Merge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus
[linux-2.6/mini2440.git] / fs / ubifs / super.c
blob00eb9c68ad03512ef40ebf98acf77e83bbc13345
1 /*
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
13 * more details.
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)
20 * Adrian Hunter
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/random.h>
34 #include <linux/kthread.h>
35 #include <linux/parser.h>
36 #include <linux/seq_file.h>
37 #include <linux/mount.h>
38 #include "ubifs.h"
40 /* Slab cache for UBIFS inodes */
41 struct kmem_cache *ubifs_inode_slab;
43 /* UBIFS TNC shrinker description */
44 static struct shrinker ubifs_shrinker_info = {
45 .shrink = ubifs_shrinker,
46 .seeks = DEFAULT_SEEKS,
49 /**
50 * validate_inode - validate inode.
51 * @c: UBIFS file-system description object
52 * @inode: the inode to validate
54 * This is a helper function for 'ubifs_iget()' which validates various fields
55 * of a newly built inode to make sure they contain sane values and prevent
56 * possible vulnerabilities. Returns zero if the inode is all right and
57 * a non-zero error code if not.
59 static int validate_inode(struct ubifs_info *c, const struct inode *inode)
61 int err;
62 const struct ubifs_inode *ui = ubifs_inode(inode);
64 if (inode->i_size > c->max_inode_sz) {
65 ubifs_err("inode is too large (%lld)",
66 (long long)inode->i_size);
67 return 1;
70 if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
71 ubifs_err("unknown compression type %d", ui->compr_type);
72 return 2;
75 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
76 return 3;
78 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
79 return 4;
81 if (ui->xattr && (inode->i_mode & S_IFMT) != S_IFREG)
82 return 5;
84 if (!ubifs_compr_present(ui->compr_type)) {
85 ubifs_warn("inode %lu uses '%s' compression, but it was not "
86 "compiled in", inode->i_ino,
87 ubifs_compr_name(ui->compr_type));
90 err = dbg_check_dir_size(c, inode);
91 return err;
94 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
96 int err;
97 union ubifs_key key;
98 struct ubifs_ino_node *ino;
99 struct ubifs_info *c = sb->s_fs_info;
100 struct inode *inode;
101 struct ubifs_inode *ui;
103 dbg_gen("inode %lu", inum);
105 inode = iget_locked(sb, inum);
106 if (!inode)
107 return ERR_PTR(-ENOMEM);
108 if (!(inode->i_state & I_NEW))
109 return inode;
110 ui = ubifs_inode(inode);
112 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
113 if (!ino) {
114 err = -ENOMEM;
115 goto out;
118 ino_key_init(c, &key, inode->i_ino);
120 err = ubifs_tnc_lookup(c, &key, ino);
121 if (err)
122 goto out_ino;
124 inode->i_flags |= (S_NOCMTIME | S_NOATIME);
125 inode->i_nlink = le32_to_cpu(ino->nlink);
126 inode->i_uid = le32_to_cpu(ino->uid);
127 inode->i_gid = le32_to_cpu(ino->gid);
128 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec);
129 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
130 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec);
131 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec);
132 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec);
133 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec);
134 inode->i_mode = le32_to_cpu(ino->mode);
135 inode->i_size = le64_to_cpu(ino->size);
137 ui->data_len = le32_to_cpu(ino->data_len);
138 ui->flags = le32_to_cpu(ino->flags);
139 ui->compr_type = le16_to_cpu(ino->compr_type);
140 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
141 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
142 ui->xattr_size = le32_to_cpu(ino->xattr_size);
143 ui->xattr_names = le32_to_cpu(ino->xattr_names);
144 ui->synced_i_size = ui->ui_size = inode->i_size;
146 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
148 err = validate_inode(c, inode);
149 if (err)
150 goto out_invalid;
152 /* Disable readahead */
153 inode->i_mapping->backing_dev_info = &c->bdi;
155 switch (inode->i_mode & S_IFMT) {
156 case S_IFREG:
157 inode->i_mapping->a_ops = &ubifs_file_address_operations;
158 inode->i_op = &ubifs_file_inode_operations;
159 inode->i_fop = &ubifs_file_operations;
160 if (ui->xattr) {
161 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
162 if (!ui->data) {
163 err = -ENOMEM;
164 goto out_ino;
166 memcpy(ui->data, ino->data, ui->data_len);
167 ((char *)ui->data)[ui->data_len] = '\0';
168 } else if (ui->data_len != 0) {
169 err = 10;
170 goto out_invalid;
172 break;
173 case S_IFDIR:
174 inode->i_op = &ubifs_dir_inode_operations;
175 inode->i_fop = &ubifs_dir_operations;
176 if (ui->data_len != 0) {
177 err = 11;
178 goto out_invalid;
180 break;
181 case S_IFLNK:
182 inode->i_op = &ubifs_symlink_inode_operations;
183 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
184 err = 12;
185 goto out_invalid;
187 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
188 if (!ui->data) {
189 err = -ENOMEM;
190 goto out_ino;
192 memcpy(ui->data, ino->data, ui->data_len);
193 ((char *)ui->data)[ui->data_len] = '\0';
194 break;
195 case S_IFBLK:
196 case S_IFCHR:
198 dev_t rdev;
199 union ubifs_dev_desc *dev;
201 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
202 if (!ui->data) {
203 err = -ENOMEM;
204 goto out_ino;
207 dev = (union ubifs_dev_desc *)ino->data;
208 if (ui->data_len == sizeof(dev->new))
209 rdev = new_decode_dev(le32_to_cpu(dev->new));
210 else if (ui->data_len == sizeof(dev->huge))
211 rdev = huge_decode_dev(le64_to_cpu(dev->huge));
212 else {
213 err = 13;
214 goto out_invalid;
216 memcpy(ui->data, ino->data, ui->data_len);
217 inode->i_op = &ubifs_file_inode_operations;
218 init_special_inode(inode, inode->i_mode, rdev);
219 break;
221 case S_IFSOCK:
222 case S_IFIFO:
223 inode->i_op = &ubifs_file_inode_operations;
224 init_special_inode(inode, inode->i_mode, 0);
225 if (ui->data_len != 0) {
226 err = 14;
227 goto out_invalid;
229 break;
230 default:
231 err = 15;
232 goto out_invalid;
235 kfree(ino);
236 ubifs_set_inode_flags(inode);
237 unlock_new_inode(inode);
238 return inode;
240 out_invalid:
241 ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
242 dbg_dump_node(c, ino);
243 dbg_dump_inode(c, inode);
244 err = -EINVAL;
245 out_ino:
246 kfree(ino);
247 out:
248 ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
249 iget_failed(inode);
250 return ERR_PTR(err);
253 static struct inode *ubifs_alloc_inode(struct super_block *sb)
255 struct ubifs_inode *ui;
257 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
258 if (!ui)
259 return NULL;
261 memset((void *)ui + sizeof(struct inode), 0,
262 sizeof(struct ubifs_inode) - sizeof(struct inode));
263 mutex_init(&ui->ui_mutex);
264 spin_lock_init(&ui->ui_lock);
265 return &ui->vfs_inode;
268 static void ubifs_destroy_inode(struct inode *inode)
270 struct ubifs_inode *ui = ubifs_inode(inode);
272 kfree(ui->data);
273 kmem_cache_free(ubifs_inode_slab, inode);
277 * Note, Linux write-back code calls this without 'i_mutex'.
279 static int ubifs_write_inode(struct inode *inode, int wait)
281 int err;
282 struct ubifs_info *c = inode->i_sb->s_fs_info;
283 struct ubifs_inode *ui = ubifs_inode(inode);
285 ubifs_assert(!ui->xattr);
286 if (is_bad_inode(inode))
287 return 0;
289 mutex_lock(&ui->ui_mutex);
291 * Due to races between write-back forced by budgeting
292 * (see 'sync_some_inodes()') and pdflush write-back, the inode may
293 * have already been synchronized, do not do this again. This might
294 * also happen if it was synchronized in an VFS operation, e.g.
295 * 'ubifs_link()'.
297 if (!ui->dirty) {
298 mutex_unlock(&ui->ui_mutex);
299 return 0;
302 dbg_gen("inode %lu", inode->i_ino);
303 err = ubifs_jnl_write_inode(c, inode, 0);
304 if (err)
305 ubifs_err("can't write inode %lu, error %d", inode->i_ino, err);
307 ui->dirty = 0;
308 mutex_unlock(&ui->ui_mutex);
309 ubifs_release_dirty_inode_budget(c, ui);
310 return err;
313 static void ubifs_delete_inode(struct inode *inode)
315 int err;
316 struct ubifs_info *c = inode->i_sb->s_fs_info;
318 if (ubifs_inode(inode)->xattr)
320 * Extended attribute inode deletions are fully handled in
321 * 'ubifs_removexattr()'. These inodes are special and have
322 * limited usage, so there is nothing to do here.
324 goto out;
326 dbg_gen("inode %lu", inode->i_ino);
327 ubifs_assert(!atomic_read(&inode->i_count));
328 ubifs_assert(inode->i_nlink == 0);
330 truncate_inode_pages(&inode->i_data, 0);
331 if (is_bad_inode(inode))
332 goto out;
334 ubifs_inode(inode)->ui_size = inode->i_size = 0;
335 err = ubifs_jnl_write_inode(c, inode, 1);
336 if (err)
338 * Worst case we have a lost orphan inode wasting space, so a
339 * simple error message is ok here.
341 ubifs_err("can't write inode %lu, error %d", inode->i_ino, err);
342 out:
343 clear_inode(inode);
346 static void ubifs_dirty_inode(struct inode *inode)
348 struct ubifs_inode *ui = ubifs_inode(inode);
350 ubifs_assert(mutex_is_locked(&ui->ui_mutex));
351 if (!ui->dirty) {
352 ui->dirty = 1;
353 dbg_gen("inode %lu", inode->i_ino);
357 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
359 struct ubifs_info *c = dentry->d_sb->s_fs_info;
360 unsigned long long free;
362 free = ubifs_budg_get_free_space(c);
363 dbg_gen("free space %lld bytes (%lld blocks)",
364 free, free >> UBIFS_BLOCK_SHIFT);
366 buf->f_type = UBIFS_SUPER_MAGIC;
367 buf->f_bsize = UBIFS_BLOCK_SIZE;
368 buf->f_blocks = c->block_cnt;
369 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
370 if (free > c->report_rp_size)
371 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
372 else
373 buf->f_bavail = 0;
374 buf->f_files = 0;
375 buf->f_ffree = 0;
376 buf->f_namelen = UBIFS_MAX_NLEN;
378 return 0;
381 static int ubifs_show_options(struct seq_file *s, struct vfsmount *mnt)
383 struct ubifs_info *c = mnt->mnt_sb->s_fs_info;
385 if (c->mount_opts.unmount_mode == 2)
386 seq_printf(s, ",fast_unmount");
387 else if (c->mount_opts.unmount_mode == 1)
388 seq_printf(s, ",norm_unmount");
390 return 0;
393 static int ubifs_sync_fs(struct super_block *sb, int wait)
395 struct ubifs_info *c = sb->s_fs_info;
396 int i, ret = 0, err;
398 if (c->jheads)
399 for (i = 0; i < c->jhead_cnt; i++) {
400 err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
401 if (err && !ret)
402 ret = err;
405 * We ought to call sync for c->ubi but it does not have one. If it had
406 * it would in turn call mtd->sync, however mtd operations are
407 * synchronous anyway, so we don't lose any sleep here.
409 return ret;
413 * init_constants_early - initialize UBIFS constants.
414 * @c: UBIFS file-system description object
416 * This function initialize UBIFS constants which do not need the superblock to
417 * be read. It also checks that the UBI volume satisfies basic UBIFS
418 * requirements. Returns zero in case of success and a negative error code in
419 * case of failure.
421 static int init_constants_early(struct ubifs_info *c)
423 if (c->vi.corrupted) {
424 ubifs_warn("UBI volume is corrupted - read-only mode");
425 c->ro_media = 1;
428 if (c->di.ro_mode) {
429 ubifs_msg("read-only UBI device");
430 c->ro_media = 1;
433 if (c->vi.vol_type == UBI_STATIC_VOLUME) {
434 ubifs_msg("static UBI volume - read-only mode");
435 c->ro_media = 1;
438 c->leb_cnt = c->vi.size;
439 c->leb_size = c->vi.usable_leb_size;
440 c->half_leb_size = c->leb_size / 2;
441 c->min_io_size = c->di.min_io_size;
442 c->min_io_shift = fls(c->min_io_size) - 1;
444 if (c->leb_size < UBIFS_MIN_LEB_SZ) {
445 ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
446 c->leb_size, UBIFS_MIN_LEB_SZ);
447 return -EINVAL;
450 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) {
451 ubifs_err("too few LEBs (%d), min. is %d",
452 c->leb_cnt, UBIFS_MIN_LEB_CNT);
453 return -EINVAL;
456 if (!is_power_of_2(c->min_io_size)) {
457 ubifs_err("bad min. I/O size %d", c->min_io_size);
458 return -EINVAL;
462 * UBIFS aligns all node to 8-byte boundary, so to make function in
463 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
464 * less than 8.
466 if (c->min_io_size < 8) {
467 c->min_io_size = 8;
468 c->min_io_shift = 3;
471 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
472 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size);
475 * Initialize node length ranges which are mostly needed for node
476 * length validation.
478 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ;
479 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ;
480 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ;
481 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ;
482 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ;
483 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ;
485 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ;
486 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ;
487 c->ranges[UBIFS_ORPH_NODE].min_len =
488 UBIFS_ORPH_NODE_SZ + sizeof(__le64);
489 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size;
490 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ;
491 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ;
492 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ;
493 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ;
494 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ;
495 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ;
497 * Minimum indexing node size is amended later when superblock is
498 * read and the key length is known.
500 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ;
502 * Maximum indexing node size is amended later when superblock is
503 * read and the fanout is known.
505 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX;
508 * Initialize dead and dark LEB space watermarks.
510 * Dead space is the space which cannot be used. Its watermark is
511 * equivalent to min. I/O unit or minimum node size if it is greater
512 * then min. I/O unit.
514 * Dark space is the space which might be used, or might not, depending
515 * on which node should be written to the LEB. Its watermark is
516 * equivalent to maximum UBIFS node size.
518 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size);
519 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size);
521 return 0;
525 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
526 * @c: UBIFS file-system description object
527 * @lnum: LEB the write-buffer was synchronized to
528 * @free: how many free bytes left in this LEB
529 * @pad: how many bytes were padded
531 * This is a callback function which is called by the I/O unit when the
532 * write-buffer is synchronized. We need this to correctly maintain space
533 * accounting in bud logical eraseblocks. This function returns zero in case of
534 * success and a negative error code in case of failure.
536 * This function actually belongs to the journal, but we keep it here because
537 * we want to keep it static.
539 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
541 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
545 * init_constants_late - initialize UBIFS constants.
546 * @c: UBIFS file-system description object
548 * This is a helper function which initializes various UBIFS constants after
549 * the superblock has been read. It also checks various UBIFS parameters and
550 * makes sure they are all right. Returns zero in case of success and a
551 * negative error code in case of failure.
553 static int init_constants_late(struct ubifs_info *c)
555 int tmp, err;
556 uint64_t tmp64;
558 c->main_bytes = (long long)c->main_lebs * c->leb_size;
559 c->max_znode_sz = sizeof(struct ubifs_znode) +
560 c->fanout * sizeof(struct ubifs_zbranch);
562 tmp = ubifs_idx_node_sz(c, 1);
563 c->ranges[UBIFS_IDX_NODE].min_len = tmp;
564 c->min_idx_node_sz = ALIGN(tmp, 8);
566 tmp = ubifs_idx_node_sz(c, c->fanout);
567 c->ranges[UBIFS_IDX_NODE].max_len = tmp;
568 c->max_idx_node_sz = ALIGN(tmp, 8);
570 /* Make sure LEB size is large enough to fit full commit */
571 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
572 tmp = ALIGN(tmp, c->min_io_size);
573 if (tmp > c->leb_size) {
574 dbg_err("too small LEB size %d, at least %d needed",
575 c->leb_size, tmp);
576 return -EINVAL;
580 * Make sure that the log is large enough to fit reference nodes for
581 * all buds plus one reserved LEB.
583 tmp64 = c->max_bud_bytes;
584 tmp = do_div(tmp64, c->leb_size);
585 c->max_bud_cnt = tmp64 + !!tmp;
586 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1);
587 tmp /= c->leb_size;
588 tmp += 1;
589 if (c->log_lebs < tmp) {
590 dbg_err("too small log %d LEBs, required min. %d LEBs",
591 c->log_lebs, tmp);
592 return -EINVAL;
596 * When budgeting we assume worst-case scenarios when the pages are not
597 * be compressed and direntries are of the maximum size.
599 * Note, data, which may be stored in inodes is budgeted separately, so
600 * it is not included into 'c->inode_budget'.
602 c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
603 c->inode_budget = UBIFS_INO_NODE_SZ;
604 c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
607 * When the amount of flash space used by buds becomes
608 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit.
609 * The writers are unblocked when the commit is finished. To avoid
610 * writers to be blocked UBIFS initiates background commit in advance,
611 * when number of bud bytes becomes above the limit defined below.
613 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4;
616 * Ensure minimum journal size. All the bytes in the journal heads are
617 * considered to be used, when calculating the current journal usage.
618 * Consequently, if the journal is too small, UBIFS will treat it as
619 * always full.
621 tmp64 = (uint64_t)(c->jhead_cnt + 1) * c->leb_size + 1;
622 if (c->bg_bud_bytes < tmp64)
623 c->bg_bud_bytes = tmp64;
624 if (c->max_bud_bytes < tmp64 + c->leb_size)
625 c->max_bud_bytes = tmp64 + c->leb_size;
627 err = ubifs_calc_lpt_geom(c);
628 if (err)
629 return err;
631 c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
634 * Calculate total amount of FS blocks. This number is not used
635 * internally because it does not make much sense for UBIFS, but it is
636 * necessary to report something for the 'statfs()' call.
638 * Subtract the LEB reserved for GC and the LEB which is reserved for
639 * deletions.
641 * Review 'ubifs_calc_available()' if changing this calculation.
643 tmp64 = c->main_lebs - 2;
644 tmp64 *= (uint64_t)c->leb_size - c->dark_wm;
645 tmp64 = ubifs_reported_space(c, tmp64);
646 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
648 return 0;
652 * take_gc_lnum - reserve GC LEB.
653 * @c: UBIFS file-system description object
655 * This function ensures that the LEB reserved for garbage collection is
656 * unmapped and is marked as "taken" in lprops. We also have to set free space
657 * to LEB size and dirty space to zero, because lprops may contain out-of-date
658 * information if the file-system was un-mounted before it has been committed.
659 * This function returns zero in case of success and a negative error code in
660 * case of failure.
662 static int take_gc_lnum(struct ubifs_info *c)
664 int err;
666 if (c->gc_lnum == -1) {
667 ubifs_err("no LEB for GC");
668 return -EINVAL;
671 err = ubifs_leb_unmap(c, c->gc_lnum);
672 if (err)
673 return err;
675 /* And we have to tell lprops that this LEB is taken */
676 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
677 LPROPS_TAKEN, 0, 0);
678 return err;
682 * alloc_wbufs - allocate write-buffers.
683 * @c: UBIFS file-system description object
685 * This helper function allocates and initializes UBIFS write-buffers. Returns
686 * zero in case of success and %-ENOMEM in case of failure.
688 static int alloc_wbufs(struct ubifs_info *c)
690 int i, err;
692 c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
693 GFP_KERNEL);
694 if (!c->jheads)
695 return -ENOMEM;
697 /* Initialize journal heads */
698 for (i = 0; i < c->jhead_cnt; i++) {
699 INIT_LIST_HEAD(&c->jheads[i].buds_list);
700 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
701 if (err)
702 return err;
704 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
705 c->jheads[i].wbuf.jhead = i;
708 c->jheads[BASEHD].wbuf.dtype = UBI_SHORTTERM;
710 * Garbage Collector head likely contains long-term data and
711 * does not need to be synchronized by timer.
713 c->jheads[GCHD].wbuf.dtype = UBI_LONGTERM;
714 c->jheads[GCHD].wbuf.timeout = 0;
716 return 0;
720 * free_wbufs - free write-buffers.
721 * @c: UBIFS file-system description object
723 static void free_wbufs(struct ubifs_info *c)
725 int i;
727 if (c->jheads) {
728 for (i = 0; i < c->jhead_cnt; i++) {
729 kfree(c->jheads[i].wbuf.buf);
730 kfree(c->jheads[i].wbuf.inodes);
732 kfree(c->jheads);
733 c->jheads = NULL;
738 * free_orphans - free orphans.
739 * @c: UBIFS file-system description object
741 static void free_orphans(struct ubifs_info *c)
743 struct ubifs_orphan *orph;
745 while (c->orph_dnext) {
746 orph = c->orph_dnext;
747 c->orph_dnext = orph->dnext;
748 list_del(&orph->list);
749 kfree(orph);
752 while (!list_empty(&c->orph_list)) {
753 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
754 list_del(&orph->list);
755 kfree(orph);
756 dbg_err("orphan list not empty at unmount");
759 vfree(c->orph_buf);
760 c->orph_buf = NULL;
764 * free_buds - free per-bud objects.
765 * @c: UBIFS file-system description object
767 static void free_buds(struct ubifs_info *c)
769 struct rb_node *this = c->buds.rb_node;
770 struct ubifs_bud *bud;
772 while (this) {
773 if (this->rb_left)
774 this = this->rb_left;
775 else if (this->rb_right)
776 this = this->rb_right;
777 else {
778 bud = rb_entry(this, struct ubifs_bud, rb);
779 this = rb_parent(this);
780 if (this) {
781 if (this->rb_left == &bud->rb)
782 this->rb_left = NULL;
783 else
784 this->rb_right = NULL;
786 kfree(bud);
792 * check_volume_empty - check if the UBI volume is empty.
793 * @c: UBIFS file-system description object
795 * This function checks if the UBIFS volume is empty by looking if its LEBs are
796 * mapped or not. The result of checking is stored in the @c->empty variable.
797 * Returns zero in case of success and a negative error code in case of
798 * failure.
800 static int check_volume_empty(struct ubifs_info *c)
802 int lnum, err;
804 c->empty = 1;
805 for (lnum = 0; lnum < c->leb_cnt; lnum++) {
806 err = ubi_is_mapped(c->ubi, lnum);
807 if (unlikely(err < 0))
808 return err;
809 if (err == 1) {
810 c->empty = 0;
811 break;
814 cond_resched();
817 return 0;
821 * UBIFS mount options.
823 * Opt_fast_unmount: do not run a journal commit before un-mounting
824 * Opt_norm_unmount: run a journal commit before un-mounting
825 * Opt_err: just end of array marker
827 enum {
828 Opt_fast_unmount,
829 Opt_norm_unmount,
830 Opt_err,
833 static match_table_t tokens = {
834 {Opt_fast_unmount, "fast_unmount"},
835 {Opt_norm_unmount, "norm_unmount"},
836 {Opt_err, NULL},
840 * ubifs_parse_options - parse mount parameters.
841 * @c: UBIFS file-system description object
842 * @options: parameters to parse
843 * @is_remount: non-zero if this is FS re-mount
845 * This function parses UBIFS mount options and returns zero in case success
846 * and a negative error code in case of failure.
848 static int ubifs_parse_options(struct ubifs_info *c, char *options,
849 int is_remount)
851 char *p;
852 substring_t args[MAX_OPT_ARGS];
854 if (!options)
855 return 0;
857 while ((p = strsep(&options, ","))) {
858 int token;
860 if (!*p)
861 continue;
863 token = match_token(p, tokens, args);
864 switch (token) {
865 case Opt_fast_unmount:
866 c->mount_opts.unmount_mode = 2;
867 c->fast_unmount = 1;
868 break;
869 case Opt_norm_unmount:
870 c->mount_opts.unmount_mode = 1;
871 c->fast_unmount = 0;
872 break;
873 default:
874 ubifs_err("unrecognized mount option \"%s\" "
875 "or missing value", p);
876 return -EINVAL;
880 return 0;
884 * destroy_journal - destroy journal data structures.
885 * @c: UBIFS file-system description object
887 * This function destroys journal data structures including those that may have
888 * been created by recovery functions.
890 static void destroy_journal(struct ubifs_info *c)
892 while (!list_empty(&c->unclean_leb_list)) {
893 struct ubifs_unclean_leb *ucleb;
895 ucleb = list_entry(c->unclean_leb_list.next,
896 struct ubifs_unclean_leb, list);
897 list_del(&ucleb->list);
898 kfree(ucleb);
900 while (!list_empty(&c->old_buds)) {
901 struct ubifs_bud *bud;
903 bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
904 list_del(&bud->list);
905 kfree(bud);
907 ubifs_destroy_idx_gc(c);
908 ubifs_destroy_size_tree(c);
909 ubifs_tnc_close(c);
910 free_buds(c);
914 * mount_ubifs - mount UBIFS file-system.
915 * @c: UBIFS file-system description object
917 * This function mounts UBIFS file system. Returns zero in case of success and
918 * a negative error code in case of failure.
920 * Note, the function does not de-allocate resources it it fails half way
921 * through, and the caller has to do this instead.
923 static int mount_ubifs(struct ubifs_info *c)
925 struct super_block *sb = c->vfs_sb;
926 int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
927 long long x;
928 size_t sz;
930 err = init_constants_early(c);
931 if (err)
932 return err;
934 #ifdef CONFIG_UBIFS_FS_DEBUG
935 c->dbg_buf = vmalloc(c->leb_size);
936 if (!c->dbg_buf)
937 return -ENOMEM;
938 #endif
940 err = check_volume_empty(c);
941 if (err)
942 goto out_free;
944 if (c->empty && (mounted_read_only || c->ro_media)) {
946 * This UBI volume is empty, and read-only, or the file system
947 * is mounted read-only - we cannot format it.
949 ubifs_err("can't format empty UBI volume: read-only %s",
950 c->ro_media ? "UBI volume" : "mount");
951 err = -EROFS;
952 goto out_free;
955 if (c->ro_media && !mounted_read_only) {
956 ubifs_err("cannot mount read-write - read-only media");
957 err = -EROFS;
958 goto out_free;
962 * The requirement for the buffer is that it should fit indexing B-tree
963 * height amount of integers. We assume the height if the TNC tree will
964 * never exceed 64.
966 err = -ENOMEM;
967 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL);
968 if (!c->bottom_up_buf)
969 goto out_free;
971 c->sbuf = vmalloc(c->leb_size);
972 if (!c->sbuf)
973 goto out_free;
975 if (!mounted_read_only) {
976 c->ileb_buf = vmalloc(c->leb_size);
977 if (!c->ileb_buf)
978 goto out_free;
981 err = ubifs_read_superblock(c);
982 if (err)
983 goto out_free;
986 * Make sure the compressor which is set as the default on in the
987 * superblock was actually compiled in.
989 if (!ubifs_compr_present(c->default_compr)) {
990 ubifs_warn("'%s' compressor is set by superblock, but not "
991 "compiled in", ubifs_compr_name(c->default_compr));
992 c->default_compr = UBIFS_COMPR_NONE;
995 dbg_failure_mode_registration(c);
997 err = init_constants_late(c);
998 if (err)
999 goto out_dereg;
1001 sz = ALIGN(c->max_idx_node_sz, c->min_io_size);
1002 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size);
1003 c->cbuf = kmalloc(sz, GFP_NOFS);
1004 if (!c->cbuf) {
1005 err = -ENOMEM;
1006 goto out_dereg;
1009 if (!mounted_read_only) {
1010 err = alloc_wbufs(c);
1011 if (err)
1012 goto out_cbuf;
1014 /* Create background thread */
1015 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num,
1016 c->vi.vol_id);
1017 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1018 if (!c->bgt)
1019 c->bgt = ERR_PTR(-EINVAL);
1020 if (IS_ERR(c->bgt)) {
1021 err = PTR_ERR(c->bgt);
1022 c->bgt = NULL;
1023 ubifs_err("cannot spawn \"%s\", error %d",
1024 c->bgt_name, err);
1025 goto out_wbufs;
1027 wake_up_process(c->bgt);
1030 err = ubifs_read_master(c);
1031 if (err)
1032 goto out_master;
1034 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) {
1035 ubifs_msg("recovery needed");
1036 c->need_recovery = 1;
1037 if (!mounted_read_only) {
1038 err = ubifs_recover_inl_heads(c, c->sbuf);
1039 if (err)
1040 goto out_master;
1042 } else if (!mounted_read_only) {
1044 * Set the "dirty" flag so that if we reboot uncleanly we
1045 * will notice this immediately on the next mount.
1047 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1048 err = ubifs_write_master(c);
1049 if (err)
1050 goto out_master;
1053 err = ubifs_lpt_init(c, 1, !mounted_read_only);
1054 if (err)
1055 goto out_lpt;
1057 err = dbg_check_idx_size(c, c->old_idx_sz);
1058 if (err)
1059 goto out_lpt;
1061 err = ubifs_replay_journal(c);
1062 if (err)
1063 goto out_journal;
1065 err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
1066 if (err)
1067 goto out_orphans;
1069 if (!mounted_read_only) {
1070 int lnum;
1072 /* Check for enough free space */
1073 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1074 ubifs_err("insufficient available space");
1075 err = -EINVAL;
1076 goto out_orphans;
1079 /* Check for enough log space */
1080 lnum = c->lhead_lnum + 1;
1081 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1082 lnum = UBIFS_LOG_LNUM;
1083 if (lnum == c->ltail_lnum) {
1084 err = ubifs_consolidate_log(c);
1085 if (err)
1086 goto out_orphans;
1089 if (c->need_recovery) {
1090 err = ubifs_recover_size(c);
1091 if (err)
1092 goto out_orphans;
1093 err = ubifs_rcvry_gc_commit(c);
1094 } else
1095 err = take_gc_lnum(c);
1096 if (err)
1097 goto out_orphans;
1099 err = dbg_check_lprops(c);
1100 if (err)
1101 goto out_orphans;
1102 } else if (c->need_recovery) {
1103 err = ubifs_recover_size(c);
1104 if (err)
1105 goto out_orphans;
1108 spin_lock(&ubifs_infos_lock);
1109 list_add_tail(&c->infos_list, &ubifs_infos);
1110 spin_unlock(&ubifs_infos_lock);
1112 if (c->need_recovery) {
1113 if (mounted_read_only)
1114 ubifs_msg("recovery deferred");
1115 else {
1116 c->need_recovery = 0;
1117 ubifs_msg("recovery completed");
1121 err = dbg_check_filesystem(c);
1122 if (err)
1123 goto out_infos;
1125 ubifs_msg("mounted UBI device %d, volume %d", c->vi.ubi_num,
1126 c->vi.vol_id);
1127 if (mounted_read_only)
1128 ubifs_msg("mounted read-only");
1129 x = (long long)c->main_lebs * c->leb_size;
1130 ubifs_msg("file system size: %lld bytes (%lld KiB, %lld MiB, %d LEBs)",
1131 x, x >> 10, x >> 20, c->main_lebs);
1132 x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
1133 ubifs_msg("journal size: %lld bytes (%lld KiB, %lld MiB, %d LEBs)",
1134 x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
1135 ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
1136 ubifs_msg("media format %d, latest format %d",
1137 c->fmt_version, UBIFS_FORMAT_VERSION);
1139 dbg_msg("compiled on: " __DATE__ " at " __TIME__);
1140 dbg_msg("min. I/O unit size: %d bytes", c->min_io_size);
1141 dbg_msg("LEB size: %d bytes (%d KiB)",
1142 c->leb_size, c->leb_size / 1024);
1143 dbg_msg("data journal heads: %d",
1144 c->jhead_cnt - NONDATA_JHEADS_CNT);
1145 dbg_msg("UUID: %02X%02X%02X%02X-%02X%02X"
1146 "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
1147 c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
1148 c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
1149 c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
1150 c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
1151 dbg_msg("fast unmount: %d", c->fast_unmount);
1152 dbg_msg("big_lpt %d", c->big_lpt);
1153 dbg_msg("log LEBs: %d (%d - %d)",
1154 c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
1155 dbg_msg("LPT area LEBs: %d (%d - %d)",
1156 c->lpt_lebs, c->lpt_first, c->lpt_last);
1157 dbg_msg("orphan area LEBs: %d (%d - %d)",
1158 c->orph_lebs, c->orph_first, c->orph_last);
1159 dbg_msg("main area LEBs: %d (%d - %d)",
1160 c->main_lebs, c->main_first, c->leb_cnt - 1);
1161 dbg_msg("index LEBs: %d", c->lst.idx_lebs);
1162 dbg_msg("total index bytes: %lld (%lld KiB, %lld MiB)",
1163 c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
1164 dbg_msg("key hash type: %d", c->key_hash_type);
1165 dbg_msg("tree fanout: %d", c->fanout);
1166 dbg_msg("reserved GC LEB: %d", c->gc_lnum);
1167 dbg_msg("first main LEB: %d", c->main_first);
1168 dbg_msg("dead watermark: %d", c->dead_wm);
1169 dbg_msg("dark watermark: %d", c->dark_wm);
1170 x = (long long)c->main_lebs * c->dark_wm;
1171 dbg_msg("max. dark space: %lld (%lld KiB, %lld MiB)",
1172 x, x >> 10, x >> 20);
1173 dbg_msg("maximum bud bytes: %lld (%lld KiB, %lld MiB)",
1174 c->max_bud_bytes, c->max_bud_bytes >> 10,
1175 c->max_bud_bytes >> 20);
1176 dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
1177 c->bg_bud_bytes, c->bg_bud_bytes >> 10,
1178 c->bg_bud_bytes >> 20);
1179 dbg_msg("current bud bytes %lld (%lld KiB, %lld MiB)",
1180 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
1181 dbg_msg("max. seq. number: %llu", c->max_sqnum);
1182 dbg_msg("commit number: %llu", c->cmt_no);
1184 return 0;
1186 out_infos:
1187 spin_lock(&ubifs_infos_lock);
1188 list_del(&c->infos_list);
1189 spin_unlock(&ubifs_infos_lock);
1190 out_orphans:
1191 free_orphans(c);
1192 out_journal:
1193 destroy_journal(c);
1194 out_lpt:
1195 ubifs_lpt_free(c, 0);
1196 out_master:
1197 kfree(c->mst_node);
1198 kfree(c->rcvrd_mst_node);
1199 if (c->bgt)
1200 kthread_stop(c->bgt);
1201 out_wbufs:
1202 free_wbufs(c);
1203 out_cbuf:
1204 kfree(c->cbuf);
1205 out_dereg:
1206 dbg_failure_mode_deregistration(c);
1207 out_free:
1208 vfree(c->ileb_buf);
1209 vfree(c->sbuf);
1210 kfree(c->bottom_up_buf);
1211 UBIFS_DBG(vfree(c->dbg_buf));
1212 return err;
1216 * ubifs_umount - un-mount UBIFS file-system.
1217 * @c: UBIFS file-system description object
1219 * Note, this function is called to free allocated resourced when un-mounting,
1220 * as well as free resources when an error occurred while we were half way
1221 * through mounting (error path cleanup function). So it has to make sure the
1222 * resource was actually allocated before freeing it.
1224 static void ubifs_umount(struct ubifs_info *c)
1226 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
1227 c->vi.vol_id);
1229 spin_lock(&ubifs_infos_lock);
1230 list_del(&c->infos_list);
1231 spin_unlock(&ubifs_infos_lock);
1233 if (c->bgt)
1234 kthread_stop(c->bgt);
1236 destroy_journal(c);
1237 free_wbufs(c);
1238 free_orphans(c);
1239 ubifs_lpt_free(c, 0);
1241 kfree(c->cbuf);
1242 kfree(c->rcvrd_mst_node);
1243 kfree(c->mst_node);
1244 vfree(c->sbuf);
1245 kfree(c->bottom_up_buf);
1246 UBIFS_DBG(vfree(c->dbg_buf));
1247 vfree(c->ileb_buf);
1248 dbg_failure_mode_deregistration(c);
1252 * ubifs_remount_rw - re-mount in read-write mode.
1253 * @c: UBIFS file-system description object
1255 * UBIFS avoids allocating many unnecessary resources when mounted in read-only
1256 * mode. This function allocates the needed resources and re-mounts UBIFS in
1257 * read-write mode.
1259 static int ubifs_remount_rw(struct ubifs_info *c)
1261 int err, lnum;
1263 if (c->ro_media)
1264 return -EINVAL;
1266 mutex_lock(&c->umount_mutex);
1267 c->remounting_rw = 1;
1269 /* Check for enough free space */
1270 if (ubifs_calc_available(c, c->min_idx_lebs) <= 0) {
1271 ubifs_err("insufficient available space");
1272 err = -EINVAL;
1273 goto out;
1276 if (c->old_leb_cnt != c->leb_cnt) {
1277 struct ubifs_sb_node *sup;
1279 sup = ubifs_read_sb_node(c);
1280 if (IS_ERR(sup)) {
1281 err = PTR_ERR(sup);
1282 goto out;
1284 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
1285 err = ubifs_write_sb_node(c, sup);
1286 if (err)
1287 goto out;
1290 if (c->need_recovery) {
1291 ubifs_msg("completing deferred recovery");
1292 err = ubifs_write_rcvrd_mst_node(c);
1293 if (err)
1294 goto out;
1295 err = ubifs_recover_size(c);
1296 if (err)
1297 goto out;
1298 err = ubifs_clean_lebs(c, c->sbuf);
1299 if (err)
1300 goto out;
1301 err = ubifs_recover_inl_heads(c, c->sbuf);
1302 if (err)
1303 goto out;
1306 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
1307 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
1308 err = ubifs_write_master(c);
1309 if (err)
1310 goto out;
1313 c->ileb_buf = vmalloc(c->leb_size);
1314 if (!c->ileb_buf) {
1315 err = -ENOMEM;
1316 goto out;
1319 err = ubifs_lpt_init(c, 0, 1);
1320 if (err)
1321 goto out;
1323 err = alloc_wbufs(c);
1324 if (err)
1325 goto out;
1327 ubifs_create_buds_lists(c);
1329 /* Create background thread */
1330 c->bgt = kthread_create(ubifs_bg_thread, c, c->bgt_name);
1331 if (!c->bgt)
1332 c->bgt = ERR_PTR(-EINVAL);
1333 if (IS_ERR(c->bgt)) {
1334 err = PTR_ERR(c->bgt);
1335 c->bgt = NULL;
1336 ubifs_err("cannot spawn \"%s\", error %d",
1337 c->bgt_name, err);
1338 return err;
1340 wake_up_process(c->bgt);
1342 c->orph_buf = vmalloc(c->leb_size);
1343 if (!c->orph_buf)
1344 return -ENOMEM;
1346 /* Check for enough log space */
1347 lnum = c->lhead_lnum + 1;
1348 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
1349 lnum = UBIFS_LOG_LNUM;
1350 if (lnum == c->ltail_lnum) {
1351 err = ubifs_consolidate_log(c);
1352 if (err)
1353 goto out;
1356 if (c->need_recovery)
1357 err = ubifs_rcvry_gc_commit(c);
1358 else
1359 err = take_gc_lnum(c);
1360 if (err)
1361 goto out;
1363 if (c->need_recovery) {
1364 c->need_recovery = 0;
1365 ubifs_msg("deferred recovery completed");
1368 dbg_gen("re-mounted read-write");
1369 c->vfs_sb->s_flags &= ~MS_RDONLY;
1370 c->remounting_rw = 0;
1371 mutex_unlock(&c->umount_mutex);
1372 return 0;
1374 out:
1375 vfree(c->orph_buf);
1376 c->orph_buf = NULL;
1377 if (c->bgt) {
1378 kthread_stop(c->bgt);
1379 c->bgt = NULL;
1381 free_wbufs(c);
1382 vfree(c->ileb_buf);
1383 c->ileb_buf = NULL;
1384 ubifs_lpt_free(c, 1);
1385 c->remounting_rw = 0;
1386 mutex_unlock(&c->umount_mutex);
1387 return err;
1391 * commit_on_unmount - commit the journal when un-mounting.
1392 * @c: UBIFS file-system description object
1394 * This function is called during un-mounting and it commits the journal unless
1395 * the "fast unmount" mode is enabled. It also avoids committing the journal if
1396 * it contains too few data.
1398 * Sometimes recovery requires the journal to be committed at least once, and
1399 * this function takes care about this.
1401 static void commit_on_unmount(struct ubifs_info *c)
1403 if (!c->fast_unmount) {
1404 long long bud_bytes;
1406 spin_lock(&c->buds_lock);
1407 bud_bytes = c->bud_bytes;
1408 spin_unlock(&c->buds_lock);
1409 if (bud_bytes > c->leb_size)
1410 ubifs_run_commit(c);
1415 * ubifs_remount_ro - re-mount in read-only mode.
1416 * @c: UBIFS file-system description object
1418 * We rely on VFS to have stopped writing. Possibly the background thread could
1419 * be running a commit, however kthread_stop will wait in that case.
1421 static void ubifs_remount_ro(struct ubifs_info *c)
1423 int i, err;
1425 ubifs_assert(!c->need_recovery);
1426 commit_on_unmount(c);
1428 mutex_lock(&c->umount_mutex);
1429 if (c->bgt) {
1430 kthread_stop(c->bgt);
1431 c->bgt = NULL;
1434 for (i = 0; i < c->jhead_cnt; i++) {
1435 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1436 del_timer_sync(&c->jheads[i].wbuf.timer);
1439 if (!c->ro_media) {
1440 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1441 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1442 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1443 err = ubifs_write_master(c);
1444 if (err)
1445 ubifs_ro_mode(c, err);
1448 ubifs_destroy_idx_gc(c);
1449 free_wbufs(c);
1450 vfree(c->orph_buf);
1451 c->orph_buf = NULL;
1452 vfree(c->ileb_buf);
1453 c->ileb_buf = NULL;
1454 ubifs_lpt_free(c, 1);
1455 mutex_unlock(&c->umount_mutex);
1458 static void ubifs_put_super(struct super_block *sb)
1460 int i;
1461 struct ubifs_info *c = sb->s_fs_info;
1463 ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
1464 c->vi.vol_id);
1466 * The following asserts are only valid if there has not been a failure
1467 * of the media. For example, there will be dirty inodes if we failed
1468 * to write them back because of I/O errors.
1470 ubifs_assert(atomic_long_read(&c->dirty_pg_cnt) == 0);
1471 ubifs_assert(c->budg_idx_growth == 0);
1472 ubifs_assert(c->budg_data_growth == 0);
1475 * The 'c->umount_lock' prevents races between UBIFS memory shrinker
1476 * and file system un-mount. Namely, it prevents the shrinker from
1477 * picking this superblock for shrinking - it will be just skipped if
1478 * the mutex is locked.
1480 mutex_lock(&c->umount_mutex);
1481 if (!(c->vfs_sb->s_flags & MS_RDONLY)) {
1483 * First of all kill the background thread to make sure it does
1484 * not interfere with un-mounting and freeing resources.
1486 if (c->bgt) {
1487 kthread_stop(c->bgt);
1488 c->bgt = NULL;
1491 /* Synchronize write-buffers */
1492 if (c->jheads)
1493 for (i = 0; i < c->jhead_cnt; i++) {
1494 ubifs_wbuf_sync(&c->jheads[i].wbuf);
1495 del_timer_sync(&c->jheads[i].wbuf.timer);
1499 * On fatal errors c->ro_media is set to 1, in which case we do
1500 * not write the master node.
1502 if (!c->ro_media) {
1504 * We are being cleanly unmounted which means the
1505 * orphans were killed - indicate this in the master
1506 * node. Also save the reserved GC LEB number.
1508 int err;
1510 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
1511 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
1512 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
1513 err = ubifs_write_master(c);
1514 if (err)
1516 * Recovery will attempt to fix the master area
1517 * next mount, so we just print a message and
1518 * continue to unmount normally.
1520 ubifs_err("failed to write master node, "
1521 "error %d", err);
1525 ubifs_umount(c);
1526 bdi_destroy(&c->bdi);
1527 ubi_close_volume(c->ubi);
1528 mutex_unlock(&c->umount_mutex);
1529 kfree(c);
1532 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
1534 int err;
1535 struct ubifs_info *c = sb->s_fs_info;
1537 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
1539 err = ubifs_parse_options(c, data, 1);
1540 if (err) {
1541 ubifs_err("invalid or unknown remount parameter");
1542 return err;
1544 if ((sb->s_flags & MS_RDONLY) && !(*flags & MS_RDONLY)) {
1545 err = ubifs_remount_rw(c);
1546 if (err)
1547 return err;
1548 } else if (!(sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
1549 ubifs_remount_ro(c);
1551 return 0;
1554 struct super_operations ubifs_super_operations = {
1555 .alloc_inode = ubifs_alloc_inode,
1556 .destroy_inode = ubifs_destroy_inode,
1557 .put_super = ubifs_put_super,
1558 .write_inode = ubifs_write_inode,
1559 .delete_inode = ubifs_delete_inode,
1560 .statfs = ubifs_statfs,
1561 .dirty_inode = ubifs_dirty_inode,
1562 .remount_fs = ubifs_remount_fs,
1563 .show_options = ubifs_show_options,
1564 .sync_fs = ubifs_sync_fs,
1568 * open_ubi - parse UBI device name string and open the UBI device.
1569 * @name: UBI volume name
1570 * @mode: UBI volume open mode
1572 * There are several ways to specify UBI volumes when mounting UBIFS:
1573 * o ubiX_Y - UBI device number X, volume Y;
1574 * o ubiY - UBI device number 0, volume Y;
1575 * o ubiX:NAME - mount UBI device X, volume with name NAME;
1576 * o ubi:NAME - mount UBI device 0, volume with name NAME.
1578 * Alternative '!' separator may be used instead of ':' (because some shells
1579 * like busybox may interpret ':' as an NFS host name separator). This function
1580 * returns ubi volume object in case of success and a negative error code in
1581 * case of failure.
1583 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
1585 int dev, vol;
1586 char *endptr;
1588 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
1589 return ERR_PTR(-EINVAL);
1591 /* ubi:NAME method */
1592 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0')
1593 return ubi_open_volume_nm(0, name + 4, mode);
1595 if (!isdigit(name[3]))
1596 return ERR_PTR(-EINVAL);
1598 dev = simple_strtoul(name + 3, &endptr, 0);
1600 /* ubiY method */
1601 if (*endptr == '\0')
1602 return ubi_open_volume(0, dev, mode);
1604 /* ubiX_Y method */
1605 if (*endptr == '_' && isdigit(endptr[1])) {
1606 vol = simple_strtoul(endptr + 1, &endptr, 0);
1607 if (*endptr != '\0')
1608 return ERR_PTR(-EINVAL);
1609 return ubi_open_volume(dev, vol, mode);
1612 /* ubiX:NAME method */
1613 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0')
1614 return ubi_open_volume_nm(dev, ++endptr, mode);
1616 return ERR_PTR(-EINVAL);
1619 static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
1621 struct ubi_volume_desc *ubi = sb->s_fs_info;
1622 struct ubifs_info *c;
1623 struct inode *root;
1624 int err;
1626 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
1627 if (!c)
1628 return -ENOMEM;
1630 spin_lock_init(&c->cnt_lock);
1631 spin_lock_init(&c->cs_lock);
1632 spin_lock_init(&c->buds_lock);
1633 spin_lock_init(&c->space_lock);
1634 spin_lock_init(&c->orphan_lock);
1635 init_rwsem(&c->commit_sem);
1636 mutex_init(&c->lp_mutex);
1637 mutex_init(&c->tnc_mutex);
1638 mutex_init(&c->log_mutex);
1639 mutex_init(&c->mst_mutex);
1640 mutex_init(&c->umount_mutex);
1641 init_waitqueue_head(&c->cmt_wq);
1642 c->buds = RB_ROOT;
1643 c->old_idx = RB_ROOT;
1644 c->size_tree = RB_ROOT;
1645 c->orph_tree = RB_ROOT;
1646 INIT_LIST_HEAD(&c->infos_list);
1647 INIT_LIST_HEAD(&c->idx_gc);
1648 INIT_LIST_HEAD(&c->replay_list);
1649 INIT_LIST_HEAD(&c->replay_buds);
1650 INIT_LIST_HEAD(&c->uncat_list);
1651 INIT_LIST_HEAD(&c->empty_list);
1652 INIT_LIST_HEAD(&c->freeable_list);
1653 INIT_LIST_HEAD(&c->frdi_idx_list);
1654 INIT_LIST_HEAD(&c->unclean_leb_list);
1655 INIT_LIST_HEAD(&c->old_buds);
1656 INIT_LIST_HEAD(&c->orph_list);
1657 INIT_LIST_HEAD(&c->orph_new);
1659 c->highest_inum = UBIFS_FIRST_INO;
1660 get_random_bytes(&c->vfs_gen, sizeof(int));
1661 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
1663 ubi_get_volume_info(ubi, &c->vi);
1664 ubi_get_device_info(c->vi.ubi_num, &c->di);
1666 /* Re-open the UBI device in read-write mode */
1667 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
1668 if (IS_ERR(c->ubi)) {
1669 err = PTR_ERR(c->ubi);
1670 goto out_free;
1674 * UBIFS provids 'backing_dev_info' in order to disable readahead. For
1675 * UBIFS, I/O is not deferred, it is done immediately in readpage,
1676 * which means the user would have to wait not just for their own I/O
1677 * but the readahead I/O as well i.e. completely pointless.
1679 * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
1681 c->bdi.capabilities = BDI_CAP_MAP_COPY;
1682 c->bdi.unplug_io_fn = default_unplug_io_fn;
1683 err = bdi_init(&c->bdi);
1684 if (err)
1685 goto out_close;
1687 err = ubifs_parse_options(c, data, 0);
1688 if (err)
1689 goto out_bdi;
1691 c->vfs_sb = sb;
1693 sb->s_fs_info = c;
1694 sb->s_magic = UBIFS_SUPER_MAGIC;
1695 sb->s_blocksize = UBIFS_BLOCK_SIZE;
1696 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
1697 sb->s_dev = c->vi.cdev;
1698 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
1699 if (c->max_inode_sz > MAX_LFS_FILESIZE)
1700 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
1701 sb->s_op = &ubifs_super_operations;
1703 mutex_lock(&c->umount_mutex);
1704 err = mount_ubifs(c);
1705 if (err) {
1706 ubifs_assert(err < 0);
1707 goto out_unlock;
1710 /* Read the root inode */
1711 root = ubifs_iget(sb, UBIFS_ROOT_INO);
1712 if (IS_ERR(root)) {
1713 err = PTR_ERR(root);
1714 goto out_umount;
1717 sb->s_root = d_alloc_root(root);
1718 if (!sb->s_root)
1719 goto out_iput;
1721 mutex_unlock(&c->umount_mutex);
1723 return 0;
1725 out_iput:
1726 iput(root);
1727 out_umount:
1728 ubifs_umount(c);
1729 out_unlock:
1730 mutex_unlock(&c->umount_mutex);
1731 out_bdi:
1732 bdi_destroy(&c->bdi);
1733 out_close:
1734 ubi_close_volume(c->ubi);
1735 out_free:
1736 kfree(c);
1737 return err;
1740 static int sb_test(struct super_block *sb, void *data)
1742 dev_t *dev = data;
1744 return sb->s_dev == *dev;
1747 static int sb_set(struct super_block *sb, void *data)
1749 dev_t *dev = data;
1751 sb->s_dev = *dev;
1752 return 0;
1755 static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
1756 const char *name, void *data, struct vfsmount *mnt)
1758 struct ubi_volume_desc *ubi;
1759 struct ubi_volume_info vi;
1760 struct super_block *sb;
1761 int err;
1763 dbg_gen("name %s, flags %#x", name, flags);
1766 * Get UBI device number and volume ID. Mount it read-only so far
1767 * because this might be a new mount point, and UBI allows only one
1768 * read-write user at a time.
1770 ubi = open_ubi(name, UBI_READONLY);
1771 if (IS_ERR(ubi)) {
1772 ubifs_err("cannot open \"%s\", error %d",
1773 name, (int)PTR_ERR(ubi));
1774 return PTR_ERR(ubi);
1776 ubi_get_volume_info(ubi, &vi);
1778 dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
1780 sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
1781 if (IS_ERR(sb)) {
1782 err = PTR_ERR(sb);
1783 goto out_close;
1786 if (sb->s_root) {
1787 /* A new mount point for already mounted UBIFS */
1788 dbg_gen("this ubi volume is already mounted");
1789 if ((flags ^ sb->s_flags) & MS_RDONLY) {
1790 err = -EBUSY;
1791 goto out_deact;
1793 } else {
1794 sb->s_flags = flags;
1796 * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
1797 * replaced by 'c'.
1799 sb->s_fs_info = ubi;
1800 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
1801 if (err)
1802 goto out_deact;
1803 /* We do not support atime */
1804 sb->s_flags |= MS_ACTIVE | MS_NOATIME;
1807 /* 'fill_super()' opens ubi again so we must close it here */
1808 ubi_close_volume(ubi);
1810 return simple_set_mnt(mnt, sb);
1812 out_deact:
1813 up_write(&sb->s_umount);
1814 deactivate_super(sb);
1815 out_close:
1816 ubi_close_volume(ubi);
1817 return err;
1820 static void ubifs_kill_sb(struct super_block *sb)
1822 struct ubifs_info *c = sb->s_fs_info;
1825 * We do 'commit_on_unmount()' here instead of 'ubifs_put_super()'
1826 * in order to be outside BKL.
1828 if (sb->s_root && !(sb->s_flags & MS_RDONLY))
1829 commit_on_unmount(c);
1830 /* The un-mount routine is actually done in put_super() */
1831 generic_shutdown_super(sb);
1834 static struct file_system_type ubifs_fs_type = {
1835 .name = "ubifs",
1836 .owner = THIS_MODULE,
1837 .get_sb = ubifs_get_sb,
1838 .kill_sb = ubifs_kill_sb
1842 * Inode slab cache constructor.
1844 static void inode_slab_ctor(struct kmem_cache *cachep, void *obj)
1846 struct ubifs_inode *ui = obj;
1847 inode_init_once(&ui->vfs_inode);
1850 static int __init ubifs_init(void)
1852 int err;
1854 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24);
1856 /* Make sure node sizes are 8-byte aligned */
1857 BUILD_BUG_ON(UBIFS_CH_SZ & 7);
1858 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7);
1859 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7);
1860 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7);
1861 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7);
1862 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7);
1863 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7);
1864 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7);
1865 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7);
1866 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7);
1867 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7);
1869 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7);
1870 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7);
1871 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7);
1872 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7);
1873 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7);
1874 BUILD_BUG_ON(MIN_WRITE_SZ & 7);
1876 /* Check min. node size */
1877 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ);
1878 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ);
1879 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ);
1880 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ);
1882 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
1883 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ);
1884 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ);
1885 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ);
1887 /* Defined node sizes */
1888 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096);
1889 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512);
1890 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160);
1891 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64);
1894 * We require that PAGE_CACHE_SIZE is greater-than-or-equal-to
1895 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
1897 if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
1898 ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
1899 " at least 4096 bytes",
1900 (unsigned int)PAGE_CACHE_SIZE);
1901 return -EINVAL;
1904 err = register_filesystem(&ubifs_fs_type);
1905 if (err) {
1906 ubifs_err("cannot register file system, error %d", err);
1907 return err;
1910 err = -ENOMEM;
1911 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
1912 sizeof(struct ubifs_inode), 0,
1913 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
1914 &inode_slab_ctor);
1915 if (!ubifs_inode_slab)
1916 goto out_reg;
1918 register_shrinker(&ubifs_shrinker_info);
1920 err = ubifs_compressors_init();
1921 if (err)
1922 goto out_compr;
1924 return 0;
1926 out_compr:
1927 unregister_shrinker(&ubifs_shrinker_info);
1928 kmem_cache_destroy(ubifs_inode_slab);
1929 out_reg:
1930 unregister_filesystem(&ubifs_fs_type);
1931 return err;
1933 /* late_initcall to let compressors initialize first */
1934 late_initcall(ubifs_init);
1936 static void __exit ubifs_exit(void)
1938 ubifs_assert(list_empty(&ubifs_infos));
1939 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
1941 ubifs_compressors_exit();
1942 unregister_shrinker(&ubifs_shrinker_info);
1943 kmem_cache_destroy(ubifs_inode_slab);
1944 unregister_filesystem(&ubifs_fs_type);
1946 module_exit(ubifs_exit);
1948 MODULE_LICENSE("GPL");
1949 MODULE_VERSION(__stringify(UBIFS_VERSION));
1950 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
1951 MODULE_DESCRIPTION("UBIFS - UBI File System");