]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blob - fs/f2fs/super.c
projects / mirror_ubuntu-hirsute-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: add a sysfs entry to control max_discards
[mirror_ubuntu-hirsute-kernel.git] / fs / f2fs / super.c
1 /*
2 * 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/buffer_head.h>
16 #include <linux/backing-dev.h>
17 #include <linux/kthread.h>
18 #include <linux/parser.h>
19 #include <linux/mount.h>
20 #include <linux/seq_file.h>
21 #include <linux/proc_fs.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/f2fs_fs.h>
26 #include <linux/sysfs.h>
27
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "gc.h"
33
34 #define CREATE_TRACE_POINTS
35 #include <trace/events/f2fs.h>
36
37 static struct proc_dir_entry *f2fs_proc_root;
38 static struct kmem_cache *f2fs_inode_cachep;
39 static struct kset *f2fs_kset;
40
41 enum {
42 Opt_gc_background,
43 Opt_disable_roll_forward,
44 Opt_discard,
45 Opt_noheap,
46 Opt_user_xattr,
47 Opt_nouser_xattr,
48 Opt_acl,
49 Opt_noacl,
50 Opt_active_logs,
51 Opt_disable_ext_identify,
52 Opt_inline_xattr,
53 Opt_err,
54 };
55
56 static match_table_t f2fs_tokens = {
57 {Opt_gc_background, "background_gc=%s"},
58 {Opt_disable_roll_forward, "disable_roll_forward"},
59 {Opt_discard, "discard"},
60 {Opt_noheap, "no_heap"},
61 {Opt_user_xattr, "user_xattr"},
62 {Opt_nouser_xattr, "nouser_xattr"},
63 {Opt_acl, "acl"},
64 {Opt_noacl, "noacl"},
65 {Opt_active_logs, "active_logs=%u"},
66 {Opt_disable_ext_identify, "disable_ext_identify"},
67 {Opt_inline_xattr, "inline_xattr"},
68 {Opt_err, NULL},
69 };
70
71 /* Sysfs support for f2fs */
72 enum {
73 GC_THREAD, /* struct f2fs_gc_thread */
74 SM_INFO, /* struct f2fs_sm_info */
75 };
76
77 struct f2fs_attr {
78 struct attribute attr;
79 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
80 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
81 const char *, size_t);
82 int struct_type;
83 int offset;
84 };
85
86 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
87 {
88 if (struct_type == GC_THREAD)
89 return (unsigned char *)sbi->gc_thread;
90 else if (struct_type == SM_INFO)
91 return (unsigned char *)SM_I(sbi);
92 return NULL;
93 }
94
95 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
96 struct f2fs_sb_info *sbi, char *buf)
97 {
98 unsigned char *ptr = NULL;
99 unsigned int *ui;
100
101 ptr = __struct_ptr(sbi, a->struct_type);
102 if (!ptr)
103 return -EINVAL;
104
105 ui = (unsigned int *)(ptr + a->offset);
106
107 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
108 }
109
110 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
111 struct f2fs_sb_info *sbi,
112 const char *buf, size_t count)
113 {
114 unsigned char *ptr;
115 unsigned long t;
116 unsigned int *ui;
117 ssize_t ret;
118
119 ptr = __struct_ptr(sbi, a->struct_type);
120 if (!ptr)
121 return -EINVAL;
122
123 ui = (unsigned int *)(ptr + a->offset);
124
125 ret = kstrtoul(skip_spaces(buf), 0, &t);
126 if (ret < 0)
127 return ret;
128 *ui = t;
129 return count;
130 }
131
132 static ssize_t f2fs_attr_show(struct kobject *kobj,
133 struct attribute *attr, char *buf)
134 {
135 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
136 s_kobj);
137 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
138
139 return a->show ? a->show(a, sbi, buf) : 0;
140 }
141
142 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
143 const char *buf, size_t len)
144 {
145 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
146 s_kobj);
147 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
148
149 return a->store ? a->store(a, sbi, buf, len) : 0;
150 }
151
152 static void f2fs_sb_release(struct kobject *kobj)
153 {
154 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
155 s_kobj);
156 complete(&sbi->s_kobj_unregister);
157 }
158
159 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
160 static struct f2fs_attr f2fs_attr_##_name = { \
161 .attr = {.name = __stringify(_name), .mode = _mode }, \
162 .show = _show, \
163 .store = _store, \
164 .struct_type = _struct_type, \
165 .offset = _offset \
166 }
167
168 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
169 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
170 f2fs_sbi_show, f2fs_sbi_store, \
171 offsetof(struct struct_name, elname))
172
173 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
174 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
175 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
176 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
177 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
178 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
179
180 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
181 static struct attribute *f2fs_attrs[] = {
182 ATTR_LIST(gc_min_sleep_time),
183 ATTR_LIST(gc_max_sleep_time),
184 ATTR_LIST(gc_no_gc_sleep_time),
185 ATTR_LIST(gc_idle),
186 ATTR_LIST(reclaim_segments),
187 ATTR_LIST(max_small_discards),
188 NULL,
189 };
190
191 static const struct sysfs_ops f2fs_attr_ops = {
192 .show = f2fs_attr_show,
193 .store = f2fs_attr_store,
194 };
195
196 static struct kobj_type f2fs_ktype = {
197 .default_attrs = f2fs_attrs,
198 .sysfs_ops = &f2fs_attr_ops,
199 .release = f2fs_sb_release,
200 };
201
202 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
203 {
204 struct va_format vaf;
205 va_list args;
206
207 va_start(args, fmt);
208 vaf.fmt = fmt;
209 vaf.va = &args;
210 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
211 va_end(args);
212 }
213
214 static void init_once(void *foo)
215 {
216 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
217
218 inode_init_once(&fi->vfs_inode);
219 }
220
221 static int parse_options(struct super_block *sb, char *options)
222 {
223 struct f2fs_sb_info *sbi = F2FS_SB(sb);
224 substring_t args[MAX_OPT_ARGS];
225 char *p, *name;
226 int arg = 0;
227
228 if (!options)
229 return 0;
230
231 while ((p = strsep(&options, ",")) != NULL) {
232 int token;
233 if (!*p)
234 continue;
235 /*
236 * Initialize args struct so we know whether arg was
237 * found; some options take optional arguments.
238 */
239 args[0].to = args[0].from = NULL;
240 token = match_token(p, f2fs_tokens, args);
241
242 switch (token) {
243 case Opt_gc_background:
244 name = match_strdup(&args[0]);
245
246 if (!name)
247 return -ENOMEM;
248 if (!strncmp(name, "on", 2))
249 set_opt(sbi, BG_GC);
250 else if (!strncmp(name, "off", 3))
251 clear_opt(sbi, BG_GC);
252 else {
253 kfree(name);
254 return -EINVAL;
255 }
256 kfree(name);
257 break;
258 case Opt_disable_roll_forward:
259 set_opt(sbi, DISABLE_ROLL_FORWARD);
260 break;
261 case Opt_discard:
262 set_opt(sbi, DISCARD);
263 break;
264 case Opt_noheap:
265 set_opt(sbi, NOHEAP);
266 break;
267 #ifdef CONFIG_F2FS_FS_XATTR
268 case Opt_user_xattr:
269 set_opt(sbi, XATTR_USER);
270 break;
271 case Opt_nouser_xattr:
272 clear_opt(sbi, XATTR_USER);
273 break;
274 case Opt_inline_xattr:
275 set_opt(sbi, INLINE_XATTR);
276 break;
277 #else
278 case Opt_user_xattr:
279 f2fs_msg(sb, KERN_INFO,
280 "user_xattr options not supported");
281 break;
282 case Opt_nouser_xattr:
283 f2fs_msg(sb, KERN_INFO,
284 "nouser_xattr options not supported");
285 break;
286 case Opt_inline_xattr:
287 f2fs_msg(sb, KERN_INFO,
288 "inline_xattr options not supported");
289 break;
290 #endif
291 #ifdef CONFIG_F2FS_FS_POSIX_ACL
292 case Opt_acl:
293 set_opt(sbi, POSIX_ACL);
294 break;
295 case Opt_noacl:
296 clear_opt(sbi, POSIX_ACL);
297 break;
298 #else
299 case Opt_acl:
300 f2fs_msg(sb, KERN_INFO, "acl options not supported");
301 break;
302 case Opt_noacl:
303 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
304 break;
305 #endif
306 case Opt_active_logs:
307 if (args->from && match_int(args, &arg))
308 return -EINVAL;
309 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
310 return -EINVAL;
311 sbi->active_logs = arg;
312 break;
313 case Opt_disable_ext_identify:
314 set_opt(sbi, DISABLE_EXT_IDENTIFY);
315 break;
316 default:
317 f2fs_msg(sb, KERN_ERR,
318 "Unrecognized mount option \"%s\" or missing value",
319 p);
320 return -EINVAL;
321 }
322 }
323 return 0;
324 }
325
326 static struct inode *f2fs_alloc_inode(struct super_block *sb)
327 {
328 struct f2fs_inode_info *fi;
329
330 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_NOFS | __GFP_ZERO);
331 if (!fi)
332 return NULL;
333
334 init_once((void *) fi);
335
336 /* Initialize f2fs-specific inode info */
337 fi->vfs_inode.i_version = 1;
338 atomic_set(&fi->dirty_dents, 0);
339 fi->i_current_depth = 1;
340 fi->i_advise = 0;
341 rwlock_init(&fi->ext.ext_lock);
342
343 set_inode_flag(fi, FI_NEW_INODE);
344
345 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
346 set_inode_flag(fi, FI_INLINE_XATTR);
347
348 return &fi->vfs_inode;
349 }
350
351 static int f2fs_drop_inode(struct inode *inode)
352 {
353 /*
354 * This is to avoid a deadlock condition like below.
355 * writeback_single_inode(inode)
356 * - f2fs_write_data_page
357 * - f2fs_gc -> iput -> evict
358 * - inode_wait_for_writeback(inode)
359 */
360 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
361 return 0;
362 return generic_drop_inode(inode);
363 }
364
365 /*
366 * f2fs_dirty_inode() is called from __mark_inode_dirty()
367 *
368 * We should call set_dirty_inode to write the dirty inode through write_inode.
369 */
370 static void f2fs_dirty_inode(struct inode *inode, int flags)
371 {
372 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
373 }
374
375 static void f2fs_i_callback(struct rcu_head *head)
376 {
377 struct inode *inode = container_of(head, struct inode, i_rcu);
378 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
379 }
380
381 static void f2fs_destroy_inode(struct inode *inode)
382 {
383 call_rcu(&inode->i_rcu, f2fs_i_callback);
384 }
385
386 static void f2fs_put_super(struct super_block *sb)
387 {
388 struct f2fs_sb_info *sbi = F2FS_SB(sb);
389
390 if (sbi->s_proc) {
391 remove_proc_entry("segment_info", sbi->s_proc);
392 remove_proc_entry(sb->s_id, f2fs_proc_root);
393 }
394 kobject_del(&sbi->s_kobj);
395
396 f2fs_destroy_stats(sbi);
397 stop_gc_thread(sbi);
398
399 /* We don't need to do checkpoint when it's clean */
400 if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
401 write_checkpoint(sbi, true);
402
403 iput(sbi->node_inode);
404 iput(sbi->meta_inode);
405
406 /* destroy f2fs internal modules */
407 destroy_node_manager(sbi);
408 destroy_segment_manager(sbi);
409
410 kfree(sbi->ckpt);
411 kobject_put(&sbi->s_kobj);
412 wait_for_completion(&sbi->s_kobj_unregister);
413
414 sb->s_fs_info = NULL;
415 brelse(sbi->raw_super_buf);
416 kfree(sbi);
417 }
418
419 int f2fs_sync_fs(struct super_block *sb, int sync)
420 {
421 struct f2fs_sb_info *sbi = F2FS_SB(sb);
422
423 trace_f2fs_sync_fs(sb, sync);
424
425 if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
426 return 0;
427
428 if (sync) {
429 mutex_lock(&sbi->gc_mutex);
430 write_checkpoint(sbi, false);
431 mutex_unlock(&sbi->gc_mutex);
432 } else {
433 f2fs_balance_fs(sbi);
434 }
435
436 return 0;
437 }
438
439 static int f2fs_freeze(struct super_block *sb)
440 {
441 int err;
442
443 if (f2fs_readonly(sb))
444 return 0;
445
446 err = f2fs_sync_fs(sb, 1);
447 return err;
448 }
449
450 static int f2fs_unfreeze(struct super_block *sb)
451 {
452 return 0;
453 }
454
455 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
456 {
457 struct super_block *sb = dentry->d_sb;
458 struct f2fs_sb_info *sbi = F2FS_SB(sb);
459 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
460 block_t total_count, user_block_count, start_count, ovp_count;
461
462 total_count = le64_to_cpu(sbi->raw_super->block_count);
463 user_block_count = sbi->user_block_count;
464 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
465 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
466 buf->f_type = F2FS_SUPER_MAGIC;
467 buf->f_bsize = sbi->blocksize;
468
469 buf->f_blocks = total_count - start_count;
470 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
471 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
472
473 buf->f_files = sbi->total_node_count;
474 buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
475
476 buf->f_namelen = F2FS_NAME_LEN;
477 buf->f_fsid.val[0] = (u32)id;
478 buf->f_fsid.val[1] = (u32)(id >> 32);
479
480 return 0;
481 }
482
483 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
484 {
485 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
486
487 if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
488 seq_printf(seq, ",background_gc=%s", "on");
489 else
490 seq_printf(seq, ",background_gc=%s", "off");
491 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
492 seq_puts(seq, ",disable_roll_forward");
493 if (test_opt(sbi, DISCARD))
494 seq_puts(seq, ",discard");
495 if (test_opt(sbi, NOHEAP))
496 seq_puts(seq, ",no_heap_alloc");
497 #ifdef CONFIG_F2FS_FS_XATTR
498 if (test_opt(sbi, XATTR_USER))
499 seq_puts(seq, ",user_xattr");
500 else
501 seq_puts(seq, ",nouser_xattr");
502 if (test_opt(sbi, INLINE_XATTR))
503 seq_puts(seq, ",inline_xattr");
504 #endif
505 #ifdef CONFIG_F2FS_FS_POSIX_ACL
506 if (test_opt(sbi, POSIX_ACL))
507 seq_puts(seq, ",acl");
508 else
509 seq_puts(seq, ",noacl");
510 #endif
511 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
512 seq_puts(seq, ",disable_ext_identify");
513
514 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
515
516 return 0;
517 }
518
519 static int segment_info_seq_show(struct seq_file *seq, void *offset)
520 {
521 struct super_block *sb = seq->private;
522 struct f2fs_sb_info *sbi = F2FS_SB(sb);
523 unsigned int total_segs = le32_to_cpu(sbi->raw_super->segment_count_main);
524 int i;
525
526 for (i = 0; i < total_segs; i++) {
527 seq_printf(seq, "%u", get_valid_blocks(sbi, i, 1));
528 if (i != 0 && (i % 10) == 0)
529 seq_puts(seq, "\n");
530 else
531 seq_puts(seq, " ");
532 }
533 return 0;
534 }
535
536 static int segment_info_open_fs(struct inode *inode, struct file *file)
537 {
538 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
539 }
540
541 static const struct file_operations f2fs_seq_segment_info_fops = {
542 .owner = THIS_MODULE,
543 .open = segment_info_open_fs,
544 .read = seq_read,
545 .llseek = seq_lseek,
546 .release = single_release,
547 };
548
549 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
550 {
551 struct f2fs_sb_info *sbi = F2FS_SB(sb);
552 struct f2fs_mount_info org_mount_opt;
553 int err, active_logs;
554
555 /*
556 * Save the old mount options in case we
557 * need to restore them.
558 */
559 org_mount_opt = sbi->mount_opt;
560 active_logs = sbi->active_logs;
561
562 /* parse mount options */
563 err = parse_options(sb, data);
564 if (err)
565 goto restore_opts;
566
567 /*
568 * Previous and new state of filesystem is RO,
569 * so no point in checking GC conditions.
570 */
571 if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
572 goto skip;
573
574 /*
575 * We stop the GC thread if FS is mounted as RO
576 * or if background_gc = off is passed in mount
577 * option. Also sync the filesystem.
578 */
579 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
580 if (sbi->gc_thread) {
581 stop_gc_thread(sbi);
582 f2fs_sync_fs(sb, 1);
583 }
584 } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
585 err = start_gc_thread(sbi);
586 if (err)
587 goto restore_opts;
588 }
589 skip:
590 /* Update the POSIXACL Flag */
591 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
592 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
593 return 0;
594
595 restore_opts:
596 sbi->mount_opt = org_mount_opt;
597 sbi->active_logs = active_logs;
598 return err;
599 }
600
601 static struct super_operations f2fs_sops = {
602 .alloc_inode = f2fs_alloc_inode,
603 .drop_inode = f2fs_drop_inode,
604 .destroy_inode = f2fs_destroy_inode,
605 .write_inode = f2fs_write_inode,
606 .dirty_inode = f2fs_dirty_inode,
607 .show_options = f2fs_show_options,
608 .evict_inode = f2fs_evict_inode,
609 .put_super = f2fs_put_super,
610 .sync_fs = f2fs_sync_fs,
611 .freeze_fs = f2fs_freeze,
612 .unfreeze_fs = f2fs_unfreeze,
613 .statfs = f2fs_statfs,
614 .remount_fs = f2fs_remount,
615 };
616
617 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
618 u64 ino, u32 generation)
619 {
620 struct f2fs_sb_info *sbi = F2FS_SB(sb);
621 struct inode *inode;
622
623 if (ino < F2FS_ROOT_INO(sbi))
624 return ERR_PTR(-ESTALE);
625
626 /*
627 * f2fs_iget isn't quite right if the inode is currently unallocated!
628 * However f2fs_iget currently does appropriate checks to handle stale
629 * inodes so everything is OK.
630 */
631 inode = f2fs_iget(sb, ino);
632 if (IS_ERR(inode))
633 return ERR_CAST(inode);
634 if (generation && inode->i_generation != generation) {
635 /* we didn't find the right inode.. */
636 iput(inode);
637 return ERR_PTR(-ESTALE);
638 }
639 return inode;
640 }
641
642 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
643 int fh_len, int fh_type)
644 {
645 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
646 f2fs_nfs_get_inode);
647 }
648
649 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
650 int fh_len, int fh_type)
651 {
652 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
653 f2fs_nfs_get_inode);
654 }
655
656 static const struct export_operations f2fs_export_ops = {
657 .fh_to_dentry = f2fs_fh_to_dentry,
658 .fh_to_parent = f2fs_fh_to_parent,
659 .get_parent = f2fs_get_parent,
660 };
661
662 static loff_t max_file_size(unsigned bits)
663 {
664 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
665 loff_t leaf_count = ADDRS_PER_BLOCK;
666
667 /* two direct node blocks */
668 result += (leaf_count * 2);
669
670 /* two indirect node blocks */
671 leaf_count *= NIDS_PER_BLOCK;
672 result += (leaf_count * 2);
673
674 /* one double indirect node block */
675 leaf_count *= NIDS_PER_BLOCK;
676 result += leaf_count;
677
678 result <<= bits;
679 return result;
680 }
681
682 static int sanity_check_raw_super(struct super_block *sb,
683 struct f2fs_super_block *raw_super)
684 {
685 unsigned int blocksize;
686
687 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
688 f2fs_msg(sb, KERN_INFO,
689 "Magic Mismatch, valid(0x%x) - read(0x%x)",
690 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
691 return 1;
692 }
693
694 /* Currently, support only 4KB page cache size */
695 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
696 f2fs_msg(sb, KERN_INFO,
697 "Invalid page_cache_size (%lu), supports only 4KB\n",
698 PAGE_CACHE_SIZE);
699 return 1;
700 }
701
702 /* Currently, support only 4KB block size */
703 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
704 if (blocksize != F2FS_BLKSIZE) {
705 f2fs_msg(sb, KERN_INFO,
706 "Invalid blocksize (%u), supports only 4KB\n",
707 blocksize);
708 return 1;
709 }
710
711 if (le32_to_cpu(raw_super->log_sectorsize) !=
712 F2FS_LOG_SECTOR_SIZE) {
713 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
714 return 1;
715 }
716 if (le32_to_cpu(raw_super->log_sectors_per_block) !=
717 F2FS_LOG_SECTORS_PER_BLOCK) {
718 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
719 return 1;
720 }
721 return 0;
722 }
723
724 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
725 {
726 unsigned int total, fsmeta;
727 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
728 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
729
730 total = le32_to_cpu(raw_super->segment_count);
731 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
732 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
733 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
734 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
735 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
736
737 if (fsmeta >= total)
738 return 1;
739
740 if (is_set_ckpt_flags(ckpt, CP_ERROR_FLAG)) {
741 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
742 return 1;
743 }
744 return 0;
745 }
746
747 static void init_sb_info(struct f2fs_sb_info *sbi)
748 {
749 struct f2fs_super_block *raw_super = sbi->raw_super;
750 int i;
751
752 sbi->log_sectors_per_block =
753 le32_to_cpu(raw_super->log_sectors_per_block);
754 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
755 sbi->blocksize = 1 << sbi->log_blocksize;
756 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
757 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
758 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
759 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
760 sbi->total_sections = le32_to_cpu(raw_super->section_count);
761 sbi->total_node_count =
762 (le32_to_cpu(raw_super->segment_count_nat) / 2)
763 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
764 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
765 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
766 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
767 sbi->cur_victim_sec = NULL_SECNO;
768
769 for (i = 0; i < NR_COUNT_TYPE; i++)
770 atomic_set(&sbi->nr_pages[i], 0);
771 }
772
773 /*
774 * Read f2fs raw super block.
775 * Because we have two copies of super block, so read the first one at first,
776 * if the first one is invalid, move to read the second one.
777 */
778 static int read_raw_super_block(struct super_block *sb,
779 struct f2fs_super_block **raw_super,
780 struct buffer_head **raw_super_buf)
781 {
782 int block = 0;
783
784 retry:
785 *raw_super_buf = sb_bread(sb, block);
786 if (!*raw_super_buf) {
787 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
788 block + 1);
789 if (block == 0) {
790 block++;
791 goto retry;
792 } else {
793 return -EIO;
794 }
795 }
796
797 *raw_super = (struct f2fs_super_block *)
798 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
799
800 /* sanity checking of raw super */
801 if (sanity_check_raw_super(sb, *raw_super)) {
802 brelse(*raw_super_buf);
803 f2fs_msg(sb, KERN_ERR, "Can't find a valid F2FS filesystem "
804 "in %dth superblock", block + 1);
805 if(block == 0) {
806 block++;
807 goto retry;
808 } else {
809 return -EINVAL;
810 }
811 }
812
813 return 0;
814 }
815
816 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
817 {
818 struct f2fs_sb_info *sbi;
819 struct f2fs_super_block *raw_super;
820 struct buffer_head *raw_super_buf;
821 struct inode *root;
822 long err = -EINVAL;
823
824 /* allocate memory for f2fs-specific super block info */
825 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
826 if (!sbi)
827 return -ENOMEM;
828
829 /* set a block size */
830 if (!sb_set_blocksize(sb, F2FS_BLKSIZE)) {
831 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
832 goto free_sbi;
833 }
834
835 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
836 if (err)
837 goto free_sbi;
838
839 sb->s_fs_info = sbi;
840 /* init some FS parameters */
841 sbi->active_logs = NR_CURSEG_TYPE;
842
843 set_opt(sbi, BG_GC);
844
845 #ifdef CONFIG_F2FS_FS_XATTR
846 set_opt(sbi, XATTR_USER);
847 #endif
848 #ifdef CONFIG_F2FS_FS_POSIX_ACL
849 set_opt(sbi, POSIX_ACL);
850 #endif
851 /* parse mount options */
852 err = parse_options(sb, (char *)data);
853 if (err)
854 goto free_sb_buf;
855
856 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
857 sb->s_max_links = F2FS_LINK_MAX;
858 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
859
860 sb->s_op = &f2fs_sops;
861 sb->s_xattr = f2fs_xattr_handlers;
862 sb->s_export_op = &f2fs_export_ops;
863 sb->s_magic = F2FS_SUPER_MAGIC;
864 sb->s_time_gran = 1;
865 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
866 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
867 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
868
869 /* init f2fs-specific super block info */
870 sbi->sb = sb;
871 sbi->raw_super = raw_super;
872 sbi->raw_super_buf = raw_super_buf;
873 mutex_init(&sbi->gc_mutex);
874 mutex_init(&sbi->writepages);
875 mutex_init(&sbi->cp_mutex);
876 mutex_init(&sbi->node_write);
877 sbi->por_doing = false;
878 spin_lock_init(&sbi->stat_lock);
879 init_rwsem(&sbi->bio_sem);
880 init_rwsem(&sbi->cp_rwsem);
881 init_waitqueue_head(&sbi->cp_wait);
882 init_sb_info(sbi);
883
884 /* get an inode for meta space */
885 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
886 if (IS_ERR(sbi->meta_inode)) {
887 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
888 err = PTR_ERR(sbi->meta_inode);
889 goto free_sb_buf;
890 }
891
892 err = get_valid_checkpoint(sbi);
893 if (err) {
894 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
895 goto free_meta_inode;
896 }
897
898 /* sanity checking of checkpoint */
899 err = -EINVAL;
900 if (sanity_check_ckpt(sbi)) {
901 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
902 goto free_cp;
903 }
904
905 sbi->total_valid_node_count =
906 le32_to_cpu(sbi->ckpt->valid_node_count);
907 sbi->total_valid_inode_count =
908 le32_to_cpu(sbi->ckpt->valid_inode_count);
909 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
910 sbi->total_valid_block_count =
911 le64_to_cpu(sbi->ckpt->valid_block_count);
912 sbi->last_valid_block_count = sbi->total_valid_block_count;
913 sbi->alloc_valid_block_count = 0;
914 INIT_LIST_HEAD(&sbi->dir_inode_list);
915 spin_lock_init(&sbi->dir_inode_lock);
916
917 init_orphan_info(sbi);
918
919 /* setup f2fs internal modules */
920 err = build_segment_manager(sbi);
921 if (err) {
922 f2fs_msg(sb, KERN_ERR,
923 "Failed to initialize F2FS segment manager");
924 goto free_sm;
925 }
926 err = build_node_manager(sbi);
927 if (err) {
928 f2fs_msg(sb, KERN_ERR,
929 "Failed to initialize F2FS node manager");
930 goto free_nm;
931 }
932
933 build_gc_manager(sbi);
934
935 /* get an inode for node space */
936 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
937 if (IS_ERR(sbi->node_inode)) {
938 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
939 err = PTR_ERR(sbi->node_inode);
940 goto free_nm;
941 }
942
943 /* if there are nt orphan nodes free them */
944 err = -EINVAL;
945 if (recover_orphan_inodes(sbi))
946 goto free_node_inode;
947
948 /* read root inode and dentry */
949 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
950 if (IS_ERR(root)) {
951 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
952 err = PTR_ERR(root);
953 goto free_node_inode;
954 }
955 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size)
956 goto free_root_inode;
957
958 sb->s_root = d_make_root(root); /* allocate root dentry */
959 if (!sb->s_root) {
960 err = -ENOMEM;
961 goto free_root_inode;
962 }
963
964 /* recover fsynced data */
965 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
966 err = recover_fsync_data(sbi);
967 if (err)
968 f2fs_msg(sb, KERN_ERR,
969 "Cannot recover all fsync data errno=%ld", err);
970 }
971
972 /*
973 * If filesystem is not mounted as read-only then
974 * do start the gc_thread.
975 */
976 if (!(sb->s_flags & MS_RDONLY)) {
977 /* After POR, we can run background GC thread.*/
978 err = start_gc_thread(sbi);
979 if (err)
980 goto free_gc;
981 }
982
983 err = f2fs_build_stats(sbi);
984 if (err)
985 goto free_gc;
986
987 if (f2fs_proc_root)
988 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
989
990 if (sbi->s_proc)
991 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
992 &f2fs_seq_segment_info_fops, sb);
993
994 if (test_opt(sbi, DISCARD)) {
995 struct request_queue *q = bdev_get_queue(sb->s_bdev);
996 if (!blk_queue_discard(q))
997 f2fs_msg(sb, KERN_WARNING,
998 "mounting with \"discard\" option, but "
999 "the device does not support discard");
1000 }
1001
1002 sbi->s_kobj.kset = f2fs_kset;
1003 init_completion(&sbi->s_kobj_unregister);
1004 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1005 "%s", sb->s_id);
1006 if (err)
1007 goto fail;
1008
1009 return 0;
1010 fail:
1011 if (sbi->s_proc) {
1012 remove_proc_entry("segment_info", sbi->s_proc);
1013 remove_proc_entry(sb->s_id, f2fs_proc_root);
1014 }
1015 f2fs_destroy_stats(sbi);
1016 free_gc:
1017 stop_gc_thread(sbi);
1018 free_root_inode:
1019 dput(sb->s_root);
1020 sb->s_root = NULL;
1021 free_node_inode:
1022 iput(sbi->node_inode);
1023 free_nm:
1024 destroy_node_manager(sbi);
1025 free_sm:
1026 destroy_segment_manager(sbi);
1027 free_cp:
1028 kfree(sbi->ckpt);
1029 free_meta_inode:
1030 make_bad_inode(sbi->meta_inode);
1031 iput(sbi->meta_inode);
1032 free_sb_buf:
1033 brelse(raw_super_buf);
1034 free_sbi:
1035 kfree(sbi);
1036 return err;
1037 }
1038
1039 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1040 const char *dev_name, void *data)
1041 {
1042 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1043 }
1044
1045 static struct file_system_type f2fs_fs_type = {
1046 .owner = THIS_MODULE,
1047 .name = "f2fs",
1048 .mount = f2fs_mount,
1049 .kill_sb = kill_block_super,
1050 .fs_flags = FS_REQUIRES_DEV,
1051 };
1052 MODULE_ALIAS_FS("f2fs");
1053
1054 static int __init init_inodecache(void)
1055 {
1056 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1057 sizeof(struct f2fs_inode_info), NULL);
1058 if (f2fs_inode_cachep == NULL)
1059 return -ENOMEM;
1060 return 0;
1061 }
1062
1063 static void destroy_inodecache(void)
1064 {
1065 /*
1066 * Make sure all delayed rcu free inodes are flushed before we
1067 * destroy cache.
1068 */
1069 rcu_barrier();
1070 kmem_cache_destroy(f2fs_inode_cachep);
1071 }
1072
1073 static int __init init_f2fs_fs(void)
1074 {
1075 int err;
1076
1077 err = init_inodecache();
1078 if (err)
1079 goto fail;
1080 err = create_node_manager_caches();
1081 if (err)
1082 goto free_inodecache;
1083 err = create_segment_manager_caches();
1084 if (err)
1085 goto free_node_manager_caches;
1086 err = create_gc_caches();
1087 if (err)
1088 goto free_segment_manager_caches;
1089 err = create_checkpoint_caches();
1090 if (err)
1091 goto free_gc_caches;
1092 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1093 if (!f2fs_kset) {
1094 err = -ENOMEM;
1095 goto free_checkpoint_caches;
1096 }
1097 err = register_filesystem(&f2fs_fs_type);
1098 if (err)
1099 goto free_kset;
1100 f2fs_create_root_stats();
1101 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1102 return 0;
1103
1104 free_kset:
1105 kset_unregister(f2fs_kset);
1106 free_checkpoint_caches:
1107 destroy_checkpoint_caches();
1108 free_gc_caches:
1109 destroy_gc_caches();
1110 free_segment_manager_caches:
1111 destroy_segment_manager_caches();
1112 free_node_manager_caches:
1113 destroy_node_manager_caches();
1114 free_inodecache:
1115 destroy_inodecache();
1116 fail:
1117 return err;
1118 }
1119
1120 static void __exit exit_f2fs_fs(void)
1121 {
1122 remove_proc_entry("fs/f2fs", NULL);
1123 f2fs_destroy_root_stats();
1124 unregister_filesystem(&f2fs_fs_type);
1125 destroy_checkpoint_caches();
1126 destroy_gc_caches();
1127 destroy_node_manager_caches();
1128 destroy_inodecache();
1129 kset_unregister(f2fs_kset);
1130 }
1131
1132 module_init(init_f2fs_fs)
1133 module_exit(exit_f2fs_fs)
1134
1135 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1136 MODULE_DESCRIPTION("Flash Friendly File System");
1137 MODULE_LICENSE("GPL");
Ubuntu Hirsute Linux kernel mirror