2 * the_nilfs.c - the_nilfs shared structure.
4 * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 * Written by Ryusuke Konishi <ryusuke@osrg.net>
24 #include <linux/buffer_head.h>
25 #include <linux/slab.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/crc32.h>
38 static int nilfs_valid_sb(struct nilfs_super_block *sbp);
40 void nilfs_set_last_segment(struct the_nilfs *nilfs,
41 sector_t start_blocknr, u64 seq, __u64 cno)
43 spin_lock(&nilfs->ns_last_segment_lock);
44 nilfs->ns_last_pseg = start_blocknr;
45 nilfs->ns_last_seq = seq;
46 nilfs->ns_last_cno = cno;
48 if (!nilfs_sb_dirty(nilfs)) {
49 if (nilfs->ns_prev_seq == nilfs->ns_last_seq)
52 set_nilfs_sb_dirty(nilfs);
54 nilfs->ns_prev_seq = nilfs->ns_last_seq;
57 spin_unlock(&nilfs->ns_last_segment_lock);
61 * alloc_nilfs - allocate a nilfs object
62 * @bdev: block device to which the_nilfs is related
64 * Return Value: On success, pointer to the_nilfs is returned.
65 * On error, NULL is returned.
67 struct the_nilfs *alloc_nilfs(struct block_device *bdev)
69 struct the_nilfs *nilfs;
71 nilfs = kzalloc(sizeof(*nilfs), GFP_KERNEL);
75 nilfs->ns_bdev = bdev;
76 atomic_set(&nilfs->ns_ndirtyblks, 0);
77 init_rwsem(&nilfs->ns_sem);
78 INIT_LIST_HEAD(&nilfs->ns_gc_inodes);
79 spin_lock_init(&nilfs->ns_last_segment_lock);
80 nilfs->ns_cptree = RB_ROOT;
81 spin_lock_init(&nilfs->ns_cptree_lock);
82 init_rwsem(&nilfs->ns_segctor_sem);
88 * destroy_nilfs - destroy nilfs object
89 * @nilfs: nilfs object to be released
91 void destroy_nilfs(struct the_nilfs *nilfs)
94 if (nilfs_init(nilfs)) {
95 brelse(nilfs->ns_sbh[0]);
96 brelse(nilfs->ns_sbh[1]);
101 static int nilfs_load_super_root(struct the_nilfs *nilfs,
102 struct super_block *sb, sector_t sr_block)
104 struct buffer_head *bh_sr;
105 struct nilfs_super_root *raw_sr;
106 struct nilfs_super_block **sbp = nilfs->ns_sbp;
107 struct nilfs_inode *rawi;
108 unsigned dat_entry_size, segment_usage_size, checkpoint_size;
112 err = nilfs_read_super_root_block(nilfs, sr_block, &bh_sr, 1);
116 down_read(&nilfs->ns_sem);
117 dat_entry_size = le16_to_cpu(sbp[0]->s_dat_entry_size);
118 checkpoint_size = le16_to_cpu(sbp[0]->s_checkpoint_size);
119 segment_usage_size = le16_to_cpu(sbp[0]->s_segment_usage_size);
120 up_read(&nilfs->ns_sem);
122 inode_size = nilfs->ns_inode_size;
124 rawi = (void *)bh_sr->b_data + NILFS_SR_DAT_OFFSET(inode_size);
125 err = nilfs_dat_read(sb, dat_entry_size, rawi, &nilfs->ns_dat);
129 rawi = (void *)bh_sr->b_data + NILFS_SR_CPFILE_OFFSET(inode_size);
130 err = nilfs_cpfile_read(sb, checkpoint_size, rawi, &nilfs->ns_cpfile);
134 rawi = (void *)bh_sr->b_data + NILFS_SR_SUFILE_OFFSET(inode_size);
135 err = nilfs_sufile_read(sb, segment_usage_size, rawi,
140 raw_sr = (struct nilfs_super_root *)bh_sr->b_data;
141 nilfs->ns_nongc_ctime = le64_to_cpu(raw_sr->sr_nongc_ctime);
148 iput(nilfs->ns_cpfile);
155 static void nilfs_init_recovery_info(struct nilfs_recovery_info *ri)
157 memset(ri, 0, sizeof(*ri));
158 INIT_LIST_HEAD(&ri->ri_used_segments);
161 static void nilfs_clear_recovery_info(struct nilfs_recovery_info *ri)
163 nilfs_dispose_segment_list(&ri->ri_used_segments);
167 * nilfs_store_log_cursor - load log cursor from a super block
168 * @nilfs: nilfs object
169 * @sbp: buffer storing super block to be read
171 * nilfs_store_log_cursor() reads the last position of the log
172 * containing a super root from a given super block, and initializes
173 * relevant information on the nilfs object preparatory for log
174 * scanning and recovery.
176 static int nilfs_store_log_cursor(struct the_nilfs *nilfs,
177 struct nilfs_super_block *sbp)
181 nilfs->ns_last_pseg = le64_to_cpu(sbp->s_last_pseg);
182 nilfs->ns_last_cno = le64_to_cpu(sbp->s_last_cno);
183 nilfs->ns_last_seq = le64_to_cpu(sbp->s_last_seq);
185 nilfs->ns_prev_seq = nilfs->ns_last_seq;
186 nilfs->ns_seg_seq = nilfs->ns_last_seq;
188 nilfs_get_segnum_of_block(nilfs, nilfs->ns_last_pseg);
189 nilfs->ns_cno = nilfs->ns_last_cno + 1;
190 if (nilfs->ns_segnum >= nilfs->ns_nsegments) {
191 printk(KERN_ERR "NILFS invalid last segment number.\n");
198 * load_nilfs - load and recover the nilfs
199 * @nilfs: the_nilfs structure to be released
200 * @sbi: nilfs_sb_info used to recover past segment
202 * load_nilfs() searches and load the latest super root,
203 * attaches the last segment, and does recovery if needed.
204 * The caller must call this exclusively for simultaneous mounts.
206 int load_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi)
208 struct nilfs_recovery_info ri;
209 unsigned int s_flags = sbi->s_super->s_flags;
210 int really_read_only = bdev_read_only(nilfs->ns_bdev);
211 int valid_fs = nilfs_valid_fs(nilfs);
215 printk(KERN_WARNING "NILFS warning: mounting unchecked fs\n");
216 if (s_flags & MS_RDONLY) {
217 printk(KERN_INFO "NILFS: INFO: recovery "
218 "required for readonly filesystem.\n");
219 printk(KERN_INFO "NILFS: write access will "
220 "be enabled during recovery.\n");
224 nilfs_init_recovery_info(&ri);
226 err = nilfs_search_super_root(nilfs, &ri);
228 struct nilfs_super_block **sbp = nilfs->ns_sbp;
234 if (!nilfs_valid_sb(sbp[1])) {
236 "NILFS warning: unable to fall back to spare"
241 "NILFS: try rollback from an earlier position\n");
244 * restore super block with its spare and reconfigure
245 * relevant states of the nilfs object.
247 memcpy(sbp[0], sbp[1], nilfs->ns_sbsize);
248 nilfs->ns_crc_seed = le32_to_cpu(sbp[0]->s_crc_seed);
249 nilfs->ns_sbwtime = le64_to_cpu(sbp[0]->s_wtime);
251 /* verify consistency between two super blocks */
252 blocksize = BLOCK_SIZE << le32_to_cpu(sbp[0]->s_log_block_size);
253 if (blocksize != nilfs->ns_blocksize) {
255 "NILFS warning: blocksize differs between "
256 "two super blocks (%d != %d)\n",
257 blocksize, nilfs->ns_blocksize);
261 err = nilfs_store_log_cursor(nilfs, sbp[0]);
265 /* drop clean flag to allow roll-forward and recovery */
266 nilfs->ns_mount_state &= ~NILFS_VALID_FS;
269 err = nilfs_search_super_root(nilfs, &ri);
274 err = nilfs_load_super_root(nilfs, sbi->s_super, ri.ri_super_root);
276 printk(KERN_ERR "NILFS: error loading super root.\n");
283 if (s_flags & MS_RDONLY) {
286 if (nilfs_test_opt(sbi, NORECOVERY)) {
287 printk(KERN_INFO "NILFS: norecovery option specified. "
288 "skipping roll-forward recovery\n");
291 features = le64_to_cpu(nilfs->ns_sbp[0]->s_feature_compat_ro) &
292 ~NILFS_FEATURE_COMPAT_RO_SUPP;
294 printk(KERN_ERR "NILFS: couldn't proceed with "
295 "recovery because of unsupported optional "
297 (unsigned long long)features);
301 if (really_read_only) {
302 printk(KERN_ERR "NILFS: write access "
303 "unavailable, cannot proceed.\n");
307 sbi->s_super->s_flags &= ~MS_RDONLY;
308 } else if (nilfs_test_opt(sbi, NORECOVERY)) {
309 printk(KERN_ERR "NILFS: recovery cancelled because norecovery "
310 "option was specified for a read/write mount\n");
315 err = nilfs_salvage_orphan_logs(nilfs, sbi, &ri);
319 down_write(&nilfs->ns_sem);
320 nilfs->ns_mount_state |= NILFS_VALID_FS; /* set "clean" flag */
321 err = nilfs_cleanup_super(sbi);
322 up_write(&nilfs->ns_sem);
325 printk(KERN_ERR "NILFS: failed to update super block. "
326 "recovery unfinished.\n");
329 printk(KERN_INFO "NILFS: recovery complete.\n");
332 nilfs_clear_recovery_info(&ri);
333 sbi->s_super->s_flags = s_flags;
337 printk(KERN_ERR "NILFS: error searching super root.\n");
341 iput(nilfs->ns_cpfile);
342 iput(nilfs->ns_sufile);
346 nilfs_clear_recovery_info(&ri);
347 sbi->s_super->s_flags = s_flags;
351 static unsigned long long nilfs_max_size(unsigned int blkbits)
353 unsigned int max_bits;
354 unsigned long long res = MAX_LFS_FILESIZE; /* page cache limit */
356 max_bits = blkbits + NILFS_BMAP_KEY_BIT; /* bmap size limit */
358 res = min_t(unsigned long long, res, (1ULL << max_bits) - 1);
362 static int nilfs_store_disk_layout(struct the_nilfs *nilfs,
363 struct nilfs_super_block *sbp)
365 if (le32_to_cpu(sbp->s_rev_level) < NILFS_MIN_SUPP_REV) {
366 printk(KERN_ERR "NILFS: unsupported revision "
367 "(superblock rev.=%d.%d, current rev.=%d.%d). "
368 "Please check the version of mkfs.nilfs.\n",
369 le32_to_cpu(sbp->s_rev_level),
370 le16_to_cpu(sbp->s_minor_rev_level),
371 NILFS_CURRENT_REV, NILFS_MINOR_REV);
374 nilfs->ns_sbsize = le16_to_cpu(sbp->s_bytes);
375 if (nilfs->ns_sbsize > BLOCK_SIZE)
378 nilfs->ns_inode_size = le16_to_cpu(sbp->s_inode_size);
379 nilfs->ns_first_ino = le32_to_cpu(sbp->s_first_ino);
381 nilfs->ns_blocks_per_segment = le32_to_cpu(sbp->s_blocks_per_segment);
382 if (nilfs->ns_blocks_per_segment < NILFS_SEG_MIN_BLOCKS) {
383 printk(KERN_ERR "NILFS: too short segment.\n");
387 nilfs->ns_first_data_block = le64_to_cpu(sbp->s_first_data_block);
388 nilfs->ns_nsegments = le64_to_cpu(sbp->s_nsegments);
389 nilfs->ns_r_segments_percentage =
390 le32_to_cpu(sbp->s_r_segments_percentage);
392 max_t(unsigned long, NILFS_MIN_NRSVSEGS,
393 DIV_ROUND_UP(nilfs->ns_nsegments *
394 nilfs->ns_r_segments_percentage, 100));
395 nilfs->ns_crc_seed = le32_to_cpu(sbp->s_crc_seed);
399 static int nilfs_valid_sb(struct nilfs_super_block *sbp)
401 static unsigned char sum[4];
402 const int sumoff = offsetof(struct nilfs_super_block, s_sum);
406 if (!sbp || le16_to_cpu(sbp->s_magic) != NILFS_SUPER_MAGIC)
408 bytes = le16_to_cpu(sbp->s_bytes);
409 if (bytes > BLOCK_SIZE)
411 crc = crc32_le(le32_to_cpu(sbp->s_crc_seed), (unsigned char *)sbp,
413 crc = crc32_le(crc, sum, 4);
414 crc = crc32_le(crc, (unsigned char *)sbp + sumoff + 4,
416 return crc == le32_to_cpu(sbp->s_sum);
419 static int nilfs_sb2_bad_offset(struct nilfs_super_block *sbp, u64 offset)
421 return offset < ((le64_to_cpu(sbp->s_nsegments) *
422 le32_to_cpu(sbp->s_blocks_per_segment)) <<
423 (le32_to_cpu(sbp->s_log_block_size) + 10));
426 static void nilfs_release_super_block(struct the_nilfs *nilfs)
430 for (i = 0; i < 2; i++) {
431 if (nilfs->ns_sbp[i]) {
432 brelse(nilfs->ns_sbh[i]);
433 nilfs->ns_sbh[i] = NULL;
434 nilfs->ns_sbp[i] = NULL;
439 void nilfs_fall_back_super_block(struct the_nilfs *nilfs)
441 brelse(nilfs->ns_sbh[0]);
442 nilfs->ns_sbh[0] = nilfs->ns_sbh[1];
443 nilfs->ns_sbp[0] = nilfs->ns_sbp[1];
444 nilfs->ns_sbh[1] = NULL;
445 nilfs->ns_sbp[1] = NULL;
448 void nilfs_swap_super_block(struct the_nilfs *nilfs)
450 struct buffer_head *tsbh = nilfs->ns_sbh[0];
451 struct nilfs_super_block *tsbp = nilfs->ns_sbp[0];
453 nilfs->ns_sbh[0] = nilfs->ns_sbh[1];
454 nilfs->ns_sbp[0] = nilfs->ns_sbp[1];
455 nilfs->ns_sbh[1] = tsbh;
456 nilfs->ns_sbp[1] = tsbp;
459 static int nilfs_load_super_block(struct the_nilfs *nilfs,
460 struct super_block *sb, int blocksize,
461 struct nilfs_super_block **sbpp)
463 struct nilfs_super_block **sbp = nilfs->ns_sbp;
464 struct buffer_head **sbh = nilfs->ns_sbh;
465 u64 sb2off = NILFS_SB2_OFFSET_BYTES(nilfs->ns_bdev->bd_inode->i_size);
466 int valid[2], swp = 0;
468 sbp[0] = nilfs_read_super_block(sb, NILFS_SB_OFFSET_BYTES, blocksize,
470 sbp[1] = nilfs_read_super_block(sb, sb2off, blocksize, &sbh[1]);
474 printk(KERN_ERR "NILFS: unable to read superblock\n");
478 "NILFS warning: unable to read primary superblock\n");
481 "NILFS warning: unable to read secondary superblock\n");
484 * Compare two super blocks and set 1 in swp if the secondary
485 * super block is valid and newer. Otherwise, set 0 in swp.
487 valid[0] = nilfs_valid_sb(sbp[0]);
488 valid[1] = nilfs_valid_sb(sbp[1]);
489 swp = valid[1] && (!valid[0] ||
490 le64_to_cpu(sbp[1]->s_last_cno) >
491 le64_to_cpu(sbp[0]->s_last_cno));
493 if (valid[swp] && nilfs_sb2_bad_offset(sbp[swp], sb2off)) {
500 nilfs_release_super_block(nilfs);
501 printk(KERN_ERR "NILFS: Can't find nilfs on dev %s.\n",
507 printk(KERN_WARNING "NILFS warning: broken superblock. "
508 "using spare superblock.\n");
510 nilfs_swap_super_block(nilfs);
512 nilfs->ns_sbwcount = 0;
513 nilfs->ns_sbwtime = le64_to_cpu(sbp[0]->s_wtime);
514 nilfs->ns_prot_seq = le64_to_cpu(sbp[valid[1] & !swp]->s_last_seq);
520 * init_nilfs - initialize a NILFS instance.
521 * @nilfs: the_nilfs structure
522 * @sbi: nilfs_sb_info
524 * @data: mount options
526 * init_nilfs() performs common initialization per block device (e.g.
527 * reading the super block, getting disk layout information, initializing
528 * shared fields in the_nilfs).
530 * Return Value: On success, 0 is returned. On error, a negative error
533 int init_nilfs(struct the_nilfs *nilfs, struct nilfs_sb_info *sbi, char *data)
535 struct super_block *sb = sbi->s_super;
536 struct nilfs_super_block *sbp;
540 down_write(&nilfs->ns_sem);
542 blocksize = sb_min_blocksize(sb, NILFS_MIN_BLOCK_SIZE);
544 printk(KERN_ERR "NILFS: unable to set blocksize\n");
548 err = nilfs_load_super_block(nilfs, sb, blocksize, &sbp);
552 err = nilfs_store_magic_and_option(sb, sbp, data);
556 err = nilfs_check_feature_compatibility(sb, sbp);
560 blocksize = BLOCK_SIZE << le32_to_cpu(sbp->s_log_block_size);
561 if (blocksize < NILFS_MIN_BLOCK_SIZE ||
562 blocksize > NILFS_MAX_BLOCK_SIZE) {
563 printk(KERN_ERR "NILFS: couldn't mount because of unsupported "
564 "filesystem blocksize %d\n", blocksize);
568 if (sb->s_blocksize != blocksize) {
569 int hw_blocksize = bdev_logical_block_size(sb->s_bdev);
571 if (blocksize < hw_blocksize) {
573 "NILFS: blocksize %d too small for device "
574 "(sector-size = %d).\n",
575 blocksize, hw_blocksize);
579 nilfs_release_super_block(nilfs);
580 sb_set_blocksize(sb, blocksize);
582 err = nilfs_load_super_block(nilfs, sb, blocksize, &sbp);
585 /* not failed_sbh; sbh is released automatically
586 when reloading fails. */
588 nilfs->ns_blocksize_bits = sb->s_blocksize_bits;
589 nilfs->ns_blocksize = blocksize;
591 err = nilfs_store_disk_layout(nilfs, sbp);
595 sb->s_maxbytes = nilfs_max_size(sb->s_blocksize_bits);
597 nilfs->ns_mount_state = le16_to_cpu(sbp->s_state);
599 err = nilfs_store_log_cursor(nilfs, sbp);
603 set_nilfs_init(nilfs);
606 up_write(&nilfs->ns_sem);
610 nilfs_release_super_block(nilfs);
614 int nilfs_discard_segments(struct the_nilfs *nilfs, __u64 *segnump,
617 sector_t seg_start, seg_end;
618 sector_t start = 0, nblocks = 0;
619 unsigned int sects_per_block;
623 sects_per_block = (1 << nilfs->ns_blocksize_bits) /
624 bdev_logical_block_size(nilfs->ns_bdev);
625 for (sn = segnump; sn < segnump + nsegs; sn++) {
626 nilfs_get_segment_range(nilfs, *sn, &seg_start, &seg_end);
630 nblocks = seg_end - seg_start + 1;
631 } else if (start + nblocks == seg_start) {
632 nblocks += seg_end - seg_start + 1;
634 ret = blkdev_issue_discard(nilfs->ns_bdev,
635 start * sects_per_block,
636 nblocks * sects_per_block,
644 ret = blkdev_issue_discard(nilfs->ns_bdev,
645 start * sects_per_block,
646 nblocks * sects_per_block,
651 int nilfs_count_free_blocks(struct the_nilfs *nilfs, sector_t *nblocks)
653 unsigned long ncleansegs;
655 down_read(&NILFS_MDT(nilfs->ns_dat)->mi_sem);
656 ncleansegs = nilfs_sufile_get_ncleansegs(nilfs->ns_sufile);
657 up_read(&NILFS_MDT(nilfs->ns_dat)->mi_sem);
658 *nblocks = (sector_t)ncleansegs * nilfs->ns_blocks_per_segment;
662 int nilfs_near_disk_full(struct the_nilfs *nilfs)
664 unsigned long ncleansegs, nincsegs;
666 ncleansegs = nilfs_sufile_get_ncleansegs(nilfs->ns_sufile);
667 nincsegs = atomic_read(&nilfs->ns_ndirtyblks) /
668 nilfs->ns_blocks_per_segment + 1;
670 return ncleansegs <= nilfs->ns_nrsvsegs + nincsegs;
673 struct nilfs_root *nilfs_lookup_root(struct the_nilfs *nilfs, __u64 cno)
676 struct nilfs_root *root;
678 spin_lock(&nilfs->ns_cptree_lock);
679 n = nilfs->ns_cptree.rb_node;
681 root = rb_entry(n, struct nilfs_root, rb_node);
683 if (cno < root->cno) {
685 } else if (cno > root->cno) {
688 atomic_inc(&root->count);
689 spin_unlock(&nilfs->ns_cptree_lock);
693 spin_unlock(&nilfs->ns_cptree_lock);
699 nilfs_find_or_create_root(struct the_nilfs *nilfs, __u64 cno)
701 struct rb_node **p, *parent;
702 struct nilfs_root *root, *new;
704 root = nilfs_lookup_root(nilfs, cno);
708 new = kmalloc(sizeof(*root), GFP_KERNEL);
712 spin_lock(&nilfs->ns_cptree_lock);
714 p = &nilfs->ns_cptree.rb_node;
719 root = rb_entry(parent, struct nilfs_root, rb_node);
721 if (cno < root->cno) {
723 } else if (cno > root->cno) {
726 atomic_inc(&root->count);
727 spin_unlock(&nilfs->ns_cptree_lock);
736 atomic_set(&new->count, 1);
737 atomic_set(&new->inodes_count, 0);
738 atomic_set(&new->blocks_count, 0);
740 rb_link_node(&new->rb_node, parent, p);
741 rb_insert_color(&new->rb_node, &nilfs->ns_cptree);
743 spin_unlock(&nilfs->ns_cptree_lock);
748 void nilfs_put_root(struct nilfs_root *root)
750 if (atomic_dec_and_test(&root->count)) {
751 struct the_nilfs *nilfs = root->nilfs;
753 spin_lock(&nilfs->ns_cptree_lock);
754 rb_erase(&root->rb_node, &nilfs->ns_cptree);
755 spin_unlock(&nilfs->ns_cptree_lock);