UBIFS: always commit in sync_fs
[linux-2.6-block.git] / fs / ubifs / super.c
CommitLineData
1e51764a
AB
1/*
2 * This file is part of UBIFS.
3 *
4 * Copyright (C) 2006-2008 Nokia Corporation.
5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License version 2 as published by
8 * the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * more details.
14 *
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc., 51
17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18 *
19 * Authors: Artem Bityutskiy (Битюцкий Артём)
20 * Adrian Hunter
21 */
22
23/*
24 * This file implements UBIFS initialization and VFS superblock operations. Some
25 * initialization stuff which is rather large and complex is placed at
26 * corresponding subsystems, but most of it is here.
27 */
28
29#include <linux/init.h>
30#include <linux/slab.h>
31#include <linux/module.h>
32#include <linux/ctype.h>
1e51764a
AB
33#include <linux/kthread.h>
34#include <linux/parser.h>
35#include <linux/seq_file.h>
36#include <linux/mount.h>
4d61db4f 37#include <linux/math64.h>
304d427c 38#include <linux/writeback.h>
1e51764a
AB
39#include "ubifs.h"
40
39ce81ce
AB
41/*
42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
43 * allocating too much.
44 */
45#define UBIFS_KMALLOC_OK (128*1024)
46
1e51764a
AB
47/* Slab cache for UBIFS inodes */
48struct kmem_cache *ubifs_inode_slab;
49
50/* UBIFS TNC shrinker description */
51static struct shrinker ubifs_shrinker_info = {
52 .shrink = ubifs_shrinker,
53 .seeks = DEFAULT_SEEKS,
54};
55
56/**
57 * validate_inode - validate inode.
58 * @c: UBIFS file-system description object
59 * @inode: the inode to validate
60 *
61 * This is a helper function for 'ubifs_iget()' which validates various fields
62 * of a newly built inode to make sure they contain sane values and prevent
63 * possible vulnerabilities. Returns zero if the inode is all right and
64 * a non-zero error code if not.
65 */
66static int validate_inode(struct ubifs_info *c, const struct inode *inode)
67{
68 int err;
69 const struct ubifs_inode *ui = ubifs_inode(inode);
70
71 if (inode->i_size > c->max_inode_sz) {
72 ubifs_err("inode is too large (%lld)",
73 (long long)inode->i_size);
74 return 1;
75 }
76
77 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
78 ubifs_err("unknown compression type %d", ui->compr_type);
79 return 2;
80 }
81
82 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
83 return 3;
84
85 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
86 return 4;
87
88 if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
89 return 5;
90
91 if (!ubifs_compr_present(ui->compr_type)) {
92 ubifs_warn("inode %lu uses '%s' compression, but it was not "
93 "compiled in", inode->i_ino,
94 ubifs_compr_name(ui->compr_type));
95 }
96
97 err = dbg_check_dir_size(c, inode);
98 return err;
99}
100
101struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
102{
103 int err;
104 union ubifs_key key;
105 struct ubifs_ino_node *ino;
106 struct ubifs_info *c = sb->s_fs_info;
107 struct inode *inode;
108 struct ubifs_inode *ui;
109
110 dbg_gen("inode %lu", inum);
111
112 inode = iget_locked(sb, inum);
113 if (!inode)
114 return ERR_PTR(-ENOMEM);
115 if (!(inode->i_state & I_NEW))
116 return inode;
117 ui = ubifs_inode(inode);
118
119 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
120 if (!ino) {
121 err = -ENOMEM;
122 goto out;
123 }
124
125 ino_key_init(c, &key, inode->i_ino);
126
127 err = ubifs_tnc_lookup(c, &key, ino);
128 if (err)
129 goto out_ino;
130
131 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
132 inode->i_nlink = le32_to_cpu(ino->nlink);
133 inode->i_uid = le32_to_cpu(ino->uid);
134 inode->i_gid = le32_to_cpu(ino->gid);
135 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
136 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
137 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
138 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
139 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
140 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
141 inode->i_mode = le32_to_cpu(ino->mode);
142 inode->i_size = le64_to_cpu(ino->size);
143
144 ui->data_len = le32_to_cpu(ino->data_len);
145 ui->flags = le32_to_cpu(ino->flags);
146 ui->compr_type = le16_to_cpu(ino->compr_type);
147 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
148 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
149 ui->xattr_size = le32_to_cpu(ino->xattr_size);
150 ui->xattr_names = le32_to_cpu(ino->xattr_names);
151 ui->synced_i_size = ui->ui_size = inode->i_size;
152
153 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
154
155 err = validate_inode(c, inode);
156 if (err)
157 goto out_invalid;
158
0a883a05 159 /* Disable read-ahead */
1e51764a
AB
160 inode->i_mapping->backing_dev_info = &c->bdi;
161
162 switch (inode->i_mode & S_IFMT) {
163 case S_IFREG:
164 inode->i_mapping->a_ops = &ubifs_file_address_operations;
165 inode->i_op = &ubifs_file_inode_operations;
166 inode->i_fop = &ubifs_file_operations;
167 if (ui->xattr) {
168 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
169 if (!ui->data) {
170 err = -ENOMEM;
171 goto out_ino;
172 }
173 memcpy(ui->data, ino->data, ui->data_len);
174 ((char *)ui->data)[ui->data_len] = '\0';
175 } else if (ui->data_len != 0) {
176 err = 10;
177 goto out_invalid;
178 }
179 break;
180 case S_IFDIR:
181 inode->i_op = &ubifs_dir_inode_operations;
182 inode->i_fop = &ubifs_dir_operations;
183 if (ui->data_len != 0) {
184 err = 11;
185 goto out_invalid;
186 }
187 break;
188 case S_IFLNK:
189 inode->i_op = &ubifs_symlink_inode_operations;
190 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
191 err = 12;
192 goto out_invalid;
193 }
194 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
195 if (!ui->data) {
196 err = -ENOMEM;
197 goto out_ino;
198 }
199 memcpy(ui->data, ino->data, ui->data_len);
200 ((char *)ui->data)[ui->data_len] = '\0';
201 break;
202 case S_IFBLK:
203 case S_IFCHR:
204 {
205 dev_t rdev;
206 union ubifs_dev_desc *dev;
207
208 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
209 if (!ui->data) {
210 err = -ENOMEM;
211 goto out_ino;
212 }
213
214 dev = (union ubifs_dev_desc *)ino->data;
215 if (ui->data_len == sizeof(dev->new))
216 rdev = new_decode_dev(le32_to_cpu(dev->new));
217 else if (ui->data_len == sizeof(dev->huge))
218 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
219 else {
220 err = 13;
221 goto out_invalid;
222 }
223 memcpy(ui->data, ino->data, ui->data_len);
224 inode->i_op = &ubifs_file_inode_operations;
225 init_special_inode(inode, inode->i_mode, rdev);
226 break;
227 }
228 case S_IFSOCK:
229 case S_IFIFO:
230 inode->i_op = &ubifs_file_inode_operations;
231 init_special_inode(inode, inode->i_mode, 0);
232 if (ui->data_len != 0) {
233 err = 14;
234 goto out_invalid;
235 }
236 break;
237 default:
238 err = 15;
239 goto out_invalid;
240 }
241
242 kfree(ino);
243 ubifs_set_inode_flags(inode);
244 unlock_new_inode(inode);
245 return inode;
246
247out_invalid:
248 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
249 dbg_dump_node(c, ino);
250 dbg_dump_inode(c, inode);
251 err = -EINVAL;
252out_ino:
253 kfree(ino);
254out:
255 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
256 iget_failed(inode);
257 return ERR_PTR(err);
258}
259
260static struct inode *ubifs_alloc_inode(struct super_block *sb)
261{
262 struct ubifs_inode *ui;
263
264 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
265 if (!ui)
266 return NULL;
267
268 memset((void *)ui + sizeof(struct inode), 0,
269 sizeof(struct ubifs_inode) - sizeof(struct inode));
270 mutex_init(&ui->ui_mutex);
271 spin_lock_init(&ui->ui_lock);
272 return &ui->vfs_inode;
273};
274
275static void ubifs_destroy_inode(struct inode *inode)
276{
277 struct ubifs_inode *ui = ubifs_inode(inode);
278
279 kfree(ui->data);
280 kmem_cache_free(ubifs_inode_slab, inode);
281}
282
283/*
284 * Note, Linux write-back code calls this without 'i_mutex'.
285 */
286static int ubifs_write_inode(struct inode *inode, int wait)
287{
fbfa6c88 288 int err = 0;
1e51764a
AB
289 struct ubifs_info *c = inode->i_sb->s_fs_info;
290 struct ubifs_inode *ui = ubifs_inode(inode);
291
292 ubifs_assert(!ui->xattr);
293 if (is_bad_inode(inode))
294 return 0;
295
296 mutex_lock(&ui->ui_mutex);
297 /*
298 * Due to races between write-back forced by budgeting
299 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
300 * have already been synchronized, do not do this again. This might
301 * also happen if it was synchronized in an VFS operation, e.g.
302 * 'ubifs_link()'.
303 */
304 if (!ui->dirty) {
305 mutex_unlock(&ui->ui_mutex);
306 return 0;
307 }
308
fbfa6c88
AB
309 /*
310 * As an optimization, do not write orphan inodes to the media just
311 * because this is not needed.
312 */
313 dbg_gen("inode %lu, mode %#x, nlink %u",
314 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
315 if (inode->i_nlink) {
1f28681a 316 err = ubifs_jnl_write_inode(c, inode);
fbfa6c88
AB
317 if (err)
318 ubifs_err("can't write inode %lu, error %d",
319 inode->i_ino, err);
320 }
1e51764a
AB
321
322 ui->dirty = 0;
323 mutex_unlock(&ui->ui_mutex);
324 ubifs_release_dirty_inode_budget(c, ui);
325 return err;
326}
327
328static void ubifs_delete_inode(struct inode *inode)
329{
330 int err;
331 struct ubifs_info *c = inode->i_sb->s_fs_info;
1e0f358e 332 struct ubifs_inode *ui = ubifs_inode(inode);
1e51764a 333
1e0f358e 334 if (ui->xattr)
1e51764a
AB
335 /*
336 * Extended attribute inode deletions are fully handled in
337 * 'ubifs_removexattr()'. These inodes are special and have
338 * limited usage, so there is nothing to do here.
339 */
340 goto out;
341
7d32c2bb 342 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
1e51764a
AB
343 ubifs_assert(!atomic_read(&inode->i_count));
344 ubifs_assert(inode->i_nlink == 0);
345
346 truncate_inode_pages(&inode->i_data, 0);
347 if (is_bad_inode(inode))
348 goto out;
349
1e0f358e 350 ui->ui_size = inode->i_size = 0;
de94eb55 351 err = ubifs_jnl_delete_inode(c, inode);
1e51764a
AB
352 if (err)
353 /*
354 * Worst case we have a lost orphan inode wasting space, so a
0a883a05 355 * simple error message is OK here.
1e51764a 356 */
de94eb55
AB
357 ubifs_err("can't delete inode %lu, error %d",
358 inode->i_ino, err);
359
1e51764a 360out:
1e0f358e
AB
361 if (ui->dirty)
362 ubifs_release_dirty_inode_budget(c, ui);
1e51764a
AB
363 clear_inode(inode);
364}
365
366static void ubifs_dirty_inode(struct inode *inode)
367{
368 struct ubifs_inode *ui = ubifs_inode(inode);
369
370 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
371 if (!ui->dirty) {
372 ui->dirty = 1;
373 dbg_gen("inode %lu", inode->i_ino);
374 }
375}
376
377static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
378{
379 struct ubifs_info *c = dentry->d_sb->s_fs_info;
380 unsigned long long free;
7c7cbadf 381 __le32 *uuid = (__le32 *)c->uuid;
1e51764a 382
7dad181b 383 free = ubifs_get_free_space(c);
1e51764a
AB
384 dbg_gen("free space %lld bytes (%lld blocks)",
385 free, free >> UBIFS_BLOCK_SHIFT);
386
387 buf->f_type = UBIFS_SUPER_MAGIC;
388 buf->f_bsize = UBIFS_BLOCK_SIZE;
389 buf->f_blocks = c->block_cnt;
390 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
391 if (free > c->report_rp_size)
392 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
393 else
394 buf->f_bavail = 0;
395 buf->f_files = 0;
396 buf->f_ffree = 0;
397 buf->f_namelen = UBIFS_MAX_NLEN;
7c7cbadf
AB
398 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
399 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
1e51764a
AB
400 return 0;
401}
402
403static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
404{
405 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
406
407 if (c->mount_opts.unmount_mode == 2)
408 seq_printf(s, ",fast_unmount");
409 else if (c->mount_opts.unmount_mode == 1)
410 seq_printf(s, ",norm_unmount");
411
4793e7c5
AH
412 if (c->mount_opts.bulk_read == 2)
413 seq_printf(s, ",bulk_read");
414 else if (c->mount_opts.bulk_read == 1)
415 seq_printf(s, ",no_bulk_read");
416
2953e73f
AH
417 if (c->mount_opts.chk_data_crc == 2)
418 seq_printf(s, ",chk_data_crc");
419 else if (c->mount_opts.chk_data_crc == 1)
420 seq_printf(s, ",no_chk_data_crc");
421
553dea4d
AB
422 if (c->mount_opts.override_compr) {
423 seq_printf(s, ",compr=");
424 seq_printf(s, ubifs_compr_name(c->mount_opts.compr_type));
425 }
426
1e51764a
AB
427 return 0;
428}
429
430static int ubifs_sync_fs(struct super_block *sb, int wait)
431{
f1038300 432 int i, err;
1e51764a 433 struct ubifs_info *c = sb->s_fs_info;
304d427c
AB
434 struct writeback_control wbc = {
435 .sync_mode = wait ? WB_SYNC_ALL : WB_SYNC_HOLD,
436 .range_start = 0,
437 .range_end = LLONG_MAX,
438 .nr_to_write = LONG_MAX,
439 };
440
f1038300
AB
441 if (sb->s_flags & MS_RDONLY)
442 return 0;
443
444 /*
445 * Synchronize write buffers, because 'ubifs_run_commit()' does not
446 * do this if it waits for an already running commit.
447 */
448 for (i = 0; i < c->jhead_cnt; i++) {
449 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
450 if (err)
451 return err;
452 }
453
304d427c
AB
454 /*
455 * VFS calls '->sync_fs()' before synchronizing all dirty inodes and
456 * pages, so synchronize them first, then commit the journal. Strictly
457 * speaking, it is not necessary to commit the journal here,
458 * synchronizing write-buffers would be enough. But committing makes
459 * UBIFS free space predictions much more accurate, so we want to let
460 * the user be able to get more accurate results of 'statfs()' after
461 * they synchronize the file system.
462 */
463 generic_sync_sb_inodes(sb, &wbc);
1e51764a 464
f1038300
AB
465 err = ubifs_run_commit(c);
466 if (err)
467 return err;
403e12ab 468
1e51764a
AB
469 /*
470 * We ought to call sync for c->ubi but it does not have one. If it had
471 * it would in turn call mtd->sync, however mtd operations are
472 * synchronous anyway, so we don't lose any sleep here.
473 */
f1038300 474 return err;
1e51764a
AB
475}
476
477/**
478 * init_constants_early - initialize UBIFS constants.
479 * @c: UBIFS file-system description object
480 *
481 * This function initialize UBIFS constants which do not need the superblock to
482 * be read. It also checks that the UBI volume satisfies basic UBIFS
483 * requirements. Returns zero in case of success and a negative error code in
484 * case of failure.
485 */
486static int init_constants_early(struct ubifs_info *c)
487{
488 if (c->vi.corrupted) {
489 ubifs_warn("UBI volume is corrupted - read-only mode");
490 c->ro_media = 1;
491 }
492
493 if (c->di.ro_mode) {
494 ubifs_msg("read-only UBI device");
495 c->ro_media = 1;
496 }
497
498 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
499 ubifs_msg("static UBI volume - read-only mode");
500 c->ro_media = 1;
501 }
502
503 c->leb_cnt = c->vi.size;
504 c->leb_size = c->vi.usable_leb_size;
505 c->half_leb_size = c->leb_size / 2;
506 c->min_io_size = c->di.min_io_size;
507 c->min_io_shift = fls(c->min_io_size) - 1;
508
509 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
510 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
511 c->leb_size, UBIFS_MIN_LEB_SZ);
512 return -EINVAL;
513 }
514
515 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
516 ubifs_err("too few LEBs (%d), min. is %d",
517 c->leb_cnt, UBIFS_MIN_LEB_CNT);
518 return -EINVAL;
519 }
520
521 if (!is_power_of_2(c->min_io_size)) {
522 ubifs_err("bad min. I/O size %d", c->min_io_size);
523 return -EINVAL;
524 }
525
526 /*
527 * UBIFS aligns all node to 8-byte boundary, so to make function in
528 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
529 * less than 8.
530 */
531 if (c->min_io_size < 8) {
532 c->min_io_size = 8;
533 c->min_io_shift = 3;
534 }
535
536 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
537 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
538
539 /*
540 * Initialize node length ranges which are mostly needed for node
541 * length validation.
542 */
543 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
544 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
545 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
546 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
547 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
548 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
549
550 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
551 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
552 c->ranges[UBIFS_ORPH_NODE].min_len =
553 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
554 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
555 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
556 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
557 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
558 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
559 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
560 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
561 /*
562 * Minimum indexing node size is amended later when superblock is
563 * read and the key length is known.
564 */
565 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
566 /*
567 * Maximum indexing node size is amended later when superblock is
568 * read and the fanout is known.
569 */
570 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
571
572 /*
573 * Initialize dead and dark LEB space watermarks.
574 *
575 * Dead space is the space which cannot be used. Its watermark is
576 * equivalent to min. I/O unit or minimum node size if it is greater
577 * then min. I/O unit.
578 *
579 * Dark space is the space which might be used, or might not, depending
580 * on which node should be written to the LEB. Its watermark is
581 * equivalent to maximum UBIFS node size.
582 */
583 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
584 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
585
9bbb5726
AB
586 /*
587 * Calculate how many bytes would be wasted at the end of LEB if it was
588 * fully filled with data nodes of maximum size. This is used in
589 * calculations when reporting free space.
590 */
591 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
39ce81ce 592
4793e7c5 593 /* Buffer size for bulk-reads */
6c0c42cd
AB
594 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
595 if (c->max_bu_buf_len > c->leb_size)
596 c->max_bu_buf_len = c->leb_size;
1e51764a
AB
597 return 0;
598}
599
600/**
601 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
602 * @c: UBIFS file-system description object
603 * @lnum: LEB the write-buffer was synchronized to
604 * @free: how many free bytes left in this LEB
605 * @pad: how many bytes were padded
606 *
607 * This is a callback function which is called by the I/O unit when the
608 * write-buffer is synchronized. We need this to correctly maintain space
609 * accounting in bud logical eraseblocks. This function returns zero in case of
610 * success and a negative error code in case of failure.
611 *
612 * This function actually belongs to the journal, but we keep it here because
613 * we want to keep it static.
614 */
615static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
616{
617 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
618}
619
620/*
79807d07 621 * init_constants_sb - initialize UBIFS constants.
1e51764a
AB
622 * @c: UBIFS file-system description object
623 *
624 * This is a helper function which initializes various UBIFS constants after
625 * the superblock has been read. It also checks various UBIFS parameters and
626 * makes sure they are all right. Returns zero in case of success and a
627 * negative error code in case of failure.
628 */
79807d07 629static int init_constants_sb(struct ubifs_info *c)
1e51764a
AB
630{
631 int tmp, err;
4d61db4f 632 long long tmp64;
1e51764a
AB
633
634 c->main_bytes = (long long)c->main_lebs * c->leb_size;
635 c->max_znode_sz = sizeof(struct ubifs_znode) +
636 c->fanout * sizeof(struct ubifs_zbranch);
637
638 tmp = ubifs_idx_node_sz(c, 1);
639 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
640 c->min_idx_node_sz = ALIGN(tmp, 8);
641
642 tmp = ubifs_idx_node_sz(c, c->fanout);
643 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
644 c->max_idx_node_sz = ALIGN(tmp, 8);
645
646 /* Make sure LEB size is large enough to fit full commit */
647 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
648 tmp = ALIGN(tmp, c->min_io_size);
649 if (tmp > c->leb_size) {
650 dbg_err("too small LEB size %d, at least %d needed",
651 c->leb_size, tmp);
652 return -EINVAL;
653 }
654
655 /*
656 * Make sure that the log is large enough to fit reference nodes for
657 * all buds plus one reserved LEB.
658 */
4d61db4f
AB
659 tmp64 = c->max_bud_bytes + c->leb_size - 1;
660 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
1e51764a
AB
661 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
662 tmp /= c->leb_size;
663 tmp += 1;
664 if (c->log_lebs < tmp) {
665 dbg_err("too small log %d LEBs, required min. %d LEBs",
666 c->log_lebs, tmp);
667 return -EINVAL;
668 }
669
670 /*
671 * When budgeting we assume worst-case scenarios when the pages are not
672 * be compressed and direntries are of the maximum size.
673 *
674 * Note, data, which may be stored in inodes is budgeted separately, so
675 * it is not included into 'c->inode_budget'.
676 */
677 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
678 c->inode_budget = UBIFS_INO_NODE_SZ;
679 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
680
681 /*
682 * When the amount of flash space used by buds becomes
683 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
684 * The writers are unblocked when the commit is finished. To avoid
685 * writers to be blocked UBIFS initiates background commit in advance,
686 * when number of bud bytes becomes above the limit defined below.
687 */
688 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
689
690 /*
691 * Ensure minimum journal size. All the bytes in the journal heads are
692 * considered to be used, when calculating the current journal usage.
693 * Consequently, if the journal is too small, UBIFS will treat it as
694 * always full.
695 */
4d61db4f 696 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
1e51764a
AB
697 if (c->bg_bud_bytes < tmp64)
698 c->bg_bud_bytes = tmp64;
699 if (c->max_bud_bytes < tmp64 + c->leb_size)
700 c->max_bud_bytes = tmp64 + c->leb_size;
701
702 err = ubifs_calc_lpt_geom(c);
703 if (err)
704 return err;
705
79807d07
AB
706 return 0;
707}
708
709/*
710 * init_constants_master - initialize UBIFS constants.
711 * @c: UBIFS file-system description object
712 *
713 * This is a helper function which initializes various UBIFS constants after
714 * the master node has been read. It also checks various UBIFS parameters and
715 * makes sure they are all right.
716 */
717static void init_constants_master(struct ubifs_info *c)
718{
719 long long tmp64;
720
1e51764a
AB
721 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
722
723 /*
724 * Calculate total amount of FS blocks. This number is not used
725 * internally because it does not make much sense for UBIFS, but it is
726 * necessary to report something for the 'statfs()' call.
727 *
7dad181b 728 * Subtract the LEB reserved for GC, the LEB which is reserved for
af14a1ad
AB
729 * deletions, minimum LEBs for the index, and assume only one journal
730 * head is available.
1e51764a 731 */
af14a1ad 732 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
4d61db4f 733 tmp64 *= (long long)c->leb_size - c->leb_overhead;
1e51764a
AB
734 tmp64 = ubifs_reported_space(c, tmp64);
735 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
1e51764a
AB
736}
737
738/**
739 * take_gc_lnum - reserve GC LEB.
740 * @c: UBIFS file-system description object
741 *
742 * This function ensures that the LEB reserved for garbage collection is
743 * unmapped and is marked as "taken" in lprops. We also have to set free space
744 * to LEB size and dirty space to zero, because lprops may contain out-of-date
745 * information if the file-system was un-mounted before it has been committed.
746 * This function returns zero in case of success and a negative error code in
747 * case of failure.
748 */
749static int take_gc_lnum(struct ubifs_info *c)
750{
751 int err;
752
753 if (c->gc_lnum == -1) {
754 ubifs_err("no LEB for GC");
755 return -EINVAL;
756 }
757
758 err = ubifs_leb_unmap(c, c->gc_lnum);
759 if (err)
760 return err;
761
762 /* And we have to tell lprops that this LEB is taken */
763 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
764 LPROPS_TAKEN, 0, 0);
765 return err;
766}
767
768/**
769 * alloc_wbufs - allocate write-buffers.
770 * @c: UBIFS file-system description object
771 *
772 * This helper function allocates and initializes UBIFS write-buffers. Returns
773 * zero in case of success and %-ENOMEM in case of failure.
774 */
775static int alloc_wbufs(struct ubifs_info *c)
776{
777 int i, err;
778
779 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
780 GFP_KERNEL);
781 if (!c->jheads)
782 return -ENOMEM;
783
784 /* Initialize journal heads */
785 for (i = 0; i < c->jhead_cnt; i++) {
786 INIT_LIST_HEAD(&c->jheads[i].buds_list);
787 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
788 if (err)
789 return err;
790
791 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
792 c->jheads[i].wbuf.jhead = i;
793 }
794
795 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
796 /*
797 * Garbage Collector head likely contains long-term data and
798 * does not need to be synchronized by timer.
799 */
800 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
801 c->jheads[GCHD].wbuf.timeout = 0;
802
803 return 0;
804}
805
806/**
807 * free_wbufs - free write-buffers.
808 * @c: UBIFS file-system description object
809 */
810static void free_wbufs(struct ubifs_info *c)
811{
812 int i;
813
814 if (c->jheads) {
815 for (i = 0; i < c->jhead_cnt; i++) {
816 kfree(c->jheads[i].wbuf.buf);
817 kfree(c->jheads[i].wbuf.inodes);
818 }
819 kfree(c->jheads);
820 c->jheads = NULL;
821 }
822}
823
824/**
825 * free_orphans - free orphans.
826 * @c: UBIFS file-system description object
827 */
828static void free_orphans(struct ubifs_info *c)
829{
830 struct ubifs_orphan *orph;
831
832 while (c->orph_dnext) {
833 orph = c->orph_dnext;
834 c->orph_dnext = orph->dnext;
835 list_del(&orph->list);
836 kfree(orph);
837 }
838
839 while (!list_empty(&c->orph_list)) {
840 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
841 list_del(&orph->list);
842 kfree(orph);
843 dbg_err("orphan list not empty at unmount");
844 }
845
846 vfree(c->orph_buf);
847 c->orph_buf = NULL;
848}
849
850/**
851 * free_buds - free per-bud objects.
852 * @c: UBIFS file-system description object
853 */
854static void free_buds(struct ubifs_info *c)
855{
856 struct rb_node *this = c->buds.rb_node;
857 struct ubifs_bud *bud;
858
859 while (this) {
860 if (this->rb_left)
861 this = this->rb_left;
862 else if (this->rb_right)
863 this = this->rb_right;
864 else {
865 bud = rb_entry(this, struct ubifs_bud, rb);
866 this = rb_parent(this);
867 if (this) {
868 if (this->rb_left == &bud->rb)
869 this->rb_left = NULL;
870 else
871 this->rb_right = NULL;
872 }
873 kfree(bud);
874 }
875 }
876}
877
878/**
879 * check_volume_empty - check if the UBI volume is empty.
880 * @c: UBIFS file-system description object
881 *
882 * This function checks if the UBIFS volume is empty by looking if its LEBs are
883 * mapped or not. The result of checking is stored in the @c->empty variable.
884 * Returns zero in case of success and a negative error code in case of
885 * failure.
886 */
887static int check_volume_empty(struct ubifs_info *c)
888{
889 int lnum, err;
890
891 c->empty = 1;
892 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
893 err = ubi_is_mapped(c->ubi, lnum);
894 if (unlikely(err < 0))
895 return err;
896 if (err == 1) {
897 c->empty = 0;
898 break;
899 }
900
901 cond_resched();
902 }
903
904 return 0;
905}
906
907/*
908 * UBIFS mount options.
909 *
910 * Opt_fast_unmount: do not run a journal commit before un-mounting
911 * Opt_norm_unmount: run a journal commit before un-mounting
4793e7c5
AH
912 * Opt_bulk_read: enable bulk-reads
913 * Opt_no_bulk_read: disable bulk-reads
2953e73f
AH
914 * Opt_chk_data_crc: check CRCs when reading data nodes
915 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
553dea4d 916 * Opt_override_compr: override default compressor
1e51764a
AB
917 * Opt_err: just end of array marker
918 */
919enum {
920 Opt_fast_unmount,
921 Opt_norm_unmount,
4793e7c5
AH
922 Opt_bulk_read,
923 Opt_no_bulk_read,
2953e73f
AH
924 Opt_chk_data_crc,
925 Opt_no_chk_data_crc,
553dea4d 926 Opt_override_compr,
1e51764a
AB
927 Opt_err,
928};
929
a447c093 930static const match_table_t tokens = {
1e51764a
AB
931 {Opt_fast_unmount, "fast_unmount"},
932 {Opt_norm_unmount, "norm_unmount"},
4793e7c5
AH
933 {Opt_bulk_read, "bulk_read"},
934 {Opt_no_bulk_read, "no_bulk_read"},
2953e73f
AH
935 {Opt_chk_data_crc, "chk_data_crc"},
936 {Opt_no_chk_data_crc, "no_chk_data_crc"},
553dea4d 937 {Opt_override_compr, "compr=%s"},
1e51764a
AB
938 {Opt_err, NULL},
939};
940
941/**
942 * ubifs_parse_options - parse mount parameters.
943 * @c: UBIFS file-system description object
944 * @options: parameters to parse
945 * @is_remount: non-zero if this is FS re-mount
946 *
947 * This function parses UBIFS mount options and returns zero in case success
948 * and a negative error code in case of failure.
949 */
950static int ubifs_parse_options(struct ubifs_info *c, char *options,
951 int is_remount)
952{
953 char *p;
954 substring_t args[MAX_OPT_ARGS];
955
956 if (!options)
957 return 0;
958
959 while ((p = strsep(&options, ","))) {
960 int token;
961
962 if (!*p)
963 continue;
964
965 token = match_token(p, tokens, args);
966 switch (token) {
967 case Opt_fast_unmount:
968 c->mount_opts.unmount_mode = 2;
969 c->fast_unmount = 1;
970 break;
971 case Opt_norm_unmount:
972 c->mount_opts.unmount_mode = 1;
973 c->fast_unmount = 0;
974 break;
4793e7c5
AH
975 case Opt_bulk_read:
976 c->mount_opts.bulk_read = 2;
977 c->bulk_read = 1;
978 break;
979 case Opt_no_bulk_read:
980 c->mount_opts.bulk_read = 1;
981 c->bulk_read = 0;
982 break;
2953e73f
AH
983 case Opt_chk_data_crc:
984 c->mount_opts.chk_data_crc = 2;
985 c->no_chk_data_crc = 0;
986 break;
987 case Opt_no_chk_data_crc:
988 c->mount_opts.chk_data_crc = 1;
989 c->no_chk_data_crc = 1;
990 break;
553dea4d
AB
991 case Opt_override_compr:
992 {
993 char *name = match_strdup(&args[0]);
994
995 if (!name)
996 return -ENOMEM;
997 if (!strcmp(name, "none"))
998 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
999 else if (!strcmp(name, "lzo"))
1000 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1001 else if (!strcmp(name, "zlib"))
1002 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1003 else {
1004 ubifs_err("unknown compressor \"%s\"", name);
1005 kfree(name);
1006 return -EINVAL;
1007 }
1008 kfree(name);
1009 c->mount_opts.override_compr = 1;
1010 c->default_compr = c->mount_opts.compr_type;
1011 break;
1012 }
1e51764a
AB
1013 default:
1014 ubifs_err("unrecognized mount option \"%s\" "
1015 "or missing value", p);
1016 return -EINVAL;
1017 }
1018 }
1019
1020 return 0;
1021}
1022
1023/**
1024 * destroy_journal - destroy journal data structures.
1025 * @c: UBIFS file-system description object
1026 *
1027 * This function destroys journal data structures including those that may have
1028 * been created by recovery functions.
1029 */
1030static void destroy_journal(struct ubifs_info *c)
1031{
1032 while (!list_empty(&c->unclean_leb_list)) {
1033 struct ubifs_unclean_leb *ucleb;
1034
1035 ucleb = list_entry(c->unclean_leb_list.next,
1036 struct ubifs_unclean_leb, list);
1037 list_del(&ucleb->list);
1038 kfree(ucleb);
1039 }
1040 while (!list_empty(&c->old_buds)) {
1041 struct ubifs_bud *bud;
1042
1043 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1044 list_del(&bud->list);
1045 kfree(bud);
1046 }
1047 ubifs_destroy_idx_gc(c);
1048 ubifs_destroy_size_tree(c);
1049 ubifs_tnc_close(c);
1050 free_buds(c);
1051}
1052
3477d204
AB
1053/**
1054 * bu_init - initialize bulk-read information.
1055 * @c: UBIFS file-system description object
1056 */
1057static void bu_init(struct ubifs_info *c)
1058{
1059 ubifs_assert(c->bulk_read == 1);
1060
1061 if (c->bu.buf)
1062 return; /* Already initialized */
1063
1064again:
1065 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1066 if (!c->bu.buf) {
1067 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1068 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1069 goto again;
1070 }
1071
1072 /* Just disable bulk-read */
1073 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1074 "disabling it", c->max_bu_buf_len);
1075 c->mount_opts.bulk_read = 1;
1076 c->bulk_read = 0;
1077 return;
1078 }
1079}
1080
1e51764a
AB
1081/**
1082 * mount_ubifs - mount UBIFS file-system.
1083 * @c: UBIFS file-system description object
1084 *
1085 * This function mounts UBIFS file system. Returns zero in case of success and
1086 * a negative error code in case of failure.
1087 *
1088 * Note, the function does not de-allocate resources it it fails half way
1089 * through, and the caller has to do this instead.
1090 */
1091static int mount_ubifs(struct ubifs_info *c)
1092{
1093 struct super_block *sb = c->vfs_sb;
1094 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1095 long long x;
1096 size_t sz;
1097
1098 err = init_constants_early(c);
1099 if (err)
1100 return err;
1101
17c2f9f8
AB
1102 err = ubifs_debugging_init(c);
1103 if (err)
1104 return err;
1e51764a
AB
1105
1106 err = check_volume_empty(c);
1107 if (err)
1108 goto out_free;
1109
1110 if (c->empty && (mounted_read_only || c->ro_media)) {
1111 /*
1112 * This UBI volume is empty, and read-only, or the file system
1113 * is mounted read-only - we cannot format it.
1114 */
1115 ubifs_err("can't format empty UBI volume: read-only %s",
1116 c->ro_media ? "UBI volume" : "mount");
1117 err = -EROFS;
1118 goto out_free;
1119 }
1120
1121 if (c->ro_media && !mounted_read_only) {
1122 ubifs_err("cannot mount read-write - read-only media");
1123 err = -EROFS;
1124 goto out_free;
1125 }
1126
1127 /*
1128 * The requirement for the buffer is that it should fit indexing B-tree
1129 * height amount of integers. We assume the height if the TNC tree will
1130 * never exceed 64.
1131 */
1132 err = -ENOMEM;
1133 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1134 if (!c->bottom_up_buf)
1135 goto out_free;
1136
1137 c->sbuf = vmalloc(c->leb_size);
1138 if (!c->sbuf)
1139 goto out_free;
1140
1141 if (!mounted_read_only) {
1142 c->ileb_buf = vmalloc(c->leb_size);
1143 if (!c->ileb_buf)
1144 goto out_free;
1145 }
1146
3477d204
AB
1147 if (c->bulk_read == 1)
1148 bu_init(c);
1149
1150 /*
1151 * We have to check all CRCs, even for data nodes, when we mount the FS
1152 * (specifically, when we are replaying).
1153 */
2953e73f
AH
1154 c->always_chk_crc = 1;
1155
1e51764a
AB
1156 err = ubifs_read_superblock(c);
1157 if (err)
1158 goto out_free;
1159
1160 /*
553dea4d
AB
1161 * Make sure the compressor which is set as default in the superblock
1162 * or overriden by mount options is actually compiled in.
1e51764a
AB
1163 */
1164 if (!ubifs_compr_present(c->default_compr)) {
553dea4d
AB
1165 ubifs_err("'compressor \"%s\" is not compiled in",
1166 ubifs_compr_name(c->default_compr));
1167 goto out_free;
1e51764a
AB
1168 }
1169
79807d07 1170 err = init_constants_sb(c);
1e51764a 1171 if (err)
17c2f9f8 1172 goto out_free;
1e51764a
AB
1173
1174 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1175 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1176 c->cbuf = kmalloc(sz, GFP_NOFS);
1177 if (!c->cbuf) {
1178 err = -ENOMEM;
17c2f9f8 1179 goto out_free;
1e51764a
AB
1180 }
1181
0855f310 1182 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1e51764a
AB
1183 if (!mounted_read_only) {
1184 err = alloc_wbufs(c);
1185 if (err)
1186 goto out_cbuf;
1187
1188 /* Create background thread */
1e51764a 1189 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1e51764a
AB
1190 if (IS_ERR(c->bgt)) {
1191 err = PTR_ERR(c->bgt);
1192 c->bgt = NULL;
1193 ubifs_err("cannot spawn \"%s\", error %d",
1194 c->bgt_name, err);
1195 goto out_wbufs;
1196 }
1197 wake_up_process(c->bgt);
1198 }
1199
1200 err = ubifs_read_master(c);
1201 if (err)
1202 goto out_master;
1203
79807d07
AB
1204 init_constants_master(c);
1205
1e51764a
AB
1206 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1207 ubifs_msg("recovery needed");
1208 c->need_recovery = 1;
1209 if (!mounted_read_only) {
1210 err = ubifs_recover_inl_heads(c, c->sbuf);
1211 if (err)
1212 goto out_master;
1213 }
1214 } else if (!mounted_read_only) {
1215 /*
1216 * Set the "dirty" flag so that if we reboot uncleanly we
1217 * will notice this immediately on the next mount.
1218 */
1219 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1220 err = ubifs_write_master(c);
1221 if (err)
1222 goto out_master;
1223 }
1224
1225 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1226 if (err)
1227 goto out_lpt;
1228
1229 err = dbg_check_idx_size(c, c->old_idx_sz);
1230 if (err)
1231 goto out_lpt;
1232
1233 err = ubifs_replay_journal(c);
1234 if (err)
1235 goto out_journal;
1236
1237 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1238 if (err)
1239 goto out_orphans;
1240
1241 if (!mounted_read_only) {
1242 int lnum;
1243
1244 /* Check for enough free space */
1245 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1246 ubifs_err("insufficient available space");
1247 err = -EINVAL;
1248 goto out_orphans;
1249 }
1250
1251 /* Check for enough log space */
1252 lnum = c->lhead_lnum + 1;
1253 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1254 lnum = UBIFS_LOG_LNUM;
1255 if (lnum == c->ltail_lnum) {
1256 err = ubifs_consolidate_log(c);
1257 if (err)
1258 goto out_orphans;
1259 }
1260
1261 if (c->need_recovery) {
1262 err = ubifs_recover_size(c);
1263 if (err)
1264 goto out_orphans;
1265 err = ubifs_rcvry_gc_commit(c);
1266 } else
1267 err = take_gc_lnum(c);
1268 if (err)
1269 goto out_orphans;
1270
1271 err = dbg_check_lprops(c);
1272 if (err)
1273 goto out_orphans;
1274 } else if (c->need_recovery) {
1275 err = ubifs_recover_size(c);
1276 if (err)
1277 goto out_orphans;
1278 }
1279
1280 spin_lock(&ubifs_infos_lock);
1281 list_add_tail(&c->infos_list, &ubifs_infos);
1282 spin_unlock(&ubifs_infos_lock);
1283
1284 if (c->need_recovery) {
1285 if (mounted_read_only)
1286 ubifs_msg("recovery deferred");
1287 else {
1288 c->need_recovery = 0;
1289 ubifs_msg("recovery completed");
1290 }
1291 }
1292
552ff317
AB
1293 err = dbg_debugfs_init_fs(c);
1294 if (err)
1295 goto out_infos;
1296
1e51764a
AB
1297 err = dbg_check_filesystem(c);
1298 if (err)
1299 goto out_infos;
1300
2953e73f
AH
1301 c->always_chk_crc = 0;
1302
ce769caa
AB
1303 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1304 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1e51764a
AB
1305 if (mounted_read_only)
1306 ubifs_msg("mounted read-only");
1307 x = (long long)c->main_lebs * c->leb_size;
948cfb21
AB
1308 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1309 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1e51764a 1310 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
948cfb21
AB
1311 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1312 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1313 ubifs_msg("media format: %d (latest is %d)",
1e51764a 1314 c->fmt_version, UBIFS_FORMAT_VERSION);
948cfb21 1315 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
fae7fb29 1316 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
948cfb21 1317 c->report_rp_size, c->report_rp_size >> 10);
1e51764a
AB
1318
1319 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1320 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1321 dbg_msg("LEB size: %d bytes (%d KiB)",
948cfb21 1322 c->leb_size, c->leb_size >> 10);
1e51764a
AB
1323 dbg_msg("data journal heads: %d",
1324 c->jhead_cnt - NONDATA_JHEADS_CNT);
1325 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1326 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1327 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1328 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1329 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1330 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1331 dbg_msg("fast unmount: %d", c->fast_unmount);
1332 dbg_msg("big_lpt %d", c->big_lpt);
1333 dbg_msg("log LEBs: %d (%d - %d)",
1334 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1335 dbg_msg("LPT area LEBs: %d (%d - %d)",
1336 c->lpt_lebs, c->lpt_first, c->lpt_last);
1337 dbg_msg("orphan area LEBs: %d (%d - %d)",
1338 c->orph_lebs, c->orph_first, c->orph_last);
1339 dbg_msg("main area LEBs: %d (%d - %d)",
1340 c->main_lebs, c->main_first, c->leb_cnt - 1);
1341 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1342 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1343 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1344 dbg_msg("key hash type: %d", c->key_hash_type);
1345 dbg_msg("tree fanout: %d", c->fanout);
1346 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1347 dbg_msg("first main LEB: %d", c->main_first);
1348 dbg_msg("dead watermark: %d", c->dead_wm);
1349 dbg_msg("dark watermark: %d", c->dark_wm);
1350 x = (long long)c->main_lebs * c->dark_wm;
1351 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1352 x, x >> 10, x >> 20);
1353 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1354 c->max_bud_bytes, c->max_bud_bytes >> 10,
1355 c->max_bud_bytes >> 20);
1356 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1357 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1358 c->bg_bud_bytes >> 20);
1359 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1360 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1361 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1362 dbg_msg("commit number: %llu", c->cmt_no);
1363
1364 return 0;
1365
1366out_infos:
1367 spin_lock(&ubifs_infos_lock);
1368 list_del(&c->infos_list);
1369 spin_unlock(&ubifs_infos_lock);
1370out_orphans:
1371 free_orphans(c);
1372out_journal:
1373 destroy_journal(c);
1374out_lpt:
1375 ubifs_lpt_free(c, 0);
1376out_master:
1377 kfree(c->mst_node);
1378 kfree(c->rcvrd_mst_node);
1379 if (c->bgt)
1380 kthread_stop(c->bgt);
1381out_wbufs:
1382 free_wbufs(c);
1383out_cbuf:
1384 kfree(c->cbuf);
1e51764a 1385out_free:
3477d204 1386 kfree(c->bu.buf);
1e51764a
AB
1387 vfree(c->ileb_buf);
1388 vfree(c->sbuf);
1389 kfree(c->bottom_up_buf);
17c2f9f8 1390 ubifs_debugging_exit(c);
1e51764a
AB
1391 return err;
1392}
1393
1394/**
1395 * ubifs_umount - un-mount UBIFS file-system.
1396 * @c: UBIFS file-system description object
1397 *
1398 * Note, this function is called to free allocated resourced when un-mounting,
1399 * as well as free resources when an error occurred while we were half way
1400 * through mounting (error path cleanup function). So it has to make sure the
1401 * resource was actually allocated before freeing it.
1402 */
1403static void ubifs_umount(struct ubifs_info *c)
1404{
1405 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1406 c->vi.vol_id);
1407
552ff317 1408 dbg_debugfs_exit_fs(c);
1e51764a
AB
1409 spin_lock(&ubifs_infos_lock);
1410 list_del(&c->infos_list);
1411 spin_unlock(&ubifs_infos_lock);
1412
1413 if (c->bgt)
1414 kthread_stop(c->bgt);
1415
1416 destroy_journal(c);
1417 free_wbufs(c);
1418 free_orphans(c);
1419 ubifs_lpt_free(c, 0);
1420
1421 kfree(c->cbuf);
1422 kfree(c->rcvrd_mst_node);
1423 kfree(c->mst_node);
3477d204
AB
1424 kfree(c->bu.buf);
1425 vfree(c->ileb_buf);
1e51764a
AB
1426 vfree(c->sbuf);
1427 kfree(c->bottom_up_buf);
17c2f9f8 1428 ubifs_debugging_exit(c);
1e51764a
AB
1429}
1430
1431/**
1432 * ubifs_remount_rw - re-mount in read-write mode.
1433 * @c: UBIFS file-system description object
1434 *
1435 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1436 * mode. This function allocates the needed resources and re-mounts UBIFS in
1437 * read-write mode.
1438 */
1439static int ubifs_remount_rw(struct ubifs_info *c)
1440{
1441 int err, lnum;
1442
1443 if (c->ro_media)
1444 return -EINVAL;
1445
1446 mutex_lock(&c->umount_mutex);
1447 c->remounting_rw = 1;
2953e73f 1448 c->always_chk_crc = 1;
1e51764a
AB
1449
1450 /* Check for enough free space */
1451 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1452 ubifs_err("insufficient available space");
1453 err = -EINVAL;
1454 goto out;
1455 }
1456
1457 if (c->old_leb_cnt != c->leb_cnt) {
1458 struct ubifs_sb_node *sup;
1459
1460 sup = ubifs_read_sb_node(c);
1461 if (IS_ERR(sup)) {
1462 err = PTR_ERR(sup);
1463 goto out;
1464 }
1465 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1466 err = ubifs_write_sb_node(c, sup);
1467 if (err)
1468 goto out;
1469 }
1470
1471 if (c->need_recovery) {
1472 ubifs_msg("completing deferred recovery");
1473 err = ubifs_write_rcvrd_mst_node(c);
1474 if (err)
1475 goto out;
1476 err = ubifs_recover_size(c);
1477 if (err)
1478 goto out;
1479 err = ubifs_clean_lebs(c, c->sbuf);
1480 if (err)
1481 goto out;
1482 err = ubifs_recover_inl_heads(c, c->sbuf);
1483 if (err)
1484 goto out;
1485 }
1486
1487 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1488 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1489 err = ubifs_write_master(c);
1490 if (err)
1491 goto out;
1492 }
1493
1494 c->ileb_buf = vmalloc(c->leb_size);
1495 if (!c->ileb_buf) {
1496 err = -ENOMEM;
1497 goto out;
1498 }
1499
1500 err = ubifs_lpt_init(c, 0, 1);
1501 if (err)
1502 goto out;
1503
1504 err = alloc_wbufs(c);
1505 if (err)
1506 goto out;
1507
1508 ubifs_create_buds_lists(c);
1509
1510 /* Create background thread */
1511 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1e51764a
AB
1512 if (IS_ERR(c->bgt)) {
1513 err = PTR_ERR(c->bgt);
1514 c->bgt = NULL;
1515 ubifs_err("cannot spawn \"%s\", error %d",
1516 c->bgt_name, err);
2953e73f 1517 goto out;
1e51764a
AB
1518 }
1519 wake_up_process(c->bgt);
1520
1521 c->orph_buf = vmalloc(c->leb_size);
2953e73f
AH
1522 if (!c->orph_buf) {
1523 err = -ENOMEM;
1524 goto out;
1525 }
1e51764a
AB
1526
1527 /* Check for enough log space */
1528 lnum = c->lhead_lnum + 1;
1529 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1530 lnum = UBIFS_LOG_LNUM;
1531 if (lnum == c->ltail_lnum) {
1532 err = ubifs_consolidate_log(c);
1533 if (err)
1534 goto out;
1535 }
1536
1537 if (c->need_recovery)
1538 err = ubifs_rcvry_gc_commit(c);
1539 else
1540 err = take_gc_lnum(c);
1541 if (err)
1542 goto out;
1543
1544 if (c->need_recovery) {
1545 c->need_recovery = 0;
1546 ubifs_msg("deferred recovery completed");
1547 }
1548
1549 dbg_gen("re-mounted read-write");
1550 c->vfs_sb->s_flags &= ~MS_RDONLY;
1551 c->remounting_rw = 0;
2953e73f 1552 c->always_chk_crc = 0;
1e51764a
AB
1553 mutex_unlock(&c->umount_mutex);
1554 return 0;
1555
1556out:
1557 vfree(c->orph_buf);
1558 c->orph_buf = NULL;
1559 if (c->bgt) {
1560 kthread_stop(c->bgt);
1561 c->bgt = NULL;
1562 }
1563 free_wbufs(c);
1564 vfree(c->ileb_buf);
1565 c->ileb_buf = NULL;
1566 ubifs_lpt_free(c, 1);
1567 c->remounting_rw = 0;
2953e73f 1568 c->always_chk_crc = 0;
1e51764a
AB
1569 mutex_unlock(&c->umount_mutex);
1570 return err;
1571}
1572
1573/**
1574 * commit_on_unmount - commit the journal when un-mounting.
1575 * @c: UBIFS file-system description object
1576 *
af2eb563
AB
1577 * This function is called during un-mounting and re-mounting, and it commits
1578 * the journal unless the "fast unmount" mode is enabled. It also avoids
1579 * committing the journal if it contains too few data.
1e51764a
AB
1580 */
1581static void commit_on_unmount(struct ubifs_info *c)
1582{
1583 if (!c->fast_unmount) {
1584 long long bud_bytes;
1585
1586 spin_lock(&c->buds_lock);
1587 bud_bytes = c->bud_bytes;
1588 spin_unlock(&c->buds_lock);
1589 if (bud_bytes > c->leb_size)
1590 ubifs_run_commit(c);
1591 }
1592}
1593
1594/**
1595 * ubifs_remount_ro - re-mount in read-only mode.
1596 * @c: UBIFS file-system description object
1597 *
1598 * We rely on VFS to have stopped writing. Possibly the background thread could
1599 * be running a commit, however kthread_stop will wait in that case.
1600 */
1601static void ubifs_remount_ro(struct ubifs_info *c)
1602{
1603 int i, err;
1604
1605 ubifs_assert(!c->need_recovery);
1606 commit_on_unmount(c);
1607
1608 mutex_lock(&c->umount_mutex);
1609 if (c->bgt) {
1610 kthread_stop(c->bgt);
1611 c->bgt = NULL;
1612 }
1613
1614 for (i = 0; i < c->jhead_cnt; i++) {
1615 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1616 del_timer_sync(&c->jheads[i].wbuf.timer);
1617 }
1618
1619 if (!c->ro_media) {
1620 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1621 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1622 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1623 err = ubifs_write_master(c);
1624 if (err)
1625 ubifs_ro_mode(c, err);
1626 }
1627
1628 ubifs_destroy_idx_gc(c);
1629 free_wbufs(c);
1630 vfree(c->orph_buf);
1631 c->orph_buf = NULL;
1632 vfree(c->ileb_buf);
1633 c->ileb_buf = NULL;
1634 ubifs_lpt_free(c, 1);
1635 mutex_unlock(&c->umount_mutex);
1636}
1637
1638static void ubifs_put_super(struct super_block *sb)
1639{
1640 int i;
1641 struct ubifs_info *c = sb->s_fs_info;
1642
1643 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1644 c->vi.vol_id);
1645 /*
1646 * The following asserts are only valid if there has not been a failure
1647 * of the media. For example, there will be dirty inodes if we failed
1648 * to write them back because of I/O errors.
1649 */
1650 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1651 ubifs_assert(c->budg_idx_growth == 0);
7d32c2bb 1652 ubifs_assert(c->budg_dd_growth == 0);
1e51764a
AB
1653 ubifs_assert(c->budg_data_growth == 0);
1654
1655 /*
1656 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1657 * and file system un-mount. Namely, it prevents the shrinker from
1658 * picking this superblock for shrinking - it will be just skipped if
1659 * the mutex is locked.
1660 */
1661 mutex_lock(&c->umount_mutex);
1662 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1663 /*
1664 * First of all kill the background thread to make sure it does
1665 * not interfere with un-mounting and freeing resources.
1666 */
1667 if (c->bgt) {
1668 kthread_stop(c->bgt);
1669 c->bgt = NULL;
1670 }
1671
1672 /* Synchronize write-buffers */
1673 if (c->jheads)
1674 for (i = 0; i < c->jhead_cnt; i++) {
1675 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1676 del_timer_sync(&c->jheads[i].wbuf.timer);
1677 }
1678
1679 /*
1680 * On fatal errors c->ro_media is set to 1, in which case we do
1681 * not write the master node.
1682 */
1683 if (!c->ro_media) {
1684 /*
1685 * We are being cleanly unmounted which means the
1686 * orphans were killed - indicate this in the master
1687 * node. Also save the reserved GC LEB number.
1688 */
1689 int err;
1690
1691 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1692 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1693 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1694 err = ubifs_write_master(c);
1695 if (err)
1696 /*
1697 * Recovery will attempt to fix the master area
1698 * next mount, so we just print a message and
1699 * continue to unmount normally.
1700 */
1701 ubifs_err("failed to write master node, "
1702 "error %d", err);
1703 }
1704 }
1705
1706 ubifs_umount(c);
1707 bdi_destroy(&c->bdi);
1708 ubi_close_volume(c->ubi);
1709 mutex_unlock(&c->umount_mutex);
1710 kfree(c);
1711}
1712
1713static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1714{
1715 int err;
1716 struct ubifs_info *c = sb->s_fs_info;
1717
1718 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1719
1720 err = ubifs_parse_options(c, data, 1);
1721 if (err) {
1722 ubifs_err("invalid or unknown remount parameter");
1723 return err;
1724 }
3477d204 1725
1e51764a
AB
1726 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1727 err = ubifs_remount_rw(c);
1728 if (err)
1729 return err;
1730 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
1731 ubifs_remount_ro(c);
1732
3477d204
AB
1733 if (c->bulk_read == 1)
1734 bu_init(c);
1735 else {
1736 dbg_gen("disable bulk-read");
1737 kfree(c->bu.buf);
1738 c->bu.buf = NULL;
1739 }
1740
1e51764a
AB
1741 return 0;
1742}
1743
1744struct super_operations ubifs_super_operations = {
1745 .alloc_inode = ubifs_alloc_inode,
1746 .destroy_inode = ubifs_destroy_inode,
1747 .put_super = ubifs_put_super,
1748 .write_inode = ubifs_write_inode,
1749 .delete_inode = ubifs_delete_inode,
1750 .statfs = ubifs_statfs,
1751 .dirty_inode = ubifs_dirty_inode,
1752 .remount_fs = ubifs_remount_fs,
1753 .show_options = ubifs_show_options,
1754 .sync_fs = ubifs_sync_fs,
1755};
1756
1757/**
1758 * open_ubi - parse UBI device name string and open the UBI device.
1759 * @name: UBI volume name
1760 * @mode: UBI volume open mode
1761 *
1762 * There are several ways to specify UBI volumes when mounting UBIFS:
1763 * o ubiX_Y - UBI device number X, volume Y;
1764 * o ubiY - UBI device number 0, volume Y;
1765 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1766 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1767 *
1768 * Alternative '!' separator may be used instead of ':' (because some shells
1769 * like busybox may interpret ':' as an NFS host name separator). This function
1770 * returns ubi volume object in case of success and a negative error code in
1771 * case of failure.
1772 */
1773static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1774{
1775 int dev, vol;
1776 char *endptr;
1777
1778 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1779 return ERR_PTR(-EINVAL);
1780
1781 /* ubi:NAME method */
1782 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1783 return ubi_open_volume_nm(0, name + 4, mode);
1784
1785 if (!isdigit(name[3]))
1786 return ERR_PTR(-EINVAL);
1787
1788 dev = simple_strtoul(name + 3, &endptr, 0);
1789
1790 /* ubiY method */
1791 if (*endptr == '\0')
1792 return ubi_open_volume(0, dev, mode);
1793
1794 /* ubiX_Y method */
1795 if (*endptr == '_' && isdigit(endptr[1])) {
1796 vol = simple_strtoul(endptr + 1, &endptr, 0);
1797 if (*endptr != '\0')
1798 return ERR_PTR(-EINVAL);
1799 return ubi_open_volume(dev, vol, mode);
1800 }
1801
1802 /* ubiX:NAME method */
1803 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1804 return ubi_open_volume_nm(dev, ++endptr, mode);
1805
1806 return ERR_PTR(-EINVAL);
1807}
1808
1809static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1810{
1811 struct ubi_volume_desc *ubi = sb->s_fs_info;
1812 struct ubifs_info *c;
1813 struct inode *root;
1814 int err;
1815
1816 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1817 if (!c)
1818 return -ENOMEM;
1819
1820 spin_lock_init(&c->cnt_lock);
1821 spin_lock_init(&c->cs_lock);
1822 spin_lock_init(&c->buds_lock);
1823 spin_lock_init(&c->space_lock);
1824 spin_lock_init(&c->orphan_lock);
1825 init_rwsem(&c->commit_sem);
1826 mutex_init(&c->lp_mutex);
1827 mutex_init(&c->tnc_mutex);
1828 mutex_init(&c->log_mutex);
1829 mutex_init(&c->mst_mutex);
1830 mutex_init(&c->umount_mutex);
3477d204 1831 mutex_init(&c->bu_mutex);
1e51764a
AB
1832 init_waitqueue_head(&c->cmt_wq);
1833 c->buds = RB_ROOT;
1834 c->old_idx = RB_ROOT;
1835 c->size_tree = RB_ROOT;
1836 c->orph_tree = RB_ROOT;
1837 INIT_LIST_HEAD(&c->infos_list);
1838 INIT_LIST_HEAD(&c->idx_gc);
1839 INIT_LIST_HEAD(&c->replay_list);
1840 INIT_LIST_HEAD(&c->replay_buds);
1841 INIT_LIST_HEAD(&c->uncat_list);
1842 INIT_LIST_HEAD(&c->empty_list);
1843 INIT_LIST_HEAD(&c->freeable_list);
1844 INIT_LIST_HEAD(&c->frdi_idx_list);
1845 INIT_LIST_HEAD(&c->unclean_leb_list);
1846 INIT_LIST_HEAD(&c->old_buds);
1847 INIT_LIST_HEAD(&c->orph_list);
1848 INIT_LIST_HEAD(&c->orph_new);
1849
1850 c->highest_inum = UBIFS_FIRST_INO;
1e51764a
AB
1851 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1852
1853 ubi_get_volume_info(ubi, &c->vi);
1854 ubi_get_device_info(c->vi.ubi_num, &c->di);
1855
1856 /* Re-open the UBI device in read-write mode */
1857 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1858 if (IS_ERR(c->ubi)) {
1859 err = PTR_ERR(c->ubi);
1860 goto out_free;
1861 }
1862
1863 /*
0a883a05 1864 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1e51764a
AB
1865 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1866 * which means the user would have to wait not just for their own I/O
0a883a05 1867 * but the read-ahead I/O as well i.e. completely pointless.
1e51764a
AB
1868 *
1869 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1870 */
1871 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1872 c->bdi.unplug_io_fn = default_unplug_io_fn;
1873 err = bdi_init(&c->bdi);
1874 if (err)
1875 goto out_close;
1876
1877 err = ubifs_parse_options(c, data, 0);
1878 if (err)
1879 goto out_bdi;
1880
1881 c->vfs_sb = sb;
1882
1883 sb->s_fs_info = c;
1884 sb->s_magic = UBIFS_SUPER_MAGIC;
1885 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1886 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1887 sb->s_dev = c->vi.cdev;
1888 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1889 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1890 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1891 sb->s_op = &ubifs_super_operations;
1892
1893 mutex_lock(&c->umount_mutex);
1894 err = mount_ubifs(c);
1895 if (err) {
1896 ubifs_assert(err < 0);
1897 goto out_unlock;
1898 }
1899
1900 /* Read the root inode */
1901 root = ubifs_iget(sb, UBIFS_ROOT_INO);
1902 if (IS_ERR(root)) {
1903 err = PTR_ERR(root);
1904 goto out_umount;
1905 }
1906
1907 sb->s_root = d_alloc_root(root);
1908 if (!sb->s_root)
1909 goto out_iput;
1910
1911 mutex_unlock(&c->umount_mutex);
1e51764a
AB
1912 return 0;
1913
1914out_iput:
1915 iput(root);
1916out_umount:
1917 ubifs_umount(c);
1918out_unlock:
1919 mutex_unlock(&c->umount_mutex);
1920out_bdi:
1921 bdi_destroy(&c->bdi);
1922out_close:
1923 ubi_close_volume(c->ubi);
1924out_free:
1925 kfree(c);
1926 return err;
1927}
1928
1929static int sb_test(struct super_block *sb, void *data)
1930{
1931 dev_t *dev = data;
1932
1933 return sb->s_dev == *dev;
1934}
1935
1936static int sb_set(struct super_block *sb, void *data)
1937{
1938 dev_t *dev = data;
1939
1940 sb->s_dev = *dev;
1941 return 0;
1942}
1943
1944static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
1945 const char *name, void *data, struct vfsmount *mnt)
1946{
1947 struct ubi_volume_desc *ubi;
1948 struct ubi_volume_info vi;
1949 struct super_block *sb;
1950 int err;
1951
1952 dbg_gen("name %s, flags %#x", name, flags);
1953
1954 /*
1955 * Get UBI device number and volume ID. Mount it read-only so far
1956 * because this might be a new mount point, and UBI allows only one
1957 * read-write user at a time.
1958 */
1959 ubi = open_ubi(name, UBI_READONLY);
1960 if (IS_ERR(ubi)) {
1961 ubifs_err("cannot open \"%s\", error %d",
1962 name, (int)PTR_ERR(ubi));
1963 return PTR_ERR(ubi);
1964 }
1965 ubi_get_volume_info(ubi, &vi);
1966
1967 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
1968
1969 sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
1970 if (IS_ERR(sb)) {
1971 err = PTR_ERR(sb);
1972 goto out_close;
1973 }
1974
1975 if (sb->s_root) {
1976 /* A new mount point for already mounted UBIFS */
1977 dbg_gen("this ubi volume is already mounted");
1978 if ((flags ^ sb->s_flags) & MS_RDONLY) {
1979 err = -EBUSY;
1980 goto out_deact;
1981 }
1982 } else {
1983 sb->s_flags = flags;
1984 /*
1985 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
1986 * replaced by 'c'.
1987 */
1988 sb->s_fs_info = ubi;
1989 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
1990 if (err)
1991 goto out_deact;
1992 /* We do not support atime */
1993 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
1994 }
1995
1996 /* 'fill_super()' opens ubi again so we must close it here */
1997 ubi_close_volume(ubi);
1998
1999 return simple_set_mnt(mnt, sb);
2000
2001out_deact:
2002 up_write(&sb->s_umount);
2003 deactivate_super(sb);
2004out_close:
2005 ubi_close_volume(ubi);
2006 return err;
2007}
2008
2009static void ubifs_kill_sb(struct super_block *sb)
2010{
2011 struct ubifs_info *c = sb->s_fs_info;
2012
2013 /*
2014 * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
2015 * in order to be outside BKL.
2016 */
2017 if (sb->s_root && !(sb->s_flags & MS_RDONLY))
2018 commit_on_unmount(c);
2019 /* The un-mount routine is actually done in put_super() */
2020 generic_shutdown_super(sb);
2021}
2022
2023static struct file_system_type ubifs_fs_type = {
2024 .name = "ubifs",
2025 .owner = THIS_MODULE,
2026 .get_sb = ubifs_get_sb,
2027 .kill_sb = ubifs_kill_sb
2028};
2029
2030/*
2031 * Inode slab cache constructor.
2032 */
51cc5068 2033static void inode_slab_ctor(void *obj)
1e51764a
AB
2034{
2035 struct ubifs_inode *ui = obj;
2036 inode_init_once(&ui->vfs_inode);
2037}
2038
2039static int __init ubifs_init(void)
2040{
2041 int err;
2042
2043 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2044
2045 /* Make sure node sizes are 8-byte aligned */
2046 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2047 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2048 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2049 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2050 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2051 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2052 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2053 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2054 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2055 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2056 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2057
2058 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2059 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2060 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2061 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2062 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2063 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2064
2065 /* Check min. node size */
2066 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2067 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2068 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2069 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2070
2071 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2072 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2073 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2074 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2075
2076 /* Defined node sizes */
2077 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2078 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2079 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2080 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2081
a1dc080c
AB
2082 /*
2083 * We use 2 bit wide bit-fields to store compression type, which should
2084 * be amended if more compressors are added. The bit-fields are:
553dea4d
AB
2085 * @compr_type in 'struct ubifs_inode', @default_compr in
2086 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
a1dc080c
AB
2087 */
2088 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2089
1e51764a
AB
2090 /*
2091 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2092 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2093 */
2094 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2095 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2096 " at least 4096 bytes",
2097 (unsigned int)PAGE_CACHE_SIZE);
2098 return -EINVAL;
2099 }
2100
2101 err = register_filesystem(&ubifs_fs_type);
2102 if (err) {
2103 ubifs_err("cannot register file system, error %d", err);
2104 return err;
2105 }
2106
2107 err = -ENOMEM;
2108 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2109 sizeof(struct ubifs_inode), 0,
2110 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2111 &inode_slab_ctor);
2112 if (!ubifs_inode_slab)
2113 goto out_reg;
2114
2115 register_shrinker(&ubifs_shrinker_info);
2116
2117 err = ubifs_compressors_init();
552ff317
AB
2118 if (err)
2119 goto out_shrinker;
2120
2121 err = dbg_debugfs_init();
1e51764a
AB
2122 if (err)
2123 goto out_compr;
2124
2125 return 0;
2126
2127out_compr:
552ff317
AB
2128 ubifs_compressors_exit();
2129out_shrinker:
1e51764a
AB
2130 unregister_shrinker(&ubifs_shrinker_info);
2131 kmem_cache_destroy(ubifs_inode_slab);
2132out_reg:
2133 unregister_filesystem(&ubifs_fs_type);
2134 return err;
2135}
2136/* late_initcall to let compressors initialize first */
2137late_initcall(ubifs_init);
2138
2139static void __exit ubifs_exit(void)
2140{
2141 ubifs_assert(list_empty(&ubifs_infos));
2142 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2143
552ff317 2144 dbg_debugfs_exit();
1e51764a
AB
2145 ubifs_compressors_exit();
2146 unregister_shrinker(&ubifs_shrinker_info);
2147 kmem_cache_destroy(ubifs_inode_slab);
2148 unregister_filesystem(&ubifs_fs_type);
2149}
2150module_exit(ubifs_exit);
2151
2152MODULE_LICENSE("GPL");
2153MODULE_VERSION(__stringify(UBIFS_VERSION));
2154MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2155MODULE_DESCRIPTION("UBIFS - UBI File System");