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