linux/fs/ubifs/replay.c
<<
>>
Prefs
   1/*
   2 * This file is part of UBIFS.
   3 *
   4 * Copyright (C) 2006-2008 Nokia Corporation.
   5 *
   6 * This program is free software; you can redistribute it and/or modify it
   7 * under the terms of the GNU General Public License version 2 as published by
   8 * the Free Software Foundation.
   9 *
  10 * This program is distributed in the hope that it will be useful, but WITHOUT
  11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  13 * more details.
  14 *
  15 * You should have received a copy of the GNU General Public License along with
  16 * this program; if not, write to the Free Software Foundation, Inc., 51
  17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  18 *
  19 * Authors: Adrian Hunter
  20 *          Artem Bityutskiy (Битюцкий Артём)
  21 */
  22
  23/*
  24 * This file contains journal replay code. It runs when the file-system is being
  25 * mounted and requires no locking.
  26 *
  27 * The larger is the journal, the longer it takes to scan it, so the longer it
  28 * takes to mount UBIFS. This is why the journal has limited size which may be
  29 * changed depending on the system requirements. But a larger journal gives
  30 * faster I/O speed because it writes the index less frequently. So this is a
  31 * trade-off. Also, the journal is indexed by the in-memory index (TNC), so the
  32 * larger is the journal, the more memory its index may consume.
  33 */
  34
  35#include "ubifs.h"
  36#include <linux/list_sort.h>
  37
  38/**
  39 * struct replay_entry - replay list entry.
  40 * @lnum: logical eraseblock number of the node
  41 * @offs: node offset
  42 * @len: node length
  43 * @deletion: non-zero if this entry corresponds to a node deletion
  44 * @sqnum: node sequence number
  45 * @list: links the replay list
  46 * @key: node key
  47 * @nm: directory entry name
  48 * @old_size: truncation old size
  49 * @new_size: truncation new size
  50 *
  51 * The replay process first scans all buds and builds the replay list, then
  52 * sorts the replay list in nodes sequence number order, and then inserts all
  53 * the replay entries to the TNC.
  54 */
  55struct replay_entry {
  56        int lnum;
  57        int offs;
  58        int len;
  59        unsigned int deletion:1;
  60        unsigned long long sqnum;
  61        struct list_head list;
  62        union ubifs_key key;
  63        union {
  64                struct qstr nm;
  65                struct {
  66                        loff_t old_size;
  67                        loff_t new_size;
  68                };
  69        };
  70};
  71
  72/**
  73 * struct bud_entry - entry in the list of buds to replay.
  74 * @list: next bud in the list
  75 * @bud: bud description object
  76 * @sqnum: reference node sequence number
  77 * @free: free bytes in the bud
  78 * @dirty: dirty bytes in the bud
  79 */
  80struct bud_entry {
  81        struct list_head list;
  82        struct ubifs_bud *bud;
  83        unsigned long long sqnum;
  84        int free;
  85        int dirty;
  86};
  87
  88/**
  89 * set_bud_lprops - set free and dirty space used by a bud.
  90 * @c: UBIFS file-system description object
  91 * @b: bud entry which describes the bud
  92 *
  93 * This function makes sure the LEB properties of bud @b are set correctly
  94 * after the replay. Returns zero in case of success and a negative error code
  95 * in case of failure.
  96 */
  97static int set_bud_lprops(struct ubifs_info *c, struct bud_entry *b)
  98{
  99        const struct ubifs_lprops *lp;
 100        int err = 0, dirty;
 101
 102        ubifs_get_lprops(c);
 103
 104        lp = ubifs_lpt_lookup_dirty(c, b->bud->lnum);
 105        if (IS_ERR(lp)) {
 106                err = PTR_ERR(lp);
 107                goto out;
 108        }
 109
 110        dirty = lp->dirty;
 111        if (b->bud->start == 0 && (lp->free != c->leb_size || lp->dirty != 0)) {
 112                /*
 113                 * The LEB was added to the journal with a starting offset of
 114                 * zero which means the LEB must have been empty. The LEB
 115                 * property values should be @lp->free == @c->leb_size and
 116                 * @lp->dirty == 0, but that is not the case. The reason is that
 117                 * the LEB had been garbage collected before it became the bud,
 118                 * and there was not commit inbetween. The garbage collector
 119                 * resets the free and dirty space without recording it
 120                 * anywhere except lprops, so if there was no commit then
 121                 * lprops does not have that information.
 122                 *
 123                 * We do not need to adjust free space because the scan has told
 124                 * us the exact value which is recorded in the replay entry as
 125                 * @b->free.
 126                 *
 127                 * However we do need to subtract from the dirty space the
 128                 * amount of space that the garbage collector reclaimed, which
 129                 * is the whole LEB minus the amount of space that was free.
 130                 */
 131                dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
 132                        lp->free, lp->dirty);
 133                dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
 134                        lp->free, lp->dirty);
 135                dirty -= c->leb_size - lp->free;
 136                /*
 137                 * If the replay order was perfect the dirty space would now be
 138                 * zero. The order is not perfect because the journal heads
 139                 * race with each other. This is not a problem but is does mean
 140                 * that the dirty space may temporarily exceed c->leb_size
 141                 * during the replay.
 142                 */
 143                if (dirty != 0)
 144                        dbg_mnt("LEB %d lp: %d free %d dirty replay: %d free %d dirty",
 145                                b->bud->lnum, lp->free, lp->dirty, b->free,
 146                                b->dirty);
 147        }
 148        lp = ubifs_change_lp(c, lp, b->free, dirty + b->dirty,
 149                             lp->flags | LPROPS_TAKEN, 0);
 150        if (IS_ERR(lp)) {
 151                err = PTR_ERR(lp);
 152                goto out;
 153        }
 154
 155        /* Make sure the journal head points to the latest bud */
 156        err = ubifs_wbuf_seek_nolock(&c->jheads[b->bud->jhead].wbuf,
 157                                     b->bud->lnum, c->leb_size - b->free);
 158
 159out:
 160        ubifs_release_lprops(c);
 161        return err;
 162}
 163
 164/**
 165 * set_buds_lprops - set free and dirty space for all replayed buds.
 166 * @c: UBIFS file-system description object
 167 *
 168 * This function sets LEB properties for all replayed buds. Returns zero in
 169 * case of success and a negative error code in case of failure.
 170 */
 171static int set_buds_lprops(struct ubifs_info *c)
 172{
 173        struct bud_entry *b;
 174        int err;
 175
 176        list_for_each_entry(b, &c->replay_buds, list) {
 177                err = set_bud_lprops(c, b);
 178                if (err)
 179                        return err;
 180        }
 181
 182        return 0;
 183}
 184
 185/**
 186 * trun_remove_range - apply a replay entry for a truncation to the TNC.
 187 * @c: UBIFS file-system description object
 188 * @r: replay entry of truncation
 189 */
 190static int trun_remove_range(struct ubifs_info *c, struct replay_entry *r)
 191{
 192        unsigned min_blk, max_blk;
 193        union ubifs_key min_key, max_key;
 194        ino_t ino;
 195
 196        min_blk = r->new_size / UBIFS_BLOCK_SIZE;
 197        if (r->new_size & (UBIFS_BLOCK_SIZE - 1))
 198                min_blk += 1;
 199
 200        max_blk = r->old_size / UBIFS_BLOCK_SIZE;
 201        if ((r->old_size & (UBIFS_BLOCK_SIZE - 1)) == 0)
 202                max_blk -= 1;
 203
 204        ino = key_inum(c, &r->key);
 205
 206        data_key_init(c, &min_key, ino, min_blk);
 207        data_key_init(c, &max_key, ino, max_blk);
 208
 209        return ubifs_tnc_remove_range(c, &min_key, &max_key);
 210}
 211
 212/**
 213 * apply_replay_entry - apply a replay entry to the TNC.
 214 * @c: UBIFS file-system description object
 215 * @r: replay entry to apply
 216 *
 217 * Apply a replay entry to the TNC.
 218 */
 219static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r)
 220{
 221        int err;
 222
 223        dbg_mntk(&r->key, "LEB %d:%d len %d deletion %d sqnum %llu key ",
 224                 r->lnum, r->offs, r->len, r->deletion, r->sqnum);
 225
 226        /* Set c->replay_sqnum to help deal with dangling branches. */
 227        c->replay_sqnum = r->sqnum;
 228
 229        if (is_hash_key(c, &r->key)) {
 230                if (r->deletion)
 231                        err = ubifs_tnc_remove_nm(c, &r->key, &r->nm);
 232                else
 233                        err = ubifs_tnc_add_nm(c, &r->key, r->lnum, r->offs,
 234                                               r->len, &r->nm);
 235        } else {
 236                if (r->deletion)
 237                        switch (key_type(c, &r->key)) {
 238                        case UBIFS_INO_KEY:
 239                        {
 240                                ino_t inum = key_inum(c, &r->key);
 241
 242                                err = ubifs_tnc_remove_ino(c, inum);
 243                                break;
 244                        }
 245                        case UBIFS_TRUN_KEY:
 246                                err = trun_remove_range(c, r);
 247                                break;
 248                        default:
 249                                err = ubifs_tnc_remove(c, &r->key);
 250                                break;
 251                        }
 252                else
 253                        err = ubifs_tnc_add(c, &r->key, r->lnum, r->offs,
 254                                            r->len);
 255                if (err)
 256                        return err;
 257
 258                if (c->need_recovery)
 259                        err = ubifs_recover_size_accum(c, &r->key, r->deletion,
 260                                                       r->new_size);
 261        }
 262
 263        return err;
 264}
 265
 266/**
 267 * replay_entries_cmp - compare 2 replay entries.
 268 * @priv: UBIFS file-system description object
 269 * @a: first replay entry
 270 * @a: second replay entry
 271 *
 272 * This is a comparios function for 'list_sort()' which compares 2 replay
 273 * entries @a and @b by comparing their sequence numer.  Returns %1 if @a has
 274 * greater sequence number and %-1 otherwise.
 275 */
 276static int replay_entries_cmp(void *priv, struct list_head *a,
 277                              struct list_head *b)
 278{
 279        struct replay_entry *ra, *rb;
 280
 281        cond_resched();
 282        if (a == b)
 283                return 0;
 284
 285        ra = list_entry(a, struct replay_entry, list);
 286        rb = list_entry(b, struct replay_entry, list);
 287        ubifs_assert(ra->sqnum != rb->sqnum);
 288        if (ra->sqnum > rb->sqnum)
 289                return 1;
 290        return -1;
 291}
 292
 293/**
 294 * apply_replay_list - apply the replay list to the TNC.
 295 * @c: UBIFS file-system description object
 296 *
 297 * Apply all entries in the replay list to the TNC. Returns zero in case of
 298 * success and a negative error code in case of failure.
 299 */
 300static int apply_replay_list(struct ubifs_info *c)
 301{
 302        struct replay_entry *r;
 303        int err;
 304
 305        list_sort(c, &c->replay_list, &replay_entries_cmp);
 306
 307        list_for_each_entry(r, &c->replay_list, list) {
 308                cond_resched();
 309
 310                err = apply_replay_entry(c, r);
 311                if (err)
 312                        return err;
 313        }
 314
 315        return 0;
 316}
 317
 318/**
 319 * destroy_replay_list - destroy the replay.
 320 * @c: UBIFS file-system description object
 321 *
 322 * Destroy the replay list.
 323 */
 324static void destroy_replay_list(struct ubifs_info *c)
 325{
 326        struct replay_entry *r, *tmp;
 327
 328        list_for_each_entry_safe(r, tmp, &c->replay_list, list) {
 329                if (is_hash_key(c, &r->key))
 330                        kfree(r->nm.name);
 331                list_del(&r->list);
 332                kfree(r);
 333        }
 334}
 335
 336/**
 337 * insert_node - insert a node to the replay list
 338 * @c: UBIFS file-system description object
 339 * @lnum: node logical eraseblock number
 340 * @offs: node offset
 341 * @len: node length
 342 * @key: node key
 343 * @sqnum: sequence number
 344 * @deletion: non-zero if this is a deletion
 345 * @used: number of bytes in use in a LEB
 346 * @old_size: truncation old size
 347 * @new_size: truncation new size
 348 *
 349 * This function inserts a scanned non-direntry node to the replay list. The
 350 * replay list contains @struct replay_entry elements, and we sort this list in
 351 * sequence number order before applying it. The replay list is applied at the
 352 * very end of the replay process. Since the list is sorted in sequence number
 353 * order, the older modifications are applied first. This function returns zero
 354 * in case of success and a negative error code in case of failure.
 355 */
 356static int insert_node(struct ubifs_info *c, int lnum, int offs, int len,
 357                       union ubifs_key *key, unsigned long long sqnum,
 358                       int deletion, int *used, loff_t old_size,
 359                       loff_t new_size)
 360{
 361        struct replay_entry *r;
 362
 363        dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
 364
 365        if (key_inum(c, key) >= c->highest_inum)
 366                c->highest_inum = key_inum(c, key);
 367
 368        r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
 369        if (!r)
 370                return -ENOMEM;
 371
 372        if (!deletion)
 373                *used += ALIGN(len, 8);
 374        r->lnum = lnum;
 375        r->offs = offs;
 376        r->len = len;
 377        r->deletion = !!deletion;
 378        r->sqnum = sqnum;
 379        key_copy(c, key, &r->key);
 380        r->old_size = old_size;
 381        r->new_size = new_size;
 382
 383        list_add_tail(&r->list, &c->replay_list);
 384        return 0;
 385}
 386
 387/**
 388 * insert_dent - insert a directory entry node into the replay list.
 389 * @c: UBIFS file-system description object
 390 * @lnum: node logical eraseblock number
 391 * @offs: node offset
 392 * @len: node length
 393 * @key: node key
 394 * @name: directory entry name
 395 * @nlen: directory entry name length
 396 * @sqnum: sequence number
 397 * @deletion: non-zero if this is a deletion
 398 * @used: number of bytes in use in a LEB
 399 *
 400 * This function inserts a scanned directory entry node or an extended
 401 * attribute entry to the replay list. Returns zero in case of success and a
 402 * negative error code in case of failure.
 403 */
 404static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len,
 405                       union ubifs_key *key, const char *name, int nlen,
 406                       unsigned long long sqnum, int deletion, int *used)
 407{
 408        struct replay_entry *r;
 409        char *nbuf;
 410
 411        dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
 412        if (key_inum(c, key) >= c->highest_inum)
 413                c->highest_inum = key_inum(c, key);
 414
 415        r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
 416        if (!r)
 417                return -ENOMEM;
 418
 419        nbuf = kmalloc(nlen + 1, GFP_KERNEL);
 420        if (!nbuf) {
 421                kfree(r);
 422                return -ENOMEM;
 423        }
 424
 425        if (!deletion)
 426                *used += ALIGN(len, 8);
 427        r->lnum = lnum;
 428        r->offs = offs;
 429        r->len = len;
 430        r->deletion = !!deletion;
 431        r->sqnum = sqnum;
 432        key_copy(c, key, &r->key);
 433        r->nm.len = nlen;
 434        memcpy(nbuf, name, nlen);
 435        nbuf[nlen] = '\0';
 436        r->nm.name = nbuf;
 437
 438        list_add_tail(&r->list, &c->replay_list);
 439        return 0;
 440}
 441
 442/**
 443 * ubifs_validate_entry - validate directory or extended attribute entry node.
 444 * @c: UBIFS file-system description object
 445 * @dent: the node to validate
 446 *
 447 * This function validates directory or extended attribute entry node @dent.
 448 * Returns zero if the node is all right and a %-EINVAL if not.
 449 */
 450int ubifs_validate_entry(struct ubifs_info *c,
 451                         const struct ubifs_dent_node *dent)
 452{
 453        int key_type = key_type_flash(c, dent->key);
 454        int nlen = le16_to_cpu(dent->nlen);
 455
 456        if (le32_to_cpu(dent->ch.len) != nlen + UBIFS_DENT_NODE_SZ + 1 ||
 457            dent->type >= UBIFS_ITYPES_CNT ||
 458            nlen > UBIFS_MAX_NLEN || dent->name[nlen] != 0 ||
 459            strnlen(dent->name, nlen) != nlen ||
 460            le64_to_cpu(dent->inum) > MAX_INUM) {
 461                ubifs_err("bad %s node", key_type == UBIFS_DENT_KEY ?
 462                          "directory entry" : "extended attribute entry");
 463                return -EINVAL;
 464        }
 465
 466        if (key_type != UBIFS_DENT_KEY && key_type != UBIFS_XENT_KEY) {
 467                ubifs_err("bad key type %d", key_type);
 468                return -EINVAL;
 469        }
 470
 471        return 0;
 472}
 473
 474/**
 475 * is_last_bud - check if the bud is the last in the journal head.
 476 * @c: UBIFS file-system description object
 477 * @bud: bud description object
 478 *
 479 * This function checks if bud @bud is the last bud in its journal head. This
 480 * information is then used by 'replay_bud()' to decide whether the bud can
 481 * have corruptions or not. Indeed, only last buds can be corrupted by power
 482 * cuts. Returns %1 if this is the last bud, and %0 if not.
 483 */
 484static int is_last_bud(struct ubifs_info *c, struct ubifs_bud *bud)
 485{
 486        struct ubifs_jhead *jh = &c->jheads[bud->jhead];
 487        struct ubifs_bud *next;
 488        uint32_t data;
 489        int err;
 490
 491        if (list_is_last(&bud->list, &jh->buds_list))
 492                return 1;
 493
 494        /*
 495         * The following is a quirk to make sure we work correctly with UBIFS
 496         * images used with older UBIFS.
 497         *
 498         * Normally, the last bud will be the last in the journal head's list
 499         * of bud. However, there is one exception if the UBIFS image belongs
 500         * to older UBIFS. This is fairly unlikely: one would need to use old
 501         * UBIFS, then have a power cut exactly at the right point, and then
 502         * try to mount this image with new UBIFS.
 503         *
 504         * The exception is: it is possible to have 2 buds A and B, A goes
 505         * before B, and B is the last, bud B is contains no data, and bud A is
 506         * corrupted at the end. The reason is that in older versions when the
 507         * journal code switched the next bud (from A to B), it first added a
 508         * log reference node for the new bud (B), and only after this it
 509         * synchronized the write-buffer of current bud (A). But later this was
 510         * changed and UBIFS started to always synchronize the write-buffer of
 511         * the bud (A) before writing the log reference for the new bud (B).
 512         *
 513         * But because older UBIFS always synchronized A's write-buffer before
 514         * writing to B, we can recognize this exceptional situation but
 515         * checking the contents of bud B - if it is empty, then A can be
 516         * treated as the last and we can recover it.
 517         *
 518         * TODO: remove this piece of code in a couple of years (today it is
 519         * 16.05.2011).
 520         */
 521        next = list_entry(bud->list.next, struct ubifs_bud, list);
 522        if (!list_is_last(&next->list, &jh->buds_list))
 523                return 0;
 524
 525        err = ubifs_leb_read(c, next->lnum, (char *)&data, next->start, 4, 1);
 526        if (err)
 527                return 0;
 528
 529        return data == 0xFFFFFFFF;
 530}
 531
 532/**
 533 * replay_bud - replay a bud logical eraseblock.
 534 * @c: UBIFS file-system description object
 535 * @b: bud entry which describes the bud
 536 *
 537 * This function replays bud @bud, recovers it if needed, and adds all nodes
 538 * from this bud to the replay list. Returns zero in case of success and a
 539 * negative error code in case of failure.
 540 */
 541static int replay_bud(struct ubifs_info *c, struct bud_entry *b)
 542{
 543        int is_last = is_last_bud(c, b->bud);
 544        int err = 0, used = 0, lnum = b->bud->lnum, offs = b->bud->start;
 545        struct ubifs_scan_leb *sleb;
 546        struct ubifs_scan_node *snod;
 547
 548        dbg_mnt("replay bud LEB %d, head %d, offs %d, is_last %d",
 549                lnum, b->bud->jhead, offs, is_last);
 550
 551        if (c->need_recovery && is_last)
 552                /*
 553                 * Recover only last LEBs in the journal heads, because power
 554                 * cuts may cause corruptions only in these LEBs, because only
 555                 * these LEBs could possibly be written to at the power cut
 556                 * time.
 557                 */
 558                sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, b->bud->jhead);
 559        else
 560                sleb = ubifs_scan(c, lnum, offs, c->sbuf, 0);
 561        if (IS_ERR(sleb))
 562                return PTR_ERR(sleb);
 563
 564        /*
 565         * The bud does not have to start from offset zero - the beginning of
 566         * the 'lnum' LEB may contain previously committed data. One of the
 567         * things we have to do in replay is to correctly update lprops with
 568         * newer information about this LEB.
 569         *
 570         * At this point lprops thinks that this LEB has 'c->leb_size - offs'
 571         * bytes of free space because it only contain information about
 572         * committed data.
 573         *
 574         * But we know that real amount of free space is 'c->leb_size -
 575         * sleb->endpt', and the space in the 'lnum' LEB between 'offs' and
 576         * 'sleb->endpt' is used by bud data. We have to correctly calculate
 577         * how much of these data are dirty and update lprops with this
 578         * information.
 579         *
 580         * The dirt in that LEB region is comprised of padding nodes, deletion
 581         * nodes, truncation nodes and nodes which are obsoleted by subsequent
 582         * nodes in this LEB. So instead of calculating clean space, we
 583         * calculate used space ('used' variable).
 584         */
 585
 586        list_for_each_entry(snod, &sleb->nodes, list) {
 587                int deletion = 0;
 588
 589                cond_resched();
 590
 591                if (snod->sqnum >= SQNUM_WATERMARK) {
 592                        ubifs_err("file system's life ended");
 593                        goto out_dump;
 594                }
 595
 596                if (snod->sqnum > c->max_sqnum)
 597                        c->max_sqnum = snod->sqnum;
 598
 599                switch (snod->type) {
 600                case UBIFS_INO_NODE:
 601                {
 602                        struct ubifs_ino_node *ino = snod->node;
 603                        loff_t new_size = le64_to_cpu(ino->size);
 604
 605                        if (le32_to_cpu(ino->nlink) == 0)
 606                                deletion = 1;
 607                        err = insert_node(c, lnum, snod->offs, snod->len,
 608                                          &snod->key, snod->sqnum, deletion,
 609                                          &used, 0, new_size);
 610                        break;
 611                }
 612                case UBIFS_DATA_NODE:
 613                {
 614                        struct ubifs_data_node *dn = snod->node;
 615                        loff_t new_size = le32_to_cpu(dn->size) +
 616                                          key_block(c, &snod->key) *
 617                                          UBIFS_BLOCK_SIZE;
 618
 619                        err = insert_node(c, lnum, snod->offs, snod->len,
 620                                          &snod->key, snod->sqnum, deletion,
 621                                          &used, 0, new_size);
 622                        break;
 623                }
 624                case UBIFS_DENT_NODE:
 625                case UBIFS_XENT_NODE:
 626                {
 627                        struct ubifs_dent_node *dent = snod->node;
 628
 629                        err = ubifs_validate_entry(c, dent);
 630                        if (err)
 631                                goto out_dump;
 632
 633                        err = insert_dent(c, lnum, snod->offs, snod->len,
 634                                          &snod->key, dent->name,
 635                                          le16_to_cpu(dent->nlen), snod->sqnum,
 636                                          !le64_to_cpu(dent->inum), &used);
 637                        break;
 638                }
 639                case UBIFS_TRUN_NODE:
 640                {
 641                        struct ubifs_trun_node *trun = snod->node;
 642                        loff_t old_size = le64_to_cpu(trun->old_size);
 643                        loff_t new_size = le64_to_cpu(trun->new_size);
 644                        union ubifs_key key;
 645
 646                        /* Validate truncation node */
 647                        if (old_size < 0 || old_size > c->max_inode_sz ||
 648                            new_size < 0 || new_size > c->max_inode_sz ||
 649                            old_size <= new_size) {
 650                                ubifs_err("bad truncation node");
 651                                goto out_dump;
 652                        }
 653
 654                        /*
 655                         * Create a fake truncation key just to use the same
 656                         * functions which expect nodes to have keys.
 657                         */
 658                        trun_key_init(c, &key, le32_to_cpu(trun->inum));
 659                        err = insert_node(c, lnum, snod->offs, snod->len,
 660                                          &key, snod->sqnum, 1, &used,
 661                                          old_size, new_size);
 662                        break;
 663                }
 664                default:
 665                        ubifs_err("unexpected node type %d in bud LEB %d:%d",
 666                                  snod->type, lnum, snod->offs);
 667                        err = -EINVAL;
 668                        goto out_dump;
 669                }
 670                if (err)
 671                        goto out;
 672        }
 673
 674        ubifs_assert(ubifs_search_bud(c, lnum));
 675        ubifs_assert(sleb->endpt - offs >= used);
 676        ubifs_assert(sleb->endpt % c->min_io_size == 0);
 677
 678        b->dirty = sleb->endpt - offs - used;
 679        b->free = c->leb_size - sleb->endpt;
 680        dbg_mnt("bud LEB %d replied: dirty %d, free %d",
 681                lnum, b->dirty, b->free);
 682
 683out:
 684        ubifs_scan_destroy(sleb);
 685        return err;
 686
 687out_dump:
 688        ubifs_err("bad node is at LEB %d:%d", lnum, snod->offs);
 689        ubifs_dump_node(c, snod->node);
 690        ubifs_scan_destroy(sleb);
 691        return -EINVAL;
 692}
 693
 694/**
 695 * replay_buds - replay all buds.
 696 * @c: UBIFS file-system description object
 697 *
 698 * This function returns zero in case of success and a negative error code in
 699 * case of failure.
 700 */
 701static int replay_buds(struct ubifs_info *c)
 702{
 703        struct bud_entry *b;
 704        int err;
 705        unsigned long long prev_sqnum = 0;
 706
 707        list_for_each_entry(b, &c->replay_buds, list) {
 708                err = replay_bud(c, b);
 709                if (err)
 710                        return err;
 711
 712                ubifs_assert(b->sqnum > prev_sqnum);
 713                prev_sqnum = b->sqnum;
 714        }
 715
 716        return 0;
 717}
 718
 719/**
 720 * destroy_bud_list - destroy the list of buds to replay.
 721 * @c: UBIFS file-system description object
 722 */
 723static void destroy_bud_list(struct ubifs_info *c)
 724{
 725        struct bud_entry *b;
 726
 727        while (!list_empty(&c->replay_buds)) {
 728                b = list_entry(c->replay_buds.next, struct bud_entry, list);
 729                list_del(&b->list);
 730                kfree(b);
 731        }
 732}
 733
 734/**
 735 * add_replay_bud - add a bud to the list of buds to replay.
 736 * @c: UBIFS file-system description object
 737 * @lnum: bud logical eraseblock number to replay
 738 * @offs: bud start offset
 739 * @jhead: journal head to which this bud belongs
 740 * @sqnum: reference node sequence number
 741 *
 742 * This function returns zero in case of success and a negative error code in
 743 * case of failure.
 744 */
 745static int add_replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead,
 746                          unsigned long long sqnum)
 747{
 748        struct ubifs_bud *bud;
 749        struct bud_entry *b;
 750
 751        dbg_mnt("add replay bud LEB %d:%d, head %d", lnum, offs, jhead);
 752
 753        bud = kmalloc(sizeof(struct ubifs_bud), GFP_KERNEL);
 754        if (!bud)
 755                return -ENOMEM;
 756
 757        b = kmalloc(sizeof(struct bud_entry), GFP_KERNEL);
 758        if (!b) {
 759                kfree(bud);
 760                return -ENOMEM;
 761        }
 762
 763        bud->lnum = lnum;
 764        bud->start = offs;
 765        bud->jhead = jhead;
 766        ubifs_add_bud(c, bud);
 767
 768        b->bud = bud;
 769        b->sqnum = sqnum;
 770        list_add_tail(&b->list, &c->replay_buds);
 771
 772        return 0;
 773}
 774
 775/**
 776 * validate_ref - validate a reference node.
 777 * @c: UBIFS file-system description object
 778 * @ref: the reference node to validate
 779 * @ref_lnum: LEB number of the reference node
 780 * @ref_offs: reference node offset
 781 *
 782 * This function returns %1 if a bud reference already exists for the LEB. %0 is
 783 * returned if the reference node is new, otherwise %-EINVAL is returned if
 784 * validation failed.
 785 */
 786static int validate_ref(struct ubifs_info *c, const struct ubifs_ref_node *ref)
 787{
 788        struct ubifs_bud *bud;
 789        int lnum = le32_to_cpu(ref->lnum);
 790        unsigned int offs = le32_to_cpu(ref->offs);
 791        unsigned int jhead = le32_to_cpu(ref->jhead);
 792
 793        /*
 794         * ref->offs may point to the end of LEB when the journal head points
 795         * to the end of LEB and we write reference node for it during commit.
 796         * So this is why we require 'offs > c->leb_size'.
 797         */
 798        if (jhead >= c->jhead_cnt || lnum >= c->leb_cnt ||
 799            lnum < c->main_first || offs > c->leb_size ||
 800            offs & (c->min_io_size - 1))
 801                return -EINVAL;
 802
 803        /* Make sure we have not already looked at this bud */
 804        bud = ubifs_search_bud(c, lnum);
 805        if (bud) {
 806                if (bud->jhead == jhead && bud->start <= offs)
 807                        return 1;
 808                ubifs_err("bud at LEB %d:%d was already referred", lnum, offs);
 809                return -EINVAL;
 810        }
 811
 812        return 0;
 813}
 814
 815/**
 816 * replay_log_leb - replay a log logical eraseblock.
 817 * @c: UBIFS file-system description object
 818 * @lnum: log logical eraseblock to replay
 819 * @offs: offset to start replaying from
 820 * @sbuf: scan buffer
 821 *
 822 * This function replays a log LEB and returns zero in case of success, %1 if
 823 * this is the last LEB in the log, and a negative error code in case of
 824 * failure.
 825 */
 826static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
 827{
 828        int err;
 829        struct ubifs_scan_leb *sleb;
 830        struct ubifs_scan_node *snod;
 831        const struct ubifs_cs_node *node;
 832
 833        dbg_mnt("replay log LEB %d:%d", lnum, offs);
 834        sleb = ubifs_scan(c, lnum, offs, sbuf, c->need_recovery);
 835        if (IS_ERR(sleb)) {
 836                if (PTR_ERR(sleb) != -EUCLEAN || !c->need_recovery)
 837                        return PTR_ERR(sleb);
 838                /*
 839                 * Note, the below function will recover this log LEB only if
 840                 * it is the last, because unclean reboots can possibly corrupt
 841                 * only the tail of the log.
 842                 */
 843                sleb = ubifs_recover_log_leb(c, lnum, offs, sbuf);
 844                if (IS_ERR(sleb))
 845                        return PTR_ERR(sleb);
 846        }
 847
 848        if (sleb->nodes_cnt == 0) {
 849                err = 1;
 850                goto out;
 851        }
 852
 853        node = sleb->buf;
 854        snod = list_entry(sleb->nodes.next, struct ubifs_scan_node, list);
 855        if (c->cs_sqnum == 0) {
 856                /*
 857                 * This is the first log LEB we are looking at, make sure that
 858                 * the first node is a commit start node. Also record its
 859                 * sequence number so that UBIFS can determine where the log
 860                 * ends, because all nodes which were have higher sequence
 861                 * numbers.
 862                 */
 863                if (snod->type != UBIFS_CS_NODE) {
 864                        ubifs_err("first log node at LEB %d:%d is not CS node",
 865                                  lnum, offs);
 866                        goto out_dump;
 867                }
 868                if (le64_to_cpu(node->cmt_no) != c->cmt_no) {
 869                        ubifs_err("first CS node at LEB %d:%d has wrong commit number %llu expected %llu",
 870                                  lnum, offs,
 871                                  (unsigned long long)le64_to_cpu(node->cmt_no),
 872                                  c->cmt_no);
 873                        goto out_dump;
 874                }
 875
 876                c->cs_sqnum = le64_to_cpu(node->ch.sqnum);
 877                dbg_mnt("commit start sqnum %llu", c->cs_sqnum);
 878        }
 879
 880        if (snod->sqnum < c->cs_sqnum) {
 881                /*
 882                 * This means that we reached end of log and now
 883                 * look to the older log data, which was already
 884                 * committed but the eraseblock was not erased (UBIFS
 885                 * only un-maps it). So this basically means we have to
 886                 * exit with "end of log" code.
 887                 */
 888                err = 1;
 889                goto out;
 890        }
 891
 892        /* Make sure the first node sits at offset zero of the LEB */
 893        if (snod->offs != 0) {
 894                ubifs_err("first node is not at zero offset");
 895                goto out_dump;
 896        }
 897
 898        list_for_each_entry(snod, &sleb->nodes, list) {
 899                cond_resched();
 900
 901                if (snod->sqnum >= SQNUM_WATERMARK) {
 902                        ubifs_err("file system's life ended");
 903                        goto out_dump;
 904                }
 905
 906                if (snod->sqnum < c->cs_sqnum) {
 907                        ubifs_err("bad sqnum %llu, commit sqnum %llu",
 908                                  snod->sqnum, c->cs_sqnum);
 909                        goto out_dump;
 910                }
 911
 912                if (snod->sqnum > c->max_sqnum)
 913                        c->max_sqnum = snod->sqnum;
 914
 915                switch (snod->type) {
 916                case UBIFS_REF_NODE: {
 917                        const struct ubifs_ref_node *ref = snod->node;
 918
 919                        err = validate_ref(c, ref);
 920                        if (err == 1)
 921                                break; /* Already have this bud */
 922                        if (err)
 923                                goto out_dump;
 924
 925                        err = add_replay_bud(c, le32_to_cpu(ref->lnum),
 926                                             le32_to_cpu(ref->offs),
 927                                             le32_to_cpu(ref->jhead),
 928                                             snod->sqnum);
 929                        if (err)
 930                                goto out;
 931
 932                        break;
 933                }
 934                case UBIFS_CS_NODE:
 935                        /* Make sure it sits at the beginning of LEB */
 936                        if (snod->offs != 0) {
 937                                ubifs_err("unexpected node in log");
 938                                goto out_dump;
 939                        }
 940                        break;
 941                default:
 942                        ubifs_err("unexpected node in log");
 943                        goto out_dump;
 944                }
 945        }
 946
 947        if (sleb->endpt || c->lhead_offs >= c->leb_size) {
 948                c->lhead_lnum = lnum;
 949                c->lhead_offs = sleb->endpt;
 950        }
 951
 952        err = !sleb->endpt;
 953out:
 954        ubifs_scan_destroy(sleb);
 955        return err;
 956
 957out_dump:
 958        ubifs_err("log error detected while replaying the log at LEB %d:%d",
 959                  lnum, offs + snod->offs);
 960        ubifs_dump_node(c, snod->node);
 961        ubifs_scan_destroy(sleb);
 962        return -EINVAL;
 963}
 964
 965/**
 966 * take_ihead - update the status of the index head in lprops to 'taken'.
 967 * @c: UBIFS file-system description object
 968 *
 969 * This function returns the amount of free space in the index head LEB or a
 970 * negative error code.
 971 */
 972static int take_ihead(struct ubifs_info *c)
 973{
 974        const struct ubifs_lprops *lp;
 975        int err, free;
 976
 977        ubifs_get_lprops(c);
 978
 979        lp = ubifs_lpt_lookup_dirty(c, c->ihead_lnum);
 980        if (IS_ERR(lp)) {
 981                err = PTR_ERR(lp);
 982                goto out;
 983        }
 984
 985        free = lp->free;
 986
 987        lp = ubifs_change_lp(c, lp, LPROPS_NC, LPROPS_NC,
 988                             lp->flags | LPROPS_TAKEN, 0);
 989        if (IS_ERR(lp)) {
 990                err = PTR_ERR(lp);
 991                goto out;
 992        }
 993
 994        err = free;
 995out:
 996        ubifs_release_lprops(c);
 997        return err;
 998}
 999
1000/**
1001 * ubifs_replay_journal - replay journal.
1002 * @c: UBIFS file-system description object
1003 *
1004 * This function scans the journal, replays and cleans it up. It makes sure all
1005 * memory data structures related to uncommitted journal are built (dirty TNC
1006 * tree, tree of buds, modified lprops, etc).
1007 */
1008int ubifs_replay_journal(struct ubifs_info *c)
1009{
1010        int err, lnum, free;
1011
1012        BUILD_BUG_ON(UBIFS_TRUN_KEY > 5);
1013
1014        /* Update the status of the index head in lprops to 'taken' */
1015        free = take_ihead(c);
1016        if (free < 0)
1017                return free; /* Error code */
1018
1019        if (c->ihead_offs != c->leb_size - free) {
1020                ubifs_err("bad index head LEB %d:%d", c->ihead_lnum,
1021                          c->ihead_offs);
1022                return -EINVAL;
1023        }
1024
1025        dbg_mnt("start replaying the journal");
1026        c->replaying = 1;
1027        lnum = c->ltail_lnum = c->lhead_lnum;
1028
1029        do {
1030                err = replay_log_leb(c, lnum, 0, c->sbuf);
1031                if (err == 1)
1032                        /* We hit the end of the log */
1033                        break;
1034                if (err)
1035                        goto out;
1036                lnum = ubifs_next_log_lnum(c, lnum);
1037        } while (lnum != c->ltail_lnum);
1038
1039        err = replay_buds(c);
1040        if (err)
1041                goto out;
1042
1043        err = apply_replay_list(c);
1044        if (err)
1045                goto out;
1046
1047        err = set_buds_lprops(c);
1048        if (err)
1049                goto out;
1050
1051        /*
1052         * UBIFS budgeting calculations use @c->bi.uncommitted_idx variable
1053         * to roughly estimate index growth. Things like @c->bi.min_idx_lebs
1054         * depend on it. This means we have to initialize it to make sure
1055         * budgeting works properly.
1056         */
1057        c->bi.uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt);
1058        c->bi.uncommitted_idx *= c->max_idx_node_sz;
1059
1060        ubifs_assert(c->bud_bytes <= c->max_bud_bytes || c->need_recovery);
1061        dbg_mnt("finished, log head LEB %d:%d, max_sqnum %llu, highest_inum %lu",
1062                c->lhead_lnum, c->lhead_offs, c->max_sqnum,
1063                (unsigned long)c->highest_inum);
1064out:
1065        destroy_replay_list(c);
1066        destroy_bud_list(c);
1067        c->replaying = 0;
1068        return err;
1069}
1070