1/*
2 * This file is part of UBIFS.
3 *
4 * Copyright (C) 2006-2008 Nokia Corporation.
5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License version 2 as published by
8 * the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13 * more details.
14 *
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc., 51
17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18 *
19 * Authors: Artem Bityutskiy (Битюцкий Артём)
20 *          Adrian Hunter
21 */
22
23/*
24 * This file implements UBIFS initialization and VFS superblock operations. Some
25 * initialization stuff which is rather large and complex is placed at
26 * corresponding subsystems, but most of it is here.
27 */
28
29#include <linux/init.h>
30#include <linux/slab.h>
31#include <linux/module.h>
32#include <linux/ctype.h>
33#include <linux/kthread.h>
34#include <linux/parser.h>
35#include <linux/seq_file.h>
36#include <linux/mount.h>
37#include <linux/math64.h>
38#include <linux/writeback.h>
39#include "ubifs.h"
40
41/*
42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
43 * allocating too much.
44 */
45#define UBIFS_KMALLOC_OK (128*1024)
46
47/* Slab cache for UBIFS inodes */
48struct kmem_cache *ubifs_inode_slab;
49
50/* UBIFS TNC shrinker description */
51static struct shrinker ubifs_shrinker_info = {
52	.scan_objects = ubifs_shrink_scan,
53	.count_objects = ubifs_shrink_count,
54	.seeks = DEFAULT_SEEKS,
55};
56
57/**
58 * validate_inode - validate inode.
59 * @c: UBIFS file-system description object
60 * @inode: the inode to validate
61 *
62 * This is a helper function for 'ubifs_iget()' which validates various fields
63 * of a newly built inode to make sure they contain sane values and prevent
64 * possible vulnerabilities. Returns zero if the inode is all right and
65 * a non-zero error code if not.
66 */
67static int validate_inode(struct ubifs_info *c, const struct inode *inode)
68{
69	int err;
70	const struct ubifs_inode *ui = ubifs_inode(inode);
71
72	if (inode->i_size > c->max_inode_sz) {
73		ubifs_err(c, "inode is too large (%lld)",
74			  (long long)inode->i_size);
75		return 1;
76	}
77
78	if (ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
79		ubifs_err(c, "unknown compression type %d", ui->compr_type);
80		return 2;
81	}
82
83	if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
84		return 3;
85
86	if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
87		return 4;
88
89	if (ui->xattr && !S_ISREG(inode->i_mode))
90		return 5;
91
92	if (!ubifs_compr_present(ui->compr_type)) {
93		ubifs_warn(c, "inode %lu uses '%s' compression, but it was not compiled in",
94			   inode->i_ino, ubifs_compr_name(ui->compr_type));
95	}
96
97	err = dbg_check_dir(c, inode);
98	return err;
99}
100
101struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
102{
103	int err;
104	union ubifs_key key;
105	struct ubifs_ino_node *ino;
106	struct ubifs_info *c = sb->s_fs_info;
107	struct inode *inode;
108	struct ubifs_inode *ui;
109
110	dbg_gen("inode %lu", inum);
111
112	inode = iget_locked(sb, inum);
113	if (!inode)
114		return ERR_PTR(-ENOMEM);
115	if (!(inode->i_state & I_NEW))
116		return inode;
117	ui = ubifs_inode(inode);
118
119	ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
120	if (!ino) {
121		err = -ENOMEM;
122		goto out;
123	}
124
125	ino_key_init(c, &key, inode->i_ino);
126
127	err = ubifs_tnc_lookup(c, &key, ino);
128	if (err)
129		goto out_ino;
130
131	inode->i_flags |= (S_NOCMTIME | S_NOATIME);
132	set_nlink(inode, le32_to_cpu(ino->nlink));
133	i_uid_write(inode, le32_to_cpu(ino->uid));
134	i_gid_write(inode, le32_to_cpu(ino->gid));
135	inode->i_atime.tv_sec  = (int64_t)le64_to_cpu(ino->atime_sec);
136	inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
137	inode->i_mtime.tv_sec  = (int64_t)le64_to_cpu(ino->mtime_sec);
138	inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
139	inode->i_ctime.tv_sec  = (int64_t)le64_to_cpu(ino->ctime_sec);
140	inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
141	inode->i_mode = le32_to_cpu(ino->mode);
142	inode->i_size = le64_to_cpu(ino->size);
143
144	ui->data_len    = le32_to_cpu(ino->data_len);
145	ui->flags       = le32_to_cpu(ino->flags);
146	ui->compr_type  = le16_to_cpu(ino->compr_type);
147	ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
148	ui->xattr_cnt   = le32_to_cpu(ino->xattr_cnt);
149	ui->xattr_size  = le32_to_cpu(ino->xattr_size);
150	ui->xattr_names = le32_to_cpu(ino->xattr_names);
151	ui->synced_i_size = ui->ui_size = inode->i_size;
152
153	ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
154
155	err = validate_inode(c, inode);
156	if (err)
157		goto out_invalid;
158
159	switch (inode->i_mode & S_IFMT) {
160	case S_IFREG:
161		inode->i_mapping->a_ops = &ubifs_file_address_operations;
162		inode->i_op = &ubifs_file_inode_operations;
163		inode->i_fop = &ubifs_file_operations;
164		if (ui->xattr) {
165			ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
166			if (!ui->data) {
167				err = -ENOMEM;
168				goto out_ino;
169			}
170			memcpy(ui->data, ino->data, ui->data_len);
171			((char *)ui->data)[ui->data_len] = '\0';
172		} else if (ui->data_len != 0) {
173			err = 10;
174			goto out_invalid;
175		}
176		break;
177	case S_IFDIR:
178		inode->i_op  = &ubifs_dir_inode_operations;
179		inode->i_fop = &ubifs_dir_operations;
180		if (ui->data_len != 0) {
181			err = 11;
182			goto out_invalid;
183		}
184		break;
185	case S_IFLNK:
186		inode->i_op = &ubifs_symlink_inode_operations;
187		if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
188			err = 12;
189			goto out_invalid;
190		}
191		ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
192		if (!ui->data) {
193			err = -ENOMEM;
194			goto out_ino;
195		}
196		memcpy(ui->data, ino->data, ui->data_len);
197		((char *)ui->data)[ui->data_len] = '\0';
198		break;
199	case S_IFBLK:
200	case S_IFCHR:
201	{
202		dev_t rdev;
203		union ubifs_dev_desc *dev;
204
205		ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
206		if (!ui->data) {
207			err = -ENOMEM;
208			goto out_ino;
209		}
210
211		dev = (union ubifs_dev_desc *)ino->data;
212		if (ui->data_len == sizeof(dev->new))
213			rdev = new_decode_dev(le32_to_cpu(dev->new));
214		else if (ui->data_len == sizeof(dev->huge))
215			rdev = huge_decode_dev(le64_to_cpu(dev->huge));
216		else {
217			err = 13;
218			goto out_invalid;
219		}
220		memcpy(ui->data, ino->data, ui->data_len);
221		inode->i_op = &ubifs_file_inode_operations;
222		init_special_inode(inode, inode->i_mode, rdev);
223		break;
224	}
225	case S_IFSOCK:
226	case S_IFIFO:
227		inode->i_op = &ubifs_file_inode_operations;
228		init_special_inode(inode, inode->i_mode, 0);
229		if (ui->data_len != 0) {
230			err = 14;
231			goto out_invalid;
232		}
233		break;
234	default:
235		err = 15;
236		goto out_invalid;
237	}
238
239	kfree(ino);
240	ubifs_set_inode_flags(inode);
241	unlock_new_inode(inode);
242	return inode;
243
244out_invalid:
245	ubifs_err(c, "inode %lu validation failed, error %d", inode->i_ino, err);
246	ubifs_dump_node(c, ino);
247	ubifs_dump_inode(c, inode);
248	err = -EINVAL;
249out_ino:
250	kfree(ino);
251out:
252	ubifs_err(c, "failed to read inode %lu, error %d", inode->i_ino, err);
253	iget_failed(inode);
254	return ERR_PTR(err);
255}
256
257static struct inode *ubifs_alloc_inode(struct super_block *sb)
258{
259	struct ubifs_inode *ui;
260
261	ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
262	if (!ui)
263		return NULL;
264
265	memset((void *)ui + sizeof(struct inode), 0,
266	       sizeof(struct ubifs_inode) - sizeof(struct inode));
267	mutex_init(&ui->ui_mutex);
268	spin_lock_init(&ui->ui_lock);
269	return &ui->vfs_inode;
270};
271
272static void ubifs_i_callback(struct rcu_head *head)
273{
274	struct inode *inode = container_of(head, struct inode, i_rcu);
275	struct ubifs_inode *ui = ubifs_inode(inode);
276	kmem_cache_free(ubifs_inode_slab, ui);
277}
278
279static void ubifs_destroy_inode(struct inode *inode)
280{
281	struct ubifs_inode *ui = ubifs_inode(inode);
282
283	kfree(ui->data);
284	call_rcu(&inode->i_rcu, ubifs_i_callback);
285}
286
287/*
288 * Note, Linux write-back code calls this without 'i_mutex'.
289 */
290static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
291{
292	int err = 0;
293	struct ubifs_info *c = inode->i_sb->s_fs_info;
294	struct ubifs_inode *ui = ubifs_inode(inode);
295
296	ubifs_assert(!ui->xattr);
297	if (is_bad_inode(inode))
298		return 0;
299
300	mutex_lock(&ui->ui_mutex);
301	/*
302	 * Due to races between write-back forced by budgeting
303	 * (see 'sync_some_inodes()') and background write-back, the inode may
304	 * have already been synchronized, do not do this again. This might
305	 * also happen if it was synchronized in an VFS operation, e.g.
306	 * 'ubifs_link()'.
307	 */
308	if (!ui->dirty) {
309		mutex_unlock(&ui->ui_mutex);
310		return 0;
311	}
312
313	/*
314	 * As an optimization, do not write orphan inodes to the media just
315	 * because this is not needed.
316	 */
317	dbg_gen("inode %lu, mode %#x, nlink %u",
318		inode->i_ino, (int)inode->i_mode, inode->i_nlink);
319	if (inode->i_nlink) {
320		err = ubifs_jnl_write_inode(c, inode);
321		if (err)
322			ubifs_err(c, "can't write inode %lu, error %d",
323				  inode->i_ino, err);
324		else
325			err = dbg_check_inode_size(c, inode, ui->ui_size);
326	}
327
328	ui->dirty = 0;
329	mutex_unlock(&ui->ui_mutex);
330	ubifs_release_dirty_inode_budget(c, ui);
331	return err;
332}
333
334static void ubifs_evict_inode(struct inode *inode)
335{
336	int err;
337	struct ubifs_info *c = inode->i_sb->s_fs_info;
338	struct ubifs_inode *ui = ubifs_inode(inode);
339
340	if (ui->xattr)
341		/*
342		 * Extended attribute inode deletions are fully handled in
343		 * 'ubifs_removexattr()'. These inodes are special and have
344		 * limited usage, so there is nothing to do here.
345		 */
346		goto out;
347
348	dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
349	ubifs_assert(!atomic_read(&inode->i_count));
350
351	truncate_inode_pages_final(&inode->i_data);
352
353	if (inode->i_nlink)
354		goto done;
355
356	if (is_bad_inode(inode))
357		goto out;
358
359	ui->ui_size = inode->i_size = 0;
360	err = ubifs_jnl_delete_inode(c, inode);
361	if (err)
362		/*
363		 * Worst case we have a lost orphan inode wasting space, so a
364		 * simple error message is OK here.
365		 */
366		ubifs_err(c, "can't delete inode %lu, error %d",
367			  inode->i_ino, err);
368
369out:
370	if (ui->dirty)
371		ubifs_release_dirty_inode_budget(c, ui);
372	else {
373		/* We've deleted something - clean the "no space" flags */
374		c->bi.nospace = c->bi.nospace_rp = 0;
375		smp_wmb();
376	}
377done:
378	clear_inode(inode);
379}
380
381static void ubifs_dirty_inode(struct inode *inode, int flags)
382{
383	struct ubifs_inode *ui = ubifs_inode(inode);
384
385	ubifs_assert(mutex_is_locked(&ui->ui_mutex));
386	if (!ui->dirty) {
387		ui->dirty = 1;
388		dbg_gen("inode %lu",  inode->i_ino);
389	}
390}
391
392static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
393{
394	struct ubifs_info *c = dentry->d_sb->s_fs_info;
395	unsigned long long free;
396	__le32 *uuid = (__le32 *)c->uuid;
397
398	free = ubifs_get_free_space(c);
399	dbg_gen("free space %lld bytes (%lld blocks)",
400		free, free >> UBIFS_BLOCK_SHIFT);
401
402	buf->f_type = UBIFS_SUPER_MAGIC;
403	buf->f_bsize = UBIFS_BLOCK_SIZE;
404	buf->f_blocks = c->block_cnt;
405	buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
406	if (free > c->report_rp_size)
407		buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
408	else
409		buf->f_bavail = 0;
410	buf->f_files = 0;
411	buf->f_ffree = 0;
412	buf->f_namelen = UBIFS_MAX_NLEN;
413	buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
414	buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
415	ubifs_assert(buf->f_bfree <= c->block_cnt);
416	return 0;
417}
418
419static int ubifs_show_options(struct seq_file *s, struct dentry *root)
420{
421	struct ubifs_info *c = root->d_sb->s_fs_info;
422
423	if (c->mount_opts.unmount_mode == 2)
424		seq_puts(s, ",fast_unmount");
425	else if (c->mount_opts.unmount_mode == 1)
426		seq_puts(s, ",norm_unmount");
427
428	if (c->mount_opts.bulk_read == 2)
429		seq_puts(s, ",bulk_read");
430	else if (c->mount_opts.bulk_read == 1)
431		seq_puts(s, ",no_bulk_read");
432
433	if (c->mount_opts.chk_data_crc == 2)
434		seq_puts(s, ",chk_data_crc");
435	else if (c->mount_opts.chk_data_crc == 1)
436		seq_puts(s, ",no_chk_data_crc");
437
438	if (c->mount_opts.override_compr) {
439		seq_printf(s, ",compr=%s",
440			   ubifs_compr_name(c->mount_opts.compr_type));
441	}
442
443	return 0;
444}
445
446static int ubifs_sync_fs(struct super_block *sb, int wait)
447{
448	int i, err;
449	struct ubifs_info *c = sb->s_fs_info;
450
451	/*
452	 * Zero @wait is just an advisory thing to help the file system shove
453	 * lots of data into the queues, and there will be the second
454	 * '->sync_fs()' call, with non-zero @wait.
455	 */
456	if (!wait)
457		return 0;
458
459	/*
460	 * Synchronize write buffers, because 'ubifs_run_commit()' does not
461	 * do this if it waits for an already running commit.
462	 */
463	for (i = 0; i < c->jhead_cnt; i++) {
464		err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
465		if (err)
466			return err;
467	}
468
469	/*
470	 * Strictly speaking, it is not necessary to commit the journal here,
471	 * synchronizing write-buffers would be enough. But committing makes
472	 * UBIFS free space predictions much more accurate, so we want to let
473	 * the user be able to get more accurate results of 'statfs()' after
474	 * they synchronize the file system.
475	 */
476	err = ubifs_run_commit(c);
477	if (err)
478		return err;
479
480	return ubi_sync(c->vi.ubi_num);
481}
482
483/**
484 * init_constants_early - initialize UBIFS constants.
485 * @c: UBIFS file-system description object
486 *
487 * This function initialize UBIFS constants which do not need the superblock to
488 * be read. It also checks that the UBI volume satisfies basic UBIFS
489 * requirements. Returns zero in case of success and a negative error code in
490 * case of failure.
491 */
492static int init_constants_early(struct ubifs_info *c)
493{
494	if (c->vi.corrupted) {
495		ubifs_warn(c, "UBI volume is corrupted - read-only mode");
496		c->ro_media = 1;
497	}
498
499	if (c->di.ro_mode) {
500		ubifs_msg(c, "read-only UBI device");
501		c->ro_media = 1;
502	}
503
504	if (c->vi.vol_type == UBI_STATIC_VOLUME) {
505		ubifs_msg(c, "static UBI volume - read-only mode");
506		c->ro_media = 1;
507	}
508
509	c->leb_cnt = c->vi.size;
510	c->leb_size = c->vi.usable_leb_size;
511	c->leb_start = c->di.leb_start;
512	c->half_leb_size = c->leb_size / 2;
513	c->min_io_size = c->di.min_io_size;
514	c->min_io_shift = fls(c->min_io_size) - 1;
515	c->max_write_size = c->di.max_write_size;
516	c->max_write_shift = fls(c->max_write_size) - 1;
517
518	if (c->leb_size < UBIFS_MIN_LEB_SZ) {
519		ubifs_err(c, "too small LEBs (%d bytes), min. is %d bytes",
520			  c->leb_size, UBIFS_MIN_LEB_SZ);
521		return -EINVAL;
522	}
523
524	if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
525		ubifs_err(c, "too few LEBs (%d), min. is %d",
526			  c->leb_cnt, UBIFS_MIN_LEB_CNT);
527		return -EINVAL;
528	}
529
530	if (!is_power_of_2(c->min_io_size)) {
531		ubifs_err(c, "bad min. I/O size %d", c->min_io_size);
532		return -EINVAL;
533	}
534
535	/*
536	 * Maximum write size has to be greater or equivalent to min. I/O
537	 * size, and be multiple of min. I/O size.
538	 */
539	if (c->max_write_size < c->min_io_size ||
540	    c->max_write_size % c->min_io_size ||
541	    !is_power_of_2(c->max_write_size)) {
542		ubifs_err(c, "bad write buffer size %d for %d min. I/O unit",
543			  c->max_write_size, c->min_io_size);
544		return -EINVAL;
545	}
546
547	/*
548	 * UBIFS aligns all node to 8-byte boundary, so to make function in
549	 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
550	 * less than 8.
551	 */
552	if (c->min_io_size < 8) {
553		c->min_io_size = 8;
554		c->min_io_shift = 3;
555		if (c->max_write_size < c->min_io_size) {
556			c->max_write_size = c->min_io_size;
557			c->max_write_shift = c->min_io_shift;
558		}
559	}
560
561	c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
562	c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
563
564	/*
565	 * Initialize node length ranges which are mostly needed for node
566	 * length validation.
567	 */
568	c->ranges[UBIFS_PAD_NODE].len  = UBIFS_PAD_NODE_SZ;
569	c->ranges[UBIFS_SB_NODE].len   = UBIFS_SB_NODE_SZ;
570	c->ranges[UBIFS_MST_NODE].len  = UBIFS_MST_NODE_SZ;
571	c->ranges[UBIFS_REF_NODE].len  = UBIFS_REF_NODE_SZ;
572	c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
573	c->ranges[UBIFS_CS_NODE].len   = UBIFS_CS_NODE_SZ;
574
575	c->ranges[UBIFS_INO_NODE].min_len  = UBIFS_INO_NODE_SZ;
576	c->ranges[UBIFS_INO_NODE].max_len  = UBIFS_MAX_INO_NODE_SZ;
577	c->ranges[UBIFS_ORPH_NODE].min_len =
578				UBIFS_ORPH_NODE_SZ + sizeof(__le64);
579	c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
580	c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
581	c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
582	c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
583	c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
584	c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
585	c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
586	/*
587	 * Minimum indexing node size is amended later when superblock is
588	 * read and the key length is known.
589	 */
590	c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
591	/*
592	 * Maximum indexing node size is amended later when superblock is
593	 * read and the fanout is known.
594	 */
595	c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
596
597	/*
598	 * Initialize dead and dark LEB space watermarks. See gc.c for comments
599	 * about these values.
600	 */
601	c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
602	c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
603
604	/*
605	 * Calculate how many bytes would be wasted at the end of LEB if it was
606	 * fully filled with data nodes of maximum size. This is used in
607	 * calculations when reporting free space.
608	 */
609	c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
610
611	/* Buffer size for bulk-reads */
612	c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
613	if (c->max_bu_buf_len > c->leb_size)
614		c->max_bu_buf_len = c->leb_size;
615	return 0;
616}
617
618/**
619 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
620 * @c: UBIFS file-system description object
621 * @lnum: LEB the write-buffer was synchronized to
622 * @free: how many free bytes left in this LEB
623 * @pad: how many bytes were padded
624 *
625 * This is a callback function which is called by the I/O unit when the
626 * write-buffer is synchronized. We need this to correctly maintain space
627 * accounting in bud logical eraseblocks. This function returns zero in case of
628 * success and a negative error code in case of failure.
629 *
630 * This function actually belongs to the journal, but we keep it here because
631 * we want to keep it static.
632 */
633static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
634{
635	return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
636}
637
638/*
639 * init_constants_sb - initialize UBIFS constants.
640 * @c: UBIFS file-system description object
641 *
642 * This is a helper function which initializes various UBIFS constants after
643 * the superblock has been read. It also checks various UBIFS parameters and
644 * makes sure they are all right. Returns zero in case of success and a
645 * negative error code in case of failure.
646 */
647static int init_constants_sb(struct ubifs_info *c)
648{
649	int tmp, err;
650	long long tmp64;
651
652	c->main_bytes = (long long)c->main_lebs * c->leb_size;
653	c->max_znode_sz = sizeof(struct ubifs_znode) +
654				c->fanout * sizeof(struct ubifs_zbranch);
655
656	tmp = ubifs_idx_node_sz(c, 1);
657	c->ranges[UBIFS_IDX_NODE].min_len = tmp;
658	c->min_idx_node_sz = ALIGN(tmp, 8);
659
660	tmp = ubifs_idx_node_sz(c, c->fanout);
661	c->ranges[UBIFS_IDX_NODE].max_len = tmp;
662	c->max_idx_node_sz = ALIGN(tmp, 8);
663
664	/* Make sure LEB size is large enough to fit full commit */
665	tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
666	tmp = ALIGN(tmp, c->min_io_size);
667	if (tmp > c->leb_size) {
668		ubifs_err(c, "too small LEB size %d, at least %d needed",
669			  c->leb_size, tmp);
670		return -EINVAL;
671	}
672
673	/*
674	 * Make sure that the log is large enough to fit reference nodes for
675	 * all buds plus one reserved LEB.
676	 */
677	tmp64 = c->max_bud_bytes + c->leb_size - 1;
678	c->max_bud_cnt = div_u64(tmp64, c->leb_size);
679	tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
680	tmp /= c->leb_size;
681	tmp += 1;
682	if (c->log_lebs < tmp) {
683		ubifs_err(c, "too small log %d LEBs, required min. %d LEBs",
684			  c->log_lebs, tmp);
685		return -EINVAL;
686	}
687
688	/*
689	 * When budgeting we assume worst-case scenarios when the pages are not
690	 * be compressed and direntries are of the maximum size.
691	 *
692	 * Note, data, which may be stored in inodes is budgeted separately, so
693	 * it is not included into 'c->bi.inode_budget'.
694	 */
695	c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
696	c->bi.inode_budget = UBIFS_INO_NODE_SZ;
697	c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
698
699	/*
700	 * When the amount of flash space used by buds becomes
701	 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
702	 * The writers are unblocked when the commit is finished. To avoid
703	 * writers to be blocked UBIFS initiates background commit in advance,
704	 * when number of bud bytes becomes above the limit defined below.
705	 */
706	c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
707
708	/*
709	 * Ensure minimum journal size. All the bytes in the journal heads are
710	 * considered to be used, when calculating the current journal usage.
711	 * Consequently, if the journal is too small, UBIFS will treat it as
712	 * always full.
713	 */
714	tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
715	if (c->bg_bud_bytes < tmp64)
716		c->bg_bud_bytes = tmp64;
717	if (c->max_bud_bytes < tmp64 + c->leb_size)
718		c->max_bud_bytes = tmp64 + c->leb_size;
719
720	err = ubifs_calc_lpt_geom(c);
721	if (err)
722		return err;
723
724	/* Initialize effective LEB size used in budgeting calculations */
725	c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
726	return 0;
727}
728
729/*
730 * init_constants_master - initialize UBIFS constants.
731 * @c: UBIFS file-system description object
732 *
733 * This is a helper function which initializes various UBIFS constants after
734 * the master node has been read. It also checks various UBIFS parameters and
735 * makes sure they are all right.
736 */
737static void init_constants_master(struct ubifs_info *c)
738{
739	long long tmp64;
740
741	c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
742	c->report_rp_size = ubifs_reported_space(c, c->rp_size);
743
744	/*
745	 * Calculate total amount of FS blocks. This number is not used
746	 * internally because it does not make much sense for UBIFS, but it is
747	 * necessary to report something for the 'statfs()' call.
748	 *
749	 * Subtract the LEB reserved for GC, the LEB which is reserved for
750	 * deletions, minimum LEBs for the index, and assume only one journal
751	 * head is available.
752	 */
753	tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
754	tmp64 *= (long long)c->leb_size - c->leb_overhead;
755	tmp64 = ubifs_reported_space(c, tmp64);
756	c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
757}
758
759/**
760 * take_gc_lnum - reserve GC LEB.
761 * @c: UBIFS file-system description object
762 *
763 * This function ensures that the LEB reserved for garbage collection is marked
764 * as "taken" in lprops. We also have to set free space to LEB size and dirty
765 * space to zero, because lprops may contain out-of-date information if the
766 * file-system was un-mounted before it has been committed. This function
767 * returns zero in case of success and a negative error code in case of
768 * failure.
769 */
770static int take_gc_lnum(struct ubifs_info *c)
771{
772	int err;
773
774	if (c->gc_lnum == -1) {
775		ubifs_err(c, "no LEB for GC");
776		return -EINVAL;
777	}
778
779	/* And we have to tell lprops that this LEB is taken */
780	err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
781				  LPROPS_TAKEN, 0, 0);
782	return err;
783}
784
785/**
786 * alloc_wbufs - allocate write-buffers.
787 * @c: UBIFS file-system description object
788 *
789 * This helper function allocates and initializes UBIFS write-buffers. Returns
790 * zero in case of success and %-ENOMEM in case of failure.
791 */
792static int alloc_wbufs(struct ubifs_info *c)
793{
794	int i, err;
795
796	c->jheads = kcalloc(c->jhead_cnt, sizeof(struct ubifs_jhead),
797			    GFP_KERNEL);
798	if (!c->jheads)
799		return -ENOMEM;
800
801	/* Initialize journal heads */
802	for (i = 0; i < c->jhead_cnt; i++) {
803		INIT_LIST_HEAD(&c->jheads[i].buds_list);
804		err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
805		if (err)
806			return err;
807
808		c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
809		c->jheads[i].wbuf.jhead = i;
810		c->jheads[i].grouped = 1;
811	}
812
813	/*
814	 * Garbage Collector head does not need to be synchronized by timer.
815	 * Also GC head nodes are not grouped.
816	 */
817	c->jheads[GCHD].wbuf.no_timer = 1;
818	c->jheads[GCHD].grouped = 0;
819
820	return 0;
821}
822
823/**
824 * free_wbufs - free write-buffers.
825 * @c: UBIFS file-system description object
826 */
827static void free_wbufs(struct ubifs_info *c)
828{
829	int i;
830
831	if (c->jheads) {
832		for (i = 0; i < c->jhead_cnt; i++) {
833			kfree(c->jheads[i].wbuf.buf);
834			kfree(c->jheads[i].wbuf.inodes);
835		}
836		kfree(c->jheads);
837		c->jheads = NULL;
838	}
839}
840
841/**
842 * free_orphans - free orphans.
843 * @c: UBIFS file-system description object
844 */
845static void free_orphans(struct ubifs_info *c)
846{
847	struct ubifs_orphan *orph;
848
849	while (c->orph_dnext) {
850		orph = c->orph_dnext;
851		c->orph_dnext = orph->dnext;
852		list_del(&orph->list);
853		kfree(orph);
854	}
855
856	while (!list_empty(&c->orph_list)) {
857		orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
858		list_del(&orph->list);
859		kfree(orph);
860		ubifs_err(c, "orphan list not empty at unmount");
861	}
862
863	vfree(c->orph_buf);
864	c->orph_buf = NULL;
865}
866
867/**
868 * free_buds - free per-bud objects.
869 * @c: UBIFS file-system description object
870 */
871static void free_buds(struct ubifs_info *c)
872{
873	struct ubifs_bud *bud, *n;
874
875	rbtree_postorder_for_each_entry_safe(bud, n, &c->buds, rb)
876		kfree(bud);
877}
878
879/**
880 * check_volume_empty - check if the UBI volume is empty.
881 * @c: UBIFS file-system description object
882 *
883 * This function checks if the UBIFS volume is empty by looking if its LEBs are
884 * mapped or not. The result of checking is stored in the @c->empty variable.
885 * Returns zero in case of success and a negative error code in case of
886 * failure.
887 */
888static int check_volume_empty(struct ubifs_info *c)
889{
890	int lnum, err;
891
892	c->empty = 1;
893	for (lnum = 0; lnum < c->leb_cnt; lnum++) {
894		err = ubifs_is_mapped(c, lnum);
895		if (unlikely(err < 0))
896			return err;
897		if (err == 1) {
898			c->empty = 0;
899			break;
900		}
901
902		cond_resched();
903	}
904
905	return 0;
906}
907
908/*
909 * UBIFS mount options.
910 *
911 * Opt_fast_unmount: do not run a journal commit before un-mounting
912 * Opt_norm_unmount: run a journal commit before un-mounting
913 * Opt_bulk_read: enable bulk-reads
914 * Opt_no_bulk_read: disable bulk-reads
915 * Opt_chk_data_crc: check CRCs when reading data nodes
916 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
917 * Opt_override_compr: override default compressor
918 * Opt_err: just end of array marker
919 */
920enum {
921	Opt_fast_unmount,
922	Opt_norm_unmount,
923	Opt_bulk_read,
924	Opt_no_bulk_read,
925	Opt_chk_data_crc,
926	Opt_no_chk_data_crc,
927	Opt_override_compr,
928	Opt_err,
929};
930
931static const match_table_t tokens = {
932	{Opt_fast_unmount, "fast_unmount"},
933	{Opt_norm_unmount, "norm_unmount"},
934	{Opt_bulk_read, "bulk_read"},
935	{Opt_no_bulk_read, "no_bulk_read"},
936	{Opt_chk_data_crc, "chk_data_crc"},
937	{Opt_no_chk_data_crc, "no_chk_data_crc"},
938	{Opt_override_compr, "compr=%s"},
939	{Opt_err, NULL},
940};
941
942/**
943 * parse_standard_option - parse a standard mount option.
944 * @option: the option to parse
945 *
946 * Normally, standard mount options like "sync" are passed to file-systems as
947 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
948 * be present in the options string. This function tries to deal with this
949 * situation and parse standard options. Returns 0 if the option was not
950 * recognized, and the corresponding integer flag if it was.
951 *
952 * UBIFS is only interested in the "sync" option, so do not check for anything
953 * else.
954 */
955static int parse_standard_option(const char *option)
956{
957
958	pr_notice("UBIFS: parse %s\n", option);
959	if (!strcmp(option, "sync"))
960		return MS_SYNCHRONOUS;
961	return 0;
962}
963
964/**
965 * ubifs_parse_options - parse mount parameters.
966 * @c: UBIFS file-system description object
967 * @options: parameters to parse
968 * @is_remount: non-zero if this is FS re-mount
969 *
970 * This function parses UBIFS mount options and returns zero in case success
971 * and a negative error code in case of failure.
972 */
973static int ubifs_parse_options(struct ubifs_info *c, char *options,
974			       int is_remount)
975{
976	char *p;
977	substring_t args[MAX_OPT_ARGS];
978
979	if (!options)
980		return 0;
981
982	while ((p = strsep(&options, ","))) {
983		int token;
984
985		if (!*p)
986			continue;
987
988		token = match_token(p, tokens, args);
989		switch (token) {
990		/*
991		 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
992		 * We accept them in order to be backward-compatible. But this
993		 * should be removed at some point.
994		 */
995		case Opt_fast_unmount:
996			c->mount_opts.unmount_mode = 2;
997			break;
998		case Opt_norm_unmount:
999			c->mount_opts.unmount_mode = 1;
1000			break;
1001		case Opt_bulk_read:
1002			c->mount_opts.bulk_read = 2;
1003			c->bulk_read = 1;
1004			break;
1005		case Opt_no_bulk_read:
1006			c->mount_opts.bulk_read = 1;
1007			c->bulk_read = 0;
1008			break;
1009		case Opt_chk_data_crc:
1010			c->mount_opts.chk_data_crc = 2;
1011			c->no_chk_data_crc = 0;
1012			break;
1013		case Opt_no_chk_data_crc:
1014			c->mount_opts.chk_data_crc = 1;
1015			c->no_chk_data_crc = 1;
1016			break;
1017		case Opt_override_compr:
1018		{
1019			char *name = match_strdup(&args[0]);
1020
1021			if (!name)
1022				return -ENOMEM;
1023			if (!strcmp(name, "none"))
1024				c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1025			else if (!strcmp(name, "lzo"))
1026				c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1027			else if (!strcmp(name, "zlib"))
1028				c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1029			else {
1030				ubifs_err(c, "unknown compressor \"%s\"", name); //FIXME: is c ready?
1031				kfree(name);
1032				return -EINVAL;
1033			}
1034			kfree(name);
1035			c->mount_opts.override_compr = 1;
1036			c->default_compr = c->mount_opts.compr_type;
1037			break;
1038		}
1039		default:
1040		{
1041			unsigned long flag;
1042			struct super_block *sb = c->vfs_sb;
1043
1044			flag = parse_standard_option(p);
1045			if (!flag) {
1046				ubifs_err(c, "unrecognized mount option \"%s\" or missing value",
1047					  p);
1048				return -EINVAL;
1049			}
1050			sb->s_flags |= flag;
1051			break;
1052		}
1053		}
1054	}
1055
1056	return 0;
1057}
1058
1059/**
1060 * destroy_journal - destroy journal data structures.
1061 * @c: UBIFS file-system description object
1062 *
1063 * This function destroys journal data structures including those that may have
1064 * been created by recovery functions.
1065 */
1066static void destroy_journal(struct ubifs_info *c)
1067{
1068	while (!list_empty(&c->unclean_leb_list)) {
1069		struct ubifs_unclean_leb *ucleb;
1070
1071		ucleb = list_entry(c->unclean_leb_list.next,
1072				   struct ubifs_unclean_leb, list);
1073		list_del(&ucleb->list);
1074		kfree(ucleb);
1075	}
1076	while (!list_empty(&c->old_buds)) {
1077		struct ubifs_bud *bud;
1078
1079		bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1080		list_del(&bud->list);
1081		kfree(bud);
1082	}
1083	ubifs_destroy_idx_gc(c);
1084	ubifs_destroy_size_tree(c);
1085	ubifs_tnc_close(c);
1086	free_buds(c);
1087}
1088
1089/**
1090 * bu_init - initialize bulk-read information.
1091 * @c: UBIFS file-system description object
1092 */
1093static void bu_init(struct ubifs_info *c)
1094{
1095	ubifs_assert(c->bulk_read == 1);
1096
1097	if (c->bu.buf)
1098		return; /* Already initialized */
1099
1100again:
1101	c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1102	if (!c->bu.buf) {
1103		if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1104			c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1105			goto again;
1106		}
1107
1108		/* Just disable bulk-read */
1109		ubifs_warn(c, "cannot allocate %d bytes of memory for bulk-read, disabling it",
1110			   c->max_bu_buf_len);
1111		c->mount_opts.bulk_read = 1;
1112		c->bulk_read = 0;
1113		return;
1114	}
1115}
1116
1117/**
1118 * check_free_space - check if there is enough free space to mount.
1119 * @c: UBIFS file-system description object
1120 *
1121 * This function makes sure UBIFS has enough free space to be mounted in
1122 * read/write mode. UBIFS must always have some free space to allow deletions.
1123 */
1124static int check_free_space(struct ubifs_info *c)
1125{
1126	ubifs_assert(c->dark_wm > 0);
1127	if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1128		ubifs_err(c, "insufficient free space to mount in R/W mode");
1129		ubifs_dump_budg(c, &c->bi);
1130		ubifs_dump_lprops(c);
1131		return -ENOSPC;
1132	}
1133	return 0;
1134}
1135
1136/**
1137 * mount_ubifs - mount UBIFS file-system.
1138 * @c: UBIFS file-system description object
1139 *
1140 * This function mounts UBIFS file system. Returns zero in case of success and
1141 * a negative error code in case of failure.
1142 */
1143static int mount_ubifs(struct ubifs_info *c)
1144{
1145	int err;
1146	long long x, y;
1147	size_t sz;
1148
1149	c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1150	/* Suppress error messages while probing if MS_SILENT is set */
1151	c->probing = !!(c->vfs_sb->s_flags & MS_SILENT);
1152
1153	err = init_constants_early(c);
1154	if (err)
1155		return err;
1156
1157	err = ubifs_debugging_init(c);
1158	if (err)
1159		return err;
1160
1161	err = check_volume_empty(c);
1162	if (err)
1163		goto out_free;
1164
1165	if (c->empty && (c->ro_mount || c->ro_media)) {
1166		/*
1167		 * This UBI volume is empty, and read-only, or the file system
1168		 * is mounted read-only - we cannot format it.
1169		 */
1170		ubifs_err(c, "can't format empty UBI volume: read-only %s",
1171			  c->ro_media ? "UBI volume" : "mount");
1172		err = -EROFS;
1173		goto out_free;
1174	}
1175
1176	if (c->ro_media && !c->ro_mount) {
1177		ubifs_err(c, "cannot mount read-write - read-only media");
1178		err = -EROFS;
1179		goto out_free;
1180	}
1181
1182	/*
1183	 * The requirement for the buffer is that it should fit indexing B-tree
1184	 * height amount of integers. We assume the height if the TNC tree will
1185	 * never exceed 64.
1186	 */
1187	err = -ENOMEM;
1188	c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1189	if (!c->bottom_up_buf)
1190		goto out_free;
1191
1192	c->sbuf = vmalloc(c->leb_size);
1193	if (!c->sbuf)
1194		goto out_free;
1195
1196	if (!c->ro_mount) {
1197		c->ileb_buf = vmalloc(c->leb_size);
1198		if (!c->ileb_buf)
1199			goto out_free;
1200	}
1201
1202	if (c->bulk_read == 1)
1203		bu_init(c);
1204
1205	if (!c->ro_mount) {
1206		c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
1207					       GFP_KERNEL);
1208		if (!c->write_reserve_buf)
1209			goto out_free;
1210	}
1211
1212	c->mounting = 1;
1213
1214	err = ubifs_read_superblock(c);
1215	if (err)
1216		goto out_free;
1217
1218	c->probing = 0;
1219
1220	/*
1221	 * Make sure the compressor which is set as default in the superblock
1222	 * or overridden by mount options is actually compiled in.
1223	 */
1224	if (!ubifs_compr_present(c->default_compr)) {
1225		ubifs_err(c, "'compressor \"%s\" is not compiled in",
1226			  ubifs_compr_name(c->default_compr));
1227		err = -ENOTSUPP;
1228		goto out_free;
1229	}
1230
1231	err = init_constants_sb(c);
1232	if (err)
1233		goto out_free;
1234
1235	sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1236	sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1237	c->cbuf = kmalloc(sz, GFP_NOFS);
1238	if (!c->cbuf) {
1239		err = -ENOMEM;
1240		goto out_free;
1241	}
1242
1243	err = alloc_wbufs(c);
1244	if (err)
1245		goto out_cbuf;
1246
1247	sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1248	if (!c->ro_mount) {
1249		/* Create background thread */
1250		c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1251		if (IS_ERR(c->bgt)) {
1252			err = PTR_ERR(c->bgt);
1253			c->bgt = NULL;
1254			ubifs_err(c, "cannot spawn \"%s\", error %d",
1255				  c->bgt_name, err);
1256			goto out_wbufs;
1257		}
1258		wake_up_process(c->bgt);
1259	}
1260
1261	err = ubifs_read_master(c);
1262	if (err)
1263		goto out_master;
1264
1265	init_constants_master(c);
1266
1267	if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1268		ubifs_msg(c, "recovery needed");
1269		c->need_recovery = 1;
1270	}
1271
1272	if (c->need_recovery && !c->ro_mount) {
1273		err = ubifs_recover_inl_heads(c, c->sbuf);
1274		if (err)
1275			goto out_master;
1276	}
1277
1278	err = ubifs_lpt_init(c, 1, !c->ro_mount);
1279	if (err)
1280		goto out_master;
1281
1282	if (!c->ro_mount && c->space_fixup) {
1283		err = ubifs_fixup_free_space(c);
1284		if (err)
1285			goto out_lpt;
1286	}
1287
1288	if (!c->ro_mount && !c->need_recovery) {
1289		/*
1290		 * Set the "dirty" flag so that if we reboot uncleanly we
1291		 * will notice this immediately on the next mount.
1292		 */
1293		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1294		err = ubifs_write_master(c);
1295		if (err)
1296			goto out_lpt;
1297	}
1298
1299	err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1300	if (err)
1301		goto out_lpt;
1302
1303	err = ubifs_replay_journal(c);
1304	if (err)
1305		goto out_journal;
1306
1307	/* Calculate 'min_idx_lebs' after journal replay */
1308	c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1309
1310	err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1311	if (err)
1312		goto out_orphans;
1313
1314	if (!c->ro_mount) {
1315		int lnum;
1316
1317		err = check_free_space(c);
1318		if (err)
1319			goto out_orphans;
1320
1321		/* Check for enough log space */
1322		lnum = c->lhead_lnum + 1;
1323		if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1324			lnum = UBIFS_LOG_LNUM;
1325		if (lnum == c->ltail_lnum) {
1326			err = ubifs_consolidate_log(c);
1327			if (err)
1328				goto out_orphans;
1329		}
1330
1331		if (c->need_recovery) {
1332			err = ubifs_recover_size(c);
1333			if (err)
1334				goto out_orphans;
1335			err = ubifs_rcvry_gc_commit(c);
1336			if (err)
1337				goto out_orphans;
1338		} else {
1339			err = take_gc_lnum(c);
1340			if (err)
1341				goto out_orphans;
1342
1343			/*
1344			 * GC LEB may contain garbage if there was an unclean
1345			 * reboot, and it should be un-mapped.
1346			 */
1347			err = ubifs_leb_unmap(c, c->gc_lnum);
1348			if (err)
1349				goto out_orphans;
1350		}
1351
1352		err = dbg_check_lprops(c);
1353		if (err)
1354			goto out_orphans;
1355	} else if (c->need_recovery) {
1356		err = ubifs_recover_size(c);
1357		if (err)
1358			goto out_orphans;
1359	} else {
1360		/*
1361		 * Even if we mount read-only, we have to set space in GC LEB
1362		 * to proper value because this affects UBIFS free space
1363		 * reporting. We do not want to have a situation when
1364		 * re-mounting from R/O to R/W changes amount of free space.
1365		 */
1366		err = take_gc_lnum(c);
1367		if (err)
1368			goto out_orphans;
1369	}
1370
1371	spin_lock(&ubifs_infos_lock);
1372	list_add_tail(&c->infos_list, &ubifs_infos);
1373	spin_unlock(&ubifs_infos_lock);
1374
1375	if (c->need_recovery) {
1376		if (c->ro_mount)
1377			ubifs_msg(c, "recovery deferred");
1378		else {
1379			c->need_recovery = 0;
1380			ubifs_msg(c, "recovery completed");
1381			/*
1382			 * GC LEB has to be empty and taken at this point. But
1383			 * the journal head LEBs may also be accounted as
1384			 * "empty taken" if they are empty.
1385			 */
1386			ubifs_assert(c->lst.taken_empty_lebs > 0);
1387		}
1388	} else
1389		ubifs_assert(c->lst.taken_empty_lebs > 0);
1390
1391	err = dbg_check_filesystem(c);
1392	if (err)
1393		goto out_infos;
1394
1395	err = dbg_debugfs_init_fs(c);
1396	if (err)
1397		goto out_infos;
1398
1399	c->mounting = 0;
1400
1401	ubifs_msg(c, "UBIFS: mounted UBI device %d, volume %d, name \"%s\"%s",
1402		  c->vi.ubi_num, c->vi.vol_id, c->vi.name,
1403		  c->ro_mount ? ", R/O mode" : "");
1404	x = (long long)c->main_lebs * c->leb_size;
1405	y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1406	ubifs_msg(c, "LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes",
1407		  c->leb_size, c->leb_size >> 10, c->min_io_size,
1408		  c->max_write_size);
1409	ubifs_msg(c, "FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)",
1410		  x, x >> 20, c->main_lebs,
1411		  y, y >> 20, c->log_lebs + c->max_bud_cnt);
1412	ubifs_msg(c, "reserved for root: %llu bytes (%llu KiB)",
1413		  c->report_rp_size, c->report_rp_size >> 10);
1414	ubifs_msg(c, "media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s",
1415		  c->fmt_version, c->ro_compat_version,
1416		  UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid,
1417		  c->big_lpt ? ", big LPT model" : ", small LPT model");
1418
1419	dbg_gen("default compressor:  %s", ubifs_compr_name(c->default_compr));
1420	dbg_gen("data journal heads:  %d",
1421		c->jhead_cnt - NONDATA_JHEADS_CNT);
1422	dbg_gen("log LEBs:            %d (%d - %d)",
1423		c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1424	dbg_gen("LPT area LEBs:       %d (%d - %d)",
1425		c->lpt_lebs, c->lpt_first, c->lpt_last);
1426	dbg_gen("orphan area LEBs:    %d (%d - %d)",
1427		c->orph_lebs, c->orph_first, c->orph_last);
1428	dbg_gen("main area LEBs:      %d (%d - %d)",
1429		c->main_lebs, c->main_first, c->leb_cnt - 1);
1430	dbg_gen("index LEBs:          %d", c->lst.idx_lebs);
1431	dbg_gen("total index bytes:   %lld (%lld KiB, %lld MiB)",
1432		c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1433		c->bi.old_idx_sz >> 20);
1434	dbg_gen("key hash type:       %d", c->key_hash_type);
1435	dbg_gen("tree fanout:         %d", c->fanout);
1436	dbg_gen("reserved GC LEB:     %d", c->gc_lnum);
1437	dbg_gen("max. znode size      %d", c->max_znode_sz);
1438	dbg_gen("max. index node size %d", c->max_idx_node_sz);
1439	dbg_gen("node sizes:          data %zu, inode %zu, dentry %zu",
1440		UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1441	dbg_gen("node sizes:          trun %zu, sb %zu, master %zu",
1442		UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1443	dbg_gen("node sizes:          ref %zu, cmt. start %zu, orph %zu",
1444		UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1445	dbg_gen("max. node sizes:     data %zu, inode %zu dentry %zu, idx %d",
1446		UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1447		UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1448	dbg_gen("dead watermark:      %d", c->dead_wm);
1449	dbg_gen("dark watermark:      %d", c->dark_wm);
1450	dbg_gen("LEB overhead:        %d", c->leb_overhead);
1451	x = (long long)c->main_lebs * c->dark_wm;
1452	dbg_gen("max. dark space:     %lld (%lld KiB, %lld MiB)",
1453		x, x >> 10, x >> 20);
1454	dbg_gen("maximum bud bytes:   %lld (%lld KiB, %lld MiB)",
1455		c->max_bud_bytes, c->max_bud_bytes >> 10,
1456		c->max_bud_bytes >> 20);
1457	dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1458		c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1459		c->bg_bud_bytes >> 20);
1460	dbg_gen("current bud bytes    %lld (%lld KiB, %lld MiB)",
1461		c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1462	dbg_gen("max. seq. number:    %llu", c->max_sqnum);
1463	dbg_gen("commit number:       %llu", c->cmt_no);
1464
1465	return 0;
1466
1467out_infos:
1468	spin_lock(&ubifs_infos_lock);
1469	list_del(&c->infos_list);
1470	spin_unlock(&ubifs_infos_lock);
1471out_orphans:
1472	free_orphans(c);
1473out_journal:
1474	destroy_journal(c);
1475out_lpt:
1476	ubifs_lpt_free(c, 0);
1477out_master:
1478	kfree(c->mst_node);
1479	kfree(c->rcvrd_mst_node);
1480	if (c->bgt)
1481		kthread_stop(c->bgt);
1482out_wbufs:
1483	free_wbufs(c);
1484out_cbuf:
1485	kfree(c->cbuf);
1486out_free:
1487	kfree(c->write_reserve_buf);
1488	kfree(c->bu.buf);
1489	vfree(c->ileb_buf);
1490	vfree(c->sbuf);
1491	kfree(c->bottom_up_buf);
1492	ubifs_debugging_exit(c);
1493	return err;
1494}
1495
1496/**
1497 * ubifs_umount - un-mount UBIFS file-system.
1498 * @c: UBIFS file-system description object
1499 *
1500 * Note, this function is called to free allocated resourced when un-mounting,
1501 * as well as free resources when an error occurred while we were half way
1502 * through mounting (error path cleanup function). So it has to make sure the
1503 * resource was actually allocated before freeing it.
1504 */
1505static void ubifs_umount(struct ubifs_info *c)
1506{
1507	dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1508		c->vi.vol_id);
1509
1510	dbg_debugfs_exit_fs(c);
1511	spin_lock(&ubifs_infos_lock);
1512	list_del(&c->infos_list);
1513	spin_unlock(&ubifs_infos_lock);
1514
1515	if (c->bgt)
1516		kthread_stop(c->bgt);
1517
1518	destroy_journal(c);
1519	free_wbufs(c);
1520	free_orphans(c);
1521	ubifs_lpt_free(c, 0);
1522
1523	kfree(c->cbuf);
1524	kfree(c->rcvrd_mst_node);
1525	kfree(c->mst_node);
1526	kfree(c->write_reserve_buf);
1527	kfree(c->bu.buf);
1528	vfree(c->ileb_buf);
1529	vfree(c->sbuf);
1530	kfree(c->bottom_up_buf);
1531	ubifs_debugging_exit(c);
1532}
1533
1534/**
1535 * ubifs_remount_rw - re-mount in read-write mode.
1536 * @c: UBIFS file-system description object
1537 *
1538 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1539 * mode. This function allocates the needed resources and re-mounts UBIFS in
1540 * read-write mode.
1541 */
1542static int ubifs_remount_rw(struct ubifs_info *c)
1543{
1544	int err, lnum;
1545
1546	if (c->rw_incompat) {
1547		ubifs_err(c, "the file-system is not R/W-compatible");
1548		ubifs_msg(c, "on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
1549			  c->fmt_version, c->ro_compat_version,
1550			  UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1551		return -EROFS;
1552	}
1553
1554	mutex_lock(&c->umount_mutex);
1555	dbg_save_space_info(c);
1556	c->remounting_rw = 1;
1557	c->ro_mount = 0;
1558
1559	if (c->space_fixup) {
1560		err = ubifs_fixup_free_space(c);
1561		if (err)
1562			goto out;
1563	}
1564
1565	err = check_free_space(c);
1566	if (err)
1567		goto out;
1568
1569	if (c->old_leb_cnt != c->leb_cnt) {
1570		struct ubifs_sb_node *sup;
1571
1572		sup = ubifs_read_sb_node(c);
1573		if (IS_ERR(sup)) {
1574			err = PTR_ERR(sup);
1575			goto out;
1576		}
1577		sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1578		err = ubifs_write_sb_node(c, sup);
1579		kfree(sup);
1580		if (err)
1581			goto out;
1582	}
1583
1584	if (c->need_recovery) {
1585		ubifs_msg(c, "completing deferred recovery");
1586		err = ubifs_write_rcvrd_mst_node(c);
1587		if (err)
1588			goto out;
1589		err = ubifs_recover_size(c);
1590		if (err)
1591			goto out;
1592		err = ubifs_clean_lebs(c, c->sbuf);
1593		if (err)
1594			goto out;
1595		err = ubifs_recover_inl_heads(c, c->sbuf);
1596		if (err)
1597			goto out;
1598	} else {
1599		/* A readonly mount is not allowed to have orphans */
1600		ubifs_assert(c->tot_orphans == 0);
1601		err = ubifs_clear_orphans(c);
1602		if (err)
1603			goto out;
1604	}
1605
1606	if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1607		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1608		err = ubifs_write_master(c);
1609		if (err)
1610			goto out;
1611	}
1612
1613	c->ileb_buf = vmalloc(c->leb_size);
1614	if (!c->ileb_buf) {
1615		err = -ENOMEM;
1616		goto out;
1617	}
1618
1619	c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
1620	if (!c->write_reserve_buf) {
1621		err = -ENOMEM;
1622		goto out;
1623	}
1624
1625	err = ubifs_lpt_init(c, 0, 1);
1626	if (err)
1627		goto out;
1628
1629	/* Create background thread */
1630	c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1631	if (IS_ERR(c->bgt)) {
1632		err = PTR_ERR(c->bgt);
1633		c->bgt = NULL;
1634		ubifs_err(c, "cannot spawn \"%s\", error %d",
1635			  c->bgt_name, err);
1636		goto out;
1637	}
1638	wake_up_process(c->bgt);
1639
1640	c->orph_buf = vmalloc(c->leb_size);
1641	if (!c->orph_buf) {
1642		err = -ENOMEM;
1643		goto out;
1644	}
1645
1646	/* Check for enough log space */
1647	lnum = c->lhead_lnum + 1;
1648	if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1649		lnum = UBIFS_LOG_LNUM;
1650	if (lnum == c->ltail_lnum) {
1651		err = ubifs_consolidate_log(c);
1652		if (err)
1653			goto out;
1654	}
1655
1656	if (c->need_recovery)
1657		err = ubifs_rcvry_gc_commit(c);
1658	else
1659		err = ubifs_leb_unmap(c, c->gc_lnum);
1660	if (err)
1661		goto out;
1662
1663	dbg_gen("re-mounted read-write");
1664	c->remounting_rw = 0;
1665
1666	if (c->need_recovery) {
1667		c->need_recovery = 0;
1668		ubifs_msg(c, "deferred recovery completed");
1669	} else {
1670		/*
1671		 * Do not run the debugging space check if the were doing
1672		 * recovery, because when we saved the information we had the
1673		 * file-system in a state where the TNC and lprops has been
1674		 * modified in memory, but all the I/O operations (including a
1675		 * commit) were deferred. So the file-system was in
1676		 * "non-committed" state. Now the file-system is in committed
1677		 * state, and of course the amount of free space will change
1678		 * because, for example, the old index size was imprecise.
1679		 */
1680		err = dbg_check_space_info(c);
1681	}
1682
1683	mutex_unlock(&c->umount_mutex);
1684	return err;
1685
1686out:
1687	c->ro_mount = 1;
1688	vfree(c->orph_buf);
1689	c->orph_buf = NULL;
1690	if (c->bgt) {
1691		kthread_stop(c->bgt);
1692		c->bgt = NULL;
1693	}
1694	free_wbufs(c);
1695	kfree(c->write_reserve_buf);
1696	c->write_reserve_buf = NULL;
1697	vfree(c->ileb_buf);
1698	c->ileb_buf = NULL;
1699	ubifs_lpt_free(c, 1);
1700	c->remounting_rw = 0;
1701	mutex_unlock(&c->umount_mutex);
1702	return err;
1703}
1704
1705/**
1706 * ubifs_remount_ro - re-mount in read-only mode.
1707 * @c: UBIFS file-system description object
1708 *
1709 * We assume VFS has stopped writing. Possibly the background thread could be
1710 * running a commit, however kthread_stop will wait in that case.
1711 */
1712static void ubifs_remount_ro(struct ubifs_info *c)
1713{
1714	int i, err;
1715
1716	ubifs_assert(!c->need_recovery);
1717	ubifs_assert(!c->ro_mount);
1718
1719	mutex_lock(&c->umount_mutex);
1720	if (c->bgt) {
1721		kthread_stop(c->bgt);
1722		c->bgt = NULL;
1723	}
1724
1725	dbg_save_space_info(c);
1726
1727	for (i = 0; i < c->jhead_cnt; i++)
1728		ubifs_wbuf_sync(&c->jheads[i].wbuf);
1729
1730	c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1731	c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1732	c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1733	err = ubifs_write_master(c);
1734	if (err)
1735		ubifs_ro_mode(c, err);
1736
1737	vfree(c->orph_buf);
1738	c->orph_buf = NULL;
1739	kfree(c->write_reserve_buf);
1740	c->write_reserve_buf = NULL;
1741	vfree(c->ileb_buf);
1742	c->ileb_buf = NULL;
1743	ubifs_lpt_free(c, 1);
1744	c->ro_mount = 1;
1745	err = dbg_check_space_info(c);
1746	if (err)
1747		ubifs_ro_mode(c, err);
1748	mutex_unlock(&c->umount_mutex);
1749}
1750
1751static void ubifs_put_super(struct super_block *sb)
1752{
1753	int i;
1754	struct ubifs_info *c = sb->s_fs_info;
1755
1756	ubifs_msg(c, "un-mount UBI device %d", c->vi.ubi_num);
1757
1758	/*
1759	 * The following asserts are only valid if there has not been a failure
1760	 * of the media. For example, there will be dirty inodes if we failed
1761	 * to write them back because of I/O errors.
1762	 */
1763	if (!c->ro_error) {
1764		ubifs_assert(c->bi.idx_growth == 0);
1765		ubifs_assert(c->bi.dd_growth == 0);
1766		ubifs_assert(c->bi.data_growth == 0);
1767	}
1768
1769	/*
1770	 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1771	 * and file system un-mount. Namely, it prevents the shrinker from
1772	 * picking this superblock for shrinking - it will be just skipped if
1773	 * the mutex is locked.
1774	 */
1775	mutex_lock(&c->umount_mutex);
1776	if (!c->ro_mount) {
1777		/*
1778		 * First of all kill the background thread to make sure it does
1779		 * not interfere with un-mounting and freeing resources.
1780		 */
1781		if (c->bgt) {
1782			kthread_stop(c->bgt);
1783			c->bgt = NULL;
1784		}
1785
1786		/*
1787		 * On fatal errors c->ro_error is set to 1, in which case we do
1788		 * not write the master node.
1789		 */
1790		if (!c->ro_error) {
1791			int err;
1792
1793			/* Synchronize write-buffers */
1794			for (i = 0; i < c->jhead_cnt; i++)
1795				ubifs_wbuf_sync(&c->jheads[i].wbuf);
1796
1797			/*
1798			 * We are being cleanly unmounted which means the
1799			 * orphans were killed - indicate this in the master
1800			 * node. Also save the reserved GC LEB number.
1801			 */
1802			c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1803			c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1804			c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1805			err = ubifs_write_master(c);
1806			if (err)
1807				/*
1808				 * Recovery will attempt to fix the master area
1809				 * next mount, so we just print a message and
1810				 * continue to unmount normally.
1811				 */
1812				ubifs_err(c, "failed to write master node, error %d",
1813					  err);
1814		} else {
1815			for (i = 0; i < c->jhead_cnt; i++)
1816				/* Make sure write-buffer timers are canceled */
1817				hrtimer_cancel(&c->jheads[i].wbuf.timer);
1818		}
1819	}
1820
1821	ubifs_umount(c);
1822	bdi_destroy(&c->bdi);
1823	ubi_close_volume(c->ubi);
1824	mutex_unlock(&c->umount_mutex);
1825}
1826
1827static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1828{
1829	int err;
1830	struct ubifs_info *c = sb->s_fs_info;
1831
1832	sync_filesystem(sb);
1833	dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1834
1835	err = ubifs_parse_options(c, data, 1);
1836	if (err) {
1837		ubifs_err(c, "invalid or unknown remount parameter");
1838		return err;
1839	}
1840
1841	if (c->ro_mount && !(*flags & MS_RDONLY)) {
1842		if (c->ro_error) {
1843			ubifs_msg(c, "cannot re-mount R/W due to prior errors");
1844			return -EROFS;
1845		}
1846		if (c->ro_media) {
1847			ubifs_msg(c, "cannot re-mount R/W - UBI volume is R/O");
1848			return -EROFS;
1849		}
1850		err = ubifs_remount_rw(c);
1851		if (err)
1852			return err;
1853	} else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1854		if (c->ro_error) {
1855			ubifs_msg(c, "cannot re-mount R/O due to prior errors");
1856			return -EROFS;
1857		}
1858		ubifs_remount_ro(c);
1859	}
1860
1861	if (c->bulk_read == 1)
1862		bu_init(c);
1863	else {
1864		dbg_gen("disable bulk-read");
1865		kfree(c->bu.buf);
1866		c->bu.buf = NULL;
1867	}
1868
1869	ubifs_assert(c->lst.taken_empty_lebs > 0);
1870	return 0;
1871}
1872
1873const struct super_operations ubifs_super_operations = {
1874	.alloc_inode   = ubifs_alloc_inode,
1875	.destroy_inode = ubifs_destroy_inode,
1876	.put_super     = ubifs_put_super,
1877	.write_inode   = ubifs_write_inode,
1878	.evict_inode   = ubifs_evict_inode,
1879	.statfs        = ubifs_statfs,
1880	.dirty_inode   = ubifs_dirty_inode,
1881	.remount_fs    = ubifs_remount_fs,
1882	.show_options  = ubifs_show_options,
1883	.sync_fs       = ubifs_sync_fs,
1884};
1885
1886/**
1887 * open_ubi - parse UBI device name string and open the UBI device.
1888 * @name: UBI volume name
1889 * @mode: UBI volume open mode
1890 *
1891 * The primary method of mounting UBIFS is by specifying the UBI volume
1892 * character device node path. However, UBIFS may also be mounted withoug any
1893 * character device node using one of the following methods:
1894 *
1895 * o ubiX_Y    - mount UBI device number X, volume Y;
1896 * o ubiY      - mount UBI device number 0, volume Y;
1897 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1898 * o ubi:NAME  - mount UBI device 0, volume with name NAME.
1899 *
1900 * Alternative '!' separator may be used instead of ':' (because some shells
1901 * like busybox may interpret ':' as an NFS host name separator). This function
1902 * returns UBI volume description object in case of success and a negative
1903 * error code in case of failure.
1904 */
1905static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1906{
1907	struct ubi_volume_desc *ubi;
1908	int dev, vol;
1909	char *endptr;
1910
1911	/* First, try to open using the device node path method */
1912	ubi = ubi_open_volume_path(name, mode);
1913	if (!IS_ERR(ubi))
1914		return ubi;
1915
1916	/* Try the "nodev" method */
1917	if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1918		return ERR_PTR(-EINVAL);
1919
1920	/* ubi:NAME method */
1921	if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1922		return ubi_open_volume_nm(0, name + 4, mode);
1923
1924	if (!isdigit(name[3]))
1925		return ERR_PTR(-EINVAL);
1926
1927	dev = simple_strtoul(name + 3, &endptr, 0);
1928
1929	/* ubiY method */
1930	if (*endptr == '\0')
1931		return ubi_open_volume(0, dev, mode);
1932
1933	/* ubiX_Y method */
1934	if (*endptr == '_' && isdigit(endptr[1])) {
1935		vol = simple_strtoul(endptr + 1, &endptr, 0);
1936		if (*endptr != '\0')
1937			return ERR_PTR(-EINVAL);
1938		return ubi_open_volume(dev, vol, mode);
1939	}
1940
1941	/* ubiX:NAME method */
1942	if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1943		return ubi_open_volume_nm(dev, ++endptr, mode);
1944
1945	return ERR_PTR(-EINVAL);
1946}
1947
1948static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1949{
1950	struct ubifs_info *c;
1951
1952	c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1953	if (c) {
1954		spin_lock_init(&c->cnt_lock);
1955		spin_lock_init(&c->cs_lock);
1956		spin_lock_init(&c->buds_lock);
1957		spin_lock_init(&c->space_lock);
1958		spin_lock_init(&c->orphan_lock);
1959		init_rwsem(&c->commit_sem);
1960		mutex_init(&c->lp_mutex);
1961		mutex_init(&c->tnc_mutex);
1962		mutex_init(&c->log_mutex);
1963		mutex_init(&c->umount_mutex);
1964		mutex_init(&c->bu_mutex);
1965		mutex_init(&c->write_reserve_mutex);
1966		init_waitqueue_head(&c->cmt_wq);
1967		c->buds = RB_ROOT;
1968		c->old_idx = RB_ROOT;
1969		c->size_tree = RB_ROOT;
1970		c->orph_tree = RB_ROOT;
1971		INIT_LIST_HEAD(&c->infos_list);
1972		INIT_LIST_HEAD(&c->idx_gc);
1973		INIT_LIST_HEAD(&c->replay_list);
1974		INIT_LIST_HEAD(&c->replay_buds);
1975		INIT_LIST_HEAD(&c->uncat_list);
1976		INIT_LIST_HEAD(&c->empty_list);
1977		INIT_LIST_HEAD(&c->freeable_list);
1978		INIT_LIST_HEAD(&c->frdi_idx_list);
1979		INIT_LIST_HEAD(&c->unclean_leb_list);
1980		INIT_LIST_HEAD(&c->old_buds);
1981		INIT_LIST_HEAD(&c->orph_list);
1982		INIT_LIST_HEAD(&c->orph_new);
1983		c->no_chk_data_crc = 1;
1984
1985		c->highest_inum = UBIFS_FIRST_INO;
1986		c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1987
1988		ubi_get_volume_info(ubi, &c->vi);
1989		ubi_get_device_info(c->vi.ubi_num, &c->di);
1990	}
1991	return c;
1992}
1993
1994static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1995{
1996	struct ubifs_info *c = sb->s_fs_info;
1997	struct inode *root;
1998	int err;
1999
2000	c->vfs_sb = sb;
2001	/* Re-open the UBI device in read-write mode */
2002	c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2003	if (IS_ERR(c->ubi)) {
2004		err = PTR_ERR(c->ubi);
2005		goto out;
2006	}
2007
2008	/*
2009	 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2010	 * UBIFS, I/O is not deferred, it is done immediately in readpage,
2011	 * which means the user would have to wait not just for their own I/O
2012	 * but the read-ahead I/O as well i.e. completely pointless.
2013	 *
2014	 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
2015	 */
2016	c->bdi.name = "ubifs",
2017	c->bdi.capabilities = 0;
2018	err  = bdi_init(&c->bdi);
2019	if (err)
2020		goto out_close;
2021	err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
2022			   c->vi.ubi_num, c->vi.vol_id);
2023	if (err)
2024		goto out_bdi;
2025
2026	err = ubifs_parse_options(c, data, 0);
2027	if (err)
2028		goto out_bdi;
2029
2030	sb->s_bdi = &c->bdi;
2031	sb->s_fs_info = c;
2032	sb->s_magic = UBIFS_SUPER_MAGIC;
2033	sb->s_blocksize = UBIFS_BLOCK_SIZE;
2034	sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2035	sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2036	if (c->max_inode_sz > MAX_LFS_FILESIZE)
2037		sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2038	sb->s_op = &ubifs_super_operations;
2039	sb->s_xattr = ubifs_xattr_handlers;
2040
2041	mutex_lock(&c->umount_mutex);
2042	err = mount_ubifs(c);
2043	if (err) {
2044		ubifs_assert(err < 0);
2045		goto out_unlock;
2046	}
2047
2048	/* Read the root inode */
2049	root = ubifs_iget(sb, UBIFS_ROOT_INO);
2050	if (IS_ERR(root)) {
2051		err = PTR_ERR(root);
2052		goto out_umount;
2053	}
2054
2055	sb->s_root = d_make_root(root);
2056	if (!sb->s_root) {
2057		err = -ENOMEM;
2058		goto out_umount;
2059	}
2060
2061	mutex_unlock(&c->umount_mutex);
2062	return 0;
2063
2064out_umount:
2065	ubifs_umount(c);
2066out_unlock:
2067	mutex_unlock(&c->umount_mutex);
2068out_bdi:
2069	bdi_destroy(&c->bdi);
2070out_close:
2071	ubi_close_volume(c->ubi);
2072out:
2073	return err;
2074}
2075
2076static int sb_test(struct super_block *sb, void *data)
2077{
2078	struct ubifs_info *c1 = data;
2079	struct ubifs_info *c = sb->s_fs_info;
2080
2081	return c->vi.cdev == c1->vi.cdev;
2082}
2083
2084static int sb_set(struct super_block *sb, void *data)
2085{
2086	sb->s_fs_info = data;
2087	return set_anon_super(sb, NULL);
2088}
2089
2090static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2091			const char *name, void *data)
2092{
2093	struct ubi_volume_desc *ubi;
2094	struct ubifs_info *c;
2095	struct super_block *sb;
2096	int err;
2097
2098	dbg_gen("name %s, flags %#x", name, flags);
2099
2100	/*
2101	 * Get UBI device number and volume ID. Mount it read-only so far
2102	 * because this might be a new mount point, and UBI allows only one
2103	 * read-write user at a time.
2104	 */
2105	ubi = open_ubi(name, UBI_READONLY);
2106	if (IS_ERR(ubi)) {
2107		pr_err("UBIFS error (pid: %d): cannot open \"%s\", error %d",
2108		       current->pid, name, (int)PTR_ERR(ubi));
2109		return ERR_CAST(ubi);
2110	}
2111
2112	c = alloc_ubifs_info(ubi);
2113	if (!c) {
2114		err = -ENOMEM;
2115		goto out_close;
2116	}
2117
2118	dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2119
2120	sb = sget(fs_type, sb_test, sb_set, flags, c);
2121	if (IS_ERR(sb)) {
2122		err = PTR_ERR(sb);
2123		kfree(c);
2124		goto out_close;
2125	}
2126
2127	if (sb->s_root) {
2128		struct ubifs_info *c1 = sb->s_fs_info;
2129		kfree(c);
2130		/* A new mount point for already mounted UBIFS */
2131		dbg_gen("this ubi volume is already mounted");
2132		if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2133			err = -EBUSY;
2134			goto out_deact;
2135		}
2136	} else {
2137		err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2138		if (err)
2139			goto out_deact;
2140		/* We do not support atime */
2141		sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2142	}
2143
2144	/* 'fill_super()' opens ubi again so we must close it here */
2145	ubi_close_volume(ubi);
2146
2147	return dget(sb->s_root);
2148
2149out_deact:
2150	deactivate_locked_super(sb);
2151out_close:
2152	ubi_close_volume(ubi);
2153	return ERR_PTR(err);
2154}
2155
2156static void kill_ubifs_super(struct super_block *s)
2157{
2158	struct ubifs_info *c = s->s_fs_info;
2159	kill_anon_super(s);
2160	kfree(c);
2161}
2162
2163static struct file_system_type ubifs_fs_type = {
2164	.name    = "ubifs",
2165	.owner   = THIS_MODULE,
2166	.mount   = ubifs_mount,
2167	.kill_sb = kill_ubifs_super,
2168};
2169MODULE_ALIAS_FS("ubifs");
2170
2171/*
2172 * Inode slab cache constructor.
2173 */
2174static void inode_slab_ctor(void *obj)
2175{
2176	struct ubifs_inode *ui = obj;
2177	inode_init_once(&ui->vfs_inode);
2178}
2179
2180static int __init ubifs_init(void)
2181{
2182	int err;
2183
2184	BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2185
2186	/* Make sure node sizes are 8-byte aligned */
2187	BUILD_BUG_ON(UBIFS_CH_SZ        & 7);
2188	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  & 7);
2189	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2190	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2191	BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2192	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2193	BUILD_BUG_ON(UBIFS_SB_NODE_SZ   & 7);
2194	BUILD_BUG_ON(UBIFS_MST_NODE_SZ  & 7);
2195	BUILD_BUG_ON(UBIFS_REF_NODE_SZ  & 7);
2196	BUILD_BUG_ON(UBIFS_CS_NODE_SZ   & 7);
2197	BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2198
2199	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2200	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2201	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2202	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  & 7);
2203	BUILD_BUG_ON(UBIFS_MAX_NODE_SZ      & 7);
2204	BUILD_BUG_ON(MIN_WRITE_SZ           & 7);
2205
2206	/* Check min. node size */
2207	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  < MIN_WRITE_SZ);
2208	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2209	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2210	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2211
2212	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2213	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2214	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2215	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  > UBIFS_MAX_NODE_SZ);
2216
2217	/* Defined node sizes */
2218	BUILD_BUG_ON(UBIFS_SB_NODE_SZ  != 4096);
2219	BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2220	BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2221	BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2222
2223	/*
2224	 * We use 2 bit wide bit-fields to store compression type, which should
2225	 * be amended if more compressors are added. The bit-fields are:
2226	 * @compr_type in 'struct ubifs_inode', @default_compr in
2227	 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2228	 */
2229	BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2230
2231	/*
2232	 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2233	 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2234	 */
2235	if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2236		pr_err("UBIFS error (pid %d): VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
2237		       current->pid, (unsigned int)PAGE_CACHE_SIZE);
2238		return -EINVAL;
2239	}
2240
2241	ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2242				sizeof(struct ubifs_inode), 0,
2243				SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2244				&inode_slab_ctor);
2245	if (!ubifs_inode_slab)
2246		return -ENOMEM;
2247
2248	register_shrinker(&ubifs_shrinker_info);
2249
2250	err = ubifs_compressors_init();
2251	if (err)
2252		goto out_shrinker;
2253
2254	err = dbg_debugfs_init();
2255	if (err)
2256		goto out_compr;
2257
2258	err = register_filesystem(&ubifs_fs_type);
2259	if (err) {
2260		pr_err("UBIFS error (pid %d): cannot register file system, error %d",
2261		       current->pid, err);
2262		goto out_dbg;
2263	}
2264	return 0;
2265
2266out_dbg:
2267	dbg_debugfs_exit();
2268out_compr:
2269	ubifs_compressors_exit();
2270out_shrinker:
2271	unregister_shrinker(&ubifs_shrinker_info);
2272	kmem_cache_destroy(ubifs_inode_slab);
2273	return err;
2274}
2275/* late_initcall to let compressors initialize first */
2276late_initcall(ubifs_init);
2277
2278static void __exit ubifs_exit(void)
2279{
2280	ubifs_assert(list_empty(&ubifs_infos));
2281	ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2282
2283	dbg_debugfs_exit();
2284	ubifs_compressors_exit();
2285	unregister_shrinker(&ubifs_shrinker_info);
2286
2287	/*
2288	 * Make sure all delayed rcu free inodes are flushed before we
2289	 * destroy cache.
2290	 */
2291	rcu_barrier();
2292	kmem_cache_destroy(ubifs_inode_slab);
2293	unregister_filesystem(&ubifs_fs_type);
2294}
2295module_exit(ubifs_exit);
2296
2297MODULE_LICENSE("GPL");
2298MODULE_VERSION(__stringify(UBIFS_VERSION));
2299MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2300MODULE_DESCRIPTION("UBIFS - UBI File System");
2301