2 * This file is part of UBIFS.
4 * Copyright (C) 2006-2008 Nokia Corporation.
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License version 2 as published by
8 * the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc., 51
17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 * Authors: Artem Bityutskiy (Битюцкий Артём)
24 * This file implements UBIFS initialization and VFS superblock operations. Some
25 * initialization stuff which is rather large and complex is placed at
26 * corresponding subsystems, but most of it is here.
29 #include <linux/init.h>
30 #include <linux/slab.h>
31 #include <linux/module.h>
32 #include <linux/ctype.h>
33 #include <linux/kthread.h>
34 #include <linux/parser.h>
35 #include <linux/seq_file.h>
36 #include <linux/mount.h>
37 #include <linux/math64.h>
38 #include <linux/writeback.h>
42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
43 * allocating too much.
45 #define UBIFS_KMALLOC_OK (128*1024)
47 /* Slab cache for UBIFS inodes */
48 struct kmem_cache
*ubifs_inode_slab
;
50 /* UBIFS TNC shrinker description */
51 static struct shrinker ubifs_shrinker_info
= {
52 .shrink
= ubifs_shrinker
,
53 .seeks
= DEFAULT_SEEKS
,
57 * validate_inode - validate inode.
58 * @c: UBIFS file-system description object
59 * @inode: the inode to validate
61 * This is a helper function for 'ubifs_iget()' which validates various fields
62 * of a newly built inode to make sure they contain sane values and prevent
63 * possible vulnerabilities. Returns zero if the inode is all right and
64 * a non-zero error code if not.
66 static int validate_inode(struct ubifs_info
*c
, const struct inode
*inode
)
69 const struct ubifs_inode
*ui
= ubifs_inode(inode
);
71 if (inode
->i_size
> c
->max_inode_sz
) {
72 ubifs_err("inode is too large (%lld)",
73 (long long)inode
->i_size
);
77 if (ui
->compr_type
< 0 || ui
->compr_type
>= UBIFS_COMPR_TYPES_CNT
) {
78 ubifs_err("unknown compression type %d", ui
->compr_type
);
82 if (ui
->xattr_names
+ ui
->xattr_cnt
> XATTR_LIST_MAX
)
85 if (ui
->data_len
< 0 || ui
->data_len
> UBIFS_MAX_INO_DATA
)
88 if (ui
->xattr
&& (inode
->i_mode
& S_IFMT
) != S_IFREG
)
91 if (!ubifs_compr_present(ui
->compr_type
)) {
92 ubifs_warn("inode %lu uses '%s' compression, but it was not "
93 "compiled in", inode
->i_ino
,
94 ubifs_compr_name(ui
->compr_type
));
97 err
= dbg_check_dir_size(c
, inode
);
101 struct inode
*ubifs_iget(struct super_block
*sb
, unsigned long inum
)
105 struct ubifs_ino_node
*ino
;
106 struct ubifs_info
*c
= sb
->s_fs_info
;
108 struct ubifs_inode
*ui
;
110 dbg_gen("inode %lu", inum
);
112 inode
= iget_locked(sb
, inum
);
114 return ERR_PTR(-ENOMEM
);
115 if (!(inode
->i_state
& I_NEW
))
117 ui
= ubifs_inode(inode
);
119 ino
= kmalloc(UBIFS_MAX_INO_NODE_SZ
, GFP_NOFS
);
125 ino_key_init(c
, &key
, inode
->i_ino
);
127 err
= ubifs_tnc_lookup(c
, &key
, ino
);
131 inode
->i_flags
|= (S_NOCMTIME
| S_NOATIME
);
132 inode
->i_nlink
= le32_to_cpu(ino
->nlink
);
133 inode
->i_uid
= le32_to_cpu(ino
->uid
);
134 inode
->i_gid
= le32_to_cpu(ino
->gid
);
135 inode
->i_atime
.tv_sec
= (int64_t)le64_to_cpu(ino
->atime_sec
);
136 inode
->i_atime
.tv_nsec
= le32_to_cpu(ino
->atime_nsec
);
137 inode
->i_mtime
.tv_sec
= (int64_t)le64_to_cpu(ino
->mtime_sec
);
138 inode
->i_mtime
.tv_nsec
= le32_to_cpu(ino
->mtime_nsec
);
139 inode
->i_ctime
.tv_sec
= (int64_t)le64_to_cpu(ino
->ctime_sec
);
140 inode
->i_ctime
.tv_nsec
= le32_to_cpu(ino
->ctime_nsec
);
141 inode
->i_mode
= le32_to_cpu(ino
->mode
);
142 inode
->i_size
= le64_to_cpu(ino
->size
);
144 ui
->data_len
= le32_to_cpu(ino
->data_len
);
145 ui
->flags
= le32_to_cpu(ino
->flags
);
146 ui
->compr_type
= le16_to_cpu(ino
->compr_type
);
147 ui
->creat_sqnum
= le64_to_cpu(ino
->creat_sqnum
);
148 ui
->xattr_cnt
= le32_to_cpu(ino
->xattr_cnt
);
149 ui
->xattr_size
= le32_to_cpu(ino
->xattr_size
);
150 ui
->xattr_names
= le32_to_cpu(ino
->xattr_names
);
151 ui
->synced_i_size
= ui
->ui_size
= inode
->i_size
;
153 ui
->xattr
= (ui
->flags
& UBIFS_XATTR_FL
) ? 1 : 0;
155 err
= validate_inode(c
, inode
);
159 /* Disable read-ahead */
160 inode
->i_mapping
->backing_dev_info
= &c
->bdi
;
162 switch (inode
->i_mode
& S_IFMT
) {
164 inode
->i_mapping
->a_ops
= &ubifs_file_address_operations
;
165 inode
->i_op
= &ubifs_file_inode_operations
;
166 inode
->i_fop
= &ubifs_file_operations
;
168 ui
->data
= kmalloc(ui
->data_len
+ 1, GFP_NOFS
);
173 memcpy(ui
->data
, ino
->data
, ui
->data_len
);
174 ((char *)ui
->data
)[ui
->data_len
] = '\0';
175 } else if (ui
->data_len
!= 0) {
181 inode
->i_op
= &ubifs_dir_inode_operations
;
182 inode
->i_fop
= &ubifs_dir_operations
;
183 if (ui
->data_len
!= 0) {
189 inode
->i_op
= &ubifs_symlink_inode_operations
;
190 if (ui
->data_len
<= 0 || ui
->data_len
> UBIFS_MAX_INO_DATA
) {
194 ui
->data
= kmalloc(ui
->data_len
+ 1, GFP_NOFS
);
199 memcpy(ui
->data
, ino
->data
, ui
->data_len
);
200 ((char *)ui
->data
)[ui
->data_len
] = '\0';
206 union ubifs_dev_desc
*dev
;
208 ui
->data
= kmalloc(sizeof(union ubifs_dev_desc
), GFP_NOFS
);
214 dev
= (union ubifs_dev_desc
*)ino
->data
;
215 if (ui
->data_len
== sizeof(dev
->new))
216 rdev
= new_decode_dev(le32_to_cpu(dev
->new));
217 else if (ui
->data_len
== sizeof(dev
->huge
))
218 rdev
= huge_decode_dev(le64_to_cpu(dev
->huge
));
223 memcpy(ui
->data
, ino
->data
, ui
->data_len
);
224 inode
->i_op
= &ubifs_file_inode_operations
;
225 init_special_inode(inode
, inode
->i_mode
, rdev
);
230 inode
->i_op
= &ubifs_file_inode_operations
;
231 init_special_inode(inode
, inode
->i_mode
, 0);
232 if (ui
->data_len
!= 0) {
243 ubifs_set_inode_flags(inode
);
244 unlock_new_inode(inode
);
248 ubifs_err("inode %lu validation failed, error %d", inode
->i_ino
, err
);
249 dbg_dump_node(c
, ino
);
250 dbg_dump_inode(c
, inode
);
255 ubifs_err("failed to read inode %lu, error %d", inode
->i_ino
, err
);
260 static struct inode
*ubifs_alloc_inode(struct super_block
*sb
)
262 struct ubifs_inode
*ui
;
264 ui
= kmem_cache_alloc(ubifs_inode_slab
, GFP_NOFS
);
268 memset((void *)ui
+ sizeof(struct inode
), 0,
269 sizeof(struct ubifs_inode
) - sizeof(struct inode
));
270 mutex_init(&ui
->ui_mutex
);
271 spin_lock_init(&ui
->ui_lock
);
272 return &ui
->vfs_inode
;
275 static void ubifs_destroy_inode(struct inode
*inode
)
277 struct ubifs_inode
*ui
= ubifs_inode(inode
);
280 kmem_cache_free(ubifs_inode_slab
, inode
);
284 * Note, Linux write-back code calls this without 'i_mutex'.
286 static int ubifs_write_inode(struct inode
*inode
, struct writeback_control
*wbc
)
289 struct ubifs_info
*c
= inode
->i_sb
->s_fs_info
;
290 struct ubifs_inode
*ui
= ubifs_inode(inode
);
292 ubifs_assert(!ui
->xattr
);
293 if (is_bad_inode(inode
))
296 mutex_lock(&ui
->ui_mutex
);
298 * Due to races between write-back forced by budgeting
299 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
300 * have already been synchronized, do not do this again. This might
301 * also happen if it was synchronized in an VFS operation, e.g.
305 mutex_unlock(&ui
->ui_mutex
);
310 * As an optimization, do not write orphan inodes to the media just
311 * because this is not needed.
313 dbg_gen("inode %lu, mode %#x, nlink %u",
314 inode
->i_ino
, (int)inode
->i_mode
, inode
->i_nlink
);
315 if (inode
->i_nlink
) {
316 err
= ubifs_jnl_write_inode(c
, inode
);
318 ubifs_err("can't write inode %lu, error %d",
321 err
= dbg_check_inode_size(c
, inode
, ui
->ui_size
);
325 mutex_unlock(&ui
->ui_mutex
);
326 ubifs_release_dirty_inode_budget(c
, ui
);
330 static void ubifs_evict_inode(struct inode
*inode
)
333 struct ubifs_info
*c
= inode
->i_sb
->s_fs_info
;
334 struct ubifs_inode
*ui
= ubifs_inode(inode
);
338 * Extended attribute inode deletions are fully handled in
339 * 'ubifs_removexattr()'. These inodes are special and have
340 * limited usage, so there is nothing to do here.
344 dbg_gen("inode %lu, mode %#x", inode
->i_ino
, (int)inode
->i_mode
);
345 ubifs_assert(!atomic_read(&inode
->i_count
));
347 truncate_inode_pages(&inode
->i_data
, 0);
352 if (is_bad_inode(inode
))
355 ui
->ui_size
= inode
->i_size
= 0;
356 err
= ubifs_jnl_delete_inode(c
, inode
);
359 * Worst case we have a lost orphan inode wasting space, so a
360 * simple error message is OK here.
362 ubifs_err("can't delete inode %lu, error %d",
367 ubifs_release_dirty_inode_budget(c
, ui
);
369 /* We've deleted something - clean the "no space" flags */
370 c
->nospace
= c
->nospace_rp
= 0;
374 end_writeback(inode
);
377 static void ubifs_dirty_inode(struct inode
*inode
)
379 struct ubifs_inode
*ui
= ubifs_inode(inode
);
381 ubifs_assert(mutex_is_locked(&ui
->ui_mutex
));
384 dbg_gen("inode %lu", inode
->i_ino
);
388 static int ubifs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
390 struct ubifs_info
*c
= dentry
->d_sb
->s_fs_info
;
391 unsigned long long free
;
392 __le32
*uuid
= (__le32
*)c
->uuid
;
394 free
= ubifs_get_free_space(c
);
395 dbg_gen("free space %lld bytes (%lld blocks)",
396 free
, free
>> UBIFS_BLOCK_SHIFT
);
398 buf
->f_type
= UBIFS_SUPER_MAGIC
;
399 buf
->f_bsize
= UBIFS_BLOCK_SIZE
;
400 buf
->f_blocks
= c
->block_cnt
;
401 buf
->f_bfree
= free
>> UBIFS_BLOCK_SHIFT
;
402 if (free
> c
->report_rp_size
)
403 buf
->f_bavail
= (free
- c
->report_rp_size
) >> UBIFS_BLOCK_SHIFT
;
408 buf
->f_namelen
= UBIFS_MAX_NLEN
;
409 buf
->f_fsid
.val
[0] = le32_to_cpu(uuid
[0]) ^ le32_to_cpu(uuid
[2]);
410 buf
->f_fsid
.val
[1] = le32_to_cpu(uuid
[1]) ^ le32_to_cpu(uuid
[3]);
411 ubifs_assert(buf
->f_bfree
<= c
->block_cnt
);
415 static int ubifs_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
417 struct ubifs_info
*c
= mnt
->mnt_sb
->s_fs_info
;
419 if (c
->mount_opts
.unmount_mode
== 2)
420 seq_printf(s
, ",fast_unmount");
421 else if (c
->mount_opts
.unmount_mode
== 1)
422 seq_printf(s
, ",norm_unmount");
424 if (c
->mount_opts
.bulk_read
== 2)
425 seq_printf(s
, ",bulk_read");
426 else if (c
->mount_opts
.bulk_read
== 1)
427 seq_printf(s
, ",no_bulk_read");
429 if (c
->mount_opts
.chk_data_crc
== 2)
430 seq_printf(s
, ",chk_data_crc");
431 else if (c
->mount_opts
.chk_data_crc
== 1)
432 seq_printf(s
, ",no_chk_data_crc");
434 if (c
->mount_opts
.override_compr
) {
435 seq_printf(s
, ",compr=%s",
436 ubifs_compr_name(c
->mount_opts
.compr_type
));
442 static int ubifs_sync_fs(struct super_block
*sb
, int wait
)
445 struct ubifs_info
*c
= sb
->s_fs_info
;
448 * Zero @wait is just an advisory thing to help the file system shove
449 * lots of data into the queues, and there will be the second
450 * '->sync_fs()' call, with non-zero @wait.
456 * Synchronize write buffers, because 'ubifs_run_commit()' does not
457 * do this if it waits for an already running commit.
459 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
460 err
= ubifs_wbuf_sync(&c
->jheads
[i
].wbuf
);
466 * Strictly speaking, it is not necessary to commit the journal here,
467 * synchronizing write-buffers would be enough. But committing makes
468 * UBIFS free space predictions much more accurate, so we want to let
469 * the user be able to get more accurate results of 'statfs()' after
470 * they synchronize the file system.
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
.no_timer
= 1;
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 * parse_standard_option - parse a standard mount option.
937 * @option: the option to parse
939 * Normally, standard mount options like "sync" are passed to file-systems as
940 * flags. However, when a "rootflags=" kernel boot parameter is used, they may
941 * be present in the options string. This function tries to deal with this
942 * situation and parse standard options. Returns 0 if the option was not
943 * recognized, and the corresponding integer flag if it was.
945 * UBIFS is only interested in the "sync" option, so do not check for anything
948 static int parse_standard_option(const char *option
)
950 ubifs_msg("parse %s", option
);
951 if (!strcmp(option
, "sync"))
952 return MS_SYNCHRONOUS
;
957 * ubifs_parse_options - parse mount parameters.
958 * @c: UBIFS file-system description object
959 * @options: parameters to parse
960 * @is_remount: non-zero if this is FS re-mount
962 * This function parses UBIFS mount options and returns zero in case success
963 * and a negative error code in case of failure.
965 static int ubifs_parse_options(struct ubifs_info
*c
, char *options
,
969 substring_t args
[MAX_OPT_ARGS
];
974 while ((p
= strsep(&options
, ","))) {
980 token
= match_token(p
, tokens
, args
);
983 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
984 * We accept them in order to be backward-compatible. But this
985 * should be removed at some point.
987 case Opt_fast_unmount
:
988 c
->mount_opts
.unmount_mode
= 2;
990 case Opt_norm_unmount
:
991 c
->mount_opts
.unmount_mode
= 1;
994 c
->mount_opts
.bulk_read
= 2;
997 case Opt_no_bulk_read
:
998 c
->mount_opts
.bulk_read
= 1;
1001 case Opt_chk_data_crc
:
1002 c
->mount_opts
.chk_data_crc
= 2;
1003 c
->no_chk_data_crc
= 0;
1005 case Opt_no_chk_data_crc
:
1006 c
->mount_opts
.chk_data_crc
= 1;
1007 c
->no_chk_data_crc
= 1;
1009 case Opt_override_compr
:
1011 char *name
= match_strdup(&args
[0]);
1015 if (!strcmp(name
, "none"))
1016 c
->mount_opts
.compr_type
= UBIFS_COMPR_NONE
;
1017 else if (!strcmp(name
, "lzo"))
1018 c
->mount_opts
.compr_type
= UBIFS_COMPR_LZO
;
1019 else if (!strcmp(name
, "zlib"))
1020 c
->mount_opts
.compr_type
= UBIFS_COMPR_ZLIB
;
1022 ubifs_err("unknown compressor \"%s\"", name
);
1027 c
->mount_opts
.override_compr
= 1;
1028 c
->default_compr
= c
->mount_opts
.compr_type
;
1034 struct super_block
*sb
= c
->vfs_sb
;
1036 flag
= parse_standard_option(p
);
1038 ubifs_err("unrecognized mount option \"%s\" "
1039 "or missing value", p
);
1042 sb
->s_flags
|= flag
;
1052 * destroy_journal - destroy journal data structures.
1053 * @c: UBIFS file-system description object
1055 * This function destroys journal data structures including those that may have
1056 * been created by recovery functions.
1058 static void destroy_journal(struct ubifs_info
*c
)
1060 while (!list_empty(&c
->unclean_leb_list
)) {
1061 struct ubifs_unclean_leb
*ucleb
;
1063 ucleb
= list_entry(c
->unclean_leb_list
.next
,
1064 struct ubifs_unclean_leb
, list
);
1065 list_del(&ucleb
->list
);
1068 while (!list_empty(&c
->old_buds
)) {
1069 struct ubifs_bud
*bud
;
1071 bud
= list_entry(c
->old_buds
.next
, struct ubifs_bud
, list
);
1072 list_del(&bud
->list
);
1075 ubifs_destroy_idx_gc(c
);
1076 ubifs_destroy_size_tree(c
);
1082 * bu_init - initialize bulk-read information.
1083 * @c: UBIFS file-system description object
1085 static void bu_init(struct ubifs_info
*c
)
1087 ubifs_assert(c
->bulk_read
== 1);
1090 return; /* Already initialized */
1093 c
->bu
.buf
= kmalloc(c
->max_bu_buf_len
, GFP_KERNEL
| __GFP_NOWARN
);
1095 if (c
->max_bu_buf_len
> UBIFS_KMALLOC_OK
) {
1096 c
->max_bu_buf_len
= UBIFS_KMALLOC_OK
;
1100 /* Just disable bulk-read */
1101 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1102 "disabling it", c
->max_bu_buf_len
);
1103 c
->mount_opts
.bulk_read
= 1;
1110 * check_free_space - check if there is enough free space to mount.
1111 * @c: UBIFS file-system description object
1113 * This function makes sure UBIFS has enough free space to be mounted in
1114 * read/write mode. UBIFS must always have some free space to allow deletions.
1116 static int check_free_space(struct ubifs_info
*c
)
1118 ubifs_assert(c
->dark_wm
> 0);
1119 if (c
->lst
.total_free
+ c
->lst
.total_dirty
< c
->dark_wm
) {
1120 ubifs_err("insufficient free space to mount in read/write mode");
1129 * mount_ubifs - mount UBIFS file-system.
1130 * @c: UBIFS file-system description object
1132 * This function mounts UBIFS file system. Returns zero in case of success and
1133 * a negative error code in case of failure.
1135 * Note, the function does not de-allocate resources it it fails half way
1136 * through, and the caller has to do this instead.
1138 static int mount_ubifs(struct ubifs_info
*c
)
1140 struct super_block
*sb
= c
->vfs_sb
;
1141 int err
, mounted_read_only
= (sb
->s_flags
& MS_RDONLY
);
1145 err
= init_constants_early(c
);
1149 err
= ubifs_debugging_init(c
);
1153 err
= check_volume_empty(c
);
1157 if (c
->empty
&& (mounted_read_only
|| c
->ro_media
)) {
1159 * This UBI volume is empty, and read-only, or the file system
1160 * is mounted read-only - we cannot format it.
1162 ubifs_err("can't format empty UBI volume: read-only %s",
1163 c
->ro_media
? "UBI volume" : "mount");
1168 if (c
->ro_media
&& !mounted_read_only
) {
1169 ubifs_err("cannot mount read-write - read-only media");
1175 * The requirement for the buffer is that it should fit indexing B-tree
1176 * height amount of integers. We assume the height if the TNC tree will
1180 c
->bottom_up_buf
= kmalloc(BOTTOM_UP_HEIGHT
* sizeof(int), GFP_KERNEL
);
1181 if (!c
->bottom_up_buf
)
1184 c
->sbuf
= vmalloc(c
->leb_size
);
1188 if (!mounted_read_only
) {
1189 c
->ileb_buf
= vmalloc(c
->leb_size
);
1194 if (c
->bulk_read
== 1)
1198 * We have to check all CRCs, even for data nodes, when we mount the FS
1199 * (specifically, when we are replaying).
1201 c
->always_chk_crc
= 1;
1203 err
= ubifs_read_superblock(c
);
1208 * Make sure the compressor which is set as default in the superblock
1209 * or overridden by mount options is actually compiled in.
1211 if (!ubifs_compr_present(c
->default_compr
)) {
1212 ubifs_err("'compressor \"%s\" is not compiled in",
1213 ubifs_compr_name(c
->default_compr
));
1218 err
= init_constants_sb(c
);
1222 sz
= ALIGN(c
->max_idx_node_sz
, c
->min_io_size
);
1223 sz
= ALIGN(sz
+ c
->max_idx_node_sz
, c
->min_io_size
);
1224 c
->cbuf
= kmalloc(sz
, GFP_NOFS
);
1230 sprintf(c
->bgt_name
, BGT_NAME_PATTERN
, c
->vi
.ubi_num
, c
->vi
.vol_id
);
1231 if (!mounted_read_only
) {
1232 err
= alloc_wbufs(c
);
1236 /* Create background thread */
1237 c
->bgt
= kthread_create(ubifs_bg_thread
, c
, "%s", c
->bgt_name
);
1238 if (IS_ERR(c
->bgt
)) {
1239 err
= PTR_ERR(c
->bgt
);
1241 ubifs_err("cannot spawn \"%s\", error %d",
1245 wake_up_process(c
->bgt
);
1248 err
= ubifs_read_master(c
);
1252 init_constants_master(c
);
1254 if ((c
->mst_node
->flags
& cpu_to_le32(UBIFS_MST_DIRTY
)) != 0) {
1255 ubifs_msg("recovery needed");
1256 c
->need_recovery
= 1;
1257 if (!mounted_read_only
) {
1258 err
= ubifs_recover_inl_heads(c
, c
->sbuf
);
1262 } else if (!mounted_read_only
) {
1264 * Set the "dirty" flag so that if we reboot uncleanly we
1265 * will notice this immediately on the next mount.
1267 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_DIRTY
);
1268 err
= ubifs_write_master(c
);
1273 err
= ubifs_lpt_init(c
, 1, !mounted_read_only
);
1277 err
= dbg_check_idx_size(c
, c
->old_idx_sz
);
1281 err
= ubifs_replay_journal(c
);
1285 /* Calculate 'min_idx_lebs' after journal replay */
1286 c
->min_idx_lebs
= ubifs_calc_min_idx_lebs(c
);
1288 err
= ubifs_mount_orphans(c
, c
->need_recovery
, mounted_read_only
);
1292 if (!mounted_read_only
) {
1295 err
= check_free_space(c
);
1299 /* Check for enough log space */
1300 lnum
= c
->lhead_lnum
+ 1;
1301 if (lnum
>= UBIFS_LOG_LNUM
+ c
->log_lebs
)
1302 lnum
= UBIFS_LOG_LNUM
;
1303 if (lnum
== c
->ltail_lnum
) {
1304 err
= ubifs_consolidate_log(c
);
1309 if (c
->need_recovery
) {
1310 err
= ubifs_recover_size(c
);
1313 err
= ubifs_rcvry_gc_commit(c
);
1317 err
= take_gc_lnum(c
);
1322 * GC LEB may contain garbage if there was an unclean
1323 * reboot, and it should be un-mapped.
1325 err
= ubifs_leb_unmap(c
, c
->gc_lnum
);
1330 err
= dbg_check_lprops(c
);
1333 } else if (c
->need_recovery
) {
1334 err
= ubifs_recover_size(c
);
1339 * Even if we mount read-only, we have to set space in GC LEB
1340 * to proper value because this affects UBIFS free space
1341 * reporting. We do not want to have a situation when
1342 * re-mounting from R/O to R/W changes amount of free space.
1344 err
= take_gc_lnum(c
);
1349 spin_lock(&ubifs_infos_lock
);
1350 list_add_tail(&c
->infos_list
, &ubifs_infos
);
1351 spin_unlock(&ubifs_infos_lock
);
1353 if (c
->need_recovery
) {
1354 if (mounted_read_only
)
1355 ubifs_msg("recovery deferred");
1357 c
->need_recovery
= 0;
1358 ubifs_msg("recovery completed");
1360 * GC LEB has to be empty and taken at this point. But
1361 * the journal head LEBs may also be accounted as
1362 * "empty taken" if they are empty.
1364 ubifs_assert(c
->lst
.taken_empty_lebs
> 0);
1367 ubifs_assert(c
->lst
.taken_empty_lebs
> 0);
1369 err
= dbg_check_filesystem(c
);
1373 err
= dbg_debugfs_init_fs(c
);
1377 c
->always_chk_crc
= 0;
1379 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1380 c
->vi
.ubi_num
, c
->vi
.vol_id
, c
->vi
.name
);
1381 if (mounted_read_only
)
1382 ubifs_msg("mounted read-only");
1383 x
= (long long)c
->main_lebs
* c
->leb_size
;
1384 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1385 "LEBs)", x
, x
>> 10, x
>> 20, c
->main_lebs
);
1386 x
= (long long)c
->log_lebs
* c
->leb_size
+ c
->max_bud_bytes
;
1387 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1388 "LEBs)", x
, x
>> 10, x
>> 20, c
->log_lebs
+ c
->max_bud_cnt
);
1389 ubifs_msg("media format: w%d/r%d (latest is w%d/r%d)",
1390 c
->fmt_version
, c
->ro_compat_version
,
1391 UBIFS_FORMAT_VERSION
, UBIFS_RO_COMPAT_VERSION
);
1392 ubifs_msg("default compressor: %s", ubifs_compr_name(c
->default_compr
));
1393 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1394 c
->report_rp_size
, c
->report_rp_size
>> 10);
1396 dbg_msg("compiled on: " __DATE__
" at " __TIME__
);
1397 dbg_msg("min. I/O unit size: %d bytes", c
->min_io_size
);
1398 dbg_msg("LEB size: %d bytes (%d KiB)",
1399 c
->leb_size
, c
->leb_size
>> 10);
1400 dbg_msg("data journal heads: %d",
1401 c
->jhead_cnt
- NONDATA_JHEADS_CNT
);
1402 dbg_msg("UUID: %pUB", c
->uuid
);
1403 dbg_msg("big_lpt %d", c
->big_lpt
);
1404 dbg_msg("log LEBs: %d (%d - %d)",
1405 c
->log_lebs
, UBIFS_LOG_LNUM
, c
->log_last
);
1406 dbg_msg("LPT area LEBs: %d (%d - %d)",
1407 c
->lpt_lebs
, c
->lpt_first
, c
->lpt_last
);
1408 dbg_msg("orphan area LEBs: %d (%d - %d)",
1409 c
->orph_lebs
, c
->orph_first
, c
->orph_last
);
1410 dbg_msg("main area LEBs: %d (%d - %d)",
1411 c
->main_lebs
, c
->main_first
, c
->leb_cnt
- 1);
1412 dbg_msg("index LEBs: %d", c
->lst
.idx_lebs
);
1413 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1414 c
->old_idx_sz
, c
->old_idx_sz
>> 10, c
->old_idx_sz
>> 20);
1415 dbg_msg("key hash type: %d", c
->key_hash_type
);
1416 dbg_msg("tree fanout: %d", c
->fanout
);
1417 dbg_msg("reserved GC LEB: %d", c
->gc_lnum
);
1418 dbg_msg("first main LEB: %d", c
->main_first
);
1419 dbg_msg("max. znode size %d", c
->max_znode_sz
);
1420 dbg_msg("max. index node size %d", c
->max_idx_node_sz
);
1421 dbg_msg("node sizes: data %zu, inode %zu, dentry %zu",
1422 UBIFS_DATA_NODE_SZ
, UBIFS_INO_NODE_SZ
, UBIFS_DENT_NODE_SZ
);
1423 dbg_msg("node sizes: trun %zu, sb %zu, master %zu",
1424 UBIFS_TRUN_NODE_SZ
, UBIFS_SB_NODE_SZ
, UBIFS_MST_NODE_SZ
);
1425 dbg_msg("node sizes: ref %zu, cmt. start %zu, orph %zu",
1426 UBIFS_REF_NODE_SZ
, UBIFS_CS_NODE_SZ
, UBIFS_ORPH_NODE_SZ
);
1427 dbg_msg("max. node sizes: data %zu, inode %zu dentry %zu",
1428 UBIFS_MAX_DATA_NODE_SZ
, UBIFS_MAX_INO_NODE_SZ
,
1429 UBIFS_MAX_DENT_NODE_SZ
);
1430 dbg_msg("dead watermark: %d", c
->dead_wm
);
1431 dbg_msg("dark watermark: %d", c
->dark_wm
);
1432 dbg_msg("LEB overhead: %d", c
->leb_overhead
);
1433 x
= (long long)c
->main_lebs
* c
->dark_wm
;
1434 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1435 x
, x
>> 10, x
>> 20);
1436 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1437 c
->max_bud_bytes
, c
->max_bud_bytes
>> 10,
1438 c
->max_bud_bytes
>> 20);
1439 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1440 c
->bg_bud_bytes
, c
->bg_bud_bytes
>> 10,
1441 c
->bg_bud_bytes
>> 20);
1442 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1443 c
->bud_bytes
, c
->bud_bytes
>> 10, c
->bud_bytes
>> 20);
1444 dbg_msg("max. seq. number: %llu", c
->max_sqnum
);
1445 dbg_msg("commit number: %llu", c
->cmt_no
);
1450 spin_lock(&ubifs_infos_lock
);
1451 list_del(&c
->infos_list
);
1452 spin_unlock(&ubifs_infos_lock
);
1458 ubifs_lpt_free(c
, 0);
1461 kfree(c
->rcvrd_mst_node
);
1463 kthread_stop(c
->bgt
);
1472 kfree(c
->bottom_up_buf
);
1473 ubifs_debugging_exit(c
);
1478 * ubifs_umount - un-mount UBIFS file-system.
1479 * @c: UBIFS file-system description object
1481 * Note, this function is called to free allocated resourced when un-mounting,
1482 * as well as free resources when an error occurred while we were half way
1483 * through mounting (error path cleanup function). So it has to make sure the
1484 * resource was actually allocated before freeing it.
1486 static void ubifs_umount(struct ubifs_info
*c
)
1488 dbg_gen("un-mounting UBI device %d, volume %d", c
->vi
.ubi_num
,
1491 dbg_debugfs_exit_fs(c
);
1492 spin_lock(&ubifs_infos_lock
);
1493 list_del(&c
->infos_list
);
1494 spin_unlock(&ubifs_infos_lock
);
1497 kthread_stop(c
->bgt
);
1502 ubifs_lpt_free(c
, 0);
1505 kfree(c
->rcvrd_mst_node
);
1510 kfree(c
->bottom_up_buf
);
1511 ubifs_debugging_exit(c
);
1515 * ubifs_remount_rw - re-mount in read-write mode.
1516 * @c: UBIFS file-system description object
1518 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1519 * mode. This function allocates the needed resources and re-mounts UBIFS in
1522 static int ubifs_remount_rw(struct ubifs_info
*c
)
1526 if (c
->rw_incompat
) {
1527 ubifs_err("the file-system is not R/W-compatible");
1528 ubifs_msg("on-flash format version is w%d/r%d, but software "
1529 "only supports up to version w%d/r%d", c
->fmt_version
,
1530 c
->ro_compat_version
, UBIFS_FORMAT_VERSION
,
1531 UBIFS_RO_COMPAT_VERSION
);
1535 mutex_lock(&c
->umount_mutex
);
1536 dbg_save_space_info(c
);
1537 c
->remounting_rw
= 1;
1538 c
->always_chk_crc
= 1;
1540 err
= check_free_space(c
);
1544 if (c
->old_leb_cnt
!= c
->leb_cnt
) {
1545 struct ubifs_sb_node
*sup
;
1547 sup
= ubifs_read_sb_node(c
);
1552 sup
->leb_cnt
= cpu_to_le32(c
->leb_cnt
);
1553 err
= ubifs_write_sb_node(c
, sup
);
1558 if (c
->need_recovery
) {
1559 ubifs_msg("completing deferred recovery");
1560 err
= ubifs_write_rcvrd_mst_node(c
);
1563 err
= ubifs_recover_size(c
);
1566 err
= ubifs_clean_lebs(c
, c
->sbuf
);
1569 err
= ubifs_recover_inl_heads(c
, c
->sbuf
);
1573 /* A readonly mount is not allowed to have orphans */
1574 ubifs_assert(c
->tot_orphans
== 0);
1575 err
= ubifs_clear_orphans(c
);
1580 if (!(c
->mst_node
->flags
& cpu_to_le32(UBIFS_MST_DIRTY
))) {
1581 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_DIRTY
);
1582 err
= ubifs_write_master(c
);
1587 c
->ileb_buf
= vmalloc(c
->leb_size
);
1593 err
= ubifs_lpt_init(c
, 0, 1);
1597 err
= alloc_wbufs(c
);
1601 ubifs_create_buds_lists(c
);
1603 /* Create background thread */
1604 c
->bgt
= kthread_create(ubifs_bg_thread
, c
, "%s", c
->bgt_name
);
1605 if (IS_ERR(c
->bgt
)) {
1606 err
= PTR_ERR(c
->bgt
);
1608 ubifs_err("cannot spawn \"%s\", error %d",
1612 wake_up_process(c
->bgt
);
1614 c
->orph_buf
= vmalloc(c
->leb_size
);
1620 /* Check for enough log space */
1621 lnum
= c
->lhead_lnum
+ 1;
1622 if (lnum
>= UBIFS_LOG_LNUM
+ c
->log_lebs
)
1623 lnum
= UBIFS_LOG_LNUM
;
1624 if (lnum
== c
->ltail_lnum
) {
1625 err
= ubifs_consolidate_log(c
);
1630 if (c
->need_recovery
)
1631 err
= ubifs_rcvry_gc_commit(c
);
1633 err
= ubifs_leb_unmap(c
, c
->gc_lnum
);
1637 if (c
->need_recovery
) {
1638 c
->need_recovery
= 0;
1639 ubifs_msg("deferred recovery completed");
1642 dbg_gen("re-mounted read-write");
1643 c
->vfs_sb
->s_flags
&= ~MS_RDONLY
;
1644 c
->remounting_rw
= 0;
1645 c
->always_chk_crc
= 0;
1646 err
= dbg_check_space_info(c
);
1647 mutex_unlock(&c
->umount_mutex
);
1654 kthread_stop(c
->bgt
);
1660 ubifs_lpt_free(c
, 1);
1661 c
->remounting_rw
= 0;
1662 c
->always_chk_crc
= 0;
1663 mutex_unlock(&c
->umount_mutex
);
1668 * ubifs_remount_ro - re-mount in read-only mode.
1669 * @c: UBIFS file-system description object
1671 * We assume VFS has stopped writing. Possibly the background thread could be
1672 * running a commit, however kthread_stop will wait in that case.
1674 static void ubifs_remount_ro(struct ubifs_info
*c
)
1678 ubifs_assert(!c
->need_recovery
);
1679 ubifs_assert(!(c
->vfs_sb
->s_flags
& MS_RDONLY
));
1681 mutex_lock(&c
->umount_mutex
);
1683 kthread_stop(c
->bgt
);
1687 dbg_save_space_info(c
);
1689 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
1690 ubifs_wbuf_sync(&c
->jheads
[i
].wbuf
);
1691 hrtimer_cancel(&c
->jheads
[i
].wbuf
.timer
);
1694 c
->mst_node
->flags
&= ~cpu_to_le32(UBIFS_MST_DIRTY
);
1695 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_NO_ORPHS
);
1696 c
->mst_node
->gc_lnum
= cpu_to_le32(c
->gc_lnum
);
1697 err
= ubifs_write_master(c
);
1699 ubifs_ro_mode(c
, err
);
1706 ubifs_lpt_free(c
, 1);
1707 err
= dbg_check_space_info(c
);
1709 ubifs_ro_mode(c
, err
);
1710 mutex_unlock(&c
->umount_mutex
);
1713 static void ubifs_put_super(struct super_block
*sb
)
1716 struct ubifs_info
*c
= sb
->s_fs_info
;
1718 ubifs_msg("un-mount UBI device %d, volume %d", c
->vi
.ubi_num
,
1722 * The following asserts are only valid if there has not been a failure
1723 * of the media. For example, there will be dirty inodes if we failed
1724 * to write them back because of I/O errors.
1726 ubifs_assert(atomic_long_read(&c
->dirty_pg_cnt
) == 0);
1727 ubifs_assert(c
->budg_idx_growth
== 0);
1728 ubifs_assert(c
->budg_dd_growth
== 0);
1729 ubifs_assert(c
->budg_data_growth
== 0);
1732 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1733 * and file system un-mount. Namely, it prevents the shrinker from
1734 * picking this superblock for shrinking - it will be just skipped if
1735 * the mutex is locked.
1737 mutex_lock(&c
->umount_mutex
);
1738 if (!(c
->vfs_sb
->s_flags
& MS_RDONLY
)) {
1740 * First of all kill the background thread to make sure it does
1741 * not interfere with un-mounting and freeing resources.
1744 kthread_stop(c
->bgt
);
1748 /* Synchronize write-buffers */
1750 for (i
= 0; i
< c
->jhead_cnt
; i
++)
1751 ubifs_wbuf_sync(&c
->jheads
[i
].wbuf
);
1754 * On fatal errors c->ro_media is set to 1, in which case we do
1755 * not write the master node.
1759 * We are being cleanly unmounted which means the
1760 * orphans were killed - indicate this in the master
1761 * node. Also save the reserved GC LEB number.
1765 c
->mst_node
->flags
&= ~cpu_to_le32(UBIFS_MST_DIRTY
);
1766 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_NO_ORPHS
);
1767 c
->mst_node
->gc_lnum
= cpu_to_le32(c
->gc_lnum
);
1768 err
= ubifs_write_master(c
);
1771 * Recovery will attempt to fix the master area
1772 * next mount, so we just print a message and
1773 * continue to unmount normally.
1775 ubifs_err("failed to write master node, "
1781 bdi_destroy(&c
->bdi
);
1782 ubi_close_volume(c
->ubi
);
1783 mutex_unlock(&c
->umount_mutex
);
1787 static int ubifs_remount_fs(struct super_block
*sb
, int *flags
, char *data
)
1790 struct ubifs_info
*c
= sb
->s_fs_info
;
1792 dbg_gen("old flags %#lx, new flags %#x", sb
->s_flags
, *flags
);
1794 err
= ubifs_parse_options(c
, data
, 1);
1796 ubifs_err("invalid or unknown remount parameter");
1800 if ((sb
->s_flags
& MS_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1802 ubifs_msg("cannot re-mount due to prior errors");
1805 err
= ubifs_remount_rw(c
);
1808 } else if (!(sb
->s_flags
& MS_RDONLY
) && (*flags
& MS_RDONLY
)) {
1810 ubifs_msg("cannot re-mount due to prior errors");
1813 ubifs_remount_ro(c
);
1816 if (c
->bulk_read
== 1)
1819 dbg_gen("disable bulk-read");
1824 ubifs_assert(c
->lst
.taken_empty_lebs
> 0);
1828 const struct super_operations ubifs_super_operations
= {
1829 .alloc_inode
= ubifs_alloc_inode
,
1830 .destroy_inode
= ubifs_destroy_inode
,
1831 .put_super
= ubifs_put_super
,
1832 .write_inode
= ubifs_write_inode
,
1833 .evict_inode
= ubifs_evict_inode
,
1834 .statfs
= ubifs_statfs
,
1835 .dirty_inode
= ubifs_dirty_inode
,
1836 .remount_fs
= ubifs_remount_fs
,
1837 .show_options
= ubifs_show_options
,
1838 .sync_fs
= ubifs_sync_fs
,
1842 * open_ubi - parse UBI device name string and open the UBI device.
1843 * @name: UBI volume name
1844 * @mode: UBI volume open mode
1846 * The primary method of mounting UBIFS is by specifying the UBI volume
1847 * character device node path. However, UBIFS may also be mounted withoug any
1848 * character device node using one of the following methods:
1850 * o ubiX_Y - mount UBI device number X, volume Y;
1851 * o ubiY - mount UBI device number 0, volume Y;
1852 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1853 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1855 * Alternative '!' separator may be used instead of ':' (because some shells
1856 * like busybox may interpret ':' as an NFS host name separator). This function
1857 * returns UBI volume description object in case of success and a negative
1858 * error code in case of failure.
1860 static struct ubi_volume_desc
*open_ubi(const char *name
, int mode
)
1862 struct ubi_volume_desc
*ubi
;
1866 /* First, try to open using the device node path method */
1867 ubi
= ubi_open_volume_path(name
, mode
);
1871 /* Try the "nodev" method */
1872 if (name
[0] != 'u' || name
[1] != 'b' || name
[2] != 'i')
1873 return ERR_PTR(-EINVAL
);
1875 /* ubi:NAME method */
1876 if ((name
[3] == ':' || name
[3] == '!') && name
[4] != '\0')
1877 return ubi_open_volume_nm(0, name
+ 4, mode
);
1879 if (!isdigit(name
[3]))
1880 return ERR_PTR(-EINVAL
);
1882 dev
= simple_strtoul(name
+ 3, &endptr
, 0);
1885 if (*endptr
== '\0')
1886 return ubi_open_volume(0, dev
, mode
);
1889 if (*endptr
== '_' && isdigit(endptr
[1])) {
1890 vol
= simple_strtoul(endptr
+ 1, &endptr
, 0);
1891 if (*endptr
!= '\0')
1892 return ERR_PTR(-EINVAL
);
1893 return ubi_open_volume(dev
, vol
, mode
);
1896 /* ubiX:NAME method */
1897 if ((*endptr
== ':' || *endptr
== '!') && endptr
[1] != '\0')
1898 return ubi_open_volume_nm(dev
, ++endptr
, mode
);
1900 return ERR_PTR(-EINVAL
);
1903 static int ubifs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1905 struct ubi_volume_desc
*ubi
= sb
->s_fs_info
;
1906 struct ubifs_info
*c
;
1910 c
= kzalloc(sizeof(struct ubifs_info
), GFP_KERNEL
);
1914 spin_lock_init(&c
->cnt_lock
);
1915 spin_lock_init(&c
->cs_lock
);
1916 spin_lock_init(&c
->buds_lock
);
1917 spin_lock_init(&c
->space_lock
);
1918 spin_lock_init(&c
->orphan_lock
);
1919 init_rwsem(&c
->commit_sem
);
1920 mutex_init(&c
->lp_mutex
);
1921 mutex_init(&c
->tnc_mutex
);
1922 mutex_init(&c
->log_mutex
);
1923 mutex_init(&c
->mst_mutex
);
1924 mutex_init(&c
->umount_mutex
);
1925 mutex_init(&c
->bu_mutex
);
1926 init_waitqueue_head(&c
->cmt_wq
);
1928 c
->old_idx
= RB_ROOT
;
1929 c
->size_tree
= RB_ROOT
;
1930 c
->orph_tree
= RB_ROOT
;
1931 INIT_LIST_HEAD(&c
->infos_list
);
1932 INIT_LIST_HEAD(&c
->idx_gc
);
1933 INIT_LIST_HEAD(&c
->replay_list
);
1934 INIT_LIST_HEAD(&c
->replay_buds
);
1935 INIT_LIST_HEAD(&c
->uncat_list
);
1936 INIT_LIST_HEAD(&c
->empty_list
);
1937 INIT_LIST_HEAD(&c
->freeable_list
);
1938 INIT_LIST_HEAD(&c
->frdi_idx_list
);
1939 INIT_LIST_HEAD(&c
->unclean_leb_list
);
1940 INIT_LIST_HEAD(&c
->old_buds
);
1941 INIT_LIST_HEAD(&c
->orph_list
);
1942 INIT_LIST_HEAD(&c
->orph_new
);
1945 c
->highest_inum
= UBIFS_FIRST_INO
;
1946 c
->lhead_lnum
= c
->ltail_lnum
= UBIFS_LOG_LNUM
;
1948 ubi_get_volume_info(ubi
, &c
->vi
);
1949 ubi_get_device_info(c
->vi
.ubi_num
, &c
->di
);
1951 /* Re-open the UBI device in read-write mode */
1952 c
->ubi
= ubi_open_volume(c
->vi
.ubi_num
, c
->vi
.vol_id
, UBI_READWRITE
);
1953 if (IS_ERR(c
->ubi
)) {
1954 err
= PTR_ERR(c
->ubi
);
1959 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1960 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1961 * which means the user would have to wait not just for their own I/O
1962 * but the read-ahead I/O as well i.e. completely pointless.
1964 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1966 c
->bdi
.name
= "ubifs",
1967 c
->bdi
.capabilities
= BDI_CAP_MAP_COPY
;
1968 c
->bdi
.unplug_io_fn
= default_unplug_io_fn
;
1969 err
= bdi_init(&c
->bdi
);
1972 err
= bdi_register(&c
->bdi
, NULL
, "ubifs_%d_%d",
1973 c
->vi
.ubi_num
, c
->vi
.vol_id
);
1977 err
= ubifs_parse_options(c
, data
, 0);
1981 sb
->s_bdi
= &c
->bdi
;
1983 sb
->s_magic
= UBIFS_SUPER_MAGIC
;
1984 sb
->s_blocksize
= UBIFS_BLOCK_SIZE
;
1985 sb
->s_blocksize_bits
= UBIFS_BLOCK_SHIFT
;
1986 sb
->s_maxbytes
= c
->max_inode_sz
= key_max_inode_size(c
);
1987 if (c
->max_inode_sz
> MAX_LFS_FILESIZE
)
1988 sb
->s_maxbytes
= c
->max_inode_sz
= MAX_LFS_FILESIZE
;
1989 sb
->s_op
= &ubifs_super_operations
;
1991 mutex_lock(&c
->umount_mutex
);
1992 err
= mount_ubifs(c
);
1994 ubifs_assert(err
< 0);
1998 /* Read the root inode */
1999 root
= ubifs_iget(sb
, UBIFS_ROOT_INO
);
2001 err
= PTR_ERR(root
);
2005 sb
->s_root
= d_alloc_root(root
);
2009 mutex_unlock(&c
->umount_mutex
);
2017 mutex_unlock(&c
->umount_mutex
);
2019 bdi_destroy(&c
->bdi
);
2021 ubi_close_volume(c
->ubi
);
2027 static int sb_test(struct super_block
*sb
, void *data
)
2030 struct ubifs_info
*c
= sb
->s_fs_info
;
2032 return c
->vi
.cdev
== *dev
;
2035 static int ubifs_get_sb(struct file_system_type
*fs_type
, int flags
,
2036 const char *name
, void *data
, struct vfsmount
*mnt
)
2038 struct ubi_volume_desc
*ubi
;
2039 struct ubi_volume_info vi
;
2040 struct super_block
*sb
;
2043 dbg_gen("name %s, flags %#x", name
, flags
);
2046 * Get UBI device number and volume ID. Mount it read-only so far
2047 * because this might be a new mount point, and UBI allows only one
2048 * read-write user at a time.
2050 ubi
= open_ubi(name
, UBI_READONLY
);
2052 ubifs_err("cannot open \"%s\", error %d",
2053 name
, (int)PTR_ERR(ubi
));
2054 return PTR_ERR(ubi
);
2056 ubi_get_volume_info(ubi
, &vi
);
2058 dbg_gen("opened ubi%d_%d", vi
.ubi_num
, vi
.vol_id
);
2060 sb
= sget(fs_type
, &sb_test
, &set_anon_super
, &vi
.cdev
);
2067 /* A new mount point for already mounted UBIFS */
2068 dbg_gen("this ubi volume is already mounted");
2069 if ((flags
^ sb
->s_flags
) & MS_RDONLY
) {
2074 sb
->s_flags
= flags
;
2076 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
2079 sb
->s_fs_info
= ubi
;
2080 err
= ubifs_fill_super(sb
, data
, flags
& MS_SILENT
? 1 : 0);
2083 /* We do not support atime */
2084 sb
->s_flags
|= MS_ACTIVE
| MS_NOATIME
;
2087 /* 'fill_super()' opens ubi again so we must close it here */
2088 ubi_close_volume(ubi
);
2090 simple_set_mnt(mnt
, sb
);
2094 deactivate_locked_super(sb
);
2096 ubi_close_volume(ubi
);
2100 static struct file_system_type ubifs_fs_type
= {
2102 .owner
= THIS_MODULE
,
2103 .get_sb
= ubifs_get_sb
,
2104 .kill_sb
= kill_anon_super
,
2108 * Inode slab cache constructor.
2110 static void inode_slab_ctor(void *obj
)
2112 struct ubifs_inode
*ui
= obj
;
2113 inode_init_once(&ui
->vfs_inode
);
2116 static int __init
ubifs_init(void)
2120 BUILD_BUG_ON(sizeof(struct ubifs_ch
) != 24);
2122 /* Make sure node sizes are 8-byte aligned */
2123 BUILD_BUG_ON(UBIFS_CH_SZ
& 7);
2124 BUILD_BUG_ON(UBIFS_INO_NODE_SZ
& 7);
2125 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ
& 7);
2126 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ
& 7);
2127 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ
& 7);
2128 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ
& 7);
2129 BUILD_BUG_ON(UBIFS_SB_NODE_SZ
& 7);
2130 BUILD_BUG_ON(UBIFS_MST_NODE_SZ
& 7);
2131 BUILD_BUG_ON(UBIFS_REF_NODE_SZ
& 7);
2132 BUILD_BUG_ON(UBIFS_CS_NODE_SZ
& 7);
2133 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ
& 7);
2135 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ
& 7);
2136 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ
& 7);
2137 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ
& 7);
2138 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ
& 7);
2139 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ
& 7);
2140 BUILD_BUG_ON(MIN_WRITE_SZ
& 7);
2142 /* Check min. node size */
2143 BUILD_BUG_ON(UBIFS_INO_NODE_SZ
< MIN_WRITE_SZ
);
2144 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ
< MIN_WRITE_SZ
);
2145 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ
< MIN_WRITE_SZ
);
2146 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ
< MIN_WRITE_SZ
);
2148 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2149 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2150 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2151 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2153 /* Defined node sizes */
2154 BUILD_BUG_ON(UBIFS_SB_NODE_SZ
!= 4096);
2155 BUILD_BUG_ON(UBIFS_MST_NODE_SZ
!= 512);
2156 BUILD_BUG_ON(UBIFS_INO_NODE_SZ
!= 160);
2157 BUILD_BUG_ON(UBIFS_REF_NODE_SZ
!= 64);
2160 * We use 2 bit wide bit-fields to store compression type, which should
2161 * be amended if more compressors are added. The bit-fields are:
2162 * @compr_type in 'struct ubifs_inode', @default_compr in
2163 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2165 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT
> 4);
2168 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2169 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2171 if (PAGE_CACHE_SIZE
< UBIFS_BLOCK_SIZE
) {
2172 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2173 " at least 4096 bytes",
2174 (unsigned int)PAGE_CACHE_SIZE
);
2178 err
= register_filesystem(&ubifs_fs_type
);
2180 ubifs_err("cannot register file system, error %d", err
);
2185 ubifs_inode_slab
= kmem_cache_create("ubifs_inode_slab",
2186 sizeof(struct ubifs_inode
), 0,
2187 SLAB_MEM_SPREAD
| SLAB_RECLAIM_ACCOUNT
,
2189 if (!ubifs_inode_slab
)
2192 register_shrinker(&ubifs_shrinker_info
);
2194 err
= ubifs_compressors_init();
2198 err
= dbg_debugfs_init();
2205 ubifs_compressors_exit();
2207 unregister_shrinker(&ubifs_shrinker_info
);
2208 kmem_cache_destroy(ubifs_inode_slab
);
2210 unregister_filesystem(&ubifs_fs_type
);
2213 /* late_initcall to let compressors initialize first */
2214 late_initcall(ubifs_init
);
2216 static void __exit
ubifs_exit(void)
2218 ubifs_assert(list_empty(&ubifs_infos
));
2219 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt
) == 0);
2222 ubifs_compressors_exit();
2223 unregister_shrinker(&ubifs_shrinker_info
);
2224 kmem_cache_destroy(ubifs_inode_slab
);
2225 unregister_filesystem(&ubifs_fs_type
);
2227 module_exit(ubifs_exit
);
2229 MODULE_LICENSE("GPL");
2230 MODULE_VERSION(__stringify(UBIFS_VERSION
));
2231 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2232 MODULE_DESCRIPTION("UBIFS - UBI File System");