Linux 2.6.31.6
[linux/fpc-iii.git] / fs / ubifs / super.c
blob26d2e0d8046598ed579a400293dbebd38eefc7e3
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 <linux/smp_lock.h>
40 #include "ubifs.h"
43 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
44 * allocating too much.
46 #define UBIFS_KMALLOC_OK (128*1024)
48 /* Slab cache for UBIFS inodes */
49 struct kmem_cache *ubifs_inode_slab;
51 /* UBIFS TNC shrinker description */
52 static struct shrinker ubifs_shrinker_info = {
53 .shrink = ubifs_shrinker,
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 && (inode->i_mode & S_IFMT) != S_IFREG)
90 return 5;
92 if (!ubifs_compr_present(ui->compr_type)) {
93 ubifs_warn("inode %lu uses '%s' compression, but it was not "
94 "compiled in", inode->i_ino,
95 ubifs_compr_name(ui->compr_type));
98 err = dbg_check_dir_size(c, inode);
99 return err;
102 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
104 int err;
105 union ubifs_key key;
106 struct ubifs_ino_node *ino;
107 struct ubifs_info *c = sb->s_fs_info;
108 struct inode *inode;
109 struct ubifs_inode *ui;
111 dbg_gen("inode %lu", inum);
113 inode = iget_locked(sb, inum);
114 if (!inode)
115 return ERR_PTR(-ENOMEM);
116 if (!(inode->i_state & I_NEW))
117 return inode;
118 ui = ubifs_inode(inode);
120 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
121 if (!ino) {
122 err = -ENOMEM;
123 goto out;
126 ino_key_init(c, &key, inode->i_ino);
128 err = ubifs_tnc_lookup(c, &key, ino);
129 if (err)
130 goto out_ino;
132 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
133 inode->i_nlink = le32_to_cpu(ino->nlink);
134 inode->i_uid = le32_to_cpu(ino->uid);
135 inode->i_gid = le32_to_cpu(ino->gid);
136 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
137 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
138 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
139 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
140 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
141 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
142 inode->i_mode = le32_to_cpu(ino->mode);
143 inode->i_size = le64_to_cpu(ino->size);
145 ui->data_len = le32_to_cpu(ino->data_len);
146 ui->flags = le32_to_cpu(ino->flags);
147 ui->compr_type = le16_to_cpu(ino->compr_type);
148 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
149 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
150 ui->xattr_size = le32_to_cpu(ino->xattr_size);
151 ui->xattr_names = le32_to_cpu(ino->xattr_names);
152 ui->synced_i_size = ui->ui_size = inode->i_size;
154 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
156 err = validate_inode(c, inode);
157 if (err)
158 goto out_invalid;
160 /* Disable read-ahead */
161 inode->i_mapping->backing_dev_info = &c->bdi;
163 switch (inode->i_mode & S_IFMT) {
164 case S_IFREG:
165 inode->i_mapping->a_ops = &ubifs_file_address_operations;
166 inode->i_op = &ubifs_file_inode_operations;
167 inode->i_fop = &ubifs_file_operations;
168 if (ui->xattr) {
169 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
170 if (!ui->data) {
171 err = -ENOMEM;
172 goto out_ino;
174 memcpy(ui->data, ino->data, ui->data_len);
175 ((char *)ui->data)[ui->data_len] = '\0';
176 } else if (ui->data_len != 0) {
177 err = 10;
178 goto out_invalid;
180 break;
181 case S_IFDIR:
182 inode->i_op = &ubifs_dir_inode_operations;
183 inode->i_fop = &ubifs_dir_operations;
184 if (ui->data_len != 0) {
185 err = 11;
186 goto out_invalid;
188 break;
189 case S_IFLNK:
190 inode->i_op = &ubifs_symlink_inode_operations;
191 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
192 err = 12;
193 goto out_invalid;
195 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
196 if (!ui->data) {
197 err = -ENOMEM;
198 goto out_ino;
200 memcpy(ui->data, ino->data, ui->data_len);
201 ((char *)ui->data)[ui->data_len] = '\0';
202 break;
203 case S_IFBLK:
204 case S_IFCHR:
206 dev_t rdev;
207 union ubifs_dev_desc *dev;
209 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
210 if (!ui->data) {
211 err = -ENOMEM;
212 goto out_ino;
215 dev = (union ubifs_dev_desc *)ino->data;
216 if (ui->data_len == sizeof(dev->new))
217 rdev = new_decode_dev(le32_to_cpu(dev->new));
218 else if (ui->data_len == sizeof(dev->huge))
219 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
220 else {
221 err = 13;
222 goto out_invalid;
224 memcpy(ui->data, ino->data, ui->data_len);
225 inode->i_op = &ubifs_file_inode_operations;
226 init_special_inode(inode, inode->i_mode, rdev);
227 break;
229 case S_IFSOCK:
230 case S_IFIFO:
231 inode->i_op = &ubifs_file_inode_operations;
232 init_special_inode(inode, inode->i_mode, 0);
233 if (ui->data_len != 0) {
234 err = 14;
235 goto out_invalid;
237 break;
238 default:
239 err = 15;
240 goto out_invalid;
243 kfree(ino);
244 ubifs_set_inode_flags(inode);
245 unlock_new_inode(inode);
246 return inode;
248 out_invalid:
249 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
250 dbg_dump_node(c, ino);
251 dbg_dump_inode(c, inode);
252 err = -EINVAL;
253 out_ino:
254 kfree(ino);
255 out:
256 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
257 iget_failed(inode);
258 return ERR_PTR(err);
261 static struct inode *ubifs_alloc_inode(struct super_block *sb)
263 struct ubifs_inode *ui;
265 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
266 if (!ui)
267 return NULL;
269 memset((void *)ui + sizeof(struct inode), 0,
270 sizeof(struct ubifs_inode) - sizeof(struct inode));
271 mutex_init(&ui->ui_mutex);
272 spin_lock_init(&ui->ui_lock);
273 return &ui->vfs_inode;
276 static void ubifs_destroy_inode(struct inode *inode)
278 struct ubifs_inode *ui = ubifs_inode(inode);
280 kfree(ui->data);
281 kmem_cache_free(ubifs_inode_slab, inode);
285 * Note, Linux write-back code calls this without 'i_mutex'.
287 static int ubifs_write_inode(struct inode *inode, int wait)
289 int err = 0;
290 struct ubifs_info *c = inode->i_sb->s_fs_info;
291 struct ubifs_inode *ui = ubifs_inode(inode);
293 ubifs_assert(!ui->xattr);
294 if (is_bad_inode(inode))
295 return 0;
297 mutex_lock(&ui->ui_mutex);
299 * Due to races between write-back forced by budgeting
300 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
301 * have already been synchronized, do not do this again. This might
302 * also happen if it was synchronized in an VFS operation, e.g.
303 * 'ubifs_link()'.
305 if (!ui->dirty) {
306 mutex_unlock(&ui->ui_mutex);
307 return 0;
311 * As an optimization, do not write orphan inodes to the media just
312 * because this is not needed.
314 dbg_gen("inode %lu, mode %#x, nlink %u",
315 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
316 if (inode->i_nlink) {
317 err = ubifs_jnl_write_inode(c, inode);
318 if (err)
319 ubifs_err("can't write inode %lu, error %d",
320 inode->i_ino, err);
323 ui->dirty = 0;
324 mutex_unlock(&ui->ui_mutex);
325 ubifs_release_dirty_inode_budget(c, ui);
326 return err;
329 static void ubifs_delete_inode(struct inode *inode)
331 int err;
332 struct ubifs_info *c = inode->i_sb->s_fs_info;
333 struct ubifs_inode *ui = ubifs_inode(inode);
335 if (ui->xattr)
337 * Extended attribute inode deletions are fully handled in
338 * 'ubifs_removexattr()'. These inodes are special and have
339 * limited usage, so there is nothing to do here.
341 goto out;
343 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
344 ubifs_assert(!atomic_read(&inode->i_count));
345 ubifs_assert(inode->i_nlink == 0);
347 truncate_inode_pages(&inode->i_data, 0);
348 if (is_bad_inode(inode))
349 goto out;
351 ui->ui_size = inode->i_size = 0;
352 err = ubifs_jnl_delete_inode(c, inode);
353 if (err)
355 * Worst case we have a lost orphan inode wasting space, so a
356 * simple error message is OK here.
358 ubifs_err("can't delete inode %lu, error %d",
359 inode->i_ino, err);
361 out:
362 if (ui->dirty)
363 ubifs_release_dirty_inode_budget(c, ui);
364 else {
365 /* We've deleted something - clean the "no space" flags */
366 c->nospace = c->nospace_rp = 0;
367 smp_wmb();
369 clear_inode(inode);
372 static void ubifs_dirty_inode(struct inode *inode)
374 struct ubifs_inode *ui = ubifs_inode(inode);
376 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
377 if (!ui->dirty) {
378 ui->dirty = 1;
379 dbg_gen("inode %lu", inode->i_ino);
383 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
385 struct ubifs_info *c = dentry->d_sb->s_fs_info;
386 unsigned long long free;
387 __le32 *uuid = (__le32 *)c->uuid;
389 free = ubifs_get_free_space(c);
390 dbg_gen("free space %lld bytes (%lld blocks)",
391 free, free >> UBIFS_BLOCK_SHIFT);
393 buf->f_type = UBIFS_SUPER_MAGIC;
394 buf->f_bsize = UBIFS_BLOCK_SIZE;
395 buf->f_blocks = c->block_cnt;
396 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
397 if (free > c->report_rp_size)
398 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
399 else
400 buf->f_bavail = 0;
401 buf->f_files = 0;
402 buf->f_ffree = 0;
403 buf->f_namelen = UBIFS_MAX_NLEN;
404 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
405 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
406 ubifs_assert(buf->f_bfree <= c->block_cnt);
407 return 0;
410 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
412 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
414 if (c->mount_opts.unmount_mode == 2)
415 seq_printf(s, ",fast_unmount");
416 else if (c->mount_opts.unmount_mode == 1)
417 seq_printf(s, ",norm_unmount");
419 if (c->mount_opts.bulk_read == 2)
420 seq_printf(s, ",bulk_read");
421 else if (c->mount_opts.bulk_read == 1)
422 seq_printf(s, ",no_bulk_read");
424 if (c->mount_opts.chk_data_crc == 2)
425 seq_printf(s, ",chk_data_crc");
426 else if (c->mount_opts.chk_data_crc == 1)
427 seq_printf(s, ",no_chk_data_crc");
429 if (c->mount_opts.override_compr) {
430 seq_printf(s, ",compr=%s",
431 ubifs_compr_name(c->mount_opts.compr_type));
434 return 0;
437 static int ubifs_sync_fs(struct super_block *sb, int wait)
439 int i, err;
440 struct ubifs_info *c = sb->s_fs_info;
441 struct writeback_control wbc = {
442 .sync_mode = WB_SYNC_ALL,
443 .range_start = 0,
444 .range_end = LLONG_MAX,
445 .nr_to_write = LONG_MAX,
449 * Zero @wait is just an advisory thing to help the file system shove
450 * lots of data into the queues, and there will be the second
451 * '->sync_fs()' call, with non-zero @wait.
453 if (!wait)
454 return 0;
457 * VFS calls '->sync_fs()' before synchronizing all dirty inodes and
458 * pages, so synchronize them first, then commit the journal. Strictly
459 * speaking, it is not necessary to commit the journal here,
460 * synchronizing write-buffers would be enough. But committing makes
461 * UBIFS free space predictions much more accurate, so we want to let
462 * the user be able to get more accurate results of 'statfs()' after
463 * they synchronize the file system.
465 generic_sync_sb_inodes(sb, &wbc);
468 * Synchronize write buffers, because 'ubifs_run_commit()' does not
469 * do this if it waits for an already running commit.
471 for (i = 0; i < c->jhead_cnt; i++) {
472 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
473 if (err)
474 return err;
477 err = ubifs_run_commit(c);
478 if (err)
479 return err;
481 return ubi_sync(c->vi.ubi_num);
485 * init_constants_early - initialize UBIFS constants.
486 * @c: UBIFS file-system description object
488 * This function initialize UBIFS constants which do not need the superblock to
489 * be read. It also checks that the UBI volume satisfies basic UBIFS
490 * requirements. Returns zero in case of success and a negative error code in
491 * case of failure.
493 static int init_constants_early(struct ubifs_info *c)
495 if (c->vi.corrupted) {
496 ubifs_warn("UBI volume is corrupted - read-only mode");
497 c->ro_media = 1;
500 if (c->di.ro_mode) {
501 ubifs_msg("read-only UBI device");
502 c->ro_media = 1;
505 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
506 ubifs_msg("static UBI volume - read-only mode");
507 c->ro_media = 1;
510 c->leb_cnt = c->vi.size;
511 c->leb_size = c->vi.usable_leb_size;
512 c->half_leb_size = c->leb_size / 2;
513 c->min_io_size = c->di.min_io_size;
514 c->min_io_shift = fls(c->min_io_size) - 1;
516 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
517 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
518 c->leb_size, UBIFS_MIN_LEB_SZ);
519 return -EINVAL;
522 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
523 ubifs_err("too few LEBs (%d), min. is %d",
524 c->leb_cnt, UBIFS_MIN_LEB_CNT);
525 return -EINVAL;
528 if (!is_power_of_2(c->min_io_size)) {
529 ubifs_err("bad min. I/O size %d", c->min_io_size);
530 return -EINVAL;
534 * UBIFS aligns all node to 8-byte boundary, so to make function in
535 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
536 * less than 8.
538 if (c->min_io_size < 8) {
539 c->min_io_size = 8;
540 c->min_io_shift = 3;
543 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
544 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
547 * Initialize node length ranges which are mostly needed for node
548 * length validation.
550 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
551 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
552 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
553 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
554 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
555 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
557 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
558 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
559 c->ranges[UBIFS_ORPH_NODE].min_len =
560 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
561 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
562 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
563 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
564 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
565 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
566 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
567 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
569 * Minimum indexing node size is amended later when superblock is
570 * read and the key length is known.
572 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
574 * Maximum indexing node size is amended later when superblock is
575 * read and the fanout is known.
577 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
580 * Initialize dead and dark LEB space watermarks. See gc.c for comments
581 * about these values.
583 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
584 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
587 * Calculate how many bytes would be wasted at the end of LEB if it was
588 * fully filled with data nodes of maximum size. This is used in
589 * calculations when reporting free space.
591 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
593 /* Buffer size for bulk-reads */
594 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
595 if (c->max_bu_buf_len > c->leb_size)
596 c->max_bu_buf_len = c->leb_size;
597 return 0;
601 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
602 * @c: UBIFS file-system description object
603 * @lnum: LEB the write-buffer was synchronized to
604 * @free: how many free bytes left in this LEB
605 * @pad: how many bytes were padded
607 * This is a callback function which is called by the I/O unit when the
608 * write-buffer is synchronized. We need this to correctly maintain space
609 * accounting in bud logical eraseblocks. This function returns zero in case of
610 * success and a negative error code in case of failure.
612 * This function actually belongs to the journal, but we keep it here because
613 * we want to keep it static.
615 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
617 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
621 * init_constants_sb - initialize UBIFS constants.
622 * @c: UBIFS file-system description object
624 * This is a helper function which initializes various UBIFS constants after
625 * the superblock has been read. It also checks various UBIFS parameters and
626 * makes sure they are all right. Returns zero in case of success and a
627 * negative error code in case of failure.
629 static int init_constants_sb(struct ubifs_info *c)
631 int tmp, err;
632 long long tmp64;
634 c->main_bytes = (long long)c->main_lebs * c->leb_size;
635 c->max_znode_sz = sizeof(struct ubifs_znode) +
636 c->fanout * sizeof(struct ubifs_zbranch);
638 tmp = ubifs_idx_node_sz(c, 1);
639 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
640 c->min_idx_node_sz = ALIGN(tmp, 8);
642 tmp = ubifs_idx_node_sz(c, c->fanout);
643 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
644 c->max_idx_node_sz = ALIGN(tmp, 8);
646 /* Make sure LEB size is large enough to fit full commit */
647 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
648 tmp = ALIGN(tmp, c->min_io_size);
649 if (tmp > c->leb_size) {
650 dbg_err("too small LEB size %d, at least %d needed",
651 c->leb_size, tmp);
652 return -EINVAL;
656 * Make sure that the log is large enough to fit reference nodes for
657 * all buds plus one reserved LEB.
659 tmp64 = c->max_bud_bytes + c->leb_size - 1;
660 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
661 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
662 tmp /= c->leb_size;
663 tmp += 1;
664 if (c->log_lebs < tmp) {
665 dbg_err("too small log %d LEBs, required min. %d LEBs",
666 c->log_lebs, tmp);
667 return -EINVAL;
671 * When budgeting we assume worst-case scenarios when the pages are not
672 * be compressed and direntries are of the maximum size.
674 * Note, data, which may be stored in inodes is budgeted separately, so
675 * it is not included into 'c->inode_budget'.
677 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
678 c->inode_budget = UBIFS_INO_NODE_SZ;
679 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
682 * When the amount of flash space used by buds becomes
683 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
684 * The writers are unblocked when the commit is finished. To avoid
685 * writers to be blocked UBIFS initiates background commit in advance,
686 * when number of bud bytes becomes above the limit defined below.
688 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
691 * Ensure minimum journal size. All the bytes in the journal heads are
692 * considered to be used, when calculating the current journal usage.
693 * Consequently, if the journal is too small, UBIFS will treat it as
694 * always full.
696 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
697 if (c->bg_bud_bytes < tmp64)
698 c->bg_bud_bytes = tmp64;
699 if (c->max_bud_bytes < tmp64 + c->leb_size)
700 c->max_bud_bytes = tmp64 + c->leb_size;
702 err = ubifs_calc_lpt_geom(c);
703 if (err)
704 return err;
706 /* Initialize effective LEB size used in budgeting calculations */
707 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
708 return 0;
712 * init_constants_master - initialize UBIFS constants.
713 * @c: UBIFS file-system description object
715 * This is a helper function which initializes various UBIFS constants after
716 * the master node has been read. It also checks various UBIFS parameters and
717 * makes sure they are all right.
719 static void init_constants_master(struct ubifs_info *c)
721 long long tmp64;
723 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
724 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
727 * Calculate total amount of FS blocks. This number is not used
728 * internally because it does not make much sense for UBIFS, but it is
729 * necessary to report something for the 'statfs()' call.
731 * Subtract the LEB reserved for GC, the LEB which is reserved for
732 * deletions, minimum LEBs for the index, and assume only one journal
733 * head is available.
735 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
736 tmp64 *= (long long)c->leb_size - c->leb_overhead;
737 tmp64 = ubifs_reported_space(c, tmp64);
738 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
742 * take_gc_lnum - reserve GC LEB.
743 * @c: UBIFS file-system description object
745 * This function ensures that the LEB reserved for garbage collection is marked
746 * as "taken" in lprops. We also have to set free space to LEB size and dirty
747 * space to zero, because lprops may contain out-of-date information if the
748 * file-system was un-mounted before it has been committed. This function
749 * returns zero in case of success and a negative error code in case of
750 * failure.
752 static int take_gc_lnum(struct ubifs_info *c)
754 int err;
756 if (c->gc_lnum == -1) {
757 ubifs_err("no LEB for GC");
758 return -EINVAL;
761 /* And we have to tell lprops that this LEB is taken */
762 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
763 LPROPS_TAKEN, 0, 0);
764 return err;
768 * alloc_wbufs - allocate write-buffers.
769 * @c: UBIFS file-system description object
771 * This helper function allocates and initializes UBIFS write-buffers. Returns
772 * zero in case of success and %-ENOMEM in case of failure.
774 static int alloc_wbufs(struct ubifs_info *c)
776 int i, err;
778 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
779 GFP_KERNEL);
780 if (!c->jheads)
781 return -ENOMEM;
783 /* Initialize journal heads */
784 for (i = 0; i < c->jhead_cnt; i++) {
785 INIT_LIST_HEAD(&c->jheads[i].buds_list);
786 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
787 if (err)
788 return err;
790 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
791 c->jheads[i].wbuf.jhead = i;
794 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
796 * Garbage Collector head likely contains long-term data and
797 * does not need to be synchronized by timer.
799 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
800 c->jheads[GCHD].wbuf.no_timer = 1;
802 return 0;
806 * free_wbufs - free write-buffers.
807 * @c: UBIFS file-system description object
809 static void free_wbufs(struct ubifs_info *c)
811 int i;
813 if (c->jheads) {
814 for (i = 0; i < c->jhead_cnt; i++) {
815 kfree(c->jheads[i].wbuf.buf);
816 kfree(c->jheads[i].wbuf.inodes);
818 kfree(c->jheads);
819 c->jheads = NULL;
824 * free_orphans - free orphans.
825 * @c: UBIFS file-system description object
827 static void free_orphans(struct ubifs_info *c)
829 struct ubifs_orphan *orph;
831 while (c->orph_dnext) {
832 orph = c->orph_dnext;
833 c->orph_dnext = orph->dnext;
834 list_del(&orph->list);
835 kfree(orph);
838 while (!list_empty(&c->orph_list)) {
839 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
840 list_del(&orph->list);
841 kfree(orph);
842 dbg_err("orphan list not empty at unmount");
845 vfree(c->orph_buf);
846 c->orph_buf = NULL;
850 * free_buds - free per-bud objects.
851 * @c: UBIFS file-system description object
853 static void free_buds(struct ubifs_info *c)
855 struct rb_node *this = c->buds.rb_node;
856 struct ubifs_bud *bud;
858 while (this) {
859 if (this->rb_left)
860 this = this->rb_left;
861 else if (this->rb_right)
862 this = this->rb_right;
863 else {
864 bud = rb_entry(this, struct ubifs_bud, rb);
865 this = rb_parent(this);
866 if (this) {
867 if (this->rb_left == &bud->rb)
868 this->rb_left = NULL;
869 else
870 this->rb_right = NULL;
872 kfree(bud);
878 * check_volume_empty - check if the UBI volume is empty.
879 * @c: UBIFS file-system description object
881 * This function checks if the UBIFS volume is empty by looking if its LEBs are
882 * mapped or not. The result of checking is stored in the @c->empty variable.
883 * Returns zero in case of success and a negative error code in case of
884 * failure.
886 static int check_volume_empty(struct ubifs_info *c)
888 int lnum, err;
890 c->empty = 1;
891 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
892 err = ubi_is_mapped(c->ubi, lnum);
893 if (unlikely(err < 0))
894 return err;
895 if (err == 1) {
896 c->empty = 0;
897 break;
900 cond_resched();
903 return 0;
907 * UBIFS mount options.
909 * Opt_fast_unmount: do not run a journal commit before un-mounting
910 * Opt_norm_unmount: run a journal commit before un-mounting
911 * Opt_bulk_read: enable bulk-reads
912 * Opt_no_bulk_read: disable bulk-reads
913 * Opt_chk_data_crc: check CRCs when reading data nodes
914 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
915 * Opt_override_compr: override default compressor
916 * Opt_err: just end of array marker
918 enum {
919 Opt_fast_unmount,
920 Opt_norm_unmount,
921 Opt_bulk_read,
922 Opt_no_bulk_read,
923 Opt_chk_data_crc,
924 Opt_no_chk_data_crc,
925 Opt_override_compr,
926 Opt_err,
929 static const match_table_t tokens = {
930 {Opt_fast_unmount, "fast_unmount"},
931 {Opt_norm_unmount, "norm_unmount"},
932 {Opt_bulk_read, "bulk_read"},
933 {Opt_no_bulk_read, "no_bulk_read"},
934 {Opt_chk_data_crc, "chk_data_crc"},
935 {Opt_no_chk_data_crc, "no_chk_data_crc"},
936 {Opt_override_compr, "compr=%s"},
937 {Opt_err, NULL},
941 * parse_standard_option - parse a standard mount option.
942 * @option: the option to parse
944 * Normally, standard mount options like "sync" are passed to file-systems as
945 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
946 * be present in the options string. This function tries to deal with this
947 * situation and parse standard options. Returns 0 if the option was not
948 * recognized, and the corresponding integer flag if it was.
950 * UBIFS is only interested in the "sync" option, so do not check for anything
951 * else.
953 static int parse_standard_option(const char *option)
955 ubifs_msg("parse %s", option);
956 if (!strcmp(option, "sync"))
957 return MS_SYNCHRONOUS;
958 return 0;
962 * ubifs_parse_options - parse mount parameters.
963 * @c: UBIFS file-system description object
964 * @options: parameters to parse
965 * @is_remount: non-zero if this is FS re-mount
967 * This function parses UBIFS mount options and returns zero in case success
968 * and a negative error code in case of failure.
970 static int ubifs_parse_options(struct ubifs_info *c, char *options,
971 int is_remount)
973 char *p;
974 substring_t args[MAX_OPT_ARGS];
976 if (!options)
977 return 0;
979 while ((p = strsep(&options, ","))) {
980 int token;
982 if (!*p)
983 continue;
985 token = match_token(p, tokens, args);
986 switch (token) {
988 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
989 * We accept them in order to be backward-compatible. But this
990 * should be removed at some point.
992 case Opt_fast_unmount:
993 c->mount_opts.unmount_mode = 2;
994 break;
995 case Opt_norm_unmount:
996 c->mount_opts.unmount_mode = 1;
997 break;
998 case Opt_bulk_read:
999 c->mount_opts.bulk_read = 2;
1000 c->bulk_read = 1;
1001 break;
1002 case Opt_no_bulk_read:
1003 c->mount_opts.bulk_read = 1;
1004 c->bulk_read = 0;
1005 break;
1006 case Opt_chk_data_crc:
1007 c->mount_opts.chk_data_crc = 2;
1008 c->no_chk_data_crc = 0;
1009 break;
1010 case Opt_no_chk_data_crc:
1011 c->mount_opts.chk_data_crc = 1;
1012 c->no_chk_data_crc = 1;
1013 break;
1014 case Opt_override_compr:
1016 char *name = match_strdup(&args[0]);
1018 if (!name)
1019 return -ENOMEM;
1020 if (!strcmp(name, "none"))
1021 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1022 else if (!strcmp(name, "lzo"))
1023 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1024 else if (!strcmp(name, "zlib"))
1025 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1026 else {
1027 ubifs_err("unknown compressor \"%s\"", name);
1028 kfree(name);
1029 return -EINVAL;
1031 kfree(name);
1032 c->mount_opts.override_compr = 1;
1033 c->default_compr = c->mount_opts.compr_type;
1034 break;
1036 default:
1038 unsigned long flag;
1039 struct super_block *sb = c->vfs_sb;
1041 flag = parse_standard_option(p);
1042 if (!flag) {
1043 ubifs_err("unrecognized mount option \"%s\" "
1044 "or missing value", p);
1045 return -EINVAL;
1047 sb->s_flags |= flag;
1048 break;
1053 return 0;
1057 * destroy_journal - destroy journal data structures.
1058 * @c: UBIFS file-system description object
1060 * This function destroys journal data structures including those that may have
1061 * been created by recovery functions.
1063 static void destroy_journal(struct ubifs_info *c)
1065 while (!list_empty(&c->unclean_leb_list)) {
1066 struct ubifs_unclean_leb *ucleb;
1068 ucleb = list_entry(c->unclean_leb_list.next,
1069 struct ubifs_unclean_leb, list);
1070 list_del(&ucleb->list);
1071 kfree(ucleb);
1073 while (!list_empty(&c->old_buds)) {
1074 struct ubifs_bud *bud;
1076 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1077 list_del(&bud->list);
1078 kfree(bud);
1080 ubifs_destroy_idx_gc(c);
1081 ubifs_destroy_size_tree(c);
1082 ubifs_tnc_close(c);
1083 free_buds(c);
1087 * bu_init - initialize bulk-read information.
1088 * @c: UBIFS file-system description object
1090 static void bu_init(struct ubifs_info *c)
1092 ubifs_assert(c->bulk_read == 1);
1094 if (c->bu.buf)
1095 return; /* Already initialized */
1097 again:
1098 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1099 if (!c->bu.buf) {
1100 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1101 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1102 goto again;
1105 /* Just disable bulk-read */
1106 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1107 "disabling it", c->max_bu_buf_len);
1108 c->mount_opts.bulk_read = 1;
1109 c->bulk_read = 0;
1110 return;
1115 * check_free_space - check if there is enough free space to mount.
1116 * @c: UBIFS file-system description object
1118 * This function makes sure UBIFS has enough free space to be mounted in
1119 * read/write mode. UBIFS must always have some free space to allow deletions.
1121 static int check_free_space(struct ubifs_info *c)
1123 ubifs_assert(c->dark_wm > 0);
1124 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1125 ubifs_err("insufficient free space to mount in read/write mode");
1126 dbg_dump_budg(c);
1127 dbg_dump_lprops(c);
1128 return -ENOSPC;
1130 return 0;
1134 * mount_ubifs - mount UBIFS file-system.
1135 * @c: UBIFS file-system description object
1137 * This function mounts UBIFS file system. Returns zero in case of success and
1138 * a negative error code in case of failure.
1140 * Note, the function does not de-allocate resources it it fails half way
1141 * through, and the caller has to do this instead.
1143 static int mount_ubifs(struct ubifs_info *c)
1145 struct super_block *sb = c->vfs_sb;
1146 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1147 long long x;
1148 size_t sz;
1150 err = init_constants_early(c);
1151 if (err)
1152 return err;
1154 err = ubifs_debugging_init(c);
1155 if (err)
1156 return err;
1158 err = check_volume_empty(c);
1159 if (err)
1160 goto out_free;
1162 if (c->empty && (mounted_read_only || c->ro_media)) {
1164 * This UBI volume is empty, and read-only, or the file system
1165 * is mounted read-only - we cannot format it.
1167 ubifs_err("can't format empty UBI volume: read-only %s",
1168 c->ro_media ? "UBI volume" : "mount");
1169 err = -EROFS;
1170 goto out_free;
1173 if (c->ro_media && !mounted_read_only) {
1174 ubifs_err("cannot mount read-write - read-only media");
1175 err = -EROFS;
1176 goto out_free;
1180 * The requirement for the buffer is that it should fit indexing B-tree
1181 * height amount of integers. We assume the height if the TNC tree will
1182 * never exceed 64.
1184 err = -ENOMEM;
1185 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1186 if (!c->bottom_up_buf)
1187 goto out_free;
1189 c->sbuf = vmalloc(c->leb_size);
1190 if (!c->sbuf)
1191 goto out_free;
1193 if (!mounted_read_only) {
1194 c->ileb_buf = vmalloc(c->leb_size);
1195 if (!c->ileb_buf)
1196 goto out_free;
1199 if (c->bulk_read == 1)
1200 bu_init(c);
1203 * We have to check all CRCs, even for data nodes, when we mount the FS
1204 * (specifically, when we are replaying).
1206 c->always_chk_crc = 1;
1208 err = ubifs_read_superblock(c);
1209 if (err)
1210 goto out_free;
1213 * Make sure the compressor which is set as default in the superblock
1214 * or overridden by mount options is actually compiled in.
1216 if (!ubifs_compr_present(c->default_compr)) {
1217 ubifs_err("'compressor \"%s\" is not compiled in",
1218 ubifs_compr_name(c->default_compr));
1219 err = -ENOTSUPP;
1220 goto out_free;
1223 err = init_constants_sb(c);
1224 if (err)
1225 goto out_free;
1227 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1228 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1229 c->cbuf = kmalloc(sz, GFP_NOFS);
1230 if (!c->cbuf) {
1231 err = -ENOMEM;
1232 goto out_free;
1235 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1236 if (!mounted_read_only) {
1237 err = alloc_wbufs(c);
1238 if (err)
1239 goto out_cbuf;
1241 /* Create background thread */
1242 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1243 if (IS_ERR(c->bgt)) {
1244 err = PTR_ERR(c->bgt);
1245 c->bgt = NULL;
1246 ubifs_err("cannot spawn \"%s\", error %d",
1247 c->bgt_name, err);
1248 goto out_wbufs;
1250 wake_up_process(c->bgt);
1253 err = ubifs_read_master(c);
1254 if (err)
1255 goto out_master;
1257 init_constants_master(c);
1259 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1260 ubifs_msg("recovery needed");
1261 c->need_recovery = 1;
1262 if (!mounted_read_only) {
1263 err = ubifs_recover_inl_heads(c, c->sbuf);
1264 if (err)
1265 goto out_master;
1267 } else if (!mounted_read_only) {
1269 * Set the "dirty" flag so that if we reboot uncleanly we
1270 * will notice this immediately on the next mount.
1272 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1273 err = ubifs_write_master(c);
1274 if (err)
1275 goto out_master;
1278 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1279 if (err)
1280 goto out_lpt;
1282 err = dbg_check_idx_size(c, c->old_idx_sz);
1283 if (err)
1284 goto out_lpt;
1286 err = ubifs_replay_journal(c);
1287 if (err)
1288 goto out_journal;
1290 /* Calculate 'min_idx_lebs' after journal replay */
1291 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1293 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1294 if (err)
1295 goto out_orphans;
1297 if (!mounted_read_only) {
1298 int lnum;
1300 err = check_free_space(c);
1301 if (err)
1302 goto out_orphans;
1304 /* Check for enough log space */
1305 lnum = c->lhead_lnum + 1;
1306 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1307 lnum = UBIFS_LOG_LNUM;
1308 if (lnum == c->ltail_lnum) {
1309 err = ubifs_consolidate_log(c);
1310 if (err)
1311 goto out_orphans;
1314 if (c->need_recovery) {
1315 err = ubifs_recover_size(c);
1316 if (err)
1317 goto out_orphans;
1318 err = ubifs_rcvry_gc_commit(c);
1319 } else {
1320 err = take_gc_lnum(c);
1321 if (err)
1322 goto out_orphans;
1325 * GC LEB may contain garbage if there was an unclean
1326 * reboot, and it should be un-mapped.
1328 err = ubifs_leb_unmap(c, c->gc_lnum);
1329 if (err)
1330 return err;
1333 err = dbg_check_lprops(c);
1334 if (err)
1335 goto out_orphans;
1336 } else if (c->need_recovery) {
1337 err = ubifs_recover_size(c);
1338 if (err)
1339 goto out_orphans;
1340 } else {
1342 * Even if we mount read-only, we have to set space in GC LEB
1343 * to proper value because this affects UBIFS free space
1344 * reporting. We do not want to have a situation when
1345 * re-mounting from R/O to R/W changes amount of free space.
1347 err = take_gc_lnum(c);
1348 if (err)
1349 goto out_orphans;
1352 spin_lock(&ubifs_infos_lock);
1353 list_add_tail(&c->infos_list, &ubifs_infos);
1354 spin_unlock(&ubifs_infos_lock);
1356 if (c->need_recovery) {
1357 if (mounted_read_only)
1358 ubifs_msg("recovery deferred");
1359 else {
1360 c->need_recovery = 0;
1361 ubifs_msg("recovery completed");
1363 * GC LEB has to be empty and taken at this point. But
1364 * the journal head LEBs may also be accounted as
1365 * "empty taken" if they are empty.
1367 ubifs_assert(c->lst.taken_empty_lebs > 0);
1369 } else
1370 ubifs_assert(c->lst.taken_empty_lebs > 0);
1372 err = dbg_check_filesystem(c);
1373 if (err)
1374 goto out_infos;
1376 err = dbg_debugfs_init_fs(c);
1377 if (err)
1378 goto out_infos;
1380 c->always_chk_crc = 0;
1382 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1383 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1384 if (mounted_read_only)
1385 ubifs_msg("mounted read-only");
1386 x = (long long)c->main_lebs * c->leb_size;
1387 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1388 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1389 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1390 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1391 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1392 ubifs_msg("media format: w%d/r%d (latest is w%d/r%d)",
1393 c->fmt_version, c->ro_compat_version,
1394 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1395 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1396 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1397 c->report_rp_size, c->report_rp_size >> 10);
1399 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1400 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1401 dbg_msg("LEB size: %d bytes (%d KiB)",
1402 c->leb_size, c->leb_size >> 10);
1403 dbg_msg("data journal heads: %d",
1404 c->jhead_cnt - NONDATA_JHEADS_CNT);
1405 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1406 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1407 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1408 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1409 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1410 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1411 dbg_msg("big_lpt %d", c->big_lpt);
1412 dbg_msg("log LEBs: %d (%d - %d)",
1413 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1414 dbg_msg("LPT area LEBs: %d (%d - %d)",
1415 c->lpt_lebs, c->lpt_first, c->lpt_last);
1416 dbg_msg("orphan area LEBs: %d (%d - %d)",
1417 c->orph_lebs, c->orph_first, c->orph_last);
1418 dbg_msg("main area LEBs: %d (%d - %d)",
1419 c->main_lebs, c->main_first, c->leb_cnt - 1);
1420 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1421 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1422 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1423 dbg_msg("key hash type: %d", c->key_hash_type);
1424 dbg_msg("tree fanout: %d", c->fanout);
1425 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1426 dbg_msg("first main LEB: %d", c->main_first);
1427 dbg_msg("max. znode size %d", c->max_znode_sz);
1428 dbg_msg("max. index node size %d", c->max_idx_node_sz);
1429 dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
1430 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1431 dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
1432 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1433 dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
1434 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1435 dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu",
1436 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1437 UBIFS_MAX_DENT_NODE_SZ);
1438 dbg_msg("dead watermark: %d", c->dead_wm);
1439 dbg_msg("dark watermark: %d", c->dark_wm);
1440 dbg_msg("LEB overhead: %d", c->leb_overhead);
1441 x = (long long)c->main_lebs * c->dark_wm;
1442 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1443 x, x >> 10, x >> 20);
1444 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1445 c->max_bud_bytes, c->max_bud_bytes >> 10,
1446 c->max_bud_bytes >> 20);
1447 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1448 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1449 c->bg_bud_bytes >> 20);
1450 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1451 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1452 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1453 dbg_msg("commit number: %llu", c->cmt_no);
1455 return 0;
1457 out_infos:
1458 spin_lock(&ubifs_infos_lock);
1459 list_del(&c->infos_list);
1460 spin_unlock(&ubifs_infos_lock);
1461 out_orphans:
1462 free_orphans(c);
1463 out_journal:
1464 destroy_journal(c);
1465 out_lpt:
1466 ubifs_lpt_free(c, 0);
1467 out_master:
1468 kfree(c->mst_node);
1469 kfree(c->rcvrd_mst_node);
1470 if (c->bgt)
1471 kthread_stop(c->bgt);
1472 out_wbufs:
1473 free_wbufs(c);
1474 out_cbuf:
1475 kfree(c->cbuf);
1476 out_free:
1477 kfree(c->bu.buf);
1478 vfree(c->ileb_buf);
1479 vfree(c->sbuf);
1480 kfree(c->bottom_up_buf);
1481 ubifs_debugging_exit(c);
1482 return err;
1486 * ubifs_umount - un-mount UBIFS file-system.
1487 * @c: UBIFS file-system description object
1489 * Note, this function is called to free allocated resourced when un-mounting,
1490 * as well as free resources when an error occurred while we were half way
1491 * through mounting (error path cleanup function). So it has to make sure the
1492 * resource was actually allocated before freeing it.
1494 static void ubifs_umount(struct ubifs_info *c)
1496 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1497 c->vi.vol_id);
1499 dbg_debugfs_exit_fs(c);
1500 spin_lock(&ubifs_infos_lock);
1501 list_del(&c->infos_list);
1502 spin_unlock(&ubifs_infos_lock);
1504 if (c->bgt)
1505 kthread_stop(c->bgt);
1507 destroy_journal(c);
1508 free_wbufs(c);
1509 free_orphans(c);
1510 ubifs_lpt_free(c, 0);
1512 kfree(c->cbuf);
1513 kfree(c->rcvrd_mst_node);
1514 kfree(c->mst_node);
1515 kfree(c->bu.buf);
1516 vfree(c->ileb_buf);
1517 vfree(c->sbuf);
1518 kfree(c->bottom_up_buf);
1519 ubifs_debugging_exit(c);
1523 * ubifs_remount_rw - re-mount in read-write mode.
1524 * @c: UBIFS file-system description object
1526 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1527 * mode. This function allocates the needed resources and re-mounts UBIFS in
1528 * read-write mode.
1530 static int ubifs_remount_rw(struct ubifs_info *c)
1532 int err, lnum;
1534 if (c->rw_incompat) {
1535 ubifs_err("the file-system is not R/W-compatible");
1536 ubifs_msg("on-flash format version is w%d/r%d, but software "
1537 "only supports up to version w%d/r%d", c->fmt_version,
1538 c->ro_compat_version, UBIFS_FORMAT_VERSION,
1539 UBIFS_RO_COMPAT_VERSION);
1540 return -EROFS;
1543 mutex_lock(&c->umount_mutex);
1544 dbg_save_space_info(c);
1545 c->remounting_rw = 1;
1546 c->always_chk_crc = 1;
1548 err = check_free_space(c);
1549 if (err)
1550 goto out;
1552 if (c->old_leb_cnt != c->leb_cnt) {
1553 struct ubifs_sb_node *sup;
1555 sup = ubifs_read_sb_node(c);
1556 if (IS_ERR(sup)) {
1557 err = PTR_ERR(sup);
1558 goto out;
1560 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1561 err = ubifs_write_sb_node(c, sup);
1562 if (err)
1563 goto out;
1566 if (c->need_recovery) {
1567 ubifs_msg("completing deferred recovery");
1568 err = ubifs_write_rcvrd_mst_node(c);
1569 if (err)
1570 goto out;
1571 err = ubifs_recover_size(c);
1572 if (err)
1573 goto out;
1574 err = ubifs_clean_lebs(c, c->sbuf);
1575 if (err)
1576 goto out;
1577 err = ubifs_recover_inl_heads(c, c->sbuf);
1578 if (err)
1579 goto out;
1580 } else {
1581 /* A readonly mount is not allowed to have orphans */
1582 ubifs_assert(c->tot_orphans == 0);
1583 err = ubifs_clear_orphans(c);
1584 if (err)
1585 goto out;
1588 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1589 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1590 err = ubifs_write_master(c);
1591 if (err)
1592 goto out;
1595 c->ileb_buf = vmalloc(c->leb_size);
1596 if (!c->ileb_buf) {
1597 err = -ENOMEM;
1598 goto out;
1601 err = ubifs_lpt_init(c, 0, 1);
1602 if (err)
1603 goto out;
1605 err = alloc_wbufs(c);
1606 if (err)
1607 goto out;
1609 ubifs_create_buds_lists(c);
1611 /* Create background thread */
1612 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1613 if (IS_ERR(c->bgt)) {
1614 err = PTR_ERR(c->bgt);
1615 c->bgt = NULL;
1616 ubifs_err("cannot spawn \"%s\", error %d",
1617 c->bgt_name, err);
1618 goto out;
1620 wake_up_process(c->bgt);
1622 c->orph_buf = vmalloc(c->leb_size);
1623 if (!c->orph_buf) {
1624 err = -ENOMEM;
1625 goto out;
1628 /* Check for enough log space */
1629 lnum = c->lhead_lnum + 1;
1630 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1631 lnum = UBIFS_LOG_LNUM;
1632 if (lnum == c->ltail_lnum) {
1633 err = ubifs_consolidate_log(c);
1634 if (err)
1635 goto out;
1638 if (c->need_recovery)
1639 err = ubifs_rcvry_gc_commit(c);
1640 else
1641 err = ubifs_leb_unmap(c, c->gc_lnum);
1642 if (err)
1643 goto out;
1645 if (c->need_recovery) {
1646 c->need_recovery = 0;
1647 ubifs_msg("deferred recovery completed");
1650 dbg_gen("re-mounted read-write");
1651 c->vfs_sb->s_flags &= ~MS_RDONLY;
1652 c->remounting_rw = 0;
1653 c->always_chk_crc = 0;
1654 err = dbg_check_space_info(c);
1655 mutex_unlock(&c->umount_mutex);
1656 return err;
1658 out:
1659 vfree(c->orph_buf);
1660 c->orph_buf = NULL;
1661 if (c->bgt) {
1662 kthread_stop(c->bgt);
1663 c->bgt = NULL;
1665 free_wbufs(c);
1666 vfree(c->ileb_buf);
1667 c->ileb_buf = NULL;
1668 ubifs_lpt_free(c, 1);
1669 c->remounting_rw = 0;
1670 c->always_chk_crc = 0;
1671 mutex_unlock(&c->umount_mutex);
1672 return err;
1676 * ubifs_remount_ro - re-mount in read-only mode.
1677 * @c: UBIFS file-system description object
1679 * We assume VFS has stopped writing. Possibly the background thread could be
1680 * running a commit, however kthread_stop will wait in that case.
1682 static void ubifs_remount_ro(struct ubifs_info *c)
1684 int i, err;
1686 ubifs_assert(!c->need_recovery);
1687 ubifs_assert(!(c->vfs_sb->s_flags & MS_RDONLY));
1689 mutex_lock(&c->umount_mutex);
1690 if (c->bgt) {
1691 kthread_stop(c->bgt);
1692 c->bgt = NULL;
1695 dbg_save_space_info(c);
1697 for (i = 0; i < c->jhead_cnt; i++) {
1698 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1699 hrtimer_cancel(&c->jheads[i].wbuf.timer);
1702 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1703 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1704 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1705 err = ubifs_write_master(c);
1706 if (err)
1707 ubifs_ro_mode(c, err);
1709 free_wbufs(c);
1710 vfree(c->orph_buf);
1711 c->orph_buf = NULL;
1712 vfree(c->ileb_buf);
1713 c->ileb_buf = NULL;
1714 ubifs_lpt_free(c, 1);
1715 err = dbg_check_space_info(c);
1716 if (err)
1717 ubifs_ro_mode(c, err);
1718 mutex_unlock(&c->umount_mutex);
1721 static void ubifs_put_super(struct super_block *sb)
1723 int i;
1724 struct ubifs_info *c = sb->s_fs_info;
1726 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1727 c->vi.vol_id);
1729 lock_kernel();
1732 * The following asserts are only valid if there has not been a failure
1733 * of the media. For example, there will be dirty inodes if we failed
1734 * to write them back because of I/O errors.
1736 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1737 ubifs_assert(c->budg_idx_growth == 0);
1738 ubifs_assert(c->budg_dd_growth == 0);
1739 ubifs_assert(c->budg_data_growth == 0);
1742 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1743 * and file system un-mount. Namely, it prevents the shrinker from
1744 * picking this superblock for shrinking - it will be just skipped if
1745 * the mutex is locked.
1747 mutex_lock(&c->umount_mutex);
1748 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1750 * First of all kill the background thread to make sure it does
1751 * not interfere with un-mounting and freeing resources.
1753 if (c->bgt) {
1754 kthread_stop(c->bgt);
1755 c->bgt = NULL;
1758 /* Synchronize write-buffers */
1759 if (c->jheads)
1760 for (i = 0; i < c->jhead_cnt; i++)
1761 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1764 * On fatal errors c->ro_media is set to 1, in which case we do
1765 * not write the master node.
1767 if (!c->ro_media) {
1769 * We are being cleanly unmounted which means the
1770 * orphans were killed - indicate this in the master
1771 * node. Also save the reserved GC LEB number.
1773 int err;
1775 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1776 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1777 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1778 err = ubifs_write_master(c);
1779 if (err)
1781 * Recovery will attempt to fix the master area
1782 * next mount, so we just print a message and
1783 * continue to unmount normally.
1785 ubifs_err("failed to write master node, "
1786 "error %d", err);
1790 ubifs_umount(c);
1791 bdi_destroy(&c->bdi);
1792 ubi_close_volume(c->ubi);
1793 mutex_unlock(&c->umount_mutex);
1794 kfree(c);
1796 unlock_kernel();
1799 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1801 int err;
1802 struct ubifs_info *c = sb->s_fs_info;
1804 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1806 err = ubifs_parse_options(c, data, 1);
1807 if (err) {
1808 ubifs_err("invalid or unknown remount parameter");
1809 return err;
1812 lock_kernel();
1813 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1814 if (c->ro_media) {
1815 ubifs_msg("cannot re-mount due to prior errors");
1816 unlock_kernel();
1817 return -EROFS;
1819 err = ubifs_remount_rw(c);
1820 if (err) {
1821 unlock_kernel();
1822 return err;
1824 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY)) {
1825 if (c->ro_media) {
1826 ubifs_msg("cannot re-mount due to prior errors");
1827 unlock_kernel();
1828 return -EROFS;
1830 ubifs_remount_ro(c);
1833 if (c->bulk_read == 1)
1834 bu_init(c);
1835 else {
1836 dbg_gen("disable bulk-read");
1837 kfree(c->bu.buf);
1838 c->bu.buf = NULL;
1841 ubifs_assert(c->lst.taken_empty_lebs > 0);
1842 unlock_kernel();
1843 return 0;
1846 const struct super_operations ubifs_super_operations = {
1847 .alloc_inode = ubifs_alloc_inode,
1848 .destroy_inode = ubifs_destroy_inode,
1849 .put_super = ubifs_put_super,
1850 .write_inode = ubifs_write_inode,
1851 .delete_inode = ubifs_delete_inode,
1852 .statfs = ubifs_statfs,
1853 .dirty_inode = ubifs_dirty_inode,
1854 .remount_fs = ubifs_remount_fs,
1855 .show_options = ubifs_show_options,
1856 .sync_fs = ubifs_sync_fs,
1860 * open_ubi - parse UBI device name string and open the UBI device.
1861 * @name: UBI volume name
1862 * @mode: UBI volume open mode
1864 * There are several ways to specify UBI volumes when mounting UBIFS:
1865 * o ubiX_Y - UBI device number X, volume Y;
1866 * o ubiY - UBI device number 0, volume Y;
1867 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1868 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1870 * Alternative '!' separator may be used instead of ':' (because some shells
1871 * like busybox may interpret ':' as an NFS host name separator). This function
1872 * returns ubi volume object in case of success and a negative error code in
1873 * case of failure.
1875 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1877 int dev, vol;
1878 char *endptr;
1880 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1881 return ERR_PTR(-EINVAL);
1883 /* ubi:NAME method */
1884 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1885 return ubi_open_volume_nm(0, name + 4, mode);
1887 if (!isdigit(name[3]))
1888 return ERR_PTR(-EINVAL);
1890 dev = simple_strtoul(name + 3, &endptr, 0);
1892 /* ubiY method */
1893 if (*endptr == '\0')
1894 return ubi_open_volume(0, dev, mode);
1896 /* ubiX_Y method */
1897 if (*endptr == '_' && isdigit(endptr[1])) {
1898 vol = simple_strtoul(endptr + 1, &endptr, 0);
1899 if (*endptr != '\0')
1900 return ERR_PTR(-EINVAL);
1901 return ubi_open_volume(dev, vol, mode);
1904 /* ubiX:NAME method */
1905 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1906 return ubi_open_volume_nm(dev, ++endptr, mode);
1908 return ERR_PTR(-EINVAL);
1911 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1913 struct ubi_volume_desc *ubi = sb->s_fs_info;
1914 struct ubifs_info *c;
1915 struct inode *root;
1916 int err;
1918 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1919 if (!c)
1920 return -ENOMEM;
1922 spin_lock_init(&c->cnt_lock);
1923 spin_lock_init(&c->cs_lock);
1924 spin_lock_init(&c->buds_lock);
1925 spin_lock_init(&c->space_lock);
1926 spin_lock_init(&c->orphan_lock);
1927 init_rwsem(&c->commit_sem);
1928 mutex_init(&c->lp_mutex);
1929 mutex_init(&c->tnc_mutex);
1930 mutex_init(&c->log_mutex);
1931 mutex_init(&c->mst_mutex);
1932 mutex_init(&c->umount_mutex);
1933 mutex_init(&c->bu_mutex);
1934 init_waitqueue_head(&c->cmt_wq);
1935 c->buds = RB_ROOT;
1936 c->old_idx = RB_ROOT;
1937 c->size_tree = RB_ROOT;
1938 c->orph_tree = RB_ROOT;
1939 INIT_LIST_HEAD(&c->infos_list);
1940 INIT_LIST_HEAD(&c->idx_gc);
1941 INIT_LIST_HEAD(&c->replay_list);
1942 INIT_LIST_HEAD(&c->replay_buds);
1943 INIT_LIST_HEAD(&c->uncat_list);
1944 INIT_LIST_HEAD(&c->empty_list);
1945 INIT_LIST_HEAD(&c->freeable_list);
1946 INIT_LIST_HEAD(&c->frdi_idx_list);
1947 INIT_LIST_HEAD(&c->unclean_leb_list);
1948 INIT_LIST_HEAD(&c->old_buds);
1949 INIT_LIST_HEAD(&c->orph_list);
1950 INIT_LIST_HEAD(&c->orph_new);
1952 c->vfs_sb = sb;
1953 c->highest_inum = UBIFS_FIRST_INO;
1954 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1956 ubi_get_volume_info(ubi, &c->vi);
1957 ubi_get_device_info(c->vi.ubi_num, &c->di);
1959 /* Re-open the UBI device in read-write mode */
1960 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1961 if (IS_ERR(c->ubi)) {
1962 err = PTR_ERR(c->ubi);
1963 goto out_free;
1967 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1968 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1969 * which means the user would have to wait not just for their own I/O
1970 * but the read-ahead I/O as well i.e. completely pointless.
1972 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1974 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1975 c->bdi.unplug_io_fn = default_unplug_io_fn;
1976 err = bdi_init(&c->bdi);
1977 if (err)
1978 goto out_close;
1979 err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
1980 c->vi.ubi_num, c->vi.vol_id);
1981 if (err)
1982 goto out_bdi;
1984 err = ubifs_parse_options(c, data, 0);
1985 if (err)
1986 goto out_bdi;
1988 sb->s_fs_info = c;
1989 sb->s_magic = UBIFS_SUPER_MAGIC;
1990 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1991 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1992 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1993 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1994 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1995 sb->s_op = &ubifs_super_operations;
1997 mutex_lock(&c->umount_mutex);
1998 err = mount_ubifs(c);
1999 if (err) {
2000 ubifs_assert(err < 0);
2001 goto out_unlock;
2004 /* Read the root inode */
2005 root = ubifs_iget(sb, UBIFS_ROOT_INO);
2006 if (IS_ERR(root)) {
2007 err = PTR_ERR(root);
2008 goto out_umount;
2011 sb->s_root = d_alloc_root(root);
2012 if (!sb->s_root)
2013 goto out_iput;
2015 mutex_unlock(&c->umount_mutex);
2016 return 0;
2018 out_iput:
2019 iput(root);
2020 out_umount:
2021 ubifs_umount(c);
2022 out_unlock:
2023 mutex_unlock(&c->umount_mutex);
2024 out_bdi:
2025 bdi_destroy(&c->bdi);
2026 out_close:
2027 ubi_close_volume(c->ubi);
2028 out_free:
2029 kfree(c);
2030 return err;
2033 static int sb_test(struct super_block *sb, void *data)
2035 dev_t *dev = data;
2036 struct ubifs_info *c = sb->s_fs_info;
2038 return c->vi.cdev == *dev;
2041 static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
2042 const char *name, void *data, struct vfsmount *mnt)
2044 struct ubi_volume_desc *ubi;
2045 struct ubi_volume_info vi;
2046 struct super_block *sb;
2047 int err;
2049 dbg_gen("name %s, flags %#x", name, flags);
2052 * Get UBI device number and volume ID. Mount it read-only so far
2053 * because this might be a new mount point, and UBI allows only one
2054 * read-write user at a time.
2056 ubi = open_ubi(name, UBI_READONLY);
2057 if (IS_ERR(ubi)) {
2058 ubifs_err("cannot open \"%s\", error %d",
2059 name, (int)PTR_ERR(ubi));
2060 return PTR_ERR(ubi);
2062 ubi_get_volume_info(ubi, &vi);
2064 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
2066 sb = sget(fs_type, &sb_test, &set_anon_super, &vi.cdev);
2067 if (IS_ERR(sb)) {
2068 err = PTR_ERR(sb);
2069 goto out_close;
2072 if (sb->s_root) {
2073 /* A new mount point for already mounted UBIFS */
2074 dbg_gen("this ubi volume is already mounted");
2075 if ((flags ^ sb->s_flags) & MS_RDONLY) {
2076 err = -EBUSY;
2077 goto out_deact;
2079 } else {
2080 sb->s_flags = flags;
2082 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
2083 * replaced by 'c'.
2085 sb->s_fs_info = ubi;
2086 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2087 if (err)
2088 goto out_deact;
2089 /* We do not support atime */
2090 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2093 /* 'fill_super()' opens ubi again so we must close it here */
2094 ubi_close_volume(ubi);
2096 simple_set_mnt(mnt, sb);
2097 return 0;
2099 out_deact:
2100 deactivate_locked_super(sb);
2101 out_close:
2102 ubi_close_volume(ubi);
2103 return err;
2106 static struct file_system_type ubifs_fs_type = {
2107 .name = "ubifs",
2108 .owner = THIS_MODULE,
2109 .get_sb = ubifs_get_sb,
2110 .kill_sb = kill_anon_super,
2114 * Inode slab cache constructor.
2116 static void inode_slab_ctor(void *obj)
2118 struct ubifs_inode *ui = obj;
2119 inode_init_once(&ui->vfs_inode);
2122 static int __init ubifs_init(void)
2124 int err;
2126 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2128 /* Make sure node sizes are 8-byte aligned */
2129 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2130 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2131 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2132 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2133 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2134 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2135 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2136 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2137 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2138 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2139 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2141 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2142 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2143 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2144 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2145 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2146 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2148 /* Check min. node size */
2149 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2150 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2151 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2152 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2154 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2155 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2156 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2157 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2159 /* Defined node sizes */
2160 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2161 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2162 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2163 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2166 * We use 2 bit wide bit-fields to store compression type, which should
2167 * be amended if more compressors are added. The bit-fields are:
2168 * @compr_type in 'struct ubifs_inode', @default_compr in
2169 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2171 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2174 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2175 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2177 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2178 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2179 " at least 4096 bytes",
2180 (unsigned int)PAGE_CACHE_SIZE);
2181 return -EINVAL;
2184 err = register_filesystem(&ubifs_fs_type);
2185 if (err) {
2186 ubifs_err("cannot register file system, error %d", err);
2187 return err;
2190 err = -ENOMEM;
2191 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2192 sizeof(struct ubifs_inode), 0,
2193 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2194 &inode_slab_ctor);
2195 if (!ubifs_inode_slab)
2196 goto out_reg;
2198 register_shrinker(&ubifs_shrinker_info);
2200 err = ubifs_compressors_init();
2201 if (err)
2202 goto out_shrinker;
2204 err = dbg_debugfs_init();
2205 if (err)
2206 goto out_compr;
2208 return 0;
2210 out_compr:
2211 ubifs_compressors_exit();
2212 out_shrinker:
2213 unregister_shrinker(&ubifs_shrinker_info);
2214 kmem_cache_destroy(ubifs_inode_slab);
2215 out_reg:
2216 unregister_filesystem(&ubifs_fs_type);
2217 return err;
2219 /* late_initcall to let compressors initialize first */
2220 late_initcall(ubifs_init);
2222 static void __exit ubifs_exit(void)
2224 ubifs_assert(list_empty(&ubifs_infos));
2225 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2227 dbg_debugfs_exit();
2228 ubifs_compressors_exit();
2229 unregister_shrinker(&ubifs_shrinker_info);
2230 kmem_cache_destroy(ubifs_inode_slab);
2231 unregister_filesystem(&ubifs_fs_type);
2233 module_exit(ubifs_exit);
2235 MODULE_LICENSE("GPL");
2236 MODULE_VERSION(__stringify(UBIFS_VERSION));
2237 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2238 MODULE_DESCRIPTION("UBIFS - UBI File System");