UBIFS: do not forget to register BDI device
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ubifs / super.c
blobb9b051a4c01e12dd853f6b5c1fea6330d778bc3f
1 /*
2 * This file is part of UBIFS.
4 * Copyright (C) 2006-2008 Nokia Corporation.
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License version 2 as published by
8 * the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * more details.
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc., 51
17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 * Authors: Artem Bityutskiy (Битюцкий Артём)
20 * Adrian Hunter
24 * This file implements UBIFS initialization and VFS superblock operations. Some
25 * initialization stuff which is rather large and complex is placed at
26 * corresponding subsystems, but most of it is here.
29 #include <linux/init.h>
30 #include <linux/slab.h>
31 #include <linux/module.h>
32 #include <linux/ctype.h>
33 #include <linux/kthread.h>
34 #include <linux/parser.h>
35 #include <linux/seq_file.h>
36 #include <linux/mount.h>
37 #include <linux/math64.h>
38 #include <linux/writeback.h>
39 #include "ubifs.h"
42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
43 * allocating too much.
45 #define UBIFS_KMALLOC_OK (128*1024)
47 /* Slab cache for UBIFS inodes */
48 struct kmem_cache *ubifs_inode_slab;
50 /* UBIFS TNC shrinker description */
51 static struct shrinker ubifs_shrinker_info = {
52 .shrink = ubifs_shrinker,
53 .seeks = DEFAULT_SEEKS,
56 /**
57 * validate_inode - validate inode.
58 * @c: UBIFS file-system description object
59 * @inode: the inode to validate
61 * This is a helper function for 'ubifs_iget()' which validates various fields
62 * of a newly built inode to make sure they contain sane values and prevent
63 * possible vulnerabilities. Returns zero if the inode is all right and
64 * a non-zero error code if not.
66 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
68 int err;
69 const struct ubifs_inode *ui = ubifs_inode(inode);
71 if (inode->i_size > c->max_inode_sz) {
72 ubifs_err("inode is too large (%lld)",
73 (long long)inode->i_size);
74 return 1;
77 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
78 ubifs_err("unknown compression type %d", ui->compr_type);
79 return 2;
82 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
83 return 3;
85 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
86 return 4;
88 if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
89 return 5;
91 if (!ubifs_compr_present(ui->compr_type)) {
92 ubifs_warn("inode %lu uses '%s' compression, but it was not "
93 "compiled in", inode->i_ino,
94 ubifs_compr_name(ui->compr_type));
97 err = dbg_check_dir_size(c, inode);
98 return err;
101 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
103 int err;
104 union ubifs_key key;
105 struct ubifs_ino_node *ino;
106 struct ubifs_info *c = sb->s_fs_info;
107 struct inode *inode;
108 struct ubifs_inode *ui;
110 dbg_gen("inode %lu", inum);
112 inode = iget_locked(sb, inum);
113 if (!inode)
114 return ERR_PTR(-ENOMEM);
115 if (!(inode->i_state & I_NEW))
116 return inode;
117 ui = ubifs_inode(inode);
119 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
120 if (!ino) {
121 err = -ENOMEM;
122 goto out;
125 ino_key_init(c, &key, inode->i_ino);
127 err = ubifs_tnc_lookup(c, &key, ino);
128 if (err)
129 goto out_ino;
131 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
132 inode->i_nlink = le32_to_cpu(ino->nlink);
133 inode->i_uid = le32_to_cpu(ino->uid);
134 inode->i_gid = le32_to_cpu(ino->gid);
135 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
136 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
137 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
138 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
139 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
140 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
141 inode->i_mode = le32_to_cpu(ino->mode);
142 inode->i_size = le64_to_cpu(ino->size);
144 ui->data_len = le32_to_cpu(ino->data_len);
145 ui->flags = le32_to_cpu(ino->flags);
146 ui->compr_type = le16_to_cpu(ino->compr_type);
147 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
148 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
149 ui->xattr_size = le32_to_cpu(ino->xattr_size);
150 ui->xattr_names = le32_to_cpu(ino->xattr_names);
151 ui->synced_i_size = ui->ui_size = inode->i_size;
153 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
155 err = validate_inode(c, inode);
156 if (err)
157 goto out_invalid;
159 /* Disable read-ahead */
160 inode->i_mapping->backing_dev_info = &c->bdi;
162 switch (inode->i_mode & S_IFMT) {
163 case S_IFREG:
164 inode->i_mapping->a_ops = &ubifs_file_address_operations;
165 inode->i_op = &ubifs_file_inode_operations;
166 inode->i_fop = &ubifs_file_operations;
167 if (ui->xattr) {
168 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
169 if (!ui->data) {
170 err = -ENOMEM;
171 goto out_ino;
173 memcpy(ui->data, ino->data, ui->data_len);
174 ((char *)ui->data)[ui->data_len] = '\0';
175 } else if (ui->data_len != 0) {
176 err = 10;
177 goto out_invalid;
179 break;
180 case S_IFDIR:
181 inode->i_op = &ubifs_dir_inode_operations;
182 inode->i_fop = &ubifs_dir_operations;
183 if (ui->data_len != 0) {
184 err = 11;
185 goto out_invalid;
187 break;
188 case S_IFLNK:
189 inode->i_op = &ubifs_symlink_inode_operations;
190 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
191 err = 12;
192 goto out_invalid;
194 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
195 if (!ui->data) {
196 err = -ENOMEM;
197 goto out_ino;
199 memcpy(ui->data, ino->data, ui->data_len);
200 ((char *)ui->data)[ui->data_len] = '\0';
201 break;
202 case S_IFBLK:
203 case S_IFCHR:
205 dev_t rdev;
206 union ubifs_dev_desc *dev;
208 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
209 if (!ui->data) {
210 err = -ENOMEM;
211 goto out_ino;
214 dev = (union ubifs_dev_desc *)ino->data;
215 if (ui->data_len == sizeof(dev->new))
216 rdev = new_decode_dev(le32_to_cpu(dev->new));
217 else if (ui->data_len == sizeof(dev->huge))
218 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
219 else {
220 err = 13;
221 goto out_invalid;
223 memcpy(ui->data, ino->data, ui->data_len);
224 inode->i_op = &ubifs_file_inode_operations;
225 init_special_inode(inode, inode->i_mode, rdev);
226 break;
228 case S_IFSOCK:
229 case S_IFIFO:
230 inode->i_op = &ubifs_file_inode_operations;
231 init_special_inode(inode, inode->i_mode, 0);
232 if (ui->data_len != 0) {
233 err = 14;
234 goto out_invalid;
236 break;
237 default:
238 err = 15;
239 goto out_invalid;
242 kfree(ino);
243 ubifs_set_inode_flags(inode);
244 unlock_new_inode(inode);
245 return inode;
247 out_invalid:
248 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
249 dbg_dump_node(c, ino);
250 dbg_dump_inode(c, inode);
251 err = -EINVAL;
252 out_ino:
253 kfree(ino);
254 out:
255 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
256 iget_failed(inode);
257 return ERR_PTR(err);
260 static struct inode *ubifs_alloc_inode(struct super_block *sb)
262 struct ubifs_inode *ui;
264 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
265 if (!ui)
266 return NULL;
268 memset((void *)ui + sizeof(struct inode), 0,
269 sizeof(struct ubifs_inode) - sizeof(struct inode));
270 mutex_init(&ui->ui_mutex);
271 spin_lock_init(&ui->ui_lock);
272 return &ui->vfs_inode;
275 static void ubifs_destroy_inode(struct inode *inode)
277 struct ubifs_inode *ui = ubifs_inode(inode);
279 kfree(ui->data);
280 kmem_cache_free(ubifs_inode_slab, inode);
284 * Note, Linux write-back code calls this without 'i_mutex'.
286 static int ubifs_write_inode(struct inode *inode, int wait)
288 int err = 0;
289 struct ubifs_info *c = inode->i_sb->s_fs_info;
290 struct ubifs_inode *ui = ubifs_inode(inode);
292 ubifs_assert(!ui->xattr);
293 if (is_bad_inode(inode))
294 return 0;
296 mutex_lock(&ui->ui_mutex);
298 * Due to races between write-back forced by budgeting
299 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
300 * have already been synchronized, do not do this again. This might
301 * also happen if it was synchronized in an VFS operation, e.g.
302 * 'ubifs_link()'.
304 if (!ui->dirty) {
305 mutex_unlock(&ui->ui_mutex);
306 return 0;
310 * As an optimization, do not write orphan inodes to the media just
311 * because this is not needed.
313 dbg_gen("inode %lu, mode %#x, nlink %u",
314 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
315 if (inode->i_nlink) {
316 err = ubifs_jnl_write_inode(c, inode);
317 if (err)
318 ubifs_err("can't write inode %lu, error %d",
319 inode->i_ino, err);
322 ui->dirty = 0;
323 mutex_unlock(&ui->ui_mutex);
324 ubifs_release_dirty_inode_budget(c, ui);
325 return err;
328 static void ubifs_delete_inode(struct inode *inode)
330 int err;
331 struct ubifs_info *c = inode->i_sb->s_fs_info;
332 struct ubifs_inode *ui = ubifs_inode(inode);
334 if (ui->xattr)
336 * Extended attribute inode deletions are fully handled in
337 * 'ubifs_removexattr()'. These inodes are special and have
338 * limited usage, so there is nothing to do here.
340 goto out;
342 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
343 ubifs_assert(!atomic_read(&inode->i_count));
344 ubifs_assert(inode->i_nlink == 0);
346 truncate_inode_pages(&inode->i_data, 0);
347 if (is_bad_inode(inode))
348 goto out;
350 ui->ui_size = inode->i_size = 0;
351 err = ubifs_jnl_delete_inode(c, inode);
352 if (err)
354 * Worst case we have a lost orphan inode wasting space, so a
355 * simple error message is OK here.
357 ubifs_err("can't delete inode %lu, error %d",
358 inode->i_ino, err);
360 out:
361 if (ui->dirty)
362 ubifs_release_dirty_inode_budget(c, ui);
363 else {
364 /* We've deleted something - clean the "no space" flags */
365 c->nospace = c->nospace_rp = 0;
366 smp_wmb();
368 clear_inode(inode);
371 static void ubifs_dirty_inode(struct inode *inode)
373 struct ubifs_inode *ui = ubifs_inode(inode);
375 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
376 if (!ui->dirty) {
377 ui->dirty = 1;
378 dbg_gen("inode %lu", inode->i_ino);
382 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
384 struct ubifs_info *c = dentry->d_sb->s_fs_info;
385 unsigned long long free;
386 __le32 *uuid = (__le32 *)c->uuid;
388 free = ubifs_get_free_space(c);
389 dbg_gen("free space %lld bytes (%lld blocks)",
390 free, free >> UBIFS_BLOCK_SHIFT);
392 buf->f_type = UBIFS_SUPER_MAGIC;
393 buf->f_bsize = UBIFS_BLOCK_SIZE;
394 buf->f_blocks = c->block_cnt;
395 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
396 if (free > c->report_rp_size)
397 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
398 else
399 buf->f_bavail = 0;
400 buf->f_files = 0;
401 buf->f_ffree = 0;
402 buf->f_namelen = UBIFS_MAX_NLEN;
403 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
404 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
405 ubifs_assert(buf->f_bfree <= c->block_cnt);
406 return 0;
409 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
411 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
413 if (c->mount_opts.unmount_mode == 2)
414 seq_printf(s, ",fast_unmount");
415 else if (c->mount_opts.unmount_mode == 1)
416 seq_printf(s, ",norm_unmount");
418 if (c->mount_opts.bulk_read == 2)
419 seq_printf(s, ",bulk_read");
420 else if (c->mount_opts.bulk_read == 1)
421 seq_printf(s, ",no_bulk_read");
423 if (c->mount_opts.chk_data_crc == 2)
424 seq_printf(s, ",chk_data_crc");
425 else if (c->mount_opts.chk_data_crc == 1)
426 seq_printf(s, ",no_chk_data_crc");
428 if (c->mount_opts.override_compr) {
429 seq_printf(s, ",compr=%s",
430 ubifs_compr_name(c->mount_opts.compr_type));
433 return 0;
436 static int ubifs_sync_fs(struct super_block *sb, int wait)
438 int i, err;
439 struct ubifs_info *c = sb->s_fs_info;
440 struct writeback_control wbc = {
441 .sync_mode = WB_SYNC_ALL,
442 .range_start = 0,
443 .range_end = LLONG_MAX,
444 .nr_to_write = LONG_MAX,
448 * Zero @wait is just an advisory thing to help the file system shove
449 * lots of data into the queues, and there will be the second
450 * '->sync_fs()' call, with non-zero @wait.
452 if (!wait)
453 return 0;
455 if (sb->s_flags & MS_RDONLY)
456 return 0;
459 * VFS calls '->sync_fs()' before synchronizing all dirty inodes and
460 * pages, so synchronize them first, then commit the journal. Strictly
461 * speaking, it is not necessary to commit the journal here,
462 * synchronizing write-buffers would be enough. But committing makes
463 * UBIFS free space predictions much more accurate, so we want to let
464 * the user be able to get more accurate results of 'statfs()' after
465 * they synchronize the file system.
467 generic_sync_sb_inodes(sb, &wbc);
470 * Synchronize write buffers, because 'ubifs_run_commit()' does not
471 * do this if it waits for an already running commit.
473 for (i = 0; i < c->jhead_cnt; i++) {
474 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
475 if (err)
476 return err;
479 err = ubifs_run_commit(c);
480 if (err)
481 return err;
483 return ubi_sync(c->vi.ubi_num);
487 * init_constants_early - initialize UBIFS constants.
488 * @c: UBIFS file-system description object
490 * This function initialize UBIFS constants which do not need the superblock to
491 * be read. It also checks that the UBI volume satisfies basic UBIFS
492 * requirements. Returns zero in case of success and a negative error code in
493 * case of failure.
495 static int init_constants_early(struct ubifs_info *c)
497 if (c->vi.corrupted) {
498 ubifs_warn("UBI volume is corrupted - read-only mode");
499 c->ro_media = 1;
502 if (c->di.ro_mode) {
503 ubifs_msg("read-only UBI device");
504 c->ro_media = 1;
507 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
508 ubifs_msg("static UBI volume - read-only mode");
509 c->ro_media = 1;
512 c->leb_cnt = c->vi.size;
513 c->leb_size = c->vi.usable_leb_size;
514 c->half_leb_size = c->leb_size / 2;
515 c->min_io_size = c->di.min_io_size;
516 c->min_io_shift = fls(c->min_io_size) - 1;
518 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
519 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
520 c->leb_size, UBIFS_MIN_LEB_SZ);
521 return -EINVAL;
524 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
525 ubifs_err("too few LEBs (%d), min. is %d",
526 c->leb_cnt, UBIFS_MIN_LEB_CNT);
527 return -EINVAL;
530 if (!is_power_of_2(c->min_io_size)) {
531 ubifs_err("bad min. I/O size %d", c->min_io_size);
532 return -EINVAL;
536 * UBIFS aligns all node to 8-byte boundary, so to make function in
537 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
538 * less than 8.
540 if (c->min_io_size < 8) {
541 c->min_io_size = 8;
542 c->min_io_shift = 3;
545 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
546 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
549 * Initialize node length ranges which are mostly needed for node
550 * length validation.
552 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
553 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
554 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
555 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
556 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
557 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
559 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
560 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
561 c->ranges[UBIFS_ORPH_NODE].min_len =
562 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
563 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
564 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
565 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
566 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
567 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
568 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
569 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
571 * Minimum indexing node size is amended later when superblock is
572 * read and the key length is known.
574 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
576 * Maximum indexing node size is amended later when superblock is
577 * read and the fanout is known.
579 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
582 * Initialize dead and dark LEB space watermarks. See gc.c for comments
583 * about these values.
585 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
586 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
589 * Calculate how many bytes would be wasted at the end of LEB if it was
590 * fully filled with data nodes of maximum size. This is used in
591 * calculations when reporting free space.
593 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
595 /* Buffer size for bulk-reads */
596 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
597 if (c->max_bu_buf_len > c->leb_size)
598 c->max_bu_buf_len = c->leb_size;
599 return 0;
603 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
604 * @c: UBIFS file-system description object
605 * @lnum: LEB the write-buffer was synchronized to
606 * @free: how many free bytes left in this LEB
607 * @pad: how many bytes were padded
609 * This is a callback function which is called by the I/O unit when the
610 * write-buffer is synchronized. We need this to correctly maintain space
611 * accounting in bud logical eraseblocks. This function returns zero in case of
612 * success and a negative error code in case of failure.
614 * This function actually belongs to the journal, but we keep it here because
615 * we want to keep it static.
617 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
619 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
623 * init_constants_sb - initialize UBIFS constants.
624 * @c: UBIFS file-system description object
626 * This is a helper function which initializes various UBIFS constants after
627 * the superblock has been read. It also checks various UBIFS parameters and
628 * makes sure they are all right. Returns zero in case of success and a
629 * negative error code in case of failure.
631 static int init_constants_sb(struct ubifs_info *c)
633 int tmp, err;
634 long long tmp64;
636 c->main_bytes = (long long)c->main_lebs * c->leb_size;
637 c->max_znode_sz = sizeof(struct ubifs_znode) +
638 c->fanout * sizeof(struct ubifs_zbranch);
640 tmp = ubifs_idx_node_sz(c, 1);
641 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
642 c->min_idx_node_sz = ALIGN(tmp, 8);
644 tmp = ubifs_idx_node_sz(c, c->fanout);
645 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
646 c->max_idx_node_sz = ALIGN(tmp, 8);
648 /* Make sure LEB size is large enough to fit full commit */
649 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
650 tmp = ALIGN(tmp, c->min_io_size);
651 if (tmp > c->leb_size) {
652 dbg_err("too small LEB size %d, at least %d needed",
653 c->leb_size, tmp);
654 return -EINVAL;
658 * Make sure that the log is large enough to fit reference nodes for
659 * all buds plus one reserved LEB.
661 tmp64 = c->max_bud_bytes + c->leb_size - 1;
662 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
663 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
664 tmp /= c->leb_size;
665 tmp += 1;
666 if (c->log_lebs < tmp) {
667 dbg_err("too small log %d LEBs, required min. %d LEBs",
668 c->log_lebs, tmp);
669 return -EINVAL;
673 * When budgeting we assume worst-case scenarios when the pages are not
674 * be compressed and direntries are of the maximum size.
676 * Note, data, which may be stored in inodes is budgeted separately, so
677 * it is not included into 'c->inode_budget'.
679 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
680 c->inode_budget = UBIFS_INO_NODE_SZ;
681 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
684 * When the amount of flash space used by buds becomes
685 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
686 * The writers are unblocked when the commit is finished. To avoid
687 * writers to be blocked UBIFS initiates background commit in advance,
688 * when number of bud bytes becomes above the limit defined below.
690 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
693 * Ensure minimum journal size. All the bytes in the journal heads are
694 * considered to be used, when calculating the current journal usage.
695 * Consequently, if the journal is too small, UBIFS will treat it as
696 * always full.
698 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
699 if (c->bg_bud_bytes < tmp64)
700 c->bg_bud_bytes = tmp64;
701 if (c->max_bud_bytes < tmp64 + c->leb_size)
702 c->max_bud_bytes = tmp64 + c->leb_size;
704 err = ubifs_calc_lpt_geom(c);
705 if (err)
706 return err;
708 /* Initialize effective LEB size used in budgeting calculations */
709 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
710 return 0;
714 * init_constants_master - initialize UBIFS constants.
715 * @c: UBIFS file-system description object
717 * This is a helper function which initializes various UBIFS constants after
718 * the master node has been read. It also checks various UBIFS parameters and
719 * makes sure they are all right.
721 static void init_constants_master(struct ubifs_info *c)
723 long long tmp64;
725 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
726 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
729 * Calculate total amount of FS blocks. This number is not used
730 * internally because it does not make much sense for UBIFS, but it is
731 * necessary to report something for the 'statfs()' call.
733 * Subtract the LEB reserved for GC, the LEB which is reserved for
734 * deletions, minimum LEBs for the index, and assume only one journal
735 * head is available.
737 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
738 tmp64 *= (long long)c->leb_size - c->leb_overhead;
739 tmp64 = ubifs_reported_space(c, tmp64);
740 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
744 * take_gc_lnum - reserve GC LEB.
745 * @c: UBIFS file-system description object
747 * This function ensures that the LEB reserved for garbage collection is marked
748 * as "taken" in lprops. We also have to set free space to LEB size and dirty
749 * space to zero, because lprops may contain out-of-date information if the
750 * file-system was un-mounted before it has been committed. This function
751 * returns zero in case of success and a negative error code in case of
752 * failure.
754 static int take_gc_lnum(struct ubifs_info *c)
756 int err;
758 if (c->gc_lnum == -1) {
759 ubifs_err("no LEB for GC");
760 return -EINVAL;
763 /* And we have to tell lprops that this LEB is taken */
764 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
765 LPROPS_TAKEN, 0, 0);
766 return err;
770 * alloc_wbufs - allocate write-buffers.
771 * @c: UBIFS file-system description object
773 * This helper function allocates and initializes UBIFS write-buffers. Returns
774 * zero in case of success and %-ENOMEM in case of failure.
776 static int alloc_wbufs(struct ubifs_info *c)
778 int i, err;
780 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
781 GFP_KERNEL);
782 if (!c->jheads)
783 return -ENOMEM;
785 /* Initialize journal heads */
786 for (i = 0; i < c->jhead_cnt; i++) {
787 INIT_LIST_HEAD(&c->jheads[i].buds_list);
788 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
789 if (err)
790 return err;
792 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
793 c->jheads[i].wbuf.jhead = i;
796 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
798 * Garbage Collector head likely contains long-term data and
799 * does not need to be synchronized by timer.
801 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
802 c->jheads[GCHD].wbuf.timeout = 0;
804 return 0;
808 * free_wbufs - free write-buffers.
809 * @c: UBIFS file-system description object
811 static void free_wbufs(struct ubifs_info *c)
813 int i;
815 if (c->jheads) {
816 for (i = 0; i < c->jhead_cnt; i++) {
817 kfree(c->jheads[i].wbuf.buf);
818 kfree(c->jheads[i].wbuf.inodes);
820 kfree(c->jheads);
821 c->jheads = NULL;
826 * free_orphans - free orphans.
827 * @c: UBIFS file-system description object
829 static void free_orphans(struct ubifs_info *c)
831 struct ubifs_orphan *orph;
833 while (c->orph_dnext) {
834 orph = c->orph_dnext;
835 c->orph_dnext = orph->dnext;
836 list_del(&orph->list);
837 kfree(orph);
840 while (!list_empty(&c->orph_list)) {
841 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
842 list_del(&orph->list);
843 kfree(orph);
844 dbg_err("orphan list not empty at unmount");
847 vfree(c->orph_buf);
848 c->orph_buf = NULL;
852 * free_buds - free per-bud objects.
853 * @c: UBIFS file-system description object
855 static void free_buds(struct ubifs_info *c)
857 struct rb_node *this = c->buds.rb_node;
858 struct ubifs_bud *bud;
860 while (this) {
861 if (this->rb_left)
862 this = this->rb_left;
863 else if (this->rb_right)
864 this = this->rb_right;
865 else {
866 bud = rb_entry(this, struct ubifs_bud, rb);
867 this = rb_parent(this);
868 if (this) {
869 if (this->rb_left == &bud->rb)
870 this->rb_left = NULL;
871 else
872 this->rb_right = NULL;
874 kfree(bud);
880 * check_volume_empty - check if the UBI volume is empty.
881 * @c: UBIFS file-system description object
883 * This function checks if the UBIFS volume is empty by looking if its LEBs are
884 * mapped or not. The result of checking is stored in the @c->empty variable.
885 * Returns zero in case of success and a negative error code in case of
886 * failure.
888 static int check_volume_empty(struct ubifs_info *c)
890 int lnum, err;
892 c->empty = 1;
893 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
894 err = ubi_is_mapped(c->ubi, lnum);
895 if (unlikely(err < 0))
896 return err;
897 if (err == 1) {
898 c->empty = 0;
899 break;
902 cond_resched();
905 return 0;
909 * UBIFS mount options.
911 * Opt_fast_unmount: do not run a journal commit before un-mounting
912 * Opt_norm_unmount: run a journal commit before un-mounting
913 * Opt_bulk_read: enable bulk-reads
914 * Opt_no_bulk_read: disable bulk-reads
915 * Opt_chk_data_crc: check CRCs when reading data nodes
916 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
917 * Opt_override_compr: override default compressor
918 * Opt_err: just end of array marker
920 enum {
921 Opt_fast_unmount,
922 Opt_norm_unmount,
923 Opt_bulk_read,
924 Opt_no_bulk_read,
925 Opt_chk_data_crc,
926 Opt_no_chk_data_crc,
927 Opt_override_compr,
928 Opt_err,
931 static const match_table_t tokens = {
932 {Opt_fast_unmount, "fast_unmount"},
933 {Opt_norm_unmount, "norm_unmount"},
934 {Opt_bulk_read, "bulk_read"},
935 {Opt_no_bulk_read, "no_bulk_read"},
936 {Opt_chk_data_crc, "chk_data_crc"},
937 {Opt_no_chk_data_crc, "no_chk_data_crc"},
938 {Opt_override_compr, "compr=%s"},
939 {Opt_err, NULL},
943 * parse_standard_option - parse a standard mount option.
944 * @option: the option to parse
946 * Normally, standard mount options like "sync" are passed to file-systems as
947 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
948 * be present in the options string. This function tries to deal with this
949 * situation and parse standard options. Returns 0 if the option was not
950 * recognized, and the corresponding integer flag if it was.
952 * UBIFS is only interested in the "sync" option, so do not check for anything
953 * else.
955 static int parse_standard_option(const char *option)
957 ubifs_msg("parse %s", option);
958 if (!strcmp(option, "sync"))
959 return MS_SYNCHRONOUS;
960 return 0;
964 * ubifs_parse_options - parse mount parameters.
965 * @c: UBIFS file-system description object
966 * @options: parameters to parse
967 * @is_remount: non-zero if this is FS re-mount
969 * This function parses UBIFS mount options and returns zero in case success
970 * and a negative error code in case of failure.
972 static int ubifs_parse_options(struct ubifs_info *c, char *options,
973 int is_remount)
975 char *p;
976 substring_t args[MAX_OPT_ARGS];
978 if (!options)
979 return 0;
981 while ((p = strsep(&options, ","))) {
982 int token;
984 if (!*p)
985 continue;
987 token = match_token(p, tokens, args);
988 switch (token) {
990 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
991 * We accepte them in order to be backware-compatible. But this
992 * should be removed at some point.
994 case Opt_fast_unmount:
995 c->mount_opts.unmount_mode = 2;
996 break;
997 case Opt_norm_unmount:
998 c->mount_opts.unmount_mode = 1;
999 break;
1000 case Opt_bulk_read:
1001 c->mount_opts.bulk_read = 2;
1002 c->bulk_read = 1;
1003 break;
1004 case Opt_no_bulk_read:
1005 c->mount_opts.bulk_read = 1;
1006 c->bulk_read = 0;
1007 break;
1008 case Opt_chk_data_crc:
1009 c->mount_opts.chk_data_crc = 2;
1010 c->no_chk_data_crc = 0;
1011 break;
1012 case Opt_no_chk_data_crc:
1013 c->mount_opts.chk_data_crc = 1;
1014 c->no_chk_data_crc = 1;
1015 break;
1016 case Opt_override_compr:
1018 char *name = match_strdup(&args[0]);
1020 if (!name)
1021 return -ENOMEM;
1022 if (!strcmp(name, "none"))
1023 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1024 else if (!strcmp(name, "lzo"))
1025 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1026 else if (!strcmp(name, "zlib"))
1027 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1028 else {
1029 ubifs_err("unknown compressor \"%s\"", name);
1030 kfree(name);
1031 return -EINVAL;
1033 kfree(name);
1034 c->mount_opts.override_compr = 1;
1035 c->default_compr = c->mount_opts.compr_type;
1036 break;
1038 default:
1040 unsigned long flag;
1041 struct super_block *sb = c->vfs_sb;
1043 flag = parse_standard_option(p);
1044 if (!flag) {
1045 ubifs_err("unrecognized mount option \"%s\" "
1046 "or missing value", p);
1047 return -EINVAL;
1049 sb->s_flags |= flag;
1050 break;
1055 return 0;
1059 * destroy_journal - destroy journal data structures.
1060 * @c: UBIFS file-system description object
1062 * This function destroys journal data structures including those that may have
1063 * been created by recovery functions.
1065 static void destroy_journal(struct ubifs_info *c)
1067 while (!list_empty(&c->unclean_leb_list)) {
1068 struct ubifs_unclean_leb *ucleb;
1070 ucleb = list_entry(c->unclean_leb_list.next,
1071 struct ubifs_unclean_leb, list);
1072 list_del(&ucleb->list);
1073 kfree(ucleb);
1075 while (!list_empty(&c->old_buds)) {
1076 struct ubifs_bud *bud;
1078 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1079 list_del(&bud->list);
1080 kfree(bud);
1082 ubifs_destroy_idx_gc(c);
1083 ubifs_destroy_size_tree(c);
1084 ubifs_tnc_close(c);
1085 free_buds(c);
1089 * bu_init - initialize bulk-read information.
1090 * @c: UBIFS file-system description object
1092 static void bu_init(struct ubifs_info *c)
1094 ubifs_assert(c->bulk_read == 1);
1096 if (c->bu.buf)
1097 return; /* Already initialized */
1099 again:
1100 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1101 if (!c->bu.buf) {
1102 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1103 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1104 goto again;
1107 /* Just disable bulk-read */
1108 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1109 "disabling it", c->max_bu_buf_len);
1110 c->mount_opts.bulk_read = 1;
1111 c->bulk_read = 0;
1112 return;
1117 * check_free_space - check if there is enough free space to mount.
1118 * @c: UBIFS file-system description object
1120 * This function makes sure UBIFS has enough free space to be mounted in
1121 * read/write mode. UBIFS must always have some free space to allow deletions.
1123 static int check_free_space(struct ubifs_info *c)
1125 ubifs_assert(c->dark_wm > 0);
1126 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1127 ubifs_err("insufficient free space to mount in read/write mode");
1128 dbg_dump_budg(c);
1129 dbg_dump_lprops(c);
1130 return -ENOSPC;
1132 return 0;
1136 * mount_ubifs - mount UBIFS file-system.
1137 * @c: UBIFS file-system description object
1139 * This function mounts UBIFS file system. Returns zero in case of success and
1140 * a negative error code in case of failure.
1142 * Note, the function does not de-allocate resources it it fails half way
1143 * through, and the caller has to do this instead.
1145 static int mount_ubifs(struct ubifs_info *c)
1147 struct super_block *sb = c->vfs_sb;
1148 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1149 long long x;
1150 size_t sz;
1152 err = init_constants_early(c);
1153 if (err)
1154 return err;
1156 err = ubifs_debugging_init(c);
1157 if (err)
1158 return err;
1160 err = check_volume_empty(c);
1161 if (err)
1162 goto out_free;
1164 if (c->empty && (mounted_read_only || c->ro_media)) {
1166 * This UBI volume is empty, and read-only, or the file system
1167 * is mounted read-only - we cannot format it.
1169 ubifs_err("can't format empty UBI volume: read-only %s",
1170 c->ro_media ? "UBI volume" : "mount");
1171 err = -EROFS;
1172 goto out_free;
1175 if (c->ro_media && !mounted_read_only) {
1176 ubifs_err("cannot mount read-write - read-only media");
1177 err = -EROFS;
1178 goto out_free;
1182 * The requirement for the buffer is that it should fit indexing B-tree
1183 * height amount of integers. We assume the height if the TNC tree will
1184 * never exceed 64.
1186 err = -ENOMEM;
1187 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1188 if (!c->bottom_up_buf)
1189 goto out_free;
1191 c->sbuf = vmalloc(c->leb_size);
1192 if (!c->sbuf)
1193 goto out_free;
1195 if (!mounted_read_only) {
1196 c->ileb_buf = vmalloc(c->leb_size);
1197 if (!c->ileb_buf)
1198 goto out_free;
1201 if (c->bulk_read == 1)
1202 bu_init(c);
1205 * We have to check all CRCs, even for data nodes, when we mount the FS
1206 * (specifically, when we are replaying).
1208 c->always_chk_crc = 1;
1210 err = ubifs_read_superblock(c);
1211 if (err)
1212 goto out_free;
1215 * Make sure the compressor which is set as default in the superblock
1216 * or overridden by mount options is actually compiled in.
1218 if (!ubifs_compr_present(c->default_compr)) {
1219 ubifs_err("'compressor \"%s\" is not compiled in",
1220 ubifs_compr_name(c->default_compr));
1221 err = -ENOTSUPP;
1222 goto out_free;
1225 err = init_constants_sb(c);
1226 if (err)
1227 goto out_free;
1229 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1230 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1231 c->cbuf = kmalloc(sz, GFP_NOFS);
1232 if (!c->cbuf) {
1233 err = -ENOMEM;
1234 goto out_free;
1237 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1238 if (!mounted_read_only) {
1239 err = alloc_wbufs(c);
1240 if (err)
1241 goto out_cbuf;
1243 /* Create background thread */
1244 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1245 if (IS_ERR(c->bgt)) {
1246 err = PTR_ERR(c->bgt);
1247 c->bgt = NULL;
1248 ubifs_err("cannot spawn \"%s\", error %d",
1249 c->bgt_name, err);
1250 goto out_wbufs;
1252 wake_up_process(c->bgt);
1255 err = ubifs_read_master(c);
1256 if (err)
1257 goto out_master;
1259 init_constants_master(c);
1261 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1262 ubifs_msg("recovery needed");
1263 c->need_recovery = 1;
1264 if (!mounted_read_only) {
1265 err = ubifs_recover_inl_heads(c, c->sbuf);
1266 if (err)
1267 goto out_master;
1269 } else if (!mounted_read_only) {
1271 * Set the "dirty" flag so that if we reboot uncleanly we
1272 * will notice this immediately on the next mount.
1274 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1275 err = ubifs_write_master(c);
1276 if (err)
1277 goto out_master;
1280 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1281 if (err)
1282 goto out_lpt;
1284 err = dbg_check_idx_size(c, c->old_idx_sz);
1285 if (err)
1286 goto out_lpt;
1288 err = ubifs_replay_journal(c);
1289 if (err)
1290 goto out_journal;
1292 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1293 if (err)
1294 goto out_orphans;
1296 if (!mounted_read_only) {
1297 int lnum;
1299 err = check_free_space(c);
1300 if (err)
1301 goto out_orphans;
1303 /* Check for enough log space */
1304 lnum = c->lhead_lnum + 1;
1305 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1306 lnum = UBIFS_LOG_LNUM;
1307 if (lnum == c->ltail_lnum) {
1308 err = ubifs_consolidate_log(c);
1309 if (err)
1310 goto out_orphans;
1313 if (c->need_recovery) {
1314 err = ubifs_recover_size(c);
1315 if (err)
1316 goto out_orphans;
1317 err = ubifs_rcvry_gc_commit(c);
1318 } else {
1319 err = take_gc_lnum(c);
1320 if (err)
1321 goto out_orphans;
1324 * GC LEB may contain garbage if there was an unclean
1325 * reboot, and it should be un-mapped.
1327 err = ubifs_leb_unmap(c, c->gc_lnum);
1328 if (err)
1329 return err;
1332 err = dbg_check_lprops(c);
1333 if (err)
1334 goto out_orphans;
1335 } else if (c->need_recovery) {
1336 err = ubifs_recover_size(c);
1337 if (err)
1338 goto out_orphans;
1339 } else {
1341 * Even if we mount read-only, we have to set space in GC LEB
1342 * to proper value because this affects UBIFS free space
1343 * reporting. We do not want to have a situation when
1344 * re-mounting from R/O to R/W changes amount of free space.
1346 err = take_gc_lnum(c);
1347 if (err)
1348 goto out_orphans;
1351 spin_lock(&ubifs_infos_lock);
1352 list_add_tail(&c->infos_list, &ubifs_infos);
1353 spin_unlock(&ubifs_infos_lock);
1355 if (c->need_recovery) {
1356 if (mounted_read_only)
1357 ubifs_msg("recovery deferred");
1358 else {
1359 c->need_recovery = 0;
1360 ubifs_msg("recovery completed");
1362 * GC LEB has to be empty and taken at this point. But
1363 * the journal head LEBs may also be accounted as
1364 * "empty taken" if they are empty.
1366 ubifs_assert(c->lst.taken_empty_lebs > 0);
1368 } else
1369 ubifs_assert(c->lst.taken_empty_lebs > 0);
1371 err = dbg_check_filesystem(c);
1372 if (err)
1373 goto out_infos;
1375 err = dbg_debugfs_init_fs(c);
1376 if (err)
1377 goto out_infos;
1379 c->always_chk_crc = 0;
1381 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1382 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1383 if (mounted_read_only)
1384 ubifs_msg("mounted read-only");
1385 x = (long long)c->main_lebs * c->leb_size;
1386 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1387 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1388 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1389 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1390 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1391 ubifs_msg("media format: w%d/r%d (latest is w%d/r%d)",
1392 c->fmt_version, c->ro_compat_version,
1393 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1394 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1395 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1396 c->report_rp_size, c->report_rp_size >> 10);
1398 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1399 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1400 dbg_msg("LEB size: %d bytes (%d KiB)",
1401 c->leb_size, c->leb_size >> 10);
1402 dbg_msg("data journal heads: %d",
1403 c->jhead_cnt - NONDATA_JHEADS_CNT);
1404 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1405 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1406 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1407 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1408 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1409 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1410 dbg_msg("big_lpt %d", c->big_lpt);
1411 dbg_msg("log LEBs: %d (%d - %d)",
1412 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1413 dbg_msg("LPT area LEBs: %d (%d - %d)",
1414 c->lpt_lebs, c->lpt_first, c->lpt_last);
1415 dbg_msg("orphan area LEBs: %d (%d - %d)",
1416 c->orph_lebs, c->orph_first, c->orph_last);
1417 dbg_msg("main area LEBs: %d (%d - %d)",
1418 c->main_lebs, c->main_first, c->leb_cnt - 1);
1419 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1420 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1421 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1422 dbg_msg("key hash type: %d", c->key_hash_type);
1423 dbg_msg("tree fanout: %d", c->fanout);
1424 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1425 dbg_msg("first main LEB: %d", c->main_first);
1426 dbg_msg("max. znode size %d", c->max_znode_sz);
1427 dbg_msg("max. index node size %d", c->max_idx_node_sz);
1428 dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
1429 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1430 dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
1431 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1432 dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
1433 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1434 dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu",
1435 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1436 UBIFS_MAX_DENT_NODE_SZ);
1437 dbg_msg("dead watermark: %d", c->dead_wm);
1438 dbg_msg("dark watermark: %d", c->dark_wm);
1439 dbg_msg("LEB overhead: %d", c->leb_overhead);
1440 x = (long long)c->main_lebs * c->dark_wm;
1441 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1442 x, x >> 10, x >> 20);
1443 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1444 c->max_bud_bytes, c->max_bud_bytes >> 10,
1445 c->max_bud_bytes >> 20);
1446 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1447 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1448 c->bg_bud_bytes >> 20);
1449 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1450 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1451 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1452 dbg_msg("commit number: %llu", c->cmt_no);
1454 return 0;
1456 out_infos:
1457 spin_lock(&ubifs_infos_lock);
1458 list_del(&c->infos_list);
1459 spin_unlock(&ubifs_infos_lock);
1460 out_orphans:
1461 free_orphans(c);
1462 out_journal:
1463 destroy_journal(c);
1464 out_lpt:
1465 ubifs_lpt_free(c, 0);
1466 out_master:
1467 kfree(c->mst_node);
1468 kfree(c->rcvrd_mst_node);
1469 if (c->bgt)
1470 kthread_stop(c->bgt);
1471 out_wbufs:
1472 free_wbufs(c);
1473 out_cbuf:
1474 kfree(c->cbuf);
1475 out_free:
1476 kfree(c->bu.buf);
1477 vfree(c->ileb_buf);
1478 vfree(c->sbuf);
1479 kfree(c->bottom_up_buf);
1480 ubifs_debugging_exit(c);
1481 return err;
1485 * ubifs_umount - un-mount UBIFS file-system.
1486 * @c: UBIFS file-system description object
1488 * Note, this function is called to free allocated resourced when un-mounting,
1489 * as well as free resources when an error occurred while we were half way
1490 * through mounting (error path cleanup function). So it has to make sure the
1491 * resource was actually allocated before freeing it.
1493 static void ubifs_umount(struct ubifs_info *c)
1495 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1496 c->vi.vol_id);
1498 dbg_debugfs_exit_fs(c);
1499 spin_lock(&ubifs_infos_lock);
1500 list_del(&c->infos_list);
1501 spin_unlock(&ubifs_infos_lock);
1503 if (c->bgt)
1504 kthread_stop(c->bgt);
1506 destroy_journal(c);
1507 free_wbufs(c);
1508 free_orphans(c);
1509 ubifs_lpt_free(c, 0);
1511 kfree(c->cbuf);
1512 kfree(c->rcvrd_mst_node);
1513 kfree(c->mst_node);
1514 kfree(c->bu.buf);
1515 vfree(c->ileb_buf);
1516 vfree(c->sbuf);
1517 kfree(c->bottom_up_buf);
1518 ubifs_debugging_exit(c);
1522 * ubifs_remount_rw - re-mount in read-write mode.
1523 * @c: UBIFS file-system description object
1525 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1526 * mode. This function allocates the needed resources and re-mounts UBIFS in
1527 * read-write mode.
1529 static int ubifs_remount_rw(struct ubifs_info *c)
1531 int err, lnum;
1533 if (c->rw_incompat) {
1534 ubifs_err("the file-system is not R/W-compatible");
1535 ubifs_msg("on-flash format version is w%d/r%d, but software "
1536 "only supports up to version w%d/r%d", c->fmt_version,
1537 c->ro_compat_version, UBIFS_FORMAT_VERSION,
1538 UBIFS_RO_COMPAT_VERSION);
1539 return -EROFS;
1542 mutex_lock(&c->umount_mutex);
1543 dbg_save_space_info(c);
1544 c->remounting_rw = 1;
1545 c->always_chk_crc = 1;
1547 err = check_free_space(c);
1548 if (err)
1549 goto out;
1551 if (c->old_leb_cnt != c->leb_cnt) {
1552 struct ubifs_sb_node *sup;
1554 sup = ubifs_read_sb_node(c);
1555 if (IS_ERR(sup)) {
1556 err = PTR_ERR(sup);
1557 goto out;
1559 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1560 err = ubifs_write_sb_node(c, sup);
1561 if (err)
1562 goto out;
1565 if (c->need_recovery) {
1566 ubifs_msg("completing deferred recovery");
1567 err = ubifs_write_rcvrd_mst_node(c);
1568 if (err)
1569 goto out;
1570 err = ubifs_recover_size(c);
1571 if (err)
1572 goto out;
1573 err = ubifs_clean_lebs(c, c->sbuf);
1574 if (err)
1575 goto out;
1576 err = ubifs_recover_inl_heads(c, c->sbuf);
1577 if (err)
1578 goto out;
1579 } else {
1580 /* A readonly mount is not allowed to have orphans */
1581 ubifs_assert(c->tot_orphans == 0);
1582 err = ubifs_clear_orphans(c);
1583 if (err)
1584 goto out;
1587 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1588 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1589 err = ubifs_write_master(c);
1590 if (err)
1591 goto out;
1594 c->ileb_buf = vmalloc(c->leb_size);
1595 if (!c->ileb_buf) {
1596 err = -ENOMEM;
1597 goto out;
1600 err = ubifs_lpt_init(c, 0, 1);
1601 if (err)
1602 goto out;
1604 err = alloc_wbufs(c);
1605 if (err)
1606 goto out;
1608 ubifs_create_buds_lists(c);
1610 /* Create background thread */
1611 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1612 if (IS_ERR(c->bgt)) {
1613 err = PTR_ERR(c->bgt);
1614 c->bgt = NULL;
1615 ubifs_err("cannot spawn \"%s\", error %d",
1616 c->bgt_name, err);
1617 goto out;
1619 wake_up_process(c->bgt);
1621 c->orph_buf = vmalloc(c->leb_size);
1622 if (!c->orph_buf) {
1623 err = -ENOMEM;
1624 goto out;
1627 /* Check for enough log space */
1628 lnum = c->lhead_lnum + 1;
1629 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1630 lnum = UBIFS_LOG_LNUM;
1631 if (lnum == c->ltail_lnum) {
1632 err = ubifs_consolidate_log(c);
1633 if (err)
1634 goto out;
1637 if (c->need_recovery)
1638 err = ubifs_rcvry_gc_commit(c);
1639 else
1640 err = ubifs_leb_unmap(c, c->gc_lnum);
1641 if (err)
1642 goto out;
1644 if (c->need_recovery) {
1645 c->need_recovery = 0;
1646 ubifs_msg("deferred recovery completed");
1649 dbg_gen("re-mounted read-write");
1650 c->vfs_sb->s_flags &= ~MS_RDONLY;
1651 c->remounting_rw = 0;
1652 c->always_chk_crc = 0;
1653 err = dbg_check_space_info(c);
1654 mutex_unlock(&c->umount_mutex);
1655 return err;
1657 out:
1658 vfree(c->orph_buf);
1659 c->orph_buf = NULL;
1660 if (c->bgt) {
1661 kthread_stop(c->bgt);
1662 c->bgt = NULL;
1664 free_wbufs(c);
1665 vfree(c->ileb_buf);
1666 c->ileb_buf = NULL;
1667 ubifs_lpt_free(c, 1);
1668 c->remounting_rw = 0;
1669 c->always_chk_crc = 0;
1670 mutex_unlock(&c->umount_mutex);
1671 return err;
1675 * ubifs_remount_ro - re-mount in read-only mode.
1676 * @c: UBIFS file-system description object
1678 * We assume VFS has stopped writing. Possibly the background thread could be
1679 * running a commit, however kthread_stop will wait in that case.
1681 static void ubifs_remount_ro(struct ubifs_info *c)
1683 int i, err;
1685 ubifs_assert(!c->need_recovery);
1686 ubifs_assert(!(c->vfs_sb->s_flags & MS_RDONLY));
1688 mutex_lock(&c->umount_mutex);
1689 if (c->bgt) {
1690 kthread_stop(c->bgt);
1691 c->bgt = NULL;
1694 dbg_save_space_info(c);
1696 for (i = 0; i < c->jhead_cnt; i++) {
1697 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1698 del_timer_sync(&c->jheads[i].wbuf.timer);
1701 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1702 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1703 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1704 err = ubifs_write_master(c);
1705 if (err)
1706 ubifs_ro_mode(c, err);
1708 free_wbufs(c);
1709 vfree(c->orph_buf);
1710 c->orph_buf = NULL;
1711 vfree(c->ileb_buf);
1712 c->ileb_buf = NULL;
1713 ubifs_lpt_free(c, 1);
1714 err = dbg_check_space_info(c);
1715 if (err)
1716 ubifs_ro_mode(c, err);
1717 mutex_unlock(&c->umount_mutex);
1720 static void ubifs_put_super(struct super_block *sb)
1722 int i;
1723 struct ubifs_info *c = sb->s_fs_info;
1725 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1726 c->vi.vol_id);
1728 * The following asserts are only valid if there has not been a failure
1729 * of the media. For example, there will be dirty inodes if we failed
1730 * to write them back because of I/O errors.
1732 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1733 ubifs_assert(c->budg_idx_growth == 0);
1734 ubifs_assert(c->budg_dd_growth == 0);
1735 ubifs_assert(c->budg_data_growth == 0);
1738 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1739 * and file system un-mount. Namely, it prevents the shrinker from
1740 * picking this superblock for shrinking - it will be just skipped if
1741 * the mutex is locked.
1743 mutex_lock(&c->umount_mutex);
1744 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1746 * First of all kill the background thread to make sure it does
1747 * not interfere with un-mounting and freeing resources.
1749 if (c->bgt) {
1750 kthread_stop(c->bgt);
1751 c->bgt = NULL;
1754 /* Synchronize write-buffers */
1755 if (c->jheads)
1756 for (i = 0; i < c->jhead_cnt; i++) {
1757 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1758 del_timer_sync(&c->jheads[i].wbuf.timer);
1762 * On fatal errors c->ro_media is set to 1, in which case we do
1763 * not write the master node.
1765 if (!c->ro_media) {
1767 * We are being cleanly unmounted which means the
1768 * orphans were killed - indicate this in the master
1769 * node. Also save the reserved GC LEB number.
1771 int err;
1773 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1774 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1775 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1776 err = ubifs_write_master(c);
1777 if (err)
1779 * Recovery will attempt to fix the master area
1780 * next mount, so we just print a message and
1781 * continue to unmount normally.
1783 ubifs_err("failed to write master node, "
1784 "error %d", err);
1788 ubifs_umount(c);
1789 bdi_destroy(&c->bdi);
1790 ubi_close_volume(c->ubi);
1791 mutex_unlock(&c->umount_mutex);
1792 kfree(c);
1795 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1797 int err;
1798 struct ubifs_info *c = sb->s_fs_info;
1800 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1802 err = ubifs_parse_options(c, data, 1);
1803 if (err) {
1804 ubifs_err("invalid or unknown remount parameter");
1805 return err;
1808 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1809 if (c->ro_media) {
1810 ubifs_msg("cannot re-mount due to prior errors");
1811 return -EROFS;
1813 err = ubifs_remount_rw(c);
1814 if (err)
1815 return err;
1816 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY)) {
1817 if (c->ro_media) {
1818 ubifs_msg("cannot re-mount due to prior errors");
1819 return -EROFS;
1821 ubifs_remount_ro(c);
1824 if (c->bulk_read == 1)
1825 bu_init(c);
1826 else {
1827 dbg_gen("disable bulk-read");
1828 kfree(c->bu.buf);
1829 c->bu.buf = NULL;
1832 ubifs_assert(c->lst.taken_empty_lebs > 0);
1833 return 0;
1836 const struct super_operations ubifs_super_operations = {
1837 .alloc_inode = ubifs_alloc_inode,
1838 .destroy_inode = ubifs_destroy_inode,
1839 .put_super = ubifs_put_super,
1840 .write_inode = ubifs_write_inode,
1841 .delete_inode = ubifs_delete_inode,
1842 .statfs = ubifs_statfs,
1843 .dirty_inode = ubifs_dirty_inode,
1844 .remount_fs = ubifs_remount_fs,
1845 .show_options = ubifs_show_options,
1846 .sync_fs = ubifs_sync_fs,
1850 * open_ubi - parse UBI device name string and open the UBI device.
1851 * @name: UBI volume name
1852 * @mode: UBI volume open mode
1854 * There are several ways to specify UBI volumes when mounting UBIFS:
1855 * o ubiX_Y - UBI device number X, volume Y;
1856 * o ubiY - UBI device number 0, volume Y;
1857 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1858 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1860 * Alternative '!' separator may be used instead of ':' (because some shells
1861 * like busybox may interpret ':' as an NFS host name separator). This function
1862 * returns ubi volume object in case of success and a negative error code in
1863 * case of failure.
1865 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1867 int dev, vol;
1868 char *endptr;
1870 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1871 return ERR_PTR(-EINVAL);
1873 /* ubi:NAME method */
1874 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1875 return ubi_open_volume_nm(0, name + 4, mode);
1877 if (!isdigit(name[3]))
1878 return ERR_PTR(-EINVAL);
1880 dev = simple_strtoul(name + 3, &endptr, 0);
1882 /* ubiY method */
1883 if (*endptr == '\0')
1884 return ubi_open_volume(0, dev, mode);
1886 /* ubiX_Y method */
1887 if (*endptr == '_' && isdigit(endptr[1])) {
1888 vol = simple_strtoul(endptr + 1, &endptr, 0);
1889 if (*endptr != '\0')
1890 return ERR_PTR(-EINVAL);
1891 return ubi_open_volume(dev, vol, mode);
1894 /* ubiX:NAME method */
1895 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1896 return ubi_open_volume_nm(dev, ++endptr, mode);
1898 return ERR_PTR(-EINVAL);
1901 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1903 struct ubi_volume_desc *ubi = sb->s_fs_info;
1904 struct ubifs_info *c;
1905 struct inode *root;
1906 int err;
1908 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1909 if (!c)
1910 return -ENOMEM;
1912 spin_lock_init(&c->cnt_lock);
1913 spin_lock_init(&c->cs_lock);
1914 spin_lock_init(&c->buds_lock);
1915 spin_lock_init(&c->space_lock);
1916 spin_lock_init(&c->orphan_lock);
1917 init_rwsem(&c->commit_sem);
1918 mutex_init(&c->lp_mutex);
1919 mutex_init(&c->tnc_mutex);
1920 mutex_init(&c->log_mutex);
1921 mutex_init(&c->mst_mutex);
1922 mutex_init(&c->umount_mutex);
1923 mutex_init(&c->bu_mutex);
1924 init_waitqueue_head(&c->cmt_wq);
1925 c->buds = RB_ROOT;
1926 c->old_idx = RB_ROOT;
1927 c->size_tree = RB_ROOT;
1928 c->orph_tree = RB_ROOT;
1929 INIT_LIST_HEAD(&c->infos_list);
1930 INIT_LIST_HEAD(&c->idx_gc);
1931 INIT_LIST_HEAD(&c->replay_list);
1932 INIT_LIST_HEAD(&c->replay_buds);
1933 INIT_LIST_HEAD(&c->uncat_list);
1934 INIT_LIST_HEAD(&c->empty_list);
1935 INIT_LIST_HEAD(&c->freeable_list);
1936 INIT_LIST_HEAD(&c->frdi_idx_list);
1937 INIT_LIST_HEAD(&c->unclean_leb_list);
1938 INIT_LIST_HEAD(&c->old_buds);
1939 INIT_LIST_HEAD(&c->orph_list);
1940 INIT_LIST_HEAD(&c->orph_new);
1942 c->vfs_sb = sb;
1943 c->highest_inum = UBIFS_FIRST_INO;
1944 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1946 ubi_get_volume_info(ubi, &c->vi);
1947 ubi_get_device_info(c->vi.ubi_num, &c->di);
1949 /* Re-open the UBI device in read-write mode */
1950 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1951 if (IS_ERR(c->ubi)) {
1952 err = PTR_ERR(c->ubi);
1953 goto out_free;
1957 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1958 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1959 * which means the user would have to wait not just for their own I/O
1960 * but the read-ahead I/O as well i.e. completely pointless.
1962 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1964 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1965 c->bdi.unplug_io_fn = default_unplug_io_fn;
1966 err = bdi_init(&c->bdi);
1967 if (err)
1968 goto out_close;
1969 err = bdi_register(&c->bdi, NULL, "ubifs");
1970 if (err)
1971 goto out_bdi;
1973 err = ubifs_parse_options(c, data, 0);
1974 if (err)
1975 goto out_bdi;
1977 sb->s_fs_info = c;
1978 sb->s_magic = UBIFS_SUPER_MAGIC;
1979 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1980 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1981 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1982 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1983 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1984 sb->s_op = &ubifs_super_operations;
1986 mutex_lock(&c->umount_mutex);
1987 err = mount_ubifs(c);
1988 if (err) {
1989 ubifs_assert(err < 0);
1990 goto out_unlock;
1993 /* Read the root inode */
1994 root = ubifs_iget(sb, UBIFS_ROOT_INO);
1995 if (IS_ERR(root)) {
1996 err = PTR_ERR(root);
1997 goto out_umount;
2000 sb->s_root = d_alloc_root(root);
2001 if (!sb->s_root)
2002 goto out_iput;
2004 mutex_unlock(&c->umount_mutex);
2005 return 0;
2007 out_iput:
2008 iput(root);
2009 out_umount:
2010 ubifs_umount(c);
2011 out_unlock:
2012 mutex_unlock(&c->umount_mutex);
2013 out_bdi:
2014 bdi_destroy(&c->bdi);
2015 out_close:
2016 ubi_close_volume(c->ubi);
2017 out_free:
2018 kfree(c);
2019 return err;
2022 static int sb_test(struct super_block *sb, void *data)
2024 dev_t *dev = data;
2025 struct ubifs_info *c = sb->s_fs_info;
2027 return c->vi.cdev == *dev;
2030 static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
2031 const char *name, void *data, struct vfsmount *mnt)
2033 struct ubi_volume_desc *ubi;
2034 struct ubi_volume_info vi;
2035 struct super_block *sb;
2036 int err;
2038 dbg_gen("name %s, flags %#x", name, flags);
2041 * Get UBI device number and volume ID. Mount it read-only so far
2042 * because this might be a new mount point, and UBI allows only one
2043 * read-write user at a time.
2045 ubi = open_ubi(name, UBI_READONLY);
2046 if (IS_ERR(ubi)) {
2047 ubifs_err("cannot open \"%s\", error %d",
2048 name, (int)PTR_ERR(ubi));
2049 return PTR_ERR(ubi);
2051 ubi_get_volume_info(ubi, &vi);
2053 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
2055 sb = sget(fs_type, &sb_test, &set_anon_super, &vi.cdev);
2056 if (IS_ERR(sb)) {
2057 err = PTR_ERR(sb);
2058 goto out_close;
2061 if (sb->s_root) {
2062 /* A new mount point for already mounted UBIFS */
2063 dbg_gen("this ubi volume is already mounted");
2064 if ((flags ^ sb->s_flags) & MS_RDONLY) {
2065 err = -EBUSY;
2066 goto out_deact;
2068 } else {
2069 sb->s_flags = flags;
2071 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
2072 * replaced by 'c'.
2074 sb->s_fs_info = ubi;
2075 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2076 if (err)
2077 goto out_deact;
2078 /* We do not support atime */
2079 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2082 /* 'fill_super()' opens ubi again so we must close it here */
2083 ubi_close_volume(ubi);
2085 simple_set_mnt(mnt, sb);
2086 return 0;
2088 out_deact:
2089 up_write(&sb->s_umount);
2090 deactivate_super(sb);
2091 out_close:
2092 ubi_close_volume(ubi);
2093 return err;
2096 static struct file_system_type ubifs_fs_type = {
2097 .name = "ubifs",
2098 .owner = THIS_MODULE,
2099 .get_sb = ubifs_get_sb,
2100 .kill_sb = kill_anon_super,
2104 * Inode slab cache constructor.
2106 static void inode_slab_ctor(void *obj)
2108 struct ubifs_inode *ui = obj;
2109 inode_init_once(&ui->vfs_inode);
2112 static int __init ubifs_init(void)
2114 int err;
2116 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2118 /* Make sure node sizes are 8-byte aligned */
2119 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2120 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2121 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2122 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2123 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2124 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2125 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2126 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2127 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2128 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2129 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2131 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2132 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2133 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2134 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2135 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2136 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2138 /* Check min. node size */
2139 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2140 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2141 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2142 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2144 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2145 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2146 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2147 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2149 /* Defined node sizes */
2150 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2151 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2152 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2153 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2156 * We use 2 bit wide bit-fields to store compression type, which should
2157 * be amended if more compressors are added. The bit-fields are:
2158 * @compr_type in 'struct ubifs_inode', @default_compr in
2159 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2161 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2164 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2165 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2167 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2168 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2169 " at least 4096 bytes",
2170 (unsigned int)PAGE_CACHE_SIZE);
2171 return -EINVAL;
2174 err = register_filesystem(&ubifs_fs_type);
2175 if (err) {
2176 ubifs_err("cannot register file system, error %d", err);
2177 return err;
2180 err = -ENOMEM;
2181 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2182 sizeof(struct ubifs_inode), 0,
2183 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2184 &inode_slab_ctor);
2185 if (!ubifs_inode_slab)
2186 goto out_reg;
2188 register_shrinker(&ubifs_shrinker_info);
2190 err = ubifs_compressors_init();
2191 if (err)
2192 goto out_shrinker;
2194 err = dbg_debugfs_init();
2195 if (err)
2196 goto out_compr;
2198 return 0;
2200 out_compr:
2201 ubifs_compressors_exit();
2202 out_shrinker:
2203 unregister_shrinker(&ubifs_shrinker_info);
2204 kmem_cache_destroy(ubifs_inode_slab);
2205 out_reg:
2206 unregister_filesystem(&ubifs_fs_type);
2207 return err;
2209 /* late_initcall to let compressors initialize first */
2210 late_initcall(ubifs_init);
2212 static void __exit ubifs_exit(void)
2214 ubifs_assert(list_empty(&ubifs_infos));
2215 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2217 dbg_debugfs_exit();
2218 ubifs_compressors_exit();
2219 unregister_shrinker(&ubifs_shrinker_info);
2220 kmem_cache_destroy(ubifs_inode_slab);
2221 unregister_filesystem(&ubifs_fs_type);
2223 module_exit(ubifs_exit);
2225 MODULE_LICENSE("GPL");
2226 MODULE_VERSION(__stringify(UBIFS_VERSION));
2227 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2228 MODULE_DESCRIPTION("UBIFS - UBI File System");