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 superblock. The superblock is stored at the first
25 * LEB of the volume and is never changed by UBIFS. Only user-space tools may
26 * change it. The superblock node mostly contains geometry information.
30 #include <linux/random.h>
31 #include <linux/math64.h>
34 * Default journal size in logical eraseblocks as a percent of total
37 #define DEFAULT_JNL_PERCENT 5
39 /* Default maximum journal size in bytes */
40 #define DEFAULT_MAX_JNL (32*1024*1024)
42 /* Default indexing tree fanout */
43 #define DEFAULT_FANOUT 8
45 /* Default number of data journal heads */
46 #define DEFAULT_JHEADS_CNT 1
48 /* Default positions of different LEBs in the main area */
49 #define DEFAULT_IDX_LEB 0
50 #define DEFAULT_DATA_LEB 1
51 #define DEFAULT_GC_LEB 2
53 /* Default number of LEB numbers in LPT's save table */
54 #define DEFAULT_LSAVE_CNT 256
56 /* Default reserved pool size as a percent of maximum free space */
57 #define DEFAULT_RP_PERCENT 5
59 /* The default maximum size of reserved pool in bytes */
60 #define DEFAULT_MAX_RP_SIZE (5*1024*1024)
62 /* Default time granularity in nanoseconds */
63 #define DEFAULT_TIME_GRAN 1000000000
66 * create_default_filesystem - format empty UBI volume.
67 * @c: UBIFS file-system description object
69 * This function creates default empty file-system. Returns zero in case of
70 * success and a negative error code in case of failure.
72 static int create_default_filesystem(struct ubifs_info *c)
74 struct ubifs_sb_node *sup;
75 struct ubifs_mst_node *mst;
76 struct ubifs_idx_node *idx;
77 struct ubifs_branch *br;
78 struct ubifs_ino_node *ino;
79 struct ubifs_cs_node *cs;
81 int err, tmp, jnl_lebs, log_lebs, max_buds, main_lebs, main_first;
82 int lpt_lebs, lpt_first, orph_lebs, big_lpt, ino_waste, sup_flags = 0;
83 int min_leb_cnt = UBIFS_MIN_LEB_CNT;
84 long long tmp64, main_bytes;
87 /* Some functions called from here depend on the @c->key_len filed */
88 c->key_len = UBIFS_SK_LEN;
91 * First of all, we have to calculate default file-system geometry -
92 * log size, journal size, etc.
94 if (c->leb_cnt < 0x7FFFFFFF / DEFAULT_JNL_PERCENT)
95 /* We can first multiply then divide and have no overflow */
96 jnl_lebs = c->leb_cnt * DEFAULT_JNL_PERCENT / 100;
98 jnl_lebs = (c->leb_cnt / 100) * DEFAULT_JNL_PERCENT;
100 if (jnl_lebs < UBIFS_MIN_JNL_LEBS)
101 jnl_lebs = UBIFS_MIN_JNL_LEBS;
102 if (jnl_lebs * c->leb_size > DEFAULT_MAX_JNL)
103 jnl_lebs = DEFAULT_MAX_JNL / c->leb_size;
106 * The log should be large enough to fit reference nodes for all bud
107 * LEBs. Because buds do not have to start from the beginning of LEBs
108 * (half of the LEB may contain committed data), the log should
109 * generally be larger, make it twice as large.
111 tmp = 2 * (c->ref_node_alsz * jnl_lebs) + c->leb_size - 1;
112 log_lebs = tmp / c->leb_size;
113 /* Plus one LEB reserved for commit */
115 if (c->leb_cnt - min_leb_cnt > 8) {
116 /* And some extra space to allow writes while committing */
121 max_buds = jnl_lebs - log_lebs;
122 if (max_buds < UBIFS_MIN_BUD_LEBS)
123 max_buds = UBIFS_MIN_BUD_LEBS;
126 * Orphan nodes are stored in a separate area. One node can store a lot
127 * of orphan inode numbers, but when new orphan comes we just add a new
128 * orphan node. At some point the nodes are consolidated into one
131 orph_lebs = UBIFS_MIN_ORPH_LEBS;
132 #ifdef CONFIG_UBIFS_FS_DEBUG
133 if (c->leb_cnt - min_leb_cnt > 1)
135 * For debugging purposes it is better to have at least 2
136 * orphan LEBs, because the orphan subsystem would need to do
137 * consolidations and would be stressed more.
142 main_lebs = c->leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS - log_lebs;
143 main_lebs -= orph_lebs;
145 lpt_first = UBIFS_LOG_LNUM + log_lebs;
146 c->lsave_cnt = DEFAULT_LSAVE_CNT;
147 c->max_leb_cnt = c->leb_cnt;
148 err = ubifs_create_dflt_lpt(c, &main_lebs, lpt_first, &lpt_lebs,
153 dbg_gen("LEB Properties Tree created (LEBs %d-%d)", lpt_first,
154 lpt_first + lpt_lebs - 1);
156 main_first = c->leb_cnt - main_lebs;
158 /* Create default superblock */
159 tmp = ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size);
160 sup = kzalloc(tmp, GFP_KERNEL);
164 tmp64 = (long long)max_buds * c->leb_size;
166 sup_flags |= UBIFS_FLG_BIGLPT;
168 sup->ch.node_type = UBIFS_SB_NODE;
169 sup->key_hash = UBIFS_KEY_HASH_R5;
170 sup->flags = cpu_to_le32(sup_flags);
171 sup->min_io_size = cpu_to_le32(c->min_io_size);
172 sup->leb_size = cpu_to_le32(c->leb_size);
173 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
174 sup->max_leb_cnt = cpu_to_le32(c->max_leb_cnt);
175 sup->max_bud_bytes = cpu_to_le64(tmp64);
176 sup->log_lebs = cpu_to_le32(log_lebs);
177 sup->lpt_lebs = cpu_to_le32(lpt_lebs);
178 sup->orph_lebs = cpu_to_le32(orph_lebs);
179 sup->jhead_cnt = cpu_to_le32(DEFAULT_JHEADS_CNT);
180 sup->fanout = cpu_to_le32(DEFAULT_FANOUT);
181 sup->lsave_cnt = cpu_to_le32(c->lsave_cnt);
182 sup->fmt_version = cpu_to_le32(UBIFS_FORMAT_VERSION);
183 sup->time_gran = cpu_to_le32(DEFAULT_TIME_GRAN);
184 if (c->mount_opts.override_compr)
185 sup->default_compr = cpu_to_le16(c->mount_opts.compr_type);
187 sup->default_compr = cpu_to_le16(UBIFS_COMPR_LZO);
189 generate_random_uuid(sup->uuid);
191 main_bytes = (long long)main_lebs * c->leb_size;
192 tmp64 = div_u64(main_bytes * DEFAULT_RP_PERCENT, 100);
193 if (tmp64 > DEFAULT_MAX_RP_SIZE)
194 tmp64 = DEFAULT_MAX_RP_SIZE;
195 sup->rp_size = cpu_to_le64(tmp64);
196 sup->ro_compat_version = cpu_to_le32(UBIFS_RO_COMPAT_VERSION);
198 err = ubifs_write_node(c, sup, UBIFS_SB_NODE_SZ, 0, 0, UBI_LONGTERM);
203 dbg_gen("default superblock created at LEB 0:0");
205 /* Create default master node */
206 mst = kzalloc(c->mst_node_alsz, GFP_KERNEL);
210 mst->ch.node_type = UBIFS_MST_NODE;
211 mst->log_lnum = cpu_to_le32(UBIFS_LOG_LNUM);
212 mst->highest_inum = cpu_to_le64(UBIFS_FIRST_INO);
214 mst->root_lnum = cpu_to_le32(main_first + DEFAULT_IDX_LEB);
216 tmp = ubifs_idx_node_sz(c, 1);
217 mst->root_len = cpu_to_le32(tmp);
218 mst->gc_lnum = cpu_to_le32(main_first + DEFAULT_GC_LEB);
219 mst->ihead_lnum = cpu_to_le32(main_first + DEFAULT_IDX_LEB);
220 mst->ihead_offs = cpu_to_le32(ALIGN(tmp, c->min_io_size));
221 mst->index_size = cpu_to_le64(ALIGN(tmp, 8));
222 mst->lpt_lnum = cpu_to_le32(c->lpt_lnum);
223 mst->lpt_offs = cpu_to_le32(c->lpt_offs);
224 mst->nhead_lnum = cpu_to_le32(c->nhead_lnum);
225 mst->nhead_offs = cpu_to_le32(c->nhead_offs);
226 mst->ltab_lnum = cpu_to_le32(c->ltab_lnum);
227 mst->ltab_offs = cpu_to_le32(c->ltab_offs);
228 mst->lsave_lnum = cpu_to_le32(c->lsave_lnum);
229 mst->lsave_offs = cpu_to_le32(c->lsave_offs);
230 mst->lscan_lnum = cpu_to_le32(main_first);
231 mst->empty_lebs = cpu_to_le32(main_lebs - 2);
232 mst->idx_lebs = cpu_to_le32(1);
233 mst->leb_cnt = cpu_to_le32(c->leb_cnt);
235 /* Calculate lprops statistics */
237 tmp64 -= ALIGN(ubifs_idx_node_sz(c, 1), c->min_io_size);
238 tmp64 -= ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size);
239 mst->total_free = cpu_to_le64(tmp64);
241 tmp64 = ALIGN(ubifs_idx_node_sz(c, 1), c->min_io_size);
242 ino_waste = ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size) -
245 tmp64 -= ALIGN(ubifs_idx_node_sz(c, 1), 8);
246 mst->total_dirty = cpu_to_le64(tmp64);
248 /* The indexing LEB does not contribute to dark space */
249 tmp64 = (c->main_lebs - 1) * c->dark_wm;
250 mst->total_dark = cpu_to_le64(tmp64);
252 mst->total_used = cpu_to_le64(UBIFS_INO_NODE_SZ);
254 err = ubifs_write_node(c, mst, UBIFS_MST_NODE_SZ, UBIFS_MST_LNUM, 0,
260 err = ubifs_write_node(c, mst, UBIFS_MST_NODE_SZ, UBIFS_MST_LNUM + 1, 0,
266 dbg_gen("default master node created at LEB %d:0", UBIFS_MST_LNUM);
268 /* Create the root indexing node */
269 tmp = ubifs_idx_node_sz(c, 1);
270 idx = kzalloc(ALIGN(tmp, c->min_io_size), GFP_KERNEL);
274 c->key_fmt = UBIFS_SIMPLE_KEY_FMT;
275 c->key_hash = key_r5_hash;
277 idx->ch.node_type = UBIFS_IDX_NODE;
278 idx->child_cnt = cpu_to_le16(1);
279 ino_key_init(c, &key, UBIFS_ROOT_INO);
280 br = ubifs_idx_branch(c, idx, 0);
281 key_write_idx(c, &key, &br->key);
282 br->lnum = cpu_to_le32(main_first + DEFAULT_DATA_LEB);
283 br->len = cpu_to_le32(UBIFS_INO_NODE_SZ);
284 err = ubifs_write_node(c, idx, tmp, main_first + DEFAULT_IDX_LEB, 0,
290 dbg_gen("default root indexing node created LEB %d:0",
291 main_first + DEFAULT_IDX_LEB);
293 /* Create default root inode */
294 tmp = ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size);
295 ino = kzalloc(tmp, GFP_KERNEL);
299 ino_key_init_flash(c, &ino->key, UBIFS_ROOT_INO);
300 ino->ch.node_type = UBIFS_INO_NODE;
301 ino->creat_sqnum = cpu_to_le64(++c->max_sqnum);
302 ino->nlink = cpu_to_le32(2);
303 tmp_le64 = cpu_to_le64(CURRENT_TIME_SEC.tv_sec);
304 ino->atime_sec = tmp_le64;
305 ino->ctime_sec = tmp_le64;
306 ino->mtime_sec = tmp_le64;
310 ino->mode = cpu_to_le32(S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO);
311 ino->size = cpu_to_le64(UBIFS_INO_NODE_SZ);
313 /* Set compression enabled by default */
314 ino->flags = cpu_to_le32(UBIFS_COMPR_FL);
316 err = ubifs_write_node(c, ino, UBIFS_INO_NODE_SZ,
317 main_first + DEFAULT_DATA_LEB, 0,
323 dbg_gen("root inode created at LEB %d:0",
324 main_first + DEFAULT_DATA_LEB);
327 * The first node in the log has to be the commit start node. This is
328 * always the case during normal file-system operation. Write a fake
329 * commit start node to the log.
331 tmp = ALIGN(UBIFS_CS_NODE_SZ, c->min_io_size);
332 cs = kzalloc(tmp, GFP_KERNEL);
336 cs->ch.node_type = UBIFS_CS_NODE;
337 err = ubifs_write_node(c, cs, UBIFS_CS_NODE_SZ, UBIFS_LOG_LNUM,
341 ubifs_msg("default file-system created");
346 * validate_sb - validate superblock node.
347 * @c: UBIFS file-system description object
348 * @sup: superblock node
350 * This function validates superblock node @sup. Since most of data was read
351 * from the superblock and stored in @c, the function validates fields in @c
352 * instead. Returns zero in case of success and %-EINVAL in case of validation
355 static int validate_sb(struct ubifs_info *c, struct ubifs_sb_node *sup)
358 int err = 1, min_leb_cnt;
365 if (sup->key_fmt != UBIFS_SIMPLE_KEY_FMT) {
370 if (le32_to_cpu(sup->min_io_size) != c->min_io_size) {
371 ubifs_err("min. I/O unit mismatch: %d in superblock, %d real",
372 le32_to_cpu(sup->min_io_size), c->min_io_size);
376 if (le32_to_cpu(sup->leb_size) != c->leb_size) {
377 ubifs_err("LEB size mismatch: %d in superblock, %d real",
378 le32_to_cpu(sup->leb_size), c->leb_size);
382 if (c->log_lebs < UBIFS_MIN_LOG_LEBS ||
383 c->lpt_lebs < UBIFS_MIN_LPT_LEBS ||
384 c->orph_lebs < UBIFS_MIN_ORPH_LEBS ||
385 c->main_lebs < UBIFS_MIN_MAIN_LEBS) {
391 * Calculate minimum allowed amount of main area LEBs. This is very
392 * similar to %UBIFS_MIN_LEB_CNT, but we take into account real what we
393 * have just read from the superblock.
395 min_leb_cnt = UBIFS_SB_LEBS + UBIFS_MST_LEBS + c->log_lebs;
396 min_leb_cnt += c->lpt_lebs + c->orph_lebs + c->jhead_cnt + 6;
398 if (c->leb_cnt < min_leb_cnt || c->leb_cnt > c->vi.size) {
399 ubifs_err("bad LEB count: %d in superblock, %d on UBI volume, "
400 "%d minimum required", c->leb_cnt, c->vi.size,
405 if (c->max_leb_cnt < c->leb_cnt) {
406 ubifs_err("max. LEB count %d less than LEB count %d",
407 c->max_leb_cnt, c->leb_cnt);
411 if (c->main_lebs < UBIFS_MIN_MAIN_LEBS) {
416 if (c->max_bud_bytes < (long long)c->leb_size * UBIFS_MIN_BUD_LEBS ||
417 c->max_bud_bytes > (long long)c->leb_size * c->main_lebs) {
422 if (c->jhead_cnt < NONDATA_JHEADS_CNT + 1 ||
423 c->jhead_cnt > NONDATA_JHEADS_CNT + UBIFS_MAX_JHEADS) {
428 if (c->fanout < UBIFS_MIN_FANOUT ||
429 ubifs_idx_node_sz(c, c->fanout) > c->leb_size) {
434 if (c->lsave_cnt < 0 || (c->lsave_cnt > DEFAULT_LSAVE_CNT &&
435 c->lsave_cnt > c->max_leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS -
436 c->log_lebs - c->lpt_lebs - c->orph_lebs)) {
441 if (UBIFS_SB_LEBS + UBIFS_MST_LEBS + c->log_lebs + c->lpt_lebs +
442 c->orph_lebs + c->main_lebs != c->leb_cnt) {
447 if (c->default_compr < 0 || c->default_compr >= UBIFS_COMPR_TYPES_CNT) {
452 max_bytes = c->main_lebs * (long long)c->leb_size;
453 if (c->rp_size < 0 || max_bytes < c->rp_size) {
458 if (le32_to_cpu(sup->time_gran) > 1000000000 ||
459 le32_to_cpu(sup->time_gran) < 1) {
467 ubifs_err("bad superblock, error %d", err);
468 dbg_dump_node(c, sup);
473 * ubifs_read_sb_node - read superblock node.
474 * @c: UBIFS file-system description object
476 * This function returns a pointer to the superblock node or a negative error
479 struct ubifs_sb_node *ubifs_read_sb_node(struct ubifs_info *c)
481 struct ubifs_sb_node *sup;
484 sup = kmalloc(ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size), GFP_NOFS);
486 return ERR_PTR(-ENOMEM);
488 err = ubifs_read_node(c, sup, UBIFS_SB_NODE, UBIFS_SB_NODE_SZ,
499 * ubifs_write_sb_node - write superblock node.
500 * @c: UBIFS file-system description object
501 * @sup: superblock node read with 'ubifs_read_sb_node()'
503 * This function returns %0 on success and a negative error code on failure.
505 int ubifs_write_sb_node(struct ubifs_info *c, struct ubifs_sb_node *sup)
507 int len = ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size);
509 ubifs_prepare_node(c, sup, UBIFS_SB_NODE_SZ, 1);
510 return ubifs_leb_change(c, UBIFS_SB_LNUM, sup, len, UBI_LONGTERM);
514 * ubifs_read_superblock - read superblock.
515 * @c: UBIFS file-system description object
517 * This function finds, reads and checks the superblock. If an empty UBI volume
518 * is being mounted, this function creates default superblock. Returns zero in
519 * case of success, and a negative error code in case of failure.
521 int ubifs_read_superblock(struct ubifs_info *c)
524 struct ubifs_sb_node *sup;
527 err = create_default_filesystem(c);
532 sup = ubifs_read_sb_node(c);
536 c->fmt_version = le32_to_cpu(sup->fmt_version);
537 c->ro_compat_version = le32_to_cpu(sup->ro_compat_version);
540 * The software supports all previous versions but not future versions,
541 * due to the unavailability of time-travelling equipment.
543 if (c->fmt_version > UBIFS_FORMAT_VERSION) {
544 struct super_block *sb = c->vfs_sb;
545 int mounting_ro = sb->s_flags & MS_RDONLY;
547 ubifs_assert(!c->ro_media || mounting_ro);
549 c->ro_compat_version > UBIFS_RO_COMPAT_VERSION) {
550 ubifs_err("on-flash format version is w%d/r%d, but "
551 "software only supports up to version "
552 "w%d/r%d", c->fmt_version,
553 c->ro_compat_version, UBIFS_FORMAT_VERSION,
554 UBIFS_RO_COMPAT_VERSION);
555 if (c->ro_compat_version <= UBIFS_RO_COMPAT_VERSION) {
556 ubifs_msg("only R/O mounting is possible");
564 * The FS is mounted R/O, and the media format is
565 * R/O-compatible with the UBIFS implementation, so we can
571 if (c->fmt_version < 3) {
572 ubifs_err("on-flash format version %d is not supported",
578 switch (sup->key_hash) {
579 case UBIFS_KEY_HASH_R5:
580 c->key_hash = key_r5_hash;
581 c->key_hash_type = UBIFS_KEY_HASH_R5;
584 case UBIFS_KEY_HASH_TEST:
585 c->key_hash = key_test_hash;
586 c->key_hash_type = UBIFS_KEY_HASH_TEST;
590 c->key_fmt = sup->key_fmt;
592 switch (c->key_fmt) {
593 case UBIFS_SIMPLE_KEY_FMT:
594 c->key_len = UBIFS_SK_LEN;
597 ubifs_err("unsupported key format");
602 c->leb_cnt = le32_to_cpu(sup->leb_cnt);
603 c->max_leb_cnt = le32_to_cpu(sup->max_leb_cnt);
604 c->max_bud_bytes = le64_to_cpu(sup->max_bud_bytes);
605 c->log_lebs = le32_to_cpu(sup->log_lebs);
606 c->lpt_lebs = le32_to_cpu(sup->lpt_lebs);
607 c->orph_lebs = le32_to_cpu(sup->orph_lebs);
608 c->jhead_cnt = le32_to_cpu(sup->jhead_cnt) + NONDATA_JHEADS_CNT;
609 c->fanout = le32_to_cpu(sup->fanout);
610 c->lsave_cnt = le32_to_cpu(sup->lsave_cnt);
611 c->rp_size = le64_to_cpu(sup->rp_size);
612 c->rp_uid = le32_to_cpu(sup->rp_uid);
613 c->rp_gid = le32_to_cpu(sup->rp_gid);
614 sup_flags = le32_to_cpu(sup->flags);
615 if (!c->mount_opts.override_compr)
616 c->default_compr = le16_to_cpu(sup->default_compr);
618 c->vfs_sb->s_time_gran = le32_to_cpu(sup->time_gran);
619 memcpy(&c->uuid, &sup->uuid, 16);
620 c->big_lpt = !!(sup_flags & UBIFS_FLG_BIGLPT);
622 /* Automatically increase file system size to the maximum size */
623 c->old_leb_cnt = c->leb_cnt;
624 if (c->leb_cnt < c->vi.size && c->leb_cnt < c->max_leb_cnt) {
625 c->leb_cnt = min_t(int, c->max_leb_cnt, c->vi.size);
626 if (c->vfs_sb->s_flags & MS_RDONLY)
627 dbg_mnt("Auto resizing (ro) from %d LEBs to %d LEBs",
628 c->old_leb_cnt, c->leb_cnt);
630 dbg_mnt("Auto resizing (sb) from %d LEBs to %d LEBs",
631 c->old_leb_cnt, c->leb_cnt);
632 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
633 err = ubifs_write_sb_node(c, sup);
636 c->old_leb_cnt = c->leb_cnt;
640 c->log_bytes = (long long)c->log_lebs * c->leb_size;
641 c->log_last = UBIFS_LOG_LNUM + c->log_lebs - 1;
642 c->lpt_first = UBIFS_LOG_LNUM + c->log_lebs;
643 c->lpt_last = c->lpt_first + c->lpt_lebs - 1;
644 c->orph_first = c->lpt_last + 1;
645 c->orph_last = c->orph_first + c->orph_lebs - 1;
646 c->main_lebs = c->leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS;
647 c->main_lebs -= c->log_lebs + c->lpt_lebs + c->orph_lebs;
648 c->main_first = c->leb_cnt - c->main_lebs;
650 err = validate_sb(c, sup);