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>
39 #include <linux/smp_lock.h>
43 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
44 * allocating too much.
46 #define UBIFS_KMALLOC_OK (128*1024)
48 /* Slab cache for UBIFS inodes */
49 struct kmem_cache *ubifs_inode_slab;
51 /* UBIFS TNC shrinker description */
52 static struct shrinker ubifs_shrinker_info = {
53 .shrink = ubifs_shrinker,
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("inode is too large (%lld)",
74 (long long)inode->i_size);
78 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
79 ubifs_err("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 && (inode->i_mode & S_IFMT) != S_IFREG)
92 if (!ubifs_compr_present(ui->compr_type)) {
93 ubifs_warn("inode %lu uses '%s' compression, but it was not "
94 "compiled in", inode->i_ino,
95 ubifs_compr_name(ui->compr_type));
98 err = dbg_check_dir_size(c, inode);
102 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
106 struct ubifs_ino_node *ino;
107 struct ubifs_info *c = sb->s_fs_info;
109 struct ubifs_inode *ui;
111 dbg_gen("inode %lu", inum);
113 inode = iget_locked(sb, inum);
115 return ERR_PTR(-ENOMEM);
116 if (!(inode->i_state & I_NEW))
118 ui = ubifs_inode(inode);
120 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
126 ino_key_init(c, &key, inode->i_ino);
128 err = ubifs_tnc_lookup(c, &key, ino);
132 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
133 inode->i_nlink = le32_to_cpu(ino->nlink);
134 inode->i_uid = le32_to_cpu(ino->uid);
135 inode->i_gid = le32_to_cpu(ino->gid);
136 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
137 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
138 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
139 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
140 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
141 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
142 inode->i_mode = le32_to_cpu(ino->mode);
143 inode->i_size = le64_to_cpu(ino->size);
145 ui->data_len = le32_to_cpu(ino->data_len);
146 ui->flags = le32_to_cpu(ino->flags);
147 ui->compr_type = le16_to_cpu(ino->compr_type);
148 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
149 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
150 ui->xattr_size = le32_to_cpu(ino->xattr_size);
151 ui->xattr_names = le32_to_cpu(ino->xattr_names);
152 ui->synced_i_size = ui->ui_size = inode->i_size;
154 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
156 err = validate_inode(c, inode);
160 /* Disable read-ahead */
161 inode->i_mapping->backing_dev_info = &c->bdi;
163 switch (inode->i_mode & S_IFMT) {
165 inode->i_mapping->a_ops = &ubifs_file_address_operations;
166 inode->i_op = &ubifs_file_inode_operations;
167 inode->i_fop = &ubifs_file_operations;
169 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
174 memcpy(ui->data, ino->data, ui->data_len);
175 ((char *)ui->data)[ui->data_len] = '\0';
176 } else if (ui->data_len != 0) {
182 inode->i_op = &ubifs_dir_inode_operations;
183 inode->i_fop = &ubifs_dir_operations;
184 if (ui->data_len != 0) {
190 inode->i_op = &ubifs_symlink_inode_operations;
191 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
195 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
200 memcpy(ui->data, ino->data, ui->data_len);
201 ((char *)ui->data)[ui->data_len] = '\0';
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("inode %lu validation failed, error %d", inode->i_ino, err);
250 dbg_dump_node(c, ino);
251 dbg_dump_inode(c, inode);
256 ubifs_err("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_destroy_inode(struct inode *inode)
278 struct ubifs_inode *ui = ubifs_inode(inode);
281 kmem_cache_free(ubifs_inode_slab, inode);
285 * Note, Linux write-back code calls this without 'i_mutex'.
287 static int ubifs_write_inode(struct inode *inode, int wait)
290 struct ubifs_info *c = inode->i_sb->s_fs_info;
291 struct ubifs_inode *ui = ubifs_inode(inode);
293 ubifs_assert(!ui->xattr);
294 if (is_bad_inode(inode))
297 mutex_lock(&ui->ui_mutex);
299 * Due to races between write-back forced by budgeting
300 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
301 * have already been synchronized, do not do this again. This might
302 * also happen if it was synchronized in an VFS operation, e.g.
306 mutex_unlock(&ui->ui_mutex);
311 * As an optimization, do not write orphan inodes to the media just
312 * because this is not needed.
314 dbg_gen("inode %lu, mode %#x, nlink %u",
315 inode->i_ino, (int)inode->i_mode, inode->i_nlink);
316 if (inode->i_nlink) {
317 err = ubifs_jnl_write_inode(c, inode);
319 ubifs_err("can't write inode %lu, error %d",
324 mutex_unlock(&ui->ui_mutex);
325 ubifs_release_dirty_inode_budget(c, ui);
329 static void ubifs_delete_inode(struct inode *inode)
332 struct ubifs_info *c = inode->i_sb->s_fs_info;
333 struct ubifs_inode *ui = ubifs_inode(inode);
337 * Extended attribute inode deletions are fully handled in
338 * 'ubifs_removexattr()'. These inodes are special and have
339 * limited usage, so there is nothing to do here.
343 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
344 ubifs_assert(!atomic_read(&inode->i_count));
345 ubifs_assert(inode->i_nlink == 0);
347 truncate_inode_pages(&inode->i_data, 0);
348 if (is_bad_inode(inode))
351 ui->ui_size = inode->i_size = 0;
352 err = ubifs_jnl_delete_inode(c, inode);
355 * Worst case we have a lost orphan inode wasting space, so a
356 * simple error message is OK here.
358 ubifs_err("can't delete inode %lu, error %d",
363 ubifs_release_dirty_inode_budget(c, ui);
367 static void ubifs_dirty_inode(struct inode *inode)
369 struct ubifs_inode *ui = ubifs_inode(inode);
371 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
374 dbg_gen("inode %lu", inode->i_ino);
378 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
380 struct ubifs_info *c = dentry->d_sb->s_fs_info;
381 unsigned long long free;
382 __le32 *uuid = (__le32 *)c->uuid;
384 free = ubifs_get_free_space(c);
385 dbg_gen("free space %lld bytes (%lld blocks)",
386 free, free >> UBIFS_BLOCK_SHIFT);
388 buf->f_type = UBIFS_SUPER_MAGIC;
389 buf->f_bsize = UBIFS_BLOCK_SIZE;
390 buf->f_blocks = c->block_cnt;
391 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
392 if (free > c->report_rp_size)
393 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
398 buf->f_namelen = UBIFS_MAX_NLEN;
399 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
400 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
401 ubifs_assert(buf->f_bfree <= c->block_cnt);
405 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
407 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
409 if (c->mount_opts.unmount_mode == 2)
410 seq_printf(s, ",fast_unmount");
411 else if (c->mount_opts.unmount_mode == 1)
412 seq_printf(s, ",norm_unmount");
414 if (c->mount_opts.bulk_read == 2)
415 seq_printf(s, ",bulk_read");
416 else if (c->mount_opts.bulk_read == 1)
417 seq_printf(s, ",no_bulk_read");
419 if (c->mount_opts.chk_data_crc == 2)
420 seq_printf(s, ",chk_data_crc");
421 else if (c->mount_opts.chk_data_crc == 1)
422 seq_printf(s, ",no_chk_data_crc");
424 if (c->mount_opts.override_compr) {
425 seq_printf(s, ",compr=%s",
426 ubifs_compr_name(c->mount_opts.compr_type));
432 static int ubifs_sync_fs(struct super_block *sb, int wait)
435 struct ubifs_info *c = sb->s_fs_info;
436 struct writeback_control wbc = {
437 .sync_mode = WB_SYNC_ALL,
439 .range_end = LLONG_MAX,
440 .nr_to_write = LONG_MAX,
444 * Zero @wait is just an advisory thing to help the file system shove
445 * lots of data into the queues, and there will be the second
446 * '->sync_fs()' call, with non-zero @wait.
452 * VFS calls '->sync_fs()' before synchronizing all dirty inodes and
453 * pages, so synchronize them first, then commit the journal. Strictly
454 * speaking, it is not necessary to commit the journal here,
455 * synchronizing write-buffers would be enough. But committing makes
456 * UBIFS free space predictions much more accurate, so we want to let
457 * the user be able to get more accurate results of 'statfs()' after
458 * they synchronize the file system.
460 generic_sync_sb_inodes(sb, &wbc);
463 * Synchronize write buffers, because 'ubifs_run_commit()' does not
464 * do this if it waits for an already running commit.
466 for (i = 0; i < c->jhead_cnt; i++) {
467 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
472 err = ubifs_run_commit(c);
476 return ubi_sync(c->vi.ubi_num);
480 * init_constants_early - initialize UBIFS constants.
481 * @c: UBIFS file-system description object
483 * This function initialize UBIFS constants which do not need the superblock to
484 * be read. It also checks that the UBI volume satisfies basic UBIFS
485 * requirements. Returns zero in case of success and a negative error code in
488 static int init_constants_early(struct ubifs_info *c)
490 if (c->vi.corrupted) {
491 ubifs_warn("UBI volume is corrupted - read-only mode");
496 ubifs_msg("read-only UBI device");
500 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
501 ubifs_msg("static UBI volume - read-only mode");
505 c->leb_cnt = c->vi.size;
506 c->leb_size = c->vi.usable_leb_size;
507 c->half_leb_size = c->leb_size / 2;
508 c->min_io_size = c->di.min_io_size;
509 c->min_io_shift = fls(c->min_io_size) - 1;
511 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
512 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
513 c->leb_size, UBIFS_MIN_LEB_SZ);
517 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
518 ubifs_err("too few LEBs (%d), min. is %d",
519 c->leb_cnt, UBIFS_MIN_LEB_CNT);
523 if (!is_power_of_2(c->min_io_size)) {
524 ubifs_err("bad min. I/O size %d", c->min_io_size);
529 * UBIFS aligns all node to 8-byte boundary, so to make function in
530 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
533 if (c->min_io_size < 8) {
538 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
539 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
542 * Initialize node length ranges which are mostly needed for node
545 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
546 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
547 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
548 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
549 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
550 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
552 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
553 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
554 c->ranges[UBIFS_ORPH_NODE].min_len =
555 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
556 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
557 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
558 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
559 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
560 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
561 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
562 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
564 * Minimum indexing node size is amended later when superblock is
565 * read and the key length is known.
567 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
569 * Maximum indexing node size is amended later when superblock is
570 * read and the fanout is known.
572 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
575 * Initialize dead and dark LEB space watermarks. See gc.c for comments
576 * about these values.
578 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
579 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
582 * Calculate how many bytes would be wasted at the end of LEB if it was
583 * fully filled with data nodes of maximum size. This is used in
584 * calculations when reporting free space.
586 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
588 /* Buffer size for bulk-reads */
589 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
590 if (c->max_bu_buf_len > c->leb_size)
591 c->max_bu_buf_len = c->leb_size;
596 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
597 * @c: UBIFS file-system description object
598 * @lnum: LEB the write-buffer was synchronized to
599 * @free: how many free bytes left in this LEB
600 * @pad: how many bytes were padded
602 * This is a callback function which is called by the I/O unit when the
603 * write-buffer is synchronized. We need this to correctly maintain space
604 * accounting in bud logical eraseblocks. This function returns zero in case of
605 * success and a negative error code in case of failure.
607 * This function actually belongs to the journal, but we keep it here because
608 * we want to keep it static.
610 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
612 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
616 * init_constants_sb - initialize UBIFS constants.
617 * @c: UBIFS file-system description object
619 * This is a helper function which initializes various UBIFS constants after
620 * the superblock has been read. It also checks various UBIFS parameters and
621 * makes sure they are all right. Returns zero in case of success and a
622 * negative error code in case of failure.
624 static int init_constants_sb(struct ubifs_info *c)
629 c->main_bytes = (long long)c->main_lebs * c->leb_size;
630 c->max_znode_sz = sizeof(struct ubifs_znode) +
631 c->fanout * sizeof(struct ubifs_zbranch);
633 tmp = ubifs_idx_node_sz(c, 1);
634 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
635 c->min_idx_node_sz = ALIGN(tmp, 8);
637 tmp = ubifs_idx_node_sz(c, c->fanout);
638 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
639 c->max_idx_node_sz = ALIGN(tmp, 8);
641 /* Make sure LEB size is large enough to fit full commit */
642 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
643 tmp = ALIGN(tmp, c->min_io_size);
644 if (tmp > c->leb_size) {
645 dbg_err("too small LEB size %d, at least %d needed",
651 * Make sure that the log is large enough to fit reference nodes for
652 * all buds plus one reserved LEB.
654 tmp64 = c->max_bud_bytes + c->leb_size - 1;
655 c->max_bud_cnt = div_u64(tmp64, c->leb_size);
656 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
659 if (c->log_lebs < tmp) {
660 dbg_err("too small log %d LEBs, required min. %d LEBs",
666 * When budgeting we assume worst-case scenarios when the pages are not
667 * be compressed and direntries are of the maximum size.
669 * Note, data, which may be stored in inodes is budgeted separately, so
670 * it is not included into 'c->inode_budget'.
672 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
673 c->inode_budget = UBIFS_INO_NODE_SZ;
674 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
677 * When the amount of flash space used by buds becomes
678 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
679 * The writers are unblocked when the commit is finished. To avoid
680 * writers to be blocked UBIFS initiates background commit in advance,
681 * when number of bud bytes becomes above the limit defined below.
683 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
686 * Ensure minimum journal size. All the bytes in the journal heads are
687 * considered to be used, when calculating the current journal usage.
688 * Consequently, if the journal is too small, UBIFS will treat it as
691 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
692 if (c->bg_bud_bytes < tmp64)
693 c->bg_bud_bytes = tmp64;
694 if (c->max_bud_bytes < tmp64 + c->leb_size)
695 c->max_bud_bytes = tmp64 + c->leb_size;
697 err = ubifs_calc_lpt_geom(c);
701 /* Initialize effective LEB size used in budgeting calculations */
702 c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
707 * init_constants_master - initialize UBIFS constants.
708 * @c: UBIFS file-system description object
710 * This is a helper function which initializes various UBIFS constants after
711 * the master node has been read. It also checks various UBIFS parameters and
712 * makes sure they are all right.
714 static void init_constants_master(struct ubifs_info *c)
718 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
719 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
722 * Calculate total amount of FS blocks. This number is not used
723 * internally because it does not make much sense for UBIFS, but it is
724 * necessary to report something for the 'statfs()' call.
726 * Subtract the LEB reserved for GC, the LEB which is reserved for
727 * deletions, minimum LEBs for the index, and assume only one journal
730 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
731 tmp64 *= (long long)c->leb_size - c->leb_overhead;
732 tmp64 = ubifs_reported_space(c, tmp64);
733 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
737 * take_gc_lnum - reserve GC LEB.
738 * @c: UBIFS file-system description object
740 * This function ensures that the LEB reserved for garbage collection is marked
741 * as "taken" in lprops. We also have to set free space to LEB size and dirty
742 * space to zero, because lprops may contain out-of-date information if the
743 * file-system was un-mounted before it has been committed. This function
744 * returns zero in case of success and a negative error code in case of
747 static int take_gc_lnum(struct ubifs_info *c)
751 if (c->gc_lnum == -1) {
752 ubifs_err("no LEB for GC");
756 /* And we have to tell lprops that this LEB is taken */
757 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
763 * alloc_wbufs - allocate write-buffers.
764 * @c: UBIFS file-system description object
766 * This helper function allocates and initializes UBIFS write-buffers. Returns
767 * zero in case of success and %-ENOMEM in case of failure.
769 static int alloc_wbufs(struct ubifs_info *c)
773 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
778 /* Initialize journal heads */
779 for (i = 0; i < c->jhead_cnt; i++) {
780 INIT_LIST_HEAD(&c->jheads[i].buds_list);
781 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
785 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
786 c->jheads[i].wbuf.jhead = i;
789 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
791 * Garbage Collector head likely contains long-term data and
792 * does not need to be synchronized by timer.
794 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
795 c->jheads[GCHD].wbuf.timeout = 0;
801 * free_wbufs - free write-buffers.
802 * @c: UBIFS file-system description object
804 static void free_wbufs(struct ubifs_info *c)
809 for (i = 0; i < c->jhead_cnt; i++) {
810 kfree(c->jheads[i].wbuf.buf);
811 kfree(c->jheads[i].wbuf.inodes);
819 * free_orphans - free orphans.
820 * @c: UBIFS file-system description object
822 static void free_orphans(struct ubifs_info *c)
824 struct ubifs_orphan *orph;
826 while (c->orph_dnext) {
827 orph = c->orph_dnext;
828 c->orph_dnext = orph->dnext;
829 list_del(&orph->list);
833 while (!list_empty(&c->orph_list)) {
834 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
835 list_del(&orph->list);
837 dbg_err("orphan list not empty at unmount");
845 * free_buds - free per-bud objects.
846 * @c: UBIFS file-system description object
848 static void free_buds(struct ubifs_info *c)
850 struct rb_node *this = c->buds.rb_node;
851 struct ubifs_bud *bud;
855 this = this->rb_left;
856 else if (this->rb_right)
857 this = this->rb_right;
859 bud = rb_entry(this, struct ubifs_bud, rb);
860 this = rb_parent(this);
862 if (this->rb_left == &bud->rb)
863 this->rb_left = NULL;
865 this->rb_right = NULL;
873 * check_volume_empty - check if the UBI volume is empty.
874 * @c: UBIFS file-system description object
876 * This function checks if the UBIFS volume is empty by looking if its LEBs are
877 * mapped or not. The result of checking is stored in the @c->empty variable.
878 * Returns zero in case of success and a negative error code in case of
881 static int check_volume_empty(struct ubifs_info *c)
886 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
887 err = ubi_is_mapped(c->ubi, lnum);
888 if (unlikely(err < 0))
902 * UBIFS mount options.
904 * Opt_fast_unmount: do not run a journal commit before un-mounting
905 * Opt_norm_unmount: run a journal commit before un-mounting
906 * Opt_bulk_read: enable bulk-reads
907 * Opt_no_bulk_read: disable bulk-reads
908 * Opt_chk_data_crc: check CRCs when reading data nodes
909 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
910 * Opt_override_compr: override default compressor
911 * Opt_err: just end of array marker
924 static const match_table_t tokens = {
925 {Opt_fast_unmount, "fast_unmount"},
926 {Opt_norm_unmount, "norm_unmount"},
927 {Opt_bulk_read, "bulk_read"},
928 {Opt_no_bulk_read, "no_bulk_read"},
929 {Opt_chk_data_crc, "chk_data_crc"},
930 {Opt_no_chk_data_crc, "no_chk_data_crc"},
931 {Opt_override_compr, "compr=%s"},
936 * ubifs_parse_options - parse mount parameters.
937 * @c: UBIFS file-system description object
938 * @options: parameters to parse
939 * @is_remount: non-zero if this is FS re-mount
941 * This function parses UBIFS mount options and returns zero in case success
942 * and a negative error code in case of failure.
944 static int ubifs_parse_options(struct ubifs_info *c, char *options,
948 substring_t args[MAX_OPT_ARGS];
953 while ((p = strsep(&options, ","))) {
959 token = match_token(p, tokens, args);
962 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
963 * We accepte them in order to be backware-compatible. But this
964 * should be removed at some point.
966 case Opt_fast_unmount:
967 c->mount_opts.unmount_mode = 2;
969 case Opt_norm_unmount:
970 c->mount_opts.unmount_mode = 1;
973 c->mount_opts.bulk_read = 2;
976 case Opt_no_bulk_read:
977 c->mount_opts.bulk_read = 1;
980 case Opt_chk_data_crc:
981 c->mount_opts.chk_data_crc = 2;
982 c->no_chk_data_crc = 0;
984 case Opt_no_chk_data_crc:
985 c->mount_opts.chk_data_crc = 1;
986 c->no_chk_data_crc = 1;
988 case Opt_override_compr:
990 char *name = match_strdup(&args[0]);
994 if (!strcmp(name, "none"))
995 c->mount_opts.compr_type = UBIFS_COMPR_NONE;
996 else if (!strcmp(name, "lzo"))
997 c->mount_opts.compr_type = UBIFS_COMPR_LZO;
998 else if (!strcmp(name, "zlib"))
999 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1001 ubifs_err("unknown compressor \"%s\"", name);
1006 c->mount_opts.override_compr = 1;
1007 c->default_compr = c->mount_opts.compr_type;
1011 ubifs_err("unrecognized mount option \"%s\" "
1012 "or missing value", p);
1021 * destroy_journal - destroy journal data structures.
1022 * @c: UBIFS file-system description object
1024 * This function destroys journal data structures including those that may have
1025 * been created by recovery functions.
1027 static void destroy_journal(struct ubifs_info *c)
1029 while (!list_empty(&c->unclean_leb_list)) {
1030 struct ubifs_unclean_leb *ucleb;
1032 ucleb = list_entry(c->unclean_leb_list.next,
1033 struct ubifs_unclean_leb, list);
1034 list_del(&ucleb->list);
1037 while (!list_empty(&c->old_buds)) {
1038 struct ubifs_bud *bud;
1040 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1041 list_del(&bud->list);
1044 ubifs_destroy_idx_gc(c);
1045 ubifs_destroy_size_tree(c);
1051 * bu_init - initialize bulk-read information.
1052 * @c: UBIFS file-system description object
1054 static void bu_init(struct ubifs_info *c)
1056 ubifs_assert(c->bulk_read == 1);
1059 return; /* Already initialized */
1062 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1064 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1065 c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1069 /* Just disable bulk-read */
1070 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1071 "disabling it", c->max_bu_buf_len);
1072 c->mount_opts.bulk_read = 1;
1079 * check_free_space - check if there is enough free space to mount.
1080 * @c: UBIFS file-system description object
1082 * This function makes sure UBIFS has enough free space to be mounted in
1083 * read/write mode. UBIFS must always have some free space to allow deletions.
1085 static int check_free_space(struct ubifs_info *c)
1087 ubifs_assert(c->dark_wm > 0);
1088 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1089 ubifs_err("insufficient free space to mount in read/write mode");
1098 * mount_ubifs - mount UBIFS file-system.
1099 * @c: UBIFS file-system description object
1101 * This function mounts UBIFS file system. Returns zero in case of success and
1102 * a negative error code in case of failure.
1104 * Note, the function does not de-allocate resources it it fails half way
1105 * through, and the caller has to do this instead.
1107 static int mount_ubifs(struct ubifs_info *c)
1109 struct super_block *sb = c->vfs_sb;
1110 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
1114 err = init_constants_early(c);
1118 err = ubifs_debugging_init(c);
1122 err = check_volume_empty(c);
1126 if (c->empty && (mounted_read_only || c->ro_media)) {
1128 * This UBI volume is empty, and read-only, or the file system
1129 * is mounted read-only - we cannot format it.
1131 ubifs_err("can't format empty UBI volume: read-only %s",
1132 c->ro_media ? "UBI volume" : "mount");
1137 if (c->ro_media && !mounted_read_only) {
1138 ubifs_err("cannot mount read-write - read-only media");
1144 * The requirement for the buffer is that it should fit indexing B-tree
1145 * height amount of integers. We assume the height if the TNC tree will
1149 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1150 if (!c->bottom_up_buf)
1153 c->sbuf = vmalloc(c->leb_size);
1157 if (!mounted_read_only) {
1158 c->ileb_buf = vmalloc(c->leb_size);
1163 if (c->bulk_read == 1)
1167 * We have to check all CRCs, even for data nodes, when we mount the FS
1168 * (specifically, when we are replaying).
1170 c->always_chk_crc = 1;
1172 err = ubifs_read_superblock(c);
1177 * Make sure the compressor which is set as default in the superblock
1178 * or overridden by mount options is actually compiled in.
1180 if (!ubifs_compr_present(c->default_compr)) {
1181 ubifs_err("'compressor \"%s\" is not compiled in",
1182 ubifs_compr_name(c->default_compr));
1186 err = init_constants_sb(c);
1190 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1191 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1192 c->cbuf = kmalloc(sz, GFP_NOFS);
1198 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1199 if (!mounted_read_only) {
1200 err = alloc_wbufs(c);
1204 /* Create background thread */
1205 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1206 if (IS_ERR(c->bgt)) {
1207 err = PTR_ERR(c->bgt);
1209 ubifs_err("cannot spawn \"%s\", error %d",
1213 wake_up_process(c->bgt);
1216 err = ubifs_read_master(c);
1220 init_constants_master(c);
1222 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1223 ubifs_msg("recovery needed");
1224 c->need_recovery = 1;
1225 if (!mounted_read_only) {
1226 err = ubifs_recover_inl_heads(c, c->sbuf);
1230 } else if (!mounted_read_only) {
1232 * Set the "dirty" flag so that if we reboot uncleanly we
1233 * will notice this immediately on the next mount.
1235 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1236 err = ubifs_write_master(c);
1241 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1245 err = dbg_check_idx_size(c, c->old_idx_sz);
1249 err = ubifs_replay_journal(c);
1253 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1257 if (!mounted_read_only) {
1260 err = check_free_space(c);
1264 /* Check for enough log space */
1265 lnum = c->lhead_lnum + 1;
1266 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1267 lnum = UBIFS_LOG_LNUM;
1268 if (lnum == c->ltail_lnum) {
1269 err = ubifs_consolidate_log(c);
1274 if (c->need_recovery) {
1275 err = ubifs_recover_size(c);
1278 err = ubifs_rcvry_gc_commit(c);
1280 err = take_gc_lnum(c);
1285 * GC LEB may contain garbage if there was an unclean
1286 * reboot, and it should be un-mapped.
1288 err = ubifs_leb_unmap(c, c->gc_lnum);
1293 err = dbg_check_lprops(c);
1296 } else if (c->need_recovery) {
1297 err = ubifs_recover_size(c);
1302 * Even if we mount read-only, we have to set space in GC LEB
1303 * to proper value because this affects UBIFS free space
1304 * reporting. We do not want to have a situation when
1305 * re-mounting from R/O to R/W changes amount of free space.
1307 err = take_gc_lnum(c);
1312 spin_lock(&ubifs_infos_lock);
1313 list_add_tail(&c->infos_list, &ubifs_infos);
1314 spin_unlock(&ubifs_infos_lock);
1316 if (c->need_recovery) {
1317 if (mounted_read_only)
1318 ubifs_msg("recovery deferred");
1320 c->need_recovery = 0;
1321 ubifs_msg("recovery completed");
1323 * GC LEB has to be empty and taken at this point. But
1324 * the journal head LEBs may also be accounted as
1325 * "empty taken" if they are empty.
1327 ubifs_assert(c->lst.taken_empty_lebs > 0);
1330 ubifs_assert(c->lst.taken_empty_lebs > 0);
1332 err = dbg_check_filesystem(c);
1336 err = dbg_debugfs_init_fs(c);
1340 c->always_chk_crc = 0;
1342 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1343 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1344 if (mounted_read_only)
1345 ubifs_msg("mounted read-only");
1346 x = (long long)c->main_lebs * c->leb_size;
1347 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1348 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1349 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1350 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1351 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1352 ubifs_msg("media format: w%d/r%d (latest is w%d/r%d)",
1353 c->fmt_version, c->ro_compat_version,
1354 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1355 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1356 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1357 c->report_rp_size, c->report_rp_size >> 10);
1359 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1360 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1361 dbg_msg("LEB size: %d bytes (%d KiB)",
1362 c->leb_size, c->leb_size >> 10);
1363 dbg_msg("data journal heads: %d",
1364 c->jhead_cnt - NONDATA_JHEADS_CNT);
1365 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1366 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1367 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1368 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1369 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1370 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1371 dbg_msg("big_lpt %d", c->big_lpt);
1372 dbg_msg("log LEBs: %d (%d - %d)",
1373 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1374 dbg_msg("LPT area LEBs: %d (%d - %d)",
1375 c->lpt_lebs, c->lpt_first, c->lpt_last);
1376 dbg_msg("orphan area LEBs: %d (%d - %d)",
1377 c->orph_lebs, c->orph_first, c->orph_last);
1378 dbg_msg("main area LEBs: %d (%d - %d)",
1379 c->main_lebs, c->main_first, c->leb_cnt - 1);
1380 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1381 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1382 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1383 dbg_msg("key hash type: %d", c->key_hash_type);
1384 dbg_msg("tree fanout: %d", c->fanout);
1385 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1386 dbg_msg("first main LEB: %d", c->main_first);
1387 dbg_msg("max. znode size %d", c->max_znode_sz);
1388 dbg_msg("max. index node size %d", c->max_idx_node_sz);
1389 dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
1390 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1391 dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
1392 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1393 dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
1394 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1395 dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu",
1396 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1397 UBIFS_MAX_DENT_NODE_SZ);
1398 dbg_msg("dead watermark: %d", c->dead_wm);
1399 dbg_msg("dark watermark: %d", c->dark_wm);
1400 dbg_msg("LEB overhead: %d", c->leb_overhead);
1401 x = (long long)c->main_lebs * c->dark_wm;
1402 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1403 x, x >> 10, x >> 20);
1404 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1405 c->max_bud_bytes, c->max_bud_bytes >> 10,
1406 c->max_bud_bytes >> 20);
1407 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1408 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1409 c->bg_bud_bytes >> 20);
1410 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1411 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1412 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1413 dbg_msg("commit number: %llu", c->cmt_no);
1418 spin_lock(&ubifs_infos_lock);
1419 list_del(&c->infos_list);
1420 spin_unlock(&ubifs_infos_lock);
1426 ubifs_lpt_free(c, 0);
1429 kfree(c->rcvrd_mst_node);
1431 kthread_stop(c->bgt);
1440 kfree(c->bottom_up_buf);
1441 ubifs_debugging_exit(c);
1446 * ubifs_umount - un-mount UBIFS file-system.
1447 * @c: UBIFS file-system description object
1449 * Note, this function is called to free allocated resourced when un-mounting,
1450 * as well as free resources when an error occurred while we were half way
1451 * through mounting (error path cleanup function). So it has to make sure the
1452 * resource was actually allocated before freeing it.
1454 static void ubifs_umount(struct ubifs_info *c)
1456 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1459 dbg_debugfs_exit_fs(c);
1460 spin_lock(&ubifs_infos_lock);
1461 list_del(&c->infos_list);
1462 spin_unlock(&ubifs_infos_lock);
1465 kthread_stop(c->bgt);
1470 ubifs_lpt_free(c, 0);
1473 kfree(c->rcvrd_mst_node);
1478 kfree(c->bottom_up_buf);
1479 ubifs_debugging_exit(c);
1483 * ubifs_remount_rw - re-mount in read-write mode.
1484 * @c: UBIFS file-system description object
1486 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1487 * mode. This function allocates the needed resources and re-mounts UBIFS in
1490 static int ubifs_remount_rw(struct ubifs_info *c)
1494 if (c->rw_incompat) {
1495 ubifs_err("the file-system is not R/W-compatible");
1496 ubifs_msg("on-flash format version is w%d/r%d, but software "
1497 "only supports up to version w%d/r%d", c->fmt_version,
1498 c->ro_compat_version, UBIFS_FORMAT_VERSION,
1499 UBIFS_RO_COMPAT_VERSION);
1503 mutex_lock(&c->umount_mutex);
1504 dbg_save_space_info(c);
1505 c->remounting_rw = 1;
1506 c->always_chk_crc = 1;
1508 err = check_free_space(c);
1512 if (c->old_leb_cnt != c->leb_cnt) {
1513 struct ubifs_sb_node *sup;
1515 sup = ubifs_read_sb_node(c);
1520 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1521 err = ubifs_write_sb_node(c, sup);
1526 if (c->need_recovery) {
1527 ubifs_msg("completing deferred recovery");
1528 err = ubifs_write_rcvrd_mst_node(c);
1531 err = ubifs_recover_size(c);
1534 err = ubifs_clean_lebs(c, c->sbuf);
1537 err = ubifs_recover_inl_heads(c, c->sbuf);
1541 /* A readonly mount is not allowed to have orphans */
1542 ubifs_assert(c->tot_orphans == 0);
1543 err = ubifs_clear_orphans(c);
1548 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1549 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1550 err = ubifs_write_master(c);
1555 c->ileb_buf = vmalloc(c->leb_size);
1561 err = ubifs_lpt_init(c, 0, 1);
1565 err = alloc_wbufs(c);
1569 ubifs_create_buds_lists(c);
1571 /* Create background thread */
1572 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1573 if (IS_ERR(c->bgt)) {
1574 err = PTR_ERR(c->bgt);
1576 ubifs_err("cannot spawn \"%s\", error %d",
1580 wake_up_process(c->bgt);
1582 c->orph_buf = vmalloc(c->leb_size);
1588 /* Check for enough log space */
1589 lnum = c->lhead_lnum + 1;
1590 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1591 lnum = UBIFS_LOG_LNUM;
1592 if (lnum == c->ltail_lnum) {
1593 err = ubifs_consolidate_log(c);
1598 if (c->need_recovery)
1599 err = ubifs_rcvry_gc_commit(c);
1601 err = ubifs_leb_unmap(c, c->gc_lnum);
1605 if (c->need_recovery) {
1606 c->need_recovery = 0;
1607 ubifs_msg("deferred recovery completed");
1610 dbg_gen("re-mounted read-write");
1611 c->vfs_sb->s_flags &= ~MS_RDONLY;
1612 c->remounting_rw = 0;
1613 c->always_chk_crc = 0;
1614 err = dbg_check_space_info(c);
1615 mutex_unlock(&c->umount_mutex);
1622 kthread_stop(c->bgt);
1628 ubifs_lpt_free(c, 1);
1629 c->remounting_rw = 0;
1630 c->always_chk_crc = 0;
1631 mutex_unlock(&c->umount_mutex);
1636 * ubifs_remount_ro - re-mount in read-only mode.
1637 * @c: UBIFS file-system description object
1639 * We assume VFS has stopped writing. Possibly the background thread could be
1640 * running a commit, however kthread_stop will wait in that case.
1642 static void ubifs_remount_ro(struct ubifs_info *c)
1646 ubifs_assert(!c->need_recovery);
1647 ubifs_assert(!(c->vfs_sb->s_flags & MS_RDONLY));
1649 mutex_lock(&c->umount_mutex);
1651 kthread_stop(c->bgt);
1655 dbg_save_space_info(c);
1657 for (i = 0; i < c->jhead_cnt; i++) {
1658 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1659 del_timer_sync(&c->jheads[i].wbuf.timer);
1662 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1663 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1664 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1665 err = ubifs_write_master(c);
1667 ubifs_ro_mode(c, err);
1674 ubifs_lpt_free(c, 1);
1675 err = dbg_check_space_info(c);
1677 ubifs_ro_mode(c, err);
1678 mutex_unlock(&c->umount_mutex);
1681 static void ubifs_put_super(struct super_block *sb)
1684 struct ubifs_info *c = sb->s_fs_info;
1686 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1692 * The following asserts are only valid if there has not been a failure
1693 * of the media. For example, there will be dirty inodes if we failed
1694 * to write them back because of I/O errors.
1696 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1697 ubifs_assert(c->budg_idx_growth == 0);
1698 ubifs_assert(c->budg_dd_growth == 0);
1699 ubifs_assert(c->budg_data_growth == 0);
1702 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1703 * and file system un-mount. Namely, it prevents the shrinker from
1704 * picking this superblock for shrinking - it will be just skipped if
1705 * the mutex is locked.
1707 mutex_lock(&c->umount_mutex);
1708 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1710 * First of all kill the background thread to make sure it does
1711 * not interfere with un-mounting and freeing resources.
1714 kthread_stop(c->bgt);
1718 /* Synchronize write-buffers */
1720 for (i = 0; i < c->jhead_cnt; i++) {
1721 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1722 del_timer_sync(&c->jheads[i].wbuf.timer);
1726 * On fatal errors c->ro_media is set to 1, in which case we do
1727 * not write the master node.
1731 * We are being cleanly unmounted which means the
1732 * orphans were killed - indicate this in the master
1733 * node. Also save the reserved GC LEB number.
1737 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1738 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1739 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1740 err = ubifs_write_master(c);
1743 * Recovery will attempt to fix the master area
1744 * next mount, so we just print a message and
1745 * continue to unmount normally.
1747 ubifs_err("failed to write master node, "
1753 bdi_destroy(&c->bdi);
1754 ubi_close_volume(c->ubi);
1755 mutex_unlock(&c->umount_mutex);
1761 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1764 struct ubifs_info *c = sb->s_fs_info;
1766 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1768 err = ubifs_parse_options(c, data, 1);
1770 ubifs_err("invalid or unknown remount parameter");
1775 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1777 ubifs_msg("cannot re-mount due to prior errors");
1781 err = ubifs_remount_rw(c);
1786 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY)) {
1788 ubifs_msg("cannot re-mount due to prior errors");
1792 ubifs_remount_ro(c);
1795 if (c->bulk_read == 1)
1798 dbg_gen("disable bulk-read");
1803 ubifs_assert(c->lst.taken_empty_lebs > 0);
1808 const struct super_operations ubifs_super_operations = {
1809 .alloc_inode = ubifs_alloc_inode,
1810 .destroy_inode = ubifs_destroy_inode,
1811 .put_super = ubifs_put_super,
1812 .write_inode = ubifs_write_inode,
1813 .delete_inode = ubifs_delete_inode,
1814 .statfs = ubifs_statfs,
1815 .dirty_inode = ubifs_dirty_inode,
1816 .remount_fs = ubifs_remount_fs,
1817 .show_options = ubifs_show_options,
1818 .sync_fs = ubifs_sync_fs,
1822 * open_ubi - parse UBI device name string and open the UBI device.
1823 * @name: UBI volume name
1824 * @mode: UBI volume open mode
1826 * There are several ways to specify UBI volumes when mounting UBIFS:
1827 * o ubiX_Y - UBI device number X, volume Y;
1828 * o ubiY - UBI device number 0, volume Y;
1829 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1830 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1832 * Alternative '!' separator may be used instead of ':' (because some shells
1833 * like busybox may interpret ':' as an NFS host name separator). This function
1834 * returns ubi volume object in case of success and a negative error code in
1837 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1842 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1843 return ERR_PTR(-EINVAL);
1845 /* ubi:NAME method */
1846 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1847 return ubi_open_volume_nm(0, name + 4, mode);
1849 if (!isdigit(name[3]))
1850 return ERR_PTR(-EINVAL);
1852 dev = simple_strtoul(name + 3, &endptr, 0);
1855 if (*endptr == '\0')
1856 return ubi_open_volume(0, dev, mode);
1859 if (*endptr == '_' && isdigit(endptr[1])) {
1860 vol = simple_strtoul(endptr + 1, &endptr, 0);
1861 if (*endptr != '\0')
1862 return ERR_PTR(-EINVAL);
1863 return ubi_open_volume(dev, vol, mode);
1866 /* ubiX:NAME method */
1867 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1868 return ubi_open_volume_nm(dev, ++endptr, mode);
1870 return ERR_PTR(-EINVAL);
1873 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1875 struct ubi_volume_desc *ubi = sb->s_fs_info;
1876 struct ubifs_info *c;
1880 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1884 spin_lock_init(&c->cnt_lock);
1885 spin_lock_init(&c->cs_lock);
1886 spin_lock_init(&c->buds_lock);
1887 spin_lock_init(&c->space_lock);
1888 spin_lock_init(&c->orphan_lock);
1889 init_rwsem(&c->commit_sem);
1890 mutex_init(&c->lp_mutex);
1891 mutex_init(&c->tnc_mutex);
1892 mutex_init(&c->log_mutex);
1893 mutex_init(&c->mst_mutex);
1894 mutex_init(&c->umount_mutex);
1895 mutex_init(&c->bu_mutex);
1896 init_waitqueue_head(&c->cmt_wq);
1898 c->old_idx = RB_ROOT;
1899 c->size_tree = RB_ROOT;
1900 c->orph_tree = RB_ROOT;
1901 INIT_LIST_HEAD(&c->infos_list);
1902 INIT_LIST_HEAD(&c->idx_gc);
1903 INIT_LIST_HEAD(&c->replay_list);
1904 INIT_LIST_HEAD(&c->replay_buds);
1905 INIT_LIST_HEAD(&c->uncat_list);
1906 INIT_LIST_HEAD(&c->empty_list);
1907 INIT_LIST_HEAD(&c->freeable_list);
1908 INIT_LIST_HEAD(&c->frdi_idx_list);
1909 INIT_LIST_HEAD(&c->unclean_leb_list);
1910 INIT_LIST_HEAD(&c->old_buds);
1911 INIT_LIST_HEAD(&c->orph_list);
1912 INIT_LIST_HEAD(&c->orph_new);
1914 c->highest_inum = UBIFS_FIRST_INO;
1915 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1917 ubi_get_volume_info(ubi, &c->vi);
1918 ubi_get_device_info(c->vi.ubi_num, &c->di);
1920 /* Re-open the UBI device in read-write mode */
1921 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1922 if (IS_ERR(c->ubi)) {
1923 err = PTR_ERR(c->ubi);
1928 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1929 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1930 * which means the user would have to wait not just for their own I/O
1931 * but the read-ahead I/O as well i.e. completely pointless.
1933 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1935 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1936 c->bdi.unplug_io_fn = default_unplug_io_fn;
1937 err = bdi_init(&c->bdi);
1940 err = bdi_register(&c->bdi, NULL, "ubifs");
1944 err = ubifs_parse_options(c, data, 0);
1951 sb->s_magic = UBIFS_SUPER_MAGIC;
1952 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1953 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1954 sb->s_dev = c->vi.cdev;
1955 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1956 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1957 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1958 sb->s_op = &ubifs_super_operations;
1960 mutex_lock(&c->umount_mutex);
1961 err = mount_ubifs(c);
1963 ubifs_assert(err < 0);
1967 /* Read the root inode */
1968 root = ubifs_iget(sb, UBIFS_ROOT_INO);
1970 err = PTR_ERR(root);
1974 sb->s_root = d_alloc_root(root);
1978 mutex_unlock(&c->umount_mutex);
1986 mutex_unlock(&c->umount_mutex);
1988 bdi_destroy(&c->bdi);
1990 ubi_close_volume(c->ubi);
1996 static int sb_test(struct super_block *sb, void *data)
2000 return sb->s_dev == *dev;
2003 static int sb_set(struct super_block *sb, void *data)
2011 static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
2012 const char *name, void *data, struct vfsmount *mnt)
2014 struct ubi_volume_desc *ubi;
2015 struct ubi_volume_info vi;
2016 struct super_block *sb;
2019 dbg_gen("name %s, flags %#x", name, flags);
2022 * Get UBI device number and volume ID. Mount it read-only so far
2023 * because this might be a new mount point, and UBI allows only one
2024 * read-write user at a time.
2026 ubi = open_ubi(name, UBI_READONLY);
2028 ubifs_err("cannot open \"%s\", error %d",
2029 name, (int)PTR_ERR(ubi));
2030 return PTR_ERR(ubi);
2032 ubi_get_volume_info(ubi, &vi);
2034 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
2036 sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
2043 /* A new mount point for already mounted UBIFS */
2044 dbg_gen("this ubi volume is already mounted");
2045 if ((flags ^ sb->s_flags) & MS_RDONLY) {
2050 sb->s_flags = flags;
2052 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
2055 sb->s_fs_info = ubi;
2056 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2059 /* We do not support atime */
2060 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2063 /* 'fill_super()' opens ubi again so we must close it here */
2064 ubi_close_volume(ubi);
2066 simple_set_mnt(mnt, sb);
2070 deactivate_locked_super(sb);
2072 ubi_close_volume(ubi);
2076 static void ubifs_kill_sb(struct super_block *sb)
2078 generic_shutdown_super(sb);
2081 static struct file_system_type ubifs_fs_type = {
2083 .owner = THIS_MODULE,
2084 .get_sb = ubifs_get_sb,
2085 .kill_sb = ubifs_kill_sb
2089 * Inode slab cache constructor.
2091 static void inode_slab_ctor(void *obj)
2093 struct ubifs_inode *ui = obj;
2094 inode_init_once(&ui->vfs_inode);
2097 static int __init ubifs_init(void)
2101 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2103 /* Make sure node sizes are 8-byte aligned */
2104 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2105 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2106 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2107 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2108 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2109 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2110 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2111 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2112 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2113 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2114 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2116 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2117 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2118 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2119 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2120 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2121 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2123 /* Check min. node size */
2124 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2125 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2126 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2127 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2129 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2130 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2131 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2132 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2134 /* Defined node sizes */
2135 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2136 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2137 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2138 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2141 * We use 2 bit wide bit-fields to store compression type, which should
2142 * be amended if more compressors are added. The bit-fields are:
2143 * @compr_type in 'struct ubifs_inode', @default_compr in
2144 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2146 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2149 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2150 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2152 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2153 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2154 " at least 4096 bytes",
2155 (unsigned int)PAGE_CACHE_SIZE);
2159 err = register_filesystem(&ubifs_fs_type);
2161 ubifs_err("cannot register file system, error %d", err);
2166 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2167 sizeof(struct ubifs_inode), 0,
2168 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2170 if (!ubifs_inode_slab)
2173 register_shrinker(&ubifs_shrinker_info);
2175 err = ubifs_compressors_init();
2179 err = dbg_debugfs_init();
2186 ubifs_compressors_exit();
2188 unregister_shrinker(&ubifs_shrinker_info);
2189 kmem_cache_destroy(ubifs_inode_slab);
2191 unregister_filesystem(&ubifs_fs_type);
2194 /* late_initcall to let compressors initialize first */
2195 late_initcall(ubifs_init);
2197 static void __exit ubifs_exit(void)
2199 ubifs_assert(list_empty(&ubifs_infos));
2200 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2203 ubifs_compressors_exit();
2204 unregister_shrinker(&ubifs_shrinker_info);
2205 kmem_cache_destroy(ubifs_inode_slab);
2206 unregister_filesystem(&ubifs_fs_type);
2208 module_exit(ubifs_exit);
2210 MODULE_LICENSE("GPL");
2211 MODULE_VERSION(__stringify(UBIFS_VERSION));
2212 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2213 MODULE_DESCRIPTION("UBIFS - UBI File System");