Commit | Line | Data |
---|---|---|
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> | |
33 | #include <linux/random.h> | |
34 | #include <linux/kthread.h> | |
35 | #include <linux/parser.h> | |
36 | #include <linux/seq_file.h> | |
37 | #include <linux/mount.h> | |
38 | #include "ubifs.h" | |
39 | ||
40 | /* Slab cache for UBIFS inodes */ | |
41 | struct kmem_cache *ubifs_inode_slab; | |
42 | ||
43 | /* UBIFS TNC shrinker description */ | |
44 | static struct shrinker ubifs_shrinker_info = { | |
45 | .shrink = ubifs_shrinker, | |
46 | .seeks = DEFAULT_SEEKS, | |
47 | }; | |
48 | ||
49 | /** | |
50 | * validate_inode - validate inode. | |
51 | * @c: UBIFS file-system description object | |
52 | * @inode: the inode to validate | |
53 | * | |
54 | * This is a helper function for 'ubifs_iget()' which validates various fields | |
55 | * of a newly built inode to make sure they contain sane values and prevent | |
56 | * possible vulnerabilities. Returns zero if the inode is all right and | |
57 | * a non-zero error code if not. | |
58 | */ | |
59 | static int validate_inode(struct ubifs_info *c, const struct inode *inode) | |
60 | { | |
61 | int err; | |
62 | const struct ubifs_inode *ui = ubifs_inode(inode); | |
63 | ||
64 | if (inode->i_size > c->max_inode_sz) { | |
65 | ubifs_err("inode is too large (%lld)", | |
66 | (long long)inode->i_size); | |
67 | return 1; | |
68 | } | |
69 | ||
70 | if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) { | |
71 | ubifs_err("unknown compression type %d", ui->compr_type); | |
72 | return 2; | |
73 | } | |
74 | ||
75 | if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX) | |
76 | return 3; | |
77 | ||
78 | if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA) | |
79 | return 4; | |
80 | ||
81 | if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG) | |
82 | return 5; | |
83 | ||
84 | if (!ubifs_compr_present(ui->compr_type)) { | |
85 | ubifs_warn("inode %lu uses '%s' compression, but it was not " | |
86 | "compiled in", inode->i_ino, | |
87 | ubifs_compr_name(ui->compr_type)); | |
88 | } | |
89 | ||
90 | err = dbg_check_dir_size(c, inode); | |
91 | return err; | |
92 | } | |
93 | ||
94 | struct inode *ubifs_iget(struct super_block *sb, unsigned long inum) | |
95 | { | |
96 | int err; | |
97 | union ubifs_key key; | |
98 | struct ubifs_ino_node *ino; | |
99 | struct ubifs_info *c = sb->s_fs_info; | |
100 | struct inode *inode; | |
101 | struct ubifs_inode *ui; | |
102 | ||
103 | dbg_gen("inode %lu", inum); | |
104 | ||
105 | inode = iget_locked(sb, inum); | |
106 | if (!inode) | |
107 | return ERR_PTR(-ENOMEM); | |
108 | if (!(inode->i_state & I_NEW)) | |
109 | return inode; | |
110 | ui = ubifs_inode(inode); | |
111 | ||
112 | ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS); | |
113 | if (!ino) { | |
114 | err = -ENOMEM; | |
115 | goto out; | |
116 | } | |
117 | ||
118 | ino_key_init(c, &key, inode->i_ino); | |
119 | ||
120 | err = ubifs_tnc_lookup(c, &key, ino); | |
121 | if (err) | |
122 | goto out_ino; | |
123 | ||
124 | inode->i_flags |= (S_NOCMTIME | S_NOATIME); | |
125 | inode->i_nlink = le32_to_cpu(ino->nlink); | |
126 | inode->i_uid = le32_to_cpu(ino->uid); | |
127 | inode->i_gid = le32_to_cpu(ino->gid); | |
128 | inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec); | |
129 | inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec); | |
130 | inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec); | |
131 | inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec); | |
132 | inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec); | |
133 | inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec); | |
134 | inode->i_mode = le32_to_cpu(ino->mode); | |
135 | inode->i_size = le64_to_cpu(ino->size); | |
136 | ||
137 | ui->data_len = le32_to_cpu(ino->data_len); | |
138 | ui->flags = le32_to_cpu(ino->flags); | |
139 | ui->compr_type = le16_to_cpu(ino->compr_type); | |
140 | ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum); | |
141 | ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt); | |
142 | ui->xattr_size = le32_to_cpu(ino->xattr_size); | |
143 | ui->xattr_names = le32_to_cpu(ino->xattr_names); | |
144 | ui->synced_i_size = ui->ui_size = inode->i_size; | |
145 | ||
146 | ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0; | |
147 | ||
148 | err = validate_inode(c, inode); | |
149 | if (err) | |
150 | goto out_invalid; | |
151 | ||
152 | /* Disable readahead */ | |
153 | inode->i_mapping->backing_dev_info = &c->bdi; | |
154 | ||
155 | switch (inode->i_mode & S_IFMT) { | |
156 | case S_IFREG: | |
157 | inode->i_mapping->a_ops = &ubifs_file_address_operations; | |
158 | inode->i_op = &ubifs_file_inode_operations; | |
159 | inode->i_fop = &ubifs_file_operations; | |
160 | if (ui->xattr) { | |
161 | ui->data = kmalloc(ui->data_len + 1, GFP_NOFS); | |
162 | if (!ui->data) { | |
163 | err = -ENOMEM; | |
164 | goto out_ino; | |
165 | } | |
166 | memcpy(ui->data, ino->data, ui->data_len); | |
167 | ((char *)ui->data)[ui->data_len] = '\0'; | |
168 | } else if (ui->data_len != 0) { | |
169 | err = 10; | |
170 | goto out_invalid; | |
171 | } | |
172 | break; | |
173 | case S_IFDIR: | |
174 | inode->i_op = &ubifs_dir_inode_operations; | |
175 | inode->i_fop = &ubifs_dir_operations; | |
176 | if (ui->data_len != 0) { | |
177 | err = 11; | |
178 | goto out_invalid; | |
179 | } | |
180 | break; | |
181 | case S_IFLNK: | |
182 | inode->i_op = &ubifs_symlink_inode_operations; | |
183 | if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) { | |
184 | err = 12; | |
185 | goto out_invalid; | |
186 | } | |
187 | ui->data = kmalloc(ui->data_len + 1, GFP_NOFS); | |
188 | if (!ui->data) { | |
189 | err = -ENOMEM; | |
190 | goto out_ino; | |
191 | } | |
192 | memcpy(ui->data, ino->data, ui->data_len); | |
193 | ((char *)ui->data)[ui->data_len] = '\0'; | |
194 | break; | |
195 | case S_IFBLK: | |
196 | case S_IFCHR: | |
197 | { | |
198 | dev_t rdev; | |
199 | union ubifs_dev_desc *dev; | |
200 | ||
201 | ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS); | |
202 | if (!ui->data) { | |
203 | err = -ENOMEM; | |
204 | goto out_ino; | |
205 | } | |
206 | ||
207 | dev = (union ubifs_dev_desc *)ino->data; | |
208 | if (ui->data_len == sizeof(dev->new)) | |
209 | rdev = new_decode_dev(le32_to_cpu(dev->new)); | |
210 | else if (ui->data_len == sizeof(dev->huge)) | |
211 | rdev = huge_decode_dev(le64_to_cpu(dev->huge)); | |
212 | else { | |
213 | err = 13; | |
214 | goto out_invalid; | |
215 | } | |
216 | memcpy(ui->data, ino->data, ui->data_len); | |
217 | inode->i_op = &ubifs_file_inode_operations; | |
218 | init_special_inode(inode, inode->i_mode, rdev); | |
219 | break; | |
220 | } | |
221 | case S_IFSOCK: | |
222 | case S_IFIFO: | |
223 | inode->i_op = &ubifs_file_inode_operations; | |
224 | init_special_inode(inode, inode->i_mode, 0); | |
225 | if (ui->data_len != 0) { | |
226 | err = 14; | |
227 | goto out_invalid; | |
228 | } | |
229 | break; | |
230 | default: | |
231 | err = 15; | |
232 | goto out_invalid; | |
233 | } | |
234 | ||
235 | kfree(ino); | |
236 | ubifs_set_inode_flags(inode); | |
237 | unlock_new_inode(inode); | |
238 | return inode; | |
239 | ||
240 | out_invalid: | |
241 | ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err); | |
242 | dbg_dump_node(c, ino); | |
243 | dbg_dump_inode(c, inode); | |
244 | err = -EINVAL; | |
245 | out_ino: | |
246 | kfree(ino); | |
247 | out: | |
248 | ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err); | |
249 | iget_failed(inode); | |
250 | return ERR_PTR(err); | |
251 | } | |
252 | ||
253 | static struct inode *ubifs_alloc_inode(struct super_block *sb) | |
254 | { | |
255 | struct ubifs_inode *ui; | |
256 | ||
257 | ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS); | |
258 | if (!ui) | |
259 | return NULL; | |
260 | ||
261 | memset((void *)ui + sizeof(struct inode), 0, | |
262 | sizeof(struct ubifs_inode) - sizeof(struct inode)); | |
263 | mutex_init(&ui->ui_mutex); | |
264 | spin_lock_init(&ui->ui_lock); | |
265 | return &ui->vfs_inode; | |
266 | }; | |
267 | ||
268 | static void ubifs_destroy_inode(struct inode *inode) | |
269 | { | |
270 | struct ubifs_inode *ui = ubifs_inode(inode); | |
271 | ||
272 | kfree(ui->data); | |
273 | kmem_cache_free(ubifs_inode_slab, inode); | |
274 | } | |
275 | ||
276 | /* | |
277 | * Note, Linux write-back code calls this without 'i_mutex'. | |
278 | */ | |
279 | static int ubifs_write_inode(struct inode *inode, int wait) | |
280 | { | |
fbfa6c88 | 281 | int err = 0; |
1e51764a AB |
282 | struct ubifs_info *c = inode->i_sb->s_fs_info; |
283 | struct ubifs_inode *ui = ubifs_inode(inode); | |
284 | ||
285 | ubifs_assert(!ui->xattr); | |
286 | if (is_bad_inode(inode)) | |
287 | return 0; | |
288 | ||
289 | mutex_lock(&ui->ui_mutex); | |
290 | /* | |
291 | * Due to races between write-back forced by budgeting | |
292 | * (see 'sync_some_inodes()') and pdflush write-back, the inode may | |
293 | * have already been synchronized, do not do this again. This might | |
294 | * also happen if it was synchronized in an VFS operation, e.g. | |
295 | * 'ubifs_link()'. | |
296 | */ | |
297 | if (!ui->dirty) { | |
298 | mutex_unlock(&ui->ui_mutex); | |
299 | return 0; | |
300 | } | |
301 | ||
fbfa6c88 AB |
302 | /* |
303 | * As an optimization, do not write orphan inodes to the media just | |
304 | * because this is not needed. | |
305 | */ | |
306 | dbg_gen("inode %lu, mode %#x, nlink %u", | |
307 | inode->i_ino, (int)inode->i_mode, inode->i_nlink); | |
308 | if (inode->i_nlink) { | |
309 | err = ubifs_jnl_write_inode(c, inode, 0); | |
310 | if (err) | |
311 | ubifs_err("can't write inode %lu, error %d", | |
312 | inode->i_ino, err); | |
313 | } | |
1e51764a AB |
314 | |
315 | ui->dirty = 0; | |
316 | mutex_unlock(&ui->ui_mutex); | |
317 | ubifs_release_dirty_inode_budget(c, ui); | |
318 | return err; | |
319 | } | |
320 | ||
321 | static void ubifs_delete_inode(struct inode *inode) | |
322 | { | |
323 | int err; | |
324 | struct ubifs_info *c = inode->i_sb->s_fs_info; | |
1e0f358e | 325 | struct ubifs_inode *ui = ubifs_inode(inode); |
1e51764a | 326 | |
1e0f358e | 327 | if (ui->xattr) |
1e51764a AB |
328 | /* |
329 | * Extended attribute inode deletions are fully handled in | |
330 | * 'ubifs_removexattr()'. These inodes are special and have | |
331 | * limited usage, so there is nothing to do here. | |
332 | */ | |
333 | goto out; | |
334 | ||
7d32c2bb | 335 | dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode); |
1e51764a AB |
336 | ubifs_assert(!atomic_read(&inode->i_count)); |
337 | ubifs_assert(inode->i_nlink == 0); | |
338 | ||
339 | truncate_inode_pages(&inode->i_data, 0); | |
340 | if (is_bad_inode(inode)) | |
341 | goto out; | |
342 | ||
1e0f358e | 343 | ui->ui_size = inode->i_size = 0; |
1e51764a AB |
344 | err = ubifs_jnl_write_inode(c, inode, 1); |
345 | if (err) | |
346 | /* | |
347 | * Worst case we have a lost orphan inode wasting space, so a | |
348 | * simple error message is ok here. | |
349 | */ | |
350 | ubifs_err("can't write inode %lu, error %d", inode->i_ino, err); | |
351 | out: | |
1e0f358e AB |
352 | if (ui->dirty) |
353 | ubifs_release_dirty_inode_budget(c, ui); | |
1e51764a AB |
354 | clear_inode(inode); |
355 | } | |
356 | ||
357 | static void ubifs_dirty_inode(struct inode *inode) | |
358 | { | |
359 | struct ubifs_inode *ui = ubifs_inode(inode); | |
360 | ||
361 | ubifs_assert(mutex_is_locked(&ui->ui_mutex)); | |
362 | if (!ui->dirty) { | |
363 | ui->dirty = 1; | |
364 | dbg_gen("inode %lu", inode->i_ino); | |
365 | } | |
366 | } | |
367 | ||
368 | static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
369 | { | |
370 | struct ubifs_info *c = dentry->d_sb->s_fs_info; | |
371 | unsigned long long free; | |
372 | ||
373 | free = ubifs_budg_get_free_space(c); | |
374 | dbg_gen("free space %lld bytes (%lld blocks)", | |
375 | free, free >> UBIFS_BLOCK_SHIFT); | |
376 | ||
377 | buf->f_type = UBIFS_SUPER_MAGIC; | |
378 | buf->f_bsize = UBIFS_BLOCK_SIZE; | |
379 | buf->f_blocks = c->block_cnt; | |
380 | buf->f_bfree = free >> UBIFS_BLOCK_SHIFT; | |
381 | if (free > c->report_rp_size) | |
382 | buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT; | |
383 | else | |
384 | buf->f_bavail = 0; | |
385 | buf->f_files = 0; | |
386 | buf->f_ffree = 0; | |
387 | buf->f_namelen = UBIFS_MAX_NLEN; | |
388 | ||
389 | return 0; | |
390 | } | |
391 | ||
392 | static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt) | |
393 | { | |
394 | struct ubifs_info *c = mnt->mnt_sb->s_fs_info; | |
395 | ||
396 | if (c->mount_opts.unmount_mode == 2) | |
397 | seq_printf(s, ",fast_unmount"); | |
398 | else if (c->mount_opts.unmount_mode == 1) | |
399 | seq_printf(s, ",norm_unmount"); | |
400 | ||
401 | return 0; | |
402 | } | |
403 | ||
404 | static int ubifs_sync_fs(struct super_block *sb, int wait) | |
405 | { | |
406 | struct ubifs_info *c = sb->s_fs_info; | |
407 | int i, ret = 0, err; | |
408 | ||
409 | if (c->jheads) | |
410 | for (i = 0; i < c->jhead_cnt; i++) { | |
411 | err = ubifs_wbuf_sync(&c->jheads[i].wbuf); | |
412 | if (err && !ret) | |
413 | ret = err; | |
414 | } | |
415 | /* | |
416 | * We ought to call sync for c->ubi but it does not have one. If it had | |
417 | * it would in turn call mtd->sync, however mtd operations are | |
418 | * synchronous anyway, so we don't lose any sleep here. | |
419 | */ | |
420 | return ret; | |
421 | } | |
422 | ||
423 | /** | |
424 | * init_constants_early - initialize UBIFS constants. | |
425 | * @c: UBIFS file-system description object | |
426 | * | |
427 | * This function initialize UBIFS constants which do not need the superblock to | |
428 | * be read. It also checks that the UBI volume satisfies basic UBIFS | |
429 | * requirements. Returns zero in case of success and a negative error code in | |
430 | * case of failure. | |
431 | */ | |
432 | static int init_constants_early(struct ubifs_info *c) | |
433 | { | |
434 | if (c->vi.corrupted) { | |
435 | ubifs_warn("UBI volume is corrupted - read-only mode"); | |
436 | c->ro_media = 1; | |
437 | } | |
438 | ||
439 | if (c->di.ro_mode) { | |
440 | ubifs_msg("read-only UBI device"); | |
441 | c->ro_media = 1; | |
442 | } | |
443 | ||
444 | if (c->vi.vol_type == UBI_STATIC_VOLUME) { | |
445 | ubifs_msg("static UBI volume - read-only mode"); | |
446 | c->ro_media = 1; | |
447 | } | |
448 | ||
449 | c->leb_cnt = c->vi.size; | |
450 | c->leb_size = c->vi.usable_leb_size; | |
451 | c->half_leb_size = c->leb_size / 2; | |
452 | c->min_io_size = c->di.min_io_size; | |
453 | c->min_io_shift = fls(c->min_io_size) - 1; | |
454 | ||
455 | if (c->leb_size < UBIFS_MIN_LEB_SZ) { | |
456 | ubifs_err("too small LEBs (%d bytes), min. is %d bytes", | |
457 | c->leb_size, UBIFS_MIN_LEB_SZ); | |
458 | return -EINVAL; | |
459 | } | |
460 | ||
461 | if (c->leb_cnt < UBIFS_MIN_LEB_CNT) { | |
462 | ubifs_err("too few LEBs (%d), min. is %d", | |
463 | c->leb_cnt, UBIFS_MIN_LEB_CNT); | |
464 | return -EINVAL; | |
465 | } | |
466 | ||
467 | if (!is_power_of_2(c->min_io_size)) { | |
468 | ubifs_err("bad min. I/O size %d", c->min_io_size); | |
469 | return -EINVAL; | |
470 | } | |
471 | ||
472 | /* | |
473 | * UBIFS aligns all node to 8-byte boundary, so to make function in | |
474 | * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is | |
475 | * less than 8. | |
476 | */ | |
477 | if (c->min_io_size < 8) { | |
478 | c->min_io_size = 8; | |
479 | c->min_io_shift = 3; | |
480 | } | |
481 | ||
482 | c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size); | |
483 | c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size); | |
484 | ||
485 | /* | |
486 | * Initialize node length ranges which are mostly needed for node | |
487 | * length validation. | |
488 | */ | |
489 | c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ; | |
490 | c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ; | |
491 | c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ; | |
492 | c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ; | |
493 | c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ; | |
494 | c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ; | |
495 | ||
496 | c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ; | |
497 | c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ; | |
498 | c->ranges[UBIFS_ORPH_NODE].min_len = | |
499 | UBIFS_ORPH_NODE_SZ + sizeof(__le64); | |
500 | c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size; | |
501 | c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ; | |
502 | c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ; | |
503 | c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ; | |
504 | c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ; | |
505 | c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ; | |
506 | c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ; | |
507 | /* | |
508 | * Minimum indexing node size is amended later when superblock is | |
509 | * read and the key length is known. | |
510 | */ | |
511 | c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ; | |
512 | /* | |
513 | * Maximum indexing node size is amended later when superblock is | |
514 | * read and the fanout is known. | |
515 | */ | |
516 | c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX; | |
517 | ||
518 | /* | |
519 | * Initialize dead and dark LEB space watermarks. | |
520 | * | |
521 | * Dead space is the space which cannot be used. Its watermark is | |
522 | * equivalent to min. I/O unit or minimum node size if it is greater | |
523 | * then min. I/O unit. | |
524 | * | |
525 | * Dark space is the space which might be used, or might not, depending | |
526 | * on which node should be written to the LEB. Its watermark is | |
527 | * equivalent to maximum UBIFS node size. | |
528 | */ | |
529 | c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size); | |
530 | c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size); | |
531 | ||
532 | return 0; | |
533 | } | |
534 | ||
535 | /** | |
536 | * bud_wbuf_callback - bud LEB write-buffer synchronization call-back. | |
537 | * @c: UBIFS file-system description object | |
538 | * @lnum: LEB the write-buffer was synchronized to | |
539 | * @free: how many free bytes left in this LEB | |
540 | * @pad: how many bytes were padded | |
541 | * | |
542 | * This is a callback function which is called by the I/O unit when the | |
543 | * write-buffer is synchronized. We need this to correctly maintain space | |
544 | * accounting in bud logical eraseblocks. This function returns zero in case of | |
545 | * success and a negative error code in case of failure. | |
546 | * | |
547 | * This function actually belongs to the journal, but we keep it here because | |
548 | * we want to keep it static. | |
549 | */ | |
550 | static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad) | |
551 | { | |
552 | return ubifs_update_one_lp(c, lnum, free, pad, 0, 0); | |
553 | } | |
554 | ||
555 | /* | |
556 | * init_constants_late - initialize UBIFS constants. | |
557 | * @c: UBIFS file-system description object | |
558 | * | |
559 | * This is a helper function which initializes various UBIFS constants after | |
560 | * the superblock has been read. It also checks various UBIFS parameters and | |
561 | * makes sure they are all right. Returns zero in case of success and a | |
562 | * negative error code in case of failure. | |
563 | */ | |
564 | static int init_constants_late(struct ubifs_info *c) | |
565 | { | |
566 | int tmp, err; | |
567 | uint64_t tmp64; | |
568 | ||
569 | c->main_bytes = (long long)c->main_lebs * c->leb_size; | |
570 | c->max_znode_sz = sizeof(struct ubifs_znode) + | |
571 | c->fanout * sizeof(struct ubifs_zbranch); | |
572 | ||
573 | tmp = ubifs_idx_node_sz(c, 1); | |
574 | c->ranges[UBIFS_IDX_NODE].min_len = tmp; | |
575 | c->min_idx_node_sz = ALIGN(tmp, 8); | |
576 | ||
577 | tmp = ubifs_idx_node_sz(c, c->fanout); | |
578 | c->ranges[UBIFS_IDX_NODE].max_len = tmp; | |
579 | c->max_idx_node_sz = ALIGN(tmp, 8); | |
580 | ||
581 | /* Make sure LEB size is large enough to fit full commit */ | |
582 | tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt; | |
583 | tmp = ALIGN(tmp, c->min_io_size); | |
584 | if (tmp > c->leb_size) { | |
585 | dbg_err("too small LEB size %d, at least %d needed", | |
586 | c->leb_size, tmp); | |
587 | return -EINVAL; | |
588 | } | |
589 | ||
590 | /* | |
591 | * Make sure that the log is large enough to fit reference nodes for | |
592 | * all buds plus one reserved LEB. | |
593 | */ | |
594 | tmp64 = c->max_bud_bytes; | |
595 | tmp = do_div(tmp64, c->leb_size); | |
596 | c->max_bud_cnt = tmp64 + !!tmp; | |
597 | tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1); | |
598 | tmp /= c->leb_size; | |
599 | tmp += 1; | |
600 | if (c->log_lebs < tmp) { | |
601 | dbg_err("too small log %d LEBs, required min. %d LEBs", | |
602 | c->log_lebs, tmp); | |
603 | return -EINVAL; | |
604 | } | |
605 | ||
606 | /* | |
607 | * When budgeting we assume worst-case scenarios when the pages are not | |
608 | * be compressed and direntries are of the maximum size. | |
609 | * | |
610 | * Note, data, which may be stored in inodes is budgeted separately, so | |
611 | * it is not included into 'c->inode_budget'. | |
612 | */ | |
613 | c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE; | |
614 | c->inode_budget = UBIFS_INO_NODE_SZ; | |
615 | c->dent_budget = UBIFS_MAX_DENT_NODE_SZ; | |
616 | ||
617 | /* | |
618 | * When the amount of flash space used by buds becomes | |
619 | * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit. | |
620 | * The writers are unblocked when the commit is finished. To avoid | |
621 | * writers to be blocked UBIFS initiates background commit in advance, | |
622 | * when number of bud bytes becomes above the limit defined below. | |
623 | */ | |
624 | c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4; | |
625 | ||
626 | /* | |
627 | * Ensure minimum journal size. All the bytes in the journal heads are | |
628 | * considered to be used, when calculating the current journal usage. | |
629 | * Consequently, if the journal is too small, UBIFS will treat it as | |
630 | * always full. | |
631 | */ | |
632 | tmp64 = (uint64_t)(c->jhead_cnt + 1) * c->leb_size + 1; | |
633 | if (c->bg_bud_bytes < tmp64) | |
634 | c->bg_bud_bytes = tmp64; | |
635 | if (c->max_bud_bytes < tmp64 + c->leb_size) | |
636 | c->max_bud_bytes = tmp64 + c->leb_size; | |
637 | ||
638 | err = ubifs_calc_lpt_geom(c); | |
639 | if (err) | |
640 | return err; | |
641 | ||
642 | c->min_idx_lebs = ubifs_calc_min_idx_lebs(c); | |
643 | ||
644 | /* | |
645 | * Calculate total amount of FS blocks. This number is not used | |
646 | * internally because it does not make much sense for UBIFS, but it is | |
647 | * necessary to report something for the 'statfs()' call. | |
648 | * | |
649 | * Subtract the LEB reserved for GC and the LEB which is reserved for | |
650 | * deletions. | |
651 | * | |
652 | * Review 'ubifs_calc_available()' if changing this calculation. | |
653 | */ | |
654 | tmp64 = c->main_lebs - 2; | |
655 | tmp64 *= (uint64_t)c->leb_size - c->dark_wm; | |
656 | tmp64 = ubifs_reported_space(c, tmp64); | |
657 | c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT; | |
658 | ||
659 | return 0; | |
660 | } | |
661 | ||
662 | /** | |
663 | * take_gc_lnum - reserve GC LEB. | |
664 | * @c: UBIFS file-system description object | |
665 | * | |
666 | * This function ensures that the LEB reserved for garbage collection is | |
667 | * unmapped and is marked as "taken" in lprops. We also have to set free space | |
668 | * to LEB size and dirty space to zero, because lprops may contain out-of-date | |
669 | * information if the file-system was un-mounted before it has been committed. | |
670 | * This function returns zero in case of success and a negative error code in | |
671 | * case of failure. | |
672 | */ | |
673 | static int take_gc_lnum(struct ubifs_info *c) | |
674 | { | |
675 | int err; | |
676 | ||
677 | if (c->gc_lnum == -1) { | |
678 | ubifs_err("no LEB for GC"); | |
679 | return -EINVAL; | |
680 | } | |
681 | ||
682 | err = ubifs_leb_unmap(c, c->gc_lnum); | |
683 | if (err) | |
684 | return err; | |
685 | ||
686 | /* And we have to tell lprops that this LEB is taken */ | |
687 | err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0, | |
688 | LPROPS_TAKEN, 0, 0); | |
689 | return err; | |
690 | } | |
691 | ||
692 | /** | |
693 | * alloc_wbufs - allocate write-buffers. | |
694 | * @c: UBIFS file-system description object | |
695 | * | |
696 | * This helper function allocates and initializes UBIFS write-buffers. Returns | |
697 | * zero in case of success and %-ENOMEM in case of failure. | |
698 | */ | |
699 | static int alloc_wbufs(struct ubifs_info *c) | |
700 | { | |
701 | int i, err; | |
702 | ||
703 | c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead), | |
704 | GFP_KERNEL); | |
705 | if (!c->jheads) | |
706 | return -ENOMEM; | |
707 | ||
708 | /* Initialize journal heads */ | |
709 | for (i = 0; i < c->jhead_cnt; i++) { | |
710 | INIT_LIST_HEAD(&c->jheads[i].buds_list); | |
711 | err = ubifs_wbuf_init(c, &c->jheads[i].wbuf); | |
712 | if (err) | |
713 | return err; | |
714 | ||
715 | c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback; | |
716 | c->jheads[i].wbuf.jhead = i; | |
717 | } | |
718 | ||
719 | c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM; | |
720 | /* | |
721 | * Garbage Collector head likely contains long-term data and | |
722 | * does not need to be synchronized by timer. | |
723 | */ | |
724 | c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM; | |
725 | c->jheads[GCHD].wbuf.timeout = 0; | |
726 | ||
727 | return 0; | |
728 | } | |
729 | ||
730 | /** | |
731 | * free_wbufs - free write-buffers. | |
732 | * @c: UBIFS file-system description object | |
733 | */ | |
734 | static void free_wbufs(struct ubifs_info *c) | |
735 | { | |
736 | int i; | |
737 | ||
738 | if (c->jheads) { | |
739 | for (i = 0; i < c->jhead_cnt; i++) { | |
740 | kfree(c->jheads[i].wbuf.buf); | |
741 | kfree(c->jheads[i].wbuf.inodes); | |
742 | } | |
743 | kfree(c->jheads); | |
744 | c->jheads = NULL; | |
745 | } | |
746 | } | |
747 | ||
748 | /** | |
749 | * free_orphans - free orphans. | |
750 | * @c: UBIFS file-system description object | |
751 | */ | |
752 | static void free_orphans(struct ubifs_info *c) | |
753 | { | |
754 | struct ubifs_orphan *orph; | |
755 | ||
756 | while (c->orph_dnext) { | |
757 | orph = c->orph_dnext; | |
758 | c->orph_dnext = orph->dnext; | |
759 | list_del(&orph->list); | |
760 | kfree(orph); | |
761 | } | |
762 | ||
763 | while (!list_empty(&c->orph_list)) { | |
764 | orph = list_entry(c->orph_list.next, struct ubifs_orphan, list); | |
765 | list_del(&orph->list); | |
766 | kfree(orph); | |
767 | dbg_err("orphan list not empty at unmount"); | |
768 | } | |
769 | ||
770 | vfree(c->orph_buf); | |
771 | c->orph_buf = NULL; | |
772 | } | |
773 | ||
774 | /** | |
775 | * free_buds - free per-bud objects. | |
776 | * @c: UBIFS file-system description object | |
777 | */ | |
778 | static void free_buds(struct ubifs_info *c) | |
779 | { | |
780 | struct rb_node *this = c->buds.rb_node; | |
781 | struct ubifs_bud *bud; | |
782 | ||
783 | while (this) { | |
784 | if (this->rb_left) | |
785 | this = this->rb_left; | |
786 | else if (this->rb_right) | |
787 | this = this->rb_right; | |
788 | else { | |
789 | bud = rb_entry(this, struct ubifs_bud, rb); | |
790 | this = rb_parent(this); | |
791 | if (this) { | |
792 | if (this->rb_left == &bud->rb) | |
793 | this->rb_left = NULL; | |
794 | else | |
795 | this->rb_right = NULL; | |
796 | } | |
797 | kfree(bud); | |
798 | } | |
799 | } | |
800 | } | |
801 | ||
802 | /** | |
803 | * check_volume_empty - check if the UBI volume is empty. | |
804 | * @c: UBIFS file-system description object | |
805 | * | |
806 | * This function checks if the UBIFS volume is empty by looking if its LEBs are | |
807 | * mapped or not. The result of checking is stored in the @c->empty variable. | |
808 | * Returns zero in case of success and a negative error code in case of | |
809 | * failure. | |
810 | */ | |
811 | static int check_volume_empty(struct ubifs_info *c) | |
812 | { | |
813 | int lnum, err; | |
814 | ||
815 | c->empty = 1; | |
816 | for (lnum = 0; lnum < c->leb_cnt; lnum++) { | |
817 | err = ubi_is_mapped(c->ubi, lnum); | |
818 | if (unlikely(err < 0)) | |
819 | return err; | |
820 | if (err == 1) { | |
821 | c->empty = 0; | |
822 | break; | |
823 | } | |
824 | ||
825 | cond_resched(); | |
826 | } | |
827 | ||
828 | return 0; | |
829 | } | |
830 | ||
831 | /* | |
832 | * UBIFS mount options. | |
833 | * | |
834 | * Opt_fast_unmount: do not run a journal commit before un-mounting | |
835 | * Opt_norm_unmount: run a journal commit before un-mounting | |
836 | * Opt_err: just end of array marker | |
837 | */ | |
838 | enum { | |
839 | Opt_fast_unmount, | |
840 | Opt_norm_unmount, | |
841 | Opt_err, | |
842 | }; | |
843 | ||
844 | static match_table_t tokens = { | |
845 | {Opt_fast_unmount, "fast_unmount"}, | |
846 | {Opt_norm_unmount, "norm_unmount"}, | |
847 | {Opt_err, NULL}, | |
848 | }; | |
849 | ||
850 | /** | |
851 | * ubifs_parse_options - parse mount parameters. | |
852 | * @c: UBIFS file-system description object | |
853 | * @options: parameters to parse | |
854 | * @is_remount: non-zero if this is FS re-mount | |
855 | * | |
856 | * This function parses UBIFS mount options and returns zero in case success | |
857 | * and a negative error code in case of failure. | |
858 | */ | |
859 | static int ubifs_parse_options(struct ubifs_info *c, char *options, | |
860 | int is_remount) | |
861 | { | |
862 | char *p; | |
863 | substring_t args[MAX_OPT_ARGS]; | |
864 | ||
865 | if (!options) | |
866 | return 0; | |
867 | ||
868 | while ((p = strsep(&options, ","))) { | |
869 | int token; | |
870 | ||
871 | if (!*p) | |
872 | continue; | |
873 | ||
874 | token = match_token(p, tokens, args); | |
875 | switch (token) { | |
876 | case Opt_fast_unmount: | |
877 | c->mount_opts.unmount_mode = 2; | |
878 | c->fast_unmount = 1; | |
879 | break; | |
880 | case Opt_norm_unmount: | |
881 | c->mount_opts.unmount_mode = 1; | |
882 | c->fast_unmount = 0; | |
883 | break; | |
884 | default: | |
885 | ubifs_err("unrecognized mount option \"%s\" " | |
886 | "or missing value", p); | |
887 | return -EINVAL; | |
888 | } | |
889 | } | |
890 | ||
891 | return 0; | |
892 | } | |
893 | ||
894 | /** | |
895 | * destroy_journal - destroy journal data structures. | |
896 | * @c: UBIFS file-system description object | |
897 | * | |
898 | * This function destroys journal data structures including those that may have | |
899 | * been created by recovery functions. | |
900 | */ | |
901 | static void destroy_journal(struct ubifs_info *c) | |
902 | { | |
903 | while (!list_empty(&c->unclean_leb_list)) { | |
904 | struct ubifs_unclean_leb *ucleb; | |
905 | ||
906 | ucleb = list_entry(c->unclean_leb_list.next, | |
907 | struct ubifs_unclean_leb, list); | |
908 | list_del(&ucleb->list); | |
909 | kfree(ucleb); | |
910 | } | |
911 | while (!list_empty(&c->old_buds)) { | |
912 | struct ubifs_bud *bud; | |
913 | ||
914 | bud = list_entry(c->old_buds.next, struct ubifs_bud, list); | |
915 | list_del(&bud->list); | |
916 | kfree(bud); | |
917 | } | |
918 | ubifs_destroy_idx_gc(c); | |
919 | ubifs_destroy_size_tree(c); | |
920 | ubifs_tnc_close(c); | |
921 | free_buds(c); | |
922 | } | |
923 | ||
924 | /** | |
925 | * mount_ubifs - mount UBIFS file-system. | |
926 | * @c: UBIFS file-system description object | |
927 | * | |
928 | * This function mounts UBIFS file system. Returns zero in case of success and | |
929 | * a negative error code in case of failure. | |
930 | * | |
931 | * Note, the function does not de-allocate resources it it fails half way | |
932 | * through, and the caller has to do this instead. | |
933 | */ | |
934 | static int mount_ubifs(struct ubifs_info *c) | |
935 | { | |
936 | struct super_block *sb = c->vfs_sb; | |
937 | int err, mounted_read_only = (sb->s_flags & MS_RDONLY); | |
938 | long long x; | |
939 | size_t sz; | |
940 | ||
941 | err = init_constants_early(c); | |
942 | if (err) | |
943 | return err; | |
944 | ||
945 | #ifdef CONFIG_UBIFS_FS_DEBUG | |
946 | c->dbg_buf = vmalloc(c->leb_size); | |
947 | if (!c->dbg_buf) | |
948 | return -ENOMEM; | |
949 | #endif | |
950 | ||
951 | err = check_volume_empty(c); | |
952 | if (err) | |
953 | goto out_free; | |
954 | ||
955 | if (c->empty && (mounted_read_only || c->ro_media)) { | |
956 | /* | |
957 | * This UBI volume is empty, and read-only, or the file system | |
958 | * is mounted read-only - we cannot format it. | |
959 | */ | |
960 | ubifs_err("can't format empty UBI volume: read-only %s", | |
961 | c->ro_media ? "UBI volume" : "mount"); | |
962 | err = -EROFS; | |
963 | goto out_free; | |
964 | } | |
965 | ||
966 | if (c->ro_media && !mounted_read_only) { | |
967 | ubifs_err("cannot mount read-write - read-only media"); | |
968 | err = -EROFS; | |
969 | goto out_free; | |
970 | } | |
971 | ||
972 | /* | |
973 | * The requirement for the buffer is that it should fit indexing B-tree | |
974 | * height amount of integers. We assume the height if the TNC tree will | |
975 | * never exceed 64. | |
976 | */ | |
977 | err = -ENOMEM; | |
978 | c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL); | |
979 | if (!c->bottom_up_buf) | |
980 | goto out_free; | |
981 | ||
982 | c->sbuf = vmalloc(c->leb_size); | |
983 | if (!c->sbuf) | |
984 | goto out_free; | |
985 | ||
986 | if (!mounted_read_only) { | |
987 | c->ileb_buf = vmalloc(c->leb_size); | |
988 | if (!c->ileb_buf) | |
989 | goto out_free; | |
990 | } | |
991 | ||
992 | err = ubifs_read_superblock(c); | |
993 | if (err) | |
994 | goto out_free; | |
995 | ||
996 | /* | |
997 | * Make sure the compressor which is set as the default on in the | |
998 | * superblock was actually compiled in. | |
999 | */ | |
1000 | if (!ubifs_compr_present(c->default_compr)) { | |
1001 | ubifs_warn("'%s' compressor is set by superblock, but not " | |
1002 | "compiled in", ubifs_compr_name(c->default_compr)); | |
1003 | c->default_compr = UBIFS_COMPR_NONE; | |
1004 | } | |
1005 | ||
1006 | dbg_failure_mode_registration(c); | |
1007 | ||
1008 | err = init_constants_late(c); | |
1009 | if (err) | |
1010 | goto out_dereg; | |
1011 | ||
1012 | sz = ALIGN(c->max_idx_node_sz, c->min_io_size); | |
1013 | sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size); | |
1014 | c->cbuf = kmalloc(sz, GFP_NOFS); | |
1015 | if (!c->cbuf) { | |
1016 | err = -ENOMEM; | |
1017 | goto out_dereg; | |
1018 | } | |
1019 | ||
1020 | if (!mounted_read_only) { | |
1021 | err = alloc_wbufs(c); | |
1022 | if (err) | |
1023 | goto out_cbuf; | |
1024 | ||
1025 | /* Create background thread */ | |
1026 | sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, | |
1027 | c->vi.vol_id); | |
1028 | c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name); | |
1029 | if (!c->bgt) | |
1030 | c->bgt = ERR_PTR(-EINVAL); | |
1031 | if (IS_ERR(c->bgt)) { | |
1032 | err = PTR_ERR(c->bgt); | |
1033 | c->bgt = NULL; | |
1034 | ubifs_err("cannot spawn \"%s\", error %d", | |
1035 | c->bgt_name, err); | |
1036 | goto out_wbufs; | |
1037 | } | |
1038 | wake_up_process(c->bgt); | |
1039 | } | |
1040 | ||
1041 | err = ubifs_read_master(c); | |
1042 | if (err) | |
1043 | goto out_master; | |
1044 | ||
1045 | if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) { | |
1046 | ubifs_msg("recovery needed"); | |
1047 | c->need_recovery = 1; | |
1048 | if (!mounted_read_only) { | |
1049 | err = ubifs_recover_inl_heads(c, c->sbuf); | |
1050 | if (err) | |
1051 | goto out_master; | |
1052 | } | |
1053 | } else if (!mounted_read_only) { | |
1054 | /* | |
1055 | * Set the "dirty" flag so that if we reboot uncleanly we | |
1056 | * will notice this immediately on the next mount. | |
1057 | */ | |
1058 | c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY); | |
1059 | err = ubifs_write_master(c); | |
1060 | if (err) | |
1061 | goto out_master; | |
1062 | } | |
1063 | ||
1064 | err = ubifs_lpt_init(c, 1, !mounted_read_only); | |
1065 | if (err) | |
1066 | goto out_lpt; | |
1067 | ||
1068 | err = dbg_check_idx_size(c, c->old_idx_sz); | |
1069 | if (err) | |
1070 | goto out_lpt; | |
1071 | ||
1072 | err = ubifs_replay_journal(c); | |
1073 | if (err) | |
1074 | goto out_journal; | |
1075 | ||
1076 | err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only); | |
1077 | if (err) | |
1078 | goto out_orphans; | |
1079 | ||
1080 | if (!mounted_read_only) { | |
1081 | int lnum; | |
1082 | ||
1083 | /* Check for enough free space */ | |
1084 | if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) { | |
1085 | ubifs_err("insufficient available space"); | |
1086 | err = -EINVAL; | |
1087 | goto out_orphans; | |
1088 | } | |
1089 | ||
1090 | /* Check for enough log space */ | |
1091 | lnum = c->lhead_lnum + 1; | |
1092 | if (lnum >= UBIFS_LOG_LNUM + c->log_lebs) | |
1093 | lnum = UBIFS_LOG_LNUM; | |
1094 | if (lnum == c->ltail_lnum) { | |
1095 | err = ubifs_consolidate_log(c); | |
1096 | if (err) | |
1097 | goto out_orphans; | |
1098 | } | |
1099 | ||
1100 | if (c->need_recovery) { | |
1101 | err = ubifs_recover_size(c); | |
1102 | if (err) | |
1103 | goto out_orphans; | |
1104 | err = ubifs_rcvry_gc_commit(c); | |
1105 | } else | |
1106 | err = take_gc_lnum(c); | |
1107 | if (err) | |
1108 | goto out_orphans; | |
1109 | ||
1110 | err = dbg_check_lprops(c); | |
1111 | if (err) | |
1112 | goto out_orphans; | |
1113 | } else if (c->need_recovery) { | |
1114 | err = ubifs_recover_size(c); | |
1115 | if (err) | |
1116 | goto out_orphans; | |
1117 | } | |
1118 | ||
1119 | spin_lock(&ubifs_infos_lock); | |
1120 | list_add_tail(&c->infos_list, &ubifs_infos); | |
1121 | spin_unlock(&ubifs_infos_lock); | |
1122 | ||
1123 | if (c->need_recovery) { | |
1124 | if (mounted_read_only) | |
1125 | ubifs_msg("recovery deferred"); | |
1126 | else { | |
1127 | c->need_recovery = 0; | |
1128 | ubifs_msg("recovery completed"); | |
1129 | } | |
1130 | } | |
1131 | ||
1132 | err = dbg_check_filesystem(c); | |
1133 | if (err) | |
1134 | goto out_infos; | |
1135 | ||
ce769caa AB |
1136 | ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"", |
1137 | c->vi.ubi_num, c->vi.vol_id, c->vi.name); | |
1e51764a AB |
1138 | if (mounted_read_only) |
1139 | ubifs_msg("mounted read-only"); | |
1140 | x = (long long)c->main_lebs * c->leb_size; | |
1141 | ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d LEBs)", | |
1142 | x, x >> 10, x >> 20, c->main_lebs); | |
1143 | x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes; | |
1144 | ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d LEBs)", | |
1145 | x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt); | |
1146 | ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr)); | |
1147 | ubifs_msg("media format %d, latest format %d", | |
1148 | c->fmt_version, UBIFS_FORMAT_VERSION); | |
1149 | ||
1150 | dbg_msg("compiled on: " __DATE__ " at " __TIME__); | |
1151 | dbg_msg("min. I/O unit size: %d bytes", c->min_io_size); | |
1152 | dbg_msg("LEB size: %d bytes (%d KiB)", | |
1153 | c->leb_size, c->leb_size / 1024); | |
1154 | dbg_msg("data journal heads: %d", | |
1155 | c->jhead_cnt - NONDATA_JHEADS_CNT); | |
1156 | dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X" | |
1157 | "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X", | |
1158 | c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3], | |
1159 | c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7], | |
1160 | c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11], | |
1161 | c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]); | |
1162 | dbg_msg("fast unmount: %d", c->fast_unmount); | |
1163 | dbg_msg("big_lpt %d", c->big_lpt); | |
1164 | dbg_msg("log LEBs: %d (%d - %d)", | |
1165 | c->log_lebs, UBIFS_LOG_LNUM, c->log_last); | |
1166 | dbg_msg("LPT area LEBs: %d (%d - %d)", | |
1167 | c->lpt_lebs, c->lpt_first, c->lpt_last); | |
1168 | dbg_msg("orphan area LEBs: %d (%d - %d)", | |
1169 | c->orph_lebs, c->orph_first, c->orph_last); | |
1170 | dbg_msg("main area LEBs: %d (%d - %d)", | |
1171 | c->main_lebs, c->main_first, c->leb_cnt - 1); | |
1172 | dbg_msg("index LEBs: %d", c->lst.idx_lebs); | |
1173 | dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)", | |
1174 | c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20); | |
1175 | dbg_msg("key hash type: %d", c->key_hash_type); | |
1176 | dbg_msg("tree fanout: %d", c->fanout); | |
1177 | dbg_msg("reserved GC LEB: %d", c->gc_lnum); | |
1178 | dbg_msg("first main LEB: %d", c->main_first); | |
1179 | dbg_msg("dead watermark: %d", c->dead_wm); | |
1180 | dbg_msg("dark watermark: %d", c->dark_wm); | |
1181 | x = (long long)c->main_lebs * c->dark_wm; | |
1182 | dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)", | |
1183 | x, x >> 10, x >> 20); | |
1184 | dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)", | |
1185 | c->max_bud_bytes, c->max_bud_bytes >> 10, | |
1186 | c->max_bud_bytes >> 20); | |
1187 | dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)", | |
1188 | c->bg_bud_bytes, c->bg_bud_bytes >> 10, | |
1189 | c->bg_bud_bytes >> 20); | |
1190 | dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)", | |
1191 | c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20); | |
1192 | dbg_msg("max. seq. number: %llu", c->max_sqnum); | |
1193 | dbg_msg("commit number: %llu", c->cmt_no); | |
1194 | ||
1195 | return 0; | |
1196 | ||
1197 | out_infos: | |
1198 | spin_lock(&ubifs_infos_lock); | |
1199 | list_del(&c->infos_list); | |
1200 | spin_unlock(&ubifs_infos_lock); | |
1201 | out_orphans: | |
1202 | free_orphans(c); | |
1203 | out_journal: | |
1204 | destroy_journal(c); | |
1205 | out_lpt: | |
1206 | ubifs_lpt_free(c, 0); | |
1207 | out_master: | |
1208 | kfree(c->mst_node); | |
1209 | kfree(c->rcvrd_mst_node); | |
1210 | if (c->bgt) | |
1211 | kthread_stop(c->bgt); | |
1212 | out_wbufs: | |
1213 | free_wbufs(c); | |
1214 | out_cbuf: | |
1215 | kfree(c->cbuf); | |
1216 | out_dereg: | |
1217 | dbg_failure_mode_deregistration(c); | |
1218 | out_free: | |
1219 | vfree(c->ileb_buf); | |
1220 | vfree(c->sbuf); | |
1221 | kfree(c->bottom_up_buf); | |
1222 | UBIFS_DBG(vfree(c->dbg_buf)); | |
1223 | return err; | |
1224 | } | |
1225 | ||
1226 | /** | |
1227 | * ubifs_umount - un-mount UBIFS file-system. | |
1228 | * @c: UBIFS file-system description object | |
1229 | * | |
1230 | * Note, this function is called to free allocated resourced when un-mounting, | |
1231 | * as well as free resources when an error occurred while we were half way | |
1232 | * through mounting (error path cleanup function). So it has to make sure the | |
1233 | * resource was actually allocated before freeing it. | |
1234 | */ | |
1235 | static void ubifs_umount(struct ubifs_info *c) | |
1236 | { | |
1237 | dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num, | |
1238 | c->vi.vol_id); | |
1239 | ||
1240 | spin_lock(&ubifs_infos_lock); | |
1241 | list_del(&c->infos_list); | |
1242 | spin_unlock(&ubifs_infos_lock); | |
1243 | ||
1244 | if (c->bgt) | |
1245 | kthread_stop(c->bgt); | |
1246 | ||
1247 | destroy_journal(c); | |
1248 | free_wbufs(c); | |
1249 | free_orphans(c); | |
1250 | ubifs_lpt_free(c, 0); | |
1251 | ||
1252 | kfree(c->cbuf); | |
1253 | kfree(c->rcvrd_mst_node); | |
1254 | kfree(c->mst_node); | |
1255 | vfree(c->sbuf); | |
1256 | kfree(c->bottom_up_buf); | |
1257 | UBIFS_DBG(vfree(c->dbg_buf)); | |
1258 | vfree(c->ileb_buf); | |
1259 | dbg_failure_mode_deregistration(c); | |
1260 | } | |
1261 | ||
1262 | /** | |
1263 | * ubifs_remount_rw - re-mount in read-write mode. | |
1264 | * @c: UBIFS file-system description object | |
1265 | * | |
1266 | * UBIFS avoids allocating many unnecessary resources when mounted in read-only | |
1267 | * mode. This function allocates the needed resources and re-mounts UBIFS in | |
1268 | * read-write mode. | |
1269 | */ | |
1270 | static int ubifs_remount_rw(struct ubifs_info *c) | |
1271 | { | |
1272 | int err, lnum; | |
1273 | ||
1274 | if (c->ro_media) | |
1275 | return -EINVAL; | |
1276 | ||
1277 | mutex_lock(&c->umount_mutex); | |
1278 | c->remounting_rw = 1; | |
1279 | ||
1280 | /* Check for enough free space */ | |
1281 | if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) { | |
1282 | ubifs_err("insufficient available space"); | |
1283 | err = -EINVAL; | |
1284 | goto out; | |
1285 | } | |
1286 | ||
1287 | if (c->old_leb_cnt != c->leb_cnt) { | |
1288 | struct ubifs_sb_node *sup; | |
1289 | ||
1290 | sup = ubifs_read_sb_node(c); | |
1291 | if (IS_ERR(sup)) { | |
1292 | err = PTR_ERR(sup); | |
1293 | goto out; | |
1294 | } | |
1295 | sup->leb_cnt = cpu_to_le32(c->leb_cnt); | |
1296 | err = ubifs_write_sb_node(c, sup); | |
1297 | if (err) | |
1298 | goto out; | |
1299 | } | |
1300 | ||
1301 | if (c->need_recovery) { | |
1302 | ubifs_msg("completing deferred recovery"); | |
1303 | err = ubifs_write_rcvrd_mst_node(c); | |
1304 | if (err) | |
1305 | goto out; | |
1306 | err = ubifs_recover_size(c); | |
1307 | if (err) | |
1308 | goto out; | |
1309 | err = ubifs_clean_lebs(c, c->sbuf); | |
1310 | if (err) | |
1311 | goto out; | |
1312 | err = ubifs_recover_inl_heads(c, c->sbuf); | |
1313 | if (err) | |
1314 | goto out; | |
1315 | } | |
1316 | ||
1317 | if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) { | |
1318 | c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY); | |
1319 | err = ubifs_write_master(c); | |
1320 | if (err) | |
1321 | goto out; | |
1322 | } | |
1323 | ||
1324 | c->ileb_buf = vmalloc(c->leb_size); | |
1325 | if (!c->ileb_buf) { | |
1326 | err = -ENOMEM; | |
1327 | goto out; | |
1328 | } | |
1329 | ||
1330 | err = ubifs_lpt_init(c, 0, 1); | |
1331 | if (err) | |
1332 | goto out; | |
1333 | ||
1334 | err = alloc_wbufs(c); | |
1335 | if (err) | |
1336 | goto out; | |
1337 | ||
1338 | ubifs_create_buds_lists(c); | |
1339 | ||
1340 | /* Create background thread */ | |
1341 | c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name); | |
1342 | if (!c->bgt) | |
1343 | c->bgt = ERR_PTR(-EINVAL); | |
1344 | if (IS_ERR(c->bgt)) { | |
1345 | err = PTR_ERR(c->bgt); | |
1346 | c->bgt = NULL; | |
1347 | ubifs_err("cannot spawn \"%s\", error %d", | |
1348 | c->bgt_name, err); | |
1349 | return err; | |
1350 | } | |
1351 | wake_up_process(c->bgt); | |
1352 | ||
1353 | c->orph_buf = vmalloc(c->leb_size); | |
1354 | if (!c->orph_buf) | |
1355 | return -ENOMEM; | |
1356 | ||
1357 | /* Check for enough log space */ | |
1358 | lnum = c->lhead_lnum + 1; | |
1359 | if (lnum >= UBIFS_LOG_LNUM + c->log_lebs) | |
1360 | lnum = UBIFS_LOG_LNUM; | |
1361 | if (lnum == c->ltail_lnum) { | |
1362 | err = ubifs_consolidate_log(c); | |
1363 | if (err) | |
1364 | goto out; | |
1365 | } | |
1366 | ||
1367 | if (c->need_recovery) | |
1368 | err = ubifs_rcvry_gc_commit(c); | |
1369 | else | |
1370 | err = take_gc_lnum(c); | |
1371 | if (err) | |
1372 | goto out; | |
1373 | ||
1374 | if (c->need_recovery) { | |
1375 | c->need_recovery = 0; | |
1376 | ubifs_msg("deferred recovery completed"); | |
1377 | } | |
1378 | ||
1379 | dbg_gen("re-mounted read-write"); | |
1380 | c->vfs_sb->s_flags &= ~MS_RDONLY; | |
1381 | c->remounting_rw = 0; | |
1382 | mutex_unlock(&c->umount_mutex); | |
1383 | return 0; | |
1384 | ||
1385 | out: | |
1386 | vfree(c->orph_buf); | |
1387 | c->orph_buf = NULL; | |
1388 | if (c->bgt) { | |
1389 | kthread_stop(c->bgt); | |
1390 | c->bgt = NULL; | |
1391 | } | |
1392 | free_wbufs(c); | |
1393 | vfree(c->ileb_buf); | |
1394 | c->ileb_buf = NULL; | |
1395 | ubifs_lpt_free(c, 1); | |
1396 | c->remounting_rw = 0; | |
1397 | mutex_unlock(&c->umount_mutex); | |
1398 | return err; | |
1399 | } | |
1400 | ||
1401 | /** | |
1402 | * commit_on_unmount - commit the journal when un-mounting. | |
1403 | * @c: UBIFS file-system description object | |
1404 | * | |
1405 | * This function is called during un-mounting and it commits the journal unless | |
1406 | * the "fast unmount" mode is enabled. It also avoids committing the journal if | |
1407 | * it contains too few data. | |
1408 | * | |
1409 | * Sometimes recovery requires the journal to be committed at least once, and | |
1410 | * this function takes care about this. | |
1411 | */ | |
1412 | static void commit_on_unmount(struct ubifs_info *c) | |
1413 | { | |
1414 | if (!c->fast_unmount) { | |
1415 | long long bud_bytes; | |
1416 | ||
1417 | spin_lock(&c->buds_lock); | |
1418 | bud_bytes = c->bud_bytes; | |
1419 | spin_unlock(&c->buds_lock); | |
1420 | if (bud_bytes > c->leb_size) | |
1421 | ubifs_run_commit(c); | |
1422 | } | |
1423 | } | |
1424 | ||
1425 | /** | |
1426 | * ubifs_remount_ro - re-mount in read-only mode. | |
1427 | * @c: UBIFS file-system description object | |
1428 | * | |
1429 | * We rely on VFS to have stopped writing. Possibly the background thread could | |
1430 | * be running a commit, however kthread_stop will wait in that case. | |
1431 | */ | |
1432 | static void ubifs_remount_ro(struct ubifs_info *c) | |
1433 | { | |
1434 | int i, err; | |
1435 | ||
1436 | ubifs_assert(!c->need_recovery); | |
1437 | commit_on_unmount(c); | |
1438 | ||
1439 | mutex_lock(&c->umount_mutex); | |
1440 | if (c->bgt) { | |
1441 | kthread_stop(c->bgt); | |
1442 | c->bgt = NULL; | |
1443 | } | |
1444 | ||
1445 | for (i = 0; i < c->jhead_cnt; i++) { | |
1446 | ubifs_wbuf_sync(&c->jheads[i].wbuf); | |
1447 | del_timer_sync(&c->jheads[i].wbuf.timer); | |
1448 | } | |
1449 | ||
1450 | if (!c->ro_media) { | |
1451 | c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY); | |
1452 | c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS); | |
1453 | c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum); | |
1454 | err = ubifs_write_master(c); | |
1455 | if (err) | |
1456 | ubifs_ro_mode(c, err); | |
1457 | } | |
1458 | ||
1459 | ubifs_destroy_idx_gc(c); | |
1460 | free_wbufs(c); | |
1461 | vfree(c->orph_buf); | |
1462 | c->orph_buf = NULL; | |
1463 | vfree(c->ileb_buf); | |
1464 | c->ileb_buf = NULL; | |
1465 | ubifs_lpt_free(c, 1); | |
1466 | mutex_unlock(&c->umount_mutex); | |
1467 | } | |
1468 | ||
1469 | static void ubifs_put_super(struct super_block *sb) | |
1470 | { | |
1471 | int i; | |
1472 | struct ubifs_info *c = sb->s_fs_info; | |
1473 | ||
1474 | ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num, | |
1475 | c->vi.vol_id); | |
1476 | /* | |
1477 | * The following asserts are only valid if there has not been a failure | |
1478 | * of the media. For example, there will be dirty inodes if we failed | |
1479 | * to write them back because of I/O errors. | |
1480 | */ | |
1481 | ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0); | |
1482 | ubifs_assert(c->budg_idx_growth == 0); | |
7d32c2bb | 1483 | ubifs_assert(c->budg_dd_growth == 0); |
1e51764a AB |
1484 | ubifs_assert(c->budg_data_growth == 0); |
1485 | ||
1486 | /* | |
1487 | * The 'c->umount_lock' prevents races between UBIFS memory shrinker | |
1488 | * and file system un-mount. Namely, it prevents the shrinker from | |
1489 | * picking this superblock for shrinking - it will be just skipped if | |
1490 | * the mutex is locked. | |
1491 | */ | |
1492 | mutex_lock(&c->umount_mutex); | |
1493 | if (!(c->vfs_sb->s_flags & MS_RDONLY)) { | |
1494 | /* | |
1495 | * First of all kill the background thread to make sure it does | |
1496 | * not interfere with un-mounting and freeing resources. | |
1497 | */ | |
1498 | if (c->bgt) { | |
1499 | kthread_stop(c->bgt); | |
1500 | c->bgt = NULL; | |
1501 | } | |
1502 | ||
1503 | /* Synchronize write-buffers */ | |
1504 | if (c->jheads) | |
1505 | for (i = 0; i < c->jhead_cnt; i++) { | |
1506 | ubifs_wbuf_sync(&c->jheads[i].wbuf); | |
1507 | del_timer_sync(&c->jheads[i].wbuf.timer); | |
1508 | } | |
1509 | ||
1510 | /* | |
1511 | * On fatal errors c->ro_media is set to 1, in which case we do | |
1512 | * not write the master node. | |
1513 | */ | |
1514 | if (!c->ro_media) { | |
1515 | /* | |
1516 | * We are being cleanly unmounted which means the | |
1517 | * orphans were killed - indicate this in the master | |
1518 | * node. Also save the reserved GC LEB number. | |
1519 | */ | |
1520 | int err; | |
1521 | ||
1522 | c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY); | |
1523 | c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS); | |
1524 | c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum); | |
1525 | err = ubifs_write_master(c); | |
1526 | if (err) | |
1527 | /* | |
1528 | * Recovery will attempt to fix the master area | |
1529 | * next mount, so we just print a message and | |
1530 | * continue to unmount normally. | |
1531 | */ | |
1532 | ubifs_err("failed to write master node, " | |
1533 | "error %d", err); | |
1534 | } | |
1535 | } | |
1536 | ||
1537 | ubifs_umount(c); | |
1538 | bdi_destroy(&c->bdi); | |
1539 | ubi_close_volume(c->ubi); | |
1540 | mutex_unlock(&c->umount_mutex); | |
1541 | kfree(c); | |
1542 | } | |
1543 | ||
1544 | static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data) | |
1545 | { | |
1546 | int err; | |
1547 | struct ubifs_info *c = sb->s_fs_info; | |
1548 | ||
1549 | dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags); | |
1550 | ||
1551 | err = ubifs_parse_options(c, data, 1); | |
1552 | if (err) { | |
1553 | ubifs_err("invalid or unknown remount parameter"); | |
1554 | return err; | |
1555 | } | |
1556 | if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) { | |
1557 | err = ubifs_remount_rw(c); | |
1558 | if (err) | |
1559 | return err; | |
1560 | } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY)) | |
1561 | ubifs_remount_ro(c); | |
1562 | ||
1563 | return 0; | |
1564 | } | |
1565 | ||
1566 | struct super_operations ubifs_super_operations = { | |
1567 | .alloc_inode = ubifs_alloc_inode, | |
1568 | .destroy_inode = ubifs_destroy_inode, | |
1569 | .put_super = ubifs_put_super, | |
1570 | .write_inode = ubifs_write_inode, | |
1571 | .delete_inode = ubifs_delete_inode, | |
1572 | .statfs = ubifs_statfs, | |
1573 | .dirty_inode = ubifs_dirty_inode, | |
1574 | .remount_fs = ubifs_remount_fs, | |
1575 | .show_options = ubifs_show_options, | |
1576 | .sync_fs = ubifs_sync_fs, | |
1577 | }; | |
1578 | ||
1579 | /** | |
1580 | * open_ubi - parse UBI device name string and open the UBI device. | |
1581 | * @name: UBI volume name | |
1582 | * @mode: UBI volume open mode | |
1583 | * | |
1584 | * There are several ways to specify UBI volumes when mounting UBIFS: | |
1585 | * o ubiX_Y - UBI device number X, volume Y; | |
1586 | * o ubiY - UBI device number 0, volume Y; | |
1587 | * o ubiX:NAME - mount UBI device X, volume with name NAME; | |
1588 | * o ubi:NAME - mount UBI device 0, volume with name NAME. | |
1589 | * | |
1590 | * Alternative '!' separator may be used instead of ':' (because some shells | |
1591 | * like busybox may interpret ':' as an NFS host name separator). This function | |
1592 | * returns ubi volume object in case of success and a negative error code in | |
1593 | * case of failure. | |
1594 | */ | |
1595 | static struct ubi_volume_desc *open_ubi(const char *name, int mode) | |
1596 | { | |
1597 | int dev, vol; | |
1598 | char *endptr; | |
1599 | ||
1600 | if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i') | |
1601 | return ERR_PTR(-EINVAL); | |
1602 | ||
1603 | /* ubi:NAME method */ | |
1604 | if ((name[3] == ':' || name[3] == '!') && name[4] != '\0') | |
1605 | return ubi_open_volume_nm(0, name + 4, mode); | |
1606 | ||
1607 | if (!isdigit(name[3])) | |
1608 | return ERR_PTR(-EINVAL); | |
1609 | ||
1610 | dev = simple_strtoul(name + 3, &endptr, 0); | |
1611 | ||
1612 | /* ubiY method */ | |
1613 | if (*endptr == '\0') | |
1614 | return ubi_open_volume(0, dev, mode); | |
1615 | ||
1616 | /* ubiX_Y method */ | |
1617 | if (*endptr == '_' && isdigit(endptr[1])) { | |
1618 | vol = simple_strtoul(endptr + 1, &endptr, 0); | |
1619 | if (*endptr != '\0') | |
1620 | return ERR_PTR(-EINVAL); | |
1621 | return ubi_open_volume(dev, vol, mode); | |
1622 | } | |
1623 | ||
1624 | /* ubiX:NAME method */ | |
1625 | if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0') | |
1626 | return ubi_open_volume_nm(dev, ++endptr, mode); | |
1627 | ||
1628 | return ERR_PTR(-EINVAL); | |
1629 | } | |
1630 | ||
1631 | static int ubifs_fill_super(struct super_block *sb, void *data, int silent) | |
1632 | { | |
1633 | struct ubi_volume_desc *ubi = sb->s_fs_info; | |
1634 | struct ubifs_info *c; | |
1635 | struct inode *root; | |
1636 | int err; | |
1637 | ||
1638 | c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL); | |
1639 | if (!c) | |
1640 | return -ENOMEM; | |
1641 | ||
1642 | spin_lock_init(&c->cnt_lock); | |
1643 | spin_lock_init(&c->cs_lock); | |
1644 | spin_lock_init(&c->buds_lock); | |
1645 | spin_lock_init(&c->space_lock); | |
1646 | spin_lock_init(&c->orphan_lock); | |
1647 | init_rwsem(&c->commit_sem); | |
1648 | mutex_init(&c->lp_mutex); | |
1649 | mutex_init(&c->tnc_mutex); | |
1650 | mutex_init(&c->log_mutex); | |
1651 | mutex_init(&c->mst_mutex); | |
1652 | mutex_init(&c->umount_mutex); | |
1653 | init_waitqueue_head(&c->cmt_wq); | |
1654 | c->buds = RB_ROOT; | |
1655 | c->old_idx = RB_ROOT; | |
1656 | c->size_tree = RB_ROOT; | |
1657 | c->orph_tree = RB_ROOT; | |
1658 | INIT_LIST_HEAD(&c->infos_list); | |
1659 | INIT_LIST_HEAD(&c->idx_gc); | |
1660 | INIT_LIST_HEAD(&c->replay_list); | |
1661 | INIT_LIST_HEAD(&c->replay_buds); | |
1662 | INIT_LIST_HEAD(&c->uncat_list); | |
1663 | INIT_LIST_HEAD(&c->empty_list); | |
1664 | INIT_LIST_HEAD(&c->freeable_list); | |
1665 | INIT_LIST_HEAD(&c->frdi_idx_list); | |
1666 | INIT_LIST_HEAD(&c->unclean_leb_list); | |
1667 | INIT_LIST_HEAD(&c->old_buds); | |
1668 | INIT_LIST_HEAD(&c->orph_list); | |
1669 | INIT_LIST_HEAD(&c->orph_new); | |
1670 | ||
1671 | c->highest_inum = UBIFS_FIRST_INO; | |
1672 | get_random_bytes(&c->vfs_gen, sizeof(int)); | |
1673 | c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM; | |
1674 | ||
1675 | ubi_get_volume_info(ubi, &c->vi); | |
1676 | ubi_get_device_info(c->vi.ubi_num, &c->di); | |
1677 | ||
1678 | /* Re-open the UBI device in read-write mode */ | |
1679 | c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE); | |
1680 | if (IS_ERR(c->ubi)) { | |
1681 | err = PTR_ERR(c->ubi); | |
1682 | goto out_free; | |
1683 | } | |
1684 | ||
1685 | /* | |
1686 | * UBIFS provids 'backing_dev_info' in order to disable readahead. For | |
1687 | * UBIFS, I/O is not deferred, it is done immediately in readpage, | |
1688 | * which means the user would have to wait not just for their own I/O | |
1689 | * but the readahead I/O as well i.e. completely pointless. | |
1690 | * | |
1691 | * Read-ahead will be disabled because @c->bdi.ra_pages is 0. | |
1692 | */ | |
1693 | c->bdi.capabilities = BDI_CAP_MAP_COPY; | |
1694 | c->bdi.unplug_io_fn = default_unplug_io_fn; | |
1695 | err = bdi_init(&c->bdi); | |
1696 | if (err) | |
1697 | goto out_close; | |
1698 | ||
1699 | err = ubifs_parse_options(c, data, 0); | |
1700 | if (err) | |
1701 | goto out_bdi; | |
1702 | ||
1703 | c->vfs_sb = sb; | |
1704 | ||
1705 | sb->s_fs_info = c; | |
1706 | sb->s_magic = UBIFS_SUPER_MAGIC; | |
1707 | sb->s_blocksize = UBIFS_BLOCK_SIZE; | |
1708 | sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT; | |
1709 | sb->s_dev = c->vi.cdev; | |
1710 | sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c); | |
1711 | if (c->max_inode_sz > MAX_LFS_FILESIZE) | |
1712 | sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE; | |
1713 | sb->s_op = &ubifs_super_operations; | |
1714 | ||
1715 | mutex_lock(&c->umount_mutex); | |
1716 | err = mount_ubifs(c); | |
1717 | if (err) { | |
1718 | ubifs_assert(err < 0); | |
1719 | goto out_unlock; | |
1720 | } | |
1721 | ||
1722 | /* Read the root inode */ | |
1723 | root = ubifs_iget(sb, UBIFS_ROOT_INO); | |
1724 | if (IS_ERR(root)) { | |
1725 | err = PTR_ERR(root); | |
1726 | goto out_umount; | |
1727 | } | |
1728 | ||
1729 | sb->s_root = d_alloc_root(root); | |
1730 | if (!sb->s_root) | |
1731 | goto out_iput; | |
1732 | ||
1733 | mutex_unlock(&c->umount_mutex); | |
1734 | ||
1735 | return 0; | |
1736 | ||
1737 | out_iput: | |
1738 | iput(root); | |
1739 | out_umount: | |
1740 | ubifs_umount(c); | |
1741 | out_unlock: | |
1742 | mutex_unlock(&c->umount_mutex); | |
1743 | out_bdi: | |
1744 | bdi_destroy(&c->bdi); | |
1745 | out_close: | |
1746 | ubi_close_volume(c->ubi); | |
1747 | out_free: | |
1748 | kfree(c); | |
1749 | return err; | |
1750 | } | |
1751 | ||
1752 | static int sb_test(struct super_block *sb, void *data) | |
1753 | { | |
1754 | dev_t *dev = data; | |
1755 | ||
1756 | return sb->s_dev == *dev; | |
1757 | } | |
1758 | ||
1759 | static int sb_set(struct super_block *sb, void *data) | |
1760 | { | |
1761 | dev_t *dev = data; | |
1762 | ||
1763 | sb->s_dev = *dev; | |
1764 | return 0; | |
1765 | } | |
1766 | ||
1767 | static int ubifs_get_sb(struct file_system_type *fs_type, int flags, | |
1768 | const char *name, void *data, struct vfsmount *mnt) | |
1769 | { | |
1770 | struct ubi_volume_desc *ubi; | |
1771 | struct ubi_volume_info vi; | |
1772 | struct super_block *sb; | |
1773 | int err; | |
1774 | ||
1775 | dbg_gen("name %s, flags %#x", name, flags); | |
1776 | ||
1777 | /* | |
1778 | * Get UBI device number and volume ID. Mount it read-only so far | |
1779 | * because this might be a new mount point, and UBI allows only one | |
1780 | * read-write user at a time. | |
1781 | */ | |
1782 | ubi = open_ubi(name, UBI_READONLY); | |
1783 | if (IS_ERR(ubi)) { | |
1784 | ubifs_err("cannot open \"%s\", error %d", | |
1785 | name, (int)PTR_ERR(ubi)); | |
1786 | return PTR_ERR(ubi); | |
1787 | } | |
1788 | ubi_get_volume_info(ubi, &vi); | |
1789 | ||
1790 | dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id); | |
1791 | ||
1792 | sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev); | |
1793 | if (IS_ERR(sb)) { | |
1794 | err = PTR_ERR(sb); | |
1795 | goto out_close; | |
1796 | } | |
1797 | ||
1798 | if (sb->s_root) { | |
1799 | /* A new mount point for already mounted UBIFS */ | |
1800 | dbg_gen("this ubi volume is already mounted"); | |
1801 | if ((flags ^ sb->s_flags) & MS_RDONLY) { | |
1802 | err = -EBUSY; | |
1803 | goto out_deact; | |
1804 | } | |
1805 | } else { | |
1806 | sb->s_flags = flags; | |
1807 | /* | |
1808 | * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is | |
1809 | * replaced by 'c'. | |
1810 | */ | |
1811 | sb->s_fs_info = ubi; | |
1812 | err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0); | |
1813 | if (err) | |
1814 | goto out_deact; | |
1815 | /* We do not support atime */ | |
1816 | sb->s_flags |= MS_ACTIVE | MS_NOATIME; | |
1817 | } | |
1818 | ||
1819 | /* 'fill_super()' opens ubi again so we must close it here */ | |
1820 | ubi_close_volume(ubi); | |
1821 | ||
1822 | return simple_set_mnt(mnt, sb); | |
1823 | ||
1824 | out_deact: | |
1825 | up_write(&sb->s_umount); | |
1826 | deactivate_super(sb); | |
1827 | out_close: | |
1828 | ubi_close_volume(ubi); | |
1829 | return err; | |
1830 | } | |
1831 | ||
1832 | static void ubifs_kill_sb(struct super_block *sb) | |
1833 | { | |
1834 | struct ubifs_info *c = sb->s_fs_info; | |
1835 | ||
1836 | /* | |
1837 | * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()' | |
1838 | * in order to be outside BKL. | |
1839 | */ | |
1840 | if (sb->s_root && !(sb->s_flags & MS_RDONLY)) | |
1841 | commit_on_unmount(c); | |
1842 | /* The un-mount routine is actually done in put_super() */ | |
1843 | generic_shutdown_super(sb); | |
1844 | } | |
1845 | ||
1846 | static struct file_system_type ubifs_fs_type = { | |
1847 | .name = "ubifs", | |
1848 | .owner = THIS_MODULE, | |
1849 | .get_sb = ubifs_get_sb, | |
1850 | .kill_sb = ubifs_kill_sb | |
1851 | }; | |
1852 | ||
1853 | /* | |
1854 | * Inode slab cache constructor. | |
1855 | */ | |
51cc5068 | 1856 | static void inode_slab_ctor(void *obj) |
1e51764a AB |
1857 | { |
1858 | struct ubifs_inode *ui = obj; | |
1859 | inode_init_once(&ui->vfs_inode); | |
1860 | } | |
1861 | ||
1862 | static int __init ubifs_init(void) | |
1863 | { | |
1864 | int err; | |
1865 | ||
1866 | BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24); | |
1867 | ||
1868 | /* Make sure node sizes are 8-byte aligned */ | |
1869 | BUILD_BUG_ON(UBIFS_CH_SZ & 7); | |
1870 | BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7); | |
1871 | BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7); | |
1872 | BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7); | |
1873 | BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7); | |
1874 | BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7); | |
1875 | BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7); | |
1876 | BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7); | |
1877 | BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7); | |
1878 | BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7); | |
1879 | BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7); | |
1880 | ||
1881 | BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7); | |
1882 | BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7); | |
1883 | BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7); | |
1884 | BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7); | |
1885 | BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7); | |
1886 | BUILD_BUG_ON(MIN_WRITE_SZ & 7); | |
1887 | ||
1888 | /* Check min. node size */ | |
1889 | BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ); | |
1890 | BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ); | |
1891 | BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ); | |
1892 | BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ); | |
1893 | ||
1894 | BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ); | |
1895 | BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ); | |
1896 | BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ); | |
1897 | BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ); | |
1898 | ||
1899 | /* Defined node sizes */ | |
1900 | BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096); | |
1901 | BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512); | |
1902 | BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160); | |
1903 | BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64); | |
1904 | ||
1905 | /* | |
1906 | * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to | |
1907 | * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2. | |
1908 | */ | |
1909 | if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) { | |
1910 | ubifs_err("VFS page cache size is %u bytes, but UBIFS requires" | |
1911 | " at least 4096 bytes", | |
1912 | (unsigned int)PAGE_CACHE_SIZE); | |
1913 | return -EINVAL; | |
1914 | } | |
1915 | ||
1916 | err = register_filesystem(&ubifs_fs_type); | |
1917 | if (err) { | |
1918 | ubifs_err("cannot register file system, error %d", err); | |
1919 | return err; | |
1920 | } | |
1921 | ||
1922 | err = -ENOMEM; | |
1923 | ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab", | |
1924 | sizeof(struct ubifs_inode), 0, | |
1925 | SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT, | |
1926 | &inode_slab_ctor); | |
1927 | if (!ubifs_inode_slab) | |
1928 | goto out_reg; | |
1929 | ||
1930 | register_shrinker(&ubifs_shrinker_info); | |
1931 | ||
1932 | err = ubifs_compressors_init(); | |
1933 | if (err) | |
1934 | goto out_compr; | |
1935 | ||
1936 | return 0; | |
1937 | ||
1938 | out_compr: | |
1939 | unregister_shrinker(&ubifs_shrinker_info); | |
1940 | kmem_cache_destroy(ubifs_inode_slab); | |
1941 | out_reg: | |
1942 | unregister_filesystem(&ubifs_fs_type); | |
1943 | return err; | |
1944 | } | |
1945 | /* late_initcall to let compressors initialize first */ | |
1946 | late_initcall(ubifs_init); | |
1947 | ||
1948 | static void __exit ubifs_exit(void) | |
1949 | { | |
1950 | ubifs_assert(list_empty(&ubifs_infos)); | |
1951 | ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0); | |
1952 | ||
1953 | ubifs_compressors_exit(); | |
1954 | unregister_shrinker(&ubifs_shrinker_info); | |
1955 | kmem_cache_destroy(ubifs_inode_slab); | |
1956 | unregister_filesystem(&ubifs_fs_type); | |
1957 | } | |
1958 | module_exit(ubifs_exit); | |
1959 | ||
1960 | MODULE_LICENSE("GPL"); | |
1961 | MODULE_VERSION(__stringify(UBIFS_VERSION)); | |
1962 | MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter"); | |
1963 | MODULE_DESCRIPTION("UBIFS - UBI File System"); |