]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - fs/f2fs/super.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
f2fs: add infra for ino management
[mirror_ubuntu-jammy-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_inline_data,
54 Opt_flush_merge,
55 Opt_nobarrier,
56 Opt_err,
57 };
58
59 static match_table_t f2fs_tokens = {
60 {Opt_gc_background, "background_gc=%s"},
61 {Opt_disable_roll_forward, "disable_roll_forward"},
62 {Opt_discard, "discard"},
63 {Opt_noheap, "no_heap"},
64 {Opt_user_xattr, "user_xattr"},
65 {Opt_nouser_xattr, "nouser_xattr"},
66 {Opt_acl, "acl"},
67 {Opt_noacl, "noacl"},
68 {Opt_active_logs, "active_logs=%u"},
69 {Opt_disable_ext_identify, "disable_ext_identify"},
70 {Opt_inline_xattr, "inline_xattr"},
71 {Opt_inline_data, "inline_data"},
72 {Opt_flush_merge, "flush_merge"},
73 {Opt_nobarrier, "nobarrier"},
74 {Opt_err, NULL},
75 };
76
77 /* Sysfs support for f2fs */
78 enum {
79 GC_THREAD, /* struct f2fs_gc_thread */
80 SM_INFO, /* struct f2fs_sm_info */
81 NM_INFO, /* struct f2fs_nm_info */
82 F2FS_SBI, /* struct f2fs_sb_info */
83 };
84
85 struct f2fs_attr {
86 struct attribute attr;
87 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
88 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
89 const char *, size_t);
90 int struct_type;
91 int offset;
92 };
93
94 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
95 {
96 if (struct_type == GC_THREAD)
97 return (unsigned char *)sbi->gc_thread;
98 else if (struct_type == SM_INFO)
99 return (unsigned char *)SM_I(sbi);
100 else if (struct_type == NM_INFO)
101 return (unsigned char *)NM_I(sbi);
102 else if (struct_type == F2FS_SBI)
103 return (unsigned char *)sbi;
104 return NULL;
105 }
106
107 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
108 struct f2fs_sb_info *sbi, char *buf)
109 {
110 unsigned char *ptr = NULL;
111 unsigned int *ui;
112
113 ptr = __struct_ptr(sbi, a->struct_type);
114 if (!ptr)
115 return -EINVAL;
116
117 ui = (unsigned int *)(ptr + a->offset);
118
119 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
120 }
121
122 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
123 struct f2fs_sb_info *sbi,
124 const char *buf, size_t count)
125 {
126 unsigned char *ptr;
127 unsigned long t;
128 unsigned int *ui;
129 ssize_t ret;
130
131 ptr = __struct_ptr(sbi, a->struct_type);
132 if (!ptr)
133 return -EINVAL;
134
135 ui = (unsigned int *)(ptr + a->offset);
136
137 ret = kstrtoul(skip_spaces(buf), 0, &t);
138 if (ret < 0)
139 return ret;
140 *ui = t;
141 return count;
142 }
143
144 static ssize_t f2fs_attr_show(struct kobject *kobj,
145 struct attribute *attr, char *buf)
146 {
147 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
148 s_kobj);
149 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
150
151 return a->show ? a->show(a, sbi, buf) : 0;
152 }
153
154 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
155 const char *buf, size_t len)
156 {
157 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
158 s_kobj);
159 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
160
161 return a->store ? a->store(a, sbi, buf, len) : 0;
162 }
163
164 static void f2fs_sb_release(struct kobject *kobj)
165 {
166 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
167 s_kobj);
168 complete(&sbi->s_kobj_unregister);
169 }
170
171 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
172 static struct f2fs_attr f2fs_attr_##_name = { \
173 .attr = {.name = __stringify(_name), .mode = _mode }, \
174 .show = _show, \
175 .store = _store, \
176 .struct_type = _struct_type, \
177 .offset = _offset \
178 }
179
180 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
181 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
182 f2fs_sbi_show, f2fs_sbi_store, \
183 offsetof(struct struct_name, elname))
184
185 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
186 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
187 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
188 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
189 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
190 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
191 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
192 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
193 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
194 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
195 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
196
197 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
198 static struct attribute *f2fs_attrs[] = {
199 ATTR_LIST(gc_min_sleep_time),
200 ATTR_LIST(gc_max_sleep_time),
201 ATTR_LIST(gc_no_gc_sleep_time),
202 ATTR_LIST(gc_idle),
203 ATTR_LIST(reclaim_segments),
204 ATTR_LIST(max_small_discards),
205 ATTR_LIST(ipu_policy),
206 ATTR_LIST(min_ipu_util),
207 ATTR_LIST(max_victim_search),
208 ATTR_LIST(dir_level),
209 ATTR_LIST(ram_thresh),
210 NULL,
211 };
212
213 static const struct sysfs_ops f2fs_attr_ops = {
214 .show = f2fs_attr_show,
215 .store = f2fs_attr_store,
216 };
217
218 static struct kobj_type f2fs_ktype = {
219 .default_attrs = f2fs_attrs,
220 .sysfs_ops = &f2fs_attr_ops,
221 .release = f2fs_sb_release,
222 };
223
224 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
225 {
226 struct va_format vaf;
227 va_list args;
228
229 va_start(args, fmt);
230 vaf.fmt = fmt;
231 vaf.va = &args;
232 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
233 va_end(args);
234 }
235
236 static void init_once(void *foo)
237 {
238 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
239
240 inode_init_once(&fi->vfs_inode);
241 }
242
243 static int parse_options(struct super_block *sb, char *options)
244 {
245 struct f2fs_sb_info *sbi = F2FS_SB(sb);
246 substring_t args[MAX_OPT_ARGS];
247 char *p, *name;
248 int arg = 0;
249
250 if (!options)
251 return 0;
252
253 while ((p = strsep(&options, ",")) != NULL) {
254 int token;
255 if (!*p)
256 continue;
257 /*
258 * Initialize args struct so we know whether arg was
259 * found; some options take optional arguments.
260 */
261 args[0].to = args[0].from = NULL;
262 token = match_token(p, f2fs_tokens, args);
263
264 switch (token) {
265 case Opt_gc_background:
266 name = match_strdup(&args[0]);
267
268 if (!name)
269 return -ENOMEM;
270 if (strlen(name) == 2 && !strncmp(name, "on", 2))
271 set_opt(sbi, BG_GC);
272 else if (strlen(name) == 3 && !strncmp(name, "off", 3))
273 clear_opt(sbi, BG_GC);
274 else {
275 kfree(name);
276 return -EINVAL;
277 }
278 kfree(name);
279 break;
280 case Opt_disable_roll_forward:
281 set_opt(sbi, DISABLE_ROLL_FORWARD);
282 break;
283 case Opt_discard:
284 set_opt(sbi, DISCARD);
285 break;
286 case Opt_noheap:
287 set_opt(sbi, NOHEAP);
288 break;
289 #ifdef CONFIG_F2FS_FS_XATTR
290 case Opt_user_xattr:
291 set_opt(sbi, XATTR_USER);
292 break;
293 case Opt_nouser_xattr:
294 clear_opt(sbi, XATTR_USER);
295 break;
296 case Opt_inline_xattr:
297 set_opt(sbi, INLINE_XATTR);
298 break;
299 #else
300 case Opt_user_xattr:
301 f2fs_msg(sb, KERN_INFO,
302 "user_xattr options not supported");
303 break;
304 case Opt_nouser_xattr:
305 f2fs_msg(sb, KERN_INFO,
306 "nouser_xattr options not supported");
307 break;
308 case Opt_inline_xattr:
309 f2fs_msg(sb, KERN_INFO,
310 "inline_xattr options not supported");
311 break;
312 #endif
313 #ifdef CONFIG_F2FS_FS_POSIX_ACL
314 case Opt_acl:
315 set_opt(sbi, POSIX_ACL);
316 break;
317 case Opt_noacl:
318 clear_opt(sbi, POSIX_ACL);
319 break;
320 #else
321 case Opt_acl:
322 f2fs_msg(sb, KERN_INFO, "acl options not supported");
323 break;
324 case Opt_noacl:
325 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
326 break;
327 #endif
328 case Opt_active_logs:
329 if (args->from && match_int(args, &arg))
330 return -EINVAL;
331 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
332 return -EINVAL;
333 sbi->active_logs = arg;
334 break;
335 case Opt_disable_ext_identify:
336 set_opt(sbi, DISABLE_EXT_IDENTIFY);
337 break;
338 case Opt_inline_data:
339 set_opt(sbi, INLINE_DATA);
340 break;
341 case Opt_flush_merge:
342 set_opt(sbi, FLUSH_MERGE);
343 break;
344 case Opt_nobarrier:
345 set_opt(sbi, NOBARRIER);
346 break;
347 default:
348 f2fs_msg(sb, KERN_ERR,
349 "Unrecognized mount option \"%s\" or missing value",
350 p);
351 return -EINVAL;
352 }
353 }
354 return 0;
355 }
356
357 static struct inode *f2fs_alloc_inode(struct super_block *sb)
358 {
359 struct f2fs_inode_info *fi;
360
361 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
362 if (!fi)
363 return NULL;
364
365 init_once((void *) fi);
366
367 /* Initialize f2fs-specific inode info */
368 fi->vfs_inode.i_version = 1;
369 atomic_set(&fi->dirty_dents, 0);
370 fi->i_current_depth = 1;
371 fi->i_advise = 0;
372 rwlock_init(&fi->ext.ext_lock);
373 init_rwsem(&fi->i_sem);
374
375 set_inode_flag(fi, FI_NEW_INODE);
376
377 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
378 set_inode_flag(fi, FI_INLINE_XATTR);
379
380 /* Will be used by directory only */
381 fi->i_dir_level = F2FS_SB(sb)->dir_level;
382
383 return &fi->vfs_inode;
384 }
385
386 static int f2fs_drop_inode(struct inode *inode)
387 {
388 /*
389 * This is to avoid a deadlock condition like below.
390 * writeback_single_inode(inode)
391 * - f2fs_write_data_page
392 * - f2fs_gc -> iput -> evict
393 * - inode_wait_for_writeback(inode)
394 */
395 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
396 return 0;
397 return generic_drop_inode(inode);
398 }
399
400 /*
401 * f2fs_dirty_inode() is called from __mark_inode_dirty()
402 *
403 * We should call set_dirty_inode to write the dirty inode through write_inode.
404 */
405 static void f2fs_dirty_inode(struct inode *inode, int flags)
406 {
407 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
408 }
409
410 static void f2fs_i_callback(struct rcu_head *head)
411 {
412 struct inode *inode = container_of(head, struct inode, i_rcu);
413 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
414 }
415
416 static void f2fs_destroy_inode(struct inode *inode)
417 {
418 call_rcu(&inode->i_rcu, f2fs_i_callback);
419 }
420
421 static void f2fs_put_super(struct super_block *sb)
422 {
423 struct f2fs_sb_info *sbi = F2FS_SB(sb);
424
425 if (sbi->s_proc) {
426 remove_proc_entry("segment_info", sbi->s_proc);
427 remove_proc_entry(sb->s_id, f2fs_proc_root);
428 }
429 kobject_del(&sbi->s_kobj);
430
431 f2fs_destroy_stats(sbi);
432 stop_gc_thread(sbi);
433
434 /* We don't need to do checkpoint when it's clean */
435 if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
436 write_checkpoint(sbi, true);
437
438 iput(sbi->node_inode);
439 iput(sbi->meta_inode);
440
441 /* destroy f2fs internal modules */
442 destroy_node_manager(sbi);
443 destroy_segment_manager(sbi);
444
445 kfree(sbi->ckpt);
446 kobject_put(&sbi->s_kobj);
447 wait_for_completion(&sbi->s_kobj_unregister);
448
449 sb->s_fs_info = NULL;
450 brelse(sbi->raw_super_buf);
451 kfree(sbi);
452 }
453
454 int f2fs_sync_fs(struct super_block *sb, int sync)
455 {
456 struct f2fs_sb_info *sbi = F2FS_SB(sb);
457
458 trace_f2fs_sync_fs(sb, sync);
459
460 if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
461 return 0;
462
463 if (sync) {
464 mutex_lock(&sbi->gc_mutex);
465 write_checkpoint(sbi, false);
466 mutex_unlock(&sbi->gc_mutex);
467 } else {
468 f2fs_balance_fs(sbi);
469 }
470
471 return 0;
472 }
473
474 static int f2fs_freeze(struct super_block *sb)
475 {
476 int err;
477
478 if (f2fs_readonly(sb))
479 return 0;
480
481 err = f2fs_sync_fs(sb, 1);
482 return err;
483 }
484
485 static int f2fs_unfreeze(struct super_block *sb)
486 {
487 return 0;
488 }
489
490 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
491 {
492 struct super_block *sb = dentry->d_sb;
493 struct f2fs_sb_info *sbi = F2FS_SB(sb);
494 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
495 block_t total_count, user_block_count, start_count, ovp_count;
496
497 total_count = le64_to_cpu(sbi->raw_super->block_count);
498 user_block_count = sbi->user_block_count;
499 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
500 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
501 buf->f_type = F2FS_SUPER_MAGIC;
502 buf->f_bsize = sbi->blocksize;
503
504 buf->f_blocks = total_count - start_count;
505 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
506 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
507
508 buf->f_files = sbi->total_node_count;
509 buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
510
511 buf->f_namelen = F2FS_NAME_LEN;
512 buf->f_fsid.val[0] = (u32)id;
513 buf->f_fsid.val[1] = (u32)(id >> 32);
514
515 return 0;
516 }
517
518 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
519 {
520 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
521
522 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
523 seq_printf(seq, ",background_gc=%s", "on");
524 else
525 seq_printf(seq, ",background_gc=%s", "off");
526 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
527 seq_puts(seq, ",disable_roll_forward");
528 if (test_opt(sbi, DISCARD))
529 seq_puts(seq, ",discard");
530 if (test_opt(sbi, NOHEAP))
531 seq_puts(seq, ",no_heap_alloc");
532 #ifdef CONFIG_F2FS_FS_XATTR
533 if (test_opt(sbi, XATTR_USER))
534 seq_puts(seq, ",user_xattr");
535 else
536 seq_puts(seq, ",nouser_xattr");
537 if (test_opt(sbi, INLINE_XATTR))
538 seq_puts(seq, ",inline_xattr");
539 #endif
540 #ifdef CONFIG_F2FS_FS_POSIX_ACL
541 if (test_opt(sbi, POSIX_ACL))
542 seq_puts(seq, ",acl");
543 else
544 seq_puts(seq, ",noacl");
545 #endif
546 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
547 seq_puts(seq, ",disable_ext_identify");
548 if (test_opt(sbi, INLINE_DATA))
549 seq_puts(seq, ",inline_data");
550 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
551 seq_puts(seq, ",flush_merge");
552 if (test_opt(sbi, NOBARRIER))
553 seq_puts(seq, ",nobarrier");
554 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
555
556 return 0;
557 }
558
559 static int segment_info_seq_show(struct seq_file *seq, void *offset)
560 {
561 struct super_block *sb = seq->private;
562 struct f2fs_sb_info *sbi = F2FS_SB(sb);
563 unsigned int total_segs =
564 le32_to_cpu(sbi->raw_super->segment_count_main);
565 int i;
566
567 seq_puts(seq, "format: segment_type|valid_blocks\n"
568 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
569
570 for (i = 0; i < total_segs; i++) {
571 struct seg_entry *se = get_seg_entry(sbi, i);
572
573 if ((i % 10) == 0)
574 seq_printf(seq, "%-5d", i);
575 seq_printf(seq, "%d|%-3u", se->type,
576 get_valid_blocks(sbi, i, 1));
577 if ((i % 10) == 9 || i == (total_segs - 1))
578 seq_putc(seq, '\n');
579 else
580 seq_putc(seq, ' ');
581 }
582
583 return 0;
584 }
585
586 static int segment_info_open_fs(struct inode *inode, struct file *file)
587 {
588 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
589 }
590
591 static const struct file_operations f2fs_seq_segment_info_fops = {
592 .owner = THIS_MODULE,
593 .open = segment_info_open_fs,
594 .read = seq_read,
595 .llseek = seq_lseek,
596 .release = single_release,
597 };
598
599 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
600 {
601 struct f2fs_sb_info *sbi = F2FS_SB(sb);
602 struct f2fs_mount_info org_mount_opt;
603 int err, active_logs;
604 bool need_restart_gc = false;
605 bool need_stop_gc = false;
606
607 sync_filesystem(sb);
608
609 /*
610 * Save the old mount options in case we
611 * need to restore them.
612 */
613 org_mount_opt = sbi->mount_opt;
614 active_logs = sbi->active_logs;
615
616 /* parse mount options */
617 err = parse_options(sb, data);
618 if (err)
619 goto restore_opts;
620
621 /*
622 * Previous and new state of filesystem is RO,
623 * so skip checking GC and FLUSH_MERGE conditions.
624 */
625 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
626 goto skip;
627
628 /*
629 * We stop the GC thread if FS is mounted as RO
630 * or if background_gc = off is passed in mount
631 * option. Also sync the filesystem.
632 */
633 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
634 if (sbi->gc_thread) {
635 stop_gc_thread(sbi);
636 f2fs_sync_fs(sb, 1);
637 need_restart_gc = true;
638 }
639 } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
640 err = start_gc_thread(sbi);
641 if (err)
642 goto restore_opts;
643 need_stop_gc = true;
644 }
645
646 /*
647 * We stop issue flush thread if FS is mounted as RO
648 * or if flush_merge is not passed in mount option.
649 */
650 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
651 destroy_flush_cmd_control(sbi);
652 } else if (test_opt(sbi, FLUSH_MERGE) && !SM_I(sbi)->cmd_control_info) {
653 err = create_flush_cmd_control(sbi);
654 if (err)
655 goto restore_gc;
656 }
657 skip:
658 /* Update the POSIXACL Flag */
659 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
660 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
661 return 0;
662 restore_gc:
663 if (need_restart_gc) {
664 if (start_gc_thread(sbi))
665 f2fs_msg(sbi->sb, KERN_WARNING,
666 "background gc thread is stop");
667 } else if (need_stop_gc) {
668 stop_gc_thread(sbi);
669 }
670 restore_opts:
671 sbi->mount_opt = org_mount_opt;
672 sbi->active_logs = active_logs;
673 return err;
674 }
675
676 static struct super_operations f2fs_sops = {
677 .alloc_inode = f2fs_alloc_inode,
678 .drop_inode = f2fs_drop_inode,
679 .destroy_inode = f2fs_destroy_inode,
680 .write_inode = f2fs_write_inode,
681 .dirty_inode = f2fs_dirty_inode,
682 .show_options = f2fs_show_options,
683 .evict_inode = f2fs_evict_inode,
684 .put_super = f2fs_put_super,
685 .sync_fs = f2fs_sync_fs,
686 .freeze_fs = f2fs_freeze,
687 .unfreeze_fs = f2fs_unfreeze,
688 .statfs = f2fs_statfs,
689 .remount_fs = f2fs_remount,
690 };
691
692 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
693 u64 ino, u32 generation)
694 {
695 struct f2fs_sb_info *sbi = F2FS_SB(sb);
696 struct inode *inode;
697
698 if (check_nid_range(sbi, ino))
699 return ERR_PTR(-ESTALE);
700
701 /*
702 * f2fs_iget isn't quite right if the inode is currently unallocated!
703 * However f2fs_iget currently does appropriate checks to handle stale
704 * inodes so everything is OK.
705 */
706 inode = f2fs_iget(sb, ino);
707 if (IS_ERR(inode))
708 return ERR_CAST(inode);
709 if (unlikely(generation && inode->i_generation != generation)) {
710 /* we didn't find the right inode.. */
711 iput(inode);
712 return ERR_PTR(-ESTALE);
713 }
714 return inode;
715 }
716
717 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
718 int fh_len, int fh_type)
719 {
720 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
721 f2fs_nfs_get_inode);
722 }
723
724 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
725 int fh_len, int fh_type)
726 {
727 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
728 f2fs_nfs_get_inode);
729 }
730
731 static const struct export_operations f2fs_export_ops = {
732 .fh_to_dentry = f2fs_fh_to_dentry,
733 .fh_to_parent = f2fs_fh_to_parent,
734 .get_parent = f2fs_get_parent,
735 };
736
737 static loff_t max_file_size(unsigned bits)
738 {
739 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
740 loff_t leaf_count = ADDRS_PER_BLOCK;
741
742 /* two direct node blocks */
743 result += (leaf_count * 2);
744
745 /* two indirect node blocks */
746 leaf_count *= NIDS_PER_BLOCK;
747 result += (leaf_count * 2);
748
749 /* one double indirect node block */
750 leaf_count *= NIDS_PER_BLOCK;
751 result += leaf_count;
752
753 result <<= bits;
754 return result;
755 }
756
757 static int sanity_check_raw_super(struct super_block *sb,
758 struct f2fs_super_block *raw_super)
759 {
760 unsigned int blocksize;
761
762 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
763 f2fs_msg(sb, KERN_INFO,
764 "Magic Mismatch, valid(0x%x) - read(0x%x)",
765 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
766 return 1;
767 }
768
769 /* Currently, support only 4KB page cache size */
770 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
771 f2fs_msg(sb, KERN_INFO,
772 "Invalid page_cache_size (%lu), supports only 4KB\n",
773 PAGE_CACHE_SIZE);
774 return 1;
775 }
776
777 /* Currently, support only 4KB block size */
778 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
779 if (blocksize != F2FS_BLKSIZE) {
780 f2fs_msg(sb, KERN_INFO,
781 "Invalid blocksize (%u), supports only 4KB\n",
782 blocksize);
783 return 1;
784 }
785
786 if (le32_to_cpu(raw_super->log_sectorsize) !=
787 F2FS_LOG_SECTOR_SIZE) {
788 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
789 return 1;
790 }
791 if (le32_to_cpu(raw_super->log_sectors_per_block) !=
792 F2FS_LOG_SECTORS_PER_BLOCK) {
793 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
794 return 1;
795 }
796 return 0;
797 }
798
799 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
800 {
801 unsigned int total, fsmeta;
802 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
803 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
804
805 total = le32_to_cpu(raw_super->segment_count);
806 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
807 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
808 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
809 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
810 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
811
812 if (unlikely(fsmeta >= total))
813 return 1;
814
815 if (unlikely(is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
816 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
817 return 1;
818 }
819 return 0;
820 }
821
822 static void init_sb_info(struct f2fs_sb_info *sbi)
823 {
824 struct f2fs_super_block *raw_super = sbi->raw_super;
825 int i;
826
827 sbi->log_sectors_per_block =
828 le32_to_cpu(raw_super->log_sectors_per_block);
829 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
830 sbi->blocksize = 1 << sbi->log_blocksize;
831 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
832 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
833 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
834 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
835 sbi->total_sections = le32_to_cpu(raw_super->section_count);
836 sbi->total_node_count =
837 (le32_to_cpu(raw_super->segment_count_nat) / 2)
838 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
839 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
840 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
841 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
842 sbi->cur_victim_sec = NULL_SECNO;
843 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
844
845 for (i = 0; i < NR_COUNT_TYPE; i++)
846 atomic_set(&sbi->nr_pages[i], 0);
847
848 sbi->dir_level = DEF_DIR_LEVEL;
849 }
850
851 /*
852 * Read f2fs raw super block.
853 * Because we have two copies of super block, so read the first one at first,
854 * if the first one is invalid, move to read the second one.
855 */
856 static int read_raw_super_block(struct super_block *sb,
857 struct f2fs_super_block **raw_super,
858 struct buffer_head **raw_super_buf)
859 {
860 int block = 0;
861
862 retry:
863 *raw_super_buf = sb_bread(sb, block);
864 if (!*raw_super_buf) {
865 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
866 block + 1);
867 if (block == 0) {
868 block++;
869 goto retry;
870 } else {
871 return -EIO;
872 }
873 }
874
875 *raw_super = (struct f2fs_super_block *)
876 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
877
878 /* sanity checking of raw super */
879 if (sanity_check_raw_super(sb, *raw_super)) {
880 brelse(*raw_super_buf);
881 f2fs_msg(sb, KERN_ERR,
882 "Can't find valid F2FS filesystem in %dth superblock",
883 block + 1);
884 if (block == 0) {
885 block++;
886 goto retry;
887 } else {
888 return -EINVAL;
889 }
890 }
891
892 return 0;
893 }
894
895 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
896 {
897 struct f2fs_sb_info *sbi;
898 struct f2fs_super_block *raw_super;
899 struct buffer_head *raw_super_buf;
900 struct inode *root;
901 long err = -EINVAL;
902 int i;
903
904 /* allocate memory for f2fs-specific super block info */
905 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
906 if (!sbi)
907 return -ENOMEM;
908
909 /* set a block size */
910 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
911 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
912 goto free_sbi;
913 }
914
915 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
916 if (err)
917 goto free_sbi;
918
919 sb->s_fs_info = sbi;
920 /* init some FS parameters */
921 sbi->active_logs = NR_CURSEG_TYPE;
922
923 set_opt(sbi, BG_GC);
924
925 #ifdef CONFIG_F2FS_FS_XATTR
926 set_opt(sbi, XATTR_USER);
927 #endif
928 #ifdef CONFIG_F2FS_FS_POSIX_ACL
929 set_opt(sbi, POSIX_ACL);
930 #endif
931 /* parse mount options */
932 err = parse_options(sb, (char *)data);
933 if (err)
934 goto free_sb_buf;
935
936 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
937 sb->s_max_links = F2FS_LINK_MAX;
938 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
939
940 sb->s_op = &f2fs_sops;
941 sb->s_xattr = f2fs_xattr_handlers;
942 sb->s_export_op = &f2fs_export_ops;
943 sb->s_magic = F2FS_SUPER_MAGIC;
944 sb->s_time_gran = 1;
945 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
946 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
947 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
948
949 /* init f2fs-specific super block info */
950 sbi->sb = sb;
951 sbi->raw_super = raw_super;
952 sbi->raw_super_buf = raw_super_buf;
953 mutex_init(&sbi->gc_mutex);
954 mutex_init(&sbi->writepages);
955 mutex_init(&sbi->cp_mutex);
956 mutex_init(&sbi->node_write);
957 sbi->por_doing = false;
958 spin_lock_init(&sbi->stat_lock);
959
960 init_rwsem(&sbi->read_io.io_rwsem);
961 sbi->read_io.sbi = sbi;
962 sbi->read_io.bio = NULL;
963 for (i = 0; i < NR_PAGE_TYPE; i++) {
964 init_rwsem(&sbi->write_io[i].io_rwsem);
965 sbi->write_io[i].sbi = sbi;
966 sbi->write_io[i].bio = NULL;
967 }
968
969 init_rwsem(&sbi->cp_rwsem);
970 init_waitqueue_head(&sbi->cp_wait);
971 init_sb_info(sbi);
972
973 /* get an inode for meta space */
974 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
975 if (IS_ERR(sbi->meta_inode)) {
976 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
977 err = PTR_ERR(sbi->meta_inode);
978 goto free_sb_buf;
979 }
980
981 err = get_valid_checkpoint(sbi);
982 if (err) {
983 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
984 goto free_meta_inode;
985 }
986
987 /* sanity checking of checkpoint */
988 err = -EINVAL;
989 if (sanity_check_ckpt(sbi)) {
990 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
991 goto free_cp;
992 }
993
994 sbi->total_valid_node_count =
995 le32_to_cpu(sbi->ckpt->valid_node_count);
996 sbi->total_valid_inode_count =
997 le32_to_cpu(sbi->ckpt->valid_inode_count);
998 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
999 sbi->total_valid_block_count =
1000 le64_to_cpu(sbi->ckpt->valid_block_count);
1001 sbi->last_valid_block_count = sbi->total_valid_block_count;
1002 sbi->alloc_valid_block_count = 0;
1003 INIT_LIST_HEAD(&sbi->dir_inode_list);
1004 spin_lock_init(&sbi->dir_inode_lock);
1005
1006 init_ino_entry_info(sbi);
1007
1008 /* setup f2fs internal modules */
1009 err = build_segment_manager(sbi);
1010 if (err) {
1011 f2fs_msg(sb, KERN_ERR,
1012 "Failed to initialize F2FS segment manager");
1013 goto free_sm;
1014 }
1015 err = build_node_manager(sbi);
1016 if (err) {
1017 f2fs_msg(sb, KERN_ERR,
1018 "Failed to initialize F2FS node manager");
1019 goto free_nm;
1020 }
1021
1022 build_gc_manager(sbi);
1023
1024 /* get an inode for node space */
1025 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1026 if (IS_ERR(sbi->node_inode)) {
1027 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1028 err = PTR_ERR(sbi->node_inode);
1029 goto free_nm;
1030 }
1031
1032 /* if there are nt orphan nodes free them */
1033 recover_orphan_inodes(sbi);
1034
1035 /* read root inode and dentry */
1036 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1037 if (IS_ERR(root)) {
1038 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1039 err = PTR_ERR(root);
1040 goto free_node_inode;
1041 }
1042 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1043 iput(root);
1044 err = -EINVAL;
1045 goto free_node_inode;
1046 }
1047
1048 sb->s_root = d_make_root(root); /* allocate root dentry */
1049 if (!sb->s_root) {
1050 err = -ENOMEM;
1051 goto free_root_inode;
1052 }
1053
1054 err = f2fs_build_stats(sbi);
1055 if (err)
1056 goto free_root_inode;
1057
1058 if (f2fs_proc_root)
1059 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1060
1061 if (sbi->s_proc)
1062 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1063 &f2fs_seq_segment_info_fops, sb);
1064
1065 if (test_opt(sbi, DISCARD)) {
1066 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1067 if (!blk_queue_discard(q))
1068 f2fs_msg(sb, KERN_WARNING,
1069 "mounting with \"discard\" option, but "
1070 "the device does not support discard");
1071 }
1072
1073 sbi->s_kobj.kset = f2fs_kset;
1074 init_completion(&sbi->s_kobj_unregister);
1075 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1076 "%s", sb->s_id);
1077 if (err)
1078 goto free_proc;
1079
1080 /* recover fsynced data */
1081 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1082 err = recover_fsync_data(sbi);
1083 if (err)
1084 f2fs_msg(sb, KERN_ERR,
1085 "Cannot recover all fsync data errno=%ld", err);
1086 }
1087
1088 /*
1089 * If filesystem is not mounted as read-only then
1090 * do start the gc_thread.
1091 */
1092 if (!f2fs_readonly(sb)) {
1093 /* After POR, we can run background GC thread.*/
1094 err = start_gc_thread(sbi);
1095 if (err)
1096 goto free_kobj;
1097 }
1098 return 0;
1099
1100 free_kobj:
1101 kobject_del(&sbi->s_kobj);
1102 free_proc:
1103 if (sbi->s_proc) {
1104 remove_proc_entry("segment_info", sbi->s_proc);
1105 remove_proc_entry(sb->s_id, f2fs_proc_root);
1106 }
1107 f2fs_destroy_stats(sbi);
1108 free_root_inode:
1109 dput(sb->s_root);
1110 sb->s_root = NULL;
1111 free_node_inode:
1112 iput(sbi->node_inode);
1113 free_nm:
1114 destroy_node_manager(sbi);
1115 free_sm:
1116 destroy_segment_manager(sbi);
1117 free_cp:
1118 kfree(sbi->ckpt);
1119 free_meta_inode:
1120 make_bad_inode(sbi->meta_inode);
1121 iput(sbi->meta_inode);
1122 free_sb_buf:
1123 brelse(raw_super_buf);
1124 free_sbi:
1125 kfree(sbi);
1126 return err;
1127 }
1128
1129 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1130 const char *dev_name, void *data)
1131 {
1132 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1133 }
1134
1135 static struct file_system_type f2fs_fs_type = {
1136 .owner = THIS_MODULE,
1137 .name = "f2fs",
1138 .mount = f2fs_mount,
1139 .kill_sb = kill_block_super,
1140 .fs_flags = FS_REQUIRES_DEV,
1141 };
1142 MODULE_ALIAS_FS("f2fs");
1143
1144 static int __init init_inodecache(void)
1145 {
1146 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1147 sizeof(struct f2fs_inode_info));
1148 if (!f2fs_inode_cachep)
1149 return -ENOMEM;
1150 return 0;
1151 }
1152
1153 static void destroy_inodecache(void)
1154 {
1155 /*
1156 * Make sure all delayed rcu free inodes are flushed before we
1157 * destroy cache.
1158 */
1159 rcu_barrier();
1160 kmem_cache_destroy(f2fs_inode_cachep);
1161 }
1162
1163 static int __init init_f2fs_fs(void)
1164 {
1165 int err;
1166
1167 err = init_inodecache();
1168 if (err)
1169 goto fail;
1170 err = create_node_manager_caches();
1171 if (err)
1172 goto free_inodecache;
1173 err = create_segment_manager_caches();
1174 if (err)
1175 goto free_node_manager_caches;
1176 err = create_gc_caches();
1177 if (err)
1178 goto free_segment_manager_caches;
1179 err = create_checkpoint_caches();
1180 if (err)
1181 goto free_gc_caches;
1182 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1183 if (!f2fs_kset) {
1184 err = -ENOMEM;
1185 goto free_checkpoint_caches;
1186 }
1187 err = register_filesystem(&f2fs_fs_type);
1188 if (err)
1189 goto free_kset;
1190 f2fs_create_root_stats();
1191 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1192 return 0;
1193
1194 free_kset:
1195 kset_unregister(f2fs_kset);
1196 free_checkpoint_caches:
1197 destroy_checkpoint_caches();
1198 free_gc_caches:
1199 destroy_gc_caches();
1200 free_segment_manager_caches:
1201 destroy_segment_manager_caches();
1202 free_node_manager_caches:
1203 destroy_node_manager_caches();
1204 free_inodecache:
1205 destroy_inodecache();
1206 fail:
1207 return err;
1208 }
1209
1210 static void __exit exit_f2fs_fs(void)
1211 {
1212 remove_proc_entry("fs/f2fs", NULL);
1213 f2fs_destroy_root_stats();
1214 unregister_filesystem(&f2fs_fs_type);
1215 destroy_checkpoint_caches();
1216 destroy_gc_caches();
1217 destroy_segment_manager_caches();
1218 destroy_node_manager_caches();
1219 destroy_inodecache();
1220 kset_unregister(f2fs_kset);
1221 }
1222
1223 module_init(init_f2fs_fs)
1224 module_exit(exit_f2fs_fs)
1225
1226 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1227 MODULE_DESCRIPTION("Flash Friendly File System");
1228 MODULE_LICENSE("GPL");
Ubuntu Jammy Linux kernel mirror