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;
86 /* Some functions called from here depend on the @c->key_len filed */
87 c->key_len = UBIFS_SK_LEN;
90 * First of all, we have to calculate default file-system geometry -
91 * log size, journal size, etc.
93 if (c->leb_cnt < 0x7FFFFFFF / DEFAULT_JNL_PERCENT)
94 /* We can first multiply then divide and have no overflow */
95 jnl_lebs = c->leb_cnt * DEFAULT_JNL_PERCENT / 100;
97 jnl_lebs = (c->leb_cnt / 100) * DEFAULT_JNL_PERCENT;
99 if (jnl_lebs < UBIFS_MIN_JNL_LEBS)
100 jnl_lebs = UBIFS_MIN_JNL_LEBS;
101 if (jnl_lebs * c->leb_size > DEFAULT_MAX_JNL)
102 jnl_lebs = DEFAULT_MAX_JNL / c->leb_size;
105 * The log should be large enough to fit reference nodes for all bud
106 * LEBs. Because buds do not have to start from the beginning of LEBs
107 * (half of the LEB may contain committed data), the log should
108 * generally be larger, make it twice as large.
110 tmp = 2 * (c->ref_node_alsz * jnl_lebs) + c->leb_size - 1;
111 log_lebs = tmp / c->leb_size;
112 /* Plus one LEB reserved for commit */
114 if (c->leb_cnt - min_leb_cnt > 8) {
115 /* And some extra space to allow writes while committing */
120 max_buds = jnl_lebs - log_lebs;
121 if (max_buds < UBIFS_MIN_BUD_LEBS)
122 max_buds = UBIFS_MIN_BUD_LEBS;
125 * Orphan nodes are stored in a separate area. One node can store a lot
126 * of orphan inode numbers, but when new orphan comes we just add a new
127 * orphan node. At some point the nodes are consolidated into one
130 orph_lebs = UBIFS_MIN_ORPH_LEBS;
131 #ifdef CONFIG_UBIFS_FS_DEBUG
132 if (c->leb_cnt - min_leb_cnt > 1)
134 * For debugging purposes it is better to have at least 2
135 * orphan LEBs, because the orphan subsystem would need to do
136 * consolidations and would be stressed more.
141 main_lebs = c->leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS - log_lebs;
142 main_lebs -= orph_lebs;
144 lpt_first = UBIFS_LOG_LNUM + log_lebs;
145 c->lsave_cnt = DEFAULT_LSAVE_CNT;
146 c->max_leb_cnt = c->leb_cnt;
147 err = ubifs_create_dflt_lpt(c, &main_lebs, lpt_first, &lpt_lebs,
152 dbg_gen("LEB Properties Tree created (LEBs %d-%d)", lpt_first,
153 lpt_first + lpt_lebs - 1);
155 main_first = c->leb_cnt - main_lebs;
157 /* Create default superblock */
158 tmp = ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size);
159 sup = kzalloc(tmp, GFP_KERNEL);
163 tmp64 = (uint64_t)max_buds * c->leb_size;
165 sup_flags |= UBIFS_FLG_BIGLPT;
167 sup->ch.node_type = UBIFS_SB_NODE;
168 sup->key_hash = UBIFS_KEY_HASH_R5;
169 sup->flags = cpu_to_le32(sup_flags);
170 sup->min_io_size = cpu_to_le32(c->min_io_size);
171 sup->leb_size = cpu_to_le32(c->leb_size);
172 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
173 sup->max_leb_cnt = cpu_to_le32(c->max_leb_cnt);
174 sup->max_bud_bytes = cpu_to_le64(tmp64);
175 sup->log_lebs = cpu_to_le32(log_lebs);
176 sup->lpt_lebs = cpu_to_le32(lpt_lebs);
177 sup->orph_lebs = cpu_to_le32(orph_lebs);
178 sup->jhead_cnt = cpu_to_le32(DEFAULT_JHEADS_CNT);
179 sup->fanout = cpu_to_le32(DEFAULT_FANOUT);
180 sup->lsave_cnt = cpu_to_le32(c->lsave_cnt);
181 sup->fmt_version = cpu_to_le32(UBIFS_FORMAT_VERSION);
182 sup->default_compr = cpu_to_le16(UBIFS_COMPR_LZO);
183 sup->time_gran = cpu_to_le32(DEFAULT_TIME_GRAN);
185 generate_random_uuid(sup->uuid);
187 main_bytes = (uint64_t)main_lebs * c->leb_size;
188 tmp64 = main_bytes * DEFAULT_RP_PERCENT;
190 if (tmp64 > DEFAULT_MAX_RP_SIZE)
191 tmp64 = DEFAULT_MAX_RP_SIZE;
192 sup->rp_size = cpu_to_le64(tmp64);
194 err = ubifs_write_node(c, sup, UBIFS_SB_NODE_SZ, 0, 0, UBI_LONGTERM);
199 dbg_gen("default superblock created at LEB 0:0");
201 /* Create default master node */
202 mst = kzalloc(c->mst_node_alsz, GFP_KERNEL);
206 mst->ch.node_type = UBIFS_MST_NODE;
207 mst->log_lnum = cpu_to_le32(UBIFS_LOG_LNUM);
208 mst->highest_inum = cpu_to_le64(UBIFS_FIRST_INO);
210 mst->root_lnum = cpu_to_le32(main_first + DEFAULT_IDX_LEB);
212 tmp = ubifs_idx_node_sz(c, 1);
213 mst->root_len = cpu_to_le32(tmp);
214 mst->gc_lnum = cpu_to_le32(main_first + DEFAULT_GC_LEB);
215 mst->ihead_lnum = cpu_to_le32(main_first + DEFAULT_IDX_LEB);
216 mst->ihead_offs = cpu_to_le32(ALIGN(tmp, c->min_io_size));
217 mst->index_size = cpu_to_le64(ALIGN(tmp, 8));
218 mst->lpt_lnum = cpu_to_le32(c->lpt_lnum);
219 mst->lpt_offs = cpu_to_le32(c->lpt_offs);
220 mst->nhead_lnum = cpu_to_le32(c->nhead_lnum);
221 mst->nhead_offs = cpu_to_le32(c->nhead_offs);
222 mst->ltab_lnum = cpu_to_le32(c->ltab_lnum);
223 mst->ltab_offs = cpu_to_le32(c->ltab_offs);
224 mst->lsave_lnum = cpu_to_le32(c->lsave_lnum);
225 mst->lsave_offs = cpu_to_le32(c->lsave_offs);
226 mst->lscan_lnum = cpu_to_le32(main_first);
227 mst->empty_lebs = cpu_to_le32(main_lebs - 2);
228 mst->idx_lebs = cpu_to_le32(1);
229 mst->leb_cnt = cpu_to_le32(c->leb_cnt);
231 /* Calculate lprops statistics */
233 tmp64 -= ALIGN(ubifs_idx_node_sz(c, 1), c->min_io_size);
234 tmp64 -= ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size);
235 mst->total_free = cpu_to_le64(tmp64);
237 tmp64 = ALIGN(ubifs_idx_node_sz(c, 1), c->min_io_size);
238 ino_waste = ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size) -
241 tmp64 -= ALIGN(ubifs_idx_node_sz(c, 1), 8);
242 mst->total_dirty = cpu_to_le64(tmp64);
244 /* The indexing LEB does not contribute to dark space */
245 tmp64 = (c->main_lebs - 1) * c->dark_wm;
246 mst->total_dark = cpu_to_le64(tmp64);
248 mst->total_used = cpu_to_le64(UBIFS_INO_NODE_SZ);
250 err = ubifs_write_node(c, mst, UBIFS_MST_NODE_SZ, UBIFS_MST_LNUM, 0,
256 err = ubifs_write_node(c, mst, UBIFS_MST_NODE_SZ, UBIFS_MST_LNUM + 1, 0,
262 dbg_gen("default master node created at LEB %d:0", UBIFS_MST_LNUM);
264 /* Create the root indexing node */
265 tmp = ubifs_idx_node_sz(c, 1);
266 idx = kzalloc(ALIGN(tmp, c->min_io_size), GFP_KERNEL);
270 c->key_fmt = UBIFS_SIMPLE_KEY_FMT;
271 c->key_hash = key_r5_hash;
273 idx->ch.node_type = UBIFS_IDX_NODE;
274 idx->child_cnt = cpu_to_le16(1);
275 ino_key_init(c, &key, UBIFS_ROOT_INO);
276 br = ubifs_idx_branch(c, idx, 0);
277 key_write_idx(c, &key, &br->key);
278 br->lnum = cpu_to_le32(main_first + DEFAULT_DATA_LEB);
279 br->len = cpu_to_le32(UBIFS_INO_NODE_SZ);
280 err = ubifs_write_node(c, idx, tmp, main_first + DEFAULT_IDX_LEB, 0,
286 dbg_gen("default root indexing node created LEB %d:0",
287 main_first + DEFAULT_IDX_LEB);
289 /* Create default root inode */
290 tmp = ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size);
291 ino = kzalloc(tmp, GFP_KERNEL);
295 ino_key_init_flash(c, &ino->key, UBIFS_ROOT_INO);
296 ino->ch.node_type = UBIFS_INO_NODE;
297 ino->creat_sqnum = cpu_to_le64(++c->max_sqnum);
298 ino->nlink = cpu_to_le32(2);
299 tmp_le64 = cpu_to_le64(CURRENT_TIME_SEC.tv_sec);
300 ino->atime_sec = tmp_le64;
301 ino->ctime_sec = tmp_le64;
302 ino->mtime_sec = tmp_le64;
306 ino->mode = cpu_to_le32(S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO);
307 ino->size = cpu_to_le64(UBIFS_INO_NODE_SZ);
309 /* Set compression enabled by default */
310 ino->flags = cpu_to_le32(UBIFS_COMPR_FL);
312 err = ubifs_write_node(c, ino, UBIFS_INO_NODE_SZ,
313 main_first + DEFAULT_DATA_LEB, 0,
319 dbg_gen("root inode created at LEB %d:0",
320 main_first + DEFAULT_DATA_LEB);
323 * The first node in the log has to be the commit start node. This is
324 * always the case during normal file-system operation. Write a fake
325 * commit start node to the log.
327 tmp = ALIGN(UBIFS_CS_NODE_SZ, c->min_io_size);
328 cs = kzalloc(tmp, GFP_KERNEL);
332 cs->ch.node_type = UBIFS_CS_NODE;
333 err = ubifs_write_node(c, cs, UBIFS_CS_NODE_SZ, UBIFS_LOG_LNUM,
337 ubifs_msg("default file-system created");
342 * validate_sb - validate superblock node.
343 * @c: UBIFS file-system description object
344 * @sup: superblock node
346 * This function validates superblock node @sup. Since most of data was read
347 * from the superblock and stored in @c, the function validates fields in @c
348 * instead. Returns zero in case of success and %-EINVAL in case of validation
351 static int validate_sb(struct ubifs_info *c, struct ubifs_sb_node *sup)
354 int err = 1, min_leb_cnt;
361 if (sup->key_fmt != UBIFS_SIMPLE_KEY_FMT) {
366 if (le32_to_cpu(sup->min_io_size) != c->min_io_size) {
367 ubifs_err("min. I/O unit mismatch: %d in superblock, %d real",
368 le32_to_cpu(sup->min_io_size), c->min_io_size);
372 if (le32_to_cpu(sup->leb_size) != c->leb_size) {
373 ubifs_err("LEB size mismatch: %d in superblock, %d real",
374 le32_to_cpu(sup->leb_size), c->leb_size);
378 if (c->log_lebs < UBIFS_MIN_LOG_LEBS ||
379 c->lpt_lebs < UBIFS_MIN_LPT_LEBS ||
380 c->orph_lebs < UBIFS_MIN_ORPH_LEBS ||
381 c->main_lebs < UBIFS_MIN_MAIN_LEBS) {
387 * Calculate minimum allowed amount of main area LEBs. This is very
388 * similar to %UBIFS_MIN_LEB_CNT, but we take into account real what we
389 * have just read from the superblock.
391 min_leb_cnt = UBIFS_SB_LEBS + UBIFS_MST_LEBS + c->log_lebs;
392 min_leb_cnt += c->lpt_lebs + c->orph_lebs + c->jhead_cnt + 6;
394 if (c->leb_cnt < min_leb_cnt || c->leb_cnt > c->vi.size) {
395 ubifs_err("bad LEB count: %d in superblock, %d on UBI volume, "
396 "%d minimum required", c->leb_cnt, c->vi.size,
401 if (c->max_leb_cnt < c->leb_cnt) {
402 ubifs_err("max. LEB count %d less than LEB count %d",
403 c->max_leb_cnt, c->leb_cnt);
407 if (c->main_lebs < UBIFS_MIN_MAIN_LEBS) {
412 if (c->max_bud_bytes < (long long)c->leb_size * UBIFS_MIN_BUD_LEBS ||
413 c->max_bud_bytes > (long long)c->leb_size * c->main_lebs) {
418 if (c->jhead_cnt < NONDATA_JHEADS_CNT + 1 ||
419 c->jhead_cnt > NONDATA_JHEADS_CNT + UBIFS_MAX_JHEADS) {
424 if (c->fanout < UBIFS_MIN_FANOUT ||
425 ubifs_idx_node_sz(c, c->fanout) > c->leb_size) {
430 if (c->lsave_cnt < 0 || (c->lsave_cnt > DEFAULT_LSAVE_CNT &&
431 c->lsave_cnt > c->max_leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS -
432 c->log_lebs - c->lpt_lebs - c->orph_lebs)) {
437 if (UBIFS_SB_LEBS + UBIFS_MST_LEBS + c->log_lebs + c->lpt_lebs +
438 c->orph_lebs + c->main_lebs != c->leb_cnt) {
443 if (c->default_compr < 0 || c->default_compr >= UBIFS_COMPR_TYPES_CNT) {
448 max_bytes = c->main_lebs * (long long)c->leb_size;
449 if (c->rp_size < 0 || max_bytes < c->rp_size) {
454 if (le32_to_cpu(sup->time_gran) > 1000000000 ||
455 le32_to_cpu(sup->time_gran) < 1) {
463 ubifs_err("bad superblock, error %d", err);
464 dbg_dump_node(c, sup);
469 * ubifs_read_sb_node - read superblock node.
470 * @c: UBIFS file-system description object
472 * This function returns a pointer to the superblock node or a negative error
475 struct ubifs_sb_node *ubifs_read_sb_node(struct ubifs_info *c)
477 struct ubifs_sb_node *sup;
480 sup = kmalloc(ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size), GFP_NOFS);
482 return ERR_PTR(-ENOMEM);
484 err = ubifs_read_node(c, sup, UBIFS_SB_NODE, UBIFS_SB_NODE_SZ,
495 * ubifs_write_sb_node - write superblock node.
496 * @c: UBIFS file-system description object
497 * @sup: superblock node read with 'ubifs_read_sb_node()'
499 * This function returns %0 on success and a negative error code on failure.
501 int ubifs_write_sb_node(struct ubifs_info *c, struct ubifs_sb_node *sup)
503 int len = ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size);
505 ubifs_prepare_node(c, sup, UBIFS_SB_NODE_SZ, 1);
506 return ubifs_leb_change(c, UBIFS_SB_LNUM, sup, len, UBI_LONGTERM);
510 * ubifs_read_superblock - read superblock.
511 * @c: UBIFS file-system description object
513 * This function finds, reads and checks the superblock. If an empty UBI volume
514 * is being mounted, this function creates default superblock. Returns zero in
515 * case of success, and a negative error code in case of failure.
517 int ubifs_read_superblock(struct ubifs_info *c)
520 struct ubifs_sb_node *sup;
523 err = create_default_filesystem(c);
528 sup = ubifs_read_sb_node(c);
533 * The software supports all previous versions but not future versions,
534 * due to the unavailability of time-travelling equipment.
536 c->fmt_version = le32_to_cpu(sup->fmt_version);
537 if (c->fmt_version > UBIFS_FORMAT_VERSION) {
538 ubifs_err("on-flash format version is %d, but software only "
539 "supports up to version %d", c->fmt_version,
540 UBIFS_FORMAT_VERSION);
545 if (c->fmt_version < 3) {
546 ubifs_err("on-flash format version %d is not supported",
552 switch (sup->key_hash) {
553 case UBIFS_KEY_HASH_R5:
554 c->key_hash = key_r5_hash;
555 c->key_hash_type = UBIFS_KEY_HASH_R5;
558 case UBIFS_KEY_HASH_TEST:
559 c->key_hash = key_test_hash;
560 c->key_hash_type = UBIFS_KEY_HASH_TEST;
564 c->key_fmt = sup->key_fmt;
566 switch (c->key_fmt) {
567 case UBIFS_SIMPLE_KEY_FMT:
568 c->key_len = UBIFS_SK_LEN;
571 ubifs_err("unsupported key format");
576 c->leb_cnt = le32_to_cpu(sup->leb_cnt);
577 c->max_leb_cnt = le32_to_cpu(sup->max_leb_cnt);
578 c->max_bud_bytes = le64_to_cpu(sup->max_bud_bytes);
579 c->log_lebs = le32_to_cpu(sup->log_lebs);
580 c->lpt_lebs = le32_to_cpu(sup->lpt_lebs);
581 c->orph_lebs = le32_to_cpu(sup->orph_lebs);
582 c->jhead_cnt = le32_to_cpu(sup->jhead_cnt) + NONDATA_JHEADS_CNT;
583 c->fanout = le32_to_cpu(sup->fanout);
584 c->lsave_cnt = le32_to_cpu(sup->lsave_cnt);
585 c->default_compr = le16_to_cpu(sup->default_compr);
586 c->rp_size = le64_to_cpu(sup->rp_size);
587 c->rp_uid = le32_to_cpu(sup->rp_uid);
588 c->rp_gid = le32_to_cpu(sup->rp_gid);
589 sup_flags = le32_to_cpu(sup->flags);
591 c->vfs_sb->s_time_gran = le32_to_cpu(sup->time_gran);
593 memcpy(&c->uuid, &sup->uuid, 16);
595 c->big_lpt = !!(sup_flags & UBIFS_FLG_BIGLPT);
597 /* Automatically increase file system size to the maximum size */
598 c->old_leb_cnt = c->leb_cnt;
599 if (c->leb_cnt < c->vi.size && c->leb_cnt < c->max_leb_cnt) {
600 c->leb_cnt = min_t(int, c->max_leb_cnt, c->vi.size);
601 if (c->vfs_sb->s_flags & MS_RDONLY)
602 dbg_mnt("Auto resizing (ro) from %d LEBs to %d LEBs",
603 c->old_leb_cnt, c->leb_cnt);
605 dbg_mnt("Auto resizing (sb) from %d LEBs to %d LEBs",
606 c->old_leb_cnt, c->leb_cnt);
607 sup->leb_cnt = cpu_to_le32(c->leb_cnt);
608 err = ubifs_write_sb_node(c, sup);
611 c->old_leb_cnt = c->leb_cnt;
615 c->log_bytes = (long long)c->log_lebs * c->leb_size;
616 c->log_last = UBIFS_LOG_LNUM + c->log_lebs - 1;
617 c->lpt_first = UBIFS_LOG_LNUM + c->log_lebs;
618 c->lpt_last = c->lpt_first + c->lpt_lebs - 1;
619 c->orph_first = c->lpt_last + 1;
620 c->orph_last = c->orph_first + c->orph_lebs - 1;
621 c->main_lebs = c->leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS;
622 c->main_lebs -= c->log_lebs + c->lpt_lebs + c->orph_lebs;
623 c->main_first = c->leb_cnt - c->main_lebs;
624 c->report_rp_size = ubifs_reported_space(c, c->rp_size);
626 err = validate_sb(c, sup);