xref: /openbmc/linux/fs/nilfs2/the_nilfs.c (revision 9ffc93f2)
1 /*
2  * the_nilfs.c - the_nilfs shared structure.
3  *
4  * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
5  *
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.
10  *
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.
15  *
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
19  *
20  * Written by Ryusuke Konishi <ryusuke@osrg.net>
21  *
22  */
23 
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/random.h>
29 #include <linux/crc32.h>
30 #include "nilfs.h"
31 #include "segment.h"
32 #include "alloc.h"
33 #include "cpfile.h"
34 #include "sufile.h"
35 #include "dat.h"
36 #include "segbuf.h"
37 
38 
39 static int nilfs_valid_sb(struct nilfs_super_block *sbp);
40 
41 void nilfs_set_last_segment(struct the_nilfs *nilfs,
42 			    sector_t start_blocknr, u64 seq, __u64 cno)
43 {
44 	spin_lock(&nilfs->ns_last_segment_lock);
45 	nilfs->ns_last_pseg = start_blocknr;
46 	nilfs->ns_last_seq = seq;
47 	nilfs->ns_last_cno = cno;
48 
49 	if (!nilfs_sb_dirty(nilfs)) {
50 		if (nilfs->ns_prev_seq == nilfs->ns_last_seq)
51 			goto stay_cursor;
52 
53 		set_nilfs_sb_dirty(nilfs);
54 	}
55 	nilfs->ns_prev_seq = nilfs->ns_last_seq;
56 
57  stay_cursor:
58 	spin_unlock(&nilfs->ns_last_segment_lock);
59 }
60 
61 /**
62  * alloc_nilfs - allocate a nilfs object
63  * @bdev: block device to which the_nilfs is related
64  *
65  * Return Value: On success, pointer to the_nilfs is returned.
66  * On error, NULL is returned.
67  */
68 struct the_nilfs *alloc_nilfs(struct block_device *bdev)
69 {
70 	struct the_nilfs *nilfs;
71 
72 	nilfs = kzalloc(sizeof(*nilfs), GFP_KERNEL);
73 	if (!nilfs)
74 		return NULL;
75 
76 	nilfs->ns_bdev = bdev;
77 	atomic_set(&nilfs->ns_ndirtyblks, 0);
78 	init_rwsem(&nilfs->ns_sem);
79 	INIT_LIST_HEAD(&nilfs->ns_dirty_files);
80 	INIT_LIST_HEAD(&nilfs->ns_gc_inodes);
81 	spin_lock_init(&nilfs->ns_inode_lock);
82 	spin_lock_init(&nilfs->ns_next_gen_lock);
83 	spin_lock_init(&nilfs->ns_last_segment_lock);
84 	nilfs->ns_cptree = RB_ROOT;
85 	spin_lock_init(&nilfs->ns_cptree_lock);
86 	init_rwsem(&nilfs->ns_segctor_sem);
87 
88 	return nilfs;
89 }
90 
91 /**
92  * destroy_nilfs - destroy nilfs object
93  * @nilfs: nilfs object to be released
94  */
95 void destroy_nilfs(struct the_nilfs *nilfs)
96 {
97 	might_sleep();
98 	if (nilfs_init(nilfs)) {
99 		brelse(nilfs->ns_sbh[0]);
100 		brelse(nilfs->ns_sbh[1]);
101 	}
102 	kfree(nilfs);
103 }
104 
105 static int nilfs_load_super_root(struct the_nilfs *nilfs,
106 				 struct super_block *sb, sector_t sr_block)
107 {
108 	struct buffer_head *bh_sr;
109 	struct nilfs_super_root *raw_sr;
110 	struct nilfs_super_block **sbp = nilfs->ns_sbp;
111 	struct nilfs_inode *rawi;
112 	unsigned dat_entry_size, segment_usage_size, checkpoint_size;
113 	unsigned inode_size;
114 	int err;
115 
116 	err = nilfs_read_super_root_block(nilfs, sr_block, &bh_sr, 1);
117 	if (unlikely(err))
118 		return err;
119 
120 	down_read(&nilfs->ns_sem);
121 	dat_entry_size = le16_to_cpu(sbp[0]->s_dat_entry_size);
122 	checkpoint_size = le16_to_cpu(sbp[0]->s_checkpoint_size);
123 	segment_usage_size = le16_to_cpu(sbp[0]->s_segment_usage_size);
124 	up_read(&nilfs->ns_sem);
125 
126 	inode_size = nilfs->ns_inode_size;
127 
128 	rawi = (void *)bh_sr->b_data + NILFS_SR_DAT_OFFSET(inode_size);
129 	err = nilfs_dat_read(sb, dat_entry_size, rawi, &nilfs->ns_dat);
130 	if (err)
131 		goto failed;
132 
133 	rawi = (void *)bh_sr->b_data + NILFS_SR_CPFILE_OFFSET(inode_size);
134 	err = nilfs_cpfile_read(sb, checkpoint_size, rawi, &nilfs->ns_cpfile);
135 	if (err)
136 		goto failed_dat;
137 
138 	rawi = (void *)bh_sr->b_data + NILFS_SR_SUFILE_OFFSET(inode_size);
139 	err = nilfs_sufile_read(sb, segment_usage_size, rawi,
140 				&nilfs->ns_sufile);
141 	if (err)
142 		goto failed_cpfile;
143 
144 	raw_sr = (struct nilfs_super_root *)bh_sr->b_data;
145 	nilfs->ns_nongc_ctime = le64_to_cpu(raw_sr->sr_nongc_ctime);
146 
147  failed:
148 	brelse(bh_sr);
149 	return err;
150 
151  failed_cpfile:
152 	iput(nilfs->ns_cpfile);
153 
154  failed_dat:
155 	iput(nilfs->ns_dat);
156 	goto failed;
157 }
158 
159 static void nilfs_init_recovery_info(struct nilfs_recovery_info *ri)
160 {
161 	memset(ri, 0, sizeof(*ri));
162 	INIT_LIST_HEAD(&ri->ri_used_segments);
163 }
164 
165 static void nilfs_clear_recovery_info(struct nilfs_recovery_info *ri)
166 {
167 	nilfs_dispose_segment_list(&ri->ri_used_segments);
168 }
169 
170 /**
171  * nilfs_store_log_cursor - load log cursor from a super block
172  * @nilfs: nilfs object
173  * @sbp: buffer storing super block to be read
174  *
175  * nilfs_store_log_cursor() reads the last position of the log
176  * containing a super root from a given super block, and initializes
177  * relevant information on the nilfs object preparatory for log
178  * scanning and recovery.
179  */
180 static int nilfs_store_log_cursor(struct the_nilfs *nilfs,
181 				  struct nilfs_super_block *sbp)
182 {
183 	int ret = 0;
184 
185 	nilfs->ns_last_pseg = le64_to_cpu(sbp->s_last_pseg);
186 	nilfs->ns_last_cno = le64_to_cpu(sbp->s_last_cno);
187 	nilfs->ns_last_seq = le64_to_cpu(sbp->s_last_seq);
188 
189 	nilfs->ns_prev_seq = nilfs->ns_last_seq;
190 	nilfs->ns_seg_seq = nilfs->ns_last_seq;
191 	nilfs->ns_segnum =
192 		nilfs_get_segnum_of_block(nilfs, nilfs->ns_last_pseg);
193 	nilfs->ns_cno = nilfs->ns_last_cno + 1;
194 	if (nilfs->ns_segnum >= nilfs->ns_nsegments) {
195 		printk(KERN_ERR "NILFS invalid last segment number.\n");
196 		ret = -EINVAL;
197 	}
198 	return ret;
199 }
200 
201 /**
202  * load_nilfs - load and recover the nilfs
203  * @nilfs: the_nilfs structure to be released
204  * @sb: super block isntance used to recover past segment
205  *
206  * load_nilfs() searches and load the latest super root,
207  * attaches the last segment, and does recovery if needed.
208  * The caller must call this exclusively for simultaneous mounts.
209  */
210 int load_nilfs(struct the_nilfs *nilfs, struct super_block *sb)
211 {
212 	struct nilfs_recovery_info ri;
213 	unsigned int s_flags = sb->s_flags;
214 	int really_read_only = bdev_read_only(nilfs->ns_bdev);
215 	int valid_fs = nilfs_valid_fs(nilfs);
216 	int err;
217 
218 	if (!valid_fs) {
219 		printk(KERN_WARNING "NILFS warning: mounting unchecked fs\n");
220 		if (s_flags & MS_RDONLY) {
221 			printk(KERN_INFO "NILFS: INFO: recovery "
222 			       "required for readonly filesystem.\n");
223 			printk(KERN_INFO "NILFS: write access will "
224 			       "be enabled during recovery.\n");
225 		}
226 	}
227 
228 	nilfs_init_recovery_info(&ri);
229 
230 	err = nilfs_search_super_root(nilfs, &ri);
231 	if (unlikely(err)) {
232 		struct nilfs_super_block **sbp = nilfs->ns_sbp;
233 		int blocksize;
234 
235 		if (err != -EINVAL)
236 			goto scan_error;
237 
238 		if (!nilfs_valid_sb(sbp[1])) {
239 			printk(KERN_WARNING
240 			       "NILFS warning: unable to fall back to spare"
241 			       "super block\n");
242 			goto scan_error;
243 		}
244 		printk(KERN_INFO
245 		       "NILFS: try rollback from an earlier position\n");
246 
247 		/*
248 		 * restore super block with its spare and reconfigure
249 		 * relevant states of the nilfs object.
250 		 */
251 		memcpy(sbp[0], sbp[1], nilfs->ns_sbsize);
252 		nilfs->ns_crc_seed = le32_to_cpu(sbp[0]->s_crc_seed);
253 		nilfs->ns_sbwtime = le64_to_cpu(sbp[0]->s_wtime);
254 
255 		/* verify consistency between two super blocks */
256 		blocksize = BLOCK_SIZE << le32_to_cpu(sbp[0]->s_log_block_size);
257 		if (blocksize != nilfs->ns_blocksize) {
258 			printk(KERN_WARNING
259 			       "NILFS warning: blocksize differs between "
260 			       "two super blocks (%d != %d)\n",
261 			       blocksize, nilfs->ns_blocksize);
262 			goto scan_error;
263 		}
264 
265 		err = nilfs_store_log_cursor(nilfs, sbp[0]);
266 		if (err)
267 			goto scan_error;
268 
269 		/* drop clean flag to allow roll-forward and recovery */
270 		nilfs->ns_mount_state &= ~NILFS_VALID_FS;
271 		valid_fs = 0;
272 
273 		err = nilfs_search_super_root(nilfs, &ri);
274 		if (err)
275 			goto scan_error;
276 	}
277 
278 	err = nilfs_load_super_root(nilfs, sb, ri.ri_super_root);
279 	if (unlikely(err)) {
280 		printk(KERN_ERR "NILFS: error loading super root.\n");
281 		goto failed;
282 	}
283 
284 	if (valid_fs)
285 		goto skip_recovery;
286 
287 	if (s_flags & MS_RDONLY) {
288 		__u64 features;
289 
290 		if (nilfs_test_opt(nilfs, NORECOVERY)) {
291 			printk(KERN_INFO "NILFS: norecovery option specified. "
292 			       "skipping roll-forward recovery\n");
293 			goto skip_recovery;
294 		}
295 		features = le64_to_cpu(nilfs->ns_sbp[0]->s_feature_compat_ro) &
296 			~NILFS_FEATURE_COMPAT_RO_SUPP;
297 		if (features) {
298 			printk(KERN_ERR "NILFS: couldn't proceed with "
299 			       "recovery because of unsupported optional "
300 			       "features (%llx)\n",
301 			       (unsigned long long)features);
302 			err = -EROFS;
303 			goto failed_unload;
304 		}
305 		if (really_read_only) {
306 			printk(KERN_ERR "NILFS: write access "
307 			       "unavailable, cannot proceed.\n");
308 			err = -EROFS;
309 			goto failed_unload;
310 		}
311 		sb->s_flags &= ~MS_RDONLY;
312 	} else if (nilfs_test_opt(nilfs, NORECOVERY)) {
313 		printk(KERN_ERR "NILFS: recovery cancelled because norecovery "
314 		       "option was specified for a read/write mount\n");
315 		err = -EINVAL;
316 		goto failed_unload;
317 	}
318 
319 	err = nilfs_salvage_orphan_logs(nilfs, sb, &ri);
320 	if (err)
321 		goto failed_unload;
322 
323 	down_write(&nilfs->ns_sem);
324 	nilfs->ns_mount_state |= NILFS_VALID_FS; /* set "clean" flag */
325 	err = nilfs_cleanup_super(sb);
326 	up_write(&nilfs->ns_sem);
327 
328 	if (err) {
329 		printk(KERN_ERR "NILFS: failed to update super block. "
330 		       "recovery unfinished.\n");
331 		goto failed_unload;
332 	}
333 	printk(KERN_INFO "NILFS: recovery complete.\n");
334 
335  skip_recovery:
336 	nilfs_clear_recovery_info(&ri);
337 	sb->s_flags = s_flags;
338 	return 0;
339 
340  scan_error:
341 	printk(KERN_ERR "NILFS: error searching super root.\n");
342 	goto failed;
343 
344  failed_unload:
345 	iput(nilfs->ns_cpfile);
346 	iput(nilfs->ns_sufile);
347 	iput(nilfs->ns_dat);
348 
349  failed:
350 	nilfs_clear_recovery_info(&ri);
351 	sb->s_flags = s_flags;
352 	return err;
353 }
354 
355 static unsigned long long nilfs_max_size(unsigned int blkbits)
356 {
357 	unsigned int max_bits;
358 	unsigned long long res = MAX_LFS_FILESIZE; /* page cache limit */
359 
360 	max_bits = blkbits + NILFS_BMAP_KEY_BIT; /* bmap size limit */
361 	if (max_bits < 64)
362 		res = min_t(unsigned long long, res, (1ULL << max_bits) - 1);
363 	return res;
364 }
365 
366 /**
367  * nilfs_nrsvsegs - calculate the number of reserved segments
368  * @nilfs: nilfs object
369  * @nsegs: total number of segments
370  */
371 unsigned long nilfs_nrsvsegs(struct the_nilfs *nilfs, unsigned long nsegs)
372 {
373 	return max_t(unsigned long, NILFS_MIN_NRSVSEGS,
374 		     DIV_ROUND_UP(nsegs * nilfs->ns_r_segments_percentage,
375 				  100));
376 }
377 
378 void nilfs_set_nsegments(struct the_nilfs *nilfs, unsigned long nsegs)
379 {
380 	nilfs->ns_nsegments = nsegs;
381 	nilfs->ns_nrsvsegs = nilfs_nrsvsegs(nilfs, nsegs);
382 }
383 
384 static int nilfs_store_disk_layout(struct the_nilfs *nilfs,
385 				   struct nilfs_super_block *sbp)
386 {
387 	if (le32_to_cpu(sbp->s_rev_level) < NILFS_MIN_SUPP_REV) {
388 		printk(KERN_ERR "NILFS: unsupported revision "
389 		       "(superblock rev.=%d.%d, current rev.=%d.%d). "
390 		       "Please check the version of mkfs.nilfs.\n",
391 		       le32_to_cpu(sbp->s_rev_level),
392 		       le16_to_cpu(sbp->s_minor_rev_level),
393 		       NILFS_CURRENT_REV, NILFS_MINOR_REV);
394 		return -EINVAL;
395 	}
396 	nilfs->ns_sbsize = le16_to_cpu(sbp->s_bytes);
397 	if (nilfs->ns_sbsize > BLOCK_SIZE)
398 		return -EINVAL;
399 
400 	nilfs->ns_inode_size = le16_to_cpu(sbp->s_inode_size);
401 	nilfs->ns_first_ino = le32_to_cpu(sbp->s_first_ino);
402 
403 	nilfs->ns_blocks_per_segment = le32_to_cpu(sbp->s_blocks_per_segment);
404 	if (nilfs->ns_blocks_per_segment < NILFS_SEG_MIN_BLOCKS) {
405 		printk(KERN_ERR "NILFS: too short segment.\n");
406 		return -EINVAL;
407 	}
408 
409 	nilfs->ns_first_data_block = le64_to_cpu(sbp->s_first_data_block);
410 	nilfs->ns_r_segments_percentage =
411 		le32_to_cpu(sbp->s_r_segments_percentage);
412 	if (nilfs->ns_r_segments_percentage < 1 ||
413 	    nilfs->ns_r_segments_percentage > 99) {
414 		printk(KERN_ERR "NILFS: invalid reserved segments percentage.\n");
415 		return -EINVAL;
416 	}
417 
418 	nilfs_set_nsegments(nilfs, le64_to_cpu(sbp->s_nsegments));
419 	nilfs->ns_crc_seed = le32_to_cpu(sbp->s_crc_seed);
420 	return 0;
421 }
422 
423 static int nilfs_valid_sb(struct nilfs_super_block *sbp)
424 {
425 	static unsigned char sum[4];
426 	const int sumoff = offsetof(struct nilfs_super_block, s_sum);
427 	size_t bytes;
428 	u32 crc;
429 
430 	if (!sbp || le16_to_cpu(sbp->s_magic) != NILFS_SUPER_MAGIC)
431 		return 0;
432 	bytes = le16_to_cpu(sbp->s_bytes);
433 	if (bytes > BLOCK_SIZE)
434 		return 0;
435 	crc = crc32_le(le32_to_cpu(sbp->s_crc_seed), (unsigned char *)sbp,
436 		       sumoff);
437 	crc = crc32_le(crc, sum, 4);
438 	crc = crc32_le(crc, (unsigned char *)sbp + sumoff + 4,
439 		       bytes - sumoff - 4);
440 	return crc == le32_to_cpu(sbp->s_sum);
441 }
442 
443 static int nilfs_sb2_bad_offset(struct nilfs_super_block *sbp, u64 offset)
444 {
445 	return offset < ((le64_to_cpu(sbp->s_nsegments) *
446 			  le32_to_cpu(sbp->s_blocks_per_segment)) <<
447 			 (le32_to_cpu(sbp->s_log_block_size) + 10));
448 }
449 
450 static void nilfs_release_super_block(struct the_nilfs *nilfs)
451 {
452 	int i;
453 
454 	for (i = 0; i < 2; i++) {
455 		if (nilfs->ns_sbp[i]) {
456 			brelse(nilfs->ns_sbh[i]);
457 			nilfs->ns_sbh[i] = NULL;
458 			nilfs->ns_sbp[i] = NULL;
459 		}
460 	}
461 }
462 
463 void nilfs_fall_back_super_block(struct the_nilfs *nilfs)
464 {
465 	brelse(nilfs->ns_sbh[0]);
466 	nilfs->ns_sbh[0] = nilfs->ns_sbh[1];
467 	nilfs->ns_sbp[0] = nilfs->ns_sbp[1];
468 	nilfs->ns_sbh[1] = NULL;
469 	nilfs->ns_sbp[1] = NULL;
470 }
471 
472 void nilfs_swap_super_block(struct the_nilfs *nilfs)
473 {
474 	struct buffer_head *tsbh = nilfs->ns_sbh[0];
475 	struct nilfs_super_block *tsbp = nilfs->ns_sbp[0];
476 
477 	nilfs->ns_sbh[0] = nilfs->ns_sbh[1];
478 	nilfs->ns_sbp[0] = nilfs->ns_sbp[1];
479 	nilfs->ns_sbh[1] = tsbh;
480 	nilfs->ns_sbp[1] = tsbp;
481 }
482 
483 static int nilfs_load_super_block(struct the_nilfs *nilfs,
484 				  struct super_block *sb, int blocksize,
485 				  struct nilfs_super_block **sbpp)
486 {
487 	struct nilfs_super_block **sbp = nilfs->ns_sbp;
488 	struct buffer_head **sbh = nilfs->ns_sbh;
489 	u64 sb2off = NILFS_SB2_OFFSET_BYTES(nilfs->ns_bdev->bd_inode->i_size);
490 	int valid[2], swp = 0;
491 
492 	sbp[0] = nilfs_read_super_block(sb, NILFS_SB_OFFSET_BYTES, blocksize,
493 					&sbh[0]);
494 	sbp[1] = nilfs_read_super_block(sb, sb2off, blocksize, &sbh[1]);
495 
496 	if (!sbp[0]) {
497 		if (!sbp[1]) {
498 			printk(KERN_ERR "NILFS: unable to read superblock\n");
499 			return -EIO;
500 		}
501 		printk(KERN_WARNING
502 		       "NILFS warning: unable to read primary superblock "
503 		       "(blocksize = %d)\n", blocksize);
504 	} else if (!sbp[1]) {
505 		printk(KERN_WARNING
506 		       "NILFS warning: unable to read secondary superblock "
507 		       "(blocksize = %d)\n", blocksize);
508 	}
509 
510 	/*
511 	 * Compare two super blocks and set 1 in swp if the secondary
512 	 * super block is valid and newer.  Otherwise, set 0 in swp.
513 	 */
514 	valid[0] = nilfs_valid_sb(sbp[0]);
515 	valid[1] = nilfs_valid_sb(sbp[1]);
516 	swp = valid[1] && (!valid[0] ||
517 			   le64_to_cpu(sbp[1]->s_last_cno) >
518 			   le64_to_cpu(sbp[0]->s_last_cno));
519 
520 	if (valid[swp] && nilfs_sb2_bad_offset(sbp[swp], sb2off)) {
521 		brelse(sbh[1]);
522 		sbh[1] = NULL;
523 		sbp[1] = NULL;
524 		valid[1] = 0;
525 		swp = 0;
526 	}
527 	if (!valid[swp]) {
528 		nilfs_release_super_block(nilfs);
529 		printk(KERN_ERR "NILFS: Can't find nilfs on dev %s.\n",
530 		       sb->s_id);
531 		return -EINVAL;
532 	}
533 
534 	if (!valid[!swp])
535 		printk(KERN_WARNING "NILFS warning: broken superblock. "
536 		       "using spare superblock (blocksize = %d).\n", blocksize);
537 	if (swp)
538 		nilfs_swap_super_block(nilfs);
539 
540 	nilfs->ns_sbwcount = 0;
541 	nilfs->ns_sbwtime = le64_to_cpu(sbp[0]->s_wtime);
542 	nilfs->ns_prot_seq = le64_to_cpu(sbp[valid[1] & !swp]->s_last_seq);
543 	*sbpp = sbp[0];
544 	return 0;
545 }
546 
547 /**
548  * init_nilfs - initialize a NILFS instance.
549  * @nilfs: the_nilfs structure
550  * @sb: super block
551  * @data: mount options
552  *
553  * init_nilfs() performs common initialization per block device (e.g.
554  * reading the super block, getting disk layout information, initializing
555  * shared fields in the_nilfs).
556  *
557  * Return Value: On success, 0 is returned. On error, a negative error
558  * code is returned.
559  */
560 int init_nilfs(struct the_nilfs *nilfs, struct super_block *sb, char *data)
561 {
562 	struct nilfs_super_block *sbp;
563 	int blocksize;
564 	int err;
565 
566 	down_write(&nilfs->ns_sem);
567 
568 	blocksize = sb_min_blocksize(sb, NILFS_MIN_BLOCK_SIZE);
569 	if (!blocksize) {
570 		printk(KERN_ERR "NILFS: unable to set blocksize\n");
571 		err = -EINVAL;
572 		goto out;
573 	}
574 	err = nilfs_load_super_block(nilfs, sb, blocksize, &sbp);
575 	if (err)
576 		goto out;
577 
578 	err = nilfs_store_magic_and_option(sb, sbp, data);
579 	if (err)
580 		goto failed_sbh;
581 
582 	err = nilfs_check_feature_compatibility(sb, sbp);
583 	if (err)
584 		goto failed_sbh;
585 
586 	blocksize = BLOCK_SIZE << le32_to_cpu(sbp->s_log_block_size);
587 	if (blocksize < NILFS_MIN_BLOCK_SIZE ||
588 	    blocksize > NILFS_MAX_BLOCK_SIZE) {
589 		printk(KERN_ERR "NILFS: couldn't mount because of unsupported "
590 		       "filesystem blocksize %d\n", blocksize);
591 		err = -EINVAL;
592 		goto failed_sbh;
593 	}
594 	if (sb->s_blocksize != blocksize) {
595 		int hw_blocksize = bdev_logical_block_size(sb->s_bdev);
596 
597 		if (blocksize < hw_blocksize) {
598 			printk(KERN_ERR
599 			       "NILFS: blocksize %d too small for device "
600 			       "(sector-size = %d).\n",
601 			       blocksize, hw_blocksize);
602 			err = -EINVAL;
603 			goto failed_sbh;
604 		}
605 		nilfs_release_super_block(nilfs);
606 		sb_set_blocksize(sb, blocksize);
607 
608 		err = nilfs_load_super_block(nilfs, sb, blocksize, &sbp);
609 		if (err)
610 			goto out;
611 			/* not failed_sbh; sbh is released automatically
612 			   when reloading fails. */
613 	}
614 	nilfs->ns_blocksize_bits = sb->s_blocksize_bits;
615 	nilfs->ns_blocksize = blocksize;
616 
617 	get_random_bytes(&nilfs->ns_next_generation,
618 			 sizeof(nilfs->ns_next_generation));
619 
620 	err = nilfs_store_disk_layout(nilfs, sbp);
621 	if (err)
622 		goto failed_sbh;
623 
624 	sb->s_maxbytes = nilfs_max_size(sb->s_blocksize_bits);
625 
626 	nilfs->ns_mount_state = le16_to_cpu(sbp->s_state);
627 
628 	err = nilfs_store_log_cursor(nilfs, sbp);
629 	if (err)
630 		goto failed_sbh;
631 
632 	set_nilfs_init(nilfs);
633 	err = 0;
634  out:
635 	up_write(&nilfs->ns_sem);
636 	return err;
637 
638  failed_sbh:
639 	nilfs_release_super_block(nilfs);
640 	goto out;
641 }
642 
643 int nilfs_discard_segments(struct the_nilfs *nilfs, __u64 *segnump,
644 			    size_t nsegs)
645 {
646 	sector_t seg_start, seg_end;
647 	sector_t start = 0, nblocks = 0;
648 	unsigned int sects_per_block;
649 	__u64 *sn;
650 	int ret = 0;
651 
652 	sects_per_block = (1 << nilfs->ns_blocksize_bits) /
653 		bdev_logical_block_size(nilfs->ns_bdev);
654 	for (sn = segnump; sn < segnump + nsegs; sn++) {
655 		nilfs_get_segment_range(nilfs, *sn, &seg_start, &seg_end);
656 
657 		if (!nblocks) {
658 			start = seg_start;
659 			nblocks = seg_end - seg_start + 1;
660 		} else if (start + nblocks == seg_start) {
661 			nblocks += seg_end - seg_start + 1;
662 		} else {
663 			ret = blkdev_issue_discard(nilfs->ns_bdev,
664 						   start * sects_per_block,
665 						   nblocks * sects_per_block,
666 						   GFP_NOFS, 0);
667 			if (ret < 0)
668 				return ret;
669 			nblocks = 0;
670 		}
671 	}
672 	if (nblocks)
673 		ret = blkdev_issue_discard(nilfs->ns_bdev,
674 					   start * sects_per_block,
675 					   nblocks * sects_per_block,
676 					   GFP_NOFS, 0);
677 	return ret;
678 }
679 
680 int nilfs_count_free_blocks(struct the_nilfs *nilfs, sector_t *nblocks)
681 {
682 	unsigned long ncleansegs;
683 
684 	down_read(&NILFS_MDT(nilfs->ns_dat)->mi_sem);
685 	ncleansegs = nilfs_sufile_get_ncleansegs(nilfs->ns_sufile);
686 	up_read(&NILFS_MDT(nilfs->ns_dat)->mi_sem);
687 	*nblocks = (sector_t)ncleansegs * nilfs->ns_blocks_per_segment;
688 	return 0;
689 }
690 
691 int nilfs_near_disk_full(struct the_nilfs *nilfs)
692 {
693 	unsigned long ncleansegs, nincsegs;
694 
695 	ncleansegs = nilfs_sufile_get_ncleansegs(nilfs->ns_sufile);
696 	nincsegs = atomic_read(&nilfs->ns_ndirtyblks) /
697 		nilfs->ns_blocks_per_segment + 1;
698 
699 	return ncleansegs <= nilfs->ns_nrsvsegs + nincsegs;
700 }
701 
702 struct nilfs_root *nilfs_lookup_root(struct the_nilfs *nilfs, __u64 cno)
703 {
704 	struct rb_node *n;
705 	struct nilfs_root *root;
706 
707 	spin_lock(&nilfs->ns_cptree_lock);
708 	n = nilfs->ns_cptree.rb_node;
709 	while (n) {
710 		root = rb_entry(n, struct nilfs_root, rb_node);
711 
712 		if (cno < root->cno) {
713 			n = n->rb_left;
714 		} else if (cno > root->cno) {
715 			n = n->rb_right;
716 		} else {
717 			atomic_inc(&root->count);
718 			spin_unlock(&nilfs->ns_cptree_lock);
719 			return root;
720 		}
721 	}
722 	spin_unlock(&nilfs->ns_cptree_lock);
723 
724 	return NULL;
725 }
726 
727 struct nilfs_root *
728 nilfs_find_or_create_root(struct the_nilfs *nilfs, __u64 cno)
729 {
730 	struct rb_node **p, *parent;
731 	struct nilfs_root *root, *new;
732 
733 	root = nilfs_lookup_root(nilfs, cno);
734 	if (root)
735 		return root;
736 
737 	new = kmalloc(sizeof(*root), GFP_KERNEL);
738 	if (!new)
739 		return NULL;
740 
741 	spin_lock(&nilfs->ns_cptree_lock);
742 
743 	p = &nilfs->ns_cptree.rb_node;
744 	parent = NULL;
745 
746 	while (*p) {
747 		parent = *p;
748 		root = rb_entry(parent, struct nilfs_root, rb_node);
749 
750 		if (cno < root->cno) {
751 			p = &(*p)->rb_left;
752 		} else if (cno > root->cno) {
753 			p = &(*p)->rb_right;
754 		} else {
755 			atomic_inc(&root->count);
756 			spin_unlock(&nilfs->ns_cptree_lock);
757 			kfree(new);
758 			return root;
759 		}
760 	}
761 
762 	new->cno = cno;
763 	new->ifile = NULL;
764 	new->nilfs = nilfs;
765 	atomic_set(&new->count, 1);
766 	atomic_set(&new->inodes_count, 0);
767 	atomic_set(&new->blocks_count, 0);
768 
769 	rb_link_node(&new->rb_node, parent, p);
770 	rb_insert_color(&new->rb_node, &nilfs->ns_cptree);
771 
772 	spin_unlock(&nilfs->ns_cptree_lock);
773 
774 	return new;
775 }
776 
777 void nilfs_put_root(struct nilfs_root *root)
778 {
779 	if (atomic_dec_and_test(&root->count)) {
780 		struct the_nilfs *nilfs = root->nilfs;
781 
782 		spin_lock(&nilfs->ns_cptree_lock);
783 		rb_erase(&root->rb_node, &nilfs->ns_cptree);
784 		spin_unlock(&nilfs->ns_cptree_lock);
785 		if (root->ifile)
786 			iput(root->ifile);
787 
788 		kfree(root);
789 	}
790 }
791