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>
41 * Maximum amount of memory we may 'kmalloc()' without worrying that we are
42 * allocating too much.
44 #define UBIFS_KMALLOC_OK (128*1024)
46 /* Slab cache for UBIFS inodes */
47 struct kmem_cache
*ubifs_inode_slab
;
49 /* UBIFS TNC shrinker description */
50 static struct shrinker ubifs_shrinker_info
= {
51 .shrink
= ubifs_shrinker
,
52 .seeks
= DEFAULT_SEEKS
,
56 * validate_inode - validate inode.
57 * @c: UBIFS file-system description object
58 * @inode: the inode to validate
60 * This is a helper function for 'ubifs_iget()' which validates various fields
61 * of a newly built inode to make sure they contain sane values and prevent
62 * possible vulnerabilities. Returns zero if the inode is all right and
63 * a non-zero error code if not.
65 static int validate_inode(struct ubifs_info
*c
, const struct inode
*inode
)
68 const struct ubifs_inode
*ui
= ubifs_inode(inode
);
70 if (inode
->i_size
> c
->max_inode_sz
) {
71 ubifs_err("inode is too large (%lld)",
72 (long long)inode
->i_size
);
76 if (ui
->compr_type
< 0 || ui
->compr_type
>= UBIFS_COMPR_TYPES_CNT
) {
77 ubifs_err("unknown compression type %d", ui
->compr_type
);
81 if (ui
->xattr_names
+ ui
->xattr_cnt
> XATTR_LIST_MAX
)
84 if (ui
->data_len
< 0 || ui
->data_len
> UBIFS_MAX_INO_DATA
)
87 if (ui
->xattr
&& (inode
->i_mode
& S_IFMT
) != S_IFREG
)
90 if (!ubifs_compr_present(ui
->compr_type
)) {
91 ubifs_warn("inode %lu uses '%s' compression, but it was not "
92 "compiled in", inode
->i_ino
,
93 ubifs_compr_name(ui
->compr_type
));
96 err
= dbg_check_dir_size(c
, inode
);
100 struct inode
*ubifs_iget(struct super_block
*sb
, unsigned long inum
)
104 struct ubifs_ino_node
*ino
;
105 struct ubifs_info
*c
= sb
->s_fs_info
;
107 struct ubifs_inode
*ui
;
109 dbg_gen("inode %lu", inum
);
111 inode
= iget_locked(sb
, inum
);
113 return ERR_PTR(-ENOMEM
);
114 if (!(inode
->i_state
& I_NEW
))
116 ui
= ubifs_inode(inode
);
118 ino
= kmalloc(UBIFS_MAX_INO_NODE_SZ
, GFP_NOFS
);
124 ino_key_init(c
, &key
, inode
->i_ino
);
126 err
= ubifs_tnc_lookup(c
, &key
, ino
);
130 inode
->i_flags
|= (S_NOCMTIME
| S_NOATIME
);
131 inode
->i_nlink
= le32_to_cpu(ino
->nlink
);
132 inode
->i_uid
= le32_to_cpu(ino
->uid
);
133 inode
->i_gid
= le32_to_cpu(ino
->gid
);
134 inode
->i_atime
.tv_sec
= (int64_t)le64_to_cpu(ino
->atime_sec
);
135 inode
->i_atime
.tv_nsec
= le32_to_cpu(ino
->atime_nsec
);
136 inode
->i_mtime
.tv_sec
= (int64_t)le64_to_cpu(ino
->mtime_sec
);
137 inode
->i_mtime
.tv_nsec
= le32_to_cpu(ino
->mtime_nsec
);
138 inode
->i_ctime
.tv_sec
= (int64_t)le64_to_cpu(ino
->ctime_sec
);
139 inode
->i_ctime
.tv_nsec
= le32_to_cpu(ino
->ctime_nsec
);
140 inode
->i_mode
= le32_to_cpu(ino
->mode
);
141 inode
->i_size
= le64_to_cpu(ino
->size
);
143 ui
->data_len
= le32_to_cpu(ino
->data_len
);
144 ui
->flags
= le32_to_cpu(ino
->flags
);
145 ui
->compr_type
= le16_to_cpu(ino
->compr_type
);
146 ui
->creat_sqnum
= le64_to_cpu(ino
->creat_sqnum
);
147 ui
->xattr_cnt
= le32_to_cpu(ino
->xattr_cnt
);
148 ui
->xattr_size
= le32_to_cpu(ino
->xattr_size
);
149 ui
->xattr_names
= le32_to_cpu(ino
->xattr_names
);
150 ui
->synced_i_size
= ui
->ui_size
= inode
->i_size
;
152 ui
->xattr
= (ui
->flags
& UBIFS_XATTR_FL
) ? 1 : 0;
154 err
= validate_inode(c
, inode
);
158 /* Disable read-ahead */
159 inode
->i_mapping
->backing_dev_info
= &c
->bdi
;
161 switch (inode
->i_mode
& S_IFMT
) {
163 inode
->i_mapping
->a_ops
= &ubifs_file_address_operations
;
164 inode
->i_op
= &ubifs_file_inode_operations
;
165 inode
->i_fop
= &ubifs_file_operations
;
167 ui
->data
= kmalloc(ui
->data_len
+ 1, GFP_NOFS
);
172 memcpy(ui
->data
, ino
->data
, ui
->data_len
);
173 ((char *)ui
->data
)[ui
->data_len
] = '\0';
174 } else if (ui
->data_len
!= 0) {
180 inode
->i_op
= &ubifs_dir_inode_operations
;
181 inode
->i_fop
= &ubifs_dir_operations
;
182 if (ui
->data_len
!= 0) {
188 inode
->i_op
= &ubifs_symlink_inode_operations
;
189 if (ui
->data_len
<= 0 || ui
->data_len
> UBIFS_MAX_INO_DATA
) {
193 ui
->data
= kmalloc(ui
->data_len
+ 1, GFP_NOFS
);
198 memcpy(ui
->data
, ino
->data
, ui
->data_len
);
199 ((char *)ui
->data
)[ui
->data_len
] = '\0';
205 union ubifs_dev_desc
*dev
;
207 ui
->data
= kmalloc(sizeof(union ubifs_dev_desc
), GFP_NOFS
);
213 dev
= (union ubifs_dev_desc
*)ino
->data
;
214 if (ui
->data_len
== sizeof(dev
->new))
215 rdev
= new_decode_dev(le32_to_cpu(dev
->new));
216 else if (ui
->data_len
== sizeof(dev
->huge
))
217 rdev
= huge_decode_dev(le64_to_cpu(dev
->huge
));
222 memcpy(ui
->data
, ino
->data
, ui
->data_len
);
223 inode
->i_op
= &ubifs_file_inode_operations
;
224 init_special_inode(inode
, inode
->i_mode
, rdev
);
229 inode
->i_op
= &ubifs_file_inode_operations
;
230 init_special_inode(inode
, inode
->i_mode
, 0);
231 if (ui
->data_len
!= 0) {
242 ubifs_set_inode_flags(inode
);
243 unlock_new_inode(inode
);
247 ubifs_err("inode %lu validation failed, error %d", inode
->i_ino
, err
);
248 dbg_dump_node(c
, ino
);
249 dbg_dump_inode(c
, inode
);
254 ubifs_err("failed to read inode %lu, error %d", inode
->i_ino
, err
);
259 static struct inode
*ubifs_alloc_inode(struct super_block
*sb
)
261 struct ubifs_inode
*ui
;
263 ui
= kmem_cache_alloc(ubifs_inode_slab
, GFP_NOFS
);
267 memset((void *)ui
+ sizeof(struct inode
), 0,
268 sizeof(struct ubifs_inode
) - sizeof(struct inode
));
269 mutex_init(&ui
->ui_mutex
);
270 spin_lock_init(&ui
->ui_lock
);
271 return &ui
->vfs_inode
;
274 static void ubifs_destroy_inode(struct inode
*inode
)
276 struct ubifs_inode
*ui
= ubifs_inode(inode
);
279 kmem_cache_free(ubifs_inode_slab
, inode
);
283 * Note, Linux write-back code calls this without 'i_mutex'.
285 static int ubifs_write_inode(struct inode
*inode
, int wait
)
288 struct ubifs_info
*c
= inode
->i_sb
->s_fs_info
;
289 struct ubifs_inode
*ui
= ubifs_inode(inode
);
291 ubifs_assert(!ui
->xattr
);
292 if (is_bad_inode(inode
))
295 mutex_lock(&ui
->ui_mutex
);
297 * Due to races between write-back forced by budgeting
298 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
299 * have already been synchronized, do not do this again. This might
300 * also happen if it was synchronized in an VFS operation, e.g.
304 mutex_unlock(&ui
->ui_mutex
);
309 * As an optimization, do not write orphan inodes to the media just
310 * because this is not needed.
312 dbg_gen("inode %lu, mode %#x, nlink %u",
313 inode
->i_ino
, (int)inode
->i_mode
, inode
->i_nlink
);
314 if (inode
->i_nlink
) {
315 err
= ubifs_jnl_write_inode(c
, inode
);
317 ubifs_err("can't write inode %lu, error %d",
322 mutex_unlock(&ui
->ui_mutex
);
323 ubifs_release_dirty_inode_budget(c
, ui
);
327 static void ubifs_delete_inode(struct inode
*inode
)
330 struct ubifs_info
*c
= inode
->i_sb
->s_fs_info
;
331 struct ubifs_inode
*ui
= ubifs_inode(inode
);
335 * Extended attribute inode deletions are fully handled in
336 * 'ubifs_removexattr()'. These inodes are special and have
337 * limited usage, so there is nothing to do here.
341 dbg_gen("inode %lu, mode %#x", inode
->i_ino
, (int)inode
->i_mode
);
342 ubifs_assert(!atomic_read(&inode
->i_count
));
343 ubifs_assert(inode
->i_nlink
== 0);
345 truncate_inode_pages(&inode
->i_data
, 0);
346 if (is_bad_inode(inode
))
349 ui
->ui_size
= inode
->i_size
= 0;
350 err
= ubifs_jnl_delete_inode(c
, inode
);
353 * Worst case we have a lost orphan inode wasting space, so a
354 * simple error message is OK here.
356 ubifs_err("can't delete inode %lu, error %d",
361 ubifs_release_dirty_inode_budget(c
, ui
);
365 static void ubifs_dirty_inode(struct inode
*inode
)
367 struct ubifs_inode
*ui
= ubifs_inode(inode
);
369 ubifs_assert(mutex_is_locked(&ui
->ui_mutex
));
372 dbg_gen("inode %lu", inode
->i_ino
);
376 static int ubifs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
378 struct ubifs_info
*c
= dentry
->d_sb
->s_fs_info
;
379 unsigned long long free
;
380 __le32
*uuid
= (__le32
*)c
->uuid
;
382 free
= ubifs_get_free_space(c
);
383 dbg_gen("free space %lld bytes (%lld blocks)",
384 free
, free
>> UBIFS_BLOCK_SHIFT
);
386 buf
->f_type
= UBIFS_SUPER_MAGIC
;
387 buf
->f_bsize
= UBIFS_BLOCK_SIZE
;
388 buf
->f_blocks
= c
->block_cnt
;
389 buf
->f_bfree
= free
>> UBIFS_BLOCK_SHIFT
;
390 if (free
> c
->report_rp_size
)
391 buf
->f_bavail
= (free
- c
->report_rp_size
) >> UBIFS_BLOCK_SHIFT
;
396 buf
->f_namelen
= UBIFS_MAX_NLEN
;
397 buf
->f_fsid
.val
[0] = le32_to_cpu(uuid
[0]) ^ le32_to_cpu(uuid
[2]);
398 buf
->f_fsid
.val
[1] = le32_to_cpu(uuid
[1]) ^ le32_to_cpu(uuid
[3]);
402 static int ubifs_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
404 struct ubifs_info
*c
= mnt
->mnt_sb
->s_fs_info
;
406 if (c
->mount_opts
.unmount_mode
== 2)
407 seq_printf(s
, ",fast_unmount");
408 else if (c
->mount_opts
.unmount_mode
== 1)
409 seq_printf(s
, ",norm_unmount");
411 if (c
->mount_opts
.bulk_read
== 2)
412 seq_printf(s
, ",bulk_read");
413 else if (c
->mount_opts
.bulk_read
== 1)
414 seq_printf(s
, ",no_bulk_read");
416 if (c
->mount_opts
.chk_data_crc
== 2)
417 seq_printf(s
, ",chk_data_crc");
418 else if (c
->mount_opts
.chk_data_crc
== 1)
419 seq_printf(s
, ",no_chk_data_crc");
421 if (c
->mount_opts
.override_compr
) {
422 seq_printf(s
, ",compr=");
423 seq_printf(s
, ubifs_compr_name(c
->mount_opts
.compr_type
));
429 static int ubifs_sync_fs(struct super_block
*sb
, int wait
)
431 struct ubifs_info
*c
= sb
->s_fs_info
;
436 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
437 err
= ubifs_wbuf_sync(&c
->jheads
[i
].wbuf
);
442 /* Commit the journal unless it has too little data */
443 spin_lock(&c
->buds_lock
);
444 bud_bytes
= c
->bud_bytes
;
445 spin_unlock(&c
->buds_lock
);
446 if (bud_bytes
> c
->leb_size
) {
447 err
= ubifs_run_commit(c
);
454 * We ought to call sync for c->ubi but it does not have one. If it had
455 * it would in turn call mtd->sync, however mtd operations are
456 * synchronous anyway, so we don't lose any sleep here.
462 * init_constants_early - initialize UBIFS constants.
463 * @c: UBIFS file-system description object
465 * This function initialize UBIFS constants which do not need the superblock to
466 * be read. It also checks that the UBI volume satisfies basic UBIFS
467 * requirements. Returns zero in case of success and a negative error code in
470 static int init_constants_early(struct ubifs_info
*c
)
472 if (c
->vi
.corrupted
) {
473 ubifs_warn("UBI volume is corrupted - read-only mode");
478 ubifs_msg("read-only UBI device");
482 if (c
->vi
.vol_type
== UBI_STATIC_VOLUME
) {
483 ubifs_msg("static UBI volume - read-only mode");
487 c
->leb_cnt
= c
->vi
.size
;
488 c
->leb_size
= c
->vi
.usable_leb_size
;
489 c
->half_leb_size
= c
->leb_size
/ 2;
490 c
->min_io_size
= c
->di
.min_io_size
;
491 c
->min_io_shift
= fls(c
->min_io_size
) - 1;
493 if (c
->leb_size
< UBIFS_MIN_LEB_SZ
) {
494 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
495 c
->leb_size
, UBIFS_MIN_LEB_SZ
);
499 if (c
->leb_cnt
< UBIFS_MIN_LEB_CNT
) {
500 ubifs_err("too few LEBs (%d), min. is %d",
501 c
->leb_cnt
, UBIFS_MIN_LEB_CNT
);
505 if (!is_power_of_2(c
->min_io_size
)) {
506 ubifs_err("bad min. I/O size %d", c
->min_io_size
);
511 * UBIFS aligns all node to 8-byte boundary, so to make function in
512 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
515 if (c
->min_io_size
< 8) {
520 c
->ref_node_alsz
= ALIGN(UBIFS_REF_NODE_SZ
, c
->min_io_size
);
521 c
->mst_node_alsz
= ALIGN(UBIFS_MST_NODE_SZ
, c
->min_io_size
);
524 * Initialize node length ranges which are mostly needed for node
527 c
->ranges
[UBIFS_PAD_NODE
].len
= UBIFS_PAD_NODE_SZ
;
528 c
->ranges
[UBIFS_SB_NODE
].len
= UBIFS_SB_NODE_SZ
;
529 c
->ranges
[UBIFS_MST_NODE
].len
= UBIFS_MST_NODE_SZ
;
530 c
->ranges
[UBIFS_REF_NODE
].len
= UBIFS_REF_NODE_SZ
;
531 c
->ranges
[UBIFS_TRUN_NODE
].len
= UBIFS_TRUN_NODE_SZ
;
532 c
->ranges
[UBIFS_CS_NODE
].len
= UBIFS_CS_NODE_SZ
;
534 c
->ranges
[UBIFS_INO_NODE
].min_len
= UBIFS_INO_NODE_SZ
;
535 c
->ranges
[UBIFS_INO_NODE
].max_len
= UBIFS_MAX_INO_NODE_SZ
;
536 c
->ranges
[UBIFS_ORPH_NODE
].min_len
=
537 UBIFS_ORPH_NODE_SZ
+ sizeof(__le64
);
538 c
->ranges
[UBIFS_ORPH_NODE
].max_len
= c
->leb_size
;
539 c
->ranges
[UBIFS_DENT_NODE
].min_len
= UBIFS_DENT_NODE_SZ
;
540 c
->ranges
[UBIFS_DENT_NODE
].max_len
= UBIFS_MAX_DENT_NODE_SZ
;
541 c
->ranges
[UBIFS_XENT_NODE
].min_len
= UBIFS_XENT_NODE_SZ
;
542 c
->ranges
[UBIFS_XENT_NODE
].max_len
= UBIFS_MAX_XENT_NODE_SZ
;
543 c
->ranges
[UBIFS_DATA_NODE
].min_len
= UBIFS_DATA_NODE_SZ
;
544 c
->ranges
[UBIFS_DATA_NODE
].max_len
= UBIFS_MAX_DATA_NODE_SZ
;
546 * Minimum indexing node size is amended later when superblock is
547 * read and the key length is known.
549 c
->ranges
[UBIFS_IDX_NODE
].min_len
= UBIFS_IDX_NODE_SZ
+ UBIFS_BRANCH_SZ
;
551 * Maximum indexing node size is amended later when superblock is
552 * read and the fanout is known.
554 c
->ranges
[UBIFS_IDX_NODE
].max_len
= INT_MAX
;
557 * Initialize dead and dark LEB space watermarks.
559 * Dead space is the space which cannot be used. Its watermark is
560 * equivalent to min. I/O unit or minimum node size if it is greater
561 * then min. I/O unit.
563 * Dark space is the space which might be used, or might not, depending
564 * on which node should be written to the LEB. Its watermark is
565 * equivalent to maximum UBIFS node size.
567 c
->dead_wm
= ALIGN(MIN_WRITE_SZ
, c
->min_io_size
);
568 c
->dark_wm
= ALIGN(UBIFS_MAX_NODE_SZ
, c
->min_io_size
);
571 * Calculate how many bytes would be wasted at the end of LEB if it was
572 * fully filled with data nodes of maximum size. This is used in
573 * calculations when reporting free space.
575 c
->leb_overhead
= c
->leb_size
% UBIFS_MAX_DATA_NODE_SZ
;
577 /* Buffer size for bulk-reads */
578 c
->max_bu_buf_len
= UBIFS_MAX_BULK_READ
* UBIFS_MAX_DATA_NODE_SZ
;
579 if (c
->max_bu_buf_len
> c
->leb_size
)
580 c
->max_bu_buf_len
= c
->leb_size
;
585 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
586 * @c: UBIFS file-system description object
587 * @lnum: LEB the write-buffer was synchronized to
588 * @free: how many free bytes left in this LEB
589 * @pad: how many bytes were padded
591 * This is a callback function which is called by the I/O unit when the
592 * write-buffer is synchronized. We need this to correctly maintain space
593 * accounting in bud logical eraseblocks. This function returns zero in case of
594 * success and a negative error code in case of failure.
596 * This function actually belongs to the journal, but we keep it here because
597 * we want to keep it static.
599 static int bud_wbuf_callback(struct ubifs_info
*c
, int lnum
, int free
, int pad
)
601 return ubifs_update_one_lp(c
, lnum
, free
, pad
, 0, 0);
605 * init_constants_sb - initialize UBIFS constants.
606 * @c: UBIFS file-system description object
608 * This is a helper function which initializes various UBIFS constants after
609 * the superblock has been read. It also checks various UBIFS parameters and
610 * makes sure they are all right. Returns zero in case of success and a
611 * negative error code in case of failure.
613 static int init_constants_sb(struct ubifs_info
*c
)
618 c
->main_bytes
= (long long)c
->main_lebs
* c
->leb_size
;
619 c
->max_znode_sz
= sizeof(struct ubifs_znode
) +
620 c
->fanout
* sizeof(struct ubifs_zbranch
);
622 tmp
= ubifs_idx_node_sz(c
, 1);
623 c
->ranges
[UBIFS_IDX_NODE
].min_len
= tmp
;
624 c
->min_idx_node_sz
= ALIGN(tmp
, 8);
626 tmp
= ubifs_idx_node_sz(c
, c
->fanout
);
627 c
->ranges
[UBIFS_IDX_NODE
].max_len
= tmp
;
628 c
->max_idx_node_sz
= ALIGN(tmp
, 8);
630 /* Make sure LEB size is large enough to fit full commit */
631 tmp
= UBIFS_CS_NODE_SZ
+ UBIFS_REF_NODE_SZ
* c
->jhead_cnt
;
632 tmp
= ALIGN(tmp
, c
->min_io_size
);
633 if (tmp
> c
->leb_size
) {
634 dbg_err("too small LEB size %d, at least %d needed",
640 * Make sure that the log is large enough to fit reference nodes for
641 * all buds plus one reserved LEB.
643 tmp64
= c
->max_bud_bytes
+ c
->leb_size
- 1;
644 c
->max_bud_cnt
= div_u64(tmp64
, c
->leb_size
);
645 tmp
= (c
->ref_node_alsz
* c
->max_bud_cnt
+ c
->leb_size
- 1);
648 if (c
->log_lebs
< tmp
) {
649 dbg_err("too small log %d LEBs, required min. %d LEBs",
655 * When budgeting we assume worst-case scenarios when the pages are not
656 * be compressed and direntries are of the maximum size.
658 * Note, data, which may be stored in inodes is budgeted separately, so
659 * it is not included into 'c->inode_budget'.
661 c
->page_budget
= UBIFS_MAX_DATA_NODE_SZ
* UBIFS_BLOCKS_PER_PAGE
;
662 c
->inode_budget
= UBIFS_INO_NODE_SZ
;
663 c
->dent_budget
= UBIFS_MAX_DENT_NODE_SZ
;
666 * When the amount of flash space used by buds becomes
667 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
668 * The writers are unblocked when the commit is finished. To avoid
669 * writers to be blocked UBIFS initiates background commit in advance,
670 * when number of bud bytes becomes above the limit defined below.
672 c
->bg_bud_bytes
= (c
->max_bud_bytes
* 13) >> 4;
675 * Ensure minimum journal size. All the bytes in the journal heads are
676 * considered to be used, when calculating the current journal usage.
677 * Consequently, if the journal is too small, UBIFS will treat it as
680 tmp64
= (long long)(c
->jhead_cnt
+ 1) * c
->leb_size
+ 1;
681 if (c
->bg_bud_bytes
< tmp64
)
682 c
->bg_bud_bytes
= tmp64
;
683 if (c
->max_bud_bytes
< tmp64
+ c
->leb_size
)
684 c
->max_bud_bytes
= tmp64
+ c
->leb_size
;
686 err
= ubifs_calc_lpt_geom(c
);
694 * init_constants_master - initialize UBIFS constants.
695 * @c: UBIFS file-system description object
697 * This is a helper function which initializes various UBIFS constants after
698 * the master node has been read. It also checks various UBIFS parameters and
699 * makes sure they are all right.
701 static void init_constants_master(struct ubifs_info
*c
)
705 c
->min_idx_lebs
= ubifs_calc_min_idx_lebs(c
);
708 * Calculate total amount of FS blocks. This number is not used
709 * internally because it does not make much sense for UBIFS, but it is
710 * necessary to report something for the 'statfs()' call.
712 * Subtract the LEB reserved for GC, the LEB which is reserved for
713 * deletions, minimum LEBs for the index, and assume only one journal
716 tmp64
= c
->main_lebs
- 1 - 1 - MIN_INDEX_LEBS
- c
->jhead_cnt
+ 1;
717 tmp64
*= (long long)c
->leb_size
- c
->leb_overhead
;
718 tmp64
= ubifs_reported_space(c
, tmp64
);
719 c
->block_cnt
= tmp64
>> UBIFS_BLOCK_SHIFT
;
723 * take_gc_lnum - reserve GC LEB.
724 * @c: UBIFS file-system description object
726 * This function ensures that the LEB reserved for garbage collection is
727 * unmapped and is marked as "taken" in lprops. We also have to set free space
728 * to LEB size and dirty space to zero, because lprops may contain out-of-date
729 * information if the file-system was un-mounted before it has been committed.
730 * This function returns zero in case of success and a negative error code in
733 static int take_gc_lnum(struct ubifs_info
*c
)
737 if (c
->gc_lnum
== -1) {
738 ubifs_err("no LEB for GC");
742 err
= ubifs_leb_unmap(c
, c
->gc_lnum
);
746 /* And we have to tell lprops that this LEB is taken */
747 err
= ubifs_change_one_lp(c
, c
->gc_lnum
, c
->leb_size
, 0,
753 * alloc_wbufs - allocate write-buffers.
754 * @c: UBIFS file-system description object
756 * This helper function allocates and initializes UBIFS write-buffers. Returns
757 * zero in case of success and %-ENOMEM in case of failure.
759 static int alloc_wbufs(struct ubifs_info
*c
)
763 c
->jheads
= kzalloc(c
->jhead_cnt
* sizeof(struct ubifs_jhead
),
768 /* Initialize journal heads */
769 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
770 INIT_LIST_HEAD(&c
->jheads
[i
].buds_list
);
771 err
= ubifs_wbuf_init(c
, &c
->jheads
[i
].wbuf
);
775 c
->jheads
[i
].wbuf
.sync_callback
= &bud_wbuf_callback
;
776 c
->jheads
[i
].wbuf
.jhead
= i
;
779 c
->jheads
[BASEHD
].wbuf
.dtype
= UBI_SHORTTERM
;
781 * Garbage Collector head likely contains long-term data and
782 * does not need to be synchronized by timer.
784 c
->jheads
[GCHD
].wbuf
.dtype
= UBI_LONGTERM
;
785 c
->jheads
[GCHD
].wbuf
.timeout
= 0;
791 * free_wbufs - free write-buffers.
792 * @c: UBIFS file-system description object
794 static void free_wbufs(struct ubifs_info
*c
)
799 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
800 kfree(c
->jheads
[i
].wbuf
.buf
);
801 kfree(c
->jheads
[i
].wbuf
.inodes
);
809 * free_orphans - free orphans.
810 * @c: UBIFS file-system description object
812 static void free_orphans(struct ubifs_info
*c
)
814 struct ubifs_orphan
*orph
;
816 while (c
->orph_dnext
) {
817 orph
= c
->orph_dnext
;
818 c
->orph_dnext
= orph
->dnext
;
819 list_del(&orph
->list
);
823 while (!list_empty(&c
->orph_list
)) {
824 orph
= list_entry(c
->orph_list
.next
, struct ubifs_orphan
, list
);
825 list_del(&orph
->list
);
827 dbg_err("orphan list not empty at unmount");
835 * free_buds - free per-bud objects.
836 * @c: UBIFS file-system description object
838 static void free_buds(struct ubifs_info
*c
)
840 struct rb_node
*this = c
->buds
.rb_node
;
841 struct ubifs_bud
*bud
;
845 this = this->rb_left
;
846 else if (this->rb_right
)
847 this = this->rb_right
;
849 bud
= rb_entry(this, struct ubifs_bud
, rb
);
850 this = rb_parent(this);
852 if (this->rb_left
== &bud
->rb
)
853 this->rb_left
= NULL
;
855 this->rb_right
= NULL
;
863 * check_volume_empty - check if the UBI volume is empty.
864 * @c: UBIFS file-system description object
866 * This function checks if the UBIFS volume is empty by looking if its LEBs are
867 * mapped or not. The result of checking is stored in the @c->empty variable.
868 * Returns zero in case of success and a negative error code in case of
871 static int check_volume_empty(struct ubifs_info
*c
)
876 for (lnum
= 0; lnum
< c
->leb_cnt
; lnum
++) {
877 err
= ubi_is_mapped(c
->ubi
, lnum
);
878 if (unlikely(err
< 0))
892 * UBIFS mount options.
894 * Opt_fast_unmount: do not run a journal commit before un-mounting
895 * Opt_norm_unmount: run a journal commit before un-mounting
896 * Opt_bulk_read: enable bulk-reads
897 * Opt_no_bulk_read: disable bulk-reads
898 * Opt_chk_data_crc: check CRCs when reading data nodes
899 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
900 * Opt_override_compr: override default compressor
901 * Opt_err: just end of array marker
914 static const match_table_t tokens
= {
915 {Opt_fast_unmount
, "fast_unmount"},
916 {Opt_norm_unmount
, "norm_unmount"},
917 {Opt_bulk_read
, "bulk_read"},
918 {Opt_no_bulk_read
, "no_bulk_read"},
919 {Opt_chk_data_crc
, "chk_data_crc"},
920 {Opt_no_chk_data_crc
, "no_chk_data_crc"},
921 {Opt_override_compr
, "compr=%s"},
926 * ubifs_parse_options - parse mount parameters.
927 * @c: UBIFS file-system description object
928 * @options: parameters to parse
929 * @is_remount: non-zero if this is FS re-mount
931 * This function parses UBIFS mount options and returns zero in case success
932 * and a negative error code in case of failure.
934 static int ubifs_parse_options(struct ubifs_info
*c
, char *options
,
938 substring_t args
[MAX_OPT_ARGS
];
943 while ((p
= strsep(&options
, ","))) {
949 token
= match_token(p
, tokens
, args
);
951 case Opt_fast_unmount
:
952 c
->mount_opts
.unmount_mode
= 2;
955 case Opt_norm_unmount
:
956 c
->mount_opts
.unmount_mode
= 1;
960 c
->mount_opts
.bulk_read
= 2;
963 case Opt_no_bulk_read
:
964 c
->mount_opts
.bulk_read
= 1;
967 case Opt_chk_data_crc
:
968 c
->mount_opts
.chk_data_crc
= 2;
969 c
->no_chk_data_crc
= 0;
971 case Opt_no_chk_data_crc
:
972 c
->mount_opts
.chk_data_crc
= 1;
973 c
->no_chk_data_crc
= 1;
975 case Opt_override_compr
:
977 char *name
= match_strdup(&args
[0]);
981 if (!strcmp(name
, "none"))
982 c
->mount_opts
.compr_type
= UBIFS_COMPR_NONE
;
983 else if (!strcmp(name
, "lzo"))
984 c
->mount_opts
.compr_type
= UBIFS_COMPR_LZO
;
985 else if (!strcmp(name
, "zlib"))
986 c
->mount_opts
.compr_type
= UBIFS_COMPR_ZLIB
;
988 ubifs_err("unknown compressor \"%s\"", name
);
993 c
->mount_opts
.override_compr
= 1;
994 c
->default_compr
= c
->mount_opts
.compr_type
;
998 ubifs_err("unrecognized mount option \"%s\" "
999 "or missing value", p
);
1008 * destroy_journal - destroy journal data structures.
1009 * @c: UBIFS file-system description object
1011 * This function destroys journal data structures including those that may have
1012 * been created by recovery functions.
1014 static void destroy_journal(struct ubifs_info
*c
)
1016 while (!list_empty(&c
->unclean_leb_list
)) {
1017 struct ubifs_unclean_leb
*ucleb
;
1019 ucleb
= list_entry(c
->unclean_leb_list
.next
,
1020 struct ubifs_unclean_leb
, list
);
1021 list_del(&ucleb
->list
);
1024 while (!list_empty(&c
->old_buds
)) {
1025 struct ubifs_bud
*bud
;
1027 bud
= list_entry(c
->old_buds
.next
, struct ubifs_bud
, list
);
1028 list_del(&bud
->list
);
1031 ubifs_destroy_idx_gc(c
);
1032 ubifs_destroy_size_tree(c
);
1038 * bu_init - initialize bulk-read information.
1039 * @c: UBIFS file-system description object
1041 static void bu_init(struct ubifs_info
*c
)
1043 ubifs_assert(c
->bulk_read
== 1);
1046 return; /* Already initialized */
1049 c
->bu
.buf
= kmalloc(c
->max_bu_buf_len
, GFP_KERNEL
| __GFP_NOWARN
);
1051 if (c
->max_bu_buf_len
> UBIFS_KMALLOC_OK
) {
1052 c
->max_bu_buf_len
= UBIFS_KMALLOC_OK
;
1056 /* Just disable bulk-read */
1057 ubifs_warn("Cannot allocate %d bytes of memory for bulk-read, "
1058 "disabling it", c
->max_bu_buf_len
);
1059 c
->mount_opts
.bulk_read
= 1;
1066 * mount_ubifs - mount UBIFS file-system.
1067 * @c: UBIFS file-system description object
1069 * This function mounts UBIFS file system. Returns zero in case of success and
1070 * a negative error code in case of failure.
1072 * Note, the function does not de-allocate resources it it fails half way
1073 * through, and the caller has to do this instead.
1075 static int mount_ubifs(struct ubifs_info
*c
)
1077 struct super_block
*sb
= c
->vfs_sb
;
1078 int err
, mounted_read_only
= (sb
->s_flags
& MS_RDONLY
);
1082 err
= init_constants_early(c
);
1086 err
= ubifs_debugging_init(c
);
1090 err
= check_volume_empty(c
);
1094 if (c
->empty
&& (mounted_read_only
|| c
->ro_media
)) {
1096 * This UBI volume is empty, and read-only, or the file system
1097 * is mounted read-only - we cannot format it.
1099 ubifs_err("can't format empty UBI volume: read-only %s",
1100 c
->ro_media
? "UBI volume" : "mount");
1105 if (c
->ro_media
&& !mounted_read_only
) {
1106 ubifs_err("cannot mount read-write - read-only media");
1112 * The requirement for the buffer is that it should fit indexing B-tree
1113 * height amount of integers. We assume the height if the TNC tree will
1117 c
->bottom_up_buf
= kmalloc(BOTTOM_UP_HEIGHT
* sizeof(int), GFP_KERNEL
);
1118 if (!c
->bottom_up_buf
)
1121 c
->sbuf
= vmalloc(c
->leb_size
);
1125 if (!mounted_read_only
) {
1126 c
->ileb_buf
= vmalloc(c
->leb_size
);
1131 if (c
->bulk_read
== 1)
1135 * We have to check all CRCs, even for data nodes, when we mount the FS
1136 * (specifically, when we are replaying).
1138 c
->always_chk_crc
= 1;
1140 err
= ubifs_read_superblock(c
);
1145 * Make sure the compressor which is set as default in the superblock
1146 * or overriden by mount options is actually compiled in.
1148 if (!ubifs_compr_present(c
->default_compr
)) {
1149 ubifs_err("'compressor \"%s\" is not compiled in",
1150 ubifs_compr_name(c
->default_compr
));
1154 err
= init_constants_sb(c
);
1158 sz
= ALIGN(c
->max_idx_node_sz
, c
->min_io_size
);
1159 sz
= ALIGN(sz
+ c
->max_idx_node_sz
, c
->min_io_size
);
1160 c
->cbuf
= kmalloc(sz
, GFP_NOFS
);
1166 sprintf(c
->bgt_name
, BGT_NAME_PATTERN
, c
->vi
.ubi_num
, c
->vi
.vol_id
);
1167 if (!mounted_read_only
) {
1168 err
= alloc_wbufs(c
);
1172 /* Create background thread */
1173 c
->bgt
= kthread_create(ubifs_bg_thread
, c
, c
->bgt_name
);
1174 if (IS_ERR(c
->bgt
)) {
1175 err
= PTR_ERR(c
->bgt
);
1177 ubifs_err("cannot spawn \"%s\", error %d",
1181 wake_up_process(c
->bgt
);
1184 err
= ubifs_read_master(c
);
1188 init_constants_master(c
);
1190 if ((c
->mst_node
->flags
& cpu_to_le32(UBIFS_MST_DIRTY
)) != 0) {
1191 ubifs_msg("recovery needed");
1192 c
->need_recovery
= 1;
1193 if (!mounted_read_only
) {
1194 err
= ubifs_recover_inl_heads(c
, c
->sbuf
);
1198 } else if (!mounted_read_only
) {
1200 * Set the "dirty" flag so that if we reboot uncleanly we
1201 * will notice this immediately on the next mount.
1203 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_DIRTY
);
1204 err
= ubifs_write_master(c
);
1209 err
= ubifs_lpt_init(c
, 1, !mounted_read_only
);
1213 err
= dbg_check_idx_size(c
, c
->old_idx_sz
);
1217 err
= ubifs_replay_journal(c
);
1221 err
= ubifs_mount_orphans(c
, c
->need_recovery
, mounted_read_only
);
1225 if (!mounted_read_only
) {
1228 /* Check for enough free space */
1229 if (ubifs_calc_available(c
, c
->min_idx_lebs
) <= 0) {
1230 ubifs_err("insufficient available space");
1235 /* Check for enough log space */
1236 lnum
= c
->lhead_lnum
+ 1;
1237 if (lnum
>= UBIFS_LOG_LNUM
+ c
->log_lebs
)
1238 lnum
= UBIFS_LOG_LNUM
;
1239 if (lnum
== c
->ltail_lnum
) {
1240 err
= ubifs_consolidate_log(c
);
1245 if (c
->need_recovery
) {
1246 err
= ubifs_recover_size(c
);
1249 err
= ubifs_rcvry_gc_commit(c
);
1251 err
= take_gc_lnum(c
);
1255 err
= dbg_check_lprops(c
);
1258 } else if (c
->need_recovery
) {
1259 err
= ubifs_recover_size(c
);
1264 spin_lock(&ubifs_infos_lock
);
1265 list_add_tail(&c
->infos_list
, &ubifs_infos
);
1266 spin_unlock(&ubifs_infos_lock
);
1268 if (c
->need_recovery
) {
1269 if (mounted_read_only
)
1270 ubifs_msg("recovery deferred");
1272 c
->need_recovery
= 0;
1273 ubifs_msg("recovery completed");
1277 err
= dbg_debugfs_init_fs(c
);
1281 err
= dbg_check_filesystem(c
);
1285 c
->always_chk_crc
= 0;
1287 ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
1288 c
->vi
.ubi_num
, c
->vi
.vol_id
, c
->vi
.name
);
1289 if (mounted_read_only
)
1290 ubifs_msg("mounted read-only");
1291 x
= (long long)c
->main_lebs
* c
->leb_size
;
1292 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d "
1293 "LEBs)", x
, x
>> 10, x
>> 20, c
->main_lebs
);
1294 x
= (long long)c
->log_lebs
* c
->leb_size
+ c
->max_bud_bytes
;
1295 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d "
1296 "LEBs)", x
, x
>> 10, x
>> 20, c
->log_lebs
+ c
->max_bud_cnt
);
1297 ubifs_msg("media format: %d (latest is %d)",
1298 c
->fmt_version
, UBIFS_FORMAT_VERSION
);
1299 ubifs_msg("default compressor: %s", ubifs_compr_name(c
->default_compr
));
1300 ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
1301 c
->report_rp_size
, c
->report_rp_size
>> 10);
1303 dbg_msg("compiled on: " __DATE__
" at " __TIME__
);
1304 dbg_msg("min. I/O unit size: %d bytes", c
->min_io_size
);
1305 dbg_msg("LEB size: %d bytes (%d KiB)",
1306 c
->leb_size
, c
->leb_size
>> 10);
1307 dbg_msg("data journal heads: %d",
1308 c
->jhead_cnt
- NONDATA_JHEADS_CNT
);
1309 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1310 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1311 c
->uuid
[0], c
->uuid
[1], c
->uuid
[2], c
->uuid
[3],
1312 c
->uuid
[4], c
->uuid
[5], c
->uuid
[6], c
->uuid
[7],
1313 c
->uuid
[8], c
->uuid
[9], c
->uuid
[10], c
->uuid
[11],
1314 c
->uuid
[12], c
->uuid
[13], c
->uuid
[14], c
->uuid
[15]);
1315 dbg_msg("fast unmount: %d", c
->fast_unmount
);
1316 dbg_msg("big_lpt %d", c
->big_lpt
);
1317 dbg_msg("log LEBs: %d (%d - %d)",
1318 c
->log_lebs
, UBIFS_LOG_LNUM
, c
->log_last
);
1319 dbg_msg("LPT area LEBs: %d (%d - %d)",
1320 c
->lpt_lebs
, c
->lpt_first
, c
->lpt_last
);
1321 dbg_msg("orphan area LEBs: %d (%d - %d)",
1322 c
->orph_lebs
, c
->orph_first
, c
->orph_last
);
1323 dbg_msg("main area LEBs: %d (%d - %d)",
1324 c
->main_lebs
, c
->main_first
, c
->leb_cnt
- 1);
1325 dbg_msg("index LEBs: %d", c
->lst
.idx_lebs
);
1326 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1327 c
->old_idx_sz
, c
->old_idx_sz
>> 10, c
->old_idx_sz
>> 20);
1328 dbg_msg("key hash type: %d", c
->key_hash_type
);
1329 dbg_msg("tree fanout: %d", c
->fanout
);
1330 dbg_msg("reserved GC LEB: %d", c
->gc_lnum
);
1331 dbg_msg("first main LEB: %d", c
->main_first
);
1332 dbg_msg("dead watermark: %d", c
->dead_wm
);
1333 dbg_msg("dark watermark: %d", c
->dark_wm
);
1334 x
= (long long)c
->main_lebs
* c
->dark_wm
;
1335 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1336 x
, x
>> 10, x
>> 20);
1337 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1338 c
->max_bud_bytes
, c
->max_bud_bytes
>> 10,
1339 c
->max_bud_bytes
>> 20);
1340 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1341 c
->bg_bud_bytes
, c
->bg_bud_bytes
>> 10,
1342 c
->bg_bud_bytes
>> 20);
1343 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1344 c
->bud_bytes
, c
->bud_bytes
>> 10, c
->bud_bytes
>> 20);
1345 dbg_msg("max. seq. number: %llu", c
->max_sqnum
);
1346 dbg_msg("commit number: %llu", c
->cmt_no
);
1351 spin_lock(&ubifs_infos_lock
);
1352 list_del(&c
->infos_list
);
1353 spin_unlock(&ubifs_infos_lock
);
1359 ubifs_lpt_free(c
, 0);
1362 kfree(c
->rcvrd_mst_node
);
1364 kthread_stop(c
->bgt
);
1373 kfree(c
->bottom_up_buf
);
1374 ubifs_debugging_exit(c
);
1379 * ubifs_umount - un-mount UBIFS file-system.
1380 * @c: UBIFS file-system description object
1382 * Note, this function is called to free allocated resourced when un-mounting,
1383 * as well as free resources when an error occurred while we were half way
1384 * through mounting (error path cleanup function). So it has to make sure the
1385 * resource was actually allocated before freeing it.
1387 static void ubifs_umount(struct ubifs_info
*c
)
1389 dbg_gen("un-mounting UBI device %d, volume %d", c
->vi
.ubi_num
,
1392 dbg_debugfs_exit_fs(c
);
1393 spin_lock(&ubifs_infos_lock
);
1394 list_del(&c
->infos_list
);
1395 spin_unlock(&ubifs_infos_lock
);
1398 kthread_stop(c
->bgt
);
1403 ubifs_lpt_free(c
, 0);
1406 kfree(c
->rcvrd_mst_node
);
1411 kfree(c
->bottom_up_buf
);
1412 ubifs_debugging_exit(c
);
1416 * ubifs_remount_rw - re-mount in read-write mode.
1417 * @c: UBIFS file-system description object
1419 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1420 * mode. This function allocates the needed resources and re-mounts UBIFS in
1423 static int ubifs_remount_rw(struct ubifs_info
*c
)
1430 mutex_lock(&c
->umount_mutex
);
1431 c
->remounting_rw
= 1;
1432 c
->always_chk_crc
= 1;
1434 /* Check for enough free space */
1435 if (ubifs_calc_available(c
, c
->min_idx_lebs
) <= 0) {
1436 ubifs_err("insufficient available space");
1441 if (c
->old_leb_cnt
!= c
->leb_cnt
) {
1442 struct ubifs_sb_node
*sup
;
1444 sup
= ubifs_read_sb_node(c
);
1449 sup
->leb_cnt
= cpu_to_le32(c
->leb_cnt
);
1450 err
= ubifs_write_sb_node(c
, sup
);
1455 if (c
->need_recovery
) {
1456 ubifs_msg("completing deferred recovery");
1457 err
= ubifs_write_rcvrd_mst_node(c
);
1460 err
= ubifs_recover_size(c
);
1463 err
= ubifs_clean_lebs(c
, c
->sbuf
);
1466 err
= ubifs_recover_inl_heads(c
, c
->sbuf
);
1471 if (!(c
->mst_node
->flags
& cpu_to_le32(UBIFS_MST_DIRTY
))) {
1472 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_DIRTY
);
1473 err
= ubifs_write_master(c
);
1478 c
->ileb_buf
= vmalloc(c
->leb_size
);
1484 err
= ubifs_lpt_init(c
, 0, 1);
1488 err
= alloc_wbufs(c
);
1492 ubifs_create_buds_lists(c
);
1494 /* Create background thread */
1495 c
->bgt
= kthread_create(ubifs_bg_thread
, c
, c
->bgt_name
);
1496 if (IS_ERR(c
->bgt
)) {
1497 err
= PTR_ERR(c
->bgt
);
1499 ubifs_err("cannot spawn \"%s\", error %d",
1503 wake_up_process(c
->bgt
);
1505 c
->orph_buf
= vmalloc(c
->leb_size
);
1511 /* Check for enough log space */
1512 lnum
= c
->lhead_lnum
+ 1;
1513 if (lnum
>= UBIFS_LOG_LNUM
+ c
->log_lebs
)
1514 lnum
= UBIFS_LOG_LNUM
;
1515 if (lnum
== c
->ltail_lnum
) {
1516 err
= ubifs_consolidate_log(c
);
1521 if (c
->need_recovery
)
1522 err
= ubifs_rcvry_gc_commit(c
);
1524 err
= take_gc_lnum(c
);
1528 if (c
->need_recovery
) {
1529 c
->need_recovery
= 0;
1530 ubifs_msg("deferred recovery completed");
1533 dbg_gen("re-mounted read-write");
1534 c
->vfs_sb
->s_flags
&= ~MS_RDONLY
;
1535 c
->remounting_rw
= 0;
1536 c
->always_chk_crc
= 0;
1537 mutex_unlock(&c
->umount_mutex
);
1544 kthread_stop(c
->bgt
);
1550 ubifs_lpt_free(c
, 1);
1551 c
->remounting_rw
= 0;
1552 c
->always_chk_crc
= 0;
1553 mutex_unlock(&c
->umount_mutex
);
1558 * commit_on_unmount - commit the journal when un-mounting.
1559 * @c: UBIFS file-system description object
1561 * This function is called during un-mounting and re-mounting, and it commits
1562 * the journal unless the "fast unmount" mode is enabled. It also avoids
1563 * committing the journal if it contains too few data.
1565 static void commit_on_unmount(struct ubifs_info
*c
)
1567 if (!c
->fast_unmount
) {
1568 long long bud_bytes
;
1570 spin_lock(&c
->buds_lock
);
1571 bud_bytes
= c
->bud_bytes
;
1572 spin_unlock(&c
->buds_lock
);
1573 if (bud_bytes
> c
->leb_size
)
1574 ubifs_run_commit(c
);
1579 * ubifs_remount_ro - re-mount in read-only mode.
1580 * @c: UBIFS file-system description object
1582 * We rely on VFS to have stopped writing. Possibly the background thread could
1583 * be running a commit, however kthread_stop will wait in that case.
1585 static void ubifs_remount_ro(struct ubifs_info
*c
)
1589 ubifs_assert(!c
->need_recovery
);
1590 commit_on_unmount(c
);
1592 mutex_lock(&c
->umount_mutex
);
1594 kthread_stop(c
->bgt
);
1598 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
1599 ubifs_wbuf_sync(&c
->jheads
[i
].wbuf
);
1600 del_timer_sync(&c
->jheads
[i
].wbuf
.timer
);
1604 c
->mst_node
->flags
&= ~cpu_to_le32(UBIFS_MST_DIRTY
);
1605 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_NO_ORPHS
);
1606 c
->mst_node
->gc_lnum
= cpu_to_le32(c
->gc_lnum
);
1607 err
= ubifs_write_master(c
);
1609 ubifs_ro_mode(c
, err
);
1612 ubifs_destroy_idx_gc(c
);
1618 ubifs_lpt_free(c
, 1);
1619 mutex_unlock(&c
->umount_mutex
);
1622 static void ubifs_put_super(struct super_block
*sb
)
1625 struct ubifs_info
*c
= sb
->s_fs_info
;
1627 ubifs_msg("un-mount UBI device %d, volume %d", c
->vi
.ubi_num
,
1630 * The following asserts are only valid if there has not been a failure
1631 * of the media. For example, there will be dirty inodes if we failed
1632 * to write them back because of I/O errors.
1634 ubifs_assert(atomic_long_read(&c
->dirty_pg_cnt
) == 0);
1635 ubifs_assert(c
->budg_idx_growth
== 0);
1636 ubifs_assert(c
->budg_dd_growth
== 0);
1637 ubifs_assert(c
->budg_data_growth
== 0);
1640 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1641 * and file system un-mount. Namely, it prevents the shrinker from
1642 * picking this superblock for shrinking - it will be just skipped if
1643 * the mutex is locked.
1645 mutex_lock(&c
->umount_mutex
);
1646 if (!(c
->vfs_sb
->s_flags
& MS_RDONLY
)) {
1648 * First of all kill the background thread to make sure it does
1649 * not interfere with un-mounting and freeing resources.
1652 kthread_stop(c
->bgt
);
1656 /* Synchronize write-buffers */
1658 for (i
= 0; i
< c
->jhead_cnt
; i
++) {
1659 ubifs_wbuf_sync(&c
->jheads
[i
].wbuf
);
1660 del_timer_sync(&c
->jheads
[i
].wbuf
.timer
);
1664 * On fatal errors c->ro_media is set to 1, in which case we do
1665 * not write the master node.
1669 * We are being cleanly unmounted which means the
1670 * orphans were killed - indicate this in the master
1671 * node. Also save the reserved GC LEB number.
1675 c
->mst_node
->flags
&= ~cpu_to_le32(UBIFS_MST_DIRTY
);
1676 c
->mst_node
->flags
|= cpu_to_le32(UBIFS_MST_NO_ORPHS
);
1677 c
->mst_node
->gc_lnum
= cpu_to_le32(c
->gc_lnum
);
1678 err
= ubifs_write_master(c
);
1681 * Recovery will attempt to fix the master area
1682 * next mount, so we just print a message and
1683 * continue to unmount normally.
1685 ubifs_err("failed to write master node, "
1691 bdi_destroy(&c
->bdi
);
1692 ubi_close_volume(c
->ubi
);
1693 mutex_unlock(&c
->umount_mutex
);
1697 static int ubifs_remount_fs(struct super_block
*sb
, int *flags
, char *data
)
1700 struct ubifs_info
*c
= sb
->s_fs_info
;
1702 dbg_gen("old flags %#lx, new flags %#x", sb
->s_flags
, *flags
);
1704 err
= ubifs_parse_options(c
, data
, 1);
1706 ubifs_err("invalid or unknown remount parameter");
1710 if ((sb
->s_flags
& MS_RDONLY
) && !(*flags
& MS_RDONLY
)) {
1711 err
= ubifs_remount_rw(c
);
1714 } else if (!(sb
->s_flags
& MS_RDONLY
) && (*flags
& MS_RDONLY
))
1715 ubifs_remount_ro(c
);
1717 if (c
->bulk_read
== 1)
1720 dbg_gen("disable bulk-read");
1728 struct super_operations ubifs_super_operations
= {
1729 .alloc_inode
= ubifs_alloc_inode
,
1730 .destroy_inode
= ubifs_destroy_inode
,
1731 .put_super
= ubifs_put_super
,
1732 .write_inode
= ubifs_write_inode
,
1733 .delete_inode
= ubifs_delete_inode
,
1734 .statfs
= ubifs_statfs
,
1735 .dirty_inode
= ubifs_dirty_inode
,
1736 .remount_fs
= ubifs_remount_fs
,
1737 .show_options
= ubifs_show_options
,
1738 .sync_fs
= ubifs_sync_fs
,
1742 * open_ubi - parse UBI device name string and open the UBI device.
1743 * @name: UBI volume name
1744 * @mode: UBI volume open mode
1746 * There are several ways to specify UBI volumes when mounting UBIFS:
1747 * o ubiX_Y - UBI device number X, volume Y;
1748 * o ubiY - UBI device number 0, volume Y;
1749 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1750 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1752 * Alternative '!' separator may be used instead of ':' (because some shells
1753 * like busybox may interpret ':' as an NFS host name separator). This function
1754 * returns ubi volume object in case of success and a negative error code in
1757 static struct ubi_volume_desc
*open_ubi(const char *name
, int mode
)
1762 if (name
[0] != 'u' || name
[1] != 'b' || name
[2] != 'i')
1763 return ERR_PTR(-EINVAL
);
1765 /* ubi:NAME method */
1766 if ((name
[3] == ':' || name
[3] == '!') && name
[4] != '\0')
1767 return ubi_open_volume_nm(0, name
+ 4, mode
);
1769 if (!isdigit(name
[3]))
1770 return ERR_PTR(-EINVAL
);
1772 dev
= simple_strtoul(name
+ 3, &endptr
, 0);
1775 if (*endptr
== '\0')
1776 return ubi_open_volume(0, dev
, mode
);
1779 if (*endptr
== '_' && isdigit(endptr
[1])) {
1780 vol
= simple_strtoul(endptr
+ 1, &endptr
, 0);
1781 if (*endptr
!= '\0')
1782 return ERR_PTR(-EINVAL
);
1783 return ubi_open_volume(dev
, vol
, mode
);
1786 /* ubiX:NAME method */
1787 if ((*endptr
== ':' || *endptr
== '!') && endptr
[1] != '\0')
1788 return ubi_open_volume_nm(dev
, ++endptr
, mode
);
1790 return ERR_PTR(-EINVAL
);
1793 static int ubifs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1795 struct ubi_volume_desc
*ubi
= sb
->s_fs_info
;
1796 struct ubifs_info
*c
;
1800 c
= kzalloc(sizeof(struct ubifs_info
), GFP_KERNEL
);
1804 spin_lock_init(&c
->cnt_lock
);
1805 spin_lock_init(&c
->cs_lock
);
1806 spin_lock_init(&c
->buds_lock
);
1807 spin_lock_init(&c
->space_lock
);
1808 spin_lock_init(&c
->orphan_lock
);
1809 init_rwsem(&c
->commit_sem
);
1810 mutex_init(&c
->lp_mutex
);
1811 mutex_init(&c
->tnc_mutex
);
1812 mutex_init(&c
->log_mutex
);
1813 mutex_init(&c
->mst_mutex
);
1814 mutex_init(&c
->umount_mutex
);
1815 mutex_init(&c
->bu_mutex
);
1816 init_waitqueue_head(&c
->cmt_wq
);
1818 c
->old_idx
= RB_ROOT
;
1819 c
->size_tree
= RB_ROOT
;
1820 c
->orph_tree
= RB_ROOT
;
1821 INIT_LIST_HEAD(&c
->infos_list
);
1822 INIT_LIST_HEAD(&c
->idx_gc
);
1823 INIT_LIST_HEAD(&c
->replay_list
);
1824 INIT_LIST_HEAD(&c
->replay_buds
);
1825 INIT_LIST_HEAD(&c
->uncat_list
);
1826 INIT_LIST_HEAD(&c
->empty_list
);
1827 INIT_LIST_HEAD(&c
->freeable_list
);
1828 INIT_LIST_HEAD(&c
->frdi_idx_list
);
1829 INIT_LIST_HEAD(&c
->unclean_leb_list
);
1830 INIT_LIST_HEAD(&c
->old_buds
);
1831 INIT_LIST_HEAD(&c
->orph_list
);
1832 INIT_LIST_HEAD(&c
->orph_new
);
1834 c
->highest_inum
= UBIFS_FIRST_INO
;
1835 c
->lhead_lnum
= c
->ltail_lnum
= UBIFS_LOG_LNUM
;
1837 ubi_get_volume_info(ubi
, &c
->vi
);
1838 ubi_get_device_info(c
->vi
.ubi_num
, &c
->di
);
1840 /* Re-open the UBI device in read-write mode */
1841 c
->ubi
= ubi_open_volume(c
->vi
.ubi_num
, c
->vi
.vol_id
, UBI_READWRITE
);
1842 if (IS_ERR(c
->ubi
)) {
1843 err
= PTR_ERR(c
->ubi
);
1848 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
1849 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1850 * which means the user would have to wait not just for their own I/O
1851 * but the read-ahead I/O as well i.e. completely pointless.
1853 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1855 c
->bdi
.capabilities
= BDI_CAP_MAP_COPY
;
1856 c
->bdi
.unplug_io_fn
= default_unplug_io_fn
;
1857 err
= bdi_init(&c
->bdi
);
1861 err
= ubifs_parse_options(c
, data
, 0);
1868 sb
->s_magic
= UBIFS_SUPER_MAGIC
;
1869 sb
->s_blocksize
= UBIFS_BLOCK_SIZE
;
1870 sb
->s_blocksize_bits
= UBIFS_BLOCK_SHIFT
;
1871 sb
->s_dev
= c
->vi
.cdev
;
1872 sb
->s_maxbytes
= c
->max_inode_sz
= key_max_inode_size(c
);
1873 if (c
->max_inode_sz
> MAX_LFS_FILESIZE
)
1874 sb
->s_maxbytes
= c
->max_inode_sz
= MAX_LFS_FILESIZE
;
1875 sb
->s_op
= &ubifs_super_operations
;
1877 mutex_lock(&c
->umount_mutex
);
1878 err
= mount_ubifs(c
);
1880 ubifs_assert(err
< 0);
1884 /* Read the root inode */
1885 root
= ubifs_iget(sb
, UBIFS_ROOT_INO
);
1887 err
= PTR_ERR(root
);
1891 sb
->s_root
= d_alloc_root(root
);
1895 mutex_unlock(&c
->umount_mutex
);
1903 mutex_unlock(&c
->umount_mutex
);
1905 bdi_destroy(&c
->bdi
);
1907 ubi_close_volume(c
->ubi
);
1913 static int sb_test(struct super_block
*sb
, void *data
)
1917 return sb
->s_dev
== *dev
;
1920 static int sb_set(struct super_block
*sb
, void *data
)
1928 static int ubifs_get_sb(struct file_system_type
*fs_type
, int flags
,
1929 const char *name
, void *data
, struct vfsmount
*mnt
)
1931 struct ubi_volume_desc
*ubi
;
1932 struct ubi_volume_info vi
;
1933 struct super_block
*sb
;
1936 dbg_gen("name %s, flags %#x", name
, flags
);
1939 * Get UBI device number and volume ID. Mount it read-only so far
1940 * because this might be a new mount point, and UBI allows only one
1941 * read-write user at a time.
1943 ubi
= open_ubi(name
, UBI_READONLY
);
1945 ubifs_err("cannot open \"%s\", error %d",
1946 name
, (int)PTR_ERR(ubi
));
1947 return PTR_ERR(ubi
);
1949 ubi_get_volume_info(ubi
, &vi
);
1951 dbg_gen("opened ubi%d_%d", vi
.ubi_num
, vi
.vol_id
);
1953 sb
= sget(fs_type
, &sb_test
, &sb_set
, &vi
.cdev
);
1960 /* A new mount point for already mounted UBIFS */
1961 dbg_gen("this ubi volume is already mounted");
1962 if ((flags
^ sb
->s_flags
) & MS_RDONLY
) {
1967 sb
->s_flags
= flags
;
1969 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
1972 sb
->s_fs_info
= ubi
;
1973 err
= ubifs_fill_super(sb
, data
, flags
& MS_SILENT
? 1 : 0);
1976 /* We do not support atime */
1977 sb
->s_flags
|= MS_ACTIVE
| MS_NOATIME
;
1980 /* 'fill_super()' opens ubi again so we must close it here */
1981 ubi_close_volume(ubi
);
1983 return simple_set_mnt(mnt
, sb
);
1986 up_write(&sb
->s_umount
);
1987 deactivate_super(sb
);
1989 ubi_close_volume(ubi
);
1993 static void ubifs_kill_sb(struct super_block
*sb
)
1995 struct ubifs_info
*c
= sb
->s_fs_info
;
1998 * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
1999 * in order to be outside BKL.
2001 if (sb
->s_root
&& !(sb
->s_flags
& MS_RDONLY
))
2002 commit_on_unmount(c
);
2003 /* The un-mount routine is actually done in put_super() */
2004 generic_shutdown_super(sb
);
2007 static struct file_system_type ubifs_fs_type
= {
2009 .owner
= THIS_MODULE
,
2010 .get_sb
= ubifs_get_sb
,
2011 .kill_sb
= ubifs_kill_sb
2015 * Inode slab cache constructor.
2017 static void inode_slab_ctor(void *obj
)
2019 struct ubifs_inode
*ui
= obj
;
2020 inode_init_once(&ui
->vfs_inode
);
2023 static int __init
ubifs_init(void)
2027 BUILD_BUG_ON(sizeof(struct ubifs_ch
) != 24);
2029 /* Make sure node sizes are 8-byte aligned */
2030 BUILD_BUG_ON(UBIFS_CH_SZ
& 7);
2031 BUILD_BUG_ON(UBIFS_INO_NODE_SZ
& 7);
2032 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ
& 7);
2033 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ
& 7);
2034 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ
& 7);
2035 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ
& 7);
2036 BUILD_BUG_ON(UBIFS_SB_NODE_SZ
& 7);
2037 BUILD_BUG_ON(UBIFS_MST_NODE_SZ
& 7);
2038 BUILD_BUG_ON(UBIFS_REF_NODE_SZ
& 7);
2039 BUILD_BUG_ON(UBIFS_CS_NODE_SZ
& 7);
2040 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ
& 7);
2042 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ
& 7);
2043 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ
& 7);
2044 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ
& 7);
2045 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ
& 7);
2046 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ
& 7);
2047 BUILD_BUG_ON(MIN_WRITE_SZ
& 7);
2049 /* Check min. node size */
2050 BUILD_BUG_ON(UBIFS_INO_NODE_SZ
< MIN_WRITE_SZ
);
2051 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ
< MIN_WRITE_SZ
);
2052 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ
< MIN_WRITE_SZ
);
2053 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ
< MIN_WRITE_SZ
);
2055 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2056 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2057 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2058 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ
> UBIFS_MAX_NODE_SZ
);
2060 /* Defined node sizes */
2061 BUILD_BUG_ON(UBIFS_SB_NODE_SZ
!= 4096);
2062 BUILD_BUG_ON(UBIFS_MST_NODE_SZ
!= 512);
2063 BUILD_BUG_ON(UBIFS_INO_NODE_SZ
!= 160);
2064 BUILD_BUG_ON(UBIFS_REF_NODE_SZ
!= 64);
2067 * We use 2 bit wide bit-fields to store compression type, which should
2068 * be amended if more compressors are added. The bit-fields are:
2069 * @compr_type in 'struct ubifs_inode', @default_compr in
2070 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'.
2072 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT
> 4);
2075 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
2076 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
2078 if (PAGE_CACHE_SIZE
< UBIFS_BLOCK_SIZE
) {
2079 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
2080 " at least 4096 bytes",
2081 (unsigned int)PAGE_CACHE_SIZE
);
2085 err
= register_filesystem(&ubifs_fs_type
);
2087 ubifs_err("cannot register file system, error %d", err
);
2092 ubifs_inode_slab
= kmem_cache_create("ubifs_inode_slab",
2093 sizeof(struct ubifs_inode
), 0,
2094 SLAB_MEM_SPREAD
| SLAB_RECLAIM_ACCOUNT
,
2096 if (!ubifs_inode_slab
)
2099 register_shrinker(&ubifs_shrinker_info
);
2101 err
= ubifs_compressors_init();
2105 err
= dbg_debugfs_init();
2112 ubifs_compressors_exit();
2114 unregister_shrinker(&ubifs_shrinker_info
);
2115 kmem_cache_destroy(ubifs_inode_slab
);
2117 unregister_filesystem(&ubifs_fs_type
);
2120 /* late_initcall to let compressors initialize first */
2121 late_initcall(ubifs_init
);
2123 static void __exit
ubifs_exit(void)
2125 ubifs_assert(list_empty(&ubifs_infos
));
2126 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt
) == 0);
2129 ubifs_compressors_exit();
2130 unregister_shrinker(&ubifs_shrinker_info
);
2131 kmem_cache_destroy(ubifs_inode_slab
);
2132 unregister_filesystem(&ubifs_fs_type
);
2134 module_exit(ubifs_exit
);
2136 MODULE_LICENSE("GPL");
2137 MODULE_VERSION(__stringify(UBIFS_VERSION
));
2138 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
2139 MODULE_DESCRIPTION("UBIFS - UBI File System");