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
, int wait
)
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",
323 mutex_unlock(&ui
->ui_mutex
);
324 ubifs_release_dirty_inode_budget(c
, ui
);
328 static void ubifs_delete_inode(struct inode
*inode
)
331 struct ubifs_info
*c
= inode
->i_sb
->s_fs_info
;
332 struct ubifs_inode
*ui
= ubifs_inode(inode
);
336 * Extended attribute inode deletions are fully handled in
337 * 'ubifs_removexattr()'. These inodes are special and have
338 * limited usage, so there is nothing to do here.
342 dbg_gen("inode %lu, mode %#x", inode
->i_ino
, (int)inode
->i_mode
);
343 ubifs_assert(!atomic_read(&inode
->i_count
));
344 ubifs_assert(inode
->i_nlink
== 0);
346 truncate_inode_pages(&inode
->i_data
, 0);
347 if (is_bad_inode(inode
))
350 ui
->ui_size
= inode
->i_size
= 0;
351 err
= ubifs_jnl_delete_inode(c
, inode
);
354 * Worst case we have a lost orphan inode wasting space, so a
355 * simple error message is OK here.
357 ubifs_err("can't delete inode %lu, error %d",
362 ubifs_release_dirty_inode_budget(c
, ui
);
366 static void ubifs_dirty_inode(struct inode
*inode
)
368 struct ubifs_inode
*ui
= ubifs_inode(inode
);
370 ubifs_assert(mutex_is_locked(&ui
->ui_mutex
));
373 dbg_gen("inode %lu", inode
->i_ino
);
377 static int ubifs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
379 struct ubifs_info
*c
= dentry
->d_sb
->s_fs_info
;
380 unsigned long long free
;
381 __le32
*uuid
= (__le32
*)c
->uuid
;
383 free
= ubifs_get_free_space(c
);
384 dbg_gen("free space %lld bytes (%lld blocks)",
385 free
, free
>> UBIFS_BLOCK_SHIFT
);
387 buf
->f_type
= UBIFS_SUPER_MAGIC
;
388 buf
->f_bsize
= UBIFS_BLOCK_SIZE
;
389 buf
->f_blocks
= c
->block_cnt
;
390 buf
->f_bfree
= free
>> UBIFS_BLOCK_SHIFT
;
391 if (free
> c
->report_rp_size
)
392 buf
->f_bavail
= (free
- c
->report_rp_size
) >> UBIFS_BLOCK_SHIFT
;
397 buf
->f_namelen
= UBIFS_MAX_NLEN
;
398 buf
->f_fsid
.val
[0] = le32_to_cpu(uuid
[0]) ^ le32_to_cpu(uuid
[2]);
399 buf
->f_fsid
.val
[1] = le32_to_cpu(uuid
[1]) ^ le32_to_cpu(uuid
[3]);
403 static int ubifs_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
405 struct ubifs_info
*c
= mnt
->mnt_sb
->s_fs_info
;
407 if (c
->mount_opts
.unmount_mode
== 2)
408 seq_printf(s
, ",fast_unmount");
409 else if (c
->mount_opts
.unmount_mode
== 1)
410 seq_printf(s
, ",norm_unmount");
412 if (c
->mount_opts
.bulk_read
== 2)
413 seq_printf(s
, ",bulk_read");
414 else if (c
->mount_opts
.bulk_read
== 1)
415 seq_printf(s
, ",no_bulk_read");
417 if (c
->mount_opts
.chk_data_crc
== 2)
418 seq_printf(s
, ",chk_data_crc");
419 else if (c
->mount_opts
.chk_data_crc
== 1)
420 seq_printf(s
, ",no_chk_data_crc");
422 if (c
->mount_opts
.override_compr
) {
423 seq_printf(s
, ",compr=");
424 seq_printf(s
, ubifs_compr_name(c
->mount_opts
.compr_type
));
430 static int ubifs_sync_fs(struct super_block
*sb
, int wait
)
433 struct ubifs_info
*c
= sb
->s_fs_info
;
434 struct writeback_control wbc
= {
435 .sync_mode
= wait
? WB_SYNC_ALL
: WB_SYNC_HOLD
,
437 .range_end
= LLONG_MAX
,
438 .nr_to_write
= LONG_MAX
,
441 if (sb
->s_flags
& MS_RDONLY
)
445 * Synchronize write buffers, because 'ubifs_run_commit()' does not
446 * do this if it waits for an already running commit.
448 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
449 err
= ubifs_wbuf_sync(&c
->jheads
[i
].wbuf
);
455 * VFS calls '->sync_fs()' before synchronizing all dirty inodes and
456 * pages, so synchronize them first, then commit the journal. Strictly
457 * speaking, it is not necessary to commit the journal here,
458 * synchronizing write-buffers would be enough. But committing makes
459 * UBIFS free space predictions much more accurate, so we want to let
460 * the user be able to get more accurate results of 'statfs()' after
461 * they synchronize the file system.
463 generic_sync_sb_inodes(sb
, &wbc
);
465 err
= ubifs_run_commit(c
);
469 return ubi_sync(c
->vi
.ubi_num
);
473 * init_constants_early - initialize UBIFS constants.
474 * @c: UBIFS file-system description object
476 * This function initialize UBIFS constants which do not need the superblock to
477 * be read. It also checks that the UBI volume satisfies basic UBIFS
478 * requirements. Returns zero in case of success and a negative error code in
481 static int init_constants_early(struct ubifs_info
*c
)
483 if (c
->vi
.corrupted
) {
484 ubifs_warn("UBI volume is corrupted - read-only mode");
489 ubifs_msg("read-only UBI device");
493 if (c
->vi
.vol_type
== UBI_STATIC_VOLUME
) {
494 ubifs_msg("static UBI volume - read-only mode");
498 c
->leb_cnt
= c
->vi
.size
;
499 c
->leb_size
= c
->vi
.usable_leb_size
;
500 c
->half_leb_size
= c
->leb_size
/ 2;
501 c
->min_io_size
= c
->di
.min_io_size
;
502 c
->min_io_shift
= fls(c
->min_io_size
) - 1;
504 if (c
->leb_size
< UBIFS_MIN_LEB_SZ
) {
505 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
506 c
->leb_size
, UBIFS_MIN_LEB_SZ
);
510 if (c
->leb_cnt
< UBIFS_MIN_LEB_CNT
) {
511 ubifs_err("too few LEBs (%d), min. is %d",
512 c
->leb_cnt
, UBIFS_MIN_LEB_CNT
);
516 if (!is_power_of_2(c
->min_io_size
)) {
517 ubifs_err("bad min. I/O size %d", c
->min_io_size
);
522 * UBIFS aligns all node to 8-byte boundary, so to make function in
523 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
526 if (c
->min_io_size
< 8) {
531 c
->ref_node_alsz
= ALIGN(UBIFS_REF_NODE_SZ
, c
->min_io_size
);
532 c
->mst_node_alsz
= ALIGN(UBIFS_MST_NODE_SZ
, c
->min_io_size
);
535 * Initialize node length ranges which are mostly needed for node
538 c
->ranges
[UBIFS_PAD_NODE
].len
= UBIFS_PAD_NODE_SZ
;
539 c
->ranges
[UBIFS_SB_NODE
].len
= UBIFS_SB_NODE_SZ
;
540 c
->ranges
[UBIFS_MST_NODE
].len
= UBIFS_MST_NODE_SZ
;
541 c
->ranges
[UBIFS_REF_NODE
].len
= UBIFS_REF_NODE_SZ
;
542 c
->ranges
[UBIFS_TRUN_NODE
].len
= UBIFS_TRUN_NODE_SZ
;
543 c
->ranges
[UBIFS_CS_NODE
].len
= UBIFS_CS_NODE_SZ
;
545 c
->ranges
[UBIFS_INO_NODE
].min_len
= UBIFS_INO_NODE_SZ
;
546 c
->ranges
[UBIFS_INO_NODE
].max_len
= UBIFS_MAX_INO_NODE_SZ
;
547 c
->ranges
[UBIFS_ORPH_NODE
].min_len
=
548 UBIFS_ORPH_NODE_SZ
+ sizeof(__le64
);
549 c
->ranges
[UBIFS_ORPH_NODE
].max_len
= c
->leb_size
;
550 c
->ranges
[UBIFS_DENT_NODE
].min_len
= UBIFS_DENT_NODE_SZ
;
551 c
->ranges
[UBIFS_DENT_NODE
].max_len
= UBIFS_MAX_DENT_NODE_SZ
;
552 c
->ranges
[UBIFS_XENT_NODE
].min_len
= UBIFS_XENT_NODE_SZ
;
553 c
->ranges
[UBIFS_XENT_NODE
].max_len
= UBIFS_MAX_XENT_NODE_SZ
;
554 c
->ranges
[UBIFS_DATA_NODE
].min_len
= UBIFS_DATA_NODE_SZ
;
555 c
->ranges
[UBIFS_DATA_NODE
].max_len
= UBIFS_MAX_DATA_NODE_SZ
;
557 * Minimum indexing node size is amended later when superblock is
558 * read and the key length is known.
560 c
->ranges
[UBIFS_IDX_NODE
].min_len
= UBIFS_IDX_NODE_SZ
+ UBIFS_BRANCH_SZ
;
562 * Maximum indexing node size is amended later when superblock is
563 * read and the fanout is known.
565 c
->ranges
[UBIFS_IDX_NODE
].max_len
= INT_MAX
;
568 * Initialize dead and dark LEB space watermarks.
570 * Dead space is the space which cannot be used. Its watermark is
571 * equivalent to min. I/O unit or minimum node size if it is greater
572 * then min. I/O unit.
574 * Dark space is the space which might be used, or might not, depending
575 * on which node should be written to the LEB. Its watermark is
576 * equivalent to maximum UBIFS node size.
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
);
705 * init_constants_master - initialize UBIFS constants.
706 * @c: UBIFS file-system description object
708 * This is a helper function which initializes various UBIFS constants after
709 * the master node has been read. It also checks various UBIFS parameters and
710 * makes sure they are all right.
712 static void init_constants_master(struct ubifs_info
*c
)
716 c
->min_idx_lebs
= ubifs_calc_min_idx_lebs(c
);
719 * Calculate total amount of FS blocks. This number is not used
720 * internally because it does not make much sense for UBIFS, but it is
721 * necessary to report something for the 'statfs()' call.
723 * Subtract the LEB reserved for GC, the LEB which is reserved for
724 * deletions, minimum LEBs for the index, and assume only one journal
727 tmp64
= c
->main_lebs
- 1 - 1 - MIN_INDEX_LEBS
- c
->jhead_cnt
+ 1;
728 tmp64
*= (long long)c
->leb_size
- c
->leb_overhead
;
729 tmp64
= ubifs_reported_space(c
, tmp64
);
730 c
->block_cnt
= tmp64
>> UBIFS_BLOCK_SHIFT
;
734 * take_gc_lnum - reserve GC LEB.
735 * @c: UBIFS file-system description object
737 * This function ensures that the LEB reserved for garbage collection is
738 * unmapped and is marked as "taken" in lprops. We also have to set free space
739 * to LEB size and dirty space to zero, because lprops may contain out-of-date
740 * information if the file-system was un-mounted before it has been committed.
741 * This function returns zero in case of success and a negative error code in
744 static int take_gc_lnum(struct ubifs_info
*c
)
748 if (c
->gc_lnum
== -1) {
749 ubifs_err("no LEB for GC");
753 err
= ubifs_leb_unmap(c
, c
->gc_lnum
);
757 /* And we have to tell lprops that this LEB is taken */
758 err
= ubifs_change_one_lp(c
, c
->gc_lnum
, c
->leb_size
, 0,
764 * alloc_wbufs - allocate write-buffers.
765 * @c: UBIFS file-system description object
767 * This helper function allocates and initializes UBIFS write-buffers. Returns
768 * zero in case of success and %-ENOMEM in case of failure.
770 static int alloc_wbufs(struct ubifs_info
*c
)
774 c
->jheads
= kzalloc(c
->jhead_cnt
* sizeof(struct ubifs_jhead
),
779 /* Initialize journal heads */
780 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
781 INIT_LIST_HEAD(&c
->jheads
[i
].buds_list
);
782 err
= ubifs_wbuf_init(c
, &c
->jheads
[i
].wbuf
);
786 c
->jheads
[i
].wbuf
.sync_callback
= &bud_wbuf_callback
;
787 c
->jheads
[i
].wbuf
.jhead
= i
;
790 c
->jheads
[BASEHD
].wbuf
.dtype
= UBI_SHORTTERM
;
792 * Garbage Collector head likely contains long-term data and
793 * does not need to be synchronized by timer.
795 c
->jheads
[GCHD
].wbuf
.dtype
= UBI_LONGTERM
;
796 c
->jheads
[GCHD
].wbuf
.timeout
= 0;
802 * free_wbufs - free write-buffers.
803 * @c: UBIFS file-system description object
805 static void free_wbufs(struct ubifs_info
*c
)
810 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
811 kfree(c
->jheads
[i
].wbuf
.buf
);
812 kfree(c
->jheads
[i
].wbuf
.inodes
);
820 * free_orphans - free orphans.
821 * @c: UBIFS file-system description object
823 static void free_orphans(struct ubifs_info
*c
)
825 struct ubifs_orphan
*orph
;
827 while (c
->orph_dnext
) {
828 orph
= c
->orph_dnext
;
829 c
->orph_dnext
= orph
->dnext
;
830 list_del(&orph
->list
);
834 while (!list_empty(&c
->orph_list
)) {
835 orph
= list_entry(c
->orph_list
.next
, struct ubifs_orphan
, list
);
836 list_del(&orph
->list
);
838 dbg_err("orphan list not empty at unmount");
846 * free_buds - free per-bud objects.
847 * @c: UBIFS file-system description object
849 static void free_buds(struct ubifs_info
*c
)
851 struct rb_node
*this = c
->buds
.rb_node
;
852 struct ubifs_bud
*bud
;
856 this = this->rb_left
;
857 else if (this->rb_right
)
858 this = this->rb_right
;
860 bud
= rb_entry(this, struct ubifs_bud
, rb
);
861 this = rb_parent(this);
863 if (this->rb_left
== &bud
->rb
)
864 this->rb_left
= NULL
;
866 this->rb_right
= NULL
;
874 * check_volume_empty - check if the UBI volume is empty.
875 * @c: UBIFS file-system description object
877 * This function checks if the UBIFS volume is empty by looking if its LEBs are
878 * mapped or not. The result of checking is stored in the @c->empty variable.
879 * Returns zero in case of success and a negative error code in case of
882 static int check_volume_empty(struct ubifs_info
*c
)
887 for (lnum
= 0; lnum
< c
->leb_cnt
; lnum
++) {
888 err
= ubi_is_mapped(c
->ubi
, lnum
);
889 if (unlikely(err
< 0))
903 * UBIFS mount options.
905 * Opt_fast_unmount: do not run a journal commit before un-mounting
906 * Opt_norm_unmount: run a journal commit before un-mounting
907 * Opt_bulk_read: enable bulk-reads
908 * Opt_no_bulk_read: disable bulk-reads
909 * Opt_chk_data_crc: check CRCs when reading data nodes
910 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
911 * Opt_override_compr: override default compressor
912 * Opt_err: just end of array marker
925 static const match_table_t tokens
= {
926 {Opt_fast_unmount
, "fast_unmount"},
927 {Opt_norm_unmount
, "norm_unmount"},
928 {Opt_bulk_read
, "bulk_read"},
929 {Opt_no_bulk_read
, "no_bulk_read"},
930 {Opt_chk_data_crc
, "chk_data_crc"},
931 {Opt_no_chk_data_crc
, "no_chk_data_crc"},
932 {Opt_override_compr
, "compr=%s"},
937 * ubifs_parse_options - parse mount parameters.
938 * @c: UBIFS file-system description object
939 * @options: parameters to parse
940 * @is_remount: non-zero if this is FS re-mount
942 * This function parses UBIFS mount options and returns zero in case success
943 * and a negative error code in case of failure.
945 static int ubifs_parse_options(struct ubifs_info
*c
, char *options
,
949 substring_t args
[MAX_OPT_ARGS
];
954 while ((p
= strsep(&options
, ","))) {
960 token
= match_token(p
, tokens
, args
);
962 case Opt_fast_unmount
:
963 c
->mount_opts
.unmount_mode
= 2;
966 case Opt_norm_unmount
:
967 c
->mount_opts
.unmount_mode
= 1;
971 c
->mount_opts
.bulk_read
= 2;
974 case Opt_no_bulk_read
:
975 c
->mount_opts
.bulk_read
= 1;
978 case Opt_chk_data_crc
:
979 c
->mount_opts
.chk_data_crc
= 2;
980 c
->no_chk_data_crc
= 0;
982 case Opt_no_chk_data_crc
:
983 c
->mount_opts
.chk_data_crc
= 1;
984 c
->no_chk_data_crc
= 1;
986 case Opt_override_compr
:
988 char *name
= match_strdup(&args
[0]);
992 if (!strcmp(name
, "none"))
993 c
->mount_opts
.compr_type
= UBIFS_COMPR_NONE
;
994 else if (!strcmp(name
, "lzo"))
995 c
->mount_opts
.compr_type
= UBIFS_COMPR_LZO
;
996 else if (!strcmp(name
, "zlib"))
997 c
->mount_opts
.compr_type
= UBIFS_COMPR_ZLIB
;
999 ubifs_err("unknown compressor \"%s\"", name
);
1004 c
->mount_opts
.override_compr
= 1;
1005 c
->default_compr
= c
->mount_opts
.compr_type
;
1009 ubifs_err("unrecognized mount option \"%s\" "
1010 "or missing value", p
);
1019 * destroy_journal - destroy journal data structures.
1020 * @c: UBIFS file-system description object
1022 * This function destroys journal data structures including those that may have
1023 * been created by recovery functions.
1025 static void destroy_journal(struct ubifs_info
*c
)
1027 while (!list_empty(&c
->unclean_leb_list
)) {
1028 struct ubifs_unclean_leb
*ucleb
;
1030 ucleb
= list_entry(c
->unclean_leb_list
.next
,
1031 struct ubifs_unclean_leb
, list
);
1032 list_del(&ucleb
->list
);
1035 while (!list_empty(&c
->old_buds
)) {
1036 struct ubifs_bud
*bud
;
1038 bud
= list_entry(c
->old_buds
.next
, struct ubifs_bud
, list
);
1039 list_del(&bud
->list
);
1042 ubifs_destroy_idx_gc(c
);
1043 ubifs_destroy_size_tree(c
);
1049 * bu_init - initialize bulk-read information.
1050 * @c: UBIFS file-system description object
1052 static void bu_init(struct ubifs_info
*c
)
1054 ubifs_assert(c
->bulk_read
== 1);
1057 return; /* Already initialized */
1060 c
->bu
.buf
= kmalloc(c
->max_bu_buf_len
, GFP_KERNEL
| __GFP_NOWARN
);
1062 if (c
->max_bu_buf_len
> UBIFS_KMALLOC_OK
) {
1063 c
->max_bu_buf_len
= UBIFS_KMALLOC_OK
;
1067 /* Just disable bulk-read */
1068 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1069 "disabling it", c
->max_bu_buf_len
);
1070 c
->mount_opts
.bulk_read
= 1;
1077 * mount_ubifs - mount UBIFS file-system.
1078 * @c: UBIFS file-system description object
1080 * This function mounts UBIFS file system. Returns zero in case of success and
1081 * a negative error code in case of failure.
1083 * Note, the function does not de-allocate resources it it fails half way
1084 * through, and the caller has to do this instead.
1086 static int mount_ubifs(struct ubifs_info
*c
)
1088 struct super_block
*sb
= c
->vfs_sb
;
1089 int err
, mounted_read_only
= (sb
->s_flags
& MS_RDONLY
);
1093 err
= init_constants_early(c
);
1097 err
= ubifs_debugging_init(c
);
1101 err
= check_volume_empty(c
);
1105 if (c
->empty
&& (mounted_read_only
|| c
->ro_media
)) {
1107 * This UBI volume is empty, and read-only, or the file system
1108 * is mounted read-only - we cannot format it.
1110 ubifs_err("can't format empty UBI volume: read-only %s",
1111 c
->ro_media
? "UBI volume" : "mount");
1116 if (c
->ro_media
&& !mounted_read_only
) {
1117 ubifs_err("cannot mount read-write - read-only media");
1123 * The requirement for the buffer is that it should fit indexing B-tree
1124 * height amount of integers. We assume the height if the TNC tree will
1128 c
->bottom_up_buf
= kmalloc(BOTTOM_UP_HEIGHT
* sizeof(int), GFP_KERNEL
);
1129 if (!c
->bottom_up_buf
)
1132 c
->sbuf
= vmalloc(c
->leb_size
);
1136 if (!mounted_read_only
) {
1137 c
->ileb_buf
= vmalloc(c
->leb_size
);
1142 if (c
->bulk_read
== 1)
1146 * We have to check all CRCs, even for data nodes, when we mount the FS
1147 * (specifically, when we are replaying).
1149 c
->always_chk_crc
= 1;
1151 err
= ubifs_read_superblock(c
);
1156 * Make sure the compressor which is set as default in the superblock
1157 * or overriden by mount options is actually compiled in.
1159 if (!ubifs_compr_present(c
->default_compr
)) {
1160 ubifs_err("'compressor \"%s\" is not compiled in",
1161 ubifs_compr_name(c
->default_compr
));
1165 err
= init_constants_sb(c
);
1169 sz
= ALIGN(c
->max_idx_node_sz
, c
->min_io_size
);
1170 sz
= ALIGN(sz
+ c
->max_idx_node_sz
, c
->min_io_size
);
1171 c
->cbuf
= kmalloc(sz
, GFP_NOFS
);
1177 sprintf(c
->bgt_name
, BGT_NAME_PATTERN
, c
->vi
.ubi_num
, c
->vi
.vol_id
);
1178 if (!mounted_read_only
) {
1179 err
= alloc_wbufs(c
);
1183 /* Create background thread */
1184 c
->bgt
= kthread_create(ubifs_bg_thread
, c
, c
->bgt_name
);
1185 if (IS_ERR(c
->bgt
)) {
1186 err
= PTR_ERR(c
->bgt
);
1188 ubifs_err("cannot spawn \"%s\", error %d",
1192 wake_up_process(c
->bgt
);
1195 err
= ubifs_read_master(c
);
1199 init_constants_master(c
);
1201 if ((c
->mst_node
->flags
& cpu_to_le32(UBIFS_MST_DIRTY
)) != 0) {
1202 ubifs_msg("recovery needed");
1203 c
->need_recovery
= 1;
1204 if (!mounted_read_only
) {
1205 err
= ubifs_recover_inl_heads(c
, c
->sbuf
);
1209 } else if (!mounted_read_only
) {
1211 * Set the "dirty" flag so that if we reboot uncleanly we
1212 * will notice this immediately on the next mount.
1214 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_DIRTY
);
1215 err
= ubifs_write_master(c
);
1220 err
= ubifs_lpt_init(c
, 1, !mounted_read_only
);
1224 err
= dbg_check_idx_size(c
, c
->old_idx_sz
);
1228 err
= ubifs_replay_journal(c
);
1232 err
= ubifs_mount_orphans(c
, c
->need_recovery
, mounted_read_only
);
1236 if (!mounted_read_only
) {
1239 /* Check for enough free space */
1240 if (ubifs_calc_available(c
, c
->min_idx_lebs
) <= 0) {
1241 ubifs_err("insufficient available space");
1246 /* Check for enough log space */
1247 lnum
= c
->lhead_lnum
+ 1;
1248 if (lnum
>= UBIFS_LOG_LNUM
+ c
->log_lebs
)
1249 lnum
= UBIFS_LOG_LNUM
;
1250 if (lnum
== c
->ltail_lnum
) {
1251 err
= ubifs_consolidate_log(c
);
1256 if (c
->need_recovery
) {
1257 err
= ubifs_recover_size(c
);
1260 err
= ubifs_rcvry_gc_commit(c
);
1262 err
= take_gc_lnum(c
);
1266 err
= dbg_check_lprops(c
);
1269 } else if (c
->need_recovery
) {
1270 err
= ubifs_recover_size(c
);
1275 spin_lock(&ubifs_infos_lock
);
1276 list_add_tail(&c
->infos_list
, &ubifs_infos
);
1277 spin_unlock(&ubifs_infos_lock
);
1279 if (c
->need_recovery
) {
1280 if (mounted_read_only
)
1281 ubifs_msg("recovery deferred");
1283 c
->need_recovery
= 0;
1284 ubifs_msg("recovery completed");
1288 err
= dbg_debugfs_init_fs(c
);
1292 err
= dbg_check_filesystem(c
);
1296 c
->always_chk_crc
= 0;
1298 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1299 c
->vi
.ubi_num
, c
->vi
.vol_id
, c
->vi
.name
);
1300 if (mounted_read_only
)
1301 ubifs_msg("mounted read-only");
1302 x
= (long long)c
->main_lebs
* c
->leb_size
;
1303 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1304 "LEBs)", x
, x
>> 10, x
>> 20, c
->main_lebs
);
1305 x
= (long long)c
->log_lebs
* c
->leb_size
+ c
->max_bud_bytes
;
1306 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1307 "LEBs)", x
, x
>> 10, x
>> 20, c
->log_lebs
+ c
->max_bud_cnt
);
1308 ubifs_msg("media format: %d (latest is %d)",
1309 c
->fmt_version
, UBIFS_FORMAT_VERSION
);
1310 ubifs_msg("default compressor: %s", ubifs_compr_name(c
->default_compr
));
1311 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1312 c
->report_rp_size
, c
->report_rp_size
>> 10);
1314 dbg_msg("compiled on: " __DATE__
" at " __TIME__
);
1315 dbg_msg("min. I/O unit size: %d bytes", c
->min_io_size
);
1316 dbg_msg("LEB size: %d bytes (%d KiB)",
1317 c
->leb_size
, c
->leb_size
>> 10);
1318 dbg_msg("data journal heads: %d",
1319 c
->jhead_cnt
- NONDATA_JHEADS_CNT
);
1320 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1321 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1322 c
->uuid
[0], c
->uuid
[1], c
->uuid
[2], c
->uuid
[3],
1323 c
->uuid
[4], c
->uuid
[5], c
->uuid
[6], c
->uuid
[7],
1324 c
->uuid
[8], c
->uuid
[9], c
->uuid
[10], c
->uuid
[11],
1325 c
->uuid
[12], c
->uuid
[13], c
->uuid
[14], c
->uuid
[15]);
1326 dbg_msg("fast unmount: %d", c
->fast_unmount
);
1327 dbg_msg("big_lpt %d", c
->big_lpt
);
1328 dbg_msg("log LEBs: %d (%d - %d)",
1329 c
->log_lebs
, UBIFS_LOG_LNUM
, c
->log_last
);
1330 dbg_msg("LPT area LEBs: %d (%d - %d)",
1331 c
->lpt_lebs
, c
->lpt_first
, c
->lpt_last
);
1332 dbg_msg("orphan area LEBs: %d (%d - %d)",
1333 c
->orph_lebs
, c
->orph_first
, c
->orph_last
);
1334 dbg_msg("main area LEBs: %d (%d - %d)",
1335 c
->main_lebs
, c
->main_first
, c
->leb_cnt
- 1);
1336 dbg_msg("index LEBs: %d", c
->lst
.idx_lebs
);
1337 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1338 c
->old_idx_sz
, c
->old_idx_sz
>> 10, c
->old_idx_sz
>> 20);
1339 dbg_msg("key hash type: %d", c
->key_hash_type
);
1340 dbg_msg("tree fanout: %d", c
->fanout
);
1341 dbg_msg("reserved GC LEB: %d", c
->gc_lnum
);
1342 dbg_msg("first main LEB: %d", c
->main_first
);
1343 dbg_msg("dead watermark: %d", c
->dead_wm
);
1344 dbg_msg("dark watermark: %d", c
->dark_wm
);
1345 x
= (long long)c
->main_lebs
* c
->dark_wm
;
1346 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1347 x
, x
>> 10, x
>> 20);
1348 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1349 c
->max_bud_bytes
, c
->max_bud_bytes
>> 10,
1350 c
->max_bud_bytes
>> 20);
1351 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1352 c
->bg_bud_bytes
, c
->bg_bud_bytes
>> 10,
1353 c
->bg_bud_bytes
>> 20);
1354 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1355 c
->bud_bytes
, c
->bud_bytes
>> 10, c
->bud_bytes
>> 20);
1356 dbg_msg("max. seq. number: %llu", c
->max_sqnum
);
1357 dbg_msg("commit number: %llu", c
->cmt_no
);
1362 spin_lock(&ubifs_infos_lock
);
1363 list_del(&c
->infos_list
);
1364 spin_unlock(&ubifs_infos_lock
);
1370 ubifs_lpt_free(c
, 0);
1373 kfree(c
->rcvrd_mst_node
);
1375 kthread_stop(c
->bgt
);
1384 kfree(c
->bottom_up_buf
);
1385 ubifs_debugging_exit(c
);
1390 * ubifs_umount - un-mount UBIFS file-system.
1391 * @c: UBIFS file-system description object
1393 * Note, this function is called to free allocated resourced when un-mounting,
1394 * as well as free resources when an error occurred while we were half way
1395 * through mounting (error path cleanup function). So it has to make sure the
1396 * resource was actually allocated before freeing it.
1398 static void ubifs_umount(struct ubifs_info
*c
)
1400 dbg_gen("un-mounting UBI device %d, volume %d", c
->vi
.ubi_num
,
1403 dbg_debugfs_exit_fs(c
);
1404 spin_lock(&ubifs_infos_lock
);
1405 list_del(&c
->infos_list
);
1406 spin_unlock(&ubifs_infos_lock
);
1409 kthread_stop(c
->bgt
);
1414 ubifs_lpt_free(c
, 0);
1417 kfree(c
->rcvrd_mst_node
);
1422 kfree(c
->bottom_up_buf
);
1423 ubifs_debugging_exit(c
);
1427 * ubifs_remount_rw - re-mount in read-write mode.
1428 * @c: UBIFS file-system description object
1430 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1431 * mode. This function allocates the needed resources and re-mounts UBIFS in
1434 static int ubifs_remount_rw(struct ubifs_info
*c
)
1441 mutex_lock(&c
->umount_mutex
);
1442 c
->remounting_rw
= 1;
1443 c
->always_chk_crc
= 1;
1445 /* Check for enough free space */
1446 if (ubifs_calc_available(c
, c
->min_idx_lebs
) <= 0) {
1447 ubifs_err("insufficient available space");
1452 if (c
->old_leb_cnt
!= c
->leb_cnt
) {
1453 struct ubifs_sb_node
*sup
;
1455 sup
= ubifs_read_sb_node(c
);
1460 sup
->leb_cnt
= cpu_to_le32(c
->leb_cnt
);
1461 err
= ubifs_write_sb_node(c
, sup
);
1466 if (c
->need_recovery
) {
1467 ubifs_msg("completing deferred recovery");
1468 err
= ubifs_write_rcvrd_mst_node(c
);
1471 err
= ubifs_recover_size(c
);
1474 err
= ubifs_clean_lebs(c
, c
->sbuf
);
1477 err
= ubifs_recover_inl_heads(c
, c
->sbuf
);
1482 if (!(c
->mst_node
->flags
& cpu_to_le32(UBIFS_MST_DIRTY
))) {
1483 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_DIRTY
);
1484 err
= ubifs_write_master(c
);
1489 c
->ileb_buf
= vmalloc(c
->leb_size
);
1495 err
= ubifs_lpt_init(c
, 0, 1);
1499 err
= alloc_wbufs(c
);
1503 ubifs_create_buds_lists(c
);
1505 /* Create background thread */
1506 c
->bgt
= kthread_create(ubifs_bg_thread
, c
, c
->bgt_name
);
1507 if (IS_ERR(c
->bgt
)) {
1508 err
= PTR_ERR(c
->bgt
);
1510 ubifs_err("cannot spawn \"%s\", error %d",
1514 wake_up_process(c
->bgt
);
1516 c
->orph_buf
= vmalloc(c
->leb_size
);
1522 /* Check for enough log space */
1523 lnum
= c
->lhead_lnum
+ 1;
1524 if (lnum
>= UBIFS_LOG_LNUM
+ c
->log_lebs
)
1525 lnum
= UBIFS_LOG_LNUM
;
1526 if (lnum
== c
->ltail_lnum
) {
1527 err
= ubifs_consolidate_log(c
);
1532 if (c
->need_recovery
)
1533 err
= ubifs_rcvry_gc_commit(c
);
1535 err
= take_gc_lnum(c
);
1539 if (c
->need_recovery
) {
1540 c
->need_recovery
= 0;
1541 ubifs_msg("deferred recovery completed");
1544 dbg_gen("re-mounted read-write");
1545 c
->vfs_sb
->s_flags
&= ~MS_RDONLY
;
1546 c
->remounting_rw
= 0;
1547 c
->always_chk_crc
= 0;
1548 mutex_unlock(&c
->umount_mutex
);
1555 kthread_stop(c
->bgt
);
1561 ubifs_lpt_free(c
, 1);
1562 c
->remounting_rw
= 0;
1563 c
->always_chk_crc
= 0;
1564 mutex_unlock(&c
->umount_mutex
);
1569 * commit_on_unmount - commit the journal when un-mounting.
1570 * @c: UBIFS file-system description object
1572 * This function is called during un-mounting and re-mounting, and it commits
1573 * the journal unless the "fast unmount" mode is enabled. It also avoids
1574 * committing the journal if it contains too few data.
1576 static void commit_on_unmount(struct ubifs_info
*c
)
1578 if (!c
->fast_unmount
) {
1579 long long bud_bytes
;
1581 spin_lock(&c
->buds_lock
);
1582 bud_bytes
= c
->bud_bytes
;
1583 spin_unlock(&c
->buds_lock
);
1584 if (bud_bytes
> c
->leb_size
)
1585 ubifs_run_commit(c
);
1590 * ubifs_remount_ro - re-mount in read-only mode.
1591 * @c: UBIFS file-system description object
1593 * We rely on VFS to have stopped writing. Possibly the background thread could
1594 * be running a commit, however kthread_stop will wait in that case.
1596 static void ubifs_remount_ro(struct ubifs_info
*c
)
1600 ubifs_assert(!c
->need_recovery
);
1601 commit_on_unmount(c
);
1603 mutex_lock(&c
->umount_mutex
);
1605 kthread_stop(c
->bgt
);
1609 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
1610 ubifs_wbuf_sync(&c
->jheads
[i
].wbuf
);
1611 del_timer_sync(&c
->jheads
[i
].wbuf
.timer
);
1615 c
->mst_node
->flags
&= ~cpu_to_le32(UBIFS_MST_DIRTY
);
1616 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_NO_ORPHS
);
1617 c
->mst_node
->gc_lnum
= cpu_to_le32(c
->gc_lnum
);
1618 err
= ubifs_write_master(c
);
1620 ubifs_ro_mode(c
, err
);
1623 ubifs_destroy_idx_gc(c
);
1629 ubifs_lpt_free(c
, 1);
1630 mutex_unlock(&c
->umount_mutex
);
1633 static void ubifs_put_super(struct super_block
*sb
)
1636 struct ubifs_info
*c
= sb
->s_fs_info
;
1638 ubifs_msg("un-mount UBI device %d, volume %d", c
->vi
.ubi_num
,
1641 * The following asserts are only valid if there has not been a failure
1642 * of the media. For example, there will be dirty inodes if we failed
1643 * to write them back because of I/O errors.
1645 ubifs_assert(atomic_long_read(&c
->dirty_pg_cnt
) == 0);
1646 ubifs_assert(c
->budg_idx_growth
== 0);
1647 ubifs_assert(c
->budg_dd_growth
== 0);
1648 ubifs_assert(c
->budg_data_growth
== 0);
1651 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1652 * and file system un-mount. Namely, it prevents the shrinker from
1653 * picking this superblock for shrinking - it will be just skipped if
1654 * the mutex is locked.
1656 mutex_lock(&c
->umount_mutex
);
1657 if (!(c
->vfs_sb
->s_flags
& MS_RDONLY
)) {
1659 * First of all kill the background thread to make sure it does
1660 * not interfere with un-mounting and freeing resources.
1663 kthread_stop(c
->bgt
);
1667 /* Synchronize write-buffers */
1669 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
1670 ubifs_wbuf_sync(&c
->jheads
[i
].wbuf
);
1671 del_timer_sync(&c
->jheads
[i
].wbuf
.timer
);
1675 * On fatal errors c->ro_media is set to 1, in which case we do
1676 * not write the master node.
1680 * We are being cleanly unmounted which means the
1681 * orphans were killed - indicate this in the master
1682 * node. Also save the reserved GC LEB number.
1686 c
->mst_node
->flags
&= ~cpu_to_le32(UBIFS_MST_DIRTY
);
1687 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_NO_ORPHS
);
1688 c
->mst_node
->gc_lnum
= cpu_to_le32(c
->gc_lnum
);
1689 err
= ubifs_write_master(c
);
1692 * Recovery will attempt to fix the master area
1693 * next mount, so we just print a message and
1694 * continue to unmount normally.
1696 ubifs_err("failed to write master node, "
1702 bdi_destroy(&c
->bdi
);
1703 ubi_close_volume(c
->ubi
);
1704 mutex_unlock(&c
->umount_mutex
);
1708 static int ubifs_remount_fs(struct super_block
*sb
, int *flags
, char *data
)
1711 struct ubifs_info
*c
= sb
->s_fs_info
;
1713 dbg_gen("old flags %#lx, new flags %#x", sb
->s_flags
, *flags
);
1715 err
= ubifs_parse_options(c
, data
, 1);
1717 ubifs_err("invalid or unknown remount parameter");
1721 if ((sb
->s_flags
& MS_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1722 err
= ubifs_remount_rw(c
);
1725 } else if (!(sb
->s_flags
& MS_RDONLY
) && (*flags
& MS_RDONLY
))
1726 ubifs_remount_ro(c
);
1728 if (c
->bulk_read
== 1)
1731 dbg_gen("disable bulk-read");
1739 struct super_operations ubifs_super_operations
= {
1740 .alloc_inode
= ubifs_alloc_inode
,
1741 .destroy_inode
= ubifs_destroy_inode
,
1742 .put_super
= ubifs_put_super
,
1743 .write_inode
= ubifs_write_inode
,
1744 .delete_inode
= ubifs_delete_inode
,
1745 .statfs
= ubifs_statfs
,
1746 .dirty_inode
= ubifs_dirty_inode
,
1747 .remount_fs
= ubifs_remount_fs
,
1748 .show_options
= ubifs_show_options
,
1749 .sync_fs
= ubifs_sync_fs
,
1753 * open_ubi - parse UBI device name string and open the UBI device.
1754 * @name: UBI volume name
1755 * @mode: UBI volume open mode
1757 * There are several ways to specify UBI volumes when mounting UBIFS:
1758 * o ubiX_Y - UBI device number X, volume Y;
1759 * o ubiY - UBI device number 0, volume Y;
1760 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1761 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1763 * Alternative '!' separator may be used instead of ':' (because some shells
1764 * like busybox may interpret ':' as an NFS host name separator). This function
1765 * returns ubi volume object in case of success and a negative error code in
1768 static struct ubi_volume_desc
*open_ubi(const char *name
, int mode
)
1773 if (name
[0] != 'u' || name
[1] != 'b' || name
[2] != 'i')
1774 return ERR_PTR(-EINVAL
);
1776 /* ubi:NAME method */
1777 if ((name
[3] == ':' || name
[3] == '!') && name
[4] != '\0')
1778 return ubi_open_volume_nm(0, name
+ 4, mode
);
1780 if (!isdigit(name
[3]))
1781 return ERR_PTR(-EINVAL
);
1783 dev
= simple_strtoul(name
+ 3, &endptr
, 0);
1786 if (*endptr
== '\0')
1787 return ubi_open_volume(0, dev
, mode
);
1790 if (*endptr
== '_' && isdigit(endptr
[1])) {
1791 vol
= simple_strtoul(endptr
+ 1, &endptr
, 0);
1792 if (*endptr
!= '\0')
1793 return ERR_PTR(-EINVAL
);
1794 return ubi_open_volume(dev
, vol
, mode
);
1797 /* ubiX:NAME method */
1798 if ((*endptr
== ':' || *endptr
== '!') && endptr
[1] != '\0')
1799 return ubi_open_volume_nm(dev
, ++endptr
, mode
);
1801 return ERR_PTR(-EINVAL
);
1804 static int ubifs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1806 struct ubi_volume_desc
*ubi
= sb
->s_fs_info
;
1807 struct ubifs_info
*c
;
1811 c
= kzalloc(sizeof(struct ubifs_info
), GFP_KERNEL
);
1815 spin_lock_init(&c
->cnt_lock
);
1816 spin_lock_init(&c
->cs_lock
);
1817 spin_lock_init(&c
->buds_lock
);
1818 spin_lock_init(&c
->space_lock
);
1819 spin_lock_init(&c
->orphan_lock
);
1820 init_rwsem(&c
->commit_sem
);
1821 mutex_init(&c
->lp_mutex
);
1822 mutex_init(&c
->tnc_mutex
);
1823 mutex_init(&c
->log_mutex
);
1824 mutex_init(&c
->mst_mutex
);
1825 mutex_init(&c
->umount_mutex
);
1826 mutex_init(&c
->bu_mutex
);
1827 init_waitqueue_head(&c
->cmt_wq
);
1829 c
->old_idx
= RB_ROOT
;
1830 c
->size_tree
= RB_ROOT
;
1831 c
->orph_tree
= RB_ROOT
;
1832 INIT_LIST_HEAD(&c
->infos_list
);
1833 INIT_LIST_HEAD(&c
->idx_gc
);
1834 INIT_LIST_HEAD(&c
->replay_list
);
1835 INIT_LIST_HEAD(&c
->replay_buds
);
1836 INIT_LIST_HEAD(&c
->uncat_list
);
1837 INIT_LIST_HEAD(&c
->empty_list
);
1838 INIT_LIST_HEAD(&c
->freeable_list
);
1839 INIT_LIST_HEAD(&c
->frdi_idx_list
);
1840 INIT_LIST_HEAD(&c
->unclean_leb_list
);
1841 INIT_LIST_HEAD(&c
->old_buds
);
1842 INIT_LIST_HEAD(&c
->orph_list
);
1843 INIT_LIST_HEAD(&c
->orph_new
);
1845 c
->highest_inum
= UBIFS_FIRST_INO
;
1846 c
->lhead_lnum
= c
->ltail_lnum
= UBIFS_LOG_LNUM
;
1848 ubi_get_volume_info(ubi
, &c
->vi
);
1849 ubi_get_device_info(c
->vi
.ubi_num
, &c
->di
);
1851 /* Re-open the UBI device in read-write mode */
1852 c
->ubi
= ubi_open_volume(c
->vi
.ubi_num
, c
->vi
.vol_id
, UBI_READWRITE
);
1853 if (IS_ERR(c
->ubi
)) {
1854 err
= PTR_ERR(c
->ubi
);
1859 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1860 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1861 * which means the user would have to wait not just for their own I/O
1862 * but the read-ahead I/O as well i.e. completely pointless.
1864 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1866 c
->bdi
.capabilities
= BDI_CAP_MAP_COPY
;
1867 c
->bdi
.unplug_io_fn
= default_unplug_io_fn
;
1868 err
= bdi_init(&c
->bdi
);
1872 err
= ubifs_parse_options(c
, data
, 0);
1879 sb
->s_magic
= UBIFS_SUPER_MAGIC
;
1880 sb
->s_blocksize
= UBIFS_BLOCK_SIZE
;
1881 sb
->s_blocksize_bits
= UBIFS_BLOCK_SHIFT
;
1882 sb
->s_dev
= c
->vi
.cdev
;
1883 sb
->s_maxbytes
= c
->max_inode_sz
= key_max_inode_size(c
);
1884 if (c
->max_inode_sz
> MAX_LFS_FILESIZE
)
1885 sb
->s_maxbytes
= c
->max_inode_sz
= MAX_LFS_FILESIZE
;
1886 sb
->s_op
= &ubifs_super_operations
;
1888 mutex_lock(&c
->umount_mutex
);
1889 err
= mount_ubifs(c
);
1891 ubifs_assert(err
< 0);
1895 /* Read the root inode */
1896 root
= ubifs_iget(sb
, UBIFS_ROOT_INO
);
1898 err
= PTR_ERR(root
);
1902 sb
->s_root
= d_alloc_root(root
);
1906 mutex_unlock(&c
->umount_mutex
);
1914 mutex_unlock(&c
->umount_mutex
);
1916 bdi_destroy(&c
->bdi
);
1918 ubi_close_volume(c
->ubi
);
1924 static int sb_test(struct super_block
*sb
, void *data
)
1928 return sb
->s_dev
== *dev
;
1931 static int sb_set(struct super_block
*sb
, void *data
)
1939 static int ubifs_get_sb(struct file_system_type
*fs_type
, int flags
,
1940 const char *name
, void *data
, struct vfsmount
*mnt
)
1942 struct ubi_volume_desc
*ubi
;
1943 struct ubi_volume_info vi
;
1944 struct super_block
*sb
;
1947 dbg_gen("name %s, flags %#x", name
, flags
);
1950 * Get UBI device number and volume ID. Mount it read-only so far
1951 * because this might be a new mount point, and UBI allows only one
1952 * read-write user at a time.
1954 ubi
= open_ubi(name
, UBI_READONLY
);
1956 ubifs_err("cannot open \"%s\", error %d",
1957 name
, (int)PTR_ERR(ubi
));
1958 return PTR_ERR(ubi
);
1960 ubi_get_volume_info(ubi
, &vi
);
1962 dbg_gen("opened ubi%d_%d", vi
.ubi_num
, vi
.vol_id
);
1964 sb
= sget(fs_type
, &sb_test
, &sb_set
, &vi
.cdev
);
1971 /* A new mount point for already mounted UBIFS */
1972 dbg_gen("this ubi volume is already mounted");
1973 if ((flags
^ sb
->s_flags
) & MS_RDONLY
) {
1978 sb
->s_flags
= flags
;
1980 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
1983 sb
->s_fs_info
= ubi
;
1984 err
= ubifs_fill_super(sb
, data
, flags
& MS_SILENT
? 1 : 0);
1987 /* We do not support atime */
1988 sb
->s_flags
|= MS_ACTIVE
| MS_NOATIME
;
1991 /* 'fill_super()' opens ubi again so we must close it here */
1992 ubi_close_volume(ubi
);
1994 return simple_set_mnt(mnt
, sb
);
1997 up_write(&sb
->s_umount
);
1998 deactivate_super(sb
);
2000 ubi_close_volume(ubi
);
2004 static void ubifs_kill_sb(struct super_block
*sb
)
2006 struct ubifs_info
*c
= sb
->s_fs_info
;
2009 * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
2010 * in order to be outside BKL.
2012 if (sb
->s_root
&& !(sb
->s_flags
& MS_RDONLY
))
2013 commit_on_unmount(c
);
2014 /* The un-mount routine is actually done in put_super() */
2015 generic_shutdown_super(sb
);
2018 static struct file_system_type ubifs_fs_type
= {
2020 .owner
= THIS_MODULE
,
2021 .get_sb
= ubifs_get_sb
,
2022 .kill_sb
= ubifs_kill_sb
2026 * Inode slab cache constructor.
2028 static void inode_slab_ctor(void *obj
)
2030 struct ubifs_inode
*ui
= obj
;
2031 inode_init_once(&ui
->vfs_inode
);
2034 static int __init
ubifs_init(void)
2038 BUILD_BUG_ON(sizeof(struct ubifs_ch
) != 24);
2040 /* Make sure node sizes are 8-byte aligned */
2041 BUILD_BUG_ON(UBIFS_CH_SZ
& 7);
2042 BUILD_BUG_ON(UBIFS_INO_NODE_SZ
& 7);
2043 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ
& 7);
2044 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ
& 7);
2045 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ
& 7);
2046 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ
& 7);
2047 BUILD_BUG_ON(UBIFS_SB_NODE_SZ
& 7);
2048 BUILD_BUG_ON(UBIFS_MST_NODE_SZ
& 7);
2049 BUILD_BUG_ON(UBIFS_REF_NODE_SZ
& 7);
2050 BUILD_BUG_ON(UBIFS_CS_NODE_SZ
& 7);
2051 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ
& 7);
2053 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ
& 7);
2054 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ
& 7);
2055 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ
& 7);
2056 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ
& 7);
2057 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ
& 7);
2058 BUILD_BUG_ON(MIN_WRITE_SZ
& 7);
2060 /* Check min. node size */
2061 BUILD_BUG_ON(UBIFS_INO_NODE_SZ
< MIN_WRITE_SZ
);
2062 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ
< MIN_WRITE_SZ
);
2063 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ
< MIN_WRITE_SZ
);
2064 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ
< MIN_WRITE_SZ
);
2066 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2067 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2068 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2069 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2071 /* Defined node sizes */
2072 BUILD_BUG_ON(UBIFS_SB_NODE_SZ
!= 4096);
2073 BUILD_BUG_ON(UBIFS_MST_NODE_SZ
!= 512);
2074 BUILD_BUG_ON(UBIFS_INO_NODE_SZ
!= 160);
2075 BUILD_BUG_ON(UBIFS_REF_NODE_SZ
!= 64);
2078 * We use 2 bit wide bit-fields to store compression type, which should
2079 * be amended if more compressors are added. The bit-fields are:
2080 * @compr_type in 'struct ubifs_inode', @default_compr in
2081 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2083 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT
> 4);
2086 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2087 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2089 if (PAGE_CACHE_SIZE
< UBIFS_BLOCK_SIZE
) {
2090 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2091 " at least 4096 bytes",
2092 (unsigned int)PAGE_CACHE_SIZE
);
2096 err
= register_filesystem(&ubifs_fs_type
);
2098 ubifs_err("cannot register file system, error %d", err
);
2103 ubifs_inode_slab
= kmem_cache_create("ubifs_inode_slab",
2104 sizeof(struct ubifs_inode
), 0,
2105 SLAB_MEM_SPREAD
| SLAB_RECLAIM_ACCOUNT
,
2107 if (!ubifs_inode_slab
)
2110 register_shrinker(&ubifs_shrinker_info
);
2112 err
= ubifs_compressors_init();
2116 err
= dbg_debugfs_init();
2123 ubifs_compressors_exit();
2125 unregister_shrinker(&ubifs_shrinker_info
);
2126 kmem_cache_destroy(ubifs_inode_slab
);
2128 unregister_filesystem(&ubifs_fs_type
);
2131 /* late_initcall to let compressors initialize first */
2132 late_initcall(ubifs_init
);
2134 static void __exit
ubifs_exit(void)
2136 ubifs_assert(list_empty(&ubifs_infos
));
2137 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt
) == 0);
2140 ubifs_compressors_exit();
2141 unregister_shrinker(&ubifs_shrinker_info
);
2142 kmem_cache_destroy(ubifs_inode_slab
);
2143 unregister_filesystem(&ubifs_fs_type
);
2145 module_exit(ubifs_exit
);
2147 MODULE_LICENSE("GPL");
2148 MODULE_VERSION(__stringify(UBIFS_VERSION
));
2149 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2150 MODULE_DESCRIPTION("UBIFS - UBI File System");