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 */
  48static struct kmem_cache *ubifs_inode_slab;
  49
  50/* UBIFS TNC shrinker description */
  51static struct shrinker ubifs_shrinker_info = {
  52        .scan_objects = ubifs_shrink_scan,
  53        .count_objects = ubifs_shrink_count,
  54        .seeks = DEFAULT_SEEKS,
  55};
  56
  57/**
  58 * validate_inode - validate inode.
  59 * @c: UBIFS file-system description object
  60 * @inode: the inode to validate
  61 *
  62 * This is a helper function for 'ubifs_iget()' which validates various fields
  63 * of a newly built inode to make sure they contain sane values and prevent
  64 * possible vulnerabilities. Returns zero if the inode is all right and
  65 * a non-zero error code if not.
  66 */
  67static int validate_inode(struct ubifs_info *c, const struct inode *inode)
  68{
  69        int err;
  70        const struct ubifs_inode *ui = ubifs_inode(inode);
  71
  72        if (inode->i_size > c->max_inode_sz) {
  73                ubifs_err(c, "inode is too large (%lld)",
  74                          (long long)inode->i_size);
  75                return 1;
  76        }
  77
  78        if (ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
  79                ubifs_err(c, "unknown compression type %d", ui->compr_type);
  80                return 2;
  81        }
  82
  83        if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
  84                return 3;
  85
  86        if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
  87                return 4;
  88
  89        if (ui->xattr && !S_ISREG(inode->i_mode))
  90                return 5;
  91
  92        if (!ubifs_compr_present(ui->compr_type)) {
  93                ubifs_warn(c, "inode %lu uses '%s' compression, but it was not compiled in",
  94                           inode->i_ino, ubifs_compr_name(ui->compr_type));
  95        }
  96
  97        err = dbg_check_dir(c, inode);
  98        return err;
  99}
 100
 101struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
 102{
 103        int err;
 104        union ubifs_key key;
 105        struct ubifs_ino_node *ino;
 106        struct ubifs_info *c = sb->s_fs_info;
 107        struct inode *inode;
 108        struct ubifs_inode *ui;
 109
 110        dbg_gen("inode %lu", inum);
 111
 112        inode = iget_locked(sb, inum);
 113        if (!inode)
 114                return ERR_PTR(-ENOMEM);
 115        if (!(inode->i_state & I_NEW))
 116                return inode;
 117        ui = ubifs_inode(inode);
 118
 119        ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
 120        if (!ino) {
 121                err = -ENOMEM;
 122                goto out;
 123        }
 124
 125        ino_key_init(c, &key, inode->i_ino);
 126
 127        err = ubifs_tnc_lookup(c, &key, ino);
 128        if (err)
 129                goto out_ino;
 130
 131        inode->i_flags |= S_NOCMTIME;
 132#ifndef CONFIG_UBIFS_ATIME_SUPPORT
 133        inode->i_flags |= S_NOATIME;
 134#endif
 135        set_nlink(inode, le32_to_cpu(ino->nlink));
 136        i_uid_write(inode, le32_to_cpu(ino->uid));
 137        i_gid_write(inode, le32_to_cpu(ino->gid));
 138        inode->i_atime.tv_sec  = (int64_t)le64_to_cpu(ino->atime_sec);
 139        inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
 140        inode->i_mtime.tv_sec  = (int64_t)le64_to_cpu(ino->mtime_sec);
 141        inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
 142        inode->i_ctime.tv_sec  = (int64_t)le64_to_cpu(ino->ctime_sec);
 143        inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
 144        inode->i_mode = le32_to_cpu(ino->mode);
 145        inode->i_size = le64_to_cpu(ino->size);
 146
 147        ui->data_len    = le32_to_cpu(ino->data_len);
 148        ui->flags       = le32_to_cpu(ino->flags);
 149        ui->compr_type  = le16_to_cpu(ino->compr_type);
 150        ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
 151        ui->xattr_cnt   = le32_to_cpu(ino->xattr_cnt);
 152        ui->xattr_size  = le32_to_cpu(ino->xattr_size);
 153        ui->xattr_names = le32_to_cpu(ino->xattr_names);
 154        ui->synced_i_size = ui->ui_size = inode->i_size;
 155
 156        ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
 157
 158        err = validate_inode(c, inode);
 159        if (err)
 160                goto out_invalid;
 161
 162        switch (inode->i_mode & S_IFMT) {
 163        case S_IFREG:
 164                inode->i_mapping->a_ops = &ubifs_file_address_operations;
 165                inode->i_op = &ubifs_file_inode_operations;
 166                inode->i_fop = &ubifs_file_operations;
 167                if (ui->xattr) {
 168                        ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
 169                        if (!ui->data) {
 170                                err = -ENOMEM;
 171                                goto out_ino;
 172                        }
 173                        memcpy(ui->data, ino->data, ui->data_len);
 174                        ((char *)ui->data)[ui->data_len] = '\0';
 175                } else if (ui->data_len != 0) {
 176                        err = 10;
 177                        goto out_invalid;
 178                }
 179                break;
 180        case S_IFDIR:
 181                inode->i_op  = &ubifs_dir_inode_operations;
 182                inode->i_fop = &ubifs_dir_operations;
 183                if (ui->data_len != 0) {
 184                        err = 11;
 185                        goto out_invalid;
 186                }
 187                break;
 188        case S_IFLNK:
 189                inode->i_op = &ubifs_symlink_inode_operations;
 190                if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
 191                        err = 12;
 192                        goto out_invalid;
 193                }
 194                ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
 195                if (!ui->data) {
 196                        err = -ENOMEM;
 197                        goto out_ino;
 198                }
 199                memcpy(ui->data, ino->data, ui->data_len);
 200                ((char *)ui->data)[ui->data_len] = '\0';
 201                break;
 202        case S_IFBLK:
 203        case S_IFCHR:
 204        {
 205                dev_t rdev;
 206                union ubifs_dev_desc *dev;
 207
 208                ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
 209                if (!ui->data) {
 210                        err = -ENOMEM;
 211                        goto out_ino;
 212                }
 213
 214                dev = (union ubifs_dev_desc *)ino->data;
 215                if (ui->data_len == sizeof(dev->new))
 216                        rdev = new_decode_dev(le32_to_cpu(dev->new));
 217                else if (ui->data_len == sizeof(dev->huge))
 218                        rdev = huge_decode_dev(le64_to_cpu(dev->huge));
 219                else {
 220                        err = 13;
 221                        goto out_invalid;
 222                }
 223                memcpy(ui->data, ino->data, ui->data_len);
 224                inode->i_op = &ubifs_file_inode_operations;
 225                init_special_inode(inode, inode->i_mode, rdev);
 226                break;
 227        }
 228        case S_IFSOCK:
 229        case S_IFIFO:
 230                inode->i_op = &ubifs_file_inode_operations;
 231                init_special_inode(inode, inode->i_mode, 0);
 232                if (ui->data_len != 0) {
 233                        err = 14;
 234                        goto out_invalid;
 235                }
 236                break;
 237        default:
 238                err = 15;
 239                goto out_invalid;
 240        }
 241
 242        kfree(ino);
 243        ubifs_set_inode_flags(inode);
 244        unlock_new_inode(inode);
 245        return inode;
 246
 247out_invalid:
 248        ubifs_err(c, "inode %lu validation failed, error %d", inode->i_ino, err);
 249        ubifs_dump_node(c, ino);
 250        ubifs_dump_inode(c, inode);
 251        err = -EINVAL;
 252out_ino:
 253        kfree(ino);
 254out:
 255        ubifs_err(c, "failed to read inode %lu, error %d", inode->i_ino, err);
 256        iget_failed(inode);
 257        return ERR_PTR(err);
 258}
 259
 260static struct inode *ubifs_alloc_inode(struct super_block *sb)
 261{
 262        struct ubifs_inode *ui;
 263
 264        ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
 265        if (!ui)
 266                return NULL;
 267
 268        memset((void *)ui + sizeof(struct inode), 0,
 269               sizeof(struct ubifs_inode) - sizeof(struct inode));
 270        mutex_init(&ui->ui_mutex);
 271        spin_lock_init(&ui->ui_lock);
 272        return &ui->vfs_inode;
 273};
 274
 275static void ubifs_i_callback(struct rcu_head *head)
 276{
 277        struct inode *inode = container_of(head, struct inode, i_rcu);
 278        struct ubifs_inode *ui = ubifs_inode(inode);
 279        kmem_cache_free(ubifs_inode_slab, ui);
 280}
 281
 282static void ubifs_destroy_inode(struct inode *inode)
 283{
 284        struct ubifs_inode *ui = ubifs_inode(inode);
 285
 286        kfree(ui->data);
 287        call_rcu(&inode->i_rcu, ubifs_i_callback);
 288}
 289
 290/*
 291 * Note, Linux write-back code calls this without 'i_mutex'.
 292 */
 293static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
 294{
 295        int err = 0;
 296        struct ubifs_info *c = inode->i_sb->s_fs_info;
 297        struct ubifs_inode *ui = ubifs_inode(inode);
 298
 299        ubifs_assert(!ui->xattr);
 300        if (is_bad_inode(inode))
 301                return 0;
 302
 303        mutex_lock(&ui->ui_mutex);
 304        /*
 305         * Due to races between write-back forced by budgeting
 306         * (see 'sync_some_inodes()') and background write-back, the inode may
 307         * have already been synchronized, do not do this again. This might
 308         * also happen if it was synchronized in an VFS operation, e.g.
 309         * 'ubifs_link()'.
 310         */
 311        if (!ui->dirty) {
 312                mutex_unlock(&ui->ui_mutex);
 313                return 0;
 314        }
 315
 316        /*
 317         * As an optimization, do not write orphan inodes to the media just
 318         * because this is not needed.
 319         */
 320        dbg_gen("inode %lu, mode %#x, nlink %u",
 321                inode->i_ino, (int)inode->i_mode, inode->i_nlink);
 322        if (inode->i_nlink) {
 323                err = ubifs_jnl_write_inode(c, inode);
 324                if (err)
 325                        ubifs_err(c, "can't write inode %lu, error %d",
 326                                  inode->i_ino, err);
 327                else
 328                        err = dbg_check_inode_size(c, inode, ui->ui_size);
 329        }
 330
 331        ui->dirty = 0;
 332        mutex_unlock(&ui->ui_mutex);
 333        ubifs_release_dirty_inode_budget(c, ui);
 334        return err;
 335}
 336
 337static void ubifs_evict_inode(struct inode *inode)
 338{
 339        int err;
 340        struct ubifs_info *c = inode->i_sb->s_fs_info;
 341        struct ubifs_inode *ui = ubifs_inode(inode);
 342
 343        if (ui->xattr)
 344                /*
 345                 * Extended attribute inode deletions are fully handled in
 346                 * 'ubifs_removexattr()'. These inodes are special and have
 347                 * limited usage, so there is nothing to do here.
 348                 */
 349                goto out;
 350
 351        dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
 352        ubifs_assert(!atomic_read(&inode->i_count));
 353
 354        truncate_inode_pages_final(&inode->i_data);
 355
 356        if (inode->i_nlink)
 357                goto done;
 358
 359        if (is_bad_inode(inode))
 360                goto out;
 361
 362        ui->ui_size = inode->i_size = 0;
 363        err = ubifs_jnl_delete_inode(c, inode);
 364        if (err)
 365                /*
 366                 * Worst case we have a lost orphan inode wasting space, so a
 367                 * simple error message is OK here.
 368                 */
 369                ubifs_err(c, "can't delete inode %lu, error %d",
 370                          inode->i_ino, err);
 371
 372out:
 373        if (ui->dirty)
 374                ubifs_release_dirty_inode_budget(c, ui);
 375        else {
 376                /* We've deleted something - clean the "no space" flags */
 377                c->bi.nospace = c->bi.nospace_rp = 0;
 378                smp_wmb();
 379        }
 380done:
 381        clear_inode(inode);
 382        fscrypt_put_encryption_info(inode);
 383}
 384
 385static void ubifs_dirty_inode(struct inode *inode, int flags)
 386{
 387        struct ubifs_inode *ui = ubifs_inode(inode);
 388
 389        ubifs_assert(mutex_is_locked(&ui->ui_mutex));
 390        if (!ui->dirty) {
 391                ui->dirty = 1;
 392                dbg_gen("inode %lu",  inode->i_ino);
 393        }
 394}
 395
 396static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
 397{
 398        struct ubifs_info *c = dentry->d_sb->s_fs_info;
 399        unsigned long long free;
 400        __le32 *uuid = (__le32 *)c->uuid;
 401
 402        free = ubifs_get_free_space(c);
 403        dbg_gen("free space %lld bytes (%lld blocks)",
 404                free, free >> UBIFS_BLOCK_SHIFT);
 405
 406        buf->f_type = UBIFS_SUPER_MAGIC;
 407        buf->f_bsize = UBIFS_BLOCK_SIZE;
 408        buf->f_blocks = c->block_cnt;
 409        buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
 410        if (free > c->report_rp_size)
 411                buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
 412        else
 413                buf->f_bavail = 0;
 414        buf->f_files = 0;
 415        buf->f_ffree = 0;
 416        buf->f_namelen = UBIFS_MAX_NLEN;
 417        buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
 418        buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
 419        ubifs_assert(buf->f_bfree <= c->block_cnt);
 420        return 0;
 421}
 422
 423static int ubifs_show_options(struct seq_file *s, struct dentry *root)
 424{
 425        struct ubifs_info *c = root->d_sb->s_fs_info;
 426
 427        if (c->mount_opts.unmount_mode == 2)
 428                seq_puts(s, ",fast_unmount");
 429        else if (c->mount_opts.unmount_mode == 1)
 430                seq_puts(s, ",norm_unmount");
 431
 432        if (c->mount_opts.bulk_read == 2)
 433                seq_puts(s, ",bulk_read");
 434        else if (c->mount_opts.bulk_read == 1)
 435                seq_puts(s, ",no_bulk_read");
 436
 437        if (c->mount_opts.chk_data_crc == 2)
 438                seq_puts(s, ",chk_data_crc");
 439        else if (c->mount_opts.chk_data_crc == 1)
 440                seq_puts(s, ",no_chk_data_crc");
 441
 442        if (c->mount_opts.override_compr) {
 443                seq_printf(s, ",compr=%s",
 444                           ubifs_compr_name(c->mount_opts.compr_type));
 445        }
 446
 447        seq_printf(s, ",ubi=%d,vol=%d", c->vi.ubi_num, c->vi.vol_id);
 448
 449        return 0;
 450}
 451
 452static int ubifs_sync_fs(struct super_block *sb, int wait)
 453{
 454        int i, err;
 455        struct ubifs_info *c = sb->s_fs_info;
 456
 457        /*
 458         * Zero @wait is just an advisory thing to help the file system shove
 459         * lots of data into the queues, and there will be the second
 460         * '->sync_fs()' call, with non-zero @wait.
 461         */
 462        if (!wait)
 463                return 0;
 464
 465        /*
 466         * Synchronize write buffers, because 'ubifs_run_commit()' does not
 467         * do this if it waits for an already running commit.
 468         */
 469        for (i = 0; i < c->jhead_cnt; i++) {
 470                err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
 471                if (err)
 472                        return err;
 473        }
 474
 475        /*
 476         * Strictly speaking, it is not necessary to commit the journal here,
 477         * synchronizing write-buffers would be enough. But committing makes
 478         * UBIFS free space predictions much more accurate, so we want to let
 479         * the user be able to get more accurate results of 'statfs()' after
 480         * they synchronize the file system.
 481         */
 482        err = ubifs_run_commit(c);
 483        if (err)
 484                return err;
 485
 486        return ubi_sync(c->vi.ubi_num);
 487}
 488
 489/**
 490 * init_constants_early - initialize UBIFS constants.
 491 * @c: UBIFS file-system description object
 492 *
 493 * This function initialize UBIFS constants which do not need the superblock to
 494 * be read. It also checks that the UBI volume satisfies basic UBIFS
 495 * requirements. Returns zero in case of success and a negative error code in
 496 * case of failure.
 497 */
 498static int init_constants_early(struct ubifs_info *c)
 499{
 500        if (c->vi.corrupted) {
 501                ubifs_warn(c, "UBI volume is corrupted - read-only mode");
 502                c->ro_media = 1;
 503        }
 504
 505        if (c->di.ro_mode) {
 506                ubifs_msg(c, "read-only UBI device");
 507                c->ro_media = 1;
 508        }
 509
 510        if (c->vi.vol_type == UBI_STATIC_VOLUME) {
 511                ubifs_msg(c, "static UBI volume - read-only mode");
 512                c->ro_media = 1;
 513        }
 514
 515        c->leb_cnt = c->vi.size;
 516        c->leb_size = c->vi.usable_leb_size;
 517        c->leb_start = c->di.leb_start;
 518        c->half_leb_size = c->leb_size / 2;
 519        c->min_io_size = c->di.min_io_size;
 520        c->min_io_shift = fls(c->min_io_size) - 1;
 521        c->max_write_size = c->di.max_write_size;
 522        c->max_write_shift = fls(c->max_write_size) - 1;
 523
 524        if (c->leb_size < UBIFS_MIN_LEB_SZ) {
 525                ubifs_errc(c, "too small LEBs (%d bytes), min. is %d bytes",
 526                           c->leb_size, UBIFS_MIN_LEB_SZ);
 527                return -EINVAL;
 528        }
 529
 530        if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
 531                ubifs_errc(c, "too few LEBs (%d), min. is %d",
 532                           c->leb_cnt, UBIFS_MIN_LEB_CNT);
 533                return -EINVAL;
 534        }
 535
 536        if (!is_power_of_2(c->min_io_size)) {
 537                ubifs_errc(c, "bad min. I/O size %d", c->min_io_size);
 538                return -EINVAL;
 539        }
 540
 541        /*
 542         * Maximum write size has to be greater or equivalent to min. I/O
 543         * size, and be multiple of min. I/O size.
 544         */
 545        if (c->max_write_size < c->min_io_size ||
 546            c->max_write_size % c->min_io_size ||
 547            !is_power_of_2(c->max_write_size)) {
 548                ubifs_errc(c, "bad write buffer size %d for %d min. I/O unit",
 549                           c->max_write_size, c->min_io_size);
 550                return -EINVAL;
 551        }
 552
 553        /*
 554         * UBIFS aligns all node to 8-byte boundary, so to make function in
 555         * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
 556         * less than 8.
 557         */
 558        if (c->min_io_size < 8) {
 559                c->min_io_size = 8;
 560                c->min_io_shift = 3;
 561                if (c->max_write_size < c->min_io_size) {
 562                        c->max_write_size = c->min_io_size;
 563                        c->max_write_shift = c->min_io_shift;
 564                }
 565        }
 566
 567        c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
 568        c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
 569
 570        /*
 571         * Initialize node length ranges which are mostly needed for node
 572         * length validation.
 573         */
 574        c->ranges[UBIFS_PAD_NODE].len  = UBIFS_PAD_NODE_SZ;
 575        c->ranges[UBIFS_SB_NODE].len   = UBIFS_SB_NODE_SZ;
 576        c->ranges[UBIFS_MST_NODE].len  = UBIFS_MST_NODE_SZ;
 577        c->ranges[UBIFS_REF_NODE].len  = UBIFS_REF_NODE_SZ;
 578        c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
 579        c->ranges[UBIFS_CS_NODE].len   = UBIFS_CS_NODE_SZ;
 580
 581        c->ranges[UBIFS_INO_NODE].min_len  = UBIFS_INO_NODE_SZ;
 582        c->ranges[UBIFS_INO_NODE].max_len  = UBIFS_MAX_INO_NODE_SZ;
 583        c->ranges[UBIFS_ORPH_NODE].min_len =
 584                                UBIFS_ORPH_NODE_SZ + sizeof(__le64);
 585        c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
 586        c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
 587        c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
 588        c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
 589        c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
 590        c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
 591        c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
 592        /*
 593         * Minimum indexing node size is amended later when superblock is
 594         * read and the key length is known.
 595         */
 596        c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
 597        /*
 598         * Maximum indexing node size is amended later when superblock is
 599         * read and the fanout is known.
 600         */
 601        c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
 602
 603        /*
 604         * Initialize dead and dark LEB space watermarks. See gc.c for comments
 605         * about these values.
 606         */
 607        c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
 608        c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
 609
 610        /*
 611         * Calculate how many bytes would be wasted at the end of LEB if it was
 612         * fully filled with data nodes of maximum size. This is used in
 613         * calculations when reporting free space.
 614         */
 615        c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
 616
 617        /* Buffer size for bulk-reads */
 618        c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
 619        if (c->max_bu_buf_len > c->leb_size)
 620                c->max_bu_buf_len = c->leb_size;
 621        return 0;
 622}
 623
 624/**
 625 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
 626 * @c: UBIFS file-system description object
 627 * @lnum: LEB the write-buffer was synchronized to
 628 * @free: how many free bytes left in this LEB
 629 * @pad: how many bytes were padded
 630 *
 631 * This is a callback function which is called by the I/O unit when the
 632 * write-buffer is synchronized. We need this to correctly maintain space
 633 * accounting in bud logical eraseblocks. This function returns zero in case of
 634 * success and a negative error code in case of failure.
 635 *
 636 * This function actually belongs to the journal, but we keep it here because
 637 * we want to keep it static.
 638 */
 639static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
 640{
 641        return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
 642}
 643
 644/*
 645 * init_constants_sb - initialize UBIFS constants.
 646 * @c: UBIFS file-system description object
 647 *
 648 * This is a helper function which initializes various UBIFS constants after
 649 * the superblock has been read. It also checks various UBIFS parameters and
 650 * makes sure they are all right. Returns zero in case of success and a
 651 * negative error code in case of failure.
 652 */
 653static int init_constants_sb(struct ubifs_info *c)
 654{
 655        int tmp, err;
 656        long long tmp64;
 657
 658        c->main_bytes = (long long)c->main_lebs * c->leb_size;
 659        c->max_znode_sz = sizeof(struct ubifs_znode) +
 660                                c->fanout * sizeof(struct ubifs_zbranch);
 661
 662        tmp = ubifs_idx_node_sz(c, 1);
 663        c->ranges[UBIFS_IDX_NODE].min_len = tmp;
 664        c->min_idx_node_sz = ALIGN(tmp, 8);
 665
 666        tmp = ubifs_idx_node_sz(c, c->fanout);
 667        c->ranges[UBIFS_IDX_NODE].max_len = tmp;
 668        c->max_idx_node_sz = ALIGN(tmp, 8);
 669
 670        /* Make sure LEB size is large enough to fit full commit */
 671        tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
 672        tmp = ALIGN(tmp, c->min_io_size);
 673        if (tmp > c->leb_size) {
 674                ubifs_err(c, "too small LEB size %d, at least %d needed",
 675                          c->leb_size, tmp);
 676                return -EINVAL;
 677        }
 678
 679        /*
 680         * Make sure that the log is large enough to fit reference nodes for
 681         * all buds plus one reserved LEB.
 682         */
 683        tmp64 = c->max_bud_bytes + c->leb_size - 1;
 684        c->max_bud_cnt = div_u64(tmp64, c->leb_size);
 685        tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
 686        tmp /= c->leb_size;
 687        tmp += 1;
 688        if (c->log_lebs < tmp) {
 689                ubifs_err(c, "too small log %d LEBs, required min. %d LEBs",
 690                          c->log_lebs, tmp);
 691                return -EINVAL;
 692        }
 693
 694        /*
 695         * When budgeting we assume worst-case scenarios when the pages are not
 696         * be compressed and direntries are of the maximum size.
 697         *
 698         * Note, data, which may be stored in inodes is budgeted separately, so
 699         * it is not included into 'c->bi.inode_budget'.
 700         */
 701        c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
 702        c->bi.inode_budget = UBIFS_INO_NODE_SZ;
 703        c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
 704
 705        /*
 706         * When the amount of flash space used by buds becomes
 707         * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
 708         * The writers are unblocked when the commit is finished. To avoid
 709         * writers to be blocked UBIFS initiates background commit in advance,
 710         * when number of bud bytes becomes above the limit defined below.
 711         */
 712        c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
 713
 714        /*
 715         * Ensure minimum journal size. All the bytes in the journal heads are
 716         * considered to be used, when calculating the current journal usage.
 717         * Consequently, if the journal is too small, UBIFS will treat it as
 718         * always full.
 719         */
 720        tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
 721        if (c->bg_bud_bytes < tmp64)
 722                c->bg_bud_bytes = tmp64;
 723        if (c->max_bud_bytes < tmp64 + c->leb_size)
 724                c->max_bud_bytes = tmp64 + c->leb_size;
 725
 726        err = ubifs_calc_lpt_geom(c);
 727        if (err)
 728                return err;
 729
 730        /* Initialize effective LEB size used in budgeting calculations */
 731        c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
 732        return 0;
 733}
 734
 735/*
 736 * init_constants_master - initialize UBIFS constants.
 737 * @c: UBIFS file-system description object
 738 *
 739 * This is a helper function which initializes various UBIFS constants after
 740 * the master node has been read. It also checks various UBIFS parameters and
 741 * makes sure they are all right.
 742 */
 743static void init_constants_master(struct ubifs_info *c)
 744{
 745        long long tmp64;
 746
 747        c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
 748        c->report_rp_size = ubifs_reported_space(c, c->rp_size);
 749
 750        /*
 751         * Calculate total amount of FS blocks. This number is not used
 752         * internally because it does not make much sense for UBIFS, but it is
 753         * necessary to report something for the 'statfs()' call.
 754         *
 755         * Subtract the LEB reserved for GC, the LEB which is reserved for
 756         * deletions, minimum LEBs for the index, and assume only one journal
 757         * head is available.
 758         */
 759        tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
 760        tmp64 *= (long long)c->leb_size - c->leb_overhead;
 761        tmp64 = ubifs_reported_space(c, tmp64);
 762        c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
 763}
 764
 765/**
 766 * take_gc_lnum - reserve GC LEB.
 767 * @c: UBIFS file-system description object
 768 *
 769 * This function ensures that the LEB reserved for garbage collection is marked
 770 * as "taken" in lprops. We also have to set free space to LEB size and dirty
 771 * space to zero, because lprops may contain out-of-date information if the
 772 * file-system was un-mounted before it has been committed. This function
 773 * returns zero in case of success and a negative error code in case of
 774 * failure.
 775 */
 776static int take_gc_lnum(struct ubifs_info *c)
 777{
 778        int err;
 779
 780        if (c->gc_lnum == -1) {
 781                ubifs_err(c, "no LEB for GC");
 782                return -EINVAL;
 783        }
 784
 785        /* And we have to tell lprops that this LEB is taken */
 786        err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
 787                                  LPROPS_TAKEN, 0, 0);
 788        return err;
 789}
 790
 791/**
 792 * alloc_wbufs - allocate write-buffers.
 793 * @c: UBIFS file-system description object
 794 *
 795 * This helper function allocates and initializes UBIFS write-buffers. Returns
 796 * zero in case of success and %-ENOMEM in case of failure.
 797 */
 798static int alloc_wbufs(struct ubifs_info *c)
 799{
 800        int i, err;
 801
 802        c->jheads = kcalloc(c->jhead_cnt, sizeof(struct ubifs_jhead),
 803                            GFP_KERNEL);
 804        if (!c->jheads)
 805                return -ENOMEM;
 806
 807        /* Initialize journal heads */
 808        for (i = 0; i < c->jhead_cnt; i++) {
 809                INIT_LIST_HEAD(&c->jheads[i].buds_list);
 810                err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
 811                if (err)
 812                        return err;
 813
 814                c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
 815                c->jheads[i].wbuf.jhead = i;
 816                c->jheads[i].grouped = 1;
 817        }
 818
 819        /*
 820         * Garbage Collector head does not need to be synchronized by timer.
 821         * Also GC head nodes are not grouped.
 822         */
 823        c->jheads[GCHD].wbuf.no_timer = 1;
 824        c->jheads[GCHD].grouped = 0;
 825
 826        return 0;
 827}
 828
 829/**
 830 * free_wbufs - free write-buffers.
 831 * @c: UBIFS file-system description object
 832 */
 833static void free_wbufs(struct ubifs_info *c)
 834{
 835        int i;
 836
 837        if (c->jheads) {
 838                for (i = 0; i < c->jhead_cnt; i++) {
 839                        kfree(c->jheads[i].wbuf.buf);
 840                        kfree(c->jheads[i].wbuf.inodes);
 841                }
 842                kfree(c->jheads);
 843                c->jheads = NULL;
 844        }
 845}
 846
 847/**
 848 * free_orphans - free orphans.
 849 * @c: UBIFS file-system description object
 850 */
 851static void free_orphans(struct ubifs_info *c)
 852{
 853        struct ubifs_orphan *orph;
 854
 855        while (c->orph_dnext) {
 856                orph = c->orph_dnext;
 857                c->orph_dnext = orph->dnext;
 858                list_del(&orph->list);
 859                kfree(orph);
 860        }
 861
 862        while (!list_empty(&c->orph_list)) {
 863                orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
 864                list_del(&orph->list);
 865                kfree(orph);
 866                ubifs_err(c, "orphan list not empty at unmount");
 867        }
 868
 869        vfree(c->orph_buf);
 870        c->orph_buf = NULL;
 871}
 872
 873/**
 874 * free_buds - free per-bud objects.
 875 * @c: UBIFS file-system description object
 876 */
 877static void free_buds(struct ubifs_info *c)
 878{
 879        struct ubifs_bud *bud, *n;
 880
 881        rbtree_postorder_for_each_entry_safe(bud, n, &c->buds, rb)
 882                kfree(bud);
 883}
 884
 885/**
 886 * check_volume_empty - check if the UBI volume is empty.
 887 * @c: UBIFS file-system description object
 888 *
 889 * This function checks if the UBIFS volume is empty by looking if its LEBs are
 890 * mapped or not. The result of checking is stored in the @c->empty variable.
 891 * Returns zero in case of success and a negative error code in case of
 892 * failure.
 893 */
 894static int check_volume_empty(struct ubifs_info *c)
 895{
 896        int lnum, err;
 897
 898        c->empty = 1;
 899        for (lnum = 0; lnum < c->leb_cnt; lnum++) {
 900                err = ubifs_is_mapped(c, lnum);
 901                if (unlikely(err < 0))
 902                        return err;
 903                if (err == 1) {
 904                        c->empty = 0;
 905                        break;
 906                }
 907
 908                cond_resched();
 909        }
 910
 911        return 0;
 912}
 913
 914/*
 915 * UBIFS mount options.
 916 *
 917 * Opt_fast_unmount: do not run a journal commit before un-mounting
 918 * Opt_norm_unmount: run a journal commit before un-mounting
 919 * Opt_bulk_read: enable bulk-reads
 920 * Opt_no_bulk_read: disable bulk-reads
 921 * Opt_chk_data_crc: check CRCs when reading data nodes
 922 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
 923 * Opt_override_compr: override default compressor
 924 * Opt_err: just end of array marker
 925 */
 926enum {
 927        Opt_fast_unmount,
 928        Opt_norm_unmount,
 929        Opt_bulk_read,
 930        Opt_no_bulk_read,
 931        Opt_chk_data_crc,
 932        Opt_no_chk_data_crc,
 933        Opt_override_compr,
 934        Opt_ignore,
 935        Opt_err,
 936};
 937
 938static const match_table_t tokens = {
 939        {Opt_fast_unmount, "fast_unmount"},
 940        {Opt_norm_unmount, "norm_unmount"},
 941        {Opt_bulk_read, "bulk_read"},
 942        {Opt_no_bulk_read, "no_bulk_read"},
 943        {Opt_chk_data_crc, "chk_data_crc"},
 944        {Opt_no_chk_data_crc, "no_chk_data_crc"},
 945        {Opt_override_compr, "compr=%s"},
 946        {Opt_ignore, "ubi=%s"},
 947        {Opt_ignore, "vol=%s"},
 948        {Opt_err, NULL},
 949};
 950
 951/**
 952 * parse_standard_option - parse a standard mount option.
 953 * @option: the option to parse
 954 *
 955 * Normally, standard mount options like "sync" are passed to file-systems as
 956 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
 957 * be present in the options string. This function tries to deal with this
 958 * situation and parse standard options. Returns 0 if the option was not
 959 * recognized, and the corresponding integer flag if it was.
 960 *
 961 * UBIFS is only interested in the "sync" option, so do not check for anything
 962 * else.
 963 */
 964static int parse_standard_option(const char *option)
 965{
 966
 967        pr_notice("UBIFS: parse %s\n", option);
 968        if (!strcmp(option, "sync"))
 969                return SB_SYNCHRONOUS;
 970        return 0;
 971}
 972
 973/**
 974 * ubifs_parse_options - parse mount parameters.
 975 * @c: UBIFS file-system description object
 976 * @options: parameters to parse
 977 * @is_remount: non-zero if this is FS re-mount
 978 *
 979 * This function parses UBIFS mount options and returns zero in case success
 980 * and a negative error code in case of failure.
 981 */
 982static int ubifs_parse_options(struct ubifs_info *c, char *options,
 983                               int is_remount)
 984{
 985        char *p;
 986        substring_t args[MAX_OPT_ARGS];
 987
 988        if (!options)
 989                return 0;
 990
 991        while ((p = strsep(&options, ","))) {
 992                int token;
 993
 994                if (!*p)
 995                        continue;
 996
 997                token = match_token(p, tokens, args);
 998                switch (token) {
 999                /*
1000                 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
1001                 * We accept them in order to be backward-compatible. But this
1002                 * should be removed at some point.
1003                 */
1004                case Opt_fast_unmount:
1005                        c->mount_opts.unmount_mode = 2;
1006                        break;
1007                case Opt_norm_unmount:
1008                        c->mount_opts.unmount_mode = 1;
1009                        break;
1010                case Opt_bulk_read:
1011                        c->mount_opts.bulk_read = 2;
1012                        c->bulk_read = 1;
1013                        break;
1014                case Opt_no_bulk_read:
1015                        c->mount_opts.bulk_read = 1;
1016                        c->bulk_read = 0;
1017                        break;
1018                case Opt_chk_data_crc:
1019                        c->mount_opts.chk_data_crc = 2;
1020                        c->no_chk_data_crc = 0;
1021                        break;
1022                case Opt_no_chk_data_crc:
1023                        c->mount_opts.chk_data_crc = 1;
1024                        c->no_chk_data_crc = 1;
1025                        break;
1026                case Opt_override_compr:
1027                {
1028                        char *name = match_strdup(&args[0]);
1029
1030                        if (!name)
1031                                return -ENOMEM;
1032                        if (!strcmp(name, "none"))
1033                                c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1034                        else if (!strcmp(name, "lzo"))
1035                                c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1036                        else if (!strcmp(name, "zlib"))
1037                                c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1038                        else {
1039                                ubifs_err(c, "unknown compressor \"%s\"", name); //FIXME: is c ready?
1040                                kfree(name);
1041                                return -EINVAL;
1042                        }
1043                        kfree(name);
1044                        c->mount_opts.override_compr = 1;
1045                        c->default_compr = c->mount_opts.compr_type;
1046                        break;
1047                }
1048                case Opt_ignore:
1049                        break;
1050                default:
1051                {
1052                        unsigned long flag;
1053                        struct super_block *sb = c->vfs_sb;
1054
1055                        flag = parse_standard_option(p);
1056                        if (!flag) {
1057                                ubifs_err(c, "unrecognized mount option \"%s\" or missing value",
1058                                          p);
1059                                return -EINVAL;
1060                        }
1061                        sb->s_flags |= flag;
1062                        break;
1063                }
1064                }
1065        }
1066
1067        return 0;
1068}
1069
1070/**
1071 * destroy_journal - destroy journal data structures.
1072 * @c: UBIFS file-system description object
1073 *
1074 * This function destroys journal data structures including those that may have
1075 * been created by recovery functions.
1076 */
1077static void destroy_journal(struct ubifs_info *c)
1078{
1079        while (!list_empty(&c->unclean_leb_list)) {
1080                struct ubifs_unclean_leb *ucleb;
1081
1082                ucleb = list_entry(c->unclean_leb_list.next,
1083                                   struct ubifs_unclean_leb, list);
1084                list_del(&ucleb->list);
1085                kfree(ucleb);
1086        }
1087        while (!list_empty(&c->old_buds)) {
1088                struct ubifs_bud *bud;
1089
1090                bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1091                list_del(&bud->list);
1092                kfree(bud);
1093        }
1094        ubifs_destroy_idx_gc(c);
1095        ubifs_destroy_size_tree(c);
1096        ubifs_tnc_close(c);
1097        free_buds(c);
1098}
1099
1100/**
1101 * bu_init - initialize bulk-read information.
1102 * @c: UBIFS file-system description object
1103 */
1104static void bu_init(struct ubifs_info *c)
1105{
1106        ubifs_assert(c->bulk_read == 1);
1107
1108        if (c->bu.buf)
1109                return; /* Already initialized */
1110
1111again:
1112        c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1113        if (!c->bu.buf) {
1114                if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1115                        c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1116                        goto again;
1117                }
1118
1119                /* Just disable bulk-read */
1120                ubifs_warn(c, "cannot allocate %d bytes of memory for bulk-read, disabling it",
1121                           c->max_bu_buf_len);
1122                c->mount_opts.bulk_read = 1;
1123                c->bulk_read = 0;
1124                return;
1125        }
1126}
1127
1128/**
1129 * check_free_space - check if there is enough free space to mount.
1130 * @c: UBIFS file-system description object
1131 *
1132 * This function makes sure UBIFS has enough free space to be mounted in
1133 * read/write mode. UBIFS must always have some free space to allow deletions.
1134 */
1135static int check_free_space(struct ubifs_info *c)
1136{
1137        ubifs_assert(c->dark_wm > 0);
1138        if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1139                ubifs_err(c, "insufficient free space to mount in R/W mode");
1140                ubifs_dump_budg(c, &c->bi);
1141                ubifs_dump_lprops(c);
1142                return -ENOSPC;
1143        }
1144        return 0;
1145}
1146
1147/**
1148 * mount_ubifs - mount UBIFS file-system.
1149 * @c: UBIFS file-system description object
1150 *
1151 * This function mounts UBIFS file system. Returns zero in case of success and
1152 * a negative error code in case of failure.
1153 */
1154static int mount_ubifs(struct ubifs_info *c)
1155{
1156        int err;
1157        long long x, y;
1158        size_t sz;
1159
1160        c->ro_mount = !!sb_rdonly(c->vfs_sb);
1161        /* Suppress error messages while probing if SB_SILENT is set */
1162        c->probing = !!(c->vfs_sb->s_flags & SB_SILENT);
1163
1164        err = init_constants_early(c);
1165        if (err)
1166                return err;
1167
1168        err = ubifs_debugging_init(c);
1169        if (err)
1170                return err;
1171
1172        err = check_volume_empty(c);
1173        if (err)
1174                goto out_free;
1175
1176        if (c->empty && (c->ro_mount || c->ro_media)) {
1177                /*
1178                 * This UBI volume is empty, and read-only, or the file system
1179                 * is mounted read-only - we cannot format it.
1180                 */
1181                ubifs_err(c, "can't format empty UBI volume: read-only %s",
1182                          c->ro_media ? "UBI volume" : "mount");
1183                err = -EROFS;
1184                goto out_free;
1185        }
1186
1187        if (c->ro_media && !c->ro_mount) {
1188                ubifs_err(c, "cannot mount read-write - read-only media");
1189                err = -EROFS;
1190                goto out_free;
1191        }
1192
1193        /*
1194         * The requirement for the buffer is that it should fit indexing B-tree
1195         * height amount of integers. We assume the height if the TNC tree will
1196         * never exceed 64.
1197         */
1198        err = -ENOMEM;
1199        c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1200        if (!c->bottom_up_buf)
1201                goto out_free;
1202
1203        c->sbuf = vmalloc(c->leb_size);
1204        if (!c->sbuf)
1205                goto out_free;
1206
1207        if (!c->ro_mount) {
1208                c->ileb_buf = vmalloc(c->leb_size);
1209                if (!c->ileb_buf)
1210                        goto out_free;
1211        }
1212
1213        if (c->bulk_read == 1)
1214                bu_init(c);
1215
1216        if (!c->ro_mount) {
1217                c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ + \
1218                                               UBIFS_CIPHER_BLOCK_SIZE,
1219                                               GFP_KERNEL);
1220                if (!c->write_reserve_buf)
1221                        goto out_free;
1222        }
1223
1224        c->mounting = 1;
1225
1226        err = ubifs_read_superblock(c);
1227        if (err)
1228                goto out_free;
1229
1230        c->probing = 0;
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(c, "'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(c, "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(c, "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 && !c->need_recovery) {
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(c, "recovery deferred");
1390                else {
1391                        c->need_recovery = 0;
1392                        ubifs_msg(c, "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(c, "UBIFS: 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(c, "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(c, "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(c, "reserved for root: %llu bytes (%llu KiB)",
1425                  c->report_rp_size, c->report_rp_size >> 10);
1426        ubifs_msg(c, "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(c, "the file-system is not R/W-compatible");
1560                ubifs_msg(c, "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                        goto out;
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(c, "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 + \
1632                                       UBIFS_CIPHER_BLOCK_SIZE, GFP_KERNEL);
1633        if (!c->write_reserve_buf) {
1634                err = -ENOMEM;
1635                goto out;
1636        }
1637
1638        err = ubifs_lpt_init(c, 0, 1);
1639        if (err)
1640                goto out;
1641
1642        /* Create background thread */
1643        c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1644        if (IS_ERR(c->bgt)) {
1645                err = PTR_ERR(c->bgt);
1646                c->bgt = NULL;
1647                ubifs_err(c, "cannot spawn \"%s\", error %d",
1648                          c->bgt_name, err);
1649                goto out;
1650        }
1651        wake_up_process(c->bgt);
1652
1653        c->orph_buf = vmalloc(c->leb_size);
1654        if (!c->orph_buf) {
1655                err = -ENOMEM;
1656                goto out;
1657        }
1658
1659        /* Check for enough log space */
1660        lnum = c->lhead_lnum + 1;
1661        if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1662                lnum = UBIFS_LOG_LNUM;
1663        if (lnum == c->ltail_lnum) {
1664                err = ubifs_consolidate_log(c);
1665                if (err)
1666                        goto out;
1667        }
1668
1669        if (c->need_recovery)
1670                err = ubifs_rcvry_gc_commit(c);
1671        else
1672                err = ubifs_leb_unmap(c, c->gc_lnum);
1673        if (err)
1674                goto out;
1675
1676        dbg_gen("re-mounted read-write");
1677        c->remounting_rw = 0;
1678
1679        if (c->need_recovery) {
1680                c->need_recovery = 0;
1681                ubifs_msg(c, "deferred recovery completed");
1682        } else {
1683                /*
1684                 * Do not run the debugging space check if the were doing
1685                 * recovery, because when we saved the information we had the
1686                 * file-system in a state where the TNC and lprops has been
1687                 * modified in memory, but all the I/O operations (including a
1688                 * commit) were deferred. So the file-system was in
1689                 * "non-committed" state. Now the file-system is in committed
1690                 * state, and of course the amount of free space will change
1691                 * because, for example, the old index size was imprecise.
1692                 */
1693                err = dbg_check_space_info(c);
1694        }
1695
1696        mutex_unlock(&c->umount_mutex);
1697        return err;
1698
1699out:
1700        c->ro_mount = 1;
1701        vfree(c->orph_buf);
1702        c->orph_buf = NULL;
1703        if (c->bgt) {
1704                kthread_stop(c->bgt);
1705                c->bgt = NULL;
1706        }
1707        free_wbufs(c);
1708        kfree(c->write_reserve_buf);
1709        c->write_reserve_buf = NULL;
1710        vfree(c->ileb_buf);
1711        c->ileb_buf = NULL;
1712        ubifs_lpt_free(c, 1);
1713        c->remounting_rw = 0;
1714        mutex_unlock(&c->umount_mutex);
1715        return err;
1716}
1717
1718/**
1719 * ubifs_remount_ro - re-mount in read-only mode.
1720 * @c: UBIFS file-system description object
1721 *
1722 * We assume VFS has stopped writing. Possibly the background thread could be
1723 * running a commit, however kthread_stop will wait in that case.
1724 */
1725static void ubifs_remount_ro(struct ubifs_info *c)
1726{
1727        int i, err;
1728
1729        ubifs_assert(!c->need_recovery);
1730        ubifs_assert(!c->ro_mount);
1731
1732        mutex_lock(&c->umount_mutex);
1733        if (c->bgt) {
1734                kthread_stop(c->bgt);
1735                c->bgt = NULL;
1736        }
1737
1738        dbg_save_space_info(c);
1739
1740        for (i = 0; i < c->jhead_cnt; i++) {
1741                err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
1742                if (err)
1743                        ubifs_ro_mode(c, err);
1744        }
1745
1746        c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1747        c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1748        c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1749        err = ubifs_write_master(c);
1750        if (err)
1751                ubifs_ro_mode(c, err);
1752
1753        vfree(c->orph_buf);
1754        c->orph_buf = NULL;
1755        kfree(c->write_reserve_buf);
1756        c->write_reserve_buf = NULL;
1757        vfree(c->ileb_buf);
1758        c->ileb_buf = NULL;
1759        ubifs_lpt_free(c, 1);
1760        c->ro_mount = 1;
1761        err = dbg_check_space_info(c);
1762        if (err)
1763                ubifs_ro_mode(c, err);
1764        mutex_unlock(&c->umount_mutex);
1765}
1766
1767static void ubifs_put_super(struct super_block *sb)
1768{
1769        int i;
1770        struct ubifs_info *c = sb->s_fs_info;
1771
1772        ubifs_msg(c, "un-mount UBI device %d", c->vi.ubi_num);
1773
1774        /*
1775         * The following asserts are only valid if there has not been a failure
1776         * of the media. For example, there will be dirty inodes if we failed
1777         * to write them back because of I/O errors.
1778         */
1779        if (!c->ro_error) {
1780                ubifs_assert(c->bi.idx_growth == 0);
1781                ubifs_assert(c->bi.dd_growth == 0);
1782                ubifs_assert(c->bi.data_growth == 0);
1783        }
1784
1785        /*
1786         * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1787         * and file system un-mount. Namely, it prevents the shrinker from
1788         * picking this superblock for shrinking - it will be just skipped if
1789         * the mutex is locked.
1790         */
1791        mutex_lock(&c->umount_mutex);
1792        if (!c->ro_mount) {
1793                /*
1794                 * First of all kill the background thread to make sure it does
1795                 * not interfere with un-mounting and freeing resources.
1796                 */
1797                if (c->bgt) {
1798                        kthread_stop(c->bgt);
1799                        c->bgt = NULL;
1800                }
1801
1802                /*
1803                 * On fatal errors c->ro_error is set to 1, in which case we do
1804                 * not write the master node.
1805                 */
1806                if (!c->ro_error) {
1807                        int err;
1808
1809                        /* Synchronize write-buffers */
1810                        for (i = 0; i < c->jhead_cnt; i++) {
1811                                err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
1812                                if (err)
1813                                        ubifs_ro_mode(c, err);
1814                        }
1815
1816                        /*
1817                         * We are being cleanly unmounted which means the
1818                         * orphans were killed - indicate this in the master
1819                         * node. Also save the reserved GC LEB number.
1820                         */
1821                        c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1822                        c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1823                        c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1824                        err = ubifs_write_master(c);
1825                        if (err)
1826                                /*
1827                                 * Recovery will attempt to fix the master area
1828                                 * next mount, so we just print a message and
1829                                 * continue to unmount normally.
1830                                 */
1831                                ubifs_err(c, "failed to write master node, error %d",
1832                                          err);
1833                } else {
1834                        for (i = 0; i < c->jhead_cnt; i++)
1835                                /* Make sure write-buffer timers are canceled */
1836                                hrtimer_cancel(&c->jheads[i].wbuf.timer);
1837                }
1838        }
1839
1840        ubifs_umount(c);
1841        ubi_close_volume(c->ubi);
1842        mutex_unlock(&c->umount_mutex);
1843}
1844
1845static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1846{
1847        int err;
1848        struct ubifs_info *c = sb->s_fs_info;
1849
1850        sync_filesystem(sb);
1851        dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1852
1853        err = ubifs_parse_options(c, data, 1);
1854        if (err) {
1855                ubifs_err(c, "invalid or unknown remount parameter");
1856                return err;
1857        }
1858
1859        if (c->ro_mount && !(*flags & SB_RDONLY)) {
1860                if (c->ro_error) {
1861                        ubifs_msg(c, "cannot re-mount R/W due to prior errors");
1862                        return -EROFS;
1863                }
1864                if (c->ro_media) {
1865                        ubifs_msg(c, "cannot re-mount R/W - UBI volume is R/O");
1866                        return -EROFS;
1867                }
1868                err = ubifs_remount_rw(c);
1869                if (err)
1870                        return err;
1871        } else if (!c->ro_mount && (*flags & SB_RDONLY)) {
1872                if (c->ro_error) {
1873                        ubifs_msg(c, "cannot re-mount R/O due to prior errors");
1874                        return -EROFS;
1875                }
1876                ubifs_remount_ro(c);
1877        }
1878
1879        if (c->bulk_read == 1)
1880                bu_init(c);
1881        else {
1882                dbg_gen("disable bulk-read");
1883                mutex_lock(&c->bu_mutex);
1884                kfree(c->bu.buf);
1885                c->bu.buf = NULL;
1886                mutex_unlock(&c->bu_mutex);
1887        }
1888
1889        ubifs_assert(c->lst.taken_empty_lebs > 0);
1890        return 0;
1891}
1892
1893const struct super_operations ubifs_super_operations = {
1894        .alloc_inode   = ubifs_alloc_inode,
1895        .destroy_inode = ubifs_destroy_inode,
1896        .put_super     = ubifs_put_super,
1897        .write_inode   = ubifs_write_inode,
1898        .evict_inode   = ubifs_evict_inode,
1899        .statfs        = ubifs_statfs,
1900        .dirty_inode   = ubifs_dirty_inode,
1901        .remount_fs    = ubifs_remount_fs,
1902        .show_options  = ubifs_show_options,
1903        .sync_fs       = ubifs_sync_fs,
1904};
1905
1906/**
1907 * open_ubi - parse UBI device name string and open the UBI device.
1908 * @name: UBI volume name
1909 * @mode: UBI volume open mode
1910 *
1911 * The primary method of mounting UBIFS is by specifying the UBI volume
1912 * character device node path. However, UBIFS may also be mounted withoug any
1913 * character device node using one of the following methods:
1914 *
1915 * o ubiX_Y    - mount UBI device number X, volume Y;
1916 * o ubiY      - mount UBI device number 0, volume Y;
1917 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1918 * o ubi:NAME  - mount UBI device 0, volume with name NAME.
1919 *
1920 * Alternative '!' separator may be used instead of ':' (because some shells
1921 * like busybox may interpret ':' as an NFS host name separator). This function
1922 * returns UBI volume description object in case of success and a negative
1923 * error code in case of failure.
1924 */
1925static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1926{
1927        struct ubi_volume_desc *ubi;
1928        int dev, vol;
1929        char *endptr;
1930
1931        /* First, try to open using the device node path method */
1932        ubi = ubi_open_volume_path(name, mode);
1933        if (!IS_ERR(ubi))
1934                return ubi;
1935
1936        /* Try the "nodev" method */
1937        if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1938                return ERR_PTR(-EINVAL);
1939
1940        /* ubi:NAME method */
1941        if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1942                return ubi_open_volume_nm(0, name + 4, mode);
1943
1944        if (!isdigit(name[3]))
1945                return ERR_PTR(-EINVAL);
1946
1947        dev = simple_strtoul(name + 3, &endptr, 0);
1948
1949        /* ubiY method */
1950        if (*endptr == '\0')
1951                return ubi_open_volume(0, dev, mode);
1952
1953        /* ubiX_Y method */
1954        if (*endptr == '_' && isdigit(endptr[1])) {
1955                vol = simple_strtoul(endptr + 1, &endptr, 0);
1956                if (*endptr != '\0')
1957                        return ERR_PTR(-EINVAL);
1958                return ubi_open_volume(dev, vol, mode);
1959        }
1960
1961        /* ubiX:NAME method */
1962        if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1963                return ubi_open_volume_nm(dev, ++endptr, mode);
1964
1965        return ERR_PTR(-EINVAL);
1966}
1967
1968static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1969{
1970        struct ubifs_info *c;
1971
1972        c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1973        if (c) {
1974                spin_lock_init(&c->cnt_lock);
1975                spin_lock_init(&c->cs_lock);
1976                spin_lock_init(&c->buds_lock);
1977                spin_lock_init(&c->space_lock);
1978                spin_lock_init(&c->orphan_lock);
1979                init_rwsem(&c->commit_sem);
1980                mutex_init(&c->lp_mutex);
1981                mutex_init(&c->tnc_mutex);
1982                mutex_init(&c->log_mutex);
1983                mutex_init(&c->umount_mutex);
1984                mutex_init(&c->bu_mutex);
1985                mutex_init(&c->write_reserve_mutex);
1986                init_waitqueue_head(&c->cmt_wq);
1987                c->buds = RB_ROOT;
1988                c->old_idx = RB_ROOT;
1989                c->size_tree = RB_ROOT;
1990                c->orph_tree = RB_ROOT;
1991                INIT_LIST_HEAD(&c->infos_list);
1992                INIT_LIST_HEAD(&c->idx_gc);
1993                INIT_LIST_HEAD(&c->replay_list);
1994                INIT_LIST_HEAD(&c->replay_buds);
1995                INIT_LIST_HEAD(&c->uncat_list);
1996                INIT_LIST_HEAD(&c->empty_list);
1997                INIT_LIST_HEAD(&c->freeable_list);
1998                INIT_LIST_HEAD(&c->frdi_idx_list);
1999                INIT_LIST_HEAD(&c->unclean_leb_list);
2000                INIT_LIST_HEAD(&c->old_buds);
2001                INIT_LIST_HEAD(&c->orph_list);
2002                INIT_LIST_HEAD(&c->orph_new);
2003                c->no_chk_data_crc = 1;
2004
2005                c->highest_inum = UBIFS_FIRST_INO;
2006                c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
2007
2008                ubi_get_volume_info(ubi, &c->vi);
2009                ubi_get_device_info(c->vi.ubi_num, &c->di);
2010        }
2011        return c;
2012}
2013
2014static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
2015{
2016        struct ubifs_info *c = sb->s_fs_info;
2017        struct inode *root;
2018        int err;
2019
2020        c->vfs_sb = sb;
2021        /* Re-open the UBI device in read-write mode */
2022        c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2023        if (IS_ERR(c->ubi)) {
2024                err = PTR_ERR(c->ubi);
2025                goto out;
2026        }
2027
2028        err = ubifs_parse_options(c, data, 0);
2029        if (err)
2030                goto out_close;
2031
2032        /*
2033         * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2034         * UBIFS, I/O is not deferred, it is done immediately in readpage,
2035         * which means the user would have to wait not just for their own I/O
2036         * but the read-ahead I/O as well i.e. completely pointless.
2037         *
2038         * Read-ahead will be disabled because @sb->s_bdi->ra_pages is 0. Also
2039         * @sb->s_bdi->capabilities are initialized to 0 so there won't be any
2040         * writeback happening.
2041         */
2042        err = super_setup_bdi_name(sb, "ubifs_%d_%d", c->vi.ubi_num,
2043                                   c->vi.vol_id);
2044        if (err)
2045                goto out_close;
2046
2047        sb->s_fs_info = c;
2048        sb->s_magic = UBIFS_SUPER_MAGIC;
2049        sb->s_blocksize = UBIFS_BLOCK_SIZE;
2050        sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2051        sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2052        if (c->max_inode_sz > MAX_LFS_FILESIZE)
2053                sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2054        sb->s_op = &ubifs_super_operations;
2055        sb->s_xattr = ubifs_xattr_handlers;
2056#ifdef CONFIG_UBIFS_FS_ENCRYPTION
2057        sb->s_cop = &ubifs_crypt_operations;
2058#endif
2059
2060        mutex_lock(&c->umount_mutex);
2061        err = mount_ubifs(c);
2062        if (err) {
2063                ubifs_assert(err < 0);
2064                goto out_unlock;
2065        }
2066
2067        /* Read the root inode */
2068        root = ubifs_iget(sb, UBIFS_ROOT_INO);
2069        if (IS_ERR(root)) {
2070                err = PTR_ERR(root);
2071                goto out_umount;
2072        }
2073
2074        sb->s_root = d_make_root(root);
2075        if (!sb->s_root) {
2076                err = -ENOMEM;
2077                goto out_umount;
2078        }
2079
2080        mutex_unlock(&c->umount_mutex);
2081        return 0;
2082
2083out_umount:
2084        ubifs_umount(c);
2085out_unlock:
2086        mutex_unlock(&c->umount_mutex);
2087out_close:
2088        ubi_close_volume(c->ubi);
2089out:
2090        return err;
2091}
2092
2093static int sb_test(struct super_block *sb, void *data)
2094{
2095        struct ubifs_info *c1 = data;
2096        struct ubifs_info *c = sb->s_fs_info;
2097
2098        return c->vi.cdev == c1->vi.cdev;
2099}
2100
2101static int sb_set(struct super_block *sb, void *data)
2102{
2103        sb->s_fs_info = data;
2104        return set_anon_super(sb, NULL);
2105}
2106
2107static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2108                        const char *name, void *data)
2109{
2110        struct ubi_volume_desc *ubi;
2111        struct ubifs_info *c;
2112        struct super_block *sb;
2113        int err;
2114
2115        dbg_gen("name %s, flags %#x", name, flags);
2116
2117        /*
2118         * Get UBI device number and volume ID. Mount it read-only so far
2119         * because this might be a new mount point, and UBI allows only one
2120         * read-write user at a time.
2121         */
2122        ubi = open_ubi(name, UBI_READONLY);
2123        if (IS_ERR(ubi)) {
2124                if (!(flags & SB_SILENT))
2125                        pr_err("UBIFS error (pid: %d): cannot open \"%s\", error %d",
2126                               current->pid, name, (int)PTR_ERR(ubi));
2127                return ERR_CAST(ubi);
2128        }
2129
2130        c = alloc_ubifs_info(ubi);
2131        if (!c) {
2132                err = -ENOMEM;
2133                goto out_close;
2134        }
2135
2136        dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2137
2138        sb = sget(fs_type, sb_test, sb_set, flags, c);
2139        if (IS_ERR(sb)) {
2140                err = PTR_ERR(sb);
2141                kfree(c);
2142                goto out_close;
2143        }
2144
2145        if (sb->s_root) {
2146                struct ubifs_info *c1 = sb->s_fs_info;
2147                kfree(c);
2148                /* A new mount point for already mounted UBIFS */
2149                dbg_gen("this ubi volume is already mounted");
2150                if (!!(flags & SB_RDONLY) != c1->ro_mount) {
2151                        err = -EBUSY;
2152                        goto out_deact;
2153                }
2154        } else {
2155                err = ubifs_fill_super(sb, data, flags & SB_SILENT ? 1 : 0);
2156                if (err)
2157                        goto out_deact;
2158                /* We do not support atime */
2159                sb->s_flags |= SB_ACTIVE;
2160#ifndef CONFIG_UBIFS_ATIME_SUPPORT
2161                sb->s_flags |= SB_NOATIME;
2162#else
2163                ubifs_msg(c, "full atime support is enabled.");
2164#endif
2165        }
2166
2167        /* 'fill_super()' opens ubi again so we must close it here */
2168        ubi_close_volume(ubi);
2169
2170        return dget(sb->s_root);
2171
2172out_deact:
2173        deactivate_locked_super(sb);
2174out_close:
2175        ubi_close_volume(ubi);
2176        return ERR_PTR(err);
2177}
2178
2179static void kill_ubifs_super(struct super_block *s)
2180{
2181        struct ubifs_info *c = s->s_fs_info;
2182        kill_anon_super(s);
2183        kfree(c);
2184}
2185
2186static struct file_system_type ubifs_fs_type = {
2187        .name    = "ubifs",
2188        .owner   = THIS_MODULE,
2189        .mount   = ubifs_mount,
2190        .kill_sb = kill_ubifs_super,
2191};
2192MODULE_ALIAS_FS("ubifs");
2193
2194/*
2195 * Inode slab cache constructor.
2196 */
2197static void inode_slab_ctor(void *obj)
2198{
2199        struct ubifs_inode *ui = obj;
2200        inode_init_once(&ui->vfs_inode);
2201}
2202
2203static int __init ubifs_init(void)
2204{
2205        int err;
2206
2207        BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2208
2209        /* Make sure node sizes are 8-byte aligned */
2210        BUILD_BUG_ON(UBIFS_CH_SZ        & 7);
2211        BUILD_BUG_ON(UBIFS_INO_NODE_SZ  & 7);
2212        BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2213        BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2214        BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2215        BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2216        BUILD_BUG_ON(UBIFS_SB_NODE_SZ   & 7);
2217        BUILD_BUG_ON(UBIFS_MST_NODE_SZ  & 7);
2218        BUILD_BUG_ON(UBIFS_REF_NODE_SZ  & 7);
2219        BUILD_BUG_ON(UBIFS_CS_NODE_SZ   & 7);
2220        BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2221
2222        BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2223        BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2224        BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2225        BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  & 7);
2226        BUILD_BUG_ON(UBIFS_MAX_NODE_SZ      & 7);
2227        BUILD_BUG_ON(MIN_WRITE_SZ           & 7);
2228
2229        /* Check min. node size */
2230        BUILD_BUG_ON(UBIFS_INO_NODE_SZ  < MIN_WRITE_SZ);
2231        BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2232        BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2233        BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2234
2235        BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2236        BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2237        BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2238        BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  > UBIFS_MAX_NODE_SZ);
2239
2240        /* Defined node sizes */
2241        BUILD_BUG_ON(UBIFS_SB_NODE_SZ  != 4096);
2242        BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2243        BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2244        BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2245
2246        /*
2247         * We use 2 bit wide bit-fields to store compression type, which should
2248         * be amended if more compressors are added. The bit-fields are:
2249         * @compr_type in 'struct ubifs_inode', @default_compr in
2250         * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2251         */
2252        BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2253
2254        /*
2255         * We require that PAGE_SIZE is greater-than-or-equal-to
2256         * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2257         */
2258        if (PAGE_SIZE < UBIFS_BLOCK_SIZE) {
2259                pr_err("UBIFS error (pid %d): VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
2260                       current->pid, (unsigned int)PAGE_SIZE);
2261                return -EINVAL;
2262        }
2263
2264        ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2265                                sizeof(struct ubifs_inode), 0,
2266                                SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT |
2267                                SLAB_ACCOUNT, &inode_slab_ctor);
2268        if (!ubifs_inode_slab)
2269                return -ENOMEM;
2270
2271        err = register_shrinker(&ubifs_shrinker_info);
2272        if (err)
2273                goto out_slab;
2274
2275        err = ubifs_compressors_init();
2276        if (err)
2277                goto out_shrinker;
2278
2279        err = dbg_debugfs_init();
2280        if (err)
2281                goto out_compr;
2282
2283        err = register_filesystem(&ubifs_fs_type);
2284        if (err) {
2285                pr_err("UBIFS error (pid %d): cannot register file system, error %d",
2286                       current->pid, err);
2287                goto out_dbg;
2288        }
2289        return 0;
2290
2291out_dbg:
2292        dbg_debugfs_exit();
2293out_compr:
2294        ubifs_compressors_exit();
2295out_shrinker:
2296        unregister_shrinker(&ubifs_shrinker_info);
2297out_slab:
2298        kmem_cache_destroy(ubifs_inode_slab);
2299        return err;
2300}
2301/* late_initcall to let compressors initialize first */
2302late_initcall(ubifs_init);
2303
2304static void __exit ubifs_exit(void)
2305{
2306        ubifs_assert(list_empty(&ubifs_infos));
2307        ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2308
2309        dbg_debugfs_exit();
2310        ubifs_compressors_exit();
2311        unregister_shrinker(&ubifs_shrinker_info);
2312
2313        /*
2314         * Make sure all delayed rcu free inodes are flushed before we
2315         * destroy cache.
2316         */
2317        rcu_barrier();
2318        kmem_cache_destroy(ubifs_inode_slab);
2319        unregister_filesystem(&ubifs_fs_type);
2320}
2321module_exit(ubifs_exit);
2322
2323MODULE_LICENSE("GPL");
2324MODULE_VERSION(__stringify(UBIFS_VERSION));
2325MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2326MODULE_DESCRIPTION("UBIFS - UBI File System");
2327