linux/fs/ubifs/super.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: Artem Bityutskiy (Битюцкий Артём)
  20 *          Adrian Hunter
  21 */
  22
  23/*
  24 * This file implements UBIFS initialization and VFS superblock operations. Some
  25 * initialization stuff which is rather large and complex is placed at
  26 * corresponding subsystems, but most of it is here.
  27 */
  28
  29#include <linux/init.h>
  30#include <linux/slab.h>
  31#include <linux/module.h>
  32#include <linux/ctype.h>
  33#include <linux/kthread.h>
  34#include <linux/parser.h>
  35#include <linux/seq_file.h>
  36#include <linux/mount.h>
  37#include <linux/math64.h>
  38#include <linux/writeback.h>
  39#include "ubifs.h"
  40
  41/*
  42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
  43 * allocating too much.
  44 */
  45#define UBIFS_KMALLOC_OK (128*1024)
  46
  47/* Slab cache for UBIFS inodes */
  48struct kmem_cache *ubifs_inode_slab;
  49
  50/* UBIFS TNC shrinker description */
  51static struct shrinker ubifs_shrinker_info = {
  52        .shrink = ubifs_shrinker,
  53        .seeks = DEFAULT_SEEKS,
  54};
  55
  56/**
  57 * validate_inode - validate inode.
  58 * @c: UBIFS file-system description object
  59 * @inode: the inode to validate
  60 *
  61 * This is a helper function for 'ubifs_iget()' which validates various fields
  62 * of a newly built inode to make sure they contain sane values and prevent
  63 * possible vulnerabilities. Returns zero if the inode is all right and
  64 * a non-zero error code if not.
  65 */
  66static int validate_inode(struct ubifs_info *c, const struct inode *inode)
  67{
  68        int err;
  69        const struct ubifs_inode *ui = ubifs_inode(inode);
  70
  71        if (inode->i_size > c->max_inode_sz) {
  72                ubifs_err("inode is too large (%lld)",
  73                          (long long)inode->i_size);
  74                return 1;
  75        }
  76
  77        if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
  78                ubifs_err("unknown compression type %d", ui->compr_type);
  79                return 2;
  80        }
  81
  82        if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
  83                return 3;
  84
  85        if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
  86                return 4;
  87
  88        if (ui->xattr && !S_ISREG(inode->i_mode))
  89                return 5;
  90
  91        if (!ubifs_compr_present(ui->compr_type)) {
  92                ubifs_warn("inode %lu uses '%s' compression, but it was not compiled in",
  93                           inode->i_ino, ubifs_compr_name(ui->compr_type));
  94        }
  95
  96        err = dbg_check_dir(c, inode);
  97        return err;
  98}
  99
 100struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
 101{
 102        int err;
 103        union ubifs_key key;
 104        struct ubifs_ino_node *ino;
 105        struct ubifs_info *c = sb->s_fs_info;
 106        struct inode *inode;
 107        struct ubifs_inode *ui;
 108
 109        dbg_gen("inode %lu", inum);
 110
 111        inode = iget_locked(sb, inum);
 112        if (!inode)
 113                return ERR_PTR(-ENOMEM);
 114        if (!(inode->i_state & I_NEW))
 115                return inode;
 116        ui = ubifs_inode(inode);
 117
 118        ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
 119        if (!ino) {
 120                err = -ENOMEM;
 121                goto out;
 122        }
 123
 124        ino_key_init(c, &key, inode->i_ino);
 125
 126        err = ubifs_tnc_lookup(c, &key, ino);
 127        if (err)
 128                goto out_ino;
 129
 130        inode->i_flags |= (S_NOCMTIME | S_NOATIME);
 131        set_nlink(inode, le32_to_cpu(ino->nlink));
 132        i_uid_write(inode, le32_to_cpu(ino->uid));
 133        i_gid_write(inode, le32_to_cpu(ino->gid));
 134        inode->i_atime.tv_sec  = (int64_t)le64_to_cpu(ino->atime_sec);
 135        inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
 136        inode->i_mtime.tv_sec  = (int64_t)le64_to_cpu(ino->mtime_sec);
 137        inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
 138        inode->i_ctime.tv_sec  = (int64_t)le64_to_cpu(ino->ctime_sec);
 139        inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
 140        inode->i_mode = le32_to_cpu(ino->mode);
 141        inode->i_size = le64_to_cpu(ino->size);
 142
 143        ui->data_len    = le32_to_cpu(ino->data_len);
 144        ui->flags       = le32_to_cpu(ino->flags);
 145        ui->compr_type  = le16_to_cpu(ino->compr_type);
 146        ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
 147        ui->xattr_cnt   = le32_to_cpu(ino->xattr_cnt);
 148        ui->xattr_size  = le32_to_cpu(ino->xattr_size);
 149        ui->xattr_names = le32_to_cpu(ino->xattr_names);
 150        ui->synced_i_size = ui->ui_size = inode->i_size;
 151
 152        ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
 153
 154        err = validate_inode(c, inode);
 155        if (err)
 156                goto out_invalid;
 157
 158        /* Disable read-ahead */
 159        inode->i_mapping->backing_dev_info = &c->bdi;
 160
 161        switch (inode->i_mode & S_IFMT) {
 162        case S_IFREG:
 163                inode->i_mapping->a_ops = &ubifs_file_address_operations;
 164                inode->i_op = &ubifs_file_inode_operations;
 165                inode->i_fop = &ubifs_file_operations;
 166                if (ui->xattr) {
 167                        ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
 168                        if (!ui->data) {
 169                                err = -ENOMEM;
 170                                goto out_ino;
 171                        }
 172                        memcpy(ui->data, ino->data, ui->data_len);
 173                        ((char *)ui->data)[ui->data_len] = '\0';
 174                } else if (ui->data_len != 0) {
 175                        err = 10;
 176                        goto out_invalid;
 177                }
 178                break;
 179        case S_IFDIR:
 180                inode->i_op  = &ubifs_dir_inode_operations;
 181                inode->i_fop = &ubifs_dir_operations;
 182                if (ui->data_len != 0) {
 183                        err = 11;
 184                        goto out_invalid;
 185                }
 186                break;
 187        case S_IFLNK:
 188                inode->i_op = &ubifs_symlink_inode_operations;
 189                if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
 190                        err = 12;
 191                        goto out_invalid;
 192                }
 193                ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
 194                if (!ui->data) {
 195                        err = -ENOMEM;
 196                        goto out_ino;
 197                }
 198                memcpy(ui->data, ino->data, ui->data_len);
 199                ((char *)ui->data)[ui->data_len] = '\0';
 200                break;
 201        case S_IFBLK:
 202        case S_IFCHR:
 203        {
 204                dev_t rdev;
 205                union ubifs_dev_desc *dev;
 206
 207                ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
 208                if (!ui->data) {
 209                        err = -ENOMEM;
 210                        goto out_ino;
 211                }
 212
 213                dev = (union ubifs_dev_desc *)ino->data;
 214                if (ui->data_len == sizeof(dev->new))
 215                        rdev = new_decode_dev(le32_to_cpu(dev->new));
 216                else if (ui->data_len == sizeof(dev->huge))
 217                        rdev = huge_decode_dev(le64_to_cpu(dev->huge));
 218                else {
 219                        err = 13;
 220                        goto out_invalid;
 221                }
 222                memcpy(ui->data, ino->data, ui->data_len);
 223                inode->i_op = &ubifs_file_inode_operations;
 224                init_special_inode(inode, inode->i_mode, rdev);
 225                break;
 226        }
 227        case S_IFSOCK:
 228        case S_IFIFO:
 229                inode->i_op = &ubifs_file_inode_operations;
 230                init_special_inode(inode, inode->i_mode, 0);
 231                if (ui->data_len != 0) {
 232                        err = 14;
 233                        goto out_invalid;
 234                }
 235                break;
 236        default:
 237                err = 15;
 238                goto out_invalid;
 239        }
 240
 241        kfree(ino);
 242        ubifs_set_inode_flags(inode);
 243        unlock_new_inode(inode);
 244        return inode;
 245
 246out_invalid:
 247        ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
 248        ubifs_dump_node(c, ino);
 249        ubifs_dump_inode(c, inode);
 250        err = -EINVAL;
 251out_ino:
 252        kfree(ino);
 253out:
 254        ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
 255        iget_failed(inode);
 256        return ERR_PTR(err);
 257}
 258
 259static struct inode *ubifs_alloc_inode(struct super_block *sb)
 260{
 261        struct ubifs_inode *ui;
 262
 263        ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
 264        if (!ui)
 265                return NULL;
 266
 267        memset((void *)ui + sizeof(struct inode), 0,
 268               sizeof(struct ubifs_inode) - sizeof(struct inode));
 269        mutex_init(&ui->ui_mutex);
 270        spin_lock_init(&ui->ui_lock);
 271        return &ui->vfs_inode;
 272};
 273
 274static void ubifs_i_callback(struct rcu_head *head)
 275{
 276        struct inode *inode = container_of(head, struct inode, i_rcu);
 277        struct ubifs_inode *ui = ubifs_inode(inode);
 278        kmem_cache_free(ubifs_inode_slab, ui);
 279}
 280
 281static void ubifs_destroy_inode(struct inode *inode)
 282{
 283        struct ubifs_inode *ui = ubifs_inode(inode);
 284
 285        kfree(ui->data);
 286        call_rcu(&inode->i_rcu, ubifs_i_callback);
 287}
 288
 289/*
 290 * Note, Linux write-back code calls this without 'i_mutex'.
 291 */
 292static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
 293{
 294        int err = 0;
 295        struct ubifs_info *c = inode->i_sb->s_fs_info;
 296        struct ubifs_inode *ui = ubifs_inode(inode);
 297
 298        ubifs_assert(!ui->xattr);
 299        if (is_bad_inode(inode))
 300                return 0;
 301
 302        mutex_lock(&ui->ui_mutex);
 303        /*
 304         * Due to races between write-back forced by budgeting
 305         * (see 'sync_some_inodes()') and background write-back, the inode may
 306         * have already been synchronized, do not do this again. This might
 307         * also happen if it was synchronized in an VFS operation, e.g.
 308         * 'ubifs_link()'.
 309         */
 310        if (!ui->dirty) {
 311                mutex_unlock(&ui->ui_mutex);
 312                return 0;
 313        }
 314
 315        /*
 316         * As an optimization, do not write orphan inodes to the media just
 317         * because this is not needed.
 318         */
 319        dbg_gen("inode %lu, mode %#x, nlink %u",
 320                inode->i_ino, (int)inode->i_mode, inode->i_nlink);
 321        if (inode->i_nlink) {
 322                err = ubifs_jnl_write_inode(c, inode);
 323                if (err)
 324                        ubifs_err("can't write inode %lu, error %d",
 325                                  inode->i_ino, err);
 326                else
 327                        err = dbg_check_inode_size(c, inode, ui->ui_size);
 328        }
 329
 330        ui->dirty = 0;
 331        mutex_unlock(&ui->ui_mutex);
 332        ubifs_release_dirty_inode_budget(c, ui);
 333        return err;
 334}
 335
 336static void ubifs_evict_inode(struct inode *inode)
 337{
 338        int err;
 339        struct ubifs_info *c = inode->i_sb->s_fs_info;
 340        struct ubifs_inode *ui = ubifs_inode(inode);
 341
 342        if (ui->xattr)
 343                /*
 344                 * Extended attribute inode deletions are fully handled in
 345                 * 'ubifs_removexattr()'. These inodes are special and have
 346                 * limited usage, so there is nothing to do here.
 347                 */
 348                goto out;
 349
 350        dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
 351        ubifs_assert(!atomic_read(&inode->i_count));
 352
 353        truncate_inode_pages(&inode->i_data, 0);
 354
 355        if (inode->i_nlink)
 356                goto done;
 357
 358        if (is_bad_inode(inode))
 359                goto out;
 360
 361        ui->ui_size = inode->i_size = 0;
 362        err = ubifs_jnl_delete_inode(c, inode);
 363        if (err)
 364                /*
 365                 * Worst case we have a lost orphan inode wasting space, so a
 366                 * simple error message is OK here.
 367                 */
 368                ubifs_err("can't delete inode %lu, error %d",
 369                          inode->i_ino, err);
 370
 371out:
 372        if (ui->dirty)
 373                ubifs_release_dirty_inode_budget(c, ui);
 374        else {
 375                /* We've deleted something - clean the "no space" flags */
 376                c->bi.nospace = c->bi.nospace_rp = 0;
 377                smp_wmb();
 378        }
 379done:
 380        clear_inode(inode);
 381}
 382
 383static void ubifs_dirty_inode(struct inode *inode, int flags)
 384{
 385        struct ubifs_inode *ui = ubifs_inode(inode);
 386
 387        ubifs_assert(mutex_is_locked(&ui->ui_mutex));
 388        if (!ui->dirty) {
 389                ui->dirty = 1;
 390                dbg_gen("inode %lu",  inode->i_ino);
 391        }
 392}
 393
 394static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
 395{
 396        struct ubifs_info *c = dentry->d_sb->s_fs_info;
 397        unsigned long long free;
 398        __le32 *uuid = (__le32 *)c->uuid;
 399
 400        free = ubifs_get_free_space(c);
 401        dbg_gen("free space %lld bytes (%lld blocks)",
 402                free, free >> UBIFS_BLOCK_SHIFT);
 403
 404        buf->f_type = UBIFS_SUPER_MAGIC;
 405        buf->f_bsize = UBIFS_BLOCK_SIZE;
 406        buf->f_blocks = c->block_cnt;
 407        buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
 408        if (free > c->report_rp_size)
 409                buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
 410        else
 411                buf->f_bavail = 0;
 412        buf->f_files = 0;
 413        buf->f_ffree = 0;
 414        buf->f_namelen = UBIFS_MAX_NLEN;
 415        buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
 416        buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
 417        ubifs_assert(buf->f_bfree <= c->block_cnt);
 418        return 0;
 419}
 420
 421static int ubifs_show_options(struct seq_file *s, struct dentry *root)
 422{
 423        struct ubifs_info *c = root->d_sb->s_fs_info;
 424
 425        if (c->mount_opts.unmount_mode == 2)
 426                seq_printf(s, ",fast_unmount");
 427        else if (c->mount_opts.unmount_mode == 1)
 428                seq_printf(s, ",norm_unmount");
 429
 430        if (c->mount_opts.bulk_read == 2)
 431                seq_printf(s, ",bulk_read");
 432        else if (c->mount_opts.bulk_read == 1)
 433                seq_printf(s, ",no_bulk_read");
 434
 435        if (c->mount_opts.chk_data_crc == 2)
 436                seq_printf(s, ",chk_data_crc");
 437        else if (c->mount_opts.chk_data_crc == 1)
 438                seq_printf(s, ",no_chk_data_crc");
 439
 440        if (c->mount_opts.override_compr) {
 441                seq_printf(s, ",compr=%s",
 442                           ubifs_compr_name(c->mount_opts.compr_type));
 443        }
 444
 445        return 0;
 446}
 447
 448static int ubifs_sync_fs(struct super_block *sb, int wait)
 449{
 450        int i, err;
 451        struct ubifs_info *c = sb->s_fs_info;
 452
 453        /*
 454         * Zero @wait is just an advisory thing to help the file system shove
 455         * lots of data into the queues, and there will be the second
 456         * '->sync_fs()' call, with non-zero @wait.
 457         */
 458        if (!wait)
 459                return 0;
 460
 461        /*
 462         * Synchronize write buffers, because 'ubifs_run_commit()' does not
 463         * do this if it waits for an already running commit.
 464         */
 465        for (i = 0; i < c->jhead_cnt; i++) {
 466                err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
 467                if (err)
 468                        return err;
 469        }
 470
 471        /*
 472         * Strictly speaking, it is not necessary to commit the journal here,
 473         * synchronizing write-buffers would be enough. But committing makes
 474         * UBIFS free space predictions much more accurate, so we want to let
 475         * the user be able to get more accurate results of 'statfs()' after
 476         * they synchronize the file system.
 477         */
 478        err = ubifs_run_commit(c);
 479        if (err)
 480                return err;
 481
 482        return ubi_sync(c->vi.ubi_num);
 483}
 484
 485/**
 486 * init_constants_early - initialize UBIFS constants.
 487 * @c: UBIFS file-system description object
 488 *
 489 * This function initialize UBIFS constants which do not need the superblock to
 490 * be read. It also checks that the UBI volume satisfies basic UBIFS
 491 * requirements. Returns zero in case of success and a negative error code in
 492 * case of failure.
 493 */
 494static int init_constants_early(struct ubifs_info *c)
 495{
 496        if (c->vi.corrupted) {
 497                ubifs_warn("UBI volume is corrupted - read-only mode");
 498                c->ro_media = 1;
 499        }
 500
 501        if (c->di.ro_mode) {
 502                ubifs_msg("read-only UBI device");
 503                c->ro_media = 1;
 504        }
 505
 506        if (c->vi.vol_type == UBI_STATIC_VOLUME) {
 507                ubifs_msg("static UBI volume - read-only mode");
 508                c->ro_media = 1;
 509        }
 510
 511        c->leb_cnt = c->vi.size;
 512        c->leb_size = c->vi.usable_leb_size;
 513        c->leb_start = c->di.leb_start;
 514        c->half_leb_size = c->leb_size / 2;
 515        c->min_io_size = c->di.min_io_size;
 516        c->min_io_shift = fls(c->min_io_size) - 1;
 517        c->max_write_size = c->di.max_write_size;
 518        c->max_write_shift = fls(c->max_write_size) - 1;
 519
 520        if (c->leb_size < UBIFS_MIN_LEB_SZ) {
 521                ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
 522                          c->leb_size, UBIFS_MIN_LEB_SZ);
 523                return -EINVAL;
 524        }
 525
 526        if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
 527                ubifs_err("too few LEBs (%d), min. is %d",
 528                          c->leb_cnt, UBIFS_MIN_LEB_CNT);
 529                return -EINVAL;
 530        }
 531
 532        if (!is_power_of_2(c->min_io_size)) {
 533                ubifs_err("bad min. I/O size %d", c->min_io_size);
 534                return -EINVAL;
 535        }
 536
 537        /*
 538         * Maximum write size has to be greater or equivalent to min. I/O
 539         * size, and be multiple of min. I/O size.
 540         */
 541        if (c->max_write_size < c->min_io_size ||
 542            c->max_write_size % c->min_io_size ||
 543            !is_power_of_2(c->max_write_size)) {
 544                ubifs_err("bad write buffer size %d for %d min. I/O unit",
 545                          c->max_write_size, c->min_io_size);
 546                return -EINVAL;
 547        }
 548
 549        /*
 550         * UBIFS aligns all node to 8-byte boundary, so to make function in
 551         * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
 552         * less than 8.
 553         */
 554        if (c->min_io_size < 8) {
 555                c->min_io_size = 8;
 556                c->min_io_shift = 3;
 557                if (c->max_write_size < c->min_io_size) {
 558                        c->max_write_size = c->min_io_size;
 559                        c->max_write_shift = c->min_io_shift;
 560                }
 561        }
 562
 563        c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
 564        c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
 565
 566        /*
 567         * Initialize node length ranges which are mostly needed for node
 568         * length validation.
 569         */
 570        c->ranges[UBIFS_PAD_NODE].len  = UBIFS_PAD_NODE_SZ;
 571        c->ranges[UBIFS_SB_NODE].len   = UBIFS_SB_NODE_SZ;
 572        c->ranges[UBIFS_MST_NODE].len  = UBIFS_MST_NODE_SZ;
 573        c->ranges[UBIFS_REF_NODE].len  = UBIFS_REF_NODE_SZ;
 574        c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
 575        c->ranges[UBIFS_CS_NODE].len   = UBIFS_CS_NODE_SZ;
 576
 577        c->ranges[UBIFS_INO_NODE].min_len  = UBIFS_INO_NODE_SZ;
 578        c->ranges[UBIFS_INO_NODE].max_len  = UBIFS_MAX_INO_NODE_SZ;
 579        c->ranges[UBIFS_ORPH_NODE].min_len =
 580                                UBIFS_ORPH_NODE_SZ + sizeof(__le64);
 581        c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
 582        c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
 583        c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
 584        c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
 585        c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
 586        c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
 587        c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
 588        /*
 589         * Minimum indexing node size is amended later when superblock is
 590         * read and the key length is known.
 591         */
 592        c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
 593        /*
 594         * Maximum indexing node size is amended later when superblock is
 595         * read and the fanout is known.
 596         */
 597        c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
 598
 599        /*
 600         * Initialize dead and dark LEB space watermarks. See gc.c for comments
 601         * about these values.
 602         */
 603        c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
 604        c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
 605
 606        /*
 607         * Calculate how many bytes would be wasted at the end of LEB if it was
 608         * fully filled with data nodes of maximum size. This is used in
 609         * calculations when reporting free space.
 610         */
 611        c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
 612
 613        /* Buffer size for bulk-reads */
 614        c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
 615        if (c->max_bu_buf_len > c->leb_size)
 616                c->max_bu_buf_len = c->leb_size;
 617        return 0;
 618}
 619
 620/**
 621 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
 622 * @c: UBIFS file-system description object
 623 * @lnum: LEB the write-buffer was synchronized to
 624 * @free: how many free bytes left in this LEB
 625 * @pad: how many bytes were padded
 626 *
 627 * This is a callback function which is called by the I/O unit when the
 628 * write-buffer is synchronized. We need this to correctly maintain space
 629 * accounting in bud logical eraseblocks. This function returns zero in case of
 630 * success and a negative error code in case of failure.
 631 *
 632 * This function actually belongs to the journal, but we keep it here because
 633 * we want to keep it static.
 634 */
 635static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
 636{
 637        return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
 638}
 639
 640/*
 641 * init_constants_sb - initialize UBIFS constants.
 642 * @c: UBIFS file-system description object
 643 *
 644 * This is a helper function which initializes various UBIFS constants after
 645 * the superblock has been read. It also checks various UBIFS parameters and
 646 * makes sure they are all right. Returns zero in case of success and a
 647 * negative error code in case of failure.
 648 */
 649static int init_constants_sb(struct ubifs_info *c)
 650{
 651        int tmp, err;
 652        long long tmp64;
 653
 654        c->main_bytes = (long long)c->main_lebs * c->leb_size;
 655        c->max_znode_sz = sizeof(struct ubifs_znode) +
 656                                c->fanout * sizeof(struct ubifs_zbranch);
 657
 658        tmp = ubifs_idx_node_sz(c, 1);
 659        c->ranges[UBIFS_IDX_NODE].min_len = tmp;
 660        c->min_idx_node_sz = ALIGN(tmp, 8);
 661
 662        tmp = ubifs_idx_node_sz(c, c->fanout);
 663        c->ranges[UBIFS_IDX_NODE].max_len = tmp;
 664        c->max_idx_node_sz = ALIGN(tmp, 8);
 665
 666        /* Make sure LEB size is large enough to fit full commit */
 667        tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
 668        tmp = ALIGN(tmp, c->min_io_size);
 669        if (tmp > c->leb_size) {
 670                ubifs_err("too small LEB size %d, at least %d needed",
 671                          c->leb_size, tmp);
 672                return -EINVAL;
 673        }
 674
 675        /*
 676         * Make sure that the log is large enough to fit reference nodes for
 677         * all buds plus one reserved LEB.
 678         */
 679        tmp64 = c->max_bud_bytes + c->leb_size - 1;
 680        c->max_bud_cnt = div_u64(tmp64, c->leb_size);
 681        tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
 682        tmp /= c->leb_size;
 683        tmp += 1;
 684        if (c->log_lebs < tmp) {
 685                ubifs_err("too small log %d LEBs, required min. %d LEBs",
 686                          c->log_lebs, tmp);
 687                return -EINVAL;
 688        }
 689
 690        /*
 691         * When budgeting we assume worst-case scenarios when the pages are not
 692         * be compressed and direntries are of the maximum size.
 693         *
 694         * Note, data, which may be stored in inodes is budgeted separately, so
 695         * it is not included into 'c->bi.inode_budget'.
 696         */
 697        c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
 698        c->bi.inode_budget = UBIFS_INO_NODE_SZ;
 699        c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
 700
 701        /*
 702         * When the amount of flash space used by buds becomes
 703         * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
 704         * The writers are unblocked when the commit is finished. To avoid
 705         * writers to be blocked UBIFS initiates background commit in advance,
 706         * when number of bud bytes becomes above the limit defined below.
 707         */
 708        c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
 709
 710        /*
 711         * Ensure minimum journal size. All the bytes in the journal heads are
 712         * considered to be used, when calculating the current journal usage.
 713         * Consequently, if the journal is too small, UBIFS will treat it as
 714         * always full.
 715         */
 716        tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
 717        if (c->bg_bud_bytes < tmp64)
 718                c->bg_bud_bytes = tmp64;
 719        if (c->max_bud_bytes < tmp64 + c->leb_size)
 720                c->max_bud_bytes = tmp64 + c->leb_size;
 721
 722        err = ubifs_calc_lpt_geom(c);
 723        if (err)
 724                return err;
 725
 726        /* Initialize effective LEB size used in budgeting calculations */
 727        c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
 728        return 0;
 729}
 730
 731/*
 732 * init_constants_master - initialize UBIFS constants.
 733 * @c: UBIFS file-system description object
 734 *
 735 * This is a helper function which initializes various UBIFS constants after
 736 * the master node has been read. It also checks various UBIFS parameters and
 737 * makes sure they are all right.
 738 */
 739static void init_constants_master(struct ubifs_info *c)
 740{
 741        long long tmp64;
 742
 743        c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
 744        c->report_rp_size = ubifs_reported_space(c, c->rp_size);
 745
 746        /*
 747         * Calculate total amount of FS blocks. This number is not used
 748         * internally because it does not make much sense for UBIFS, but it is
 749         * necessary to report something for the 'statfs()' call.
 750         *
 751         * Subtract the LEB reserved for GC, the LEB which is reserved for
 752         * deletions, minimum LEBs for the index, and assume only one journal
 753         * head is available.
 754         */
 755        tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
 756        tmp64 *= (long long)c->leb_size - c->leb_overhead;
 757        tmp64 = ubifs_reported_space(c, tmp64);
 758        c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
 759}
 760
 761/**
 762 * take_gc_lnum - reserve GC LEB.
 763 * @c: UBIFS file-system description object
 764 *
 765 * This function ensures that the LEB reserved for garbage collection is marked
 766 * as "taken" in lprops. We also have to set free space to LEB size and dirty
 767 * space to zero, because lprops may contain out-of-date information if the
 768 * file-system was un-mounted before it has been committed. This function
 769 * returns zero in case of success and a negative error code in case of
 770 * failure.
 771 */
 772static int take_gc_lnum(struct ubifs_info *c)
 773{
 774        int err;
 775
 776        if (c->gc_lnum == -1) {
 777                ubifs_err("no LEB for GC");
 778                return -EINVAL;
 779        }
 780
 781        /* And we have to tell lprops that this LEB is taken */
 782        err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
 783                                  LPROPS_TAKEN, 0, 0);
 784        return err;
 785}
 786
 787/**
 788 * alloc_wbufs - allocate write-buffers.
 789 * @c: UBIFS file-system description object
 790 *
 791 * This helper function allocates and initializes UBIFS write-buffers. Returns
 792 * zero in case of success and %-ENOMEM in case of failure.
 793 */
 794static int alloc_wbufs(struct ubifs_info *c)
 795{
 796        int i, err;
 797
 798        c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
 799                           GFP_KERNEL);
 800        if (!c->jheads)
 801                return -ENOMEM;
 802
 803        /* Initialize journal heads */
 804        for (i = 0; i < c->jhead_cnt; i++) {
 805                INIT_LIST_HEAD(&c->jheads[i].buds_list);
 806                err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
 807                if (err)
 808                        return err;
 809
 810                c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
 811                c->jheads[i].wbuf.jhead = i;
 812                c->jheads[i].grouped = 1;
 813        }
 814
 815        /*
 816         * Garbage Collector head does not need to be synchronized by timer.
 817         * Also GC head nodes are not grouped.
 818         */
 819        c->jheads[GCHD].wbuf.no_timer = 1;
 820        c->jheads[GCHD].grouped = 0;
 821
 822        return 0;
 823}
 824
 825/**
 826 * free_wbufs - free write-buffers.
 827 * @c: UBIFS file-system description object
 828 */
 829static void free_wbufs(struct ubifs_info *c)
 830{
 831        int i;
 832
 833        if (c->jheads) {
 834                for (i = 0; i < c->jhead_cnt; i++) {
 835                        kfree(c->jheads[i].wbuf.buf);
 836                        kfree(c->jheads[i].wbuf.inodes);
 837                }
 838                kfree(c->jheads);
 839                c->jheads = NULL;
 840        }
 841}
 842
 843/**
 844 * free_orphans - free orphans.
 845 * @c: UBIFS file-system description object
 846 */
 847static void free_orphans(struct ubifs_info *c)
 848{
 849        struct ubifs_orphan *orph;
 850
 851        while (c->orph_dnext) {
 852                orph = c->orph_dnext;
 853                c->orph_dnext = orph->dnext;
 854                list_del(&orph->list);
 855                kfree(orph);
 856        }
 857
 858        while (!list_empty(&c->orph_list)) {
 859                orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
 860                list_del(&orph->list);
 861                kfree(orph);
 862                ubifs_err("orphan list not empty at unmount");
 863        }
 864
 865        vfree(c->orph_buf);
 866        c->orph_buf = NULL;
 867}
 868
 869/**
 870 * free_buds - free per-bud objects.
 871 * @c: UBIFS file-system description object
 872 */
 873static void free_buds(struct ubifs_info *c)
 874{
 875        struct rb_node *this = c->buds.rb_node;
 876        struct ubifs_bud *bud;
 877
 878        while (this) {
 879                if (this->rb_left)
 880                        this = this->rb_left;
 881                else if (this->rb_right)
 882                        this = this->rb_right;
 883                else {
 884                        bud = rb_entry(this, struct ubifs_bud, rb);
 885                        this = rb_parent(this);
 886                        if (this) {
 887                                if (this->rb_left == &bud->rb)
 888                                        this->rb_left = NULL;
 889                                else
 890                                        this->rb_right = NULL;
 891                        }
 892                        kfree(bud);
 893                }
 894        }
 895}
 896
 897/**
 898 * check_volume_empty - check if the UBI volume is empty.
 899 * @c: UBIFS file-system description object
 900 *
 901 * This function checks if the UBIFS volume is empty by looking if its LEBs are
 902 * mapped or not. The result of checking is stored in the @c->empty variable.
 903 * Returns zero in case of success and a negative error code in case of
 904 * failure.
 905 */
 906static int check_volume_empty(struct ubifs_info *c)
 907{
 908        int lnum, err;
 909
 910        c->empty = 1;
 911        for (lnum = 0; lnum < c->leb_cnt; lnum++) {
 912                err = ubifs_is_mapped(c, lnum);
 913                if (unlikely(err < 0))
 914                        return err;
 915                if (err == 1) {
 916                        c->empty = 0;
 917                        break;
 918                }
 919
 920                cond_resched();
 921        }
 922
 923        return 0;
 924}
 925
 926/*
 927 * UBIFS mount options.
 928 *
 929 * Opt_fast_unmount: do not run a journal commit before un-mounting
 930 * Opt_norm_unmount: run a journal commit before un-mounting
 931 * Opt_bulk_read: enable bulk-reads
 932 * Opt_no_bulk_read: disable bulk-reads
 933 * Opt_chk_data_crc: check CRCs when reading data nodes
 934 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
 935 * Opt_override_compr: override default compressor
 936 * Opt_err: just end of array marker
 937 */
 938enum {
 939        Opt_fast_unmount,
 940        Opt_norm_unmount,
 941        Opt_bulk_read,
 942        Opt_no_bulk_read,
 943        Opt_chk_data_crc,
 944        Opt_no_chk_data_crc,
 945        Opt_override_compr,
 946        Opt_err,
 947};
 948
 949static const match_table_t tokens = {
 950        {Opt_fast_unmount, "fast_unmount"},
 951        {Opt_norm_unmount, "norm_unmount"},
 952        {Opt_bulk_read, "bulk_read"},
 953        {Opt_no_bulk_read, "no_bulk_read"},
 954        {Opt_chk_data_crc, "chk_data_crc"},
 955        {Opt_no_chk_data_crc, "no_chk_data_crc"},
 956        {Opt_override_compr, "compr=%s"},
 957        {Opt_err, NULL},
 958};
 959
 960/**
 961 * parse_standard_option - parse a standard mount option.
 962 * @option: the option to parse
 963 *
 964 * Normally, standard mount options like "sync" are passed to file-systems as
 965 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
 966 * be present in the options string. This function tries to deal with this
 967 * situation and parse standard options. Returns 0 if the option was not
 968 * recognized, and the corresponding integer flag if it was.
 969 *
 970 * UBIFS is only interested in the "sync" option, so do not check for anything
 971 * else.
 972 */
 973static int parse_standard_option(const char *option)
 974{
 975        ubifs_msg("parse %s", option);
 976        if (!strcmp(option, "sync"))
 977                return MS_SYNCHRONOUS;
 978        return 0;
 979}
 980
 981/**
 982 * ubifs_parse_options - parse mount parameters.
 983 * @c: UBIFS file-system description object
 984 * @options: parameters to parse
 985 * @is_remount: non-zero if this is FS re-mount
 986 *
 987 * This function parses UBIFS mount options and returns zero in case success
 988 * and a negative error code in case of failure.
 989 */
 990static int ubifs_parse_options(struct ubifs_info *c, char *options,
 991                               int is_remount)
 992{
 993        char *p;
 994        substring_t args[MAX_OPT_ARGS];
 995
 996        if (!options)
 997                return 0;
 998
 999        while ((p = strsep(&options, ","))) {
1000                int token;
1001
1002                if (!*p)
1003                        continue;
1004
1005                token = match_token(p, tokens, args);
1006                switch (token) {
1007                /*
1008                 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
1009                 * We accept them in order to be backward-compatible. But this
1010                 * should be removed at some point.
1011                 */
1012                case Opt_fast_unmount:
1013                        c->mount_opts.unmount_mode = 2;
1014                        break;
1015                case Opt_norm_unmount:
1016                        c->mount_opts.unmount_mode = 1;
1017                        break;
1018                case Opt_bulk_read:
1019                        c->mount_opts.bulk_read = 2;
1020                        c->bulk_read = 1;
1021                        break;
1022                case Opt_no_bulk_read:
1023                        c->mount_opts.bulk_read = 1;
1024                        c->bulk_read = 0;
1025                        break;
1026                case Opt_chk_data_crc:
1027                        c->mount_opts.chk_data_crc = 2;
1028                        c->no_chk_data_crc = 0;
1029                        break;
1030                case Opt_no_chk_data_crc:
1031                        c->mount_opts.chk_data_crc = 1;
1032                        c->no_chk_data_crc = 1;
1033                        break;
1034                case Opt_override_compr:
1035                {
1036                        char *name = match_strdup(&args[0]);
1037
1038                        if (!name)
1039                                return -ENOMEM;
1040                        if (!strcmp(name, "none"))
1041                                c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1042                        else if (!strcmp(name, "lzo"))
1043                                c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1044                        else if (!strcmp(name, "zlib"))
1045                                c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1046                        else {
1047                                ubifs_err("unknown compressor \"%s\"", name);
1048                                kfree(name);
1049                                return -EINVAL;
1050                        }
1051                        kfree(name);
1052                        c->mount_opts.override_compr = 1;
1053                        c->default_compr = c->mount_opts.compr_type;
1054                        break;
1055                }
1056                default:
1057                {
1058                        unsigned long flag;
1059                        struct super_block *sb = c->vfs_sb;
1060
1061                        flag = parse_standard_option(p);
1062                        if (!flag) {
1063                                ubifs_err("unrecognized mount option \"%s\" or missing value",
1064                                          p);
1065                                return -EINVAL;
1066                        }
1067                        sb->s_flags |= flag;
1068                        break;
1069                }
1070                }
1071        }
1072
1073        return 0;
1074}
1075
1076/**
1077 * destroy_journal - destroy journal data structures.
1078 * @c: UBIFS file-system description object
1079 *
1080 * This function destroys journal data structures including those that may have
1081 * been created by recovery functions.
1082 */
1083static void destroy_journal(struct ubifs_info *c)
1084{
1085        while (!list_empty(&c->unclean_leb_list)) {
1086                struct ubifs_unclean_leb *ucleb;
1087
1088                ucleb = list_entry(c->unclean_leb_list.next,
1089                                   struct ubifs_unclean_leb, list);
1090                list_del(&ucleb->list);
1091                kfree(ucleb);
1092        }
1093        while (!list_empty(&c->old_buds)) {
1094                struct ubifs_bud *bud;
1095
1096                bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1097                list_del(&bud->list);
1098                kfree(bud);
1099        }
1100        ubifs_destroy_idx_gc(c);
1101        ubifs_destroy_size_tree(c);
1102        ubifs_tnc_close(c);
1103        free_buds(c);
1104}
1105
1106/**
1107 * bu_init - initialize bulk-read information.
1108 * @c: UBIFS file-system description object
1109 */
1110static void bu_init(struct ubifs_info *c)
1111{
1112        ubifs_assert(c->bulk_read == 1);
1113
1114        if (c->bu.buf)
1115                return; /* Already initialized */
1116
1117again:
1118        c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1119        if (!c->bu.buf) {
1120                if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1121                        c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1122                        goto again;
1123                }
1124
1125                /* Just disable bulk-read */
1126                ubifs_warn("cannot allocate %d bytes of memory for bulk-read, disabling it",
1127                           c->max_bu_buf_len);
1128                c->mount_opts.bulk_read = 1;
1129                c->bulk_read = 0;
1130                return;
1131        }
1132}
1133
1134/**
1135 * check_free_space - check if there is enough free space to mount.
1136 * @c: UBIFS file-system description object
1137 *
1138 * This function makes sure UBIFS has enough free space to be mounted in
1139 * read/write mode. UBIFS must always have some free space to allow deletions.
1140 */
1141static int check_free_space(struct ubifs_info *c)
1142{
1143        ubifs_assert(c->dark_wm > 0);
1144        if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1145                ubifs_err("insufficient free space to mount in R/W mode");
1146                ubifs_dump_budg(c, &c->bi);
1147                ubifs_dump_lprops(c);
1148                return -ENOSPC;
1149        }
1150        return 0;
1151}
1152
1153/**
1154 * mount_ubifs - mount UBIFS file-system.
1155 * @c: UBIFS file-system description object
1156 *
1157 * This function mounts UBIFS file system. Returns zero in case of success and
1158 * a negative error code in case of failure.
1159 */
1160static int mount_ubifs(struct ubifs_info *c)
1161{
1162        int err;
1163        long long x, y;
1164        size_t sz;
1165
1166        c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1167        err = init_constants_early(c);
1168        if (err)
1169                return err;
1170
1171        err = ubifs_debugging_init(c);
1172        if (err)
1173                return err;
1174
1175        err = check_volume_empty(c);
1176        if (err)
1177                goto out_free;
1178
1179        if (c->empty && (c->ro_mount || c->ro_media)) {
1180                /*
1181                 * This UBI volume is empty, and read-only, or the file system
1182                 * is mounted read-only - we cannot format it.
1183                 */
1184                ubifs_err("can't format empty UBI volume: read-only %s",
1185                          c->ro_media ? "UBI volume" : "mount");
1186                err = -EROFS;
1187                goto out_free;
1188        }
1189
1190        if (c->ro_media && !c->ro_mount) {
1191                ubifs_err("cannot mount read-write - read-only media");
1192                err = -EROFS;
1193                goto out_free;
1194        }
1195
1196        /*
1197         * The requirement for the buffer is that it should fit indexing B-tree
1198         * height amount of integers. We assume the height if the TNC tree will
1199         * never exceed 64.
1200         */
1201        err = -ENOMEM;
1202        c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1203        if (!c->bottom_up_buf)
1204                goto out_free;
1205
1206        c->sbuf = vmalloc(c->leb_size);
1207        if (!c->sbuf)
1208                goto out_free;
1209
1210        if (!c->ro_mount) {
1211                c->ileb_buf = vmalloc(c->leb_size);
1212                if (!c->ileb_buf)
1213                        goto out_free;
1214        }
1215
1216        if (c->bulk_read == 1)
1217                bu_init(c);
1218
1219        if (!c->ro_mount) {
1220                c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
1221                                               GFP_KERNEL);
1222                if (!c->write_reserve_buf)
1223                        goto out_free;
1224        }
1225
1226        c->mounting = 1;
1227
1228        err = ubifs_read_superblock(c);
1229        if (err)
1230                goto out_free;
1231
1232        /*
1233         * Make sure the compressor which is set as default in the superblock
1234         * or overridden by mount options is actually compiled in.
1235         */
1236        if (!ubifs_compr_present(c->default_compr)) {
1237                ubifs_err("'compressor \"%s\" is not compiled in",
1238                          ubifs_compr_name(c->default_compr));
1239                err = -ENOTSUPP;
1240                goto out_free;
1241        }
1242
1243        err = init_constants_sb(c);
1244        if (err)
1245                goto out_free;
1246
1247        sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1248        sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1249        c->cbuf = kmalloc(sz, GFP_NOFS);
1250        if (!c->cbuf) {
1251                err = -ENOMEM;
1252                goto out_free;
1253        }
1254
1255        err = alloc_wbufs(c);
1256        if (err)
1257                goto out_cbuf;
1258
1259        sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1260        if (!c->ro_mount) {
1261                /* Create background thread */
1262                c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1263                if (IS_ERR(c->bgt)) {
1264                        err = PTR_ERR(c->bgt);
1265                        c->bgt = NULL;
1266                        ubifs_err("cannot spawn \"%s\", error %d",
1267                                  c->bgt_name, err);
1268                        goto out_wbufs;
1269                }
1270                wake_up_process(c->bgt);
1271        }
1272
1273        err = ubifs_read_master(c);
1274        if (err)
1275                goto out_master;
1276
1277        init_constants_master(c);
1278
1279        if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1280                ubifs_msg("recovery needed");
1281                c->need_recovery = 1;
1282        }
1283
1284        if (c->need_recovery && !c->ro_mount) {
1285                err = ubifs_recover_inl_heads(c, c->sbuf);
1286                if (err)
1287                        goto out_master;
1288        }
1289
1290        err = ubifs_lpt_init(c, 1, !c->ro_mount);
1291        if (err)
1292                goto out_master;
1293
1294        if (!c->ro_mount && c->space_fixup) {
1295                err = ubifs_fixup_free_space(c);
1296                if (err)
1297                        goto out_lpt;
1298        }
1299
1300        if (!c->ro_mount) {
1301                /*
1302                 * Set the "dirty" flag so that if we reboot uncleanly we
1303                 * will notice this immediately on the next mount.
1304                 */
1305                c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1306                err = ubifs_write_master(c);
1307                if (err)
1308                        goto out_lpt;
1309        }
1310
1311        err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1312        if (err)
1313                goto out_lpt;
1314
1315        err = ubifs_replay_journal(c);
1316        if (err)
1317                goto out_journal;
1318
1319        /* Calculate 'min_idx_lebs' after journal replay */
1320        c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1321
1322        err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1323        if (err)
1324                goto out_orphans;
1325
1326        if (!c->ro_mount) {
1327                int lnum;
1328
1329                err = check_free_space(c);
1330                if (err)
1331                        goto out_orphans;
1332
1333                /* Check for enough log space */
1334                lnum = c->lhead_lnum + 1;
1335                if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1336                        lnum = UBIFS_LOG_LNUM;
1337                if (lnum == c->ltail_lnum) {
1338                        err = ubifs_consolidate_log(c);
1339                        if (err)
1340                                goto out_orphans;
1341                }
1342
1343                if (c->need_recovery) {
1344                        err = ubifs_recover_size(c);
1345                        if (err)
1346                                goto out_orphans;
1347                        err = ubifs_rcvry_gc_commit(c);
1348                        if (err)
1349                                goto out_orphans;
1350                } else {
1351                        err = take_gc_lnum(c);
1352                        if (err)
1353                                goto out_orphans;
1354
1355                        /*
1356                         * GC LEB may contain garbage if there was an unclean
1357                         * reboot, and it should be un-mapped.
1358                         */
1359                        err = ubifs_leb_unmap(c, c->gc_lnum);
1360                        if (err)
1361                                goto out_orphans;
1362                }
1363
1364                err = dbg_check_lprops(c);
1365                if (err)
1366                        goto out_orphans;
1367        } else if (c->need_recovery) {
1368                err = ubifs_recover_size(c);
1369                if (err)
1370                        goto out_orphans;
1371        } else {
1372                /*
1373                 * Even if we mount read-only, we have to set space in GC LEB
1374                 * to proper value because this affects UBIFS free space
1375                 * reporting. We do not want to have a situation when
1376                 * re-mounting from R/O to R/W changes amount of free space.
1377                 */
1378                err = take_gc_lnum(c);
1379                if (err)
1380                        goto out_orphans;
1381        }
1382
1383        spin_lock(&ubifs_infos_lock);
1384        list_add_tail(&c->infos_list, &ubifs_infos);
1385        spin_unlock(&ubifs_infos_lock);
1386
1387        if (c->need_recovery) {
1388                if (c->ro_mount)
1389                        ubifs_msg("recovery deferred");
1390                else {
1391                        c->need_recovery = 0;
1392                        ubifs_msg("recovery completed");
1393                        /*
1394                         * GC LEB has to be empty and taken at this point. But
1395                         * the journal head LEBs may also be accounted as
1396                         * "empty taken" if they are empty.
1397                         */
1398                        ubifs_assert(c->lst.taken_empty_lebs > 0);
1399                }
1400        } else
1401                ubifs_assert(c->lst.taken_empty_lebs > 0);
1402
1403        err = dbg_check_filesystem(c);
1404        if (err)
1405                goto out_infos;
1406
1407        err = dbg_debugfs_init_fs(c);
1408        if (err)
1409                goto out_infos;
1410
1411        c->mounting = 0;
1412
1413        ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"%s",
1414                  c->vi.ubi_num, c->vi.vol_id, c->vi.name,
1415                  c->ro_mount ? ", R/O mode" : "");
1416        x = (long long)c->main_lebs * c->leb_size;
1417        y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1418        ubifs_msg("LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes",
1419                  c->leb_size, c->leb_size >> 10, c->min_io_size,
1420                  c->max_write_size);
1421        ubifs_msg("FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)",
1422                  x, x >> 20, c->main_lebs,
1423                  y, y >> 20, c->log_lebs + c->max_bud_cnt);
1424        ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1425                  c->report_rp_size, c->report_rp_size >> 10);
1426        ubifs_msg("media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s",
1427                  c->fmt_version, c->ro_compat_version,
1428                  UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid,
1429                  c->big_lpt ? ", big LPT model" : ", small LPT model");
1430
1431        dbg_gen("default compressor:  %s", ubifs_compr_name(c->default_compr));
1432        dbg_gen("data journal heads:  %d",
1433                c->jhead_cnt - NONDATA_JHEADS_CNT);
1434        dbg_gen("log LEBs:            %d (%d - %d)",
1435                c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1436        dbg_gen("LPT area LEBs:       %d (%d - %d)",
1437                c->lpt_lebs, c->lpt_first, c->lpt_last);
1438        dbg_gen("orphan area LEBs:    %d (%d - %d)",
1439                c->orph_lebs, c->orph_first, c->orph_last);
1440        dbg_gen("main area LEBs:      %d (%d - %d)",
1441                c->main_lebs, c->main_first, c->leb_cnt - 1);
1442        dbg_gen("index LEBs:          %d", c->lst.idx_lebs);
1443        dbg_gen("total index bytes:   %lld (%lld KiB, %lld MiB)",
1444                c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1445                c->bi.old_idx_sz >> 20);
1446        dbg_gen("key hash type:       %d", c->key_hash_type);
1447        dbg_gen("tree fanout:         %d", c->fanout);
1448        dbg_gen("reserved GC LEB:     %d", c->gc_lnum);
1449        dbg_gen("max. znode size      %d", c->max_znode_sz);
1450        dbg_gen("max. index node size %d", c->max_idx_node_sz);
1451        dbg_gen("node sizes:          data %zu, inode %zu, dentry %zu",
1452                UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1453        dbg_gen("node sizes:          trun %zu, sb %zu, master %zu",
1454                UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1455        dbg_gen("node sizes:          ref %zu, cmt. start %zu, orph %zu",
1456                UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1457        dbg_gen("max. node sizes:     data %zu, inode %zu dentry %zu, idx %d",
1458                UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1459                UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1460        dbg_gen("dead watermark:      %d", c->dead_wm);
1461        dbg_gen("dark watermark:      %d", c->dark_wm);
1462        dbg_gen("LEB overhead:        %d", c->leb_overhead);
1463        x = (long long)c->main_lebs * c->dark_wm;
1464        dbg_gen("max. dark space:     %lld (%lld KiB, %lld MiB)",
1465                x, x >> 10, x >> 20);
1466        dbg_gen("maximum bud bytes:   %lld (%lld KiB, %lld MiB)",
1467                c->max_bud_bytes, c->max_bud_bytes >> 10,
1468                c->max_bud_bytes >> 20);
1469        dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1470                c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1471                c->bg_bud_bytes >> 20);
1472        dbg_gen("current bud bytes    %lld (%lld KiB, %lld MiB)",
1473                c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1474        dbg_gen("max. seq. number:    %llu", c->max_sqnum);
1475        dbg_gen("commit number:       %llu", c->cmt_no);
1476
1477        return 0;
1478
1479out_infos:
1480        spin_lock(&ubifs_infos_lock);
1481        list_del(&c->infos_list);
1482        spin_unlock(&ubifs_infos_lock);
1483out_orphans:
1484        free_orphans(c);
1485out_journal:
1486        destroy_journal(c);
1487out_lpt:
1488        ubifs_lpt_free(c, 0);
1489out_master:
1490        kfree(c->mst_node);
1491        kfree(c->rcvrd_mst_node);
1492        if (c->bgt)
1493                kthread_stop(c->bgt);
1494out_wbufs:
1495        free_wbufs(c);
1496out_cbuf:
1497        kfree(c->cbuf);
1498out_free:
1499        kfree(c->write_reserve_buf);
1500        kfree(c->bu.buf);
1501        vfree(c->ileb_buf);
1502        vfree(c->sbuf);
1503        kfree(c->bottom_up_buf);
1504        ubifs_debugging_exit(c);
1505        return err;
1506}
1507
1508/**
1509 * ubifs_umount - un-mount UBIFS file-system.
1510 * @c: UBIFS file-system description object
1511 *
1512 * Note, this function is called to free allocated resourced when un-mounting,
1513 * as well as free resources when an error occurred while we were half way
1514 * through mounting (error path cleanup function). So it has to make sure the
1515 * resource was actually allocated before freeing it.
1516 */
1517static void ubifs_umount(struct ubifs_info *c)
1518{
1519        dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1520                c->vi.vol_id);
1521
1522        dbg_debugfs_exit_fs(c);
1523        spin_lock(&ubifs_infos_lock);
1524        list_del(&c->infos_list);
1525        spin_unlock(&ubifs_infos_lock);
1526
1527        if (c->bgt)
1528                kthread_stop(c->bgt);
1529
1530        destroy_journal(c);
1531        free_wbufs(c);
1532        free_orphans(c);
1533        ubifs_lpt_free(c, 0);
1534
1535        kfree(c->cbuf);
1536        kfree(c->rcvrd_mst_node);
1537        kfree(c->mst_node);
1538        kfree(c->write_reserve_buf);
1539        kfree(c->bu.buf);
1540        vfree(c->ileb_buf);
1541        vfree(c->sbuf);
1542        kfree(c->bottom_up_buf);
1543        ubifs_debugging_exit(c);
1544}
1545
1546/**
1547 * ubifs_remount_rw - re-mount in read-write mode.
1548 * @c: UBIFS file-system description object
1549 *
1550 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1551 * mode. This function allocates the needed resources and re-mounts UBIFS in
1552 * read-write mode.
1553 */
1554static int ubifs_remount_rw(struct ubifs_info *c)
1555{
1556        int err, lnum;
1557
1558        if (c->rw_incompat) {
1559                ubifs_err("the file-system is not R/W-compatible");
1560                ubifs_msg("on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
1561                          c->fmt_version, c->ro_compat_version,
1562                          UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1563                return -EROFS;
1564        }
1565
1566        mutex_lock(&c->umount_mutex);
1567        dbg_save_space_info(c);
1568        c->remounting_rw = 1;
1569        c->ro_mount = 0;
1570
1571        if (c->space_fixup) {
1572                err = ubifs_fixup_free_space(c);
1573                if (err)
1574                        return err;
1575        }
1576
1577        err = check_free_space(c);
1578        if (err)
1579                goto out;
1580
1581        if (c->old_leb_cnt != c->leb_cnt) {
1582                struct ubifs_sb_node *sup;
1583
1584                sup = ubifs_read_sb_node(c);
1585                if (IS_ERR(sup)) {
1586                        err = PTR_ERR(sup);
1587                        goto out;
1588                }
1589                sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1590                err = ubifs_write_sb_node(c, sup);
1591                kfree(sup);
1592                if (err)
1593                        goto out;
1594        }
1595
1596        if (c->need_recovery) {
1597                ubifs_msg("completing deferred recovery");
1598                err = ubifs_write_rcvrd_mst_node(c);
1599                if (err)
1600                        goto out;
1601                err = ubifs_recover_size(c);
1602                if (err)
1603                        goto out;
1604                err = ubifs_clean_lebs(c, c->sbuf);
1605                if (err)
1606                        goto out;
1607                err = ubifs_recover_inl_heads(c, c->sbuf);
1608                if (err)
1609                        goto out;
1610        } else {
1611                /* A readonly mount is not allowed to have orphans */
1612                ubifs_assert(c->tot_orphans == 0);
1613                err = ubifs_clear_orphans(c);
1614                if (err)
1615                        goto out;
1616        }
1617
1618        if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1619                c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1620                err = ubifs_write_master(c);
1621                if (err)
1622                        goto out;
1623        }
1624
1625        c->ileb_buf = vmalloc(c->leb_size);
1626        if (!c->ileb_buf) {
1627                err = -ENOMEM;
1628                goto out;
1629        }
1630
1631        c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
1632        if (!c->write_reserve_buf)
1633                goto out;
1634
1635        err = ubifs_lpt_init(c, 0, 1);
1636        if (err)
1637                goto out;
1638
1639        /* Create background thread */
1640        c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1641        if (IS_ERR(c->bgt)) {
1642                err = PTR_ERR(c->bgt);
1643                c->bgt = NULL;
1644                ubifs_err("cannot spawn \"%s\", error %d",
1645                          c->bgt_name, err);
1646                goto out;
1647        }
1648        wake_up_process(c->bgt);
1649
1650        c->orph_buf = vmalloc(c->leb_size);
1651        if (!c->orph_buf) {
1652                err = -ENOMEM;
1653                goto out;
1654        }
1655
1656        /* Check for enough log space */
1657        lnum = c->lhead_lnum + 1;
1658        if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1659                lnum = UBIFS_LOG_LNUM;
1660        if (lnum == c->ltail_lnum) {
1661                err = ubifs_consolidate_log(c);
1662                if (err)
1663                        goto out;
1664        }
1665
1666        if (c->need_recovery)
1667                err = ubifs_rcvry_gc_commit(c);
1668        else
1669                err = ubifs_leb_unmap(c, c->gc_lnum);
1670        if (err)
1671                goto out;
1672
1673        dbg_gen("re-mounted read-write");
1674        c->remounting_rw = 0;
1675
1676        if (c->need_recovery) {
1677                c->need_recovery = 0;
1678                ubifs_msg("deferred recovery completed");
1679        } else {
1680                /*
1681                 * Do not run the debugging space check if the were doing
1682                 * recovery, because when we saved the information we had the
1683                 * file-system in a state where the TNC and lprops has been
1684                 * modified in memory, but all the I/O operations (including a
1685                 * commit) were deferred. So the file-system was in
1686                 * "non-committed" state. Now the file-system is in committed
1687                 * state, and of course the amount of free space will change
1688                 * because, for example, the old index size was imprecise.
1689                 */
1690                err = dbg_check_space_info(c);
1691        }
1692
1693        mutex_unlock(&c->umount_mutex);
1694        return err;
1695
1696out:
1697        c->ro_mount = 1;
1698        vfree(c->orph_buf);
1699        c->orph_buf = NULL;
1700        if (c->bgt) {
1701                kthread_stop(c->bgt);
1702                c->bgt = NULL;
1703        }
1704        free_wbufs(c);
1705        kfree(c->write_reserve_buf);
1706        c->write_reserve_buf = NULL;
1707        vfree(c->ileb_buf);
1708        c->ileb_buf = NULL;
1709        ubifs_lpt_free(c, 1);
1710        c->remounting_rw = 0;
1711        mutex_unlock(&c->umount_mutex);
1712        return err;
1713}
1714
1715/**
1716 * ubifs_remount_ro - re-mount in read-only mode.
1717 * @c: UBIFS file-system description object
1718 *
1719 * We assume VFS has stopped writing. Possibly the background thread could be
1720 * running a commit, however kthread_stop will wait in that case.
1721 */
1722static void ubifs_remount_ro(struct ubifs_info *c)
1723{
1724        int i, err;
1725
1726        ubifs_assert(!c->need_recovery);
1727        ubifs_assert(!c->ro_mount);
1728
1729        mutex_lock(&c->umount_mutex);
1730        if (c->bgt) {
1731                kthread_stop(c->bgt);
1732                c->bgt = NULL;
1733        }
1734
1735        dbg_save_space_info(c);
1736
1737        for (i = 0; i < c->jhead_cnt; i++)
1738                ubifs_wbuf_sync(&c->jheads[i].wbuf);
1739
1740        c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1741        c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1742        c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1743        err = ubifs_write_master(c);
1744        if (err)
1745                ubifs_ro_mode(c, err);
1746
1747        vfree(c->orph_buf);
1748        c->orph_buf = NULL;
1749        kfree(c->write_reserve_buf);
1750        c->write_reserve_buf = NULL;
1751        vfree(c->ileb_buf);
1752        c->ileb_buf = NULL;
1753        ubifs_lpt_free(c, 1);
1754        c->ro_mount = 1;
1755        err = dbg_check_space_info(c);
1756        if (err)
1757                ubifs_ro_mode(c, err);
1758        mutex_unlock(&c->umount_mutex);
1759}
1760
1761static void ubifs_put_super(struct super_block *sb)
1762{
1763        int i;
1764        struct ubifs_info *c = sb->s_fs_info;
1765
1766        ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1767                  c->vi.vol_id);
1768
1769        /*
1770         * The following asserts are only valid if there has not been a failure
1771         * of the media. For example, there will be dirty inodes if we failed
1772         * to write them back because of I/O errors.
1773         */
1774        if (!c->ro_error) {
1775                ubifs_assert(c->bi.idx_growth == 0);
1776                ubifs_assert(c->bi.dd_growth == 0);
1777                ubifs_assert(c->bi.data_growth == 0);
1778        }
1779
1780        /*
1781         * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1782         * and file system un-mount. Namely, it prevents the shrinker from
1783         * picking this superblock for shrinking - it will be just skipped if
1784         * the mutex is locked.
1785         */
1786        mutex_lock(&c->umount_mutex);
1787        if (!c->ro_mount) {
1788                /*
1789                 * First of all kill the background thread to make sure it does
1790                 * not interfere with un-mounting and freeing resources.
1791                 */
1792                if (c->bgt) {
1793                        kthread_stop(c->bgt);
1794                        c->bgt = NULL;
1795                }
1796
1797                /*
1798                 * On fatal errors c->ro_error is set to 1, in which case we do
1799                 * not write the master node.
1800                 */
1801                if (!c->ro_error) {
1802                        int err;
1803
1804                        /* Synchronize write-buffers */
1805                        for (i = 0; i < c->jhead_cnt; i++)
1806                                ubifs_wbuf_sync(&c->jheads[i].wbuf);
1807
1808                        /*
1809                         * We are being cleanly unmounted which means the
1810                         * orphans were killed - indicate this in the master
1811                         * node. Also save the reserved GC LEB number.
1812                         */
1813                        c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1814                        c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1815                        c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1816                        err = ubifs_write_master(c);
1817                        if (err)
1818                                /*
1819                                 * Recovery will attempt to fix the master area
1820                                 * next mount, so we just print a message and
1821                                 * continue to unmount normally.
1822                                 */
1823                                ubifs_err("failed to write master node, error %d",
1824                                          err);
1825                } else {
1826                        for (i = 0; i < c->jhead_cnt; i++)
1827                                /* Make sure write-buffer timers are canceled */
1828                                hrtimer_cancel(&c->jheads[i].wbuf.timer);
1829                }
1830        }
1831
1832        ubifs_umount(c);
1833        bdi_destroy(&c->bdi);
1834        ubi_close_volume(c->ubi);
1835        mutex_unlock(&c->umount_mutex);
1836}
1837
1838static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1839{
1840        int err;
1841        struct ubifs_info *c = sb->s_fs_info;
1842
1843        dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1844
1845        err = ubifs_parse_options(c, data, 1);
1846        if (err) {
1847                ubifs_err("invalid or unknown remount parameter");
1848                return err;
1849        }
1850
1851        if (c->ro_mount && !(*flags & MS_RDONLY)) {
1852                if (c->ro_error) {
1853                        ubifs_msg("cannot re-mount R/W due to prior errors");
1854                        return -EROFS;
1855                }
1856                if (c->ro_media) {
1857                        ubifs_msg("cannot re-mount R/W - UBI volume is R/O");
1858                        return -EROFS;
1859                }
1860                err = ubifs_remount_rw(c);
1861                if (err)
1862                        return err;
1863        } else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1864                if (c->ro_error) {
1865                        ubifs_msg("cannot re-mount R/O due to prior errors");
1866                        return -EROFS;
1867                }
1868                ubifs_remount_ro(c);
1869        }
1870
1871        if (c->bulk_read == 1)
1872                bu_init(c);
1873        else {
1874                dbg_gen("disable bulk-read");
1875                kfree(c->bu.buf);
1876                c->bu.buf = NULL;
1877        }
1878
1879        ubifs_assert(c->lst.taken_empty_lebs > 0);
1880        return 0;
1881}
1882
1883const struct super_operations ubifs_super_operations = {
1884        .alloc_inode   = ubifs_alloc_inode,
1885        .destroy_inode = ubifs_destroy_inode,
1886        .put_super     = ubifs_put_super,
1887        .write_inode   = ubifs_write_inode,
1888        .evict_inode   = ubifs_evict_inode,
1889        .statfs        = ubifs_statfs,
1890        .dirty_inode   = ubifs_dirty_inode,
1891        .remount_fs    = ubifs_remount_fs,
1892        .show_options  = ubifs_show_options,
1893        .sync_fs       = ubifs_sync_fs,
1894};
1895
1896/**
1897 * open_ubi - parse UBI device name string and open the UBI device.
1898 * @name: UBI volume name
1899 * @mode: UBI volume open mode
1900 *
1901 * The primary method of mounting UBIFS is by specifying the UBI volume
1902 * character device node path. However, UBIFS may also be mounted withoug any
1903 * character device node using one of the following methods:
1904 *
1905 * o ubiX_Y    - mount UBI device number X, volume Y;
1906 * o ubiY      - mount UBI device number 0, volume Y;
1907 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1908 * o ubi:NAME  - mount UBI device 0, volume with name NAME.
1909 *
1910 * Alternative '!' separator may be used instead of ':' (because some shells
1911 * like busybox may interpret ':' as an NFS host name separator). This function
1912 * returns UBI volume description object in case of success and a negative
1913 * error code in case of failure.
1914 */
1915static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1916{
1917        struct ubi_volume_desc *ubi;
1918        int dev, vol;
1919        char *endptr;
1920
1921        /* First, try to open using the device node path method */
1922        ubi = ubi_open_volume_path(name, mode);
1923        if (!IS_ERR(ubi))
1924                return ubi;
1925
1926        /* Try the "nodev" method */
1927        if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1928                return ERR_PTR(-EINVAL);
1929
1930        /* ubi:NAME method */
1931        if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1932                return ubi_open_volume_nm(0, name + 4, mode);
1933
1934        if (!isdigit(name[3]))
1935                return ERR_PTR(-EINVAL);
1936
1937        dev = simple_strtoul(name + 3, &endptr, 0);
1938
1939        /* ubiY method */
1940        if (*endptr == '\0')
1941                return ubi_open_volume(0, dev, mode);
1942
1943        /* ubiX_Y method */
1944        if (*endptr == '_' && isdigit(endptr[1])) {
1945                vol = simple_strtoul(endptr + 1, &endptr, 0);
1946                if (*endptr != '\0')
1947                        return ERR_PTR(-EINVAL);
1948                return ubi_open_volume(dev, vol, mode);
1949        }
1950
1951        /* ubiX:NAME method */
1952        if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1953                return ubi_open_volume_nm(dev, ++endptr, mode);
1954
1955        return ERR_PTR(-EINVAL);
1956}
1957
1958static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1959{
1960        struct ubifs_info *c;
1961
1962        c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1963        if (c) {
1964                spin_lock_init(&c->cnt_lock);
1965                spin_lock_init(&c->cs_lock);
1966                spin_lock_init(&c->buds_lock);
1967                spin_lock_init(&c->space_lock);
1968                spin_lock_init(&c->orphan_lock);
1969                init_rwsem(&c->commit_sem);
1970                mutex_init(&c->lp_mutex);
1971                mutex_init(&c->tnc_mutex);
1972                mutex_init(&c->log_mutex);
1973                mutex_init(&c->mst_mutex);
1974                mutex_init(&c->umount_mutex);
1975                mutex_init(&c->bu_mutex);
1976                mutex_init(&c->write_reserve_mutex);
1977                init_waitqueue_head(&c->cmt_wq);
1978                c->buds = RB_ROOT;
1979                c->old_idx = RB_ROOT;
1980                c->size_tree = RB_ROOT;
1981                c->orph_tree = RB_ROOT;
1982                INIT_LIST_HEAD(&c->infos_list);
1983                INIT_LIST_HEAD(&c->idx_gc);
1984                INIT_LIST_HEAD(&c->replay_list);
1985                INIT_LIST_HEAD(&c->replay_buds);
1986                INIT_LIST_HEAD(&c->uncat_list);
1987                INIT_LIST_HEAD(&c->empty_list);
1988                INIT_LIST_HEAD(&c->freeable_list);
1989                INIT_LIST_HEAD(&c->frdi_idx_list);
1990                INIT_LIST_HEAD(&c->unclean_leb_list);
1991                INIT_LIST_HEAD(&c->old_buds);
1992                INIT_LIST_HEAD(&c->orph_list);
1993                INIT_LIST_HEAD(&c->orph_new);
1994                c->no_chk_data_crc = 1;
1995
1996                c->highest_inum = UBIFS_FIRST_INO;
1997                c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1998
1999                ubi_get_volume_info(ubi, &c->vi);
2000                ubi_get_device_info(c->vi.ubi_num, &c->di);
2001        }
2002        return c;
2003}
2004
2005static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
2006{
2007        struct ubifs_info *c = sb->s_fs_info;
2008        struct inode *root;
2009        int err;
2010
2011        c->vfs_sb = sb;
2012        /* Re-open the UBI device in read-write mode */
2013        c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2014        if (IS_ERR(c->ubi)) {
2015                err = PTR_ERR(c->ubi);
2016                goto out;
2017        }
2018
2019        /*
2020         * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2021         * UBIFS, I/O is not deferred, it is done immediately in readpage,
2022         * which means the user would have to wait not just for their own I/O
2023         * but the read-ahead I/O as well i.e. completely pointless.
2024         *
2025         * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
2026         */
2027        c->bdi.name = "ubifs",
2028        c->bdi.capabilities = BDI_CAP_MAP_COPY;
2029        err  = bdi_init(&c->bdi);
2030        if (err)
2031                goto out_close;
2032        err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
2033                           c->vi.ubi_num, c->vi.vol_id);
2034        if (err)
2035                goto out_bdi;
2036
2037        err = ubifs_parse_options(c, data, 0);
2038        if (err)
2039                goto out_bdi;
2040
2041        sb->s_bdi = &c->bdi;
2042        sb->s_fs_info = c;
2043        sb->s_magic = UBIFS_SUPER_MAGIC;
2044        sb->s_blocksize = UBIFS_BLOCK_SIZE;
2045        sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2046        sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2047        if (c->max_inode_sz > MAX_LFS_FILESIZE)
2048                sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2049        sb->s_op = &ubifs_super_operations;
2050
2051        mutex_lock(&c->umount_mutex);
2052        err = mount_ubifs(c);
2053        if (err) {
2054                ubifs_assert(err < 0);
2055                goto out_unlock;
2056        }
2057
2058        /* Read the root inode */
2059        root = ubifs_iget(sb, UBIFS_ROOT_INO);
2060        if (IS_ERR(root)) {
2061                err = PTR_ERR(root);
2062                goto out_umount;
2063        }
2064
2065        sb->s_root = d_make_root(root);
2066        if (!sb->s_root)
2067                goto out_umount;
2068
2069        mutex_unlock(&c->umount_mutex);
2070        return 0;
2071
2072out_umount:
2073        ubifs_umount(c);
2074out_unlock:
2075        mutex_unlock(&c->umount_mutex);
2076out_bdi:
2077        bdi_destroy(&c->bdi);
2078out_close:
2079        ubi_close_volume(c->ubi);
2080out:
2081        return err;
2082}
2083
2084static int sb_test(struct super_block *sb, void *data)
2085{
2086        struct ubifs_info *c1 = data;
2087        struct ubifs_info *c = sb->s_fs_info;
2088
2089        return c->vi.cdev == c1->vi.cdev;
2090}
2091
2092static int sb_set(struct super_block *sb, void *data)
2093{
2094        sb->s_fs_info = data;
2095        return set_anon_super(sb, NULL);
2096}
2097
2098static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2099                        const char *name, void *data)
2100{
2101        struct ubi_volume_desc *ubi;
2102        struct ubifs_info *c;
2103        struct super_block *sb;
2104        int err;
2105
2106        dbg_gen("name %s, flags %#x", name, flags);
2107
2108        /*
2109         * Get UBI device number and volume ID. Mount it read-only so far
2110         * because this might be a new mount point, and UBI allows only one
2111         * read-write user at a time.
2112         */
2113        ubi = open_ubi(name, UBI_READONLY);
2114        if (IS_ERR(ubi)) {
2115                ubifs_err("cannot open \"%s\", error %d",
2116                          name, (int)PTR_ERR(ubi));
2117                return ERR_CAST(ubi);
2118        }
2119
2120        c = alloc_ubifs_info(ubi);
2121        if (!c) {
2122                err = -ENOMEM;
2123                goto out_close;
2124        }
2125
2126        dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2127
2128        sb = sget(fs_type, sb_test, sb_set, flags, c);
2129        if (IS_ERR(sb)) {
2130                err = PTR_ERR(sb);
2131                kfree(c);
2132                goto out_close;
2133        }
2134
2135        if (sb->s_root) {
2136                struct ubifs_info *c1 = sb->s_fs_info;
2137                kfree(c);
2138                /* A new mount point for already mounted UBIFS */
2139                dbg_gen("this ubi volume is already mounted");
2140                if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2141                        err = -EBUSY;
2142                        goto out_deact;
2143                }
2144        } else {
2145                err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2146                if (err)
2147                        goto out_deact;
2148                /* We do not support atime */
2149                sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2150        }
2151
2152        /* 'fill_super()' opens ubi again so we must close it here */
2153        ubi_close_volume(ubi);
2154
2155        return dget(sb->s_root);
2156
2157out_deact:
2158        deactivate_locked_super(sb);
2159out_close:
2160        ubi_close_volume(ubi);
2161        return ERR_PTR(err);
2162}
2163
2164static void kill_ubifs_super(struct super_block *s)
2165{
2166        struct ubifs_info *c = s->s_fs_info;
2167        kill_anon_super(s);
2168        kfree(c);
2169}
2170
2171static struct file_system_type ubifs_fs_type = {
2172        .name    = "ubifs",
2173        .owner   = THIS_MODULE,
2174        .mount   = ubifs_mount,
2175        .kill_sb = kill_ubifs_super,
2176};
2177MODULE_ALIAS_FS("ubifs");
2178
2179/*
2180 * Inode slab cache constructor.
2181 */
2182static void inode_slab_ctor(void *obj)
2183{
2184        struct ubifs_inode *ui = obj;
2185        inode_init_once(&ui->vfs_inode);
2186}
2187
2188static int __init ubifs_init(void)
2189{
2190        int err;
2191
2192        BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2193
2194        /* Make sure node sizes are 8-byte aligned */
2195        BUILD_BUG_ON(UBIFS_CH_SZ        & 7);
2196        BUILD_BUG_ON(UBIFS_INO_NODE_SZ  & 7);
2197        BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2198        BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2199        BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2200        BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2201        BUILD_BUG_ON(UBIFS_SB_NODE_SZ   & 7);
2202        BUILD_BUG_ON(UBIFS_MST_NODE_SZ  & 7);
2203        BUILD_BUG_ON(UBIFS_REF_NODE_SZ  & 7);
2204        BUILD_BUG_ON(UBIFS_CS_NODE_SZ   & 7);
2205        BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2206
2207        BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2208        BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2209        BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2210        BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  & 7);
2211        BUILD_BUG_ON(UBIFS_MAX_NODE_SZ      & 7);
2212        BUILD_BUG_ON(MIN_WRITE_SZ           & 7);
2213
2214        /* Check min. node size */
2215        BUILD_BUG_ON(UBIFS_INO_NODE_SZ  < MIN_WRITE_SZ);
2216        BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2217        BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2218        BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2219
2220        BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2221        BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2222        BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2223        BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  > UBIFS_MAX_NODE_SZ);
2224
2225        /* Defined node sizes */
2226        BUILD_BUG_ON(UBIFS_SB_NODE_SZ  != 4096);
2227        BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2228        BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2229        BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2230
2231        /*
2232         * We use 2 bit wide bit-fields to store compression type, which should
2233         * be amended if more compressors are added. The bit-fields are:
2234         * @compr_type in 'struct ubifs_inode', @default_compr in
2235         * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2236         */
2237        BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2238
2239        /*
2240         * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2241         * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2242         */
2243        if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2244                ubifs_err("VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
2245                          (unsigned int)PAGE_CACHE_SIZE);
2246                return -EINVAL;
2247        }
2248
2249        ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2250                                sizeof(struct ubifs_inode), 0,
2251                                SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2252                                &inode_slab_ctor);
2253        if (!ubifs_inode_slab)
2254                return -ENOMEM;
2255
2256        register_shrinker(&ubifs_shrinker_info);
2257
2258        err = ubifs_compressors_init();
2259        if (err)
2260                goto out_shrinker;
2261
2262        err = dbg_debugfs_init();
2263        if (err)
2264                goto out_compr;
2265
2266        err = register_filesystem(&ubifs_fs_type);
2267        if (err) {
2268                ubifs_err("cannot register file system, error %d", err);
2269                goto out_dbg;
2270        }
2271        return 0;
2272
2273out_dbg:
2274        dbg_debugfs_exit();
2275out_compr:
2276        ubifs_compressors_exit();
2277out_shrinker:
2278        unregister_shrinker(&ubifs_shrinker_info);
2279        kmem_cache_destroy(ubifs_inode_slab);
2280        return err;
2281}
2282/* late_initcall to let compressors initialize first */
2283late_initcall(ubifs_init);
2284
2285static void __exit ubifs_exit(void)
2286{
2287        ubifs_assert(list_empty(&ubifs_infos));
2288        ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2289
2290        dbg_debugfs_exit();
2291        ubifs_compressors_exit();
2292        unregister_shrinker(&ubifs_shrinker_info);
2293
2294        /*
2295         * Make sure all delayed rcu free inodes are flushed before we
2296         * destroy cache.
2297         */
2298        rcu_barrier();
2299        kmem_cache_destroy(ubifs_inode_slab);
2300        unregister_filesystem(&ubifs_fs_type);
2301}
2302module_exit(ubifs_exit);
2303
2304MODULE_LICENSE("GPL");
2305MODULE_VERSION(__stringify(UBIFS_VERSION));
2306MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2307MODULE_DESCRIPTION("UBIFS - UBI File System");
2308