1/*
   2 * This file is part of UBIFS.
   3 *
   4 * Copyright (C) 2006-2008 Nokia Corporation.
   5 *
   6 * This program is free software; you can redistribute it and/or modify it
   7 * under the terms of the GNU General Public License version 2 as published by
   8 * the Free Software Foundation.
   9 *
  10 * This program is distributed in the hope that it will be useful, but WITHOUT
  11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  13 * more details.
  14 *
  15 * You should have received a copy of the GNU General Public License along with
  16 * this program; if not, write to the Free Software Foundation, Inc., 51
  17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  18 *
  19 * Authors: Artem Bityutskiy (Битюцкий Артём)
  20 *          Adrian Hunter
  21 */
  22
  23/*
  24 * This file implements UBIFS initialization and VFS superblock operations. Some
  25 * initialization stuff which is rather large and complex is placed at
  26 * corresponding subsystems, but most of it is here.
  27 */
  28
  29#include <linux/init.h>
  30#include <linux/slab.h>
  31#include <linux/module.h>
  32#include <linux/ctype.h>
  33#include <linux/kthread.h>
  34#include <linux/parser.h>
  35#include <linux/seq_file.h>
  36#include <linux/mount.h>
  37#include <linux/math64.h>
  38#include <linux/writeback.h>
  39#include "ubifs.h"
  40
  41/*
  42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
  43 * allocating too much.
  44 */
  45#define UBIFS_KMALLOC_OK (128*1024)
  46
  47/* Slab cache for UBIFS inodes */
  48struct kmem_cache *ubifs_inode_slab;
  49
  50/* UBIFS TNC shrinker description */
  51static struct shrinker ubifs_shrinker_info = {
  52	.shrink = ubifs_shrinker,
  53	.seeks = DEFAULT_SEEKS,
  54};
  55
  56/**
  57 * validate_inode - validate inode.
  58 * @c: UBIFS file-system description object
  59 * @inode: the inode to validate
  60 *
  61 * This is a helper function for 'ubifs_iget()' which validates various fields
  62 * of a newly built inode to make sure they contain sane values and prevent
  63 * possible vulnerabilities. Returns zero if the inode is all right and
  64 * a non-zero error code if not.
  65 */
  66static int validate_inode(struct ubifs_info *c, const struct inode *inode)
  67{
  68	int err;
  69	const struct ubifs_inode *ui = ubifs_inode(inode);
  70
  71	if (inode->i_size > c->max_inode_sz) {
  72		ubifs_err("inode is too large (%lld)",
  73			  (long long)inode->i_size);
  74		return 1;
  75	}
  76
  77	if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
  78		ubifs_err("unknown compression type %d", ui->compr_type);
  79		return 2;
  80	}
  81
  82	if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
  83		return 3;
  84
  85	if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
  86		return 4;
  87
  88	if (ui->xattr && !S_ISREG(inode->i_mode))
  89		return 5;
  90
  91	if (!ubifs_compr_present(ui->compr_type)) {
  92		ubifs_warn("inode %lu uses '%s' compression, but it was not "
  93			   "compiled in", inode->i_ino,
  94			   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	inode->i_nlink = le32_to_cpu(ino->nlink);
 133	inode->i_uid   = le32_to_cpu(ino->uid);
 134	inode->i_gid   = 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	dbg_dump_node(c, ino);
 250	dbg_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	INIT_LIST_HEAD(&inode->i_dentry);
 280	kmem_cache_free(ubifs_inode_slab, ui);
 281}
 282
 283static void ubifs_destroy_inode(struct inode *inode)
 284{
 285	struct ubifs_inode *ui = ubifs_inode(inode);
 286
 287	kfree(ui->data);
 288	call_rcu(&inode->i_rcu, ubifs_i_callback);
 289}
 290
 291/*
 292 * Note, Linux write-back code calls this without 'i_mutex'.
 293 */
 294static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
 295{
 296	int err = 0;
 297	struct ubifs_info *c = inode->i_sb->s_fs_info;
 298	struct ubifs_inode *ui = ubifs_inode(inode);
 299
 300	ubifs_assert(!ui->xattr);
 301	if (is_bad_inode(inode))
 302		return 0;
 303
 304	mutex_lock(&ui->ui_mutex);
 305	/*
 306	 * Due to races between write-back forced by budgeting
 307	 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
 308	 * have already been synchronized, do not do this again. This might
 309	 * also happen if it was synchronized in an VFS operation, e.g.
 310	 * 'ubifs_link()'.
 311	 */
 312	if (!ui->dirty) {
 313		mutex_unlock(&ui->ui_mutex);
 314		return 0;
 315	}
 316
 317	/*
 318	 * As an optimization, do not write orphan inodes to the media just
 319	 * because this is not needed.
 320	 */
 321	dbg_gen("inode %lu, mode %#x, nlink %u",
 322		inode->i_ino, (int)inode->i_mode, inode->i_nlink);
 323	if (inode->i_nlink) {
 324		err = ubifs_jnl_write_inode(c, inode);
 325		if (err)
 326			ubifs_err("can't write inode %lu, error %d",
 327				  inode->i_ino, err);
 328		else
 329			err = dbg_check_inode_size(c, inode, ui->ui_size);
 330	}
 331
 332	ui->dirty = 0;
 333	mutex_unlock(&ui->ui_mutex);
 334	ubifs_release_dirty_inode_budget(c, ui);
 335	return err;
 336}
 337
 338static void ubifs_evict_inode(struct inode *inode)
 339{
 340	int err;
 341	struct ubifs_info *c = inode->i_sb->s_fs_info;
 342	struct ubifs_inode *ui = ubifs_inode(inode);
 343
 344	if (ui->xattr)
 345		/*
 346		 * Extended attribute inode deletions are fully handled in
 347		 * 'ubifs_removexattr()'. These inodes are special and have
 348		 * limited usage, so there is nothing to do here.
 349		 */
 350		goto out;
 351
 352	dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
 353	ubifs_assert(!atomic_read(&inode->i_count));
 354
 355	truncate_inode_pages(&inode->i_data, 0);
 356
 357	if (inode->i_nlink)
 358		goto done;
 359
 360	if (is_bad_inode(inode))
 361		goto out;
 362
 363	ui->ui_size = inode->i_size = 0;
 364	err = ubifs_jnl_delete_inode(c, inode);
 365	if (err)
 366		/*
 367		 * Worst case we have a lost orphan inode wasting space, so a
 368		 * simple error message is OK here.
 369		 */
 370		ubifs_err("can't delete inode %lu, error %d",
 371			  inode->i_ino, err);
 372
 373out:
 374	if (ui->dirty)
 375		ubifs_release_dirty_inode_budget(c, ui);
 376	else {
 377		/* We've deleted something - clean the "no space" flags */
 378		c->bi.nospace = c->bi.nospace_rp = 0;
 379		smp_wmb();
 380	}
 381done:
 382	end_writeback(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 vfsmount *mnt)
 424{
 425	struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
 426
 427	if (c->mount_opts.unmount_mode == 2)
 428		seq_printf(s, ",fast_unmount");
 429	else if (c->mount_opts.unmount_mode == 1)
 430		seq_printf(s, ",norm_unmount");
 431
 432	if (c->mount_opts.bulk_read == 2)
 433		seq_printf(s, ",bulk_read");
 434	else if (c->mount_opts.bulk_read == 1)
 435		seq_printf(s, ",no_bulk_read");
 436
 437	if (c->mount_opts.chk_data_crc == 2)
 438		seq_printf(s, ",chk_data_crc");
 439	else if (c->mount_opts.chk_data_crc == 1)
 440		seq_printf(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	return 0;
 448}
 449
 450static int ubifs_sync_fs(struct super_block *sb, int wait)
 451{
 452	int i, err;
 453	struct ubifs_info *c = sb->s_fs_info;
 454
 455	/*
 456	 * Zero @wait is just an advisory thing to help the file system shove
 457	 * lots of data into the queues, and there will be the second
 458	 * '->sync_fs()' call, with non-zero @wait.
 459	 */
 460	if (!wait)
 461		return 0;
 462
 463	/*
 464	 * Synchronize write buffers, because 'ubifs_run_commit()' does not
 465	 * do this if it waits for an already running commit.
 466	 */
 467	for (i = 0; i < c->jhead_cnt; i++) {
 468		err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
 469		if (err)
 470			return err;
 471	}
 472
 473	/*
 474	 * Strictly speaking, it is not necessary to commit the journal here,
 475	 * synchronizing write-buffers would be enough. But committing makes
 476	 * UBIFS free space predictions much more accurate, so we want to let
 477	 * the user be able to get more accurate results of 'statfs()' after
 478	 * they synchronize the file system.
 479	 */
 480	err = ubifs_run_commit(c);
 481	if (err)
 482		return err;
 483
 484	return ubi_sync(c->vi.ubi_num);
 485}
 486
 487/**
 488 * init_constants_early - initialize UBIFS constants.
 489 * @c: UBIFS file-system description object
 490 *
 491 * This function initialize UBIFS constants which do not need the superblock to
 492 * be read. It also checks that the UBI volume satisfies basic UBIFS
 493 * requirements. Returns zero in case of success and a negative error code in
 494 * case of failure.
 495 */
 496static int init_constants_early(struct ubifs_info *c)
 497{
 498	if (c->vi.corrupted) {
 499		ubifs_warn("UBI volume is corrupted - read-only mode");
 500		c->ro_media = 1;
 501	}
 502
 503	if (c->di.ro_mode) {
 504		ubifs_msg("read-only UBI device");
 505		c->ro_media = 1;
 506	}
 507
 508	if (c->vi.vol_type == UBI_STATIC_VOLUME) {
 509		ubifs_msg("static UBI volume - read-only mode");
 510		c->ro_media = 1;
 511	}
 512
 513	c->leb_cnt = c->vi.size;
 514	c->leb_size = c->vi.usable_leb_size;
 515	c->leb_start = c->di.leb_start;
 516	c->half_leb_size = c->leb_size / 2;
 517	c->min_io_size = c->di.min_io_size;
 518	c->min_io_shift = fls(c->min_io_size) - 1;
 519	c->max_write_size = c->di.max_write_size;
 520	c->max_write_shift = fls(c->max_write_size) - 1;
 521
 522	if (c->leb_size < UBIFS_MIN_LEB_SZ) {
 523		ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
 524			  c->leb_size, UBIFS_MIN_LEB_SZ);
 525		return -EINVAL;
 526	}
 527
 528	if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
 529		ubifs_err("too few LEBs (%d), min. is %d",
 530			  c->leb_cnt, UBIFS_MIN_LEB_CNT);
 531		return -EINVAL;
 532	}
 533
 534	if (!is_power_of_2(c->min_io_size)) {
 535		ubifs_err("bad min. I/O size %d", c->min_io_size);
 536		return -EINVAL;
 537	}
 538
 539	/*
 540	 * Maximum write size has to be greater or equivalent to min. I/O
 541	 * size, and be multiple of min. I/O size.
 542	 */
 543	if (c->max_write_size < c->min_io_size ||
 544	    c->max_write_size % c->min_io_size ||
 545	    !is_power_of_2(c->max_write_size)) {
 546		ubifs_err("bad write buffer size %d for %d min. I/O unit",
 547			  c->max_write_size, c->min_io_size);
 548		return -EINVAL;
 549	}
 550
 551	/*
 552	 * UBIFS aligns all node to 8-byte boundary, so to make function in
 553	 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
 554	 * less than 8.
 555	 */
 556	if (c->min_io_size < 8) {
 557		c->min_io_size = 8;
 558		c->min_io_shift = 3;
 559		if (c->max_write_size < c->min_io_size) {
 560			c->max_write_size = c->min_io_size;
 561			c->max_write_shift = c->min_io_shift;
 562		}
 563	}
 564
 565	c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
 566	c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
 567
 568	/*
 569	 * Initialize node length ranges which are mostly needed for node
 570	 * length validation.
 571	 */
 572	c->ranges[UBIFS_PAD_NODE].len  = UBIFS_PAD_NODE_SZ;
 573	c->ranges[UBIFS_SB_NODE].len   = UBIFS_SB_NODE_SZ;
 574	c->ranges[UBIFS_MST_NODE].len  = UBIFS_MST_NODE_SZ;
 575	c->ranges[UBIFS_REF_NODE].len  = UBIFS_REF_NODE_SZ;
 576	c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
 577	c->ranges[UBIFS_CS_NODE].len   = UBIFS_CS_NODE_SZ;
 578
 579	c->ranges[UBIFS_INO_NODE].min_len  = UBIFS_INO_NODE_SZ;
 580	c->ranges[UBIFS_INO_NODE].max_len  = UBIFS_MAX_INO_NODE_SZ;
 581	c->ranges[UBIFS_ORPH_NODE].min_len =
 582				UBIFS_ORPH_NODE_SZ + sizeof(__le64);
 583	c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
 584	c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
 585	c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
 586	c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
 587	c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
 588	c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
 589	c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
 590	/*
 591	 * Minimum indexing node size is amended later when superblock is
 592	 * read and the key length is known.
 593	 */
 594	c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
 595	/*
 596	 * Maximum indexing node size is amended later when superblock is
 597	 * read and the fanout is known.
 598	 */
 599	c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
 600
 601	/*
 602	 * Initialize dead and dark LEB space watermarks. See gc.c for comments
 603	 * about these values.
 604	 */
 605	c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
 606	c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
 607
 608	/*
 609	 * Calculate how many bytes would be wasted at the end of LEB if it was
 610	 * fully filled with data nodes of maximum size. This is used in
 611	 * calculations when reporting free space.
 612	 */
 613	c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
 614
 615	/* Buffer size for bulk-reads */
 616	c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
 617	if (c->max_bu_buf_len > c->leb_size)
 618		c->max_bu_buf_len = c->leb_size;
 619	return 0;
 620}
 621
 622/**
 623 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
 624 * @c: UBIFS file-system description object
 625 * @lnum: LEB the write-buffer was synchronized to
 626 * @free: how many free bytes left in this LEB
 627 * @pad: how many bytes were padded
 628 *
 629 * This is a callback function which is called by the I/O unit when the
 630 * write-buffer is synchronized. We need this to correctly maintain space
 631 * accounting in bud logical eraseblocks. This function returns zero in case of
 632 * success and a negative error code in case of failure.
 633 *
 634 * This function actually belongs to the journal, but we keep it here because
 635 * we want to keep it static.
 636 */
 637static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
 638{
 639	return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
 640}
 641
 642/*
 643 * init_constants_sb - initialize UBIFS constants.
 644 * @c: UBIFS file-system description object
 645 *
 646 * This is a helper function which initializes various UBIFS constants after
 647 * the superblock has been read. It also checks various UBIFS parameters and
 648 * makes sure they are all right. Returns zero in case of success and a
 649 * negative error code in case of failure.
 650 */
 651static int init_constants_sb(struct ubifs_info *c)
 652{
 653	int tmp, err;
 654	long long tmp64;
 655
 656	c->main_bytes = (long long)c->main_lebs * c->leb_size;
 657	c->max_znode_sz = sizeof(struct ubifs_znode) +
 658				c->fanout * sizeof(struct ubifs_zbranch);
 659
 660	tmp = ubifs_idx_node_sz(c, 1);
 661	c->ranges[UBIFS_IDX_NODE].min_len = tmp;
 662	c->min_idx_node_sz = ALIGN(tmp, 8);
 663
 664	tmp = ubifs_idx_node_sz(c, c->fanout);
 665	c->ranges[UBIFS_IDX_NODE].max_len = tmp;
 666	c->max_idx_node_sz = ALIGN(tmp, 8);
 667
 668	/* Make sure LEB size is large enough to fit full commit */
 669	tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
 670	tmp = ALIGN(tmp, c->min_io_size);
 671	if (tmp > c->leb_size) {
 672		dbg_err("too small LEB size %d, at least %d needed",
 673			c->leb_size, tmp);
 674		return -EINVAL;
 675	}
 676
 677	/*
 678	 * Make sure that the log is large enough to fit reference nodes for
 679	 * all buds plus one reserved LEB.
 680	 */
 681	tmp64 = c->max_bud_bytes + c->leb_size - 1;
 682	c->max_bud_cnt = div_u64(tmp64, c->leb_size);
 683	tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
 684	tmp /= c->leb_size;
 685	tmp += 1;
 686	if (c->log_lebs < tmp) {
 687		dbg_err("too small log %d LEBs, required min. %d LEBs",
 688			c->log_lebs, tmp);
 689		return -EINVAL;
 690	}
 691
 692	/*
 693	 * When budgeting we assume worst-case scenarios when the pages are not
 694	 * be compressed and direntries are of the maximum size.
 695	 *
 696	 * Note, data, which may be stored in inodes is budgeted separately, so
 697	 * it is not included into 'c->bi.inode_budget'.
 698	 */
 699	c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
 700	c->bi.inode_budget = UBIFS_INO_NODE_SZ;
 701	c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
 702
 703	/*
 704	 * When the amount of flash space used by buds becomes
 705	 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
 706	 * The writers are unblocked when the commit is finished. To avoid
 707	 * writers to be blocked UBIFS initiates background commit in advance,
 708	 * when number of bud bytes becomes above the limit defined below.
 709	 */
 710	c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
 711
 712	/*
 713	 * Ensure minimum journal size. All the bytes in the journal heads are
 714	 * considered to be used, when calculating the current journal usage.
 715	 * Consequently, if the journal is too small, UBIFS will treat it as
 716	 * always full.
 717	 */
 718	tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
 719	if (c->bg_bud_bytes < tmp64)
 720		c->bg_bud_bytes = tmp64;
 721	if (c->max_bud_bytes < tmp64 + c->leb_size)
 722		c->max_bud_bytes = tmp64 + c->leb_size;
 723
 724	err = ubifs_calc_lpt_geom(c);
 725	if (err)
 726		return err;
 727
 728	/* Initialize effective LEB size used in budgeting calculations */
 729	c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
 730	return 0;
 731}
 732
 733/*
 734 * init_constants_master - initialize UBIFS constants.
 735 * @c: UBIFS file-system description object
 736 *
 737 * This is a helper function which initializes various UBIFS constants after
 738 * the master node has been read. It also checks various UBIFS parameters and
 739 * makes sure they are all right.
 740 */
 741static void init_constants_master(struct ubifs_info *c)
 742{
 743	long long tmp64;
 744
 745	c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
 746	c->report_rp_size = ubifs_reported_space(c, c->rp_size);
 747
 748	/*
 749	 * Calculate total amount of FS blocks. This number is not used
 750	 * internally because it does not make much sense for UBIFS, but it is
 751	 * necessary to report something for the 'statfs()' call.
 752	 *
 753	 * Subtract the LEB reserved for GC, the LEB which is reserved for
 754	 * deletions, minimum LEBs for the index, and assume only one journal
 755	 * head is available.
 756	 */
 757	tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
 758	tmp64 *= (long long)c->leb_size - c->leb_overhead;
 759	tmp64 = ubifs_reported_space(c, tmp64);
 760	c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
 761}
 762
 763/**
 764 * take_gc_lnum - reserve GC LEB.
 765 * @c: UBIFS file-system description object
 766 *
 767 * This function ensures that the LEB reserved for garbage collection is marked
 768 * as "taken" in lprops. We also have to set free space to LEB size and dirty
 769 * space to zero, because lprops may contain out-of-date information if the
 770 * file-system was un-mounted before it has been committed. This function
 771 * returns zero in case of success and a negative error code in case of
 772 * failure.
 773 */
 774static int take_gc_lnum(struct ubifs_info *c)
 775{
 776	int err;
 777
 778	if (c->gc_lnum == -1) {
 779		ubifs_err("no LEB for GC");
 780		return -EINVAL;
 781	}
 782
 783	/* And we have to tell lprops that this LEB is taken */
 784	err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
 785				  LPROPS_TAKEN, 0, 0);
 786	return err;
 787}
 788
 789/**
 790 * alloc_wbufs - allocate write-buffers.
 791 * @c: UBIFS file-system description object
 792 *
 793 * This helper function allocates and initializes UBIFS write-buffers. Returns
 794 * zero in case of success and %-ENOMEM in case of failure.
 795 */
 796static int alloc_wbufs(struct ubifs_info *c)
 797{
 798	int i, err;
 799
 800	c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
 801			   GFP_KERNEL);
 802	if (!c->jheads)
 803		return -ENOMEM;
 804
 805	/* Initialize journal heads */
 806	for (i = 0; i < c->jhead_cnt; i++) {
 807		INIT_LIST_HEAD(&c->jheads[i].buds_list);
 808		err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
 809		if (err)
 810			return err;
 811
 812		c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
 813		c->jheads[i].wbuf.jhead = i;
 814		c->jheads[i].grouped = 1;
 815	}
 816
 817	c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
 818	/*
 819	 * Garbage Collector head likely contains long-term data and
 820	 * does not need to be synchronized by timer. Also GC head nodes are
 821	 * not grouped.
 822	 */
 823	c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
 824	c->jheads[GCHD].wbuf.no_timer = 1;
 825	c->jheads[GCHD].grouped = 0;
 826
 827	return 0;
 828}
 829
 830/**
 831 * free_wbufs - free write-buffers.
 832 * @c: UBIFS file-system description object
 833 */
 834static void free_wbufs(struct ubifs_info *c)
 835{
 836	int i;
 837
 838	if (c->jheads) {
 839		for (i = 0; i < c->jhead_cnt; i++) {
 840			kfree(c->jheads[i].wbuf.buf);
 841			kfree(c->jheads[i].wbuf.inodes);
 842		}
 843		kfree(c->jheads);
 844		c->jheads = NULL;
 845	}
 846}
 847
 848/**
 849 * free_orphans - free orphans.
 850 * @c: UBIFS file-system description object
 851 */
 852static void free_orphans(struct ubifs_info *c)
 853{
 854	struct ubifs_orphan *orph;
 855
 856	while (c->orph_dnext) {
 857		orph = c->orph_dnext;
 858		c->orph_dnext = orph->dnext;
 859		list_del(&orph->list);
 860		kfree(orph);
 861	}
 862
 863	while (!list_empty(&c->orph_list)) {
 864		orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
 865		list_del(&orph->list);
 866		kfree(orph);
 867		dbg_err("orphan list not empty at unmount");
 868	}
 869
 870	vfree(c->orph_buf);
 871	c->orph_buf = NULL;
 872}
 873
 874/**
 875 * free_buds - free per-bud objects.
 876 * @c: UBIFS file-system description object
 877 */
 878static void free_buds(struct ubifs_info *c)
 879{
 880	struct rb_node *this = c->buds.rb_node;
 881	struct ubifs_bud *bud;
 882
 883	while (this) {
 884		if (this->rb_left)
 885			this = this->rb_left;
 886		else if (this->rb_right)
 887			this = this->rb_right;
 888		else {
 889			bud = rb_entry(this, struct ubifs_bud, rb);
 890			this = rb_parent(this);
 891			if (this) {
 892				if (this->rb_left == &bud->rb)
 893					this->rb_left = NULL;
 894				else
 895					this->rb_right = NULL;
 896			}
 897			kfree(bud);
 898		}
 899	}
 900}
 901
 902/**
 903 * check_volume_empty - check if the UBI volume is empty.
 904 * @c: UBIFS file-system description object
 905 *
 906 * This function checks if the UBIFS volume is empty by looking if its LEBs are
 907 * mapped or not. The result of checking is stored in the @c->empty variable.
 908 * Returns zero in case of success and a negative error code in case of
 909 * failure.
 910 */
 911static int check_volume_empty(struct ubifs_info *c)
 912{
 913	int lnum, err;
 914
 915	c->empty = 1;
 916	for (lnum = 0; lnum < c->leb_cnt; lnum++) {
 917		err = ubifs_is_mapped(c, lnum);
 918		if (unlikely(err < 0))
 919			return err;
 920		if (err == 1) {
 921			c->empty = 0;
 922			break;
 923		}
 924
 925		cond_resched();
 926	}
 927
 928	return 0;
 929}
 930
 931/*
 932 * UBIFS mount options.
 933 *
 934 * Opt_fast_unmount: do not run a journal commit before un-mounting
 935 * Opt_norm_unmount: run a journal commit before un-mounting
 936 * Opt_bulk_read: enable bulk-reads
 937 * Opt_no_bulk_read: disable bulk-reads
 938 * Opt_chk_data_crc: check CRCs when reading data nodes
 939 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
 940 * Opt_override_compr: override default compressor
 941 * Opt_err: just end of array marker
 942 */
 943enum {
 944	Opt_fast_unmount,
 945	Opt_norm_unmount,
 946	Opt_bulk_read,
 947	Opt_no_bulk_read,
 948	Opt_chk_data_crc,
 949	Opt_no_chk_data_crc,
 950	Opt_override_compr,
 951	Opt_err,
 952};
 953
 954static const match_table_t tokens = {
 955	{Opt_fast_unmount, "fast_unmount"},
 956	{Opt_norm_unmount, "norm_unmount"},
 957	{Opt_bulk_read, "bulk_read"},
 958	{Opt_no_bulk_read, "no_bulk_read"},
 959	{Opt_chk_data_crc, "chk_data_crc"},
 960	{Opt_no_chk_data_crc, "no_chk_data_crc"},
 961	{Opt_override_compr, "compr=%s"},
 962	{Opt_err, NULL},
 963};
 964
 965/**
 966 * parse_standard_option - parse a standard mount option.
 967 * @option: the option to parse
 968 *
 969 * Normally, standard mount options like "sync" are passed to file-systems as
 970 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
 971 * be present in the options string. This function tries to deal with this
 972 * situation and parse standard options. Returns 0 if the option was not
 973 * recognized, and the corresponding integer flag if it was.
 974 *
 975 * UBIFS is only interested in the "sync" option, so do not check for anything
 976 * else.
 977 */
 978static int parse_standard_option(const char *option)
 979{
 980	ubifs_msg("parse %s", option);
 981	if (!strcmp(option, "sync"))
 982		return MS_SYNCHRONOUS;
 983	return 0;
 984}
 985
 986/**
 987 * ubifs_parse_options - parse mount parameters.
 988 * @c: UBIFS file-system description object
 989 * @options: parameters to parse
 990 * @is_remount: non-zero if this is FS re-mount
 991 *
 992 * This function parses UBIFS mount options and returns zero in case success
 993 * and a negative error code in case of failure.
 994 */
 995static int ubifs_parse_options(struct ubifs_info *c, char *options,
 996			       int is_remount)
 997{
 998	char *p;
 999	substring_t args[MAX_OPT_ARGS];
1000
1001	if (!options)
1002		return 0;
1003
1004	while ((p = strsep(&options, ","))) {
1005		int token;
1006
1007		if (!*p)
1008			continue;
1009
1010		token = match_token(p, tokens, args);
1011		switch (token) {
1012		/*
1013		 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
1014		 * We accept them in order to be backward-compatible. But this
1015		 * should be removed at some point.
1016		 */
1017		case Opt_fast_unmount:
1018			c->mount_opts.unmount_mode = 2;
1019			break;
1020		case Opt_norm_unmount:
1021			c->mount_opts.unmount_mode = 1;
1022			break;
1023		case Opt_bulk_read:
1024			c->mount_opts.bulk_read = 2;
1025			c->bulk_read = 1;
1026			break;
1027		case Opt_no_bulk_read:
1028			c->mount_opts.bulk_read = 1;
1029			c->bulk_read = 0;
1030			break;
1031		case Opt_chk_data_crc:
1032			c->mount_opts.chk_data_crc = 2;
1033			c->no_chk_data_crc = 0;
1034			break;
1035		case Opt_no_chk_data_crc:
1036			c->mount_opts.chk_data_crc = 1;
1037			c->no_chk_data_crc = 1;
1038			break;
1039		case Opt_override_compr:
1040		{
1041			char *name = match_strdup(&args[0]);
1042
1043			if (!name)
1044				return -ENOMEM;
1045			if (!strcmp(name, "none"))
1046				c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1047			else if (!strcmp(name, "lzo"))
1048				c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1049			else if (!strcmp(name, "zlib"))
1050				c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1051			else {
1052				ubifs_err("unknown compressor \"%s\"", name);
1053				kfree(name);
1054				return -EINVAL;
1055			}
1056			kfree(name);
1057			c->mount_opts.override_compr = 1;
1058			c->default_compr = c->mount_opts.compr_type;
1059			break;
1060		}
1061		default:
1062		{
1063			unsigned long flag;
1064			struct super_block *sb = c->vfs_sb;
1065
1066			flag = parse_standard_option(p);
1067			if (!flag) {
1068				ubifs_err("unrecognized mount option \"%s\" "
1069					  "or missing value", p);
1070				return -EINVAL;
1071			}
1072			sb->s_flags |= flag;
1073			break;
1074		}
1075		}
1076	}
1077
1078	return 0;
1079}
1080
1081/**
1082 * destroy_journal - destroy journal data structures.
1083 * @c: UBIFS file-system description object
1084 *
1085 * This function destroys journal data structures including those that may have
1086 * been created by recovery functions.
1087 */
1088static void destroy_journal(struct ubifs_info *c)
1089{
1090	while (!list_empty(&c->unclean_leb_list)) {
1091		struct ubifs_unclean_leb *ucleb;
1092
1093		ucleb = list_entry(c->unclean_leb_list.next,
1094				   struct ubifs_unclean_leb, list);
1095		list_del(&ucleb->list);
1096		kfree(ucleb);
1097	}
1098	while (!list_empty(&c->old_buds)) {
1099		struct ubifs_bud *bud;
1100
1101		bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1102		list_del(&bud->list);
1103		kfree(bud);
1104	}
1105	ubifs_destroy_idx_gc(c);
1106	ubifs_destroy_size_tree(c);
1107	ubifs_tnc_close(c);
1108	free_buds(c);
1109}
1110
1111/**
1112 * bu_init - initialize bulk-read information.
1113 * @c: UBIFS file-system description object
1114 */
1115static void bu_init(struct ubifs_info *c)
1116{
1117	ubifs_assert(c->bulk_read == 1);
1118
1119	if (c->bu.buf)
1120		return; /* Already initialized */
1121
1122again:
1123	c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1124	if (!c->bu.buf) {
1125		if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1126			c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1127			goto again;
1128		}
1129
1130		/* Just disable bulk-read */
1131		ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1132			   "disabling it", c->max_bu_buf_len);
1133		c->mount_opts.bulk_read = 1;
1134		c->bulk_read = 0;
1135		return;
1136	}
1137}
1138
1139/**
1140 * check_free_space - check if there is enough free space to mount.
1141 * @c: UBIFS file-system description object
1142 *
1143 * This function makes sure UBIFS has enough free space to be mounted in
1144 * read/write mode. UBIFS must always have some free space to allow deletions.
1145 */
1146static int check_free_space(struct ubifs_info *c)
1147{
1148	ubifs_assert(c->dark_wm > 0);
1149	if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1150		ubifs_err("insufficient free space to mount in R/W mode");
1151		dbg_dump_budg(c, &c->bi);
1152		dbg_dump_lprops(c);
1153		return -ENOSPC;
1154	}
1155	return 0;
1156}
1157
1158/**
1159 * mount_ubifs - mount UBIFS file-system.
1160 * @c: UBIFS file-system description object
1161 *
1162 * This function mounts UBIFS file system. Returns zero in case of success and
1163 * a negative error code in case of failure.
1164 *
1165 * Note, the function does not de-allocate resources it it fails half way
1166 * through, and the caller has to do this instead.
1167 */
1168static int mount_ubifs(struct ubifs_info *c)
1169{
1170	int err;
1171	long long x;
1172	size_t sz;
1173
1174	c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1175	err = init_constants_early(c);
1176	if (err)
1177		return err;
1178
1179	err = ubifs_debugging_init(c);
1180	if (err)
1181		return err;
1182
1183	err = check_volume_empty(c);
1184	if (err)
1185		goto out_free;
1186
1187	if (c->empty && (c->ro_mount || c->ro_media)) {
1188		/*
1189		 * This UBI volume is empty, and read-only, or the file system
1190		 * is mounted read-only - we cannot format it.
1191		 */
1192		ubifs_err("can't format empty UBI volume: read-only %s",
1193			  c->ro_media ? "UBI volume" : "mount");
1194		err = -EROFS;
1195		goto out_free;
1196	}
1197
1198	if (c->ro_media && !c->ro_mount) {
1199		ubifs_err("cannot mount read-write - read-only media");
1200		err = -EROFS;
1201		goto out_free;
1202	}
1203
1204	/*
1205	 * The requirement for the buffer is that it should fit indexing B-tree
1206	 * height amount of integers. We assume the height if the TNC tree will
1207	 * never exceed 64.
1208	 */
1209	err = -ENOMEM;
1210	c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1211	if (!c->bottom_up_buf)
1212		goto out_free;
1213
1214	c->sbuf = vmalloc(c->leb_size);
1215	if (!c->sbuf)
1216		goto out_free;
1217
1218	if (!c->ro_mount) {
1219		c->ileb_buf = vmalloc(c->leb_size);
1220		if (!c->ileb_buf)
1221			goto out_free;
1222	}
1223
1224	if (c->bulk_read == 1)
1225		bu_init(c);
1226
1227	if (!c->ro_mount) {
1228		c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
1229					       GFP_KERNEL);
1230		if (!c->write_reserve_buf)
1231			goto out_free;
1232	}
1233
1234	c->mounting = 1;
1235
1236	err = ubifs_read_superblock(c);
1237	if (err)
1238		goto out_free;
1239
1240	/*
1241	 * Make sure the compressor which is set as default in the superblock
1242	 * or overridden by mount options is actually compiled in.
1243	 */
1244	if (!ubifs_compr_present(c->default_compr)) {
1245		ubifs_err("'compressor \"%s\" is not compiled in",
1246			  ubifs_compr_name(c->default_compr));
1247		err = -ENOTSUPP;
1248		goto out_free;
1249	}
1250
1251	err = init_constants_sb(c);
1252	if (err)
1253		goto out_free;
1254
1255	sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1256	sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1257	c->cbuf = kmalloc(sz, GFP_NOFS);
1258	if (!c->cbuf) {
1259		err = -ENOMEM;
1260		goto out_free;
1261	}
1262
1263	err = alloc_wbufs(c);
1264	if (err)
1265		goto out_cbuf;
1266
1267	sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1268	if (!c->ro_mount) {
1269		/* Create background thread */
1270		c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1271		if (IS_ERR(c->bgt)) {
1272			err = PTR_ERR(c->bgt);
1273			c->bgt = NULL;
1274			ubifs_err("cannot spawn \"%s\", error %d",
1275				  c->bgt_name, err);
1276			goto out_wbufs;
1277		}
1278		wake_up_process(c->bgt);
1279	}
1280
1281	err = ubifs_read_master(c);
1282	if (err)
1283		goto out_master;
1284
1285	init_constants_master(c);
1286
1287	if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1288		ubifs_msg("recovery needed");
1289		c->need_recovery = 1;
1290	}
1291
1292	if (c->need_recovery && !c->ro_mount) {
1293		err = ubifs_recover_inl_heads(c, c->sbuf);
1294		if (err)
1295			goto out_master;
1296	}
1297
1298	err = ubifs_lpt_init(c, 1, !c->ro_mount);
1299	if (err)
1300		goto out_master;
1301
1302	if (!c->ro_mount && c->space_fixup) {
1303		err = ubifs_fixup_free_space(c);
1304		if (err)
1305			goto out_master;
1306	}
1307
1308	if (!c->ro_mount) {
1309		/*
1310		 * Set the "dirty" flag so that if we reboot uncleanly we
1311		 * will notice this immediately on the next mount.
1312		 */
1313		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1314		err = ubifs_write_master(c);
1315		if (err)
1316			goto out_lpt;
1317	}
1318
1319	err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1320	if (err)
1321		goto out_lpt;
1322
1323	err = ubifs_replay_journal(c);
1324	if (err)
1325		goto out_journal;
1326
1327	/* Calculate 'min_idx_lebs' after journal replay */
1328	c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1329
1330	err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1331	if (err)
1332		goto out_orphans;
1333
1334	if (!c->ro_mount) {
1335		int lnum;
1336
1337		err = check_free_space(c);
1338		if (err)
1339			goto out_orphans;
1340
1341		/* Check for enough log space */
1342		lnum = c->lhead_lnum + 1;
1343		if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1344			lnum = UBIFS_LOG_LNUM;
1345		if (lnum == c->ltail_lnum) {
1346			err = ubifs_consolidate_log(c);
1347			if (err)
1348				goto out_orphans;
1349		}
1350
1351		if (c->need_recovery) {
1352			err = ubifs_recover_size(c);
1353			if (err)
1354				goto out_orphans;
1355			err = ubifs_rcvry_gc_commit(c);
1356			if (err)
1357				goto out_orphans;
1358		} else {
1359			err = take_gc_lnum(c);
1360			if (err)
1361				goto out_orphans;
1362
1363			/*
1364			 * GC LEB may contain garbage if there was an unclean
1365			 * reboot, and it should be un-mapped.
1366			 */
1367			err = ubifs_leb_unmap(c, c->gc_lnum);
1368			if (err)
1369				goto out_orphans;
1370		}
1371
1372		err = dbg_check_lprops(c);
1373		if (err)
1374			goto out_orphans;
1375	} else if (c->need_recovery) {
1376		err = ubifs_recover_size(c);
1377		if (err)
1378			goto out_orphans;
1379	} else {
1380		/*
1381		 * Even if we mount read-only, we have to set space in GC LEB
1382		 * to proper value because this affects UBIFS free space
1383		 * reporting. We do not want to have a situation when
1384		 * re-mounting from R/O to R/W changes amount of free space.
1385		 */
1386		err = take_gc_lnum(c);
1387		if (err)
1388			goto out_orphans;
1389	}
1390
1391	spin_lock(&ubifs_infos_lock);
1392	list_add_tail(&c->infos_list, &ubifs_infos);
1393	spin_unlock(&ubifs_infos_lock);
1394
1395	if (c->need_recovery) {
1396		if (c->ro_mount)
1397			ubifs_msg("recovery deferred");
1398		else {
1399			c->need_recovery = 0;
1400			ubifs_msg("recovery completed");
1401			/*
1402			 * GC LEB has to be empty and taken at this point. But
1403			 * the journal head LEBs may also be accounted as
1404			 * "empty taken" if they are empty.
1405			 */
1406			ubifs_assert(c->lst.taken_empty_lebs > 0);
1407		}
1408	} else
1409		ubifs_assert(c->lst.taken_empty_lebs > 0);
1410
1411	err = dbg_check_filesystem(c);
1412	if (err)
1413		goto out_infos;
1414
1415	err = dbg_debugfs_init_fs(c);
1416	if (err)
1417		goto out_infos;
1418
1419	c->mounting = 0;
1420
1421	ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1422		  c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1423	if (c->ro_mount)
1424		ubifs_msg("mounted read-only");
1425	x = (long long)c->main_lebs * c->leb_size;
1426	ubifs_msg("file system size:   %lld bytes (%lld KiB, %lld MiB, %d "
1427		  "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1428	x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1429	ubifs_msg("journal size:       %lld bytes (%lld KiB, %lld MiB, %d "
1430		  "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1431	ubifs_msg("media format:       w%d/r%d (latest is w%d/r%d)",
1432		  c->fmt_version, c->ro_compat_version,
1433		  UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1434	ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1435	ubifs_msg("reserved for root:  %llu bytes (%llu KiB)",
1436		c->report_rp_size, c->report_rp_size >> 10);
1437
1438	dbg_msg("compiled on:         " __DATE__ " at " __TIME__);
1439	dbg_msg("min. I/O unit size:  %d bytes", c->min_io_size);
1440	dbg_msg("max. write size:     %d bytes", c->max_write_size);
1441	dbg_msg("LEB size:            %d bytes (%d KiB)",
1442		c->leb_size, c->leb_size >> 10);
1443	dbg_msg("data journal heads:  %d",
1444		c->jhead_cnt - NONDATA_JHEADS_CNT);
1445	dbg_msg("UUID:                %pUB", c->uuid);
1446	dbg_msg("big_lpt              %d", c->big_lpt);
1447	dbg_msg("log LEBs:            %d (%d - %d)",
1448		c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1449	dbg_msg("LPT area LEBs:       %d (%d - %d)",
1450		c->lpt_lebs, c->lpt_first, c->lpt_last);
1451	dbg_msg("orphan area LEBs:    %d (%d - %d)",
1452		c->orph_lebs, c->orph_first, c->orph_last);
1453	dbg_msg("main area LEBs:      %d (%d - %d)",
1454		c->main_lebs, c->main_first, c->leb_cnt - 1);
1455	dbg_msg("index LEBs:          %d", c->lst.idx_lebs);
1456	dbg_msg("total index bytes:   %lld (%lld KiB, %lld MiB)",
1457		c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1458		c->bi.old_idx_sz >> 20);
1459	dbg_msg("key hash type:       %d", c->key_hash_type);
1460	dbg_msg("tree fanout:         %d", c->fanout);
1461	dbg_msg("reserved GC LEB:     %d", c->gc_lnum);
1462	dbg_msg("first main LEB:      %d", c->main_first);
1463	dbg_msg("max. znode size      %d", c->max_znode_sz);
1464	dbg_msg("max. index node size %d", c->max_idx_node_sz);
1465	dbg_msg("node sizes:          data %zu, inode %zu, dentry %zu",
1466		UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1467	dbg_msg("node sizes:          trun %zu, sb %zu, master %zu",
1468		UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1469	dbg_msg("node sizes:          ref %zu, cmt. start %zu, orph %zu",
1470		UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1471	dbg_msg("max. node sizes:     data %zu, inode %zu dentry %zu, idx %d",
1472		UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1473		UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1474	dbg_msg("dead watermark:      %d", c->dead_wm);
1475	dbg_msg("dark watermark:      %d", c->dark_wm);
1476	dbg_msg("LEB overhead:        %d", c->leb_overhead);
1477	x = (long long)c->main_lebs * c->dark_wm;
1478	dbg_msg("max. dark space:     %lld (%lld KiB, %lld MiB)",
1479		x, x >> 10, x >> 20);
1480	dbg_msg("maximum bud bytes:   %lld (%lld KiB, %lld MiB)",
1481		c->max_bud_bytes, c->max_bud_bytes >> 10,
1482		c->max_bud_bytes >> 20);
1483	dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1484		c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1485		c->bg_bud_bytes >> 20);
1486	dbg_msg("current bud bytes    %lld (%lld KiB, %lld MiB)",
1487		c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1488	dbg_msg("max. seq. number:    %llu", c->max_sqnum);
1489	dbg_msg("commit number:       %llu", c->cmt_no);
1490
1491	return 0;
1492
1493out_infos:
1494	spin_lock(&ubifs_infos_lock);
1495	list_del(&c->infos_list);
1496	spin_unlock(&ubifs_infos_lock);
1497out_orphans:
1498	free_orphans(c);
1499out_journal:
1500	destroy_journal(c);
1501out_lpt:
1502	ubifs_lpt_free(c, 0);
1503out_master:
1504	kfree(c->mst_node);
1505	kfree(c->rcvrd_mst_node);
1506	if (c->bgt)
1507		kthread_stop(c->bgt);
1508out_wbufs:
1509	free_wbufs(c);
1510out_cbuf:
1511	kfree(c->cbuf);
1512out_free:
1513	kfree(c->write_reserve_buf);
1514	kfree(c->bu.buf);
1515	vfree(c->ileb_buf);
1516	vfree(c->sbuf);
1517	kfree(c->bottom_up_buf);
1518	ubifs_debugging_exit(c);
1519	return err;
1520}
1521
1522/**
1523 * ubifs_umount - un-mount UBIFS file-system.
1524 * @c: UBIFS file-system description object
1525 *
1526 * Note, this function is called to free allocated resourced when un-mounting,
1527 * as well as free resources when an error occurred while we were half way
1528 * through mounting (error path cleanup function). So it has to make sure the
1529 * resource was actually allocated before freeing it.
1530 */
1531static void ubifs_umount(struct ubifs_info *c)
1532{
1533	dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1534		c->vi.vol_id);
1535
1536	dbg_debugfs_exit_fs(c);
1537	spin_lock(&ubifs_infos_lock);
1538	list_del(&c->infos_list);
1539	spin_unlock(&ubifs_infos_lock);
1540
1541	if (c->bgt)
1542		kthread_stop(c->bgt);
1543
1544	destroy_journal(c);
1545	free_wbufs(c);
1546	free_orphans(c);
1547	ubifs_lpt_free(c, 0);
1548
1549	kfree(c->cbuf);
1550	kfree(c->rcvrd_mst_node);
1551	kfree(c->mst_node);
1552	kfree(c->write_reserve_buf);
1553	kfree(c->bu.buf);
1554	vfree(c->ileb_buf);
1555	vfree(c->sbuf);
1556	kfree(c->bottom_up_buf);
1557	ubifs_debugging_exit(c);
1558}
1559
1560/**
1561 * ubifs_remount_rw - re-mount in read-write mode.
1562 * @c: UBIFS file-system description object
1563 *
1564 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1565 * mode. This function allocates the needed resources and re-mounts UBIFS in
1566 * read-write mode.
1567 */
1568static int ubifs_remount_rw(struct ubifs_info *c)
1569{
1570	int err, lnum;
1571
1572	if (c->rw_incompat) {
1573		ubifs_err("the file-system is not R/W-compatible");
1574		ubifs_msg("on-flash format version is w%d/r%d, but software "
1575			  "only supports up to version w%d/r%d", c->fmt_version,
1576			  c->ro_compat_version, UBIFS_FORMAT_VERSION,
1577			  UBIFS_RO_COMPAT_VERSION);
1578		return -EROFS;
1579	}
1580
1581	mutex_lock(&c->umount_mutex);
1582	dbg_save_space_info(c);
1583	c->remounting_rw = 1;
1584	c->ro_mount = 0;
1585
1586	err = check_free_space(c);
1587	if (err)
1588		goto out;
1589
1590	if (c->old_leb_cnt != c->leb_cnt) {
1591		struct ubifs_sb_node *sup;
1592
1593		sup = ubifs_read_sb_node(c);
1594		if (IS_ERR(sup)) {
1595			err = PTR_ERR(sup);
1596			goto out;
1597		}
1598		sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1599		err = ubifs_write_sb_node(c, sup);
1600		kfree(sup);
1601		if (err)
1602			goto out;
1603	}
1604
1605	if (c->need_recovery) {
1606		ubifs_msg("completing deferred recovery");
1607		err = ubifs_write_rcvrd_mst_node(c);
1608		if (err)
1609			goto out;
1610		err = ubifs_recover_size(c);
1611		if (err)
1612			goto out;
1613		err = ubifs_clean_lebs(c, c->sbuf);
1614		if (err)
1615			goto out;
1616		err = ubifs_recover_inl_heads(c, c->sbuf);
1617		if (err)
1618			goto out;
1619	} else {
1620		/* A readonly mount is not allowed to have orphans */
1621		ubifs_assert(c->tot_orphans == 0);
1622		err = ubifs_clear_orphans(c);
1623		if (err)
1624			goto out;
1625	}
1626
1627	if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1628		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1629		err = ubifs_write_master(c);
1630		if (err)
1631			goto out;
1632	}
1633
1634	c->ileb_buf = vmalloc(c->leb_size);
1635	if (!c->ileb_buf) {
1636		err = -ENOMEM;
1637		goto out;
1638	}
1639
1640	c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
1641	if (!c->write_reserve_buf)
1642		goto out;
1643
1644	err = ubifs_lpt_init(c, 0, 1);
1645	if (err)
1646		goto out;
1647
1648	/* Create background thread */
1649	c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1650	if (IS_ERR(c->bgt)) {
1651		err = PTR_ERR(c->bgt);
1652		c->bgt = NULL;
1653		ubifs_err("cannot spawn \"%s\", error %d",
1654			  c->bgt_name, err);
1655		goto out;
1656	}
1657	wake_up_process(c->bgt);
1658
1659	c->orph_buf = vmalloc(c->leb_size);
1660	if (!c->orph_buf) {
1661		err = -ENOMEM;
1662		goto out;
1663	}
1664
1665	/* Check for enough log space */
1666	lnum = c->lhead_lnum + 1;
1667	if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1668		lnum = UBIFS_LOG_LNUM;
1669	if (lnum == c->ltail_lnum) {
1670		err = ubifs_consolidate_log(c);
1671		if (err)
1672			goto out;
1673	}
1674
1675	if (c->need_recovery)
1676		err = ubifs_rcvry_gc_commit(c);
1677	else
1678		err = ubifs_leb_unmap(c, c->gc_lnum);
1679	if (err)
1680		goto out;
1681
1682	dbg_gen("re-mounted read-write");
1683	c->remounting_rw = 0;
1684
1685	if (c->need_recovery) {
1686		c->need_recovery = 0;
1687		ubifs_msg("deferred recovery completed");
1688	} else {
1689		/*
1690		 * Do not run the debugging space check if the were doing
1691		 * recovery, because when we saved the information we had the
1692		 * file-system in a state where the TNC and lprops has been
1693		 * modified in memory, but all the I/O operations (including a
1694		 * commit) were deferred. So the file-system was in
1695		 * "non-committed" state. Now the file-system is in committed
1696		 * state, and of course the amount of free space will change
1697		 * because, for example, the old index size was imprecise.
1698		 */
1699		err = dbg_check_space_info(c);
1700	}
1701
1702	if (c->space_fixup) {
1703		err = ubifs_fixup_free_space(c);
1704		if (err)
1705			goto out;
1706	}
1707
1708	mutex_unlock(&c->umount_mutex);
1709	return err;
1710
1711out:
1712	c->ro_mount = 1;
1713	vfree(c->orph_buf);
1714	c->orph_buf = NULL;
1715	if (c->bgt) {
1716		kthread_stop(c->bgt);
1717		c->bgt = NULL;
1718	}
1719	free_wbufs(c);
1720	kfree(c->write_reserve_buf);
1721	c->write_reserve_buf = NULL;
1722	vfree(c->ileb_buf);
1723	c->ileb_buf = NULL;
1724	ubifs_lpt_free(c, 1);
1725	c->remounting_rw = 0;
1726	mutex_unlock(&c->umount_mutex);
1727	return err;
1728}
1729
1730/**
1731 * ubifs_remount_ro - re-mount in read-only mode.
1732 * @c: UBIFS file-system description object
1733 *
1734 * We assume VFS has stopped writing. Possibly the background thread could be
1735 * running a commit, however kthread_stop will wait in that case.
1736 */
1737static void ubifs_remount_ro(struct ubifs_info *c)
1738{
1739	int i, err;
1740
1741	ubifs_assert(!c->need_recovery);
1742	ubifs_assert(!c->ro_mount);
1743
1744	mutex_lock(&c->umount_mutex);
1745	if (c->bgt) {
1746		kthread_stop(c->bgt);
1747		c->bgt = NULL;
1748	}
1749
1750	dbg_save_space_info(c);
1751
1752	for (i = 0; i < c->jhead_cnt; i++)
1753		ubifs_wbuf_sync(&c->jheads[i].wbuf);
1754
1755	c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1756	c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1757	c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1758	err = ubifs_write_master(c);
1759	if (err)
1760		ubifs_ro_mode(c, err);
1761
1762	vfree(c->orph_buf);
1763	c->orph_buf = NULL;
1764	kfree(c->write_reserve_buf);
1765	c->write_reserve_buf = NULL;
1766	vfree(c->ileb_buf);
1767	c->ileb_buf = NULL;
1768	ubifs_lpt_free(c, 1);
1769	c->ro_mount = 1;
1770	err = dbg_check_space_info(c);
1771	if (err)
1772		ubifs_ro_mode(c, err);
1773	mutex_unlock(&c->umount_mutex);
1774}
1775
1776static void ubifs_put_super(struct super_block *sb)
1777{
1778	int i;
1779	struct ubifs_info *c = sb->s_fs_info;
1780
1781	ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1782		  c->vi.vol_id);
1783
1784	/*
1785	 * The following asserts are only valid if there has not been a failure
1786	 * of the media. For example, there will be dirty inodes if we failed
1787	 * to write them back because of I/O errors.
1788	 */
1789	if (!c->ro_error) {
1790		ubifs_assert(c->bi.idx_growth == 0);
1791		ubifs_assert(c->bi.dd_growth == 0);
1792		ubifs_assert(c->bi.data_growth == 0);
1793	}
1794
1795	/*
1796	 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1797	 * and file system un-mount. Namely, it prevents the shrinker from
1798	 * picking this superblock for shrinking - it will be just skipped if
1799	 * the mutex is locked.
1800	 */
1801	mutex_lock(&c->umount_mutex);
1802	if (!c->ro_mount) {
1803		/*
1804		 * First of all kill the background thread to make sure it does
1805		 * not interfere with un-mounting and freeing resources.
1806		 */
1807		if (c->bgt) {
1808			kthread_stop(c->bgt);
1809			c->bgt = NULL;
1810		}
1811
1812		/*
1813		 * On fatal errors c->ro_error is set to 1, in which case we do
1814		 * not write the master node.
1815		 */
1816		if (!c->ro_error) {
1817			int err;
1818
1819			/* Synchronize write-buffers */
1820			for (i = 0; i < c->jhead_cnt; i++)
1821				ubifs_wbuf_sync(&c->jheads[i].wbuf);
1822
1823			/*
1824			 * We are being cleanly unmounted which means the
1825			 * orphans were killed - indicate this in the master
1826			 * node. Also save the reserved GC LEB number.
1827			 */
1828			c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1829			c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1830			c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1831			err = ubifs_write_master(c);
1832			if (err)
1833				/*
1834				 * Recovery will attempt to fix the master area
1835				 * next mount, so we just print a message and
1836				 * continue to unmount normally.
1837				 */
1838				ubifs_err("failed to write master node, "
1839					  "error %d", err);
1840		} else {
1841			for (i = 0; i < c->jhead_cnt; i++)
1842				/* Make sure write-buffer timers are canceled */
1843				hrtimer_cancel(&c->jheads[i].wbuf.timer);
1844		}
1845	}
1846
1847	ubifs_umount(c);
1848	bdi_destroy(&c->bdi);
1849	ubi_close_volume(c->ubi);
1850	mutex_unlock(&c->umount_mutex);
1851}
1852
1853static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1854{
1855	int err;
1856	struct ubifs_info *c = sb->s_fs_info;
1857
1858	dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1859
1860	err = ubifs_parse_options(c, data, 1);
1861	if (err) {
1862		ubifs_err("invalid or unknown remount parameter");
1863		return err;
1864	}
1865
1866	if (c->ro_mount && !(*flags & MS_RDONLY)) {
1867		if (c->ro_error) {
1868			ubifs_msg("cannot re-mount R/W due to prior errors");
1869			return -EROFS;
1870		}
1871		if (c->ro_media) {
1872			ubifs_msg("cannot re-mount R/W - UBI volume is R/O");
1873			return -EROFS;
1874		}
1875		err = ubifs_remount_rw(c);
1876		if (err)
1877			return err;
1878	} else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1879		if (c->ro_error) {
1880			ubifs_msg("cannot re-mount R/O due to prior errors");
1881			return -EROFS;
1882		}
1883		ubifs_remount_ro(c);
1884	}
1885
1886	if (c->bulk_read == 1)
1887		bu_init(c);
1888	else {
1889		dbg_gen("disable bulk-read");
1890		kfree(c->bu.buf);
1891		c->bu.buf = NULL;
1892	}
1893
1894	ubifs_assert(c->lst.taken_empty_lebs > 0);
1895	return 0;
1896}
1897
1898const struct super_operations ubifs_super_operations = {
1899	.alloc_inode   = ubifs_alloc_inode,
1900	.destroy_inode = ubifs_destroy_inode,
1901	.put_super     = ubifs_put_super,
1902	.write_inode   = ubifs_write_inode,
1903	.evict_inode   = ubifs_evict_inode,
1904	.statfs        = ubifs_statfs,
1905	.dirty_inode   = ubifs_dirty_inode,
1906	.remount_fs    = ubifs_remount_fs,
1907	.show_options  = ubifs_show_options,
1908	.sync_fs       = ubifs_sync_fs,
1909};
1910
1911/**
1912 * open_ubi - parse UBI device name string and open the UBI device.
1913 * @name: UBI volume name
1914 * @mode: UBI volume open mode
1915 *
1916 * The primary method of mounting UBIFS is by specifying the UBI volume
1917 * character device node path. However, UBIFS may also be mounted withoug any
1918 * character device node using one of the following methods:
1919 *
1920 * o ubiX_Y    - mount UBI device number X, volume Y;
1921 * o ubiY      - mount UBI device number 0, volume Y;
1922 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1923 * o ubi:NAME  - mount UBI device 0, volume with name NAME.
1924 *
1925 * Alternative '!' separator may be used instead of ':' (because some shells
1926 * like busybox may interpret ':' as an NFS host name separator). This function
1927 * returns UBI volume description object in case of success and a negative
1928 * error code in case of failure.
1929 */
1930static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1931{
1932	struct ubi_volume_desc *ubi;
1933	int dev, vol;
1934	char *endptr;
1935
1936	/* First, try to open using the device node path method */
1937	ubi = ubi_open_volume_path(name, mode);
1938	if (!IS_ERR(ubi))
1939		return ubi;
1940
1941	/* Try the "nodev" method */
1942	if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1943		return ERR_PTR(-EINVAL);
1944
1945	/* ubi:NAME method */
1946	if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1947		return ubi_open_volume_nm(0, name + 4, mode);
1948
1949	if (!isdigit(name[3]))
1950		return ERR_PTR(-EINVAL);
1951
1952	dev = simple_strtoul(name + 3, &endptr, 0);
1953
1954	/* ubiY method */
1955	if (*endptr == '\0')
1956		return ubi_open_volume(0, dev, mode);
1957
1958	/* ubiX_Y method */
1959	if (*endptr == '_' && isdigit(endptr[1])) {
1960		vol = simple_strtoul(endptr + 1, &endptr, 0);
1961		if (*endptr != '\0')
1962			return ERR_PTR(-EINVAL);
1963		return ubi_open_volume(dev, vol, mode);
1964	}
1965
1966	/* ubiX:NAME method */
1967	if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1968		return ubi_open_volume_nm(dev, ++endptr, mode);
1969
1970	return ERR_PTR(-EINVAL);
1971}
1972
1973static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1974{
1975	struct ubifs_info *c;
1976
1977	c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1978	if (c) {
1979		spin_lock_init(&c->cnt_lock);
1980		spin_lock_init(&c->cs_lock);
1981		spin_lock_init(&c->buds_lock);
1982		spin_lock_init(&c->space_lock);
1983		spin_lock_init(&c->orphan_lock);
1984		init_rwsem(&c->commit_sem);
1985		mutex_init(&c->lp_mutex);
1986		mutex_init(&c->tnc_mutex);
1987		mutex_init(&c->log_mutex);
1988		mutex_init(&c->mst_mutex);
1989		mutex_init(&c->umount_mutex);
1990		mutex_init(&c->bu_mutex);
1991		mutex_init(&c->write_reserve_mutex);
1992		init_waitqueue_head(&c->cmt_wq);
1993		c->buds = RB_ROOT;
1994		c->old_idx = RB_ROOT;
1995		c->size_tree = RB_ROOT;
1996		c->orph_tree = RB_ROOT;
1997		INIT_LIST_HEAD(&c->infos_list);
1998		INIT_LIST_HEAD(&c->idx_gc);
1999		INIT_LIST_HEAD(&c->replay_list);
2000		INIT_LIST_HEAD(&c->replay_buds);
2001		INIT_LIST_HEAD(&c->uncat_list);
2002		INIT_LIST_HEAD(&c->empty_list);
2003		INIT_LIST_HEAD(&c->freeable_list);
2004		INIT_LIST_HEAD(&c->frdi_idx_list);
2005		INIT_LIST_HEAD(&c->unclean_leb_list);
2006		INIT_LIST_HEAD(&c->old_buds);
2007		INIT_LIST_HEAD(&c->orph_list);
2008		INIT_LIST_HEAD(&c->orph_new);
2009		c->no_chk_data_crc = 1;
2010
2011		c->highest_inum = UBIFS_FIRST_INO;
2012		c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
2013
2014		ubi_get_volume_info(ubi, &c->vi);
2015		ubi_get_device_info(c->vi.ubi_num, &c->di);
2016	}
2017	return c;
2018}
2019
2020static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
2021{
2022	struct ubifs_info *c = sb->s_fs_info;
2023	struct inode *root;
2024	int err;
2025
2026	c->vfs_sb = sb;
2027	/* Re-open the UBI device in read-write mode */
2028	c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2029	if (IS_ERR(c->ubi)) {
2030		err = PTR_ERR(c->ubi);
2031		goto out;
2032	}
2033
2034	/*
2035	 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2036	 * UBIFS, I/O is not deferred, it is done immediately in readpage,
2037	 * which means the user would have to wait not just for their own I/O
2038	 * but the read-ahead I/O as well i.e. completely pointless.
2039	 *
2040	 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
2041	 */
2042	c->bdi.name = "ubifs",
2043	c->bdi.capabilities = BDI_CAP_MAP_COPY;
2044	err  = bdi_init(&c->bdi);
2045	if (err)
2046		goto out_close;
2047	err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
2048			   c->vi.ubi_num, c->vi.vol_id);
2049	if (err)
2050		goto out_bdi;
2051
2052	err = ubifs_parse_options(c, data, 0);
2053	if (err)
2054		goto out_bdi;
2055
2056	sb->s_bdi = &c->bdi;
2057	sb->s_fs_info = c;
2058	sb->s_magic = UBIFS_SUPER_MAGIC;
2059	sb->s_blocksize = UBIFS_BLOCK_SIZE;
2060	sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2061	sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2062	if (c->max_inode_sz > MAX_LFS_FILESIZE)
2063		sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2064	sb->s_op = &ubifs_super_operations;
2065
2066	mutex_lock(&c->umount_mutex);
2067	err = mount_ubifs(c);
2068	if (err) {
2069		ubifs_assert(err < 0);
2070		goto out_unlock;
2071	}
2072
2073	/* Read the root inode */
2074	root = ubifs_iget(sb, UBIFS_ROOT_INO);
2075	if (IS_ERR(root)) {
2076		err = PTR_ERR(root);
2077		goto out_umount;
2078	}
2079
2080	sb->s_root = d_alloc_root(root);
2081	if (!sb->s_root)
2082		goto out_iput;
2083
2084	mutex_unlock(&c->umount_mutex);
2085	return 0;
2086
2087out_iput:
2088	iput(root);
2089out_umount:
2090	ubifs_umount(c);
2091out_unlock:
2092	mutex_unlock(&c->umount_mutex);
2093out_bdi:
2094	bdi_destroy(&c->bdi);
2095out_close:
2096	ubi_close_volume(c->ubi);
2097out:
2098	return err;
2099}
2100
2101static int sb_test(struct super_block *sb, void *data)
2102{
2103	struct ubifs_info *c1 = data;
2104	struct ubifs_info *c = sb->s_fs_info;
2105
2106	return c->vi.cdev == c1->vi.cdev;
2107}
2108
2109static int sb_set(struct super_block *sb, void *data)
2110{
2111	sb->s_fs_info = data;
2112	return set_anon_super(sb, NULL);
2113}
2114
2115static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2116			const char *name, void *data)
2117{
2118	struct ubi_volume_desc *ubi;
2119	struct ubifs_info *c;
2120	struct super_block *sb;
2121	int err;
2122
2123	dbg_gen("name %s, flags %#x", name, flags);
2124
2125	/*
2126	 * Get UBI device number and volume ID. Mount it read-only so far
2127	 * because this might be a new mount point, and UBI allows only one
2128	 * read-write user at a time.
2129	 */
2130	ubi = open_ubi(name, UBI_READONLY);
2131	if (IS_ERR(ubi)) {
2132		dbg_err("cannot open \"%s\", error %d",
2133			name, (int)PTR_ERR(ubi));
2134		return ERR_CAST(ubi);
2135	}
2136
2137	c = alloc_ubifs_info(ubi);
2138	if (!c) {
2139		err = -ENOMEM;
2140		goto out_close;
2141	}
2142
2143	dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2144
2145	sb = sget(fs_type, sb_test, sb_set, c);
2146	if (IS_ERR(sb)) {
2147		err = PTR_ERR(sb);
2148		kfree(c);
2149		goto out_close;
2150	}
2151
2152	if (sb->s_root) {
2153		struct ubifs_info *c1 = sb->s_fs_info;
2154		kfree(c);
2155		/* A new mount point for already mounted UBIFS */
2156		dbg_gen("this ubi volume is already mounted");
2157		if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2158			err = -EBUSY;
2159			goto out_deact;
2160		}
2161	} else {
2162		sb->s_flags = flags;
2163		err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2164		if (err)
2165			goto out_deact;
2166		/* We do not support atime */
2167		sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2168	}
2169
2170	/* 'fill_super()' opens ubi again so we must close it here */
2171	ubi_close_volume(ubi);
2172
2173	return dget(sb->s_root);
2174
2175out_deact:
2176	deactivate_locked_super(sb);
2177out_close:
2178	ubi_close_volume(ubi);
2179	return ERR_PTR(err);
2180}
2181
2182static void kill_ubifs_super(struct super_block *s)
2183{
2184	struct ubifs_info *c = s->s_fs_info;
2185	kill_anon_super(s);
2186	kfree(c);
2187}
2188
2189static struct file_system_type ubifs_fs_type = {
2190	.name    = "ubifs",
2191	.owner   = THIS_MODULE,
2192	.mount   = ubifs_mount,
2193	.kill_sb = kill_ubifs_super,
2194};
2195
2196/*
2197 * Inode slab cache constructor.
2198 */
2199static void inode_slab_ctor(void *obj)
2200{
2201	struct ubifs_inode *ui = obj;
2202	inode_init_once(&ui->vfs_inode);
2203}
2204
2205static int __init ubifs_init(void)
2206{
2207	int err;
2208
2209	BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2210
2211	/* Make sure node sizes are 8-byte aligned */
2212	BUILD_BUG_ON(UBIFS_CH_SZ        & 7);
2213	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  & 7);
2214	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2215	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2216	BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2217	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2218	BUILD_BUG_ON(UBIFS_SB_NODE_SZ   & 7);
2219	BUILD_BUG_ON(UBIFS_MST_NODE_SZ  & 7);
2220	BUILD_BUG_ON(UBIFS_REF_NODE_SZ  & 7);
2221	BUILD_BUG_ON(UBIFS_CS_NODE_SZ   & 7);
2222	BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2223
2224	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2225	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2226	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2227	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  & 7);
2228	BUILD_BUG_ON(UBIFS_MAX_NODE_SZ      & 7);
2229	BUILD_BUG_ON(MIN_WRITE_SZ           & 7);
2230
2231	/* Check min. node size */
2232	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  < MIN_WRITE_SZ);
2233	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2234	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2235	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2236
2237	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2238	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2239	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2240	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  > UBIFS_MAX_NODE_SZ);
2241
2242	/* Defined node sizes */
2243	BUILD_BUG_ON(UBIFS_SB_NODE_SZ  != 4096);
2244	BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2245	BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2246	BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2247
2248	/*
2249	 * We use 2 bit wide bit-fields to store compression type, which should
2250	 * be amended if more compressors are added. The bit-fields are:
2251	 * @compr_type in 'struct ubifs_inode', @default_compr in
2252	 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2253	 */
2254	BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2255
2256	/*
2257	 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2258	 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2259	 */
2260	if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2261		ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2262			  " at least 4096 bytes",
2263			  (unsigned int)PAGE_CACHE_SIZE);
2264		return -EINVAL;
2265	}
2266
2267	err = register_filesystem(&ubifs_fs_type);
2268	if (err) {
2269		ubifs_err("cannot register file system, error %d", err);
2270		return err;
2271	}
2272
2273	err = -ENOMEM;
2274	ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2275				sizeof(struct ubifs_inode), 0,
2276				SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2277				&inode_slab_ctor);
2278	if (!ubifs_inode_slab)
2279		goto out_reg;
2280
2281	register_shrinker(&ubifs_shrinker_info);
2282
2283	err = ubifs_compressors_init();
2284	if (err)
2285		goto out_shrinker;
2286
2287	err = dbg_debugfs_init();
2288	if (err)
2289		goto out_compr;
2290
 
 
 
 
 
2291	return 0;
2292
 
 
2293out_compr:
2294	ubifs_compressors_exit();
2295out_shrinker:
2296	unregister_shrinker(&ubifs_shrinker_info);
2297	kmem_cache_destroy(ubifs_inode_slab);
2298out_reg:
2299	unregister_filesystem(&ubifs_fs_type);
2300	return err;
2301}
2302/* late_initcall to let compressors initialize first */
2303late_initcall(ubifs_init);
2304
2305static void __exit ubifs_exit(void)
2306{
2307	ubifs_assert(list_empty(&ubifs_infos));
2308	ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2309
2310	dbg_debugfs_exit();
2311	ubifs_compressors_exit();
2312	unregister_shrinker(&ubifs_shrinker_info);
2313	kmem_cache_destroy(ubifs_inode_slab);
2314	unregister_filesystem(&ubifs_fs_type);
2315}
2316module_exit(ubifs_exit);
2317
2318MODULE_LICENSE("GPL");
2319MODULE_VERSION(__stringify(UBIFS_VERSION));
2320MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2321MODULE_DESCRIPTION("UBIFS - UBI File System");