "%s", " fastboot");
 583         if (flag & CP_FSCK_FLAG)
 584                 MSG(0, "%s", " fsck");
 585         if (flag & CP_ERROR_FLAG)
 586                 MSG(0, "%s", " error");
 587         if (flag & CP_COMPACT_SUM_FLAG)
 588                 MSG(0, "%s", " compacted_summary");
 589         if (flag & CP_ORPHAN_PRESENT_FLAG)
 590                 MSG(0, "%s", " orphan_inodes");
 591         if (flag & CP_DISABLED_FLAG)
 592                 MSG(0, "%s", " disabled");
 593         if (flag & CP_RESIZEFS_FLAG)
 594                 MSG(0, "%s", " resizefs");
 595         if (flag & CP_UMOUNT_FLAG)
 596                 MSG(0, "%s", " unmount");
 597         else
 598                 MSG(0, "%s", " sudden-power-off");
 599         MSG(0, "\n");
 600 }
 601
 602 void print_sb_state(struct f2fs_super_block *sb)
 603 {
 604         __le32 f = sb->feature;
 605         int i;
 606
 607         MSG(0, "Info: superblock features = %x : ", f);
 608         if (f & cpu_to_le32(F2FS_FEATURE_ENCRYPT)) {
 609                 MSG(0, "%s", " encrypt");
 610         }
 611         if (f & cpu_to_le32(F2FS_FEATURE_VERITY)) {
 612                 MSG(0, "%s", " verity");
 613         }
 614         if (f & cpu_to_le32(F2FS_FEATURE_BLKZONED)) {
 615                 MSG(0, "%s", " blkzoned");
 616         }
 617         if (f & cpu_to_le32(F2FS_FEATURE_EXTRA_ATTR)) {
 618                 MSG(0, "%s", " extra_attr");
 619         }
 620         if (f & cpu_to_le32(F2FS_FEATURE_PRJQUOTA)) {
 621                 MSG(0, "%s", " project_quota");
 622         }
 623         if (f & cpu_to_le32(F2FS_FEATURE_INODE_CHKSUM)) {
 624                 MSG(0, "%s", " inode_checksum");
 625         }
 626         if (f & cpu_to_le32(F2FS_FEATURE_FLEXIBLE_INLINE_XATTR)) {
 627                 MSG(0, "%s", " flexible_inline_xattr");
 628         }
 629         if (f & cpu_to_le32(F2FS_FEATURE_QUOTA_INO)) {
 630                 MSG(0, "%s", " quota_ino");
 631         }
 632         if (f & cpu_to_le32(F2FS_FEATURE_INODE_CRTIME)) {
 633                 MSG(0, "%s", " inode_crtime");
 634         }
 635         if (f & cpu_to_le32(F2FS_FEATURE_LOST_FOUND)) {
 636                 MSG(0, "%s", " lost_found");
 637         }
 638         if (f & cpu_to_le32(F2FS_FEATURE_SB_CHKSUM)) {
 639                 MSG(0, "%s", " sb_checksum");
 640         }
 641         if (f & cpu_to_le32(F2FS_FEATURE_CASEFOLD)) {
 642                 MSG(0, "%s", " casefold");
 643         }
 644         if (f & cpu_to_le32(F2FS_FEATURE_COMPRESSION)) {
 645                 MSG(0, "%s", " compression");
 646         }
 647         if (f & cpu_to_le32(F2FS_FEATURE_RO)) {
 648                 MSG(0, "%s", " ro");
 649         }
 650         MSG(0, "\n");
 651         MSG(0, "Info: superblock encrypt level = %d, salt = ",
 652                                         sb->encryption_level);
 653         for (i = 0; i < 16; i++)
 654                 MSG(0, "%02x", sb->encrypt_pw_salt[i]);
 655         MSG(0, "\n");
 656 }
 657
 658 static char *stop_reason_str[] = {
 659         [STOP_CP_REASON_SHUTDOWN]               = "shutdown",
 660         [STOP_CP_REASON_FAULT_INJECT]           = "fault_inject",
 661         [STOP_CP_REASON_META_PAGE]              = "meta_page",
 662         [STOP_CP_REASON_WRITE_FAIL]             = "write_fail",
 663         [STOP_CP_REASON_CORRUPTED_SUMMARY]      = "corrupted_summary",
 664         [STOP_CP_REASON_UPDATE_INODE]           = "update_inode",
 665         [STOP_CP_REASON_FLUSH_FAIL]             = "flush_fail",
 666 };
 667
 668 void print_sb_stop_reason(struct f2fs_super_block *sb)
 669 {
 670         u8 *reason = sb->s_stop_reason;
 671         int i;
 672
 673         if (!c.force_stop)
 674                 return;
 675
 676         MSG(0, "Info: checkpoint stop reason: ");
 677
 678         for (i = 0; i < STOP_CP_REASON_MAX; i++) {
 679                 if (reason[i])
 680                         MSG(0, "%s(%d) ", stop_reason_str[i], reason[i]);
 681         }
 682
 683         MSG(0, "\n");
 684 }
 685
 686 static char *errors_str[] = {
 687         [ERROR_CORRUPTED_CLUSTER]               = "corrupted_cluster",
 688         [ERROR_FAIL_DECOMPRESSION]              = "fail_decompression",
 689         [ERROR_INVALID_BLKADDR]                 = "invalid_blkaddr",
 690         [ERROR_CORRUPTED_DIRENT]                = "corrupted_dirent",
 691         [ERROR_CORRUPTED_INODE]                 = "corrupted_inode",
 692         [ERROR_INCONSISTENT_SUMMARY]            = "inconsistent_summary",
 693         [ERROR_INCONSISTENT_FOOTER]             = "inconsistent_footer",
 694         [ERROR_INCONSISTENT_SUM_TYPE]           = "inconsistent_sum_type",
 695         [ERROR_CORRUPTED_JOURNAL]               = "corrupted_journal",
 696         [ERROR_INCONSISTENT_NODE_COUNT]         = "inconsistent_node_count",
 697         [ERROR_INCONSISTENT_BLOCK_COUNT]        = "inconsistent_block_count",
 698         [ERROR_INVALID_CURSEG]                  = "invalid_curseg",
 699         [ERROR_INCONSISTENT_SIT]                = "inconsistent_sit",
 700         [ERROR_CORRUPTED_VERITY_XATTR]          = "corrupted_verity_xattr",
 701         [ERROR_CORRUPTED_XATTR]                 = "corrupted_xattr",
 702 };
 703
 704 void print_sb_errors(struct f2fs_super_block *sb)
 705 {
 706         u8 *errors = sb->s_errors;
 707         int i;
 708
 709         if (!c.fs_errors)
 710                 return;
 711
 712         MSG(0, "Info: fs errors: ");
 713
 714         for (i = 0; i < ERROR_MAX; i++) {
 715                 if (test_bit_le(i, errors))
 716                         MSG(0, "%s ",  errors_str[i]);
 717         }
 718
 719         MSG(0, "\n");
 720 }
 721
 722 void print_sb_debug_info(struct f2fs_super_block *sb)
 723 {
 724         u8 *reason = sb->s_stop_reason;
 725         u8 *errors = sb->s_errors;
 726         int i;
 727
 728         for (i = 0; i < STOP_CP_REASON_MAX; i++) {
 729                 if (!reason[i])
 730                         continue;
 731                 if (c.layout)
 732                         printf("%-30s %s(%s, %d)\n", "", "stop_reason",
 733                                 stop_reason_str[i], reason[i]);
 734                 else
 735                         printf("%-30s\t\t[%-20s : %u]\n", "",
 736                                 stop_reason_str[i], reason[i]);
 737         }
 738
 739         for (i = 0; i < ERROR_MAX; i++) {
 740                 if (!test_bit_le(i, errors))
 741                         continue;
 742                 if (c.layout)
 743                         printf("%-30s %s(%s)\n", "", "errors", errors_str[i]);
 744                 else
 745                         printf("%-30s\t\t[%-20s]\n", "", errors_str[i]);
 746         }
 747 }
 748
 749 bool f2fs_is_valid_blkaddr(struct f2fs_sb_info *sbi,
 750                                         block_t blkaddr, int type)
 751 {
 752         switch (type) {
 753         case META_NAT:
 754                 break;
 755         case META_SIT:
 756                 if (blkaddr >= SIT_BLK_CNT(sbi))
 757                         return 0;
 758                 break;
 759         case META_SSA:
 760                 if (blkaddr >= MAIN_BLKADDR(sbi) ||
 761                         blkaddr < SM_I(sbi)->ssa_blkaddr)
 762                         return 0;
 763                 break;
 764         case META_CP:
 765                 if (blkaddr >= SIT_I(sbi)->sit_base_addr ||
 766                         blkaddr < __start_cp_addr(sbi))
 767                         return 0;
 768                 break;
 769         case META_POR:
 770                 if (blkaddr >= MAX_BLKADDR(sbi) ||
 771                         blkaddr < MAIN_BLKADDR(sbi))
 772                         return 0;
 773                 break;
 774         default:
 775                 ASSERT(0);
 776         }
 777
 778         return 1;
 779 }
 780
 781 static inline block_t current_sit_addr(struct f2fs_sb_info *sbi,
 782                                                 unsigned int start);
 783
 784 /*
 785  * Readahead CP/NAT/SIT/SSA pages
 786  */
 787 int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
 788                                                         int type)
 789 {
 790         block_t blkno = start;
 791         block_t blkaddr, start_blk = 0, len = 0;
 792
 793         for (; nrpages-- > 0; blkno++) {
 794
 795                 if (!f2fs_is_valid_blkaddr(sbi, blkno, type))
 796                         goto out;
 797
 798                 switch (type) {
 799                 case META_NAT:
 800                         if (blkno >= NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))
 801                                 blkno = 0;
 802                         /* get nat block addr */
 803                         blkaddr = current_nat_addr(sbi,
 804                                         blkno * NAT_ENTRY_PER_BLOCK, NULL);
 805                         break;
 806                 case META_SIT:
 807                         /* get sit block addr */
 808                         blkaddr = current_sit_addr(sbi,
 809                                         blkno * SIT_ENTRY_PER_BLOCK);
 810                         break;
 811                 case META_SSA:
 812                 case META_CP:
 813                 case META_POR:
 814                         blkaddr = blkno;
 815                         break;
 816                 default:
 817                         ASSERT(0);
 818                 }
 819
 820                 if (!len) {
 821                         start_blk = blkaddr;
 822                         len = 1;
 823                 } else if (start_blk + len == blkaddr) {
 824                         len++;
 825                 } else {
 826                         dev_readahead(start_blk << F2FS_BLKSIZE_BITS,
 827                                                 len << F2FS_BLKSIZE_BITS);
 828                 }
 829         }
 830 out:
 831         if (len)
 832                 dev_readahead(start_blk << F2FS_BLKSIZE_BITS,
 833                                         len << F2FS_BLKSIZE_BITS);
 834         return blkno - start;
 835 }
 836
 837 void update_superblock(struct f2fs_super_block *sb, int sb_mask)
 838 {
 839         int addr, ret;
 840         uint8_t *buf;
 841         u32 old_crc, new_crc;
 842
 843         buf = calloc(BLOCK_SZ, 1);
 844         ASSERT(buf);
 845
 846         if (get_sb(feature) & F2FS_FEATURE_SB_CHKSUM) {
 847                 old_crc = get_sb(crc);
 848                 new_crc = f2fs_cal_crc32(F2FS_SUPER_MAGIC, sb,
 849                                                 SB_CHKSUM_OFFSET);
 850                 set_sb(crc, new_crc);
 851                 MSG(1, "Info: SB CRC is updated (0x%x -> 0x%x)\n",
 852                                                         old_crc, new_crc);
 853         }
 854
 855         memcpy(buf + F2FS_SUPER_OFFSET, sb, sizeof(*sb));
 856         for (addr = SB0_ADDR; addr < SB_MAX_ADDR; addr++) {
 857                 if (SB_MASK(addr) & sb_mask) {
 858                         ret = dev_write_block(buf, addr);
 859                         ASSERT(ret >= 0);
 860                 }
 861         }
 862
 863         free(buf);
 864         DBG(0, "Info: Done to update superblock\n");
 865 }
 866
 867 static inline int sanity_check_area_boundary(struct f2fs_super_block *sb,
 868                                                         enum SB_ADDR sb_addr)
 869 {
 870         u32 segment0_blkaddr = get_sb(segment0_blkaddr);
 871         u32 cp_blkaddr = get_sb(cp_blkaddr);
 872         u32 sit_blkaddr = get_sb(sit_blkaddr);
 873         u32 nat_blkaddr = get_sb(nat_blkaddr);
 874         u32 ssa_blkaddr = get_sb(ssa_blkaddr);
 875         u32 main_blkaddr = get_sb(main_blkaddr);
 876         u32 segment_count_ckpt = get_sb(segment_count_ckpt);
 877         u32 segment_count_sit = get_sb(segment_count_sit);
 878         u32 segment_count_nat = get_sb(segment_count_nat);
 879         u32 segment_count_ssa = get_sb(segment_count_ssa);
 880         u32 segment_count_main = get_sb(segment_count_main);
 881         u32 segment_count = get_sb(segment_count);
 882         u32 log_blocks_per_seg = get_sb(log_blocks_per_seg);
 883         u64 main_end_blkaddr = main_blkaddr +
 884                                 (segment_count_main << log_blocks_per_seg);
 885         u64 seg_end_blkaddr = segment0_blkaddr +
 886                                 (segment_count << log_blocks_per_seg);
 887
 888         if (segment0_blkaddr != cp_blkaddr) {
 889                 MSG(0, "\tMismatch segment0(%u) cp_blkaddr(%u)\n",
 890                                 segment0_blkaddr, cp_blkaddr);
 891                 return -1;
 892         }
 893
 894         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
 895                                                         sit_blkaddr) {
 896                 MSG(0, "\tWrong CP boundary, start(%u) end(%u) blocks(%u)\n",
 897                         cp_blkaddr, sit_blkaddr,
 898                         segment_count_ckpt << log_blocks_per_seg);
 899                 return -1;
 900         }
 901
 902         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
 903                                                         nat_blkaddr) {
 904                 MSG(0, "\tWrong SIT boundary, start(%u) end(%u) blocks(%u)\n",
 905                         sit_blkaddr, nat_blkaddr,
 906                         segment_count_sit << log_blocks_per_seg);
 907                 return -1;
 908         }
 909
 910         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
 911                                                         ssa_blkaddr) {
 912                 MSG(0, "\tWrong NAT boundary, start(%u) end(%u) blocks(%u)\n",
 913                         nat_blkaddr, ssa_blkaddr,
 914                         segment_count_nat << log_blocks_per_seg);
 915                 return -1;
 916         }
 917
 918         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
 919                                                         main_blkaddr) {
 920                 MSG(0, "\tWrong SSA boundary, start(%u) end(%u) blocks(%u)\n",
 921                         ssa_blkaddr, main_blkaddr,
 922                         segment_count_ssa << log_blocks_per_seg);
 923                 return -1;
 924         }
 925
 926         if (main_end_blkaddr > seg_end_blkaddr) {
 927                 MSG(0, "\tWrong MAIN_AREA, start(%u) end(%u) block(%u)\n",
 928                         main_blkaddr,
 929                         segment0_blkaddr +
 930                                 (segment_count << log_blocks_per_seg),
 931                         segment_count_main << log_blocks_per_seg);
 932                 return -1;
 933         } else if (main_end_blkaddr < seg_end_blkaddr) {
 934                 set_sb(segment_count, (main_end_blkaddr -
 935                                 segment0_blkaddr) >> log_blocks_per_seg);
 936
 937                 update_superblock(sb, SB_MASK(sb_addr));
 938                 MSG(0, "Info: Fix alignment: start(%u) end(%u) block(%u)\n",
 939                         main_blkaddr,
 940                         segment0_blkaddr +
 941                                 (segment_count << log_blocks_per_seg),
 942                         segment_count_main << log_blocks_per_seg);
 943         }
 944         return 0;
 945 }
 946
 947 static int verify_sb_chksum(struct f2fs_super_block *sb)
 948 {
 949         if (SB_CHKSUM_OFFSET != get_sb(checksum_offset)) {
 950                 MSG(0, "\tInvalid SB CRC offset: %u\n",
 951                                         get_sb(checksum_offset));
 952                 return -1;
 953         }
 954         if (f2fs_crc_valid(get_sb(crc), sb,
 955                         get_sb(checksum_offset))) {
 956                 MSG(0, "\tInvalid SB CRC: 0x%x\n", get_sb(crc));
 957                 return -1;
 958         }
 959         return 0;
 960 }
 961
 962 int sanity_check_raw_super(struct f2fs_super_block *sb, enum SB_ADDR sb_addr)
 963 {
 964         unsigned int blocksize;
 965         unsigned int segment_count, segs_per_sec, secs_per_zone, segs_per_zone;
 966         unsigned int total_sections, blocks_per_seg;
 967
 968         if (F2FS_SUPER_MAGIC != get_sb(magic)) {
 969                 MSG(0, "Magic Mismatch, valid(0x%x) - read(0x%x)\n",
 970                         F2FS_SUPER_MAGIC, get_sb(magic));
 971                 return -1;
 972         }
 973
 974         if ((get_sb(feature) & F2FS_FEATURE_SB_CHKSUM) &&
 975                                         verify_sb_chksum(sb))
 976                 return -1;
 977
 978         blocksize = 1 << get_sb(log_blocksize);
 979         if (F2FS_BLKSIZE != blocksize) {
 980                 MSG(0, "Invalid blocksize (%u), supports only 4KB\n",
 981                         blocksize);
 982                 return -1;
 983         }
 984
 985         /* check log blocks per segment */
 986         if (get_sb(log_blocks_per_seg) != 9) {
 987                 MSG(0, "Invalid log blocks per segment (%u)\n",
 988                         get_sb(log_blocks_per_seg));
 989                 return -1;
 990         }
 991
 992         /* Currently, support 512/1024/2048/4096 bytes sector size */
 993         if (get_sb(log_sectorsize) > F2FS_MAX_LOG_SECTOR_SIZE ||
 994                         get_sb(log_sectorsize) < F2FS_MIN_LOG_SECTOR_SIZE) {
 995                 MSG(0, "Invalid log sectorsize (%u)\n", get_sb(log_sectorsize));
 996                 return -1;
 997         }
 998
 999         if (get_sb(log_sectors_per_block) + get_sb(log_sectorsize) !=
1000                                                 F2FS_MAX_LOG_SECTOR_SIZE) {
1001                 MSG(0, "Invalid log sectors per block(%u) log sectorsize(%u)\n",
1002                         get_sb(log_sectors_per_block),
1003                         get_sb(log_sectorsize));
1004                 return -1;
1005         }
1006
1007         segment_count = get_sb(segment_count);
1008         segs_per_sec = get_sb(segs_per_sec);
1009         secs_per_zone = get_sb(secs_per_zone);
1010         total_sections = get_sb(section_count);
1011         segs_per_zone = segs_per_sec * secs_per_zone;
1012
1013         /* blocks_per_seg should be 512, given the above check */
1014         blocks_per_seg = 1 << get_sb(log_blocks_per_seg);
1015
1016         if (segment_count > F2FS_MAX_SEGMENT ||
1017                         segment_count < F2FS_MIN_SEGMENTS) {
1018                 MSG(0, "\tInvalid segment count (%u)\n", segment_count);
1019                 return -1;
1020         }
1021
1022         if (!(get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO)) &&
1023                         (total_sections > segment_count ||
1024                         total_sections < F2FS_MIN_SEGMENTS ||
1025                         segs_per_sec > segment_count || !segs_per_sec)) {
1026                 MSG(0, "\tInvalid segment/section count (%u, %u x %u)\n",
1027                         segment_count, total_sections, segs_per_sec);
1028                 return 1;
1029         }
1030
1031         if ((segment_count / segs_per_sec) < total_sections) {
1032                 MSG(0, "Small segment_count (%u < %u * %u)\n",
1033                         segment_count, segs_per_sec, total_sections);
1034                 return 1;
1035         }
1036
1037         if (segment_count > (get_sb(block_count) >> 9)) {
1038                 MSG(0, "Wrong segment_count / block_count (%u > %llu)\n",
1039                         segment_count, get_sb(block_count));
1040                 return 1;
1041         }
1042
1043         if (sb->devs[0].path[0]) {
1044                 unsigned int dev_segs = le32_to_cpu(sb->devs[0].total_segments);
1045                 int i = 1;
1046
1047                 while (i < MAX_DEVICES && sb->devs[i].path[0]) {
1048                         dev_segs += le32_to_cpu(sb->devs[i].total_segments);
1049                         i++;
1050                 }
1051                 if (segment_count != dev_segs / segs_per_zone * segs_per_zone) {
1052                         MSG(0, "Segment count (%u) mismatch with total segments from devices (%u)",
1053                                 segment_count, dev_segs);
1054                         return 1;
1055                 }
1056         }
1057
1058         if (secs_per_zone > total_sections || !secs_per_zone) {
1059                 MSG(0, "Wrong secs_per_zone / total_sections (%u, %u)\n",
1060                         secs_per_zone, total_sections);
1061                 return 1;
1062         }
1063         if (get_sb(extension_count) > F2FS_MAX_EXTENSION ||
1064                         sb->hot_ext_count > F2FS_MAX_EXTENSION ||
1065                         get_sb(extension_count) +
1066                         sb->hot_ext_count > F2FS_MAX_EXTENSION) {
1067                 MSG(0, "Corrupted extension count (%u + %u > %u)\n",
1068                         get_sb(extension_count),
1069                         sb->hot_ext_count,
1070                         F2FS_MAX_EXTENSION);
1071                 return 1;
1072         }
1073
1074         if (get_sb(cp_payload) > (blocks_per_seg - F2FS_CP_PACKS)) {
1075                 MSG(0, "Insane cp_payload (%u > %u)\n",
1076                         get_sb(cp_payload), blocks_per_seg - F2FS_CP_PACKS);
1077                 return 1;
1078         }
1079
1080         /* check reserved ino info */
1081         if (get_sb(node_ino) != 1 || get_sb(meta_ino) != 2 ||
1082                                                 get_sb(root_ino) != 3) {
1083                 MSG(0, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)\n",
1084                         get_sb(node_ino), get_sb(meta_ino), get_sb(root_ino));
1085                 return -1;
1086         }
1087
1088         /* Check zoned block device feature */
1089         if (c.devices[0].zoned_model != F2FS_ZONED_NONE &&
1090                         !(sb->feature & cpu_to_le32(F2FS_FEATURE_BLKZONED))) {
1091                 MSG(0, "\tMissing zoned block device feature\n");
1092                 return -1;
1093         }
1094
1095         if (sanity_check_area_boundary(sb, sb_addr))
1096                 return -1;
1097         return 0;
1098 }
1099
1100 #define CHECK_PERIOD (3600 * 24 * 30)   // one month by default
1101
1102 int validate_super_block(struct f2fs_sb_info *sbi, enum SB_ADDR sb_addr)
1103 {
1104         char buf[F2FS_BLKSIZE];
1105
1106         sbi->raw_super = malloc(sizeof(struct f2fs_super_block));
1107         if (!sbi->raw_super)
1108                 return -ENOMEM;
1109
1110         if (dev_read_block(buf, sb_addr))
1111                 return -1;
1112
1113         memcpy(sbi->raw_super, buf + F2FS_SUPER_OFFSET,
1114                                         sizeof(struct f2fs_super_block));
1115
1116         if (!sanity_check_raw_super(sbi->raw_super, sb_addr)) {
1117                 /* get kernel version */
1118                 if (c.kd >= 0) {
1119                         dev_read_version(c.version, 0, VERSION_NAME_LEN);
1120                         get_kernel_version(c.version);
1121                 } else {
1122                         get_kernel_uname_version(c.version);
1123                 }
1124
1125                 /* build sb version */
1126                 memcpy(c.sb_version, sbi->raw_super->version, VERSION_NAME_LEN);
1127                 get_kernel_version(c.sb_version);
1128                 memcpy(c.init_version, sbi->raw_super->init_version,
1129                                 VERSION_NAME_LEN);
1130                 get_kernel_version(c.init_version);
1131
1132                 c.force_stop = is_checkpoint_stop(sbi->raw_super, false);
1133                 c.abnormal_stop = is_checkpoint_stop(sbi->raw_super, true);
1134                 c.fs_errors = is_inconsistent_error(sbi->raw_super);
1135
1136                 MSG(0, "Info: MKFS version\n  \"%s\"\n", c.init_version);
1137                 MSG(0, "Info: FSCK version\n  from \"%s\"\n    to \"%s\"\n",
1138                                         c.sb_version, c.version);
1139                 print_sb_state(sbi->raw_super);
1140                 print_sb_stop_reason(sbi->raw_super);
1141                 print_sb_errors(sbi->raw_super);
1142                 return 0;
1143         }
1144
1145         free(sbi->raw_super);
1146         sbi->raw_super = NULL;
1147         MSG(0, "\tCan't find a valid F2FS superblock at 0x%x\n", sb_addr);
1148
1149         return -EINVAL;
1150 }
1151
1152 int init_sb_info(struct f2fs_sb_info *sbi)
1153 {
1154         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1155         u64 total_sectors;
1156         int i;
1157
1158         sbi->log_sectors_per_block = get_sb(log_sectors_per_block);
1159         sbi->log_blocksize = get_sb(log_blocksize);
1160         sbi->blocksize = 1 << sbi->log_blocksize;
1161         sbi->log_blocks_per_seg = get_sb(log_blocks_per_seg);
1162         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1163         sbi->segs_per_sec = get_sb(segs_per_sec);
1164         sbi->secs_per_zone = get_sb(secs_per_zone);
1165         sbi->total_sections = get_sb(section_count);
1166         sbi->total_node_count = (get_sb(segment_count_nat) / 2) *
1167                                 sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1168         sbi->root_ino_num = get_sb(root_ino);
1169         sbi->node_ino_num = get_sb(node_ino);
1170         sbi->meta_ino_num = get_sb(meta_ino);
1171         sbi->cur_victim_sec = NULL_SEGNO;
1172
1173         for (i = 0; i < MAX_DEVICES; i++) {
1174                 if (!sb->devs[i].path[0])
1175                         break;
1176
1177                 if (i) {
1178                         c.devices[i].path = strdup((char *)sb->devs[i].path);
1179                         if (get_device_info(i))
1180                                 ASSERT(0);
1181                 } else {
1182                         ASSERT(!strcmp((char *)sb->devs[i].path,
1183                                                 (char *)c.devices[i].path));
1184                 }
1185
1186                 c.devices[i].total_segments =
1187                         le32_to_cpu(sb->devs[i].total_segments);
1188                 if (i)
1189                         c.devices[i].start_blkaddr =
1190                                 c.devices[i - 1].end_blkaddr + 1;
1191                 c.devices[i].end_blkaddr = c.devices[i].start_blkaddr +
1192                         c.devices[i].total_segments *
1193                         c.blks_per_seg - 1;
1194                 if (i == 0)
1195                         c.devices[i].end_blkaddr += get_sb(segment0_blkaddr);
1196
1197                 if (c.zoned_model == F2FS_ZONED_NONE) {
1198                         if (c.devices[i].zoned_model == F2FS_ZONED_HM)
1199                                 c.zoned_model = F2FS_ZONED_HM;
1200                         else if (c.devices[i].zoned_model == F2FS_ZONED_HA &&
1201                                         c.zoned_model != F2FS_ZONED_HM)
1202                                 c.zoned_model = F2FS_ZONED_HA;
1203                 }
1204
1205                 c.ndevs = i + 1;
1206                 MSG(0, "Info: Device[%d] : %s blkaddr = %"PRIx64"--%"PRIx64"\n",
1207                                 i, c.devices[i].path,
1208                                 c.devices[i].start_blkaddr,
1209                                 c.devices[i].end_blkaddr);
1210         }
1211
1212         total_sectors = get_sb(block_count) << sbi->log_sectors_per_block;
1213         MSG(0, "Info: Segments per section = %d\n", sbi->segs_per_sec);
1214         MSG(0, "Info: Sections per zone = %d\n", sbi->secs_per_zone);
1215         MSG(0, "Info: total FS sectors = %"PRIu64" (%"PRIu64" MB)\n",
1216                                 total_sectors, total_sectors >>
1217                                                 (20 - get_sb(log_sectorsize)));
1218         return 0;
1219 }
1220
1221 static int verify_checksum_chksum(struct f2fs_checkpoint *cp)
1222 {
1223         unsigned int chksum_offset = get_cp(checksum_offset);
1224         unsigned int crc, cal_crc;
1225
1226         if (chksum_offset < CP_MIN_CHKSUM_OFFSET ||
1227                         chksum_offset > CP_CHKSUM_OFFSET) {
1228                 MSG(0, "\tInvalid CP CRC offset: %u\n", chksum_offset);
1229                 return -1;
1230         }
1231
1232         crc = le32_to_cpu(*(__le32 *)((unsigned char *)cp + chksum_offset));
1233         cal_crc = f2fs_checkpoint_chksum(cp);
1234         if (cal_crc != crc) {
1235                 MSG(0, "\tInvalid CP CRC: offset:%u, crc:0x%x, calc:0x%x\n",
1236                         chksum_offset, crc, cal_crc);
1237                 return -1;
1238         }
1239         return 0;
1240 }
1241
1242 static void *get_checkpoint_version(block_t cp_addr)
1243 {
1244         void *cp_page;
1245
1246         cp_page = malloc(F2FS_BLKSIZE);
1247         ASSERT(cp_page);
1248
1249         if (dev_read_block(cp_page, cp_addr) < 0)
1250                 ASSERT(0);
1251
1252         if (verify_checksum_chksum((struct f2fs_checkpoint *)cp_page))
1253                 goto out;
1254         return cp_page;
1255 out:
1256         free(cp_page);
1257         return NULL;
1258 }
1259
1260 void *validate_checkpoint(struct f2fs_sb_info *sbi, block_t cp_addr,
1261                                 unsigned long long *version)
1262 {
1263         void *cp_page_1, *cp_page_2;
1264         struct f2fs_checkpoint *cp;
1265         unsigned long long cur_version = 0, pre_version = 0;
1266
1267         /* Read the 1st cp block in this CP pack */
1268         cp_page_1 = get_checkpoint_version(cp_addr);
1269         if (!cp_page_1)
1270                 return NULL;
1271
1272         cp = (struct f2fs_checkpoint *)cp_page_1;
1273         if (get_cp(cp_pack_total_block_count) > sbi->blocks_per_seg)
1274                 goto invalid_cp1;
1275
1276         pre_version = get_cp(checkpoint_ver);
1277
1278         /* Read the 2nd cp block in this CP pack */
1279         cp_addr += get_cp(cp_pack_total_block_count) - 1;
1280         cp_page_2 = get_checkpoint_version(cp_addr);
1281         if (!cp_page_2)
1282                 goto invalid_cp1;
1283
1284         cp = (struct f2fs_checkpoint *)cp_page_2;
1285         cur_version = get_cp(checkpoint_ver);
1286
1287         if (cur_version == pre_version) {
1288                 *version = cur_version;
1289                 free(cp_page_2);
1290                 return cp_page_1;
1291         }
1292
1293         free(cp_page_2);
1294 invalid_cp1:
1295         free(cp_page_1);
1296         return NULL;
1297 }
1298
1299 int get_valid_checkpoint(struct f2fs_sb_info *sbi)
1300 {
1301         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1302         void *cp1, *cp2, *cur_page;
1303         unsigned long blk_size = sbi->blocksize;
1304         unsigned long long cp1_version = 0, cp2_version = 0, version;
1305         unsigned long long cp_start_blk_no;
1306         unsigned int cp_payload, cp_blks;
1307         int ret;
1308
1309         cp_payload = get_sb(cp_payload);
1310         if (cp_payload > F2FS_BLK_ALIGN(MAX_CP_PAYLOAD))
1311                 return -EINVAL;
1312
1313         cp_blks = 1 + cp_payload;
1314         sbi->ckpt = malloc(cp_blks * blk_size);
1315         if (!sbi->ckpt)
1316                 return -ENOMEM;
1317         /*
1318          * Finding out valid cp block involves read both
1319          * sets( cp pack1 and cp pack 2)
1320          */
1321         cp_start_blk_no = get_sb(cp_blkaddr);
1322         cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version);
1323
1324         /* The second checkpoint pack should start at the next segment */
1325         cp_start_blk_no += 1 << get_sb(log_blocks_per_seg);
1326         cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version);
1327
1328         if (cp1 && cp2) {
1329                 if (ver_after(cp2_version, cp1_version)) {
1330                         cur_page = cp2;
1331                         sbi->cur_cp = 2;
1332                         version = cp2_version;
1333                 } else {
1334                         cur_page = cp1;
1335                         sbi->cur_cp = 1;
1336                         version = cp1_version;
1337                 }
1338         } else if (cp1) {
1339                 cur_page = cp1;
1340                 sbi->cur_cp = 1;
1341                 version = cp1_version;
1342         } else if (cp2) {
1343                 cur_page = cp2;
1344                 sbi->cur_cp = 2;
1345                 version = cp2_version;
1346         } else
1347                 goto fail_no_cp;
1348
1349         MSG(0, "Info: CKPT version = %llx\n", version);
1350
1351         memcpy(sbi->ckpt, cur_page, blk_size);
1352
1353         if (cp_blks > 1) {
1354                 unsigned int i;
1355                 unsigned long long cp_blk_no;
1356
1357                 cp_blk_no = get_sb(cp_blkaddr);
1358                 if (cur_page == cp2)
1359                         cp_blk_no += 1 << get_sb(log_blocks_per_seg);
1360
1361                 /* copy sit bitmap */
1362                 for (i = 1; i < cp_blks; i++) {
1363                         unsigned char *ckpt = (unsigned char *)sbi->ckpt;
1364                         ret = dev_read_block(cur_page, cp_blk_no + i);
1365                         ASSERT(ret >= 0);
1366                         memcpy(ckpt + i * blk_size, cur_page, blk_size);
1367                 }
1368         }
1369         if (cp1)
1370                 free(cp1);
1371         if (cp2)
1372                 free(cp2);
1373         return 0;
1374
1375 fail_no_cp:
1376         free(sbi->ckpt);
1377         sbi->ckpt = NULL;
1378         return -EINVAL;
1379 }
1380
1381 bool is_checkpoint_stop(struct f2fs_super_block *sb, bool abnormal)
1382 {
1383         int i;
1384
1385         for (i = 0; i < STOP_CP_REASON_MAX; i++) {
1386                 if (abnormal && i == STOP_CP_REASON_SHUTDOWN)
1387                         continue;
1388                 if (sb->s_stop_reason[i])
1389                         return true;
1390         }
1391
1392         return false;
1393 }
1394
1395 bool is_inconsistent_error(struct f2fs_super_block *sb)
1396 {
1397         int i;
1398
1399         for (i = 0; i < MAX_F2FS_ERRORS; i++) {
1400                 if (sb->s_errors[i])
1401                         return true;
1402         }
1403
1404         return false;
1405 }
1406
1407 /*
1408  * For a return value of 1, caller should further check for c.fix_on state
1409  * and take appropriate action.
1410  */
1411 static int f2fs_should_proceed(struct f2fs_super_block *sb, u32 flag)
1412 {
1413         if (!c.fix_on && (c.auto_fix || c.preen_mode)) {
1414                 if (flag & CP_FSCK_FLAG ||
1415                         flag & CP_QUOTA_NEED_FSCK_FLAG ||
1416                         c.abnormal_stop || c.fs_errors ||
1417                         (exist_qf_ino(sb) && (flag & CP_ERROR_FLAG))) {
1418                         c.fix_on = 1;
1419                 } else if (!c.preen_mode) {
1420                         print_cp_state(flag);
1421                         return 0;
1422                 }
1423         }
1424         return 1;
1425 }
1426
1427 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1428 {
1429         unsigned int total, fsmeta;
1430         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1431         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
1432         unsigned int flag = get_cp(ckpt_flags);
1433         unsigned int ovp_segments, reserved_segments;
1434         unsigned int main_segs, blocks_per_seg;
1435         unsigned int sit_segs, nat_segs;
1436         unsigned int sit_bitmap_size, nat_bitmap_size;
1437         unsigned int log_blocks_per_seg;
1438         unsigned int segment_count_main;
1439         unsigned int cp_pack_start_sum, cp_payload;
1440         block_t user_block_count;
1441         int i;
1442
1443         total = get_sb(segment_count);
1444         fsmeta = get_sb(segment_count_ckpt);
1445         sit_segs = get_sb(segment_count_sit);
1446         fsmeta += sit_segs;
1447         nat_segs = get_sb(segment_count_nat);
1448         fsmeta += nat_segs;
1449         fsmeta += get_cp(rsvd_segment_count);
1450         fsmeta += get_sb(segment_count_ssa);
1451
1452         if (fsmeta >= total)
1453                 return 1;
1454
1455         ovp_segments = get_cp(overprov_segment_count);
1456         reserved_segments = get_cp(rsvd_segment_count);
1457
1458         if (!(get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO)) &&
1459                 (fsmeta < F2FS_MIN_SEGMENT || ovp_segments == 0 ||
1460                                         reserved_segments == 0)) {
1461                 MSG(0, "\tWrong layout: check mkfs.f2fs version\n");
1462                 return 1;
1463         }
1464
1465         user_block_count = get_cp(user_block_count);
1466         segment_count_main = get_sb(segment_count_main) +
1467                                 (cpu_to_le32(F2FS_FEATURE_RO) ? 1 : 0);
1468         log_blocks_per_seg = get_sb(log_blocks_per_seg);
1469         if (!user_block_count || user_block_count >=
1470                         segment_count_main << log_blocks_per_seg) {
1471                 ASSERT_MSG("\tWrong user_block_count(%u)\n", user_block_count);
1472
1473                 if (!f2fs_should_proceed(sb, flag))
1474                         return 1;
1475                 if (!c.fix_on)
1476                         return 1;
1477
1478                 if (flag & (CP_FSCK_FLAG | CP_RESIZEFS_FLAG)) {
1479                         u32 valid_user_block_cnt;
1480                         u32 seg_cnt_main = get_sb(segment_count) -
1481                                         (get_sb(segment_count_ckpt) +
1482                                          get_sb(segment_count_sit) +
1483                                          get_sb(segment_count_nat) +
1484                                          get_sb(segment_count_ssa));
1485
1486                         /* validate segment_count_main in sb first */
1487                         if (seg_cnt_main != get_sb(segment_count_main)) {
1488                                 MSG(0, "Inconsistent segment_cnt_main %u in sb\n",
1489                                                 segment_count_main << log_blocks_per_seg);
1490                                 return 1;
1491                         }
1492                         valid_user_block_cnt = ((get_sb(segment_count_main) -
1493                                                 get_cp(overprov_segment_count)) * c.blks_per_seg);
1494                         MSG(0, "Info: Fix wrong user_block_count in CP: (%u) -> (%u)\n",
1495                                         user_block_count, valid_user_block_cnt);
1496                         set_cp(user_block_count, valid_user_block_cnt);
1497                         c.bug_on = 1;
1498                 }
1499         }
1500
1501         main_segs = get_sb(segment_count_main);
1502         blocks_per_seg = sbi->blocks_per_seg;
1503
1504         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
1505                 if (get_cp(cur_node_segno[i]) >= main_segs ||
1506                         get_cp(cur_node_blkoff[i]) >= blocks_per_seg)
1507                         return 1;
1508         }
1509         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
1510                 if (get_cp(cur_data_segno[i]) >= main_segs ||
1511                         get_cp(cur_data_blkoff[i]) >= blocks_per_seg)
1512                         return 1;
1513         }
1514
1515         sit_bitmap_size = get_cp(sit_ver_bitmap_bytesize);
1516         nat_bitmap_size = get_cp(nat_ver_bitmap_bytesize);
1517
1518         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
1519                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
1520                 MSG(0, "\tWrong bitmap size: sit(%u), nat(%u)\n",
1521                                 sit_bitmap_size, nat_bitmap_size);
1522                 return 1;
1523         }
1524
1525         cp_pack_start_sum = __start_sum_addr(sbi);
1526         cp_payload = __cp_payload(sbi);
1527         if (cp_pack_start_sum < cp_payload + 1 ||
1528                 cp_pack_start_sum > blocks_per_seg - 1 -
1529                         NR_CURSEG_TYPE) {
1530                 MSG(0, "\tWrong cp_pack_start_sum(%u) or cp_payload(%u)\n",
1531                         cp_pack_start_sum, cp_payload);
1532                 if ((get_sb(feature) & F2FS_FEATURE_SB_CHKSUM))
1533                         return 1;
1534                 set_sb(cp_payload, cp_pack_start_sum - 1);
1535                 update_superblock(sb, SB_MASK_ALL);
1536         }
1537
1538         return 0;
1539 }
1540
1541 pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start, int *pack)
1542 {
1543         struct f2fs_nm_info *nm_i = NM_I(sbi);
1544         pgoff_t block_off;
1545         pgoff_t block_addr;
1546         int seg_off;
1547
1548         block_off = NAT_BLOCK_OFFSET(start);
1549         seg_off = block_off >> sbi->log_blocks_per_seg;
1550
1551         block_addr = (pgoff_t)(nm_i->nat_blkaddr +
1552                         (seg_off << sbi->log_blocks_per_seg << 1) +
1553                         (block_off & ((1 << sbi->log_blocks_per_seg) -1)));
1554         if (pack)
1555                 *pack = 1;
1556
1557         if (f2fs_test_bit(block_off, nm_i->nat_bitmap)) {
1558                 block_addr += sbi->blocks_per_seg;
1559                 if (pack)
1560                         *pack = 2;
1561         }
1562
1563         return block_addr;
1564 }
1565
1566 /* will not init nid_bitmap from nat */
1567 static int f2fs_early_init_nid_bitmap(struct f2fs_sb_info *sbi)
1568 {
1569         struct f2fs_nm_info *nm_i = NM_I(sbi);
1570         int nid_bitmap_size = (nm_i->max_nid + BITS_PER_BYTE - 1) / BITS_PER_BYTE;
1571         struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1572         struct f2fs_summary_block *sum = curseg->sum_blk;
1573         struct f2fs_journal *journal = &sum->journal;
1574         nid_t nid;
1575         int i;
1576
1577         if (!(c.func == SLOAD || c.func == FSCK))
1578                 return 0;
1579
1580         nm_i->nid_bitmap = (char *)calloc(nid_bitmap_size, 1);
1581         if (!nm_i->nid_bitmap)
1582                 return -ENOMEM;
1583
1584         /* arbitrarily set 0 bit */
1585         f2fs_set_bit(0, nm_i->nid_bitmap);
1586
1587         if (nats_in_cursum(journal) > NAT_JOURNAL_ENTRIES) {
1588                 MSG(0, "\tError: f2fs_init_nid_bitmap truncate n_nats(%u) to "
1589                         "NAT_JOURNAL_ENTRIES(%zu)\n",
1590                         nats_in_cursum(journal), NAT_JOURNAL_ENTRIES);
1591                 journal->n_nats = cpu_to_le16(NAT_JOURNAL_ENTRIES);
1592                 c.fix_on = 1;
1593         }
1594
1595         for (i = 0; i < nats_in_cursum(journal); i++) {
1596                 block_t addr;
1597
1598                 addr = le32_to_cpu(nat_in_journal(journal, i).block_addr);
1599                 if (!IS_VALID_BLK_ADDR(sbi, addr)) {
1600                         MSG(0, "\tError: f2fs_init_nid_bitmap: addr(%u) is invalid!!!\n", addr);
1601                         journal->n_nats = cpu_to_le16(i);
1602                         c.fix_on = 1;
1603                         continue;
1604                 }
1605
1606                 nid = le32_to_cpu(nid_in_journal(journal, i));
1607                 if (!IS_VALID_NID(sbi, nid)) {
1608                         MSG(0, "\tError: f2fs_init_nid_bitmap: nid(%u) is invalid!!!\n", nid);
1609                         journal->n_nats = cpu_to_le16(i);
1610                         c.fix_on = 1;
1611                         continue;
1612                 }
1613                 if (addr != NULL_ADDR)
1614                         f2fs_set_bit(nid, nm_i->nid_bitmap);
1615         }
1616         return 0;
1617 }
1618
1619 /* will init nid_bitmap from nat */
1620 static int f2fs_late_init_nid_bitmap(struct f2fs_sb_info *sbi)
1621 {
1622         struct f2fs_nm_info *nm_i = NM_I(sbi);
1623         struct f2fs_nat_block *nat_block;
1624         block_t start_blk;
1625         nid_t nid;
1626
1627         if (!(c.func == SLOAD || c.func == FSCK))
1628                 return 0;
1629
1630         nat_block = malloc(F2FS_BLKSIZE);
1631         if (!nat_block) {
1632                 free(nm_i->nid_bitmap);
1633                 return -ENOMEM;
1634         }
1635
1636         f2fs_ra_meta_pages(sbi, 0, NAT_BLOCK_OFFSET(nm_i->max_nid),
1637                                                         META_NAT);
1638         for (nid = 0; nid < nm_i->max_nid; nid++) {
1639                 if (!(nid % NAT_ENTRY_PER_BLOCK)) {
1640                         int ret;
1641
1642                         start_blk = current_nat_addr(sbi, nid, NULL);
1643                         ret = dev_read_block(nat_block, start_blk);
1644                         ASSERT(ret >= 0);
1645                 }
1646
1647                 if (nat_block->entries[nid % NAT_ENTRY_PER_BLOCK].block_addr)
1648                         f2fs_set_bit(nid, nm_i->nid_bitmap);
1649         }
1650
1651         free(nat_block);
1652         return 0;
1653 }
1654
1655 u32 update_nat_bits_flags(struct f2fs_super_block *sb,
1656                                 struct f2fs_checkpoint *cp, u32 flags)
1657 {
1658         uint32_t nat_bits_bytes, nat_bits_blocks;
1659
1660         nat_bits_bytes = get_sb(segment_count_nat) << 5;
1661         nat_bits_blocks = F2FS_BYTES_TO_BLK((nat_bits_bytes << 1) + 8 +
1662                                                 F2FS_BLKSIZE - 1);
1663         if (get_cp(cp_pack_total_block_count) <=
1664                         (1 << get_sb(log_blocks_per_seg)) - nat_bits_blocks)
1665                 flags |= CP_NAT_BITS_FLAG;
1666         else
1667                 flags &= (~CP_NAT_BITS_FLAG);
1668
1669         return flags;
1670 }
1671
1672 /* should call flush_journal_entries() bfore this */
1673 void write_nat_bits(struct f2fs_sb_info *sbi,
1674         struct f2fs_super_block *sb, struct f2fs_checkpoint *cp, int set)
1675 {
1676         struct f2fs_nm_info *nm_i = NM_I(sbi);
1677         uint32_t nat_blocks = get_sb(segment_count_nat) <<
1678                                 (get_sb(log_blocks_per_seg) - 1);
1679         uint32_t nat_bits_bytes = nat_blocks >> 3;
1680         uint32_t nat_bits_blocks = F2FS_BYTES_TO_BLK((nat_bits_bytes << 1) +
1681                                         8 + F2FS_BLKSIZE - 1);
1682         unsigned char *nat_bits, *full_nat_bits, *empty_nat_bits;
1683         struct f2fs_nat_block *nat_block;
1684         uint32_t i, j;
1685         block_t blkaddr;
1686         int ret;
1687
1688         nat_bits = calloc(F2FS_BLKSIZE, nat_bits_blocks);
1689         ASSERT(nat_bits);
1690
1691         nat_block = malloc(F2FS_BLKSIZE);
1692         ASSERT(nat_block);
1693
1694         full_nat_bits = nat_bits + 8;
1695         empty_nat_bits = full_nat_bits + nat_bits_bytes;
1696
1697         memset(full_nat_bits, 0, nat_bits_bytes);
1698         memset(empty_nat_bits, 0, nat_bits_bytes);
1699
1700         for (i = 0; i < nat_blocks; i++) {
1701                 int seg_off = i >> get_sb(log_blocks_per_seg);
1702                 int valid = 0;
1703
1704                 blkaddr = (pgoff_t)(get_sb(nat_blkaddr) +
1705                                 (seg_off << get_sb(log_blocks_per_seg) << 1) +
1706                                 (i & ((1 << get_sb(log_blocks_per_seg)) - 1)));
1707
1708                 /*
1709                  * Should consider new nat_blocks is larger than old
1710                  * nm_i->nat_blocks, since nm_i->nat_bitmap is based on
1711                  * old one.
1712                  */
1713                 if (i < nm_i->nat_blocks && f2fs_test_bit(i, nm_i->nat_bitmap))
1714                         blkaddr += (1 << get_sb(log_blocks_per_seg));
1715
1716                 ret = dev_read_block(nat_block, blkaddr);
1717                 ASSERT(ret >= 0);
1718
1719                 for (j = 0; j < NAT_ENTRY_PER_BLOCK; j++) {
1720                         if ((i == 0 && j == 0) ||
1721                                 nat_block->entries[j].block_addr != NULL_ADDR)
1722                                 valid++;
1723                 }
1724                 if (valid == 0)
1725                         test_and_set_bit_le(i, empty_nat_bits);
1726                 else if (valid == NAT_ENTRY_PER_BLOCK)
1727                         test_and_set_bit_le(i, full_nat_bits);
1728         }
1729         *(__le64 *)nat_bits = get_cp_crc(cp);
1730         free(nat_block);
1731
1732         blkaddr = get_sb(segment0_blkaddr) + (set <<
1733                                 get_sb(log_blocks_per_seg)) - nat_bits_blocks;
1734
1735         DBG(1, "\tWriting NAT bits pages, at offset 0x%08x\n", blkaddr);
1736
1737         for (i = 0; i < nat_bits_blocks; i++) {
1738                 if (dev_write_block(nat_bits + i * F2FS_BLKSIZE, blkaddr + i))
1739                         ASSERT_MSG("\tError: write NAT bits to disk!!!\n");
1740         }
1741         MSG(0, "Info: Write valid nat_bits in checkpoint\n");
1742
1743         free(nat_bits);
1744 }
1745
1746 static int check_nat_bits(struct f2fs_sb_info *sbi,
1747         struct f2fs_super_block *sb, struct f2fs_checkpoint *cp)
1748 {
1749         struct f2fs_nm_info *nm_i = NM_I(sbi);
1750         uint32_t nat_blocks = get_sb(segment_count_nat) <<
1751                                 (get_sb(log_blocks_per_seg) - 1);
1752         uint32_t nat_bits_bytes = nat_blocks >> 3;
1753         uint32_t nat_bits_blocks = F2FS_BYTES_TO_BLK((nat_bits_bytes << 1) +
1754                                         8 + F2FS_BLKSIZE - 1);
1755         unsigned char *nat_bits, *full_nat_bits, *empty_nat_bits;
1756         struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1757         struct f2fs_journal *journal = &curseg->sum_blk->journal;
1758         uint32_t i, j;
1759         block_t blkaddr;
1760         int err = 0;
1761
1762         nat_bits = calloc(F2FS_BLKSIZE, nat_bits_blocks);
1763         ASSERT(nat_bits);
1764
1765         full_nat_bits = nat_bits + 8;
1766         empty_nat_bits = full_nat_bits + nat_bits_bytes;
1767
1768         blkaddr = get_sb(segment0_blkaddr) + (sbi->cur_cp <<
1769                                 get_sb(log_blocks_per_seg)) - nat_bits_blocks;
1770
1771         for (i = 0; i < nat_bits_blocks; i++) {
1772                 if (dev_read_block(nat_bits + i * F2FS_BLKSIZE, blkaddr + i))
1773                         ASSERT_MSG("\tError: read NAT bits to disk!!!\n");
1774         }
1775
1776         if (*(__le64 *)nat_bits != get_cp_crc(cp) || nats_in_cursum(journal)) {
1777                 /*
1778                  * if there is a journal, f2fs was not shutdown cleanly. Let's
1779                  * flush them with nat_bits.
1780                  */
1781                 if (c.fix_on)
1782                         err = -1;
1783                 /* Otherwise, kernel will disable nat_bits */
1784                 goto out;
1785         }
1786
1787         for (i = 0; i < nat_blocks; i++) {
1788                 uint32_t start_nid = i * NAT_ENTRY_PER_BLOCK;
1789                 uint32_t valid = 0;
1790                 int empty = test_bit_le(i, empty_nat_bits);
1791                 int full = test_bit_le(i, full_nat_bits);
1792
1793                 for (j = 0; j < NAT_ENTRY_PER_BLOCK; j++) {
1794                         if (f2fs_test_bit(start_nid + j, nm_i->nid_bitmap))
1795                                 valid++;
1796                 }
1797                 if (valid == 0) {
1798                         if (!empty || full) {
1799                                 err = -1;
1800                                 goto out;
1801                         }
1802                 } else if (valid == NAT_ENTRY_PER_BLOCK) {
1803                         if (empty || !full) {
1804                                 err = -1;
1805                                 goto out;
1806                         }
1807                 } else {
1808                         if (empty || full) {
1809                                 err = -1;
1810                                 goto out;
1811                         }
1812                 }
1813         }
1814 out:
1815         free(nat_bits);
1816         if (!err) {
1817                 MSG(0, "Info: Checked valid nat_bits in checkpoint\n");
1818         } else {
1819                 c.bug_nat_bits = 1;
1820                 MSG(0, "Info: Corrupted valid nat_bits in checkpoint\n");
1821         }
1822         return err;
1823 }
1824
1825 int init_node_manager(struct f2fs_sb_info *sbi)
1826 {
1827         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1828         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
1829         struct f2fs_nm_info *nm_i = NM_I(sbi);
1830         unsigned char *version_bitmap;
1831         unsigned int nat_segs;
1832
1833         nm_i->nat_blkaddr = get_sb(nat_blkaddr);
1834
1835         /* segment_count_nat includes pair segment so divide to 2. */
1836         nat_segs = get_sb(segment_count_nat) >> 1;
1837         nm_i->nat_blocks = nat_segs << get_sb(log_blocks_per_seg);
1838         nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nm_i->nat_blocks;
1839         nm_i->fcnt = 0;
1840         nm_i->nat_cnt = 0;
1841         nm_i->init_scan_nid = get_cp(next_free_nid);
1842         nm_i->next_scan_nid = get_cp(next_free_nid);
1843
1844         nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP);
1845
1846         nm_i->nat_bitmap = malloc(nm_i->bitmap_size);
1847         if (!nm_i->nat_bitmap)
1848                 return -ENOMEM;
1849         version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP);
1850         if (!version_bitmap)
1851                 return -EFAULT;
1852
1853         /* copy version bitmap */
1854         memcpy(nm_i->nat_bitmap, version_bitmap, nm_i->bitmap_size);
1855         return f2fs_early_init_nid_bitmap(sbi);
1856 }
1857
1858 int build_node_manager(struct f2fs_sb_info *sbi)
1859 {
1860         int err;
1861         sbi->nm_info = malloc(sizeof(struct f2fs_nm_info));
1862         if (!sbi->nm_info)
1863                 return -ENOMEM;
1864
1865         err = init_node_manager(sbi);
1866         if (err)
1867                 return err;
1868
1869         return 0;
1870 }
1871
1872 int build_sit_info(struct f2fs_sb_info *sbi)
1873 {
1874         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1875         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
1876         struct sit_info *sit_i;
1877         unsigned int sit_segs;
1878         int start;
1879         char *src_bitmap, *dst_bitmap;
1880         unsigned char *bitmap;
1881         unsigned int bitmap_size;
1882
1883         sit_i = malloc(sizeof(struct sit_info));
1884         if (!sit_i) {
1885                 MSG(1, "\tError: Malloc failed for build_sit_info!\n");
1886                 return -ENOMEM;
1887         }
1888
1889         SM_I(sbi)->sit_info = sit_i;
1890
1891         sit_i->sentries = calloc(MAIN_SEGS(sbi) * sizeof(struct seg_entry), 1);
1892         if (!sit_i->sentries) {
1893                 MSG(1, "\tError: Calloc failed for build_sit_info!\n");
1894                 goto free_sit_info;
1895         }
1896
1897         bitmap_size = MAIN_SEGS(sbi) * SIT_VBLOCK_MAP_SIZE;
1898
1899         if (need_fsync_data_record(sbi))
1900                 bitmap_size += bitmap_size;
1901
1902         sit_i->bitmap = calloc(bitmap_size, 1);
1903         if (!sit_i->bitmap) {
1904                 MSG(1, "\tError: Calloc failed for build_sit_info!!\n");
1905                 goto free_sentries;
1906         }
1907
1908         bitmap = sit_i->bitmap;
1909
1910         for (start = 0; start < MAIN_SEGS(sbi); start++) {
1911                 sit_i->sentries[start].cur_valid_map = bitmap;
1912                 bitmap += SIT_VBLOCK_MAP_SIZE;
1913
1914                 if (need_fsync_data_record(sbi)) {
1915                         sit_i->sentries[start].ckpt_valid_map = bitmap;
1916                         bitmap += SIT_VBLOCK_MAP_SIZE;
1917                 }
1918         }
1919
1920         sit_segs = get_sb(segment_count_sit) >> 1;
1921         bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
1922         src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
1923
1924         dst_bitmap = malloc(bitmap_size);
1925         if (!dst_bitmap) {
1926                 MSG(1, "\tError: Malloc failed for build_sit_info!!\n");
1927                 goto free_validity_maps;
1928         }
1929
1930         memcpy(dst_bitmap, src_bitmap, bitmap_size);
1931
1932         sit_i->sit_base_addr = get_sb(sit_blkaddr);
1933         sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
1934         sit_i->written_valid_blocks = get_cp(valid_block_count);
1935         sit_i->sit_bitmap = dst_bitmap;
1936         sit_i->bitmap_size = bitmap_size;
1937         sit_i->dirty_sentries = 0;
1938         sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
1939         sit_i->elapsed_time = get_cp(elapsed_time);
1940         return 0;
1941
1942 free_validity_maps:
1943         free(sit_i->bitmap);
1944 free_sentries:
1945         free(sit_i->sentries);
1946 free_sit_info:
1947         free(sit_i);
1948
1949         return -ENOMEM;
1950 }
1951
1952 void reset_curseg(struct f2fs_sb_info *sbi, int type)
1953 {
1954         struct curseg_info *curseg = CURSEG_I(sbi, type);
1955         struct summary_footer *sum_footer;
1956         struct seg_entry *se;
1957
1958         sum_footer = &(curseg->sum_blk->footer);
1959         memset(sum_footer, 0, sizeof(struct summary_footer));
1960         if (IS_DATASEG(type))
1961                 SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
1962         if (IS_NODESEG(type))
1963                 SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
1964         se = get_seg_entry(sbi, curseg->segno);
1965         se->type = type;
1966         se->dirty = 1;
1967 }
1968
1969 static void read_compacted_summaries(struct f2fs_sb_info *sbi)
1970 {
1971         struct curseg_info *curseg;
1972         unsigned int i, j, offset;
1973         block_t start;
1974         char *kaddr;
1975         int ret;
1976
1977         start = start_sum_block(sbi);
1978
1979         kaddr = malloc(F2FS_BLKSIZE);
1980         ASSERT(kaddr);
1981
1982         ret = dev_read_block(kaddr, start++);
1983         ASSERT(ret >= 0);
1984
1985         curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1986         memcpy(&curseg->sum_blk->journal.n_nats, kaddr, SUM_JOURNAL_SIZE);
1987
1988         curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
1989         memcpy(&curseg->sum_blk->journal.n_sits, kaddr + SUM_JOURNAL_SIZE,
1990                                                 SUM_JOURNAL_SIZE);
1991
1992         offset = 2 * SUM_JOURNAL_SIZE;
1993         for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1994                 unsigned short blk_off;
1995                 struct curseg_info *curseg = CURSEG_I(sbi, i);
1996
1997                 reset_curseg(sbi, i);
1998
1999                 if (curseg->alloc_type == SSR)
2000                         blk_off = sbi->blocks_per_seg;
2001                 else
2002                         blk_off = curseg->next_blkoff;
2003
2004                 ASSERT(blk_off <= ENTRIES_IN_SUM);
2005
2006                 for (j = 0; j < blk_off; j++) {
2007                         struct f2fs_summary *s;
2008                         s = (struct f2fs_summary *)(kaddr + offset);
2009                         curseg->sum_blk->entries[j] = *s;
2010                         offset += SUMMARY_SIZE;
2011                         if (offset + SUMMARY_SIZE <=
2012                                         F2FS_BLKSIZE - SUM_FOOTER_SIZE)
2013                                 continue;
2014                         memset(kaddr, 0, F2FS_BLKSIZE);
2015                         ret = dev_read_block(kaddr, start++);
2016                         ASSERT(ret >= 0);
2017                         offset = 0;
2018                 }
2019         }
2020         free(kaddr);
2021 }
2022
2023 static void restore_node_summary(struct f2fs_sb_info *sbi,
2024                 unsigned int segno, struct f2fs_summary_block *sum_blk)
2025 {
2026         struct f2fs_node *node_blk;
2027         struct f2fs_summary *sum_entry;
2028         block_t addr;
2029         unsigned int i;
2030         int ret;
2031
2032         node_blk = malloc(F2FS_BLKSIZE);
2033         ASSERT(node_blk);
2034
2035         /* scan the node segment */
2036         addr = START_BLOCK(sbi, segno);
2037         sum_entry = &sum_blk->entries[0];
2038
2039         for (i = 0; i < sbi->blocks_per_seg; i++, sum_entry++) {
2040                 ret = dev_read_block(node_blk, addr);
2041                 ASSERT(ret >= 0);
2042                 sum_entry->nid = node_blk->footer.nid;
2043                 addr++;
2044         }
2045         free(node_blk);
2046 }
2047
2048 static void read_normal_summaries(struct f2fs_sb_info *sbi, int type)
2049 {
2050         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2051         struct f2fs_summary_block *sum_blk;
2052         struct curseg_info *curseg;
2053         unsigned int segno = 0;
2054         block_t blk_addr = 0;
2055         int ret;
2056
2057         if (IS_DATASEG(type)) {
2058                 segno = get_cp(cur_data_segno[type]);
2059                 if (is_set_ckpt_flags(cp, CP_UMOUNT_FLAG))
2060                         blk_addr = sum_blk_addr(sbi, NR_CURSEG_TYPE, type);
2061                 else
2062                         blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
2063         } else {
2064                 segno = get_cp(cur_node_segno[type - CURSEG_HOT_NODE]);
2065                 if (is_set_ckpt_flags(cp, CP_UMOUNT_FLAG))
2066                         blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
2067                                                         type - CURSEG_HOT_NODE);
2068                 else
2069                         blk_addr = GET_SUM_BLKADDR(sbi, segno);
2070         }
2071
2072         sum_blk = malloc(sizeof(*sum_blk));
2073         ASSERT(sum_blk);
2074
2075         ret = dev_read_block(sum_blk, blk_addr);
2076         ASSERT(ret >= 0);
2077
2078         if (IS_NODESEG(type) && !is_set_ckpt_flags(cp, CP_UMOUNT_FLAG))
2079                 restore_node_summary(sbi, segno, sum_blk);
2080
2081         curseg = CURSEG_I(sbi, type);
2082         memcpy(curseg->sum_blk, sum_blk, sizeof(*sum_blk));
2083         reset_curseg(sbi, type);
2084         free(sum_blk);
2085 }
2086
2087 void update_sum_entry(struct f2fs_sb_info *sbi, block_t blk_addr,
2088                                         struct f2fs_summary *sum)
2089 {
2090         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2091         struct f2fs_summary_block *sum_blk;
2092         u32 segno, offset;
2093         int type, ret;
2094         struct seg_entry *se;
2095
2096         if (get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO))
2097                 return;
2098
2099         segno = GET_SEGNO(sbi, blk_addr);
2100         offset = OFFSET_IN_SEG(sbi, blk_addr);
2101
2102         se = get_seg_entry(sbi, segno);
2103
2104         sum_blk = get_sum_block(sbi, segno, &type);
2105         memcpy(&sum_blk->entries[offset], sum, sizeof(*sum));
2106         sum_blk->footer.entry_type = IS_NODESEG(se->type) ? SUM_TYPE_NODE :
2107                                                         SUM_TYPE_DATA;
2108
2109         /* write SSA all the time */
2110         ret = dev_write_block(sum_blk, GET_SUM_BLKADDR(sbi, segno));
2111         ASSERT(ret >= 0);
2112
2113         if (type == SEG_TYPE_NODE || type == SEG_TYPE_DATA ||
2114                                         type == SEG_TYPE_MAX)
2115                 free(sum_blk);
2116 }
2117
2118 static void restore_curseg_summaries(struct f2fs_sb_info *sbi)
2119 {
2120         int type = CURSEG_HOT_DATA;
2121
2122         if (is_set_ckpt_flags(F2FS_CKPT(sbi), CP_COMPACT_SUM_FLAG)) {
2123                 read_compacted_summaries(sbi);
2124                 type = CURSEG_HOT_NODE;
2125         }
2126
2127         for (; type <= CURSEG_COLD_NODE; type++)
2128                 read_normal_summaries(sbi, type);
2129 }
2130
2131 static int build_curseg(struct f2fs_sb_info *sbi)
2132 {
2133         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2134         struct curseg_info *array;
2135         unsigned short blk_off;
2136         unsigned int segno;
2137         int i;
2138
2139         array = malloc(sizeof(*array) * NR_CURSEG_TYPE);
2140         if (!array) {
2141                 MSG(1, "\tError: Malloc failed for build_curseg!\n");
2142                 return -ENOMEM;
2143         }
2144
2145         SM_I(sbi)->curseg_array = array;
2146
2147         for (i = 0; i < NR_CURSEG_TYPE; i++) {
2148                 array[i].sum_blk = calloc(sizeof(*(array[i].sum_blk)), 1);
2149                 if (!array[i].sum_blk) {
2150                         MSG(1, "\tError: Calloc failed for build_curseg!!\n");
2151                         goto seg_cleanup;
2152                 }
2153
2154                 if (i <= CURSEG_COLD_DATA) {
2155                         blk_off = get_cp(cur_data_blkoff[i]);
2156                         segno = get_cp(cur_data_segno[i]);
2157                 }
2158                 if (i > CURSEG_COLD_DATA) {
2159                         blk_off = get_cp(cur_node_blkoff[i - CURSEG_HOT_NODE]);
2160                         segno = get_cp(cur_node_segno[i - CURSEG_HOT_NODE]);
2161                 }
2162                 ASSERT(segno < MAIN_SEGS(sbi));
2163                 ASSERT(blk_off < DEFAULT_BLOCKS_PER_SEGMENT);
2164
2165                 array[i].segno = segno;
2166                 array[i].zone = GET_ZONENO_FROM_SEGNO(sbi, segno);
2167                 array[i].next_segno = NULL_SEGNO;
2168                 array[i].next_blkoff = blk_off;
2169                 array[i].alloc_type = cp->alloc_type[i];
2170         }
2171         restore_curseg_summaries(sbi);
2172         return 0;
2173
2174 seg_cleanup:
2175         for(--i ; i >=0; --i)
2176                 free(array[i].sum_blk);
2177         free(array);
2178
2179         return -ENOMEM;
2180 }
2181
2182 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
2183 {
2184         unsigned int end_segno = SM_I(sbi)->segment_count - 1;
2185         ASSERT(segno <= end_segno);
2186 }
2187
2188 static inline block_t current_sit_addr(struct f2fs_sb_info *sbi,
2189                                                 unsigned int segno)
2190 {
2191         struct sit_info *sit_i = SIT_I(sbi);
2192         unsigned int offset = SIT_BLOCK_OFFSET(sit_i, segno);
2193         block_t blk_addr = sit_i->sit_base_addr + offset;
2194
2195         check_seg_range(sbi, segno);
2196
2197         /* calculate sit block address */
2198         if (f2fs_test_bit(offset, sit_i->sit_bitmap))
2199                 blk_addr += sit_i->sit_blocks;
2200
2201         return blk_addr;
2202 }
2203
2204 void get_current_sit_page(struct f2fs_sb_info *sbi,
2205                         unsigned int segno, struct f2fs_sit_block *sit_blk)
2206 {
2207         block_t blk_addr = current_sit_addr(sbi, segno);
2208
2209         ASSERT(dev_read_block(sit_blk, blk_addr) >= 0);
2210 }
2211
2212 void rewrite_current_sit_page(struct f2fs_sb_info *sbi,
2213                         unsigned int segno, struct f2fs_sit_block *sit_blk)
2214 {
2215         block_t blk_addr = current_sit_addr(sbi, segno);
2216
2217         ASSERT(dev_write_block(sit_blk, blk_addr) >= 0);
2218 }
2219
2220 void check_block_count(struct f2fs_sb_info *sbi,
2221                 unsigned int segno, struct f2fs_sit_entry *raw_sit)
2222 {
2223         struct f2fs_sm_info *sm_info = SM_I(sbi);
2224         unsigned int end_segno = sm_info->segment_count - 1;
2225         int valid_blocks = 0;
2226         unsigned int i;
2227
2228         /* check segment usage */
2229         if (GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg)
2230                 ASSERT_MSG("Invalid SIT vblocks: segno=0x%x, %u",
2231                                 segno, GET_SIT_VBLOCKS(raw_sit));
2232
2233         /* check boundary of a given segment number */
2234         if (segno > end_segno)
2235                 ASSERT_MSG("Invalid SEGNO: 0x%x", segno);
2236
2237         /* check bitmap with valid block count */
2238         for (i = 0; i < SIT_VBLOCK_MAP_SIZE; i++)
2239                 valid_blocks += get_bits_in_byte(raw_sit->valid_map[i]);
2240
2241         if (GET_SIT_VBLOCKS(raw_sit) != valid_blocks)
2242                 ASSERT_MSG("Wrong SIT valid blocks: segno=0x%x, %u vs. %u",
2243                                 segno, GET_SIT_VBLOCKS(raw_sit), valid_blocks);
2244
2245         if (GET_SIT_TYPE(raw_sit) >= NO_CHECK_TYPE)
2246                 ASSERT_MSG("Wrong SIT type: segno=0x%x, %u",
2247                                 segno, GET_SIT_TYPE(raw_sit));
2248 }
2249
2250 void __seg_info_from_raw_sit(struct seg_entry *se,
2251                 struct f2fs_sit_entry *raw_sit)
2252 {
2253         se->valid_blocks = GET_SIT_VBLOCKS(raw_sit);
2254         memcpy(se->cur_valid_map, raw_sit->valid_map, SIT_VBLOCK_MAP_SIZE);
2255         se->type = GET_SIT_TYPE(raw_sit);
2256         se->orig_type = GET_SIT_TYPE(raw_sit);
2257         se->mtime = le64_to_cpu(raw_sit->mtime);
2258 }
2259
2260 void seg_info_from_raw_sit(struct f2fs_sb_info *sbi, struct seg_entry *se,
2261                                                 struct f2fs_sit_entry *raw_sit)
2262 {
2263         __seg_info_from_raw_sit(se, raw_sit);
2264
2265         if (!need_fsync_data_record(sbi))
2266                 return;
2267         se->ckpt_valid_blocks = se->valid_blocks;
2268         memcpy(se->ckpt_valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
2269         se->ckpt_type = se->type;
2270 }
2271
2272 struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
2273                 unsigned int segno)
2274 {
2275         struct sit_info *sit_i = SIT_I(sbi);
2276         return &sit_i->sentries[segno];
2277 }
2278
2279 unsigned short get_seg_vblocks(struct f2fs_sb_info *sbi, struct seg_entry *se)
2280 {
2281         if (!need_fsync_data_record(sbi))
2282                 return se->valid_blocks;
2283         else
2284                 return se->ckpt_valid_blocks;
2285 }
2286
2287 unsigned char *get_seg_bitmap(struct f2fs_sb_info *sbi, struct seg_entry *se)
2288 {
2289         if (!need_fsync_data_record(sbi))
2290                 return se->cur_valid_map;
2291         else
2292                 return se->ckpt_valid_map;
2293 }
2294
2295 unsigned char get_seg_type(struct f2fs_sb_info *sbi, struct seg_entry *se)
2296 {
2297         if (!need_fsync_data_record(sbi))
2298                 return se->type;
2299         else
2300                 return se->ckpt_type;
2301 }
2302
2303 struct f2fs_summary_block *get_sum_block(struct f2fs_sb_info *sbi,
2304                                 unsigned int segno, int *ret_type)
2305 {
2306         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2307         struct f2fs_summary_block *sum_blk;
2308         struct curseg_info *curseg;
2309         int type, ret;
2310         u64 ssa_blk;
2311
2312         *ret_type= SEG_TYPE_MAX;
2313
2314         ssa_blk = GET_SUM_BLKADDR(sbi, segno);
2315         for (type = 0; type < NR_CURSEG_NODE_TYPE; type++) {
2316                 if (segno == get_cp(cur_node_segno[type])) {
2317                         curseg = CURSEG_I(sbi, CURSEG_HOT_NODE + type);
2318                         if (!IS_SUM_NODE_SEG(curseg->sum_blk->footer)) {
2319                                 ASSERT_MSG("segno [0x%x] indicates a data "
2320                                                 "segment, but should be node",
2321                                                 segno);
2322                                 *ret_type = -SEG_TYPE_CUR_NODE;
2323                         } else {
2324                                 *ret_type = SEG_TYPE_CUR_NODE;
2325                         }
2326                         return curseg->sum_blk;
2327                 }
2328         }
2329
2330         for (type = 0; type < NR_CURSEG_DATA_TYPE; type++) {
2331                 if (segno == get_cp(cur_data_segno[type])) {
2332                         curseg = CURSEG_I(sbi, type);
2333                         if (IS_SUM_NODE_SEG(curseg->sum_blk->footer)) {
2334                                 ASSERT_MSG("segno [0x%x] indicates a node "
2335                                                 "segment, but should be data",
2336                                                 segno);
2337                                 *ret_type = -SEG_TYPE_CUR_DATA;
2338                         } else {
2339                                 *ret_type = SEG_TYPE_CUR_DATA;
2340                         }
2341                         return curseg->sum_blk;
2342                 }
2343         }
2344
2345         sum_blk = calloc(BLOCK_SZ, 1);
2346         ASSERT(sum_blk);
2347
2348         ret = dev_read_block(sum_blk, ssa_blk);
2349         ASSERT(ret >= 0);
2350
2351         if (IS_SUM_NODE_SEG(sum_blk->footer))
2352                 *ret_type = SEG_TYPE_NODE;
2353         else if (IS_SUM_DATA_SEG(sum_blk->footer))
2354                 *ret_type = SEG_TYPE_DATA;
2355
2356         return sum_blk;
2357 }
2358
2359 int get_sum_entry(struct f2fs_sb_info *sbi, u32 blk_addr,
2360                                 struct f2fs_summary *sum_entry)
2361 {
2362         struct f2fs_summary_block *sum_blk;
2363         u32 segno, offset;
2364         int type;
2365
2366         segno = GET_SEGNO(sbi, blk_addr);
2367         offset = OFFSET_IN_SEG(sbi, blk_addr);
2368
2369         sum_blk = get_sum_block(sbi, segno, &type);
2370         memcpy(sum_entry, &(sum_blk->entries[offset]),
2371                                 sizeof(struct f2fs_summary));
2372         if (type == SEG_TYPE_NODE || type == SEG_TYPE_DATA ||
2373                                         type == SEG_TYPE_MAX)
2374                 free(sum_blk);
2375         return type;
2376 }
2377
2378 static void get_nat_entry(struct f2fs_sb_info *sbi, nid_t nid,
2379                                 struct f2fs_nat_entry *raw_nat)
2380 {
2381         struct f2fs_nat_block *nat_block;
2382         pgoff_t block_addr;
2383         int entry_off;
2384         int ret;
2385
2386         if (lookup_nat_in_journal(sbi, nid, raw_nat) >= 0)
2387                 return;
2388
2389         nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
2390         ASSERT(nat_block);
2391
2392         entry_off = nid % NAT_ENTRY_PER_BLOCK;
2393         block_addr = current_nat_addr(sbi, nid, NULL);
2394
2395         ret = dev_read_block(nat_block, block_addr);
2396         ASSERT(ret >= 0);
2397
2398         memcpy(raw_nat, &nat_block->entries[entry_off],
2399                                         sizeof(struct f2fs_nat_entry));
2400         free(nat_block);
2401 }
2402
2403 void update_data_blkaddr(struct f2fs_sb_info *sbi, nid_t nid,
2404                                 u16 ofs_in_node, block_t newaddr)
2405 {
2406         struct f2fs_node *node_blk = NULL;
2407         struct node_info ni;
2408         block_t oldaddr, startaddr, endaddr;
2409         int ret;
2410
2411         node_blk = (struct f2fs_node *)calloc(BLOCK_SZ, 1);
2412         ASSERT(node_blk);
2413
2414         get_node_info(sbi, nid, &ni);
2415
2416         /* read node_block */
2417         ret = dev_read_block(node_blk, ni.blk_addr);
2418         ASSERT(ret >= 0);
2419
2420         /* check its block address */
2421         if (node_blk->footer.nid == node_blk->footer.ino) {
2422                 int ofs = get_extra_isize(node_blk);
2423
2424                 oldaddr = le32_to_cpu(node_blk->i.i_addr[ofs + ofs_in_node]);
2425                 node_blk->i.i_addr[ofs + ofs_in_node] = cpu_to_le32(newaddr);
2426                 ret = write_inode(node_blk, ni.blk_addr);
2427                 ASSERT(ret >= 0);
2428         } else {
2429                 oldaddr = le32_to_cpu(node_blk->dn.addr[ofs_in_node]);
2430                 node_blk->dn.addr[ofs_in_node] = cpu_to_le32(newaddr);
2431                 ret = dev_write_block(node_blk, ni.blk_addr);
2432                 ASSERT(ret >= 0);
2433         }
2434
2435         /* check extent cache entry */
2436         if (node_blk->footer.nid != node_blk->footer.ino) {
2437                 get_node_info(sbi, le32_to_cpu(node_blk->footer.ino), &ni);
2438
2439                 /* read inode block */
2440                 ret = dev_read_block(node_blk, ni.blk_addr);
2441                 ASSERT(ret >= 0);
2442         }
2443
2444         startaddr = le32_to_cpu(node_blk->i.i_ext.blk_addr);
2445         endaddr = startaddr + le32_to_cpu(node_blk->i.i_ext.len);
2446         if (oldaddr >= startaddr && oldaddr < endaddr) {
2447                 node_blk->i.i_ext.len = 0;
2448
2449                 /* update inode block */
2450                 ASSERT(write_inode(node_blk, ni.blk_addr) >= 0);
2451         }
2452         free(node_blk);
2453 }
2454
2455 void update_nat_blkaddr(struct f2fs_sb_info *sbi, nid_t ino,
2456                                         nid_t nid, block_t newaddr)
2457 {
2458         struct f2fs_nat_block *nat_block;
2459         pgoff_t block_addr;
2460         int entry_off;
2461         int ret;
2462
2463         nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
2464         ASSERT(nat_block);
2465
2466         entry_off = nid % NAT_ENTRY_PER_BLOCK;
2467         block_addr = current_nat_addr(sbi, nid, NULL);
2468
2469         ret = dev_read_block(nat_block, block_addr);
2470         ASSERT(ret >= 0);
2471
2472         if (ino)
2473                 nat_block->entries[entry_off].ino = cpu_to_le32(ino);
2474         nat_block->entries[entry_off].block_addr = cpu_to_le32(newaddr);
2475         if (c.func == FSCK)
2476                 F2FS_FSCK(sbi)->entries[nid] = nat_block->entries[entry_off];
2477
2478         ret = dev_write_block(nat_block, block_addr);
2479         ASSERT(ret >= 0);
2480         free(nat_block);
2481 }
2482
2483 void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni)
2484 {
2485         struct f2fs_nat_entry raw_nat;
2486
2487         ni->nid = nid;
2488         if (c.func == FSCK && F2FS_FSCK(sbi)->nr_nat_entries) {
2489                 node_info_from_raw_nat(ni, &(F2FS_FSCK(sbi)->entries[nid]));
2490                 if (ni->blk_addr)
2491                         return;
2492                 /* nat entry is not cached, read it */
2493         }
2494
2495         get_nat_entry(sbi, nid, &raw_nat);
2496         node_info_from_raw_nat(ni, &raw_nat);
2497 }
2498
2499 static int build_sit_entries(struct f2fs_sb_info *sbi)
2500 {
2501         struct sit_info *sit_i = SIT_I(sbi);
2502         struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
2503         struct f2fs_journal *journal = &curseg->sum_blk->journal;
2504         struct f2fs_sit_block *sit_blk;
2505         struct seg_entry *se;
2506         struct f2fs_sit_entry sit;
2507         int sit_blk_cnt = SIT_BLK_CNT(sbi);
2508         unsigned int i, segno, end;
2509         unsigned int readed, start_blk = 0;
2510
2511         sit_blk = calloc(BLOCK_SZ, 1);
2512         if (!sit_blk) {
2513                 MSG(1, "\tError: Calloc failed for build_sit_entries!\n");
2514                 return -ENOMEM;
2515         }
2516
2517         do {
2518                 readed = f2fs_ra_meta_pages(sbi, start_blk, MAX_RA_BLOCKS,
2519                                                                 META_SIT);
2520
2521                 segno = start_blk * sit_i->sents_per_block;
2522                 end = (start_blk + readed) * sit_i->sents_per_block;
2523
2524                 for (; segno < end && segno < MAIN_SEGS(sbi); segno++) {
2525                         se = &sit_i->sentries[segno];
2526
2527                         get_current_sit_page(sbi, segno, sit_blk);
2528                         sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, segno)];
2529
2530                         check_block_count(sbi, segno, &sit);
2531                         seg_info_from_raw_sit(sbi, se, &sit);
2532                 }
2533                 start_blk += readed;
2534         } while (start_blk < sit_blk_cnt);
2535
2536
2537         free(sit_blk);
2538
2539         if (sits_in_cursum(journal) > SIT_JOURNAL_ENTRIES) {
2540                 MSG(0, "\tError: build_sit_entries truncate n_sits(%u) to "
2541                         "SIT_JOURNAL_ENTRIES(%zu)\n",
2542                         sits_in_cursum(journal), SIT_JOURNAL_ENTRIES);
2543                 journal->n_sits = cpu_to_le16(SIT_JOURNAL_ENTRIES);
2544                 c.fix_on = 1;
2545         }
2546
2547         for (i = 0; i < sits_in_cursum(journal); i++) {
2548                 segno = le32_to_cpu(segno_in_journal(journal, i));
2549
2550                 if (segno >= MAIN_SEGS(sbi)) {
2551                         MSG(0, "\tError: build_sit_entries: segno(%u) is invalid!!!\n", segno);
2552                         journal->n_sits = cpu_to_le16(i);
2553                         c.fix_on = 1;
2554                         continue;
2555                 }
2556
2557                 se = &sit_i->sentries[segno];
2558                 sit = sit_in_journal(journal, i);
2559
2560                 check_block_count(sbi, segno, &sit);
2561                 seg_info_from_raw_sit(sbi, se, &sit);
2562         }
2563         return 0;
2564 }
2565
2566 static int early_build_segment_manager(struct f2fs_sb_info *sbi)
2567 {
2568         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2569         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2570         struct f2fs_sm_info *sm_info;
2571
2572         sm_info = malloc(sizeof(struct f2fs_sm_info));
2573         if (!sm_info) {
2574                 MSG(1, "\tError: Malloc failed for build_segment_manager!\n");
2575                 return -ENOMEM;
2576         }
2577
2578         /* init sm info */
2579         sbi->sm_info = sm_info;
2580         sm_info->seg0_blkaddr = get_sb(segment0_blkaddr);
2581         sm_info->main_blkaddr = get_sb(main_blkaddr);
2582         sm_info->segment_count = get_sb(segment_count);
2583         sm_info->reserved_segments = get_cp(rsvd_segment_count);
2584         sm_info->ovp_segments = get_cp(overprov_segment_count);
2585         sm_info->main_segments = get_sb(segment_count_main);
2586         sm_info->ssa_blkaddr = get_sb(ssa_blkaddr);
2587
2588         if (build_sit_info(sbi) || build_curseg(sbi)) {
2589                 free(sm_info);
2590                 return -ENOMEM;
2591         }
2592
2593         return 0;
2594 }
2595
2596 static int late_build_segment_manager(struct f2fs_sb_info *sbi)
2597 {
2598         if (sbi->seg_manager_done)
2599                 return 1; /* this function was already called */
2600
2601         sbi->seg_manager_done = true;
2602         if (build_sit_entries(sbi)) {
2603                 free (sbi->sm_info);
2604                 return -ENOMEM;
2605         }
2606
2607         return 0;
2608 }
2609
2610 void build_sit_area_bitmap(struct f2fs_sb_info *sbi)
2611 {
2612         struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2613         struct f2fs_sm_info *sm_i = SM_I(sbi);
2614         unsigned int segno = 0;
2615         char *ptr = NULL;
2616         u32 sum_vblocks = 0;
2617         u32 free_segs = 0;
2618         struct seg_entry *se;
2619
2620         fsck->sit_area_bitmap_sz = sm_i->main_segments * SIT_VBLOCK_MAP_SIZE;
2621         fsck->sit_area_bitmap = calloc(1, fsck->sit_area_bitmap_sz);
2622         ASSERT(fsck->sit_area_bitmap);
2623         ptr = fsck->sit_area_bitmap;
2624
2625         ASSERT(fsck->sit_area_bitmap_sz == fsck->main_area_bitmap_sz);
2626
2627         for (segno = 0; segno < MAIN_SEGS(sbi); segno++) {
2628                 se = get_seg_entry(sbi, segno);
2629
2630                 memcpy(ptr, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
2631                 ptr += SIT_VBLOCK_MAP_SIZE;
2632
2633                 if (se->valid_blocks == 0x0 && is_usable_seg(sbi, segno)) {
2634                         if (le32_to_cpu(sbi->ckpt->cur_node_segno[0]) == segno ||
2635                                 le32_to_cpu(sbi->ckpt->cur_data_segno[0]) == segno ||
2636                                 le32_to_cpu(sbi->ckpt->cur_node_segno[1]) == segno ||
2637                                 le32_to_cpu(sbi->ckpt->cur_data_segno[1]) == segno ||
2638                                 le32_to_cpu(sbi->ckpt->cur_node_segno[2]) == segno ||
2639                                 le32_to_cpu(sbi->ckpt->cur_data_segno[2]) == segno) {
2640                                 continue;
2641                         } else {
2642                                 free_segs++;
2643                         }
2644                 } else {
2645                         sum_vblocks += se->valid_blocks;
2646                 }
2647         }
2648         fsck->chk.sit_valid_blocks = sum_vblocks;
2649         fsck->chk.sit_free_segs = free_segs;
2650
2651         DBG(1, "Blocks [0x%x : %d] Free Segs [0x%x : %d]\n\n",
2652                         sum_vblocks, sum_vblocks,
2653                         free_segs, free_segs);
2654 }
2655
2656 void rewrite_sit_area_bitmap(struct f2fs_sb_info *sbi)
2657 {
2658         struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2659         struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
2660         struct sit_info *sit_i = SIT_I(sbi);
2661         struct f2fs_sit_block *sit_blk;
2662         unsigned int segno = 0;
2663         struct f2fs_summary_block *sum = curseg->sum_blk;
2664         char *ptr = NULL;
2665
2666         sit_blk = calloc(BLOCK_SZ, 1);
2667         ASSERT(sit_blk);
2668         /* remove sit journal */
2669         sum->journal.n_sits = 0;
2670
2671         ptr = fsck->main_area_bitmap;
2672
2673         for (segno = 0; segno < MAIN_SEGS(sbi); segno++) {
2674                 struct f2fs_sit_entry *sit;
2675                 struct seg_entry *se;
2676                 u16 valid_blocks = 0;
2677                 u16 type;
2678                 int i;
2679
2680                 get_current_sit_page(sbi, segno, sit_blk);
2681                 sit = &sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, segno)];
2682                 memcpy(sit->valid_map, ptr, SIT_VBLOCK_MAP_SIZE);
2683
2684                 /* update valid block count */
2685                 for (i = 0; i < SIT_VBLOCK_MAP_SIZE; i++)
2686                         valid_blocks += get_bits_in_byte(sit->valid_map[i]);
2687
2688                 se = get_seg_entry(sbi, segno);
2689                 memcpy(se->cur_valid_map, ptr, SIT_VBLOCK_MAP_SIZE);
2690                 se->valid_blocks = valid_blocks;
2691                 type = se->type;
2692                 if (type >= NO_CHECK_TYPE) {
2693                         ASSERT_MSG("Invalid type and valid blocks=%x,%x",
2694                                         segno, valid_blocks);
2695                         type = 0;
2696                 }
2697                 sit->vblocks = cpu_to_le16((type << SIT_VBLOCKS_SHIFT) |
2698                                                                 valid_blocks);
2699                 rewrite_current_sit_page(sbi, segno, sit_blk);
2700
2701                 ptr += SIT_VBLOCK_MAP_SIZE;
2702         }
2703
2704         free(sit_blk);
2705 }
2706
2707 static int flush_sit_journal_entries(struct f2fs_sb_info *sbi)
2708 {
2709         struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
2710         struct f2fs_journal *journal = &curseg->sum_blk->journal;
2711         struct sit_info *sit_i = SIT_I(sbi);
2712         struct f2fs_sit_block *sit_blk;
2713         unsigned int segno;
2714         int i;
2715
2716         sit_blk = calloc(BLOCK_SZ, 1);
2717         ASSERT(sit_blk);
2718         for (i = 0; i < sits_in_cursum(journal); i++) {
2719                 struct f2fs_sit_entry *sit;
2720                 struct seg_entry *se;
2721
2722                 segno = segno_in_journal(journal, i);
2723                 se = get_seg_entry(sbi, segno);
2724
2725                 get_current_sit_page(sbi, segno, sit_blk);
2726                 sit = &sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, segno)];
2727
2728                 memcpy(sit->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
2729                 sit->vblocks = cpu_to_le16((se->type << SIT_VBLOCKS_SHIFT) |
2730                                                         se->valid_blocks);
2731                 sit->mtime = cpu_to_le64(se->mtime);
2732
2733                 rewrite_current_sit_page(sbi, segno, sit_blk);
2734         }
2735
2736         free(sit_blk);
2737         journal->n_sits = 0;
2738         return i;
2739 }
2740
2741 static int flush_nat_journal_entries(struct f2fs_sb_info *sbi)
2742 {
2743         struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2744         struct f2fs_journal *journal = &curseg->sum_blk->journal;
2745         struct f2fs_nat_block *nat_block;
2746         pgoff_t block_addr;
2747         int entry_off;
2748         nid_t nid;
2749         int ret;
2750         int i = 0;
2751
2752         nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
2753         ASSERT(nat_block);
2754 next:
2755         if (i >= nats_in_cursum(journal)) {
2756                 free(nat_block);
2757                 journal->n_nats = 0;
2758                 return i;
2759         }
2760
2761         nid = le32_to_cpu(nid_in_journal(journal, i));
2762
2763         entry_off = nid % NAT_ENTRY_PER_BLOCK;
2764         block_addr = current_nat_addr(sbi, nid, NULL);
2765
2766         ret = dev_read_block(nat_block, block_addr);
2767         ASSERT(ret >= 0);
2768
2769         memcpy(&nat_block->entries[entry_off], &nat_in_journal(journal, i),
2770                                         sizeof(struct f2fs_nat_entry));
2771
2772         ret = dev_write_block(nat_block, block_addr);
2773         ASSERT(ret >= 0);
2774         i++;
2775         goto next;
2776 }
2777
2778 void flush_journal_entries(struct f2fs_sb_info *sbi)
2779 {
2780         int n_nats = flush_nat_journal_entries(sbi);
2781         int n_sits = flush_sit_journal_entries(sbi);
2782
2783         if (n_nats || n_sits)
2784                 write_checkpoints(sbi);
2785 }
2786
2787 void flush_sit_entries(struct f2fs_sb_info *sbi)
2788 {
2789         struct sit_info *sit_i = SIT_I(sbi);
2790         struct f2fs_sit_block *sit_blk;
2791         unsigned int segno = 0;
2792
2793         sit_blk = calloc(BLOCK_SZ, 1);
2794         ASSERT(sit_blk);
2795         /* update free segments */
2796         for (segno = 0; segno < MAIN_SEGS(sbi); segno++) {
2797                 struct f2fs_sit_entry *sit;
2798                 struct seg_entry *se;
2799
2800                 se = get_seg_entry(sbi, segno);
2801
2802                 if (!se->dirty)
2803                         continue;
2804
2805                 get_current_sit_page(sbi, segno, sit_blk);
2806                 sit = &sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, segno)];
2807                 memcpy(sit->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
2808                 sit->vblocks = cpu_to_le16((se->type << SIT_VBLOCKS_SHIFT) |
2809                                                         se->valid_blocks);
2810                 rewrite_current_sit_page(sbi, segno, sit_blk);
2811         }
2812
2813         free(sit_blk);
2814 }
2815
2816 int relocate_curseg_offset(struct f2fs_sb_info *sbi, int type)
2817 {
2818         struct curseg_info *curseg = CURSEG_I(sbi, type);
2819         struct seg_entry *se = get_seg_entry(sbi, curseg->segno);
2820         unsigned int i;
2821
2822         if (c.zoned_model == F2FS_ZONED_HM)
2823                 return -EINVAL;
2824
2825         for (i = 0; i < sbi->blocks_per_seg; i++) {
2826                 if (!f2fs_test_bit(i, (const char *)se->cur_valid_map))
2827                         break;
2828         }
2829
2830         if (i == sbi->blocks_per_seg)
2831                 return -EINVAL;
2832
2833         DBG(1, "Update curseg[%d].next_blkoff %u -> %u, alloc_type %s -> SSR\n",
2834                         type, curseg->next_blkoff, i,
2835                         curseg->alloc_type == LFS ? "LFS" : "SSR");
2836
2837         curseg->next_blkoff = i;
2838         curseg->alloc_type = SSR;
2839
2840         return 0;
2841 }
2842
2843 void set_section_type(struct f2fs_sb_info *sbi, unsigned int segno, int type)
2844 {
2845         unsigned int i;
2846
2847         if (sbi->segs_per_sec == 1)
2848                 return;
2849
2850         for (i = 0; i < sbi->segs_per_sec; i++) {
2851                 struct seg_entry *se = get_seg_entry(sbi, segno + i);
2852
2853                 se->type = type;
2854         }
2855 }
2856
2857 #ifdef HAVE_LINUX_BLKZONED_H
2858
2859 static bool write_pointer_at_zone_start(struct f2fs_sb_info *sbi,
2860                                         unsigned int zone_segno)
2861 {
2862         uint64_t sector;
2863         struct blk_zone blkz;
2864         block_t block = START_BLOCK(sbi, zone_segno);
2865         int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
2866         int ret, j;
2867
2868         if (c.zoned_model != F2FS_ZONED_HM)
2869                 return true;
2870
2871         for (j = 0; j < MAX_DEVICES; j++) {
2872                 if (!c.devices[j].path)
2873                         break;
2874                 if (c.devices[j].start_blkaddr <= block &&
2875                     block <= c.devices[j].end_blkaddr)
2876                         break;
2877         }
2878
2879         if (j >= MAX_DEVICES)
2880                 return false;
2881
2882         sector = (block - c.devices[j].start_blkaddr) << log_sectors_per_block;
2883         ret = f2fs_report_zone(j, sector, &blkz);
2884         if (ret)
2885                 return false;
2886
2887         if (blk_zone_type(&blkz) != BLK_ZONE_TYPE_SEQWRITE_REQ)
2888                 return true;
2889
2890         return blk_zone_sector(&blkz) == blk_zone_wp_sector(&blkz);
2891 }
2892
2893 #else
2894
2895 static bool write_pointer_at_zone_start(struct f2fs_sb_info *UNUSED(sbi),
2896                                         unsigned int UNUSED(zone_segno))
2897 {
2898         return true;
2899 }
2900
2901 #endif
2902
2903 int find_next_free_block(struct f2fs_sb_info *sbi, u64 *to, int left,
2904                                                 int want_type, bool new_sec)
2905 {
2906         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2907         struct seg_entry *se;
2908         u32 segno;
2909         u32 offset;
2910         int not_enough = 0;
2911         u64 end_blkaddr = (get_sb(segment_count_main) <<
2912                         get_sb(log_blocks_per_seg)) + get_sb(main_blkaddr);
2913
2914         if (*to > 0)
2915                 *to -= left;
2916         if (get_free_segments(sbi) <= SM_I(sbi)->reserved_segments + 1)
2917                 not_enough = 1;
2918
2919         while (*to >= SM_I(sbi)->main_blkaddr && *to < end_blkaddr) {
2920                 unsigned short vblocks;
2921                 unsigned char *bitmap;
2922                 unsigned char type;
2923
2924                 segno = GET_SEGNO(sbi, *to);
2925                 offset = OFFSET_IN_SEG(sbi, *to);
2926
2927                 se = get_seg_entry(sbi, segno);
2928
2929                 vblocks = get_seg_vblocks(sbi, se);
2930                 bitmap = get_seg_bitmap(sbi, se);
2931                 type = get_seg_type(sbi, se);
2932
2933                 if (vblocks == sbi->blocks_per_seg) {
2934 next_segment:
2935                         *to = left ? START_BLOCK(sbi, segno) - 1:
2936                                                 START_BLOCK(sbi, segno + 1);
2937                         continue;
2938                 }
2939                 if (!(get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO)) &&
2940                                                 IS_CUR_SEGNO(sbi, segno))
2941                         goto next_segment;
2942                 if (vblocks == 0 && not_enough)
2943                         goto next_segment;
2944
2945                 if (vblocks == 0 && !(segno % sbi->segs_per_sec)) {
2946                         struct seg_entry *se2;
2947                         unsigned int i;
2948
2949                         for (i = 1; i < sbi->segs_per_sec; i++) {
2950                                 se2 = get_seg_entry(sbi, segno + i);
2951                                 if (get_seg_vblocks(sbi, se2))
2952                                         break;
2953                         }
2954
2955                         if (i == sbi->segs_per_sec &&
2956                             write_pointer_at_zone_start(sbi, segno)) {
2957                                 set_section_type(sbi, segno, want_type);
2958                                 return 0;
2959                         }
2960                 }
2961
2962                 if (type == want_type && !new_sec &&
2963                         !f2fs_test_bit(offset, (const char *)bitmap))
2964                         return 0;
2965
2966                 *to = left ? *to - 1: *to + 1;
2967         }
2968         return -1;
2969 }
2970
2971 static void move_one_curseg_info(struct f2fs_sb_info *sbi, u64 from, int left,
2972                                  int i)
2973 {
2974         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2975         struct curseg_info *curseg = CURSEG_I(sbi, i);
2976         struct f2fs_summary_block buf;
2977         u32 old_segno;
2978         u64 ssa_blk, to;
2979         int ret;
2980
2981         if ((get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO))) {
2982                 if (i != CURSEG_HOT_DATA && i != CURSEG_HOT_NODE)
2983                         return;
2984
2985                 if (i == CURSEG_HOT_DATA) {
2986                         left = 0;
2987                         from = SM_I(sbi)->main_blkaddr;
2988                 } else {
2989                         left = 1;
2990                         from = __end_block_addr(sbi);
2991                 }
2992                 goto bypass_ssa;
2993         }
2994
2995         /* update original SSA too */
2996         ssa_blk = GET_SUM_BLKADDR(sbi, curseg->segno);
2997         ret = dev_write_block(curseg->sum_blk, ssa_blk);
2998         ASSERT(ret >= 0);
2999 bypass_ssa:
3000         to = from;
3001         ret = find_next_free_block(sbi, &to, left, i,
3002                                    c.zoned_model == F2FS_ZONED_HM);
3003         ASSERT(ret == 0);
3004
3005         old_segno = curseg->segno;
3006         curseg->segno = GET_SEGNO(sbi, to);
3007         curseg->next_blkoff = OFFSET_IN_SEG(sbi, to);
3008         curseg->alloc_type = c.zoned_model == F2FS_ZONED_HM ? LFS : SSR;
3009
3010         /* update new segno */
3011         ssa_blk = GET_SUM_BLKADDR(sbi, curseg->segno);
3012         ret = dev_read_block(&buf, ssa_blk);
3013         ASSERT(ret >= 0);
3014
3015         memcpy(curseg->sum_blk, &buf, SUM_ENTRIES_SIZE);
3016
3017         /* update se->types */
3018         reset_curseg(sbi, i);
3019
3020         FIX_MSG("Move curseg[%d] %x -> %x after %"PRIx64"\n",
3021                 i, old_segno, curseg->segno, from);
3022 }
3023
3024 void move_curseg_info(struct f2fs_sb_info *sbi, u64 from, int left)
3025 {
3026         int i;
3027
3028         /* update summary blocks having nullified journal entries */
3029         for (i = 0; i < NO_CHECK_TYPE; i++)
3030                 move_one_curseg_info(sbi, from, left, i);
3031 }
3032
3033 void update_curseg_info(struct f2fs_sb_info *sbi, int type)
3034 {
3035         if (!relocate_curseg_offset(sbi, type))
3036                 return;
3037         move_one_curseg_info(sbi, SM_I(sbi)->main_blkaddr, 0, type);
3038 }
3039
3040 void zero_journal_entries(struct f2fs_sb_info *sbi)
3041 {
3042         int i;
3043
3044         for (i = 0; i < NO_CHECK_TYPE; i++)
3045                 CURSEG_I(sbi, i)->sum_blk->journal.n_nats = 0;
3046 }
3047
3048 void write_curseg_info(struct f2fs_sb_info *sbi)
3049 {
3050         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
3051         int i;
3052
3053         for (i = 0; i < NO_CHECK_TYPE; i++) {
3054                 cp->alloc_type[i] = CURSEG_I(sbi, i)->alloc_type;
3055                 if (i < CURSEG_HOT_NODE) {
3056                         set_cp(cur_data_segno[i], CURSEG_I(sbi, i)->segno);
3057                         set_cp(cur_data_blkoff[i],
3058                                         CURSEG_I(sbi, i)->next_blkoff);
3059                 } else {
3060                         int n = i - CURSEG_HOT_NODE;
3061
3062                         set_cp(cur_node_segno[n], CURSEG_I(sbi, i)->segno);
3063                         set_cp(cur_node_blkoff[n],
3064                                         CURSEG_I(sbi, i)->next_blkoff);
3065                 }
3066         }
3067 }
3068
3069 int lookup_nat_in_journal(struct f2fs_sb_info *sbi, u32 nid,
3070                                         struct f2fs_nat_entry *raw_nat)
3071 {
3072         struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
3073         struct f2fs_journal *journal = &curseg->sum_blk->journal;
3074         int i = 0;
3075
3076         for (i = 0; i < nats_in_cursum(journal); i++) {
3077                 if (le32_to_cpu(nid_in_journal(journal, i)) == nid) {
3078                         memcpy(raw_nat, &nat_in_journal(journal, i),
3079                                                 sizeof(struct f2fs_nat_entry));
3080                         DBG(3, "==> Found nid [0x%x] in nat cache\n", nid);
3081                         return i;
3082                 }
3083         }
3084         return -1;
3085 }
3086
3087 void nullify_nat_entry(struct f2fs_sb_info *sbi, u32 nid)
3088 {
3089         struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
3090         struct f2fs_journal *journal = &curseg->sum_blk->journal;
3091         struct f2fs_nat_block *nat_block;
3092         pgoff_t block_addr;
3093         int entry_off;
3094         int ret;
3095         int i = 0;
3096
3097         /* check in journal */
3098         for (i = 0; i < nats_in_cursum(journal); i++) {
3099                 if (le32_to_cpu(nid_in_journal(journal, i)) == nid) {
3100                         memset(&nat_in_journal(journal, i), 0,
3101                                         sizeof(struct f2fs_nat_entry));
3102                         FIX_MSG("Remove nid [0x%x] in nat journal", nid);
3103                         return;
3104                 }
3105         }
3106         nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
3107         ASSERT(nat_block);
3108
3109         entry_off = nid % NAT_ENTRY_PER_BLOCK;
3110         block_addr = current_nat_addr(sbi, nid, NULL);
3111
3112         ret = dev_read_block(nat_block, block_addr);
3113         ASSERT(ret >= 0);
3114
3115         if (nid == F2FS_NODE_INO(sbi) || nid == F2FS_META_INO(sbi)) {
3116                 FIX_MSG("nid [0x%x] block_addr= 0x%x -> 0x1", nid,
3117                         le32_to_cpu(nat_block->entries[entry_off].block_addr));
3118                 nat_block->entries[entry_off].block_addr = cpu_to_le32(0x1);
3119         } else {
3120                 memset(&nat_block->entries[entry_off], 0,
3121                                         sizeof(struct f2fs_nat_entry));
3122                 FIX_MSG("Remove nid [0x%x] in NAT", nid);
3123         }
3124
3125         ret = dev_write_block(nat_block, block_addr);
3126         ASSERT(ret >= 0);
3127         free(nat_block);
3128 }
3129
3130 void duplicate_checkpoint(struct f2fs_sb_info *sbi)
3131 {
3132         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
3133         unsigned long long dst, src;
3134         void *buf;
3135         unsigned int seg_size = 1 << get_sb(log_blocks_per_seg);
3136         int ret;
3137
3138         if (sbi->cp_backuped)
3139                 return;
3140
3141         buf = malloc(F2FS_BLKSIZE * seg_size);
3142         ASSERT(buf);
3143
3144         if (sbi->cur_cp == 1) {
3145                 src = get_sb(cp_blkaddr);
3146                 dst = src + seg_size;
3147         } else {
3148                 dst = get_sb(cp_blkaddr);
3149                 src = dst + seg_size;
3150         }
3151
3152         ret = dev_read(buf, src << F2FS_BLKSIZE_BITS,
3153                                 seg_size << F2FS_BLKSIZE_BITS);
3154         ASSERT(ret >= 0);
3155
3156         ret = dev_write(buf, dst << F2FS_BLKSIZE_BITS,
3157                                 seg_size << F2FS_BLKSIZE_BITS);
3158         ASSERT(ret >= 0);
3159
3160         free(buf);
3161
3162         ret = f2fs_fsync_device();
3163         ASSERT(ret >= 0);
3164
3165         sbi->cp_backuped = 1;
3166
3167         MSG(0, "Info: Duplicate valid checkpoint to mirror position "
3168                 "%llu -> %llu\n", src, dst);
3169 }
3170
3171 void write_checkpoint(struct f2fs_sb_info *sbi)
3172 {
3173         struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
3174         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
3175         block_t orphan_blks = 0;
3176         unsigned long long cp_blk_no;
3177         u32 flags = CP_UMOUNT_FLAG;
3178         int i, ret;
3179         uint32_t crc = 0;
3180
3181         if (is_set_ckpt_flags(cp, CP_ORPHAN_PRESENT_FLAG)) {
3182                 orphan_blks = __start_sum_addr(sbi) - 1;
3183                 flags |= CP_ORPHAN_PRESENT_FLAG;
3184         }
3185         if (is_set_ckpt_flags(cp, CP_TRIMMED_FLAG))
3186                 flags |= CP_TRIMMED_FLAG;
3187         if (is_set_ckpt_flags(cp, CP_DISABLED_FLAG))
3188                 flags |= CP_DISABLED_FLAG;
3189         if (is_set_ckpt_flags(cp, CP_LARGE_NAT_BITMAP_FLAG)) {
3190                 flags |= CP_LARGE_NAT_BITMAP_FLAG;
3191                 set_cp(checksum_offset, CP_MIN_CHKSUM_OFFSET);
3192         } else {
3193                 set_cp(checksum_offset, CP_CHKSUM_OFFSET);
3194         }
3195
3196         set_cp(free_segment_count, get_free_segments(sbi));
3197         if (c.func == FSCK) {
3198                 struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
3199
3200                 set_cp(valid_block_count, fsck->chk.valid_blk_cnt);
3201                 set_cp(valid_node_count, fsck->chk.valid_node_cnt);
3202                 set_cp(valid_inode_count, fsck->chk.valid_inode_cnt);
3203         } else {
3204                 set_cp(valid_block_count, sbi->total_valid_block_count);
3205                 set_cp(valid_node_count, sbi->total_valid_node_count);
3206                 set_cp(valid_inode_count, sbi->total_valid_inode_count);
3207         }
3208         set_cp(cp_pack_total_block_count, 8 + orphan_blks + get_sb(cp_payload));
3209
3210         flags = update_nat_bits_flags(sb, cp, flags);
3211         set_cp(ckpt_flags, flags);
3212
3213         crc = f2fs_checkpoint_chksum(cp);
3214         *((__le32 *)((unsigned char *)cp + get_cp(checksum_offset))) =
3215                                                         cpu_to_le32(crc);
3216
3217         cp_blk_no = get_sb(cp_blkaddr);
3218         if (sbi->cur_cp == 2)
3219                 cp_blk_no += 1 << get_sb(log_blocks_per_seg);
3220
3221         /* write the first cp */
3222         ret = dev_write_block(cp, cp_blk_no++);
3223         ASSERT(ret >= 0);
3224
3225         /* skip payload */
3226         cp_blk_no += get_sb(cp_payload);
3227         /* skip orphan blocks */
3228         cp_blk_no += orphan_blks;
3229
3230         /* update summary blocks having nullified journal entries */
3231         for (i = 0; i < NO_CHECK_TYPE; i++) {
3232                 struct curseg_info *curseg = CURSEG_I(sbi, i);
3233                 u64 ssa_blk;
3234
3235                 ret = dev_write_block(curseg->sum_blk, cp_blk_no++);
3236                 ASSERT(ret >= 0);
3237
3238                 if (!(get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO))) {
3239                         /* update original SSA too */
3240                         ssa_blk = GET_SUM_BLKADDR(sbi, curseg->segno);
3241                         ret = dev_write_block(curseg->sum_blk, ssa_blk);
3242                         ASSERT(ret >= 0);
3243                 }
3244         }
3245
3246         /* Write nat bits */
3247         if (flags & CP_NAT_BITS_FLAG)
3248                 write_nat_bits(sbi, sb, cp, sbi->cur_cp);
3249
3250         /* in case of sudden power off */
3251         ret = f2fs_fsync_device();
3252         ASSERT(ret >= 0);
3253
3254         /* write the last cp */
3255         ret = dev_write_block(cp, cp_blk_no++);
3256         ASSERT(ret >= 0);
3257
3258         ret = f2fs_fsync_device();
3259         ASSERT(ret >= 0);
3260 }
3261
3262 void write_checkpoints(struct f2fs_sb_info *sbi)
3263 {
3264         /* copy valid checkpoint to its mirror position */
3265         duplicate_checkpoint(sbi);
3266
3267         /* repair checkpoint at CP #0 position */
3268         sbi->cur_cp = 1;
3269         write_checkpoint(sbi);
3270 }
3271
3272 void build_nat_area_bitmap(struct f2fs_sb_info *sbi)
3273 {
3274         struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
3275         struct f2fs_journal *journal = &curseg->sum_blk->journal;
3276         struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
3277         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
3278         struct f2fs_nm_info *nm_i = NM_I(sbi);
3279         struct f2fs_nat_block *nat_block;
3280         struct node_info ni;
3281         u32 nid, nr_nat_blks;
3282         pgoff_t block_off;
3283         pgoff_t block_addr;
3284         int seg_off;
3285         int ret;
3286         unsigned int i;
3287
3288         nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
3289         ASSERT(nat_block);
3290
3291         /* Alloc & build nat entry bitmap */
3292         nr_nat_blks = (get_sb(segment_count_nat) / 2) <<
3293                                         sbi->log_blocks_per_seg;
3294
3295         fsck->nr_nat_entries = nr_nat_blks * NAT_ENTRY_PER_BLOCK;
3296         fsck->nat_area_bitmap_sz = (fsck->nr_nat_entries + 7) / 8;
3297         fsck->nat_area_bitmap = calloc(fsck->nat_area_bitmap_sz, 1);
3298         ASSERT(fsck->nat_area_bitmap);
3299
3300         fsck->entries = calloc(sizeof(struct f2fs_nat_entry),
3301                                         fsck->nr_nat_entries);
3302         ASSERT(fsck->entries);
3303
3304         for (block_off = 0; block_off < nr_nat_blks; block_off++) {
3305
3306                 seg_off = block_off >> sbi->log_blocks_per_seg;
3307                 block_addr = (pgoff_t)(nm_i->nat_blkaddr +
3308                         (seg_off << sbi->log_blocks_per_seg << 1) +
3309                         (block_off & ((1 << sbi->log_blocks_per_seg) - 1)));
3310
3311                 if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
3312                         block_addr += sbi->blocks_per_seg;
3313
3314                 ret = dev_read_block(nat_block, block_addr);
3315                 ASSERT(ret >= 0);
3316
3317                 nid = block_off * NAT_ENTRY_PER_BLOCK;
3318                 for (i = 0; i < NAT_ENTRY_PER_BLOCK; i++) {
3319                         ni.nid = nid + i;
3320
3321                         if ((nid + i) == F2FS_NODE_INO(sbi) ||
3322                                         (nid + i) == F2FS_META_INO(sbi)) {
3323                                 /*
3324                                  * block_addr of node/meta inode should be 0x1.
3325                                  * Set this bit, and fsck_verify will fix it.
3326                                  */
3327                                 if (le32_to_cpu(nat_block->entries[i].block_addr) != 0x1) {
3328                                         ASSERT_MSG("\tError: ino[0x%x] block_addr[0x%x] is invalid\n",
3329                                                         nid + i, le32_to_cpu(nat_block->entries[i].block_addr));
3330                                         f2fs_set_bit(nid + i, fsck->nat_area_bitmap);
3331                                 }
3332                                 continue;
3333                         }
3334
3335                         node_info_from_raw_nat(&ni, &nat_block->entries[i]);
3336                         if (ni.blk_addr == 0x0)
3337                                 continue;
3338                         if (ni.ino == 0x0) {
3339                                 ASSERT_MSG("\tError: ino[0x%8x] or blk_addr[0x%16x]"
3340                                         " is invalid\n", ni.ino, ni.blk_addr);
3341                         }
3342                         if (ni.ino == (nid + i)) {
3343                                 fsck->nat_valid_inode_cnt++;
3344                                 DBG(3, "ino[0x%8x] maybe is inode\n", ni.ino);
3345                         }
3346                         if (nid + i == 0) {
3347                                 /*
3348                                  * nat entry [0] must be null.  If
3349                                  * it is corrupted, set its bit in
3350                                  * nat_area_bitmap, fsck_verify will
3351                                  * nullify it
3352                                  */
3353                                 ASSERT_MSG("Invalid nat entry[0]: "
3354                                         "blk_addr[0x%x]\n", ni.blk_addr);
3355                                 fsck->chk.valid_nat_entry_cnt--;
3356                         }
3357
3358                         DBG(3, "nid[0x%8x] addr[0x%16x] ino[0x%8x]\n",
3359                                 nid + i, ni.blk_addr, ni.ino);
3360                         f2fs_set_bit(nid + i, fsck->nat_area_bitmap);
3361                         fsck->chk.valid_nat_entry_cnt++;
3362
3363                         fsck->entries[nid + i] = nat_block->entries[i];
3364                 }
3365         }
3366
3367         /* Traverse nat journal, update the corresponding entries */
3368         for (i = 0; i < nats_in_cursum(journal); i++) {
3369                 struct f2fs_nat_entry raw_nat;
3370                 nid = le32_to_cpu(nid_in_journal(journal, i));
3371                 ni.nid = nid;
3372
3373                 DBG(3, "==> Found nid [0x%x] in nat cache, update it\n", nid);
3374
3375                 /* Clear the original bit and count */
3376                 if (fsck->entries[nid].block_addr != 0x0) {
3377                         fsck->chk.valid_nat_entry_cnt--;
3378                         f2fs_clear_bit(nid, fsck->nat_area_bitmap);
3379                         if (fsck->entries[nid].ino == nid)
3380                                 fsck->nat_valid_inode_cnt--;
3381                 }
3382
3383                 /* Use nat entries in journal */
3384                 memcpy(&raw_nat, &nat_in_journal(journal, i),
3385                                         sizeof(struct f2fs_nat_entry));
3386                 node_info_from_raw_nat(&ni, &raw_nat);
3387                 if (ni.blk_addr != 0x0) {
3388                         if (ni.ino == 0x0)
3389                                 ASSERT_MSG("\tError: ino[0x%8x] or blk_addr[0x%16x]"
3390                                         " is invalid\n", ni.ino, ni.blk_addr);
3391                         if (ni.ino == nid) {
3392                                 fsck->nat_valid_inode_cnt++;
3393                                 DBG(3, "ino[0x%8x] maybe is inode\n", ni.ino);
3394                         }
3395                         f2fs_set_bit(nid, fsck->nat_area_bitmap);
3396                         fsck->chk.valid_nat_entry_cnt++;
3397                         DBG(3, "nid[0x%x] in nat cache\n", nid);
3398                 }
3399                 fsck->entries[nid] = raw_nat;
3400         }
3401         free(nat_block);
3402
3403         DBG(1, "valid nat entries (block_addr != 0x0) [0x%8x : %u]\n",
3404                         fsck->chk.valid_nat_entry_cnt,
3405                         fsck->chk.valid_nat_entry_cnt);
3406 }
3407
3408 static int check_sector_size(struct f2fs_super_block *sb)
3409 {
3410         uint32_t log_sectorsize, log_sectors_per_block;
3411
3412         log_sectorsize = log_base_2(c.sector_size);
3413         log_sectors_per_block = log_base_2(c.sectors_per_blk);
3414
3415         if (log_sectorsize == get_sb(log_sectorsize) &&
3416                         log_sectors_per_block == get_sb(log_sectors_per_block))
3417                 return 0;
3418
3419         set_sb(log_sectorsize, log_sectorsize);
3420         set_sb(log_sectors_per_block, log_sectors_per_block);
3421
3422         update_superblock(sb, SB_MASK_ALL);
3423         return 0;
3424 }
3425
3426 static int tune_sb_features(struct f2fs_sb_info *sbi)
3427 {
3428         int sb_changed = 0;
3429         struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
3430
3431         if (!(sb->feature & cpu_to_le32(F2FS_FEATURE_ENCRYPT)) &&
3432                         c.feature & cpu_to_le32(F2FS_FEATURE_ENCRYPT)) {
3433                 sb->feature |= cpu_to_le32(F2FS_FEATURE_ENCRYPT);
3434                 MSG(0, "Info: Set Encryption feature\n");
3435                 sb_changed = 1;
3436         }
3437         if (!(sb->feature & cpu_to_le32(F2FS_FEATURE_CASEFOLD)) &&
3438                 c.feature & cpu_to_le32(F2FS_FEATURE_CASEFOLD)) {
3439                 if (!c.s_encoding) {
3440                         ERR_MSG("ERROR: Must specify encoding to enable casefolding.\n");
3441                         return -1;
3442                 }
3443                 sb->feature |= cpu_to_le32(F2FS_FEATURE_CASEFOLD);
3444                 MSG(0, "Info: Set Casefold feature\n");
3445                 sb_changed = 1;
3446         }
3447         /* TODO: quota needs to allocate inode numbers */
3448
3449         c.feature = sb->feature;
3450         if (!sb_changed)
3451                 return 0;
3452
3453         update_superblock(sb, SB_MASK_ALL);
3454         return 0;
3455 }
3456
3457 static struct fsync_inode_entry *get_fsync_inode(struct list_head *head,
3458                                                                 nid_t ino)
3459 {
3460         struct fsync_inode_entry *entry;
3461
3462         list_for_each_entry(entry, head, list)
3463                 if (entry->ino == ino)
3464                         return entry;
3465
3466         return NULL;
3467 }
3468
3469 static struct fsync_inode_entry *add_fsync_inode(struct list_head *head,
3470                                                                 nid_t ino)
3471 {
3472         struct fsync_inode_entry *entry;
3473
3474         entry = calloc(sizeof(struct fsync_inode_entry), 1);
3475         if (!entry)
3476                 return NULL;
3477         entry->ino = ino;
3478         list_add_tail(&entry->list, head);
3479         return entry;
3480 }
3481
3482 static void del_fsync_inode(struct fsync_inode_entry *entry)
3483 {
3484         list_del(&entry->list);
3485         free(entry);
3486 }
3487
3488 static void destroy_fsync_dnodes(struct list_head *head)
3489 {
3490         struct fsync_inode_entry *entry, *tmp;
3491
3492         list_for_each_entry_safe(entry, tmp, head, list)
3493                 del_fsync_inode(entry);
3494 }
3495
3496 static int find_fsync_inode(struct f2fs_sb_info *sbi, struct list_head *head)
3497 {
3498         struct curseg_info *curseg;
3499         struct f2fs_node *node_blk;
3500         block_t blkaddr;
3501         unsigned int loop_cnt = 0;
3502         unsigned int free_blocks = MAIN_SEGS(sbi) * sbi->blocks_per_seg -
3503                                                 sbi->total_valid_block_count;
3504         int err = 0;
3505
3506         /* get node pages in the current segment */
3507         curseg = CURSEG_I(sbi, CURSEG_WARM_NODE);
3508         blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
3509
3510         node_blk = calloc(F2FS_BLKSIZE, 1);
3511         ASSERT(node_blk);
3512
3513         while (1) {
3514                 struct fsync_inode_entry *entry;
3515
3516                 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR))
3517                         break;
3518
3519                 err = dev_read_block(node_blk, blkaddr);
3520                 if (err)
3521                         break;
3522
3523                 if (!is_recoverable_dnode(sbi, node_blk))
3524                         break;
3525
3526                 if (!is_fsync_dnode(node_blk))
3527                         goto next;
3528
3529                 entry = get_fsync_inode(head, ino_of_node(node_blk));
3530                 if (!entry) {
3531                         entry = add_fsync_inode(head, ino_of_node(node_blk));
3532                         if (!entry) {
3533                                 err = -1;
3534                                 break;
3535                         }
3536                 }
3537                 entry->blkaddr = blkaddr;
3538
3539                 if (IS_INODE(node_blk) && is_dent_dnode(node_blk))
3540                         entry->last_dentry = blkaddr;
3541 next:
3542                 /* sanity check in order to detect looped node chain */
3543                 if (++loop_cnt >= free_blocks ||
3544                         blkaddr == next_blkaddr_of_node(node_blk)) {
3545                         MSG(0, "\tdetect looped node chain, blkaddr:%u, next:%u\n",
3546                                     blkaddr,
3547                                     next_blkaddr_of_node(node_blk));
3548                         err = -1;
3549                         break;
3550                 }
3551
3552                 blkaddr = next_blkaddr_of_node(node_blk);
3553         }
3554
3555         free(node_blk);
3556         return err;
3557 }
3558
3559 static int do_record_fsync_data(struct f2fs_sb_info *sbi,
3560                                         struct f2fs_node *node_blk,
3561                                         block_t blkaddr)
3562 {
3563         unsigned int segno, offset;
3564         struct seg_entry *se;
3565         unsigned int ofs_in_node = 0;
3566         unsigned int start, end;
3567         int err = 0, recorded = 0;
3568
3569         segno = GET_SEGNO(sbi, blkaddr);
3570         se = get_seg_entry(sbi, segno);
3571         offset = OFFSET_IN_SEG(sbi, blkaddr);
3572
3573         if (f2fs_test_bit(offset, (char *)se->cur_valid_map)) {
3574                 ASSERT(0);
3575                 return -1;
3576         }
3577         if (f2fs_test_bit(offset, (char *)se->ckpt_valid_map)) {
3578                 ASSERT(0);
3579                 return -1;
3580         }
3581
3582         if (!se->ckpt_valid_blocks)
3583                 se->ckpt_type = CURSEG_WARM_NODE;
3584
3585         se->ckpt_valid_blocks++;
3586         f2fs_set_bit(offset, (char *)se->ckpt_valid_map);
3587
3588         MSG(1, "do_record_fsync_data: [node] ino = %u, nid = %u, blkaddr = %u\n",
3589             ino_of_node(node_blk), ofs_of_node(node_blk), blkaddr);
3590
3591         /* inline data */
3592         if (IS_INODE(node_blk) && (node_blk->i.i_inline & F2FS_INLINE_DATA))
3593                 return 0;
3594         /* xattr node */
3595         if (ofs_of_node(node_blk) == XATTR_NODE_OFFSET)
3596                 return 0;
3597
3598         /* step 3: recover data indices */
3599         start = start_bidx_of_node(ofs_of_node(node_blk), node_blk);
3600         end = start + ADDRS_PER_PAGE(sbi, node_blk, NULL);
3601
3602         for (; start < end; start++, ofs_in_node++) {
3603                 blkaddr = datablock_addr(node_blk, ofs_in_node);
3604
3605                 if (!is_valid_data_blkaddr(blkaddr))
3606                         continue;
3607
3608                 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR)) {
3609                         err = -1;
3610                         goto out;
3611                 }
3612
3613                 segno = GET_SEGNO(sbi, blkaddr);
3614                 se = get_seg_entry(sbi, segno);
3615                 offset = OFFSET_IN_SEG(sbi, blkaddr);
3616
3617                 if (f2fs_test_bit(offset, (char *)se->cur_valid_map))
3618                         continue;
3619                 if (f2fs_test_bit(offset, (char *)se->ckpt_valid_map))
3620                         continue;
3621
3622                 if (!se->ckpt_valid_blocks)
3623                         se->ckpt_type = CURSEG_WARM_DATA;
3624
3625                 se->ckpt_valid_blocks++;
3626                 f2fs_set_bit(offset, (char *)se->ckpt_valid_map);
3627
3628                 MSG(1, "do_record_fsync_data: [data] ino = %u, nid = %u, blkaddr = %u\n",
3629                     ino_of_node(node_blk), ofs_of_node(node_blk), blkaddr);
3630
3631                 recorded++;
3632         }
3633 out:
3634         MSG(1, "recover_data: ino = %u, nid = %u, recorded = %d, err = %d\n",
3635                     ino_of_node(node_blk), ofs_of_node(node_blk),
3636                     recorded, err);
3637         return err;
3638 }
3639
3640 static int traverse_dnodes(struct f2fs_sb_info *sbi,
3641                                 struct list_head *inode_list)
3642 {
3643         struct curseg_info *curseg;
3644         struct f2fs_node *node_blk;
3645         block_t blkaddr;
3646         int err = 0;
3647
3648         /* get node pages in the current segment */
3649         curseg = CURSEG_I(sbi, CURSEG_WARM_NODE);
3650         blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
3651
3652         node_blk = calloc(F2FS_BLKSIZE, 1);
3653         ASSERT(node_blk);
3654
3655         while (1) {
3656                 struct fsync_inode_entry *entry;
3657
3658                 if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR))
3659                         break;
3660
3661                 err = dev_read_block(node_blk, blkaddr);
3662                 if (err)
3663                         break;
3664
3665                 if (!is_recoverable_dnode(sbi, node_blk))
3666                         break;
3667
3668                 entry = get_fsync_inode(inode_list,
3669                                         ino_of_node(node_blk));
3670                 if (!entry)
3671                         goto next;
3672
3673                 err = do_record_fsync_data(sbi, node_blk, blkaddr);
3674                 if (err)
3675                         break;
3676
3677                 if (entry->blkaddr == blkaddr)
3678                         del_fsync_inode(entry);
3679 next:
3680                 blkaddr = next_blkaddr_of_node(node_blk);
3681         }
3682
3683         free(node_blk);
3684         return err;
3685 }
3686
3687 static int record_fsync_data(struct f2fs_sb_info *sbi)
3688 {
3689         struct list_head inode_list = LIST_HEAD_INIT(inode_list);
3690         int ret;
3691
3692         if (!need_fsync_data_record(sbi))
3693                 return 0;
3694
3695         ret = find_fsync_inode(sbi, &inode_list);
3696         if (ret)
3697                 goto out;
3698
3699         ret = late_build_segment_manager(sbi);
3700         if (ret < 0) {
3701                 ERR_MSG("late_build_segment_manager failed\n");
3702                 goto out;
3703         }
3704
3705         ret = traverse_dnodes(sbi, &inode_list);
3706 out:
3707         destroy_fsync_dnodes(&inode_list);
3708         return ret;
3709 }
3710
3711 int f2fs_do_mount(struct f2fs_sb_info *sbi)
3712 {
3713         struct f2fs_checkpoint *cp = NULL;
3714         struct f2fs_super_block *sb = NULL;
3715         int ret;
3716
3717         sbi->active_logs = NR_CURSEG_TYPE;
3718         ret = validate_super_block(sbi, SB0_ADDR);
3719         if (ret) {
3720                 ret = validate_super_block(sbi, SB1_ADDR);
3721                 if (ret)
3722                         return -1;
3723         }
3724         sb = F2FS_RAW_SUPER(sbi);
3725
3726         ret = check_sector_size(sb);
3727         if (ret)
3728                 return -1;
3729
3730         print_raw_sb_info(sb);
3731
3732         init_sb_info(sbi);
3733
3734         ret = get_valid_checkpoint(sbi);
3735         if (ret) {
3736                 ERR_MSG("Can't find valid checkpoint\n");
3737                 return -1;
3738         }
3739
3740         c.bug_on = 0;
3741
3742         if (sanity_check_ckpt(sbi)) {
3743                 ERR_MSG("Checkpoint is polluted\n");
3744                 return -1;
3745         }
3746         cp = F2FS_CKPT(sbi);
3747
3748         if (c.func != FSCK && c.func != DUMP &&
3749                 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
3750                 ERR_MSG("Mount unclean image to replay log first\n");
3751                 return -1;
3752         }
3753
3754         if (c.func == FSCK) {
3755 #if defined(__APPLE__)
3756                 if (!c.no_kernel_check &&
3757                         memcmp(c.sb_version, c.version, VERSION_NAME_LEN)) {
3758                         c.auto_fix = 0;
3759                         c.fix_on = 1;
3760                         memcpy(sbi->raw_super->version,
3761                                         c.version, VERSION_NAME_LEN);
3762                         update_superblock(sbi->raw_super, SB_MASK_ALL);
3763                 }
3764 #else
3765                 if (!c.no_kernel_check) {
3766                         u32 prev_time, cur_time, time_diff;
3767                         __le32 *ver_ts_ptr = (__le32 *)(sbi->raw_super->version
3768                                                 + VERSION_NAME_LEN);
3769
3770                         cur_time = (u32)get_cp(elapsed_time);
3771                         prev_time = le32_to_cpu(*ver_ts_ptr);
3772
3773                         MSG(0, "Info: version timestamp cur: %u, prev: %u\n",
3774                                         cur_time, prev_time);
3775                         if (!memcmp(c.sb_version, c.version,
3776                                                 VERSION_NAME_LEN)) {
3777                                 /* valid prev_time */
3778                                 if (prev_time != 0 && cur_time > prev_time) {
3779                                         time_diff = cur_time - prev_time;
3780                                         if (time_diff < CHECK_PERIOD)
3781                                                 goto out;
3782                                         c.auto_fix = 0;
3783                                         c.fix_on = 1;
3784                                 }
3785                         } else {
3786                                 memcpy(sbi->raw_super->version,
3787                                                 c.version, VERSION_NAME_LEN);
3788                         }
3789
3790                         *ver_ts_ptr = cpu_to_le32(cur_time);
3791                         update_superblock(sbi->raw_super, SB_MASK_ALL);
3792                 }
3793 #endif
3794         }
3795 out:
3796         print_ckpt_info(sbi);
3797
3798         if (c.quota_fix) {
3799                 if (get_cp(ckpt_flags) & CP_QUOTA_NEED_FSCK_FLAG)
3800                         c.fix_on = 1;
3801         }
3802         if (c.layout)
3803                 return 1;
3804
3805         if (tune_sb_features(sbi))
3806                 return -1;
3807
3808         /* precompute checksum seed for metadata */
3809         if (c.feature & cpu_to_le32(F2FS_FEATURE_INODE_CHKSUM))
3810                 c.chksum_seed = f2fs_cal_crc32(~0, sb->uuid, sizeof(sb->uuid));
3811
3812         sbi->total_valid_node_count = get_cp(valid_node_count);
3813         sbi->total_valid_inode_count = get_cp(valid_inode_count);
3814         sbi->user_block_count = get_cp(user_block_count);
3815         sbi->total_valid_block_count = get_cp(valid_block_count);
3816         sbi->last_valid_block_count = sbi->total_valid_block_count;
3817         sbi->alloc_valid_block_count = 0;
3818
3819         if (early_build_segment_manager(sbi)) {
3820                 ERR_MSG("early_build_segment_manager failed\n");
3821                 return -1;
3822         }
3823
3824         if (build_node_manager(sbi)) {
3825                 ERR_MSG("build_node_manager failed\n");
3826                 return -1;
3827         }
3828
3829         if (record_fsync_data(sbi)) {
3830                 ERR_MSG("record_fsync_data failed\n");
3831                 return -1;
3832         }
3833
3834         if (!f2fs_should_proceed(sb, get_cp(ckpt_flags)))
3835                 return 1;
3836
3837         if (late_build_segment_manager(sbi) < 0) {
3838                 ERR_MSG("late_build_segment_manager failed\n");
3839                 return -1;
3840         }
3841
3842         if (f2fs_late_init_nid_bitmap(sbi)) {
3843                 ERR_MSG("f2fs_late_init_nid_bitmap failed\n");
3844                 return -1;
3845         }
3846
3847         /* Check nat_bits */
3848         if (c.func == FSCK && is_set_ckpt_flags(cp, CP_NAT_BITS_FLAG)) {
3849                 if (check_nat_bits(sbi, sb, cp) && c.fix_on)
3850                         write_nat_bits(sbi, sb, cp, sbi->cur_cp);
3851         }
3852         return 0;
3853 }
3854
3855 void f2fs_do_umount(struct f2fs_sb_info *sbi)
3856 {
3857         struct sit_info *sit_i = SIT_I(sbi);
3858         struct f2fs_sm_info *sm_i = SM_I(sbi);
3859         struct f2fs_nm_info *nm_i = NM_I(sbi);
3860         unsigned int i;
3861
3862         /* free nm_info */
3863         if (c.func == SLOAD || c.func == FSCK)
3864                 free(nm_i->nid_bitmap);
3865         free(nm_i->nat_bitmap);
3866         free(sbi->nm_info);
3867
3868         /* free sit_info */
3869         free(sit_i->bitmap);
3870         free(sit_i->sit_bitmap);
3871         free(sit_i->sentries);
3872         free(sm_i->sit_info);
3873
3874         /* free sm_info */
3875         for (i = 0; i < NR_CURSEG_TYPE; i++)
3876                 free(sm_i->curseg_array[i].sum_blk);
3877
3878         free(sm_i->curseg_array);
3879         free(sbi->sm_info);
3880
3881         free(sbi->ckpt);
3882         free(sbi->raw_super);
3883 }
3884
3885 #ifdef WITH_ANDROID
3886 int f2fs_sparse_initialize_meta(struct f2fs_sb_info *sbi)
3887 {
3888         struct f2fs_super_block *sb = sbi->raw_super;
3889         uint32_t sit_seg_count, sit_size;
3890         uint32_t nat_seg_count, nat_size;
3891         uint64_t sit_seg_addr, nat_seg_addr, payload_addr;
3892         uint32_t seg_size = 1 << get_sb(log_blocks_per_seg);
3893         int ret;
3894
3895         if (!c.sparse_mode)
3896                 return 0;
3897
3898         sit_seg_addr = get_sb(sit_blkaddr);
3899         sit_seg_count = get_sb(segment_count_sit);
3900         sit_size = sit_seg_count * seg_size;
3901
3902         DBG(1, "\tSparse: filling sit area at block offset: 0x%08"PRIx64" len: %u\n",
3903                                                         sit_seg_addr, sit_size);
3904         ret = dev_fill(NULL, sit_seg_addr * F2FS_BLKSIZE,
3905                                         sit_size * F2FS_BLKSIZE);
3906         if (ret) {
3907                 MSG(1, "\tError: While zeroing out the sit area "
3908                                 "on disk!!!\n");
3909                 return -1;
3910         }
3911
3912         nat_seg_addr = get_sb(nat_blkaddr);
3913         nat_seg_count = get_sb(segment_count_nat);
3914         nat_size = nat_seg_count * seg_size;
3915
3916         DBG(1, "\tSparse: filling nat area at block offset 0x%08"PRIx64" len: %u\n",
3917                                                         nat_seg_addr, nat_size);
3918         ret = dev_fill(NULL, nat_seg_addr * F2FS_BLKSIZE,
3919                                         nat_size * F2FS_BLKSIZE);
3920         if (ret) {
3921                 MSG(1, "\tError: While zeroing out the nat area "
3922                                 "on disk!!!\n");
3923                 return -1;
3924         }
3925
3926         payload_addr = get_sb(segment0_blkaddr) + 1;
3927
3928         DBG(1, "\tSparse: filling bitmap area at block offset 0x%08"PRIx64" len: %u\n",
3929                                         payload_addr, get_sb(cp_payload));
3930         ret = dev_fill(NULL, payload_addr * F2FS_BLKSIZE,
3931                                         get_sb(cp_payload) * F2FS_BLKSIZE);
3932         if (ret) {
3933                 MSG(1, "\tError: While zeroing out the nat/sit bitmap area "
3934                                 "on disk!!!\n");
3935                 return -1;
3936         }
3937
3938         payload_addr += seg_size;
3939
3940         DBG(1, "\tSparse: filling bitmap area at block offset 0x%08"PRIx64" len: %u\n",
3941                                         payload_addr, get_sb(cp_payload));
3942         ret = dev_fill(NULL, payload_addr * F2FS_BLKSIZE,
3943                                         get_sb(cp_payload) * F2FS_BLKSIZE);
3944         if (ret) {
3945                 MSG(1, "\tError: While zeroing out the nat/sit bitmap area "
3946                                 "on disk!!!\n");
3947                 return -1;
3948         }
3949         return 0;
3950 }
3951 #else
3952 int f2fs_sparse_initialize_meta(struct f2fs_sb_info *sbi) { return 0; }
3953 #endif