blob: 08a94c814bdc9a9a5047e632863ac1d03bdfcd8d [file] [log] [blame]
Jaegeuk Kim0a8165d2012-11-29 13:28:09 +09001/*
Jaegeuk Kimaff063e2012-11-02 17:07:47 +09002 * fs/f2fs/super.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11#include <linux/module.h>
12#include <linux/init.h>
13#include <linux/fs.h>
14#include <linux/statfs.h>
15#include <linux/proc_fs.h>
16#include <linux/buffer_head.h>
17#include <linux/backing-dev.h>
18#include <linux/kthread.h>
19#include <linux/parser.h>
20#include <linux/mount.h>
21#include <linux/seq_file.h>
22#include <linux/random.h>
23#include <linux/exportfs.h>
24#include <linux/f2fs_fs.h>
25
26#include "f2fs.h"
27#include "node.h"
28#include "xattr.h"
29
30static struct kmem_cache *f2fs_inode_cachep;
31
32enum {
33 Opt_gc_background_off,
34 Opt_disable_roll_forward,
35 Opt_discard,
36 Opt_noheap,
37 Opt_nouser_xattr,
38 Opt_noacl,
39 Opt_active_logs,
40 Opt_disable_ext_identify,
41 Opt_err,
42};
43
44static match_table_t f2fs_tokens = {
45 {Opt_gc_background_off, "background_gc_off"},
46 {Opt_disable_roll_forward, "disable_roll_forward"},
47 {Opt_discard, "discard"},
48 {Opt_noheap, "no_heap"},
49 {Opt_nouser_xattr, "nouser_xattr"},
50 {Opt_noacl, "noacl"},
51 {Opt_active_logs, "active_logs=%u"},
52 {Opt_disable_ext_identify, "disable_ext_identify"},
53 {Opt_err, NULL},
54};
55
56static void init_once(void *foo)
57{
58 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
59
Jaegeuk Kimaff063e2012-11-02 17:07:47 +090060 inode_init_once(&fi->vfs_inode);
61}
62
63static struct inode *f2fs_alloc_inode(struct super_block *sb)
64{
65 struct f2fs_inode_info *fi;
66
67 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_NOFS | __GFP_ZERO);
68 if (!fi)
69 return NULL;
70
71 init_once((void *) fi);
72
73 /* Initilize f2fs-specific inode info */
74 fi->vfs_inode.i_version = 1;
75 atomic_set(&fi->dirty_dents, 0);
76 fi->i_current_depth = 1;
77 fi->i_advise = 0;
78 rwlock_init(&fi->ext.ext_lock);
79
80 set_inode_flag(fi, FI_NEW_INODE);
81
82 return &fi->vfs_inode;
83}
84
85static void f2fs_i_callback(struct rcu_head *head)
86{
87 struct inode *inode = container_of(head, struct inode, i_rcu);
88 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
89}
90
Jaegeuk Kim25ca9232012-11-28 16:12:41 +090091static void f2fs_destroy_inode(struct inode *inode)
Jaegeuk Kimaff063e2012-11-02 17:07:47 +090092{
93 call_rcu(&inode->i_rcu, f2fs_i_callback);
94}
95
96static void f2fs_put_super(struct super_block *sb)
97{
98 struct f2fs_sb_info *sbi = F2FS_SB(sb);
99
100 f2fs_destroy_stats(sbi);
101 stop_gc_thread(sbi);
102
103 write_checkpoint(sbi, false, true);
104
105 iput(sbi->node_inode);
106 iput(sbi->meta_inode);
107
108 /* destroy f2fs internal modules */
109 destroy_node_manager(sbi);
110 destroy_segment_manager(sbi);
111
112 kfree(sbi->ckpt);
113
114 sb->s_fs_info = NULL;
115 brelse(sbi->raw_super_buf);
116 kfree(sbi);
117}
118
119int f2fs_sync_fs(struct super_block *sb, int sync)
120{
121 struct f2fs_sb_info *sbi = F2FS_SB(sb);
Jaegeuk Kimaff063e2012-11-02 17:07:47 +0900122
123 if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
124 return 0;
125
126 if (sync)
127 write_checkpoint(sbi, false, false);
128
Namjae Jeon64c576f2012-12-22 12:10:27 +0900129 return 0;
Jaegeuk Kimaff063e2012-11-02 17:07:47 +0900130}
131
132static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
133{
134 struct super_block *sb = dentry->d_sb;
135 struct f2fs_sb_info *sbi = F2FS_SB(sb);
136 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
137 block_t total_count, user_block_count, start_count, ovp_count;
138
139 total_count = le64_to_cpu(sbi->raw_super->block_count);
140 user_block_count = sbi->user_block_count;
141 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
142 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
143 buf->f_type = F2FS_SUPER_MAGIC;
144 buf->f_bsize = sbi->blocksize;
145
146 buf->f_blocks = total_count - start_count;
147 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
148 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
149
Jaegeuk Kim1362b5e2012-12-12 19:45:49 +0900150 buf->f_files = sbi->total_node_count;
151 buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
Jaegeuk Kimaff063e2012-11-02 17:07:47 +0900152
153 buf->f_namelen = F2FS_MAX_NAME_LEN;
154 buf->f_fsid.val[0] = (u32)id;
155 buf->f_fsid.val[1] = (u32)(id >> 32);
156
157 return 0;
158}
159
160static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
161{
162 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
163
164 if (test_opt(sbi, BG_GC))
165 seq_puts(seq, ",background_gc_on");
166 else
167 seq_puts(seq, ",background_gc_off");
168 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
169 seq_puts(seq, ",disable_roll_forward");
170 if (test_opt(sbi, DISCARD))
171 seq_puts(seq, ",discard");
172 if (test_opt(sbi, NOHEAP))
173 seq_puts(seq, ",no_heap_alloc");
174#ifdef CONFIG_F2FS_FS_XATTR
175 if (test_opt(sbi, XATTR_USER))
176 seq_puts(seq, ",user_xattr");
177 else
178 seq_puts(seq, ",nouser_xattr");
179#endif
180#ifdef CONFIG_F2FS_FS_POSIX_ACL
181 if (test_opt(sbi, POSIX_ACL))
182 seq_puts(seq, ",acl");
183 else
184 seq_puts(seq, ",noacl");
185#endif
186 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
187 seq_puts(seq, ",disable_ext_indentify");
188
189 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
190
191 return 0;
192}
193
194static struct super_operations f2fs_sops = {
195 .alloc_inode = f2fs_alloc_inode,
196 .destroy_inode = f2fs_destroy_inode,
197 .write_inode = f2fs_write_inode,
198 .show_options = f2fs_show_options,
199 .evict_inode = f2fs_evict_inode,
200 .put_super = f2fs_put_super,
201 .sync_fs = f2fs_sync_fs,
202 .statfs = f2fs_statfs,
203};
204
205static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
206 u64 ino, u32 generation)
207{
208 struct f2fs_sb_info *sbi = F2FS_SB(sb);
209 struct inode *inode;
210
211 if (ino < F2FS_ROOT_INO(sbi))
212 return ERR_PTR(-ESTALE);
213
214 /*
215 * f2fs_iget isn't quite right if the inode is currently unallocated!
216 * However f2fs_iget currently does appropriate checks to handle stale
217 * inodes so everything is OK.
218 */
219 inode = f2fs_iget(sb, ino);
220 if (IS_ERR(inode))
221 return ERR_CAST(inode);
222 if (generation && inode->i_generation != generation) {
223 /* we didn't find the right inode.. */
224 iput(inode);
225 return ERR_PTR(-ESTALE);
226 }
227 return inode;
228}
229
230static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
231 int fh_len, int fh_type)
232{
233 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
234 f2fs_nfs_get_inode);
235}
236
237static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
238 int fh_len, int fh_type)
239{
240 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
241 f2fs_nfs_get_inode);
242}
243
244static const struct export_operations f2fs_export_ops = {
245 .fh_to_dentry = f2fs_fh_to_dentry,
246 .fh_to_parent = f2fs_fh_to_parent,
247 .get_parent = f2fs_get_parent,
248};
249
250static int parse_options(struct f2fs_sb_info *sbi, char *options)
251{
252 substring_t args[MAX_OPT_ARGS];
253 char *p;
254 int arg = 0;
255
256 if (!options)
257 return 0;
258
259 while ((p = strsep(&options, ",")) != NULL) {
260 int token;
261 if (!*p)
262 continue;
263 /*
264 * Initialize args struct so we know whether arg was
265 * found; some options take optional arguments.
266 */
267 args[0].to = args[0].from = NULL;
268 token = match_token(p, f2fs_tokens, args);
269
270 switch (token) {
271 case Opt_gc_background_off:
272 clear_opt(sbi, BG_GC);
273 break;
274 case Opt_disable_roll_forward:
275 set_opt(sbi, DISABLE_ROLL_FORWARD);
276 break;
277 case Opt_discard:
278 set_opt(sbi, DISCARD);
279 break;
280 case Opt_noheap:
281 set_opt(sbi, NOHEAP);
282 break;
283#ifdef CONFIG_F2FS_FS_XATTR
284 case Opt_nouser_xattr:
285 clear_opt(sbi, XATTR_USER);
286 break;
287#else
288 case Opt_nouser_xattr:
289 pr_info("nouser_xattr options not supported\n");
290 break;
291#endif
292#ifdef CONFIG_F2FS_FS_POSIX_ACL
293 case Opt_noacl:
294 clear_opt(sbi, POSIX_ACL);
295 break;
296#else
297 case Opt_noacl:
298 pr_info("noacl options not supported\n");
299 break;
300#endif
301 case Opt_active_logs:
302 if (args->from && match_int(args, &arg))
303 return -EINVAL;
Jaegeuk Kim12a67142012-12-21 11:47:05 +0900304 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
Jaegeuk Kimaff063e2012-11-02 17:07:47 +0900305 return -EINVAL;
306 sbi->active_logs = arg;
307 break;
308 case Opt_disable_ext_identify:
309 set_opt(sbi, DISABLE_EXT_IDENTIFY);
310 break;
311 default:
Namjae Jeon72ce6092012-12-01 10:55:32 +0900312 pr_err("Unrecognized mount option \"%s\" or missing value\n",
313 p);
Jaegeuk Kimaff063e2012-11-02 17:07:47 +0900314 return -EINVAL;
315 }
316 }
317 return 0;
318}
319
320static loff_t max_file_size(unsigned bits)
321{
322 loff_t result = ADDRS_PER_INODE;
323 loff_t leaf_count = ADDRS_PER_BLOCK;
324
325 /* two direct node blocks */
326 result += (leaf_count * 2);
327
328 /* two indirect node blocks */
329 leaf_count *= NIDS_PER_BLOCK;
330 result += (leaf_count * 2);
331
332 /* one double indirect node block */
333 leaf_count *= NIDS_PER_BLOCK;
334 result += leaf_count;
335
336 result <<= bits;
337 return result;
338}
339
340static int sanity_check_raw_super(struct f2fs_super_block *raw_super)
341{
342 unsigned int blocksize;
343
344 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic))
345 return 1;
346
347 /* Currently, support only 4KB block size */
348 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
349 if (blocksize != PAGE_CACHE_SIZE)
350 return 1;
351 if (le32_to_cpu(raw_super->log_sectorsize) !=
352 F2FS_LOG_SECTOR_SIZE)
353 return 1;
354 if (le32_to_cpu(raw_super->log_sectors_per_block) !=
355 F2FS_LOG_SECTORS_PER_BLOCK)
356 return 1;
357 return 0;
358}
359
360static int sanity_check_ckpt(struct f2fs_super_block *raw_super,
361 struct f2fs_checkpoint *ckpt)
362{
363 unsigned int total, fsmeta;
364
365 total = le32_to_cpu(raw_super->segment_count);
366 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
367 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
368 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
369 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
370 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
371
372 if (fsmeta >= total)
373 return 1;
374 return 0;
375}
376
377static void init_sb_info(struct f2fs_sb_info *sbi)
378{
379 struct f2fs_super_block *raw_super = sbi->raw_super;
380 int i;
381
382 sbi->log_sectors_per_block =
383 le32_to_cpu(raw_super->log_sectors_per_block);
384 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
385 sbi->blocksize = 1 << sbi->log_blocksize;
386 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
387 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
388 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
389 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
390 sbi->total_sections = le32_to_cpu(raw_super->section_count);
391 sbi->total_node_count =
392 (le32_to_cpu(raw_super->segment_count_nat) / 2)
393 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
394 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
395 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
396 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
397
398 for (i = 0; i < NR_COUNT_TYPE; i++)
399 atomic_set(&sbi->nr_pages[i], 0);
400}
401
402static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
403{
404 struct f2fs_sb_info *sbi;
405 struct f2fs_super_block *raw_super;
406 struct buffer_head *raw_super_buf;
407 struct inode *root;
408 long err = -EINVAL;
409 int i;
410
411 /* allocate memory for f2fs-specific super block info */
412 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
413 if (!sbi)
414 return -ENOMEM;
415
416 /* set a temporary block size */
417 if (!sb_set_blocksize(sb, F2FS_BLKSIZE))
418 goto free_sbi;
419
420 /* read f2fs raw super block */
421 raw_super_buf = sb_bread(sb, 0);
422 if (!raw_super_buf) {
423 err = -EIO;
424 goto free_sbi;
425 }
426 raw_super = (struct f2fs_super_block *)
427 ((char *)raw_super_buf->b_data + F2FS_SUPER_OFFSET);
428
429 /* init some FS parameters */
430 sbi->active_logs = NR_CURSEG_TYPE;
431
432 set_opt(sbi, BG_GC);
433
434#ifdef CONFIG_F2FS_FS_XATTR
435 set_opt(sbi, XATTR_USER);
436#endif
437#ifdef CONFIG_F2FS_FS_POSIX_ACL
438 set_opt(sbi, POSIX_ACL);
439#endif
440 /* parse mount options */
441 if (parse_options(sbi, (char *)data))
442 goto free_sb_buf;
443
444 /* sanity checking of raw super */
445 if (sanity_check_raw_super(raw_super))
446 goto free_sb_buf;
447
Jaegeuk Kim25ca9232012-11-28 16:12:41 +0900448 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
Jaegeuk Kimaff063e2012-11-02 17:07:47 +0900449 sb->s_max_links = F2FS_LINK_MAX;
450 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
451
452 sb->s_op = &f2fs_sops;
453 sb->s_xattr = f2fs_xattr_handlers;
454 sb->s_export_op = &f2fs_export_ops;
455 sb->s_magic = F2FS_SUPER_MAGIC;
456 sb->s_fs_info = sbi;
457 sb->s_time_gran = 1;
458 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
459 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
460 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
461
462 /* init f2fs-specific super block info */
463 sbi->sb = sb;
464 sbi->raw_super = raw_super;
465 sbi->raw_super_buf = raw_super_buf;
466 mutex_init(&sbi->gc_mutex);
467 mutex_init(&sbi->write_inode);
468 mutex_init(&sbi->writepages);
469 mutex_init(&sbi->cp_mutex);
470 for (i = 0; i < NR_LOCK_TYPE; i++)
471 mutex_init(&sbi->fs_lock[i]);
472 sbi->por_doing = 0;
473 spin_lock_init(&sbi->stat_lock);
474 init_rwsem(&sbi->bio_sem);
475 init_sb_info(sbi);
476
477 /* get an inode for meta space */
478 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
479 if (IS_ERR(sbi->meta_inode)) {
480 err = PTR_ERR(sbi->meta_inode);
481 goto free_sb_buf;
482 }
483
484 err = get_valid_checkpoint(sbi);
485 if (err)
486 goto free_meta_inode;
487
488 /* sanity checking of checkpoint */
489 err = -EINVAL;
490 if (sanity_check_ckpt(raw_super, sbi->ckpt))
491 goto free_cp;
492
493 sbi->total_valid_node_count =
494 le32_to_cpu(sbi->ckpt->valid_node_count);
495 sbi->total_valid_inode_count =
496 le32_to_cpu(sbi->ckpt->valid_inode_count);
497 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
498 sbi->total_valid_block_count =
499 le64_to_cpu(sbi->ckpt->valid_block_count);
500 sbi->last_valid_block_count = sbi->total_valid_block_count;
501 sbi->alloc_valid_block_count = 0;
502 INIT_LIST_HEAD(&sbi->dir_inode_list);
503 spin_lock_init(&sbi->dir_inode_lock);
504
505 /* init super block */
506 if (!sb_set_blocksize(sb, sbi->blocksize))
507 goto free_cp;
508
509 init_orphan_info(sbi);
510
511 /* setup f2fs internal modules */
512 err = build_segment_manager(sbi);
513 if (err)
514 goto free_sm;
515 err = build_node_manager(sbi);
516 if (err)
517 goto free_nm;
518
519 build_gc_manager(sbi);
520
521 /* get an inode for node space */
522 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
523 if (IS_ERR(sbi->node_inode)) {
524 err = PTR_ERR(sbi->node_inode);
525 goto free_nm;
526 }
527
528 /* if there are nt orphan nodes free them */
529 err = -EINVAL;
Jaegeuk Kim30f0c752012-12-19 16:09:19 +0900530 if (recover_orphan_inodes(sbi))
Jaegeuk Kimaff063e2012-11-02 17:07:47 +0900531 goto free_node_inode;
532
533 /* read root inode and dentry */
534 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
535 if (IS_ERR(root)) {
536 err = PTR_ERR(root);
537 goto free_node_inode;
538 }
539 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size)
540 goto free_root_inode;
541
542 sb->s_root = d_make_root(root); /* allocate root dentry */
543 if (!sb->s_root) {
544 err = -ENOMEM;
545 goto free_root_inode;
546 }
547
548 /* recover fsynced data */
Jaegeuk Kim30f0c752012-12-19 16:09:19 +0900549 if (!test_opt(sbi, DISABLE_ROLL_FORWARD))
Jaegeuk Kimaff063e2012-11-02 17:07:47 +0900550 recover_fsync_data(sbi);
551
552 /* After POR, we can run background GC thread */
553 err = start_gc_thread(sbi);
554 if (err)
555 goto fail;
556
557 err = f2fs_build_stats(sbi);
558 if (err)
559 goto fail;
560
561 return 0;
562fail:
563 stop_gc_thread(sbi);
564free_root_inode:
565 dput(sb->s_root);
566 sb->s_root = NULL;
567free_node_inode:
568 iput(sbi->node_inode);
569free_nm:
570 destroy_node_manager(sbi);
571free_sm:
572 destroy_segment_manager(sbi);
573free_cp:
574 kfree(sbi->ckpt);
575free_meta_inode:
576 make_bad_inode(sbi->meta_inode);
577 iput(sbi->meta_inode);
578free_sb_buf:
579 brelse(raw_super_buf);
580free_sbi:
581 kfree(sbi);
582 return err;
583}
584
585static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
586 const char *dev_name, void *data)
587{
588 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
589}
590
591static struct file_system_type f2fs_fs_type = {
592 .owner = THIS_MODULE,
593 .name = "f2fs",
594 .mount = f2fs_mount,
595 .kill_sb = kill_block_super,
596 .fs_flags = FS_REQUIRES_DEV,
597};
598
599static int init_inodecache(void)
600{
601 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
602 sizeof(struct f2fs_inode_info), NULL);
603 if (f2fs_inode_cachep == NULL)
604 return -ENOMEM;
605 return 0;
606}
607
608static void destroy_inodecache(void)
609{
610 /*
611 * Make sure all delayed rcu free inodes are flushed before we
612 * destroy cache.
613 */
614 rcu_barrier();
615 kmem_cache_destroy(f2fs_inode_cachep);
616}
617
618static int __init init_f2fs_fs(void)
619{
620 int err;
621
622 err = init_inodecache();
623 if (err)
624 goto fail;
625 err = create_node_manager_caches();
626 if (err)
627 goto fail;
628 err = create_gc_caches();
629 if (err)
630 goto fail;
631 err = create_checkpoint_caches();
632 if (err)
633 goto fail;
634 return register_filesystem(&f2fs_fs_type);
635fail:
636 return err;
637}
638
639static void __exit exit_f2fs_fs(void)
640{
641 destroy_root_stats();
642 unregister_filesystem(&f2fs_fs_type);
643 destroy_checkpoint_caches();
644 destroy_gc_caches();
645 destroy_node_manager_caches();
646 destroy_inodecache();
647}
648
649module_init(init_f2fs_fs)
650module_exit(exit_f2fs_fs)
651
652MODULE_AUTHOR("Samsung Electronics's Praesto Team");
653MODULE_DESCRIPTION("Flash Friendly File System");
654MODULE_LICENSE("GPL");