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>
37 #include <linux/math64.h>
41 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
42 * allocating too much.
44 #define UBIFS_KMALLOC_OK (128*1024)
46 /* Slab cache for UBIFS inodes */
47 struct kmem_cache *ubifs_inode_slab;
49 /* UBIFS TNC shrinker description */
50 static struct shrinker ubifs_shrinker_info = {
51 .shrink = ubifs_shrinker,
52 .seeks = DEFAULT_SEEKS,
56 * validate_inode - validate inode.
57 * @c: UBIFS file-system description object
58 * @inode: the inode to validate
60 * This is a helper function for 'ubifs_iget()' which validates various fields
61 * of a newly built inode to make sure they contain sane values and prevent
62 * possible vulnerabilities. Returns zero if the inode is all right and
63 * a non-zero error code if not.
65 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
68 const struct ubifs_inode *ui = ubifs_inode(inode);
70 if (inode->i_size > c->max_inode_sz) {
71 ubifs_err("inode is too large (%lld)",
72 (long long)inode->i_size);
76 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
77 ubifs_err("unknown compression type %d", ui->compr_type);
81 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
84 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
87 if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
90 if (!ubifs_compr_present(ui->compr_type)) {
91 ubifs_warn("inode %lu uses '%s' compression, but it was not "
92 "compiled in", inode->i_ino,
93 ubifs_compr_name(ui->compr_type));
96 err = dbg_check_dir_size(c, inode);
100 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
104 struct ubifs_ino_node *ino;
105 struct ubifs_info *c = sb->s_fs_info;
107 struct ubifs_inode *ui;
109 dbg_gen("inode %lu", inum);
111 inode = iget_locked(sb, inum);
113 return ERR_PTR(-ENOMEM);
114 if (!(inode->i_state & I_NEW))
116 ui = ubifs_inode(inode);
118 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
124 ino_key_init(c, &key, inode->i_ino);
126 err = ubifs_tnc_lookup(c, &key, ino);
130 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
131 inode->i_nlink = le32_to_cpu(ino->nlink);
132 inode->i_uid = le32_to_cpu(ino->uid);
133 inode->i_gid = le32_to_cpu(ino->gid);
134 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
135 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
136 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
137 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
138 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
139 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
140 inode->i_mode = le32_to_cpu(ino->mode);
141 inode->i_size = le64_to_cpu(ino->size);
143 ui->data_len = le32_to_cpu(ino->data_len);
144 ui->flags = le32_to_cpu(ino->flags);
145 ui->compr_type = le16_to_cpu(ino->compr_type);
146 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
147 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
148 ui->xattr_size = le32_to_cpu(ino->xattr_size);
149 ui->xattr_names = le32_to_cpu(ino->xattr_names);
150 ui->synced_i_size = ui->ui_size = inode->i_size;
152 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
154 err = validate_inode(c, inode);
158 /* Disable read-ahead */
159 inode->i_mapping->backing_dev_info = &c->bdi;
161 switch (inode->i_mode & S_IFMT) {
163 inode->i_mapping->a_ops = &ubifs_file_address_operations;
164 inode->i_op = &ubifs_file_inode_operations;
165 inode->i_fop = &ubifs_file_operations;
167 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
172 memcpy(ui->data, ino->data, ui->data_len);
173 ((char *)ui->data)[ui->data_len] = '\0';
174 } else if (ui->data_len != 0) {
180 inode->i_op = &ubifs_dir_inode_operations;
181 inode->i_fop = &ubifs_dir_operations;
182 if (ui->data_len != 0) {
188 inode->i_op = &ubifs_symlink_inode_operations;
189 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
193 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
198 memcpy(ui->data, ino->data, ui->data_len);
199 ((char *)ui->data)[ui->data_len] = '\0';
205 union ubifs_dev_desc *dev;
207 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
213 dev = (union ubifs_dev_desc *)ino->data;
214 if (ui->data_len == sizeof(dev->new))
215 rdev = new_decode_dev(le32_to_cpu(dev->new));
216 else if (ui->data_len == sizeof(dev->huge))
217 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
222 memcpy(ui->data, ino->data, ui->data_len);
223 inode->i_op = &ubifs_file_inode_operations;
224 init_special_inode(inode, inode->i_mode, rdev);
229 inode->i_op = &ubifs_file_inode_operations;
230 init_special_inode(inode, inode->i_mode, 0);
231 if (ui->data_len != 0) {
242 ubifs_set_inode_flags(inode);
243 unlock_new_inode(inode);
247 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
248 dbg_dump_node(c, ino);
249 dbg_dump_inode(c, inode);
254 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
259 static struct inode *ubifs_alloc_inode(struct super_block *sb)
261 struct ubifs_inode *ui;
263 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
267 memset((void *)ui + sizeof(struct inode), 0,
268 sizeof(struct ubifs_inode) - sizeof(struct inode));
269 mutex_init(&ui->ui_mutex);
270 spin_lock_init(&ui->ui_lock);
271 return &ui->vfs_inode;
274 static void ubifs_destroy_inode(struct inode *inode)
276 struct ubifs_inode *ui = ubifs_inode(inode);
279 kmem_cache_free(ubifs_inode_slab, inode);
283 * Note, Linux write-back code calls this without 'i_mutex'.
285 static int ubifs_write_inode(struct inode *inode, int wait)
288 struct ubifs_info *c = inode->i_sb->s_fs_info;
289 struct ubifs_inode *ui = ubifs_inode(inode);
291 ubifs_assert(!ui->xattr);
292 if (is_bad_inode(inode))
295 mutex_lock(&ui->ui_mutex);
297 * Due to races between write-back forced by budgeting
298 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
299 * have already been synchronized, do not do this again. This might
300 * also happen if it was synchronized in an VFS operation, e.g.
304 mutex_unlock(&ui->ui_mutex);
309 * As an optimization, do not write orphan inodes to the media just
310 * because this is not needed.
312 dbg_gen("inode %lu, mode %#x, nlink %u",
313 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
314 if (inode->i_nlink) {
315 err = ubifs_jnl_write_inode(c, inode);
317 ubifs_err("can't write inode %lu, error %d",
322 mutex_unlock(&ui->ui_mutex);
323 ubifs_release_dirty_inode_budget(c, ui);
327 static void ubifs_delete_inode(struct inode *inode)
330 struct ubifs_info *c = inode->i_sb->s_fs_info;
331 struct ubifs_inode *ui = ubifs_inode(inode);
335 * Extended attribute inode deletions are fully handled in
336 * 'ubifs_removexattr()'. These inodes are special and have
337 * limited usage, so there is nothing to do here.
341 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
342 ubifs_assert(!atomic_read(&inode->i_count));
343 ubifs_assert(inode->i_nlink == 0);
345 truncate_inode_pages(&inode->i_data, 0);
346 if (is_bad_inode(inode))
349 ui->ui_size = inode->i_size = 0;
350 err = ubifs_jnl_delete_inode(c, inode);
353 * Worst case we have a lost orphan inode wasting space, so a
354 * simple error message is OK here.
356 ubifs_err("can't delete inode %lu, error %d",
361 ubifs_release_dirty_inode_budget(c, ui);
365 static void ubifs_dirty_inode(struct inode *inode)
367 struct ubifs_inode *ui = ubifs_inode(inode);
369 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
372 dbg_gen("inode %lu", inode->i_ino);
376 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
378 struct ubifs_info *c = dentry->d_sb->s_fs_info;
379 unsigned long long free;
380 __le32 *uuid = (__le32 *)c->uuid;
382 free = ubifs_get_free_space(c);
383 dbg_gen("free space %lld bytes (%lld blocks)",
384 free, free >> UBIFS_BLOCK_SHIFT);
386 buf->f_type = UBIFS_SUPER_MAGIC;
387 buf->f_bsize = UBIFS_BLOCK_SIZE;
388 buf->f_blocks = c->block_cnt;
389 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
390 if (free > c->report_rp_size)
391 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
396 buf->f_namelen = UBIFS_MAX_NLEN;
397 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
398 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
402 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
404 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
406 if (c->mount_opts.unmount_mode == 2)
407 seq_printf(s, ",fast_unmount");
408 else if (c->mount_opts.unmount_mode == 1)
409 seq_printf(s, ",norm_unmount");
411 if (c->mount_opts.bulk_read == 2)
412 seq_printf(s, ",bulk_read");
413 else if (c->mount_opts.bulk_read == 1)
414 seq_printf(s, ",no_bulk_read");
416 if (c->mount_opts.chk_data_crc == 2)
417 seq_printf(s, ",chk_data_crc");
418 else if (c->mount_opts.chk_data_crc == 1)
419 seq_printf(s, ",no_chk_data_crc");
421 if (c->mount_opts.override_compr) {
422 seq_printf(s, ",compr=");
423 seq_printf(s, ubifs_compr_name(c->mount_opts.compr_type));
429 static int ubifs_sync_fs(struct super_block *sb, int wait)
431 struct ubifs_info *c = sb->s_fs_info;
436 for (i = 0; i < c->jhead_cnt; i++) {
437 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
442 /* Commit the journal unless it has too little data */
443 spin_lock(&c->buds_lock);
444 bud_bytes = c->bud_bytes;
445 spin_unlock(&c->buds_lock);
446 if (bud_bytes > c->leb_size) {
447 err = ubifs_run_commit(c);
454 * We ought to call sync for c->ubi but it does not have one. If it had
455 * it would in turn call mtd->sync, however mtd operations are
456 * synchronous anyway, so we don't lose any sleep here.
462 * init_constants_early - initialize UBIFS constants.
463 * @c: UBIFS file-system description object
465 * This function initialize UBIFS constants which do not need the superblock to
466 * be read. It also checks that the UBI volume satisfies basic UBIFS
467 * requirements. Returns zero in case of success and a negative error code in
470 static int init_constants_early(struct ubifs_info *c)
472 if (c->vi.corrupted) {
473 ubifs_warn("UBI volume is corrupted - read-only mode");
478 ubifs_msg("read-only UBI device");
482 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
483 ubifs_msg("static UBI volume - read-only mode");
487 c->leb_cnt = c->vi.size;
488 c->leb_size = c->vi.usable_leb_size;
489 c->half_leb_size = c->leb_size / 2;
490 c->min_io_size = c->di.min_io_size;
491 c->min_io_shift = fls(c->min_io_size) - 1;
493 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
494 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
495 c->leb_size, UBIFS_MIN_LEB_SZ);
499 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
500 ubifs_err("too few LEBs (%d), min. is %d",
501 c->leb_cnt, UBIFS_MIN_LEB_CNT);
505 if (!is_power_of_2(c->min_io_size)) {
506 ubifs_err("bad min. I/O size %d", c->min_io_size);
511 * UBIFS aligns all node to 8-byte boundary, so to make function in
512 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
515 if (c->min_io_size < 8) {
520 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
521 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
524 * Initialize node length ranges which are mostly needed for node
527 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
528 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
529 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
530 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
531 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
532 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
534 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
535 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
536 c->ranges[UBIFS_ORPH_NODE].min_len =
537 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
538 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
539 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
540 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
541 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
542 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
543 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
544 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
546 * Minimum indexing node size is amended later when superblock is
547 * read and the key length is known.
549 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
551 * Maximum indexing node size is amended later when superblock is
552 * read and the fanout is known.
554 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
557 * Initialize dead and dark LEB space watermarks.
559 * Dead space is the space which cannot be used. Its watermark is
560 * equivalent to min. I/O unit or minimum node size if it is greater
561 * then min. I/O unit.
563 * Dark space is the space which might be used, or might not, depending
564 * on which node should be written to the LEB. Its watermark is
565 * equivalent to maximum UBIFS node size.
567 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
568 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
571 * Calculate how many bytes would be wasted at the end of LEB if it was
572 * fully filled with data nodes of maximum size. This is used in
573 * calculations when reporting free space.
575 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
577 /* Buffer size for bulk-reads */
578 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
579 if (c->max_bu_buf_len > c->leb_size)
580 c->max_bu_buf_len = c->leb_size;
585 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
586 * @c: UBIFS file-system description object
587 * @lnum: LEB the write-buffer was synchronized to
588 * @free: how many free bytes left in this LEB
589 * @pad: how many bytes were padded
591 * This is a callback function which is called by the I/O unit when the
592 * write-buffer is synchronized. We need this to correctly maintain space
593 * accounting in bud logical eraseblocks. This function returns zero in case of
594 * success and a negative error code in case of failure.
596 * This function actually belongs to the journal, but we keep it here because
597 * we want to keep it static.
599 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
601 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
605 * init_constants_late - initialize UBIFS constants.
606 * @c: UBIFS file-system description object
608 * This is a helper function which initializes various UBIFS constants after
609 * the superblock has been read. It also checks various UBIFS parameters and
610 * makes sure they are all right. Returns zero in case of success and a
611 * negative error code in case of failure.
613 static int init_constants_late(struct ubifs_info *c)
618 c->main_bytes = (long long)c->main_lebs * c->leb_size;
619 c->max_znode_sz = sizeof(struct ubifs_znode) +
620 c->fanout * sizeof(struct ubifs_zbranch);
622 tmp = ubifs_idx_node_sz(c, 1);
623 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
624 c->min_idx_node_sz = ALIGN(tmp, 8);
626 tmp = ubifs_idx_node_sz(c, c->fanout);
627 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
628 c->max_idx_node_sz = ALIGN(tmp, 8);
630 /* Make sure LEB size is large enough to fit full commit */
631 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
632 tmp = ALIGN(tmp, c->min_io_size);
633 if (tmp > c->leb_size) {
634 dbg_err("too small LEB size %d, at least %d needed",
640 * Make sure that the log is large enough to fit reference nodes for
641 * all buds plus one reserved LEB.
643 tmp64 = c->max_bud_bytes + c->leb_size - 1;
644 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
645 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
648 if (c->log_lebs < tmp) {
649 dbg_err("too small log %d LEBs, required min. %d LEBs",
655 * When budgeting we assume worst-case scenarios when the pages are not
656 * be compressed and direntries are of the maximum size.
658 * Note, data, which may be stored in inodes is budgeted separately, so
659 * it is not included into 'c->inode_budget'.
661 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
662 c->inode_budget = UBIFS_INO_NODE_SZ;
663 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
666 * When the amount of flash space used by buds becomes
667 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
668 * The writers are unblocked when the commit is finished. To avoid
669 * writers to be blocked UBIFS initiates background commit in advance,
670 * when number of bud bytes becomes above the limit defined below.
672 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
675 * Ensure minimum journal size. All the bytes in the journal heads are
676 * considered to be used, when calculating the current journal usage.
677 * Consequently, if the journal is too small, UBIFS will treat it as
680 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
681 if (c->bg_bud_bytes < tmp64)
682 c->bg_bud_bytes = tmp64;
683 if (c->max_bud_bytes < tmp64 + c->leb_size)
684 c->max_bud_bytes = tmp64 + c->leb_size;
686 err = ubifs_calc_lpt_geom(c);
690 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
693 * Calculate total amount of FS blocks. This number is not used
694 * internally because it does not make much sense for UBIFS, but it is
695 * necessary to report something for the 'statfs()' call.
697 * Subtract the LEB reserved for GC, the LEB which is reserved for
698 * deletions, minimum LEBs for the index, and assume only one journal
701 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
702 tmp64 *= (long long)c->leb_size - c->leb_overhead;
703 tmp64 = ubifs_reported_space(c, tmp64);
704 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
710 * take_gc_lnum - reserve GC LEB.
711 * @c: UBIFS file-system description object
713 * This function ensures that the LEB reserved for garbage collection is
714 * unmapped and is marked as "taken" in lprops. We also have to set free space
715 * to LEB size and dirty space to zero, because lprops may contain out-of-date
716 * information if the file-system was un-mounted before it has been committed.
717 * This function returns zero in case of success and a negative error code in
720 static int take_gc_lnum(struct ubifs_info *c)
724 if (c->gc_lnum == -1) {
725 ubifs_err("no LEB for GC");
729 err = ubifs_leb_unmap(c, c->gc_lnum);
733 /* And we have to tell lprops that this LEB is taken */
734 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
740 * alloc_wbufs - allocate write-buffers.
741 * @c: UBIFS file-system description object
743 * This helper function allocates and initializes UBIFS write-buffers. Returns
744 * zero in case of success and %-ENOMEM in case of failure.
746 static int alloc_wbufs(struct ubifs_info *c)
750 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
755 /* Initialize journal heads */
756 for (i = 0; i < c->jhead_cnt; i++) {
757 INIT_LIST_HEAD(&c->jheads[i].buds_list);
758 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
762 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
763 c->jheads[i].wbuf.jhead = i;
766 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
768 * Garbage Collector head likely contains long-term data and
769 * does not need to be synchronized by timer.
771 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
772 c->jheads[GCHD].wbuf.timeout = 0;
778 * free_wbufs - free write-buffers.
779 * @c: UBIFS file-system description object
781 static void free_wbufs(struct ubifs_info *c)
786 for (i = 0; i < c->jhead_cnt; i++) {
787 kfree(c->jheads[i].wbuf.buf);
788 kfree(c->jheads[i].wbuf.inodes);
796 * free_orphans - free orphans.
797 * @c: UBIFS file-system description object
799 static void free_orphans(struct ubifs_info *c)
801 struct ubifs_orphan *orph;
803 while (c->orph_dnext) {
804 orph = c->orph_dnext;
805 c->orph_dnext = orph->dnext;
806 list_del(&orph->list);
810 while (!list_empty(&c->orph_list)) {
811 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
812 list_del(&orph->list);
814 dbg_err("orphan list not empty at unmount");
822 * free_buds - free per-bud objects.
823 * @c: UBIFS file-system description object
825 static void free_buds(struct ubifs_info *c)
827 struct rb_node *this = c->buds.rb_node;
828 struct ubifs_bud *bud;
832 this = this->rb_left;
833 else if (this->rb_right)
834 this = this->rb_right;
836 bud = rb_entry(this, struct ubifs_bud, rb);
837 this = rb_parent(this);
839 if (this->rb_left == &bud->rb)
840 this->rb_left = NULL;
842 this->rb_right = NULL;
850 * check_volume_empty - check if the UBI volume is empty.
851 * @c: UBIFS file-system description object
853 * This function checks if the UBIFS volume is empty by looking if its LEBs are
854 * mapped or not. The result of checking is stored in the @c->empty variable.
855 * Returns zero in case of success and a negative error code in case of
858 static int check_volume_empty(struct ubifs_info *c)
863 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
864 err = ubi_is_mapped(c->ubi, lnum);
865 if (unlikely(err < 0))
879 * UBIFS mount options.
881 * Opt_fast_unmount: do not run a journal commit before un-mounting
882 * Opt_norm_unmount: run a journal commit before un-mounting
883 * Opt_bulk_read: enable bulk-reads
884 * Opt_no_bulk_read: disable bulk-reads
885 * Opt_chk_data_crc: check CRCs when reading data nodes
886 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
887 * Opt_override_compr: override default compressor
888 * Opt_err: just end of array marker
901 static const match_table_t tokens = {
902 {Opt_fast_unmount, "fast_unmount"},
903 {Opt_norm_unmount, "norm_unmount"},
904 {Opt_bulk_read, "bulk_read"},
905 {Opt_no_bulk_read, "no_bulk_read"},
906 {Opt_chk_data_crc, "chk_data_crc"},
907 {Opt_no_chk_data_crc, "no_chk_data_crc"},
908 {Opt_override_compr, "compr=%s"},
913 * ubifs_parse_options - parse mount parameters.
914 * @c: UBIFS file-system description object
915 * @options: parameters to parse
916 * @is_remount: non-zero if this is FS re-mount
918 * This function parses UBIFS mount options and returns zero in case success
919 * and a negative error code in case of failure.
921 static int ubifs_parse_options(struct ubifs_info *c, char *options,
925 substring_t args[MAX_OPT_ARGS];
930 while ((p = strsep(&options, ","))) {
936 token = match_token(p, tokens, args);
938 case Opt_fast_unmount:
939 c->mount_opts.unmount_mode = 2;
942 case Opt_norm_unmount:
943 c->mount_opts.unmount_mode = 1;
947 c->mount_opts.bulk_read = 2;
950 case Opt_no_bulk_read:
951 c->mount_opts.bulk_read = 1;
954 case Opt_chk_data_crc:
955 c->mount_opts.chk_data_crc = 2;
956 c->no_chk_data_crc = 0;
958 case Opt_no_chk_data_crc:
959 c->mount_opts.chk_data_crc = 1;
960 c->no_chk_data_crc = 1;
962 case Opt_override_compr:
964 char *name = match_strdup(&args[0]);
968 if (!strcmp(name, "none"))
969 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
970 else if (!strcmp(name, "lzo"))
971 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
972 else if (!strcmp(name, "zlib"))
973 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
975 ubifs_err("unknown compressor \"%s\"", name);
980 c->mount_opts.override_compr = 1;
981 c->default_compr = c->mount_opts.compr_type;
985 ubifs_err("unrecognized mount option \"%s\" "
986 "or missing value", p);
995 * destroy_journal - destroy journal data structures.
996 * @c: UBIFS file-system description object
998 * This function destroys journal data structures including those that may have
999 * been created by recovery functions.
1001 static void destroy_journal(struct ubifs_info *c)
1003 while (!list_empty(&c->unclean_leb_list)) {
1004 struct ubifs_unclean_leb *ucleb;
1006 ucleb = list_entry(c->unclean_leb_list.next,
1007 struct ubifs_unclean_leb, list);
1008 list_del(&ucleb->list);
1011 while (!list_empty(&c->old_buds)) {
1012 struct ubifs_bud *bud;
1014 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1015 list_del(&bud->list);
1018 ubifs_destroy_idx_gc(c);
1019 ubifs_destroy_size_tree(c);
1025 * bu_init - initialize bulk-read information.
1026 * @c: UBIFS file-system description object
1028 static void bu_init(struct ubifs_info *c)
1030 ubifs_assert(c->bulk_read == 1);
1033 return; /* Already initialized */
1036 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1038 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1039 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1043 /* Just disable bulk-read */
1044 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1045 "disabling it", c->max_bu_buf_len);
1046 c->mount_opts.bulk_read = 1;
1053 * mount_ubifs - mount UBIFS file-system.
1054 * @c: UBIFS file-system description object
1056 * This function mounts UBIFS file system. Returns zero in case of success and
1057 * a negative error code in case of failure.
1059 * Note, the function does not de-allocate resources it it fails half way
1060 * through, and the caller has to do this instead.
1062 static int mount_ubifs(struct ubifs_info *c)
1064 struct super_block *sb = c->vfs_sb;
1065 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1069 err = init_constants_early(c);
1073 err = ubifs_debugging_init(c);
1077 err = check_volume_empty(c);
1081 if (c->empty && (mounted_read_only || c->ro_media)) {
1083 * This UBI volume is empty, and read-only, or the file system
1084 * is mounted read-only - we cannot format it.
1086 ubifs_err("can't format empty UBI volume: read-only %s",
1087 c->ro_media ? "UBI volume" : "mount");
1092 if (c->ro_media && !mounted_read_only) {
1093 ubifs_err("cannot mount read-write - read-only media");
1099 * The requirement for the buffer is that it should fit indexing B-tree
1100 * height amount of integers. We assume the height if the TNC tree will
1104 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1105 if (!c->bottom_up_buf)
1108 c->sbuf = vmalloc(c->leb_size);
1112 if (!mounted_read_only) {
1113 c->ileb_buf = vmalloc(c->leb_size);
1118 if (c->bulk_read == 1)
1122 * We have to check all CRCs, even for data nodes, when we mount the FS
1123 * (specifically, when we are replaying).
1125 c->always_chk_crc = 1;
1127 err = ubifs_read_superblock(c);
1132 * Make sure the compressor which is set as default in the superblock
1133 * or overriden by mount options is actually compiled in.
1135 if (!ubifs_compr_present(c->default_compr)) {
1136 ubifs_err("'compressor \"%s\" is not compiled in",
1137 ubifs_compr_name(c->default_compr));
1141 err = init_constants_late(c);
1145 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1146 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1147 c->cbuf = kmalloc(sz, GFP_NOFS);
1153 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1154 if (!mounted_read_only) {
1155 err = alloc_wbufs(c);
1159 /* Create background thread */
1160 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1161 if (IS_ERR(c->bgt)) {
1162 err = PTR_ERR(c->bgt);
1164 ubifs_err("cannot spawn \"%s\", error %d",
1168 wake_up_process(c->bgt);
1171 err = ubifs_read_master(c);
1175 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1176 ubifs_msg("recovery needed");
1177 c->need_recovery = 1;
1178 if (!mounted_read_only) {
1179 err = ubifs_recover_inl_heads(c, c->sbuf);
1183 } else if (!mounted_read_only) {
1185 * Set the "dirty" flag so that if we reboot uncleanly we
1186 * will notice this immediately on the next mount.
1188 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1189 err = ubifs_write_master(c);
1194 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1198 err = dbg_check_idx_size(c, c->old_idx_sz);
1202 err = ubifs_replay_journal(c);
1206 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1210 if (!mounted_read_only) {
1213 /* Check for enough free space */
1214 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1215 ubifs_err("insufficient available space");
1220 /* Check for enough log space */
1221 lnum = c->lhead_lnum + 1;
1222 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1223 lnum = UBIFS_LOG_LNUM;
1224 if (lnum == c->ltail_lnum) {
1225 err = ubifs_consolidate_log(c);
1230 if (c->need_recovery) {
1231 err = ubifs_recover_size(c);
1234 err = ubifs_rcvry_gc_commit(c);
1236 err = take_gc_lnum(c);
1240 err = dbg_check_lprops(c);
1243 } else if (c->need_recovery) {
1244 err = ubifs_recover_size(c);
1249 spin_lock(&ubifs_infos_lock);
1250 list_add_tail(&c->infos_list, &ubifs_infos);
1251 spin_unlock(&ubifs_infos_lock);
1253 if (c->need_recovery) {
1254 if (mounted_read_only)
1255 ubifs_msg("recovery deferred");
1257 c->need_recovery = 0;
1258 ubifs_msg("recovery completed");
1262 err = dbg_debugfs_init_fs(c);
1266 err = dbg_check_filesystem(c);
1270 c->always_chk_crc = 0;
1272 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1273 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1274 if (mounted_read_only)
1275 ubifs_msg("mounted read-only");
1276 x = (long long)c->main_lebs * c->leb_size;
1277 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1278 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1279 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1280 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1281 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1282 ubifs_msg("media format: %d (latest is %d)",
1283 c->fmt_version, UBIFS_FORMAT_VERSION);
1284 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1285 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1286 c->report_rp_size, c->report_rp_size >> 10);
1288 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1289 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1290 dbg_msg("LEB size: %d bytes (%d KiB)",
1291 c->leb_size, c->leb_size >> 10);
1292 dbg_msg("data journal heads: %d",
1293 c->jhead_cnt - NONDATA_JHEADS_CNT);
1294 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1295 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1296 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1297 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1298 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1299 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1300 dbg_msg("fast unmount: %d", c->fast_unmount);
1301 dbg_msg("big_lpt %d", c->big_lpt);
1302 dbg_msg("log LEBs: %d (%d - %d)",
1303 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1304 dbg_msg("LPT area LEBs: %d (%d - %d)",
1305 c->lpt_lebs, c->lpt_first, c->lpt_last);
1306 dbg_msg("orphan area LEBs: %d (%d - %d)",
1307 c->orph_lebs, c->orph_first, c->orph_last);
1308 dbg_msg("main area LEBs: %d (%d - %d)",
1309 c->main_lebs, c->main_first, c->leb_cnt - 1);
1310 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1311 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1312 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1313 dbg_msg("key hash type: %d", c->key_hash_type);
1314 dbg_msg("tree fanout: %d", c->fanout);
1315 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1316 dbg_msg("first main LEB: %d", c->main_first);
1317 dbg_msg("dead watermark: %d", c->dead_wm);
1318 dbg_msg("dark watermark: %d", c->dark_wm);
1319 x = (long long)c->main_lebs * c->dark_wm;
1320 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1321 x, x >> 10, x >> 20);
1322 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1323 c->max_bud_bytes, c->max_bud_bytes >> 10,
1324 c->max_bud_bytes >> 20);
1325 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1326 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1327 c->bg_bud_bytes >> 20);
1328 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1329 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1330 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1331 dbg_msg("commit number: %llu", c->cmt_no);
1336 spin_lock(&ubifs_infos_lock);
1337 list_del(&c->infos_list);
1338 spin_unlock(&ubifs_infos_lock);
1344 ubifs_lpt_free(c, 0);
1347 kfree(c->rcvrd_mst_node);
1349 kthread_stop(c->bgt);
1358 kfree(c->bottom_up_buf);
1359 ubifs_debugging_exit(c);
1364 * ubifs_umount - un-mount UBIFS file-system.
1365 * @c: UBIFS file-system description object
1367 * Note, this function is called to free allocated resourced when un-mounting,
1368 * as well as free resources when an error occurred while we were half way
1369 * through mounting (error path cleanup function). So it has to make sure the
1370 * resource was actually allocated before freeing it.
1372 static void ubifs_umount(struct ubifs_info *c)
1374 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1377 dbg_debugfs_exit_fs(c);
1378 spin_lock(&ubifs_infos_lock);
1379 list_del(&c->infos_list);
1380 spin_unlock(&ubifs_infos_lock);
1383 kthread_stop(c->bgt);
1388 ubifs_lpt_free(c, 0);
1391 kfree(c->rcvrd_mst_node);
1396 kfree(c->bottom_up_buf);
1397 ubifs_debugging_exit(c);
1401 * ubifs_remount_rw - re-mount in read-write mode.
1402 * @c: UBIFS file-system description object
1404 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1405 * mode. This function allocates the needed resources and re-mounts UBIFS in
1408 static int ubifs_remount_rw(struct ubifs_info *c)
1415 mutex_lock(&c->umount_mutex);
1416 c->remounting_rw = 1;
1417 c->always_chk_crc = 1;
1419 /* Check for enough free space */
1420 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1421 ubifs_err("insufficient available space");
1426 if (c->old_leb_cnt != c->leb_cnt) {
1427 struct ubifs_sb_node *sup;
1429 sup = ubifs_read_sb_node(c);
1434 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1435 err = ubifs_write_sb_node(c, sup);
1440 if (c->need_recovery) {
1441 ubifs_msg("completing deferred recovery");
1442 err = ubifs_write_rcvrd_mst_node(c);
1445 err = ubifs_recover_size(c);
1448 err = ubifs_clean_lebs(c, c->sbuf);
1451 err = ubifs_recover_inl_heads(c, c->sbuf);
1456 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1457 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1458 err = ubifs_write_master(c);
1463 c->ileb_buf = vmalloc(c->leb_size);
1469 err = ubifs_lpt_init(c, 0, 1);
1473 err = alloc_wbufs(c);
1477 ubifs_create_buds_lists(c);
1479 /* Create background thread */
1480 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1481 if (IS_ERR(c->bgt)) {
1482 err = PTR_ERR(c->bgt);
1484 ubifs_err("cannot spawn \"%s\", error %d",
1488 wake_up_process(c->bgt);
1490 c->orph_buf = vmalloc(c->leb_size);
1496 /* Check for enough log space */
1497 lnum = c->lhead_lnum + 1;
1498 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1499 lnum = UBIFS_LOG_LNUM;
1500 if (lnum == c->ltail_lnum) {
1501 err = ubifs_consolidate_log(c);
1506 if (c->need_recovery)
1507 err = ubifs_rcvry_gc_commit(c);
1509 err = take_gc_lnum(c);
1513 if (c->need_recovery) {
1514 c->need_recovery = 0;
1515 ubifs_msg("deferred recovery completed");
1518 dbg_gen("re-mounted read-write");
1519 c->vfs_sb->s_flags &= ~MS_RDONLY;
1520 c->remounting_rw = 0;
1521 c->always_chk_crc = 0;
1522 mutex_unlock(&c->umount_mutex);
1529 kthread_stop(c->bgt);
1535 ubifs_lpt_free(c, 1);
1536 c->remounting_rw = 0;
1537 c->always_chk_crc = 0;
1538 mutex_unlock(&c->umount_mutex);
1543 * commit_on_unmount - commit the journal when un-mounting.
1544 * @c: UBIFS file-system description object
1546 * This function is called during un-mounting and re-mounting, and it commits
1547 * the journal unless the "fast unmount" mode is enabled. It also avoids
1548 * committing the journal if it contains too few data.
1550 static void commit_on_unmount(struct ubifs_info *c)
1552 if (!c->fast_unmount) {
1553 long long bud_bytes;
1555 spin_lock(&c->buds_lock);
1556 bud_bytes = c->bud_bytes;
1557 spin_unlock(&c->buds_lock);
1558 if (bud_bytes > c->leb_size)
1559 ubifs_run_commit(c);
1564 * ubifs_remount_ro - re-mount in read-only mode.
1565 * @c: UBIFS file-system description object
1567 * We rely on VFS to have stopped writing. Possibly the background thread could
1568 * be running a commit, however kthread_stop will wait in that case.
1570 static void ubifs_remount_ro(struct ubifs_info *c)
1574 ubifs_assert(!c->need_recovery);
1575 commit_on_unmount(c);
1577 mutex_lock(&c->umount_mutex);
1579 kthread_stop(c->bgt);
1583 for (i = 0; i < c->jhead_cnt; i++) {
1584 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1585 del_timer_sync(&c->jheads[i].wbuf.timer);
1589 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1590 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1591 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1592 err = ubifs_write_master(c);
1594 ubifs_ro_mode(c, err);
1597 ubifs_destroy_idx_gc(c);
1603 ubifs_lpt_free(c, 1);
1604 mutex_unlock(&c->umount_mutex);
1607 static void ubifs_put_super(struct super_block *sb)
1610 struct ubifs_info *c = sb->s_fs_info;
1612 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1615 * The following asserts are only valid if there has not been a failure
1616 * of the media. For example, there will be dirty inodes if we failed
1617 * to write them back because of I/O errors.
1619 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1620 ubifs_assert(c->budg_idx_growth == 0);
1621 ubifs_assert(c->budg_dd_growth == 0);
1622 ubifs_assert(c->budg_data_growth == 0);
1625 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1626 * and file system un-mount. Namely, it prevents the shrinker from
1627 * picking this superblock for shrinking - it will be just skipped if
1628 * the mutex is locked.
1630 mutex_lock(&c->umount_mutex);
1631 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1633 * First of all kill the background thread to make sure it does
1634 * not interfere with un-mounting and freeing resources.
1637 kthread_stop(c->bgt);
1641 /* Synchronize write-buffers */
1643 for (i = 0; i < c->jhead_cnt; i++) {
1644 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1645 del_timer_sync(&c->jheads[i].wbuf.timer);
1649 * On fatal errors c->ro_media is set to 1, in which case we do
1650 * not write the master node.
1654 * We are being cleanly unmounted which means the
1655 * orphans were killed - indicate this in the master
1656 * node. Also save the reserved GC LEB number.
1660 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1661 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1662 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1663 err = ubifs_write_master(c);
1666 * Recovery will attempt to fix the master area
1667 * next mount, so we just print a message and
1668 * continue to unmount normally.
1670 ubifs_err("failed to write master node, "
1676 bdi_destroy(&c->bdi);
1677 ubi_close_volume(c->ubi);
1678 mutex_unlock(&c->umount_mutex);
1682 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1685 struct ubifs_info *c = sb->s_fs_info;
1687 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1689 err = ubifs_parse_options(c, data, 1);
1691 ubifs_err("invalid or unknown remount parameter");
1695 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1696 err = ubifs_remount_rw(c);
1699 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
1700 ubifs_remount_ro(c);
1702 if (c->bulk_read == 1)
1705 dbg_gen("disable bulk-read");
1713 struct super_operations ubifs_super_operations = {
1714 .alloc_inode = ubifs_alloc_inode,
1715 .destroy_inode = ubifs_destroy_inode,
1716 .put_super = ubifs_put_super,
1717 .write_inode = ubifs_write_inode,
1718 .delete_inode = ubifs_delete_inode,
1719 .statfs = ubifs_statfs,
1720 .dirty_inode = ubifs_dirty_inode,
1721 .remount_fs = ubifs_remount_fs,
1722 .show_options = ubifs_show_options,
1723 .sync_fs = ubifs_sync_fs,
1727 * open_ubi - parse UBI device name string and open the UBI device.
1728 * @name: UBI volume name
1729 * @mode: UBI volume open mode
1731 * There are several ways to specify UBI volumes when mounting UBIFS:
1732 * o ubiX_Y - UBI device number X, volume Y;
1733 * o ubiY - UBI device number 0, volume Y;
1734 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1735 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1737 * Alternative '!' separator may be used instead of ':' (because some shells
1738 * like busybox may interpret ':' as an NFS host name separator). This function
1739 * returns ubi volume object in case of success and a negative error code in
1742 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1747 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1748 return ERR_PTR(-EINVAL);
1750 /* ubi:NAME method */
1751 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1752 return ubi_open_volume_nm(0, name + 4, mode);
1754 if (!isdigit(name[3]))
1755 return ERR_PTR(-EINVAL);
1757 dev = simple_strtoul(name + 3, &endptr, 0);
1760 if (*endptr == '\0')
1761 return ubi_open_volume(0, dev, mode);
1764 if (*endptr == '_' && isdigit(endptr[1])) {
1765 vol = simple_strtoul(endptr + 1, &endptr, 0);
1766 if (*endptr != '\0')
1767 return ERR_PTR(-EINVAL);
1768 return ubi_open_volume(dev, vol, mode);
1771 /* ubiX:NAME method */
1772 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1773 return ubi_open_volume_nm(dev, ++endptr, mode);
1775 return ERR_PTR(-EINVAL);
1778 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1780 struct ubi_volume_desc *ubi = sb->s_fs_info;
1781 struct ubifs_info *c;
1785 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1789 spin_lock_init(&c->cnt_lock);
1790 spin_lock_init(&c->cs_lock);
1791 spin_lock_init(&c->buds_lock);
1792 spin_lock_init(&c->space_lock);
1793 spin_lock_init(&c->orphan_lock);
1794 init_rwsem(&c->commit_sem);
1795 mutex_init(&c->lp_mutex);
1796 mutex_init(&c->tnc_mutex);
1797 mutex_init(&c->log_mutex);
1798 mutex_init(&c->mst_mutex);
1799 mutex_init(&c->umount_mutex);
1800 mutex_init(&c->bu_mutex);
1801 init_waitqueue_head(&c->cmt_wq);
1803 c->old_idx = RB_ROOT;
1804 c->size_tree = RB_ROOT;
1805 c->orph_tree = RB_ROOT;
1806 INIT_LIST_HEAD(&c->infos_list);
1807 INIT_LIST_HEAD(&c->idx_gc);
1808 INIT_LIST_HEAD(&c->replay_list);
1809 INIT_LIST_HEAD(&c->replay_buds);
1810 INIT_LIST_HEAD(&c->uncat_list);
1811 INIT_LIST_HEAD(&c->empty_list);
1812 INIT_LIST_HEAD(&c->freeable_list);
1813 INIT_LIST_HEAD(&c->frdi_idx_list);
1814 INIT_LIST_HEAD(&c->unclean_leb_list);
1815 INIT_LIST_HEAD(&c->old_buds);
1816 INIT_LIST_HEAD(&c->orph_list);
1817 INIT_LIST_HEAD(&c->orph_new);
1819 c->highest_inum = UBIFS_FIRST_INO;
1820 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1822 ubi_get_volume_info(ubi, &c->vi);
1823 ubi_get_device_info(c->vi.ubi_num, &c->di);
1825 /* Re-open the UBI device in read-write mode */
1826 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1827 if (IS_ERR(c->ubi)) {
1828 err = PTR_ERR(c->ubi);
1833 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1834 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1835 * which means the user would have to wait not just for their own I/O
1836 * but the read-ahead I/O as well i.e. completely pointless.
1838 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1840 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1841 c->bdi.unplug_io_fn = default_unplug_io_fn;
1842 err = bdi_init(&c->bdi);
1846 err = ubifs_parse_options(c, data, 0);
1853 sb->s_magic = UBIFS_SUPER_MAGIC;
1854 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1855 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1856 sb->s_dev = c->vi.cdev;
1857 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1858 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1859 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1860 sb->s_op = &ubifs_super_operations;
1862 mutex_lock(&c->umount_mutex);
1863 err = mount_ubifs(c);
1865 ubifs_assert(err < 0);
1869 /* Read the root inode */
1870 root = ubifs_iget(sb, UBIFS_ROOT_INO);
1872 err = PTR_ERR(root);
1876 sb->s_root = d_alloc_root(root);
1880 mutex_unlock(&c->umount_mutex);
1888 mutex_unlock(&c->umount_mutex);
1890 bdi_destroy(&c->bdi);
1892 ubi_close_volume(c->ubi);
1898 static int sb_test(struct super_block *sb, void *data)
1902 return sb->s_dev == *dev;
1905 static int sb_set(struct super_block *sb, void *data)
1913 static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
1914 const char *name, void *data, struct vfsmount *mnt)
1916 struct ubi_volume_desc *ubi;
1917 struct ubi_volume_info vi;
1918 struct super_block *sb;
1921 dbg_gen("name %s, flags %#x", name, flags);
1924 * Get UBI device number and volume ID. Mount it read-only so far
1925 * because this might be a new mount point, and UBI allows only one
1926 * read-write user at a time.
1928 ubi = open_ubi(name, UBI_READONLY);
1930 ubifs_err("cannot open \"%s\", error %d",
1931 name, (int)PTR_ERR(ubi));
1932 return PTR_ERR(ubi);
1934 ubi_get_volume_info(ubi, &vi);
1936 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
1938 sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
1945 /* A new mount point for already mounted UBIFS */
1946 dbg_gen("this ubi volume is already mounted");
1947 if ((flags ^ sb->s_flags) & MS_RDONLY) {
1952 sb->s_flags = flags;
1954 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
1957 sb->s_fs_info = ubi;
1958 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
1961 /* We do not support atime */
1962 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
1965 /* 'fill_super()' opens ubi again so we must close it here */
1966 ubi_close_volume(ubi);
1968 return simple_set_mnt(mnt, sb);
1971 up_write(&sb->s_umount);
1972 deactivate_super(sb);
1974 ubi_close_volume(ubi);
1978 static void ubifs_kill_sb(struct super_block *sb)
1980 struct ubifs_info *c = sb->s_fs_info;
1983 * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
1984 * in order to be outside BKL.
1986 if (sb->s_root && !(sb->s_flags & MS_RDONLY))
1987 commit_on_unmount(c);
1988 /* The un-mount routine is actually done in put_super() */
1989 generic_shutdown_super(sb);
1992 static struct file_system_type ubifs_fs_type = {
1994 .owner = THIS_MODULE,
1995 .get_sb = ubifs_get_sb,
1996 .kill_sb = ubifs_kill_sb
2000 * Inode slab cache constructor.
2002 static void inode_slab_ctor(void *obj)
2004 struct ubifs_inode *ui = obj;
2005 inode_init_once(&ui->vfs_inode);
2008 static int __init ubifs_init(void)
2012 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2014 /* Make sure node sizes are 8-byte aligned */
2015 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2016 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2017 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2018 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2019 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2020 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2021 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2022 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2023 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2024 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2025 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2027 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2028 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2029 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2030 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2031 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2032 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2034 /* Check min. node size */
2035 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2036 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2037 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2038 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2040 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2041 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2042 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2043 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2045 /* Defined node sizes */
2046 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2047 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2048 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2049 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2052 * We use 2 bit wide bit-fields to store compression type, which should
2053 * be amended if more compressors are added. The bit-fields are:
2054 * @compr_type in 'struct ubifs_inode', @default_compr in
2055 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2057 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2060 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2061 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2063 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2064 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2065 " at least 4096 bytes",
2066 (unsigned int)PAGE_CACHE_SIZE);
2070 err = register_filesystem(&ubifs_fs_type);
2072 ubifs_err("cannot register file system, error %d", err);
2077 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2078 sizeof(struct ubifs_inode), 0,
2079 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2081 if (!ubifs_inode_slab)
2084 register_shrinker(&ubifs_shrinker_info);
2086 err = ubifs_compressors_init();
2090 err = dbg_debugfs_init();
2097 ubifs_compressors_exit();
2099 unregister_shrinker(&ubifs_shrinker_info);
2100 kmem_cache_destroy(ubifs_inode_slab);
2102 unregister_filesystem(&ubifs_fs_type);
2105 /* late_initcall to let compressors initialize first */
2106 late_initcall(ubifs_init);
2108 static void __exit ubifs_exit(void)
2110 ubifs_assert(list_empty(&ubifs_infos));
2111 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2114 ubifs_compressors_exit();
2115 unregister_shrinker(&ubifs_shrinker_info);
2116 kmem_cache_destroy(ubifs_inode_slab);
2117 unregister_filesystem(&ubifs_fs_type);
2119 module_exit(ubifs_exit);
2121 MODULE_LICENSE("GPL");
2122 MODULE_VERSION(__stringify(UBIFS_VERSION));
2123 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2124 MODULE_DESCRIPTION("UBIFS - UBI File System");