Loading...
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
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/fs_context.h>
12#include <linux/sched/mm.h>
13#include <linux/statfs.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#include <linux/part_stat.h>
28#include <linux/zstd.h>
29#include <linux/lz4.h>
30
31#include "f2fs.h"
32#include "node.h"
33#include "segment.h"
34#include "xattr.h"
35#include "gc.h"
36#include "iostat.h"
37
38#define CREATE_TRACE_POINTS
39#include <trace/events/f2fs.h>
40
41static struct kmem_cache *f2fs_inode_cachep;
42
43#ifdef CONFIG_F2FS_FAULT_INJECTION
44
45const char *f2fs_fault_name[FAULT_MAX] = {
46 [FAULT_KMALLOC] = "kmalloc",
47 [FAULT_KVMALLOC] = "kvmalloc",
48 [FAULT_PAGE_ALLOC] = "page alloc",
49 [FAULT_PAGE_GET] = "page get",
50 [FAULT_ALLOC_NID] = "alloc nid",
51 [FAULT_ORPHAN] = "orphan",
52 [FAULT_BLOCK] = "no more block",
53 [FAULT_DIR_DEPTH] = "too big dir depth",
54 [FAULT_EVICT_INODE] = "evict_inode fail",
55 [FAULT_TRUNCATE] = "truncate fail",
56 [FAULT_READ_IO] = "read IO error",
57 [FAULT_CHECKPOINT] = "checkpoint error",
58 [FAULT_DISCARD] = "discard error",
59 [FAULT_WRITE_IO] = "write IO error",
60 [FAULT_SLAB_ALLOC] = "slab alloc",
61 [FAULT_DQUOT_INIT] = "dquot initialize",
62 [FAULT_LOCK_OP] = "lock_op",
63 [FAULT_BLKADDR_VALIDITY] = "invalid blkaddr",
64 [FAULT_BLKADDR_CONSISTENCE] = "inconsistent blkaddr",
65 [FAULT_NO_SEGMENT] = "no free segment",
66};
67
68int f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned long rate,
69 unsigned long type)
70{
71 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
72
73 if (rate) {
74 if (rate > INT_MAX)
75 return -EINVAL;
76 atomic_set(&ffi->inject_ops, 0);
77 ffi->inject_rate = (int)rate;
78 }
79
80 if (type) {
81 if (type >= BIT(FAULT_MAX))
82 return -EINVAL;
83 ffi->inject_type = (unsigned int)type;
84 }
85
86 if (!rate && !type)
87 memset(ffi, 0, sizeof(struct f2fs_fault_info));
88 else
89 f2fs_info(sbi,
90 "build fault injection attr: rate: %lu, type: 0x%lx",
91 rate, type);
92 return 0;
93}
94#endif
95
96/* f2fs-wide shrinker description */
97static struct shrinker *f2fs_shrinker_info;
98
99static int __init f2fs_init_shrinker(void)
100{
101 f2fs_shrinker_info = shrinker_alloc(0, "f2fs-shrinker");
102 if (!f2fs_shrinker_info)
103 return -ENOMEM;
104
105 f2fs_shrinker_info->count_objects = f2fs_shrink_count;
106 f2fs_shrinker_info->scan_objects = f2fs_shrink_scan;
107
108 shrinker_register(f2fs_shrinker_info);
109
110 return 0;
111}
112
113static void f2fs_exit_shrinker(void)
114{
115 shrinker_free(f2fs_shrinker_info);
116}
117
118enum {
119 Opt_gc_background,
120 Opt_disable_roll_forward,
121 Opt_norecovery,
122 Opt_discard,
123 Opt_nodiscard,
124 Opt_noheap,
125 Opt_heap,
126 Opt_user_xattr,
127 Opt_nouser_xattr,
128 Opt_acl,
129 Opt_noacl,
130 Opt_active_logs,
131 Opt_disable_ext_identify,
132 Opt_inline_xattr,
133 Opt_noinline_xattr,
134 Opt_inline_xattr_size,
135 Opt_inline_data,
136 Opt_inline_dentry,
137 Opt_noinline_dentry,
138 Opt_flush_merge,
139 Opt_noflush_merge,
140 Opt_barrier,
141 Opt_nobarrier,
142 Opt_fastboot,
143 Opt_extent_cache,
144 Opt_noextent_cache,
145 Opt_noinline_data,
146 Opt_data_flush,
147 Opt_reserve_root,
148 Opt_resgid,
149 Opt_resuid,
150 Opt_mode,
151 Opt_fault_injection,
152 Opt_fault_type,
153 Opt_lazytime,
154 Opt_nolazytime,
155 Opt_quota,
156 Opt_noquota,
157 Opt_usrquota,
158 Opt_grpquota,
159 Opt_prjquota,
160 Opt_usrjquota,
161 Opt_grpjquota,
162 Opt_prjjquota,
163 Opt_offusrjquota,
164 Opt_offgrpjquota,
165 Opt_offprjjquota,
166 Opt_jqfmt_vfsold,
167 Opt_jqfmt_vfsv0,
168 Opt_jqfmt_vfsv1,
169 Opt_alloc,
170 Opt_fsync,
171 Opt_test_dummy_encryption,
172 Opt_inlinecrypt,
173 Opt_checkpoint_disable,
174 Opt_checkpoint_disable_cap,
175 Opt_checkpoint_disable_cap_perc,
176 Opt_checkpoint_enable,
177 Opt_checkpoint_merge,
178 Opt_nocheckpoint_merge,
179 Opt_compress_algorithm,
180 Opt_compress_log_size,
181 Opt_compress_extension,
182 Opt_nocompress_extension,
183 Opt_compress_chksum,
184 Opt_compress_mode,
185 Opt_compress_cache,
186 Opt_atgc,
187 Opt_gc_merge,
188 Opt_nogc_merge,
189 Opt_discard_unit,
190 Opt_memory_mode,
191 Opt_age_extent_cache,
192 Opt_errors,
193 Opt_err,
194};
195
196static match_table_t f2fs_tokens = {
197 {Opt_gc_background, "background_gc=%s"},
198 {Opt_disable_roll_forward, "disable_roll_forward"},
199 {Opt_norecovery, "norecovery"},
200 {Opt_discard, "discard"},
201 {Opt_nodiscard, "nodiscard"},
202 {Opt_noheap, "no_heap"},
203 {Opt_heap, "heap"},
204 {Opt_user_xattr, "user_xattr"},
205 {Opt_nouser_xattr, "nouser_xattr"},
206 {Opt_acl, "acl"},
207 {Opt_noacl, "noacl"},
208 {Opt_active_logs, "active_logs=%u"},
209 {Opt_disable_ext_identify, "disable_ext_identify"},
210 {Opt_inline_xattr, "inline_xattr"},
211 {Opt_noinline_xattr, "noinline_xattr"},
212 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
213 {Opt_inline_data, "inline_data"},
214 {Opt_inline_dentry, "inline_dentry"},
215 {Opt_noinline_dentry, "noinline_dentry"},
216 {Opt_flush_merge, "flush_merge"},
217 {Opt_noflush_merge, "noflush_merge"},
218 {Opt_barrier, "barrier"},
219 {Opt_nobarrier, "nobarrier"},
220 {Opt_fastboot, "fastboot"},
221 {Opt_extent_cache, "extent_cache"},
222 {Opt_noextent_cache, "noextent_cache"},
223 {Opt_noinline_data, "noinline_data"},
224 {Opt_data_flush, "data_flush"},
225 {Opt_reserve_root, "reserve_root=%u"},
226 {Opt_resgid, "resgid=%u"},
227 {Opt_resuid, "resuid=%u"},
228 {Opt_mode, "mode=%s"},
229 {Opt_fault_injection, "fault_injection=%u"},
230 {Opt_fault_type, "fault_type=%u"},
231 {Opt_lazytime, "lazytime"},
232 {Opt_nolazytime, "nolazytime"},
233 {Opt_quota, "quota"},
234 {Opt_noquota, "noquota"},
235 {Opt_usrquota, "usrquota"},
236 {Opt_grpquota, "grpquota"},
237 {Opt_prjquota, "prjquota"},
238 {Opt_usrjquota, "usrjquota=%s"},
239 {Opt_grpjquota, "grpjquota=%s"},
240 {Opt_prjjquota, "prjjquota=%s"},
241 {Opt_offusrjquota, "usrjquota="},
242 {Opt_offgrpjquota, "grpjquota="},
243 {Opt_offprjjquota, "prjjquota="},
244 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
245 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
246 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
247 {Opt_alloc, "alloc_mode=%s"},
248 {Opt_fsync, "fsync_mode=%s"},
249 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
250 {Opt_test_dummy_encryption, "test_dummy_encryption"},
251 {Opt_inlinecrypt, "inlinecrypt"},
252 {Opt_checkpoint_disable, "checkpoint=disable"},
253 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
254 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
255 {Opt_checkpoint_enable, "checkpoint=enable"},
256 {Opt_checkpoint_merge, "checkpoint_merge"},
257 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
258 {Opt_compress_algorithm, "compress_algorithm=%s"},
259 {Opt_compress_log_size, "compress_log_size=%u"},
260 {Opt_compress_extension, "compress_extension=%s"},
261 {Opt_nocompress_extension, "nocompress_extension=%s"},
262 {Opt_compress_chksum, "compress_chksum"},
263 {Opt_compress_mode, "compress_mode=%s"},
264 {Opt_compress_cache, "compress_cache"},
265 {Opt_atgc, "atgc"},
266 {Opt_gc_merge, "gc_merge"},
267 {Opt_nogc_merge, "nogc_merge"},
268 {Opt_discard_unit, "discard_unit=%s"},
269 {Opt_memory_mode, "memory=%s"},
270 {Opt_age_extent_cache, "age_extent_cache"},
271 {Opt_errors, "errors=%s"},
272 {Opt_err, NULL},
273};
274
275void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate,
276 const char *fmt, ...)
277{
278 struct va_format vaf;
279 va_list args;
280 int level;
281
282 va_start(args, fmt);
283
284 level = printk_get_level(fmt);
285 vaf.fmt = printk_skip_level(fmt);
286 vaf.va = &args;
287 if (limit_rate)
288 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n",
289 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
290 else
291 printk("%c%cF2FS-fs (%s): %pV\n",
292 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
293
294 va_end(args);
295}
296
297#if IS_ENABLED(CONFIG_UNICODE)
298static const struct f2fs_sb_encodings {
299 __u16 magic;
300 char *name;
301 unsigned int version;
302} f2fs_sb_encoding_map[] = {
303 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
304};
305
306static const struct f2fs_sb_encodings *
307f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
308{
309 __u16 magic = le16_to_cpu(sb->s_encoding);
310 int i;
311
312 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
313 if (magic == f2fs_sb_encoding_map[i].magic)
314 return &f2fs_sb_encoding_map[i];
315
316 return NULL;
317}
318
319struct kmem_cache *f2fs_cf_name_slab;
320static int __init f2fs_create_casefold_cache(void)
321{
322 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
323 F2FS_NAME_LEN);
324 return f2fs_cf_name_slab ? 0 : -ENOMEM;
325}
326
327static void f2fs_destroy_casefold_cache(void)
328{
329 kmem_cache_destroy(f2fs_cf_name_slab);
330}
331#else
332static int __init f2fs_create_casefold_cache(void) { return 0; }
333static void f2fs_destroy_casefold_cache(void) { }
334#endif
335
336static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
337{
338 block_t limit = min((sbi->user_block_count >> 3),
339 sbi->user_block_count - sbi->reserved_blocks);
340
341 /* limit is 12.5% */
342 if (test_opt(sbi, RESERVE_ROOT) &&
343 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
344 F2FS_OPTION(sbi).root_reserved_blocks = limit;
345 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
346 F2FS_OPTION(sbi).root_reserved_blocks);
347 }
348 if (!test_opt(sbi, RESERVE_ROOT) &&
349 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
350 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
351 !gid_eq(F2FS_OPTION(sbi).s_resgid,
352 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
353 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
354 from_kuid_munged(&init_user_ns,
355 F2FS_OPTION(sbi).s_resuid),
356 from_kgid_munged(&init_user_ns,
357 F2FS_OPTION(sbi).s_resgid));
358}
359
360static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
361{
362 if (!F2FS_OPTION(sbi).unusable_cap_perc)
363 return;
364
365 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
366 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
367 else
368 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
369 F2FS_OPTION(sbi).unusable_cap_perc;
370
371 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
372 F2FS_OPTION(sbi).unusable_cap,
373 F2FS_OPTION(sbi).unusable_cap_perc);
374}
375
376static void init_once(void *foo)
377{
378 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
379
380 inode_init_once(&fi->vfs_inode);
381}
382
383#ifdef CONFIG_QUOTA
384static const char * const quotatypes[] = INITQFNAMES;
385#define QTYPE2NAME(t) (quotatypes[t])
386static int f2fs_set_qf_name(struct super_block *sb, int qtype,
387 substring_t *args)
388{
389 struct f2fs_sb_info *sbi = F2FS_SB(sb);
390 char *qname;
391 int ret = -EINVAL;
392
393 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
394 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
395 return -EINVAL;
396 }
397 if (f2fs_sb_has_quota_ino(sbi)) {
398 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
399 return 0;
400 }
401
402 qname = match_strdup(args);
403 if (!qname) {
404 f2fs_err(sbi, "Not enough memory for storing quotafile name");
405 return -ENOMEM;
406 }
407 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
408 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
409 ret = 0;
410 else
411 f2fs_err(sbi, "%s quota file already specified",
412 QTYPE2NAME(qtype));
413 goto errout;
414 }
415 if (strchr(qname, '/')) {
416 f2fs_err(sbi, "quotafile must be on filesystem root");
417 goto errout;
418 }
419 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
420 set_opt(sbi, QUOTA);
421 return 0;
422errout:
423 kfree(qname);
424 return ret;
425}
426
427static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
428{
429 struct f2fs_sb_info *sbi = F2FS_SB(sb);
430
431 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
432 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
433 return -EINVAL;
434 }
435 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
436 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
437 return 0;
438}
439
440static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
441{
442 /*
443 * We do the test below only for project quotas. 'usrquota' and
444 * 'grpquota' mount options are allowed even without quota feature
445 * to support legacy quotas in quota files.
446 */
447 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
448 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
449 return -1;
450 }
451 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
452 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
453 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
454 if (test_opt(sbi, USRQUOTA) &&
455 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
456 clear_opt(sbi, USRQUOTA);
457
458 if (test_opt(sbi, GRPQUOTA) &&
459 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
460 clear_opt(sbi, GRPQUOTA);
461
462 if (test_opt(sbi, PRJQUOTA) &&
463 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
464 clear_opt(sbi, PRJQUOTA);
465
466 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
467 test_opt(sbi, PRJQUOTA)) {
468 f2fs_err(sbi, "old and new quota format mixing");
469 return -1;
470 }
471
472 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
473 f2fs_err(sbi, "journaled quota format not specified");
474 return -1;
475 }
476 }
477
478 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
479 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
480 F2FS_OPTION(sbi).s_jquota_fmt = 0;
481 }
482 return 0;
483}
484#endif
485
486static int f2fs_set_test_dummy_encryption(struct super_block *sb,
487 const char *opt,
488 const substring_t *arg,
489 bool is_remount)
490{
491 struct f2fs_sb_info *sbi = F2FS_SB(sb);
492 struct fs_parameter param = {
493 .type = fs_value_is_string,
494 .string = arg->from ? arg->from : "",
495 };
496 struct fscrypt_dummy_policy *policy =
497 &F2FS_OPTION(sbi).dummy_enc_policy;
498 int err;
499
500 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
501 f2fs_warn(sbi, "test_dummy_encryption option not supported");
502 return -EINVAL;
503 }
504
505 if (!f2fs_sb_has_encrypt(sbi)) {
506 f2fs_err(sbi, "Encrypt feature is off");
507 return -EINVAL;
508 }
509
510 /*
511 * This mount option is just for testing, and it's not worthwhile to
512 * implement the extra complexity (e.g. RCU protection) that would be
513 * needed to allow it to be set or changed during remount. We do allow
514 * it to be specified during remount, but only if there is no change.
515 */
516 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
517 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
518 return -EINVAL;
519 }
520
521 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
522 if (err) {
523 if (err == -EEXIST)
524 f2fs_warn(sbi,
525 "Can't change test_dummy_encryption on remount");
526 else if (err == -EINVAL)
527 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
528 opt);
529 else
530 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
531 opt, err);
532 return -EINVAL;
533 }
534 f2fs_warn(sbi, "Test dummy encryption mode enabled");
535 return 0;
536}
537
538#ifdef CONFIG_F2FS_FS_COMPRESSION
539static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
540 const char *new_ext, bool is_ext)
541{
542 unsigned char (*ext)[F2FS_EXTENSION_LEN];
543 int ext_cnt;
544 int i;
545
546 if (is_ext) {
547 ext = F2FS_OPTION(sbi).extensions;
548 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
549 } else {
550 ext = F2FS_OPTION(sbi).noextensions;
551 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
552 }
553
554 for (i = 0; i < ext_cnt; i++) {
555 if (!strcasecmp(new_ext, ext[i]))
556 return true;
557 }
558
559 return false;
560}
561
562/*
563 * 1. The same extension name cannot not appear in both compress and non-compress extension
564 * at the same time.
565 * 2. If the compress extension specifies all files, the types specified by the non-compress
566 * extension will be treated as special cases and will not be compressed.
567 * 3. Don't allow the non-compress extension specifies all files.
568 */
569static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
570{
571 unsigned char (*ext)[F2FS_EXTENSION_LEN];
572 unsigned char (*noext)[F2FS_EXTENSION_LEN];
573 int ext_cnt, noext_cnt, index = 0, no_index = 0;
574
575 ext = F2FS_OPTION(sbi).extensions;
576 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
577 noext = F2FS_OPTION(sbi).noextensions;
578 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
579
580 if (!noext_cnt)
581 return 0;
582
583 for (no_index = 0; no_index < noext_cnt; no_index++) {
584 if (!strcasecmp("*", noext[no_index])) {
585 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
586 return -EINVAL;
587 }
588 for (index = 0; index < ext_cnt; index++) {
589 if (!strcasecmp(ext[index], noext[no_index])) {
590 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
591 ext[index]);
592 return -EINVAL;
593 }
594 }
595 }
596 return 0;
597}
598
599#ifdef CONFIG_F2FS_FS_LZ4
600static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
601{
602#ifdef CONFIG_F2FS_FS_LZ4HC
603 unsigned int level;
604
605 if (strlen(str) == 3) {
606 F2FS_OPTION(sbi).compress_level = 0;
607 return 0;
608 }
609
610 str += 3;
611
612 if (str[0] != ':') {
613 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
614 return -EINVAL;
615 }
616 if (kstrtouint(str + 1, 10, &level))
617 return -EINVAL;
618
619 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
620 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
621 return -EINVAL;
622 }
623
624 F2FS_OPTION(sbi).compress_level = level;
625 return 0;
626#else
627 if (strlen(str) == 3) {
628 F2FS_OPTION(sbi).compress_level = 0;
629 return 0;
630 }
631 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
632 return -EINVAL;
633#endif
634}
635#endif
636
637#ifdef CONFIG_F2FS_FS_ZSTD
638static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
639{
640 int level;
641 int len = 4;
642
643 if (strlen(str) == len) {
644 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
645 return 0;
646 }
647
648 str += len;
649
650 if (str[0] != ':') {
651 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
652 return -EINVAL;
653 }
654 if (kstrtoint(str + 1, 10, &level))
655 return -EINVAL;
656
657 /* f2fs does not support negative compress level now */
658 if (level < 0) {
659 f2fs_info(sbi, "do not support negative compress level: %d", level);
660 return -ERANGE;
661 }
662
663 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
664 f2fs_info(sbi, "invalid zstd compress level: %d", level);
665 return -EINVAL;
666 }
667
668 F2FS_OPTION(sbi).compress_level = level;
669 return 0;
670}
671#endif
672#endif
673
674static int parse_options(struct super_block *sb, char *options, bool is_remount)
675{
676 struct f2fs_sb_info *sbi = F2FS_SB(sb);
677 substring_t args[MAX_OPT_ARGS];
678#ifdef CONFIG_F2FS_FS_COMPRESSION
679 unsigned char (*ext)[F2FS_EXTENSION_LEN];
680 unsigned char (*noext)[F2FS_EXTENSION_LEN];
681 int ext_cnt, noext_cnt;
682#endif
683 char *p, *name;
684 int arg = 0;
685 kuid_t uid;
686 kgid_t gid;
687 int ret;
688
689 if (!options)
690 goto default_check;
691
692 while ((p = strsep(&options, ",")) != NULL) {
693 int token;
694
695 if (!*p)
696 continue;
697 /*
698 * Initialize args struct so we know whether arg was
699 * found; some options take optional arguments.
700 */
701 args[0].to = args[0].from = NULL;
702 token = match_token(p, f2fs_tokens, args);
703
704 switch (token) {
705 case Opt_gc_background:
706 name = match_strdup(&args[0]);
707
708 if (!name)
709 return -ENOMEM;
710 if (!strcmp(name, "on")) {
711 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
712 } else if (!strcmp(name, "off")) {
713 if (f2fs_sb_has_blkzoned(sbi)) {
714 f2fs_warn(sbi, "zoned devices need bggc");
715 kfree(name);
716 return -EINVAL;
717 }
718 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
719 } else if (!strcmp(name, "sync")) {
720 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
721 } else {
722 kfree(name);
723 return -EINVAL;
724 }
725 kfree(name);
726 break;
727 case Opt_disable_roll_forward:
728 set_opt(sbi, DISABLE_ROLL_FORWARD);
729 break;
730 case Opt_norecovery:
731 /* this option mounts f2fs with ro */
732 set_opt(sbi, NORECOVERY);
733 if (!f2fs_readonly(sb))
734 return -EINVAL;
735 break;
736 case Opt_discard:
737 if (!f2fs_hw_support_discard(sbi)) {
738 f2fs_warn(sbi, "device does not support discard");
739 break;
740 }
741 set_opt(sbi, DISCARD);
742 break;
743 case Opt_nodiscard:
744 if (f2fs_hw_should_discard(sbi)) {
745 f2fs_warn(sbi, "discard is required for zoned block devices");
746 return -EINVAL;
747 }
748 clear_opt(sbi, DISCARD);
749 break;
750 case Opt_noheap:
751 case Opt_heap:
752 f2fs_warn(sbi, "heap/no_heap options were deprecated");
753 break;
754#ifdef CONFIG_F2FS_FS_XATTR
755 case Opt_user_xattr:
756 set_opt(sbi, XATTR_USER);
757 break;
758 case Opt_nouser_xattr:
759 clear_opt(sbi, XATTR_USER);
760 break;
761 case Opt_inline_xattr:
762 set_opt(sbi, INLINE_XATTR);
763 break;
764 case Opt_noinline_xattr:
765 clear_opt(sbi, INLINE_XATTR);
766 break;
767 case Opt_inline_xattr_size:
768 if (args->from && match_int(args, &arg))
769 return -EINVAL;
770 set_opt(sbi, INLINE_XATTR_SIZE);
771 F2FS_OPTION(sbi).inline_xattr_size = arg;
772 break;
773#else
774 case Opt_user_xattr:
775 f2fs_info(sbi, "user_xattr options not supported");
776 break;
777 case Opt_nouser_xattr:
778 f2fs_info(sbi, "nouser_xattr options not supported");
779 break;
780 case Opt_inline_xattr:
781 f2fs_info(sbi, "inline_xattr options not supported");
782 break;
783 case Opt_noinline_xattr:
784 f2fs_info(sbi, "noinline_xattr options not supported");
785 break;
786#endif
787#ifdef CONFIG_F2FS_FS_POSIX_ACL
788 case Opt_acl:
789 set_opt(sbi, POSIX_ACL);
790 break;
791 case Opt_noacl:
792 clear_opt(sbi, POSIX_ACL);
793 break;
794#else
795 case Opt_acl:
796 f2fs_info(sbi, "acl options not supported");
797 break;
798 case Opt_noacl:
799 f2fs_info(sbi, "noacl options not supported");
800 break;
801#endif
802 case Opt_active_logs:
803 if (args->from && match_int(args, &arg))
804 return -EINVAL;
805 if (arg != 2 && arg != 4 &&
806 arg != NR_CURSEG_PERSIST_TYPE)
807 return -EINVAL;
808 F2FS_OPTION(sbi).active_logs = arg;
809 break;
810 case Opt_disable_ext_identify:
811 set_opt(sbi, DISABLE_EXT_IDENTIFY);
812 break;
813 case Opt_inline_data:
814 set_opt(sbi, INLINE_DATA);
815 break;
816 case Opt_inline_dentry:
817 set_opt(sbi, INLINE_DENTRY);
818 break;
819 case Opt_noinline_dentry:
820 clear_opt(sbi, INLINE_DENTRY);
821 break;
822 case Opt_flush_merge:
823 set_opt(sbi, FLUSH_MERGE);
824 break;
825 case Opt_noflush_merge:
826 clear_opt(sbi, FLUSH_MERGE);
827 break;
828 case Opt_nobarrier:
829 set_opt(sbi, NOBARRIER);
830 break;
831 case Opt_barrier:
832 clear_opt(sbi, NOBARRIER);
833 break;
834 case Opt_fastboot:
835 set_opt(sbi, FASTBOOT);
836 break;
837 case Opt_extent_cache:
838 set_opt(sbi, READ_EXTENT_CACHE);
839 break;
840 case Opt_noextent_cache:
841 if (F2FS_HAS_FEATURE(sbi, F2FS_FEATURE_DEVICE_ALIAS)) {
842 f2fs_err(sbi, "device aliasing requires extent cache");
843 return -EINVAL;
844 }
845 clear_opt(sbi, READ_EXTENT_CACHE);
846 break;
847 case Opt_noinline_data:
848 clear_opt(sbi, INLINE_DATA);
849 break;
850 case Opt_data_flush:
851 set_opt(sbi, DATA_FLUSH);
852 break;
853 case Opt_reserve_root:
854 if (args->from && match_int(args, &arg))
855 return -EINVAL;
856 if (test_opt(sbi, RESERVE_ROOT)) {
857 f2fs_info(sbi, "Preserve previous reserve_root=%u",
858 F2FS_OPTION(sbi).root_reserved_blocks);
859 } else {
860 F2FS_OPTION(sbi).root_reserved_blocks = arg;
861 set_opt(sbi, RESERVE_ROOT);
862 }
863 break;
864 case Opt_resuid:
865 if (args->from && match_int(args, &arg))
866 return -EINVAL;
867 uid = make_kuid(current_user_ns(), arg);
868 if (!uid_valid(uid)) {
869 f2fs_err(sbi, "Invalid uid value %d", arg);
870 return -EINVAL;
871 }
872 F2FS_OPTION(sbi).s_resuid = uid;
873 break;
874 case Opt_resgid:
875 if (args->from && match_int(args, &arg))
876 return -EINVAL;
877 gid = make_kgid(current_user_ns(), arg);
878 if (!gid_valid(gid)) {
879 f2fs_err(sbi, "Invalid gid value %d", arg);
880 return -EINVAL;
881 }
882 F2FS_OPTION(sbi).s_resgid = gid;
883 break;
884 case Opt_mode:
885 name = match_strdup(&args[0]);
886
887 if (!name)
888 return -ENOMEM;
889 if (!strcmp(name, "adaptive")) {
890 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
891 } else if (!strcmp(name, "lfs")) {
892 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
893 } else if (!strcmp(name, "fragment:segment")) {
894 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
895 } else if (!strcmp(name, "fragment:block")) {
896 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
897 } else {
898 kfree(name);
899 return -EINVAL;
900 }
901 kfree(name);
902 break;
903#ifdef CONFIG_F2FS_FAULT_INJECTION
904 case Opt_fault_injection:
905 if (args->from && match_int(args, &arg))
906 return -EINVAL;
907 if (f2fs_build_fault_attr(sbi, arg,
908 F2FS_ALL_FAULT_TYPE))
909 return -EINVAL;
910 set_opt(sbi, FAULT_INJECTION);
911 break;
912
913 case Opt_fault_type:
914 if (args->from && match_int(args, &arg))
915 return -EINVAL;
916 if (f2fs_build_fault_attr(sbi, 0, arg))
917 return -EINVAL;
918 set_opt(sbi, FAULT_INJECTION);
919 break;
920#else
921 case Opt_fault_injection:
922 f2fs_info(sbi, "fault_injection options not supported");
923 break;
924
925 case Opt_fault_type:
926 f2fs_info(sbi, "fault_type options not supported");
927 break;
928#endif
929 case Opt_lazytime:
930 sb->s_flags |= SB_LAZYTIME;
931 break;
932 case Opt_nolazytime:
933 sb->s_flags &= ~SB_LAZYTIME;
934 break;
935#ifdef CONFIG_QUOTA
936 case Opt_quota:
937 case Opt_usrquota:
938 set_opt(sbi, USRQUOTA);
939 break;
940 case Opt_grpquota:
941 set_opt(sbi, GRPQUOTA);
942 break;
943 case Opt_prjquota:
944 set_opt(sbi, PRJQUOTA);
945 break;
946 case Opt_usrjquota:
947 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
948 if (ret)
949 return ret;
950 break;
951 case Opt_grpjquota:
952 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
953 if (ret)
954 return ret;
955 break;
956 case Opt_prjjquota:
957 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
958 if (ret)
959 return ret;
960 break;
961 case Opt_offusrjquota:
962 ret = f2fs_clear_qf_name(sb, USRQUOTA);
963 if (ret)
964 return ret;
965 break;
966 case Opt_offgrpjquota:
967 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
968 if (ret)
969 return ret;
970 break;
971 case Opt_offprjjquota:
972 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
973 if (ret)
974 return ret;
975 break;
976 case Opt_jqfmt_vfsold:
977 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
978 break;
979 case Opt_jqfmt_vfsv0:
980 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
981 break;
982 case Opt_jqfmt_vfsv1:
983 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
984 break;
985 case Opt_noquota:
986 clear_opt(sbi, QUOTA);
987 clear_opt(sbi, USRQUOTA);
988 clear_opt(sbi, GRPQUOTA);
989 clear_opt(sbi, PRJQUOTA);
990 break;
991#else
992 case Opt_quota:
993 case Opt_usrquota:
994 case Opt_grpquota:
995 case Opt_prjquota:
996 case Opt_usrjquota:
997 case Opt_grpjquota:
998 case Opt_prjjquota:
999 case Opt_offusrjquota:
1000 case Opt_offgrpjquota:
1001 case Opt_offprjjquota:
1002 case Opt_jqfmt_vfsold:
1003 case Opt_jqfmt_vfsv0:
1004 case Opt_jqfmt_vfsv1:
1005 case Opt_noquota:
1006 f2fs_info(sbi, "quota operations not supported");
1007 break;
1008#endif
1009 case Opt_alloc:
1010 name = match_strdup(&args[0]);
1011 if (!name)
1012 return -ENOMEM;
1013
1014 if (!strcmp(name, "default")) {
1015 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1016 } else if (!strcmp(name, "reuse")) {
1017 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1018 } else {
1019 kfree(name);
1020 return -EINVAL;
1021 }
1022 kfree(name);
1023 break;
1024 case Opt_fsync:
1025 name = match_strdup(&args[0]);
1026 if (!name)
1027 return -ENOMEM;
1028 if (!strcmp(name, "posix")) {
1029 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1030 } else if (!strcmp(name, "strict")) {
1031 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1032 } else if (!strcmp(name, "nobarrier")) {
1033 F2FS_OPTION(sbi).fsync_mode =
1034 FSYNC_MODE_NOBARRIER;
1035 } else {
1036 kfree(name);
1037 return -EINVAL;
1038 }
1039 kfree(name);
1040 break;
1041 case Opt_test_dummy_encryption:
1042 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1043 is_remount);
1044 if (ret)
1045 return ret;
1046 break;
1047 case Opt_inlinecrypt:
1048#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1049 sb->s_flags |= SB_INLINECRYPT;
1050#else
1051 f2fs_info(sbi, "inline encryption not supported");
1052#endif
1053 break;
1054 case Opt_checkpoint_disable_cap_perc:
1055 if (args->from && match_int(args, &arg))
1056 return -EINVAL;
1057 if (arg < 0 || arg > 100)
1058 return -EINVAL;
1059 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1060 set_opt(sbi, DISABLE_CHECKPOINT);
1061 break;
1062 case Opt_checkpoint_disable_cap:
1063 if (args->from && match_int(args, &arg))
1064 return -EINVAL;
1065 F2FS_OPTION(sbi).unusable_cap = arg;
1066 set_opt(sbi, DISABLE_CHECKPOINT);
1067 break;
1068 case Opt_checkpoint_disable:
1069 set_opt(sbi, DISABLE_CHECKPOINT);
1070 break;
1071 case Opt_checkpoint_enable:
1072 clear_opt(sbi, DISABLE_CHECKPOINT);
1073 break;
1074 case Opt_checkpoint_merge:
1075 set_opt(sbi, MERGE_CHECKPOINT);
1076 break;
1077 case Opt_nocheckpoint_merge:
1078 clear_opt(sbi, MERGE_CHECKPOINT);
1079 break;
1080#ifdef CONFIG_F2FS_FS_COMPRESSION
1081 case Opt_compress_algorithm:
1082 if (!f2fs_sb_has_compression(sbi)) {
1083 f2fs_info(sbi, "Image doesn't support compression");
1084 break;
1085 }
1086 name = match_strdup(&args[0]);
1087 if (!name)
1088 return -ENOMEM;
1089 if (!strcmp(name, "lzo")) {
1090#ifdef CONFIG_F2FS_FS_LZO
1091 F2FS_OPTION(sbi).compress_level = 0;
1092 F2FS_OPTION(sbi).compress_algorithm =
1093 COMPRESS_LZO;
1094#else
1095 f2fs_info(sbi, "kernel doesn't support lzo compression");
1096#endif
1097 } else if (!strncmp(name, "lz4", 3)) {
1098#ifdef CONFIG_F2FS_FS_LZ4
1099 ret = f2fs_set_lz4hc_level(sbi, name);
1100 if (ret) {
1101 kfree(name);
1102 return -EINVAL;
1103 }
1104 F2FS_OPTION(sbi).compress_algorithm =
1105 COMPRESS_LZ4;
1106#else
1107 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1108#endif
1109 } else if (!strncmp(name, "zstd", 4)) {
1110#ifdef CONFIG_F2FS_FS_ZSTD
1111 ret = f2fs_set_zstd_level(sbi, name);
1112 if (ret) {
1113 kfree(name);
1114 return -EINVAL;
1115 }
1116 F2FS_OPTION(sbi).compress_algorithm =
1117 COMPRESS_ZSTD;
1118#else
1119 f2fs_info(sbi, "kernel doesn't support zstd compression");
1120#endif
1121 } else if (!strcmp(name, "lzo-rle")) {
1122#ifdef CONFIG_F2FS_FS_LZORLE
1123 F2FS_OPTION(sbi).compress_level = 0;
1124 F2FS_OPTION(sbi).compress_algorithm =
1125 COMPRESS_LZORLE;
1126#else
1127 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1128#endif
1129 } else {
1130 kfree(name);
1131 return -EINVAL;
1132 }
1133 kfree(name);
1134 break;
1135 case Opt_compress_log_size:
1136 if (!f2fs_sb_has_compression(sbi)) {
1137 f2fs_info(sbi, "Image doesn't support compression");
1138 break;
1139 }
1140 if (args->from && match_int(args, &arg))
1141 return -EINVAL;
1142 if (arg < MIN_COMPRESS_LOG_SIZE ||
1143 arg > MAX_COMPRESS_LOG_SIZE) {
1144 f2fs_err(sbi,
1145 "Compress cluster log size is out of range");
1146 return -EINVAL;
1147 }
1148 F2FS_OPTION(sbi).compress_log_size = arg;
1149 break;
1150 case Opt_compress_extension:
1151 if (!f2fs_sb_has_compression(sbi)) {
1152 f2fs_info(sbi, "Image doesn't support compression");
1153 break;
1154 }
1155 name = match_strdup(&args[0]);
1156 if (!name)
1157 return -ENOMEM;
1158
1159 ext = F2FS_OPTION(sbi).extensions;
1160 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1161
1162 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1163 ext_cnt >= COMPRESS_EXT_NUM) {
1164 f2fs_err(sbi,
1165 "invalid extension length/number");
1166 kfree(name);
1167 return -EINVAL;
1168 }
1169
1170 if (is_compress_extension_exist(sbi, name, true)) {
1171 kfree(name);
1172 break;
1173 }
1174
1175 ret = strscpy(ext[ext_cnt], name);
1176 if (ret < 0) {
1177 kfree(name);
1178 return ret;
1179 }
1180 F2FS_OPTION(sbi).compress_ext_cnt++;
1181 kfree(name);
1182 break;
1183 case Opt_nocompress_extension:
1184 if (!f2fs_sb_has_compression(sbi)) {
1185 f2fs_info(sbi, "Image doesn't support compression");
1186 break;
1187 }
1188 name = match_strdup(&args[0]);
1189 if (!name)
1190 return -ENOMEM;
1191
1192 noext = F2FS_OPTION(sbi).noextensions;
1193 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1194
1195 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1196 noext_cnt >= COMPRESS_EXT_NUM) {
1197 f2fs_err(sbi,
1198 "invalid extension length/number");
1199 kfree(name);
1200 return -EINVAL;
1201 }
1202
1203 if (is_compress_extension_exist(sbi, name, false)) {
1204 kfree(name);
1205 break;
1206 }
1207
1208 ret = strscpy(noext[noext_cnt], name);
1209 if (ret < 0) {
1210 kfree(name);
1211 return ret;
1212 }
1213 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1214 kfree(name);
1215 break;
1216 case Opt_compress_chksum:
1217 if (!f2fs_sb_has_compression(sbi)) {
1218 f2fs_info(sbi, "Image doesn't support compression");
1219 break;
1220 }
1221 F2FS_OPTION(sbi).compress_chksum = true;
1222 break;
1223 case Opt_compress_mode:
1224 if (!f2fs_sb_has_compression(sbi)) {
1225 f2fs_info(sbi, "Image doesn't support compression");
1226 break;
1227 }
1228 name = match_strdup(&args[0]);
1229 if (!name)
1230 return -ENOMEM;
1231 if (!strcmp(name, "fs")) {
1232 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1233 } else if (!strcmp(name, "user")) {
1234 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1235 } else {
1236 kfree(name);
1237 return -EINVAL;
1238 }
1239 kfree(name);
1240 break;
1241 case Opt_compress_cache:
1242 if (!f2fs_sb_has_compression(sbi)) {
1243 f2fs_info(sbi, "Image doesn't support compression");
1244 break;
1245 }
1246 set_opt(sbi, COMPRESS_CACHE);
1247 break;
1248#else
1249 case Opt_compress_algorithm:
1250 case Opt_compress_log_size:
1251 case Opt_compress_extension:
1252 case Opt_nocompress_extension:
1253 case Opt_compress_chksum:
1254 case Opt_compress_mode:
1255 case Opt_compress_cache:
1256 f2fs_info(sbi, "compression options not supported");
1257 break;
1258#endif
1259 case Opt_atgc:
1260 set_opt(sbi, ATGC);
1261 break;
1262 case Opt_gc_merge:
1263 set_opt(sbi, GC_MERGE);
1264 break;
1265 case Opt_nogc_merge:
1266 clear_opt(sbi, GC_MERGE);
1267 break;
1268 case Opt_discard_unit:
1269 name = match_strdup(&args[0]);
1270 if (!name)
1271 return -ENOMEM;
1272 if (!strcmp(name, "block")) {
1273 F2FS_OPTION(sbi).discard_unit =
1274 DISCARD_UNIT_BLOCK;
1275 } else if (!strcmp(name, "segment")) {
1276 F2FS_OPTION(sbi).discard_unit =
1277 DISCARD_UNIT_SEGMENT;
1278 } else if (!strcmp(name, "section")) {
1279 F2FS_OPTION(sbi).discard_unit =
1280 DISCARD_UNIT_SECTION;
1281 } else {
1282 kfree(name);
1283 return -EINVAL;
1284 }
1285 kfree(name);
1286 break;
1287 case Opt_memory_mode:
1288 name = match_strdup(&args[0]);
1289 if (!name)
1290 return -ENOMEM;
1291 if (!strcmp(name, "normal")) {
1292 F2FS_OPTION(sbi).memory_mode =
1293 MEMORY_MODE_NORMAL;
1294 } else if (!strcmp(name, "low")) {
1295 F2FS_OPTION(sbi).memory_mode =
1296 MEMORY_MODE_LOW;
1297 } else {
1298 kfree(name);
1299 return -EINVAL;
1300 }
1301 kfree(name);
1302 break;
1303 case Opt_age_extent_cache:
1304 set_opt(sbi, AGE_EXTENT_CACHE);
1305 break;
1306 case Opt_errors:
1307 name = match_strdup(&args[0]);
1308 if (!name)
1309 return -ENOMEM;
1310 if (!strcmp(name, "remount-ro")) {
1311 F2FS_OPTION(sbi).errors =
1312 MOUNT_ERRORS_READONLY;
1313 } else if (!strcmp(name, "continue")) {
1314 F2FS_OPTION(sbi).errors =
1315 MOUNT_ERRORS_CONTINUE;
1316 } else if (!strcmp(name, "panic")) {
1317 F2FS_OPTION(sbi).errors =
1318 MOUNT_ERRORS_PANIC;
1319 } else {
1320 kfree(name);
1321 return -EINVAL;
1322 }
1323 kfree(name);
1324 break;
1325 default:
1326 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1327 p);
1328 return -EINVAL;
1329 }
1330 }
1331default_check:
1332#ifdef CONFIG_QUOTA
1333 if (f2fs_check_quota_options(sbi))
1334 return -EINVAL;
1335#else
1336 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1337 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1338 return -EINVAL;
1339 }
1340 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1341 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1342 return -EINVAL;
1343 }
1344#endif
1345
1346 if (!IS_ENABLED(CONFIG_UNICODE) && f2fs_sb_has_casefold(sbi)) {
1347 f2fs_err(sbi,
1348 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1349 return -EINVAL;
1350 }
1351
1352 /*
1353 * The BLKZONED feature indicates that the drive was formatted with
1354 * zone alignment optimization. This is optional for host-aware
1355 * devices, but mandatory for host-managed zoned block devices.
1356 */
1357 if (f2fs_sb_has_blkzoned(sbi)) {
1358#ifdef CONFIG_BLK_DEV_ZONED
1359 if (F2FS_OPTION(sbi).discard_unit !=
1360 DISCARD_UNIT_SECTION) {
1361 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1362 F2FS_OPTION(sbi).discard_unit =
1363 DISCARD_UNIT_SECTION;
1364 }
1365
1366 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1367 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1368 return -EINVAL;
1369 }
1370#else
1371 f2fs_err(sbi, "Zoned block device support is not enabled");
1372 return -EINVAL;
1373#endif
1374 }
1375
1376#ifdef CONFIG_F2FS_FS_COMPRESSION
1377 if (f2fs_test_compress_extension(sbi)) {
1378 f2fs_err(sbi, "invalid compress or nocompress extension");
1379 return -EINVAL;
1380 }
1381#endif
1382
1383 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1384 int min_size, max_size;
1385
1386 if (!f2fs_sb_has_extra_attr(sbi) ||
1387 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1388 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1389 return -EINVAL;
1390 }
1391 if (!test_opt(sbi, INLINE_XATTR)) {
1392 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1393 return -EINVAL;
1394 }
1395
1396 min_size = MIN_INLINE_XATTR_SIZE;
1397 max_size = MAX_INLINE_XATTR_SIZE;
1398
1399 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1400 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1401 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1402 min_size, max_size);
1403 return -EINVAL;
1404 }
1405 }
1406
1407 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1408 f2fs_err(sbi, "LFS is not compatible with ATGC");
1409 return -EINVAL;
1410 }
1411
1412 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1413 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1414 return -EINVAL;
1415 }
1416
1417 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1418 f2fs_err(sbi, "Allow to mount readonly mode only");
1419 return -EROFS;
1420 }
1421 return 0;
1422}
1423
1424static struct inode *f2fs_alloc_inode(struct super_block *sb)
1425{
1426 struct f2fs_inode_info *fi;
1427
1428 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1429 return NULL;
1430
1431 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1432 if (!fi)
1433 return NULL;
1434
1435 init_once((void *) fi);
1436
1437 /* Initialize f2fs-specific inode info */
1438 atomic_set(&fi->dirty_pages, 0);
1439 atomic_set(&fi->i_compr_blocks, 0);
1440 init_f2fs_rwsem(&fi->i_sem);
1441 spin_lock_init(&fi->i_size_lock);
1442 INIT_LIST_HEAD(&fi->dirty_list);
1443 INIT_LIST_HEAD(&fi->gdirty_list);
1444 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1445 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1446 init_f2fs_rwsem(&fi->i_xattr_sem);
1447
1448 /* Will be used by directory only */
1449 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1450
1451 return &fi->vfs_inode;
1452}
1453
1454static int f2fs_drop_inode(struct inode *inode)
1455{
1456 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1457 int ret;
1458
1459 /*
1460 * during filesystem shutdown, if checkpoint is disabled,
1461 * drop useless meta/node dirty pages.
1462 */
1463 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1464 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1465 inode->i_ino == F2FS_META_INO(sbi)) {
1466 trace_f2fs_drop_inode(inode, 1);
1467 return 1;
1468 }
1469 }
1470
1471 /*
1472 * This is to avoid a deadlock condition like below.
1473 * writeback_single_inode(inode)
1474 * - f2fs_write_data_page
1475 * - f2fs_gc -> iput -> evict
1476 * - inode_wait_for_writeback(inode)
1477 */
1478 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1479 if (!inode->i_nlink && !is_bad_inode(inode)) {
1480 /* to avoid evict_inode call simultaneously */
1481 atomic_inc(&inode->i_count);
1482 spin_unlock(&inode->i_lock);
1483
1484 /* should remain fi->extent_tree for writepage */
1485 f2fs_destroy_extent_node(inode);
1486
1487 sb_start_intwrite(inode->i_sb);
1488 f2fs_i_size_write(inode, 0);
1489
1490 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1491 inode, NULL, 0, DATA);
1492 truncate_inode_pages_final(inode->i_mapping);
1493
1494 if (F2FS_HAS_BLOCKS(inode))
1495 f2fs_truncate(inode);
1496
1497 sb_end_intwrite(inode->i_sb);
1498
1499 spin_lock(&inode->i_lock);
1500 atomic_dec(&inode->i_count);
1501 }
1502 trace_f2fs_drop_inode(inode, 0);
1503 return 0;
1504 }
1505 ret = generic_drop_inode(inode);
1506 if (!ret)
1507 ret = fscrypt_drop_inode(inode);
1508 trace_f2fs_drop_inode(inode, ret);
1509 return ret;
1510}
1511
1512int f2fs_inode_dirtied(struct inode *inode, bool sync)
1513{
1514 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1515 int ret = 0;
1516
1517 spin_lock(&sbi->inode_lock[DIRTY_META]);
1518 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1519 ret = 1;
1520 } else {
1521 set_inode_flag(inode, FI_DIRTY_INODE);
1522 stat_inc_dirty_inode(sbi, DIRTY_META);
1523 }
1524 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1525 list_add_tail(&F2FS_I(inode)->gdirty_list,
1526 &sbi->inode_list[DIRTY_META]);
1527 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1528 }
1529 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1530 return ret;
1531}
1532
1533void f2fs_inode_synced(struct inode *inode)
1534{
1535 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1536
1537 spin_lock(&sbi->inode_lock[DIRTY_META]);
1538 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1539 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1540 return;
1541 }
1542 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1543 list_del_init(&F2FS_I(inode)->gdirty_list);
1544 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1545 }
1546 clear_inode_flag(inode, FI_DIRTY_INODE);
1547 clear_inode_flag(inode, FI_AUTO_RECOVER);
1548 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1549 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1550}
1551
1552/*
1553 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1554 *
1555 * We should call set_dirty_inode to write the dirty inode through write_inode.
1556 */
1557static void f2fs_dirty_inode(struct inode *inode, int flags)
1558{
1559 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1560
1561 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1562 inode->i_ino == F2FS_META_INO(sbi))
1563 return;
1564
1565 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1566 clear_inode_flag(inode, FI_AUTO_RECOVER);
1567
1568 f2fs_inode_dirtied(inode, false);
1569}
1570
1571static void f2fs_free_inode(struct inode *inode)
1572{
1573 fscrypt_free_inode(inode);
1574 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1575}
1576
1577static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1578{
1579 percpu_counter_destroy(&sbi->total_valid_inode_count);
1580 percpu_counter_destroy(&sbi->rf_node_block_count);
1581 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1582}
1583
1584static void destroy_device_list(struct f2fs_sb_info *sbi)
1585{
1586 int i;
1587
1588 for (i = 0; i < sbi->s_ndevs; i++) {
1589 if (i > 0)
1590 bdev_fput(FDEV(i).bdev_file);
1591#ifdef CONFIG_BLK_DEV_ZONED
1592 kvfree(FDEV(i).blkz_seq);
1593#endif
1594 }
1595 kvfree(sbi->devs);
1596}
1597
1598static void f2fs_put_super(struct super_block *sb)
1599{
1600 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1601 int i;
1602 int err = 0;
1603 bool done;
1604
1605 /* unregister procfs/sysfs entries in advance to avoid race case */
1606 f2fs_unregister_sysfs(sbi);
1607
1608 f2fs_quota_off_umount(sb);
1609
1610 /* prevent remaining shrinker jobs */
1611 mutex_lock(&sbi->umount_mutex);
1612
1613 /*
1614 * flush all issued checkpoints and stop checkpoint issue thread.
1615 * after then, all checkpoints should be done by each process context.
1616 */
1617 f2fs_stop_ckpt_thread(sbi);
1618
1619 /*
1620 * We don't need to do checkpoint when superblock is clean.
1621 * But, the previous checkpoint was not done by umount, it needs to do
1622 * clean checkpoint again.
1623 */
1624 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1625 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1626 struct cp_control cpc = {
1627 .reason = CP_UMOUNT,
1628 };
1629 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1630 err = f2fs_write_checkpoint(sbi, &cpc);
1631 }
1632
1633 /* be sure to wait for any on-going discard commands */
1634 done = f2fs_issue_discard_timeout(sbi);
1635 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1636 struct cp_control cpc = {
1637 .reason = CP_UMOUNT | CP_TRIMMED,
1638 };
1639 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1640 err = f2fs_write_checkpoint(sbi, &cpc);
1641 }
1642
1643 /*
1644 * normally superblock is clean, so we need to release this.
1645 * In addition, EIO will skip do checkpoint, we need this as well.
1646 */
1647 f2fs_release_ino_entry(sbi, true);
1648
1649 f2fs_leave_shrinker(sbi);
1650 mutex_unlock(&sbi->umount_mutex);
1651
1652 /* our cp_error case, we can wait for any writeback page */
1653 f2fs_flush_merged_writes(sbi);
1654
1655 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1656
1657 if (err || f2fs_cp_error(sbi)) {
1658 truncate_inode_pages_final(NODE_MAPPING(sbi));
1659 truncate_inode_pages_final(META_MAPPING(sbi));
1660 }
1661
1662 for (i = 0; i < NR_COUNT_TYPE; i++) {
1663 if (!get_pages(sbi, i))
1664 continue;
1665 f2fs_err(sbi, "detect filesystem reference count leak during "
1666 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1667 f2fs_bug_on(sbi, 1);
1668 }
1669
1670 f2fs_bug_on(sbi, sbi->fsync_node_num);
1671
1672 f2fs_destroy_compress_inode(sbi);
1673
1674 iput(sbi->node_inode);
1675 sbi->node_inode = NULL;
1676
1677 iput(sbi->meta_inode);
1678 sbi->meta_inode = NULL;
1679
1680 /*
1681 * iput() can update stat information, if f2fs_write_checkpoint()
1682 * above failed with error.
1683 */
1684 f2fs_destroy_stats(sbi);
1685
1686 /* destroy f2fs internal modules */
1687 f2fs_destroy_node_manager(sbi);
1688 f2fs_destroy_segment_manager(sbi);
1689
1690 /* flush s_error_work before sbi destroy */
1691 flush_work(&sbi->s_error_work);
1692
1693 f2fs_destroy_post_read_wq(sbi);
1694
1695 kvfree(sbi->ckpt);
1696
1697 if (sbi->s_chksum_driver)
1698 crypto_free_shash(sbi->s_chksum_driver);
1699 kfree(sbi->raw_super);
1700
1701 f2fs_destroy_page_array_cache(sbi);
1702 f2fs_destroy_xattr_caches(sbi);
1703#ifdef CONFIG_QUOTA
1704 for (i = 0; i < MAXQUOTAS; i++)
1705 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1706#endif
1707 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1708 destroy_percpu_info(sbi);
1709 f2fs_destroy_iostat(sbi);
1710 for (i = 0; i < NR_PAGE_TYPE; i++)
1711 kvfree(sbi->write_io[i]);
1712#if IS_ENABLED(CONFIG_UNICODE)
1713 utf8_unload(sb->s_encoding);
1714#endif
1715}
1716
1717int f2fs_sync_fs(struct super_block *sb, int sync)
1718{
1719 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1720 int err = 0;
1721
1722 if (unlikely(f2fs_cp_error(sbi)))
1723 return 0;
1724 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1725 return 0;
1726
1727 trace_f2fs_sync_fs(sb, sync);
1728
1729 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1730 return -EAGAIN;
1731
1732 if (sync) {
1733 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1734 err = f2fs_issue_checkpoint(sbi);
1735 }
1736
1737 return err;
1738}
1739
1740static int f2fs_freeze(struct super_block *sb)
1741{
1742 if (f2fs_readonly(sb))
1743 return 0;
1744
1745 /* IO error happened before */
1746 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1747 return -EIO;
1748
1749 /* must be clean, since sync_filesystem() was already called */
1750 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1751 return -EINVAL;
1752
1753 /* Let's flush checkpoints and stop the thread. */
1754 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1755
1756 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1757 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1758 return 0;
1759}
1760
1761static int f2fs_unfreeze(struct super_block *sb)
1762{
1763 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1764
1765 /*
1766 * It will update discard_max_bytes of mounted lvm device to zero
1767 * after creating snapshot on this lvm device, let's drop all
1768 * remained discards.
1769 * We don't need to disable real-time discard because discard_max_bytes
1770 * will recover after removal of snapshot.
1771 */
1772 if (test_opt(sbi, DISCARD) && !f2fs_hw_support_discard(sbi))
1773 f2fs_issue_discard_timeout(sbi);
1774
1775 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1776 return 0;
1777}
1778
1779#ifdef CONFIG_QUOTA
1780static int f2fs_statfs_project(struct super_block *sb,
1781 kprojid_t projid, struct kstatfs *buf)
1782{
1783 struct kqid qid;
1784 struct dquot *dquot;
1785 u64 limit;
1786 u64 curblock;
1787
1788 qid = make_kqid_projid(projid);
1789 dquot = dqget(sb, qid);
1790 if (IS_ERR(dquot))
1791 return PTR_ERR(dquot);
1792 spin_lock(&dquot->dq_dqb_lock);
1793
1794 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1795 dquot->dq_dqb.dqb_bhardlimit);
1796 if (limit)
1797 limit >>= sb->s_blocksize_bits;
1798
1799 if (limit && buf->f_blocks > limit) {
1800 curblock = (dquot->dq_dqb.dqb_curspace +
1801 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1802 buf->f_blocks = limit;
1803 buf->f_bfree = buf->f_bavail =
1804 (buf->f_blocks > curblock) ?
1805 (buf->f_blocks - curblock) : 0;
1806 }
1807
1808 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1809 dquot->dq_dqb.dqb_ihardlimit);
1810
1811 if (limit && buf->f_files > limit) {
1812 buf->f_files = limit;
1813 buf->f_ffree =
1814 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1815 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1816 }
1817
1818 spin_unlock(&dquot->dq_dqb_lock);
1819 dqput(dquot);
1820 return 0;
1821}
1822#endif
1823
1824static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1825{
1826 struct super_block *sb = dentry->d_sb;
1827 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1828 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1829 block_t total_count, user_block_count, start_count;
1830 u64 avail_node_count;
1831 unsigned int total_valid_node_count;
1832
1833 total_count = le64_to_cpu(sbi->raw_super->block_count);
1834 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1835 buf->f_type = F2FS_SUPER_MAGIC;
1836 buf->f_bsize = sbi->blocksize;
1837
1838 buf->f_blocks = total_count - start_count;
1839
1840 spin_lock(&sbi->stat_lock);
1841
1842 user_block_count = sbi->user_block_count;
1843 total_valid_node_count = valid_node_count(sbi);
1844 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1845 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1846 sbi->current_reserved_blocks;
1847
1848 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1849 buf->f_bfree = 0;
1850 else
1851 buf->f_bfree -= sbi->unusable_block_count;
1852 spin_unlock(&sbi->stat_lock);
1853
1854 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1855 buf->f_bavail = buf->f_bfree -
1856 F2FS_OPTION(sbi).root_reserved_blocks;
1857 else
1858 buf->f_bavail = 0;
1859
1860 if (avail_node_count > user_block_count) {
1861 buf->f_files = user_block_count;
1862 buf->f_ffree = buf->f_bavail;
1863 } else {
1864 buf->f_files = avail_node_count;
1865 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1866 buf->f_bavail);
1867 }
1868
1869 buf->f_namelen = F2FS_NAME_LEN;
1870 buf->f_fsid = u64_to_fsid(id);
1871
1872#ifdef CONFIG_QUOTA
1873 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1874 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1875 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1876 }
1877#endif
1878 return 0;
1879}
1880
1881static inline void f2fs_show_quota_options(struct seq_file *seq,
1882 struct super_block *sb)
1883{
1884#ifdef CONFIG_QUOTA
1885 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1886
1887 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1888 char *fmtname = "";
1889
1890 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1891 case QFMT_VFS_OLD:
1892 fmtname = "vfsold";
1893 break;
1894 case QFMT_VFS_V0:
1895 fmtname = "vfsv0";
1896 break;
1897 case QFMT_VFS_V1:
1898 fmtname = "vfsv1";
1899 break;
1900 }
1901 seq_printf(seq, ",jqfmt=%s", fmtname);
1902 }
1903
1904 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1905 seq_show_option(seq, "usrjquota",
1906 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1907
1908 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1909 seq_show_option(seq, "grpjquota",
1910 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1911
1912 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1913 seq_show_option(seq, "prjjquota",
1914 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1915#endif
1916}
1917
1918#ifdef CONFIG_F2FS_FS_COMPRESSION
1919static inline void f2fs_show_compress_options(struct seq_file *seq,
1920 struct super_block *sb)
1921{
1922 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1923 char *algtype = "";
1924 int i;
1925
1926 if (!f2fs_sb_has_compression(sbi))
1927 return;
1928
1929 switch (F2FS_OPTION(sbi).compress_algorithm) {
1930 case COMPRESS_LZO:
1931 algtype = "lzo";
1932 break;
1933 case COMPRESS_LZ4:
1934 algtype = "lz4";
1935 break;
1936 case COMPRESS_ZSTD:
1937 algtype = "zstd";
1938 break;
1939 case COMPRESS_LZORLE:
1940 algtype = "lzo-rle";
1941 break;
1942 }
1943 seq_printf(seq, ",compress_algorithm=%s", algtype);
1944
1945 if (F2FS_OPTION(sbi).compress_level)
1946 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1947
1948 seq_printf(seq, ",compress_log_size=%u",
1949 F2FS_OPTION(sbi).compress_log_size);
1950
1951 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1952 seq_printf(seq, ",compress_extension=%s",
1953 F2FS_OPTION(sbi).extensions[i]);
1954 }
1955
1956 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1957 seq_printf(seq, ",nocompress_extension=%s",
1958 F2FS_OPTION(sbi).noextensions[i]);
1959 }
1960
1961 if (F2FS_OPTION(sbi).compress_chksum)
1962 seq_puts(seq, ",compress_chksum");
1963
1964 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1965 seq_printf(seq, ",compress_mode=%s", "fs");
1966 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1967 seq_printf(seq, ",compress_mode=%s", "user");
1968
1969 if (test_opt(sbi, COMPRESS_CACHE))
1970 seq_puts(seq, ",compress_cache");
1971}
1972#endif
1973
1974static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1975{
1976 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1977
1978 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1979 seq_printf(seq, ",background_gc=%s", "sync");
1980 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1981 seq_printf(seq, ",background_gc=%s", "on");
1982 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1983 seq_printf(seq, ",background_gc=%s", "off");
1984
1985 if (test_opt(sbi, GC_MERGE))
1986 seq_puts(seq, ",gc_merge");
1987 else
1988 seq_puts(seq, ",nogc_merge");
1989
1990 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1991 seq_puts(seq, ",disable_roll_forward");
1992 if (test_opt(sbi, NORECOVERY))
1993 seq_puts(seq, ",norecovery");
1994 if (test_opt(sbi, DISCARD)) {
1995 seq_puts(seq, ",discard");
1996 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1997 seq_printf(seq, ",discard_unit=%s", "block");
1998 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1999 seq_printf(seq, ",discard_unit=%s", "segment");
2000 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2001 seq_printf(seq, ",discard_unit=%s", "section");
2002 } else {
2003 seq_puts(seq, ",nodiscard");
2004 }
2005#ifdef CONFIG_F2FS_FS_XATTR
2006 if (test_opt(sbi, XATTR_USER))
2007 seq_puts(seq, ",user_xattr");
2008 else
2009 seq_puts(seq, ",nouser_xattr");
2010 if (test_opt(sbi, INLINE_XATTR))
2011 seq_puts(seq, ",inline_xattr");
2012 else
2013 seq_puts(seq, ",noinline_xattr");
2014 if (test_opt(sbi, INLINE_XATTR_SIZE))
2015 seq_printf(seq, ",inline_xattr_size=%u",
2016 F2FS_OPTION(sbi).inline_xattr_size);
2017#endif
2018#ifdef CONFIG_F2FS_FS_POSIX_ACL
2019 if (test_opt(sbi, POSIX_ACL))
2020 seq_puts(seq, ",acl");
2021 else
2022 seq_puts(seq, ",noacl");
2023#endif
2024 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
2025 seq_puts(seq, ",disable_ext_identify");
2026 if (test_opt(sbi, INLINE_DATA))
2027 seq_puts(seq, ",inline_data");
2028 else
2029 seq_puts(seq, ",noinline_data");
2030 if (test_opt(sbi, INLINE_DENTRY))
2031 seq_puts(seq, ",inline_dentry");
2032 else
2033 seq_puts(seq, ",noinline_dentry");
2034 if (test_opt(sbi, FLUSH_MERGE))
2035 seq_puts(seq, ",flush_merge");
2036 else
2037 seq_puts(seq, ",noflush_merge");
2038 if (test_opt(sbi, NOBARRIER))
2039 seq_puts(seq, ",nobarrier");
2040 else
2041 seq_puts(seq, ",barrier");
2042 if (test_opt(sbi, FASTBOOT))
2043 seq_puts(seq, ",fastboot");
2044 if (test_opt(sbi, READ_EXTENT_CACHE))
2045 seq_puts(seq, ",extent_cache");
2046 else
2047 seq_puts(seq, ",noextent_cache");
2048 if (test_opt(sbi, AGE_EXTENT_CACHE))
2049 seq_puts(seq, ",age_extent_cache");
2050 if (test_opt(sbi, DATA_FLUSH))
2051 seq_puts(seq, ",data_flush");
2052
2053 seq_puts(seq, ",mode=");
2054 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2055 seq_puts(seq, "adaptive");
2056 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2057 seq_puts(seq, "lfs");
2058 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2059 seq_puts(seq, "fragment:segment");
2060 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2061 seq_puts(seq, "fragment:block");
2062 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2063 if (test_opt(sbi, RESERVE_ROOT))
2064 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2065 F2FS_OPTION(sbi).root_reserved_blocks,
2066 from_kuid_munged(&init_user_ns,
2067 F2FS_OPTION(sbi).s_resuid),
2068 from_kgid_munged(&init_user_ns,
2069 F2FS_OPTION(sbi).s_resgid));
2070#ifdef CONFIG_F2FS_FAULT_INJECTION
2071 if (test_opt(sbi, FAULT_INJECTION)) {
2072 seq_printf(seq, ",fault_injection=%u",
2073 F2FS_OPTION(sbi).fault_info.inject_rate);
2074 seq_printf(seq, ",fault_type=%u",
2075 F2FS_OPTION(sbi).fault_info.inject_type);
2076 }
2077#endif
2078#ifdef CONFIG_QUOTA
2079 if (test_opt(sbi, QUOTA))
2080 seq_puts(seq, ",quota");
2081 if (test_opt(sbi, USRQUOTA))
2082 seq_puts(seq, ",usrquota");
2083 if (test_opt(sbi, GRPQUOTA))
2084 seq_puts(seq, ",grpquota");
2085 if (test_opt(sbi, PRJQUOTA))
2086 seq_puts(seq, ",prjquota");
2087#endif
2088 f2fs_show_quota_options(seq, sbi->sb);
2089
2090 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2091
2092 if (sbi->sb->s_flags & SB_INLINECRYPT)
2093 seq_puts(seq, ",inlinecrypt");
2094
2095 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2096 seq_printf(seq, ",alloc_mode=%s", "default");
2097 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2098 seq_printf(seq, ",alloc_mode=%s", "reuse");
2099
2100 if (test_opt(sbi, DISABLE_CHECKPOINT))
2101 seq_printf(seq, ",checkpoint=disable:%u",
2102 F2FS_OPTION(sbi).unusable_cap);
2103 if (test_opt(sbi, MERGE_CHECKPOINT))
2104 seq_puts(seq, ",checkpoint_merge");
2105 else
2106 seq_puts(seq, ",nocheckpoint_merge");
2107 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2108 seq_printf(seq, ",fsync_mode=%s", "posix");
2109 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2110 seq_printf(seq, ",fsync_mode=%s", "strict");
2111 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2112 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2113
2114#ifdef CONFIG_F2FS_FS_COMPRESSION
2115 f2fs_show_compress_options(seq, sbi->sb);
2116#endif
2117
2118 if (test_opt(sbi, ATGC))
2119 seq_puts(seq, ",atgc");
2120
2121 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2122 seq_printf(seq, ",memory=%s", "normal");
2123 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2124 seq_printf(seq, ",memory=%s", "low");
2125
2126 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2127 seq_printf(seq, ",errors=%s", "remount-ro");
2128 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2129 seq_printf(seq, ",errors=%s", "continue");
2130 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2131 seq_printf(seq, ",errors=%s", "panic");
2132
2133 return 0;
2134}
2135
2136static void default_options(struct f2fs_sb_info *sbi, bool remount)
2137{
2138 /* init some FS parameters */
2139 if (!remount) {
2140 set_opt(sbi, READ_EXTENT_CACHE);
2141 clear_opt(sbi, DISABLE_CHECKPOINT);
2142
2143 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2144 set_opt(sbi, DISCARD);
2145
2146 if (f2fs_sb_has_blkzoned(sbi))
2147 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2148 else
2149 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2150 }
2151
2152 if (f2fs_sb_has_readonly(sbi))
2153 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2154 else
2155 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2156
2157 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2158 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2159 SMALL_VOLUME_SEGMENTS)
2160 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2161 else
2162 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2163 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2164 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2165 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2166 if (f2fs_sb_has_compression(sbi)) {
2167 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2168 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2169 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2170 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2171 }
2172 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2173 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2174 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2175
2176 set_opt(sbi, INLINE_XATTR);
2177 set_opt(sbi, INLINE_DATA);
2178 set_opt(sbi, INLINE_DENTRY);
2179 set_opt(sbi, MERGE_CHECKPOINT);
2180 F2FS_OPTION(sbi).unusable_cap = 0;
2181 sbi->sb->s_flags |= SB_LAZYTIME;
2182 if (!f2fs_is_readonly(sbi))
2183 set_opt(sbi, FLUSH_MERGE);
2184 if (f2fs_sb_has_blkzoned(sbi))
2185 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2186 else
2187 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2188
2189#ifdef CONFIG_F2FS_FS_XATTR
2190 set_opt(sbi, XATTR_USER);
2191#endif
2192#ifdef CONFIG_F2FS_FS_POSIX_ACL
2193 set_opt(sbi, POSIX_ACL);
2194#endif
2195
2196 f2fs_build_fault_attr(sbi, 0, 0);
2197}
2198
2199#ifdef CONFIG_QUOTA
2200static int f2fs_enable_quotas(struct super_block *sb);
2201#endif
2202
2203static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2204{
2205 unsigned int s_flags = sbi->sb->s_flags;
2206 struct cp_control cpc;
2207 unsigned int gc_mode = sbi->gc_mode;
2208 int err = 0;
2209 int ret;
2210 block_t unusable;
2211
2212 if (s_flags & SB_RDONLY) {
2213 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2214 return -EINVAL;
2215 }
2216 sbi->sb->s_flags |= SB_ACTIVE;
2217
2218 /* check if we need more GC first */
2219 unusable = f2fs_get_unusable_blocks(sbi);
2220 if (!f2fs_disable_cp_again(sbi, unusable))
2221 goto skip_gc;
2222
2223 f2fs_update_time(sbi, DISABLE_TIME);
2224
2225 sbi->gc_mode = GC_URGENT_HIGH;
2226
2227 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2228 struct f2fs_gc_control gc_control = {
2229 .victim_segno = NULL_SEGNO,
2230 .init_gc_type = FG_GC,
2231 .should_migrate_blocks = false,
2232 .err_gc_skipped = true,
2233 .no_bg_gc = true,
2234 .nr_free_secs = 1 };
2235
2236 f2fs_down_write(&sbi->gc_lock);
2237 stat_inc_gc_call_count(sbi, FOREGROUND);
2238 err = f2fs_gc(sbi, &gc_control);
2239 if (err == -ENODATA) {
2240 err = 0;
2241 break;
2242 }
2243 if (err && err != -EAGAIN)
2244 break;
2245 }
2246
2247 ret = sync_filesystem(sbi->sb);
2248 if (ret || err) {
2249 err = ret ? ret : err;
2250 goto restore_flag;
2251 }
2252
2253 unusable = f2fs_get_unusable_blocks(sbi);
2254 if (f2fs_disable_cp_again(sbi, unusable)) {
2255 err = -EAGAIN;
2256 goto restore_flag;
2257 }
2258
2259skip_gc:
2260 f2fs_down_write(&sbi->gc_lock);
2261 cpc.reason = CP_PAUSE;
2262 set_sbi_flag(sbi, SBI_CP_DISABLED);
2263 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2264 err = f2fs_write_checkpoint(sbi, &cpc);
2265 if (err)
2266 goto out_unlock;
2267
2268 spin_lock(&sbi->stat_lock);
2269 sbi->unusable_block_count = unusable;
2270 spin_unlock(&sbi->stat_lock);
2271
2272out_unlock:
2273 f2fs_up_write(&sbi->gc_lock);
2274restore_flag:
2275 sbi->gc_mode = gc_mode;
2276 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2277 return err;
2278}
2279
2280static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2281{
2282 int retry = DEFAULT_RETRY_IO_COUNT;
2283
2284 /* we should flush all the data to keep data consistency */
2285 do {
2286 sync_inodes_sb(sbi->sb);
2287 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2288 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2289
2290 if (unlikely(retry < 0))
2291 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2292
2293 f2fs_down_write(&sbi->gc_lock);
2294 f2fs_dirty_to_prefree(sbi);
2295
2296 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2297 set_sbi_flag(sbi, SBI_IS_DIRTY);
2298 f2fs_up_write(&sbi->gc_lock);
2299
2300 f2fs_sync_fs(sbi->sb, 1);
2301
2302 /* Let's ensure there's no pending checkpoint anymore */
2303 f2fs_flush_ckpt_thread(sbi);
2304}
2305
2306static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2307{
2308 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2309 struct f2fs_mount_info org_mount_opt;
2310 unsigned long old_sb_flags;
2311 int err;
2312 bool need_restart_gc = false, need_stop_gc = false;
2313 bool need_restart_flush = false, need_stop_flush = false;
2314 bool need_restart_discard = false, need_stop_discard = false;
2315 bool need_enable_checkpoint = false, need_disable_checkpoint = false;
2316 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2317 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2318 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2319 bool no_atgc = !test_opt(sbi, ATGC);
2320 bool no_discard = !test_opt(sbi, DISCARD);
2321 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2322 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2323#ifdef CONFIG_QUOTA
2324 int i, j;
2325#endif
2326
2327 /*
2328 * Save the old mount options in case we
2329 * need to restore them.
2330 */
2331 org_mount_opt = sbi->mount_opt;
2332 old_sb_flags = sb->s_flags;
2333
2334#ifdef CONFIG_QUOTA
2335 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2336 for (i = 0; i < MAXQUOTAS; i++) {
2337 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2338 org_mount_opt.s_qf_names[i] =
2339 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2340 GFP_KERNEL);
2341 if (!org_mount_opt.s_qf_names[i]) {
2342 for (j = 0; j < i; j++)
2343 kfree(org_mount_opt.s_qf_names[j]);
2344 return -ENOMEM;
2345 }
2346 } else {
2347 org_mount_opt.s_qf_names[i] = NULL;
2348 }
2349 }
2350#endif
2351
2352 /* recover superblocks we couldn't write due to previous RO mount */
2353 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2354 err = f2fs_commit_super(sbi, false);
2355 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2356 err);
2357 if (!err)
2358 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2359 }
2360
2361 default_options(sbi, true);
2362
2363 /* parse mount options */
2364 err = parse_options(sb, data, true);
2365 if (err)
2366 goto restore_opts;
2367
2368#ifdef CONFIG_BLK_DEV_ZONED
2369 if (f2fs_sb_has_blkzoned(sbi) &&
2370 sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
2371 f2fs_err(sbi,
2372 "zoned: max open zones %u is too small, need at least %u open zones",
2373 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
2374 err = -EINVAL;
2375 goto restore_opts;
2376 }
2377#endif
2378
2379 /* flush outstanding errors before changing fs state */
2380 flush_work(&sbi->s_error_work);
2381
2382 /*
2383 * Previous and new state of filesystem is RO,
2384 * so skip checking GC and FLUSH_MERGE conditions.
2385 */
2386 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2387 goto skip;
2388
2389 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2390 err = -EROFS;
2391 goto restore_opts;
2392 }
2393
2394#ifdef CONFIG_QUOTA
2395 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2396 err = dquot_suspend(sb, -1);
2397 if (err < 0)
2398 goto restore_opts;
2399 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2400 /* dquot_resume needs RW */
2401 sb->s_flags &= ~SB_RDONLY;
2402 if (sb_any_quota_suspended(sb)) {
2403 dquot_resume(sb, -1);
2404 } else if (f2fs_sb_has_quota_ino(sbi)) {
2405 err = f2fs_enable_quotas(sb);
2406 if (err)
2407 goto restore_opts;
2408 }
2409 }
2410#endif
2411 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2412 err = -EINVAL;
2413 f2fs_warn(sbi, "LFS is not compatible with IPU");
2414 goto restore_opts;
2415 }
2416
2417 /* disallow enable atgc dynamically */
2418 if (no_atgc == !!test_opt(sbi, ATGC)) {
2419 err = -EINVAL;
2420 f2fs_warn(sbi, "switch atgc option is not allowed");
2421 goto restore_opts;
2422 }
2423
2424 /* disallow enable/disable extent_cache dynamically */
2425 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2426 err = -EINVAL;
2427 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2428 goto restore_opts;
2429 }
2430 /* disallow enable/disable age extent_cache dynamically */
2431 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2432 err = -EINVAL;
2433 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2434 goto restore_opts;
2435 }
2436
2437 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2438 err = -EINVAL;
2439 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2440 goto restore_opts;
2441 }
2442
2443 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2444 err = -EINVAL;
2445 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2446 goto restore_opts;
2447 }
2448
2449 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2450 err = -EINVAL;
2451 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2452 goto restore_opts;
2453 }
2454
2455 /*
2456 * We stop the GC thread if FS is mounted as RO
2457 * or if background_gc = off is passed in mount
2458 * option. Also sync the filesystem.
2459 */
2460 if ((*flags & SB_RDONLY) ||
2461 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2462 !test_opt(sbi, GC_MERGE))) {
2463 if (sbi->gc_thread) {
2464 f2fs_stop_gc_thread(sbi);
2465 need_restart_gc = true;
2466 }
2467 } else if (!sbi->gc_thread) {
2468 err = f2fs_start_gc_thread(sbi);
2469 if (err)
2470 goto restore_opts;
2471 need_stop_gc = true;
2472 }
2473
2474 if (*flags & SB_RDONLY) {
2475 sync_inodes_sb(sb);
2476
2477 set_sbi_flag(sbi, SBI_IS_DIRTY);
2478 set_sbi_flag(sbi, SBI_IS_CLOSE);
2479 f2fs_sync_fs(sb, 1);
2480 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2481 }
2482
2483 /*
2484 * We stop issue flush thread if FS is mounted as RO
2485 * or if flush_merge is not passed in mount option.
2486 */
2487 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2488 clear_opt(sbi, FLUSH_MERGE);
2489 f2fs_destroy_flush_cmd_control(sbi, false);
2490 need_restart_flush = true;
2491 } else {
2492 err = f2fs_create_flush_cmd_control(sbi);
2493 if (err)
2494 goto restore_gc;
2495 need_stop_flush = true;
2496 }
2497
2498 if (no_discard == !!test_opt(sbi, DISCARD)) {
2499 if (test_opt(sbi, DISCARD)) {
2500 err = f2fs_start_discard_thread(sbi);
2501 if (err)
2502 goto restore_flush;
2503 need_stop_discard = true;
2504 } else {
2505 f2fs_stop_discard_thread(sbi);
2506 f2fs_issue_discard_timeout(sbi);
2507 need_restart_discard = true;
2508 }
2509 }
2510
2511 adjust_unusable_cap_perc(sbi);
2512 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2513 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2514 err = f2fs_disable_checkpoint(sbi);
2515 if (err)
2516 goto restore_discard;
2517 need_enable_checkpoint = true;
2518 } else {
2519 f2fs_enable_checkpoint(sbi);
2520 need_disable_checkpoint = true;
2521 }
2522 }
2523
2524 /*
2525 * Place this routine at the end, since a new checkpoint would be
2526 * triggered while remount and we need to take care of it before
2527 * returning from remount.
2528 */
2529 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2530 !test_opt(sbi, MERGE_CHECKPOINT)) {
2531 f2fs_stop_ckpt_thread(sbi);
2532 } else {
2533 /* Flush if the prevous checkpoint, if exists. */
2534 f2fs_flush_ckpt_thread(sbi);
2535
2536 err = f2fs_start_ckpt_thread(sbi);
2537 if (err) {
2538 f2fs_err(sbi,
2539 "Failed to start F2FS issue_checkpoint_thread (%d)",
2540 err);
2541 goto restore_checkpoint;
2542 }
2543 }
2544
2545skip:
2546#ifdef CONFIG_QUOTA
2547 /* Release old quota file names */
2548 for (i = 0; i < MAXQUOTAS; i++)
2549 kfree(org_mount_opt.s_qf_names[i]);
2550#endif
2551 /* Update the POSIXACL Flag */
2552 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2553 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2554
2555 limit_reserve_root(sbi);
2556 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2557 return 0;
2558restore_checkpoint:
2559 if (need_enable_checkpoint) {
2560 f2fs_enable_checkpoint(sbi);
2561 } else if (need_disable_checkpoint) {
2562 if (f2fs_disable_checkpoint(sbi))
2563 f2fs_warn(sbi, "checkpoint has not been disabled");
2564 }
2565restore_discard:
2566 if (need_restart_discard) {
2567 if (f2fs_start_discard_thread(sbi))
2568 f2fs_warn(sbi, "discard has been stopped");
2569 } else if (need_stop_discard) {
2570 f2fs_stop_discard_thread(sbi);
2571 }
2572restore_flush:
2573 if (need_restart_flush) {
2574 if (f2fs_create_flush_cmd_control(sbi))
2575 f2fs_warn(sbi, "background flush thread has stopped");
2576 } else if (need_stop_flush) {
2577 clear_opt(sbi, FLUSH_MERGE);
2578 f2fs_destroy_flush_cmd_control(sbi, false);
2579 }
2580restore_gc:
2581 if (need_restart_gc) {
2582 if (f2fs_start_gc_thread(sbi))
2583 f2fs_warn(sbi, "background gc thread has stopped");
2584 } else if (need_stop_gc) {
2585 f2fs_stop_gc_thread(sbi);
2586 }
2587restore_opts:
2588#ifdef CONFIG_QUOTA
2589 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2590 for (i = 0; i < MAXQUOTAS; i++) {
2591 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2592 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2593 }
2594#endif
2595 sbi->mount_opt = org_mount_opt;
2596 sb->s_flags = old_sb_flags;
2597 return err;
2598}
2599
2600static void f2fs_shutdown(struct super_block *sb)
2601{
2602 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false, false);
2603}
2604
2605#ifdef CONFIG_QUOTA
2606static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2607{
2608 /* need to recovery orphan */
2609 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2610 return true;
2611 /* need to recovery data */
2612 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2613 return false;
2614 if (test_opt(sbi, NORECOVERY))
2615 return false;
2616 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2617}
2618
2619static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2620{
2621 bool readonly = f2fs_readonly(sbi->sb);
2622
2623 if (!f2fs_need_recovery(sbi))
2624 return false;
2625
2626 /* it doesn't need to check f2fs_sb_has_readonly() */
2627 if (f2fs_hw_is_readonly(sbi))
2628 return false;
2629
2630 if (readonly) {
2631 sbi->sb->s_flags &= ~SB_RDONLY;
2632 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2633 }
2634
2635 /*
2636 * Turn on quotas which were not enabled for read-only mounts if
2637 * filesystem has quota feature, so that they are updated correctly.
2638 */
2639 return f2fs_enable_quota_files(sbi, readonly);
2640}
2641
2642static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2643 bool quota_enabled)
2644{
2645 if (quota_enabled)
2646 f2fs_quota_off_umount(sbi->sb);
2647
2648 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2649 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2650 sbi->sb->s_flags |= SB_RDONLY;
2651 }
2652}
2653
2654/* Read data from quotafile */
2655static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2656 size_t len, loff_t off)
2657{
2658 struct inode *inode = sb_dqopt(sb)->files[type];
2659 struct address_space *mapping = inode->i_mapping;
2660 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2661 int offset = off & (sb->s_blocksize - 1);
2662 int tocopy;
2663 size_t toread;
2664 loff_t i_size = i_size_read(inode);
2665 struct page *page;
2666
2667 if (off > i_size)
2668 return 0;
2669
2670 if (off + len > i_size)
2671 len = i_size - off;
2672 toread = len;
2673 while (toread > 0) {
2674 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2675repeat:
2676 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2677 if (IS_ERR(page)) {
2678 if (PTR_ERR(page) == -ENOMEM) {
2679 memalloc_retry_wait(GFP_NOFS);
2680 goto repeat;
2681 }
2682 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2683 return PTR_ERR(page);
2684 }
2685
2686 lock_page(page);
2687
2688 if (unlikely(page->mapping != mapping)) {
2689 f2fs_put_page(page, 1);
2690 goto repeat;
2691 }
2692 if (unlikely(!PageUptodate(page))) {
2693 f2fs_put_page(page, 1);
2694 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2695 return -EIO;
2696 }
2697
2698 memcpy_from_page(data, page, offset, tocopy);
2699 f2fs_put_page(page, 1);
2700
2701 offset = 0;
2702 toread -= tocopy;
2703 data += tocopy;
2704 blkidx++;
2705 }
2706 return len;
2707}
2708
2709/* Write to quotafile */
2710static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2711 const char *data, size_t len, loff_t off)
2712{
2713 struct inode *inode = sb_dqopt(sb)->files[type];
2714 struct address_space *mapping = inode->i_mapping;
2715 const struct address_space_operations *a_ops = mapping->a_ops;
2716 int offset = off & (sb->s_blocksize - 1);
2717 size_t towrite = len;
2718 struct folio *folio;
2719 void *fsdata = NULL;
2720 int err = 0;
2721 int tocopy;
2722
2723 while (towrite > 0) {
2724 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2725 towrite);
2726retry:
2727 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2728 &folio, &fsdata);
2729 if (unlikely(err)) {
2730 if (err == -ENOMEM) {
2731 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2732 goto retry;
2733 }
2734 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2735 break;
2736 }
2737
2738 memcpy_to_folio(folio, offset_in_folio(folio, off), data, tocopy);
2739
2740 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2741 folio, fsdata);
2742 offset = 0;
2743 towrite -= tocopy;
2744 off += tocopy;
2745 data += tocopy;
2746 cond_resched();
2747 }
2748
2749 if (len == towrite)
2750 return err;
2751 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2752 f2fs_mark_inode_dirty_sync(inode, false);
2753 return len - towrite;
2754}
2755
2756int f2fs_dquot_initialize(struct inode *inode)
2757{
2758 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2759 return -ESRCH;
2760
2761 return dquot_initialize(inode);
2762}
2763
2764static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2765{
2766 return F2FS_I(inode)->i_dquot;
2767}
2768
2769static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2770{
2771 return &F2FS_I(inode)->i_reserved_quota;
2772}
2773
2774static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2775{
2776 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2777 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2778 return 0;
2779 }
2780
2781 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2782 F2FS_OPTION(sbi).s_jquota_fmt, type);
2783}
2784
2785int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2786{
2787 int enabled = 0;
2788 int i, err;
2789
2790 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2791 err = f2fs_enable_quotas(sbi->sb);
2792 if (err) {
2793 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2794 return 0;
2795 }
2796 return 1;
2797 }
2798
2799 for (i = 0; i < MAXQUOTAS; i++) {
2800 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2801 err = f2fs_quota_on_mount(sbi, i);
2802 if (!err) {
2803 enabled = 1;
2804 continue;
2805 }
2806 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2807 err, i);
2808 }
2809 }
2810 return enabled;
2811}
2812
2813static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2814 unsigned int flags)
2815{
2816 struct inode *qf_inode;
2817 unsigned long qf_inum;
2818 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2819 int err;
2820
2821 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2822
2823 qf_inum = f2fs_qf_ino(sb, type);
2824 if (!qf_inum)
2825 return -EPERM;
2826
2827 qf_inode = f2fs_iget(sb, qf_inum);
2828 if (IS_ERR(qf_inode)) {
2829 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2830 return PTR_ERR(qf_inode);
2831 }
2832
2833 /* Don't account quota for quota files to avoid recursion */
2834 inode_lock(qf_inode);
2835 qf_inode->i_flags |= S_NOQUOTA;
2836
2837 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2838 F2FS_I(qf_inode)->i_flags |= qf_flag;
2839 f2fs_set_inode_flags(qf_inode);
2840 }
2841 inode_unlock(qf_inode);
2842
2843 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2844 iput(qf_inode);
2845 return err;
2846}
2847
2848static int f2fs_enable_quotas(struct super_block *sb)
2849{
2850 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2851 int type, err = 0;
2852 unsigned long qf_inum;
2853 bool quota_mopt[MAXQUOTAS] = {
2854 test_opt(sbi, USRQUOTA),
2855 test_opt(sbi, GRPQUOTA),
2856 test_opt(sbi, PRJQUOTA),
2857 };
2858
2859 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2860 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2861 return 0;
2862 }
2863
2864 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2865
2866 for (type = 0; type < MAXQUOTAS; type++) {
2867 qf_inum = f2fs_qf_ino(sb, type);
2868 if (qf_inum) {
2869 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2870 DQUOT_USAGE_ENABLED |
2871 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2872 if (err) {
2873 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2874 type, err);
2875 for (type--; type >= 0; type--)
2876 dquot_quota_off(sb, type);
2877 set_sbi_flag(F2FS_SB(sb),
2878 SBI_QUOTA_NEED_REPAIR);
2879 return err;
2880 }
2881 }
2882 }
2883 return 0;
2884}
2885
2886static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2887{
2888 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2889 struct address_space *mapping = dqopt->files[type]->i_mapping;
2890 int ret = 0;
2891
2892 ret = dquot_writeback_dquots(sbi->sb, type);
2893 if (ret)
2894 goto out;
2895
2896 ret = filemap_fdatawrite(mapping);
2897 if (ret)
2898 goto out;
2899
2900 /* if we are using journalled quota */
2901 if (is_journalled_quota(sbi))
2902 goto out;
2903
2904 ret = filemap_fdatawait(mapping);
2905
2906 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2907out:
2908 if (ret)
2909 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2910 return ret;
2911}
2912
2913int f2fs_quota_sync(struct super_block *sb, int type)
2914{
2915 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2916 struct quota_info *dqopt = sb_dqopt(sb);
2917 int cnt;
2918 int ret = 0;
2919
2920 /*
2921 * Now when everything is written we can discard the pagecache so
2922 * that userspace sees the changes.
2923 */
2924 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2925
2926 if (type != -1 && cnt != type)
2927 continue;
2928
2929 if (!sb_has_quota_active(sb, cnt))
2930 continue;
2931
2932 if (!f2fs_sb_has_quota_ino(sbi))
2933 inode_lock(dqopt->files[cnt]);
2934
2935 /*
2936 * do_quotactl
2937 * f2fs_quota_sync
2938 * f2fs_down_read(quota_sem)
2939 * dquot_writeback_dquots()
2940 * f2fs_dquot_commit
2941 * block_operation
2942 * f2fs_down_read(quota_sem)
2943 */
2944 f2fs_lock_op(sbi);
2945 f2fs_down_read(&sbi->quota_sem);
2946
2947 ret = f2fs_quota_sync_file(sbi, cnt);
2948
2949 f2fs_up_read(&sbi->quota_sem);
2950 f2fs_unlock_op(sbi);
2951
2952 if (!f2fs_sb_has_quota_ino(sbi))
2953 inode_unlock(dqopt->files[cnt]);
2954
2955 if (ret)
2956 break;
2957 }
2958 return ret;
2959}
2960
2961static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2962 const struct path *path)
2963{
2964 struct inode *inode;
2965 int err;
2966
2967 /* if quota sysfile exists, deny enabling quota with specific file */
2968 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2969 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2970 return -EBUSY;
2971 }
2972
2973 if (path->dentry->d_sb != sb)
2974 return -EXDEV;
2975
2976 err = f2fs_quota_sync(sb, type);
2977 if (err)
2978 return err;
2979
2980 inode = d_inode(path->dentry);
2981
2982 err = filemap_fdatawrite(inode->i_mapping);
2983 if (err)
2984 return err;
2985
2986 err = filemap_fdatawait(inode->i_mapping);
2987 if (err)
2988 return err;
2989
2990 err = dquot_quota_on(sb, type, format_id, path);
2991 if (err)
2992 return err;
2993
2994 inode_lock(inode);
2995 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2996 f2fs_set_inode_flags(inode);
2997 inode_unlock(inode);
2998 f2fs_mark_inode_dirty_sync(inode, false);
2999
3000 return 0;
3001}
3002
3003static int __f2fs_quota_off(struct super_block *sb, int type)
3004{
3005 struct inode *inode = sb_dqopt(sb)->files[type];
3006 int err;
3007
3008 if (!inode || !igrab(inode))
3009 return dquot_quota_off(sb, type);
3010
3011 err = f2fs_quota_sync(sb, type);
3012 if (err)
3013 goto out_put;
3014
3015 err = dquot_quota_off(sb, type);
3016 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
3017 goto out_put;
3018
3019 inode_lock(inode);
3020 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
3021 f2fs_set_inode_flags(inode);
3022 inode_unlock(inode);
3023 f2fs_mark_inode_dirty_sync(inode, false);
3024out_put:
3025 iput(inode);
3026 return err;
3027}
3028
3029static int f2fs_quota_off(struct super_block *sb, int type)
3030{
3031 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3032 int err;
3033
3034 err = __f2fs_quota_off(sb, type);
3035
3036 /*
3037 * quotactl can shutdown journalled quota, result in inconsistence
3038 * between quota record and fs data by following updates, tag the
3039 * flag to let fsck be aware of it.
3040 */
3041 if (is_journalled_quota(sbi))
3042 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3043 return err;
3044}
3045
3046void f2fs_quota_off_umount(struct super_block *sb)
3047{
3048 int type;
3049 int err;
3050
3051 for (type = 0; type < MAXQUOTAS; type++) {
3052 err = __f2fs_quota_off(sb, type);
3053 if (err) {
3054 int ret = dquot_quota_off(sb, type);
3055
3056 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3057 type, err, ret);
3058 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3059 }
3060 }
3061 /*
3062 * In case of checkpoint=disable, we must flush quota blocks.
3063 * This can cause NULL exception for node_inode in end_io, since
3064 * put_super already dropped it.
3065 */
3066 sync_filesystem(sb);
3067}
3068
3069static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3070{
3071 struct quota_info *dqopt = sb_dqopt(sb);
3072 int type;
3073
3074 for (type = 0; type < MAXQUOTAS; type++) {
3075 if (!dqopt->files[type])
3076 continue;
3077 f2fs_inode_synced(dqopt->files[type]);
3078 }
3079}
3080
3081static int f2fs_dquot_commit(struct dquot *dquot)
3082{
3083 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3084 int ret;
3085
3086 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3087 ret = dquot_commit(dquot);
3088 if (ret < 0)
3089 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3090 f2fs_up_read(&sbi->quota_sem);
3091 return ret;
3092}
3093
3094static int f2fs_dquot_acquire(struct dquot *dquot)
3095{
3096 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3097 int ret;
3098
3099 f2fs_down_read(&sbi->quota_sem);
3100 ret = dquot_acquire(dquot);
3101 if (ret < 0)
3102 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3103 f2fs_up_read(&sbi->quota_sem);
3104 return ret;
3105}
3106
3107static int f2fs_dquot_release(struct dquot *dquot)
3108{
3109 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3110 int ret = dquot_release(dquot);
3111
3112 if (ret < 0)
3113 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3114 return ret;
3115}
3116
3117static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3118{
3119 struct super_block *sb = dquot->dq_sb;
3120 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3121 int ret = dquot_mark_dquot_dirty(dquot);
3122
3123 /* if we are using journalled quota */
3124 if (is_journalled_quota(sbi))
3125 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3126
3127 return ret;
3128}
3129
3130static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3131{
3132 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3133 int ret = dquot_commit_info(sb, type);
3134
3135 if (ret < 0)
3136 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3137 return ret;
3138}
3139
3140static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3141{
3142 *projid = F2FS_I(inode)->i_projid;
3143 return 0;
3144}
3145
3146static const struct dquot_operations f2fs_quota_operations = {
3147 .get_reserved_space = f2fs_get_reserved_space,
3148 .write_dquot = f2fs_dquot_commit,
3149 .acquire_dquot = f2fs_dquot_acquire,
3150 .release_dquot = f2fs_dquot_release,
3151 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3152 .write_info = f2fs_dquot_commit_info,
3153 .alloc_dquot = dquot_alloc,
3154 .destroy_dquot = dquot_destroy,
3155 .get_projid = f2fs_get_projid,
3156 .get_next_id = dquot_get_next_id,
3157};
3158
3159static const struct quotactl_ops f2fs_quotactl_ops = {
3160 .quota_on = f2fs_quota_on,
3161 .quota_off = f2fs_quota_off,
3162 .quota_sync = f2fs_quota_sync,
3163 .get_state = dquot_get_state,
3164 .set_info = dquot_set_dqinfo,
3165 .get_dqblk = dquot_get_dqblk,
3166 .set_dqblk = dquot_set_dqblk,
3167 .get_nextdqblk = dquot_get_next_dqblk,
3168};
3169#else
3170int f2fs_dquot_initialize(struct inode *inode)
3171{
3172 return 0;
3173}
3174
3175int f2fs_quota_sync(struct super_block *sb, int type)
3176{
3177 return 0;
3178}
3179
3180void f2fs_quota_off_umount(struct super_block *sb)
3181{
3182}
3183#endif
3184
3185static const struct super_operations f2fs_sops = {
3186 .alloc_inode = f2fs_alloc_inode,
3187 .free_inode = f2fs_free_inode,
3188 .drop_inode = f2fs_drop_inode,
3189 .write_inode = f2fs_write_inode,
3190 .dirty_inode = f2fs_dirty_inode,
3191 .show_options = f2fs_show_options,
3192#ifdef CONFIG_QUOTA
3193 .quota_read = f2fs_quota_read,
3194 .quota_write = f2fs_quota_write,
3195 .get_dquots = f2fs_get_dquots,
3196#endif
3197 .evict_inode = f2fs_evict_inode,
3198 .put_super = f2fs_put_super,
3199 .sync_fs = f2fs_sync_fs,
3200 .freeze_fs = f2fs_freeze,
3201 .unfreeze_fs = f2fs_unfreeze,
3202 .statfs = f2fs_statfs,
3203 .remount_fs = f2fs_remount,
3204 .shutdown = f2fs_shutdown,
3205};
3206
3207#ifdef CONFIG_FS_ENCRYPTION
3208static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3209{
3210 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3211 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3212 ctx, len, NULL);
3213}
3214
3215static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3216 void *fs_data)
3217{
3218 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3219
3220 /*
3221 * Encrypting the root directory is not allowed because fsck
3222 * expects lost+found directory to exist and remain unencrypted
3223 * if LOST_FOUND feature is enabled.
3224 *
3225 */
3226 if (f2fs_sb_has_lost_found(sbi) &&
3227 inode->i_ino == F2FS_ROOT_INO(sbi))
3228 return -EPERM;
3229
3230 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3231 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3232 ctx, len, fs_data, XATTR_CREATE);
3233}
3234
3235static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3236{
3237 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3238}
3239
3240static bool f2fs_has_stable_inodes(struct super_block *sb)
3241{
3242 return true;
3243}
3244
3245static struct block_device **f2fs_get_devices(struct super_block *sb,
3246 unsigned int *num_devs)
3247{
3248 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3249 struct block_device **devs;
3250 int i;
3251
3252 if (!f2fs_is_multi_device(sbi))
3253 return NULL;
3254
3255 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3256 if (!devs)
3257 return ERR_PTR(-ENOMEM);
3258
3259 for (i = 0; i < sbi->s_ndevs; i++)
3260 devs[i] = FDEV(i).bdev;
3261 *num_devs = sbi->s_ndevs;
3262 return devs;
3263}
3264
3265static const struct fscrypt_operations f2fs_cryptops = {
3266 .needs_bounce_pages = 1,
3267 .has_32bit_inodes = 1,
3268 .supports_subblock_data_units = 1,
3269 .legacy_key_prefix = "f2fs:",
3270 .get_context = f2fs_get_context,
3271 .set_context = f2fs_set_context,
3272 .get_dummy_policy = f2fs_get_dummy_policy,
3273 .empty_dir = f2fs_empty_dir,
3274 .has_stable_inodes = f2fs_has_stable_inodes,
3275 .get_devices = f2fs_get_devices,
3276};
3277#endif
3278
3279static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3280 u64 ino, u32 generation)
3281{
3282 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3283 struct inode *inode;
3284
3285 if (f2fs_check_nid_range(sbi, ino))
3286 return ERR_PTR(-ESTALE);
3287
3288 /*
3289 * f2fs_iget isn't quite right if the inode is currently unallocated!
3290 * However f2fs_iget currently does appropriate checks to handle stale
3291 * inodes so everything is OK.
3292 */
3293 inode = f2fs_iget(sb, ino);
3294 if (IS_ERR(inode))
3295 return ERR_CAST(inode);
3296 if (unlikely(generation && inode->i_generation != generation)) {
3297 /* we didn't find the right inode.. */
3298 iput(inode);
3299 return ERR_PTR(-ESTALE);
3300 }
3301 return inode;
3302}
3303
3304static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3305 int fh_len, int fh_type)
3306{
3307 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3308 f2fs_nfs_get_inode);
3309}
3310
3311static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3312 int fh_len, int fh_type)
3313{
3314 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3315 f2fs_nfs_get_inode);
3316}
3317
3318static const struct export_operations f2fs_export_ops = {
3319 .encode_fh = generic_encode_ino32_fh,
3320 .fh_to_dentry = f2fs_fh_to_dentry,
3321 .fh_to_parent = f2fs_fh_to_parent,
3322 .get_parent = f2fs_get_parent,
3323};
3324
3325loff_t max_file_blocks(struct inode *inode)
3326{
3327 loff_t result = 0;
3328 loff_t leaf_count;
3329
3330 /*
3331 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3332 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3333 * space in inode.i_addr, it will be more safe to reassign
3334 * result as zero.
3335 */
3336
3337 if (inode && f2fs_compressed_file(inode))
3338 leaf_count = ADDRS_PER_BLOCK(inode);
3339 else
3340 leaf_count = DEF_ADDRS_PER_BLOCK;
3341
3342 /* two direct node blocks */
3343 result += (leaf_count * 2);
3344
3345 /* two indirect node blocks */
3346 leaf_count *= NIDS_PER_BLOCK;
3347 result += (leaf_count * 2);
3348
3349 /* one double indirect node block */
3350 leaf_count *= NIDS_PER_BLOCK;
3351 result += leaf_count;
3352
3353 /*
3354 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with
3355 * a 4K crypto data unit, we must restrict the max filesize to what can
3356 * fit within U32_MAX + 1 data units.
3357 */
3358
3359 result = umin(result, F2FS_BYTES_TO_BLK(((loff_t)U32_MAX + 1) * 4096));
3360
3361 return result;
3362}
3363
3364static int __f2fs_commit_super(struct f2fs_sb_info *sbi, struct folio *folio,
3365 pgoff_t index, bool update)
3366{
3367 struct bio *bio;
3368 /* it's rare case, we can do fua all the time */
3369 blk_opf_t opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH | REQ_FUA;
3370 int ret;
3371
3372 folio_lock(folio);
3373 folio_wait_writeback(folio);
3374 if (update)
3375 memcpy(F2FS_SUPER_BLOCK(folio, index), F2FS_RAW_SUPER(sbi),
3376 sizeof(struct f2fs_super_block));
3377 folio_mark_dirty(folio);
3378 folio_clear_dirty_for_io(folio);
3379 folio_start_writeback(folio);
3380 folio_unlock(folio);
3381
3382 bio = bio_alloc(sbi->sb->s_bdev, 1, opf, GFP_NOFS);
3383
3384 /* it doesn't need to set crypto context for superblock update */
3385 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(folio_index(folio));
3386
3387 if (!bio_add_folio(bio, folio, folio_size(folio), 0))
3388 f2fs_bug_on(sbi, 1);
3389
3390 ret = submit_bio_wait(bio);
3391 folio_end_writeback(folio);
3392
3393 return ret;
3394}
3395
3396static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3397 struct folio *folio, pgoff_t index)
3398{
3399 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index);
3400 struct super_block *sb = sbi->sb;
3401 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3402 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3403 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3404 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3405 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3406 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3407 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3408 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3409 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3410 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3411 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3412 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3413 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3414 u64 main_end_blkaddr = main_blkaddr +
3415 ((u64)segment_count_main << log_blocks_per_seg);
3416 u64 seg_end_blkaddr = segment0_blkaddr +
3417 ((u64)segment_count << log_blocks_per_seg);
3418
3419 if (segment0_blkaddr != cp_blkaddr) {
3420 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3421 segment0_blkaddr, cp_blkaddr);
3422 return true;
3423 }
3424
3425 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3426 sit_blkaddr) {
3427 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3428 cp_blkaddr, sit_blkaddr,
3429 segment_count_ckpt << log_blocks_per_seg);
3430 return true;
3431 }
3432
3433 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3434 nat_blkaddr) {
3435 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3436 sit_blkaddr, nat_blkaddr,
3437 segment_count_sit << log_blocks_per_seg);
3438 return true;
3439 }
3440
3441 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3442 ssa_blkaddr) {
3443 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3444 nat_blkaddr, ssa_blkaddr,
3445 segment_count_nat << log_blocks_per_seg);
3446 return true;
3447 }
3448
3449 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3450 main_blkaddr) {
3451 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3452 ssa_blkaddr, main_blkaddr,
3453 segment_count_ssa << log_blocks_per_seg);
3454 return true;
3455 }
3456
3457 if (main_end_blkaddr > seg_end_blkaddr) {
3458 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3459 main_blkaddr, seg_end_blkaddr,
3460 segment_count_main << log_blocks_per_seg);
3461 return true;
3462 } else if (main_end_blkaddr < seg_end_blkaddr) {
3463 int err = 0;
3464 char *res;
3465
3466 /* fix in-memory information all the time */
3467 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3468 segment0_blkaddr) >> log_blocks_per_seg);
3469
3470 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3471 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3472 res = "internally";
3473 } else {
3474 err = __f2fs_commit_super(sbi, folio, index, false);
3475 res = err ? "failed" : "done";
3476 }
3477 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3478 res, main_blkaddr, seg_end_blkaddr,
3479 segment_count_main << log_blocks_per_seg);
3480 if (err)
3481 return true;
3482 }
3483 return false;
3484}
3485
3486static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3487 struct folio *folio, pgoff_t index)
3488{
3489 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3490 block_t total_sections, blocks_per_seg;
3491 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index);
3492 size_t crc_offset = 0;
3493 __u32 crc = 0;
3494
3495 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3496 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3497 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3498 return -EINVAL;
3499 }
3500
3501 /* Check checksum_offset and crc in superblock */
3502 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3503 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3504 if (crc_offset !=
3505 offsetof(struct f2fs_super_block, crc)) {
3506 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3507 crc_offset);
3508 return -EFSCORRUPTED;
3509 }
3510 crc = le32_to_cpu(raw_super->crc);
3511 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3512 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3513 return -EFSCORRUPTED;
3514 }
3515 }
3516
3517 /* only support block_size equals to PAGE_SIZE */
3518 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3519 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3520 le32_to_cpu(raw_super->log_blocksize),
3521 F2FS_BLKSIZE_BITS);
3522 return -EFSCORRUPTED;
3523 }
3524
3525 /* check log blocks per segment */
3526 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3527 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3528 le32_to_cpu(raw_super->log_blocks_per_seg));
3529 return -EFSCORRUPTED;
3530 }
3531
3532 /* Currently, support 512/1024/2048/4096/16K bytes sector size */
3533 if (le32_to_cpu(raw_super->log_sectorsize) >
3534 F2FS_MAX_LOG_SECTOR_SIZE ||
3535 le32_to_cpu(raw_super->log_sectorsize) <
3536 F2FS_MIN_LOG_SECTOR_SIZE) {
3537 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3538 le32_to_cpu(raw_super->log_sectorsize));
3539 return -EFSCORRUPTED;
3540 }
3541 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3542 le32_to_cpu(raw_super->log_sectorsize) !=
3543 F2FS_MAX_LOG_SECTOR_SIZE) {
3544 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3545 le32_to_cpu(raw_super->log_sectors_per_block),
3546 le32_to_cpu(raw_super->log_sectorsize));
3547 return -EFSCORRUPTED;
3548 }
3549
3550 segment_count = le32_to_cpu(raw_super->segment_count);
3551 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3552 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3553 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3554 total_sections = le32_to_cpu(raw_super->section_count);
3555
3556 /* blocks_per_seg should be 512, given the above check */
3557 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3558
3559 if (segment_count > F2FS_MAX_SEGMENT ||
3560 segment_count < F2FS_MIN_SEGMENTS) {
3561 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3562 return -EFSCORRUPTED;
3563 }
3564
3565 if (total_sections > segment_count_main || total_sections < 1 ||
3566 segs_per_sec > segment_count || !segs_per_sec) {
3567 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3568 segment_count, total_sections, segs_per_sec);
3569 return -EFSCORRUPTED;
3570 }
3571
3572 if (segment_count_main != total_sections * segs_per_sec) {
3573 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3574 segment_count_main, total_sections, segs_per_sec);
3575 return -EFSCORRUPTED;
3576 }
3577
3578 if ((segment_count / segs_per_sec) < total_sections) {
3579 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3580 segment_count, segs_per_sec, total_sections);
3581 return -EFSCORRUPTED;
3582 }
3583
3584 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3585 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3586 segment_count, le64_to_cpu(raw_super->block_count));
3587 return -EFSCORRUPTED;
3588 }
3589
3590 if (RDEV(0).path[0]) {
3591 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3592 int i = 1;
3593
3594 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3595 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3596 i++;
3597 }
3598 if (segment_count != dev_seg_count) {
3599 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3600 segment_count, dev_seg_count);
3601 return -EFSCORRUPTED;
3602 }
3603 } else {
3604 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3605 !bdev_is_zoned(sbi->sb->s_bdev)) {
3606 f2fs_info(sbi, "Zoned block device path is missing");
3607 return -EFSCORRUPTED;
3608 }
3609 }
3610
3611 if (secs_per_zone > total_sections || !secs_per_zone) {
3612 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3613 secs_per_zone, total_sections);
3614 return -EFSCORRUPTED;
3615 }
3616 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3617 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3618 (le32_to_cpu(raw_super->extension_count) +
3619 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3620 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3621 le32_to_cpu(raw_super->extension_count),
3622 raw_super->hot_ext_count,
3623 F2FS_MAX_EXTENSION);
3624 return -EFSCORRUPTED;
3625 }
3626
3627 if (le32_to_cpu(raw_super->cp_payload) >=
3628 (blocks_per_seg - F2FS_CP_PACKS -
3629 NR_CURSEG_PERSIST_TYPE)) {
3630 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3631 le32_to_cpu(raw_super->cp_payload),
3632 blocks_per_seg - F2FS_CP_PACKS -
3633 NR_CURSEG_PERSIST_TYPE);
3634 return -EFSCORRUPTED;
3635 }
3636
3637 /* check reserved ino info */
3638 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3639 le32_to_cpu(raw_super->meta_ino) != 2 ||
3640 le32_to_cpu(raw_super->root_ino) != 3) {
3641 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3642 le32_to_cpu(raw_super->node_ino),
3643 le32_to_cpu(raw_super->meta_ino),
3644 le32_to_cpu(raw_super->root_ino));
3645 return -EFSCORRUPTED;
3646 }
3647
3648 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3649 if (sanity_check_area_boundary(sbi, folio, index))
3650 return -EFSCORRUPTED;
3651
3652 return 0;
3653}
3654
3655int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3656{
3657 unsigned int total, fsmeta;
3658 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3659 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3660 unsigned int ovp_segments, reserved_segments;
3661 unsigned int main_segs, blocks_per_seg;
3662 unsigned int sit_segs, nat_segs;
3663 unsigned int sit_bitmap_size, nat_bitmap_size;
3664 unsigned int log_blocks_per_seg;
3665 unsigned int segment_count_main;
3666 unsigned int cp_pack_start_sum, cp_payload;
3667 block_t user_block_count, valid_user_blocks;
3668 block_t avail_node_count, valid_node_count;
3669 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3670 int i, j;
3671
3672 total = le32_to_cpu(raw_super->segment_count);
3673 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3674 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3675 fsmeta += sit_segs;
3676 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3677 fsmeta += nat_segs;
3678 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3679 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3680
3681 if (unlikely(fsmeta >= total))
3682 return 1;
3683
3684 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3685 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3686
3687 if (!f2fs_sb_has_readonly(sbi) &&
3688 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3689 ovp_segments == 0 || reserved_segments == 0)) {
3690 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3691 return 1;
3692 }
3693 user_block_count = le64_to_cpu(ckpt->user_block_count);
3694 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3695 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3696 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3697 if (!user_block_count || user_block_count >=
3698 segment_count_main << log_blocks_per_seg) {
3699 f2fs_err(sbi, "Wrong user_block_count: %u",
3700 user_block_count);
3701 return 1;
3702 }
3703
3704 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3705 if (valid_user_blocks > user_block_count) {
3706 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3707 valid_user_blocks, user_block_count);
3708 return 1;
3709 }
3710
3711 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3712 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3713 if (valid_node_count > avail_node_count) {
3714 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3715 valid_node_count, avail_node_count);
3716 return 1;
3717 }
3718
3719 main_segs = le32_to_cpu(raw_super->segment_count_main);
3720 blocks_per_seg = BLKS_PER_SEG(sbi);
3721
3722 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3723 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3724 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3725 return 1;
3726
3727 if (f2fs_sb_has_readonly(sbi))
3728 goto check_data;
3729
3730 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3731 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3732 le32_to_cpu(ckpt->cur_node_segno[j])) {
3733 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3734 i, j,
3735 le32_to_cpu(ckpt->cur_node_segno[i]));
3736 return 1;
3737 }
3738 }
3739 }
3740check_data:
3741 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3742 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3743 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3744 return 1;
3745
3746 if (f2fs_sb_has_readonly(sbi))
3747 goto skip_cross;
3748
3749 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3750 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3751 le32_to_cpu(ckpt->cur_data_segno[j])) {
3752 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3753 i, j,
3754 le32_to_cpu(ckpt->cur_data_segno[i]));
3755 return 1;
3756 }
3757 }
3758 }
3759 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3760 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3761 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3762 le32_to_cpu(ckpt->cur_data_segno[j])) {
3763 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3764 i, j,
3765 le32_to_cpu(ckpt->cur_node_segno[i]));
3766 return 1;
3767 }
3768 }
3769 }
3770skip_cross:
3771 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3772 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3773
3774 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3775 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3776 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3777 sit_bitmap_size, nat_bitmap_size);
3778 return 1;
3779 }
3780
3781 cp_pack_start_sum = __start_sum_addr(sbi);
3782 cp_payload = __cp_payload(sbi);
3783 if (cp_pack_start_sum < cp_payload + 1 ||
3784 cp_pack_start_sum > blocks_per_seg - 1 -
3785 NR_CURSEG_PERSIST_TYPE) {
3786 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3787 cp_pack_start_sum);
3788 return 1;
3789 }
3790
3791 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3792 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3793 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3794 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3795 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3796 le32_to_cpu(ckpt->checksum_offset));
3797 return 1;
3798 }
3799
3800 nat_blocks = nat_segs << log_blocks_per_seg;
3801 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3802 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3803 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3804 (cp_payload + F2FS_CP_PACKS +
3805 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3806 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3807 cp_payload, nat_bits_blocks);
3808 return 1;
3809 }
3810
3811 if (unlikely(f2fs_cp_error(sbi))) {
3812 f2fs_err(sbi, "A bug case: need to run fsck");
3813 return 1;
3814 }
3815 return 0;
3816}
3817
3818static void init_sb_info(struct f2fs_sb_info *sbi)
3819{
3820 struct f2fs_super_block *raw_super = sbi->raw_super;
3821 int i;
3822
3823 sbi->log_sectors_per_block =
3824 le32_to_cpu(raw_super->log_sectors_per_block);
3825 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3826 sbi->blocksize = BIT(sbi->log_blocksize);
3827 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3828 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3829 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3830 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3831 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3832 sbi->total_node_count = SEGS_TO_BLKS(sbi,
3833 ((le32_to_cpu(raw_super->segment_count_nat) / 2) *
3834 NAT_ENTRY_PER_BLOCK));
3835 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3836 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3837 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3838 sbi->cur_victim_sec = NULL_SECNO;
3839 sbi->gc_mode = GC_NORMAL;
3840 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3841 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3842 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3843 sbi->migration_granularity = SEGS_PER_SEC(sbi);
3844 sbi->migration_window_granularity = f2fs_sb_has_blkzoned(sbi) ?
3845 DEF_MIGRATION_WINDOW_GRANULARITY_ZONED : SEGS_PER_SEC(sbi);
3846 sbi->seq_file_ra_mul = MIN_RA_MUL;
3847 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3848 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3849 spin_lock_init(&sbi->gc_remaining_trials_lock);
3850 atomic64_set(&sbi->current_atomic_write, 0);
3851
3852 sbi->dir_level = DEF_DIR_LEVEL;
3853 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3854 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3855 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3856 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3857 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3858 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3859 DEF_UMOUNT_DISCARD_TIMEOUT;
3860 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3861
3862 for (i = 0; i < NR_COUNT_TYPE; i++)
3863 atomic_set(&sbi->nr_pages[i], 0);
3864
3865 for (i = 0; i < META; i++)
3866 atomic_set(&sbi->wb_sync_req[i], 0);
3867
3868 INIT_LIST_HEAD(&sbi->s_list);
3869 mutex_init(&sbi->umount_mutex);
3870 init_f2fs_rwsem(&sbi->io_order_lock);
3871 spin_lock_init(&sbi->cp_lock);
3872
3873 sbi->dirty_device = 0;
3874 spin_lock_init(&sbi->dev_lock);
3875
3876 init_f2fs_rwsem(&sbi->sb_lock);
3877 init_f2fs_rwsem(&sbi->pin_sem);
3878}
3879
3880static int init_percpu_info(struct f2fs_sb_info *sbi)
3881{
3882 int err;
3883
3884 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3885 if (err)
3886 return err;
3887
3888 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3889 if (err)
3890 goto err_valid_block;
3891
3892 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3893 GFP_KERNEL);
3894 if (err)
3895 goto err_node_block;
3896 return 0;
3897
3898err_node_block:
3899 percpu_counter_destroy(&sbi->rf_node_block_count);
3900err_valid_block:
3901 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3902 return err;
3903}
3904
3905#ifdef CONFIG_BLK_DEV_ZONED
3906
3907struct f2fs_report_zones_args {
3908 struct f2fs_sb_info *sbi;
3909 struct f2fs_dev_info *dev;
3910};
3911
3912static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3913 void *data)
3914{
3915 struct f2fs_report_zones_args *rz_args = data;
3916 block_t unusable_blocks = (zone->len - zone->capacity) >>
3917 F2FS_LOG_SECTORS_PER_BLOCK;
3918
3919 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3920 return 0;
3921
3922 set_bit(idx, rz_args->dev->blkz_seq);
3923 if (!rz_args->sbi->unusable_blocks_per_sec) {
3924 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3925 return 0;
3926 }
3927 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3928 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3929 return -EINVAL;
3930 }
3931 return 0;
3932}
3933
3934static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3935{
3936 struct block_device *bdev = FDEV(devi).bdev;
3937 sector_t nr_sectors = bdev_nr_sectors(bdev);
3938 struct f2fs_report_zones_args rep_zone_arg;
3939 u64 zone_sectors;
3940 unsigned int max_open_zones;
3941 int ret;
3942
3943 if (!f2fs_sb_has_blkzoned(sbi))
3944 return 0;
3945
3946 if (bdev_is_zoned(FDEV(devi).bdev)) {
3947 max_open_zones = bdev_max_open_zones(bdev);
3948 if (max_open_zones && (max_open_zones < sbi->max_open_zones))
3949 sbi->max_open_zones = max_open_zones;
3950 if (sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
3951 f2fs_err(sbi,
3952 "zoned: max open zones %u is too small, need at least %u open zones",
3953 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
3954 return -EINVAL;
3955 }
3956 }
3957
3958 zone_sectors = bdev_zone_sectors(bdev);
3959 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3960 SECTOR_TO_BLOCK(zone_sectors))
3961 return -EINVAL;
3962 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3963 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3964 sbi->blocks_per_blkz);
3965 if (nr_sectors & (zone_sectors - 1))
3966 FDEV(devi).nr_blkz++;
3967
3968 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3969 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3970 * sizeof(unsigned long),
3971 GFP_KERNEL);
3972 if (!FDEV(devi).blkz_seq)
3973 return -ENOMEM;
3974
3975 rep_zone_arg.sbi = sbi;
3976 rep_zone_arg.dev = &FDEV(devi);
3977
3978 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3979 &rep_zone_arg);
3980 if (ret < 0)
3981 return ret;
3982 return 0;
3983}
3984#endif
3985
3986/*
3987 * Read f2fs raw super block.
3988 * Because we have two copies of super block, so read both of them
3989 * to get the first valid one. If any one of them is broken, we pass
3990 * them recovery flag back to the caller.
3991 */
3992static int read_raw_super_block(struct f2fs_sb_info *sbi,
3993 struct f2fs_super_block **raw_super,
3994 int *valid_super_block, int *recovery)
3995{
3996 struct super_block *sb = sbi->sb;
3997 int block;
3998 struct folio *folio;
3999 struct f2fs_super_block *super;
4000 int err = 0;
4001
4002 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
4003 if (!super)
4004 return -ENOMEM;
4005
4006 for (block = 0; block < 2; block++) {
4007 folio = read_mapping_folio(sb->s_bdev->bd_mapping, block, NULL);
4008 if (IS_ERR(folio)) {
4009 f2fs_err(sbi, "Unable to read %dth superblock",
4010 block + 1);
4011 err = PTR_ERR(folio);
4012 *recovery = 1;
4013 continue;
4014 }
4015
4016 /* sanity checking of raw super */
4017 err = sanity_check_raw_super(sbi, folio, block);
4018 if (err) {
4019 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
4020 block + 1);
4021 folio_put(folio);
4022 *recovery = 1;
4023 continue;
4024 }
4025
4026 if (!*raw_super) {
4027 memcpy(super, F2FS_SUPER_BLOCK(folio, block),
4028 sizeof(*super));
4029 *valid_super_block = block;
4030 *raw_super = super;
4031 }
4032 folio_put(folio);
4033 }
4034
4035 /* No valid superblock */
4036 if (!*raw_super)
4037 kfree(super);
4038 else
4039 err = 0;
4040
4041 return err;
4042}
4043
4044int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
4045{
4046 struct folio *folio;
4047 pgoff_t index;
4048 __u32 crc = 0;
4049 int err;
4050
4051 if ((recover && f2fs_readonly(sbi->sb)) ||
4052 f2fs_hw_is_readonly(sbi)) {
4053 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
4054 return -EROFS;
4055 }
4056
4057 /* we should update superblock crc here */
4058 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
4059 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
4060 offsetof(struct f2fs_super_block, crc));
4061 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
4062 }
4063
4064 /* write back-up superblock first */
4065 index = sbi->valid_super_block ? 0 : 1;
4066 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL);
4067 if (IS_ERR(folio))
4068 return PTR_ERR(folio);
4069 err = __f2fs_commit_super(sbi, folio, index, true);
4070 folio_put(folio);
4071
4072 /* if we are in recovery path, skip writing valid superblock */
4073 if (recover || err)
4074 return err;
4075
4076 /* write current valid superblock */
4077 index = sbi->valid_super_block;
4078 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL);
4079 if (IS_ERR(folio))
4080 return PTR_ERR(folio);
4081 err = __f2fs_commit_super(sbi, folio, index, true);
4082 folio_put(folio);
4083 return err;
4084}
4085
4086static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4087{
4088 unsigned long flags;
4089
4090 spin_lock_irqsave(&sbi->error_lock, flags);
4091 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4092 sbi->stop_reason[reason]++;
4093 spin_unlock_irqrestore(&sbi->error_lock, flags);
4094}
4095
4096static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4097{
4098 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4099 unsigned long flags;
4100 int err;
4101
4102 f2fs_down_write(&sbi->sb_lock);
4103
4104 spin_lock_irqsave(&sbi->error_lock, flags);
4105 if (sbi->error_dirty) {
4106 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4107 MAX_F2FS_ERRORS);
4108 sbi->error_dirty = false;
4109 }
4110 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4111 spin_unlock_irqrestore(&sbi->error_lock, flags);
4112
4113 err = f2fs_commit_super(sbi, false);
4114
4115 f2fs_up_write(&sbi->sb_lock);
4116 if (err)
4117 f2fs_err_ratelimited(sbi,
4118 "f2fs_commit_super fails to record stop_reason, err:%d",
4119 err);
4120}
4121
4122void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4123{
4124 unsigned long flags;
4125
4126 spin_lock_irqsave(&sbi->error_lock, flags);
4127 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4128 set_bit(flag, (unsigned long *)sbi->errors);
4129 sbi->error_dirty = true;
4130 }
4131 spin_unlock_irqrestore(&sbi->error_lock, flags);
4132}
4133
4134static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4135{
4136 unsigned long flags;
4137 bool need_update = false;
4138
4139 spin_lock_irqsave(&sbi->error_lock, flags);
4140 if (sbi->error_dirty) {
4141 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4142 MAX_F2FS_ERRORS);
4143 sbi->error_dirty = false;
4144 need_update = true;
4145 }
4146 spin_unlock_irqrestore(&sbi->error_lock, flags);
4147
4148 return need_update;
4149}
4150
4151static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4152{
4153 int err;
4154
4155 f2fs_down_write(&sbi->sb_lock);
4156
4157 if (!f2fs_update_errors(sbi))
4158 goto out_unlock;
4159
4160 err = f2fs_commit_super(sbi, false);
4161 if (err)
4162 f2fs_err_ratelimited(sbi,
4163 "f2fs_commit_super fails to record errors:%u, err:%d",
4164 error, err);
4165out_unlock:
4166 f2fs_up_write(&sbi->sb_lock);
4167}
4168
4169void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4170{
4171 f2fs_save_errors(sbi, error);
4172 f2fs_record_errors(sbi, error);
4173}
4174
4175void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4176{
4177 f2fs_save_errors(sbi, error);
4178
4179 if (!sbi->error_dirty)
4180 return;
4181 if (!test_bit(error, (unsigned long *)sbi->errors))
4182 return;
4183 schedule_work(&sbi->s_error_work);
4184}
4185
4186static bool system_going_down(void)
4187{
4188 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4189 || system_state == SYSTEM_RESTART;
4190}
4191
4192void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason)
4193{
4194 struct super_block *sb = sbi->sb;
4195 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4196 bool continue_fs = !shutdown &&
4197 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4198
4199 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4200
4201 if (!f2fs_hw_is_readonly(sbi)) {
4202 save_stop_reason(sbi, reason);
4203
4204 /*
4205 * always create an asynchronous task to record stop_reason
4206 * in order to avoid potential deadlock when running into
4207 * f2fs_record_stop_reason() synchronously.
4208 */
4209 schedule_work(&sbi->s_error_work);
4210 }
4211
4212 /*
4213 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4214 * could panic during 'reboot -f' as the underlying device got already
4215 * disabled.
4216 */
4217 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4218 !shutdown && !system_going_down() &&
4219 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4220 panic("F2FS-fs (device %s): panic forced after error\n",
4221 sb->s_id);
4222
4223 if (shutdown)
4224 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4225
4226 /*
4227 * Continue filesystem operators if errors=continue. Should not set
4228 * RO by shutdown, since RO bypasses thaw_super which can hang the
4229 * system.
4230 */
4231 if (continue_fs || f2fs_readonly(sb) || shutdown) {
4232 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason);
4233 return;
4234 }
4235
4236 f2fs_warn(sbi, "Remounting filesystem read-only");
4237
4238 /*
4239 * We have already set CP_ERROR_FLAG flag to stop all updates
4240 * to filesystem, so it doesn't need to set SB_RDONLY flag here
4241 * because the flag should be set covered w/ sb->s_umount semaphore
4242 * via remount procedure, otherwise, it will confuse code like
4243 * freeze_super() which will lead to deadlocks and other problems.
4244 */
4245}
4246
4247static void f2fs_record_error_work(struct work_struct *work)
4248{
4249 struct f2fs_sb_info *sbi = container_of(work,
4250 struct f2fs_sb_info, s_error_work);
4251
4252 f2fs_record_stop_reason(sbi);
4253}
4254
4255static inline unsigned int get_first_zoned_segno(struct f2fs_sb_info *sbi)
4256{
4257 int devi;
4258
4259 for (devi = 0; devi < sbi->s_ndevs; devi++)
4260 if (bdev_is_zoned(FDEV(devi).bdev))
4261 return GET_SEGNO(sbi, FDEV(devi).start_blk);
4262 return 0;
4263}
4264
4265static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4266{
4267 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4268 unsigned int max_devices = MAX_DEVICES;
4269 unsigned int logical_blksize;
4270 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4271 int i;
4272
4273 /* Initialize single device information */
4274 if (!RDEV(0).path[0]) {
4275 if (!bdev_is_zoned(sbi->sb->s_bdev))
4276 return 0;
4277 max_devices = 1;
4278 }
4279
4280 /*
4281 * Initialize multiple devices information, or single
4282 * zoned block device information.
4283 */
4284 sbi->devs = f2fs_kzalloc(sbi,
4285 array_size(max_devices,
4286 sizeof(struct f2fs_dev_info)),
4287 GFP_KERNEL);
4288 if (!sbi->devs)
4289 return -ENOMEM;
4290
4291 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4292 sbi->aligned_blksize = true;
4293#ifdef CONFIG_BLK_DEV_ZONED
4294 sbi->max_open_zones = UINT_MAX;
4295 sbi->blkzone_alloc_policy = BLKZONE_ALLOC_PRIOR_SEQ;
4296#endif
4297
4298 for (i = 0; i < max_devices; i++) {
4299 if (i == 0)
4300 FDEV(0).bdev_file = sbi->sb->s_bdev_file;
4301 else if (!RDEV(i).path[0])
4302 break;
4303
4304 if (max_devices > 1) {
4305 /* Multi-device mount */
4306 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4307 FDEV(i).total_segments =
4308 le32_to_cpu(RDEV(i).total_segments);
4309 if (i == 0) {
4310 FDEV(i).start_blk = 0;
4311 FDEV(i).end_blk = FDEV(i).start_blk +
4312 SEGS_TO_BLKS(sbi,
4313 FDEV(i).total_segments) - 1 +
4314 le32_to_cpu(raw_super->segment0_blkaddr);
4315 } else {
4316 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4317 FDEV(i).end_blk = FDEV(i).start_blk +
4318 SEGS_TO_BLKS(sbi,
4319 FDEV(i).total_segments) - 1;
4320 FDEV(i).bdev_file = bdev_file_open_by_path(
4321 FDEV(i).path, mode, sbi->sb, NULL);
4322 }
4323 }
4324 if (IS_ERR(FDEV(i).bdev_file))
4325 return PTR_ERR(FDEV(i).bdev_file);
4326
4327 FDEV(i).bdev = file_bdev(FDEV(i).bdev_file);
4328 /* to release errored devices */
4329 sbi->s_ndevs = i + 1;
4330
4331 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4332 sbi->aligned_blksize = false;
4333
4334#ifdef CONFIG_BLK_DEV_ZONED
4335 if (bdev_is_zoned(FDEV(i).bdev)) {
4336 if (!f2fs_sb_has_blkzoned(sbi)) {
4337 f2fs_err(sbi, "Zoned block device feature not enabled");
4338 return -EINVAL;
4339 }
4340 if (init_blkz_info(sbi, i)) {
4341 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4342 return -EINVAL;
4343 }
4344 if (max_devices == 1)
4345 break;
4346 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)",
4347 i, FDEV(i).path,
4348 FDEV(i).total_segments,
4349 FDEV(i).start_blk, FDEV(i).end_blk);
4350 continue;
4351 }
4352#endif
4353 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4354 i, FDEV(i).path,
4355 FDEV(i).total_segments,
4356 FDEV(i).start_blk, FDEV(i).end_blk);
4357 }
4358 return 0;
4359}
4360
4361static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4362{
4363#if IS_ENABLED(CONFIG_UNICODE)
4364 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4365 const struct f2fs_sb_encodings *encoding_info;
4366 struct unicode_map *encoding;
4367 __u16 encoding_flags;
4368
4369 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4370 if (!encoding_info) {
4371 f2fs_err(sbi,
4372 "Encoding requested by superblock is unknown");
4373 return -EINVAL;
4374 }
4375
4376 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4377 encoding = utf8_load(encoding_info->version);
4378 if (IS_ERR(encoding)) {
4379 f2fs_err(sbi,
4380 "can't mount with superblock charset: %s-%u.%u.%u "
4381 "not supported by the kernel. flags: 0x%x.",
4382 encoding_info->name,
4383 unicode_major(encoding_info->version),
4384 unicode_minor(encoding_info->version),
4385 unicode_rev(encoding_info->version),
4386 encoding_flags);
4387 return PTR_ERR(encoding);
4388 }
4389 f2fs_info(sbi, "Using encoding defined by superblock: "
4390 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4391 unicode_major(encoding_info->version),
4392 unicode_minor(encoding_info->version),
4393 unicode_rev(encoding_info->version),
4394 encoding_flags);
4395
4396 sbi->sb->s_encoding = encoding;
4397 sbi->sb->s_encoding_flags = encoding_flags;
4398 }
4399#else
4400 if (f2fs_sb_has_casefold(sbi)) {
4401 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4402 return -EINVAL;
4403 }
4404#endif
4405 return 0;
4406}
4407
4408static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4409{
4410 /* adjust parameters according to the volume size */
4411 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4412 if (f2fs_block_unit_discard(sbi))
4413 SM_I(sbi)->dcc_info->discard_granularity =
4414 MIN_DISCARD_GRANULARITY;
4415 if (!f2fs_lfs_mode(sbi))
4416 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4417 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4418 }
4419
4420 sbi->readdir_ra = true;
4421}
4422
4423static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4424{
4425 struct f2fs_sb_info *sbi;
4426 struct f2fs_super_block *raw_super;
4427 struct inode *root;
4428 int err;
4429 bool skip_recovery = false, need_fsck = false;
4430 char *options = NULL;
4431 int recovery, i, valid_super_block;
4432 struct curseg_info *seg_i;
4433 int retry_cnt = 1;
4434#ifdef CONFIG_QUOTA
4435 bool quota_enabled = false;
4436#endif
4437
4438try_onemore:
4439 err = -EINVAL;
4440 raw_super = NULL;
4441 valid_super_block = -1;
4442 recovery = 0;
4443
4444 /* allocate memory for f2fs-specific super block info */
4445 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4446 if (!sbi)
4447 return -ENOMEM;
4448
4449 sbi->sb = sb;
4450
4451 /* initialize locks within allocated memory */
4452 init_f2fs_rwsem(&sbi->gc_lock);
4453 mutex_init(&sbi->writepages);
4454 init_f2fs_rwsem(&sbi->cp_global_sem);
4455 init_f2fs_rwsem(&sbi->node_write);
4456 init_f2fs_rwsem(&sbi->node_change);
4457 spin_lock_init(&sbi->stat_lock);
4458 init_f2fs_rwsem(&sbi->cp_rwsem);
4459 init_f2fs_rwsem(&sbi->quota_sem);
4460 init_waitqueue_head(&sbi->cp_wait);
4461 spin_lock_init(&sbi->error_lock);
4462
4463 for (i = 0; i < NR_INODE_TYPE; i++) {
4464 INIT_LIST_HEAD(&sbi->inode_list[i]);
4465 spin_lock_init(&sbi->inode_lock[i]);
4466 }
4467 mutex_init(&sbi->flush_lock);
4468
4469 /* Load the checksum driver */
4470 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4471 if (IS_ERR(sbi->s_chksum_driver)) {
4472 f2fs_err(sbi, "Cannot load crc32 driver.");
4473 err = PTR_ERR(sbi->s_chksum_driver);
4474 sbi->s_chksum_driver = NULL;
4475 goto free_sbi;
4476 }
4477
4478 /* set a block size */
4479 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4480 f2fs_err(sbi, "unable to set blocksize");
4481 goto free_sbi;
4482 }
4483
4484 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4485 &recovery);
4486 if (err)
4487 goto free_sbi;
4488
4489 sb->s_fs_info = sbi;
4490 sbi->raw_super = raw_super;
4491
4492 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4493 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4494 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4495
4496 /* precompute checksum seed for metadata */
4497 if (f2fs_sb_has_inode_chksum(sbi))
4498 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4499 sizeof(raw_super->uuid));
4500
4501 default_options(sbi, false);
4502 /* parse mount options */
4503 options = kstrdup((const char *)data, GFP_KERNEL);
4504 if (data && !options) {
4505 err = -ENOMEM;
4506 goto free_sb_buf;
4507 }
4508
4509 err = parse_options(sb, options, false);
4510 if (err)
4511 goto free_options;
4512
4513 sb->s_maxbytes = max_file_blocks(NULL) <<
4514 le32_to_cpu(raw_super->log_blocksize);
4515 sb->s_max_links = F2FS_LINK_MAX;
4516
4517 err = f2fs_setup_casefold(sbi);
4518 if (err)
4519 goto free_options;
4520
4521#ifdef CONFIG_QUOTA
4522 sb->dq_op = &f2fs_quota_operations;
4523 sb->s_qcop = &f2fs_quotactl_ops;
4524 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4525
4526 if (f2fs_sb_has_quota_ino(sbi)) {
4527 for (i = 0; i < MAXQUOTAS; i++) {
4528 if (f2fs_qf_ino(sbi->sb, i))
4529 sbi->nquota_files++;
4530 }
4531 }
4532#endif
4533
4534 sb->s_op = &f2fs_sops;
4535#ifdef CONFIG_FS_ENCRYPTION
4536 sb->s_cop = &f2fs_cryptops;
4537#endif
4538#ifdef CONFIG_FS_VERITY
4539 sb->s_vop = &f2fs_verityops;
4540#endif
4541 sb->s_xattr = f2fs_xattr_handlers;
4542 sb->s_export_op = &f2fs_export_ops;
4543 sb->s_magic = F2FS_SUPER_MAGIC;
4544 sb->s_time_gran = 1;
4545 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4546 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4547 super_set_uuid(sb, (void *) raw_super->uuid, sizeof(raw_super->uuid));
4548 super_set_sysfs_name_bdev(sb);
4549 sb->s_iflags |= SB_I_CGROUPWB;
4550
4551 /* init f2fs-specific super block info */
4552 sbi->valid_super_block = valid_super_block;
4553
4554 /* disallow all the data/node/meta page writes */
4555 set_sbi_flag(sbi, SBI_POR_DOING);
4556
4557 err = f2fs_init_write_merge_io(sbi);
4558 if (err)
4559 goto free_bio_info;
4560
4561 init_sb_info(sbi);
4562
4563 err = f2fs_init_iostat(sbi);
4564 if (err)
4565 goto free_bio_info;
4566
4567 err = init_percpu_info(sbi);
4568 if (err)
4569 goto free_iostat;
4570
4571 /* init per sbi slab cache */
4572 err = f2fs_init_xattr_caches(sbi);
4573 if (err)
4574 goto free_percpu;
4575 err = f2fs_init_page_array_cache(sbi);
4576 if (err)
4577 goto free_xattr_cache;
4578
4579 /* get an inode for meta space */
4580 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4581 if (IS_ERR(sbi->meta_inode)) {
4582 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4583 err = PTR_ERR(sbi->meta_inode);
4584 goto free_page_array_cache;
4585 }
4586
4587 err = f2fs_get_valid_checkpoint(sbi);
4588 if (err) {
4589 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4590 goto free_meta_inode;
4591 }
4592
4593 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4594 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4595 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4596 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4597 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4598 }
4599
4600 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4601 set_sbi_flag(sbi, SBI_NEED_FSCK);
4602
4603 /* Initialize device list */
4604 err = f2fs_scan_devices(sbi);
4605 if (err) {
4606 f2fs_err(sbi, "Failed to find devices");
4607 goto free_devices;
4608 }
4609
4610 err = f2fs_init_post_read_wq(sbi);
4611 if (err) {
4612 f2fs_err(sbi, "Failed to initialize post read workqueue");
4613 goto free_devices;
4614 }
4615
4616 sbi->total_valid_node_count =
4617 le32_to_cpu(sbi->ckpt->valid_node_count);
4618 percpu_counter_set(&sbi->total_valid_inode_count,
4619 le32_to_cpu(sbi->ckpt->valid_inode_count));
4620 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4621 sbi->total_valid_block_count =
4622 le64_to_cpu(sbi->ckpt->valid_block_count);
4623 sbi->last_valid_block_count = sbi->total_valid_block_count;
4624 sbi->reserved_blocks = 0;
4625 sbi->current_reserved_blocks = 0;
4626 limit_reserve_root(sbi);
4627 adjust_unusable_cap_perc(sbi);
4628
4629 f2fs_init_extent_cache_info(sbi);
4630
4631 f2fs_init_ino_entry_info(sbi);
4632
4633 f2fs_init_fsync_node_info(sbi);
4634
4635 /* setup checkpoint request control and start checkpoint issue thread */
4636 f2fs_init_ckpt_req_control(sbi);
4637 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4638 test_opt(sbi, MERGE_CHECKPOINT)) {
4639 err = f2fs_start_ckpt_thread(sbi);
4640 if (err) {
4641 f2fs_err(sbi,
4642 "Failed to start F2FS issue_checkpoint_thread (%d)",
4643 err);
4644 goto stop_ckpt_thread;
4645 }
4646 }
4647
4648 /* setup f2fs internal modules */
4649 err = f2fs_build_segment_manager(sbi);
4650 if (err) {
4651 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4652 err);
4653 goto free_sm;
4654 }
4655 err = f2fs_build_node_manager(sbi);
4656 if (err) {
4657 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4658 err);
4659 goto free_nm;
4660 }
4661
4662 /* For write statistics */
4663 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4664
4665 /* get segno of first zoned block device */
4666 sbi->first_zoned_segno = get_first_zoned_segno(sbi);
4667
4668 /* Read accumulated write IO statistics if exists */
4669 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4670 if (__exist_node_summaries(sbi))
4671 sbi->kbytes_written =
4672 le64_to_cpu(seg_i->journal->info.kbytes_written);
4673
4674 f2fs_build_gc_manager(sbi);
4675
4676 err = f2fs_build_stats(sbi);
4677 if (err)
4678 goto free_nm;
4679
4680 /* get an inode for node space */
4681 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4682 if (IS_ERR(sbi->node_inode)) {
4683 f2fs_err(sbi, "Failed to read node inode");
4684 err = PTR_ERR(sbi->node_inode);
4685 goto free_stats;
4686 }
4687
4688 /* read root inode and dentry */
4689 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4690 if (IS_ERR(root)) {
4691 f2fs_err(sbi, "Failed to read root inode");
4692 err = PTR_ERR(root);
4693 goto free_node_inode;
4694 }
4695 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4696 !root->i_size || !root->i_nlink) {
4697 iput(root);
4698 err = -EINVAL;
4699 goto free_node_inode;
4700 }
4701
4702 generic_set_sb_d_ops(sb);
4703 sb->s_root = d_make_root(root); /* allocate root dentry */
4704 if (!sb->s_root) {
4705 err = -ENOMEM;
4706 goto free_node_inode;
4707 }
4708
4709 err = f2fs_init_compress_inode(sbi);
4710 if (err)
4711 goto free_root_inode;
4712
4713 err = f2fs_register_sysfs(sbi);
4714 if (err)
4715 goto free_compress_inode;
4716
4717#ifdef CONFIG_QUOTA
4718 /* Enable quota usage during mount */
4719 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4720 err = f2fs_enable_quotas(sb);
4721 if (err)
4722 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4723 }
4724
4725 quota_enabled = f2fs_recover_quota_begin(sbi);
4726#endif
4727 /* if there are any orphan inodes, free them */
4728 err = f2fs_recover_orphan_inodes(sbi);
4729 if (err)
4730 goto free_meta;
4731
4732 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4733 goto reset_checkpoint;
4734
4735 /* recover fsynced data */
4736 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4737 !test_opt(sbi, NORECOVERY)) {
4738 /*
4739 * mount should be failed, when device has readonly mode, and
4740 * previous checkpoint was not done by clean system shutdown.
4741 */
4742 if (f2fs_hw_is_readonly(sbi)) {
4743 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4744 err = f2fs_recover_fsync_data(sbi, true);
4745 if (err > 0) {
4746 err = -EROFS;
4747 f2fs_err(sbi, "Need to recover fsync data, but "
4748 "write access unavailable, please try "
4749 "mount w/ disable_roll_forward or norecovery");
4750 }
4751 if (err < 0)
4752 goto free_meta;
4753 }
4754 f2fs_info(sbi, "write access unavailable, skipping recovery");
4755 goto reset_checkpoint;
4756 }
4757
4758 if (need_fsck)
4759 set_sbi_flag(sbi, SBI_NEED_FSCK);
4760
4761 if (skip_recovery)
4762 goto reset_checkpoint;
4763
4764 err = f2fs_recover_fsync_data(sbi, false);
4765 if (err < 0) {
4766 if (err != -ENOMEM)
4767 skip_recovery = true;
4768 need_fsck = true;
4769 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4770 err);
4771 goto free_meta;
4772 }
4773 } else {
4774 err = f2fs_recover_fsync_data(sbi, true);
4775
4776 if (!f2fs_readonly(sb) && err > 0) {
4777 err = -EINVAL;
4778 f2fs_err(sbi, "Need to recover fsync data");
4779 goto free_meta;
4780 }
4781 }
4782
4783#ifdef CONFIG_QUOTA
4784 f2fs_recover_quota_end(sbi, quota_enabled);
4785#endif
4786reset_checkpoint:
4787 /*
4788 * If the f2fs is not readonly and fsync data recovery succeeds,
4789 * write pointer consistency of cursegs and other zones are already
4790 * checked and fixed during recovery. However, if recovery fails,
4791 * write pointers are left untouched, and retry-mount should check
4792 * them here.
4793 */
4794 if (skip_recovery)
4795 err = f2fs_check_and_fix_write_pointer(sbi);
4796 if (err)
4797 goto free_meta;
4798
4799 /* f2fs_recover_fsync_data() cleared this already */
4800 clear_sbi_flag(sbi, SBI_POR_DOING);
4801
4802 err = f2fs_init_inmem_curseg(sbi);
4803 if (err)
4804 goto sync_free_meta;
4805
4806 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4807 err = f2fs_disable_checkpoint(sbi);
4808 if (err)
4809 goto sync_free_meta;
4810 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4811 f2fs_enable_checkpoint(sbi);
4812 }
4813
4814 /*
4815 * If filesystem is not mounted as read-only then
4816 * do start the gc_thread.
4817 */
4818 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4819 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4820 /* After POR, we can run background GC thread.*/
4821 err = f2fs_start_gc_thread(sbi);
4822 if (err)
4823 goto sync_free_meta;
4824 }
4825 kvfree(options);
4826
4827 /* recover broken superblock */
4828 if (recovery) {
4829 err = f2fs_commit_super(sbi, true);
4830 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4831 sbi->valid_super_block ? 1 : 2, err);
4832 }
4833
4834 f2fs_join_shrinker(sbi);
4835
4836 f2fs_tuning_parameters(sbi);
4837
4838 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4839 cur_cp_version(F2FS_CKPT(sbi)));
4840 f2fs_update_time(sbi, CP_TIME);
4841 f2fs_update_time(sbi, REQ_TIME);
4842 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4843 return 0;
4844
4845sync_free_meta:
4846 /* safe to flush all the data */
4847 sync_filesystem(sbi->sb);
4848 retry_cnt = 0;
4849
4850free_meta:
4851#ifdef CONFIG_QUOTA
4852 f2fs_truncate_quota_inode_pages(sb);
4853 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4854 f2fs_quota_off_umount(sbi->sb);
4855#endif
4856 /*
4857 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4858 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4859 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4860 * falls into an infinite loop in f2fs_sync_meta_pages().
4861 */
4862 truncate_inode_pages_final(META_MAPPING(sbi));
4863 /* evict some inodes being cached by GC */
4864 evict_inodes(sb);
4865 f2fs_unregister_sysfs(sbi);
4866free_compress_inode:
4867 f2fs_destroy_compress_inode(sbi);
4868free_root_inode:
4869 dput(sb->s_root);
4870 sb->s_root = NULL;
4871free_node_inode:
4872 f2fs_release_ino_entry(sbi, true);
4873 truncate_inode_pages_final(NODE_MAPPING(sbi));
4874 iput(sbi->node_inode);
4875 sbi->node_inode = NULL;
4876free_stats:
4877 f2fs_destroy_stats(sbi);
4878free_nm:
4879 /* stop discard thread before destroying node manager */
4880 f2fs_stop_discard_thread(sbi);
4881 f2fs_destroy_node_manager(sbi);
4882free_sm:
4883 f2fs_destroy_segment_manager(sbi);
4884stop_ckpt_thread:
4885 f2fs_stop_ckpt_thread(sbi);
4886 /* flush s_error_work before sbi destroy */
4887 flush_work(&sbi->s_error_work);
4888 f2fs_destroy_post_read_wq(sbi);
4889free_devices:
4890 destroy_device_list(sbi);
4891 kvfree(sbi->ckpt);
4892free_meta_inode:
4893 make_bad_inode(sbi->meta_inode);
4894 iput(sbi->meta_inode);
4895 sbi->meta_inode = NULL;
4896free_page_array_cache:
4897 f2fs_destroy_page_array_cache(sbi);
4898free_xattr_cache:
4899 f2fs_destroy_xattr_caches(sbi);
4900free_percpu:
4901 destroy_percpu_info(sbi);
4902free_iostat:
4903 f2fs_destroy_iostat(sbi);
4904free_bio_info:
4905 for (i = 0; i < NR_PAGE_TYPE; i++)
4906 kvfree(sbi->write_io[i]);
4907
4908#if IS_ENABLED(CONFIG_UNICODE)
4909 utf8_unload(sb->s_encoding);
4910 sb->s_encoding = NULL;
4911#endif
4912free_options:
4913#ifdef CONFIG_QUOTA
4914 for (i = 0; i < MAXQUOTAS; i++)
4915 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4916#endif
4917 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4918 kvfree(options);
4919free_sb_buf:
4920 kfree(raw_super);
4921free_sbi:
4922 if (sbi->s_chksum_driver)
4923 crypto_free_shash(sbi->s_chksum_driver);
4924 kfree(sbi);
4925 sb->s_fs_info = NULL;
4926
4927 /* give only one another chance */
4928 if (retry_cnt > 0 && skip_recovery) {
4929 retry_cnt--;
4930 shrink_dcache_sb(sb);
4931 goto try_onemore;
4932 }
4933 return err;
4934}
4935
4936static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4937 const char *dev_name, void *data)
4938{
4939 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4940}
4941
4942static void kill_f2fs_super(struct super_block *sb)
4943{
4944 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4945
4946 if (sb->s_root) {
4947 set_sbi_flag(sbi, SBI_IS_CLOSE);
4948 f2fs_stop_gc_thread(sbi);
4949 f2fs_stop_discard_thread(sbi);
4950
4951#ifdef CONFIG_F2FS_FS_COMPRESSION
4952 /*
4953 * latter evict_inode() can bypass checking and invalidating
4954 * compress inode cache.
4955 */
4956 if (test_opt(sbi, COMPRESS_CACHE))
4957 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4958#endif
4959
4960 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4961 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4962 struct cp_control cpc = {
4963 .reason = CP_UMOUNT,
4964 };
4965 stat_inc_cp_call_count(sbi, TOTAL_CALL);
4966 f2fs_write_checkpoint(sbi, &cpc);
4967 }
4968
4969 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4970 sb->s_flags &= ~SB_RDONLY;
4971 }
4972 kill_block_super(sb);
4973 /* Release block devices last, after fscrypt_destroy_keyring(). */
4974 if (sbi) {
4975 destroy_device_list(sbi);
4976 kfree(sbi);
4977 sb->s_fs_info = NULL;
4978 }
4979}
4980
4981static struct file_system_type f2fs_fs_type = {
4982 .owner = THIS_MODULE,
4983 .name = "f2fs",
4984 .mount = f2fs_mount,
4985 .kill_sb = kill_f2fs_super,
4986 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4987};
4988MODULE_ALIAS_FS("f2fs");
4989
4990static int __init init_inodecache(void)
4991{
4992 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4993 sizeof(struct f2fs_inode_info), 0,
4994 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4995 return f2fs_inode_cachep ? 0 : -ENOMEM;
4996}
4997
4998static void destroy_inodecache(void)
4999{
5000 /*
5001 * Make sure all delayed rcu free inodes are flushed before we
5002 * destroy cache.
5003 */
5004 rcu_barrier();
5005 kmem_cache_destroy(f2fs_inode_cachep);
5006}
5007
5008static int __init init_f2fs_fs(void)
5009{
5010 int err;
5011
5012 err = init_inodecache();
5013 if (err)
5014 goto fail;
5015 err = f2fs_create_node_manager_caches();
5016 if (err)
5017 goto free_inodecache;
5018 err = f2fs_create_segment_manager_caches();
5019 if (err)
5020 goto free_node_manager_caches;
5021 err = f2fs_create_checkpoint_caches();
5022 if (err)
5023 goto free_segment_manager_caches;
5024 err = f2fs_create_recovery_cache();
5025 if (err)
5026 goto free_checkpoint_caches;
5027 err = f2fs_create_extent_cache();
5028 if (err)
5029 goto free_recovery_cache;
5030 err = f2fs_create_garbage_collection_cache();
5031 if (err)
5032 goto free_extent_cache;
5033 err = f2fs_init_sysfs();
5034 if (err)
5035 goto free_garbage_collection_cache;
5036 err = f2fs_init_shrinker();
5037 if (err)
5038 goto free_sysfs;
5039 f2fs_create_root_stats();
5040 err = f2fs_init_post_read_processing();
5041 if (err)
5042 goto free_root_stats;
5043 err = f2fs_init_iostat_processing();
5044 if (err)
5045 goto free_post_read;
5046 err = f2fs_init_bio_entry_cache();
5047 if (err)
5048 goto free_iostat;
5049 err = f2fs_init_bioset();
5050 if (err)
5051 goto free_bio_entry_cache;
5052 err = f2fs_init_compress_mempool();
5053 if (err)
5054 goto free_bioset;
5055 err = f2fs_init_compress_cache();
5056 if (err)
5057 goto free_compress_mempool;
5058 err = f2fs_create_casefold_cache();
5059 if (err)
5060 goto free_compress_cache;
5061 err = register_filesystem(&f2fs_fs_type);
5062 if (err)
5063 goto free_casefold_cache;
5064 return 0;
5065free_casefold_cache:
5066 f2fs_destroy_casefold_cache();
5067free_compress_cache:
5068 f2fs_destroy_compress_cache();
5069free_compress_mempool:
5070 f2fs_destroy_compress_mempool();
5071free_bioset:
5072 f2fs_destroy_bioset();
5073free_bio_entry_cache:
5074 f2fs_destroy_bio_entry_cache();
5075free_iostat:
5076 f2fs_destroy_iostat_processing();
5077free_post_read:
5078 f2fs_destroy_post_read_processing();
5079free_root_stats:
5080 f2fs_destroy_root_stats();
5081 f2fs_exit_shrinker();
5082free_sysfs:
5083 f2fs_exit_sysfs();
5084free_garbage_collection_cache:
5085 f2fs_destroy_garbage_collection_cache();
5086free_extent_cache:
5087 f2fs_destroy_extent_cache();
5088free_recovery_cache:
5089 f2fs_destroy_recovery_cache();
5090free_checkpoint_caches:
5091 f2fs_destroy_checkpoint_caches();
5092free_segment_manager_caches:
5093 f2fs_destroy_segment_manager_caches();
5094free_node_manager_caches:
5095 f2fs_destroy_node_manager_caches();
5096free_inodecache:
5097 destroy_inodecache();
5098fail:
5099 return err;
5100}
5101
5102static void __exit exit_f2fs_fs(void)
5103{
5104 unregister_filesystem(&f2fs_fs_type);
5105 f2fs_destroy_casefold_cache();
5106 f2fs_destroy_compress_cache();
5107 f2fs_destroy_compress_mempool();
5108 f2fs_destroy_bioset();
5109 f2fs_destroy_bio_entry_cache();
5110 f2fs_destroy_iostat_processing();
5111 f2fs_destroy_post_read_processing();
5112 f2fs_destroy_root_stats();
5113 f2fs_exit_shrinker();
5114 f2fs_exit_sysfs();
5115 f2fs_destroy_garbage_collection_cache();
5116 f2fs_destroy_extent_cache();
5117 f2fs_destroy_recovery_cache();
5118 f2fs_destroy_checkpoint_caches();
5119 f2fs_destroy_segment_manager_caches();
5120 f2fs_destroy_node_manager_caches();
5121 destroy_inodecache();
5122}
5123
5124module_init(init_f2fs_fs)
5125module_exit(exit_f2fs_fs)
5126
5127MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5128MODULE_DESCRIPTION("Flash Friendly File System");
5129MODULE_LICENSE("GPL");
5130MODULE_SOFTDEP("pre: crc32");
5131