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 (ŠŠøŃŃŃŠŗŠøŠ¹ ŠŃŃŃŠ¼)
24
* This file implements UBIFS initialization and VFS superblock operations. Some
25
* initialization stuff which is rather large and complex is placed at
26
* corresponding subsystems, but most of it is here.
29
#include <linux/init.h>
30
#include <linux/slab.h>
31
#include <linux/module.h>
32
#include <linux/ctype.h>
33
#include <linux/kthread.h>
34
#include <linux/parser.h>
35
#include <linux/seq_file.h>
36
#include <linux/mount.h>
37
#include <linux/math64.h>
38
#include <linux/writeback.h>
42
* Maximum amount of memory we may 'kmalloc()' without worrying that we are
43
* allocating too much.
45
#define UBIFS_KMALLOC_OK (128*1024)
47
/* Slab cache for UBIFS inodes */
48
struct kmem_cache *ubifs_inode_slab;
50
/* UBIFS TNC shrinker description */
51
static struct shrinker ubifs_shrinker_info = {
52
.shrink = ubifs_shrinker,
53
.seeks = DEFAULT_SEEKS,
57
* validate_inode - validate inode.
58
* @c: UBIFS file-system description object
59
* @inode: the inode to validate
61
* This is a helper function for 'ubifs_iget()' which validates various fields
62
* of a newly built inode to make sure they contain sane values and prevent
63
* possible vulnerabilities. Returns zero if the inode is all right and
64
* a non-zero error code if not.
66
static int validate_inode(struct ubifs_info *c, const struct inode *inode)
69
const struct ubifs_inode *ui = ubifs_inode(inode);
71
if (inode->i_size > c->max_inode_sz) {
72
ubifs_err("inode is too large (%lld)",
73
(long long)inode->i_size);
77
if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
78
ubifs_err("unknown compression type %d", ui->compr_type);
82
if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
85
if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
88
if (ui->xattr && !S_ISREG(inode->i_mode))
91
if (!ubifs_compr_present(ui->compr_type)) {
92
ubifs_warn("inode %lu uses '%s' compression, but it was not "
93
"compiled in", inode->i_ino,
94
ubifs_compr_name(ui->compr_type));
97
err = dbg_check_dir(c, inode);
101
struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
105
struct ubifs_ino_node *ino;
106
struct ubifs_info *c = sb->s_fs_info;
108
struct ubifs_inode *ui;
110
dbg_gen("inode %lu", inum);
112
inode = iget_locked(sb, inum);
114
return ERR_PTR(-ENOMEM);
115
if (!(inode->i_state & I_NEW))
117
ui = ubifs_inode(inode);
119
ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
125
ino_key_init(c, &key, inode->i_ino);
127
err = ubifs_tnc_lookup(c, &key, ino);
131
inode->i_flags |= (S_NOCMTIME | S_NOATIME);
132
set_nlink(inode, le32_to_cpu(ino->nlink));
133
inode->i_uid = le32_to_cpu(ino->uid);
134
inode->i_gid = le32_to_cpu(ino->gid);
135
inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
136
inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
137
inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
138
inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
139
inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
140
inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
141
inode->i_mode = le32_to_cpu(ino->mode);
142
inode->i_size = le64_to_cpu(ino->size);
144
ui->data_len = le32_to_cpu(ino->data_len);
145
ui->flags = le32_to_cpu(ino->flags);
146
ui->compr_type = le16_to_cpu(ino->compr_type);
147
ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
148
ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
149
ui->xattr_size = le32_to_cpu(ino->xattr_size);
150
ui->xattr_names = le32_to_cpu(ino->xattr_names);
151
ui->synced_i_size = ui->ui_size = inode->i_size;
153
ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
155
err = validate_inode(c, inode);
159
/* Disable read-ahead */
160
inode->i_mapping->backing_dev_info = &c->bdi;
162
switch (inode->i_mode & S_IFMT) {
164
inode->i_mapping->a_ops = &ubifs_file_address_operations;
165
inode->i_op = &ubifs_file_inode_operations;
166
inode->i_fop = &ubifs_file_operations;
168
ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
173
memcpy(ui->data, ino->data, ui->data_len);
174
((char *)ui->data)[ui->data_len] = '\0';
175
} else if (ui->data_len != 0) {
181
inode->i_op = &ubifs_dir_inode_operations;
182
inode->i_fop = &ubifs_dir_operations;
183
if (ui->data_len != 0) {
189
inode->i_op = &ubifs_symlink_inode_operations;
190
if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
194
ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
199
memcpy(ui->data, ino->data, ui->data_len);
200
((char *)ui->data)[ui->data_len] = '\0';
206
union ubifs_dev_desc *dev;
208
ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
214
dev = (union ubifs_dev_desc *)ino->data;
215
if (ui->data_len == sizeof(dev->new))
216
rdev = new_decode_dev(le32_to_cpu(dev->new));
217
else if (ui->data_len == sizeof(dev->huge))
218
rdev = huge_decode_dev(le64_to_cpu(dev->huge));
223
memcpy(ui->data, ino->data, ui->data_len);
224
inode->i_op = &ubifs_file_inode_operations;
225
init_special_inode(inode, inode->i_mode, rdev);
230
inode->i_op = &ubifs_file_inode_operations;
231
init_special_inode(inode, inode->i_mode, 0);
232
if (ui->data_len != 0) {
243
ubifs_set_inode_flags(inode);
244
unlock_new_inode(inode);
248
ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
249
dbg_dump_node(c, ino);
250
dbg_dump_inode(c, inode);
255
ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
260
static struct inode *ubifs_alloc_inode(struct super_block *sb)
262
struct ubifs_inode *ui;
264
ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
268
memset((void *)ui + sizeof(struct inode), 0,
269
sizeof(struct ubifs_inode) - sizeof(struct inode));
270
mutex_init(&ui->ui_mutex);
271
spin_lock_init(&ui->ui_lock);
272
return &ui->vfs_inode;
275
static void ubifs_i_callback(struct rcu_head *head)
277
struct inode *inode = container_of(head, struct inode, i_rcu);
278
struct ubifs_inode *ui = ubifs_inode(inode);
279
INIT_LIST_HEAD(&inode->i_dentry);
280
kmem_cache_free(ubifs_inode_slab, ui);
283
static void ubifs_destroy_inode(struct inode *inode)
285
struct ubifs_inode *ui = ubifs_inode(inode);
288
call_rcu(&inode->i_rcu, ubifs_i_callback);
292
* Note, Linux write-back code calls this without 'i_mutex'.
294
static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
297
struct ubifs_info *c = inode->i_sb->s_fs_info;
298
struct ubifs_inode *ui = ubifs_inode(inode);
300
ubifs_assert(!ui->xattr);
301
if (is_bad_inode(inode))
304
mutex_lock(&ui->ui_mutex);
306
* Due to races between write-back forced by budgeting
307
* (see 'sync_some_inodes()') and pdflush write-back, the inode may
308
* have already been synchronized, do not do this again. This might
309
* also happen if it was synchronized in an VFS operation, e.g.
313
mutex_unlock(&ui->ui_mutex);
318
* As an optimization, do not write orphan inodes to the media just
319
* because this is not needed.
321
dbg_gen("inode %lu, mode %#x, nlink %u",
322
inode->i_ino, (int)inode->i_mode, inode->i_nlink);
323
if (inode->i_nlink) {
324
err = ubifs_jnl_write_inode(c, inode);
326
ubifs_err("can't write inode %lu, error %d",
329
err = dbg_check_inode_size(c, inode, ui->ui_size);
333
mutex_unlock(&ui->ui_mutex);
334
ubifs_release_dirty_inode_budget(c, ui);
338
static void ubifs_evict_inode(struct inode *inode)
341
struct ubifs_info *c = inode->i_sb->s_fs_info;
342
struct ubifs_inode *ui = ubifs_inode(inode);
346
* Extended attribute inode deletions are fully handled in
347
* 'ubifs_removexattr()'. These inodes are special and have
348
* limited usage, so there is nothing to do here.
352
dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
353
ubifs_assert(!atomic_read(&inode->i_count));
355
truncate_inode_pages(&inode->i_data, 0);
360
if (is_bad_inode(inode))
363
ui->ui_size = inode->i_size = 0;
364
err = ubifs_jnl_delete_inode(c, inode);
367
* Worst case we have a lost orphan inode wasting space, so a
368
* simple error message is OK here.
370
ubifs_err("can't delete inode %lu, error %d",
375
ubifs_release_dirty_inode_budget(c, ui);
377
/* We've deleted something - clean the "no space" flags */
378
c->bi.nospace = c->bi.nospace_rp = 0;
382
end_writeback(inode);
385
static void ubifs_dirty_inode(struct inode *inode, int flags)
387
struct ubifs_inode *ui = ubifs_inode(inode);
389
ubifs_assert(mutex_is_locked(&ui->ui_mutex));
392
dbg_gen("inode %lu", inode->i_ino);
396
static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
398
struct ubifs_info *c = dentry->d_sb->s_fs_info;
399
unsigned long long free;
400
__le32 *uuid = (__le32 *)c->uuid;
402
free = ubifs_get_free_space(c);
403
dbg_gen("free space %lld bytes (%lld blocks)",
404
free, free >> UBIFS_BLOCK_SHIFT);
406
buf->f_type = UBIFS_SUPER_MAGIC;
407
buf->f_bsize = UBIFS_BLOCK_SIZE;
408
buf->f_blocks = c->block_cnt;
409
buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
410
if (free > c->report_rp_size)
411
buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
416
buf->f_namelen = UBIFS_MAX_NLEN;
417
buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
418
buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
419
ubifs_assert(buf->f_bfree <= c->block_cnt);
423
static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
425
struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
427
if (c->mount_opts.unmount_mode == 2)
428
seq_printf(s, ",fast_unmount");
429
else if (c->mount_opts.unmount_mode == 1)
430
seq_printf(s, ",norm_unmount");
432
if (c->mount_opts.bulk_read == 2)
433
seq_printf(s, ",bulk_read");
434
else if (c->mount_opts.bulk_read == 1)
435
seq_printf(s, ",no_bulk_read");
437
if (c->mount_opts.chk_data_crc == 2)
438
seq_printf(s, ",chk_data_crc");
439
else if (c->mount_opts.chk_data_crc == 1)
440
seq_printf(s, ",no_chk_data_crc");
442
if (c->mount_opts.override_compr) {
443
seq_printf(s, ",compr=%s",
444
ubifs_compr_name(c->mount_opts.compr_type));
450
static int ubifs_sync_fs(struct super_block *sb, int wait)
453
struct ubifs_info *c = sb->s_fs_info;
456
* Zero @wait is just an advisory thing to help the file system shove
457
* lots of data into the queues, and there will be the second
458
* '->sync_fs()' call, with non-zero @wait.
464
* Synchronize write buffers, because 'ubifs_run_commit()' does not
465
* do this if it waits for an already running commit.
467
for (i = 0; i < c->jhead_cnt; i++) {
468
err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
474
* Strictly speaking, it is not necessary to commit the journal here,
475
* synchronizing write-buffers would be enough. But committing makes
476
* UBIFS free space predictions much more accurate, so we want to let
477
* the user be able to get more accurate results of 'statfs()' after
478
* they synchronize the file system.
480
err = ubifs_run_commit(c);
484
return ubi_sync(c->vi.ubi_num);
488
* init_constants_early - initialize UBIFS constants.
489
* @c: UBIFS file-system description object
491
* This function initialize UBIFS constants which do not need the superblock to
492
* be read. It also checks that the UBI volume satisfies basic UBIFS
493
* requirements. Returns zero in case of success and a negative error code in
496
static int init_constants_early(struct ubifs_info *c)
498
if (c->vi.corrupted) {
499
ubifs_warn("UBI volume is corrupted - read-only mode");
504
ubifs_msg("read-only UBI device");
508
if (c->vi.vol_type == UBI_STATIC_VOLUME) {
509
ubifs_msg("static UBI volume - read-only mode");
513
c->leb_cnt = c->vi.size;
514
c->leb_size = c->vi.usable_leb_size;
515
c->leb_start = c->di.leb_start;
516
c->half_leb_size = c->leb_size / 2;
517
c->min_io_size = c->di.min_io_size;
518
c->min_io_shift = fls(c->min_io_size) - 1;
519
c->max_write_size = c->di.max_write_size;
520
c->max_write_shift = fls(c->max_write_size) - 1;
522
if (c->leb_size < UBIFS_MIN_LEB_SZ) {
523
ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
524
c->leb_size, UBIFS_MIN_LEB_SZ);
528
if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
529
ubifs_err("too few LEBs (%d), min. is %d",
530
c->leb_cnt, UBIFS_MIN_LEB_CNT);
534
if (!is_power_of_2(c->min_io_size)) {
535
ubifs_err("bad min. I/O size %d", c->min_io_size);
540
* Maximum write size has to be greater or equivalent to min. I/O
541
* size, and be multiple of min. I/O size.
543
if (c->max_write_size < c->min_io_size ||
544
c->max_write_size % c->min_io_size ||
545
!is_power_of_2(c->max_write_size)) {
546
ubifs_err("bad write buffer size %d for %d min. I/O unit",
547
c->max_write_size, c->min_io_size);
552
* UBIFS aligns all node to 8-byte boundary, so to make function in
553
* io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
556
if (c->min_io_size < 8) {
559
if (c->max_write_size < c->min_io_size) {
560
c->max_write_size = c->min_io_size;
561
c->max_write_shift = c->min_io_shift;
565
c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
566
c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
569
* Initialize node length ranges which are mostly needed for node
572
c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
573
c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
574
c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
575
c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
576
c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
577
c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
579
c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
580
c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
581
c->ranges[UBIFS_ORPH_NODE].min_len =
582
UBIFS_ORPH_NODE_SZ + sizeof(__le64);
583
c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
584
c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
585
c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
586
c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
587
c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
588
c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
589
c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
591
* Minimum indexing node size is amended later when superblock is
592
* read and the key length is known.
594
c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
596
* Maximum indexing node size is amended later when superblock is
597
* read and the fanout is known.
599
c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
602
* Initialize dead and dark LEB space watermarks. See gc.c for comments
603
* about these values.
605
c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
606
c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
609
* Calculate how many bytes would be wasted at the end of LEB if it was
610
* fully filled with data nodes of maximum size. This is used in
611
* calculations when reporting free space.
613
c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
615
/* Buffer size for bulk-reads */
616
c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
617
if (c->max_bu_buf_len > c->leb_size)
618
c->max_bu_buf_len = c->leb_size;
623
* bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
624
* @c: UBIFS file-system description object
625
* @lnum: LEB the write-buffer was synchronized to
626
* @free: how many free bytes left in this LEB
627
* @pad: how many bytes were padded
629
* This is a callback function which is called by the I/O unit when the
630
* write-buffer is synchronized. We need this to correctly maintain space
631
* accounting in bud logical eraseblocks. This function returns zero in case of
632
* success and a negative error code in case of failure.
634
* This function actually belongs to the journal, but we keep it here because
635
* we want to keep it static.
637
static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
639
return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
643
* init_constants_sb - initialize UBIFS constants.
644
* @c: UBIFS file-system description object
646
* This is a helper function which initializes various UBIFS constants after
647
* the superblock has been read. It also checks various UBIFS parameters and
648
* makes sure they are all right. Returns zero in case of success and a
649
* negative error code in case of failure.
651
static int init_constants_sb(struct ubifs_info *c)
656
c->main_bytes = (long long)c->main_lebs * c->leb_size;
657
c->max_znode_sz = sizeof(struct ubifs_znode) +
658
c->fanout * sizeof(struct ubifs_zbranch);
660
tmp = ubifs_idx_node_sz(c, 1);
661
c->ranges[UBIFS_IDX_NODE].min_len = tmp;
662
c->min_idx_node_sz = ALIGN(tmp, 8);
664
tmp = ubifs_idx_node_sz(c, c->fanout);
665
c->ranges[UBIFS_IDX_NODE].max_len = tmp;
666
c->max_idx_node_sz = ALIGN(tmp, 8);
668
/* Make sure LEB size is large enough to fit full commit */
669
tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
670
tmp = ALIGN(tmp, c->min_io_size);
671
if (tmp > c->leb_size) {
672
dbg_err("too small LEB size %d, at least %d needed",
678
* Make sure that the log is large enough to fit reference nodes for
679
* all buds plus one reserved LEB.
681
tmp64 = c->max_bud_bytes + c->leb_size - 1;
682
c->max_bud_cnt = div_u64(tmp64, c->leb_size);
683
tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
686
if (c->log_lebs < tmp) {
687
dbg_err("too small log %d LEBs, required min. %d LEBs",
693
* When budgeting we assume worst-case scenarios when the pages are not
694
* be compressed and direntries are of the maximum size.
696
* Note, data, which may be stored in inodes is budgeted separately, so
697
* it is not included into 'c->bi.inode_budget'.
699
c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
700
c->bi.inode_budget = UBIFS_INO_NODE_SZ;
701
c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
704
* When the amount of flash space used by buds becomes
705
* 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
706
* The writers are unblocked when the commit is finished. To avoid
707
* writers to be blocked UBIFS initiates background commit in advance,
708
* when number of bud bytes becomes above the limit defined below.
710
c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
713
* Ensure minimum journal size. All the bytes in the journal heads are
714
* considered to be used, when calculating the current journal usage.
715
* Consequently, if the journal is too small, UBIFS will treat it as
718
tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1;
719
if (c->bg_bud_bytes < tmp64)
720
c->bg_bud_bytes = tmp64;
721
if (c->max_bud_bytes < tmp64 + c->leb_size)
722
c->max_bud_bytes = tmp64 + c->leb_size;
724
err = ubifs_calc_lpt_geom(c);
728
/* Initialize effective LEB size used in budgeting calculations */
729
c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
734
* init_constants_master - initialize UBIFS constants.
735
* @c: UBIFS file-system description object
737
* This is a helper function which initializes various UBIFS constants after
738
* the master node has been read. It also checks various UBIFS parameters and
739
* makes sure they are all right.
741
static void init_constants_master(struct ubifs_info *c)
745
c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
746
c->report_rp_size = ubifs_reported_space(c, c->rp_size);
749
* Calculate total amount of FS blocks. This number is not used
750
* internally because it does not make much sense for UBIFS, but it is
751
* necessary to report something for the 'statfs()' call.
753
* Subtract the LEB reserved for GC, the LEB which is reserved for
754
* deletions, minimum LEBs for the index, and assume only one journal
757
tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1;
758
tmp64 *= (long long)c->leb_size - c->leb_overhead;
759
tmp64 = ubifs_reported_space(c, tmp64);
760
c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
764
* take_gc_lnum - reserve GC LEB.
765
* @c: UBIFS file-system description object
767
* This function ensures that the LEB reserved for garbage collection is marked
768
* as "taken" in lprops. We also have to set free space to LEB size and dirty
769
* space to zero, because lprops may contain out-of-date information if the
770
* file-system was un-mounted before it has been committed. This function
771
* returns zero in case of success and a negative error code in case of
774
static int take_gc_lnum(struct ubifs_info *c)
778
if (c->gc_lnum == -1) {
779
ubifs_err("no LEB for GC");
783
/* And we have to tell lprops that this LEB is taken */
784
err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
790
* alloc_wbufs - allocate write-buffers.
791
* @c: UBIFS file-system description object
793
* This helper function allocates and initializes UBIFS write-buffers. Returns
794
* zero in case of success and %-ENOMEM in case of failure.
796
static int alloc_wbufs(struct ubifs_info *c)
800
c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
805
/* Initialize journal heads */
806
for (i = 0; i < c->jhead_cnt; i++) {
807
INIT_LIST_HEAD(&c->jheads[i].buds_list);
808
err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
812
c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
813
c->jheads[i].wbuf.jhead = i;
814
c->jheads[i].grouped = 1;
817
c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
819
* Garbage Collector head likely contains long-term data and
820
* does not need to be synchronized by timer. Also GC head nodes are
823
c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
824
c->jheads[GCHD].wbuf.no_timer = 1;
825
c->jheads[GCHD].grouped = 0;
831
* free_wbufs - free write-buffers.
832
* @c: UBIFS file-system description object
834
static void free_wbufs(struct ubifs_info *c)
839
for (i = 0; i < c->jhead_cnt; i++) {
840
kfree(c->jheads[i].wbuf.buf);
841
kfree(c->jheads[i].wbuf.inodes);
849
* free_orphans - free orphans.
850
* @c: UBIFS file-system description object
852
static void free_orphans(struct ubifs_info *c)
854
struct ubifs_orphan *orph;
856
while (c->orph_dnext) {
857
orph = c->orph_dnext;
858
c->orph_dnext = orph->dnext;
859
list_del(&orph->list);
863
while (!list_empty(&c->orph_list)) {
864
orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
865
list_del(&orph->list);
867
dbg_err("orphan list not empty at unmount");
875
* free_buds - free per-bud objects.
876
* @c: UBIFS file-system description object
878
static void free_buds(struct ubifs_info *c)
880
struct rb_node *this = c->buds.rb_node;
881
struct ubifs_bud *bud;
885
this = this->rb_left;
886
else if (this->rb_right)
887
this = this->rb_right;
889
bud = rb_entry(this, struct ubifs_bud, rb);
890
this = rb_parent(this);
892
if (this->rb_left == &bud->rb)
893
this->rb_left = NULL;
895
this->rb_right = NULL;
903
* check_volume_empty - check if the UBI volume is empty.
904
* @c: UBIFS file-system description object
906
* This function checks if the UBIFS volume is empty by looking if its LEBs are
907
* mapped or not. The result of checking is stored in the @c->empty variable.
908
* Returns zero in case of success and a negative error code in case of
911
static int check_volume_empty(struct ubifs_info *c)
916
for (lnum = 0; lnum < c->leb_cnt; lnum++) {
917
err = ubifs_is_mapped(c, lnum);
918
if (unlikely(err < 0))
932
* UBIFS mount options.
934
* Opt_fast_unmount: do not run a journal commit before un-mounting
935
* Opt_norm_unmount: run a journal commit before un-mounting
936
* Opt_bulk_read: enable bulk-reads
937
* Opt_no_bulk_read: disable bulk-reads
938
* Opt_chk_data_crc: check CRCs when reading data nodes
939
* Opt_no_chk_data_crc: do not check CRCs when reading data nodes
940
* Opt_override_compr: override default compressor
941
* Opt_err: just end of array marker
954
static const match_table_t tokens = {
955
{Opt_fast_unmount, "fast_unmount"},
956
{Opt_norm_unmount, "norm_unmount"},
957
{Opt_bulk_read, "bulk_read"},
958
{Opt_no_bulk_read, "no_bulk_read"},
959
{Opt_chk_data_crc, "chk_data_crc"},
960
{Opt_no_chk_data_crc, "no_chk_data_crc"},
961
{Opt_override_compr, "compr=%s"},
966
* parse_standard_option - parse a standard mount option.
967
* @option: the option to parse
969
* Normally, standard mount options like "sync" are passed to file-systems as
970
* flags. However, when a "rootflags=" kernel boot parameter is used, they may
971
* be present in the options string. This function tries to deal with this
972
* situation and parse standard options. Returns 0 if the option was not
973
* recognized, and the corresponding integer flag if it was.
975
* UBIFS is only interested in the "sync" option, so do not check for anything
978
static int parse_standard_option(const char *option)
980
ubifs_msg("parse %s", option);
981
if (!strcmp(option, "sync"))
982
return MS_SYNCHRONOUS;
987
* ubifs_parse_options - parse mount parameters.
988
* @c: UBIFS file-system description object
989
* @options: parameters to parse
990
* @is_remount: non-zero if this is FS re-mount
992
* This function parses UBIFS mount options and returns zero in case success
993
* and a negative error code in case of failure.
995
static int ubifs_parse_options(struct ubifs_info *c, char *options,
999
substring_t args[MAX_OPT_ARGS];
1004
while ((p = strsep(&options, ","))) {
1010
token = match_token(p, tokens, args);
1013
* %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
1014
* We accept them in order to be backward-compatible. But this
1015
* should be removed at some point.
1017
case Opt_fast_unmount:
1018
c->mount_opts.unmount_mode = 2;
1020
case Opt_norm_unmount:
1021
c->mount_opts.unmount_mode = 1;
1024
c->mount_opts.bulk_read = 2;
1027
case Opt_no_bulk_read:
1028
c->mount_opts.bulk_read = 1;
1031
case Opt_chk_data_crc:
1032
c->mount_opts.chk_data_crc = 2;
1033
c->no_chk_data_crc = 0;
1035
case Opt_no_chk_data_crc:
1036
c->mount_opts.chk_data_crc = 1;
1037
c->no_chk_data_crc = 1;
1039
case Opt_override_compr:
1041
char *name = match_strdup(&args[0]);
1045
if (!strcmp(name, "none"))
1046
c->mount_opts.compr_type = UBIFS_COMPR_NONE;
1047
else if (!strcmp(name, "lzo"))
1048
c->mount_opts.compr_type = UBIFS_COMPR_LZO;
1049
else if (!strcmp(name, "zlib"))
1050
c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
1052
ubifs_err("unknown compressor \"%s\"", name);
1057
c->mount_opts.override_compr = 1;
1058
c->default_compr = c->mount_opts.compr_type;
1064
struct super_block *sb = c->vfs_sb;
1066
flag = parse_standard_option(p);
1068
ubifs_err("unrecognized mount option \"%s\" "
1069
"or missing value", p);
1072
sb->s_flags |= flag;
1082
* destroy_journal - destroy journal data structures.
1083
* @c: UBIFS file-system description object
1085
* This function destroys journal data structures including those that may have
1086
* been created by recovery functions.
1088
static void destroy_journal(struct ubifs_info *c)
1090
while (!list_empty(&c->unclean_leb_list)) {
1091
struct ubifs_unclean_leb *ucleb;
1093
ucleb = list_entry(c->unclean_leb_list.next,
1094
struct ubifs_unclean_leb, list);
1095
list_del(&ucleb->list);
1098
while (!list_empty(&c->old_buds)) {
1099
struct ubifs_bud *bud;
1101
bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
1102
list_del(&bud->list);
1105
ubifs_destroy_idx_gc(c);
1106
ubifs_destroy_size_tree(c);
1112
* bu_init - initialize bulk-read information.
1113
* @c: UBIFS file-system description object
1115
static void bu_init(struct ubifs_info *c)
1117
ubifs_assert(c->bulk_read == 1);
1120
return; /* Already initialized */
1123
c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
1125
if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
1126
c->max_bu_buf_len = UBIFS_KMALLOC_OK;
1130
/* Just disable bulk-read */
1131
ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1132
"disabling it", c->max_bu_buf_len);
1133
c->mount_opts.bulk_read = 1;
1140
* check_free_space - check if there is enough free space to mount.
1141
* @c: UBIFS file-system description object
1143
* This function makes sure UBIFS has enough free space to be mounted in
1144
* read/write mode. UBIFS must always have some free space to allow deletions.
1146
static int check_free_space(struct ubifs_info *c)
1148
ubifs_assert(c->dark_wm > 0);
1149
if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
1150
ubifs_err("insufficient free space to mount in R/W mode");
1151
dbg_dump_budg(c, &c->bi);
1159
* mount_ubifs - mount UBIFS file-system.
1160
* @c: UBIFS file-system description object
1162
* This function mounts UBIFS file system. Returns zero in case of success and
1163
* a negative error code in case of failure.
1165
* Note, the function does not de-allocate resources it it fails half way
1166
* through, and the caller has to do this instead.
1168
static int mount_ubifs(struct ubifs_info *c)
1174
c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
1175
err = init_constants_early(c);
1179
err = ubifs_debugging_init(c);
1183
err = check_volume_empty(c);
1187
if (c->empty && (c->ro_mount || c->ro_media)) {
1189
* This UBI volume is empty, and read-only, or the file system
1190
* is mounted read-only - we cannot format it.
1192
ubifs_err("can't format empty UBI volume: read-only %s",
1193
c->ro_media ? "UBI volume" : "mount");
1198
if (c->ro_media && !c->ro_mount) {
1199
ubifs_err("cannot mount read-write - read-only media");
1205
* The requirement for the buffer is that it should fit indexing B-tree
1206
* height amount of integers. We assume the height if the TNC tree will
1210
c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
1211
if (!c->bottom_up_buf)
1214
c->sbuf = vmalloc(c->leb_size);
1219
c->ileb_buf = vmalloc(c->leb_size);
1224
if (c->bulk_read == 1)
1228
c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
1230
if (!c->write_reserve_buf)
1236
err = ubifs_read_superblock(c);
1241
* Make sure the compressor which is set as default in the superblock
1242
* or overridden by mount options is actually compiled in.
1244
if (!ubifs_compr_present(c->default_compr)) {
1245
ubifs_err("'compressor \"%s\" is not compiled in",
1246
ubifs_compr_name(c->default_compr));
1251
err = init_constants_sb(c);
1255
sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1256
sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1257
c->cbuf = kmalloc(sz, GFP_NOFS);
1263
err = alloc_wbufs(c);
1267
sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
1269
/* Create background thread */
1270
c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1271
if (IS_ERR(c->bgt)) {
1272
err = PTR_ERR(c->bgt);
1274
ubifs_err("cannot spawn \"%s\", error %d",
1278
wake_up_process(c->bgt);
1281
err = ubifs_read_master(c);
1285
init_constants_master(c);
1287
if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1288
ubifs_msg("recovery needed");
1289
c->need_recovery = 1;
1292
if (c->need_recovery && !c->ro_mount) {
1293
err = ubifs_recover_inl_heads(c, c->sbuf);
1298
err = ubifs_lpt_init(c, 1, !c->ro_mount);
1302
if (!c->ro_mount && c->space_fixup) {
1303
err = ubifs_fixup_free_space(c);
1310
* Set the "dirty" flag so that if we reboot uncleanly we
1311
* will notice this immediately on the next mount.
1313
c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1314
err = ubifs_write_master(c);
1319
err = dbg_check_idx_size(c, c->bi.old_idx_sz);
1323
err = ubifs_replay_journal(c);
1327
/* Calculate 'min_idx_lebs' after journal replay */
1328
c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
1330
err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
1337
err = check_free_space(c);
1341
/* Check for enough log space */
1342
lnum = c->lhead_lnum + 1;
1343
if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1344
lnum = UBIFS_LOG_LNUM;
1345
if (lnum == c->ltail_lnum) {
1346
err = ubifs_consolidate_log(c);
1351
if (c->need_recovery) {
1352
err = ubifs_recover_size(c);
1355
err = ubifs_rcvry_gc_commit(c);
1359
err = take_gc_lnum(c);
1364
* GC LEB may contain garbage if there was an unclean
1365
* reboot, and it should be un-mapped.
1367
err = ubifs_leb_unmap(c, c->gc_lnum);
1372
err = dbg_check_lprops(c);
1375
} else if (c->need_recovery) {
1376
err = ubifs_recover_size(c);
1381
* Even if we mount read-only, we have to set space in GC LEB
1382
* to proper value because this affects UBIFS free space
1383
* reporting. We do not want to have a situation when
1384
* re-mounting from R/O to R/W changes amount of free space.
1386
err = take_gc_lnum(c);
1391
spin_lock(&ubifs_infos_lock);
1392
list_add_tail(&c->infos_list, &ubifs_infos);
1393
spin_unlock(&ubifs_infos_lock);
1395
if (c->need_recovery) {
1397
ubifs_msg("recovery deferred");
1399
c->need_recovery = 0;
1400
ubifs_msg("recovery completed");
1402
* GC LEB has to be empty and taken at this point. But
1403
* the journal head LEBs may also be accounted as
1404
* "empty taken" if they are empty.
1406
ubifs_assert(c->lst.taken_empty_lebs > 0);
1409
ubifs_assert(c->lst.taken_empty_lebs > 0);
1411
err = dbg_check_filesystem(c);
1415
err = dbg_debugfs_init_fs(c);
1421
ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1422
c->vi.ubi_num, c->vi.vol_id, c->vi.name);
1424
ubifs_msg("mounted read-only");
1425
x = (long long)c->main_lebs * c->leb_size;
1426
ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1427
"LEBs)", x, x >> 10, x >> 20, c->main_lebs);
1428
x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1429
ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1430
"LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1431
ubifs_msg("media format: w%d/r%d (latest is w%d/r%d)",
1432
c->fmt_version, c->ro_compat_version,
1433
UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
1434
ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1435
ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1436
c->report_rp_size, c->report_rp_size >> 10);
1438
dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1439
dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1440
dbg_msg("max. write size: %d bytes", c->max_write_size);
1441
dbg_msg("LEB size: %d bytes (%d KiB)",
1442
c->leb_size, c->leb_size >> 10);
1443
dbg_msg("data journal heads: %d",
1444
c->jhead_cnt - NONDATA_JHEADS_CNT);
1445
dbg_msg("UUID: %pUB", c->uuid);
1446
dbg_msg("big_lpt %d", c->big_lpt);
1447
dbg_msg("log LEBs: %d (%d - %d)",
1448
c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1449
dbg_msg("LPT area LEBs: %d (%d - %d)",
1450
c->lpt_lebs, c->lpt_first, c->lpt_last);
1451
dbg_msg("orphan area LEBs: %d (%d - %d)",
1452
c->orph_lebs, c->orph_first, c->orph_last);
1453
dbg_msg("main area LEBs: %d (%d - %d)",
1454
c->main_lebs, c->main_first, c->leb_cnt - 1);
1455
dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1456
dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1457
c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
1458
c->bi.old_idx_sz >> 20);
1459
dbg_msg("key hash type: %d", c->key_hash_type);
1460
dbg_msg("tree fanout: %d", c->fanout);
1461
dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1462
dbg_msg("first main LEB: %d", c->main_first);
1463
dbg_msg("max. znode size %d", c->max_znode_sz);
1464
dbg_msg("max. index node size %d", c->max_idx_node_sz);
1465
dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
1466
UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
1467
dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
1468
UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
1469
dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
1470
UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
1471
dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu, idx %d",
1472
UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
1473
UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
1474
dbg_msg("dead watermark: %d", c->dead_wm);
1475
dbg_msg("dark watermark: %d", c->dark_wm);
1476
dbg_msg("LEB overhead: %d", c->leb_overhead);
1477
x = (long long)c->main_lebs * c->dark_wm;
1478
dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1479
x, x >> 10, x >> 20);
1480
dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1481
c->max_bud_bytes, c->max_bud_bytes >> 10,
1482
c->max_bud_bytes >> 20);
1483
dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1484
c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1485
c->bg_bud_bytes >> 20);
1486
dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1487
c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1488
dbg_msg("max. seq. number: %llu", c->max_sqnum);
1489
dbg_msg("commit number: %llu", c->cmt_no);
1494
spin_lock(&ubifs_infos_lock);
1495
list_del(&c->infos_list);
1496
spin_unlock(&ubifs_infos_lock);
1502
ubifs_lpt_free(c, 0);
1505
kfree(c->rcvrd_mst_node);
1507
kthread_stop(c->bgt);
1513
kfree(c->write_reserve_buf);
1517
kfree(c->bottom_up_buf);
1518
ubifs_debugging_exit(c);
1523
* ubifs_umount - un-mount UBIFS file-system.
1524
* @c: UBIFS file-system description object
1526
* Note, this function is called to free allocated resourced when un-mounting,
1527
* as well as free resources when an error occurred while we were half way
1528
* through mounting (error path cleanup function). So it has to make sure the
1529
* resource was actually allocated before freeing it.
1531
static void ubifs_umount(struct ubifs_info *c)
1533
dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1536
dbg_debugfs_exit_fs(c);
1537
spin_lock(&ubifs_infos_lock);
1538
list_del(&c->infos_list);
1539
spin_unlock(&ubifs_infos_lock);
1542
kthread_stop(c->bgt);
1547
ubifs_lpt_free(c, 0);
1550
kfree(c->rcvrd_mst_node);
1552
kfree(c->write_reserve_buf);
1556
kfree(c->bottom_up_buf);
1557
ubifs_debugging_exit(c);
1561
* ubifs_remount_rw - re-mount in read-write mode.
1562
* @c: UBIFS file-system description object
1564
* UBIFS avoids allocating many unnecessary resources when mounted in read-only
1565
* mode. This function allocates the needed resources and re-mounts UBIFS in
1568
static int ubifs_remount_rw(struct ubifs_info *c)
1572
if (c->rw_incompat) {
1573
ubifs_err("the file-system is not R/W-compatible");
1574
ubifs_msg("on-flash format version is w%d/r%d, but software "
1575
"only supports up to version w%d/r%d", c->fmt_version,
1576
c->ro_compat_version, UBIFS_FORMAT_VERSION,
1577
UBIFS_RO_COMPAT_VERSION);
1581
mutex_lock(&c->umount_mutex);
1582
dbg_save_space_info(c);
1583
c->remounting_rw = 1;
1586
err = check_free_space(c);
1590
if (c->old_leb_cnt != c->leb_cnt) {
1591
struct ubifs_sb_node *sup;
1593
sup = ubifs_read_sb_node(c);
1598
sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1599
err = ubifs_write_sb_node(c, sup);
1605
if (c->need_recovery) {
1606
ubifs_msg("completing deferred recovery");
1607
err = ubifs_write_rcvrd_mst_node(c);
1610
err = ubifs_recover_size(c);
1613
err = ubifs_clean_lebs(c, c->sbuf);
1616
err = ubifs_recover_inl_heads(c, c->sbuf);
1620
/* A readonly mount is not allowed to have orphans */
1621
ubifs_assert(c->tot_orphans == 0);
1622
err = ubifs_clear_orphans(c);
1627
if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1628
c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1629
err = ubifs_write_master(c);
1634
c->ileb_buf = vmalloc(c->leb_size);
1640
c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
1641
if (!c->write_reserve_buf)
1644
err = ubifs_lpt_init(c, 0, 1);
1648
/* Create background thread */
1649
c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
1650
if (IS_ERR(c->bgt)) {
1651
err = PTR_ERR(c->bgt);
1653
ubifs_err("cannot spawn \"%s\", error %d",
1657
wake_up_process(c->bgt);
1659
c->orph_buf = vmalloc(c->leb_size);
1665
/* Check for enough log space */
1666
lnum = c->lhead_lnum + 1;
1667
if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1668
lnum = UBIFS_LOG_LNUM;
1669
if (lnum == c->ltail_lnum) {
1670
err = ubifs_consolidate_log(c);
1675
if (c->need_recovery)
1676
err = ubifs_rcvry_gc_commit(c);
1678
err = ubifs_leb_unmap(c, c->gc_lnum);
1682
dbg_gen("re-mounted read-write");
1683
c->remounting_rw = 0;
1685
if (c->need_recovery) {
1686
c->need_recovery = 0;
1687
ubifs_msg("deferred recovery completed");
1690
* Do not run the debugging space check if the were doing
1691
* recovery, because when we saved the information we had the
1692
* file-system in a state where the TNC and lprops has been
1693
* modified in memory, but all the I/O operations (including a
1694
* commit) were deferred. So the file-system was in
1695
* "non-committed" state. Now the file-system is in committed
1696
* state, and of course the amount of free space will change
1697
* because, for example, the old index size was imprecise.
1699
err = dbg_check_space_info(c);
1702
if (c->space_fixup) {
1703
err = ubifs_fixup_free_space(c);
1708
mutex_unlock(&c->umount_mutex);
1716
kthread_stop(c->bgt);
1720
kfree(c->write_reserve_buf);
1721
c->write_reserve_buf = NULL;
1724
ubifs_lpt_free(c, 1);
1725
c->remounting_rw = 0;
1726
mutex_unlock(&c->umount_mutex);
1731
* ubifs_remount_ro - re-mount in read-only mode.
1732
* @c: UBIFS file-system description object
1734
* We assume VFS has stopped writing. Possibly the background thread could be
1735
* running a commit, however kthread_stop will wait in that case.
1737
static void ubifs_remount_ro(struct ubifs_info *c)
1741
ubifs_assert(!c->need_recovery);
1742
ubifs_assert(!c->ro_mount);
1744
mutex_lock(&c->umount_mutex);
1746
kthread_stop(c->bgt);
1750
dbg_save_space_info(c);
1752
for (i = 0; i < c->jhead_cnt; i++)
1753
ubifs_wbuf_sync(&c->jheads[i].wbuf);
1755
c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1756
c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1757
c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1758
err = ubifs_write_master(c);
1760
ubifs_ro_mode(c, err);
1764
kfree(c->write_reserve_buf);
1765
c->write_reserve_buf = NULL;
1768
ubifs_lpt_free(c, 1);
1770
err = dbg_check_space_info(c);
1772
ubifs_ro_mode(c, err);
1773
mutex_unlock(&c->umount_mutex);
1776
static void ubifs_put_super(struct super_block *sb)
1779
struct ubifs_info *c = sb->s_fs_info;
1781
ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1785
* The following asserts are only valid if there has not been a failure
1786
* of the media. For example, there will be dirty inodes if we failed
1787
* to write them back because of I/O errors.
1790
ubifs_assert(c->bi.idx_growth == 0);
1791
ubifs_assert(c->bi.dd_growth == 0);
1792
ubifs_assert(c->bi.data_growth == 0);
1796
* The 'c->umount_lock' prevents races between UBIFS memory shrinker
1797
* and file system un-mount. Namely, it prevents the shrinker from
1798
* picking this superblock for shrinking - it will be just skipped if
1799
* the mutex is locked.
1801
mutex_lock(&c->umount_mutex);
1804
* First of all kill the background thread to make sure it does
1805
* not interfere with un-mounting and freeing resources.
1808
kthread_stop(c->bgt);
1813
* On fatal errors c->ro_error is set to 1, in which case we do
1814
* not write the master node.
1819
/* Synchronize write-buffers */
1820
for (i = 0; i < c->jhead_cnt; i++)
1821
ubifs_wbuf_sync(&c->jheads[i].wbuf);
1824
* We are being cleanly unmounted which means the
1825
* orphans were killed - indicate this in the master
1826
* node. Also save the reserved GC LEB number.
1828
c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1829
c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1830
c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1831
err = ubifs_write_master(c);
1834
* Recovery will attempt to fix the master area
1835
* next mount, so we just print a message and
1836
* continue to unmount normally.
1838
ubifs_err("failed to write master node, "
1841
for (i = 0; i < c->jhead_cnt; i++)
1842
/* Make sure write-buffer timers are canceled */
1843
hrtimer_cancel(&c->jheads[i].wbuf.timer);
1848
bdi_destroy(&c->bdi);
1849
ubi_close_volume(c->ubi);
1850
mutex_unlock(&c->umount_mutex);
1853
static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1856
struct ubifs_info *c = sb->s_fs_info;
1858
dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1860
err = ubifs_parse_options(c, data, 1);
1862
ubifs_err("invalid or unknown remount parameter");
1866
if (c->ro_mount && !(*flags & MS_RDONLY)) {
1868
ubifs_msg("cannot re-mount R/W due to prior errors");
1872
ubifs_msg("cannot re-mount R/W - UBI volume is R/O");
1875
err = ubifs_remount_rw(c);
1878
} else if (!c->ro_mount && (*flags & MS_RDONLY)) {
1880
ubifs_msg("cannot re-mount R/O due to prior errors");
1883
ubifs_remount_ro(c);
1886
if (c->bulk_read == 1)
1889
dbg_gen("disable bulk-read");
1894
ubifs_assert(c->lst.taken_empty_lebs > 0);
1898
const struct super_operations ubifs_super_operations = {
1899
.alloc_inode = ubifs_alloc_inode,
1900
.destroy_inode = ubifs_destroy_inode,
1901
.put_super = ubifs_put_super,
1902
.write_inode = ubifs_write_inode,
1903
.evict_inode = ubifs_evict_inode,
1904
.statfs = ubifs_statfs,
1905
.dirty_inode = ubifs_dirty_inode,
1906
.remount_fs = ubifs_remount_fs,
1907
.show_options = ubifs_show_options,
1908
.sync_fs = ubifs_sync_fs,
1912
* open_ubi - parse UBI device name string and open the UBI device.
1913
* @name: UBI volume name
1914
* @mode: UBI volume open mode
1916
* The primary method of mounting UBIFS is by specifying the UBI volume
1917
* character device node path. However, UBIFS may also be mounted withoug any
1918
* character device node using one of the following methods:
1920
* o ubiX_Y - mount UBI device number X, volume Y;
1921
* o ubiY - mount UBI device number 0, volume Y;
1922
* o ubiX:NAME - mount UBI device X, volume with name NAME;
1923
* o ubi:NAME - mount UBI device 0, volume with name NAME.
1925
* Alternative '!' separator may be used instead of ':' (because some shells
1926
* like busybox may interpret ':' as an NFS host name separator). This function
1927
* returns UBI volume description object in case of success and a negative
1928
* error code in case of failure.
1930
static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1932
struct ubi_volume_desc *ubi;
1936
/* First, try to open using the device node path method */
1937
ubi = ubi_open_volume_path(name, mode);
1941
/* Try the "nodev" method */
1942
if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1943
return ERR_PTR(-EINVAL);
1945
/* ubi:NAME method */
1946
if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1947
return ubi_open_volume_nm(0, name + 4, mode);
1949
if (!isdigit(name[3]))
1950
return ERR_PTR(-EINVAL);
1952
dev = simple_strtoul(name + 3, &endptr, 0);
1955
if (*endptr == '\0')
1956
return ubi_open_volume(0, dev, mode);
1959
if (*endptr == '_' && isdigit(endptr[1])) {
1960
vol = simple_strtoul(endptr + 1, &endptr, 0);
1961
if (*endptr != '\0')
1962
return ERR_PTR(-EINVAL);
1963
return ubi_open_volume(dev, vol, mode);
1966
/* ubiX:NAME method */
1967
if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1968
return ubi_open_volume_nm(dev, ++endptr, mode);
1970
return ERR_PTR(-EINVAL);
1973
static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
1975
struct ubifs_info *c;
1977
c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1979
spin_lock_init(&c->cnt_lock);
1980
spin_lock_init(&c->cs_lock);
1981
spin_lock_init(&c->buds_lock);
1982
spin_lock_init(&c->space_lock);
1983
spin_lock_init(&c->orphan_lock);
1984
init_rwsem(&c->commit_sem);
1985
mutex_init(&c->lp_mutex);
1986
mutex_init(&c->tnc_mutex);
1987
mutex_init(&c->log_mutex);
1988
mutex_init(&c->mst_mutex);
1989
mutex_init(&c->umount_mutex);
1990
mutex_init(&c->bu_mutex);
1991
mutex_init(&c->write_reserve_mutex);
1992
init_waitqueue_head(&c->cmt_wq);
1994
c->old_idx = RB_ROOT;
1995
c->size_tree = RB_ROOT;
1996
c->orph_tree = RB_ROOT;
1997
INIT_LIST_HEAD(&c->infos_list);
1998
INIT_LIST_HEAD(&c->idx_gc);
1999
INIT_LIST_HEAD(&c->replay_list);
2000
INIT_LIST_HEAD(&c->replay_buds);
2001
INIT_LIST_HEAD(&c->uncat_list);
2002
INIT_LIST_HEAD(&c->empty_list);
2003
INIT_LIST_HEAD(&c->freeable_list);
2004
INIT_LIST_HEAD(&c->frdi_idx_list);
2005
INIT_LIST_HEAD(&c->unclean_leb_list);
2006
INIT_LIST_HEAD(&c->old_buds);
2007
INIT_LIST_HEAD(&c->orph_list);
2008
INIT_LIST_HEAD(&c->orph_new);
2009
c->no_chk_data_crc = 1;
2011
c->highest_inum = UBIFS_FIRST_INO;
2012
c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
2014
ubi_get_volume_info(ubi, &c->vi);
2015
ubi_get_device_info(c->vi.ubi_num, &c->di);
2020
static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
2022
struct ubifs_info *c = sb->s_fs_info;
2027
/* Re-open the UBI device in read-write mode */
2028
c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
2029
if (IS_ERR(c->ubi)) {
2030
err = PTR_ERR(c->ubi);
2035
* UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
2036
* UBIFS, I/O is not deferred, it is done immediately in readpage,
2037
* which means the user would have to wait not just for their own I/O
2038
* but the read-ahead I/O as well i.e. completely pointless.
2040
* Read-ahead will be disabled because @c->bdi.ra_pages is 0.
2042
c->bdi.name = "ubifs",
2043
c->bdi.capabilities = BDI_CAP_MAP_COPY;
2044
err = bdi_init(&c->bdi);
2047
err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
2048
c->vi.ubi_num, c->vi.vol_id);
2052
err = ubifs_parse_options(c, data, 0);
2056
sb->s_bdi = &c->bdi;
2058
sb->s_magic = UBIFS_SUPER_MAGIC;
2059
sb->s_blocksize = UBIFS_BLOCK_SIZE;
2060
sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
2061
sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
2062
if (c->max_inode_sz > MAX_LFS_FILESIZE)
2063
sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
2064
sb->s_op = &ubifs_super_operations;
2066
mutex_lock(&c->umount_mutex);
2067
err = mount_ubifs(c);
2069
ubifs_assert(err < 0);
2073
/* Read the root inode */
2074
root = ubifs_iget(sb, UBIFS_ROOT_INO);
2076
err = PTR_ERR(root);
2080
sb->s_root = d_alloc_root(root);
2084
mutex_unlock(&c->umount_mutex);
2092
mutex_unlock(&c->umount_mutex);
2094
bdi_destroy(&c->bdi);
2096
ubi_close_volume(c->ubi);
2101
static int sb_test(struct super_block *sb, void *data)
2103
struct ubifs_info *c1 = data;
2104
struct ubifs_info *c = sb->s_fs_info;
2106
return c->vi.cdev == c1->vi.cdev;
2109
static int sb_set(struct super_block *sb, void *data)
2111
sb->s_fs_info = data;
2112
return set_anon_super(sb, NULL);
2115
static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
2116
const char *name, void *data)
2118
struct ubi_volume_desc *ubi;
2119
struct ubifs_info *c;
2120
struct super_block *sb;
2123
dbg_gen("name %s, flags %#x", name, flags);
2126
* Get UBI device number and volume ID. Mount it read-only so far
2127
* because this might be a new mount point, and UBI allows only one
2128
* read-write user at a time.
2130
ubi = open_ubi(name, UBI_READONLY);
2132
dbg_err("cannot open \"%s\", error %d",
2133
name, (int)PTR_ERR(ubi));
2134
return ERR_CAST(ubi);
2137
c = alloc_ubifs_info(ubi);
2143
dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
2145
sb = sget(fs_type, sb_test, sb_set, c);
2153
struct ubifs_info *c1 = sb->s_fs_info;
2155
/* A new mount point for already mounted UBIFS */
2156
dbg_gen("this ubi volume is already mounted");
2157
if (!!(flags & MS_RDONLY) != c1->ro_mount) {
2162
sb->s_flags = flags;
2163
err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
2166
/* We do not support atime */
2167
sb->s_flags |= MS_ACTIVE | MS_NOATIME;
2170
/* 'fill_super()' opens ubi again so we must close it here */
2171
ubi_close_volume(ubi);
2173
return dget(sb->s_root);
2176
deactivate_locked_super(sb);
2178
ubi_close_volume(ubi);
2179
return ERR_PTR(err);
2182
static void kill_ubifs_super(struct super_block *s)
2184
struct ubifs_info *c = s->s_fs_info;
2189
static struct file_system_type ubifs_fs_type = {
2191
.owner = THIS_MODULE,
2192
.mount = ubifs_mount,
2193
.kill_sb = kill_ubifs_super,
2197
* Inode slab cache constructor.
2199
static void inode_slab_ctor(void *obj)
2201
struct ubifs_inode *ui = obj;
2202
inode_init_once(&ui->vfs_inode);
2205
static int __init ubifs_init(void)
2209
BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
2211
/* Make sure node sizes are 8-byte aligned */
2212
BUILD_BUG_ON(UBIFS_CH_SZ & 7);
2213
BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
2214
BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
2215
BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
2216
BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
2217
BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
2218
BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
2219
BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
2220
BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
2221
BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
2222
BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
2224
BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
2225
BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
2226
BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
2227
BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
2228
BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
2229
BUILD_BUG_ON(MIN_WRITE_SZ & 7);
2231
/* Check min. node size */
2232
BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
2233
BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
2234
BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
2235
BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
2237
BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2238
BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
2239
BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
2240
BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
2242
/* Defined node sizes */
2243
BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
2244
BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
2245
BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
2246
BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
2249
* We use 2 bit wide bit-fields to store compression type, which should
2250
* be amended if more compressors are added. The bit-fields are:
2251
* @compr_type in 'struct ubifs_inode', @default_compr in
2252
* 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2254
BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4);
2257
* We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2258
* UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2260
if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
2261
ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2262
" at least 4096 bytes",
2263
(unsigned int)PAGE_CACHE_SIZE);
2267
ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
2268
sizeof(struct ubifs_inode), 0,
2269
SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
2271
if (!ubifs_inode_slab)
2274
register_shrinker(&ubifs_shrinker_info);
2276
err = ubifs_compressors_init();
2280
err = dbg_debugfs_init();
2284
err = register_filesystem(&ubifs_fs_type);
2286
ubifs_err("cannot register file system, error %d", err);
2294
ubifs_compressors_exit();
2296
unregister_shrinker(&ubifs_shrinker_info);
2297
kmem_cache_destroy(ubifs_inode_slab);
2300
/* late_initcall to let compressors initialize first */
2301
late_initcall(ubifs_init);
2303
static void __exit ubifs_exit(void)
2305
ubifs_assert(list_empty(&ubifs_infos));
2306
ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
2309
ubifs_compressors_exit();
2310
unregister_shrinker(&ubifs_shrinker_info);
2311
kmem_cache_destroy(ubifs_inode_slab);
2312
unregister_filesystem(&ubifs_fs_type);
2314
module_exit(ubifs_exit);
2316
MODULE_LICENSE("GPL");
2317
MODULE_VERSION(__stringify(UBIFS_VERSION));
2318
MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2319
MODULE_DESCRIPTION("UBIFS - UBI File System");