Merge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
[linux-2.6.git] / fs / ubifs / super.c
blobf69daa514a57a0827a55c47f4680dcb5e69ec746
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("inode is too large (%lld)",
74 (long long)inode->i_size);
75 return 1;
78 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
79 ubifs_err("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("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 /* 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 ubifs_dump_node(c, ino);
250 ubifs_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 background 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 clear_inode(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 ubifs_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 ubifs_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;
817 * Garbage Collector head does not need to be synchronized by timer.
818 * Also GC head nodes are not grouped.
820 c->jheads[GCHD].wbuf.no_timer = 1;
821 c->jheads[GCHD].grouped = 0;
823 return 0;
827 * free_wbufs - free write-buffers.
828 * @c: UBIFS file-system description object
830 static void free_wbufs(struct ubifs_info *c)
832 int i;
834 if (c->jheads) {
835 for (i = 0; i < c->jhead_cnt; i++) {
836 kfree(c->jheads[i].wbuf.buf);
837 kfree(c->jheads[i].wbuf.inodes);
839 kfree(c->jheads);
840 c->jheads = NULL;
845 * free_orphans - free orphans.
846 * @c: UBIFS file-system description object
848 static void free_orphans(struct ubifs_info *c)
850 struct ubifs_orphan *orph;
852 while (c->orph_dnext) {
853 orph = c->orph_dnext;
854 c->orph_dnext = orph->dnext;
855 list_del(&orph->list);
856 kfree(orph);
859 while (!list_empty(&c->orph_list)) {
860 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
861 list_del(&orph->list);
862 kfree(orph);
863 ubifs_err("orphan list not empty at unmount");
866 vfree(c->orph_buf);
867 c->orph_buf = NULL;
871 * free_buds - free per-bud objects.
872 * @c: UBIFS file-system description object
874 static void free_buds(struct ubifs_info *c)
876 struct rb_node *this = c->buds.rb_node;
877 struct ubifs_bud *bud;
879 while (this) {
880 if (this->rb_left)
881 this = this->rb_left;
882 else if (this->rb_right)
883 this = this->rb_right;
884 else {
885 bud = rb_entry(this, struct ubifs_bud, rb);
886 this = rb_parent(this);
887 if (this) {
888 if (this->rb_left == &bud->rb)
889 this->rb_left = NULL;
890 else
891 this->rb_right = NULL;
893 kfree(bud);
899 * check_volume_empty - check if the UBI volume is empty.
900 * @c: UBIFS file-system description object
902 * This function checks if the UBIFS volume is empty by looking if its LEBs are
903 * mapped or not. The result of checking is stored in the @c->empty variable.
904 * Returns zero in case of success and a negative error code in case of
905 * failure.
907 static int check_volume_empty(struct ubifs_info *c)
909 int lnum, err;
911 c->empty = 1;
912 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
913 err = ubifs_is_mapped(c, lnum);
914 if (unlikely(err < 0))
915 return err;
916 if (err == 1) {
917 c->empty = 0;
918 break;
921 cond_resched();
924 return 0;
928 * UBIFS mount options.
930 * Opt_fast_unmount: do not run a journal commit before un-mounting
931 * Opt_norm_unmount: run a journal commit before un-mounting
932 * Opt_bulk_read: enable bulk-reads
933 * Opt_no_bulk_read: disable bulk-reads
934 * Opt_chk_data_crc: check CRCs when reading data nodes
935 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
936 * Opt_override_compr: override default compressor
937 * Opt_err: just end of array marker
939 enum {
940 Opt_fast_unmount,
941 Opt_norm_unmount,
942 Opt_bulk_read,
943 Opt_no_bulk_read,
944 Opt_chk_data_crc,
945 Opt_no_chk_data_crc,
946 Opt_override_compr,
947 Opt_err,
950 static const match_table_t tokens = {
951 {Opt_fast_unmount, "fast_unmount"},
952 {Opt_norm_unmount, "norm_unmount"},
953 {Opt_bulk_read, "bulk_read"},
954 {Opt_no_bulk_read, "no_bulk_read"},
955 {Opt_chk_data_crc, "chk_data_crc"},
956 {Opt_no_chk_data_crc, "no_chk_data_crc"},
957 {Opt_override_compr, "compr=%s"},
958 {Opt_err, NULL},
962 * parse_standard_option - parse a standard mount option.
963 * @option: the option to parse
965 * Normally, standard mount options like "sync" are passed to file-systems as
966 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
967 * be present in the options string. This function tries to deal with this
968 * situation and parse standard options. Returns 0 if the option was not
969 * recognized, and the corresponding integer flag if it was.
971 * UBIFS is only interested in the "sync" option, so do not check for anything
972 * else.
974 static int parse_standard_option(const char *option)
976 ubifs_msg("parse %s", option);
977 if (!strcmp(option, "sync"))
978 return MS_SYNCHRONOUS;
979 return 0;
983 * ubifs_parse_options - parse mount parameters.
984 * @c: UBIFS file-system description object
985 * @options: parameters to parse
986 * @is_remount: non-zero if this is FS re-mount
988 * This function parses UBIFS mount options and returns zero in case success
989 * and a negative error code in case of failure.
991 static int ubifs_parse_options(struct ubifs_info *c, char *options,
992 int is_remount)
994 char *p;
995 substring_t args[MAX_OPT_ARGS];
997 if (!options)
998 return 0;
1000 while ((p = strsep(&options, ","))) {
1001 int token;
1003 if (!*p)
1004 continue;
1006 token = match_token(p, tokens, args);
1007 switch (token) {
1009 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
1010 * We accept them in order to be backward-compatible. But this
1011 * should be removed at some point.
1013 case Opt_fast_unmount:
1014 c->mount_opts.unmount_mode = 2;
1015 break;
1016 case Opt_norm_unmount:
1017 c->mount_opts.unmount_mode = 1;
1018 break;
1019 case Opt_bulk_read:
1020 c->mount_opts.bulk_read = 2;
1021 c->bulk_read = 1;
1022 break;
1023 case Opt_no_bulk_read:
1024 c->mount_opts.bulk_read = 1;
1025 c->bulk_read = 0;
1026 break;
1027 case Opt_chk_data_crc:
1028 c->mount_opts.chk_data_crc = 2;
1029 c->no_chk_data_crc = 0;
1030 break;
1031 case Opt_no_chk_data_crc:
1032 c->mount_opts.chk_data_crc = 1;
1033 c->no_chk_data_crc = 1;
1034 break;
1035 case Opt_override_compr:
1037 char *name = match_strdup(&args[0]);
1039 if (!name)
1040 return -ENOMEM;
1041 if (!strcmp(name, "none"))
1042 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1043 else if (!strcmp(name, "lzo"))
1044 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1045 else if (!strcmp(name, "zlib"))
1046 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1047 else {
1048 ubifs_err("unknown compressor \"%s\"", name);
1049 kfree(name);
1050 return -EINVAL;
1052 kfree(name);
1053 c->mount_opts.override_compr = 1;
1054 c->default_compr = c->mount_opts.compr_type;
1055 break;
1057 default:
1059 unsigned long flag;
1060 struct super_block *sb = c->vfs_sb;
1062 flag = parse_standard_option(p);
1063 if (!flag) {
1064 ubifs_err("unrecognized mount option \"%s\" or missing value",
1066 return -EINVAL;
1068 sb->s_flags |= flag;
1069 break;
1074 return 0;
1078 * destroy_journal - destroy journal data structures.
1079 * @c: UBIFS file-system description object
1081 * This function destroys journal data structures including those that may have
1082 * been created by recovery functions.
1084 static void destroy_journal(struct ubifs_info *c)
1086 while (!list_empty(&c->unclean_leb_list)) {
1087 struct ubifs_unclean_leb *ucleb;
1089 ucleb = list_entry(c->unclean_leb_list.next,
1090 struct ubifs_unclean_leb, list);
1091 list_del(&ucleb->list);
1092 kfree(ucleb);
1094 while (!list_empty(&c->old_buds)) {
1095 struct ubifs_bud *bud;
1097 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1098 list_del(&bud->list);
1099 kfree(bud);
1101 ubifs_destroy_idx_gc(c);
1102 ubifs_destroy_size_tree(c);
1103 ubifs_tnc_close(c);
1104 free_buds(c);
1108 * bu_init - initialize bulk-read information.
1109 * @c: UBIFS file-system description object
1111 static void bu_init(struct ubifs_info *c)
1113 ubifs_assert(c->bulk_read == 1);
1115 if (c->bu.buf)
1116 return; /* Already initialized */
1118 again:
1119 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1120 if (!c->bu.buf) {
1121 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1122 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1123 goto again;
1126 /* Just disable bulk-read */
1127 ubifs_warn("cannot allocate %d bytes of memory for bulk-read, disabling it",
1128 c->max_bu_buf_len);
1129 c->mount_opts.bulk_read = 1;
1130 c->bulk_read = 0;
1131 return;
1136 * check_free_space - check if there is enough free space to mount.
1137 * @c: UBIFS file-system description object
1139 * This function makes sure UBIFS has enough free space to be mounted in
1140 * read/write mode. UBIFS must always have some free space to allow deletions.
1142 static int check_free_space(struct ubifs_info *c)
1144 ubifs_assert(c->dark_wm > 0);
1145 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1146 ubifs_err("insufficient free space to mount in R/W mode");
1147 ubifs_dump_budg(c, &c->bi);
1148 ubifs_dump_lprops(c);
1149 return -ENOSPC;
1151 return 0;
1155 * mount_ubifs - mount UBIFS file-system.
1156 * @c: UBIFS file-system description object
1158 * This function mounts UBIFS file system. Returns zero in case of success and
1159 * a negative error code in case of failure.
1161 static int mount_ubifs(struct ubifs_info *c)
1163 int err;
1164 long long x, y;
1165 size_t sz;
1167 c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1168 err = init_constants_early(c);
1169 if (err)
1170 return err;
1172 err = ubifs_debugging_init(c);
1173 if (err)
1174 return err;
1176 err = check_volume_empty(c);
1177 if (err)
1178 goto out_free;
1180 if (c->empty && (c->ro_mount || c->ro_media)) {
1182 * This UBI volume is empty, and read-only, or the file system
1183 * is mounted read-only - we cannot format it.
1185 ubifs_err("can't format empty UBI volume: read-only %s",
1186 c->ro_media ? "UBI volume" : "mount");
1187 err = -EROFS;
1188 goto out_free;
1191 if (c->ro_media && !c->ro_mount) {
1192 ubifs_err("cannot mount read-write - read-only media");
1193 err = -EROFS;
1194 goto out_free;
1198 * The requirement for the buffer is that it should fit indexing B-tree
1199 * height amount of integers. We assume the height if the TNC tree will
1200 * never exceed 64.
1202 err = -ENOMEM;
1203 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1204 if (!c->bottom_up_buf)
1205 goto out_free;
1207 c->sbuf = vmalloc(c->leb_size);
1208 if (!c->sbuf)
1209 goto out_free;
1211 if (!c->ro_mount) {
1212 c->ileb_buf = vmalloc(c->leb_size);
1213 if (!c->ileb_buf)
1214 goto out_free;
1217 if (c->bulk_read == 1)
1218 bu_init(c);
1220 if (!c->ro_mount) {
1221 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
1222 GFP_KERNEL);
1223 if (!c->write_reserve_buf)
1224 goto out_free;
1227 c->mounting = 1;
1229 err = ubifs_read_superblock(c);
1230 if (err)
1231 goto out_free;
1234 * Make sure the compressor which is set as default in the superblock
1235 * or overridden by mount options is actually compiled in.
1237 if (!ubifs_compr_present(c->default_compr)) {
1238 ubifs_err("'compressor \"%s\" is not compiled in",
1239 ubifs_compr_name(c->default_compr));
1240 err = -ENOTSUPP;
1241 goto out_free;
1244 err = init_constants_sb(c);
1245 if (err)
1246 goto out_free;
1248 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1249 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1250 c->cbuf = kmalloc(sz, GFP_NOFS);
1251 if (!c->cbuf) {
1252 err = -ENOMEM;
1253 goto out_free;
1256 err = alloc_wbufs(c);
1257 if (err)
1258 goto out_cbuf;
1260 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1261 if (!c->ro_mount) {
1262 /* Create background thread */
1263 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1264 if (IS_ERR(c->bgt)) {
1265 err = PTR_ERR(c->bgt);
1266 c->bgt = NULL;
1267 ubifs_err("cannot spawn \"%s\", error %d",
1268 c->bgt_name, err);
1269 goto out_wbufs;
1271 wake_up_process(c->bgt);
1274 err = ubifs_read_master(c);
1275 if (err)
1276 goto out_master;
1278 init_constants_master(c);
1280 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1281 ubifs_msg("recovery needed");
1282 c->need_recovery = 1;
1285 if (c->need_recovery && !c->ro_mount) {
1286 err = ubifs_recover_inl_heads(c, c->sbuf);
1287 if (err)
1288 goto out_master;
1291 err = ubifs_lpt_init(c, 1, !c->ro_mount);
1292 if (err)
1293 goto out_master;
1295 if (!c->ro_mount && c->space_fixup) {
1296 err = ubifs_fixup_free_space(c);
1297 if (err)
1298 goto out_lpt;
1301 if (!c->ro_mount) {
1303 * Set the "dirty" flag so that if we reboot uncleanly we
1304 * will notice this immediately on the next mount.
1306 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1307 err = ubifs_write_master(c);
1308 if (err)
1309 goto out_lpt;
1312 err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1313 if (err)
1314 goto out_lpt;
1316 err = ubifs_replay_journal(c);
1317 if (err)
1318 goto out_journal;
1320 /* Calculate 'min_idx_lebs' after journal replay */
1321 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1323 err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1324 if (err)
1325 goto out_orphans;
1327 if (!c->ro_mount) {
1328 int lnum;
1330 err = check_free_space(c);
1331 if (err)
1332 goto out_orphans;
1334 /* Check for enough log space */
1335 lnum = c->lhead_lnum + 1;
1336 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1337 lnum = UBIFS_LOG_LNUM;
1338 if (lnum == c->ltail_lnum) {
1339 err = ubifs_consolidate_log(c);
1340 if (err)
1341 goto out_orphans;
1344 if (c->need_recovery) {
1345 err = ubifs_recover_size(c);
1346 if (err)
1347 goto out_orphans;
1348 err = ubifs_rcvry_gc_commit(c);
1349 if (err)
1350 goto out_orphans;
1351 } else {
1352 err = take_gc_lnum(c);
1353 if (err)
1354 goto out_orphans;
1357 * GC LEB may contain garbage if there was an unclean
1358 * reboot, and it should be un-mapped.
1360 err = ubifs_leb_unmap(c, c->gc_lnum);
1361 if (err)
1362 goto out_orphans;
1365 err = dbg_check_lprops(c);
1366 if (err)
1367 goto out_orphans;
1368 } else if (c->need_recovery) {
1369 err = ubifs_recover_size(c);
1370 if (err)
1371 goto out_orphans;
1372 } else {
1374 * Even if we mount read-only, we have to set space in GC LEB
1375 * to proper value because this affects UBIFS free space
1376 * reporting. We do not want to have a situation when
1377 * re-mounting from R/O to R/W changes amount of free space.
1379 err = take_gc_lnum(c);
1380 if (err)
1381 goto out_orphans;
1384 spin_lock(&ubifs_infos_lock);
1385 list_add_tail(&c->infos_list, &ubifs_infos);
1386 spin_unlock(&ubifs_infos_lock);
1388 if (c->need_recovery) {
1389 if (c->ro_mount)
1390 ubifs_msg("recovery deferred");
1391 else {
1392 c->need_recovery = 0;
1393 ubifs_msg("recovery completed");
1395 * GC LEB has to be empty and taken at this point. But
1396 * the journal head LEBs may also be accounted as
1397 * "empty taken" if they are empty.
1399 ubifs_assert(c->lst.taken_empty_lebs > 0);
1401 } else
1402 ubifs_assert(c->lst.taken_empty_lebs > 0);
1404 err = dbg_check_filesystem(c);
1405 if (err)
1406 goto out_infos;
1408 err = dbg_debugfs_init_fs(c);
1409 if (err)
1410 goto out_infos;
1412 c->mounting = 0;
1414 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"%s",
1415 c->vi.ubi_num, c->vi.vol_id, c->vi.name,
1416 c->ro_mount ? ", R/O mode" : "");
1417 x = (long long)c->main_lebs * c->leb_size;
1418 y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1419 ubifs_msg("LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes",
1420 c->leb_size, c->leb_size >> 10, c->min_io_size,
1421 c->max_write_size);
1422 ubifs_msg("FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)",
1423 x, x >> 20, c->main_lebs,
1424 y, y >> 20, c->log_lebs + c->max_bud_cnt);
1425 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1426 c->report_rp_size, c->report_rp_size >> 10);
1427 ubifs_msg("media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s",
1428 c->fmt_version, c->ro_compat_version,
1429 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid,
1430 c->big_lpt ? ", big LPT model" : ", small LPT model");
1432 dbg_gen("default compressor: %s", ubifs_compr_name(c->default_compr));
1433 dbg_gen("data journal heads: %d",
1434 c->jhead_cnt - NONDATA_JHEADS_CNT);
1435 dbg_gen("log LEBs: %d (%d - %d)",
1436 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1437 dbg_gen("LPT area LEBs: %d (%d - %d)",
1438 c->lpt_lebs, c->lpt_first, c->lpt_last);
1439 dbg_gen("orphan area LEBs: %d (%d - %d)",
1440 c->orph_lebs, c->orph_first, c->orph_last);
1441 dbg_gen("main area LEBs: %d (%d - %d)",
1442 c->main_lebs, c->main_first, c->leb_cnt - 1);
1443 dbg_gen("index LEBs: %d", c->lst.idx_lebs);
1444 dbg_gen("total index bytes: %lld (%lld KiB, %lld MiB)",
1445 c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1446 c->bi.old_idx_sz >> 20);
1447 dbg_gen("key hash type: %d", c->key_hash_type);
1448 dbg_gen("tree fanout: %d", c->fanout);
1449 dbg_gen("reserved GC LEB: %d", c->gc_lnum);
1450 dbg_gen("max. znode size %d", c->max_znode_sz);
1451 dbg_gen("max. index node size %d", c->max_idx_node_sz);
1452 dbg_gen("node sizes: data %zu, inode %zu, dentry %zu",
1453 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1454 dbg_gen("node sizes: trun %zu, sb %zu, master %zu",
1455 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1456 dbg_gen("node sizes: ref %zu, cmt. start %zu, orph %zu",
1457 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1458 dbg_gen("max. node sizes: data %zu, inode %zu dentry %zu, idx %d",
1459 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1460 UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1461 dbg_gen("dead watermark: %d", c->dead_wm);
1462 dbg_gen("dark watermark: %d", c->dark_wm);
1463 dbg_gen("LEB overhead: %d", c->leb_overhead);
1464 x = (long long)c->main_lebs * c->dark_wm;
1465 dbg_gen("max. dark space: %lld (%lld KiB, %lld MiB)",
1466 x, x >> 10, x >> 20);
1467 dbg_gen("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1468 c->max_bud_bytes, c->max_bud_bytes >> 10,
1469 c->max_bud_bytes >> 20);
1470 dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1471 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1472 c->bg_bud_bytes >> 20);
1473 dbg_gen("current bud bytes %lld (%lld KiB, %lld MiB)",
1474 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1475 dbg_gen("max. seq. number: %llu", c->max_sqnum);
1476 dbg_gen("commit number: %llu", c->cmt_no);
1478 return 0;
1480 out_infos:
1481 spin_lock(&ubifs_infos_lock);
1482 list_del(&c->infos_list);
1483 spin_unlock(&ubifs_infos_lock);
1484 out_orphans:
1485 free_orphans(c);
1486 out_journal:
1487 destroy_journal(c);
1488 out_lpt:
1489 ubifs_lpt_free(c, 0);
1490 out_master:
1491 kfree(c->mst_node);
1492 kfree(c->rcvrd_mst_node);
1493 if (c->bgt)
1494 kthread_stop(c->bgt);
1495 out_wbufs:
1496 free_wbufs(c);
1497 out_cbuf:
1498 kfree(c->cbuf);
1499 out_free:
1500 kfree(c->write_reserve_buf);
1501 kfree(c->bu.buf);
1502 vfree(c->ileb_buf);
1503 vfree(c->sbuf);
1504 kfree(c->bottom_up_buf);
1505 ubifs_debugging_exit(c);
1506 return err;
1510 * ubifs_umount - un-mount UBIFS file-system.
1511 * @c: UBIFS file-system description object
1513 * Note, this function is called to free allocated resourced when un-mounting,
1514 * as well as free resources when an error occurred while we were half way
1515 * through mounting (error path cleanup function). So it has to make sure the
1516 * resource was actually allocated before freeing it.
1518 static void ubifs_umount(struct ubifs_info *c)
1520 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1521 c->vi.vol_id);
1523 dbg_debugfs_exit_fs(c);
1524 spin_lock(&ubifs_infos_lock);
1525 list_del(&c->infos_list);
1526 spin_unlock(&ubifs_infos_lock);
1528 if (c->bgt)
1529 kthread_stop(c->bgt);
1531 destroy_journal(c);
1532 free_wbufs(c);
1533 free_orphans(c);
1534 ubifs_lpt_free(c, 0);
1536 kfree(c->cbuf);
1537 kfree(c->rcvrd_mst_node);
1538 kfree(c->mst_node);
1539 kfree(c->write_reserve_buf);
1540 kfree(c->bu.buf);
1541 vfree(c->ileb_buf);
1542 vfree(c->sbuf);
1543 kfree(c->bottom_up_buf);
1544 ubifs_debugging_exit(c);
1548 * ubifs_remount_rw - re-mount in read-write mode.
1549 * @c: UBIFS file-system description object
1551 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1552 * mode. This function allocates the needed resources and re-mounts UBIFS in
1553 * read-write mode.
1555 static int ubifs_remount_rw(struct ubifs_info *c)
1557 int err, lnum;
1559 if (c->rw_incompat) {
1560 ubifs_err("the file-system is not R/W-compatible");
1561 ubifs_msg("on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
1562 c->fmt_version, c->ro_compat_version,
1563 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1564 return -EROFS;
1567 mutex_lock(&c->umount_mutex);
1568 dbg_save_space_info(c);
1569 c->remounting_rw = 1;
1570 c->ro_mount = 0;
1572 if (c->space_fixup) {
1573 err = ubifs_fixup_free_space(c);
1574 if (err)
1575 return err;
1578 err = check_free_space(c);
1579 if (err)
1580 goto out;
1582 if (c->old_leb_cnt != c->leb_cnt) {
1583 struct ubifs_sb_node *sup;
1585 sup = ubifs_read_sb_node(c);
1586 if (IS_ERR(sup)) {
1587 err = PTR_ERR(sup);
1588 goto out;
1590 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1591 err = ubifs_write_sb_node(c, sup);
1592 kfree(sup);
1593 if (err)
1594 goto out;
1597 if (c->need_recovery) {
1598 ubifs_msg("completing deferred recovery");
1599 err = ubifs_write_rcvrd_mst_node(c);
1600 if (err)
1601 goto out;
1602 err = ubifs_recover_size(c);
1603 if (err)
1604 goto out;
1605 err = ubifs_clean_lebs(c, c->sbuf);
1606 if (err)
1607 goto out;
1608 err = ubifs_recover_inl_heads(c, c->sbuf);
1609 if (err)
1610 goto out;
1611 } else {
1612 /* A readonly mount is not allowed to have orphans */
1613 ubifs_assert(c->tot_orphans == 0);
1614 err = ubifs_clear_orphans(c);
1615 if (err)
1616 goto out;
1619 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1620 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1621 err = ubifs_write_master(c);
1622 if (err)
1623 goto out;
1626 c->ileb_buf = vmalloc(c->leb_size);
1627 if (!c->ileb_buf) {
1628 err = -ENOMEM;
1629 goto out;
1632 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
1633 if (!c->write_reserve_buf) {
1634 err = -ENOMEM;
1635 goto out;
1638 err = ubifs_lpt_init(c, 0, 1);
1639 if (err)
1640 goto out;
1642 /* Create background thread */
1643 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1644 if (IS_ERR(c->bgt)) {
1645 err = PTR_ERR(c->bgt);
1646 c->bgt = NULL;
1647 ubifs_err("cannot spawn \"%s\", error %d",
1648 c->bgt_name, err);
1649 goto out;
1651 wake_up_process(c->bgt);
1653 c->orph_buf = vmalloc(c->leb_size);
1654 if (!c->orph_buf) {
1655 err = -ENOMEM;
1656 goto out;
1659 /* Check for enough log space */
1660 lnum = c->lhead_lnum + 1;
1661 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1662 lnum = UBIFS_LOG_LNUM;
1663 if (lnum == c->ltail_lnum) {
1664 err = ubifs_consolidate_log(c);
1665 if (err)
1666 goto out;
1669 if (c->need_recovery)
1670 err = ubifs_rcvry_gc_commit(c);
1671 else
1672 err = ubifs_leb_unmap(c, c->gc_lnum);
1673 if (err)
1674 goto out;
1676 dbg_gen("re-mounted read-write");
1677 c->remounting_rw = 0;
1679 if (c->need_recovery) {
1680 c->need_recovery = 0;
1681 ubifs_msg("deferred recovery completed");
1682 } else {
1684 * Do not run the debugging space check if the were doing
1685 * recovery, because when we saved the information we had the
1686 * file-system in a state where the TNC and lprops has been
1687 * modified in memory, but all the I/O operations (including a
1688 * commit) were deferred. So the file-system was in
1689 * "non-committed" state. Now the file-system is in committed
1690 * state, and of course the amount of free space will change
1691 * because, for example, the old index size was imprecise.
1693 err = dbg_check_space_info(c);
1696 mutex_unlock(&c->umount_mutex);
1697 return err;
1699 out:
1700 c->ro_mount = 1;
1701 vfree(c->orph_buf);
1702 c->orph_buf = NULL;
1703 if (c->bgt) {
1704 kthread_stop(c->bgt);
1705 c->bgt = NULL;
1707 free_wbufs(c);
1708 kfree(c->write_reserve_buf);
1709 c->write_reserve_buf = NULL;
1710 vfree(c->ileb_buf);
1711 c->ileb_buf = NULL;
1712 ubifs_lpt_free(c, 1);
1713 c->remounting_rw = 0;
1714 mutex_unlock(&c->umount_mutex);
1715 return err;
1719 * ubifs_remount_ro - re-mount in read-only mode.
1720 * @c: UBIFS file-system description object
1722 * We assume VFS has stopped writing. Possibly the background thread could be
1723 * running a commit, however kthread_stop will wait in that case.
1725 static void ubifs_remount_ro(struct ubifs_info *c)
1727 int i, err;
1729 ubifs_assert(!c->need_recovery);
1730 ubifs_assert(!c->ro_mount);
1732 mutex_lock(&c->umount_mutex);
1733 if (c->bgt) {
1734 kthread_stop(c->bgt);
1735 c->bgt = NULL;
1738 dbg_save_space_info(c);
1740 for (i = 0; i < c->jhead_cnt; i++)
1741 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1743 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1744 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1745 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1746 err = ubifs_write_master(c);
1747 if (err)
1748 ubifs_ro_mode(c, err);
1750 vfree(c->orph_buf);
1751 c->orph_buf = NULL;
1752 kfree(c->write_reserve_buf);
1753 c->write_reserve_buf = NULL;
1754 vfree(c->ileb_buf);
1755 c->ileb_buf = NULL;
1756 ubifs_lpt_free(c, 1);
1757 c->ro_mount = 1;
1758 err = dbg_check_space_info(c);
1759 if (err)
1760 ubifs_ro_mode(c, err);
1761 mutex_unlock(&c->umount_mutex);
1764 static void ubifs_put_super(struct super_block *sb)
1766 int i;
1767 struct ubifs_info *c = sb->s_fs_info;
1769 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1770 c->vi.vol_id);
1773 * The following asserts are only valid if there has not been a failure
1774 * of the media. For example, there will be dirty inodes if we failed
1775 * to write them back because of I/O errors.
1777 if (!c->ro_error) {
1778 ubifs_assert(c->bi.idx_growth == 0);
1779 ubifs_assert(c->bi.dd_growth == 0);
1780 ubifs_assert(c->bi.data_growth == 0);
1784 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1785 * and file system un-mount. Namely, it prevents the shrinker from
1786 * picking this superblock for shrinking - it will be just skipped if
1787 * the mutex is locked.
1789 mutex_lock(&c->umount_mutex);
1790 if (!c->ro_mount) {
1792 * First of all kill the background thread to make sure it does
1793 * not interfere with un-mounting and freeing resources.
1795 if (c->bgt) {
1796 kthread_stop(c->bgt);
1797 c->bgt = NULL;
1801 * On fatal errors c->ro_error is set to 1, in which case we do
1802 * not write the master node.
1804 if (!c->ro_error) {
1805 int err;
1807 /* Synchronize write-buffers */
1808 for (i = 0; i < c->jhead_cnt; i++)
1809 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1812 * We are being cleanly unmounted which means the
1813 * orphans were killed - indicate this in the master
1814 * node. Also save the reserved GC LEB number.
1816 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1817 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1818 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1819 err = ubifs_write_master(c);
1820 if (err)
1822 * Recovery will attempt to fix the master area
1823 * next mount, so we just print a message and
1824 * continue to unmount normally.
1826 ubifs_err("failed to write master node, error %d",
1827 err);
1828 } else {
1829 for (i = 0; i < c->jhead_cnt; i++)
1830 /* Make sure write-buffer timers are canceled */
1831 hrtimer_cancel(&c->jheads[i].wbuf.timer);
1835 ubifs_umount(c);
1836 bdi_destroy(&c->bdi);
1837 ubi_close_volume(c->ubi);
1838 mutex_unlock(&c->umount_mutex);
1841 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1843 int err;
1844 struct ubifs_info *c = sb->s_fs_info;
1846 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1848 err = ubifs_parse_options(c, data, 1);
1849 if (err) {
1850 ubifs_err("invalid or unknown remount parameter");
1851 return err;
1854 if (c->ro_mount && !(*flags & MS_RDONLY)) {
1855 if (c->ro_error) {
1856 ubifs_msg("cannot re-mount R/W due to prior errors");
1857 return -EROFS;
1859 if (c->ro_media) {
1860 ubifs_msg("cannot re-mount R/W - UBI volume is R/O");
1861 return -EROFS;
1863 err = ubifs_remount_rw(c);
1864 if (err)
1865 return err;
1866 } else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1867 if (c->ro_error) {
1868 ubifs_msg("cannot re-mount R/O due to prior errors");
1869 return -EROFS;
1871 ubifs_remount_ro(c);
1874 if (c->bulk_read == 1)
1875 bu_init(c);
1876 else {
1877 dbg_gen("disable bulk-read");
1878 kfree(c->bu.buf);
1879 c->bu.buf = NULL;
1882 ubifs_assert(c->lst.taken_empty_lebs > 0);
1883 return 0;
1886 const struct super_operations ubifs_super_operations = {
1887 .alloc_inode = ubifs_alloc_inode,
1888 .destroy_inode = ubifs_destroy_inode,
1889 .put_super = ubifs_put_super,
1890 .write_inode = ubifs_write_inode,
1891 .evict_inode = ubifs_evict_inode,
1892 .statfs = ubifs_statfs,
1893 .dirty_inode = ubifs_dirty_inode,
1894 .remount_fs = ubifs_remount_fs,
1895 .show_options = ubifs_show_options,
1896 .sync_fs = ubifs_sync_fs,
1900 * open_ubi - parse UBI device name string and open the UBI device.
1901 * @name: UBI volume name
1902 * @mode: UBI volume open mode
1904 * The primary method of mounting UBIFS is by specifying the UBI volume
1905 * character device node path. However, UBIFS may also be mounted withoug any
1906 * character device node using one of the following methods:
1908 * o ubiX_Y - mount UBI device number X, volume Y;
1909 * o ubiY - mount UBI device number 0, volume Y;
1910 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1911 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1913 * Alternative '!' separator may be used instead of ':' (because some shells
1914 * like busybox may interpret ':' as an NFS host name separator). This function
1915 * returns UBI volume description object in case of success and a negative
1916 * error code in case of failure.
1918 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1920 struct ubi_volume_desc *ubi;
1921 int dev, vol;
1922 char *endptr;
1924 /* First, try to open using the device node path method */
1925 ubi = ubi_open_volume_path(name, mode);
1926 if (!IS_ERR(ubi))
1927 return ubi;
1929 /* Try the "nodev" method */
1930 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1931 return ERR_PTR(-EINVAL);
1933 /* ubi:NAME method */
1934 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1935 return ubi_open_volume_nm(0, name + 4, mode);
1937 if (!isdigit(name[3]))
1938 return ERR_PTR(-EINVAL);
1940 dev = simple_strtoul(name + 3, &endptr, 0);
1942 /* ubiY method */
1943 if (*endptr == '\0')
1944 return ubi_open_volume(0, dev, mode);
1946 /* ubiX_Y method */
1947 if (*endptr == '_' && isdigit(endptr[1])) {
1948 vol = simple_strtoul(endptr + 1, &endptr, 0);
1949 if (*endptr != '\0')
1950 return ERR_PTR(-EINVAL);
1951 return ubi_open_volume(dev, vol, mode);
1954 /* ubiX:NAME method */
1955 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1956 return ubi_open_volume_nm(dev, ++endptr, mode);
1958 return ERR_PTR(-EINVAL);
1961 static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1963 struct ubifs_info *c;
1965 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1966 if (c) {
1967 spin_lock_init(&c->cnt_lock);
1968 spin_lock_init(&c->cs_lock);
1969 spin_lock_init(&c->buds_lock);
1970 spin_lock_init(&c->space_lock);
1971 spin_lock_init(&c->orphan_lock);
1972 init_rwsem(&c->commit_sem);
1973 mutex_init(&c->lp_mutex);
1974 mutex_init(&c->tnc_mutex);
1975 mutex_init(&c->log_mutex);
1976 mutex_init(&c->mst_mutex);
1977 mutex_init(&c->umount_mutex);
1978 mutex_init(&c->bu_mutex);
1979 mutex_init(&c->write_reserve_mutex);
1980 init_waitqueue_head(&c->cmt_wq);
1981 c->buds = RB_ROOT;
1982 c->old_idx = RB_ROOT;
1983 c->size_tree = RB_ROOT;
1984 c->orph_tree = RB_ROOT;
1985 INIT_LIST_HEAD(&c->infos_list);
1986 INIT_LIST_HEAD(&c->idx_gc);
1987 INIT_LIST_HEAD(&c->replay_list);
1988 INIT_LIST_HEAD(&c->replay_buds);
1989 INIT_LIST_HEAD(&c->uncat_list);
1990 INIT_LIST_HEAD(&c->empty_list);
1991 INIT_LIST_HEAD(&c->freeable_list);
1992 INIT_LIST_HEAD(&c->frdi_idx_list);
1993 INIT_LIST_HEAD(&c->unclean_leb_list);
1994 INIT_LIST_HEAD(&c->old_buds);
1995 INIT_LIST_HEAD(&c->orph_list);
1996 INIT_LIST_HEAD(&c->orph_new);
1997 c->no_chk_data_crc = 1;
1999 c->highest_inum = UBIFS_FIRST_INO;
2000 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
2002 ubi_get_volume_info(ubi, &c->vi);
2003 ubi_get_device_info(c->vi.ubi_num, &c->di);
2005 return c;
2008 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
2010 struct ubifs_info *c = sb->s_fs_info;
2011 struct inode *root;
2012 int err;
2014 c->vfs_sb = sb;
2015 /* Re-open the UBI device in read-write mode */
2016 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2017 if (IS_ERR(c->ubi)) {
2018 err = PTR_ERR(c->ubi);
2019 goto out;
2023 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2024 * UBIFS, I/O is not deferred, it is done immediately in readpage,
2025 * which means the user would have to wait not just for their own I/O
2026 * but the read-ahead I/O as well i.e. completely pointless.
2028 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
2030 c->bdi.name = "ubifs",
2031 c->bdi.capabilities = BDI_CAP_MAP_COPY;
2032 err = bdi_init(&c->bdi);
2033 if (err)
2034 goto out_close;
2035 err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
2036 c->vi.ubi_num, c->vi.vol_id);
2037 if (err)
2038 goto out_bdi;
2040 err = ubifs_parse_options(c, data, 0);
2041 if (err)
2042 goto out_bdi;
2044 sb->s_bdi = &c->bdi;
2045 sb->s_fs_info = c;
2046 sb->s_magic = UBIFS_SUPER_MAGIC;
2047 sb->s_blocksize = UBIFS_BLOCK_SIZE;
2048 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2049 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2050 if (c->max_inode_sz > MAX_LFS_FILESIZE)
2051 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2052 sb->s_op = &ubifs_super_operations;
2054 mutex_lock(&c->umount_mutex);
2055 err = mount_ubifs(c);
2056 if (err) {
2057 ubifs_assert(err < 0);
2058 goto out_unlock;
2061 /* Read the root inode */
2062 root = ubifs_iget(sb, UBIFS_ROOT_INO);
2063 if (IS_ERR(root)) {
2064 err = PTR_ERR(root);
2065 goto out_umount;
2068 sb->s_root = d_make_root(root);
2069 if (!sb->s_root) {
2070 err = -ENOMEM;
2071 goto out_umount;
2074 mutex_unlock(&c->umount_mutex);
2075 return 0;
2077 out_umount:
2078 ubifs_umount(c);
2079 out_unlock:
2080 mutex_unlock(&c->umount_mutex);
2081 out_bdi:
2082 bdi_destroy(&c->bdi);
2083 out_close:
2084 ubi_close_volume(c->ubi);
2085 out:
2086 return err;
2089 static int sb_test(struct super_block *sb, void *data)
2091 struct ubifs_info *c1 = data;
2092 struct ubifs_info *c = sb->s_fs_info;
2094 return c->vi.cdev == c1->vi.cdev;
2097 static int sb_set(struct super_block *sb, void *data)
2099 sb->s_fs_info = data;
2100 return set_anon_super(sb, NULL);
2103 static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2104 const char *name, void *data)
2106 struct ubi_volume_desc *ubi;
2107 struct ubifs_info *c;
2108 struct super_block *sb;
2109 int err;
2111 dbg_gen("name %s, flags %#x", name, flags);
2114 * Get UBI device number and volume ID. Mount it read-only so far
2115 * because this might be a new mount point, and UBI allows only one
2116 * read-write user at a time.
2118 ubi = open_ubi(name, UBI_READONLY);
2119 if (IS_ERR(ubi)) {
2120 ubifs_err("cannot open \"%s\", error %d",
2121 name, (int)PTR_ERR(ubi));
2122 return ERR_CAST(ubi);
2125 c = alloc_ubifs_info(ubi);
2126 if (!c) {
2127 err = -ENOMEM;
2128 goto out_close;
2131 dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2133 sb = sget(fs_type, sb_test, sb_set, flags, c);
2134 if (IS_ERR(sb)) {
2135 err = PTR_ERR(sb);
2136 kfree(c);
2137 goto out_close;
2140 if (sb->s_root) {
2141 struct ubifs_info *c1 = sb->s_fs_info;
2142 kfree(c);
2143 /* A new mount point for already mounted UBIFS */
2144 dbg_gen("this ubi volume is already mounted");
2145 if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2146 err = -EBUSY;
2147 goto out_deact;
2149 } else {
2150 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2151 if (err)
2152 goto out_deact;
2153 /* We do not support atime */
2154 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2157 /* 'fill_super()' opens ubi again so we must close it here */
2158 ubi_close_volume(ubi);
2160 return dget(sb->s_root);
2162 out_deact:
2163 deactivate_locked_super(sb);
2164 out_close:
2165 ubi_close_volume(ubi);
2166 return ERR_PTR(err);
2169 static void kill_ubifs_super(struct super_block *s)
2171 struct ubifs_info *c = s->s_fs_info;
2172 kill_anon_super(s);
2173 kfree(c);
2176 static struct file_system_type ubifs_fs_type = {
2177 .name = "ubifs",
2178 .owner = THIS_MODULE,
2179 .mount = ubifs_mount,
2180 .kill_sb = kill_ubifs_super,
2182 MODULE_ALIAS_FS("ubifs");
2185 * Inode slab cache constructor.
2187 static void inode_slab_ctor(void *obj)
2189 struct ubifs_inode *ui = obj;
2190 inode_init_once(&ui->vfs_inode);
2193 static int __init ubifs_init(void)
2195 int err;
2197 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2199 /* Make sure node sizes are 8-byte aligned */
2200 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2201 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2202 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2203 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2204 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2205 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2206 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2207 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2208 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2209 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2210 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2212 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2213 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2214 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2215 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2216 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2217 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2219 /* Check min. node size */
2220 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2221 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2222 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2223 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2225 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2226 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2227 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2228 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2230 /* Defined node sizes */
2231 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2232 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2233 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2234 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2237 * We use 2 bit wide bit-fields to store compression type, which should
2238 * be amended if more compressors are added. The bit-fields are:
2239 * @compr_type in 'struct ubifs_inode', @default_compr in
2240 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2242 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2245 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2246 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2248 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2249 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
2250 (unsigned int)PAGE_CACHE_SIZE);
2251 return -EINVAL;
2254 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2255 sizeof(struct ubifs_inode), 0,
2256 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2257 &inode_slab_ctor);
2258 if (!ubifs_inode_slab)
2259 return -ENOMEM;
2261 register_shrinker(&ubifs_shrinker_info);
2263 err = ubifs_compressors_init();
2264 if (err)
2265 goto out_shrinker;
2267 err = dbg_debugfs_init();
2268 if (err)
2269 goto out_compr;
2271 err = register_filesystem(&ubifs_fs_type);
2272 if (err) {
2273 ubifs_err("cannot register file system, error %d", err);
2274 goto out_dbg;
2276 return 0;
2278 out_dbg:
2279 dbg_debugfs_exit();
2280 out_compr:
2281 ubifs_compressors_exit();
2282 out_shrinker:
2283 unregister_shrinker(&ubifs_shrinker_info);
2284 kmem_cache_destroy(ubifs_inode_slab);
2285 return err;
2287 /* late_initcall to let compressors initialize first */
2288 late_initcall(ubifs_init);
2290 static void __exit ubifs_exit(void)
2292 ubifs_assert(list_empty(&ubifs_infos));
2293 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2295 dbg_debugfs_exit();
2296 ubifs_compressors_exit();
2297 unregister_shrinker(&ubifs_shrinker_info);
2300 * Make sure all delayed rcu free inodes are flushed before we
2301 * destroy cache.
2303 rcu_barrier();
2304 kmem_cache_destroy(ubifs_inode_slab);
2305 unregister_filesystem(&ubifs_fs_type);
2307 module_exit(ubifs_exit);
2309 MODULE_LICENSE("GPL");
2310 MODULE_VERSION(__stringify(UBIFS_VERSION));
2311 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2312 MODULE_DESCRIPTION("UBIFS - UBI File System");