xref: /kernel/linux/linux-5.10/fs/ubifs/super.c (revision 8c2ecf20)
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * This file is part of UBIFS.
4 *
5 * Copyright (C) 2006-2008 Nokia Corporation.
6 *
7 * Authors: Artem Bityutskiy (Битюцкий Артём)
8 *          Adrian Hunter
9 */
10
11/*
12 * This file implements UBIFS initialization and VFS superblock operations. Some
13 * initialization stuff which is rather large and complex is placed at
14 * corresponding subsystems, but most of it is here.
15 */
16
17#include <linux/init.h>
18#include <linux/slab.h>
19#include <linux/module.h>
20#include <linux/ctype.h>
21#include <linux/kthread.h>
22#include <linux/parser.h>
23#include <linux/seq_file.h>
24#include <linux/mount.h>
25#include <linux/math64.h>
26#include <linux/writeback.h>
27#include "ubifs.h"
28
29static int ubifs_default_version_set(const char *val, const struct kernel_param *kp)
30{
31	int n = 0, ret;
32
33	ret = kstrtoint(val, 10, &n);
34	if (ret != 0 || n < 4 || n > UBIFS_FORMAT_VERSION)
35		return -EINVAL;
36	return param_set_int(val, kp);
37}
38
39static const struct kernel_param_ops ubifs_default_version_ops = {
40	.set = ubifs_default_version_set,
41	.get = param_get_int,
42};
43
44int ubifs_default_version = UBIFS_FORMAT_VERSION;
45module_param_cb(default_version, &ubifs_default_version_ops, &ubifs_default_version, 0600);
46
47/*
48 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
49 * allocating too much.
50 */
51#define UBIFS_KMALLOC_OK (128*1024)
52
53/* Slab cache for UBIFS inodes */
54static struct kmem_cache *ubifs_inode_slab;
55
56/* UBIFS TNC shrinker description */
57static struct shrinker ubifs_shrinker_info = {
58	.scan_objects = ubifs_shrink_scan,
59	.count_objects = ubifs_shrink_count,
60	.seeks = DEFAULT_SEEKS,
61};
62
63/**
64 * validate_inode - validate inode.
65 * @c: UBIFS file-system description object
66 * @inode: the inode to validate
67 *
68 * This is a helper function for 'ubifs_iget()' which validates various fields
69 * of a newly built inode to make sure they contain sane values and prevent
70 * possible vulnerabilities. Returns zero if the inode is all right and
71 * a non-zero error code if not.
72 */
73static int validate_inode(struct ubifs_info *c, const struct inode *inode)
74{
75	int err;
76	const struct ubifs_inode *ui = ubifs_inode(inode);
77
78	if (inode->i_size > c->max_inode_sz) {
79		ubifs_err(c, "inode is too large (%lld)",
80			  (long long)inode->i_size);
81		return 1;
82	}
83
84	if (ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
85		ubifs_err(c, "unknown compression type %d", ui->compr_type);
86		return 2;
87	}
88
89	if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
90		return 3;
91
92	if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
93		return 4;
94
95	if (ui->xattr && !S_ISREG(inode->i_mode))
96		return 5;
97
98	if (!ubifs_compr_present(c, ui->compr_type)) {
99		ubifs_warn(c, "inode %lu uses '%s' compression, but it was not compiled in",
100			   inode->i_ino, ubifs_compr_name(c, ui->compr_type));
101	}
102
103	err = dbg_check_dir(c, inode);
104	return err;
105}
106
107struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
108{
109	int err;
110	union ubifs_key key;
111	struct ubifs_ino_node *ino;
112	struct ubifs_info *c = sb->s_fs_info;
113	struct inode *inode;
114	struct ubifs_inode *ui;
115
116	dbg_gen("inode %lu", inum);
117
118	inode = iget_locked(sb, inum);
119	if (!inode)
120		return ERR_PTR(-ENOMEM);
121	if (!(inode->i_state & I_NEW))
122		return inode;
123	ui = ubifs_inode(inode);
124
125	ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
126	if (!ino) {
127		err = -ENOMEM;
128		goto out;
129	}
130
131	ino_key_init(c, &key, inode->i_ino);
132
133	err = ubifs_tnc_lookup(c, &key, ino);
134	if (err)
135		goto out_ino;
136
137	inode->i_flags |= S_NOCMTIME;
138
139	if (!IS_ENABLED(CONFIG_UBIFS_ATIME_SUPPORT))
140		inode->i_flags |= S_NOATIME;
141
142	set_nlink(inode, le32_to_cpu(ino->nlink));
143	i_uid_write(inode, le32_to_cpu(ino->uid));
144	i_gid_write(inode, le32_to_cpu(ino->gid));
145	inode->i_atime.tv_sec  = (int64_t)le64_to_cpu(ino->atime_sec);
146	inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
147	inode->i_mtime.tv_sec  = (int64_t)le64_to_cpu(ino->mtime_sec);
148	inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
149	inode->i_ctime.tv_sec  = (int64_t)le64_to_cpu(ino->ctime_sec);
150	inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
151	inode->i_mode = le32_to_cpu(ino->mode);
152	inode->i_size = le64_to_cpu(ino->size);
153
154	ui->data_len    = le32_to_cpu(ino->data_len);
155	ui->flags       = le32_to_cpu(ino->flags);
156	ui->compr_type  = le16_to_cpu(ino->compr_type);
157	ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
158	ui->xattr_cnt   = le32_to_cpu(ino->xattr_cnt);
159	ui->xattr_size  = le32_to_cpu(ino->xattr_size);
160	ui->xattr_names = le32_to_cpu(ino->xattr_names);
161	ui->synced_i_size = ui->ui_size = inode->i_size;
162
163	ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
164
165	err = validate_inode(c, inode);
166	if (err)
167		goto out_invalid;
168
169	switch (inode->i_mode & S_IFMT) {
170	case S_IFREG:
171		inode->i_mapping->a_ops = &ubifs_file_address_operations;
172		inode->i_op = &ubifs_file_inode_operations;
173		inode->i_fop = &ubifs_file_operations;
174		if (ui->xattr) {
175			ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
176			if (!ui->data) {
177				err = -ENOMEM;
178				goto out_ino;
179			}
180			memcpy(ui->data, ino->data, ui->data_len);
181			((char *)ui->data)[ui->data_len] = '\0';
182		} else if (ui->data_len != 0) {
183			err = 10;
184			goto out_invalid;
185		}
186		break;
187	case S_IFDIR:
188		inode->i_op  = &ubifs_dir_inode_operations;
189		inode->i_fop = &ubifs_dir_operations;
190		if (ui->data_len != 0) {
191			err = 11;
192			goto out_invalid;
193		}
194		break;
195	case S_IFLNK:
196		inode->i_op = &ubifs_symlink_inode_operations;
197		if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
198			err = 12;
199			goto out_invalid;
200		}
201		ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
202		if (!ui->data) {
203			err = -ENOMEM;
204			goto out_ino;
205		}
206		memcpy(ui->data, ino->data, ui->data_len);
207		((char *)ui->data)[ui->data_len] = '\0';
208		break;
209	case S_IFBLK:
210	case S_IFCHR:
211	{
212		dev_t rdev;
213		union ubifs_dev_desc *dev;
214
215		ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
216		if (!ui->data) {
217			err = -ENOMEM;
218			goto out_ino;
219		}
220
221		dev = (union ubifs_dev_desc *)ino->data;
222		if (ui->data_len == sizeof(dev->new))
223			rdev = new_decode_dev(le32_to_cpu(dev->new));
224		else if (ui->data_len == sizeof(dev->huge))
225			rdev = huge_decode_dev(le64_to_cpu(dev->huge));
226		else {
227			err = 13;
228			goto out_invalid;
229		}
230		memcpy(ui->data, ino->data, ui->data_len);
231		inode->i_op = &ubifs_file_inode_operations;
232		init_special_inode(inode, inode->i_mode, rdev);
233		break;
234	}
235	case S_IFSOCK:
236	case S_IFIFO:
237		inode->i_op = &ubifs_file_inode_operations;
238		init_special_inode(inode, inode->i_mode, 0);
239		if (ui->data_len != 0) {
240			err = 14;
241			goto out_invalid;
242		}
243		break;
244	default:
245		err = 15;
246		goto out_invalid;
247	}
248
249	kfree(ino);
250	ubifs_set_inode_flags(inode);
251	unlock_new_inode(inode);
252	return inode;
253
254out_invalid:
255	ubifs_err(c, "inode %lu validation failed, error %d", inode->i_ino, err);
256	ubifs_dump_node(c, ino, UBIFS_MAX_INO_NODE_SZ);
257	ubifs_dump_inode(c, inode);
258	err = -EINVAL;
259out_ino:
260	kfree(ino);
261out:
262	ubifs_err(c, "failed to read inode %lu, error %d", inode->i_ino, err);
263	iget_failed(inode);
264	return ERR_PTR(err);
265}
266
267static struct inode *ubifs_alloc_inode(struct super_block *sb)
268{
269	struct ubifs_inode *ui;
270
271	ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
272	if (!ui)
273		return NULL;
274
275	memset((void *)ui + sizeof(struct inode), 0,
276	       sizeof(struct ubifs_inode) - sizeof(struct inode));
277	mutex_init(&ui->ui_mutex);
278	init_rwsem(&ui->xattr_sem);
279	spin_lock_init(&ui->ui_lock);
280	return &ui->vfs_inode;
281};
282
283static void ubifs_free_inode(struct inode *inode)
284{
285	struct ubifs_inode *ui = ubifs_inode(inode);
286
287	kfree(ui->data);
288	fscrypt_free_inode(inode);
289
290	kmem_cache_free(ubifs_inode_slab, ui);
291}
292
293/*
294 * Note, Linux write-back code calls this without 'i_mutex'.
295 */
296static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
297{
298	int err = 0;
299	struct ubifs_info *c = inode->i_sb->s_fs_info;
300	struct ubifs_inode *ui = ubifs_inode(inode);
301
302	ubifs_assert(c, !ui->xattr);
303	if (is_bad_inode(inode))
304		return 0;
305
306	mutex_lock(&ui->ui_mutex);
307	/*
308	 * Due to races between write-back forced by budgeting
309	 * (see 'sync_some_inodes()') and background write-back, the inode may
310	 * have already been synchronized, do not do this again. This might
311	 * also happen if it was synchronized in an VFS operation, e.g.
312	 * 'ubifs_link()'.
313	 */
314	if (!ui->dirty) {
315		mutex_unlock(&ui->ui_mutex);
316		return 0;
317	}
318
319	/*
320	 * As an optimization, do not write orphan inodes to the media just
321	 * because this is not needed.
322	 */
323	dbg_gen("inode %lu, mode %#x, nlink %u",
324		inode->i_ino, (int)inode->i_mode, inode->i_nlink);
325	if (inode->i_nlink) {
326		err = ubifs_jnl_write_inode(c, inode);
327		if (err)
328			ubifs_err(c, "can't write inode %lu, error %d",
329				  inode->i_ino, err);
330		else
331			err = dbg_check_inode_size(c, inode, ui->ui_size);
332	}
333
334	ui->dirty = 0;
335	mutex_unlock(&ui->ui_mutex);
336	ubifs_release_dirty_inode_budget(c, ui);
337	return err;
338}
339
340static int ubifs_drop_inode(struct inode *inode)
341{
342	int drop = generic_drop_inode(inode);
343
344	if (!drop)
345		drop = fscrypt_drop_inode(inode);
346
347	return drop;
348}
349
350static void ubifs_evict_inode(struct inode *inode)
351{
352	int err;
353	struct ubifs_info *c = inode->i_sb->s_fs_info;
354	struct ubifs_inode *ui = ubifs_inode(inode);
355
356	if (ui->xattr)
357		/*
358		 * Extended attribute inode deletions are fully handled in
359		 * 'ubifs_removexattr()'. These inodes are special and have
360		 * limited usage, so there is nothing to do here.
361		 */
362		goto out;
363
364	dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
365	ubifs_assert(c, !atomic_read(&inode->i_count));
366
367	truncate_inode_pages_final(&inode->i_data);
368
369	if (inode->i_nlink)
370		goto done;
371
372	if (is_bad_inode(inode))
373		goto out;
374
375	ui->ui_size = inode->i_size = 0;
376	err = ubifs_jnl_delete_inode(c, inode);
377	if (err)
378		/*
379		 * Worst case we have a lost orphan inode wasting space, so a
380		 * simple error message is OK here.
381		 */
382		ubifs_err(c, "can't delete inode %lu, error %d",
383			  inode->i_ino, err);
384
385out:
386	if (ui->dirty)
387		ubifs_release_dirty_inode_budget(c, ui);
388	else {
389		/* We've deleted something - clean the "no space" flags */
390		c->bi.nospace = c->bi.nospace_rp = 0;
391		smp_wmb();
392	}
393done:
394	clear_inode(inode);
395	fscrypt_put_encryption_info(inode);
396}
397
398static void ubifs_dirty_inode(struct inode *inode, int flags)
399{
400	struct ubifs_info *c = inode->i_sb->s_fs_info;
401	struct ubifs_inode *ui = ubifs_inode(inode);
402
403	ubifs_assert(c, mutex_is_locked(&ui->ui_mutex));
404	if (!ui->dirty) {
405		ui->dirty = 1;
406		dbg_gen("inode %lu",  inode->i_ino);
407	}
408}
409
410static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
411{
412	struct ubifs_info *c = dentry->d_sb->s_fs_info;
413	unsigned long long free;
414	__le32 *uuid = (__le32 *)c->uuid;
415
416	free = ubifs_get_free_space(c);
417	dbg_gen("free space %lld bytes (%lld blocks)",
418		free, free >> UBIFS_BLOCK_SHIFT);
419
420	buf->f_type = UBIFS_SUPER_MAGIC;
421	buf->f_bsize = UBIFS_BLOCK_SIZE;
422	buf->f_blocks = c->block_cnt;
423	buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
424	if (free > c->report_rp_size)
425		buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
426	else
427		buf->f_bavail = 0;
428	buf->f_files = 0;
429	buf->f_ffree = 0;
430	buf->f_namelen = UBIFS_MAX_NLEN;
431	buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
432	buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
433	ubifs_assert(c, buf->f_bfree <= c->block_cnt);
434	return 0;
435}
436
437static int ubifs_show_options(struct seq_file *s, struct dentry *root)
438{
439	struct ubifs_info *c = root->d_sb->s_fs_info;
440
441	if (c->mount_opts.unmount_mode == 2)
442		seq_puts(s, ",fast_unmount");
443	else if (c->mount_opts.unmount_mode == 1)
444		seq_puts(s, ",norm_unmount");
445
446	if (c->mount_opts.bulk_read == 2)
447		seq_puts(s, ",bulk_read");
448	else if (c->mount_opts.bulk_read == 1)
449		seq_puts(s, ",no_bulk_read");
450
451	if (c->mount_opts.chk_data_crc == 2)
452		seq_puts(s, ",chk_data_crc");
453	else if (c->mount_opts.chk_data_crc == 1)
454		seq_puts(s, ",no_chk_data_crc");
455
456	if (c->mount_opts.override_compr) {
457		seq_printf(s, ",compr=%s",
458			   ubifs_compr_name(c, c->mount_opts.compr_type));
459	}
460
461	seq_printf(s, ",assert=%s", ubifs_assert_action_name(c));
462	seq_printf(s, ",ubi=%d,vol=%d", c->vi.ubi_num, c->vi.vol_id);
463
464	return 0;
465}
466
467static int ubifs_sync_fs(struct super_block *sb, int wait)
468{
469	int i, err;
470	struct ubifs_info *c = sb->s_fs_info;
471
472	/*
473	 * Zero @wait is just an advisory thing to help the file system shove
474	 * lots of data into the queues, and there will be the second
475	 * '->sync_fs()' call, with non-zero @wait.
476	 */
477	if (!wait)
478		return 0;
479
480	/*
481	 * Synchronize write buffers, because 'ubifs_run_commit()' does not
482	 * do this if it waits for an already running commit.
483	 */
484	for (i = 0; i < c->jhead_cnt; i++) {
485		err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
486		if (err)
487			return err;
488	}
489
490	/*
491	 * Strictly speaking, it is not necessary to commit the journal here,
492	 * synchronizing write-buffers would be enough. But committing makes
493	 * UBIFS free space predictions much more accurate, so we want to let
494	 * the user be able to get more accurate results of 'statfs()' after
495	 * they synchronize the file system.
496	 */
497	err = ubifs_run_commit(c);
498	if (err)
499		return err;
500
501	return ubi_sync(c->vi.ubi_num);
502}
503
504/**
505 * init_constants_early - initialize UBIFS constants.
506 * @c: UBIFS file-system description object
507 *
508 * This function initialize UBIFS constants which do not need the superblock to
509 * be read. It also checks that the UBI volume satisfies basic UBIFS
510 * requirements. Returns zero in case of success and a negative error code in
511 * case of failure.
512 */
513static int init_constants_early(struct ubifs_info *c)
514{
515	if (c->vi.corrupted) {
516		ubifs_warn(c, "UBI volume is corrupted - read-only mode");
517		c->ro_media = 1;
518	}
519
520	if (c->di.ro_mode) {
521		ubifs_msg(c, "read-only UBI device");
522		c->ro_media = 1;
523	}
524
525	if (c->vi.vol_type == UBI_STATIC_VOLUME) {
526		ubifs_msg(c, "static UBI volume - read-only mode");
527		c->ro_media = 1;
528	}
529
530	c->leb_cnt = c->vi.size;
531	c->leb_size = c->vi.usable_leb_size;
532	c->leb_start = c->di.leb_start;
533	c->half_leb_size = c->leb_size / 2;
534	c->min_io_size = c->di.min_io_size;
535	c->min_io_shift = fls(c->min_io_size) - 1;
536	c->max_write_size = c->di.max_write_size;
537	c->max_write_shift = fls(c->max_write_size) - 1;
538
539	if (c->leb_size < UBIFS_MIN_LEB_SZ) {
540		ubifs_errc(c, "too small LEBs (%d bytes), min. is %d bytes",
541			   c->leb_size, UBIFS_MIN_LEB_SZ);
542		return -EINVAL;
543	}
544
545	if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
546		ubifs_errc(c, "too few LEBs (%d), min. is %d",
547			   c->leb_cnt, UBIFS_MIN_LEB_CNT);
548		return -EINVAL;
549	}
550
551	if (!is_power_of_2(c->min_io_size)) {
552		ubifs_errc(c, "bad min. I/O size %d", c->min_io_size);
553		return -EINVAL;
554	}
555
556	/*
557	 * Maximum write size has to be greater or equivalent to min. I/O
558	 * size, and be multiple of min. I/O size.
559	 */
560	if (c->max_write_size < c->min_io_size ||
561	    c->max_write_size % c->min_io_size ||
562	    !is_power_of_2(c->max_write_size)) {
563		ubifs_errc(c, "bad write buffer size %d for %d min. I/O unit",
564			   c->max_write_size, c->min_io_size);
565		return -EINVAL;
566	}
567
568	/*
569	 * UBIFS aligns all node to 8-byte boundary, so to make function in
570	 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
571	 * less than 8.
572	 */
573	if (c->min_io_size < 8) {
574		c->min_io_size = 8;
575		c->min_io_shift = 3;
576		if (c->max_write_size < c->min_io_size) {
577			c->max_write_size = c->min_io_size;
578			c->max_write_shift = c->min_io_shift;
579		}
580	}
581
582	c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
583	c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
584
585	/*
586	 * Initialize node length ranges which are mostly needed for node
587	 * length validation.
588	 */
589	c->ranges[UBIFS_PAD_NODE].len  = UBIFS_PAD_NODE_SZ;
590	c->ranges[UBIFS_SB_NODE].len   = UBIFS_SB_NODE_SZ;
591	c->ranges[UBIFS_MST_NODE].len  = UBIFS_MST_NODE_SZ;
592	c->ranges[UBIFS_REF_NODE].len  = UBIFS_REF_NODE_SZ;
593	c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
594	c->ranges[UBIFS_CS_NODE].len   = UBIFS_CS_NODE_SZ;
595	c->ranges[UBIFS_AUTH_NODE].min_len = UBIFS_AUTH_NODE_SZ;
596	c->ranges[UBIFS_AUTH_NODE].max_len = UBIFS_AUTH_NODE_SZ +
597				UBIFS_MAX_HMAC_LEN;
598	c->ranges[UBIFS_SIG_NODE].min_len = UBIFS_SIG_NODE_SZ;
599	c->ranges[UBIFS_SIG_NODE].max_len = c->leb_size - UBIFS_SB_NODE_SZ;
600
601	c->ranges[UBIFS_INO_NODE].min_len  = UBIFS_INO_NODE_SZ;
602	c->ranges[UBIFS_INO_NODE].max_len  = UBIFS_MAX_INO_NODE_SZ;
603	c->ranges[UBIFS_ORPH_NODE].min_len =
604				UBIFS_ORPH_NODE_SZ + sizeof(__le64);
605	c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
606	c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
607	c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
608	c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
609	c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
610	c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
611	c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
612	/*
613	 * Minimum indexing node size is amended later when superblock is
614	 * read and the key length is known.
615	 */
616	c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
617	/*
618	 * Maximum indexing node size is amended later when superblock is
619	 * read and the fanout is known.
620	 */
621	c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
622
623	/*
624	 * Initialize dead and dark LEB space watermarks. See gc.c for comments
625	 * about these values.
626	 */
627	c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
628	c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
629
630	/*
631	 * Calculate how many bytes would be wasted at the end of LEB if it was
632	 * fully filled with data nodes of maximum size. This is used in
633	 * calculations when reporting free space.
634	 */
635	c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
636
637	/* Buffer size for bulk-reads */
638	c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
639	if (c->max_bu_buf_len > c->leb_size)
640		c->max_bu_buf_len = c->leb_size;
641
642	/* Log is ready, preserve one LEB for commits. */
643	c->min_log_bytes = c->leb_size;
644
645	return 0;
646}
647
648/**
649 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
650 * @c: UBIFS file-system description object
651 * @lnum: LEB the write-buffer was synchronized to
652 * @free: how many free bytes left in this LEB
653 * @pad: how many bytes were padded
654 *
655 * This is a callback function which is called by the I/O unit when the
656 * write-buffer is synchronized. We need this to correctly maintain space
657 * accounting in bud logical eraseblocks. This function returns zero in case of
658 * success and a negative error code in case of failure.
659 *
660 * This function actually belongs to the journal, but we keep it here because
661 * we want to keep it static.
662 */
663static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
664{
665	return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
666}
667
668/*
669 * init_constants_sb - initialize UBIFS constants.
670 * @c: UBIFS file-system description object
671 *
672 * This is a helper function which initializes various UBIFS constants after
673 * the superblock has been read. It also checks various UBIFS parameters and
674 * makes sure they are all right. Returns zero in case of success and a
675 * negative error code in case of failure.
676 */
677static int init_constants_sb(struct ubifs_info *c)
678{
679	int tmp, err;
680	long long tmp64;
681
682	c->main_bytes = (long long)c->main_lebs * c->leb_size;
683	c->max_znode_sz = sizeof(struct ubifs_znode) +
684				c->fanout * sizeof(struct ubifs_zbranch);
685
686	tmp = ubifs_idx_node_sz(c, 1);
687	c->ranges[UBIFS_IDX_NODE].min_len = tmp;
688	c->min_idx_node_sz = ALIGN(tmp, 8);
689
690	tmp = ubifs_idx_node_sz(c, c->fanout);
691	c->ranges[UBIFS_IDX_NODE].max_len = tmp;
692	c->max_idx_node_sz = ALIGN(tmp, 8);
693
694	/* Make sure LEB size is large enough to fit full commit */
695	tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
696	tmp = ALIGN(tmp, c->min_io_size);
697	if (tmp > c->leb_size) {
698		ubifs_err(c, "too small LEB size %d, at least %d needed",
699			  c->leb_size, tmp);
700		return -EINVAL;
701	}
702
703	/*
704	 * Make sure that the log is large enough to fit reference nodes for
705	 * all buds plus one reserved LEB.
706	 */
707	tmp64 = c->max_bud_bytes + c->leb_size - 1;
708	c->max_bud_cnt = div_u64(tmp64, c->leb_size);
709	tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
710	tmp /= c->leb_size;
711	tmp += 1;
712	if (c->log_lebs < tmp) {
713		ubifs_err(c, "too small log %d LEBs, required min. %d LEBs",
714			  c->log_lebs, tmp);
715		return -EINVAL;
716	}
717
718	/*
719	 * When budgeting we assume worst-case scenarios when the pages are not
720	 * be compressed and direntries are of the maximum size.
721	 *
722	 * Note, data, which may be stored in inodes is budgeted separately, so
723	 * it is not included into 'c->bi.inode_budget'.
724	 */
725	c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
726	c->bi.inode_budget = UBIFS_INO_NODE_SZ;
727	c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
728
729	/*
730	 * When the amount of flash space used by buds becomes
731	 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
732	 * The writers are unblocked when the commit is finished. To avoid
733	 * writers to be blocked UBIFS initiates background commit in advance,
734	 * when number of bud bytes becomes above the limit defined below.
735	 */
736	c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
737
738	/*
739	 * Ensure minimum journal size. All the bytes in the journal heads are
740	 * considered to be used, when calculating the current journal usage.
741	 * Consequently, if the journal is too small, UBIFS will treat it as
742	 * always full.
743	 */
744	tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
745	if (c->bg_bud_bytes < tmp64)
746		c->bg_bud_bytes = tmp64;
747	if (c->max_bud_bytes < tmp64 + c->leb_size)
748		c->max_bud_bytes = tmp64 + c->leb_size;
749
750	err = ubifs_calc_lpt_geom(c);
751	if (err)
752		return err;
753
754	/* Initialize effective LEB size used in budgeting calculations */
755	c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
756	return 0;
757}
758
759/*
760 * init_constants_master - initialize UBIFS constants.
761 * @c: UBIFS file-system description object
762 *
763 * This is a helper function which initializes various UBIFS constants after
764 * the master node has been read. It also checks various UBIFS parameters and
765 * makes sure they are all right.
766 */
767static void init_constants_master(struct ubifs_info *c)
768{
769	long long tmp64;
770
771	c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
772	c->report_rp_size = ubifs_reported_space(c, c->rp_size);
773
774	/*
775	 * Calculate total amount of FS blocks. This number is not used
776	 * internally because it does not make much sense for UBIFS, but it is
777	 * necessary to report something for the 'statfs()' call.
778	 *
779	 * Subtract the LEB reserved for GC, the LEB which is reserved for
780	 * deletions, minimum LEBs for the index, and assume only one journal
781	 * head is available.
782	 */
783	tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
784	tmp64 *= (long long)c->leb_size - c->leb_overhead;
785	tmp64 = ubifs_reported_space(c, tmp64);
786	c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
787}
788
789/**
790 * take_gc_lnum - reserve GC LEB.
791 * @c: UBIFS file-system description object
792 *
793 * This function ensures that the LEB reserved for garbage collection is marked
794 * as "taken" in lprops. We also have to set free space to LEB size and dirty
795 * space to zero, because lprops may contain out-of-date information if the
796 * file-system was un-mounted before it has been committed. This function
797 * returns zero in case of success and a negative error code in case of
798 * failure.
799 */
800static int take_gc_lnum(struct ubifs_info *c)
801{
802	int err;
803
804	if (c->gc_lnum == -1) {
805		ubifs_err(c, "no LEB for GC");
806		return -EINVAL;
807	}
808
809	/* And we have to tell lprops that this LEB is taken */
810	err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
811				  LPROPS_TAKEN, 0, 0);
812	return err;
813}
814
815/**
816 * alloc_wbufs - allocate write-buffers.
817 * @c: UBIFS file-system description object
818 *
819 * This helper function allocates and initializes UBIFS write-buffers. Returns
820 * zero in case of success and %-ENOMEM in case of failure.
821 */
822static int alloc_wbufs(struct ubifs_info *c)
823{
824	int i, err;
825
826	c->jheads = kcalloc(c->jhead_cnt, sizeof(struct ubifs_jhead),
827			    GFP_KERNEL);
828	if (!c->jheads)
829		return -ENOMEM;
830
831	/* Initialize journal heads */
832	for (i = 0; i < c->jhead_cnt; i++) {
833		INIT_LIST_HEAD(&c->jheads[i].buds_list);
834		err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
835		if (err)
836			goto out_wbuf;
837
838		c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
839		c->jheads[i].wbuf.jhead = i;
840		c->jheads[i].grouped = 1;
841		c->jheads[i].log_hash = ubifs_hash_get_desc(c);
842		if (IS_ERR(c->jheads[i].log_hash)) {
843			err = PTR_ERR(c->jheads[i].log_hash);
844			goto out_log_hash;
845		}
846	}
847
848	/*
849	 * Garbage Collector head does not need to be synchronized by timer.
850	 * Also GC head nodes are not grouped.
851	 */
852	c->jheads[GCHD].wbuf.no_timer = 1;
853	c->jheads[GCHD].grouped = 0;
854
855	return 0;
856
857out_log_hash:
858	kfree(c->jheads[i].wbuf.buf);
859	kfree(c->jheads[i].wbuf.inodes);
860
861out_wbuf:
862	while (i--) {
863		kfree(c->jheads[i].wbuf.buf);
864		kfree(c->jheads[i].wbuf.inodes);
865		kfree(c->jheads[i].log_hash);
866	}
867	kfree(c->jheads);
868	c->jheads = NULL;
869
870	return err;
871}
872
873/**
874 * free_wbufs - free write-buffers.
875 * @c: UBIFS file-system description object
876 */
877static void free_wbufs(struct ubifs_info *c)
878{
879	int i;
880
881	if (c->jheads) {
882		for (i = 0; i < c->jhead_cnt; i++) {
883			kfree(c->jheads[i].wbuf.buf);
884			kfree(c->jheads[i].wbuf.inodes);
885			kfree(c->jheads[i].log_hash);
886		}
887		kfree(c->jheads);
888		c->jheads = NULL;
889	}
890}
891
892/**
893 * free_orphans - free orphans.
894 * @c: UBIFS file-system description object
895 */
896static void free_orphans(struct ubifs_info *c)
897{
898	struct ubifs_orphan *orph;
899
900	while (c->orph_dnext) {
901		orph = c->orph_dnext;
902		c->orph_dnext = orph->dnext;
903		list_del(&orph->list);
904		kfree(orph);
905	}
906
907	while (!list_empty(&c->orph_list)) {
908		orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
909		list_del(&orph->list);
910		kfree(orph);
911		ubifs_err(c, "orphan list not empty at unmount");
912	}
913
914	vfree(c->orph_buf);
915	c->orph_buf = NULL;
916}
917
918/**
919 * free_buds - free per-bud objects.
920 * @c: UBIFS file-system description object
921 */
922static void free_buds(struct ubifs_info *c)
923{
924	struct ubifs_bud *bud, *n;
925
926	rbtree_postorder_for_each_entry_safe(bud, n, &c->buds, rb)
927		kfree(bud);
928}
929
930/**
931 * check_volume_empty - check if the UBI volume is empty.
932 * @c: UBIFS file-system description object
933 *
934 * This function checks if the UBIFS volume is empty by looking if its LEBs are
935 * mapped or not. The result of checking is stored in the @c->empty variable.
936 * Returns zero in case of success and a negative error code in case of
937 * failure.
938 */
939static int check_volume_empty(struct ubifs_info *c)
940{
941	int lnum, err;
942
943	c->empty = 1;
944	for (lnum = 0; lnum < c->leb_cnt; lnum++) {
945		err = ubifs_is_mapped(c, lnum);
946		if (unlikely(err < 0))
947			return err;
948		if (err == 1) {
949			c->empty = 0;
950			break;
951		}
952
953		cond_resched();
954	}
955
956	return 0;
957}
958
959/*
960 * UBIFS mount options.
961 *
962 * Opt_fast_unmount: do not run a journal commit before un-mounting
963 * Opt_norm_unmount: run a journal commit before un-mounting
964 * Opt_bulk_read: enable bulk-reads
965 * Opt_no_bulk_read: disable bulk-reads
966 * Opt_chk_data_crc: check CRCs when reading data nodes
967 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
968 * Opt_override_compr: override default compressor
969 * Opt_assert: set ubifs_assert() action
970 * Opt_auth_key: The key name used for authentication
971 * Opt_auth_hash_name: The hash type used for authentication
972 * Opt_err: just end of array marker
973 */
974enum {
975	Opt_fast_unmount,
976	Opt_norm_unmount,
977	Opt_bulk_read,
978	Opt_no_bulk_read,
979	Opt_chk_data_crc,
980	Opt_no_chk_data_crc,
981	Opt_override_compr,
982	Opt_assert,
983	Opt_auth_key,
984	Opt_auth_hash_name,
985	Opt_ignore,
986	Opt_err,
987};
988
989static const match_table_t tokens = {
990	{Opt_fast_unmount, "fast_unmount"},
991	{Opt_norm_unmount, "norm_unmount"},
992	{Opt_bulk_read, "bulk_read"},
993	{Opt_no_bulk_read, "no_bulk_read"},
994	{Opt_chk_data_crc, "chk_data_crc"},
995	{Opt_no_chk_data_crc, "no_chk_data_crc"},
996	{Opt_override_compr, "compr=%s"},
997	{Opt_auth_key, "auth_key=%s"},
998	{Opt_auth_hash_name, "auth_hash_name=%s"},
999	{Opt_ignore, "ubi=%s"},
1000	{Opt_ignore, "vol=%s"},
1001	{Opt_assert, "assert=%s"},
1002	{Opt_err, NULL},
1003};
1004
1005/**
1006 * parse_standard_option - parse a standard mount option.
1007 * @option: the option to parse
1008 *
1009 * Normally, standard mount options like "sync" are passed to file-systems as
1010 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
1011 * be present in the options string. This function tries to deal with this
1012 * situation and parse standard options. Returns 0 if the option was not
1013 * recognized, and the corresponding integer flag if it was.
1014 *
1015 * UBIFS is only interested in the "sync" option, so do not check for anything
1016 * else.
1017 */
1018static int parse_standard_option(const char *option)
1019{
1020
1021	pr_notice("UBIFS: parse %s\n", option);
1022	if (!strcmp(option, "sync"))
1023		return SB_SYNCHRONOUS;
1024	return 0;
1025}
1026
1027/**
1028 * ubifs_parse_options - parse mount parameters.
1029 * @c: UBIFS file-system description object
1030 * @options: parameters to parse
1031 * @is_remount: non-zero if this is FS re-mount
1032 *
1033 * This function parses UBIFS mount options and returns zero in case success
1034 * and a negative error code in case of failure.
1035 */
1036static int ubifs_parse_options(struct ubifs_info *c, char *options,
1037			       int is_remount)
1038{
1039	char *p;
1040	substring_t args[MAX_OPT_ARGS];
1041
1042	if (!options)
1043		return 0;
1044
1045	while ((p = strsep(&options, ","))) {
1046		int token;
1047
1048		if (!*p)
1049			continue;
1050
1051		token = match_token(p, tokens, args);
1052		switch (token) {
1053		/*
1054		 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
1055		 * We accept them in order to be backward-compatible. But this
1056		 * should be removed at some point.
1057		 */
1058		case Opt_fast_unmount:
1059			c->mount_opts.unmount_mode = 2;
1060			break;
1061		case Opt_norm_unmount:
1062			c->mount_opts.unmount_mode = 1;
1063			break;
1064		case Opt_bulk_read:
1065			c->mount_opts.bulk_read = 2;
1066			c->bulk_read = 1;
1067			break;
1068		case Opt_no_bulk_read:
1069			c->mount_opts.bulk_read = 1;
1070			c->bulk_read = 0;
1071			break;
1072		case Opt_chk_data_crc:
1073			c->mount_opts.chk_data_crc = 2;
1074			c->no_chk_data_crc = 0;
1075			break;
1076		case Opt_no_chk_data_crc:
1077			c->mount_opts.chk_data_crc = 1;
1078			c->no_chk_data_crc = 1;
1079			break;
1080		case Opt_override_compr:
1081		{
1082			char *name = match_strdup(&args[0]);
1083
1084			if (!name)
1085				return -ENOMEM;
1086			if (!strcmp(name, "none"))
1087				c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1088			else if (!strcmp(name, "lzo"))
1089				c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1090			else if (!strcmp(name, "zlib"))
1091				c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1092			else if (!strcmp(name, "zstd"))
1093				c->mount_opts.compr_type = UBIFS_COMPR_ZSTD;
1094			else {
1095				ubifs_err(c, "unknown compressor \"%s\"", name); //FIXME: is c ready?
1096				kfree(name);
1097				return -EINVAL;
1098			}
1099			kfree(name);
1100			c->mount_opts.override_compr = 1;
1101			c->default_compr = c->mount_opts.compr_type;
1102			break;
1103		}
1104		case Opt_assert:
1105		{
1106			char *act = match_strdup(&args[0]);
1107
1108			if (!act)
1109				return -ENOMEM;
1110			if (!strcmp(act, "report"))
1111				c->assert_action = ASSACT_REPORT;
1112			else if (!strcmp(act, "read-only"))
1113				c->assert_action = ASSACT_RO;
1114			else if (!strcmp(act, "panic"))
1115				c->assert_action = ASSACT_PANIC;
1116			else {
1117				ubifs_err(c, "unknown assert action \"%s\"", act);
1118				kfree(act);
1119				return -EINVAL;
1120			}
1121			kfree(act);
1122			break;
1123		}
1124		case Opt_auth_key:
1125			if (!is_remount) {
1126				c->auth_key_name = kstrdup(args[0].from,
1127								GFP_KERNEL);
1128				if (!c->auth_key_name)
1129					return -ENOMEM;
1130			}
1131			break;
1132		case Opt_auth_hash_name:
1133			if (!is_remount) {
1134				c->auth_hash_name = kstrdup(args[0].from,
1135								GFP_KERNEL);
1136				if (!c->auth_hash_name)
1137					return -ENOMEM;
1138			}
1139			break;
1140		case Opt_ignore:
1141			break;
1142		default:
1143		{
1144			unsigned long flag;
1145			struct super_block *sb = c->vfs_sb;
1146
1147			flag = parse_standard_option(p);
1148			if (!flag) {
1149				ubifs_err(c, "unrecognized mount option \"%s\" or missing value",
1150					  p);
1151				return -EINVAL;
1152			}
1153			sb->s_flags |= flag;
1154			break;
1155		}
1156		}
1157	}
1158
1159	return 0;
1160}
1161
1162/*
1163 * ubifs_release_options - release mount parameters which have been dumped.
1164 * @c: UBIFS file-system description object
1165 */
1166static void ubifs_release_options(struct ubifs_info *c)
1167{
1168	kfree(c->auth_key_name);
1169	c->auth_key_name = NULL;
1170	kfree(c->auth_hash_name);
1171	c->auth_hash_name = NULL;
1172}
1173
1174/**
1175 * destroy_journal - destroy journal data structures.
1176 * @c: UBIFS file-system description object
1177 *
1178 * This function destroys journal data structures including those that may have
1179 * been created by recovery functions.
1180 */
1181static void destroy_journal(struct ubifs_info *c)
1182{
1183	while (!list_empty(&c->unclean_leb_list)) {
1184		struct ubifs_unclean_leb *ucleb;
1185
1186		ucleb = list_entry(c->unclean_leb_list.next,
1187				   struct ubifs_unclean_leb, list);
1188		list_del(&ucleb->list);
1189		kfree(ucleb);
1190	}
1191	while (!list_empty(&c->old_buds)) {
1192		struct ubifs_bud *bud;
1193
1194		bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1195		list_del(&bud->list);
1196		kfree(bud);
1197	}
1198	ubifs_destroy_idx_gc(c);
1199	ubifs_destroy_size_tree(c);
1200	ubifs_tnc_close(c);
1201	free_buds(c);
1202}
1203
1204/**
1205 * bu_init - initialize bulk-read information.
1206 * @c: UBIFS file-system description object
1207 */
1208static void bu_init(struct ubifs_info *c)
1209{
1210	ubifs_assert(c, c->bulk_read == 1);
1211
1212	if (c->bu.buf)
1213		return; /* Already initialized */
1214
1215again:
1216	c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1217	if (!c->bu.buf) {
1218		if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1219			c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1220			goto again;
1221		}
1222
1223		/* Just disable bulk-read */
1224		ubifs_warn(c, "cannot allocate %d bytes of memory for bulk-read, disabling it",
1225			   c->max_bu_buf_len);
1226		c->mount_opts.bulk_read = 1;
1227		c->bulk_read = 0;
1228		return;
1229	}
1230}
1231
1232/**
1233 * check_free_space - check if there is enough free space to mount.
1234 * @c: UBIFS file-system description object
1235 *
1236 * This function makes sure UBIFS has enough free space to be mounted in
1237 * read/write mode. UBIFS must always have some free space to allow deletions.
1238 */
1239static int check_free_space(struct ubifs_info *c)
1240{
1241	ubifs_assert(c, c->dark_wm > 0);
1242	if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1243		ubifs_err(c, "insufficient free space to mount in R/W mode");
1244		ubifs_dump_budg(c, &c->bi);
1245		ubifs_dump_lprops(c);
1246		return -ENOSPC;
1247	}
1248	return 0;
1249}
1250
1251/**
1252 * mount_ubifs - mount UBIFS file-system.
1253 * @c: UBIFS file-system description object
1254 *
1255 * This function mounts UBIFS file system. Returns zero in case of success and
1256 * a negative error code in case of failure.
1257 */
1258static int mount_ubifs(struct ubifs_info *c)
1259{
1260	int err;
1261	long long x, y;
1262	size_t sz;
1263
1264	c->ro_mount = !!sb_rdonly(c->vfs_sb);
1265	/* Suppress error messages while probing if SB_SILENT is set */
1266	c->probing = !!(c->vfs_sb->s_flags & SB_SILENT);
1267
1268	err = init_constants_early(c);
1269	if (err)
1270		return err;
1271
1272	err = ubifs_debugging_init(c);
1273	if (err)
1274		return err;
1275
1276	err = check_volume_empty(c);
1277	if (err)
1278		goto out_free;
1279
1280	if (c->empty && (c->ro_mount || c->ro_media)) {
1281		/*
1282		 * This UBI volume is empty, and read-only, or the file system
1283		 * is mounted read-only - we cannot format it.
1284		 */
1285		ubifs_err(c, "can't format empty UBI volume: read-only %s",
1286			  c->ro_media ? "UBI volume" : "mount");
1287		err = -EROFS;
1288		goto out_free;
1289	}
1290
1291	if (c->ro_media && !c->ro_mount) {
1292		ubifs_err(c, "cannot mount read-write - read-only media");
1293		err = -EROFS;
1294		goto out_free;
1295	}
1296
1297	/*
1298	 * The requirement for the buffer is that it should fit indexing B-tree
1299	 * height amount of integers. We assume the height if the TNC tree will
1300	 * never exceed 64.
1301	 */
1302	err = -ENOMEM;
1303	c->bottom_up_buf = kmalloc_array(BOTTOM_UP_HEIGHT, sizeof(int),
1304					 GFP_KERNEL);
1305	if (!c->bottom_up_buf)
1306		goto out_free;
1307
1308	c->sbuf = vmalloc(c->leb_size);
1309	if (!c->sbuf)
1310		goto out_free;
1311
1312	if (!c->ro_mount) {
1313		c->ileb_buf = vmalloc(c->leb_size);
1314		if (!c->ileb_buf)
1315			goto out_free;
1316	}
1317
1318	if (c->bulk_read == 1)
1319		bu_init(c);
1320
1321	if (!c->ro_mount) {
1322		c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ + \
1323					       UBIFS_CIPHER_BLOCK_SIZE,
1324					       GFP_KERNEL);
1325		if (!c->write_reserve_buf)
1326			goto out_free;
1327	}
1328
1329	c->mounting = 1;
1330
1331	if (c->auth_key_name) {
1332		if (IS_ENABLED(CONFIG_UBIFS_FS_AUTHENTICATION)) {
1333			err = ubifs_init_authentication(c);
1334			if (err)
1335				goto out_free;
1336		} else {
1337			ubifs_err(c, "auth_key_name, but UBIFS is built without"
1338				  " authentication support");
1339			err = -EINVAL;
1340			goto out_free;
1341		}
1342	}
1343
1344	err = ubifs_read_superblock(c);
1345	if (err)
1346		goto out_auth;
1347
1348	c->probing = 0;
1349
1350	/*
1351	 * Make sure the compressor which is set as default in the superblock
1352	 * or overridden by mount options is actually compiled in.
1353	 */
1354	if (!ubifs_compr_present(c, c->default_compr)) {
1355		ubifs_err(c, "'compressor \"%s\" is not compiled in",
1356			  ubifs_compr_name(c, c->default_compr));
1357		err = -ENOTSUPP;
1358		goto out_auth;
1359	}
1360
1361	err = init_constants_sb(c);
1362	if (err)
1363		goto out_auth;
1364
1365	sz = ALIGN(c->max_idx_node_sz, c->min_io_size) * 2;
1366	c->cbuf = kmalloc(sz, GFP_NOFS);
1367	if (!c->cbuf) {
1368		err = -ENOMEM;
1369		goto out_auth;
1370	}
1371
1372	err = alloc_wbufs(c);
1373	if (err)
1374		goto out_cbuf;
1375
1376	sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1377	if (!c->ro_mount) {
1378		/* Create background thread */
1379		c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1380		if (IS_ERR(c->bgt)) {
1381			err = PTR_ERR(c->bgt);
1382			c->bgt = NULL;
1383			ubifs_err(c, "cannot spawn \"%s\", error %d",
1384				  c->bgt_name, err);
1385			goto out_wbufs;
1386		}
1387		wake_up_process(c->bgt);
1388	}
1389
1390	err = ubifs_read_master(c);
1391	if (err)
1392		goto out_master;
1393
1394	init_constants_master(c);
1395
1396	if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1397		ubifs_msg(c, "recovery needed");
1398		c->need_recovery = 1;
1399	}
1400
1401	if (c->need_recovery && !c->ro_mount) {
1402		err = ubifs_recover_inl_heads(c, c->sbuf);
1403		if (err)
1404			goto out_master;
1405	}
1406
1407	err = ubifs_lpt_init(c, 1, !c->ro_mount);
1408	if (err)
1409		goto out_master;
1410
1411	if (!c->ro_mount && c->space_fixup) {
1412		err = ubifs_fixup_free_space(c);
1413		if (err)
1414			goto out_lpt;
1415	}
1416
1417	if (!c->ro_mount && !c->need_recovery) {
1418		/*
1419		 * Set the "dirty" flag so that if we reboot uncleanly we
1420		 * will notice this immediately on the next mount.
1421		 */
1422		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1423		err = ubifs_write_master(c);
1424		if (err)
1425			goto out_lpt;
1426	}
1427
1428	/*
1429	 * Handle offline signed images: Now that the master node is
1430	 * written and its validation no longer depends on the hash
1431	 * in the superblock, we can update the offline signed
1432	 * superblock with a HMAC version,
1433	 */
1434	if (ubifs_authenticated(c) && ubifs_hmac_zero(c, c->sup_node->hmac)) {
1435		err = ubifs_hmac_wkm(c, c->sup_node->hmac_wkm);
1436		if (err)
1437			goto out_lpt;
1438		c->superblock_need_write = 1;
1439	}
1440
1441	if (!c->ro_mount && c->superblock_need_write) {
1442		err = ubifs_write_sb_node(c, c->sup_node);
1443		if (err)
1444			goto out_lpt;
1445		c->superblock_need_write = 0;
1446	}
1447
1448	err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1449	if (err)
1450		goto out_lpt;
1451
1452	err = ubifs_replay_journal(c);
1453	if (err)
1454		goto out_journal;
1455
1456	/* Calculate 'min_idx_lebs' after journal replay */
1457	c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1458
1459	err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1460	if (err)
1461		goto out_orphans;
1462
1463	if (!c->ro_mount) {
1464		int lnum;
1465
1466		err = check_free_space(c);
1467		if (err)
1468			goto out_orphans;
1469
1470		/* Check for enough log space */
1471		lnum = c->lhead_lnum + 1;
1472		if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1473			lnum = UBIFS_LOG_LNUM;
1474		if (lnum == c->ltail_lnum) {
1475			err = ubifs_consolidate_log(c);
1476			if (err)
1477				goto out_orphans;
1478		}
1479
1480		if (c->need_recovery) {
1481			if (!ubifs_authenticated(c)) {
1482				err = ubifs_recover_size(c, true);
1483				if (err)
1484					goto out_orphans;
1485			}
1486
1487			err = ubifs_rcvry_gc_commit(c);
1488			if (err)
1489				goto out_orphans;
1490
1491			if (ubifs_authenticated(c)) {
1492				err = ubifs_recover_size(c, false);
1493				if (err)
1494					goto out_orphans;
1495			}
1496		} else {
1497			err = take_gc_lnum(c);
1498			if (err)
1499				goto out_orphans;
1500
1501			/*
1502			 * GC LEB may contain garbage if there was an unclean
1503			 * reboot, and it should be un-mapped.
1504			 */
1505			err = ubifs_leb_unmap(c, c->gc_lnum);
1506			if (err)
1507				goto out_orphans;
1508		}
1509
1510		err = dbg_check_lprops(c);
1511		if (err)
1512			goto out_orphans;
1513	} else if (c->need_recovery) {
1514		err = ubifs_recover_size(c, false);
1515		if (err)
1516			goto out_orphans;
1517	} else {
1518		/*
1519		 * Even if we mount read-only, we have to set space in GC LEB
1520		 * to proper value because this affects UBIFS free space
1521		 * reporting. We do not want to have a situation when
1522		 * re-mounting from R/O to R/W changes amount of free space.
1523		 */
1524		err = take_gc_lnum(c);
1525		if (err)
1526			goto out_orphans;
1527	}
1528
1529	spin_lock(&ubifs_infos_lock);
1530	list_add_tail(&c->infos_list, &ubifs_infos);
1531	spin_unlock(&ubifs_infos_lock);
1532
1533	if (c->need_recovery) {
1534		if (c->ro_mount)
1535			ubifs_msg(c, "recovery deferred");
1536		else {
1537			c->need_recovery = 0;
1538			ubifs_msg(c, "recovery completed");
1539			/*
1540			 * GC LEB has to be empty and taken at this point. But
1541			 * the journal head LEBs may also be accounted as
1542			 * "empty taken" if they are empty.
1543			 */
1544			ubifs_assert(c, c->lst.taken_empty_lebs > 0);
1545		}
1546	} else
1547		ubifs_assert(c, c->lst.taken_empty_lebs > 0);
1548
1549	err = dbg_check_filesystem(c);
1550	if (err)
1551		goto out_infos;
1552
1553	dbg_debugfs_init_fs(c);
1554
1555	c->mounting = 0;
1556
1557	ubifs_msg(c, "UBIFS: mounted UBI device %d, volume %d, name \"%s\"%s",
1558		  c->vi.ubi_num, c->vi.vol_id, c->vi.name,
1559		  c->ro_mount ? ", R/O mode" : "");
1560	x = (long long)c->main_lebs * c->leb_size;
1561	y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1562	ubifs_msg(c, "LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes",
1563		  c->leb_size, c->leb_size >> 10, c->min_io_size,
1564		  c->max_write_size);
1565	ubifs_msg(c, "FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)",
1566		  x, x >> 20, c->main_lebs,
1567		  y, y >> 20, c->log_lebs + c->max_bud_cnt);
1568	ubifs_msg(c, "reserved for root: %llu bytes (%llu KiB)",
1569		  c->report_rp_size, c->report_rp_size >> 10);
1570	ubifs_msg(c, "media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s",
1571		  c->fmt_version, c->ro_compat_version,
1572		  UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid,
1573		  c->big_lpt ? ", big LPT model" : ", small LPT model");
1574
1575	dbg_gen("default compressor:  %s", ubifs_compr_name(c, c->default_compr));
1576	dbg_gen("data journal heads:  %d",
1577		c->jhead_cnt - NONDATA_JHEADS_CNT);
1578	dbg_gen("log LEBs:            %d (%d - %d)",
1579		c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1580	dbg_gen("LPT area LEBs:       %d (%d - %d)",
1581		c->lpt_lebs, c->lpt_first, c->lpt_last);
1582	dbg_gen("orphan area LEBs:    %d (%d - %d)",
1583		c->orph_lebs, c->orph_first, c->orph_last);
1584	dbg_gen("main area LEBs:      %d (%d - %d)",
1585		c->main_lebs, c->main_first, c->leb_cnt - 1);
1586	dbg_gen("index LEBs:          %d", c->lst.idx_lebs);
1587	dbg_gen("total index bytes:   %lld (%lld KiB, %lld MiB)",
1588		c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1589		c->bi.old_idx_sz >> 20);
1590	dbg_gen("key hash type:       %d", c->key_hash_type);
1591	dbg_gen("tree fanout:         %d", c->fanout);
1592	dbg_gen("reserved GC LEB:     %d", c->gc_lnum);
1593	dbg_gen("max. znode size      %d", c->max_znode_sz);
1594	dbg_gen("max. index node size %d", c->max_idx_node_sz);
1595	dbg_gen("node sizes:          data %zu, inode %zu, dentry %zu",
1596		UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1597	dbg_gen("node sizes:          trun %zu, sb %zu, master %zu",
1598		UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1599	dbg_gen("node sizes:          ref %zu, cmt. start %zu, orph %zu",
1600		UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1601	dbg_gen("max. node sizes:     data %zu, inode %zu dentry %zu, idx %d",
1602		UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1603		UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1604	dbg_gen("dead watermark:      %d", c->dead_wm);
1605	dbg_gen("dark watermark:      %d", c->dark_wm);
1606	dbg_gen("LEB overhead:        %d", c->leb_overhead);
1607	x = (long long)c->main_lebs * c->dark_wm;
1608	dbg_gen("max. dark space:     %lld (%lld KiB, %lld MiB)",
1609		x, x >> 10, x >> 20);
1610	dbg_gen("maximum bud bytes:   %lld (%lld KiB, %lld MiB)",
1611		c->max_bud_bytes, c->max_bud_bytes >> 10,
1612		c->max_bud_bytes >> 20);
1613	dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1614		c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1615		c->bg_bud_bytes >> 20);
1616	dbg_gen("current bud bytes    %lld (%lld KiB, %lld MiB)",
1617		c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1618	dbg_gen("max. seq. number:    %llu", c->max_sqnum);
1619	dbg_gen("commit number:       %llu", c->cmt_no);
1620	dbg_gen("max. xattrs per inode: %d", ubifs_xattr_max_cnt(c));
1621	dbg_gen("max orphans:           %d", c->max_orphans);
1622
1623	return 0;
1624
1625out_infos:
1626	spin_lock(&ubifs_infos_lock);
1627	list_del(&c->infos_list);
1628	spin_unlock(&ubifs_infos_lock);
1629out_orphans:
1630	free_orphans(c);
1631out_journal:
1632	destroy_journal(c);
1633out_lpt:
1634	ubifs_lpt_free(c, 0);
1635out_master:
1636	kfree(c->mst_node);
1637	kfree(c->rcvrd_mst_node);
1638	if (c->bgt)
1639		kthread_stop(c->bgt);
1640out_wbufs:
1641	free_wbufs(c);
1642out_cbuf:
1643	kfree(c->cbuf);
1644out_auth:
1645	ubifs_exit_authentication(c);
1646out_free:
1647	kfree(c->write_reserve_buf);
1648	kfree(c->bu.buf);
1649	vfree(c->ileb_buf);
1650	vfree(c->sbuf);
1651	kfree(c->bottom_up_buf);
1652	kfree(c->sup_node);
1653	ubifs_debugging_exit(c);
1654	return err;
1655}
1656
1657/**
1658 * ubifs_umount - un-mount UBIFS file-system.
1659 * @c: UBIFS file-system description object
1660 *
1661 * Note, this function is called to free allocated resourced when un-mounting,
1662 * as well as free resources when an error occurred while we were half way
1663 * through mounting (error path cleanup function). So it has to make sure the
1664 * resource was actually allocated before freeing it.
1665 */
1666static void ubifs_umount(struct ubifs_info *c)
1667{
1668	dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1669		c->vi.vol_id);
1670
1671	dbg_debugfs_exit_fs(c);
1672	spin_lock(&ubifs_infos_lock);
1673	list_del(&c->infos_list);
1674	spin_unlock(&ubifs_infos_lock);
1675
1676	if (c->bgt)
1677		kthread_stop(c->bgt);
1678
1679	destroy_journal(c);
1680	free_wbufs(c);
1681	free_orphans(c);
1682	ubifs_lpt_free(c, 0);
1683	ubifs_exit_authentication(c);
1684
1685	ubifs_release_options(c);
1686	kfree(c->cbuf);
1687	kfree(c->rcvrd_mst_node);
1688	kfree(c->mst_node);
1689	kfree(c->write_reserve_buf);
1690	kfree(c->bu.buf);
1691	vfree(c->ileb_buf);
1692	vfree(c->sbuf);
1693	kfree(c->bottom_up_buf);
1694	kfree(c->sup_node);
1695	ubifs_debugging_exit(c);
1696}
1697
1698/**
1699 * ubifs_remount_rw - re-mount in read-write mode.
1700 * @c: UBIFS file-system description object
1701 *
1702 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1703 * mode. This function allocates the needed resources and re-mounts UBIFS in
1704 * read-write mode.
1705 */
1706static int ubifs_remount_rw(struct ubifs_info *c)
1707{
1708	int err, lnum;
1709
1710	if (c->rw_incompat) {
1711		ubifs_err(c, "the file-system is not R/W-compatible");
1712		ubifs_msg(c, "on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
1713			  c->fmt_version, c->ro_compat_version,
1714			  UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1715		return -EROFS;
1716	}
1717
1718	mutex_lock(&c->umount_mutex);
1719	dbg_save_space_info(c);
1720	c->remounting_rw = 1;
1721	c->ro_mount = 0;
1722
1723	if (c->space_fixup) {
1724		err = ubifs_fixup_free_space(c);
1725		if (err)
1726			goto out;
1727	}
1728
1729	err = check_free_space(c);
1730	if (err)
1731		goto out;
1732
1733	if (c->need_recovery) {
1734		ubifs_msg(c, "completing deferred recovery");
1735		err = ubifs_write_rcvrd_mst_node(c);
1736		if (err)
1737			goto out;
1738		if (!ubifs_authenticated(c)) {
1739			err = ubifs_recover_size(c, true);
1740			if (err)
1741				goto out;
1742		}
1743		err = ubifs_clean_lebs(c, c->sbuf);
1744		if (err)
1745			goto out;
1746		err = ubifs_recover_inl_heads(c, c->sbuf);
1747		if (err)
1748			goto out;
1749	} else {
1750		/* A readonly mount is not allowed to have orphans */
1751		ubifs_assert(c, c->tot_orphans == 0);
1752		err = ubifs_clear_orphans(c);
1753		if (err)
1754			goto out;
1755	}
1756
1757	if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1758		c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1759		err = ubifs_write_master(c);
1760		if (err)
1761			goto out;
1762	}
1763
1764	if (c->superblock_need_write) {
1765		struct ubifs_sb_node *sup = c->sup_node;
1766
1767		err = ubifs_write_sb_node(c, sup);
1768		if (err)
1769			goto out;
1770
1771		c->superblock_need_write = 0;
1772	}
1773
1774	c->ileb_buf = vmalloc(c->leb_size);
1775	if (!c->ileb_buf) {
1776		err = -ENOMEM;
1777		goto out;
1778	}
1779
1780	c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ + \
1781				       UBIFS_CIPHER_BLOCK_SIZE, GFP_KERNEL);
1782	if (!c->write_reserve_buf) {
1783		err = -ENOMEM;
1784		goto out;
1785	}
1786
1787	err = ubifs_lpt_init(c, 0, 1);
1788	if (err)
1789		goto out;
1790
1791	/* Create background thread */
1792	c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1793	if (IS_ERR(c->bgt)) {
1794		err = PTR_ERR(c->bgt);
1795		c->bgt = NULL;
1796		ubifs_err(c, "cannot spawn \"%s\", error %d",
1797			  c->bgt_name, err);
1798		goto out;
1799	}
1800	wake_up_process(c->bgt);
1801
1802	c->orph_buf = vmalloc(c->leb_size);
1803	if (!c->orph_buf) {
1804		err = -ENOMEM;
1805		goto out;
1806	}
1807
1808	/* Check for enough log space */
1809	lnum = c->lhead_lnum + 1;
1810	if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1811		lnum = UBIFS_LOG_LNUM;
1812	if (lnum == c->ltail_lnum) {
1813		err = ubifs_consolidate_log(c);
1814		if (err)
1815			goto out;
1816	}
1817
1818	if (c->need_recovery) {
1819		err = ubifs_rcvry_gc_commit(c);
1820		if (err)
1821			goto out;
1822
1823		if (ubifs_authenticated(c)) {
1824			err = ubifs_recover_size(c, false);
1825			if (err)
1826				goto out;
1827		}
1828	} else {
1829		err = ubifs_leb_unmap(c, c->gc_lnum);
1830	}
1831	if (err)
1832		goto out;
1833
1834	dbg_gen("re-mounted read-write");
1835	c->remounting_rw = 0;
1836
1837	if (c->need_recovery) {
1838		c->need_recovery = 0;
1839		ubifs_msg(c, "deferred recovery completed");
1840	} else {
1841		/*
1842		 * Do not run the debugging space check if the were doing
1843		 * recovery, because when we saved the information we had the
1844		 * file-system in a state where the TNC and lprops has been
1845		 * modified in memory, but all the I/O operations (including a
1846		 * commit) were deferred. So the file-system was in
1847		 * "non-committed" state. Now the file-system is in committed
1848		 * state, and of course the amount of free space will change
1849		 * because, for example, the old index size was imprecise.
1850		 */
1851		err = dbg_check_space_info(c);
1852	}
1853
1854	mutex_unlock(&c->umount_mutex);
1855	return err;
1856
1857out:
1858	c->ro_mount = 1;
1859	vfree(c->orph_buf);
1860	c->orph_buf = NULL;
1861	if (c->bgt) {
1862		kthread_stop(c->bgt);
1863		c->bgt = NULL;
1864	}
1865	kfree(c->write_reserve_buf);
1866	c->write_reserve_buf = NULL;
1867	vfree(c->ileb_buf);
1868	c->ileb_buf = NULL;
1869	ubifs_lpt_free(c, 1);
1870	c->remounting_rw = 0;
1871	mutex_unlock(&c->umount_mutex);
1872	return err;
1873}
1874
1875/**
1876 * ubifs_remount_ro - re-mount in read-only mode.
1877 * @c: UBIFS file-system description object
1878 *
1879 * We assume VFS has stopped writing. Possibly the background thread could be
1880 * running a commit, however kthread_stop will wait in that case.
1881 */
1882static void ubifs_remount_ro(struct ubifs_info *c)
1883{
1884	int i, err;
1885
1886	ubifs_assert(c, !c->need_recovery);
1887	ubifs_assert(c, !c->ro_mount);
1888
1889	mutex_lock(&c->umount_mutex);
1890	if (c->bgt) {
1891		kthread_stop(c->bgt);
1892		c->bgt = NULL;
1893	}
1894
1895	dbg_save_space_info(c);
1896
1897	for (i = 0; i < c->jhead_cnt; i++) {
1898		err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
1899		if (err)
1900			ubifs_ro_mode(c, err);
1901	}
1902
1903	c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1904	c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1905	c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1906	err = ubifs_write_master(c);
1907	if (err)
1908		ubifs_ro_mode(c, err);
1909
1910	vfree(c->orph_buf);
1911	c->orph_buf = NULL;
1912	kfree(c->write_reserve_buf);
1913	c->write_reserve_buf = NULL;
1914	vfree(c->ileb_buf);
1915	c->ileb_buf = NULL;
1916	ubifs_lpt_free(c, 1);
1917	c->ro_mount = 1;
1918	err = dbg_check_space_info(c);
1919	if (err)
1920		ubifs_ro_mode(c, err);
1921	mutex_unlock(&c->umount_mutex);
1922}
1923
1924static void ubifs_put_super(struct super_block *sb)
1925{
1926	int i;
1927	struct ubifs_info *c = sb->s_fs_info;
1928
1929	ubifs_msg(c, "un-mount UBI device %d", c->vi.ubi_num);
1930
1931	/*
1932	 * The following asserts are only valid if there has not been a failure
1933	 * of the media. For example, there will be dirty inodes if we failed
1934	 * to write them back because of I/O errors.
1935	 */
1936	if (!c->ro_error) {
1937		ubifs_assert(c, c->bi.idx_growth == 0);
1938		ubifs_assert(c, c->bi.dd_growth == 0);
1939		ubifs_assert(c, c->bi.data_growth == 0);
1940	}
1941
1942	/*
1943	 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1944	 * and file system un-mount. Namely, it prevents the shrinker from
1945	 * picking this superblock for shrinking - it will be just skipped if
1946	 * the mutex is locked.
1947	 */
1948	mutex_lock(&c->umount_mutex);
1949	if (!c->ro_mount) {
1950		/*
1951		 * First of all kill the background thread to make sure it does
1952		 * not interfere with un-mounting and freeing resources.
1953		 */
1954		if (c->bgt) {
1955			kthread_stop(c->bgt);
1956			c->bgt = NULL;
1957		}
1958
1959		/*
1960		 * On fatal errors c->ro_error is set to 1, in which case we do
1961		 * not write the master node.
1962		 */
1963		if (!c->ro_error) {
1964			int err;
1965
1966			/* Synchronize write-buffers */
1967			for (i = 0; i < c->jhead_cnt; i++) {
1968				err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
1969				if (err)
1970					ubifs_ro_mode(c, err);
1971			}
1972
1973			/*
1974			 * We are being cleanly unmounted which means the
1975			 * orphans were killed - indicate this in the master
1976			 * node. Also save the reserved GC LEB number.
1977			 */
1978			c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1979			c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1980			c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1981			err = ubifs_write_master(c);
1982			if (err)
1983				/*
1984				 * Recovery will attempt to fix the master area
1985				 * next mount, so we just print a message and
1986				 * continue to unmount normally.
1987				 */
1988				ubifs_err(c, "failed to write master node, error %d",
1989					  err);
1990		} else {
1991			for (i = 0; i < c->jhead_cnt; i++)
1992				/* Make sure write-buffer timers are canceled */
1993				hrtimer_cancel(&c->jheads[i].wbuf.timer);
1994		}
1995	}
1996
1997	ubifs_umount(c);
1998	ubi_close_volume(c->ubi);
1999	mutex_unlock(&c->umount_mutex);
2000}
2001
2002static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
2003{
2004	int err;
2005	struct ubifs_info *c = sb->s_fs_info;
2006
2007	sync_filesystem(sb);
2008	dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
2009
2010	err = ubifs_parse_options(c, data, 1);
2011	if (err) {
2012		ubifs_err(c, "invalid or unknown remount parameter");
2013		return err;
2014	}
2015
2016	if (c->ro_mount && !(*flags & SB_RDONLY)) {
2017		if (c->ro_error) {
2018			ubifs_msg(c, "cannot re-mount R/W due to prior errors");
2019			return -EROFS;
2020		}
2021		if (c->ro_media) {
2022			ubifs_msg(c, "cannot re-mount R/W - UBI volume is R/O");
2023			return -EROFS;
2024		}
2025		err = ubifs_remount_rw(c);
2026		if (err)
2027			return err;
2028	} else if (!c->ro_mount && (*flags & SB_RDONLY)) {
2029		if (c->ro_error) {
2030			ubifs_msg(c, "cannot re-mount R/O due to prior errors");
2031			return -EROFS;
2032		}
2033		ubifs_remount_ro(c);
2034	}
2035
2036	if (c->bulk_read == 1)
2037		bu_init(c);
2038	else {
2039		dbg_gen("disable bulk-read");
2040		mutex_lock(&c->bu_mutex);
2041		kfree(c->bu.buf);
2042		c->bu.buf = NULL;
2043		mutex_unlock(&c->bu_mutex);
2044	}
2045
2046	if (!c->need_recovery)
2047		ubifs_assert(c, c->lst.taken_empty_lebs > 0);
2048
2049	return 0;
2050}
2051
2052const struct super_operations ubifs_super_operations = {
2053	.alloc_inode   = ubifs_alloc_inode,
2054	.free_inode    = ubifs_free_inode,
2055	.put_super     = ubifs_put_super,
2056	.write_inode   = ubifs_write_inode,
2057	.drop_inode    = ubifs_drop_inode,
2058	.evict_inode   = ubifs_evict_inode,
2059	.statfs        = ubifs_statfs,
2060	.dirty_inode   = ubifs_dirty_inode,
2061	.remount_fs    = ubifs_remount_fs,
2062	.show_options  = ubifs_show_options,
2063	.sync_fs       = ubifs_sync_fs,
2064};
2065
2066/**
2067 * open_ubi - parse UBI device name string and open the UBI device.
2068 * @name: UBI volume name
2069 * @mode: UBI volume open mode
2070 *
2071 * The primary method of mounting UBIFS is by specifying the UBI volume
2072 * character device node path. However, UBIFS may also be mounted withoug any
2073 * character device node using one of the following methods:
2074 *
2075 * o ubiX_Y    - mount UBI device number X, volume Y;
2076 * o ubiY      - mount UBI device number 0, volume Y;
2077 * o ubiX:NAME - mount UBI device X, volume with name NAME;
2078 * o ubi:NAME  - mount UBI device 0, volume with name NAME.
2079 *
2080 * Alternative '!' separator may be used instead of ':' (because some shells
2081 * like busybox may interpret ':' as an NFS host name separator). This function
2082 * returns UBI volume description object in case of success and a negative
2083 * error code in case of failure.
2084 */
2085static struct ubi_volume_desc *open_ubi(const char *name, int mode)
2086{
2087	struct ubi_volume_desc *ubi;
2088	int dev, vol;
2089	char *endptr;
2090
2091	if (!name || !*name)
2092		return ERR_PTR(-EINVAL);
2093
2094	/* First, try to open using the device node path method */
2095	ubi = ubi_open_volume_path(name, mode);
2096	if (!IS_ERR(ubi))
2097		return ubi;
2098
2099	/* Try the "nodev" method */
2100	if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
2101		return ERR_PTR(-EINVAL);
2102
2103	/* ubi:NAME method */
2104	if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
2105		return ubi_open_volume_nm(0, name + 4, mode);
2106
2107	if (!isdigit(name[3]))
2108		return ERR_PTR(-EINVAL);
2109
2110	dev = simple_strtoul(name + 3, &endptr, 0);
2111
2112	/* ubiY method */
2113	if (*endptr == '\0')
2114		return ubi_open_volume(0, dev, mode);
2115
2116	/* ubiX_Y method */
2117	if (*endptr == '_' && isdigit(endptr[1])) {
2118		vol = simple_strtoul(endptr + 1, &endptr, 0);
2119		if (*endptr != '\0')
2120			return ERR_PTR(-EINVAL);
2121		return ubi_open_volume(dev, vol, mode);
2122	}
2123
2124	/* ubiX:NAME method */
2125	if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
2126		return ubi_open_volume_nm(dev, ++endptr, mode);
2127
2128	return ERR_PTR(-EINVAL);
2129}
2130
2131static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
2132{
2133	struct ubifs_info *c;
2134
2135	c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
2136	if (c) {
2137		spin_lock_init(&c->cnt_lock);
2138		spin_lock_init(&c->cs_lock);
2139		spin_lock_init(&c->buds_lock);
2140		spin_lock_init(&c->space_lock);
2141		spin_lock_init(&c->orphan_lock);
2142		init_rwsem(&c->commit_sem);
2143		mutex_init(&c->lp_mutex);
2144		mutex_init(&c->tnc_mutex);
2145		mutex_init(&c->log_mutex);
2146		mutex_init(&c->umount_mutex);
2147		mutex_init(&c->bu_mutex);
2148		mutex_init(&c->write_reserve_mutex);
2149		init_waitqueue_head(&c->cmt_wq);
2150		c->buds = RB_ROOT;
2151		c->old_idx = RB_ROOT;
2152		c->size_tree = RB_ROOT;
2153		c->orph_tree = RB_ROOT;
2154		INIT_LIST_HEAD(&c->infos_list);
2155		INIT_LIST_HEAD(&c->idx_gc);
2156		INIT_LIST_HEAD(&c->replay_list);
2157		INIT_LIST_HEAD(&c->replay_buds);
2158		INIT_LIST_HEAD(&c->uncat_list);
2159		INIT_LIST_HEAD(&c->empty_list);
2160		INIT_LIST_HEAD(&c->freeable_list);
2161		INIT_LIST_HEAD(&c->frdi_idx_list);
2162		INIT_LIST_HEAD(&c->unclean_leb_list);
2163		INIT_LIST_HEAD(&c->old_buds);
2164		INIT_LIST_HEAD(&c->orph_list);
2165		INIT_LIST_HEAD(&c->orph_new);
2166		c->no_chk_data_crc = 1;
2167		c->assert_action = ASSACT_RO;
2168
2169		c->highest_inum = UBIFS_FIRST_INO;
2170		c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
2171
2172		ubi_get_volume_info(ubi, &c->vi);
2173		ubi_get_device_info(c->vi.ubi_num, &c->di);
2174	}
2175	return c;
2176}
2177
2178static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
2179{
2180	struct ubifs_info *c = sb->s_fs_info;
2181	struct inode *root;
2182	int err;
2183
2184	c->vfs_sb = sb;
2185	/* Re-open the UBI device in read-write mode */
2186	c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2187	if (IS_ERR(c->ubi)) {
2188		err = PTR_ERR(c->ubi);
2189		goto out;
2190	}
2191
2192	err = ubifs_parse_options(c, data, 0);
2193	if (err)
2194		goto out_close;
2195
2196	/*
2197	 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2198	 * UBIFS, I/O is not deferred, it is done immediately in readpage,
2199	 * which means the user would have to wait not just for their own I/O
2200	 * but the read-ahead I/O as well i.e. completely pointless.
2201	 *
2202	 * Read-ahead will be disabled because @sb->s_bdi->ra_pages is 0. Also
2203	 * @sb->s_bdi->capabilities are initialized to 0 so there won't be any
2204	 * writeback happening.
2205	 */
2206	err = super_setup_bdi_name(sb, "ubifs_%d_%d", c->vi.ubi_num,
2207				   c->vi.vol_id);
2208	if (err)
2209		goto out_close;
2210	sb->s_bdi->ra_pages = 0;
2211	sb->s_bdi->io_pages = 0;
2212
2213	sb->s_fs_info = c;
2214	sb->s_magic = UBIFS_SUPER_MAGIC;
2215	sb->s_blocksize = UBIFS_BLOCK_SIZE;
2216	sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2217	sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2218	if (c->max_inode_sz > MAX_LFS_FILESIZE)
2219		sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2220	sb->s_op = &ubifs_super_operations;
2221#ifdef CONFIG_UBIFS_FS_XATTR
2222	sb->s_xattr = ubifs_xattr_handlers;
2223#endif
2224	fscrypt_set_ops(sb, &ubifs_crypt_operations);
2225
2226	mutex_lock(&c->umount_mutex);
2227	err = mount_ubifs(c);
2228	if (err) {
2229		ubifs_assert(c, err < 0);
2230		goto out_unlock;
2231	}
2232
2233	/* Read the root inode */
2234	root = ubifs_iget(sb, UBIFS_ROOT_INO);
2235	if (IS_ERR(root)) {
2236		err = PTR_ERR(root);
2237		goto out_umount;
2238	}
2239
2240	sb->s_root = d_make_root(root);
2241	if (!sb->s_root) {
2242		err = -ENOMEM;
2243		goto out_umount;
2244	}
2245
2246	mutex_unlock(&c->umount_mutex);
2247	return 0;
2248
2249out_umount:
2250	ubifs_umount(c);
2251out_unlock:
2252	mutex_unlock(&c->umount_mutex);
2253out_close:
2254	ubifs_release_options(c);
2255	ubi_close_volume(c->ubi);
2256out:
2257	return err;
2258}
2259
2260static int sb_test(struct super_block *sb, void *data)
2261{
2262	struct ubifs_info *c1 = data;
2263	struct ubifs_info *c = sb->s_fs_info;
2264
2265	return c->vi.cdev == c1->vi.cdev;
2266}
2267
2268static int sb_set(struct super_block *sb, void *data)
2269{
2270	sb->s_fs_info = data;
2271	return set_anon_super(sb, NULL);
2272}
2273
2274static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2275			const char *name, void *data)
2276{
2277	struct ubi_volume_desc *ubi;
2278	struct ubifs_info *c;
2279	struct super_block *sb;
2280	int err;
2281
2282	dbg_gen("name %s, flags %#x", name, flags);
2283
2284	/*
2285	 * Get UBI device number and volume ID. Mount it read-only so far
2286	 * because this might be a new mount point, and UBI allows only one
2287	 * read-write user at a time.
2288	 */
2289	ubi = open_ubi(name, UBI_READONLY);
2290	if (IS_ERR(ubi)) {
2291		if (!(flags & SB_SILENT))
2292			pr_err("UBIFS error (pid: %d): cannot open \"%s\", error %d",
2293			       current->pid, name, (int)PTR_ERR(ubi));
2294		return ERR_CAST(ubi);
2295	}
2296
2297	c = alloc_ubifs_info(ubi);
2298	if (!c) {
2299		err = -ENOMEM;
2300		goto out_close;
2301	}
2302
2303	dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2304
2305	sb = sget(fs_type, sb_test, sb_set, flags, c);
2306	if (IS_ERR(sb)) {
2307		err = PTR_ERR(sb);
2308		kfree(c);
2309		goto out_close;
2310	}
2311
2312	if (sb->s_root) {
2313		struct ubifs_info *c1 = sb->s_fs_info;
2314		kfree(c);
2315		/* A new mount point for already mounted UBIFS */
2316		dbg_gen("this ubi volume is already mounted");
2317		if (!!(flags & SB_RDONLY) != c1->ro_mount) {
2318			err = -EBUSY;
2319			goto out_deact;
2320		}
2321	} else {
2322		err = ubifs_fill_super(sb, data, flags & SB_SILENT ? 1 : 0);
2323		if (err)
2324			goto out_deact;
2325		/* We do not support atime */
2326		sb->s_flags |= SB_ACTIVE;
2327		if (IS_ENABLED(CONFIG_UBIFS_ATIME_SUPPORT))
2328			ubifs_msg(c, "full atime support is enabled.");
2329		else
2330			sb->s_flags |= SB_NOATIME;
2331	}
2332
2333	/* 'fill_super()' opens ubi again so we must close it here */
2334	ubi_close_volume(ubi);
2335
2336	return dget(sb->s_root);
2337
2338out_deact:
2339	deactivate_locked_super(sb);
2340out_close:
2341	ubi_close_volume(ubi);
2342	return ERR_PTR(err);
2343}
2344
2345static void kill_ubifs_super(struct super_block *s)
2346{
2347	struct ubifs_info *c = s->s_fs_info;
2348	kill_anon_super(s);
2349	kfree(c);
2350}
2351
2352static struct file_system_type ubifs_fs_type = {
2353	.name    = "ubifs",
2354	.owner   = THIS_MODULE,
2355	.mount   = ubifs_mount,
2356	.kill_sb = kill_ubifs_super,
2357};
2358MODULE_ALIAS_FS("ubifs");
2359
2360/*
2361 * Inode slab cache constructor.
2362 */
2363static void inode_slab_ctor(void *obj)
2364{
2365	struct ubifs_inode *ui = obj;
2366	inode_init_once(&ui->vfs_inode);
2367}
2368
2369static int __init ubifs_init(void)
2370{
2371	int err;
2372
2373	BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2374
2375	/* Make sure node sizes are 8-byte aligned */
2376	BUILD_BUG_ON(UBIFS_CH_SZ        & 7);
2377	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  & 7);
2378	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2379	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2380	BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2381	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2382	BUILD_BUG_ON(UBIFS_SB_NODE_SZ   & 7);
2383	BUILD_BUG_ON(UBIFS_MST_NODE_SZ  & 7);
2384	BUILD_BUG_ON(UBIFS_REF_NODE_SZ  & 7);
2385	BUILD_BUG_ON(UBIFS_CS_NODE_SZ   & 7);
2386	BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2387
2388	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2389	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2390	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2391	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  & 7);
2392	BUILD_BUG_ON(UBIFS_MAX_NODE_SZ      & 7);
2393	BUILD_BUG_ON(MIN_WRITE_SZ           & 7);
2394
2395	/* Check min. node size */
2396	BUILD_BUG_ON(UBIFS_INO_NODE_SZ  < MIN_WRITE_SZ);
2397	BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2398	BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2399	BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2400
2401	BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2402	BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2403	BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2404	BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ  > UBIFS_MAX_NODE_SZ);
2405
2406	/* Defined node sizes */
2407	BUILD_BUG_ON(UBIFS_SB_NODE_SZ  != 4096);
2408	BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2409	BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2410	BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2411
2412	/*
2413	 * We use 2 bit wide bit-fields to store compression type, which should
2414	 * be amended if more compressors are added. The bit-fields are:
2415	 * @compr_type in 'struct ubifs_inode', @default_compr in
2416	 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2417	 */
2418	BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2419
2420	/*
2421	 * We require that PAGE_SIZE is greater-than-or-equal-to
2422	 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2423	 */
2424	if (PAGE_SIZE < UBIFS_BLOCK_SIZE) {
2425		pr_err("UBIFS error (pid %d): VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
2426		       current->pid, (unsigned int)PAGE_SIZE);
2427		return -EINVAL;
2428	}
2429
2430	ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2431				sizeof(struct ubifs_inode), 0,
2432				SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT |
2433				SLAB_ACCOUNT, &inode_slab_ctor);
2434	if (!ubifs_inode_slab)
2435		return -ENOMEM;
2436
2437	err = register_shrinker(&ubifs_shrinker_info);
2438	if (err)
2439		goto out_slab;
2440
2441	err = ubifs_compressors_init();
2442	if (err)
2443		goto out_shrinker;
2444
2445	dbg_debugfs_init();
2446
2447	err = register_filesystem(&ubifs_fs_type);
2448	if (err) {
2449		pr_err("UBIFS error (pid %d): cannot register file system, error %d",
2450		       current->pid, err);
2451		goto out_dbg;
2452	}
2453	return 0;
2454
2455out_dbg:
2456	dbg_debugfs_exit();
2457	ubifs_compressors_exit();
2458out_shrinker:
2459	unregister_shrinker(&ubifs_shrinker_info);
2460out_slab:
2461	kmem_cache_destroy(ubifs_inode_slab);
2462	return err;
2463}
2464/* late_initcall to let compressors initialize first */
2465late_initcall(ubifs_init);
2466
2467static void __exit ubifs_exit(void)
2468{
2469	WARN_ON(!list_empty(&ubifs_infos));
2470	WARN_ON(atomic_long_read(&ubifs_clean_zn_cnt) != 0);
2471
2472	dbg_debugfs_exit();
2473	ubifs_compressors_exit();
2474	unregister_shrinker(&ubifs_shrinker_info);
2475
2476	/*
2477	 * Make sure all delayed rcu free inodes are flushed before we
2478	 * destroy cache.
2479	 */
2480	rcu_barrier();
2481	kmem_cache_destroy(ubifs_inode_slab);
2482	unregister_filesystem(&ubifs_fs_type);
2483}
2484module_exit(ubifs_exit);
2485
2486MODULE_LICENSE("GPL");
2487MODULE_VERSION(__stringify(UBIFS_VERSION));
2488MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2489MODULE_DESCRIPTION("UBIFS - UBI File System");
2490