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