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>
38 #include <linux/writeback.h>
42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
43 * allocating too much.
45 #define UBIFS_KMALLOC_OK (128*1024)
47 /* Slab cache for UBIFS inodes */
48 struct kmem_cache *ubifs_inode_slab;
50 /* UBIFS TNC shrinker description */
51 static struct shrinker ubifs_shrinker_info = {
52 .shrink = ubifs_shrinker,
53 .seeks = DEFAULT_SEEKS,
57 * validate_inode - validate inode.
58 * @c: UBIFS file-system description object
59 * @inode: the inode to validate
61 * This is a helper function for 'ubifs_iget()' which validates various fields
62 * of a newly built inode to make sure they contain sane values and prevent
63 * possible vulnerabilities. Returns zero if the inode is all right and
64 * a non-zero error code if not.
66 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
69 const struct ubifs_inode *ui = ubifs_inode(inode);
71 if (inode->i_size > c->max_inode_sz) {
72 ubifs_err("inode is too large (%lld)",
73 (long long)inode->i_size);
77 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
78 ubifs_err("unknown compression type %d", ui->compr_type);
82 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
85 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
88 if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
91 if (!ubifs_compr_present(ui->compr_type)) {
92 ubifs_warn("inode %lu uses '%s' compression, but it was not "
93 "compiled in", inode->i_ino,
94 ubifs_compr_name(ui->compr_type));
97 err = dbg_check_dir_size(c, inode);
101 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
105 struct ubifs_ino_node *ino;
106 struct ubifs_info *c = sb->s_fs_info;
108 struct ubifs_inode *ui;
110 dbg_gen("inode %lu", inum);
112 inode = iget_locked(sb, inum);
114 return ERR_PTR(-ENOMEM);
115 if (!(inode->i_state & I_NEW))
117 ui = ubifs_inode(inode);
119 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
125 ino_key_init(c, &key, inode->i_ino);
127 err = ubifs_tnc_lookup(c, &key, ino);
131 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
132 inode->i_nlink = le32_to_cpu(ino->nlink);
133 inode->i_uid = le32_to_cpu(ino->uid);
134 inode->i_gid = le32_to_cpu(ino->gid);
135 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
136 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
137 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
138 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
139 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
140 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
141 inode->i_mode = le32_to_cpu(ino->mode);
142 inode->i_size = le64_to_cpu(ino->size);
144 ui->data_len = le32_to_cpu(ino->data_len);
145 ui->flags = le32_to_cpu(ino->flags);
146 ui->compr_type = le16_to_cpu(ino->compr_type);
147 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
148 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
149 ui->xattr_size = le32_to_cpu(ino->xattr_size);
150 ui->xattr_names = le32_to_cpu(ino->xattr_names);
151 ui->synced_i_size = ui->ui_size = inode->i_size;
153 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
155 err = validate_inode(c, inode);
159 /* Disable read-ahead */
160 inode->i_mapping->backing_dev_info = &c->bdi;
162 switch (inode->i_mode & S_IFMT) {
164 inode->i_mapping->a_ops = &ubifs_file_address_operations;
165 inode->i_op = &ubifs_file_inode_operations;
166 inode->i_fop = &ubifs_file_operations;
168 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
173 memcpy(ui->data, ino->data, ui->data_len);
174 ((char *)ui->data)[ui->data_len] = '\0';
175 } else if (ui->data_len != 0) {
181 inode->i_op = &ubifs_dir_inode_operations;
182 inode->i_fop = &ubifs_dir_operations;
183 if (ui->data_len != 0) {
189 inode->i_op = &ubifs_symlink_inode_operations;
190 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
194 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
199 memcpy(ui->data, ino->data, ui->data_len);
200 ((char *)ui->data)[ui->data_len] = '\0';
206 union ubifs_dev_desc *dev;
208 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
214 dev = (union ubifs_dev_desc *)ino->data;
215 if (ui->data_len == sizeof(dev->new))
216 rdev = new_decode_dev(le32_to_cpu(dev->new));
217 else if (ui->data_len == sizeof(dev->huge))
218 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
223 memcpy(ui->data, ino->data, ui->data_len);
224 inode->i_op = &ubifs_file_inode_operations;
225 init_special_inode(inode, inode->i_mode, rdev);
230 inode->i_op = &ubifs_file_inode_operations;
231 init_special_inode(inode, inode->i_mode, 0);
232 if (ui->data_len != 0) {
243 ubifs_set_inode_flags(inode);
244 unlock_new_inode(inode);
248 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
249 dbg_dump_node(c, ino);
250 dbg_dump_inode(c, inode);
255 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
260 static struct inode *ubifs_alloc_inode(struct super_block *sb)
262 struct ubifs_inode *ui;
264 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
268 memset((void *)ui + sizeof(struct inode), 0,
269 sizeof(struct ubifs_inode) - sizeof(struct inode));
270 mutex_init(&ui->ui_mutex);
271 spin_lock_init(&ui->ui_lock);
272 return &ui->vfs_inode;
275 static void ubifs_i_callback(struct rcu_head *head)
277 struct inode *inode = container_of(head, struct inode, i_rcu);
278 struct ubifs_inode *ui = ubifs_inode(inode);
279 INIT_LIST_HEAD(&inode->i_dentry);
280 kmem_cache_free(ubifs_inode_slab, ui);
283 static void ubifs_destroy_inode(struct inode *inode)
285 struct ubifs_inode *ui = ubifs_inode(inode);
288 call_rcu(&inode->i_rcu, ubifs_i_callback);
292 * Note, Linux write-back code calls this without 'i_mutex'.
294 static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
297 struct ubifs_info *c = inode->i_sb->s_fs_info;
298 struct ubifs_inode *ui = ubifs_inode(inode);
300 ubifs_assert(!ui->xattr);
301 if (is_bad_inode(inode))
304 mutex_lock(&ui->ui_mutex);
306 * Due to races between write-back forced by budgeting
307 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
308 * have already been synchronized, do not do this again. This might
309 * also happen if it was synchronized in an VFS operation, e.g.
313 mutex_unlock(&ui->ui_mutex);
318 * As an optimization, do not write orphan inodes to the media just
319 * because this is not needed.
321 dbg_gen("inode %lu, mode %#x, nlink %u",
322 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
323 if (inode->i_nlink) {
324 err = ubifs_jnl_write_inode(c, inode);
326 ubifs_err("can't write inode %lu, error %d",
329 err = dbg_check_inode_size(c, inode, ui->ui_size);
333 mutex_unlock(&ui->ui_mutex);
334 ubifs_release_dirty_inode_budget(c, ui);
338 static void ubifs_evict_inode(struct inode *inode)
341 struct ubifs_info *c = inode->i_sb->s_fs_info;
342 struct ubifs_inode *ui = ubifs_inode(inode);
346 * Extended attribute inode deletions are fully handled in
347 * 'ubifs_removexattr()'. These inodes are special and have
348 * limited usage, so there is nothing to do here.
352 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
353 ubifs_assert(!atomic_read(&inode->i_count));
355 truncate_inode_pages(&inode->i_data, 0);
360 if (is_bad_inode(inode))
363 ui->ui_size = inode->i_size = 0;
364 err = ubifs_jnl_delete_inode(c, inode);
367 * Worst case we have a lost orphan inode wasting space, so a
368 * simple error message is OK here.
370 ubifs_err("can't delete inode %lu, error %d",
375 ubifs_release_dirty_inode_budget(c, ui);
377 /* We've deleted something - clean the "no space" flags */
378 c->bi.nospace = c->bi.nospace_rp = 0;
382 end_writeback(inode);
385 static void ubifs_dirty_inode(struct inode *inode)
387 struct ubifs_inode *ui = ubifs_inode(inode);
389 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
392 dbg_gen("inode %lu", inode->i_ino);
396 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
398 struct ubifs_info *c = dentry->d_sb->s_fs_info;
399 unsigned long long free;
400 __le32 *uuid = (__le32 *)c->uuid;
402 free = ubifs_get_free_space(c);
403 dbg_gen("free space %lld bytes (%lld blocks)",
404 free, free >> UBIFS_BLOCK_SHIFT);
406 buf->f_type = UBIFS_SUPER_MAGIC;
407 buf->f_bsize = UBIFS_BLOCK_SIZE;
408 buf->f_blocks = c->block_cnt;
409 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
410 if (free > c->report_rp_size)
411 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
416 buf->f_namelen = UBIFS_MAX_NLEN;
417 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
418 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
419 ubifs_assert(buf->f_bfree <= c->block_cnt);
423 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
425 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
427 if (c->mount_opts.unmount_mode == 2)
428 seq_printf(s, ",fast_unmount");
429 else if (c->mount_opts.unmount_mode == 1)
430 seq_printf(s, ",norm_unmount");
432 if (c->mount_opts.bulk_read == 2)
433 seq_printf(s, ",bulk_read");
434 else if (c->mount_opts.bulk_read == 1)
435 seq_printf(s, ",no_bulk_read");
437 if (c->mount_opts.chk_data_crc == 2)
438 seq_printf(s, ",chk_data_crc");
439 else if (c->mount_opts.chk_data_crc == 1)
440 seq_printf(s, ",no_chk_data_crc");
442 if (c->mount_opts.override_compr) {
443 seq_printf(s, ",compr=%s",
444 ubifs_compr_name(c->mount_opts.compr_type));
450 static int ubifs_sync_fs(struct super_block *sb, int wait)
453 struct ubifs_info *c = sb->s_fs_info;
456 * Zero @wait is just an advisory thing to help the file system shove
457 * lots of data into the queues, and there will be the second
458 * '->sync_fs()' call, with non-zero @wait.
464 * Synchronize write buffers, because 'ubifs_run_commit()' does not
465 * do this if it waits for an already running commit.
467 for (i = 0; i < c->jhead_cnt; i++) {
468 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
474 * Strictly speaking, it is not necessary to commit the journal here,
475 * synchronizing write-buffers would be enough. But committing makes
476 * UBIFS free space predictions much more accurate, so we want to let
477 * the user be able to get more accurate results of 'statfs()' after
478 * they synchronize the file system.
480 err = ubifs_run_commit(c);
484 return ubi_sync(c->vi.ubi_num);
488 * init_constants_early - initialize UBIFS constants.
489 * @c: UBIFS file-system description object
491 * This function initialize UBIFS constants which do not need the superblock to
492 * be read. It also checks that the UBI volume satisfies basic UBIFS
493 * requirements. Returns zero in case of success and a negative error code in
496 static int init_constants_early(struct ubifs_info *c)
498 if (c->vi.corrupted) {
499 ubifs_warn("UBI volume is corrupted - read-only mode");
504 ubifs_msg("read-only UBI device");
508 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
509 ubifs_msg("static UBI volume - read-only mode");
513 c->leb_cnt = c->vi.size;
514 c->leb_size = c->vi.usable_leb_size;
515 c->leb_start = c->di.leb_start;
516 c->half_leb_size = c->leb_size / 2;
517 c->min_io_size = c->di.min_io_size;
518 c->min_io_shift = fls(c->min_io_size) - 1;
519 c->max_write_size = c->di.max_write_size;
520 c->max_write_shift = fls(c->max_write_size) - 1;
522 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
523 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
524 c->leb_size, UBIFS_MIN_LEB_SZ);
528 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
529 ubifs_err("too few LEBs (%d), min. is %d",
530 c->leb_cnt, UBIFS_MIN_LEB_CNT);
534 if (!is_power_of_2(c->min_io_size)) {
535 ubifs_err("bad min. I/O size %d", c->min_io_size);
540 * Maximum write size has to be greater or equivalent to min. I/O
541 * size, and be multiple of min. I/O size.
543 if (c->max_write_size < c->min_io_size ||
544 c->max_write_size % c->min_io_size ||
545 !is_power_of_2(c->max_write_size)) {
546 ubifs_err("bad write buffer size %d for %d min. I/O unit",
547 c->max_write_size, c->min_io_size);
552 * UBIFS aligns all node to 8-byte boundary, so to make function in
553 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
556 if (c->min_io_size < 8) {
559 if (c->max_write_size < c->min_io_size) {
560 c->max_write_size = c->min_io_size;
561 c->max_write_shift = c->min_io_shift;
565 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
566 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
569 * Initialize node length ranges which are mostly needed for node
572 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
573 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
574 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
575 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
576 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
577 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
579 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
580 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
581 c->ranges[UBIFS_ORPH_NODE].min_len =
582 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
583 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
584 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
585 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
586 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
587 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
588 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
589 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
591 * Minimum indexing node size is amended later when superblock is
592 * read and the key length is known.
594 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
596 * Maximum indexing node size is amended later when superblock is
597 * read and the fanout is known.
599 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
602 * Initialize dead and dark LEB space watermarks. See gc.c for comments
603 * about these values.
605 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
606 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
609 * Calculate how many bytes would be wasted at the end of LEB if it was
610 * fully filled with data nodes of maximum size. This is used in
611 * calculations when reporting free space.
613 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
615 /* Buffer size for bulk-reads */
616 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
617 if (c->max_bu_buf_len > c->leb_size)
618 c->max_bu_buf_len = c->leb_size;
623 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
624 * @c: UBIFS file-system description object
625 * @lnum: LEB the write-buffer was synchronized to
626 * @free: how many free bytes left in this LEB
627 * @pad: how many bytes were padded
629 * This is a callback function which is called by the I/O unit when the
630 * write-buffer is synchronized. We need this to correctly maintain space
631 * accounting in bud logical eraseblocks. This function returns zero in case of
632 * success and a negative error code in case of failure.
634 * This function actually belongs to the journal, but we keep it here because
635 * we want to keep it static.
637 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
639 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
643 * init_constants_sb - initialize UBIFS constants.
644 * @c: UBIFS file-system description object
646 * This is a helper function which initializes various UBIFS constants after
647 * the superblock has been read. It also checks various UBIFS parameters and
648 * makes sure they are all right. Returns zero in case of success and a
649 * negative error code in case of failure.
651 static int init_constants_sb(struct ubifs_info *c)
656 c->main_bytes = (long long)c->main_lebs * c->leb_size;
657 c->max_znode_sz = sizeof(struct ubifs_znode) +
658 c->fanout * sizeof(struct ubifs_zbranch);
660 tmp = ubifs_idx_node_sz(c, 1);
661 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
662 c->min_idx_node_sz = ALIGN(tmp, 8);
664 tmp = ubifs_idx_node_sz(c, c->fanout);
665 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
666 c->max_idx_node_sz = ALIGN(tmp, 8);
668 /* Make sure LEB size is large enough to fit full commit */
669 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
670 tmp = ALIGN(tmp, c->min_io_size);
671 if (tmp > c->leb_size) {
672 dbg_err("too small LEB size %d, at least %d needed",
678 * Make sure that the log is large enough to fit reference nodes for
679 * all buds plus one reserved LEB.
681 tmp64 = c->max_bud_bytes + c->leb_size - 1;
682 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
683 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
686 if (c->log_lebs < tmp) {
687 dbg_err("too small log %d LEBs, required min. %d LEBs",
693 * When budgeting we assume worst-case scenarios when the pages are not
694 * be compressed and direntries are of the maximum size.
696 * Note, data, which may be stored in inodes is budgeted separately, so
697 * it is not included into 'c->bi.inode_budget'.
699 c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
700 c->bi.inode_budget = UBIFS_INO_NODE_SZ;
701 c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
704 * When the amount of flash space used by buds becomes
705 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
706 * The writers are unblocked when the commit is finished. To avoid
707 * writers to be blocked UBIFS initiates background commit in advance,
708 * when number of bud bytes becomes above the limit defined below.
710 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
713 * Ensure minimum journal size. All the bytes in the journal heads are
714 * considered to be used, when calculating the current journal usage.
715 * Consequently, if the journal is too small, UBIFS will treat it as
718 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
719 if (c->bg_bud_bytes < tmp64)
720 c->bg_bud_bytes = tmp64;
721 if (c->max_bud_bytes < tmp64 + c->leb_size)
722 c->max_bud_bytes = tmp64 + c->leb_size;
724 err = ubifs_calc_lpt_geom(c);
728 /* Initialize effective LEB size used in budgeting calculations */
729 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
734 * init_constants_master - initialize UBIFS constants.
735 * @c: UBIFS file-system description object
737 * This is a helper function which initializes various UBIFS constants after
738 * the master node has been read. It also checks various UBIFS parameters and
739 * makes sure they are all right.
741 static void init_constants_master(struct ubifs_info *c)
745 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
746 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
749 * Calculate total amount of FS blocks. This number is not used
750 * internally because it does not make much sense for UBIFS, but it is
751 * necessary to report something for the 'statfs()' call.
753 * Subtract the LEB reserved for GC, the LEB which is reserved for
754 * deletions, minimum LEBs for the index, and assume only one journal
757 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
758 tmp64 *= (long long)c->leb_size - c->leb_overhead;
759 tmp64 = ubifs_reported_space(c, tmp64);
760 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
764 * take_gc_lnum - reserve GC LEB.
765 * @c: UBIFS file-system description object
767 * This function ensures that the LEB reserved for garbage collection is marked
768 * as "taken" in lprops. We also have to set free space to LEB size and dirty
769 * space to zero, because lprops may contain out-of-date information if the
770 * file-system was un-mounted before it has been committed. This function
771 * returns zero in case of success and a negative error code in case of
774 static int take_gc_lnum(struct ubifs_info *c)
778 if (c->gc_lnum == -1) {
779 ubifs_err("no LEB for GC");
783 /* And we have to tell lprops that this LEB is taken */
784 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
790 * alloc_wbufs - allocate write-buffers.
791 * @c: UBIFS file-system description object
793 * This helper function allocates and initializes UBIFS write-buffers. Returns
794 * zero in case of success and %-ENOMEM in case of failure.
796 static int alloc_wbufs(struct ubifs_info *c)
800 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
805 /* Initialize journal heads */
806 for (i = 0; i < c->jhead_cnt; i++) {
807 INIT_LIST_HEAD(&c->jheads[i].buds_list);
808 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
812 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
813 c->jheads[i].wbuf.jhead = i;
816 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
818 * Garbage Collector head likely contains long-term data and
819 * does not need to be synchronized by timer.
821 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
822 c->jheads[GCHD].wbuf.no_timer = 1;
828 * free_wbufs - free write-buffers.
829 * @c: UBIFS file-system description object
831 static void free_wbufs(struct ubifs_info *c)
836 for (i = 0; i < c->jhead_cnt; i++) {
837 kfree(c->jheads[i].wbuf.buf);
838 kfree(c->jheads[i].wbuf.inodes);
846 * free_orphans - free orphans.
847 * @c: UBIFS file-system description object
849 static void free_orphans(struct ubifs_info *c)
851 struct ubifs_orphan *orph;
853 while (c->orph_dnext) {
854 orph = c->orph_dnext;
855 c->orph_dnext = orph->dnext;
856 list_del(&orph->list);
860 while (!list_empty(&c->orph_list)) {
861 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
862 list_del(&orph->list);
864 dbg_err("orphan list not empty at unmount");
872 * free_buds - free per-bud objects.
873 * @c: UBIFS file-system description object
875 static void free_buds(struct ubifs_info *c)
877 struct rb_node *this = c->buds.rb_node;
878 struct ubifs_bud *bud;
882 this = this->rb_left;
883 else if (this->rb_right)
884 this = this->rb_right;
886 bud = rb_entry(this, struct ubifs_bud, rb);
887 this = rb_parent(this);
889 if (this->rb_left == &bud->rb)
890 this->rb_left = NULL;
892 this->rb_right = NULL;
900 * check_volume_empty - check if the UBI volume is empty.
901 * @c: UBIFS file-system description object
903 * This function checks if the UBIFS volume is empty by looking if its LEBs are
904 * mapped or not. The result of checking is stored in the @c->empty variable.
905 * Returns zero in case of success and a negative error code in case of
908 static int check_volume_empty(struct ubifs_info *c)
913 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
914 err = ubi_is_mapped(c->ubi, lnum);
915 if (unlikely(err < 0))
929 * UBIFS mount options.
931 * Opt_fast_unmount: do not run a journal commit before un-mounting
932 * Opt_norm_unmount: run a journal commit before un-mounting
933 * Opt_bulk_read: enable bulk-reads
934 * Opt_no_bulk_read: disable bulk-reads
935 * Opt_chk_data_crc: check CRCs when reading data nodes
936 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
937 * Opt_override_compr: override default compressor
938 * Opt_err: just end of array marker
951 static const match_table_t tokens = {
952 {Opt_fast_unmount, "fast_unmount"},
953 {Opt_norm_unmount, "norm_unmount"},
954 {Opt_bulk_read, "bulk_read"},
955 {Opt_no_bulk_read, "no_bulk_read"},
956 {Opt_chk_data_crc, "chk_data_crc"},
957 {Opt_no_chk_data_crc, "no_chk_data_crc"},
958 {Opt_override_compr, "compr=%s"},
963 * parse_standard_option - parse a standard mount option.
964 * @option: the option to parse
966 * Normally, standard mount options like "sync" are passed to file-systems as
967 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
968 * be present in the options string. This function tries to deal with this
969 * situation and parse standard options. Returns 0 if the option was not
970 * recognized, and the corresponding integer flag if it was.
972 * UBIFS is only interested in the "sync" option, so do not check for anything
975 static int parse_standard_option(const char *option)
977 ubifs_msg("parse %s", option);
978 if (!strcmp(option, "sync"))
979 return MS_SYNCHRONOUS;
984 * ubifs_parse_options - parse mount parameters.
985 * @c: UBIFS file-system description object
986 * @options: parameters to parse
987 * @is_remount: non-zero if this is FS re-mount
989 * This function parses UBIFS mount options and returns zero in case success
990 * and a negative error code in case of failure.
992 static int ubifs_parse_options(struct ubifs_info *c, char *options,
996 substring_t args[MAX_OPT_ARGS];
1001 while ((p = strsep(&options, ","))) {
1007 token = match_token(p, tokens, args);
1010 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
1011 * We accept them in order to be backward-compatible. But this
1012 * should be removed at some point.
1014 case Opt_fast_unmount:
1015 c->mount_opts.unmount_mode = 2;
1017 case Opt_norm_unmount:
1018 c->mount_opts.unmount_mode = 1;
1021 c->mount_opts.bulk_read = 2;
1024 case Opt_no_bulk_read:
1025 c->mount_opts.bulk_read = 1;
1028 case Opt_chk_data_crc:
1029 c->mount_opts.chk_data_crc = 2;
1030 c->no_chk_data_crc = 0;
1032 case Opt_no_chk_data_crc:
1033 c->mount_opts.chk_data_crc = 1;
1034 c->no_chk_data_crc = 1;
1036 case Opt_override_compr:
1038 char *name = match_strdup(&args[0]);
1042 if (!strcmp(name, "none"))
1043 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1044 else if (!strcmp(name, "lzo"))
1045 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1046 else if (!strcmp(name, "zlib"))
1047 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1049 ubifs_err("unknown compressor \"%s\"", name);
1054 c->mount_opts.override_compr = 1;
1055 c->default_compr = c->mount_opts.compr_type;
1061 struct super_block *sb = c->vfs_sb;
1063 flag = parse_standard_option(p);
1065 ubifs_err("unrecognized mount option \"%s\" "
1066 "or missing value", p);
1069 sb->s_flags |= flag;
1079 * destroy_journal - destroy journal data structures.
1080 * @c: UBIFS file-system description object
1082 * This function destroys journal data structures including those that may have
1083 * been created by recovery functions.
1085 static void destroy_journal(struct ubifs_info *c)
1087 while (!list_empty(&c->unclean_leb_list)) {
1088 struct ubifs_unclean_leb *ucleb;
1090 ucleb = list_entry(c->unclean_leb_list.next,
1091 struct ubifs_unclean_leb, list);
1092 list_del(&ucleb->list);
1095 while (!list_empty(&c->old_buds)) {
1096 struct ubifs_bud *bud;
1098 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1099 list_del(&bud->list);
1102 ubifs_destroy_idx_gc(c);
1103 ubifs_destroy_size_tree(c);
1109 * bu_init - initialize bulk-read information.
1110 * @c: UBIFS file-system description object
1112 static void bu_init(struct ubifs_info *c)
1114 ubifs_assert(c->bulk_read == 1);
1117 return; /* Already initialized */
1120 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1122 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1123 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1127 /* Just disable bulk-read */
1128 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1129 "disabling it", c->max_bu_buf_len);
1130 c->mount_opts.bulk_read = 1;
1137 * check_free_space - check if there is enough free space to mount.
1138 * @c: UBIFS file-system description object
1140 * This function makes sure UBIFS has enough free space to be mounted in
1141 * read/write mode. UBIFS must always have some free space to allow deletions.
1143 static int check_free_space(struct ubifs_info *c)
1145 ubifs_assert(c->dark_wm > 0);
1146 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1147 ubifs_err("insufficient free space to mount in R/W mode");
1148 dbg_dump_budg(c, &c->bi);
1156 * mount_ubifs - mount UBIFS file-system.
1157 * @c: UBIFS file-system description object
1159 * This function mounts UBIFS file system. Returns zero in case of success and
1160 * a negative error code in case of failure.
1162 * Note, the function does not de-allocate resources it it fails half way
1163 * through, and the caller has to do this instead.
1165 static int mount_ubifs(struct ubifs_info *c)
1171 c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1172 err = init_constants_early(c);
1176 err = ubifs_debugging_init(c);
1180 err = check_volume_empty(c);
1184 if (c->empty && (c->ro_mount || c->ro_media)) {
1186 * This UBI volume is empty, and read-only, or the file system
1187 * is mounted read-only - we cannot format it.
1189 ubifs_err("can't format empty UBI volume: read-only %s",
1190 c->ro_media ? "UBI volume" : "mount");
1195 if (c->ro_media && !c->ro_mount) {
1196 ubifs_err("cannot mount read-write - read-only media");
1202 * The requirement for the buffer is that it should fit indexing B-tree
1203 * height amount of integers. We assume the height if the TNC tree will
1207 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1208 if (!c->bottom_up_buf)
1211 c->sbuf = vmalloc(c->leb_size);
1216 c->ileb_buf = vmalloc(c->leb_size);
1221 if (c->bulk_read == 1)
1225 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
1227 if (!c->write_reserve_buf)
1233 err = ubifs_read_superblock(c);
1238 * Make sure the compressor which is set as default in the superblock
1239 * or overridden by mount options is actually compiled in.
1241 if (!ubifs_compr_present(c->default_compr)) {
1242 ubifs_err("'compressor \"%s\" is not compiled in",
1243 ubifs_compr_name(c->default_compr));
1248 err = init_constants_sb(c);
1252 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1253 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1254 c->cbuf = kmalloc(sz, GFP_NOFS);
1260 err = alloc_wbufs(c);
1264 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1266 /* Create background thread */
1267 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1268 if (IS_ERR(c->bgt)) {
1269 err = PTR_ERR(c->bgt);
1271 ubifs_err("cannot spawn \"%s\", error %d",
1275 wake_up_process(c->bgt);
1278 err = ubifs_read_master(c);
1282 init_constants_master(c);
1284 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1285 ubifs_msg("recovery needed");
1286 c->need_recovery = 1;
1288 err = ubifs_recover_inl_heads(c, c->sbuf);
1292 } else if (!c->ro_mount) {
1294 * Set the "dirty" flag so that if we reboot uncleanly we
1295 * will notice this immediately on the next mount.
1297 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1298 err = ubifs_write_master(c);
1303 err = ubifs_lpt_init(c, 1, !c->ro_mount);
1307 err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1311 err = ubifs_replay_journal(c);
1315 /* Calculate 'min_idx_lebs' after journal replay */
1316 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1318 err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1325 err = check_free_space(c);
1329 /* Check for enough log space */
1330 lnum = c->lhead_lnum + 1;
1331 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1332 lnum = UBIFS_LOG_LNUM;
1333 if (lnum == c->ltail_lnum) {
1334 err = ubifs_consolidate_log(c);
1339 if (c->need_recovery) {
1340 err = ubifs_recover_size(c);
1343 err = ubifs_rcvry_gc_commit(c);
1347 err = take_gc_lnum(c);
1352 * GC LEB may contain garbage if there was an unclean
1353 * reboot, and it should be un-mapped.
1355 err = ubifs_leb_unmap(c, c->gc_lnum);
1360 err = dbg_check_lprops(c);
1363 } else if (c->need_recovery) {
1364 err = ubifs_recover_size(c);
1369 * Even if we mount read-only, we have to set space in GC LEB
1370 * to proper value because this affects UBIFS free space
1371 * reporting. We do not want to have a situation when
1372 * re-mounting from R/O to R/W changes amount of free space.
1374 err = take_gc_lnum(c);
1379 spin_lock(&ubifs_infos_lock);
1380 list_add_tail(&c->infos_list, &ubifs_infos);
1381 spin_unlock(&ubifs_infos_lock);
1383 if (c->need_recovery) {
1385 ubifs_msg("recovery deferred");
1387 c->need_recovery = 0;
1388 ubifs_msg("recovery completed");
1390 * GC LEB has to be empty and taken at this point. But
1391 * the journal head LEBs may also be accounted as
1392 * "empty taken" if they are empty.
1394 ubifs_assert(c->lst.taken_empty_lebs > 0);
1397 ubifs_assert(c->lst.taken_empty_lebs > 0);
1399 if (!c->ro_mount && c->space_fixup) {
1400 err = ubifs_fixup_free_space(c);
1405 err = dbg_check_filesystem(c);
1409 err = dbg_debugfs_init_fs(c);
1415 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1416 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1418 ubifs_msg("mounted read-only");
1419 x = (long long)c->main_lebs * c->leb_size;
1420 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1421 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1422 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1423 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1424 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1425 ubifs_msg("media format: w%d/r%d (latest is w%d/r%d)",
1426 c->fmt_version, c->ro_compat_version,
1427 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1428 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1429 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1430 c->report_rp_size, c->report_rp_size >> 10);
1432 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1433 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1434 dbg_msg("max. write size: %d bytes", c->max_write_size);
1435 dbg_msg("LEB size: %d bytes (%d KiB)",
1436 c->leb_size, c->leb_size >> 10);
1437 dbg_msg("data journal heads: %d",
1438 c->jhead_cnt - NONDATA_JHEADS_CNT);
1439 dbg_msg("UUID: %pUB", c->uuid);
1440 dbg_msg("big_lpt %d", c->big_lpt);
1441 dbg_msg("log LEBs: %d (%d - %d)",
1442 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1443 dbg_msg("LPT area LEBs: %d (%d - %d)",
1444 c->lpt_lebs, c->lpt_first, c->lpt_last);
1445 dbg_msg("orphan area LEBs: %d (%d - %d)",
1446 c->orph_lebs, c->orph_first, c->orph_last);
1447 dbg_msg("main area LEBs: %d (%d - %d)",
1448 c->main_lebs, c->main_first, c->leb_cnt - 1);
1449 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1450 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1451 c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1452 c->bi.old_idx_sz >> 20);
1453 dbg_msg("key hash type: %d", c->key_hash_type);
1454 dbg_msg("tree fanout: %d", c->fanout);
1455 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1456 dbg_msg("first main LEB: %d", c->main_first);
1457 dbg_msg("max. znode size %d", c->max_znode_sz);
1458 dbg_msg("max. index node size %d", c->max_idx_node_sz);
1459 dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
1460 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1461 dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
1462 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1463 dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
1464 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1465 dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu, idx %d",
1466 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1467 UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1468 dbg_msg("dead watermark: %d", c->dead_wm);
1469 dbg_msg("dark watermark: %d", c->dark_wm);
1470 dbg_msg("LEB overhead: %d", c->leb_overhead);
1471 x = (long long)c->main_lebs * c->dark_wm;
1472 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1473 x, x >> 10, x >> 20);
1474 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1475 c->max_bud_bytes, c->max_bud_bytes >> 10,
1476 c->max_bud_bytes >> 20);
1477 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1478 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1479 c->bg_bud_bytes >> 20);
1480 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1481 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1482 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1483 dbg_msg("commit number: %llu", c->cmt_no);
1488 spin_lock(&ubifs_infos_lock);
1489 list_del(&c->infos_list);
1490 spin_unlock(&ubifs_infos_lock);
1496 ubifs_lpt_free(c, 0);
1499 kfree(c->rcvrd_mst_node);
1501 kthread_stop(c->bgt);
1507 kfree(c->write_reserve_buf);
1511 kfree(c->bottom_up_buf);
1512 ubifs_debugging_exit(c);
1517 * ubifs_umount - un-mount UBIFS file-system.
1518 * @c: UBIFS file-system description object
1520 * Note, this function is called to free allocated resourced when un-mounting,
1521 * as well as free resources when an error occurred while we were half way
1522 * through mounting (error path cleanup function). So it has to make sure the
1523 * resource was actually allocated before freeing it.
1525 static void ubifs_umount(struct ubifs_info *c)
1527 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1530 dbg_debugfs_exit_fs(c);
1531 spin_lock(&ubifs_infos_lock);
1532 list_del(&c->infos_list);
1533 spin_unlock(&ubifs_infos_lock);
1536 kthread_stop(c->bgt);
1541 ubifs_lpt_free(c, 0);
1544 kfree(c->rcvrd_mst_node);
1546 kfree(c->write_reserve_buf);
1550 kfree(c->bottom_up_buf);
1551 ubifs_debugging_exit(c);
1555 * ubifs_remount_rw - re-mount in read-write mode.
1556 * @c: UBIFS file-system description object
1558 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1559 * mode. This function allocates the needed resources and re-mounts UBIFS in
1562 static int ubifs_remount_rw(struct ubifs_info *c)
1566 if (c->rw_incompat) {
1567 ubifs_err("the file-system is not R/W-compatible");
1568 ubifs_msg("on-flash format version is w%d/r%d, but software "
1569 "only supports up to version w%d/r%d", c->fmt_version,
1570 c->ro_compat_version, UBIFS_FORMAT_VERSION,
1571 UBIFS_RO_COMPAT_VERSION);
1575 mutex_lock(&c->umount_mutex);
1576 dbg_save_space_info(c);
1577 c->remounting_rw = 1;
1580 err = check_free_space(c);
1584 if (c->old_leb_cnt != c->leb_cnt) {
1585 struct ubifs_sb_node *sup;
1587 sup = ubifs_read_sb_node(c);
1592 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1593 err = ubifs_write_sb_node(c, sup);
1599 if (c->need_recovery) {
1600 ubifs_msg("completing deferred recovery");
1601 err = ubifs_write_rcvrd_mst_node(c);
1604 err = ubifs_recover_size(c);
1607 err = ubifs_clean_lebs(c, c->sbuf);
1610 err = ubifs_recover_inl_heads(c, c->sbuf);
1614 /* A readonly mount is not allowed to have orphans */
1615 ubifs_assert(c->tot_orphans == 0);
1616 err = ubifs_clear_orphans(c);
1621 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1622 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1623 err = ubifs_write_master(c);
1628 c->ileb_buf = vmalloc(c->leb_size);
1634 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
1635 if (!c->write_reserve_buf)
1638 err = ubifs_lpt_init(c, 0, 1);
1642 /* Create background thread */
1643 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1644 if (IS_ERR(c->bgt)) {
1645 err = PTR_ERR(c->bgt);
1647 ubifs_err("cannot spawn \"%s\", error %d",
1651 wake_up_process(c->bgt);
1653 c->orph_buf = vmalloc(c->leb_size);
1659 /* Check for enough log space */
1660 lnum = c->lhead_lnum + 1;
1661 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1662 lnum = UBIFS_LOG_LNUM;
1663 if (lnum == c->ltail_lnum) {
1664 err = ubifs_consolidate_log(c);
1669 if (c->need_recovery)
1670 err = ubifs_rcvry_gc_commit(c);
1672 err = ubifs_leb_unmap(c, c->gc_lnum);
1676 dbg_gen("re-mounted read-write");
1677 c->remounting_rw = 0;
1679 if (c->need_recovery) {
1680 c->need_recovery = 0;
1681 ubifs_msg("deferred recovery completed");
1684 * Do not run the debugging space check if the were doing
1685 * recovery, because when we saved the information we had the
1686 * file-system in a state where the TNC and lprops has been
1687 * modified in memory, but all the I/O operations (including a
1688 * commit) were deferred. So the file-system was in
1689 * "non-committed" state. Now the file-system is in committed
1690 * state, and of course the amount of free space will change
1691 * because, for example, the old index size was imprecise.
1693 err = dbg_check_space_info(c);
1696 if (c->space_fixup) {
1697 err = ubifs_fixup_free_space(c);
1702 mutex_unlock(&c->umount_mutex);
1710 kthread_stop(c->bgt);
1714 kfree(c->write_reserve_buf);
1715 c->write_reserve_buf = NULL;
1718 ubifs_lpt_free(c, 1);
1719 c->remounting_rw = 0;
1720 mutex_unlock(&c->umount_mutex);
1725 * ubifs_remount_ro - re-mount in read-only mode.
1726 * @c: UBIFS file-system description object
1728 * We assume VFS has stopped writing. Possibly the background thread could be
1729 * running a commit, however kthread_stop will wait in that case.
1731 static void ubifs_remount_ro(struct ubifs_info *c)
1735 ubifs_assert(!c->need_recovery);
1736 ubifs_assert(!c->ro_mount);
1738 mutex_lock(&c->umount_mutex);
1740 kthread_stop(c->bgt);
1744 dbg_save_space_info(c);
1746 for (i = 0; i < c->jhead_cnt; i++)
1747 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1749 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1750 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1751 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1752 err = ubifs_write_master(c);
1754 ubifs_ro_mode(c, err);
1758 kfree(c->write_reserve_buf);
1759 c->write_reserve_buf = NULL;
1762 ubifs_lpt_free(c, 1);
1764 err = dbg_check_space_info(c);
1766 ubifs_ro_mode(c, err);
1767 mutex_unlock(&c->umount_mutex);
1770 static void ubifs_put_super(struct super_block *sb)
1773 struct ubifs_info *c = sb->s_fs_info;
1775 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1779 * The following asserts are only valid if there has not been a failure
1780 * of the media. For example, there will be dirty inodes if we failed
1781 * to write them back because of I/O errors.
1784 ubifs_assert(c->bi.idx_growth == 0);
1785 ubifs_assert(c->bi.dd_growth == 0);
1786 ubifs_assert(c->bi.data_growth == 0);
1790 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1791 * and file system un-mount. Namely, it prevents the shrinker from
1792 * picking this superblock for shrinking - it will be just skipped if
1793 * the mutex is locked.
1795 mutex_lock(&c->umount_mutex);
1798 * First of all kill the background thread to make sure it does
1799 * not interfere with un-mounting and freeing resources.
1802 kthread_stop(c->bgt);
1807 * On fatal errors c->ro_error is set to 1, in which case we do
1808 * not write the master node.
1813 /* Synchronize write-buffers */
1814 for (i = 0; i < c->jhead_cnt; i++)
1815 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1818 * We are being cleanly unmounted which means the
1819 * orphans were killed - indicate this in the master
1820 * node. Also save the reserved GC LEB number.
1822 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1823 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1824 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1825 err = ubifs_write_master(c);
1828 * Recovery will attempt to fix the master area
1829 * next mount, so we just print a message and
1830 * continue to unmount normally.
1832 ubifs_err("failed to write master node, "
1835 for (i = 0; i < c->jhead_cnt; i++)
1836 /* Make sure write-buffer timers are canceled */
1837 hrtimer_cancel(&c->jheads[i].wbuf.timer);
1842 bdi_destroy(&c->bdi);
1843 ubi_close_volume(c->ubi);
1844 mutex_unlock(&c->umount_mutex);
1848 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1851 struct ubifs_info *c = sb->s_fs_info;
1853 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1855 err = ubifs_parse_options(c, data, 1);
1857 ubifs_err("invalid or unknown remount parameter");
1861 if (c->ro_mount && !(*flags & MS_RDONLY)) {
1863 ubifs_msg("cannot re-mount R/W due to prior errors");
1867 ubifs_msg("cannot re-mount R/W - UBI volume is R/O");
1870 err = ubifs_remount_rw(c);
1873 } else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1875 ubifs_msg("cannot re-mount R/O due to prior errors");
1878 ubifs_remount_ro(c);
1881 if (c->bulk_read == 1)
1884 dbg_gen("disable bulk-read");
1889 ubifs_assert(c->lst.taken_empty_lebs > 0);
1893 const struct super_operations ubifs_super_operations = {
1894 .alloc_inode = ubifs_alloc_inode,
1895 .destroy_inode = ubifs_destroy_inode,
1896 .put_super = ubifs_put_super,
1897 .write_inode = ubifs_write_inode,
1898 .evict_inode = ubifs_evict_inode,
1899 .statfs = ubifs_statfs,
1900 .dirty_inode = ubifs_dirty_inode,
1901 .remount_fs = ubifs_remount_fs,
1902 .show_options = ubifs_show_options,
1903 .sync_fs = ubifs_sync_fs,
1907 * open_ubi - parse UBI device name string and open the UBI device.
1908 * @name: UBI volume name
1909 * @mode: UBI volume open mode
1911 * The primary method of mounting UBIFS is by specifying the UBI volume
1912 * character device node path. However, UBIFS may also be mounted withoug any
1913 * character device node using one of the following methods:
1915 * o ubiX_Y - mount UBI device number X, volume Y;
1916 * o ubiY - mount UBI device number 0, volume Y;
1917 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1918 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1920 * Alternative '!' separator may be used instead of ':' (because some shells
1921 * like busybox may interpret ':' as an NFS host name separator). This function
1922 * returns UBI volume description object in case of success and a negative
1923 * error code in case of failure.
1925 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1927 struct ubi_volume_desc *ubi;
1931 /* First, try to open using the device node path method */
1932 ubi = ubi_open_volume_path(name, mode);
1936 /* Try the "nodev" method */
1937 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1938 return ERR_PTR(-EINVAL);
1940 /* ubi:NAME method */
1941 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1942 return ubi_open_volume_nm(0, name + 4, mode);
1944 if (!isdigit(name[3]))
1945 return ERR_PTR(-EINVAL);
1947 dev = simple_strtoul(name + 3, &endptr, 0);
1950 if (*endptr == '\0')
1951 return ubi_open_volume(0, dev, mode);
1954 if (*endptr == '_' && isdigit(endptr[1])) {
1955 vol = simple_strtoul(endptr + 1, &endptr, 0);
1956 if (*endptr != '\0')
1957 return ERR_PTR(-EINVAL);
1958 return ubi_open_volume(dev, vol, mode);
1961 /* ubiX:NAME method */
1962 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1963 return ubi_open_volume_nm(dev, ++endptr, mode);
1965 return ERR_PTR(-EINVAL);
1968 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1970 struct ubi_volume_desc *ubi = sb->s_fs_info;
1971 struct ubifs_info *c;
1975 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1979 spin_lock_init(&c->cnt_lock);
1980 spin_lock_init(&c->cs_lock);
1981 spin_lock_init(&c->buds_lock);
1982 spin_lock_init(&c->space_lock);
1983 spin_lock_init(&c->orphan_lock);
1984 init_rwsem(&c->commit_sem);
1985 mutex_init(&c->lp_mutex);
1986 mutex_init(&c->tnc_mutex);
1987 mutex_init(&c->log_mutex);
1988 mutex_init(&c->mst_mutex);
1989 mutex_init(&c->umount_mutex);
1990 mutex_init(&c->bu_mutex);
1991 mutex_init(&c->write_reserve_mutex);
1992 init_waitqueue_head(&c->cmt_wq);
1994 c->old_idx = RB_ROOT;
1995 c->size_tree = RB_ROOT;
1996 c->orph_tree = RB_ROOT;
1997 INIT_LIST_HEAD(&c->infos_list);
1998 INIT_LIST_HEAD(&c->idx_gc);
1999 INIT_LIST_HEAD(&c->replay_list);
2000 INIT_LIST_HEAD(&c->replay_buds);
2001 INIT_LIST_HEAD(&c->uncat_list);
2002 INIT_LIST_HEAD(&c->empty_list);
2003 INIT_LIST_HEAD(&c->freeable_list);
2004 INIT_LIST_HEAD(&c->frdi_idx_list);
2005 INIT_LIST_HEAD(&c->unclean_leb_list);
2006 INIT_LIST_HEAD(&c->old_buds);
2007 INIT_LIST_HEAD(&c->orph_list);
2008 INIT_LIST_HEAD(&c->orph_new);
2009 c->no_chk_data_crc = 1;
2012 c->highest_inum = UBIFS_FIRST_INO;
2013 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
2015 ubi_get_volume_info(ubi, &c->vi);
2016 ubi_get_device_info(c->vi.ubi_num, &c->di);
2018 /* Re-open the UBI device in read-write mode */
2019 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2020 if (IS_ERR(c->ubi)) {
2021 err = PTR_ERR(c->ubi);
2026 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2027 * UBIFS, I/O is not deferred, it is done immediately in readpage,
2028 * which means the user would have to wait not just for their own I/O
2029 * but the read-ahead I/O as well i.e. completely pointless.
2031 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
2033 c->bdi.name = "ubifs",
2034 c->bdi.capabilities = BDI_CAP_MAP_COPY;
2035 err = bdi_init(&c->bdi);
2038 err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
2039 c->vi.ubi_num, c->vi.vol_id);
2043 err = ubifs_parse_options(c, data, 0);
2047 sb->s_bdi = &c->bdi;
2049 sb->s_magic = UBIFS_SUPER_MAGIC;
2050 sb->s_blocksize = UBIFS_BLOCK_SIZE;
2051 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2052 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2053 if (c->max_inode_sz > MAX_LFS_FILESIZE)
2054 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2055 sb->s_op = &ubifs_super_operations;
2057 mutex_lock(&c->umount_mutex);
2058 err = mount_ubifs(c);
2060 ubifs_assert(err < 0);
2064 /* Read the root inode */
2065 root = ubifs_iget(sb, UBIFS_ROOT_INO);
2067 err = PTR_ERR(root);
2071 sb->s_root = d_alloc_root(root);
2075 mutex_unlock(&c->umount_mutex);
2083 mutex_unlock(&c->umount_mutex);
2085 bdi_destroy(&c->bdi);
2087 ubi_close_volume(c->ubi);
2093 static int sb_test(struct super_block *sb, void *data)
2096 struct ubifs_info *c = sb->s_fs_info;
2098 return c->vi.cdev == *dev;
2101 static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2102 const char *name, void *data)
2104 struct ubi_volume_desc *ubi;
2105 struct ubi_volume_info vi;
2106 struct super_block *sb;
2109 dbg_gen("name %s, flags %#x", name, flags);
2112 * Get UBI device number and volume ID. Mount it read-only so far
2113 * because this might be a new mount point, and UBI allows only one
2114 * read-write user at a time.
2116 ubi = open_ubi(name, UBI_READONLY);
2118 dbg_err("cannot open \"%s\", error %d",
2119 name, (int)PTR_ERR(ubi));
2120 return ERR_CAST(ubi);
2122 ubi_get_volume_info(ubi, &vi);
2124 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
2126 sb = sget(fs_type, &sb_test, &set_anon_super, &vi.cdev);
2133 struct ubifs_info *c1 = sb->s_fs_info;
2135 /* A new mount point for already mounted UBIFS */
2136 dbg_gen("this ubi volume is already mounted");
2137 if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2142 sb->s_flags = flags;
2144 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
2147 sb->s_fs_info = ubi;
2148 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2151 /* We do not support atime */
2152 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2155 /* 'fill_super()' opens ubi again so we must close it here */
2156 ubi_close_volume(ubi);
2158 return dget(sb->s_root);
2161 deactivate_locked_super(sb);
2163 ubi_close_volume(ubi);
2164 return ERR_PTR(err);
2167 static struct file_system_type ubifs_fs_type = {
2169 .owner = THIS_MODULE,
2170 .mount = ubifs_mount,
2171 .kill_sb = kill_anon_super,
2175 * Inode slab cache constructor.
2177 static void inode_slab_ctor(void *obj)
2179 struct ubifs_inode *ui = obj;
2180 inode_init_once(&ui->vfs_inode);
2183 static int __init ubifs_init(void)
2187 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2189 /* Make sure node sizes are 8-byte aligned */
2190 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2191 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2192 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2193 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2194 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2195 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2196 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2197 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2198 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2199 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2200 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2202 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2203 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2204 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2205 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2206 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2207 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2209 /* Check min. node size */
2210 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2211 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2212 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2213 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2215 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2216 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2217 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2218 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2220 /* Defined node sizes */
2221 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2222 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2223 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2224 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2227 * We use 2 bit wide bit-fields to store compression type, which should
2228 * be amended if more compressors are added. The bit-fields are:
2229 * @compr_type in 'struct ubifs_inode', @default_compr in
2230 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2232 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2235 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2236 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2238 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2239 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2240 " at least 4096 bytes",
2241 (unsigned int)PAGE_CACHE_SIZE);
2245 err = register_filesystem(&ubifs_fs_type);
2247 ubifs_err("cannot register file system, error %d", err);
2252 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2253 sizeof(struct ubifs_inode), 0,
2254 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2256 if (!ubifs_inode_slab)
2259 register_shrinker(&ubifs_shrinker_info);
2261 err = ubifs_compressors_init();
2265 err = dbg_debugfs_init();
2272 ubifs_compressors_exit();
2274 unregister_shrinker(&ubifs_shrinker_info);
2275 kmem_cache_destroy(ubifs_inode_slab);
2277 unregister_filesystem(&ubifs_fs_type);
2280 /* late_initcall to let compressors initialize first */
2281 late_initcall(ubifs_init);
2283 static void __exit ubifs_exit(void)
2285 ubifs_assert(list_empty(&ubifs_infos));
2286 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2289 ubifs_compressors_exit();
2290 unregister_shrinker(&ubifs_shrinker_info);
2291 kmem_cache_destroy(ubifs_inode_slab);
2292 unregister_filesystem(&ubifs_fs_type);
2294 module_exit(ubifs_exit);
2296 MODULE_LICENSE("GPL");
2297 MODULE_VERSION(__stringify(UBIFS_VERSION));
2298 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2299 MODULE_DESCRIPTION("UBIFS - UBI File System");