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>
40 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
41 * allocating too much.
43 #define UBIFS_KMALLOC_OK (128*1024)
45 /* Slab cache for UBIFS inodes */
46 struct kmem_cache *ubifs_inode_slab;
48 /* UBIFS TNC shrinker description */
49 static struct shrinker ubifs_shrinker_info = {
50 .shrink = ubifs_shrinker,
51 .seeks = DEFAULT_SEEKS,
55 * validate_inode - validate inode.
56 * @c: UBIFS file-system description object
57 * @inode: the inode to validate
59 * This is a helper function for 'ubifs_iget()' which validates various fields
60 * of a newly built inode to make sure they contain sane values and prevent
61 * possible vulnerabilities. Returns zero if the inode is all right and
62 * a non-zero error code if not.
64 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
67 const struct ubifs_inode *ui = ubifs_inode(inode);
69 if (inode->i_size > c->max_inode_sz) {
70 ubifs_err("inode is too large (%lld)",
71 (long long)inode->i_size);
75 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
76 ubifs_err("unknown compression type %d", ui->compr_type);
80 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
83 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
86 if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
89 if (!ubifs_compr_present(ui->compr_type)) {
90 ubifs_warn("inode %lu uses '%s' compression, but it was not "
91 "compiled in", inode->i_ino,
92 ubifs_compr_name(ui->compr_type));
95 err = dbg_check_dir_size(c, inode);
99 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
103 struct ubifs_ino_node *ino;
104 struct ubifs_info *c = sb->s_fs_info;
106 struct ubifs_inode *ui;
108 dbg_gen("inode %lu", inum);
110 inode = iget_locked(sb, inum);
112 return ERR_PTR(-ENOMEM);
113 if (!(inode->i_state & I_NEW))
115 ui = ubifs_inode(inode);
117 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
123 ino_key_init(c, &key, inode->i_ino);
125 err = ubifs_tnc_lookup(c, &key, ino);
129 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
130 inode->i_nlink = le32_to_cpu(ino->nlink);
131 inode->i_uid = le32_to_cpu(ino->uid);
132 inode->i_gid = le32_to_cpu(ino->gid);
133 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
134 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
135 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
136 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
137 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
138 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
139 inode->i_mode = le32_to_cpu(ino->mode);
140 inode->i_size = le64_to_cpu(ino->size);
142 ui->data_len = le32_to_cpu(ino->data_len);
143 ui->flags = le32_to_cpu(ino->flags);
144 ui->compr_type = le16_to_cpu(ino->compr_type);
145 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
146 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
147 ui->xattr_size = le32_to_cpu(ino->xattr_size);
148 ui->xattr_names = le32_to_cpu(ino->xattr_names);
149 ui->synced_i_size = ui->ui_size = inode->i_size;
151 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
153 err = validate_inode(c, inode);
157 /* Disable read-ahead */
158 inode->i_mapping->backing_dev_info = &c->bdi;
160 switch (inode->i_mode & S_IFMT) {
162 inode->i_mapping->a_ops = &ubifs_file_address_operations;
163 inode->i_op = &ubifs_file_inode_operations;
164 inode->i_fop = &ubifs_file_operations;
166 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
171 memcpy(ui->data, ino->data, ui->data_len);
172 ((char *)ui->data)[ui->data_len] = '\0';
173 } else if (ui->data_len != 0) {
179 inode->i_op = &ubifs_dir_inode_operations;
180 inode->i_fop = &ubifs_dir_operations;
181 if (ui->data_len != 0) {
187 inode->i_op = &ubifs_symlink_inode_operations;
188 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
192 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
197 memcpy(ui->data, ino->data, ui->data_len);
198 ((char *)ui->data)[ui->data_len] = '\0';
204 union ubifs_dev_desc *dev;
206 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
212 dev = (union ubifs_dev_desc *)ino->data;
213 if (ui->data_len == sizeof(dev->new))
214 rdev = new_decode_dev(le32_to_cpu(dev->new));
215 else if (ui->data_len == sizeof(dev->huge))
216 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
221 memcpy(ui->data, ino->data, ui->data_len);
222 inode->i_op = &ubifs_file_inode_operations;
223 init_special_inode(inode, inode->i_mode, rdev);
228 inode->i_op = &ubifs_file_inode_operations;
229 init_special_inode(inode, inode->i_mode, 0);
230 if (ui->data_len != 0) {
241 ubifs_set_inode_flags(inode);
242 unlock_new_inode(inode);
246 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
247 dbg_dump_node(c, ino);
248 dbg_dump_inode(c, inode);
253 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
258 static struct inode *ubifs_alloc_inode(struct super_block *sb)
260 struct ubifs_inode *ui;
262 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
266 memset((void *)ui + sizeof(struct inode), 0,
267 sizeof(struct ubifs_inode) - sizeof(struct inode));
268 mutex_init(&ui->ui_mutex);
269 spin_lock_init(&ui->ui_lock);
270 return &ui->vfs_inode;
273 static void ubifs_destroy_inode(struct inode *inode)
275 struct ubifs_inode *ui = ubifs_inode(inode);
278 kmem_cache_free(ubifs_inode_slab, inode);
282 * Note, Linux write-back code calls this without 'i_mutex'.
284 static int ubifs_write_inode(struct inode *inode, int wait)
287 struct ubifs_info *c = inode->i_sb->s_fs_info;
288 struct ubifs_inode *ui = ubifs_inode(inode);
290 ubifs_assert(!ui->xattr);
291 if (is_bad_inode(inode))
294 mutex_lock(&ui->ui_mutex);
296 * Due to races between write-back forced by budgeting
297 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
298 * have already been synchronized, do not do this again. This might
299 * also happen if it was synchronized in an VFS operation, e.g.
303 mutex_unlock(&ui->ui_mutex);
308 * As an optimization, do not write orphan inodes to the media just
309 * because this is not needed.
311 dbg_gen("inode %lu, mode %#x, nlink %u",
312 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
313 if (inode->i_nlink) {
314 err = ubifs_jnl_write_inode(c, inode);
316 ubifs_err("can't write inode %lu, error %d",
321 mutex_unlock(&ui->ui_mutex);
322 ubifs_release_dirty_inode_budget(c, ui);
326 static void ubifs_delete_inode(struct inode *inode)
329 struct ubifs_info *c = inode->i_sb->s_fs_info;
330 struct ubifs_inode *ui = ubifs_inode(inode);
334 * Extended attribute inode deletions are fully handled in
335 * 'ubifs_removexattr()'. These inodes are special and have
336 * limited usage, so there is nothing to do here.
340 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
341 ubifs_assert(!atomic_read(&inode->i_count));
342 ubifs_assert(inode->i_nlink == 0);
344 truncate_inode_pages(&inode->i_data, 0);
345 if (is_bad_inode(inode))
348 ui->ui_size = inode->i_size = 0;
349 err = ubifs_jnl_delete_inode(c, inode);
352 * Worst case we have a lost orphan inode wasting space, so a
353 * simple error message is OK here.
355 ubifs_err("can't delete inode %lu, error %d",
360 ubifs_release_dirty_inode_budget(c, ui);
364 static void ubifs_dirty_inode(struct inode *inode)
366 struct ubifs_inode *ui = ubifs_inode(inode);
368 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
371 dbg_gen("inode %lu", inode->i_ino);
375 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
377 struct ubifs_info *c = dentry->d_sb->s_fs_info;
378 unsigned long long free;
379 __le32 *uuid = (__le32 *)c->uuid;
381 free = ubifs_get_free_space(c);
382 dbg_gen("free space %lld bytes (%lld blocks)",
383 free, free >> UBIFS_BLOCK_SHIFT);
385 buf->f_type = UBIFS_SUPER_MAGIC;
386 buf->f_bsize = UBIFS_BLOCK_SIZE;
387 buf->f_blocks = c->block_cnt;
388 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
389 if (free > c->report_rp_size)
390 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
395 buf->f_namelen = UBIFS_MAX_NLEN;
396 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
397 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
401 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
403 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
405 if (c->mount_opts.unmount_mode == 2)
406 seq_printf(s, ",fast_unmount");
407 else if (c->mount_opts.unmount_mode == 1)
408 seq_printf(s, ",norm_unmount");
410 if (c->mount_opts.bulk_read == 2)
411 seq_printf(s, ",bulk_read");
412 else if (c->mount_opts.bulk_read == 1)
413 seq_printf(s, ",no_bulk_read");
415 if (c->mount_opts.chk_data_crc == 2)
416 seq_printf(s, ",chk_data_crc");
417 else if (c->mount_opts.chk_data_crc == 1)
418 seq_printf(s, ",no_chk_data_crc");
420 if (c->mount_opts.override_compr) {
421 seq_printf(s, ",compr=");
422 seq_printf(s, ubifs_compr_name(c->mount_opts.compr_type));
428 static int ubifs_sync_fs(struct super_block *sb, int wait)
430 struct ubifs_info *c = sb->s_fs_info;
435 for (i = 0; i < c->jhead_cnt; i++) {
436 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
441 /* Commit the journal unless it has too little data */
442 spin_lock(&c->buds_lock);
443 bud_bytes = c->bud_bytes;
444 spin_unlock(&c->buds_lock);
445 if (bud_bytes > c->leb_size) {
446 err = ubifs_run_commit(c);
453 * We ought to call sync for c->ubi but it does not have one. If it had
454 * it would in turn call mtd->sync, however mtd operations are
455 * synchronous anyway, so we don't lose any sleep here.
461 * init_constants_early - initialize UBIFS constants.
462 * @c: UBIFS file-system description object
464 * This function initialize UBIFS constants which do not need the superblock to
465 * be read. It also checks that the UBI volume satisfies basic UBIFS
466 * requirements. Returns zero in case of success and a negative error code in
469 static int init_constants_early(struct ubifs_info *c)
471 if (c->vi.corrupted) {
472 ubifs_warn("UBI volume is corrupted - read-only mode");
477 ubifs_msg("read-only UBI device");
481 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
482 ubifs_msg("static UBI volume - read-only mode");
486 c->leb_cnt = c->vi.size;
487 c->leb_size = c->vi.usable_leb_size;
488 c->half_leb_size = c->leb_size / 2;
489 c->min_io_size = c->di.min_io_size;
490 c->min_io_shift = fls(c->min_io_size) - 1;
492 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
493 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
494 c->leb_size, UBIFS_MIN_LEB_SZ);
498 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
499 ubifs_err("too few LEBs (%d), min. is %d",
500 c->leb_cnt, UBIFS_MIN_LEB_CNT);
504 if (!is_power_of_2(c->min_io_size)) {
505 ubifs_err("bad min. I/O size %d", c->min_io_size);
510 * UBIFS aligns all node to 8-byte boundary, so to make function in
511 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
514 if (c->min_io_size < 8) {
519 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
520 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
523 * Initialize node length ranges which are mostly needed for node
526 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
527 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
528 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
529 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
530 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
531 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
533 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
534 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
535 c->ranges[UBIFS_ORPH_NODE].min_len =
536 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
537 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
538 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
539 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
540 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
541 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
542 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
543 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
545 * Minimum indexing node size is amended later when superblock is
546 * read and the key length is known.
548 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
550 * Maximum indexing node size is amended later when superblock is
551 * read and the fanout is known.
553 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
556 * Initialize dead and dark LEB space watermarks.
558 * Dead space is the space which cannot be used. Its watermark is
559 * equivalent to min. I/O unit or minimum node size if it is greater
560 * then min. I/O unit.
562 * Dark space is the space which might be used, or might not, depending
563 * on which node should be written to the LEB. Its watermark is
564 * equivalent to maximum UBIFS node size.
566 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
567 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
570 * Calculate how many bytes would be wasted at the end of LEB if it was
571 * fully filled with data nodes of maximum size. This is used in
572 * calculations when reporting free space.
574 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
576 /* Buffer size for bulk-reads */
577 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
578 if (c->max_bu_buf_len > c->leb_size)
579 c->max_bu_buf_len = c->leb_size;
584 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
585 * @c: UBIFS file-system description object
586 * @lnum: LEB the write-buffer was synchronized to
587 * @free: how many free bytes left in this LEB
588 * @pad: how many bytes were padded
590 * This is a callback function which is called by the I/O unit when the
591 * write-buffer is synchronized. We need this to correctly maintain space
592 * accounting in bud logical eraseblocks. This function returns zero in case of
593 * success and a negative error code in case of failure.
595 * This function actually belongs to the journal, but we keep it here because
596 * we want to keep it static.
598 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
600 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
604 * init_constants_late - initialize UBIFS constants.
605 * @c: UBIFS file-system description object
607 * This is a helper function which initializes various UBIFS constants after
608 * the superblock has been read. It also checks various UBIFS parameters and
609 * makes sure they are all right. Returns zero in case of success and a
610 * negative error code in case of failure.
612 static int init_constants_late(struct ubifs_info *c)
617 c->main_bytes = (long long)c->main_lebs * c->leb_size;
618 c->max_znode_sz = sizeof(struct ubifs_znode) +
619 c->fanout * sizeof(struct ubifs_zbranch);
621 tmp = ubifs_idx_node_sz(c, 1);
622 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
623 c->min_idx_node_sz = ALIGN(tmp, 8);
625 tmp = ubifs_idx_node_sz(c, c->fanout);
626 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
627 c->max_idx_node_sz = ALIGN(tmp, 8);
629 /* Make sure LEB size is large enough to fit full commit */
630 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
631 tmp = ALIGN(tmp, c->min_io_size);
632 if (tmp > c->leb_size) {
633 dbg_err("too small LEB size %d, at least %d needed",
639 * Make sure that the log is large enough to fit reference nodes for
640 * all buds plus one reserved LEB.
642 tmp64 = c->max_bud_bytes;
643 tmp = do_div(tmp64, c->leb_size);
644 c->max_bud_cnt = tmp64 + !!tmp;
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 = (uint64_t)(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, and assume only one journal head is available.
700 tmp64 = c->main_lebs - 2 - c->jhead_cnt + 1;
701 tmp64 *= (uint64_t)c->leb_size - c->leb_overhead;
702 tmp64 = ubifs_reported_space(c, tmp64);
703 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
709 * take_gc_lnum - reserve GC LEB.
710 * @c: UBIFS file-system description object
712 * This function ensures that the LEB reserved for garbage collection is
713 * unmapped and is marked as "taken" in lprops. We also have to set free space
714 * to LEB size and dirty space to zero, because lprops may contain out-of-date
715 * information if the file-system was un-mounted before it has been committed.
716 * This function returns zero in case of success and a negative error code in
719 static int take_gc_lnum(struct ubifs_info *c)
723 if (c->gc_lnum == -1) {
724 ubifs_err("no LEB for GC");
728 err = ubifs_leb_unmap(c, c->gc_lnum);
732 /* And we have to tell lprops that this LEB is taken */
733 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
739 * alloc_wbufs - allocate write-buffers.
740 * @c: UBIFS file-system description object
742 * This helper function allocates and initializes UBIFS write-buffers. Returns
743 * zero in case of success and %-ENOMEM in case of failure.
745 static int alloc_wbufs(struct ubifs_info *c)
749 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
754 /* Initialize journal heads */
755 for (i = 0; i < c->jhead_cnt; i++) {
756 INIT_LIST_HEAD(&c->jheads[i].buds_list);
757 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
761 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
762 c->jheads[i].wbuf.jhead = i;
765 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
767 * Garbage Collector head likely contains long-term data and
768 * does not need to be synchronized by timer.
770 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
771 c->jheads[GCHD].wbuf.timeout = 0;
777 * free_wbufs - free write-buffers.
778 * @c: UBIFS file-system description object
780 static void free_wbufs(struct ubifs_info *c)
785 for (i = 0; i < c->jhead_cnt; i++) {
786 kfree(c->jheads[i].wbuf.buf);
787 kfree(c->jheads[i].wbuf.inodes);
795 * free_orphans - free orphans.
796 * @c: UBIFS file-system description object
798 static void free_orphans(struct ubifs_info *c)
800 struct ubifs_orphan *orph;
802 while (c->orph_dnext) {
803 orph = c->orph_dnext;
804 c->orph_dnext = orph->dnext;
805 list_del(&orph->list);
809 while (!list_empty(&c->orph_list)) {
810 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
811 list_del(&orph->list);
813 dbg_err("orphan list not empty at unmount");
821 * free_buds - free per-bud objects.
822 * @c: UBIFS file-system description object
824 static void free_buds(struct ubifs_info *c)
826 struct rb_node *this = c->buds.rb_node;
827 struct ubifs_bud *bud;
831 this = this->rb_left;
832 else if (this->rb_right)
833 this = this->rb_right;
835 bud = rb_entry(this, struct ubifs_bud, rb);
836 this = rb_parent(this);
838 if (this->rb_left == &bud->rb)
839 this->rb_left = NULL;
841 this->rb_right = NULL;
849 * check_volume_empty - check if the UBI volume is empty.
850 * @c: UBIFS file-system description object
852 * This function checks if the UBIFS volume is empty by looking if its LEBs are
853 * mapped or not. The result of checking is stored in the @c->empty variable.
854 * Returns zero in case of success and a negative error code in case of
857 static int check_volume_empty(struct ubifs_info *c)
862 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
863 err = ubi_is_mapped(c->ubi, lnum);
864 if (unlikely(err < 0))
878 * UBIFS mount options.
880 * Opt_fast_unmount: do not run a journal commit before un-mounting
881 * Opt_norm_unmount: run a journal commit before un-mounting
882 * Opt_bulk_read: enable bulk-reads
883 * Opt_no_bulk_read: disable bulk-reads
884 * Opt_chk_data_crc: check CRCs when reading data nodes
885 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
886 * Opt_override_compr: override default compressor
887 * Opt_err: just end of array marker
900 static const match_table_t tokens = {
901 {Opt_fast_unmount, "fast_unmount"},
902 {Opt_norm_unmount, "norm_unmount"},
903 {Opt_bulk_read, "bulk_read"},
904 {Opt_no_bulk_read, "no_bulk_read"},
905 {Opt_chk_data_crc, "chk_data_crc"},
906 {Opt_no_chk_data_crc, "no_chk_data_crc"},
907 {Opt_override_compr, "compr=%s"},
912 * ubifs_parse_options - parse mount parameters.
913 * @c: UBIFS file-system description object
914 * @options: parameters to parse
915 * @is_remount: non-zero if this is FS re-mount
917 * This function parses UBIFS mount options and returns zero in case success
918 * and a negative error code in case of failure.
920 static int ubifs_parse_options(struct ubifs_info *c, char *options,
924 substring_t args[MAX_OPT_ARGS];
929 while ((p = strsep(&options, ","))) {
935 token = match_token(p, tokens, args);
937 case Opt_fast_unmount:
938 c->mount_opts.unmount_mode = 2;
941 case Opt_norm_unmount:
942 c->mount_opts.unmount_mode = 1;
946 c->mount_opts.bulk_read = 2;
949 case Opt_no_bulk_read:
950 c->mount_opts.bulk_read = 1;
953 case Opt_chk_data_crc:
954 c->mount_opts.chk_data_crc = 2;
955 c->no_chk_data_crc = 0;
957 case Opt_no_chk_data_crc:
958 c->mount_opts.chk_data_crc = 1;
959 c->no_chk_data_crc = 1;
961 case Opt_override_compr:
963 char *name = match_strdup(&args[0]);
967 if (!strcmp(name, "none"))
968 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
969 else if (!strcmp(name, "lzo"))
970 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
971 else if (!strcmp(name, "zlib"))
972 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
974 ubifs_err("unknown compressor \"%s\"", name);
979 c->mount_opts.override_compr = 1;
980 c->default_compr = c->mount_opts.compr_type;
984 ubifs_err("unrecognized mount option \"%s\" "
985 "or missing value", p);
994 * destroy_journal - destroy journal data structures.
995 * @c: UBIFS file-system description object
997 * This function destroys journal data structures including those that may have
998 * been created by recovery functions.
1000 static void destroy_journal(struct ubifs_info *c)
1002 while (!list_empty(&c->unclean_leb_list)) {
1003 struct ubifs_unclean_leb *ucleb;
1005 ucleb = list_entry(c->unclean_leb_list.next,
1006 struct ubifs_unclean_leb, list);
1007 list_del(&ucleb->list);
1010 while (!list_empty(&c->old_buds)) {
1011 struct ubifs_bud *bud;
1013 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1014 list_del(&bud->list);
1017 ubifs_destroy_idx_gc(c);
1018 ubifs_destroy_size_tree(c);
1024 * bu_init - initialize bulk-read information.
1025 * @c: UBIFS file-system description object
1027 static void bu_init(struct ubifs_info *c)
1029 ubifs_assert(c->bulk_read == 1);
1032 return; /* Already initialized */
1035 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1037 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1038 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1042 /* Just disable bulk-read */
1043 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1044 "disabling it", c->max_bu_buf_len);
1045 c->mount_opts.bulk_read = 1;
1052 * mount_ubifs - mount UBIFS file-system.
1053 * @c: UBIFS file-system description object
1055 * This function mounts UBIFS file system. Returns zero in case of success and
1056 * a negative error code in case of failure.
1058 * Note, the function does not de-allocate resources it it fails half way
1059 * through, and the caller has to do this instead.
1061 static int mount_ubifs(struct ubifs_info *c)
1063 struct super_block *sb = c->vfs_sb;
1064 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1068 err = init_constants_early(c);
1072 #ifdef CONFIG_UBIFS_FS_DEBUG
1073 c->dbg_buf = vmalloc(c->leb_size);
1078 err = check_volume_empty(c);
1082 if (c->empty && (mounted_read_only || c->ro_media)) {
1084 * This UBI volume is empty, and read-only, or the file system
1085 * is mounted read-only - we cannot format it.
1087 ubifs_err("can't format empty UBI volume: read-only %s",
1088 c->ro_media ? "UBI volume" : "mount");
1093 if (c->ro_media && !mounted_read_only) {
1094 ubifs_err("cannot mount read-write - read-only media");
1100 * The requirement for the buffer is that it should fit indexing B-tree
1101 * height amount of integers. We assume the height if the TNC tree will
1105 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1106 if (!c->bottom_up_buf)
1109 c->sbuf = vmalloc(c->leb_size);
1113 if (!mounted_read_only) {
1114 c->ileb_buf = vmalloc(c->leb_size);
1119 if (c->bulk_read == 1)
1123 * We have to check all CRCs, even for data nodes, when we mount the FS
1124 * (specifically, when we are replaying).
1126 c->always_chk_crc = 1;
1128 err = ubifs_read_superblock(c);
1133 * Make sure the compressor which is set as default in the superblock
1134 * or overriden by mount options is actually compiled in.
1136 if (!ubifs_compr_present(c->default_compr)) {
1137 ubifs_err("'compressor \"%s\" is not compiled in",
1138 ubifs_compr_name(c->default_compr));
1142 dbg_failure_mode_registration(c);
1144 err = init_constants_late(c);
1148 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1149 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1150 c->cbuf = kmalloc(sz, GFP_NOFS);
1156 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1157 if (!mounted_read_only) {
1158 err = alloc_wbufs(c);
1162 /* Create background thread */
1163 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1164 if (IS_ERR(c->bgt)) {
1165 err = PTR_ERR(c->bgt);
1167 ubifs_err("cannot spawn \"%s\", error %d",
1171 wake_up_process(c->bgt);
1174 err = ubifs_read_master(c);
1178 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1179 ubifs_msg("recovery needed");
1180 c->need_recovery = 1;
1181 if (!mounted_read_only) {
1182 err = ubifs_recover_inl_heads(c, c->sbuf);
1186 } else if (!mounted_read_only) {
1188 * Set the "dirty" flag so that if we reboot uncleanly we
1189 * will notice this immediately on the next mount.
1191 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1192 err = ubifs_write_master(c);
1197 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1201 err = dbg_check_idx_size(c, c->old_idx_sz);
1205 err = ubifs_replay_journal(c);
1209 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1213 if (!mounted_read_only) {
1216 /* Check for enough free space */
1217 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1218 ubifs_err("insufficient available space");
1223 /* Check for enough log space */
1224 lnum = c->lhead_lnum + 1;
1225 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1226 lnum = UBIFS_LOG_LNUM;
1227 if (lnum == c->ltail_lnum) {
1228 err = ubifs_consolidate_log(c);
1233 if (c->need_recovery) {
1234 err = ubifs_recover_size(c);
1237 err = ubifs_rcvry_gc_commit(c);
1239 err = take_gc_lnum(c);
1243 err = dbg_check_lprops(c);
1246 } else if (c->need_recovery) {
1247 err = ubifs_recover_size(c);
1252 spin_lock(&ubifs_infos_lock);
1253 list_add_tail(&c->infos_list, &ubifs_infos);
1254 spin_unlock(&ubifs_infos_lock);
1256 if (c->need_recovery) {
1257 if (mounted_read_only)
1258 ubifs_msg("recovery deferred");
1260 c->need_recovery = 0;
1261 ubifs_msg("recovery completed");
1265 err = dbg_check_filesystem(c);
1269 c->always_chk_crc = 0;
1271 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1272 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1273 if (mounted_read_only)
1274 ubifs_msg("mounted read-only");
1275 x = (long long)c->main_lebs * c->leb_size;
1276 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1277 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1278 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1279 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1280 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1281 ubifs_msg("media format: %d (latest is %d)",
1282 c->fmt_version, UBIFS_FORMAT_VERSION);
1283 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1284 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1285 c->report_rp_size, c->report_rp_size >> 10);
1287 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1288 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1289 dbg_msg("LEB size: %d bytes (%d KiB)",
1290 c->leb_size, c->leb_size >> 10);
1291 dbg_msg("data journal heads: %d",
1292 c->jhead_cnt - NONDATA_JHEADS_CNT);
1293 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1294 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1295 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1296 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1297 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1298 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1299 dbg_msg("fast unmount: %d", c->fast_unmount);
1300 dbg_msg("big_lpt %d", c->big_lpt);
1301 dbg_msg("log LEBs: %d (%d - %d)",
1302 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1303 dbg_msg("LPT area LEBs: %d (%d - %d)",
1304 c->lpt_lebs, c->lpt_first, c->lpt_last);
1305 dbg_msg("orphan area LEBs: %d (%d - %d)",
1306 c->orph_lebs, c->orph_first, c->orph_last);
1307 dbg_msg("main area LEBs: %d (%d - %d)",
1308 c->main_lebs, c->main_first, c->leb_cnt - 1);
1309 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1310 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1311 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1312 dbg_msg("key hash type: %d", c->key_hash_type);
1313 dbg_msg("tree fanout: %d", c->fanout);
1314 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1315 dbg_msg("first main LEB: %d", c->main_first);
1316 dbg_msg("dead watermark: %d", c->dead_wm);
1317 dbg_msg("dark watermark: %d", c->dark_wm);
1318 x = (long long)c->main_lebs * c->dark_wm;
1319 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1320 x, x >> 10, x >> 20);
1321 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1322 c->max_bud_bytes, c->max_bud_bytes >> 10,
1323 c->max_bud_bytes >> 20);
1324 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1325 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1326 c->bg_bud_bytes >> 20);
1327 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1328 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1329 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1330 dbg_msg("commit number: %llu", c->cmt_no);
1335 spin_lock(&ubifs_infos_lock);
1336 list_del(&c->infos_list);
1337 spin_unlock(&ubifs_infos_lock);
1343 ubifs_lpt_free(c, 0);
1346 kfree(c->rcvrd_mst_node);
1348 kthread_stop(c->bgt);
1354 dbg_failure_mode_deregistration(c);
1359 kfree(c->bottom_up_buf);
1360 UBIFS_DBG(vfree(c->dbg_buf));
1365 * ubifs_umount - un-mount UBIFS file-system.
1366 * @c: UBIFS file-system description object
1368 * Note, this function is called to free allocated resourced when un-mounting,
1369 * as well as free resources when an error occurred while we were half way
1370 * through mounting (error path cleanup function). So it has to make sure the
1371 * resource was actually allocated before freeing it.
1373 static void ubifs_umount(struct ubifs_info *c)
1375 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
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_DBG(vfree(c->dbg_buf));
1398 dbg_failure_mode_deregistration(c);
1402 * ubifs_remount_rw - re-mount in read-write mode.
1403 * @c: UBIFS file-system description object
1405 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1406 * mode. This function allocates the needed resources and re-mounts UBIFS in
1409 static int ubifs_remount_rw(struct ubifs_info *c)
1416 mutex_lock(&c->umount_mutex);
1417 c->remounting_rw = 1;
1418 c->always_chk_crc = 1;
1420 /* Check for enough free space */
1421 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1422 ubifs_err("insufficient available space");
1427 if (c->old_leb_cnt != c->leb_cnt) {
1428 struct ubifs_sb_node *sup;
1430 sup = ubifs_read_sb_node(c);
1435 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1436 err = ubifs_write_sb_node(c, sup);
1441 if (c->need_recovery) {
1442 ubifs_msg("completing deferred recovery");
1443 err = ubifs_write_rcvrd_mst_node(c);
1446 err = ubifs_recover_size(c);
1449 err = ubifs_clean_lebs(c, c->sbuf);
1452 err = ubifs_recover_inl_heads(c, c->sbuf);
1457 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1458 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1459 err = ubifs_write_master(c);
1464 c->ileb_buf = vmalloc(c->leb_size);
1470 err = ubifs_lpt_init(c, 0, 1);
1474 err = alloc_wbufs(c);
1478 ubifs_create_buds_lists(c);
1480 /* Create background thread */
1481 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1482 if (IS_ERR(c->bgt)) {
1483 err = PTR_ERR(c->bgt);
1485 ubifs_err("cannot spawn \"%s\", error %d",
1489 wake_up_process(c->bgt);
1491 c->orph_buf = vmalloc(c->leb_size);
1497 /* Check for enough log space */
1498 lnum = c->lhead_lnum + 1;
1499 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1500 lnum = UBIFS_LOG_LNUM;
1501 if (lnum == c->ltail_lnum) {
1502 err = ubifs_consolidate_log(c);
1507 if (c->need_recovery)
1508 err = ubifs_rcvry_gc_commit(c);
1510 err = take_gc_lnum(c);
1514 if (c->need_recovery) {
1515 c->need_recovery = 0;
1516 ubifs_msg("deferred recovery completed");
1519 dbg_gen("re-mounted read-write");
1520 c->vfs_sb->s_flags &= ~MS_RDONLY;
1521 c->remounting_rw = 0;
1522 c->always_chk_crc = 0;
1523 mutex_unlock(&c->umount_mutex);
1530 kthread_stop(c->bgt);
1536 ubifs_lpt_free(c, 1);
1537 c->remounting_rw = 0;
1538 c->always_chk_crc = 0;
1539 mutex_unlock(&c->umount_mutex);
1544 * commit_on_unmount - commit the journal when un-mounting.
1545 * @c: UBIFS file-system description object
1547 * This function is called during un-mounting and re-mounting, and it commits
1548 * the journal unless the "fast unmount" mode is enabled. It also avoids
1549 * committing the journal if it contains too few data.
1551 static void commit_on_unmount(struct ubifs_info *c)
1553 if (!c->fast_unmount) {
1554 long long bud_bytes;
1556 spin_lock(&c->buds_lock);
1557 bud_bytes = c->bud_bytes;
1558 spin_unlock(&c->buds_lock);
1559 if (bud_bytes > c->leb_size)
1560 ubifs_run_commit(c);
1565 * ubifs_remount_ro - re-mount in read-only mode.
1566 * @c: UBIFS file-system description object
1568 * We rely on VFS to have stopped writing. Possibly the background thread could
1569 * be running a commit, however kthread_stop will wait in that case.
1571 static void ubifs_remount_ro(struct ubifs_info *c)
1575 ubifs_assert(!c->need_recovery);
1576 commit_on_unmount(c);
1578 mutex_lock(&c->umount_mutex);
1580 kthread_stop(c->bgt);
1584 for (i = 0; i < c->jhead_cnt; i++) {
1585 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1586 del_timer_sync(&c->jheads[i].wbuf.timer);
1590 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1591 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1592 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1593 err = ubifs_write_master(c);
1595 ubifs_ro_mode(c, err);
1598 ubifs_destroy_idx_gc(c);
1604 ubifs_lpt_free(c, 1);
1605 mutex_unlock(&c->umount_mutex);
1608 static void ubifs_put_super(struct super_block *sb)
1611 struct ubifs_info *c = sb->s_fs_info;
1613 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1616 * The following asserts are only valid if there has not been a failure
1617 * of the media. For example, there will be dirty inodes if we failed
1618 * to write them back because of I/O errors.
1620 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1621 ubifs_assert(c->budg_idx_growth == 0);
1622 ubifs_assert(c->budg_dd_growth == 0);
1623 ubifs_assert(c->budg_data_growth == 0);
1626 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1627 * and file system un-mount. Namely, it prevents the shrinker from
1628 * picking this superblock for shrinking - it will be just skipped if
1629 * the mutex is locked.
1631 mutex_lock(&c->umount_mutex);
1632 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1634 * First of all kill the background thread to make sure it does
1635 * not interfere with un-mounting and freeing resources.
1638 kthread_stop(c->bgt);
1642 /* Synchronize write-buffers */
1644 for (i = 0; i < c->jhead_cnt; i++) {
1645 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1646 del_timer_sync(&c->jheads[i].wbuf.timer);
1650 * On fatal errors c->ro_media is set to 1, in which case we do
1651 * not write the master node.
1655 * We are being cleanly unmounted which means the
1656 * orphans were killed - indicate this in the master
1657 * node. Also save the reserved GC LEB number.
1661 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1662 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1663 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1664 err = ubifs_write_master(c);
1667 * Recovery will attempt to fix the master area
1668 * next mount, so we just print a message and
1669 * continue to unmount normally.
1671 ubifs_err("failed to write master node, "
1677 bdi_destroy(&c->bdi);
1678 ubi_close_volume(c->ubi);
1679 mutex_unlock(&c->umount_mutex);
1683 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1686 struct ubifs_info *c = sb->s_fs_info;
1688 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1690 err = ubifs_parse_options(c, data, 1);
1692 ubifs_err("invalid or unknown remount parameter");
1696 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1697 err = ubifs_remount_rw(c);
1700 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
1701 ubifs_remount_ro(c);
1703 if (c->bulk_read == 1)
1706 dbg_gen("disable bulk-read");
1714 struct super_operations ubifs_super_operations = {
1715 .alloc_inode = ubifs_alloc_inode,
1716 .destroy_inode = ubifs_destroy_inode,
1717 .put_super = ubifs_put_super,
1718 .write_inode = ubifs_write_inode,
1719 .delete_inode = ubifs_delete_inode,
1720 .statfs = ubifs_statfs,
1721 .dirty_inode = ubifs_dirty_inode,
1722 .remount_fs = ubifs_remount_fs,
1723 .show_options = ubifs_show_options,
1724 .sync_fs = ubifs_sync_fs,
1728 * open_ubi - parse UBI device name string and open the UBI device.
1729 * @name: UBI volume name
1730 * @mode: UBI volume open mode
1732 * There are several ways to specify UBI volumes when mounting UBIFS:
1733 * o ubiX_Y - UBI device number X, volume Y;
1734 * o ubiY - UBI device number 0, volume Y;
1735 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1736 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1738 * Alternative '!' separator may be used instead of ':' (because some shells
1739 * like busybox may interpret ':' as an NFS host name separator). This function
1740 * returns ubi volume object in case of success and a negative error code in
1743 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1748 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1749 return ERR_PTR(-EINVAL);
1751 /* ubi:NAME method */
1752 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1753 return ubi_open_volume_nm(0, name + 4, mode);
1755 if (!isdigit(name[3]))
1756 return ERR_PTR(-EINVAL);
1758 dev = simple_strtoul(name + 3, &endptr, 0);
1761 if (*endptr == '\0')
1762 return ubi_open_volume(0, dev, mode);
1765 if (*endptr == '_' && isdigit(endptr[1])) {
1766 vol = simple_strtoul(endptr + 1, &endptr, 0);
1767 if (*endptr != '\0')
1768 return ERR_PTR(-EINVAL);
1769 return ubi_open_volume(dev, vol, mode);
1772 /* ubiX:NAME method */
1773 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1774 return ubi_open_volume_nm(dev, ++endptr, mode);
1776 return ERR_PTR(-EINVAL);
1779 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1781 struct ubi_volume_desc *ubi = sb->s_fs_info;
1782 struct ubifs_info *c;
1786 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1790 spin_lock_init(&c->cnt_lock);
1791 spin_lock_init(&c->cs_lock);
1792 spin_lock_init(&c->buds_lock);
1793 spin_lock_init(&c->space_lock);
1794 spin_lock_init(&c->orphan_lock);
1795 init_rwsem(&c->commit_sem);
1796 mutex_init(&c->lp_mutex);
1797 mutex_init(&c->tnc_mutex);
1798 mutex_init(&c->log_mutex);
1799 mutex_init(&c->mst_mutex);
1800 mutex_init(&c->umount_mutex);
1801 mutex_init(&c->bu_mutex);
1802 init_waitqueue_head(&c->cmt_wq);
1804 c->old_idx = RB_ROOT;
1805 c->size_tree = RB_ROOT;
1806 c->orph_tree = RB_ROOT;
1807 INIT_LIST_HEAD(&c->infos_list);
1808 INIT_LIST_HEAD(&c->idx_gc);
1809 INIT_LIST_HEAD(&c->replay_list);
1810 INIT_LIST_HEAD(&c->replay_buds);
1811 INIT_LIST_HEAD(&c->uncat_list);
1812 INIT_LIST_HEAD(&c->empty_list);
1813 INIT_LIST_HEAD(&c->freeable_list);
1814 INIT_LIST_HEAD(&c->frdi_idx_list);
1815 INIT_LIST_HEAD(&c->unclean_leb_list);
1816 INIT_LIST_HEAD(&c->old_buds);
1817 INIT_LIST_HEAD(&c->orph_list);
1818 INIT_LIST_HEAD(&c->orph_new);
1820 c->highest_inum = UBIFS_FIRST_INO;
1821 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1823 ubi_get_volume_info(ubi, &c->vi);
1824 ubi_get_device_info(c->vi.ubi_num, &c->di);
1826 /* Re-open the UBI device in read-write mode */
1827 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1828 if (IS_ERR(c->ubi)) {
1829 err = PTR_ERR(c->ubi);
1834 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1835 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1836 * which means the user would have to wait not just for their own I/O
1837 * but the read-ahead I/O as well i.e. completely pointless.
1839 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1841 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1842 c->bdi.unplug_io_fn = default_unplug_io_fn;
1843 err = bdi_init(&c->bdi);
1847 err = ubifs_parse_options(c, data, 0);
1854 sb->s_magic = UBIFS_SUPER_MAGIC;
1855 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1856 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1857 sb->s_dev = c->vi.cdev;
1858 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1859 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1860 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1861 sb->s_op = &ubifs_super_operations;
1863 mutex_lock(&c->umount_mutex);
1864 err = mount_ubifs(c);
1866 ubifs_assert(err < 0);
1870 /* Read the root inode */
1871 root = ubifs_iget(sb, UBIFS_ROOT_INO);
1873 err = PTR_ERR(root);
1877 sb->s_root = d_alloc_root(root);
1881 mutex_unlock(&c->umount_mutex);
1890 mutex_unlock(&c->umount_mutex);
1892 bdi_destroy(&c->bdi);
1894 ubi_close_volume(c->ubi);
1900 static int sb_test(struct super_block *sb, void *data)
1904 return sb->s_dev == *dev;
1907 static int sb_set(struct super_block *sb, void *data)
1915 static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
1916 const char *name, void *data, struct vfsmount *mnt)
1918 struct ubi_volume_desc *ubi;
1919 struct ubi_volume_info vi;
1920 struct super_block *sb;
1923 dbg_gen("name %s, flags %#x", name, flags);
1926 * Get UBI device number and volume ID. Mount it read-only so far
1927 * because this might be a new mount point, and UBI allows only one
1928 * read-write user at a time.
1930 ubi = open_ubi(name, UBI_READONLY);
1932 ubifs_err("cannot open \"%s\", error %d",
1933 name, (int)PTR_ERR(ubi));
1934 return PTR_ERR(ubi);
1936 ubi_get_volume_info(ubi, &vi);
1938 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
1940 sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
1947 /* A new mount point for already mounted UBIFS */
1948 dbg_gen("this ubi volume is already mounted");
1949 if ((flags ^ sb->s_flags) & MS_RDONLY) {
1954 sb->s_flags = flags;
1956 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
1959 sb->s_fs_info = ubi;
1960 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
1963 /* We do not support atime */
1964 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
1967 /* 'fill_super()' opens ubi again so we must close it here */
1968 ubi_close_volume(ubi);
1970 return simple_set_mnt(mnt, sb);
1973 up_write(&sb->s_umount);
1974 deactivate_super(sb);
1976 ubi_close_volume(ubi);
1980 static void ubifs_kill_sb(struct super_block *sb)
1982 struct ubifs_info *c = sb->s_fs_info;
1985 * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
1986 * in order to be outside BKL.
1988 if (sb->s_root && !(sb->s_flags & MS_RDONLY))
1989 commit_on_unmount(c);
1990 /* The un-mount routine is actually done in put_super() */
1991 generic_shutdown_super(sb);
1994 static struct file_system_type ubifs_fs_type = {
1996 .owner = THIS_MODULE,
1997 .get_sb = ubifs_get_sb,
1998 .kill_sb = ubifs_kill_sb
2002 * Inode slab cache constructor.
2004 static void inode_slab_ctor(void *obj)
2006 struct ubifs_inode *ui = obj;
2007 inode_init_once(&ui->vfs_inode);
2010 static int __init ubifs_init(void)
2014 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2016 /* Make sure node sizes are 8-byte aligned */
2017 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2018 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2019 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2020 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2021 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2022 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2023 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2024 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2025 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2026 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2027 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2029 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2030 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2031 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2032 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2033 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2034 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2036 /* Check min. node size */
2037 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2038 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2039 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2040 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2042 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2043 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2044 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2045 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2047 /* Defined node sizes */
2048 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2049 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2050 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2051 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2054 * We use 2 bit wide bit-fields to store compression type, which should
2055 * be amended if more compressors are added. The bit-fields are:
2056 * @compr_type in 'struct ubifs_inode', @default_compr in
2057 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2059 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2062 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2063 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2065 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2066 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2067 " at least 4096 bytes",
2068 (unsigned int)PAGE_CACHE_SIZE);
2072 err = register_filesystem(&ubifs_fs_type);
2074 ubifs_err("cannot register file system, error %d", err);
2079 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2080 sizeof(struct ubifs_inode), 0,
2081 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2083 if (!ubifs_inode_slab)
2086 register_shrinker(&ubifs_shrinker_info);
2088 err = ubifs_compressors_init();
2095 unregister_shrinker(&ubifs_shrinker_info);
2096 kmem_cache_destroy(ubifs_inode_slab);
2098 unregister_filesystem(&ubifs_fs_type);
2101 /* late_initcall to let compressors initialize first */
2102 late_initcall(ubifs_init);
2104 static void __exit ubifs_exit(void)
2106 ubifs_assert(list_empty(&ubifs_infos));
2107 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2109 ubifs_compressors_exit();
2110 unregister_shrinker(&ubifs_shrinker_info);
2111 kmem_cache_destroy(ubifs_inode_slab);
2112 unregister_filesystem(&ubifs_fs_type);
2114 module_exit(ubifs_exit);
2116 MODULE_LICENSE("GPL");
2117 MODULE_VERSION(__stringify(UBIFS_VERSION));
2118 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2119 MODULE_DESCRIPTION("UBIFS - UBI File System");