]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ubifs/super.c
xps: Transmit Packet Steering
[net-next-2.6.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 */
a9185b41 286static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
1e51764a 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);
e3c3efc2
AB
320 else
321 err = dbg_check_inode_size(c, inode, ui->ui_size);
fbfa6c88 322 }
1e51764a
AB
323
324 ui->dirty = 0;
325 mutex_unlock(&ui->ui_mutex);
326 ubifs_release_dirty_inode_budget(c, ui);
327 return err;
328}
329
d640e1b5 330static void ubifs_evict_inode(struct inode *inode)
1e51764a
AB
331{
332 int err;
333 struct ubifs_info *c = inode->i_sb->s_fs_info;
1e0f358e 334 struct ubifs_inode *ui = ubifs_inode(inode);
1e51764a 335
1e0f358e 336 if (ui->xattr)
1e51764a
AB
337 /*
338 * Extended attribute inode deletions are fully handled in
339 * 'ubifs_removexattr()'. These inodes are special and have
340 * limited usage, so there is nothing to do here.
341 */
342 goto out;
343
7d32c2bb 344 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
1e51764a 345 ubifs_assert(!atomic_read(&inode->i_count));
1e51764a
AB
346
347 truncate_inode_pages(&inode->i_data, 0);
d640e1b5
AV
348
349 if (inode->i_nlink)
350 goto done;
351
1e51764a
AB
352 if (is_bad_inode(inode))
353 goto out;
354
1e0f358e 355 ui->ui_size = inode->i_size = 0;
de94eb55 356 err = ubifs_jnl_delete_inode(c, inode);
1e51764a
AB
357 if (err)
358 /*
359 * Worst case we have a lost orphan inode wasting space, so a
0a883a05 360 * simple error message is OK here.
1e51764a 361 */
de94eb55
AB
362 ubifs_err("can't delete inode %lu, error %d",
363 inode->i_ino, err);
364
1e51764a 365out:
1e0f358e
AB
366 if (ui->dirty)
367 ubifs_release_dirty_inode_budget(c, ui);
6d6cb0d6
AH
368 else {
369 /* We've deleted something - clean the "no space" flags */
370 c->nospace = c->nospace_rp = 0;
371 smp_wmb();
372 }
d640e1b5
AV
373done:
374 end_writeback(inode);
1e51764a
AB
375}
376
377static void ubifs_dirty_inode(struct inode *inode)
378{
379 struct ubifs_inode *ui = ubifs_inode(inode);
380
381 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
382 if (!ui->dirty) {
383 ui->dirty = 1;
384 dbg_gen("inode %lu", inode->i_ino);
385 }
386}
387
388static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
389{
390 struct ubifs_info *c = dentry->d_sb->s_fs_info;
391 unsigned long long free;
7c7cbadf 392 __le32 *uuid = (__le32 *)c->uuid;
1e51764a 393
7dad181b 394 free = ubifs_get_free_space(c);
1e51764a
AB
395 dbg_gen("free space %lld bytes (%lld blocks)",
396 free, free >> UBIFS_BLOCK_SHIFT);
397
398 buf->f_type = UBIFS_SUPER_MAGIC;
399 buf->f_bsize = UBIFS_BLOCK_SIZE;
400 buf->f_blocks = c->block_cnt;
401 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
402 if (free > c->report_rp_size)
403 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
404 else
405 buf->f_bavail = 0;
406 buf->f_files = 0;
407 buf->f_ffree = 0;
408 buf->f_namelen = UBIFS_MAX_NLEN;
7c7cbadf
AB
409 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
410 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
b4978e94 411 ubifs_assert(buf->f_bfree <= c->block_cnt);
1e51764a
AB
412 return 0;
413}
414
415static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
416{
417 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
418
419 if (c->mount_opts.unmount_mode == 2)
420 seq_printf(s, ",fast_unmount");
421 else if (c->mount_opts.unmount_mode == 1)
422 seq_printf(s, ",norm_unmount");
423
4793e7c5
AH
424 if (c->mount_opts.bulk_read == 2)
425 seq_printf(s, ",bulk_read");
426 else if (c->mount_opts.bulk_read == 1)
427 seq_printf(s, ",no_bulk_read");
428
2953e73f
AH
429 if (c->mount_opts.chk_data_crc == 2)
430 seq_printf(s, ",chk_data_crc");
431 else if (c->mount_opts.chk_data_crc == 1)
432 seq_printf(s, ",no_chk_data_crc");
433
553dea4d 434 if (c->mount_opts.override_compr) {
fcabb347
HA
435 seq_printf(s, ",compr=%s",
436 ubifs_compr_name(c->mount_opts.compr_type));
553dea4d
AB
437 }
438
1e51764a
AB
439 return 0;
440}
441
442static int ubifs_sync_fs(struct super_block *sb, int wait)
443{
f1038300 444 int i, err;
1e51764a 445 struct ubifs_info *c = sb->s_fs_info;
304d427c 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
3eb14297
AH
455 /*
456 * Synchronize write buffers, because 'ubifs_run_commit()' does not
457 * do this if it waits for an already running commit.
458 */
459 for (i = 0; i < c->jhead_cnt; i++) {
460 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
461 if (err)
462 return err;
463 }
464
887ee171
AB
465 /*
466 * Strictly speaking, it is not necessary to commit the journal here,
467 * synchronizing write-buffers would be enough. But committing makes
468 * UBIFS free space predictions much more accurate, so we want to let
469 * the user be able to get more accurate results of 'statfs()' after
470 * they synchronize the file system.
471 */
f1038300
AB
472 err = ubifs_run_commit(c);
473 if (err)
474 return err;
403e12ab 475
cb5c6a2b 476 return ubi_sync(c->vi.ubi_num);
1e51764a
AB
477}
478
479/**
480 * init_constants_early - initialize UBIFS constants.
481 * @c: UBIFS file-system description object
482 *
483 * This function initialize UBIFS constants which do not need the superblock to
484 * be read. It also checks that the UBI volume satisfies basic UBIFS
485 * requirements. Returns zero in case of success and a negative error code in
486 * case of failure.
487 */
488static int init_constants_early(struct ubifs_info *c)
489{
490 if (c->vi.corrupted) {
491 ubifs_warn("UBI volume is corrupted - read-only mode");
492 c->ro_media = 1;
493 }
494
495 if (c->di.ro_mode) {
496 ubifs_msg("read-only UBI device");
497 c->ro_media = 1;
498 }
499
500 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
501 ubifs_msg("static UBI volume - read-only mode");
502 c->ro_media = 1;
503 }
504
505 c->leb_cnt = c->vi.size;
506 c->leb_size = c->vi.usable_leb_size;
507 c->half_leb_size = c->leb_size / 2;
508 c->min_io_size = c->di.min_io_size;
509 c->min_io_shift = fls(c->min_io_size) - 1;
510
511 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
512 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
513 c->leb_size, UBIFS_MIN_LEB_SZ);
514 return -EINVAL;
515 }
516
517 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
518 ubifs_err("too few LEBs (%d), min. is %d",
519 c->leb_cnt, UBIFS_MIN_LEB_CNT);
520 return -EINVAL;
521 }
522
523 if (!is_power_of_2(c->min_io_size)) {
524 ubifs_err("bad min. I/O size %d", c->min_io_size);
525 return -EINVAL;
526 }
527
528 /*
529 * UBIFS aligns all node to 8-byte boundary, so to make function in
530 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
531 * less than 8.
532 */
533 if (c->min_io_size < 8) {
534 c->min_io_size = 8;
535 c->min_io_shift = 3;
536 }
537
538 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
539 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
540
541 /*
542 * Initialize node length ranges which are mostly needed for node
543 * length validation.
544 */
545 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
546 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
547 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
548 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
549 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
550 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
551
552 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
553 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
554 c->ranges[UBIFS_ORPH_NODE].min_len =
555 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
556 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
557 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
558 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
559 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
560 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
561 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
562 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
563 /*
564 * Minimum indexing node size is amended later when superblock is
565 * read and the key length is known.
566 */
567 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
568 /*
569 * Maximum indexing node size is amended later when superblock is
570 * read and the fanout is known.
571 */
572 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
573
574 /*
7078202e
AB
575 * Initialize dead and dark LEB space watermarks. See gc.c for comments
576 * about these values.
1e51764a
AB
577 */
578 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
579 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
580
9bbb5726
AB
581 /*
582 * Calculate how many bytes would be wasted at the end of LEB if it was
583 * fully filled with data nodes of maximum size. This is used in
584 * calculations when reporting free space.
585 */
586 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
39ce81ce 587
4793e7c5 588 /* Buffer size for bulk-reads */
6c0c42cd
AB
589 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
590 if (c->max_bu_buf_len > c->leb_size)
591 c->max_bu_buf_len = c->leb_size;
1e51764a
AB
592 return 0;
593}
594
595/**
596 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
597 * @c: UBIFS file-system description object
598 * @lnum: LEB the write-buffer was synchronized to
599 * @free: how many free bytes left in this LEB
600 * @pad: how many bytes were padded
601 *
602 * This is a callback function which is called by the I/O unit when the
603 * write-buffer is synchronized. We need this to correctly maintain space
604 * accounting in bud logical eraseblocks. This function returns zero in case of
605 * success and a negative error code in case of failure.
606 *
607 * This function actually belongs to the journal, but we keep it here because
608 * we want to keep it static.
609 */
610static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
611{
612 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
613}
614
615/*
79807d07 616 * init_constants_sb - initialize UBIFS constants.
1e51764a
AB
617 * @c: UBIFS file-system description object
618 *
619 * This is a helper function which initializes various UBIFS constants after
620 * the superblock has been read. It also checks various UBIFS parameters and
621 * makes sure they are all right. Returns zero in case of success and a
622 * negative error code in case of failure.
623 */
79807d07 624static int init_constants_sb(struct ubifs_info *c)
1e51764a
AB
625{
626 int tmp, err;
4d61db4f 627 long long tmp64;
1e51764a
AB
628
629 c->main_bytes = (long long)c->main_lebs * c->leb_size;
630 c->max_znode_sz = sizeof(struct ubifs_znode) +
631 c->fanout * sizeof(struct ubifs_zbranch);
632
633 tmp = ubifs_idx_node_sz(c, 1);
634 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
635 c->min_idx_node_sz = ALIGN(tmp, 8);
636
637 tmp = ubifs_idx_node_sz(c, c->fanout);
638 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
639 c->max_idx_node_sz = ALIGN(tmp, 8);
640
641 /* Make sure LEB size is large enough to fit full commit */
642 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
643 tmp = ALIGN(tmp, c->min_io_size);
644 if (tmp > c->leb_size) {
645 dbg_err("too small LEB size %d, at least %d needed",
646 c->leb_size, tmp);
647 return -EINVAL;
648 }
649
650 /*
651 * Make sure that the log is large enough to fit reference nodes for
652 * all buds plus one reserved LEB.
653 */
4d61db4f
AB
654 tmp64 = c->max_bud_bytes + c->leb_size - 1;
655 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
1e51764a
AB
656 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
657 tmp /= c->leb_size;
658 tmp += 1;
659 if (c->log_lebs < tmp) {
660 dbg_err("too small log %d LEBs, required min. %d LEBs",
661 c->log_lebs, tmp);
662 return -EINVAL;
663 }
664
665 /*
666 * When budgeting we assume worst-case scenarios when the pages are not
667 * be compressed and direntries are of the maximum size.
668 *
669 * Note, data, which may be stored in inodes is budgeted separately, so
670 * it is not included into 'c->inode_budget'.
671 */
672 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
673 c->inode_budget = UBIFS_INO_NODE_SZ;
674 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
675
676 /*
677 * When the amount of flash space used by buds becomes
678 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
679 * The writers are unblocked when the commit is finished. To avoid
680 * writers to be blocked UBIFS initiates background commit in advance,
681 * when number of bud bytes becomes above the limit defined below.
682 */
683 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
684
685 /*
686 * Ensure minimum journal size. All the bytes in the journal heads are
687 * considered to be used, when calculating the current journal usage.
688 * Consequently, if the journal is too small, UBIFS will treat it as
689 * always full.
690 */
4d61db4f 691 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
1e51764a
AB
692 if (c->bg_bud_bytes < tmp64)
693 c->bg_bud_bytes = tmp64;
694 if (c->max_bud_bytes < tmp64 + c->leb_size)
695 c->max_bud_bytes = tmp64 + c->leb_size;
696
697 err = ubifs_calc_lpt_geom(c);
698 if (err)
699 return err;
700
fb1cd01a
AB
701 /* Initialize effective LEB size used in budgeting calculations */
702 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
79807d07
AB
703 return 0;
704}
705
706/*
707 * init_constants_master - initialize UBIFS constants.
708 * @c: UBIFS file-system description object
709 *
710 * This is a helper function which initializes various UBIFS constants after
711 * the master node has been read. It also checks various UBIFS parameters and
712 * makes sure they are all right.
713 */
714static void init_constants_master(struct ubifs_info *c)
715{
716 long long tmp64;
717
1e51764a 718 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
fb1cd01a 719 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
1e51764a
AB
720
721 /*
722 * Calculate total amount of FS blocks. This number is not used
723 * internally because it does not make much sense for UBIFS, but it is
724 * necessary to report something for the 'statfs()' call.
725 *
7dad181b 726 * Subtract the LEB reserved for GC, the LEB which is reserved for
af14a1ad
AB
727 * deletions, minimum LEBs for the index, and assume only one journal
728 * head is available.
1e51764a 729 */
af14a1ad 730 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
4d61db4f 731 tmp64 *= (long long)c->leb_size - c->leb_overhead;
1e51764a
AB
732 tmp64 = ubifs_reported_space(c, tmp64);
733 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
1e51764a
AB
734}
735
736/**
737 * take_gc_lnum - reserve GC LEB.
738 * @c: UBIFS file-system description object
739 *
b4978e94
AB
740 * This function ensures that the LEB reserved for garbage collection is marked
741 * as "taken" in lprops. We also have to set free space to LEB size and dirty
742 * space to zero, because lprops may contain out-of-date information if the
743 * file-system was un-mounted before it has been committed. This function
744 * returns zero in case of success and a negative error code in case of
745 * failure.
1e51764a
AB
746 */
747static int take_gc_lnum(struct ubifs_info *c)
748{
749 int err;
750
751 if (c->gc_lnum == -1) {
752 ubifs_err("no LEB for GC");
753 return -EINVAL;
754 }
755
1e51764a
AB
756 /* And we have to tell lprops that this LEB is taken */
757 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
758 LPROPS_TAKEN, 0, 0);
759 return err;
760}
761
762/**
763 * alloc_wbufs - allocate write-buffers.
764 * @c: UBIFS file-system description object
765 *
766 * This helper function allocates and initializes UBIFS write-buffers. Returns
767 * zero in case of success and %-ENOMEM in case of failure.
768 */
769static int alloc_wbufs(struct ubifs_info *c)
770{
771 int i, err;
772
773 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
774 GFP_KERNEL);
775 if (!c->jheads)
776 return -ENOMEM;
777
778 /* Initialize journal heads */
779 for (i = 0; i < c->jhead_cnt; i++) {
780 INIT_LIST_HEAD(&c->jheads[i].buds_list);
781 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
782 if (err)
783 return err;
784
785 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
786 c->jheads[i].wbuf.jhead = i;
787 }
788
789 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
790 /*
791 * Garbage Collector head likely contains long-term data and
792 * does not need to be synchronized by timer.
793 */
794 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
0b335b9d 795 c->jheads[GCHD].wbuf.no_timer = 1;
1e51764a
AB
796
797 return 0;
798}
799
800/**
801 * free_wbufs - free write-buffers.
802 * @c: UBIFS file-system description object
803 */
804static void free_wbufs(struct ubifs_info *c)
805{
806 int i;
807
808 if (c->jheads) {
809 for (i = 0; i < c->jhead_cnt; i++) {
810 kfree(c->jheads[i].wbuf.buf);
811 kfree(c->jheads[i].wbuf.inodes);
812 }
813 kfree(c->jheads);
814 c->jheads = NULL;
815 }
816}
817
818/**
819 * free_orphans - free orphans.
820 * @c: UBIFS file-system description object
821 */
822static void free_orphans(struct ubifs_info *c)
823{
824 struct ubifs_orphan *orph;
825
826 while (c->orph_dnext) {
827 orph = c->orph_dnext;
828 c->orph_dnext = orph->dnext;
829 list_del(&orph->list);
830 kfree(orph);
831 }
832
833 while (!list_empty(&c->orph_list)) {
834 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
835 list_del(&orph->list);
836 kfree(orph);
837 dbg_err("orphan list not empty at unmount");
838 }
839
840 vfree(c->orph_buf);
841 c->orph_buf = NULL;
842}
843
844/**
845 * free_buds - free per-bud objects.
846 * @c: UBIFS file-system description object
847 */
848static void free_buds(struct ubifs_info *c)
849{
850 struct rb_node *this = c->buds.rb_node;
851 struct ubifs_bud *bud;
852
853 while (this) {
854 if (this->rb_left)
855 this = this->rb_left;
856 else if (this->rb_right)
857 this = this->rb_right;
858 else {
859 bud = rb_entry(this, struct ubifs_bud, rb);
860 this = rb_parent(this);
861 if (this) {
862 if (this->rb_left == &bud->rb)
863 this->rb_left = NULL;
864 else
865 this->rb_right = NULL;
866 }
867 kfree(bud);
868 }
869 }
870}
871
872/**
873 * check_volume_empty - check if the UBI volume is empty.
874 * @c: UBIFS file-system description object
875 *
876 * This function checks if the UBIFS volume is empty by looking if its LEBs are
877 * mapped or not. The result of checking is stored in the @c->empty variable.
878 * Returns zero in case of success and a negative error code in case of
879 * failure.
880 */
881static int check_volume_empty(struct ubifs_info *c)
882{
883 int lnum, err;
884
885 c->empty = 1;
886 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
887 err = ubi_is_mapped(c->ubi, lnum);
888 if (unlikely(err < 0))
889 return err;
890 if (err == 1) {
891 c->empty = 0;
892 break;
893 }
894
895 cond_resched();
896 }
897
898 return 0;
899}
900
901/*
902 * UBIFS mount options.
903 *
904 * Opt_fast_unmount: do not run a journal commit before un-mounting
905 * Opt_norm_unmount: run a journal commit before un-mounting
4793e7c5
AH
906 * Opt_bulk_read: enable bulk-reads
907 * Opt_no_bulk_read: disable bulk-reads
2953e73f
AH
908 * Opt_chk_data_crc: check CRCs when reading data nodes
909 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
553dea4d 910 * Opt_override_compr: override default compressor
1e51764a
AB
911 * Opt_err: just end of array marker
912 */
913enum {
914 Opt_fast_unmount,
915 Opt_norm_unmount,
4793e7c5
AH
916 Opt_bulk_read,
917 Opt_no_bulk_read,
2953e73f
AH
918 Opt_chk_data_crc,
919 Opt_no_chk_data_crc,
553dea4d 920 Opt_override_compr,
1e51764a
AB
921 Opt_err,
922};
923
a447c093 924static const match_table_t tokens = {
1e51764a
AB
925 {Opt_fast_unmount, "fast_unmount"},
926 {Opt_norm_unmount, "norm_unmount"},
4793e7c5
AH
927 {Opt_bulk_read, "bulk_read"},
928 {Opt_no_bulk_read, "no_bulk_read"},
2953e73f
AH
929 {Opt_chk_data_crc, "chk_data_crc"},
930 {Opt_no_chk_data_crc, "no_chk_data_crc"},
553dea4d 931 {Opt_override_compr, "compr=%s"},
1e51764a
AB
932 {Opt_err, NULL},
933};
934
8379ea31
AB
935/**
936 * parse_standard_option - parse a standard mount option.
937 * @option: the option to parse
938 *
939 * Normally, standard mount options like "sync" are passed to file-systems as
940 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
941 * be present in the options string. This function tries to deal with this
942 * situation and parse standard options. Returns 0 if the option was not
943 * recognized, and the corresponding integer flag if it was.
944 *
945 * UBIFS is only interested in the "sync" option, so do not check for anything
946 * else.
947 */
948static int parse_standard_option(const char *option)
949{
950 ubifs_msg("parse %s", option);
951 if (!strcmp(option, "sync"))
952 return MS_SYNCHRONOUS;
953 return 0;
954}
955
1e51764a
AB
956/**
957 * ubifs_parse_options - parse mount parameters.
958 * @c: UBIFS file-system description object
959 * @options: parameters to parse
960 * @is_remount: non-zero if this is FS re-mount
961 *
962 * This function parses UBIFS mount options and returns zero in case success
963 * and a negative error code in case of failure.
964 */
965static int ubifs_parse_options(struct ubifs_info *c, char *options,
966 int is_remount)
967{
968 char *p;
969 substring_t args[MAX_OPT_ARGS];
970
971 if (!options)
972 return 0;
973
974 while ((p = strsep(&options, ","))) {
975 int token;
976
977 if (!*p)
978 continue;
979
980 token = match_token(p, tokens, args);
981 switch (token) {
27ad2799
AB
982 /*
983 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
cb54ef8b 984 * We accept them in order to be backward-compatible. But this
27ad2799
AB
985 * should be removed at some point.
986 */
1e51764a
AB
987 case Opt_fast_unmount:
988 c->mount_opts.unmount_mode = 2;
1e51764a
AB
989 break;
990 case Opt_norm_unmount:
991 c->mount_opts.unmount_mode = 1;
1e51764a 992 break;
4793e7c5
AH
993 case Opt_bulk_read:
994 c->mount_opts.bulk_read = 2;
995 c->bulk_read = 1;
996 break;
997 case Opt_no_bulk_read:
998 c->mount_opts.bulk_read = 1;
999 c->bulk_read = 0;
1000 break;
2953e73f
AH
1001 case Opt_chk_data_crc:
1002 c->mount_opts.chk_data_crc = 2;
1003 c->no_chk_data_crc = 0;
1004 break;
1005 case Opt_no_chk_data_crc:
1006 c->mount_opts.chk_data_crc = 1;
1007 c->no_chk_data_crc = 1;
1008 break;
553dea4d
AB
1009 case Opt_override_compr:
1010 {
1011 char *name = match_strdup(&args[0]);
1012
1013 if (!name)
1014 return -ENOMEM;
1015 if (!strcmp(name, "none"))
1016 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1017 else if (!strcmp(name, "lzo"))
1018 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1019 else if (!strcmp(name, "zlib"))
1020 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1021 else {
1022 ubifs_err("unknown compressor \"%s\"", name);
1023 kfree(name);
1024 return -EINVAL;
1025 }
1026 kfree(name);
1027 c->mount_opts.override_compr = 1;
1028 c->default_compr = c->mount_opts.compr_type;
1029 break;
1030 }
1e51764a 1031 default:
8379ea31
AB
1032 {
1033 unsigned long flag;
1034 struct super_block *sb = c->vfs_sb;
1035
1036 flag = parse_standard_option(p);
1037 if (!flag) {
1038 ubifs_err("unrecognized mount option \"%s\" "
1039 "or missing value", p);
1040 return -EINVAL;
1041 }
1042 sb->s_flags |= flag;
1043 break;
1044 }
1e51764a
AB
1045 }
1046 }
1047
1048 return 0;
1049}
1050
1051/**
1052 * destroy_journal - destroy journal data structures.
1053 * @c: UBIFS file-system description object
1054 *
1055 * This function destroys journal data structures including those that may have
1056 * been created by recovery functions.
1057 */
1058static void destroy_journal(struct ubifs_info *c)
1059{
1060 while (!list_empty(&c->unclean_leb_list)) {
1061 struct ubifs_unclean_leb *ucleb;
1062
1063 ucleb = list_entry(c->unclean_leb_list.next,
1064 struct ubifs_unclean_leb, list);
1065 list_del(&ucleb->list);
1066 kfree(ucleb);
1067 }
1068 while (!list_empty(&c->old_buds)) {
1069 struct ubifs_bud *bud;
1070
1071 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1072 list_del(&bud->list);
1073 kfree(bud);
1074 }
1075 ubifs_destroy_idx_gc(c);
1076 ubifs_destroy_size_tree(c);
1077 ubifs_tnc_close(c);
1078 free_buds(c);
1079}
1080
3477d204
AB
1081/**
1082 * bu_init - initialize bulk-read information.
1083 * @c: UBIFS file-system description object
1084 */
1085static void bu_init(struct ubifs_info *c)
1086{
1087 ubifs_assert(c->bulk_read == 1);
1088
1089 if (c->bu.buf)
1090 return; /* Already initialized */
1091
1092again:
1093 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1094 if (!c->bu.buf) {
1095 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1096 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1097 goto again;
1098 }
1099
1100 /* Just disable bulk-read */
1101 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1102 "disabling it", c->max_bu_buf_len);
1103 c->mount_opts.bulk_read = 1;
1104 c->bulk_read = 0;
1105 return;
1106 }
1107}
1108
57a450e9
AB
1109/**
1110 * check_free_space - check if there is enough free space to mount.
1111 * @c: UBIFS file-system description object
1112 *
1113 * This function makes sure UBIFS has enough free space to be mounted in
1114 * read/write mode. UBIFS must always have some free space to allow deletions.
1115 */
1116static int check_free_space(struct ubifs_info *c)
1117{
1118 ubifs_assert(c->dark_wm > 0);
1119 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1120 ubifs_err("insufficient free space to mount in read/write mode");
1121 dbg_dump_budg(c);
1122 dbg_dump_lprops(c);
a2b9df3f 1123 return -ENOSPC;
57a450e9
AB
1124 }
1125 return 0;
1126}
1127
1e51764a
AB
1128/**
1129 * mount_ubifs - mount UBIFS file-system.
1130 * @c: UBIFS file-system description object
1131 *
1132 * This function mounts UBIFS file system. Returns zero in case of success and
1133 * a negative error code in case of failure.
1134 *
1135 * Note, the function does not de-allocate resources it it fails half way
1136 * through, and the caller has to do this instead.
1137 */
1138static int mount_ubifs(struct ubifs_info *c)
1139{
2ef13294 1140 int err;
1e51764a
AB
1141 long long x;
1142 size_t sz;
1143
2ef13294 1144 c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1e51764a
AB
1145 err = init_constants_early(c);
1146 if (err)
1147 return err;
1148
17c2f9f8
AB
1149 err = ubifs_debugging_init(c);
1150 if (err)
1151 return err;
1e51764a
AB
1152
1153 err = check_volume_empty(c);
1154 if (err)
1155 goto out_free;
1156
2ef13294 1157 if (c->empty && (c->ro_mount || c->ro_media)) {
1e51764a
AB
1158 /*
1159 * This UBI volume is empty, and read-only, or the file system
1160 * is mounted read-only - we cannot format it.
1161 */
1162 ubifs_err("can't format empty UBI volume: read-only %s",
1163 c->ro_media ? "UBI volume" : "mount");
1164 err = -EROFS;
1165 goto out_free;
1166 }
1167
2ef13294 1168 if (c->ro_media && !c->ro_mount) {
1e51764a
AB
1169 ubifs_err("cannot mount read-write - read-only media");
1170 err = -EROFS;
1171 goto out_free;
1172 }
1173
1174 /*
1175 * The requirement for the buffer is that it should fit indexing B-tree
1176 * height amount of integers. We assume the height if the TNC tree will
1177 * never exceed 64.
1178 */
1179 err = -ENOMEM;
1180 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1181 if (!c->bottom_up_buf)
1182 goto out_free;
1183
1184 c->sbuf = vmalloc(c->leb_size);
1185 if (!c->sbuf)
1186 goto out_free;
1187
2ef13294 1188 if (!c->ro_mount) {
1e51764a
AB
1189 c->ileb_buf = vmalloc(c->leb_size);
1190 if (!c->ileb_buf)
1191 goto out_free;
1192 }
1193
3477d204
AB
1194 if (c->bulk_read == 1)
1195 bu_init(c);
1196
1197 /*
1198 * We have to check all CRCs, even for data nodes, when we mount the FS
1199 * (specifically, when we are replaying).
1200 */
2953e73f
AH
1201 c->always_chk_crc = 1;
1202
1e51764a
AB
1203 err = ubifs_read_superblock(c);
1204 if (err)
1205 goto out_free;
1206
1207 /*
553dea4d 1208 * Make sure the compressor which is set as default in the superblock
57a450e9 1209 * or overridden by mount options is actually compiled in.
1e51764a
AB
1210 */
1211 if (!ubifs_compr_present(c->default_compr)) {
553dea4d
AB
1212 ubifs_err("'compressor \"%s\" is not compiled in",
1213 ubifs_compr_name(c->default_compr));
8eec2f36 1214 err = -ENOTSUPP;
553dea4d 1215 goto out_free;
1e51764a
AB
1216 }
1217
79807d07 1218 err = init_constants_sb(c);
1e51764a 1219 if (err)
17c2f9f8 1220 goto out_free;
1e51764a
AB
1221
1222 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1223 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1224 c->cbuf = kmalloc(sz, GFP_NOFS);
1225 if (!c->cbuf) {
1226 err = -ENOMEM;
17c2f9f8 1227 goto out_free;
1e51764a
AB
1228 }
1229
0855f310 1230 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
2ef13294 1231 if (!c->ro_mount) {
1e51764a
AB
1232 err = alloc_wbufs(c);
1233 if (err)
1234 goto out_cbuf;
1235
1236 /* Create background thread */
fcabb347 1237 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1e51764a
AB
1238 if (IS_ERR(c->bgt)) {
1239 err = PTR_ERR(c->bgt);
1240 c->bgt = NULL;
1241 ubifs_err("cannot spawn \"%s\", error %d",
1242 c->bgt_name, err);
1243 goto out_wbufs;
1244 }
1245 wake_up_process(c->bgt);
1246 }
1247
1248 err = ubifs_read_master(c);
1249 if (err)
1250 goto out_master;
1251
79807d07
AB
1252 init_constants_master(c);
1253
1e51764a
AB
1254 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1255 ubifs_msg("recovery needed");
1256 c->need_recovery = 1;
2ef13294 1257 if (!c->ro_mount) {
1e51764a
AB
1258 err = ubifs_recover_inl_heads(c, c->sbuf);
1259 if (err)
1260 goto out_master;
1261 }
2ef13294 1262 } else if (!c->ro_mount) {
1e51764a
AB
1263 /*
1264 * Set the "dirty" flag so that if we reboot uncleanly we
1265 * will notice this immediately on the next mount.
1266 */
1267 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1268 err = ubifs_write_master(c);
1269 if (err)
1270 goto out_master;
1271 }
1272
2ef13294 1273 err = ubifs_lpt_init(c, 1, !c->ro_mount);
1e51764a
AB
1274 if (err)
1275 goto out_lpt;
1276
1277 err = dbg_check_idx_size(c, c->old_idx_sz);
1278 if (err)
1279 goto out_lpt;
1280
1281 err = ubifs_replay_journal(c);
1282 if (err)
1283 goto out_journal;
1284
1fb8bd01
AB
1285 /* Calculate 'min_idx_lebs' after journal replay */
1286 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1287
2ef13294 1288 err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1e51764a
AB
1289 if (err)
1290 goto out_orphans;
1291
2ef13294 1292 if (!c->ro_mount) {
1e51764a
AB
1293 int lnum;
1294
57a450e9
AB
1295 err = check_free_space(c);
1296 if (err)
1e51764a 1297 goto out_orphans;
1e51764a
AB
1298
1299 /* Check for enough log space */
1300 lnum = c->lhead_lnum + 1;
1301 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1302 lnum = UBIFS_LOG_LNUM;
1303 if (lnum == c->ltail_lnum) {
1304 err = ubifs_consolidate_log(c);
1305 if (err)
1306 goto out_orphans;
1307 }
1308
1309 if (c->need_recovery) {
1310 err = ubifs_recover_size(c);
1311 if (err)
1312 goto out_orphans;
1313 err = ubifs_rcvry_gc_commit(c);
276de5d2
AB
1314 if (err)
1315 goto out_orphans;
b4978e94 1316 } else {
1e51764a 1317 err = take_gc_lnum(c);
b4978e94
AB
1318 if (err)
1319 goto out_orphans;
1320
1321 /*
1322 * GC LEB may contain garbage if there was an unclean
1323 * reboot, and it should be un-mapped.
1324 */
1325 err = ubifs_leb_unmap(c, c->gc_lnum);
1326 if (err)
c18de72f 1327 goto out_orphans;
b4978e94 1328 }
1e51764a
AB
1329
1330 err = dbg_check_lprops(c);
1331 if (err)
1332 goto out_orphans;
1333 } else if (c->need_recovery) {
1334 err = ubifs_recover_size(c);
1335 if (err)
1336 goto out_orphans;
b4978e94
AB
1337 } else {
1338 /*
1339 * Even if we mount read-only, we have to set space in GC LEB
1340 * to proper value because this affects UBIFS free space
1341 * reporting. We do not want to have a situation when
1342 * re-mounting from R/O to R/W changes amount of free space.
1343 */
1344 err = take_gc_lnum(c);
1345 if (err)
1346 goto out_orphans;
1e51764a
AB
1347 }
1348
1349 spin_lock(&ubifs_infos_lock);
1350 list_add_tail(&c->infos_list, &ubifs_infos);
1351 spin_unlock(&ubifs_infos_lock);
1352
1353 if (c->need_recovery) {
2ef13294 1354 if (c->ro_mount)
1e51764a
AB
1355 ubifs_msg("recovery deferred");
1356 else {
1357 c->need_recovery = 0;
1358 ubifs_msg("recovery completed");
b221337a
AB
1359 /*
1360 * GC LEB has to be empty and taken at this point. But
1361 * the journal head LEBs may also be accounted as
1362 * "empty taken" if they are empty.
1363 */
1364 ubifs_assert(c->lst.taken_empty_lebs > 0);
1e51764a 1365 }
6ba87c9b 1366 } else
b221337a 1367 ubifs_assert(c->lst.taken_empty_lebs > 0);
1e51764a 1368
6ba87c9b 1369 err = dbg_check_filesystem(c);
552ff317
AB
1370 if (err)
1371 goto out_infos;
1372
6ba87c9b 1373 err = dbg_debugfs_init_fs(c);
1e51764a
AB
1374 if (err)
1375 goto out_infos;
1376
2953e73f
AH
1377 c->always_chk_crc = 0;
1378
ce769caa
AB
1379 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1380 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
2ef13294 1381 if (c->ro_mount)
1e51764a
AB
1382 ubifs_msg("mounted read-only");
1383 x = (long long)c->main_lebs * c->leb_size;
948cfb21
AB
1384 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1385 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1e51764a 1386 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
948cfb21
AB
1387 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1388 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
963f0cf6
AB
1389 ubifs_msg("media format: w%d/r%d (latest is w%d/r%d)",
1390 c->fmt_version, c->ro_compat_version,
1391 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
948cfb21 1392 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
fae7fb29 1393 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
948cfb21 1394 c->report_rp_size, c->report_rp_size >> 10);
1e51764a
AB
1395
1396 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1397 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1398 dbg_msg("LEB size: %d bytes (%d KiB)",
948cfb21 1399 c->leb_size, c->leb_size >> 10);
1e51764a
AB
1400 dbg_msg("data journal heads: %d",
1401 c->jhead_cnt - NONDATA_JHEADS_CNT);
7f2f4e72 1402 dbg_msg("UUID: %pUB", c->uuid);
1e51764a
AB
1403 dbg_msg("big_lpt %d", c->big_lpt);
1404 dbg_msg("log LEBs: %d (%d - %d)",
1405 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1406 dbg_msg("LPT area LEBs: %d (%d - %d)",
1407 c->lpt_lebs, c->lpt_first, c->lpt_last);
1408 dbg_msg("orphan area LEBs: %d (%d - %d)",
1409 c->orph_lebs, c->orph_first, c->orph_last);
1410 dbg_msg("main area LEBs: %d (%d - %d)",
1411 c->main_lebs, c->main_first, c->leb_cnt - 1);
1412 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1413 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1414 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1415 dbg_msg("key hash type: %d", c->key_hash_type);
1416 dbg_msg("tree fanout: %d", c->fanout);
1417 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1418 dbg_msg("first main LEB: %d", c->main_first);
8e5033ad
AB
1419 dbg_msg("max. znode size %d", c->max_znode_sz);
1420 dbg_msg("max. index node size %d", c->max_idx_node_sz);
1421 dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
1422 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1423 dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
1424 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1425 dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
1426 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1427 dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu",
1428 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1429 UBIFS_MAX_DENT_NODE_SZ);
1e51764a
AB
1430 dbg_msg("dead watermark: %d", c->dead_wm);
1431 dbg_msg("dark watermark: %d", c->dark_wm);
8e5033ad 1432 dbg_msg("LEB overhead: %d", c->leb_overhead);
1e51764a
AB
1433 x = (long long)c->main_lebs * c->dark_wm;
1434 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1435 x, x >> 10, x >> 20);
1436 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1437 c->max_bud_bytes, c->max_bud_bytes >> 10,
1438 c->max_bud_bytes >> 20);
1439 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1440 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1441 c->bg_bud_bytes >> 20);
1442 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1443 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1444 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1445 dbg_msg("commit number: %llu", c->cmt_no);
1446
1447 return 0;
1448
1449out_infos:
1450 spin_lock(&ubifs_infos_lock);
1451 list_del(&c->infos_list);
1452 spin_unlock(&ubifs_infos_lock);
1453out_orphans:
1454 free_orphans(c);
1455out_journal:
1456 destroy_journal(c);
1457out_lpt:
1458 ubifs_lpt_free(c, 0);
1459out_master:
1460 kfree(c->mst_node);
1461 kfree(c->rcvrd_mst_node);
1462 if (c->bgt)
1463 kthread_stop(c->bgt);
1464out_wbufs:
1465 free_wbufs(c);
1466out_cbuf:
1467 kfree(c->cbuf);
1e51764a 1468out_free:
3477d204 1469 kfree(c->bu.buf);
1e51764a
AB
1470 vfree(c->ileb_buf);
1471 vfree(c->sbuf);
1472 kfree(c->bottom_up_buf);
17c2f9f8 1473 ubifs_debugging_exit(c);
1e51764a
AB
1474 return err;
1475}
1476
1477/**
1478 * ubifs_umount - un-mount UBIFS file-system.
1479 * @c: UBIFS file-system description object
1480 *
1481 * Note, this function is called to free allocated resourced when un-mounting,
1482 * as well as free resources when an error occurred while we were half way
1483 * through mounting (error path cleanup function). So it has to make sure the
1484 * resource was actually allocated before freeing it.
1485 */
1486static void ubifs_umount(struct ubifs_info *c)
1487{
1488 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1489 c->vi.vol_id);
1490
552ff317 1491 dbg_debugfs_exit_fs(c);
1e51764a
AB
1492 spin_lock(&ubifs_infos_lock);
1493 list_del(&c->infos_list);
1494 spin_unlock(&ubifs_infos_lock);
1495
1496 if (c->bgt)
1497 kthread_stop(c->bgt);
1498
1499 destroy_journal(c);
1500 free_wbufs(c);
1501 free_orphans(c);
1502 ubifs_lpt_free(c, 0);
1503
1504 kfree(c->cbuf);
1505 kfree(c->rcvrd_mst_node);
1506 kfree(c->mst_node);
3477d204
AB
1507 kfree(c->bu.buf);
1508 vfree(c->ileb_buf);
1e51764a
AB
1509 vfree(c->sbuf);
1510 kfree(c->bottom_up_buf);
17c2f9f8 1511 ubifs_debugging_exit(c);
1e51764a
AB
1512}
1513
1514/**
1515 * ubifs_remount_rw - re-mount in read-write mode.
1516 * @c: UBIFS file-system description object
1517 *
1518 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1519 * mode. This function allocates the needed resources and re-mounts UBIFS in
1520 * read-write mode.
1521 */
1522static int ubifs_remount_rw(struct ubifs_info *c)
1523{
1524 int err, lnum;
1525
963f0cf6
AB
1526 if (c->rw_incompat) {
1527 ubifs_err("the file-system is not R/W-compatible");
1528 ubifs_msg("on-flash format version is w%d/r%d, but software "
1529 "only supports up to version w%d/r%d", c->fmt_version,
1530 c->ro_compat_version, UBIFS_FORMAT_VERSION,
1531 UBIFS_RO_COMPAT_VERSION);
1532 return -EROFS;
1533 }
1534
1e51764a 1535 mutex_lock(&c->umount_mutex);
84abf972 1536 dbg_save_space_info(c);
1e51764a 1537 c->remounting_rw = 1;
2953e73f 1538 c->always_chk_crc = 1;
1e51764a 1539
57a450e9
AB
1540 err = check_free_space(c);
1541 if (err)
1e51764a 1542 goto out;
1e51764a
AB
1543
1544 if (c->old_leb_cnt != c->leb_cnt) {
1545 struct ubifs_sb_node *sup;
1546
1547 sup = ubifs_read_sb_node(c);
1548 if (IS_ERR(sup)) {
1549 err = PTR_ERR(sup);
1550 goto out;
1551 }
1552 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1553 err = ubifs_write_sb_node(c, sup);
1554 if (err)
1555 goto out;
1556 }
1557
1558 if (c->need_recovery) {
1559 ubifs_msg("completing deferred recovery");
1560 err = ubifs_write_rcvrd_mst_node(c);
1561 if (err)
1562 goto out;
1563 err = ubifs_recover_size(c);
1564 if (err)
1565 goto out;
1566 err = ubifs_clean_lebs(c, c->sbuf);
1567 if (err)
1568 goto out;
1569 err = ubifs_recover_inl_heads(c, c->sbuf);
1570 if (err)
1571 goto out;
49d128aa
AH
1572 } else {
1573 /* A readonly mount is not allowed to have orphans */
1574 ubifs_assert(c->tot_orphans == 0);
1575 err = ubifs_clear_orphans(c);
1576 if (err)
1577 goto out;
1e51764a
AB
1578 }
1579
1580 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1581 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1582 err = ubifs_write_master(c);
1583 if (err)
1584 goto out;
1585 }
1586
1587 c->ileb_buf = vmalloc(c->leb_size);
1588 if (!c->ileb_buf) {
1589 err = -ENOMEM;
1590 goto out;
1591 }
1592
1593 err = ubifs_lpt_init(c, 0, 1);
1594 if (err)
1595 goto out;
1596
1597 err = alloc_wbufs(c);
1598 if (err)
1599 goto out;
1600
1601 ubifs_create_buds_lists(c);
1602
1603 /* Create background thread */
fcabb347 1604 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1e51764a
AB
1605 if (IS_ERR(c->bgt)) {
1606 err = PTR_ERR(c->bgt);
1607 c->bgt = NULL;
1608 ubifs_err("cannot spawn \"%s\", error %d",
1609 c->bgt_name, err);
2953e73f 1610 goto out;
1e51764a
AB
1611 }
1612 wake_up_process(c->bgt);
1613
1614 c->orph_buf = vmalloc(c->leb_size);
2953e73f
AH
1615 if (!c->orph_buf) {
1616 err = -ENOMEM;
1617 goto out;
1618 }
1e51764a
AB
1619
1620 /* Check for enough log space */
1621 lnum = c->lhead_lnum + 1;
1622 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1623 lnum = UBIFS_LOG_LNUM;
1624 if (lnum == c->ltail_lnum) {
1625 err = ubifs_consolidate_log(c);
1626 if (err)
1627 goto out;
1628 }
1629
1630 if (c->need_recovery)
1631 err = ubifs_rcvry_gc_commit(c);
1632 else
b4978e94 1633 err = ubifs_leb_unmap(c, c->gc_lnum);
1e51764a
AB
1634 if (err)
1635 goto out;
1636
1637 if (c->need_recovery) {
1638 c->need_recovery = 0;
1639 ubifs_msg("deferred recovery completed");
1640 }
1641
1642 dbg_gen("re-mounted read-write");
2ef13294 1643 c->ro_mount = 0;
1e51764a 1644 c->remounting_rw = 0;
2953e73f 1645 c->always_chk_crc = 0;
84abf972 1646 err = dbg_check_space_info(c);
1e51764a 1647 mutex_unlock(&c->umount_mutex);
84abf972 1648 return err;
1e51764a
AB
1649
1650out:
1651 vfree(c->orph_buf);
1652 c->orph_buf = NULL;
1653 if (c->bgt) {
1654 kthread_stop(c->bgt);
1655 c->bgt = NULL;
1656 }
1657 free_wbufs(c);
1658 vfree(c->ileb_buf);
1659 c->ileb_buf = NULL;
1660 ubifs_lpt_free(c, 1);
1661 c->remounting_rw = 0;
2953e73f 1662 c->always_chk_crc = 0;
1e51764a
AB
1663 mutex_unlock(&c->umount_mutex);
1664 return err;
1665}
1666
1e51764a
AB
1667/**
1668 * ubifs_remount_ro - re-mount in read-only mode.
1669 * @c: UBIFS file-system description object
1670 *
84abf972
AB
1671 * We assume VFS has stopped writing. Possibly the background thread could be
1672 * running a commit, however kthread_stop will wait in that case.
1e51764a
AB
1673 */
1674static void ubifs_remount_ro(struct ubifs_info *c)
1675{
1676 int i, err;
1677
1678 ubifs_assert(!c->need_recovery);
2ef13294 1679 ubifs_assert(!c->ro_mount);
e4d9b6cb 1680
1e51764a
AB
1681 mutex_lock(&c->umount_mutex);
1682 if (c->bgt) {
1683 kthread_stop(c->bgt);
1684 c->bgt = NULL;
1685 }
1686
84abf972
AB
1687 dbg_save_space_info(c);
1688
39037559 1689 for (i = 0; i < c->jhead_cnt; i++)
1e51764a 1690 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1e51764a 1691
e4d9b6cb
AB
1692 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1693 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1694 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1695 err = ubifs_write_master(c);
1696 if (err)
1697 ubifs_ro_mode(c, err);
1698
1e51764a
AB
1699 free_wbufs(c);
1700 vfree(c->orph_buf);
1701 c->orph_buf = NULL;
1702 vfree(c->ileb_buf);
1703 c->ileb_buf = NULL;
1704 ubifs_lpt_free(c, 1);
2ef13294 1705 c->ro_mount = 1;
84abf972
AB
1706 err = dbg_check_space_info(c);
1707 if (err)
1708 ubifs_ro_mode(c, err);
1e51764a
AB
1709 mutex_unlock(&c->umount_mutex);
1710}
1711
1712static void ubifs_put_super(struct super_block *sb)
1713{
1714 int i;
1715 struct ubifs_info *c = sb->s_fs_info;
1716
1717 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1718 c->vi.vol_id);
6cfd0148 1719
1e51764a
AB
1720 /*
1721 * The following asserts are only valid if there has not been a failure
1722 * of the media. For example, there will be dirty inodes if we failed
1723 * to write them back because of I/O errors.
1724 */
1725 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1726 ubifs_assert(c->budg_idx_growth == 0);
7d32c2bb 1727 ubifs_assert(c->budg_dd_growth == 0);
1e51764a
AB
1728 ubifs_assert(c->budg_data_growth == 0);
1729
1730 /*
1731 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1732 * and file system un-mount. Namely, it prevents the shrinker from
1733 * picking this superblock for shrinking - it will be just skipped if
1734 * the mutex is locked.
1735 */
1736 mutex_lock(&c->umount_mutex);
2ef13294 1737 if (!c->ro_mount) {
1e51764a
AB
1738 /*
1739 * First of all kill the background thread to make sure it does
1740 * not interfere with un-mounting and freeing resources.
1741 */
1742 if (c->bgt) {
1743 kthread_stop(c->bgt);
1744 c->bgt = NULL;
1745 }
1746
1e51764a 1747 /*
2680d722 1748 * On fatal errors c->ro_error is set to 1, in which case we do
1e51764a
AB
1749 * not write the master node.
1750 */
2680d722 1751 if (!c->ro_error) {
2ef13294
AB
1752 int err;
1753
1754 /* Synchronize write-buffers */
39037559
AB
1755 for (i = 0; i < c->jhead_cnt; i++)
1756 ubifs_wbuf_sync(&c->jheads[i].wbuf);
2ef13294 1757
1e51764a
AB
1758 /*
1759 * We are being cleanly unmounted which means the
1760 * orphans were killed - indicate this in the master
1761 * node. Also save the reserved GC LEB number.
1762 */
1e51764a
AB
1763 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1764 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1765 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1766 err = ubifs_write_master(c);
1767 if (err)
1768 /*
1769 * Recovery will attempt to fix the master area
1770 * next mount, so we just print a message and
1771 * continue to unmount normally.
1772 */
1773 ubifs_err("failed to write master node, "
1774 "error %d", err);
3601ba27
AB
1775 } else {
1776 for (i = 0; i < c->jhead_cnt; i++)
1777 /* Make sure write-buffer timers are canceled */
1778 hrtimer_cancel(&c->jheads[i].wbuf.timer);
1e51764a
AB
1779 }
1780 }
1781
1782 ubifs_umount(c);
1783 bdi_destroy(&c->bdi);
1784 ubi_close_volume(c->ubi);
1785 mutex_unlock(&c->umount_mutex);
1786 kfree(c);
1787}
1788
1789static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1790{
1791 int err;
1792 struct ubifs_info *c = sb->s_fs_info;
1793
1794 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1795
1796 err = ubifs_parse_options(c, data, 1);
1797 if (err) {
1798 ubifs_err("invalid or unknown remount parameter");
1799 return err;
1800 }
3477d204 1801
2ef13294 1802 if (c->ro_mount && !(*flags & MS_RDONLY)) {
2680d722
AB
1803 if (c->ro_error) {
1804 ubifs_msg("cannot re-mount R/W due to prior errors");
1805 return -EROFS;
1806 }
e4d9b6cb 1807 if (c->ro_media) {
2680d722 1808 ubifs_msg("cannot re-mount R/W - UBI volume is R/O");
a2b9df3f 1809 return -EROFS;
e4d9b6cb 1810 }
1e51764a 1811 err = ubifs_remount_rw(c);
e9d6bbc4 1812 if (err)
1e51764a 1813 return err;
2ef13294 1814 } else if (!c->ro_mount && (*flags & MS_RDONLY)) {
2680d722
AB
1815 if (c->ro_error) {
1816 ubifs_msg("cannot re-mount R/O due to prior errors");
a2b9df3f 1817 return -EROFS;
b466f17d 1818 }
1e51764a 1819 ubifs_remount_ro(c);
b466f17d 1820 }
1e51764a 1821
3477d204
AB
1822 if (c->bulk_read == 1)
1823 bu_init(c);
1824 else {
1825 dbg_gen("disable bulk-read");
1826 kfree(c->bu.buf);
1827 c->bu.buf = NULL;
1828 }
1829
b221337a 1830 ubifs_assert(c->lst.taken_empty_lebs > 0);
1e51764a
AB
1831 return 0;
1832}
1833
e8b81566 1834const struct super_operations ubifs_super_operations = {
1e51764a
AB
1835 .alloc_inode = ubifs_alloc_inode,
1836 .destroy_inode = ubifs_destroy_inode,
1837 .put_super = ubifs_put_super,
1838 .write_inode = ubifs_write_inode,
d640e1b5 1839 .evict_inode = ubifs_evict_inode,
1e51764a
AB
1840 .statfs = ubifs_statfs,
1841 .dirty_inode = ubifs_dirty_inode,
1842 .remount_fs = ubifs_remount_fs,
1843 .show_options = ubifs_show_options,
1844 .sync_fs = ubifs_sync_fs,
1845};
1846
1847/**
1848 * open_ubi - parse UBI device name string and open the UBI device.
1849 * @name: UBI volume name
1850 * @mode: UBI volume open mode
1851 *
9722324e
CC
1852 * The primary method of mounting UBIFS is by specifying the UBI volume
1853 * character device node path. However, UBIFS may also be mounted withoug any
1854 * character device node using one of the following methods:
1855 *
1856 * o ubiX_Y - mount UBI device number X, volume Y;
1857 * o ubiY - mount UBI device number 0, volume Y;
1e51764a
AB
1858 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1859 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1860 *
1861 * Alternative '!' separator may be used instead of ':' (because some shells
1862 * like busybox may interpret ':' as an NFS host name separator). This function
9722324e
CC
1863 * returns UBI volume description object in case of success and a negative
1864 * error code in case of failure.
1e51764a
AB
1865 */
1866static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1867{
9722324e 1868 struct ubi_volume_desc *ubi;
1e51764a
AB
1869 int dev, vol;
1870 char *endptr;
1871
9722324e
CC
1872 /* First, try to open using the device node path method */
1873 ubi = ubi_open_volume_path(name, mode);
1874 if (!IS_ERR(ubi))
1875 return ubi;
1876
1877 /* Try the "nodev" method */
1e51764a
AB
1878 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1879 return ERR_PTR(-EINVAL);
1880
1881 /* ubi:NAME method */
1882 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1883 return ubi_open_volume_nm(0, name + 4, mode);
1884
1885 if (!isdigit(name[3]))
1886 return ERR_PTR(-EINVAL);
1887
1888 dev = simple_strtoul(name + 3, &endptr, 0);
1889
1890 /* ubiY method */
1891 if (*endptr == '\0')
1892 return ubi_open_volume(0, dev, mode);
1893
1894 /* ubiX_Y method */
1895 if (*endptr == '_' && isdigit(endptr[1])) {
1896 vol = simple_strtoul(endptr + 1, &endptr, 0);
1897 if (*endptr != '\0')
1898 return ERR_PTR(-EINVAL);
1899 return ubi_open_volume(dev, vol, mode);
1900 }
1901
1902 /* ubiX:NAME method */
1903 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1904 return ubi_open_volume_nm(dev, ++endptr, mode);
1905
1906 return ERR_PTR(-EINVAL);
1907}
1908
1909static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1910{
1911 struct ubi_volume_desc *ubi = sb->s_fs_info;
1912 struct ubifs_info *c;
1913 struct inode *root;
1914 int err;
1915
1916 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1917 if (!c)
1918 return -ENOMEM;
1919
1920 spin_lock_init(&c->cnt_lock);
1921 spin_lock_init(&c->cs_lock);
1922 spin_lock_init(&c->buds_lock);
1923 spin_lock_init(&c->space_lock);
1924 spin_lock_init(&c->orphan_lock);
1925 init_rwsem(&c->commit_sem);
1926 mutex_init(&c->lp_mutex);
1927 mutex_init(&c->tnc_mutex);
1928 mutex_init(&c->log_mutex);
1929 mutex_init(&c->mst_mutex);
1930 mutex_init(&c->umount_mutex);
3477d204 1931 mutex_init(&c->bu_mutex);
1e51764a
AB
1932 init_waitqueue_head(&c->cmt_wq);
1933 c->buds = RB_ROOT;
1934 c->old_idx = RB_ROOT;
1935 c->size_tree = RB_ROOT;
1936 c->orph_tree = RB_ROOT;
1937 INIT_LIST_HEAD(&c->infos_list);
1938 INIT_LIST_HEAD(&c->idx_gc);
1939 INIT_LIST_HEAD(&c->replay_list);
1940 INIT_LIST_HEAD(&c->replay_buds);
1941 INIT_LIST_HEAD(&c->uncat_list);
1942 INIT_LIST_HEAD(&c->empty_list);
1943 INIT_LIST_HEAD(&c->freeable_list);
1944 INIT_LIST_HEAD(&c->frdi_idx_list);
1945 INIT_LIST_HEAD(&c->unclean_leb_list);
1946 INIT_LIST_HEAD(&c->old_buds);
1947 INIT_LIST_HEAD(&c->orph_list);
1948 INIT_LIST_HEAD(&c->orph_new);
1949
8379ea31 1950 c->vfs_sb = sb;
1e51764a 1951 c->highest_inum = UBIFS_FIRST_INO;
1e51764a
AB
1952 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1953
1954 ubi_get_volume_info(ubi, &c->vi);
1955 ubi_get_device_info(c->vi.ubi_num, &c->di);
1956
1957 /* Re-open the UBI device in read-write mode */
1958 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1959 if (IS_ERR(c->ubi)) {
1960 err = PTR_ERR(c->ubi);
1961 goto out_free;
1962 }
1963
1964 /*
0a883a05 1965 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1e51764a
AB
1966 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1967 * which means the user would have to wait not just for their own I/O
0a883a05 1968 * but the read-ahead I/O as well i.e. completely pointless.
1e51764a
AB
1969 *
1970 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1971 */
d993831f 1972 c->bdi.name = "ubifs",
1e51764a
AB
1973 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1974 c->bdi.unplug_io_fn = default_unplug_io_fn;
1975 err = bdi_init(&c->bdi);
1976 if (err)
1977 goto out_close;
7fcd9c3e
DM
1978 err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
1979 c->vi.ubi_num, c->vi.vol_id);
a979eff1
JA
1980 if (err)
1981 goto out_bdi;
1e51764a
AB
1982
1983 err = ubifs_parse_options(c, data, 0);
1984 if (err)
1985 goto out_bdi;
1986
32a88aa1 1987 sb->s_bdi = &c->bdi;
1e51764a
AB
1988 sb->s_fs_info = c;
1989 sb->s_magic = UBIFS_SUPER_MAGIC;
1990 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1991 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1e51764a
AB
1992 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1993 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1994 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1995 sb->s_op = &ubifs_super_operations;
1996
1997 mutex_lock(&c->umount_mutex);
1998 err = mount_ubifs(c);
1999 if (err) {
2000 ubifs_assert(err < 0);
2001 goto out_unlock;
2002 }
2003
2004 /* Read the root inode */
2005 root = ubifs_iget(sb, UBIFS_ROOT_INO);
2006 if (IS_ERR(root)) {
2007 err = PTR_ERR(root);
2008 goto out_umount;
2009 }
2010
2011 sb->s_root = d_alloc_root(root);
2012 if (!sb->s_root)
2013 goto out_iput;
2014
2015 mutex_unlock(&c->umount_mutex);
1e51764a
AB
2016 return 0;
2017
2018out_iput:
2019 iput(root);
2020out_umount:
2021 ubifs_umount(c);
2022out_unlock:
2023 mutex_unlock(&c->umount_mutex);
2024out_bdi:
2025 bdi_destroy(&c->bdi);
2026out_close:
2027 ubi_close_volume(c->ubi);
2028out_free:
2029 kfree(c);
2030 return err;
2031}
2032
2033static int sb_test(struct super_block *sb, void *data)
2034{
2035 dev_t *dev = data;
7c83f5cb 2036 struct ubifs_info *c = sb->s_fs_info;
1e51764a 2037
7c83f5cb 2038 return c->vi.cdev == *dev;
1e51764a
AB
2039}
2040
157d81e7
AV
2041static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2042 const char *name, void *data)
1e51764a
AB
2043{
2044 struct ubi_volume_desc *ubi;
2045 struct ubi_volume_info vi;
2046 struct super_block *sb;
2047 int err;
2048
2049 dbg_gen("name %s, flags %#x", name, flags);
2050
2051 /*
2052 * Get UBI device number and volume ID. Mount it read-only so far
2053 * because this might be a new mount point, and UBI allows only one
2054 * read-write user at a time.
2055 */
2056 ubi = open_ubi(name, UBI_READONLY);
2057 if (IS_ERR(ubi)) {
54dd55a4
SS
2058 dbg_err("cannot open \"%s\", error %d",
2059 name, (int)PTR_ERR(ubi));
157d81e7 2060 return ERR_CAST(ubi);
1e51764a
AB
2061 }
2062 ubi_get_volume_info(ubi, &vi);
2063
2064 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
2065
7c83f5cb 2066 sb = sget(fs_type, &sb_test, &set_anon_super, &vi.cdev);
1e51764a
AB
2067 if (IS_ERR(sb)) {
2068 err = PTR_ERR(sb);
2069 goto out_close;
2070 }
2071
2072 if (sb->s_root) {
2ef13294
AB
2073 struct ubifs_info *c1 = sb->s_fs_info;
2074
1e51764a
AB
2075 /* A new mount point for already mounted UBIFS */
2076 dbg_gen("this ubi volume is already mounted");
2ef13294 2077 if (!!(flags & MS_RDONLY) != c1->ro_mount) {
1e51764a
AB
2078 err = -EBUSY;
2079 goto out_deact;
2080 }
2081 } else {
2082 sb->s_flags = flags;
2083 /*
2084 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
2085 * replaced by 'c'.
2086 */
2087 sb->s_fs_info = ubi;
2088 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2089 if (err)
2090 goto out_deact;
2091 /* We do not support atime */
2092 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2093 }
2094
2095 /* 'fill_super()' opens ubi again so we must close it here */
2096 ubi_close_volume(ubi);
2097
157d81e7 2098 return dget(sb->s_root);
1e51764a
AB
2099
2100out_deact:
6f5bbff9 2101 deactivate_locked_super(sb);
1e51764a
AB
2102out_close:
2103 ubi_close_volume(ubi);
157d81e7 2104 return ERR_PTR(err);
1e51764a
AB
2105}
2106
1e51764a
AB
2107static struct file_system_type ubifs_fs_type = {
2108 .name = "ubifs",
2109 .owner = THIS_MODULE,
157d81e7 2110 .mount = ubifs_mount,
7c83f5cb 2111 .kill_sb = kill_anon_super,
1e51764a
AB
2112};
2113
2114/*
2115 * Inode slab cache constructor.
2116 */
51cc5068 2117static void inode_slab_ctor(void *obj)
1e51764a
AB
2118{
2119 struct ubifs_inode *ui = obj;
2120 inode_init_once(&ui->vfs_inode);
2121}
2122
2123static int __init ubifs_init(void)
2124{
2125 int err;
2126
2127 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2128
2129 /* Make sure node sizes are 8-byte aligned */
2130 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2131 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2132 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2133 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2134 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2135 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2136 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2137 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2138 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2139 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2140 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2141
2142 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2143 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2144 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2145 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2146 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2147 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2148
2149 /* Check min. node size */
2150 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2151 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2152 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2153 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2154
2155 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2156 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2157 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2158 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2159
2160 /* Defined node sizes */
2161 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2162 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2163 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2164 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2165
a1dc080c
AB
2166 /*
2167 * We use 2 bit wide bit-fields to store compression type, which should
2168 * be amended if more compressors are added. The bit-fields are:
553dea4d
AB
2169 * @compr_type in 'struct ubifs_inode', @default_compr in
2170 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
a1dc080c
AB
2171 */
2172 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2173
1e51764a
AB
2174 /*
2175 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2176 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2177 */
2178 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2179 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2180 " at least 4096 bytes",
2181 (unsigned int)PAGE_CACHE_SIZE);
2182 return -EINVAL;
2183 }
2184
2185 err = register_filesystem(&ubifs_fs_type);
2186 if (err) {
2187 ubifs_err("cannot register file system, error %d", err);
2188 return err;
2189 }
2190
2191 err = -ENOMEM;
2192 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2193 sizeof(struct ubifs_inode), 0,
2194 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2195 &inode_slab_ctor);
2196 if (!ubifs_inode_slab)
2197 goto out_reg;
2198
2199 register_shrinker(&ubifs_shrinker_info);
2200
2201 err = ubifs_compressors_init();
552ff317
AB
2202 if (err)
2203 goto out_shrinker;
2204
2205 err = dbg_debugfs_init();
1e51764a
AB
2206 if (err)
2207 goto out_compr;
2208
2209 return 0;
2210
2211out_compr:
552ff317
AB
2212 ubifs_compressors_exit();
2213out_shrinker:
1e51764a
AB
2214 unregister_shrinker(&ubifs_shrinker_info);
2215 kmem_cache_destroy(ubifs_inode_slab);
2216out_reg:
2217 unregister_filesystem(&ubifs_fs_type);
2218 return err;
2219}
2220/* late_initcall to let compressors initialize first */
2221late_initcall(ubifs_init);
2222
2223static void __exit ubifs_exit(void)
2224{
2225 ubifs_assert(list_empty(&ubifs_infos));
2226 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2227
552ff317 2228 dbg_debugfs_exit();
1e51764a
AB
2229 ubifs_compressors_exit();
2230 unregister_shrinker(&ubifs_shrinker_info);
2231 kmem_cache_destroy(ubifs_inode_slab);
2232 unregister_filesystem(&ubifs_fs_type);
2233}
2234module_exit(ubifs_exit);
2235
2236MODULE_LICENSE("GPL");
2237MODULE_VERSION(__stringify(UBIFS_VERSION));
2238MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2239MODULE_DESCRIPTION("UBIFS - UBI File System");