Linux 4.2.1
[linux/fpc-iii.git] / fs / ubifs / super.c
blob9547a27868ad49e11151c324c8f60a7db2bc92f6
1 /*
2 * This file is part of UBIFS.
4 * Copyright (C) 2006-2008 Nokia Corporation.
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.
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.
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
19 * Authors: Artem Bityutskiy (Битюцкий Артём)
20 * Adrian Hunter
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.
29 #include <linux/init.h>
30 #include <linux/slab.h>
31 #include <linux/module.h>
32 #include <linux/ctype.h>
33 #include <linux/kthread.h>
34 #include <linux/parser.h>
35 #include <linux/seq_file.h>
36 #include <linux/mount.h>
37 #include <linux/math64.h>
38 #include <linux/writeback.h>
39 #include "ubifs.h"
42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
43 * allocating too much.
45 #define UBIFS_KMALLOC_OK (128*1024)
47 /* Slab cache for UBIFS inodes */
48 struct kmem_cache *ubifs_inode_slab;
50 /* UBIFS TNC shrinker description */
51 static struct shrinker ubifs_shrinker_info = {
52 .scan_objects = ubifs_shrink_scan,
53 .count_objects = ubifs_shrink_count,
54 .seeks = DEFAULT_SEEKS,
57 /**
58 * validate_inode - validate inode.
59 * @c: UBIFS file-system description object
60 * @inode: the inode to validate
62 * This is a helper function for 'ubifs_iget()' which validates various fields
63 * of a newly built inode to make sure they contain sane values and prevent
64 * possible vulnerabilities. Returns zero if the inode is all right and
65 * a non-zero error code if not.
67 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
69 int err;
70 const struct ubifs_inode *ui = ubifs_inode(inode);
72 if (inode->i_size > c->max_inode_sz) {
73 ubifs_err(c, "inode is too large (%lld)",
74 (long long)inode->i_size);
75 return 1;
78 if (ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
79 ubifs_err(c, "unknown compression type %d", ui->compr_type);
80 return 2;
83 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
84 return 3;
86 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
87 return 4;
89 if (ui->xattr && !S_ISREG(inode->i_mode))
90 return 5;
92 if (!ubifs_compr_present(ui->compr_type)) {
93 ubifs_warn(c, "inode %lu uses '%s' compression, but it was not compiled in",
94 inode->i_ino, ubifs_compr_name(ui->compr_type));
97 err = dbg_check_dir(c, inode);
98 return err;
101 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
103 int err;
104 union ubifs_key key;
105 struct ubifs_ino_node *ino;
106 struct ubifs_info *c = sb->s_fs_info;
107 struct inode *inode;
108 struct ubifs_inode *ui;
110 dbg_gen("inode %lu", inum);
112 inode = iget_locked(sb, inum);
113 if (!inode)
114 return ERR_PTR(-ENOMEM);
115 if (!(inode->i_state & I_NEW))
116 return inode;
117 ui = ubifs_inode(inode);
119 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
120 if (!ino) {
121 err = -ENOMEM;
122 goto out;
125 ino_key_init(c, &key, inode->i_ino);
127 err = ubifs_tnc_lookup(c, &key, ino);
128 if (err)
129 goto out_ino;
131 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
132 set_nlink(inode, le32_to_cpu(ino->nlink));
133 i_uid_write(inode, le32_to_cpu(ino->uid));
134 i_gid_write(inode, le32_to_cpu(ino->gid));
135 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
136 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
137 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
138 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
139 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
140 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
141 inode->i_mode = le32_to_cpu(ino->mode);
142 inode->i_size = le64_to_cpu(ino->size);
144 ui->data_len = le32_to_cpu(ino->data_len);
145 ui->flags = le32_to_cpu(ino->flags);
146 ui->compr_type = le16_to_cpu(ino->compr_type);
147 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
148 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
149 ui->xattr_size = le32_to_cpu(ino->xattr_size);
150 ui->xattr_names = le32_to_cpu(ino->xattr_names);
151 ui->synced_i_size = ui->ui_size = inode->i_size;
153 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
155 err = validate_inode(c, inode);
156 if (err)
157 goto out_invalid;
159 switch (inode->i_mode & S_IFMT) {
160 case S_IFREG:
161 inode->i_mapping->a_ops = &ubifs_file_address_operations;
162 inode->i_op = &ubifs_file_inode_operations;
163 inode->i_fop = &ubifs_file_operations;
164 if (ui->xattr) {
165 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
166 if (!ui->data) {
167 err = -ENOMEM;
168 goto out_ino;
170 memcpy(ui->data, ino->data, ui->data_len);
171 ((char *)ui->data)[ui->data_len] = '\0';
172 } else if (ui->data_len != 0) {
173 err = 10;
174 goto out_invalid;
176 break;
177 case S_IFDIR:
178 inode->i_op = &ubifs_dir_inode_operations;
179 inode->i_fop = &ubifs_dir_operations;
180 if (ui->data_len != 0) {
181 err = 11;
182 goto out_invalid;
184 break;
185 case S_IFLNK:
186 inode->i_op = &ubifs_symlink_inode_operations;
187 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
188 err = 12;
189 goto out_invalid;
191 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
192 if (!ui->data) {
193 err = -ENOMEM;
194 goto out_ino;
196 memcpy(ui->data, ino->data, ui->data_len);
197 ((char *)ui->data)[ui->data_len] = '\0';
198 inode->i_link = ui->data;
199 break;
200 case S_IFBLK:
201 case S_IFCHR:
203 dev_t rdev;
204 union ubifs_dev_desc *dev;
206 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
207 if (!ui->data) {
208 err = -ENOMEM;
209 goto out_ino;
212 dev = (union ubifs_dev_desc *)ino->data;
213 if (ui->data_len == sizeof(dev->new))
214 rdev = new_decode_dev(le32_to_cpu(dev->new));
215 else if (ui->data_len == sizeof(dev->huge))
216 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
217 else {
218 err = 13;
219 goto out_invalid;
221 memcpy(ui->data, ino->data, ui->data_len);
222 inode->i_op = &ubifs_file_inode_operations;
223 init_special_inode(inode, inode->i_mode, rdev);
224 break;
226 case S_IFSOCK:
227 case S_IFIFO:
228 inode->i_op = &ubifs_file_inode_operations;
229 init_special_inode(inode, inode->i_mode, 0);
230 if (ui->data_len != 0) {
231 err = 14;
232 goto out_invalid;
234 break;
235 default:
236 err = 15;
237 goto out_invalid;
240 kfree(ino);
241 ubifs_set_inode_flags(inode);
242 unlock_new_inode(inode);
243 return inode;
245 out_invalid:
246 ubifs_err(c, "inode %lu validation failed, error %d", inode->i_ino, err);
247 ubifs_dump_node(c, ino);
248 ubifs_dump_inode(c, inode);
249 err = -EINVAL;
250 out_ino:
251 kfree(ino);
252 out:
253 ubifs_err(c, "failed to read inode %lu, error %d", inode->i_ino, err);
254 iget_failed(inode);
255 return ERR_PTR(err);
258 static struct inode *ubifs_alloc_inode(struct super_block *sb)
260 struct ubifs_inode *ui;
262 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
263 if (!ui)
264 return NULL;
266 memset((void *)ui + sizeof(struct inode), 0,
267 sizeof(struct ubifs_inode) - sizeof(struct inode));
268 mutex_init(&ui->ui_mutex);
269 spin_lock_init(&ui->ui_lock);
270 return &ui->vfs_inode;
273 static void ubifs_i_callback(struct rcu_head *head)
275 struct inode *inode = container_of(head, struct inode, i_rcu);
276 struct ubifs_inode *ui = ubifs_inode(inode);
277 kmem_cache_free(ubifs_inode_slab, ui);
280 static void ubifs_destroy_inode(struct inode *inode)
282 struct ubifs_inode *ui = ubifs_inode(inode);
284 kfree(ui->data);
285 call_rcu(&inode->i_rcu, ubifs_i_callback);
289 * Note, Linux write-back code calls this without 'i_mutex'.
291 static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
293 int err = 0;
294 struct ubifs_info *c = inode->i_sb->s_fs_info;
295 struct ubifs_inode *ui = ubifs_inode(inode);
297 ubifs_assert(!ui->xattr);
298 if (is_bad_inode(inode))
299 return 0;
301 mutex_lock(&ui->ui_mutex);
303 * Due to races between write-back forced by budgeting
304 * (see 'sync_some_inodes()') and background write-back, the inode may
305 * have already been synchronized, do not do this again. This might
306 * also happen if it was synchronized in an VFS operation, e.g.
307 * 'ubifs_link()'.
309 if (!ui->dirty) {
310 mutex_unlock(&ui->ui_mutex);
311 return 0;
315 * As an optimization, do not write orphan inodes to the media just
316 * because this is not needed.
318 dbg_gen("inode %lu, mode %#x, nlink %u",
319 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
320 if (inode->i_nlink) {
321 err = ubifs_jnl_write_inode(c, inode);
322 if (err)
323 ubifs_err(c, "can't write inode %lu, error %d",
324 inode->i_ino, err);
325 else
326 err = dbg_check_inode_size(c, inode, ui->ui_size);
329 ui->dirty = 0;
330 mutex_unlock(&ui->ui_mutex);
331 ubifs_release_dirty_inode_budget(c, ui);
332 return err;
335 static void ubifs_evict_inode(struct inode *inode)
337 int err;
338 struct ubifs_info *c = inode->i_sb->s_fs_info;
339 struct ubifs_inode *ui = ubifs_inode(inode);
341 if (ui->xattr)
343 * Extended attribute inode deletions are fully handled in
344 * 'ubifs_removexattr()'. These inodes are special and have
345 * limited usage, so there is nothing to do here.
347 goto out;
349 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
350 ubifs_assert(!atomic_read(&inode->i_count));
352 truncate_inode_pages_final(&inode->i_data);
354 if (inode->i_nlink)
355 goto done;
357 if (is_bad_inode(inode))
358 goto out;
360 ui->ui_size = inode->i_size = 0;
361 err = ubifs_jnl_delete_inode(c, inode);
362 if (err)
364 * Worst case we have a lost orphan inode wasting space, so a
365 * simple error message is OK here.
367 ubifs_err(c, "can't delete inode %lu, error %d",
368 inode->i_ino, err);
370 out:
371 if (ui->dirty)
372 ubifs_release_dirty_inode_budget(c, ui);
373 else {
374 /* We've deleted something - clean the "no space" flags */
375 c->bi.nospace = c->bi.nospace_rp = 0;
376 smp_wmb();
378 done:
379 clear_inode(inode);
382 static void ubifs_dirty_inode(struct inode *inode, int flags)
384 struct ubifs_inode *ui = ubifs_inode(inode);
386 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
387 if (!ui->dirty) {
388 ui->dirty = 1;
389 dbg_gen("inode %lu", inode->i_ino);
393 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
395 struct ubifs_info *c = dentry->d_sb->s_fs_info;
396 unsigned long long free;
397 __le32 *uuid = (__le32 *)c->uuid;
399 free = ubifs_get_free_space(c);
400 dbg_gen("free space %lld bytes (%lld blocks)",
401 free, free >> UBIFS_BLOCK_SHIFT);
403 buf->f_type = UBIFS_SUPER_MAGIC;
404 buf->f_bsize = UBIFS_BLOCK_SIZE;
405 buf->f_blocks = c->block_cnt;
406 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
407 if (free > c->report_rp_size)
408 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
409 else
410 buf->f_bavail = 0;
411 buf->f_files = 0;
412 buf->f_ffree = 0;
413 buf->f_namelen = UBIFS_MAX_NLEN;
414 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
415 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
416 ubifs_assert(buf->f_bfree <= c->block_cnt);
417 return 0;
420 static int ubifs_show_options(struct seq_file *s, struct dentry *root)
422 struct ubifs_info *c = root->d_sb->s_fs_info;
424 if (c->mount_opts.unmount_mode == 2)
425 seq_puts(s, ",fast_unmount");
426 else if (c->mount_opts.unmount_mode == 1)
427 seq_puts(s, ",norm_unmount");
429 if (c->mount_opts.bulk_read == 2)
430 seq_puts(s, ",bulk_read");
431 else if (c->mount_opts.bulk_read == 1)
432 seq_puts(s, ",no_bulk_read");
434 if (c->mount_opts.chk_data_crc == 2)
435 seq_puts(s, ",chk_data_crc");
436 else if (c->mount_opts.chk_data_crc == 1)
437 seq_puts(s, ",no_chk_data_crc");
439 if (c->mount_opts.override_compr) {
440 seq_printf(s, ",compr=%s",
441 ubifs_compr_name(c->mount_opts.compr_type));
444 return 0;
447 static int ubifs_sync_fs(struct super_block *sb, int wait)
449 int i, err;
450 struct ubifs_info *c = sb->s_fs_info;
453 * Zero @wait is just an advisory thing to help the file system shove
454 * lots of data into the queues, and there will be the second
455 * '->sync_fs()' call, with non-zero @wait.
457 if (!wait)
458 return 0;
461 * Synchronize write buffers, because 'ubifs_run_commit()' does not
462 * do this if it waits for an already running commit.
464 for (i = 0; i < c->jhead_cnt; i++) {
465 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
466 if (err)
467 return err;
471 * Strictly speaking, it is not necessary to commit the journal here,
472 * synchronizing write-buffers would be enough. But committing makes
473 * UBIFS free space predictions much more accurate, so we want to let
474 * the user be able to get more accurate results of 'statfs()' after
475 * they synchronize the file system.
477 err = ubifs_run_commit(c);
478 if (err)
479 return err;
481 return ubi_sync(c->vi.ubi_num);
485 * init_constants_early - initialize UBIFS constants.
486 * @c: UBIFS file-system description object
488 * This function initialize UBIFS constants which do not need the superblock to
489 * be read. It also checks that the UBI volume satisfies basic UBIFS
490 * requirements. Returns zero in case of success and a negative error code in
491 * case of failure.
493 static int init_constants_early(struct ubifs_info *c)
495 if (c->vi.corrupted) {
496 ubifs_warn(c, "UBI volume is corrupted - read-only mode");
497 c->ro_media = 1;
500 if (c->di.ro_mode) {
501 ubifs_msg(c, "read-only UBI device");
502 c->ro_media = 1;
505 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
506 ubifs_msg(c, "static UBI volume - read-only mode");
507 c->ro_media = 1;
510 c->leb_cnt = c->vi.size;
511 c->leb_size = c->vi.usable_leb_size;
512 c->leb_start = c->di.leb_start;
513 c->half_leb_size = c->leb_size / 2;
514 c->min_io_size = c->di.min_io_size;
515 c->min_io_shift = fls(c->min_io_size) - 1;
516 c->max_write_size = c->di.max_write_size;
517 c->max_write_shift = fls(c->max_write_size) - 1;
519 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
520 ubifs_err(c, "too small LEBs (%d bytes), min. is %d bytes",
521 c->leb_size, UBIFS_MIN_LEB_SZ);
522 return -EINVAL;
525 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
526 ubifs_err(c, "too few LEBs (%d), min. is %d",
527 c->leb_cnt, UBIFS_MIN_LEB_CNT);
528 return -EINVAL;
531 if (!is_power_of_2(c->min_io_size)) {
532 ubifs_err(c, "bad min. I/O size %d", c->min_io_size);
533 return -EINVAL;
537 * Maximum write size has to be greater or equivalent to min. I/O
538 * size, and be multiple of min. I/O size.
540 if (c->max_write_size < c->min_io_size ||
541 c->max_write_size % c->min_io_size ||
542 !is_power_of_2(c->max_write_size)) {
543 ubifs_err(c, "bad write buffer size %d for %d min. I/O unit",
544 c->max_write_size, c->min_io_size);
545 return -EINVAL;
549 * UBIFS aligns all node to 8-byte boundary, so to make function in
550 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
551 * less than 8.
553 if (c->min_io_size < 8) {
554 c->min_io_size = 8;
555 c->min_io_shift = 3;
556 if (c->max_write_size < c->min_io_size) {
557 c->max_write_size = c->min_io_size;
558 c->max_write_shift = c->min_io_shift;
562 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
563 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
566 * Initialize node length ranges which are mostly needed for node
567 * length validation.
569 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
570 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
571 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
572 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
573 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
574 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
576 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
577 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
578 c->ranges[UBIFS_ORPH_NODE].min_len =
579 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
580 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
581 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
582 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
583 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
584 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
585 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
586 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
588 * Minimum indexing node size is amended later when superblock is
589 * read and the key length is known.
591 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
593 * Maximum indexing node size is amended later when superblock is
594 * read and the fanout is known.
596 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
599 * Initialize dead and dark LEB space watermarks. See gc.c for comments
600 * about these values.
602 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
603 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
606 * Calculate how many bytes would be wasted at the end of LEB if it was
607 * fully filled with data nodes of maximum size. This is used in
608 * calculations when reporting free space.
610 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
612 /* Buffer size for bulk-reads */
613 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
614 if (c->max_bu_buf_len > c->leb_size)
615 c->max_bu_buf_len = c->leb_size;
616 return 0;
620 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
621 * @c: UBIFS file-system description object
622 * @lnum: LEB the write-buffer was synchronized to
623 * @free: how many free bytes left in this LEB
624 * @pad: how many bytes were padded
626 * This is a callback function which is called by the I/O unit when the
627 * write-buffer is synchronized. We need this to correctly maintain space
628 * accounting in bud logical eraseblocks. This function returns zero in case of
629 * success and a negative error code in case of failure.
631 * This function actually belongs to the journal, but we keep it here because
632 * we want to keep it static.
634 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
636 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
640 * init_constants_sb - initialize UBIFS constants.
641 * @c: UBIFS file-system description object
643 * This is a helper function which initializes various UBIFS constants after
644 * the superblock has been read. It also checks various UBIFS parameters and
645 * makes sure they are all right. Returns zero in case of success and a
646 * negative error code in case of failure.
648 static int init_constants_sb(struct ubifs_info *c)
650 int tmp, err;
651 long long tmp64;
653 c->main_bytes = (long long)c->main_lebs * c->leb_size;
654 c->max_znode_sz = sizeof(struct ubifs_znode) +
655 c->fanout * sizeof(struct ubifs_zbranch);
657 tmp = ubifs_idx_node_sz(c, 1);
658 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
659 c->min_idx_node_sz = ALIGN(tmp, 8);
661 tmp = ubifs_idx_node_sz(c, c->fanout);
662 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
663 c->max_idx_node_sz = ALIGN(tmp, 8);
665 /* Make sure LEB size is large enough to fit full commit */
666 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
667 tmp = ALIGN(tmp, c->min_io_size);
668 if (tmp > c->leb_size) {
669 ubifs_err(c, "too small LEB size %d, at least %d needed",
670 c->leb_size, tmp);
671 return -EINVAL;
675 * Make sure that the log is large enough to fit reference nodes for
676 * all buds plus one reserved LEB.
678 tmp64 = c->max_bud_bytes + c->leb_size - 1;
679 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
680 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
681 tmp /= c->leb_size;
682 tmp += 1;
683 if (c->log_lebs < tmp) {
684 ubifs_err(c, "too small log %d LEBs, required min. %d LEBs",
685 c->log_lebs, tmp);
686 return -EINVAL;
690 * When budgeting we assume worst-case scenarios when the pages are not
691 * be compressed and direntries are of the maximum size.
693 * Note, data, which may be stored in inodes is budgeted separately, so
694 * it is not included into 'c->bi.inode_budget'.
696 c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
697 c->bi.inode_budget = UBIFS_INO_NODE_SZ;
698 c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
701 * When the amount of flash space used by buds becomes
702 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
703 * The writers are unblocked when the commit is finished. To avoid
704 * writers to be blocked UBIFS initiates background commit in advance,
705 * when number of bud bytes becomes above the limit defined below.
707 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
710 * Ensure minimum journal size. All the bytes in the journal heads are
711 * considered to be used, when calculating the current journal usage.
712 * Consequently, if the journal is too small, UBIFS will treat it as
713 * always full.
715 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
716 if (c->bg_bud_bytes < tmp64)
717 c->bg_bud_bytes = tmp64;
718 if (c->max_bud_bytes < tmp64 + c->leb_size)
719 c->max_bud_bytes = tmp64 + c->leb_size;
721 err = ubifs_calc_lpt_geom(c);
722 if (err)
723 return err;
725 /* Initialize effective LEB size used in budgeting calculations */
726 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
727 return 0;
731 * init_constants_master - initialize UBIFS constants.
732 * @c: UBIFS file-system description object
734 * This is a helper function which initializes various UBIFS constants after
735 * the master node has been read. It also checks various UBIFS parameters and
736 * makes sure they are all right.
738 static void init_constants_master(struct ubifs_info *c)
740 long long tmp64;
742 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
743 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
746 * Calculate total amount of FS blocks. This number is not used
747 * internally because it does not make much sense for UBIFS, but it is
748 * necessary to report something for the 'statfs()' call.
750 * Subtract the LEB reserved for GC, the LEB which is reserved for
751 * deletions, minimum LEBs for the index, and assume only one journal
752 * head is available.
754 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
755 tmp64 *= (long long)c->leb_size - c->leb_overhead;
756 tmp64 = ubifs_reported_space(c, tmp64);
757 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
761 * take_gc_lnum - reserve GC LEB.
762 * @c: UBIFS file-system description object
764 * This function ensures that the LEB reserved for garbage collection is marked
765 * as "taken" in lprops. We also have to set free space to LEB size and dirty
766 * space to zero, because lprops may contain out-of-date information if the
767 * file-system was un-mounted before it has been committed. This function
768 * returns zero in case of success and a negative error code in case of
769 * failure.
771 static int take_gc_lnum(struct ubifs_info *c)
773 int err;
775 if (c->gc_lnum == -1) {
776 ubifs_err(c, "no LEB for GC");
777 return -EINVAL;
780 /* And we have to tell lprops that this LEB is taken */
781 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
782 LPROPS_TAKEN, 0, 0);
783 return err;
787 * alloc_wbufs - allocate write-buffers.
788 * @c: UBIFS file-system description object
790 * This helper function allocates and initializes UBIFS write-buffers. Returns
791 * zero in case of success and %-ENOMEM in case of failure.
793 static int alloc_wbufs(struct ubifs_info *c)
795 int i, err;
797 c->jheads = kcalloc(c->jhead_cnt, sizeof(struct ubifs_jhead),
798 GFP_KERNEL);
799 if (!c->jheads)
800 return -ENOMEM;
802 /* Initialize journal heads */
803 for (i = 0; i < c->jhead_cnt; i++) {
804 INIT_LIST_HEAD(&c->jheads[i].buds_list);
805 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
806 if (err)
807 return err;
809 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
810 c->jheads[i].wbuf.jhead = i;
811 c->jheads[i].grouped = 1;
815 * Garbage Collector head does not need to be synchronized by timer.
816 * Also GC head nodes are not grouped.
818 c->jheads[GCHD].wbuf.no_timer = 1;
819 c->jheads[GCHD].grouped = 0;
821 return 0;
825 * free_wbufs - free write-buffers.
826 * @c: UBIFS file-system description object
828 static void free_wbufs(struct ubifs_info *c)
830 int i;
832 if (c->jheads) {
833 for (i = 0; i < c->jhead_cnt; i++) {
834 kfree(c->jheads[i].wbuf.buf);
835 kfree(c->jheads[i].wbuf.inodes);
837 kfree(c->jheads);
838 c->jheads = NULL;
843 * free_orphans - free orphans.
844 * @c: UBIFS file-system description object
846 static void free_orphans(struct ubifs_info *c)
848 struct ubifs_orphan *orph;
850 while (c->orph_dnext) {
851 orph = c->orph_dnext;
852 c->orph_dnext = orph->dnext;
853 list_del(&orph->list);
854 kfree(orph);
857 while (!list_empty(&c->orph_list)) {
858 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
859 list_del(&orph->list);
860 kfree(orph);
861 ubifs_err(c, "orphan list not empty at unmount");
864 vfree(c->orph_buf);
865 c->orph_buf = NULL;
869 * free_buds - free per-bud objects.
870 * @c: UBIFS file-system description object
872 static void free_buds(struct ubifs_info *c)
874 struct ubifs_bud *bud, *n;
876 rbtree_postorder_for_each_entry_safe(bud, n, &c->buds, rb)
877 kfree(bud);
881 * check_volume_empty - check if the UBI volume is empty.
882 * @c: UBIFS file-system description object
884 * This function checks if the UBIFS volume is empty by looking if its LEBs are
885 * mapped or not. The result of checking is stored in the @c->empty variable.
886 * Returns zero in case of success and a negative error code in case of
887 * failure.
889 static int check_volume_empty(struct ubifs_info *c)
891 int lnum, err;
893 c->empty = 1;
894 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
895 err = ubifs_is_mapped(c, lnum);
896 if (unlikely(err < 0))
897 return err;
898 if (err == 1) {
899 c->empty = 0;
900 break;
903 cond_resched();
906 return 0;
910 * UBIFS mount options.
912 * Opt_fast_unmount: do not run a journal commit before un-mounting
913 * Opt_norm_unmount: run a journal commit before un-mounting
914 * Opt_bulk_read: enable bulk-reads
915 * Opt_no_bulk_read: disable bulk-reads
916 * Opt_chk_data_crc: check CRCs when reading data nodes
917 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
918 * Opt_override_compr: override default compressor
919 * Opt_err: just end of array marker
921 enum {
922 Opt_fast_unmount,
923 Opt_norm_unmount,
924 Opt_bulk_read,
925 Opt_no_bulk_read,
926 Opt_chk_data_crc,
927 Opt_no_chk_data_crc,
928 Opt_override_compr,
929 Opt_err,
932 static const match_table_t tokens = {
933 {Opt_fast_unmount, "fast_unmount"},
934 {Opt_norm_unmount, "norm_unmount"},
935 {Opt_bulk_read, "bulk_read"},
936 {Opt_no_bulk_read, "no_bulk_read"},
937 {Opt_chk_data_crc, "chk_data_crc"},
938 {Opt_no_chk_data_crc, "no_chk_data_crc"},
939 {Opt_override_compr, "compr=%s"},
940 {Opt_err, NULL},
944 * parse_standard_option - parse a standard mount option.
945 * @option: the option to parse
947 * Normally, standard mount options like "sync" are passed to file-systems as
948 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
949 * be present in the options string. This function tries to deal with this
950 * situation and parse standard options. Returns 0 if the option was not
951 * recognized, and the corresponding integer flag if it was.
953 * UBIFS is only interested in the "sync" option, so do not check for anything
954 * else.
956 static int parse_standard_option(const char *option)
959 pr_notice("UBIFS: parse %s\n", option);
960 if (!strcmp(option, "sync"))
961 return MS_SYNCHRONOUS;
962 return 0;
966 * ubifs_parse_options - parse mount parameters.
967 * @c: UBIFS file-system description object
968 * @options: parameters to parse
969 * @is_remount: non-zero if this is FS re-mount
971 * This function parses UBIFS mount options and returns zero in case success
972 * and a negative error code in case of failure.
974 static int ubifs_parse_options(struct ubifs_info *c, char *options,
975 int is_remount)
977 char *p;
978 substring_t args[MAX_OPT_ARGS];
980 if (!options)
981 return 0;
983 while ((p = strsep(&options, ","))) {
984 int token;
986 if (!*p)
987 continue;
989 token = match_token(p, tokens, args);
990 switch (token) {
992 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
993 * We accept them in order to be backward-compatible. But this
994 * should be removed at some point.
996 case Opt_fast_unmount:
997 c->mount_opts.unmount_mode = 2;
998 break;
999 case Opt_norm_unmount:
1000 c->mount_opts.unmount_mode = 1;
1001 break;
1002 case Opt_bulk_read:
1003 c->mount_opts.bulk_read = 2;
1004 c->bulk_read = 1;
1005 break;
1006 case Opt_no_bulk_read:
1007 c->mount_opts.bulk_read = 1;
1008 c->bulk_read = 0;
1009 break;
1010 case Opt_chk_data_crc:
1011 c->mount_opts.chk_data_crc = 2;
1012 c->no_chk_data_crc = 0;
1013 break;
1014 case Opt_no_chk_data_crc:
1015 c->mount_opts.chk_data_crc = 1;
1016 c->no_chk_data_crc = 1;
1017 break;
1018 case Opt_override_compr:
1020 char *name = match_strdup(&args[0]);
1022 if (!name)
1023 return -ENOMEM;
1024 if (!strcmp(name, "none"))
1025 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1026 else if (!strcmp(name, "lzo"))
1027 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1028 else if (!strcmp(name, "zlib"))
1029 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1030 else {
1031 ubifs_err(c, "unknown compressor \"%s\"", name); //FIXME: is c ready?
1032 kfree(name);
1033 return -EINVAL;
1035 kfree(name);
1036 c->mount_opts.override_compr = 1;
1037 c->default_compr = c->mount_opts.compr_type;
1038 break;
1040 default:
1042 unsigned long flag;
1043 struct super_block *sb = c->vfs_sb;
1045 flag = parse_standard_option(p);
1046 if (!flag) {
1047 ubifs_err(c, "unrecognized mount option \"%s\" or missing value",
1049 return -EINVAL;
1051 sb->s_flags |= flag;
1052 break;
1057 return 0;
1061 * destroy_journal - destroy journal data structures.
1062 * @c: UBIFS file-system description object
1064 * This function destroys journal data structures including those that may have
1065 * been created by recovery functions.
1067 static void destroy_journal(struct ubifs_info *c)
1069 while (!list_empty(&c->unclean_leb_list)) {
1070 struct ubifs_unclean_leb *ucleb;
1072 ucleb = list_entry(c->unclean_leb_list.next,
1073 struct ubifs_unclean_leb, list);
1074 list_del(&ucleb->list);
1075 kfree(ucleb);
1077 while (!list_empty(&c->old_buds)) {
1078 struct ubifs_bud *bud;
1080 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1081 list_del(&bud->list);
1082 kfree(bud);
1084 ubifs_destroy_idx_gc(c);
1085 ubifs_destroy_size_tree(c);
1086 ubifs_tnc_close(c);
1087 free_buds(c);
1091 * bu_init - initialize bulk-read information.
1092 * @c: UBIFS file-system description object
1094 static void bu_init(struct ubifs_info *c)
1096 ubifs_assert(c->bulk_read == 1);
1098 if (c->bu.buf)
1099 return; /* Already initialized */
1101 again:
1102 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1103 if (!c->bu.buf) {
1104 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1105 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1106 goto again;
1109 /* Just disable bulk-read */
1110 ubifs_warn(c, "cannot allocate %d bytes of memory for bulk-read, disabling it",
1111 c->max_bu_buf_len);
1112 c->mount_opts.bulk_read = 1;
1113 c->bulk_read = 0;
1114 return;
1119 * check_free_space - check if there is enough free space to mount.
1120 * @c: UBIFS file-system description object
1122 * This function makes sure UBIFS has enough free space to be mounted in
1123 * read/write mode. UBIFS must always have some free space to allow deletions.
1125 static int check_free_space(struct ubifs_info *c)
1127 ubifs_assert(c->dark_wm > 0);
1128 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1129 ubifs_err(c, "insufficient free space to mount in R/W mode");
1130 ubifs_dump_budg(c, &c->bi);
1131 ubifs_dump_lprops(c);
1132 return -ENOSPC;
1134 return 0;
1138 * mount_ubifs - mount UBIFS file-system.
1139 * @c: UBIFS file-system description object
1141 * This function mounts UBIFS file system. Returns zero in case of success and
1142 * a negative error code in case of failure.
1144 static int mount_ubifs(struct ubifs_info *c)
1146 int err;
1147 long long x, y;
1148 size_t sz;
1150 c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1151 /* Suppress error messages while probing if MS_SILENT is set */
1152 c->probing = !!(c->vfs_sb->s_flags & MS_SILENT);
1154 err = init_constants_early(c);
1155 if (err)
1156 return err;
1158 err = ubifs_debugging_init(c);
1159 if (err)
1160 return err;
1162 err = check_volume_empty(c);
1163 if (err)
1164 goto out_free;
1166 if (c->empty && (c->ro_mount || c->ro_media)) {
1168 * This UBI volume is empty, and read-only, or the file system
1169 * is mounted read-only - we cannot format it.
1171 ubifs_err(c, "can't format empty UBI volume: read-only %s",
1172 c->ro_media ? "UBI volume" : "mount");
1173 err = -EROFS;
1174 goto out_free;
1177 if (c->ro_media && !c->ro_mount) {
1178 ubifs_err(c, "cannot mount read-write - read-only media");
1179 err = -EROFS;
1180 goto out_free;
1184 * The requirement for the buffer is that it should fit indexing B-tree
1185 * height amount of integers. We assume the height if the TNC tree will
1186 * never exceed 64.
1188 err = -ENOMEM;
1189 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1190 if (!c->bottom_up_buf)
1191 goto out_free;
1193 c->sbuf = vmalloc(c->leb_size);
1194 if (!c->sbuf)
1195 goto out_free;
1197 if (!c->ro_mount) {
1198 c->ileb_buf = vmalloc(c->leb_size);
1199 if (!c->ileb_buf)
1200 goto out_free;
1203 if (c->bulk_read == 1)
1204 bu_init(c);
1206 if (!c->ro_mount) {
1207 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
1208 GFP_KERNEL);
1209 if (!c->write_reserve_buf)
1210 goto out_free;
1213 c->mounting = 1;
1215 err = ubifs_read_superblock(c);
1216 if (err)
1217 goto out_free;
1219 c->probing = 0;
1222 * Make sure the compressor which is set as default in the superblock
1223 * or overridden by mount options is actually compiled in.
1225 if (!ubifs_compr_present(c->default_compr)) {
1226 ubifs_err(c, "'compressor \"%s\" is not compiled in",
1227 ubifs_compr_name(c->default_compr));
1228 err = -ENOTSUPP;
1229 goto out_free;
1232 err = init_constants_sb(c);
1233 if (err)
1234 goto out_free;
1236 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1237 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1238 c->cbuf = kmalloc(sz, GFP_NOFS);
1239 if (!c->cbuf) {
1240 err = -ENOMEM;
1241 goto out_free;
1244 err = alloc_wbufs(c);
1245 if (err)
1246 goto out_cbuf;
1248 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1249 if (!c->ro_mount) {
1250 /* Create background thread */
1251 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1252 if (IS_ERR(c->bgt)) {
1253 err = PTR_ERR(c->bgt);
1254 c->bgt = NULL;
1255 ubifs_err(c, "cannot spawn \"%s\", error %d",
1256 c->bgt_name, err);
1257 goto out_wbufs;
1259 wake_up_process(c->bgt);
1262 err = ubifs_read_master(c);
1263 if (err)
1264 goto out_master;
1266 init_constants_master(c);
1268 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1269 ubifs_msg(c, "recovery needed");
1270 c->need_recovery = 1;
1273 if (c->need_recovery && !c->ro_mount) {
1274 err = ubifs_recover_inl_heads(c, c->sbuf);
1275 if (err)
1276 goto out_master;
1279 err = ubifs_lpt_init(c, 1, !c->ro_mount);
1280 if (err)
1281 goto out_master;
1283 if (!c->ro_mount && c->space_fixup) {
1284 err = ubifs_fixup_free_space(c);
1285 if (err)
1286 goto out_lpt;
1289 if (!c->ro_mount && !c->need_recovery) {
1291 * Set the "dirty" flag so that if we reboot uncleanly we
1292 * will notice this immediately on the next mount.
1294 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1295 err = ubifs_write_master(c);
1296 if (err)
1297 goto out_lpt;
1300 err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1301 if (err)
1302 goto out_lpt;
1304 err = ubifs_replay_journal(c);
1305 if (err)
1306 goto out_journal;
1308 /* Calculate 'min_idx_lebs' after journal replay */
1309 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1311 err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1312 if (err)
1313 goto out_orphans;
1315 if (!c->ro_mount) {
1316 int lnum;
1318 err = check_free_space(c);
1319 if (err)
1320 goto out_orphans;
1322 /* Check for enough log space */
1323 lnum = c->lhead_lnum + 1;
1324 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1325 lnum = UBIFS_LOG_LNUM;
1326 if (lnum == c->ltail_lnum) {
1327 err = ubifs_consolidate_log(c);
1328 if (err)
1329 goto out_orphans;
1332 if (c->need_recovery) {
1333 err = ubifs_recover_size(c);
1334 if (err)
1335 goto out_orphans;
1336 err = ubifs_rcvry_gc_commit(c);
1337 if (err)
1338 goto out_orphans;
1339 } else {
1340 err = take_gc_lnum(c);
1341 if (err)
1342 goto out_orphans;
1345 * GC LEB may contain garbage if there was an unclean
1346 * reboot, and it should be un-mapped.
1348 err = ubifs_leb_unmap(c, c->gc_lnum);
1349 if (err)
1350 goto out_orphans;
1353 err = dbg_check_lprops(c);
1354 if (err)
1355 goto out_orphans;
1356 } else if (c->need_recovery) {
1357 err = ubifs_recover_size(c);
1358 if (err)
1359 goto out_orphans;
1360 } else {
1362 * Even if we mount read-only, we have to set space in GC LEB
1363 * to proper value because this affects UBIFS free space
1364 * reporting. We do not want to have a situation when
1365 * re-mounting from R/O to R/W changes amount of free space.
1367 err = take_gc_lnum(c);
1368 if (err)
1369 goto out_orphans;
1372 spin_lock(&ubifs_infos_lock);
1373 list_add_tail(&c->infos_list, &ubifs_infos);
1374 spin_unlock(&ubifs_infos_lock);
1376 if (c->need_recovery) {
1377 if (c->ro_mount)
1378 ubifs_msg(c, "recovery deferred");
1379 else {
1380 c->need_recovery = 0;
1381 ubifs_msg(c, "recovery completed");
1383 * GC LEB has to be empty and taken at this point. But
1384 * the journal head LEBs may also be accounted as
1385 * "empty taken" if they are empty.
1387 ubifs_assert(c->lst.taken_empty_lebs > 0);
1389 } else
1390 ubifs_assert(c->lst.taken_empty_lebs > 0);
1392 err = dbg_check_filesystem(c);
1393 if (err)
1394 goto out_infos;
1396 err = dbg_debugfs_init_fs(c);
1397 if (err)
1398 goto out_infos;
1400 c->mounting = 0;
1402 ubifs_msg(c, "UBIFS: mounted UBI device %d, volume %d, name \"%s\"%s",
1403 c->vi.ubi_num, c->vi.vol_id, c->vi.name,
1404 c->ro_mount ? ", R/O mode" : "");
1405 x = (long long)c->main_lebs * c->leb_size;
1406 y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1407 ubifs_msg(c, "LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes",
1408 c->leb_size, c->leb_size >> 10, c->min_io_size,
1409 c->max_write_size);
1410 ubifs_msg(c, "FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)",
1411 x, x >> 20, c->main_lebs,
1412 y, y >> 20, c->log_lebs + c->max_bud_cnt);
1413 ubifs_msg(c, "reserved for root: %llu bytes (%llu KiB)",
1414 c->report_rp_size, c->report_rp_size >> 10);
1415 ubifs_msg(c, "media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s",
1416 c->fmt_version, c->ro_compat_version,
1417 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid,
1418 c->big_lpt ? ", big LPT model" : ", small LPT model");
1420 dbg_gen("default compressor: %s", ubifs_compr_name(c->default_compr));
1421 dbg_gen("data journal heads: %d",
1422 c->jhead_cnt - NONDATA_JHEADS_CNT);
1423 dbg_gen("log LEBs: %d (%d - %d)",
1424 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1425 dbg_gen("LPT area LEBs: %d (%d - %d)",
1426 c->lpt_lebs, c->lpt_first, c->lpt_last);
1427 dbg_gen("orphan area LEBs: %d (%d - %d)",
1428 c->orph_lebs, c->orph_first, c->orph_last);
1429 dbg_gen("main area LEBs: %d (%d - %d)",
1430 c->main_lebs, c->main_first, c->leb_cnt - 1);
1431 dbg_gen("index LEBs: %d", c->lst.idx_lebs);
1432 dbg_gen("total index bytes: %lld (%lld KiB, %lld MiB)",
1433 c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1434 c->bi.old_idx_sz >> 20);
1435 dbg_gen("key hash type: %d", c->key_hash_type);
1436 dbg_gen("tree fanout: %d", c->fanout);
1437 dbg_gen("reserved GC LEB: %d", c->gc_lnum);
1438 dbg_gen("max. znode size %d", c->max_znode_sz);
1439 dbg_gen("max. index node size %d", c->max_idx_node_sz);
1440 dbg_gen("node sizes: data %zu, inode %zu, dentry %zu",
1441 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1442 dbg_gen("node sizes: trun %zu, sb %zu, master %zu",
1443 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1444 dbg_gen("node sizes: ref %zu, cmt. start %zu, orph %zu",
1445 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1446 dbg_gen("max. node sizes: data %zu, inode %zu dentry %zu, idx %d",
1447 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1448 UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1449 dbg_gen("dead watermark: %d", c->dead_wm);
1450 dbg_gen("dark watermark: %d", c->dark_wm);
1451 dbg_gen("LEB overhead: %d", c->leb_overhead);
1452 x = (long long)c->main_lebs * c->dark_wm;
1453 dbg_gen("max. dark space: %lld (%lld KiB, %lld MiB)",
1454 x, x >> 10, x >> 20);
1455 dbg_gen("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1456 c->max_bud_bytes, c->max_bud_bytes >> 10,
1457 c->max_bud_bytes >> 20);
1458 dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1459 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1460 c->bg_bud_bytes >> 20);
1461 dbg_gen("current bud bytes %lld (%lld KiB, %lld MiB)",
1462 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1463 dbg_gen("max. seq. number: %llu", c->max_sqnum);
1464 dbg_gen("commit number: %llu", c->cmt_no);
1466 return 0;
1468 out_infos:
1469 spin_lock(&ubifs_infos_lock);
1470 list_del(&c->infos_list);
1471 spin_unlock(&ubifs_infos_lock);
1472 out_orphans:
1473 free_orphans(c);
1474 out_journal:
1475 destroy_journal(c);
1476 out_lpt:
1477 ubifs_lpt_free(c, 0);
1478 out_master:
1479 kfree(c->mst_node);
1480 kfree(c->rcvrd_mst_node);
1481 if (c->bgt)
1482 kthread_stop(c->bgt);
1483 out_wbufs:
1484 free_wbufs(c);
1485 out_cbuf:
1486 kfree(c->cbuf);
1487 out_free:
1488 kfree(c->write_reserve_buf);
1489 kfree(c->bu.buf);
1490 vfree(c->ileb_buf);
1491 vfree(c->sbuf);
1492 kfree(c->bottom_up_buf);
1493 ubifs_debugging_exit(c);
1494 return err;
1498 * ubifs_umount - un-mount UBIFS file-system.
1499 * @c: UBIFS file-system description object
1501 * Note, this function is called to free allocated resourced when un-mounting,
1502 * as well as free resources when an error occurred while we were half way
1503 * through mounting (error path cleanup function). So it has to make sure the
1504 * resource was actually allocated before freeing it.
1506 static void ubifs_umount(struct ubifs_info *c)
1508 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1509 c->vi.vol_id);
1511 dbg_debugfs_exit_fs(c);
1512 spin_lock(&ubifs_infos_lock);
1513 list_del(&c->infos_list);
1514 spin_unlock(&ubifs_infos_lock);
1516 if (c->bgt)
1517 kthread_stop(c->bgt);
1519 destroy_journal(c);
1520 free_wbufs(c);
1521 free_orphans(c);
1522 ubifs_lpt_free(c, 0);
1524 kfree(c->cbuf);
1525 kfree(c->rcvrd_mst_node);
1526 kfree(c->mst_node);
1527 kfree(c->write_reserve_buf);
1528 kfree(c->bu.buf);
1529 vfree(c->ileb_buf);
1530 vfree(c->sbuf);
1531 kfree(c->bottom_up_buf);
1532 ubifs_debugging_exit(c);
1536 * ubifs_remount_rw - re-mount in read-write mode.
1537 * @c: UBIFS file-system description object
1539 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1540 * mode. This function allocates the needed resources and re-mounts UBIFS in
1541 * read-write mode.
1543 static int ubifs_remount_rw(struct ubifs_info *c)
1545 int err, lnum;
1547 if (c->rw_incompat) {
1548 ubifs_err(c, "the file-system is not R/W-compatible");
1549 ubifs_msg(c, "on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
1550 c->fmt_version, c->ro_compat_version,
1551 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1552 return -EROFS;
1555 mutex_lock(&c->umount_mutex);
1556 dbg_save_space_info(c);
1557 c->remounting_rw = 1;
1558 c->ro_mount = 0;
1560 if (c->space_fixup) {
1561 err = ubifs_fixup_free_space(c);
1562 if (err)
1563 goto out;
1566 err = check_free_space(c);
1567 if (err)
1568 goto out;
1570 if (c->old_leb_cnt != c->leb_cnt) {
1571 struct ubifs_sb_node *sup;
1573 sup = ubifs_read_sb_node(c);
1574 if (IS_ERR(sup)) {
1575 err = PTR_ERR(sup);
1576 goto out;
1578 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1579 err = ubifs_write_sb_node(c, sup);
1580 kfree(sup);
1581 if (err)
1582 goto out;
1585 if (c->need_recovery) {
1586 ubifs_msg(c, "completing deferred recovery");
1587 err = ubifs_write_rcvrd_mst_node(c);
1588 if (err)
1589 goto out;
1590 err = ubifs_recover_size(c);
1591 if (err)
1592 goto out;
1593 err = ubifs_clean_lebs(c, c->sbuf);
1594 if (err)
1595 goto out;
1596 err = ubifs_recover_inl_heads(c, c->sbuf);
1597 if (err)
1598 goto out;
1599 } else {
1600 /* A readonly mount is not allowed to have orphans */
1601 ubifs_assert(c->tot_orphans == 0);
1602 err = ubifs_clear_orphans(c);
1603 if (err)
1604 goto out;
1607 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1608 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1609 err = ubifs_write_master(c);
1610 if (err)
1611 goto out;
1614 c->ileb_buf = vmalloc(c->leb_size);
1615 if (!c->ileb_buf) {
1616 err = -ENOMEM;
1617 goto out;
1620 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
1621 if (!c->write_reserve_buf) {
1622 err = -ENOMEM;
1623 goto out;
1626 err = ubifs_lpt_init(c, 0, 1);
1627 if (err)
1628 goto out;
1630 /* Create background thread */
1631 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1632 if (IS_ERR(c->bgt)) {
1633 err = PTR_ERR(c->bgt);
1634 c->bgt = NULL;
1635 ubifs_err(c, "cannot spawn \"%s\", error %d",
1636 c->bgt_name, err);
1637 goto out;
1639 wake_up_process(c->bgt);
1641 c->orph_buf = vmalloc(c->leb_size);
1642 if (!c->orph_buf) {
1643 err = -ENOMEM;
1644 goto out;
1647 /* Check for enough log space */
1648 lnum = c->lhead_lnum + 1;
1649 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1650 lnum = UBIFS_LOG_LNUM;
1651 if (lnum == c->ltail_lnum) {
1652 err = ubifs_consolidate_log(c);
1653 if (err)
1654 goto out;
1657 if (c->need_recovery)
1658 err = ubifs_rcvry_gc_commit(c);
1659 else
1660 err = ubifs_leb_unmap(c, c->gc_lnum);
1661 if (err)
1662 goto out;
1664 dbg_gen("re-mounted read-write");
1665 c->remounting_rw = 0;
1667 if (c->need_recovery) {
1668 c->need_recovery = 0;
1669 ubifs_msg(c, "deferred recovery completed");
1670 } else {
1672 * Do not run the debugging space check if the were doing
1673 * recovery, because when we saved the information we had the
1674 * file-system in a state where the TNC and lprops has been
1675 * modified in memory, but all the I/O operations (including a
1676 * commit) were deferred. So the file-system was in
1677 * "non-committed" state. Now the file-system is in committed
1678 * state, and of course the amount of free space will change
1679 * because, for example, the old index size was imprecise.
1681 err = dbg_check_space_info(c);
1684 mutex_unlock(&c->umount_mutex);
1685 return err;
1687 out:
1688 c->ro_mount = 1;
1689 vfree(c->orph_buf);
1690 c->orph_buf = NULL;
1691 if (c->bgt) {
1692 kthread_stop(c->bgt);
1693 c->bgt = NULL;
1695 free_wbufs(c);
1696 kfree(c->write_reserve_buf);
1697 c->write_reserve_buf = NULL;
1698 vfree(c->ileb_buf);
1699 c->ileb_buf = NULL;
1700 ubifs_lpt_free(c, 1);
1701 c->remounting_rw = 0;
1702 mutex_unlock(&c->umount_mutex);
1703 return err;
1707 * ubifs_remount_ro - re-mount in read-only mode.
1708 * @c: UBIFS file-system description object
1710 * We assume VFS has stopped writing. Possibly the background thread could be
1711 * running a commit, however kthread_stop will wait in that case.
1713 static void ubifs_remount_ro(struct ubifs_info *c)
1715 int i, err;
1717 ubifs_assert(!c->need_recovery);
1718 ubifs_assert(!c->ro_mount);
1720 mutex_lock(&c->umount_mutex);
1721 if (c->bgt) {
1722 kthread_stop(c->bgt);
1723 c->bgt = NULL;
1726 dbg_save_space_info(c);
1728 for (i = 0; i < c->jhead_cnt; i++)
1729 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1731 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1732 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1733 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1734 err = ubifs_write_master(c);
1735 if (err)
1736 ubifs_ro_mode(c, err);
1738 vfree(c->orph_buf);
1739 c->orph_buf = NULL;
1740 kfree(c->write_reserve_buf);
1741 c->write_reserve_buf = NULL;
1742 vfree(c->ileb_buf);
1743 c->ileb_buf = NULL;
1744 ubifs_lpt_free(c, 1);
1745 c->ro_mount = 1;
1746 err = dbg_check_space_info(c);
1747 if (err)
1748 ubifs_ro_mode(c, err);
1749 mutex_unlock(&c->umount_mutex);
1752 static void ubifs_put_super(struct super_block *sb)
1754 int i;
1755 struct ubifs_info *c = sb->s_fs_info;
1757 ubifs_msg(c, "un-mount UBI device %d", c->vi.ubi_num);
1760 * The following asserts are only valid if there has not been a failure
1761 * of the media. For example, there will be dirty inodes if we failed
1762 * to write them back because of I/O errors.
1764 if (!c->ro_error) {
1765 ubifs_assert(c->bi.idx_growth == 0);
1766 ubifs_assert(c->bi.dd_growth == 0);
1767 ubifs_assert(c->bi.data_growth == 0);
1771 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1772 * and file system un-mount. Namely, it prevents the shrinker from
1773 * picking this superblock for shrinking - it will be just skipped if
1774 * the mutex is locked.
1776 mutex_lock(&c->umount_mutex);
1777 if (!c->ro_mount) {
1779 * First of all kill the background thread to make sure it does
1780 * not interfere with un-mounting and freeing resources.
1782 if (c->bgt) {
1783 kthread_stop(c->bgt);
1784 c->bgt = NULL;
1788 * On fatal errors c->ro_error is set to 1, in which case we do
1789 * not write the master node.
1791 if (!c->ro_error) {
1792 int err;
1794 /* Synchronize write-buffers */
1795 for (i = 0; i < c->jhead_cnt; i++)
1796 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1799 * We are being cleanly unmounted which means the
1800 * orphans were killed - indicate this in the master
1801 * node. Also save the reserved GC LEB number.
1803 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1804 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1805 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1806 err = ubifs_write_master(c);
1807 if (err)
1809 * Recovery will attempt to fix the master area
1810 * next mount, so we just print a message and
1811 * continue to unmount normally.
1813 ubifs_err(c, "failed to write master node, error %d",
1814 err);
1815 } else {
1816 for (i = 0; i < c->jhead_cnt; i++)
1817 /* Make sure write-buffer timers are canceled */
1818 hrtimer_cancel(&c->jheads[i].wbuf.timer);
1822 ubifs_umount(c);
1823 bdi_destroy(&c->bdi);
1824 ubi_close_volume(c->ubi);
1825 mutex_unlock(&c->umount_mutex);
1828 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1830 int err;
1831 struct ubifs_info *c = sb->s_fs_info;
1833 sync_filesystem(sb);
1834 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1836 err = ubifs_parse_options(c, data, 1);
1837 if (err) {
1838 ubifs_err(c, "invalid or unknown remount parameter");
1839 return err;
1842 if (c->ro_mount && !(*flags & MS_RDONLY)) {
1843 if (c->ro_error) {
1844 ubifs_msg(c, "cannot re-mount R/W due to prior errors");
1845 return -EROFS;
1847 if (c->ro_media) {
1848 ubifs_msg(c, "cannot re-mount R/W - UBI volume is R/O");
1849 return -EROFS;
1851 err = ubifs_remount_rw(c);
1852 if (err)
1853 return err;
1854 } else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1855 if (c->ro_error) {
1856 ubifs_msg(c, "cannot re-mount R/O due to prior errors");
1857 return -EROFS;
1859 ubifs_remount_ro(c);
1862 if (c->bulk_read == 1)
1863 bu_init(c);
1864 else {
1865 dbg_gen("disable bulk-read");
1866 kfree(c->bu.buf);
1867 c->bu.buf = NULL;
1870 ubifs_assert(c->lst.taken_empty_lebs > 0);
1871 return 0;
1874 const struct super_operations ubifs_super_operations = {
1875 .alloc_inode = ubifs_alloc_inode,
1876 .destroy_inode = ubifs_destroy_inode,
1877 .put_super = ubifs_put_super,
1878 .write_inode = ubifs_write_inode,
1879 .evict_inode = ubifs_evict_inode,
1880 .statfs = ubifs_statfs,
1881 .dirty_inode = ubifs_dirty_inode,
1882 .remount_fs = ubifs_remount_fs,
1883 .show_options = ubifs_show_options,
1884 .sync_fs = ubifs_sync_fs,
1888 * open_ubi - parse UBI device name string and open the UBI device.
1889 * @name: UBI volume name
1890 * @mode: UBI volume open mode
1892 * The primary method of mounting UBIFS is by specifying the UBI volume
1893 * character device node path. However, UBIFS may also be mounted withoug any
1894 * character device node using one of the following methods:
1896 * o ubiX_Y - mount UBI device number X, volume Y;
1897 * o ubiY - mount UBI device number 0, volume Y;
1898 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1899 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1901 * Alternative '!' separator may be used instead of ':' (because some shells
1902 * like busybox may interpret ':' as an NFS host name separator). This function
1903 * returns UBI volume description object in case of success and a negative
1904 * error code in case of failure.
1906 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1908 struct ubi_volume_desc *ubi;
1909 int dev, vol;
1910 char *endptr;
1912 /* First, try to open using the device node path method */
1913 ubi = ubi_open_volume_path(name, mode);
1914 if (!IS_ERR(ubi))
1915 return ubi;
1917 /* Try the "nodev" method */
1918 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1919 return ERR_PTR(-EINVAL);
1921 /* ubi:NAME method */
1922 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1923 return ubi_open_volume_nm(0, name + 4, mode);
1925 if (!isdigit(name[3]))
1926 return ERR_PTR(-EINVAL);
1928 dev = simple_strtoul(name + 3, &endptr, 0);
1930 /* ubiY method */
1931 if (*endptr == '\0')
1932 return ubi_open_volume(0, dev, mode);
1934 /* ubiX_Y method */
1935 if (*endptr == '_' && isdigit(endptr[1])) {
1936 vol = simple_strtoul(endptr + 1, &endptr, 0);
1937 if (*endptr != '\0')
1938 return ERR_PTR(-EINVAL);
1939 return ubi_open_volume(dev, vol, mode);
1942 /* ubiX:NAME method */
1943 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1944 return ubi_open_volume_nm(dev, ++endptr, mode);
1946 return ERR_PTR(-EINVAL);
1949 static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1951 struct ubifs_info *c;
1953 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1954 if (c) {
1955 spin_lock_init(&c->cnt_lock);
1956 spin_lock_init(&c->cs_lock);
1957 spin_lock_init(&c->buds_lock);
1958 spin_lock_init(&c->space_lock);
1959 spin_lock_init(&c->orphan_lock);
1960 init_rwsem(&c->commit_sem);
1961 mutex_init(&c->lp_mutex);
1962 mutex_init(&c->tnc_mutex);
1963 mutex_init(&c->log_mutex);
1964 mutex_init(&c->umount_mutex);
1965 mutex_init(&c->bu_mutex);
1966 mutex_init(&c->write_reserve_mutex);
1967 init_waitqueue_head(&c->cmt_wq);
1968 c->buds = RB_ROOT;
1969 c->old_idx = RB_ROOT;
1970 c->size_tree = RB_ROOT;
1971 c->orph_tree = RB_ROOT;
1972 INIT_LIST_HEAD(&c->infos_list);
1973 INIT_LIST_HEAD(&c->idx_gc);
1974 INIT_LIST_HEAD(&c->replay_list);
1975 INIT_LIST_HEAD(&c->replay_buds);
1976 INIT_LIST_HEAD(&c->uncat_list);
1977 INIT_LIST_HEAD(&c->empty_list);
1978 INIT_LIST_HEAD(&c->freeable_list);
1979 INIT_LIST_HEAD(&c->frdi_idx_list);
1980 INIT_LIST_HEAD(&c->unclean_leb_list);
1981 INIT_LIST_HEAD(&c->old_buds);
1982 INIT_LIST_HEAD(&c->orph_list);
1983 INIT_LIST_HEAD(&c->orph_new);
1984 c->no_chk_data_crc = 1;
1986 c->highest_inum = UBIFS_FIRST_INO;
1987 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1989 ubi_get_volume_info(ubi, &c->vi);
1990 ubi_get_device_info(c->vi.ubi_num, &c->di);
1992 return c;
1995 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1997 struct ubifs_info *c = sb->s_fs_info;
1998 struct inode *root;
1999 int err;
2001 c->vfs_sb = sb;
2002 /* Re-open the UBI device in read-write mode */
2003 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2004 if (IS_ERR(c->ubi)) {
2005 err = PTR_ERR(c->ubi);
2006 goto out;
2010 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2011 * UBIFS, I/O is not deferred, it is done immediately in readpage,
2012 * which means the user would have to wait not just for their own I/O
2013 * but the read-ahead I/O as well i.e. completely pointless.
2015 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
2017 c->bdi.name = "ubifs",
2018 c->bdi.capabilities = 0;
2019 err = bdi_init(&c->bdi);
2020 if (err)
2021 goto out_close;
2022 err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
2023 c->vi.ubi_num, c->vi.vol_id);
2024 if (err)
2025 goto out_bdi;
2027 err = ubifs_parse_options(c, data, 0);
2028 if (err)
2029 goto out_bdi;
2031 sb->s_bdi = &c->bdi;
2032 sb->s_fs_info = c;
2033 sb->s_magic = UBIFS_SUPER_MAGIC;
2034 sb->s_blocksize = UBIFS_BLOCK_SIZE;
2035 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2036 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2037 if (c->max_inode_sz > MAX_LFS_FILESIZE)
2038 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2039 sb->s_op = &ubifs_super_operations;
2040 sb->s_xattr = ubifs_xattr_handlers;
2042 mutex_lock(&c->umount_mutex);
2043 err = mount_ubifs(c);
2044 if (err) {
2045 ubifs_assert(err < 0);
2046 goto out_unlock;
2049 /* Read the root inode */
2050 root = ubifs_iget(sb, UBIFS_ROOT_INO);
2051 if (IS_ERR(root)) {
2052 err = PTR_ERR(root);
2053 goto out_umount;
2056 sb->s_root = d_make_root(root);
2057 if (!sb->s_root) {
2058 err = -ENOMEM;
2059 goto out_umount;
2062 mutex_unlock(&c->umount_mutex);
2063 return 0;
2065 out_umount:
2066 ubifs_umount(c);
2067 out_unlock:
2068 mutex_unlock(&c->umount_mutex);
2069 out_bdi:
2070 bdi_destroy(&c->bdi);
2071 out_close:
2072 ubi_close_volume(c->ubi);
2073 out:
2074 return err;
2077 static int sb_test(struct super_block *sb, void *data)
2079 struct ubifs_info *c1 = data;
2080 struct ubifs_info *c = sb->s_fs_info;
2082 return c->vi.cdev == c1->vi.cdev;
2085 static int sb_set(struct super_block *sb, void *data)
2087 sb->s_fs_info = data;
2088 return set_anon_super(sb, NULL);
2091 static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2092 const char *name, void *data)
2094 struct ubi_volume_desc *ubi;
2095 struct ubifs_info *c;
2096 struct super_block *sb;
2097 int err;
2099 dbg_gen("name %s, flags %#x", name, flags);
2102 * Get UBI device number and volume ID. Mount it read-only so far
2103 * because this might be a new mount point, and UBI allows only one
2104 * read-write user at a time.
2106 ubi = open_ubi(name, UBI_READONLY);
2107 if (IS_ERR(ubi)) {
2108 pr_err("UBIFS error (pid: %d): cannot open \"%s\", error %d",
2109 current->pid, name, (int)PTR_ERR(ubi));
2110 return ERR_CAST(ubi);
2113 c = alloc_ubifs_info(ubi);
2114 if (!c) {
2115 err = -ENOMEM;
2116 goto out_close;
2119 dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2121 sb = sget(fs_type, sb_test, sb_set, flags, c);
2122 if (IS_ERR(sb)) {
2123 err = PTR_ERR(sb);
2124 kfree(c);
2125 goto out_close;
2128 if (sb->s_root) {
2129 struct ubifs_info *c1 = sb->s_fs_info;
2130 kfree(c);
2131 /* A new mount point for already mounted UBIFS */
2132 dbg_gen("this ubi volume is already mounted");
2133 if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2134 err = -EBUSY;
2135 goto out_deact;
2137 } else {
2138 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2139 if (err)
2140 goto out_deact;
2141 /* We do not support atime */
2142 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2145 /* 'fill_super()' opens ubi again so we must close it here */
2146 ubi_close_volume(ubi);
2148 return dget(sb->s_root);
2150 out_deact:
2151 deactivate_locked_super(sb);
2152 out_close:
2153 ubi_close_volume(ubi);
2154 return ERR_PTR(err);
2157 static void kill_ubifs_super(struct super_block *s)
2159 struct ubifs_info *c = s->s_fs_info;
2160 kill_anon_super(s);
2161 kfree(c);
2164 static struct file_system_type ubifs_fs_type = {
2165 .name = "ubifs",
2166 .owner = THIS_MODULE,
2167 .mount = ubifs_mount,
2168 .kill_sb = kill_ubifs_super,
2170 MODULE_ALIAS_FS("ubifs");
2173 * Inode slab cache constructor.
2175 static void inode_slab_ctor(void *obj)
2177 struct ubifs_inode *ui = obj;
2178 inode_init_once(&ui->vfs_inode);
2181 static int __init ubifs_init(void)
2183 int err;
2185 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2187 /* Make sure node sizes are 8-byte aligned */
2188 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2189 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2190 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2191 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2192 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2193 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2194 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2195 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2196 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2197 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2198 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2200 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2201 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2202 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2203 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2204 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2205 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2207 /* Check min. node size */
2208 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2209 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2210 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2211 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2213 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2214 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2215 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2216 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2218 /* Defined node sizes */
2219 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2220 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2221 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2222 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2225 * We use 2 bit wide bit-fields to store compression type, which should
2226 * be amended if more compressors are added. The bit-fields are:
2227 * @compr_type in 'struct ubifs_inode', @default_compr in
2228 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2230 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2233 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2234 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2236 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2237 pr_err("UBIFS error (pid %d): VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
2238 current->pid, (unsigned int)PAGE_CACHE_SIZE);
2239 return -EINVAL;
2242 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2243 sizeof(struct ubifs_inode), 0,
2244 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2245 &inode_slab_ctor);
2246 if (!ubifs_inode_slab)
2247 return -ENOMEM;
2249 err = register_shrinker(&ubifs_shrinker_info);
2250 if (err)
2251 goto out_slab;
2253 err = ubifs_compressors_init();
2254 if (err)
2255 goto out_shrinker;
2257 err = dbg_debugfs_init();
2258 if (err)
2259 goto out_compr;
2261 err = register_filesystem(&ubifs_fs_type);
2262 if (err) {
2263 pr_err("UBIFS error (pid %d): cannot register file system, error %d",
2264 current->pid, err);
2265 goto out_dbg;
2267 return 0;
2269 out_dbg:
2270 dbg_debugfs_exit();
2271 out_compr:
2272 ubifs_compressors_exit();
2273 out_shrinker:
2274 unregister_shrinker(&ubifs_shrinker_info);
2275 out_slab:
2276 kmem_cache_destroy(ubifs_inode_slab);
2277 return err;
2279 /* late_initcall to let compressors initialize first */
2280 late_initcall(ubifs_init);
2282 static void __exit ubifs_exit(void)
2284 ubifs_assert(list_empty(&ubifs_infos));
2285 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2287 dbg_debugfs_exit();
2288 ubifs_compressors_exit();
2289 unregister_shrinker(&ubifs_shrinker_info);
2292 * Make sure all delayed rcu free inodes are flushed before we
2293 * destroy cache.
2295 rcu_barrier();
2296 kmem_cache_destroy(ubifs_inode_slab);
2297 unregister_filesystem(&ubifs_fs_type);
2299 module_exit(ubifs_exit);
2301 MODULE_LICENSE("GPL");
2302 MODULE_VERSION(__stringify(UBIFS_VERSION));
2303 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2304 MODULE_DESCRIPTION("UBIFS - UBI File System");