xref: /linux/fs/ubifs/super.c (revision b8bb76713ec50df2f11efee386e16f93d51e1076)
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 */
48 struct kmem_cache *ubifs_inode_slab;
49 
50 /* UBIFS TNC shrinker description */
51 static 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  */
66 static 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 && (inode->i_mode & S_IFMT) != S_IFREG)
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_size(c, inode);
98 	return err;
99 }
100 
101 struct 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 
247 out_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;
252 out_ino:
253 	kfree(ino);
254 out:
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 
260 static 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 
275 static void ubifs_destroy_inode(struct inode *inode)
276 {
277 	struct ubifs_inode *ui = ubifs_inode(inode);
278 
279 	kfree(ui->data);
280 	kmem_cache_free(ubifs_inode_slab, inode);
281 }
282 
283 /*
284  * Note, Linux write-back code calls this without 'i_mutex'.
285  */
286 static int ubifs_write_inode(struct inode *inode, int wait)
287 {
288 	int err = 0;
289 	struct ubifs_info *c = inode->i_sb->s_fs_info;
290 	struct ubifs_inode *ui = ubifs_inode(inode);
291 
292 	ubifs_assert(!ui->xattr);
293 	if (is_bad_inode(inode))
294 		return 0;
295 
296 	mutex_lock(&ui->ui_mutex);
297 	/*
298 	 * Due to races between write-back forced by budgeting
299 	 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
300 	 * have already been synchronized, do not do this again. This might
301 	 * also happen if it was synchronized in an VFS operation, e.g.
302 	 * 'ubifs_link()'.
303 	 */
304 	if (!ui->dirty) {
305 		mutex_unlock(&ui->ui_mutex);
306 		return 0;
307 	}
308 
309 	/*
310 	 * As an optimization, do not write orphan inodes to the media just
311 	 * because this is not needed.
312 	 */
313 	dbg_gen("inode %lu, mode %#x, nlink %u",
314 		inode->i_ino, (int)inode->i_mode, inode->i_nlink);
315 	if (inode->i_nlink) {
316 		err = ubifs_jnl_write_inode(c, inode);
317 		if (err)
318 			ubifs_err("can't write inode %lu, error %d",
319 				  inode->i_ino, err);
320 	}
321 
322 	ui->dirty = 0;
323 	mutex_unlock(&ui->ui_mutex);
324 	ubifs_release_dirty_inode_budget(c, ui);
325 	return err;
326 }
327 
328 static void ubifs_delete_inode(struct inode *inode)
329 {
330 	int err;
331 	struct ubifs_info *c = inode->i_sb->s_fs_info;
332 	struct ubifs_inode *ui = ubifs_inode(inode);
333 
334 	if (ui->xattr)
335 		/*
336 		 * Extended attribute inode deletions are fully handled in
337 		 * 'ubifs_removexattr()'. These inodes are special and have
338 		 * limited usage, so there is nothing to do here.
339 		 */
340 		goto out;
341 
342 	dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
343 	ubifs_assert(!atomic_read(&inode->i_count));
344 	ubifs_assert(inode->i_nlink == 0);
345 
346 	truncate_inode_pages(&inode->i_data, 0);
347 	if (is_bad_inode(inode))
348 		goto out;
349 
350 	ui->ui_size = inode->i_size = 0;
351 	err = ubifs_jnl_delete_inode(c, inode);
352 	if (err)
353 		/*
354 		 * Worst case we have a lost orphan inode wasting space, so a
355 		 * simple error message is OK here.
356 		 */
357 		ubifs_err("can't delete inode %lu, error %d",
358 			  inode->i_ino, err);
359 
360 out:
361 	if (ui->dirty)
362 		ubifs_release_dirty_inode_budget(c, ui);
363 	clear_inode(inode);
364 }
365 
366 static void ubifs_dirty_inode(struct inode *inode)
367 {
368 	struct ubifs_inode *ui = ubifs_inode(inode);
369 
370 	ubifs_assert(mutex_is_locked(&ui->ui_mutex));
371 	if (!ui->dirty) {
372 		ui->dirty = 1;
373 		dbg_gen("inode %lu",  inode->i_ino);
374 	}
375 }
376 
377 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
378 {
379 	struct ubifs_info *c = dentry->d_sb->s_fs_info;
380 	unsigned long long free;
381 	__le32 *uuid = (__le32 *)c->uuid;
382 
383 	free = ubifs_get_free_space(c);
384 	dbg_gen("free space %lld bytes (%lld blocks)",
385 		free, free >> UBIFS_BLOCK_SHIFT);
386 
387 	buf->f_type = UBIFS_SUPER_MAGIC;
388 	buf->f_bsize = UBIFS_BLOCK_SIZE;
389 	buf->f_blocks = c->block_cnt;
390 	buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
391 	if (free > c->report_rp_size)
392 		buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
393 	else
394 		buf->f_bavail = 0;
395 	buf->f_files = 0;
396 	buf->f_ffree = 0;
397 	buf->f_namelen = UBIFS_MAX_NLEN;
398 	buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
399 	buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
400 	ubifs_assert(buf->f_bfree <= c->block_cnt);
401 	return 0;
402 }
403 
404 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
405 {
406 	struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
407 
408 	if (c->mount_opts.unmount_mode == 2)
409 		seq_printf(s, ",fast_unmount");
410 	else if (c->mount_opts.unmount_mode == 1)
411 		seq_printf(s, ",norm_unmount");
412 
413 	if (c->mount_opts.bulk_read == 2)
414 		seq_printf(s, ",bulk_read");
415 	else if (c->mount_opts.bulk_read == 1)
416 		seq_printf(s, ",no_bulk_read");
417 
418 	if (c->mount_opts.chk_data_crc == 2)
419 		seq_printf(s, ",chk_data_crc");
420 	else if (c->mount_opts.chk_data_crc == 1)
421 		seq_printf(s, ",no_chk_data_crc");
422 
423 	if (c->mount_opts.override_compr) {
424 		seq_printf(s, ",compr=");
425 		seq_printf(s, ubifs_compr_name(c->mount_opts.compr_type));
426 	}
427 
428 	return 0;
429 }
430 
431 static int ubifs_sync_fs(struct super_block *sb, int wait)
432 {
433 	int i, err;
434 	struct ubifs_info *c = sb->s_fs_info;
435 	struct writeback_control wbc = {
436 		.sync_mode   = WB_SYNC_ALL,
437 		.range_start = 0,
438 		.range_end   = LLONG_MAX,
439 		.nr_to_write = LONG_MAX,
440 	};
441 
442 	/*
443 	 * Zero @wait is just an advisory thing to help the file system shove
444 	 * lots of data into the queues, and there will be the second
445 	 * '->sync_fs()' call, with non-zero @wait.
446 	 */
447 	if (!wait)
448 		return 0;
449 
450 	if (sb->s_flags & MS_RDONLY)
451 		return 0;
452 
453 	/*
454 	 * VFS calls '->sync_fs()' before synchronizing all dirty inodes and
455 	 * pages, so synchronize them first, then commit the journal. Strictly
456 	 * speaking, it is not necessary to commit the journal here,
457 	 * synchronizing write-buffers would be enough. But committing makes
458 	 * UBIFS free space predictions much more accurate, so we want to let
459 	 * the user be able to get more accurate results of 'statfs()' after
460 	 * they synchronize the file system.
461 	 */
462 	generic_sync_sb_inodes(sb, &wbc);
463 
464 	/*
465 	 * Synchronize write buffers, because 'ubifs_run_commit()' does not
466 	 * do this if it waits for an already running commit.
467 	 */
468 	for (i = 0; i < c->jhead_cnt; i++) {
469 		err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
470 		if (err)
471 			return err;
472 	}
473 
474 	err = ubifs_run_commit(c);
475 	if (err)
476 		return err;
477 
478 	return ubi_sync(c->vi.ubi_num);
479 }
480 
481 /**
482  * init_constants_early - initialize UBIFS constants.
483  * @c: UBIFS file-system description object
484  *
485  * This function initialize UBIFS constants which do not need the superblock to
486  * be read. It also checks that the UBI volume satisfies basic UBIFS
487  * requirements. Returns zero in case of success and a negative error code in
488  * case of failure.
489  */
490 static int init_constants_early(struct ubifs_info *c)
491 {
492 	if (c->vi.corrupted) {
493 		ubifs_warn("UBI volume is corrupted - read-only mode");
494 		c->ro_media = 1;
495 	}
496 
497 	if (c->di.ro_mode) {
498 		ubifs_msg("read-only UBI device");
499 		c->ro_media = 1;
500 	}
501 
502 	if (c->vi.vol_type == UBI_STATIC_VOLUME) {
503 		ubifs_msg("static UBI volume - read-only mode");
504 		c->ro_media = 1;
505 	}
506 
507 	c->leb_cnt = c->vi.size;
508 	c->leb_size = c->vi.usable_leb_size;
509 	c->half_leb_size = c->leb_size / 2;
510 	c->min_io_size = c->di.min_io_size;
511 	c->min_io_shift = fls(c->min_io_size) - 1;
512 
513 	if (c->leb_size < UBIFS_MIN_LEB_SZ) {
514 		ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
515 			  c->leb_size, UBIFS_MIN_LEB_SZ);
516 		return -EINVAL;
517 	}
518 
519 	if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
520 		ubifs_err("too few LEBs (%d), min. is %d",
521 			  c->leb_cnt, UBIFS_MIN_LEB_CNT);
522 		return -EINVAL;
523 	}
524 
525 	if (!is_power_of_2(c->min_io_size)) {
526 		ubifs_err("bad min. I/O size %d", c->min_io_size);
527 		return -EINVAL;
528 	}
529 
530 	/*
531 	 * UBIFS aligns all node to 8-byte boundary, so to make function in
532 	 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
533 	 * less than 8.
534 	 */
535 	if (c->min_io_size < 8) {
536 		c->min_io_size = 8;
537 		c->min_io_shift = 3;
538 	}
539 
540 	c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
541 	c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
542 
543 	/*
544 	 * Initialize node length ranges which are mostly needed for node
545 	 * length validation.
546 	 */
547 	c->ranges[UBIFS_PAD_NODE].len  = UBIFS_PAD_NODE_SZ;
548 	c->ranges[UBIFS_SB_NODE].len   = UBIFS_SB_NODE_SZ;
549 	c->ranges[UBIFS_MST_NODE].len  = UBIFS_MST_NODE_SZ;
550 	c->ranges[UBIFS_REF_NODE].len  = UBIFS_REF_NODE_SZ;
551 	c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
552 	c->ranges[UBIFS_CS_NODE].len   = UBIFS_CS_NODE_SZ;
553 
554 	c->ranges[UBIFS_INO_NODE].min_len  = UBIFS_INO_NODE_SZ;
555 	c->ranges[UBIFS_INO_NODE].max_len  = UBIFS_MAX_INO_NODE_SZ;
556 	c->ranges[UBIFS_ORPH_NODE].min_len =
557 				UBIFS_ORPH_NODE_SZ + sizeof(__le64);
558 	c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
559 	c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
560 	c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
561 	c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
562 	c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
563 	c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
564 	c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
565 	/*
566 	 * Minimum indexing node size is amended later when superblock is
567 	 * read and the key length is known.
568 	 */
569 	c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
570 	/*
571 	 * Maximum indexing node size is amended later when superblock is
572 	 * read and the fanout is known.
573 	 */
574 	c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
575 
576 	/*
577 	 * Initialize dead and dark LEB space watermarks. See gc.c for comments
578 	 * about these values.
579 	 */
580 	c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
581 	c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
582 
583 	/*
584 	 * Calculate how many bytes would be wasted at the end of LEB if it was
585 	 * fully filled with data nodes of maximum size. This is used in
586 	 * calculations when reporting free space.
587 	 */
588 	c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
589 
590 	/* Buffer size for bulk-reads */
591 	c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
592 	if (c->max_bu_buf_len > c->leb_size)
593 		c->max_bu_buf_len = c->leb_size;
594 	return 0;
595 }
596 
597 /**
598  * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
599  * @c: UBIFS file-system description object
600  * @lnum: LEB the write-buffer was synchronized to
601  * @free: how many free bytes left in this LEB
602  * @pad: how many bytes were padded
603  *
604  * This is a callback function which is called by the I/O unit when the
605  * write-buffer is synchronized. We need this to correctly maintain space
606  * accounting in bud logical eraseblocks. This function returns zero in case of
607  * success and a negative error code in case of failure.
608  *
609  * This function actually belongs to the journal, but we keep it here because
610  * we want to keep it static.
611  */
612 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
613 {
614 	return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
615 }
616 
617 /*
618  * init_constants_sb - initialize UBIFS constants.
619  * @c: UBIFS file-system description object
620  *
621  * This is a helper function which initializes various UBIFS constants after
622  * the superblock has been read. It also checks various UBIFS parameters and
623  * makes sure they are all right. Returns zero in case of success and a
624  * negative error code in case of failure.
625  */
626 static int init_constants_sb(struct ubifs_info *c)
627 {
628 	int tmp, err;
629 	long long tmp64;
630 
631 	c->main_bytes = (long long)c->main_lebs * c->leb_size;
632 	c->max_znode_sz = sizeof(struct ubifs_znode) +
633 				c->fanout * sizeof(struct ubifs_zbranch);
634 
635 	tmp = ubifs_idx_node_sz(c, 1);
636 	c->ranges[UBIFS_IDX_NODE].min_len = tmp;
637 	c->min_idx_node_sz = ALIGN(tmp, 8);
638 
639 	tmp = ubifs_idx_node_sz(c, c->fanout);
640 	c->ranges[UBIFS_IDX_NODE].max_len = tmp;
641 	c->max_idx_node_sz = ALIGN(tmp, 8);
642 
643 	/* Make sure LEB size is large enough to fit full commit */
644 	tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
645 	tmp = ALIGN(tmp, c->min_io_size);
646 	if (tmp > c->leb_size) {
647 		dbg_err("too small LEB size %d, at least %d needed",
648 			c->leb_size, tmp);
649 		return -EINVAL;
650 	}
651 
652 	/*
653 	 * Make sure that the log is large enough to fit reference nodes for
654 	 * all buds plus one reserved LEB.
655 	 */
656 	tmp64 = c->max_bud_bytes + c->leb_size - 1;
657 	c->max_bud_cnt = div_u64(tmp64, c->leb_size);
658 	tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
659 	tmp /= c->leb_size;
660 	tmp += 1;
661 	if (c->log_lebs < tmp) {
662 		dbg_err("too small log %d LEBs, required min. %d LEBs",
663 			c->log_lebs, tmp);
664 		return -EINVAL;
665 	}
666 
667 	/*
668 	 * When budgeting we assume worst-case scenarios when the pages are not
669 	 * be compressed and direntries are of the maximum size.
670 	 *
671 	 * Note, data, which may be stored in inodes is budgeted separately, so
672 	 * it is not included into 'c->inode_budget'.
673 	 */
674 	c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
675 	c->inode_budget = UBIFS_INO_NODE_SZ;
676 	c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
677 
678 	/*
679 	 * When the amount of flash space used by buds becomes
680 	 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
681 	 * The writers are unblocked when the commit is finished. To avoid
682 	 * writers to be blocked UBIFS initiates background commit in advance,
683 	 * when number of bud bytes becomes above the limit defined below.
684 	 */
685 	c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
686 
687 	/*
688 	 * Ensure minimum journal size. All the bytes in the journal heads are
689 	 * considered to be used, when calculating the current journal usage.
690 	 * Consequently, if the journal is too small, UBIFS will treat it as
691 	 * always full.
692 	 */
693 	tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
694 	if (c->bg_bud_bytes < tmp64)
695 		c->bg_bud_bytes = tmp64;
696 	if (c->max_bud_bytes < tmp64 + c->leb_size)
697 		c->max_bud_bytes = tmp64 + c->leb_size;
698 
699 	err = ubifs_calc_lpt_geom(c);
700 	if (err)
701 		return err;
702 
703 	return 0;
704 }
705 
706 /*
707  * init_constants_master - initialize UBIFS constants.
708  * @c: UBIFS file-system description object
709  *
710  * This is a helper function which initializes various UBIFS constants after
711  * the master node has been read. It also checks various UBIFS parameters and
712  * makes sure they are all right.
713  */
714 static void init_constants_master(struct ubifs_info *c)
715 {
716 	long long tmp64;
717 
718 	c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
719 
720 	/*
721 	 * Calculate total amount of FS blocks. This number is not used
722 	 * internally because it does not make much sense for UBIFS, but it is
723 	 * necessary to report something for the 'statfs()' call.
724 	 *
725 	 * Subtract the LEB reserved for GC, the LEB which is reserved for
726 	 * deletions, minimum LEBs for the index, and assume only one journal
727 	 * head is available.
728 	 */
729 	tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
730 	tmp64 *= (long long)c->leb_size - c->leb_overhead;
731 	tmp64 = ubifs_reported_space(c, tmp64);
732 	c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
733 }
734 
735 /**
736  * take_gc_lnum - reserve GC LEB.
737  * @c: UBIFS file-system description object
738  *
739  * This function ensures that the LEB reserved for garbage collection is marked
740  * as "taken" in lprops. We also have to set free space to LEB size and dirty
741  * space to zero, because lprops may contain out-of-date information if the
742  * file-system was un-mounted before it has been committed. This function
743  * returns zero in case of success and a negative error code in case of
744  * failure.
745  */
746 static int take_gc_lnum(struct ubifs_info *c)
747 {
748 	int err;
749 
750 	if (c->gc_lnum == -1) {
751 		ubifs_err("no LEB for GC");
752 		return -EINVAL;
753 	}
754 
755 	/* And we have to tell lprops that this LEB is taken */
756 	err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
757 				  LPROPS_TAKEN, 0, 0);
758 	return err;
759 }
760 
761 /**
762  * alloc_wbufs - allocate write-buffers.
763  * @c: UBIFS file-system description object
764  *
765  * This helper function allocates and initializes UBIFS write-buffers. Returns
766  * zero in case of success and %-ENOMEM in case of failure.
767  */
768 static int alloc_wbufs(struct ubifs_info *c)
769 {
770 	int i, err;
771 
772 	c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
773 			   GFP_KERNEL);
774 	if (!c->jheads)
775 		return -ENOMEM;
776 
777 	/* Initialize journal heads */
778 	for (i = 0; i < c->jhead_cnt; i++) {
779 		INIT_LIST_HEAD(&c->jheads[i].buds_list);
780 		err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
781 		if (err)
782 			return err;
783 
784 		c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
785 		c->jheads[i].wbuf.jhead = i;
786 	}
787 
788 	c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
789 	/*
790 	 * Garbage Collector head likely contains long-term data and
791 	 * does not need to be synchronized by timer.
792 	 */
793 	c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
794 	c->jheads[GCHD].wbuf.timeout = 0;
795 
796 	return 0;
797 }
798 
799 /**
800  * free_wbufs - free write-buffers.
801  * @c: UBIFS file-system description object
802  */
803 static void free_wbufs(struct ubifs_info *c)
804 {
805 	int i;
806 
807 	if (c->jheads) {
808 		for (i = 0; i < c->jhead_cnt; i++) {
809 			kfree(c->jheads[i].wbuf.buf);
810 			kfree(c->jheads[i].wbuf.inodes);
811 		}
812 		kfree(c->jheads);
813 		c->jheads = NULL;
814 	}
815 }
816 
817 /**
818  * free_orphans - free orphans.
819  * @c: UBIFS file-system description object
820  */
821 static void free_orphans(struct ubifs_info *c)
822 {
823 	struct ubifs_orphan *orph;
824 
825 	while (c->orph_dnext) {
826 		orph = c->orph_dnext;
827 		c->orph_dnext = orph->dnext;
828 		list_del(&orph->list);
829 		kfree(orph);
830 	}
831 
832 	while (!list_empty(&c->orph_list)) {
833 		orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
834 		list_del(&orph->list);
835 		kfree(orph);
836 		dbg_err("orphan list not empty at unmount");
837 	}
838 
839 	vfree(c->orph_buf);
840 	c->orph_buf = NULL;
841 }
842 
843 /**
844  * free_buds - free per-bud objects.
845  * @c: UBIFS file-system description object
846  */
847 static void free_buds(struct ubifs_info *c)
848 {
849 	struct rb_node *this = c->buds.rb_node;
850 	struct ubifs_bud *bud;
851 
852 	while (this) {
853 		if (this->rb_left)
854 			this = this->rb_left;
855 		else if (this->rb_right)
856 			this = this->rb_right;
857 		else {
858 			bud = rb_entry(this, struct ubifs_bud, rb);
859 			this = rb_parent(this);
860 			if (this) {
861 				if (this->rb_left == &bud->rb)
862 					this->rb_left = NULL;
863 				else
864 					this->rb_right = NULL;
865 			}
866 			kfree(bud);
867 		}
868 	}
869 }
870 
871 /**
872  * check_volume_empty - check if the UBI volume is empty.
873  * @c: UBIFS file-system description object
874  *
875  * This function checks if the UBIFS volume is empty by looking if its LEBs are
876  * mapped or not. The result of checking is stored in the @c->empty variable.
877  * Returns zero in case of success and a negative error code in case of
878  * failure.
879  */
880 static int check_volume_empty(struct ubifs_info *c)
881 {
882 	int lnum, err;
883 
884 	c->empty = 1;
885 	for (lnum = 0; lnum < c->leb_cnt; lnum++) {
886 		err = ubi_is_mapped(c->ubi, lnum);
887 		if (unlikely(err < 0))
888 			return err;
889 		if (err == 1) {
890 			c->empty = 0;
891 			break;
892 		}
893 
894 		cond_resched();
895 	}
896 
897 	return 0;
898 }
899 
900 /*
901  * UBIFS mount options.
902  *
903  * Opt_fast_unmount: do not run a journal commit before un-mounting
904  * Opt_norm_unmount: run a journal commit before un-mounting
905  * Opt_bulk_read: enable bulk-reads
906  * Opt_no_bulk_read: disable bulk-reads
907  * Opt_chk_data_crc: check CRCs when reading data nodes
908  * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
909  * Opt_override_compr: override default compressor
910  * Opt_err: just end of array marker
911  */
912 enum {
913 	Opt_fast_unmount,
914 	Opt_norm_unmount,
915 	Opt_bulk_read,
916 	Opt_no_bulk_read,
917 	Opt_chk_data_crc,
918 	Opt_no_chk_data_crc,
919 	Opt_override_compr,
920 	Opt_err,
921 };
922 
923 static const match_table_t tokens = {
924 	{Opt_fast_unmount, "fast_unmount"},
925 	{Opt_norm_unmount, "norm_unmount"},
926 	{Opt_bulk_read, "bulk_read"},
927 	{Opt_no_bulk_read, "no_bulk_read"},
928 	{Opt_chk_data_crc, "chk_data_crc"},
929 	{Opt_no_chk_data_crc, "no_chk_data_crc"},
930 	{Opt_override_compr, "compr=%s"},
931 	{Opt_err, NULL},
932 };
933 
934 /**
935  * ubifs_parse_options - parse mount parameters.
936  * @c: UBIFS file-system description object
937  * @options: parameters to parse
938  * @is_remount: non-zero if this is FS re-mount
939  *
940  * This function parses UBIFS mount options and returns zero in case success
941  * and a negative error code in case of failure.
942  */
943 static int ubifs_parse_options(struct ubifs_info *c, char *options,
944 			       int is_remount)
945 {
946 	char *p;
947 	substring_t args[MAX_OPT_ARGS];
948 
949 	if (!options)
950 		return 0;
951 
952 	while ((p = strsep(&options, ","))) {
953 		int token;
954 
955 		if (!*p)
956 			continue;
957 
958 		token = match_token(p, tokens, args);
959 		switch (token) {
960 		/*
961 		 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
962 		 * We accepte them in order to be backware-compatible. But this
963 		 * should be removed at some point.
964 		 */
965 		case Opt_fast_unmount:
966 			c->mount_opts.unmount_mode = 2;
967 			break;
968 		case Opt_norm_unmount:
969 			c->mount_opts.unmount_mode = 1;
970 			break;
971 		case Opt_bulk_read:
972 			c->mount_opts.bulk_read = 2;
973 			c->bulk_read = 1;
974 			break;
975 		case Opt_no_bulk_read:
976 			c->mount_opts.bulk_read = 1;
977 			c->bulk_read = 0;
978 			break;
979 		case Opt_chk_data_crc:
980 			c->mount_opts.chk_data_crc = 2;
981 			c->no_chk_data_crc = 0;
982 			break;
983 		case Opt_no_chk_data_crc:
984 			c->mount_opts.chk_data_crc = 1;
985 			c->no_chk_data_crc = 1;
986 			break;
987 		case Opt_override_compr:
988 		{
989 			char *name = match_strdup(&args[0]);
990 
991 			if (!name)
992 				return -ENOMEM;
993 			if (!strcmp(name, "none"))
994 				c->mount_opts.compr_type = UBIFS_COMPR_NONE;
995 			else if (!strcmp(name, "lzo"))
996 				c->mount_opts.compr_type = UBIFS_COMPR_LZO;
997 			else if (!strcmp(name, "zlib"))
998 				c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
999 			else {
1000 				ubifs_err("unknown compressor \"%s\"", name);
1001 				kfree(name);
1002 				return -EINVAL;
1003 			}
1004 			kfree(name);
1005 			c->mount_opts.override_compr = 1;
1006 			c->default_compr = c->mount_opts.compr_type;
1007 			break;
1008 		}
1009 		default:
1010 			ubifs_err("unrecognized mount option \"%s\" "
1011 				  "or missing value", p);
1012 			return -EINVAL;
1013 		}
1014 	}
1015 
1016 	return 0;
1017 }
1018 
1019 /**
1020  * destroy_journal - destroy journal data structures.
1021  * @c: UBIFS file-system description object
1022  *
1023  * This function destroys journal data structures including those that may have
1024  * been created by recovery functions.
1025  */
1026 static void destroy_journal(struct ubifs_info *c)
1027 {
1028 	while (!list_empty(&c->unclean_leb_list)) {
1029 		struct ubifs_unclean_leb *ucleb;
1030 
1031 		ucleb = list_entry(c->unclean_leb_list.next,
1032 				   struct ubifs_unclean_leb, list);
1033 		list_del(&ucleb->list);
1034 		kfree(ucleb);
1035 	}
1036 	while (!list_empty(&c->old_buds)) {
1037 		struct ubifs_bud *bud;
1038 
1039 		bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1040 		list_del(&bud->list);
1041 		kfree(bud);
1042 	}
1043 	ubifs_destroy_idx_gc(c);
1044 	ubifs_destroy_size_tree(c);
1045 	ubifs_tnc_close(c);
1046 	free_buds(c);
1047 }
1048 
1049 /**
1050  * bu_init - initialize bulk-read information.
1051  * @c: UBIFS file-system description object
1052  */
1053 static void bu_init(struct ubifs_info *c)
1054 {
1055 	ubifs_assert(c->bulk_read == 1);
1056 
1057 	if (c->bu.buf)
1058 		return; /* Already initialized */
1059 
1060 again:
1061 	c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1062 	if (!c->bu.buf) {
1063 		if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1064 			c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1065 			goto again;
1066 		}
1067 
1068 		/* Just disable bulk-read */
1069 		ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1070 			   "disabling it", c->max_bu_buf_len);
1071 		c->mount_opts.bulk_read = 1;
1072 		c->bulk_read = 0;
1073 		return;
1074 	}
1075 }
1076 
1077 /**
1078  * check_free_space - check if there is enough free space to mount.
1079  * @c: UBIFS file-system description object
1080  *
1081  * This function makes sure UBIFS has enough free space to be mounted in
1082  * read/write mode. UBIFS must always have some free space to allow deletions.
1083  */
1084 static int check_free_space(struct ubifs_info *c)
1085 {
1086 	ubifs_assert(c->dark_wm > 0);
1087 	if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1088 		ubifs_err("insufficient free space to mount in read/write mode");
1089 		dbg_dump_budg(c);
1090 		dbg_dump_lprops(c);
1091 		return -ENOSPC;
1092 	}
1093 	return 0;
1094 }
1095 
1096 /**
1097  * mount_ubifs - mount UBIFS file-system.
1098  * @c: UBIFS file-system description object
1099  *
1100  * This function mounts UBIFS file system. Returns zero in case of success and
1101  * a negative error code in case of failure.
1102  *
1103  * Note, the function does not de-allocate resources it it fails half way
1104  * through, and the caller has to do this instead.
1105  */
1106 static int mount_ubifs(struct ubifs_info *c)
1107 {
1108 	struct super_block *sb = c->vfs_sb;
1109 	int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1110 	long long x;
1111 	size_t sz;
1112 
1113 	err = init_constants_early(c);
1114 	if (err)
1115 		return err;
1116 
1117 	err = ubifs_debugging_init(c);
1118 	if (err)
1119 		return err;
1120 
1121 	err = check_volume_empty(c);
1122 	if (err)
1123 		goto out_free;
1124 
1125 	if (c->empty && (mounted_read_only || c->ro_media)) {
1126 		/*
1127 		 * This UBI volume is empty, and read-only, or the file system
1128 		 * is mounted read-only - we cannot format it.
1129 		 */
1130 		ubifs_err("can't format empty UBI volume: read-only %s",
1131 			  c->ro_media ? "UBI volume" : "mount");
1132 		err = -EROFS;
1133 		goto out_free;
1134 	}
1135 
1136 	if (c->ro_media && !mounted_read_only) {
1137 		ubifs_err("cannot mount read-write - read-only media");
1138 		err = -EROFS;
1139 		goto out_free;
1140 	}
1141 
1142 	/*
1143 	 * The requirement for the buffer is that it should fit indexing B-tree
1144 	 * height amount of integers. We assume the height if the TNC tree will
1145 	 * never exceed 64.
1146 	 */
1147 	err = -ENOMEM;
1148 	c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1149 	if (!c->bottom_up_buf)
1150 		goto out_free;
1151 
1152 	c->sbuf = vmalloc(c->leb_size);
1153 	if (!c->sbuf)
1154 		goto out_free;
1155 
1156 	if (!mounted_read_only) {
1157 		c->ileb_buf = vmalloc(c->leb_size);
1158 		if (!c->ileb_buf)
1159 			goto out_free;
1160 	}
1161 
1162 	if (c->bulk_read == 1)
1163 		bu_init(c);
1164 
1165 	/*
1166 	 * We have to check all CRCs, even for data nodes, when we mount the FS
1167 	 * (specifically, when we are replaying).
1168 	 */
1169 	c->always_chk_crc = 1;
1170 
1171 	err = ubifs_read_superblock(c);
1172 	if (err)
1173 		goto out_free;
1174 
1175 	/*
1176 	 * Make sure the compressor which is set as default in the superblock
1177 	 * or overridden by mount options is actually compiled in.
1178 	 */
1179 	if (!ubifs_compr_present(c->default_compr)) {
1180 		ubifs_err("'compressor \"%s\" is not compiled in",
1181 			  ubifs_compr_name(c->default_compr));
1182 		goto out_free;
1183 	}
1184 
1185 	err = init_constants_sb(c);
1186 	if (err)
1187 		goto out_free;
1188 
1189 	sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1190 	sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1191 	c->cbuf = kmalloc(sz, GFP_NOFS);
1192 	if (!c->cbuf) {
1193 		err = -ENOMEM;
1194 		goto out_free;
1195 	}
1196 
1197 	sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1198 	if (!mounted_read_only) {
1199 		err = alloc_wbufs(c);
1200 		if (err)
1201 			goto out_cbuf;
1202 
1203 		/* Create background thread */
1204 		c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1205 		if (IS_ERR(c->bgt)) {
1206 			err = PTR_ERR(c->bgt);
1207 			c->bgt = NULL;
1208 			ubifs_err("cannot spawn \"%s\", error %d",
1209 				  c->bgt_name, err);
1210 			goto out_wbufs;
1211 		}
1212 		wake_up_process(c->bgt);
1213 	}
1214 
1215 	err = ubifs_read_master(c);
1216 	if (err)
1217 		goto out_master;
1218 
1219 	init_constants_master(c);
1220 
1221 	if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1222 		ubifs_msg("recovery needed");
1223 		c->need_recovery = 1;
1224 		if (!mounted_read_only) {
1225 			err = ubifs_recover_inl_heads(c, c->sbuf);
1226 			if (err)
1227 				goto out_master;
1228 		}
1229 	} else if (!mounted_read_only) {
1230 		/*
1231 		 * Set the "dirty" flag so that if we reboot uncleanly we
1232 		 * will notice this immediately on the next mount.
1233 		 */
1234 		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1235 		err = ubifs_write_master(c);
1236 		if (err)
1237 			goto out_master;
1238 	}
1239 
1240 	err = ubifs_lpt_init(c, 1, !mounted_read_only);
1241 	if (err)
1242 		goto out_lpt;
1243 
1244 	err = dbg_check_idx_size(c, c->old_idx_sz);
1245 	if (err)
1246 		goto out_lpt;
1247 
1248 	err = ubifs_replay_journal(c);
1249 	if (err)
1250 		goto out_journal;
1251 
1252 	err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1253 	if (err)
1254 		goto out_orphans;
1255 
1256 	if (!mounted_read_only) {
1257 		int lnum;
1258 
1259 		err = check_free_space(c);
1260 		if (err)
1261 			goto out_orphans;
1262 
1263 		/* Check for enough log space */
1264 		lnum = c->lhead_lnum + 1;
1265 		if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1266 			lnum = UBIFS_LOG_LNUM;
1267 		if (lnum == c->ltail_lnum) {
1268 			err = ubifs_consolidate_log(c);
1269 			if (err)
1270 				goto out_orphans;
1271 		}
1272 
1273 		if (c->need_recovery) {
1274 			err = ubifs_recover_size(c);
1275 			if (err)
1276 				goto out_orphans;
1277 			err = ubifs_rcvry_gc_commit(c);
1278 		} else {
1279 			err = take_gc_lnum(c);
1280 			if (err)
1281 				goto out_orphans;
1282 
1283 			/*
1284 			 * GC LEB may contain garbage if there was an unclean
1285 			 * reboot, and it should be un-mapped.
1286 			 */
1287 			err = ubifs_leb_unmap(c, c->gc_lnum);
1288 			if (err)
1289 				return err;
1290 		}
1291 
1292 		err = dbg_check_lprops(c);
1293 		if (err)
1294 			goto out_orphans;
1295 	} else if (c->need_recovery) {
1296 		err = ubifs_recover_size(c);
1297 		if (err)
1298 			goto out_orphans;
1299 	} else {
1300 		/*
1301 		 * Even if we mount read-only, we have to set space in GC LEB
1302 		 * to proper value because this affects UBIFS free space
1303 		 * reporting. We do not want to have a situation when
1304 		 * re-mounting from R/O to R/W changes amount of free space.
1305 		 */
1306 		err = take_gc_lnum(c);
1307 		if (err)
1308 			goto out_orphans;
1309 	}
1310 
1311 	spin_lock(&ubifs_infos_lock);
1312 	list_add_tail(&c->infos_list, &ubifs_infos);
1313 	spin_unlock(&ubifs_infos_lock);
1314 
1315 	if (c->need_recovery) {
1316 		if (mounted_read_only)
1317 			ubifs_msg("recovery deferred");
1318 		else {
1319 			c->need_recovery = 0;
1320 			ubifs_msg("recovery completed");
1321 			/* GC LEB has to be empty and taken at this point */
1322 			ubifs_assert(c->lst.taken_empty_lebs == 1);
1323 		}
1324 	} else
1325 		ubifs_assert(c->lst.taken_empty_lebs == 1);
1326 
1327 	err = dbg_check_filesystem(c);
1328 	if (err)
1329 		goto out_infos;
1330 
1331 	err = dbg_debugfs_init_fs(c);
1332 	if (err)
1333 		goto out_infos;
1334 
1335 	c->always_chk_crc = 0;
1336 
1337 	ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1338 		  c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1339 	if (mounted_read_only)
1340 		ubifs_msg("mounted read-only");
1341 	x = (long long)c->main_lebs * c->leb_size;
1342 	ubifs_msg("file system size:   %lld bytes (%lld KiB, %lld MiB, %d "
1343 		  "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1344 	x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1345 	ubifs_msg("journal size:       %lld bytes (%lld KiB, %lld MiB, %d "
1346 		  "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1347 	ubifs_msg("media format:       %d (latest is %d)",
1348 		  c->fmt_version, UBIFS_FORMAT_VERSION);
1349 	ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1350 	ubifs_msg("reserved for root:  %llu bytes (%llu KiB)",
1351 		c->report_rp_size, c->report_rp_size >> 10);
1352 
1353 	dbg_msg("compiled on:         " __DATE__ " at " __TIME__);
1354 	dbg_msg("min. I/O unit size:  %d bytes", c->min_io_size);
1355 	dbg_msg("LEB size:            %d bytes (%d KiB)",
1356 		c->leb_size, c->leb_size >> 10);
1357 	dbg_msg("data journal heads:  %d",
1358 		c->jhead_cnt - NONDATA_JHEADS_CNT);
1359 	dbg_msg("UUID:                %02X%02X%02X%02X-%02X%02X"
1360 	       "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1361 	       c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1362 	       c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1363 	       c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1364 	       c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1365 	dbg_msg("big_lpt              %d", c->big_lpt);
1366 	dbg_msg("log LEBs:            %d (%d - %d)",
1367 		c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1368 	dbg_msg("LPT area LEBs:       %d (%d - %d)",
1369 		c->lpt_lebs, c->lpt_first, c->lpt_last);
1370 	dbg_msg("orphan area LEBs:    %d (%d - %d)",
1371 		c->orph_lebs, c->orph_first, c->orph_last);
1372 	dbg_msg("main area LEBs:      %d (%d - %d)",
1373 		c->main_lebs, c->main_first, c->leb_cnt - 1);
1374 	dbg_msg("index LEBs:          %d", c->lst.idx_lebs);
1375 	dbg_msg("total index bytes:   %lld (%lld KiB, %lld MiB)",
1376 		c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1377 	dbg_msg("key hash type:       %d", c->key_hash_type);
1378 	dbg_msg("tree fanout:         %d", c->fanout);
1379 	dbg_msg("reserved GC LEB:     %d", c->gc_lnum);
1380 	dbg_msg("first main LEB:      %d", c->main_first);
1381 	dbg_msg("max. znode size      %d", c->max_znode_sz);
1382 	dbg_msg("max. index node size %d", c->max_idx_node_sz);
1383 	dbg_msg("node sizes:          data %zu, inode %zu, dentry %zu",
1384 		UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1385 	dbg_msg("node sizes:          trun %zu, sb %zu, master %zu",
1386 		UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1387 	dbg_msg("node sizes:          ref %zu, cmt. start %zu, orph %zu",
1388 		UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1389 	dbg_msg("max. node sizes:     data %zu, inode %zu dentry %zu",
1390 	        UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1391 		UBIFS_MAX_DENT_NODE_SZ);
1392 	dbg_msg("dead watermark:      %d", c->dead_wm);
1393 	dbg_msg("dark watermark:      %d", c->dark_wm);
1394 	dbg_msg("LEB overhead:        %d", c->leb_overhead);
1395 	x = (long long)c->main_lebs * c->dark_wm;
1396 	dbg_msg("max. dark space:     %lld (%lld KiB, %lld MiB)",
1397 		x, x >> 10, x >> 20);
1398 	dbg_msg("maximum bud bytes:   %lld (%lld KiB, %lld MiB)",
1399 		c->max_bud_bytes, c->max_bud_bytes >> 10,
1400 		c->max_bud_bytes >> 20);
1401 	dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1402 		c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1403 		c->bg_bud_bytes >> 20);
1404 	dbg_msg("current bud bytes    %lld (%lld KiB, %lld MiB)",
1405 		c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1406 	dbg_msg("max. seq. number:    %llu", c->max_sqnum);
1407 	dbg_msg("commit number:       %llu", c->cmt_no);
1408 
1409 	return 0;
1410 
1411 out_infos:
1412 	spin_lock(&ubifs_infos_lock);
1413 	list_del(&c->infos_list);
1414 	spin_unlock(&ubifs_infos_lock);
1415 out_orphans:
1416 	free_orphans(c);
1417 out_journal:
1418 	destroy_journal(c);
1419 out_lpt:
1420 	ubifs_lpt_free(c, 0);
1421 out_master:
1422 	kfree(c->mst_node);
1423 	kfree(c->rcvrd_mst_node);
1424 	if (c->bgt)
1425 		kthread_stop(c->bgt);
1426 out_wbufs:
1427 	free_wbufs(c);
1428 out_cbuf:
1429 	kfree(c->cbuf);
1430 out_free:
1431 	kfree(c->bu.buf);
1432 	vfree(c->ileb_buf);
1433 	vfree(c->sbuf);
1434 	kfree(c->bottom_up_buf);
1435 	ubifs_debugging_exit(c);
1436 	return err;
1437 }
1438 
1439 /**
1440  * ubifs_umount - un-mount UBIFS file-system.
1441  * @c: UBIFS file-system description object
1442  *
1443  * Note, this function is called to free allocated resourced when un-mounting,
1444  * as well as free resources when an error occurred while we were half way
1445  * through mounting (error path cleanup function). So it has to make sure the
1446  * resource was actually allocated before freeing it.
1447  */
1448 static void ubifs_umount(struct ubifs_info *c)
1449 {
1450 	dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1451 		c->vi.vol_id);
1452 
1453 	dbg_debugfs_exit_fs(c);
1454 	spin_lock(&ubifs_infos_lock);
1455 	list_del(&c->infos_list);
1456 	spin_unlock(&ubifs_infos_lock);
1457 
1458 	if (c->bgt)
1459 		kthread_stop(c->bgt);
1460 
1461 	destroy_journal(c);
1462 	free_wbufs(c);
1463 	free_orphans(c);
1464 	ubifs_lpt_free(c, 0);
1465 
1466 	kfree(c->cbuf);
1467 	kfree(c->rcvrd_mst_node);
1468 	kfree(c->mst_node);
1469 	kfree(c->bu.buf);
1470 	vfree(c->ileb_buf);
1471 	vfree(c->sbuf);
1472 	kfree(c->bottom_up_buf);
1473 	ubifs_debugging_exit(c);
1474 }
1475 
1476 /**
1477  * ubifs_remount_rw - re-mount in read-write mode.
1478  * @c: UBIFS file-system description object
1479  *
1480  * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1481  * mode. This function allocates the needed resources and re-mounts UBIFS in
1482  * read-write mode.
1483  */
1484 static int ubifs_remount_rw(struct ubifs_info *c)
1485 {
1486 	int err, lnum;
1487 
1488 	mutex_lock(&c->umount_mutex);
1489 	dbg_save_space_info(c);
1490 	c->remounting_rw = 1;
1491 	c->always_chk_crc = 1;
1492 
1493 	err = check_free_space(c);
1494 	if (err)
1495 		goto out;
1496 
1497 	if (c->old_leb_cnt != c->leb_cnt) {
1498 		struct ubifs_sb_node *sup;
1499 
1500 		sup = ubifs_read_sb_node(c);
1501 		if (IS_ERR(sup)) {
1502 			err = PTR_ERR(sup);
1503 			goto out;
1504 		}
1505 		sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1506 		err = ubifs_write_sb_node(c, sup);
1507 		if (err)
1508 			goto out;
1509 	}
1510 
1511 	if (c->need_recovery) {
1512 		ubifs_msg("completing deferred recovery");
1513 		err = ubifs_write_rcvrd_mst_node(c);
1514 		if (err)
1515 			goto out;
1516 		err = ubifs_recover_size(c);
1517 		if (err)
1518 			goto out;
1519 		err = ubifs_clean_lebs(c, c->sbuf);
1520 		if (err)
1521 			goto out;
1522 		err = ubifs_recover_inl_heads(c, c->sbuf);
1523 		if (err)
1524 			goto out;
1525 	} else {
1526 		/* A readonly mount is not allowed to have orphans */
1527 		ubifs_assert(c->tot_orphans == 0);
1528 		err = ubifs_clear_orphans(c);
1529 		if (err)
1530 			goto out;
1531 	}
1532 
1533 	if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1534 		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1535 		err = ubifs_write_master(c);
1536 		if (err)
1537 			goto out;
1538 	}
1539 
1540 	c->ileb_buf = vmalloc(c->leb_size);
1541 	if (!c->ileb_buf) {
1542 		err = -ENOMEM;
1543 		goto out;
1544 	}
1545 
1546 	err = ubifs_lpt_init(c, 0, 1);
1547 	if (err)
1548 		goto out;
1549 
1550 	err = alloc_wbufs(c);
1551 	if (err)
1552 		goto out;
1553 
1554 	ubifs_create_buds_lists(c);
1555 
1556 	/* Create background thread */
1557 	c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1558 	if (IS_ERR(c->bgt)) {
1559 		err = PTR_ERR(c->bgt);
1560 		c->bgt = NULL;
1561 		ubifs_err("cannot spawn \"%s\", error %d",
1562 			  c->bgt_name, err);
1563 		goto out;
1564 	}
1565 	wake_up_process(c->bgt);
1566 
1567 	c->orph_buf = vmalloc(c->leb_size);
1568 	if (!c->orph_buf) {
1569 		err = -ENOMEM;
1570 		goto out;
1571 	}
1572 
1573 	/* Check for enough log space */
1574 	lnum = c->lhead_lnum + 1;
1575 	if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1576 		lnum = UBIFS_LOG_LNUM;
1577 	if (lnum == c->ltail_lnum) {
1578 		err = ubifs_consolidate_log(c);
1579 		if (err)
1580 			goto out;
1581 	}
1582 
1583 	if (c->need_recovery)
1584 		err = ubifs_rcvry_gc_commit(c);
1585 	else
1586 		err = ubifs_leb_unmap(c, c->gc_lnum);
1587 	if (err)
1588 		goto out;
1589 
1590 	if (c->need_recovery) {
1591 		c->need_recovery = 0;
1592 		ubifs_msg("deferred recovery completed");
1593 	}
1594 
1595 	dbg_gen("re-mounted read-write");
1596 	c->vfs_sb->s_flags &= ~MS_RDONLY;
1597 	c->remounting_rw = 0;
1598 	c->always_chk_crc = 0;
1599 	err = dbg_check_space_info(c);
1600 	mutex_unlock(&c->umount_mutex);
1601 	return err;
1602 
1603 out:
1604 	vfree(c->orph_buf);
1605 	c->orph_buf = NULL;
1606 	if (c->bgt) {
1607 		kthread_stop(c->bgt);
1608 		c->bgt = NULL;
1609 	}
1610 	free_wbufs(c);
1611 	vfree(c->ileb_buf);
1612 	c->ileb_buf = NULL;
1613 	ubifs_lpt_free(c, 1);
1614 	c->remounting_rw = 0;
1615 	c->always_chk_crc = 0;
1616 	mutex_unlock(&c->umount_mutex);
1617 	return err;
1618 }
1619 
1620 /**
1621  * ubifs_remount_ro - re-mount in read-only mode.
1622  * @c: UBIFS file-system description object
1623  *
1624  * We assume VFS has stopped writing. Possibly the background thread could be
1625  * running a commit, however kthread_stop will wait in that case.
1626  */
1627 static void ubifs_remount_ro(struct ubifs_info *c)
1628 {
1629 	int i, err;
1630 
1631 	ubifs_assert(!c->need_recovery);
1632 	ubifs_assert(!(c->vfs_sb->s_flags & MS_RDONLY));
1633 
1634 	mutex_lock(&c->umount_mutex);
1635 	if (c->bgt) {
1636 		kthread_stop(c->bgt);
1637 		c->bgt = NULL;
1638 	}
1639 
1640 	dbg_save_space_info(c);
1641 
1642 	for (i = 0; i < c->jhead_cnt; i++) {
1643 		ubifs_wbuf_sync(&c->jheads[i].wbuf);
1644 		del_timer_sync(&c->jheads[i].wbuf.timer);
1645 	}
1646 
1647 	c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1648 	c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1649 	c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1650 	err = ubifs_write_master(c);
1651 	if (err)
1652 		ubifs_ro_mode(c, err);
1653 
1654 	free_wbufs(c);
1655 	vfree(c->orph_buf);
1656 	c->orph_buf = NULL;
1657 	vfree(c->ileb_buf);
1658 	c->ileb_buf = NULL;
1659 	ubifs_lpt_free(c, 1);
1660 	err = dbg_check_space_info(c);
1661 	if (err)
1662 		ubifs_ro_mode(c, err);
1663 	mutex_unlock(&c->umount_mutex);
1664 }
1665 
1666 static void ubifs_put_super(struct super_block *sb)
1667 {
1668 	int i;
1669 	struct ubifs_info *c = sb->s_fs_info;
1670 
1671 	ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1672 		  c->vi.vol_id);
1673 	/*
1674 	 * The following asserts are only valid if there has not been a failure
1675 	 * of the media. For example, there will be dirty inodes if we failed
1676 	 * to write them back because of I/O errors.
1677 	 */
1678 	ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1679 	ubifs_assert(c->budg_idx_growth == 0);
1680 	ubifs_assert(c->budg_dd_growth == 0);
1681 	ubifs_assert(c->budg_data_growth == 0);
1682 
1683 	/*
1684 	 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1685 	 * and file system un-mount. Namely, it prevents the shrinker from
1686 	 * picking this superblock for shrinking - it will be just skipped if
1687 	 * the mutex is locked.
1688 	 */
1689 	mutex_lock(&c->umount_mutex);
1690 	if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1691 		/*
1692 		 * First of all kill the background thread to make sure it does
1693 		 * not interfere with un-mounting and freeing resources.
1694 		 */
1695 		if (c->bgt) {
1696 			kthread_stop(c->bgt);
1697 			c->bgt = NULL;
1698 		}
1699 
1700 		/* Synchronize write-buffers */
1701 		if (c->jheads)
1702 			for (i = 0; i < c->jhead_cnt; i++) {
1703 				ubifs_wbuf_sync(&c->jheads[i].wbuf);
1704 				del_timer_sync(&c->jheads[i].wbuf.timer);
1705 			}
1706 
1707 		/*
1708 		 * On fatal errors c->ro_media is set to 1, in which case we do
1709 		 * not write the master node.
1710 		 */
1711 		if (!c->ro_media) {
1712 			/*
1713 			 * We are being cleanly unmounted which means the
1714 			 * orphans were killed - indicate this in the master
1715 			 * node. Also save the reserved GC LEB number.
1716 			 */
1717 			int err;
1718 
1719 			c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1720 			c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1721 			c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1722 			err = ubifs_write_master(c);
1723 			if (err)
1724 				/*
1725 				 * Recovery will attempt to fix the master area
1726 				 * next mount, so we just print a message and
1727 				 * continue to unmount normally.
1728 				 */
1729 				ubifs_err("failed to write master node, "
1730 					  "error %d", err);
1731 		}
1732 	}
1733 
1734 	ubifs_umount(c);
1735 	bdi_destroy(&c->bdi);
1736 	ubi_close_volume(c->ubi);
1737 	mutex_unlock(&c->umount_mutex);
1738 	kfree(c);
1739 }
1740 
1741 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1742 {
1743 	int err;
1744 	struct ubifs_info *c = sb->s_fs_info;
1745 
1746 	dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1747 
1748 	err = ubifs_parse_options(c, data, 1);
1749 	if (err) {
1750 		ubifs_err("invalid or unknown remount parameter");
1751 		return err;
1752 	}
1753 
1754 	if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1755 		if (c->ro_media) {
1756 			ubifs_msg("cannot re-mount due to prior errors");
1757 			return -EROFS;
1758 		}
1759 		err = ubifs_remount_rw(c);
1760 		if (err)
1761 			return err;
1762 	} else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY)) {
1763 		if (c->ro_media) {
1764 			ubifs_msg("cannot re-mount due to prior errors");
1765 			return -EROFS;
1766 		}
1767 		ubifs_remount_ro(c);
1768 	}
1769 
1770 	if (c->bulk_read == 1)
1771 		bu_init(c);
1772 	else {
1773 		dbg_gen("disable bulk-read");
1774 		kfree(c->bu.buf);
1775 		c->bu.buf = NULL;
1776 	}
1777 
1778 	ubifs_assert(c->lst.taken_empty_lebs == 1);
1779 	return 0;
1780 }
1781 
1782 const struct super_operations ubifs_super_operations = {
1783 	.alloc_inode   = ubifs_alloc_inode,
1784 	.destroy_inode = ubifs_destroy_inode,
1785 	.put_super     = ubifs_put_super,
1786 	.write_inode   = ubifs_write_inode,
1787 	.delete_inode  = ubifs_delete_inode,
1788 	.statfs        = ubifs_statfs,
1789 	.dirty_inode   = ubifs_dirty_inode,
1790 	.remount_fs    = ubifs_remount_fs,
1791 	.show_options  = ubifs_show_options,
1792 	.sync_fs       = ubifs_sync_fs,
1793 };
1794 
1795 /**
1796  * open_ubi - parse UBI device name string and open the UBI device.
1797  * @name: UBI volume name
1798  * @mode: UBI volume open mode
1799  *
1800  * There are several ways to specify UBI volumes when mounting UBIFS:
1801  * o ubiX_Y    - UBI device number X, volume Y;
1802  * o ubiY      - UBI device number 0, volume Y;
1803  * o ubiX:NAME - mount UBI device X, volume with name NAME;
1804  * o ubi:NAME  - mount UBI device 0, volume with name NAME.
1805  *
1806  * Alternative '!' separator may be used instead of ':' (because some shells
1807  * like busybox may interpret ':' as an NFS host name separator). This function
1808  * returns ubi volume object in case of success and a negative error code in
1809  * case of failure.
1810  */
1811 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1812 {
1813 	int dev, vol;
1814 	char *endptr;
1815 
1816 	if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1817 		return ERR_PTR(-EINVAL);
1818 
1819 	/* ubi:NAME method */
1820 	if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1821 		return ubi_open_volume_nm(0, name + 4, mode);
1822 
1823 	if (!isdigit(name[3]))
1824 		return ERR_PTR(-EINVAL);
1825 
1826 	dev = simple_strtoul(name + 3, &endptr, 0);
1827 
1828 	/* ubiY method */
1829 	if (*endptr == '\0')
1830 		return ubi_open_volume(0, dev, mode);
1831 
1832 	/* ubiX_Y method */
1833 	if (*endptr == '_' && isdigit(endptr[1])) {
1834 		vol = simple_strtoul(endptr + 1, &endptr, 0);
1835 		if (*endptr != '\0')
1836 			return ERR_PTR(-EINVAL);
1837 		return ubi_open_volume(dev, vol, mode);
1838 	}
1839 
1840 	/* ubiX:NAME method */
1841 	if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1842 		return ubi_open_volume_nm(dev, ++endptr, mode);
1843 
1844 	return ERR_PTR(-EINVAL);
1845 }
1846 
1847 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1848 {
1849 	struct ubi_volume_desc *ubi = sb->s_fs_info;
1850 	struct ubifs_info *c;
1851 	struct inode *root;
1852 	int err;
1853 
1854 	c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1855 	if (!c)
1856 		return -ENOMEM;
1857 
1858 	spin_lock_init(&c->cnt_lock);
1859 	spin_lock_init(&c->cs_lock);
1860 	spin_lock_init(&c->buds_lock);
1861 	spin_lock_init(&c->space_lock);
1862 	spin_lock_init(&c->orphan_lock);
1863 	init_rwsem(&c->commit_sem);
1864 	mutex_init(&c->lp_mutex);
1865 	mutex_init(&c->tnc_mutex);
1866 	mutex_init(&c->log_mutex);
1867 	mutex_init(&c->mst_mutex);
1868 	mutex_init(&c->umount_mutex);
1869 	mutex_init(&c->bu_mutex);
1870 	init_waitqueue_head(&c->cmt_wq);
1871 	c->buds = RB_ROOT;
1872 	c->old_idx = RB_ROOT;
1873 	c->size_tree = RB_ROOT;
1874 	c->orph_tree = RB_ROOT;
1875 	INIT_LIST_HEAD(&c->infos_list);
1876 	INIT_LIST_HEAD(&c->idx_gc);
1877 	INIT_LIST_HEAD(&c->replay_list);
1878 	INIT_LIST_HEAD(&c->replay_buds);
1879 	INIT_LIST_HEAD(&c->uncat_list);
1880 	INIT_LIST_HEAD(&c->empty_list);
1881 	INIT_LIST_HEAD(&c->freeable_list);
1882 	INIT_LIST_HEAD(&c->frdi_idx_list);
1883 	INIT_LIST_HEAD(&c->unclean_leb_list);
1884 	INIT_LIST_HEAD(&c->old_buds);
1885 	INIT_LIST_HEAD(&c->orph_list);
1886 	INIT_LIST_HEAD(&c->orph_new);
1887 
1888 	c->highest_inum = UBIFS_FIRST_INO;
1889 	c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1890 
1891 	ubi_get_volume_info(ubi, &c->vi);
1892 	ubi_get_device_info(c->vi.ubi_num, &c->di);
1893 
1894 	/* Re-open the UBI device in read-write mode */
1895 	c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1896 	if (IS_ERR(c->ubi)) {
1897 		err = PTR_ERR(c->ubi);
1898 		goto out_free;
1899 	}
1900 
1901 	/*
1902 	 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1903 	 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1904 	 * which means the user would have to wait not just for their own I/O
1905 	 * but the read-ahead I/O as well i.e. completely pointless.
1906 	 *
1907 	 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1908 	 */
1909 	c->bdi.capabilities = BDI_CAP_MAP_COPY;
1910 	c->bdi.unplug_io_fn = default_unplug_io_fn;
1911 	err  = bdi_init(&c->bdi);
1912 	if (err)
1913 		goto out_close;
1914 
1915 	err = ubifs_parse_options(c, data, 0);
1916 	if (err)
1917 		goto out_bdi;
1918 
1919 	c->vfs_sb = sb;
1920 
1921 	sb->s_fs_info = c;
1922 	sb->s_magic = UBIFS_SUPER_MAGIC;
1923 	sb->s_blocksize = UBIFS_BLOCK_SIZE;
1924 	sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1925 	sb->s_dev = c->vi.cdev;
1926 	sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1927 	if (c->max_inode_sz > MAX_LFS_FILESIZE)
1928 		sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1929 	sb->s_op = &ubifs_super_operations;
1930 
1931 	mutex_lock(&c->umount_mutex);
1932 	err = mount_ubifs(c);
1933 	if (err) {
1934 		ubifs_assert(err < 0);
1935 		goto out_unlock;
1936 	}
1937 
1938 	/* Read the root inode */
1939 	root = ubifs_iget(sb, UBIFS_ROOT_INO);
1940 	if (IS_ERR(root)) {
1941 		err = PTR_ERR(root);
1942 		goto out_umount;
1943 	}
1944 
1945 	sb->s_root = d_alloc_root(root);
1946 	if (!sb->s_root)
1947 		goto out_iput;
1948 
1949 	mutex_unlock(&c->umount_mutex);
1950 	return 0;
1951 
1952 out_iput:
1953 	iput(root);
1954 out_umount:
1955 	ubifs_umount(c);
1956 out_unlock:
1957 	mutex_unlock(&c->umount_mutex);
1958 out_bdi:
1959 	bdi_destroy(&c->bdi);
1960 out_close:
1961 	ubi_close_volume(c->ubi);
1962 out_free:
1963 	kfree(c);
1964 	return err;
1965 }
1966 
1967 static int sb_test(struct super_block *sb, void *data)
1968 {
1969 	dev_t *dev = data;
1970 
1971 	return sb->s_dev == *dev;
1972 }
1973 
1974 static int sb_set(struct super_block *sb, void *data)
1975 {
1976 	dev_t *dev = data;
1977 
1978 	sb->s_dev = *dev;
1979 	return 0;
1980 }
1981 
1982 static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
1983 			const char *name, void *data, struct vfsmount *mnt)
1984 {
1985 	struct ubi_volume_desc *ubi;
1986 	struct ubi_volume_info vi;
1987 	struct super_block *sb;
1988 	int err;
1989 
1990 	dbg_gen("name %s, flags %#x", name, flags);
1991 
1992 	/*
1993 	 * Get UBI device number and volume ID. Mount it read-only so far
1994 	 * because this might be a new mount point, and UBI allows only one
1995 	 * read-write user at a time.
1996 	 */
1997 	ubi = open_ubi(name, UBI_READONLY);
1998 	if (IS_ERR(ubi)) {
1999 		ubifs_err("cannot open \"%s\", error %d",
2000 			  name, (int)PTR_ERR(ubi));
2001 		return PTR_ERR(ubi);
2002 	}
2003 	ubi_get_volume_info(ubi, &vi);
2004 
2005 	dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
2006 
2007 	sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
2008 	if (IS_ERR(sb)) {
2009 		err = PTR_ERR(sb);
2010 		goto out_close;
2011 	}
2012 
2013 	if (sb->s_root) {
2014 		/* A new mount point for already mounted UBIFS */
2015 		dbg_gen("this ubi volume is already mounted");
2016 		if ((flags ^ sb->s_flags) & MS_RDONLY) {
2017 			err = -EBUSY;
2018 			goto out_deact;
2019 		}
2020 	} else {
2021 		sb->s_flags = flags;
2022 		/*
2023 		 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
2024 		 * replaced by 'c'.
2025 		 */
2026 		sb->s_fs_info = ubi;
2027 		err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2028 		if (err)
2029 			goto out_deact;
2030 		/* We do not support atime */
2031 		sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2032 	}
2033 
2034 	/* 'fill_super()' opens ubi again so we must close it here */
2035 	ubi_close_volume(ubi);
2036 
2037 	simple_set_mnt(mnt, sb);
2038 	return 0;
2039 
2040 out_deact:
2041 	up_write(&sb->s_umount);
2042 	deactivate_super(sb);
2043 out_close:
2044 	ubi_close_volume(ubi);
2045 	return err;
2046 }
2047 
2048 static void ubifs_kill_sb(struct super_block *sb)
2049 {
2050 	generic_shutdown_super(sb);
2051 }
2052 
2053 static struct file_system_type ubifs_fs_type = {
2054 	.name    = "ubifs",
2055 	.owner   = THIS_MODULE,
2056 	.get_sb  = ubifs_get_sb,
2057 	.kill_sb = ubifs_kill_sb
2058 };
2059 
2060 /*
2061  * Inode slab cache constructor.
2062  */
2063 static void inode_slab_ctor(void *obj)
2064 {
2065 	struct ubifs_inode *ui = obj;
2066 	inode_init_once(&ui->vfs_inode);
2067 }
2068 
2069 static int __init ubifs_init(void)
2070 {
2071 	int err;
2072 
2073 	BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2074 
2075 	/* Make sure node sizes are 8-byte aligned */
2076 	BUILD_BUG_ON(UBIFS_CH_SZ        & 7);
2077 	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  & 7);
2078 	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2079 	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2080 	BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2081 	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2082 	BUILD_BUG_ON(UBIFS_SB_NODE_SZ   & 7);
2083 	BUILD_BUG_ON(UBIFS_MST_NODE_SZ  & 7);
2084 	BUILD_BUG_ON(UBIFS_REF_NODE_SZ  & 7);
2085 	BUILD_BUG_ON(UBIFS_CS_NODE_SZ   & 7);
2086 	BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2087 
2088 	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2089 	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2090 	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2091 	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  & 7);
2092 	BUILD_BUG_ON(UBIFS_MAX_NODE_SZ      & 7);
2093 	BUILD_BUG_ON(MIN_WRITE_SZ           & 7);
2094 
2095 	/* Check min. node size */
2096 	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  < MIN_WRITE_SZ);
2097 	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2098 	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2099 	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2100 
2101 	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2102 	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2103 	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2104 	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  > UBIFS_MAX_NODE_SZ);
2105 
2106 	/* Defined node sizes */
2107 	BUILD_BUG_ON(UBIFS_SB_NODE_SZ  != 4096);
2108 	BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2109 	BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2110 	BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2111 
2112 	/*
2113 	 * We use 2 bit wide bit-fields to store compression type, which should
2114 	 * be amended if more compressors are added. The bit-fields are:
2115 	 * @compr_type in 'struct ubifs_inode', @default_compr in
2116 	 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2117 	 */
2118 	BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2119 
2120 	/*
2121 	 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2122 	 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2123 	 */
2124 	if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2125 		ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2126 			  " at least 4096 bytes",
2127 			  (unsigned int)PAGE_CACHE_SIZE);
2128 		return -EINVAL;
2129 	}
2130 
2131 	err = register_filesystem(&ubifs_fs_type);
2132 	if (err) {
2133 		ubifs_err("cannot register file system, error %d", err);
2134 		return err;
2135 	}
2136 
2137 	err = -ENOMEM;
2138 	ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2139 				sizeof(struct ubifs_inode), 0,
2140 				SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2141 				&inode_slab_ctor);
2142 	if (!ubifs_inode_slab)
2143 		goto out_reg;
2144 
2145 	register_shrinker(&ubifs_shrinker_info);
2146 
2147 	err = ubifs_compressors_init();
2148 	if (err)
2149 		goto out_shrinker;
2150 
2151 	err = dbg_debugfs_init();
2152 	if (err)
2153 		goto out_compr;
2154 
2155 	return 0;
2156 
2157 out_compr:
2158 	ubifs_compressors_exit();
2159 out_shrinker:
2160 	unregister_shrinker(&ubifs_shrinker_info);
2161 	kmem_cache_destroy(ubifs_inode_slab);
2162 out_reg:
2163 	unregister_filesystem(&ubifs_fs_type);
2164 	return err;
2165 }
2166 /* late_initcall to let compressors initialize first */
2167 late_initcall(ubifs_init);
2168 
2169 static void __exit ubifs_exit(void)
2170 {
2171 	ubifs_assert(list_empty(&ubifs_infos));
2172 	ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2173 
2174 	dbg_debugfs_exit();
2175 	ubifs_compressors_exit();
2176 	unregister_shrinker(&ubifs_shrinker_info);
2177 	kmem_cache_destroy(ubifs_inode_slab);
2178 	unregister_filesystem(&ubifs_fs_type);
2179 }
2180 module_exit(ubifs_exit);
2181 
2182 MODULE_LICENSE("GPL");
2183 MODULE_VERSION(__stringify(UBIFS_VERSION));
2184 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2185 MODULE_DESCRIPTION("UBIFS - UBI File System");
2186