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/buffer_head.h>
15 #include <linux/kthread.h>
16 #include <linux/parser.h>
17 #include <linux/mount.h>
18 #include <linux/seq_file.h>
19 #include <linux/proc_fs.h>
20 #include <linux/random.h>
21 #include <linux/exportfs.h>
22 #include <linux/blkdev.h>
23 #include <linux/quotaops.h>
24 #include <linux/f2fs_fs.h>
25 #include <linux/sysfs.h>
26 #include <linux/quota.h>
27 #include <linux/unicode.h>
28 #include <linux/part_stat.h>
29 #include <linux/zstd.h>
30 #include <linux/lz4.h>
31
32 #include "f2fs.h"
33 #include "node.h"
34 #include "segment.h"
35 #include "xattr.h"
36 #include "gc.h"
37 #include "iostat.h"
38
39 #define CREATE_TRACE_POINTS
40 #include <trace/events/f2fs.h>
41
42 static struct kmem_cache *f2fs_inode_cachep;
43
44 #ifdef CONFIG_F2FS_FAULT_INJECTION
45
46 const char *f2fs_fault_name[FAULT_MAX] = {
47 [FAULT_KMALLOC] = "kmalloc",
48 [FAULT_KVMALLOC] = "kvmalloc",
49 [FAULT_PAGE_ALLOC] = "page alloc",
50 [FAULT_PAGE_GET] = "page get",
51 [FAULT_ALLOC_NID] = "alloc nid",
52 [FAULT_ORPHAN] = "orphan",
53 [FAULT_BLOCK] = "no more block",
54 [FAULT_DIR_DEPTH] = "too big dir depth",
55 [FAULT_EVICT_INODE] = "evict_inode fail",
56 [FAULT_TRUNCATE] = "truncate fail",
57 [FAULT_READ_IO] = "read IO error",
58 [FAULT_CHECKPOINT] = "checkpoint error",
59 [FAULT_DISCARD] = "discard error",
60 [FAULT_WRITE_IO] = "write IO error",
61 [FAULT_SLAB_ALLOC] = "slab alloc",
62 [FAULT_DQUOT_INIT] = "dquot initialize",
63 [FAULT_LOCK_OP] = "lock_op",
64 [FAULT_BLKADDR_VALIDITY] = "invalid blkaddr",
65 [FAULT_BLKADDR_CONSISTENCE] = "inconsistent blkaddr",
66 [FAULT_NO_SEGMENT] = "no free segment",
67 };
68
69 int f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned long rate,
70 unsigned long type)
71 {
72 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
73
74 if (rate) {
75 if (rate > INT_MAX)
76 return -EINVAL;
77 atomic_set(&ffi->inject_ops, 0);
78 ffi->inject_rate = (int)rate;
79 }
80
81 if (type) {
82 if (type >= BIT(FAULT_MAX))
83 return -EINVAL;
84 ffi->inject_type = (unsigned int)type;
85 }
86
87 if (!rate && !type)
88 memset(ffi, 0, sizeof(struct f2fs_fault_info));
89 else
90 f2fs_info(sbi,
91 "build fault injection attr: rate: %lu, type: 0x%lx",
92 rate, type);
93 return 0;
94 }
95 #endif
96
97 /* f2fs-wide shrinker description */
98 static struct shrinker *f2fs_shrinker_info;
99
100 static int __init f2fs_init_shrinker(void)
101 {
102 f2fs_shrinker_info = shrinker_alloc(0, "f2fs-shrinker");
103 if (!f2fs_shrinker_info)
104 return -ENOMEM;
105
106 f2fs_shrinker_info->count_objects = f2fs_shrink_count;
107 f2fs_shrinker_info->scan_objects = f2fs_shrink_scan;
108
109 shrinker_register(f2fs_shrinker_info);
110
111 return 0;
112 }
113
114 static void f2fs_exit_shrinker(void)
115 {
116 shrinker_free(f2fs_shrinker_info);
117 }
118
119 enum {
120 Opt_gc_background,
121 Opt_disable_roll_forward,
122 Opt_norecovery,
123 Opt_discard,
124 Opt_nodiscard,
125 Opt_noheap,
126 Opt_heap,
127 Opt_user_xattr,
128 Opt_nouser_xattr,
129 Opt_acl,
130 Opt_noacl,
131 Opt_active_logs,
132 Opt_disable_ext_identify,
133 Opt_inline_xattr,
134 Opt_noinline_xattr,
135 Opt_inline_xattr_size,
136 Opt_inline_data,
137 Opt_inline_dentry,
138 Opt_noinline_dentry,
139 Opt_flush_merge,
140 Opt_noflush_merge,
141 Opt_barrier,
142 Opt_nobarrier,
143 Opt_fastboot,
144 Opt_extent_cache,
145 Opt_noextent_cache,
146 Opt_noinline_data,
147 Opt_data_flush,
148 Opt_reserve_root,
149 Opt_resgid,
150 Opt_resuid,
151 Opt_mode,
152 Opt_fault_injection,
153 Opt_fault_type,
154 Opt_quota,
155 Opt_noquota,
156 Opt_usrquota,
157 Opt_grpquota,
158 Opt_prjquota,
159 Opt_usrjquota,
160 Opt_grpjquota,
161 Opt_prjjquota,
162 Opt_offusrjquota,
163 Opt_offgrpjquota,
164 Opt_offprjjquota,
165 Opt_jqfmt_vfsold,
166 Opt_jqfmt_vfsv0,
167 Opt_jqfmt_vfsv1,
168 Opt_alloc,
169 Opt_fsync,
170 Opt_test_dummy_encryption,
171 Opt_inlinecrypt,
172 Opt_checkpoint_disable,
173 Opt_checkpoint_disable_cap,
174 Opt_checkpoint_disable_cap_perc,
175 Opt_checkpoint_enable,
176 Opt_checkpoint_merge,
177 Opt_nocheckpoint_merge,
178 Opt_compress_algorithm,
179 Opt_compress_log_size,
180 Opt_compress_extension,
181 Opt_nocompress_extension,
182 Opt_compress_chksum,
183 Opt_compress_mode,
184 Opt_compress_cache,
185 Opt_atgc,
186 Opt_gc_merge,
187 Opt_nogc_merge,
188 Opt_discard_unit,
189 Opt_memory_mode,
190 Opt_age_extent_cache,
191 Opt_errors,
192 Opt_err,
193 };
194
195 static match_table_t f2fs_tokens = {
196 {Opt_gc_background, "background_gc=%s"},
197 {Opt_disable_roll_forward, "disable_roll_forward"},
198 {Opt_norecovery, "norecovery"},
199 {Opt_discard, "discard"},
200 {Opt_nodiscard, "nodiscard"},
201 {Opt_noheap, "no_heap"},
202 {Opt_heap, "heap"},
203 {Opt_user_xattr, "user_xattr"},
204 {Opt_nouser_xattr, "nouser_xattr"},
205 {Opt_acl, "acl"},
206 {Opt_noacl, "noacl"},
207 {Opt_active_logs, "active_logs=%u"},
208 {Opt_disable_ext_identify, "disable_ext_identify"},
209 {Opt_inline_xattr, "inline_xattr"},
210 {Opt_noinline_xattr, "noinline_xattr"},
211 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
212 {Opt_inline_data, "inline_data"},
213 {Opt_inline_dentry, "inline_dentry"},
214 {Opt_noinline_dentry, "noinline_dentry"},
215 {Opt_flush_merge, "flush_merge"},
216 {Opt_noflush_merge, "noflush_merge"},
217 {Opt_barrier, "barrier"},
218 {Opt_nobarrier, "nobarrier"},
219 {Opt_fastboot, "fastboot"},
220 {Opt_extent_cache, "extent_cache"},
221 {Opt_noextent_cache, "noextent_cache"},
222 {Opt_noinline_data, "noinline_data"},
223 {Opt_data_flush, "data_flush"},
224 {Opt_reserve_root, "reserve_root=%u"},
225 {Opt_resgid, "resgid=%u"},
226 {Opt_resuid, "resuid=%u"},
227 {Opt_mode, "mode=%s"},
228 {Opt_fault_injection, "fault_injection=%u"},
229 {Opt_fault_type, "fault_type=%u"},
230 {Opt_quota, "quota"},
231 {Opt_noquota, "noquota"},
232 {Opt_usrquota, "usrquota"},
233 {Opt_grpquota, "grpquota"},
234 {Opt_prjquota, "prjquota"},
235 {Opt_usrjquota, "usrjquota=%s"},
236 {Opt_grpjquota, "grpjquota=%s"},
237 {Opt_prjjquota, "prjjquota=%s"},
238 {Opt_offusrjquota, "usrjquota="},
239 {Opt_offgrpjquota, "grpjquota="},
240 {Opt_offprjjquota, "prjjquota="},
241 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
242 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
243 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
244 {Opt_alloc, "alloc_mode=%s"},
245 {Opt_fsync, "fsync_mode=%s"},
246 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
247 {Opt_test_dummy_encryption, "test_dummy_encryption"},
248 {Opt_inlinecrypt, "inlinecrypt"},
249 {Opt_checkpoint_disable, "checkpoint=disable"},
250 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
251 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
252 {Opt_checkpoint_enable, "checkpoint=enable"},
253 {Opt_checkpoint_merge, "checkpoint_merge"},
254 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
255 {Opt_compress_algorithm, "compress_algorithm=%s"},
256 {Opt_compress_log_size, "compress_log_size=%u"},
257 {Opt_compress_extension, "compress_extension=%s"},
258 {Opt_nocompress_extension, "nocompress_extension=%s"},
259 {Opt_compress_chksum, "compress_chksum"},
260 {Opt_compress_mode, "compress_mode=%s"},
261 {Opt_compress_cache, "compress_cache"},
262 {Opt_atgc, "atgc"},
263 {Opt_gc_merge, "gc_merge"},
264 {Opt_nogc_merge, "nogc_merge"},
265 {Opt_discard_unit, "discard_unit=%s"},
266 {Opt_memory_mode, "memory=%s"},
267 {Opt_age_extent_cache, "age_extent_cache"},
268 {Opt_errors, "errors=%s"},
269 {Opt_err, NULL},
270 };
271
272 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate,
273 const char *fmt, ...)
274 {
275 struct va_format vaf;
276 va_list args;
277 int level;
278
279 va_start(args, fmt);
280
281 level = printk_get_level(fmt);
282 vaf.fmt = printk_skip_level(fmt);
283 vaf.va = &args;
284 if (limit_rate)
285 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n",
286 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
287 else
288 printk("%c%cF2FS-fs (%s): %pV\n",
289 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
290
291 va_end(args);
292 }
293
294 #if IS_ENABLED(CONFIG_UNICODE)
295 static const struct f2fs_sb_encodings {
296 __u16 magic;
297 char *name;
298 unsigned int version;
299 } f2fs_sb_encoding_map[] = {
300 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
301 };
302
303 static const struct f2fs_sb_encodings *
304 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
305 {
306 __u16 magic = le16_to_cpu(sb->s_encoding);
307 int i;
308
309 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
310 if (magic == f2fs_sb_encoding_map[i].magic)
311 return &f2fs_sb_encoding_map[i];
312
313 return NULL;
314 }
315
316 struct kmem_cache *f2fs_cf_name_slab;
317 static int __init f2fs_create_casefold_cache(void)
318 {
319 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
320 F2FS_NAME_LEN);
321 return f2fs_cf_name_slab ? 0 : -ENOMEM;
322 }
323
324 static void f2fs_destroy_casefold_cache(void)
325 {
326 kmem_cache_destroy(f2fs_cf_name_slab);
327 }
328 #else
329 static int __init f2fs_create_casefold_cache(void) { return 0; }
330 static void f2fs_destroy_casefold_cache(void) { }
331 #endif
332
333 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
334 {
335 block_t limit = min((sbi->user_block_count >> 3),
336 sbi->user_block_count - sbi->reserved_blocks);
337
338 /* limit is 12.5% */
339 if (test_opt(sbi, RESERVE_ROOT) &&
340 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
341 F2FS_OPTION(sbi).root_reserved_blocks = limit;
342 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
343 F2FS_OPTION(sbi).root_reserved_blocks);
344 }
345 if (!test_opt(sbi, RESERVE_ROOT) &&
346 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
347 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
348 !gid_eq(F2FS_OPTION(sbi).s_resgid,
349 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
350 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
351 from_kuid_munged(&init_user_ns,
352 F2FS_OPTION(sbi).s_resuid),
353 from_kgid_munged(&init_user_ns,
354 F2FS_OPTION(sbi).s_resgid));
355 }
356
357 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
358 {
359 if (!F2FS_OPTION(sbi).unusable_cap_perc)
360 return;
361
362 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
363 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
364 else
365 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
366 F2FS_OPTION(sbi).unusable_cap_perc;
367
368 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
369 F2FS_OPTION(sbi).unusable_cap,
370 F2FS_OPTION(sbi).unusable_cap_perc);
371 }
372
373 static void init_once(void *foo)
374 {
375 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
376
377 inode_init_once(&fi->vfs_inode);
378 }
379
380 #ifdef CONFIG_QUOTA
381 static const char * const quotatypes[] = INITQFNAMES;
382 #define QTYPE2NAME(t) (quotatypes[t])
383 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
384 substring_t *args)
385 {
386 struct f2fs_sb_info *sbi = F2FS_SB(sb);
387 char *qname;
388 int ret = -EINVAL;
389
390 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
391 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
392 return -EINVAL;
393 }
394 if (f2fs_sb_has_quota_ino(sbi)) {
395 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
396 return 0;
397 }
398
399 qname = match_strdup(args);
400 if (!qname) {
401 f2fs_err(sbi, "Not enough memory for storing quotafile name");
402 return -ENOMEM;
403 }
404 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
405 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
406 ret = 0;
407 else
408 f2fs_err(sbi, "%s quota file already specified",
409 QTYPE2NAME(qtype));
410 goto errout;
411 }
412 if (strchr(qname, '/')) {
413 f2fs_err(sbi, "quotafile must be on filesystem root");
414 goto errout;
415 }
416 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
417 set_opt(sbi, QUOTA);
418 return 0;
419 errout:
420 kfree(qname);
421 return ret;
422 }
423
424 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
425 {
426 struct f2fs_sb_info *sbi = F2FS_SB(sb);
427
428 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
429 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
430 return -EINVAL;
431 }
432 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
433 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
434 return 0;
435 }
436
437 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
438 {
439 /*
440 * We do the test below only for project quotas. 'usrquota' and
441 * 'grpquota' mount options are allowed even without quota feature
442 * to support legacy quotas in quota files.
443 */
444 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
445 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
446 return -1;
447 }
448 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
449 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
450 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
451 if (test_opt(sbi, USRQUOTA) &&
452 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
453 clear_opt(sbi, USRQUOTA);
454
455 if (test_opt(sbi, GRPQUOTA) &&
456 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
457 clear_opt(sbi, GRPQUOTA);
458
459 if (test_opt(sbi, PRJQUOTA) &&
460 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
461 clear_opt(sbi, PRJQUOTA);
462
463 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
464 test_opt(sbi, PRJQUOTA)) {
465 f2fs_err(sbi, "old and new quota format mixing");
466 return -1;
467 }
468
469 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
470 f2fs_err(sbi, "journaled quota format not specified");
471 return -1;
472 }
473 }
474
475 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
476 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
477 F2FS_OPTION(sbi).s_jquota_fmt = 0;
478 }
479 return 0;
480 }
481 #endif
482
483 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
484 const char *opt,
485 const substring_t *arg,
486 bool is_remount)
487 {
488 struct f2fs_sb_info *sbi = F2FS_SB(sb);
489 struct fs_parameter param = {
490 .type = fs_value_is_string,
491 .string = arg->from ? arg->from : "",
492 };
493 struct fscrypt_dummy_policy *policy =
494 &F2FS_OPTION(sbi).dummy_enc_policy;
495 int err;
496
497 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
498 f2fs_warn(sbi, "test_dummy_encryption option not supported");
499 return -EINVAL;
500 }
501
502 if (!f2fs_sb_has_encrypt(sbi)) {
503 f2fs_err(sbi, "Encrypt feature is off");
504 return -EINVAL;
505 }
506
507 /*
508 * This mount option is just for testing, and it's not worthwhile to
509 * implement the extra complexity (e.g. RCU protection) that would be
510 * needed to allow it to be set or changed during remount. We do allow
511 * it to be specified during remount, but only if there is no change.
512 */
513 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
514 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
515 return -EINVAL;
516 }
517
518 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
519 if (err) {
520 if (err == -EEXIST)
521 f2fs_warn(sbi,
522 "Can't change test_dummy_encryption on remount");
523 else if (err == -EINVAL)
524 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
525 opt);
526 else
527 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
528 opt, err);
529 return -EINVAL;
530 }
531 f2fs_warn(sbi, "Test dummy encryption mode enabled");
532 return 0;
533 }
534
535 #ifdef CONFIG_F2FS_FS_COMPRESSION
536 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
537 const char *new_ext, bool is_ext)
538 {
539 unsigned char (*ext)[F2FS_EXTENSION_LEN];
540 int ext_cnt;
541 int i;
542
543 if (is_ext) {
544 ext = F2FS_OPTION(sbi).extensions;
545 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
546 } else {
547 ext = F2FS_OPTION(sbi).noextensions;
548 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
549 }
550
551 for (i = 0; i < ext_cnt; i++) {
552 if (!strcasecmp(new_ext, ext[i]))
553 return true;
554 }
555
556 return false;
557 }
558
559 /*
560 * 1. The same extension name cannot not appear in both compress and non-compress extension
561 * at the same time.
562 * 2. If the compress extension specifies all files, the types specified by the non-compress
563 * extension will be treated as special cases and will not be compressed.
564 * 3. Don't allow the non-compress extension specifies all files.
565 */
566 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
567 {
568 unsigned char (*ext)[F2FS_EXTENSION_LEN];
569 unsigned char (*noext)[F2FS_EXTENSION_LEN];
570 int ext_cnt, noext_cnt, index = 0, no_index = 0;
571
572 ext = F2FS_OPTION(sbi).extensions;
573 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
574 noext = F2FS_OPTION(sbi).noextensions;
575 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
576
577 if (!noext_cnt)
578 return 0;
579
580 for (no_index = 0; no_index < noext_cnt; no_index++) {
581 if (!strcasecmp("*", noext[no_index])) {
582 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
583 return -EINVAL;
584 }
585 for (index = 0; index < ext_cnt; index++) {
586 if (!strcasecmp(ext[index], noext[no_index])) {
587 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
588 ext[index]);
589 return -EINVAL;
590 }
591 }
592 }
593 return 0;
594 }
595
596 #ifdef CONFIG_F2FS_FS_LZ4
597 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
598 {
599 #ifdef CONFIG_F2FS_FS_LZ4HC
600 unsigned int level;
601
602 if (strlen(str) == 3) {
603 F2FS_OPTION(sbi).compress_level = 0;
604 return 0;
605 }
606
607 str += 3;
608
609 if (str[0] != ':') {
610 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
611 return -EINVAL;
612 }
613 if (kstrtouint(str + 1, 10, &level))
614 return -EINVAL;
615
616 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
617 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
618 return -EINVAL;
619 }
620
621 F2FS_OPTION(sbi).compress_level = level;
622 return 0;
623 #else
624 if (strlen(str) == 3) {
625 F2FS_OPTION(sbi).compress_level = 0;
626 return 0;
627 }
628 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
629 return -EINVAL;
630 #endif
631 }
632 #endif
633
634 #ifdef CONFIG_F2FS_FS_ZSTD
635 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
636 {
637 int level;
638 int len = 4;
639
640 if (strlen(str) == len) {
641 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
642 return 0;
643 }
644
645 str += len;
646
647 if (str[0] != ':') {
648 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
649 return -EINVAL;
650 }
651 if (kstrtoint(str + 1, 10, &level))
652 return -EINVAL;
653
654 /* f2fs does not support negative compress level now */
655 if (level < 0) {
656 f2fs_info(sbi, "do not support negative compress level: %d", level);
657 return -ERANGE;
658 }
659
660 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
661 f2fs_info(sbi, "invalid zstd compress level: %d", level);
662 return -EINVAL;
663 }
664
665 F2FS_OPTION(sbi).compress_level = level;
666 return 0;
667 }
668 #endif
669 #endif
670
671 static int parse_options(struct super_block *sb, char *options, bool is_remount)
672 {
673 struct f2fs_sb_info *sbi = F2FS_SB(sb);
674 substring_t args[MAX_OPT_ARGS];
675 #ifdef CONFIG_F2FS_FS_COMPRESSION
676 unsigned char (*ext)[F2FS_EXTENSION_LEN];
677 unsigned char (*noext)[F2FS_EXTENSION_LEN];
678 int ext_cnt, noext_cnt;
679 #endif
680 char *p, *name;
681 int arg = 0;
682 kuid_t uid;
683 kgid_t gid;
684 int ret;
685
686 if (!options)
687 goto default_check;
688
689 while ((p = strsep(&options, ",")) != NULL) {
690 int token;
691
692 if (!*p)
693 continue;
694 /*
695 * Initialize args struct so we know whether arg was
696 * found; some options take optional arguments.
697 */
698 args[0].to = args[0].from = NULL;
699 token = match_token(p, f2fs_tokens, args);
700
701 switch (token) {
702 case Opt_gc_background:
703 name = match_strdup(&args[0]);
704
705 if (!name)
706 return -ENOMEM;
707 if (!strcmp(name, "on")) {
708 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
709 } else if (!strcmp(name, "off")) {
710 if (f2fs_sb_has_blkzoned(sbi)) {
711 f2fs_warn(sbi, "zoned devices need bggc");
712 kfree(name);
713 return -EINVAL;
714 }
715 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
716 } else if (!strcmp(name, "sync")) {
717 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
718 } else {
719 kfree(name);
720 return -EINVAL;
721 }
722 kfree(name);
723 break;
724 case Opt_disable_roll_forward:
725 set_opt(sbi, DISABLE_ROLL_FORWARD);
726 break;
727 case Opt_norecovery:
728 /* this option mounts f2fs with ro */
729 set_opt(sbi, NORECOVERY);
730 if (!f2fs_readonly(sb))
731 return -EINVAL;
732 break;
733 case Opt_discard:
734 if (!f2fs_hw_support_discard(sbi)) {
735 f2fs_warn(sbi, "device does not support discard");
736 break;
737 }
738 set_opt(sbi, DISCARD);
739 break;
740 case Opt_nodiscard:
741 if (f2fs_hw_should_discard(sbi)) {
742 f2fs_warn(sbi, "discard is required for zoned block devices");
743 return -EINVAL;
744 }
745 clear_opt(sbi, DISCARD);
746 break;
747 case Opt_noheap:
748 case Opt_heap:
749 f2fs_warn(sbi, "heap/no_heap options were deprecated");
750 break;
751 #ifdef CONFIG_F2FS_FS_XATTR
752 case Opt_user_xattr:
753 set_opt(sbi, XATTR_USER);
754 break;
755 case Opt_nouser_xattr:
756 clear_opt(sbi, XATTR_USER);
757 break;
758 case Opt_inline_xattr:
759 set_opt(sbi, INLINE_XATTR);
760 break;
761 case Opt_noinline_xattr:
762 clear_opt(sbi, INLINE_XATTR);
763 break;
764 case Opt_inline_xattr_size:
765 if (args->from && match_int(args, &arg))
766 return -EINVAL;
767 set_opt(sbi, INLINE_XATTR_SIZE);
768 F2FS_OPTION(sbi).inline_xattr_size = arg;
769 break;
770 #else
771 case Opt_user_xattr:
772 f2fs_info(sbi, "user_xattr options not supported");
773 break;
774 case Opt_nouser_xattr:
775 f2fs_info(sbi, "nouser_xattr options not supported");
776 break;
777 case Opt_inline_xattr:
778 f2fs_info(sbi, "inline_xattr options not supported");
779 break;
780 case Opt_noinline_xattr:
781 f2fs_info(sbi, "noinline_xattr options not supported");
782 break;
783 #endif
784 #ifdef CONFIG_F2FS_FS_POSIX_ACL
785 case Opt_acl:
786 set_opt(sbi, POSIX_ACL);
787 break;
788 case Opt_noacl:
789 clear_opt(sbi, POSIX_ACL);
790 break;
791 #else
792 case Opt_acl:
793 f2fs_info(sbi, "acl options not supported");
794 break;
795 case Opt_noacl:
796 f2fs_info(sbi, "noacl options not supported");
797 break;
798 #endif
799 case Opt_active_logs:
800 if (args->from && match_int(args, &arg))
801 return -EINVAL;
802 if (arg != 2 && arg != 4 &&
803 arg != NR_CURSEG_PERSIST_TYPE)
804 return -EINVAL;
805 F2FS_OPTION(sbi).active_logs = arg;
806 break;
807 case Opt_disable_ext_identify:
808 set_opt(sbi, DISABLE_EXT_IDENTIFY);
809 break;
810 case Opt_inline_data:
811 set_opt(sbi, INLINE_DATA);
812 break;
813 case Opt_inline_dentry:
814 set_opt(sbi, INLINE_DENTRY);
815 break;
816 case Opt_noinline_dentry:
817 clear_opt(sbi, INLINE_DENTRY);
818 break;
819 case Opt_flush_merge:
820 set_opt(sbi, FLUSH_MERGE);
821 break;
822 case Opt_noflush_merge:
823 clear_opt(sbi, FLUSH_MERGE);
824 break;
825 case Opt_nobarrier:
826 set_opt(sbi, NOBARRIER);
827 break;
828 case Opt_barrier:
829 clear_opt(sbi, NOBARRIER);
830 break;
831 case Opt_fastboot:
832 set_opt(sbi, FASTBOOT);
833 break;
834 case Opt_extent_cache:
835 set_opt(sbi, READ_EXTENT_CACHE);
836 break;
837 case Opt_noextent_cache:
838 clear_opt(sbi, READ_EXTENT_CACHE);
839 break;
840 case Opt_noinline_data:
841 clear_opt(sbi, INLINE_DATA);
842 break;
843 case Opt_data_flush:
844 set_opt(sbi, DATA_FLUSH);
845 break;
846 case Opt_reserve_root:
847 if (args->from && match_int(args, &arg))
848 return -EINVAL;
849 if (test_opt(sbi, RESERVE_ROOT)) {
850 f2fs_info(sbi, "Preserve previous reserve_root=%u",
851 F2FS_OPTION(sbi).root_reserved_blocks);
852 } else {
853 F2FS_OPTION(sbi).root_reserved_blocks = arg;
854 set_opt(sbi, RESERVE_ROOT);
855 }
856 break;
857 case Opt_resuid:
858 if (args->from && match_int(args, &arg))
859 return -EINVAL;
860 uid = make_kuid(current_user_ns(), arg);
861 if (!uid_valid(uid)) {
862 f2fs_err(sbi, "Invalid uid value %d", arg);
863 return -EINVAL;
864 }
865 F2FS_OPTION(sbi).s_resuid = uid;
866 break;
867 case Opt_resgid:
868 if (args->from && match_int(args, &arg))
869 return -EINVAL;
870 gid = make_kgid(current_user_ns(), arg);
871 if (!gid_valid(gid)) {
872 f2fs_err(sbi, "Invalid gid value %d", arg);
873 return -EINVAL;
874 }
875 F2FS_OPTION(sbi).s_resgid = gid;
876 break;
877 case Opt_mode:
878 name = match_strdup(&args[0]);
879
880 if (!name)
881 return -ENOMEM;
882 if (!strcmp(name, "adaptive")) {
883 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
884 } else if (!strcmp(name, "lfs")) {
885 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
886 } else if (!strcmp(name, "fragment:segment")) {
887 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
888 } else if (!strcmp(name, "fragment:block")) {
889 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
890 } else {
891 kfree(name);
892 return -EINVAL;
893 }
894 kfree(name);
895 break;
896 #ifdef CONFIG_F2FS_FAULT_INJECTION
897 case Opt_fault_injection:
898 if (args->from && match_int(args, &arg))
899 return -EINVAL;
900 if (f2fs_build_fault_attr(sbi, arg,
901 F2FS_ALL_FAULT_TYPE))
902 return -EINVAL;
903 set_opt(sbi, FAULT_INJECTION);
904 break;
905
906 case Opt_fault_type:
907 if (args->from && match_int(args, &arg))
908 return -EINVAL;
909 if (f2fs_build_fault_attr(sbi, 0, arg))
910 return -EINVAL;
911 set_opt(sbi, FAULT_INJECTION);
912 break;
913 #else
914 case Opt_fault_injection:
915 f2fs_info(sbi, "fault_injection options not supported");
916 break;
917
918 case Opt_fault_type:
919 f2fs_info(sbi, "fault_type options not supported");
920 break;
921 #endif
922 #ifdef CONFIG_QUOTA
923 case Opt_quota:
924 case Opt_usrquota:
925 set_opt(sbi, USRQUOTA);
926 break;
927 case Opt_grpquota:
928 set_opt(sbi, GRPQUOTA);
929 break;
930 case Opt_prjquota:
931 set_opt(sbi, PRJQUOTA);
932 break;
933 case Opt_usrjquota:
934 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
935 if (ret)
936 return ret;
937 break;
938 case Opt_grpjquota:
939 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
940 if (ret)
941 return ret;
942 break;
943 case Opt_prjjquota:
944 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
945 if (ret)
946 return ret;
947 break;
948 case Opt_offusrjquota:
949 ret = f2fs_clear_qf_name(sb, USRQUOTA);
950 if (ret)
951 return ret;
952 break;
953 case Opt_offgrpjquota:
954 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
955 if (ret)
956 return ret;
957 break;
958 case Opt_offprjjquota:
959 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
960 if (ret)
961 return ret;
962 break;
963 case Opt_jqfmt_vfsold:
964 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
965 break;
966 case Opt_jqfmt_vfsv0:
967 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
968 break;
969 case Opt_jqfmt_vfsv1:
970 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
971 break;
972 case Opt_noquota:
973 clear_opt(sbi, QUOTA);
974 clear_opt(sbi, USRQUOTA);
975 clear_opt(sbi, GRPQUOTA);
976 clear_opt(sbi, PRJQUOTA);
977 break;
978 #else
979 case Opt_quota:
980 case Opt_usrquota:
981 case Opt_grpquota:
982 case Opt_prjquota:
983 case Opt_usrjquota:
984 case Opt_grpjquota:
985 case Opt_prjjquota:
986 case Opt_offusrjquota:
987 case Opt_offgrpjquota:
988 case Opt_offprjjquota:
989 case Opt_jqfmt_vfsold:
990 case Opt_jqfmt_vfsv0:
991 case Opt_jqfmt_vfsv1:
992 case Opt_noquota:
993 f2fs_info(sbi, "quota operations not supported");
994 break;
995 #endif
996 case Opt_alloc:
997 name = match_strdup(&args[0]);
998 if (!name)
999 return -ENOMEM;
1000
1001 if (!strcmp(name, "default")) {
1002 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1003 } else if (!strcmp(name, "reuse")) {
1004 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1005 } else {
1006 kfree(name);
1007 return -EINVAL;
1008 }
1009 kfree(name);
1010 break;
1011 case Opt_fsync:
1012 name = match_strdup(&args[0]);
1013 if (!name)
1014 return -ENOMEM;
1015 if (!strcmp(name, "posix")) {
1016 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1017 } else if (!strcmp(name, "strict")) {
1018 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1019 } else if (!strcmp(name, "nobarrier")) {
1020 F2FS_OPTION(sbi).fsync_mode =
1021 FSYNC_MODE_NOBARRIER;
1022 } else {
1023 kfree(name);
1024 return -EINVAL;
1025 }
1026 kfree(name);
1027 break;
1028 case Opt_test_dummy_encryption:
1029 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1030 is_remount);
1031 if (ret)
1032 return ret;
1033 break;
1034 case Opt_inlinecrypt:
1035 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1036 sb->s_flags |= SB_INLINECRYPT;
1037 #else
1038 f2fs_info(sbi, "inline encryption not supported");
1039 #endif
1040 break;
1041 case Opt_checkpoint_disable_cap_perc:
1042 if (args->from && match_int(args, &arg))
1043 return -EINVAL;
1044 if (arg < 0 || arg > 100)
1045 return -EINVAL;
1046 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1047 set_opt(sbi, DISABLE_CHECKPOINT);
1048 break;
1049 case Opt_checkpoint_disable_cap:
1050 if (args->from && match_int(args, &arg))
1051 return -EINVAL;
1052 F2FS_OPTION(sbi).unusable_cap = arg;
1053 set_opt(sbi, DISABLE_CHECKPOINT);
1054 break;
1055 case Opt_checkpoint_disable:
1056 set_opt(sbi, DISABLE_CHECKPOINT);
1057 break;
1058 case Opt_checkpoint_enable:
1059 clear_opt(sbi, DISABLE_CHECKPOINT);
1060 break;
1061 case Opt_checkpoint_merge:
1062 set_opt(sbi, MERGE_CHECKPOINT);
1063 break;
1064 case Opt_nocheckpoint_merge:
1065 clear_opt(sbi, MERGE_CHECKPOINT);
1066 break;
1067 #ifdef CONFIG_F2FS_FS_COMPRESSION
1068 case Opt_compress_algorithm:
1069 if (!f2fs_sb_has_compression(sbi)) {
1070 f2fs_info(sbi, "Image doesn't support compression");
1071 break;
1072 }
1073 name = match_strdup(&args[0]);
1074 if (!name)
1075 return -ENOMEM;
1076 if (!strcmp(name, "lzo")) {
1077 #ifdef CONFIG_F2FS_FS_LZO
1078 F2FS_OPTION(sbi).compress_level = 0;
1079 F2FS_OPTION(sbi).compress_algorithm =
1080 COMPRESS_LZO;
1081 #else
1082 f2fs_info(sbi, "kernel doesn't support lzo compression");
1083 #endif
1084 } else if (!strncmp(name, "lz4", 3)) {
1085 #ifdef CONFIG_F2FS_FS_LZ4
1086 ret = f2fs_set_lz4hc_level(sbi, name);
1087 if (ret) {
1088 kfree(name);
1089 return -EINVAL;
1090 }
1091 F2FS_OPTION(sbi).compress_algorithm =
1092 COMPRESS_LZ4;
1093 #else
1094 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1095 #endif
1096 } else if (!strncmp(name, "zstd", 4)) {
1097 #ifdef CONFIG_F2FS_FS_ZSTD
1098 ret = f2fs_set_zstd_level(sbi, name);
1099 if (ret) {
1100 kfree(name);
1101 return -EINVAL;
1102 }
1103 F2FS_OPTION(sbi).compress_algorithm =
1104 COMPRESS_ZSTD;
1105 #else
1106 f2fs_info(sbi, "kernel doesn't support zstd compression");
1107 #endif
1108 } else if (!strcmp(name, "lzo-rle")) {
1109 #ifdef CONFIG_F2FS_FS_LZORLE
1110 F2FS_OPTION(sbi).compress_level = 0;
1111 F2FS_OPTION(sbi).compress_algorithm =
1112 COMPRESS_LZORLE;
1113 #else
1114 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1115 #endif
1116 } else {
1117 kfree(name);
1118 return -EINVAL;
1119 }
1120 kfree(name);
1121 break;
1122 case Opt_compress_log_size:
1123 if (!f2fs_sb_has_compression(sbi)) {
1124 f2fs_info(sbi, "Image doesn't support compression");
1125 break;
1126 }
1127 if (args->from && match_int(args, &arg))
1128 return -EINVAL;
1129 if (arg < MIN_COMPRESS_LOG_SIZE ||
1130 arg > MAX_COMPRESS_LOG_SIZE) {
1131 f2fs_err(sbi,
1132 "Compress cluster log size is out of range");
1133 return -EINVAL;
1134 }
1135 F2FS_OPTION(sbi).compress_log_size = arg;
1136 break;
1137 case Opt_compress_extension:
1138 if (!f2fs_sb_has_compression(sbi)) {
1139 f2fs_info(sbi, "Image doesn't support compression");
1140 break;
1141 }
1142 name = match_strdup(&args[0]);
1143 if (!name)
1144 return -ENOMEM;
1145
1146 ext = F2FS_OPTION(sbi).extensions;
1147 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1148
1149 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1150 ext_cnt >= COMPRESS_EXT_NUM) {
1151 f2fs_err(sbi,
1152 "invalid extension length/number");
1153 kfree(name);
1154 return -EINVAL;
1155 }
1156
1157 if (is_compress_extension_exist(sbi, name, true)) {
1158 kfree(name);
1159 break;
1160 }
1161
1162 strcpy(ext[ext_cnt], name);
1163 F2FS_OPTION(sbi).compress_ext_cnt++;
1164 kfree(name);
1165 break;
1166 case Opt_nocompress_extension:
1167 if (!f2fs_sb_has_compression(sbi)) {
1168 f2fs_info(sbi, "Image doesn't support compression");
1169 break;
1170 }
1171 name = match_strdup(&args[0]);
1172 if (!name)
1173 return -ENOMEM;
1174
1175 noext = F2FS_OPTION(sbi).noextensions;
1176 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1177
1178 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1179 noext_cnt >= COMPRESS_EXT_NUM) {
1180 f2fs_err(sbi,
1181 "invalid extension length/number");
1182 kfree(name);
1183 return -EINVAL;
1184 }
1185
1186 if (is_compress_extension_exist(sbi, name, false)) {
1187 kfree(name);
1188 break;
1189 }
1190
1191 strcpy(noext[noext_cnt], name);
1192 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1193 kfree(name);
1194 break;
1195 case Opt_compress_chksum:
1196 if (!f2fs_sb_has_compression(sbi)) {
1197 f2fs_info(sbi, "Image doesn't support compression");
1198 break;
1199 }
1200 F2FS_OPTION(sbi).compress_chksum = true;
1201 break;
1202 case Opt_compress_mode:
1203 if (!f2fs_sb_has_compression(sbi)) {
1204 f2fs_info(sbi, "Image doesn't support compression");
1205 break;
1206 }
1207 name = match_strdup(&args[0]);
1208 if (!name)
1209 return -ENOMEM;
1210 if (!strcmp(name, "fs")) {
1211 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1212 } else if (!strcmp(name, "user")) {
1213 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1214 } else {
1215 kfree(name);
1216 return -EINVAL;
1217 }
1218 kfree(name);
1219 break;
1220 case Opt_compress_cache:
1221 if (!f2fs_sb_has_compression(sbi)) {
1222 f2fs_info(sbi, "Image doesn't support compression");
1223 break;
1224 }
1225 set_opt(sbi, COMPRESS_CACHE);
1226 break;
1227 #else
1228 case Opt_compress_algorithm:
1229 case Opt_compress_log_size:
1230 case Opt_compress_extension:
1231 case Opt_nocompress_extension:
1232 case Opt_compress_chksum:
1233 case Opt_compress_mode:
1234 case Opt_compress_cache:
1235 f2fs_info(sbi, "compression options not supported");
1236 break;
1237 #endif
1238 case Opt_atgc:
1239 set_opt(sbi, ATGC);
1240 break;
1241 case Opt_gc_merge:
1242 set_opt(sbi, GC_MERGE);
1243 break;
1244 case Opt_nogc_merge:
1245 clear_opt(sbi, GC_MERGE);
1246 break;
1247 case Opt_discard_unit:
1248 name = match_strdup(&args[0]);
1249 if (!name)
1250 return -ENOMEM;
1251 if (!strcmp(name, "block")) {
1252 F2FS_OPTION(sbi).discard_unit =
1253 DISCARD_UNIT_BLOCK;
1254 } else if (!strcmp(name, "segment")) {
1255 F2FS_OPTION(sbi).discard_unit =
1256 DISCARD_UNIT_SEGMENT;
1257 } else if (!strcmp(name, "section")) {
1258 F2FS_OPTION(sbi).discard_unit =
1259 DISCARD_UNIT_SECTION;
1260 } else {
1261 kfree(name);
1262 return -EINVAL;
1263 }
1264 kfree(name);
1265 break;
1266 case Opt_memory_mode:
1267 name = match_strdup(&args[0]);
1268 if (!name)
1269 return -ENOMEM;
1270 if (!strcmp(name, "normal")) {
1271 F2FS_OPTION(sbi).memory_mode =
1272 MEMORY_MODE_NORMAL;
1273 } else if (!strcmp(name, "low")) {
1274 F2FS_OPTION(sbi).memory_mode =
1275 MEMORY_MODE_LOW;
1276 } else {
1277 kfree(name);
1278 return -EINVAL;
1279 }
1280 kfree(name);
1281 break;
1282 case Opt_age_extent_cache:
1283 set_opt(sbi, AGE_EXTENT_CACHE);
1284 break;
1285 case Opt_errors:
1286 name = match_strdup(&args[0]);
1287 if (!name)
1288 return -ENOMEM;
1289 if (!strcmp(name, "remount-ro")) {
1290 F2FS_OPTION(sbi).errors =
1291 MOUNT_ERRORS_READONLY;
1292 } else if (!strcmp(name, "continue")) {
1293 F2FS_OPTION(sbi).errors =
1294 MOUNT_ERRORS_CONTINUE;
1295 } else if (!strcmp(name, "panic")) {
1296 F2FS_OPTION(sbi).errors =
1297 MOUNT_ERRORS_PANIC;
1298 } else {
1299 kfree(name);
1300 return -EINVAL;
1301 }
1302 kfree(name);
1303 break;
1304 default:
1305 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1306 p);
1307 return -EINVAL;
1308 }
1309 }
1310 default_check:
1311 #ifdef CONFIG_QUOTA
1312 if (f2fs_check_quota_options(sbi))
1313 return -EINVAL;
1314 #else
1315 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1316 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1317 return -EINVAL;
1318 }
1319 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1320 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1321 return -EINVAL;
1322 }
1323 #endif
1324
1325 if (!IS_ENABLED(CONFIG_UNICODE) && f2fs_sb_has_casefold(sbi)) {
1326 f2fs_err(sbi,
1327 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1328 return -EINVAL;
1329 }
1330
1331 /*
1332 * The BLKZONED feature indicates that the drive was formatted with
1333 * zone alignment optimization. This is optional for host-aware
1334 * devices, but mandatory for host-managed zoned block devices.
1335 */
1336 if (f2fs_sb_has_blkzoned(sbi)) {
1337 #ifdef CONFIG_BLK_DEV_ZONED
1338 if (F2FS_OPTION(sbi).discard_unit !=
1339 DISCARD_UNIT_SECTION) {
1340 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1341 F2FS_OPTION(sbi).discard_unit =
1342 DISCARD_UNIT_SECTION;
1343 }
1344
1345 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1346 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1347 return -EINVAL;
1348 }
1349 #else
1350 f2fs_err(sbi, "Zoned block device support is not enabled");
1351 return -EINVAL;
1352 #endif
1353 }
1354
1355 #ifdef CONFIG_F2FS_FS_COMPRESSION
1356 if (f2fs_test_compress_extension(sbi)) {
1357 f2fs_err(sbi, "invalid compress or nocompress extension");
1358 return -EINVAL;
1359 }
1360 #endif
1361
1362 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1363 int min_size, max_size;
1364
1365 if (!f2fs_sb_has_extra_attr(sbi) ||
1366 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1367 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1368 return -EINVAL;
1369 }
1370 if (!test_opt(sbi, INLINE_XATTR)) {
1371 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1372 return -EINVAL;
1373 }
1374
1375 min_size = MIN_INLINE_XATTR_SIZE;
1376 max_size = MAX_INLINE_XATTR_SIZE;
1377
1378 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1379 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1380 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1381 min_size, max_size);
1382 return -EINVAL;
1383 }
1384 }
1385
1386 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1387 f2fs_err(sbi, "LFS is not compatible with ATGC");
1388 return -EINVAL;
1389 }
1390
1391 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1392 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1393 return -EINVAL;
1394 }
1395
1396 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1397 f2fs_err(sbi, "Allow to mount readonly mode only");
1398 return -EROFS;
1399 }
1400 return 0;
1401 }
1402
1403 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1404 {
1405 struct f2fs_inode_info *fi;
1406
1407 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1408 return NULL;
1409
1410 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1411 if (!fi)
1412 return NULL;
1413
1414 init_once((void *) fi);
1415
1416 /* Initialize f2fs-specific inode info */
1417 atomic_set(&fi->dirty_pages, 0);
1418 atomic_set(&fi->i_compr_blocks, 0);
1419 init_f2fs_rwsem(&fi->i_sem);
1420 spin_lock_init(&fi->i_size_lock);
1421 INIT_LIST_HEAD(&fi->dirty_list);
1422 INIT_LIST_HEAD(&fi->gdirty_list);
1423 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1424 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1425 init_f2fs_rwsem(&fi->i_xattr_sem);
1426
1427 /* Will be used by directory only */
1428 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1429
1430 return &fi->vfs_inode;
1431 }
1432
1433 static int f2fs_drop_inode(struct inode *inode)
1434 {
1435 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1436 int ret;
1437
1438 /*
1439 * during filesystem shutdown, if checkpoint is disabled,
1440 * drop useless meta/node dirty pages.
1441 */
1442 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1443 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1444 inode->i_ino == F2FS_META_INO(sbi)) {
1445 trace_f2fs_drop_inode(inode, 1);
1446 return 1;
1447 }
1448 }
1449
1450 /*
1451 * This is to avoid a deadlock condition like below.
1452 * writeback_single_inode(inode)
1453 * - f2fs_write_data_page
1454 * - f2fs_gc -> iput -> evict
1455 * - inode_wait_for_writeback(inode)
1456 */
1457 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1458 if (!inode->i_nlink && !is_bad_inode(inode)) {
1459 /* to avoid evict_inode call simultaneously */
1460 atomic_inc(&inode->i_count);
1461 spin_unlock(&inode->i_lock);
1462
1463 /* should remain fi->extent_tree for writepage */
1464 f2fs_destroy_extent_node(inode);
1465
1466 sb_start_intwrite(inode->i_sb);
1467 f2fs_i_size_write(inode, 0);
1468
1469 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1470 inode, NULL, 0, DATA);
1471 truncate_inode_pages_final(inode->i_mapping);
1472
1473 if (F2FS_HAS_BLOCKS(inode))
1474 f2fs_truncate(inode);
1475
1476 sb_end_intwrite(inode->i_sb);
1477
1478 spin_lock(&inode->i_lock);
1479 atomic_dec(&inode->i_count);
1480 }
1481 trace_f2fs_drop_inode(inode, 0);
1482 return 0;
1483 }
1484 ret = generic_drop_inode(inode);
1485 if (!ret)
1486 ret = fscrypt_drop_inode(inode);
1487 trace_f2fs_drop_inode(inode, ret);
1488 return ret;
1489 }
1490
1491 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1492 {
1493 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1494 int ret = 0;
1495
1496 spin_lock(&sbi->inode_lock[DIRTY_META]);
1497 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1498 ret = 1;
1499 } else {
1500 set_inode_flag(inode, FI_DIRTY_INODE);
1501 stat_inc_dirty_inode(sbi, DIRTY_META);
1502 }
1503 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1504 list_add_tail(&F2FS_I(inode)->gdirty_list,
1505 &sbi->inode_list[DIRTY_META]);
1506 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1507 }
1508 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1509 return ret;
1510 }
1511
1512 void f2fs_inode_synced(struct inode *inode)
1513 {
1514 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1515
1516 spin_lock(&sbi->inode_lock[DIRTY_META]);
1517 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1518 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1519 return;
1520 }
1521 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1522 list_del_init(&F2FS_I(inode)->gdirty_list);
1523 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1524 }
1525 clear_inode_flag(inode, FI_DIRTY_INODE);
1526 clear_inode_flag(inode, FI_AUTO_RECOVER);
1527 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1528 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1529 }
1530
1531 /*
1532 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1533 *
1534 * We should call set_dirty_inode to write the dirty inode through write_inode.
1535 */
1536 static void f2fs_dirty_inode(struct inode *inode, int flags)
1537 {
1538 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1539
1540 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1541 inode->i_ino == F2FS_META_INO(sbi))
1542 return;
1543
1544 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1545 clear_inode_flag(inode, FI_AUTO_RECOVER);
1546
1547 f2fs_inode_dirtied(inode, false);
1548 }
1549
1550 static void f2fs_free_inode(struct inode *inode)
1551 {
1552 fscrypt_free_inode(inode);
1553 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1554 }
1555
1556 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1557 {
1558 percpu_counter_destroy(&sbi->total_valid_inode_count);
1559 percpu_counter_destroy(&sbi->rf_node_block_count);
1560 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1561 }
1562
1563 static void destroy_device_list(struct f2fs_sb_info *sbi)
1564 {
1565 int i;
1566
1567 for (i = 0; i < sbi->s_ndevs; i++) {
1568 if (i > 0)
1569 bdev_fput(FDEV(i).bdev_file);
1570 #ifdef CONFIG_BLK_DEV_ZONED
1571 kvfree(FDEV(i).blkz_seq);
1572 #endif
1573 }
1574 kvfree(sbi->devs);
1575 }
1576
1577 static void f2fs_put_super(struct super_block *sb)
1578 {
1579 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1580 int i;
1581 int err = 0;
1582 bool done;
1583
1584 /* unregister procfs/sysfs entries in advance to avoid race case */
1585 f2fs_unregister_sysfs(sbi);
1586
1587 f2fs_quota_off_umount(sb);
1588
1589 /* prevent remaining shrinker jobs */
1590 mutex_lock(&sbi->umount_mutex);
1591
1592 /*
1593 * flush all issued checkpoints and stop checkpoint issue thread.
1594 * after then, all checkpoints should be done by each process context.
1595 */
1596 f2fs_stop_ckpt_thread(sbi);
1597
1598 /*
1599 * We don't need to do checkpoint when superblock is clean.
1600 * But, the previous checkpoint was not done by umount, it needs to do
1601 * clean checkpoint again.
1602 */
1603 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1604 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1605 struct cp_control cpc = {
1606 .reason = CP_UMOUNT,
1607 };
1608 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1609 err = f2fs_write_checkpoint(sbi, &cpc);
1610 }
1611
1612 /* be sure to wait for any on-going discard commands */
1613 done = f2fs_issue_discard_timeout(sbi);
1614 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1615 struct cp_control cpc = {
1616 .reason = CP_UMOUNT | CP_TRIMMED,
1617 };
1618 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1619 err = f2fs_write_checkpoint(sbi, &cpc);
1620 }
1621
1622 /*
1623 * normally superblock is clean, so we need to release this.
1624 * In addition, EIO will skip do checkpoint, we need this as well.
1625 */
1626 f2fs_release_ino_entry(sbi, true);
1627
1628 f2fs_leave_shrinker(sbi);
1629 mutex_unlock(&sbi->umount_mutex);
1630
1631 /* our cp_error case, we can wait for any writeback page */
1632 f2fs_flush_merged_writes(sbi);
1633
1634 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1635
1636 if (err || f2fs_cp_error(sbi)) {
1637 truncate_inode_pages_final(NODE_MAPPING(sbi));
1638 truncate_inode_pages_final(META_MAPPING(sbi));
1639 }
1640
1641 for (i = 0; i < NR_COUNT_TYPE; i++) {
1642 if (!get_pages(sbi, i))
1643 continue;
1644 f2fs_err(sbi, "detect filesystem reference count leak during "
1645 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1646 f2fs_bug_on(sbi, 1);
1647 }
1648
1649 f2fs_bug_on(sbi, sbi->fsync_node_num);
1650
1651 f2fs_destroy_compress_inode(sbi);
1652
1653 iput(sbi->node_inode);
1654 sbi->node_inode = NULL;
1655
1656 iput(sbi->meta_inode);
1657 sbi->meta_inode = NULL;
1658
1659 /*
1660 * iput() can update stat information, if f2fs_write_checkpoint()
1661 * above failed with error.
1662 */
1663 f2fs_destroy_stats(sbi);
1664
1665 /* destroy f2fs internal modules */
1666 f2fs_destroy_node_manager(sbi);
1667 f2fs_destroy_segment_manager(sbi);
1668
1669 /* flush s_error_work before sbi destroy */
1670 flush_work(&sbi->s_error_work);
1671
1672 f2fs_destroy_post_read_wq(sbi);
1673
1674 kvfree(sbi->ckpt);
1675
1676 if (sbi->s_chksum_driver)
1677 crypto_free_shash(sbi->s_chksum_driver);
1678 kfree(sbi->raw_super);
1679
1680 f2fs_destroy_page_array_cache(sbi);
1681 f2fs_destroy_xattr_caches(sbi);
1682 #ifdef CONFIG_QUOTA
1683 for (i = 0; i < MAXQUOTAS; i++)
1684 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1685 #endif
1686 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1687 destroy_percpu_info(sbi);
1688 f2fs_destroy_iostat(sbi);
1689 for (i = 0; i < NR_PAGE_TYPE; i++)
1690 kvfree(sbi->write_io[i]);
1691 #if IS_ENABLED(CONFIG_UNICODE)
1692 utf8_unload(sb->s_encoding);
1693 #endif
1694 }
1695
1696 int f2fs_sync_fs(struct super_block *sb, int sync)
1697 {
1698 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1699 int err = 0;
1700
1701 if (unlikely(f2fs_cp_error(sbi)))
1702 return 0;
1703 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1704 return 0;
1705
1706 trace_f2fs_sync_fs(sb, sync);
1707
1708 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1709 return -EAGAIN;
1710
1711 if (sync) {
1712 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1713 err = f2fs_issue_checkpoint(sbi);
1714 }
1715
1716 return err;
1717 }
1718
1719 static int f2fs_freeze(struct super_block *sb)
1720 {
1721 if (f2fs_readonly(sb))
1722 return 0;
1723
1724 /* IO error happened before */
1725 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1726 return -EIO;
1727
1728 /* must be clean, since sync_filesystem() was already called */
1729 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1730 return -EINVAL;
1731
1732 /* Let's flush checkpoints and stop the thread. */
1733 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1734
1735 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1736 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1737 return 0;
1738 }
1739
1740 static int f2fs_unfreeze(struct super_block *sb)
1741 {
1742 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1743 return 0;
1744 }
1745
1746 #ifdef CONFIG_QUOTA
1747 static int f2fs_statfs_project(struct super_block *sb,
1748 kprojid_t projid, struct kstatfs *buf)
1749 {
1750 struct kqid qid;
1751 struct dquot *dquot;
1752 u64 limit;
1753 u64 curblock;
1754
1755 qid = make_kqid_projid(projid);
1756 dquot = dqget(sb, qid);
1757 if (IS_ERR(dquot))
1758 return PTR_ERR(dquot);
1759 spin_lock(&dquot->dq_dqb_lock);
1760
1761 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1762 dquot->dq_dqb.dqb_bhardlimit);
1763 if (limit)
1764 limit >>= sb->s_blocksize_bits;
1765
1766 if (limit && buf->f_blocks > limit) {
1767 curblock = (dquot->dq_dqb.dqb_curspace +
1768 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1769 buf->f_blocks = limit;
1770 buf->f_bfree = buf->f_bavail =
1771 (buf->f_blocks > curblock) ?
1772 (buf->f_blocks - curblock) : 0;
1773 }
1774
1775 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1776 dquot->dq_dqb.dqb_ihardlimit);
1777
1778 if (limit && buf->f_files > limit) {
1779 buf->f_files = limit;
1780 buf->f_ffree =
1781 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1782 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1783 }
1784
1785 spin_unlock(&dquot->dq_dqb_lock);
1786 dqput(dquot);
1787 return 0;
1788 }
1789 #endif
1790
1791 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1792 {
1793 struct super_block *sb = dentry->d_sb;
1794 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1795 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1796 block_t total_count, user_block_count, start_count;
1797 u64 avail_node_count;
1798 unsigned int total_valid_node_count;
1799
1800 total_count = le64_to_cpu(sbi->raw_super->block_count);
1801 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1802 buf->f_type = F2FS_SUPER_MAGIC;
1803 buf->f_bsize = sbi->blocksize;
1804
1805 buf->f_blocks = total_count - start_count;
1806
1807 spin_lock(&sbi->stat_lock);
1808
1809 user_block_count = sbi->user_block_count;
1810 total_valid_node_count = valid_node_count(sbi);
1811 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1812 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1813 sbi->current_reserved_blocks;
1814
1815 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1816 buf->f_bfree = 0;
1817 else
1818 buf->f_bfree -= sbi->unusable_block_count;
1819 spin_unlock(&sbi->stat_lock);
1820
1821 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1822 buf->f_bavail = buf->f_bfree -
1823 F2FS_OPTION(sbi).root_reserved_blocks;
1824 else
1825 buf->f_bavail = 0;
1826
1827 if (avail_node_count > user_block_count) {
1828 buf->f_files = user_block_count;
1829 buf->f_ffree = buf->f_bavail;
1830 } else {
1831 buf->f_files = avail_node_count;
1832 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1833 buf->f_bavail);
1834 }
1835
1836 buf->f_namelen = F2FS_NAME_LEN;
1837 buf->f_fsid = u64_to_fsid(id);
1838
1839 #ifdef CONFIG_QUOTA
1840 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1841 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1842 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1843 }
1844 #endif
1845 return 0;
1846 }
1847
1848 static inline void f2fs_show_quota_options(struct seq_file *seq,
1849 struct super_block *sb)
1850 {
1851 #ifdef CONFIG_QUOTA
1852 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1853
1854 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1855 char *fmtname = "";
1856
1857 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1858 case QFMT_VFS_OLD:
1859 fmtname = "vfsold";
1860 break;
1861 case QFMT_VFS_V0:
1862 fmtname = "vfsv0";
1863 break;
1864 case QFMT_VFS_V1:
1865 fmtname = "vfsv1";
1866 break;
1867 }
1868 seq_printf(seq, ",jqfmt=%s", fmtname);
1869 }
1870
1871 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1872 seq_show_option(seq, "usrjquota",
1873 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1874
1875 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1876 seq_show_option(seq, "grpjquota",
1877 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1878
1879 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1880 seq_show_option(seq, "prjjquota",
1881 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1882 #endif
1883 }
1884
1885 #ifdef CONFIG_F2FS_FS_COMPRESSION
1886 static inline void f2fs_show_compress_options(struct seq_file *seq,
1887 struct super_block *sb)
1888 {
1889 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1890 char *algtype = "";
1891 int i;
1892
1893 if (!f2fs_sb_has_compression(sbi))
1894 return;
1895
1896 switch (F2FS_OPTION(sbi).compress_algorithm) {
1897 case COMPRESS_LZO:
1898 algtype = "lzo";
1899 break;
1900 case COMPRESS_LZ4:
1901 algtype = "lz4";
1902 break;
1903 case COMPRESS_ZSTD:
1904 algtype = "zstd";
1905 break;
1906 case COMPRESS_LZORLE:
1907 algtype = "lzo-rle";
1908 break;
1909 }
1910 seq_printf(seq, ",compress_algorithm=%s", algtype);
1911
1912 if (F2FS_OPTION(sbi).compress_level)
1913 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1914
1915 seq_printf(seq, ",compress_log_size=%u",
1916 F2FS_OPTION(sbi).compress_log_size);
1917
1918 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1919 seq_printf(seq, ",compress_extension=%s",
1920 F2FS_OPTION(sbi).extensions[i]);
1921 }
1922
1923 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1924 seq_printf(seq, ",nocompress_extension=%s",
1925 F2FS_OPTION(sbi).noextensions[i]);
1926 }
1927
1928 if (F2FS_OPTION(sbi).compress_chksum)
1929 seq_puts(seq, ",compress_chksum");
1930
1931 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1932 seq_printf(seq, ",compress_mode=%s", "fs");
1933 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1934 seq_printf(seq, ",compress_mode=%s", "user");
1935
1936 if (test_opt(sbi, COMPRESS_CACHE))
1937 seq_puts(seq, ",compress_cache");
1938 }
1939 #endif
1940
1941 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1942 {
1943 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1944
1945 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1946 seq_printf(seq, ",background_gc=%s", "sync");
1947 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1948 seq_printf(seq, ",background_gc=%s", "on");
1949 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1950 seq_printf(seq, ",background_gc=%s", "off");
1951
1952 if (test_opt(sbi, GC_MERGE))
1953 seq_puts(seq, ",gc_merge");
1954 else
1955 seq_puts(seq, ",nogc_merge");
1956
1957 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1958 seq_puts(seq, ",disable_roll_forward");
1959 if (test_opt(sbi, NORECOVERY))
1960 seq_puts(seq, ",norecovery");
1961 if (test_opt(sbi, DISCARD)) {
1962 seq_puts(seq, ",discard");
1963 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1964 seq_printf(seq, ",discard_unit=%s", "block");
1965 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1966 seq_printf(seq, ",discard_unit=%s", "segment");
1967 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1968 seq_printf(seq, ",discard_unit=%s", "section");
1969 } else {
1970 seq_puts(seq, ",nodiscard");
1971 }
1972 #ifdef CONFIG_F2FS_FS_XATTR
1973 if (test_opt(sbi, XATTR_USER))
1974 seq_puts(seq, ",user_xattr");
1975 else
1976 seq_puts(seq, ",nouser_xattr");
1977 if (test_opt(sbi, INLINE_XATTR))
1978 seq_puts(seq, ",inline_xattr");
1979 else
1980 seq_puts(seq, ",noinline_xattr");
1981 if (test_opt(sbi, INLINE_XATTR_SIZE))
1982 seq_printf(seq, ",inline_xattr_size=%u",
1983 F2FS_OPTION(sbi).inline_xattr_size);
1984 #endif
1985 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1986 if (test_opt(sbi, POSIX_ACL))
1987 seq_puts(seq, ",acl");
1988 else
1989 seq_puts(seq, ",noacl");
1990 #endif
1991 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1992 seq_puts(seq, ",disable_ext_identify");
1993 if (test_opt(sbi, INLINE_DATA))
1994 seq_puts(seq, ",inline_data");
1995 else
1996 seq_puts(seq, ",noinline_data");
1997 if (test_opt(sbi, INLINE_DENTRY))
1998 seq_puts(seq, ",inline_dentry");
1999 else
2000 seq_puts(seq, ",noinline_dentry");
2001 if (test_opt(sbi, FLUSH_MERGE))
2002 seq_puts(seq, ",flush_merge");
2003 else
2004 seq_puts(seq, ",noflush_merge");
2005 if (test_opt(sbi, NOBARRIER))
2006 seq_puts(seq, ",nobarrier");
2007 else
2008 seq_puts(seq, ",barrier");
2009 if (test_opt(sbi, FASTBOOT))
2010 seq_puts(seq, ",fastboot");
2011 if (test_opt(sbi, READ_EXTENT_CACHE))
2012 seq_puts(seq, ",extent_cache");
2013 else
2014 seq_puts(seq, ",noextent_cache");
2015 if (test_opt(sbi, AGE_EXTENT_CACHE))
2016 seq_puts(seq, ",age_extent_cache");
2017 if (test_opt(sbi, DATA_FLUSH))
2018 seq_puts(seq, ",data_flush");
2019
2020 seq_puts(seq, ",mode=");
2021 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2022 seq_puts(seq, "adaptive");
2023 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2024 seq_puts(seq, "lfs");
2025 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2026 seq_puts(seq, "fragment:segment");
2027 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2028 seq_puts(seq, "fragment:block");
2029 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2030 if (test_opt(sbi, RESERVE_ROOT))
2031 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2032 F2FS_OPTION(sbi).root_reserved_blocks,
2033 from_kuid_munged(&init_user_ns,
2034 F2FS_OPTION(sbi).s_resuid),
2035 from_kgid_munged(&init_user_ns,
2036 F2FS_OPTION(sbi).s_resgid));
2037 #ifdef CONFIG_F2FS_FAULT_INJECTION
2038 if (test_opt(sbi, FAULT_INJECTION)) {
2039 seq_printf(seq, ",fault_injection=%u",
2040 F2FS_OPTION(sbi).fault_info.inject_rate);
2041 seq_printf(seq, ",fault_type=%u",
2042 F2FS_OPTION(sbi).fault_info.inject_type);
2043 }
2044 #endif
2045 #ifdef CONFIG_QUOTA
2046 if (test_opt(sbi, QUOTA))
2047 seq_puts(seq, ",quota");
2048 if (test_opt(sbi, USRQUOTA))
2049 seq_puts(seq, ",usrquota");
2050 if (test_opt(sbi, GRPQUOTA))
2051 seq_puts(seq, ",grpquota");
2052 if (test_opt(sbi, PRJQUOTA))
2053 seq_puts(seq, ",prjquota");
2054 #endif
2055 f2fs_show_quota_options(seq, sbi->sb);
2056
2057 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2058
2059 if (sbi->sb->s_flags & SB_INLINECRYPT)
2060 seq_puts(seq, ",inlinecrypt");
2061
2062 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2063 seq_printf(seq, ",alloc_mode=%s", "default");
2064 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2065 seq_printf(seq, ",alloc_mode=%s", "reuse");
2066
2067 if (test_opt(sbi, DISABLE_CHECKPOINT))
2068 seq_printf(seq, ",checkpoint=disable:%u",
2069 F2FS_OPTION(sbi).unusable_cap);
2070 if (test_opt(sbi, MERGE_CHECKPOINT))
2071 seq_puts(seq, ",checkpoint_merge");
2072 else
2073 seq_puts(seq, ",nocheckpoint_merge");
2074 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2075 seq_printf(seq, ",fsync_mode=%s", "posix");
2076 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2077 seq_printf(seq, ",fsync_mode=%s", "strict");
2078 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2079 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2080
2081 #ifdef CONFIG_F2FS_FS_COMPRESSION
2082 f2fs_show_compress_options(seq, sbi->sb);
2083 #endif
2084
2085 if (test_opt(sbi, ATGC))
2086 seq_puts(seq, ",atgc");
2087
2088 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2089 seq_printf(seq, ",memory=%s", "normal");
2090 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2091 seq_printf(seq, ",memory=%s", "low");
2092
2093 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2094 seq_printf(seq, ",errors=%s", "remount-ro");
2095 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2096 seq_printf(seq, ",errors=%s", "continue");
2097 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2098 seq_printf(seq, ",errors=%s", "panic");
2099
2100 return 0;
2101 }
2102
2103 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2104 {
2105 /* init some FS parameters */
2106 if (!remount) {
2107 set_opt(sbi, READ_EXTENT_CACHE);
2108 clear_opt(sbi, DISABLE_CHECKPOINT);
2109
2110 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2111 set_opt(sbi, DISCARD);
2112
2113 if (f2fs_sb_has_blkzoned(sbi))
2114 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2115 else
2116 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2117 }
2118
2119 if (f2fs_sb_has_readonly(sbi))
2120 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2121 else
2122 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2123
2124 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2125 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2126 SMALL_VOLUME_SEGMENTS)
2127 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2128 else
2129 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2130 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2131 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2132 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2133 if (f2fs_sb_has_compression(sbi)) {
2134 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2135 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2136 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2137 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2138 }
2139 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2140 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2141 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2142
2143 set_opt(sbi, INLINE_XATTR);
2144 set_opt(sbi, INLINE_DATA);
2145 set_opt(sbi, INLINE_DENTRY);
2146 set_opt(sbi, MERGE_CHECKPOINT);
2147 F2FS_OPTION(sbi).unusable_cap = 0;
2148 sbi->sb->s_flags |= SB_LAZYTIME;
2149 if (!f2fs_is_readonly(sbi))
2150 set_opt(sbi, FLUSH_MERGE);
2151 if (f2fs_sb_has_blkzoned(sbi))
2152 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2153 else
2154 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2155
2156 #ifdef CONFIG_F2FS_FS_XATTR
2157 set_opt(sbi, XATTR_USER);
2158 #endif
2159 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2160 set_opt(sbi, POSIX_ACL);
2161 #endif
2162
2163 f2fs_build_fault_attr(sbi, 0, 0);
2164 }
2165
2166 #ifdef CONFIG_QUOTA
2167 static int f2fs_enable_quotas(struct super_block *sb);
2168 #endif
2169
2170 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2171 {
2172 unsigned int s_flags = sbi->sb->s_flags;
2173 struct cp_control cpc;
2174 unsigned int gc_mode = sbi->gc_mode;
2175 int err = 0;
2176 int ret;
2177 block_t unusable;
2178
2179 if (s_flags & SB_RDONLY) {
2180 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2181 return -EINVAL;
2182 }
2183 sbi->sb->s_flags |= SB_ACTIVE;
2184
2185 /* check if we need more GC first */
2186 unusable = f2fs_get_unusable_blocks(sbi);
2187 if (!f2fs_disable_cp_again(sbi, unusable))
2188 goto skip_gc;
2189
2190 f2fs_update_time(sbi, DISABLE_TIME);
2191
2192 sbi->gc_mode = GC_URGENT_HIGH;
2193
2194 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2195 struct f2fs_gc_control gc_control = {
2196 .victim_segno = NULL_SEGNO,
2197 .init_gc_type = FG_GC,
2198 .should_migrate_blocks = false,
2199 .err_gc_skipped = true,
2200 .no_bg_gc = true,
2201 .nr_free_secs = 1 };
2202
2203 f2fs_down_write(&sbi->gc_lock);
2204 stat_inc_gc_call_count(sbi, FOREGROUND);
2205 err = f2fs_gc(sbi, &gc_control);
2206 if (err == -ENODATA) {
2207 err = 0;
2208 break;
2209 }
2210 if (err && err != -EAGAIN)
2211 break;
2212 }
2213
2214 ret = sync_filesystem(sbi->sb);
2215 if (ret || err) {
2216 err = ret ? ret : err;
2217 goto restore_flag;
2218 }
2219
2220 unusable = f2fs_get_unusable_blocks(sbi);
2221 if (f2fs_disable_cp_again(sbi, unusable)) {
2222 err = -EAGAIN;
2223 goto restore_flag;
2224 }
2225
2226 skip_gc:
2227 f2fs_down_write(&sbi->gc_lock);
2228 cpc.reason = CP_PAUSE;
2229 set_sbi_flag(sbi, SBI_CP_DISABLED);
2230 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2231 err = f2fs_write_checkpoint(sbi, &cpc);
2232 if (err)
2233 goto out_unlock;
2234
2235 spin_lock(&sbi->stat_lock);
2236 sbi->unusable_block_count = unusable;
2237 spin_unlock(&sbi->stat_lock);
2238
2239 out_unlock:
2240 f2fs_up_write(&sbi->gc_lock);
2241 restore_flag:
2242 sbi->gc_mode = gc_mode;
2243 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2244 return err;
2245 }
2246
2247 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2248 {
2249 int retry = DEFAULT_RETRY_IO_COUNT;
2250
2251 /* we should flush all the data to keep data consistency */
2252 do {
2253 sync_inodes_sb(sbi->sb);
2254 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2255 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2256
2257 if (unlikely(retry < 0))
2258 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2259
2260 f2fs_down_write(&sbi->gc_lock);
2261 f2fs_dirty_to_prefree(sbi);
2262
2263 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2264 set_sbi_flag(sbi, SBI_IS_DIRTY);
2265 f2fs_up_write(&sbi->gc_lock);
2266
2267 f2fs_sync_fs(sbi->sb, 1);
2268
2269 /* Let's ensure there's no pending checkpoint anymore */
2270 f2fs_flush_ckpt_thread(sbi);
2271 }
2272
2273 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2274 {
2275 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2276 struct f2fs_mount_info org_mount_opt;
2277 unsigned long old_sb_flags;
2278 int err;
2279 bool need_restart_gc = false, need_stop_gc = false;
2280 bool need_restart_flush = false, need_stop_flush = false;
2281 bool need_restart_discard = false, need_stop_discard = false;
2282 bool need_enable_checkpoint = false, need_disable_checkpoint = false;
2283 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2284 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2285 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2286 bool no_atgc = !test_opt(sbi, ATGC);
2287 bool no_discard = !test_opt(sbi, DISCARD);
2288 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2289 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2290 #ifdef CONFIG_QUOTA
2291 int i, j;
2292 #endif
2293
2294 /*
2295 * Save the old mount options in case we
2296 * need to restore them.
2297 */
2298 org_mount_opt = sbi->mount_opt;
2299 old_sb_flags = sb->s_flags;
2300
2301 #ifdef CONFIG_QUOTA
2302 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2303 for (i = 0; i < MAXQUOTAS; i++) {
2304 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2305 org_mount_opt.s_qf_names[i] =
2306 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2307 GFP_KERNEL);
2308 if (!org_mount_opt.s_qf_names[i]) {
2309 for (j = 0; j < i; j++)
2310 kfree(org_mount_opt.s_qf_names[j]);
2311 return -ENOMEM;
2312 }
2313 } else {
2314 org_mount_opt.s_qf_names[i] = NULL;
2315 }
2316 }
2317 #endif
2318
2319 /* recover superblocks we couldn't write due to previous RO mount */
2320 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2321 err = f2fs_commit_super(sbi, false);
2322 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2323 err);
2324 if (!err)
2325 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2326 }
2327
2328 default_options(sbi, true);
2329
2330 /* parse mount options */
2331 err = parse_options(sb, data, true);
2332 if (err)
2333 goto restore_opts;
2334
2335 #ifdef CONFIG_BLK_DEV_ZONED
2336 if (f2fs_sb_has_blkzoned(sbi) &&
2337 sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
2338 f2fs_err(sbi,
2339 "zoned: max open zones %u is too small, need at least %u open zones",
2340 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
2341 err = -EINVAL;
2342 goto restore_opts;
2343 }
2344 #endif
2345
2346 /* flush outstanding errors before changing fs state */
2347 flush_work(&sbi->s_error_work);
2348
2349 /*
2350 * Previous and new state of filesystem is RO,
2351 * so skip checking GC and FLUSH_MERGE conditions.
2352 */
2353 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2354 goto skip;
2355
2356 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2357 err = -EROFS;
2358 goto restore_opts;
2359 }
2360
2361 #ifdef CONFIG_QUOTA
2362 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2363 err = dquot_suspend(sb, -1);
2364 if (err < 0)
2365 goto restore_opts;
2366 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2367 /* dquot_resume needs RW */
2368 sb->s_flags &= ~SB_RDONLY;
2369 if (sb_any_quota_suspended(sb)) {
2370 dquot_resume(sb, -1);
2371 } else if (f2fs_sb_has_quota_ino(sbi)) {
2372 err = f2fs_enable_quotas(sb);
2373 if (err)
2374 goto restore_opts;
2375 }
2376 }
2377 #endif
2378 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2379 err = -EINVAL;
2380 f2fs_warn(sbi, "LFS is not compatible with IPU");
2381 goto restore_opts;
2382 }
2383
2384 /* disallow enable atgc dynamically */
2385 if (no_atgc == !!test_opt(sbi, ATGC)) {
2386 err = -EINVAL;
2387 f2fs_warn(sbi, "switch atgc option is not allowed");
2388 goto restore_opts;
2389 }
2390
2391 /* disallow enable/disable extent_cache dynamically */
2392 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2393 err = -EINVAL;
2394 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2395 goto restore_opts;
2396 }
2397 /* disallow enable/disable age extent_cache dynamically */
2398 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2399 err = -EINVAL;
2400 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2401 goto restore_opts;
2402 }
2403
2404 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2405 err = -EINVAL;
2406 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2407 goto restore_opts;
2408 }
2409
2410 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2411 err = -EINVAL;
2412 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2413 goto restore_opts;
2414 }
2415
2416 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2417 err = -EINVAL;
2418 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2419 goto restore_opts;
2420 }
2421
2422 /*
2423 * We stop the GC thread if FS is mounted as RO
2424 * or if background_gc = off is passed in mount
2425 * option. Also sync the filesystem.
2426 */
2427 if ((*flags & SB_RDONLY) ||
2428 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2429 !test_opt(sbi, GC_MERGE))) {
2430 if (sbi->gc_thread) {
2431 f2fs_stop_gc_thread(sbi);
2432 need_restart_gc = true;
2433 }
2434 } else if (!sbi->gc_thread) {
2435 err = f2fs_start_gc_thread(sbi);
2436 if (err)
2437 goto restore_opts;
2438 need_stop_gc = true;
2439 }
2440
2441 if (*flags & SB_RDONLY) {
2442 sync_inodes_sb(sb);
2443
2444 set_sbi_flag(sbi, SBI_IS_DIRTY);
2445 set_sbi_flag(sbi, SBI_IS_CLOSE);
2446 f2fs_sync_fs(sb, 1);
2447 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2448 }
2449
2450 /*
2451 * We stop issue flush thread if FS is mounted as RO
2452 * or if flush_merge is not passed in mount option.
2453 */
2454 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2455 clear_opt(sbi, FLUSH_MERGE);
2456 f2fs_destroy_flush_cmd_control(sbi, false);
2457 need_restart_flush = true;
2458 } else {
2459 err = f2fs_create_flush_cmd_control(sbi);
2460 if (err)
2461 goto restore_gc;
2462 need_stop_flush = true;
2463 }
2464
2465 if (no_discard == !!test_opt(sbi, DISCARD)) {
2466 if (test_opt(sbi, DISCARD)) {
2467 err = f2fs_start_discard_thread(sbi);
2468 if (err)
2469 goto restore_flush;
2470 need_stop_discard = true;
2471 } else {
2472 f2fs_stop_discard_thread(sbi);
2473 f2fs_issue_discard_timeout(sbi);
2474 need_restart_discard = true;
2475 }
2476 }
2477
2478 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2479 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2480 err = f2fs_disable_checkpoint(sbi);
2481 if (err)
2482 goto restore_discard;
2483 need_enable_checkpoint = true;
2484 } else {
2485 f2fs_enable_checkpoint(sbi);
2486 need_disable_checkpoint = true;
2487 }
2488 }
2489
2490 /*
2491 * Place this routine at the end, since a new checkpoint would be
2492 * triggered while remount and we need to take care of it before
2493 * returning from remount.
2494 */
2495 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2496 !test_opt(sbi, MERGE_CHECKPOINT)) {
2497 f2fs_stop_ckpt_thread(sbi);
2498 } else {
2499 /* Flush if the prevous checkpoint, if exists. */
2500 f2fs_flush_ckpt_thread(sbi);
2501
2502 err = f2fs_start_ckpt_thread(sbi);
2503 if (err) {
2504 f2fs_err(sbi,
2505 "Failed to start F2FS issue_checkpoint_thread (%d)",
2506 err);
2507 goto restore_checkpoint;
2508 }
2509 }
2510
2511 skip:
2512 #ifdef CONFIG_QUOTA
2513 /* Release old quota file names */
2514 for (i = 0; i < MAXQUOTAS; i++)
2515 kfree(org_mount_opt.s_qf_names[i]);
2516 #endif
2517 /* Update the POSIXACL Flag */
2518 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2519 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2520
2521 limit_reserve_root(sbi);
2522 adjust_unusable_cap_perc(sbi);
2523 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2524 return 0;
2525 restore_checkpoint:
2526 if (need_enable_checkpoint) {
2527 f2fs_enable_checkpoint(sbi);
2528 } else if (need_disable_checkpoint) {
2529 if (f2fs_disable_checkpoint(sbi))
2530 f2fs_warn(sbi, "checkpoint has not been disabled");
2531 }
2532 restore_discard:
2533 if (need_restart_discard) {
2534 if (f2fs_start_discard_thread(sbi))
2535 f2fs_warn(sbi, "discard has been stopped");
2536 } else if (need_stop_discard) {
2537 f2fs_stop_discard_thread(sbi);
2538 }
2539 restore_flush:
2540 if (need_restart_flush) {
2541 if (f2fs_create_flush_cmd_control(sbi))
2542 f2fs_warn(sbi, "background flush thread has stopped");
2543 } else if (need_stop_flush) {
2544 clear_opt(sbi, FLUSH_MERGE);
2545 f2fs_destroy_flush_cmd_control(sbi, false);
2546 }
2547 restore_gc:
2548 if (need_restart_gc) {
2549 if (f2fs_start_gc_thread(sbi))
2550 f2fs_warn(sbi, "background gc thread has stopped");
2551 } else if (need_stop_gc) {
2552 f2fs_stop_gc_thread(sbi);
2553 }
2554 restore_opts:
2555 #ifdef CONFIG_QUOTA
2556 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2557 for (i = 0; i < MAXQUOTAS; i++) {
2558 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2559 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2560 }
2561 #endif
2562 sbi->mount_opt = org_mount_opt;
2563 sb->s_flags = old_sb_flags;
2564 return err;
2565 }
2566
2567 static void f2fs_shutdown(struct super_block *sb)
2568 {
2569 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false, false);
2570 }
2571
2572 #ifdef CONFIG_QUOTA
2573 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2574 {
2575 /* need to recovery orphan */
2576 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2577 return true;
2578 /* need to recovery data */
2579 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2580 return false;
2581 if (test_opt(sbi, NORECOVERY))
2582 return false;
2583 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2584 }
2585
2586 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2587 {
2588 bool readonly = f2fs_readonly(sbi->sb);
2589
2590 if (!f2fs_need_recovery(sbi))
2591 return false;
2592
2593 /* it doesn't need to check f2fs_sb_has_readonly() */
2594 if (f2fs_hw_is_readonly(sbi))
2595 return false;
2596
2597 if (readonly) {
2598 sbi->sb->s_flags &= ~SB_RDONLY;
2599 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2600 }
2601
2602 /*
2603 * Turn on quotas which were not enabled for read-only mounts if
2604 * filesystem has quota feature, so that they are updated correctly.
2605 */
2606 return f2fs_enable_quota_files(sbi, readonly);
2607 }
2608
2609 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2610 bool quota_enabled)
2611 {
2612 if (quota_enabled)
2613 f2fs_quota_off_umount(sbi->sb);
2614
2615 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2616 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2617 sbi->sb->s_flags |= SB_RDONLY;
2618 }
2619 }
2620
2621 /* Read data from quotafile */
2622 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2623 size_t len, loff_t off)
2624 {
2625 struct inode *inode = sb_dqopt(sb)->files[type];
2626 struct address_space *mapping = inode->i_mapping;
2627 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2628 int offset = off & (sb->s_blocksize - 1);
2629 int tocopy;
2630 size_t toread;
2631 loff_t i_size = i_size_read(inode);
2632 struct page *page;
2633
2634 if (off > i_size)
2635 return 0;
2636
2637 if (off + len > i_size)
2638 len = i_size - off;
2639 toread = len;
2640 while (toread > 0) {
2641 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2642 repeat:
2643 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2644 if (IS_ERR(page)) {
2645 if (PTR_ERR(page) == -ENOMEM) {
2646 memalloc_retry_wait(GFP_NOFS);
2647 goto repeat;
2648 }
2649 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2650 return PTR_ERR(page);
2651 }
2652
2653 lock_page(page);
2654
2655 if (unlikely(page->mapping != mapping)) {
2656 f2fs_put_page(page, 1);
2657 goto repeat;
2658 }
2659 if (unlikely(!PageUptodate(page))) {
2660 f2fs_put_page(page, 1);
2661 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2662 return -EIO;
2663 }
2664
2665 memcpy_from_page(data, page, offset, tocopy);
2666 f2fs_put_page(page, 1);
2667
2668 offset = 0;
2669 toread -= tocopy;
2670 data += tocopy;
2671 blkidx++;
2672 }
2673 return len;
2674 }
2675
2676 /* Write to quotafile */
2677 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2678 const char *data, size_t len, loff_t off)
2679 {
2680 struct inode *inode = sb_dqopt(sb)->files[type];
2681 struct address_space *mapping = inode->i_mapping;
2682 const struct address_space_operations *a_ops = mapping->a_ops;
2683 int offset = off & (sb->s_blocksize - 1);
2684 size_t towrite = len;
2685 struct page *page;
2686 void *fsdata = NULL;
2687 int err = 0;
2688 int tocopy;
2689
2690 while (towrite > 0) {
2691 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2692 towrite);
2693 retry:
2694 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2695 &page, &fsdata);
2696 if (unlikely(err)) {
2697 if (err == -ENOMEM) {
2698 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2699 goto retry;
2700 }
2701 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2702 break;
2703 }
2704
2705 memcpy_to_page(page, offset, data, tocopy);
2706
2707 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2708 page, fsdata);
2709 offset = 0;
2710 towrite -= tocopy;
2711 off += tocopy;
2712 data += tocopy;
2713 cond_resched();
2714 }
2715
2716 if (len == towrite)
2717 return err;
2718 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2719 f2fs_mark_inode_dirty_sync(inode, false);
2720 return len - towrite;
2721 }
2722
2723 int f2fs_dquot_initialize(struct inode *inode)
2724 {
2725 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2726 return -ESRCH;
2727
2728 return dquot_initialize(inode);
2729 }
2730
2731 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2732 {
2733 return F2FS_I(inode)->i_dquot;
2734 }
2735
2736 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2737 {
2738 return &F2FS_I(inode)->i_reserved_quota;
2739 }
2740
2741 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2742 {
2743 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2744 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2745 return 0;
2746 }
2747
2748 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2749 F2FS_OPTION(sbi).s_jquota_fmt, type);
2750 }
2751
2752 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2753 {
2754 int enabled = 0;
2755 int i, err;
2756
2757 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2758 err = f2fs_enable_quotas(sbi->sb);
2759 if (err) {
2760 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2761 return 0;
2762 }
2763 return 1;
2764 }
2765
2766 for (i = 0; i < MAXQUOTAS; i++) {
2767 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2768 err = f2fs_quota_on_mount(sbi, i);
2769 if (!err) {
2770 enabled = 1;
2771 continue;
2772 }
2773 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2774 err, i);
2775 }
2776 }
2777 return enabled;
2778 }
2779
2780 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2781 unsigned int flags)
2782 {
2783 struct inode *qf_inode;
2784 unsigned long qf_inum;
2785 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2786 int err;
2787
2788 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2789
2790 qf_inum = f2fs_qf_ino(sb, type);
2791 if (!qf_inum)
2792 return -EPERM;
2793
2794 qf_inode = f2fs_iget(sb, qf_inum);
2795 if (IS_ERR(qf_inode)) {
2796 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2797 return PTR_ERR(qf_inode);
2798 }
2799
2800 /* Don't account quota for quota files to avoid recursion */
2801 inode_lock(qf_inode);
2802 qf_inode->i_flags |= S_NOQUOTA;
2803
2804 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2805 F2FS_I(qf_inode)->i_flags |= qf_flag;
2806 f2fs_set_inode_flags(qf_inode);
2807 }
2808 inode_unlock(qf_inode);
2809
2810 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2811 iput(qf_inode);
2812 return err;
2813 }
2814
2815 static int f2fs_enable_quotas(struct super_block *sb)
2816 {
2817 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2818 int type, err = 0;
2819 unsigned long qf_inum;
2820 bool quota_mopt[MAXQUOTAS] = {
2821 test_opt(sbi, USRQUOTA),
2822 test_opt(sbi, GRPQUOTA),
2823 test_opt(sbi, PRJQUOTA),
2824 };
2825
2826 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2827 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2828 return 0;
2829 }
2830
2831 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2832
2833 for (type = 0; type < MAXQUOTAS; type++) {
2834 qf_inum = f2fs_qf_ino(sb, type);
2835 if (qf_inum) {
2836 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2837 DQUOT_USAGE_ENABLED |
2838 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2839 if (err) {
2840 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2841 type, err);
2842 for (type--; type >= 0; type--)
2843 dquot_quota_off(sb, type);
2844 set_sbi_flag(F2FS_SB(sb),
2845 SBI_QUOTA_NEED_REPAIR);
2846 return err;
2847 }
2848 }
2849 }
2850 return 0;
2851 }
2852
2853 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2854 {
2855 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2856 struct address_space *mapping = dqopt->files[type]->i_mapping;
2857 int ret = 0;
2858
2859 ret = dquot_writeback_dquots(sbi->sb, type);
2860 if (ret)
2861 goto out;
2862
2863 ret = filemap_fdatawrite(mapping);
2864 if (ret)
2865 goto out;
2866
2867 /* if we are using journalled quota */
2868 if (is_journalled_quota(sbi))
2869 goto out;
2870
2871 ret = filemap_fdatawait(mapping);
2872
2873 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2874 out:
2875 if (ret)
2876 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2877 return ret;
2878 }
2879
2880 int f2fs_quota_sync(struct super_block *sb, int type)
2881 {
2882 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2883 struct quota_info *dqopt = sb_dqopt(sb);
2884 int cnt;
2885 int ret = 0;
2886
2887 /*
2888 * Now when everything is written we can discard the pagecache so
2889 * that userspace sees the changes.
2890 */
2891 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2892
2893 if (type != -1 && cnt != type)
2894 continue;
2895
2896 if (!sb_has_quota_active(sb, cnt))
2897 continue;
2898
2899 if (!f2fs_sb_has_quota_ino(sbi))
2900 inode_lock(dqopt->files[cnt]);
2901
2902 /*
2903 * do_quotactl
2904 * f2fs_quota_sync
2905 * f2fs_down_read(quota_sem)
2906 * dquot_writeback_dquots()
2907 * f2fs_dquot_commit
2908 * block_operation
2909 * f2fs_down_read(quota_sem)
2910 */
2911 f2fs_lock_op(sbi);
2912 f2fs_down_read(&sbi->quota_sem);
2913
2914 ret = f2fs_quota_sync_file(sbi, cnt);
2915
2916 f2fs_up_read(&sbi->quota_sem);
2917 f2fs_unlock_op(sbi);
2918
2919 if (!f2fs_sb_has_quota_ino(sbi))
2920 inode_unlock(dqopt->files[cnt]);
2921
2922 if (ret)
2923 break;
2924 }
2925 return ret;
2926 }
2927
2928 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2929 const struct path *path)
2930 {
2931 struct inode *inode;
2932 int err;
2933
2934 /* if quota sysfile exists, deny enabling quota with specific file */
2935 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2936 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2937 return -EBUSY;
2938 }
2939
2940 if (path->dentry->d_sb != sb)
2941 return -EXDEV;
2942
2943 err = f2fs_quota_sync(sb, type);
2944 if (err)
2945 return err;
2946
2947 inode = d_inode(path->dentry);
2948
2949 err = filemap_fdatawrite(inode->i_mapping);
2950 if (err)
2951 return err;
2952
2953 err = filemap_fdatawait(inode->i_mapping);
2954 if (err)
2955 return err;
2956
2957 err = dquot_quota_on(sb, type, format_id, path);
2958 if (err)
2959 return err;
2960
2961 inode_lock(inode);
2962 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2963 f2fs_set_inode_flags(inode);
2964 inode_unlock(inode);
2965 f2fs_mark_inode_dirty_sync(inode, false);
2966
2967 return 0;
2968 }
2969
2970 static int __f2fs_quota_off(struct super_block *sb, int type)
2971 {
2972 struct inode *inode = sb_dqopt(sb)->files[type];
2973 int err;
2974
2975 if (!inode || !igrab(inode))
2976 return dquot_quota_off(sb, type);
2977
2978 err = f2fs_quota_sync(sb, type);
2979 if (err)
2980 goto out_put;
2981
2982 err = dquot_quota_off(sb, type);
2983 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2984 goto out_put;
2985
2986 inode_lock(inode);
2987 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
2988 f2fs_set_inode_flags(inode);
2989 inode_unlock(inode);
2990 f2fs_mark_inode_dirty_sync(inode, false);
2991 out_put:
2992 iput(inode);
2993 return err;
2994 }
2995
2996 static int f2fs_quota_off(struct super_block *sb, int type)
2997 {
2998 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2999 int err;
3000
3001 err = __f2fs_quota_off(sb, type);
3002
3003 /*
3004 * quotactl can shutdown journalled quota, result in inconsistence
3005 * between quota record and fs data by following updates, tag the
3006 * flag to let fsck be aware of it.
3007 */
3008 if (is_journalled_quota(sbi))
3009 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3010 return err;
3011 }
3012
3013 void f2fs_quota_off_umount(struct super_block *sb)
3014 {
3015 int type;
3016 int err;
3017
3018 for (type = 0; type < MAXQUOTAS; type++) {
3019 err = __f2fs_quota_off(sb, type);
3020 if (err) {
3021 int ret = dquot_quota_off(sb, type);
3022
3023 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3024 type, err, ret);
3025 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3026 }
3027 }
3028 /*
3029 * In case of checkpoint=disable, we must flush quota blocks.
3030 * This can cause NULL exception for node_inode in end_io, since
3031 * put_super already dropped it.
3032 */
3033 sync_filesystem(sb);
3034 }
3035
3036 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3037 {
3038 struct quota_info *dqopt = sb_dqopt(sb);
3039 int type;
3040
3041 for (type = 0; type < MAXQUOTAS; type++) {
3042 if (!dqopt->files[type])
3043 continue;
3044 f2fs_inode_synced(dqopt->files[type]);
3045 }
3046 }
3047
3048 static int f2fs_dquot_commit(struct dquot *dquot)
3049 {
3050 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3051 int ret;
3052
3053 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3054 ret = dquot_commit(dquot);
3055 if (ret < 0)
3056 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3057 f2fs_up_read(&sbi->quota_sem);
3058 return ret;
3059 }
3060
3061 static int f2fs_dquot_acquire(struct dquot *dquot)
3062 {
3063 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3064 int ret;
3065
3066 f2fs_down_read(&sbi->quota_sem);
3067 ret = dquot_acquire(dquot);
3068 if (ret < 0)
3069 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3070 f2fs_up_read(&sbi->quota_sem);
3071 return ret;
3072 }
3073
3074 static int f2fs_dquot_release(struct dquot *dquot)
3075 {
3076 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3077 int ret = dquot_release(dquot);
3078
3079 if (ret < 0)
3080 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3081 return ret;
3082 }
3083
3084 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3085 {
3086 struct super_block *sb = dquot->dq_sb;
3087 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3088 int ret = dquot_mark_dquot_dirty(dquot);
3089
3090 /* if we are using journalled quota */
3091 if (is_journalled_quota(sbi))
3092 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3093
3094 return ret;
3095 }
3096
3097 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3098 {
3099 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3100 int ret = dquot_commit_info(sb, type);
3101
3102 if (ret < 0)
3103 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3104 return ret;
3105 }
3106
3107 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3108 {
3109 *projid = F2FS_I(inode)->i_projid;
3110 return 0;
3111 }
3112
3113 static const struct dquot_operations f2fs_quota_operations = {
3114 .get_reserved_space = f2fs_get_reserved_space,
3115 .write_dquot = f2fs_dquot_commit,
3116 .acquire_dquot = f2fs_dquot_acquire,
3117 .release_dquot = f2fs_dquot_release,
3118 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3119 .write_info = f2fs_dquot_commit_info,
3120 .alloc_dquot = dquot_alloc,
3121 .destroy_dquot = dquot_destroy,
3122 .get_projid = f2fs_get_projid,
3123 .get_next_id = dquot_get_next_id,
3124 };
3125
3126 static const struct quotactl_ops f2fs_quotactl_ops = {
3127 .quota_on = f2fs_quota_on,
3128 .quota_off = f2fs_quota_off,
3129 .quota_sync = f2fs_quota_sync,
3130 .get_state = dquot_get_state,
3131 .set_info = dquot_set_dqinfo,
3132 .get_dqblk = dquot_get_dqblk,
3133 .set_dqblk = dquot_set_dqblk,
3134 .get_nextdqblk = dquot_get_next_dqblk,
3135 };
3136 #else
3137 int f2fs_dquot_initialize(struct inode *inode)
3138 {
3139 return 0;
3140 }
3141
3142 int f2fs_quota_sync(struct super_block *sb, int type)
3143 {
3144 return 0;
3145 }
3146
3147 void f2fs_quota_off_umount(struct super_block *sb)
3148 {
3149 }
3150 #endif
3151
3152 static const struct super_operations f2fs_sops = {
3153 .alloc_inode = f2fs_alloc_inode,
3154 .free_inode = f2fs_free_inode,
3155 .drop_inode = f2fs_drop_inode,
3156 .write_inode = f2fs_write_inode,
3157 .dirty_inode = f2fs_dirty_inode,
3158 .show_options = f2fs_show_options,
3159 #ifdef CONFIG_QUOTA
3160 .quota_read = f2fs_quota_read,
3161 .quota_write = f2fs_quota_write,
3162 .get_dquots = f2fs_get_dquots,
3163 #endif
3164 .evict_inode = f2fs_evict_inode,
3165 .put_super = f2fs_put_super,
3166 .sync_fs = f2fs_sync_fs,
3167 .freeze_fs = f2fs_freeze,
3168 .unfreeze_fs = f2fs_unfreeze,
3169 .statfs = f2fs_statfs,
3170 .remount_fs = f2fs_remount,
3171 .shutdown = f2fs_shutdown,
3172 };
3173
3174 #ifdef CONFIG_FS_ENCRYPTION
3175 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3176 {
3177 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3178 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3179 ctx, len, NULL);
3180 }
3181
3182 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3183 void *fs_data)
3184 {
3185 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3186
3187 /*
3188 * Encrypting the root directory is not allowed because fsck
3189 * expects lost+found directory to exist and remain unencrypted
3190 * if LOST_FOUND feature is enabled.
3191 *
3192 */
3193 if (f2fs_sb_has_lost_found(sbi) &&
3194 inode->i_ino == F2FS_ROOT_INO(sbi))
3195 return -EPERM;
3196
3197 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3198 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3199 ctx, len, fs_data, XATTR_CREATE);
3200 }
3201
3202 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3203 {
3204 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3205 }
3206
3207 static bool f2fs_has_stable_inodes(struct super_block *sb)
3208 {
3209 return true;
3210 }
3211
3212 static struct block_device **f2fs_get_devices(struct super_block *sb,
3213 unsigned int *num_devs)
3214 {
3215 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3216 struct block_device **devs;
3217 int i;
3218
3219 if (!f2fs_is_multi_device(sbi))
3220 return NULL;
3221
3222 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3223 if (!devs)
3224 return ERR_PTR(-ENOMEM);
3225
3226 for (i = 0; i < sbi->s_ndevs; i++)
3227 devs[i] = FDEV(i).bdev;
3228 *num_devs = sbi->s_ndevs;
3229 return devs;
3230 }
3231
3232 static const struct fscrypt_operations f2fs_cryptops = {
3233 .needs_bounce_pages = 1,
3234 .has_32bit_inodes = 1,
3235 .supports_subblock_data_units = 1,
3236 .legacy_key_prefix = "f2fs:",
3237 .get_context = f2fs_get_context,
3238 .set_context = f2fs_set_context,
3239 .get_dummy_policy = f2fs_get_dummy_policy,
3240 .empty_dir = f2fs_empty_dir,
3241 .has_stable_inodes = f2fs_has_stable_inodes,
3242 .get_devices = f2fs_get_devices,
3243 };
3244 #endif
3245
3246 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3247 u64 ino, u32 generation)
3248 {
3249 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3250 struct inode *inode;
3251
3252 if (f2fs_check_nid_range(sbi, ino))
3253 return ERR_PTR(-ESTALE);
3254
3255 /*
3256 * f2fs_iget isn't quite right if the inode is currently unallocated!
3257 * However f2fs_iget currently does appropriate checks to handle stale
3258 * inodes so everything is OK.
3259 */
3260 inode = f2fs_iget(sb, ino);
3261 if (IS_ERR(inode))
3262 return ERR_CAST(inode);
3263 if (unlikely(generation && inode->i_generation != generation)) {
3264 /* we didn't find the right inode.. */
3265 iput(inode);
3266 return ERR_PTR(-ESTALE);
3267 }
3268 return inode;
3269 }
3270
3271 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3272 int fh_len, int fh_type)
3273 {
3274 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3275 f2fs_nfs_get_inode);
3276 }
3277
3278 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3279 int fh_len, int fh_type)
3280 {
3281 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3282 f2fs_nfs_get_inode);
3283 }
3284
3285 static const struct export_operations f2fs_export_ops = {
3286 .encode_fh = generic_encode_ino32_fh,
3287 .fh_to_dentry = f2fs_fh_to_dentry,
3288 .fh_to_parent = f2fs_fh_to_parent,
3289 .get_parent = f2fs_get_parent,
3290 };
3291
3292 loff_t max_file_blocks(struct inode *inode)
3293 {
3294 loff_t result = 0;
3295 loff_t leaf_count;
3296
3297 /*
3298 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3299 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3300 * space in inode.i_addr, it will be more safe to reassign
3301 * result as zero.
3302 */
3303
3304 if (inode && f2fs_compressed_file(inode))
3305 leaf_count = ADDRS_PER_BLOCK(inode);
3306 else
3307 leaf_count = DEF_ADDRS_PER_BLOCK;
3308
3309 /* two direct node blocks */
3310 result += (leaf_count * 2);
3311
3312 /* two indirect node blocks */
3313 leaf_count *= NIDS_PER_BLOCK;
3314 result += (leaf_count * 2);
3315
3316 /* one double indirect node block */
3317 leaf_count *= NIDS_PER_BLOCK;
3318 result += leaf_count;
3319
3320 /*
3321 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with
3322 * a 4K crypto data unit, we must restrict the max filesize to what can
3323 * fit within U32_MAX + 1 data units.
3324 */
3325
3326 result = min(result, (((loff_t)U32_MAX + 1) * 4096) >> F2FS_BLKSIZE_BITS);
3327
3328 return result;
3329 }
3330
3331 static int __f2fs_commit_super(struct buffer_head *bh,
3332 struct f2fs_super_block *super)
3333 {
3334 lock_buffer(bh);
3335 if (super)
3336 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3337 set_buffer_dirty(bh);
3338 unlock_buffer(bh);
3339
3340 /* it's rare case, we can do fua all the time */
3341 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3342 }
3343
3344 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3345 struct buffer_head *bh)
3346 {
3347 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3348 (bh->b_data + F2FS_SUPER_OFFSET);
3349 struct super_block *sb = sbi->sb;
3350 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3351 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3352 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3353 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3354 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3355 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3356 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3357 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3358 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3359 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3360 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3361 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3362 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3363 u64 main_end_blkaddr = main_blkaddr +
3364 ((u64)segment_count_main << log_blocks_per_seg);
3365 u64 seg_end_blkaddr = segment0_blkaddr +
3366 ((u64)segment_count << log_blocks_per_seg);
3367
3368 if (segment0_blkaddr != cp_blkaddr) {
3369 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3370 segment0_blkaddr, cp_blkaddr);
3371 return true;
3372 }
3373
3374 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3375 sit_blkaddr) {
3376 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3377 cp_blkaddr, sit_blkaddr,
3378 segment_count_ckpt << log_blocks_per_seg);
3379 return true;
3380 }
3381
3382 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3383 nat_blkaddr) {
3384 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3385 sit_blkaddr, nat_blkaddr,
3386 segment_count_sit << log_blocks_per_seg);
3387 return true;
3388 }
3389
3390 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3391 ssa_blkaddr) {
3392 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3393 nat_blkaddr, ssa_blkaddr,
3394 segment_count_nat << log_blocks_per_seg);
3395 return true;
3396 }
3397
3398 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3399 main_blkaddr) {
3400 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3401 ssa_blkaddr, main_blkaddr,
3402 segment_count_ssa << log_blocks_per_seg);
3403 return true;
3404 }
3405
3406 if (main_end_blkaddr > seg_end_blkaddr) {
3407 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3408 main_blkaddr, seg_end_blkaddr,
3409 segment_count_main << log_blocks_per_seg);
3410 return true;
3411 } else if (main_end_blkaddr < seg_end_blkaddr) {
3412 int err = 0;
3413 char *res;
3414
3415 /* fix in-memory information all the time */
3416 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3417 segment0_blkaddr) >> log_blocks_per_seg);
3418
3419 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3420 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3421 res = "internally";
3422 } else {
3423 err = __f2fs_commit_super(bh, NULL);
3424 res = err ? "failed" : "done";
3425 }
3426 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3427 res, main_blkaddr, seg_end_blkaddr,
3428 segment_count_main << log_blocks_per_seg);
3429 if (err)
3430 return true;
3431 }
3432 return false;
3433 }
3434
3435 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3436 struct buffer_head *bh)
3437 {
3438 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3439 block_t total_sections, blocks_per_seg;
3440 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3441 (bh->b_data + F2FS_SUPER_OFFSET);
3442 size_t crc_offset = 0;
3443 __u32 crc = 0;
3444
3445 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3446 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3447 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3448 return -EINVAL;
3449 }
3450
3451 /* Check checksum_offset and crc in superblock */
3452 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3453 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3454 if (crc_offset !=
3455 offsetof(struct f2fs_super_block, crc)) {
3456 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3457 crc_offset);
3458 return -EFSCORRUPTED;
3459 }
3460 crc = le32_to_cpu(raw_super->crc);
3461 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3462 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3463 return -EFSCORRUPTED;
3464 }
3465 }
3466
3467 /* only support block_size equals to PAGE_SIZE */
3468 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3469 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3470 le32_to_cpu(raw_super->log_blocksize),
3471 F2FS_BLKSIZE_BITS);
3472 return -EFSCORRUPTED;
3473 }
3474
3475 /* check log blocks per segment */
3476 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3477 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3478 le32_to_cpu(raw_super->log_blocks_per_seg));
3479 return -EFSCORRUPTED;
3480 }
3481
3482 /* Currently, support 512/1024/2048/4096/16K bytes sector size */
3483 if (le32_to_cpu(raw_super->log_sectorsize) >
3484 F2FS_MAX_LOG_SECTOR_SIZE ||
3485 le32_to_cpu(raw_super->log_sectorsize) <
3486 F2FS_MIN_LOG_SECTOR_SIZE) {
3487 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3488 le32_to_cpu(raw_super->log_sectorsize));
3489 return -EFSCORRUPTED;
3490 }
3491 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3492 le32_to_cpu(raw_super->log_sectorsize) !=
3493 F2FS_MAX_LOG_SECTOR_SIZE) {
3494 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3495 le32_to_cpu(raw_super->log_sectors_per_block),
3496 le32_to_cpu(raw_super->log_sectorsize));
3497 return -EFSCORRUPTED;
3498 }
3499
3500 segment_count = le32_to_cpu(raw_super->segment_count);
3501 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3502 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3503 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3504 total_sections = le32_to_cpu(raw_super->section_count);
3505
3506 /* blocks_per_seg should be 512, given the above check */
3507 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3508
3509 if (segment_count > F2FS_MAX_SEGMENT ||
3510 segment_count < F2FS_MIN_SEGMENTS) {
3511 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3512 return -EFSCORRUPTED;
3513 }
3514
3515 if (total_sections > segment_count_main || total_sections < 1 ||
3516 segs_per_sec > segment_count || !segs_per_sec) {
3517 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3518 segment_count, total_sections, segs_per_sec);
3519 return -EFSCORRUPTED;
3520 }
3521
3522 if (segment_count_main != total_sections * segs_per_sec) {
3523 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3524 segment_count_main, total_sections, segs_per_sec);
3525 return -EFSCORRUPTED;
3526 }
3527
3528 if ((segment_count / segs_per_sec) < total_sections) {
3529 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3530 segment_count, segs_per_sec, total_sections);
3531 return -EFSCORRUPTED;
3532 }
3533
3534 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3535 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3536 segment_count, le64_to_cpu(raw_super->block_count));
3537 return -EFSCORRUPTED;
3538 }
3539
3540 if (RDEV(0).path[0]) {
3541 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3542 int i = 1;
3543
3544 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3545 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3546 i++;
3547 }
3548 if (segment_count != dev_seg_count) {
3549 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3550 segment_count, dev_seg_count);
3551 return -EFSCORRUPTED;
3552 }
3553 } else {
3554 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3555 !bdev_is_zoned(sbi->sb->s_bdev)) {
3556 f2fs_info(sbi, "Zoned block device path is missing");
3557 return -EFSCORRUPTED;
3558 }
3559 }
3560
3561 if (secs_per_zone > total_sections || !secs_per_zone) {
3562 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3563 secs_per_zone, total_sections);
3564 return -EFSCORRUPTED;
3565 }
3566 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3567 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3568 (le32_to_cpu(raw_super->extension_count) +
3569 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3570 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3571 le32_to_cpu(raw_super->extension_count),
3572 raw_super->hot_ext_count,
3573 F2FS_MAX_EXTENSION);
3574 return -EFSCORRUPTED;
3575 }
3576
3577 if (le32_to_cpu(raw_super->cp_payload) >=
3578 (blocks_per_seg - F2FS_CP_PACKS -
3579 NR_CURSEG_PERSIST_TYPE)) {
3580 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3581 le32_to_cpu(raw_super->cp_payload),
3582 blocks_per_seg - F2FS_CP_PACKS -
3583 NR_CURSEG_PERSIST_TYPE);
3584 return -EFSCORRUPTED;
3585 }
3586
3587 /* check reserved ino info */
3588 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3589 le32_to_cpu(raw_super->meta_ino) != 2 ||
3590 le32_to_cpu(raw_super->root_ino) != 3) {
3591 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3592 le32_to_cpu(raw_super->node_ino),
3593 le32_to_cpu(raw_super->meta_ino),
3594 le32_to_cpu(raw_super->root_ino));
3595 return -EFSCORRUPTED;
3596 }
3597
3598 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3599 if (sanity_check_area_boundary(sbi, bh))
3600 return -EFSCORRUPTED;
3601
3602 return 0;
3603 }
3604
3605 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3606 {
3607 unsigned int total, fsmeta;
3608 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3609 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3610 unsigned int ovp_segments, reserved_segments;
3611 unsigned int main_segs, blocks_per_seg;
3612 unsigned int sit_segs, nat_segs;
3613 unsigned int sit_bitmap_size, nat_bitmap_size;
3614 unsigned int log_blocks_per_seg;
3615 unsigned int segment_count_main;
3616 unsigned int cp_pack_start_sum, cp_payload;
3617 block_t user_block_count, valid_user_blocks;
3618 block_t avail_node_count, valid_node_count;
3619 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3620 int i, j;
3621
3622 total = le32_to_cpu(raw_super->segment_count);
3623 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3624 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3625 fsmeta += sit_segs;
3626 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3627 fsmeta += nat_segs;
3628 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3629 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3630
3631 if (unlikely(fsmeta >= total))
3632 return 1;
3633
3634 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3635 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3636
3637 if (!f2fs_sb_has_readonly(sbi) &&
3638 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3639 ovp_segments == 0 || reserved_segments == 0)) {
3640 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3641 return 1;
3642 }
3643 user_block_count = le64_to_cpu(ckpt->user_block_count);
3644 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3645 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3646 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3647 if (!user_block_count || user_block_count >=
3648 segment_count_main << log_blocks_per_seg) {
3649 f2fs_err(sbi, "Wrong user_block_count: %u",
3650 user_block_count);
3651 return 1;
3652 }
3653
3654 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3655 if (valid_user_blocks > user_block_count) {
3656 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3657 valid_user_blocks, user_block_count);
3658 return 1;
3659 }
3660
3661 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3662 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3663 if (valid_node_count > avail_node_count) {
3664 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3665 valid_node_count, avail_node_count);
3666 return 1;
3667 }
3668
3669 main_segs = le32_to_cpu(raw_super->segment_count_main);
3670 blocks_per_seg = BLKS_PER_SEG(sbi);
3671
3672 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3673 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3674 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3675 return 1;
3676
3677 if (f2fs_sb_has_readonly(sbi))
3678 goto check_data;
3679
3680 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3681 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3682 le32_to_cpu(ckpt->cur_node_segno[j])) {
3683 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3684 i, j,
3685 le32_to_cpu(ckpt->cur_node_segno[i]));
3686 return 1;
3687 }
3688 }
3689 }
3690 check_data:
3691 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3692 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3693 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3694 return 1;
3695
3696 if (f2fs_sb_has_readonly(sbi))
3697 goto skip_cross;
3698
3699 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3700 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3701 le32_to_cpu(ckpt->cur_data_segno[j])) {
3702 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3703 i, j,
3704 le32_to_cpu(ckpt->cur_data_segno[i]));
3705 return 1;
3706 }
3707 }
3708 }
3709 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3710 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3711 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3712 le32_to_cpu(ckpt->cur_data_segno[j])) {
3713 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3714 i, j,
3715 le32_to_cpu(ckpt->cur_node_segno[i]));
3716 return 1;
3717 }
3718 }
3719 }
3720 skip_cross:
3721 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3722 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3723
3724 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3725 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3726 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3727 sit_bitmap_size, nat_bitmap_size);
3728 return 1;
3729 }
3730
3731 cp_pack_start_sum = __start_sum_addr(sbi);
3732 cp_payload = __cp_payload(sbi);
3733 if (cp_pack_start_sum < cp_payload + 1 ||
3734 cp_pack_start_sum > blocks_per_seg - 1 -
3735 NR_CURSEG_PERSIST_TYPE) {
3736 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3737 cp_pack_start_sum);
3738 return 1;
3739 }
3740
3741 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3742 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3743 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3744 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3745 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3746 le32_to_cpu(ckpt->checksum_offset));
3747 return 1;
3748 }
3749
3750 nat_blocks = nat_segs << log_blocks_per_seg;
3751 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3752 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3753 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3754 (cp_payload + F2FS_CP_PACKS +
3755 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3756 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3757 cp_payload, nat_bits_blocks);
3758 return 1;
3759 }
3760
3761 if (unlikely(f2fs_cp_error(sbi))) {
3762 f2fs_err(sbi, "A bug case: need to run fsck");
3763 return 1;
3764 }
3765 return 0;
3766 }
3767
3768 static void init_sb_info(struct f2fs_sb_info *sbi)
3769 {
3770 struct f2fs_super_block *raw_super = sbi->raw_super;
3771 int i;
3772
3773 sbi->log_sectors_per_block =
3774 le32_to_cpu(raw_super->log_sectors_per_block);
3775 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3776 sbi->blocksize = BIT(sbi->log_blocksize);
3777 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3778 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3779 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3780 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3781 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3782 sbi->total_node_count = SEGS_TO_BLKS(sbi,
3783 ((le32_to_cpu(raw_super->segment_count_nat) / 2) *
3784 NAT_ENTRY_PER_BLOCK));
3785 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3786 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3787 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3788 sbi->cur_victim_sec = NULL_SECNO;
3789 sbi->gc_mode = GC_NORMAL;
3790 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3791 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3792 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3793 sbi->migration_granularity = SEGS_PER_SEC(sbi);
3794 sbi->migration_window_granularity = f2fs_sb_has_blkzoned(sbi) ?
3795 DEF_MIGRATION_WINDOW_GRANULARITY_ZONED : SEGS_PER_SEC(sbi);
3796 sbi->seq_file_ra_mul = MIN_RA_MUL;
3797 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3798 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3799 spin_lock_init(&sbi->gc_remaining_trials_lock);
3800 atomic64_set(&sbi->current_atomic_write, 0);
3801
3802 sbi->dir_level = DEF_DIR_LEVEL;
3803 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3804 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3805 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3806 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3807 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3808 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3809 DEF_UMOUNT_DISCARD_TIMEOUT;
3810 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3811
3812 for (i = 0; i < NR_COUNT_TYPE; i++)
3813 atomic_set(&sbi->nr_pages[i], 0);
3814
3815 for (i = 0; i < META; i++)
3816 atomic_set(&sbi->wb_sync_req[i], 0);
3817
3818 INIT_LIST_HEAD(&sbi->s_list);
3819 mutex_init(&sbi->umount_mutex);
3820 init_f2fs_rwsem(&sbi->io_order_lock);
3821 spin_lock_init(&sbi->cp_lock);
3822
3823 sbi->dirty_device = 0;
3824 spin_lock_init(&sbi->dev_lock);
3825
3826 init_f2fs_rwsem(&sbi->sb_lock);
3827 init_f2fs_rwsem(&sbi->pin_sem);
3828 }
3829
3830 static int init_percpu_info(struct f2fs_sb_info *sbi)
3831 {
3832 int err;
3833
3834 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3835 if (err)
3836 return err;
3837
3838 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3839 if (err)
3840 goto err_valid_block;
3841
3842 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3843 GFP_KERNEL);
3844 if (err)
3845 goto err_node_block;
3846 return 0;
3847
3848 err_node_block:
3849 percpu_counter_destroy(&sbi->rf_node_block_count);
3850 err_valid_block:
3851 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3852 return err;
3853 }
3854
3855 #ifdef CONFIG_BLK_DEV_ZONED
3856
3857 struct f2fs_report_zones_args {
3858 struct f2fs_sb_info *sbi;
3859 struct f2fs_dev_info *dev;
3860 };
3861
3862 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3863 void *data)
3864 {
3865 struct f2fs_report_zones_args *rz_args = data;
3866 block_t unusable_blocks = (zone->len - zone->capacity) >>
3867 F2FS_LOG_SECTORS_PER_BLOCK;
3868
3869 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3870 return 0;
3871
3872 set_bit(idx, rz_args->dev->blkz_seq);
3873 if (!rz_args->sbi->unusable_blocks_per_sec) {
3874 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3875 return 0;
3876 }
3877 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3878 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3879 return -EINVAL;
3880 }
3881 return 0;
3882 }
3883
3884 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3885 {
3886 struct block_device *bdev = FDEV(devi).bdev;
3887 sector_t nr_sectors = bdev_nr_sectors(bdev);
3888 struct f2fs_report_zones_args rep_zone_arg;
3889 u64 zone_sectors;
3890 unsigned int max_open_zones;
3891 int ret;
3892
3893 if (!f2fs_sb_has_blkzoned(sbi))
3894 return 0;
3895
3896 if (bdev_is_zoned(FDEV(devi).bdev)) {
3897 max_open_zones = bdev_max_open_zones(bdev);
3898 if (max_open_zones && (max_open_zones < sbi->max_open_zones))
3899 sbi->max_open_zones = max_open_zones;
3900 if (sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
3901 f2fs_err(sbi,
3902 "zoned: max open zones %u is too small, need at least %u open zones",
3903 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
3904 return -EINVAL;
3905 }
3906 }
3907
3908 zone_sectors = bdev_zone_sectors(bdev);
3909 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3910 SECTOR_TO_BLOCK(zone_sectors))
3911 return -EINVAL;
3912 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3913 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3914 sbi->blocks_per_blkz);
3915 if (nr_sectors & (zone_sectors - 1))
3916 FDEV(devi).nr_blkz++;
3917
3918 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3919 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3920 * sizeof(unsigned long),
3921 GFP_KERNEL);
3922 if (!FDEV(devi).blkz_seq)
3923 return -ENOMEM;
3924
3925 rep_zone_arg.sbi = sbi;
3926 rep_zone_arg.dev = &FDEV(devi);
3927
3928 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3929 &rep_zone_arg);
3930 if (ret < 0)
3931 return ret;
3932 return 0;
3933 }
3934 #endif
3935
3936 /*
3937 * Read f2fs raw super block.
3938 * Because we have two copies of super block, so read both of them
3939 * to get the first valid one. If any one of them is broken, we pass
3940 * them recovery flag back to the caller.
3941 */
3942 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3943 struct f2fs_super_block **raw_super,
3944 int *valid_super_block, int *recovery)
3945 {
3946 struct super_block *sb = sbi->sb;
3947 int block;
3948 struct buffer_head *bh;
3949 struct f2fs_super_block *super;
3950 int err = 0;
3951
3952 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3953 if (!super)
3954 return -ENOMEM;
3955
3956 for (block = 0; block < 2; block++) {
3957 bh = sb_bread(sb, block);
3958 if (!bh) {
3959 f2fs_err(sbi, "Unable to read %dth superblock",
3960 block + 1);
3961 err = -EIO;
3962 *recovery = 1;
3963 continue;
3964 }
3965
3966 /* sanity checking of raw super */
3967 err = sanity_check_raw_super(sbi, bh);
3968 if (err) {
3969 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3970 block + 1);
3971 brelse(bh);
3972 *recovery = 1;
3973 continue;
3974 }
3975
3976 if (!*raw_super) {
3977 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3978 sizeof(*super));
3979 *valid_super_block = block;
3980 *raw_super = super;
3981 }
3982 brelse(bh);
3983 }
3984
3985 /* No valid superblock */
3986 if (!*raw_super)
3987 kfree(super);
3988 else
3989 err = 0;
3990
3991 return err;
3992 }
3993
3994 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3995 {
3996 struct buffer_head *bh;
3997 __u32 crc = 0;
3998 int err;
3999
4000 if ((recover && f2fs_readonly(sbi->sb)) ||
4001 f2fs_hw_is_readonly(sbi)) {
4002 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
4003 return -EROFS;
4004 }
4005
4006 /* we should update superblock crc here */
4007 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
4008 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
4009 offsetof(struct f2fs_super_block, crc));
4010 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
4011 }
4012
4013 /* write back-up superblock first */
4014 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
4015 if (!bh)
4016 return -EIO;
4017 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4018 brelse(bh);
4019
4020 /* if we are in recovery path, skip writing valid superblock */
4021 if (recover || err)
4022 return err;
4023
4024 /* write current valid superblock */
4025 bh = sb_bread(sbi->sb, sbi->valid_super_block);
4026 if (!bh)
4027 return -EIO;
4028 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4029 brelse(bh);
4030 return err;
4031 }
4032
4033 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4034 {
4035 unsigned long flags;
4036
4037 spin_lock_irqsave(&sbi->error_lock, flags);
4038 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4039 sbi->stop_reason[reason]++;
4040 spin_unlock_irqrestore(&sbi->error_lock, flags);
4041 }
4042
4043 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4044 {
4045 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4046 unsigned long flags;
4047 int err;
4048
4049 f2fs_down_write(&sbi->sb_lock);
4050
4051 spin_lock_irqsave(&sbi->error_lock, flags);
4052 if (sbi->error_dirty) {
4053 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4054 MAX_F2FS_ERRORS);
4055 sbi->error_dirty = false;
4056 }
4057 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4058 spin_unlock_irqrestore(&sbi->error_lock, flags);
4059
4060 err = f2fs_commit_super(sbi, false);
4061
4062 f2fs_up_write(&sbi->sb_lock);
4063 if (err)
4064 f2fs_err_ratelimited(sbi,
4065 "f2fs_commit_super fails to record stop_reason, err:%d",
4066 err);
4067 }
4068
4069 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4070 {
4071 unsigned long flags;
4072
4073 spin_lock_irqsave(&sbi->error_lock, flags);
4074 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4075 set_bit(flag, (unsigned long *)sbi->errors);
4076 sbi->error_dirty = true;
4077 }
4078 spin_unlock_irqrestore(&sbi->error_lock, flags);
4079 }
4080
4081 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4082 {
4083 unsigned long flags;
4084 bool need_update = false;
4085
4086 spin_lock_irqsave(&sbi->error_lock, flags);
4087 if (sbi->error_dirty) {
4088 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4089 MAX_F2FS_ERRORS);
4090 sbi->error_dirty = false;
4091 need_update = true;
4092 }
4093 spin_unlock_irqrestore(&sbi->error_lock, flags);
4094
4095 return need_update;
4096 }
4097
4098 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4099 {
4100 int err;
4101
4102 f2fs_down_write(&sbi->sb_lock);
4103
4104 if (!f2fs_update_errors(sbi))
4105 goto out_unlock;
4106
4107 err = f2fs_commit_super(sbi, false);
4108 if (err)
4109 f2fs_err_ratelimited(sbi,
4110 "f2fs_commit_super fails to record errors:%u, err:%d",
4111 error, err);
4112 out_unlock:
4113 f2fs_up_write(&sbi->sb_lock);
4114 }
4115
4116 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4117 {
4118 f2fs_save_errors(sbi, error);
4119 f2fs_record_errors(sbi, error);
4120 }
4121
4122 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4123 {
4124 f2fs_save_errors(sbi, error);
4125
4126 if (!sbi->error_dirty)
4127 return;
4128 if (!test_bit(error, (unsigned long *)sbi->errors))
4129 return;
4130 schedule_work(&sbi->s_error_work);
4131 }
4132
4133 static bool system_going_down(void)
4134 {
4135 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4136 || system_state == SYSTEM_RESTART;
4137 }
4138
4139 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason,
4140 bool irq_context)
4141 {
4142 struct super_block *sb = sbi->sb;
4143 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4144 bool continue_fs = !shutdown &&
4145 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4146
4147 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4148
4149 if (!f2fs_hw_is_readonly(sbi)) {
4150 save_stop_reason(sbi, reason);
4151
4152 if (irq_context && !shutdown)
4153 schedule_work(&sbi->s_error_work);
4154 else
4155 f2fs_record_stop_reason(sbi);
4156 }
4157
4158 /*
4159 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4160 * could panic during 'reboot -f' as the underlying device got already
4161 * disabled.
4162 */
4163 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4164 !shutdown && !system_going_down() &&
4165 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4166 panic("F2FS-fs (device %s): panic forced after error\n",
4167 sb->s_id);
4168
4169 if (shutdown)
4170 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4171
4172 /*
4173 * Continue filesystem operators if errors=continue. Should not set
4174 * RO by shutdown, since RO bypasses thaw_super which can hang the
4175 * system.
4176 */
4177 if (continue_fs || f2fs_readonly(sb) || shutdown) {
4178 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason);
4179 return;
4180 }
4181
4182 f2fs_warn(sbi, "Remounting filesystem read-only");
4183
4184 /*
4185 * We have already set CP_ERROR_FLAG flag to stop all updates
4186 * to filesystem, so it doesn't need to set SB_RDONLY flag here
4187 * because the flag should be set covered w/ sb->s_umount semaphore
4188 * via remount procedure, otherwise, it will confuse code like
4189 * freeze_super() which will lead to deadlocks and other problems.
4190 */
4191 }
4192
4193 static void f2fs_record_error_work(struct work_struct *work)
4194 {
4195 struct f2fs_sb_info *sbi = container_of(work,
4196 struct f2fs_sb_info, s_error_work);
4197
4198 f2fs_record_stop_reason(sbi);
4199 }
4200
4201 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4202 {
4203 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4204 unsigned int max_devices = MAX_DEVICES;
4205 unsigned int logical_blksize;
4206 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4207 int i;
4208
4209 /* Initialize single device information */
4210 if (!RDEV(0).path[0]) {
4211 if (!bdev_is_zoned(sbi->sb->s_bdev))
4212 return 0;
4213 max_devices = 1;
4214 }
4215
4216 /*
4217 * Initialize multiple devices information, or single
4218 * zoned block device information.
4219 */
4220 sbi->devs = f2fs_kzalloc(sbi,
4221 array_size(max_devices,
4222 sizeof(struct f2fs_dev_info)),
4223 GFP_KERNEL);
4224 if (!sbi->devs)
4225 return -ENOMEM;
4226
4227 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4228 sbi->aligned_blksize = true;
4229 #ifdef CONFIG_BLK_DEV_ZONED
4230 sbi->max_open_zones = UINT_MAX;
4231 sbi->blkzone_alloc_policy = BLKZONE_ALLOC_PRIOR_SEQ;
4232 #endif
4233
4234 for (i = 0; i < max_devices; i++) {
4235 if (i == 0)
4236 FDEV(0).bdev_file = sbi->sb->s_bdev_file;
4237 else if (!RDEV(i).path[0])
4238 break;
4239
4240 if (max_devices > 1) {
4241 /* Multi-device mount */
4242 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4243 FDEV(i).total_segments =
4244 le32_to_cpu(RDEV(i).total_segments);
4245 if (i == 0) {
4246 FDEV(i).start_blk = 0;
4247 FDEV(i).end_blk = FDEV(i).start_blk +
4248 SEGS_TO_BLKS(sbi,
4249 FDEV(i).total_segments) - 1 +
4250 le32_to_cpu(raw_super->segment0_blkaddr);
4251 } else {
4252 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4253 FDEV(i).end_blk = FDEV(i).start_blk +
4254 SEGS_TO_BLKS(sbi,
4255 FDEV(i).total_segments) - 1;
4256 FDEV(i).bdev_file = bdev_file_open_by_path(
4257 FDEV(i).path, mode, sbi->sb, NULL);
4258 }
4259 }
4260 if (IS_ERR(FDEV(i).bdev_file))
4261 return PTR_ERR(FDEV(i).bdev_file);
4262
4263 FDEV(i).bdev = file_bdev(FDEV(i).bdev_file);
4264 /* to release errored devices */
4265 sbi->s_ndevs = i + 1;
4266
4267 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4268 sbi->aligned_blksize = false;
4269
4270 #ifdef CONFIG_BLK_DEV_ZONED
4271 if (bdev_is_zoned(FDEV(i).bdev)) {
4272 if (!f2fs_sb_has_blkzoned(sbi)) {
4273 f2fs_err(sbi, "Zoned block device feature not enabled");
4274 return -EINVAL;
4275 }
4276 if (init_blkz_info(sbi, i)) {
4277 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4278 return -EINVAL;
4279 }
4280 if (max_devices == 1)
4281 break;
4282 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)",
4283 i, FDEV(i).path,
4284 FDEV(i).total_segments,
4285 FDEV(i).start_blk, FDEV(i).end_blk);
4286 continue;
4287 }
4288 #endif
4289 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4290 i, FDEV(i).path,
4291 FDEV(i).total_segments,
4292 FDEV(i).start_blk, FDEV(i).end_blk);
4293 }
4294 return 0;
4295 }
4296
4297 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4298 {
4299 #if IS_ENABLED(CONFIG_UNICODE)
4300 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4301 const struct f2fs_sb_encodings *encoding_info;
4302 struct unicode_map *encoding;
4303 __u16 encoding_flags;
4304
4305 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4306 if (!encoding_info) {
4307 f2fs_err(sbi,
4308 "Encoding requested by superblock is unknown");
4309 return -EINVAL;
4310 }
4311
4312 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4313 encoding = utf8_load(encoding_info->version);
4314 if (IS_ERR(encoding)) {
4315 f2fs_err(sbi,
4316 "can't mount with superblock charset: %s-%u.%u.%u "
4317 "not supported by the kernel. flags: 0x%x.",
4318 encoding_info->name,
4319 unicode_major(encoding_info->version),
4320 unicode_minor(encoding_info->version),
4321 unicode_rev(encoding_info->version),
4322 encoding_flags);
4323 return PTR_ERR(encoding);
4324 }
4325 f2fs_info(sbi, "Using encoding defined by superblock: "
4326 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4327 unicode_major(encoding_info->version),
4328 unicode_minor(encoding_info->version),
4329 unicode_rev(encoding_info->version),
4330 encoding_flags);
4331
4332 sbi->sb->s_encoding = encoding;
4333 sbi->sb->s_encoding_flags = encoding_flags;
4334 }
4335 #else
4336 if (f2fs_sb_has_casefold(sbi)) {
4337 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4338 return -EINVAL;
4339 }
4340 #endif
4341 return 0;
4342 }
4343
4344 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4345 {
4346 /* adjust parameters according to the volume size */
4347 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4348 if (f2fs_block_unit_discard(sbi))
4349 SM_I(sbi)->dcc_info->discard_granularity =
4350 MIN_DISCARD_GRANULARITY;
4351 if (!f2fs_lfs_mode(sbi))
4352 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4353 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4354 }
4355
4356 sbi->readdir_ra = true;
4357 }
4358
4359 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4360 {
4361 struct f2fs_sb_info *sbi;
4362 struct f2fs_super_block *raw_super;
4363 struct inode *root;
4364 int err;
4365 bool skip_recovery = false, need_fsck = false;
4366 char *options = NULL;
4367 int recovery, i, valid_super_block;
4368 struct curseg_info *seg_i;
4369 int retry_cnt = 1;
4370 #ifdef CONFIG_QUOTA
4371 bool quota_enabled = false;
4372 #endif
4373
4374 try_onemore:
4375 err = -EINVAL;
4376 raw_super = NULL;
4377 valid_super_block = -1;
4378 recovery = 0;
4379
4380 /* allocate memory for f2fs-specific super block info */
4381 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4382 if (!sbi)
4383 return -ENOMEM;
4384
4385 sbi->sb = sb;
4386
4387 /* initialize locks within allocated memory */
4388 init_f2fs_rwsem(&sbi->gc_lock);
4389 mutex_init(&sbi->writepages);
4390 init_f2fs_rwsem(&sbi->cp_global_sem);
4391 init_f2fs_rwsem(&sbi->node_write);
4392 init_f2fs_rwsem(&sbi->node_change);
4393 spin_lock_init(&sbi->stat_lock);
4394 init_f2fs_rwsem(&sbi->cp_rwsem);
4395 init_f2fs_rwsem(&sbi->quota_sem);
4396 init_waitqueue_head(&sbi->cp_wait);
4397 spin_lock_init(&sbi->error_lock);
4398
4399 for (i = 0; i < NR_INODE_TYPE; i++) {
4400 INIT_LIST_HEAD(&sbi->inode_list[i]);
4401 spin_lock_init(&sbi->inode_lock[i]);
4402 }
4403 mutex_init(&sbi->flush_lock);
4404
4405 /* Load the checksum driver */
4406 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4407 if (IS_ERR(sbi->s_chksum_driver)) {
4408 f2fs_err(sbi, "Cannot load crc32 driver.");
4409 err = PTR_ERR(sbi->s_chksum_driver);
4410 sbi->s_chksum_driver = NULL;
4411 goto free_sbi;
4412 }
4413
4414 /* set a block size */
4415 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4416 f2fs_err(sbi, "unable to set blocksize");
4417 goto free_sbi;
4418 }
4419
4420 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4421 &recovery);
4422 if (err)
4423 goto free_sbi;
4424
4425 sb->s_fs_info = sbi;
4426 sbi->raw_super = raw_super;
4427
4428 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4429 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4430 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4431
4432 /* precompute checksum seed for metadata */
4433 if (f2fs_sb_has_inode_chksum(sbi))
4434 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4435 sizeof(raw_super->uuid));
4436
4437 default_options(sbi, false);
4438 /* parse mount options */
4439 options = kstrdup((const char *)data, GFP_KERNEL);
4440 if (data && !options) {
4441 err = -ENOMEM;
4442 goto free_sb_buf;
4443 }
4444
4445 err = parse_options(sb, options, false);
4446 if (err)
4447 goto free_options;
4448
4449 sb->s_maxbytes = max_file_blocks(NULL) <<
4450 le32_to_cpu(raw_super->log_blocksize);
4451 sb->s_max_links = F2FS_LINK_MAX;
4452
4453 err = f2fs_setup_casefold(sbi);
4454 if (err)
4455 goto free_options;
4456
4457 #ifdef CONFIG_QUOTA
4458 sb->dq_op = &f2fs_quota_operations;
4459 sb->s_qcop = &f2fs_quotactl_ops;
4460 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4461
4462 if (f2fs_sb_has_quota_ino(sbi)) {
4463 for (i = 0; i < MAXQUOTAS; i++) {
4464 if (f2fs_qf_ino(sbi->sb, i))
4465 sbi->nquota_files++;
4466 }
4467 }
4468 #endif
4469
4470 sb->s_op = &f2fs_sops;
4471 #ifdef CONFIG_FS_ENCRYPTION
4472 sb->s_cop = &f2fs_cryptops;
4473 #endif
4474 #ifdef CONFIG_FS_VERITY
4475 sb->s_vop = &f2fs_verityops;
4476 #endif
4477 sb->s_xattr = f2fs_xattr_handlers;
4478 sb->s_export_op = &f2fs_export_ops;
4479 sb->s_magic = F2FS_SUPER_MAGIC;
4480 sb->s_time_gran = 1;
4481 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4482 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4483 super_set_uuid(sb, (void *) raw_super->uuid, sizeof(raw_super->uuid));
4484 super_set_sysfs_name_bdev(sb);
4485 sb->s_iflags |= SB_I_CGROUPWB;
4486
4487 /* init f2fs-specific super block info */
4488 sbi->valid_super_block = valid_super_block;
4489
4490 /* disallow all the data/node/meta page writes */
4491 set_sbi_flag(sbi, SBI_POR_DOING);
4492
4493 err = f2fs_init_write_merge_io(sbi);
4494 if (err)
4495 goto free_bio_info;
4496
4497 init_sb_info(sbi);
4498
4499 err = f2fs_init_iostat(sbi);
4500 if (err)
4501 goto free_bio_info;
4502
4503 err = init_percpu_info(sbi);
4504 if (err)
4505 goto free_iostat;
4506
4507 /* init per sbi slab cache */
4508 err = f2fs_init_xattr_caches(sbi);
4509 if (err)
4510 goto free_percpu;
4511 err = f2fs_init_page_array_cache(sbi);
4512 if (err)
4513 goto free_xattr_cache;
4514
4515 /* get an inode for meta space */
4516 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4517 if (IS_ERR(sbi->meta_inode)) {
4518 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4519 err = PTR_ERR(sbi->meta_inode);
4520 goto free_page_array_cache;
4521 }
4522
4523 err = f2fs_get_valid_checkpoint(sbi);
4524 if (err) {
4525 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4526 goto free_meta_inode;
4527 }
4528
4529 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4530 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4531 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4532 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4533 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4534 }
4535
4536 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4537 set_sbi_flag(sbi, SBI_NEED_FSCK);
4538
4539 /* Initialize device list */
4540 err = f2fs_scan_devices(sbi);
4541 if (err) {
4542 f2fs_err(sbi, "Failed to find devices");
4543 goto free_devices;
4544 }
4545
4546 err = f2fs_init_post_read_wq(sbi);
4547 if (err) {
4548 f2fs_err(sbi, "Failed to initialize post read workqueue");
4549 goto free_devices;
4550 }
4551
4552 sbi->total_valid_node_count =
4553 le32_to_cpu(sbi->ckpt->valid_node_count);
4554 percpu_counter_set(&sbi->total_valid_inode_count,
4555 le32_to_cpu(sbi->ckpt->valid_inode_count));
4556 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4557 sbi->total_valid_block_count =
4558 le64_to_cpu(sbi->ckpt->valid_block_count);
4559 sbi->last_valid_block_count = sbi->total_valid_block_count;
4560 sbi->reserved_blocks = 0;
4561 sbi->current_reserved_blocks = 0;
4562 limit_reserve_root(sbi);
4563 adjust_unusable_cap_perc(sbi);
4564
4565 f2fs_init_extent_cache_info(sbi);
4566
4567 f2fs_init_ino_entry_info(sbi);
4568
4569 f2fs_init_fsync_node_info(sbi);
4570
4571 /* setup checkpoint request control and start checkpoint issue thread */
4572 f2fs_init_ckpt_req_control(sbi);
4573 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4574 test_opt(sbi, MERGE_CHECKPOINT)) {
4575 err = f2fs_start_ckpt_thread(sbi);
4576 if (err) {
4577 f2fs_err(sbi,
4578 "Failed to start F2FS issue_checkpoint_thread (%d)",
4579 err);
4580 goto stop_ckpt_thread;
4581 }
4582 }
4583
4584 /* setup f2fs internal modules */
4585 err = f2fs_build_segment_manager(sbi);
4586 if (err) {
4587 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4588 err);
4589 goto free_sm;
4590 }
4591 err = f2fs_build_node_manager(sbi);
4592 if (err) {
4593 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4594 err);
4595 goto free_nm;
4596 }
4597
4598 /* For write statistics */
4599 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4600
4601 /* Read accumulated write IO statistics if exists */
4602 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4603 if (__exist_node_summaries(sbi))
4604 sbi->kbytes_written =
4605 le64_to_cpu(seg_i->journal->info.kbytes_written);
4606
4607 f2fs_build_gc_manager(sbi);
4608
4609 err = f2fs_build_stats(sbi);
4610 if (err)
4611 goto free_nm;
4612
4613 /* get an inode for node space */
4614 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4615 if (IS_ERR(sbi->node_inode)) {
4616 f2fs_err(sbi, "Failed to read node inode");
4617 err = PTR_ERR(sbi->node_inode);
4618 goto free_stats;
4619 }
4620
4621 /* read root inode and dentry */
4622 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4623 if (IS_ERR(root)) {
4624 f2fs_err(sbi, "Failed to read root inode");
4625 err = PTR_ERR(root);
4626 goto free_node_inode;
4627 }
4628 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4629 !root->i_size || !root->i_nlink) {
4630 iput(root);
4631 err = -EINVAL;
4632 goto free_node_inode;
4633 }
4634
4635 generic_set_sb_d_ops(sb);
4636 sb->s_root = d_make_root(root); /* allocate root dentry */
4637 if (!sb->s_root) {
4638 err = -ENOMEM;
4639 goto free_node_inode;
4640 }
4641
4642 err = f2fs_init_compress_inode(sbi);
4643 if (err)
4644 goto free_root_inode;
4645
4646 err = f2fs_register_sysfs(sbi);
4647 if (err)
4648 goto free_compress_inode;
4649
4650 #ifdef CONFIG_QUOTA
4651 /* Enable quota usage during mount */
4652 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4653 err = f2fs_enable_quotas(sb);
4654 if (err)
4655 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4656 }
4657
4658 quota_enabled = f2fs_recover_quota_begin(sbi);
4659 #endif
4660 /* if there are any orphan inodes, free them */
4661 err = f2fs_recover_orphan_inodes(sbi);
4662 if (err)
4663 goto free_meta;
4664
4665 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4666 goto reset_checkpoint;
4667
4668 /* recover fsynced data */
4669 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4670 !test_opt(sbi, NORECOVERY)) {
4671 /*
4672 * mount should be failed, when device has readonly mode, and
4673 * previous checkpoint was not done by clean system shutdown.
4674 */
4675 if (f2fs_hw_is_readonly(sbi)) {
4676 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4677 err = f2fs_recover_fsync_data(sbi, true);
4678 if (err > 0) {
4679 err = -EROFS;
4680 f2fs_err(sbi, "Need to recover fsync data, but "
4681 "write access unavailable, please try "
4682 "mount w/ disable_roll_forward or norecovery");
4683 }
4684 if (err < 0)
4685 goto free_meta;
4686 }
4687 f2fs_info(sbi, "write access unavailable, skipping recovery");
4688 goto reset_checkpoint;
4689 }
4690
4691 if (need_fsck)
4692 set_sbi_flag(sbi, SBI_NEED_FSCK);
4693
4694 if (skip_recovery)
4695 goto reset_checkpoint;
4696
4697 err = f2fs_recover_fsync_data(sbi, false);
4698 if (err < 0) {
4699 if (err != -ENOMEM)
4700 skip_recovery = true;
4701 need_fsck = true;
4702 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4703 err);
4704 goto free_meta;
4705 }
4706 } else {
4707 err = f2fs_recover_fsync_data(sbi, true);
4708
4709 if (!f2fs_readonly(sb) && err > 0) {
4710 err = -EINVAL;
4711 f2fs_err(sbi, "Need to recover fsync data");
4712 goto free_meta;
4713 }
4714 }
4715
4716 #ifdef CONFIG_QUOTA
4717 f2fs_recover_quota_end(sbi, quota_enabled);
4718 #endif
4719 reset_checkpoint:
4720 /*
4721 * If the f2fs is not readonly and fsync data recovery succeeds,
4722 * check zoned block devices' write pointer consistency.
4723 */
4724 if (f2fs_sb_has_blkzoned(sbi) && !f2fs_readonly(sb)) {
4725 int err2;
4726
4727 f2fs_notice(sbi, "Checking entire write pointers");
4728 err2 = f2fs_check_write_pointer(sbi);
4729 if (err2)
4730 err = err2;
4731 }
4732 if (err)
4733 goto free_meta;
4734
4735 err = f2fs_init_inmem_curseg(sbi);
4736 if (err)
4737 goto sync_free_meta;
4738
4739 /* f2fs_recover_fsync_data() cleared this already */
4740 clear_sbi_flag(sbi, SBI_POR_DOING);
4741
4742 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4743 err = f2fs_disable_checkpoint(sbi);
4744 if (err)
4745 goto sync_free_meta;
4746 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4747 f2fs_enable_checkpoint(sbi);
4748 }
4749
4750 /*
4751 * If filesystem is not mounted as read-only then
4752 * do start the gc_thread.
4753 */
4754 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4755 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4756 /* After POR, we can run background GC thread.*/
4757 err = f2fs_start_gc_thread(sbi);
4758 if (err)
4759 goto sync_free_meta;
4760 }
4761 kvfree(options);
4762
4763 /* recover broken superblock */
4764 if (recovery) {
4765 err = f2fs_commit_super(sbi, true);
4766 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4767 sbi->valid_super_block ? 1 : 2, err);
4768 }
4769
4770 f2fs_join_shrinker(sbi);
4771
4772 f2fs_tuning_parameters(sbi);
4773
4774 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4775 cur_cp_version(F2FS_CKPT(sbi)));
4776 f2fs_update_time(sbi, CP_TIME);
4777 f2fs_update_time(sbi, REQ_TIME);
4778 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4779 return 0;
4780
4781 sync_free_meta:
4782 /* safe to flush all the data */
4783 sync_filesystem(sbi->sb);
4784 retry_cnt = 0;
4785
4786 free_meta:
4787 #ifdef CONFIG_QUOTA
4788 f2fs_truncate_quota_inode_pages(sb);
4789 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4790 f2fs_quota_off_umount(sbi->sb);
4791 #endif
4792 /*
4793 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4794 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4795 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4796 * falls into an infinite loop in f2fs_sync_meta_pages().
4797 */
4798 truncate_inode_pages_final(META_MAPPING(sbi));
4799 /* evict some inodes being cached by GC */
4800 evict_inodes(sb);
4801 f2fs_unregister_sysfs(sbi);
4802 free_compress_inode:
4803 f2fs_destroy_compress_inode(sbi);
4804 free_root_inode:
4805 dput(sb->s_root);
4806 sb->s_root = NULL;
4807 free_node_inode:
4808 f2fs_release_ino_entry(sbi, true);
4809 truncate_inode_pages_final(NODE_MAPPING(sbi));
4810 iput(sbi->node_inode);
4811 sbi->node_inode = NULL;
4812 free_stats:
4813 f2fs_destroy_stats(sbi);
4814 free_nm:
4815 /* stop discard thread before destroying node manager */
4816 f2fs_stop_discard_thread(sbi);
4817 f2fs_destroy_node_manager(sbi);
4818 free_sm:
4819 f2fs_destroy_segment_manager(sbi);
4820 stop_ckpt_thread:
4821 f2fs_stop_ckpt_thread(sbi);
4822 /* flush s_error_work before sbi destroy */
4823 flush_work(&sbi->s_error_work);
4824 f2fs_destroy_post_read_wq(sbi);
4825 free_devices:
4826 destroy_device_list(sbi);
4827 kvfree(sbi->ckpt);
4828 free_meta_inode:
4829 make_bad_inode(sbi->meta_inode);
4830 iput(sbi->meta_inode);
4831 sbi->meta_inode = NULL;
4832 free_page_array_cache:
4833 f2fs_destroy_page_array_cache(sbi);
4834 free_xattr_cache:
4835 f2fs_destroy_xattr_caches(sbi);
4836 free_percpu:
4837 destroy_percpu_info(sbi);
4838 free_iostat:
4839 f2fs_destroy_iostat(sbi);
4840 free_bio_info:
4841 for (i = 0; i < NR_PAGE_TYPE; i++)
4842 kvfree(sbi->write_io[i]);
4843
4844 #if IS_ENABLED(CONFIG_UNICODE)
4845 utf8_unload(sb->s_encoding);
4846 sb->s_encoding = NULL;
4847 #endif
4848 free_options:
4849 #ifdef CONFIG_QUOTA
4850 for (i = 0; i < MAXQUOTAS; i++)
4851 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4852 #endif
4853 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4854 kvfree(options);
4855 free_sb_buf:
4856 kfree(raw_super);
4857 free_sbi:
4858 if (sbi->s_chksum_driver)
4859 crypto_free_shash(sbi->s_chksum_driver);
4860 kfree(sbi);
4861 sb->s_fs_info = NULL;
4862
4863 /* give only one another chance */
4864 if (retry_cnt > 0 && skip_recovery) {
4865 retry_cnt--;
4866 shrink_dcache_sb(sb);
4867 goto try_onemore;
4868 }
4869 return err;
4870 }
4871
4872 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4873 const char *dev_name, void *data)
4874 {
4875 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4876 }
4877
4878 static void kill_f2fs_super(struct super_block *sb)
4879 {
4880 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4881
4882 if (sb->s_root) {
4883 set_sbi_flag(sbi, SBI_IS_CLOSE);
4884 f2fs_stop_gc_thread(sbi);
4885 f2fs_stop_discard_thread(sbi);
4886
4887 #ifdef CONFIG_F2FS_FS_COMPRESSION
4888 /*
4889 * latter evict_inode() can bypass checking and invalidating
4890 * compress inode cache.
4891 */
4892 if (test_opt(sbi, COMPRESS_CACHE))
4893 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4894 #endif
4895
4896 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4897 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4898 struct cp_control cpc = {
4899 .reason = CP_UMOUNT,
4900 };
4901 stat_inc_cp_call_count(sbi, TOTAL_CALL);
4902 f2fs_write_checkpoint(sbi, &cpc);
4903 }
4904
4905 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4906 sb->s_flags &= ~SB_RDONLY;
4907 }
4908 kill_block_super(sb);
4909 /* Release block devices last, after fscrypt_destroy_keyring(). */
4910 if (sbi) {
4911 destroy_device_list(sbi);
4912 kfree(sbi);
4913 sb->s_fs_info = NULL;
4914 }
4915 }
4916
4917 static struct file_system_type f2fs_fs_type = {
4918 .owner = THIS_MODULE,
4919 .name = "f2fs",
4920 .mount = f2fs_mount,
4921 .kill_sb = kill_f2fs_super,
4922 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4923 };
4924 MODULE_ALIAS_FS("f2fs");
4925
4926 static int __init init_inodecache(void)
4927 {
4928 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4929 sizeof(struct f2fs_inode_info), 0,
4930 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4931 return f2fs_inode_cachep ? 0 : -ENOMEM;
4932 }
4933
4934 static void destroy_inodecache(void)
4935 {
4936 /*
4937 * Make sure all delayed rcu free inodes are flushed before we
4938 * destroy cache.
4939 */
4940 rcu_barrier();
4941 kmem_cache_destroy(f2fs_inode_cachep);
4942 }
4943
4944 static int __init init_f2fs_fs(void)
4945 {
4946 int err;
4947
4948 err = init_inodecache();
4949 if (err)
4950 goto fail;
4951 err = f2fs_create_node_manager_caches();
4952 if (err)
4953 goto free_inodecache;
4954 err = f2fs_create_segment_manager_caches();
4955 if (err)
4956 goto free_node_manager_caches;
4957 err = f2fs_create_checkpoint_caches();
4958 if (err)
4959 goto free_segment_manager_caches;
4960 err = f2fs_create_recovery_cache();
4961 if (err)
4962 goto free_checkpoint_caches;
4963 err = f2fs_create_extent_cache();
4964 if (err)
4965 goto free_recovery_cache;
4966 err = f2fs_create_garbage_collection_cache();
4967 if (err)
4968 goto free_extent_cache;
4969 err = f2fs_init_sysfs();
4970 if (err)
4971 goto free_garbage_collection_cache;
4972 err = f2fs_init_shrinker();
4973 if (err)
4974 goto free_sysfs;
4975 err = register_filesystem(&f2fs_fs_type);
4976 if (err)
4977 goto free_shrinker;
4978 f2fs_create_root_stats();
4979 err = f2fs_init_post_read_processing();
4980 if (err)
4981 goto free_root_stats;
4982 err = f2fs_init_iostat_processing();
4983 if (err)
4984 goto free_post_read;
4985 err = f2fs_init_bio_entry_cache();
4986 if (err)
4987 goto free_iostat;
4988 err = f2fs_init_bioset();
4989 if (err)
4990 goto free_bio_entry_cache;
4991 err = f2fs_init_compress_mempool();
4992 if (err)
4993 goto free_bioset;
4994 err = f2fs_init_compress_cache();
4995 if (err)
4996 goto free_compress_mempool;
4997 err = f2fs_create_casefold_cache();
4998 if (err)
4999 goto free_compress_cache;
5000 return 0;
5001 free_compress_cache:
5002 f2fs_destroy_compress_cache();
5003 free_compress_mempool:
5004 f2fs_destroy_compress_mempool();
5005 free_bioset:
5006 f2fs_destroy_bioset();
5007 free_bio_entry_cache:
5008 f2fs_destroy_bio_entry_cache();
5009 free_iostat:
5010 f2fs_destroy_iostat_processing();
5011 free_post_read:
5012 f2fs_destroy_post_read_processing();
5013 free_root_stats:
5014 f2fs_destroy_root_stats();
5015 unregister_filesystem(&f2fs_fs_type);
5016 free_shrinker:
5017 f2fs_exit_shrinker();
5018 free_sysfs:
5019 f2fs_exit_sysfs();
5020 free_garbage_collection_cache:
5021 f2fs_destroy_garbage_collection_cache();
5022 free_extent_cache:
5023 f2fs_destroy_extent_cache();
5024 free_recovery_cache:
5025 f2fs_destroy_recovery_cache();
5026 free_checkpoint_caches:
5027 f2fs_destroy_checkpoint_caches();
5028 free_segment_manager_caches:
5029 f2fs_destroy_segment_manager_caches();
5030 free_node_manager_caches:
5031 f2fs_destroy_node_manager_caches();
5032 free_inodecache:
5033 destroy_inodecache();
5034 fail:
5035 return err;
5036 }
5037
5038 static void __exit exit_f2fs_fs(void)
5039 {
5040 f2fs_destroy_casefold_cache();
5041 f2fs_destroy_compress_cache();
5042 f2fs_destroy_compress_mempool();
5043 f2fs_destroy_bioset();
5044 f2fs_destroy_bio_entry_cache();
5045 f2fs_destroy_iostat_processing();
5046 f2fs_destroy_post_read_processing();
5047 f2fs_destroy_root_stats();
5048 unregister_filesystem(&f2fs_fs_type);
5049 f2fs_exit_shrinker();
5050 f2fs_exit_sysfs();
5051 f2fs_destroy_garbage_collection_cache();
5052 f2fs_destroy_extent_cache();
5053 f2fs_destroy_recovery_cache();
5054 f2fs_destroy_checkpoint_caches();
5055 f2fs_destroy_segment_manager_caches();
5056 f2fs_destroy_node_manager_caches();
5057 destroy_inodecache();
5058 }
5059
5060 module_init(init_f2fs_fs)
5061 module_exit(exit_f2fs_fs)
5062
5063 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5064 MODULE_DESCRIPTION("Flash Friendly File System");
5065 MODULE_LICENSE("GPL");
5066 MODULE_SOFTDEP("pre: crc32");
5067
5068
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.