Loading...
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#include "trace.h"
34
35#define CREATE_TRACE_POINTS
36#include <trace/events/f2fs.h>
37
38static struct proc_dir_entry *f2fs_proc_root;
39static struct kmem_cache *f2fs_inode_cachep;
40static struct kset *f2fs_kset;
41
42#ifdef CONFIG_F2FS_FAULT_INJECTION
43
44char *fault_name[FAULT_MAX] = {
45 [FAULT_KMALLOC] = "kmalloc",
46 [FAULT_PAGE_ALLOC] = "page alloc",
47 [FAULT_ALLOC_NID] = "alloc nid",
48 [FAULT_ORPHAN] = "orphan",
49 [FAULT_BLOCK] = "no more block",
50 [FAULT_DIR_DEPTH] = "too big dir depth",
51 [FAULT_EVICT_INODE] = "evict_inode fail",
52 [FAULT_IO] = "IO error",
53 [FAULT_CHECKPOINT] = "checkpoint error",
54};
55
56static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
57 unsigned int rate)
58{
59 struct f2fs_fault_info *ffi = &sbi->fault_info;
60
61 if (rate) {
62 atomic_set(&ffi->inject_ops, 0);
63 ffi->inject_rate = rate;
64 ffi->inject_type = (1 << FAULT_MAX) - 1;
65 } else {
66 memset(ffi, 0, sizeof(struct f2fs_fault_info));
67 }
68}
69#endif
70
71/* f2fs-wide shrinker description */
72static struct shrinker f2fs_shrinker_info = {
73 .scan_objects = f2fs_shrink_scan,
74 .count_objects = f2fs_shrink_count,
75 .seeks = DEFAULT_SEEKS,
76};
77
78enum {
79 Opt_gc_background,
80 Opt_disable_roll_forward,
81 Opt_norecovery,
82 Opt_discard,
83 Opt_nodiscard,
84 Opt_noheap,
85 Opt_user_xattr,
86 Opt_nouser_xattr,
87 Opt_acl,
88 Opt_noacl,
89 Opt_active_logs,
90 Opt_disable_ext_identify,
91 Opt_inline_xattr,
92 Opt_inline_data,
93 Opt_inline_dentry,
94 Opt_noinline_dentry,
95 Opt_flush_merge,
96 Opt_noflush_merge,
97 Opt_nobarrier,
98 Opt_fastboot,
99 Opt_extent_cache,
100 Opt_noextent_cache,
101 Opt_noinline_data,
102 Opt_data_flush,
103 Opt_mode,
104 Opt_fault_injection,
105 Opt_lazytime,
106 Opt_nolazytime,
107 Opt_err,
108};
109
110static match_table_t f2fs_tokens = {
111 {Opt_gc_background, "background_gc=%s"},
112 {Opt_disable_roll_forward, "disable_roll_forward"},
113 {Opt_norecovery, "norecovery"},
114 {Opt_discard, "discard"},
115 {Opt_nodiscard, "nodiscard"},
116 {Opt_noheap, "no_heap"},
117 {Opt_user_xattr, "user_xattr"},
118 {Opt_nouser_xattr, "nouser_xattr"},
119 {Opt_acl, "acl"},
120 {Opt_noacl, "noacl"},
121 {Opt_active_logs, "active_logs=%u"},
122 {Opt_disable_ext_identify, "disable_ext_identify"},
123 {Opt_inline_xattr, "inline_xattr"},
124 {Opt_inline_data, "inline_data"},
125 {Opt_inline_dentry, "inline_dentry"},
126 {Opt_noinline_dentry, "noinline_dentry"},
127 {Opt_flush_merge, "flush_merge"},
128 {Opt_noflush_merge, "noflush_merge"},
129 {Opt_nobarrier, "nobarrier"},
130 {Opt_fastboot, "fastboot"},
131 {Opt_extent_cache, "extent_cache"},
132 {Opt_noextent_cache, "noextent_cache"},
133 {Opt_noinline_data, "noinline_data"},
134 {Opt_data_flush, "data_flush"},
135 {Opt_mode, "mode=%s"},
136 {Opt_fault_injection, "fault_injection=%u"},
137 {Opt_lazytime, "lazytime"},
138 {Opt_nolazytime, "nolazytime"},
139 {Opt_err, NULL},
140};
141
142/* Sysfs support for f2fs */
143enum {
144 GC_THREAD, /* struct f2fs_gc_thread */
145 SM_INFO, /* struct f2fs_sm_info */
146 NM_INFO, /* struct f2fs_nm_info */
147 F2FS_SBI, /* struct f2fs_sb_info */
148#ifdef CONFIG_F2FS_FAULT_INJECTION
149 FAULT_INFO_RATE, /* struct f2fs_fault_info */
150 FAULT_INFO_TYPE, /* struct f2fs_fault_info */
151#endif
152};
153
154struct f2fs_attr {
155 struct attribute attr;
156 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
157 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
158 const char *, size_t);
159 int struct_type;
160 int offset;
161};
162
163static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
164{
165 if (struct_type == GC_THREAD)
166 return (unsigned char *)sbi->gc_thread;
167 else if (struct_type == SM_INFO)
168 return (unsigned char *)SM_I(sbi);
169 else if (struct_type == NM_INFO)
170 return (unsigned char *)NM_I(sbi);
171 else if (struct_type == F2FS_SBI)
172 return (unsigned char *)sbi;
173#ifdef CONFIG_F2FS_FAULT_INJECTION
174 else if (struct_type == FAULT_INFO_RATE ||
175 struct_type == FAULT_INFO_TYPE)
176 return (unsigned char *)&sbi->fault_info;
177#endif
178 return NULL;
179}
180
181static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
182 struct f2fs_sb_info *sbi, char *buf)
183{
184 struct super_block *sb = sbi->sb;
185
186 if (!sb->s_bdev->bd_part)
187 return snprintf(buf, PAGE_SIZE, "0\n");
188
189 return snprintf(buf, PAGE_SIZE, "%llu\n",
190 (unsigned long long)(sbi->kbytes_written +
191 BD_PART_WRITTEN(sbi)));
192}
193
194static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
195 struct f2fs_sb_info *sbi, char *buf)
196{
197 unsigned char *ptr = NULL;
198 unsigned int *ui;
199
200 ptr = __struct_ptr(sbi, a->struct_type);
201 if (!ptr)
202 return -EINVAL;
203
204 ui = (unsigned int *)(ptr + a->offset);
205
206 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
207}
208
209static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
210 struct f2fs_sb_info *sbi,
211 const char *buf, size_t count)
212{
213 unsigned char *ptr;
214 unsigned long t;
215 unsigned int *ui;
216 ssize_t ret;
217
218 ptr = __struct_ptr(sbi, a->struct_type);
219 if (!ptr)
220 return -EINVAL;
221
222 ui = (unsigned int *)(ptr + a->offset);
223
224 ret = kstrtoul(skip_spaces(buf), 0, &t);
225 if (ret < 0)
226 return ret;
227#ifdef CONFIG_F2FS_FAULT_INJECTION
228 if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
229 return -EINVAL;
230#endif
231 *ui = t;
232 return count;
233}
234
235static ssize_t f2fs_attr_show(struct kobject *kobj,
236 struct attribute *attr, char *buf)
237{
238 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
239 s_kobj);
240 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
241
242 return a->show ? a->show(a, sbi, buf) : 0;
243}
244
245static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
246 const char *buf, size_t len)
247{
248 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
249 s_kobj);
250 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
251
252 return a->store ? a->store(a, sbi, buf, len) : 0;
253}
254
255static void f2fs_sb_release(struct kobject *kobj)
256{
257 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
258 s_kobj);
259 complete(&sbi->s_kobj_unregister);
260}
261
262#define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
263static struct f2fs_attr f2fs_attr_##_name = { \
264 .attr = {.name = __stringify(_name), .mode = _mode }, \
265 .show = _show, \
266 .store = _store, \
267 .struct_type = _struct_type, \
268 .offset = _offset \
269}
270
271#define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
272 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
273 f2fs_sbi_show, f2fs_sbi_store, \
274 offsetof(struct struct_name, elname))
275
276#define F2FS_GENERAL_RO_ATTR(name) \
277static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
278
279F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
280F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
281F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
282F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
283F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
284F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
285F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
286F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
287F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
288F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
289F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
290F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
291F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
292F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
293F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
294F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
295F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
296#ifdef CONFIG_F2FS_FAULT_INJECTION
297F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
298F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
299#endif
300F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
301
302#define ATTR_LIST(name) (&f2fs_attr_##name.attr)
303static struct attribute *f2fs_attrs[] = {
304 ATTR_LIST(gc_min_sleep_time),
305 ATTR_LIST(gc_max_sleep_time),
306 ATTR_LIST(gc_no_gc_sleep_time),
307 ATTR_LIST(gc_idle),
308 ATTR_LIST(reclaim_segments),
309 ATTR_LIST(max_small_discards),
310 ATTR_LIST(batched_trim_sections),
311 ATTR_LIST(ipu_policy),
312 ATTR_LIST(min_ipu_util),
313 ATTR_LIST(min_fsync_blocks),
314 ATTR_LIST(max_victim_search),
315 ATTR_LIST(dir_level),
316 ATTR_LIST(ram_thresh),
317 ATTR_LIST(ra_nid_pages),
318 ATTR_LIST(dirty_nats_ratio),
319 ATTR_LIST(cp_interval),
320 ATTR_LIST(idle_interval),
321#ifdef CONFIG_F2FS_FAULT_INJECTION
322 ATTR_LIST(inject_rate),
323 ATTR_LIST(inject_type),
324#endif
325 ATTR_LIST(lifetime_write_kbytes),
326 NULL,
327};
328
329static const struct sysfs_ops f2fs_attr_ops = {
330 .show = f2fs_attr_show,
331 .store = f2fs_attr_store,
332};
333
334static struct kobj_type f2fs_ktype = {
335 .default_attrs = f2fs_attrs,
336 .sysfs_ops = &f2fs_attr_ops,
337 .release = f2fs_sb_release,
338};
339
340void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
341{
342 struct va_format vaf;
343 va_list args;
344
345 va_start(args, fmt);
346 vaf.fmt = fmt;
347 vaf.va = &args;
348 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
349 va_end(args);
350}
351
352static void init_once(void *foo)
353{
354 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
355
356 inode_init_once(&fi->vfs_inode);
357}
358
359static int parse_options(struct super_block *sb, char *options)
360{
361 struct f2fs_sb_info *sbi = F2FS_SB(sb);
362 struct request_queue *q;
363 substring_t args[MAX_OPT_ARGS];
364 char *p, *name;
365 int arg = 0;
366
367 if (!options)
368 return 0;
369
370 while ((p = strsep(&options, ",")) != NULL) {
371 int token;
372 if (!*p)
373 continue;
374 /*
375 * Initialize args struct so we know whether arg was
376 * found; some options take optional arguments.
377 */
378 args[0].to = args[0].from = NULL;
379 token = match_token(p, f2fs_tokens, args);
380
381 switch (token) {
382 case Opt_gc_background:
383 name = match_strdup(&args[0]);
384
385 if (!name)
386 return -ENOMEM;
387 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
388 set_opt(sbi, BG_GC);
389 clear_opt(sbi, FORCE_FG_GC);
390 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
391 clear_opt(sbi, BG_GC);
392 clear_opt(sbi, FORCE_FG_GC);
393 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
394 set_opt(sbi, BG_GC);
395 set_opt(sbi, FORCE_FG_GC);
396 } else {
397 kfree(name);
398 return -EINVAL;
399 }
400 kfree(name);
401 break;
402 case Opt_disable_roll_forward:
403 set_opt(sbi, DISABLE_ROLL_FORWARD);
404 break;
405 case Opt_norecovery:
406 /* this option mounts f2fs with ro */
407 set_opt(sbi, DISABLE_ROLL_FORWARD);
408 if (!f2fs_readonly(sb))
409 return -EINVAL;
410 break;
411 case Opt_discard:
412 q = bdev_get_queue(sb->s_bdev);
413 if (blk_queue_discard(q)) {
414 set_opt(sbi, DISCARD);
415 } else if (!f2fs_sb_mounted_blkzoned(sb)) {
416 f2fs_msg(sb, KERN_WARNING,
417 "mounting with \"discard\" option, but "
418 "the device does not support discard");
419 }
420 break;
421 case Opt_nodiscard:
422 if (f2fs_sb_mounted_blkzoned(sb)) {
423 f2fs_msg(sb, KERN_WARNING,
424 "discard is required for zoned block devices");
425 return -EINVAL;
426 }
427 clear_opt(sbi, DISCARD);
428 break;
429 case Opt_noheap:
430 set_opt(sbi, NOHEAP);
431 break;
432#ifdef CONFIG_F2FS_FS_XATTR
433 case Opt_user_xattr:
434 set_opt(sbi, XATTR_USER);
435 break;
436 case Opt_nouser_xattr:
437 clear_opt(sbi, XATTR_USER);
438 break;
439 case Opt_inline_xattr:
440 set_opt(sbi, INLINE_XATTR);
441 break;
442#else
443 case Opt_user_xattr:
444 f2fs_msg(sb, KERN_INFO,
445 "user_xattr options not supported");
446 break;
447 case Opt_nouser_xattr:
448 f2fs_msg(sb, KERN_INFO,
449 "nouser_xattr options not supported");
450 break;
451 case Opt_inline_xattr:
452 f2fs_msg(sb, KERN_INFO,
453 "inline_xattr options not supported");
454 break;
455#endif
456#ifdef CONFIG_F2FS_FS_POSIX_ACL
457 case Opt_acl:
458 set_opt(sbi, POSIX_ACL);
459 break;
460 case Opt_noacl:
461 clear_opt(sbi, POSIX_ACL);
462 break;
463#else
464 case Opt_acl:
465 f2fs_msg(sb, KERN_INFO, "acl options not supported");
466 break;
467 case Opt_noacl:
468 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
469 break;
470#endif
471 case Opt_active_logs:
472 if (args->from && match_int(args, &arg))
473 return -EINVAL;
474 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
475 return -EINVAL;
476 sbi->active_logs = arg;
477 break;
478 case Opt_disable_ext_identify:
479 set_opt(sbi, DISABLE_EXT_IDENTIFY);
480 break;
481 case Opt_inline_data:
482 set_opt(sbi, INLINE_DATA);
483 break;
484 case Opt_inline_dentry:
485 set_opt(sbi, INLINE_DENTRY);
486 break;
487 case Opt_noinline_dentry:
488 clear_opt(sbi, INLINE_DENTRY);
489 break;
490 case Opt_flush_merge:
491 set_opt(sbi, FLUSH_MERGE);
492 break;
493 case Opt_noflush_merge:
494 clear_opt(sbi, FLUSH_MERGE);
495 break;
496 case Opt_nobarrier:
497 set_opt(sbi, NOBARRIER);
498 break;
499 case Opt_fastboot:
500 set_opt(sbi, FASTBOOT);
501 break;
502 case Opt_extent_cache:
503 set_opt(sbi, EXTENT_CACHE);
504 break;
505 case Opt_noextent_cache:
506 clear_opt(sbi, EXTENT_CACHE);
507 break;
508 case Opt_noinline_data:
509 clear_opt(sbi, INLINE_DATA);
510 break;
511 case Opt_data_flush:
512 set_opt(sbi, DATA_FLUSH);
513 break;
514 case Opt_mode:
515 name = match_strdup(&args[0]);
516
517 if (!name)
518 return -ENOMEM;
519 if (strlen(name) == 8 &&
520 !strncmp(name, "adaptive", 8)) {
521 if (f2fs_sb_mounted_blkzoned(sb)) {
522 f2fs_msg(sb, KERN_WARNING,
523 "adaptive mode is not allowed with "
524 "zoned block device feature");
525 kfree(name);
526 return -EINVAL;
527 }
528 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
529 } else if (strlen(name) == 3 &&
530 !strncmp(name, "lfs", 3)) {
531 set_opt_mode(sbi, F2FS_MOUNT_LFS);
532 } else {
533 kfree(name);
534 return -EINVAL;
535 }
536 kfree(name);
537 break;
538 case Opt_fault_injection:
539 if (args->from && match_int(args, &arg))
540 return -EINVAL;
541#ifdef CONFIG_F2FS_FAULT_INJECTION
542 f2fs_build_fault_attr(sbi, arg);
543#else
544 f2fs_msg(sb, KERN_INFO,
545 "FAULT_INJECTION was not selected");
546#endif
547 break;
548 case Opt_lazytime:
549 sb->s_flags |= MS_LAZYTIME;
550 break;
551 case Opt_nolazytime:
552 sb->s_flags &= ~MS_LAZYTIME;
553 break;
554 default:
555 f2fs_msg(sb, KERN_ERR,
556 "Unrecognized mount option \"%s\" or missing value",
557 p);
558 return -EINVAL;
559 }
560 }
561 return 0;
562}
563
564static struct inode *f2fs_alloc_inode(struct super_block *sb)
565{
566 struct f2fs_inode_info *fi;
567
568 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
569 if (!fi)
570 return NULL;
571
572 init_once((void *) fi);
573
574 /* Initialize f2fs-specific inode info */
575 fi->vfs_inode.i_version = 1;
576 atomic_set(&fi->dirty_pages, 0);
577 fi->i_current_depth = 1;
578 fi->i_advise = 0;
579 init_rwsem(&fi->i_sem);
580 INIT_LIST_HEAD(&fi->dirty_list);
581 INIT_LIST_HEAD(&fi->gdirty_list);
582 INIT_LIST_HEAD(&fi->inmem_pages);
583 mutex_init(&fi->inmem_lock);
584 init_rwsem(&fi->dio_rwsem[READ]);
585 init_rwsem(&fi->dio_rwsem[WRITE]);
586
587 /* Will be used by directory only */
588 fi->i_dir_level = F2FS_SB(sb)->dir_level;
589 return &fi->vfs_inode;
590}
591
592static int f2fs_drop_inode(struct inode *inode)
593{
594 /*
595 * This is to avoid a deadlock condition like below.
596 * writeback_single_inode(inode)
597 * - f2fs_write_data_page
598 * - f2fs_gc -> iput -> evict
599 * - inode_wait_for_writeback(inode)
600 */
601 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
602 if (!inode->i_nlink && !is_bad_inode(inode)) {
603 /* to avoid evict_inode call simultaneously */
604 atomic_inc(&inode->i_count);
605 spin_unlock(&inode->i_lock);
606
607 /* some remained atomic pages should discarded */
608 if (f2fs_is_atomic_file(inode))
609 drop_inmem_pages(inode);
610
611 /* should remain fi->extent_tree for writepage */
612 f2fs_destroy_extent_node(inode);
613
614 sb_start_intwrite(inode->i_sb);
615 f2fs_i_size_write(inode, 0);
616
617 if (F2FS_HAS_BLOCKS(inode))
618 f2fs_truncate(inode);
619
620 sb_end_intwrite(inode->i_sb);
621
622 fscrypt_put_encryption_info(inode, NULL);
623 spin_lock(&inode->i_lock);
624 atomic_dec(&inode->i_count);
625 }
626 return 0;
627 }
628
629 return generic_drop_inode(inode);
630}
631
632int f2fs_inode_dirtied(struct inode *inode, bool sync)
633{
634 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
635 int ret = 0;
636
637 spin_lock(&sbi->inode_lock[DIRTY_META]);
638 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
639 ret = 1;
640 } else {
641 set_inode_flag(inode, FI_DIRTY_INODE);
642 stat_inc_dirty_inode(sbi, DIRTY_META);
643 }
644 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
645 list_add_tail(&F2FS_I(inode)->gdirty_list,
646 &sbi->inode_list[DIRTY_META]);
647 inc_page_count(sbi, F2FS_DIRTY_IMETA);
648 }
649 spin_unlock(&sbi->inode_lock[DIRTY_META]);
650 return ret;
651}
652
653void f2fs_inode_synced(struct inode *inode)
654{
655 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
656
657 spin_lock(&sbi->inode_lock[DIRTY_META]);
658 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
659 spin_unlock(&sbi->inode_lock[DIRTY_META]);
660 return;
661 }
662 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
663 list_del_init(&F2FS_I(inode)->gdirty_list);
664 dec_page_count(sbi, F2FS_DIRTY_IMETA);
665 }
666 clear_inode_flag(inode, FI_DIRTY_INODE);
667 clear_inode_flag(inode, FI_AUTO_RECOVER);
668 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
669 spin_unlock(&sbi->inode_lock[DIRTY_META]);
670}
671
672/*
673 * f2fs_dirty_inode() is called from __mark_inode_dirty()
674 *
675 * We should call set_dirty_inode to write the dirty inode through write_inode.
676 */
677static void f2fs_dirty_inode(struct inode *inode, int flags)
678{
679 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
680
681 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
682 inode->i_ino == F2FS_META_INO(sbi))
683 return;
684
685 if (flags == I_DIRTY_TIME)
686 return;
687
688 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
689 clear_inode_flag(inode, FI_AUTO_RECOVER);
690
691 f2fs_inode_dirtied(inode, false);
692}
693
694static void f2fs_i_callback(struct rcu_head *head)
695{
696 struct inode *inode = container_of(head, struct inode, i_rcu);
697 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
698}
699
700static void f2fs_destroy_inode(struct inode *inode)
701{
702 call_rcu(&inode->i_rcu, f2fs_i_callback);
703}
704
705static void destroy_percpu_info(struct f2fs_sb_info *sbi)
706{
707 percpu_counter_destroy(&sbi->alloc_valid_block_count);
708 percpu_counter_destroy(&sbi->total_valid_inode_count);
709}
710
711static void destroy_device_list(struct f2fs_sb_info *sbi)
712{
713 int i;
714
715 for (i = 0; i < sbi->s_ndevs; i++) {
716 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
717#ifdef CONFIG_BLK_DEV_ZONED
718 kfree(FDEV(i).blkz_type);
719#endif
720 }
721 kfree(sbi->devs);
722}
723
724static void f2fs_put_super(struct super_block *sb)
725{
726 struct f2fs_sb_info *sbi = F2FS_SB(sb);
727
728 if (sbi->s_proc) {
729 remove_proc_entry("segment_info", sbi->s_proc);
730 remove_proc_entry("segment_bits", sbi->s_proc);
731 remove_proc_entry(sb->s_id, f2fs_proc_root);
732 }
733 kobject_del(&sbi->s_kobj);
734
735 stop_gc_thread(sbi);
736
737 /* prevent remaining shrinker jobs */
738 mutex_lock(&sbi->umount_mutex);
739
740 /*
741 * We don't need to do checkpoint when superblock is clean.
742 * But, the previous checkpoint was not done by umount, it needs to do
743 * clean checkpoint again.
744 */
745 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
746 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
747 struct cp_control cpc = {
748 .reason = CP_UMOUNT,
749 };
750 write_checkpoint(sbi, &cpc);
751 }
752
753 /* write_checkpoint can update stat informaion */
754 f2fs_destroy_stats(sbi);
755
756 /*
757 * normally superblock is clean, so we need to release this.
758 * In addition, EIO will skip do checkpoint, we need this as well.
759 */
760 release_ino_entry(sbi, true);
761
762 f2fs_leave_shrinker(sbi);
763 mutex_unlock(&sbi->umount_mutex);
764
765 /* our cp_error case, we can wait for any writeback page */
766 f2fs_flush_merged_bios(sbi);
767
768 iput(sbi->node_inode);
769 iput(sbi->meta_inode);
770
771 /* destroy f2fs internal modules */
772 destroy_node_manager(sbi);
773 destroy_segment_manager(sbi);
774
775 kfree(sbi->ckpt);
776 kobject_put(&sbi->s_kobj);
777 wait_for_completion(&sbi->s_kobj_unregister);
778
779 sb->s_fs_info = NULL;
780 if (sbi->s_chksum_driver)
781 crypto_free_shash(sbi->s_chksum_driver);
782 kfree(sbi->raw_super);
783
784 destroy_device_list(sbi);
785
786 destroy_percpu_info(sbi);
787 kfree(sbi);
788}
789
790int f2fs_sync_fs(struct super_block *sb, int sync)
791{
792 struct f2fs_sb_info *sbi = F2FS_SB(sb);
793 int err = 0;
794
795 trace_f2fs_sync_fs(sb, sync);
796
797 if (sync) {
798 struct cp_control cpc;
799
800 cpc.reason = __get_cp_reason(sbi);
801
802 mutex_lock(&sbi->gc_mutex);
803 err = write_checkpoint(sbi, &cpc);
804 mutex_unlock(&sbi->gc_mutex);
805 }
806 f2fs_trace_ios(NULL, 1);
807
808 return err;
809}
810
811static int f2fs_freeze(struct super_block *sb)
812{
813 if (f2fs_readonly(sb))
814 return 0;
815
816 /* IO error happened before */
817 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
818 return -EIO;
819
820 /* must be clean, since sync_filesystem() was already called */
821 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
822 return -EINVAL;
823 return 0;
824}
825
826static int f2fs_unfreeze(struct super_block *sb)
827{
828 return 0;
829}
830
831static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
832{
833 struct super_block *sb = dentry->d_sb;
834 struct f2fs_sb_info *sbi = F2FS_SB(sb);
835 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
836 block_t total_count, user_block_count, start_count, ovp_count;
837
838 total_count = le64_to_cpu(sbi->raw_super->block_count);
839 user_block_count = sbi->user_block_count;
840 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
841 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
842 buf->f_type = F2FS_SUPER_MAGIC;
843 buf->f_bsize = sbi->blocksize;
844
845 buf->f_blocks = total_count - start_count;
846 buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
847 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
848
849 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
850 buf->f_ffree = min(buf->f_files - valid_node_count(sbi),
851 buf->f_bavail);
852
853 buf->f_namelen = F2FS_NAME_LEN;
854 buf->f_fsid.val[0] = (u32)id;
855 buf->f_fsid.val[1] = (u32)(id >> 32);
856
857 return 0;
858}
859
860static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
861{
862 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
863
864 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
865 if (test_opt(sbi, FORCE_FG_GC))
866 seq_printf(seq, ",background_gc=%s", "sync");
867 else
868 seq_printf(seq, ",background_gc=%s", "on");
869 } else {
870 seq_printf(seq, ",background_gc=%s", "off");
871 }
872 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
873 seq_puts(seq, ",disable_roll_forward");
874 if (test_opt(sbi, DISCARD))
875 seq_puts(seq, ",discard");
876 if (test_opt(sbi, NOHEAP))
877 seq_puts(seq, ",no_heap_alloc");
878#ifdef CONFIG_F2FS_FS_XATTR
879 if (test_opt(sbi, XATTR_USER))
880 seq_puts(seq, ",user_xattr");
881 else
882 seq_puts(seq, ",nouser_xattr");
883 if (test_opt(sbi, INLINE_XATTR))
884 seq_puts(seq, ",inline_xattr");
885#endif
886#ifdef CONFIG_F2FS_FS_POSIX_ACL
887 if (test_opt(sbi, POSIX_ACL))
888 seq_puts(seq, ",acl");
889 else
890 seq_puts(seq, ",noacl");
891#endif
892 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
893 seq_puts(seq, ",disable_ext_identify");
894 if (test_opt(sbi, INLINE_DATA))
895 seq_puts(seq, ",inline_data");
896 else
897 seq_puts(seq, ",noinline_data");
898 if (test_opt(sbi, INLINE_DENTRY))
899 seq_puts(seq, ",inline_dentry");
900 else
901 seq_puts(seq, ",noinline_dentry");
902 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
903 seq_puts(seq, ",flush_merge");
904 if (test_opt(sbi, NOBARRIER))
905 seq_puts(seq, ",nobarrier");
906 if (test_opt(sbi, FASTBOOT))
907 seq_puts(seq, ",fastboot");
908 if (test_opt(sbi, EXTENT_CACHE))
909 seq_puts(seq, ",extent_cache");
910 else
911 seq_puts(seq, ",noextent_cache");
912 if (test_opt(sbi, DATA_FLUSH))
913 seq_puts(seq, ",data_flush");
914
915 seq_puts(seq, ",mode=");
916 if (test_opt(sbi, ADAPTIVE))
917 seq_puts(seq, "adaptive");
918 else if (test_opt(sbi, LFS))
919 seq_puts(seq, "lfs");
920 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
921
922 return 0;
923}
924
925static int segment_info_seq_show(struct seq_file *seq, void *offset)
926{
927 struct super_block *sb = seq->private;
928 struct f2fs_sb_info *sbi = F2FS_SB(sb);
929 unsigned int total_segs =
930 le32_to_cpu(sbi->raw_super->segment_count_main);
931 int i;
932
933 seq_puts(seq, "format: segment_type|valid_blocks\n"
934 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
935
936 for (i = 0; i < total_segs; i++) {
937 struct seg_entry *se = get_seg_entry(sbi, i);
938
939 if ((i % 10) == 0)
940 seq_printf(seq, "%-10d", i);
941 seq_printf(seq, "%d|%-3u", se->type,
942 get_valid_blocks(sbi, i, 1));
943 if ((i % 10) == 9 || i == (total_segs - 1))
944 seq_putc(seq, '\n');
945 else
946 seq_putc(seq, ' ');
947 }
948
949 return 0;
950}
951
952static int segment_bits_seq_show(struct seq_file *seq, void *offset)
953{
954 struct super_block *sb = seq->private;
955 struct f2fs_sb_info *sbi = F2FS_SB(sb);
956 unsigned int total_segs =
957 le32_to_cpu(sbi->raw_super->segment_count_main);
958 int i, j;
959
960 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
961 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
962
963 for (i = 0; i < total_segs; i++) {
964 struct seg_entry *se = get_seg_entry(sbi, i);
965
966 seq_printf(seq, "%-10d", i);
967 seq_printf(seq, "%d|%-3u|", se->type,
968 get_valid_blocks(sbi, i, 1));
969 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
970 seq_printf(seq, " %.2x", se->cur_valid_map[j]);
971 seq_putc(seq, '\n');
972 }
973 return 0;
974}
975
976#define F2FS_PROC_FILE_DEF(_name) \
977static int _name##_open_fs(struct inode *inode, struct file *file) \
978{ \
979 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
980} \
981 \
982static const struct file_operations f2fs_seq_##_name##_fops = { \
983 .open = _name##_open_fs, \
984 .read = seq_read, \
985 .llseek = seq_lseek, \
986 .release = single_release, \
987};
988
989F2FS_PROC_FILE_DEF(segment_info);
990F2FS_PROC_FILE_DEF(segment_bits);
991
992static void default_options(struct f2fs_sb_info *sbi)
993{
994 /* init some FS parameters */
995 sbi->active_logs = NR_CURSEG_TYPE;
996
997 set_opt(sbi, BG_GC);
998 set_opt(sbi, INLINE_DATA);
999 set_opt(sbi, INLINE_DENTRY);
1000 set_opt(sbi, EXTENT_CACHE);
1001 sbi->sb->s_flags |= MS_LAZYTIME;
1002 set_opt(sbi, FLUSH_MERGE);
1003 if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1004 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1005 set_opt(sbi, DISCARD);
1006 } else {
1007 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1008 }
1009
1010#ifdef CONFIG_F2FS_FS_XATTR
1011 set_opt(sbi, XATTR_USER);
1012#endif
1013#ifdef CONFIG_F2FS_FS_POSIX_ACL
1014 set_opt(sbi, POSIX_ACL);
1015#endif
1016
1017#ifdef CONFIG_F2FS_FAULT_INJECTION
1018 f2fs_build_fault_attr(sbi, 0);
1019#endif
1020}
1021
1022static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1023{
1024 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1025 struct f2fs_mount_info org_mount_opt;
1026 int err, active_logs;
1027 bool need_restart_gc = false;
1028 bool need_stop_gc = false;
1029 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1030#ifdef CONFIG_F2FS_FAULT_INJECTION
1031 struct f2fs_fault_info ffi = sbi->fault_info;
1032#endif
1033
1034 /*
1035 * Save the old mount options in case we
1036 * need to restore them.
1037 */
1038 org_mount_opt = sbi->mount_opt;
1039 active_logs = sbi->active_logs;
1040
1041 /* recover superblocks we couldn't write due to previous RO mount */
1042 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1043 err = f2fs_commit_super(sbi, false);
1044 f2fs_msg(sb, KERN_INFO,
1045 "Try to recover all the superblocks, ret: %d", err);
1046 if (!err)
1047 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1048 }
1049
1050 sbi->mount_opt.opt = 0;
1051 default_options(sbi);
1052
1053 /* parse mount options */
1054 err = parse_options(sb, data);
1055 if (err)
1056 goto restore_opts;
1057
1058 /*
1059 * Previous and new state of filesystem is RO,
1060 * so skip checking GC and FLUSH_MERGE conditions.
1061 */
1062 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1063 goto skip;
1064
1065 /* disallow enable/disable extent_cache dynamically */
1066 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1067 err = -EINVAL;
1068 f2fs_msg(sbi->sb, KERN_WARNING,
1069 "switch extent_cache option is not allowed");
1070 goto restore_opts;
1071 }
1072
1073 /*
1074 * We stop the GC thread if FS is mounted as RO
1075 * or if background_gc = off is passed in mount
1076 * option. Also sync the filesystem.
1077 */
1078 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1079 if (sbi->gc_thread) {
1080 stop_gc_thread(sbi);
1081 need_restart_gc = true;
1082 }
1083 } else if (!sbi->gc_thread) {
1084 err = start_gc_thread(sbi);
1085 if (err)
1086 goto restore_opts;
1087 need_stop_gc = true;
1088 }
1089
1090 if (*flags & MS_RDONLY) {
1091 writeback_inodes_sb(sb, WB_REASON_SYNC);
1092 sync_inodes_sb(sb);
1093
1094 set_sbi_flag(sbi, SBI_IS_DIRTY);
1095 set_sbi_flag(sbi, SBI_IS_CLOSE);
1096 f2fs_sync_fs(sb, 1);
1097 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1098 }
1099
1100 /*
1101 * We stop issue flush thread if FS is mounted as RO
1102 * or if flush_merge is not passed in mount option.
1103 */
1104 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1105 clear_opt(sbi, FLUSH_MERGE);
1106 destroy_flush_cmd_control(sbi, false);
1107 } else {
1108 err = create_flush_cmd_control(sbi);
1109 if (err)
1110 goto restore_gc;
1111 }
1112skip:
1113 /* Update the POSIXACL Flag */
1114 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1115 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1116
1117 return 0;
1118restore_gc:
1119 if (need_restart_gc) {
1120 if (start_gc_thread(sbi))
1121 f2fs_msg(sbi->sb, KERN_WARNING,
1122 "background gc thread has stopped");
1123 } else if (need_stop_gc) {
1124 stop_gc_thread(sbi);
1125 }
1126restore_opts:
1127 sbi->mount_opt = org_mount_opt;
1128 sbi->active_logs = active_logs;
1129#ifdef CONFIG_F2FS_FAULT_INJECTION
1130 sbi->fault_info = ffi;
1131#endif
1132 return err;
1133}
1134
1135static struct super_operations f2fs_sops = {
1136 .alloc_inode = f2fs_alloc_inode,
1137 .drop_inode = f2fs_drop_inode,
1138 .destroy_inode = f2fs_destroy_inode,
1139 .write_inode = f2fs_write_inode,
1140 .dirty_inode = f2fs_dirty_inode,
1141 .show_options = f2fs_show_options,
1142 .evict_inode = f2fs_evict_inode,
1143 .put_super = f2fs_put_super,
1144 .sync_fs = f2fs_sync_fs,
1145 .freeze_fs = f2fs_freeze,
1146 .unfreeze_fs = f2fs_unfreeze,
1147 .statfs = f2fs_statfs,
1148 .remount_fs = f2fs_remount,
1149};
1150
1151#ifdef CONFIG_F2FS_FS_ENCRYPTION
1152static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1153{
1154 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1155 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1156 ctx, len, NULL);
1157}
1158
1159static int f2fs_key_prefix(struct inode *inode, u8 **key)
1160{
1161 *key = F2FS_I_SB(inode)->key_prefix;
1162 return F2FS_I_SB(inode)->key_prefix_size;
1163}
1164
1165static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1166 void *fs_data)
1167{
1168 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1169 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1170 ctx, len, fs_data, XATTR_CREATE);
1171}
1172
1173static unsigned f2fs_max_namelen(struct inode *inode)
1174{
1175 return S_ISLNK(inode->i_mode) ?
1176 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1177}
1178
1179static struct fscrypt_operations f2fs_cryptops = {
1180 .get_context = f2fs_get_context,
1181 .key_prefix = f2fs_key_prefix,
1182 .set_context = f2fs_set_context,
1183 .is_encrypted = f2fs_encrypted_inode,
1184 .empty_dir = f2fs_empty_dir,
1185 .max_namelen = f2fs_max_namelen,
1186};
1187#else
1188static struct fscrypt_operations f2fs_cryptops = {
1189 .is_encrypted = f2fs_encrypted_inode,
1190};
1191#endif
1192
1193static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1194 u64 ino, u32 generation)
1195{
1196 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1197 struct inode *inode;
1198
1199 if (check_nid_range(sbi, ino))
1200 return ERR_PTR(-ESTALE);
1201
1202 /*
1203 * f2fs_iget isn't quite right if the inode is currently unallocated!
1204 * However f2fs_iget currently does appropriate checks to handle stale
1205 * inodes so everything is OK.
1206 */
1207 inode = f2fs_iget(sb, ino);
1208 if (IS_ERR(inode))
1209 return ERR_CAST(inode);
1210 if (unlikely(generation && inode->i_generation != generation)) {
1211 /* we didn't find the right inode.. */
1212 iput(inode);
1213 return ERR_PTR(-ESTALE);
1214 }
1215 return inode;
1216}
1217
1218static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1219 int fh_len, int fh_type)
1220{
1221 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1222 f2fs_nfs_get_inode);
1223}
1224
1225static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1226 int fh_len, int fh_type)
1227{
1228 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1229 f2fs_nfs_get_inode);
1230}
1231
1232static const struct export_operations f2fs_export_ops = {
1233 .fh_to_dentry = f2fs_fh_to_dentry,
1234 .fh_to_parent = f2fs_fh_to_parent,
1235 .get_parent = f2fs_get_parent,
1236};
1237
1238static loff_t max_file_blocks(void)
1239{
1240 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1241 loff_t leaf_count = ADDRS_PER_BLOCK;
1242
1243 /* two direct node blocks */
1244 result += (leaf_count * 2);
1245
1246 /* two indirect node blocks */
1247 leaf_count *= NIDS_PER_BLOCK;
1248 result += (leaf_count * 2);
1249
1250 /* one double indirect node block */
1251 leaf_count *= NIDS_PER_BLOCK;
1252 result += leaf_count;
1253
1254 return result;
1255}
1256
1257static int __f2fs_commit_super(struct buffer_head *bh,
1258 struct f2fs_super_block *super)
1259{
1260 lock_buffer(bh);
1261 if (super)
1262 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1263 set_buffer_uptodate(bh);
1264 set_buffer_dirty(bh);
1265 unlock_buffer(bh);
1266
1267 /* it's rare case, we can do fua all the time */
1268 return __sync_dirty_buffer(bh, REQ_PREFLUSH | REQ_FUA);
1269}
1270
1271static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1272 struct buffer_head *bh)
1273{
1274 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1275 (bh->b_data + F2FS_SUPER_OFFSET);
1276 struct super_block *sb = sbi->sb;
1277 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1278 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1279 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1280 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1281 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1282 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1283 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1284 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1285 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1286 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1287 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1288 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1289 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1290 u64 main_end_blkaddr = main_blkaddr +
1291 (segment_count_main << log_blocks_per_seg);
1292 u64 seg_end_blkaddr = segment0_blkaddr +
1293 (segment_count << log_blocks_per_seg);
1294
1295 if (segment0_blkaddr != cp_blkaddr) {
1296 f2fs_msg(sb, KERN_INFO,
1297 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1298 segment0_blkaddr, cp_blkaddr);
1299 return true;
1300 }
1301
1302 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1303 sit_blkaddr) {
1304 f2fs_msg(sb, KERN_INFO,
1305 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1306 cp_blkaddr, sit_blkaddr,
1307 segment_count_ckpt << log_blocks_per_seg);
1308 return true;
1309 }
1310
1311 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1312 nat_blkaddr) {
1313 f2fs_msg(sb, KERN_INFO,
1314 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1315 sit_blkaddr, nat_blkaddr,
1316 segment_count_sit << log_blocks_per_seg);
1317 return true;
1318 }
1319
1320 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1321 ssa_blkaddr) {
1322 f2fs_msg(sb, KERN_INFO,
1323 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1324 nat_blkaddr, ssa_blkaddr,
1325 segment_count_nat << log_blocks_per_seg);
1326 return true;
1327 }
1328
1329 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1330 main_blkaddr) {
1331 f2fs_msg(sb, KERN_INFO,
1332 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1333 ssa_blkaddr, main_blkaddr,
1334 segment_count_ssa << log_blocks_per_seg);
1335 return true;
1336 }
1337
1338 if (main_end_blkaddr > seg_end_blkaddr) {
1339 f2fs_msg(sb, KERN_INFO,
1340 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1341 main_blkaddr,
1342 segment0_blkaddr +
1343 (segment_count << log_blocks_per_seg),
1344 segment_count_main << log_blocks_per_seg);
1345 return true;
1346 } else if (main_end_blkaddr < seg_end_blkaddr) {
1347 int err = 0;
1348 char *res;
1349
1350 /* fix in-memory information all the time */
1351 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1352 segment0_blkaddr) >> log_blocks_per_seg);
1353
1354 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1355 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1356 res = "internally";
1357 } else {
1358 err = __f2fs_commit_super(bh, NULL);
1359 res = err ? "failed" : "done";
1360 }
1361 f2fs_msg(sb, KERN_INFO,
1362 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1363 res, main_blkaddr,
1364 segment0_blkaddr +
1365 (segment_count << log_blocks_per_seg),
1366 segment_count_main << log_blocks_per_seg);
1367 if (err)
1368 return true;
1369 }
1370 return false;
1371}
1372
1373static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1374 struct buffer_head *bh)
1375{
1376 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1377 (bh->b_data + F2FS_SUPER_OFFSET);
1378 struct super_block *sb = sbi->sb;
1379 unsigned int blocksize;
1380
1381 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1382 f2fs_msg(sb, KERN_INFO,
1383 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1384 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1385 return 1;
1386 }
1387
1388 /* Currently, support only 4KB page cache size */
1389 if (F2FS_BLKSIZE != PAGE_SIZE) {
1390 f2fs_msg(sb, KERN_INFO,
1391 "Invalid page_cache_size (%lu), supports only 4KB\n",
1392 PAGE_SIZE);
1393 return 1;
1394 }
1395
1396 /* Currently, support only 4KB block size */
1397 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1398 if (blocksize != F2FS_BLKSIZE) {
1399 f2fs_msg(sb, KERN_INFO,
1400 "Invalid blocksize (%u), supports only 4KB\n",
1401 blocksize);
1402 return 1;
1403 }
1404
1405 /* check log blocks per segment */
1406 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1407 f2fs_msg(sb, KERN_INFO,
1408 "Invalid log blocks per segment (%u)\n",
1409 le32_to_cpu(raw_super->log_blocks_per_seg));
1410 return 1;
1411 }
1412
1413 /* Currently, support 512/1024/2048/4096 bytes sector size */
1414 if (le32_to_cpu(raw_super->log_sectorsize) >
1415 F2FS_MAX_LOG_SECTOR_SIZE ||
1416 le32_to_cpu(raw_super->log_sectorsize) <
1417 F2FS_MIN_LOG_SECTOR_SIZE) {
1418 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1419 le32_to_cpu(raw_super->log_sectorsize));
1420 return 1;
1421 }
1422 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1423 le32_to_cpu(raw_super->log_sectorsize) !=
1424 F2FS_MAX_LOG_SECTOR_SIZE) {
1425 f2fs_msg(sb, KERN_INFO,
1426 "Invalid log sectors per block(%u) log sectorsize(%u)",
1427 le32_to_cpu(raw_super->log_sectors_per_block),
1428 le32_to_cpu(raw_super->log_sectorsize));
1429 return 1;
1430 }
1431
1432 /* check reserved ino info */
1433 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1434 le32_to_cpu(raw_super->meta_ino) != 2 ||
1435 le32_to_cpu(raw_super->root_ino) != 3) {
1436 f2fs_msg(sb, KERN_INFO,
1437 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1438 le32_to_cpu(raw_super->node_ino),
1439 le32_to_cpu(raw_super->meta_ino),
1440 le32_to_cpu(raw_super->root_ino));
1441 return 1;
1442 }
1443
1444 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1445 if (sanity_check_area_boundary(sbi, bh))
1446 return 1;
1447
1448 return 0;
1449}
1450
1451int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1452{
1453 unsigned int total, fsmeta;
1454 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1455 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1456 unsigned int ovp_segments, reserved_segments;
1457
1458 total = le32_to_cpu(raw_super->segment_count);
1459 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1460 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1461 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1462 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1463 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1464
1465 if (unlikely(fsmeta >= total))
1466 return 1;
1467
1468 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1469 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1470
1471 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1472 ovp_segments == 0 || reserved_segments == 0)) {
1473 f2fs_msg(sbi->sb, KERN_ERR,
1474 "Wrong layout: check mkfs.f2fs version");
1475 return 1;
1476 }
1477
1478 if (unlikely(f2fs_cp_error(sbi))) {
1479 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1480 return 1;
1481 }
1482 return 0;
1483}
1484
1485static void init_sb_info(struct f2fs_sb_info *sbi)
1486{
1487 struct f2fs_super_block *raw_super = sbi->raw_super;
1488 int i;
1489
1490 sbi->log_sectors_per_block =
1491 le32_to_cpu(raw_super->log_sectors_per_block);
1492 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1493 sbi->blocksize = 1 << sbi->log_blocksize;
1494 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1495 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1496 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1497 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1498 sbi->total_sections = le32_to_cpu(raw_super->section_count);
1499 sbi->total_node_count =
1500 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1501 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1502 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1503 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1504 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1505 sbi->cur_victim_sec = NULL_SECNO;
1506 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1507
1508 sbi->dir_level = DEF_DIR_LEVEL;
1509 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1510 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1511 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1512
1513 for (i = 0; i < NR_COUNT_TYPE; i++)
1514 atomic_set(&sbi->nr_pages[i], 0);
1515
1516 INIT_LIST_HEAD(&sbi->s_list);
1517 mutex_init(&sbi->umount_mutex);
1518 mutex_init(&sbi->wio_mutex[NODE]);
1519 mutex_init(&sbi->wio_mutex[DATA]);
1520 spin_lock_init(&sbi->cp_lock);
1521
1522#ifdef CONFIG_F2FS_FS_ENCRYPTION
1523 memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1524 F2FS_KEY_DESC_PREFIX_SIZE);
1525 sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1526#endif
1527}
1528
1529static int init_percpu_info(struct f2fs_sb_info *sbi)
1530{
1531 int err;
1532
1533 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1534 if (err)
1535 return err;
1536
1537 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1538 GFP_KERNEL);
1539}
1540
1541#ifdef CONFIG_BLK_DEV_ZONED
1542static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
1543{
1544 struct block_device *bdev = FDEV(devi).bdev;
1545 sector_t nr_sectors = bdev->bd_part->nr_sects;
1546 sector_t sector = 0;
1547 struct blk_zone *zones;
1548 unsigned int i, nr_zones;
1549 unsigned int n = 0;
1550 int err = -EIO;
1551
1552 if (!f2fs_sb_mounted_blkzoned(sbi->sb))
1553 return 0;
1554
1555 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
1556 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
1557 return -EINVAL;
1558 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
1559 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
1560 __ilog2_u32(sbi->blocks_per_blkz))
1561 return -EINVAL;
1562 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
1563 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
1564 sbi->log_blocks_per_blkz;
1565 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
1566 FDEV(devi).nr_blkz++;
1567
1568 FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
1569 if (!FDEV(devi).blkz_type)
1570 return -ENOMEM;
1571
1572#define F2FS_REPORT_NR_ZONES 4096
1573
1574 zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
1575 GFP_KERNEL);
1576 if (!zones)
1577 return -ENOMEM;
1578
1579 /* Get block zones type */
1580 while (zones && sector < nr_sectors) {
1581
1582 nr_zones = F2FS_REPORT_NR_ZONES;
1583 err = blkdev_report_zones(bdev, sector,
1584 zones, &nr_zones,
1585 GFP_KERNEL);
1586 if (err)
1587 break;
1588 if (!nr_zones) {
1589 err = -EIO;
1590 break;
1591 }
1592
1593 for (i = 0; i < nr_zones; i++) {
1594 FDEV(devi).blkz_type[n] = zones[i].type;
1595 sector += zones[i].len;
1596 n++;
1597 }
1598 }
1599
1600 kfree(zones);
1601
1602 return err;
1603}
1604#endif
1605
1606/*
1607 * Read f2fs raw super block.
1608 * Because we have two copies of super block, so read both of them
1609 * to get the first valid one. If any one of them is broken, we pass
1610 * them recovery flag back to the caller.
1611 */
1612static int read_raw_super_block(struct f2fs_sb_info *sbi,
1613 struct f2fs_super_block **raw_super,
1614 int *valid_super_block, int *recovery)
1615{
1616 struct super_block *sb = sbi->sb;
1617 int block;
1618 struct buffer_head *bh;
1619 struct f2fs_super_block *super;
1620 int err = 0;
1621
1622 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1623 if (!super)
1624 return -ENOMEM;
1625
1626 for (block = 0; block < 2; block++) {
1627 bh = sb_bread(sb, block);
1628 if (!bh) {
1629 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1630 block + 1);
1631 err = -EIO;
1632 continue;
1633 }
1634
1635 /* sanity checking of raw super */
1636 if (sanity_check_raw_super(sbi, bh)) {
1637 f2fs_msg(sb, KERN_ERR,
1638 "Can't find valid F2FS filesystem in %dth superblock",
1639 block + 1);
1640 err = -EINVAL;
1641 brelse(bh);
1642 continue;
1643 }
1644
1645 if (!*raw_super) {
1646 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1647 sizeof(*super));
1648 *valid_super_block = block;
1649 *raw_super = super;
1650 }
1651 brelse(bh);
1652 }
1653
1654 /* Fail to read any one of the superblocks*/
1655 if (err < 0)
1656 *recovery = 1;
1657
1658 /* No valid superblock */
1659 if (!*raw_super)
1660 kfree(super);
1661 else
1662 err = 0;
1663
1664 return err;
1665}
1666
1667int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1668{
1669 struct buffer_head *bh;
1670 int err;
1671
1672 if ((recover && f2fs_readonly(sbi->sb)) ||
1673 bdev_read_only(sbi->sb->s_bdev)) {
1674 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1675 return -EROFS;
1676 }
1677
1678 /* write back-up superblock first */
1679 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1680 if (!bh)
1681 return -EIO;
1682 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1683 brelse(bh);
1684
1685 /* if we are in recovery path, skip writing valid superblock */
1686 if (recover || err)
1687 return err;
1688
1689 /* write current valid superblock */
1690 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1691 if (!bh)
1692 return -EIO;
1693 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1694 brelse(bh);
1695 return err;
1696}
1697
1698static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
1699{
1700 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1701 unsigned int max_devices = MAX_DEVICES;
1702 int i;
1703
1704 /* Initialize single device information */
1705 if (!RDEV(0).path[0]) {
1706 if (!bdev_is_zoned(sbi->sb->s_bdev))
1707 return 0;
1708 max_devices = 1;
1709 }
1710
1711 /*
1712 * Initialize multiple devices information, or single
1713 * zoned block device information.
1714 */
1715 sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info),
1716 GFP_KERNEL);
1717 if (!sbi->devs)
1718 return -ENOMEM;
1719
1720 for (i = 0; i < max_devices; i++) {
1721
1722 if (i > 0 && !RDEV(i).path[0])
1723 break;
1724
1725 if (max_devices == 1) {
1726 /* Single zoned block device mount */
1727 FDEV(0).bdev =
1728 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
1729 sbi->sb->s_mode, sbi->sb->s_type);
1730 } else {
1731 /* Multi-device mount */
1732 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
1733 FDEV(i).total_segments =
1734 le32_to_cpu(RDEV(i).total_segments);
1735 if (i == 0) {
1736 FDEV(i).start_blk = 0;
1737 FDEV(i).end_blk = FDEV(i).start_blk +
1738 (FDEV(i).total_segments <<
1739 sbi->log_blocks_per_seg) - 1 +
1740 le32_to_cpu(raw_super->segment0_blkaddr);
1741 } else {
1742 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
1743 FDEV(i).end_blk = FDEV(i).start_blk +
1744 (FDEV(i).total_segments <<
1745 sbi->log_blocks_per_seg) - 1;
1746 }
1747 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
1748 sbi->sb->s_mode, sbi->sb->s_type);
1749 }
1750 if (IS_ERR(FDEV(i).bdev))
1751 return PTR_ERR(FDEV(i).bdev);
1752
1753 /* to release errored devices */
1754 sbi->s_ndevs = i + 1;
1755
1756#ifdef CONFIG_BLK_DEV_ZONED
1757 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
1758 !f2fs_sb_mounted_blkzoned(sbi->sb)) {
1759 f2fs_msg(sbi->sb, KERN_ERR,
1760 "Zoned block device feature not enabled\n");
1761 return -EINVAL;
1762 }
1763 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
1764 if (init_blkz_info(sbi, i)) {
1765 f2fs_msg(sbi->sb, KERN_ERR,
1766 "Failed to initialize F2FS blkzone information");
1767 return -EINVAL;
1768 }
1769 if (max_devices == 1)
1770 break;
1771 f2fs_msg(sbi->sb, KERN_INFO,
1772 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
1773 i, FDEV(i).path,
1774 FDEV(i).total_segments,
1775 FDEV(i).start_blk, FDEV(i).end_blk,
1776 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
1777 "Host-aware" : "Host-managed");
1778 continue;
1779 }
1780#endif
1781 f2fs_msg(sbi->sb, KERN_INFO,
1782 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
1783 i, FDEV(i).path,
1784 FDEV(i).total_segments,
1785 FDEV(i).start_blk, FDEV(i).end_blk);
1786 }
1787 return 0;
1788}
1789
1790static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1791{
1792 struct f2fs_sb_info *sbi;
1793 struct f2fs_super_block *raw_super;
1794 struct inode *root;
1795 int err;
1796 bool retry = true, need_fsck = false;
1797 char *options = NULL;
1798 int recovery, i, valid_super_block;
1799 struct curseg_info *seg_i;
1800
1801try_onemore:
1802 err = -EINVAL;
1803 raw_super = NULL;
1804 valid_super_block = -1;
1805 recovery = 0;
1806
1807 /* allocate memory for f2fs-specific super block info */
1808 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1809 if (!sbi)
1810 return -ENOMEM;
1811
1812 sbi->sb = sb;
1813
1814 /* Load the checksum driver */
1815 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1816 if (IS_ERR(sbi->s_chksum_driver)) {
1817 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1818 err = PTR_ERR(sbi->s_chksum_driver);
1819 sbi->s_chksum_driver = NULL;
1820 goto free_sbi;
1821 }
1822
1823 /* set a block size */
1824 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1825 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1826 goto free_sbi;
1827 }
1828
1829 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1830 &recovery);
1831 if (err)
1832 goto free_sbi;
1833
1834 sb->s_fs_info = sbi;
1835 sbi->raw_super = raw_super;
1836
1837 /*
1838 * The BLKZONED feature indicates that the drive was formatted with
1839 * zone alignment optimization. This is optional for host-aware
1840 * devices, but mandatory for host-managed zoned block devices.
1841 */
1842#ifndef CONFIG_BLK_DEV_ZONED
1843 if (f2fs_sb_mounted_blkzoned(sb)) {
1844 f2fs_msg(sb, KERN_ERR,
1845 "Zoned block device support is not enabled\n");
1846 goto free_sb_buf;
1847 }
1848#endif
1849 default_options(sbi);
1850 /* parse mount options */
1851 options = kstrdup((const char *)data, GFP_KERNEL);
1852 if (data && !options) {
1853 err = -ENOMEM;
1854 goto free_sb_buf;
1855 }
1856
1857 err = parse_options(sb, options);
1858 if (err)
1859 goto free_options;
1860
1861 sbi->max_file_blocks = max_file_blocks();
1862 sb->s_maxbytes = sbi->max_file_blocks <<
1863 le32_to_cpu(raw_super->log_blocksize);
1864 sb->s_max_links = F2FS_LINK_MAX;
1865 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1866
1867 sb->s_op = &f2fs_sops;
1868 sb->s_cop = &f2fs_cryptops;
1869 sb->s_xattr = f2fs_xattr_handlers;
1870 sb->s_export_op = &f2fs_export_ops;
1871 sb->s_magic = F2FS_SUPER_MAGIC;
1872 sb->s_time_gran = 1;
1873 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1874 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1875 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1876
1877 /* init f2fs-specific super block info */
1878 sbi->valid_super_block = valid_super_block;
1879 mutex_init(&sbi->gc_mutex);
1880 mutex_init(&sbi->cp_mutex);
1881 init_rwsem(&sbi->node_write);
1882
1883 /* disallow all the data/node/meta page writes */
1884 set_sbi_flag(sbi, SBI_POR_DOING);
1885 spin_lock_init(&sbi->stat_lock);
1886
1887 init_rwsem(&sbi->read_io.io_rwsem);
1888 sbi->read_io.sbi = sbi;
1889 sbi->read_io.bio = NULL;
1890 for (i = 0; i < NR_PAGE_TYPE; i++) {
1891 init_rwsem(&sbi->write_io[i].io_rwsem);
1892 sbi->write_io[i].sbi = sbi;
1893 sbi->write_io[i].bio = NULL;
1894 }
1895
1896 init_rwsem(&sbi->cp_rwsem);
1897 init_waitqueue_head(&sbi->cp_wait);
1898 init_sb_info(sbi);
1899
1900 err = init_percpu_info(sbi);
1901 if (err)
1902 goto free_options;
1903
1904 /* get an inode for meta space */
1905 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1906 if (IS_ERR(sbi->meta_inode)) {
1907 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1908 err = PTR_ERR(sbi->meta_inode);
1909 goto free_options;
1910 }
1911
1912 err = get_valid_checkpoint(sbi);
1913 if (err) {
1914 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1915 goto free_meta_inode;
1916 }
1917
1918 /* Initialize device list */
1919 err = f2fs_scan_devices(sbi);
1920 if (err) {
1921 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
1922 goto free_devices;
1923 }
1924
1925 sbi->total_valid_node_count =
1926 le32_to_cpu(sbi->ckpt->valid_node_count);
1927 percpu_counter_set(&sbi->total_valid_inode_count,
1928 le32_to_cpu(sbi->ckpt->valid_inode_count));
1929 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1930 sbi->total_valid_block_count =
1931 le64_to_cpu(sbi->ckpt->valid_block_count);
1932 sbi->last_valid_block_count = sbi->total_valid_block_count;
1933
1934 for (i = 0; i < NR_INODE_TYPE; i++) {
1935 INIT_LIST_HEAD(&sbi->inode_list[i]);
1936 spin_lock_init(&sbi->inode_lock[i]);
1937 }
1938
1939 init_extent_cache_info(sbi);
1940
1941 init_ino_entry_info(sbi);
1942
1943 /* setup f2fs internal modules */
1944 err = build_segment_manager(sbi);
1945 if (err) {
1946 f2fs_msg(sb, KERN_ERR,
1947 "Failed to initialize F2FS segment manager");
1948 goto free_sm;
1949 }
1950 err = build_node_manager(sbi);
1951 if (err) {
1952 f2fs_msg(sb, KERN_ERR,
1953 "Failed to initialize F2FS node manager");
1954 goto free_nm;
1955 }
1956
1957 /* For write statistics */
1958 if (sb->s_bdev->bd_part)
1959 sbi->sectors_written_start =
1960 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1961
1962 /* Read accumulated write IO statistics if exists */
1963 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1964 if (__exist_node_summaries(sbi))
1965 sbi->kbytes_written =
1966 le64_to_cpu(seg_i->journal->info.kbytes_written);
1967
1968 build_gc_manager(sbi);
1969
1970 /* get an inode for node space */
1971 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1972 if (IS_ERR(sbi->node_inode)) {
1973 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1974 err = PTR_ERR(sbi->node_inode);
1975 goto free_nm;
1976 }
1977
1978 f2fs_join_shrinker(sbi);
1979
1980 /* if there are nt orphan nodes free them */
1981 err = recover_orphan_inodes(sbi);
1982 if (err)
1983 goto free_node_inode;
1984
1985 /* read root inode and dentry */
1986 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1987 if (IS_ERR(root)) {
1988 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1989 err = PTR_ERR(root);
1990 goto free_node_inode;
1991 }
1992 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1993 iput(root);
1994 err = -EINVAL;
1995 goto free_node_inode;
1996 }
1997
1998 sb->s_root = d_make_root(root); /* allocate root dentry */
1999 if (!sb->s_root) {
2000 err = -ENOMEM;
2001 goto free_root_inode;
2002 }
2003
2004 err = f2fs_build_stats(sbi);
2005 if (err)
2006 goto free_root_inode;
2007
2008 if (f2fs_proc_root)
2009 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
2010
2011 if (sbi->s_proc) {
2012 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
2013 &f2fs_seq_segment_info_fops, sb);
2014 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
2015 &f2fs_seq_segment_bits_fops, sb);
2016 }
2017
2018 sbi->s_kobj.kset = f2fs_kset;
2019 init_completion(&sbi->s_kobj_unregister);
2020 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
2021 "%s", sb->s_id);
2022 if (err)
2023 goto free_proc;
2024
2025 /* recover fsynced data */
2026 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2027 /*
2028 * mount should be failed, when device has readonly mode, and
2029 * previous checkpoint was not done by clean system shutdown.
2030 */
2031 if (bdev_read_only(sb->s_bdev) &&
2032 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2033 err = -EROFS;
2034 goto free_kobj;
2035 }
2036
2037 if (need_fsck)
2038 set_sbi_flag(sbi, SBI_NEED_FSCK);
2039
2040 if (!retry)
2041 goto skip_recovery;
2042
2043 err = recover_fsync_data(sbi, false);
2044 if (err < 0) {
2045 need_fsck = true;
2046 f2fs_msg(sb, KERN_ERR,
2047 "Cannot recover all fsync data errno=%d", err);
2048 goto free_kobj;
2049 }
2050 } else {
2051 err = recover_fsync_data(sbi, true);
2052
2053 if (!f2fs_readonly(sb) && err > 0) {
2054 err = -EINVAL;
2055 f2fs_msg(sb, KERN_ERR,
2056 "Need to recover fsync data");
2057 goto free_kobj;
2058 }
2059 }
2060skip_recovery:
2061 /* recover_fsync_data() cleared this already */
2062 clear_sbi_flag(sbi, SBI_POR_DOING);
2063
2064 /*
2065 * If filesystem is not mounted as read-only then
2066 * do start the gc_thread.
2067 */
2068 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2069 /* After POR, we can run background GC thread.*/
2070 err = start_gc_thread(sbi);
2071 if (err)
2072 goto free_kobj;
2073 }
2074 kfree(options);
2075
2076 /* recover broken superblock */
2077 if (recovery) {
2078 err = f2fs_commit_super(sbi, true);
2079 f2fs_msg(sb, KERN_INFO,
2080 "Try to recover %dth superblock, ret: %d",
2081 sbi->valid_super_block ? 1 : 2, err);
2082 }
2083
2084 f2fs_update_time(sbi, CP_TIME);
2085 f2fs_update_time(sbi, REQ_TIME);
2086 return 0;
2087
2088free_kobj:
2089 f2fs_sync_inode_meta(sbi);
2090 kobject_del(&sbi->s_kobj);
2091 kobject_put(&sbi->s_kobj);
2092 wait_for_completion(&sbi->s_kobj_unregister);
2093free_proc:
2094 if (sbi->s_proc) {
2095 remove_proc_entry("segment_info", sbi->s_proc);
2096 remove_proc_entry("segment_bits", sbi->s_proc);
2097 remove_proc_entry(sb->s_id, f2fs_proc_root);
2098 }
2099 f2fs_destroy_stats(sbi);
2100free_root_inode:
2101 dput(sb->s_root);
2102 sb->s_root = NULL;
2103free_node_inode:
2104 truncate_inode_pages_final(NODE_MAPPING(sbi));
2105 mutex_lock(&sbi->umount_mutex);
2106 release_ino_entry(sbi, true);
2107 f2fs_leave_shrinker(sbi);
2108 /*
2109 * Some dirty meta pages can be produced by recover_orphan_inodes()
2110 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2111 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2112 * falls into an infinite loop in sync_meta_pages().
2113 */
2114 truncate_inode_pages_final(META_MAPPING(sbi));
2115 iput(sbi->node_inode);
2116 mutex_unlock(&sbi->umount_mutex);
2117free_nm:
2118 destroy_node_manager(sbi);
2119free_sm:
2120 destroy_segment_manager(sbi);
2121free_devices:
2122 destroy_device_list(sbi);
2123 kfree(sbi->ckpt);
2124free_meta_inode:
2125 make_bad_inode(sbi->meta_inode);
2126 iput(sbi->meta_inode);
2127free_options:
2128 destroy_percpu_info(sbi);
2129 kfree(options);
2130free_sb_buf:
2131 kfree(raw_super);
2132free_sbi:
2133 if (sbi->s_chksum_driver)
2134 crypto_free_shash(sbi->s_chksum_driver);
2135 kfree(sbi);
2136
2137 /* give only one another chance */
2138 if (retry) {
2139 retry = false;
2140 shrink_dcache_sb(sb);
2141 goto try_onemore;
2142 }
2143 return err;
2144}
2145
2146static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2147 const char *dev_name, void *data)
2148{
2149 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2150}
2151
2152static void kill_f2fs_super(struct super_block *sb)
2153{
2154 if (sb->s_root)
2155 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2156 kill_block_super(sb);
2157}
2158
2159static struct file_system_type f2fs_fs_type = {
2160 .owner = THIS_MODULE,
2161 .name = "f2fs",
2162 .mount = f2fs_mount,
2163 .kill_sb = kill_f2fs_super,
2164 .fs_flags = FS_REQUIRES_DEV,
2165};
2166MODULE_ALIAS_FS("f2fs");
2167
2168static int __init init_inodecache(void)
2169{
2170 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2171 sizeof(struct f2fs_inode_info), 0,
2172 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2173 if (!f2fs_inode_cachep)
2174 return -ENOMEM;
2175 return 0;
2176}
2177
2178static void destroy_inodecache(void)
2179{
2180 /*
2181 * Make sure all delayed rcu free inodes are flushed before we
2182 * destroy cache.
2183 */
2184 rcu_barrier();
2185 kmem_cache_destroy(f2fs_inode_cachep);
2186}
2187
2188static int __init init_f2fs_fs(void)
2189{
2190 int err;
2191
2192 f2fs_build_trace_ios();
2193
2194 err = init_inodecache();
2195 if (err)
2196 goto fail;
2197 err = create_node_manager_caches();
2198 if (err)
2199 goto free_inodecache;
2200 err = create_segment_manager_caches();
2201 if (err)
2202 goto free_node_manager_caches;
2203 err = create_checkpoint_caches();
2204 if (err)
2205 goto free_segment_manager_caches;
2206 err = create_extent_cache();
2207 if (err)
2208 goto free_checkpoint_caches;
2209 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
2210 if (!f2fs_kset) {
2211 err = -ENOMEM;
2212 goto free_extent_cache;
2213 }
2214 err = register_shrinker(&f2fs_shrinker_info);
2215 if (err)
2216 goto free_kset;
2217
2218 err = register_filesystem(&f2fs_fs_type);
2219 if (err)
2220 goto free_shrinker;
2221 err = f2fs_create_root_stats();
2222 if (err)
2223 goto free_filesystem;
2224 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2225 return 0;
2226
2227free_filesystem:
2228 unregister_filesystem(&f2fs_fs_type);
2229free_shrinker:
2230 unregister_shrinker(&f2fs_shrinker_info);
2231free_kset:
2232 kset_unregister(f2fs_kset);
2233free_extent_cache:
2234 destroy_extent_cache();
2235free_checkpoint_caches:
2236 destroy_checkpoint_caches();
2237free_segment_manager_caches:
2238 destroy_segment_manager_caches();
2239free_node_manager_caches:
2240 destroy_node_manager_caches();
2241free_inodecache:
2242 destroy_inodecache();
2243fail:
2244 return err;
2245}
2246
2247static void __exit exit_f2fs_fs(void)
2248{
2249 remove_proc_entry("fs/f2fs", NULL);
2250 f2fs_destroy_root_stats();
2251 unregister_filesystem(&f2fs_fs_type);
2252 unregister_shrinker(&f2fs_shrinker_info);
2253 kset_unregister(f2fs_kset);
2254 destroy_extent_cache();
2255 destroy_checkpoint_caches();
2256 destroy_segment_manager_caches();
2257 destroy_node_manager_caches();
2258 destroy_inodecache();
2259 f2fs_destroy_trace_ios();
2260}
2261
2262module_init(init_f2fs_fs)
2263module_exit(exit_f2fs_fs)
2264
2265MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2266MODULE_DESCRIPTION("Flash Friendly File System");
2267MODULE_LICENSE("GPL");
2268
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * fs/f2fs/super.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8#include <linux/module.h>
9#include <linux/init.h>
10#include <linux/fs.h>
11#include <linux/statfs.h>
12#include <linux/buffer_head.h>
13#include <linux/backing-dev.h>
14#include <linux/kthread.h>
15#include <linux/parser.h>
16#include <linux/mount.h>
17#include <linux/seq_file.h>
18#include <linux/proc_fs.h>
19#include <linux/random.h>
20#include <linux/exportfs.h>
21#include <linux/blkdev.h>
22#include <linux/quotaops.h>
23#include <linux/f2fs_fs.h>
24#include <linux/sysfs.h>
25#include <linux/quota.h>
26#include <linux/unicode.h>
27
28#include "f2fs.h"
29#include "node.h"
30#include "segment.h"
31#include "xattr.h"
32#include "gc.h"
33#include "trace.h"
34
35#define CREATE_TRACE_POINTS
36#include <trace/events/f2fs.h>
37
38static struct kmem_cache *f2fs_inode_cachep;
39
40#ifdef CONFIG_F2FS_FAULT_INJECTION
41
42const char *f2fs_fault_name[FAULT_MAX] = {
43 [FAULT_KMALLOC] = "kmalloc",
44 [FAULT_KVMALLOC] = "kvmalloc",
45 [FAULT_PAGE_ALLOC] = "page alloc",
46 [FAULT_PAGE_GET] = "page get",
47 [FAULT_ALLOC_BIO] = "alloc bio",
48 [FAULT_ALLOC_NID] = "alloc nid",
49 [FAULT_ORPHAN] = "orphan",
50 [FAULT_BLOCK] = "no more block",
51 [FAULT_DIR_DEPTH] = "too big dir depth",
52 [FAULT_EVICT_INODE] = "evict_inode fail",
53 [FAULT_TRUNCATE] = "truncate fail",
54 [FAULT_READ_IO] = "read IO error",
55 [FAULT_CHECKPOINT] = "checkpoint error",
56 [FAULT_DISCARD] = "discard error",
57 [FAULT_WRITE_IO] = "write IO error",
58};
59
60void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
61 unsigned int type)
62{
63 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
64
65 if (rate) {
66 atomic_set(&ffi->inject_ops, 0);
67 ffi->inject_rate = rate;
68 }
69
70 if (type)
71 ffi->inject_type = type;
72
73 if (!rate && !type)
74 memset(ffi, 0, sizeof(struct f2fs_fault_info));
75}
76#endif
77
78/* f2fs-wide shrinker description */
79static struct shrinker f2fs_shrinker_info = {
80 .scan_objects = f2fs_shrink_scan,
81 .count_objects = f2fs_shrink_count,
82 .seeks = DEFAULT_SEEKS,
83};
84
85enum {
86 Opt_gc_background,
87 Opt_disable_roll_forward,
88 Opt_norecovery,
89 Opt_discard,
90 Opt_nodiscard,
91 Opt_noheap,
92 Opt_heap,
93 Opt_user_xattr,
94 Opt_nouser_xattr,
95 Opt_acl,
96 Opt_noacl,
97 Opt_active_logs,
98 Opt_disable_ext_identify,
99 Opt_inline_xattr,
100 Opt_noinline_xattr,
101 Opt_inline_xattr_size,
102 Opt_inline_data,
103 Opt_inline_dentry,
104 Opt_noinline_dentry,
105 Opt_flush_merge,
106 Opt_noflush_merge,
107 Opt_nobarrier,
108 Opt_fastboot,
109 Opt_extent_cache,
110 Opt_noextent_cache,
111 Opt_noinline_data,
112 Opt_data_flush,
113 Opt_reserve_root,
114 Opt_resgid,
115 Opt_resuid,
116 Opt_mode,
117 Opt_io_size_bits,
118 Opt_fault_injection,
119 Opt_fault_type,
120 Opt_lazytime,
121 Opt_nolazytime,
122 Opt_quota,
123 Opt_noquota,
124 Opt_usrquota,
125 Opt_grpquota,
126 Opt_prjquota,
127 Opt_usrjquota,
128 Opt_grpjquota,
129 Opt_prjjquota,
130 Opt_offusrjquota,
131 Opt_offgrpjquota,
132 Opt_offprjjquota,
133 Opt_jqfmt_vfsold,
134 Opt_jqfmt_vfsv0,
135 Opt_jqfmt_vfsv1,
136 Opt_whint,
137 Opt_alloc,
138 Opt_fsync,
139 Opt_test_dummy_encryption,
140 Opt_checkpoint_disable,
141 Opt_checkpoint_disable_cap,
142 Opt_checkpoint_disable_cap_perc,
143 Opt_checkpoint_enable,
144 Opt_err,
145};
146
147static match_table_t f2fs_tokens = {
148 {Opt_gc_background, "background_gc=%s"},
149 {Opt_disable_roll_forward, "disable_roll_forward"},
150 {Opt_norecovery, "norecovery"},
151 {Opt_discard, "discard"},
152 {Opt_nodiscard, "nodiscard"},
153 {Opt_noheap, "no_heap"},
154 {Opt_heap, "heap"},
155 {Opt_user_xattr, "user_xattr"},
156 {Opt_nouser_xattr, "nouser_xattr"},
157 {Opt_acl, "acl"},
158 {Opt_noacl, "noacl"},
159 {Opt_active_logs, "active_logs=%u"},
160 {Opt_disable_ext_identify, "disable_ext_identify"},
161 {Opt_inline_xattr, "inline_xattr"},
162 {Opt_noinline_xattr, "noinline_xattr"},
163 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
164 {Opt_inline_data, "inline_data"},
165 {Opt_inline_dentry, "inline_dentry"},
166 {Opt_noinline_dentry, "noinline_dentry"},
167 {Opt_flush_merge, "flush_merge"},
168 {Opt_noflush_merge, "noflush_merge"},
169 {Opt_nobarrier, "nobarrier"},
170 {Opt_fastboot, "fastboot"},
171 {Opt_extent_cache, "extent_cache"},
172 {Opt_noextent_cache, "noextent_cache"},
173 {Opt_noinline_data, "noinline_data"},
174 {Opt_data_flush, "data_flush"},
175 {Opt_reserve_root, "reserve_root=%u"},
176 {Opt_resgid, "resgid=%u"},
177 {Opt_resuid, "resuid=%u"},
178 {Opt_mode, "mode=%s"},
179 {Opt_io_size_bits, "io_bits=%u"},
180 {Opt_fault_injection, "fault_injection=%u"},
181 {Opt_fault_type, "fault_type=%u"},
182 {Opt_lazytime, "lazytime"},
183 {Opt_nolazytime, "nolazytime"},
184 {Opt_quota, "quota"},
185 {Opt_noquota, "noquota"},
186 {Opt_usrquota, "usrquota"},
187 {Opt_grpquota, "grpquota"},
188 {Opt_prjquota, "prjquota"},
189 {Opt_usrjquota, "usrjquota=%s"},
190 {Opt_grpjquota, "grpjquota=%s"},
191 {Opt_prjjquota, "prjjquota=%s"},
192 {Opt_offusrjquota, "usrjquota="},
193 {Opt_offgrpjquota, "grpjquota="},
194 {Opt_offprjjquota, "prjjquota="},
195 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
196 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
197 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
198 {Opt_whint, "whint_mode=%s"},
199 {Opt_alloc, "alloc_mode=%s"},
200 {Opt_fsync, "fsync_mode=%s"},
201 {Opt_test_dummy_encryption, "test_dummy_encryption"},
202 {Opt_checkpoint_disable, "checkpoint=disable"},
203 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
204 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
205 {Opt_checkpoint_enable, "checkpoint=enable"},
206 {Opt_err, NULL},
207};
208
209void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
210{
211 struct va_format vaf;
212 va_list args;
213 int level;
214
215 va_start(args, fmt);
216
217 level = printk_get_level(fmt);
218 vaf.fmt = printk_skip_level(fmt);
219 vaf.va = &args;
220 printk("%c%cF2FS-fs (%s): %pV\n",
221 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
222
223 va_end(args);
224}
225
226#ifdef CONFIG_UNICODE
227static const struct f2fs_sb_encodings {
228 __u16 magic;
229 char *name;
230 char *version;
231} f2fs_sb_encoding_map[] = {
232 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
233};
234
235static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
236 const struct f2fs_sb_encodings **encoding,
237 __u16 *flags)
238{
239 __u16 magic = le16_to_cpu(sb->s_encoding);
240 int i;
241
242 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
243 if (magic == f2fs_sb_encoding_map[i].magic)
244 break;
245
246 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
247 return -EINVAL;
248
249 *encoding = &f2fs_sb_encoding_map[i];
250 *flags = le16_to_cpu(sb->s_encoding_flags);
251
252 return 0;
253}
254#endif
255
256static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
257{
258 block_t limit = min((sbi->user_block_count << 1) / 1000,
259 sbi->user_block_count - sbi->reserved_blocks);
260
261 /* limit is 0.2% */
262 if (test_opt(sbi, RESERVE_ROOT) &&
263 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
264 F2FS_OPTION(sbi).root_reserved_blocks = limit;
265 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
266 F2FS_OPTION(sbi).root_reserved_blocks);
267 }
268 if (!test_opt(sbi, RESERVE_ROOT) &&
269 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
270 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
271 !gid_eq(F2FS_OPTION(sbi).s_resgid,
272 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
273 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
274 from_kuid_munged(&init_user_ns,
275 F2FS_OPTION(sbi).s_resuid),
276 from_kgid_munged(&init_user_ns,
277 F2FS_OPTION(sbi).s_resgid));
278}
279
280static void init_once(void *foo)
281{
282 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
283
284 inode_init_once(&fi->vfs_inode);
285}
286
287#ifdef CONFIG_QUOTA
288static const char * const quotatypes[] = INITQFNAMES;
289#define QTYPE2NAME(t) (quotatypes[t])
290static int f2fs_set_qf_name(struct super_block *sb, int qtype,
291 substring_t *args)
292{
293 struct f2fs_sb_info *sbi = F2FS_SB(sb);
294 char *qname;
295 int ret = -EINVAL;
296
297 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
298 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
299 return -EINVAL;
300 }
301 if (f2fs_sb_has_quota_ino(sbi)) {
302 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
303 return 0;
304 }
305
306 qname = match_strdup(args);
307 if (!qname) {
308 f2fs_err(sbi, "Not enough memory for storing quotafile name");
309 return -ENOMEM;
310 }
311 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
312 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
313 ret = 0;
314 else
315 f2fs_err(sbi, "%s quota file already specified",
316 QTYPE2NAME(qtype));
317 goto errout;
318 }
319 if (strchr(qname, '/')) {
320 f2fs_err(sbi, "quotafile must be on filesystem root");
321 goto errout;
322 }
323 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
324 set_opt(sbi, QUOTA);
325 return 0;
326errout:
327 kvfree(qname);
328 return ret;
329}
330
331static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
332{
333 struct f2fs_sb_info *sbi = F2FS_SB(sb);
334
335 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
336 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
337 return -EINVAL;
338 }
339 kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
340 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
341 return 0;
342}
343
344static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
345{
346 /*
347 * We do the test below only for project quotas. 'usrquota' and
348 * 'grpquota' mount options are allowed even without quota feature
349 * to support legacy quotas in quota files.
350 */
351 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
352 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
353 return -1;
354 }
355 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
356 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
357 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
358 if (test_opt(sbi, USRQUOTA) &&
359 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
360 clear_opt(sbi, USRQUOTA);
361
362 if (test_opt(sbi, GRPQUOTA) &&
363 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
364 clear_opt(sbi, GRPQUOTA);
365
366 if (test_opt(sbi, PRJQUOTA) &&
367 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
368 clear_opt(sbi, PRJQUOTA);
369
370 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
371 test_opt(sbi, PRJQUOTA)) {
372 f2fs_err(sbi, "old and new quota format mixing");
373 return -1;
374 }
375
376 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
377 f2fs_err(sbi, "journaled quota format not specified");
378 return -1;
379 }
380 }
381
382 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
383 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
384 F2FS_OPTION(sbi).s_jquota_fmt = 0;
385 }
386 return 0;
387}
388#endif
389
390static int parse_options(struct super_block *sb, char *options)
391{
392 struct f2fs_sb_info *sbi = F2FS_SB(sb);
393 substring_t args[MAX_OPT_ARGS];
394 char *p, *name;
395 int arg = 0;
396 kuid_t uid;
397 kgid_t gid;
398#ifdef CONFIG_QUOTA
399 int ret;
400#endif
401
402 if (!options)
403 return 0;
404
405 while ((p = strsep(&options, ",")) != NULL) {
406 int token;
407 if (!*p)
408 continue;
409 /*
410 * Initialize args struct so we know whether arg was
411 * found; some options take optional arguments.
412 */
413 args[0].to = args[0].from = NULL;
414 token = match_token(p, f2fs_tokens, args);
415
416 switch (token) {
417 case Opt_gc_background:
418 name = match_strdup(&args[0]);
419
420 if (!name)
421 return -ENOMEM;
422 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
423 set_opt(sbi, BG_GC);
424 clear_opt(sbi, FORCE_FG_GC);
425 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
426 clear_opt(sbi, BG_GC);
427 clear_opt(sbi, FORCE_FG_GC);
428 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
429 set_opt(sbi, BG_GC);
430 set_opt(sbi, FORCE_FG_GC);
431 } else {
432 kvfree(name);
433 return -EINVAL;
434 }
435 kvfree(name);
436 break;
437 case Opt_disable_roll_forward:
438 set_opt(sbi, DISABLE_ROLL_FORWARD);
439 break;
440 case Opt_norecovery:
441 /* this option mounts f2fs with ro */
442 set_opt(sbi, DISABLE_ROLL_FORWARD);
443 if (!f2fs_readonly(sb))
444 return -EINVAL;
445 break;
446 case Opt_discard:
447 set_opt(sbi, DISCARD);
448 break;
449 case Opt_nodiscard:
450 if (f2fs_sb_has_blkzoned(sbi)) {
451 f2fs_warn(sbi, "discard is required for zoned block devices");
452 return -EINVAL;
453 }
454 clear_opt(sbi, DISCARD);
455 break;
456 case Opt_noheap:
457 set_opt(sbi, NOHEAP);
458 break;
459 case Opt_heap:
460 clear_opt(sbi, NOHEAP);
461 break;
462#ifdef CONFIG_F2FS_FS_XATTR
463 case Opt_user_xattr:
464 set_opt(sbi, XATTR_USER);
465 break;
466 case Opt_nouser_xattr:
467 clear_opt(sbi, XATTR_USER);
468 break;
469 case Opt_inline_xattr:
470 set_opt(sbi, INLINE_XATTR);
471 break;
472 case Opt_noinline_xattr:
473 clear_opt(sbi, INLINE_XATTR);
474 break;
475 case Opt_inline_xattr_size:
476 if (args->from && match_int(args, &arg))
477 return -EINVAL;
478 set_opt(sbi, INLINE_XATTR_SIZE);
479 F2FS_OPTION(sbi).inline_xattr_size = arg;
480 break;
481#else
482 case Opt_user_xattr:
483 f2fs_info(sbi, "user_xattr options not supported");
484 break;
485 case Opt_nouser_xattr:
486 f2fs_info(sbi, "nouser_xattr options not supported");
487 break;
488 case Opt_inline_xattr:
489 f2fs_info(sbi, "inline_xattr options not supported");
490 break;
491 case Opt_noinline_xattr:
492 f2fs_info(sbi, "noinline_xattr options not supported");
493 break;
494#endif
495#ifdef CONFIG_F2FS_FS_POSIX_ACL
496 case Opt_acl:
497 set_opt(sbi, POSIX_ACL);
498 break;
499 case Opt_noacl:
500 clear_opt(sbi, POSIX_ACL);
501 break;
502#else
503 case Opt_acl:
504 f2fs_info(sbi, "acl options not supported");
505 break;
506 case Opt_noacl:
507 f2fs_info(sbi, "noacl options not supported");
508 break;
509#endif
510 case Opt_active_logs:
511 if (args->from && match_int(args, &arg))
512 return -EINVAL;
513 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
514 return -EINVAL;
515 F2FS_OPTION(sbi).active_logs = arg;
516 break;
517 case Opt_disable_ext_identify:
518 set_opt(sbi, DISABLE_EXT_IDENTIFY);
519 break;
520 case Opt_inline_data:
521 set_opt(sbi, INLINE_DATA);
522 break;
523 case Opt_inline_dentry:
524 set_opt(sbi, INLINE_DENTRY);
525 break;
526 case Opt_noinline_dentry:
527 clear_opt(sbi, INLINE_DENTRY);
528 break;
529 case Opt_flush_merge:
530 set_opt(sbi, FLUSH_MERGE);
531 break;
532 case Opt_noflush_merge:
533 clear_opt(sbi, FLUSH_MERGE);
534 break;
535 case Opt_nobarrier:
536 set_opt(sbi, NOBARRIER);
537 break;
538 case Opt_fastboot:
539 set_opt(sbi, FASTBOOT);
540 break;
541 case Opt_extent_cache:
542 set_opt(sbi, EXTENT_CACHE);
543 break;
544 case Opt_noextent_cache:
545 clear_opt(sbi, EXTENT_CACHE);
546 break;
547 case Opt_noinline_data:
548 clear_opt(sbi, INLINE_DATA);
549 break;
550 case Opt_data_flush:
551 set_opt(sbi, DATA_FLUSH);
552 break;
553 case Opt_reserve_root:
554 if (args->from && match_int(args, &arg))
555 return -EINVAL;
556 if (test_opt(sbi, RESERVE_ROOT)) {
557 f2fs_info(sbi, "Preserve previous reserve_root=%u",
558 F2FS_OPTION(sbi).root_reserved_blocks);
559 } else {
560 F2FS_OPTION(sbi).root_reserved_blocks = arg;
561 set_opt(sbi, RESERVE_ROOT);
562 }
563 break;
564 case Opt_resuid:
565 if (args->from && match_int(args, &arg))
566 return -EINVAL;
567 uid = make_kuid(current_user_ns(), arg);
568 if (!uid_valid(uid)) {
569 f2fs_err(sbi, "Invalid uid value %d", arg);
570 return -EINVAL;
571 }
572 F2FS_OPTION(sbi).s_resuid = uid;
573 break;
574 case Opt_resgid:
575 if (args->from && match_int(args, &arg))
576 return -EINVAL;
577 gid = make_kgid(current_user_ns(), arg);
578 if (!gid_valid(gid)) {
579 f2fs_err(sbi, "Invalid gid value %d", arg);
580 return -EINVAL;
581 }
582 F2FS_OPTION(sbi).s_resgid = gid;
583 break;
584 case Opt_mode:
585 name = match_strdup(&args[0]);
586
587 if (!name)
588 return -ENOMEM;
589 if (strlen(name) == 8 &&
590 !strncmp(name, "adaptive", 8)) {
591 if (f2fs_sb_has_blkzoned(sbi)) {
592 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
593 kvfree(name);
594 return -EINVAL;
595 }
596 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
597 } else if (strlen(name) == 3 &&
598 !strncmp(name, "lfs", 3)) {
599 set_opt_mode(sbi, F2FS_MOUNT_LFS);
600 } else {
601 kvfree(name);
602 return -EINVAL;
603 }
604 kvfree(name);
605 break;
606 case Opt_io_size_bits:
607 if (args->from && match_int(args, &arg))
608 return -EINVAL;
609 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
610 f2fs_warn(sbi, "Not support %d, larger than %d",
611 1 << arg, BIO_MAX_PAGES);
612 return -EINVAL;
613 }
614 F2FS_OPTION(sbi).write_io_size_bits = arg;
615 break;
616#ifdef CONFIG_F2FS_FAULT_INJECTION
617 case Opt_fault_injection:
618 if (args->from && match_int(args, &arg))
619 return -EINVAL;
620 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
621 set_opt(sbi, FAULT_INJECTION);
622 break;
623
624 case Opt_fault_type:
625 if (args->from && match_int(args, &arg))
626 return -EINVAL;
627 f2fs_build_fault_attr(sbi, 0, arg);
628 set_opt(sbi, FAULT_INJECTION);
629 break;
630#else
631 case Opt_fault_injection:
632 f2fs_info(sbi, "fault_injection options not supported");
633 break;
634
635 case Opt_fault_type:
636 f2fs_info(sbi, "fault_type options not supported");
637 break;
638#endif
639 case Opt_lazytime:
640 sb->s_flags |= SB_LAZYTIME;
641 break;
642 case Opt_nolazytime:
643 sb->s_flags &= ~SB_LAZYTIME;
644 break;
645#ifdef CONFIG_QUOTA
646 case Opt_quota:
647 case Opt_usrquota:
648 set_opt(sbi, USRQUOTA);
649 break;
650 case Opt_grpquota:
651 set_opt(sbi, GRPQUOTA);
652 break;
653 case Opt_prjquota:
654 set_opt(sbi, PRJQUOTA);
655 break;
656 case Opt_usrjquota:
657 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
658 if (ret)
659 return ret;
660 break;
661 case Opt_grpjquota:
662 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
663 if (ret)
664 return ret;
665 break;
666 case Opt_prjjquota:
667 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
668 if (ret)
669 return ret;
670 break;
671 case Opt_offusrjquota:
672 ret = f2fs_clear_qf_name(sb, USRQUOTA);
673 if (ret)
674 return ret;
675 break;
676 case Opt_offgrpjquota:
677 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
678 if (ret)
679 return ret;
680 break;
681 case Opt_offprjjquota:
682 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
683 if (ret)
684 return ret;
685 break;
686 case Opt_jqfmt_vfsold:
687 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
688 break;
689 case Opt_jqfmt_vfsv0:
690 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
691 break;
692 case Opt_jqfmt_vfsv1:
693 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
694 break;
695 case Opt_noquota:
696 clear_opt(sbi, QUOTA);
697 clear_opt(sbi, USRQUOTA);
698 clear_opt(sbi, GRPQUOTA);
699 clear_opt(sbi, PRJQUOTA);
700 break;
701#else
702 case Opt_quota:
703 case Opt_usrquota:
704 case Opt_grpquota:
705 case Opt_prjquota:
706 case Opt_usrjquota:
707 case Opt_grpjquota:
708 case Opt_prjjquota:
709 case Opt_offusrjquota:
710 case Opt_offgrpjquota:
711 case Opt_offprjjquota:
712 case Opt_jqfmt_vfsold:
713 case Opt_jqfmt_vfsv0:
714 case Opt_jqfmt_vfsv1:
715 case Opt_noquota:
716 f2fs_info(sbi, "quota operations not supported");
717 break;
718#endif
719 case Opt_whint:
720 name = match_strdup(&args[0]);
721 if (!name)
722 return -ENOMEM;
723 if (strlen(name) == 10 &&
724 !strncmp(name, "user-based", 10)) {
725 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
726 } else if (strlen(name) == 3 &&
727 !strncmp(name, "off", 3)) {
728 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
729 } else if (strlen(name) == 8 &&
730 !strncmp(name, "fs-based", 8)) {
731 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
732 } else {
733 kvfree(name);
734 return -EINVAL;
735 }
736 kvfree(name);
737 break;
738 case Opt_alloc:
739 name = match_strdup(&args[0]);
740 if (!name)
741 return -ENOMEM;
742
743 if (strlen(name) == 7 &&
744 !strncmp(name, "default", 7)) {
745 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
746 } else if (strlen(name) == 5 &&
747 !strncmp(name, "reuse", 5)) {
748 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
749 } else {
750 kvfree(name);
751 return -EINVAL;
752 }
753 kvfree(name);
754 break;
755 case Opt_fsync:
756 name = match_strdup(&args[0]);
757 if (!name)
758 return -ENOMEM;
759 if (strlen(name) == 5 &&
760 !strncmp(name, "posix", 5)) {
761 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
762 } else if (strlen(name) == 6 &&
763 !strncmp(name, "strict", 6)) {
764 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
765 } else if (strlen(name) == 9 &&
766 !strncmp(name, "nobarrier", 9)) {
767 F2FS_OPTION(sbi).fsync_mode =
768 FSYNC_MODE_NOBARRIER;
769 } else {
770 kvfree(name);
771 return -EINVAL;
772 }
773 kvfree(name);
774 break;
775 case Opt_test_dummy_encryption:
776#ifdef CONFIG_FS_ENCRYPTION
777 if (!f2fs_sb_has_encrypt(sbi)) {
778 f2fs_err(sbi, "Encrypt feature is off");
779 return -EINVAL;
780 }
781
782 F2FS_OPTION(sbi).test_dummy_encryption = true;
783 f2fs_info(sbi, "Test dummy encryption mode enabled");
784#else
785 f2fs_info(sbi, "Test dummy encryption mount option ignored");
786#endif
787 break;
788 case Opt_checkpoint_disable_cap_perc:
789 if (args->from && match_int(args, &arg))
790 return -EINVAL;
791 if (arg < 0 || arg > 100)
792 return -EINVAL;
793 if (arg == 100)
794 F2FS_OPTION(sbi).unusable_cap =
795 sbi->user_block_count;
796 else
797 F2FS_OPTION(sbi).unusable_cap =
798 (sbi->user_block_count / 100) * arg;
799 set_opt(sbi, DISABLE_CHECKPOINT);
800 break;
801 case Opt_checkpoint_disable_cap:
802 if (args->from && match_int(args, &arg))
803 return -EINVAL;
804 F2FS_OPTION(sbi).unusable_cap = arg;
805 set_opt(sbi, DISABLE_CHECKPOINT);
806 break;
807 case Opt_checkpoint_disable:
808 set_opt(sbi, DISABLE_CHECKPOINT);
809 break;
810 case Opt_checkpoint_enable:
811 clear_opt(sbi, DISABLE_CHECKPOINT);
812 break;
813 default:
814 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
815 p);
816 return -EINVAL;
817 }
818 }
819#ifdef CONFIG_QUOTA
820 if (f2fs_check_quota_options(sbi))
821 return -EINVAL;
822#else
823 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
824 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
825 return -EINVAL;
826 }
827 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
828 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
829 return -EINVAL;
830 }
831#endif
832#ifndef CONFIG_UNICODE
833 if (f2fs_sb_has_casefold(sbi)) {
834 f2fs_err(sbi,
835 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
836 return -EINVAL;
837 }
838#endif
839
840 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
841 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
842 F2FS_IO_SIZE_KB(sbi));
843 return -EINVAL;
844 }
845
846 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
847 int min_size, max_size;
848
849 if (!f2fs_sb_has_extra_attr(sbi) ||
850 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
851 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
852 return -EINVAL;
853 }
854 if (!test_opt(sbi, INLINE_XATTR)) {
855 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
856 return -EINVAL;
857 }
858
859 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
860 max_size = MAX_INLINE_XATTR_SIZE;
861
862 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
863 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
864 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
865 min_size, max_size);
866 return -EINVAL;
867 }
868 }
869
870 if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
871 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
872 return -EINVAL;
873 }
874
875 /* Not pass down write hints if the number of active logs is lesser
876 * than NR_CURSEG_TYPE.
877 */
878 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
879 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
880 return 0;
881}
882
883static struct inode *f2fs_alloc_inode(struct super_block *sb)
884{
885 struct f2fs_inode_info *fi;
886
887 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
888 if (!fi)
889 return NULL;
890
891 init_once((void *) fi);
892
893 /* Initialize f2fs-specific inode info */
894 atomic_set(&fi->dirty_pages, 0);
895 init_rwsem(&fi->i_sem);
896 INIT_LIST_HEAD(&fi->dirty_list);
897 INIT_LIST_HEAD(&fi->gdirty_list);
898 INIT_LIST_HEAD(&fi->inmem_ilist);
899 INIT_LIST_HEAD(&fi->inmem_pages);
900 mutex_init(&fi->inmem_lock);
901 init_rwsem(&fi->i_gc_rwsem[READ]);
902 init_rwsem(&fi->i_gc_rwsem[WRITE]);
903 init_rwsem(&fi->i_mmap_sem);
904 init_rwsem(&fi->i_xattr_sem);
905
906 /* Will be used by directory only */
907 fi->i_dir_level = F2FS_SB(sb)->dir_level;
908
909 return &fi->vfs_inode;
910}
911
912static int f2fs_drop_inode(struct inode *inode)
913{
914 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
915 int ret;
916
917 /*
918 * during filesystem shutdown, if checkpoint is disabled,
919 * drop useless meta/node dirty pages.
920 */
921 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
922 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
923 inode->i_ino == F2FS_META_INO(sbi)) {
924 trace_f2fs_drop_inode(inode, 1);
925 return 1;
926 }
927 }
928
929 /*
930 * This is to avoid a deadlock condition like below.
931 * writeback_single_inode(inode)
932 * - f2fs_write_data_page
933 * - f2fs_gc -> iput -> evict
934 * - inode_wait_for_writeback(inode)
935 */
936 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
937 if (!inode->i_nlink && !is_bad_inode(inode)) {
938 /* to avoid evict_inode call simultaneously */
939 atomic_inc(&inode->i_count);
940 spin_unlock(&inode->i_lock);
941
942 /* some remained atomic pages should discarded */
943 if (f2fs_is_atomic_file(inode))
944 f2fs_drop_inmem_pages(inode);
945
946 /* should remain fi->extent_tree for writepage */
947 f2fs_destroy_extent_node(inode);
948
949 sb_start_intwrite(inode->i_sb);
950 f2fs_i_size_write(inode, 0);
951
952 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
953 inode, NULL, 0, DATA);
954 truncate_inode_pages_final(inode->i_mapping);
955
956 if (F2FS_HAS_BLOCKS(inode))
957 f2fs_truncate(inode);
958
959 sb_end_intwrite(inode->i_sb);
960
961 spin_lock(&inode->i_lock);
962 atomic_dec(&inode->i_count);
963 }
964 trace_f2fs_drop_inode(inode, 0);
965 return 0;
966 }
967 ret = generic_drop_inode(inode);
968 if (!ret)
969 ret = fscrypt_drop_inode(inode);
970 trace_f2fs_drop_inode(inode, ret);
971 return ret;
972}
973
974int f2fs_inode_dirtied(struct inode *inode, bool sync)
975{
976 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
977 int ret = 0;
978
979 spin_lock(&sbi->inode_lock[DIRTY_META]);
980 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
981 ret = 1;
982 } else {
983 set_inode_flag(inode, FI_DIRTY_INODE);
984 stat_inc_dirty_inode(sbi, DIRTY_META);
985 }
986 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
987 list_add_tail(&F2FS_I(inode)->gdirty_list,
988 &sbi->inode_list[DIRTY_META]);
989 inc_page_count(sbi, F2FS_DIRTY_IMETA);
990 }
991 spin_unlock(&sbi->inode_lock[DIRTY_META]);
992 return ret;
993}
994
995void f2fs_inode_synced(struct inode *inode)
996{
997 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
998
999 spin_lock(&sbi->inode_lock[DIRTY_META]);
1000 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1001 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1002 return;
1003 }
1004 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1005 list_del_init(&F2FS_I(inode)->gdirty_list);
1006 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1007 }
1008 clear_inode_flag(inode, FI_DIRTY_INODE);
1009 clear_inode_flag(inode, FI_AUTO_RECOVER);
1010 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1011 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1012}
1013
1014/*
1015 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1016 *
1017 * We should call set_dirty_inode to write the dirty inode through write_inode.
1018 */
1019static void f2fs_dirty_inode(struct inode *inode, int flags)
1020{
1021 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1022
1023 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1024 inode->i_ino == F2FS_META_INO(sbi))
1025 return;
1026
1027 if (flags == I_DIRTY_TIME)
1028 return;
1029
1030 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1031 clear_inode_flag(inode, FI_AUTO_RECOVER);
1032
1033 f2fs_inode_dirtied(inode, false);
1034}
1035
1036static void f2fs_free_inode(struct inode *inode)
1037{
1038 fscrypt_free_inode(inode);
1039 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1040}
1041
1042static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1043{
1044 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1045 percpu_counter_destroy(&sbi->total_valid_inode_count);
1046}
1047
1048static void destroy_device_list(struct f2fs_sb_info *sbi)
1049{
1050 int i;
1051
1052 for (i = 0; i < sbi->s_ndevs; i++) {
1053 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1054#ifdef CONFIG_BLK_DEV_ZONED
1055 kvfree(FDEV(i).blkz_seq);
1056#endif
1057 }
1058 kvfree(sbi->devs);
1059}
1060
1061static void f2fs_put_super(struct super_block *sb)
1062{
1063 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1064 int i;
1065 bool dropped;
1066
1067 f2fs_quota_off_umount(sb);
1068
1069 /* prevent remaining shrinker jobs */
1070 mutex_lock(&sbi->umount_mutex);
1071
1072 /*
1073 * We don't need to do checkpoint when superblock is clean.
1074 * But, the previous checkpoint was not done by umount, it needs to do
1075 * clean checkpoint again.
1076 */
1077 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1078 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1079 struct cp_control cpc = {
1080 .reason = CP_UMOUNT,
1081 };
1082 f2fs_write_checkpoint(sbi, &cpc);
1083 }
1084
1085 /* be sure to wait for any on-going discard commands */
1086 dropped = f2fs_issue_discard_timeout(sbi);
1087
1088 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1089 !sbi->discard_blks && !dropped) {
1090 struct cp_control cpc = {
1091 .reason = CP_UMOUNT | CP_TRIMMED,
1092 };
1093 f2fs_write_checkpoint(sbi, &cpc);
1094 }
1095
1096 /*
1097 * normally superblock is clean, so we need to release this.
1098 * In addition, EIO will skip do checkpoint, we need this as well.
1099 */
1100 f2fs_release_ino_entry(sbi, true);
1101
1102 f2fs_leave_shrinker(sbi);
1103 mutex_unlock(&sbi->umount_mutex);
1104
1105 /* our cp_error case, we can wait for any writeback page */
1106 f2fs_flush_merged_writes(sbi);
1107
1108 f2fs_wait_on_all_pages_writeback(sbi);
1109
1110 f2fs_bug_on(sbi, sbi->fsync_node_num);
1111
1112 iput(sbi->node_inode);
1113 sbi->node_inode = NULL;
1114
1115 iput(sbi->meta_inode);
1116 sbi->meta_inode = NULL;
1117
1118 /*
1119 * iput() can update stat information, if f2fs_write_checkpoint()
1120 * above failed with error.
1121 */
1122 f2fs_destroy_stats(sbi);
1123
1124 /* destroy f2fs internal modules */
1125 f2fs_destroy_node_manager(sbi);
1126 f2fs_destroy_segment_manager(sbi);
1127
1128 kvfree(sbi->ckpt);
1129
1130 f2fs_unregister_sysfs(sbi);
1131
1132 sb->s_fs_info = NULL;
1133 if (sbi->s_chksum_driver)
1134 crypto_free_shash(sbi->s_chksum_driver);
1135 kvfree(sbi->raw_super);
1136
1137 destroy_device_list(sbi);
1138 mempool_destroy(sbi->write_io_dummy);
1139#ifdef CONFIG_QUOTA
1140 for (i = 0; i < MAXQUOTAS; i++)
1141 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1142#endif
1143 destroy_percpu_info(sbi);
1144 for (i = 0; i < NR_PAGE_TYPE; i++)
1145 kvfree(sbi->write_io[i]);
1146#ifdef CONFIG_UNICODE
1147 utf8_unload(sbi->s_encoding);
1148#endif
1149 kvfree(sbi);
1150}
1151
1152int f2fs_sync_fs(struct super_block *sb, int sync)
1153{
1154 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1155 int err = 0;
1156
1157 if (unlikely(f2fs_cp_error(sbi)))
1158 return 0;
1159 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1160 return 0;
1161
1162 trace_f2fs_sync_fs(sb, sync);
1163
1164 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1165 return -EAGAIN;
1166
1167 if (sync) {
1168 struct cp_control cpc;
1169
1170 cpc.reason = __get_cp_reason(sbi);
1171
1172 mutex_lock(&sbi->gc_mutex);
1173 err = f2fs_write_checkpoint(sbi, &cpc);
1174 mutex_unlock(&sbi->gc_mutex);
1175 }
1176 f2fs_trace_ios(NULL, 1);
1177
1178 return err;
1179}
1180
1181static int f2fs_freeze(struct super_block *sb)
1182{
1183 if (f2fs_readonly(sb))
1184 return 0;
1185
1186 /* IO error happened before */
1187 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1188 return -EIO;
1189
1190 /* must be clean, since sync_filesystem() was already called */
1191 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1192 return -EINVAL;
1193 return 0;
1194}
1195
1196static int f2fs_unfreeze(struct super_block *sb)
1197{
1198 return 0;
1199}
1200
1201#ifdef CONFIG_QUOTA
1202static int f2fs_statfs_project(struct super_block *sb,
1203 kprojid_t projid, struct kstatfs *buf)
1204{
1205 struct kqid qid;
1206 struct dquot *dquot;
1207 u64 limit;
1208 u64 curblock;
1209
1210 qid = make_kqid_projid(projid);
1211 dquot = dqget(sb, qid);
1212 if (IS_ERR(dquot))
1213 return PTR_ERR(dquot);
1214 spin_lock(&dquot->dq_dqb_lock);
1215
1216 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1217 dquot->dq_dqb.dqb_bsoftlimit :
1218 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1219 if (limit && buf->f_blocks > limit) {
1220 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1221 buf->f_blocks = limit;
1222 buf->f_bfree = buf->f_bavail =
1223 (buf->f_blocks > curblock) ?
1224 (buf->f_blocks - curblock) : 0;
1225 }
1226
1227 limit = dquot->dq_dqb.dqb_isoftlimit ?
1228 dquot->dq_dqb.dqb_isoftlimit :
1229 dquot->dq_dqb.dqb_ihardlimit;
1230 if (limit && buf->f_files > limit) {
1231 buf->f_files = limit;
1232 buf->f_ffree =
1233 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1234 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1235 }
1236
1237 spin_unlock(&dquot->dq_dqb_lock);
1238 dqput(dquot);
1239 return 0;
1240}
1241#endif
1242
1243static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1244{
1245 struct super_block *sb = dentry->d_sb;
1246 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1247 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1248 block_t total_count, user_block_count, start_count;
1249 u64 avail_node_count;
1250
1251 total_count = le64_to_cpu(sbi->raw_super->block_count);
1252 user_block_count = sbi->user_block_count;
1253 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1254 buf->f_type = F2FS_SUPER_MAGIC;
1255 buf->f_bsize = sbi->blocksize;
1256
1257 buf->f_blocks = total_count - start_count;
1258 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1259 sbi->current_reserved_blocks;
1260
1261 spin_lock(&sbi->stat_lock);
1262 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1263 buf->f_bfree = 0;
1264 else
1265 buf->f_bfree -= sbi->unusable_block_count;
1266 spin_unlock(&sbi->stat_lock);
1267
1268 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1269 buf->f_bavail = buf->f_bfree -
1270 F2FS_OPTION(sbi).root_reserved_blocks;
1271 else
1272 buf->f_bavail = 0;
1273
1274 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1275
1276 if (avail_node_count > user_block_count) {
1277 buf->f_files = user_block_count;
1278 buf->f_ffree = buf->f_bavail;
1279 } else {
1280 buf->f_files = avail_node_count;
1281 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1282 buf->f_bavail);
1283 }
1284
1285 buf->f_namelen = F2FS_NAME_LEN;
1286 buf->f_fsid.val[0] = (u32)id;
1287 buf->f_fsid.val[1] = (u32)(id >> 32);
1288
1289#ifdef CONFIG_QUOTA
1290 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1291 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1292 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1293 }
1294#endif
1295 return 0;
1296}
1297
1298static inline void f2fs_show_quota_options(struct seq_file *seq,
1299 struct super_block *sb)
1300{
1301#ifdef CONFIG_QUOTA
1302 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1303
1304 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1305 char *fmtname = "";
1306
1307 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1308 case QFMT_VFS_OLD:
1309 fmtname = "vfsold";
1310 break;
1311 case QFMT_VFS_V0:
1312 fmtname = "vfsv0";
1313 break;
1314 case QFMT_VFS_V1:
1315 fmtname = "vfsv1";
1316 break;
1317 }
1318 seq_printf(seq, ",jqfmt=%s", fmtname);
1319 }
1320
1321 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1322 seq_show_option(seq, "usrjquota",
1323 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1324
1325 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1326 seq_show_option(seq, "grpjquota",
1327 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1328
1329 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1330 seq_show_option(seq, "prjjquota",
1331 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1332#endif
1333}
1334
1335static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1336{
1337 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1338
1339 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1340 if (test_opt(sbi, FORCE_FG_GC))
1341 seq_printf(seq, ",background_gc=%s", "sync");
1342 else
1343 seq_printf(seq, ",background_gc=%s", "on");
1344 } else {
1345 seq_printf(seq, ",background_gc=%s", "off");
1346 }
1347 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1348 seq_puts(seq, ",disable_roll_forward");
1349 if (test_opt(sbi, DISCARD))
1350 seq_puts(seq, ",discard");
1351 else
1352 seq_puts(seq, ",nodiscard");
1353 if (test_opt(sbi, NOHEAP))
1354 seq_puts(seq, ",no_heap");
1355 else
1356 seq_puts(seq, ",heap");
1357#ifdef CONFIG_F2FS_FS_XATTR
1358 if (test_opt(sbi, XATTR_USER))
1359 seq_puts(seq, ",user_xattr");
1360 else
1361 seq_puts(seq, ",nouser_xattr");
1362 if (test_opt(sbi, INLINE_XATTR))
1363 seq_puts(seq, ",inline_xattr");
1364 else
1365 seq_puts(seq, ",noinline_xattr");
1366 if (test_opt(sbi, INLINE_XATTR_SIZE))
1367 seq_printf(seq, ",inline_xattr_size=%u",
1368 F2FS_OPTION(sbi).inline_xattr_size);
1369#endif
1370#ifdef CONFIG_F2FS_FS_POSIX_ACL
1371 if (test_opt(sbi, POSIX_ACL))
1372 seq_puts(seq, ",acl");
1373 else
1374 seq_puts(seq, ",noacl");
1375#endif
1376 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1377 seq_puts(seq, ",disable_ext_identify");
1378 if (test_opt(sbi, INLINE_DATA))
1379 seq_puts(seq, ",inline_data");
1380 else
1381 seq_puts(seq, ",noinline_data");
1382 if (test_opt(sbi, INLINE_DENTRY))
1383 seq_puts(seq, ",inline_dentry");
1384 else
1385 seq_puts(seq, ",noinline_dentry");
1386 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1387 seq_puts(seq, ",flush_merge");
1388 if (test_opt(sbi, NOBARRIER))
1389 seq_puts(seq, ",nobarrier");
1390 if (test_opt(sbi, FASTBOOT))
1391 seq_puts(seq, ",fastboot");
1392 if (test_opt(sbi, EXTENT_CACHE))
1393 seq_puts(seq, ",extent_cache");
1394 else
1395 seq_puts(seq, ",noextent_cache");
1396 if (test_opt(sbi, DATA_FLUSH))
1397 seq_puts(seq, ",data_flush");
1398
1399 seq_puts(seq, ",mode=");
1400 if (test_opt(sbi, ADAPTIVE))
1401 seq_puts(seq, "adaptive");
1402 else if (test_opt(sbi, LFS))
1403 seq_puts(seq, "lfs");
1404 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1405 if (test_opt(sbi, RESERVE_ROOT))
1406 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1407 F2FS_OPTION(sbi).root_reserved_blocks,
1408 from_kuid_munged(&init_user_ns,
1409 F2FS_OPTION(sbi).s_resuid),
1410 from_kgid_munged(&init_user_ns,
1411 F2FS_OPTION(sbi).s_resgid));
1412 if (F2FS_IO_SIZE_BITS(sbi))
1413 seq_printf(seq, ",io_bits=%u",
1414 F2FS_OPTION(sbi).write_io_size_bits);
1415#ifdef CONFIG_F2FS_FAULT_INJECTION
1416 if (test_opt(sbi, FAULT_INJECTION)) {
1417 seq_printf(seq, ",fault_injection=%u",
1418 F2FS_OPTION(sbi).fault_info.inject_rate);
1419 seq_printf(seq, ",fault_type=%u",
1420 F2FS_OPTION(sbi).fault_info.inject_type);
1421 }
1422#endif
1423#ifdef CONFIG_QUOTA
1424 if (test_opt(sbi, QUOTA))
1425 seq_puts(seq, ",quota");
1426 if (test_opt(sbi, USRQUOTA))
1427 seq_puts(seq, ",usrquota");
1428 if (test_opt(sbi, GRPQUOTA))
1429 seq_puts(seq, ",grpquota");
1430 if (test_opt(sbi, PRJQUOTA))
1431 seq_puts(seq, ",prjquota");
1432#endif
1433 f2fs_show_quota_options(seq, sbi->sb);
1434 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1435 seq_printf(seq, ",whint_mode=%s", "user-based");
1436 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1437 seq_printf(seq, ",whint_mode=%s", "fs-based");
1438#ifdef CONFIG_FS_ENCRYPTION
1439 if (F2FS_OPTION(sbi).test_dummy_encryption)
1440 seq_puts(seq, ",test_dummy_encryption");
1441#endif
1442
1443 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1444 seq_printf(seq, ",alloc_mode=%s", "default");
1445 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1446 seq_printf(seq, ",alloc_mode=%s", "reuse");
1447
1448 if (test_opt(sbi, DISABLE_CHECKPOINT))
1449 seq_printf(seq, ",checkpoint=disable:%u",
1450 F2FS_OPTION(sbi).unusable_cap);
1451 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1452 seq_printf(seq, ",fsync_mode=%s", "posix");
1453 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1454 seq_printf(seq, ",fsync_mode=%s", "strict");
1455 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1456 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1457 return 0;
1458}
1459
1460static void default_options(struct f2fs_sb_info *sbi)
1461{
1462 /* init some FS parameters */
1463 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1464 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1465 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1466 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1467 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1468 F2FS_OPTION(sbi).test_dummy_encryption = false;
1469 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1470 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1471
1472 set_opt(sbi, BG_GC);
1473 set_opt(sbi, INLINE_XATTR);
1474 set_opt(sbi, INLINE_DATA);
1475 set_opt(sbi, INLINE_DENTRY);
1476 set_opt(sbi, EXTENT_CACHE);
1477 set_opt(sbi, NOHEAP);
1478 clear_opt(sbi, DISABLE_CHECKPOINT);
1479 F2FS_OPTION(sbi).unusable_cap = 0;
1480 sbi->sb->s_flags |= SB_LAZYTIME;
1481 set_opt(sbi, FLUSH_MERGE);
1482 set_opt(sbi, DISCARD);
1483 if (f2fs_sb_has_blkzoned(sbi))
1484 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1485 else
1486 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1487
1488#ifdef CONFIG_F2FS_FS_XATTR
1489 set_opt(sbi, XATTR_USER);
1490#endif
1491#ifdef CONFIG_F2FS_FS_POSIX_ACL
1492 set_opt(sbi, POSIX_ACL);
1493#endif
1494
1495 f2fs_build_fault_attr(sbi, 0, 0);
1496}
1497
1498#ifdef CONFIG_QUOTA
1499static int f2fs_enable_quotas(struct super_block *sb);
1500#endif
1501
1502static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1503{
1504 unsigned int s_flags = sbi->sb->s_flags;
1505 struct cp_control cpc;
1506 int err = 0;
1507 int ret;
1508 block_t unusable;
1509
1510 if (s_flags & SB_RDONLY) {
1511 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1512 return -EINVAL;
1513 }
1514 sbi->sb->s_flags |= SB_ACTIVE;
1515
1516 f2fs_update_time(sbi, DISABLE_TIME);
1517
1518 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1519 mutex_lock(&sbi->gc_mutex);
1520 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1521 if (err == -ENODATA) {
1522 err = 0;
1523 break;
1524 }
1525 if (err && err != -EAGAIN)
1526 break;
1527 }
1528
1529 ret = sync_filesystem(sbi->sb);
1530 if (ret || err) {
1531 err = ret ? ret: err;
1532 goto restore_flag;
1533 }
1534
1535 unusable = f2fs_get_unusable_blocks(sbi);
1536 if (f2fs_disable_cp_again(sbi, unusable)) {
1537 err = -EAGAIN;
1538 goto restore_flag;
1539 }
1540
1541 mutex_lock(&sbi->gc_mutex);
1542 cpc.reason = CP_PAUSE;
1543 set_sbi_flag(sbi, SBI_CP_DISABLED);
1544 err = f2fs_write_checkpoint(sbi, &cpc);
1545 if (err)
1546 goto out_unlock;
1547
1548 spin_lock(&sbi->stat_lock);
1549 sbi->unusable_block_count = unusable;
1550 spin_unlock(&sbi->stat_lock);
1551
1552out_unlock:
1553 mutex_unlock(&sbi->gc_mutex);
1554restore_flag:
1555 sbi->sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1556 return err;
1557}
1558
1559static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1560{
1561 mutex_lock(&sbi->gc_mutex);
1562 f2fs_dirty_to_prefree(sbi);
1563
1564 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1565 set_sbi_flag(sbi, SBI_IS_DIRTY);
1566 mutex_unlock(&sbi->gc_mutex);
1567
1568 f2fs_sync_fs(sbi->sb, 1);
1569}
1570
1571static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1572{
1573 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1574 struct f2fs_mount_info org_mount_opt;
1575 unsigned long old_sb_flags;
1576 int err;
1577 bool need_restart_gc = false;
1578 bool need_stop_gc = false;
1579 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1580 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1581 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1582 bool checkpoint_changed;
1583#ifdef CONFIG_QUOTA
1584 int i, j;
1585#endif
1586
1587 /*
1588 * Save the old mount options in case we
1589 * need to restore them.
1590 */
1591 org_mount_opt = sbi->mount_opt;
1592 old_sb_flags = sb->s_flags;
1593
1594#ifdef CONFIG_QUOTA
1595 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1596 for (i = 0; i < MAXQUOTAS; i++) {
1597 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1598 org_mount_opt.s_qf_names[i] =
1599 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1600 GFP_KERNEL);
1601 if (!org_mount_opt.s_qf_names[i]) {
1602 for (j = 0; j < i; j++)
1603 kvfree(org_mount_opt.s_qf_names[j]);
1604 return -ENOMEM;
1605 }
1606 } else {
1607 org_mount_opt.s_qf_names[i] = NULL;
1608 }
1609 }
1610#endif
1611
1612 /* recover superblocks we couldn't write due to previous RO mount */
1613 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1614 err = f2fs_commit_super(sbi, false);
1615 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1616 err);
1617 if (!err)
1618 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1619 }
1620
1621 default_options(sbi);
1622
1623 /* parse mount options */
1624 err = parse_options(sb, data);
1625 if (err)
1626 goto restore_opts;
1627 checkpoint_changed =
1628 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1629
1630 /*
1631 * Previous and new state of filesystem is RO,
1632 * so skip checking GC and FLUSH_MERGE conditions.
1633 */
1634 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1635 goto skip;
1636
1637#ifdef CONFIG_QUOTA
1638 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1639 err = dquot_suspend(sb, -1);
1640 if (err < 0)
1641 goto restore_opts;
1642 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1643 /* dquot_resume needs RW */
1644 sb->s_flags &= ~SB_RDONLY;
1645 if (sb_any_quota_suspended(sb)) {
1646 dquot_resume(sb, -1);
1647 } else if (f2fs_sb_has_quota_ino(sbi)) {
1648 err = f2fs_enable_quotas(sb);
1649 if (err)
1650 goto restore_opts;
1651 }
1652 }
1653#endif
1654 /* disallow enable/disable extent_cache dynamically */
1655 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1656 err = -EINVAL;
1657 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1658 goto restore_opts;
1659 }
1660
1661 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1662 err = -EINVAL;
1663 f2fs_warn(sbi, "switch io_bits option is not allowed");
1664 goto restore_opts;
1665 }
1666
1667 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1668 err = -EINVAL;
1669 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1670 goto restore_opts;
1671 }
1672
1673 /*
1674 * We stop the GC thread if FS is mounted as RO
1675 * or if background_gc = off is passed in mount
1676 * option. Also sync the filesystem.
1677 */
1678 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1679 if (sbi->gc_thread) {
1680 f2fs_stop_gc_thread(sbi);
1681 need_restart_gc = true;
1682 }
1683 } else if (!sbi->gc_thread) {
1684 err = f2fs_start_gc_thread(sbi);
1685 if (err)
1686 goto restore_opts;
1687 need_stop_gc = true;
1688 }
1689
1690 if (*flags & SB_RDONLY ||
1691 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1692 writeback_inodes_sb(sb, WB_REASON_SYNC);
1693 sync_inodes_sb(sb);
1694
1695 set_sbi_flag(sbi, SBI_IS_DIRTY);
1696 set_sbi_flag(sbi, SBI_IS_CLOSE);
1697 f2fs_sync_fs(sb, 1);
1698 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1699 }
1700
1701 if (checkpoint_changed) {
1702 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1703 err = f2fs_disable_checkpoint(sbi);
1704 if (err)
1705 goto restore_gc;
1706 } else {
1707 f2fs_enable_checkpoint(sbi);
1708 }
1709 }
1710
1711 /*
1712 * We stop issue flush thread if FS is mounted as RO
1713 * or if flush_merge is not passed in mount option.
1714 */
1715 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1716 clear_opt(sbi, FLUSH_MERGE);
1717 f2fs_destroy_flush_cmd_control(sbi, false);
1718 } else {
1719 err = f2fs_create_flush_cmd_control(sbi);
1720 if (err)
1721 goto restore_gc;
1722 }
1723skip:
1724#ifdef CONFIG_QUOTA
1725 /* Release old quota file names */
1726 for (i = 0; i < MAXQUOTAS; i++)
1727 kvfree(org_mount_opt.s_qf_names[i]);
1728#endif
1729 /* Update the POSIXACL Flag */
1730 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1731 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1732
1733 limit_reserve_root(sbi);
1734 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1735 return 0;
1736restore_gc:
1737 if (need_restart_gc) {
1738 if (f2fs_start_gc_thread(sbi))
1739 f2fs_warn(sbi, "background gc thread has stopped");
1740 } else if (need_stop_gc) {
1741 f2fs_stop_gc_thread(sbi);
1742 }
1743restore_opts:
1744#ifdef CONFIG_QUOTA
1745 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1746 for (i = 0; i < MAXQUOTAS; i++) {
1747 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1748 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1749 }
1750#endif
1751 sbi->mount_opt = org_mount_opt;
1752 sb->s_flags = old_sb_flags;
1753 return err;
1754}
1755
1756#ifdef CONFIG_QUOTA
1757/* Read data from quotafile */
1758static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1759 size_t len, loff_t off)
1760{
1761 struct inode *inode = sb_dqopt(sb)->files[type];
1762 struct address_space *mapping = inode->i_mapping;
1763 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1764 int offset = off & (sb->s_blocksize - 1);
1765 int tocopy;
1766 size_t toread;
1767 loff_t i_size = i_size_read(inode);
1768 struct page *page;
1769 char *kaddr;
1770
1771 if (off > i_size)
1772 return 0;
1773
1774 if (off + len > i_size)
1775 len = i_size - off;
1776 toread = len;
1777 while (toread > 0) {
1778 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1779repeat:
1780 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1781 if (IS_ERR(page)) {
1782 if (PTR_ERR(page) == -ENOMEM) {
1783 congestion_wait(BLK_RW_ASYNC, HZ/50);
1784 goto repeat;
1785 }
1786 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1787 return PTR_ERR(page);
1788 }
1789
1790 lock_page(page);
1791
1792 if (unlikely(page->mapping != mapping)) {
1793 f2fs_put_page(page, 1);
1794 goto repeat;
1795 }
1796 if (unlikely(!PageUptodate(page))) {
1797 f2fs_put_page(page, 1);
1798 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1799 return -EIO;
1800 }
1801
1802 kaddr = kmap_atomic(page);
1803 memcpy(data, kaddr + offset, tocopy);
1804 kunmap_atomic(kaddr);
1805 f2fs_put_page(page, 1);
1806
1807 offset = 0;
1808 toread -= tocopy;
1809 data += tocopy;
1810 blkidx++;
1811 }
1812 return len;
1813}
1814
1815/* Write to quotafile */
1816static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1817 const char *data, size_t len, loff_t off)
1818{
1819 struct inode *inode = sb_dqopt(sb)->files[type];
1820 struct address_space *mapping = inode->i_mapping;
1821 const struct address_space_operations *a_ops = mapping->a_ops;
1822 int offset = off & (sb->s_blocksize - 1);
1823 size_t towrite = len;
1824 struct page *page;
1825 char *kaddr;
1826 int err = 0;
1827 int tocopy;
1828
1829 while (towrite > 0) {
1830 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1831 towrite);
1832retry:
1833 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1834 &page, NULL);
1835 if (unlikely(err)) {
1836 if (err == -ENOMEM) {
1837 congestion_wait(BLK_RW_ASYNC, HZ/50);
1838 goto retry;
1839 }
1840 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1841 break;
1842 }
1843
1844 kaddr = kmap_atomic(page);
1845 memcpy(kaddr + offset, data, tocopy);
1846 kunmap_atomic(kaddr);
1847 flush_dcache_page(page);
1848
1849 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1850 page, NULL);
1851 offset = 0;
1852 towrite -= tocopy;
1853 off += tocopy;
1854 data += tocopy;
1855 cond_resched();
1856 }
1857
1858 if (len == towrite)
1859 return err;
1860 inode->i_mtime = inode->i_ctime = current_time(inode);
1861 f2fs_mark_inode_dirty_sync(inode, false);
1862 return len - towrite;
1863}
1864
1865static struct dquot **f2fs_get_dquots(struct inode *inode)
1866{
1867 return F2FS_I(inode)->i_dquot;
1868}
1869
1870static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1871{
1872 return &F2FS_I(inode)->i_reserved_quota;
1873}
1874
1875static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1876{
1877 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1878 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1879 return 0;
1880 }
1881
1882 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1883 F2FS_OPTION(sbi).s_jquota_fmt, type);
1884}
1885
1886int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1887{
1888 int enabled = 0;
1889 int i, err;
1890
1891 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1892 err = f2fs_enable_quotas(sbi->sb);
1893 if (err) {
1894 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
1895 return 0;
1896 }
1897 return 1;
1898 }
1899
1900 for (i = 0; i < MAXQUOTAS; i++) {
1901 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1902 err = f2fs_quota_on_mount(sbi, i);
1903 if (!err) {
1904 enabled = 1;
1905 continue;
1906 }
1907 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
1908 err, i);
1909 }
1910 }
1911 return enabled;
1912}
1913
1914static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1915 unsigned int flags)
1916{
1917 struct inode *qf_inode;
1918 unsigned long qf_inum;
1919 int err;
1920
1921 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1922
1923 qf_inum = f2fs_qf_ino(sb, type);
1924 if (!qf_inum)
1925 return -EPERM;
1926
1927 qf_inode = f2fs_iget(sb, qf_inum);
1928 if (IS_ERR(qf_inode)) {
1929 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
1930 return PTR_ERR(qf_inode);
1931 }
1932
1933 /* Don't account quota for quota files to avoid recursion */
1934 qf_inode->i_flags |= S_NOQUOTA;
1935 err = dquot_enable(qf_inode, type, format_id, flags);
1936 iput(qf_inode);
1937 return err;
1938}
1939
1940static int f2fs_enable_quotas(struct super_block *sb)
1941{
1942 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1943 int type, err = 0;
1944 unsigned long qf_inum;
1945 bool quota_mopt[MAXQUOTAS] = {
1946 test_opt(sbi, USRQUOTA),
1947 test_opt(sbi, GRPQUOTA),
1948 test_opt(sbi, PRJQUOTA),
1949 };
1950
1951 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1952 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
1953 return 0;
1954 }
1955
1956 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1957
1958 for (type = 0; type < MAXQUOTAS; type++) {
1959 qf_inum = f2fs_qf_ino(sb, type);
1960 if (qf_inum) {
1961 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1962 DQUOT_USAGE_ENABLED |
1963 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1964 if (err) {
1965 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
1966 type, err);
1967 for (type--; type >= 0; type--)
1968 dquot_quota_off(sb, type);
1969 set_sbi_flag(F2FS_SB(sb),
1970 SBI_QUOTA_NEED_REPAIR);
1971 return err;
1972 }
1973 }
1974 }
1975 return 0;
1976}
1977
1978int f2fs_quota_sync(struct super_block *sb, int type)
1979{
1980 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1981 struct quota_info *dqopt = sb_dqopt(sb);
1982 int cnt;
1983 int ret;
1984
1985 /*
1986 * do_quotactl
1987 * f2fs_quota_sync
1988 * down_read(quota_sem)
1989 * dquot_writeback_dquots()
1990 * f2fs_dquot_commit
1991 * block_operation
1992 * down_read(quota_sem)
1993 */
1994 f2fs_lock_op(sbi);
1995
1996 down_read(&sbi->quota_sem);
1997 ret = dquot_writeback_dquots(sb, type);
1998 if (ret)
1999 goto out;
2000
2001 /*
2002 * Now when everything is written we can discard the pagecache so
2003 * that userspace sees the changes.
2004 */
2005 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2006 struct address_space *mapping;
2007
2008 if (type != -1 && cnt != type)
2009 continue;
2010 if (!sb_has_quota_active(sb, cnt))
2011 continue;
2012
2013 mapping = dqopt->files[cnt]->i_mapping;
2014
2015 ret = filemap_fdatawrite(mapping);
2016 if (ret)
2017 goto out;
2018
2019 /* if we are using journalled quota */
2020 if (is_journalled_quota(sbi))
2021 continue;
2022
2023 ret = filemap_fdatawait(mapping);
2024 if (ret)
2025 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2026
2027 inode_lock(dqopt->files[cnt]);
2028 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2029 inode_unlock(dqopt->files[cnt]);
2030 }
2031out:
2032 if (ret)
2033 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2034 up_read(&sbi->quota_sem);
2035 f2fs_unlock_op(sbi);
2036 return ret;
2037}
2038
2039static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2040 const struct path *path)
2041{
2042 struct inode *inode;
2043 int err;
2044
2045 /* if quota sysfile exists, deny enabling quota with specific file */
2046 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2047 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2048 return -EBUSY;
2049 }
2050
2051 err = f2fs_quota_sync(sb, type);
2052 if (err)
2053 return err;
2054
2055 err = dquot_quota_on(sb, type, format_id, path);
2056 if (err)
2057 return err;
2058
2059 inode = d_inode(path->dentry);
2060
2061 inode_lock(inode);
2062 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2063 f2fs_set_inode_flags(inode);
2064 inode_unlock(inode);
2065 f2fs_mark_inode_dirty_sync(inode, false);
2066
2067 return 0;
2068}
2069
2070static int __f2fs_quota_off(struct super_block *sb, int type)
2071{
2072 struct inode *inode = sb_dqopt(sb)->files[type];
2073 int err;
2074
2075 if (!inode || !igrab(inode))
2076 return dquot_quota_off(sb, type);
2077
2078 err = f2fs_quota_sync(sb, type);
2079 if (err)
2080 goto out_put;
2081
2082 err = dquot_quota_off(sb, type);
2083 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2084 goto out_put;
2085
2086 inode_lock(inode);
2087 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2088 f2fs_set_inode_flags(inode);
2089 inode_unlock(inode);
2090 f2fs_mark_inode_dirty_sync(inode, false);
2091out_put:
2092 iput(inode);
2093 return err;
2094}
2095
2096static int f2fs_quota_off(struct super_block *sb, int type)
2097{
2098 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2099 int err;
2100
2101 err = __f2fs_quota_off(sb, type);
2102
2103 /*
2104 * quotactl can shutdown journalled quota, result in inconsistence
2105 * between quota record and fs data by following updates, tag the
2106 * flag to let fsck be aware of it.
2107 */
2108 if (is_journalled_quota(sbi))
2109 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2110 return err;
2111}
2112
2113void f2fs_quota_off_umount(struct super_block *sb)
2114{
2115 int type;
2116 int err;
2117
2118 for (type = 0; type < MAXQUOTAS; type++) {
2119 err = __f2fs_quota_off(sb, type);
2120 if (err) {
2121 int ret = dquot_quota_off(sb, type);
2122
2123 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2124 type, err, ret);
2125 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2126 }
2127 }
2128 /*
2129 * In case of checkpoint=disable, we must flush quota blocks.
2130 * This can cause NULL exception for node_inode in end_io, since
2131 * put_super already dropped it.
2132 */
2133 sync_filesystem(sb);
2134}
2135
2136static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2137{
2138 struct quota_info *dqopt = sb_dqopt(sb);
2139 int type;
2140
2141 for (type = 0; type < MAXQUOTAS; type++) {
2142 if (!dqopt->files[type])
2143 continue;
2144 f2fs_inode_synced(dqopt->files[type]);
2145 }
2146}
2147
2148static int f2fs_dquot_commit(struct dquot *dquot)
2149{
2150 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2151 int ret;
2152
2153 down_read(&sbi->quota_sem);
2154 ret = dquot_commit(dquot);
2155 if (ret < 0)
2156 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2157 up_read(&sbi->quota_sem);
2158 return ret;
2159}
2160
2161static int f2fs_dquot_acquire(struct dquot *dquot)
2162{
2163 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2164 int ret;
2165
2166 down_read(&sbi->quota_sem);
2167 ret = dquot_acquire(dquot);
2168 if (ret < 0)
2169 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2170 up_read(&sbi->quota_sem);
2171 return ret;
2172}
2173
2174static int f2fs_dquot_release(struct dquot *dquot)
2175{
2176 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2177 int ret;
2178
2179 down_read(&sbi->quota_sem);
2180 ret = dquot_release(dquot);
2181 if (ret < 0)
2182 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2183 up_read(&sbi->quota_sem);
2184 return ret;
2185}
2186
2187static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2188{
2189 struct super_block *sb = dquot->dq_sb;
2190 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2191 int ret;
2192
2193 down_read(&sbi->quota_sem);
2194 ret = dquot_mark_dquot_dirty(dquot);
2195
2196 /* if we are using journalled quota */
2197 if (is_journalled_quota(sbi))
2198 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2199
2200 up_read(&sbi->quota_sem);
2201 return ret;
2202}
2203
2204static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2205{
2206 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2207 int ret;
2208
2209 down_read(&sbi->quota_sem);
2210 ret = dquot_commit_info(sb, type);
2211 if (ret < 0)
2212 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2213 up_read(&sbi->quota_sem);
2214 return ret;
2215}
2216
2217static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2218{
2219 *projid = F2FS_I(inode)->i_projid;
2220 return 0;
2221}
2222
2223static const struct dquot_operations f2fs_quota_operations = {
2224 .get_reserved_space = f2fs_get_reserved_space,
2225 .write_dquot = f2fs_dquot_commit,
2226 .acquire_dquot = f2fs_dquot_acquire,
2227 .release_dquot = f2fs_dquot_release,
2228 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2229 .write_info = f2fs_dquot_commit_info,
2230 .alloc_dquot = dquot_alloc,
2231 .destroy_dquot = dquot_destroy,
2232 .get_projid = f2fs_get_projid,
2233 .get_next_id = dquot_get_next_id,
2234};
2235
2236static const struct quotactl_ops f2fs_quotactl_ops = {
2237 .quota_on = f2fs_quota_on,
2238 .quota_off = f2fs_quota_off,
2239 .quota_sync = f2fs_quota_sync,
2240 .get_state = dquot_get_state,
2241 .set_info = dquot_set_dqinfo,
2242 .get_dqblk = dquot_get_dqblk,
2243 .set_dqblk = dquot_set_dqblk,
2244 .get_nextdqblk = dquot_get_next_dqblk,
2245};
2246#else
2247int f2fs_quota_sync(struct super_block *sb, int type)
2248{
2249 return 0;
2250}
2251
2252void f2fs_quota_off_umount(struct super_block *sb)
2253{
2254}
2255#endif
2256
2257static const struct super_operations f2fs_sops = {
2258 .alloc_inode = f2fs_alloc_inode,
2259 .free_inode = f2fs_free_inode,
2260 .drop_inode = f2fs_drop_inode,
2261 .write_inode = f2fs_write_inode,
2262 .dirty_inode = f2fs_dirty_inode,
2263 .show_options = f2fs_show_options,
2264#ifdef CONFIG_QUOTA
2265 .quota_read = f2fs_quota_read,
2266 .quota_write = f2fs_quota_write,
2267 .get_dquots = f2fs_get_dquots,
2268#endif
2269 .evict_inode = f2fs_evict_inode,
2270 .put_super = f2fs_put_super,
2271 .sync_fs = f2fs_sync_fs,
2272 .freeze_fs = f2fs_freeze,
2273 .unfreeze_fs = f2fs_unfreeze,
2274 .statfs = f2fs_statfs,
2275 .remount_fs = f2fs_remount,
2276};
2277
2278#ifdef CONFIG_FS_ENCRYPTION
2279static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2280{
2281 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2282 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2283 ctx, len, NULL);
2284}
2285
2286static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2287 void *fs_data)
2288{
2289 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2290
2291 /*
2292 * Encrypting the root directory is not allowed because fsck
2293 * expects lost+found directory to exist and remain unencrypted
2294 * if LOST_FOUND feature is enabled.
2295 *
2296 */
2297 if (f2fs_sb_has_lost_found(sbi) &&
2298 inode->i_ino == F2FS_ROOT_INO(sbi))
2299 return -EPERM;
2300
2301 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2302 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2303 ctx, len, fs_data, XATTR_CREATE);
2304}
2305
2306static bool f2fs_dummy_context(struct inode *inode)
2307{
2308 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2309}
2310
2311static const struct fscrypt_operations f2fs_cryptops = {
2312 .key_prefix = "f2fs:",
2313 .get_context = f2fs_get_context,
2314 .set_context = f2fs_set_context,
2315 .dummy_context = f2fs_dummy_context,
2316 .empty_dir = f2fs_empty_dir,
2317 .max_namelen = F2FS_NAME_LEN,
2318};
2319#endif
2320
2321static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2322 u64 ino, u32 generation)
2323{
2324 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2325 struct inode *inode;
2326
2327 if (f2fs_check_nid_range(sbi, ino))
2328 return ERR_PTR(-ESTALE);
2329
2330 /*
2331 * f2fs_iget isn't quite right if the inode is currently unallocated!
2332 * However f2fs_iget currently does appropriate checks to handle stale
2333 * inodes so everything is OK.
2334 */
2335 inode = f2fs_iget(sb, ino);
2336 if (IS_ERR(inode))
2337 return ERR_CAST(inode);
2338 if (unlikely(generation && inode->i_generation != generation)) {
2339 /* we didn't find the right inode.. */
2340 iput(inode);
2341 return ERR_PTR(-ESTALE);
2342 }
2343 return inode;
2344}
2345
2346static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2347 int fh_len, int fh_type)
2348{
2349 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2350 f2fs_nfs_get_inode);
2351}
2352
2353static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2354 int fh_len, int fh_type)
2355{
2356 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2357 f2fs_nfs_get_inode);
2358}
2359
2360static const struct export_operations f2fs_export_ops = {
2361 .fh_to_dentry = f2fs_fh_to_dentry,
2362 .fh_to_parent = f2fs_fh_to_parent,
2363 .get_parent = f2fs_get_parent,
2364};
2365
2366static loff_t max_file_blocks(void)
2367{
2368 loff_t result = 0;
2369 loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2370
2371 /*
2372 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2373 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2374 * space in inode.i_addr, it will be more safe to reassign
2375 * result as zero.
2376 */
2377
2378 /* two direct node blocks */
2379 result += (leaf_count * 2);
2380
2381 /* two indirect node blocks */
2382 leaf_count *= NIDS_PER_BLOCK;
2383 result += (leaf_count * 2);
2384
2385 /* one double indirect node block */
2386 leaf_count *= NIDS_PER_BLOCK;
2387 result += leaf_count;
2388
2389 return result;
2390}
2391
2392static int __f2fs_commit_super(struct buffer_head *bh,
2393 struct f2fs_super_block *super)
2394{
2395 lock_buffer(bh);
2396 if (super)
2397 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2398 set_buffer_dirty(bh);
2399 unlock_buffer(bh);
2400
2401 /* it's rare case, we can do fua all the time */
2402 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2403}
2404
2405static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2406 struct buffer_head *bh)
2407{
2408 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2409 (bh->b_data + F2FS_SUPER_OFFSET);
2410 struct super_block *sb = sbi->sb;
2411 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2412 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2413 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2414 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2415 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2416 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2417 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2418 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2419 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2420 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2421 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2422 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2423 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2424 u64 main_end_blkaddr = main_blkaddr +
2425 (segment_count_main << log_blocks_per_seg);
2426 u64 seg_end_blkaddr = segment0_blkaddr +
2427 (segment_count << log_blocks_per_seg);
2428
2429 if (segment0_blkaddr != cp_blkaddr) {
2430 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2431 segment0_blkaddr, cp_blkaddr);
2432 return true;
2433 }
2434
2435 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2436 sit_blkaddr) {
2437 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2438 cp_blkaddr, sit_blkaddr,
2439 segment_count_ckpt << log_blocks_per_seg);
2440 return true;
2441 }
2442
2443 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2444 nat_blkaddr) {
2445 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2446 sit_blkaddr, nat_blkaddr,
2447 segment_count_sit << log_blocks_per_seg);
2448 return true;
2449 }
2450
2451 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2452 ssa_blkaddr) {
2453 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2454 nat_blkaddr, ssa_blkaddr,
2455 segment_count_nat << log_blocks_per_seg);
2456 return true;
2457 }
2458
2459 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2460 main_blkaddr) {
2461 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2462 ssa_blkaddr, main_blkaddr,
2463 segment_count_ssa << log_blocks_per_seg);
2464 return true;
2465 }
2466
2467 if (main_end_blkaddr > seg_end_blkaddr) {
2468 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2469 main_blkaddr,
2470 segment0_blkaddr +
2471 (segment_count << log_blocks_per_seg),
2472 segment_count_main << log_blocks_per_seg);
2473 return true;
2474 } else if (main_end_blkaddr < seg_end_blkaddr) {
2475 int err = 0;
2476 char *res;
2477
2478 /* fix in-memory information all the time */
2479 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2480 segment0_blkaddr) >> log_blocks_per_seg);
2481
2482 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2483 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2484 res = "internally";
2485 } else {
2486 err = __f2fs_commit_super(bh, NULL);
2487 res = err ? "failed" : "done";
2488 }
2489 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2490 res, main_blkaddr,
2491 segment0_blkaddr +
2492 (segment_count << log_blocks_per_seg),
2493 segment_count_main << log_blocks_per_seg);
2494 if (err)
2495 return true;
2496 }
2497 return false;
2498}
2499
2500static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2501 struct buffer_head *bh)
2502{
2503 block_t segment_count, segs_per_sec, secs_per_zone;
2504 block_t total_sections, blocks_per_seg;
2505 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2506 (bh->b_data + F2FS_SUPER_OFFSET);
2507 unsigned int blocksize;
2508 size_t crc_offset = 0;
2509 __u32 crc = 0;
2510
2511 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2512 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2513 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2514 return -EINVAL;
2515 }
2516
2517 /* Check checksum_offset and crc in superblock */
2518 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2519 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2520 if (crc_offset !=
2521 offsetof(struct f2fs_super_block, crc)) {
2522 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2523 crc_offset);
2524 return -EFSCORRUPTED;
2525 }
2526 crc = le32_to_cpu(raw_super->crc);
2527 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2528 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2529 return -EFSCORRUPTED;
2530 }
2531 }
2532
2533 /* Currently, support only 4KB page cache size */
2534 if (F2FS_BLKSIZE != PAGE_SIZE) {
2535 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2536 PAGE_SIZE);
2537 return -EFSCORRUPTED;
2538 }
2539
2540 /* Currently, support only 4KB block size */
2541 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2542 if (blocksize != F2FS_BLKSIZE) {
2543 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2544 blocksize);
2545 return -EFSCORRUPTED;
2546 }
2547
2548 /* check log blocks per segment */
2549 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2550 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2551 le32_to_cpu(raw_super->log_blocks_per_seg));
2552 return -EFSCORRUPTED;
2553 }
2554
2555 /* Currently, support 512/1024/2048/4096 bytes sector size */
2556 if (le32_to_cpu(raw_super->log_sectorsize) >
2557 F2FS_MAX_LOG_SECTOR_SIZE ||
2558 le32_to_cpu(raw_super->log_sectorsize) <
2559 F2FS_MIN_LOG_SECTOR_SIZE) {
2560 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2561 le32_to_cpu(raw_super->log_sectorsize));
2562 return -EFSCORRUPTED;
2563 }
2564 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2565 le32_to_cpu(raw_super->log_sectorsize) !=
2566 F2FS_MAX_LOG_SECTOR_SIZE) {
2567 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2568 le32_to_cpu(raw_super->log_sectors_per_block),
2569 le32_to_cpu(raw_super->log_sectorsize));
2570 return -EFSCORRUPTED;
2571 }
2572
2573 segment_count = le32_to_cpu(raw_super->segment_count);
2574 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2575 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2576 total_sections = le32_to_cpu(raw_super->section_count);
2577
2578 /* blocks_per_seg should be 512, given the above check */
2579 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2580
2581 if (segment_count > F2FS_MAX_SEGMENT ||
2582 segment_count < F2FS_MIN_SEGMENTS) {
2583 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2584 return -EFSCORRUPTED;
2585 }
2586
2587 if (total_sections > segment_count ||
2588 total_sections < F2FS_MIN_SEGMENTS ||
2589 segs_per_sec > segment_count || !segs_per_sec) {
2590 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2591 segment_count, total_sections, segs_per_sec);
2592 return -EFSCORRUPTED;
2593 }
2594
2595 if ((segment_count / segs_per_sec) < total_sections) {
2596 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2597 segment_count, segs_per_sec, total_sections);
2598 return -EFSCORRUPTED;
2599 }
2600
2601 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2602 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2603 segment_count, le64_to_cpu(raw_super->block_count));
2604 return -EFSCORRUPTED;
2605 }
2606
2607 if (secs_per_zone > total_sections || !secs_per_zone) {
2608 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2609 secs_per_zone, total_sections);
2610 return -EFSCORRUPTED;
2611 }
2612 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2613 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2614 (le32_to_cpu(raw_super->extension_count) +
2615 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2616 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2617 le32_to_cpu(raw_super->extension_count),
2618 raw_super->hot_ext_count,
2619 F2FS_MAX_EXTENSION);
2620 return -EFSCORRUPTED;
2621 }
2622
2623 if (le32_to_cpu(raw_super->cp_payload) >
2624 (blocks_per_seg - F2FS_CP_PACKS)) {
2625 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2626 le32_to_cpu(raw_super->cp_payload),
2627 blocks_per_seg - F2FS_CP_PACKS);
2628 return -EFSCORRUPTED;
2629 }
2630
2631 /* check reserved ino info */
2632 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2633 le32_to_cpu(raw_super->meta_ino) != 2 ||
2634 le32_to_cpu(raw_super->root_ino) != 3) {
2635 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2636 le32_to_cpu(raw_super->node_ino),
2637 le32_to_cpu(raw_super->meta_ino),
2638 le32_to_cpu(raw_super->root_ino));
2639 return -EFSCORRUPTED;
2640 }
2641
2642 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2643 if (sanity_check_area_boundary(sbi, bh))
2644 return -EFSCORRUPTED;
2645
2646 return 0;
2647}
2648
2649int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2650{
2651 unsigned int total, fsmeta;
2652 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2653 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2654 unsigned int ovp_segments, reserved_segments;
2655 unsigned int main_segs, blocks_per_seg;
2656 unsigned int sit_segs, nat_segs;
2657 unsigned int sit_bitmap_size, nat_bitmap_size;
2658 unsigned int log_blocks_per_seg;
2659 unsigned int segment_count_main;
2660 unsigned int cp_pack_start_sum, cp_payload;
2661 block_t user_block_count, valid_user_blocks;
2662 block_t avail_node_count, valid_node_count;
2663 int i, j;
2664
2665 total = le32_to_cpu(raw_super->segment_count);
2666 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2667 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2668 fsmeta += sit_segs;
2669 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2670 fsmeta += nat_segs;
2671 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2672 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2673
2674 if (unlikely(fsmeta >= total))
2675 return 1;
2676
2677 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2678 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2679
2680 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2681 ovp_segments == 0 || reserved_segments == 0)) {
2682 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2683 return 1;
2684 }
2685
2686 user_block_count = le64_to_cpu(ckpt->user_block_count);
2687 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2688 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2689 if (!user_block_count || user_block_count >=
2690 segment_count_main << log_blocks_per_seg) {
2691 f2fs_err(sbi, "Wrong user_block_count: %u",
2692 user_block_count);
2693 return 1;
2694 }
2695
2696 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2697 if (valid_user_blocks > user_block_count) {
2698 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2699 valid_user_blocks, user_block_count);
2700 return 1;
2701 }
2702
2703 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2704 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2705 if (valid_node_count > avail_node_count) {
2706 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2707 valid_node_count, avail_node_count);
2708 return 1;
2709 }
2710
2711 main_segs = le32_to_cpu(raw_super->segment_count_main);
2712 blocks_per_seg = sbi->blocks_per_seg;
2713
2714 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2715 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2716 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2717 return 1;
2718 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2719 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2720 le32_to_cpu(ckpt->cur_node_segno[j])) {
2721 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2722 i, j,
2723 le32_to_cpu(ckpt->cur_node_segno[i]));
2724 return 1;
2725 }
2726 }
2727 }
2728 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2729 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2730 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2731 return 1;
2732 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2733 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2734 le32_to_cpu(ckpt->cur_data_segno[j])) {
2735 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2736 i, j,
2737 le32_to_cpu(ckpt->cur_data_segno[i]));
2738 return 1;
2739 }
2740 }
2741 }
2742 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2743 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2744 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2745 le32_to_cpu(ckpt->cur_data_segno[j])) {
2746 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2747 i, j,
2748 le32_to_cpu(ckpt->cur_node_segno[i]));
2749 return 1;
2750 }
2751 }
2752 }
2753
2754 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2755 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2756
2757 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2758 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2759 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2760 sit_bitmap_size, nat_bitmap_size);
2761 return 1;
2762 }
2763
2764 cp_pack_start_sum = __start_sum_addr(sbi);
2765 cp_payload = __cp_payload(sbi);
2766 if (cp_pack_start_sum < cp_payload + 1 ||
2767 cp_pack_start_sum > blocks_per_seg - 1 -
2768 NR_CURSEG_TYPE) {
2769 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2770 cp_pack_start_sum);
2771 return 1;
2772 }
2773
2774 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2775 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2776 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2777 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2778 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2779 le32_to_cpu(ckpt->checksum_offset));
2780 return 1;
2781 }
2782
2783 if (unlikely(f2fs_cp_error(sbi))) {
2784 f2fs_err(sbi, "A bug case: need to run fsck");
2785 return 1;
2786 }
2787 return 0;
2788}
2789
2790static void init_sb_info(struct f2fs_sb_info *sbi)
2791{
2792 struct f2fs_super_block *raw_super = sbi->raw_super;
2793 int i;
2794
2795 sbi->log_sectors_per_block =
2796 le32_to_cpu(raw_super->log_sectors_per_block);
2797 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2798 sbi->blocksize = 1 << sbi->log_blocksize;
2799 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2800 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2801 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2802 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2803 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2804 sbi->total_node_count =
2805 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2806 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2807 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2808 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2809 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2810 sbi->cur_victim_sec = NULL_SECNO;
2811 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2812 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2813 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2814 sbi->migration_granularity = sbi->segs_per_sec;
2815
2816 sbi->dir_level = DEF_DIR_LEVEL;
2817 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2818 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2819 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2820 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2821 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2822 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2823 DEF_UMOUNT_DISCARD_TIMEOUT;
2824 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2825
2826 for (i = 0; i < NR_COUNT_TYPE; i++)
2827 atomic_set(&sbi->nr_pages[i], 0);
2828
2829 for (i = 0; i < META; i++)
2830 atomic_set(&sbi->wb_sync_req[i], 0);
2831
2832 INIT_LIST_HEAD(&sbi->s_list);
2833 mutex_init(&sbi->umount_mutex);
2834 init_rwsem(&sbi->io_order_lock);
2835 spin_lock_init(&sbi->cp_lock);
2836
2837 sbi->dirty_device = 0;
2838 spin_lock_init(&sbi->dev_lock);
2839
2840 init_rwsem(&sbi->sb_lock);
2841}
2842
2843static int init_percpu_info(struct f2fs_sb_info *sbi)
2844{
2845 int err;
2846
2847 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2848 if (err)
2849 return err;
2850
2851 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2852 GFP_KERNEL);
2853 if (err)
2854 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2855
2856 return err;
2857}
2858
2859#ifdef CONFIG_BLK_DEV_ZONED
2860static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2861{
2862 struct block_device *bdev = FDEV(devi).bdev;
2863 sector_t nr_sectors = bdev->bd_part->nr_sects;
2864 sector_t sector = 0;
2865 struct blk_zone *zones;
2866 unsigned int i, nr_zones;
2867 unsigned int n = 0;
2868 int err = -EIO;
2869
2870 if (!f2fs_sb_has_blkzoned(sbi))
2871 return 0;
2872
2873 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2874 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2875 return -EINVAL;
2876 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2877 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2878 __ilog2_u32(sbi->blocks_per_blkz))
2879 return -EINVAL;
2880 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2881 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2882 sbi->log_blocks_per_blkz;
2883 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2884 FDEV(devi).nr_blkz++;
2885
2886 FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
2887 BITS_TO_LONGS(FDEV(devi).nr_blkz)
2888 * sizeof(unsigned long),
2889 GFP_KERNEL);
2890 if (!FDEV(devi).blkz_seq)
2891 return -ENOMEM;
2892
2893#define F2FS_REPORT_NR_ZONES 4096
2894
2895 zones = f2fs_kzalloc(sbi,
2896 array_size(F2FS_REPORT_NR_ZONES,
2897 sizeof(struct blk_zone)),
2898 GFP_KERNEL);
2899 if (!zones)
2900 return -ENOMEM;
2901
2902 /* Get block zones type */
2903 while (zones && sector < nr_sectors) {
2904
2905 nr_zones = F2FS_REPORT_NR_ZONES;
2906 err = blkdev_report_zones(bdev, sector, zones, &nr_zones);
2907 if (err)
2908 break;
2909 if (!nr_zones) {
2910 err = -EIO;
2911 break;
2912 }
2913
2914 for (i = 0; i < nr_zones; i++) {
2915 if (zones[i].type != BLK_ZONE_TYPE_CONVENTIONAL)
2916 set_bit(n, FDEV(devi).blkz_seq);
2917 sector += zones[i].len;
2918 n++;
2919 }
2920 }
2921
2922 kvfree(zones);
2923
2924 return err;
2925}
2926#endif
2927
2928/*
2929 * Read f2fs raw super block.
2930 * Because we have two copies of super block, so read both of them
2931 * to get the first valid one. If any one of them is broken, we pass
2932 * them recovery flag back to the caller.
2933 */
2934static int read_raw_super_block(struct f2fs_sb_info *sbi,
2935 struct f2fs_super_block **raw_super,
2936 int *valid_super_block, int *recovery)
2937{
2938 struct super_block *sb = sbi->sb;
2939 int block;
2940 struct buffer_head *bh;
2941 struct f2fs_super_block *super;
2942 int err = 0;
2943
2944 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2945 if (!super)
2946 return -ENOMEM;
2947
2948 for (block = 0; block < 2; block++) {
2949 bh = sb_bread(sb, block);
2950 if (!bh) {
2951 f2fs_err(sbi, "Unable to read %dth superblock",
2952 block + 1);
2953 err = -EIO;
2954 continue;
2955 }
2956
2957 /* sanity checking of raw super */
2958 err = sanity_check_raw_super(sbi, bh);
2959 if (err) {
2960 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
2961 block + 1);
2962 brelse(bh);
2963 continue;
2964 }
2965
2966 if (!*raw_super) {
2967 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2968 sizeof(*super));
2969 *valid_super_block = block;
2970 *raw_super = super;
2971 }
2972 brelse(bh);
2973 }
2974
2975 /* Fail to read any one of the superblocks*/
2976 if (err < 0)
2977 *recovery = 1;
2978
2979 /* No valid superblock */
2980 if (!*raw_super)
2981 kvfree(super);
2982 else
2983 err = 0;
2984
2985 return err;
2986}
2987
2988int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2989{
2990 struct buffer_head *bh;
2991 __u32 crc = 0;
2992 int err;
2993
2994 if ((recover && f2fs_readonly(sbi->sb)) ||
2995 bdev_read_only(sbi->sb->s_bdev)) {
2996 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2997 return -EROFS;
2998 }
2999
3000 /* we should update superblock crc here */
3001 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3002 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3003 offsetof(struct f2fs_super_block, crc));
3004 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3005 }
3006
3007 /* write back-up superblock first */
3008 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3009 if (!bh)
3010 return -EIO;
3011 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3012 brelse(bh);
3013
3014 /* if we are in recovery path, skip writing valid superblock */
3015 if (recover || err)
3016 return err;
3017
3018 /* write current valid superblock */
3019 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3020 if (!bh)
3021 return -EIO;
3022 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3023 brelse(bh);
3024 return err;
3025}
3026
3027static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3028{
3029 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3030 unsigned int max_devices = MAX_DEVICES;
3031 int i;
3032
3033 /* Initialize single device information */
3034 if (!RDEV(0).path[0]) {
3035 if (!bdev_is_zoned(sbi->sb->s_bdev))
3036 return 0;
3037 max_devices = 1;
3038 }
3039
3040 /*
3041 * Initialize multiple devices information, or single
3042 * zoned block device information.
3043 */
3044 sbi->devs = f2fs_kzalloc(sbi,
3045 array_size(max_devices,
3046 sizeof(struct f2fs_dev_info)),
3047 GFP_KERNEL);
3048 if (!sbi->devs)
3049 return -ENOMEM;
3050
3051 for (i = 0; i < max_devices; i++) {
3052
3053 if (i > 0 && !RDEV(i).path[0])
3054 break;
3055
3056 if (max_devices == 1) {
3057 /* Single zoned block device mount */
3058 FDEV(0).bdev =
3059 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3060 sbi->sb->s_mode, sbi->sb->s_type);
3061 } else {
3062 /* Multi-device mount */
3063 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3064 FDEV(i).total_segments =
3065 le32_to_cpu(RDEV(i).total_segments);
3066 if (i == 0) {
3067 FDEV(i).start_blk = 0;
3068 FDEV(i).end_blk = FDEV(i).start_blk +
3069 (FDEV(i).total_segments <<
3070 sbi->log_blocks_per_seg) - 1 +
3071 le32_to_cpu(raw_super->segment0_blkaddr);
3072 } else {
3073 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3074 FDEV(i).end_blk = FDEV(i).start_blk +
3075 (FDEV(i).total_segments <<
3076 sbi->log_blocks_per_seg) - 1;
3077 }
3078 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3079 sbi->sb->s_mode, sbi->sb->s_type);
3080 }
3081 if (IS_ERR(FDEV(i).bdev))
3082 return PTR_ERR(FDEV(i).bdev);
3083
3084 /* to release errored devices */
3085 sbi->s_ndevs = i + 1;
3086
3087#ifdef CONFIG_BLK_DEV_ZONED
3088 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3089 !f2fs_sb_has_blkzoned(sbi)) {
3090 f2fs_err(sbi, "Zoned block device feature not enabled\n");
3091 return -EINVAL;
3092 }
3093 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3094 if (init_blkz_info(sbi, i)) {
3095 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3096 return -EINVAL;
3097 }
3098 if (max_devices == 1)
3099 break;
3100 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3101 i, FDEV(i).path,
3102 FDEV(i).total_segments,
3103 FDEV(i).start_blk, FDEV(i).end_blk,
3104 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3105 "Host-aware" : "Host-managed");
3106 continue;
3107 }
3108#endif
3109 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3110 i, FDEV(i).path,
3111 FDEV(i).total_segments,
3112 FDEV(i).start_blk, FDEV(i).end_blk);
3113 }
3114 f2fs_info(sbi,
3115 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3116 return 0;
3117}
3118
3119static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3120{
3121#ifdef CONFIG_UNICODE
3122 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3123 const struct f2fs_sb_encodings *encoding_info;
3124 struct unicode_map *encoding;
3125 __u16 encoding_flags;
3126
3127 if (f2fs_sb_has_encrypt(sbi)) {
3128 f2fs_err(sbi,
3129 "Can't mount with encoding and encryption");
3130 return -EINVAL;
3131 }
3132
3133 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3134 &encoding_flags)) {
3135 f2fs_err(sbi,
3136 "Encoding requested by superblock is unknown");
3137 return -EINVAL;
3138 }
3139
3140 encoding = utf8_load(encoding_info->version);
3141 if (IS_ERR(encoding)) {
3142 f2fs_err(sbi,
3143 "can't mount with superblock charset: %s-%s "
3144 "not supported by the kernel. flags: 0x%x.",
3145 encoding_info->name, encoding_info->version,
3146 encoding_flags);
3147 return PTR_ERR(encoding);
3148 }
3149 f2fs_info(sbi, "Using encoding defined by superblock: "
3150 "%s-%s with flags 0x%hx", encoding_info->name,
3151 encoding_info->version?:"\b", encoding_flags);
3152
3153 sbi->s_encoding = encoding;
3154 sbi->s_encoding_flags = encoding_flags;
3155 sbi->sb->s_d_op = &f2fs_dentry_ops;
3156 }
3157#else
3158 if (f2fs_sb_has_casefold(sbi)) {
3159 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3160 return -EINVAL;
3161 }
3162#endif
3163 return 0;
3164}
3165
3166static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3167{
3168 struct f2fs_sm_info *sm_i = SM_I(sbi);
3169
3170 /* adjust parameters according to the volume size */
3171 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3172 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3173 sm_i->dcc_info->discard_granularity = 1;
3174 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3175 }
3176
3177 sbi->readdir_ra = 1;
3178}
3179
3180static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3181{
3182 struct f2fs_sb_info *sbi;
3183 struct f2fs_super_block *raw_super;
3184 struct inode *root;
3185 int err;
3186 bool skip_recovery = false, need_fsck = false;
3187 char *options = NULL;
3188 int recovery, i, valid_super_block;
3189 struct curseg_info *seg_i;
3190 int retry_cnt = 1;
3191
3192try_onemore:
3193 err = -EINVAL;
3194 raw_super = NULL;
3195 valid_super_block = -1;
3196 recovery = 0;
3197
3198 /* allocate memory for f2fs-specific super block info */
3199 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3200 if (!sbi)
3201 return -ENOMEM;
3202
3203 sbi->sb = sb;
3204
3205 /* Load the checksum driver */
3206 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3207 if (IS_ERR(sbi->s_chksum_driver)) {
3208 f2fs_err(sbi, "Cannot load crc32 driver.");
3209 err = PTR_ERR(sbi->s_chksum_driver);
3210 sbi->s_chksum_driver = NULL;
3211 goto free_sbi;
3212 }
3213
3214 /* set a block size */
3215 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3216 f2fs_err(sbi, "unable to set blocksize");
3217 goto free_sbi;
3218 }
3219
3220 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3221 &recovery);
3222 if (err)
3223 goto free_sbi;
3224
3225 sb->s_fs_info = sbi;
3226 sbi->raw_super = raw_super;
3227
3228 /* precompute checksum seed for metadata */
3229 if (f2fs_sb_has_inode_chksum(sbi))
3230 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3231 sizeof(raw_super->uuid));
3232
3233 /*
3234 * The BLKZONED feature indicates that the drive was formatted with
3235 * zone alignment optimization. This is optional for host-aware
3236 * devices, but mandatory for host-managed zoned block devices.
3237 */
3238#ifndef CONFIG_BLK_DEV_ZONED
3239 if (f2fs_sb_has_blkzoned(sbi)) {
3240 f2fs_err(sbi, "Zoned block device support is not enabled");
3241 err = -EOPNOTSUPP;
3242 goto free_sb_buf;
3243 }
3244#endif
3245 default_options(sbi);
3246 /* parse mount options */
3247 options = kstrdup((const char *)data, GFP_KERNEL);
3248 if (data && !options) {
3249 err = -ENOMEM;
3250 goto free_sb_buf;
3251 }
3252
3253 err = parse_options(sb, options);
3254 if (err)
3255 goto free_options;
3256
3257 sbi->max_file_blocks = max_file_blocks();
3258 sb->s_maxbytes = sbi->max_file_blocks <<
3259 le32_to_cpu(raw_super->log_blocksize);
3260 sb->s_max_links = F2FS_LINK_MAX;
3261
3262 err = f2fs_setup_casefold(sbi);
3263 if (err)
3264 goto free_options;
3265
3266#ifdef CONFIG_QUOTA
3267 sb->dq_op = &f2fs_quota_operations;
3268 sb->s_qcop = &f2fs_quotactl_ops;
3269 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3270
3271 if (f2fs_sb_has_quota_ino(sbi)) {
3272 for (i = 0; i < MAXQUOTAS; i++) {
3273 if (f2fs_qf_ino(sbi->sb, i))
3274 sbi->nquota_files++;
3275 }
3276 }
3277#endif
3278
3279 sb->s_op = &f2fs_sops;
3280#ifdef CONFIG_FS_ENCRYPTION
3281 sb->s_cop = &f2fs_cryptops;
3282#endif
3283#ifdef CONFIG_FS_VERITY
3284 sb->s_vop = &f2fs_verityops;
3285#endif
3286 sb->s_xattr = f2fs_xattr_handlers;
3287 sb->s_export_op = &f2fs_export_ops;
3288 sb->s_magic = F2FS_SUPER_MAGIC;
3289 sb->s_time_gran = 1;
3290 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3291 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3292 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3293 sb->s_iflags |= SB_I_CGROUPWB;
3294
3295 /* init f2fs-specific super block info */
3296 sbi->valid_super_block = valid_super_block;
3297 mutex_init(&sbi->gc_mutex);
3298 mutex_init(&sbi->writepages);
3299 mutex_init(&sbi->cp_mutex);
3300 mutex_init(&sbi->resize_mutex);
3301 init_rwsem(&sbi->node_write);
3302 init_rwsem(&sbi->node_change);
3303
3304 /* disallow all the data/node/meta page writes */
3305 set_sbi_flag(sbi, SBI_POR_DOING);
3306 spin_lock_init(&sbi->stat_lock);
3307
3308 /* init iostat info */
3309 spin_lock_init(&sbi->iostat_lock);
3310 sbi->iostat_enable = false;
3311
3312 for (i = 0; i < NR_PAGE_TYPE; i++) {
3313 int n = (i == META) ? 1: NR_TEMP_TYPE;
3314 int j;
3315
3316 sbi->write_io[i] =
3317 f2fs_kmalloc(sbi,
3318 array_size(n,
3319 sizeof(struct f2fs_bio_info)),
3320 GFP_KERNEL);
3321 if (!sbi->write_io[i]) {
3322 err = -ENOMEM;
3323 goto free_bio_info;
3324 }
3325
3326 for (j = HOT; j < n; j++) {
3327 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3328 sbi->write_io[i][j].sbi = sbi;
3329 sbi->write_io[i][j].bio = NULL;
3330 spin_lock_init(&sbi->write_io[i][j].io_lock);
3331 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3332 }
3333 }
3334
3335 init_rwsem(&sbi->cp_rwsem);
3336 init_rwsem(&sbi->quota_sem);
3337 init_waitqueue_head(&sbi->cp_wait);
3338 init_sb_info(sbi);
3339
3340 err = init_percpu_info(sbi);
3341 if (err)
3342 goto free_bio_info;
3343
3344 if (F2FS_IO_ALIGNED(sbi)) {
3345 sbi->write_io_dummy =
3346 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3347 if (!sbi->write_io_dummy) {
3348 err = -ENOMEM;
3349 goto free_percpu;
3350 }
3351 }
3352
3353 /* get an inode for meta space */
3354 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3355 if (IS_ERR(sbi->meta_inode)) {
3356 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3357 err = PTR_ERR(sbi->meta_inode);
3358 goto free_io_dummy;
3359 }
3360
3361 err = f2fs_get_valid_checkpoint(sbi);
3362 if (err) {
3363 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3364 goto free_meta_inode;
3365 }
3366
3367 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3368 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3369 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3370 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3371 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3372 }
3373
3374 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3375 set_sbi_flag(sbi, SBI_NEED_FSCK);
3376
3377 /* Initialize device list */
3378 err = f2fs_scan_devices(sbi);
3379 if (err) {
3380 f2fs_err(sbi, "Failed to find devices");
3381 goto free_devices;
3382 }
3383
3384 sbi->total_valid_node_count =
3385 le32_to_cpu(sbi->ckpt->valid_node_count);
3386 percpu_counter_set(&sbi->total_valid_inode_count,
3387 le32_to_cpu(sbi->ckpt->valid_inode_count));
3388 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3389 sbi->total_valid_block_count =
3390 le64_to_cpu(sbi->ckpt->valid_block_count);
3391 sbi->last_valid_block_count = sbi->total_valid_block_count;
3392 sbi->reserved_blocks = 0;
3393 sbi->current_reserved_blocks = 0;
3394 limit_reserve_root(sbi);
3395
3396 for (i = 0; i < NR_INODE_TYPE; i++) {
3397 INIT_LIST_HEAD(&sbi->inode_list[i]);
3398 spin_lock_init(&sbi->inode_lock[i]);
3399 }
3400 mutex_init(&sbi->flush_lock);
3401
3402 f2fs_init_extent_cache_info(sbi);
3403
3404 f2fs_init_ino_entry_info(sbi);
3405
3406 f2fs_init_fsync_node_info(sbi);
3407
3408 /* setup f2fs internal modules */
3409 err = f2fs_build_segment_manager(sbi);
3410 if (err) {
3411 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3412 err);
3413 goto free_sm;
3414 }
3415 err = f2fs_build_node_manager(sbi);
3416 if (err) {
3417 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3418 err);
3419 goto free_nm;
3420 }
3421
3422 /* For write statistics */
3423 if (sb->s_bdev->bd_part)
3424 sbi->sectors_written_start =
3425 (u64)part_stat_read(sb->s_bdev->bd_part,
3426 sectors[STAT_WRITE]);
3427
3428 /* Read accumulated write IO statistics if exists */
3429 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3430 if (__exist_node_summaries(sbi))
3431 sbi->kbytes_written =
3432 le64_to_cpu(seg_i->journal->info.kbytes_written);
3433
3434 f2fs_build_gc_manager(sbi);
3435
3436 err = f2fs_build_stats(sbi);
3437 if (err)
3438 goto free_nm;
3439
3440 /* get an inode for node space */
3441 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3442 if (IS_ERR(sbi->node_inode)) {
3443 f2fs_err(sbi, "Failed to read node inode");
3444 err = PTR_ERR(sbi->node_inode);
3445 goto free_stats;
3446 }
3447
3448 /* read root inode and dentry */
3449 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3450 if (IS_ERR(root)) {
3451 f2fs_err(sbi, "Failed to read root inode");
3452 err = PTR_ERR(root);
3453 goto free_node_inode;
3454 }
3455 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3456 !root->i_size || !root->i_nlink) {
3457 iput(root);
3458 err = -EINVAL;
3459 goto free_node_inode;
3460 }
3461
3462 sb->s_root = d_make_root(root); /* allocate root dentry */
3463 if (!sb->s_root) {
3464 err = -ENOMEM;
3465 goto free_node_inode;
3466 }
3467
3468 err = f2fs_register_sysfs(sbi);
3469 if (err)
3470 goto free_root_inode;
3471
3472#ifdef CONFIG_QUOTA
3473 /* Enable quota usage during mount */
3474 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3475 err = f2fs_enable_quotas(sb);
3476 if (err)
3477 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3478 }
3479#endif
3480 /* if there are nt orphan nodes free them */
3481 err = f2fs_recover_orphan_inodes(sbi);
3482 if (err)
3483 goto free_meta;
3484
3485 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3486 goto reset_checkpoint;
3487
3488 /* recover fsynced data */
3489 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3490 /*
3491 * mount should be failed, when device has readonly mode, and
3492 * previous checkpoint was not done by clean system shutdown.
3493 */
3494 if (f2fs_hw_is_readonly(sbi)) {
3495 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3496 err = -EROFS;
3497 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3498 goto free_meta;
3499 }
3500 f2fs_info(sbi, "write access unavailable, skipping recovery");
3501 goto reset_checkpoint;
3502 }
3503
3504 if (need_fsck)
3505 set_sbi_flag(sbi, SBI_NEED_FSCK);
3506
3507 if (skip_recovery)
3508 goto reset_checkpoint;
3509
3510 err = f2fs_recover_fsync_data(sbi, false);
3511 if (err < 0) {
3512 if (err != -ENOMEM)
3513 skip_recovery = true;
3514 need_fsck = true;
3515 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3516 err);
3517 goto free_meta;
3518 }
3519 } else {
3520 err = f2fs_recover_fsync_data(sbi, true);
3521
3522 if (!f2fs_readonly(sb) && err > 0) {
3523 err = -EINVAL;
3524 f2fs_err(sbi, "Need to recover fsync data");
3525 goto free_meta;
3526 }
3527 }
3528reset_checkpoint:
3529 /* f2fs_recover_fsync_data() cleared this already */
3530 clear_sbi_flag(sbi, SBI_POR_DOING);
3531
3532 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3533 err = f2fs_disable_checkpoint(sbi);
3534 if (err)
3535 goto sync_free_meta;
3536 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3537 f2fs_enable_checkpoint(sbi);
3538 }
3539
3540 /*
3541 * If filesystem is not mounted as read-only then
3542 * do start the gc_thread.
3543 */
3544 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3545 /* After POR, we can run background GC thread.*/
3546 err = f2fs_start_gc_thread(sbi);
3547 if (err)
3548 goto sync_free_meta;
3549 }
3550 kvfree(options);
3551
3552 /* recover broken superblock */
3553 if (recovery) {
3554 err = f2fs_commit_super(sbi, true);
3555 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3556 sbi->valid_super_block ? 1 : 2, err);
3557 }
3558
3559 f2fs_join_shrinker(sbi);
3560
3561 f2fs_tuning_parameters(sbi);
3562
3563 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3564 cur_cp_version(F2FS_CKPT(sbi)));
3565 f2fs_update_time(sbi, CP_TIME);
3566 f2fs_update_time(sbi, REQ_TIME);
3567 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3568 return 0;
3569
3570sync_free_meta:
3571 /* safe to flush all the data */
3572 sync_filesystem(sbi->sb);
3573 retry_cnt = 0;
3574
3575free_meta:
3576#ifdef CONFIG_QUOTA
3577 f2fs_truncate_quota_inode_pages(sb);
3578 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3579 f2fs_quota_off_umount(sbi->sb);
3580#endif
3581 /*
3582 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3583 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3584 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3585 * falls into an infinite loop in f2fs_sync_meta_pages().
3586 */
3587 truncate_inode_pages_final(META_MAPPING(sbi));
3588 /* evict some inodes being cached by GC */
3589 evict_inodes(sb);
3590 f2fs_unregister_sysfs(sbi);
3591free_root_inode:
3592 dput(sb->s_root);
3593 sb->s_root = NULL;
3594free_node_inode:
3595 f2fs_release_ino_entry(sbi, true);
3596 truncate_inode_pages_final(NODE_MAPPING(sbi));
3597 iput(sbi->node_inode);
3598 sbi->node_inode = NULL;
3599free_stats:
3600 f2fs_destroy_stats(sbi);
3601free_nm:
3602 f2fs_destroy_node_manager(sbi);
3603free_sm:
3604 f2fs_destroy_segment_manager(sbi);
3605free_devices:
3606 destroy_device_list(sbi);
3607 kvfree(sbi->ckpt);
3608free_meta_inode:
3609 make_bad_inode(sbi->meta_inode);
3610 iput(sbi->meta_inode);
3611 sbi->meta_inode = NULL;
3612free_io_dummy:
3613 mempool_destroy(sbi->write_io_dummy);
3614free_percpu:
3615 destroy_percpu_info(sbi);
3616free_bio_info:
3617 for (i = 0; i < NR_PAGE_TYPE; i++)
3618 kvfree(sbi->write_io[i]);
3619
3620#ifdef CONFIG_UNICODE
3621 utf8_unload(sbi->s_encoding);
3622#endif
3623free_options:
3624#ifdef CONFIG_QUOTA
3625 for (i = 0; i < MAXQUOTAS; i++)
3626 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3627#endif
3628 kvfree(options);
3629free_sb_buf:
3630 kvfree(raw_super);
3631free_sbi:
3632 if (sbi->s_chksum_driver)
3633 crypto_free_shash(sbi->s_chksum_driver);
3634 kvfree(sbi);
3635
3636 /* give only one another chance */
3637 if (retry_cnt > 0 && skip_recovery) {
3638 retry_cnt--;
3639 shrink_dcache_sb(sb);
3640 goto try_onemore;
3641 }
3642 return err;
3643}
3644
3645static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3646 const char *dev_name, void *data)
3647{
3648 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3649}
3650
3651static void kill_f2fs_super(struct super_block *sb)
3652{
3653 if (sb->s_root) {
3654 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3655
3656 set_sbi_flag(sbi, SBI_IS_CLOSE);
3657 f2fs_stop_gc_thread(sbi);
3658 f2fs_stop_discard_thread(sbi);
3659
3660 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3661 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3662 struct cp_control cpc = {
3663 .reason = CP_UMOUNT,
3664 };
3665 f2fs_write_checkpoint(sbi, &cpc);
3666 }
3667
3668 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3669 sb->s_flags &= ~SB_RDONLY;
3670 }
3671 kill_block_super(sb);
3672}
3673
3674static struct file_system_type f2fs_fs_type = {
3675 .owner = THIS_MODULE,
3676 .name = "f2fs",
3677 .mount = f2fs_mount,
3678 .kill_sb = kill_f2fs_super,
3679 .fs_flags = FS_REQUIRES_DEV,
3680};
3681MODULE_ALIAS_FS("f2fs");
3682
3683static int __init init_inodecache(void)
3684{
3685 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3686 sizeof(struct f2fs_inode_info), 0,
3687 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3688 if (!f2fs_inode_cachep)
3689 return -ENOMEM;
3690 return 0;
3691}
3692
3693static void destroy_inodecache(void)
3694{
3695 /*
3696 * Make sure all delayed rcu free inodes are flushed before we
3697 * destroy cache.
3698 */
3699 rcu_barrier();
3700 kmem_cache_destroy(f2fs_inode_cachep);
3701}
3702
3703static int __init init_f2fs_fs(void)
3704{
3705 int err;
3706
3707 if (PAGE_SIZE != F2FS_BLKSIZE) {
3708 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3709 PAGE_SIZE, F2FS_BLKSIZE);
3710 return -EINVAL;
3711 }
3712
3713 f2fs_build_trace_ios();
3714
3715 err = init_inodecache();
3716 if (err)
3717 goto fail;
3718 err = f2fs_create_node_manager_caches();
3719 if (err)
3720 goto free_inodecache;
3721 err = f2fs_create_segment_manager_caches();
3722 if (err)
3723 goto free_node_manager_caches;
3724 err = f2fs_create_checkpoint_caches();
3725 if (err)
3726 goto free_segment_manager_caches;
3727 err = f2fs_create_extent_cache();
3728 if (err)
3729 goto free_checkpoint_caches;
3730 err = f2fs_init_sysfs();
3731 if (err)
3732 goto free_extent_cache;
3733 err = register_shrinker(&f2fs_shrinker_info);
3734 if (err)
3735 goto free_sysfs;
3736 err = register_filesystem(&f2fs_fs_type);
3737 if (err)
3738 goto free_shrinker;
3739 f2fs_create_root_stats();
3740 err = f2fs_init_post_read_processing();
3741 if (err)
3742 goto free_root_stats;
3743 return 0;
3744
3745free_root_stats:
3746 f2fs_destroy_root_stats();
3747 unregister_filesystem(&f2fs_fs_type);
3748free_shrinker:
3749 unregister_shrinker(&f2fs_shrinker_info);
3750free_sysfs:
3751 f2fs_exit_sysfs();
3752free_extent_cache:
3753 f2fs_destroy_extent_cache();
3754free_checkpoint_caches:
3755 f2fs_destroy_checkpoint_caches();
3756free_segment_manager_caches:
3757 f2fs_destroy_segment_manager_caches();
3758free_node_manager_caches:
3759 f2fs_destroy_node_manager_caches();
3760free_inodecache:
3761 destroy_inodecache();
3762fail:
3763 return err;
3764}
3765
3766static void __exit exit_f2fs_fs(void)
3767{
3768 f2fs_destroy_post_read_processing();
3769 f2fs_destroy_root_stats();
3770 unregister_filesystem(&f2fs_fs_type);
3771 unregister_shrinker(&f2fs_shrinker_info);
3772 f2fs_exit_sysfs();
3773 f2fs_destroy_extent_cache();
3774 f2fs_destroy_checkpoint_caches();
3775 f2fs_destroy_segment_manager_caches();
3776 f2fs_destroy_node_manager_caches();
3777 destroy_inodecache();
3778 f2fs_destroy_trace_ios();
3779}
3780
3781module_init(init_f2fs_fs)
3782module_exit(exit_f2fs_fs)
3783
3784MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3785MODULE_DESCRIPTION("Flash Friendly File System");
3786MODULE_LICENSE("GPL");
3787