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>
33 * Default journal size in logical eraseblocks as a percent of total
36 #define DEFAULT_JNL_PERCENT 5
38 /* Default maximum journal size in bytes */
39 #define DEFAULT_MAX_JNL (32*1024*1024)
41 /* Default indexing tree fanout */
42 #define DEFAULT_FANOUT 8
44 /* Default number of data journal heads */
45 #define DEFAULT_JHEADS_CNT 1
47 /* Default positions of different LEBs in the main area */
48 #define DEFAULT_IDX_LEB 0
49 #define DEFAULT_DATA_LEB 1
50 #define DEFAULT_GC_LEB 2
52 /* Default number of LEB numbers in LPT's save table */
53 #define DEFAULT_LSAVE_CNT 256
55 /* Default reserved pool size as a percent of maximum free space */
56 #define DEFAULT_RP_PERCENT 5
58 /* The default maximum size of reserved pool in bytes */
59 #define DEFAULT_MAX_RP_SIZE (5*1024*1024)
61 /* Default time granularity in nanoseconds */
62 #define DEFAULT_TIME_GRAN 1000000000
65 * create_default_filesystem - format empty UBI volume.
66 * @c: UBIFS file-system description object
68 * This function creates default empty file-system. Returns zero in case of
69 * success and a negative error code in case of failure.
71 static int create_default_filesystem(struct ubifs_info *c)
73 struct ubifs_sb_node *sup;
74 struct ubifs_mst_node *mst;
75 struct ubifs_idx_node *idx;
76 struct ubifs_branch *br;
77 struct ubifs_ino_node *ino;
78 struct ubifs_cs_node *cs;
80 int err, tmp, jnl_lebs, log_lebs, max_buds, main_lebs, main_first;
81 int lpt_lebs, lpt_first, orph_lebs, big_lpt, ino_waste, sup_flags = 0;
82 int min_leb_cnt = UBIFS_MIN_LEB_CNT;
83 uint64_t tmp64, main_bytes;
85 /* Some functions called from here depend on the @c->key_len filed */
86 c->key_len = UBIFS_SK_LEN;
89 * First of all, we have to calculate default file-system geometry -
90 * log size, journal size, etc.
92 if (c->leb_cnt < 0x7FFFFFFF / DEFAULT_JNL_PERCENT)
93 /* We can first multiply then divide and have no overflow */
94 jnl_lebs = c->leb_cnt * DEFAULT_JNL_PERCENT / 100;
96 jnl_lebs = (c->leb_cnt / 100) * DEFAULT_JNL_PERCENT;
98 if (jnl_lebs < UBIFS_MIN_JNL_LEBS)
99 jnl_lebs = UBIFS_MIN_JNL_LEBS;
100 if (jnl_lebs * c->leb_size > DEFAULT_MAX_JNL)
101 jnl_lebs = DEFAULT_MAX_JNL / c->leb_size;
104 * The log should be large enough to fit reference nodes for all bud
105 * LEBs. Because buds do not have to start from the beginning of LEBs
106 * (half of the LEB may contain committed data), the log should
107 * generally be larger, make it twice as large.
109 tmp = 2 * (c->ref_node_alsz * jnl_lebs) + c->leb_size - 1;
110 log_lebs = tmp / c->leb_size;
111 /* Plus one LEB reserved for commit */
113 if (c->leb_cnt - min_leb_cnt > 8) {
114 /* And some extra space to allow writes while committing */
119 max_buds = jnl_lebs - log_lebs;
120 if (max_buds < UBIFS_MIN_BUD_LEBS)
121 max_buds = UBIFS_MIN_BUD_LEBS;
124 * Orphan nodes are stored in a separate area. One node can store a lot
125 * of orphan inode numbers, but when new orphan comes we just add a new
126 * orphan node. At some point the nodes are consolidated into one
129 orph_lebs = UBIFS_MIN_ORPH_LEBS;
130 #ifdef CONFIG_UBIFS_FS_DEBUG
131 if (c->leb_cnt - min_leb_cnt > 1)
133 * For debugging purposes it is better to have at least 2
134 * orphan LEBs, because the orphan subsystem would need to do
135 * consolidations and would be stressed more.
140 main_lebs = c->leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS - log_lebs;
141 main_lebs -= orph_lebs;
143 lpt_first = UBIFS_LOG_LNUM + log_lebs;
144 c->lsave_cnt = DEFAULT_LSAVE_CNT;
145 c->max_leb_cnt = c->leb_cnt;
146 err = ubifs_create_dflt_lpt(c, &main_lebs, lpt_first, &lpt_lebs,
151 dbg_gen("LEB Properties Tree created (LEBs %d-%d)", lpt_first,
152 lpt_first + lpt_lebs - 1);
154 main_first = c->leb_cnt - main_lebs;
156 /* Create default superblock */
157 tmp = ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size);
158 sup = kzalloc(tmp, GFP_KERNEL);
162 tmp64 = (uint64_t)max_buds * c->leb_size;
164 sup_flags |= UBIFS_FLG_BIGLPT;
166 sup->ch.node_type = UBIFS_SB_NODE;
167 sup->key_hash = UBIFS_KEY_HASH_R5;
168 sup->flags = cpu_to_le32(sup_flags);
169 sup->min_io_size = cpu_to_le32(c->min_io_size);
170 sup->leb_size = cpu_to_le32(c->leb_size);
171 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
172 sup->max_leb_cnt = cpu_to_le32(c->max_leb_cnt);
173 sup->max_bud_bytes = cpu_to_le64(tmp64);
174 sup->log_lebs = cpu_to_le32(log_lebs);
175 sup->lpt_lebs = cpu_to_le32(lpt_lebs);
176 sup->orph_lebs = cpu_to_le32(orph_lebs);
177 sup->jhead_cnt = cpu_to_le32(DEFAULT_JHEADS_CNT);
178 sup->fanout = cpu_to_le32(DEFAULT_FANOUT);
179 sup->lsave_cnt = cpu_to_le32(c->lsave_cnt);
180 sup->fmt_version = cpu_to_le32(UBIFS_FORMAT_VERSION);
181 sup->default_compr = cpu_to_le16(UBIFS_COMPR_LZO);
182 sup->time_gran = cpu_to_le32(DEFAULT_TIME_GRAN);
184 generate_random_uuid(sup->uuid);
186 main_bytes = (uint64_t)main_lebs * c->leb_size;
187 tmp64 = main_bytes * DEFAULT_RP_PERCENT;
189 if (tmp64 > DEFAULT_MAX_RP_SIZE)
190 tmp64 = DEFAULT_MAX_RP_SIZE;
191 sup->rp_size = cpu_to_le64(tmp64);
193 err = ubifs_write_node(c, sup, UBIFS_SB_NODE_SZ, 0, 0, UBI_LONGTERM);
198 dbg_gen("default superblock created at LEB 0:0");
200 /* Create default master node */
201 mst = kzalloc(c->mst_node_alsz, GFP_KERNEL);
205 mst->ch.node_type = UBIFS_MST_NODE;
206 mst->log_lnum = cpu_to_le32(UBIFS_LOG_LNUM);
207 mst->highest_inum = cpu_to_le64(UBIFS_FIRST_INO);
209 mst->root_lnum = cpu_to_le32(main_first + DEFAULT_IDX_LEB);
211 tmp = ubifs_idx_node_sz(c, 1);
212 mst->root_len = cpu_to_le32(tmp);
213 mst->gc_lnum = cpu_to_le32(main_first + DEFAULT_GC_LEB);
214 mst->ihead_lnum = cpu_to_le32(main_first + DEFAULT_IDX_LEB);
215 mst->ihead_offs = cpu_to_le32(ALIGN(tmp, c->min_io_size));
216 mst->index_size = cpu_to_le64(ALIGN(tmp, 8));
217 mst->lpt_lnum = cpu_to_le32(c->lpt_lnum);
218 mst->lpt_offs = cpu_to_le32(c->lpt_offs);
219 mst->nhead_lnum = cpu_to_le32(c->nhead_lnum);
220 mst->nhead_offs = cpu_to_le32(c->nhead_offs);
221 mst->ltab_lnum = cpu_to_le32(c->ltab_lnum);
222 mst->ltab_offs = cpu_to_le32(c->ltab_offs);
223 mst->lsave_lnum = cpu_to_le32(c->lsave_lnum);
224 mst->lsave_offs = cpu_to_le32(c->lsave_offs);
225 mst->lscan_lnum = cpu_to_le32(main_first);
226 mst->empty_lebs = cpu_to_le32(main_lebs - 2);
227 mst->idx_lebs = cpu_to_le32(1);
228 mst->leb_cnt = cpu_to_le32(c->leb_cnt);
230 /* Calculate lprops statistics */
232 tmp64 -= ALIGN(ubifs_idx_node_sz(c, 1), c->min_io_size);
233 tmp64 -= ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size);
234 mst->total_free = cpu_to_le64(tmp64);
236 tmp64 = ALIGN(ubifs_idx_node_sz(c, 1), c->min_io_size);
237 ino_waste = ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size) -
240 tmp64 -= ALIGN(ubifs_idx_node_sz(c, 1), 8);
241 mst->total_dirty = cpu_to_le64(tmp64);
243 /* The indexing LEB does not contribute to dark space */
244 tmp64 = (c->main_lebs - 1) * c->dark_wm;
245 mst->total_dark = cpu_to_le64(tmp64);
247 mst->total_used = cpu_to_le64(UBIFS_INO_NODE_SZ);
249 err = ubifs_write_node(c, mst, UBIFS_MST_NODE_SZ, UBIFS_MST_LNUM, 0,
255 err = ubifs_write_node(c, mst, UBIFS_MST_NODE_SZ, UBIFS_MST_LNUM + 1, 0,
261 dbg_gen("default master node created at LEB %d:0", UBIFS_MST_LNUM);
263 /* Create the root indexing node */
264 tmp = ubifs_idx_node_sz(c, 1);
265 idx = kzalloc(ALIGN(tmp, c->min_io_size), GFP_KERNEL);
269 c->key_fmt = UBIFS_SIMPLE_KEY_FMT;
270 c->key_hash = key_r5_hash;
272 idx->ch.node_type = UBIFS_IDX_NODE;
273 idx->child_cnt = cpu_to_le16(1);
274 ino_key_init(c, &key, UBIFS_ROOT_INO);
275 br = ubifs_idx_branch(c, idx, 0);
276 key_write_idx(c, &key, &br->key);
277 br->lnum = cpu_to_le32(main_first + DEFAULT_DATA_LEB);
278 br->len = cpu_to_le32(UBIFS_INO_NODE_SZ);
279 err = ubifs_write_node(c, idx, tmp, main_first + DEFAULT_IDX_LEB, 0,
285 dbg_gen("default root indexing node created LEB %d:0",
286 main_first + DEFAULT_IDX_LEB);
288 /* Create default root inode */
289 tmp = ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size);
290 ino = kzalloc(tmp, GFP_KERNEL);
294 ino_key_init_flash(c, &ino->key, UBIFS_ROOT_INO);
295 ino->ch.node_type = UBIFS_INO_NODE;
296 ino->creat_sqnum = cpu_to_le64(++c->max_sqnum);
297 ino->nlink = cpu_to_le32(2);
298 tmp = cpu_to_le64(CURRENT_TIME_SEC.tv_sec);
299 ino->atime_sec = tmp;
300 ino->ctime_sec = tmp;
301 ino->mtime_sec = tmp;
305 ino->mode = cpu_to_le32(S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO);
306 ino->size = cpu_to_le64(UBIFS_INO_NODE_SZ);
308 /* Set compression enabled by default */
309 ino->flags = cpu_to_le32(UBIFS_COMPR_FL);
311 err = ubifs_write_node(c, ino, UBIFS_INO_NODE_SZ,
312 main_first + DEFAULT_DATA_LEB, 0,
318 dbg_gen("root inode created at LEB %d:0",
319 main_first + DEFAULT_DATA_LEB);
322 * The first node in the log has to be the commit start node. This is
323 * always the case during normal file-system operation. Write a fake
324 * commit start node to the log.
326 tmp = ALIGN(UBIFS_CS_NODE_SZ, c->min_io_size);
327 cs = kzalloc(tmp, GFP_KERNEL);
331 cs->ch.node_type = UBIFS_CS_NODE;
332 err = ubifs_write_node(c, cs, UBIFS_CS_NODE_SZ, UBIFS_LOG_LNUM,
336 ubifs_msg("default file-system created");
341 * validate_sb - validate superblock node.
342 * @c: UBIFS file-system description object
343 * @sup: superblock node
345 * This function validates superblock node @sup. Since most of data was read
346 * from the superblock and stored in @c, the function validates fields in @c
347 * instead. Returns zero in case of success and %-EINVAL in case of validation
350 static int validate_sb(struct ubifs_info *c, struct ubifs_sb_node *sup)
353 int err = 1, min_leb_cnt;
360 if (sup->key_fmt != UBIFS_SIMPLE_KEY_FMT) {
365 if (le32_to_cpu(sup->min_io_size) != c->min_io_size) {
366 ubifs_err("min. I/O unit mismatch: %d in superblock, %d real",
367 le32_to_cpu(sup->min_io_size), c->min_io_size);
371 if (le32_to_cpu(sup->leb_size) != c->leb_size) {
372 ubifs_err("LEB size mismatch: %d in superblock, %d real",
373 le32_to_cpu(sup->leb_size), c->leb_size);
377 if (c->log_lebs < UBIFS_MIN_LOG_LEBS ||
378 c->lpt_lebs < UBIFS_MIN_LPT_LEBS ||
379 c->orph_lebs < UBIFS_MIN_ORPH_LEBS ||
380 c->main_lebs < UBIFS_MIN_MAIN_LEBS) {
386 * Calculate minimum allowed amount of main area LEBs. This is very
387 * similar to %UBIFS_MIN_LEB_CNT, but we take into account real what we
388 * have just read from the superblock.
390 min_leb_cnt = UBIFS_SB_LEBS + UBIFS_MST_LEBS + c->log_lebs;
391 min_leb_cnt += c->lpt_lebs + c->orph_lebs + c->jhead_cnt + 6;
393 if (c->leb_cnt < min_leb_cnt || c->leb_cnt > c->vi.size) {
394 ubifs_err("bad LEB count: %d in superblock, %d on UBI volume, "
395 "%d minimum required", c->leb_cnt, c->vi.size,
400 if (c->max_leb_cnt < c->leb_cnt) {
401 ubifs_err("max. LEB count %d less than LEB count %d",
402 c->max_leb_cnt, c->leb_cnt);
406 if (c->main_lebs < UBIFS_MIN_MAIN_LEBS) {
411 if (c->max_bud_bytes < (long long)c->leb_size * UBIFS_MIN_BUD_LEBS ||
412 c->max_bud_bytes > (long long)c->leb_size * c->main_lebs) {
417 if (c->jhead_cnt < NONDATA_JHEADS_CNT + 1 ||
418 c->jhead_cnt > NONDATA_JHEADS_CNT + UBIFS_MAX_JHEADS) {
423 if (c->fanout < UBIFS_MIN_FANOUT ||
424 ubifs_idx_node_sz(c, c->fanout) > c->leb_size) {
429 if (c->lsave_cnt < 0 || (c->lsave_cnt > DEFAULT_LSAVE_CNT &&
430 c->lsave_cnt > c->max_leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS -
431 c->log_lebs - c->lpt_lebs - c->orph_lebs)) {
436 if (UBIFS_SB_LEBS + UBIFS_MST_LEBS + c->log_lebs + c->lpt_lebs +
437 c->orph_lebs + c->main_lebs != c->leb_cnt) {
442 if (c->default_compr < 0 || c->default_compr >= UBIFS_COMPR_TYPES_CNT) {
447 max_bytes = c->main_lebs * (long long)c->leb_size;
448 if (c->rp_size < 0 || max_bytes < c->rp_size) {
453 if (le32_to_cpu(sup->time_gran) > 1000000000 ||
454 le32_to_cpu(sup->time_gran) < 1) {
462 ubifs_err("bad superblock, error %d", err);
463 dbg_dump_node(c, sup);
468 * ubifs_read_sb_node - read superblock node.
469 * @c: UBIFS file-system description object
471 * This function returns a pointer to the superblock node or a negative error
474 struct ubifs_sb_node *ubifs_read_sb_node(struct ubifs_info *c)
476 struct ubifs_sb_node *sup;
479 sup = kmalloc(ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size), GFP_NOFS);
481 return ERR_PTR(-ENOMEM);
483 err = ubifs_read_node(c, sup, UBIFS_SB_NODE, UBIFS_SB_NODE_SZ,
494 * ubifs_write_sb_node - write superblock node.
495 * @c: UBIFS file-system description object
496 * @sup: superblock node read with 'ubifs_read_sb_node()'
498 * This function returns %0 on success and a negative error code on failure.
500 int ubifs_write_sb_node(struct ubifs_info *c, struct ubifs_sb_node *sup)
502 int len = ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size);
504 ubifs_prepare_node(c, sup, UBIFS_SB_NODE_SZ, 1);
505 return ubifs_leb_change(c, UBIFS_SB_LNUM, sup, len, UBI_LONGTERM);
509 * ubifs_read_superblock - read superblock.
510 * @c: UBIFS file-system description object
512 * This function finds, reads and checks the superblock. If an empty UBI volume
513 * is being mounted, this function creates default superblock. Returns zero in
514 * case of success, and a negative error code in case of failure.
516 int ubifs_read_superblock(struct ubifs_info *c)
519 struct ubifs_sb_node *sup;
522 err = create_default_filesystem(c);
527 sup = ubifs_read_sb_node(c);
532 * The software supports all previous versions but not future versions,
533 * due to the unavailability of time-travelling equipment.
535 c->fmt_version = le32_to_cpu(sup->fmt_version);
536 if (c->fmt_version > UBIFS_FORMAT_VERSION) {
537 ubifs_err("on-flash format version is %d, but software only "
538 "supports up to version %d", c->fmt_version,
539 UBIFS_FORMAT_VERSION);
544 if (c->fmt_version < 3) {
545 ubifs_err("on-flash format version %d is not supported",
551 switch (sup->key_hash) {
552 case UBIFS_KEY_HASH_R5:
553 c->key_hash = key_r5_hash;
554 c->key_hash_type = UBIFS_KEY_HASH_R5;
557 case UBIFS_KEY_HASH_TEST:
558 c->key_hash = key_test_hash;
559 c->key_hash_type = UBIFS_KEY_HASH_TEST;
563 c->key_fmt = sup->key_fmt;
565 switch (c->key_fmt) {
566 case UBIFS_SIMPLE_KEY_FMT:
567 c->key_len = UBIFS_SK_LEN;
570 ubifs_err("unsupported key format");
575 c->leb_cnt = le32_to_cpu(sup->leb_cnt);
576 c->max_leb_cnt = le32_to_cpu(sup->max_leb_cnt);
577 c->max_bud_bytes = le64_to_cpu(sup->max_bud_bytes);
578 c->log_lebs = le32_to_cpu(sup->log_lebs);
579 c->lpt_lebs = le32_to_cpu(sup->lpt_lebs);
580 c->orph_lebs = le32_to_cpu(sup->orph_lebs);
581 c->jhead_cnt = le32_to_cpu(sup->jhead_cnt) + NONDATA_JHEADS_CNT;
582 c->fanout = le32_to_cpu(sup->fanout);
583 c->lsave_cnt = le32_to_cpu(sup->lsave_cnt);
584 c->default_compr = le16_to_cpu(sup->default_compr);
585 c->rp_size = le64_to_cpu(sup->rp_size);
586 c->rp_uid = le32_to_cpu(sup->rp_uid);
587 c->rp_gid = le32_to_cpu(sup->rp_gid);
588 sup_flags = le32_to_cpu(sup->flags);
590 c->vfs_sb->s_time_gran = le32_to_cpu(sup->time_gran);
592 memcpy(&c->uuid, &sup->uuid, 16);
594 c->big_lpt = !!(sup_flags & UBIFS_FLG_BIGLPT);
596 /* Automatically increase file system size to the maximum size */
597 c->old_leb_cnt = c->leb_cnt;
598 if (c->leb_cnt < c->vi.size && c->leb_cnt < c->max_leb_cnt) {
599 c->leb_cnt = min_t(int, c->max_leb_cnt, c->vi.size);
600 if (c->vfs_sb->s_flags & MS_RDONLY)
601 dbg_mnt("Auto resizing (ro) from %d LEBs to %d LEBs",
602 c->old_leb_cnt, c->leb_cnt);
604 dbg_mnt("Auto resizing (sb) from %d LEBs to %d LEBs",
605 c->old_leb_cnt, c->leb_cnt);
606 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
607 err = ubifs_write_sb_node(c, sup);
610 c->old_leb_cnt = c->leb_cnt;
614 c->log_bytes = (long long)c->log_lebs * c->leb_size;
615 c->log_last = UBIFS_LOG_LNUM + c->log_lebs - 1;
616 c->lpt_first = UBIFS_LOG_LNUM + c->log_lebs;
617 c->lpt_last = c->lpt_first + c->lpt_lebs - 1;
618 c->orph_first = c->lpt_last + 1;
619 c->orph_last = c->orph_first + c->orph_lebs - 1;
620 c->main_lebs = c->leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS;
621 c->main_lebs -= c->log_lebs + c->lpt_lebs + c->orph_lebs;
622 c->main_first = c->leb_cnt - c->main_lebs;
623 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
625 err = validate_sb(c, sup);