audit: complex interfield comparison helper
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ubifs / super.c
blob63765d58445b61e7b871e7e1722fedd1668a23b9
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 .shrink = ubifs_shrinker,
53 .seeks = DEFAULT_SEEKS,
56 /**
57 * validate_inode - validate inode.
58 * @c: UBIFS file-system description object
59 * @inode: the inode to validate
61 * This is a helper function for 'ubifs_iget()' which validates various fields
62 * of a newly built inode to make sure they contain sane values and prevent
63 * possible vulnerabilities. Returns zero if the inode is all right and
64 * a non-zero error code if not.
66 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
68 int err;
69 const struct ubifs_inode *ui = ubifs_inode(inode);
71 if (inode->i_size > c->max_inode_sz) {
72 ubifs_err("inode is too large (%lld)",
73 (long long)inode->i_size);
74 return 1;
77 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
78 ubifs_err("unknown compression type %d", ui->compr_type);
79 return 2;
82 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
83 return 3;
85 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
86 return 4;
88 if (ui->xattr && !S_ISREG(inode->i_mode))
89 return 5;
91 if (!ubifs_compr_present(ui->compr_type)) {
92 ubifs_warn("inode %lu uses '%s' compression, but it was not "
93 "compiled in", inode->i_ino,
94 ubifs_compr_name(ui->compr_type));
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 inode->i_uid = le32_to_cpu(ino->uid);
134 inode->i_gid = le32_to_cpu(ino->gid);
135 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
136 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
137 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
138 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
139 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
140 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
141 inode->i_mode = le32_to_cpu(ino->mode);
142 inode->i_size = le64_to_cpu(ino->size);
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 /* Disable read-ahead */
160 inode->i_mapping->backing_dev_info = &c->bdi;
162 switch (inode->i_mode & S_IFMT) {
163 case S_IFREG:
164 inode->i_mapping->a_ops = &ubifs_file_address_operations;
165 inode->i_op = &ubifs_file_inode_operations;
166 inode->i_fop = &ubifs_file_operations;
167 if (ui->xattr) {
168 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
169 if (!ui->data) {
170 err = -ENOMEM;
171 goto out_ino;
173 memcpy(ui->data, ino->data, ui->data_len);
174 ((char *)ui->data)[ui->data_len] = '\0';
175 } else if (ui->data_len != 0) {
176 err = 10;
177 goto out_invalid;
179 break;
180 case S_IFDIR:
181 inode->i_op = &ubifs_dir_inode_operations;
182 inode->i_fop = &ubifs_dir_operations;
183 if (ui->data_len != 0) {
184 err = 11;
185 goto out_invalid;
187 break;
188 case S_IFLNK:
189 inode->i_op = &ubifs_symlink_inode_operations;
190 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
191 err = 12;
192 goto out_invalid;
194 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
195 if (!ui->data) {
196 err = -ENOMEM;
197 goto out_ino;
199 memcpy(ui->data, ino->data, ui->data_len);
200 ((char *)ui->data)[ui->data_len] = '\0';
201 break;
202 case S_IFBLK:
203 case S_IFCHR:
205 dev_t rdev;
206 union ubifs_dev_desc *dev;
208 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
209 if (!ui->data) {
210 err = -ENOMEM;
211 goto out_ino;
214 dev = (union ubifs_dev_desc *)ino->data;
215 if (ui->data_len == sizeof(dev->new))
216 rdev = new_decode_dev(le32_to_cpu(dev->new));
217 else if (ui->data_len == sizeof(dev->huge))
218 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
219 else {
220 err = 13;
221 goto out_invalid;
223 memcpy(ui->data, ino->data, ui->data_len);
224 inode->i_op = &ubifs_file_inode_operations;
225 init_special_inode(inode, inode->i_mode, rdev);
226 break;
228 case S_IFSOCK:
229 case S_IFIFO:
230 inode->i_op = &ubifs_file_inode_operations;
231 init_special_inode(inode, inode->i_mode, 0);
232 if (ui->data_len != 0) {
233 err = 14;
234 goto out_invalid;
236 break;
237 default:
238 err = 15;
239 goto out_invalid;
242 kfree(ino);
243 ubifs_set_inode_flags(inode);
244 unlock_new_inode(inode);
245 return inode;
247 out_invalid:
248 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
249 dbg_dump_node(c, ino);
250 dbg_dump_inode(c, inode);
251 err = -EINVAL;
252 out_ino:
253 kfree(ino);
254 out:
255 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
256 iget_failed(inode);
257 return ERR_PTR(err);
260 static struct inode *ubifs_alloc_inode(struct super_block *sb)
262 struct ubifs_inode *ui;
264 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
265 if (!ui)
266 return NULL;
268 memset((void *)ui + sizeof(struct inode), 0,
269 sizeof(struct ubifs_inode) - sizeof(struct inode));
270 mutex_init(&ui->ui_mutex);
271 spin_lock_init(&ui->ui_lock);
272 return &ui->vfs_inode;
275 static void ubifs_i_callback(struct rcu_head *head)
277 struct inode *inode = container_of(head, struct inode, i_rcu);
278 struct ubifs_inode *ui = ubifs_inode(inode);
279 kmem_cache_free(ubifs_inode_slab, ui);
282 static void ubifs_destroy_inode(struct inode *inode)
284 struct ubifs_inode *ui = ubifs_inode(inode);
286 kfree(ui->data);
287 call_rcu(&inode->i_rcu, ubifs_i_callback);
291 * Note, Linux write-back code calls this without 'i_mutex'.
293 static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
295 int err = 0;
296 struct ubifs_info *c = inode->i_sb->s_fs_info;
297 struct ubifs_inode *ui = ubifs_inode(inode);
299 ubifs_assert(!ui->xattr);
300 if (is_bad_inode(inode))
301 return 0;
303 mutex_lock(&ui->ui_mutex);
305 * Due to races between write-back forced by budgeting
306 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
307 * have already been synchronized, do not do this again. This might
308 * also happen if it was synchronized in an VFS operation, e.g.
309 * 'ubifs_link()'.
311 if (!ui->dirty) {
312 mutex_unlock(&ui->ui_mutex);
313 return 0;
317 * As an optimization, do not write orphan inodes to the media just
318 * because this is not needed.
320 dbg_gen("inode %lu, mode %#x, nlink %u",
321 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
322 if (inode->i_nlink) {
323 err = ubifs_jnl_write_inode(c, inode);
324 if (err)
325 ubifs_err("can't write inode %lu, error %d",
326 inode->i_ino, err);
327 else
328 err = dbg_check_inode_size(c, inode, ui->ui_size);
331 ui->dirty = 0;
332 mutex_unlock(&ui->ui_mutex);
333 ubifs_release_dirty_inode_budget(c, ui);
334 return err;
337 static void ubifs_evict_inode(struct inode *inode)
339 int err;
340 struct ubifs_info *c = inode->i_sb->s_fs_info;
341 struct ubifs_inode *ui = ubifs_inode(inode);
343 if (ui->xattr)
345 * Extended attribute inode deletions are fully handled in
346 * 'ubifs_removexattr()'. These inodes are special and have
347 * limited usage, so there is nothing to do here.
349 goto out;
351 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
352 ubifs_assert(!atomic_read(&inode->i_count));
354 truncate_inode_pages(&inode->i_data, 0);
356 if (inode->i_nlink)
357 goto done;
359 if (is_bad_inode(inode))
360 goto out;
362 ui->ui_size = inode->i_size = 0;
363 err = ubifs_jnl_delete_inode(c, inode);
364 if (err)
366 * Worst case we have a lost orphan inode wasting space, so a
367 * simple error message is OK here.
369 ubifs_err("can't delete inode %lu, error %d",
370 inode->i_ino, err);
372 out:
373 if (ui->dirty)
374 ubifs_release_dirty_inode_budget(c, ui);
375 else {
376 /* We've deleted something - clean the "no space" flags */
377 c->bi.nospace = c->bi.nospace_rp = 0;
378 smp_wmb();
380 done:
381 end_writeback(inode);
384 static void ubifs_dirty_inode(struct inode *inode, int flags)
386 struct ubifs_inode *ui = ubifs_inode(inode);
388 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
389 if (!ui->dirty) {
390 ui->dirty = 1;
391 dbg_gen("inode %lu", inode->i_ino);
395 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
397 struct ubifs_info *c = dentry->d_sb->s_fs_info;
398 unsigned long long free;
399 __le32 *uuid = (__le32 *)c->uuid;
401 free = ubifs_get_free_space(c);
402 dbg_gen("free space %lld bytes (%lld blocks)",
403 free, free >> UBIFS_BLOCK_SHIFT);
405 buf->f_type = UBIFS_SUPER_MAGIC;
406 buf->f_bsize = UBIFS_BLOCK_SIZE;
407 buf->f_blocks = c->block_cnt;
408 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
409 if (free > c->report_rp_size)
410 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
411 else
412 buf->f_bavail = 0;
413 buf->f_files = 0;
414 buf->f_ffree = 0;
415 buf->f_namelen = UBIFS_MAX_NLEN;
416 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
417 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
418 ubifs_assert(buf->f_bfree <= c->block_cnt);
419 return 0;
422 static int ubifs_show_options(struct seq_file *s, struct dentry *root)
424 struct ubifs_info *c = root->d_sb->s_fs_info;
426 if (c->mount_opts.unmount_mode == 2)
427 seq_printf(s, ",fast_unmount");
428 else if (c->mount_opts.unmount_mode == 1)
429 seq_printf(s, ",norm_unmount");
431 if (c->mount_opts.bulk_read == 2)
432 seq_printf(s, ",bulk_read");
433 else if (c->mount_opts.bulk_read == 1)
434 seq_printf(s, ",no_bulk_read");
436 if (c->mount_opts.chk_data_crc == 2)
437 seq_printf(s, ",chk_data_crc");
438 else if (c->mount_opts.chk_data_crc == 1)
439 seq_printf(s, ",no_chk_data_crc");
441 if (c->mount_opts.override_compr) {
442 seq_printf(s, ",compr=%s",
443 ubifs_compr_name(c->mount_opts.compr_type));
446 return 0;
449 static int ubifs_sync_fs(struct super_block *sb, int wait)
451 int i, err;
452 struct ubifs_info *c = sb->s_fs_info;
455 * Zero @wait is just an advisory thing to help the file system shove
456 * lots of data into the queues, and there will be the second
457 * '->sync_fs()' call, with non-zero @wait.
459 if (!wait)
460 return 0;
463 * Synchronize write buffers, because 'ubifs_run_commit()' does not
464 * do this if it waits for an already running commit.
466 for (i = 0; i < c->jhead_cnt; i++) {
467 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
468 if (err)
469 return err;
473 * Strictly speaking, it is not necessary to commit the journal here,
474 * synchronizing write-buffers would be enough. But committing makes
475 * UBIFS free space predictions much more accurate, so we want to let
476 * the user be able to get more accurate results of 'statfs()' after
477 * they synchronize the file system.
479 err = ubifs_run_commit(c);
480 if (err)
481 return err;
483 return ubi_sync(c->vi.ubi_num);
487 * init_constants_early - initialize UBIFS constants.
488 * @c: UBIFS file-system description object
490 * This function initialize UBIFS constants which do not need the superblock to
491 * be read. It also checks that the UBI volume satisfies basic UBIFS
492 * requirements. Returns zero in case of success and a negative error code in
493 * case of failure.
495 static int init_constants_early(struct ubifs_info *c)
497 if (c->vi.corrupted) {
498 ubifs_warn("UBI volume is corrupted - read-only mode");
499 c->ro_media = 1;
502 if (c->di.ro_mode) {
503 ubifs_msg("read-only UBI device");
504 c->ro_media = 1;
507 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
508 ubifs_msg("static UBI volume - read-only mode");
509 c->ro_media = 1;
512 c->leb_cnt = c->vi.size;
513 c->leb_size = c->vi.usable_leb_size;
514 c->leb_start = c->di.leb_start;
515 c->half_leb_size = c->leb_size / 2;
516 c->min_io_size = c->di.min_io_size;
517 c->min_io_shift = fls(c->min_io_size) - 1;
518 c->max_write_size = c->di.max_write_size;
519 c->max_write_shift = fls(c->max_write_size) - 1;
521 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
522 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
523 c->leb_size, UBIFS_MIN_LEB_SZ);
524 return -EINVAL;
527 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
528 ubifs_err("too few LEBs (%d), min. is %d",
529 c->leb_cnt, UBIFS_MIN_LEB_CNT);
530 return -EINVAL;
533 if (!is_power_of_2(c->min_io_size)) {
534 ubifs_err("bad min. I/O size %d", c->min_io_size);
535 return -EINVAL;
539 * Maximum write size has to be greater or equivalent to min. I/O
540 * size, and be multiple of min. I/O size.
542 if (c->max_write_size < c->min_io_size ||
543 c->max_write_size % c->min_io_size ||
544 !is_power_of_2(c->max_write_size)) {
545 ubifs_err("bad write buffer size %d for %d min. I/O unit",
546 c->max_write_size, c->min_io_size);
547 return -EINVAL;
551 * UBIFS aligns all node to 8-byte boundary, so to make function in
552 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
553 * less than 8.
555 if (c->min_io_size < 8) {
556 c->min_io_size = 8;
557 c->min_io_shift = 3;
558 if (c->max_write_size < c->min_io_size) {
559 c->max_write_size = c->min_io_size;
560 c->max_write_shift = c->min_io_shift;
564 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
565 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
568 * Initialize node length ranges which are mostly needed for node
569 * length validation.
571 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
572 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
573 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
574 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
575 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
576 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
578 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
579 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
580 c->ranges[UBIFS_ORPH_NODE].min_len =
581 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
582 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
583 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
584 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
585 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
586 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
587 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
588 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
590 * Minimum indexing node size is amended later when superblock is
591 * read and the key length is known.
593 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
595 * Maximum indexing node size is amended later when superblock is
596 * read and the fanout is known.
598 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
601 * Initialize dead and dark LEB space watermarks. See gc.c for comments
602 * about these values.
604 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
605 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
608 * Calculate how many bytes would be wasted at the end of LEB if it was
609 * fully filled with data nodes of maximum size. This is used in
610 * calculations when reporting free space.
612 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
614 /* Buffer size for bulk-reads */
615 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
616 if (c->max_bu_buf_len > c->leb_size)
617 c->max_bu_buf_len = c->leb_size;
618 return 0;
622 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
623 * @c: UBIFS file-system description object
624 * @lnum: LEB the write-buffer was synchronized to
625 * @free: how many free bytes left in this LEB
626 * @pad: how many bytes were padded
628 * This is a callback function which is called by the I/O unit when the
629 * write-buffer is synchronized. We need this to correctly maintain space
630 * accounting in bud logical eraseblocks. This function returns zero in case of
631 * success and a negative error code in case of failure.
633 * This function actually belongs to the journal, but we keep it here because
634 * we want to keep it static.
636 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
638 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
642 * init_constants_sb - initialize UBIFS constants.
643 * @c: UBIFS file-system description object
645 * This is a helper function which initializes various UBIFS constants after
646 * the superblock has been read. It also checks various UBIFS parameters and
647 * makes sure they are all right. Returns zero in case of success and a
648 * negative error code in case of failure.
650 static int init_constants_sb(struct ubifs_info *c)
652 int tmp, err;
653 long long tmp64;
655 c->main_bytes = (long long)c->main_lebs * c->leb_size;
656 c->max_znode_sz = sizeof(struct ubifs_znode) +
657 c->fanout * sizeof(struct ubifs_zbranch);
659 tmp = ubifs_idx_node_sz(c, 1);
660 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
661 c->min_idx_node_sz = ALIGN(tmp, 8);
663 tmp = ubifs_idx_node_sz(c, c->fanout);
664 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
665 c->max_idx_node_sz = ALIGN(tmp, 8);
667 /* Make sure LEB size is large enough to fit full commit */
668 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
669 tmp = ALIGN(tmp, c->min_io_size);
670 if (tmp > c->leb_size) {
671 dbg_err("too small LEB size %d, at least %d needed",
672 c->leb_size, tmp);
673 return -EINVAL;
677 * Make sure that the log is large enough to fit reference nodes for
678 * all buds plus one reserved LEB.
680 tmp64 = c->max_bud_bytes + c->leb_size - 1;
681 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
682 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
683 tmp /= c->leb_size;
684 tmp += 1;
685 if (c->log_lebs < tmp) {
686 dbg_err("too small log %d LEBs, required min. %d LEBs",
687 c->log_lebs, tmp);
688 return -EINVAL;
692 * When budgeting we assume worst-case scenarios when the pages are not
693 * be compressed and direntries are of the maximum size.
695 * Note, data, which may be stored in inodes is budgeted separately, so
696 * it is not included into 'c->bi.inode_budget'.
698 c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
699 c->bi.inode_budget = UBIFS_INO_NODE_SZ;
700 c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
703 * When the amount of flash space used by buds becomes
704 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
705 * The writers are unblocked when the commit is finished. To avoid
706 * writers to be blocked UBIFS initiates background commit in advance,
707 * when number of bud bytes becomes above the limit defined below.
709 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
712 * Ensure minimum journal size. All the bytes in the journal heads are
713 * considered to be used, when calculating the current journal usage.
714 * Consequently, if the journal is too small, UBIFS will treat it as
715 * always full.
717 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
718 if (c->bg_bud_bytes < tmp64)
719 c->bg_bud_bytes = tmp64;
720 if (c->max_bud_bytes < tmp64 + c->leb_size)
721 c->max_bud_bytes = tmp64 + c->leb_size;
723 err = ubifs_calc_lpt_geom(c);
724 if (err)
725 return err;
727 /* Initialize effective LEB size used in budgeting calculations */
728 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
729 return 0;
733 * init_constants_master - initialize UBIFS constants.
734 * @c: UBIFS file-system description object
736 * This is a helper function which initializes various UBIFS constants after
737 * the master node has been read. It also checks various UBIFS parameters and
738 * makes sure they are all right.
740 static void init_constants_master(struct ubifs_info *c)
742 long long tmp64;
744 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
745 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
748 * Calculate total amount of FS blocks. This number is not used
749 * internally because it does not make much sense for UBIFS, but it is
750 * necessary to report something for the 'statfs()' call.
752 * Subtract the LEB reserved for GC, the LEB which is reserved for
753 * deletions, minimum LEBs for the index, and assume only one journal
754 * head is available.
756 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
757 tmp64 *= (long long)c->leb_size - c->leb_overhead;
758 tmp64 = ubifs_reported_space(c, tmp64);
759 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
763 * take_gc_lnum - reserve GC LEB.
764 * @c: UBIFS file-system description object
766 * This function ensures that the LEB reserved for garbage collection is marked
767 * as "taken" in lprops. We also have to set free space to LEB size and dirty
768 * space to zero, because lprops may contain out-of-date information if the
769 * file-system was un-mounted before it has been committed. This function
770 * returns zero in case of success and a negative error code in case of
771 * failure.
773 static int take_gc_lnum(struct ubifs_info *c)
775 int err;
777 if (c->gc_lnum == -1) {
778 ubifs_err("no LEB for GC");
779 return -EINVAL;
782 /* And we have to tell lprops that this LEB is taken */
783 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
784 LPROPS_TAKEN, 0, 0);
785 return err;
789 * alloc_wbufs - allocate write-buffers.
790 * @c: UBIFS file-system description object
792 * This helper function allocates and initializes UBIFS write-buffers. Returns
793 * zero in case of success and %-ENOMEM in case of failure.
795 static int alloc_wbufs(struct ubifs_info *c)
797 int i, err;
799 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
800 GFP_KERNEL);
801 if (!c->jheads)
802 return -ENOMEM;
804 /* Initialize journal heads */
805 for (i = 0; i < c->jhead_cnt; i++) {
806 INIT_LIST_HEAD(&c->jheads[i].buds_list);
807 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
808 if (err)
809 return err;
811 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
812 c->jheads[i].wbuf.jhead = i;
813 c->jheads[i].grouped = 1;
816 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
818 * Garbage Collector head likely contains long-term data and
819 * does not need to be synchronized by timer. Also GC head nodes are
820 * not grouped.
822 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
823 c->jheads[GCHD].wbuf.no_timer = 1;
824 c->jheads[GCHD].grouped = 0;
826 return 0;
830 * free_wbufs - free write-buffers.
831 * @c: UBIFS file-system description object
833 static void free_wbufs(struct ubifs_info *c)
835 int i;
837 if (c->jheads) {
838 for (i = 0; i < c->jhead_cnt; i++) {
839 kfree(c->jheads[i].wbuf.buf);
840 kfree(c->jheads[i].wbuf.inodes);
842 kfree(c->jheads);
843 c->jheads = NULL;
848 * free_orphans - free orphans.
849 * @c: UBIFS file-system description object
851 static void free_orphans(struct ubifs_info *c)
853 struct ubifs_orphan *orph;
855 while (c->orph_dnext) {
856 orph = c->orph_dnext;
857 c->orph_dnext = orph->dnext;
858 list_del(&orph->list);
859 kfree(orph);
862 while (!list_empty(&c->orph_list)) {
863 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
864 list_del(&orph->list);
865 kfree(orph);
866 dbg_err("orphan list not empty at unmount");
869 vfree(c->orph_buf);
870 c->orph_buf = NULL;
874 * free_buds - free per-bud objects.
875 * @c: UBIFS file-system description object
877 static void free_buds(struct ubifs_info *c)
879 struct rb_node *this = c->buds.rb_node;
880 struct ubifs_bud *bud;
882 while (this) {
883 if (this->rb_left)
884 this = this->rb_left;
885 else if (this->rb_right)
886 this = this->rb_right;
887 else {
888 bud = rb_entry(this, struct ubifs_bud, rb);
889 this = rb_parent(this);
890 if (this) {
891 if (this->rb_left == &bud->rb)
892 this->rb_left = NULL;
893 else
894 this->rb_right = NULL;
896 kfree(bud);
902 * check_volume_empty - check if the UBI volume is empty.
903 * @c: UBIFS file-system description object
905 * This function checks if the UBIFS volume is empty by looking if its LEBs are
906 * mapped or not. The result of checking is stored in the @c->empty variable.
907 * Returns zero in case of success and a negative error code in case of
908 * failure.
910 static int check_volume_empty(struct ubifs_info *c)
912 int lnum, err;
914 c->empty = 1;
915 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
916 err = ubifs_is_mapped(c, lnum);
917 if (unlikely(err < 0))
918 return err;
919 if (err == 1) {
920 c->empty = 0;
921 break;
924 cond_resched();
927 return 0;
931 * UBIFS mount options.
933 * Opt_fast_unmount: do not run a journal commit before un-mounting
934 * Opt_norm_unmount: run a journal commit before un-mounting
935 * Opt_bulk_read: enable bulk-reads
936 * Opt_no_bulk_read: disable bulk-reads
937 * Opt_chk_data_crc: check CRCs when reading data nodes
938 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
939 * Opt_override_compr: override default compressor
940 * Opt_err: just end of array marker
942 enum {
943 Opt_fast_unmount,
944 Opt_norm_unmount,
945 Opt_bulk_read,
946 Opt_no_bulk_read,
947 Opt_chk_data_crc,
948 Opt_no_chk_data_crc,
949 Opt_override_compr,
950 Opt_err,
953 static const match_table_t tokens = {
954 {Opt_fast_unmount, "fast_unmount"},
955 {Opt_norm_unmount, "norm_unmount"},
956 {Opt_bulk_read, "bulk_read"},
957 {Opt_no_bulk_read, "no_bulk_read"},
958 {Opt_chk_data_crc, "chk_data_crc"},
959 {Opt_no_chk_data_crc, "no_chk_data_crc"},
960 {Opt_override_compr, "compr=%s"},
961 {Opt_err, NULL},
965 * parse_standard_option - parse a standard mount option.
966 * @option: the option to parse
968 * Normally, standard mount options like "sync" are passed to file-systems as
969 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
970 * be present in the options string. This function tries to deal with this
971 * situation and parse standard options. Returns 0 if the option was not
972 * recognized, and the corresponding integer flag if it was.
974 * UBIFS is only interested in the "sync" option, so do not check for anything
975 * else.
977 static int parse_standard_option(const char *option)
979 ubifs_msg("parse %s", option);
980 if (!strcmp(option, "sync"))
981 return MS_SYNCHRONOUS;
982 return 0;
986 * ubifs_parse_options - parse mount parameters.
987 * @c: UBIFS file-system description object
988 * @options: parameters to parse
989 * @is_remount: non-zero if this is FS re-mount
991 * This function parses UBIFS mount options and returns zero in case success
992 * and a negative error code in case of failure.
994 static int ubifs_parse_options(struct ubifs_info *c, char *options,
995 int is_remount)
997 char *p;
998 substring_t args[MAX_OPT_ARGS];
1000 if (!options)
1001 return 0;
1003 while ((p = strsep(&options, ","))) {
1004 int token;
1006 if (!*p)
1007 continue;
1009 token = match_token(p, tokens, args);
1010 switch (token) {
1012 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
1013 * We accept them in order to be backward-compatible. But this
1014 * should be removed at some point.
1016 case Opt_fast_unmount:
1017 c->mount_opts.unmount_mode = 2;
1018 break;
1019 case Opt_norm_unmount:
1020 c->mount_opts.unmount_mode = 1;
1021 break;
1022 case Opt_bulk_read:
1023 c->mount_opts.bulk_read = 2;
1024 c->bulk_read = 1;
1025 break;
1026 case Opt_no_bulk_read:
1027 c->mount_opts.bulk_read = 1;
1028 c->bulk_read = 0;
1029 break;
1030 case Opt_chk_data_crc:
1031 c->mount_opts.chk_data_crc = 2;
1032 c->no_chk_data_crc = 0;
1033 break;
1034 case Opt_no_chk_data_crc:
1035 c->mount_opts.chk_data_crc = 1;
1036 c->no_chk_data_crc = 1;
1037 break;
1038 case Opt_override_compr:
1040 char *name = match_strdup(&args[0]);
1042 if (!name)
1043 return -ENOMEM;
1044 if (!strcmp(name, "none"))
1045 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1046 else if (!strcmp(name, "lzo"))
1047 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1048 else if (!strcmp(name, "zlib"))
1049 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1050 else {
1051 ubifs_err("unknown compressor \"%s\"", name);
1052 kfree(name);
1053 return -EINVAL;
1055 kfree(name);
1056 c->mount_opts.override_compr = 1;
1057 c->default_compr = c->mount_opts.compr_type;
1058 break;
1060 default:
1062 unsigned long flag;
1063 struct super_block *sb = c->vfs_sb;
1065 flag = parse_standard_option(p);
1066 if (!flag) {
1067 ubifs_err("unrecognized mount option \"%s\" "
1068 "or missing value", p);
1069 return -EINVAL;
1071 sb->s_flags |= flag;
1072 break;
1077 return 0;
1081 * destroy_journal - destroy journal data structures.
1082 * @c: UBIFS file-system description object
1084 * This function destroys journal data structures including those that may have
1085 * been created by recovery functions.
1087 static void destroy_journal(struct ubifs_info *c)
1089 while (!list_empty(&c->unclean_leb_list)) {
1090 struct ubifs_unclean_leb *ucleb;
1092 ucleb = list_entry(c->unclean_leb_list.next,
1093 struct ubifs_unclean_leb, list);
1094 list_del(&ucleb->list);
1095 kfree(ucleb);
1097 while (!list_empty(&c->old_buds)) {
1098 struct ubifs_bud *bud;
1100 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1101 list_del(&bud->list);
1102 kfree(bud);
1104 ubifs_destroy_idx_gc(c);
1105 ubifs_destroy_size_tree(c);
1106 ubifs_tnc_close(c);
1107 free_buds(c);
1111 * bu_init - initialize bulk-read information.
1112 * @c: UBIFS file-system description object
1114 static void bu_init(struct ubifs_info *c)
1116 ubifs_assert(c->bulk_read == 1);
1118 if (c->bu.buf)
1119 return; /* Already initialized */
1121 again:
1122 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1123 if (!c->bu.buf) {
1124 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1125 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1126 goto again;
1129 /* Just disable bulk-read */
1130 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1131 "disabling it", c->max_bu_buf_len);
1132 c->mount_opts.bulk_read = 1;
1133 c->bulk_read = 0;
1134 return;
1139 * check_free_space - check if there is enough free space to mount.
1140 * @c: UBIFS file-system description object
1142 * This function makes sure UBIFS has enough free space to be mounted in
1143 * read/write mode. UBIFS must always have some free space to allow deletions.
1145 static int check_free_space(struct ubifs_info *c)
1147 ubifs_assert(c->dark_wm > 0);
1148 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1149 ubifs_err("insufficient free space to mount in R/W mode");
1150 dbg_dump_budg(c, &c->bi);
1151 dbg_dump_lprops(c);
1152 return -ENOSPC;
1154 return 0;
1158 * mount_ubifs - mount UBIFS file-system.
1159 * @c: UBIFS file-system description object
1161 * This function mounts UBIFS file system. Returns zero in case of success and
1162 * a negative error code in case of failure.
1164 * Note, the function does not de-allocate resources it it fails half way
1165 * through, and the caller has to do this instead.
1167 static int mount_ubifs(struct ubifs_info *c)
1169 int err;
1170 long long x;
1171 size_t sz;
1173 c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1174 err = init_constants_early(c);
1175 if (err)
1176 return err;
1178 err = ubifs_debugging_init(c);
1179 if (err)
1180 return err;
1182 err = check_volume_empty(c);
1183 if (err)
1184 goto out_free;
1186 if (c->empty && (c->ro_mount || c->ro_media)) {
1188 * This UBI volume is empty, and read-only, or the file system
1189 * is mounted read-only - we cannot format it.
1191 ubifs_err("can't format empty UBI volume: read-only %s",
1192 c->ro_media ? "UBI volume" : "mount");
1193 err = -EROFS;
1194 goto out_free;
1197 if (c->ro_media && !c->ro_mount) {
1198 ubifs_err("cannot mount read-write - read-only media");
1199 err = -EROFS;
1200 goto out_free;
1204 * The requirement for the buffer is that it should fit indexing B-tree
1205 * height amount of integers. We assume the height if the TNC tree will
1206 * never exceed 64.
1208 err = -ENOMEM;
1209 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1210 if (!c->bottom_up_buf)
1211 goto out_free;
1213 c->sbuf = vmalloc(c->leb_size);
1214 if (!c->sbuf)
1215 goto out_free;
1217 if (!c->ro_mount) {
1218 c->ileb_buf = vmalloc(c->leb_size);
1219 if (!c->ileb_buf)
1220 goto out_free;
1223 if (c->bulk_read == 1)
1224 bu_init(c);
1226 if (!c->ro_mount) {
1227 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
1228 GFP_KERNEL);
1229 if (!c->write_reserve_buf)
1230 goto out_free;
1233 c->mounting = 1;
1235 err = ubifs_read_superblock(c);
1236 if (err)
1237 goto out_free;
1240 * Make sure the compressor which is set as default in the superblock
1241 * or overridden by mount options is actually compiled in.
1243 if (!ubifs_compr_present(c->default_compr)) {
1244 ubifs_err("'compressor \"%s\" is not compiled in",
1245 ubifs_compr_name(c->default_compr));
1246 err = -ENOTSUPP;
1247 goto out_free;
1250 err = init_constants_sb(c);
1251 if (err)
1252 goto out_free;
1254 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1255 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1256 c->cbuf = kmalloc(sz, GFP_NOFS);
1257 if (!c->cbuf) {
1258 err = -ENOMEM;
1259 goto out_free;
1262 err = alloc_wbufs(c);
1263 if (err)
1264 goto out_cbuf;
1266 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1267 if (!c->ro_mount) {
1268 /* Create background thread */
1269 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1270 if (IS_ERR(c->bgt)) {
1271 err = PTR_ERR(c->bgt);
1272 c->bgt = NULL;
1273 ubifs_err("cannot spawn \"%s\", error %d",
1274 c->bgt_name, err);
1275 goto out_wbufs;
1277 wake_up_process(c->bgt);
1280 err = ubifs_read_master(c);
1281 if (err)
1282 goto out_master;
1284 init_constants_master(c);
1286 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1287 ubifs_msg("recovery needed");
1288 c->need_recovery = 1;
1291 if (c->need_recovery && !c->ro_mount) {
1292 err = ubifs_recover_inl_heads(c, c->sbuf);
1293 if (err)
1294 goto out_master;
1297 err = ubifs_lpt_init(c, 1, !c->ro_mount);
1298 if (err)
1299 goto out_master;
1301 if (!c->ro_mount && c->space_fixup) {
1302 err = ubifs_fixup_free_space(c);
1303 if (err)
1304 goto out_master;
1307 if (!c->ro_mount) {
1309 * Set the "dirty" flag so that if we reboot uncleanly we
1310 * will notice this immediately on the next mount.
1312 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1313 err = ubifs_write_master(c);
1314 if (err)
1315 goto out_lpt;
1318 err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1319 if (err)
1320 goto out_lpt;
1322 err = ubifs_replay_journal(c);
1323 if (err)
1324 goto out_journal;
1326 /* Calculate 'min_idx_lebs' after journal replay */
1327 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1329 err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1330 if (err)
1331 goto out_orphans;
1333 if (!c->ro_mount) {
1334 int lnum;
1336 err = check_free_space(c);
1337 if (err)
1338 goto out_orphans;
1340 /* Check for enough log space */
1341 lnum = c->lhead_lnum + 1;
1342 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1343 lnum = UBIFS_LOG_LNUM;
1344 if (lnum == c->ltail_lnum) {
1345 err = ubifs_consolidate_log(c);
1346 if (err)
1347 goto out_orphans;
1350 if (c->need_recovery) {
1351 err = ubifs_recover_size(c);
1352 if (err)
1353 goto out_orphans;
1354 err = ubifs_rcvry_gc_commit(c);
1355 if (err)
1356 goto out_orphans;
1357 } else {
1358 err = take_gc_lnum(c);
1359 if (err)
1360 goto out_orphans;
1363 * GC LEB may contain garbage if there was an unclean
1364 * reboot, and it should be un-mapped.
1366 err = ubifs_leb_unmap(c, c->gc_lnum);
1367 if (err)
1368 goto out_orphans;
1371 err = dbg_check_lprops(c);
1372 if (err)
1373 goto out_orphans;
1374 } else if (c->need_recovery) {
1375 err = ubifs_recover_size(c);
1376 if (err)
1377 goto out_orphans;
1378 } else {
1380 * Even if we mount read-only, we have to set space in GC LEB
1381 * to proper value because this affects UBIFS free space
1382 * reporting. We do not want to have a situation when
1383 * re-mounting from R/O to R/W changes amount of free space.
1385 err = take_gc_lnum(c);
1386 if (err)
1387 goto out_orphans;
1390 spin_lock(&ubifs_infos_lock);
1391 list_add_tail(&c->infos_list, &ubifs_infos);
1392 spin_unlock(&ubifs_infos_lock);
1394 if (c->need_recovery) {
1395 if (c->ro_mount)
1396 ubifs_msg("recovery deferred");
1397 else {
1398 c->need_recovery = 0;
1399 ubifs_msg("recovery completed");
1401 * GC LEB has to be empty and taken at this point. But
1402 * the journal head LEBs may also be accounted as
1403 * "empty taken" if they are empty.
1405 ubifs_assert(c->lst.taken_empty_lebs > 0);
1407 } else
1408 ubifs_assert(c->lst.taken_empty_lebs > 0);
1410 err = dbg_check_filesystem(c);
1411 if (err)
1412 goto out_infos;
1414 err = dbg_debugfs_init_fs(c);
1415 if (err)
1416 goto out_infos;
1418 c->mounting = 0;
1420 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1421 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1422 if (c->ro_mount)
1423 ubifs_msg("mounted read-only");
1424 x = (long long)c->main_lebs * c->leb_size;
1425 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1426 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1427 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1428 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1429 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1430 ubifs_msg("media format: w%d/r%d (latest is w%d/r%d)",
1431 c->fmt_version, c->ro_compat_version,
1432 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1433 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1434 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1435 c->report_rp_size, c->report_rp_size >> 10);
1437 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1438 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1439 dbg_msg("max. write size: %d bytes", c->max_write_size);
1440 dbg_msg("LEB size: %d bytes (%d KiB)",
1441 c->leb_size, c->leb_size >> 10);
1442 dbg_msg("data journal heads: %d",
1443 c->jhead_cnt - NONDATA_JHEADS_CNT);
1444 dbg_msg("UUID: %pUB", c->uuid);
1445 dbg_msg("big_lpt %d", c->big_lpt);
1446 dbg_msg("log LEBs: %d (%d - %d)",
1447 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1448 dbg_msg("LPT area LEBs: %d (%d - %d)",
1449 c->lpt_lebs, c->lpt_first, c->lpt_last);
1450 dbg_msg("orphan area LEBs: %d (%d - %d)",
1451 c->orph_lebs, c->orph_first, c->orph_last);
1452 dbg_msg("main area LEBs: %d (%d - %d)",
1453 c->main_lebs, c->main_first, c->leb_cnt - 1);
1454 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1455 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1456 c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1457 c->bi.old_idx_sz >> 20);
1458 dbg_msg("key hash type: %d", c->key_hash_type);
1459 dbg_msg("tree fanout: %d", c->fanout);
1460 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1461 dbg_msg("first main LEB: %d", c->main_first);
1462 dbg_msg("max. znode size %d", c->max_znode_sz);
1463 dbg_msg("max. index node size %d", c->max_idx_node_sz);
1464 dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
1465 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1466 dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
1467 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1468 dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
1469 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1470 dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu, idx %d",
1471 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1472 UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1473 dbg_msg("dead watermark: %d", c->dead_wm);
1474 dbg_msg("dark watermark: %d", c->dark_wm);
1475 dbg_msg("LEB overhead: %d", c->leb_overhead);
1476 x = (long long)c->main_lebs * c->dark_wm;
1477 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1478 x, x >> 10, x >> 20);
1479 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1480 c->max_bud_bytes, c->max_bud_bytes >> 10,
1481 c->max_bud_bytes >> 20);
1482 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1483 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1484 c->bg_bud_bytes >> 20);
1485 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1486 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1487 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1488 dbg_msg("commit number: %llu", c->cmt_no);
1490 return 0;
1492 out_infos:
1493 spin_lock(&ubifs_infos_lock);
1494 list_del(&c->infos_list);
1495 spin_unlock(&ubifs_infos_lock);
1496 out_orphans:
1497 free_orphans(c);
1498 out_journal:
1499 destroy_journal(c);
1500 out_lpt:
1501 ubifs_lpt_free(c, 0);
1502 out_master:
1503 kfree(c->mst_node);
1504 kfree(c->rcvrd_mst_node);
1505 if (c->bgt)
1506 kthread_stop(c->bgt);
1507 out_wbufs:
1508 free_wbufs(c);
1509 out_cbuf:
1510 kfree(c->cbuf);
1511 out_free:
1512 kfree(c->write_reserve_buf);
1513 kfree(c->bu.buf);
1514 vfree(c->ileb_buf);
1515 vfree(c->sbuf);
1516 kfree(c->bottom_up_buf);
1517 ubifs_debugging_exit(c);
1518 return err;
1522 * ubifs_umount - un-mount UBIFS file-system.
1523 * @c: UBIFS file-system description object
1525 * Note, this function is called to free allocated resourced when un-mounting,
1526 * as well as free resources when an error occurred while we were half way
1527 * through mounting (error path cleanup function). So it has to make sure the
1528 * resource was actually allocated before freeing it.
1530 static void ubifs_umount(struct ubifs_info *c)
1532 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1533 c->vi.vol_id);
1535 dbg_debugfs_exit_fs(c);
1536 spin_lock(&ubifs_infos_lock);
1537 list_del(&c->infos_list);
1538 spin_unlock(&ubifs_infos_lock);
1540 if (c->bgt)
1541 kthread_stop(c->bgt);
1543 destroy_journal(c);
1544 free_wbufs(c);
1545 free_orphans(c);
1546 ubifs_lpt_free(c, 0);
1548 kfree(c->cbuf);
1549 kfree(c->rcvrd_mst_node);
1550 kfree(c->mst_node);
1551 kfree(c->write_reserve_buf);
1552 kfree(c->bu.buf);
1553 vfree(c->ileb_buf);
1554 vfree(c->sbuf);
1555 kfree(c->bottom_up_buf);
1556 ubifs_debugging_exit(c);
1560 * ubifs_remount_rw - re-mount in read-write mode.
1561 * @c: UBIFS file-system description object
1563 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1564 * mode. This function allocates the needed resources and re-mounts UBIFS in
1565 * read-write mode.
1567 static int ubifs_remount_rw(struct ubifs_info *c)
1569 int err, lnum;
1571 if (c->rw_incompat) {
1572 ubifs_err("the file-system is not R/W-compatible");
1573 ubifs_msg("on-flash format version is w%d/r%d, but software "
1574 "only supports up to version w%d/r%d", c->fmt_version,
1575 c->ro_compat_version, UBIFS_FORMAT_VERSION,
1576 UBIFS_RO_COMPAT_VERSION);
1577 return -EROFS;
1580 mutex_lock(&c->umount_mutex);
1581 dbg_save_space_info(c);
1582 c->remounting_rw = 1;
1583 c->ro_mount = 0;
1585 err = check_free_space(c);
1586 if (err)
1587 goto out;
1589 if (c->old_leb_cnt != c->leb_cnt) {
1590 struct ubifs_sb_node *sup;
1592 sup = ubifs_read_sb_node(c);
1593 if (IS_ERR(sup)) {
1594 err = PTR_ERR(sup);
1595 goto out;
1597 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1598 err = ubifs_write_sb_node(c, sup);
1599 kfree(sup);
1600 if (err)
1601 goto out;
1604 if (c->need_recovery) {
1605 ubifs_msg("completing deferred recovery");
1606 err = ubifs_write_rcvrd_mst_node(c);
1607 if (err)
1608 goto out;
1609 err = ubifs_recover_size(c);
1610 if (err)
1611 goto out;
1612 err = ubifs_clean_lebs(c, c->sbuf);
1613 if (err)
1614 goto out;
1615 err = ubifs_recover_inl_heads(c, c->sbuf);
1616 if (err)
1617 goto out;
1618 } else {
1619 /* A readonly mount is not allowed to have orphans */
1620 ubifs_assert(c->tot_orphans == 0);
1621 err = ubifs_clear_orphans(c);
1622 if (err)
1623 goto out;
1626 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1627 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1628 err = ubifs_write_master(c);
1629 if (err)
1630 goto out;
1633 c->ileb_buf = vmalloc(c->leb_size);
1634 if (!c->ileb_buf) {
1635 err = -ENOMEM;
1636 goto out;
1639 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
1640 if (!c->write_reserve_buf)
1641 goto out;
1643 err = ubifs_lpt_init(c, 0, 1);
1644 if (err)
1645 goto out;
1647 /* Create background thread */
1648 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1649 if (IS_ERR(c->bgt)) {
1650 err = PTR_ERR(c->bgt);
1651 c->bgt = NULL;
1652 ubifs_err("cannot spawn \"%s\", error %d",
1653 c->bgt_name, err);
1654 goto out;
1656 wake_up_process(c->bgt);
1658 c->orph_buf = vmalloc(c->leb_size);
1659 if (!c->orph_buf) {
1660 err = -ENOMEM;
1661 goto out;
1664 /* Check for enough log space */
1665 lnum = c->lhead_lnum + 1;
1666 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1667 lnum = UBIFS_LOG_LNUM;
1668 if (lnum == c->ltail_lnum) {
1669 err = ubifs_consolidate_log(c);
1670 if (err)
1671 goto out;
1674 if (c->need_recovery)
1675 err = ubifs_rcvry_gc_commit(c);
1676 else
1677 err = ubifs_leb_unmap(c, c->gc_lnum);
1678 if (err)
1679 goto out;
1681 dbg_gen("re-mounted read-write");
1682 c->remounting_rw = 0;
1684 if (c->need_recovery) {
1685 c->need_recovery = 0;
1686 ubifs_msg("deferred recovery completed");
1687 } else {
1689 * Do not run the debugging space check if the were doing
1690 * recovery, because when we saved the information we had the
1691 * file-system in a state where the TNC and lprops has been
1692 * modified in memory, but all the I/O operations (including a
1693 * commit) were deferred. So the file-system was in
1694 * "non-committed" state. Now the file-system is in committed
1695 * state, and of course the amount of free space will change
1696 * because, for example, the old index size was imprecise.
1698 err = dbg_check_space_info(c);
1701 if (c->space_fixup) {
1702 err = ubifs_fixup_free_space(c);
1703 if (err)
1704 goto out;
1707 mutex_unlock(&c->umount_mutex);
1708 return err;
1710 out:
1711 c->ro_mount = 1;
1712 vfree(c->orph_buf);
1713 c->orph_buf = NULL;
1714 if (c->bgt) {
1715 kthread_stop(c->bgt);
1716 c->bgt = NULL;
1718 free_wbufs(c);
1719 kfree(c->write_reserve_buf);
1720 c->write_reserve_buf = NULL;
1721 vfree(c->ileb_buf);
1722 c->ileb_buf = NULL;
1723 ubifs_lpt_free(c, 1);
1724 c->remounting_rw = 0;
1725 mutex_unlock(&c->umount_mutex);
1726 return err;
1730 * ubifs_remount_ro - re-mount in read-only mode.
1731 * @c: UBIFS file-system description object
1733 * We assume VFS has stopped writing. Possibly the background thread could be
1734 * running a commit, however kthread_stop will wait in that case.
1736 static void ubifs_remount_ro(struct ubifs_info *c)
1738 int i, err;
1740 ubifs_assert(!c->need_recovery);
1741 ubifs_assert(!c->ro_mount);
1743 mutex_lock(&c->umount_mutex);
1744 if (c->bgt) {
1745 kthread_stop(c->bgt);
1746 c->bgt = NULL;
1749 dbg_save_space_info(c);
1751 for (i = 0; i < c->jhead_cnt; i++)
1752 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1754 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1755 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1756 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1757 err = ubifs_write_master(c);
1758 if (err)
1759 ubifs_ro_mode(c, err);
1761 vfree(c->orph_buf);
1762 c->orph_buf = NULL;
1763 kfree(c->write_reserve_buf);
1764 c->write_reserve_buf = NULL;
1765 vfree(c->ileb_buf);
1766 c->ileb_buf = NULL;
1767 ubifs_lpt_free(c, 1);
1768 c->ro_mount = 1;
1769 err = dbg_check_space_info(c);
1770 if (err)
1771 ubifs_ro_mode(c, err);
1772 mutex_unlock(&c->umount_mutex);
1775 static void ubifs_put_super(struct super_block *sb)
1777 int i;
1778 struct ubifs_info *c = sb->s_fs_info;
1780 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1781 c->vi.vol_id);
1784 * The following asserts are only valid if there has not been a failure
1785 * of the media. For example, there will be dirty inodes if we failed
1786 * to write them back because of I/O errors.
1788 if (!c->ro_error) {
1789 ubifs_assert(c->bi.idx_growth == 0);
1790 ubifs_assert(c->bi.dd_growth == 0);
1791 ubifs_assert(c->bi.data_growth == 0);
1795 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1796 * and file system un-mount. Namely, it prevents the shrinker from
1797 * picking this superblock for shrinking - it will be just skipped if
1798 * the mutex is locked.
1800 mutex_lock(&c->umount_mutex);
1801 if (!c->ro_mount) {
1803 * First of all kill the background thread to make sure it does
1804 * not interfere with un-mounting and freeing resources.
1806 if (c->bgt) {
1807 kthread_stop(c->bgt);
1808 c->bgt = NULL;
1812 * On fatal errors c->ro_error is set to 1, in which case we do
1813 * not write the master node.
1815 if (!c->ro_error) {
1816 int err;
1818 /* Synchronize write-buffers */
1819 for (i = 0; i < c->jhead_cnt; i++)
1820 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1823 * We are being cleanly unmounted which means the
1824 * orphans were killed - indicate this in the master
1825 * node. Also save the reserved GC LEB number.
1827 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1828 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1829 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1830 err = ubifs_write_master(c);
1831 if (err)
1833 * Recovery will attempt to fix the master area
1834 * next mount, so we just print a message and
1835 * continue to unmount normally.
1837 ubifs_err("failed to write master node, "
1838 "error %d", err);
1839 } else {
1840 for (i = 0; i < c->jhead_cnt; i++)
1841 /* Make sure write-buffer timers are canceled */
1842 hrtimer_cancel(&c->jheads[i].wbuf.timer);
1846 ubifs_umount(c);
1847 bdi_destroy(&c->bdi);
1848 ubi_close_volume(c->ubi);
1849 mutex_unlock(&c->umount_mutex);
1852 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1854 int err;
1855 struct ubifs_info *c = sb->s_fs_info;
1857 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1859 err = ubifs_parse_options(c, data, 1);
1860 if (err) {
1861 ubifs_err("invalid or unknown remount parameter");
1862 return err;
1865 if (c->ro_mount && !(*flags & MS_RDONLY)) {
1866 if (c->ro_error) {
1867 ubifs_msg("cannot re-mount R/W due to prior errors");
1868 return -EROFS;
1870 if (c->ro_media) {
1871 ubifs_msg("cannot re-mount R/W - UBI volume is R/O");
1872 return -EROFS;
1874 err = ubifs_remount_rw(c);
1875 if (err)
1876 return err;
1877 } else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1878 if (c->ro_error) {
1879 ubifs_msg("cannot re-mount R/O due to prior errors");
1880 return -EROFS;
1882 ubifs_remount_ro(c);
1885 if (c->bulk_read == 1)
1886 bu_init(c);
1887 else {
1888 dbg_gen("disable bulk-read");
1889 kfree(c->bu.buf);
1890 c->bu.buf = NULL;
1893 ubifs_assert(c->lst.taken_empty_lebs > 0);
1894 return 0;
1897 const struct super_operations ubifs_super_operations = {
1898 .alloc_inode = ubifs_alloc_inode,
1899 .destroy_inode = ubifs_destroy_inode,
1900 .put_super = ubifs_put_super,
1901 .write_inode = ubifs_write_inode,
1902 .evict_inode = ubifs_evict_inode,
1903 .statfs = ubifs_statfs,
1904 .dirty_inode = ubifs_dirty_inode,
1905 .remount_fs = ubifs_remount_fs,
1906 .show_options = ubifs_show_options,
1907 .sync_fs = ubifs_sync_fs,
1911 * open_ubi - parse UBI device name string and open the UBI device.
1912 * @name: UBI volume name
1913 * @mode: UBI volume open mode
1915 * The primary method of mounting UBIFS is by specifying the UBI volume
1916 * character device node path. However, UBIFS may also be mounted withoug any
1917 * character device node using one of the following methods:
1919 * o ubiX_Y - mount UBI device number X, volume Y;
1920 * o ubiY - mount UBI device number 0, volume Y;
1921 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1922 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1924 * Alternative '!' separator may be used instead of ':' (because some shells
1925 * like busybox may interpret ':' as an NFS host name separator). This function
1926 * returns UBI volume description object in case of success and a negative
1927 * error code in case of failure.
1929 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1931 struct ubi_volume_desc *ubi;
1932 int dev, vol;
1933 char *endptr;
1935 /* First, try to open using the device node path method */
1936 ubi = ubi_open_volume_path(name, mode);
1937 if (!IS_ERR(ubi))
1938 return ubi;
1940 /* Try the "nodev" method */
1941 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1942 return ERR_PTR(-EINVAL);
1944 /* ubi:NAME method */
1945 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1946 return ubi_open_volume_nm(0, name + 4, mode);
1948 if (!isdigit(name[3]))
1949 return ERR_PTR(-EINVAL);
1951 dev = simple_strtoul(name + 3, &endptr, 0);
1953 /* ubiY method */
1954 if (*endptr == '\0')
1955 return ubi_open_volume(0, dev, mode);
1957 /* ubiX_Y method */
1958 if (*endptr == '_' && isdigit(endptr[1])) {
1959 vol = simple_strtoul(endptr + 1, &endptr, 0);
1960 if (*endptr != '\0')
1961 return ERR_PTR(-EINVAL);
1962 return ubi_open_volume(dev, vol, mode);
1965 /* ubiX:NAME method */
1966 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1967 return ubi_open_volume_nm(dev, ++endptr, mode);
1969 return ERR_PTR(-EINVAL);
1972 static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1974 struct ubifs_info *c;
1976 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1977 if (c) {
1978 spin_lock_init(&c->cnt_lock);
1979 spin_lock_init(&c->cs_lock);
1980 spin_lock_init(&c->buds_lock);
1981 spin_lock_init(&c->space_lock);
1982 spin_lock_init(&c->orphan_lock);
1983 init_rwsem(&c->commit_sem);
1984 mutex_init(&c->lp_mutex);
1985 mutex_init(&c->tnc_mutex);
1986 mutex_init(&c->log_mutex);
1987 mutex_init(&c->mst_mutex);
1988 mutex_init(&c->umount_mutex);
1989 mutex_init(&c->bu_mutex);
1990 mutex_init(&c->write_reserve_mutex);
1991 init_waitqueue_head(&c->cmt_wq);
1992 c->buds = RB_ROOT;
1993 c->old_idx = RB_ROOT;
1994 c->size_tree = RB_ROOT;
1995 c->orph_tree = RB_ROOT;
1996 INIT_LIST_HEAD(&c->infos_list);
1997 INIT_LIST_HEAD(&c->idx_gc);
1998 INIT_LIST_HEAD(&c->replay_list);
1999 INIT_LIST_HEAD(&c->replay_buds);
2000 INIT_LIST_HEAD(&c->uncat_list);
2001 INIT_LIST_HEAD(&c->empty_list);
2002 INIT_LIST_HEAD(&c->freeable_list);
2003 INIT_LIST_HEAD(&c->frdi_idx_list);
2004 INIT_LIST_HEAD(&c->unclean_leb_list);
2005 INIT_LIST_HEAD(&c->old_buds);
2006 INIT_LIST_HEAD(&c->orph_list);
2007 INIT_LIST_HEAD(&c->orph_new);
2008 c->no_chk_data_crc = 1;
2010 c->highest_inum = UBIFS_FIRST_INO;
2011 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
2013 ubi_get_volume_info(ubi, &c->vi);
2014 ubi_get_device_info(c->vi.ubi_num, &c->di);
2016 return c;
2019 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
2021 struct ubifs_info *c = sb->s_fs_info;
2022 struct inode *root;
2023 int err;
2025 c->vfs_sb = sb;
2026 /* Re-open the UBI device in read-write mode */
2027 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2028 if (IS_ERR(c->ubi)) {
2029 err = PTR_ERR(c->ubi);
2030 goto out;
2034 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2035 * UBIFS, I/O is not deferred, it is done immediately in readpage,
2036 * which means the user would have to wait not just for their own I/O
2037 * but the read-ahead I/O as well i.e. completely pointless.
2039 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
2041 c->bdi.name = "ubifs",
2042 c->bdi.capabilities = BDI_CAP_MAP_COPY;
2043 err = bdi_init(&c->bdi);
2044 if (err)
2045 goto out_close;
2046 err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
2047 c->vi.ubi_num, c->vi.vol_id);
2048 if (err)
2049 goto out_bdi;
2051 err = ubifs_parse_options(c, data, 0);
2052 if (err)
2053 goto out_bdi;
2055 sb->s_bdi = &c->bdi;
2056 sb->s_fs_info = c;
2057 sb->s_magic = UBIFS_SUPER_MAGIC;
2058 sb->s_blocksize = UBIFS_BLOCK_SIZE;
2059 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2060 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2061 if (c->max_inode_sz > MAX_LFS_FILESIZE)
2062 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2063 sb->s_op = &ubifs_super_operations;
2065 mutex_lock(&c->umount_mutex);
2066 err = mount_ubifs(c);
2067 if (err) {
2068 ubifs_assert(err < 0);
2069 goto out_unlock;
2072 /* Read the root inode */
2073 root = ubifs_iget(sb, UBIFS_ROOT_INO);
2074 if (IS_ERR(root)) {
2075 err = PTR_ERR(root);
2076 goto out_umount;
2079 sb->s_root = d_alloc_root(root);
2080 if (!sb->s_root)
2081 goto out_iput;
2083 mutex_unlock(&c->umount_mutex);
2084 return 0;
2086 out_iput:
2087 iput(root);
2088 out_umount:
2089 ubifs_umount(c);
2090 out_unlock:
2091 mutex_unlock(&c->umount_mutex);
2092 out_bdi:
2093 bdi_destroy(&c->bdi);
2094 out_close:
2095 ubi_close_volume(c->ubi);
2096 out:
2097 return err;
2100 static int sb_test(struct super_block *sb, void *data)
2102 struct ubifs_info *c1 = data;
2103 struct ubifs_info *c = sb->s_fs_info;
2105 return c->vi.cdev == c1->vi.cdev;
2108 static int sb_set(struct super_block *sb, void *data)
2110 sb->s_fs_info = data;
2111 return set_anon_super(sb, NULL);
2114 static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2115 const char *name, void *data)
2117 struct ubi_volume_desc *ubi;
2118 struct ubifs_info *c;
2119 struct super_block *sb;
2120 int err;
2122 dbg_gen("name %s, flags %#x", name, flags);
2125 * Get UBI device number and volume ID. Mount it read-only so far
2126 * because this might be a new mount point, and UBI allows only one
2127 * read-write user at a time.
2129 ubi = open_ubi(name, UBI_READONLY);
2130 if (IS_ERR(ubi)) {
2131 dbg_err("cannot open \"%s\", error %d",
2132 name, (int)PTR_ERR(ubi));
2133 return ERR_CAST(ubi);
2136 c = alloc_ubifs_info(ubi);
2137 if (!c) {
2138 err = -ENOMEM;
2139 goto out_close;
2142 dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2144 sb = sget(fs_type, sb_test, sb_set, c);
2145 if (IS_ERR(sb)) {
2146 err = PTR_ERR(sb);
2147 kfree(c);
2148 goto out_close;
2151 if (sb->s_root) {
2152 struct ubifs_info *c1 = sb->s_fs_info;
2153 kfree(c);
2154 /* A new mount point for already mounted UBIFS */
2155 dbg_gen("this ubi volume is already mounted");
2156 if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2157 err = -EBUSY;
2158 goto out_deact;
2160 } else {
2161 sb->s_flags = flags;
2162 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2163 if (err)
2164 goto out_deact;
2165 /* We do not support atime */
2166 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2169 /* 'fill_super()' opens ubi again so we must close it here */
2170 ubi_close_volume(ubi);
2172 return dget(sb->s_root);
2174 out_deact:
2175 deactivate_locked_super(sb);
2176 out_close:
2177 ubi_close_volume(ubi);
2178 return ERR_PTR(err);
2181 static void kill_ubifs_super(struct super_block *s)
2183 struct ubifs_info *c = s->s_fs_info;
2184 kill_anon_super(s);
2185 kfree(c);
2188 static struct file_system_type ubifs_fs_type = {
2189 .name = "ubifs",
2190 .owner = THIS_MODULE,
2191 .mount = ubifs_mount,
2192 .kill_sb = kill_ubifs_super,
2196 * Inode slab cache constructor.
2198 static void inode_slab_ctor(void *obj)
2200 struct ubifs_inode *ui = obj;
2201 inode_init_once(&ui->vfs_inode);
2204 static int __init ubifs_init(void)
2206 int err;
2208 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2210 /* Make sure node sizes are 8-byte aligned */
2211 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2212 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2213 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2214 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2215 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2216 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2217 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2218 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2219 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2220 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2221 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2223 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2224 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2225 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2226 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2227 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2228 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2230 /* Check min. node size */
2231 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2232 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2233 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2234 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2236 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2237 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2238 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2239 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2241 /* Defined node sizes */
2242 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2243 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2244 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2245 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2248 * We use 2 bit wide bit-fields to store compression type, which should
2249 * be amended if more compressors are added. The bit-fields are:
2250 * @compr_type in 'struct ubifs_inode', @default_compr in
2251 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2253 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2256 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2257 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2259 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2260 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2261 " at least 4096 bytes",
2262 (unsigned int)PAGE_CACHE_SIZE);
2263 return -EINVAL;
2266 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2267 sizeof(struct ubifs_inode), 0,
2268 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2269 &inode_slab_ctor);
2270 if (!ubifs_inode_slab)
2271 return -ENOMEM;
2273 register_shrinker(&ubifs_shrinker_info);
2275 err = ubifs_compressors_init();
2276 if (err)
2277 goto out_shrinker;
2279 err = dbg_debugfs_init();
2280 if (err)
2281 goto out_compr;
2283 err = register_filesystem(&ubifs_fs_type);
2284 if (err) {
2285 ubifs_err("cannot register file system, error %d", err);
2286 goto out_dbg;
2288 return 0;
2290 out_dbg:
2291 dbg_debugfs_exit();
2292 out_compr:
2293 ubifs_compressors_exit();
2294 out_shrinker:
2295 unregister_shrinker(&ubifs_shrinker_info);
2296 kmem_cache_destroy(ubifs_inode_slab);
2297 return err;
2299 /* late_initcall to let compressors initialize first */
2300 late_initcall(ubifs_init);
2302 static void __exit ubifs_exit(void)
2304 ubifs_assert(list_empty(&ubifs_infos));
2305 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2307 dbg_debugfs_exit();
2308 ubifs_compressors_exit();
2309 unregister_shrinker(&ubifs_shrinker_info);
2310 kmem_cache_destroy(ubifs_inode_slab);
2311 unregister_filesystem(&ubifs_fs_type);
2313 module_exit(ubifs_exit);
2315 MODULE_LICENSE("GPL");
2316 MODULE_VERSION(__stringify(UBIFS_VERSION));
2317 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2318 MODULE_DESCRIPTION("UBIFS - UBI File System");