blob: d80b2aef42b661001b2c87ac645f9a8ac29fbfd7 [file] [log] [blame]
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001/*
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>
Artem Bityutskiy1e517642008-07-14 19:08:37 +030033#include <linux/kthread.h>
34#include <linux/parser.h>
35#include <linux/seq_file.h>
36#include <linux/mount.h>
37#include "ubifs.h"
38
Artem Bityutskiy39ce81c2008-11-18 18:09:49 +020039/*
40 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
41 * allocating too much.
42 */
43#define UBIFS_KMALLOC_OK (128*1024)
44
Artem Bityutskiy1e517642008-07-14 19:08:37 +030045/* Slab cache for UBIFS inodes */
46struct kmem_cache *ubifs_inode_slab;
47
48/* UBIFS TNC shrinker description */
49static struct shrinker ubifs_shrinker_info = {
50 .shrink = ubifs_shrinker,
51 .seeks = DEFAULT_SEEKS,
52};
53
54/**
55 * validate_inode - validate inode.
56 * @c: UBIFS file-system description object
57 * @inode: the inode to validate
58 *
59 * This is a helper function for 'ubifs_iget()' which validates various fields
60 * of a newly built inode to make sure they contain sane values and prevent
61 * possible vulnerabilities. Returns zero if the inode is all right and
62 * a non-zero error code if not.
63 */
64static int validate_inode(struct ubifs_info *c, const struct inode *inode)
65{
66 int err;
67 const struct ubifs_inode *ui = ubifs_inode(inode);
68
69 if (inode->i_size > c->max_inode_sz) {
70 ubifs_err("inode is too large (%lld)",
71 (long long)inode->i_size);
72 return 1;
73 }
74
75 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
76 ubifs_err("unknown compression type %d", ui->compr_type);
77 return 2;
78 }
79
80 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
81 return 3;
82
83 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
84 return 4;
85
86 if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
87 return 5;
88
89 if (!ubifs_compr_present(ui->compr_type)) {
90 ubifs_warn("inode %lu uses '%s' compression, but it was not "
91 "compiled in", inode->i_ino,
92 ubifs_compr_name(ui->compr_type));
93 }
94
95 err = dbg_check_dir_size(c, inode);
96 return err;
97}
98
99struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
100{
101 int err;
102 union ubifs_key key;
103 struct ubifs_ino_node *ino;
104 struct ubifs_info *c = sb->s_fs_info;
105 struct inode *inode;
106 struct ubifs_inode *ui;
107
108 dbg_gen("inode %lu", inum);
109
110 inode = iget_locked(sb, inum);
111 if (!inode)
112 return ERR_PTR(-ENOMEM);
113 if (!(inode->i_state & I_NEW))
114 return inode;
115 ui = ubifs_inode(inode);
116
117 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
118 if (!ino) {
119 err = -ENOMEM;
120 goto out;
121 }
122
123 ino_key_init(c, &key, inode->i_ino);
124
125 err = ubifs_tnc_lookup(c, &key, ino);
126 if (err)
127 goto out_ino;
128
129 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
130 inode->i_nlink = le32_to_cpu(ino->nlink);
131 inode->i_uid = le32_to_cpu(ino->uid);
132 inode->i_gid = le32_to_cpu(ino->gid);
133 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
134 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
135 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
136 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
137 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
138 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
139 inode->i_mode = le32_to_cpu(ino->mode);
140 inode->i_size = le64_to_cpu(ino->size);
141
142 ui->data_len = le32_to_cpu(ino->data_len);
143 ui->flags = le32_to_cpu(ino->flags);
144 ui->compr_type = le16_to_cpu(ino->compr_type);
145 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
146 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
147 ui->xattr_size = le32_to_cpu(ino->xattr_size);
148 ui->xattr_names = le32_to_cpu(ino->xattr_names);
149 ui->synced_i_size = ui->ui_size = inode->i_size;
150
151 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
152
153 err = validate_inode(c, inode);
154 if (err)
155 goto out_invalid;
156
Artem Bityutskiy0a883a02008-08-13 14:13:26 +0300157 /* Disable read-ahead */
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300158 inode->i_mapping->backing_dev_info = &c->bdi;
159
160 switch (inode->i_mode & S_IFMT) {
161 case S_IFREG:
162 inode->i_mapping->a_ops = &ubifs_file_address_operations;
163 inode->i_op = &ubifs_file_inode_operations;
164 inode->i_fop = &ubifs_file_operations;
165 if (ui->xattr) {
166 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
167 if (!ui->data) {
168 err = -ENOMEM;
169 goto out_ino;
170 }
171 memcpy(ui->data, ino->data, ui->data_len);
172 ((char *)ui->data)[ui->data_len] = '\0';
173 } else if (ui->data_len != 0) {
174 err = 10;
175 goto out_invalid;
176 }
177 break;
178 case S_IFDIR:
179 inode->i_op = &ubifs_dir_inode_operations;
180 inode->i_fop = &ubifs_dir_operations;
181 if (ui->data_len != 0) {
182 err = 11;
183 goto out_invalid;
184 }
185 break;
186 case S_IFLNK:
187 inode->i_op = &ubifs_symlink_inode_operations;
188 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
189 err = 12;
190 goto out_invalid;
191 }
192 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
193 if (!ui->data) {
194 err = -ENOMEM;
195 goto out_ino;
196 }
197 memcpy(ui->data, ino->data, ui->data_len);
198 ((char *)ui->data)[ui->data_len] = '\0';
199 break;
200 case S_IFBLK:
201 case S_IFCHR:
202 {
203 dev_t rdev;
204 union ubifs_dev_desc *dev;
205
206 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
207 if (!ui->data) {
208 err = -ENOMEM;
209 goto out_ino;
210 }
211
212 dev = (union ubifs_dev_desc *)ino->data;
213 if (ui->data_len == sizeof(dev->new))
214 rdev = new_decode_dev(le32_to_cpu(dev->new));
215 else if (ui->data_len == sizeof(dev->huge))
216 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
217 else {
218 err = 13;
219 goto out_invalid;
220 }
221 memcpy(ui->data, ino->data, ui->data_len);
222 inode->i_op = &ubifs_file_inode_operations;
223 init_special_inode(inode, inode->i_mode, rdev);
224 break;
225 }
226 case S_IFSOCK:
227 case S_IFIFO:
228 inode->i_op = &ubifs_file_inode_operations;
229 init_special_inode(inode, inode->i_mode, 0);
230 if (ui->data_len != 0) {
231 err = 14;
232 goto out_invalid;
233 }
234 break;
235 default:
236 err = 15;
237 goto out_invalid;
238 }
239
240 kfree(ino);
241 ubifs_set_inode_flags(inode);
242 unlock_new_inode(inode);
243 return inode;
244
245out_invalid:
246 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
247 dbg_dump_node(c, ino);
248 dbg_dump_inode(c, inode);
249 err = -EINVAL;
250out_ino:
251 kfree(ino);
252out:
253 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
254 iget_failed(inode);
255 return ERR_PTR(err);
256}
257
258static struct inode *ubifs_alloc_inode(struct super_block *sb)
259{
260 struct ubifs_inode *ui;
261
262 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
263 if (!ui)
264 return NULL;
265
266 memset((void *)ui + sizeof(struct inode), 0,
267 sizeof(struct ubifs_inode) - sizeof(struct inode));
268 mutex_init(&ui->ui_mutex);
269 spin_lock_init(&ui->ui_lock);
270 return &ui->vfs_inode;
271};
272
273static void ubifs_destroy_inode(struct inode *inode)
274{
275 struct ubifs_inode *ui = ubifs_inode(inode);
276
277 kfree(ui->data);
278 kmem_cache_free(ubifs_inode_slab, inode);
279}
280
281/*
282 * Note, Linux write-back code calls this without 'i_mutex'.
283 */
284static int ubifs_write_inode(struct inode *inode, int wait)
285{
Artem Bityutskiyfbfa6c82008-07-22 11:52:52 +0300286 int err = 0;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300287 struct ubifs_info *c = inode->i_sb->s_fs_info;
288 struct ubifs_inode *ui = ubifs_inode(inode);
289
290 ubifs_assert(!ui->xattr);
291 if (is_bad_inode(inode))
292 return 0;
293
294 mutex_lock(&ui->ui_mutex);
295 /*
296 * Due to races between write-back forced by budgeting
297 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
298 * have already been synchronized, do not do this again. This might
299 * also happen if it was synchronized in an VFS operation, e.g.
300 * 'ubifs_link()'.
301 */
302 if (!ui->dirty) {
303 mutex_unlock(&ui->ui_mutex);
304 return 0;
305 }
306
Artem Bityutskiyfbfa6c82008-07-22 11:52:52 +0300307 /*
308 * As an optimization, do not write orphan inodes to the media just
309 * because this is not needed.
310 */
311 dbg_gen("inode %lu, mode %#x, nlink %u",
312 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
313 if (inode->i_nlink) {
Artem Bityutskiy1f286812008-07-22 12:06:13 +0300314 err = ubifs_jnl_write_inode(c, inode);
Artem Bityutskiyfbfa6c82008-07-22 11:52:52 +0300315 if (err)
316 ubifs_err("can't write inode %lu, error %d",
317 inode->i_ino, err);
318 }
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300319
320 ui->dirty = 0;
321 mutex_unlock(&ui->ui_mutex);
322 ubifs_release_dirty_inode_budget(c, ui);
323 return err;
324}
325
326static void ubifs_delete_inode(struct inode *inode)
327{
328 int err;
329 struct ubifs_info *c = inode->i_sb->s_fs_info;
Artem Bityutskiy1e0f3582008-07-21 10:59:53 +0300330 struct ubifs_inode *ui = ubifs_inode(inode);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300331
Artem Bityutskiy1e0f3582008-07-21 10:59:53 +0300332 if (ui->xattr)
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300333 /*
334 * Extended attribute inode deletions are fully handled in
335 * 'ubifs_removexattr()'. These inodes are special and have
336 * limited usage, so there is nothing to do here.
337 */
338 goto out;
339
Artem Bityutskiy7d32c2b2008-07-18 18:54:29 +0300340 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300341 ubifs_assert(!atomic_read(&inode->i_count));
342 ubifs_assert(inode->i_nlink == 0);
343
344 truncate_inode_pages(&inode->i_data, 0);
345 if (is_bad_inode(inode))
346 goto out;
347
Artem Bityutskiy1e0f3582008-07-21 10:59:53 +0300348 ui->ui_size = inode->i_size = 0;
Artem Bityutskiyde94eb52008-07-22 13:06:20 +0300349 err = ubifs_jnl_delete_inode(c, inode);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300350 if (err)
351 /*
352 * Worst case we have a lost orphan inode wasting space, so a
Artem Bityutskiy0a883a02008-08-13 14:13:26 +0300353 * simple error message is OK here.
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300354 */
Artem Bityutskiyde94eb52008-07-22 13:06:20 +0300355 ubifs_err("can't delete inode %lu, error %d",
356 inode->i_ino, err);
357
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300358out:
Artem Bityutskiy1e0f3582008-07-21 10:59:53 +0300359 if (ui->dirty)
360 ubifs_release_dirty_inode_budget(c, ui);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300361 clear_inode(inode);
362}
363
364static void ubifs_dirty_inode(struct inode *inode)
365{
366 struct ubifs_inode *ui = ubifs_inode(inode);
367
368 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
369 if (!ui->dirty) {
370 ui->dirty = 1;
371 dbg_gen("inode %lu", inode->i_ino);
372 }
373}
374
375static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
376{
377 struct ubifs_info *c = dentry->d_sb->s_fs_info;
378 unsigned long long free;
Artem Bityutskiy7c7cbad2008-09-03 14:16:42 +0300379 __le32 *uuid = (__le32 *)c->uuid;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300380
Artem Bityutskiy7dad1812008-08-25 18:58:19 +0300381 free = ubifs_get_free_space(c);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300382 dbg_gen("free space %lld bytes (%lld blocks)",
383 free, free >> UBIFS_BLOCK_SHIFT);
384
385 buf->f_type = UBIFS_SUPER_MAGIC;
386 buf->f_bsize = UBIFS_BLOCK_SIZE;
387 buf->f_blocks = c->block_cnt;
388 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
389 if (free > c->report_rp_size)
390 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
391 else
392 buf->f_bavail = 0;
393 buf->f_files = 0;
394 buf->f_ffree = 0;
395 buf->f_namelen = UBIFS_MAX_NLEN;
Artem Bityutskiy7c7cbad2008-09-03 14:16:42 +0300396 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
397 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300398 return 0;
399}
400
401static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
402{
403 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
404
405 if (c->mount_opts.unmount_mode == 2)
406 seq_printf(s, ",fast_unmount");
407 else if (c->mount_opts.unmount_mode == 1)
408 seq_printf(s, ",norm_unmount");
409
Adrian Hunter4793e7c2008-09-02 16:29:46 +0300410 if (c->mount_opts.bulk_read == 2)
411 seq_printf(s, ",bulk_read");
412 else if (c->mount_opts.bulk_read == 1)
413 seq_printf(s, ",no_bulk_read");
414
Adrian Hunter2953e732008-09-04 16:26:00 +0300415 if (c->mount_opts.chk_data_crc == 2)
416 seq_printf(s, ",chk_data_crc");
417 else if (c->mount_opts.chk_data_crc == 1)
418 seq_printf(s, ",no_chk_data_crc");
419
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300420 return 0;
421}
422
423static int ubifs_sync_fs(struct super_block *sb, int wait)
424{
425 struct ubifs_info *c = sb->s_fs_info;
426 int i, ret = 0, err;
Artem Bityutskiy403e12a2008-09-09 12:31:37 +0300427 long long bud_bytes;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300428
Adrian Hunterbed79932008-09-11 14:25:44 +0300429 if (c->jheads) {
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300430 for (i = 0; i < c->jhead_cnt; i++) {
431 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
432 if (err && !ret)
433 ret = err;
434 }
Artem Bityutskiy403e12a2008-09-09 12:31:37 +0300435
Adrian Hunterbed79932008-09-11 14:25:44 +0300436 /* Commit the journal unless it has too little data */
437 spin_lock(&c->buds_lock);
438 bud_bytes = c->bud_bytes;
439 spin_unlock(&c->buds_lock);
440 if (bud_bytes > c->leb_size) {
441 err = ubifs_run_commit(c);
442 if (err)
443 return err;
444 }
Artem Bityutskiy403e12a2008-09-09 12:31:37 +0300445 }
446
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300447 /*
448 * We ought to call sync for c->ubi but it does not have one. If it had
449 * it would in turn call mtd->sync, however mtd operations are
450 * synchronous anyway, so we don't lose any sleep here.
451 */
452 return ret;
453}
454
455/**
456 * init_constants_early - initialize UBIFS constants.
457 * @c: UBIFS file-system description object
458 *
459 * This function initialize UBIFS constants which do not need the superblock to
460 * be read. It also checks that the UBI volume satisfies basic UBIFS
461 * requirements. Returns zero in case of success and a negative error code in
462 * case of failure.
463 */
464static int init_constants_early(struct ubifs_info *c)
465{
466 if (c->vi.corrupted) {
467 ubifs_warn("UBI volume is corrupted - read-only mode");
468 c->ro_media = 1;
469 }
470
471 if (c->di.ro_mode) {
472 ubifs_msg("read-only UBI device");
473 c->ro_media = 1;
474 }
475
476 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
477 ubifs_msg("static UBI volume - read-only mode");
478 c->ro_media = 1;
479 }
480
481 c->leb_cnt = c->vi.size;
482 c->leb_size = c->vi.usable_leb_size;
483 c->half_leb_size = c->leb_size / 2;
484 c->min_io_size = c->di.min_io_size;
485 c->min_io_shift = fls(c->min_io_size) - 1;
486
487 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
488 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
489 c->leb_size, UBIFS_MIN_LEB_SZ);
490 return -EINVAL;
491 }
492
493 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
494 ubifs_err("too few LEBs (%d), min. is %d",
495 c->leb_cnt, UBIFS_MIN_LEB_CNT);
496 return -EINVAL;
497 }
498
499 if (!is_power_of_2(c->min_io_size)) {
500 ubifs_err("bad min. I/O size %d", c->min_io_size);
501 return -EINVAL;
502 }
503
504 /*
505 * UBIFS aligns all node to 8-byte boundary, so to make function in
506 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
507 * less than 8.
508 */
509 if (c->min_io_size < 8) {
510 c->min_io_size = 8;
511 c->min_io_shift = 3;
512 }
513
514 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
515 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
516
517 /*
518 * Initialize node length ranges which are mostly needed for node
519 * length validation.
520 */
521 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
522 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
523 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
524 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
525 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
526 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
527
528 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
529 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
530 c->ranges[UBIFS_ORPH_NODE].min_len =
531 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
532 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
533 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
534 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
535 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
536 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
537 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
538 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
539 /*
540 * Minimum indexing node size is amended later when superblock is
541 * read and the key length is known.
542 */
543 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
544 /*
545 * Maximum indexing node size is amended later when superblock is
546 * read and the fanout is known.
547 */
548 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
549
550 /*
551 * Initialize dead and dark LEB space watermarks.
552 *
553 * Dead space is the space which cannot be used. Its watermark is
554 * equivalent to min. I/O unit or minimum node size if it is greater
555 * then min. I/O unit.
556 *
557 * Dark space is the space which might be used, or might not, depending
558 * on which node should be written to the LEB. Its watermark is
559 * equivalent to maximum UBIFS node size.
560 */
561 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
562 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
563
Artem Bityutskiy9bbb5722008-08-22 18:23:22 +0300564 /*
565 * Calculate how many bytes would be wasted at the end of LEB if it was
566 * fully filled with data nodes of maximum size. This is used in
567 * calculations when reporting free space.
568 */
569 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
Artem Bityutskiy39ce81c2008-11-18 18:09:49 +0200570
Adrian Hunter4793e7c2008-09-02 16:29:46 +0300571 /* Buffer size for bulk-reads */
Artem Bityutskiy6c0c42c2008-11-18 20:20:05 +0200572 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
573 if (c->max_bu_buf_len > c->leb_size)
574 c->max_bu_buf_len = c->leb_size;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300575 return 0;
576}
577
578/**
579 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
580 * @c: UBIFS file-system description object
581 * @lnum: LEB the write-buffer was synchronized to
582 * @free: how many free bytes left in this LEB
583 * @pad: how many bytes were padded
584 *
585 * This is a callback function which is called by the I/O unit when the
586 * write-buffer is synchronized. We need this to correctly maintain space
587 * accounting in bud logical eraseblocks. This function returns zero in case of
588 * success and a negative error code in case of failure.
589 *
590 * This function actually belongs to the journal, but we keep it here because
591 * we want to keep it static.
592 */
593static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
594{
595 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
596}
597
598/*
599 * init_constants_late - initialize UBIFS constants.
600 * @c: UBIFS file-system description object
601 *
602 * This is a helper function which initializes various UBIFS constants after
603 * the superblock has been read. It also checks various UBIFS parameters and
604 * makes sure they are all right. Returns zero in case of success and a
605 * negative error code in case of failure.
606 */
607static int init_constants_late(struct ubifs_info *c)
608{
609 int tmp, err;
610 uint64_t tmp64;
611
612 c->main_bytes = (long long)c->main_lebs * c->leb_size;
613 c->max_znode_sz = sizeof(struct ubifs_znode) +
614 c->fanout * sizeof(struct ubifs_zbranch);
615
616 tmp = ubifs_idx_node_sz(c, 1);
617 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
618 c->min_idx_node_sz = ALIGN(tmp, 8);
619
620 tmp = ubifs_idx_node_sz(c, c->fanout);
621 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
622 c->max_idx_node_sz = ALIGN(tmp, 8);
623
624 /* Make sure LEB size is large enough to fit full commit */
625 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
626 tmp = ALIGN(tmp, c->min_io_size);
627 if (tmp > c->leb_size) {
628 dbg_err("too small LEB size %d, at least %d needed",
629 c->leb_size, tmp);
630 return -EINVAL;
631 }
632
633 /*
634 * Make sure that the log is large enough to fit reference nodes for
635 * all buds plus one reserved LEB.
636 */
637 tmp64 = c->max_bud_bytes;
638 tmp = do_div(tmp64, c->leb_size);
639 c->max_bud_cnt = tmp64 + !!tmp;
640 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
641 tmp /= c->leb_size;
642 tmp += 1;
643 if (c->log_lebs < tmp) {
644 dbg_err("too small log %d LEBs, required min. %d LEBs",
645 c->log_lebs, tmp);
646 return -EINVAL;
647 }
648
649 /*
650 * When budgeting we assume worst-case scenarios when the pages are not
651 * be compressed and direntries are of the maximum size.
652 *
653 * Note, data, which may be stored in inodes is budgeted separately, so
654 * it is not included into 'c->inode_budget'.
655 */
656 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
657 c->inode_budget = UBIFS_INO_NODE_SZ;
658 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
659
660 /*
661 * When the amount of flash space used by buds becomes
662 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
663 * The writers are unblocked when the commit is finished. To avoid
664 * writers to be blocked UBIFS initiates background commit in advance,
665 * when number of bud bytes becomes above the limit defined below.
666 */
667 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
668
669 /*
670 * Ensure minimum journal size. All the bytes in the journal heads are
671 * considered to be used, when calculating the current journal usage.
672 * Consequently, if the journal is too small, UBIFS will treat it as
673 * always full.
674 */
675 tmp64 = (uint64_t)(c->jhead_cnt + 1) * c->leb_size + 1;
676 if (c->bg_bud_bytes < tmp64)
677 c->bg_bud_bytes = tmp64;
678 if (c->max_bud_bytes < tmp64 + c->leb_size)
679 c->max_bud_bytes = tmp64 + c->leb_size;
680
681 err = ubifs_calc_lpt_geom(c);
682 if (err)
683 return err;
684
685 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
686
687 /*
688 * Calculate total amount of FS blocks. This number is not used
689 * internally because it does not make much sense for UBIFS, but it is
690 * necessary to report something for the 'statfs()' call.
691 *
Artem Bityutskiy7dad1812008-08-25 18:58:19 +0300692 * Subtract the LEB reserved for GC, the LEB which is reserved for
693 * deletions, and assume only one journal head is available.
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300694 */
Artem Bityutskiy7dad1812008-08-25 18:58:19 +0300695 tmp64 = c->main_lebs - 2 - c->jhead_cnt + 1;
696 tmp64 *= (uint64_t)c->leb_size - c->leb_overhead;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300697 tmp64 = ubifs_reported_space(c, tmp64);
698 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
699
700 return 0;
701}
702
703/**
704 * take_gc_lnum - reserve GC LEB.
705 * @c: UBIFS file-system description object
706 *
707 * This function ensures that the LEB reserved for garbage collection is
708 * unmapped and is marked as "taken" in lprops. We also have to set free space
709 * to LEB size and dirty space to zero, because lprops may contain out-of-date
710 * information if the file-system was un-mounted before it has been committed.
711 * This function returns zero in case of success and a negative error code in
712 * case of failure.
713 */
714static int take_gc_lnum(struct ubifs_info *c)
715{
716 int err;
717
718 if (c->gc_lnum == -1) {
719 ubifs_err("no LEB for GC");
720 return -EINVAL;
721 }
722
723 err = ubifs_leb_unmap(c, c->gc_lnum);
724 if (err)
725 return err;
726
727 /* And we have to tell lprops that this LEB is taken */
728 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
729 LPROPS_TAKEN, 0, 0);
730 return err;
731}
732
733/**
734 * alloc_wbufs - allocate write-buffers.
735 * @c: UBIFS file-system description object
736 *
737 * This helper function allocates and initializes UBIFS write-buffers. Returns
738 * zero in case of success and %-ENOMEM in case of failure.
739 */
740static int alloc_wbufs(struct ubifs_info *c)
741{
742 int i, err;
743
744 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
745 GFP_KERNEL);
746 if (!c->jheads)
747 return -ENOMEM;
748
749 /* Initialize journal heads */
750 for (i = 0; i < c->jhead_cnt; i++) {
751 INIT_LIST_HEAD(&c->jheads[i].buds_list);
752 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
753 if (err)
754 return err;
755
756 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
757 c->jheads[i].wbuf.jhead = i;
758 }
759
760 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
761 /*
762 * Garbage Collector head likely contains long-term data and
763 * does not need to be synchronized by timer.
764 */
765 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
766 c->jheads[GCHD].wbuf.timeout = 0;
767
768 return 0;
769}
770
771/**
772 * free_wbufs - free write-buffers.
773 * @c: UBIFS file-system description object
774 */
775static void free_wbufs(struct ubifs_info *c)
776{
777 int i;
778
779 if (c->jheads) {
780 for (i = 0; i < c->jhead_cnt; i++) {
781 kfree(c->jheads[i].wbuf.buf);
782 kfree(c->jheads[i].wbuf.inodes);
783 }
784 kfree(c->jheads);
785 c->jheads = NULL;
786 }
787}
788
789/**
790 * free_orphans - free orphans.
791 * @c: UBIFS file-system description object
792 */
793static void free_orphans(struct ubifs_info *c)
794{
795 struct ubifs_orphan *orph;
796
797 while (c->orph_dnext) {
798 orph = c->orph_dnext;
799 c->orph_dnext = orph->dnext;
800 list_del(&orph->list);
801 kfree(orph);
802 }
803
804 while (!list_empty(&c->orph_list)) {
805 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
806 list_del(&orph->list);
807 kfree(orph);
808 dbg_err("orphan list not empty at unmount");
809 }
810
811 vfree(c->orph_buf);
812 c->orph_buf = NULL;
813}
814
815/**
816 * free_buds - free per-bud objects.
817 * @c: UBIFS file-system description object
818 */
819static void free_buds(struct ubifs_info *c)
820{
821 struct rb_node *this = c->buds.rb_node;
822 struct ubifs_bud *bud;
823
824 while (this) {
825 if (this->rb_left)
826 this = this->rb_left;
827 else if (this->rb_right)
828 this = this->rb_right;
829 else {
830 bud = rb_entry(this, struct ubifs_bud, rb);
831 this = rb_parent(this);
832 if (this) {
833 if (this->rb_left == &bud->rb)
834 this->rb_left = NULL;
835 else
836 this->rb_right = NULL;
837 }
838 kfree(bud);
839 }
840 }
841}
842
843/**
844 * check_volume_empty - check if the UBI volume is empty.
845 * @c: UBIFS file-system description object
846 *
847 * This function checks if the UBIFS volume is empty by looking if its LEBs are
848 * mapped or not. The result of checking is stored in the @c->empty variable.
849 * Returns zero in case of success and a negative error code in case of
850 * failure.
851 */
852static int check_volume_empty(struct ubifs_info *c)
853{
854 int lnum, err;
855
856 c->empty = 1;
857 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
858 err = ubi_is_mapped(c->ubi, lnum);
859 if (unlikely(err < 0))
860 return err;
861 if (err == 1) {
862 c->empty = 0;
863 break;
864 }
865
866 cond_resched();
867 }
868
869 return 0;
870}
871
872/*
873 * UBIFS mount options.
874 *
875 * Opt_fast_unmount: do not run a journal commit before un-mounting
876 * Opt_norm_unmount: run a journal commit before un-mounting
Adrian Hunter4793e7c2008-09-02 16:29:46 +0300877 * Opt_bulk_read: enable bulk-reads
878 * Opt_no_bulk_read: disable bulk-reads
Adrian Hunter2953e732008-09-04 16:26:00 +0300879 * Opt_chk_data_crc: check CRCs when reading data nodes
880 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300881 * Opt_err: just end of array marker
882 */
883enum {
884 Opt_fast_unmount,
885 Opt_norm_unmount,
Adrian Hunter4793e7c2008-09-02 16:29:46 +0300886 Opt_bulk_read,
887 Opt_no_bulk_read,
Adrian Hunter2953e732008-09-04 16:26:00 +0300888 Opt_chk_data_crc,
889 Opt_no_chk_data_crc,
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300890 Opt_err,
891};
892
Steven Whitehousea447c092008-10-13 10:46:57 +0100893static const match_table_t tokens = {
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300894 {Opt_fast_unmount, "fast_unmount"},
895 {Opt_norm_unmount, "norm_unmount"},
Adrian Hunter4793e7c2008-09-02 16:29:46 +0300896 {Opt_bulk_read, "bulk_read"},
897 {Opt_no_bulk_read, "no_bulk_read"},
Adrian Hunter2953e732008-09-04 16:26:00 +0300898 {Opt_chk_data_crc, "chk_data_crc"},
899 {Opt_no_chk_data_crc, "no_chk_data_crc"},
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300900 {Opt_err, NULL},
901};
902
903/**
904 * ubifs_parse_options - parse mount parameters.
905 * @c: UBIFS file-system description object
906 * @options: parameters to parse
907 * @is_remount: non-zero if this is FS re-mount
908 *
909 * This function parses UBIFS mount options and returns zero in case success
910 * and a negative error code in case of failure.
911 */
912static int ubifs_parse_options(struct ubifs_info *c, char *options,
913 int is_remount)
914{
915 char *p;
916 substring_t args[MAX_OPT_ARGS];
917
918 if (!options)
919 return 0;
920
921 while ((p = strsep(&options, ","))) {
922 int token;
923
924 if (!*p)
925 continue;
926
927 token = match_token(p, tokens, args);
928 switch (token) {
929 case Opt_fast_unmount:
930 c->mount_opts.unmount_mode = 2;
931 c->fast_unmount = 1;
932 break;
933 case Opt_norm_unmount:
934 c->mount_opts.unmount_mode = 1;
935 c->fast_unmount = 0;
936 break;
Adrian Hunter4793e7c2008-09-02 16:29:46 +0300937 case Opt_bulk_read:
938 c->mount_opts.bulk_read = 2;
939 c->bulk_read = 1;
940 break;
941 case Opt_no_bulk_read:
942 c->mount_opts.bulk_read = 1;
943 c->bulk_read = 0;
944 break;
Adrian Hunter2953e732008-09-04 16:26:00 +0300945 case Opt_chk_data_crc:
946 c->mount_opts.chk_data_crc = 2;
947 c->no_chk_data_crc = 0;
948 break;
949 case Opt_no_chk_data_crc:
950 c->mount_opts.chk_data_crc = 1;
951 c->no_chk_data_crc = 1;
952 break;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300953 default:
954 ubifs_err("unrecognized mount option \"%s\" "
955 "or missing value", p);
956 return -EINVAL;
957 }
958 }
959
960 return 0;
961}
962
963/**
964 * destroy_journal - destroy journal data structures.
965 * @c: UBIFS file-system description object
966 *
967 * This function destroys journal data structures including those that may have
968 * been created by recovery functions.
969 */
970static void destroy_journal(struct ubifs_info *c)
971{
972 while (!list_empty(&c->unclean_leb_list)) {
973 struct ubifs_unclean_leb *ucleb;
974
975 ucleb = list_entry(c->unclean_leb_list.next,
976 struct ubifs_unclean_leb, list);
977 list_del(&ucleb->list);
978 kfree(ucleb);
979 }
980 while (!list_empty(&c->old_buds)) {
981 struct ubifs_bud *bud;
982
983 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
984 list_del(&bud->list);
985 kfree(bud);
986 }
987 ubifs_destroy_idx_gc(c);
988 ubifs_destroy_size_tree(c);
989 ubifs_tnc_close(c);
990 free_buds(c);
991}
992
993/**
Artem Bityutskiy3477d202008-11-19 11:53:15 +0200994 * bu_init - initialize bulk-read information.
995 * @c: UBIFS file-system description object
996 */
997static void bu_init(struct ubifs_info *c)
998{
999 ubifs_assert(c->bulk_read == 1);
1000
1001 if (c->bu.buf)
1002 return; /* Already initialized */
1003
1004again:
1005 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1006 if (!c->bu.buf) {
1007 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1008 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1009 goto again;
1010 }
1011
1012 /* Just disable bulk-read */
1013 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1014 "disabling it", c->max_bu_buf_len);
1015 c->mount_opts.bulk_read = 1;
1016 c->bulk_read = 0;
1017 return;
1018 }
1019}
1020
1021/**
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001022 * mount_ubifs - mount UBIFS file-system.
1023 * @c: UBIFS file-system description object
1024 *
1025 * This function mounts UBIFS file system. Returns zero in case of success and
1026 * a negative error code in case of failure.
1027 *
1028 * Note, the function does not de-allocate resources it it fails half way
1029 * through, and the caller has to do this instead.
1030 */
1031static int mount_ubifs(struct ubifs_info *c)
1032{
1033 struct super_block *sb = c->vfs_sb;
1034 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1035 long long x;
1036 size_t sz;
1037
1038 err = init_constants_early(c);
1039 if (err)
1040 return err;
1041
1042#ifdef CONFIG_UBIFS_FS_DEBUG
1043 c->dbg_buf = vmalloc(c->leb_size);
1044 if (!c->dbg_buf)
1045 return -ENOMEM;
1046#endif
1047
1048 err = check_volume_empty(c);
1049 if (err)
1050 goto out_free;
1051
1052 if (c->empty && (mounted_read_only || c->ro_media)) {
1053 /*
1054 * This UBI volume is empty, and read-only, or the file system
1055 * is mounted read-only - we cannot format it.
1056 */
1057 ubifs_err("can't format empty UBI volume: read-only %s",
1058 c->ro_media ? "UBI volume" : "mount");
1059 err = -EROFS;
1060 goto out_free;
1061 }
1062
1063 if (c->ro_media && !mounted_read_only) {
1064 ubifs_err("cannot mount read-write - read-only media");
1065 err = -EROFS;
1066 goto out_free;
1067 }
1068
1069 /*
1070 * The requirement for the buffer is that it should fit indexing B-tree
1071 * height amount of integers. We assume the height if the TNC tree will
1072 * never exceed 64.
1073 */
1074 err = -ENOMEM;
1075 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1076 if (!c->bottom_up_buf)
1077 goto out_free;
1078
1079 c->sbuf = vmalloc(c->leb_size);
1080 if (!c->sbuf)
1081 goto out_free;
1082
1083 if (!mounted_read_only) {
1084 c->ileb_buf = vmalloc(c->leb_size);
1085 if (!c->ileb_buf)
1086 goto out_free;
1087 }
1088
Artem Bityutskiy3477d202008-11-19 11:53:15 +02001089 if (c->bulk_read == 1)
1090 bu_init(c);
1091
1092 /*
1093 * We have to check all CRCs, even for data nodes, when we mount the FS
1094 * (specifically, when we are replaying).
1095 */
Adrian Hunter2953e732008-09-04 16:26:00 +03001096 c->always_chk_crc = 1;
1097
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001098 err = ubifs_read_superblock(c);
1099 if (err)
1100 goto out_free;
1101
1102 /*
1103 * Make sure the compressor which is set as the default on in the
1104 * superblock was actually compiled in.
1105 */
1106 if (!ubifs_compr_present(c->default_compr)) {
1107 ubifs_warn("'%s' compressor is set by superblock, but not "
1108 "compiled in", ubifs_compr_name(c->default_compr));
1109 c->default_compr = UBIFS_COMPR_NONE;
1110 }
1111
1112 dbg_failure_mode_registration(c);
1113
1114 err = init_constants_late(c);
1115 if (err)
1116 goto out_dereg;
1117
1118 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1119 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1120 c->cbuf = kmalloc(sz, GFP_NOFS);
1121 if (!c->cbuf) {
1122 err = -ENOMEM;
1123 goto out_dereg;
1124 }
1125
Sebastian Siewior0855f312008-09-09 11:17:29 +02001126 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001127 if (!mounted_read_only) {
1128 err = alloc_wbufs(c);
1129 if (err)
1130 goto out_cbuf;
1131
1132 /* Create background thread */
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001133 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001134 if (IS_ERR(c->bgt)) {
1135 err = PTR_ERR(c->bgt);
1136 c->bgt = NULL;
1137 ubifs_err("cannot spawn \"%s\", error %d",
1138 c->bgt_name, err);
1139 goto out_wbufs;
1140 }
1141 wake_up_process(c->bgt);
1142 }
1143
1144 err = ubifs_read_master(c);
1145 if (err)
1146 goto out_master;
1147
1148 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1149 ubifs_msg("recovery needed");
1150 c->need_recovery = 1;
1151 if (!mounted_read_only) {
1152 err = ubifs_recover_inl_heads(c, c->sbuf);
1153 if (err)
1154 goto out_master;
1155 }
1156 } else if (!mounted_read_only) {
1157 /*
1158 * Set the "dirty" flag so that if we reboot uncleanly we
1159 * will notice this immediately on the next mount.
1160 */
1161 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1162 err = ubifs_write_master(c);
1163 if (err)
1164 goto out_master;
1165 }
1166
1167 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1168 if (err)
1169 goto out_lpt;
1170
1171 err = dbg_check_idx_size(c, c->old_idx_sz);
1172 if (err)
1173 goto out_lpt;
1174
1175 err = ubifs_replay_journal(c);
1176 if (err)
1177 goto out_journal;
1178
1179 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1180 if (err)
1181 goto out_orphans;
1182
1183 if (!mounted_read_only) {
1184 int lnum;
1185
1186 /* Check for enough free space */
1187 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1188 ubifs_err("insufficient available space");
1189 err = -EINVAL;
1190 goto out_orphans;
1191 }
1192
1193 /* Check for enough log space */
1194 lnum = c->lhead_lnum + 1;
1195 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1196 lnum = UBIFS_LOG_LNUM;
1197 if (lnum == c->ltail_lnum) {
1198 err = ubifs_consolidate_log(c);
1199 if (err)
1200 goto out_orphans;
1201 }
1202
1203 if (c->need_recovery) {
1204 err = ubifs_recover_size(c);
1205 if (err)
1206 goto out_orphans;
1207 err = ubifs_rcvry_gc_commit(c);
1208 } else
1209 err = take_gc_lnum(c);
1210 if (err)
1211 goto out_orphans;
1212
1213 err = dbg_check_lprops(c);
1214 if (err)
1215 goto out_orphans;
1216 } else if (c->need_recovery) {
1217 err = ubifs_recover_size(c);
1218 if (err)
1219 goto out_orphans;
1220 }
1221
1222 spin_lock(&ubifs_infos_lock);
1223 list_add_tail(&c->infos_list, &ubifs_infos);
1224 spin_unlock(&ubifs_infos_lock);
1225
1226 if (c->need_recovery) {
1227 if (mounted_read_only)
1228 ubifs_msg("recovery deferred");
1229 else {
1230 c->need_recovery = 0;
1231 ubifs_msg("recovery completed");
1232 }
1233 }
1234
1235 err = dbg_check_filesystem(c);
1236 if (err)
1237 goto out_infos;
1238
Adrian Hunter2953e732008-09-04 16:26:00 +03001239 c->always_chk_crc = 0;
1240
Artem Bityutskiyce769ca2008-07-18 12:54:21 +03001241 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1242 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001243 if (mounted_read_only)
1244 ubifs_msg("mounted read-only");
1245 x = (long long)c->main_lebs * c->leb_size;
Artem Bityutskiy948cfb22008-08-20 11:56:33 +03001246 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1247 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001248 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
Artem Bityutskiy948cfb22008-08-20 11:56:33 +03001249 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1250 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1251 ubifs_msg("media format: %d (latest is %d)",
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001252 c->fmt_version, UBIFS_FORMAT_VERSION);
Artem Bityutskiy948cfb22008-08-20 11:56:33 +03001253 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
Artem Bityutskiyfae7fb22008-10-17 18:49:23 +03001254 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
Artem Bityutskiy948cfb22008-08-20 11:56:33 +03001255 c->report_rp_size, c->report_rp_size >> 10);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001256
1257 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1258 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1259 dbg_msg("LEB size: %d bytes (%d KiB)",
Artem Bityutskiy948cfb22008-08-20 11:56:33 +03001260 c->leb_size, c->leb_size >> 10);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001261 dbg_msg("data journal heads: %d",
1262 c->jhead_cnt - NONDATA_JHEADS_CNT);
1263 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1264 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1265 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1266 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1267 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1268 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1269 dbg_msg("fast unmount: %d", c->fast_unmount);
1270 dbg_msg("big_lpt %d", c->big_lpt);
1271 dbg_msg("log LEBs: %d (%d - %d)",
1272 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1273 dbg_msg("LPT area LEBs: %d (%d - %d)",
1274 c->lpt_lebs, c->lpt_first, c->lpt_last);
1275 dbg_msg("orphan area LEBs: %d (%d - %d)",
1276 c->orph_lebs, c->orph_first, c->orph_last);
1277 dbg_msg("main area LEBs: %d (%d - %d)",
1278 c->main_lebs, c->main_first, c->leb_cnt - 1);
1279 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1280 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1281 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1282 dbg_msg("key hash type: %d", c->key_hash_type);
1283 dbg_msg("tree fanout: %d", c->fanout);
1284 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1285 dbg_msg("first main LEB: %d", c->main_first);
1286 dbg_msg("dead watermark: %d", c->dead_wm);
1287 dbg_msg("dark watermark: %d", c->dark_wm);
1288 x = (long long)c->main_lebs * c->dark_wm;
1289 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1290 x, x >> 10, x >> 20);
1291 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1292 c->max_bud_bytes, c->max_bud_bytes >> 10,
1293 c->max_bud_bytes >> 20);
1294 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1295 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1296 c->bg_bud_bytes >> 20);
1297 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1298 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1299 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1300 dbg_msg("commit number: %llu", c->cmt_no);
1301
1302 return 0;
1303
1304out_infos:
1305 spin_lock(&ubifs_infos_lock);
1306 list_del(&c->infos_list);
1307 spin_unlock(&ubifs_infos_lock);
1308out_orphans:
1309 free_orphans(c);
1310out_journal:
1311 destroy_journal(c);
1312out_lpt:
1313 ubifs_lpt_free(c, 0);
1314out_master:
1315 kfree(c->mst_node);
1316 kfree(c->rcvrd_mst_node);
1317 if (c->bgt)
1318 kthread_stop(c->bgt);
1319out_wbufs:
1320 free_wbufs(c);
1321out_cbuf:
1322 kfree(c->cbuf);
1323out_dereg:
1324 dbg_failure_mode_deregistration(c);
1325out_free:
Artem Bityutskiy3477d202008-11-19 11:53:15 +02001326 kfree(c->bu.buf);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001327 vfree(c->ileb_buf);
1328 vfree(c->sbuf);
1329 kfree(c->bottom_up_buf);
1330 UBIFS_DBG(vfree(c->dbg_buf));
1331 return err;
1332}
1333
1334/**
1335 * ubifs_umount - un-mount UBIFS file-system.
1336 * @c: UBIFS file-system description object
1337 *
1338 * Note, this function is called to free allocated resourced when un-mounting,
1339 * as well as free resources when an error occurred while we were half way
1340 * through mounting (error path cleanup function). So it has to make sure the
1341 * resource was actually allocated before freeing it.
1342 */
1343static void ubifs_umount(struct ubifs_info *c)
1344{
1345 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1346 c->vi.vol_id);
1347
1348 spin_lock(&ubifs_infos_lock);
1349 list_del(&c->infos_list);
1350 spin_unlock(&ubifs_infos_lock);
1351
1352 if (c->bgt)
1353 kthread_stop(c->bgt);
1354
1355 destroy_journal(c);
1356 free_wbufs(c);
1357 free_orphans(c);
1358 ubifs_lpt_free(c, 0);
1359
1360 kfree(c->cbuf);
1361 kfree(c->rcvrd_mst_node);
1362 kfree(c->mst_node);
Artem Bityutskiy3477d202008-11-19 11:53:15 +02001363 kfree(c->bu.buf);
1364 vfree(c->ileb_buf);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001365 vfree(c->sbuf);
1366 kfree(c->bottom_up_buf);
1367 UBIFS_DBG(vfree(c->dbg_buf));
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001368 dbg_failure_mode_deregistration(c);
1369}
1370
1371/**
1372 * ubifs_remount_rw - re-mount in read-write mode.
1373 * @c: UBIFS file-system description object
1374 *
1375 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1376 * mode. This function allocates the needed resources and re-mounts UBIFS in
1377 * read-write mode.
1378 */
1379static int ubifs_remount_rw(struct ubifs_info *c)
1380{
1381 int err, lnum;
1382
1383 if (c->ro_media)
1384 return -EINVAL;
1385
1386 mutex_lock(&c->umount_mutex);
1387 c->remounting_rw = 1;
Adrian Hunter2953e732008-09-04 16:26:00 +03001388 c->always_chk_crc = 1;
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001389
1390 /* Check for enough free space */
1391 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1392 ubifs_err("insufficient available space");
1393 err = -EINVAL;
1394 goto out;
1395 }
1396
1397 if (c->old_leb_cnt != c->leb_cnt) {
1398 struct ubifs_sb_node *sup;
1399
1400 sup = ubifs_read_sb_node(c);
1401 if (IS_ERR(sup)) {
1402 err = PTR_ERR(sup);
1403 goto out;
1404 }
1405 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1406 err = ubifs_write_sb_node(c, sup);
1407 if (err)
1408 goto out;
1409 }
1410
1411 if (c->need_recovery) {
1412 ubifs_msg("completing deferred recovery");
1413 err = ubifs_write_rcvrd_mst_node(c);
1414 if (err)
1415 goto out;
1416 err = ubifs_recover_size(c);
1417 if (err)
1418 goto out;
1419 err = ubifs_clean_lebs(c, c->sbuf);
1420 if (err)
1421 goto out;
1422 err = ubifs_recover_inl_heads(c, c->sbuf);
1423 if (err)
1424 goto out;
1425 }
1426
1427 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1428 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1429 err = ubifs_write_master(c);
1430 if (err)
1431 goto out;
1432 }
1433
1434 c->ileb_buf = vmalloc(c->leb_size);
1435 if (!c->ileb_buf) {
1436 err = -ENOMEM;
1437 goto out;
1438 }
1439
1440 err = ubifs_lpt_init(c, 0, 1);
1441 if (err)
1442 goto out;
1443
1444 err = alloc_wbufs(c);
1445 if (err)
1446 goto out;
1447
1448 ubifs_create_buds_lists(c);
1449
1450 /* Create background thread */
1451 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001452 if (IS_ERR(c->bgt)) {
1453 err = PTR_ERR(c->bgt);
1454 c->bgt = NULL;
1455 ubifs_err("cannot spawn \"%s\", error %d",
1456 c->bgt_name, err);
Adrian Hunter2953e732008-09-04 16:26:00 +03001457 goto out;
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001458 }
1459 wake_up_process(c->bgt);
1460
1461 c->orph_buf = vmalloc(c->leb_size);
Adrian Hunter2953e732008-09-04 16:26:00 +03001462 if (!c->orph_buf) {
1463 err = -ENOMEM;
1464 goto out;
1465 }
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001466
1467 /* Check for enough log space */
1468 lnum = c->lhead_lnum + 1;
1469 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1470 lnum = UBIFS_LOG_LNUM;
1471 if (lnum == c->ltail_lnum) {
1472 err = ubifs_consolidate_log(c);
1473 if (err)
1474 goto out;
1475 }
1476
1477 if (c->need_recovery)
1478 err = ubifs_rcvry_gc_commit(c);
1479 else
1480 err = take_gc_lnum(c);
1481 if (err)
1482 goto out;
1483
1484 if (c->need_recovery) {
1485 c->need_recovery = 0;
1486 ubifs_msg("deferred recovery completed");
1487 }
1488
1489 dbg_gen("re-mounted read-write");
1490 c->vfs_sb->s_flags &= ~MS_RDONLY;
1491 c->remounting_rw = 0;
Adrian Hunter2953e732008-09-04 16:26:00 +03001492 c->always_chk_crc = 0;
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001493 mutex_unlock(&c->umount_mutex);
1494 return 0;
1495
1496out:
1497 vfree(c->orph_buf);
1498 c->orph_buf = NULL;
1499 if (c->bgt) {
1500 kthread_stop(c->bgt);
1501 c->bgt = NULL;
1502 }
1503 free_wbufs(c);
1504 vfree(c->ileb_buf);
1505 c->ileb_buf = NULL;
1506 ubifs_lpt_free(c, 1);
1507 c->remounting_rw = 0;
Adrian Hunter2953e732008-09-04 16:26:00 +03001508 c->always_chk_crc = 0;
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001509 mutex_unlock(&c->umount_mutex);
1510 return err;
1511}
1512
1513/**
1514 * commit_on_unmount - commit the journal when un-mounting.
1515 * @c: UBIFS file-system description object
1516 *
Artem Bityutskiyaf2eb562008-09-09 12:23:50 +03001517 * This function is called during un-mounting and re-mounting, and it commits
1518 * the journal unless the "fast unmount" mode is enabled. It also avoids
1519 * committing the journal if it contains too few data.
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001520 */
1521static void commit_on_unmount(struct ubifs_info *c)
1522{
1523 if (!c->fast_unmount) {
1524 long long bud_bytes;
1525
1526 spin_lock(&c->buds_lock);
1527 bud_bytes = c->bud_bytes;
1528 spin_unlock(&c->buds_lock);
1529 if (bud_bytes > c->leb_size)
1530 ubifs_run_commit(c);
1531 }
1532}
1533
1534/**
1535 * ubifs_remount_ro - re-mount in read-only mode.
1536 * @c: UBIFS file-system description object
1537 *
1538 * We rely on VFS to have stopped writing. Possibly the background thread could
1539 * be running a commit, however kthread_stop will wait in that case.
1540 */
1541static void ubifs_remount_ro(struct ubifs_info *c)
1542{
1543 int i, err;
1544
1545 ubifs_assert(!c->need_recovery);
1546 commit_on_unmount(c);
1547
1548 mutex_lock(&c->umount_mutex);
1549 if (c->bgt) {
1550 kthread_stop(c->bgt);
1551 c->bgt = NULL;
1552 }
1553
1554 for (i = 0; i < c->jhead_cnt; i++) {
1555 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1556 del_timer_sync(&c->jheads[i].wbuf.timer);
1557 }
1558
1559 if (!c->ro_media) {
1560 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1561 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1562 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1563 err = ubifs_write_master(c);
1564 if (err)
1565 ubifs_ro_mode(c, err);
1566 }
1567
1568 ubifs_destroy_idx_gc(c);
1569 free_wbufs(c);
1570 vfree(c->orph_buf);
1571 c->orph_buf = NULL;
1572 vfree(c->ileb_buf);
1573 c->ileb_buf = NULL;
1574 ubifs_lpt_free(c, 1);
1575 mutex_unlock(&c->umount_mutex);
1576}
1577
1578static void ubifs_put_super(struct super_block *sb)
1579{
1580 int i;
1581 struct ubifs_info *c = sb->s_fs_info;
1582
1583 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1584 c->vi.vol_id);
1585 /*
1586 * The following asserts are only valid if there has not been a failure
1587 * of the media. For example, there will be dirty inodes if we failed
1588 * to write them back because of I/O errors.
1589 */
1590 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1591 ubifs_assert(c->budg_idx_growth == 0);
Artem Bityutskiy7d32c2b2008-07-18 18:54:29 +03001592 ubifs_assert(c->budg_dd_growth == 0);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001593 ubifs_assert(c->budg_data_growth == 0);
1594
1595 /*
1596 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1597 * and file system un-mount. Namely, it prevents the shrinker from
1598 * picking this superblock for shrinking - it will be just skipped if
1599 * the mutex is locked.
1600 */
1601 mutex_lock(&c->umount_mutex);
1602 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1603 /*
1604 * First of all kill the background thread to make sure it does
1605 * not interfere with un-mounting and freeing resources.
1606 */
1607 if (c->bgt) {
1608 kthread_stop(c->bgt);
1609 c->bgt = NULL;
1610 }
1611
1612 /* Synchronize write-buffers */
1613 if (c->jheads)
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 /*
1620 * On fatal errors c->ro_media is set to 1, in which case we do
1621 * not write the master node.
1622 */
1623 if (!c->ro_media) {
1624 /*
1625 * We are being cleanly unmounted which means the
1626 * orphans were killed - indicate this in the master
1627 * node. Also save the reserved GC LEB number.
1628 */
1629 int err;
1630
1631 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1632 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1633 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1634 err = ubifs_write_master(c);
1635 if (err)
1636 /*
1637 * Recovery will attempt to fix the master area
1638 * next mount, so we just print a message and
1639 * continue to unmount normally.
1640 */
1641 ubifs_err("failed to write master node, "
1642 "error %d", err);
1643 }
1644 }
1645
1646 ubifs_umount(c);
1647 bdi_destroy(&c->bdi);
1648 ubi_close_volume(c->ubi);
1649 mutex_unlock(&c->umount_mutex);
1650 kfree(c);
1651}
1652
1653static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1654{
1655 int err;
1656 struct ubifs_info *c = sb->s_fs_info;
1657
1658 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1659
1660 err = ubifs_parse_options(c, data, 1);
1661 if (err) {
1662 ubifs_err("invalid or unknown remount parameter");
1663 return err;
1664 }
Artem Bityutskiy3477d202008-11-19 11:53:15 +02001665
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001666 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1667 err = ubifs_remount_rw(c);
1668 if (err)
1669 return err;
1670 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
1671 ubifs_remount_ro(c);
1672
Artem Bityutskiy3477d202008-11-19 11:53:15 +02001673 if (c->bulk_read == 1)
1674 bu_init(c);
1675 else {
1676 dbg_gen("disable bulk-read");
1677 kfree(c->bu.buf);
1678 c->bu.buf = NULL;
1679 }
1680
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001681 return 0;
1682}
1683
1684struct super_operations ubifs_super_operations = {
1685 .alloc_inode = ubifs_alloc_inode,
1686 .destroy_inode = ubifs_destroy_inode,
1687 .put_super = ubifs_put_super,
1688 .write_inode = ubifs_write_inode,
1689 .delete_inode = ubifs_delete_inode,
1690 .statfs = ubifs_statfs,
1691 .dirty_inode = ubifs_dirty_inode,
1692 .remount_fs = ubifs_remount_fs,
1693 .show_options = ubifs_show_options,
1694 .sync_fs = ubifs_sync_fs,
1695};
1696
1697/**
1698 * open_ubi - parse UBI device name string and open the UBI device.
1699 * @name: UBI volume name
1700 * @mode: UBI volume open mode
1701 *
1702 * There are several ways to specify UBI volumes when mounting UBIFS:
1703 * o ubiX_Y - UBI device number X, volume Y;
1704 * o ubiY - UBI device number 0, volume Y;
1705 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1706 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1707 *
1708 * Alternative '!' separator may be used instead of ':' (because some shells
1709 * like busybox may interpret ':' as an NFS host name separator). This function
1710 * returns ubi volume object in case of success and a negative error code in
1711 * case of failure.
1712 */
1713static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1714{
1715 int dev, vol;
1716 char *endptr;
1717
1718 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1719 return ERR_PTR(-EINVAL);
1720
1721 /* ubi:NAME method */
1722 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1723 return ubi_open_volume_nm(0, name + 4, mode);
1724
1725 if (!isdigit(name[3]))
1726 return ERR_PTR(-EINVAL);
1727
1728 dev = simple_strtoul(name + 3, &endptr, 0);
1729
1730 /* ubiY method */
1731 if (*endptr == '\0')
1732 return ubi_open_volume(0, dev, mode);
1733
1734 /* ubiX_Y method */
1735 if (*endptr == '_' && isdigit(endptr[1])) {
1736 vol = simple_strtoul(endptr + 1, &endptr, 0);
1737 if (*endptr != '\0')
1738 return ERR_PTR(-EINVAL);
1739 return ubi_open_volume(dev, vol, mode);
1740 }
1741
1742 /* ubiX:NAME method */
1743 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1744 return ubi_open_volume_nm(dev, ++endptr, mode);
1745
1746 return ERR_PTR(-EINVAL);
1747}
1748
1749static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1750{
1751 struct ubi_volume_desc *ubi = sb->s_fs_info;
1752 struct ubifs_info *c;
1753 struct inode *root;
1754 int err;
1755
1756 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1757 if (!c)
1758 return -ENOMEM;
1759
1760 spin_lock_init(&c->cnt_lock);
1761 spin_lock_init(&c->cs_lock);
1762 spin_lock_init(&c->buds_lock);
1763 spin_lock_init(&c->space_lock);
1764 spin_lock_init(&c->orphan_lock);
1765 init_rwsem(&c->commit_sem);
1766 mutex_init(&c->lp_mutex);
1767 mutex_init(&c->tnc_mutex);
1768 mutex_init(&c->log_mutex);
1769 mutex_init(&c->mst_mutex);
1770 mutex_init(&c->umount_mutex);
Artem Bityutskiy3477d202008-11-19 11:53:15 +02001771 mutex_init(&c->bu_mutex);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001772 init_waitqueue_head(&c->cmt_wq);
1773 c->buds = RB_ROOT;
1774 c->old_idx = RB_ROOT;
1775 c->size_tree = RB_ROOT;
1776 c->orph_tree = RB_ROOT;
1777 INIT_LIST_HEAD(&c->infos_list);
1778 INIT_LIST_HEAD(&c->idx_gc);
1779 INIT_LIST_HEAD(&c->replay_list);
1780 INIT_LIST_HEAD(&c->replay_buds);
1781 INIT_LIST_HEAD(&c->uncat_list);
1782 INIT_LIST_HEAD(&c->empty_list);
1783 INIT_LIST_HEAD(&c->freeable_list);
1784 INIT_LIST_HEAD(&c->frdi_idx_list);
1785 INIT_LIST_HEAD(&c->unclean_leb_list);
1786 INIT_LIST_HEAD(&c->old_buds);
1787 INIT_LIST_HEAD(&c->orph_list);
1788 INIT_LIST_HEAD(&c->orph_new);
1789
1790 c->highest_inum = UBIFS_FIRST_INO;
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001791 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1792
1793 ubi_get_volume_info(ubi, &c->vi);
1794 ubi_get_device_info(c->vi.ubi_num, &c->di);
1795
1796 /* Re-open the UBI device in read-write mode */
1797 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1798 if (IS_ERR(c->ubi)) {
1799 err = PTR_ERR(c->ubi);
1800 goto out_free;
1801 }
1802
1803 /*
Artem Bityutskiy0a883a02008-08-13 14:13:26 +03001804 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001805 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1806 * which means the user would have to wait not just for their own I/O
Artem Bityutskiy0a883a02008-08-13 14:13:26 +03001807 * but the read-ahead I/O as well i.e. completely pointless.
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001808 *
1809 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1810 */
1811 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1812 c->bdi.unplug_io_fn = default_unplug_io_fn;
1813 err = bdi_init(&c->bdi);
1814 if (err)
1815 goto out_close;
1816
1817 err = ubifs_parse_options(c, data, 0);
1818 if (err)
1819 goto out_bdi;
1820
1821 c->vfs_sb = sb;
1822
1823 sb->s_fs_info = c;
1824 sb->s_magic = UBIFS_SUPER_MAGIC;
1825 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1826 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1827 sb->s_dev = c->vi.cdev;
1828 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1829 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1830 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1831 sb->s_op = &ubifs_super_operations;
1832
1833 mutex_lock(&c->umount_mutex);
1834 err = mount_ubifs(c);
1835 if (err) {
1836 ubifs_assert(err < 0);
1837 goto out_unlock;
1838 }
1839
1840 /* Read the root inode */
1841 root = ubifs_iget(sb, UBIFS_ROOT_INO);
1842 if (IS_ERR(root)) {
1843 err = PTR_ERR(root);
1844 goto out_umount;
1845 }
1846
1847 sb->s_root = d_alloc_root(root);
1848 if (!sb->s_root)
1849 goto out_iput;
1850
1851 mutex_unlock(&c->umount_mutex);
1852
1853 return 0;
1854
1855out_iput:
1856 iput(root);
1857out_umount:
1858 ubifs_umount(c);
1859out_unlock:
1860 mutex_unlock(&c->umount_mutex);
1861out_bdi:
1862 bdi_destroy(&c->bdi);
1863out_close:
1864 ubi_close_volume(c->ubi);
1865out_free:
1866 kfree(c);
1867 return err;
1868}
1869
1870static int sb_test(struct super_block *sb, void *data)
1871{
1872 dev_t *dev = data;
1873
1874 return sb->s_dev == *dev;
1875}
1876
1877static int sb_set(struct super_block *sb, void *data)
1878{
1879 dev_t *dev = data;
1880
1881 sb->s_dev = *dev;
1882 return 0;
1883}
1884
1885static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
1886 const char *name, void *data, struct vfsmount *mnt)
1887{
1888 struct ubi_volume_desc *ubi;
1889 struct ubi_volume_info vi;
1890 struct super_block *sb;
1891 int err;
1892
1893 dbg_gen("name %s, flags %#x", name, flags);
1894
1895 /*
1896 * Get UBI device number and volume ID. Mount it read-only so far
1897 * because this might be a new mount point, and UBI allows only one
1898 * read-write user at a time.
1899 */
1900 ubi = open_ubi(name, UBI_READONLY);
1901 if (IS_ERR(ubi)) {
1902 ubifs_err("cannot open \"%s\", error %d",
1903 name, (int)PTR_ERR(ubi));
1904 return PTR_ERR(ubi);
1905 }
1906 ubi_get_volume_info(ubi, &vi);
1907
1908 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
1909
1910 sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
1911 if (IS_ERR(sb)) {
1912 err = PTR_ERR(sb);
1913 goto out_close;
1914 }
1915
1916 if (sb->s_root) {
1917 /* A new mount point for already mounted UBIFS */
1918 dbg_gen("this ubi volume is already mounted");
1919 if ((flags ^ sb->s_flags) & MS_RDONLY) {
1920 err = -EBUSY;
1921 goto out_deact;
1922 }
1923 } else {
1924 sb->s_flags = flags;
1925 /*
1926 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
1927 * replaced by 'c'.
1928 */
1929 sb->s_fs_info = ubi;
1930 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
1931 if (err)
1932 goto out_deact;
1933 /* We do not support atime */
1934 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
1935 }
1936
1937 /* 'fill_super()' opens ubi again so we must close it here */
1938 ubi_close_volume(ubi);
1939
1940 return simple_set_mnt(mnt, sb);
1941
1942out_deact:
1943 up_write(&sb->s_umount);
1944 deactivate_super(sb);
1945out_close:
1946 ubi_close_volume(ubi);
1947 return err;
1948}
1949
1950static void ubifs_kill_sb(struct super_block *sb)
1951{
1952 struct ubifs_info *c = sb->s_fs_info;
1953
1954 /*
1955 * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
1956 * in order to be outside BKL.
1957 */
1958 if (sb->s_root && !(sb->s_flags & MS_RDONLY))
1959 commit_on_unmount(c);
1960 /* The un-mount routine is actually done in put_super() */
1961 generic_shutdown_super(sb);
1962}
1963
1964static struct file_system_type ubifs_fs_type = {
1965 .name = "ubifs",
1966 .owner = THIS_MODULE,
1967 .get_sb = ubifs_get_sb,
1968 .kill_sb = ubifs_kill_sb
1969};
1970
1971/*
1972 * Inode slab cache constructor.
1973 */
Alexey Dobriyan51cc5062008-07-25 19:45:34 -07001974static void inode_slab_ctor(void *obj)
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001975{
1976 struct ubifs_inode *ui = obj;
1977 inode_init_once(&ui->vfs_inode);
1978}
1979
1980static int __init ubifs_init(void)
1981{
1982 int err;
1983
1984 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
1985
1986 /* Make sure node sizes are 8-byte aligned */
1987 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
1988 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
1989 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
1990 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
1991 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
1992 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
1993 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
1994 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
1995 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
1996 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
1997 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
1998
1999 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2000 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2001 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2002 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2003 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2004 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2005
2006 /* Check min. node size */
2007 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2008 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2009 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2010 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2011
2012 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2013 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2014 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2015 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2016
2017 /* Defined node sizes */
2018 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2019 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2020 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2021 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2022
2023 /*
2024 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2025 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2026 */
2027 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2028 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2029 " at least 4096 bytes",
2030 (unsigned int)PAGE_CACHE_SIZE);
2031 return -EINVAL;
2032 }
2033
2034 err = register_filesystem(&ubifs_fs_type);
2035 if (err) {
2036 ubifs_err("cannot register file system, error %d", err);
2037 return err;
2038 }
2039
2040 err = -ENOMEM;
2041 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2042 sizeof(struct ubifs_inode), 0,
2043 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2044 &inode_slab_ctor);
2045 if (!ubifs_inode_slab)
2046 goto out_reg;
2047
2048 register_shrinker(&ubifs_shrinker_info);
2049
2050 err = ubifs_compressors_init();
2051 if (err)
2052 goto out_compr;
2053
2054 return 0;
2055
2056out_compr:
2057 unregister_shrinker(&ubifs_shrinker_info);
2058 kmem_cache_destroy(ubifs_inode_slab);
2059out_reg:
2060 unregister_filesystem(&ubifs_fs_type);
2061 return err;
2062}
2063/* late_initcall to let compressors initialize first */
2064late_initcall(ubifs_init);
2065
2066static void __exit ubifs_exit(void)
2067{
2068 ubifs_assert(list_empty(&ubifs_infos));
2069 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2070
2071 ubifs_compressors_exit();
2072 unregister_shrinker(&ubifs_shrinker_info);
2073 kmem_cache_destroy(ubifs_inode_slab);
2074 unregister_filesystem(&ubifs_fs_type);
2075}
2076module_exit(ubifs_exit);
2077
2078MODULE_LICENSE("GPL");
2079MODULE_VERSION(__stringify(UBIFS_VERSION));
2080MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2081MODULE_DESCRIPTION("UBIFS - UBI File System");