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 static 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(c, 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(c, 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';
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(c, "inode %lu validation failed, error %d", inode->i_ino, err);
249 ubifs_dump_node(c, ino);
250 ubifs_dump_inode(c, inode);
255 ubifs_err(c, "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);
281 fscrypt_free_inode(inode);
283 kmem_cache_free(ubifs_inode_slab, ui);
286 static void ubifs_destroy_inode(struct 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(c, !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(c, !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;
383 fscrypt_put_encryption_info(inode);
386 static void ubifs_dirty_inode(struct inode *inode, int flags)
388 struct ubifs_info *c = inode->i_sb->s_fs_info;
389 struct ubifs_inode *ui = ubifs_inode(inode);
391 ubifs_assert(c, mutex_is_locked(&ui->ui_mutex));
394 dbg_gen("inode %lu", inode->i_ino);
398 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
400 struct ubifs_info *c = dentry->d_sb->s_fs_info;
401 unsigned long long free;
402 __le32 *uuid = (__le32 *)c->uuid;
404 free = ubifs_get_free_space(c);
405 dbg_gen("free space %lld bytes (%lld blocks)",
406 free, free >> UBIFS_BLOCK_SHIFT);
408 buf->f_type = UBIFS_SUPER_MAGIC;
409 buf->f_bsize = UBIFS_BLOCK_SIZE;
410 buf->f_blocks = c->block_cnt;
411 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
412 if (free > c->report_rp_size)
413 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
418 buf->f_namelen = UBIFS_MAX_NLEN;
419 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
420 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
421 ubifs_assert(c, buf->f_bfree <= c->block_cnt);
425 static int ubifs_show_options(struct seq_file *s, struct dentry *root)
427 struct ubifs_info *c = root->d_sb->s_fs_info;
429 if (c->mount_opts.unmount_mode == 2)
430 seq_puts(s, ",fast_unmount");
431 else if (c->mount_opts.unmount_mode == 1)
432 seq_puts(s, ",norm_unmount");
434 if (c->mount_opts.bulk_read == 2)
435 seq_puts(s, ",bulk_read");
436 else if (c->mount_opts.bulk_read == 1)
437 seq_puts(s, ",no_bulk_read");
439 if (c->mount_opts.chk_data_crc == 2)
440 seq_puts(s, ",chk_data_crc");
441 else if (c->mount_opts.chk_data_crc == 1)
442 seq_puts(s, ",no_chk_data_crc");
444 if (c->mount_opts.override_compr) {
445 seq_printf(s, ",compr=%s",
446 ubifs_compr_name(c, c->mount_opts.compr_type));
449 seq_printf(s, ",assert=%s", ubifs_assert_action_name(c));
450 seq_printf(s, ",ubi=%d,vol=%d", c->vi.ubi_num, c->vi.vol_id);
455 static int ubifs_sync_fs(struct super_block *sb, int wait)
458 struct ubifs_info *c = sb->s_fs_info;
461 * Zero @wait is just an advisory thing to help the file system shove
462 * lots of data into the queues, and there will be the second
463 * '->sync_fs()' call, with non-zero @wait.
469 * Synchronize write buffers, because 'ubifs_run_commit()' does not
470 * do this if it waits for an already running commit.
472 for (i = 0; i < c->jhead_cnt; i++) {
473 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
479 * Strictly speaking, it is not necessary to commit the journal here,
480 * synchronizing write-buffers would be enough. But committing makes
481 * UBIFS free space predictions much more accurate, so we want to let
482 * the user be able to get more accurate results of 'statfs()' after
483 * they synchronize the file system.
485 err = ubifs_run_commit(c);
489 return ubi_sync(c->vi.ubi_num);
493 * init_constants_early - initialize UBIFS constants.
494 * @c: UBIFS file-system description object
496 * This function initialize UBIFS constants which do not need the superblock to
497 * be read. It also checks that the UBI volume satisfies basic UBIFS
498 * requirements. Returns zero in case of success and a negative error code in
501 static int init_constants_early(struct ubifs_info *c)
503 if (c->vi.corrupted) {
504 ubifs_warn(c, "UBI volume is corrupted - read-only mode");
509 ubifs_msg(c, "read-only UBI device");
513 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
514 ubifs_msg(c, "static UBI volume - read-only mode");
518 c->leb_cnt = c->vi.size;
519 c->leb_size = c->vi.usable_leb_size;
520 c->leb_start = c->di.leb_start;
521 c->half_leb_size = c->leb_size / 2;
522 c->min_io_size = c->di.min_io_size;
523 c->min_io_shift = fls(c->min_io_size) - 1;
524 c->max_write_size = c->di.max_write_size;
525 c->max_write_shift = fls(c->max_write_size) - 1;
527 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
528 ubifs_errc(c, "too small LEBs (%d bytes), min. is %d bytes",
529 c->leb_size, UBIFS_MIN_LEB_SZ);
533 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
534 ubifs_errc(c, "too few LEBs (%d), min. is %d",
535 c->leb_cnt, UBIFS_MIN_LEB_CNT);
539 if (!is_power_of_2(c->min_io_size)) {
540 ubifs_errc(c, "bad min. I/O size %d", c->min_io_size);
545 * Maximum write size has to be greater or equivalent to min. I/O
546 * size, and be multiple of min. I/O size.
548 if (c->max_write_size < c->min_io_size ||
549 c->max_write_size % c->min_io_size ||
550 !is_power_of_2(c->max_write_size)) {
551 ubifs_errc(c, "bad write buffer size %d for %d min. I/O unit",
552 c->max_write_size, c->min_io_size);
557 * UBIFS aligns all node to 8-byte boundary, so to make function in
558 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
561 if (c->min_io_size < 8) {
564 if (c->max_write_size < c->min_io_size) {
565 c->max_write_size = c->min_io_size;
566 c->max_write_shift = c->min_io_shift;
570 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
571 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
574 * Initialize node length ranges which are mostly needed for node
577 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
578 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
579 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
580 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
581 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
582 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
583 c->ranges[UBIFS_AUTH_NODE].min_len = UBIFS_AUTH_NODE_SZ;
584 c->ranges[UBIFS_AUTH_NODE].max_len = UBIFS_AUTH_NODE_SZ +
587 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
588 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
589 c->ranges[UBIFS_ORPH_NODE].min_len =
590 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
591 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
592 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
593 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
594 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
595 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
596 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
597 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
599 * Minimum indexing node size is amended later when superblock is
600 * read and the key length is known.
602 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
604 * Maximum indexing node size is amended later when superblock is
605 * read and the fanout is known.
607 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
610 * Initialize dead and dark LEB space watermarks. See gc.c for comments
611 * about these values.
613 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
614 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
617 * Calculate how many bytes would be wasted at the end of LEB if it was
618 * fully filled with data nodes of maximum size. This is used in
619 * calculations when reporting free space.
621 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
623 /* Buffer size for bulk-reads */
624 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
625 if (c->max_bu_buf_len > c->leb_size)
626 c->max_bu_buf_len = c->leb_size;
631 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
632 * @c: UBIFS file-system description object
633 * @lnum: LEB the write-buffer was synchronized to
634 * @free: how many free bytes left in this LEB
635 * @pad: how many bytes were padded
637 * This is a callback function which is called by the I/O unit when the
638 * write-buffer is synchronized. We need this to correctly maintain space
639 * accounting in bud logical eraseblocks. This function returns zero in case of
640 * success and a negative error code in case of failure.
642 * This function actually belongs to the journal, but we keep it here because
643 * we want to keep it static.
645 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
647 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
651 * init_constants_sb - initialize UBIFS constants.
652 * @c: UBIFS file-system description object
654 * This is a helper function which initializes various UBIFS constants after
655 * the superblock has been read. It also checks various UBIFS parameters and
656 * makes sure they are all right. Returns zero in case of success and a
657 * negative error code in case of failure.
659 static int init_constants_sb(struct ubifs_info *c)
664 c->main_bytes = (long long)c->main_lebs * c->leb_size;
665 c->max_znode_sz = sizeof(struct ubifs_znode) +
666 c->fanout * sizeof(struct ubifs_zbranch);
668 tmp = ubifs_idx_node_sz(c, 1);
669 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
670 c->min_idx_node_sz = ALIGN(tmp, 8);
672 tmp = ubifs_idx_node_sz(c, c->fanout);
673 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
674 c->max_idx_node_sz = ALIGN(tmp, 8);
676 /* Make sure LEB size is large enough to fit full commit */
677 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
678 tmp = ALIGN(tmp, c->min_io_size);
679 if (tmp > c->leb_size) {
680 ubifs_err(c, "too small LEB size %d, at least %d needed",
686 * Make sure that the log is large enough to fit reference nodes for
687 * all buds plus one reserved LEB.
689 tmp64 = c->max_bud_bytes + c->leb_size - 1;
690 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
691 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
694 if (c->log_lebs < tmp) {
695 ubifs_err(c, "too small log %d LEBs, required min. %d LEBs",
701 * When budgeting we assume worst-case scenarios when the pages are not
702 * be compressed and direntries are of the maximum size.
704 * Note, data, which may be stored in inodes is budgeted separately, so
705 * it is not included into 'c->bi.inode_budget'.
707 c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
708 c->bi.inode_budget = UBIFS_INO_NODE_SZ;
709 c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
712 * When the amount of flash space used by buds becomes
713 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
714 * The writers are unblocked when the commit is finished. To avoid
715 * writers to be blocked UBIFS initiates background commit in advance,
716 * when number of bud bytes becomes above the limit defined below.
718 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
721 * Ensure minimum journal size. All the bytes in the journal heads are
722 * considered to be used, when calculating the current journal usage.
723 * Consequently, if the journal is too small, UBIFS will treat it as
726 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
727 if (c->bg_bud_bytes < tmp64)
728 c->bg_bud_bytes = tmp64;
729 if (c->max_bud_bytes < tmp64 + c->leb_size)
730 c->max_bud_bytes = tmp64 + c->leb_size;
732 err = ubifs_calc_lpt_geom(c);
736 /* Initialize effective LEB size used in budgeting calculations */
737 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
742 * init_constants_master - initialize UBIFS constants.
743 * @c: UBIFS file-system description object
745 * This is a helper function which initializes various UBIFS constants after
746 * the master node has been read. It also checks various UBIFS parameters and
747 * makes sure they are all right.
749 static void init_constants_master(struct ubifs_info *c)
753 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
754 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
757 * Calculate total amount of FS blocks. This number is not used
758 * internally because it does not make much sense for UBIFS, but it is
759 * necessary to report something for the 'statfs()' call.
761 * Subtract the LEB reserved for GC, the LEB which is reserved for
762 * deletions, minimum LEBs for the index, and assume only one journal
765 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
766 tmp64 *= (long long)c->leb_size - c->leb_overhead;
767 tmp64 = ubifs_reported_space(c, tmp64);
768 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
772 * take_gc_lnum - reserve GC LEB.
773 * @c: UBIFS file-system description object
775 * This function ensures that the LEB reserved for garbage collection is marked
776 * as "taken" in lprops. We also have to set free space to LEB size and dirty
777 * space to zero, because lprops may contain out-of-date information if the
778 * file-system was un-mounted before it has been committed. This function
779 * returns zero in case of success and a negative error code in case of
782 static int take_gc_lnum(struct ubifs_info *c)
786 if (c->gc_lnum == -1) {
787 ubifs_err(c, "no LEB for GC");
791 /* And we have to tell lprops that this LEB is taken */
792 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
798 * alloc_wbufs - allocate write-buffers.
799 * @c: UBIFS file-system description object
801 * This helper function allocates and initializes UBIFS write-buffers. Returns
802 * zero in case of success and %-ENOMEM in case of failure.
804 static int alloc_wbufs(struct ubifs_info *c)
808 c->jheads = kcalloc(c->jhead_cnt, sizeof(struct ubifs_jhead),
813 /* Initialize journal heads */
814 for (i = 0; i < c->jhead_cnt; i++) {
815 INIT_LIST_HEAD(&c->jheads[i].buds_list);
816 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
820 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
821 c->jheads[i].wbuf.jhead = i;
822 c->jheads[i].grouped = 1;
823 c->jheads[i].log_hash = ubifs_hash_get_desc(c);
824 if (IS_ERR(c->jheads[i].log_hash))
829 * Garbage Collector head does not need to be synchronized by timer.
830 * Also GC head nodes are not grouped.
832 c->jheads[GCHD].wbuf.no_timer = 1;
833 c->jheads[GCHD].grouped = 0;
839 kfree(c->jheads[i].log_hash);
845 * free_wbufs - free write-buffers.
846 * @c: UBIFS file-system description object
848 static void free_wbufs(struct ubifs_info *c)
853 for (i = 0; i < c->jhead_cnt; i++) {
854 kfree(c->jheads[i].wbuf.buf);
855 kfree(c->jheads[i].wbuf.inodes);
856 kfree(c->jheads[i].log_hash);
864 * free_orphans - free orphans.
865 * @c: UBIFS file-system description object
867 static void free_orphans(struct ubifs_info *c)
869 struct ubifs_orphan *orph;
871 while (c->orph_dnext) {
872 orph = c->orph_dnext;
873 c->orph_dnext = orph->dnext;
874 list_del(&orph->list);
878 while (!list_empty(&c->orph_list)) {
879 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
880 list_del(&orph->list);
882 ubifs_err(c, "orphan list not empty at unmount");
890 * free_buds - free per-bud objects.
891 * @c: UBIFS file-system description object
893 static void free_buds(struct ubifs_info *c)
895 struct ubifs_bud *bud, *n;
897 rbtree_postorder_for_each_entry_safe(bud, n, &c->buds, rb)
902 * check_volume_empty - check if the UBI volume is empty.
903 * @c: UBIFS file-system description object
905 * This function checks if the UBIFS volume is empty by looking if its LEBs are
906 * mapped or not. The result of checking is stored in the @c->empty variable.
907 * Returns zero in case of success and a negative error code in case of
910 static int check_volume_empty(struct ubifs_info *c)
915 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
916 err = ubifs_is_mapped(c, lnum);
917 if (unlikely(err < 0))
931 * UBIFS mount options.
933 * Opt_fast_unmount: do not run a journal commit before un-mounting
934 * Opt_norm_unmount: run a journal commit before un-mounting
935 * Opt_bulk_read: enable bulk-reads
936 * Opt_no_bulk_read: disable bulk-reads
937 * Opt_chk_data_crc: check CRCs when reading data nodes
938 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
939 * Opt_override_compr: override default compressor
940 * Opt_assert: set ubifs_assert() action
941 * Opt_auth_key: The key name used for authentication
942 * Opt_auth_hash_name: The hash type used for authentication
943 * Opt_err: just end of array marker
960 static const match_table_t tokens = {
961 {Opt_fast_unmount, "fast_unmount"},
962 {Opt_norm_unmount, "norm_unmount"},
963 {Opt_bulk_read, "bulk_read"},
964 {Opt_no_bulk_read, "no_bulk_read"},
965 {Opt_chk_data_crc, "chk_data_crc"},
966 {Opt_no_chk_data_crc, "no_chk_data_crc"},
967 {Opt_override_compr, "compr=%s"},
968 {Opt_auth_key, "auth_key=%s"},
969 {Opt_auth_hash_name, "auth_hash_name=%s"},
970 {Opt_ignore, "ubi=%s"},
971 {Opt_ignore, "vol=%s"},
972 {Opt_assert, "assert=%s"},
977 * parse_standard_option - parse a standard mount option.
978 * @option: the option to parse
980 * Normally, standard mount options like "sync" are passed to file-systems as
981 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
982 * be present in the options string. This function tries to deal with this
983 * situation and parse standard options. Returns 0 if the option was not
984 * recognized, and the corresponding integer flag if it was.
986 * UBIFS is only interested in the "sync" option, so do not check for anything
989 static int parse_standard_option(const char *option)
992 pr_notice("UBIFS: parse %s\n", option);
993 if (!strcmp(option, "sync"))
994 return SB_SYNCHRONOUS;
999 * ubifs_parse_options - parse mount parameters.
1000 * @c: UBIFS file-system description object
1001 * @options: parameters to parse
1002 * @is_remount: non-zero if this is FS re-mount
1004 * This function parses UBIFS mount options and returns zero in case success
1005 * and a negative error code in case of failure.
1007 static int ubifs_parse_options(struct ubifs_info *c, char *options,
1011 substring_t args[MAX_OPT_ARGS];
1016 while ((p = strsep(&options, ","))) {
1022 token = match_token(p, tokens, args);
1025 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
1026 * We accept them in order to be backward-compatible. But this
1027 * should be removed at some point.
1029 case Opt_fast_unmount:
1030 c->mount_opts.unmount_mode = 2;
1032 case Opt_norm_unmount:
1033 c->mount_opts.unmount_mode = 1;
1036 c->mount_opts.bulk_read = 2;
1039 case Opt_no_bulk_read:
1040 c->mount_opts.bulk_read = 1;
1043 case Opt_chk_data_crc:
1044 c->mount_opts.chk_data_crc = 2;
1045 c->no_chk_data_crc = 0;
1047 case Opt_no_chk_data_crc:
1048 c->mount_opts.chk_data_crc = 1;
1049 c->no_chk_data_crc = 1;
1051 case Opt_override_compr:
1053 char *name = match_strdup(&args[0]);
1057 if (!strcmp(name, "none"))
1058 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1059 else if (!strcmp(name, "lzo"))
1060 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1061 else if (!strcmp(name, "zlib"))
1062 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1064 ubifs_err(c, "unknown compressor \"%s\"", name); //FIXME: is c ready?
1069 c->mount_opts.override_compr = 1;
1070 c->default_compr = c->mount_opts.compr_type;
1075 char *act = match_strdup(&args[0]);
1079 if (!strcmp(act, "report"))
1080 c->assert_action = ASSACT_REPORT;
1081 else if (!strcmp(act, "read-only"))
1082 c->assert_action = ASSACT_RO;
1083 else if (!strcmp(act, "panic"))
1084 c->assert_action = ASSACT_PANIC;
1086 ubifs_err(c, "unknown assert action \"%s\"", act);
1094 c->auth_key_name = kstrdup(args[0].from, GFP_KERNEL);
1095 if (!c->auth_key_name)
1098 case Opt_auth_hash_name:
1099 c->auth_hash_name = kstrdup(args[0].from, GFP_KERNEL);
1100 if (!c->auth_hash_name)
1108 struct super_block *sb = c->vfs_sb;
1110 flag = parse_standard_option(p);
1112 ubifs_err(c, "unrecognized mount option \"%s\" or missing value",
1116 sb->s_flags |= flag;
1126 * destroy_journal - destroy journal data structures.
1127 * @c: UBIFS file-system description object
1129 * This function destroys journal data structures including those that may have
1130 * been created by recovery functions.
1132 static void destroy_journal(struct ubifs_info *c)
1134 while (!list_empty(&c->unclean_leb_list)) {
1135 struct ubifs_unclean_leb *ucleb;
1137 ucleb = list_entry(c->unclean_leb_list.next,
1138 struct ubifs_unclean_leb, list);
1139 list_del(&ucleb->list);
1142 while (!list_empty(&c->old_buds)) {
1143 struct ubifs_bud *bud;
1145 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1146 list_del(&bud->list);
1149 ubifs_destroy_idx_gc(c);
1150 ubifs_destroy_size_tree(c);
1156 * bu_init - initialize bulk-read information.
1157 * @c: UBIFS file-system description object
1159 static void bu_init(struct ubifs_info *c)
1161 ubifs_assert(c, c->bulk_read == 1);
1164 return; /* Already initialized */
1167 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1169 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1170 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1174 /* Just disable bulk-read */
1175 ubifs_warn(c, "cannot allocate %d bytes of memory for bulk-read, disabling it",
1177 c->mount_opts.bulk_read = 1;
1184 * check_free_space - check if there is enough free space to mount.
1185 * @c: UBIFS file-system description object
1187 * This function makes sure UBIFS has enough free space to be mounted in
1188 * read/write mode. UBIFS must always have some free space to allow deletions.
1190 static int check_free_space(struct ubifs_info *c)
1192 ubifs_assert(c, c->dark_wm > 0);
1193 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1194 ubifs_err(c, "insufficient free space to mount in R/W mode");
1195 ubifs_dump_budg(c, &c->bi);
1196 ubifs_dump_lprops(c);
1203 * mount_ubifs - mount UBIFS file-system.
1204 * @c: UBIFS file-system description object
1206 * This function mounts UBIFS file system. Returns zero in case of success and
1207 * a negative error code in case of failure.
1209 static int mount_ubifs(struct ubifs_info *c)
1215 c->ro_mount = !!sb_rdonly(c->vfs_sb);
1216 /* Suppress error messages while probing if SB_SILENT is set */
1217 c->probing = !!(c->vfs_sb->s_flags & SB_SILENT);
1219 err = init_constants_early(c);
1223 err = ubifs_debugging_init(c);
1227 err = check_volume_empty(c);
1231 if (c->empty && (c->ro_mount || c->ro_media)) {
1233 * This UBI volume is empty, and read-only, or the file system
1234 * is mounted read-only - we cannot format it.
1236 ubifs_err(c, "can't format empty UBI volume: read-only %s",
1237 c->ro_media ? "UBI volume" : "mount");
1242 if (c->ro_media && !c->ro_mount) {
1243 ubifs_err(c, "cannot mount read-write - read-only media");
1249 * The requirement for the buffer is that it should fit indexing B-tree
1250 * height amount of integers. We assume the height if the TNC tree will
1254 c->bottom_up_buf = kmalloc_array(BOTTOM_UP_HEIGHT, sizeof(int),
1256 if (!c->bottom_up_buf)
1259 c->sbuf = vmalloc(c->leb_size);
1264 c->ileb_buf = vmalloc(c->leb_size);
1269 if (c->bulk_read == 1)
1273 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ + \
1274 UBIFS_CIPHER_BLOCK_SIZE,
1276 if (!c->write_reserve_buf)
1282 if (c->auth_key_name) {
1283 if (IS_ENABLED(CONFIG_UBIFS_FS_AUTHENTICATION)) {
1284 err = ubifs_init_authentication(c);
1288 ubifs_err(c, "auth_key_name, but UBIFS is built without"
1289 " authentication support");
1295 err = ubifs_read_superblock(c);
1302 * Make sure the compressor which is set as default in the superblock
1303 * or overridden by mount options is actually compiled in.
1305 if (!ubifs_compr_present(c, c->default_compr)) {
1306 ubifs_err(c, "'compressor \"%s\" is not compiled in",
1307 ubifs_compr_name(c, c->default_compr));
1312 err = init_constants_sb(c);
1316 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1317 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1318 c->cbuf = kmalloc(sz, GFP_NOFS);
1324 err = alloc_wbufs(c);
1328 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1330 /* Create background thread */
1331 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1332 if (IS_ERR(c->bgt)) {
1333 err = PTR_ERR(c->bgt);
1335 ubifs_err(c, "cannot spawn \"%s\", error %d",
1339 wake_up_process(c->bgt);
1342 err = ubifs_read_master(c);
1346 init_constants_master(c);
1348 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1349 ubifs_msg(c, "recovery needed");
1350 c->need_recovery = 1;
1353 if (c->need_recovery && !c->ro_mount) {
1354 err = ubifs_recover_inl_heads(c, c->sbuf);
1359 err = ubifs_lpt_init(c, 1, !c->ro_mount);
1363 if (!c->ro_mount && c->space_fixup) {
1364 err = ubifs_fixup_free_space(c);
1369 if (!c->ro_mount && !c->need_recovery) {
1371 * Set the "dirty" flag so that if we reboot uncleanly we
1372 * will notice this immediately on the next mount.
1374 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1375 err = ubifs_write_master(c);
1380 err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1384 err = ubifs_replay_journal(c);
1388 /* Calculate 'min_idx_lebs' after journal replay */
1389 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1391 err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1398 err = check_free_space(c);
1402 /* Check for enough log space */
1403 lnum = c->lhead_lnum + 1;
1404 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1405 lnum = UBIFS_LOG_LNUM;
1406 if (lnum == c->ltail_lnum) {
1407 err = ubifs_consolidate_log(c);
1412 if (c->need_recovery) {
1413 if (!ubifs_authenticated(c)) {
1414 err = ubifs_recover_size(c, true);
1419 err = ubifs_rcvry_gc_commit(c);
1423 if (ubifs_authenticated(c)) {
1424 err = ubifs_recover_size(c, false);
1429 err = take_gc_lnum(c);
1434 * GC LEB may contain garbage if there was an unclean
1435 * reboot, and it should be un-mapped.
1437 err = ubifs_leb_unmap(c, c->gc_lnum);
1442 err = dbg_check_lprops(c);
1445 } else if (c->need_recovery) {
1446 err = ubifs_recover_size(c, false);
1451 * Even if we mount read-only, we have to set space in GC LEB
1452 * to proper value because this affects UBIFS free space
1453 * reporting. We do not want to have a situation when
1454 * re-mounting from R/O to R/W changes amount of free space.
1456 err = take_gc_lnum(c);
1461 spin_lock(&ubifs_infos_lock);
1462 list_add_tail(&c->infos_list, &ubifs_infos);
1463 spin_unlock(&ubifs_infos_lock);
1465 if (c->need_recovery) {
1467 ubifs_msg(c, "recovery deferred");
1469 c->need_recovery = 0;
1470 ubifs_msg(c, "recovery completed");
1472 * GC LEB has to be empty and taken at this point. But
1473 * the journal head LEBs may also be accounted as
1474 * "empty taken" if they are empty.
1476 ubifs_assert(c, c->lst.taken_empty_lebs > 0);
1479 ubifs_assert(c, c->lst.taken_empty_lebs > 0);
1481 err = dbg_check_filesystem(c);
1485 err = dbg_debugfs_init_fs(c);
1491 ubifs_msg(c, "UBIFS: mounted UBI device %d, volume %d, name \"%s\"%s",
1492 c->vi.ubi_num, c->vi.vol_id, c->vi.name,
1493 c->ro_mount ? ", R/O mode" : "");
1494 x = (long long)c->main_lebs * c->leb_size;
1495 y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1496 ubifs_msg(c, "LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes",
1497 c->leb_size, c->leb_size >> 10, c->min_io_size,
1499 ubifs_msg(c, "FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)",
1500 x, x >> 20, c->main_lebs,
1501 y, y >> 20, c->log_lebs + c->max_bud_cnt);
1502 ubifs_msg(c, "reserved for root: %llu bytes (%llu KiB)",
1503 c->report_rp_size, c->report_rp_size >> 10);
1504 ubifs_msg(c, "media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s",
1505 c->fmt_version, c->ro_compat_version,
1506 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid,
1507 c->big_lpt ? ", big LPT model" : ", small LPT model");
1509 dbg_gen("default compressor: %s", ubifs_compr_name(c, c->default_compr));
1510 dbg_gen("data journal heads: %d",
1511 c->jhead_cnt - NONDATA_JHEADS_CNT);
1512 dbg_gen("log LEBs: %d (%d - %d)",
1513 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1514 dbg_gen("LPT area LEBs: %d (%d - %d)",
1515 c->lpt_lebs, c->lpt_first, c->lpt_last);
1516 dbg_gen("orphan area LEBs: %d (%d - %d)",
1517 c->orph_lebs, c->orph_first, c->orph_last);
1518 dbg_gen("main area LEBs: %d (%d - %d)",
1519 c->main_lebs, c->main_first, c->leb_cnt - 1);
1520 dbg_gen("index LEBs: %d", c->lst.idx_lebs);
1521 dbg_gen("total index bytes: %lld (%lld KiB, %lld MiB)",
1522 c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1523 c->bi.old_idx_sz >> 20);
1524 dbg_gen("key hash type: %d", c->key_hash_type);
1525 dbg_gen("tree fanout: %d", c->fanout);
1526 dbg_gen("reserved GC LEB: %d", c->gc_lnum);
1527 dbg_gen("max. znode size %d", c->max_znode_sz);
1528 dbg_gen("max. index node size %d", c->max_idx_node_sz);
1529 dbg_gen("node sizes: data %zu, inode %zu, dentry %zu",
1530 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1531 dbg_gen("node sizes: trun %zu, sb %zu, master %zu",
1532 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1533 dbg_gen("node sizes: ref %zu, cmt. start %zu, orph %zu",
1534 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1535 dbg_gen("max. node sizes: data %zu, inode %zu dentry %zu, idx %d",
1536 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1537 UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1538 dbg_gen("dead watermark: %d", c->dead_wm);
1539 dbg_gen("dark watermark: %d", c->dark_wm);
1540 dbg_gen("LEB overhead: %d", c->leb_overhead);
1541 x = (long long)c->main_lebs * c->dark_wm;
1542 dbg_gen("max. dark space: %lld (%lld KiB, %lld MiB)",
1543 x, x >> 10, x >> 20);
1544 dbg_gen("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1545 c->max_bud_bytes, c->max_bud_bytes >> 10,
1546 c->max_bud_bytes >> 20);
1547 dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1548 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1549 c->bg_bud_bytes >> 20);
1550 dbg_gen("current bud bytes %lld (%lld KiB, %lld MiB)",
1551 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1552 dbg_gen("max. seq. number: %llu", c->max_sqnum);
1553 dbg_gen("commit number: %llu", c->cmt_no);
1558 spin_lock(&ubifs_infos_lock);
1559 list_del(&c->infos_list);
1560 spin_unlock(&ubifs_infos_lock);
1566 ubifs_lpt_free(c, 0);
1569 kfree(c->rcvrd_mst_node);
1571 kthread_stop(c->bgt);
1577 kfree(c->write_reserve_buf);
1581 kfree(c->bottom_up_buf);
1582 ubifs_debugging_exit(c);
1587 * ubifs_umount - un-mount UBIFS file-system.
1588 * @c: UBIFS file-system description object
1590 * Note, this function is called to free allocated resourced when un-mounting,
1591 * as well as free resources when an error occurred while we were half way
1592 * through mounting (error path cleanup function). So it has to make sure the
1593 * resource was actually allocated before freeing it.
1595 static void ubifs_umount(struct ubifs_info *c)
1597 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1600 dbg_debugfs_exit_fs(c);
1601 spin_lock(&ubifs_infos_lock);
1602 list_del(&c->infos_list);
1603 spin_unlock(&ubifs_infos_lock);
1606 kthread_stop(c->bgt);
1611 ubifs_lpt_free(c, 0);
1612 ubifs_exit_authentication(c);
1614 kfree(c->auth_key_name);
1615 kfree(c->auth_hash_name);
1617 kfree(c->rcvrd_mst_node);
1619 kfree(c->write_reserve_buf);
1623 kfree(c->bottom_up_buf);
1624 ubifs_debugging_exit(c);
1628 * ubifs_remount_rw - re-mount in read-write mode.
1629 * @c: UBIFS file-system description object
1631 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1632 * mode. This function allocates the needed resources and re-mounts UBIFS in
1635 static int ubifs_remount_rw(struct ubifs_info *c)
1639 if (c->rw_incompat) {
1640 ubifs_err(c, "the file-system is not R/W-compatible");
1641 ubifs_msg(c, "on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
1642 c->fmt_version, c->ro_compat_version,
1643 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1647 mutex_lock(&c->umount_mutex);
1648 dbg_save_space_info(c);
1649 c->remounting_rw = 1;
1652 if (c->space_fixup) {
1653 err = ubifs_fixup_free_space(c);
1658 err = check_free_space(c);
1662 if (c->old_leb_cnt != c->leb_cnt) {
1663 struct ubifs_sb_node *sup = c->sup_node;
1665 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1666 err = ubifs_write_sb_node(c, sup);
1671 if (c->need_recovery) {
1672 ubifs_msg(c, "completing deferred recovery");
1673 err = ubifs_write_rcvrd_mst_node(c);
1676 if (!ubifs_authenticated(c)) {
1677 err = ubifs_recover_size(c, true);
1681 err = ubifs_clean_lebs(c, c->sbuf);
1684 err = ubifs_recover_inl_heads(c, c->sbuf);
1688 /* A readonly mount is not allowed to have orphans */
1689 ubifs_assert(c, c->tot_orphans == 0);
1690 err = ubifs_clear_orphans(c);
1695 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1696 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1697 err = ubifs_write_master(c);
1702 c->ileb_buf = vmalloc(c->leb_size);
1708 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ + \
1709 UBIFS_CIPHER_BLOCK_SIZE, GFP_KERNEL);
1710 if (!c->write_reserve_buf) {
1715 err = ubifs_lpt_init(c, 0, 1);
1719 /* Create background thread */
1720 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1721 if (IS_ERR(c->bgt)) {
1722 err = PTR_ERR(c->bgt);
1724 ubifs_err(c, "cannot spawn \"%s\", error %d",
1728 wake_up_process(c->bgt);
1730 c->orph_buf = vmalloc(c->leb_size);
1736 /* Check for enough log space */
1737 lnum = c->lhead_lnum + 1;
1738 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1739 lnum = UBIFS_LOG_LNUM;
1740 if (lnum == c->ltail_lnum) {
1741 err = ubifs_consolidate_log(c);
1746 if (c->need_recovery) {
1747 err = ubifs_rcvry_gc_commit(c);
1751 if (ubifs_authenticated(c)) {
1752 err = ubifs_recover_size(c, false);
1757 err = ubifs_leb_unmap(c, c->gc_lnum);
1762 dbg_gen("re-mounted read-write");
1763 c->remounting_rw = 0;
1765 if (c->need_recovery) {
1766 c->need_recovery = 0;
1767 ubifs_msg(c, "deferred recovery completed");
1770 * Do not run the debugging space check if the were doing
1771 * recovery, because when we saved the information we had the
1772 * file-system in a state where the TNC and lprops has been
1773 * modified in memory, but all the I/O operations (including a
1774 * commit) were deferred. So the file-system was in
1775 * "non-committed" state. Now the file-system is in committed
1776 * state, and of course the amount of free space will change
1777 * because, for example, the old index size was imprecise.
1779 err = dbg_check_space_info(c);
1782 mutex_unlock(&c->umount_mutex);
1790 kthread_stop(c->bgt);
1794 kfree(c->write_reserve_buf);
1795 c->write_reserve_buf = NULL;
1798 ubifs_lpt_free(c, 1);
1799 c->remounting_rw = 0;
1800 mutex_unlock(&c->umount_mutex);
1805 * ubifs_remount_ro - re-mount in read-only mode.
1806 * @c: UBIFS file-system description object
1808 * We assume VFS has stopped writing. Possibly the background thread could be
1809 * running a commit, however kthread_stop will wait in that case.
1811 static void ubifs_remount_ro(struct ubifs_info *c)
1815 ubifs_assert(c, !c->need_recovery);
1816 ubifs_assert(c, !c->ro_mount);
1818 mutex_lock(&c->umount_mutex);
1820 kthread_stop(c->bgt);
1824 dbg_save_space_info(c);
1826 for (i = 0; i < c->jhead_cnt; i++) {
1827 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
1829 ubifs_ro_mode(c, err);
1832 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1833 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1834 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1835 err = ubifs_write_master(c);
1837 ubifs_ro_mode(c, err);
1841 kfree(c->write_reserve_buf);
1842 c->write_reserve_buf = NULL;
1845 ubifs_lpt_free(c, 1);
1847 err = dbg_check_space_info(c);
1849 ubifs_ro_mode(c, err);
1850 mutex_unlock(&c->umount_mutex);
1853 static void ubifs_put_super(struct super_block *sb)
1856 struct ubifs_info *c = sb->s_fs_info;
1858 ubifs_msg(c, "un-mount UBI device %d", c->vi.ubi_num);
1861 * The following asserts are only valid if there has not been a failure
1862 * of the media. For example, there will be dirty inodes if we failed
1863 * to write them back because of I/O errors.
1866 ubifs_assert(c, c->bi.idx_growth == 0);
1867 ubifs_assert(c, c->bi.dd_growth == 0);
1868 ubifs_assert(c, c->bi.data_growth == 0);
1872 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1873 * and file system un-mount. Namely, it prevents the shrinker from
1874 * picking this superblock for shrinking - it will be just skipped if
1875 * the mutex is locked.
1877 mutex_lock(&c->umount_mutex);
1880 * First of all kill the background thread to make sure it does
1881 * not interfere with un-mounting and freeing resources.
1884 kthread_stop(c->bgt);
1889 * On fatal errors c->ro_error is set to 1, in which case we do
1890 * not write the master node.
1895 /* Synchronize write-buffers */
1896 for (i = 0; i < c->jhead_cnt; i++) {
1897 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
1899 ubifs_ro_mode(c, err);
1903 * We are being cleanly unmounted which means the
1904 * orphans were killed - indicate this in the master
1905 * node. Also save the reserved GC LEB number.
1907 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1908 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1909 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1910 err = ubifs_write_master(c);
1913 * Recovery will attempt to fix the master area
1914 * next mount, so we just print a message and
1915 * continue to unmount normally.
1917 ubifs_err(c, "failed to write master node, error %d",
1920 for (i = 0; i < c->jhead_cnt; i++)
1921 /* Make sure write-buffer timers are canceled */
1922 hrtimer_cancel(&c->jheads[i].wbuf.timer);
1927 ubi_close_volume(c->ubi);
1928 mutex_unlock(&c->umount_mutex);
1931 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1934 struct ubifs_info *c = sb->s_fs_info;
1936 sync_filesystem(sb);
1937 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1939 err = ubifs_parse_options(c, data, 1);
1941 ubifs_err(c, "invalid or unknown remount parameter");
1945 if (c->ro_mount && !(*flags & SB_RDONLY)) {
1947 ubifs_msg(c, "cannot re-mount R/W due to prior errors");
1951 ubifs_msg(c, "cannot re-mount R/W - UBI volume is R/O");
1954 err = ubifs_remount_rw(c);
1957 } else if (!c->ro_mount && (*flags & SB_RDONLY)) {
1959 ubifs_msg(c, "cannot re-mount R/O due to prior errors");
1962 ubifs_remount_ro(c);
1965 if (c->bulk_read == 1)
1968 dbg_gen("disable bulk-read");
1969 mutex_lock(&c->bu_mutex);
1972 mutex_unlock(&c->bu_mutex);
1975 if (!c->need_recovery)
1976 ubifs_assert(c, c->lst.taken_empty_lebs > 0);
1981 const struct super_operations ubifs_super_operations = {
1982 .alloc_inode = ubifs_alloc_inode,
1983 .destroy_inode = ubifs_destroy_inode,
1984 .put_super = ubifs_put_super,
1985 .write_inode = ubifs_write_inode,
1986 .evict_inode = ubifs_evict_inode,
1987 .statfs = ubifs_statfs,
1988 .dirty_inode = ubifs_dirty_inode,
1989 .remount_fs = ubifs_remount_fs,
1990 .show_options = ubifs_show_options,
1991 .sync_fs = ubifs_sync_fs,
1995 * open_ubi - parse UBI device name string and open the UBI device.
1996 * @name: UBI volume name
1997 * @mode: UBI volume open mode
1999 * The primary method of mounting UBIFS is by specifying the UBI volume
2000 * character device node path. However, UBIFS may also be mounted withoug any
2001 * character device node using one of the following methods:
2003 * o ubiX_Y - mount UBI device number X, volume Y;
2004 * o ubiY - mount UBI device number 0, volume Y;
2005 * o ubiX:NAME - mount UBI device X, volume with name NAME;
2006 * o ubi:NAME - mount UBI device 0, volume with name NAME.
2008 * Alternative '!' separator may be used instead of ':' (because some shells
2009 * like busybox may interpret ':' as an NFS host name separator). This function
2010 * returns UBI volume description object in case of success and a negative
2011 * error code in case of failure.
2013 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
2015 struct ubi_volume_desc *ubi;
2019 if (!name || !*name)
2020 return ERR_PTR(-EINVAL);
2022 /* First, try to open using the device node path method */
2023 ubi = ubi_open_volume_path(name, mode);
2027 /* Try the "nodev" method */
2028 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
2029 return ERR_PTR(-EINVAL);
2031 /* ubi:NAME method */
2032 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
2033 return ubi_open_volume_nm(0, name + 4, mode);
2035 if (!isdigit(name[3]))
2036 return ERR_PTR(-EINVAL);
2038 dev = simple_strtoul(name + 3, &endptr, 0);
2041 if (*endptr == '\0')
2042 return ubi_open_volume(0, dev, mode);
2045 if (*endptr == '_' && isdigit(endptr[1])) {
2046 vol = simple_strtoul(endptr + 1, &endptr, 0);
2047 if (*endptr != '\0')
2048 return ERR_PTR(-EINVAL);
2049 return ubi_open_volume(dev, vol, mode);
2052 /* ubiX:NAME method */
2053 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
2054 return ubi_open_volume_nm(dev, ++endptr, mode);
2056 return ERR_PTR(-EINVAL);
2059 static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
2061 struct ubifs_info *c;
2063 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
2065 spin_lock_init(&c->cnt_lock);
2066 spin_lock_init(&c->cs_lock);
2067 spin_lock_init(&c->buds_lock);
2068 spin_lock_init(&c->space_lock);
2069 spin_lock_init(&c->orphan_lock);
2070 init_rwsem(&c->commit_sem);
2071 mutex_init(&c->lp_mutex);
2072 mutex_init(&c->tnc_mutex);
2073 mutex_init(&c->log_mutex);
2074 mutex_init(&c->umount_mutex);
2075 mutex_init(&c->bu_mutex);
2076 mutex_init(&c->write_reserve_mutex);
2077 init_waitqueue_head(&c->cmt_wq);
2079 c->old_idx = RB_ROOT;
2080 c->size_tree = RB_ROOT;
2081 c->orph_tree = RB_ROOT;
2082 INIT_LIST_HEAD(&c->infos_list);
2083 INIT_LIST_HEAD(&c->idx_gc);
2084 INIT_LIST_HEAD(&c->replay_list);
2085 INIT_LIST_HEAD(&c->replay_buds);
2086 INIT_LIST_HEAD(&c->uncat_list);
2087 INIT_LIST_HEAD(&c->empty_list);
2088 INIT_LIST_HEAD(&c->freeable_list);
2089 INIT_LIST_HEAD(&c->frdi_idx_list);
2090 INIT_LIST_HEAD(&c->unclean_leb_list);
2091 INIT_LIST_HEAD(&c->old_buds);
2092 INIT_LIST_HEAD(&c->orph_list);
2093 INIT_LIST_HEAD(&c->orph_new);
2094 c->no_chk_data_crc = 1;
2095 c->assert_action = ASSACT_RO;
2097 c->highest_inum = UBIFS_FIRST_INO;
2098 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
2100 ubi_get_volume_info(ubi, &c->vi);
2101 ubi_get_device_info(c->vi.ubi_num, &c->di);
2106 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
2108 struct ubifs_info *c = sb->s_fs_info;
2113 /* Re-open the UBI device in read-write mode */
2114 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2115 if (IS_ERR(c->ubi)) {
2116 err = PTR_ERR(c->ubi);
2120 err = ubifs_parse_options(c, data, 0);
2125 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2126 * UBIFS, I/O is not deferred, it is done immediately in readpage,
2127 * which means the user would have to wait not just for their own I/O
2128 * but the read-ahead I/O as well i.e. completely pointless.
2130 * Read-ahead will be disabled because @sb->s_bdi->ra_pages is 0. Also
2131 * @sb->s_bdi->capabilities are initialized to 0 so there won't be any
2132 * writeback happening.
2134 err = super_setup_bdi_name(sb, "ubifs_%d_%d", c->vi.ubi_num,
2140 sb->s_magic = UBIFS_SUPER_MAGIC;
2141 sb->s_blocksize = UBIFS_BLOCK_SIZE;
2142 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2143 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2144 if (c->max_inode_sz > MAX_LFS_FILESIZE)
2145 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2146 sb->s_op = &ubifs_super_operations;
2147 #ifdef CONFIG_UBIFS_FS_XATTR
2148 sb->s_xattr = ubifs_xattr_handlers;
2150 #ifdef CONFIG_FS_ENCRYPTION
2151 sb->s_cop = &ubifs_crypt_operations;
2154 mutex_lock(&c->umount_mutex);
2155 err = mount_ubifs(c);
2157 ubifs_assert(c, err < 0);
2161 /* Read the root inode */
2162 root = ubifs_iget(sb, UBIFS_ROOT_INO);
2164 err = PTR_ERR(root);
2168 sb->s_root = d_make_root(root);
2174 mutex_unlock(&c->umount_mutex);
2180 mutex_unlock(&c->umount_mutex);
2182 ubi_close_volume(c->ubi);
2187 static int sb_test(struct super_block *sb, void *data)
2189 struct ubifs_info *c1 = data;
2190 struct ubifs_info *c = sb->s_fs_info;
2192 return c->vi.cdev == c1->vi.cdev;
2195 static int sb_set(struct super_block *sb, void *data)
2197 sb->s_fs_info = data;
2198 return set_anon_super(sb, NULL);
2201 static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2202 const char *name, void *data)
2204 struct ubi_volume_desc *ubi;
2205 struct ubifs_info *c;
2206 struct super_block *sb;
2209 dbg_gen("name %s, flags %#x", name, flags);
2212 * Get UBI device number and volume ID. Mount it read-only so far
2213 * because this might be a new mount point, and UBI allows only one
2214 * read-write user at a time.
2216 ubi = open_ubi(name, UBI_READONLY);
2218 if (!(flags & SB_SILENT))
2219 pr_err("UBIFS error (pid: %d): cannot open \"%s\", error %d",
2220 current->pid, name, (int)PTR_ERR(ubi));
2221 return ERR_CAST(ubi);
2224 c = alloc_ubifs_info(ubi);
2230 dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2232 sb = sget(fs_type, sb_test, sb_set, flags, c);
2240 struct ubifs_info *c1 = sb->s_fs_info;
2242 /* A new mount point for already mounted UBIFS */
2243 dbg_gen("this ubi volume is already mounted");
2244 if (!!(flags & SB_RDONLY) != c1->ro_mount) {
2249 err = ubifs_fill_super(sb, data, flags & SB_SILENT ? 1 : 0);
2252 /* We do not support atime */
2253 sb->s_flags |= SB_ACTIVE;
2254 #ifndef CONFIG_UBIFS_ATIME_SUPPORT
2255 sb->s_flags |= SB_NOATIME;
2257 ubifs_msg(c, "full atime support is enabled.");
2261 /* 'fill_super()' opens ubi again so we must close it here */
2262 ubi_close_volume(ubi);
2264 return dget(sb->s_root);
2267 deactivate_locked_super(sb);
2269 ubi_close_volume(ubi);
2270 return ERR_PTR(err);
2273 static void kill_ubifs_super(struct super_block *s)
2275 struct ubifs_info *c = s->s_fs_info;
2280 static struct file_system_type ubifs_fs_type = {
2282 .owner = THIS_MODULE,
2283 .mount = ubifs_mount,
2284 .kill_sb = kill_ubifs_super,
2286 MODULE_ALIAS_FS("ubifs");
2289 * Inode slab cache constructor.
2291 static void inode_slab_ctor(void *obj)
2293 struct ubifs_inode *ui = obj;
2294 inode_init_once(&ui->vfs_inode);
2297 static int __init ubifs_init(void)
2301 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2303 /* Make sure node sizes are 8-byte aligned */
2304 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2305 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2306 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2307 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2308 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2309 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2310 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2311 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2312 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2313 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2314 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2316 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2317 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2318 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2319 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2320 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2321 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2323 /* Check min. node size */
2324 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2325 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2326 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2327 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2329 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2330 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2331 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2332 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2334 /* Defined node sizes */
2335 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2336 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2337 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2338 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2341 * We use 2 bit wide bit-fields to store compression type, which should
2342 * be amended if more compressors are added. The bit-fields are:
2343 * @compr_type in 'struct ubifs_inode', @default_compr in
2344 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2346 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2349 * We require that PAGE_SIZE is greater-than-or-equal-to
2350 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2352 if (PAGE_SIZE < UBIFS_BLOCK_SIZE) {
2353 pr_err("UBIFS error (pid %d): VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
2354 current->pid, (unsigned int)PAGE_SIZE);
2358 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2359 sizeof(struct ubifs_inode), 0,
2360 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT |
2361 SLAB_ACCOUNT, &inode_slab_ctor);
2362 if (!ubifs_inode_slab)
2365 err = register_shrinker(&ubifs_shrinker_info);
2369 err = ubifs_compressors_init();
2373 err = dbg_debugfs_init();
2377 err = register_filesystem(&ubifs_fs_type);
2379 pr_err("UBIFS error (pid %d): cannot register file system, error %d",
2388 ubifs_compressors_exit();
2390 unregister_shrinker(&ubifs_shrinker_info);
2392 kmem_cache_destroy(ubifs_inode_slab);
2395 /* late_initcall to let compressors initialize first */
2396 late_initcall(ubifs_init);
2398 static void __exit ubifs_exit(void)
2400 WARN_ON(!list_empty(&ubifs_infos));
2401 WARN_ON(atomic_long_read(&ubifs_clean_zn_cnt) != 0);
2404 ubifs_compressors_exit();
2405 unregister_shrinker(&ubifs_shrinker_info);
2408 * Make sure all delayed rcu free inodes are flushed before we
2412 kmem_cache_destroy(ubifs_inode_slab);
2413 unregister_filesystem(&ubifs_fs_type);
2415 module_exit(ubifs_exit);
2417 MODULE_LICENSE("GPL");
2418 MODULE_VERSION(__stringify(UBIFS_VERSION));
2419 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2420 MODULE_DESCRIPTION("UBIFS - UBI File System");