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
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)
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>
39 /* Slab cache for UBIFS inodes */
40 struct kmem_cache *ubifs_inode_slab;
42 /* UBIFS TNC shrinker description */
43 static struct shrinker ubifs_shrinker_info = {
44 .shrink = ubifs_shrinker,
45 .seeks = DEFAULT_SEEKS,
49 * validate_inode - validate inode.
50 * @c: UBIFS file-system description object
51 * @inode: the inode to validate
53 * This is a helper function for 'ubifs_iget()' which validates various fields
54 * of a newly built inode to make sure they contain sane values and prevent
55 * possible vulnerabilities. Returns zero if the inode is all right and
56 * a non-zero error code if not.
58 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
61 const struct ubifs_inode *ui = ubifs_inode(inode);
63 if (inode->i_size > c->max_inode_sz) {
64 ubifs_err("inode is too large (%lld)",
65 (long long)inode->i_size);
69 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
70 ubifs_err("unknown compression type %d", ui->compr_type);
74 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
77 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
80 if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
83 if (!ubifs_compr_present(ui->compr_type)) {
84 ubifs_warn("inode %lu uses '%s' compression, but it was not "
85 "compiled in", inode->i_ino,
86 ubifs_compr_name(ui->compr_type));
89 err = dbg_check_dir_size(c, inode);
93 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
97 struct ubifs_ino_node *ino;
98 struct ubifs_info *c = sb->s_fs_info;
100 struct ubifs_inode *ui;
102 dbg_gen("inode %lu", inum);
104 inode = iget_locked(sb, inum);
106 return ERR_PTR(-ENOMEM);
107 if (!(inode->i_state & I_NEW))
109 ui = ubifs_inode(inode);
111 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
117 ino_key_init(c, &key, inode->i_ino);
119 err = ubifs_tnc_lookup(c, &key, ino);
123 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
124 inode->i_nlink = le32_to_cpu(ino->nlink);
125 inode->i_uid = le32_to_cpu(ino->uid);
126 inode->i_gid = le32_to_cpu(ino->gid);
127 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
128 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
129 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
130 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
131 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
132 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
133 inode->i_mode = le32_to_cpu(ino->mode);
134 inode->i_size = le64_to_cpu(ino->size);
136 ui->data_len = le32_to_cpu(ino->data_len);
137 ui->flags = le32_to_cpu(ino->flags);
138 ui->compr_type = le16_to_cpu(ino->compr_type);
139 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
140 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
141 ui->xattr_size = le32_to_cpu(ino->xattr_size);
142 ui->xattr_names = le32_to_cpu(ino->xattr_names);
143 ui->synced_i_size = ui->ui_size = inode->i_size;
145 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
147 err = validate_inode(c, inode);
151 /* Disable read-ahead */
152 inode->i_mapping->backing_dev_info = &c->bdi;
154 switch (inode->i_mode & S_IFMT) {
156 inode->i_mapping->a_ops = &ubifs_file_address_operations;
157 inode->i_op = &ubifs_file_inode_operations;
158 inode->i_fop = &ubifs_file_operations;
160 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
165 memcpy(ui->data, ino->data, ui->data_len);
166 ((char *)ui->data)[ui->data_len] = '\0';
167 } else if (ui->data_len != 0) {
173 inode->i_op = &ubifs_dir_inode_operations;
174 inode->i_fop = &ubifs_dir_operations;
175 if (ui->data_len != 0) {
181 inode->i_op = &ubifs_symlink_inode_operations;
182 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
186 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
191 memcpy(ui->data, ino->data, ui->data_len);
192 ((char *)ui->data)[ui->data_len] = '\0';
198 union ubifs_dev_desc *dev;
200 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
206 dev = (union ubifs_dev_desc *)ino->data;
207 if (ui->data_len == sizeof(dev->new))
208 rdev = new_decode_dev(le32_to_cpu(dev->new));
209 else if (ui->data_len == sizeof(dev->huge))
210 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
215 memcpy(ui->data, ino->data, ui->data_len);
216 inode->i_op = &ubifs_file_inode_operations;
217 init_special_inode(inode, inode->i_mode, rdev);
222 inode->i_op = &ubifs_file_inode_operations;
223 init_special_inode(inode, inode->i_mode, 0);
224 if (ui->data_len != 0) {
235 ubifs_set_inode_flags(inode);
236 unlock_new_inode(inode);
240 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
241 dbg_dump_node(c, ino);
242 dbg_dump_inode(c, inode);
247 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
252 static struct inode *ubifs_alloc_inode(struct super_block *sb)
254 struct ubifs_inode *ui;
256 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
260 memset((void *)ui + sizeof(struct inode), 0,
261 sizeof(struct ubifs_inode) - sizeof(struct inode));
262 mutex_init(&ui->ui_mutex);
263 spin_lock_init(&ui->ui_lock);
264 return &ui->vfs_inode;
267 static void ubifs_destroy_inode(struct inode *inode)
269 struct ubifs_inode *ui = ubifs_inode(inode);
272 kmem_cache_free(ubifs_inode_slab, inode);
276 * Note, Linux write-back code calls this without 'i_mutex'.
278 static int ubifs_write_inode(struct inode *inode, int wait)
281 struct ubifs_info *c = inode->i_sb->s_fs_info;
282 struct ubifs_inode *ui = ubifs_inode(inode);
284 ubifs_assert(!ui->xattr);
285 if (is_bad_inode(inode))
288 mutex_lock(&ui->ui_mutex);
290 * Due to races between write-back forced by budgeting
291 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
292 * have already been synchronized, do not do this again. This might
293 * also happen if it was synchronized in an VFS operation, e.g.
297 mutex_unlock(&ui->ui_mutex);
302 * As an optimization, do not write orphan inodes to the media just
303 * because this is not needed.
305 dbg_gen("inode %lu, mode %#x, nlink %u",
306 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
307 if (inode->i_nlink) {
308 err = ubifs_jnl_write_inode(c, inode);
310 ubifs_err("can't write inode %lu, error %d",
315 mutex_unlock(&ui->ui_mutex);
316 ubifs_release_dirty_inode_budget(c, ui);
320 static void ubifs_delete_inode(struct inode *inode)
323 struct ubifs_info *c = inode->i_sb->s_fs_info;
324 struct ubifs_inode *ui = ubifs_inode(inode);
328 * Extended attribute inode deletions are fully handled in
329 * 'ubifs_removexattr()'. These inodes are special and have
330 * limited usage, so there is nothing to do here.
334 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
335 ubifs_assert(!atomic_read(&inode->i_count));
336 ubifs_assert(inode->i_nlink == 0);
338 truncate_inode_pages(&inode->i_data, 0);
339 if (is_bad_inode(inode))
342 ui->ui_size = inode->i_size = 0;
343 err = ubifs_jnl_delete_inode(c, inode);
346 * Worst case we have a lost orphan inode wasting space, so a
347 * simple error message is OK here.
349 ubifs_err("can't delete inode %lu, error %d",
354 ubifs_release_dirty_inode_budget(c, ui);
358 static void ubifs_dirty_inode(struct inode *inode)
360 struct ubifs_inode *ui = ubifs_inode(inode);
362 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
365 dbg_gen("inode %lu", inode->i_ino);
369 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
371 struct ubifs_info *c = dentry->d_sb->s_fs_info;
372 unsigned long long free;
374 free = ubifs_budg_get_free_space(c);
375 dbg_gen("free space %lld bytes (%lld blocks)",
376 free, free >> UBIFS_BLOCK_SHIFT);
378 buf->f_type = UBIFS_SUPER_MAGIC;
379 buf->f_bsize = UBIFS_BLOCK_SIZE;
380 buf->f_blocks = c->block_cnt;
381 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
382 if (free > c->report_rp_size)
383 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
388 buf->f_namelen = UBIFS_MAX_NLEN;
393 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
395 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
397 if (c->mount_opts.unmount_mode == 2)
398 seq_printf(s, ",fast_unmount");
399 else if (c->mount_opts.unmount_mode == 1)
400 seq_printf(s, ",norm_unmount");
405 static int ubifs_sync_fs(struct super_block *sb, int wait)
407 struct ubifs_info *c = sb->s_fs_info;
411 for (i = 0; i < c->jhead_cnt; i++) {
412 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
417 * We ought to call sync for c->ubi but it does not have one. If it had
418 * it would in turn call mtd->sync, however mtd operations are
419 * synchronous anyway, so we don't lose any sleep here.
425 * init_constants_early - initialize UBIFS constants.
426 * @c: UBIFS file-system description object
428 * This function initialize UBIFS constants which do not need the superblock to
429 * be read. It also checks that the UBI volume satisfies basic UBIFS
430 * requirements. Returns zero in case of success and a negative error code in
433 static int init_constants_early(struct ubifs_info *c)
435 if (c->vi.corrupted) {
436 ubifs_warn("UBI volume is corrupted - read-only mode");
441 ubifs_msg("read-only UBI device");
445 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
446 ubifs_msg("static UBI volume - read-only mode");
450 c->leb_cnt = c->vi.size;
451 c->leb_size = c->vi.usable_leb_size;
452 c->half_leb_size = c->leb_size / 2;
453 c->min_io_size = c->di.min_io_size;
454 c->min_io_shift = fls(c->min_io_size) - 1;
456 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
457 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
458 c->leb_size, UBIFS_MIN_LEB_SZ);
462 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
463 ubifs_err("too few LEBs (%d), min. is %d",
464 c->leb_cnt, UBIFS_MIN_LEB_CNT);
468 if (!is_power_of_2(c->min_io_size)) {
469 ubifs_err("bad min. I/O size %d", c->min_io_size);
474 * UBIFS aligns all node to 8-byte boundary, so to make function in
475 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
478 if (c->min_io_size < 8) {
483 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
484 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
487 * Initialize node length ranges which are mostly needed for node
490 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
491 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
492 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
493 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
494 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
495 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
497 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
498 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
499 c->ranges[UBIFS_ORPH_NODE].min_len =
500 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
501 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
502 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
503 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
504 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
505 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
506 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
507 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
509 * Minimum indexing node size is amended later when superblock is
510 * read and the key length is known.
512 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
514 * Maximum indexing node size is amended later when superblock is
515 * read and the fanout is known.
517 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
520 * Initialize dead and dark LEB space watermarks.
522 * Dead space is the space which cannot be used. Its watermark is
523 * equivalent to min. I/O unit or minimum node size if it is greater
524 * then min. I/O unit.
526 * Dark space is the space which might be used, or might not, depending
527 * on which node should be written to the LEB. Its watermark is
528 * equivalent to maximum UBIFS node size.
530 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
531 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
537 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
538 * @c: UBIFS file-system description object
539 * @lnum: LEB the write-buffer was synchronized to
540 * @free: how many free bytes left in this LEB
541 * @pad: how many bytes were padded
543 * This is a callback function which is called by the I/O unit when the
544 * write-buffer is synchronized. We need this to correctly maintain space
545 * accounting in bud logical eraseblocks. This function returns zero in case of
546 * success and a negative error code in case of failure.
548 * This function actually belongs to the journal, but we keep it here because
549 * we want to keep it static.
551 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
553 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
557 * init_constants_late - initialize UBIFS constants.
558 * @c: UBIFS file-system description object
560 * This is a helper function which initializes various UBIFS constants after
561 * the superblock has been read. It also checks various UBIFS parameters and
562 * makes sure they are all right. Returns zero in case of success and a
563 * negative error code in case of failure.
565 static int init_constants_late(struct ubifs_info *c)
570 c->main_bytes = (long long)c->main_lebs * c->leb_size;
571 c->max_znode_sz = sizeof(struct ubifs_znode) +
572 c->fanout * sizeof(struct ubifs_zbranch);
574 tmp = ubifs_idx_node_sz(c, 1);
575 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
576 c->min_idx_node_sz = ALIGN(tmp, 8);
578 tmp = ubifs_idx_node_sz(c, c->fanout);
579 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
580 c->max_idx_node_sz = ALIGN(tmp, 8);
582 /* Make sure LEB size is large enough to fit full commit */
583 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
584 tmp = ALIGN(tmp, c->min_io_size);
585 if (tmp > c->leb_size) {
586 dbg_err("too small LEB size %d, at least %d needed",
592 * Make sure that the log is large enough to fit reference nodes for
593 * all buds plus one reserved LEB.
595 tmp64 = c->max_bud_bytes;
596 tmp = do_div(tmp64, c->leb_size);
597 c->max_bud_cnt = tmp64 + !!tmp;
598 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
601 if (c->log_lebs < tmp) {
602 dbg_err("too small log %d LEBs, required min. %d LEBs",
608 * When budgeting we assume worst-case scenarios when the pages are not
609 * be compressed and direntries are of the maximum size.
611 * Note, data, which may be stored in inodes is budgeted separately, so
612 * it is not included into 'c->inode_budget'.
614 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
615 c->inode_budget = UBIFS_INO_NODE_SZ;
616 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
619 * When the amount of flash space used by buds becomes
620 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
621 * The writers are unblocked when the commit is finished. To avoid
622 * writers to be blocked UBIFS initiates background commit in advance,
623 * when number of bud bytes becomes above the limit defined below.
625 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
628 * Ensure minimum journal size. All the bytes in the journal heads are
629 * considered to be used, when calculating the current journal usage.
630 * Consequently, if the journal is too small, UBIFS will treat it as
633 tmp64 = (uint64_t)(c->jhead_cnt + 1) * c->leb_size + 1;
634 if (c->bg_bud_bytes < tmp64)
635 c->bg_bud_bytes = tmp64;
636 if (c->max_bud_bytes < tmp64 + c->leb_size)
637 c->max_bud_bytes = tmp64 + c->leb_size;
639 err = ubifs_calc_lpt_geom(c);
643 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
646 * Calculate total amount of FS blocks. This number is not used
647 * internally because it does not make much sense for UBIFS, but it is
648 * necessary to report something for the 'statfs()' call.
650 * Subtract the LEB reserved for GC and the LEB which is reserved for
653 * Review 'ubifs_calc_available()' if changing this calculation.
655 tmp64 = c->main_lebs - 2;
656 tmp64 *= (uint64_t)c->leb_size - c->dark_wm;
657 tmp64 = ubifs_reported_space(c, tmp64);
658 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
664 * take_gc_lnum - reserve GC LEB.
665 * @c: UBIFS file-system description object
667 * This function ensures that the LEB reserved for garbage collection is
668 * unmapped and is marked as "taken" in lprops. We also have to set free space
669 * to LEB size and dirty space to zero, because lprops may contain out-of-date
670 * information if the file-system was un-mounted before it has been committed.
671 * This function returns zero in case of success and a negative error code in
674 static int take_gc_lnum(struct ubifs_info *c)
678 if (c->gc_lnum == -1) {
679 ubifs_err("no LEB for GC");
683 err = ubifs_leb_unmap(c, c->gc_lnum);
687 /* And we have to tell lprops that this LEB is taken */
688 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
694 * alloc_wbufs - allocate write-buffers.
695 * @c: UBIFS file-system description object
697 * This helper function allocates and initializes UBIFS write-buffers. Returns
698 * zero in case of success and %-ENOMEM in case of failure.
700 static int alloc_wbufs(struct ubifs_info *c)
704 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
709 /* Initialize journal heads */
710 for (i = 0; i < c->jhead_cnt; i++) {
711 INIT_LIST_HEAD(&c->jheads[i].buds_list);
712 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
716 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
717 c->jheads[i].wbuf.jhead = i;
720 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
722 * Garbage Collector head likely contains long-term data and
723 * does not need to be synchronized by timer.
725 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
726 c->jheads[GCHD].wbuf.timeout = 0;
732 * free_wbufs - free write-buffers.
733 * @c: UBIFS file-system description object
735 static void free_wbufs(struct ubifs_info *c)
740 for (i = 0; i < c->jhead_cnt; i++) {
741 kfree(c->jheads[i].wbuf.buf);
742 kfree(c->jheads[i].wbuf.inodes);
750 * free_orphans - free orphans.
751 * @c: UBIFS file-system description object
753 static void free_orphans(struct ubifs_info *c)
755 struct ubifs_orphan *orph;
757 while (c->orph_dnext) {
758 orph = c->orph_dnext;
759 c->orph_dnext = orph->dnext;
760 list_del(&orph->list);
764 while (!list_empty(&c->orph_list)) {
765 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
766 list_del(&orph->list);
768 dbg_err("orphan list not empty at unmount");
776 * free_buds - free per-bud objects.
777 * @c: UBIFS file-system description object
779 static void free_buds(struct ubifs_info *c)
781 struct rb_node *this = c->buds.rb_node;
782 struct ubifs_bud *bud;
786 this = this->rb_left;
787 else if (this->rb_right)
788 this = this->rb_right;
790 bud = rb_entry(this, struct ubifs_bud, rb);
791 this = rb_parent(this);
793 if (this->rb_left == &bud->rb)
794 this->rb_left = NULL;
796 this->rb_right = NULL;
804 * check_volume_empty - check if the UBI volume is empty.
805 * @c: UBIFS file-system description object
807 * This function checks if the UBIFS volume is empty by looking if its LEBs are
808 * mapped or not. The result of checking is stored in the @c->empty variable.
809 * Returns zero in case of success and a negative error code in case of
812 static int check_volume_empty(struct ubifs_info *c)
817 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
818 err = ubi_is_mapped(c->ubi, lnum);
819 if (unlikely(err < 0))
833 * UBIFS mount options.
835 * Opt_fast_unmount: do not run a journal commit before un-mounting
836 * Opt_norm_unmount: run a journal commit before un-mounting
837 * Opt_err: just end of array marker
845 static match_table_t tokens = {
846 {Opt_fast_unmount, "fast_unmount"},
847 {Opt_norm_unmount, "norm_unmount"},
852 * ubifs_parse_options - parse mount parameters.
853 * @c: UBIFS file-system description object
854 * @options: parameters to parse
855 * @is_remount: non-zero if this is FS re-mount
857 * This function parses UBIFS mount options and returns zero in case success
858 * and a negative error code in case of failure.
860 static int ubifs_parse_options(struct ubifs_info *c, char *options,
864 substring_t args[MAX_OPT_ARGS];
869 while ((p = strsep(&options, ","))) {
875 token = match_token(p, tokens, args);
877 case Opt_fast_unmount:
878 c->mount_opts.unmount_mode = 2;
881 case Opt_norm_unmount:
882 c->mount_opts.unmount_mode = 1;
886 ubifs_err("unrecognized mount option \"%s\" "
887 "or missing value", p);
896 * destroy_journal - destroy journal data structures.
897 * @c: UBIFS file-system description object
899 * This function destroys journal data structures including those that may have
900 * been created by recovery functions.
902 static void destroy_journal(struct ubifs_info *c)
904 while (!list_empty(&c->unclean_leb_list)) {
905 struct ubifs_unclean_leb *ucleb;
907 ucleb = list_entry(c->unclean_leb_list.next,
908 struct ubifs_unclean_leb, list);
909 list_del(&ucleb->list);
912 while (!list_empty(&c->old_buds)) {
913 struct ubifs_bud *bud;
915 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
916 list_del(&bud->list);
919 ubifs_destroy_idx_gc(c);
920 ubifs_destroy_size_tree(c);
926 * mount_ubifs - mount UBIFS file-system.
927 * @c: UBIFS file-system description object
929 * This function mounts UBIFS file system. Returns zero in case of success and
930 * a negative error code in case of failure.
932 * Note, the function does not de-allocate resources it it fails half way
933 * through, and the caller has to do this instead.
935 static int mount_ubifs(struct ubifs_info *c)
937 struct super_block *sb = c->vfs_sb;
938 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
942 err = init_constants_early(c);
946 #ifdef CONFIG_UBIFS_FS_DEBUG
947 c->dbg_buf = vmalloc(c->leb_size);
952 err = check_volume_empty(c);
956 if (c->empty && (mounted_read_only || c->ro_media)) {
958 * This UBI volume is empty, and read-only, or the file system
959 * is mounted read-only - we cannot format it.
961 ubifs_err("can't format empty UBI volume: read-only %s",
962 c->ro_media ? "UBI volume" : "mount");
967 if (c->ro_media && !mounted_read_only) {
968 ubifs_err("cannot mount read-write - read-only media");
974 * The requirement for the buffer is that it should fit indexing B-tree
975 * height amount of integers. We assume the height if the TNC tree will
979 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
980 if (!c->bottom_up_buf)
983 c->sbuf = vmalloc(c->leb_size);
987 if (!mounted_read_only) {
988 c->ileb_buf = vmalloc(c->leb_size);
993 err = ubifs_read_superblock(c);
998 * Make sure the compressor which is set as the default on in the
999 * superblock was actually compiled in.
1001 if (!ubifs_compr_present(c->default_compr)) {
1002 ubifs_warn("'%s' compressor is set by superblock, but not "
1003 "compiled in", ubifs_compr_name(c->default_compr));
1004 c->default_compr = UBIFS_COMPR_NONE;
1007 dbg_failure_mode_registration(c);
1009 err = init_constants_late(c);
1013 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1014 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1015 c->cbuf = kmalloc(sz, GFP_NOFS);
1021 if (!mounted_read_only) {
1022 err = alloc_wbufs(c);
1026 /* Create background thread */
1027 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num,
1029 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1031 c->bgt = ERR_PTR(-EINVAL);
1032 if (IS_ERR(c->bgt)) {
1033 err = PTR_ERR(c->bgt);
1035 ubifs_err("cannot spawn \"%s\", error %d",
1039 wake_up_process(c->bgt);
1042 err = ubifs_read_master(c);
1046 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1047 ubifs_msg("recovery needed");
1048 c->need_recovery = 1;
1049 if (!mounted_read_only) {
1050 err = ubifs_recover_inl_heads(c, c->sbuf);
1054 } else if (!mounted_read_only) {
1056 * Set the "dirty" flag so that if we reboot uncleanly we
1057 * will notice this immediately on the next mount.
1059 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1060 err = ubifs_write_master(c);
1065 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1069 err = dbg_check_idx_size(c, c->old_idx_sz);
1073 err = ubifs_replay_journal(c);
1077 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1081 if (!mounted_read_only) {
1084 /* Check for enough free space */
1085 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1086 ubifs_err("insufficient available space");
1091 /* Check for enough log space */
1092 lnum = c->lhead_lnum + 1;
1093 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1094 lnum = UBIFS_LOG_LNUM;
1095 if (lnum == c->ltail_lnum) {
1096 err = ubifs_consolidate_log(c);
1101 if (c->need_recovery) {
1102 err = ubifs_recover_size(c);
1105 err = ubifs_rcvry_gc_commit(c);
1107 err = take_gc_lnum(c);
1111 err = dbg_check_lprops(c);
1114 } else if (c->need_recovery) {
1115 err = ubifs_recover_size(c);
1120 spin_lock(&ubifs_infos_lock);
1121 list_add_tail(&c->infos_list, &ubifs_infos);
1122 spin_unlock(&ubifs_infos_lock);
1124 if (c->need_recovery) {
1125 if (mounted_read_only)
1126 ubifs_msg("recovery deferred");
1128 c->need_recovery = 0;
1129 ubifs_msg("recovery completed");
1133 err = dbg_check_filesystem(c);
1137 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1138 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1139 if (mounted_read_only)
1140 ubifs_msg("mounted read-only");
1141 x = (long long)c->main_lebs * c->leb_size;
1142 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d LEBs)",
1143 x, x >> 10, x >> 20, c->main_lebs);
1144 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1145 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d LEBs)",
1146 x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1147 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1148 ubifs_msg("media format %d, latest format %d",
1149 c->fmt_version, UBIFS_FORMAT_VERSION);
1151 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1152 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1153 dbg_msg("LEB size: %d bytes (%d KiB)",
1154 c->leb_size, c->leb_size / 1024);
1155 dbg_msg("data journal heads: %d",
1156 c->jhead_cnt - NONDATA_JHEADS_CNT);
1157 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1158 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1159 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1160 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1161 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1162 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1163 dbg_msg("fast unmount: %d", c->fast_unmount);
1164 dbg_msg("big_lpt %d", c->big_lpt);
1165 dbg_msg("log LEBs: %d (%d - %d)",
1166 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1167 dbg_msg("LPT area LEBs: %d (%d - %d)",
1168 c->lpt_lebs, c->lpt_first, c->lpt_last);
1169 dbg_msg("orphan area LEBs: %d (%d - %d)",
1170 c->orph_lebs, c->orph_first, c->orph_last);
1171 dbg_msg("main area LEBs: %d (%d - %d)",
1172 c->main_lebs, c->main_first, c->leb_cnt - 1);
1173 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1174 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1175 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1176 dbg_msg("key hash type: %d", c->key_hash_type);
1177 dbg_msg("tree fanout: %d", c->fanout);
1178 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1179 dbg_msg("first main LEB: %d", c->main_first);
1180 dbg_msg("dead watermark: %d", c->dead_wm);
1181 dbg_msg("dark watermark: %d", c->dark_wm);
1182 x = (long long)c->main_lebs * c->dark_wm;
1183 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1184 x, x >> 10, x >> 20);
1185 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1186 c->max_bud_bytes, c->max_bud_bytes >> 10,
1187 c->max_bud_bytes >> 20);
1188 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1189 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1190 c->bg_bud_bytes >> 20);
1191 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1192 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1193 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1194 dbg_msg("commit number: %llu", c->cmt_no);
1199 spin_lock(&ubifs_infos_lock);
1200 list_del(&c->infos_list);
1201 spin_unlock(&ubifs_infos_lock);
1207 ubifs_lpt_free(c, 0);
1210 kfree(c->rcvrd_mst_node);
1212 kthread_stop(c->bgt);
1218 dbg_failure_mode_deregistration(c);
1222 kfree(c->bottom_up_buf);
1223 UBIFS_DBG(vfree(c->dbg_buf));
1228 * ubifs_umount - un-mount UBIFS file-system.
1229 * @c: UBIFS file-system description object
1231 * Note, this function is called to free allocated resourced when un-mounting,
1232 * as well as free resources when an error occurred while we were half way
1233 * through mounting (error path cleanup function). So it has to make sure the
1234 * resource was actually allocated before freeing it.
1236 static void ubifs_umount(struct ubifs_info *c)
1238 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1241 spin_lock(&ubifs_infos_lock);
1242 list_del(&c->infos_list);
1243 spin_unlock(&ubifs_infos_lock);
1246 kthread_stop(c->bgt);
1251 ubifs_lpt_free(c, 0);
1254 kfree(c->rcvrd_mst_node);
1257 kfree(c->bottom_up_buf);
1258 UBIFS_DBG(vfree(c->dbg_buf));
1260 dbg_failure_mode_deregistration(c);
1264 * ubifs_remount_rw - re-mount in read-write mode.
1265 * @c: UBIFS file-system description object
1267 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1268 * mode. This function allocates the needed resources and re-mounts UBIFS in
1271 static int ubifs_remount_rw(struct ubifs_info *c)
1278 mutex_lock(&c->umount_mutex);
1279 c->remounting_rw = 1;
1281 /* Check for enough free space */
1282 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1283 ubifs_err("insufficient available space");
1288 if (c->old_leb_cnt != c->leb_cnt) {
1289 struct ubifs_sb_node *sup;
1291 sup = ubifs_read_sb_node(c);
1296 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1297 err = ubifs_write_sb_node(c, sup);
1302 if (c->need_recovery) {
1303 ubifs_msg("completing deferred recovery");
1304 err = ubifs_write_rcvrd_mst_node(c);
1307 err = ubifs_recover_size(c);
1310 err = ubifs_clean_lebs(c, c->sbuf);
1313 err = ubifs_recover_inl_heads(c, c->sbuf);
1318 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1319 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1320 err = ubifs_write_master(c);
1325 c->ileb_buf = vmalloc(c->leb_size);
1331 err = ubifs_lpt_init(c, 0, 1);
1335 err = alloc_wbufs(c);
1339 ubifs_create_buds_lists(c);
1341 /* Create background thread */
1342 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1344 c->bgt = ERR_PTR(-EINVAL);
1345 if (IS_ERR(c->bgt)) {
1346 err = PTR_ERR(c->bgt);
1348 ubifs_err("cannot spawn \"%s\", error %d",
1352 wake_up_process(c->bgt);
1354 c->orph_buf = vmalloc(c->leb_size);
1358 /* Check for enough log space */
1359 lnum = c->lhead_lnum + 1;
1360 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1361 lnum = UBIFS_LOG_LNUM;
1362 if (lnum == c->ltail_lnum) {
1363 err = ubifs_consolidate_log(c);
1368 if (c->need_recovery)
1369 err = ubifs_rcvry_gc_commit(c);
1371 err = take_gc_lnum(c);
1375 if (c->need_recovery) {
1376 c->need_recovery = 0;
1377 ubifs_msg("deferred recovery completed");
1380 dbg_gen("re-mounted read-write");
1381 c->vfs_sb->s_flags &= ~MS_RDONLY;
1382 c->remounting_rw = 0;
1383 mutex_unlock(&c->umount_mutex);
1390 kthread_stop(c->bgt);
1396 ubifs_lpt_free(c, 1);
1397 c->remounting_rw = 0;
1398 mutex_unlock(&c->umount_mutex);
1403 * commit_on_unmount - commit the journal when un-mounting.
1404 * @c: UBIFS file-system description object
1406 * This function is called during un-mounting and it commits the journal unless
1407 * the "fast unmount" mode is enabled. It also avoids committing the journal if
1408 * it contains too few data.
1410 * Sometimes recovery requires the journal to be committed at least once, and
1411 * this function takes care about this.
1413 static void commit_on_unmount(struct ubifs_info *c)
1415 if (!c->fast_unmount) {
1416 long long bud_bytes;
1418 spin_lock(&c->buds_lock);
1419 bud_bytes = c->bud_bytes;
1420 spin_unlock(&c->buds_lock);
1421 if (bud_bytes > c->leb_size)
1422 ubifs_run_commit(c);
1427 * ubifs_remount_ro - re-mount in read-only mode.
1428 * @c: UBIFS file-system description object
1430 * We rely on VFS to have stopped writing. Possibly the background thread could
1431 * be running a commit, however kthread_stop will wait in that case.
1433 static void ubifs_remount_ro(struct ubifs_info *c)
1437 ubifs_assert(!c->need_recovery);
1438 commit_on_unmount(c);
1440 mutex_lock(&c->umount_mutex);
1442 kthread_stop(c->bgt);
1446 for (i = 0; i < c->jhead_cnt; i++) {
1447 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1448 del_timer_sync(&c->jheads[i].wbuf.timer);
1452 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1453 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1454 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1455 err = ubifs_write_master(c);
1457 ubifs_ro_mode(c, err);
1460 ubifs_destroy_idx_gc(c);
1466 ubifs_lpt_free(c, 1);
1467 mutex_unlock(&c->umount_mutex);
1470 static void ubifs_put_super(struct super_block *sb)
1473 struct ubifs_info *c = sb->s_fs_info;
1475 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1478 * The following asserts are only valid if there has not been a failure
1479 * of the media. For example, there will be dirty inodes if we failed
1480 * to write them back because of I/O errors.
1482 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1483 ubifs_assert(c->budg_idx_growth == 0);
1484 ubifs_assert(c->budg_dd_growth == 0);
1485 ubifs_assert(c->budg_data_growth == 0);
1488 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1489 * and file system un-mount. Namely, it prevents the shrinker from
1490 * picking this superblock for shrinking - it will be just skipped if
1491 * the mutex is locked.
1493 mutex_lock(&c->umount_mutex);
1494 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1496 * First of all kill the background thread to make sure it does
1497 * not interfere with un-mounting and freeing resources.
1500 kthread_stop(c->bgt);
1504 /* Synchronize write-buffers */
1506 for (i = 0; i < c->jhead_cnt; i++) {
1507 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1508 del_timer_sync(&c->jheads[i].wbuf.timer);
1512 * On fatal errors c->ro_media is set to 1, in which case we do
1513 * not write the master node.
1517 * We are being cleanly unmounted which means the
1518 * orphans were killed - indicate this in the master
1519 * node. Also save the reserved GC LEB number.
1523 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1524 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1525 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1526 err = ubifs_write_master(c);
1529 * Recovery will attempt to fix the master area
1530 * next mount, so we just print a message and
1531 * continue to unmount normally.
1533 ubifs_err("failed to write master node, "
1539 bdi_destroy(&c->bdi);
1540 ubi_close_volume(c->ubi);
1541 mutex_unlock(&c->umount_mutex);
1545 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1548 struct ubifs_info *c = sb->s_fs_info;
1550 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1552 err = ubifs_parse_options(c, data, 1);
1554 ubifs_err("invalid or unknown remount parameter");
1557 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1558 err = ubifs_remount_rw(c);
1561 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
1562 ubifs_remount_ro(c);
1567 struct super_operations ubifs_super_operations = {
1568 .alloc_inode = ubifs_alloc_inode,
1569 .destroy_inode = ubifs_destroy_inode,
1570 .put_super = ubifs_put_super,
1571 .write_inode = ubifs_write_inode,
1572 .delete_inode = ubifs_delete_inode,
1573 .statfs = ubifs_statfs,
1574 .dirty_inode = ubifs_dirty_inode,
1575 .remount_fs = ubifs_remount_fs,
1576 .show_options = ubifs_show_options,
1577 .sync_fs = ubifs_sync_fs,
1581 * open_ubi - parse UBI device name string and open the UBI device.
1582 * @name: UBI volume name
1583 * @mode: UBI volume open mode
1585 * There are several ways to specify UBI volumes when mounting UBIFS:
1586 * o ubiX_Y - UBI device number X, volume Y;
1587 * o ubiY - UBI device number 0, volume Y;
1588 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1589 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1591 * Alternative '!' separator may be used instead of ':' (because some shells
1592 * like busybox may interpret ':' as an NFS host name separator). This function
1593 * returns ubi volume object in case of success and a negative error code in
1596 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1601 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1602 return ERR_PTR(-EINVAL);
1604 /* ubi:NAME method */
1605 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1606 return ubi_open_volume_nm(0, name + 4, mode);
1608 if (!isdigit(name[3]))
1609 return ERR_PTR(-EINVAL);
1611 dev = simple_strtoul(name + 3, &endptr, 0);
1614 if (*endptr == '\0')
1615 return ubi_open_volume(0, dev, mode);
1618 if (*endptr == '_' && isdigit(endptr[1])) {
1619 vol = simple_strtoul(endptr + 1, &endptr, 0);
1620 if (*endptr != '\0')
1621 return ERR_PTR(-EINVAL);
1622 return ubi_open_volume(dev, vol, mode);
1625 /* ubiX:NAME method */
1626 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1627 return ubi_open_volume_nm(dev, ++endptr, mode);
1629 return ERR_PTR(-EINVAL);
1632 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1634 struct ubi_volume_desc *ubi = sb->s_fs_info;
1635 struct ubifs_info *c;
1639 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1643 spin_lock_init(&c->cnt_lock);
1644 spin_lock_init(&c->cs_lock);
1645 spin_lock_init(&c->buds_lock);
1646 spin_lock_init(&c->space_lock);
1647 spin_lock_init(&c->orphan_lock);
1648 init_rwsem(&c->commit_sem);
1649 mutex_init(&c->lp_mutex);
1650 mutex_init(&c->tnc_mutex);
1651 mutex_init(&c->log_mutex);
1652 mutex_init(&c->mst_mutex);
1653 mutex_init(&c->umount_mutex);
1654 init_waitqueue_head(&c->cmt_wq);
1656 c->old_idx = RB_ROOT;
1657 c->size_tree = RB_ROOT;
1658 c->orph_tree = RB_ROOT;
1659 INIT_LIST_HEAD(&c->infos_list);
1660 INIT_LIST_HEAD(&c->idx_gc);
1661 INIT_LIST_HEAD(&c->replay_list);
1662 INIT_LIST_HEAD(&c->replay_buds);
1663 INIT_LIST_HEAD(&c->uncat_list);
1664 INIT_LIST_HEAD(&c->empty_list);
1665 INIT_LIST_HEAD(&c->freeable_list);
1666 INIT_LIST_HEAD(&c->frdi_idx_list);
1667 INIT_LIST_HEAD(&c->unclean_leb_list);
1668 INIT_LIST_HEAD(&c->old_buds);
1669 INIT_LIST_HEAD(&c->orph_list);
1670 INIT_LIST_HEAD(&c->orph_new);
1672 c->highest_inum = UBIFS_FIRST_INO;
1673 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1675 ubi_get_volume_info(ubi, &c->vi);
1676 ubi_get_device_info(c->vi.ubi_num, &c->di);
1678 /* Re-open the UBI device in read-write mode */
1679 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1680 if (IS_ERR(c->ubi)) {
1681 err = PTR_ERR(c->ubi);
1686 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1687 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1688 * which means the user would have to wait not just for their own I/O
1689 * but the read-ahead I/O as well i.e. completely pointless.
1691 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1693 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1694 c->bdi.unplug_io_fn = default_unplug_io_fn;
1695 err = bdi_init(&c->bdi);
1699 err = ubifs_parse_options(c, data, 0);
1706 sb->s_magic = UBIFS_SUPER_MAGIC;
1707 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1708 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1709 sb->s_dev = c->vi.cdev;
1710 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1711 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1712 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1713 sb->s_op = &ubifs_super_operations;
1715 mutex_lock(&c->umount_mutex);
1716 err = mount_ubifs(c);
1718 ubifs_assert(err < 0);
1722 /* Read the root inode */
1723 root = ubifs_iget(sb, UBIFS_ROOT_INO);
1725 err = PTR_ERR(root);
1729 sb->s_root = d_alloc_root(root);
1733 mutex_unlock(&c->umount_mutex);
1742 mutex_unlock(&c->umount_mutex);
1744 bdi_destroy(&c->bdi);
1746 ubi_close_volume(c->ubi);
1752 static int sb_test(struct super_block *sb, void *data)
1756 return sb->s_dev == *dev;
1759 static int sb_set(struct super_block *sb, void *data)
1767 static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
1768 const char *name, void *data, struct vfsmount *mnt)
1770 struct ubi_volume_desc *ubi;
1771 struct ubi_volume_info vi;
1772 struct super_block *sb;
1775 dbg_gen("name %s, flags %#x", name, flags);
1778 * Get UBI device number and volume ID. Mount it read-only so far
1779 * because this might be a new mount point, and UBI allows only one
1780 * read-write user at a time.
1782 ubi = open_ubi(name, UBI_READONLY);
1784 ubifs_err("cannot open \"%s\", error %d",
1785 name, (int)PTR_ERR(ubi));
1786 return PTR_ERR(ubi);
1788 ubi_get_volume_info(ubi, &vi);
1790 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
1792 sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
1799 /* A new mount point for already mounted UBIFS */
1800 dbg_gen("this ubi volume is already mounted");
1801 if ((flags ^ sb->s_flags) & MS_RDONLY) {
1806 sb->s_flags = flags;
1808 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
1811 sb->s_fs_info = ubi;
1812 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
1815 /* We do not support atime */
1816 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
1819 /* 'fill_super()' opens ubi again so we must close it here */
1820 ubi_close_volume(ubi);
1822 return simple_set_mnt(mnt, sb);
1825 up_write(&sb->s_umount);
1826 deactivate_super(sb);
1828 ubi_close_volume(ubi);
1832 static void ubifs_kill_sb(struct super_block *sb)
1834 struct ubifs_info *c = sb->s_fs_info;
1837 * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
1838 * in order to be outside BKL.
1840 if (sb->s_root && !(sb->s_flags & MS_RDONLY))
1841 commit_on_unmount(c);
1842 /* The un-mount routine is actually done in put_super() */
1843 generic_shutdown_super(sb);
1846 static struct file_system_type ubifs_fs_type = {
1848 .owner = THIS_MODULE,
1849 .get_sb = ubifs_get_sb,
1850 .kill_sb = ubifs_kill_sb
1854 * Inode slab cache constructor.
1856 static void inode_slab_ctor(void *obj)
1858 struct ubifs_inode *ui = obj;
1859 inode_init_once(&ui->vfs_inode);
1862 static int __init ubifs_init(void)
1866 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
1868 /* Make sure node sizes are 8-byte aligned */
1869 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
1870 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
1871 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
1872 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
1873 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
1874 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
1875 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
1876 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
1877 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
1878 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
1879 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
1881 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
1882 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
1883 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
1884 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
1885 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
1886 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
1888 /* Check min. node size */
1889 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
1890 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
1891 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
1892 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
1894 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
1895 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
1896 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
1897 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
1899 /* Defined node sizes */
1900 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
1901 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
1902 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
1903 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
1906 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
1907 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
1909 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
1910 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
1911 " at least 4096 bytes",
1912 (unsigned int)PAGE_CACHE_SIZE);
1916 err = register_filesystem(&ubifs_fs_type);
1918 ubifs_err("cannot register file system, error %d", err);
1923 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
1924 sizeof(struct ubifs_inode), 0,
1925 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
1927 if (!ubifs_inode_slab)
1930 register_shrinker(&ubifs_shrinker_info);
1932 err = ubifs_compressors_init();
1939 unregister_shrinker(&ubifs_shrinker_info);
1940 kmem_cache_destroy(ubifs_inode_slab);
1942 unregister_filesystem(&ubifs_fs_type);
1945 /* late_initcall to let compressors initialize first */
1946 late_initcall(ubifs_init);
1948 static void __exit ubifs_exit(void)
1950 ubifs_assert(list_empty(&ubifs_infos));
1951 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
1953 ubifs_compressors_exit();
1954 unregister_shrinker(&ubifs_shrinker_info);
1955 kmem_cache_destroy(ubifs_inode_slab);
1956 unregister_filesystem(&ubifs_fs_type);
1958 module_exit(ubifs_exit);
1960 MODULE_LICENSE("GPL");
1961 MODULE_VERSION(__stringify(UBIFS_VERSION));
1962 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
1963 MODULE_DESCRIPTION("UBIFS - UBI File System");