]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/ubifs/super.c
UBIFS: return proper error code if the compr is not present
[mirror_ubuntu-artful-kernel.git] / fs / ubifs / super.c
CommitLineData
1e51764a
AB
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>
1e51764a
AB
33#include <linux/kthread.h>
34#include <linux/parser.h>
35#include <linux/seq_file.h>
36#include <linux/mount.h>
4d61db4f 37#include <linux/math64.h>
304d427c 38#include <linux/writeback.h>
1e51764a
AB
39#include "ubifs.h"
40
39ce81ce
AB
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
1e51764a
AB
47/* Slab cache for UBIFS inodes */
48struct kmem_cache *ubifs_inode_slab;
49
50/* UBIFS TNC shrinker description */
51static struct shrinker ubifs_shrinker_info = {
52 .shrink = ubifs_shrinker,
53 .seeks = DEFAULT_SEEKS,
54};
55
56/**
57 * validate_inode - validate inode.
58 * @c: UBIFS file-system description object
59 * @inode: the inode to validate
60 *
61 * This is a helper function for 'ubifs_iget()' which validates various fields
62 * of a newly built inode to make sure they contain sane values and prevent
63 * possible vulnerabilities. Returns zero if the inode is all right and
64 * a non-zero error code if not.
65 */
66static int validate_inode(struct ubifs_info *c, const struct inode *inode)
67{
68 int err;
69 const struct ubifs_inode *ui = ubifs_inode(inode);
70
71 if (inode->i_size > c->max_inode_sz) {
72 ubifs_err("inode is too large (%lld)",
73 (long long)inode->i_size);
74 return 1;
75 }
76
77 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
78 ubifs_err("unknown compression type %d", ui->compr_type);
79 return 2;
80 }
81
82 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
83 return 3;
84
85 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
86 return 4;
87
88 if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
89 return 5;
90
91 if (!ubifs_compr_present(ui->compr_type)) {
92 ubifs_warn("inode %lu uses '%s' compression, but it was not "
93 "compiled in", inode->i_ino,
94 ubifs_compr_name(ui->compr_type));
95 }
96
97 err = dbg_check_dir_size(c, inode);
98 return err;
99}
100
101struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
102{
103 int err;
104 union ubifs_key key;
105 struct ubifs_ino_node *ino;
106 struct ubifs_info *c = sb->s_fs_info;
107 struct inode *inode;
108 struct ubifs_inode *ui;
109
110 dbg_gen("inode %lu", inum);
111
112 inode = iget_locked(sb, inum);
113 if (!inode)
114 return ERR_PTR(-ENOMEM);
115 if (!(inode->i_state & I_NEW))
116 return inode;
117 ui = ubifs_inode(inode);
118
119 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
120 if (!ino) {
121 err = -ENOMEM;
122 goto out;
123 }
124
125 ino_key_init(c, &key, inode->i_ino);
126
127 err = ubifs_tnc_lookup(c, &key, ino);
128 if (err)
129 goto out_ino;
130
131 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
132 inode->i_nlink = le32_to_cpu(ino->nlink);
133 inode->i_uid = le32_to_cpu(ino->uid);
134 inode->i_gid = le32_to_cpu(ino->gid);
135 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
136 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
137 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
138 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
139 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
140 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
141 inode->i_mode = le32_to_cpu(ino->mode);
142 inode->i_size = le64_to_cpu(ino->size);
143
144 ui->data_len = le32_to_cpu(ino->data_len);
145 ui->flags = le32_to_cpu(ino->flags);
146 ui->compr_type = le16_to_cpu(ino->compr_type);
147 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
148 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
149 ui->xattr_size = le32_to_cpu(ino->xattr_size);
150 ui->xattr_names = le32_to_cpu(ino->xattr_names);
151 ui->synced_i_size = ui->ui_size = inode->i_size;
152
153 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
154
155 err = validate_inode(c, inode);
156 if (err)
157 goto out_invalid;
158
0a883a05 159 /* Disable read-ahead */
1e51764a
AB
160 inode->i_mapping->backing_dev_info = &c->bdi;
161
162 switch (inode->i_mode & S_IFMT) {
163 case S_IFREG:
164 inode->i_mapping->a_ops = &ubifs_file_address_operations;
165 inode->i_op = &ubifs_file_inode_operations;
166 inode->i_fop = &ubifs_file_operations;
167 if (ui->xattr) {
168 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
169 if (!ui->data) {
170 err = -ENOMEM;
171 goto out_ino;
172 }
173 memcpy(ui->data, ino->data, ui->data_len);
174 ((char *)ui->data)[ui->data_len] = '\0';
175 } else if (ui->data_len != 0) {
176 err = 10;
177 goto out_invalid;
178 }
179 break;
180 case S_IFDIR:
181 inode->i_op = &ubifs_dir_inode_operations;
182 inode->i_fop = &ubifs_dir_operations;
183 if (ui->data_len != 0) {
184 err = 11;
185 goto out_invalid;
186 }
187 break;
188 case S_IFLNK:
189 inode->i_op = &ubifs_symlink_inode_operations;
190 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
191 err = 12;
192 goto out_invalid;
193 }
194 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
195 if (!ui->data) {
196 err = -ENOMEM;
197 goto out_ino;
198 }
199 memcpy(ui->data, ino->data, ui->data_len);
200 ((char *)ui->data)[ui->data_len] = '\0';
201 break;
202 case S_IFBLK:
203 case S_IFCHR:
204 {
205 dev_t rdev;
206 union ubifs_dev_desc *dev;
207
208 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
209 if (!ui->data) {
210 err = -ENOMEM;
211 goto out_ino;
212 }
213
214 dev = (union ubifs_dev_desc *)ino->data;
215 if (ui->data_len == sizeof(dev->new))
216 rdev = new_decode_dev(le32_to_cpu(dev->new));
217 else if (ui->data_len == sizeof(dev->huge))
218 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
219 else {
220 err = 13;
221 goto out_invalid;
222 }
223 memcpy(ui->data, ino->data, ui->data_len);
224 inode->i_op = &ubifs_file_inode_operations;
225 init_special_inode(inode, inode->i_mode, rdev);
226 break;
227 }
228 case S_IFSOCK:
229 case S_IFIFO:
230 inode->i_op = &ubifs_file_inode_operations;
231 init_special_inode(inode, inode->i_mode, 0);
232 if (ui->data_len != 0) {
233 err = 14;
234 goto out_invalid;
235 }
236 break;
237 default:
238 err = 15;
239 goto out_invalid;
240 }
241
242 kfree(ino);
243 ubifs_set_inode_flags(inode);
244 unlock_new_inode(inode);
245 return inode;
246
247out_invalid:
248 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
249 dbg_dump_node(c, ino);
250 dbg_dump_inode(c, inode);
251 err = -EINVAL;
252out_ino:
253 kfree(ino);
254out:
255 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
256 iget_failed(inode);
257 return ERR_PTR(err);
258}
259
260static struct inode *ubifs_alloc_inode(struct super_block *sb)
261{
262 struct ubifs_inode *ui;
263
264 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
265 if (!ui)
266 return NULL;
267
268 memset((void *)ui + sizeof(struct inode), 0,
269 sizeof(struct ubifs_inode) - sizeof(struct inode));
270 mutex_init(&ui->ui_mutex);
271 spin_lock_init(&ui->ui_lock);
272 return &ui->vfs_inode;
273};
274
275static void ubifs_destroy_inode(struct inode *inode)
276{
277 struct ubifs_inode *ui = ubifs_inode(inode);
278
279 kfree(ui->data);
280 kmem_cache_free(ubifs_inode_slab, inode);
281}
282
283/*
284 * Note, Linux write-back code calls this without 'i_mutex'.
285 */
286static int ubifs_write_inode(struct inode *inode, int wait)
287{
fbfa6c88 288 int err = 0;
1e51764a
AB
289 struct ubifs_info *c = inode->i_sb->s_fs_info;
290 struct ubifs_inode *ui = ubifs_inode(inode);
291
292 ubifs_assert(!ui->xattr);
293 if (is_bad_inode(inode))
294 return 0;
295
296 mutex_lock(&ui->ui_mutex);
297 /*
298 * Due to races between write-back forced by budgeting
299 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
300 * have already been synchronized, do not do this again. This might
301 * also happen if it was synchronized in an VFS operation, e.g.
302 * 'ubifs_link()'.
303 */
304 if (!ui->dirty) {
305 mutex_unlock(&ui->ui_mutex);
306 return 0;
307 }
308
fbfa6c88
AB
309 /*
310 * As an optimization, do not write orphan inodes to the media just
311 * because this is not needed.
312 */
313 dbg_gen("inode %lu, mode %#x, nlink %u",
314 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
315 if (inode->i_nlink) {
1f28681a 316 err = ubifs_jnl_write_inode(c, inode);
fbfa6c88
AB
317 if (err)
318 ubifs_err("can't write inode %lu, error %d",
319 inode->i_ino, err);
320 }
1e51764a
AB
321
322 ui->dirty = 0;
323 mutex_unlock(&ui->ui_mutex);
324 ubifs_release_dirty_inode_budget(c, ui);
325 return err;
326}
327
328static void ubifs_delete_inode(struct inode *inode)
329{
330 int err;
331 struct ubifs_info *c = inode->i_sb->s_fs_info;
1e0f358e 332 struct ubifs_inode *ui = ubifs_inode(inode);
1e51764a 333
1e0f358e 334 if (ui->xattr)
1e51764a
AB
335 /*
336 * Extended attribute inode deletions are fully handled in
337 * 'ubifs_removexattr()'. These inodes are special and have
338 * limited usage, so there is nothing to do here.
339 */
340 goto out;
341
7d32c2bb 342 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
1e51764a
AB
343 ubifs_assert(!atomic_read(&inode->i_count));
344 ubifs_assert(inode->i_nlink == 0);
345
346 truncate_inode_pages(&inode->i_data, 0);
347 if (is_bad_inode(inode))
348 goto out;
349
1e0f358e 350 ui->ui_size = inode->i_size = 0;
de94eb55 351 err = ubifs_jnl_delete_inode(c, inode);
1e51764a
AB
352 if (err)
353 /*
354 * Worst case we have a lost orphan inode wasting space, so a
0a883a05 355 * simple error message is OK here.
1e51764a 356 */
de94eb55
AB
357 ubifs_err("can't delete inode %lu, error %d",
358 inode->i_ino, err);
359
1e51764a 360out:
1e0f358e
AB
361 if (ui->dirty)
362 ubifs_release_dirty_inode_budget(c, ui);
6d6cb0d6
AH
363 else {
364 /* We've deleted something - clean the "no space" flags */
365 c->nospace = c->nospace_rp = 0;
366 smp_wmb();
367 }
1e51764a
AB
368 clear_inode(inode);
369}
370
371static void ubifs_dirty_inode(struct inode *inode)
372{
373 struct ubifs_inode *ui = ubifs_inode(inode);
374
375 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
376 if (!ui->dirty) {
377 ui->dirty = 1;
378 dbg_gen("inode %lu", inode->i_ino);
379 }
380}
381
382static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
383{
384 struct ubifs_info *c = dentry->d_sb->s_fs_info;
385 unsigned long long free;
7c7cbadf 386 __le32 *uuid = (__le32 *)c->uuid;
1e51764a 387
7dad181b 388 free = ubifs_get_free_space(c);
1e51764a
AB
389 dbg_gen("free space %lld bytes (%lld blocks)",
390 free, free >> UBIFS_BLOCK_SHIFT);
391
392 buf->f_type = UBIFS_SUPER_MAGIC;
393 buf->f_bsize = UBIFS_BLOCK_SIZE;
394 buf->f_blocks = c->block_cnt;
395 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
396 if (free > c->report_rp_size)
397 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
398 else
399 buf->f_bavail = 0;
400 buf->f_files = 0;
401 buf->f_ffree = 0;
402 buf->f_namelen = UBIFS_MAX_NLEN;
7c7cbadf
AB
403 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
404 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
b4978e94 405 ubifs_assert(buf->f_bfree <= c->block_cnt);
1e51764a
AB
406 return 0;
407}
408
409static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
410{
411 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
412
413 if (c->mount_opts.unmount_mode == 2)
414 seq_printf(s, ",fast_unmount");
415 else if (c->mount_opts.unmount_mode == 1)
416 seq_printf(s, ",norm_unmount");
417
4793e7c5
AH
418 if (c->mount_opts.bulk_read == 2)
419 seq_printf(s, ",bulk_read");
420 else if (c->mount_opts.bulk_read == 1)
421 seq_printf(s, ",no_bulk_read");
422
2953e73f
AH
423 if (c->mount_opts.chk_data_crc == 2)
424 seq_printf(s, ",chk_data_crc");
425 else if (c->mount_opts.chk_data_crc == 1)
426 seq_printf(s, ",no_chk_data_crc");
427
553dea4d 428 if (c->mount_opts.override_compr) {
fcabb347
HA
429 seq_printf(s, ",compr=%s",
430 ubifs_compr_name(c->mount_opts.compr_type));
553dea4d
AB
431 }
432
1e51764a
AB
433 return 0;
434}
435
436static int ubifs_sync_fs(struct super_block *sb, int wait)
437{
f1038300 438 int i, err;
1e51764a 439 struct ubifs_info *c = sb->s_fs_info;
304d427c 440 struct writeback_control wbc = {
dedb0d48 441 .sync_mode = WB_SYNC_ALL,
304d427c
AB
442 .range_start = 0,
443 .range_end = LLONG_MAX,
444 .nr_to_write = LONG_MAX,
445 };
446
e8ea1759 447 /*
dedb0d48
AB
448 * Zero @wait is just an advisory thing to help the file system shove
449 * lots of data into the queues, and there will be the second
e8ea1759
AB
450 * '->sync_fs()' call, with non-zero @wait.
451 */
dedb0d48
AB
452 if (!wait)
453 return 0;
e8ea1759 454
f1038300
AB
455 if (sb->s_flags & MS_RDONLY)
456 return 0;
457
304d427c
AB
458 /*
459 * VFS calls '->sync_fs()' before synchronizing all dirty inodes and
460 * pages, so synchronize them first, then commit the journal. Strictly
461 * speaking, it is not necessary to commit the journal here,
462 * synchronizing write-buffers would be enough. But committing makes
463 * UBIFS free space predictions much more accurate, so we want to let
464 * the user be able to get more accurate results of 'statfs()' after
465 * they synchronize the file system.
466 */
467 generic_sync_sb_inodes(sb, &wbc);
1e51764a 468
3eb14297
AH
469 /*
470 * Synchronize write buffers, because 'ubifs_run_commit()' does not
471 * do this if it waits for an already running commit.
472 */
473 for (i = 0; i < c->jhead_cnt; i++) {
474 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
475 if (err)
476 return err;
477 }
478
f1038300
AB
479 err = ubifs_run_commit(c);
480 if (err)
481 return err;
403e12ab 482
cb5c6a2b 483 return ubi_sync(c->vi.ubi_num);
1e51764a
AB
484}
485
486/**
487 * init_constants_early - initialize UBIFS constants.
488 * @c: UBIFS file-system description object
489 *
490 * This function initialize UBIFS constants which do not need the superblock to
491 * be read. It also checks that the UBI volume satisfies basic UBIFS
492 * requirements. Returns zero in case of success and a negative error code in
493 * case of failure.
494 */
495static int init_constants_early(struct ubifs_info *c)
496{
497 if (c->vi.corrupted) {
498 ubifs_warn("UBI volume is corrupted - read-only mode");
499 c->ro_media = 1;
500 }
501
502 if (c->di.ro_mode) {
503 ubifs_msg("read-only UBI device");
504 c->ro_media = 1;
505 }
506
507 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
508 ubifs_msg("static UBI volume - read-only mode");
509 c->ro_media = 1;
510 }
511
512 c->leb_cnt = c->vi.size;
513 c->leb_size = c->vi.usable_leb_size;
514 c->half_leb_size = c->leb_size / 2;
515 c->min_io_size = c->di.min_io_size;
516 c->min_io_shift = fls(c->min_io_size) - 1;
517
518 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
519 ubifs_err("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("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("bad min. I/O size %d", c->min_io_size);
532 return -EINVAL;
533 }
534
535 /*
536 * UBIFS aligns all node to 8-byte boundary, so to make function in
537 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
538 * less than 8.
539 */
540 if (c->min_io_size < 8) {
541 c->min_io_size = 8;
542 c->min_io_shift = 3;
543 }
544
545 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
546 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
547
548 /*
549 * Initialize node length ranges which are mostly needed for node
550 * length validation.
551 */
552 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
553 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
554 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
555 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
556 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
557 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
558
559 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
560 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
561 c->ranges[UBIFS_ORPH_NODE].min_len =
562 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
563 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
564 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
565 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
566 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
567 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
568 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
569 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
570 /*
571 * Minimum indexing node size is amended later when superblock is
572 * read and the key length is known.
573 */
574 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
575 /*
576 * Maximum indexing node size is amended later when superblock is
577 * read and the fanout is known.
578 */
579 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
580
581 /*
7078202e
AB
582 * Initialize dead and dark LEB space watermarks. See gc.c for comments
583 * about these values.
1e51764a
AB
584 */
585 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
586 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
587
9bbb5726
AB
588 /*
589 * Calculate how many bytes would be wasted at the end of LEB if it was
590 * fully filled with data nodes of maximum size. This is used in
591 * calculations when reporting free space.
592 */
593 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
39ce81ce 594
4793e7c5 595 /* Buffer size for bulk-reads */
6c0c42cd
AB
596 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
597 if (c->max_bu_buf_len > c->leb_size)
598 c->max_bu_buf_len = c->leb_size;
1e51764a
AB
599 return 0;
600}
601
602/**
603 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
604 * @c: UBIFS file-system description object
605 * @lnum: LEB the write-buffer was synchronized to
606 * @free: how many free bytes left in this LEB
607 * @pad: how many bytes were padded
608 *
609 * This is a callback function which is called by the I/O unit when the
610 * write-buffer is synchronized. We need this to correctly maintain space
611 * accounting in bud logical eraseblocks. This function returns zero in case of
612 * success and a negative error code in case of failure.
613 *
614 * This function actually belongs to the journal, but we keep it here because
615 * we want to keep it static.
616 */
617static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
618{
619 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
620}
621
622/*
79807d07 623 * init_constants_sb - initialize UBIFS constants.
1e51764a
AB
624 * @c: UBIFS file-system description object
625 *
626 * This is a helper function which initializes various UBIFS constants after
627 * the superblock has been read. It also checks various UBIFS parameters and
628 * makes sure they are all right. Returns zero in case of success and a
629 * negative error code in case of failure.
630 */
79807d07 631static int init_constants_sb(struct ubifs_info *c)
1e51764a
AB
632{
633 int tmp, err;
4d61db4f 634 long long tmp64;
1e51764a
AB
635
636 c->main_bytes = (long long)c->main_lebs * c->leb_size;
637 c->max_znode_sz = sizeof(struct ubifs_znode) +
638 c->fanout * sizeof(struct ubifs_zbranch);
639
640 tmp = ubifs_idx_node_sz(c, 1);
641 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
642 c->min_idx_node_sz = ALIGN(tmp, 8);
643
644 tmp = ubifs_idx_node_sz(c, c->fanout);
645 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
646 c->max_idx_node_sz = ALIGN(tmp, 8);
647
648 /* Make sure LEB size is large enough to fit full commit */
649 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
650 tmp = ALIGN(tmp, c->min_io_size);
651 if (tmp > c->leb_size) {
652 dbg_err("too small LEB size %d, at least %d needed",
653 c->leb_size, tmp);
654 return -EINVAL;
655 }
656
657 /*
658 * Make sure that the log is large enough to fit reference nodes for
659 * all buds plus one reserved LEB.
660 */
4d61db4f
AB
661 tmp64 = c->max_bud_bytes + c->leb_size - 1;
662 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
1e51764a
AB
663 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
664 tmp /= c->leb_size;
665 tmp += 1;
666 if (c->log_lebs < tmp) {
667 dbg_err("too small log %d LEBs, required min. %d LEBs",
668 c->log_lebs, tmp);
669 return -EINVAL;
670 }
671
672 /*
673 * When budgeting we assume worst-case scenarios when the pages are not
674 * be compressed and direntries are of the maximum size.
675 *
676 * Note, data, which may be stored in inodes is budgeted separately, so
677 * it is not included into 'c->inode_budget'.
678 */
679 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
680 c->inode_budget = UBIFS_INO_NODE_SZ;
681 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
682
683 /*
684 * When the amount of flash space used by buds becomes
685 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
686 * The writers are unblocked when the commit is finished. To avoid
687 * writers to be blocked UBIFS initiates background commit in advance,
688 * when number of bud bytes becomes above the limit defined below.
689 */
690 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
691
692 /*
693 * Ensure minimum journal size. All the bytes in the journal heads are
694 * considered to be used, when calculating the current journal usage.
695 * Consequently, if the journal is too small, UBIFS will treat it as
696 * always full.
697 */
4d61db4f 698 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
1e51764a
AB
699 if (c->bg_bud_bytes < tmp64)
700 c->bg_bud_bytes = tmp64;
701 if (c->max_bud_bytes < tmp64 + c->leb_size)
702 c->max_bud_bytes = tmp64 + c->leb_size;
703
704 err = ubifs_calc_lpt_geom(c);
705 if (err)
706 return err;
707
fb1cd01a
AB
708 /* Initialize effective LEB size used in budgeting calculations */
709 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
79807d07
AB
710 return 0;
711}
712
713/*
714 * init_constants_master - initialize UBIFS constants.
715 * @c: UBIFS file-system description object
716 *
717 * This is a helper function which initializes various UBIFS constants after
718 * the master node has been read. It also checks various UBIFS parameters and
719 * makes sure they are all right.
720 */
721static void init_constants_master(struct ubifs_info *c)
722{
723 long long tmp64;
724
1e51764a 725 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
fb1cd01a 726 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
1e51764a
AB
727
728 /*
729 * Calculate total amount of FS blocks. This number is not used
730 * internally because it does not make much sense for UBIFS, but it is
731 * necessary to report something for the 'statfs()' call.
732 *
7dad181b 733 * Subtract the LEB reserved for GC, the LEB which is reserved for
af14a1ad
AB
734 * deletions, minimum LEBs for the index, and assume only one journal
735 * head is available.
1e51764a 736 */
af14a1ad 737 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
4d61db4f 738 tmp64 *= (long long)c->leb_size - c->leb_overhead;
1e51764a
AB
739 tmp64 = ubifs_reported_space(c, tmp64);
740 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
1e51764a
AB
741}
742
743/**
744 * take_gc_lnum - reserve GC LEB.
745 * @c: UBIFS file-system description object
746 *
b4978e94
AB
747 * This function ensures that the LEB reserved for garbage collection is marked
748 * as "taken" in lprops. We also have to set free space to LEB size and dirty
749 * space to zero, because lprops may contain out-of-date information if the
750 * file-system was un-mounted before it has been committed. This function
751 * returns zero in case of success and a negative error code in case of
752 * failure.
1e51764a
AB
753 */
754static int take_gc_lnum(struct ubifs_info *c)
755{
756 int err;
757
758 if (c->gc_lnum == -1) {
759 ubifs_err("no LEB for GC");
760 return -EINVAL;
761 }
762
1e51764a
AB
763 /* And we have to tell lprops that this LEB is taken */
764 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
765 LPROPS_TAKEN, 0, 0);
766 return err;
767}
768
769/**
770 * alloc_wbufs - allocate write-buffers.
771 * @c: UBIFS file-system description object
772 *
773 * This helper function allocates and initializes UBIFS write-buffers. Returns
774 * zero in case of success and %-ENOMEM in case of failure.
775 */
776static int alloc_wbufs(struct ubifs_info *c)
777{
778 int i, err;
779
780 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
781 GFP_KERNEL);
782 if (!c->jheads)
783 return -ENOMEM;
784
785 /* Initialize journal heads */
786 for (i = 0; i < c->jhead_cnt; i++) {
787 INIT_LIST_HEAD(&c->jheads[i].buds_list);
788 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
789 if (err)
790 return err;
791
792 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
793 c->jheads[i].wbuf.jhead = i;
794 }
795
796 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
797 /*
798 * Garbage Collector head likely contains long-term data and
799 * does not need to be synchronized by timer.
800 */
801 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
802 c->jheads[GCHD].wbuf.timeout = 0;
803
804 return 0;
805}
806
807/**
808 * free_wbufs - free write-buffers.
809 * @c: UBIFS file-system description object
810 */
811static void free_wbufs(struct ubifs_info *c)
812{
813 int i;
814
815 if (c->jheads) {
816 for (i = 0; i < c->jhead_cnt; i++) {
817 kfree(c->jheads[i].wbuf.buf);
818 kfree(c->jheads[i].wbuf.inodes);
819 }
820 kfree(c->jheads);
821 c->jheads = NULL;
822 }
823}
824
825/**
826 * free_orphans - free orphans.
827 * @c: UBIFS file-system description object
828 */
829static void free_orphans(struct ubifs_info *c)
830{
831 struct ubifs_orphan *orph;
832
833 while (c->orph_dnext) {
834 orph = c->orph_dnext;
835 c->orph_dnext = orph->dnext;
836 list_del(&orph->list);
837 kfree(orph);
838 }
839
840 while (!list_empty(&c->orph_list)) {
841 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
842 list_del(&orph->list);
843 kfree(orph);
844 dbg_err("orphan list not empty at unmount");
845 }
846
847 vfree(c->orph_buf);
848 c->orph_buf = NULL;
849}
850
851/**
852 * free_buds - free per-bud objects.
853 * @c: UBIFS file-system description object
854 */
855static void free_buds(struct ubifs_info *c)
856{
857 struct rb_node *this = c->buds.rb_node;
858 struct ubifs_bud *bud;
859
860 while (this) {
861 if (this->rb_left)
862 this = this->rb_left;
863 else if (this->rb_right)
864 this = this->rb_right;
865 else {
866 bud = rb_entry(this, struct ubifs_bud, rb);
867 this = rb_parent(this);
868 if (this) {
869 if (this->rb_left == &bud->rb)
870 this->rb_left = NULL;
871 else
872 this->rb_right = NULL;
873 }
874 kfree(bud);
875 }
876 }
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 = ubi_is_mapped(c->ubi, 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
4793e7c5
AH
913 * Opt_bulk_read: enable bulk-reads
914 * Opt_no_bulk_read: disable bulk-reads
2953e73f
AH
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
553dea4d 917 * Opt_override_compr: override default compressor
1e51764a
AB
918 * Opt_err: just end of array marker
919 */
920enum {
921 Opt_fast_unmount,
922 Opt_norm_unmount,
4793e7c5
AH
923 Opt_bulk_read,
924 Opt_no_bulk_read,
2953e73f
AH
925 Opt_chk_data_crc,
926 Opt_no_chk_data_crc,
553dea4d 927 Opt_override_compr,
1e51764a
AB
928 Opt_err,
929};
930
a447c093 931static const match_table_t tokens = {
1e51764a
AB
932 {Opt_fast_unmount, "fast_unmount"},
933 {Opt_norm_unmount, "norm_unmount"},
4793e7c5
AH
934 {Opt_bulk_read, "bulk_read"},
935 {Opt_no_bulk_read, "no_bulk_read"},
2953e73f
AH
936 {Opt_chk_data_crc, "chk_data_crc"},
937 {Opt_no_chk_data_crc, "no_chk_data_crc"},
553dea4d 938 {Opt_override_compr, "compr=%s"},
1e51764a
AB
939 {Opt_err, NULL},
940};
941
942/**
943 * ubifs_parse_options - parse mount parameters.
944 * @c: UBIFS file-system description object
945 * @options: parameters to parse
946 * @is_remount: non-zero if this is FS re-mount
947 *
948 * This function parses UBIFS mount options and returns zero in case success
949 * and a negative error code in case of failure.
950 */
951static int ubifs_parse_options(struct ubifs_info *c, char *options,
952 int is_remount)
953{
954 char *p;
955 substring_t args[MAX_OPT_ARGS];
956
957 if (!options)
958 return 0;
959
960 while ((p = strsep(&options, ","))) {
961 int token;
962
963 if (!*p)
964 continue;
965
966 token = match_token(p, tokens, args);
967 switch (token) {
27ad2799
AB
968 /*
969 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
970 * We accepte them in order to be backware-compatible. But this
971 * should be removed at some point.
972 */
1e51764a
AB
973 case Opt_fast_unmount:
974 c->mount_opts.unmount_mode = 2;
1e51764a
AB
975 break;
976 case Opt_norm_unmount:
977 c->mount_opts.unmount_mode = 1;
1e51764a 978 break;
4793e7c5
AH
979 case Opt_bulk_read:
980 c->mount_opts.bulk_read = 2;
981 c->bulk_read = 1;
982 break;
983 case Opt_no_bulk_read:
984 c->mount_opts.bulk_read = 1;
985 c->bulk_read = 0;
986 break;
2953e73f
AH
987 case Opt_chk_data_crc:
988 c->mount_opts.chk_data_crc = 2;
989 c->no_chk_data_crc = 0;
990 break;
991 case Opt_no_chk_data_crc:
992 c->mount_opts.chk_data_crc = 1;
993 c->no_chk_data_crc = 1;
994 break;
553dea4d
AB
995 case Opt_override_compr:
996 {
997 char *name = match_strdup(&args[0]);
998
999 if (!name)
1000 return -ENOMEM;
1001 if (!strcmp(name, "none"))
1002 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1003 else if (!strcmp(name, "lzo"))
1004 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1005 else if (!strcmp(name, "zlib"))
1006 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1007 else {
1008 ubifs_err("unknown compressor \"%s\"", name);
1009 kfree(name);
1010 return -EINVAL;
1011 }
1012 kfree(name);
1013 c->mount_opts.override_compr = 1;
1014 c->default_compr = c->mount_opts.compr_type;
1015 break;
1016 }
1e51764a
AB
1017 default:
1018 ubifs_err("unrecognized mount option \"%s\" "
1019 "or missing value", p);
1020 return -EINVAL;
1021 }
1022 }
1023
1024 return 0;
1025}
1026
1027/**
1028 * destroy_journal - destroy journal data structures.
1029 * @c: UBIFS file-system description object
1030 *
1031 * This function destroys journal data structures including those that may have
1032 * been created by recovery functions.
1033 */
1034static void destroy_journal(struct ubifs_info *c)
1035{
1036 while (!list_empty(&c->unclean_leb_list)) {
1037 struct ubifs_unclean_leb *ucleb;
1038
1039 ucleb = list_entry(c->unclean_leb_list.next,
1040 struct ubifs_unclean_leb, list);
1041 list_del(&ucleb->list);
1042 kfree(ucleb);
1043 }
1044 while (!list_empty(&c->old_buds)) {
1045 struct ubifs_bud *bud;
1046
1047 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1048 list_del(&bud->list);
1049 kfree(bud);
1050 }
1051 ubifs_destroy_idx_gc(c);
1052 ubifs_destroy_size_tree(c);
1053 ubifs_tnc_close(c);
1054 free_buds(c);
1055}
1056
3477d204
AB
1057/**
1058 * bu_init - initialize bulk-read information.
1059 * @c: UBIFS file-system description object
1060 */
1061static void bu_init(struct ubifs_info *c)
1062{
1063 ubifs_assert(c->bulk_read == 1);
1064
1065 if (c->bu.buf)
1066 return; /* Already initialized */
1067
1068again:
1069 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1070 if (!c->bu.buf) {
1071 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1072 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1073 goto again;
1074 }
1075
1076 /* Just disable bulk-read */
1077 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1078 "disabling it", c->max_bu_buf_len);
1079 c->mount_opts.bulk_read = 1;
1080 c->bulk_read = 0;
1081 return;
1082 }
1083}
1084
57a450e9
AB
1085/**
1086 * check_free_space - check if there is enough free space to mount.
1087 * @c: UBIFS file-system description object
1088 *
1089 * This function makes sure UBIFS has enough free space to be mounted in
1090 * read/write mode. UBIFS must always have some free space to allow deletions.
1091 */
1092static int check_free_space(struct ubifs_info *c)
1093{
1094 ubifs_assert(c->dark_wm > 0);
1095 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1096 ubifs_err("insufficient free space to mount in read/write mode");
1097 dbg_dump_budg(c);
1098 dbg_dump_lprops(c);
a2b9df3f 1099 return -ENOSPC;
57a450e9
AB
1100 }
1101 return 0;
1102}
1103
1e51764a
AB
1104/**
1105 * mount_ubifs - mount UBIFS file-system.
1106 * @c: UBIFS file-system description object
1107 *
1108 * This function mounts UBIFS file system. Returns zero in case of success and
1109 * a negative error code in case of failure.
1110 *
1111 * Note, the function does not de-allocate resources it it fails half way
1112 * through, and the caller has to do this instead.
1113 */
1114static int mount_ubifs(struct ubifs_info *c)
1115{
1116 struct super_block *sb = c->vfs_sb;
1117 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1118 long long x;
1119 size_t sz;
1120
1121 err = init_constants_early(c);
1122 if (err)
1123 return err;
1124
17c2f9f8
AB
1125 err = ubifs_debugging_init(c);
1126 if (err)
1127 return err;
1e51764a
AB
1128
1129 err = check_volume_empty(c);
1130 if (err)
1131 goto out_free;
1132
1133 if (c->empty && (mounted_read_only || c->ro_media)) {
1134 /*
1135 * This UBI volume is empty, and read-only, or the file system
1136 * is mounted read-only - we cannot format it.
1137 */
1138 ubifs_err("can't format empty UBI volume: read-only %s",
1139 c->ro_media ? "UBI volume" : "mount");
1140 err = -EROFS;
1141 goto out_free;
1142 }
1143
1144 if (c->ro_media && !mounted_read_only) {
1145 ubifs_err("cannot mount read-write - read-only media");
1146 err = -EROFS;
1147 goto out_free;
1148 }
1149
1150 /*
1151 * The requirement for the buffer is that it should fit indexing B-tree
1152 * height amount of integers. We assume the height if the TNC tree will
1153 * never exceed 64.
1154 */
1155 err = -ENOMEM;
1156 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1157 if (!c->bottom_up_buf)
1158 goto out_free;
1159
1160 c->sbuf = vmalloc(c->leb_size);
1161 if (!c->sbuf)
1162 goto out_free;
1163
1164 if (!mounted_read_only) {
1165 c->ileb_buf = vmalloc(c->leb_size);
1166 if (!c->ileb_buf)
1167 goto out_free;
1168 }
1169
3477d204
AB
1170 if (c->bulk_read == 1)
1171 bu_init(c);
1172
1173 /*
1174 * We have to check all CRCs, even for data nodes, when we mount the FS
1175 * (specifically, when we are replaying).
1176 */
2953e73f
AH
1177 c->always_chk_crc = 1;
1178
1e51764a
AB
1179 err = ubifs_read_superblock(c);
1180 if (err)
1181 goto out_free;
1182
1183 /*
553dea4d 1184 * Make sure the compressor which is set as default in the superblock
57a450e9 1185 * or overridden by mount options is actually compiled in.
1e51764a
AB
1186 */
1187 if (!ubifs_compr_present(c->default_compr)) {
553dea4d
AB
1188 ubifs_err("'compressor \"%s\" is not compiled in",
1189 ubifs_compr_name(c->default_compr));
8eec2f36 1190 err = -ENOTSUPP;
553dea4d 1191 goto out_free;
1e51764a
AB
1192 }
1193
79807d07 1194 err = init_constants_sb(c);
1e51764a 1195 if (err)
17c2f9f8 1196 goto out_free;
1e51764a
AB
1197
1198 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1199 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1200 c->cbuf = kmalloc(sz, GFP_NOFS);
1201 if (!c->cbuf) {
1202 err = -ENOMEM;
17c2f9f8 1203 goto out_free;
1e51764a
AB
1204 }
1205
0855f310 1206 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1e51764a
AB
1207 if (!mounted_read_only) {
1208 err = alloc_wbufs(c);
1209 if (err)
1210 goto out_cbuf;
1211
1212 /* Create background thread */
fcabb347 1213 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1e51764a
AB
1214 if (IS_ERR(c->bgt)) {
1215 err = PTR_ERR(c->bgt);
1216 c->bgt = NULL;
1217 ubifs_err("cannot spawn \"%s\", error %d",
1218 c->bgt_name, err);
1219 goto out_wbufs;
1220 }
1221 wake_up_process(c->bgt);
1222 }
1223
1224 err = ubifs_read_master(c);
1225 if (err)
1226 goto out_master;
1227
79807d07
AB
1228 init_constants_master(c);
1229
1e51764a
AB
1230 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1231 ubifs_msg("recovery needed");
1232 c->need_recovery = 1;
1233 if (!mounted_read_only) {
1234 err = ubifs_recover_inl_heads(c, c->sbuf);
1235 if (err)
1236 goto out_master;
1237 }
1238 } else if (!mounted_read_only) {
1239 /*
1240 * Set the "dirty" flag so that if we reboot uncleanly we
1241 * will notice this immediately on the next mount.
1242 */
1243 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1244 err = ubifs_write_master(c);
1245 if (err)
1246 goto out_master;
1247 }
1248
1249 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1250 if (err)
1251 goto out_lpt;
1252
1253 err = dbg_check_idx_size(c, c->old_idx_sz);
1254 if (err)
1255 goto out_lpt;
1256
1257 err = ubifs_replay_journal(c);
1258 if (err)
1259 goto out_journal;
1260
1261 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1262 if (err)
1263 goto out_orphans;
1264
1265 if (!mounted_read_only) {
1266 int lnum;
1267
57a450e9
AB
1268 err = check_free_space(c);
1269 if (err)
1e51764a 1270 goto out_orphans;
1e51764a
AB
1271
1272 /* Check for enough log space */
1273 lnum = c->lhead_lnum + 1;
1274 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1275 lnum = UBIFS_LOG_LNUM;
1276 if (lnum == c->ltail_lnum) {
1277 err = ubifs_consolidate_log(c);
1278 if (err)
1279 goto out_orphans;
1280 }
1281
1282 if (c->need_recovery) {
1283 err = ubifs_recover_size(c);
1284 if (err)
1285 goto out_orphans;
1286 err = ubifs_rcvry_gc_commit(c);
b4978e94 1287 } else {
1e51764a 1288 err = take_gc_lnum(c);
b4978e94
AB
1289 if (err)
1290 goto out_orphans;
1291
1292 /*
1293 * GC LEB may contain garbage if there was an unclean
1294 * reboot, and it should be un-mapped.
1295 */
1296 err = ubifs_leb_unmap(c, c->gc_lnum);
1297 if (err)
1298 return err;
1299 }
1e51764a
AB
1300
1301 err = dbg_check_lprops(c);
1302 if (err)
1303 goto out_orphans;
1304 } else if (c->need_recovery) {
1305 err = ubifs_recover_size(c);
1306 if (err)
1307 goto out_orphans;
b4978e94
AB
1308 } else {
1309 /*
1310 * Even if we mount read-only, we have to set space in GC LEB
1311 * to proper value because this affects UBIFS free space
1312 * reporting. We do not want to have a situation when
1313 * re-mounting from R/O to R/W changes amount of free space.
1314 */
1315 err = take_gc_lnum(c);
1316 if (err)
1317 goto out_orphans;
1e51764a
AB
1318 }
1319
1320 spin_lock(&ubifs_infos_lock);
1321 list_add_tail(&c->infos_list, &ubifs_infos);
1322 spin_unlock(&ubifs_infos_lock);
1323
1324 if (c->need_recovery) {
1325 if (mounted_read_only)
1326 ubifs_msg("recovery deferred");
1327 else {
1328 c->need_recovery = 0;
1329 ubifs_msg("recovery completed");
b221337a
AB
1330 /*
1331 * GC LEB has to be empty and taken at this point. But
1332 * the journal head LEBs may also be accounted as
1333 * "empty taken" if they are empty.
1334 */
1335 ubifs_assert(c->lst.taken_empty_lebs > 0);
1e51764a 1336 }
6ba87c9b 1337 } else
b221337a 1338 ubifs_assert(c->lst.taken_empty_lebs > 0);
1e51764a 1339
6ba87c9b 1340 err = dbg_check_filesystem(c);
552ff317
AB
1341 if (err)
1342 goto out_infos;
1343
6ba87c9b 1344 err = dbg_debugfs_init_fs(c);
1e51764a
AB
1345 if (err)
1346 goto out_infos;
1347
2953e73f
AH
1348 c->always_chk_crc = 0;
1349
ce769caa
AB
1350 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1351 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1e51764a
AB
1352 if (mounted_read_only)
1353 ubifs_msg("mounted read-only");
1354 x = (long long)c->main_lebs * c->leb_size;
948cfb21
AB
1355 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1356 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1e51764a 1357 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
948cfb21
AB
1358 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1359 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
963f0cf6
AB
1360 ubifs_msg("media format: w%d/r%d (latest is w%d/r%d)",
1361 c->fmt_version, c->ro_compat_version,
1362 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
948cfb21 1363 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
fae7fb29 1364 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
948cfb21 1365 c->report_rp_size, c->report_rp_size >> 10);
1e51764a
AB
1366
1367 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1368 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1369 dbg_msg("LEB size: %d bytes (%d KiB)",
948cfb21 1370 c->leb_size, c->leb_size >> 10);
1e51764a
AB
1371 dbg_msg("data journal heads: %d",
1372 c->jhead_cnt - NONDATA_JHEADS_CNT);
1373 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1374 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1375 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1376 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1377 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1378 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1e51764a
AB
1379 dbg_msg("big_lpt %d", c->big_lpt);
1380 dbg_msg("log LEBs: %d (%d - %d)",
1381 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1382 dbg_msg("LPT area LEBs: %d (%d - %d)",
1383 c->lpt_lebs, c->lpt_first, c->lpt_last);
1384 dbg_msg("orphan area LEBs: %d (%d - %d)",
1385 c->orph_lebs, c->orph_first, c->orph_last);
1386 dbg_msg("main area LEBs: %d (%d - %d)",
1387 c->main_lebs, c->main_first, c->leb_cnt - 1);
1388 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1389 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1390 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1391 dbg_msg("key hash type: %d", c->key_hash_type);
1392 dbg_msg("tree fanout: %d", c->fanout);
1393 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1394 dbg_msg("first main LEB: %d", c->main_first);
8e5033ad
AB
1395 dbg_msg("max. znode size %d", c->max_znode_sz);
1396 dbg_msg("max. index node size %d", c->max_idx_node_sz);
1397 dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
1398 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1399 dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
1400 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1401 dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
1402 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1403 dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu",
1404 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1405 UBIFS_MAX_DENT_NODE_SZ);
1e51764a
AB
1406 dbg_msg("dead watermark: %d", c->dead_wm);
1407 dbg_msg("dark watermark: %d", c->dark_wm);
8e5033ad 1408 dbg_msg("LEB overhead: %d", c->leb_overhead);
1e51764a
AB
1409 x = (long long)c->main_lebs * c->dark_wm;
1410 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1411 x, x >> 10, x >> 20);
1412 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1413 c->max_bud_bytes, c->max_bud_bytes >> 10,
1414 c->max_bud_bytes >> 20);
1415 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1416 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1417 c->bg_bud_bytes >> 20);
1418 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1419 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1420 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1421 dbg_msg("commit number: %llu", c->cmt_no);
1422
1423 return 0;
1424
1425out_infos:
1426 spin_lock(&ubifs_infos_lock);
1427 list_del(&c->infos_list);
1428 spin_unlock(&ubifs_infos_lock);
1429out_orphans:
1430 free_orphans(c);
1431out_journal:
1432 destroy_journal(c);
1433out_lpt:
1434 ubifs_lpt_free(c, 0);
1435out_master:
1436 kfree(c->mst_node);
1437 kfree(c->rcvrd_mst_node);
1438 if (c->bgt)
1439 kthread_stop(c->bgt);
1440out_wbufs:
1441 free_wbufs(c);
1442out_cbuf:
1443 kfree(c->cbuf);
1e51764a 1444out_free:
3477d204 1445 kfree(c->bu.buf);
1e51764a
AB
1446 vfree(c->ileb_buf);
1447 vfree(c->sbuf);
1448 kfree(c->bottom_up_buf);
17c2f9f8 1449 ubifs_debugging_exit(c);
1e51764a
AB
1450 return err;
1451}
1452
1453/**
1454 * ubifs_umount - un-mount UBIFS file-system.
1455 * @c: UBIFS file-system description object
1456 *
1457 * Note, this function is called to free allocated resourced when un-mounting,
1458 * as well as free resources when an error occurred while we were half way
1459 * through mounting (error path cleanup function). So it has to make sure the
1460 * resource was actually allocated before freeing it.
1461 */
1462static void ubifs_umount(struct ubifs_info *c)
1463{
1464 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1465 c->vi.vol_id);
1466
552ff317 1467 dbg_debugfs_exit_fs(c);
1e51764a
AB
1468 spin_lock(&ubifs_infos_lock);
1469 list_del(&c->infos_list);
1470 spin_unlock(&ubifs_infos_lock);
1471
1472 if (c->bgt)
1473 kthread_stop(c->bgt);
1474
1475 destroy_journal(c);
1476 free_wbufs(c);
1477 free_orphans(c);
1478 ubifs_lpt_free(c, 0);
1479
1480 kfree(c->cbuf);
1481 kfree(c->rcvrd_mst_node);
1482 kfree(c->mst_node);
3477d204
AB
1483 kfree(c->bu.buf);
1484 vfree(c->ileb_buf);
1e51764a
AB
1485 vfree(c->sbuf);
1486 kfree(c->bottom_up_buf);
17c2f9f8 1487 ubifs_debugging_exit(c);
1e51764a
AB
1488}
1489
1490/**
1491 * ubifs_remount_rw - re-mount in read-write mode.
1492 * @c: UBIFS file-system description object
1493 *
1494 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1495 * mode. This function allocates the needed resources and re-mounts UBIFS in
1496 * read-write mode.
1497 */
1498static int ubifs_remount_rw(struct ubifs_info *c)
1499{
1500 int err, lnum;
1501
963f0cf6
AB
1502 if (c->rw_incompat) {
1503 ubifs_err("the file-system is not R/W-compatible");
1504 ubifs_msg("on-flash format version is w%d/r%d, but software "
1505 "only supports up to version w%d/r%d", c->fmt_version,
1506 c->ro_compat_version, UBIFS_FORMAT_VERSION,
1507 UBIFS_RO_COMPAT_VERSION);
1508 return -EROFS;
1509 }
1510
1e51764a 1511 mutex_lock(&c->umount_mutex);
84abf972 1512 dbg_save_space_info(c);
1e51764a 1513 c->remounting_rw = 1;
2953e73f 1514 c->always_chk_crc = 1;
1e51764a 1515
57a450e9
AB
1516 err = check_free_space(c);
1517 if (err)
1e51764a 1518 goto out;
1e51764a
AB
1519
1520 if (c->old_leb_cnt != c->leb_cnt) {
1521 struct ubifs_sb_node *sup;
1522
1523 sup = ubifs_read_sb_node(c);
1524 if (IS_ERR(sup)) {
1525 err = PTR_ERR(sup);
1526 goto out;
1527 }
1528 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1529 err = ubifs_write_sb_node(c, sup);
1530 if (err)
1531 goto out;
1532 }
1533
1534 if (c->need_recovery) {
1535 ubifs_msg("completing deferred recovery");
1536 err = ubifs_write_rcvrd_mst_node(c);
1537 if (err)
1538 goto out;
1539 err = ubifs_recover_size(c);
1540 if (err)
1541 goto out;
1542 err = ubifs_clean_lebs(c, c->sbuf);
1543 if (err)
1544 goto out;
1545 err = ubifs_recover_inl_heads(c, c->sbuf);
1546 if (err)
1547 goto out;
49d128aa
AH
1548 } else {
1549 /* A readonly mount is not allowed to have orphans */
1550 ubifs_assert(c->tot_orphans == 0);
1551 err = ubifs_clear_orphans(c);
1552 if (err)
1553 goto out;
1e51764a
AB
1554 }
1555
1556 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1557 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1558 err = ubifs_write_master(c);
1559 if (err)
1560 goto out;
1561 }
1562
1563 c->ileb_buf = vmalloc(c->leb_size);
1564 if (!c->ileb_buf) {
1565 err = -ENOMEM;
1566 goto out;
1567 }
1568
1569 err = ubifs_lpt_init(c, 0, 1);
1570 if (err)
1571 goto out;
1572
1573 err = alloc_wbufs(c);
1574 if (err)
1575 goto out;
1576
1577 ubifs_create_buds_lists(c);
1578
1579 /* Create background thread */
fcabb347 1580 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1e51764a
AB
1581 if (IS_ERR(c->bgt)) {
1582 err = PTR_ERR(c->bgt);
1583 c->bgt = NULL;
1584 ubifs_err("cannot spawn \"%s\", error %d",
1585 c->bgt_name, err);
2953e73f 1586 goto out;
1e51764a
AB
1587 }
1588 wake_up_process(c->bgt);
1589
1590 c->orph_buf = vmalloc(c->leb_size);
2953e73f
AH
1591 if (!c->orph_buf) {
1592 err = -ENOMEM;
1593 goto out;
1594 }
1e51764a
AB
1595
1596 /* Check for enough log space */
1597 lnum = c->lhead_lnum + 1;
1598 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1599 lnum = UBIFS_LOG_LNUM;
1600 if (lnum == c->ltail_lnum) {
1601 err = ubifs_consolidate_log(c);
1602 if (err)
1603 goto out;
1604 }
1605
1606 if (c->need_recovery)
1607 err = ubifs_rcvry_gc_commit(c);
1608 else
b4978e94 1609 err = ubifs_leb_unmap(c, c->gc_lnum);
1e51764a
AB
1610 if (err)
1611 goto out;
1612
1613 if (c->need_recovery) {
1614 c->need_recovery = 0;
1615 ubifs_msg("deferred recovery completed");
1616 }
1617
1618 dbg_gen("re-mounted read-write");
1619 c->vfs_sb->s_flags &= ~MS_RDONLY;
1620 c->remounting_rw = 0;
2953e73f 1621 c->always_chk_crc = 0;
84abf972 1622 err = dbg_check_space_info(c);
1e51764a 1623 mutex_unlock(&c->umount_mutex);
84abf972 1624 return err;
1e51764a
AB
1625
1626out:
1627 vfree(c->orph_buf);
1628 c->orph_buf = NULL;
1629 if (c->bgt) {
1630 kthread_stop(c->bgt);
1631 c->bgt = NULL;
1632 }
1633 free_wbufs(c);
1634 vfree(c->ileb_buf);
1635 c->ileb_buf = NULL;
1636 ubifs_lpt_free(c, 1);
1637 c->remounting_rw = 0;
2953e73f 1638 c->always_chk_crc = 0;
1e51764a
AB
1639 mutex_unlock(&c->umount_mutex);
1640 return err;
1641}
1642
1e51764a
AB
1643/**
1644 * ubifs_remount_ro - re-mount in read-only mode.
1645 * @c: UBIFS file-system description object
1646 *
84abf972
AB
1647 * We assume VFS has stopped writing. Possibly the background thread could be
1648 * running a commit, however kthread_stop will wait in that case.
1e51764a
AB
1649 */
1650static void ubifs_remount_ro(struct ubifs_info *c)
1651{
1652 int i, err;
1653
1654 ubifs_assert(!c->need_recovery);
6ba87c9b 1655 ubifs_assert(!(c->vfs_sb->s_flags & MS_RDONLY));
e4d9b6cb 1656
1e51764a
AB
1657 mutex_lock(&c->umount_mutex);
1658 if (c->bgt) {
1659 kthread_stop(c->bgt);
1660 c->bgt = NULL;
1661 }
1662
84abf972
AB
1663 dbg_save_space_info(c);
1664
1e51764a
AB
1665 for (i = 0; i < c->jhead_cnt; i++) {
1666 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1667 del_timer_sync(&c->jheads[i].wbuf.timer);
1668 }
1669
e4d9b6cb
AB
1670 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1671 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1672 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1673 err = ubifs_write_master(c);
1674 if (err)
1675 ubifs_ro_mode(c, err);
1676
1e51764a
AB
1677 free_wbufs(c);
1678 vfree(c->orph_buf);
1679 c->orph_buf = NULL;
1680 vfree(c->ileb_buf);
1681 c->ileb_buf = NULL;
1682 ubifs_lpt_free(c, 1);
84abf972
AB
1683 err = dbg_check_space_info(c);
1684 if (err)
1685 ubifs_ro_mode(c, err);
1e51764a
AB
1686 mutex_unlock(&c->umount_mutex);
1687}
1688
1689static void ubifs_put_super(struct super_block *sb)
1690{
1691 int i;
1692 struct ubifs_info *c = sb->s_fs_info;
1693
1694 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1695 c->vi.vol_id);
1696 /*
1697 * The following asserts are only valid if there has not been a failure
1698 * of the media. For example, there will be dirty inodes if we failed
1699 * to write them back because of I/O errors.
1700 */
1701 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1702 ubifs_assert(c->budg_idx_growth == 0);
7d32c2bb 1703 ubifs_assert(c->budg_dd_growth == 0);
1e51764a
AB
1704 ubifs_assert(c->budg_data_growth == 0);
1705
1706 /*
1707 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1708 * and file system un-mount. Namely, it prevents the shrinker from
1709 * picking this superblock for shrinking - it will be just skipped if
1710 * the mutex is locked.
1711 */
1712 mutex_lock(&c->umount_mutex);
1713 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1714 /*
1715 * First of all kill the background thread to make sure it does
1716 * not interfere with un-mounting and freeing resources.
1717 */
1718 if (c->bgt) {
1719 kthread_stop(c->bgt);
1720 c->bgt = NULL;
1721 }
1722
1723 /* Synchronize write-buffers */
1724 if (c->jheads)
1725 for (i = 0; i < c->jhead_cnt; i++) {
1726 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1727 del_timer_sync(&c->jheads[i].wbuf.timer);
1728 }
1729
1730 /*
1731 * On fatal errors c->ro_media is set to 1, in which case we do
1732 * not write the master node.
1733 */
1734 if (!c->ro_media) {
1735 /*
1736 * We are being cleanly unmounted which means the
1737 * orphans were killed - indicate this in the master
1738 * node. Also save the reserved GC LEB number.
1739 */
1740 int err;
1741
1742 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1743 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1744 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1745 err = ubifs_write_master(c);
1746 if (err)
1747 /*
1748 * Recovery will attempt to fix the master area
1749 * next mount, so we just print a message and
1750 * continue to unmount normally.
1751 */
1752 ubifs_err("failed to write master node, "
1753 "error %d", err);
1754 }
1755 }
1756
1757 ubifs_umount(c);
1758 bdi_destroy(&c->bdi);
1759 ubi_close_volume(c->ubi);
1760 mutex_unlock(&c->umount_mutex);
1761 kfree(c);
1762}
1763
1764static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1765{
1766 int err;
1767 struct ubifs_info *c = sb->s_fs_info;
1768
1769 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1770
1771 err = ubifs_parse_options(c, data, 1);
1772 if (err) {
1773 ubifs_err("invalid or unknown remount parameter");
1774 return err;
1775 }
3477d204 1776
1e51764a 1777 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
e4d9b6cb 1778 if (c->ro_media) {
b466f17d 1779 ubifs_msg("cannot re-mount due to prior errors");
a2b9df3f 1780 return -EROFS;
e4d9b6cb 1781 }
1e51764a
AB
1782 err = ubifs_remount_rw(c);
1783 if (err)
1784 return err;
b466f17d
AH
1785 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY)) {
1786 if (c->ro_media) {
1787 ubifs_msg("cannot re-mount due to prior errors");
a2b9df3f 1788 return -EROFS;
b466f17d 1789 }
1e51764a 1790 ubifs_remount_ro(c);
b466f17d 1791 }
1e51764a 1792
3477d204
AB
1793 if (c->bulk_read == 1)
1794 bu_init(c);
1795 else {
1796 dbg_gen("disable bulk-read");
1797 kfree(c->bu.buf);
1798 c->bu.buf = NULL;
1799 }
1800
b221337a 1801 ubifs_assert(c->lst.taken_empty_lebs > 0);
1e51764a
AB
1802 return 0;
1803}
1804
e8b81566 1805const struct super_operations ubifs_super_operations = {
1e51764a
AB
1806 .alloc_inode = ubifs_alloc_inode,
1807 .destroy_inode = ubifs_destroy_inode,
1808 .put_super = ubifs_put_super,
1809 .write_inode = ubifs_write_inode,
1810 .delete_inode = ubifs_delete_inode,
1811 .statfs = ubifs_statfs,
1812 .dirty_inode = ubifs_dirty_inode,
1813 .remount_fs = ubifs_remount_fs,
1814 .show_options = ubifs_show_options,
1815 .sync_fs = ubifs_sync_fs,
1816};
1817
1818/**
1819 * open_ubi - parse UBI device name string and open the UBI device.
1820 * @name: UBI volume name
1821 * @mode: UBI volume open mode
1822 *
1823 * There are several ways to specify UBI volumes when mounting UBIFS:
1824 * o ubiX_Y - UBI device number X, volume Y;
1825 * o ubiY - UBI device number 0, volume Y;
1826 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1827 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1828 *
1829 * Alternative '!' separator may be used instead of ':' (because some shells
1830 * like busybox may interpret ':' as an NFS host name separator). This function
1831 * returns ubi volume object in case of success and a negative error code in
1832 * case of failure.
1833 */
1834static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1835{
1836 int dev, vol;
1837 char *endptr;
1838
1839 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1840 return ERR_PTR(-EINVAL);
1841
1842 /* ubi:NAME method */
1843 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1844 return ubi_open_volume_nm(0, name + 4, mode);
1845
1846 if (!isdigit(name[3]))
1847 return ERR_PTR(-EINVAL);
1848
1849 dev = simple_strtoul(name + 3, &endptr, 0);
1850
1851 /* ubiY method */
1852 if (*endptr == '\0')
1853 return ubi_open_volume(0, dev, mode);
1854
1855 /* ubiX_Y method */
1856 if (*endptr == '_' && isdigit(endptr[1])) {
1857 vol = simple_strtoul(endptr + 1, &endptr, 0);
1858 if (*endptr != '\0')
1859 return ERR_PTR(-EINVAL);
1860 return ubi_open_volume(dev, vol, mode);
1861 }
1862
1863 /* ubiX:NAME method */
1864 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1865 return ubi_open_volume_nm(dev, ++endptr, mode);
1866
1867 return ERR_PTR(-EINVAL);
1868}
1869
1870static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1871{
1872 struct ubi_volume_desc *ubi = sb->s_fs_info;
1873 struct ubifs_info *c;
1874 struct inode *root;
1875 int err;
1876
1877 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1878 if (!c)
1879 return -ENOMEM;
1880
1881 spin_lock_init(&c->cnt_lock);
1882 spin_lock_init(&c->cs_lock);
1883 spin_lock_init(&c->buds_lock);
1884 spin_lock_init(&c->space_lock);
1885 spin_lock_init(&c->orphan_lock);
1886 init_rwsem(&c->commit_sem);
1887 mutex_init(&c->lp_mutex);
1888 mutex_init(&c->tnc_mutex);
1889 mutex_init(&c->log_mutex);
1890 mutex_init(&c->mst_mutex);
1891 mutex_init(&c->umount_mutex);
3477d204 1892 mutex_init(&c->bu_mutex);
1e51764a
AB
1893 init_waitqueue_head(&c->cmt_wq);
1894 c->buds = RB_ROOT;
1895 c->old_idx = RB_ROOT;
1896 c->size_tree = RB_ROOT;
1897 c->orph_tree = RB_ROOT;
1898 INIT_LIST_HEAD(&c->infos_list);
1899 INIT_LIST_HEAD(&c->idx_gc);
1900 INIT_LIST_HEAD(&c->replay_list);
1901 INIT_LIST_HEAD(&c->replay_buds);
1902 INIT_LIST_HEAD(&c->uncat_list);
1903 INIT_LIST_HEAD(&c->empty_list);
1904 INIT_LIST_HEAD(&c->freeable_list);
1905 INIT_LIST_HEAD(&c->frdi_idx_list);
1906 INIT_LIST_HEAD(&c->unclean_leb_list);
1907 INIT_LIST_HEAD(&c->old_buds);
1908 INIT_LIST_HEAD(&c->orph_list);
1909 INIT_LIST_HEAD(&c->orph_new);
1910
1911 c->highest_inum = UBIFS_FIRST_INO;
1e51764a
AB
1912 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1913
1914 ubi_get_volume_info(ubi, &c->vi);
1915 ubi_get_device_info(c->vi.ubi_num, &c->di);
1916
1917 /* Re-open the UBI device in read-write mode */
1918 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1919 if (IS_ERR(c->ubi)) {
1920 err = PTR_ERR(c->ubi);
1921 goto out_free;
1922 }
1923
1924 /*
0a883a05 1925 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1e51764a
AB
1926 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1927 * which means the user would have to wait not just for their own I/O
0a883a05 1928 * but the read-ahead I/O as well i.e. completely pointless.
1e51764a
AB
1929 *
1930 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1931 */
1932 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1933 c->bdi.unplug_io_fn = default_unplug_io_fn;
1934 err = bdi_init(&c->bdi);
1935 if (err)
1936 goto out_close;
1937
1938 err = ubifs_parse_options(c, data, 0);
1939 if (err)
1940 goto out_bdi;
1941
1942 c->vfs_sb = sb;
1943
1944 sb->s_fs_info = c;
1945 sb->s_magic = UBIFS_SUPER_MAGIC;
1946 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1947 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1948 sb->s_dev = c->vi.cdev;
1949 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1950 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1951 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1952 sb->s_op = &ubifs_super_operations;
1953
1954 mutex_lock(&c->umount_mutex);
1955 err = mount_ubifs(c);
1956 if (err) {
1957 ubifs_assert(err < 0);
1958 goto out_unlock;
1959 }
1960
1961 /* Read the root inode */
1962 root = ubifs_iget(sb, UBIFS_ROOT_INO);
1963 if (IS_ERR(root)) {
1964 err = PTR_ERR(root);
1965 goto out_umount;
1966 }
1967
1968 sb->s_root = d_alloc_root(root);
1969 if (!sb->s_root)
1970 goto out_iput;
1971
1972 mutex_unlock(&c->umount_mutex);
1e51764a
AB
1973 return 0;
1974
1975out_iput:
1976 iput(root);
1977out_umount:
1978 ubifs_umount(c);
1979out_unlock:
1980 mutex_unlock(&c->umount_mutex);
1981out_bdi:
1982 bdi_destroy(&c->bdi);
1983out_close:
1984 ubi_close_volume(c->ubi);
1985out_free:
1986 kfree(c);
1987 return err;
1988}
1989
1990static int sb_test(struct super_block *sb, void *data)
1991{
1992 dev_t *dev = data;
1993
1994 return sb->s_dev == *dev;
1995}
1996
1997static int sb_set(struct super_block *sb, void *data)
1998{
1999 dev_t *dev = data;
2000
2001 sb->s_dev = *dev;
2002 return 0;
2003}
2004
2005static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
2006 const char *name, void *data, struct vfsmount *mnt)
2007{
2008 struct ubi_volume_desc *ubi;
2009 struct ubi_volume_info vi;
2010 struct super_block *sb;
2011 int err;
2012
2013 dbg_gen("name %s, flags %#x", name, flags);
2014
2015 /*
2016 * Get UBI device number and volume ID. Mount it read-only so far
2017 * because this might be a new mount point, and UBI allows only one
2018 * read-write user at a time.
2019 */
2020 ubi = open_ubi(name, UBI_READONLY);
2021 if (IS_ERR(ubi)) {
2022 ubifs_err("cannot open \"%s\", error %d",
2023 name, (int)PTR_ERR(ubi));
2024 return PTR_ERR(ubi);
2025 }
2026 ubi_get_volume_info(ubi, &vi);
2027
2028 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
2029
2030 sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
2031 if (IS_ERR(sb)) {
2032 err = PTR_ERR(sb);
2033 goto out_close;
2034 }
2035
2036 if (sb->s_root) {
2037 /* A new mount point for already mounted UBIFS */
2038 dbg_gen("this ubi volume is already mounted");
2039 if ((flags ^ sb->s_flags) & MS_RDONLY) {
2040 err = -EBUSY;
2041 goto out_deact;
2042 }
2043 } else {
2044 sb->s_flags = flags;
2045 /*
2046 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
2047 * replaced by 'c'.
2048 */
2049 sb->s_fs_info = ubi;
2050 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2051 if (err)
2052 goto out_deact;
2053 /* We do not support atime */
2054 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2055 }
2056
2057 /* 'fill_super()' opens ubi again so we must close it here */
2058 ubi_close_volume(ubi);
2059
a3ec947c
SB
2060 simple_set_mnt(mnt, sb);
2061 return 0;
1e51764a
AB
2062
2063out_deact:
2064 up_write(&sb->s_umount);
2065 deactivate_super(sb);
2066out_close:
2067 ubi_close_volume(ubi);
2068 return err;
2069}
2070
2071static void ubifs_kill_sb(struct super_block *sb)
2072{
1e51764a
AB
2073 generic_shutdown_super(sb);
2074}
2075
2076static struct file_system_type ubifs_fs_type = {
2077 .name = "ubifs",
2078 .owner = THIS_MODULE,
2079 .get_sb = ubifs_get_sb,
2080 .kill_sb = ubifs_kill_sb
2081};
2082
2083/*
2084 * Inode slab cache constructor.
2085 */
51cc5068 2086static void inode_slab_ctor(void *obj)
1e51764a
AB
2087{
2088 struct ubifs_inode *ui = obj;
2089 inode_init_once(&ui->vfs_inode);
2090}
2091
2092static int __init ubifs_init(void)
2093{
2094 int err;
2095
2096 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2097
2098 /* Make sure node sizes are 8-byte aligned */
2099 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2100 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2101 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2102 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2103 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2104 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2105 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2106 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2107 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2108 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2109 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2110
2111 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2112 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2113 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2114 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2115 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2116 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2117
2118 /* Check min. node size */
2119 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2120 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2121 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2122 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2123
2124 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2125 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2126 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2127 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2128
2129 /* Defined node sizes */
2130 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2131 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2132 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2133 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2134
a1dc080c
AB
2135 /*
2136 * We use 2 bit wide bit-fields to store compression type, which should
2137 * be amended if more compressors are added. The bit-fields are:
553dea4d
AB
2138 * @compr_type in 'struct ubifs_inode', @default_compr in
2139 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
a1dc080c
AB
2140 */
2141 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2142
1e51764a
AB
2143 /*
2144 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2145 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2146 */
2147 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2148 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2149 " at least 4096 bytes",
2150 (unsigned int)PAGE_CACHE_SIZE);
2151 return -EINVAL;
2152 }
2153
2154 err = register_filesystem(&ubifs_fs_type);
2155 if (err) {
2156 ubifs_err("cannot register file system, error %d", err);
2157 return err;
2158 }
2159
2160 err = -ENOMEM;
2161 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2162 sizeof(struct ubifs_inode), 0,
2163 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2164 &inode_slab_ctor);
2165 if (!ubifs_inode_slab)
2166 goto out_reg;
2167
2168 register_shrinker(&ubifs_shrinker_info);
2169
2170 err = ubifs_compressors_init();
552ff317
AB
2171 if (err)
2172 goto out_shrinker;
2173
2174 err = dbg_debugfs_init();
1e51764a
AB
2175 if (err)
2176 goto out_compr;
2177
2178 return 0;
2179
2180out_compr:
552ff317
AB
2181 ubifs_compressors_exit();
2182out_shrinker:
1e51764a
AB
2183 unregister_shrinker(&ubifs_shrinker_info);
2184 kmem_cache_destroy(ubifs_inode_slab);
2185out_reg:
2186 unregister_filesystem(&ubifs_fs_type);
2187 return err;
2188}
2189/* late_initcall to let compressors initialize first */
2190late_initcall(ubifs_init);
2191
2192static void __exit ubifs_exit(void)
2193{
2194 ubifs_assert(list_empty(&ubifs_infos));
2195 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2196
552ff317 2197 dbg_debugfs_exit();
1e51764a
AB
2198 ubifs_compressors_exit();
2199 unregister_shrinker(&ubifs_shrinker_info);
2200 kmem_cache_destroy(ubifs_inode_slab);
2201 unregister_filesystem(&ubifs_fs_type);
2202}
2203module_exit(ubifs_exit);
2204
2205MODULE_LICENSE("GPL");
2206MODULE_VERSION(__stringify(UBIFS_VERSION));
2207MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2208MODULE_DESCRIPTION("UBIFS - UBI File System");