ARM: dts: dm816x: Add an empty chosen node to top level DTSI
[linux-2.6/btrfs-unstable.git] / fs / ubifs / super.c
blobe08aa04fc8351ca03db6479f94a33c0fe1ff845b
1 /*
2 * This file is part of UBIFS.
4 * Copyright (C) 2006-2008 Nokia Corporation.
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License version 2 as published by
8 * the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * more details.
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc., 51
17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 * Authors: Artem Bityutskiy (Битюцкий Артём)
20 * Adrian Hunter
24 * This file implements UBIFS initialization and VFS superblock operations. Some
25 * initialization stuff which is rather large and complex is placed at
26 * corresponding subsystems, but most of it is here.
29 #include <linux/init.h>
30 #include <linux/slab.h>
31 #include <linux/module.h>
32 #include <linux/ctype.h>
33 #include <linux/kthread.h>
34 #include <linux/parser.h>
35 #include <linux/seq_file.h>
36 #include <linux/mount.h>
37 #include <linux/math64.h>
38 #include <linux/writeback.h>
39 #include "ubifs.h"
42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
43 * allocating too much.
45 #define UBIFS_KMALLOC_OK (128*1024)
47 /* Slab cache for UBIFS inodes */
48 struct kmem_cache *ubifs_inode_slab;
50 /* UBIFS TNC shrinker description */
51 static struct shrinker ubifs_shrinker_info = {
52 .scan_objects = ubifs_shrink_scan,
53 .count_objects = ubifs_shrink_count,
54 .seeks = DEFAULT_SEEKS,
57 /**
58 * validate_inode - validate inode.
59 * @c: UBIFS file-system description object
60 * @inode: the inode to validate
62 * This is a helper function for 'ubifs_iget()' which validates various fields
63 * of a newly built inode to make sure they contain sane values and prevent
64 * possible vulnerabilities. Returns zero if the inode is all right and
65 * a non-zero error code if not.
67 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
69 int err;
70 const struct ubifs_inode *ui = ubifs_inode(inode);
72 if (inode->i_size > c->max_inode_sz) {
73 ubifs_err(c, "inode is too large (%lld)",
74 (long long)inode->i_size);
75 return 1;
78 if (ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
79 ubifs_err(c, "unknown compression type %d", ui->compr_type);
80 return 2;
83 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
84 return 3;
86 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
87 return 4;
89 if (ui->xattr && !S_ISREG(inode->i_mode))
90 return 5;
92 if (!ubifs_compr_present(ui->compr_type)) {
93 ubifs_warn(c, "inode %lu uses '%s' compression, but it was not compiled in",
94 inode->i_ino, ubifs_compr_name(ui->compr_type));
97 err = dbg_check_dir(c, inode);
98 return err;
101 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
103 int err;
104 union ubifs_key key;
105 struct ubifs_ino_node *ino;
106 struct ubifs_info *c = sb->s_fs_info;
107 struct inode *inode;
108 struct ubifs_inode *ui;
110 dbg_gen("inode %lu", inum);
112 inode = iget_locked(sb, inum);
113 if (!inode)
114 return ERR_PTR(-ENOMEM);
115 if (!(inode->i_state & I_NEW))
116 return inode;
117 ui = ubifs_inode(inode);
119 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
120 if (!ino) {
121 err = -ENOMEM;
122 goto out;
125 ino_key_init(c, &key, inode->i_ino);
127 err = ubifs_tnc_lookup(c, &key, ino);
128 if (err)
129 goto out_ino;
131 inode->i_flags |= S_NOCMTIME;
132 #ifndef CONFIG_UBIFS_ATIME_SUPPORT
133 inode->i_flags |= S_NOATIME;
134 #endif
135 set_nlink(inode, le32_to_cpu(ino->nlink));
136 i_uid_write(inode, le32_to_cpu(ino->uid));
137 i_gid_write(inode, le32_to_cpu(ino->gid));
138 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
139 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
140 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
141 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
142 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
143 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
144 inode->i_mode = le32_to_cpu(ino->mode);
145 inode->i_size = le64_to_cpu(ino->size);
147 ui->data_len = le32_to_cpu(ino->data_len);
148 ui->flags = le32_to_cpu(ino->flags);
149 ui->compr_type = le16_to_cpu(ino->compr_type);
150 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
151 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
152 ui->xattr_size = le32_to_cpu(ino->xattr_size);
153 ui->xattr_names = le32_to_cpu(ino->xattr_names);
154 ui->synced_i_size = ui->ui_size = inode->i_size;
156 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
158 err = validate_inode(c, inode);
159 if (err)
160 goto out_invalid;
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(c, "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(c, "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(c, "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_final(&inode->i_data);
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(c, "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);
382 #ifdef CONFIG_UBIFS_FS_ENCRYPTION
383 fscrypt_put_encryption_info(inode, NULL);
384 #endif
387 static void ubifs_dirty_inode(struct inode *inode, int flags)
389 struct ubifs_inode *ui = ubifs_inode(inode);
391 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
392 if (!ui->dirty) {
393 ui->dirty = 1;
394 dbg_gen("inode %lu", inode->i_ino);
398 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
400 struct ubifs_info *c = dentry->d_sb->s_fs_info;
401 unsigned long long free;
402 __le32 *uuid = (__le32 *)c->uuid;
404 free = ubifs_get_free_space(c);
405 dbg_gen("free space %lld bytes (%lld blocks)",
406 free, free >> UBIFS_BLOCK_SHIFT);
408 buf->f_type = UBIFS_SUPER_MAGIC;
409 buf->f_bsize = UBIFS_BLOCK_SIZE;
410 buf->f_blocks = c->block_cnt;
411 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
412 if (free > c->report_rp_size)
413 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
414 else
415 buf->f_bavail = 0;
416 buf->f_files = 0;
417 buf->f_ffree = 0;
418 buf->f_namelen = UBIFS_MAX_NLEN;
419 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
420 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
421 ubifs_assert(buf->f_bfree <= c->block_cnt);
422 return 0;
425 static int ubifs_show_options(struct seq_file *s, struct dentry *root)
427 struct ubifs_info *c = root->d_sb->s_fs_info;
429 if (c->mount_opts.unmount_mode == 2)
430 seq_puts(s, ",fast_unmount");
431 else if (c->mount_opts.unmount_mode == 1)
432 seq_puts(s, ",norm_unmount");
434 if (c->mount_opts.bulk_read == 2)
435 seq_puts(s, ",bulk_read");
436 else if (c->mount_opts.bulk_read == 1)
437 seq_puts(s, ",no_bulk_read");
439 if (c->mount_opts.chk_data_crc == 2)
440 seq_puts(s, ",chk_data_crc");
441 else if (c->mount_opts.chk_data_crc == 1)
442 seq_puts(s, ",no_chk_data_crc");
444 if (c->mount_opts.override_compr) {
445 seq_printf(s, ",compr=%s",
446 ubifs_compr_name(c->mount_opts.compr_type));
449 return 0;
452 static int ubifs_sync_fs(struct super_block *sb, int wait)
454 int i, err;
455 struct ubifs_info *c = sb->s_fs_info;
458 * Zero @wait is just an advisory thing to help the file system shove
459 * lots of data into the queues, and there will be the second
460 * '->sync_fs()' call, with non-zero @wait.
462 if (!wait)
463 return 0;
466 * Synchronize write buffers, because 'ubifs_run_commit()' does not
467 * do this if it waits for an already running commit.
469 for (i = 0; i < c->jhead_cnt; i++) {
470 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
471 if (err)
472 return err;
476 * Strictly speaking, it is not necessary to commit the journal here,
477 * synchronizing write-buffers would be enough. But committing makes
478 * UBIFS free space predictions much more accurate, so we want to let
479 * the user be able to get more accurate results of 'statfs()' after
480 * they synchronize the file system.
482 err = ubifs_run_commit(c);
483 if (err)
484 return err;
486 return ubi_sync(c->vi.ubi_num);
490 * init_constants_early - initialize UBIFS constants.
491 * @c: UBIFS file-system description object
493 * This function initialize UBIFS constants which do not need the superblock to
494 * be read. It also checks that the UBI volume satisfies basic UBIFS
495 * requirements. Returns zero in case of success and a negative error code in
496 * case of failure.
498 static int init_constants_early(struct ubifs_info *c)
500 if (c->vi.corrupted) {
501 ubifs_warn(c, "UBI volume is corrupted - read-only mode");
502 c->ro_media = 1;
505 if (c->di.ro_mode) {
506 ubifs_msg(c, "read-only UBI device");
507 c->ro_media = 1;
510 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
511 ubifs_msg(c, "static UBI volume - read-only mode");
512 c->ro_media = 1;
515 c->leb_cnt = c->vi.size;
516 c->leb_size = c->vi.usable_leb_size;
517 c->leb_start = c->di.leb_start;
518 c->half_leb_size = c->leb_size / 2;
519 c->min_io_size = c->di.min_io_size;
520 c->min_io_shift = fls(c->min_io_size) - 1;
521 c->max_write_size = c->di.max_write_size;
522 c->max_write_shift = fls(c->max_write_size) - 1;
524 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
525 ubifs_errc(c, "too small LEBs (%d bytes), min. is %d bytes",
526 c->leb_size, UBIFS_MIN_LEB_SZ);
527 return -EINVAL;
530 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
531 ubifs_errc(c, "too few LEBs (%d), min. is %d",
532 c->leb_cnt, UBIFS_MIN_LEB_CNT);
533 return -EINVAL;
536 if (!is_power_of_2(c->min_io_size)) {
537 ubifs_errc(c, "bad min. I/O size %d", c->min_io_size);
538 return -EINVAL;
542 * Maximum write size has to be greater or equivalent to min. I/O
543 * size, and be multiple of min. I/O size.
545 if (c->max_write_size < c->min_io_size ||
546 c->max_write_size % c->min_io_size ||
547 !is_power_of_2(c->max_write_size)) {
548 ubifs_errc(c, "bad write buffer size %d for %d min. I/O unit",
549 c->max_write_size, c->min_io_size);
550 return -EINVAL;
554 * UBIFS aligns all node to 8-byte boundary, so to make function in
555 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
556 * less than 8.
558 if (c->min_io_size < 8) {
559 c->min_io_size = 8;
560 c->min_io_shift = 3;
561 if (c->max_write_size < c->min_io_size) {
562 c->max_write_size = c->min_io_size;
563 c->max_write_shift = c->min_io_shift;
567 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
568 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
571 * Initialize node length ranges which are mostly needed for node
572 * length validation.
574 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
575 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
576 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
577 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
578 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
579 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
581 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
582 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
583 c->ranges[UBIFS_ORPH_NODE].min_len =
584 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
585 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
586 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
587 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
588 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
589 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
590 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
591 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
593 * Minimum indexing node size is amended later when superblock is
594 * read and the key length is known.
596 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
598 * Maximum indexing node size is amended later when superblock is
599 * read and the fanout is known.
601 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
604 * Initialize dead and dark LEB space watermarks. See gc.c for comments
605 * about these values.
607 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
608 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
611 * Calculate how many bytes would be wasted at the end of LEB if it was
612 * fully filled with data nodes of maximum size. This is used in
613 * calculations when reporting free space.
615 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
617 /* Buffer size for bulk-reads */
618 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
619 if (c->max_bu_buf_len > c->leb_size)
620 c->max_bu_buf_len = c->leb_size;
621 return 0;
625 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
626 * @c: UBIFS file-system description object
627 * @lnum: LEB the write-buffer was synchronized to
628 * @free: how many free bytes left in this LEB
629 * @pad: how many bytes were padded
631 * This is a callback function which is called by the I/O unit when the
632 * write-buffer is synchronized. We need this to correctly maintain space
633 * accounting in bud logical eraseblocks. This function returns zero in case of
634 * success and a negative error code in case of failure.
636 * This function actually belongs to the journal, but we keep it here because
637 * we want to keep it static.
639 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
641 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
645 * init_constants_sb - initialize UBIFS constants.
646 * @c: UBIFS file-system description object
648 * This is a helper function which initializes various UBIFS constants after
649 * the superblock has been read. It also checks various UBIFS parameters and
650 * makes sure they are all right. Returns zero in case of success and a
651 * negative error code in case of failure.
653 static int init_constants_sb(struct ubifs_info *c)
655 int tmp, err;
656 long long tmp64;
658 c->main_bytes = (long long)c->main_lebs * c->leb_size;
659 c->max_znode_sz = sizeof(struct ubifs_znode) +
660 c->fanout * sizeof(struct ubifs_zbranch);
662 tmp = ubifs_idx_node_sz(c, 1);
663 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
664 c->min_idx_node_sz = ALIGN(tmp, 8);
666 tmp = ubifs_idx_node_sz(c, c->fanout);
667 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
668 c->max_idx_node_sz = ALIGN(tmp, 8);
670 /* Make sure LEB size is large enough to fit full commit */
671 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
672 tmp = ALIGN(tmp, c->min_io_size);
673 if (tmp > c->leb_size) {
674 ubifs_err(c, "too small LEB size %d, at least %d needed",
675 c->leb_size, tmp);
676 return -EINVAL;
680 * Make sure that the log is large enough to fit reference nodes for
681 * all buds plus one reserved LEB.
683 tmp64 = c->max_bud_bytes + c->leb_size - 1;
684 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
685 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
686 tmp /= c->leb_size;
687 tmp += 1;
688 if (c->log_lebs < tmp) {
689 ubifs_err(c, "too small log %d LEBs, required min. %d LEBs",
690 c->log_lebs, tmp);
691 return -EINVAL;
695 * When budgeting we assume worst-case scenarios when the pages are not
696 * be compressed and direntries are of the maximum size.
698 * Note, data, which may be stored in inodes is budgeted separately, so
699 * it is not included into 'c->bi.inode_budget'.
701 c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
702 c->bi.inode_budget = UBIFS_INO_NODE_SZ;
703 c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
706 * When the amount of flash space used by buds becomes
707 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
708 * The writers are unblocked when the commit is finished. To avoid
709 * writers to be blocked UBIFS initiates background commit in advance,
710 * when number of bud bytes becomes above the limit defined below.
712 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
715 * Ensure minimum journal size. All the bytes in the journal heads are
716 * considered to be used, when calculating the current journal usage.
717 * Consequently, if the journal is too small, UBIFS will treat it as
718 * always full.
720 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
721 if (c->bg_bud_bytes < tmp64)
722 c->bg_bud_bytes = tmp64;
723 if (c->max_bud_bytes < tmp64 + c->leb_size)
724 c->max_bud_bytes = tmp64 + c->leb_size;
726 err = ubifs_calc_lpt_geom(c);
727 if (err)
728 return err;
730 /* Initialize effective LEB size used in budgeting calculations */
731 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
732 return 0;
736 * init_constants_master - initialize UBIFS constants.
737 * @c: UBIFS file-system description object
739 * This is a helper function which initializes various UBIFS constants after
740 * the master node has been read. It also checks various UBIFS parameters and
741 * makes sure they are all right.
743 static void init_constants_master(struct ubifs_info *c)
745 long long tmp64;
747 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
748 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
751 * Calculate total amount of FS blocks. This number is not used
752 * internally because it does not make much sense for UBIFS, but it is
753 * necessary to report something for the 'statfs()' call.
755 * Subtract the LEB reserved for GC, the LEB which is reserved for
756 * deletions, minimum LEBs for the index, and assume only one journal
757 * head is available.
759 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
760 tmp64 *= (long long)c->leb_size - c->leb_overhead;
761 tmp64 = ubifs_reported_space(c, tmp64);
762 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
766 * take_gc_lnum - reserve GC LEB.
767 * @c: UBIFS file-system description object
769 * This function ensures that the LEB reserved for garbage collection is marked
770 * as "taken" in lprops. We also have to set free space to LEB size and dirty
771 * space to zero, because lprops may contain out-of-date information if the
772 * file-system was un-mounted before it has been committed. This function
773 * returns zero in case of success and a negative error code in case of
774 * failure.
776 static int take_gc_lnum(struct ubifs_info *c)
778 int err;
780 if (c->gc_lnum == -1) {
781 ubifs_err(c, "no LEB for GC");
782 return -EINVAL;
785 /* And we have to tell lprops that this LEB is taken */
786 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
787 LPROPS_TAKEN, 0, 0);
788 return err;
792 * alloc_wbufs - allocate write-buffers.
793 * @c: UBIFS file-system description object
795 * This helper function allocates and initializes UBIFS write-buffers. Returns
796 * zero in case of success and %-ENOMEM in case of failure.
798 static int alloc_wbufs(struct ubifs_info *c)
800 int i, err;
802 c->jheads = kcalloc(c->jhead_cnt, sizeof(struct ubifs_jhead),
803 GFP_KERNEL);
804 if (!c->jheads)
805 return -ENOMEM;
807 /* Initialize journal heads */
808 for (i = 0; i < c->jhead_cnt; i++) {
809 INIT_LIST_HEAD(&c->jheads[i].buds_list);
810 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
811 if (err)
812 return err;
814 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
815 c->jheads[i].wbuf.jhead = i;
816 c->jheads[i].grouped = 1;
820 * Garbage Collector head does not need to be synchronized by timer.
821 * Also GC head nodes are not grouped.
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 ubifs_err(c, "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 ubifs_bud *bud, *n;
881 rbtree_postorder_for_each_entry_safe(bud, n, &c->buds, rb)
882 kfree(bud);
886 * check_volume_empty - check if the UBI volume is empty.
887 * @c: UBIFS file-system description object
889 * This function checks if the UBIFS volume is empty by looking if its LEBs are
890 * mapped or not. The result of checking is stored in the @c->empty variable.
891 * Returns zero in case of success and a negative error code in case of
892 * failure.
894 static int check_volume_empty(struct ubifs_info *c)
896 int lnum, err;
898 c->empty = 1;
899 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
900 err = ubifs_is_mapped(c, lnum);
901 if (unlikely(err < 0))
902 return err;
903 if (err == 1) {
904 c->empty = 0;
905 break;
908 cond_resched();
911 return 0;
915 * UBIFS mount options.
917 * Opt_fast_unmount: do not run a journal commit before un-mounting
918 * Opt_norm_unmount: run a journal commit before un-mounting
919 * Opt_bulk_read: enable bulk-reads
920 * Opt_no_bulk_read: disable bulk-reads
921 * Opt_chk_data_crc: check CRCs when reading data nodes
922 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
923 * Opt_override_compr: override default compressor
924 * Opt_err: just end of array marker
926 enum {
927 Opt_fast_unmount,
928 Opt_norm_unmount,
929 Opt_bulk_read,
930 Opt_no_bulk_read,
931 Opt_chk_data_crc,
932 Opt_no_chk_data_crc,
933 Opt_override_compr,
934 Opt_err,
937 static const match_table_t tokens = {
938 {Opt_fast_unmount, "fast_unmount"},
939 {Opt_norm_unmount, "norm_unmount"},
940 {Opt_bulk_read, "bulk_read"},
941 {Opt_no_bulk_read, "no_bulk_read"},
942 {Opt_chk_data_crc, "chk_data_crc"},
943 {Opt_no_chk_data_crc, "no_chk_data_crc"},
944 {Opt_override_compr, "compr=%s"},
945 {Opt_err, NULL},
949 * parse_standard_option - parse a standard mount option.
950 * @option: the option to parse
952 * Normally, standard mount options like "sync" are passed to file-systems as
953 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
954 * be present in the options string. This function tries to deal with this
955 * situation and parse standard options. Returns 0 if the option was not
956 * recognized, and the corresponding integer flag if it was.
958 * UBIFS is only interested in the "sync" option, so do not check for anything
959 * else.
961 static int parse_standard_option(const char *option)
964 pr_notice("UBIFS: parse %s\n", option);
965 if (!strcmp(option, "sync"))
966 return MS_SYNCHRONOUS;
967 return 0;
971 * ubifs_parse_options - parse mount parameters.
972 * @c: UBIFS file-system description object
973 * @options: parameters to parse
974 * @is_remount: non-zero if this is FS re-mount
976 * This function parses UBIFS mount options and returns zero in case success
977 * and a negative error code in case of failure.
979 static int ubifs_parse_options(struct ubifs_info *c, char *options,
980 int is_remount)
982 char *p;
983 substring_t args[MAX_OPT_ARGS];
985 if (!options)
986 return 0;
988 while ((p = strsep(&options, ","))) {
989 int token;
991 if (!*p)
992 continue;
994 token = match_token(p, tokens, args);
995 switch (token) {
997 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
998 * We accept them in order to be backward-compatible. But this
999 * should be removed at some point.
1001 case Opt_fast_unmount:
1002 c->mount_opts.unmount_mode = 2;
1003 break;
1004 case Opt_norm_unmount:
1005 c->mount_opts.unmount_mode = 1;
1006 break;
1007 case Opt_bulk_read:
1008 c->mount_opts.bulk_read = 2;
1009 c->bulk_read = 1;
1010 break;
1011 case Opt_no_bulk_read:
1012 c->mount_opts.bulk_read = 1;
1013 c->bulk_read = 0;
1014 break;
1015 case Opt_chk_data_crc:
1016 c->mount_opts.chk_data_crc = 2;
1017 c->no_chk_data_crc = 0;
1018 break;
1019 case Opt_no_chk_data_crc:
1020 c->mount_opts.chk_data_crc = 1;
1021 c->no_chk_data_crc = 1;
1022 break;
1023 case Opt_override_compr:
1025 char *name = match_strdup(&args[0]);
1027 if (!name)
1028 return -ENOMEM;
1029 if (!strcmp(name, "none"))
1030 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1031 else if (!strcmp(name, "lzo"))
1032 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1033 else if (!strcmp(name, "zlib"))
1034 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1035 else {
1036 ubifs_err(c, "unknown compressor \"%s\"", name); //FIXME: is c ready?
1037 kfree(name);
1038 return -EINVAL;
1040 kfree(name);
1041 c->mount_opts.override_compr = 1;
1042 c->default_compr = c->mount_opts.compr_type;
1043 break;
1045 default:
1047 unsigned long flag;
1048 struct super_block *sb = c->vfs_sb;
1050 flag = parse_standard_option(p);
1051 if (!flag) {
1052 ubifs_err(c, "unrecognized mount option \"%s\" or missing value",
1054 return -EINVAL;
1056 sb->s_flags |= flag;
1057 break;
1062 return 0;
1066 * destroy_journal - destroy journal data structures.
1067 * @c: UBIFS file-system description object
1069 * This function destroys journal data structures including those that may have
1070 * been created by recovery functions.
1072 static void destroy_journal(struct ubifs_info *c)
1074 while (!list_empty(&c->unclean_leb_list)) {
1075 struct ubifs_unclean_leb *ucleb;
1077 ucleb = list_entry(c->unclean_leb_list.next,
1078 struct ubifs_unclean_leb, list);
1079 list_del(&ucleb->list);
1080 kfree(ucleb);
1082 while (!list_empty(&c->old_buds)) {
1083 struct ubifs_bud *bud;
1085 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1086 list_del(&bud->list);
1087 kfree(bud);
1089 ubifs_destroy_idx_gc(c);
1090 ubifs_destroy_size_tree(c);
1091 ubifs_tnc_close(c);
1092 free_buds(c);
1096 * bu_init - initialize bulk-read information.
1097 * @c: UBIFS file-system description object
1099 static void bu_init(struct ubifs_info *c)
1101 ubifs_assert(c->bulk_read == 1);
1103 if (c->bu.buf)
1104 return; /* Already initialized */
1106 again:
1107 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1108 if (!c->bu.buf) {
1109 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1110 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1111 goto again;
1114 /* Just disable bulk-read */
1115 ubifs_warn(c, "cannot allocate %d bytes of memory for bulk-read, disabling it",
1116 c->max_bu_buf_len);
1117 c->mount_opts.bulk_read = 1;
1118 c->bulk_read = 0;
1119 return;
1124 * check_free_space - check if there is enough free space to mount.
1125 * @c: UBIFS file-system description object
1127 * This function makes sure UBIFS has enough free space to be mounted in
1128 * read/write mode. UBIFS must always have some free space to allow deletions.
1130 static int check_free_space(struct ubifs_info *c)
1132 ubifs_assert(c->dark_wm > 0);
1133 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1134 ubifs_err(c, "insufficient free space to mount in R/W mode");
1135 ubifs_dump_budg(c, &c->bi);
1136 ubifs_dump_lprops(c);
1137 return -ENOSPC;
1139 return 0;
1143 * mount_ubifs - mount UBIFS file-system.
1144 * @c: UBIFS file-system description object
1146 * This function mounts UBIFS file system. Returns zero in case of success and
1147 * a negative error code in case of failure.
1149 static int mount_ubifs(struct ubifs_info *c)
1151 int err;
1152 long long x, y;
1153 size_t sz;
1155 c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1156 /* Suppress error messages while probing if MS_SILENT is set */
1157 c->probing = !!(c->vfs_sb->s_flags & MS_SILENT);
1159 err = init_constants_early(c);
1160 if (err)
1161 return err;
1163 err = ubifs_debugging_init(c);
1164 if (err)
1165 return err;
1167 err = check_volume_empty(c);
1168 if (err)
1169 goto out_free;
1171 if (c->empty && (c->ro_mount || c->ro_media)) {
1173 * This UBI volume is empty, and read-only, or the file system
1174 * is mounted read-only - we cannot format it.
1176 ubifs_err(c, "can't format empty UBI volume: read-only %s",
1177 c->ro_media ? "UBI volume" : "mount");
1178 err = -EROFS;
1179 goto out_free;
1182 if (c->ro_media && !c->ro_mount) {
1183 ubifs_err(c, "cannot mount read-write - read-only media");
1184 err = -EROFS;
1185 goto out_free;
1189 * The requirement for the buffer is that it should fit indexing B-tree
1190 * height amount of integers. We assume the height if the TNC tree will
1191 * never exceed 64.
1193 err = -ENOMEM;
1194 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1195 if (!c->bottom_up_buf)
1196 goto out_free;
1198 c->sbuf = vmalloc(c->leb_size);
1199 if (!c->sbuf)
1200 goto out_free;
1202 if (!c->ro_mount) {
1203 c->ileb_buf = vmalloc(c->leb_size);
1204 if (!c->ileb_buf)
1205 goto out_free;
1208 if (c->bulk_read == 1)
1209 bu_init(c);
1211 if (!c->ro_mount) {
1212 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ + \
1213 UBIFS_CIPHER_BLOCK_SIZE,
1214 GFP_KERNEL);
1215 if (!c->write_reserve_buf)
1216 goto out_free;
1219 c->mounting = 1;
1221 err = ubifs_read_superblock(c);
1222 if (err)
1223 goto out_free;
1225 c->probing = 0;
1228 * Make sure the compressor which is set as default in the superblock
1229 * or overridden by mount options is actually compiled in.
1231 if (!ubifs_compr_present(c->default_compr)) {
1232 ubifs_err(c, "'compressor \"%s\" is not compiled in",
1233 ubifs_compr_name(c->default_compr));
1234 err = -ENOTSUPP;
1235 goto out_free;
1238 err = init_constants_sb(c);
1239 if (err)
1240 goto out_free;
1242 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1243 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1244 c->cbuf = kmalloc(sz, GFP_NOFS);
1245 if (!c->cbuf) {
1246 err = -ENOMEM;
1247 goto out_free;
1250 err = alloc_wbufs(c);
1251 if (err)
1252 goto out_cbuf;
1254 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1255 if (!c->ro_mount) {
1256 /* Create background thread */
1257 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1258 if (IS_ERR(c->bgt)) {
1259 err = PTR_ERR(c->bgt);
1260 c->bgt = NULL;
1261 ubifs_err(c, "cannot spawn \"%s\", error %d",
1262 c->bgt_name, err);
1263 goto out_wbufs;
1265 wake_up_process(c->bgt);
1268 err = ubifs_read_master(c);
1269 if (err)
1270 goto out_master;
1272 init_constants_master(c);
1274 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1275 ubifs_msg(c, "recovery needed");
1276 c->need_recovery = 1;
1279 if (c->need_recovery && !c->ro_mount) {
1280 err = ubifs_recover_inl_heads(c, c->sbuf);
1281 if (err)
1282 goto out_master;
1285 err = ubifs_lpt_init(c, 1, !c->ro_mount);
1286 if (err)
1287 goto out_master;
1289 if (!c->ro_mount && c->space_fixup) {
1290 err = ubifs_fixup_free_space(c);
1291 if (err)
1292 goto out_lpt;
1295 if (!c->ro_mount && !c->need_recovery) {
1297 * Set the "dirty" flag so that if we reboot uncleanly we
1298 * will notice this immediately on the next mount.
1300 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1301 err = ubifs_write_master(c);
1302 if (err)
1303 goto out_lpt;
1306 err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1307 if (err)
1308 goto out_lpt;
1310 err = ubifs_replay_journal(c);
1311 if (err)
1312 goto out_journal;
1314 /* Calculate 'min_idx_lebs' after journal replay */
1315 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1317 err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1318 if (err)
1319 goto out_orphans;
1321 if (!c->ro_mount) {
1322 int lnum;
1324 err = check_free_space(c);
1325 if (err)
1326 goto out_orphans;
1328 /* Check for enough log space */
1329 lnum = c->lhead_lnum + 1;
1330 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1331 lnum = UBIFS_LOG_LNUM;
1332 if (lnum == c->ltail_lnum) {
1333 err = ubifs_consolidate_log(c);
1334 if (err)
1335 goto out_orphans;
1338 if (c->need_recovery) {
1339 err = ubifs_recover_size(c);
1340 if (err)
1341 goto out_orphans;
1342 err = ubifs_rcvry_gc_commit(c);
1343 if (err)
1344 goto out_orphans;
1345 } else {
1346 err = take_gc_lnum(c);
1347 if (err)
1348 goto out_orphans;
1351 * GC LEB may contain garbage if there was an unclean
1352 * reboot, and it should be un-mapped.
1354 err = ubifs_leb_unmap(c, c->gc_lnum);
1355 if (err)
1356 goto out_orphans;
1359 err = dbg_check_lprops(c);
1360 if (err)
1361 goto out_orphans;
1362 } else if (c->need_recovery) {
1363 err = ubifs_recover_size(c);
1364 if (err)
1365 goto out_orphans;
1366 } else {
1368 * Even if we mount read-only, we have to set space in GC LEB
1369 * to proper value because this affects UBIFS free space
1370 * reporting. We do not want to have a situation when
1371 * re-mounting from R/O to R/W changes amount of free space.
1373 err = take_gc_lnum(c);
1374 if (err)
1375 goto out_orphans;
1378 spin_lock(&ubifs_infos_lock);
1379 list_add_tail(&c->infos_list, &ubifs_infos);
1380 spin_unlock(&ubifs_infos_lock);
1382 if (c->need_recovery) {
1383 if (c->ro_mount)
1384 ubifs_msg(c, "recovery deferred");
1385 else {
1386 c->need_recovery = 0;
1387 ubifs_msg(c, "recovery completed");
1389 * GC LEB has to be empty and taken at this point. But
1390 * the journal head LEBs may also be accounted as
1391 * "empty taken" if they are empty.
1393 ubifs_assert(c->lst.taken_empty_lebs > 0);
1395 } else
1396 ubifs_assert(c->lst.taken_empty_lebs > 0);
1398 err = dbg_check_filesystem(c);
1399 if (err)
1400 goto out_infos;
1402 err = dbg_debugfs_init_fs(c);
1403 if (err)
1404 goto out_infos;
1406 c->mounting = 0;
1408 ubifs_msg(c, "UBIFS: mounted UBI device %d, volume %d, name \"%s\"%s",
1409 c->vi.ubi_num, c->vi.vol_id, c->vi.name,
1410 c->ro_mount ? ", R/O mode" : "");
1411 x = (long long)c->main_lebs * c->leb_size;
1412 y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1413 ubifs_msg(c, "LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes",
1414 c->leb_size, c->leb_size >> 10, c->min_io_size,
1415 c->max_write_size);
1416 ubifs_msg(c, "FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)",
1417 x, x >> 20, c->main_lebs,
1418 y, y >> 20, c->log_lebs + c->max_bud_cnt);
1419 ubifs_msg(c, "reserved for root: %llu bytes (%llu KiB)",
1420 c->report_rp_size, c->report_rp_size >> 10);
1421 ubifs_msg(c, "media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s",
1422 c->fmt_version, c->ro_compat_version,
1423 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid,
1424 c->big_lpt ? ", big LPT model" : ", small LPT model");
1426 dbg_gen("default compressor: %s", ubifs_compr_name(c->default_compr));
1427 dbg_gen("data journal heads: %d",
1428 c->jhead_cnt - NONDATA_JHEADS_CNT);
1429 dbg_gen("log LEBs: %d (%d - %d)",
1430 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1431 dbg_gen("LPT area LEBs: %d (%d - %d)",
1432 c->lpt_lebs, c->lpt_first, c->lpt_last);
1433 dbg_gen("orphan area LEBs: %d (%d - %d)",
1434 c->orph_lebs, c->orph_first, c->orph_last);
1435 dbg_gen("main area LEBs: %d (%d - %d)",
1436 c->main_lebs, c->main_first, c->leb_cnt - 1);
1437 dbg_gen("index LEBs: %d", c->lst.idx_lebs);
1438 dbg_gen("total index bytes: %lld (%lld KiB, %lld MiB)",
1439 c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1440 c->bi.old_idx_sz >> 20);
1441 dbg_gen("key hash type: %d", c->key_hash_type);
1442 dbg_gen("tree fanout: %d", c->fanout);
1443 dbg_gen("reserved GC LEB: %d", c->gc_lnum);
1444 dbg_gen("max. znode size %d", c->max_znode_sz);
1445 dbg_gen("max. index node size %d", c->max_idx_node_sz);
1446 dbg_gen("node sizes: data %zu, inode %zu, dentry %zu",
1447 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1448 dbg_gen("node sizes: trun %zu, sb %zu, master %zu",
1449 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1450 dbg_gen("node sizes: ref %zu, cmt. start %zu, orph %zu",
1451 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1452 dbg_gen("max. node sizes: data %zu, inode %zu dentry %zu, idx %d",
1453 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1454 UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1455 dbg_gen("dead watermark: %d", c->dead_wm);
1456 dbg_gen("dark watermark: %d", c->dark_wm);
1457 dbg_gen("LEB overhead: %d", c->leb_overhead);
1458 x = (long long)c->main_lebs * c->dark_wm;
1459 dbg_gen("max. dark space: %lld (%lld KiB, %lld MiB)",
1460 x, x >> 10, x >> 20);
1461 dbg_gen("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1462 c->max_bud_bytes, c->max_bud_bytes >> 10,
1463 c->max_bud_bytes >> 20);
1464 dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1465 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1466 c->bg_bud_bytes >> 20);
1467 dbg_gen("current bud bytes %lld (%lld KiB, %lld MiB)",
1468 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1469 dbg_gen("max. seq. number: %llu", c->max_sqnum);
1470 dbg_gen("commit number: %llu", c->cmt_no);
1472 return 0;
1474 out_infos:
1475 spin_lock(&ubifs_infos_lock);
1476 list_del(&c->infos_list);
1477 spin_unlock(&ubifs_infos_lock);
1478 out_orphans:
1479 free_orphans(c);
1480 out_journal:
1481 destroy_journal(c);
1482 out_lpt:
1483 ubifs_lpt_free(c, 0);
1484 out_master:
1485 kfree(c->mst_node);
1486 kfree(c->rcvrd_mst_node);
1487 if (c->bgt)
1488 kthread_stop(c->bgt);
1489 out_wbufs:
1490 free_wbufs(c);
1491 out_cbuf:
1492 kfree(c->cbuf);
1493 out_free:
1494 kfree(c->write_reserve_buf);
1495 kfree(c->bu.buf);
1496 vfree(c->ileb_buf);
1497 vfree(c->sbuf);
1498 kfree(c->bottom_up_buf);
1499 ubifs_debugging_exit(c);
1500 return err;
1504 * ubifs_umount - un-mount UBIFS file-system.
1505 * @c: UBIFS file-system description object
1507 * Note, this function is called to free allocated resourced when un-mounting,
1508 * as well as free resources when an error occurred while we were half way
1509 * through mounting (error path cleanup function). So it has to make sure the
1510 * resource was actually allocated before freeing it.
1512 static void ubifs_umount(struct ubifs_info *c)
1514 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1515 c->vi.vol_id);
1517 dbg_debugfs_exit_fs(c);
1518 spin_lock(&ubifs_infos_lock);
1519 list_del(&c->infos_list);
1520 spin_unlock(&ubifs_infos_lock);
1522 if (c->bgt)
1523 kthread_stop(c->bgt);
1525 destroy_journal(c);
1526 free_wbufs(c);
1527 free_orphans(c);
1528 ubifs_lpt_free(c, 0);
1530 kfree(c->cbuf);
1531 kfree(c->rcvrd_mst_node);
1532 kfree(c->mst_node);
1533 kfree(c->write_reserve_buf);
1534 kfree(c->bu.buf);
1535 vfree(c->ileb_buf);
1536 vfree(c->sbuf);
1537 kfree(c->bottom_up_buf);
1538 ubifs_debugging_exit(c);
1542 * ubifs_remount_rw - re-mount in read-write mode.
1543 * @c: UBIFS file-system description object
1545 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1546 * mode. This function allocates the needed resources and re-mounts UBIFS in
1547 * read-write mode.
1549 static int ubifs_remount_rw(struct ubifs_info *c)
1551 int err, lnum;
1553 if (c->rw_incompat) {
1554 ubifs_err(c, "the file-system is not R/W-compatible");
1555 ubifs_msg(c, "on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
1556 c->fmt_version, c->ro_compat_version,
1557 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1558 return -EROFS;
1561 mutex_lock(&c->umount_mutex);
1562 dbg_save_space_info(c);
1563 c->remounting_rw = 1;
1564 c->ro_mount = 0;
1566 if (c->space_fixup) {
1567 err = ubifs_fixup_free_space(c);
1568 if (err)
1569 goto out;
1572 err = check_free_space(c);
1573 if (err)
1574 goto out;
1576 if (c->old_leb_cnt != c->leb_cnt) {
1577 struct ubifs_sb_node *sup;
1579 sup = ubifs_read_sb_node(c);
1580 if (IS_ERR(sup)) {
1581 err = PTR_ERR(sup);
1582 goto out;
1584 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1585 err = ubifs_write_sb_node(c, sup);
1586 kfree(sup);
1587 if (err)
1588 goto out;
1591 if (c->need_recovery) {
1592 ubifs_msg(c, "completing deferred recovery");
1593 err = ubifs_write_rcvrd_mst_node(c);
1594 if (err)
1595 goto out;
1596 err = ubifs_recover_size(c);
1597 if (err)
1598 goto out;
1599 err = ubifs_clean_lebs(c, c->sbuf);
1600 if (err)
1601 goto out;
1602 err = ubifs_recover_inl_heads(c, c->sbuf);
1603 if (err)
1604 goto out;
1605 } else {
1606 /* A readonly mount is not allowed to have orphans */
1607 ubifs_assert(c->tot_orphans == 0);
1608 err = ubifs_clear_orphans(c);
1609 if (err)
1610 goto out;
1613 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1614 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1615 err = ubifs_write_master(c);
1616 if (err)
1617 goto out;
1620 c->ileb_buf = vmalloc(c->leb_size);
1621 if (!c->ileb_buf) {
1622 err = -ENOMEM;
1623 goto out;
1626 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ + \
1627 UBIFS_CIPHER_BLOCK_SIZE, GFP_KERNEL);
1628 if (!c->write_reserve_buf) {
1629 err = -ENOMEM;
1630 goto out;
1633 err = ubifs_lpt_init(c, 0, 1);
1634 if (err)
1635 goto out;
1637 /* Create background thread */
1638 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1639 if (IS_ERR(c->bgt)) {
1640 err = PTR_ERR(c->bgt);
1641 c->bgt = NULL;
1642 ubifs_err(c, "cannot spawn \"%s\", error %d",
1643 c->bgt_name, err);
1644 goto out;
1646 wake_up_process(c->bgt);
1648 c->orph_buf = vmalloc(c->leb_size);
1649 if (!c->orph_buf) {
1650 err = -ENOMEM;
1651 goto out;
1654 /* Check for enough log space */
1655 lnum = c->lhead_lnum + 1;
1656 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1657 lnum = UBIFS_LOG_LNUM;
1658 if (lnum == c->ltail_lnum) {
1659 err = ubifs_consolidate_log(c);
1660 if (err)
1661 goto out;
1664 if (c->need_recovery)
1665 err = ubifs_rcvry_gc_commit(c);
1666 else
1667 err = ubifs_leb_unmap(c, c->gc_lnum);
1668 if (err)
1669 goto out;
1671 dbg_gen("re-mounted read-write");
1672 c->remounting_rw = 0;
1674 if (c->need_recovery) {
1675 c->need_recovery = 0;
1676 ubifs_msg(c, "deferred recovery completed");
1677 } else {
1679 * Do not run the debugging space check if the were doing
1680 * recovery, because when we saved the information we had the
1681 * file-system in a state where the TNC and lprops has been
1682 * modified in memory, but all the I/O operations (including a
1683 * commit) were deferred. So the file-system was in
1684 * "non-committed" state. Now the file-system is in committed
1685 * state, and of course the amount of free space will change
1686 * because, for example, the old index size was imprecise.
1688 err = dbg_check_space_info(c);
1691 mutex_unlock(&c->umount_mutex);
1692 return err;
1694 out:
1695 c->ro_mount = 1;
1696 vfree(c->orph_buf);
1697 c->orph_buf = NULL;
1698 if (c->bgt) {
1699 kthread_stop(c->bgt);
1700 c->bgt = NULL;
1702 free_wbufs(c);
1703 kfree(c->write_reserve_buf);
1704 c->write_reserve_buf = NULL;
1705 vfree(c->ileb_buf);
1706 c->ileb_buf = NULL;
1707 ubifs_lpt_free(c, 1);
1708 c->remounting_rw = 0;
1709 mutex_unlock(&c->umount_mutex);
1710 return err;
1714 * ubifs_remount_ro - re-mount in read-only mode.
1715 * @c: UBIFS file-system description object
1717 * We assume VFS has stopped writing. Possibly the background thread could be
1718 * running a commit, however kthread_stop will wait in that case.
1720 static void ubifs_remount_ro(struct ubifs_info *c)
1722 int i, err;
1724 ubifs_assert(!c->need_recovery);
1725 ubifs_assert(!c->ro_mount);
1727 mutex_lock(&c->umount_mutex);
1728 if (c->bgt) {
1729 kthread_stop(c->bgt);
1730 c->bgt = NULL;
1733 dbg_save_space_info(c);
1735 for (i = 0; i < c->jhead_cnt; i++)
1736 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1738 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1739 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1740 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1741 err = ubifs_write_master(c);
1742 if (err)
1743 ubifs_ro_mode(c, err);
1745 vfree(c->orph_buf);
1746 c->orph_buf = NULL;
1747 kfree(c->write_reserve_buf);
1748 c->write_reserve_buf = NULL;
1749 vfree(c->ileb_buf);
1750 c->ileb_buf = NULL;
1751 ubifs_lpt_free(c, 1);
1752 c->ro_mount = 1;
1753 err = dbg_check_space_info(c);
1754 if (err)
1755 ubifs_ro_mode(c, err);
1756 mutex_unlock(&c->umount_mutex);
1759 static void ubifs_put_super(struct super_block *sb)
1761 int i;
1762 struct ubifs_info *c = sb->s_fs_info;
1764 ubifs_msg(c, "un-mount UBI device %d", c->vi.ubi_num);
1767 * The following asserts are only valid if there has not been a failure
1768 * of the media. For example, there will be dirty inodes if we failed
1769 * to write them back because of I/O errors.
1771 if (!c->ro_error) {
1772 ubifs_assert(c->bi.idx_growth == 0);
1773 ubifs_assert(c->bi.dd_growth == 0);
1774 ubifs_assert(c->bi.data_growth == 0);
1778 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1779 * and file system un-mount. Namely, it prevents the shrinker from
1780 * picking this superblock for shrinking - it will be just skipped if
1781 * the mutex is locked.
1783 mutex_lock(&c->umount_mutex);
1784 if (!c->ro_mount) {
1786 * First of all kill the background thread to make sure it does
1787 * not interfere with un-mounting and freeing resources.
1789 if (c->bgt) {
1790 kthread_stop(c->bgt);
1791 c->bgt = NULL;
1795 * On fatal errors c->ro_error is set to 1, in which case we do
1796 * not write the master node.
1798 if (!c->ro_error) {
1799 int err;
1801 /* Synchronize write-buffers */
1802 for (i = 0; i < c->jhead_cnt; i++)
1803 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1806 * We are being cleanly unmounted which means the
1807 * orphans were killed - indicate this in the master
1808 * node. Also save the reserved GC LEB number.
1810 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1811 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1812 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1813 err = ubifs_write_master(c);
1814 if (err)
1816 * Recovery will attempt to fix the master area
1817 * next mount, so we just print a message and
1818 * continue to unmount normally.
1820 ubifs_err(c, "failed to write master node, error %d",
1821 err);
1822 } else {
1823 for (i = 0; i < c->jhead_cnt; i++)
1824 /* Make sure write-buffer timers are canceled */
1825 hrtimer_cancel(&c->jheads[i].wbuf.timer);
1829 ubifs_umount(c);
1830 bdi_destroy(&c->bdi);
1831 ubi_close_volume(c->ubi);
1832 mutex_unlock(&c->umount_mutex);
1835 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1837 int err;
1838 struct ubifs_info *c = sb->s_fs_info;
1840 sync_filesystem(sb);
1841 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1843 err = ubifs_parse_options(c, data, 1);
1844 if (err) {
1845 ubifs_err(c, "invalid or unknown remount parameter");
1846 return err;
1849 if (c->ro_mount && !(*flags & MS_RDONLY)) {
1850 if (c->ro_error) {
1851 ubifs_msg(c, "cannot re-mount R/W due to prior errors");
1852 return -EROFS;
1854 if (c->ro_media) {
1855 ubifs_msg(c, "cannot re-mount R/W - UBI volume is R/O");
1856 return -EROFS;
1858 err = ubifs_remount_rw(c);
1859 if (err)
1860 return err;
1861 } else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1862 if (c->ro_error) {
1863 ubifs_msg(c, "cannot re-mount R/O due to prior errors");
1864 return -EROFS;
1866 ubifs_remount_ro(c);
1869 if (c->bulk_read == 1)
1870 bu_init(c);
1871 else {
1872 dbg_gen("disable bulk-read");
1873 kfree(c->bu.buf);
1874 c->bu.buf = NULL;
1877 ubifs_assert(c->lst.taken_empty_lebs > 0);
1878 return 0;
1881 const struct super_operations ubifs_super_operations = {
1882 .alloc_inode = ubifs_alloc_inode,
1883 .destroy_inode = ubifs_destroy_inode,
1884 .put_super = ubifs_put_super,
1885 .write_inode = ubifs_write_inode,
1886 .evict_inode = ubifs_evict_inode,
1887 .statfs = ubifs_statfs,
1888 .dirty_inode = ubifs_dirty_inode,
1889 .remount_fs = ubifs_remount_fs,
1890 .show_options = ubifs_show_options,
1891 .sync_fs = ubifs_sync_fs,
1895 * open_ubi - parse UBI device name string and open the UBI device.
1896 * @name: UBI volume name
1897 * @mode: UBI volume open mode
1899 * The primary method of mounting UBIFS is by specifying the UBI volume
1900 * character device node path. However, UBIFS may also be mounted withoug any
1901 * character device node using one of the following methods:
1903 * o ubiX_Y - mount UBI device number X, volume Y;
1904 * o ubiY - mount UBI device number 0, volume Y;
1905 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1906 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1908 * Alternative '!' separator may be used instead of ':' (because some shells
1909 * like busybox may interpret ':' as an NFS host name separator). This function
1910 * returns UBI volume description object in case of success and a negative
1911 * error code in case of failure.
1913 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1915 struct ubi_volume_desc *ubi;
1916 int dev, vol;
1917 char *endptr;
1919 /* First, try to open using the device node path method */
1920 ubi = ubi_open_volume_path(name, mode);
1921 if (!IS_ERR(ubi))
1922 return ubi;
1924 /* Try the "nodev" method */
1925 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1926 return ERR_PTR(-EINVAL);
1928 /* ubi:NAME method */
1929 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1930 return ubi_open_volume_nm(0, name + 4, mode);
1932 if (!isdigit(name[3]))
1933 return ERR_PTR(-EINVAL);
1935 dev = simple_strtoul(name + 3, &endptr, 0);
1937 /* ubiY method */
1938 if (*endptr == '\0')
1939 return ubi_open_volume(0, dev, mode);
1941 /* ubiX_Y method */
1942 if (*endptr == '_' && isdigit(endptr[1])) {
1943 vol = simple_strtoul(endptr + 1, &endptr, 0);
1944 if (*endptr != '\0')
1945 return ERR_PTR(-EINVAL);
1946 return ubi_open_volume(dev, vol, mode);
1949 /* ubiX:NAME method */
1950 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1951 return ubi_open_volume_nm(dev, ++endptr, mode);
1953 return ERR_PTR(-EINVAL);
1956 static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1958 struct ubifs_info *c;
1960 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1961 if (c) {
1962 spin_lock_init(&c->cnt_lock);
1963 spin_lock_init(&c->cs_lock);
1964 spin_lock_init(&c->buds_lock);
1965 spin_lock_init(&c->space_lock);
1966 spin_lock_init(&c->orphan_lock);
1967 init_rwsem(&c->commit_sem);
1968 mutex_init(&c->lp_mutex);
1969 mutex_init(&c->tnc_mutex);
1970 mutex_init(&c->log_mutex);
1971 mutex_init(&c->umount_mutex);
1972 mutex_init(&c->bu_mutex);
1973 mutex_init(&c->write_reserve_mutex);
1974 init_waitqueue_head(&c->cmt_wq);
1975 c->buds = RB_ROOT;
1976 c->old_idx = RB_ROOT;
1977 c->size_tree = RB_ROOT;
1978 c->orph_tree = RB_ROOT;
1979 INIT_LIST_HEAD(&c->infos_list);
1980 INIT_LIST_HEAD(&c->idx_gc);
1981 INIT_LIST_HEAD(&c->replay_list);
1982 INIT_LIST_HEAD(&c->replay_buds);
1983 INIT_LIST_HEAD(&c->uncat_list);
1984 INIT_LIST_HEAD(&c->empty_list);
1985 INIT_LIST_HEAD(&c->freeable_list);
1986 INIT_LIST_HEAD(&c->frdi_idx_list);
1987 INIT_LIST_HEAD(&c->unclean_leb_list);
1988 INIT_LIST_HEAD(&c->old_buds);
1989 INIT_LIST_HEAD(&c->orph_list);
1990 INIT_LIST_HEAD(&c->orph_new);
1991 c->no_chk_data_crc = 1;
1993 c->highest_inum = UBIFS_FIRST_INO;
1994 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1996 ubi_get_volume_info(ubi, &c->vi);
1997 ubi_get_device_info(c->vi.ubi_num, &c->di);
1999 return c;
2002 #ifndef CONFIG_UBIFS_FS_ENCRYPTION
2003 struct fscrypt_operations ubifs_crypt_operations = {
2004 .is_encrypted = __ubifs_crypt_is_encrypted,
2006 #endif
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 = 0;
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;
2053 sb->s_xattr = ubifs_xattr_handlers;
2054 sb->s_cop = &ubifs_crypt_operations;
2056 mutex_lock(&c->umount_mutex);
2057 err = mount_ubifs(c);
2058 if (err) {
2059 ubifs_assert(err < 0);
2060 goto out_unlock;
2063 /* Read the root inode */
2064 root = ubifs_iget(sb, UBIFS_ROOT_INO);
2065 if (IS_ERR(root)) {
2066 err = PTR_ERR(root);
2067 goto out_umount;
2070 sb->s_root = d_make_root(root);
2071 if (!sb->s_root) {
2072 err = -ENOMEM;
2073 goto out_umount;
2076 mutex_unlock(&c->umount_mutex);
2077 return 0;
2079 out_umount:
2080 ubifs_umount(c);
2081 out_unlock:
2082 mutex_unlock(&c->umount_mutex);
2083 out_bdi:
2084 bdi_destroy(&c->bdi);
2085 out_close:
2086 ubi_close_volume(c->ubi);
2087 out:
2088 return err;
2091 static int sb_test(struct super_block *sb, void *data)
2093 struct ubifs_info *c1 = data;
2094 struct ubifs_info *c = sb->s_fs_info;
2096 return c->vi.cdev == c1->vi.cdev;
2099 static int sb_set(struct super_block *sb, void *data)
2101 sb->s_fs_info = data;
2102 return set_anon_super(sb, NULL);
2105 static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2106 const char *name, void *data)
2108 struct ubi_volume_desc *ubi;
2109 struct ubifs_info *c;
2110 struct super_block *sb;
2111 int err;
2113 dbg_gen("name %s, flags %#x", name, flags);
2116 * Get UBI device number and volume ID. Mount it read-only so far
2117 * because this might be a new mount point, and UBI allows only one
2118 * read-write user at a time.
2120 ubi = open_ubi(name, UBI_READONLY);
2121 if (IS_ERR(ubi)) {
2122 if (!(flags & MS_SILENT))
2123 pr_err("UBIFS error (pid: %d): cannot open \"%s\", error %d",
2124 current->pid, name, (int)PTR_ERR(ubi));
2125 return ERR_CAST(ubi);
2128 c = alloc_ubifs_info(ubi);
2129 if (!c) {
2130 err = -ENOMEM;
2131 goto out_close;
2134 dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2136 sb = sget(fs_type, sb_test, sb_set, flags, c);
2137 if (IS_ERR(sb)) {
2138 err = PTR_ERR(sb);
2139 kfree(c);
2140 goto out_close;
2143 if (sb->s_root) {
2144 struct ubifs_info *c1 = sb->s_fs_info;
2145 kfree(c);
2146 /* A new mount point for already mounted UBIFS */
2147 dbg_gen("this ubi volume is already mounted");
2148 if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2149 err = -EBUSY;
2150 goto out_deact;
2152 } else {
2153 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2154 if (err)
2155 goto out_deact;
2156 /* We do not support atime */
2157 sb->s_flags |= MS_ACTIVE;
2158 #ifndef CONFIG_UBIFS_ATIME_SUPPORT
2159 sb->s_flags |= MS_NOATIME;
2160 #else
2161 ubifs_msg(c, "full atime support is enabled.");
2162 #endif
2165 /* 'fill_super()' opens ubi again so we must close it here */
2166 ubi_close_volume(ubi);
2168 return dget(sb->s_root);
2170 out_deact:
2171 deactivate_locked_super(sb);
2172 out_close:
2173 ubi_close_volume(ubi);
2174 return ERR_PTR(err);
2177 static void kill_ubifs_super(struct super_block *s)
2179 struct ubifs_info *c = s->s_fs_info;
2180 kill_anon_super(s);
2181 kfree(c);
2184 static struct file_system_type ubifs_fs_type = {
2185 .name = "ubifs",
2186 .owner = THIS_MODULE,
2187 .mount = ubifs_mount,
2188 .kill_sb = kill_ubifs_super,
2190 MODULE_ALIAS_FS("ubifs");
2193 * Inode slab cache constructor.
2195 static void inode_slab_ctor(void *obj)
2197 struct ubifs_inode *ui = obj;
2198 inode_init_once(&ui->vfs_inode);
2201 static int __init ubifs_init(void)
2203 int err;
2205 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2207 /* Make sure node sizes are 8-byte aligned */
2208 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2209 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2210 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2211 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2212 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2213 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2214 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2215 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2216 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2217 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2218 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2220 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2221 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2222 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2223 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2224 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2225 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2227 /* Check min. node size */
2228 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2229 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2230 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2231 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2233 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2234 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2235 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2236 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2238 /* Defined node sizes */
2239 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2240 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2241 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2242 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2245 * We use 2 bit wide bit-fields to store compression type, which should
2246 * be amended if more compressors are added. The bit-fields are:
2247 * @compr_type in 'struct ubifs_inode', @default_compr in
2248 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2250 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2253 * We require that PAGE_SIZE is greater-than-or-equal-to
2254 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2256 if (PAGE_SIZE < UBIFS_BLOCK_SIZE) {
2257 pr_err("UBIFS error (pid %d): VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
2258 current->pid, (unsigned int)PAGE_SIZE);
2259 return -EINVAL;
2262 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2263 sizeof(struct ubifs_inode), 0,
2264 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT |
2265 SLAB_ACCOUNT, &inode_slab_ctor);
2266 if (!ubifs_inode_slab)
2267 return -ENOMEM;
2269 err = register_shrinker(&ubifs_shrinker_info);
2270 if (err)
2271 goto out_slab;
2273 err = ubifs_compressors_init();
2274 if (err)
2275 goto out_shrinker;
2277 err = dbg_debugfs_init();
2278 if (err)
2279 goto out_compr;
2281 err = register_filesystem(&ubifs_fs_type);
2282 if (err) {
2283 pr_err("UBIFS error (pid %d): cannot register file system, error %d",
2284 current->pid, err);
2285 goto out_dbg;
2287 return 0;
2289 out_dbg:
2290 dbg_debugfs_exit();
2291 out_compr:
2292 ubifs_compressors_exit();
2293 out_shrinker:
2294 unregister_shrinker(&ubifs_shrinker_info);
2295 out_slab:
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);
2312 * Make sure all delayed rcu free inodes are flushed before we
2313 * destroy cache.
2315 rcu_barrier();
2316 kmem_cache_destroy(ubifs_inode_slab);
2317 unregister_filesystem(&ubifs_fs_type);
2319 module_exit(ubifs_exit);
2321 MODULE_LICENSE("GPL");
2322 MODULE_VERSION(__stringify(UBIFS_VERSION));
2323 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2324 MODULE_DESCRIPTION("UBIFS - UBI File System");