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> | |
1e51764a AB |
33 | #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 | ||
39 | /* Slab cache for UBIFS inodes */ | |
40 | struct kmem_cache *ubifs_inode_slab; | |
41 | ||
42 | /* UBIFS TNC shrinker description */ | |
43 | static struct shrinker ubifs_shrinker_info = { | |
44 | .shrink = ubifs_shrinker, | |
45 | .seeks = DEFAULT_SEEKS, | |
46 | }; | |
47 | ||
48 | /** | |
49 | * validate_inode - validate inode. | |
50 | * @c: UBIFS file-system description object | |
51 | * @inode: the inode to validate | |
52 | * | |
53 | * This is a helper function for 'ubifs_iget()' which validates various fields | |
54 | * of a newly built inode to make sure they contain sane values and prevent | |
55 | * possible vulnerabilities. Returns zero if the inode is all right and | |
56 | * a non-zero error code if not. | |
57 | */ | |
58 | static int validate_inode(struct ubifs_info *c, const struct inode *inode) | |
59 | { | |
60 | int err; | |
61 | const struct ubifs_inode *ui = ubifs_inode(inode); | |
62 | ||
63 | if (inode->i_size > c->max_inode_sz) { | |
64 | ubifs_err("inode is too large (%lld)", | |
65 | (long long)inode->i_size); | |
66 | return 1; | |
67 | } | |
68 | ||
69 | if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) { | |
70 | ubifs_err("unknown compression type %d", ui->compr_type); | |
71 | return 2; | |
72 | } | |
73 | ||
74 | if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX) | |
75 | return 3; | |
76 | ||
77 | if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA) | |
78 | return 4; | |
79 | ||
80 | if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG) | |
81 | return 5; | |
82 | ||
83 | if (!ubifs_compr_present(ui->compr_type)) { | |
84 | ubifs_warn("inode %lu uses '%s' compression, but it was not " | |
85 | "compiled in", inode->i_ino, | |
86 | ubifs_compr_name(ui->compr_type)); | |
87 | } | |
88 | ||
89 | err = dbg_check_dir_size(c, inode); | |
90 | return err; | |
91 | } | |
92 | ||
93 | struct inode *ubifs_iget(struct super_block *sb, unsigned long inum) | |
94 | { | |
95 | int err; | |
96 | union ubifs_key key; | |
97 | struct ubifs_ino_node *ino; | |
98 | struct ubifs_info *c = sb->s_fs_info; | |
99 | struct inode *inode; | |
100 | struct ubifs_inode *ui; | |
101 | ||
102 | dbg_gen("inode %lu", inum); | |
103 | ||
104 | inode = iget_locked(sb, inum); | |
105 | if (!inode) | |
106 | return ERR_PTR(-ENOMEM); | |
107 | if (!(inode->i_state & I_NEW)) | |
108 | return inode; | |
109 | ui = ubifs_inode(inode); | |
110 | ||
111 | ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS); | |
112 | if (!ino) { | |
113 | err = -ENOMEM; | |
114 | goto out; | |
115 | } | |
116 | ||
117 | ino_key_init(c, &key, inode->i_ino); | |
118 | ||
119 | err = ubifs_tnc_lookup(c, &key, ino); | |
120 | if (err) | |
121 | goto out_ino; | |
122 | ||
123 | inode->i_flags |= (S_NOCMTIME | S_NOATIME); | |
124 | inode->i_nlink = le32_to_cpu(ino->nlink); | |
125 | inode->i_uid = le32_to_cpu(ino->uid); | |
126 | inode->i_gid = le32_to_cpu(ino->gid); | |
127 | inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec); | |
128 | inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec); | |
129 | inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec); | |
130 | inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec); | |
131 | inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec); | |
132 | inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec); | |
133 | inode->i_mode = le32_to_cpu(ino->mode); | |
134 | inode->i_size = le64_to_cpu(ino->size); | |
135 | ||
136 | ui->data_len = le32_to_cpu(ino->data_len); | |
137 | ui->flags = le32_to_cpu(ino->flags); | |
138 | ui->compr_type = le16_to_cpu(ino->compr_type); | |
139 | ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum); | |
140 | ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt); | |
141 | ui->xattr_size = le32_to_cpu(ino->xattr_size); | |
142 | ui->xattr_names = le32_to_cpu(ino->xattr_names); | |
143 | ui->synced_i_size = ui->ui_size = inode->i_size; | |
144 | ||
145 | ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0; | |
146 | ||
147 | err = validate_inode(c, inode); | |
148 | if (err) | |
149 | goto out_invalid; | |
150 | ||
0a883a05 | 151 | /* Disable read-ahead */ |
1e51764a AB |
152 | inode->i_mapping->backing_dev_info = &c->bdi; |
153 | ||
154 | switch (inode->i_mode & S_IFMT) { | |
155 | case S_IFREG: | |
156 | inode->i_mapping->a_ops = &ubifs_file_address_operations; | |
157 | inode->i_op = &ubifs_file_inode_operations; | |
158 | inode->i_fop = &ubifs_file_operations; | |
159 | if (ui->xattr) { | |
160 | ui->data = kmalloc(ui->data_len + 1, GFP_NOFS); | |
161 | if (!ui->data) { | |
162 | err = -ENOMEM; | |
163 | goto out_ino; | |
164 | } | |
165 | memcpy(ui->data, ino->data, ui->data_len); | |
166 | ((char *)ui->data)[ui->data_len] = '\0'; | |
167 | } else if (ui->data_len != 0) { | |
168 | err = 10; | |
169 | goto out_invalid; | |
170 | } | |
171 | break; | |
172 | case S_IFDIR: | |
173 | inode->i_op = &ubifs_dir_inode_operations; | |
174 | inode->i_fop = &ubifs_dir_operations; | |
175 | if (ui->data_len != 0) { | |
176 | err = 11; | |
177 | goto out_invalid; | |
178 | } | |
179 | break; | |
180 | case S_IFLNK: | |
181 | inode->i_op = &ubifs_symlink_inode_operations; | |
182 | if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) { | |
183 | err = 12; | |
184 | goto out_invalid; | |
185 | } | |
186 | ui->data = kmalloc(ui->data_len + 1, GFP_NOFS); | |
187 | if (!ui->data) { | |
188 | err = -ENOMEM; | |
189 | goto out_ino; | |
190 | } | |
191 | memcpy(ui->data, ino->data, ui->data_len); | |
192 | ((char *)ui->data)[ui->data_len] = '\0'; | |
193 | break; | |
194 | case S_IFBLK: | |
195 | case S_IFCHR: | |
196 | { | |
197 | dev_t rdev; | |
198 | union ubifs_dev_desc *dev; | |
199 | ||
200 | ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS); | |
201 | if (!ui->data) { | |
202 | err = -ENOMEM; | |
203 | goto out_ino; | |
204 | } | |
205 | ||
206 | dev = (union ubifs_dev_desc *)ino->data; | |
207 | if (ui->data_len == sizeof(dev->new)) | |
208 | rdev = new_decode_dev(le32_to_cpu(dev->new)); | |
209 | else if (ui->data_len == sizeof(dev->huge)) | |
210 | rdev = huge_decode_dev(le64_to_cpu(dev->huge)); | |
211 | else { | |
212 | err = 13; | |
213 | goto out_invalid; | |
214 | } | |
215 | memcpy(ui->data, ino->data, ui->data_len); | |
216 | inode->i_op = &ubifs_file_inode_operations; | |
217 | init_special_inode(inode, inode->i_mode, rdev); | |
218 | break; | |
219 | } | |
220 | case S_IFSOCK: | |
221 | case S_IFIFO: | |
222 | inode->i_op = &ubifs_file_inode_operations; | |
223 | init_special_inode(inode, inode->i_mode, 0); | |
224 | if (ui->data_len != 0) { | |
225 | err = 14; | |
226 | goto out_invalid; | |
227 | } | |
228 | break; | |
229 | default: | |
230 | err = 15; | |
231 | goto out_invalid; | |
232 | } | |
233 | ||
234 | kfree(ino); | |
235 | ubifs_set_inode_flags(inode); | |
236 | unlock_new_inode(inode); | |
237 | return inode; | |
238 | ||
239 | out_invalid: | |
240 | ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err); | |
241 | dbg_dump_node(c, ino); | |
242 | dbg_dump_inode(c, inode); | |
243 | err = -EINVAL; | |
244 | out_ino: | |
245 | kfree(ino); | |
246 | out: | |
247 | ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err); | |
248 | iget_failed(inode); | |
249 | return ERR_PTR(err); | |
250 | } | |
251 | ||
252 | static struct inode *ubifs_alloc_inode(struct super_block *sb) | |
253 | { | |
254 | struct ubifs_inode *ui; | |
255 | ||
256 | ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS); | |
257 | if (!ui) | |
258 | return NULL; | |
259 | ||
260 | memset((void *)ui + sizeof(struct inode), 0, | |
261 | sizeof(struct ubifs_inode) - sizeof(struct inode)); | |
262 | mutex_init(&ui->ui_mutex); | |
263 | spin_lock_init(&ui->ui_lock); | |
264 | return &ui->vfs_inode; | |
265 | }; | |
266 | ||
267 | static void ubifs_destroy_inode(struct inode *inode) | |
268 | { | |
269 | struct ubifs_inode *ui = ubifs_inode(inode); | |
270 | ||
271 | kfree(ui->data); | |
272 | kmem_cache_free(ubifs_inode_slab, inode); | |
273 | } | |
274 | ||
275 | /* | |
276 | * Note, Linux write-back code calls this without 'i_mutex'. | |
277 | */ | |
278 | static int ubifs_write_inode(struct inode *inode, int wait) | |
279 | { | |
fbfa6c88 | 280 | int err = 0; |
1e51764a AB |
281 | struct ubifs_info *c = inode->i_sb->s_fs_info; |
282 | struct ubifs_inode *ui = ubifs_inode(inode); | |
283 | ||
284 | ubifs_assert(!ui->xattr); | |
285 | if (is_bad_inode(inode)) | |
286 | return 0; | |
287 | ||
288 | mutex_lock(&ui->ui_mutex); | |
289 | /* | |
290 | * Due to races between write-back forced by budgeting | |
291 | * (see 'sync_some_inodes()') and pdflush write-back, the inode may | |
292 | * have already been synchronized, do not do this again. This might | |
293 | * also happen if it was synchronized in an VFS operation, e.g. | |
294 | * 'ubifs_link()'. | |
295 | */ | |
296 | if (!ui->dirty) { | |
297 | mutex_unlock(&ui->ui_mutex); | |
298 | return 0; | |
299 | } | |
300 | ||
fbfa6c88 AB |
301 | /* |
302 | * As an optimization, do not write orphan inodes to the media just | |
303 | * because this is not needed. | |
304 | */ | |
305 | dbg_gen("inode %lu, mode %#x, nlink %u", | |
306 | inode->i_ino, (int)inode->i_mode, inode->i_nlink); | |
307 | if (inode->i_nlink) { | |
1f28681a | 308 | err = ubifs_jnl_write_inode(c, inode); |
fbfa6c88 AB |
309 | if (err) |
310 | ubifs_err("can't write inode %lu, error %d", | |
311 | inode->i_ino, err); | |
312 | } | |
1e51764a AB |
313 | |
314 | ui->dirty = 0; | |
315 | mutex_unlock(&ui->ui_mutex); | |
316 | ubifs_release_dirty_inode_budget(c, ui); | |
317 | return err; | |
318 | } | |
319 | ||
320 | static void ubifs_delete_inode(struct inode *inode) | |
321 | { | |
322 | int err; | |
323 | struct ubifs_info *c = inode->i_sb->s_fs_info; | |
1e0f358e | 324 | struct ubifs_inode *ui = ubifs_inode(inode); |
1e51764a | 325 | |
1e0f358e | 326 | if (ui->xattr) |
1e51764a AB |
327 | /* |
328 | * Extended attribute inode deletions are fully handled in | |
329 | * 'ubifs_removexattr()'. These inodes are special and have | |
330 | * limited usage, so there is nothing to do here. | |
331 | */ | |
332 | goto out; | |
333 | ||
7d32c2bb | 334 | dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode); |
1e51764a AB |
335 | ubifs_assert(!atomic_read(&inode->i_count)); |
336 | ubifs_assert(inode->i_nlink == 0); | |
337 | ||
338 | truncate_inode_pages(&inode->i_data, 0); | |
339 | if (is_bad_inode(inode)) | |
340 | goto out; | |
341 | ||
1e0f358e | 342 | ui->ui_size = inode->i_size = 0; |
de94eb55 | 343 | err = ubifs_jnl_delete_inode(c, inode); |
1e51764a AB |
344 | if (err) |
345 | /* | |
346 | * Worst case we have a lost orphan inode wasting space, so a | |
0a883a05 | 347 | * simple error message is OK here. |
1e51764a | 348 | */ |
de94eb55 AB |
349 | ubifs_err("can't delete inode %lu, error %d", |
350 | inode->i_ino, err); | |
351 | ||
1e51764a | 352 | out: |
1e0f358e AB |
353 | if (ui->dirty) |
354 | ubifs_release_dirty_inode_budget(c, ui); | |
1e51764a AB |
355 | clear_inode(inode); |
356 | } | |
357 | ||
358 | static void ubifs_dirty_inode(struct inode *inode) | |
359 | { | |
360 | struct ubifs_inode *ui = ubifs_inode(inode); | |
361 | ||
362 | ubifs_assert(mutex_is_locked(&ui->ui_mutex)); | |
363 | if (!ui->dirty) { | |
364 | ui->dirty = 1; | |
365 | dbg_gen("inode %lu", inode->i_ino); | |
366 | } | |
367 | } | |
368 | ||
369 | static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
370 | { | |
371 | struct ubifs_info *c = dentry->d_sb->s_fs_info; | |
372 | unsigned long long free; | |
7c7cbadf | 373 | __le32 *uuid = (__le32 *)c->uuid; |
1e51764a | 374 | |
7dad181b | 375 | free = ubifs_get_free_space(c); |
1e51764a AB |
376 | dbg_gen("free space %lld bytes (%lld blocks)", |
377 | free, free >> UBIFS_BLOCK_SHIFT); | |
378 | ||
379 | buf->f_type = UBIFS_SUPER_MAGIC; | |
380 | buf->f_bsize = UBIFS_BLOCK_SIZE; | |
381 | buf->f_blocks = c->block_cnt; | |
382 | buf->f_bfree = free >> UBIFS_BLOCK_SHIFT; | |
383 | if (free > c->report_rp_size) | |
384 | buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT; | |
385 | else | |
386 | buf->f_bavail = 0; | |
387 | buf->f_files = 0; | |
388 | buf->f_ffree = 0; | |
389 | buf->f_namelen = UBIFS_MAX_NLEN; | |
7c7cbadf AB |
390 | buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]); |
391 | buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]); | |
1e51764a AB |
392 | return 0; |
393 | } | |
394 | ||
395 | static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt) | |
396 | { | |
397 | struct ubifs_info *c = mnt->mnt_sb->s_fs_info; | |
398 | ||
399 | if (c->mount_opts.unmount_mode == 2) | |
400 | seq_printf(s, ",fast_unmount"); | |
401 | else if (c->mount_opts.unmount_mode == 1) | |
402 | seq_printf(s, ",norm_unmount"); | |
403 | ||
4793e7c5 AH |
404 | if (c->mount_opts.bulk_read == 2) |
405 | seq_printf(s, ",bulk_read"); | |
406 | else if (c->mount_opts.bulk_read == 1) | |
407 | seq_printf(s, ",no_bulk_read"); | |
408 | ||
1e51764a AB |
409 | return 0; |
410 | } | |
411 | ||
412 | static int ubifs_sync_fs(struct super_block *sb, int wait) | |
413 | { | |
414 | struct ubifs_info *c = sb->s_fs_info; | |
415 | int i, ret = 0, err; | |
416 | ||
417 | if (c->jheads) | |
418 | for (i = 0; i < c->jhead_cnt; i++) { | |
419 | err = ubifs_wbuf_sync(&c->jheads[i].wbuf); | |
420 | if (err && !ret) | |
421 | ret = err; | |
422 | } | |
423 | /* | |
424 | * We ought to call sync for c->ubi but it does not have one. If it had | |
425 | * it would in turn call mtd->sync, however mtd operations are | |
426 | * synchronous anyway, so we don't lose any sleep here. | |
427 | */ | |
428 | return ret; | |
429 | } | |
430 | ||
431 | /** | |
432 | * init_constants_early - initialize UBIFS constants. | |
433 | * @c: UBIFS file-system description object | |
434 | * | |
435 | * This function initialize UBIFS constants which do not need the superblock to | |
436 | * be read. It also checks that the UBI volume satisfies basic UBIFS | |
437 | * requirements. Returns zero in case of success and a negative error code in | |
438 | * case of failure. | |
439 | */ | |
440 | static int init_constants_early(struct ubifs_info *c) | |
441 | { | |
442 | if (c->vi.corrupted) { | |
443 | ubifs_warn("UBI volume is corrupted - read-only mode"); | |
444 | c->ro_media = 1; | |
445 | } | |
446 | ||
447 | if (c->di.ro_mode) { | |
448 | ubifs_msg("read-only UBI device"); | |
449 | c->ro_media = 1; | |
450 | } | |
451 | ||
452 | if (c->vi.vol_type == UBI_STATIC_VOLUME) { | |
453 | ubifs_msg("static UBI volume - read-only mode"); | |
454 | c->ro_media = 1; | |
455 | } | |
456 | ||
457 | c->leb_cnt = c->vi.size; | |
458 | c->leb_size = c->vi.usable_leb_size; | |
459 | c->half_leb_size = c->leb_size / 2; | |
460 | c->min_io_size = c->di.min_io_size; | |
461 | c->min_io_shift = fls(c->min_io_size) - 1; | |
462 | ||
463 | if (c->leb_size < UBIFS_MIN_LEB_SZ) { | |
464 | ubifs_err("too small LEBs (%d bytes), min. is %d bytes", | |
465 | c->leb_size, UBIFS_MIN_LEB_SZ); | |
466 | return -EINVAL; | |
467 | } | |
468 | ||
469 | if (c->leb_cnt < UBIFS_MIN_LEB_CNT) { | |
470 | ubifs_err("too few LEBs (%d), min. is %d", | |
471 | c->leb_cnt, UBIFS_MIN_LEB_CNT); | |
472 | return -EINVAL; | |
473 | } | |
474 | ||
475 | if (!is_power_of_2(c->min_io_size)) { | |
476 | ubifs_err("bad min. I/O size %d", c->min_io_size); | |
477 | return -EINVAL; | |
478 | } | |
479 | ||
480 | /* | |
481 | * UBIFS aligns all node to 8-byte boundary, so to make function in | |
482 | * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is | |
483 | * less than 8. | |
484 | */ | |
485 | if (c->min_io_size < 8) { | |
486 | c->min_io_size = 8; | |
487 | c->min_io_shift = 3; | |
488 | } | |
489 | ||
490 | c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size); | |
491 | c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size); | |
492 | ||
493 | /* | |
494 | * Initialize node length ranges which are mostly needed for node | |
495 | * length validation. | |
496 | */ | |
497 | c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ; | |
498 | c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ; | |
499 | c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ; | |
500 | c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ; | |
501 | c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ; | |
502 | c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ; | |
503 | ||
504 | c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ; | |
505 | c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ; | |
506 | c->ranges[UBIFS_ORPH_NODE].min_len = | |
507 | UBIFS_ORPH_NODE_SZ + sizeof(__le64); | |
508 | c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size; | |
509 | c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ; | |
510 | c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ; | |
511 | c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ; | |
512 | c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ; | |
513 | c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ; | |
514 | c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ; | |
515 | /* | |
516 | * Minimum indexing node size is amended later when superblock is | |
517 | * read and the key length is known. | |
518 | */ | |
519 | c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ; | |
520 | /* | |
521 | * Maximum indexing node size is amended later when superblock is | |
522 | * read and the fanout is known. | |
523 | */ | |
524 | c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX; | |
525 | ||
526 | /* | |
527 | * Initialize dead and dark LEB space watermarks. | |
528 | * | |
529 | * Dead space is the space which cannot be used. Its watermark is | |
530 | * equivalent to min. I/O unit or minimum node size if it is greater | |
531 | * then min. I/O unit. | |
532 | * | |
533 | * Dark space is the space which might be used, or might not, depending | |
534 | * on which node should be written to the LEB. Its watermark is | |
535 | * equivalent to maximum UBIFS node size. | |
536 | */ | |
537 | c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size); | |
538 | c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size); | |
539 | ||
9bbb5726 AB |
540 | /* |
541 | * Calculate how many bytes would be wasted at the end of LEB if it was | |
542 | * fully filled with data nodes of maximum size. This is used in | |
543 | * calculations when reporting free space. | |
544 | */ | |
545 | c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ; | |
4793e7c5 AH |
546 | /* Buffer size for bulk-reads */ |
547 | c->bulk_read_buf_size = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ; | |
548 | if (c->bulk_read_buf_size > c->leb_size) | |
549 | c->bulk_read_buf_size = c->leb_size; | |
550 | if (c->bulk_read_buf_size > 128 * 1024) { | |
551 | /* Check if we can kmalloc more than 128KiB */ | |
552 | void *try = kmalloc(c->bulk_read_buf_size, GFP_KERNEL); | |
553 | ||
554 | kfree(try); | |
555 | if (!try) | |
556 | c->bulk_read_buf_size = 128 * 1024; | |
557 | } | |
1e51764a AB |
558 | return 0; |
559 | } | |
560 | ||
561 | /** | |
562 | * bud_wbuf_callback - bud LEB write-buffer synchronization call-back. | |
563 | * @c: UBIFS file-system description object | |
564 | * @lnum: LEB the write-buffer was synchronized to | |
565 | * @free: how many free bytes left in this LEB | |
566 | * @pad: how many bytes were padded | |
567 | * | |
568 | * This is a callback function which is called by the I/O unit when the | |
569 | * write-buffer is synchronized. We need this to correctly maintain space | |
570 | * accounting in bud logical eraseblocks. This function returns zero in case of | |
571 | * success and a negative error code in case of failure. | |
572 | * | |
573 | * This function actually belongs to the journal, but we keep it here because | |
574 | * we want to keep it static. | |
575 | */ | |
576 | static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad) | |
577 | { | |
578 | return ubifs_update_one_lp(c, lnum, free, pad, 0, 0); | |
579 | } | |
580 | ||
581 | /* | |
582 | * init_constants_late - initialize UBIFS constants. | |
583 | * @c: UBIFS file-system description object | |
584 | * | |
585 | * This is a helper function which initializes various UBIFS constants after | |
586 | * the superblock has been read. It also checks various UBIFS parameters and | |
587 | * makes sure they are all right. Returns zero in case of success and a | |
588 | * negative error code in case of failure. | |
589 | */ | |
590 | static int init_constants_late(struct ubifs_info *c) | |
591 | { | |
592 | int tmp, err; | |
593 | uint64_t tmp64; | |
594 | ||
595 | c->main_bytes = (long long)c->main_lebs * c->leb_size; | |
596 | c->max_znode_sz = sizeof(struct ubifs_znode) + | |
597 | c->fanout * sizeof(struct ubifs_zbranch); | |
598 | ||
599 | tmp = ubifs_idx_node_sz(c, 1); | |
600 | c->ranges[UBIFS_IDX_NODE].min_len = tmp; | |
601 | c->min_idx_node_sz = ALIGN(tmp, 8); | |
602 | ||
603 | tmp = ubifs_idx_node_sz(c, c->fanout); | |
604 | c->ranges[UBIFS_IDX_NODE].max_len = tmp; | |
605 | c->max_idx_node_sz = ALIGN(tmp, 8); | |
606 | ||
607 | /* Make sure LEB size is large enough to fit full commit */ | |
608 | tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt; | |
609 | tmp = ALIGN(tmp, c->min_io_size); | |
610 | if (tmp > c->leb_size) { | |
611 | dbg_err("too small LEB size %d, at least %d needed", | |
612 | c->leb_size, tmp); | |
613 | return -EINVAL; | |
614 | } | |
615 | ||
616 | /* | |
617 | * Make sure that the log is large enough to fit reference nodes for | |
618 | * all buds plus one reserved LEB. | |
619 | */ | |
620 | tmp64 = c->max_bud_bytes; | |
621 | tmp = do_div(tmp64, c->leb_size); | |
622 | c->max_bud_cnt = tmp64 + !!tmp; | |
623 | tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1); | |
624 | tmp /= c->leb_size; | |
625 | tmp += 1; | |
626 | if (c->log_lebs < tmp) { | |
627 | dbg_err("too small log %d LEBs, required min. %d LEBs", | |
628 | c->log_lebs, tmp); | |
629 | return -EINVAL; | |
630 | } | |
631 | ||
632 | /* | |
633 | * When budgeting we assume worst-case scenarios when the pages are not | |
634 | * be compressed and direntries are of the maximum size. | |
635 | * | |
636 | * Note, data, which may be stored in inodes is budgeted separately, so | |
637 | * it is not included into 'c->inode_budget'. | |
638 | */ | |
639 | c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE; | |
640 | c->inode_budget = UBIFS_INO_NODE_SZ; | |
641 | c->dent_budget = UBIFS_MAX_DENT_NODE_SZ; | |
642 | ||
643 | /* | |
644 | * When the amount of flash space used by buds becomes | |
645 | * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit. | |
646 | * The writers are unblocked when the commit is finished. To avoid | |
647 | * writers to be blocked UBIFS initiates background commit in advance, | |
648 | * when number of bud bytes becomes above the limit defined below. | |
649 | */ | |
650 | c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4; | |
651 | ||
652 | /* | |
653 | * Ensure minimum journal size. All the bytes in the journal heads are | |
654 | * considered to be used, when calculating the current journal usage. | |
655 | * Consequently, if the journal is too small, UBIFS will treat it as | |
656 | * always full. | |
657 | */ | |
658 | tmp64 = (uint64_t)(c->jhead_cnt + 1) * c->leb_size + 1; | |
659 | if (c->bg_bud_bytes < tmp64) | |
660 | c->bg_bud_bytes = tmp64; | |
661 | if (c->max_bud_bytes < tmp64 + c->leb_size) | |
662 | c->max_bud_bytes = tmp64 + c->leb_size; | |
663 | ||
664 | err = ubifs_calc_lpt_geom(c); | |
665 | if (err) | |
666 | return err; | |
667 | ||
668 | c->min_idx_lebs = ubifs_calc_min_idx_lebs(c); | |
669 | ||
670 | /* | |
671 | * Calculate total amount of FS blocks. This number is not used | |
672 | * internally because it does not make much sense for UBIFS, but it is | |
673 | * necessary to report something for the 'statfs()' call. | |
674 | * | |
7dad181b AB |
675 | * Subtract the LEB reserved for GC, the LEB which is reserved for |
676 | * deletions, and assume only one journal head is available. | |
1e51764a | 677 | */ |
7dad181b AB |
678 | tmp64 = c->main_lebs - 2 - c->jhead_cnt + 1; |
679 | tmp64 *= (uint64_t)c->leb_size - c->leb_overhead; | |
1e51764a AB |
680 | tmp64 = ubifs_reported_space(c, tmp64); |
681 | c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT; | |
682 | ||
683 | return 0; | |
684 | } | |
685 | ||
686 | /** | |
687 | * take_gc_lnum - reserve GC LEB. | |
688 | * @c: UBIFS file-system description object | |
689 | * | |
690 | * This function ensures that the LEB reserved for garbage collection is | |
691 | * unmapped and is marked as "taken" in lprops. We also have to set free space | |
692 | * to LEB size and dirty space to zero, because lprops may contain out-of-date | |
693 | * information if the file-system was un-mounted before it has been committed. | |
694 | * This function returns zero in case of success and a negative error code in | |
695 | * case of failure. | |
696 | */ | |
697 | static int take_gc_lnum(struct ubifs_info *c) | |
698 | { | |
699 | int err; | |
700 | ||
701 | if (c->gc_lnum == -1) { | |
702 | ubifs_err("no LEB for GC"); | |
703 | return -EINVAL; | |
704 | } | |
705 | ||
706 | err = ubifs_leb_unmap(c, c->gc_lnum); | |
707 | if (err) | |
708 | return err; | |
709 | ||
710 | /* And we have to tell lprops that this LEB is taken */ | |
711 | err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0, | |
712 | LPROPS_TAKEN, 0, 0); | |
713 | return err; | |
714 | } | |
715 | ||
716 | /** | |
717 | * alloc_wbufs - allocate write-buffers. | |
718 | * @c: UBIFS file-system description object | |
719 | * | |
720 | * This helper function allocates and initializes UBIFS write-buffers. Returns | |
721 | * zero in case of success and %-ENOMEM in case of failure. | |
722 | */ | |
723 | static int alloc_wbufs(struct ubifs_info *c) | |
724 | { | |
725 | int i, err; | |
726 | ||
727 | c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead), | |
728 | GFP_KERNEL); | |
729 | if (!c->jheads) | |
730 | return -ENOMEM; | |
731 | ||
732 | /* Initialize journal heads */ | |
733 | for (i = 0; i < c->jhead_cnt; i++) { | |
734 | INIT_LIST_HEAD(&c->jheads[i].buds_list); | |
735 | err = ubifs_wbuf_init(c, &c->jheads[i].wbuf); | |
736 | if (err) | |
737 | return err; | |
738 | ||
739 | c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback; | |
740 | c->jheads[i].wbuf.jhead = i; | |
741 | } | |
742 | ||
743 | c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM; | |
744 | /* | |
745 | * Garbage Collector head likely contains long-term data and | |
746 | * does not need to be synchronized by timer. | |
747 | */ | |
748 | c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM; | |
749 | c->jheads[GCHD].wbuf.timeout = 0; | |
750 | ||
751 | return 0; | |
752 | } | |
753 | ||
754 | /** | |
755 | * free_wbufs - free write-buffers. | |
756 | * @c: UBIFS file-system description object | |
757 | */ | |
758 | static void free_wbufs(struct ubifs_info *c) | |
759 | { | |
760 | int i; | |
761 | ||
762 | if (c->jheads) { | |
763 | for (i = 0; i < c->jhead_cnt; i++) { | |
764 | kfree(c->jheads[i].wbuf.buf); | |
765 | kfree(c->jheads[i].wbuf.inodes); | |
766 | } | |
767 | kfree(c->jheads); | |
768 | c->jheads = NULL; | |
769 | } | |
770 | } | |
771 | ||
772 | /** | |
773 | * free_orphans - free orphans. | |
774 | * @c: UBIFS file-system description object | |
775 | */ | |
776 | static void free_orphans(struct ubifs_info *c) | |
777 | { | |
778 | struct ubifs_orphan *orph; | |
779 | ||
780 | while (c->orph_dnext) { | |
781 | orph = c->orph_dnext; | |
782 | c->orph_dnext = orph->dnext; | |
783 | list_del(&orph->list); | |
784 | kfree(orph); | |
785 | } | |
786 | ||
787 | while (!list_empty(&c->orph_list)) { | |
788 | orph = list_entry(c->orph_list.next, struct ubifs_orphan, list); | |
789 | list_del(&orph->list); | |
790 | kfree(orph); | |
791 | dbg_err("orphan list not empty at unmount"); | |
792 | } | |
793 | ||
794 | vfree(c->orph_buf); | |
795 | c->orph_buf = NULL; | |
796 | } | |
797 | ||
798 | /** | |
799 | * free_buds - free per-bud objects. | |
800 | * @c: UBIFS file-system description object | |
801 | */ | |
802 | static void free_buds(struct ubifs_info *c) | |
803 | { | |
804 | struct rb_node *this = c->buds.rb_node; | |
805 | struct ubifs_bud *bud; | |
806 | ||
807 | while (this) { | |
808 | if (this->rb_left) | |
809 | this = this->rb_left; | |
810 | else if (this->rb_right) | |
811 | this = this->rb_right; | |
812 | else { | |
813 | bud = rb_entry(this, struct ubifs_bud, rb); | |
814 | this = rb_parent(this); | |
815 | if (this) { | |
816 | if (this->rb_left == &bud->rb) | |
817 | this->rb_left = NULL; | |
818 | else | |
819 | this->rb_right = NULL; | |
820 | } | |
821 | kfree(bud); | |
822 | } | |
823 | } | |
824 | } | |
825 | ||
826 | /** | |
827 | * check_volume_empty - check if the UBI volume is empty. | |
828 | * @c: UBIFS file-system description object | |
829 | * | |
830 | * This function checks if the UBIFS volume is empty by looking if its LEBs are | |
831 | * mapped or not. The result of checking is stored in the @c->empty variable. | |
832 | * Returns zero in case of success and a negative error code in case of | |
833 | * failure. | |
834 | */ | |
835 | static int check_volume_empty(struct ubifs_info *c) | |
836 | { | |
837 | int lnum, err; | |
838 | ||
839 | c->empty = 1; | |
840 | for (lnum = 0; lnum < c->leb_cnt; lnum++) { | |
841 | err = ubi_is_mapped(c->ubi, lnum); | |
842 | if (unlikely(err < 0)) | |
843 | return err; | |
844 | if (err == 1) { | |
845 | c->empty = 0; | |
846 | break; | |
847 | } | |
848 | ||
849 | cond_resched(); | |
850 | } | |
851 | ||
852 | return 0; | |
853 | } | |
854 | ||
855 | /* | |
856 | * UBIFS mount options. | |
857 | * | |
858 | * Opt_fast_unmount: do not run a journal commit before un-mounting | |
859 | * Opt_norm_unmount: run a journal commit before un-mounting | |
4793e7c5 AH |
860 | * Opt_bulk_read: enable bulk-reads |
861 | * Opt_no_bulk_read: disable bulk-reads | |
1e51764a AB |
862 | * Opt_err: just end of array marker |
863 | */ | |
864 | enum { | |
865 | Opt_fast_unmount, | |
866 | Opt_norm_unmount, | |
4793e7c5 AH |
867 | Opt_bulk_read, |
868 | Opt_no_bulk_read, | |
1e51764a AB |
869 | Opt_err, |
870 | }; | |
871 | ||
872 | static match_table_t tokens = { | |
873 | {Opt_fast_unmount, "fast_unmount"}, | |
874 | {Opt_norm_unmount, "norm_unmount"}, | |
4793e7c5 AH |
875 | {Opt_bulk_read, "bulk_read"}, |
876 | {Opt_no_bulk_read, "no_bulk_read"}, | |
1e51764a AB |
877 | {Opt_err, NULL}, |
878 | }; | |
879 | ||
880 | /** | |
881 | * ubifs_parse_options - parse mount parameters. | |
882 | * @c: UBIFS file-system description object | |
883 | * @options: parameters to parse | |
884 | * @is_remount: non-zero if this is FS re-mount | |
885 | * | |
886 | * This function parses UBIFS mount options and returns zero in case success | |
887 | * and a negative error code in case of failure. | |
888 | */ | |
889 | static int ubifs_parse_options(struct ubifs_info *c, char *options, | |
890 | int is_remount) | |
891 | { | |
892 | char *p; | |
893 | substring_t args[MAX_OPT_ARGS]; | |
894 | ||
895 | if (!options) | |
896 | return 0; | |
897 | ||
898 | while ((p = strsep(&options, ","))) { | |
899 | int token; | |
900 | ||
901 | if (!*p) | |
902 | continue; | |
903 | ||
904 | token = match_token(p, tokens, args); | |
905 | switch (token) { | |
906 | case Opt_fast_unmount: | |
907 | c->mount_opts.unmount_mode = 2; | |
908 | c->fast_unmount = 1; | |
909 | break; | |
910 | case Opt_norm_unmount: | |
911 | c->mount_opts.unmount_mode = 1; | |
912 | c->fast_unmount = 0; | |
913 | break; | |
4793e7c5 AH |
914 | case Opt_bulk_read: |
915 | c->mount_opts.bulk_read = 2; | |
916 | c->bulk_read = 1; | |
917 | break; | |
918 | case Opt_no_bulk_read: | |
919 | c->mount_opts.bulk_read = 1; | |
920 | c->bulk_read = 0; | |
921 | break; | |
1e51764a AB |
922 | default: |
923 | ubifs_err("unrecognized mount option \"%s\" " | |
924 | "or missing value", p); | |
925 | return -EINVAL; | |
926 | } | |
927 | } | |
928 | ||
929 | return 0; | |
930 | } | |
931 | ||
932 | /** | |
933 | * destroy_journal - destroy journal data structures. | |
934 | * @c: UBIFS file-system description object | |
935 | * | |
936 | * This function destroys journal data structures including those that may have | |
937 | * been created by recovery functions. | |
938 | */ | |
939 | static void destroy_journal(struct ubifs_info *c) | |
940 | { | |
941 | while (!list_empty(&c->unclean_leb_list)) { | |
942 | struct ubifs_unclean_leb *ucleb; | |
943 | ||
944 | ucleb = list_entry(c->unclean_leb_list.next, | |
945 | struct ubifs_unclean_leb, list); | |
946 | list_del(&ucleb->list); | |
947 | kfree(ucleb); | |
948 | } | |
949 | while (!list_empty(&c->old_buds)) { | |
950 | struct ubifs_bud *bud; | |
951 | ||
952 | bud = list_entry(c->old_buds.next, struct ubifs_bud, list); | |
953 | list_del(&bud->list); | |
954 | kfree(bud); | |
955 | } | |
956 | ubifs_destroy_idx_gc(c); | |
957 | ubifs_destroy_size_tree(c); | |
958 | ubifs_tnc_close(c); | |
959 | free_buds(c); | |
960 | } | |
961 | ||
962 | /** | |
963 | * mount_ubifs - mount UBIFS file-system. | |
964 | * @c: UBIFS file-system description object | |
965 | * | |
966 | * This function mounts UBIFS file system. Returns zero in case of success and | |
967 | * a negative error code in case of failure. | |
968 | * | |
969 | * Note, the function does not de-allocate resources it it fails half way | |
970 | * through, and the caller has to do this instead. | |
971 | */ | |
972 | static int mount_ubifs(struct ubifs_info *c) | |
973 | { | |
974 | struct super_block *sb = c->vfs_sb; | |
975 | int err, mounted_read_only = (sb->s_flags & MS_RDONLY); | |
976 | long long x; | |
977 | size_t sz; | |
978 | ||
979 | err = init_constants_early(c); | |
980 | if (err) | |
981 | return err; | |
982 | ||
983 | #ifdef CONFIG_UBIFS_FS_DEBUG | |
984 | c->dbg_buf = vmalloc(c->leb_size); | |
985 | if (!c->dbg_buf) | |
986 | return -ENOMEM; | |
987 | #endif | |
988 | ||
989 | err = check_volume_empty(c); | |
990 | if (err) | |
991 | goto out_free; | |
992 | ||
993 | if (c->empty && (mounted_read_only || c->ro_media)) { | |
994 | /* | |
995 | * This UBI volume is empty, and read-only, or the file system | |
996 | * is mounted read-only - we cannot format it. | |
997 | */ | |
998 | ubifs_err("can't format empty UBI volume: read-only %s", | |
999 | c->ro_media ? "UBI volume" : "mount"); | |
1000 | err = -EROFS; | |
1001 | goto out_free; | |
1002 | } | |
1003 | ||
1004 | if (c->ro_media && !mounted_read_only) { | |
1005 | ubifs_err("cannot mount read-write - read-only media"); | |
1006 | err = -EROFS; | |
1007 | goto out_free; | |
1008 | } | |
1009 | ||
1010 | /* | |
1011 | * The requirement for the buffer is that it should fit indexing B-tree | |
1012 | * height amount of integers. We assume the height if the TNC tree will | |
1013 | * never exceed 64. | |
1014 | */ | |
1015 | err = -ENOMEM; | |
1016 | c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL); | |
1017 | if (!c->bottom_up_buf) | |
1018 | goto out_free; | |
1019 | ||
1020 | c->sbuf = vmalloc(c->leb_size); | |
1021 | if (!c->sbuf) | |
1022 | goto out_free; | |
1023 | ||
1024 | if (!mounted_read_only) { | |
1025 | c->ileb_buf = vmalloc(c->leb_size); | |
1026 | if (!c->ileb_buf) | |
1027 | goto out_free; | |
1028 | } | |
1029 | ||
1030 | err = ubifs_read_superblock(c); | |
1031 | if (err) | |
1032 | goto out_free; | |
1033 | ||
1034 | /* | |
1035 | * Make sure the compressor which is set as the default on in the | |
1036 | * superblock was actually compiled in. | |
1037 | */ | |
1038 | if (!ubifs_compr_present(c->default_compr)) { | |
1039 | ubifs_warn("'%s' compressor is set by superblock, but not " | |
1040 | "compiled in", ubifs_compr_name(c->default_compr)); | |
1041 | c->default_compr = UBIFS_COMPR_NONE; | |
1042 | } | |
1043 | ||
1044 | dbg_failure_mode_registration(c); | |
1045 | ||
1046 | err = init_constants_late(c); | |
1047 | if (err) | |
1048 | goto out_dereg; | |
1049 | ||
1050 | sz = ALIGN(c->max_idx_node_sz, c->min_io_size); | |
1051 | sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size); | |
1052 | c->cbuf = kmalloc(sz, GFP_NOFS); | |
1053 | if (!c->cbuf) { | |
1054 | err = -ENOMEM; | |
1055 | goto out_dereg; | |
1056 | } | |
1057 | ||
0855f310 | 1058 | sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id); |
1e51764a AB |
1059 | if (!mounted_read_only) { |
1060 | err = alloc_wbufs(c); | |
1061 | if (err) | |
1062 | goto out_cbuf; | |
1063 | ||
1064 | /* Create background thread */ | |
1e51764a | 1065 | c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name); |
1e51764a AB |
1066 | if (IS_ERR(c->bgt)) { |
1067 | err = PTR_ERR(c->bgt); | |
1068 | c->bgt = NULL; | |
1069 | ubifs_err("cannot spawn \"%s\", error %d", | |
1070 | c->bgt_name, err); | |
1071 | goto out_wbufs; | |
1072 | } | |
1073 | wake_up_process(c->bgt); | |
1074 | } | |
1075 | ||
1076 | err = ubifs_read_master(c); | |
1077 | if (err) | |
1078 | goto out_master; | |
1079 | ||
1080 | if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) { | |
1081 | ubifs_msg("recovery needed"); | |
1082 | c->need_recovery = 1; | |
1083 | if (!mounted_read_only) { | |
1084 | err = ubifs_recover_inl_heads(c, c->sbuf); | |
1085 | if (err) | |
1086 | goto out_master; | |
1087 | } | |
1088 | } else if (!mounted_read_only) { | |
1089 | /* | |
1090 | * Set the "dirty" flag so that if we reboot uncleanly we | |
1091 | * will notice this immediately on the next mount. | |
1092 | */ | |
1093 | c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY); | |
1094 | err = ubifs_write_master(c); | |
1095 | if (err) | |
1096 | goto out_master; | |
1097 | } | |
1098 | ||
1099 | err = ubifs_lpt_init(c, 1, !mounted_read_only); | |
1100 | if (err) | |
1101 | goto out_lpt; | |
1102 | ||
1103 | err = dbg_check_idx_size(c, c->old_idx_sz); | |
1104 | if (err) | |
1105 | goto out_lpt; | |
1106 | ||
1107 | err = ubifs_replay_journal(c); | |
1108 | if (err) | |
1109 | goto out_journal; | |
1110 | ||
1111 | err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only); | |
1112 | if (err) | |
1113 | goto out_orphans; | |
1114 | ||
1115 | if (!mounted_read_only) { | |
1116 | int lnum; | |
1117 | ||
1118 | /* Check for enough free space */ | |
1119 | if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) { | |
1120 | ubifs_err("insufficient available space"); | |
1121 | err = -EINVAL; | |
1122 | goto out_orphans; | |
1123 | } | |
1124 | ||
1125 | /* Check for enough log space */ | |
1126 | lnum = c->lhead_lnum + 1; | |
1127 | if (lnum >= UBIFS_LOG_LNUM + c->log_lebs) | |
1128 | lnum = UBIFS_LOG_LNUM; | |
1129 | if (lnum == c->ltail_lnum) { | |
1130 | err = ubifs_consolidate_log(c); | |
1131 | if (err) | |
1132 | goto out_orphans; | |
1133 | } | |
1134 | ||
1135 | if (c->need_recovery) { | |
1136 | err = ubifs_recover_size(c); | |
1137 | if (err) | |
1138 | goto out_orphans; | |
1139 | err = ubifs_rcvry_gc_commit(c); | |
1140 | } else | |
1141 | err = take_gc_lnum(c); | |
1142 | if (err) | |
1143 | goto out_orphans; | |
1144 | ||
1145 | err = dbg_check_lprops(c); | |
1146 | if (err) | |
1147 | goto out_orphans; | |
1148 | } else if (c->need_recovery) { | |
1149 | err = ubifs_recover_size(c); | |
1150 | if (err) | |
1151 | goto out_orphans; | |
1152 | } | |
1153 | ||
1154 | spin_lock(&ubifs_infos_lock); | |
1155 | list_add_tail(&c->infos_list, &ubifs_infos); | |
1156 | spin_unlock(&ubifs_infos_lock); | |
1157 | ||
1158 | if (c->need_recovery) { | |
1159 | if (mounted_read_only) | |
1160 | ubifs_msg("recovery deferred"); | |
1161 | else { | |
1162 | c->need_recovery = 0; | |
1163 | ubifs_msg("recovery completed"); | |
1164 | } | |
1165 | } | |
1166 | ||
1167 | err = dbg_check_filesystem(c); | |
1168 | if (err) | |
1169 | goto out_infos; | |
1170 | ||
ce769caa AB |
1171 | ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"", |
1172 | c->vi.ubi_num, c->vi.vol_id, c->vi.name); | |
1e51764a AB |
1173 | if (mounted_read_only) |
1174 | ubifs_msg("mounted read-only"); | |
1175 | x = (long long)c->main_lebs * c->leb_size; | |
948cfb21 AB |
1176 | ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d " |
1177 | "LEBs)", x, x >> 10, x >> 20, c->main_lebs); | |
1e51764a | 1178 | x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes; |
948cfb21 AB |
1179 | ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d " |
1180 | "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt); | |
1181 | ubifs_msg("media format: %d (latest is %d)", | |
1e51764a | 1182 | c->fmt_version, UBIFS_FORMAT_VERSION); |
948cfb21 AB |
1183 | ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr)); |
1184 | ubifs_msg("reserved pool size: %llu bytes (%llu KiB)", | |
1185 | c->report_rp_size, c->report_rp_size >> 10); | |
1e51764a AB |
1186 | |
1187 | dbg_msg("compiled on: " __DATE__ " at " __TIME__); | |
1188 | dbg_msg("min. I/O unit size: %d bytes", c->min_io_size); | |
1189 | dbg_msg("LEB size: %d bytes (%d KiB)", | |
948cfb21 | 1190 | c->leb_size, c->leb_size >> 10); |
1e51764a AB |
1191 | dbg_msg("data journal heads: %d", |
1192 | c->jhead_cnt - NONDATA_JHEADS_CNT); | |
1193 | dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X" | |
1194 | "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X", | |
1195 | c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3], | |
1196 | c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7], | |
1197 | c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11], | |
1198 | c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]); | |
1199 | dbg_msg("fast unmount: %d", c->fast_unmount); | |
1200 | dbg_msg("big_lpt %d", c->big_lpt); | |
1201 | dbg_msg("log LEBs: %d (%d - %d)", | |
1202 | c->log_lebs, UBIFS_LOG_LNUM, c->log_last); | |
1203 | dbg_msg("LPT area LEBs: %d (%d - %d)", | |
1204 | c->lpt_lebs, c->lpt_first, c->lpt_last); | |
1205 | dbg_msg("orphan area LEBs: %d (%d - %d)", | |
1206 | c->orph_lebs, c->orph_first, c->orph_last); | |
1207 | dbg_msg("main area LEBs: %d (%d - %d)", | |
1208 | c->main_lebs, c->main_first, c->leb_cnt - 1); | |
1209 | dbg_msg("index LEBs: %d", c->lst.idx_lebs); | |
1210 | dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)", | |
1211 | c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20); | |
1212 | dbg_msg("key hash type: %d", c->key_hash_type); | |
1213 | dbg_msg("tree fanout: %d", c->fanout); | |
1214 | dbg_msg("reserved GC LEB: %d", c->gc_lnum); | |
1215 | dbg_msg("first main LEB: %d", c->main_first); | |
1216 | dbg_msg("dead watermark: %d", c->dead_wm); | |
1217 | dbg_msg("dark watermark: %d", c->dark_wm); | |
1218 | x = (long long)c->main_lebs * c->dark_wm; | |
1219 | dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)", | |
1220 | x, x >> 10, x >> 20); | |
1221 | dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)", | |
1222 | c->max_bud_bytes, c->max_bud_bytes >> 10, | |
1223 | c->max_bud_bytes >> 20); | |
1224 | dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)", | |
1225 | c->bg_bud_bytes, c->bg_bud_bytes >> 10, | |
1226 | c->bg_bud_bytes >> 20); | |
1227 | dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)", | |
1228 | c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20); | |
1229 | dbg_msg("max. seq. number: %llu", c->max_sqnum); | |
1230 | dbg_msg("commit number: %llu", c->cmt_no); | |
1231 | ||
1232 | return 0; | |
1233 | ||
1234 | out_infos: | |
1235 | spin_lock(&ubifs_infos_lock); | |
1236 | list_del(&c->infos_list); | |
1237 | spin_unlock(&ubifs_infos_lock); | |
1238 | out_orphans: | |
1239 | free_orphans(c); | |
1240 | out_journal: | |
1241 | destroy_journal(c); | |
1242 | out_lpt: | |
1243 | ubifs_lpt_free(c, 0); | |
1244 | out_master: | |
1245 | kfree(c->mst_node); | |
1246 | kfree(c->rcvrd_mst_node); | |
1247 | if (c->bgt) | |
1248 | kthread_stop(c->bgt); | |
1249 | out_wbufs: | |
1250 | free_wbufs(c); | |
1251 | out_cbuf: | |
1252 | kfree(c->cbuf); | |
1253 | out_dereg: | |
1254 | dbg_failure_mode_deregistration(c); | |
1255 | out_free: | |
1256 | vfree(c->ileb_buf); | |
1257 | vfree(c->sbuf); | |
1258 | kfree(c->bottom_up_buf); | |
1259 | UBIFS_DBG(vfree(c->dbg_buf)); | |
1260 | return err; | |
1261 | } | |
1262 | ||
1263 | /** | |
1264 | * ubifs_umount - un-mount UBIFS file-system. | |
1265 | * @c: UBIFS file-system description object | |
1266 | * | |
1267 | * Note, this function is called to free allocated resourced when un-mounting, | |
1268 | * as well as free resources when an error occurred while we were half way | |
1269 | * through mounting (error path cleanup function). So it has to make sure the | |
1270 | * resource was actually allocated before freeing it. | |
1271 | */ | |
1272 | static void ubifs_umount(struct ubifs_info *c) | |
1273 | { | |
1274 | dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num, | |
1275 | c->vi.vol_id); | |
1276 | ||
1277 | spin_lock(&ubifs_infos_lock); | |
1278 | list_del(&c->infos_list); | |
1279 | spin_unlock(&ubifs_infos_lock); | |
1280 | ||
1281 | if (c->bgt) | |
1282 | kthread_stop(c->bgt); | |
1283 | ||
1284 | destroy_journal(c); | |
1285 | free_wbufs(c); | |
1286 | free_orphans(c); | |
1287 | ubifs_lpt_free(c, 0); | |
1288 | ||
1289 | kfree(c->cbuf); | |
1290 | kfree(c->rcvrd_mst_node); | |
1291 | kfree(c->mst_node); | |
1292 | vfree(c->sbuf); | |
1293 | kfree(c->bottom_up_buf); | |
1294 | UBIFS_DBG(vfree(c->dbg_buf)); | |
1295 | vfree(c->ileb_buf); | |
1296 | dbg_failure_mode_deregistration(c); | |
1297 | } | |
1298 | ||
1299 | /** | |
1300 | * ubifs_remount_rw - re-mount in read-write mode. | |
1301 | * @c: UBIFS file-system description object | |
1302 | * | |
1303 | * UBIFS avoids allocating many unnecessary resources when mounted in read-only | |
1304 | * mode. This function allocates the needed resources and re-mounts UBIFS in | |
1305 | * read-write mode. | |
1306 | */ | |
1307 | static int ubifs_remount_rw(struct ubifs_info *c) | |
1308 | { | |
1309 | int err, lnum; | |
1310 | ||
1311 | if (c->ro_media) | |
1312 | return -EINVAL; | |
1313 | ||
1314 | mutex_lock(&c->umount_mutex); | |
1315 | c->remounting_rw = 1; | |
1316 | ||
1317 | /* Check for enough free space */ | |
1318 | if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) { | |
1319 | ubifs_err("insufficient available space"); | |
1320 | err = -EINVAL; | |
1321 | goto out; | |
1322 | } | |
1323 | ||
1324 | if (c->old_leb_cnt != c->leb_cnt) { | |
1325 | struct ubifs_sb_node *sup; | |
1326 | ||
1327 | sup = ubifs_read_sb_node(c); | |
1328 | if (IS_ERR(sup)) { | |
1329 | err = PTR_ERR(sup); | |
1330 | goto out; | |
1331 | } | |
1332 | sup->leb_cnt = cpu_to_le32(c->leb_cnt); | |
1333 | err = ubifs_write_sb_node(c, sup); | |
1334 | if (err) | |
1335 | goto out; | |
1336 | } | |
1337 | ||
1338 | if (c->need_recovery) { | |
1339 | ubifs_msg("completing deferred recovery"); | |
1340 | err = ubifs_write_rcvrd_mst_node(c); | |
1341 | if (err) | |
1342 | goto out; | |
1343 | err = ubifs_recover_size(c); | |
1344 | if (err) | |
1345 | goto out; | |
1346 | err = ubifs_clean_lebs(c, c->sbuf); | |
1347 | if (err) | |
1348 | goto out; | |
1349 | err = ubifs_recover_inl_heads(c, c->sbuf); | |
1350 | if (err) | |
1351 | goto out; | |
1352 | } | |
1353 | ||
1354 | if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) { | |
1355 | c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY); | |
1356 | err = ubifs_write_master(c); | |
1357 | if (err) | |
1358 | goto out; | |
1359 | } | |
1360 | ||
1361 | c->ileb_buf = vmalloc(c->leb_size); | |
1362 | if (!c->ileb_buf) { | |
1363 | err = -ENOMEM; | |
1364 | goto out; | |
1365 | } | |
1366 | ||
1367 | err = ubifs_lpt_init(c, 0, 1); | |
1368 | if (err) | |
1369 | goto out; | |
1370 | ||
1371 | err = alloc_wbufs(c); | |
1372 | if (err) | |
1373 | goto out; | |
1374 | ||
1375 | ubifs_create_buds_lists(c); | |
1376 | ||
1377 | /* Create background thread */ | |
1378 | c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name); | |
1e51764a AB |
1379 | if (IS_ERR(c->bgt)) { |
1380 | err = PTR_ERR(c->bgt); | |
1381 | c->bgt = NULL; | |
1382 | ubifs_err("cannot spawn \"%s\", error %d", | |
1383 | c->bgt_name, err); | |
1384 | return err; | |
1385 | } | |
1386 | wake_up_process(c->bgt); | |
1387 | ||
1388 | c->orph_buf = vmalloc(c->leb_size); | |
1389 | if (!c->orph_buf) | |
1390 | return -ENOMEM; | |
1391 | ||
1392 | /* Check for enough log space */ | |
1393 | lnum = c->lhead_lnum + 1; | |
1394 | if (lnum >= UBIFS_LOG_LNUM + c->log_lebs) | |
1395 | lnum = UBIFS_LOG_LNUM; | |
1396 | if (lnum == c->ltail_lnum) { | |
1397 | err = ubifs_consolidate_log(c); | |
1398 | if (err) | |
1399 | goto out; | |
1400 | } | |
1401 | ||
1402 | if (c->need_recovery) | |
1403 | err = ubifs_rcvry_gc_commit(c); | |
1404 | else | |
1405 | err = take_gc_lnum(c); | |
1406 | if (err) | |
1407 | goto out; | |
1408 | ||
1409 | if (c->need_recovery) { | |
1410 | c->need_recovery = 0; | |
1411 | ubifs_msg("deferred recovery completed"); | |
1412 | } | |
1413 | ||
1414 | dbg_gen("re-mounted read-write"); | |
1415 | c->vfs_sb->s_flags &= ~MS_RDONLY; | |
1416 | c->remounting_rw = 0; | |
1417 | mutex_unlock(&c->umount_mutex); | |
1418 | return 0; | |
1419 | ||
1420 | out: | |
1421 | vfree(c->orph_buf); | |
1422 | c->orph_buf = NULL; | |
1423 | if (c->bgt) { | |
1424 | kthread_stop(c->bgt); | |
1425 | c->bgt = NULL; | |
1426 | } | |
1427 | free_wbufs(c); | |
1428 | vfree(c->ileb_buf); | |
1429 | c->ileb_buf = NULL; | |
1430 | ubifs_lpt_free(c, 1); | |
1431 | c->remounting_rw = 0; | |
1432 | mutex_unlock(&c->umount_mutex); | |
1433 | return err; | |
1434 | } | |
1435 | ||
1436 | /** | |
1437 | * commit_on_unmount - commit the journal when un-mounting. | |
1438 | * @c: UBIFS file-system description object | |
1439 | * | |
1440 | * This function is called during un-mounting and it commits the journal unless | |
1441 | * the "fast unmount" mode is enabled. It also avoids committing the journal if | |
1442 | * it contains too few data. | |
1443 | * | |
1444 | * Sometimes recovery requires the journal to be committed at least once, and | |
1445 | * this function takes care about this. | |
1446 | */ | |
1447 | static void commit_on_unmount(struct ubifs_info *c) | |
1448 | { | |
1449 | if (!c->fast_unmount) { | |
1450 | long long bud_bytes; | |
1451 | ||
1452 | spin_lock(&c->buds_lock); | |
1453 | bud_bytes = c->bud_bytes; | |
1454 | spin_unlock(&c->buds_lock); | |
1455 | if (bud_bytes > c->leb_size) | |
1456 | ubifs_run_commit(c); | |
1457 | } | |
1458 | } | |
1459 | ||
1460 | /** | |
1461 | * ubifs_remount_ro - re-mount in read-only mode. | |
1462 | * @c: UBIFS file-system description object | |
1463 | * | |
1464 | * We rely on VFS to have stopped writing. Possibly the background thread could | |
1465 | * be running a commit, however kthread_stop will wait in that case. | |
1466 | */ | |
1467 | static void ubifs_remount_ro(struct ubifs_info *c) | |
1468 | { | |
1469 | int i, err; | |
1470 | ||
1471 | ubifs_assert(!c->need_recovery); | |
1472 | commit_on_unmount(c); | |
1473 | ||
1474 | mutex_lock(&c->umount_mutex); | |
1475 | if (c->bgt) { | |
1476 | kthread_stop(c->bgt); | |
1477 | c->bgt = NULL; | |
1478 | } | |
1479 | ||
1480 | for (i = 0; i < c->jhead_cnt; i++) { | |
1481 | ubifs_wbuf_sync(&c->jheads[i].wbuf); | |
1482 | del_timer_sync(&c->jheads[i].wbuf.timer); | |
1483 | } | |
1484 | ||
1485 | if (!c->ro_media) { | |
1486 | c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY); | |
1487 | c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS); | |
1488 | c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum); | |
1489 | err = ubifs_write_master(c); | |
1490 | if (err) | |
1491 | ubifs_ro_mode(c, err); | |
1492 | } | |
1493 | ||
1494 | ubifs_destroy_idx_gc(c); | |
1495 | free_wbufs(c); | |
1496 | vfree(c->orph_buf); | |
1497 | c->orph_buf = NULL; | |
1498 | vfree(c->ileb_buf); | |
1499 | c->ileb_buf = NULL; | |
1500 | ubifs_lpt_free(c, 1); | |
1501 | mutex_unlock(&c->umount_mutex); | |
1502 | } | |
1503 | ||
1504 | static void ubifs_put_super(struct super_block *sb) | |
1505 | { | |
1506 | int i; | |
1507 | struct ubifs_info *c = sb->s_fs_info; | |
1508 | ||
1509 | ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num, | |
1510 | c->vi.vol_id); | |
1511 | /* | |
1512 | * The following asserts are only valid if there has not been a failure | |
1513 | * of the media. For example, there will be dirty inodes if we failed | |
1514 | * to write them back because of I/O errors. | |
1515 | */ | |
1516 | ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0); | |
1517 | ubifs_assert(c->budg_idx_growth == 0); | |
7d32c2bb | 1518 | ubifs_assert(c->budg_dd_growth == 0); |
1e51764a AB |
1519 | ubifs_assert(c->budg_data_growth == 0); |
1520 | ||
1521 | /* | |
1522 | * The 'c->umount_lock' prevents races between UBIFS memory shrinker | |
1523 | * and file system un-mount. Namely, it prevents the shrinker from | |
1524 | * picking this superblock for shrinking - it will be just skipped if | |
1525 | * the mutex is locked. | |
1526 | */ | |
1527 | mutex_lock(&c->umount_mutex); | |
1528 | if (!(c->vfs_sb->s_flags & MS_RDONLY)) { | |
1529 | /* | |
1530 | * First of all kill the background thread to make sure it does | |
1531 | * not interfere with un-mounting and freeing resources. | |
1532 | */ | |
1533 | if (c->bgt) { | |
1534 | kthread_stop(c->bgt); | |
1535 | c->bgt = NULL; | |
1536 | } | |
1537 | ||
1538 | /* Synchronize write-buffers */ | |
1539 | if (c->jheads) | |
1540 | for (i = 0; i < c->jhead_cnt; i++) { | |
1541 | ubifs_wbuf_sync(&c->jheads[i].wbuf); | |
1542 | del_timer_sync(&c->jheads[i].wbuf.timer); | |
1543 | } | |
1544 | ||
1545 | /* | |
1546 | * On fatal errors c->ro_media is set to 1, in which case we do | |
1547 | * not write the master node. | |
1548 | */ | |
1549 | if (!c->ro_media) { | |
1550 | /* | |
1551 | * We are being cleanly unmounted which means the | |
1552 | * orphans were killed - indicate this in the master | |
1553 | * node. Also save the reserved GC LEB number. | |
1554 | */ | |
1555 | int err; | |
1556 | ||
1557 | c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY); | |
1558 | c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS); | |
1559 | c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum); | |
1560 | err = ubifs_write_master(c); | |
1561 | if (err) | |
1562 | /* | |
1563 | * Recovery will attempt to fix the master area | |
1564 | * next mount, so we just print a message and | |
1565 | * continue to unmount normally. | |
1566 | */ | |
1567 | ubifs_err("failed to write master node, " | |
1568 | "error %d", err); | |
1569 | } | |
1570 | } | |
1571 | ||
1572 | ubifs_umount(c); | |
1573 | bdi_destroy(&c->bdi); | |
1574 | ubi_close_volume(c->ubi); | |
1575 | mutex_unlock(&c->umount_mutex); | |
1576 | kfree(c); | |
1577 | } | |
1578 | ||
1579 | static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data) | |
1580 | { | |
1581 | int err; | |
1582 | struct ubifs_info *c = sb->s_fs_info; | |
1583 | ||
1584 | dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags); | |
1585 | ||
1586 | err = ubifs_parse_options(c, data, 1); | |
1587 | if (err) { | |
1588 | ubifs_err("invalid or unknown remount parameter"); | |
1589 | return err; | |
1590 | } | |
1591 | if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) { | |
1592 | err = ubifs_remount_rw(c); | |
1593 | if (err) | |
1594 | return err; | |
1595 | } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY)) | |
1596 | ubifs_remount_ro(c); | |
1597 | ||
1598 | return 0; | |
1599 | } | |
1600 | ||
1601 | struct super_operations ubifs_super_operations = { | |
1602 | .alloc_inode = ubifs_alloc_inode, | |
1603 | .destroy_inode = ubifs_destroy_inode, | |
1604 | .put_super = ubifs_put_super, | |
1605 | .write_inode = ubifs_write_inode, | |
1606 | .delete_inode = ubifs_delete_inode, | |
1607 | .statfs = ubifs_statfs, | |
1608 | .dirty_inode = ubifs_dirty_inode, | |
1609 | .remount_fs = ubifs_remount_fs, | |
1610 | .show_options = ubifs_show_options, | |
1611 | .sync_fs = ubifs_sync_fs, | |
1612 | }; | |
1613 | ||
1614 | /** | |
1615 | * open_ubi - parse UBI device name string and open the UBI device. | |
1616 | * @name: UBI volume name | |
1617 | * @mode: UBI volume open mode | |
1618 | * | |
1619 | * There are several ways to specify UBI volumes when mounting UBIFS: | |
1620 | * o ubiX_Y - UBI device number X, volume Y; | |
1621 | * o ubiY - UBI device number 0, volume Y; | |
1622 | * o ubiX:NAME - mount UBI device X, volume with name NAME; | |
1623 | * o ubi:NAME - mount UBI device 0, volume with name NAME. | |
1624 | * | |
1625 | * Alternative '!' separator may be used instead of ':' (because some shells | |
1626 | * like busybox may interpret ':' as an NFS host name separator). This function | |
1627 | * returns ubi volume object in case of success and a negative error code in | |
1628 | * case of failure. | |
1629 | */ | |
1630 | static struct ubi_volume_desc *open_ubi(const char *name, int mode) | |
1631 | { | |
1632 | int dev, vol; | |
1633 | char *endptr; | |
1634 | ||
1635 | if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i') | |
1636 | return ERR_PTR(-EINVAL); | |
1637 | ||
1638 | /* ubi:NAME method */ | |
1639 | if ((name[3] == ':' || name[3] == '!') && name[4] != '\0') | |
1640 | return ubi_open_volume_nm(0, name + 4, mode); | |
1641 | ||
1642 | if (!isdigit(name[3])) | |
1643 | return ERR_PTR(-EINVAL); | |
1644 | ||
1645 | dev = simple_strtoul(name + 3, &endptr, 0); | |
1646 | ||
1647 | /* ubiY method */ | |
1648 | if (*endptr == '\0') | |
1649 | return ubi_open_volume(0, dev, mode); | |
1650 | ||
1651 | /* ubiX_Y method */ | |
1652 | if (*endptr == '_' && isdigit(endptr[1])) { | |
1653 | vol = simple_strtoul(endptr + 1, &endptr, 0); | |
1654 | if (*endptr != '\0') | |
1655 | return ERR_PTR(-EINVAL); | |
1656 | return ubi_open_volume(dev, vol, mode); | |
1657 | } | |
1658 | ||
1659 | /* ubiX:NAME method */ | |
1660 | if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0') | |
1661 | return ubi_open_volume_nm(dev, ++endptr, mode); | |
1662 | ||
1663 | return ERR_PTR(-EINVAL); | |
1664 | } | |
1665 | ||
1666 | static int ubifs_fill_super(struct super_block *sb, void *data, int silent) | |
1667 | { | |
1668 | struct ubi_volume_desc *ubi = sb->s_fs_info; | |
1669 | struct ubifs_info *c; | |
1670 | struct inode *root; | |
1671 | int err; | |
1672 | ||
1673 | c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL); | |
1674 | if (!c) | |
1675 | return -ENOMEM; | |
1676 | ||
1677 | spin_lock_init(&c->cnt_lock); | |
1678 | spin_lock_init(&c->cs_lock); | |
1679 | spin_lock_init(&c->buds_lock); | |
1680 | spin_lock_init(&c->space_lock); | |
1681 | spin_lock_init(&c->orphan_lock); | |
1682 | init_rwsem(&c->commit_sem); | |
1683 | mutex_init(&c->lp_mutex); | |
1684 | mutex_init(&c->tnc_mutex); | |
1685 | mutex_init(&c->log_mutex); | |
1686 | mutex_init(&c->mst_mutex); | |
1687 | mutex_init(&c->umount_mutex); | |
1688 | init_waitqueue_head(&c->cmt_wq); | |
1689 | c->buds = RB_ROOT; | |
1690 | c->old_idx = RB_ROOT; | |
1691 | c->size_tree = RB_ROOT; | |
1692 | c->orph_tree = RB_ROOT; | |
1693 | INIT_LIST_HEAD(&c->infos_list); | |
1694 | INIT_LIST_HEAD(&c->idx_gc); | |
1695 | INIT_LIST_HEAD(&c->replay_list); | |
1696 | INIT_LIST_HEAD(&c->replay_buds); | |
1697 | INIT_LIST_HEAD(&c->uncat_list); | |
1698 | INIT_LIST_HEAD(&c->empty_list); | |
1699 | INIT_LIST_HEAD(&c->freeable_list); | |
1700 | INIT_LIST_HEAD(&c->frdi_idx_list); | |
1701 | INIT_LIST_HEAD(&c->unclean_leb_list); | |
1702 | INIT_LIST_HEAD(&c->old_buds); | |
1703 | INIT_LIST_HEAD(&c->orph_list); | |
1704 | INIT_LIST_HEAD(&c->orph_new); | |
1705 | ||
1706 | c->highest_inum = UBIFS_FIRST_INO; | |
1e51764a AB |
1707 | c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM; |
1708 | ||
1709 | ubi_get_volume_info(ubi, &c->vi); | |
1710 | ubi_get_device_info(c->vi.ubi_num, &c->di); | |
1711 | ||
1712 | /* Re-open the UBI device in read-write mode */ | |
1713 | c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE); | |
1714 | if (IS_ERR(c->ubi)) { | |
1715 | err = PTR_ERR(c->ubi); | |
1716 | goto out_free; | |
1717 | } | |
1718 | ||
1719 | /* | |
0a883a05 | 1720 | * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For |
1e51764a AB |
1721 | * UBIFS, I/O is not deferred, it is done immediately in readpage, |
1722 | * which means the user would have to wait not just for their own I/O | |
0a883a05 | 1723 | * but the read-ahead I/O as well i.e. completely pointless. |
1e51764a AB |
1724 | * |
1725 | * Read-ahead will be disabled because @c->bdi.ra_pages is 0. | |
1726 | */ | |
1727 | c->bdi.capabilities = BDI_CAP_MAP_COPY; | |
1728 | c->bdi.unplug_io_fn = default_unplug_io_fn; | |
1729 | err = bdi_init(&c->bdi); | |
1730 | if (err) | |
1731 | goto out_close; | |
1732 | ||
1733 | err = ubifs_parse_options(c, data, 0); | |
1734 | if (err) | |
1735 | goto out_bdi; | |
1736 | ||
1737 | c->vfs_sb = sb; | |
1738 | ||
1739 | sb->s_fs_info = c; | |
1740 | sb->s_magic = UBIFS_SUPER_MAGIC; | |
1741 | sb->s_blocksize = UBIFS_BLOCK_SIZE; | |
1742 | sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT; | |
1743 | sb->s_dev = c->vi.cdev; | |
1744 | sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c); | |
1745 | if (c->max_inode_sz > MAX_LFS_FILESIZE) | |
1746 | sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE; | |
1747 | sb->s_op = &ubifs_super_operations; | |
1748 | ||
1749 | mutex_lock(&c->umount_mutex); | |
1750 | err = mount_ubifs(c); | |
1751 | if (err) { | |
1752 | ubifs_assert(err < 0); | |
1753 | goto out_unlock; | |
1754 | } | |
1755 | ||
1756 | /* Read the root inode */ | |
1757 | root = ubifs_iget(sb, UBIFS_ROOT_INO); | |
1758 | if (IS_ERR(root)) { | |
1759 | err = PTR_ERR(root); | |
1760 | goto out_umount; | |
1761 | } | |
1762 | ||
1763 | sb->s_root = d_alloc_root(root); | |
1764 | if (!sb->s_root) | |
1765 | goto out_iput; | |
1766 | ||
1767 | mutex_unlock(&c->umount_mutex); | |
1768 | ||
1769 | return 0; | |
1770 | ||
1771 | out_iput: | |
1772 | iput(root); | |
1773 | out_umount: | |
1774 | ubifs_umount(c); | |
1775 | out_unlock: | |
1776 | mutex_unlock(&c->umount_mutex); | |
1777 | out_bdi: | |
1778 | bdi_destroy(&c->bdi); | |
1779 | out_close: | |
1780 | ubi_close_volume(c->ubi); | |
1781 | out_free: | |
1782 | kfree(c); | |
1783 | return err; | |
1784 | } | |
1785 | ||
1786 | static int sb_test(struct super_block *sb, void *data) | |
1787 | { | |
1788 | dev_t *dev = data; | |
1789 | ||
1790 | return sb->s_dev == *dev; | |
1791 | } | |
1792 | ||
1793 | static int sb_set(struct super_block *sb, void *data) | |
1794 | { | |
1795 | dev_t *dev = data; | |
1796 | ||
1797 | sb->s_dev = *dev; | |
1798 | return 0; | |
1799 | } | |
1800 | ||
1801 | static int ubifs_get_sb(struct file_system_type *fs_type, int flags, | |
1802 | const char *name, void *data, struct vfsmount *mnt) | |
1803 | { | |
1804 | struct ubi_volume_desc *ubi; | |
1805 | struct ubi_volume_info vi; | |
1806 | struct super_block *sb; | |
1807 | int err; | |
1808 | ||
1809 | dbg_gen("name %s, flags %#x", name, flags); | |
1810 | ||
1811 | /* | |
1812 | * Get UBI device number and volume ID. Mount it read-only so far | |
1813 | * because this might be a new mount point, and UBI allows only one | |
1814 | * read-write user at a time. | |
1815 | */ | |
1816 | ubi = open_ubi(name, UBI_READONLY); | |
1817 | if (IS_ERR(ubi)) { | |
1818 | ubifs_err("cannot open \"%s\", error %d", | |
1819 | name, (int)PTR_ERR(ubi)); | |
1820 | return PTR_ERR(ubi); | |
1821 | } | |
1822 | ubi_get_volume_info(ubi, &vi); | |
1823 | ||
1824 | dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id); | |
1825 | ||
1826 | sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev); | |
1827 | if (IS_ERR(sb)) { | |
1828 | err = PTR_ERR(sb); | |
1829 | goto out_close; | |
1830 | } | |
1831 | ||
1832 | if (sb->s_root) { | |
1833 | /* A new mount point for already mounted UBIFS */ | |
1834 | dbg_gen("this ubi volume is already mounted"); | |
1835 | if ((flags ^ sb->s_flags) & MS_RDONLY) { | |
1836 | err = -EBUSY; | |
1837 | goto out_deact; | |
1838 | } | |
1839 | } else { | |
1840 | sb->s_flags = flags; | |
1841 | /* | |
1842 | * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is | |
1843 | * replaced by 'c'. | |
1844 | */ | |
1845 | sb->s_fs_info = ubi; | |
1846 | err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0); | |
1847 | if (err) | |
1848 | goto out_deact; | |
1849 | /* We do not support atime */ | |
1850 | sb->s_flags |= MS_ACTIVE | MS_NOATIME; | |
1851 | } | |
1852 | ||
1853 | /* 'fill_super()' opens ubi again so we must close it here */ | |
1854 | ubi_close_volume(ubi); | |
1855 | ||
1856 | return simple_set_mnt(mnt, sb); | |
1857 | ||
1858 | out_deact: | |
1859 | up_write(&sb->s_umount); | |
1860 | deactivate_super(sb); | |
1861 | out_close: | |
1862 | ubi_close_volume(ubi); | |
1863 | return err; | |
1864 | } | |
1865 | ||
1866 | static void ubifs_kill_sb(struct super_block *sb) | |
1867 | { | |
1868 | struct ubifs_info *c = sb->s_fs_info; | |
1869 | ||
1870 | /* | |
1871 | * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()' | |
1872 | * in order to be outside BKL. | |
1873 | */ | |
1874 | if (sb->s_root && !(sb->s_flags & MS_RDONLY)) | |
1875 | commit_on_unmount(c); | |
1876 | /* The un-mount routine is actually done in put_super() */ | |
1877 | generic_shutdown_super(sb); | |
1878 | } | |
1879 | ||
1880 | static struct file_system_type ubifs_fs_type = { | |
1881 | .name = "ubifs", | |
1882 | .owner = THIS_MODULE, | |
1883 | .get_sb = ubifs_get_sb, | |
1884 | .kill_sb = ubifs_kill_sb | |
1885 | }; | |
1886 | ||
1887 | /* | |
1888 | * Inode slab cache constructor. | |
1889 | */ | |
51cc5068 | 1890 | static void inode_slab_ctor(void *obj) |
1e51764a AB |
1891 | { |
1892 | struct ubifs_inode *ui = obj; | |
1893 | inode_init_once(&ui->vfs_inode); | |
1894 | } | |
1895 | ||
1896 | static int __init ubifs_init(void) | |
1897 | { | |
1898 | int err; | |
1899 | ||
1900 | BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24); | |
1901 | ||
1902 | /* Make sure node sizes are 8-byte aligned */ | |
1903 | BUILD_BUG_ON(UBIFS_CH_SZ & 7); | |
1904 | BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7); | |
1905 | BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7); | |
1906 | BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7); | |
1907 | BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7); | |
1908 | BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7); | |
1909 | BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7); | |
1910 | BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7); | |
1911 | BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7); | |
1912 | BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7); | |
1913 | BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7); | |
1914 | ||
1915 | BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7); | |
1916 | BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7); | |
1917 | BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7); | |
1918 | BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7); | |
1919 | BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7); | |
1920 | BUILD_BUG_ON(MIN_WRITE_SZ & 7); | |
1921 | ||
1922 | /* Check min. node size */ | |
1923 | BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ); | |
1924 | BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ); | |
1925 | BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ); | |
1926 | BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ); | |
1927 | ||
1928 | BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ); | |
1929 | BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ); | |
1930 | BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ); | |
1931 | BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ); | |
1932 | ||
1933 | /* Defined node sizes */ | |
1934 | BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096); | |
1935 | BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512); | |
1936 | BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160); | |
1937 | BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64); | |
1938 | ||
1939 | /* | |
1940 | * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to | |
1941 | * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2. | |
1942 | */ | |
1943 | if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) { | |
1944 | ubifs_err("VFS page cache size is %u bytes, but UBIFS requires" | |
1945 | " at least 4096 bytes", | |
1946 | (unsigned int)PAGE_CACHE_SIZE); | |
1947 | return -EINVAL; | |
1948 | } | |
1949 | ||
1950 | err = register_filesystem(&ubifs_fs_type); | |
1951 | if (err) { | |
1952 | ubifs_err("cannot register file system, error %d", err); | |
1953 | return err; | |
1954 | } | |
1955 | ||
1956 | err = -ENOMEM; | |
1957 | ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab", | |
1958 | sizeof(struct ubifs_inode), 0, | |
1959 | SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT, | |
1960 | &inode_slab_ctor); | |
1961 | if (!ubifs_inode_slab) | |
1962 | goto out_reg; | |
1963 | ||
1964 | register_shrinker(&ubifs_shrinker_info); | |
1965 | ||
1966 | err = ubifs_compressors_init(); | |
1967 | if (err) | |
1968 | goto out_compr; | |
1969 | ||
1970 | return 0; | |
1971 | ||
1972 | out_compr: | |
1973 | unregister_shrinker(&ubifs_shrinker_info); | |
1974 | kmem_cache_destroy(ubifs_inode_slab); | |
1975 | out_reg: | |
1976 | unregister_filesystem(&ubifs_fs_type); | |
1977 | return err; | |
1978 | } | |
1979 | /* late_initcall to let compressors initialize first */ | |
1980 | late_initcall(ubifs_init); | |
1981 | ||
1982 | static void __exit ubifs_exit(void) | |
1983 | { | |
1984 | ubifs_assert(list_empty(&ubifs_infos)); | |
1985 | ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0); | |
1986 | ||
1987 | ubifs_compressors_exit(); | |
1988 | unregister_shrinker(&ubifs_shrinker_info); | |
1989 | kmem_cache_destroy(ubifs_inode_slab); | |
1990 | unregister_filesystem(&ubifs_fs_type); | |
1991 | } | |
1992 | module_exit(ubifs_exit); | |
1993 | ||
1994 | MODULE_LICENSE("GPL"); | |
1995 | MODULE_VERSION(__stringify(UBIFS_VERSION)); | |
1996 | MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter"); | |
1997 | MODULE_DESCRIPTION("UBIFS - UBI File System"); |