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 .scan_objects = ubifs_shrink_scan,
53 .count_objects = ubifs_shrink_count,
54 .seeks = DEFAULT_SEEKS,
58 * validate_inode - validate inode.
59 * @c: UBIFS file-system description object
60 * @inode: the inode to validate
62 * This is a helper function for 'ubifs_iget()' which validates various fields
63 * of a newly built inode to make sure they contain sane values and prevent
64 * possible vulnerabilities. Returns zero if the inode is all right and
65 * a non-zero error code if not.
67 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
70 const struct ubifs_inode *ui = ubifs_inode(inode);
72 if (inode->i_size > c->max_inode_sz) {
73 ubifs_err(c, "inode is too large (%lld)",
74 (long long)inode->i_size);
78 if (ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
79 ubifs_err(c, "unknown compression type %d", ui->compr_type);
83 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
86 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
89 if (ui->xattr && !S_ISREG(inode->i_mode))
92 if (!ubifs_compr_present(ui->compr_type)) {
93 ubifs_warn(c, "inode %lu uses '%s' compression, but it was not compiled in",
94 inode->i_ino, ubifs_compr_name(ui->compr_type));
97 err = dbg_check_dir(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;
132 #ifndef CONFIG_UBIFS_ATIME_SUPPORT
133 inode->i_flags |= S_NOATIME;
135 set_nlink(inode, le32_to_cpu(ino->nlink));
136 i_uid_write(inode, le32_to_cpu(ino->uid));
137 i_gid_write(inode, le32_to_cpu(ino->gid));
138 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
139 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
140 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
141 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
142 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
143 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
144 inode->i_mode = le32_to_cpu(ino->mode);
145 inode->i_size = le64_to_cpu(ino->size);
147 ui->data_len = le32_to_cpu(ino->data_len);
148 ui->flags = le32_to_cpu(ino->flags);
149 ui->compr_type = le16_to_cpu(ino->compr_type);
150 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
151 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
152 ui->xattr_size = le32_to_cpu(ino->xattr_size);
153 ui->xattr_names = le32_to_cpu(ino->xattr_names);
154 ui->synced_i_size = ui->ui_size = inode->i_size;
156 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
158 err = validate_inode(c, inode);
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';
201 inode->i_link = ui->data;
207 union ubifs_dev_desc *dev;
209 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
215 dev = (union ubifs_dev_desc *)ino->data;
216 if (ui->data_len == sizeof(dev->new))
217 rdev = new_decode_dev(le32_to_cpu(dev->new));
218 else if (ui->data_len == sizeof(dev->huge))
219 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
224 memcpy(ui->data, ino->data, ui->data_len);
225 inode->i_op = &ubifs_file_inode_operations;
226 init_special_inode(inode, inode->i_mode, rdev);
231 inode->i_op = &ubifs_file_inode_operations;
232 init_special_inode(inode, inode->i_mode, 0);
233 if (ui->data_len != 0) {
244 ubifs_set_inode_flags(inode);
245 unlock_new_inode(inode);
249 ubifs_err(c, "inode %lu validation failed, error %d", inode->i_ino, err);
250 ubifs_dump_node(c, ino);
251 ubifs_dump_inode(c, inode);
256 ubifs_err(c, "failed to read inode %lu, error %d", inode->i_ino, err);
261 static struct inode *ubifs_alloc_inode(struct super_block *sb)
263 struct ubifs_inode *ui;
265 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
269 memset((void *)ui + sizeof(struct inode), 0,
270 sizeof(struct ubifs_inode) - sizeof(struct inode));
271 mutex_init(&ui->ui_mutex);
272 spin_lock_init(&ui->ui_lock);
273 return &ui->vfs_inode;
276 static void ubifs_i_callback(struct rcu_head *head)
278 struct inode *inode = container_of(head, struct inode, i_rcu);
279 struct ubifs_inode *ui = ubifs_inode(inode);
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 background 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(c, "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_final(&inode->i_data);
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(c, "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;
385 static void ubifs_dirty_inode(struct inode *inode, int flags)
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 dentry *root)
425 struct ubifs_info *c = root->d_sb->s_fs_info;
427 if (c->mount_opts.unmount_mode == 2)
428 seq_puts(s, ",fast_unmount");
429 else if (c->mount_opts.unmount_mode == 1)
430 seq_puts(s, ",norm_unmount");
432 if (c->mount_opts.bulk_read == 2)
433 seq_puts(s, ",bulk_read");
434 else if (c->mount_opts.bulk_read == 1)
435 seq_puts(s, ",no_bulk_read");
437 if (c->mount_opts.chk_data_crc == 2)
438 seq_puts(s, ",chk_data_crc");
439 else if (c->mount_opts.chk_data_crc == 1)
440 seq_puts(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(c, "UBI volume is corrupted - read-only mode");
504 ubifs_msg(c, "read-only UBI device");
508 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
509 ubifs_msg(c, "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(c, "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(c, "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(c, "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(c, "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 ubifs_err(c, "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 ubifs_err(c, "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(c, "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 = kcalloc(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;
814 c->jheads[i].grouped = 1;
818 * Garbage Collector head does not need to be synchronized by timer.
819 * Also GC head nodes are not grouped.
821 c->jheads[GCHD].wbuf.no_timer = 1;
822 c->jheads[GCHD].grouped = 0;
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 ubifs_err(c, "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 ubifs_bud *bud, *n;
879 rbtree_postorder_for_each_entry_safe(bud, n, &c->buds, rb)
884 * check_volume_empty - check if the UBI volume is empty.
885 * @c: UBIFS file-system description object
887 * This function checks if the UBIFS volume is empty by looking if its LEBs are
888 * mapped or not. The result of checking is stored in the @c->empty variable.
889 * Returns zero in case of success and a negative error code in case of
892 static int check_volume_empty(struct ubifs_info *c)
897 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
898 err = ubifs_is_mapped(c, lnum);
899 if (unlikely(err < 0))
913 * UBIFS mount options.
915 * Opt_fast_unmount: do not run a journal commit before un-mounting
916 * Opt_norm_unmount: run a journal commit before un-mounting
917 * Opt_bulk_read: enable bulk-reads
918 * Opt_no_bulk_read: disable bulk-reads
919 * Opt_chk_data_crc: check CRCs when reading data nodes
920 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
921 * Opt_override_compr: override default compressor
922 * Opt_err: just end of array marker
935 static const match_table_t tokens = {
936 {Opt_fast_unmount, "fast_unmount"},
937 {Opt_norm_unmount, "norm_unmount"},
938 {Opt_bulk_read, "bulk_read"},
939 {Opt_no_bulk_read, "no_bulk_read"},
940 {Opt_chk_data_crc, "chk_data_crc"},
941 {Opt_no_chk_data_crc, "no_chk_data_crc"},
942 {Opt_override_compr, "compr=%s"},
947 * parse_standard_option - parse a standard mount option.
948 * @option: the option to parse
950 * Normally, standard mount options like "sync" are passed to file-systems as
951 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
952 * be present in the options string. This function tries to deal with this
953 * situation and parse standard options. Returns 0 if the option was not
954 * recognized, and the corresponding integer flag if it was.
956 * UBIFS is only interested in the "sync" option, so do not check for anything
959 static int parse_standard_option(const char *option)
962 pr_notice("UBIFS: parse %s\n", option);
963 if (!strcmp(option, "sync"))
964 return MS_SYNCHRONOUS;
969 * ubifs_parse_options - parse mount parameters.
970 * @c: UBIFS file-system description object
971 * @options: parameters to parse
972 * @is_remount: non-zero if this is FS re-mount
974 * This function parses UBIFS mount options and returns zero in case success
975 * and a negative error code in case of failure.
977 static int ubifs_parse_options(struct ubifs_info *c, char *options,
981 substring_t args[MAX_OPT_ARGS];
986 while ((p = strsep(&options, ","))) {
992 token = match_token(p, tokens, args);
995 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
996 * We accept them in order to be backward-compatible. But this
997 * should be removed at some point.
999 case Opt_fast_unmount:
1000 c->mount_opts.unmount_mode = 2;
1002 case Opt_norm_unmount:
1003 c->mount_opts.unmount_mode = 1;
1006 c->mount_opts.bulk_read = 2;
1009 case Opt_no_bulk_read:
1010 c->mount_opts.bulk_read = 1;
1013 case Opt_chk_data_crc:
1014 c->mount_opts.chk_data_crc = 2;
1015 c->no_chk_data_crc = 0;
1017 case Opt_no_chk_data_crc:
1018 c->mount_opts.chk_data_crc = 1;
1019 c->no_chk_data_crc = 1;
1021 case Opt_override_compr:
1023 char *name = match_strdup(&args[0]);
1027 if (!strcmp(name, "none"))
1028 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1029 else if (!strcmp(name, "lzo"))
1030 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1031 else if (!strcmp(name, "zlib"))
1032 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1034 ubifs_err(c, "unknown compressor \"%s\"", name); //FIXME: is c ready?
1039 c->mount_opts.override_compr = 1;
1040 c->default_compr = c->mount_opts.compr_type;
1046 struct super_block *sb = c->vfs_sb;
1048 flag = parse_standard_option(p);
1050 ubifs_err(c, "unrecognized mount option \"%s\" or missing value",
1054 sb->s_flags |= flag;
1064 * destroy_journal - destroy journal data structures.
1065 * @c: UBIFS file-system description object
1067 * This function destroys journal data structures including those that may have
1068 * been created by recovery functions.
1070 static void destroy_journal(struct ubifs_info *c)
1072 while (!list_empty(&c->unclean_leb_list)) {
1073 struct ubifs_unclean_leb *ucleb;
1075 ucleb = list_entry(c->unclean_leb_list.next,
1076 struct ubifs_unclean_leb, list);
1077 list_del(&ucleb->list);
1080 while (!list_empty(&c->old_buds)) {
1081 struct ubifs_bud *bud;
1083 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1084 list_del(&bud->list);
1087 ubifs_destroy_idx_gc(c);
1088 ubifs_destroy_size_tree(c);
1094 * bu_init - initialize bulk-read information.
1095 * @c: UBIFS file-system description object
1097 static void bu_init(struct ubifs_info *c)
1099 ubifs_assert(c->bulk_read == 1);
1102 return; /* Already initialized */
1105 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1107 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1108 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1112 /* Just disable bulk-read */
1113 ubifs_warn(c, "cannot allocate %d bytes of memory for bulk-read, disabling it",
1115 c->mount_opts.bulk_read = 1;
1122 * check_free_space - check if there is enough free space to mount.
1123 * @c: UBIFS file-system description object
1125 * This function makes sure UBIFS has enough free space to be mounted in
1126 * read/write mode. UBIFS must always have some free space to allow deletions.
1128 static int check_free_space(struct ubifs_info *c)
1130 ubifs_assert(c->dark_wm > 0);
1131 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1132 ubifs_err(c, "insufficient free space to mount in R/W mode");
1133 ubifs_dump_budg(c, &c->bi);
1134 ubifs_dump_lprops(c);
1141 * mount_ubifs - mount UBIFS file-system.
1142 * @c: UBIFS file-system description object
1144 * This function mounts UBIFS file system. Returns zero in case of success and
1145 * a negative error code in case of failure.
1147 static int mount_ubifs(struct ubifs_info *c)
1153 c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1154 /* Suppress error messages while probing if MS_SILENT is set */
1155 c->probing = !!(c->vfs_sb->s_flags & MS_SILENT);
1157 err = init_constants_early(c);
1161 err = ubifs_debugging_init(c);
1165 err = check_volume_empty(c);
1169 if (c->empty && (c->ro_mount || c->ro_media)) {
1171 * This UBI volume is empty, and read-only, or the file system
1172 * is mounted read-only - we cannot format it.
1174 ubifs_err(c, "can't format empty UBI volume: read-only %s",
1175 c->ro_media ? "UBI volume" : "mount");
1180 if (c->ro_media && !c->ro_mount) {
1181 ubifs_err(c, "cannot mount read-write - read-only media");
1187 * The requirement for the buffer is that it should fit indexing B-tree
1188 * height amount of integers. We assume the height if the TNC tree will
1192 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1193 if (!c->bottom_up_buf)
1196 c->sbuf = vmalloc(c->leb_size);
1201 c->ileb_buf = vmalloc(c->leb_size);
1206 if (c->bulk_read == 1)
1210 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
1212 if (!c->write_reserve_buf)
1218 err = ubifs_read_superblock(c);
1225 * Make sure the compressor which is set as default in the superblock
1226 * or overridden by mount options is actually compiled in.
1228 if (!ubifs_compr_present(c->default_compr)) {
1229 ubifs_err(c, "'compressor \"%s\" is not compiled in",
1230 ubifs_compr_name(c->default_compr));
1235 err = init_constants_sb(c);
1239 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1240 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1241 c->cbuf = kmalloc(sz, GFP_NOFS);
1247 err = alloc_wbufs(c);
1251 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1253 /* Create background thread */
1254 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1255 if (IS_ERR(c->bgt)) {
1256 err = PTR_ERR(c->bgt);
1258 ubifs_err(c, "cannot spawn \"%s\", error %d",
1262 wake_up_process(c->bgt);
1265 err = ubifs_read_master(c);
1269 init_constants_master(c);
1271 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1272 ubifs_msg(c, "recovery needed");
1273 c->need_recovery = 1;
1276 if (c->need_recovery && !c->ro_mount) {
1277 err = ubifs_recover_inl_heads(c, c->sbuf);
1282 err = ubifs_lpt_init(c, 1, !c->ro_mount);
1286 if (!c->ro_mount && c->space_fixup) {
1287 err = ubifs_fixup_free_space(c);
1292 if (!c->ro_mount && !c->need_recovery) {
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 = dbg_check_idx_size(c, c->bi.old_idx_sz);
1307 err = ubifs_replay_journal(c);
1311 /* Calculate 'min_idx_lebs' after journal replay */
1312 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1314 err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1321 err = check_free_space(c);
1325 /* Check for enough log space */
1326 lnum = c->lhead_lnum + 1;
1327 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1328 lnum = UBIFS_LOG_LNUM;
1329 if (lnum == c->ltail_lnum) {
1330 err = ubifs_consolidate_log(c);
1335 if (c->need_recovery) {
1336 err = ubifs_recover_size(c);
1339 err = ubifs_rcvry_gc_commit(c);
1343 err = take_gc_lnum(c);
1348 * GC LEB may contain garbage if there was an unclean
1349 * reboot, and it should be un-mapped.
1351 err = ubifs_leb_unmap(c, c->gc_lnum);
1356 err = dbg_check_lprops(c);
1359 } else if (c->need_recovery) {
1360 err = ubifs_recover_size(c);
1365 * Even if we mount read-only, we have to set space in GC LEB
1366 * to proper value because this affects UBIFS free space
1367 * reporting. We do not want to have a situation when
1368 * re-mounting from R/O to R/W changes amount of free space.
1370 err = take_gc_lnum(c);
1375 spin_lock(&ubifs_infos_lock);
1376 list_add_tail(&c->infos_list, &ubifs_infos);
1377 spin_unlock(&ubifs_infos_lock);
1379 if (c->need_recovery) {
1381 ubifs_msg(c, "recovery deferred");
1383 c->need_recovery = 0;
1384 ubifs_msg(c, "recovery completed");
1386 * GC LEB has to be empty and taken at this point. But
1387 * the journal head LEBs may also be accounted as
1388 * "empty taken" if they are empty.
1390 ubifs_assert(c->lst.taken_empty_lebs > 0);
1393 ubifs_assert(c->lst.taken_empty_lebs > 0);
1395 err = dbg_check_filesystem(c);
1399 err = dbg_debugfs_init_fs(c);
1405 ubifs_msg(c, "UBIFS: mounted UBI device %d, volume %d, name \"%s\"%s",
1406 c->vi.ubi_num, c->vi.vol_id, c->vi.name,
1407 c->ro_mount ? ", R/O mode" : "");
1408 x = (long long)c->main_lebs * c->leb_size;
1409 y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1410 ubifs_msg(c, "LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes",
1411 c->leb_size, c->leb_size >> 10, c->min_io_size,
1413 ubifs_msg(c, "FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)",
1414 x, x >> 20, c->main_lebs,
1415 y, y >> 20, c->log_lebs + c->max_bud_cnt);
1416 ubifs_msg(c, "reserved for root: %llu bytes (%llu KiB)",
1417 c->report_rp_size, c->report_rp_size >> 10);
1418 ubifs_msg(c, "media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s",
1419 c->fmt_version, c->ro_compat_version,
1420 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid,
1421 c->big_lpt ? ", big LPT model" : ", small LPT model");
1423 dbg_gen("default compressor: %s", ubifs_compr_name(c->default_compr));
1424 dbg_gen("data journal heads: %d",
1425 c->jhead_cnt - NONDATA_JHEADS_CNT);
1426 dbg_gen("log LEBs: %d (%d - %d)",
1427 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1428 dbg_gen("LPT area LEBs: %d (%d - %d)",
1429 c->lpt_lebs, c->lpt_first, c->lpt_last);
1430 dbg_gen("orphan area LEBs: %d (%d - %d)",
1431 c->orph_lebs, c->orph_first, c->orph_last);
1432 dbg_gen("main area LEBs: %d (%d - %d)",
1433 c->main_lebs, c->main_first, c->leb_cnt - 1);
1434 dbg_gen("index LEBs: %d", c->lst.idx_lebs);
1435 dbg_gen("total index bytes: %lld (%lld KiB, %lld MiB)",
1436 c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1437 c->bi.old_idx_sz >> 20);
1438 dbg_gen("key hash type: %d", c->key_hash_type);
1439 dbg_gen("tree fanout: %d", c->fanout);
1440 dbg_gen("reserved GC LEB: %d", c->gc_lnum);
1441 dbg_gen("max. znode size %d", c->max_znode_sz);
1442 dbg_gen("max. index node size %d", c->max_idx_node_sz);
1443 dbg_gen("node sizes: data %zu, inode %zu, dentry %zu",
1444 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1445 dbg_gen("node sizes: trun %zu, sb %zu, master %zu",
1446 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1447 dbg_gen("node sizes: ref %zu, cmt. start %zu, orph %zu",
1448 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1449 dbg_gen("max. node sizes: data %zu, inode %zu dentry %zu, idx %d",
1450 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1451 UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1452 dbg_gen("dead watermark: %d", c->dead_wm);
1453 dbg_gen("dark watermark: %d", c->dark_wm);
1454 dbg_gen("LEB overhead: %d", c->leb_overhead);
1455 x = (long long)c->main_lebs * c->dark_wm;
1456 dbg_gen("max. dark space: %lld (%lld KiB, %lld MiB)",
1457 x, x >> 10, x >> 20);
1458 dbg_gen("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1459 c->max_bud_bytes, c->max_bud_bytes >> 10,
1460 c->max_bud_bytes >> 20);
1461 dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1462 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1463 c->bg_bud_bytes >> 20);
1464 dbg_gen("current bud bytes %lld (%lld KiB, %lld MiB)",
1465 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1466 dbg_gen("max. seq. number: %llu", c->max_sqnum);
1467 dbg_gen("commit number: %llu", c->cmt_no);
1472 spin_lock(&ubifs_infos_lock);
1473 list_del(&c->infos_list);
1474 spin_unlock(&ubifs_infos_lock);
1480 ubifs_lpt_free(c, 0);
1483 kfree(c->rcvrd_mst_node);
1485 kthread_stop(c->bgt);
1491 kfree(c->write_reserve_buf);
1495 kfree(c->bottom_up_buf);
1496 ubifs_debugging_exit(c);
1501 * ubifs_umount - un-mount UBIFS file-system.
1502 * @c: UBIFS file-system description object
1504 * Note, this function is called to free allocated resourced when un-mounting,
1505 * as well as free resources when an error occurred while we were half way
1506 * through mounting (error path cleanup function). So it has to make sure the
1507 * resource was actually allocated before freeing it.
1509 static void ubifs_umount(struct ubifs_info *c)
1511 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1514 dbg_debugfs_exit_fs(c);
1515 spin_lock(&ubifs_infos_lock);
1516 list_del(&c->infos_list);
1517 spin_unlock(&ubifs_infos_lock);
1520 kthread_stop(c->bgt);
1525 ubifs_lpt_free(c, 0);
1528 kfree(c->rcvrd_mst_node);
1530 kfree(c->write_reserve_buf);
1534 kfree(c->bottom_up_buf);
1535 ubifs_debugging_exit(c);
1539 * ubifs_remount_rw - re-mount in read-write mode.
1540 * @c: UBIFS file-system description object
1542 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1543 * mode. This function allocates the needed resources and re-mounts UBIFS in
1546 static int ubifs_remount_rw(struct ubifs_info *c)
1550 if (c->rw_incompat) {
1551 ubifs_err(c, "the file-system is not R/W-compatible");
1552 ubifs_msg(c, "on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
1553 c->fmt_version, c->ro_compat_version,
1554 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1558 mutex_lock(&c->umount_mutex);
1559 dbg_save_space_info(c);
1560 c->remounting_rw = 1;
1563 if (c->space_fixup) {
1564 err = ubifs_fixup_free_space(c);
1569 err = check_free_space(c);
1573 if (c->old_leb_cnt != c->leb_cnt) {
1574 struct ubifs_sb_node *sup;
1576 sup = ubifs_read_sb_node(c);
1581 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1582 err = ubifs_write_sb_node(c, sup);
1588 if (c->need_recovery) {
1589 ubifs_msg(c, "completing deferred recovery");
1590 err = ubifs_write_rcvrd_mst_node(c);
1593 err = ubifs_recover_size(c);
1596 err = ubifs_clean_lebs(c, c->sbuf);
1599 err = ubifs_recover_inl_heads(c, c->sbuf);
1603 /* A readonly mount is not allowed to have orphans */
1604 ubifs_assert(c->tot_orphans == 0);
1605 err = ubifs_clear_orphans(c);
1610 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1611 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1612 err = ubifs_write_master(c);
1617 c->ileb_buf = vmalloc(c->leb_size);
1623 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
1624 if (!c->write_reserve_buf) {
1629 err = ubifs_lpt_init(c, 0, 1);
1633 /* Create background thread */
1634 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1635 if (IS_ERR(c->bgt)) {
1636 err = PTR_ERR(c->bgt);
1638 ubifs_err(c, "cannot spawn \"%s\", error %d",
1642 wake_up_process(c->bgt);
1644 c->orph_buf = vmalloc(c->leb_size);
1650 /* Check for enough log space */
1651 lnum = c->lhead_lnum + 1;
1652 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1653 lnum = UBIFS_LOG_LNUM;
1654 if (lnum == c->ltail_lnum) {
1655 err = ubifs_consolidate_log(c);
1660 if (c->need_recovery)
1661 err = ubifs_rcvry_gc_commit(c);
1663 err = ubifs_leb_unmap(c, c->gc_lnum);
1667 dbg_gen("re-mounted read-write");
1668 c->remounting_rw = 0;
1670 if (c->need_recovery) {
1671 c->need_recovery = 0;
1672 ubifs_msg(c, "deferred recovery completed");
1675 * Do not run the debugging space check if the were doing
1676 * recovery, because when we saved the information we had the
1677 * file-system in a state where the TNC and lprops has been
1678 * modified in memory, but all the I/O operations (including a
1679 * commit) were deferred. So the file-system was in
1680 * "non-committed" state. Now the file-system is in committed
1681 * state, and of course the amount of free space will change
1682 * because, for example, the old index size was imprecise.
1684 err = dbg_check_space_info(c);
1687 mutex_unlock(&c->umount_mutex);
1695 kthread_stop(c->bgt);
1699 kfree(c->write_reserve_buf);
1700 c->write_reserve_buf = NULL;
1703 ubifs_lpt_free(c, 1);
1704 c->remounting_rw = 0;
1705 mutex_unlock(&c->umount_mutex);
1710 * ubifs_remount_ro - re-mount in read-only mode.
1711 * @c: UBIFS file-system description object
1713 * We assume VFS has stopped writing. Possibly the background thread could be
1714 * running a commit, however kthread_stop will wait in that case.
1716 static void ubifs_remount_ro(struct ubifs_info *c)
1720 ubifs_assert(!c->need_recovery);
1721 ubifs_assert(!c->ro_mount);
1723 mutex_lock(&c->umount_mutex);
1725 kthread_stop(c->bgt);
1729 dbg_save_space_info(c);
1731 for (i = 0; i < c->jhead_cnt; i++)
1732 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1734 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1735 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1736 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1737 err = ubifs_write_master(c);
1739 ubifs_ro_mode(c, err);
1743 kfree(c->write_reserve_buf);
1744 c->write_reserve_buf = NULL;
1747 ubifs_lpt_free(c, 1);
1749 err = dbg_check_space_info(c);
1751 ubifs_ro_mode(c, err);
1752 mutex_unlock(&c->umount_mutex);
1755 static void ubifs_put_super(struct super_block *sb)
1758 struct ubifs_info *c = sb->s_fs_info;
1760 ubifs_msg(c, "un-mount UBI device %d", c->vi.ubi_num);
1763 * The following asserts are only valid if there has not been a failure
1764 * of the media. For example, there will be dirty inodes if we failed
1765 * to write them back because of I/O errors.
1768 ubifs_assert(c->bi.idx_growth == 0);
1769 ubifs_assert(c->bi.dd_growth == 0);
1770 ubifs_assert(c->bi.data_growth == 0);
1774 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1775 * and file system un-mount. Namely, it prevents the shrinker from
1776 * picking this superblock for shrinking - it will be just skipped if
1777 * the mutex is locked.
1779 mutex_lock(&c->umount_mutex);
1782 * First of all kill the background thread to make sure it does
1783 * not interfere with un-mounting and freeing resources.
1786 kthread_stop(c->bgt);
1791 * On fatal errors c->ro_error is set to 1, in which case we do
1792 * not write the master node.
1797 /* Synchronize write-buffers */
1798 for (i = 0; i < c->jhead_cnt; i++)
1799 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1802 * We are being cleanly unmounted which means the
1803 * orphans were killed - indicate this in the master
1804 * node. Also save the reserved GC LEB number.
1806 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1807 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1808 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1809 err = ubifs_write_master(c);
1812 * Recovery will attempt to fix the master area
1813 * next mount, so we just print a message and
1814 * continue to unmount normally.
1816 ubifs_err(c, "failed to write master node, error %d",
1819 for (i = 0; i < c->jhead_cnt; i++)
1820 /* Make sure write-buffer timers are canceled */
1821 hrtimer_cancel(&c->jheads[i].wbuf.timer);
1826 bdi_destroy(&c->bdi);
1827 ubi_close_volume(c->ubi);
1828 mutex_unlock(&c->umount_mutex);
1831 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1834 struct ubifs_info *c = sb->s_fs_info;
1836 sync_filesystem(sb);
1837 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1839 err = ubifs_parse_options(c, data, 1);
1841 ubifs_err(c, "invalid or unknown remount parameter");
1845 if (c->ro_mount && !(*flags & MS_RDONLY)) {
1847 ubifs_msg(c, "cannot re-mount R/W due to prior errors");
1851 ubifs_msg(c, "cannot re-mount R/W - UBI volume is R/O");
1854 err = ubifs_remount_rw(c);
1857 } else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1859 ubifs_msg(c, "cannot re-mount R/O due to prior errors");
1862 ubifs_remount_ro(c);
1865 if (c->bulk_read == 1)
1868 dbg_gen("disable bulk-read");
1873 ubifs_assert(c->lst.taken_empty_lebs > 0);
1877 const struct super_operations ubifs_super_operations = {
1878 .alloc_inode = ubifs_alloc_inode,
1879 .destroy_inode = ubifs_destroy_inode,
1880 .put_super = ubifs_put_super,
1881 .write_inode = ubifs_write_inode,
1882 .evict_inode = ubifs_evict_inode,
1883 .statfs = ubifs_statfs,
1884 .dirty_inode = ubifs_dirty_inode,
1885 .remount_fs = ubifs_remount_fs,
1886 .show_options = ubifs_show_options,
1887 .sync_fs = ubifs_sync_fs,
1891 * open_ubi - parse UBI device name string and open the UBI device.
1892 * @name: UBI volume name
1893 * @mode: UBI volume open mode
1895 * The primary method of mounting UBIFS is by specifying the UBI volume
1896 * character device node path. However, UBIFS may also be mounted withoug any
1897 * character device node using one of the following methods:
1899 * o ubiX_Y - mount UBI device number X, volume Y;
1900 * o ubiY - mount UBI device number 0, volume Y;
1901 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1902 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1904 * Alternative '!' separator may be used instead of ':' (because some shells
1905 * like busybox may interpret ':' as an NFS host name separator). This function
1906 * returns UBI volume description object in case of success and a negative
1907 * error code in case of failure.
1909 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1911 struct ubi_volume_desc *ubi;
1915 /* First, try to open using the device node path method */
1916 ubi = ubi_open_volume_path(name, mode);
1920 /* Try the "nodev" method */
1921 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1922 return ERR_PTR(-EINVAL);
1924 /* ubi:NAME method */
1925 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1926 return ubi_open_volume_nm(0, name + 4, mode);
1928 if (!isdigit(name[3]))
1929 return ERR_PTR(-EINVAL);
1931 dev = simple_strtoul(name + 3, &endptr, 0);
1934 if (*endptr == '\0')
1935 return ubi_open_volume(0, dev, mode);
1938 if (*endptr == '_' && isdigit(endptr[1])) {
1939 vol = simple_strtoul(endptr + 1, &endptr, 0);
1940 if (*endptr != '\0')
1941 return ERR_PTR(-EINVAL);
1942 return ubi_open_volume(dev, vol, mode);
1945 /* ubiX:NAME method */
1946 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1947 return ubi_open_volume_nm(dev, ++endptr, mode);
1949 return ERR_PTR(-EINVAL);
1952 static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1954 struct ubifs_info *c;
1956 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1958 spin_lock_init(&c->cnt_lock);
1959 spin_lock_init(&c->cs_lock);
1960 spin_lock_init(&c->buds_lock);
1961 spin_lock_init(&c->space_lock);
1962 spin_lock_init(&c->orphan_lock);
1963 init_rwsem(&c->commit_sem);
1964 mutex_init(&c->lp_mutex);
1965 mutex_init(&c->tnc_mutex);
1966 mutex_init(&c->log_mutex);
1967 mutex_init(&c->umount_mutex);
1968 mutex_init(&c->bu_mutex);
1969 mutex_init(&c->write_reserve_mutex);
1970 init_waitqueue_head(&c->cmt_wq);
1972 c->old_idx = RB_ROOT;
1973 c->size_tree = RB_ROOT;
1974 c->orph_tree = RB_ROOT;
1975 INIT_LIST_HEAD(&c->infos_list);
1976 INIT_LIST_HEAD(&c->idx_gc);
1977 INIT_LIST_HEAD(&c->replay_list);
1978 INIT_LIST_HEAD(&c->replay_buds);
1979 INIT_LIST_HEAD(&c->uncat_list);
1980 INIT_LIST_HEAD(&c->empty_list);
1981 INIT_LIST_HEAD(&c->freeable_list);
1982 INIT_LIST_HEAD(&c->frdi_idx_list);
1983 INIT_LIST_HEAD(&c->unclean_leb_list);
1984 INIT_LIST_HEAD(&c->old_buds);
1985 INIT_LIST_HEAD(&c->orph_list);
1986 INIT_LIST_HEAD(&c->orph_new);
1987 c->no_chk_data_crc = 1;
1989 c->highest_inum = UBIFS_FIRST_INO;
1990 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1992 ubi_get_volume_info(ubi, &c->vi);
1993 ubi_get_device_info(c->vi.ubi_num, &c->di);
1998 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
2000 struct ubifs_info *c = sb->s_fs_info;
2005 /* Re-open the UBI device in read-write mode */
2006 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2007 if (IS_ERR(c->ubi)) {
2008 err = PTR_ERR(c->ubi);
2013 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2014 * UBIFS, I/O is not deferred, it is done immediately in readpage,
2015 * which means the user would have to wait not just for their own I/O
2016 * but the read-ahead I/O as well i.e. completely pointless.
2018 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
2020 c->bdi.name = "ubifs",
2021 c->bdi.capabilities = 0;
2022 err = bdi_init(&c->bdi);
2025 err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
2026 c->vi.ubi_num, c->vi.vol_id);
2030 err = ubifs_parse_options(c, data, 0);
2034 sb->s_bdi = &c->bdi;
2036 sb->s_magic = UBIFS_SUPER_MAGIC;
2037 sb->s_blocksize = UBIFS_BLOCK_SIZE;
2038 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2039 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2040 if (c->max_inode_sz > MAX_LFS_FILESIZE)
2041 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2042 sb->s_op = &ubifs_super_operations;
2044 mutex_lock(&c->umount_mutex);
2045 err = mount_ubifs(c);
2047 ubifs_assert(err < 0);
2051 /* Read the root inode */
2052 root = ubifs_iget(sb, UBIFS_ROOT_INO);
2054 err = PTR_ERR(root);
2058 sb->s_root = d_make_root(root);
2064 mutex_unlock(&c->umount_mutex);
2070 mutex_unlock(&c->umount_mutex);
2072 bdi_destroy(&c->bdi);
2074 ubi_close_volume(c->ubi);
2079 static int sb_test(struct super_block *sb, void *data)
2081 struct ubifs_info *c1 = data;
2082 struct ubifs_info *c = sb->s_fs_info;
2084 return c->vi.cdev == c1->vi.cdev;
2087 static int sb_set(struct super_block *sb, void *data)
2089 sb->s_fs_info = data;
2090 return set_anon_super(sb, NULL);
2093 static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2094 const char *name, void *data)
2096 struct ubi_volume_desc *ubi;
2097 struct ubifs_info *c;
2098 struct super_block *sb;
2101 dbg_gen("name %s, flags %#x", name, flags);
2104 * Get UBI device number and volume ID. Mount it read-only so far
2105 * because this might be a new mount point, and UBI allows only one
2106 * read-write user at a time.
2108 ubi = open_ubi(name, UBI_READONLY);
2110 pr_err("UBIFS error (pid: %d): cannot open \"%s\", error %d",
2111 current->pid, name, (int)PTR_ERR(ubi));
2112 return ERR_CAST(ubi);
2115 c = alloc_ubifs_info(ubi);
2121 dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2123 sb = sget(fs_type, sb_test, sb_set, flags, c);
2131 struct ubifs_info *c1 = sb->s_fs_info;
2133 /* A new mount point for already mounted UBIFS */
2134 dbg_gen("this ubi volume is already mounted");
2135 if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2140 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2143 /* We do not support atime */
2144 sb->s_flags |= MS_ACTIVE;
2145 #ifndef CONFIG_UBIFS_ATIME_SUPPORT
2146 sb->s_flags |= MS_NOATIME;
2148 ubifs_msg(c, "full atime support is enabled.");
2152 /* 'fill_super()' opens ubi again so we must close it here */
2153 ubi_close_volume(ubi);
2155 return dget(sb->s_root);
2158 deactivate_locked_super(sb);
2160 ubi_close_volume(ubi);
2161 return ERR_PTR(err);
2164 static void kill_ubifs_super(struct super_block *s)
2166 struct ubifs_info *c = s->s_fs_info;
2171 static struct file_system_type ubifs_fs_type = {
2173 .owner = THIS_MODULE,
2174 .mount = ubifs_mount,
2175 .kill_sb = kill_ubifs_super,
2177 MODULE_ALIAS_FS("ubifs");
2180 * Inode slab cache constructor.
2182 static void inode_slab_ctor(void *obj)
2184 struct ubifs_inode *ui = obj;
2185 inode_init_once(&ui->vfs_inode);
2188 static int __init ubifs_init(void)
2192 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2194 /* Make sure node sizes are 8-byte aligned */
2195 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2196 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2197 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2198 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2199 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2200 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2201 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2202 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2203 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2204 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2205 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2207 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2208 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2209 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2210 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2211 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2212 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2214 /* Check min. node size */
2215 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2216 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2217 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2218 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2220 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2221 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2222 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2223 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2225 /* Defined node sizes */
2226 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2227 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2228 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2229 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2232 * We use 2 bit wide bit-fields to store compression type, which should
2233 * be amended if more compressors are added. The bit-fields are:
2234 * @compr_type in 'struct ubifs_inode', @default_compr in
2235 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2237 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2240 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2241 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2243 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2244 pr_err("UBIFS error (pid %d): VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
2245 current->pid, (unsigned int)PAGE_CACHE_SIZE);
2249 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2250 sizeof(struct ubifs_inode), 0,
2251 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT |
2252 SLAB_ACCOUNT, &inode_slab_ctor);
2253 if (!ubifs_inode_slab)
2256 err = register_shrinker(&ubifs_shrinker_info);
2260 err = ubifs_compressors_init();
2264 err = dbg_debugfs_init();
2268 err = register_filesystem(&ubifs_fs_type);
2270 pr_err("UBIFS error (pid %d): cannot register file system, error %d",
2279 ubifs_compressors_exit();
2281 unregister_shrinker(&ubifs_shrinker_info);
2283 kmem_cache_destroy(ubifs_inode_slab);
2286 /* late_initcall to let compressors initialize first */
2287 late_initcall(ubifs_init);
2289 static void __exit ubifs_exit(void)
2291 ubifs_assert(list_empty(&ubifs_infos));
2292 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2295 ubifs_compressors_exit();
2296 unregister_shrinker(&ubifs_shrinker_info);
2299 * Make sure all delayed rcu free inodes are flushed before we
2303 kmem_cache_destroy(ubifs_inode_slab);
2304 unregister_filesystem(&ubifs_fs_type);
2306 module_exit(ubifs_exit);
2308 MODULE_LICENSE("GPL");
2309 MODULE_VERSION(__stringify(UBIFS_VERSION));
2310 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2311 MODULE_DESCRIPTION("UBIFS - UBI File System");