1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * fs/f2fs/super.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8 #include <linux/module.h>
9 #include <linux/init.h>
10 #include <linux/fs.h>
11 #include <linux/fs_context.h>
12 #include <linux/sched/mm.h>
13 #include <linux/statfs.h>
14 #include <linux/buffer_head.h>
15 #include <linux/kthread.h>
16 #include <linux/parser.h>
17 #include <linux/mount.h>
18 #include <linux/seq_file.h>
19 #include <linux/proc_fs.h>
20 #include <linux/random.h>
21 #include <linux/exportfs.h>
22 #include <linux/blkdev.h>
23 #include <linux/quotaops.h>
24 #include <linux/f2fs_fs.h>
25 #include <linux/sysfs.h>
26 #include <linux/quota.h>
27 #include <linux/unicode.h>
28 #include <linux/part_stat.h>
29 #include <linux/zstd.h>
30 #include <linux/lz4.h>
31 #include <linux/cleancache.h>
32
33 #include "f2fs.h"
34 #include "node.h"
35 #include "segment.h"
36 #include "xattr.h"
37 #include "gc.h"
38 #include "iostat.h"
39
40 #define CREATE_TRACE_POINTS
41 #include <trace/events/f2fs.h>
42
43 static struct kmem_cache *f2fs_inode_cachep;
44
45 #ifdef CONFIG_F2FS_FAULT_INJECTION
46
47 const char *f2fs_fault_name[FAULT_MAX] = {
48 [FAULT_KMALLOC] = "kmalloc",
49 [FAULT_KVMALLOC] = "kvmalloc",
50 [FAULT_PAGE_ALLOC] = "page alloc",
51 [FAULT_PAGE_GET] = "page get",
52 [FAULT_ALLOC_NID] = "alloc nid",
53 [FAULT_ORPHAN] = "orphan",
54 [FAULT_BLOCK] = "no more block",
55 [FAULT_DIR_DEPTH] = "too big dir depth",
56 [FAULT_EVICT_INODE] = "evict_inode fail",
57 [FAULT_TRUNCATE] = "truncate fail",
58 [FAULT_READ_IO] = "read IO error",
59 [FAULT_CHECKPOINT] = "checkpoint error",
60 [FAULT_DISCARD] = "discard error",
61 [FAULT_WRITE_IO] = "write IO error",
62 [FAULT_SLAB_ALLOC] = "slab alloc",
63 [FAULT_DQUOT_INIT] = "dquot initialize",
64 [FAULT_LOCK_OP] = "lock_op",
65 [FAULT_BLKADDR_VALIDITY] = "invalid blkaddr",
66 [FAULT_BLKADDR_CONSISTENCE] = "inconsistent blkaddr",
67 [FAULT_NO_SEGMENT] = "no free segment",
68 };
69
f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned long rate,unsigned long type)70 int f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned long rate,
71 unsigned long type)
72 {
73 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
74
75 if (rate) {
76 if (rate > INT_MAX)
77 return -EINVAL;
78 atomic_set(&ffi->inject_ops, 0);
79 ffi->inject_rate = (int)rate;
80 }
81
82 if (type) {
83 if (type >= BIT(FAULT_MAX))
84 return -EINVAL;
85 ffi->inject_type = (unsigned int)type;
86 }
87
88 if (!rate && !type)
89 memset(ffi, 0, sizeof(struct f2fs_fault_info));
90 else
91 f2fs_info(sbi,
92 "build fault injection attr: rate: %lu, type: 0x%lx",
93 rate, type);
94 return 0;
95 }
96 #endif
97
98 /* f2fs-wide shrinker description */
99 static struct shrinker f2fs_shrinker_info = {
100 .scan_objects = f2fs_shrink_scan,
101 .count_objects = f2fs_shrink_count,
102 .seeks = DEFAULT_SEEKS,
103 };
104
105 enum {
106 Opt_gc_background,
107 Opt_disable_roll_forward,
108 Opt_norecovery,
109 Opt_discard,
110 Opt_nodiscard,
111 Opt_noheap,
112 Opt_heap,
113 Opt_user_xattr,
114 Opt_nouser_xattr,
115 Opt_acl,
116 Opt_noacl,
117 Opt_active_logs,
118 Opt_disable_ext_identify,
119 Opt_inline_xattr,
120 Opt_noinline_xattr,
121 Opt_inline_xattr_size,
122 Opt_inline_data,
123 Opt_inline_dentry,
124 Opt_noinline_dentry,
125 Opt_flush_merge,
126 Opt_noflush_merge,
127 Opt_barrier,
128 Opt_nobarrier,
129 Opt_fastboot,
130 Opt_extent_cache,
131 Opt_noextent_cache,
132 Opt_noinline_data,
133 Opt_data_flush,
134 Opt_reserve_root,
135 Opt_resgid,
136 Opt_resuid,
137 Opt_mode,
138 Opt_fault_injection,
139 Opt_fault_type,
140 Opt_lazytime,
141 Opt_nolazytime,
142 Opt_quota,
143 Opt_noquota,
144 Opt_usrquota,
145 Opt_grpquota,
146 Opt_prjquota,
147 Opt_usrjquota,
148 Opt_grpjquota,
149 Opt_prjjquota,
150 Opt_offusrjquota,
151 Opt_offgrpjquota,
152 Opt_offprjjquota,
153 Opt_jqfmt_vfsold,
154 Opt_jqfmt_vfsv0,
155 Opt_jqfmt_vfsv1,
156 Opt_alloc,
157 Opt_fsync,
158 Opt_test_dummy_encryption,
159 Opt_inlinecrypt,
160 Opt_checkpoint_disable,
161 Opt_checkpoint_disable_cap,
162 Opt_checkpoint_disable_cap_perc,
163 Opt_checkpoint_enable,
164 Opt_checkpoint_merge,
165 Opt_nocheckpoint_merge,
166 Opt_compress_algorithm,
167 Opt_compress_log_size,
168 Opt_compress_extension,
169 Opt_nocompress_extension,
170 Opt_compress_chksum,
171 Opt_compress_mode,
172 Opt_compress_cache,
173 Opt_atgc,
174 Opt_gc_merge,
175 Opt_nogc_merge,
176 Opt_discard_unit,
177 Opt_memory_mode,
178 Opt_age_extent_cache,
179 Opt_errors,
180 Opt_err,
181 };
182
183 static match_table_t f2fs_tokens = {
184 {Opt_gc_background, "background_gc=%s"},
185 {Opt_disable_roll_forward, "disable_roll_forward"},
186 {Opt_norecovery, "norecovery"},
187 {Opt_discard, "discard"},
188 {Opt_nodiscard, "nodiscard"},
189 {Opt_noheap, "no_heap"},
190 {Opt_heap, "heap"},
191 {Opt_user_xattr, "user_xattr"},
192 {Opt_nouser_xattr, "nouser_xattr"},
193 {Opt_acl, "acl"},
194 {Opt_noacl, "noacl"},
195 {Opt_active_logs, "active_logs=%u"},
196 {Opt_disable_ext_identify, "disable_ext_identify"},
197 {Opt_inline_xattr, "inline_xattr"},
198 {Opt_noinline_xattr, "noinline_xattr"},
199 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
200 {Opt_inline_data, "inline_data"},
201 {Opt_inline_dentry, "inline_dentry"},
202 {Opt_noinline_dentry, "noinline_dentry"},
203 {Opt_flush_merge, "flush_merge"},
204 {Opt_noflush_merge, "noflush_merge"},
205 {Opt_barrier, "barrier"},
206 {Opt_nobarrier, "nobarrier"},
207 {Opt_fastboot, "fastboot"},
208 {Opt_extent_cache, "extent_cache"},
209 {Opt_noextent_cache, "noextent_cache"},
210 {Opt_noinline_data, "noinline_data"},
211 {Opt_data_flush, "data_flush"},
212 {Opt_reserve_root, "reserve_root=%u"},
213 {Opt_resgid, "resgid=%u"},
214 {Opt_resuid, "resuid=%u"},
215 {Opt_mode, "mode=%s"},
216 {Opt_fault_injection, "fault_injection=%u"},
217 {Opt_fault_type, "fault_type=%u"},
218 {Opt_lazytime, "lazytime"},
219 {Opt_nolazytime, "nolazytime"},
220 {Opt_quota, "quota"},
221 {Opt_noquota, "noquota"},
222 {Opt_usrquota, "usrquota"},
223 {Opt_grpquota, "grpquota"},
224 {Opt_prjquota, "prjquota"},
225 {Opt_usrjquota, "usrjquota=%s"},
226 {Opt_grpjquota, "grpjquota=%s"},
227 {Opt_prjjquota, "prjjquota=%s"},
228 {Opt_offusrjquota, "usrjquota="},
229 {Opt_offgrpjquota, "grpjquota="},
230 {Opt_offprjjquota, "prjjquota="},
231 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
232 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
233 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
234 {Opt_alloc, "alloc_mode=%s"},
235 {Opt_fsync, "fsync_mode=%s"},
236 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
237 {Opt_test_dummy_encryption, "test_dummy_encryption"},
238 {Opt_inlinecrypt, "inlinecrypt"},
239 {Opt_checkpoint_disable, "checkpoint=disable"},
240 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
241 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
242 {Opt_checkpoint_enable, "checkpoint=enable"},
243 {Opt_checkpoint_merge, "checkpoint_merge"},
244 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
245 {Opt_compress_algorithm, "compress_algorithm=%s"},
246 {Opt_compress_log_size, "compress_log_size=%u"},
247 {Opt_compress_extension, "compress_extension=%s"},
248 {Opt_nocompress_extension, "nocompress_extension=%s"},
249 {Opt_compress_chksum, "compress_chksum"},
250 {Opt_compress_mode, "compress_mode=%s"},
251 {Opt_compress_cache, "compress_cache"},
252 {Opt_atgc, "atgc"},
253 {Opt_gc_merge, "gc_merge"},
254 {Opt_nogc_merge, "nogc_merge"},
255 {Opt_discard_unit, "discard_unit=%s"},
256 {Opt_memory_mode, "memory=%s"},
257 {Opt_age_extent_cache, "age_extent_cache"},
258 {Opt_errors, "errors=%s"},
259 {Opt_err, NULL},
260 };
261
f2fs_printk(struct f2fs_sb_info * sbi,bool limit_rate,const char * fmt,...)262 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate,
263 const char *fmt, ...)
264 {
265 struct va_format vaf;
266 va_list args;
267 int level;
268
269 va_start(args, fmt);
270
271 level = printk_get_level(fmt);
272 vaf.fmt = printk_skip_level(fmt);
273 vaf.va = &args;
274 if (limit_rate)
275 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n",
276 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
277 else
278 printk("%c%cF2FS-fs (%s): %pV\n",
279 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
280
281 va_end(args);
282 }
283
284 #if IS_ENABLED(CONFIG_UNICODE)
285 static const struct f2fs_sb_encodings {
286 __u16 magic;
287 char *name;
288 unsigned int version;
289 } f2fs_sb_encoding_map[] = {
290 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
291 };
292
293 static const struct f2fs_sb_encodings *
f2fs_sb_read_encoding(const struct f2fs_super_block * sb)294 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
295 {
296 __u16 magic = le16_to_cpu(sb->s_encoding);
297 int i;
298
299 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
300 if (magic == f2fs_sb_encoding_map[i].magic)
301 return &f2fs_sb_encoding_map[i];
302
303 return NULL;
304 }
305
306 struct kmem_cache *f2fs_cf_name_slab;
f2fs_create_casefold_cache(void)307 static int __init f2fs_create_casefold_cache(void)
308 {
309 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
310 F2FS_NAME_LEN);
311 return f2fs_cf_name_slab ? 0 : -ENOMEM;
312 }
313
f2fs_destroy_casefold_cache(void)314 static void f2fs_destroy_casefold_cache(void)
315 {
316 kmem_cache_destroy(f2fs_cf_name_slab);
317 }
318 #else
f2fs_create_casefold_cache(void)319 static int __init f2fs_create_casefold_cache(void) { return 0; }
f2fs_destroy_casefold_cache(void)320 static void f2fs_destroy_casefold_cache(void) { }
321 #endif
322
limit_reserve_root(struct f2fs_sb_info * sbi)323 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
324 {
325 block_t limit = min((sbi->user_block_count >> 3),
326 sbi->user_block_count - sbi->reserved_blocks);
327
328 /* limit is 12.5% */
329 if (test_opt(sbi, RESERVE_ROOT) &&
330 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
331 F2FS_OPTION(sbi).root_reserved_blocks = limit;
332 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
333 F2FS_OPTION(sbi).root_reserved_blocks);
334 }
335 if (!test_opt(sbi, RESERVE_ROOT) &&
336 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
337 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
338 !gid_eq(F2FS_OPTION(sbi).s_resgid,
339 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
340 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
341 from_kuid_munged(&init_user_ns,
342 F2FS_OPTION(sbi).s_resuid),
343 from_kgid_munged(&init_user_ns,
344 F2FS_OPTION(sbi).s_resgid));
345 }
346
adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)347 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
348 {
349 if (!F2FS_OPTION(sbi).unusable_cap_perc)
350 return;
351
352 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
353 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
354 else
355 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
356 F2FS_OPTION(sbi).unusable_cap_perc;
357
358 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
359 F2FS_OPTION(sbi).unusable_cap,
360 F2FS_OPTION(sbi).unusable_cap_perc);
361 }
362
init_once(void * foo)363 static void init_once(void *foo)
364 {
365 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
366
367 inode_init_once(&fi->vfs_inode);
368 }
369
370 #ifdef CONFIG_QUOTA
371 static const char * const quotatypes[] = INITQFNAMES;
372 #define QTYPE2NAME(t) (quotatypes[t])
f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)373 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
374 substring_t *args)
375 {
376 struct f2fs_sb_info *sbi = F2FS_SB(sb);
377 char *qname;
378 int ret = -EINVAL;
379
380 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
381 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
382 return -EINVAL;
383 }
384 if (f2fs_sb_has_quota_ino(sbi)) {
385 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
386 return 0;
387 }
388
389 qname = match_strdup(args);
390 if (!qname) {
391 f2fs_err(sbi, "Not enough memory for storing quotafile name");
392 return -ENOMEM;
393 }
394 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
395 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
396 ret = 0;
397 else
398 f2fs_err(sbi, "%s quota file already specified",
399 QTYPE2NAME(qtype));
400 goto errout;
401 }
402 if (strchr(qname, '/')) {
403 f2fs_err(sbi, "quotafile must be on filesystem root");
404 goto errout;
405 }
406 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
407 set_opt(sbi, QUOTA);
408 return 0;
409 errout:
410 kfree(qname);
411 return ret;
412 }
413
f2fs_clear_qf_name(struct super_block * sb,int qtype)414 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
415 {
416 struct f2fs_sb_info *sbi = F2FS_SB(sb);
417
418 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
419 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
420 return -EINVAL;
421 }
422 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
423 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
424 return 0;
425 }
426
f2fs_check_quota_options(struct f2fs_sb_info * sbi)427 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
428 {
429 /*
430 * We do the test below only for project quotas. 'usrquota' and
431 * 'grpquota' mount options are allowed even without quota feature
432 * to support legacy quotas in quota files.
433 */
434 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
435 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
436 return -1;
437 }
438 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
439 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
440 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
441 if (test_opt(sbi, USRQUOTA) &&
442 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
443 clear_opt(sbi, USRQUOTA);
444
445 if (test_opt(sbi, GRPQUOTA) &&
446 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
447 clear_opt(sbi, GRPQUOTA);
448
449 if (test_opt(sbi, PRJQUOTA) &&
450 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
451 clear_opt(sbi, PRJQUOTA);
452
453 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
454 test_opt(sbi, PRJQUOTA)) {
455 f2fs_err(sbi, "old and new quota format mixing");
456 return -1;
457 }
458
459 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
460 f2fs_err(sbi, "journaled quota format not specified");
461 return -1;
462 }
463 }
464
465 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
466 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
467 F2FS_OPTION(sbi).s_jquota_fmt = 0;
468 }
469 return 0;
470 }
471 #endif
472
f2fs_set_test_dummy_encryption(struct super_block * sb,const char * opt,const substring_t * arg,bool is_remount)473 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
474 const char *opt,
475 const substring_t *arg,
476 bool is_remount)
477 {
478 struct f2fs_sb_info *sbi = F2FS_SB(sb);
479 struct fs_parameter param = {
480 .type = fs_value_is_string,
481 .string = arg->from ? arg->from : "",
482 };
483 struct fscrypt_dummy_policy *policy =
484 &F2FS_OPTION(sbi).dummy_enc_policy;
485 int err;
486
487 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
488 f2fs_warn(sbi, "test_dummy_encryption option not supported");
489 return -EINVAL;
490 }
491
492 if (!f2fs_sb_has_encrypt(sbi)) {
493 f2fs_err(sbi, "Encrypt feature is off");
494 return -EINVAL;
495 }
496
497 /*
498 * This mount option is just for testing, and it's not worthwhile to
499 * implement the extra complexity (e.g. RCU protection) that would be
500 * needed to allow it to be set or changed during remount. We do allow
501 * it to be specified during remount, but only if there is no change.
502 */
503 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
504 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
505 return -EINVAL;
506 }
507
508 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
509 if (err) {
510 if (err == -EEXIST)
511 f2fs_warn(sbi,
512 "Can't change test_dummy_encryption on remount");
513 else if (err == -EINVAL)
514 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
515 opt);
516 else
517 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
518 opt, err);
519 return -EINVAL;
520 }
521 f2fs_warn(sbi, "Test dummy encryption mode enabled");
522 return 0;
523 }
524
525 #ifdef CONFIG_F2FS_FS_COMPRESSION
is_compress_extension_exist(struct f2fs_sb_info * sbi,const char * new_ext,bool is_ext)526 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
527 const char *new_ext, bool is_ext)
528 {
529 unsigned char (*ext)[F2FS_EXTENSION_LEN];
530 int ext_cnt;
531 int i;
532
533 if (is_ext) {
534 ext = F2FS_OPTION(sbi).extensions;
535 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
536 } else {
537 ext = F2FS_OPTION(sbi).noextensions;
538 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
539 }
540
541 for (i = 0; i < ext_cnt; i++) {
542 if (!strcasecmp(new_ext, ext[i]))
543 return true;
544 }
545
546 return false;
547 }
548
549 /*
550 * 1. The same extension name cannot not appear in both compress and non-compress extension
551 * at the same time.
552 * 2. If the compress extension specifies all files, the types specified by the non-compress
553 * extension will be treated as special cases and will not be compressed.
554 * 3. Don't allow the non-compress extension specifies all files.
555 */
f2fs_test_compress_extension(struct f2fs_sb_info * sbi)556 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
557 {
558 unsigned char (*ext)[F2FS_EXTENSION_LEN];
559 unsigned char (*noext)[F2FS_EXTENSION_LEN];
560 int ext_cnt, noext_cnt, index = 0, no_index = 0;
561
562 ext = F2FS_OPTION(sbi).extensions;
563 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
564 noext = F2FS_OPTION(sbi).noextensions;
565 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
566
567 if (!noext_cnt)
568 return 0;
569
570 for (no_index = 0; no_index < noext_cnt; no_index++) {
571 if (!strcasecmp("*", noext[no_index])) {
572 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
573 return -EINVAL;
574 }
575 for (index = 0; index < ext_cnt; index++) {
576 if (!strcasecmp(ext[index], noext[no_index])) {
577 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
578 ext[index]);
579 return -EINVAL;
580 }
581 }
582 }
583 return 0;
584 }
585
586 #ifdef CONFIG_F2FS_FS_LZ4
f2fs_set_lz4hc_level(struct f2fs_sb_info * sbi,const char * str)587 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
588 {
589 #ifdef CONFIG_F2FS_FS_LZ4HC
590 unsigned int level;
591
592 if (strlen(str) == 3) {
593 F2FS_OPTION(sbi).compress_level = 0;
594 return 0;
595 }
596
597 str += 3;
598
599 if (str[0] != ':') {
600 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
601 return -EINVAL;
602 }
603 if (kstrtouint(str + 1, 10, &level))
604 return -EINVAL;
605
606 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
607 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
608 return -EINVAL;
609 }
610
611 F2FS_OPTION(sbi).compress_level = level;
612 return 0;
613 #else
614 if (strlen(str) == 3) {
615 F2FS_OPTION(sbi).compress_level = 0;
616 return 0;
617 }
618 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
619 return -EINVAL;
620 #endif
621 }
622 #endif
623
624 #ifdef CONFIG_F2FS_FS_ZSTD
f2fs_set_zstd_level(struct f2fs_sb_info * sbi,const char * str)625 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
626 {
627 int level;
628 int len = 4;
629
630 if (strlen(str) == len) {
631 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
632 return 0;
633 }
634
635 str += len;
636
637 if (str[0] != ':') {
638 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
639 return -EINVAL;
640 }
641 if (kstrtoint(str + 1, 10, &level))
642 return -EINVAL;
643
644 /* f2fs does not support negative compress level now */
645 if (level < 0) {
646 f2fs_info(sbi, "do not support negative compress level: %d", level);
647 return -ERANGE;
648 }
649
650 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
651 f2fs_info(sbi, "invalid zstd compress level: %d", level);
652 return -EINVAL;
653 }
654
655 F2FS_OPTION(sbi).compress_level = level;
656 return 0;
657 }
658 #endif
659 #endif
660
parse_options(struct super_block * sb,char * options,bool is_remount)661 static int parse_options(struct super_block *sb, char *options, bool is_remount)
662 {
663 struct f2fs_sb_info *sbi = F2FS_SB(sb);
664 substring_t args[MAX_OPT_ARGS];
665 #ifdef CONFIG_F2FS_FS_COMPRESSION
666 unsigned char (*ext)[F2FS_EXTENSION_LEN];
667 unsigned char (*noext)[F2FS_EXTENSION_LEN];
668 int ext_cnt, noext_cnt;
669 #endif
670 char *p, *name;
671 int arg = 0;
672 kuid_t uid;
673 kgid_t gid;
674 int ret;
675
676 if (!options)
677 goto default_check;
678
679 while ((p = strsep(&options, ",")) != NULL) {
680 int token;
681
682 if (!*p)
683 continue;
684 /*
685 * Initialize args struct so we know whether arg was
686 * found; some options take optional arguments.
687 */
688 args[0].to = args[0].from = NULL;
689 token = match_token(p, f2fs_tokens, args);
690
691 switch (token) {
692 case Opt_gc_background:
693 name = match_strdup(&args[0]);
694
695 if (!name)
696 return -ENOMEM;
697 if (!strcmp(name, "on")) {
698 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
699 } else if (!strcmp(name, "off")) {
700 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
701 } else if (!strcmp(name, "sync")) {
702 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
703 } else {
704 kfree(name);
705 return -EINVAL;
706 }
707 kfree(name);
708 break;
709 case Opt_disable_roll_forward:
710 set_opt(sbi, DISABLE_ROLL_FORWARD);
711 break;
712 case Opt_norecovery:
713 /* this option mounts f2fs with ro */
714 set_opt(sbi, NORECOVERY);
715 if (!f2fs_readonly(sb))
716 return -EINVAL;
717 break;
718 case Opt_discard:
719 if (!f2fs_hw_support_discard(sbi)) {
720 f2fs_warn(sbi, "device does not support discard");
721 break;
722 }
723 set_opt(sbi, DISCARD);
724 break;
725 case Opt_nodiscard:
726 if (f2fs_hw_should_discard(sbi)) {
727 f2fs_warn(sbi, "discard is required for zoned block devices");
728 return -EINVAL;
729 }
730 clear_opt(sbi, DISCARD);
731 break;
732 case Opt_noheap:
733 case Opt_heap:
734 f2fs_warn(sbi, "heap/no_heap options were deprecated");
735 break;
736 #ifdef CONFIG_F2FS_FS_XATTR
737 case Opt_user_xattr:
738 set_opt(sbi, XATTR_USER);
739 break;
740 case Opt_nouser_xattr:
741 clear_opt(sbi, XATTR_USER);
742 break;
743 case Opt_inline_xattr:
744 set_opt(sbi, INLINE_XATTR);
745 break;
746 case Opt_noinline_xattr:
747 clear_opt(sbi, INLINE_XATTR);
748 break;
749 case Opt_inline_xattr_size:
750 if (args->from && match_int(args, &arg))
751 return -EINVAL;
752 set_opt(sbi, INLINE_XATTR_SIZE);
753 F2FS_OPTION(sbi).inline_xattr_size = arg;
754 break;
755 #else
756 case Opt_user_xattr:
757 f2fs_info(sbi, "user_xattr options not supported");
758 break;
759 case Opt_nouser_xattr:
760 f2fs_info(sbi, "nouser_xattr options not supported");
761 break;
762 case Opt_inline_xattr:
763 f2fs_info(sbi, "inline_xattr options not supported");
764 break;
765 case Opt_noinline_xattr:
766 f2fs_info(sbi, "noinline_xattr options not supported");
767 break;
768 #endif
769 #ifdef CONFIG_F2FS_FS_POSIX_ACL
770 case Opt_acl:
771 set_opt(sbi, POSIX_ACL);
772 break;
773 case Opt_noacl:
774 clear_opt(sbi, POSIX_ACL);
775 break;
776 #else
777 case Opt_acl:
778 f2fs_info(sbi, "acl options not supported");
779 break;
780 case Opt_noacl:
781 f2fs_info(sbi, "noacl options not supported");
782 break;
783 #endif
784 case Opt_active_logs:
785 if (args->from && match_int(args, &arg))
786 return -EINVAL;
787 if (arg != 2 && arg != 4 &&
788 arg != NR_CURSEG_PERSIST_TYPE)
789 return -EINVAL;
790 F2FS_OPTION(sbi).active_logs = arg;
791 break;
792 case Opt_disable_ext_identify:
793 set_opt(sbi, DISABLE_EXT_IDENTIFY);
794 break;
795 case Opt_inline_data:
796 set_opt(sbi, INLINE_DATA);
797 break;
798 case Opt_inline_dentry:
799 set_opt(sbi, INLINE_DENTRY);
800 break;
801 case Opt_noinline_dentry:
802 clear_opt(sbi, INLINE_DENTRY);
803 break;
804 case Opt_flush_merge:
805 set_opt(sbi, FLUSH_MERGE);
806 break;
807 case Opt_noflush_merge:
808 clear_opt(sbi, FLUSH_MERGE);
809 break;
810 case Opt_nobarrier:
811 set_opt(sbi, NOBARRIER);
812 break;
813 case Opt_barrier:
814 clear_opt(sbi, NOBARRIER);
815 break;
816 case Opt_fastboot:
817 set_opt(sbi, FASTBOOT);
818 break;
819 case Opt_extent_cache:
820 set_opt(sbi, READ_EXTENT_CACHE);
821 break;
822 case Opt_noextent_cache:
823 clear_opt(sbi, READ_EXTENT_CACHE);
824 break;
825 case Opt_noinline_data:
826 clear_opt(sbi, INLINE_DATA);
827 break;
828 case Opt_data_flush:
829 set_opt(sbi, DATA_FLUSH);
830 break;
831 case Opt_reserve_root:
832 if (args->from && match_int(args, &arg))
833 return -EINVAL;
834 if (test_opt(sbi, RESERVE_ROOT)) {
835 f2fs_info(sbi, "Preserve previous reserve_root=%u",
836 F2FS_OPTION(sbi).root_reserved_blocks);
837 } else {
838 F2FS_OPTION(sbi).root_reserved_blocks = arg;
839 set_opt(sbi, RESERVE_ROOT);
840 }
841 break;
842 case Opt_resuid:
843 if (args->from && match_int(args, &arg))
844 return -EINVAL;
845 uid = make_kuid(current_user_ns(), arg);
846 if (!uid_valid(uid)) {
847 f2fs_err(sbi, "Invalid uid value %d", arg);
848 return -EINVAL;
849 }
850 F2FS_OPTION(sbi).s_resuid = uid;
851 break;
852 case Opt_resgid:
853 if (args->from && match_int(args, &arg))
854 return -EINVAL;
855 gid = make_kgid(current_user_ns(), arg);
856 if (!gid_valid(gid)) {
857 f2fs_err(sbi, "Invalid gid value %d", arg);
858 return -EINVAL;
859 }
860 F2FS_OPTION(sbi).s_resgid = gid;
861 break;
862 case Opt_mode:
863 name = match_strdup(&args[0]);
864
865 if (!name)
866 return -ENOMEM;
867 if (!strcmp(name, "adaptive")) {
868 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
869 } else if (!strcmp(name, "lfs")) {
870 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
871 } else if (!strcmp(name, "fragment:segment")) {
872 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
873 } else if (!strcmp(name, "fragment:block")) {
874 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
875 } else {
876 kfree(name);
877 return -EINVAL;
878 }
879 kfree(name);
880 break;
881 #ifdef CONFIG_F2FS_FAULT_INJECTION
882 case Opt_fault_injection:
883 if (args->from && match_int(args, &arg))
884 return -EINVAL;
885 if (f2fs_build_fault_attr(sbi, arg,
886 F2FS_ALL_FAULT_TYPE))
887 return -EINVAL;
888 set_opt(sbi, FAULT_INJECTION);
889 break;
890
891 case Opt_fault_type:
892 if (args->from && match_int(args, &arg))
893 return -EINVAL;
894 if (f2fs_build_fault_attr(sbi, 0, arg))
895 return -EINVAL;
896 set_opt(sbi, FAULT_INJECTION);
897 break;
898 #else
899 case Opt_fault_injection:
900 f2fs_info(sbi, "fault_injection options not supported");
901 break;
902
903 case Opt_fault_type:
904 f2fs_info(sbi, "fault_type options not supported");
905 break;
906 #endif
907 case Opt_lazytime:
908 sb->s_flags |= SB_LAZYTIME;
909 break;
910 case Opt_nolazytime:
911 sb->s_flags &= ~SB_LAZYTIME;
912 break;
913 #ifdef CONFIG_QUOTA
914 case Opt_quota:
915 case Opt_usrquota:
916 set_opt(sbi, USRQUOTA);
917 break;
918 case Opt_grpquota:
919 set_opt(sbi, GRPQUOTA);
920 break;
921 case Opt_prjquota:
922 set_opt(sbi, PRJQUOTA);
923 break;
924 case Opt_usrjquota:
925 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
926 if (ret)
927 return ret;
928 break;
929 case Opt_grpjquota:
930 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
931 if (ret)
932 return ret;
933 break;
934 case Opt_prjjquota:
935 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
936 if (ret)
937 return ret;
938 break;
939 case Opt_offusrjquota:
940 ret = f2fs_clear_qf_name(sb, USRQUOTA);
941 if (ret)
942 return ret;
943 break;
944 case Opt_offgrpjquota:
945 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
946 if (ret)
947 return ret;
948 break;
949 case Opt_offprjjquota:
950 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
951 if (ret)
952 return ret;
953 break;
954 case Opt_jqfmt_vfsold:
955 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
956 break;
957 case Opt_jqfmt_vfsv0:
958 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
959 break;
960 case Opt_jqfmt_vfsv1:
961 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
962 break;
963 case Opt_noquota:
964 clear_opt(sbi, QUOTA);
965 clear_opt(sbi, USRQUOTA);
966 clear_opt(sbi, GRPQUOTA);
967 clear_opt(sbi, PRJQUOTA);
968 break;
969 #else
970 case Opt_quota:
971 case Opt_usrquota:
972 case Opt_grpquota:
973 case Opt_prjquota:
974 case Opt_usrjquota:
975 case Opt_grpjquota:
976 case Opt_prjjquota:
977 case Opt_offusrjquota:
978 case Opt_offgrpjquota:
979 case Opt_offprjjquota:
980 case Opt_jqfmt_vfsold:
981 case Opt_jqfmt_vfsv0:
982 case Opt_jqfmt_vfsv1:
983 case Opt_noquota:
984 f2fs_info(sbi, "quota operations not supported");
985 break;
986 #endif
987 case Opt_alloc:
988 name = match_strdup(&args[0]);
989 if (!name)
990 return -ENOMEM;
991
992 if (!strcmp(name, "default")) {
993 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
994 } else if (!strcmp(name, "reuse")) {
995 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
996 } else {
997 kfree(name);
998 return -EINVAL;
999 }
1000 kfree(name);
1001 break;
1002 case Opt_fsync:
1003 name = match_strdup(&args[0]);
1004 if (!name)
1005 return -ENOMEM;
1006 if (!strcmp(name, "posix")) {
1007 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1008 } else if (!strcmp(name, "strict")) {
1009 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1010 } else if (!strcmp(name, "nobarrier")) {
1011 F2FS_OPTION(sbi).fsync_mode =
1012 FSYNC_MODE_NOBARRIER;
1013 } else {
1014 kfree(name);
1015 return -EINVAL;
1016 }
1017 kfree(name);
1018 break;
1019 case Opt_test_dummy_encryption:
1020 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1021 is_remount);
1022 if (ret)
1023 return ret;
1024 break;
1025 case Opt_inlinecrypt:
1026 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1027 sb->s_flags |= SB_INLINECRYPT;
1028 #else
1029 f2fs_info(sbi, "inline encryption not supported");
1030 #endif
1031 break;
1032 case Opt_checkpoint_disable_cap_perc:
1033 if (args->from && match_int(args, &arg))
1034 return -EINVAL;
1035 if (arg < 0 || arg > 100)
1036 return -EINVAL;
1037 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1038 set_opt(sbi, DISABLE_CHECKPOINT);
1039 break;
1040 case Opt_checkpoint_disable_cap:
1041 if (args->from && match_int(args, &arg))
1042 return -EINVAL;
1043 F2FS_OPTION(sbi).unusable_cap = arg;
1044 set_opt(sbi, DISABLE_CHECKPOINT);
1045 break;
1046 case Opt_checkpoint_disable:
1047 set_opt(sbi, DISABLE_CHECKPOINT);
1048 break;
1049 case Opt_checkpoint_enable:
1050 clear_opt(sbi, DISABLE_CHECKPOINT);
1051 break;
1052 case Opt_checkpoint_merge:
1053 set_opt(sbi, MERGE_CHECKPOINT);
1054 break;
1055 case Opt_nocheckpoint_merge:
1056 clear_opt(sbi, MERGE_CHECKPOINT);
1057 break;
1058 #ifdef CONFIG_F2FS_FS_COMPRESSION
1059 case Opt_compress_algorithm:
1060 if (!f2fs_sb_has_compression(sbi)) {
1061 f2fs_info(sbi, "Image doesn't support compression");
1062 break;
1063 }
1064 name = match_strdup(&args[0]);
1065 if (!name)
1066 return -ENOMEM;
1067 if (!strcmp(name, "lzo")) {
1068 #ifdef CONFIG_F2FS_FS_LZO
1069 F2FS_OPTION(sbi).compress_level = 0;
1070 F2FS_OPTION(sbi).compress_algorithm =
1071 COMPRESS_LZO;
1072 #else
1073 f2fs_info(sbi, "kernel doesn't support lzo compression");
1074 #endif
1075 } else if (!strncmp(name, "lz4", 3)) {
1076 #ifdef CONFIG_F2FS_FS_LZ4
1077 ret = f2fs_set_lz4hc_level(sbi, name);
1078 if (ret) {
1079 kfree(name);
1080 return -EINVAL;
1081 }
1082 F2FS_OPTION(sbi).compress_algorithm =
1083 COMPRESS_LZ4;
1084 #else
1085 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1086 #endif
1087 } else if (!strncmp(name, "zstd", 4)) {
1088 #ifdef CONFIG_F2FS_FS_ZSTD
1089 ret = f2fs_set_zstd_level(sbi, name);
1090 if (ret) {
1091 kfree(name);
1092 return -EINVAL;
1093 }
1094 F2FS_OPTION(sbi).compress_algorithm =
1095 COMPRESS_ZSTD;
1096 #else
1097 f2fs_info(sbi, "kernel doesn't support zstd compression");
1098 #endif
1099 } else if (!strcmp(name, "lzo-rle")) {
1100 #ifdef CONFIG_F2FS_FS_LZORLE
1101 F2FS_OPTION(sbi).compress_level = 0;
1102 F2FS_OPTION(sbi).compress_algorithm =
1103 COMPRESS_LZORLE;
1104 #else
1105 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1106 #endif
1107 } else {
1108 kfree(name);
1109 return -EINVAL;
1110 }
1111 kfree(name);
1112 break;
1113 case Opt_compress_log_size:
1114 if (!f2fs_sb_has_compression(sbi)) {
1115 f2fs_info(sbi, "Image doesn't support compression");
1116 break;
1117 }
1118 if (args->from && match_int(args, &arg))
1119 return -EINVAL;
1120 if (arg < MIN_COMPRESS_LOG_SIZE ||
1121 arg > MAX_COMPRESS_LOG_SIZE) {
1122 f2fs_err(sbi,
1123 "Compress cluster log size is out of range");
1124 return -EINVAL;
1125 }
1126 F2FS_OPTION(sbi).compress_log_size = arg;
1127 break;
1128 case Opt_compress_extension:
1129 if (!f2fs_sb_has_compression(sbi)) {
1130 f2fs_info(sbi, "Image doesn't support compression");
1131 break;
1132 }
1133 name = match_strdup(&args[0]);
1134 if (!name)
1135 return -ENOMEM;
1136
1137 ext = F2FS_OPTION(sbi).extensions;
1138 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1139
1140 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1141 ext_cnt >= COMPRESS_EXT_NUM) {
1142 f2fs_err(sbi,
1143 "invalid extension length/number");
1144 kfree(name);
1145 return -EINVAL;
1146 }
1147
1148 if (is_compress_extension_exist(sbi, name, true)) {
1149 kfree(name);
1150 break;
1151 }
1152
1153 strcpy(ext[ext_cnt], name);
1154 F2FS_OPTION(sbi).compress_ext_cnt++;
1155 kfree(name);
1156 break;
1157 case Opt_nocompress_extension:
1158 if (!f2fs_sb_has_compression(sbi)) {
1159 f2fs_info(sbi, "Image doesn't support compression");
1160 break;
1161 }
1162 name = match_strdup(&args[0]);
1163 if (!name)
1164 return -ENOMEM;
1165
1166 noext = F2FS_OPTION(sbi).noextensions;
1167 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1168
1169 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1170 noext_cnt >= COMPRESS_EXT_NUM) {
1171 f2fs_err(sbi,
1172 "invalid extension length/number");
1173 kfree(name);
1174 return -EINVAL;
1175 }
1176
1177 if (is_compress_extension_exist(sbi, name, false)) {
1178 kfree(name);
1179 break;
1180 }
1181
1182 strcpy(noext[noext_cnt], name);
1183 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1184 kfree(name);
1185 break;
1186 case Opt_compress_chksum:
1187 if (!f2fs_sb_has_compression(sbi)) {
1188 f2fs_info(sbi, "Image doesn't support compression");
1189 break;
1190 }
1191 F2FS_OPTION(sbi).compress_chksum = true;
1192 break;
1193 case Opt_compress_mode:
1194 if (!f2fs_sb_has_compression(sbi)) {
1195 f2fs_info(sbi, "Image doesn't support compression");
1196 break;
1197 }
1198 name = match_strdup(&args[0]);
1199 if (!name)
1200 return -ENOMEM;
1201 if (!strcmp(name, "fs")) {
1202 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1203 } else if (!strcmp(name, "user")) {
1204 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1205 } else {
1206 kfree(name);
1207 return -EINVAL;
1208 }
1209 kfree(name);
1210 break;
1211 case Opt_compress_cache:
1212 if (!f2fs_sb_has_compression(sbi)) {
1213 f2fs_info(sbi, "Image doesn't support compression");
1214 break;
1215 }
1216 set_opt(sbi, COMPRESS_CACHE);
1217 break;
1218 #else
1219 case Opt_compress_algorithm:
1220 case Opt_compress_log_size:
1221 case Opt_compress_extension:
1222 case Opt_nocompress_extension:
1223 case Opt_compress_chksum:
1224 case Opt_compress_mode:
1225 case Opt_compress_cache:
1226 f2fs_info(sbi, "compression options not supported");
1227 break;
1228 #endif
1229 case Opt_atgc:
1230 set_opt(sbi, ATGC);
1231 break;
1232 case Opt_gc_merge:
1233 set_opt(sbi, GC_MERGE);
1234 break;
1235 case Opt_nogc_merge:
1236 clear_opt(sbi, GC_MERGE);
1237 break;
1238 case Opt_discard_unit:
1239 name = match_strdup(&args[0]);
1240 if (!name)
1241 return -ENOMEM;
1242 if (!strcmp(name, "block")) {
1243 F2FS_OPTION(sbi).discard_unit =
1244 DISCARD_UNIT_BLOCK;
1245 } else if (!strcmp(name, "segment")) {
1246 F2FS_OPTION(sbi).discard_unit =
1247 DISCARD_UNIT_SEGMENT;
1248 } else if (!strcmp(name, "section")) {
1249 F2FS_OPTION(sbi).discard_unit =
1250 DISCARD_UNIT_SECTION;
1251 } else {
1252 kfree(name);
1253 return -EINVAL;
1254 }
1255 kfree(name);
1256 break;
1257 case Opt_memory_mode:
1258 name = match_strdup(&args[0]);
1259 if (!name)
1260 return -ENOMEM;
1261 if (!strcmp(name, "normal")) {
1262 F2FS_OPTION(sbi).memory_mode =
1263 MEMORY_MODE_NORMAL;
1264 } else if (!strcmp(name, "low")) {
1265 F2FS_OPTION(sbi).memory_mode =
1266 MEMORY_MODE_LOW;
1267 } else {
1268 kfree(name);
1269 return -EINVAL;
1270 }
1271 kfree(name);
1272 break;
1273 case Opt_age_extent_cache:
1274 set_opt(sbi, AGE_EXTENT_CACHE);
1275 break;
1276 case Opt_errors:
1277 name = match_strdup(&args[0]);
1278 if (!name)
1279 return -ENOMEM;
1280 if (!strcmp(name, "remount-ro")) {
1281 F2FS_OPTION(sbi).errors =
1282 MOUNT_ERRORS_READONLY;
1283 } else if (!strcmp(name, "continue")) {
1284 F2FS_OPTION(sbi).errors =
1285 MOUNT_ERRORS_CONTINUE;
1286 } else if (!strcmp(name, "panic")) {
1287 F2FS_OPTION(sbi).errors =
1288 MOUNT_ERRORS_PANIC;
1289 } else {
1290 kfree(name);
1291 return -EINVAL;
1292 }
1293 kfree(name);
1294 break;
1295 default:
1296 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1297 p);
1298 return -EINVAL;
1299 }
1300 }
1301 default_check:
1302 #ifdef CONFIG_QUOTA
1303 if (f2fs_check_quota_options(sbi))
1304 return -EINVAL;
1305 #else
1306 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1307 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1308 return -EINVAL;
1309 }
1310 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1311 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1312 return -EINVAL;
1313 }
1314 #endif
1315 #if !IS_ENABLED(CONFIG_UNICODE)
1316 if (f2fs_sb_has_casefold(sbi)) {
1317 f2fs_err(sbi,
1318 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1319 return -EINVAL;
1320 }
1321 #endif
1322 /*
1323 * The BLKZONED feature indicates that the drive was formatted with
1324 * zone alignment optimization. This is optional for host-aware
1325 * devices, but mandatory for host-managed zoned block devices.
1326 */
1327 if (f2fs_sb_has_blkzoned(sbi)) {
1328 #ifdef CONFIG_BLK_DEV_ZONED
1329 if (F2FS_OPTION(sbi).discard_unit !=
1330 DISCARD_UNIT_SECTION) {
1331 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1332 F2FS_OPTION(sbi).discard_unit =
1333 DISCARD_UNIT_SECTION;
1334 }
1335
1336 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1337 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1338 return -EINVAL;
1339 }
1340 #else
1341 f2fs_err(sbi, "Zoned block device support is not enabled");
1342 return -EINVAL;
1343 #endif
1344 }
1345
1346 #ifdef CONFIG_F2FS_FS_COMPRESSION
1347 if (f2fs_test_compress_extension(sbi)) {
1348 f2fs_err(sbi, "invalid compress or nocompress extension");
1349 return -EINVAL;
1350 }
1351 #endif
1352
1353 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1354 int min_size, max_size;
1355
1356 if (!f2fs_sb_has_extra_attr(sbi) ||
1357 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1358 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1359 return -EINVAL;
1360 }
1361 if (!test_opt(sbi, INLINE_XATTR)) {
1362 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1363 return -EINVAL;
1364 }
1365
1366 min_size = MIN_INLINE_XATTR_SIZE;
1367 max_size = MAX_INLINE_XATTR_SIZE;
1368
1369 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1370 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1371 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1372 min_size, max_size);
1373 return -EINVAL;
1374 }
1375 }
1376
1377 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1378 f2fs_err(sbi, "LFS is not compatible with ATGC");
1379 return -EINVAL;
1380 }
1381
1382 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1383 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1384 return -EINVAL;
1385 }
1386
1387 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1388 f2fs_err(sbi, "Allow to mount readonly mode only");
1389 return -EROFS;
1390 }
1391 return 0;
1392 }
1393
f2fs_alloc_inode(struct super_block * sb)1394 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1395 {
1396 struct f2fs_inode_info *fi;
1397
1398 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1399 return NULL;
1400
1401 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1402 if (!fi)
1403 return NULL;
1404
1405 init_once((void *) fi);
1406
1407 /* Initialize f2fs-specific inode info */
1408 atomic_set(&fi->dirty_pages, 0);
1409 atomic_set(&fi->i_compr_blocks, 0);
1410 init_f2fs_rwsem(&fi->i_sem);
1411 spin_lock_init(&fi->i_size_lock);
1412 INIT_LIST_HEAD(&fi->dirty_list);
1413 INIT_LIST_HEAD(&fi->gdirty_list);
1414 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1415 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1416 init_f2fs_rwsem(&fi->i_xattr_sem);
1417
1418 /* Will be used by directory only */
1419 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1420
1421 return &fi->vfs_inode;
1422 }
1423
f2fs_drop_inode(struct inode * inode)1424 static int f2fs_drop_inode(struct inode *inode)
1425 {
1426 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1427 int ret;
1428
1429 /*
1430 * during filesystem shutdown, if checkpoint is disabled,
1431 * drop useless meta/node dirty pages.
1432 */
1433 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1434 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1435 inode->i_ino == F2FS_META_INO(sbi)) {
1436 trace_f2fs_drop_inode(inode, 1);
1437 return 1;
1438 }
1439 }
1440
1441 /*
1442 * This is to avoid a deadlock condition like below.
1443 * writeback_single_inode(inode)
1444 * - f2fs_write_data_page
1445 * - f2fs_gc -> iput -> evict
1446 * - inode_wait_for_writeback(inode)
1447 */
1448 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1449 if (!inode->i_nlink && !is_bad_inode(inode)) {
1450 /* to avoid evict_inode call simultaneously */
1451 atomic_inc(&inode->i_count);
1452 spin_unlock(&inode->i_lock);
1453
1454 /* should remain fi->extent_tree for writepage */
1455 f2fs_destroy_extent_node(inode);
1456
1457 sb_start_intwrite(inode->i_sb);
1458 f2fs_i_size_write(inode, 0);
1459
1460 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1461 inode, NULL, 0, DATA);
1462 truncate_inode_pages_final(inode->i_mapping);
1463
1464 if (F2FS_HAS_BLOCKS(inode))
1465 f2fs_truncate(inode);
1466
1467 sb_end_intwrite(inode->i_sb);
1468
1469 spin_lock(&inode->i_lock);
1470 atomic_dec(&inode->i_count);
1471 }
1472 trace_f2fs_drop_inode(inode, 0);
1473 return 0;
1474 }
1475 ret = generic_drop_inode(inode);
1476 if (!ret)
1477 ret = fscrypt_drop_inode(inode);
1478 trace_f2fs_drop_inode(inode, ret);
1479 return ret;
1480 }
1481
f2fs_inode_dirtied(struct inode * inode,bool sync)1482 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1483 {
1484 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1485 int ret = 0;
1486
1487 spin_lock(&sbi->inode_lock[DIRTY_META]);
1488 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1489 ret = 1;
1490 } else {
1491 set_inode_flag(inode, FI_DIRTY_INODE);
1492 stat_inc_dirty_inode(sbi, DIRTY_META);
1493 }
1494 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1495 list_add_tail(&F2FS_I(inode)->gdirty_list,
1496 &sbi->inode_list[DIRTY_META]);
1497 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1498 }
1499 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1500 return ret;
1501 }
1502
f2fs_inode_synced(struct inode * inode)1503 void f2fs_inode_synced(struct inode *inode)
1504 {
1505 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1506
1507 spin_lock(&sbi->inode_lock[DIRTY_META]);
1508 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1509 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1510 return;
1511 }
1512 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1513 list_del_init(&F2FS_I(inode)->gdirty_list);
1514 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1515 }
1516 clear_inode_flag(inode, FI_DIRTY_INODE);
1517 clear_inode_flag(inode, FI_AUTO_RECOVER);
1518 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1519 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1520 }
1521
1522 /*
1523 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1524 *
1525 * We should call set_dirty_inode to write the dirty inode through write_inode.
1526 */
f2fs_dirty_inode(struct inode * inode,int flags)1527 static void f2fs_dirty_inode(struct inode *inode, int flags)
1528 {
1529 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1530
1531 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1532 inode->i_ino == F2FS_META_INO(sbi))
1533 return;
1534
1535 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1536 clear_inode_flag(inode, FI_AUTO_RECOVER);
1537
1538 f2fs_inode_dirtied(inode, false);
1539 }
1540
f2fs_free_inode(struct inode * inode)1541 static void f2fs_free_inode(struct inode *inode)
1542 {
1543 fscrypt_free_inode(inode);
1544 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1545 }
1546
destroy_percpu_info(struct f2fs_sb_info * sbi)1547 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1548 {
1549 percpu_counter_destroy(&sbi->total_valid_inode_count);
1550 percpu_counter_destroy(&sbi->rf_node_block_count);
1551 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1552 }
1553
destroy_device_list(struct f2fs_sb_info * sbi)1554 static void destroy_device_list(struct f2fs_sb_info *sbi)
1555 {
1556 int i;
1557
1558 for (i = 0; i < sbi->s_ndevs; i++) {
1559 if (i > 0)
1560 blkdev_put(FDEV(i).bdev, sbi->sb);
1561 #ifdef CONFIG_BLK_DEV_ZONED
1562 kvfree(FDEV(i).blkz_seq);
1563 #endif
1564 }
1565 kvfree(sbi->devs);
1566 }
1567
f2fs_put_super(struct super_block * sb)1568 static void f2fs_put_super(struct super_block *sb)
1569 {
1570 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1571 int i;
1572 int err = 0;
1573 bool done;
1574
1575 /* unregister procfs/sysfs entries in advance to avoid race case */
1576 f2fs_unregister_sysfs(sbi);
1577
1578 f2fs_quota_off_umount(sb);
1579
1580 /* prevent remaining shrinker jobs */
1581 mutex_lock(&sbi->umount_mutex);
1582
1583 /*
1584 * flush all issued checkpoints and stop checkpoint issue thread.
1585 * after then, all checkpoints should be done by each process context.
1586 */
1587 f2fs_stop_ckpt_thread(sbi);
1588
1589 /*
1590 * We don't need to do checkpoint when superblock is clean.
1591 * But, the previous checkpoint was not done by umount, it needs to do
1592 * clean checkpoint again.
1593 */
1594 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1595 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1596 struct cp_control cpc = {
1597 .reason = CP_UMOUNT,
1598 };
1599 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1600 err = f2fs_write_checkpoint(sbi, &cpc);
1601 }
1602
1603 /* be sure to wait for any on-going discard commands */
1604 done = f2fs_issue_discard_timeout(sbi);
1605 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1606 struct cp_control cpc = {
1607 .reason = CP_UMOUNT | CP_TRIMMED,
1608 };
1609 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1610 err = f2fs_write_checkpoint(sbi, &cpc);
1611 }
1612
1613 /*
1614 * normally superblock is clean, so we need to release this.
1615 * In addition, EIO will skip do checkpoint, we need this as well.
1616 */
1617 f2fs_release_ino_entry(sbi, true);
1618
1619 f2fs_leave_shrinker(sbi);
1620 mutex_unlock(&sbi->umount_mutex);
1621
1622 /* our cp_error case, we can wait for any writeback page */
1623 f2fs_flush_merged_writes(sbi);
1624
1625 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1626
1627 if (err || f2fs_cp_error(sbi)) {
1628 truncate_inode_pages_final(NODE_MAPPING(sbi));
1629 truncate_inode_pages_final(META_MAPPING(sbi));
1630 }
1631
1632 for (i = 0; i < NR_COUNT_TYPE; i++) {
1633 if (!get_pages(sbi, i))
1634 continue;
1635 f2fs_err(sbi, "detect filesystem reference count leak during "
1636 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1637 f2fs_bug_on(sbi, 1);
1638 }
1639
1640 f2fs_bug_on(sbi, sbi->fsync_node_num);
1641
1642 f2fs_destroy_compress_inode(sbi);
1643
1644 iput(sbi->node_inode);
1645 sbi->node_inode = NULL;
1646
1647 iput(sbi->meta_inode);
1648 sbi->meta_inode = NULL;
1649
1650 /*
1651 * iput() can update stat information, if f2fs_write_checkpoint()
1652 * above failed with error.
1653 */
1654 f2fs_destroy_stats(sbi);
1655
1656 /* destroy f2fs internal modules */
1657 f2fs_destroy_node_manager(sbi);
1658 f2fs_destroy_segment_manager(sbi);
1659
1660 /* flush s_error_work before sbi destroy */
1661 flush_work(&sbi->s_error_work);
1662
1663 f2fs_destroy_post_read_wq(sbi);
1664
1665 kvfree(sbi->ckpt);
1666
1667 if (sbi->s_chksum_driver)
1668 crypto_free_shash(sbi->s_chksum_driver);
1669 kfree(sbi->raw_super);
1670
1671 f2fs_destroy_page_array_cache(sbi);
1672 f2fs_destroy_xattr_caches(sbi);
1673 #ifdef CONFIG_QUOTA
1674 for (i = 0; i < MAXQUOTAS; i++)
1675 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1676 #endif
1677 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1678 destroy_percpu_info(sbi);
1679 f2fs_destroy_iostat(sbi);
1680 for (i = 0; i < NR_PAGE_TYPE; i++)
1681 kvfree(sbi->write_io[i]);
1682 #if IS_ENABLED(CONFIG_UNICODE)
1683 utf8_unload(sb->s_encoding);
1684 #endif
1685 }
1686
f2fs_sync_fs(struct super_block * sb,int sync)1687 int f2fs_sync_fs(struct super_block *sb, int sync)
1688 {
1689 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1690 int err = 0;
1691
1692 if (unlikely(f2fs_cp_error(sbi)))
1693 return 0;
1694 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1695 return 0;
1696
1697 trace_f2fs_sync_fs(sb, sync);
1698
1699 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1700 return -EAGAIN;
1701
1702 if (sync) {
1703 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1704 err = f2fs_issue_checkpoint(sbi);
1705 }
1706
1707 return err;
1708 }
1709
f2fs_freeze(struct super_block * sb)1710 static int f2fs_freeze(struct super_block *sb)
1711 {
1712 if (f2fs_readonly(sb))
1713 return 0;
1714
1715 /* IO error happened before */
1716 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1717 return -EIO;
1718
1719 /* must be clean, since sync_filesystem() was already called */
1720 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1721 return -EINVAL;
1722
1723 /* Let's flush checkpoints and stop the thread. */
1724 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1725
1726 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1727 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1728 return 0;
1729 }
1730
f2fs_unfreeze(struct super_block * sb)1731 static int f2fs_unfreeze(struct super_block *sb)
1732 {
1733 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1734 return 0;
1735 }
1736
1737 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1738 static int f2fs_statfs_project(struct super_block *sb,
1739 kprojid_t projid, struct kstatfs *buf)
1740 {
1741 struct kqid qid;
1742 struct dquot *dquot;
1743 u64 limit;
1744 u64 curblock;
1745
1746 qid = make_kqid_projid(projid);
1747 dquot = dqget(sb, qid);
1748 if (IS_ERR(dquot))
1749 return PTR_ERR(dquot);
1750 spin_lock(&dquot->dq_dqb_lock);
1751
1752 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1753 dquot->dq_dqb.dqb_bhardlimit);
1754 if (limit)
1755 limit >>= sb->s_blocksize_bits;
1756
1757 if (limit && buf->f_blocks > limit) {
1758 curblock = (dquot->dq_dqb.dqb_curspace +
1759 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1760 buf->f_blocks = limit;
1761 buf->f_bfree = buf->f_bavail =
1762 (buf->f_blocks > curblock) ?
1763 (buf->f_blocks - curblock) : 0;
1764 }
1765
1766 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1767 dquot->dq_dqb.dqb_ihardlimit);
1768
1769 if (limit && buf->f_files > limit) {
1770 buf->f_files = limit;
1771 buf->f_ffree =
1772 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1773 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1774 }
1775
1776 spin_unlock(&dquot->dq_dqb_lock);
1777 dqput(dquot);
1778 return 0;
1779 }
1780 #endif
1781
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1782 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1783 {
1784 struct super_block *sb = dentry->d_sb;
1785 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1786 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1787 block_t total_count, user_block_count, start_count;
1788 u64 avail_node_count;
1789 unsigned int total_valid_node_count;
1790
1791 total_count = le64_to_cpu(sbi->raw_super->block_count);
1792 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1793 buf->f_type = F2FS_SUPER_MAGIC;
1794 buf->f_bsize = sbi->blocksize;
1795
1796 buf->f_blocks = total_count - start_count;
1797
1798 spin_lock(&sbi->stat_lock);
1799
1800 user_block_count = sbi->user_block_count;
1801 total_valid_node_count = valid_node_count(sbi);
1802 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1803 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1804 sbi->current_reserved_blocks;
1805
1806 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1807 buf->f_bfree = 0;
1808 else
1809 buf->f_bfree -= sbi->unusable_block_count;
1810 spin_unlock(&sbi->stat_lock);
1811
1812 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1813 buf->f_bavail = buf->f_bfree -
1814 F2FS_OPTION(sbi).root_reserved_blocks;
1815 else
1816 buf->f_bavail = 0;
1817
1818 if (avail_node_count > user_block_count) {
1819 buf->f_files = user_block_count;
1820 buf->f_ffree = buf->f_bavail;
1821 } else {
1822 buf->f_files = avail_node_count;
1823 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1824 buf->f_bavail);
1825 }
1826
1827 buf->f_namelen = F2FS_NAME_LEN;
1828 buf->f_fsid = u64_to_fsid(id);
1829
1830 #ifdef CONFIG_QUOTA
1831 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1832 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1833 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1834 }
1835 #endif
1836 return 0;
1837 }
1838
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1839 static inline void f2fs_show_quota_options(struct seq_file *seq,
1840 struct super_block *sb)
1841 {
1842 #ifdef CONFIG_QUOTA
1843 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1844
1845 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1846 char *fmtname = "";
1847
1848 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1849 case QFMT_VFS_OLD:
1850 fmtname = "vfsold";
1851 break;
1852 case QFMT_VFS_V0:
1853 fmtname = "vfsv0";
1854 break;
1855 case QFMT_VFS_V1:
1856 fmtname = "vfsv1";
1857 break;
1858 }
1859 seq_printf(seq, ",jqfmt=%s", fmtname);
1860 }
1861
1862 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1863 seq_show_option(seq, "usrjquota",
1864 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1865
1866 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1867 seq_show_option(seq, "grpjquota",
1868 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1869
1870 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1871 seq_show_option(seq, "prjjquota",
1872 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1873 #endif
1874 }
1875
1876 #ifdef CONFIG_F2FS_FS_COMPRESSION
f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1877 static inline void f2fs_show_compress_options(struct seq_file *seq,
1878 struct super_block *sb)
1879 {
1880 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1881 char *algtype = "";
1882 int i;
1883
1884 if (!f2fs_sb_has_compression(sbi))
1885 return;
1886
1887 switch (F2FS_OPTION(sbi).compress_algorithm) {
1888 case COMPRESS_LZO:
1889 algtype = "lzo";
1890 break;
1891 case COMPRESS_LZ4:
1892 algtype = "lz4";
1893 break;
1894 case COMPRESS_ZSTD:
1895 algtype = "zstd";
1896 break;
1897 case COMPRESS_LZORLE:
1898 algtype = "lzo-rle";
1899 break;
1900 }
1901 seq_printf(seq, ",compress_algorithm=%s", algtype);
1902
1903 if (F2FS_OPTION(sbi).compress_level)
1904 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1905
1906 seq_printf(seq, ",compress_log_size=%u",
1907 F2FS_OPTION(sbi).compress_log_size);
1908
1909 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1910 seq_printf(seq, ",compress_extension=%s",
1911 F2FS_OPTION(sbi).extensions[i]);
1912 }
1913
1914 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1915 seq_printf(seq, ",nocompress_extension=%s",
1916 F2FS_OPTION(sbi).noextensions[i]);
1917 }
1918
1919 if (F2FS_OPTION(sbi).compress_chksum)
1920 seq_puts(seq, ",compress_chksum");
1921
1922 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1923 seq_printf(seq, ",compress_mode=%s", "fs");
1924 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1925 seq_printf(seq, ",compress_mode=%s", "user");
1926
1927 if (test_opt(sbi, COMPRESS_CACHE))
1928 seq_puts(seq, ",compress_cache");
1929 }
1930 #endif
1931
f2fs_show_options(struct seq_file * seq,struct dentry * root)1932 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1933 {
1934 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1935
1936 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1937 seq_printf(seq, ",background_gc=%s", "sync");
1938 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1939 seq_printf(seq, ",background_gc=%s", "on");
1940 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1941 seq_printf(seq, ",background_gc=%s", "off");
1942
1943 if (test_opt(sbi, GC_MERGE))
1944 seq_puts(seq, ",gc_merge");
1945 else
1946 seq_puts(seq, ",nogc_merge");
1947
1948 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1949 seq_puts(seq, ",disable_roll_forward");
1950 if (test_opt(sbi, NORECOVERY))
1951 seq_puts(seq, ",norecovery");
1952 if (test_opt(sbi, DISCARD)) {
1953 seq_puts(seq, ",discard");
1954 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1955 seq_printf(seq, ",discard_unit=%s", "block");
1956 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1957 seq_printf(seq, ",discard_unit=%s", "segment");
1958 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1959 seq_printf(seq, ",discard_unit=%s", "section");
1960 } else {
1961 seq_puts(seq, ",nodiscard");
1962 }
1963 #ifdef CONFIG_F2FS_FS_XATTR
1964 if (test_opt(sbi, XATTR_USER))
1965 seq_puts(seq, ",user_xattr");
1966 else
1967 seq_puts(seq, ",nouser_xattr");
1968 if (test_opt(sbi, INLINE_XATTR))
1969 seq_puts(seq, ",inline_xattr");
1970 else
1971 seq_puts(seq, ",noinline_xattr");
1972 if (test_opt(sbi, INLINE_XATTR_SIZE))
1973 seq_printf(seq, ",inline_xattr_size=%u",
1974 F2FS_OPTION(sbi).inline_xattr_size);
1975 #endif
1976 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1977 if (test_opt(sbi, POSIX_ACL))
1978 seq_puts(seq, ",acl");
1979 else
1980 seq_puts(seq, ",noacl");
1981 #endif
1982 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1983 seq_puts(seq, ",disable_ext_identify");
1984 if (test_opt(sbi, INLINE_DATA))
1985 seq_puts(seq, ",inline_data");
1986 else
1987 seq_puts(seq, ",noinline_data");
1988 if (test_opt(sbi, INLINE_DENTRY))
1989 seq_puts(seq, ",inline_dentry");
1990 else
1991 seq_puts(seq, ",noinline_dentry");
1992 if (test_opt(sbi, FLUSH_MERGE))
1993 seq_puts(seq, ",flush_merge");
1994 else
1995 seq_puts(seq, ",noflush_merge");
1996 if (test_opt(sbi, NOBARRIER))
1997 seq_puts(seq, ",nobarrier");
1998 else
1999 seq_puts(seq, ",barrier");
2000 if (test_opt(sbi, FASTBOOT))
2001 seq_puts(seq, ",fastboot");
2002 if (test_opt(sbi, READ_EXTENT_CACHE))
2003 seq_puts(seq, ",extent_cache");
2004 else
2005 seq_puts(seq, ",noextent_cache");
2006 if (test_opt(sbi, AGE_EXTENT_CACHE))
2007 seq_puts(seq, ",age_extent_cache");
2008 if (test_opt(sbi, DATA_FLUSH))
2009 seq_puts(seq, ",data_flush");
2010
2011 seq_puts(seq, ",mode=");
2012 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2013 seq_puts(seq, "adaptive");
2014 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2015 seq_puts(seq, "lfs");
2016 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2017 seq_puts(seq, "fragment:segment");
2018 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2019 seq_puts(seq, "fragment:block");
2020 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2021 if (test_opt(sbi, RESERVE_ROOT))
2022 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2023 F2FS_OPTION(sbi).root_reserved_blocks,
2024 from_kuid_munged(&init_user_ns,
2025 F2FS_OPTION(sbi).s_resuid),
2026 from_kgid_munged(&init_user_ns,
2027 F2FS_OPTION(sbi).s_resgid));
2028 #ifdef CONFIG_F2FS_FAULT_INJECTION
2029 if (test_opt(sbi, FAULT_INJECTION)) {
2030 seq_printf(seq, ",fault_injection=%u",
2031 F2FS_OPTION(sbi).fault_info.inject_rate);
2032 seq_printf(seq, ",fault_type=%u",
2033 F2FS_OPTION(sbi).fault_info.inject_type);
2034 }
2035 #endif
2036 #ifdef CONFIG_QUOTA
2037 if (test_opt(sbi, QUOTA))
2038 seq_puts(seq, ",quota");
2039 if (test_opt(sbi, USRQUOTA))
2040 seq_puts(seq, ",usrquota");
2041 if (test_opt(sbi, GRPQUOTA))
2042 seq_puts(seq, ",grpquota");
2043 if (test_opt(sbi, PRJQUOTA))
2044 seq_puts(seq, ",prjquota");
2045 #endif
2046 f2fs_show_quota_options(seq, sbi->sb);
2047
2048 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2049
2050 if (sbi->sb->s_flags & SB_INLINECRYPT)
2051 seq_puts(seq, ",inlinecrypt");
2052
2053 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2054 seq_printf(seq, ",alloc_mode=%s", "default");
2055 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2056 seq_printf(seq, ",alloc_mode=%s", "reuse");
2057
2058 if (test_opt(sbi, DISABLE_CHECKPOINT))
2059 seq_printf(seq, ",checkpoint=disable:%u",
2060 F2FS_OPTION(sbi).unusable_cap);
2061 if (test_opt(sbi, MERGE_CHECKPOINT))
2062 seq_puts(seq, ",checkpoint_merge");
2063 else
2064 seq_puts(seq, ",nocheckpoint_merge");
2065 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2066 seq_printf(seq, ",fsync_mode=%s", "posix");
2067 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2068 seq_printf(seq, ",fsync_mode=%s", "strict");
2069 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2070 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2071
2072 #ifdef CONFIG_F2FS_FS_COMPRESSION
2073 f2fs_show_compress_options(seq, sbi->sb);
2074 #endif
2075
2076 if (test_opt(sbi, ATGC))
2077 seq_puts(seq, ",atgc");
2078
2079 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2080 seq_printf(seq, ",memory=%s", "normal");
2081 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2082 seq_printf(seq, ",memory=%s", "low");
2083
2084 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2085 seq_printf(seq, ",errors=%s", "remount-ro");
2086 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2087 seq_printf(seq, ",errors=%s", "continue");
2088 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2089 seq_printf(seq, ",errors=%s", "panic");
2090
2091 return 0;
2092 }
2093
default_options(struct f2fs_sb_info * sbi,bool remount)2094 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2095 {
2096 /* init some FS parameters */
2097 if (!remount) {
2098 set_opt(sbi, READ_EXTENT_CACHE);
2099 clear_opt(sbi, DISABLE_CHECKPOINT);
2100
2101 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2102 set_opt(sbi, DISCARD);
2103
2104 if (f2fs_sb_has_blkzoned(sbi))
2105 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2106 else
2107 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2108 }
2109
2110 if (f2fs_sb_has_readonly(sbi))
2111 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2112 else
2113 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2114
2115 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2116 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2117 SMALL_VOLUME_SEGMENTS)
2118 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2119 else
2120 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2121 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2122 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2123 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2124 if (f2fs_sb_has_compression(sbi)) {
2125 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2126 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2127 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2128 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2129 }
2130 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2131 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2132 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2133
2134 set_opt(sbi, INLINE_XATTR);
2135 set_opt(sbi, INLINE_DATA);
2136 set_opt(sbi, INLINE_DENTRY);
2137 set_opt(sbi, MERGE_CHECKPOINT);
2138 F2FS_OPTION(sbi).unusable_cap = 0;
2139 sbi->sb->s_flags |= SB_LAZYTIME;
2140 if (!f2fs_is_readonly(sbi))
2141 set_opt(sbi, FLUSH_MERGE);
2142 if (f2fs_sb_has_blkzoned(sbi))
2143 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2144 else
2145 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2146
2147 #ifdef CONFIG_F2FS_FS_XATTR
2148 set_opt(sbi, XATTR_USER);
2149 #endif
2150 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2151 set_opt(sbi, POSIX_ACL);
2152 #endif
2153
2154 f2fs_build_fault_attr(sbi, 0, 0);
2155 }
2156
2157 #ifdef CONFIG_QUOTA
2158 static int f2fs_enable_quotas(struct super_block *sb);
2159 #endif
2160
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)2161 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2162 {
2163 unsigned int s_flags = sbi->sb->s_flags;
2164 struct cp_control cpc;
2165 unsigned int gc_mode = sbi->gc_mode;
2166 int err = 0;
2167 int ret;
2168 block_t unusable;
2169
2170 if (s_flags & SB_RDONLY) {
2171 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2172 return -EINVAL;
2173 }
2174 sbi->sb->s_flags |= SB_ACTIVE;
2175
2176 /* check if we need more GC first */
2177 unusable = f2fs_get_unusable_blocks(sbi);
2178 if (!f2fs_disable_cp_again(sbi, unusable))
2179 goto skip_gc;
2180
2181 f2fs_update_time(sbi, DISABLE_TIME);
2182
2183 sbi->gc_mode = GC_URGENT_HIGH;
2184
2185 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2186 struct f2fs_gc_control gc_control = {
2187 .victim_segno = NULL_SEGNO,
2188 .init_gc_type = FG_GC,
2189 .should_migrate_blocks = false,
2190 .err_gc_skipped = true,
2191 .no_bg_gc = true,
2192 .nr_free_secs = 1 };
2193
2194 f2fs_down_write(&sbi->gc_lock);
2195 stat_inc_gc_call_count(sbi, FOREGROUND);
2196 err = f2fs_gc(sbi, &gc_control);
2197 if (err == -ENODATA) {
2198 err = 0;
2199 break;
2200 }
2201 if (err && err != -EAGAIN)
2202 break;
2203 }
2204
2205 ret = sync_filesystem(sbi->sb);
2206 if (ret || err) {
2207 err = ret ? ret : err;
2208 goto restore_flag;
2209 }
2210
2211 unusable = f2fs_get_unusable_blocks(sbi);
2212 if (f2fs_disable_cp_again(sbi, unusable)) {
2213 err = -EAGAIN;
2214 goto restore_flag;
2215 }
2216
2217 skip_gc:
2218 f2fs_down_write(&sbi->gc_lock);
2219 cpc.reason = CP_PAUSE;
2220 set_sbi_flag(sbi, SBI_CP_DISABLED);
2221 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2222 err = f2fs_write_checkpoint(sbi, &cpc);
2223 if (err)
2224 goto out_unlock;
2225
2226 spin_lock(&sbi->stat_lock);
2227 sbi->unusable_block_count = unusable;
2228 spin_unlock(&sbi->stat_lock);
2229
2230 out_unlock:
2231 f2fs_up_write(&sbi->gc_lock);
2232 restore_flag:
2233 sbi->gc_mode = gc_mode;
2234 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2235 return err;
2236 }
2237
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2238 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2239 {
2240 int retry = DEFAULT_RETRY_IO_COUNT;
2241
2242 /* we should flush all the data to keep data consistency */
2243 do {
2244 sync_inodes_sb(sbi->sb);
2245 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2246 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2247
2248 if (unlikely(retry < 0))
2249 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2250
2251 f2fs_down_write(&sbi->gc_lock);
2252 f2fs_dirty_to_prefree(sbi);
2253
2254 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2255 set_sbi_flag(sbi, SBI_IS_DIRTY);
2256 f2fs_up_write(&sbi->gc_lock);
2257
2258 f2fs_sync_fs(sbi->sb, 1);
2259
2260 /* Let's ensure there's no pending checkpoint anymore */
2261 f2fs_flush_ckpt_thread(sbi);
2262 }
2263
f2fs_remount(struct super_block * sb,int * flags,char * data)2264 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2265 {
2266 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2267 struct f2fs_mount_info org_mount_opt;
2268 unsigned long old_sb_flags;
2269 int err;
2270 bool need_restart_gc = false, need_stop_gc = false;
2271 bool need_restart_flush = false, need_stop_flush = false;
2272 bool need_restart_discard = false, need_stop_discard = false;
2273 bool need_enable_checkpoint = false, need_disable_checkpoint = false;
2274 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2275 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2276 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2277 bool no_atgc = !test_opt(sbi, ATGC);
2278 bool no_discard = !test_opt(sbi, DISCARD);
2279 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2280 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2281 #ifdef CONFIG_QUOTA
2282 int i, j;
2283 #endif
2284
2285 /*
2286 * Save the old mount options in case we
2287 * need to restore them.
2288 */
2289 org_mount_opt = sbi->mount_opt;
2290 old_sb_flags = sb->s_flags;
2291
2292 #ifdef CONFIG_QUOTA
2293 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2294 for (i = 0; i < MAXQUOTAS; i++) {
2295 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2296 org_mount_opt.s_qf_names[i] =
2297 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2298 GFP_KERNEL);
2299 if (!org_mount_opt.s_qf_names[i]) {
2300 for (j = 0; j < i; j++)
2301 kfree(org_mount_opt.s_qf_names[j]);
2302 return -ENOMEM;
2303 }
2304 } else {
2305 org_mount_opt.s_qf_names[i] = NULL;
2306 }
2307 }
2308 #endif
2309
2310 /* recover superblocks we couldn't write due to previous RO mount */
2311 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2312 err = f2fs_commit_super(sbi, false);
2313 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2314 err);
2315 if (!err)
2316 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2317 }
2318
2319 default_options(sbi, true);
2320
2321 /* parse mount options */
2322 err = parse_options(sb, data, true);
2323 if (err)
2324 goto restore_opts;
2325
2326 #ifdef CONFIG_BLK_DEV_ZONED
2327 if (f2fs_sb_has_blkzoned(sbi) &&
2328 sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
2329 f2fs_err(sbi,
2330 "zoned: max open zones %u is too small, need at least %u open zones",
2331 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
2332 err = -EINVAL;
2333 goto restore_opts;
2334 }
2335 #endif
2336
2337 /* flush outstanding errors before changing fs state */
2338 flush_work(&sbi->s_error_work);
2339
2340 /*
2341 * Previous and new state of filesystem is RO,
2342 * so skip checking GC and FLUSH_MERGE conditions.
2343 */
2344 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2345 goto skip;
2346
2347 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2348 err = -EROFS;
2349 goto restore_opts;
2350 }
2351
2352 #ifdef CONFIG_QUOTA
2353 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2354 err = dquot_suspend(sb, -1);
2355 if (err < 0)
2356 goto restore_opts;
2357 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2358 /* dquot_resume needs RW */
2359 sb->s_flags &= ~SB_RDONLY;
2360 if (sb_any_quota_suspended(sb)) {
2361 dquot_resume(sb, -1);
2362 } else if (f2fs_sb_has_quota_ino(sbi)) {
2363 err = f2fs_enable_quotas(sb);
2364 if (err)
2365 goto restore_opts;
2366 }
2367 }
2368 #endif
2369 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2370 err = -EINVAL;
2371 f2fs_warn(sbi, "LFS is not compatible with IPU");
2372 goto restore_opts;
2373 }
2374
2375 /* disallow enable atgc dynamically */
2376 if (no_atgc == !!test_opt(sbi, ATGC)) {
2377 err = -EINVAL;
2378 f2fs_warn(sbi, "switch atgc option is not allowed");
2379 goto restore_opts;
2380 }
2381
2382 /* disallow enable/disable extent_cache dynamically */
2383 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2384 err = -EINVAL;
2385 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2386 goto restore_opts;
2387 }
2388 /* disallow enable/disable age extent_cache dynamically */
2389 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2390 err = -EINVAL;
2391 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2392 goto restore_opts;
2393 }
2394
2395 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2396 err = -EINVAL;
2397 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2398 goto restore_opts;
2399 }
2400
2401 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2402 err = -EINVAL;
2403 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2404 goto restore_opts;
2405 }
2406
2407 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2408 err = -EINVAL;
2409 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2410 goto restore_opts;
2411 }
2412
2413 /*
2414 * We stop the GC thread if FS is mounted as RO
2415 * or if background_gc = off is passed in mount
2416 * option. Also sync the filesystem.
2417 */
2418 if ((*flags & SB_RDONLY) ||
2419 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2420 !test_opt(sbi, GC_MERGE))) {
2421 if (sbi->gc_thread) {
2422 f2fs_stop_gc_thread(sbi);
2423 need_restart_gc = true;
2424 }
2425 } else if (!sbi->gc_thread) {
2426 err = f2fs_start_gc_thread(sbi);
2427 if (err)
2428 goto restore_opts;
2429 need_stop_gc = true;
2430 }
2431
2432 if (*flags & SB_RDONLY) {
2433 sync_inodes_sb(sb);
2434
2435 set_sbi_flag(sbi, SBI_IS_DIRTY);
2436 set_sbi_flag(sbi, SBI_IS_CLOSE);
2437 f2fs_sync_fs(sb, 1);
2438 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2439 }
2440
2441 /*
2442 * We stop issue flush thread if FS is mounted as RO
2443 * or if flush_merge is not passed in mount option.
2444 */
2445 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2446 clear_opt(sbi, FLUSH_MERGE);
2447 f2fs_destroy_flush_cmd_control(sbi, false);
2448 need_restart_flush = true;
2449 } else {
2450 err = f2fs_create_flush_cmd_control(sbi);
2451 if (err)
2452 goto restore_gc;
2453 need_stop_flush = true;
2454 }
2455
2456 if (no_discard == !!test_opt(sbi, DISCARD)) {
2457 if (test_opt(sbi, DISCARD)) {
2458 err = f2fs_start_discard_thread(sbi);
2459 if (err)
2460 goto restore_flush;
2461 need_stop_discard = true;
2462 } else {
2463 f2fs_stop_discard_thread(sbi);
2464 f2fs_issue_discard_timeout(sbi);
2465 need_restart_discard = true;
2466 }
2467 }
2468
2469 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2470 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2471 err = f2fs_disable_checkpoint(sbi);
2472 if (err)
2473 goto restore_discard;
2474 need_enable_checkpoint = true;
2475 } else {
2476 f2fs_enable_checkpoint(sbi);
2477 need_disable_checkpoint = true;
2478 }
2479 }
2480
2481 /*
2482 * Place this routine at the end, since a new checkpoint would be
2483 * triggered while remount and we need to take care of it before
2484 * returning from remount.
2485 */
2486 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2487 !test_opt(sbi, MERGE_CHECKPOINT)) {
2488 f2fs_stop_ckpt_thread(sbi);
2489 } else {
2490 /* Flush if the prevous checkpoint, if exists. */
2491 f2fs_flush_ckpt_thread(sbi);
2492
2493 err = f2fs_start_ckpt_thread(sbi);
2494 if (err) {
2495 f2fs_err(sbi,
2496 "Failed to start F2FS issue_checkpoint_thread (%d)",
2497 err);
2498 goto restore_checkpoint;
2499 }
2500 }
2501
2502 skip:
2503 #ifdef CONFIG_QUOTA
2504 /* Release old quota file names */
2505 for (i = 0; i < MAXQUOTAS; i++)
2506 kfree(org_mount_opt.s_qf_names[i]);
2507 #endif
2508 /* Update the POSIXACL Flag */
2509 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2510 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2511
2512 limit_reserve_root(sbi);
2513 adjust_unusable_cap_perc(sbi);
2514 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2515 return 0;
2516 restore_checkpoint:
2517 if (need_enable_checkpoint) {
2518 f2fs_enable_checkpoint(sbi);
2519 } else if (need_disable_checkpoint) {
2520 if (f2fs_disable_checkpoint(sbi))
2521 f2fs_warn(sbi, "checkpoint has not been disabled");
2522 }
2523 restore_discard:
2524 if (need_restart_discard) {
2525 if (f2fs_start_discard_thread(sbi))
2526 f2fs_warn(sbi, "discard has been stopped");
2527 } else if (need_stop_discard) {
2528 f2fs_stop_discard_thread(sbi);
2529 }
2530 restore_flush:
2531 if (need_restart_flush) {
2532 if (f2fs_create_flush_cmd_control(sbi))
2533 f2fs_warn(sbi, "background flush thread has stopped");
2534 } else if (need_stop_flush) {
2535 clear_opt(sbi, FLUSH_MERGE);
2536 f2fs_destroy_flush_cmd_control(sbi, false);
2537 }
2538 restore_gc:
2539 if (need_restart_gc) {
2540 if (f2fs_start_gc_thread(sbi))
2541 f2fs_warn(sbi, "background gc thread has stopped");
2542 } else if (need_stop_gc) {
2543 f2fs_stop_gc_thread(sbi);
2544 }
2545 restore_opts:
2546 #ifdef CONFIG_QUOTA
2547 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2548 for (i = 0; i < MAXQUOTAS; i++) {
2549 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2550 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2551 }
2552 #endif
2553 sbi->mount_opt = org_mount_opt;
2554 sb->s_flags = old_sb_flags;
2555 return err;
2556 }
2557
f2fs_shutdown(struct super_block * sb)2558 static void f2fs_shutdown(struct super_block *sb)
2559 {
2560 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false);
2561 }
2562
2563 #ifdef CONFIG_QUOTA
f2fs_need_recovery(struct f2fs_sb_info * sbi)2564 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2565 {
2566 /* need to recovery orphan */
2567 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2568 return true;
2569 /* need to recovery data */
2570 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2571 return false;
2572 if (test_opt(sbi, NORECOVERY))
2573 return false;
2574 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2575 }
2576
f2fs_recover_quota_begin(struct f2fs_sb_info * sbi)2577 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2578 {
2579 bool readonly = f2fs_readonly(sbi->sb);
2580
2581 if (!f2fs_need_recovery(sbi))
2582 return false;
2583
2584 /* it doesn't need to check f2fs_sb_has_readonly() */
2585 if (f2fs_hw_is_readonly(sbi))
2586 return false;
2587
2588 if (readonly) {
2589 sbi->sb->s_flags &= ~SB_RDONLY;
2590 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2591 }
2592
2593 /*
2594 * Turn on quotas which were not enabled for read-only mounts if
2595 * filesystem has quota feature, so that they are updated correctly.
2596 */
2597 return f2fs_enable_quota_files(sbi, readonly);
2598 }
2599
f2fs_recover_quota_end(struct f2fs_sb_info * sbi,bool quota_enabled)2600 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2601 bool quota_enabled)
2602 {
2603 if (quota_enabled)
2604 f2fs_quota_off_umount(sbi->sb);
2605
2606 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2607 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2608 sbi->sb->s_flags |= SB_RDONLY;
2609 }
2610 }
2611
2612 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2613 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2614 size_t len, loff_t off)
2615 {
2616 struct inode *inode = sb_dqopt(sb)->files[type];
2617 struct address_space *mapping = inode->i_mapping;
2618 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2619 int offset = off & (sb->s_blocksize - 1);
2620 int tocopy;
2621 size_t toread;
2622 loff_t i_size = i_size_read(inode);
2623 struct page *page;
2624
2625 if (off > i_size)
2626 return 0;
2627
2628 if (off + len > i_size)
2629 len = i_size - off;
2630 toread = len;
2631 while (toread > 0) {
2632 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2633 repeat:
2634 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2635 if (IS_ERR(page)) {
2636 if (PTR_ERR(page) == -ENOMEM) {
2637 memalloc_retry_wait(GFP_NOFS);
2638 goto repeat;
2639 }
2640 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2641 return PTR_ERR(page);
2642 }
2643
2644 lock_page(page);
2645
2646 if (unlikely(page->mapping != mapping)) {
2647 f2fs_put_page(page, 1);
2648 goto repeat;
2649 }
2650 if (unlikely(!PageUptodate(page))) {
2651 f2fs_put_page(page, 1);
2652 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2653 return -EIO;
2654 }
2655
2656 memcpy_from_page(data, page, offset, tocopy);
2657 f2fs_put_page(page, 1);
2658
2659 offset = 0;
2660 toread -= tocopy;
2661 data += tocopy;
2662 blkidx++;
2663 }
2664 return len;
2665 }
2666
2667 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2668 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2669 const char *data, size_t len, loff_t off)
2670 {
2671 struct inode *inode = sb_dqopt(sb)->files[type];
2672 struct address_space *mapping = inode->i_mapping;
2673 const struct address_space_operations *a_ops = mapping->a_ops;
2674 int offset = off & (sb->s_blocksize - 1);
2675 size_t towrite = len;
2676 struct page *page;
2677 void *fsdata = NULL;
2678 int err = 0;
2679 int tocopy;
2680
2681 while (towrite > 0) {
2682 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2683 towrite);
2684 retry:
2685 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2686 &page, &fsdata);
2687 if (unlikely(err)) {
2688 if (err == -ENOMEM) {
2689 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2690 goto retry;
2691 }
2692 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2693 break;
2694 }
2695
2696 memcpy_to_page(page, offset, data, tocopy);
2697
2698 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2699 page, fsdata);
2700 offset = 0;
2701 towrite -= tocopy;
2702 off += tocopy;
2703 data += tocopy;
2704 cond_resched();
2705 }
2706
2707 if (len == towrite)
2708 return err;
2709 inode->i_mtime = inode_set_ctime_current(inode);
2710 f2fs_mark_inode_dirty_sync(inode, false);
2711 return len - towrite;
2712 }
2713
f2fs_dquot_initialize(struct inode * inode)2714 int f2fs_dquot_initialize(struct inode *inode)
2715 {
2716 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2717 return -ESRCH;
2718
2719 return dquot_initialize(inode);
2720 }
2721
f2fs_get_dquots(struct inode * inode)2722 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2723 {
2724 return F2FS_I(inode)->i_dquot;
2725 }
2726
f2fs_get_reserved_space(struct inode * inode)2727 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2728 {
2729 return &F2FS_I(inode)->i_reserved_quota;
2730 }
2731
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2732 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2733 {
2734 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2735 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2736 return 0;
2737 }
2738
2739 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2740 F2FS_OPTION(sbi).s_jquota_fmt, type);
2741 }
2742
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2743 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2744 {
2745 int enabled = 0;
2746 int i, err;
2747
2748 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2749 err = f2fs_enable_quotas(sbi->sb);
2750 if (err) {
2751 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2752 return 0;
2753 }
2754 return 1;
2755 }
2756
2757 for (i = 0; i < MAXQUOTAS; i++) {
2758 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2759 err = f2fs_quota_on_mount(sbi, i);
2760 if (!err) {
2761 enabled = 1;
2762 continue;
2763 }
2764 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2765 err, i);
2766 }
2767 }
2768 return enabled;
2769 }
2770
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2771 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2772 unsigned int flags)
2773 {
2774 struct inode *qf_inode;
2775 unsigned long qf_inum;
2776 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2777 int err;
2778
2779 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2780
2781 qf_inum = f2fs_qf_ino(sb, type);
2782 if (!qf_inum)
2783 return -EPERM;
2784
2785 qf_inode = f2fs_iget(sb, qf_inum);
2786 if (IS_ERR(qf_inode)) {
2787 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2788 return PTR_ERR(qf_inode);
2789 }
2790
2791 /* Don't account quota for quota files to avoid recursion */
2792 inode_lock(qf_inode);
2793 qf_inode->i_flags |= S_NOQUOTA;
2794
2795 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2796 F2FS_I(qf_inode)->i_flags |= qf_flag;
2797 f2fs_set_inode_flags(qf_inode);
2798 }
2799 inode_unlock(qf_inode);
2800
2801 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2802 iput(qf_inode);
2803 return err;
2804 }
2805
f2fs_enable_quotas(struct super_block * sb)2806 static int f2fs_enable_quotas(struct super_block *sb)
2807 {
2808 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2809 int type, err = 0;
2810 unsigned long qf_inum;
2811 bool quota_mopt[MAXQUOTAS] = {
2812 test_opt(sbi, USRQUOTA),
2813 test_opt(sbi, GRPQUOTA),
2814 test_opt(sbi, PRJQUOTA),
2815 };
2816
2817 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2818 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2819 return 0;
2820 }
2821
2822 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2823
2824 for (type = 0; type < MAXQUOTAS; type++) {
2825 qf_inum = f2fs_qf_ino(sb, type);
2826 if (qf_inum) {
2827 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2828 DQUOT_USAGE_ENABLED |
2829 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2830 if (err) {
2831 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2832 type, err);
2833 for (type--; type >= 0; type--)
2834 dquot_quota_off(sb, type);
2835 set_sbi_flag(F2FS_SB(sb),
2836 SBI_QUOTA_NEED_REPAIR);
2837 return err;
2838 }
2839 }
2840 }
2841 return 0;
2842 }
2843
f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2844 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2845 {
2846 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2847 struct address_space *mapping = dqopt->files[type]->i_mapping;
2848 int ret = 0;
2849
2850 ret = dquot_writeback_dquots(sbi->sb, type);
2851 if (ret)
2852 goto out;
2853
2854 ret = filemap_fdatawrite(mapping);
2855 if (ret)
2856 goto out;
2857
2858 /* if we are using journalled quota */
2859 if (is_journalled_quota(sbi))
2860 goto out;
2861
2862 ret = filemap_fdatawait(mapping);
2863
2864 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2865 out:
2866 if (ret)
2867 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2868 return ret;
2869 }
2870
f2fs_quota_sync(struct super_block * sb,int type)2871 int f2fs_quota_sync(struct super_block *sb, int type)
2872 {
2873 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2874 struct quota_info *dqopt = sb_dqopt(sb);
2875 int cnt;
2876 int ret = 0;
2877
2878 /*
2879 * Now when everything is written we can discard the pagecache so
2880 * that userspace sees the changes.
2881 */
2882 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2883
2884 if (type != -1 && cnt != type)
2885 continue;
2886
2887 if (!sb_has_quota_active(sb, cnt))
2888 continue;
2889
2890 if (!f2fs_sb_has_quota_ino(sbi))
2891 inode_lock(dqopt->files[cnt]);
2892
2893 /*
2894 * do_quotactl
2895 * f2fs_quota_sync
2896 * f2fs_down_read(quota_sem)
2897 * dquot_writeback_dquots()
2898 * f2fs_dquot_commit
2899 * block_operation
2900 * f2fs_down_read(quota_sem)
2901 */
2902 f2fs_lock_op(sbi);
2903 f2fs_down_read(&sbi->quota_sem);
2904
2905 ret = f2fs_quota_sync_file(sbi, cnt);
2906
2907 f2fs_up_read(&sbi->quota_sem);
2908 f2fs_unlock_op(sbi);
2909
2910 if (!f2fs_sb_has_quota_ino(sbi))
2911 inode_unlock(dqopt->files[cnt]);
2912
2913 if (ret)
2914 break;
2915 }
2916 return ret;
2917 }
2918
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2919 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2920 const struct path *path)
2921 {
2922 struct inode *inode;
2923 int err;
2924
2925 /* if quota sysfile exists, deny enabling quota with specific file */
2926 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2927 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2928 return -EBUSY;
2929 }
2930
2931 if (path->dentry->d_sb != sb)
2932 return -EXDEV;
2933
2934 err = f2fs_quota_sync(sb, type);
2935 if (err)
2936 return err;
2937
2938 inode = d_inode(path->dentry);
2939
2940 err = filemap_fdatawrite(inode->i_mapping);
2941 if (err)
2942 return err;
2943
2944 err = filemap_fdatawait(inode->i_mapping);
2945 if (err)
2946 return err;
2947
2948 err = dquot_quota_on(sb, type, format_id, path);
2949 if (err)
2950 return err;
2951
2952 inode_lock(inode);
2953 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2954 f2fs_set_inode_flags(inode);
2955 inode_unlock(inode);
2956 f2fs_mark_inode_dirty_sync(inode, false);
2957
2958 return 0;
2959 }
2960
__f2fs_quota_off(struct super_block * sb,int type)2961 static int __f2fs_quota_off(struct super_block *sb, int type)
2962 {
2963 struct inode *inode = sb_dqopt(sb)->files[type];
2964 int err;
2965
2966 if (!inode || !igrab(inode))
2967 return dquot_quota_off(sb, type);
2968
2969 err = f2fs_quota_sync(sb, type);
2970 if (err)
2971 goto out_put;
2972
2973 err = dquot_quota_off(sb, type);
2974 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2975 goto out_put;
2976
2977 inode_lock(inode);
2978 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
2979 f2fs_set_inode_flags(inode);
2980 inode_unlock(inode);
2981 f2fs_mark_inode_dirty_sync(inode, false);
2982 out_put:
2983 iput(inode);
2984 return err;
2985 }
2986
f2fs_quota_off(struct super_block * sb,int type)2987 static int f2fs_quota_off(struct super_block *sb, int type)
2988 {
2989 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2990 int err;
2991
2992 err = __f2fs_quota_off(sb, type);
2993
2994 /*
2995 * quotactl can shutdown journalled quota, result in inconsistence
2996 * between quota record and fs data by following updates, tag the
2997 * flag to let fsck be aware of it.
2998 */
2999 if (is_journalled_quota(sbi))
3000 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3001 return err;
3002 }
3003
f2fs_quota_off_umount(struct super_block * sb)3004 void f2fs_quota_off_umount(struct super_block *sb)
3005 {
3006 int type;
3007 int err;
3008
3009 for (type = 0; type < MAXQUOTAS; type++) {
3010 err = __f2fs_quota_off(sb, type);
3011 if (err) {
3012 int ret = dquot_quota_off(sb, type);
3013
3014 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3015 type, err, ret);
3016 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3017 }
3018 }
3019 /*
3020 * In case of checkpoint=disable, we must flush quota blocks.
3021 * This can cause NULL exception for node_inode in end_io, since
3022 * put_super already dropped it.
3023 */
3024 sync_filesystem(sb);
3025 }
3026
f2fs_truncate_quota_inode_pages(struct super_block * sb)3027 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3028 {
3029 struct quota_info *dqopt = sb_dqopt(sb);
3030 int type;
3031
3032 for (type = 0; type < MAXQUOTAS; type++) {
3033 if (!dqopt->files[type])
3034 continue;
3035 f2fs_inode_synced(dqopt->files[type]);
3036 }
3037 }
3038
f2fs_dquot_commit(struct dquot * dquot)3039 static int f2fs_dquot_commit(struct dquot *dquot)
3040 {
3041 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3042 int ret;
3043
3044 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3045 ret = dquot_commit(dquot);
3046 if (ret < 0)
3047 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3048 f2fs_up_read(&sbi->quota_sem);
3049 return ret;
3050 }
3051
f2fs_dquot_acquire(struct dquot * dquot)3052 static int f2fs_dquot_acquire(struct dquot *dquot)
3053 {
3054 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3055 int ret;
3056
3057 f2fs_down_read(&sbi->quota_sem);
3058 ret = dquot_acquire(dquot);
3059 if (ret < 0)
3060 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3061 f2fs_up_read(&sbi->quota_sem);
3062 return ret;
3063 }
3064
f2fs_dquot_release(struct dquot * dquot)3065 static int f2fs_dquot_release(struct dquot *dquot)
3066 {
3067 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3068 int ret = dquot_release(dquot);
3069
3070 if (ret < 0)
3071 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3072 return ret;
3073 }
3074
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)3075 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3076 {
3077 struct super_block *sb = dquot->dq_sb;
3078 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3079 int ret = dquot_mark_dquot_dirty(dquot);
3080
3081 /* if we are using journalled quota */
3082 if (is_journalled_quota(sbi))
3083 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3084
3085 return ret;
3086 }
3087
f2fs_dquot_commit_info(struct super_block * sb,int type)3088 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3089 {
3090 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3091 int ret = dquot_commit_info(sb, type);
3092
3093 if (ret < 0)
3094 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3095 return ret;
3096 }
3097
f2fs_get_projid(struct inode * inode,kprojid_t * projid)3098 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3099 {
3100 *projid = F2FS_I(inode)->i_projid;
3101 return 0;
3102 }
3103
3104 static const struct dquot_operations f2fs_quota_operations = {
3105 .get_reserved_space = f2fs_get_reserved_space,
3106 .write_dquot = f2fs_dquot_commit,
3107 .acquire_dquot = f2fs_dquot_acquire,
3108 .release_dquot = f2fs_dquot_release,
3109 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3110 .write_info = f2fs_dquot_commit_info,
3111 .alloc_dquot = dquot_alloc,
3112 .destroy_dquot = dquot_destroy,
3113 .get_projid = f2fs_get_projid,
3114 .get_next_id = dquot_get_next_id,
3115 };
3116
3117 static const struct quotactl_ops f2fs_quotactl_ops = {
3118 .quota_on = f2fs_quota_on,
3119 .quota_off = f2fs_quota_off,
3120 .quota_sync = f2fs_quota_sync,
3121 .get_state = dquot_get_state,
3122 .set_info = dquot_set_dqinfo,
3123 .get_dqblk = dquot_get_dqblk,
3124 .set_dqblk = dquot_set_dqblk,
3125 .get_nextdqblk = dquot_get_next_dqblk,
3126 };
3127 #else
f2fs_dquot_initialize(struct inode * inode)3128 int f2fs_dquot_initialize(struct inode *inode)
3129 {
3130 return 0;
3131 }
3132
f2fs_quota_sync(struct super_block * sb,int type)3133 int f2fs_quota_sync(struct super_block *sb, int type)
3134 {
3135 return 0;
3136 }
3137
f2fs_quota_off_umount(struct super_block * sb)3138 void f2fs_quota_off_umount(struct super_block *sb)
3139 {
3140 }
3141 #endif
3142
3143 static const struct super_operations f2fs_sops = {
3144 .alloc_inode = f2fs_alloc_inode,
3145 .free_inode = f2fs_free_inode,
3146 .drop_inode = f2fs_drop_inode,
3147 .write_inode = f2fs_write_inode,
3148 .dirty_inode = f2fs_dirty_inode,
3149 .show_options = f2fs_show_options,
3150 #ifdef CONFIG_QUOTA
3151 .quota_read = f2fs_quota_read,
3152 .quota_write = f2fs_quota_write,
3153 .get_dquots = f2fs_get_dquots,
3154 #endif
3155 .evict_inode = f2fs_evict_inode,
3156 .put_super = f2fs_put_super,
3157 .sync_fs = f2fs_sync_fs,
3158 .freeze_fs = f2fs_freeze,
3159 .unfreeze_fs = f2fs_unfreeze,
3160 .statfs = f2fs_statfs,
3161 .remount_fs = f2fs_remount,
3162 .shutdown = f2fs_shutdown,
3163 };
3164
3165 #ifdef CONFIG_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)3166 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3167 {
3168 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3169 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3170 ctx, len, NULL);
3171 }
3172
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)3173 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3174 void *fs_data)
3175 {
3176 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3177
3178 /*
3179 * Encrypting the root directory is not allowed because fsck
3180 * expects lost+found directory to exist and remain unencrypted
3181 * if LOST_FOUND feature is enabled.
3182 *
3183 */
3184 if (f2fs_sb_has_lost_found(sbi) &&
3185 inode->i_ino == F2FS_ROOT_INO(sbi))
3186 return -EPERM;
3187
3188 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3189 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3190 ctx, len, fs_data, XATTR_CREATE);
3191 }
3192
f2fs_get_dummy_policy(struct super_block * sb)3193 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3194 {
3195 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3196 }
3197
f2fs_has_stable_inodes(struct super_block * sb)3198 static bool f2fs_has_stable_inodes(struct super_block *sb)
3199 {
3200 return true;
3201 }
3202
f2fs_get_devices(struct super_block * sb,unsigned int * num_devs)3203 static struct block_device **f2fs_get_devices(struct super_block *sb,
3204 unsigned int *num_devs)
3205 {
3206 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3207 struct block_device **devs;
3208 int i;
3209
3210 if (!f2fs_is_multi_device(sbi))
3211 return NULL;
3212
3213 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3214 if (!devs)
3215 return ERR_PTR(-ENOMEM);
3216
3217 for (i = 0; i < sbi->s_ndevs; i++)
3218 devs[i] = FDEV(i).bdev;
3219 *num_devs = sbi->s_ndevs;
3220 return devs;
3221 }
3222
3223 static const struct fscrypt_operations f2fs_cryptops = {
3224 .needs_bounce_pages = 1,
3225 .has_32bit_inodes = 1,
3226 .supports_subblock_data_units = 1,
3227 .legacy_key_prefix = "f2fs:",
3228 .get_context = f2fs_get_context,
3229 .set_context = f2fs_set_context,
3230 .get_dummy_policy = f2fs_get_dummy_policy,
3231 .empty_dir = f2fs_empty_dir,
3232 .has_stable_inodes = f2fs_has_stable_inodes,
3233 .get_devices = f2fs_get_devices,
3234 };
3235 #endif
3236
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)3237 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3238 u64 ino, u32 generation)
3239 {
3240 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3241 struct inode *inode;
3242
3243 if (f2fs_check_nid_range(sbi, ino))
3244 return ERR_PTR(-ESTALE);
3245
3246 /*
3247 * f2fs_iget isn't quite right if the inode is currently unallocated!
3248 * However f2fs_iget currently does appropriate checks to handle stale
3249 * inodes so everything is OK.
3250 */
3251 inode = f2fs_iget(sb, ino);
3252 if (IS_ERR(inode))
3253 return ERR_CAST(inode);
3254 if (unlikely(generation && inode->i_generation != generation)) {
3255 /* we didn't find the right inode.. */
3256 iput(inode);
3257 return ERR_PTR(-ESTALE);
3258 }
3259 return inode;
3260 }
3261
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3262 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3263 int fh_len, int fh_type)
3264 {
3265 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3266 f2fs_nfs_get_inode);
3267 }
3268
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3269 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3270 int fh_len, int fh_type)
3271 {
3272 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3273 f2fs_nfs_get_inode);
3274 }
3275
3276 static const struct export_operations f2fs_export_ops = {
3277 .fh_to_dentry = f2fs_fh_to_dentry,
3278 .fh_to_parent = f2fs_fh_to_parent,
3279 .get_parent = f2fs_get_parent,
3280 };
3281
max_file_blocks(struct inode * inode)3282 loff_t max_file_blocks(struct inode *inode)
3283 {
3284 loff_t result = 0;
3285 loff_t leaf_count;
3286
3287 /*
3288 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3289 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3290 * space in inode.i_addr, it will be more safe to reassign
3291 * result as zero.
3292 */
3293
3294 if (inode && f2fs_compressed_file(inode))
3295 leaf_count = ADDRS_PER_BLOCK(inode);
3296 else
3297 leaf_count = DEF_ADDRS_PER_BLOCK;
3298
3299 /* two direct node blocks */
3300 result += (leaf_count * 2);
3301
3302 /* two indirect node blocks */
3303 leaf_count *= NIDS_PER_BLOCK;
3304 result += (leaf_count * 2);
3305
3306 /* one double indirect node block */
3307 leaf_count *= NIDS_PER_BLOCK;
3308 result += leaf_count;
3309
3310 /*
3311 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with
3312 * a 4K crypto data unit, we must restrict the max filesize to what can
3313 * fit within U32_MAX + 1 data units.
3314 */
3315
3316 result = min(result, (((loff_t)U32_MAX + 1) * 4096) >> F2FS_BLKSIZE_BITS);
3317
3318 return result;
3319 }
3320
__f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)3321 static int __f2fs_commit_super(struct buffer_head *bh,
3322 struct f2fs_super_block *super)
3323 {
3324 lock_buffer(bh);
3325 if (super)
3326 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3327 set_buffer_dirty(bh);
3328 unlock_buffer(bh);
3329
3330 /* it's rare case, we can do fua all the time */
3331 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3332 }
3333
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)3334 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3335 struct buffer_head *bh)
3336 {
3337 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3338 (bh->b_data + F2FS_SUPER_OFFSET);
3339 struct super_block *sb = sbi->sb;
3340 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3341 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3342 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3343 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3344 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3345 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3346 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3347 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3348 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3349 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3350 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3351 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3352 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3353 u64 main_end_blkaddr = main_blkaddr +
3354 (segment_count_main << log_blocks_per_seg);
3355 u64 seg_end_blkaddr = segment0_blkaddr +
3356 (segment_count << log_blocks_per_seg);
3357
3358 if (segment0_blkaddr != cp_blkaddr) {
3359 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3360 segment0_blkaddr, cp_blkaddr);
3361 return true;
3362 }
3363
3364 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3365 sit_blkaddr) {
3366 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3367 cp_blkaddr, sit_blkaddr,
3368 segment_count_ckpt << log_blocks_per_seg);
3369 return true;
3370 }
3371
3372 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3373 nat_blkaddr) {
3374 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3375 sit_blkaddr, nat_blkaddr,
3376 segment_count_sit << log_blocks_per_seg);
3377 return true;
3378 }
3379
3380 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3381 ssa_blkaddr) {
3382 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3383 nat_blkaddr, ssa_blkaddr,
3384 segment_count_nat << log_blocks_per_seg);
3385 return true;
3386 }
3387
3388 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3389 main_blkaddr) {
3390 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3391 ssa_blkaddr, main_blkaddr,
3392 segment_count_ssa << log_blocks_per_seg);
3393 return true;
3394 }
3395
3396 if (main_end_blkaddr > seg_end_blkaddr) {
3397 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3398 main_blkaddr, seg_end_blkaddr,
3399 segment_count_main << log_blocks_per_seg);
3400 return true;
3401 } else if (main_end_blkaddr < seg_end_blkaddr) {
3402 int err = 0;
3403 char *res;
3404
3405 /* fix in-memory information all the time */
3406 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3407 segment0_blkaddr) >> log_blocks_per_seg);
3408
3409 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3410 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3411 res = "internally";
3412 } else {
3413 err = __f2fs_commit_super(bh, NULL);
3414 res = err ? "failed" : "done";
3415 }
3416 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3417 res, main_blkaddr, seg_end_blkaddr,
3418 segment_count_main << log_blocks_per_seg);
3419 if (err)
3420 return true;
3421 }
3422 return false;
3423 }
3424
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)3425 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3426 struct buffer_head *bh)
3427 {
3428 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3429 block_t total_sections, blocks_per_seg;
3430 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3431 (bh->b_data + F2FS_SUPER_OFFSET);
3432 size_t crc_offset = 0;
3433 __u32 crc = 0;
3434
3435 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3436 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3437 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3438 return -EINVAL;
3439 }
3440
3441 /* Check checksum_offset and crc in superblock */
3442 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3443 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3444 if (crc_offset !=
3445 offsetof(struct f2fs_super_block, crc)) {
3446 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3447 crc_offset);
3448 return -EFSCORRUPTED;
3449 }
3450 crc = le32_to_cpu(raw_super->crc);
3451 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3452 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3453 return -EFSCORRUPTED;
3454 }
3455 }
3456
3457 /* only support block_size equals to PAGE_SIZE */
3458 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3459 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3460 le32_to_cpu(raw_super->log_blocksize),
3461 F2FS_BLKSIZE_BITS);
3462 return -EFSCORRUPTED;
3463 }
3464
3465 /* check log blocks per segment */
3466 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3467 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3468 le32_to_cpu(raw_super->log_blocks_per_seg));
3469 return -EFSCORRUPTED;
3470 }
3471
3472 /* Currently, support 512/1024/2048/4096/16K bytes sector size */
3473 if (le32_to_cpu(raw_super->log_sectorsize) >
3474 F2FS_MAX_LOG_SECTOR_SIZE ||
3475 le32_to_cpu(raw_super->log_sectorsize) <
3476 F2FS_MIN_LOG_SECTOR_SIZE) {
3477 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3478 le32_to_cpu(raw_super->log_sectorsize));
3479 return -EFSCORRUPTED;
3480 }
3481 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3482 le32_to_cpu(raw_super->log_sectorsize) !=
3483 F2FS_MAX_LOG_SECTOR_SIZE) {
3484 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3485 le32_to_cpu(raw_super->log_sectors_per_block),
3486 le32_to_cpu(raw_super->log_sectorsize));
3487 return -EFSCORRUPTED;
3488 }
3489
3490 segment_count = le32_to_cpu(raw_super->segment_count);
3491 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3492 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3493 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3494 total_sections = le32_to_cpu(raw_super->section_count);
3495
3496 /* blocks_per_seg should be 512, given the above check */
3497 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3498
3499 if (segment_count > F2FS_MAX_SEGMENT ||
3500 segment_count < F2FS_MIN_SEGMENTS) {
3501 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3502 return -EFSCORRUPTED;
3503 }
3504
3505 if (total_sections > segment_count_main || total_sections < 1 ||
3506 segs_per_sec > segment_count || !segs_per_sec) {
3507 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3508 segment_count, total_sections, segs_per_sec);
3509 return -EFSCORRUPTED;
3510 }
3511
3512 if (segment_count_main != total_sections * segs_per_sec) {
3513 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3514 segment_count_main, total_sections, segs_per_sec);
3515 return -EFSCORRUPTED;
3516 }
3517
3518 if ((segment_count / segs_per_sec) < total_sections) {
3519 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3520 segment_count, segs_per_sec, total_sections);
3521 return -EFSCORRUPTED;
3522 }
3523
3524 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3525 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3526 segment_count, le64_to_cpu(raw_super->block_count));
3527 return -EFSCORRUPTED;
3528 }
3529
3530 if (RDEV(0).path[0]) {
3531 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3532 int i = 1;
3533
3534 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3535 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3536 i++;
3537 }
3538 if (segment_count != dev_seg_count) {
3539 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3540 segment_count, dev_seg_count);
3541 return -EFSCORRUPTED;
3542 }
3543 } else {
3544 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3545 !bdev_is_zoned(sbi->sb->s_bdev)) {
3546 f2fs_info(sbi, "Zoned block device path is missing");
3547 return -EFSCORRUPTED;
3548 }
3549 }
3550
3551 if (secs_per_zone > total_sections || !secs_per_zone) {
3552 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3553 secs_per_zone, total_sections);
3554 return -EFSCORRUPTED;
3555 }
3556 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3557 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3558 (le32_to_cpu(raw_super->extension_count) +
3559 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3560 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3561 le32_to_cpu(raw_super->extension_count),
3562 raw_super->hot_ext_count,
3563 F2FS_MAX_EXTENSION);
3564 return -EFSCORRUPTED;
3565 }
3566
3567 if (le32_to_cpu(raw_super->cp_payload) >=
3568 (blocks_per_seg - F2FS_CP_PACKS -
3569 NR_CURSEG_PERSIST_TYPE)) {
3570 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3571 le32_to_cpu(raw_super->cp_payload),
3572 blocks_per_seg - F2FS_CP_PACKS -
3573 NR_CURSEG_PERSIST_TYPE);
3574 return -EFSCORRUPTED;
3575 }
3576
3577 /* check reserved ino info */
3578 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3579 le32_to_cpu(raw_super->meta_ino) != 2 ||
3580 le32_to_cpu(raw_super->root_ino) != 3) {
3581 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3582 le32_to_cpu(raw_super->node_ino),
3583 le32_to_cpu(raw_super->meta_ino),
3584 le32_to_cpu(raw_super->root_ino));
3585 return -EFSCORRUPTED;
3586 }
3587
3588 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3589 if (sanity_check_area_boundary(sbi, bh))
3590 return -EFSCORRUPTED;
3591
3592 return 0;
3593 }
3594
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3595 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3596 {
3597 unsigned int total, fsmeta;
3598 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3599 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3600 unsigned int ovp_segments, reserved_segments;
3601 unsigned int main_segs, blocks_per_seg;
3602 unsigned int sit_segs, nat_segs;
3603 unsigned int sit_bitmap_size, nat_bitmap_size;
3604 unsigned int log_blocks_per_seg;
3605 unsigned int segment_count_main;
3606 unsigned int cp_pack_start_sum, cp_payload;
3607 block_t user_block_count, valid_user_blocks;
3608 block_t avail_node_count, valid_node_count;
3609 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3610 int i, j;
3611
3612 total = le32_to_cpu(raw_super->segment_count);
3613 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3614 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3615 fsmeta += sit_segs;
3616 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3617 fsmeta += nat_segs;
3618 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3619 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3620
3621 if (unlikely(fsmeta >= total))
3622 return 1;
3623
3624 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3625 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3626
3627 if (!f2fs_sb_has_readonly(sbi) &&
3628 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3629 ovp_segments == 0 || reserved_segments == 0)) {
3630 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3631 return 1;
3632 }
3633 user_block_count = le64_to_cpu(ckpt->user_block_count);
3634 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3635 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3636 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3637 if (!user_block_count || user_block_count >=
3638 segment_count_main << log_blocks_per_seg) {
3639 f2fs_err(sbi, "Wrong user_block_count: %u",
3640 user_block_count);
3641 return 1;
3642 }
3643
3644 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3645 if (valid_user_blocks > user_block_count) {
3646 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3647 valid_user_blocks, user_block_count);
3648 return 1;
3649 }
3650
3651 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3652 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3653 if (valid_node_count > avail_node_count) {
3654 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3655 valid_node_count, avail_node_count);
3656 return 1;
3657 }
3658
3659 main_segs = le32_to_cpu(raw_super->segment_count_main);
3660 blocks_per_seg = BLKS_PER_SEG(sbi);
3661
3662 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3663 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3664 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3665 return 1;
3666
3667 if (f2fs_sb_has_readonly(sbi))
3668 goto check_data;
3669
3670 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3671 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3672 le32_to_cpu(ckpt->cur_node_segno[j])) {
3673 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3674 i, j,
3675 le32_to_cpu(ckpt->cur_node_segno[i]));
3676 return 1;
3677 }
3678 }
3679 }
3680 check_data:
3681 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3682 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3683 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3684 return 1;
3685
3686 if (f2fs_sb_has_readonly(sbi))
3687 goto skip_cross;
3688
3689 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3690 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3691 le32_to_cpu(ckpt->cur_data_segno[j])) {
3692 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3693 i, j,
3694 le32_to_cpu(ckpt->cur_data_segno[i]));
3695 return 1;
3696 }
3697 }
3698 }
3699 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3700 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3701 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3702 le32_to_cpu(ckpt->cur_data_segno[j])) {
3703 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3704 i, j,
3705 le32_to_cpu(ckpt->cur_node_segno[i]));
3706 return 1;
3707 }
3708 }
3709 }
3710 skip_cross:
3711 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3712 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3713
3714 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3715 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3716 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3717 sit_bitmap_size, nat_bitmap_size);
3718 return 1;
3719 }
3720
3721 cp_pack_start_sum = __start_sum_addr(sbi);
3722 cp_payload = __cp_payload(sbi);
3723 if (cp_pack_start_sum < cp_payload + 1 ||
3724 cp_pack_start_sum > blocks_per_seg - 1 -
3725 NR_CURSEG_PERSIST_TYPE) {
3726 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3727 cp_pack_start_sum);
3728 return 1;
3729 }
3730
3731 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3732 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3733 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3734 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3735 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3736 le32_to_cpu(ckpt->checksum_offset));
3737 return 1;
3738 }
3739
3740 nat_blocks = nat_segs << log_blocks_per_seg;
3741 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3742 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3743 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3744 (cp_payload + F2FS_CP_PACKS +
3745 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3746 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3747 cp_payload, nat_bits_blocks);
3748 return 1;
3749 }
3750
3751 if (unlikely(f2fs_cp_error(sbi))) {
3752 f2fs_err(sbi, "A bug case: need to run fsck");
3753 return 1;
3754 }
3755 return 0;
3756 }
3757
init_sb_info(struct f2fs_sb_info * sbi)3758 static void init_sb_info(struct f2fs_sb_info *sbi)
3759 {
3760 struct f2fs_super_block *raw_super = sbi->raw_super;
3761 int i;
3762
3763 sbi->log_sectors_per_block =
3764 le32_to_cpu(raw_super->log_sectors_per_block);
3765 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3766 sbi->blocksize = BIT(sbi->log_blocksize);
3767 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3768 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3769 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3770 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3771 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3772 sbi->total_node_count = SEGS_TO_BLKS(sbi,
3773 ((le32_to_cpu(raw_super->segment_count_nat) / 2) *
3774 NAT_ENTRY_PER_BLOCK));
3775 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3776 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3777 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3778 sbi->cur_victim_sec = NULL_SECNO;
3779 sbi->gc_mode = GC_NORMAL;
3780 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3781 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3782 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3783 sbi->migration_granularity = SEGS_PER_SEC(sbi);
3784 sbi->seq_file_ra_mul = MIN_RA_MUL;
3785 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3786 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3787 spin_lock_init(&sbi->gc_remaining_trials_lock);
3788 atomic64_set(&sbi->current_atomic_write, 0);
3789
3790 sbi->dir_level = DEF_DIR_LEVEL;
3791 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3792 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3793 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3794 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3795 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3796 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3797 DEF_UMOUNT_DISCARD_TIMEOUT;
3798 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3799
3800 for (i = 0; i < NR_COUNT_TYPE; i++)
3801 atomic_set(&sbi->nr_pages[i], 0);
3802
3803 for (i = 0; i < META; i++)
3804 atomic_set(&sbi->wb_sync_req[i], 0);
3805
3806 INIT_LIST_HEAD(&sbi->s_list);
3807 mutex_init(&sbi->umount_mutex);
3808 init_f2fs_rwsem(&sbi->io_order_lock);
3809 spin_lock_init(&sbi->cp_lock);
3810
3811 sbi->dirty_device = 0;
3812 spin_lock_init(&sbi->dev_lock);
3813
3814 init_f2fs_rwsem(&sbi->sb_lock);
3815 init_f2fs_rwsem(&sbi->pin_sem);
3816 }
3817
init_percpu_info(struct f2fs_sb_info * sbi)3818 static int init_percpu_info(struct f2fs_sb_info *sbi)
3819 {
3820 int err;
3821
3822 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3823 if (err)
3824 return err;
3825
3826 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3827 if (err)
3828 goto err_valid_block;
3829
3830 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3831 GFP_KERNEL);
3832 if (err)
3833 goto err_node_block;
3834 return 0;
3835
3836 err_node_block:
3837 percpu_counter_destroy(&sbi->rf_node_block_count);
3838 err_valid_block:
3839 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3840 return err;
3841 }
3842
3843 #ifdef CONFIG_BLK_DEV_ZONED
3844
3845 struct f2fs_report_zones_args {
3846 struct f2fs_sb_info *sbi;
3847 struct f2fs_dev_info *dev;
3848 };
3849
f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3850 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3851 void *data)
3852 {
3853 struct f2fs_report_zones_args *rz_args = data;
3854 block_t unusable_blocks = (zone->len - zone->capacity) >>
3855 F2FS_LOG_SECTORS_PER_BLOCK;
3856
3857 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3858 return 0;
3859
3860 set_bit(idx, rz_args->dev->blkz_seq);
3861 if (!rz_args->sbi->unusable_blocks_per_sec) {
3862 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3863 return 0;
3864 }
3865 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3866 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3867 return -EINVAL;
3868 }
3869 return 0;
3870 }
3871
init_blkz_info(struct f2fs_sb_info * sbi,int devi)3872 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3873 {
3874 struct block_device *bdev = FDEV(devi).bdev;
3875 sector_t nr_sectors = bdev_nr_sectors(bdev);
3876 struct f2fs_report_zones_args rep_zone_arg;
3877 u64 zone_sectors;
3878 unsigned int max_open_zones;
3879 int ret;
3880
3881 if (!f2fs_sb_has_blkzoned(sbi))
3882 return 0;
3883
3884 if (bdev_is_zoned(FDEV(devi).bdev)) {
3885 max_open_zones = bdev_max_open_zones(bdev);
3886 if (max_open_zones && (max_open_zones < sbi->max_open_zones))
3887 sbi->max_open_zones = max_open_zones;
3888 if (sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
3889 f2fs_err(sbi,
3890 "zoned: max open zones %u is too small, need at least %u open zones",
3891 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
3892 return -EINVAL;
3893 }
3894 }
3895
3896 zone_sectors = bdev_zone_sectors(bdev);
3897 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3898 SECTOR_TO_BLOCK(zone_sectors))
3899 return -EINVAL;
3900 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3901 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3902 sbi->blocks_per_blkz);
3903 if (!bdev_is_zone_start(bdev, nr_sectors))
3904 FDEV(devi).nr_blkz++;
3905
3906 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3907 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3908 * sizeof(unsigned long),
3909 GFP_KERNEL);
3910 if (!FDEV(devi).blkz_seq)
3911 return -ENOMEM;
3912
3913 rep_zone_arg.sbi = sbi;
3914 rep_zone_arg.dev = &FDEV(devi);
3915
3916 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3917 &rep_zone_arg);
3918 if (ret < 0)
3919 return ret;
3920 return 0;
3921 }
3922 #endif
3923
3924 /*
3925 * Read f2fs raw super block.
3926 * Because we have two copies of super block, so read both of them
3927 * to get the first valid one. If any one of them is broken, we pass
3928 * them recovery flag back to the caller.
3929 */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3930 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3931 struct f2fs_super_block **raw_super,
3932 int *valid_super_block, int *recovery)
3933 {
3934 struct super_block *sb = sbi->sb;
3935 int block;
3936 struct buffer_head *bh;
3937 struct f2fs_super_block *super;
3938 int err = 0;
3939
3940 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3941 if (!super)
3942 return -ENOMEM;
3943
3944 for (block = 0; block < 2; block++) {
3945 bh = sb_bread(sb, block);
3946 if (!bh) {
3947 f2fs_err(sbi, "Unable to read %dth superblock",
3948 block + 1);
3949 err = -EIO;
3950 *recovery = 1;
3951 continue;
3952 }
3953
3954 /* sanity checking of raw super */
3955 err = sanity_check_raw_super(sbi, bh);
3956 if (err) {
3957 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3958 block + 1);
3959 brelse(bh);
3960 *recovery = 1;
3961 continue;
3962 }
3963
3964 if (!*raw_super) {
3965 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3966 sizeof(*super));
3967 *valid_super_block = block;
3968 *raw_super = super;
3969 }
3970 brelse(bh);
3971 }
3972
3973 /* No valid superblock */
3974 if (!*raw_super)
3975 kfree(super);
3976 else
3977 err = 0;
3978
3979 return err;
3980 }
3981
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)3982 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3983 {
3984 struct buffer_head *bh;
3985 __u32 crc = 0;
3986 int err;
3987
3988 if ((recover && f2fs_readonly(sbi->sb)) ||
3989 f2fs_hw_is_readonly(sbi)) {
3990 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3991 return -EROFS;
3992 }
3993
3994 /* we should update superblock crc here */
3995 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3996 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3997 offsetof(struct f2fs_super_block, crc));
3998 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3999 }
4000
4001 /* write back-up superblock first */
4002 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
4003 if (!bh)
4004 return -EIO;
4005 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4006 brelse(bh);
4007
4008 /* if we are in recovery path, skip writing valid superblock */
4009 if (recover || err)
4010 return err;
4011
4012 /* write current valid superblock */
4013 bh = sb_bread(sbi->sb, sbi->valid_super_block);
4014 if (!bh)
4015 return -EIO;
4016 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4017 brelse(bh);
4018 return err;
4019 }
4020
save_stop_reason(struct f2fs_sb_info * sbi,unsigned char reason)4021 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4022 {
4023 unsigned long flags;
4024
4025 spin_lock_irqsave(&sbi->error_lock, flags);
4026 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4027 sbi->stop_reason[reason]++;
4028 spin_unlock_irqrestore(&sbi->error_lock, flags);
4029 }
4030
f2fs_record_stop_reason(struct f2fs_sb_info * sbi)4031 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4032 {
4033 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4034 unsigned long flags;
4035 int err;
4036
4037 f2fs_down_write(&sbi->sb_lock);
4038
4039 spin_lock_irqsave(&sbi->error_lock, flags);
4040 if (sbi->error_dirty) {
4041 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4042 MAX_F2FS_ERRORS);
4043 sbi->error_dirty = false;
4044 }
4045 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4046 spin_unlock_irqrestore(&sbi->error_lock, flags);
4047
4048 err = f2fs_commit_super(sbi, false);
4049
4050 f2fs_up_write(&sbi->sb_lock);
4051 if (err)
4052 f2fs_err_ratelimited(sbi,
4053 "f2fs_commit_super fails to record stop_reason, err:%d",
4054 err);
4055 }
4056
f2fs_save_errors(struct f2fs_sb_info * sbi,unsigned char flag)4057 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4058 {
4059 unsigned long flags;
4060
4061 spin_lock_irqsave(&sbi->error_lock, flags);
4062 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4063 set_bit(flag, (unsigned long *)sbi->errors);
4064 sbi->error_dirty = true;
4065 }
4066 spin_unlock_irqrestore(&sbi->error_lock, flags);
4067 }
4068
f2fs_update_errors(struct f2fs_sb_info * sbi)4069 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4070 {
4071 unsigned long flags;
4072 bool need_update = false;
4073
4074 spin_lock_irqsave(&sbi->error_lock, flags);
4075 if (sbi->error_dirty) {
4076 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4077 MAX_F2FS_ERRORS);
4078 sbi->error_dirty = false;
4079 need_update = true;
4080 }
4081 spin_unlock_irqrestore(&sbi->error_lock, flags);
4082
4083 return need_update;
4084 }
4085
f2fs_record_errors(struct f2fs_sb_info * sbi,unsigned char error)4086 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4087 {
4088 int err;
4089
4090 f2fs_down_write(&sbi->sb_lock);
4091
4092 if (!f2fs_update_errors(sbi))
4093 goto out_unlock;
4094
4095 err = f2fs_commit_super(sbi, false);
4096 if (err)
4097 f2fs_err_ratelimited(sbi,
4098 "f2fs_commit_super fails to record errors:%u, err:%d",
4099 error, err);
4100 out_unlock:
4101 f2fs_up_write(&sbi->sb_lock);
4102 }
4103
f2fs_handle_error(struct f2fs_sb_info * sbi,unsigned char error)4104 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4105 {
4106 f2fs_save_errors(sbi, error);
4107 f2fs_record_errors(sbi, error);
4108 }
4109
f2fs_handle_error_async(struct f2fs_sb_info * sbi,unsigned char error)4110 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4111 {
4112 f2fs_save_errors(sbi, error);
4113
4114 if (!sbi->error_dirty)
4115 return;
4116 if (!test_bit(error, (unsigned long *)sbi->errors))
4117 return;
4118 schedule_work(&sbi->s_error_work);
4119 }
4120
system_going_down(void)4121 static bool system_going_down(void)
4122 {
4123 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4124 || system_state == SYSTEM_RESTART;
4125 }
4126
f2fs_handle_critical_error(struct f2fs_sb_info * sbi,unsigned char reason,bool irq_context)4127 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason,
4128 bool irq_context)
4129 {
4130 struct super_block *sb = sbi->sb;
4131 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4132 bool continue_fs = !shutdown &&
4133 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4134
4135 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4136
4137 if (!f2fs_hw_is_readonly(sbi)) {
4138 save_stop_reason(sbi, reason);
4139
4140 if (irq_context && !shutdown)
4141 schedule_work(&sbi->s_error_work);
4142 else
4143 f2fs_record_stop_reason(sbi);
4144 }
4145
4146 /*
4147 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4148 * could panic during 'reboot -f' as the underlying device got already
4149 * disabled.
4150 */
4151 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4152 !shutdown && !system_going_down() &&
4153 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4154 panic("F2FS-fs (device %s): panic forced after error\n",
4155 sb->s_id);
4156
4157 if (shutdown)
4158 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4159
4160 /*
4161 * Continue filesystem operators if errors=continue. Should not set
4162 * RO by shutdown, since RO bypasses thaw_super which can hang the
4163 * system.
4164 */
4165 if (continue_fs || f2fs_readonly(sb) || shutdown) {
4166 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason);
4167 return;
4168 }
4169
4170 f2fs_warn(sbi, "Remounting filesystem read-only");
4171 /*
4172 * Make sure updated value of ->s_mount_flags will be visible before
4173 * ->s_flags update
4174 */
4175 smp_wmb();
4176 sb->s_flags |= SB_RDONLY;
4177 }
4178
f2fs_record_error_work(struct work_struct * work)4179 static void f2fs_record_error_work(struct work_struct *work)
4180 {
4181 struct f2fs_sb_info *sbi = container_of(work,
4182 struct f2fs_sb_info, s_error_work);
4183
4184 f2fs_record_stop_reason(sbi);
4185 }
4186
f2fs_scan_devices(struct f2fs_sb_info * sbi)4187 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4188 {
4189 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4190 unsigned int max_devices = MAX_DEVICES;
4191 unsigned int logical_blksize;
4192 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4193 int i;
4194
4195 /* Initialize single device information */
4196 if (!RDEV(0).path[0]) {
4197 if (!bdev_is_zoned(sbi->sb->s_bdev))
4198 return 0;
4199 max_devices = 1;
4200 }
4201
4202 /*
4203 * Initialize multiple devices information, or single
4204 * zoned block device information.
4205 */
4206 sbi->devs = f2fs_kzalloc(sbi,
4207 array_size(max_devices,
4208 sizeof(struct f2fs_dev_info)),
4209 GFP_KERNEL);
4210 if (!sbi->devs)
4211 return -ENOMEM;
4212
4213 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4214 sbi->aligned_blksize = true;
4215 #ifdef CONFIG_BLK_DEV_ZONED
4216 sbi->max_open_zones = UINT_MAX;
4217 #endif
4218
4219 for (i = 0; i < max_devices; i++) {
4220 if (i == 0)
4221 FDEV(0).bdev = sbi->sb->s_bdev;
4222 else if (!RDEV(i).path[0])
4223 break;
4224
4225 if (max_devices > 1) {
4226 /* Multi-device mount */
4227 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4228 FDEV(i).total_segments =
4229 le32_to_cpu(RDEV(i).total_segments);
4230 if (i == 0) {
4231 FDEV(i).start_blk = 0;
4232 FDEV(i).end_blk = FDEV(i).start_blk +
4233 SEGS_TO_BLKS(sbi,
4234 FDEV(i).total_segments) - 1 +
4235 le32_to_cpu(raw_super->segment0_blkaddr);
4236 } else {
4237 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4238 FDEV(i).end_blk = FDEV(i).start_blk +
4239 SEGS_TO_BLKS(sbi,
4240 FDEV(i).total_segments) - 1;
4241 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
4242 mode, sbi->sb, NULL);
4243 }
4244 }
4245 if (IS_ERR(FDEV(i).bdev))
4246 return PTR_ERR(FDEV(i).bdev);
4247
4248 /* to release errored devices */
4249 sbi->s_ndevs = i + 1;
4250
4251 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4252 sbi->aligned_blksize = false;
4253
4254 #ifdef CONFIG_BLK_DEV_ZONED
4255 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
4256 !f2fs_sb_has_blkzoned(sbi)) {
4257 f2fs_err(sbi, "Zoned block device feature not enabled");
4258 return -EINVAL;
4259 }
4260 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
4261 if (init_blkz_info(sbi, i)) {
4262 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4263 return -EINVAL;
4264 }
4265 if (max_devices == 1)
4266 break;
4267 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
4268 i, FDEV(i).path,
4269 FDEV(i).total_segments,
4270 FDEV(i).start_blk, FDEV(i).end_blk,
4271 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
4272 "Host-aware" : "Host-managed");
4273 continue;
4274 }
4275 #endif
4276 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4277 i, FDEV(i).path,
4278 FDEV(i).total_segments,
4279 FDEV(i).start_blk, FDEV(i).end_blk);
4280 }
4281 return 0;
4282 }
4283
f2fs_setup_casefold(struct f2fs_sb_info * sbi)4284 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4285 {
4286 #if IS_ENABLED(CONFIG_UNICODE)
4287 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4288 const struct f2fs_sb_encodings *encoding_info;
4289 struct unicode_map *encoding;
4290 __u16 encoding_flags;
4291
4292 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4293 if (!encoding_info) {
4294 f2fs_err(sbi,
4295 "Encoding requested by superblock is unknown");
4296 return -EINVAL;
4297 }
4298
4299 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4300 encoding = utf8_load(encoding_info->version);
4301 if (IS_ERR(encoding)) {
4302 f2fs_err(sbi,
4303 "can't mount with superblock charset: %s-%u.%u.%u "
4304 "not supported by the kernel. flags: 0x%x.",
4305 encoding_info->name,
4306 unicode_major(encoding_info->version),
4307 unicode_minor(encoding_info->version),
4308 unicode_rev(encoding_info->version),
4309 encoding_flags);
4310 return PTR_ERR(encoding);
4311 }
4312 f2fs_info(sbi, "Using encoding defined by superblock: "
4313 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4314 unicode_major(encoding_info->version),
4315 unicode_minor(encoding_info->version),
4316 unicode_rev(encoding_info->version),
4317 encoding_flags);
4318
4319 sbi->sb->s_encoding = encoding;
4320 sbi->sb->s_encoding_flags = encoding_flags;
4321 }
4322 #else
4323 if (f2fs_sb_has_casefold(sbi)) {
4324 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4325 return -EINVAL;
4326 }
4327 #endif
4328 return 0;
4329 }
4330
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)4331 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4332 {
4333 /* adjust parameters according to the volume size */
4334 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4335 if (f2fs_block_unit_discard(sbi))
4336 SM_I(sbi)->dcc_info->discard_granularity =
4337 MIN_DISCARD_GRANULARITY;
4338 if (!f2fs_lfs_mode(sbi))
4339 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4340 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4341 }
4342
4343 sbi->readdir_ra = true;
4344 }
4345
f2fs_fill_super(struct super_block * sb,void * data,int silent)4346 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4347 {
4348 struct f2fs_sb_info *sbi;
4349 struct f2fs_super_block *raw_super;
4350 struct inode *root;
4351 int err;
4352 bool skip_recovery = false, need_fsck = false;
4353 char *options = NULL;
4354 int recovery, i, valid_super_block;
4355 struct curseg_info *seg_i;
4356 int retry_cnt = 1;
4357 #ifdef CONFIG_QUOTA
4358 bool quota_enabled = false;
4359 #endif
4360
4361 try_onemore:
4362 err = -EINVAL;
4363 raw_super = NULL;
4364 valid_super_block = -1;
4365 recovery = 0;
4366
4367 /* allocate memory for f2fs-specific super block info */
4368 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4369 if (!sbi)
4370 return -ENOMEM;
4371
4372 sbi->sb = sb;
4373
4374 /* initialize locks within allocated memory */
4375 init_f2fs_rwsem(&sbi->gc_lock);
4376 mutex_init(&sbi->writepages);
4377 init_f2fs_rwsem(&sbi->cp_global_sem);
4378 init_f2fs_rwsem(&sbi->node_write);
4379 init_f2fs_rwsem(&sbi->node_change);
4380 spin_lock_init(&sbi->stat_lock);
4381 init_f2fs_rwsem(&sbi->cp_rwsem);
4382 init_f2fs_rwsem(&sbi->quota_sem);
4383 init_waitqueue_head(&sbi->cp_wait);
4384 spin_lock_init(&sbi->error_lock);
4385
4386 for (i = 0; i < NR_INODE_TYPE; i++) {
4387 INIT_LIST_HEAD(&sbi->inode_list[i]);
4388 spin_lock_init(&sbi->inode_lock[i]);
4389 }
4390 mutex_init(&sbi->flush_lock);
4391
4392 /* Load the checksum driver */
4393 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4394 if (IS_ERR(sbi->s_chksum_driver)) {
4395 f2fs_err(sbi, "Cannot load crc32 driver.");
4396 err = PTR_ERR(sbi->s_chksum_driver);
4397 sbi->s_chksum_driver = NULL;
4398 goto free_sbi;
4399 }
4400
4401 /* set a block size */
4402 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4403 f2fs_err(sbi, "unable to set blocksize");
4404 goto free_sbi;
4405 }
4406
4407 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4408 &recovery);
4409 if (err)
4410 goto free_sbi;
4411
4412 sb->s_fs_info = sbi;
4413 sbi->raw_super = raw_super;
4414
4415 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4416 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4417 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4418
4419 /* precompute checksum seed for metadata */
4420 if (f2fs_sb_has_inode_chksum(sbi))
4421 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4422 sizeof(raw_super->uuid));
4423
4424 default_options(sbi, false);
4425 /* parse mount options */
4426 options = kstrdup((const char *)data, GFP_KERNEL);
4427 if (data && !options) {
4428 err = -ENOMEM;
4429 goto free_sb_buf;
4430 }
4431
4432 err = parse_options(sb, options, false);
4433 if (err)
4434 goto free_options;
4435
4436 sb->s_maxbytes = max_file_blocks(NULL) <<
4437 le32_to_cpu(raw_super->log_blocksize);
4438 sb->s_max_links = F2FS_LINK_MAX;
4439
4440 err = f2fs_setup_casefold(sbi);
4441 if (err)
4442 goto free_options;
4443
4444 #ifdef CONFIG_QUOTA
4445 sb->dq_op = &f2fs_quota_operations;
4446 sb->s_qcop = &f2fs_quotactl_ops;
4447 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4448
4449 if (f2fs_sb_has_quota_ino(sbi)) {
4450 for (i = 0; i < MAXQUOTAS; i++) {
4451 if (f2fs_qf_ino(sbi->sb, i))
4452 sbi->nquota_files++;
4453 }
4454 }
4455 #endif
4456
4457 sb->s_op = &f2fs_sops;
4458 #ifdef CONFIG_FS_ENCRYPTION
4459 sb->s_cop = &f2fs_cryptops;
4460 #endif
4461 #ifdef CONFIG_FS_VERITY
4462 sb->s_vop = &f2fs_verityops;
4463 #endif
4464 sb->s_xattr = f2fs_xattr_handlers;
4465 sb->s_export_op = &f2fs_export_ops;
4466 sb->s_magic = F2FS_SUPER_MAGIC;
4467 sb->s_time_gran = 1;
4468 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4469 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4470 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4471 sb->s_iflags |= SB_I_CGROUPWB;
4472
4473 /* init f2fs-specific super block info */
4474 sbi->valid_super_block = valid_super_block;
4475
4476 /* disallow all the data/node/meta page writes */
4477 set_sbi_flag(sbi, SBI_POR_DOING);
4478
4479 err = f2fs_init_write_merge_io(sbi);
4480 if (err)
4481 goto free_bio_info;
4482
4483 init_sb_info(sbi);
4484
4485 err = f2fs_init_iostat(sbi);
4486 if (err)
4487 goto free_bio_info;
4488
4489 err = init_percpu_info(sbi);
4490 if (err)
4491 goto free_iostat;
4492
4493 /* init per sbi slab cache */
4494 err = f2fs_init_xattr_caches(sbi);
4495 if (err)
4496 goto free_percpu;
4497 err = f2fs_init_page_array_cache(sbi);
4498 if (err)
4499 goto free_xattr_cache;
4500
4501 /* get an inode for meta space */
4502 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4503 if (IS_ERR(sbi->meta_inode)) {
4504 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4505 err = PTR_ERR(sbi->meta_inode);
4506 goto free_page_array_cache;
4507 }
4508
4509 err = f2fs_get_valid_checkpoint(sbi);
4510 if (err) {
4511 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4512 goto free_meta_inode;
4513 }
4514
4515 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4516 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4517 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4518 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4519 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4520 }
4521
4522 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4523 set_sbi_flag(sbi, SBI_NEED_FSCK);
4524
4525 /* Initialize device list */
4526 err = f2fs_scan_devices(sbi);
4527 if (err) {
4528 f2fs_err(sbi, "Failed to find devices");
4529 goto free_devices;
4530 }
4531
4532 err = f2fs_init_post_read_wq(sbi);
4533 if (err) {
4534 f2fs_err(sbi, "Failed to initialize post read workqueue");
4535 goto free_devices;
4536 }
4537
4538 sbi->total_valid_node_count =
4539 le32_to_cpu(sbi->ckpt->valid_node_count);
4540 percpu_counter_set(&sbi->total_valid_inode_count,
4541 le32_to_cpu(sbi->ckpt->valid_inode_count));
4542 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4543 sbi->total_valid_block_count =
4544 le64_to_cpu(sbi->ckpt->valid_block_count);
4545 sbi->last_valid_block_count = sbi->total_valid_block_count;
4546 sbi->reserved_blocks = 0;
4547 sbi->current_reserved_blocks = 0;
4548 limit_reserve_root(sbi);
4549 adjust_unusable_cap_perc(sbi);
4550
4551 f2fs_init_extent_cache_info(sbi);
4552
4553 f2fs_init_ino_entry_info(sbi);
4554
4555 f2fs_init_fsync_node_info(sbi);
4556
4557 /* setup checkpoint request control and start checkpoint issue thread */
4558 f2fs_init_ckpt_req_control(sbi);
4559 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4560 test_opt(sbi, MERGE_CHECKPOINT)) {
4561 err = f2fs_start_ckpt_thread(sbi);
4562 if (err) {
4563 f2fs_err(sbi,
4564 "Failed to start F2FS issue_checkpoint_thread (%d)",
4565 err);
4566 goto stop_ckpt_thread;
4567 }
4568 }
4569
4570 /* setup f2fs internal modules */
4571 err = f2fs_build_segment_manager(sbi);
4572 if (err) {
4573 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4574 err);
4575 goto free_sm;
4576 }
4577 err = f2fs_build_node_manager(sbi);
4578 if (err) {
4579 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4580 err);
4581 goto free_nm;
4582 }
4583
4584 /* For write statistics */
4585 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4586
4587 /* Read accumulated write IO statistics if exists */
4588 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4589 if (__exist_node_summaries(sbi))
4590 sbi->kbytes_written =
4591 le64_to_cpu(seg_i->journal->info.kbytes_written);
4592
4593 f2fs_build_gc_manager(sbi);
4594
4595 err = f2fs_build_stats(sbi);
4596 if (err)
4597 goto free_nm;
4598
4599 /* get an inode for node space */
4600 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4601 if (IS_ERR(sbi->node_inode)) {
4602 f2fs_err(sbi, "Failed to read node inode");
4603 err = PTR_ERR(sbi->node_inode);
4604 goto free_stats;
4605 }
4606
4607 /* read root inode and dentry */
4608 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4609 if (IS_ERR(root)) {
4610 f2fs_err(sbi, "Failed to read root inode");
4611 err = PTR_ERR(root);
4612 goto free_node_inode;
4613 }
4614 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4615 !root->i_size || !root->i_nlink) {
4616 iput(root);
4617 err = -EINVAL;
4618 goto free_node_inode;
4619 }
4620
4621 sb->s_root = d_make_root(root); /* allocate root dentry */
4622 if (!sb->s_root) {
4623 err = -ENOMEM;
4624 goto free_node_inode;
4625 }
4626
4627 err = f2fs_init_compress_inode(sbi);
4628 if (err)
4629 goto free_root_inode;
4630
4631 err = f2fs_register_sysfs(sbi);
4632 if (err)
4633 goto free_compress_inode;
4634
4635 #ifdef CONFIG_QUOTA
4636 /* Enable quota usage during mount */
4637 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4638 err = f2fs_enable_quotas(sb);
4639 if (err)
4640 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4641 }
4642
4643 quota_enabled = f2fs_recover_quota_begin(sbi);
4644 #endif
4645 /* if there are any orphan inodes, free them */
4646 err = f2fs_recover_orphan_inodes(sbi);
4647 if (err)
4648 goto free_meta;
4649
4650 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4651 goto reset_checkpoint;
4652
4653 /* recover fsynced data */
4654 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4655 !test_opt(sbi, NORECOVERY)) {
4656 /*
4657 * mount should be failed, when device has readonly mode, and
4658 * previous checkpoint was not done by clean system shutdown.
4659 */
4660 if (f2fs_hw_is_readonly(sbi)) {
4661 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4662 err = f2fs_recover_fsync_data(sbi, true);
4663 if (err > 0) {
4664 err = -EROFS;
4665 f2fs_err(sbi, "Need to recover fsync data, but "
4666 "write access unavailable, please try "
4667 "mount w/ disable_roll_forward or norecovery");
4668 }
4669 if (err < 0)
4670 goto free_meta;
4671 }
4672 f2fs_info(sbi, "write access unavailable, skipping recovery");
4673 goto reset_checkpoint;
4674 }
4675
4676 if (need_fsck)
4677 set_sbi_flag(sbi, SBI_NEED_FSCK);
4678
4679 if (skip_recovery)
4680 goto reset_checkpoint;
4681
4682 err = f2fs_recover_fsync_data(sbi, false);
4683 if (err < 0) {
4684 if (err != -ENOMEM)
4685 skip_recovery = true;
4686 need_fsck = true;
4687 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4688 err);
4689 goto free_meta;
4690 }
4691 } else {
4692 err = f2fs_recover_fsync_data(sbi, true);
4693
4694 if (!f2fs_readonly(sb) && err > 0) {
4695 err = -EINVAL;
4696 f2fs_err(sbi, "Need to recover fsync data");
4697 goto free_meta;
4698 }
4699 }
4700
4701 #ifdef CONFIG_QUOTA
4702 f2fs_recover_quota_end(sbi, quota_enabled);
4703 #endif
4704 reset_checkpoint:
4705 /*
4706 * If the f2fs is not readonly and fsync data recovery succeeds,
4707 * check zoned block devices' write pointer consistency.
4708 */
4709 if (f2fs_sb_has_blkzoned(sbi) && !f2fs_readonly(sb)) {
4710 int err2;
4711
4712 f2fs_notice(sbi, "Checking entire write pointers");
4713 err2 = f2fs_check_write_pointer(sbi);
4714 if (err2)
4715 err = err2;
4716 }
4717 if (err)
4718 goto free_meta;
4719
4720 err = f2fs_init_inmem_curseg(sbi);
4721 if (err)
4722 goto sync_free_meta;
4723
4724 /* f2fs_recover_fsync_data() cleared this already */
4725 clear_sbi_flag(sbi, SBI_POR_DOING);
4726
4727 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4728 err = f2fs_disable_checkpoint(sbi);
4729 if (err)
4730 goto sync_free_meta;
4731 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4732 f2fs_enable_checkpoint(sbi);
4733 }
4734
4735 /*
4736 * If filesystem is not mounted as read-only then
4737 * do start the gc_thread.
4738 */
4739 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4740 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4741 /* After POR, we can run background GC thread.*/
4742 err = f2fs_start_gc_thread(sbi);
4743 if (err)
4744 goto sync_free_meta;
4745 }
4746 kvfree(options);
4747
4748 /* recover broken superblock */
4749 if (recovery) {
4750 err = f2fs_commit_super(sbi, true);
4751 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4752 sbi->valid_super_block ? 1 : 2, err);
4753 }
4754
4755 f2fs_join_shrinker(sbi);
4756
4757 f2fs_tuning_parameters(sbi);
4758
4759 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4760 cur_cp_version(F2FS_CKPT(sbi)));
4761 f2fs_update_time(sbi, CP_TIME);
4762 f2fs_update_time(sbi, REQ_TIME);
4763 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4764
4765 cleancache_init_fs(sb);
4766 return 0;
4767
4768 sync_free_meta:
4769 /* safe to flush all the data */
4770 sync_filesystem(sbi->sb);
4771 retry_cnt = 0;
4772
4773 free_meta:
4774 #ifdef CONFIG_QUOTA
4775 f2fs_truncate_quota_inode_pages(sb);
4776 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4777 f2fs_quota_off_umount(sbi->sb);
4778 #endif
4779 /*
4780 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4781 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4782 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4783 * falls into an infinite loop in f2fs_sync_meta_pages().
4784 */
4785 truncate_inode_pages_final(META_MAPPING(sbi));
4786 /* evict some inodes being cached by GC */
4787 evict_inodes(sb);
4788 f2fs_unregister_sysfs(sbi);
4789 free_compress_inode:
4790 f2fs_destroy_compress_inode(sbi);
4791 free_root_inode:
4792 dput(sb->s_root);
4793 sb->s_root = NULL;
4794 free_node_inode:
4795 f2fs_release_ino_entry(sbi, true);
4796 truncate_inode_pages_final(NODE_MAPPING(sbi));
4797 iput(sbi->node_inode);
4798 sbi->node_inode = NULL;
4799 free_stats:
4800 f2fs_destroy_stats(sbi);
4801 free_nm:
4802 /* stop discard thread before destroying node manager */
4803 f2fs_stop_discard_thread(sbi);
4804 f2fs_destroy_node_manager(sbi);
4805 free_sm:
4806 f2fs_destroy_segment_manager(sbi);
4807 stop_ckpt_thread:
4808 f2fs_stop_ckpt_thread(sbi);
4809 /* flush s_error_work before sbi destroy */
4810 flush_work(&sbi->s_error_work);
4811 f2fs_destroy_post_read_wq(sbi);
4812 free_devices:
4813 destroy_device_list(sbi);
4814 kvfree(sbi->ckpt);
4815 free_meta_inode:
4816 make_bad_inode(sbi->meta_inode);
4817 iput(sbi->meta_inode);
4818 sbi->meta_inode = NULL;
4819 free_page_array_cache:
4820 f2fs_destroy_page_array_cache(sbi);
4821 free_xattr_cache:
4822 f2fs_destroy_xattr_caches(sbi);
4823 free_percpu:
4824 destroy_percpu_info(sbi);
4825 free_iostat:
4826 f2fs_destroy_iostat(sbi);
4827 free_bio_info:
4828 for (i = 0; i < NR_PAGE_TYPE; i++)
4829 kvfree(sbi->write_io[i]);
4830
4831 #if IS_ENABLED(CONFIG_UNICODE)
4832 utf8_unload(sb->s_encoding);
4833 sb->s_encoding = NULL;
4834 #endif
4835 free_options:
4836 #ifdef CONFIG_QUOTA
4837 for (i = 0; i < MAXQUOTAS; i++)
4838 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4839 #endif
4840 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4841 kvfree(options);
4842 free_sb_buf:
4843 kfree(raw_super);
4844 free_sbi:
4845 if (sbi->s_chksum_driver)
4846 crypto_free_shash(sbi->s_chksum_driver);
4847 kfree(sbi);
4848 sb->s_fs_info = NULL;
4849
4850 /* give only one another chance */
4851 if (retry_cnt > 0 && skip_recovery) {
4852 retry_cnt--;
4853 shrink_dcache_sb(sb);
4854 goto try_onemore;
4855 }
4856 return err;
4857 }
4858
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4859 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4860 const char *dev_name, void *data)
4861 {
4862 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4863 }
4864
kill_f2fs_super(struct super_block * sb)4865 static void kill_f2fs_super(struct super_block *sb)
4866 {
4867 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4868
4869 if (sb->s_root) {
4870 set_sbi_flag(sbi, SBI_IS_CLOSE);
4871 f2fs_stop_gc_thread(sbi);
4872 f2fs_stop_discard_thread(sbi);
4873
4874 #ifdef CONFIG_F2FS_FS_COMPRESSION
4875 /*
4876 * latter evict_inode() can bypass checking and invalidating
4877 * compress inode cache.
4878 */
4879 if (test_opt(sbi, COMPRESS_CACHE))
4880 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4881 #endif
4882
4883 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4884 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4885 struct cp_control cpc = {
4886 .reason = CP_UMOUNT,
4887 };
4888 stat_inc_cp_call_count(sbi, TOTAL_CALL);
4889 f2fs_write_checkpoint(sbi, &cpc);
4890 }
4891
4892 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4893 sb->s_flags &= ~SB_RDONLY;
4894 }
4895 kill_block_super(sb);
4896 /* Release block devices last, after fscrypt_destroy_keyring(). */
4897 if (sbi) {
4898 destroy_device_list(sbi);
4899 kfree(sbi);
4900 sb->s_fs_info = NULL;
4901 }
4902 }
4903
4904 static struct file_system_type f2fs_fs_type = {
4905 .owner = THIS_MODULE,
4906 .name = "f2fs",
4907 .mount = f2fs_mount,
4908 .kill_sb = kill_f2fs_super,
4909 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4910 };
4911 MODULE_ALIAS_FS("f2fs");
4912
init_inodecache(void)4913 static int __init init_inodecache(void)
4914 {
4915 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4916 sizeof(struct f2fs_inode_info), 0,
4917 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4918 return f2fs_inode_cachep ? 0 : -ENOMEM;
4919 }
4920
destroy_inodecache(void)4921 static void destroy_inodecache(void)
4922 {
4923 /*
4924 * Make sure all delayed rcu free inodes are flushed before we
4925 * destroy cache.
4926 */
4927 rcu_barrier();
4928 kmem_cache_destroy(f2fs_inode_cachep);
4929 }
4930
init_f2fs_fs(void)4931 static int __init init_f2fs_fs(void)
4932 {
4933 int err;
4934
4935 err = init_inodecache();
4936 if (err)
4937 goto fail;
4938 err = f2fs_create_node_manager_caches();
4939 if (err)
4940 goto free_inodecache;
4941 err = f2fs_create_segment_manager_caches();
4942 if (err)
4943 goto free_node_manager_caches;
4944 err = f2fs_create_checkpoint_caches();
4945 if (err)
4946 goto free_segment_manager_caches;
4947 err = f2fs_create_recovery_cache();
4948 if (err)
4949 goto free_checkpoint_caches;
4950 err = f2fs_create_extent_cache();
4951 if (err)
4952 goto free_recovery_cache;
4953 err = f2fs_create_garbage_collection_cache();
4954 if (err)
4955 goto free_extent_cache;
4956 err = f2fs_init_sysfs();
4957 if (err)
4958 goto free_garbage_collection_cache;
4959 err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4960 if (err)
4961 goto free_sysfs;
4962 err = register_filesystem(&f2fs_fs_type);
4963 if (err)
4964 goto free_shrinker;
4965 f2fs_create_root_stats();
4966 err = f2fs_init_post_read_processing();
4967 if (err)
4968 goto free_root_stats;
4969 err = f2fs_init_iostat_processing();
4970 if (err)
4971 goto free_post_read;
4972 err = f2fs_init_bio_entry_cache();
4973 if (err)
4974 goto free_iostat;
4975 err = f2fs_init_bioset();
4976 if (err)
4977 goto free_bio_entry_cache;
4978 err = f2fs_init_compress_mempool();
4979 if (err)
4980 goto free_bioset;
4981 err = f2fs_init_compress_cache();
4982 if (err)
4983 goto free_compress_mempool;
4984 err = f2fs_create_casefold_cache();
4985 if (err)
4986 goto free_compress_cache;
4987 return 0;
4988 free_compress_cache:
4989 f2fs_destroy_compress_cache();
4990 free_compress_mempool:
4991 f2fs_destroy_compress_mempool();
4992 free_bioset:
4993 f2fs_destroy_bioset();
4994 free_bio_entry_cache:
4995 f2fs_destroy_bio_entry_cache();
4996 free_iostat:
4997 f2fs_destroy_iostat_processing();
4998 free_post_read:
4999 f2fs_destroy_post_read_processing();
5000 free_root_stats:
5001 f2fs_destroy_root_stats();
5002 unregister_filesystem(&f2fs_fs_type);
5003 free_shrinker:
5004 unregister_shrinker(&f2fs_shrinker_info);
5005 free_sysfs:
5006 f2fs_exit_sysfs();
5007 free_garbage_collection_cache:
5008 f2fs_destroy_garbage_collection_cache();
5009 free_extent_cache:
5010 f2fs_destroy_extent_cache();
5011 free_recovery_cache:
5012 f2fs_destroy_recovery_cache();
5013 free_checkpoint_caches:
5014 f2fs_destroy_checkpoint_caches();
5015 free_segment_manager_caches:
5016 f2fs_destroy_segment_manager_caches();
5017 free_node_manager_caches:
5018 f2fs_destroy_node_manager_caches();
5019 free_inodecache:
5020 destroy_inodecache();
5021 fail:
5022 return err;
5023 }
5024
exit_f2fs_fs(void)5025 static void __exit exit_f2fs_fs(void)
5026 {
5027 f2fs_destroy_casefold_cache();
5028 f2fs_destroy_compress_cache();
5029 f2fs_destroy_compress_mempool();
5030 f2fs_destroy_bioset();
5031 f2fs_destroy_bio_entry_cache();
5032 f2fs_destroy_iostat_processing();
5033 f2fs_destroy_post_read_processing();
5034 f2fs_destroy_root_stats();
5035 unregister_filesystem(&f2fs_fs_type);
5036 unregister_shrinker(&f2fs_shrinker_info);
5037 f2fs_exit_sysfs();
5038 f2fs_destroy_garbage_collection_cache();
5039 f2fs_destroy_extent_cache();
5040 f2fs_destroy_recovery_cache();
5041 f2fs_destroy_checkpoint_caches();
5042 f2fs_destroy_segment_manager_caches();
5043 f2fs_destroy_node_manager_caches();
5044 destroy_inodecache();
5045 }
5046
5047 module_init(init_f2fs_fs)
5048 module_exit(exit_f2fs_fs)
5049
5050 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5051 MODULE_DESCRIPTION("Flash Friendly File System");
5052 MODULE_LICENSE("GPL");
5053 MODULE_IMPORT_NS(ANDROID_GKI_VFS_EXPORT_ONLY);
5054 MODULE_SOFTDEP("pre: crc32");
5055
5056