1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * fs/f2fs/super.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8 #include <linux/module.h>
9 #include <linux/init.h>
10 #include <linux/fs.h>
11 #include <linux/fs_context.h>
12 #include <linux/sched/mm.h>
13 #include <linux/statfs.h>
14 #include <linux/kthread.h>
15 #include <linux/parser.h>
16 #include <linux/mount.h>
17 #include <linux/seq_file.h>
18 #include <linux/proc_fs.h>
19 #include <linux/random.h>
20 #include <linux/exportfs.h>
21 #include <linux/blkdev.h>
22 #include <linux/quotaops.h>
23 #include <linux/f2fs_fs.h>
24 #include <linux/sysfs.h>
25 #include <linux/quota.h>
26 #include <linux/unicode.h>
27 #include <linux/part_stat.h>
28 #include <linux/zstd.h>
29 #include <linux/lz4.h>
30
31 #include "f2fs.h"
32 #include "node.h"
33 #include "segment.h"
34 #include "xattr.h"
35 #include "gc.h"
36 #include "iostat.h"
37
38 #define CREATE_TRACE_POINTS
39 #include <trace/events/f2fs.h>
40
41 static struct kmem_cache *f2fs_inode_cachep;
42
43 #ifdef CONFIG_F2FS_FAULT_INJECTION
44
45 const char *f2fs_fault_name[FAULT_MAX] = {
46 [FAULT_KMALLOC] = "kmalloc",
47 [FAULT_KVMALLOC] = "kvmalloc",
48 [FAULT_PAGE_ALLOC] = "page alloc",
49 [FAULT_PAGE_GET] = "page get",
50 [FAULT_ALLOC_NID] = "alloc nid",
51 [FAULT_ORPHAN] = "orphan",
52 [FAULT_BLOCK] = "no more block",
53 [FAULT_DIR_DEPTH] = "too big dir depth",
54 [FAULT_EVICT_INODE] = "evict_inode fail",
55 [FAULT_TRUNCATE] = "truncate fail",
56 [FAULT_READ_IO] = "read IO error",
57 [FAULT_CHECKPOINT] = "checkpoint error",
58 [FAULT_DISCARD] = "discard error",
59 [FAULT_WRITE_IO] = "write IO error",
60 [FAULT_SLAB_ALLOC] = "slab alloc",
61 [FAULT_DQUOT_INIT] = "dquot initialize",
62 [FAULT_LOCK_OP] = "lock_op",
63 [FAULT_BLKADDR_VALIDITY] = "invalid blkaddr",
64 [FAULT_BLKADDR_CONSISTENCE] = "inconsistent blkaddr",
65 [FAULT_NO_SEGMENT] = "no free segment",
66 [FAULT_INCONSISTENT_FOOTER] = "inconsistent footer",
67 };
68
f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned long rate,unsigned long type)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
f2fs_init_shrinker(void)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
f2fs_exit_shrinker(void)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_lazytime,
155 Opt_nolazytime,
156 Opt_quota,
157 Opt_noquota,
158 Opt_usrquota,
159 Opt_grpquota,
160 Opt_prjquota,
161 Opt_usrjquota,
162 Opt_grpjquota,
163 Opt_prjjquota,
164 Opt_offusrjquota,
165 Opt_offgrpjquota,
166 Opt_offprjjquota,
167 Opt_jqfmt_vfsold,
168 Opt_jqfmt_vfsv0,
169 Opt_jqfmt_vfsv1,
170 Opt_alloc,
171 Opt_fsync,
172 Opt_test_dummy_encryption,
173 Opt_inlinecrypt,
174 Opt_checkpoint_disable,
175 Opt_checkpoint_disable_cap,
176 Opt_checkpoint_disable_cap_perc,
177 Opt_checkpoint_enable,
178 Opt_checkpoint_merge,
179 Opt_nocheckpoint_merge,
180 Opt_compress_algorithm,
181 Opt_compress_log_size,
182 Opt_compress_extension,
183 Opt_nocompress_extension,
184 Opt_compress_chksum,
185 Opt_compress_mode,
186 Opt_compress_cache,
187 Opt_atgc,
188 Opt_gc_merge,
189 Opt_nogc_merge,
190 Opt_discard_unit,
191 Opt_memory_mode,
192 Opt_age_extent_cache,
193 Opt_errors,
194 Opt_nat_bits,
195 Opt_err,
196 };
197
198 static match_table_t f2fs_tokens = {
199 {Opt_gc_background, "background_gc=%s"},
200 {Opt_disable_roll_forward, "disable_roll_forward"},
201 {Opt_norecovery, "norecovery"},
202 {Opt_discard, "discard"},
203 {Opt_nodiscard, "nodiscard"},
204 {Opt_noheap, "no_heap"},
205 {Opt_heap, "heap"},
206 {Opt_user_xattr, "user_xattr"},
207 {Opt_nouser_xattr, "nouser_xattr"},
208 {Opt_acl, "acl"},
209 {Opt_noacl, "noacl"},
210 {Opt_active_logs, "active_logs=%u"},
211 {Opt_disable_ext_identify, "disable_ext_identify"},
212 {Opt_inline_xattr, "inline_xattr"},
213 {Opt_noinline_xattr, "noinline_xattr"},
214 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
215 {Opt_inline_data, "inline_data"},
216 {Opt_inline_dentry, "inline_dentry"},
217 {Opt_noinline_dentry, "noinline_dentry"},
218 {Opt_flush_merge, "flush_merge"},
219 {Opt_noflush_merge, "noflush_merge"},
220 {Opt_barrier, "barrier"},
221 {Opt_nobarrier, "nobarrier"},
222 {Opt_fastboot, "fastboot"},
223 {Opt_extent_cache, "extent_cache"},
224 {Opt_noextent_cache, "noextent_cache"},
225 {Opt_noinline_data, "noinline_data"},
226 {Opt_data_flush, "data_flush"},
227 {Opt_reserve_root, "reserve_root=%u"},
228 {Opt_resgid, "resgid=%u"},
229 {Opt_resuid, "resuid=%u"},
230 {Opt_mode, "mode=%s"},
231 {Opt_fault_injection, "fault_injection=%u"},
232 {Opt_fault_type, "fault_type=%u"},
233 {Opt_lazytime, "lazytime"},
234 {Opt_nolazytime, "nolazytime"},
235 {Opt_quota, "quota"},
236 {Opt_noquota, "noquota"},
237 {Opt_usrquota, "usrquota"},
238 {Opt_grpquota, "grpquota"},
239 {Opt_prjquota, "prjquota"},
240 {Opt_usrjquota, "usrjquota=%s"},
241 {Opt_grpjquota, "grpjquota=%s"},
242 {Opt_prjjquota, "prjjquota=%s"},
243 {Opt_offusrjquota, "usrjquota="},
244 {Opt_offgrpjquota, "grpjquota="},
245 {Opt_offprjjquota, "prjjquota="},
246 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
247 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
248 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
249 {Opt_alloc, "alloc_mode=%s"},
250 {Opt_fsync, "fsync_mode=%s"},
251 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
252 {Opt_test_dummy_encryption, "test_dummy_encryption"},
253 {Opt_inlinecrypt, "inlinecrypt"},
254 {Opt_checkpoint_disable, "checkpoint=disable"},
255 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
256 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
257 {Opt_checkpoint_enable, "checkpoint=enable"},
258 {Opt_checkpoint_merge, "checkpoint_merge"},
259 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
260 {Opt_compress_algorithm, "compress_algorithm=%s"},
261 {Opt_compress_log_size, "compress_log_size=%u"},
262 {Opt_compress_extension, "compress_extension=%s"},
263 {Opt_nocompress_extension, "nocompress_extension=%s"},
264 {Opt_compress_chksum, "compress_chksum"},
265 {Opt_compress_mode, "compress_mode=%s"},
266 {Opt_compress_cache, "compress_cache"},
267 {Opt_atgc, "atgc"},
268 {Opt_gc_merge, "gc_merge"},
269 {Opt_nogc_merge, "nogc_merge"},
270 {Opt_discard_unit, "discard_unit=%s"},
271 {Opt_memory_mode, "memory=%s"},
272 {Opt_age_extent_cache, "age_extent_cache"},
273 {Opt_errors, "errors=%s"},
274 {Opt_nat_bits, "nat_bits"},
275 {Opt_err, NULL},
276 };
277
f2fs_printk(struct f2fs_sb_info * sbi,bool limit_rate,const char * fmt,...)278 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate,
279 const char *fmt, ...)
280 {
281 struct va_format vaf;
282 va_list args;
283 int level;
284
285 va_start(args, fmt);
286
287 level = printk_get_level(fmt);
288 vaf.fmt = printk_skip_level(fmt);
289 vaf.va = &args;
290 if (limit_rate)
291 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n",
292 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
293 else
294 printk("%c%cF2FS-fs (%s): %pV\n",
295 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
296
297 va_end(args);
298 }
299
300 #if IS_ENABLED(CONFIG_UNICODE)
301 static const struct f2fs_sb_encodings {
302 __u16 magic;
303 char *name;
304 unsigned int version;
305 } f2fs_sb_encoding_map[] = {
306 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
307 };
308
309 static const struct f2fs_sb_encodings *
f2fs_sb_read_encoding(const struct f2fs_super_block * sb)310 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
311 {
312 __u16 magic = le16_to_cpu(sb->s_encoding);
313 int i;
314
315 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
316 if (magic == f2fs_sb_encoding_map[i].magic)
317 return &f2fs_sb_encoding_map[i];
318
319 return NULL;
320 }
321
322 struct kmem_cache *f2fs_cf_name_slab;
f2fs_create_casefold_cache(void)323 static int __init f2fs_create_casefold_cache(void)
324 {
325 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
326 F2FS_NAME_LEN);
327 return f2fs_cf_name_slab ? 0 : -ENOMEM;
328 }
329
f2fs_destroy_casefold_cache(void)330 static void f2fs_destroy_casefold_cache(void)
331 {
332 kmem_cache_destroy(f2fs_cf_name_slab);
333 }
334 #else
f2fs_create_casefold_cache(void)335 static int __init f2fs_create_casefold_cache(void) { return 0; }
f2fs_destroy_casefold_cache(void)336 static void f2fs_destroy_casefold_cache(void) { }
337 #endif
338
limit_reserve_root(struct f2fs_sb_info * sbi)339 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
340 {
341 block_t limit = min((sbi->user_block_count >> 3),
342 sbi->user_block_count - sbi->reserved_blocks);
343
344 /* limit is 12.5% */
345 if (test_opt(sbi, RESERVE_ROOT) &&
346 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
347 F2FS_OPTION(sbi).root_reserved_blocks = limit;
348 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
349 F2FS_OPTION(sbi).root_reserved_blocks);
350 }
351 if (!test_opt(sbi, RESERVE_ROOT) &&
352 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
353 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
354 !gid_eq(F2FS_OPTION(sbi).s_resgid,
355 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
356 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
357 from_kuid_munged(&init_user_ns,
358 F2FS_OPTION(sbi).s_resuid),
359 from_kgid_munged(&init_user_ns,
360 F2FS_OPTION(sbi).s_resgid));
361 }
362
adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)363 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
364 {
365 if (!F2FS_OPTION(sbi).unusable_cap_perc)
366 return;
367
368 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
369 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
370 else
371 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
372 F2FS_OPTION(sbi).unusable_cap_perc;
373
374 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
375 F2FS_OPTION(sbi).unusable_cap,
376 F2FS_OPTION(sbi).unusable_cap_perc);
377 }
378
init_once(void * foo)379 static void init_once(void *foo)
380 {
381 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
382
383 inode_init_once(&fi->vfs_inode);
384 }
385
386 #ifdef CONFIG_QUOTA
387 static const char * const quotatypes[] = INITQFNAMES;
388 #define QTYPE2NAME(t) (quotatypes[t])
f2fs_set_qf_name(struct f2fs_sb_info * sbi,int qtype,substring_t * args)389 static int f2fs_set_qf_name(struct f2fs_sb_info *sbi, int qtype,
390 substring_t *args)
391 {
392 struct super_block *sb = sbi->sb;
393 char *qname;
394 int ret = -EINVAL;
395
396 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
397 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
398 return -EINVAL;
399 }
400 if (f2fs_sb_has_quota_ino(sbi)) {
401 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
402 return 0;
403 }
404
405 qname = match_strdup(args);
406 if (!qname) {
407 f2fs_err(sbi, "Not enough memory for storing quotafile name");
408 return -ENOMEM;
409 }
410 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
411 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
412 ret = 0;
413 else
414 f2fs_err(sbi, "%s quota file already specified",
415 QTYPE2NAME(qtype));
416 goto errout;
417 }
418 if (strchr(qname, '/')) {
419 f2fs_err(sbi, "quotafile must be on filesystem root");
420 goto errout;
421 }
422 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
423 set_opt(sbi, QUOTA);
424 return 0;
425 errout:
426 kfree(qname);
427 return ret;
428 }
429
f2fs_clear_qf_name(struct f2fs_sb_info * sbi,int qtype)430 static int f2fs_clear_qf_name(struct f2fs_sb_info *sbi, int qtype)
431 {
432 struct super_block *sb = sbi->sb;
433
434 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
435 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
436 return -EINVAL;
437 }
438 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
439 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
440 return 0;
441 }
442
f2fs_check_quota_options(struct f2fs_sb_info * sbi)443 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
444 {
445 /*
446 * We do the test below only for project quotas. 'usrquota' and
447 * 'grpquota' mount options are allowed even without quota feature
448 * to support legacy quotas in quota files.
449 */
450 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
451 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
452 return -1;
453 }
454 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
455 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
456 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
457 if (test_opt(sbi, USRQUOTA) &&
458 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
459 clear_opt(sbi, USRQUOTA);
460
461 if (test_opt(sbi, GRPQUOTA) &&
462 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
463 clear_opt(sbi, GRPQUOTA);
464
465 if (test_opt(sbi, PRJQUOTA) &&
466 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
467 clear_opt(sbi, PRJQUOTA);
468
469 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
470 test_opt(sbi, PRJQUOTA)) {
471 f2fs_err(sbi, "old and new quota format mixing");
472 return -1;
473 }
474
475 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
476 f2fs_err(sbi, "journaled quota format not specified");
477 return -1;
478 }
479 }
480
481 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
482 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
483 F2FS_OPTION(sbi).s_jquota_fmt = 0;
484 }
485 return 0;
486 }
487 #endif
488
f2fs_set_test_dummy_encryption(struct f2fs_sb_info * sbi,const char * opt,const substring_t * arg,bool is_remount)489 static int f2fs_set_test_dummy_encryption(struct f2fs_sb_info *sbi,
490 const char *opt,
491 const substring_t *arg,
492 bool is_remount)
493 {
494 struct fs_parameter param = {
495 .type = fs_value_is_string,
496 .string = arg->from ? arg->from : "",
497 };
498 struct fscrypt_dummy_policy *policy =
499 &F2FS_OPTION(sbi).dummy_enc_policy;
500 int err;
501
502 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
503 f2fs_warn(sbi, "test_dummy_encryption option not supported");
504 return -EINVAL;
505 }
506
507 if (!f2fs_sb_has_encrypt(sbi)) {
508 f2fs_err(sbi, "Encrypt feature is off");
509 return -EINVAL;
510 }
511
512 /*
513 * This mount option is just for testing, and it's not worthwhile to
514 * implement the extra complexity (e.g. RCU protection) that would be
515 * needed to allow it to be set or changed during remount. We do allow
516 * it to be specified during remount, but only if there is no change.
517 */
518 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
519 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
520 return -EINVAL;
521 }
522
523 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
524 if (err) {
525 if (err == -EEXIST)
526 f2fs_warn(sbi,
527 "Can't change test_dummy_encryption on remount");
528 else if (err == -EINVAL)
529 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
530 opt);
531 else
532 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
533 opt, err);
534 return -EINVAL;
535 }
536 f2fs_warn(sbi, "Test dummy encryption mode enabled");
537 return 0;
538 }
539
540 #ifdef CONFIG_F2FS_FS_COMPRESSION
is_compress_extension_exist(struct f2fs_sb_info * sbi,const char * new_ext,bool is_ext)541 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
542 const char *new_ext, bool is_ext)
543 {
544 unsigned char (*ext)[F2FS_EXTENSION_LEN];
545 int ext_cnt;
546 int i;
547
548 if (is_ext) {
549 ext = F2FS_OPTION(sbi).extensions;
550 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
551 } else {
552 ext = F2FS_OPTION(sbi).noextensions;
553 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
554 }
555
556 for (i = 0; i < ext_cnt; i++) {
557 if (!strcasecmp(new_ext, ext[i]))
558 return true;
559 }
560
561 return false;
562 }
563
564 /*
565 * 1. The same extension name cannot not appear in both compress and non-compress extension
566 * at the same time.
567 * 2. If the compress extension specifies all files, the types specified by the non-compress
568 * extension will be treated as special cases and will not be compressed.
569 * 3. Don't allow the non-compress extension specifies all files.
570 */
f2fs_test_compress_extension(struct f2fs_sb_info * sbi)571 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
572 {
573 unsigned char (*ext)[F2FS_EXTENSION_LEN];
574 unsigned char (*noext)[F2FS_EXTENSION_LEN];
575 int ext_cnt, noext_cnt, index = 0, no_index = 0;
576
577 ext = F2FS_OPTION(sbi).extensions;
578 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
579 noext = F2FS_OPTION(sbi).noextensions;
580 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
581
582 if (!noext_cnt)
583 return 0;
584
585 for (no_index = 0; no_index < noext_cnt; no_index++) {
586 if (!strcasecmp("*", noext[no_index])) {
587 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
588 return -EINVAL;
589 }
590 for (index = 0; index < ext_cnt; index++) {
591 if (!strcasecmp(ext[index], noext[no_index])) {
592 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
593 ext[index]);
594 return -EINVAL;
595 }
596 }
597 }
598 return 0;
599 }
600
601 #ifdef CONFIG_F2FS_FS_LZ4
f2fs_set_lz4hc_level(struct f2fs_sb_info * sbi,const char * str)602 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
603 {
604 #ifdef CONFIG_F2FS_FS_LZ4HC
605 unsigned int level;
606
607 if (strlen(str) == 3) {
608 F2FS_OPTION(sbi).compress_level = 0;
609 return 0;
610 }
611
612 str += 3;
613
614 if (str[0] != ':') {
615 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
616 return -EINVAL;
617 }
618 if (kstrtouint(str + 1, 10, &level))
619 return -EINVAL;
620
621 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
622 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
623 return -EINVAL;
624 }
625
626 F2FS_OPTION(sbi).compress_level = level;
627 return 0;
628 #else
629 if (strlen(str) == 3) {
630 F2FS_OPTION(sbi).compress_level = 0;
631 return 0;
632 }
633 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
634 return -EINVAL;
635 #endif
636 }
637 #endif
638
639 #ifdef CONFIG_F2FS_FS_ZSTD
f2fs_set_zstd_level(struct f2fs_sb_info * sbi,const char * str)640 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
641 {
642 int level;
643 int len = 4;
644
645 if (strlen(str) == len) {
646 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
647 return 0;
648 }
649
650 str += len;
651
652 if (str[0] != ':') {
653 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
654 return -EINVAL;
655 }
656 if (kstrtoint(str + 1, 10, &level))
657 return -EINVAL;
658
659 /* f2fs does not support negative compress level now */
660 if (level < 0) {
661 f2fs_info(sbi, "do not support negative compress level: %d", level);
662 return -ERANGE;
663 }
664
665 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
666 f2fs_info(sbi, "invalid zstd compress level: %d", level);
667 return -EINVAL;
668 }
669
670 F2FS_OPTION(sbi).compress_level = level;
671 return 0;
672 }
673 #endif
674 #endif
675
parse_options(struct f2fs_sb_info * sbi,char * options,bool is_remount)676 static int parse_options(struct f2fs_sb_info *sbi, char *options, bool is_remount)
677 {
678 substring_t args[MAX_OPT_ARGS];
679 #ifdef CONFIG_F2FS_FS_COMPRESSION
680 unsigned char (*ext)[F2FS_EXTENSION_LEN];
681 unsigned char (*noext)[F2FS_EXTENSION_LEN];
682 int ext_cnt, noext_cnt;
683 #endif
684 char *p, *name;
685 int arg = 0;
686 kuid_t uid;
687 kgid_t gid;
688 int ret;
689
690 if (!options)
691 return 0;
692
693 while ((p = strsep(&options, ",")) != NULL) {
694 int token;
695
696 if (!*p)
697 continue;
698 /*
699 * Initialize args struct so we know whether arg was
700 * found; some options take optional arguments.
701 */
702 args[0].to = args[0].from = NULL;
703 token = match_token(p, f2fs_tokens, args);
704
705 switch (token) {
706 case Opt_gc_background:
707 name = match_strdup(&args[0]);
708
709 if (!name)
710 return -ENOMEM;
711 if (!strcmp(name, "on")) {
712 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
713 } else if (!strcmp(name, "off")) {
714 if (f2fs_sb_has_blkzoned(sbi)) {
715 f2fs_warn(sbi, "zoned devices need bggc");
716 kfree(name);
717 return -EINVAL;
718 }
719 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
720 } else if (!strcmp(name, "sync")) {
721 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
722 } else {
723 kfree(name);
724 return -EINVAL;
725 }
726 kfree(name);
727 break;
728 case Opt_disable_roll_forward:
729 set_opt(sbi, DISABLE_ROLL_FORWARD);
730 break;
731 case Opt_norecovery:
732 /* requires ro mount, checked in f2fs_default_check */
733 set_opt(sbi, NORECOVERY);
734 break;
735 case Opt_discard:
736 if (!f2fs_hw_support_discard(sbi)) {
737 f2fs_warn(sbi, "device does not support discard");
738 break;
739 }
740 set_opt(sbi, DISCARD);
741 break;
742 case Opt_nodiscard:
743 if (f2fs_hw_should_discard(sbi)) {
744 f2fs_warn(sbi, "discard is required for zoned block devices");
745 return -EINVAL;
746 }
747 clear_opt(sbi, DISCARD);
748 break;
749 case Opt_noheap:
750 case Opt_heap:
751 f2fs_warn(sbi, "heap/no_heap options were deprecated");
752 break;
753 #ifdef CONFIG_F2FS_FS_XATTR
754 case Opt_user_xattr:
755 set_opt(sbi, XATTR_USER);
756 break;
757 case Opt_nouser_xattr:
758 clear_opt(sbi, XATTR_USER);
759 break;
760 case Opt_inline_xattr:
761 set_opt(sbi, INLINE_XATTR);
762 break;
763 case Opt_noinline_xattr:
764 clear_opt(sbi, INLINE_XATTR);
765 break;
766 case Opt_inline_xattr_size:
767 if (args->from && match_int(args, &arg))
768 return -EINVAL;
769 set_opt(sbi, INLINE_XATTR_SIZE);
770 F2FS_OPTION(sbi).inline_xattr_size = arg;
771 break;
772 #else
773 case Opt_user_xattr:
774 case Opt_nouser_xattr:
775 case Opt_inline_xattr:
776 case Opt_noinline_xattr:
777 case Opt_inline_xattr_size:
778 f2fs_info(sbi, "xattr options not supported");
779 break;
780 #endif
781 #ifdef CONFIG_F2FS_FS_POSIX_ACL
782 case Opt_acl:
783 set_opt(sbi, POSIX_ACL);
784 break;
785 case Opt_noacl:
786 clear_opt(sbi, POSIX_ACL);
787 break;
788 #else
789 case Opt_acl:
790 case Opt_noacl:
791 f2fs_info(sbi, "acl options not supported");
792 break;
793 #endif
794 case Opt_active_logs:
795 if (args->from && match_int(args, &arg))
796 return -EINVAL;
797 if (arg != 2 && arg != 4 &&
798 arg != NR_CURSEG_PERSIST_TYPE)
799 return -EINVAL;
800 F2FS_OPTION(sbi).active_logs = arg;
801 break;
802 case Opt_disable_ext_identify:
803 set_opt(sbi, DISABLE_EXT_IDENTIFY);
804 break;
805 case Opt_inline_data:
806 set_opt(sbi, INLINE_DATA);
807 break;
808 case Opt_inline_dentry:
809 set_opt(sbi, INLINE_DENTRY);
810 break;
811 case Opt_noinline_dentry:
812 clear_opt(sbi, INLINE_DENTRY);
813 break;
814 case Opt_flush_merge:
815 set_opt(sbi, FLUSH_MERGE);
816 break;
817 case Opt_noflush_merge:
818 clear_opt(sbi, FLUSH_MERGE);
819 break;
820 case Opt_nobarrier:
821 set_opt(sbi, NOBARRIER);
822 break;
823 case Opt_barrier:
824 clear_opt(sbi, NOBARRIER);
825 break;
826 case Opt_fastboot:
827 set_opt(sbi, FASTBOOT);
828 break;
829 case Opt_extent_cache:
830 set_opt(sbi, READ_EXTENT_CACHE);
831 break;
832 case Opt_noextent_cache:
833 if (f2fs_sb_has_device_alias(sbi)) {
834 f2fs_err(sbi, "device aliasing requires extent cache");
835 return -EINVAL;
836 }
837 clear_opt(sbi, READ_EXTENT_CACHE);
838 break;
839 case Opt_noinline_data:
840 clear_opt(sbi, INLINE_DATA);
841 break;
842 case Opt_data_flush:
843 set_opt(sbi, DATA_FLUSH);
844 break;
845 case Opt_reserve_root:
846 if (args->from && match_int(args, &arg))
847 return -EINVAL;
848 if (test_opt(sbi, RESERVE_ROOT)) {
849 f2fs_info(sbi, "Preserve previous reserve_root=%u",
850 F2FS_OPTION(sbi).root_reserved_blocks);
851 } else {
852 F2FS_OPTION(sbi).root_reserved_blocks = arg;
853 set_opt(sbi, RESERVE_ROOT);
854 }
855 break;
856 case Opt_resuid:
857 if (args->from && match_int(args, &arg))
858 return -EINVAL;
859 uid = make_kuid(current_user_ns(), arg);
860 if (!uid_valid(uid)) {
861 f2fs_err(sbi, "Invalid uid value %d", arg);
862 return -EINVAL;
863 }
864 F2FS_OPTION(sbi).s_resuid = uid;
865 break;
866 case Opt_resgid:
867 if (args->from && match_int(args, &arg))
868 return -EINVAL;
869 gid = make_kgid(current_user_ns(), arg);
870 if (!gid_valid(gid)) {
871 f2fs_err(sbi, "Invalid gid value %d", arg);
872 return -EINVAL;
873 }
874 F2FS_OPTION(sbi).s_resgid = gid;
875 break;
876 case Opt_mode:
877 name = match_strdup(&args[0]);
878
879 if (!name)
880 return -ENOMEM;
881 if (!strcmp(name, "adaptive")) {
882 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
883 } else if (!strcmp(name, "lfs")) {
884 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
885 } else if (!strcmp(name, "fragment:segment")) {
886 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
887 } else if (!strcmp(name, "fragment:block")) {
888 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
889 } else {
890 kfree(name);
891 return -EINVAL;
892 }
893 kfree(name);
894 break;
895 #ifdef CONFIG_F2FS_FAULT_INJECTION
896 case Opt_fault_injection:
897 if (args->from && match_int(args, &arg))
898 return -EINVAL;
899 if (f2fs_build_fault_attr(sbi, arg,
900 F2FS_ALL_FAULT_TYPE))
901 return -EINVAL;
902 set_opt(sbi, FAULT_INJECTION);
903 break;
904
905 case Opt_fault_type:
906 if (args->from && match_int(args, &arg))
907 return -EINVAL;
908 if (f2fs_build_fault_attr(sbi, 0, arg))
909 return -EINVAL;
910 set_opt(sbi, FAULT_INJECTION);
911 break;
912 #else
913 case Opt_fault_injection:
914 case Opt_fault_type:
915 f2fs_info(sbi, "fault injection options not supported");
916 break;
917 #endif
918 case Opt_lazytime:
919 set_opt(sbi, LAZYTIME);
920 break;
921 case Opt_nolazytime:
922 clear_opt(sbi, LAZYTIME);
923 break;
924 #ifdef CONFIG_QUOTA
925 case Opt_quota:
926 case Opt_usrquota:
927 set_opt(sbi, USRQUOTA);
928 break;
929 case Opt_grpquota:
930 set_opt(sbi, GRPQUOTA);
931 break;
932 case Opt_prjquota:
933 set_opt(sbi, PRJQUOTA);
934 break;
935 case Opt_usrjquota:
936 ret = f2fs_set_qf_name(sbi, USRQUOTA, &args[0]);
937 if (ret)
938 return ret;
939 break;
940 case Opt_grpjquota:
941 ret = f2fs_set_qf_name(sbi, GRPQUOTA, &args[0]);
942 if (ret)
943 return ret;
944 break;
945 case Opt_prjjquota:
946 ret = f2fs_set_qf_name(sbi, PRJQUOTA, &args[0]);
947 if (ret)
948 return ret;
949 break;
950 case Opt_offusrjquota:
951 ret = f2fs_clear_qf_name(sbi, USRQUOTA);
952 if (ret)
953 return ret;
954 break;
955 case Opt_offgrpjquota:
956 ret = f2fs_clear_qf_name(sbi, GRPQUOTA);
957 if (ret)
958 return ret;
959 break;
960 case Opt_offprjjquota:
961 ret = f2fs_clear_qf_name(sbi, PRJQUOTA);
962 if (ret)
963 return ret;
964 break;
965 case Opt_jqfmt_vfsold:
966 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
967 break;
968 case Opt_jqfmt_vfsv0:
969 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
970 break;
971 case Opt_jqfmt_vfsv1:
972 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
973 break;
974 case Opt_noquota:
975 clear_opt(sbi, QUOTA);
976 clear_opt(sbi, USRQUOTA);
977 clear_opt(sbi, GRPQUOTA);
978 clear_opt(sbi, PRJQUOTA);
979 break;
980 #else
981 case Opt_quota:
982 case Opt_usrquota:
983 case Opt_grpquota:
984 case Opt_prjquota:
985 case Opt_usrjquota:
986 case Opt_grpjquota:
987 case Opt_prjjquota:
988 case Opt_offusrjquota:
989 case Opt_offgrpjquota:
990 case Opt_offprjjquota:
991 case Opt_jqfmt_vfsold:
992 case Opt_jqfmt_vfsv0:
993 case Opt_jqfmt_vfsv1:
994 case Opt_noquota:
995 f2fs_info(sbi, "quota operations not supported");
996 break;
997 #endif
998 case Opt_alloc:
999 name = match_strdup(&args[0]);
1000 if (!name)
1001 return -ENOMEM;
1002
1003 if (!strcmp(name, "default")) {
1004 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1005 } else if (!strcmp(name, "reuse")) {
1006 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1007 } else {
1008 kfree(name);
1009 return -EINVAL;
1010 }
1011 kfree(name);
1012 break;
1013 case Opt_fsync:
1014 name = match_strdup(&args[0]);
1015 if (!name)
1016 return -ENOMEM;
1017 if (!strcmp(name, "posix")) {
1018 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1019 } else if (!strcmp(name, "strict")) {
1020 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1021 } else if (!strcmp(name, "nobarrier")) {
1022 F2FS_OPTION(sbi).fsync_mode =
1023 FSYNC_MODE_NOBARRIER;
1024 } else {
1025 kfree(name);
1026 return -EINVAL;
1027 }
1028 kfree(name);
1029 break;
1030 case Opt_test_dummy_encryption:
1031 ret = f2fs_set_test_dummy_encryption(sbi, p, &args[0],
1032 is_remount);
1033 if (ret)
1034 return ret;
1035 break;
1036 case Opt_inlinecrypt:
1037 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1038 set_opt(sbi, INLINECRYPT);
1039 #else
1040 f2fs_info(sbi, "inline encryption not supported");
1041 #endif
1042 break;
1043 case Opt_checkpoint_disable_cap_perc:
1044 if (args->from && match_int(args, &arg))
1045 return -EINVAL;
1046 if (arg < 0 || arg > 100)
1047 return -EINVAL;
1048 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1049 set_opt(sbi, DISABLE_CHECKPOINT);
1050 break;
1051 case Opt_checkpoint_disable_cap:
1052 if (args->from && match_int(args, &arg))
1053 return -EINVAL;
1054 F2FS_OPTION(sbi).unusable_cap = arg;
1055 set_opt(sbi, DISABLE_CHECKPOINT);
1056 break;
1057 case Opt_checkpoint_disable:
1058 set_opt(sbi, DISABLE_CHECKPOINT);
1059 break;
1060 case Opt_checkpoint_enable:
1061 clear_opt(sbi, DISABLE_CHECKPOINT);
1062 break;
1063 case Opt_checkpoint_merge:
1064 set_opt(sbi, MERGE_CHECKPOINT);
1065 break;
1066 case Opt_nocheckpoint_merge:
1067 clear_opt(sbi, MERGE_CHECKPOINT);
1068 break;
1069 #ifdef CONFIG_F2FS_FS_COMPRESSION
1070 case Opt_compress_algorithm:
1071 if (!f2fs_sb_has_compression(sbi)) {
1072 f2fs_info(sbi, "Image doesn't support compression");
1073 break;
1074 }
1075 name = match_strdup(&args[0]);
1076 if (!name)
1077 return -ENOMEM;
1078 if (!strcmp(name, "lzo")) {
1079 #ifdef CONFIG_F2FS_FS_LZO
1080 F2FS_OPTION(sbi).compress_level = 0;
1081 F2FS_OPTION(sbi).compress_algorithm =
1082 COMPRESS_LZO;
1083 #else
1084 f2fs_info(sbi, "kernel doesn't support lzo compression");
1085 #endif
1086 } else if (!strncmp(name, "lz4", 3)) {
1087 #ifdef CONFIG_F2FS_FS_LZ4
1088 ret = f2fs_set_lz4hc_level(sbi, name);
1089 if (ret) {
1090 kfree(name);
1091 return -EINVAL;
1092 }
1093 F2FS_OPTION(sbi).compress_algorithm =
1094 COMPRESS_LZ4;
1095 #else
1096 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1097 #endif
1098 } else if (!strncmp(name, "zstd", 4)) {
1099 #ifdef CONFIG_F2FS_FS_ZSTD
1100 ret = f2fs_set_zstd_level(sbi, name);
1101 if (ret) {
1102 kfree(name);
1103 return -EINVAL;
1104 }
1105 F2FS_OPTION(sbi).compress_algorithm =
1106 COMPRESS_ZSTD;
1107 #else
1108 f2fs_info(sbi, "kernel doesn't support zstd compression");
1109 #endif
1110 } else if (!strcmp(name, "lzo-rle")) {
1111 #ifdef CONFIG_F2FS_FS_LZORLE
1112 F2FS_OPTION(sbi).compress_level = 0;
1113 F2FS_OPTION(sbi).compress_algorithm =
1114 COMPRESS_LZORLE;
1115 #else
1116 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1117 #endif
1118 } else {
1119 kfree(name);
1120 return -EINVAL;
1121 }
1122 kfree(name);
1123 break;
1124 case Opt_compress_log_size:
1125 if (!f2fs_sb_has_compression(sbi)) {
1126 f2fs_info(sbi, "Image doesn't support compression");
1127 break;
1128 }
1129 if (args->from && match_int(args, &arg))
1130 return -EINVAL;
1131 if (arg < MIN_COMPRESS_LOG_SIZE ||
1132 arg > MAX_COMPRESS_LOG_SIZE) {
1133 f2fs_err(sbi,
1134 "Compress cluster log size is out of range");
1135 return -EINVAL;
1136 }
1137 F2FS_OPTION(sbi).compress_log_size = arg;
1138 break;
1139 case Opt_compress_extension:
1140 if (!f2fs_sb_has_compression(sbi)) {
1141 f2fs_info(sbi, "Image doesn't support compression");
1142 break;
1143 }
1144 name = match_strdup(&args[0]);
1145 if (!name)
1146 return -ENOMEM;
1147
1148 ext = F2FS_OPTION(sbi).extensions;
1149 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1150
1151 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1152 ext_cnt >= COMPRESS_EXT_NUM) {
1153 f2fs_err(sbi,
1154 "invalid extension length/number");
1155 kfree(name);
1156 return -EINVAL;
1157 }
1158
1159 if (is_compress_extension_exist(sbi, name, true)) {
1160 kfree(name);
1161 break;
1162 }
1163
1164 ret = strscpy(ext[ext_cnt], name);
1165 if (ret < 0) {
1166 kfree(name);
1167 return ret;
1168 }
1169 F2FS_OPTION(sbi).compress_ext_cnt++;
1170 kfree(name);
1171 break;
1172 case Opt_nocompress_extension:
1173 if (!f2fs_sb_has_compression(sbi)) {
1174 f2fs_info(sbi, "Image doesn't support compression");
1175 break;
1176 }
1177 name = match_strdup(&args[0]);
1178 if (!name)
1179 return -ENOMEM;
1180
1181 noext = F2FS_OPTION(sbi).noextensions;
1182 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1183
1184 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1185 noext_cnt >= COMPRESS_EXT_NUM) {
1186 f2fs_err(sbi,
1187 "invalid extension length/number");
1188 kfree(name);
1189 return -EINVAL;
1190 }
1191
1192 if (is_compress_extension_exist(sbi, name, false)) {
1193 kfree(name);
1194 break;
1195 }
1196
1197 ret = strscpy(noext[noext_cnt], name);
1198 if (ret < 0) {
1199 kfree(name);
1200 return ret;
1201 }
1202 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1203 kfree(name);
1204 break;
1205 case Opt_compress_chksum:
1206 if (!f2fs_sb_has_compression(sbi)) {
1207 f2fs_info(sbi, "Image doesn't support compression");
1208 break;
1209 }
1210 F2FS_OPTION(sbi).compress_chksum = true;
1211 break;
1212 case Opt_compress_mode:
1213 if (!f2fs_sb_has_compression(sbi)) {
1214 f2fs_info(sbi, "Image doesn't support compression");
1215 break;
1216 }
1217 name = match_strdup(&args[0]);
1218 if (!name)
1219 return -ENOMEM;
1220 if (!strcmp(name, "fs")) {
1221 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1222 } else if (!strcmp(name, "user")) {
1223 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1224 } else {
1225 kfree(name);
1226 return -EINVAL;
1227 }
1228 kfree(name);
1229 break;
1230 case Opt_compress_cache:
1231 if (!f2fs_sb_has_compression(sbi)) {
1232 f2fs_info(sbi, "Image doesn't support compression");
1233 break;
1234 }
1235 set_opt(sbi, COMPRESS_CACHE);
1236 break;
1237 #else
1238 case Opt_compress_algorithm:
1239 case Opt_compress_log_size:
1240 case Opt_compress_extension:
1241 case Opt_nocompress_extension:
1242 case Opt_compress_chksum:
1243 case Opt_compress_mode:
1244 case Opt_compress_cache:
1245 f2fs_info(sbi, "compression options not supported");
1246 break;
1247 #endif
1248 case Opt_atgc:
1249 set_opt(sbi, ATGC);
1250 break;
1251 case Opt_gc_merge:
1252 set_opt(sbi, GC_MERGE);
1253 break;
1254 case Opt_nogc_merge:
1255 clear_opt(sbi, GC_MERGE);
1256 break;
1257 case Opt_discard_unit:
1258 name = match_strdup(&args[0]);
1259 if (!name)
1260 return -ENOMEM;
1261 if (!strcmp(name, "block")) {
1262 F2FS_OPTION(sbi).discard_unit =
1263 DISCARD_UNIT_BLOCK;
1264 } else if (!strcmp(name, "segment")) {
1265 F2FS_OPTION(sbi).discard_unit =
1266 DISCARD_UNIT_SEGMENT;
1267 } else if (!strcmp(name, "section")) {
1268 F2FS_OPTION(sbi).discard_unit =
1269 DISCARD_UNIT_SECTION;
1270 } else {
1271 kfree(name);
1272 return -EINVAL;
1273 }
1274 kfree(name);
1275 break;
1276 case Opt_memory_mode:
1277 name = match_strdup(&args[0]);
1278 if (!name)
1279 return -ENOMEM;
1280 if (!strcmp(name, "normal")) {
1281 F2FS_OPTION(sbi).memory_mode =
1282 MEMORY_MODE_NORMAL;
1283 } else if (!strcmp(name, "low")) {
1284 F2FS_OPTION(sbi).memory_mode =
1285 MEMORY_MODE_LOW;
1286 } else {
1287 kfree(name);
1288 return -EINVAL;
1289 }
1290 kfree(name);
1291 break;
1292 case Opt_age_extent_cache:
1293 set_opt(sbi, AGE_EXTENT_CACHE);
1294 break;
1295 case Opt_errors:
1296 name = match_strdup(&args[0]);
1297 if (!name)
1298 return -ENOMEM;
1299 if (!strcmp(name, "remount-ro")) {
1300 F2FS_OPTION(sbi).errors =
1301 MOUNT_ERRORS_READONLY;
1302 } else if (!strcmp(name, "continue")) {
1303 F2FS_OPTION(sbi).errors =
1304 MOUNT_ERRORS_CONTINUE;
1305 } else if (!strcmp(name, "panic")) {
1306 F2FS_OPTION(sbi).errors =
1307 MOUNT_ERRORS_PANIC;
1308 } else {
1309 kfree(name);
1310 return -EINVAL;
1311 }
1312 kfree(name);
1313 break;
1314 case Opt_nat_bits:
1315 set_opt(sbi, NAT_BITS);
1316 break;
1317 default:
1318 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1319 p);
1320 return -EINVAL;
1321 }
1322 }
1323 return 0;
1324 }
1325
f2fs_default_check(struct f2fs_sb_info * sbi)1326 static int f2fs_default_check(struct f2fs_sb_info *sbi)
1327 {
1328 #ifdef CONFIG_QUOTA
1329 if (f2fs_check_quota_options(sbi))
1330 return -EINVAL;
1331 #else
1332 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1333 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1334 return -EINVAL;
1335 }
1336 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1337 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1338 return -EINVAL;
1339 }
1340 #endif
1341
1342 if (!IS_ENABLED(CONFIG_UNICODE) && f2fs_sb_has_casefold(sbi)) {
1343 f2fs_err(sbi,
1344 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1345 return -EINVAL;
1346 }
1347
1348 /*
1349 * The BLKZONED feature indicates that the drive was formatted with
1350 * zone alignment optimization. This is optional for host-aware
1351 * devices, but mandatory for host-managed zoned block devices.
1352 */
1353 if (f2fs_sb_has_blkzoned(sbi)) {
1354 #ifdef CONFIG_BLK_DEV_ZONED
1355 if (F2FS_OPTION(sbi).discard_unit !=
1356 DISCARD_UNIT_SECTION) {
1357 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1358 F2FS_OPTION(sbi).discard_unit =
1359 DISCARD_UNIT_SECTION;
1360 }
1361
1362 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1363 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1364 return -EINVAL;
1365 }
1366 #else
1367 f2fs_err(sbi, "Zoned block device support is not enabled");
1368 return -EINVAL;
1369 #endif
1370 }
1371
1372 #ifdef CONFIG_F2FS_FS_COMPRESSION
1373 if (f2fs_test_compress_extension(sbi)) {
1374 f2fs_err(sbi, "invalid compress or nocompress extension");
1375 return -EINVAL;
1376 }
1377 #endif
1378
1379 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1380 int min_size, max_size;
1381
1382 if (!f2fs_sb_has_extra_attr(sbi) ||
1383 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1384 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1385 return -EINVAL;
1386 }
1387 if (!test_opt(sbi, INLINE_XATTR)) {
1388 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1389 return -EINVAL;
1390 }
1391
1392 min_size = MIN_INLINE_XATTR_SIZE;
1393 max_size = MAX_INLINE_XATTR_SIZE;
1394
1395 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1396 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1397 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1398 min_size, max_size);
1399 return -EINVAL;
1400 }
1401 }
1402
1403 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1404 f2fs_err(sbi, "LFS is not compatible with ATGC");
1405 return -EINVAL;
1406 }
1407
1408 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1409 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1410 return -EINVAL;
1411 }
1412
1413 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1414 f2fs_err(sbi, "Allow to mount readonly mode only");
1415 return -EROFS;
1416 }
1417
1418 if (test_opt(sbi, NORECOVERY) && !f2fs_readonly(sbi->sb)) {
1419 f2fs_err(sbi, "norecovery requires readonly mount");
1420 return -EINVAL;
1421 }
1422
1423 return 0;
1424 }
1425
f2fs_alloc_inode(struct super_block * sb)1426 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1427 {
1428 struct f2fs_inode_info *fi;
1429
1430 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1431 return NULL;
1432
1433 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1434 if (!fi)
1435 return NULL;
1436
1437 init_once((void *) fi);
1438
1439 /* Initialize f2fs-specific inode info */
1440 atomic_set(&fi->dirty_pages, 0);
1441 atomic_set(&fi->i_compr_blocks, 0);
1442 init_f2fs_rwsem(&fi->i_sem);
1443 spin_lock_init(&fi->i_size_lock);
1444 INIT_LIST_HEAD(&fi->dirty_list);
1445 INIT_LIST_HEAD(&fi->gdirty_list);
1446 INIT_LIST_HEAD(&fi->gdonate_list);
1447 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1448 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1449 init_f2fs_rwsem(&fi->i_xattr_sem);
1450
1451 /* Will be used by directory only */
1452 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1453
1454 return &fi->vfs_inode;
1455 }
1456
f2fs_drop_inode(struct inode * inode)1457 static int f2fs_drop_inode(struct inode *inode)
1458 {
1459 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1460 int ret;
1461
1462 /*
1463 * during filesystem shutdown, if checkpoint is disabled,
1464 * drop useless meta/node dirty pages.
1465 */
1466 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1467 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1468 inode->i_ino == F2FS_META_INO(sbi)) {
1469 trace_f2fs_drop_inode(inode, 1);
1470 return 1;
1471 }
1472 }
1473
1474 /*
1475 * This is to avoid a deadlock condition like below.
1476 * writeback_single_inode(inode)
1477 * - f2fs_write_data_page
1478 * - f2fs_gc -> iput -> evict
1479 * - inode_wait_for_writeback(inode)
1480 */
1481 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1482 if (!inode->i_nlink && !is_bad_inode(inode)) {
1483 /* to avoid evict_inode call simultaneously */
1484 atomic_inc(&inode->i_count);
1485 spin_unlock(&inode->i_lock);
1486
1487 /* should remain fi->extent_tree for writepage */
1488 f2fs_destroy_extent_node(inode);
1489
1490 sb_start_intwrite(inode->i_sb);
1491 f2fs_i_size_write(inode, 0);
1492
1493 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1494 inode, NULL, 0, DATA);
1495 truncate_inode_pages_final(inode->i_mapping);
1496
1497 if (F2FS_HAS_BLOCKS(inode))
1498 f2fs_truncate(inode);
1499
1500 sb_end_intwrite(inode->i_sb);
1501
1502 spin_lock(&inode->i_lock);
1503 atomic_dec(&inode->i_count);
1504 }
1505 trace_f2fs_drop_inode(inode, 0);
1506 return 0;
1507 }
1508 ret = generic_drop_inode(inode);
1509 if (!ret)
1510 ret = fscrypt_drop_inode(inode);
1511 trace_f2fs_drop_inode(inode, ret);
1512 return ret;
1513 }
1514
f2fs_inode_dirtied(struct inode * inode,bool sync)1515 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1516 {
1517 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1518 int ret = 0;
1519
1520 spin_lock(&sbi->inode_lock[DIRTY_META]);
1521 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1522 ret = 1;
1523 } else {
1524 set_inode_flag(inode, FI_DIRTY_INODE);
1525 stat_inc_dirty_inode(sbi, DIRTY_META);
1526 }
1527 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1528 list_add_tail(&F2FS_I(inode)->gdirty_list,
1529 &sbi->inode_list[DIRTY_META]);
1530 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1531 }
1532 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1533
1534 if (!ret && f2fs_is_atomic_file(inode))
1535 set_inode_flag(inode, FI_ATOMIC_DIRTIED);
1536
1537 return ret;
1538 }
1539
f2fs_inode_synced(struct inode * inode)1540 void f2fs_inode_synced(struct inode *inode)
1541 {
1542 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1543
1544 spin_lock(&sbi->inode_lock[DIRTY_META]);
1545 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1546 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1547 return;
1548 }
1549 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1550 list_del_init(&F2FS_I(inode)->gdirty_list);
1551 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1552 }
1553 clear_inode_flag(inode, FI_DIRTY_INODE);
1554 clear_inode_flag(inode, FI_AUTO_RECOVER);
1555 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1556 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1557 }
1558
1559 /*
1560 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1561 *
1562 * We should call set_dirty_inode to write the dirty inode through write_inode.
1563 */
f2fs_dirty_inode(struct inode * inode,int flags)1564 static void f2fs_dirty_inode(struct inode *inode, int flags)
1565 {
1566 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1567
1568 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1569 inode->i_ino == F2FS_META_INO(sbi))
1570 return;
1571
1572 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1573 clear_inode_flag(inode, FI_AUTO_RECOVER);
1574
1575 f2fs_inode_dirtied(inode, false);
1576 }
1577
f2fs_free_inode(struct inode * inode)1578 static void f2fs_free_inode(struct inode *inode)
1579 {
1580 fscrypt_free_inode(inode);
1581 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1582 }
1583
destroy_percpu_info(struct f2fs_sb_info * sbi)1584 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1585 {
1586 percpu_counter_destroy(&sbi->total_valid_inode_count);
1587 percpu_counter_destroy(&sbi->rf_node_block_count);
1588 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1589 }
1590
destroy_device_list(struct f2fs_sb_info * sbi)1591 static void destroy_device_list(struct f2fs_sb_info *sbi)
1592 {
1593 int i;
1594
1595 for (i = 0; i < sbi->s_ndevs; i++) {
1596 if (i > 0)
1597 bdev_fput(FDEV(i).bdev_file);
1598 #ifdef CONFIG_BLK_DEV_ZONED
1599 kvfree(FDEV(i).blkz_seq);
1600 #endif
1601 }
1602 kvfree(sbi->devs);
1603 }
1604
f2fs_put_super(struct super_block * sb)1605 static void f2fs_put_super(struct super_block *sb)
1606 {
1607 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1608 int i;
1609 int err = 0;
1610 bool done;
1611
1612 /* unregister procfs/sysfs entries in advance to avoid race case */
1613 f2fs_unregister_sysfs(sbi);
1614
1615 f2fs_quota_off_umount(sb);
1616
1617 /* prevent remaining shrinker jobs */
1618 mutex_lock(&sbi->umount_mutex);
1619
1620 /*
1621 * flush all issued checkpoints and stop checkpoint issue thread.
1622 * after then, all checkpoints should be done by each process context.
1623 */
1624 f2fs_stop_ckpt_thread(sbi);
1625
1626 /*
1627 * We don't need to do checkpoint when superblock is clean.
1628 * But, the previous checkpoint was not done by umount, it needs to do
1629 * clean checkpoint again.
1630 */
1631 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1632 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1633 struct cp_control cpc = {
1634 .reason = CP_UMOUNT,
1635 };
1636 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1637 err = f2fs_write_checkpoint(sbi, &cpc);
1638 }
1639
1640 /* be sure to wait for any on-going discard commands */
1641 done = f2fs_issue_discard_timeout(sbi);
1642 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1643 struct cp_control cpc = {
1644 .reason = CP_UMOUNT | CP_TRIMMED,
1645 };
1646 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1647 err = f2fs_write_checkpoint(sbi, &cpc);
1648 }
1649
1650 /*
1651 * normally superblock is clean, so we need to release this.
1652 * In addition, EIO will skip do checkpoint, we need this as well.
1653 */
1654 f2fs_release_ino_entry(sbi, true);
1655
1656 f2fs_leave_shrinker(sbi);
1657 mutex_unlock(&sbi->umount_mutex);
1658
1659 /* our cp_error case, we can wait for any writeback page */
1660 f2fs_flush_merged_writes(sbi);
1661
1662 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1663
1664 if (err || f2fs_cp_error(sbi)) {
1665 truncate_inode_pages_final(NODE_MAPPING(sbi));
1666 truncate_inode_pages_final(META_MAPPING(sbi));
1667 }
1668
1669 for (i = 0; i < NR_COUNT_TYPE; i++) {
1670 if (!get_pages(sbi, i))
1671 continue;
1672 f2fs_err(sbi, "detect filesystem reference count leak during "
1673 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1674 f2fs_bug_on(sbi, 1);
1675 }
1676
1677 f2fs_bug_on(sbi, sbi->fsync_node_num);
1678
1679 f2fs_destroy_compress_inode(sbi);
1680
1681 iput(sbi->node_inode);
1682 sbi->node_inode = NULL;
1683
1684 iput(sbi->meta_inode);
1685 sbi->meta_inode = NULL;
1686
1687 /*
1688 * iput() can update stat information, if f2fs_write_checkpoint()
1689 * above failed with error.
1690 */
1691 f2fs_destroy_stats(sbi);
1692
1693 /* destroy f2fs internal modules */
1694 f2fs_destroy_node_manager(sbi);
1695 f2fs_destroy_segment_manager(sbi);
1696
1697 /* flush s_error_work before sbi destroy */
1698 flush_work(&sbi->s_error_work);
1699
1700 f2fs_destroy_post_read_wq(sbi);
1701
1702 kvfree(sbi->ckpt);
1703
1704 kfree(sbi->raw_super);
1705
1706 f2fs_destroy_page_array_cache(sbi);
1707 f2fs_destroy_xattr_caches(sbi);
1708 #ifdef CONFIG_QUOTA
1709 for (i = 0; i < MAXQUOTAS; i++)
1710 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1711 #endif
1712 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1713 destroy_percpu_info(sbi);
1714 f2fs_destroy_iostat(sbi);
1715 for (i = 0; i < NR_PAGE_TYPE; i++)
1716 kvfree(sbi->write_io[i]);
1717 #if IS_ENABLED(CONFIG_UNICODE)
1718 utf8_unload(sb->s_encoding);
1719 #endif
1720 }
1721
f2fs_sync_fs(struct super_block * sb,int sync)1722 int f2fs_sync_fs(struct super_block *sb, int sync)
1723 {
1724 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1725 int err = 0;
1726
1727 if (unlikely(f2fs_cp_error(sbi)))
1728 return 0;
1729 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1730 return 0;
1731
1732 trace_f2fs_sync_fs(sb, sync);
1733
1734 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1735 return -EAGAIN;
1736
1737 if (sync) {
1738 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1739 err = f2fs_issue_checkpoint(sbi);
1740 }
1741
1742 return err;
1743 }
1744
f2fs_freeze(struct super_block * sb)1745 static int f2fs_freeze(struct super_block *sb)
1746 {
1747 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1748
1749 if (f2fs_readonly(sb))
1750 return 0;
1751
1752 /* IO error happened before */
1753 if (unlikely(f2fs_cp_error(sbi)))
1754 return -EIO;
1755
1756 /* must be clean, since sync_filesystem() was already called */
1757 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY))
1758 return -EINVAL;
1759
1760 sbi->umount_lock_holder = current;
1761
1762 /* Let's flush checkpoints and stop the thread. */
1763 f2fs_flush_ckpt_thread(sbi);
1764
1765 sbi->umount_lock_holder = NULL;
1766
1767 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1768 set_sbi_flag(sbi, SBI_IS_FREEZING);
1769 return 0;
1770 }
1771
f2fs_unfreeze(struct super_block * sb)1772 static int f2fs_unfreeze(struct super_block *sb)
1773 {
1774 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1775
1776 /*
1777 * It will update discard_max_bytes of mounted lvm device to zero
1778 * after creating snapshot on this lvm device, let's drop all
1779 * remained discards.
1780 * We don't need to disable real-time discard because discard_max_bytes
1781 * will recover after removal of snapshot.
1782 */
1783 if (test_opt(sbi, DISCARD) && !f2fs_hw_support_discard(sbi))
1784 f2fs_issue_discard_timeout(sbi);
1785
1786 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1787 return 0;
1788 }
1789
1790 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1791 static int f2fs_statfs_project(struct super_block *sb,
1792 kprojid_t projid, struct kstatfs *buf)
1793 {
1794 struct kqid qid;
1795 struct dquot *dquot;
1796 u64 limit;
1797 u64 curblock;
1798
1799 qid = make_kqid_projid(projid);
1800 dquot = dqget(sb, qid);
1801 if (IS_ERR(dquot))
1802 return PTR_ERR(dquot);
1803 spin_lock(&dquot->dq_dqb_lock);
1804
1805 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1806 dquot->dq_dqb.dqb_bhardlimit);
1807 if (limit)
1808 limit >>= sb->s_blocksize_bits;
1809
1810 if (limit && buf->f_blocks > limit) {
1811 curblock = (dquot->dq_dqb.dqb_curspace +
1812 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1813 buf->f_blocks = limit;
1814 buf->f_bfree = buf->f_bavail =
1815 (buf->f_blocks > curblock) ?
1816 (buf->f_blocks - curblock) : 0;
1817 }
1818
1819 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1820 dquot->dq_dqb.dqb_ihardlimit);
1821
1822 if (limit && buf->f_files > limit) {
1823 buf->f_files = limit;
1824 buf->f_ffree =
1825 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1826 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1827 }
1828
1829 spin_unlock(&dquot->dq_dqb_lock);
1830 dqput(dquot);
1831 return 0;
1832 }
1833 #endif
1834
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1835 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1836 {
1837 struct super_block *sb = dentry->d_sb;
1838 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1839 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1840 block_t total_count, user_block_count, start_count;
1841 u64 avail_node_count;
1842 unsigned int total_valid_node_count;
1843
1844 total_count = le64_to_cpu(sbi->raw_super->block_count);
1845 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1846 buf->f_type = F2FS_SUPER_MAGIC;
1847 buf->f_bsize = sbi->blocksize;
1848
1849 buf->f_blocks = total_count - start_count;
1850
1851 spin_lock(&sbi->stat_lock);
1852 if (sbi->carve_out)
1853 buf->f_blocks -= sbi->current_reserved_blocks;
1854 user_block_count = sbi->user_block_count;
1855 total_valid_node_count = valid_node_count(sbi);
1856 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1857 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1858 sbi->current_reserved_blocks;
1859
1860 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1861 buf->f_bfree = 0;
1862 else
1863 buf->f_bfree -= sbi->unusable_block_count;
1864 spin_unlock(&sbi->stat_lock);
1865
1866 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1867 buf->f_bavail = buf->f_bfree -
1868 F2FS_OPTION(sbi).root_reserved_blocks;
1869 else
1870 buf->f_bavail = 0;
1871
1872 if (avail_node_count > user_block_count) {
1873 buf->f_files = user_block_count;
1874 buf->f_ffree = buf->f_bavail;
1875 } else {
1876 buf->f_files = avail_node_count;
1877 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1878 buf->f_bavail);
1879 }
1880
1881 buf->f_namelen = F2FS_NAME_LEN;
1882 buf->f_fsid = u64_to_fsid(id);
1883
1884 #ifdef CONFIG_QUOTA
1885 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1886 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1887 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1888 }
1889 #endif
1890 return 0;
1891 }
1892
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1893 static inline void f2fs_show_quota_options(struct seq_file *seq,
1894 struct super_block *sb)
1895 {
1896 #ifdef CONFIG_QUOTA
1897 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1898
1899 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1900 char *fmtname = "";
1901
1902 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1903 case QFMT_VFS_OLD:
1904 fmtname = "vfsold";
1905 break;
1906 case QFMT_VFS_V0:
1907 fmtname = "vfsv0";
1908 break;
1909 case QFMT_VFS_V1:
1910 fmtname = "vfsv1";
1911 break;
1912 }
1913 seq_printf(seq, ",jqfmt=%s", fmtname);
1914 }
1915
1916 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1917 seq_show_option(seq, "usrjquota",
1918 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1919
1920 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1921 seq_show_option(seq, "grpjquota",
1922 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1923
1924 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1925 seq_show_option(seq, "prjjquota",
1926 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1927 #endif
1928 }
1929
1930 #ifdef CONFIG_F2FS_FS_COMPRESSION
f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1931 static inline void f2fs_show_compress_options(struct seq_file *seq,
1932 struct super_block *sb)
1933 {
1934 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1935 char *algtype = "";
1936 int i;
1937
1938 if (!f2fs_sb_has_compression(sbi))
1939 return;
1940
1941 switch (F2FS_OPTION(sbi).compress_algorithm) {
1942 case COMPRESS_LZO:
1943 algtype = "lzo";
1944 break;
1945 case COMPRESS_LZ4:
1946 algtype = "lz4";
1947 break;
1948 case COMPRESS_ZSTD:
1949 algtype = "zstd";
1950 break;
1951 case COMPRESS_LZORLE:
1952 algtype = "lzo-rle";
1953 break;
1954 }
1955 seq_printf(seq, ",compress_algorithm=%s", algtype);
1956
1957 if (F2FS_OPTION(sbi).compress_level)
1958 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1959
1960 seq_printf(seq, ",compress_log_size=%u",
1961 F2FS_OPTION(sbi).compress_log_size);
1962
1963 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1964 seq_printf(seq, ",compress_extension=%s",
1965 F2FS_OPTION(sbi).extensions[i]);
1966 }
1967
1968 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1969 seq_printf(seq, ",nocompress_extension=%s",
1970 F2FS_OPTION(sbi).noextensions[i]);
1971 }
1972
1973 if (F2FS_OPTION(sbi).compress_chksum)
1974 seq_puts(seq, ",compress_chksum");
1975
1976 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1977 seq_printf(seq, ",compress_mode=%s", "fs");
1978 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1979 seq_printf(seq, ",compress_mode=%s", "user");
1980
1981 if (test_opt(sbi, COMPRESS_CACHE))
1982 seq_puts(seq, ",compress_cache");
1983 }
1984 #endif
1985
f2fs_show_options(struct seq_file * seq,struct dentry * root)1986 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1987 {
1988 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1989
1990 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1991 seq_printf(seq, ",background_gc=%s", "sync");
1992 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1993 seq_printf(seq, ",background_gc=%s", "on");
1994 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1995 seq_printf(seq, ",background_gc=%s", "off");
1996
1997 if (test_opt(sbi, GC_MERGE))
1998 seq_puts(seq, ",gc_merge");
1999 else
2000 seq_puts(seq, ",nogc_merge");
2001
2002 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2003 seq_puts(seq, ",disable_roll_forward");
2004 if (test_opt(sbi, NORECOVERY))
2005 seq_puts(seq, ",norecovery");
2006 if (test_opt(sbi, DISCARD)) {
2007 seq_puts(seq, ",discard");
2008 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2009 seq_printf(seq, ",discard_unit=%s", "block");
2010 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2011 seq_printf(seq, ",discard_unit=%s", "segment");
2012 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2013 seq_printf(seq, ",discard_unit=%s", "section");
2014 } else {
2015 seq_puts(seq, ",nodiscard");
2016 }
2017 #ifdef CONFIG_F2FS_FS_XATTR
2018 if (test_opt(sbi, XATTR_USER))
2019 seq_puts(seq, ",user_xattr");
2020 else
2021 seq_puts(seq, ",nouser_xattr");
2022 if (test_opt(sbi, INLINE_XATTR))
2023 seq_puts(seq, ",inline_xattr");
2024 else
2025 seq_puts(seq, ",noinline_xattr");
2026 if (test_opt(sbi, INLINE_XATTR_SIZE))
2027 seq_printf(seq, ",inline_xattr_size=%u",
2028 F2FS_OPTION(sbi).inline_xattr_size);
2029 #endif
2030 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2031 if (test_opt(sbi, POSIX_ACL))
2032 seq_puts(seq, ",acl");
2033 else
2034 seq_puts(seq, ",noacl");
2035 #endif
2036 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
2037 seq_puts(seq, ",disable_ext_identify");
2038 if (test_opt(sbi, INLINE_DATA))
2039 seq_puts(seq, ",inline_data");
2040 else
2041 seq_puts(seq, ",noinline_data");
2042 if (test_opt(sbi, INLINE_DENTRY))
2043 seq_puts(seq, ",inline_dentry");
2044 else
2045 seq_puts(seq, ",noinline_dentry");
2046 if (test_opt(sbi, FLUSH_MERGE))
2047 seq_puts(seq, ",flush_merge");
2048 else
2049 seq_puts(seq, ",noflush_merge");
2050 if (test_opt(sbi, NOBARRIER))
2051 seq_puts(seq, ",nobarrier");
2052 else
2053 seq_puts(seq, ",barrier");
2054 if (test_opt(sbi, FASTBOOT))
2055 seq_puts(seq, ",fastboot");
2056 if (test_opt(sbi, READ_EXTENT_CACHE))
2057 seq_puts(seq, ",extent_cache");
2058 else
2059 seq_puts(seq, ",noextent_cache");
2060 if (test_opt(sbi, AGE_EXTENT_CACHE))
2061 seq_puts(seq, ",age_extent_cache");
2062 if (test_opt(sbi, DATA_FLUSH))
2063 seq_puts(seq, ",data_flush");
2064
2065 seq_puts(seq, ",mode=");
2066 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2067 seq_puts(seq, "adaptive");
2068 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2069 seq_puts(seq, "lfs");
2070 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2071 seq_puts(seq, "fragment:segment");
2072 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2073 seq_puts(seq, "fragment:block");
2074 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2075 if (test_opt(sbi, RESERVE_ROOT))
2076 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2077 F2FS_OPTION(sbi).root_reserved_blocks,
2078 from_kuid_munged(&init_user_ns,
2079 F2FS_OPTION(sbi).s_resuid),
2080 from_kgid_munged(&init_user_ns,
2081 F2FS_OPTION(sbi).s_resgid));
2082 #ifdef CONFIG_F2FS_FAULT_INJECTION
2083 if (test_opt(sbi, FAULT_INJECTION)) {
2084 seq_printf(seq, ",fault_injection=%u",
2085 F2FS_OPTION(sbi).fault_info.inject_rate);
2086 seq_printf(seq, ",fault_type=%u",
2087 F2FS_OPTION(sbi).fault_info.inject_type);
2088 }
2089 #endif
2090 #ifdef CONFIG_QUOTA
2091 if (test_opt(sbi, QUOTA))
2092 seq_puts(seq, ",quota");
2093 if (test_opt(sbi, USRQUOTA))
2094 seq_puts(seq, ",usrquota");
2095 if (test_opt(sbi, GRPQUOTA))
2096 seq_puts(seq, ",grpquota");
2097 if (test_opt(sbi, PRJQUOTA))
2098 seq_puts(seq, ",prjquota");
2099 #endif
2100 f2fs_show_quota_options(seq, sbi->sb);
2101
2102 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2103
2104 if (sbi->sb->s_flags & SB_INLINECRYPT)
2105 seq_puts(seq, ",inlinecrypt");
2106
2107 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2108 seq_printf(seq, ",alloc_mode=%s", "default");
2109 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2110 seq_printf(seq, ",alloc_mode=%s", "reuse");
2111
2112 if (test_opt(sbi, DISABLE_CHECKPOINT))
2113 seq_printf(seq, ",checkpoint=disable:%u",
2114 F2FS_OPTION(sbi).unusable_cap);
2115 if (test_opt(sbi, MERGE_CHECKPOINT))
2116 seq_puts(seq, ",checkpoint_merge");
2117 else
2118 seq_puts(seq, ",nocheckpoint_merge");
2119 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2120 seq_printf(seq, ",fsync_mode=%s", "posix");
2121 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2122 seq_printf(seq, ",fsync_mode=%s", "strict");
2123 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2124 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2125
2126 #ifdef CONFIG_F2FS_FS_COMPRESSION
2127 f2fs_show_compress_options(seq, sbi->sb);
2128 #endif
2129
2130 if (test_opt(sbi, ATGC))
2131 seq_puts(seq, ",atgc");
2132
2133 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2134 seq_printf(seq, ",memory=%s", "normal");
2135 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2136 seq_printf(seq, ",memory=%s", "low");
2137
2138 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2139 seq_printf(seq, ",errors=%s", "remount-ro");
2140 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2141 seq_printf(seq, ",errors=%s", "continue");
2142 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2143 seq_printf(seq, ",errors=%s", "panic");
2144
2145 if (test_opt(sbi, NAT_BITS))
2146 seq_puts(seq, ",nat_bits");
2147
2148 return 0;
2149 }
2150
default_options(struct f2fs_sb_info * sbi,bool remount)2151 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2152 {
2153 /* init some FS parameters */
2154 if (!remount) {
2155 set_opt(sbi, READ_EXTENT_CACHE);
2156 clear_opt(sbi, DISABLE_CHECKPOINT);
2157
2158 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2159 set_opt(sbi, DISCARD);
2160
2161 if (f2fs_sb_has_blkzoned(sbi))
2162 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2163 else
2164 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2165 }
2166
2167 if (f2fs_sb_has_readonly(sbi))
2168 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2169 else
2170 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2171
2172 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2173 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2174 SMALL_VOLUME_SEGMENTS)
2175 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2176 else
2177 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2178 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2179 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2180 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2181 if (f2fs_sb_has_compression(sbi)) {
2182 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2183 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2184 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2185 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2186 }
2187 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2188 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2189 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2190
2191 set_opt(sbi, INLINE_XATTR);
2192 set_opt(sbi, INLINE_DATA);
2193 set_opt(sbi, INLINE_DENTRY);
2194 set_opt(sbi, MERGE_CHECKPOINT);
2195 set_opt(sbi, LAZYTIME);
2196 F2FS_OPTION(sbi).unusable_cap = 0;
2197 if (!f2fs_is_readonly(sbi))
2198 set_opt(sbi, FLUSH_MERGE);
2199 if (f2fs_sb_has_blkzoned(sbi))
2200 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2201 else
2202 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2203
2204 #ifdef CONFIG_F2FS_FS_XATTR
2205 set_opt(sbi, XATTR_USER);
2206 #endif
2207 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2208 set_opt(sbi, POSIX_ACL);
2209 #endif
2210
2211 f2fs_build_fault_attr(sbi, 0, 0);
2212 }
2213
2214 #ifdef CONFIG_QUOTA
2215 static int f2fs_enable_quotas(struct super_block *sb);
2216 #endif
2217
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)2218 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2219 {
2220 unsigned int s_flags = sbi->sb->s_flags;
2221 struct cp_control cpc;
2222 unsigned int gc_mode = sbi->gc_mode;
2223 int err = 0;
2224 int ret;
2225 block_t unusable;
2226
2227 if (s_flags & SB_RDONLY) {
2228 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2229 return -EINVAL;
2230 }
2231 sbi->sb->s_flags |= SB_ACTIVE;
2232
2233 /* check if we need more GC first */
2234 unusable = f2fs_get_unusable_blocks(sbi);
2235 if (!f2fs_disable_cp_again(sbi, unusable))
2236 goto skip_gc;
2237
2238 f2fs_update_time(sbi, DISABLE_TIME);
2239
2240 sbi->gc_mode = GC_URGENT_HIGH;
2241
2242 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2243 struct f2fs_gc_control gc_control = {
2244 .victim_segno = NULL_SEGNO,
2245 .init_gc_type = FG_GC,
2246 .should_migrate_blocks = false,
2247 .err_gc_skipped = true,
2248 .no_bg_gc = true,
2249 .nr_free_secs = 1 };
2250
2251 f2fs_down_write(&sbi->gc_lock);
2252 stat_inc_gc_call_count(sbi, FOREGROUND);
2253 err = f2fs_gc(sbi, &gc_control);
2254 if (err == -ENODATA) {
2255 err = 0;
2256 break;
2257 }
2258 if (err && err != -EAGAIN)
2259 break;
2260 }
2261
2262 ret = sync_filesystem(sbi->sb);
2263 if (ret || err) {
2264 err = ret ? ret : err;
2265 goto restore_flag;
2266 }
2267
2268 unusable = f2fs_get_unusable_blocks(sbi);
2269 if (f2fs_disable_cp_again(sbi, unusable)) {
2270 err = -EAGAIN;
2271 goto restore_flag;
2272 }
2273
2274 skip_gc:
2275 f2fs_down_write(&sbi->gc_lock);
2276 cpc.reason = CP_PAUSE;
2277 set_sbi_flag(sbi, SBI_CP_DISABLED);
2278 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2279 err = f2fs_write_checkpoint(sbi, &cpc);
2280 if (err)
2281 goto out_unlock;
2282
2283 spin_lock(&sbi->stat_lock);
2284 sbi->unusable_block_count = unusable;
2285 spin_unlock(&sbi->stat_lock);
2286
2287 out_unlock:
2288 f2fs_up_write(&sbi->gc_lock);
2289 restore_flag:
2290 sbi->gc_mode = gc_mode;
2291 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2292 return err;
2293 }
2294
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2295 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2296 {
2297 int retry = DEFAULT_RETRY_IO_COUNT;
2298
2299 /* we should flush all the data to keep data consistency */
2300 do {
2301 sync_inodes_sb(sbi->sb);
2302 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2303 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2304
2305 if (unlikely(retry < 0))
2306 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2307
2308 f2fs_down_write(&sbi->gc_lock);
2309 f2fs_dirty_to_prefree(sbi);
2310
2311 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2312 set_sbi_flag(sbi, SBI_IS_DIRTY);
2313 f2fs_up_write(&sbi->gc_lock);
2314
2315 f2fs_sync_fs(sbi->sb, 1);
2316
2317 /* Let's ensure there's no pending checkpoint anymore */
2318 f2fs_flush_ckpt_thread(sbi);
2319 }
2320
f2fs_remount(struct super_block * sb,int * flags,char * data)2321 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2322 {
2323 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2324 struct f2fs_mount_info org_mount_opt;
2325 unsigned long old_sb_flags;
2326 int err;
2327 bool need_restart_gc = false, need_stop_gc = false;
2328 bool need_restart_flush = false, need_stop_flush = false;
2329 bool need_restart_discard = false, need_stop_discard = false;
2330 bool need_enable_checkpoint = false, need_disable_checkpoint = false;
2331 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2332 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2333 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2334 bool no_atgc = !test_opt(sbi, ATGC);
2335 bool no_discard = !test_opt(sbi, DISCARD);
2336 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2337 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2338 bool no_nat_bits = !test_opt(sbi, NAT_BITS);
2339 #ifdef CONFIG_QUOTA
2340 int i, j;
2341 #endif
2342
2343 /*
2344 * Save the old mount options in case we
2345 * need to restore them.
2346 */
2347 org_mount_opt = sbi->mount_opt;
2348 old_sb_flags = sb->s_flags;
2349
2350 sbi->umount_lock_holder = current;
2351
2352 #ifdef CONFIG_QUOTA
2353 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2354 for (i = 0; i < MAXQUOTAS; i++) {
2355 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2356 org_mount_opt.s_qf_names[i] =
2357 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2358 GFP_KERNEL);
2359 if (!org_mount_opt.s_qf_names[i]) {
2360 for (j = 0; j < i; j++)
2361 kfree(org_mount_opt.s_qf_names[j]);
2362 return -ENOMEM;
2363 }
2364 } else {
2365 org_mount_opt.s_qf_names[i] = NULL;
2366 }
2367 }
2368 #endif
2369
2370 /* recover superblocks we couldn't write due to previous RO mount */
2371 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2372 err = f2fs_commit_super(sbi, false);
2373 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2374 err);
2375 if (!err)
2376 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2377 }
2378
2379 default_options(sbi, true);
2380
2381 /* parse mount options */
2382 err = parse_options(sbi, data, true);
2383 if (err)
2384 goto restore_opts;
2385
2386 #ifdef CONFIG_BLK_DEV_ZONED
2387 if (f2fs_sb_has_blkzoned(sbi) &&
2388 sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
2389 f2fs_err(sbi,
2390 "zoned: max open zones %u is too small, need at least %u open zones",
2391 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
2392 err = -EINVAL;
2393 goto restore_opts;
2394 }
2395 #endif
2396
2397 err = f2fs_default_check(sbi);
2398 if (err)
2399 goto restore_opts;
2400
2401 /* flush outstanding errors before changing fs state */
2402 flush_work(&sbi->s_error_work);
2403
2404 /*
2405 * Previous and new state of filesystem is RO,
2406 * so skip checking GC and FLUSH_MERGE conditions.
2407 */
2408 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2409 goto skip;
2410
2411 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2412 err = -EROFS;
2413 goto restore_opts;
2414 }
2415
2416 #ifdef CONFIG_QUOTA
2417 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2418 err = dquot_suspend(sb, -1);
2419 if (err < 0)
2420 goto restore_opts;
2421 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2422 /* dquot_resume needs RW */
2423 sb->s_flags &= ~SB_RDONLY;
2424 if (sb_any_quota_suspended(sb)) {
2425 dquot_resume(sb, -1);
2426 } else if (f2fs_sb_has_quota_ino(sbi)) {
2427 err = f2fs_enable_quotas(sb);
2428 if (err)
2429 goto restore_opts;
2430 }
2431 }
2432 #endif
2433 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2434 err = -EINVAL;
2435 f2fs_warn(sbi, "LFS is not compatible with IPU");
2436 goto restore_opts;
2437 }
2438
2439 /* disallow enable atgc dynamically */
2440 if (no_atgc == !!test_opt(sbi, ATGC)) {
2441 err = -EINVAL;
2442 f2fs_warn(sbi, "switch atgc option is not allowed");
2443 goto restore_opts;
2444 }
2445
2446 /* disallow enable/disable extent_cache dynamically */
2447 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2448 err = -EINVAL;
2449 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2450 goto restore_opts;
2451 }
2452 /* disallow enable/disable age extent_cache dynamically */
2453 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2454 err = -EINVAL;
2455 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2456 goto restore_opts;
2457 }
2458
2459 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2460 err = -EINVAL;
2461 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2462 goto restore_opts;
2463 }
2464
2465 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2466 err = -EINVAL;
2467 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2468 goto restore_opts;
2469 }
2470
2471 if (no_nat_bits == !!test_opt(sbi, NAT_BITS)) {
2472 err = -EINVAL;
2473 f2fs_warn(sbi, "switch nat_bits option is not allowed");
2474 goto restore_opts;
2475 }
2476
2477 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2478 err = -EINVAL;
2479 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2480 goto restore_opts;
2481 }
2482
2483 /*
2484 * We stop the GC thread if FS is mounted as RO
2485 * or if background_gc = off is passed in mount
2486 * option. Also sync the filesystem.
2487 */
2488 if ((*flags & SB_RDONLY) ||
2489 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2490 !test_opt(sbi, GC_MERGE))) {
2491 if (sbi->gc_thread) {
2492 f2fs_stop_gc_thread(sbi);
2493 need_restart_gc = true;
2494 }
2495 } else if (!sbi->gc_thread) {
2496 err = f2fs_start_gc_thread(sbi);
2497 if (err)
2498 goto restore_opts;
2499 need_stop_gc = true;
2500 }
2501
2502 if (*flags & SB_RDONLY) {
2503 sync_inodes_sb(sb);
2504
2505 set_sbi_flag(sbi, SBI_IS_DIRTY);
2506 set_sbi_flag(sbi, SBI_IS_CLOSE);
2507 f2fs_sync_fs(sb, 1);
2508 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2509 }
2510
2511 /*
2512 * We stop issue flush thread if FS is mounted as RO
2513 * or if flush_merge is not passed in mount option.
2514 */
2515 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2516 clear_opt(sbi, FLUSH_MERGE);
2517 f2fs_destroy_flush_cmd_control(sbi, false);
2518 need_restart_flush = true;
2519 } else {
2520 err = f2fs_create_flush_cmd_control(sbi);
2521 if (err)
2522 goto restore_gc;
2523 need_stop_flush = true;
2524 }
2525
2526 if (no_discard == !!test_opt(sbi, DISCARD)) {
2527 if (test_opt(sbi, DISCARD)) {
2528 err = f2fs_start_discard_thread(sbi);
2529 if (err)
2530 goto restore_flush;
2531 need_stop_discard = true;
2532 } else {
2533 f2fs_stop_discard_thread(sbi);
2534 f2fs_issue_discard_timeout(sbi);
2535 need_restart_discard = true;
2536 }
2537 }
2538
2539 adjust_unusable_cap_perc(sbi);
2540 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2541 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2542 err = f2fs_disable_checkpoint(sbi);
2543 if (err)
2544 goto restore_discard;
2545 need_enable_checkpoint = true;
2546 } else {
2547 f2fs_enable_checkpoint(sbi);
2548 need_disable_checkpoint = true;
2549 }
2550 }
2551
2552 /*
2553 * Place this routine at the end, since a new checkpoint would be
2554 * triggered while remount and we need to take care of it before
2555 * returning from remount.
2556 */
2557 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2558 !test_opt(sbi, MERGE_CHECKPOINT)) {
2559 f2fs_stop_ckpt_thread(sbi);
2560 } else {
2561 /* Flush if the prevous checkpoint, if exists. */
2562 f2fs_flush_ckpt_thread(sbi);
2563
2564 err = f2fs_start_ckpt_thread(sbi);
2565 if (err) {
2566 f2fs_err(sbi,
2567 "Failed to start F2FS issue_checkpoint_thread (%d)",
2568 err);
2569 goto restore_checkpoint;
2570 }
2571 }
2572
2573 skip:
2574 #ifdef CONFIG_QUOTA
2575 /* Release old quota file names */
2576 for (i = 0; i < MAXQUOTAS; i++)
2577 kfree(org_mount_opt.s_qf_names[i]);
2578 #endif
2579 /* Update the POSIXACL Flag */
2580 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2581 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2582
2583 limit_reserve_root(sbi);
2584 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2585
2586 sbi->umount_lock_holder = NULL;
2587 return 0;
2588 restore_checkpoint:
2589 if (need_enable_checkpoint) {
2590 f2fs_enable_checkpoint(sbi);
2591 } else if (need_disable_checkpoint) {
2592 if (f2fs_disable_checkpoint(sbi))
2593 f2fs_warn(sbi, "checkpoint has not been disabled");
2594 }
2595 restore_discard:
2596 if (need_restart_discard) {
2597 if (f2fs_start_discard_thread(sbi))
2598 f2fs_warn(sbi, "discard has been stopped");
2599 } else if (need_stop_discard) {
2600 f2fs_stop_discard_thread(sbi);
2601 }
2602 restore_flush:
2603 if (need_restart_flush) {
2604 if (f2fs_create_flush_cmd_control(sbi))
2605 f2fs_warn(sbi, "background flush thread has stopped");
2606 } else if (need_stop_flush) {
2607 clear_opt(sbi, FLUSH_MERGE);
2608 f2fs_destroy_flush_cmd_control(sbi, false);
2609 }
2610 restore_gc:
2611 if (need_restart_gc) {
2612 if (f2fs_start_gc_thread(sbi))
2613 f2fs_warn(sbi, "background gc thread has stopped");
2614 } else if (need_stop_gc) {
2615 f2fs_stop_gc_thread(sbi);
2616 }
2617 restore_opts:
2618 #ifdef CONFIG_QUOTA
2619 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2620 for (i = 0; i < MAXQUOTAS; i++) {
2621 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2622 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2623 }
2624 #endif
2625 sbi->mount_opt = org_mount_opt;
2626 sb->s_flags = old_sb_flags;
2627
2628 sbi->umount_lock_holder = NULL;
2629 return err;
2630 }
2631
f2fs_shutdown(struct super_block * sb)2632 static void f2fs_shutdown(struct super_block *sb)
2633 {
2634 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false, false);
2635 }
2636
2637 #ifdef CONFIG_QUOTA
f2fs_need_recovery(struct f2fs_sb_info * sbi)2638 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2639 {
2640 /* need to recovery orphan */
2641 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2642 return true;
2643 /* need to recovery data */
2644 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2645 return false;
2646 if (test_opt(sbi, NORECOVERY))
2647 return false;
2648 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2649 }
2650
f2fs_recover_quota_begin(struct f2fs_sb_info * sbi)2651 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2652 {
2653 bool readonly = f2fs_readonly(sbi->sb);
2654
2655 if (!f2fs_need_recovery(sbi))
2656 return false;
2657
2658 /* it doesn't need to check f2fs_sb_has_readonly() */
2659 if (f2fs_hw_is_readonly(sbi))
2660 return false;
2661
2662 if (readonly) {
2663 sbi->sb->s_flags &= ~SB_RDONLY;
2664 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2665 }
2666
2667 /*
2668 * Turn on quotas which were not enabled for read-only mounts if
2669 * filesystem has quota feature, so that they are updated correctly.
2670 */
2671 return f2fs_enable_quota_files(sbi, readonly);
2672 }
2673
f2fs_recover_quota_end(struct f2fs_sb_info * sbi,bool quota_enabled)2674 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2675 bool quota_enabled)
2676 {
2677 if (quota_enabled)
2678 f2fs_quota_off_umount(sbi->sb);
2679
2680 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2681 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2682 sbi->sb->s_flags |= SB_RDONLY;
2683 }
2684 }
2685
2686 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2687 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2688 size_t len, loff_t off)
2689 {
2690 struct inode *inode = sb_dqopt(sb)->files[type];
2691 struct address_space *mapping = inode->i_mapping;
2692 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2693 int offset = off & (sb->s_blocksize - 1);
2694 int tocopy;
2695 size_t toread;
2696 loff_t i_size = i_size_read(inode);
2697 struct page *page;
2698
2699 if (off > i_size)
2700 return 0;
2701
2702 if (off + len > i_size)
2703 len = i_size - off;
2704 toread = len;
2705 while (toread > 0) {
2706 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2707 repeat:
2708 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2709 if (IS_ERR(page)) {
2710 if (PTR_ERR(page) == -ENOMEM) {
2711 memalloc_retry_wait(GFP_NOFS);
2712 goto repeat;
2713 }
2714 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2715 return PTR_ERR(page);
2716 }
2717
2718 lock_page(page);
2719
2720 if (unlikely(page->mapping != mapping)) {
2721 f2fs_put_page(page, 1);
2722 goto repeat;
2723 }
2724 if (unlikely(!PageUptodate(page))) {
2725 f2fs_put_page(page, 1);
2726 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2727 return -EIO;
2728 }
2729
2730 memcpy_from_page(data, page, offset, tocopy);
2731 f2fs_put_page(page, 1);
2732
2733 offset = 0;
2734 toread -= tocopy;
2735 data += tocopy;
2736 blkidx++;
2737 }
2738 return len;
2739 }
2740
2741 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2742 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2743 const char *data, size_t len, loff_t off)
2744 {
2745 struct inode *inode = sb_dqopt(sb)->files[type];
2746 struct address_space *mapping = inode->i_mapping;
2747 const struct address_space_operations *a_ops = mapping->a_ops;
2748 int offset = off & (sb->s_blocksize - 1);
2749 size_t towrite = len;
2750 struct folio *folio;
2751 void *fsdata = NULL;
2752 int err = 0;
2753 int tocopy;
2754
2755 while (towrite > 0) {
2756 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2757 towrite);
2758 retry:
2759 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2760 &folio, &fsdata);
2761 if (unlikely(err)) {
2762 if (err == -ENOMEM) {
2763 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2764 goto retry;
2765 }
2766 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2767 break;
2768 }
2769
2770 memcpy_to_folio(folio, offset_in_folio(folio, off), data, tocopy);
2771
2772 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2773 folio, fsdata);
2774 offset = 0;
2775 towrite -= tocopy;
2776 off += tocopy;
2777 data += tocopy;
2778 cond_resched();
2779 }
2780
2781 if (len == towrite)
2782 return err;
2783 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2784 f2fs_mark_inode_dirty_sync(inode, false);
2785 return len - towrite;
2786 }
2787
f2fs_dquot_initialize(struct inode * inode)2788 int f2fs_dquot_initialize(struct inode *inode)
2789 {
2790 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2791 return -ESRCH;
2792
2793 return dquot_initialize(inode);
2794 }
2795
f2fs_get_dquots(struct inode * inode)2796 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2797 {
2798 return F2FS_I(inode)->i_dquot;
2799 }
2800
f2fs_get_reserved_space(struct inode * inode)2801 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2802 {
2803 return &F2FS_I(inode)->i_reserved_quota;
2804 }
2805
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2806 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2807 {
2808 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2809 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2810 return 0;
2811 }
2812
2813 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2814 F2FS_OPTION(sbi).s_jquota_fmt, type);
2815 }
2816
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2817 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2818 {
2819 int enabled = 0;
2820 int i, err;
2821
2822 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2823 err = f2fs_enable_quotas(sbi->sb);
2824 if (err) {
2825 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2826 return 0;
2827 }
2828 return 1;
2829 }
2830
2831 for (i = 0; i < MAXQUOTAS; i++) {
2832 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2833 err = f2fs_quota_on_mount(sbi, i);
2834 if (!err) {
2835 enabled = 1;
2836 continue;
2837 }
2838 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2839 err, i);
2840 }
2841 }
2842 return enabled;
2843 }
2844
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2845 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2846 unsigned int flags)
2847 {
2848 struct inode *qf_inode;
2849 unsigned long qf_inum;
2850 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2851 int err;
2852
2853 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2854
2855 qf_inum = f2fs_qf_ino(sb, type);
2856 if (!qf_inum)
2857 return -EPERM;
2858
2859 qf_inode = f2fs_iget(sb, qf_inum);
2860 if (IS_ERR(qf_inode)) {
2861 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2862 return PTR_ERR(qf_inode);
2863 }
2864
2865 /* Don't account quota for quota files to avoid recursion */
2866 inode_lock(qf_inode);
2867 qf_inode->i_flags |= S_NOQUOTA;
2868
2869 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2870 F2FS_I(qf_inode)->i_flags |= qf_flag;
2871 f2fs_set_inode_flags(qf_inode);
2872 }
2873 inode_unlock(qf_inode);
2874
2875 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2876 iput(qf_inode);
2877 return err;
2878 }
2879
f2fs_enable_quotas(struct super_block * sb)2880 static int f2fs_enable_quotas(struct super_block *sb)
2881 {
2882 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2883 int type, err = 0;
2884 unsigned long qf_inum;
2885 bool quota_mopt[MAXQUOTAS] = {
2886 test_opt(sbi, USRQUOTA),
2887 test_opt(sbi, GRPQUOTA),
2888 test_opt(sbi, PRJQUOTA),
2889 };
2890
2891 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2892 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2893 return 0;
2894 }
2895
2896 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2897
2898 for (type = 0; type < MAXQUOTAS; type++) {
2899 qf_inum = f2fs_qf_ino(sb, type);
2900 if (qf_inum) {
2901 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2902 DQUOT_USAGE_ENABLED |
2903 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2904 if (err) {
2905 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2906 type, err);
2907 for (type--; type >= 0; type--)
2908 dquot_quota_off(sb, type);
2909 set_sbi_flag(F2FS_SB(sb),
2910 SBI_QUOTA_NEED_REPAIR);
2911 return err;
2912 }
2913 }
2914 }
2915 return 0;
2916 }
2917
f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2918 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2919 {
2920 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2921 struct address_space *mapping = dqopt->files[type]->i_mapping;
2922 int ret = 0;
2923
2924 ret = dquot_writeback_dquots(sbi->sb, type);
2925 if (ret)
2926 goto out;
2927
2928 ret = filemap_fdatawrite(mapping);
2929 if (ret)
2930 goto out;
2931
2932 /* if we are using journalled quota */
2933 if (is_journalled_quota(sbi))
2934 goto out;
2935
2936 ret = filemap_fdatawait(mapping);
2937
2938 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2939 out:
2940 if (ret)
2941 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2942 return ret;
2943 }
2944
f2fs_do_quota_sync(struct super_block * sb,int type)2945 int f2fs_do_quota_sync(struct super_block *sb, int type)
2946 {
2947 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2948 struct quota_info *dqopt = sb_dqopt(sb);
2949 int cnt;
2950 int ret = 0;
2951
2952 /*
2953 * Now when everything is written we can discard the pagecache so
2954 * that userspace sees the changes.
2955 */
2956 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2957
2958 if (type != -1 && cnt != type)
2959 continue;
2960
2961 if (!sb_has_quota_active(sb, cnt))
2962 continue;
2963
2964 if (!f2fs_sb_has_quota_ino(sbi))
2965 inode_lock(dqopt->files[cnt]);
2966
2967 /*
2968 * do_quotactl
2969 * f2fs_quota_sync
2970 * f2fs_down_read(quota_sem)
2971 * dquot_writeback_dquots()
2972 * f2fs_dquot_commit
2973 * block_operation
2974 * f2fs_down_read(quota_sem)
2975 */
2976 f2fs_lock_op(sbi);
2977 f2fs_down_read(&sbi->quota_sem);
2978
2979 ret = f2fs_quota_sync_file(sbi, cnt);
2980
2981 f2fs_up_read(&sbi->quota_sem);
2982 f2fs_unlock_op(sbi);
2983
2984 if (!f2fs_sb_has_quota_ino(sbi))
2985 inode_unlock(dqopt->files[cnt]);
2986
2987 if (ret)
2988 break;
2989 }
2990 return ret;
2991 }
2992
f2fs_quota_sync(struct super_block * sb,int type)2993 static int f2fs_quota_sync(struct super_block *sb, int type)
2994 {
2995 int ret;
2996
2997 F2FS_SB(sb)->umount_lock_holder = current;
2998 ret = f2fs_do_quota_sync(sb, type);
2999 F2FS_SB(sb)->umount_lock_holder = NULL;
3000 return ret;
3001 }
3002
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)3003 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
3004 const struct path *path)
3005 {
3006 struct inode *inode;
3007 int err = 0;
3008
3009 /* if quota sysfile exists, deny enabling quota with specific file */
3010 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
3011 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
3012 return -EBUSY;
3013 }
3014
3015 if (path->dentry->d_sb != sb)
3016 return -EXDEV;
3017
3018 F2FS_SB(sb)->umount_lock_holder = current;
3019
3020 err = f2fs_do_quota_sync(sb, type);
3021 if (err)
3022 goto out;
3023
3024 inode = d_inode(path->dentry);
3025
3026 err = filemap_fdatawrite(inode->i_mapping);
3027 if (err)
3028 goto out;
3029
3030 err = filemap_fdatawait(inode->i_mapping);
3031 if (err)
3032 goto out;
3033
3034 err = dquot_quota_on(sb, type, format_id, path);
3035 if (err)
3036 goto out;
3037
3038 inode_lock(inode);
3039 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
3040 f2fs_set_inode_flags(inode);
3041 inode_unlock(inode);
3042 f2fs_mark_inode_dirty_sync(inode, false);
3043 out:
3044 F2FS_SB(sb)->umount_lock_holder = NULL;
3045 return err;
3046 }
3047
__f2fs_quota_off(struct super_block * sb,int type)3048 static int __f2fs_quota_off(struct super_block *sb, int type)
3049 {
3050 struct inode *inode = sb_dqopt(sb)->files[type];
3051 int err;
3052
3053 if (!inode || !igrab(inode))
3054 return dquot_quota_off(sb, type);
3055
3056 err = f2fs_do_quota_sync(sb, type);
3057 if (err)
3058 goto out_put;
3059
3060 err = dquot_quota_off(sb, type);
3061 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
3062 goto out_put;
3063
3064 inode_lock(inode);
3065 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
3066 f2fs_set_inode_flags(inode);
3067 inode_unlock(inode);
3068 f2fs_mark_inode_dirty_sync(inode, false);
3069 out_put:
3070 iput(inode);
3071 return err;
3072 }
3073
f2fs_quota_off(struct super_block * sb,int type)3074 static int f2fs_quota_off(struct super_block *sb, int type)
3075 {
3076 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3077 int err;
3078
3079 F2FS_SB(sb)->umount_lock_holder = current;
3080
3081 err = __f2fs_quota_off(sb, type);
3082
3083 /*
3084 * quotactl can shutdown journalled quota, result in inconsistence
3085 * between quota record and fs data by following updates, tag the
3086 * flag to let fsck be aware of it.
3087 */
3088 if (is_journalled_quota(sbi))
3089 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3090
3091 F2FS_SB(sb)->umount_lock_holder = NULL;
3092
3093 return err;
3094 }
3095
f2fs_quota_off_umount(struct super_block * sb)3096 void f2fs_quota_off_umount(struct super_block *sb)
3097 {
3098 int type;
3099 int err;
3100
3101 for (type = 0; type < MAXQUOTAS; type++) {
3102 err = __f2fs_quota_off(sb, type);
3103 if (err) {
3104 int ret = dquot_quota_off(sb, type);
3105
3106 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3107 type, err, ret);
3108 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3109 }
3110 }
3111 /*
3112 * In case of checkpoint=disable, we must flush quota blocks.
3113 * This can cause NULL exception for node_inode in end_io, since
3114 * put_super already dropped it.
3115 */
3116 sync_filesystem(sb);
3117 }
3118
f2fs_truncate_quota_inode_pages(struct super_block * sb)3119 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3120 {
3121 struct quota_info *dqopt = sb_dqopt(sb);
3122 int type;
3123
3124 for (type = 0; type < MAXQUOTAS; type++) {
3125 if (!dqopt->files[type])
3126 continue;
3127 f2fs_inode_synced(dqopt->files[type]);
3128 }
3129 }
3130
f2fs_dquot_commit(struct dquot * dquot)3131 static int f2fs_dquot_commit(struct dquot *dquot)
3132 {
3133 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3134 int ret;
3135
3136 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3137 ret = dquot_commit(dquot);
3138 if (ret < 0)
3139 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3140 f2fs_up_read(&sbi->quota_sem);
3141 return ret;
3142 }
3143
f2fs_dquot_acquire(struct dquot * dquot)3144 static int f2fs_dquot_acquire(struct dquot *dquot)
3145 {
3146 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3147 int ret;
3148
3149 f2fs_down_read(&sbi->quota_sem);
3150 ret = dquot_acquire(dquot);
3151 if (ret < 0)
3152 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3153 f2fs_up_read(&sbi->quota_sem);
3154 return ret;
3155 }
3156
f2fs_dquot_release(struct dquot * dquot)3157 static int f2fs_dquot_release(struct dquot *dquot)
3158 {
3159 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3160 int ret = dquot_release(dquot);
3161
3162 if (ret < 0)
3163 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3164 return ret;
3165 }
3166
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)3167 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3168 {
3169 struct super_block *sb = dquot->dq_sb;
3170 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3171 int ret = dquot_mark_dquot_dirty(dquot);
3172
3173 /* if we are using journalled quota */
3174 if (is_journalled_quota(sbi))
3175 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3176
3177 return ret;
3178 }
3179
f2fs_dquot_commit_info(struct super_block * sb,int type)3180 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3181 {
3182 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3183 int ret = dquot_commit_info(sb, type);
3184
3185 if (ret < 0)
3186 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3187 return ret;
3188 }
3189
f2fs_get_projid(struct inode * inode,kprojid_t * projid)3190 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3191 {
3192 *projid = F2FS_I(inode)->i_projid;
3193 return 0;
3194 }
3195
3196 static const struct dquot_operations f2fs_quota_operations = {
3197 .get_reserved_space = f2fs_get_reserved_space,
3198 .write_dquot = f2fs_dquot_commit,
3199 .acquire_dquot = f2fs_dquot_acquire,
3200 .release_dquot = f2fs_dquot_release,
3201 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3202 .write_info = f2fs_dquot_commit_info,
3203 .alloc_dquot = dquot_alloc,
3204 .destroy_dquot = dquot_destroy,
3205 .get_projid = f2fs_get_projid,
3206 .get_next_id = dquot_get_next_id,
3207 };
3208
3209 static const struct quotactl_ops f2fs_quotactl_ops = {
3210 .quota_on = f2fs_quota_on,
3211 .quota_off = f2fs_quota_off,
3212 .quota_sync = f2fs_quota_sync,
3213 .get_state = dquot_get_state,
3214 .set_info = dquot_set_dqinfo,
3215 .get_dqblk = dquot_get_dqblk,
3216 .set_dqblk = dquot_set_dqblk,
3217 .get_nextdqblk = dquot_get_next_dqblk,
3218 };
3219 #else
f2fs_dquot_initialize(struct inode * inode)3220 int f2fs_dquot_initialize(struct inode *inode)
3221 {
3222 return 0;
3223 }
3224
f2fs_do_quota_sync(struct super_block * sb,int type)3225 int f2fs_do_quota_sync(struct super_block *sb, int type)
3226 {
3227 return 0;
3228 }
3229
f2fs_quota_off_umount(struct super_block * sb)3230 void f2fs_quota_off_umount(struct super_block *sb)
3231 {
3232 }
3233 #endif
3234
3235 static const struct super_operations f2fs_sops = {
3236 .alloc_inode = f2fs_alloc_inode,
3237 .free_inode = f2fs_free_inode,
3238 .drop_inode = f2fs_drop_inode,
3239 .write_inode = f2fs_write_inode,
3240 .dirty_inode = f2fs_dirty_inode,
3241 .show_options = f2fs_show_options,
3242 #ifdef CONFIG_QUOTA
3243 .quota_read = f2fs_quota_read,
3244 .quota_write = f2fs_quota_write,
3245 .get_dquots = f2fs_get_dquots,
3246 #endif
3247 .evict_inode = f2fs_evict_inode,
3248 .put_super = f2fs_put_super,
3249 .sync_fs = f2fs_sync_fs,
3250 .freeze_fs = f2fs_freeze,
3251 .unfreeze_fs = f2fs_unfreeze,
3252 .statfs = f2fs_statfs,
3253 .remount_fs = f2fs_remount,
3254 .shutdown = f2fs_shutdown,
3255 };
3256
3257 #ifdef CONFIG_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)3258 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3259 {
3260 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3261 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3262 ctx, len, NULL);
3263 }
3264
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)3265 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3266 void *fs_data)
3267 {
3268 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3269
3270 /*
3271 * Encrypting the root directory is not allowed because fsck
3272 * expects lost+found directory to exist and remain unencrypted
3273 * if LOST_FOUND feature is enabled.
3274 *
3275 */
3276 if (f2fs_sb_has_lost_found(sbi) &&
3277 inode->i_ino == F2FS_ROOT_INO(sbi))
3278 return -EPERM;
3279
3280 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3281 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3282 ctx, len, fs_data, XATTR_CREATE);
3283 }
3284
f2fs_get_dummy_policy(struct super_block * sb)3285 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3286 {
3287 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3288 }
3289
f2fs_has_stable_inodes(struct super_block * sb)3290 static bool f2fs_has_stable_inodes(struct super_block *sb)
3291 {
3292 return true;
3293 }
3294
f2fs_get_devices(struct super_block * sb,unsigned int * num_devs)3295 static struct block_device **f2fs_get_devices(struct super_block *sb,
3296 unsigned int *num_devs)
3297 {
3298 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3299 struct block_device **devs;
3300 int i;
3301
3302 if (!f2fs_is_multi_device(sbi))
3303 return NULL;
3304
3305 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3306 if (!devs)
3307 return ERR_PTR(-ENOMEM);
3308
3309 for (i = 0; i < sbi->s_ndevs; i++)
3310 devs[i] = FDEV(i).bdev;
3311 *num_devs = sbi->s_ndevs;
3312 return devs;
3313 }
3314
3315 static const struct fscrypt_operations f2fs_cryptops = {
3316 .needs_bounce_pages = 1,
3317 .has_32bit_inodes = 1,
3318 .supports_subblock_data_units = 1,
3319 .legacy_key_prefix = "f2fs:",
3320 .get_context = f2fs_get_context,
3321 .set_context = f2fs_set_context,
3322 .get_dummy_policy = f2fs_get_dummy_policy,
3323 .empty_dir = f2fs_empty_dir,
3324 .has_stable_inodes = f2fs_has_stable_inodes,
3325 .get_devices = f2fs_get_devices,
3326 };
3327 #endif
3328
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)3329 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3330 u64 ino, u32 generation)
3331 {
3332 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3333 struct inode *inode;
3334
3335 if (f2fs_check_nid_range(sbi, ino))
3336 return ERR_PTR(-ESTALE);
3337
3338 /*
3339 * f2fs_iget isn't quite right if the inode is currently unallocated!
3340 * However f2fs_iget currently does appropriate checks to handle stale
3341 * inodes so everything is OK.
3342 */
3343 inode = f2fs_iget(sb, ino);
3344 if (IS_ERR(inode))
3345 return ERR_CAST(inode);
3346 if (unlikely(generation && inode->i_generation != generation)) {
3347 /* we didn't find the right inode.. */
3348 iput(inode);
3349 return ERR_PTR(-ESTALE);
3350 }
3351 return inode;
3352 }
3353
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3354 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3355 int fh_len, int fh_type)
3356 {
3357 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3358 f2fs_nfs_get_inode);
3359 }
3360
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3361 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3362 int fh_len, int fh_type)
3363 {
3364 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3365 f2fs_nfs_get_inode);
3366 }
3367
3368 static const struct export_operations f2fs_export_ops = {
3369 .encode_fh = generic_encode_ino32_fh,
3370 .fh_to_dentry = f2fs_fh_to_dentry,
3371 .fh_to_parent = f2fs_fh_to_parent,
3372 .get_parent = f2fs_get_parent,
3373 };
3374
max_file_blocks(struct inode * inode)3375 loff_t max_file_blocks(struct inode *inode)
3376 {
3377 loff_t result = 0;
3378 loff_t leaf_count;
3379
3380 /*
3381 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3382 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3383 * space in inode.i_addr, it will be more safe to reassign
3384 * result as zero.
3385 */
3386
3387 if (inode && f2fs_compressed_file(inode))
3388 leaf_count = ADDRS_PER_BLOCK(inode);
3389 else
3390 leaf_count = DEF_ADDRS_PER_BLOCK;
3391
3392 /* two direct node blocks */
3393 result += (leaf_count * 2);
3394
3395 /* two indirect node blocks */
3396 leaf_count *= NIDS_PER_BLOCK;
3397 result += (leaf_count * 2);
3398
3399 /* one double indirect node block */
3400 leaf_count *= NIDS_PER_BLOCK;
3401 result += leaf_count;
3402
3403 /*
3404 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with
3405 * a 4K crypto data unit, we must restrict the max filesize to what can
3406 * fit within U32_MAX + 1 data units.
3407 */
3408
3409 result = umin(result, F2FS_BYTES_TO_BLK(((loff_t)U32_MAX + 1) * 4096));
3410
3411 return result;
3412 }
3413
__f2fs_commit_super(struct f2fs_sb_info * sbi,struct folio * folio,pgoff_t index,bool update)3414 static int __f2fs_commit_super(struct f2fs_sb_info *sbi, struct folio *folio,
3415 pgoff_t index, bool update)
3416 {
3417 struct bio *bio;
3418 /* it's rare case, we can do fua all the time */
3419 blk_opf_t opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH | REQ_FUA;
3420 int ret;
3421
3422 folio_lock(folio);
3423 folio_wait_writeback(folio);
3424 if (update)
3425 memcpy(F2FS_SUPER_BLOCK(folio, index), F2FS_RAW_SUPER(sbi),
3426 sizeof(struct f2fs_super_block));
3427 folio_mark_dirty(folio);
3428 folio_clear_dirty_for_io(folio);
3429 folio_start_writeback(folio);
3430 folio_unlock(folio);
3431
3432 bio = bio_alloc(sbi->sb->s_bdev, 1, opf, GFP_NOFS);
3433
3434 /* it doesn't need to set crypto context for superblock update */
3435 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(folio_index(folio));
3436
3437 if (!bio_add_folio(bio, folio, folio_size(folio), 0))
3438 f2fs_bug_on(sbi, 1);
3439
3440 ret = submit_bio_wait(bio);
3441 folio_end_writeback(folio);
3442
3443 return ret;
3444 }
3445
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct folio * folio,pgoff_t index)3446 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3447 struct folio *folio, pgoff_t index)
3448 {
3449 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index);
3450 struct super_block *sb = sbi->sb;
3451 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3452 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3453 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3454 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3455 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3456 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3457 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3458 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3459 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3460 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3461 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3462 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3463 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3464 u64 main_end_blkaddr = main_blkaddr +
3465 ((u64)segment_count_main << log_blocks_per_seg);
3466 u64 seg_end_blkaddr = segment0_blkaddr +
3467 ((u64)segment_count << log_blocks_per_seg);
3468
3469 if (segment0_blkaddr != cp_blkaddr) {
3470 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3471 segment0_blkaddr, cp_blkaddr);
3472 return true;
3473 }
3474
3475 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3476 sit_blkaddr) {
3477 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3478 cp_blkaddr, sit_blkaddr,
3479 segment_count_ckpt << log_blocks_per_seg);
3480 return true;
3481 }
3482
3483 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3484 nat_blkaddr) {
3485 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3486 sit_blkaddr, nat_blkaddr,
3487 segment_count_sit << log_blocks_per_seg);
3488 return true;
3489 }
3490
3491 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3492 ssa_blkaddr) {
3493 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3494 nat_blkaddr, ssa_blkaddr,
3495 segment_count_nat << log_blocks_per_seg);
3496 return true;
3497 }
3498
3499 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3500 main_blkaddr) {
3501 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3502 ssa_blkaddr, main_blkaddr,
3503 segment_count_ssa << log_blocks_per_seg);
3504 return true;
3505 }
3506
3507 if (main_end_blkaddr > seg_end_blkaddr) {
3508 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3509 main_blkaddr, seg_end_blkaddr,
3510 segment_count_main << log_blocks_per_seg);
3511 return true;
3512 } else if (main_end_blkaddr < seg_end_blkaddr) {
3513 int err = 0;
3514 char *res;
3515
3516 /* fix in-memory information all the time */
3517 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3518 segment0_blkaddr) >> log_blocks_per_seg);
3519
3520 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3521 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3522 res = "internally";
3523 } else {
3524 err = __f2fs_commit_super(sbi, folio, index, false);
3525 res = err ? "failed" : "done";
3526 }
3527 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3528 res, main_blkaddr, seg_end_blkaddr,
3529 segment_count_main << log_blocks_per_seg);
3530 if (err)
3531 return true;
3532 }
3533 return false;
3534 }
3535
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct folio * folio,pgoff_t index)3536 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3537 struct folio *folio, pgoff_t index)
3538 {
3539 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3540 block_t total_sections, blocks_per_seg;
3541 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index);
3542 size_t crc_offset = 0;
3543 __u32 crc = 0;
3544
3545 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3546 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3547 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3548 return -EINVAL;
3549 }
3550
3551 /* Check checksum_offset and crc in superblock */
3552 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3553 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3554 if (crc_offset !=
3555 offsetof(struct f2fs_super_block, crc)) {
3556 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3557 crc_offset);
3558 return -EFSCORRUPTED;
3559 }
3560 crc = le32_to_cpu(raw_super->crc);
3561 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3562 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3563 return -EFSCORRUPTED;
3564 }
3565 }
3566
3567 /* only support block_size equals to PAGE_SIZE */
3568 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3569 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3570 le32_to_cpu(raw_super->log_blocksize),
3571 F2FS_BLKSIZE_BITS);
3572 return -EFSCORRUPTED;
3573 }
3574
3575 /* check log blocks per segment */
3576 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3577 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3578 le32_to_cpu(raw_super->log_blocks_per_seg));
3579 return -EFSCORRUPTED;
3580 }
3581
3582 /* Currently, support 512/1024/2048/4096/16K bytes sector size */
3583 if (le32_to_cpu(raw_super->log_sectorsize) >
3584 F2FS_MAX_LOG_SECTOR_SIZE ||
3585 le32_to_cpu(raw_super->log_sectorsize) <
3586 F2FS_MIN_LOG_SECTOR_SIZE) {
3587 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3588 le32_to_cpu(raw_super->log_sectorsize));
3589 return -EFSCORRUPTED;
3590 }
3591 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3592 le32_to_cpu(raw_super->log_sectorsize) !=
3593 F2FS_MAX_LOG_SECTOR_SIZE) {
3594 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3595 le32_to_cpu(raw_super->log_sectors_per_block),
3596 le32_to_cpu(raw_super->log_sectorsize));
3597 return -EFSCORRUPTED;
3598 }
3599
3600 segment_count = le32_to_cpu(raw_super->segment_count);
3601 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3602 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3603 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3604 total_sections = le32_to_cpu(raw_super->section_count);
3605
3606 /* blocks_per_seg should be 512, given the above check */
3607 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3608
3609 if (segment_count > F2FS_MAX_SEGMENT ||
3610 segment_count < F2FS_MIN_SEGMENTS) {
3611 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3612 return -EFSCORRUPTED;
3613 }
3614
3615 if (total_sections > segment_count_main || total_sections < 1 ||
3616 segs_per_sec > segment_count || !segs_per_sec) {
3617 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3618 segment_count, total_sections, segs_per_sec);
3619 return -EFSCORRUPTED;
3620 }
3621
3622 if (segment_count_main != total_sections * segs_per_sec) {
3623 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3624 segment_count_main, total_sections, segs_per_sec);
3625 return -EFSCORRUPTED;
3626 }
3627
3628 if ((segment_count / segs_per_sec) < total_sections) {
3629 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3630 segment_count, segs_per_sec, total_sections);
3631 return -EFSCORRUPTED;
3632 }
3633
3634 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3635 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3636 segment_count, le64_to_cpu(raw_super->block_count));
3637 return -EFSCORRUPTED;
3638 }
3639
3640 if (RDEV(0).path[0]) {
3641 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3642 int i = 1;
3643
3644 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3645 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3646 i++;
3647 }
3648 if (segment_count != dev_seg_count) {
3649 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3650 segment_count, dev_seg_count);
3651 return -EFSCORRUPTED;
3652 }
3653 } else {
3654 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3655 !bdev_is_zoned(sbi->sb->s_bdev)) {
3656 f2fs_info(sbi, "Zoned block device path is missing");
3657 return -EFSCORRUPTED;
3658 }
3659 }
3660
3661 if (secs_per_zone > total_sections || !secs_per_zone) {
3662 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3663 secs_per_zone, total_sections);
3664 return -EFSCORRUPTED;
3665 }
3666 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3667 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3668 (le32_to_cpu(raw_super->extension_count) +
3669 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3670 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3671 le32_to_cpu(raw_super->extension_count),
3672 raw_super->hot_ext_count,
3673 F2FS_MAX_EXTENSION);
3674 return -EFSCORRUPTED;
3675 }
3676
3677 if (le32_to_cpu(raw_super->cp_payload) >=
3678 (blocks_per_seg - F2FS_CP_PACKS -
3679 NR_CURSEG_PERSIST_TYPE)) {
3680 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3681 le32_to_cpu(raw_super->cp_payload),
3682 blocks_per_seg - F2FS_CP_PACKS -
3683 NR_CURSEG_PERSIST_TYPE);
3684 return -EFSCORRUPTED;
3685 }
3686
3687 /* check reserved ino info */
3688 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3689 le32_to_cpu(raw_super->meta_ino) != 2 ||
3690 le32_to_cpu(raw_super->root_ino) != 3) {
3691 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3692 le32_to_cpu(raw_super->node_ino),
3693 le32_to_cpu(raw_super->meta_ino),
3694 le32_to_cpu(raw_super->root_ino));
3695 return -EFSCORRUPTED;
3696 }
3697
3698 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3699 if (sanity_check_area_boundary(sbi, folio, index))
3700 return -EFSCORRUPTED;
3701
3702 return 0;
3703 }
3704
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3705 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3706 {
3707 unsigned int total, fsmeta;
3708 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3709 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3710 unsigned int ovp_segments, reserved_segments;
3711 unsigned int main_segs, blocks_per_seg;
3712 unsigned int sit_segs, nat_segs;
3713 unsigned int sit_bitmap_size, nat_bitmap_size;
3714 unsigned int log_blocks_per_seg;
3715 unsigned int segment_count_main;
3716 unsigned int cp_pack_start_sum, cp_payload;
3717 block_t user_block_count, valid_user_blocks;
3718 block_t avail_node_count, valid_node_count;
3719 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3720 int i, j;
3721
3722 total = le32_to_cpu(raw_super->segment_count);
3723 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3724 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3725 fsmeta += sit_segs;
3726 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3727 fsmeta += nat_segs;
3728 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3729 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3730
3731 if (unlikely(fsmeta >= total))
3732 return 1;
3733
3734 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3735 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3736
3737 if (!f2fs_sb_has_readonly(sbi) &&
3738 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3739 ovp_segments == 0 || reserved_segments == 0)) {
3740 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3741 return 1;
3742 }
3743 user_block_count = le64_to_cpu(ckpt->user_block_count);
3744 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3745 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3746 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3747 if (!user_block_count || user_block_count >=
3748 segment_count_main << log_blocks_per_seg) {
3749 f2fs_err(sbi, "Wrong user_block_count: %u",
3750 user_block_count);
3751 return 1;
3752 }
3753
3754 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3755 if (valid_user_blocks > user_block_count) {
3756 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3757 valid_user_blocks, user_block_count);
3758 return 1;
3759 }
3760
3761 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3762 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3763 if (valid_node_count > avail_node_count) {
3764 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3765 valid_node_count, avail_node_count);
3766 return 1;
3767 }
3768
3769 main_segs = le32_to_cpu(raw_super->segment_count_main);
3770 blocks_per_seg = BLKS_PER_SEG(sbi);
3771
3772 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3773 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3774 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3775 return 1;
3776
3777 if (f2fs_sb_has_readonly(sbi))
3778 goto check_data;
3779
3780 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3781 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3782 le32_to_cpu(ckpt->cur_node_segno[j])) {
3783 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3784 i, j,
3785 le32_to_cpu(ckpt->cur_node_segno[i]));
3786 return 1;
3787 }
3788 }
3789 }
3790 check_data:
3791 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3792 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3793 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3794 return 1;
3795
3796 if (f2fs_sb_has_readonly(sbi))
3797 goto skip_cross;
3798
3799 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3800 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3801 le32_to_cpu(ckpt->cur_data_segno[j])) {
3802 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3803 i, j,
3804 le32_to_cpu(ckpt->cur_data_segno[i]));
3805 return 1;
3806 }
3807 }
3808 }
3809 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3810 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3811 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3812 le32_to_cpu(ckpt->cur_data_segno[j])) {
3813 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3814 i, j,
3815 le32_to_cpu(ckpt->cur_node_segno[i]));
3816 return 1;
3817 }
3818 }
3819 }
3820 skip_cross:
3821 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3822 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3823
3824 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3825 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3826 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3827 sit_bitmap_size, nat_bitmap_size);
3828 return 1;
3829 }
3830
3831 cp_pack_start_sum = __start_sum_addr(sbi);
3832 cp_payload = __cp_payload(sbi);
3833 if (cp_pack_start_sum < cp_payload + 1 ||
3834 cp_pack_start_sum > blocks_per_seg - 1 -
3835 NR_CURSEG_PERSIST_TYPE) {
3836 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3837 cp_pack_start_sum);
3838 return 1;
3839 }
3840
3841 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3842 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3843 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3844 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3845 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3846 le32_to_cpu(ckpt->checksum_offset));
3847 return 1;
3848 }
3849
3850 nat_blocks = nat_segs << log_blocks_per_seg;
3851 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3852 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3853 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3854 (cp_payload + F2FS_CP_PACKS +
3855 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3856 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3857 cp_payload, nat_bits_blocks);
3858 return 1;
3859 }
3860
3861 if (unlikely(f2fs_cp_error(sbi))) {
3862 f2fs_err(sbi, "A bug case: need to run fsck");
3863 return 1;
3864 }
3865 return 0;
3866 }
3867
init_sb_info(struct f2fs_sb_info * sbi)3868 static void init_sb_info(struct f2fs_sb_info *sbi)
3869 {
3870 struct f2fs_super_block *raw_super = sbi->raw_super;
3871 int i;
3872
3873 sbi->log_sectors_per_block =
3874 le32_to_cpu(raw_super->log_sectors_per_block);
3875 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3876 sbi->blocksize = BIT(sbi->log_blocksize);
3877 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3878 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3879 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3880 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3881 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3882 sbi->total_node_count = SEGS_TO_BLKS(sbi,
3883 ((le32_to_cpu(raw_super->segment_count_nat) / 2) *
3884 NAT_ENTRY_PER_BLOCK));
3885 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3886 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3887 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3888 sbi->cur_victim_sec = NULL_SECNO;
3889 sbi->gc_mode = GC_NORMAL;
3890 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3891 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3892 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3893 sbi->migration_granularity = SEGS_PER_SEC(sbi);
3894 sbi->migration_window_granularity = f2fs_sb_has_blkzoned(sbi) ?
3895 DEF_MIGRATION_WINDOW_GRANULARITY_ZONED : SEGS_PER_SEC(sbi);
3896 sbi->seq_file_ra_mul = MIN_RA_MUL;
3897 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3898 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3899 spin_lock_init(&sbi->gc_remaining_trials_lock);
3900 atomic64_set(&sbi->current_atomic_write, 0);
3901
3902 sbi->dir_level = DEF_DIR_LEVEL;
3903 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3904 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3905 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3906 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3907 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3908 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3909 DEF_UMOUNT_DISCARD_TIMEOUT;
3910 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3911
3912 for (i = 0; i < NR_COUNT_TYPE; i++)
3913 atomic_set(&sbi->nr_pages[i], 0);
3914
3915 for (i = 0; i < META; i++)
3916 atomic_set(&sbi->wb_sync_req[i], 0);
3917
3918 INIT_LIST_HEAD(&sbi->s_list);
3919 mutex_init(&sbi->umount_mutex);
3920 init_f2fs_rwsem(&sbi->io_order_lock);
3921 spin_lock_init(&sbi->cp_lock);
3922
3923 sbi->dirty_device = 0;
3924 spin_lock_init(&sbi->dev_lock);
3925
3926 init_f2fs_rwsem(&sbi->sb_lock);
3927 init_f2fs_rwsem(&sbi->pin_sem);
3928 }
3929
init_percpu_info(struct f2fs_sb_info * sbi)3930 static int init_percpu_info(struct f2fs_sb_info *sbi)
3931 {
3932 int err;
3933
3934 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3935 if (err)
3936 return err;
3937
3938 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3939 if (err)
3940 goto err_valid_block;
3941
3942 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3943 GFP_KERNEL);
3944 if (err)
3945 goto err_node_block;
3946 return 0;
3947
3948 err_node_block:
3949 percpu_counter_destroy(&sbi->rf_node_block_count);
3950 err_valid_block:
3951 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3952 return err;
3953 }
3954
3955 #ifdef CONFIG_BLK_DEV_ZONED
3956
3957 struct f2fs_report_zones_args {
3958 struct f2fs_sb_info *sbi;
3959 struct f2fs_dev_info *dev;
3960 };
3961
f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3962 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3963 void *data)
3964 {
3965 struct f2fs_report_zones_args *rz_args = data;
3966 block_t unusable_blocks = (zone->len - zone->capacity) >>
3967 F2FS_LOG_SECTORS_PER_BLOCK;
3968
3969 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3970 return 0;
3971
3972 set_bit(idx, rz_args->dev->blkz_seq);
3973 if (!rz_args->sbi->unusable_blocks_per_sec) {
3974 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3975 return 0;
3976 }
3977 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3978 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3979 return -EINVAL;
3980 }
3981 return 0;
3982 }
3983
init_blkz_info(struct f2fs_sb_info * sbi,int devi)3984 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3985 {
3986 struct block_device *bdev = FDEV(devi).bdev;
3987 sector_t nr_sectors = bdev_nr_sectors(bdev);
3988 struct f2fs_report_zones_args rep_zone_arg;
3989 u64 zone_sectors;
3990 unsigned int max_open_zones;
3991 int ret;
3992
3993 if (!f2fs_sb_has_blkzoned(sbi))
3994 return 0;
3995
3996 if (bdev_is_zoned(FDEV(devi).bdev)) {
3997 max_open_zones = bdev_max_open_zones(bdev);
3998 if (max_open_zones && (max_open_zones < sbi->max_open_zones))
3999 sbi->max_open_zones = max_open_zones;
4000 if (sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
4001 f2fs_err(sbi,
4002 "zoned: max open zones %u is too small, need at least %u open zones",
4003 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
4004 return -EINVAL;
4005 }
4006 }
4007
4008 zone_sectors = bdev_zone_sectors(bdev);
4009 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
4010 SECTOR_TO_BLOCK(zone_sectors))
4011 return -EINVAL;
4012 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
4013 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
4014 sbi->blocks_per_blkz);
4015 if (nr_sectors & (zone_sectors - 1))
4016 FDEV(devi).nr_blkz++;
4017
4018 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
4019 BITS_TO_LONGS(FDEV(devi).nr_blkz)
4020 * sizeof(unsigned long),
4021 GFP_KERNEL);
4022 if (!FDEV(devi).blkz_seq)
4023 return -ENOMEM;
4024
4025 rep_zone_arg.sbi = sbi;
4026 rep_zone_arg.dev = &FDEV(devi);
4027
4028 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
4029 &rep_zone_arg);
4030 if (ret < 0)
4031 return ret;
4032 return 0;
4033 }
4034 #endif
4035
4036 /*
4037 * Read f2fs raw super block.
4038 * Because we have two copies of super block, so read both of them
4039 * to get the first valid one. If any one of them is broken, we pass
4040 * them recovery flag back to the caller.
4041 */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)4042 static int read_raw_super_block(struct f2fs_sb_info *sbi,
4043 struct f2fs_super_block **raw_super,
4044 int *valid_super_block, int *recovery)
4045 {
4046 struct super_block *sb = sbi->sb;
4047 int block;
4048 struct folio *folio;
4049 struct f2fs_super_block *super;
4050 int err = 0;
4051
4052 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
4053 if (!super)
4054 return -ENOMEM;
4055
4056 for (block = 0; block < 2; block++) {
4057 folio = read_mapping_folio(sb->s_bdev->bd_mapping, block, NULL);
4058 if (IS_ERR(folio)) {
4059 f2fs_err(sbi, "Unable to read %dth superblock",
4060 block + 1);
4061 err = PTR_ERR(folio);
4062 *recovery = 1;
4063 continue;
4064 }
4065
4066 /* sanity checking of raw super */
4067 err = sanity_check_raw_super(sbi, folio, block);
4068 if (err) {
4069 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
4070 block + 1);
4071 folio_put(folio);
4072 *recovery = 1;
4073 continue;
4074 }
4075
4076 if (!*raw_super) {
4077 memcpy(super, F2FS_SUPER_BLOCK(folio, block),
4078 sizeof(*super));
4079 *valid_super_block = block;
4080 *raw_super = super;
4081 }
4082 folio_put(folio);
4083 }
4084
4085 /* No valid superblock */
4086 if (!*raw_super)
4087 kfree(super);
4088 else
4089 err = 0;
4090
4091 return err;
4092 }
4093
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)4094 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
4095 {
4096 struct folio *folio;
4097 pgoff_t index;
4098 __u32 crc = 0;
4099 int err;
4100
4101 if ((recover && f2fs_readonly(sbi->sb)) ||
4102 f2fs_hw_is_readonly(sbi)) {
4103 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
4104 return -EROFS;
4105 }
4106
4107 /* we should update superblock crc here */
4108 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
4109 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
4110 offsetof(struct f2fs_super_block, crc));
4111 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
4112 }
4113
4114 /* write back-up superblock first */
4115 index = sbi->valid_super_block ? 0 : 1;
4116 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL);
4117 if (IS_ERR(folio))
4118 return PTR_ERR(folio);
4119 err = __f2fs_commit_super(sbi, folio, index, true);
4120 folio_put(folio);
4121
4122 /* if we are in recovery path, skip writing valid superblock */
4123 if (recover || err)
4124 return err;
4125
4126 /* write current valid superblock */
4127 index = sbi->valid_super_block;
4128 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL);
4129 if (IS_ERR(folio))
4130 return PTR_ERR(folio);
4131 err = __f2fs_commit_super(sbi, folio, index, true);
4132 folio_put(folio);
4133 return err;
4134 }
4135
save_stop_reason(struct f2fs_sb_info * sbi,unsigned char reason)4136 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4137 {
4138 unsigned long flags;
4139
4140 spin_lock_irqsave(&sbi->error_lock, flags);
4141 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4142 sbi->stop_reason[reason]++;
4143 spin_unlock_irqrestore(&sbi->error_lock, flags);
4144 }
4145
f2fs_record_stop_reason(struct f2fs_sb_info * sbi)4146 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4147 {
4148 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4149 unsigned long flags;
4150 int err;
4151
4152 f2fs_down_write(&sbi->sb_lock);
4153
4154 spin_lock_irqsave(&sbi->error_lock, flags);
4155 if (sbi->error_dirty) {
4156 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4157 MAX_F2FS_ERRORS);
4158 sbi->error_dirty = false;
4159 }
4160 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4161 spin_unlock_irqrestore(&sbi->error_lock, flags);
4162
4163 err = f2fs_commit_super(sbi, false);
4164
4165 f2fs_up_write(&sbi->sb_lock);
4166 if (err)
4167 f2fs_err_ratelimited(sbi,
4168 "f2fs_commit_super fails to record stop_reason, err:%d",
4169 err);
4170 }
4171
f2fs_save_errors(struct f2fs_sb_info * sbi,unsigned char flag)4172 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4173 {
4174 unsigned long flags;
4175
4176 spin_lock_irqsave(&sbi->error_lock, flags);
4177 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4178 set_bit(flag, (unsigned long *)sbi->errors);
4179 sbi->error_dirty = true;
4180 }
4181 spin_unlock_irqrestore(&sbi->error_lock, flags);
4182 }
4183
f2fs_update_errors(struct f2fs_sb_info * sbi)4184 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4185 {
4186 unsigned long flags;
4187 bool need_update = false;
4188
4189 spin_lock_irqsave(&sbi->error_lock, flags);
4190 if (sbi->error_dirty) {
4191 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4192 MAX_F2FS_ERRORS);
4193 sbi->error_dirty = false;
4194 need_update = true;
4195 }
4196 spin_unlock_irqrestore(&sbi->error_lock, flags);
4197
4198 return need_update;
4199 }
4200
f2fs_record_errors(struct f2fs_sb_info * sbi,unsigned char error)4201 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4202 {
4203 int err;
4204
4205 f2fs_down_write(&sbi->sb_lock);
4206
4207 if (!f2fs_update_errors(sbi))
4208 goto out_unlock;
4209
4210 err = f2fs_commit_super(sbi, false);
4211 if (err)
4212 f2fs_err_ratelimited(sbi,
4213 "f2fs_commit_super fails to record errors:%u, err:%d",
4214 error, err);
4215 out_unlock:
4216 f2fs_up_write(&sbi->sb_lock);
4217 }
4218
f2fs_handle_error(struct f2fs_sb_info * sbi,unsigned char error)4219 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4220 {
4221 f2fs_save_errors(sbi, error);
4222 f2fs_record_errors(sbi, error);
4223 }
4224
f2fs_handle_error_async(struct f2fs_sb_info * sbi,unsigned char error)4225 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4226 {
4227 f2fs_save_errors(sbi, error);
4228
4229 if (!sbi->error_dirty)
4230 return;
4231 if (!test_bit(error, (unsigned long *)sbi->errors))
4232 return;
4233 schedule_work(&sbi->s_error_work);
4234 }
4235
system_going_down(void)4236 static bool system_going_down(void)
4237 {
4238 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4239 || system_state == SYSTEM_RESTART;
4240 }
4241
f2fs_handle_critical_error(struct f2fs_sb_info * sbi,unsigned char reason)4242 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason)
4243 {
4244 struct super_block *sb = sbi->sb;
4245 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4246 bool continue_fs = !shutdown &&
4247 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4248
4249 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4250
4251 if (!f2fs_hw_is_readonly(sbi)) {
4252 save_stop_reason(sbi, reason);
4253
4254 /*
4255 * always create an asynchronous task to record stop_reason
4256 * in order to avoid potential deadlock when running into
4257 * f2fs_record_stop_reason() synchronously.
4258 */
4259 schedule_work(&sbi->s_error_work);
4260 }
4261
4262 /*
4263 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4264 * could panic during 'reboot -f' as the underlying device got already
4265 * disabled.
4266 */
4267 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4268 !shutdown && !system_going_down() &&
4269 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4270 panic("F2FS-fs (device %s): panic forced after error\n",
4271 sb->s_id);
4272
4273 if (shutdown)
4274 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4275 else
4276 dump_stack();
4277
4278 /*
4279 * Continue filesystem operators if errors=continue. Should not set
4280 * RO by shutdown, since RO bypasses thaw_super which can hang the
4281 * system.
4282 */
4283 if (continue_fs || f2fs_readonly(sb) || shutdown) {
4284 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason);
4285 return;
4286 }
4287
4288 f2fs_warn(sbi, "Remounting filesystem read-only");
4289
4290 /*
4291 * We have already set CP_ERROR_FLAG flag to stop all updates
4292 * to filesystem, so it doesn't need to set SB_RDONLY flag here
4293 * because the flag should be set covered w/ sb->s_umount semaphore
4294 * via remount procedure, otherwise, it will confuse code like
4295 * freeze_super() which will lead to deadlocks and other problems.
4296 */
4297 }
4298
f2fs_record_error_work(struct work_struct * work)4299 static void f2fs_record_error_work(struct work_struct *work)
4300 {
4301 struct f2fs_sb_info *sbi = container_of(work,
4302 struct f2fs_sb_info, s_error_work);
4303
4304 f2fs_record_stop_reason(sbi);
4305 }
4306
get_first_zoned_segno(struct f2fs_sb_info * sbi)4307 static inline unsigned int get_first_zoned_segno(struct f2fs_sb_info *sbi)
4308 {
4309 int devi;
4310
4311 for (devi = 0; devi < sbi->s_ndevs; devi++)
4312 if (bdev_is_zoned(FDEV(devi).bdev))
4313 return GET_SEGNO(sbi, FDEV(devi).start_blk);
4314 return 0;
4315 }
4316
f2fs_scan_devices(struct f2fs_sb_info * sbi)4317 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4318 {
4319 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4320 unsigned int max_devices = MAX_DEVICES;
4321 unsigned int logical_blksize;
4322 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4323 int i;
4324
4325 /* Initialize single device information */
4326 if (!RDEV(0).path[0]) {
4327 if (!bdev_is_zoned(sbi->sb->s_bdev))
4328 return 0;
4329 max_devices = 1;
4330 }
4331
4332 /*
4333 * Initialize multiple devices information, or single
4334 * zoned block device information.
4335 */
4336 sbi->devs = f2fs_kzalloc(sbi,
4337 array_size(max_devices,
4338 sizeof(struct f2fs_dev_info)),
4339 GFP_KERNEL);
4340 if (!sbi->devs)
4341 return -ENOMEM;
4342
4343 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4344 sbi->aligned_blksize = true;
4345 #ifdef CONFIG_BLK_DEV_ZONED
4346 sbi->max_open_zones = UINT_MAX;
4347 sbi->blkzone_alloc_policy = BLKZONE_ALLOC_PRIOR_SEQ;
4348 #endif
4349
4350 for (i = 0; i < max_devices; i++) {
4351 if (i == 0)
4352 FDEV(0).bdev_file = sbi->sb->s_bdev_file;
4353 else if (!RDEV(i).path[0])
4354 break;
4355
4356 if (max_devices > 1) {
4357 /* Multi-device mount */
4358 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4359 FDEV(i).total_segments =
4360 le32_to_cpu(RDEV(i).total_segments);
4361 if (i == 0) {
4362 FDEV(i).start_blk = 0;
4363 FDEV(i).end_blk = FDEV(i).start_blk +
4364 SEGS_TO_BLKS(sbi,
4365 FDEV(i).total_segments) - 1 +
4366 le32_to_cpu(raw_super->segment0_blkaddr);
4367 } else {
4368 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4369 FDEV(i).end_blk = FDEV(i).start_blk +
4370 SEGS_TO_BLKS(sbi,
4371 FDEV(i).total_segments) - 1;
4372 FDEV(i).bdev_file = bdev_file_open_by_path(
4373 FDEV(i).path, mode, sbi->sb, NULL);
4374 }
4375 }
4376 if (IS_ERR(FDEV(i).bdev_file))
4377 return PTR_ERR(FDEV(i).bdev_file);
4378
4379 FDEV(i).bdev = file_bdev(FDEV(i).bdev_file);
4380 /* to release errored devices */
4381 sbi->s_ndevs = i + 1;
4382
4383 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4384 sbi->aligned_blksize = false;
4385
4386 #ifdef CONFIG_BLK_DEV_ZONED
4387 if (bdev_is_zoned(FDEV(i).bdev)) {
4388 if (!f2fs_sb_has_blkzoned(sbi)) {
4389 f2fs_err(sbi, "Zoned block device feature not enabled");
4390 return -EINVAL;
4391 }
4392 if (init_blkz_info(sbi, i)) {
4393 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4394 return -EINVAL;
4395 }
4396 if (max_devices == 1)
4397 break;
4398 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)",
4399 i, FDEV(i).path,
4400 FDEV(i).total_segments,
4401 FDEV(i).start_blk, FDEV(i).end_blk);
4402 continue;
4403 }
4404 #endif
4405 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4406 i, FDEV(i).path,
4407 FDEV(i).total_segments,
4408 FDEV(i).start_blk, FDEV(i).end_blk);
4409 }
4410 return 0;
4411 }
4412
f2fs_setup_casefold(struct f2fs_sb_info * sbi)4413 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4414 {
4415 #if IS_ENABLED(CONFIG_UNICODE)
4416 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4417 const struct f2fs_sb_encodings *encoding_info;
4418 struct unicode_map *encoding;
4419 __u16 encoding_flags;
4420
4421 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4422 if (!encoding_info) {
4423 f2fs_err(sbi,
4424 "Encoding requested by superblock is unknown");
4425 return -EINVAL;
4426 }
4427
4428 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4429 encoding = utf8_load(encoding_info->version);
4430 if (IS_ERR(encoding)) {
4431 f2fs_err(sbi,
4432 "can't mount with superblock charset: %s-%u.%u.%u "
4433 "not supported by the kernel. flags: 0x%x.",
4434 encoding_info->name,
4435 unicode_major(encoding_info->version),
4436 unicode_minor(encoding_info->version),
4437 unicode_rev(encoding_info->version),
4438 encoding_flags);
4439 return PTR_ERR(encoding);
4440 }
4441 f2fs_info(sbi, "Using encoding defined by superblock: "
4442 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4443 unicode_major(encoding_info->version),
4444 unicode_minor(encoding_info->version),
4445 unicode_rev(encoding_info->version),
4446 encoding_flags);
4447
4448 sbi->sb->s_encoding = encoding;
4449 sbi->sb->s_encoding_flags = encoding_flags;
4450 }
4451 #else
4452 if (f2fs_sb_has_casefold(sbi)) {
4453 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4454 return -EINVAL;
4455 }
4456 #endif
4457 return 0;
4458 }
4459
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)4460 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4461 {
4462 /* adjust parameters according to the volume size */
4463 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4464 if (f2fs_block_unit_discard(sbi))
4465 SM_I(sbi)->dcc_info->discard_granularity =
4466 MIN_DISCARD_GRANULARITY;
4467 if (!f2fs_lfs_mode(sbi))
4468 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4469 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4470 }
4471
4472 sbi->readdir_ra = true;
4473 }
4474
f2fs_fill_super(struct super_block * sb,void * data,int silent)4475 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4476 {
4477 struct f2fs_sb_info *sbi;
4478 struct f2fs_super_block *raw_super;
4479 struct inode *root;
4480 int err;
4481 bool skip_recovery = false, need_fsck = false;
4482 char *options = NULL;
4483 int recovery, i, valid_super_block;
4484 struct curseg_info *seg_i;
4485 int retry_cnt = 1;
4486 #ifdef CONFIG_QUOTA
4487 bool quota_enabled = false;
4488 #endif
4489
4490 try_onemore:
4491 err = -EINVAL;
4492 raw_super = NULL;
4493 valid_super_block = -1;
4494 recovery = 0;
4495
4496 /* allocate memory for f2fs-specific super block info */
4497 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4498 if (!sbi)
4499 return -ENOMEM;
4500
4501 sbi->sb = sb;
4502
4503 /* initialize locks within allocated memory */
4504 init_f2fs_rwsem(&sbi->gc_lock);
4505 mutex_init(&sbi->writepages);
4506 init_f2fs_rwsem(&sbi->cp_global_sem);
4507 init_f2fs_rwsem(&sbi->node_write);
4508 init_f2fs_rwsem(&sbi->node_change);
4509 spin_lock_init(&sbi->stat_lock);
4510 init_f2fs_rwsem(&sbi->cp_rwsem);
4511 init_f2fs_rwsem(&sbi->quota_sem);
4512 init_waitqueue_head(&sbi->cp_wait);
4513 spin_lock_init(&sbi->error_lock);
4514
4515 for (i = 0; i < NR_INODE_TYPE; i++) {
4516 INIT_LIST_HEAD(&sbi->inode_list[i]);
4517 spin_lock_init(&sbi->inode_lock[i]);
4518 }
4519 mutex_init(&sbi->flush_lock);
4520
4521 /* set a block size */
4522 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4523 f2fs_err(sbi, "unable to set blocksize");
4524 goto free_sbi;
4525 }
4526
4527 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4528 &recovery);
4529 if (err)
4530 goto free_sbi;
4531
4532 sb->s_fs_info = sbi;
4533 sbi->raw_super = raw_super;
4534
4535 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4536 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4537 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4538
4539 /* precompute checksum seed for metadata */
4540 if (f2fs_sb_has_inode_chksum(sbi))
4541 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4542 sizeof(raw_super->uuid));
4543
4544 default_options(sbi, false);
4545 /* parse mount options */
4546 options = kstrdup((const char *)data, GFP_KERNEL);
4547 if (data && !options) {
4548 err = -ENOMEM;
4549 goto free_sb_buf;
4550 }
4551
4552 err = parse_options(sbi, options, false);
4553 if (err)
4554 goto free_options;
4555
4556 err = f2fs_default_check(sbi);
4557 if (err)
4558 goto free_options;
4559
4560 sb->s_maxbytes = max_file_blocks(NULL) <<
4561 le32_to_cpu(raw_super->log_blocksize);
4562 sb->s_max_links = F2FS_LINK_MAX;
4563
4564 err = f2fs_setup_casefold(sbi);
4565 if (err)
4566 goto free_options;
4567
4568 #ifdef CONFIG_QUOTA
4569 sb->dq_op = &f2fs_quota_operations;
4570 sb->s_qcop = &f2fs_quotactl_ops;
4571 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4572
4573 if (f2fs_sb_has_quota_ino(sbi)) {
4574 for (i = 0; i < MAXQUOTAS; i++) {
4575 if (f2fs_qf_ino(sbi->sb, i))
4576 sbi->nquota_files++;
4577 }
4578 }
4579 #endif
4580
4581 sb->s_op = &f2fs_sops;
4582 #ifdef CONFIG_FS_ENCRYPTION
4583 sb->s_cop = &f2fs_cryptops;
4584 #endif
4585 #ifdef CONFIG_FS_VERITY
4586 sb->s_vop = &f2fs_verityops;
4587 #endif
4588 sb->s_xattr = f2fs_xattr_handlers;
4589 sb->s_export_op = &f2fs_export_ops;
4590 sb->s_magic = F2FS_SUPER_MAGIC;
4591 sb->s_time_gran = 1;
4592 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4593 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4594 if (test_opt(sbi, INLINECRYPT))
4595 sb->s_flags |= SB_INLINECRYPT;
4596
4597 if (test_opt(sbi, LAZYTIME))
4598 sb->s_flags |= SB_LAZYTIME;
4599 else
4600 sb->s_flags &= ~SB_LAZYTIME;
4601
4602 super_set_uuid(sb, (void *) raw_super->uuid, sizeof(raw_super->uuid));
4603 super_set_sysfs_name_bdev(sb);
4604 sb->s_iflags |= SB_I_CGROUPWB;
4605
4606 /* init f2fs-specific super block info */
4607 sbi->valid_super_block = valid_super_block;
4608
4609 /* disallow all the data/node/meta page writes */
4610 set_sbi_flag(sbi, SBI_POR_DOING);
4611
4612 err = f2fs_init_write_merge_io(sbi);
4613 if (err)
4614 goto free_bio_info;
4615
4616 init_sb_info(sbi);
4617
4618 err = f2fs_init_iostat(sbi);
4619 if (err)
4620 goto free_bio_info;
4621
4622 err = init_percpu_info(sbi);
4623 if (err)
4624 goto free_iostat;
4625
4626 /* init per sbi slab cache */
4627 err = f2fs_init_xattr_caches(sbi);
4628 if (err)
4629 goto free_percpu;
4630 err = f2fs_init_page_array_cache(sbi);
4631 if (err)
4632 goto free_xattr_cache;
4633
4634 /* get an inode for meta space */
4635 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4636 if (IS_ERR(sbi->meta_inode)) {
4637 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4638 err = PTR_ERR(sbi->meta_inode);
4639 goto free_page_array_cache;
4640 }
4641
4642 err = f2fs_get_valid_checkpoint(sbi);
4643 if (err) {
4644 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4645 goto free_meta_inode;
4646 }
4647
4648 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4649 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4650 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4651 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4652 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4653 }
4654
4655 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4656 set_sbi_flag(sbi, SBI_NEED_FSCK);
4657
4658 /* Initialize device list */
4659 err = f2fs_scan_devices(sbi);
4660 if (err) {
4661 f2fs_err(sbi, "Failed to find devices");
4662 goto free_devices;
4663 }
4664
4665 err = f2fs_init_post_read_wq(sbi);
4666 if (err) {
4667 f2fs_err(sbi, "Failed to initialize post read workqueue");
4668 goto free_devices;
4669 }
4670
4671 sbi->total_valid_node_count =
4672 le32_to_cpu(sbi->ckpt->valid_node_count);
4673 percpu_counter_set(&sbi->total_valid_inode_count,
4674 le32_to_cpu(sbi->ckpt->valid_inode_count));
4675 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4676 sbi->total_valid_block_count =
4677 le64_to_cpu(sbi->ckpt->valid_block_count);
4678 sbi->last_valid_block_count = sbi->total_valid_block_count;
4679 sbi->reserved_blocks = 0;
4680 sbi->current_reserved_blocks = 0;
4681 limit_reserve_root(sbi);
4682 adjust_unusable_cap_perc(sbi);
4683
4684 f2fs_init_extent_cache_info(sbi);
4685
4686 f2fs_init_ino_entry_info(sbi);
4687
4688 f2fs_init_fsync_node_info(sbi);
4689
4690 /* setup checkpoint request control and start checkpoint issue thread */
4691 f2fs_init_ckpt_req_control(sbi);
4692 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4693 test_opt(sbi, MERGE_CHECKPOINT)) {
4694 err = f2fs_start_ckpt_thread(sbi);
4695 if (err) {
4696 f2fs_err(sbi,
4697 "Failed to start F2FS issue_checkpoint_thread (%d)",
4698 err);
4699 goto stop_ckpt_thread;
4700 }
4701 }
4702
4703 /* setup f2fs internal modules */
4704 err = f2fs_build_segment_manager(sbi);
4705 if (err) {
4706 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4707 err);
4708 goto free_sm;
4709 }
4710 err = f2fs_build_node_manager(sbi);
4711 if (err) {
4712 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4713 err);
4714 goto free_nm;
4715 }
4716
4717 /* For write statistics */
4718 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4719
4720 /* get segno of first zoned block device */
4721 sbi->first_zoned_segno = get_first_zoned_segno(sbi);
4722
4723 /* Read accumulated write IO statistics if exists */
4724 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4725 if (__exist_node_summaries(sbi))
4726 sbi->kbytes_written =
4727 le64_to_cpu(seg_i->journal->info.kbytes_written);
4728
4729 f2fs_build_gc_manager(sbi);
4730
4731 err = f2fs_build_stats(sbi);
4732 if (err)
4733 goto free_nm;
4734
4735 /* get an inode for node space */
4736 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4737 if (IS_ERR(sbi->node_inode)) {
4738 f2fs_err(sbi, "Failed to read node inode");
4739 err = PTR_ERR(sbi->node_inode);
4740 goto free_stats;
4741 }
4742
4743 /* read root inode and dentry */
4744 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4745 if (IS_ERR(root)) {
4746 f2fs_err(sbi, "Failed to read root inode");
4747 err = PTR_ERR(root);
4748 goto free_node_inode;
4749 }
4750 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4751 !root->i_size || !root->i_nlink) {
4752 iput(root);
4753 err = -EINVAL;
4754 goto free_node_inode;
4755 }
4756
4757 generic_set_sb_d_ops(sb);
4758 sb->s_root = d_make_root(root); /* allocate root dentry */
4759 if (!sb->s_root) {
4760 err = -ENOMEM;
4761 goto free_node_inode;
4762 }
4763
4764 err = f2fs_init_compress_inode(sbi);
4765 if (err)
4766 goto free_root_inode;
4767
4768 err = f2fs_register_sysfs(sbi);
4769 if (err)
4770 goto free_compress_inode;
4771
4772 sbi->umount_lock_holder = current;
4773 #ifdef CONFIG_QUOTA
4774 /* Enable quota usage during mount */
4775 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4776 err = f2fs_enable_quotas(sb);
4777 if (err)
4778 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4779 }
4780
4781 quota_enabled = f2fs_recover_quota_begin(sbi);
4782 #endif
4783 /* if there are any orphan inodes, free them */
4784 err = f2fs_recover_orphan_inodes(sbi);
4785 if (err)
4786 goto free_meta;
4787
4788 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) {
4789 skip_recovery = true;
4790 goto reset_checkpoint;
4791 }
4792
4793 /* recover fsynced data */
4794 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4795 !test_opt(sbi, NORECOVERY)) {
4796 /*
4797 * mount should be failed, when device has readonly mode, and
4798 * previous checkpoint was not done by clean system shutdown.
4799 */
4800 if (f2fs_hw_is_readonly(sbi)) {
4801 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4802 err = f2fs_recover_fsync_data(sbi, true);
4803 if (err > 0) {
4804 err = -EROFS;
4805 f2fs_err(sbi, "Need to recover fsync data, but "
4806 "write access unavailable, please try "
4807 "mount w/ disable_roll_forward or norecovery");
4808 }
4809 if (err < 0)
4810 goto free_meta;
4811 }
4812 f2fs_info(sbi, "write access unavailable, skipping recovery");
4813 goto reset_checkpoint;
4814 }
4815
4816 if (need_fsck)
4817 set_sbi_flag(sbi, SBI_NEED_FSCK);
4818
4819 if (skip_recovery)
4820 goto reset_checkpoint;
4821
4822 err = f2fs_recover_fsync_data(sbi, false);
4823 if (err < 0) {
4824 if (err != -ENOMEM)
4825 skip_recovery = true;
4826 need_fsck = true;
4827 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4828 err);
4829 goto free_meta;
4830 }
4831 } else {
4832 err = f2fs_recover_fsync_data(sbi, true);
4833
4834 if (!f2fs_readonly(sb) && err > 0) {
4835 err = -EINVAL;
4836 f2fs_err(sbi, "Need to recover fsync data");
4837 goto free_meta;
4838 }
4839 }
4840
4841 reset_checkpoint:
4842 #ifdef CONFIG_QUOTA
4843 f2fs_recover_quota_end(sbi, quota_enabled);
4844 #endif
4845 /*
4846 * If the f2fs is not readonly and fsync data recovery succeeds,
4847 * write pointer consistency of cursegs and other zones are already
4848 * checked and fixed during recovery. However, if recovery fails,
4849 * write pointers are left untouched, and retry-mount should check
4850 * them here.
4851 */
4852 if (skip_recovery)
4853 err = f2fs_check_and_fix_write_pointer(sbi);
4854 if (err)
4855 goto free_meta;
4856
4857 /* f2fs_recover_fsync_data() cleared this already */
4858 clear_sbi_flag(sbi, SBI_POR_DOING);
4859
4860 err = f2fs_init_inmem_curseg(sbi);
4861 if (err)
4862 goto sync_free_meta;
4863
4864 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4865 err = f2fs_disable_checkpoint(sbi);
4866 if (err)
4867 goto sync_free_meta;
4868 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4869 f2fs_enable_checkpoint(sbi);
4870 }
4871
4872 /*
4873 * If filesystem is not mounted as read-only then
4874 * do start the gc_thread.
4875 */
4876 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4877 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4878 /* After POR, we can run background GC thread.*/
4879 err = f2fs_start_gc_thread(sbi);
4880 if (err)
4881 goto sync_free_meta;
4882 }
4883 kvfree(options);
4884
4885 /* recover broken superblock */
4886 if (recovery) {
4887 err = f2fs_commit_super(sbi, true);
4888 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4889 sbi->valid_super_block ? 1 : 2, err);
4890 }
4891
4892 f2fs_join_shrinker(sbi);
4893
4894 f2fs_tuning_parameters(sbi);
4895
4896 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4897 cur_cp_version(F2FS_CKPT(sbi)));
4898 f2fs_update_time(sbi, CP_TIME);
4899 f2fs_update_time(sbi, REQ_TIME);
4900 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4901
4902 sbi->umount_lock_holder = NULL;
4903 return 0;
4904
4905 sync_free_meta:
4906 /* safe to flush all the data */
4907 sync_filesystem(sbi->sb);
4908 retry_cnt = 0;
4909
4910 free_meta:
4911 #ifdef CONFIG_QUOTA
4912 f2fs_truncate_quota_inode_pages(sb);
4913 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4914 f2fs_quota_off_umount(sbi->sb);
4915 #endif
4916 /*
4917 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4918 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4919 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4920 * falls into an infinite loop in f2fs_sync_meta_pages().
4921 */
4922 truncate_inode_pages_final(META_MAPPING(sbi));
4923 /* evict some inodes being cached by GC */
4924 evict_inodes(sb);
4925 f2fs_unregister_sysfs(sbi);
4926 free_compress_inode:
4927 f2fs_destroy_compress_inode(sbi);
4928 free_root_inode:
4929 dput(sb->s_root);
4930 sb->s_root = NULL;
4931 free_node_inode:
4932 f2fs_release_ino_entry(sbi, true);
4933 truncate_inode_pages_final(NODE_MAPPING(sbi));
4934 iput(sbi->node_inode);
4935 sbi->node_inode = NULL;
4936 free_stats:
4937 f2fs_destroy_stats(sbi);
4938 free_nm:
4939 /* stop discard thread before destroying node manager */
4940 f2fs_stop_discard_thread(sbi);
4941 f2fs_destroy_node_manager(sbi);
4942 free_sm:
4943 f2fs_destroy_segment_manager(sbi);
4944 stop_ckpt_thread:
4945 f2fs_stop_ckpt_thread(sbi);
4946 /* flush s_error_work before sbi destroy */
4947 flush_work(&sbi->s_error_work);
4948 f2fs_destroy_post_read_wq(sbi);
4949 free_devices:
4950 destroy_device_list(sbi);
4951 kvfree(sbi->ckpt);
4952 free_meta_inode:
4953 make_bad_inode(sbi->meta_inode);
4954 iput(sbi->meta_inode);
4955 sbi->meta_inode = NULL;
4956 free_page_array_cache:
4957 f2fs_destroy_page_array_cache(sbi);
4958 free_xattr_cache:
4959 f2fs_destroy_xattr_caches(sbi);
4960 free_percpu:
4961 destroy_percpu_info(sbi);
4962 free_iostat:
4963 f2fs_destroy_iostat(sbi);
4964 free_bio_info:
4965 for (i = 0; i < NR_PAGE_TYPE; i++)
4966 kvfree(sbi->write_io[i]);
4967
4968 #if IS_ENABLED(CONFIG_UNICODE)
4969 utf8_unload(sb->s_encoding);
4970 sb->s_encoding = NULL;
4971 #endif
4972 free_options:
4973 #ifdef CONFIG_QUOTA
4974 for (i = 0; i < MAXQUOTAS; i++)
4975 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4976 #endif
4977 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4978 kvfree(options);
4979 free_sb_buf:
4980 kfree(raw_super);
4981 free_sbi:
4982 kfree(sbi);
4983 sb->s_fs_info = NULL;
4984
4985 /* give only one another chance */
4986 if (retry_cnt > 0 && skip_recovery) {
4987 retry_cnt--;
4988 shrink_dcache_sb(sb);
4989 goto try_onemore;
4990 }
4991 return err;
4992 }
4993
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4994 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4995 const char *dev_name, void *data)
4996 {
4997 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4998 }
4999
kill_f2fs_super(struct super_block * sb)5000 static void kill_f2fs_super(struct super_block *sb)
5001 {
5002 struct f2fs_sb_info *sbi = F2FS_SB(sb);
5003
5004 if (sb->s_root) {
5005 sbi->umount_lock_holder = current;
5006
5007 set_sbi_flag(sbi, SBI_IS_CLOSE);
5008 f2fs_stop_gc_thread(sbi);
5009 f2fs_stop_discard_thread(sbi);
5010
5011 #ifdef CONFIG_F2FS_FS_COMPRESSION
5012 /*
5013 * latter evict_inode() can bypass checking and invalidating
5014 * compress inode cache.
5015 */
5016 if (test_opt(sbi, COMPRESS_CACHE))
5017 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
5018 #endif
5019
5020 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
5021 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
5022 struct cp_control cpc = {
5023 .reason = CP_UMOUNT,
5024 };
5025 stat_inc_cp_call_count(sbi, TOTAL_CALL);
5026 f2fs_write_checkpoint(sbi, &cpc);
5027 }
5028
5029 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
5030 sb->s_flags &= ~SB_RDONLY;
5031 }
5032 kill_block_super(sb);
5033 /* Release block devices last, after fscrypt_destroy_keyring(). */
5034 if (sbi) {
5035 destroy_device_list(sbi);
5036 kfree(sbi);
5037 sb->s_fs_info = NULL;
5038 }
5039 }
5040
5041 static struct file_system_type f2fs_fs_type = {
5042 .owner = THIS_MODULE,
5043 .name = "f2fs",
5044 .mount = f2fs_mount,
5045 .kill_sb = kill_f2fs_super,
5046 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
5047 };
5048 MODULE_ALIAS_FS("f2fs");
5049
init_inodecache(void)5050 static int __init init_inodecache(void)
5051 {
5052 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
5053 sizeof(struct f2fs_inode_info), 0,
5054 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
5055 return f2fs_inode_cachep ? 0 : -ENOMEM;
5056 }
5057
destroy_inodecache(void)5058 static void destroy_inodecache(void)
5059 {
5060 /*
5061 * Make sure all delayed rcu free inodes are flushed before we
5062 * destroy cache.
5063 */
5064 rcu_barrier();
5065 kmem_cache_destroy(f2fs_inode_cachep);
5066 }
5067
init_f2fs_fs(void)5068 static int __init init_f2fs_fs(void)
5069 {
5070 int err;
5071
5072 err = init_inodecache();
5073 if (err)
5074 goto fail;
5075 err = f2fs_create_node_manager_caches();
5076 if (err)
5077 goto free_inodecache;
5078 err = f2fs_create_segment_manager_caches();
5079 if (err)
5080 goto free_node_manager_caches;
5081 err = f2fs_create_checkpoint_caches();
5082 if (err)
5083 goto free_segment_manager_caches;
5084 err = f2fs_create_recovery_cache();
5085 if (err)
5086 goto free_checkpoint_caches;
5087 err = f2fs_create_extent_cache();
5088 if (err)
5089 goto free_recovery_cache;
5090 err = f2fs_create_garbage_collection_cache();
5091 if (err)
5092 goto free_extent_cache;
5093 err = f2fs_init_sysfs();
5094 if (err)
5095 goto free_garbage_collection_cache;
5096 err = f2fs_init_shrinker();
5097 if (err)
5098 goto free_sysfs;
5099 f2fs_create_root_stats();
5100 err = f2fs_init_post_read_processing();
5101 if (err)
5102 goto free_root_stats;
5103 err = f2fs_init_iostat_processing();
5104 if (err)
5105 goto free_post_read;
5106 err = f2fs_init_bio_entry_cache();
5107 if (err)
5108 goto free_iostat;
5109 err = f2fs_init_bioset();
5110 if (err)
5111 goto free_bio_entry_cache;
5112 err = f2fs_init_compress_mempool();
5113 if (err)
5114 goto free_bioset;
5115 err = f2fs_init_compress_cache();
5116 if (err)
5117 goto free_compress_mempool;
5118 err = f2fs_create_casefold_cache();
5119 if (err)
5120 goto free_compress_cache;
5121 err = register_filesystem(&f2fs_fs_type);
5122 if (err)
5123 goto free_casefold_cache;
5124 return 0;
5125 free_casefold_cache:
5126 f2fs_destroy_casefold_cache();
5127 free_compress_cache:
5128 f2fs_destroy_compress_cache();
5129 free_compress_mempool:
5130 f2fs_destroy_compress_mempool();
5131 free_bioset:
5132 f2fs_destroy_bioset();
5133 free_bio_entry_cache:
5134 f2fs_destroy_bio_entry_cache();
5135 free_iostat:
5136 f2fs_destroy_iostat_processing();
5137 free_post_read:
5138 f2fs_destroy_post_read_processing();
5139 free_root_stats:
5140 f2fs_destroy_root_stats();
5141 f2fs_exit_shrinker();
5142 free_sysfs:
5143 f2fs_exit_sysfs();
5144 free_garbage_collection_cache:
5145 f2fs_destroy_garbage_collection_cache();
5146 free_extent_cache:
5147 f2fs_destroy_extent_cache();
5148 free_recovery_cache:
5149 f2fs_destroy_recovery_cache();
5150 free_checkpoint_caches:
5151 f2fs_destroy_checkpoint_caches();
5152 free_segment_manager_caches:
5153 f2fs_destroy_segment_manager_caches();
5154 free_node_manager_caches:
5155 f2fs_destroy_node_manager_caches();
5156 free_inodecache:
5157 destroy_inodecache();
5158 fail:
5159 return err;
5160 }
5161
exit_f2fs_fs(void)5162 static void __exit exit_f2fs_fs(void)
5163 {
5164 unregister_filesystem(&f2fs_fs_type);
5165 f2fs_destroy_casefold_cache();
5166 f2fs_destroy_compress_cache();
5167 f2fs_destroy_compress_mempool();
5168 f2fs_destroy_bioset();
5169 f2fs_destroy_bio_entry_cache();
5170 f2fs_destroy_iostat_processing();
5171 f2fs_destroy_post_read_processing();
5172 f2fs_destroy_root_stats();
5173 f2fs_exit_shrinker();
5174 f2fs_exit_sysfs();
5175 f2fs_destroy_garbage_collection_cache();
5176 f2fs_destroy_extent_cache();
5177 f2fs_destroy_recovery_cache();
5178 f2fs_destroy_checkpoint_caches();
5179 f2fs_destroy_segment_manager_caches();
5180 f2fs_destroy_node_manager_caches();
5181 destroy_inodecache();
5182 }
5183
5184 module_init(init_f2fs_fs)
5185 module_exit(exit_f2fs_fs)
5186
5187 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5188 MODULE_DESCRIPTION("Flash Friendly File System");
5189 MODULE_LICENSE("GPL");
5190