xref: /openbmc/linux/fs/ubifs/super.c (revision fd589a8f)
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 <linux/smp_lock.h>
40 #include "ubifs.h"
41 
42 /*
43  * Maximum amount of memory we may 'kmalloc()' without worrying that we are
44  * allocating too much.
45  */
46 #define UBIFS_KMALLOC_OK (128*1024)
47 
48 /* Slab cache for UBIFS inodes */
49 struct kmem_cache *ubifs_inode_slab;
50 
51 /* UBIFS TNC shrinker description */
52 static struct shrinker ubifs_shrinker_info = {
53 	.shrink = ubifs_shrinker,
54 	.seeks = DEFAULT_SEEKS,
55 };
56 
57 /**
58  * validate_inode - validate inode.
59  * @c: UBIFS file-system description object
60  * @inode: the inode to validate
61  *
62  * This is a helper function for 'ubifs_iget()' which validates various fields
63  * of a newly built inode to make sure they contain sane values and prevent
64  * possible vulnerabilities. Returns zero if the inode is all right and
65  * a non-zero error code if not.
66  */
67 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
68 {
69 	int err;
70 	const struct ubifs_inode *ui = ubifs_inode(inode);
71 
72 	if (inode->i_size > c->max_inode_sz) {
73 		ubifs_err("inode is too large (%lld)",
74 			  (long long)inode->i_size);
75 		return 1;
76 	}
77 
78 	if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
79 		ubifs_err("unknown compression type %d", ui->compr_type);
80 		return 2;
81 	}
82 
83 	if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
84 		return 3;
85 
86 	if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
87 		return 4;
88 
89 	if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
90 		return 5;
91 
92 	if (!ubifs_compr_present(ui->compr_type)) {
93 		ubifs_warn("inode %lu uses '%s' compression, but it was not "
94 			   "compiled in", inode->i_ino,
95 			   ubifs_compr_name(ui->compr_type));
96 	}
97 
98 	err = dbg_check_dir_size(c, inode);
99 	return err;
100 }
101 
102 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
103 {
104 	int err;
105 	union ubifs_key key;
106 	struct ubifs_ino_node *ino;
107 	struct ubifs_info *c = sb->s_fs_info;
108 	struct inode *inode;
109 	struct ubifs_inode *ui;
110 
111 	dbg_gen("inode %lu", inum);
112 
113 	inode = iget_locked(sb, inum);
114 	if (!inode)
115 		return ERR_PTR(-ENOMEM);
116 	if (!(inode->i_state & I_NEW))
117 		return inode;
118 	ui = ubifs_inode(inode);
119 
120 	ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
121 	if (!ino) {
122 		err = -ENOMEM;
123 		goto out;
124 	}
125 
126 	ino_key_init(c, &key, inode->i_ino);
127 
128 	err = ubifs_tnc_lookup(c, &key, ino);
129 	if (err)
130 		goto out_ino;
131 
132 	inode->i_flags |= (S_NOCMTIME | S_NOATIME);
133 	inode->i_nlink = le32_to_cpu(ino->nlink);
134 	inode->i_uid   = le32_to_cpu(ino->uid);
135 	inode->i_gid   = le32_to_cpu(ino->gid);
136 	inode->i_atime.tv_sec  = (int64_t)le64_to_cpu(ino->atime_sec);
137 	inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
138 	inode->i_mtime.tv_sec  = (int64_t)le64_to_cpu(ino->mtime_sec);
139 	inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
140 	inode->i_ctime.tv_sec  = (int64_t)le64_to_cpu(ino->ctime_sec);
141 	inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
142 	inode->i_mode = le32_to_cpu(ino->mode);
143 	inode->i_size = le64_to_cpu(ino->size);
144 
145 	ui->data_len    = le32_to_cpu(ino->data_len);
146 	ui->flags       = le32_to_cpu(ino->flags);
147 	ui->compr_type  = le16_to_cpu(ino->compr_type);
148 	ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
149 	ui->xattr_cnt   = le32_to_cpu(ino->xattr_cnt);
150 	ui->xattr_size  = le32_to_cpu(ino->xattr_size);
151 	ui->xattr_names = le32_to_cpu(ino->xattr_names);
152 	ui->synced_i_size = ui->ui_size = inode->i_size;
153 
154 	ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
155 
156 	err = validate_inode(c, inode);
157 	if (err)
158 		goto out_invalid;
159 
160 	/* Disable read-ahead */
161 	inode->i_mapping->backing_dev_info = &c->bdi;
162 
163 	switch (inode->i_mode & S_IFMT) {
164 	case S_IFREG:
165 		inode->i_mapping->a_ops = &ubifs_file_address_operations;
166 		inode->i_op = &ubifs_file_inode_operations;
167 		inode->i_fop = &ubifs_file_operations;
168 		if (ui->xattr) {
169 			ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
170 			if (!ui->data) {
171 				err = -ENOMEM;
172 				goto out_ino;
173 			}
174 			memcpy(ui->data, ino->data, ui->data_len);
175 			((char *)ui->data)[ui->data_len] = '\0';
176 		} else if (ui->data_len != 0) {
177 			err = 10;
178 			goto out_invalid;
179 		}
180 		break;
181 	case S_IFDIR:
182 		inode->i_op  = &ubifs_dir_inode_operations;
183 		inode->i_fop = &ubifs_dir_operations;
184 		if (ui->data_len != 0) {
185 			err = 11;
186 			goto out_invalid;
187 		}
188 		break;
189 	case S_IFLNK:
190 		inode->i_op = &ubifs_symlink_inode_operations;
191 		if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
192 			err = 12;
193 			goto out_invalid;
194 		}
195 		ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
196 		if (!ui->data) {
197 			err = -ENOMEM;
198 			goto out_ino;
199 		}
200 		memcpy(ui->data, ino->data, ui->data_len);
201 		((char *)ui->data)[ui->data_len] = '\0';
202 		break;
203 	case S_IFBLK:
204 	case S_IFCHR:
205 	{
206 		dev_t rdev;
207 		union ubifs_dev_desc *dev;
208 
209 		ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
210 		if (!ui->data) {
211 			err = -ENOMEM;
212 			goto out_ino;
213 		}
214 
215 		dev = (union ubifs_dev_desc *)ino->data;
216 		if (ui->data_len == sizeof(dev->new))
217 			rdev = new_decode_dev(le32_to_cpu(dev->new));
218 		else if (ui->data_len == sizeof(dev->huge))
219 			rdev = huge_decode_dev(le64_to_cpu(dev->huge));
220 		else {
221 			err = 13;
222 			goto out_invalid;
223 		}
224 		memcpy(ui->data, ino->data, ui->data_len);
225 		inode->i_op = &ubifs_file_inode_operations;
226 		init_special_inode(inode, inode->i_mode, rdev);
227 		break;
228 	}
229 	case S_IFSOCK:
230 	case S_IFIFO:
231 		inode->i_op = &ubifs_file_inode_operations;
232 		init_special_inode(inode, inode->i_mode, 0);
233 		if (ui->data_len != 0) {
234 			err = 14;
235 			goto out_invalid;
236 		}
237 		break;
238 	default:
239 		err = 15;
240 		goto out_invalid;
241 	}
242 
243 	kfree(ino);
244 	ubifs_set_inode_flags(inode);
245 	unlock_new_inode(inode);
246 	return inode;
247 
248 out_invalid:
249 	ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
250 	dbg_dump_node(c, ino);
251 	dbg_dump_inode(c, inode);
252 	err = -EINVAL;
253 out_ino:
254 	kfree(ino);
255 out:
256 	ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
257 	iget_failed(inode);
258 	return ERR_PTR(err);
259 }
260 
261 static struct inode *ubifs_alloc_inode(struct super_block *sb)
262 {
263 	struct ubifs_inode *ui;
264 
265 	ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
266 	if (!ui)
267 		return NULL;
268 
269 	memset((void *)ui + sizeof(struct inode), 0,
270 	       sizeof(struct ubifs_inode) - sizeof(struct inode));
271 	mutex_init(&ui->ui_mutex);
272 	spin_lock_init(&ui->ui_lock);
273 	return &ui->vfs_inode;
274 };
275 
276 static void ubifs_destroy_inode(struct inode *inode)
277 {
278 	struct ubifs_inode *ui = ubifs_inode(inode);
279 
280 	kfree(ui->data);
281 	kmem_cache_free(ubifs_inode_slab, inode);
282 }
283 
284 /*
285  * Note, Linux write-back code calls this without 'i_mutex'.
286  */
287 static int ubifs_write_inode(struct inode *inode, int wait)
288 {
289 	int err = 0;
290 	struct ubifs_info *c = inode->i_sb->s_fs_info;
291 	struct ubifs_inode *ui = ubifs_inode(inode);
292 
293 	ubifs_assert(!ui->xattr);
294 	if (is_bad_inode(inode))
295 		return 0;
296 
297 	mutex_lock(&ui->ui_mutex);
298 	/*
299 	 * Due to races between write-back forced by budgeting
300 	 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
301 	 * have already been synchronized, do not do this again. This might
302 	 * also happen if it was synchronized in an VFS operation, e.g.
303 	 * 'ubifs_link()'.
304 	 */
305 	if (!ui->dirty) {
306 		mutex_unlock(&ui->ui_mutex);
307 		return 0;
308 	}
309 
310 	/*
311 	 * As an optimization, do not write orphan inodes to the media just
312 	 * because this is not needed.
313 	 */
314 	dbg_gen("inode %lu, mode %#x, nlink %u",
315 		inode->i_ino, (int)inode->i_mode, inode->i_nlink);
316 	if (inode->i_nlink) {
317 		err = ubifs_jnl_write_inode(c, inode);
318 		if (err)
319 			ubifs_err("can't write inode %lu, error %d",
320 				  inode->i_ino, err);
321 	}
322 
323 	ui->dirty = 0;
324 	mutex_unlock(&ui->ui_mutex);
325 	ubifs_release_dirty_inode_budget(c, ui);
326 	return err;
327 }
328 
329 static void ubifs_delete_inode(struct inode *inode)
330 {
331 	int err;
332 	struct ubifs_info *c = inode->i_sb->s_fs_info;
333 	struct ubifs_inode *ui = ubifs_inode(inode);
334 
335 	if (ui->xattr)
336 		/*
337 		 * Extended attribute inode deletions are fully handled in
338 		 * 'ubifs_removexattr()'. These inodes are special and have
339 		 * limited usage, so there is nothing to do here.
340 		 */
341 		goto out;
342 
343 	dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
344 	ubifs_assert(!atomic_read(&inode->i_count));
345 	ubifs_assert(inode->i_nlink == 0);
346 
347 	truncate_inode_pages(&inode->i_data, 0);
348 	if (is_bad_inode(inode))
349 		goto out;
350 
351 	ui->ui_size = inode->i_size = 0;
352 	err = ubifs_jnl_delete_inode(c, inode);
353 	if (err)
354 		/*
355 		 * Worst case we have a lost orphan inode wasting space, so a
356 		 * simple error message is OK here.
357 		 */
358 		ubifs_err("can't delete inode %lu, error %d",
359 			  inode->i_ino, err);
360 
361 out:
362 	if (ui->dirty)
363 		ubifs_release_dirty_inode_budget(c, ui);
364 	else {
365 		/* We've deleted something - clean the "no space" flags */
366 		c->nospace = c->nospace_rp = 0;
367 		smp_wmb();
368 	}
369 	clear_inode(inode);
370 }
371 
372 static void ubifs_dirty_inode(struct inode *inode)
373 {
374 	struct ubifs_inode *ui = ubifs_inode(inode);
375 
376 	ubifs_assert(mutex_is_locked(&ui->ui_mutex));
377 	if (!ui->dirty) {
378 		ui->dirty = 1;
379 		dbg_gen("inode %lu",  inode->i_ino);
380 	}
381 }
382 
383 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
384 {
385 	struct ubifs_info *c = dentry->d_sb->s_fs_info;
386 	unsigned long long free;
387 	__le32 *uuid = (__le32 *)c->uuid;
388 
389 	free = ubifs_get_free_space(c);
390 	dbg_gen("free space %lld bytes (%lld blocks)",
391 		free, free >> UBIFS_BLOCK_SHIFT);
392 
393 	buf->f_type = UBIFS_SUPER_MAGIC;
394 	buf->f_bsize = UBIFS_BLOCK_SIZE;
395 	buf->f_blocks = c->block_cnt;
396 	buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
397 	if (free > c->report_rp_size)
398 		buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
399 	else
400 		buf->f_bavail = 0;
401 	buf->f_files = 0;
402 	buf->f_ffree = 0;
403 	buf->f_namelen = UBIFS_MAX_NLEN;
404 	buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
405 	buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
406 	ubifs_assert(buf->f_bfree <= c->block_cnt);
407 	return 0;
408 }
409 
410 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
411 {
412 	struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
413 
414 	if (c->mount_opts.unmount_mode == 2)
415 		seq_printf(s, ",fast_unmount");
416 	else if (c->mount_opts.unmount_mode == 1)
417 		seq_printf(s, ",norm_unmount");
418 
419 	if (c->mount_opts.bulk_read == 2)
420 		seq_printf(s, ",bulk_read");
421 	else if (c->mount_opts.bulk_read == 1)
422 		seq_printf(s, ",no_bulk_read");
423 
424 	if (c->mount_opts.chk_data_crc == 2)
425 		seq_printf(s, ",chk_data_crc");
426 	else if (c->mount_opts.chk_data_crc == 1)
427 		seq_printf(s, ",no_chk_data_crc");
428 
429 	if (c->mount_opts.override_compr) {
430 		seq_printf(s, ",compr=%s",
431 			   ubifs_compr_name(c->mount_opts.compr_type));
432 	}
433 
434 	return 0;
435 }
436 
437 static int ubifs_sync_fs(struct super_block *sb, int wait)
438 {
439 	int i, err;
440 	struct ubifs_info *c = sb->s_fs_info;
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 	/*
451 	 * VFS calls '->sync_fs()' before synchronizing all dirty inodes and
452 	 * pages, so synchronize them first, then commit the journal. Strictly
453 	 * speaking, it is not necessary to commit the journal here,
454 	 * synchronizing write-buffers would be enough. But committing makes
455 	 * UBIFS free space predictions much more accurate, so we want to let
456 	 * the user be able to get more accurate results of 'statfs()' after
457 	 * they synchronize the file system.
458 	 */
459 	sync_inodes_sb(sb);
460 
461 	/*
462 	 * Synchronize write buffers, because 'ubifs_run_commit()' does not
463 	 * do this if it waits for an already running commit.
464 	 */
465 	for (i = 0; i < c->jhead_cnt; i++) {
466 		err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
467 		if (err)
468 			return err;
469 	}
470 
471 	err = ubifs_run_commit(c);
472 	if (err)
473 		return err;
474 
475 	return ubi_sync(c->vi.ubi_num);
476 }
477 
478 /**
479  * init_constants_early - initialize UBIFS constants.
480  * @c: UBIFS file-system description object
481  *
482  * This function initialize UBIFS constants which do not need the superblock to
483  * be read. It also checks that the UBI volume satisfies basic UBIFS
484  * requirements. Returns zero in case of success and a negative error code in
485  * case of failure.
486  */
487 static int init_constants_early(struct ubifs_info *c)
488 {
489 	if (c->vi.corrupted) {
490 		ubifs_warn("UBI volume is corrupted - read-only mode");
491 		c->ro_media = 1;
492 	}
493 
494 	if (c->di.ro_mode) {
495 		ubifs_msg("read-only UBI device");
496 		c->ro_media = 1;
497 	}
498 
499 	if (c->vi.vol_type == UBI_STATIC_VOLUME) {
500 		ubifs_msg("static UBI volume - read-only mode");
501 		c->ro_media = 1;
502 	}
503 
504 	c->leb_cnt = c->vi.size;
505 	c->leb_size = c->vi.usable_leb_size;
506 	c->half_leb_size = c->leb_size / 2;
507 	c->min_io_size = c->di.min_io_size;
508 	c->min_io_shift = fls(c->min_io_size) - 1;
509 
510 	if (c->leb_size < UBIFS_MIN_LEB_SZ) {
511 		ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
512 			  c->leb_size, UBIFS_MIN_LEB_SZ);
513 		return -EINVAL;
514 	}
515 
516 	if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
517 		ubifs_err("too few LEBs (%d), min. is %d",
518 			  c->leb_cnt, UBIFS_MIN_LEB_CNT);
519 		return -EINVAL;
520 	}
521 
522 	if (!is_power_of_2(c->min_io_size)) {
523 		ubifs_err("bad min. I/O size %d", c->min_io_size);
524 		return -EINVAL;
525 	}
526 
527 	/*
528 	 * UBIFS aligns all node to 8-byte boundary, so to make function in
529 	 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
530 	 * less than 8.
531 	 */
532 	if (c->min_io_size < 8) {
533 		c->min_io_size = 8;
534 		c->min_io_shift = 3;
535 	}
536 
537 	c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
538 	c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
539 
540 	/*
541 	 * Initialize node length ranges which are mostly needed for node
542 	 * length validation.
543 	 */
544 	c->ranges[UBIFS_PAD_NODE].len  = UBIFS_PAD_NODE_SZ;
545 	c->ranges[UBIFS_SB_NODE].len   = UBIFS_SB_NODE_SZ;
546 	c->ranges[UBIFS_MST_NODE].len  = UBIFS_MST_NODE_SZ;
547 	c->ranges[UBIFS_REF_NODE].len  = UBIFS_REF_NODE_SZ;
548 	c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
549 	c->ranges[UBIFS_CS_NODE].len   = UBIFS_CS_NODE_SZ;
550 
551 	c->ranges[UBIFS_INO_NODE].min_len  = UBIFS_INO_NODE_SZ;
552 	c->ranges[UBIFS_INO_NODE].max_len  = UBIFS_MAX_INO_NODE_SZ;
553 	c->ranges[UBIFS_ORPH_NODE].min_len =
554 				UBIFS_ORPH_NODE_SZ + sizeof(__le64);
555 	c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
556 	c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
557 	c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
558 	c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
559 	c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
560 	c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
561 	c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
562 	/*
563 	 * Minimum indexing node size is amended later when superblock is
564 	 * read and the key length is known.
565 	 */
566 	c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
567 	/*
568 	 * Maximum indexing node size is amended later when superblock is
569 	 * read and the fanout is known.
570 	 */
571 	c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
572 
573 	/*
574 	 * Initialize dead and dark LEB space watermarks. See gc.c for comments
575 	 * about these values.
576 	 */
577 	c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
578 	c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
579 
580 	/*
581 	 * Calculate how many bytes would be wasted at the end of LEB if it was
582 	 * fully filled with data nodes of maximum size. This is used in
583 	 * calculations when reporting free space.
584 	 */
585 	c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
586 
587 	/* Buffer size for bulk-reads */
588 	c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
589 	if (c->max_bu_buf_len > c->leb_size)
590 		c->max_bu_buf_len = c->leb_size;
591 	return 0;
592 }
593 
594 /**
595  * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
596  * @c: UBIFS file-system description object
597  * @lnum: LEB the write-buffer was synchronized to
598  * @free: how many free bytes left in this LEB
599  * @pad: how many bytes were padded
600  *
601  * This is a callback function which is called by the I/O unit when the
602  * write-buffer is synchronized. We need this to correctly maintain space
603  * accounting in bud logical eraseblocks. This function returns zero in case of
604  * success and a negative error code in case of failure.
605  *
606  * This function actually belongs to the journal, but we keep it here because
607  * we want to keep it static.
608  */
609 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
610 {
611 	return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
612 }
613 
614 /*
615  * init_constants_sb - initialize UBIFS constants.
616  * @c: UBIFS file-system description object
617  *
618  * This is a helper function which initializes various UBIFS constants after
619  * the superblock has been read. It also checks various UBIFS parameters and
620  * makes sure they are all right. Returns zero in case of success and a
621  * negative error code in case of failure.
622  */
623 static int init_constants_sb(struct ubifs_info *c)
624 {
625 	int tmp, err;
626 	long long tmp64;
627 
628 	c->main_bytes = (long long)c->main_lebs * c->leb_size;
629 	c->max_znode_sz = sizeof(struct ubifs_znode) +
630 				c->fanout * sizeof(struct ubifs_zbranch);
631 
632 	tmp = ubifs_idx_node_sz(c, 1);
633 	c->ranges[UBIFS_IDX_NODE].min_len = tmp;
634 	c->min_idx_node_sz = ALIGN(tmp, 8);
635 
636 	tmp = ubifs_idx_node_sz(c, c->fanout);
637 	c->ranges[UBIFS_IDX_NODE].max_len = tmp;
638 	c->max_idx_node_sz = ALIGN(tmp, 8);
639 
640 	/* Make sure LEB size is large enough to fit full commit */
641 	tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
642 	tmp = ALIGN(tmp, c->min_io_size);
643 	if (tmp > c->leb_size) {
644 		dbg_err("too small LEB size %d, at least %d needed",
645 			c->leb_size, tmp);
646 		return -EINVAL;
647 	}
648 
649 	/*
650 	 * Make sure that the log is large enough to fit reference nodes for
651 	 * all buds plus one reserved LEB.
652 	 */
653 	tmp64 = c->max_bud_bytes + c->leb_size - 1;
654 	c->max_bud_cnt = div_u64(tmp64, c->leb_size);
655 	tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
656 	tmp /= c->leb_size;
657 	tmp += 1;
658 	if (c->log_lebs < tmp) {
659 		dbg_err("too small log %d LEBs, required min. %d LEBs",
660 			c->log_lebs, tmp);
661 		return -EINVAL;
662 	}
663 
664 	/*
665 	 * When budgeting we assume worst-case scenarios when the pages are not
666 	 * be compressed and direntries are of the maximum size.
667 	 *
668 	 * Note, data, which may be stored in inodes is budgeted separately, so
669 	 * it is not included into 'c->inode_budget'.
670 	 */
671 	c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
672 	c->inode_budget = UBIFS_INO_NODE_SZ;
673 	c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
674 
675 	/*
676 	 * When the amount of flash space used by buds becomes
677 	 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
678 	 * The writers are unblocked when the commit is finished. To avoid
679 	 * writers to be blocked UBIFS initiates background commit in advance,
680 	 * when number of bud bytes becomes above the limit defined below.
681 	 */
682 	c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
683 
684 	/*
685 	 * Ensure minimum journal size. All the bytes in the journal heads are
686 	 * considered to be used, when calculating the current journal usage.
687 	 * Consequently, if the journal is too small, UBIFS will treat it as
688 	 * always full.
689 	 */
690 	tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
691 	if (c->bg_bud_bytes < tmp64)
692 		c->bg_bud_bytes = tmp64;
693 	if (c->max_bud_bytes < tmp64 + c->leb_size)
694 		c->max_bud_bytes = tmp64 + c->leb_size;
695 
696 	err = ubifs_calc_lpt_geom(c);
697 	if (err)
698 		return err;
699 
700 	/* Initialize effective LEB size used in budgeting calculations */
701 	c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
702 	return 0;
703 }
704 
705 /*
706  * init_constants_master - initialize UBIFS constants.
707  * @c: UBIFS file-system description object
708  *
709  * This is a helper function which initializes various UBIFS constants after
710  * the master node has been read. It also checks various UBIFS parameters and
711  * makes sure they are all right.
712  */
713 static void init_constants_master(struct ubifs_info *c)
714 {
715 	long long tmp64;
716 
717 	c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
718 	c->report_rp_size = ubifs_reported_space(c, c->rp_size);
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.no_timer = 1;
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  * parse_standard_option - parse a standard mount option.
936  * @option: the option to parse
937  *
938  * Normally, standard mount options like "sync" are passed to file-systems as
939  * flags. However, when a "rootflags=" kernel boot parameter is used, they may
940  * be present in the options string. This function tries to deal with this
941  * situation and parse standard options. Returns 0 if the option was not
942  * recognized, and the corresponding integer flag if it was.
943  *
944  * UBIFS is only interested in the "sync" option, so do not check for anything
945  * else.
946  */
947 static int parse_standard_option(const char *option)
948 {
949 	ubifs_msg("parse %s", option);
950 	if (!strcmp(option, "sync"))
951 		return MS_SYNCHRONOUS;
952 	return 0;
953 }
954 
955 /**
956  * ubifs_parse_options - parse mount parameters.
957  * @c: UBIFS file-system description object
958  * @options: parameters to parse
959  * @is_remount: non-zero if this is FS re-mount
960  *
961  * This function parses UBIFS mount options and returns zero in case success
962  * and a negative error code in case of failure.
963  */
964 static int ubifs_parse_options(struct ubifs_info *c, char *options,
965 			       int is_remount)
966 {
967 	char *p;
968 	substring_t args[MAX_OPT_ARGS];
969 
970 	if (!options)
971 		return 0;
972 
973 	while ((p = strsep(&options, ","))) {
974 		int token;
975 
976 		if (!*p)
977 			continue;
978 
979 		token = match_token(p, tokens, args);
980 		switch (token) {
981 		/*
982 		 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
983 		 * We accept them in order to be backward-compatible. But this
984 		 * should be removed at some point.
985 		 */
986 		case Opt_fast_unmount:
987 			c->mount_opts.unmount_mode = 2;
988 			break;
989 		case Opt_norm_unmount:
990 			c->mount_opts.unmount_mode = 1;
991 			break;
992 		case Opt_bulk_read:
993 			c->mount_opts.bulk_read = 2;
994 			c->bulk_read = 1;
995 			break;
996 		case Opt_no_bulk_read:
997 			c->mount_opts.bulk_read = 1;
998 			c->bulk_read = 0;
999 			break;
1000 		case Opt_chk_data_crc:
1001 			c->mount_opts.chk_data_crc = 2;
1002 			c->no_chk_data_crc = 0;
1003 			break;
1004 		case Opt_no_chk_data_crc:
1005 			c->mount_opts.chk_data_crc = 1;
1006 			c->no_chk_data_crc = 1;
1007 			break;
1008 		case Opt_override_compr:
1009 		{
1010 			char *name = match_strdup(&args[0]);
1011 
1012 			if (!name)
1013 				return -ENOMEM;
1014 			if (!strcmp(name, "none"))
1015 				c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1016 			else if (!strcmp(name, "lzo"))
1017 				c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1018 			else if (!strcmp(name, "zlib"))
1019 				c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1020 			else {
1021 				ubifs_err("unknown compressor \"%s\"", name);
1022 				kfree(name);
1023 				return -EINVAL;
1024 			}
1025 			kfree(name);
1026 			c->mount_opts.override_compr = 1;
1027 			c->default_compr = c->mount_opts.compr_type;
1028 			break;
1029 		}
1030 		default:
1031 		{
1032 			unsigned long flag;
1033 			struct super_block *sb = c->vfs_sb;
1034 
1035 			flag = parse_standard_option(p);
1036 			if (!flag) {
1037 				ubifs_err("unrecognized mount option \"%s\" "
1038 					  "or missing value", p);
1039 				return -EINVAL;
1040 			}
1041 			sb->s_flags |= flag;
1042 			break;
1043 		}
1044 		}
1045 	}
1046 
1047 	return 0;
1048 }
1049 
1050 /**
1051  * destroy_journal - destroy journal data structures.
1052  * @c: UBIFS file-system description object
1053  *
1054  * This function destroys journal data structures including those that may have
1055  * been created by recovery functions.
1056  */
1057 static void destroy_journal(struct ubifs_info *c)
1058 {
1059 	while (!list_empty(&c->unclean_leb_list)) {
1060 		struct ubifs_unclean_leb *ucleb;
1061 
1062 		ucleb = list_entry(c->unclean_leb_list.next,
1063 				   struct ubifs_unclean_leb, list);
1064 		list_del(&ucleb->list);
1065 		kfree(ucleb);
1066 	}
1067 	while (!list_empty(&c->old_buds)) {
1068 		struct ubifs_bud *bud;
1069 
1070 		bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1071 		list_del(&bud->list);
1072 		kfree(bud);
1073 	}
1074 	ubifs_destroy_idx_gc(c);
1075 	ubifs_destroy_size_tree(c);
1076 	ubifs_tnc_close(c);
1077 	free_buds(c);
1078 }
1079 
1080 /**
1081  * bu_init - initialize bulk-read information.
1082  * @c: UBIFS file-system description object
1083  */
1084 static void bu_init(struct ubifs_info *c)
1085 {
1086 	ubifs_assert(c->bulk_read == 1);
1087 
1088 	if (c->bu.buf)
1089 		return; /* Already initialized */
1090 
1091 again:
1092 	c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1093 	if (!c->bu.buf) {
1094 		if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1095 			c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1096 			goto again;
1097 		}
1098 
1099 		/* Just disable bulk-read */
1100 		ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1101 			   "disabling it", c->max_bu_buf_len);
1102 		c->mount_opts.bulk_read = 1;
1103 		c->bulk_read = 0;
1104 		return;
1105 	}
1106 }
1107 
1108 /**
1109  * check_free_space - check if there is enough free space to mount.
1110  * @c: UBIFS file-system description object
1111  *
1112  * This function makes sure UBIFS has enough free space to be mounted in
1113  * read/write mode. UBIFS must always have some free space to allow deletions.
1114  */
1115 static int check_free_space(struct ubifs_info *c)
1116 {
1117 	ubifs_assert(c->dark_wm > 0);
1118 	if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1119 		ubifs_err("insufficient free space to mount in read/write mode");
1120 		dbg_dump_budg(c);
1121 		dbg_dump_lprops(c);
1122 		return -ENOSPC;
1123 	}
1124 	return 0;
1125 }
1126 
1127 /**
1128  * mount_ubifs - mount UBIFS file-system.
1129  * @c: UBIFS file-system description object
1130  *
1131  * This function mounts UBIFS file system. Returns zero in case of success and
1132  * a negative error code in case of failure.
1133  *
1134  * Note, the function does not de-allocate resources it it fails half way
1135  * through, and the caller has to do this instead.
1136  */
1137 static int mount_ubifs(struct ubifs_info *c)
1138 {
1139 	struct super_block *sb = c->vfs_sb;
1140 	int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1141 	long long x;
1142 	size_t sz;
1143 
1144 	err = init_constants_early(c);
1145 	if (err)
1146 		return err;
1147 
1148 	err = ubifs_debugging_init(c);
1149 	if (err)
1150 		return err;
1151 
1152 	err = check_volume_empty(c);
1153 	if (err)
1154 		goto out_free;
1155 
1156 	if (c->empty && (mounted_read_only || c->ro_media)) {
1157 		/*
1158 		 * This UBI volume is empty, and read-only, or the file system
1159 		 * is mounted read-only - we cannot format it.
1160 		 */
1161 		ubifs_err("can't format empty UBI volume: read-only %s",
1162 			  c->ro_media ? "UBI volume" : "mount");
1163 		err = -EROFS;
1164 		goto out_free;
1165 	}
1166 
1167 	if (c->ro_media && !mounted_read_only) {
1168 		ubifs_err("cannot mount read-write - read-only media");
1169 		err = -EROFS;
1170 		goto out_free;
1171 	}
1172 
1173 	/*
1174 	 * The requirement for the buffer is that it should fit indexing B-tree
1175 	 * height amount of integers. We assume the height if the TNC tree will
1176 	 * never exceed 64.
1177 	 */
1178 	err = -ENOMEM;
1179 	c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1180 	if (!c->bottom_up_buf)
1181 		goto out_free;
1182 
1183 	c->sbuf = vmalloc(c->leb_size);
1184 	if (!c->sbuf)
1185 		goto out_free;
1186 
1187 	if (!mounted_read_only) {
1188 		c->ileb_buf = vmalloc(c->leb_size);
1189 		if (!c->ileb_buf)
1190 			goto out_free;
1191 	}
1192 
1193 	if (c->bulk_read == 1)
1194 		bu_init(c);
1195 
1196 	/*
1197 	 * We have to check all CRCs, even for data nodes, when we mount the FS
1198 	 * (specifically, when we are replaying).
1199 	 */
1200 	c->always_chk_crc = 1;
1201 
1202 	err = ubifs_read_superblock(c);
1203 	if (err)
1204 		goto out_free;
1205 
1206 	/*
1207 	 * Make sure the compressor which is set as default in the superblock
1208 	 * or overridden by mount options is actually compiled in.
1209 	 */
1210 	if (!ubifs_compr_present(c->default_compr)) {
1211 		ubifs_err("'compressor \"%s\" is not compiled in",
1212 			  ubifs_compr_name(c->default_compr));
1213 		err = -ENOTSUPP;
1214 		goto out_free;
1215 	}
1216 
1217 	err = init_constants_sb(c);
1218 	if (err)
1219 		goto out_free;
1220 
1221 	sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1222 	sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1223 	c->cbuf = kmalloc(sz, GFP_NOFS);
1224 	if (!c->cbuf) {
1225 		err = -ENOMEM;
1226 		goto out_free;
1227 	}
1228 
1229 	sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1230 	if (!mounted_read_only) {
1231 		err = alloc_wbufs(c);
1232 		if (err)
1233 			goto out_cbuf;
1234 
1235 		/* Create background thread */
1236 		c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1237 		if (IS_ERR(c->bgt)) {
1238 			err = PTR_ERR(c->bgt);
1239 			c->bgt = NULL;
1240 			ubifs_err("cannot spawn \"%s\", error %d",
1241 				  c->bgt_name, err);
1242 			goto out_wbufs;
1243 		}
1244 		wake_up_process(c->bgt);
1245 	}
1246 
1247 	err = ubifs_read_master(c);
1248 	if (err)
1249 		goto out_master;
1250 
1251 	init_constants_master(c);
1252 
1253 	if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1254 		ubifs_msg("recovery needed");
1255 		c->need_recovery = 1;
1256 		if (!mounted_read_only) {
1257 			err = ubifs_recover_inl_heads(c, c->sbuf);
1258 			if (err)
1259 				goto out_master;
1260 		}
1261 	} else if (!mounted_read_only) {
1262 		/*
1263 		 * Set the "dirty" flag so that if we reboot uncleanly we
1264 		 * will notice this immediately on the next mount.
1265 		 */
1266 		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1267 		err = ubifs_write_master(c);
1268 		if (err)
1269 			goto out_master;
1270 	}
1271 
1272 	err = ubifs_lpt_init(c, 1, !mounted_read_only);
1273 	if (err)
1274 		goto out_lpt;
1275 
1276 	err = dbg_check_idx_size(c, c->old_idx_sz);
1277 	if (err)
1278 		goto out_lpt;
1279 
1280 	err = ubifs_replay_journal(c);
1281 	if (err)
1282 		goto out_journal;
1283 
1284 	/* Calculate 'min_idx_lebs' after journal replay */
1285 	c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1286 
1287 	err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1288 	if (err)
1289 		goto out_orphans;
1290 
1291 	if (!mounted_read_only) {
1292 		int lnum;
1293 
1294 		err = check_free_space(c);
1295 		if (err)
1296 			goto out_orphans;
1297 
1298 		/* Check for enough log space */
1299 		lnum = c->lhead_lnum + 1;
1300 		if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1301 			lnum = UBIFS_LOG_LNUM;
1302 		if (lnum == c->ltail_lnum) {
1303 			err = ubifs_consolidate_log(c);
1304 			if (err)
1305 				goto out_orphans;
1306 		}
1307 
1308 		if (c->need_recovery) {
1309 			err = ubifs_recover_size(c);
1310 			if (err)
1311 				goto out_orphans;
1312 			err = ubifs_rcvry_gc_commit(c);
1313 		} else {
1314 			err = take_gc_lnum(c);
1315 			if (err)
1316 				goto out_orphans;
1317 
1318 			/*
1319 			 * GC LEB may contain garbage if there was an unclean
1320 			 * reboot, and it should be un-mapped.
1321 			 */
1322 			err = ubifs_leb_unmap(c, c->gc_lnum);
1323 			if (err)
1324 				return err;
1325 		}
1326 
1327 		err = dbg_check_lprops(c);
1328 		if (err)
1329 			goto out_orphans;
1330 	} else if (c->need_recovery) {
1331 		err = ubifs_recover_size(c);
1332 		if (err)
1333 			goto out_orphans;
1334 	} else {
1335 		/*
1336 		 * Even if we mount read-only, we have to set space in GC LEB
1337 		 * to proper value because this affects UBIFS free space
1338 		 * reporting. We do not want to have a situation when
1339 		 * re-mounting from R/O to R/W changes amount of free space.
1340 		 */
1341 		err = take_gc_lnum(c);
1342 		if (err)
1343 			goto out_orphans;
1344 	}
1345 
1346 	spin_lock(&ubifs_infos_lock);
1347 	list_add_tail(&c->infos_list, &ubifs_infos);
1348 	spin_unlock(&ubifs_infos_lock);
1349 
1350 	if (c->need_recovery) {
1351 		if (mounted_read_only)
1352 			ubifs_msg("recovery deferred");
1353 		else {
1354 			c->need_recovery = 0;
1355 			ubifs_msg("recovery completed");
1356 			/*
1357 			 * GC LEB has to be empty and taken at this point. But
1358 			 * the journal head LEBs may also be accounted as
1359 			 * "empty taken" if they are empty.
1360 			 */
1361 			ubifs_assert(c->lst.taken_empty_lebs > 0);
1362 		}
1363 	} else
1364 		ubifs_assert(c->lst.taken_empty_lebs > 0);
1365 
1366 	err = dbg_check_filesystem(c);
1367 	if (err)
1368 		goto out_infos;
1369 
1370 	err = dbg_debugfs_init_fs(c);
1371 	if (err)
1372 		goto out_infos;
1373 
1374 	c->always_chk_crc = 0;
1375 
1376 	ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1377 		  c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1378 	if (mounted_read_only)
1379 		ubifs_msg("mounted read-only");
1380 	x = (long long)c->main_lebs * c->leb_size;
1381 	ubifs_msg("file system size:   %lld bytes (%lld KiB, %lld MiB, %d "
1382 		  "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1383 	x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1384 	ubifs_msg("journal size:       %lld bytes (%lld KiB, %lld MiB, %d "
1385 		  "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1386 	ubifs_msg("media format:       w%d/r%d (latest is w%d/r%d)",
1387 		  c->fmt_version, c->ro_compat_version,
1388 		  UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1389 	ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1390 	ubifs_msg("reserved for root:  %llu bytes (%llu KiB)",
1391 		c->report_rp_size, c->report_rp_size >> 10);
1392 
1393 	dbg_msg("compiled on:         " __DATE__ " at " __TIME__);
1394 	dbg_msg("min. I/O unit size:  %d bytes", c->min_io_size);
1395 	dbg_msg("LEB size:            %d bytes (%d KiB)",
1396 		c->leb_size, c->leb_size >> 10);
1397 	dbg_msg("data journal heads:  %d",
1398 		c->jhead_cnt - NONDATA_JHEADS_CNT);
1399 	dbg_msg("UUID:                %02X%02X%02X%02X-%02X%02X"
1400 	       "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1401 	       c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1402 	       c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1403 	       c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1404 	       c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1405 	dbg_msg("big_lpt              %d", c->big_lpt);
1406 	dbg_msg("log LEBs:            %d (%d - %d)",
1407 		c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1408 	dbg_msg("LPT area LEBs:       %d (%d - %d)",
1409 		c->lpt_lebs, c->lpt_first, c->lpt_last);
1410 	dbg_msg("orphan area LEBs:    %d (%d - %d)",
1411 		c->orph_lebs, c->orph_first, c->orph_last);
1412 	dbg_msg("main area LEBs:      %d (%d - %d)",
1413 		c->main_lebs, c->main_first, c->leb_cnt - 1);
1414 	dbg_msg("index LEBs:          %d", c->lst.idx_lebs);
1415 	dbg_msg("total index bytes:   %lld (%lld KiB, %lld MiB)",
1416 		c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1417 	dbg_msg("key hash type:       %d", c->key_hash_type);
1418 	dbg_msg("tree fanout:         %d", c->fanout);
1419 	dbg_msg("reserved GC LEB:     %d", c->gc_lnum);
1420 	dbg_msg("first main LEB:      %d", c->main_first);
1421 	dbg_msg("max. znode size      %d", c->max_znode_sz);
1422 	dbg_msg("max. index node size %d", c->max_idx_node_sz);
1423 	dbg_msg("node sizes:          data %zu, inode %zu, dentry %zu",
1424 		UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1425 	dbg_msg("node sizes:          trun %zu, sb %zu, master %zu",
1426 		UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1427 	dbg_msg("node sizes:          ref %zu, cmt. start %zu, orph %zu",
1428 		UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1429 	dbg_msg("max. node sizes:     data %zu, inode %zu dentry %zu",
1430 	        UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1431 		UBIFS_MAX_DENT_NODE_SZ);
1432 	dbg_msg("dead watermark:      %d", c->dead_wm);
1433 	dbg_msg("dark watermark:      %d", c->dark_wm);
1434 	dbg_msg("LEB overhead:        %d", c->leb_overhead);
1435 	x = (long long)c->main_lebs * c->dark_wm;
1436 	dbg_msg("max. dark space:     %lld (%lld KiB, %lld MiB)",
1437 		x, x >> 10, x >> 20);
1438 	dbg_msg("maximum bud bytes:   %lld (%lld KiB, %lld MiB)",
1439 		c->max_bud_bytes, c->max_bud_bytes >> 10,
1440 		c->max_bud_bytes >> 20);
1441 	dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1442 		c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1443 		c->bg_bud_bytes >> 20);
1444 	dbg_msg("current bud bytes    %lld (%lld KiB, %lld MiB)",
1445 		c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1446 	dbg_msg("max. seq. number:    %llu", c->max_sqnum);
1447 	dbg_msg("commit number:       %llu", c->cmt_no);
1448 
1449 	return 0;
1450 
1451 out_infos:
1452 	spin_lock(&ubifs_infos_lock);
1453 	list_del(&c->infos_list);
1454 	spin_unlock(&ubifs_infos_lock);
1455 out_orphans:
1456 	free_orphans(c);
1457 out_journal:
1458 	destroy_journal(c);
1459 out_lpt:
1460 	ubifs_lpt_free(c, 0);
1461 out_master:
1462 	kfree(c->mst_node);
1463 	kfree(c->rcvrd_mst_node);
1464 	if (c->bgt)
1465 		kthread_stop(c->bgt);
1466 out_wbufs:
1467 	free_wbufs(c);
1468 out_cbuf:
1469 	kfree(c->cbuf);
1470 out_free:
1471 	kfree(c->bu.buf);
1472 	vfree(c->ileb_buf);
1473 	vfree(c->sbuf);
1474 	kfree(c->bottom_up_buf);
1475 	ubifs_debugging_exit(c);
1476 	return err;
1477 }
1478 
1479 /**
1480  * ubifs_umount - un-mount UBIFS file-system.
1481  * @c: UBIFS file-system description object
1482  *
1483  * Note, this function is called to free allocated resourced when un-mounting,
1484  * as well as free resources when an error occurred while we were half way
1485  * through mounting (error path cleanup function). So it has to make sure the
1486  * resource was actually allocated before freeing it.
1487  */
1488 static void ubifs_umount(struct ubifs_info *c)
1489 {
1490 	dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1491 		c->vi.vol_id);
1492 
1493 	dbg_debugfs_exit_fs(c);
1494 	spin_lock(&ubifs_infos_lock);
1495 	list_del(&c->infos_list);
1496 	spin_unlock(&ubifs_infos_lock);
1497 
1498 	if (c->bgt)
1499 		kthread_stop(c->bgt);
1500 
1501 	destroy_journal(c);
1502 	free_wbufs(c);
1503 	free_orphans(c);
1504 	ubifs_lpt_free(c, 0);
1505 
1506 	kfree(c->cbuf);
1507 	kfree(c->rcvrd_mst_node);
1508 	kfree(c->mst_node);
1509 	kfree(c->bu.buf);
1510 	vfree(c->ileb_buf);
1511 	vfree(c->sbuf);
1512 	kfree(c->bottom_up_buf);
1513 	ubifs_debugging_exit(c);
1514 }
1515 
1516 /**
1517  * ubifs_remount_rw - re-mount in read-write mode.
1518  * @c: UBIFS file-system description object
1519  *
1520  * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1521  * mode. This function allocates the needed resources and re-mounts UBIFS in
1522  * read-write mode.
1523  */
1524 static int ubifs_remount_rw(struct ubifs_info *c)
1525 {
1526 	int err, lnum;
1527 
1528 	if (c->rw_incompat) {
1529 		ubifs_err("the file-system is not R/W-compatible");
1530 		ubifs_msg("on-flash format version is w%d/r%d, but software "
1531 			  "only supports up to version w%d/r%d", c->fmt_version,
1532 			  c->ro_compat_version, UBIFS_FORMAT_VERSION,
1533 			  UBIFS_RO_COMPAT_VERSION);
1534 		return -EROFS;
1535 	}
1536 
1537 	mutex_lock(&c->umount_mutex);
1538 	dbg_save_space_info(c);
1539 	c->remounting_rw = 1;
1540 	c->always_chk_crc = 1;
1541 
1542 	err = check_free_space(c);
1543 	if (err)
1544 		goto out;
1545 
1546 	if (c->old_leb_cnt != c->leb_cnt) {
1547 		struct ubifs_sb_node *sup;
1548 
1549 		sup = ubifs_read_sb_node(c);
1550 		if (IS_ERR(sup)) {
1551 			err = PTR_ERR(sup);
1552 			goto out;
1553 		}
1554 		sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1555 		err = ubifs_write_sb_node(c, sup);
1556 		if (err)
1557 			goto out;
1558 	}
1559 
1560 	if (c->need_recovery) {
1561 		ubifs_msg("completing deferred recovery");
1562 		err = ubifs_write_rcvrd_mst_node(c);
1563 		if (err)
1564 			goto out;
1565 		err = ubifs_recover_size(c);
1566 		if (err)
1567 			goto out;
1568 		err = ubifs_clean_lebs(c, c->sbuf);
1569 		if (err)
1570 			goto out;
1571 		err = ubifs_recover_inl_heads(c, c->sbuf);
1572 		if (err)
1573 			goto out;
1574 	} else {
1575 		/* A readonly mount is not allowed to have orphans */
1576 		ubifs_assert(c->tot_orphans == 0);
1577 		err = ubifs_clear_orphans(c);
1578 		if (err)
1579 			goto out;
1580 	}
1581 
1582 	if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1583 		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1584 		err = ubifs_write_master(c);
1585 		if (err)
1586 			goto out;
1587 	}
1588 
1589 	c->ileb_buf = vmalloc(c->leb_size);
1590 	if (!c->ileb_buf) {
1591 		err = -ENOMEM;
1592 		goto out;
1593 	}
1594 
1595 	err = ubifs_lpt_init(c, 0, 1);
1596 	if (err)
1597 		goto out;
1598 
1599 	err = alloc_wbufs(c);
1600 	if (err)
1601 		goto out;
1602 
1603 	ubifs_create_buds_lists(c);
1604 
1605 	/* Create background thread */
1606 	c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1607 	if (IS_ERR(c->bgt)) {
1608 		err = PTR_ERR(c->bgt);
1609 		c->bgt = NULL;
1610 		ubifs_err("cannot spawn \"%s\", error %d",
1611 			  c->bgt_name, err);
1612 		goto out;
1613 	}
1614 	wake_up_process(c->bgt);
1615 
1616 	c->orph_buf = vmalloc(c->leb_size);
1617 	if (!c->orph_buf) {
1618 		err = -ENOMEM;
1619 		goto out;
1620 	}
1621 
1622 	/* Check for enough log space */
1623 	lnum = c->lhead_lnum + 1;
1624 	if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1625 		lnum = UBIFS_LOG_LNUM;
1626 	if (lnum == c->ltail_lnum) {
1627 		err = ubifs_consolidate_log(c);
1628 		if (err)
1629 			goto out;
1630 	}
1631 
1632 	if (c->need_recovery)
1633 		err = ubifs_rcvry_gc_commit(c);
1634 	else
1635 		err = ubifs_leb_unmap(c, c->gc_lnum);
1636 	if (err)
1637 		goto out;
1638 
1639 	if (c->need_recovery) {
1640 		c->need_recovery = 0;
1641 		ubifs_msg("deferred recovery completed");
1642 	}
1643 
1644 	dbg_gen("re-mounted read-write");
1645 	c->vfs_sb->s_flags &= ~MS_RDONLY;
1646 	c->remounting_rw = 0;
1647 	c->always_chk_crc = 0;
1648 	err = dbg_check_space_info(c);
1649 	mutex_unlock(&c->umount_mutex);
1650 	return err;
1651 
1652 out:
1653 	vfree(c->orph_buf);
1654 	c->orph_buf = NULL;
1655 	if (c->bgt) {
1656 		kthread_stop(c->bgt);
1657 		c->bgt = NULL;
1658 	}
1659 	free_wbufs(c);
1660 	vfree(c->ileb_buf);
1661 	c->ileb_buf = NULL;
1662 	ubifs_lpt_free(c, 1);
1663 	c->remounting_rw = 0;
1664 	c->always_chk_crc = 0;
1665 	mutex_unlock(&c->umount_mutex);
1666 	return err;
1667 }
1668 
1669 /**
1670  * ubifs_remount_ro - re-mount in read-only mode.
1671  * @c: UBIFS file-system description object
1672  *
1673  * We assume VFS has stopped writing. Possibly the background thread could be
1674  * running a commit, however kthread_stop will wait in that case.
1675  */
1676 static void ubifs_remount_ro(struct ubifs_info *c)
1677 {
1678 	int i, err;
1679 
1680 	ubifs_assert(!c->need_recovery);
1681 	ubifs_assert(!(c->vfs_sb->s_flags & MS_RDONLY));
1682 
1683 	mutex_lock(&c->umount_mutex);
1684 	if (c->bgt) {
1685 		kthread_stop(c->bgt);
1686 		c->bgt = NULL;
1687 	}
1688 
1689 	dbg_save_space_info(c);
1690 
1691 	for (i = 0; i < c->jhead_cnt; i++) {
1692 		ubifs_wbuf_sync(&c->jheads[i].wbuf);
1693 		hrtimer_cancel(&c->jheads[i].wbuf.timer);
1694 	}
1695 
1696 	c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1697 	c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1698 	c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1699 	err = ubifs_write_master(c);
1700 	if (err)
1701 		ubifs_ro_mode(c, err);
1702 
1703 	free_wbufs(c);
1704 	vfree(c->orph_buf);
1705 	c->orph_buf = NULL;
1706 	vfree(c->ileb_buf);
1707 	c->ileb_buf = NULL;
1708 	ubifs_lpt_free(c, 1);
1709 	err = dbg_check_space_info(c);
1710 	if (err)
1711 		ubifs_ro_mode(c, err);
1712 	mutex_unlock(&c->umount_mutex);
1713 }
1714 
1715 static void ubifs_put_super(struct super_block *sb)
1716 {
1717 	int i;
1718 	struct ubifs_info *c = sb->s_fs_info;
1719 
1720 	ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1721 		  c->vi.vol_id);
1722 
1723 	lock_kernel();
1724 
1725 	/*
1726 	 * The following asserts are only valid if there has not been a failure
1727 	 * of the media. For example, there will be dirty inodes if we failed
1728 	 * to write them back because of I/O errors.
1729 	 */
1730 	ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1731 	ubifs_assert(c->budg_idx_growth == 0);
1732 	ubifs_assert(c->budg_dd_growth == 0);
1733 	ubifs_assert(c->budg_data_growth == 0);
1734 
1735 	/*
1736 	 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1737 	 * and file system un-mount. Namely, it prevents the shrinker from
1738 	 * picking this superblock for shrinking - it will be just skipped if
1739 	 * the mutex is locked.
1740 	 */
1741 	mutex_lock(&c->umount_mutex);
1742 	if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1743 		/*
1744 		 * First of all kill the background thread to make sure it does
1745 		 * not interfere with un-mounting and freeing resources.
1746 		 */
1747 		if (c->bgt) {
1748 			kthread_stop(c->bgt);
1749 			c->bgt = NULL;
1750 		}
1751 
1752 		/* Synchronize write-buffers */
1753 		if (c->jheads)
1754 			for (i = 0; i < c->jhead_cnt; i++)
1755 				ubifs_wbuf_sync(&c->jheads[i].wbuf);
1756 
1757 		/*
1758 		 * On fatal errors c->ro_media is set to 1, in which case we do
1759 		 * not write the master node.
1760 		 */
1761 		if (!c->ro_media) {
1762 			/*
1763 			 * We are being cleanly unmounted which means the
1764 			 * orphans were killed - indicate this in the master
1765 			 * node. Also save the reserved GC LEB number.
1766 			 */
1767 			int err;
1768 
1769 			c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1770 			c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1771 			c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1772 			err = ubifs_write_master(c);
1773 			if (err)
1774 				/*
1775 				 * Recovery will attempt to fix the master area
1776 				 * next mount, so we just print a message and
1777 				 * continue to unmount normally.
1778 				 */
1779 				ubifs_err("failed to write master node, "
1780 					  "error %d", err);
1781 		}
1782 	}
1783 
1784 	ubifs_umount(c);
1785 	bdi_destroy(&c->bdi);
1786 	ubi_close_volume(c->ubi);
1787 	mutex_unlock(&c->umount_mutex);
1788 	kfree(c);
1789 
1790 	unlock_kernel();
1791 }
1792 
1793 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1794 {
1795 	int err;
1796 	struct ubifs_info *c = sb->s_fs_info;
1797 
1798 	dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1799 
1800 	err = ubifs_parse_options(c, data, 1);
1801 	if (err) {
1802 		ubifs_err("invalid or unknown remount parameter");
1803 		return err;
1804 	}
1805 
1806 	lock_kernel();
1807 	if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1808 		if (c->ro_media) {
1809 			ubifs_msg("cannot re-mount due to prior errors");
1810 			unlock_kernel();
1811 			return -EROFS;
1812 		}
1813 		err = ubifs_remount_rw(c);
1814 		if (err) {
1815 			unlock_kernel();
1816 			return err;
1817 		}
1818 	} else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY)) {
1819 		if (c->ro_media) {
1820 			ubifs_msg("cannot re-mount due to prior errors");
1821 			unlock_kernel();
1822 			return -EROFS;
1823 		}
1824 		ubifs_remount_ro(c);
1825 	}
1826 
1827 	if (c->bulk_read == 1)
1828 		bu_init(c);
1829 	else {
1830 		dbg_gen("disable bulk-read");
1831 		kfree(c->bu.buf);
1832 		c->bu.buf = NULL;
1833 	}
1834 
1835 	ubifs_assert(c->lst.taken_empty_lebs > 0);
1836 	unlock_kernel();
1837 	return 0;
1838 }
1839 
1840 const struct super_operations ubifs_super_operations = {
1841 	.alloc_inode   = ubifs_alloc_inode,
1842 	.destroy_inode = ubifs_destroy_inode,
1843 	.put_super     = ubifs_put_super,
1844 	.write_inode   = ubifs_write_inode,
1845 	.delete_inode  = ubifs_delete_inode,
1846 	.statfs        = ubifs_statfs,
1847 	.dirty_inode   = ubifs_dirty_inode,
1848 	.remount_fs    = ubifs_remount_fs,
1849 	.show_options  = ubifs_show_options,
1850 	.sync_fs       = ubifs_sync_fs,
1851 };
1852 
1853 /**
1854  * open_ubi - parse UBI device name string and open the UBI device.
1855  * @name: UBI volume name
1856  * @mode: UBI volume open mode
1857  *
1858  * There are several ways to specify UBI volumes when mounting UBIFS:
1859  * o ubiX_Y    - UBI device number X, volume Y;
1860  * o ubiY      - UBI device number 0, volume Y;
1861  * o ubiX:NAME - mount UBI device X, volume with name NAME;
1862  * o ubi:NAME  - mount UBI device 0, volume with name NAME.
1863  *
1864  * Alternative '!' separator may be used instead of ':' (because some shells
1865  * like busybox may interpret ':' as an NFS host name separator). This function
1866  * returns ubi volume object in case of success and a negative error code in
1867  * case of failure.
1868  */
1869 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1870 {
1871 	int dev, vol;
1872 	char *endptr;
1873 
1874 	if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1875 		return ERR_PTR(-EINVAL);
1876 
1877 	/* ubi:NAME method */
1878 	if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1879 		return ubi_open_volume_nm(0, name + 4, mode);
1880 
1881 	if (!isdigit(name[3]))
1882 		return ERR_PTR(-EINVAL);
1883 
1884 	dev = simple_strtoul(name + 3, &endptr, 0);
1885 
1886 	/* ubiY method */
1887 	if (*endptr == '\0')
1888 		return ubi_open_volume(0, dev, mode);
1889 
1890 	/* ubiX_Y method */
1891 	if (*endptr == '_' && isdigit(endptr[1])) {
1892 		vol = simple_strtoul(endptr + 1, &endptr, 0);
1893 		if (*endptr != '\0')
1894 			return ERR_PTR(-EINVAL);
1895 		return ubi_open_volume(dev, vol, mode);
1896 	}
1897 
1898 	/* ubiX:NAME method */
1899 	if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1900 		return ubi_open_volume_nm(dev, ++endptr, mode);
1901 
1902 	return ERR_PTR(-EINVAL);
1903 }
1904 
1905 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1906 {
1907 	struct ubi_volume_desc *ubi = sb->s_fs_info;
1908 	struct ubifs_info *c;
1909 	struct inode *root;
1910 	int err;
1911 
1912 	c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1913 	if (!c)
1914 		return -ENOMEM;
1915 
1916 	spin_lock_init(&c->cnt_lock);
1917 	spin_lock_init(&c->cs_lock);
1918 	spin_lock_init(&c->buds_lock);
1919 	spin_lock_init(&c->space_lock);
1920 	spin_lock_init(&c->orphan_lock);
1921 	init_rwsem(&c->commit_sem);
1922 	mutex_init(&c->lp_mutex);
1923 	mutex_init(&c->tnc_mutex);
1924 	mutex_init(&c->log_mutex);
1925 	mutex_init(&c->mst_mutex);
1926 	mutex_init(&c->umount_mutex);
1927 	mutex_init(&c->bu_mutex);
1928 	init_waitqueue_head(&c->cmt_wq);
1929 	c->buds = RB_ROOT;
1930 	c->old_idx = RB_ROOT;
1931 	c->size_tree = RB_ROOT;
1932 	c->orph_tree = RB_ROOT;
1933 	INIT_LIST_HEAD(&c->infos_list);
1934 	INIT_LIST_HEAD(&c->idx_gc);
1935 	INIT_LIST_HEAD(&c->replay_list);
1936 	INIT_LIST_HEAD(&c->replay_buds);
1937 	INIT_LIST_HEAD(&c->uncat_list);
1938 	INIT_LIST_HEAD(&c->empty_list);
1939 	INIT_LIST_HEAD(&c->freeable_list);
1940 	INIT_LIST_HEAD(&c->frdi_idx_list);
1941 	INIT_LIST_HEAD(&c->unclean_leb_list);
1942 	INIT_LIST_HEAD(&c->old_buds);
1943 	INIT_LIST_HEAD(&c->orph_list);
1944 	INIT_LIST_HEAD(&c->orph_new);
1945 
1946 	c->vfs_sb = sb;
1947 	c->highest_inum = UBIFS_FIRST_INO;
1948 	c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1949 
1950 	ubi_get_volume_info(ubi, &c->vi);
1951 	ubi_get_device_info(c->vi.ubi_num, &c->di);
1952 
1953 	/* Re-open the UBI device in read-write mode */
1954 	c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1955 	if (IS_ERR(c->ubi)) {
1956 		err = PTR_ERR(c->ubi);
1957 		goto out_free;
1958 	}
1959 
1960 	/*
1961 	 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1962 	 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1963 	 * which means the user would have to wait not just for their own I/O
1964 	 * but the read-ahead I/O as well i.e. completely pointless.
1965 	 *
1966 	 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1967 	 */
1968 	c->bdi.name = "ubifs",
1969 	c->bdi.capabilities = BDI_CAP_MAP_COPY;
1970 	c->bdi.unplug_io_fn = default_unplug_io_fn;
1971 	err  = bdi_init(&c->bdi);
1972 	if (err)
1973 		goto out_close;
1974 	err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
1975 			   c->vi.ubi_num, c->vi.vol_id);
1976 	if (err)
1977 		goto out_bdi;
1978 
1979 	err = ubifs_parse_options(c, data, 0);
1980 	if (err)
1981 		goto out_bdi;
1982 
1983 	sb->s_bdi = &c->bdi;
1984 	sb->s_fs_info = c;
1985 	sb->s_magic = UBIFS_SUPER_MAGIC;
1986 	sb->s_blocksize = UBIFS_BLOCK_SIZE;
1987 	sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1988 	sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1989 	if (c->max_inode_sz > MAX_LFS_FILESIZE)
1990 		sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1991 	sb->s_op = &ubifs_super_operations;
1992 
1993 	mutex_lock(&c->umount_mutex);
1994 	err = mount_ubifs(c);
1995 	if (err) {
1996 		ubifs_assert(err < 0);
1997 		goto out_unlock;
1998 	}
1999 
2000 	/* Read the root inode */
2001 	root = ubifs_iget(sb, UBIFS_ROOT_INO);
2002 	if (IS_ERR(root)) {
2003 		err = PTR_ERR(root);
2004 		goto out_umount;
2005 	}
2006 
2007 	sb->s_root = d_alloc_root(root);
2008 	if (!sb->s_root)
2009 		goto out_iput;
2010 
2011 	mutex_unlock(&c->umount_mutex);
2012 	return 0;
2013 
2014 out_iput:
2015 	iput(root);
2016 out_umount:
2017 	ubifs_umount(c);
2018 out_unlock:
2019 	mutex_unlock(&c->umount_mutex);
2020 out_bdi:
2021 	bdi_destroy(&c->bdi);
2022 out_close:
2023 	ubi_close_volume(c->ubi);
2024 out_free:
2025 	kfree(c);
2026 	return err;
2027 }
2028 
2029 static int sb_test(struct super_block *sb, void *data)
2030 {
2031 	dev_t *dev = data;
2032 	struct ubifs_info *c = sb->s_fs_info;
2033 
2034 	return c->vi.cdev == *dev;
2035 }
2036 
2037 static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
2038 			const char *name, void *data, struct vfsmount *mnt)
2039 {
2040 	struct ubi_volume_desc *ubi;
2041 	struct ubi_volume_info vi;
2042 	struct super_block *sb;
2043 	int err;
2044 
2045 	dbg_gen("name %s, flags %#x", name, flags);
2046 
2047 	/*
2048 	 * Get UBI device number and volume ID. Mount it read-only so far
2049 	 * because this might be a new mount point, and UBI allows only one
2050 	 * read-write user at a time.
2051 	 */
2052 	ubi = open_ubi(name, UBI_READONLY);
2053 	if (IS_ERR(ubi)) {
2054 		ubifs_err("cannot open \"%s\", error %d",
2055 			  name, (int)PTR_ERR(ubi));
2056 		return PTR_ERR(ubi);
2057 	}
2058 	ubi_get_volume_info(ubi, &vi);
2059 
2060 	dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
2061 
2062 	sb = sget(fs_type, &sb_test, &set_anon_super, &vi.cdev);
2063 	if (IS_ERR(sb)) {
2064 		err = PTR_ERR(sb);
2065 		goto out_close;
2066 	}
2067 
2068 	if (sb->s_root) {
2069 		/* A new mount point for already mounted UBIFS */
2070 		dbg_gen("this ubi volume is already mounted");
2071 		if ((flags ^ sb->s_flags) & MS_RDONLY) {
2072 			err = -EBUSY;
2073 			goto out_deact;
2074 		}
2075 	} else {
2076 		sb->s_flags = flags;
2077 		/*
2078 		 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
2079 		 * replaced by 'c'.
2080 		 */
2081 		sb->s_fs_info = ubi;
2082 		err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2083 		if (err)
2084 			goto out_deact;
2085 		/* We do not support atime */
2086 		sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2087 	}
2088 
2089 	/* 'fill_super()' opens ubi again so we must close it here */
2090 	ubi_close_volume(ubi);
2091 
2092 	simple_set_mnt(mnt, sb);
2093 	return 0;
2094 
2095 out_deact:
2096 	deactivate_locked_super(sb);
2097 out_close:
2098 	ubi_close_volume(ubi);
2099 	return err;
2100 }
2101 
2102 static struct file_system_type ubifs_fs_type = {
2103 	.name    = "ubifs",
2104 	.owner   = THIS_MODULE,
2105 	.get_sb  = ubifs_get_sb,
2106 	.kill_sb = kill_anon_super,
2107 };
2108 
2109 /*
2110  * Inode slab cache constructor.
2111  */
2112 static void inode_slab_ctor(void *obj)
2113 {
2114 	struct ubifs_inode *ui = obj;
2115 	inode_init_once(&ui->vfs_inode);
2116 }
2117 
2118 static int __init ubifs_init(void)
2119 {
2120 	int err;
2121 
2122 	BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2123 
2124 	/* Make sure node sizes are 8-byte aligned */
2125 	BUILD_BUG_ON(UBIFS_CH_SZ        & 7);
2126 	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  & 7);
2127 	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2128 	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2129 	BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2130 	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2131 	BUILD_BUG_ON(UBIFS_SB_NODE_SZ   & 7);
2132 	BUILD_BUG_ON(UBIFS_MST_NODE_SZ  & 7);
2133 	BUILD_BUG_ON(UBIFS_REF_NODE_SZ  & 7);
2134 	BUILD_BUG_ON(UBIFS_CS_NODE_SZ   & 7);
2135 	BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2136 
2137 	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2138 	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2139 	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2140 	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  & 7);
2141 	BUILD_BUG_ON(UBIFS_MAX_NODE_SZ      & 7);
2142 	BUILD_BUG_ON(MIN_WRITE_SZ           & 7);
2143 
2144 	/* Check min. node size */
2145 	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  < MIN_WRITE_SZ);
2146 	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2147 	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2148 	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2149 
2150 	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2151 	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2152 	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2153 	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  > UBIFS_MAX_NODE_SZ);
2154 
2155 	/* Defined node sizes */
2156 	BUILD_BUG_ON(UBIFS_SB_NODE_SZ  != 4096);
2157 	BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2158 	BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2159 	BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2160 
2161 	/*
2162 	 * We use 2 bit wide bit-fields to store compression type, which should
2163 	 * be amended if more compressors are added. The bit-fields are:
2164 	 * @compr_type in 'struct ubifs_inode', @default_compr in
2165 	 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2166 	 */
2167 	BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2168 
2169 	/*
2170 	 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2171 	 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2172 	 */
2173 	if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2174 		ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2175 			  " at least 4096 bytes",
2176 			  (unsigned int)PAGE_CACHE_SIZE);
2177 		return -EINVAL;
2178 	}
2179 
2180 	err = register_filesystem(&ubifs_fs_type);
2181 	if (err) {
2182 		ubifs_err("cannot register file system, error %d", err);
2183 		return err;
2184 	}
2185 
2186 	err = -ENOMEM;
2187 	ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2188 				sizeof(struct ubifs_inode), 0,
2189 				SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2190 				&inode_slab_ctor);
2191 	if (!ubifs_inode_slab)
2192 		goto out_reg;
2193 
2194 	register_shrinker(&ubifs_shrinker_info);
2195 
2196 	err = ubifs_compressors_init();
2197 	if (err)
2198 		goto out_shrinker;
2199 
2200 	err = dbg_debugfs_init();
2201 	if (err)
2202 		goto out_compr;
2203 
2204 	return 0;
2205 
2206 out_compr:
2207 	ubifs_compressors_exit();
2208 out_shrinker:
2209 	unregister_shrinker(&ubifs_shrinker_info);
2210 	kmem_cache_destroy(ubifs_inode_slab);
2211 out_reg:
2212 	unregister_filesystem(&ubifs_fs_type);
2213 	return err;
2214 }
2215 /* late_initcall to let compressors initialize first */
2216 late_initcall(ubifs_init);
2217 
2218 static void __exit ubifs_exit(void)
2219 {
2220 	ubifs_assert(list_empty(&ubifs_infos));
2221 	ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2222 
2223 	dbg_debugfs_exit();
2224 	ubifs_compressors_exit();
2225 	unregister_shrinker(&ubifs_shrinker_info);
2226 	kmem_cache_destroy(ubifs_inode_slab);
2227 	unregister_filesystem(&ubifs_fs_type);
2228 }
2229 module_exit(ubifs_exit);
2230 
2231 MODULE_LICENSE("GPL");
2232 MODULE_VERSION(__stringify(UBIFS_VERSION));
2233 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2234 MODULE_DESCRIPTION("UBIFS - UBI File System");
2235