1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * fs/f2fs/super.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8 #include <linux/module.h>
9 #include <linux/init.h>
10 #include <linux/fs.h>
11 #include <linux/fs_context.h>
12 #include <linux/sched/mm.h>
13 #include <linux/statfs.h>
14 #include <linux/buffer_head.h>
15 #include <linux/kthread.h>
16 #include <linux/parser.h>
17 #include <linux/mount.h>
18 #include <linux/seq_file.h>
19 #include <linux/proc_fs.h>
20 #include <linux/random.h>
21 #include <linux/exportfs.h>
22 #include <linux/blkdev.h>
23 #include <linux/quotaops.h>
24 #include <linux/f2fs_fs.h>
25 #include <linux/sysfs.h>
26 #include <linux/quota.h>
27 #include <linux/unicode.h>
28 #include <linux/part_stat.h>
29 #include <linux/zstd.h>
30 #include <linux/lz4.h>
31
32 #include "f2fs.h"
33 #include "node.h"
34 #include "segment.h"
35 #include "xattr.h"
36 #include "gc.h"
37 #include "iostat.h"
38
39 #define CREATE_TRACE_POINTS
40 #include <trace/events/f2fs.h>
41
42 static struct kmem_cache *f2fs_inode_cachep;
43
44 #ifdef CONFIG_F2FS_FAULT_INJECTION
45
46 const char *f2fs_fault_name[FAULT_MAX] = {
47 [FAULT_KMALLOC] = "kmalloc",
48 [FAULT_KVMALLOC] = "kvmalloc",
49 [FAULT_PAGE_ALLOC] = "page alloc",
50 [FAULT_PAGE_GET] = "page get",
51 [FAULT_ALLOC_NID] = "alloc nid",
52 [FAULT_ORPHAN] = "orphan",
53 [FAULT_BLOCK] = "no more block",
54 [FAULT_DIR_DEPTH] = "too big dir depth",
55 [FAULT_EVICT_INODE] = "evict_inode fail",
56 [FAULT_TRUNCATE] = "truncate fail",
57 [FAULT_READ_IO] = "read IO error",
58 [FAULT_CHECKPOINT] = "checkpoint error",
59 [FAULT_DISCARD] = "discard error",
60 [FAULT_WRITE_IO] = "write IO error",
61 [FAULT_SLAB_ALLOC] = "slab alloc",
62 [FAULT_DQUOT_INIT] = "dquot initialize",
63 [FAULT_LOCK_OP] = "lock_op",
64 [FAULT_BLKADDR_VALIDITY] = "invalid blkaddr",
65 [FAULT_BLKADDR_CONSISTENCE] = "inconsistent blkaddr",
66 [FAULT_NO_SEGMENT] = "no free segment",
67 };
68
f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned long rate,unsigned long type)69 int f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned long rate,
70 unsigned long type)
71 {
72 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
73
74 if (rate) {
75 if (rate > INT_MAX)
76 return -EINVAL;
77 atomic_set(&ffi->inject_ops, 0);
78 ffi->inject_rate = (int)rate;
79 }
80
81 if (type) {
82 if (type >= BIT(FAULT_MAX))
83 return -EINVAL;
84 ffi->inject_type = (unsigned int)type;
85 }
86
87 if (!rate && !type)
88 memset(ffi, 0, sizeof(struct f2fs_fault_info));
89 else
90 f2fs_info(sbi,
91 "build fault injection attr: rate: %lu, type: 0x%lx",
92 rate, type);
93 return 0;
94 }
95 #endif
96
97 /* f2fs-wide shrinker description */
98 static struct shrinker *f2fs_shrinker_info;
99
f2fs_init_shrinker(void)100 static int __init f2fs_init_shrinker(void)
101 {
102 f2fs_shrinker_info = shrinker_alloc(0, "f2fs-shrinker");
103 if (!f2fs_shrinker_info)
104 return -ENOMEM;
105
106 f2fs_shrinker_info->count_objects = f2fs_shrink_count;
107 f2fs_shrinker_info->scan_objects = f2fs_shrink_scan;
108
109 shrinker_register(f2fs_shrinker_info);
110
111 return 0;
112 }
113
f2fs_exit_shrinker(void)114 static void f2fs_exit_shrinker(void)
115 {
116 shrinker_free(f2fs_shrinker_info);
117 }
118
119 enum {
120 Opt_gc_background,
121 Opt_disable_roll_forward,
122 Opt_norecovery,
123 Opt_discard,
124 Opt_nodiscard,
125 Opt_noheap,
126 Opt_heap,
127 Opt_user_xattr,
128 Opt_nouser_xattr,
129 Opt_acl,
130 Opt_noacl,
131 Opt_active_logs,
132 Opt_disable_ext_identify,
133 Opt_inline_xattr,
134 Opt_noinline_xattr,
135 Opt_inline_xattr_size,
136 Opt_inline_data,
137 Opt_inline_dentry,
138 Opt_noinline_dentry,
139 Opt_flush_merge,
140 Opt_noflush_merge,
141 Opt_barrier,
142 Opt_nobarrier,
143 Opt_fastboot,
144 Opt_extent_cache,
145 Opt_noextent_cache,
146 Opt_noinline_data,
147 Opt_data_flush,
148 Opt_reserve_root,
149 Opt_resgid,
150 Opt_resuid,
151 Opt_mode,
152 Opt_fault_injection,
153 Opt_fault_type,
154 Opt_quota,
155 Opt_noquota,
156 Opt_usrquota,
157 Opt_grpquota,
158 Opt_prjquota,
159 Opt_usrjquota,
160 Opt_grpjquota,
161 Opt_prjjquota,
162 Opt_offusrjquota,
163 Opt_offgrpjquota,
164 Opt_offprjjquota,
165 Opt_jqfmt_vfsold,
166 Opt_jqfmt_vfsv0,
167 Opt_jqfmt_vfsv1,
168 Opt_alloc,
169 Opt_fsync,
170 Opt_test_dummy_encryption,
171 Opt_inlinecrypt,
172 Opt_checkpoint_disable,
173 Opt_checkpoint_disable_cap,
174 Opt_checkpoint_disable_cap_perc,
175 Opt_checkpoint_enable,
176 Opt_checkpoint_merge,
177 Opt_nocheckpoint_merge,
178 Opt_compress_algorithm,
179 Opt_compress_log_size,
180 Opt_compress_extension,
181 Opt_nocompress_extension,
182 Opt_compress_chksum,
183 Opt_compress_mode,
184 Opt_compress_cache,
185 Opt_atgc,
186 Opt_gc_merge,
187 Opt_nogc_merge,
188 Opt_discard_unit,
189 Opt_memory_mode,
190 Opt_age_extent_cache,
191 Opt_errors,
192 Opt_err,
193 };
194
195 static match_table_t f2fs_tokens = {
196 {Opt_gc_background, "background_gc=%s"},
197 {Opt_disable_roll_forward, "disable_roll_forward"},
198 {Opt_norecovery, "norecovery"},
199 {Opt_discard, "discard"},
200 {Opt_nodiscard, "nodiscard"},
201 {Opt_noheap, "no_heap"},
202 {Opt_heap, "heap"},
203 {Opt_user_xattr, "user_xattr"},
204 {Opt_nouser_xattr, "nouser_xattr"},
205 {Opt_acl, "acl"},
206 {Opt_noacl, "noacl"},
207 {Opt_active_logs, "active_logs=%u"},
208 {Opt_disable_ext_identify, "disable_ext_identify"},
209 {Opt_inline_xattr, "inline_xattr"},
210 {Opt_noinline_xattr, "noinline_xattr"},
211 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
212 {Opt_inline_data, "inline_data"},
213 {Opt_inline_dentry, "inline_dentry"},
214 {Opt_noinline_dentry, "noinline_dentry"},
215 {Opt_flush_merge, "flush_merge"},
216 {Opt_noflush_merge, "noflush_merge"},
217 {Opt_barrier, "barrier"},
218 {Opt_nobarrier, "nobarrier"},
219 {Opt_fastboot, "fastboot"},
220 {Opt_extent_cache, "extent_cache"},
221 {Opt_noextent_cache, "noextent_cache"},
222 {Opt_noinline_data, "noinline_data"},
223 {Opt_data_flush, "data_flush"},
224 {Opt_reserve_root, "reserve_root=%u"},
225 {Opt_resgid, "resgid=%u"},
226 {Opt_resuid, "resuid=%u"},
227 {Opt_mode, "mode=%s"},
228 {Opt_fault_injection, "fault_injection=%u"},
229 {Opt_fault_type, "fault_type=%u"},
230 {Opt_quota, "quota"},
231 {Opt_noquota, "noquota"},
232 {Opt_usrquota, "usrquota"},
233 {Opt_grpquota, "grpquota"},
234 {Opt_prjquota, "prjquota"},
235 {Opt_usrjquota, "usrjquota=%s"},
236 {Opt_grpjquota, "grpjquota=%s"},
237 {Opt_prjjquota, "prjjquota=%s"},
238 {Opt_offusrjquota, "usrjquota="},
239 {Opt_offgrpjquota, "grpjquota="},
240 {Opt_offprjjquota, "prjjquota="},
241 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
242 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
243 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
244 {Opt_alloc, "alloc_mode=%s"},
245 {Opt_fsync, "fsync_mode=%s"},
246 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
247 {Opt_test_dummy_encryption, "test_dummy_encryption"},
248 {Opt_inlinecrypt, "inlinecrypt"},
249 {Opt_checkpoint_disable, "checkpoint=disable"},
250 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
251 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
252 {Opt_checkpoint_enable, "checkpoint=enable"},
253 {Opt_checkpoint_merge, "checkpoint_merge"},
254 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
255 {Opt_compress_algorithm, "compress_algorithm=%s"},
256 {Opt_compress_log_size, "compress_log_size=%u"},
257 {Opt_compress_extension, "compress_extension=%s"},
258 {Opt_nocompress_extension, "nocompress_extension=%s"},
259 {Opt_compress_chksum, "compress_chksum"},
260 {Opt_compress_mode, "compress_mode=%s"},
261 {Opt_compress_cache, "compress_cache"},
262 {Opt_atgc, "atgc"},
263 {Opt_gc_merge, "gc_merge"},
264 {Opt_nogc_merge, "nogc_merge"},
265 {Opt_discard_unit, "discard_unit=%s"},
266 {Opt_memory_mode, "memory=%s"},
267 {Opt_age_extent_cache, "age_extent_cache"},
268 {Opt_errors, "errors=%s"},
269 {Opt_err, NULL},
270 };
271
f2fs_printk(struct f2fs_sb_info * sbi,bool limit_rate,const char * fmt,...)272 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate,
273 const char *fmt, ...)
274 {
275 struct va_format vaf;
276 va_list args;
277 int level;
278
279 va_start(args, fmt);
280
281 level = printk_get_level(fmt);
282 vaf.fmt = printk_skip_level(fmt);
283 vaf.va = &args;
284 if (limit_rate)
285 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n",
286 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
287 else
288 printk("%c%cF2FS-fs (%s): %pV\n",
289 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
290
291 va_end(args);
292 }
293
294 #if IS_ENABLED(CONFIG_UNICODE)
295 static const struct f2fs_sb_encodings {
296 __u16 magic;
297 char *name;
298 unsigned int version;
299 } f2fs_sb_encoding_map[] = {
300 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
301 };
302
303 static const struct f2fs_sb_encodings *
f2fs_sb_read_encoding(const struct f2fs_super_block * sb)304 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
305 {
306 __u16 magic = le16_to_cpu(sb->s_encoding);
307 int i;
308
309 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
310 if (magic == f2fs_sb_encoding_map[i].magic)
311 return &f2fs_sb_encoding_map[i];
312
313 return NULL;
314 }
315
316 struct kmem_cache *f2fs_cf_name_slab;
f2fs_create_casefold_cache(void)317 static int __init f2fs_create_casefold_cache(void)
318 {
319 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
320 F2FS_NAME_LEN);
321 return f2fs_cf_name_slab ? 0 : -ENOMEM;
322 }
323
f2fs_destroy_casefold_cache(void)324 static void f2fs_destroy_casefold_cache(void)
325 {
326 kmem_cache_destroy(f2fs_cf_name_slab);
327 }
328 #else
f2fs_create_casefold_cache(void)329 static int __init f2fs_create_casefold_cache(void) { return 0; }
f2fs_destroy_casefold_cache(void)330 static void f2fs_destroy_casefold_cache(void) { }
331 #endif
332
limit_reserve_root(struct f2fs_sb_info * sbi)333 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
334 {
335 block_t limit = min((sbi->user_block_count >> 3),
336 sbi->user_block_count - sbi->reserved_blocks);
337
338 /* limit is 12.5% */
339 if (test_opt(sbi, RESERVE_ROOT) &&
340 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
341 F2FS_OPTION(sbi).root_reserved_blocks = limit;
342 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
343 F2FS_OPTION(sbi).root_reserved_blocks);
344 }
345 if (!test_opt(sbi, RESERVE_ROOT) &&
346 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
347 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
348 !gid_eq(F2FS_OPTION(sbi).s_resgid,
349 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
350 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
351 from_kuid_munged(&init_user_ns,
352 F2FS_OPTION(sbi).s_resuid),
353 from_kgid_munged(&init_user_ns,
354 F2FS_OPTION(sbi).s_resgid));
355 }
356
adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)357 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
358 {
359 if (!F2FS_OPTION(sbi).unusable_cap_perc)
360 return;
361
362 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
363 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
364 else
365 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
366 F2FS_OPTION(sbi).unusable_cap_perc;
367
368 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
369 F2FS_OPTION(sbi).unusable_cap,
370 F2FS_OPTION(sbi).unusable_cap_perc);
371 }
372
init_once(void * foo)373 static void init_once(void *foo)
374 {
375 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
376
377 inode_init_once(&fi->vfs_inode);
378 }
379
380 #ifdef CONFIG_QUOTA
381 static const char * const quotatypes[] = INITQFNAMES;
382 #define QTYPE2NAME(t) (quotatypes[t])
f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)383 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
384 substring_t *args)
385 {
386 struct f2fs_sb_info *sbi = F2FS_SB(sb);
387 char *qname;
388 int ret = -EINVAL;
389
390 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
391 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
392 return -EINVAL;
393 }
394 if (f2fs_sb_has_quota_ino(sbi)) {
395 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
396 return 0;
397 }
398
399 qname = match_strdup(args);
400 if (!qname) {
401 f2fs_err(sbi, "Not enough memory for storing quotafile name");
402 return -ENOMEM;
403 }
404 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
405 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
406 ret = 0;
407 else
408 f2fs_err(sbi, "%s quota file already specified",
409 QTYPE2NAME(qtype));
410 goto errout;
411 }
412 if (strchr(qname, '/')) {
413 f2fs_err(sbi, "quotafile must be on filesystem root");
414 goto errout;
415 }
416 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
417 set_opt(sbi, QUOTA);
418 return 0;
419 errout:
420 kfree(qname);
421 return ret;
422 }
423
f2fs_clear_qf_name(struct super_block * sb,int qtype)424 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
425 {
426 struct f2fs_sb_info *sbi = F2FS_SB(sb);
427
428 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
429 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
430 return -EINVAL;
431 }
432 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
433 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
434 return 0;
435 }
436
f2fs_check_quota_options(struct f2fs_sb_info * sbi)437 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
438 {
439 /*
440 * We do the test below only for project quotas. 'usrquota' and
441 * 'grpquota' mount options are allowed even without quota feature
442 * to support legacy quotas in quota files.
443 */
444 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
445 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
446 return -1;
447 }
448 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
449 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
450 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
451 if (test_opt(sbi, USRQUOTA) &&
452 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
453 clear_opt(sbi, USRQUOTA);
454
455 if (test_opt(sbi, GRPQUOTA) &&
456 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
457 clear_opt(sbi, GRPQUOTA);
458
459 if (test_opt(sbi, PRJQUOTA) &&
460 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
461 clear_opt(sbi, PRJQUOTA);
462
463 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
464 test_opt(sbi, PRJQUOTA)) {
465 f2fs_err(sbi, "old and new quota format mixing");
466 return -1;
467 }
468
469 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
470 f2fs_err(sbi, "journaled quota format not specified");
471 return -1;
472 }
473 }
474
475 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
476 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
477 F2FS_OPTION(sbi).s_jquota_fmt = 0;
478 }
479 return 0;
480 }
481 #endif
482
f2fs_set_test_dummy_encryption(struct super_block * sb,const char * opt,const substring_t * arg,bool is_remount)483 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
484 const char *opt,
485 const substring_t *arg,
486 bool is_remount)
487 {
488 struct f2fs_sb_info *sbi = F2FS_SB(sb);
489 struct fs_parameter param = {
490 .type = fs_value_is_string,
491 .string = arg->from ? arg->from : "",
492 };
493 struct fscrypt_dummy_policy *policy =
494 &F2FS_OPTION(sbi).dummy_enc_policy;
495 int err;
496
497 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
498 f2fs_warn(sbi, "test_dummy_encryption option not supported");
499 return -EINVAL;
500 }
501
502 if (!f2fs_sb_has_encrypt(sbi)) {
503 f2fs_err(sbi, "Encrypt feature is off");
504 return -EINVAL;
505 }
506
507 /*
508 * This mount option is just for testing, and it's not worthwhile to
509 * implement the extra complexity (e.g. RCU protection) that would be
510 * needed to allow it to be set or changed during remount. We do allow
511 * it to be specified during remount, but only if there is no change.
512 */
513 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
514 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
515 return -EINVAL;
516 }
517
518 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
519 if (err) {
520 if (err == -EEXIST)
521 f2fs_warn(sbi,
522 "Can't change test_dummy_encryption on remount");
523 else if (err == -EINVAL)
524 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
525 opt);
526 else
527 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
528 opt, err);
529 return -EINVAL;
530 }
531 f2fs_warn(sbi, "Test dummy encryption mode enabled");
532 return 0;
533 }
534
535 #ifdef CONFIG_F2FS_FS_COMPRESSION
is_compress_extension_exist(struct f2fs_sb_info * sbi,const char * new_ext,bool is_ext)536 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
537 const char *new_ext, bool is_ext)
538 {
539 unsigned char (*ext)[F2FS_EXTENSION_LEN];
540 int ext_cnt;
541 int i;
542
543 if (is_ext) {
544 ext = F2FS_OPTION(sbi).extensions;
545 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
546 } else {
547 ext = F2FS_OPTION(sbi).noextensions;
548 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
549 }
550
551 for (i = 0; i < ext_cnt; i++) {
552 if (!strcasecmp(new_ext, ext[i]))
553 return true;
554 }
555
556 return false;
557 }
558
559 /*
560 * 1. The same extension name cannot not appear in both compress and non-compress extension
561 * at the same time.
562 * 2. If the compress extension specifies all files, the types specified by the non-compress
563 * extension will be treated as special cases and will not be compressed.
564 * 3. Don't allow the non-compress extension specifies all files.
565 */
f2fs_test_compress_extension(struct f2fs_sb_info * sbi)566 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
567 {
568 unsigned char (*ext)[F2FS_EXTENSION_LEN];
569 unsigned char (*noext)[F2FS_EXTENSION_LEN];
570 int ext_cnt, noext_cnt, index = 0, no_index = 0;
571
572 ext = F2FS_OPTION(sbi).extensions;
573 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
574 noext = F2FS_OPTION(sbi).noextensions;
575 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
576
577 if (!noext_cnt)
578 return 0;
579
580 for (no_index = 0; no_index < noext_cnt; no_index++) {
581 if (!strcasecmp("*", noext[no_index])) {
582 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
583 return -EINVAL;
584 }
585 for (index = 0; index < ext_cnt; index++) {
586 if (!strcasecmp(ext[index], noext[no_index])) {
587 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
588 ext[index]);
589 return -EINVAL;
590 }
591 }
592 }
593 return 0;
594 }
595
596 #ifdef CONFIG_F2FS_FS_LZ4
f2fs_set_lz4hc_level(struct f2fs_sb_info * sbi,const char * str)597 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
598 {
599 #ifdef CONFIG_F2FS_FS_LZ4HC
600 unsigned int level;
601
602 if (strlen(str) == 3) {
603 F2FS_OPTION(sbi).compress_level = 0;
604 return 0;
605 }
606
607 str += 3;
608
609 if (str[0] != ':') {
610 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
611 return -EINVAL;
612 }
613 if (kstrtouint(str + 1, 10, &level))
614 return -EINVAL;
615
616 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
617 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
618 return -EINVAL;
619 }
620
621 F2FS_OPTION(sbi).compress_level = level;
622 return 0;
623 #else
624 if (strlen(str) == 3) {
625 F2FS_OPTION(sbi).compress_level = 0;
626 return 0;
627 }
628 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
629 return -EINVAL;
630 #endif
631 }
632 #endif
633
634 #ifdef CONFIG_F2FS_FS_ZSTD
f2fs_set_zstd_level(struct f2fs_sb_info * sbi,const char * str)635 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
636 {
637 int level;
638 int len = 4;
639
640 if (strlen(str) == len) {
641 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
642 return 0;
643 }
644
645 str += len;
646
647 if (str[0] != ':') {
648 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
649 return -EINVAL;
650 }
651 if (kstrtoint(str + 1, 10, &level))
652 return -EINVAL;
653
654 /* f2fs does not support negative compress level now */
655 if (level < 0) {
656 f2fs_info(sbi, "do not support negative compress level: %d", level);
657 return -ERANGE;
658 }
659
660 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
661 f2fs_info(sbi, "invalid zstd compress level: %d", level);
662 return -EINVAL;
663 }
664
665 F2FS_OPTION(sbi).compress_level = level;
666 return 0;
667 }
668 #endif
669 #endif
670
parse_options(struct super_block * sb,char * options,bool is_remount)671 static int parse_options(struct super_block *sb, char *options, bool is_remount)
672 {
673 struct f2fs_sb_info *sbi = F2FS_SB(sb);
674 substring_t args[MAX_OPT_ARGS];
675 #ifdef CONFIG_F2FS_FS_COMPRESSION
676 unsigned char (*ext)[F2FS_EXTENSION_LEN];
677 unsigned char (*noext)[F2FS_EXTENSION_LEN];
678 int ext_cnt, noext_cnt;
679 #endif
680 char *p, *name;
681 int arg = 0;
682 kuid_t uid;
683 kgid_t gid;
684 int ret;
685
686 if (!options)
687 goto default_check;
688
689 while ((p = strsep(&options, ",")) != NULL) {
690 int token;
691
692 if (!*p)
693 continue;
694 /*
695 * Initialize args struct so we know whether arg was
696 * found; some options take optional arguments.
697 */
698 args[0].to = args[0].from = NULL;
699 token = match_token(p, f2fs_tokens, args);
700
701 switch (token) {
702 case Opt_gc_background:
703 name = match_strdup(&args[0]);
704
705 if (!name)
706 return -ENOMEM;
707 if (!strcmp(name, "on")) {
708 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
709 } else if (!strcmp(name, "off")) {
710 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
711 } else if (!strcmp(name, "sync")) {
712 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
713 } else {
714 kfree(name);
715 return -EINVAL;
716 }
717 kfree(name);
718 break;
719 case Opt_disable_roll_forward:
720 set_opt(sbi, DISABLE_ROLL_FORWARD);
721 break;
722 case Opt_norecovery:
723 /* this option mounts f2fs with ro */
724 set_opt(sbi, NORECOVERY);
725 if (!f2fs_readonly(sb))
726 return -EINVAL;
727 break;
728 case Opt_discard:
729 if (!f2fs_hw_support_discard(sbi)) {
730 f2fs_warn(sbi, "device does not support discard");
731 break;
732 }
733 set_opt(sbi, DISCARD);
734 break;
735 case Opt_nodiscard:
736 if (f2fs_hw_should_discard(sbi)) {
737 f2fs_warn(sbi, "discard is required for zoned block devices");
738 return -EINVAL;
739 }
740 clear_opt(sbi, DISCARD);
741 break;
742 case Opt_noheap:
743 case Opt_heap:
744 f2fs_warn(sbi, "heap/no_heap options were deprecated");
745 break;
746 #ifdef CONFIG_F2FS_FS_XATTR
747 case Opt_user_xattr:
748 set_opt(sbi, XATTR_USER);
749 break;
750 case Opt_nouser_xattr:
751 clear_opt(sbi, XATTR_USER);
752 break;
753 case Opt_inline_xattr:
754 set_opt(sbi, INLINE_XATTR);
755 break;
756 case Opt_noinline_xattr:
757 clear_opt(sbi, INLINE_XATTR);
758 break;
759 case Opt_inline_xattr_size:
760 if (args->from && match_int(args, &arg))
761 return -EINVAL;
762 set_opt(sbi, INLINE_XATTR_SIZE);
763 F2FS_OPTION(sbi).inline_xattr_size = arg;
764 break;
765 #else
766 case Opt_user_xattr:
767 f2fs_info(sbi, "user_xattr options not supported");
768 break;
769 case Opt_nouser_xattr:
770 f2fs_info(sbi, "nouser_xattr options not supported");
771 break;
772 case Opt_inline_xattr:
773 f2fs_info(sbi, "inline_xattr options not supported");
774 break;
775 case Opt_noinline_xattr:
776 f2fs_info(sbi, "noinline_xattr options not supported");
777 break;
778 #endif
779 #ifdef CONFIG_F2FS_FS_POSIX_ACL
780 case Opt_acl:
781 set_opt(sbi, POSIX_ACL);
782 break;
783 case Opt_noacl:
784 clear_opt(sbi, POSIX_ACL);
785 break;
786 #else
787 case Opt_acl:
788 f2fs_info(sbi, "acl options not supported");
789 break;
790 case Opt_noacl:
791 f2fs_info(sbi, "noacl options not supported");
792 break;
793 #endif
794 case Opt_active_logs:
795 if (args->from && match_int(args, &arg))
796 return -EINVAL;
797 if (arg != 2 && arg != 4 &&
798 arg != NR_CURSEG_PERSIST_TYPE)
799 return -EINVAL;
800 F2FS_OPTION(sbi).active_logs = arg;
801 break;
802 case Opt_disable_ext_identify:
803 set_opt(sbi, DISABLE_EXT_IDENTIFY);
804 break;
805 case Opt_inline_data:
806 set_opt(sbi, INLINE_DATA);
807 break;
808 case Opt_inline_dentry:
809 set_opt(sbi, INLINE_DENTRY);
810 break;
811 case Opt_noinline_dentry:
812 clear_opt(sbi, INLINE_DENTRY);
813 break;
814 case Opt_flush_merge:
815 set_opt(sbi, FLUSH_MERGE);
816 break;
817 case Opt_noflush_merge:
818 clear_opt(sbi, FLUSH_MERGE);
819 break;
820 case Opt_nobarrier:
821 set_opt(sbi, NOBARRIER);
822 break;
823 case Opt_barrier:
824 clear_opt(sbi, NOBARRIER);
825 break;
826 case Opt_fastboot:
827 set_opt(sbi, FASTBOOT);
828 break;
829 case Opt_extent_cache:
830 set_opt(sbi, READ_EXTENT_CACHE);
831 break;
832 case Opt_noextent_cache:
833 clear_opt(sbi, READ_EXTENT_CACHE);
834 break;
835 case Opt_noinline_data:
836 clear_opt(sbi, INLINE_DATA);
837 break;
838 case Opt_data_flush:
839 set_opt(sbi, DATA_FLUSH);
840 break;
841 case Opt_reserve_root:
842 if (args->from && match_int(args, &arg))
843 return -EINVAL;
844 if (test_opt(sbi, RESERVE_ROOT)) {
845 f2fs_info(sbi, "Preserve previous reserve_root=%u",
846 F2FS_OPTION(sbi).root_reserved_blocks);
847 } else {
848 F2FS_OPTION(sbi).root_reserved_blocks = arg;
849 set_opt(sbi, RESERVE_ROOT);
850 }
851 break;
852 case Opt_resuid:
853 if (args->from && match_int(args, &arg))
854 return -EINVAL;
855 uid = make_kuid(current_user_ns(), arg);
856 if (!uid_valid(uid)) {
857 f2fs_err(sbi, "Invalid uid value %d", arg);
858 return -EINVAL;
859 }
860 F2FS_OPTION(sbi).s_resuid = uid;
861 break;
862 case Opt_resgid:
863 if (args->from && match_int(args, &arg))
864 return -EINVAL;
865 gid = make_kgid(current_user_ns(), arg);
866 if (!gid_valid(gid)) {
867 f2fs_err(sbi, "Invalid gid value %d", arg);
868 return -EINVAL;
869 }
870 F2FS_OPTION(sbi).s_resgid = gid;
871 break;
872 case Opt_mode:
873 name = match_strdup(&args[0]);
874
875 if (!name)
876 return -ENOMEM;
877 if (!strcmp(name, "adaptive")) {
878 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
879 } else if (!strcmp(name, "lfs")) {
880 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
881 } else if (!strcmp(name, "fragment:segment")) {
882 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
883 } else if (!strcmp(name, "fragment:block")) {
884 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
885 } else {
886 kfree(name);
887 return -EINVAL;
888 }
889 kfree(name);
890 break;
891 #ifdef CONFIG_F2FS_FAULT_INJECTION
892 case Opt_fault_injection:
893 if (args->from && match_int(args, &arg))
894 return -EINVAL;
895 if (f2fs_build_fault_attr(sbi, arg,
896 F2FS_ALL_FAULT_TYPE))
897 return -EINVAL;
898 set_opt(sbi, FAULT_INJECTION);
899 break;
900
901 case Opt_fault_type:
902 if (args->from && match_int(args, &arg))
903 return -EINVAL;
904 if (f2fs_build_fault_attr(sbi, 0, arg))
905 return -EINVAL;
906 set_opt(sbi, FAULT_INJECTION);
907 break;
908 #else
909 case Opt_fault_injection:
910 f2fs_info(sbi, "fault_injection options not supported");
911 break;
912
913 case Opt_fault_type:
914 f2fs_info(sbi, "fault_type options not supported");
915 break;
916 #endif
917 #ifdef CONFIG_QUOTA
918 case Opt_quota:
919 case Opt_usrquota:
920 set_opt(sbi, USRQUOTA);
921 break;
922 case Opt_grpquota:
923 set_opt(sbi, GRPQUOTA);
924 break;
925 case Opt_prjquota:
926 set_opt(sbi, PRJQUOTA);
927 break;
928 case Opt_usrjquota:
929 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
930 if (ret)
931 return ret;
932 break;
933 case Opt_grpjquota:
934 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
935 if (ret)
936 return ret;
937 break;
938 case Opt_prjjquota:
939 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
940 if (ret)
941 return ret;
942 break;
943 case Opt_offusrjquota:
944 ret = f2fs_clear_qf_name(sb, USRQUOTA);
945 if (ret)
946 return ret;
947 break;
948 case Opt_offgrpjquota:
949 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
950 if (ret)
951 return ret;
952 break;
953 case Opt_offprjjquota:
954 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
955 if (ret)
956 return ret;
957 break;
958 case Opt_jqfmt_vfsold:
959 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
960 break;
961 case Opt_jqfmt_vfsv0:
962 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
963 break;
964 case Opt_jqfmt_vfsv1:
965 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
966 break;
967 case Opt_noquota:
968 clear_opt(sbi, QUOTA);
969 clear_opt(sbi, USRQUOTA);
970 clear_opt(sbi, GRPQUOTA);
971 clear_opt(sbi, PRJQUOTA);
972 break;
973 #else
974 case Opt_quota:
975 case Opt_usrquota:
976 case Opt_grpquota:
977 case Opt_prjquota:
978 case Opt_usrjquota:
979 case Opt_grpjquota:
980 case Opt_prjjquota:
981 case Opt_offusrjquota:
982 case Opt_offgrpjquota:
983 case Opt_offprjjquota:
984 case Opt_jqfmt_vfsold:
985 case Opt_jqfmt_vfsv0:
986 case Opt_jqfmt_vfsv1:
987 case Opt_noquota:
988 f2fs_info(sbi, "quota operations not supported");
989 break;
990 #endif
991 case Opt_alloc:
992 name = match_strdup(&args[0]);
993 if (!name)
994 return -ENOMEM;
995
996 if (!strcmp(name, "default")) {
997 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
998 } else if (!strcmp(name, "reuse")) {
999 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1000 } else {
1001 kfree(name);
1002 return -EINVAL;
1003 }
1004 kfree(name);
1005 break;
1006 case Opt_fsync:
1007 name = match_strdup(&args[0]);
1008 if (!name)
1009 return -ENOMEM;
1010 if (!strcmp(name, "posix")) {
1011 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1012 } else if (!strcmp(name, "strict")) {
1013 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1014 } else if (!strcmp(name, "nobarrier")) {
1015 F2FS_OPTION(sbi).fsync_mode =
1016 FSYNC_MODE_NOBARRIER;
1017 } else {
1018 kfree(name);
1019 return -EINVAL;
1020 }
1021 kfree(name);
1022 break;
1023 case Opt_test_dummy_encryption:
1024 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1025 is_remount);
1026 if (ret)
1027 return ret;
1028 break;
1029 case Opt_inlinecrypt:
1030 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1031 sb->s_flags |= SB_INLINECRYPT;
1032 #else
1033 f2fs_info(sbi, "inline encryption not supported");
1034 #endif
1035 break;
1036 case Opt_checkpoint_disable_cap_perc:
1037 if (args->from && match_int(args, &arg))
1038 return -EINVAL;
1039 if (arg < 0 || arg > 100)
1040 return -EINVAL;
1041 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1042 set_opt(sbi, DISABLE_CHECKPOINT);
1043 break;
1044 case Opt_checkpoint_disable_cap:
1045 if (args->from && match_int(args, &arg))
1046 return -EINVAL;
1047 F2FS_OPTION(sbi).unusable_cap = arg;
1048 set_opt(sbi, DISABLE_CHECKPOINT);
1049 break;
1050 case Opt_checkpoint_disable:
1051 set_opt(sbi, DISABLE_CHECKPOINT);
1052 break;
1053 case Opt_checkpoint_enable:
1054 clear_opt(sbi, DISABLE_CHECKPOINT);
1055 break;
1056 case Opt_checkpoint_merge:
1057 set_opt(sbi, MERGE_CHECKPOINT);
1058 break;
1059 case Opt_nocheckpoint_merge:
1060 clear_opt(sbi, MERGE_CHECKPOINT);
1061 break;
1062 #ifdef CONFIG_F2FS_FS_COMPRESSION
1063 case Opt_compress_algorithm:
1064 if (!f2fs_sb_has_compression(sbi)) {
1065 f2fs_info(sbi, "Image doesn't support compression");
1066 break;
1067 }
1068 name = match_strdup(&args[0]);
1069 if (!name)
1070 return -ENOMEM;
1071 if (!strcmp(name, "lzo")) {
1072 #ifdef CONFIG_F2FS_FS_LZO
1073 F2FS_OPTION(sbi).compress_level = 0;
1074 F2FS_OPTION(sbi).compress_algorithm =
1075 COMPRESS_LZO;
1076 #else
1077 f2fs_info(sbi, "kernel doesn't support lzo compression");
1078 #endif
1079 } else if (!strncmp(name, "lz4", 3)) {
1080 #ifdef CONFIG_F2FS_FS_LZ4
1081 ret = f2fs_set_lz4hc_level(sbi, name);
1082 if (ret) {
1083 kfree(name);
1084 return -EINVAL;
1085 }
1086 F2FS_OPTION(sbi).compress_algorithm =
1087 COMPRESS_LZ4;
1088 #else
1089 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1090 #endif
1091 } else if (!strncmp(name, "zstd", 4)) {
1092 #ifdef CONFIG_F2FS_FS_ZSTD
1093 ret = f2fs_set_zstd_level(sbi, name);
1094 if (ret) {
1095 kfree(name);
1096 return -EINVAL;
1097 }
1098 F2FS_OPTION(sbi).compress_algorithm =
1099 COMPRESS_ZSTD;
1100 #else
1101 f2fs_info(sbi, "kernel doesn't support zstd compression");
1102 #endif
1103 } else if (!strcmp(name, "lzo-rle")) {
1104 #ifdef CONFIG_F2FS_FS_LZORLE
1105 F2FS_OPTION(sbi).compress_level = 0;
1106 F2FS_OPTION(sbi).compress_algorithm =
1107 COMPRESS_LZORLE;
1108 #else
1109 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1110 #endif
1111 } else {
1112 kfree(name);
1113 return -EINVAL;
1114 }
1115 kfree(name);
1116 break;
1117 case Opt_compress_log_size:
1118 if (!f2fs_sb_has_compression(sbi)) {
1119 f2fs_info(sbi, "Image doesn't support compression");
1120 break;
1121 }
1122 if (args->from && match_int(args, &arg))
1123 return -EINVAL;
1124 if (arg < MIN_COMPRESS_LOG_SIZE ||
1125 arg > MAX_COMPRESS_LOG_SIZE) {
1126 f2fs_err(sbi,
1127 "Compress cluster log size is out of range");
1128 return -EINVAL;
1129 }
1130 F2FS_OPTION(sbi).compress_log_size = arg;
1131 break;
1132 case Opt_compress_extension:
1133 if (!f2fs_sb_has_compression(sbi)) {
1134 f2fs_info(sbi, "Image doesn't support compression");
1135 break;
1136 }
1137 name = match_strdup(&args[0]);
1138 if (!name)
1139 return -ENOMEM;
1140
1141 ext = F2FS_OPTION(sbi).extensions;
1142 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1143
1144 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1145 ext_cnt >= COMPRESS_EXT_NUM) {
1146 f2fs_err(sbi,
1147 "invalid extension length/number");
1148 kfree(name);
1149 return -EINVAL;
1150 }
1151
1152 if (is_compress_extension_exist(sbi, name, true)) {
1153 kfree(name);
1154 break;
1155 }
1156
1157 strcpy(ext[ext_cnt], name);
1158 F2FS_OPTION(sbi).compress_ext_cnt++;
1159 kfree(name);
1160 break;
1161 case Opt_nocompress_extension:
1162 if (!f2fs_sb_has_compression(sbi)) {
1163 f2fs_info(sbi, "Image doesn't support compression");
1164 break;
1165 }
1166 name = match_strdup(&args[0]);
1167 if (!name)
1168 return -ENOMEM;
1169
1170 noext = F2FS_OPTION(sbi).noextensions;
1171 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1172
1173 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1174 noext_cnt >= COMPRESS_EXT_NUM) {
1175 f2fs_err(sbi,
1176 "invalid extension length/number");
1177 kfree(name);
1178 return -EINVAL;
1179 }
1180
1181 if (is_compress_extension_exist(sbi, name, false)) {
1182 kfree(name);
1183 break;
1184 }
1185
1186 strcpy(noext[noext_cnt], name);
1187 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1188 kfree(name);
1189 break;
1190 case Opt_compress_chksum:
1191 if (!f2fs_sb_has_compression(sbi)) {
1192 f2fs_info(sbi, "Image doesn't support compression");
1193 break;
1194 }
1195 F2FS_OPTION(sbi).compress_chksum = true;
1196 break;
1197 case Opt_compress_mode:
1198 if (!f2fs_sb_has_compression(sbi)) {
1199 f2fs_info(sbi, "Image doesn't support compression");
1200 break;
1201 }
1202 name = match_strdup(&args[0]);
1203 if (!name)
1204 return -ENOMEM;
1205 if (!strcmp(name, "fs")) {
1206 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1207 } else if (!strcmp(name, "user")) {
1208 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1209 } else {
1210 kfree(name);
1211 return -EINVAL;
1212 }
1213 kfree(name);
1214 break;
1215 case Opt_compress_cache:
1216 if (!f2fs_sb_has_compression(sbi)) {
1217 f2fs_info(sbi, "Image doesn't support compression");
1218 break;
1219 }
1220 set_opt(sbi, COMPRESS_CACHE);
1221 break;
1222 #else
1223 case Opt_compress_algorithm:
1224 case Opt_compress_log_size:
1225 case Opt_compress_extension:
1226 case Opt_nocompress_extension:
1227 case Opt_compress_chksum:
1228 case Opt_compress_mode:
1229 case Opt_compress_cache:
1230 f2fs_info(sbi, "compression options not supported");
1231 break;
1232 #endif
1233 case Opt_atgc:
1234 set_opt(sbi, ATGC);
1235 break;
1236 case Opt_gc_merge:
1237 set_opt(sbi, GC_MERGE);
1238 break;
1239 case Opt_nogc_merge:
1240 clear_opt(sbi, GC_MERGE);
1241 break;
1242 case Opt_discard_unit:
1243 name = match_strdup(&args[0]);
1244 if (!name)
1245 return -ENOMEM;
1246 if (!strcmp(name, "block")) {
1247 F2FS_OPTION(sbi).discard_unit =
1248 DISCARD_UNIT_BLOCK;
1249 } else if (!strcmp(name, "segment")) {
1250 F2FS_OPTION(sbi).discard_unit =
1251 DISCARD_UNIT_SEGMENT;
1252 } else if (!strcmp(name, "section")) {
1253 F2FS_OPTION(sbi).discard_unit =
1254 DISCARD_UNIT_SECTION;
1255 } else {
1256 kfree(name);
1257 return -EINVAL;
1258 }
1259 kfree(name);
1260 break;
1261 case Opt_memory_mode:
1262 name = match_strdup(&args[0]);
1263 if (!name)
1264 return -ENOMEM;
1265 if (!strcmp(name, "normal")) {
1266 F2FS_OPTION(sbi).memory_mode =
1267 MEMORY_MODE_NORMAL;
1268 } else if (!strcmp(name, "low")) {
1269 F2FS_OPTION(sbi).memory_mode =
1270 MEMORY_MODE_LOW;
1271 } else {
1272 kfree(name);
1273 return -EINVAL;
1274 }
1275 kfree(name);
1276 break;
1277 case Opt_age_extent_cache:
1278 set_opt(sbi, AGE_EXTENT_CACHE);
1279 break;
1280 case Opt_errors:
1281 name = match_strdup(&args[0]);
1282 if (!name)
1283 return -ENOMEM;
1284 if (!strcmp(name, "remount-ro")) {
1285 F2FS_OPTION(sbi).errors =
1286 MOUNT_ERRORS_READONLY;
1287 } else if (!strcmp(name, "continue")) {
1288 F2FS_OPTION(sbi).errors =
1289 MOUNT_ERRORS_CONTINUE;
1290 } else if (!strcmp(name, "panic")) {
1291 F2FS_OPTION(sbi).errors =
1292 MOUNT_ERRORS_PANIC;
1293 } else {
1294 kfree(name);
1295 return -EINVAL;
1296 }
1297 kfree(name);
1298 break;
1299 default:
1300 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1301 p);
1302 return -EINVAL;
1303 }
1304 }
1305 default_check:
1306 #ifdef CONFIG_QUOTA
1307 if (f2fs_check_quota_options(sbi))
1308 return -EINVAL;
1309 #else
1310 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1311 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1312 return -EINVAL;
1313 }
1314 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1315 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1316 return -EINVAL;
1317 }
1318 #endif
1319
1320 if (!IS_ENABLED(CONFIG_UNICODE) && f2fs_sb_has_casefold(sbi)) {
1321 f2fs_err(sbi,
1322 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1323 return -EINVAL;
1324 }
1325
1326 /*
1327 * The BLKZONED feature indicates that the drive was formatted with
1328 * zone alignment optimization. This is optional for host-aware
1329 * devices, but mandatory for host-managed zoned block devices.
1330 */
1331 if (f2fs_sb_has_blkzoned(sbi)) {
1332 #ifdef CONFIG_BLK_DEV_ZONED
1333 if (F2FS_OPTION(sbi).discard_unit !=
1334 DISCARD_UNIT_SECTION) {
1335 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1336 F2FS_OPTION(sbi).discard_unit =
1337 DISCARD_UNIT_SECTION;
1338 }
1339
1340 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1341 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1342 return -EINVAL;
1343 }
1344 #else
1345 f2fs_err(sbi, "Zoned block device support is not enabled");
1346 return -EINVAL;
1347 #endif
1348 }
1349
1350 #ifdef CONFIG_F2FS_FS_COMPRESSION
1351 if (f2fs_test_compress_extension(sbi)) {
1352 f2fs_err(sbi, "invalid compress or nocompress extension");
1353 return -EINVAL;
1354 }
1355 #endif
1356
1357 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1358 int min_size, max_size;
1359
1360 if (!f2fs_sb_has_extra_attr(sbi) ||
1361 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1362 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1363 return -EINVAL;
1364 }
1365 if (!test_opt(sbi, INLINE_XATTR)) {
1366 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1367 return -EINVAL;
1368 }
1369
1370 min_size = MIN_INLINE_XATTR_SIZE;
1371 max_size = MAX_INLINE_XATTR_SIZE;
1372
1373 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1374 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1375 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1376 min_size, max_size);
1377 return -EINVAL;
1378 }
1379 }
1380
1381 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1382 f2fs_err(sbi, "LFS is not compatible with ATGC");
1383 return -EINVAL;
1384 }
1385
1386 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1387 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1388 return -EINVAL;
1389 }
1390
1391 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1392 f2fs_err(sbi, "Allow to mount readonly mode only");
1393 return -EROFS;
1394 }
1395 return 0;
1396 }
1397
f2fs_alloc_inode(struct super_block * sb)1398 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1399 {
1400 struct f2fs_inode_info *fi;
1401
1402 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1403 return NULL;
1404
1405 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1406 if (!fi)
1407 return NULL;
1408
1409 init_once((void *) fi);
1410
1411 /* Initialize f2fs-specific inode info */
1412 atomic_set(&fi->dirty_pages, 0);
1413 atomic_set(&fi->i_compr_blocks, 0);
1414 init_f2fs_rwsem(&fi->i_sem);
1415 spin_lock_init(&fi->i_size_lock);
1416 INIT_LIST_HEAD(&fi->dirty_list);
1417 INIT_LIST_HEAD(&fi->gdirty_list);
1418 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1419 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1420 init_f2fs_rwsem(&fi->i_xattr_sem);
1421
1422 /* Will be used by directory only */
1423 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1424
1425 return &fi->vfs_inode;
1426 }
1427
f2fs_drop_inode(struct inode * inode)1428 static int f2fs_drop_inode(struct inode *inode)
1429 {
1430 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1431 int ret;
1432
1433 /*
1434 * during filesystem shutdown, if checkpoint is disabled,
1435 * drop useless meta/node dirty pages.
1436 */
1437 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1438 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1439 inode->i_ino == F2FS_META_INO(sbi)) {
1440 trace_f2fs_drop_inode(inode, 1);
1441 return 1;
1442 }
1443 }
1444
1445 /*
1446 * This is to avoid a deadlock condition like below.
1447 * writeback_single_inode(inode)
1448 * - f2fs_write_data_page
1449 * - f2fs_gc -> iput -> evict
1450 * - inode_wait_for_writeback(inode)
1451 */
1452 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1453 if (!inode->i_nlink && !is_bad_inode(inode)) {
1454 /* to avoid evict_inode call simultaneously */
1455 atomic_inc(&inode->i_count);
1456 spin_unlock(&inode->i_lock);
1457
1458 /* should remain fi->extent_tree for writepage */
1459 f2fs_destroy_extent_node(inode);
1460
1461 sb_start_intwrite(inode->i_sb);
1462 f2fs_i_size_write(inode, 0);
1463
1464 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1465 inode, NULL, 0, DATA);
1466 truncate_inode_pages_final(inode->i_mapping);
1467
1468 if (F2FS_HAS_BLOCKS(inode))
1469 f2fs_truncate(inode);
1470
1471 sb_end_intwrite(inode->i_sb);
1472
1473 spin_lock(&inode->i_lock);
1474 atomic_dec(&inode->i_count);
1475 }
1476 trace_f2fs_drop_inode(inode, 0);
1477 return 0;
1478 }
1479 ret = generic_drop_inode(inode);
1480 if (!ret)
1481 ret = fscrypt_drop_inode(inode);
1482 trace_f2fs_drop_inode(inode, ret);
1483 return ret;
1484 }
1485
f2fs_inode_dirtied(struct inode * inode,bool sync)1486 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1487 {
1488 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1489 int ret = 0;
1490
1491 spin_lock(&sbi->inode_lock[DIRTY_META]);
1492 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1493 ret = 1;
1494 } else {
1495 set_inode_flag(inode, FI_DIRTY_INODE);
1496 stat_inc_dirty_inode(sbi, DIRTY_META);
1497 }
1498 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1499 list_add_tail(&F2FS_I(inode)->gdirty_list,
1500 &sbi->inode_list[DIRTY_META]);
1501 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1502 }
1503 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1504 return ret;
1505 }
1506
f2fs_inode_synced(struct inode * inode)1507 void f2fs_inode_synced(struct inode *inode)
1508 {
1509 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1510
1511 spin_lock(&sbi->inode_lock[DIRTY_META]);
1512 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1513 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1514 return;
1515 }
1516 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1517 list_del_init(&F2FS_I(inode)->gdirty_list);
1518 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1519 }
1520 clear_inode_flag(inode, FI_DIRTY_INODE);
1521 clear_inode_flag(inode, FI_AUTO_RECOVER);
1522 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1523 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1524 }
1525
1526 /*
1527 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1528 *
1529 * We should call set_dirty_inode to write the dirty inode through write_inode.
1530 */
f2fs_dirty_inode(struct inode * inode,int flags)1531 static void f2fs_dirty_inode(struct inode *inode, int flags)
1532 {
1533 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1534
1535 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1536 inode->i_ino == F2FS_META_INO(sbi))
1537 return;
1538
1539 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1540 clear_inode_flag(inode, FI_AUTO_RECOVER);
1541
1542 f2fs_inode_dirtied(inode, false);
1543 }
1544
f2fs_free_inode(struct inode * inode)1545 static void f2fs_free_inode(struct inode *inode)
1546 {
1547 fscrypt_free_inode(inode);
1548 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1549 }
1550
destroy_percpu_info(struct f2fs_sb_info * sbi)1551 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1552 {
1553 percpu_counter_destroy(&sbi->total_valid_inode_count);
1554 percpu_counter_destroy(&sbi->rf_node_block_count);
1555 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1556 }
1557
destroy_device_list(struct f2fs_sb_info * sbi)1558 static void destroy_device_list(struct f2fs_sb_info *sbi)
1559 {
1560 int i;
1561
1562 for (i = 0; i < sbi->s_ndevs; i++) {
1563 if (i > 0)
1564 bdev_fput(FDEV(i).bdev_file);
1565 #ifdef CONFIG_BLK_DEV_ZONED
1566 kvfree(FDEV(i).blkz_seq);
1567 #endif
1568 }
1569 kvfree(sbi->devs);
1570 }
1571
f2fs_put_super(struct super_block * sb)1572 static void f2fs_put_super(struct super_block *sb)
1573 {
1574 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1575 int i;
1576 int err = 0;
1577 bool done;
1578
1579 /* unregister procfs/sysfs entries in advance to avoid race case */
1580 f2fs_unregister_sysfs(sbi);
1581
1582 f2fs_quota_off_umount(sb);
1583
1584 /* prevent remaining shrinker jobs */
1585 mutex_lock(&sbi->umount_mutex);
1586
1587 /*
1588 * flush all issued checkpoints and stop checkpoint issue thread.
1589 * after then, all checkpoints should be done by each process context.
1590 */
1591 f2fs_stop_ckpt_thread(sbi);
1592
1593 /*
1594 * We don't need to do checkpoint when superblock is clean.
1595 * But, the previous checkpoint was not done by umount, it needs to do
1596 * clean checkpoint again.
1597 */
1598 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1599 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1600 struct cp_control cpc = {
1601 .reason = CP_UMOUNT,
1602 };
1603 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1604 err = f2fs_write_checkpoint(sbi, &cpc);
1605 }
1606
1607 /* be sure to wait for any on-going discard commands */
1608 done = f2fs_issue_discard_timeout(sbi);
1609 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1610 struct cp_control cpc = {
1611 .reason = CP_UMOUNT | CP_TRIMMED,
1612 };
1613 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1614 err = f2fs_write_checkpoint(sbi, &cpc);
1615 }
1616
1617 /*
1618 * normally superblock is clean, so we need to release this.
1619 * In addition, EIO will skip do checkpoint, we need this as well.
1620 */
1621 f2fs_release_ino_entry(sbi, true);
1622
1623 f2fs_leave_shrinker(sbi);
1624 mutex_unlock(&sbi->umount_mutex);
1625
1626 /* our cp_error case, we can wait for any writeback page */
1627 f2fs_flush_merged_writes(sbi);
1628
1629 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1630
1631 if (err || f2fs_cp_error(sbi)) {
1632 truncate_inode_pages_final(NODE_MAPPING(sbi));
1633 truncate_inode_pages_final(META_MAPPING(sbi));
1634 }
1635
1636 for (i = 0; i < NR_COUNT_TYPE; i++) {
1637 if (!get_pages(sbi, i))
1638 continue;
1639 f2fs_err(sbi, "detect filesystem reference count leak during "
1640 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1641 f2fs_bug_on(sbi, 1);
1642 }
1643
1644 f2fs_bug_on(sbi, sbi->fsync_node_num);
1645
1646 f2fs_destroy_compress_inode(sbi);
1647
1648 iput(sbi->node_inode);
1649 sbi->node_inode = NULL;
1650
1651 iput(sbi->meta_inode);
1652 sbi->meta_inode = NULL;
1653
1654 /*
1655 * iput() can update stat information, if f2fs_write_checkpoint()
1656 * above failed with error.
1657 */
1658 f2fs_destroy_stats(sbi);
1659
1660 /* destroy f2fs internal modules */
1661 f2fs_destroy_node_manager(sbi);
1662 f2fs_destroy_segment_manager(sbi);
1663
1664 /* flush s_error_work before sbi destroy */
1665 flush_work(&sbi->s_error_work);
1666
1667 f2fs_destroy_post_read_wq(sbi);
1668
1669 kvfree(sbi->ckpt);
1670
1671 if (sbi->s_chksum_driver)
1672 crypto_free_shash(sbi->s_chksum_driver);
1673 kfree(sbi->raw_super);
1674
1675 f2fs_destroy_page_array_cache(sbi);
1676 f2fs_destroy_xattr_caches(sbi);
1677 #ifdef CONFIG_QUOTA
1678 for (i = 0; i < MAXQUOTAS; i++)
1679 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1680 #endif
1681 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1682 destroy_percpu_info(sbi);
1683 f2fs_destroy_iostat(sbi);
1684 for (i = 0; i < NR_PAGE_TYPE; i++)
1685 kvfree(sbi->write_io[i]);
1686 #if IS_ENABLED(CONFIG_UNICODE)
1687 utf8_unload(sb->s_encoding);
1688 #endif
1689 }
1690
f2fs_sync_fs(struct super_block * sb,int sync)1691 int f2fs_sync_fs(struct super_block *sb, int sync)
1692 {
1693 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1694 int err = 0;
1695
1696 if (unlikely(f2fs_cp_error(sbi)))
1697 return 0;
1698 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1699 return 0;
1700
1701 trace_f2fs_sync_fs(sb, sync);
1702
1703 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1704 return -EAGAIN;
1705
1706 if (sync) {
1707 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1708 err = f2fs_issue_checkpoint(sbi);
1709 }
1710
1711 return err;
1712 }
1713
f2fs_freeze(struct super_block * sb)1714 static int f2fs_freeze(struct super_block *sb)
1715 {
1716 if (f2fs_readonly(sb))
1717 return 0;
1718
1719 /* IO error happened before */
1720 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1721 return -EIO;
1722
1723 /* must be clean, since sync_filesystem() was already called */
1724 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1725 return -EINVAL;
1726
1727 /* Let's flush checkpoints and stop the thread. */
1728 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1729
1730 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1731 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1732 return 0;
1733 }
1734
f2fs_unfreeze(struct super_block * sb)1735 static int f2fs_unfreeze(struct super_block *sb)
1736 {
1737 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1738 return 0;
1739 }
1740
1741 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1742 static int f2fs_statfs_project(struct super_block *sb,
1743 kprojid_t projid, struct kstatfs *buf)
1744 {
1745 struct kqid qid;
1746 struct dquot *dquot;
1747 u64 limit;
1748 u64 curblock;
1749
1750 qid = make_kqid_projid(projid);
1751 dquot = dqget(sb, qid);
1752 if (IS_ERR(dquot))
1753 return PTR_ERR(dquot);
1754 spin_lock(&dquot->dq_dqb_lock);
1755
1756 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1757 dquot->dq_dqb.dqb_bhardlimit);
1758 if (limit)
1759 limit >>= sb->s_blocksize_bits;
1760
1761 if (limit && buf->f_blocks > limit) {
1762 curblock = (dquot->dq_dqb.dqb_curspace +
1763 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1764 buf->f_blocks = limit;
1765 buf->f_bfree = buf->f_bavail =
1766 (buf->f_blocks > curblock) ?
1767 (buf->f_blocks - curblock) : 0;
1768 }
1769
1770 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1771 dquot->dq_dqb.dqb_ihardlimit);
1772
1773 if (limit && buf->f_files > limit) {
1774 buf->f_files = limit;
1775 buf->f_ffree =
1776 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1777 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1778 }
1779
1780 spin_unlock(&dquot->dq_dqb_lock);
1781 dqput(dquot);
1782 return 0;
1783 }
1784 #endif
1785
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1786 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1787 {
1788 struct super_block *sb = dentry->d_sb;
1789 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1790 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1791 block_t total_count, user_block_count, start_count;
1792 u64 avail_node_count;
1793 unsigned int total_valid_node_count;
1794
1795 total_count = le64_to_cpu(sbi->raw_super->block_count);
1796 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1797 buf->f_type = F2FS_SUPER_MAGIC;
1798 buf->f_bsize = sbi->blocksize;
1799
1800 buf->f_blocks = total_count - start_count;
1801
1802 spin_lock(&sbi->stat_lock);
1803
1804 user_block_count = sbi->user_block_count;
1805 total_valid_node_count = valid_node_count(sbi);
1806 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1807 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1808 sbi->current_reserved_blocks;
1809
1810 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1811 buf->f_bfree = 0;
1812 else
1813 buf->f_bfree -= sbi->unusable_block_count;
1814 spin_unlock(&sbi->stat_lock);
1815
1816 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1817 buf->f_bavail = buf->f_bfree -
1818 F2FS_OPTION(sbi).root_reserved_blocks;
1819 else
1820 buf->f_bavail = 0;
1821
1822 if (avail_node_count > user_block_count) {
1823 buf->f_files = user_block_count;
1824 buf->f_ffree = buf->f_bavail;
1825 } else {
1826 buf->f_files = avail_node_count;
1827 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1828 buf->f_bavail);
1829 }
1830
1831 buf->f_namelen = F2FS_NAME_LEN;
1832 buf->f_fsid = u64_to_fsid(id);
1833
1834 #ifdef CONFIG_QUOTA
1835 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1836 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1837 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1838 }
1839 #endif
1840 return 0;
1841 }
1842
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1843 static inline void f2fs_show_quota_options(struct seq_file *seq,
1844 struct super_block *sb)
1845 {
1846 #ifdef CONFIG_QUOTA
1847 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1848
1849 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1850 char *fmtname = "";
1851
1852 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1853 case QFMT_VFS_OLD:
1854 fmtname = "vfsold";
1855 break;
1856 case QFMT_VFS_V0:
1857 fmtname = "vfsv0";
1858 break;
1859 case QFMT_VFS_V1:
1860 fmtname = "vfsv1";
1861 break;
1862 }
1863 seq_printf(seq, ",jqfmt=%s", fmtname);
1864 }
1865
1866 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1867 seq_show_option(seq, "usrjquota",
1868 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1869
1870 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1871 seq_show_option(seq, "grpjquota",
1872 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1873
1874 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1875 seq_show_option(seq, "prjjquota",
1876 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1877 #endif
1878 }
1879
1880 #ifdef CONFIG_F2FS_FS_COMPRESSION
f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1881 static inline void f2fs_show_compress_options(struct seq_file *seq,
1882 struct super_block *sb)
1883 {
1884 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1885 char *algtype = "";
1886 int i;
1887
1888 if (!f2fs_sb_has_compression(sbi))
1889 return;
1890
1891 switch (F2FS_OPTION(sbi).compress_algorithm) {
1892 case COMPRESS_LZO:
1893 algtype = "lzo";
1894 break;
1895 case COMPRESS_LZ4:
1896 algtype = "lz4";
1897 break;
1898 case COMPRESS_ZSTD:
1899 algtype = "zstd";
1900 break;
1901 case COMPRESS_LZORLE:
1902 algtype = "lzo-rle";
1903 break;
1904 }
1905 seq_printf(seq, ",compress_algorithm=%s", algtype);
1906
1907 if (F2FS_OPTION(sbi).compress_level)
1908 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1909
1910 seq_printf(seq, ",compress_log_size=%u",
1911 F2FS_OPTION(sbi).compress_log_size);
1912
1913 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1914 seq_printf(seq, ",compress_extension=%s",
1915 F2FS_OPTION(sbi).extensions[i]);
1916 }
1917
1918 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1919 seq_printf(seq, ",nocompress_extension=%s",
1920 F2FS_OPTION(sbi).noextensions[i]);
1921 }
1922
1923 if (F2FS_OPTION(sbi).compress_chksum)
1924 seq_puts(seq, ",compress_chksum");
1925
1926 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1927 seq_printf(seq, ",compress_mode=%s", "fs");
1928 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1929 seq_printf(seq, ",compress_mode=%s", "user");
1930
1931 if (test_opt(sbi, COMPRESS_CACHE))
1932 seq_puts(seq, ",compress_cache");
1933 }
1934 #endif
1935
f2fs_show_options(struct seq_file * seq,struct dentry * root)1936 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1937 {
1938 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1939
1940 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1941 seq_printf(seq, ",background_gc=%s", "sync");
1942 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1943 seq_printf(seq, ",background_gc=%s", "on");
1944 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1945 seq_printf(seq, ",background_gc=%s", "off");
1946
1947 if (test_opt(sbi, GC_MERGE))
1948 seq_puts(seq, ",gc_merge");
1949 else
1950 seq_puts(seq, ",nogc_merge");
1951
1952 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1953 seq_puts(seq, ",disable_roll_forward");
1954 if (test_opt(sbi, NORECOVERY))
1955 seq_puts(seq, ",norecovery");
1956 if (test_opt(sbi, DISCARD)) {
1957 seq_puts(seq, ",discard");
1958 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1959 seq_printf(seq, ",discard_unit=%s", "block");
1960 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1961 seq_printf(seq, ",discard_unit=%s", "segment");
1962 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1963 seq_printf(seq, ",discard_unit=%s", "section");
1964 } else {
1965 seq_puts(seq, ",nodiscard");
1966 }
1967 #ifdef CONFIG_F2FS_FS_XATTR
1968 if (test_opt(sbi, XATTR_USER))
1969 seq_puts(seq, ",user_xattr");
1970 else
1971 seq_puts(seq, ",nouser_xattr");
1972 if (test_opt(sbi, INLINE_XATTR))
1973 seq_puts(seq, ",inline_xattr");
1974 else
1975 seq_puts(seq, ",noinline_xattr");
1976 if (test_opt(sbi, INLINE_XATTR_SIZE))
1977 seq_printf(seq, ",inline_xattr_size=%u",
1978 F2FS_OPTION(sbi).inline_xattr_size);
1979 #endif
1980 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1981 if (test_opt(sbi, POSIX_ACL))
1982 seq_puts(seq, ",acl");
1983 else
1984 seq_puts(seq, ",noacl");
1985 #endif
1986 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1987 seq_puts(seq, ",disable_ext_identify");
1988 if (test_opt(sbi, INLINE_DATA))
1989 seq_puts(seq, ",inline_data");
1990 else
1991 seq_puts(seq, ",noinline_data");
1992 if (test_opt(sbi, INLINE_DENTRY))
1993 seq_puts(seq, ",inline_dentry");
1994 else
1995 seq_puts(seq, ",noinline_dentry");
1996 if (test_opt(sbi, FLUSH_MERGE))
1997 seq_puts(seq, ",flush_merge");
1998 else
1999 seq_puts(seq, ",noflush_merge");
2000 if (test_opt(sbi, NOBARRIER))
2001 seq_puts(seq, ",nobarrier");
2002 else
2003 seq_puts(seq, ",barrier");
2004 if (test_opt(sbi, FASTBOOT))
2005 seq_puts(seq, ",fastboot");
2006 if (test_opt(sbi, READ_EXTENT_CACHE))
2007 seq_puts(seq, ",extent_cache");
2008 else
2009 seq_puts(seq, ",noextent_cache");
2010 if (test_opt(sbi, AGE_EXTENT_CACHE))
2011 seq_puts(seq, ",age_extent_cache");
2012 if (test_opt(sbi, DATA_FLUSH))
2013 seq_puts(seq, ",data_flush");
2014
2015 seq_puts(seq, ",mode=");
2016 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2017 seq_puts(seq, "adaptive");
2018 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2019 seq_puts(seq, "lfs");
2020 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2021 seq_puts(seq, "fragment:segment");
2022 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2023 seq_puts(seq, "fragment:block");
2024 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2025 if (test_opt(sbi, RESERVE_ROOT))
2026 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2027 F2FS_OPTION(sbi).root_reserved_blocks,
2028 from_kuid_munged(&init_user_ns,
2029 F2FS_OPTION(sbi).s_resuid),
2030 from_kgid_munged(&init_user_ns,
2031 F2FS_OPTION(sbi).s_resgid));
2032 #ifdef CONFIG_F2FS_FAULT_INJECTION
2033 if (test_opt(sbi, FAULT_INJECTION)) {
2034 seq_printf(seq, ",fault_injection=%u",
2035 F2FS_OPTION(sbi).fault_info.inject_rate);
2036 seq_printf(seq, ",fault_type=%u",
2037 F2FS_OPTION(sbi).fault_info.inject_type);
2038 }
2039 #endif
2040 #ifdef CONFIG_QUOTA
2041 if (test_opt(sbi, QUOTA))
2042 seq_puts(seq, ",quota");
2043 if (test_opt(sbi, USRQUOTA))
2044 seq_puts(seq, ",usrquota");
2045 if (test_opt(sbi, GRPQUOTA))
2046 seq_puts(seq, ",grpquota");
2047 if (test_opt(sbi, PRJQUOTA))
2048 seq_puts(seq, ",prjquota");
2049 #endif
2050 f2fs_show_quota_options(seq, sbi->sb);
2051
2052 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2053
2054 if (sbi->sb->s_flags & SB_INLINECRYPT)
2055 seq_puts(seq, ",inlinecrypt");
2056
2057 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2058 seq_printf(seq, ",alloc_mode=%s", "default");
2059 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2060 seq_printf(seq, ",alloc_mode=%s", "reuse");
2061
2062 if (test_opt(sbi, DISABLE_CHECKPOINT))
2063 seq_printf(seq, ",checkpoint=disable:%u",
2064 F2FS_OPTION(sbi).unusable_cap);
2065 if (test_opt(sbi, MERGE_CHECKPOINT))
2066 seq_puts(seq, ",checkpoint_merge");
2067 else
2068 seq_puts(seq, ",nocheckpoint_merge");
2069 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2070 seq_printf(seq, ",fsync_mode=%s", "posix");
2071 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2072 seq_printf(seq, ",fsync_mode=%s", "strict");
2073 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2074 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2075
2076 #ifdef CONFIG_F2FS_FS_COMPRESSION
2077 f2fs_show_compress_options(seq, sbi->sb);
2078 #endif
2079
2080 if (test_opt(sbi, ATGC))
2081 seq_puts(seq, ",atgc");
2082
2083 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2084 seq_printf(seq, ",memory=%s", "normal");
2085 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2086 seq_printf(seq, ",memory=%s", "low");
2087
2088 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2089 seq_printf(seq, ",errors=%s", "remount-ro");
2090 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2091 seq_printf(seq, ",errors=%s", "continue");
2092 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2093 seq_printf(seq, ",errors=%s", "panic");
2094
2095 return 0;
2096 }
2097
default_options(struct f2fs_sb_info * sbi,bool remount)2098 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2099 {
2100 /* init some FS parameters */
2101 if (!remount) {
2102 set_opt(sbi, READ_EXTENT_CACHE);
2103 clear_opt(sbi, DISABLE_CHECKPOINT);
2104
2105 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2106 set_opt(sbi, DISCARD);
2107
2108 if (f2fs_sb_has_blkzoned(sbi))
2109 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2110 else
2111 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2112 }
2113
2114 if (f2fs_sb_has_readonly(sbi))
2115 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2116 else
2117 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2118
2119 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2120 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2121 SMALL_VOLUME_SEGMENTS)
2122 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2123 else
2124 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2125 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2126 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2127 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2128 if (f2fs_sb_has_compression(sbi)) {
2129 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2130 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2131 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2132 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2133 }
2134 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2135 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2136 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2137
2138 set_opt(sbi, INLINE_XATTR);
2139 set_opt(sbi, INLINE_DATA);
2140 set_opt(sbi, INLINE_DENTRY);
2141 set_opt(sbi, MERGE_CHECKPOINT);
2142 F2FS_OPTION(sbi).unusable_cap = 0;
2143 sbi->sb->s_flags |= SB_LAZYTIME;
2144 if (!f2fs_is_readonly(sbi))
2145 set_opt(sbi, FLUSH_MERGE);
2146 if (f2fs_sb_has_blkzoned(sbi))
2147 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2148 else
2149 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2150
2151 #ifdef CONFIG_F2FS_FS_XATTR
2152 set_opt(sbi, XATTR_USER);
2153 #endif
2154 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2155 set_opt(sbi, POSIX_ACL);
2156 #endif
2157
2158 f2fs_build_fault_attr(sbi, 0, 0);
2159 }
2160
2161 #ifdef CONFIG_QUOTA
2162 static int f2fs_enable_quotas(struct super_block *sb);
2163 #endif
2164
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)2165 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2166 {
2167 unsigned int s_flags = sbi->sb->s_flags;
2168 struct cp_control cpc;
2169 unsigned int gc_mode = sbi->gc_mode;
2170 int err = 0;
2171 int ret;
2172 block_t unusable;
2173
2174 if (s_flags & SB_RDONLY) {
2175 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2176 return -EINVAL;
2177 }
2178 sbi->sb->s_flags |= SB_ACTIVE;
2179
2180 /* check if we need more GC first */
2181 unusable = f2fs_get_unusable_blocks(sbi);
2182 if (!f2fs_disable_cp_again(sbi, unusable))
2183 goto skip_gc;
2184
2185 f2fs_update_time(sbi, DISABLE_TIME);
2186
2187 sbi->gc_mode = GC_URGENT_HIGH;
2188
2189 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2190 struct f2fs_gc_control gc_control = {
2191 .victim_segno = NULL_SEGNO,
2192 .init_gc_type = FG_GC,
2193 .should_migrate_blocks = false,
2194 .err_gc_skipped = true,
2195 .no_bg_gc = true,
2196 .nr_free_secs = 1 };
2197
2198 f2fs_down_write(&sbi->gc_lock);
2199 stat_inc_gc_call_count(sbi, FOREGROUND);
2200 err = f2fs_gc(sbi, &gc_control);
2201 if (err == -ENODATA) {
2202 err = 0;
2203 break;
2204 }
2205 if (err && err != -EAGAIN)
2206 break;
2207 }
2208
2209 ret = sync_filesystem(sbi->sb);
2210 if (ret || err) {
2211 err = ret ? ret : err;
2212 goto restore_flag;
2213 }
2214
2215 unusable = f2fs_get_unusable_blocks(sbi);
2216 if (f2fs_disable_cp_again(sbi, unusable)) {
2217 err = -EAGAIN;
2218 goto restore_flag;
2219 }
2220
2221 skip_gc:
2222 f2fs_down_write(&sbi->gc_lock);
2223 cpc.reason = CP_PAUSE;
2224 set_sbi_flag(sbi, SBI_CP_DISABLED);
2225 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2226 err = f2fs_write_checkpoint(sbi, &cpc);
2227 if (err)
2228 goto out_unlock;
2229
2230 spin_lock(&sbi->stat_lock);
2231 sbi->unusable_block_count = unusable;
2232 spin_unlock(&sbi->stat_lock);
2233
2234 out_unlock:
2235 f2fs_up_write(&sbi->gc_lock);
2236 restore_flag:
2237 sbi->gc_mode = gc_mode;
2238 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2239 return err;
2240 }
2241
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2242 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2243 {
2244 int retry = DEFAULT_RETRY_IO_COUNT;
2245
2246 /* we should flush all the data to keep data consistency */
2247 do {
2248 sync_inodes_sb(sbi->sb);
2249 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2250 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2251
2252 if (unlikely(retry < 0))
2253 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2254
2255 f2fs_down_write(&sbi->gc_lock);
2256 f2fs_dirty_to_prefree(sbi);
2257
2258 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2259 set_sbi_flag(sbi, SBI_IS_DIRTY);
2260 f2fs_up_write(&sbi->gc_lock);
2261
2262 f2fs_sync_fs(sbi->sb, 1);
2263
2264 /* Let's ensure there's no pending checkpoint anymore */
2265 f2fs_flush_ckpt_thread(sbi);
2266 }
2267
f2fs_remount(struct super_block * sb,int * flags,char * data)2268 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2269 {
2270 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2271 struct f2fs_mount_info org_mount_opt;
2272 unsigned long old_sb_flags;
2273 int err;
2274 bool need_restart_gc = false, need_stop_gc = false;
2275 bool need_restart_flush = false, need_stop_flush = false;
2276 bool need_restart_discard = false, need_stop_discard = false;
2277 bool need_enable_checkpoint = false, need_disable_checkpoint = false;
2278 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2279 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2280 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2281 bool no_atgc = !test_opt(sbi, ATGC);
2282 bool no_discard = !test_opt(sbi, DISCARD);
2283 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2284 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2285 #ifdef CONFIG_QUOTA
2286 int i, j;
2287 #endif
2288
2289 /*
2290 * Save the old mount options in case we
2291 * need to restore them.
2292 */
2293 org_mount_opt = sbi->mount_opt;
2294 old_sb_flags = sb->s_flags;
2295
2296 #ifdef CONFIG_QUOTA
2297 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2298 for (i = 0; i < MAXQUOTAS; i++) {
2299 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2300 org_mount_opt.s_qf_names[i] =
2301 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2302 GFP_KERNEL);
2303 if (!org_mount_opt.s_qf_names[i]) {
2304 for (j = 0; j < i; j++)
2305 kfree(org_mount_opt.s_qf_names[j]);
2306 return -ENOMEM;
2307 }
2308 } else {
2309 org_mount_opt.s_qf_names[i] = NULL;
2310 }
2311 }
2312 #endif
2313
2314 /* recover superblocks we couldn't write due to previous RO mount */
2315 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2316 err = f2fs_commit_super(sbi, false);
2317 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2318 err);
2319 if (!err)
2320 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2321 }
2322
2323 default_options(sbi, true);
2324
2325 /* parse mount options */
2326 err = parse_options(sb, data, true);
2327 if (err)
2328 goto restore_opts;
2329
2330 #ifdef CONFIG_BLK_DEV_ZONED
2331 if (f2fs_sb_has_blkzoned(sbi) &&
2332 sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
2333 f2fs_err(sbi,
2334 "zoned: max open zones %u is too small, need at least %u open zones",
2335 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
2336 err = -EINVAL;
2337 goto restore_opts;
2338 }
2339 #endif
2340
2341 /* flush outstanding errors before changing fs state */
2342 flush_work(&sbi->s_error_work);
2343
2344 /*
2345 * Previous and new state of filesystem is RO,
2346 * so skip checking GC and FLUSH_MERGE conditions.
2347 */
2348 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2349 goto skip;
2350
2351 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2352 err = -EROFS;
2353 goto restore_opts;
2354 }
2355
2356 #ifdef CONFIG_QUOTA
2357 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2358 err = dquot_suspend(sb, -1);
2359 if (err < 0)
2360 goto restore_opts;
2361 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2362 /* dquot_resume needs RW */
2363 sb->s_flags &= ~SB_RDONLY;
2364 if (sb_any_quota_suspended(sb)) {
2365 dquot_resume(sb, -1);
2366 } else if (f2fs_sb_has_quota_ino(sbi)) {
2367 err = f2fs_enable_quotas(sb);
2368 if (err)
2369 goto restore_opts;
2370 }
2371 }
2372 #endif
2373 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2374 err = -EINVAL;
2375 f2fs_warn(sbi, "LFS is not compatible with IPU");
2376 goto restore_opts;
2377 }
2378
2379 /* disallow enable atgc dynamically */
2380 if (no_atgc == !!test_opt(sbi, ATGC)) {
2381 err = -EINVAL;
2382 f2fs_warn(sbi, "switch atgc option is not allowed");
2383 goto restore_opts;
2384 }
2385
2386 /* disallow enable/disable extent_cache dynamically */
2387 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2388 err = -EINVAL;
2389 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2390 goto restore_opts;
2391 }
2392 /* disallow enable/disable age extent_cache dynamically */
2393 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2394 err = -EINVAL;
2395 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2396 goto restore_opts;
2397 }
2398
2399 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2400 err = -EINVAL;
2401 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2402 goto restore_opts;
2403 }
2404
2405 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2406 err = -EINVAL;
2407 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2408 goto restore_opts;
2409 }
2410
2411 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2412 err = -EINVAL;
2413 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2414 goto restore_opts;
2415 }
2416
2417 /*
2418 * We stop the GC thread if FS is mounted as RO
2419 * or if background_gc = off is passed in mount
2420 * option. Also sync the filesystem.
2421 */
2422 if ((*flags & SB_RDONLY) ||
2423 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2424 !test_opt(sbi, GC_MERGE))) {
2425 if (sbi->gc_thread) {
2426 f2fs_stop_gc_thread(sbi);
2427 need_restart_gc = true;
2428 }
2429 } else if (!sbi->gc_thread) {
2430 err = f2fs_start_gc_thread(sbi);
2431 if (err)
2432 goto restore_opts;
2433 need_stop_gc = true;
2434 }
2435
2436 if (*flags & SB_RDONLY) {
2437 sync_inodes_sb(sb);
2438
2439 set_sbi_flag(sbi, SBI_IS_DIRTY);
2440 set_sbi_flag(sbi, SBI_IS_CLOSE);
2441 f2fs_sync_fs(sb, 1);
2442 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2443 }
2444
2445 /*
2446 * We stop issue flush thread if FS is mounted as RO
2447 * or if flush_merge is not passed in mount option.
2448 */
2449 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2450 clear_opt(sbi, FLUSH_MERGE);
2451 f2fs_destroy_flush_cmd_control(sbi, false);
2452 need_restart_flush = true;
2453 } else {
2454 err = f2fs_create_flush_cmd_control(sbi);
2455 if (err)
2456 goto restore_gc;
2457 need_stop_flush = true;
2458 }
2459
2460 if (no_discard == !!test_opt(sbi, DISCARD)) {
2461 if (test_opt(sbi, DISCARD)) {
2462 err = f2fs_start_discard_thread(sbi);
2463 if (err)
2464 goto restore_flush;
2465 need_stop_discard = true;
2466 } else {
2467 f2fs_stop_discard_thread(sbi);
2468 f2fs_issue_discard_timeout(sbi);
2469 need_restart_discard = true;
2470 }
2471 }
2472
2473 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2474 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2475 err = f2fs_disable_checkpoint(sbi);
2476 if (err)
2477 goto restore_discard;
2478 need_enable_checkpoint = true;
2479 } else {
2480 f2fs_enable_checkpoint(sbi);
2481 need_disable_checkpoint = true;
2482 }
2483 }
2484
2485 /*
2486 * Place this routine at the end, since a new checkpoint would be
2487 * triggered while remount and we need to take care of it before
2488 * returning from remount.
2489 */
2490 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2491 !test_opt(sbi, MERGE_CHECKPOINT)) {
2492 f2fs_stop_ckpt_thread(sbi);
2493 } else {
2494 /* Flush if the prevous checkpoint, if exists. */
2495 f2fs_flush_ckpt_thread(sbi);
2496
2497 err = f2fs_start_ckpt_thread(sbi);
2498 if (err) {
2499 f2fs_err(sbi,
2500 "Failed to start F2FS issue_checkpoint_thread (%d)",
2501 err);
2502 goto restore_checkpoint;
2503 }
2504 }
2505
2506 skip:
2507 #ifdef CONFIG_QUOTA
2508 /* Release old quota file names */
2509 for (i = 0; i < MAXQUOTAS; i++)
2510 kfree(org_mount_opt.s_qf_names[i]);
2511 #endif
2512 /* Update the POSIXACL Flag */
2513 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2514 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2515
2516 limit_reserve_root(sbi);
2517 adjust_unusable_cap_perc(sbi);
2518 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2519 return 0;
2520 restore_checkpoint:
2521 if (need_enable_checkpoint) {
2522 f2fs_enable_checkpoint(sbi);
2523 } else if (need_disable_checkpoint) {
2524 if (f2fs_disable_checkpoint(sbi))
2525 f2fs_warn(sbi, "checkpoint has not been disabled");
2526 }
2527 restore_discard:
2528 if (need_restart_discard) {
2529 if (f2fs_start_discard_thread(sbi))
2530 f2fs_warn(sbi, "discard has been stopped");
2531 } else if (need_stop_discard) {
2532 f2fs_stop_discard_thread(sbi);
2533 }
2534 restore_flush:
2535 if (need_restart_flush) {
2536 if (f2fs_create_flush_cmd_control(sbi))
2537 f2fs_warn(sbi, "background flush thread has stopped");
2538 } else if (need_stop_flush) {
2539 clear_opt(sbi, FLUSH_MERGE);
2540 f2fs_destroy_flush_cmd_control(sbi, false);
2541 }
2542 restore_gc:
2543 if (need_restart_gc) {
2544 if (f2fs_start_gc_thread(sbi))
2545 f2fs_warn(sbi, "background gc thread has stopped");
2546 } else if (need_stop_gc) {
2547 f2fs_stop_gc_thread(sbi);
2548 }
2549 restore_opts:
2550 #ifdef CONFIG_QUOTA
2551 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2552 for (i = 0; i < MAXQUOTAS; i++) {
2553 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2554 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2555 }
2556 #endif
2557 sbi->mount_opt = org_mount_opt;
2558 sb->s_flags = old_sb_flags;
2559 return err;
2560 }
2561
f2fs_shutdown(struct super_block * sb)2562 static void f2fs_shutdown(struct super_block *sb)
2563 {
2564 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false);
2565 }
2566
2567 #ifdef CONFIG_QUOTA
f2fs_need_recovery(struct f2fs_sb_info * sbi)2568 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2569 {
2570 /* need to recovery orphan */
2571 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2572 return true;
2573 /* need to recovery data */
2574 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2575 return false;
2576 if (test_opt(sbi, NORECOVERY))
2577 return false;
2578 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2579 }
2580
f2fs_recover_quota_begin(struct f2fs_sb_info * sbi)2581 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2582 {
2583 bool readonly = f2fs_readonly(sbi->sb);
2584
2585 if (!f2fs_need_recovery(sbi))
2586 return false;
2587
2588 /* it doesn't need to check f2fs_sb_has_readonly() */
2589 if (f2fs_hw_is_readonly(sbi))
2590 return false;
2591
2592 if (readonly) {
2593 sbi->sb->s_flags &= ~SB_RDONLY;
2594 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2595 }
2596
2597 /*
2598 * Turn on quotas which were not enabled for read-only mounts if
2599 * filesystem has quota feature, so that they are updated correctly.
2600 */
2601 return f2fs_enable_quota_files(sbi, readonly);
2602 }
2603
f2fs_recover_quota_end(struct f2fs_sb_info * sbi,bool quota_enabled)2604 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2605 bool quota_enabled)
2606 {
2607 if (quota_enabled)
2608 f2fs_quota_off_umount(sbi->sb);
2609
2610 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2611 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2612 sbi->sb->s_flags |= SB_RDONLY;
2613 }
2614 }
2615
2616 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2617 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2618 size_t len, loff_t off)
2619 {
2620 struct inode *inode = sb_dqopt(sb)->files[type];
2621 struct address_space *mapping = inode->i_mapping;
2622 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2623 int offset = off & (sb->s_blocksize - 1);
2624 int tocopy;
2625 size_t toread;
2626 loff_t i_size = i_size_read(inode);
2627 struct page *page;
2628
2629 if (off > i_size)
2630 return 0;
2631
2632 if (off + len > i_size)
2633 len = i_size - off;
2634 toread = len;
2635 while (toread > 0) {
2636 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2637 repeat:
2638 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2639 if (IS_ERR(page)) {
2640 if (PTR_ERR(page) == -ENOMEM) {
2641 memalloc_retry_wait(GFP_NOFS);
2642 goto repeat;
2643 }
2644 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2645 return PTR_ERR(page);
2646 }
2647
2648 lock_page(page);
2649
2650 if (unlikely(page->mapping != mapping)) {
2651 f2fs_put_page(page, 1);
2652 goto repeat;
2653 }
2654 if (unlikely(!PageUptodate(page))) {
2655 f2fs_put_page(page, 1);
2656 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2657 return -EIO;
2658 }
2659
2660 memcpy_from_page(data, page, offset, tocopy);
2661 f2fs_put_page(page, 1);
2662
2663 offset = 0;
2664 toread -= tocopy;
2665 data += tocopy;
2666 blkidx++;
2667 }
2668 return len;
2669 }
2670
2671 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2672 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2673 const char *data, size_t len, loff_t off)
2674 {
2675 struct inode *inode = sb_dqopt(sb)->files[type];
2676 struct address_space *mapping = inode->i_mapping;
2677 const struct address_space_operations *a_ops = mapping->a_ops;
2678 int offset = off & (sb->s_blocksize - 1);
2679 size_t towrite = len;
2680 struct folio *folio;
2681 void *fsdata = NULL;
2682 int err = 0;
2683 int tocopy;
2684
2685 while (towrite > 0) {
2686 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2687 towrite);
2688 retry:
2689 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2690 &folio, &fsdata);
2691 if (unlikely(err)) {
2692 if (err == -ENOMEM) {
2693 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2694 goto retry;
2695 }
2696 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2697 break;
2698 }
2699
2700 memcpy_to_folio(folio, offset_in_folio(folio, off), data, tocopy);
2701
2702 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2703 folio, fsdata);
2704 offset = 0;
2705 towrite -= tocopy;
2706 off += tocopy;
2707 data += tocopy;
2708 cond_resched();
2709 }
2710
2711 if (len == towrite)
2712 return err;
2713 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2714 f2fs_mark_inode_dirty_sync(inode, false);
2715 return len - towrite;
2716 }
2717
f2fs_dquot_initialize(struct inode * inode)2718 int f2fs_dquot_initialize(struct inode *inode)
2719 {
2720 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2721 return -ESRCH;
2722
2723 return dquot_initialize(inode);
2724 }
2725
f2fs_get_dquots(struct inode * inode)2726 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2727 {
2728 return F2FS_I(inode)->i_dquot;
2729 }
2730
f2fs_get_reserved_space(struct inode * inode)2731 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2732 {
2733 return &F2FS_I(inode)->i_reserved_quota;
2734 }
2735
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2736 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2737 {
2738 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2739 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2740 return 0;
2741 }
2742
2743 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2744 F2FS_OPTION(sbi).s_jquota_fmt, type);
2745 }
2746
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2747 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2748 {
2749 int enabled = 0;
2750 int i, err;
2751
2752 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2753 err = f2fs_enable_quotas(sbi->sb);
2754 if (err) {
2755 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2756 return 0;
2757 }
2758 return 1;
2759 }
2760
2761 for (i = 0; i < MAXQUOTAS; i++) {
2762 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2763 err = f2fs_quota_on_mount(sbi, i);
2764 if (!err) {
2765 enabled = 1;
2766 continue;
2767 }
2768 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2769 err, i);
2770 }
2771 }
2772 return enabled;
2773 }
2774
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2775 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2776 unsigned int flags)
2777 {
2778 struct inode *qf_inode;
2779 unsigned long qf_inum;
2780 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2781 int err;
2782
2783 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2784
2785 qf_inum = f2fs_qf_ino(sb, type);
2786 if (!qf_inum)
2787 return -EPERM;
2788
2789 qf_inode = f2fs_iget(sb, qf_inum);
2790 if (IS_ERR(qf_inode)) {
2791 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2792 return PTR_ERR(qf_inode);
2793 }
2794
2795 /* Don't account quota for quota files to avoid recursion */
2796 inode_lock(qf_inode);
2797 qf_inode->i_flags |= S_NOQUOTA;
2798
2799 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2800 F2FS_I(qf_inode)->i_flags |= qf_flag;
2801 f2fs_set_inode_flags(qf_inode);
2802 }
2803 inode_unlock(qf_inode);
2804
2805 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2806 iput(qf_inode);
2807 return err;
2808 }
2809
f2fs_enable_quotas(struct super_block * sb)2810 static int f2fs_enable_quotas(struct super_block *sb)
2811 {
2812 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2813 int type, err = 0;
2814 unsigned long qf_inum;
2815 bool quota_mopt[MAXQUOTAS] = {
2816 test_opt(sbi, USRQUOTA),
2817 test_opt(sbi, GRPQUOTA),
2818 test_opt(sbi, PRJQUOTA),
2819 };
2820
2821 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2822 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2823 return 0;
2824 }
2825
2826 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2827
2828 for (type = 0; type < MAXQUOTAS; type++) {
2829 qf_inum = f2fs_qf_ino(sb, type);
2830 if (qf_inum) {
2831 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2832 DQUOT_USAGE_ENABLED |
2833 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2834 if (err) {
2835 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2836 type, err);
2837 for (type--; type >= 0; type--)
2838 dquot_quota_off(sb, type);
2839 set_sbi_flag(F2FS_SB(sb),
2840 SBI_QUOTA_NEED_REPAIR);
2841 return err;
2842 }
2843 }
2844 }
2845 return 0;
2846 }
2847
f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2848 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2849 {
2850 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2851 struct address_space *mapping = dqopt->files[type]->i_mapping;
2852 int ret = 0;
2853
2854 ret = dquot_writeback_dquots(sbi->sb, type);
2855 if (ret)
2856 goto out;
2857
2858 ret = filemap_fdatawrite(mapping);
2859 if (ret)
2860 goto out;
2861
2862 /* if we are using journalled quota */
2863 if (is_journalled_quota(sbi))
2864 goto out;
2865
2866 ret = filemap_fdatawait(mapping);
2867
2868 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2869 out:
2870 if (ret)
2871 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2872 return ret;
2873 }
2874
f2fs_quota_sync(struct super_block * sb,int type)2875 int f2fs_quota_sync(struct super_block *sb, int type)
2876 {
2877 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2878 struct quota_info *dqopt = sb_dqopt(sb);
2879 int cnt;
2880 int ret = 0;
2881
2882 /*
2883 * Now when everything is written we can discard the pagecache so
2884 * that userspace sees the changes.
2885 */
2886 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2887
2888 if (type != -1 && cnt != type)
2889 continue;
2890
2891 if (!sb_has_quota_active(sb, cnt))
2892 continue;
2893
2894 if (!f2fs_sb_has_quota_ino(sbi))
2895 inode_lock(dqopt->files[cnt]);
2896
2897 /*
2898 * do_quotactl
2899 * f2fs_quota_sync
2900 * f2fs_down_read(quota_sem)
2901 * dquot_writeback_dquots()
2902 * f2fs_dquot_commit
2903 * block_operation
2904 * f2fs_down_read(quota_sem)
2905 */
2906 f2fs_lock_op(sbi);
2907 f2fs_down_read(&sbi->quota_sem);
2908
2909 ret = f2fs_quota_sync_file(sbi, cnt);
2910
2911 f2fs_up_read(&sbi->quota_sem);
2912 f2fs_unlock_op(sbi);
2913
2914 if (!f2fs_sb_has_quota_ino(sbi))
2915 inode_unlock(dqopt->files[cnt]);
2916
2917 if (ret)
2918 break;
2919 }
2920 return ret;
2921 }
2922
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2923 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2924 const struct path *path)
2925 {
2926 struct inode *inode;
2927 int err;
2928
2929 /* if quota sysfile exists, deny enabling quota with specific file */
2930 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2931 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2932 return -EBUSY;
2933 }
2934
2935 if (path->dentry->d_sb != sb)
2936 return -EXDEV;
2937
2938 err = f2fs_quota_sync(sb, type);
2939 if (err)
2940 return err;
2941
2942 inode = d_inode(path->dentry);
2943
2944 err = filemap_fdatawrite(inode->i_mapping);
2945 if (err)
2946 return err;
2947
2948 err = filemap_fdatawait(inode->i_mapping);
2949 if (err)
2950 return err;
2951
2952 err = dquot_quota_on(sb, type, format_id, path);
2953 if (err)
2954 return err;
2955
2956 inode_lock(inode);
2957 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2958 f2fs_set_inode_flags(inode);
2959 inode_unlock(inode);
2960 f2fs_mark_inode_dirty_sync(inode, false);
2961
2962 return 0;
2963 }
2964
__f2fs_quota_off(struct super_block * sb,int type)2965 static int __f2fs_quota_off(struct super_block *sb, int type)
2966 {
2967 struct inode *inode = sb_dqopt(sb)->files[type];
2968 int err;
2969
2970 if (!inode || !igrab(inode))
2971 return dquot_quota_off(sb, type);
2972
2973 err = f2fs_quota_sync(sb, type);
2974 if (err)
2975 goto out_put;
2976
2977 err = dquot_quota_off(sb, type);
2978 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2979 goto out_put;
2980
2981 inode_lock(inode);
2982 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
2983 f2fs_set_inode_flags(inode);
2984 inode_unlock(inode);
2985 f2fs_mark_inode_dirty_sync(inode, false);
2986 out_put:
2987 iput(inode);
2988 return err;
2989 }
2990
f2fs_quota_off(struct super_block * sb,int type)2991 static int f2fs_quota_off(struct super_block *sb, int type)
2992 {
2993 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2994 int err;
2995
2996 err = __f2fs_quota_off(sb, type);
2997
2998 /*
2999 * quotactl can shutdown journalled quota, result in inconsistence
3000 * between quota record and fs data by following updates, tag the
3001 * flag to let fsck be aware of it.
3002 */
3003 if (is_journalled_quota(sbi))
3004 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3005 return err;
3006 }
3007
f2fs_quota_off_umount(struct super_block * sb)3008 void f2fs_quota_off_umount(struct super_block *sb)
3009 {
3010 int type;
3011 int err;
3012
3013 for (type = 0; type < MAXQUOTAS; type++) {
3014 err = __f2fs_quota_off(sb, type);
3015 if (err) {
3016 int ret = dquot_quota_off(sb, type);
3017
3018 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3019 type, err, ret);
3020 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3021 }
3022 }
3023 /*
3024 * In case of checkpoint=disable, we must flush quota blocks.
3025 * This can cause NULL exception for node_inode in end_io, since
3026 * put_super already dropped it.
3027 */
3028 sync_filesystem(sb);
3029 }
3030
f2fs_truncate_quota_inode_pages(struct super_block * sb)3031 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3032 {
3033 struct quota_info *dqopt = sb_dqopt(sb);
3034 int type;
3035
3036 for (type = 0; type < MAXQUOTAS; type++) {
3037 if (!dqopt->files[type])
3038 continue;
3039 f2fs_inode_synced(dqopt->files[type]);
3040 }
3041 }
3042
f2fs_dquot_commit(struct dquot * dquot)3043 static int f2fs_dquot_commit(struct dquot *dquot)
3044 {
3045 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3046 int ret;
3047
3048 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3049 ret = dquot_commit(dquot);
3050 if (ret < 0)
3051 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3052 f2fs_up_read(&sbi->quota_sem);
3053 return ret;
3054 }
3055
f2fs_dquot_acquire(struct dquot * dquot)3056 static int f2fs_dquot_acquire(struct dquot *dquot)
3057 {
3058 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3059 int ret;
3060
3061 f2fs_down_read(&sbi->quota_sem);
3062 ret = dquot_acquire(dquot);
3063 if (ret < 0)
3064 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3065 f2fs_up_read(&sbi->quota_sem);
3066 return ret;
3067 }
3068
f2fs_dquot_release(struct dquot * dquot)3069 static int f2fs_dquot_release(struct dquot *dquot)
3070 {
3071 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3072 int ret = dquot_release(dquot);
3073
3074 if (ret < 0)
3075 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3076 return ret;
3077 }
3078
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)3079 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3080 {
3081 struct super_block *sb = dquot->dq_sb;
3082 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3083 int ret = dquot_mark_dquot_dirty(dquot);
3084
3085 /* if we are using journalled quota */
3086 if (is_journalled_quota(sbi))
3087 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3088
3089 return ret;
3090 }
3091
f2fs_dquot_commit_info(struct super_block * sb,int type)3092 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3093 {
3094 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3095 int ret = dquot_commit_info(sb, type);
3096
3097 if (ret < 0)
3098 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3099 return ret;
3100 }
3101
f2fs_get_projid(struct inode * inode,kprojid_t * projid)3102 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3103 {
3104 *projid = F2FS_I(inode)->i_projid;
3105 return 0;
3106 }
3107
3108 static const struct dquot_operations f2fs_quota_operations = {
3109 .get_reserved_space = f2fs_get_reserved_space,
3110 .write_dquot = f2fs_dquot_commit,
3111 .acquire_dquot = f2fs_dquot_acquire,
3112 .release_dquot = f2fs_dquot_release,
3113 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3114 .write_info = f2fs_dquot_commit_info,
3115 .alloc_dquot = dquot_alloc,
3116 .destroy_dquot = dquot_destroy,
3117 .get_projid = f2fs_get_projid,
3118 .get_next_id = dquot_get_next_id,
3119 };
3120
3121 static const struct quotactl_ops f2fs_quotactl_ops = {
3122 .quota_on = f2fs_quota_on,
3123 .quota_off = f2fs_quota_off,
3124 .quota_sync = f2fs_quota_sync,
3125 .get_state = dquot_get_state,
3126 .set_info = dquot_set_dqinfo,
3127 .get_dqblk = dquot_get_dqblk,
3128 .set_dqblk = dquot_set_dqblk,
3129 .get_nextdqblk = dquot_get_next_dqblk,
3130 };
3131 #else
f2fs_dquot_initialize(struct inode * inode)3132 int f2fs_dquot_initialize(struct inode *inode)
3133 {
3134 return 0;
3135 }
3136
f2fs_quota_sync(struct super_block * sb,int type)3137 int f2fs_quota_sync(struct super_block *sb, int type)
3138 {
3139 return 0;
3140 }
3141
f2fs_quota_off_umount(struct super_block * sb)3142 void f2fs_quota_off_umount(struct super_block *sb)
3143 {
3144 }
3145 #endif
3146
3147 static const struct super_operations f2fs_sops = {
3148 .alloc_inode = f2fs_alloc_inode,
3149 .free_inode = f2fs_free_inode,
3150 .drop_inode = f2fs_drop_inode,
3151 .write_inode = f2fs_write_inode,
3152 .dirty_inode = f2fs_dirty_inode,
3153 .show_options = f2fs_show_options,
3154 #ifdef CONFIG_QUOTA
3155 .quota_read = f2fs_quota_read,
3156 .quota_write = f2fs_quota_write,
3157 .get_dquots = f2fs_get_dquots,
3158 #endif
3159 .evict_inode = f2fs_evict_inode,
3160 .put_super = f2fs_put_super,
3161 .sync_fs = f2fs_sync_fs,
3162 .freeze_fs = f2fs_freeze,
3163 .unfreeze_fs = f2fs_unfreeze,
3164 .statfs = f2fs_statfs,
3165 .remount_fs = f2fs_remount,
3166 .shutdown = f2fs_shutdown,
3167 };
3168
3169 #ifdef CONFIG_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)3170 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3171 {
3172 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3173 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3174 ctx, len, NULL);
3175 }
3176
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)3177 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3178 void *fs_data)
3179 {
3180 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3181
3182 /*
3183 * Encrypting the root directory is not allowed because fsck
3184 * expects lost+found directory to exist and remain unencrypted
3185 * if LOST_FOUND feature is enabled.
3186 *
3187 */
3188 if (f2fs_sb_has_lost_found(sbi) &&
3189 inode->i_ino == F2FS_ROOT_INO(sbi))
3190 return -EPERM;
3191
3192 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3193 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3194 ctx, len, fs_data, XATTR_CREATE);
3195 }
3196
f2fs_get_dummy_policy(struct super_block * sb)3197 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3198 {
3199 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3200 }
3201
f2fs_has_stable_inodes(struct super_block * sb)3202 static bool f2fs_has_stable_inodes(struct super_block *sb)
3203 {
3204 return true;
3205 }
3206
f2fs_get_devices(struct super_block * sb,unsigned int * num_devs)3207 static struct block_device **f2fs_get_devices(struct super_block *sb,
3208 unsigned int *num_devs)
3209 {
3210 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3211 struct block_device **devs;
3212 int i;
3213
3214 if (!f2fs_is_multi_device(sbi))
3215 return NULL;
3216
3217 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3218 if (!devs)
3219 return ERR_PTR(-ENOMEM);
3220
3221 for (i = 0; i < sbi->s_ndevs; i++)
3222 devs[i] = FDEV(i).bdev;
3223 *num_devs = sbi->s_ndevs;
3224 return devs;
3225 }
3226
3227 static const struct fscrypt_operations f2fs_cryptops = {
3228 .needs_bounce_pages = 1,
3229 .has_32bit_inodes = 1,
3230 .supports_subblock_data_units = 1,
3231 .legacy_key_prefix = "f2fs:",
3232 .get_context = f2fs_get_context,
3233 .set_context = f2fs_set_context,
3234 .get_dummy_policy = f2fs_get_dummy_policy,
3235 .empty_dir = f2fs_empty_dir,
3236 .has_stable_inodes = f2fs_has_stable_inodes,
3237 .get_devices = f2fs_get_devices,
3238 };
3239 #endif
3240
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)3241 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3242 u64 ino, u32 generation)
3243 {
3244 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3245 struct inode *inode;
3246
3247 if (f2fs_check_nid_range(sbi, ino))
3248 return ERR_PTR(-ESTALE);
3249
3250 /*
3251 * f2fs_iget isn't quite right if the inode is currently unallocated!
3252 * However f2fs_iget currently does appropriate checks to handle stale
3253 * inodes so everything is OK.
3254 */
3255 inode = f2fs_iget(sb, ino);
3256 if (IS_ERR(inode))
3257 return ERR_CAST(inode);
3258 if (unlikely(generation && inode->i_generation != generation)) {
3259 /* we didn't find the right inode.. */
3260 iput(inode);
3261 return ERR_PTR(-ESTALE);
3262 }
3263 return inode;
3264 }
3265
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3266 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3267 int fh_len, int fh_type)
3268 {
3269 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3270 f2fs_nfs_get_inode);
3271 }
3272
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3273 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3274 int fh_len, int fh_type)
3275 {
3276 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3277 f2fs_nfs_get_inode);
3278 }
3279
3280 static const struct export_operations f2fs_export_ops = {
3281 .encode_fh = generic_encode_ino32_fh,
3282 .fh_to_dentry = f2fs_fh_to_dentry,
3283 .fh_to_parent = f2fs_fh_to_parent,
3284 .get_parent = f2fs_get_parent,
3285 };
3286
max_file_blocks(struct inode * inode)3287 loff_t max_file_blocks(struct inode *inode)
3288 {
3289 loff_t result = 0;
3290 loff_t leaf_count;
3291
3292 /*
3293 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3294 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3295 * space in inode.i_addr, it will be more safe to reassign
3296 * result as zero.
3297 */
3298
3299 if (inode && f2fs_compressed_file(inode))
3300 leaf_count = ADDRS_PER_BLOCK(inode);
3301 else
3302 leaf_count = DEF_ADDRS_PER_BLOCK;
3303
3304 /* two direct node blocks */
3305 result += (leaf_count * 2);
3306
3307 /* two indirect node blocks */
3308 leaf_count *= NIDS_PER_BLOCK;
3309 result += (leaf_count * 2);
3310
3311 /* one double indirect node block */
3312 leaf_count *= NIDS_PER_BLOCK;
3313 result += leaf_count;
3314
3315 /*
3316 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with
3317 * a 4K crypto data unit, we must restrict the max filesize to what can
3318 * fit within U32_MAX + 1 data units.
3319 */
3320
3321 result = min(result, (((loff_t)U32_MAX + 1) * 4096) >> F2FS_BLKSIZE_BITS);
3322
3323 return result;
3324 }
3325
__f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)3326 static int __f2fs_commit_super(struct buffer_head *bh,
3327 struct f2fs_super_block *super)
3328 {
3329 lock_buffer(bh);
3330 if (super)
3331 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3332 set_buffer_dirty(bh);
3333 unlock_buffer(bh);
3334
3335 /* it's rare case, we can do fua all the time */
3336 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3337 }
3338
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)3339 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3340 struct buffer_head *bh)
3341 {
3342 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3343 (bh->b_data + F2FS_SUPER_OFFSET);
3344 struct super_block *sb = sbi->sb;
3345 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3346 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3347 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3348 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3349 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3350 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3351 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3352 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3353 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3354 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3355 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3356 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3357 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3358 u64 main_end_blkaddr = main_blkaddr +
3359 (segment_count_main << log_blocks_per_seg);
3360 u64 seg_end_blkaddr = segment0_blkaddr +
3361 (segment_count << log_blocks_per_seg);
3362
3363 if (segment0_blkaddr != cp_blkaddr) {
3364 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3365 segment0_blkaddr, cp_blkaddr);
3366 return true;
3367 }
3368
3369 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3370 sit_blkaddr) {
3371 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3372 cp_blkaddr, sit_blkaddr,
3373 segment_count_ckpt << log_blocks_per_seg);
3374 return true;
3375 }
3376
3377 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3378 nat_blkaddr) {
3379 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3380 sit_blkaddr, nat_blkaddr,
3381 segment_count_sit << log_blocks_per_seg);
3382 return true;
3383 }
3384
3385 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3386 ssa_blkaddr) {
3387 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3388 nat_blkaddr, ssa_blkaddr,
3389 segment_count_nat << log_blocks_per_seg);
3390 return true;
3391 }
3392
3393 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3394 main_blkaddr) {
3395 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3396 ssa_blkaddr, main_blkaddr,
3397 segment_count_ssa << log_blocks_per_seg);
3398 return true;
3399 }
3400
3401 if (main_end_blkaddr > seg_end_blkaddr) {
3402 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3403 main_blkaddr, seg_end_blkaddr,
3404 segment_count_main << log_blocks_per_seg);
3405 return true;
3406 } else if (main_end_blkaddr < seg_end_blkaddr) {
3407 int err = 0;
3408 char *res;
3409
3410 /* fix in-memory information all the time */
3411 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3412 segment0_blkaddr) >> log_blocks_per_seg);
3413
3414 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3415 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3416 res = "internally";
3417 } else {
3418 err = __f2fs_commit_super(bh, NULL);
3419 res = err ? "failed" : "done";
3420 }
3421 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3422 res, main_blkaddr, seg_end_blkaddr,
3423 segment_count_main << log_blocks_per_seg);
3424 if (err)
3425 return true;
3426 }
3427 return false;
3428 }
3429
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)3430 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3431 struct buffer_head *bh)
3432 {
3433 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3434 block_t total_sections, blocks_per_seg;
3435 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3436 (bh->b_data + F2FS_SUPER_OFFSET);
3437 size_t crc_offset = 0;
3438 __u32 crc = 0;
3439
3440 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3441 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3442 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3443 return -EINVAL;
3444 }
3445
3446 /* Check checksum_offset and crc in superblock */
3447 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3448 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3449 if (crc_offset !=
3450 offsetof(struct f2fs_super_block, crc)) {
3451 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3452 crc_offset);
3453 return -EFSCORRUPTED;
3454 }
3455 crc = le32_to_cpu(raw_super->crc);
3456 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3457 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3458 return -EFSCORRUPTED;
3459 }
3460 }
3461
3462 /* only support block_size equals to PAGE_SIZE */
3463 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3464 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3465 le32_to_cpu(raw_super->log_blocksize),
3466 F2FS_BLKSIZE_BITS);
3467 return -EFSCORRUPTED;
3468 }
3469
3470 /* check log blocks per segment */
3471 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3472 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3473 le32_to_cpu(raw_super->log_blocks_per_seg));
3474 return -EFSCORRUPTED;
3475 }
3476
3477 /* Currently, support 512/1024/2048/4096/16K bytes sector size */
3478 if (le32_to_cpu(raw_super->log_sectorsize) >
3479 F2FS_MAX_LOG_SECTOR_SIZE ||
3480 le32_to_cpu(raw_super->log_sectorsize) <
3481 F2FS_MIN_LOG_SECTOR_SIZE) {
3482 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3483 le32_to_cpu(raw_super->log_sectorsize));
3484 return -EFSCORRUPTED;
3485 }
3486 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3487 le32_to_cpu(raw_super->log_sectorsize) !=
3488 F2FS_MAX_LOG_SECTOR_SIZE) {
3489 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3490 le32_to_cpu(raw_super->log_sectors_per_block),
3491 le32_to_cpu(raw_super->log_sectorsize));
3492 return -EFSCORRUPTED;
3493 }
3494
3495 segment_count = le32_to_cpu(raw_super->segment_count);
3496 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3497 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3498 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3499 total_sections = le32_to_cpu(raw_super->section_count);
3500
3501 /* blocks_per_seg should be 512, given the above check */
3502 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3503
3504 if (segment_count > F2FS_MAX_SEGMENT ||
3505 segment_count < F2FS_MIN_SEGMENTS) {
3506 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3507 return -EFSCORRUPTED;
3508 }
3509
3510 if (total_sections > segment_count_main || total_sections < 1 ||
3511 segs_per_sec > segment_count || !segs_per_sec) {
3512 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3513 segment_count, total_sections, segs_per_sec);
3514 return -EFSCORRUPTED;
3515 }
3516
3517 if (segment_count_main != total_sections * segs_per_sec) {
3518 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3519 segment_count_main, total_sections, segs_per_sec);
3520 return -EFSCORRUPTED;
3521 }
3522
3523 if ((segment_count / segs_per_sec) < total_sections) {
3524 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3525 segment_count, segs_per_sec, total_sections);
3526 return -EFSCORRUPTED;
3527 }
3528
3529 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3530 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3531 segment_count, le64_to_cpu(raw_super->block_count));
3532 return -EFSCORRUPTED;
3533 }
3534
3535 if (RDEV(0).path[0]) {
3536 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3537 int i = 1;
3538
3539 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3540 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3541 i++;
3542 }
3543 if (segment_count != dev_seg_count) {
3544 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3545 segment_count, dev_seg_count);
3546 return -EFSCORRUPTED;
3547 }
3548 } else {
3549 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3550 !bdev_is_zoned(sbi->sb->s_bdev)) {
3551 f2fs_info(sbi, "Zoned block device path is missing");
3552 return -EFSCORRUPTED;
3553 }
3554 }
3555
3556 if (secs_per_zone > total_sections || !secs_per_zone) {
3557 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3558 secs_per_zone, total_sections);
3559 return -EFSCORRUPTED;
3560 }
3561 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3562 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3563 (le32_to_cpu(raw_super->extension_count) +
3564 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3565 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3566 le32_to_cpu(raw_super->extension_count),
3567 raw_super->hot_ext_count,
3568 F2FS_MAX_EXTENSION);
3569 return -EFSCORRUPTED;
3570 }
3571
3572 if (le32_to_cpu(raw_super->cp_payload) >=
3573 (blocks_per_seg - F2FS_CP_PACKS -
3574 NR_CURSEG_PERSIST_TYPE)) {
3575 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3576 le32_to_cpu(raw_super->cp_payload),
3577 blocks_per_seg - F2FS_CP_PACKS -
3578 NR_CURSEG_PERSIST_TYPE);
3579 return -EFSCORRUPTED;
3580 }
3581
3582 /* check reserved ino info */
3583 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3584 le32_to_cpu(raw_super->meta_ino) != 2 ||
3585 le32_to_cpu(raw_super->root_ino) != 3) {
3586 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3587 le32_to_cpu(raw_super->node_ino),
3588 le32_to_cpu(raw_super->meta_ino),
3589 le32_to_cpu(raw_super->root_ino));
3590 return -EFSCORRUPTED;
3591 }
3592
3593 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3594 if (sanity_check_area_boundary(sbi, bh))
3595 return -EFSCORRUPTED;
3596
3597 return 0;
3598 }
3599
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3600 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3601 {
3602 unsigned int total, fsmeta;
3603 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3604 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3605 unsigned int ovp_segments, reserved_segments;
3606 unsigned int main_segs, blocks_per_seg;
3607 unsigned int sit_segs, nat_segs;
3608 unsigned int sit_bitmap_size, nat_bitmap_size;
3609 unsigned int log_blocks_per_seg;
3610 unsigned int segment_count_main;
3611 unsigned int cp_pack_start_sum, cp_payload;
3612 block_t user_block_count, valid_user_blocks;
3613 block_t avail_node_count, valid_node_count;
3614 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3615 int i, j;
3616
3617 total = le32_to_cpu(raw_super->segment_count);
3618 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3619 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3620 fsmeta += sit_segs;
3621 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3622 fsmeta += nat_segs;
3623 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3624 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3625
3626 if (unlikely(fsmeta >= total))
3627 return 1;
3628
3629 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3630 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3631
3632 if (!f2fs_sb_has_readonly(sbi) &&
3633 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3634 ovp_segments == 0 || reserved_segments == 0)) {
3635 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3636 return 1;
3637 }
3638 user_block_count = le64_to_cpu(ckpt->user_block_count);
3639 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3640 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3641 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3642 if (!user_block_count || user_block_count >=
3643 segment_count_main << log_blocks_per_seg) {
3644 f2fs_err(sbi, "Wrong user_block_count: %u",
3645 user_block_count);
3646 return 1;
3647 }
3648
3649 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3650 if (valid_user_blocks > user_block_count) {
3651 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3652 valid_user_blocks, user_block_count);
3653 return 1;
3654 }
3655
3656 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3657 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3658 if (valid_node_count > avail_node_count) {
3659 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3660 valid_node_count, avail_node_count);
3661 return 1;
3662 }
3663
3664 main_segs = le32_to_cpu(raw_super->segment_count_main);
3665 blocks_per_seg = BLKS_PER_SEG(sbi);
3666
3667 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3668 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3669 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3670 return 1;
3671
3672 if (f2fs_sb_has_readonly(sbi))
3673 goto check_data;
3674
3675 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3676 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3677 le32_to_cpu(ckpt->cur_node_segno[j])) {
3678 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3679 i, j,
3680 le32_to_cpu(ckpt->cur_node_segno[i]));
3681 return 1;
3682 }
3683 }
3684 }
3685 check_data:
3686 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3687 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3688 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3689 return 1;
3690
3691 if (f2fs_sb_has_readonly(sbi))
3692 goto skip_cross;
3693
3694 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3695 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3696 le32_to_cpu(ckpt->cur_data_segno[j])) {
3697 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3698 i, j,
3699 le32_to_cpu(ckpt->cur_data_segno[i]));
3700 return 1;
3701 }
3702 }
3703 }
3704 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3705 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3706 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3707 le32_to_cpu(ckpt->cur_data_segno[j])) {
3708 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3709 i, j,
3710 le32_to_cpu(ckpt->cur_node_segno[i]));
3711 return 1;
3712 }
3713 }
3714 }
3715 skip_cross:
3716 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3717 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3718
3719 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3720 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3721 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3722 sit_bitmap_size, nat_bitmap_size);
3723 return 1;
3724 }
3725
3726 cp_pack_start_sum = __start_sum_addr(sbi);
3727 cp_payload = __cp_payload(sbi);
3728 if (cp_pack_start_sum < cp_payload + 1 ||
3729 cp_pack_start_sum > blocks_per_seg - 1 -
3730 NR_CURSEG_PERSIST_TYPE) {
3731 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3732 cp_pack_start_sum);
3733 return 1;
3734 }
3735
3736 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3737 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3738 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3739 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3740 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3741 le32_to_cpu(ckpt->checksum_offset));
3742 return 1;
3743 }
3744
3745 nat_blocks = nat_segs << log_blocks_per_seg;
3746 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3747 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3748 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3749 (cp_payload + F2FS_CP_PACKS +
3750 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3751 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3752 cp_payload, nat_bits_blocks);
3753 return 1;
3754 }
3755
3756 if (unlikely(f2fs_cp_error(sbi))) {
3757 f2fs_err(sbi, "A bug case: need to run fsck");
3758 return 1;
3759 }
3760 return 0;
3761 }
3762
init_sb_info(struct f2fs_sb_info * sbi)3763 static void init_sb_info(struct f2fs_sb_info *sbi)
3764 {
3765 struct f2fs_super_block *raw_super = sbi->raw_super;
3766 int i;
3767
3768 sbi->log_sectors_per_block =
3769 le32_to_cpu(raw_super->log_sectors_per_block);
3770 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3771 sbi->blocksize = BIT(sbi->log_blocksize);
3772 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3773 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3774 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3775 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3776 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3777 sbi->total_node_count = SEGS_TO_BLKS(sbi,
3778 ((le32_to_cpu(raw_super->segment_count_nat) / 2) *
3779 NAT_ENTRY_PER_BLOCK));
3780 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3781 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3782 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3783 sbi->cur_victim_sec = NULL_SECNO;
3784 sbi->gc_mode = GC_NORMAL;
3785 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3786 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3787 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3788 sbi->migration_granularity = SEGS_PER_SEC(sbi);
3789 sbi->seq_file_ra_mul = MIN_RA_MUL;
3790 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3791 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3792 spin_lock_init(&sbi->gc_remaining_trials_lock);
3793 atomic64_set(&sbi->current_atomic_write, 0);
3794
3795 sbi->dir_level = DEF_DIR_LEVEL;
3796 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3797 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3798 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3799 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3800 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3801 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3802 DEF_UMOUNT_DISCARD_TIMEOUT;
3803 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3804
3805 for (i = 0; i < NR_COUNT_TYPE; i++)
3806 atomic_set(&sbi->nr_pages[i], 0);
3807
3808 for (i = 0; i < META; i++)
3809 atomic_set(&sbi->wb_sync_req[i], 0);
3810
3811 INIT_LIST_HEAD(&sbi->s_list);
3812 mutex_init(&sbi->umount_mutex);
3813 init_f2fs_rwsem(&sbi->io_order_lock);
3814 spin_lock_init(&sbi->cp_lock);
3815
3816 sbi->dirty_device = 0;
3817 spin_lock_init(&sbi->dev_lock);
3818
3819 init_f2fs_rwsem(&sbi->sb_lock);
3820 init_f2fs_rwsem(&sbi->pin_sem);
3821 }
3822
init_percpu_info(struct f2fs_sb_info * sbi)3823 static int init_percpu_info(struct f2fs_sb_info *sbi)
3824 {
3825 int err;
3826
3827 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3828 if (err)
3829 return err;
3830
3831 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3832 if (err)
3833 goto err_valid_block;
3834
3835 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3836 GFP_KERNEL);
3837 if (err)
3838 goto err_node_block;
3839 return 0;
3840
3841 err_node_block:
3842 percpu_counter_destroy(&sbi->rf_node_block_count);
3843 err_valid_block:
3844 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3845 return err;
3846 }
3847
3848 #ifdef CONFIG_BLK_DEV_ZONED
3849
3850 struct f2fs_report_zones_args {
3851 struct f2fs_sb_info *sbi;
3852 struct f2fs_dev_info *dev;
3853 };
3854
f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3855 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3856 void *data)
3857 {
3858 struct f2fs_report_zones_args *rz_args = data;
3859 block_t unusable_blocks = (zone->len - zone->capacity) >>
3860 F2FS_LOG_SECTORS_PER_BLOCK;
3861
3862 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3863 return 0;
3864
3865 set_bit(idx, rz_args->dev->blkz_seq);
3866 if (!rz_args->sbi->unusable_blocks_per_sec) {
3867 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3868 return 0;
3869 }
3870 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3871 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3872 return -EINVAL;
3873 }
3874 return 0;
3875 }
3876
init_blkz_info(struct f2fs_sb_info * sbi,int devi)3877 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3878 {
3879 struct block_device *bdev = FDEV(devi).bdev;
3880 sector_t nr_sectors = bdev_nr_sectors(bdev);
3881 struct f2fs_report_zones_args rep_zone_arg;
3882 u64 zone_sectors;
3883 unsigned int max_open_zones;
3884 int ret;
3885
3886 if (!f2fs_sb_has_blkzoned(sbi))
3887 return 0;
3888
3889 if (bdev_is_zoned(FDEV(devi).bdev)) {
3890 max_open_zones = bdev_max_open_zones(bdev);
3891 if (max_open_zones && (max_open_zones < sbi->max_open_zones))
3892 sbi->max_open_zones = max_open_zones;
3893 if (sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
3894 f2fs_err(sbi,
3895 "zoned: max open zones %u is too small, need at least %u open zones",
3896 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
3897 return -EINVAL;
3898 }
3899 }
3900
3901 zone_sectors = bdev_zone_sectors(bdev);
3902 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3903 SECTOR_TO_BLOCK(zone_sectors))
3904 return -EINVAL;
3905 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3906 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3907 sbi->blocks_per_blkz);
3908 if (nr_sectors & (zone_sectors - 1))
3909 FDEV(devi).nr_blkz++;
3910
3911 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3912 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3913 * sizeof(unsigned long),
3914 GFP_KERNEL);
3915 if (!FDEV(devi).blkz_seq)
3916 return -ENOMEM;
3917
3918 rep_zone_arg.sbi = sbi;
3919 rep_zone_arg.dev = &FDEV(devi);
3920
3921 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3922 &rep_zone_arg);
3923 if (ret < 0)
3924 return ret;
3925 return 0;
3926 }
3927 #endif
3928
3929 /*
3930 * Read f2fs raw super block.
3931 * Because we have two copies of super block, so read both of them
3932 * to get the first valid one. If any one of them is broken, we pass
3933 * them recovery flag back to the caller.
3934 */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3935 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3936 struct f2fs_super_block **raw_super,
3937 int *valid_super_block, int *recovery)
3938 {
3939 struct super_block *sb = sbi->sb;
3940 int block;
3941 struct buffer_head *bh;
3942 struct f2fs_super_block *super;
3943 int err = 0;
3944
3945 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3946 if (!super)
3947 return -ENOMEM;
3948
3949 for (block = 0; block < 2; block++) {
3950 bh = sb_bread(sb, block);
3951 if (!bh) {
3952 f2fs_err(sbi, "Unable to read %dth superblock",
3953 block + 1);
3954 err = -EIO;
3955 *recovery = 1;
3956 continue;
3957 }
3958
3959 /* sanity checking of raw super */
3960 err = sanity_check_raw_super(sbi, bh);
3961 if (err) {
3962 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3963 block + 1);
3964 brelse(bh);
3965 *recovery = 1;
3966 continue;
3967 }
3968
3969 if (!*raw_super) {
3970 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3971 sizeof(*super));
3972 *valid_super_block = block;
3973 *raw_super = super;
3974 }
3975 brelse(bh);
3976 }
3977
3978 /* No valid superblock */
3979 if (!*raw_super)
3980 kfree(super);
3981 else
3982 err = 0;
3983
3984 return err;
3985 }
3986
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)3987 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3988 {
3989 struct buffer_head *bh;
3990 __u32 crc = 0;
3991 int err;
3992
3993 if ((recover && f2fs_readonly(sbi->sb)) ||
3994 f2fs_hw_is_readonly(sbi)) {
3995 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3996 return -EROFS;
3997 }
3998
3999 /* we should update superblock crc here */
4000 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
4001 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
4002 offsetof(struct f2fs_super_block, crc));
4003 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
4004 }
4005
4006 /* write back-up superblock first */
4007 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
4008 if (!bh)
4009 return -EIO;
4010 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4011 brelse(bh);
4012
4013 /* if we are in recovery path, skip writing valid superblock */
4014 if (recover || err)
4015 return err;
4016
4017 /* write current valid superblock */
4018 bh = sb_bread(sbi->sb, sbi->valid_super_block);
4019 if (!bh)
4020 return -EIO;
4021 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4022 brelse(bh);
4023 return err;
4024 }
4025
save_stop_reason(struct f2fs_sb_info * sbi,unsigned char reason)4026 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4027 {
4028 unsigned long flags;
4029
4030 spin_lock_irqsave(&sbi->error_lock, flags);
4031 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4032 sbi->stop_reason[reason]++;
4033 spin_unlock_irqrestore(&sbi->error_lock, flags);
4034 }
4035
f2fs_record_stop_reason(struct f2fs_sb_info * sbi)4036 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4037 {
4038 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4039 unsigned long flags;
4040 int err;
4041
4042 f2fs_down_write(&sbi->sb_lock);
4043
4044 spin_lock_irqsave(&sbi->error_lock, flags);
4045 if (sbi->error_dirty) {
4046 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4047 MAX_F2FS_ERRORS);
4048 sbi->error_dirty = false;
4049 }
4050 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4051 spin_unlock_irqrestore(&sbi->error_lock, flags);
4052
4053 err = f2fs_commit_super(sbi, false);
4054
4055 f2fs_up_write(&sbi->sb_lock);
4056 if (err)
4057 f2fs_err_ratelimited(sbi,
4058 "f2fs_commit_super fails to record stop_reason, err:%d",
4059 err);
4060 }
4061
f2fs_save_errors(struct f2fs_sb_info * sbi,unsigned char flag)4062 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4063 {
4064 unsigned long flags;
4065
4066 spin_lock_irqsave(&sbi->error_lock, flags);
4067 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4068 set_bit(flag, (unsigned long *)sbi->errors);
4069 sbi->error_dirty = true;
4070 }
4071 spin_unlock_irqrestore(&sbi->error_lock, flags);
4072 }
4073
f2fs_update_errors(struct f2fs_sb_info * sbi)4074 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4075 {
4076 unsigned long flags;
4077 bool need_update = false;
4078
4079 spin_lock_irqsave(&sbi->error_lock, flags);
4080 if (sbi->error_dirty) {
4081 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4082 MAX_F2FS_ERRORS);
4083 sbi->error_dirty = false;
4084 need_update = true;
4085 }
4086 spin_unlock_irqrestore(&sbi->error_lock, flags);
4087
4088 return need_update;
4089 }
4090
f2fs_record_errors(struct f2fs_sb_info * sbi,unsigned char error)4091 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4092 {
4093 int err;
4094
4095 f2fs_down_write(&sbi->sb_lock);
4096
4097 if (!f2fs_update_errors(sbi))
4098 goto out_unlock;
4099
4100 err = f2fs_commit_super(sbi, false);
4101 if (err)
4102 f2fs_err_ratelimited(sbi,
4103 "f2fs_commit_super fails to record errors:%u, err:%d",
4104 error, err);
4105 out_unlock:
4106 f2fs_up_write(&sbi->sb_lock);
4107 }
4108
f2fs_handle_error(struct f2fs_sb_info * sbi,unsigned char error)4109 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4110 {
4111 f2fs_save_errors(sbi, error);
4112 f2fs_record_errors(sbi, error);
4113 }
4114
f2fs_handle_error_async(struct f2fs_sb_info * sbi,unsigned char error)4115 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4116 {
4117 f2fs_save_errors(sbi, error);
4118
4119 if (!sbi->error_dirty)
4120 return;
4121 if (!test_bit(error, (unsigned long *)sbi->errors))
4122 return;
4123 schedule_work(&sbi->s_error_work);
4124 }
4125
system_going_down(void)4126 static bool system_going_down(void)
4127 {
4128 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4129 || system_state == SYSTEM_RESTART;
4130 }
4131
f2fs_handle_critical_error(struct f2fs_sb_info * sbi,unsigned char reason,bool irq_context)4132 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason,
4133 bool irq_context)
4134 {
4135 struct super_block *sb = sbi->sb;
4136 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4137 bool continue_fs = !shutdown &&
4138 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4139
4140 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4141
4142 if (!f2fs_hw_is_readonly(sbi)) {
4143 save_stop_reason(sbi, reason);
4144
4145 if (irq_context && !shutdown)
4146 schedule_work(&sbi->s_error_work);
4147 else
4148 f2fs_record_stop_reason(sbi);
4149 }
4150
4151 /*
4152 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4153 * could panic during 'reboot -f' as the underlying device got already
4154 * disabled.
4155 */
4156 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4157 !shutdown && !system_going_down() &&
4158 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4159 panic("F2FS-fs (device %s): panic forced after error\n",
4160 sb->s_id);
4161
4162 if (shutdown)
4163 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4164
4165 /*
4166 * Continue filesystem operators if errors=continue. Should not set
4167 * RO by shutdown, since RO bypasses thaw_super which can hang the
4168 * system.
4169 */
4170 if (continue_fs || f2fs_readonly(sb) || shutdown) {
4171 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason);
4172 return;
4173 }
4174
4175 f2fs_warn(sbi, "Remounting filesystem read-only");
4176 /*
4177 * Make sure updated value of ->s_mount_flags will be visible before
4178 * ->s_flags update
4179 */
4180 smp_wmb();
4181 sb->s_flags |= SB_RDONLY;
4182 }
4183
f2fs_record_error_work(struct work_struct * work)4184 static void f2fs_record_error_work(struct work_struct *work)
4185 {
4186 struct f2fs_sb_info *sbi = container_of(work,
4187 struct f2fs_sb_info, s_error_work);
4188
4189 f2fs_record_stop_reason(sbi);
4190 }
4191
f2fs_scan_devices(struct f2fs_sb_info * sbi)4192 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4193 {
4194 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4195 unsigned int max_devices = MAX_DEVICES;
4196 unsigned int logical_blksize;
4197 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4198 int i;
4199
4200 /* Initialize single device information */
4201 if (!RDEV(0).path[0]) {
4202 if (!bdev_is_zoned(sbi->sb->s_bdev))
4203 return 0;
4204 max_devices = 1;
4205 }
4206
4207 /*
4208 * Initialize multiple devices information, or single
4209 * zoned block device information.
4210 */
4211 sbi->devs = f2fs_kzalloc(sbi,
4212 array_size(max_devices,
4213 sizeof(struct f2fs_dev_info)),
4214 GFP_KERNEL);
4215 if (!sbi->devs)
4216 return -ENOMEM;
4217
4218 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4219 sbi->aligned_blksize = true;
4220 #ifdef CONFIG_BLK_DEV_ZONED
4221 sbi->max_open_zones = UINT_MAX;
4222 #endif
4223
4224 for (i = 0; i < max_devices; i++) {
4225 if (i == 0)
4226 FDEV(0).bdev_file = sbi->sb->s_bdev_file;
4227 else if (!RDEV(i).path[0])
4228 break;
4229
4230 if (max_devices > 1) {
4231 /* Multi-device mount */
4232 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4233 FDEV(i).total_segments =
4234 le32_to_cpu(RDEV(i).total_segments);
4235 if (i == 0) {
4236 FDEV(i).start_blk = 0;
4237 FDEV(i).end_blk = FDEV(i).start_blk +
4238 SEGS_TO_BLKS(sbi,
4239 FDEV(i).total_segments) - 1 +
4240 le32_to_cpu(raw_super->segment0_blkaddr);
4241 } else {
4242 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4243 FDEV(i).end_blk = FDEV(i).start_blk +
4244 SEGS_TO_BLKS(sbi,
4245 FDEV(i).total_segments) - 1;
4246 FDEV(i).bdev_file = bdev_file_open_by_path(
4247 FDEV(i).path, mode, sbi->sb, NULL);
4248 }
4249 }
4250 if (IS_ERR(FDEV(i).bdev_file))
4251 return PTR_ERR(FDEV(i).bdev_file);
4252
4253 FDEV(i).bdev = file_bdev(FDEV(i).bdev_file);
4254 /* to release errored devices */
4255 sbi->s_ndevs = i + 1;
4256
4257 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4258 sbi->aligned_blksize = false;
4259
4260 #ifdef CONFIG_BLK_DEV_ZONED
4261 if (bdev_is_zoned(FDEV(i).bdev)) {
4262 if (!f2fs_sb_has_blkzoned(sbi)) {
4263 f2fs_err(sbi, "Zoned block device feature not enabled");
4264 return -EINVAL;
4265 }
4266 if (init_blkz_info(sbi, i)) {
4267 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4268 return -EINVAL;
4269 }
4270 if (max_devices == 1)
4271 break;
4272 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)",
4273 i, FDEV(i).path,
4274 FDEV(i).total_segments,
4275 FDEV(i).start_blk, FDEV(i).end_blk);
4276 continue;
4277 }
4278 #endif
4279 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4280 i, FDEV(i).path,
4281 FDEV(i).total_segments,
4282 FDEV(i).start_blk, FDEV(i).end_blk);
4283 }
4284 return 0;
4285 }
4286
f2fs_setup_casefold(struct f2fs_sb_info * sbi)4287 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4288 {
4289 #if IS_ENABLED(CONFIG_UNICODE)
4290 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4291 const struct f2fs_sb_encodings *encoding_info;
4292 struct unicode_map *encoding;
4293 __u16 encoding_flags;
4294
4295 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4296 if (!encoding_info) {
4297 f2fs_err(sbi,
4298 "Encoding requested by superblock is unknown");
4299 return -EINVAL;
4300 }
4301
4302 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4303 encoding = utf8_load(encoding_info->version);
4304 if (IS_ERR(encoding)) {
4305 f2fs_err(sbi,
4306 "can't mount with superblock charset: %s-%u.%u.%u "
4307 "not supported by the kernel. flags: 0x%x.",
4308 encoding_info->name,
4309 unicode_major(encoding_info->version),
4310 unicode_minor(encoding_info->version),
4311 unicode_rev(encoding_info->version),
4312 encoding_flags);
4313 return PTR_ERR(encoding);
4314 }
4315 f2fs_info(sbi, "Using encoding defined by superblock: "
4316 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4317 unicode_major(encoding_info->version),
4318 unicode_minor(encoding_info->version),
4319 unicode_rev(encoding_info->version),
4320 encoding_flags);
4321
4322 sbi->sb->s_encoding = encoding;
4323 sbi->sb->s_encoding_flags = encoding_flags;
4324 }
4325 #else
4326 if (f2fs_sb_has_casefold(sbi)) {
4327 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4328 return -EINVAL;
4329 }
4330 #endif
4331 return 0;
4332 }
4333
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)4334 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4335 {
4336 /* adjust parameters according to the volume size */
4337 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4338 if (f2fs_block_unit_discard(sbi))
4339 SM_I(sbi)->dcc_info->discard_granularity =
4340 MIN_DISCARD_GRANULARITY;
4341 if (!f2fs_lfs_mode(sbi))
4342 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4343 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4344 }
4345
4346 sbi->readdir_ra = true;
4347 }
4348
f2fs_fill_super(struct super_block * sb,void * data,int silent)4349 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4350 {
4351 struct f2fs_sb_info *sbi;
4352 struct f2fs_super_block *raw_super;
4353 struct inode *root;
4354 int err;
4355 bool skip_recovery = false, need_fsck = false;
4356 char *options = NULL;
4357 int recovery, i, valid_super_block;
4358 struct curseg_info *seg_i;
4359 int retry_cnt = 1;
4360 #ifdef CONFIG_QUOTA
4361 bool quota_enabled = false;
4362 #endif
4363
4364 try_onemore:
4365 err = -EINVAL;
4366 raw_super = NULL;
4367 valid_super_block = -1;
4368 recovery = 0;
4369
4370 /* allocate memory for f2fs-specific super block info */
4371 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4372 if (!sbi)
4373 return -ENOMEM;
4374
4375 sbi->sb = sb;
4376
4377 /* initialize locks within allocated memory */
4378 init_f2fs_rwsem(&sbi->gc_lock);
4379 mutex_init(&sbi->writepages);
4380 init_f2fs_rwsem(&sbi->cp_global_sem);
4381 init_f2fs_rwsem(&sbi->node_write);
4382 init_f2fs_rwsem(&sbi->node_change);
4383 spin_lock_init(&sbi->stat_lock);
4384 init_f2fs_rwsem(&sbi->cp_rwsem);
4385 init_f2fs_rwsem(&sbi->quota_sem);
4386 init_waitqueue_head(&sbi->cp_wait);
4387 spin_lock_init(&sbi->error_lock);
4388
4389 for (i = 0; i < NR_INODE_TYPE; i++) {
4390 INIT_LIST_HEAD(&sbi->inode_list[i]);
4391 spin_lock_init(&sbi->inode_lock[i]);
4392 }
4393 mutex_init(&sbi->flush_lock);
4394
4395 /* Load the checksum driver */
4396 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4397 if (IS_ERR(sbi->s_chksum_driver)) {
4398 f2fs_err(sbi, "Cannot load crc32 driver.");
4399 err = PTR_ERR(sbi->s_chksum_driver);
4400 sbi->s_chksum_driver = NULL;
4401 goto free_sbi;
4402 }
4403
4404 /* set a block size */
4405 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4406 f2fs_err(sbi, "unable to set blocksize");
4407 goto free_sbi;
4408 }
4409
4410 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4411 &recovery);
4412 if (err)
4413 goto free_sbi;
4414
4415 sb->s_fs_info = sbi;
4416 sbi->raw_super = raw_super;
4417
4418 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4419 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4420 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4421
4422 /* precompute checksum seed for metadata */
4423 if (f2fs_sb_has_inode_chksum(sbi))
4424 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4425 sizeof(raw_super->uuid));
4426
4427 default_options(sbi, false);
4428 /* parse mount options */
4429 options = kstrdup((const char *)data, GFP_KERNEL);
4430 if (data && !options) {
4431 err = -ENOMEM;
4432 goto free_sb_buf;
4433 }
4434
4435 err = parse_options(sb, options, false);
4436 if (err)
4437 goto free_options;
4438
4439 sb->s_maxbytes = max_file_blocks(NULL) <<
4440 le32_to_cpu(raw_super->log_blocksize);
4441 sb->s_max_links = F2FS_LINK_MAX;
4442
4443 err = f2fs_setup_casefold(sbi);
4444 if (err)
4445 goto free_options;
4446
4447 #ifdef CONFIG_QUOTA
4448 sb->dq_op = &f2fs_quota_operations;
4449 sb->s_qcop = &f2fs_quotactl_ops;
4450 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4451
4452 if (f2fs_sb_has_quota_ino(sbi)) {
4453 for (i = 0; i < MAXQUOTAS; i++) {
4454 if (f2fs_qf_ino(sbi->sb, i))
4455 sbi->nquota_files++;
4456 }
4457 }
4458 #endif
4459
4460 sb->s_op = &f2fs_sops;
4461 #ifdef CONFIG_FS_ENCRYPTION
4462 sb->s_cop = &f2fs_cryptops;
4463 #endif
4464 #ifdef CONFIG_FS_VERITY
4465 sb->s_vop = &f2fs_verityops;
4466 #endif
4467 sb->s_xattr = f2fs_xattr_handlers;
4468 sb->s_export_op = &f2fs_export_ops;
4469 sb->s_magic = F2FS_SUPER_MAGIC;
4470 sb->s_time_gran = 1;
4471 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4472 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4473 super_set_uuid(sb, (void *) raw_super->uuid, sizeof(raw_super->uuid));
4474 super_set_sysfs_name_bdev(sb);
4475 sb->s_iflags |= SB_I_CGROUPWB;
4476
4477 /* init f2fs-specific super block info */
4478 sbi->valid_super_block = valid_super_block;
4479
4480 /* disallow all the data/node/meta page writes */
4481 set_sbi_flag(sbi, SBI_POR_DOING);
4482
4483 err = f2fs_init_write_merge_io(sbi);
4484 if (err)
4485 goto free_bio_info;
4486
4487 init_sb_info(sbi);
4488
4489 err = f2fs_init_iostat(sbi);
4490 if (err)
4491 goto free_bio_info;
4492
4493 err = init_percpu_info(sbi);
4494 if (err)
4495 goto free_iostat;
4496
4497 /* init per sbi slab cache */
4498 err = f2fs_init_xattr_caches(sbi);
4499 if (err)
4500 goto free_percpu;
4501 err = f2fs_init_page_array_cache(sbi);
4502 if (err)
4503 goto free_xattr_cache;
4504
4505 /* get an inode for meta space */
4506 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4507 if (IS_ERR(sbi->meta_inode)) {
4508 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4509 err = PTR_ERR(sbi->meta_inode);
4510 goto free_page_array_cache;
4511 }
4512
4513 err = f2fs_get_valid_checkpoint(sbi);
4514 if (err) {
4515 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4516 goto free_meta_inode;
4517 }
4518
4519 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4520 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4521 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4522 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4523 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4524 }
4525
4526 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4527 set_sbi_flag(sbi, SBI_NEED_FSCK);
4528
4529 /* Initialize device list */
4530 err = f2fs_scan_devices(sbi);
4531 if (err) {
4532 f2fs_err(sbi, "Failed to find devices");
4533 goto free_devices;
4534 }
4535
4536 err = f2fs_init_post_read_wq(sbi);
4537 if (err) {
4538 f2fs_err(sbi, "Failed to initialize post read workqueue");
4539 goto free_devices;
4540 }
4541
4542 sbi->total_valid_node_count =
4543 le32_to_cpu(sbi->ckpt->valid_node_count);
4544 percpu_counter_set(&sbi->total_valid_inode_count,
4545 le32_to_cpu(sbi->ckpt->valid_inode_count));
4546 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4547 sbi->total_valid_block_count =
4548 le64_to_cpu(sbi->ckpt->valid_block_count);
4549 sbi->last_valid_block_count = sbi->total_valid_block_count;
4550 sbi->reserved_blocks = 0;
4551 sbi->current_reserved_blocks = 0;
4552 limit_reserve_root(sbi);
4553 adjust_unusable_cap_perc(sbi);
4554
4555 f2fs_init_extent_cache_info(sbi);
4556
4557 f2fs_init_ino_entry_info(sbi);
4558
4559 f2fs_init_fsync_node_info(sbi);
4560
4561 /* setup checkpoint request control and start checkpoint issue thread */
4562 f2fs_init_ckpt_req_control(sbi);
4563 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4564 test_opt(sbi, MERGE_CHECKPOINT)) {
4565 err = f2fs_start_ckpt_thread(sbi);
4566 if (err) {
4567 f2fs_err(sbi,
4568 "Failed to start F2FS issue_checkpoint_thread (%d)",
4569 err);
4570 goto stop_ckpt_thread;
4571 }
4572 }
4573
4574 /* setup f2fs internal modules */
4575 err = f2fs_build_segment_manager(sbi);
4576 if (err) {
4577 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4578 err);
4579 goto free_sm;
4580 }
4581 err = f2fs_build_node_manager(sbi);
4582 if (err) {
4583 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4584 err);
4585 goto free_nm;
4586 }
4587
4588 /* For write statistics */
4589 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4590
4591 /* Read accumulated write IO statistics if exists */
4592 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4593 if (__exist_node_summaries(sbi))
4594 sbi->kbytes_written =
4595 le64_to_cpu(seg_i->journal->info.kbytes_written);
4596
4597 f2fs_build_gc_manager(sbi);
4598
4599 err = f2fs_build_stats(sbi);
4600 if (err)
4601 goto free_nm;
4602
4603 /* get an inode for node space */
4604 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4605 if (IS_ERR(sbi->node_inode)) {
4606 f2fs_err(sbi, "Failed to read node inode");
4607 err = PTR_ERR(sbi->node_inode);
4608 goto free_stats;
4609 }
4610
4611 /* read root inode and dentry */
4612 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4613 if (IS_ERR(root)) {
4614 f2fs_err(sbi, "Failed to read root inode");
4615 err = PTR_ERR(root);
4616 goto free_node_inode;
4617 }
4618 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4619 !root->i_size || !root->i_nlink) {
4620 iput(root);
4621 err = -EINVAL;
4622 goto free_node_inode;
4623 }
4624
4625 generic_set_sb_d_ops(sb);
4626 sb->s_root = d_make_root(root); /* allocate root dentry */
4627 if (!sb->s_root) {
4628 err = -ENOMEM;
4629 goto free_node_inode;
4630 }
4631
4632 err = f2fs_init_compress_inode(sbi);
4633 if (err)
4634 goto free_root_inode;
4635
4636 err = f2fs_register_sysfs(sbi);
4637 if (err)
4638 goto free_compress_inode;
4639
4640 #ifdef CONFIG_QUOTA
4641 /* Enable quota usage during mount */
4642 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4643 err = f2fs_enable_quotas(sb);
4644 if (err)
4645 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4646 }
4647
4648 quota_enabled = f2fs_recover_quota_begin(sbi);
4649 #endif
4650 /* if there are any orphan inodes, free them */
4651 err = f2fs_recover_orphan_inodes(sbi);
4652 if (err)
4653 goto free_meta;
4654
4655 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4656 goto reset_checkpoint;
4657
4658 /* recover fsynced data */
4659 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4660 !test_opt(sbi, NORECOVERY)) {
4661 /*
4662 * mount should be failed, when device has readonly mode, and
4663 * previous checkpoint was not done by clean system shutdown.
4664 */
4665 if (f2fs_hw_is_readonly(sbi)) {
4666 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4667 err = f2fs_recover_fsync_data(sbi, true);
4668 if (err > 0) {
4669 err = -EROFS;
4670 f2fs_err(sbi, "Need to recover fsync data, but "
4671 "write access unavailable, please try "
4672 "mount w/ disable_roll_forward or norecovery");
4673 }
4674 if (err < 0)
4675 goto free_meta;
4676 }
4677 f2fs_info(sbi, "write access unavailable, skipping recovery");
4678 goto reset_checkpoint;
4679 }
4680
4681 if (need_fsck)
4682 set_sbi_flag(sbi, SBI_NEED_FSCK);
4683
4684 if (skip_recovery)
4685 goto reset_checkpoint;
4686
4687 err = f2fs_recover_fsync_data(sbi, false);
4688 if (err < 0) {
4689 if (err != -ENOMEM)
4690 skip_recovery = true;
4691 need_fsck = true;
4692 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4693 err);
4694 goto free_meta;
4695 }
4696 } else {
4697 err = f2fs_recover_fsync_data(sbi, true);
4698
4699 if (!f2fs_readonly(sb) && err > 0) {
4700 err = -EINVAL;
4701 f2fs_err(sbi, "Need to recover fsync data");
4702 goto free_meta;
4703 }
4704 }
4705
4706 #ifdef CONFIG_QUOTA
4707 f2fs_recover_quota_end(sbi, quota_enabled);
4708 #endif
4709 reset_checkpoint:
4710 /*
4711 * If the f2fs is not readonly and fsync data recovery succeeds,
4712 * check zoned block devices' write pointer consistency.
4713 */
4714 if (f2fs_sb_has_blkzoned(sbi) && !f2fs_readonly(sb)) {
4715 int err2;
4716
4717 f2fs_notice(sbi, "Checking entire write pointers");
4718 err2 = f2fs_check_write_pointer(sbi);
4719 if (err2)
4720 err = err2;
4721 }
4722 if (err)
4723 goto free_meta;
4724
4725 err = f2fs_init_inmem_curseg(sbi);
4726 if (err)
4727 goto sync_free_meta;
4728
4729 /* f2fs_recover_fsync_data() cleared this already */
4730 clear_sbi_flag(sbi, SBI_POR_DOING);
4731
4732 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4733 err = f2fs_disable_checkpoint(sbi);
4734 if (err)
4735 goto sync_free_meta;
4736 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4737 f2fs_enable_checkpoint(sbi);
4738 }
4739
4740 /*
4741 * If filesystem is not mounted as read-only then
4742 * do start the gc_thread.
4743 */
4744 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4745 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4746 /* After POR, we can run background GC thread.*/
4747 err = f2fs_start_gc_thread(sbi);
4748 if (err)
4749 goto sync_free_meta;
4750 }
4751 kvfree(options);
4752
4753 /* recover broken superblock */
4754 if (recovery) {
4755 err = f2fs_commit_super(sbi, true);
4756 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4757 sbi->valid_super_block ? 1 : 2, err);
4758 }
4759
4760 f2fs_join_shrinker(sbi);
4761
4762 f2fs_tuning_parameters(sbi);
4763
4764 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4765 cur_cp_version(F2FS_CKPT(sbi)));
4766 f2fs_update_time(sbi, CP_TIME);
4767 f2fs_update_time(sbi, REQ_TIME);
4768 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4769 return 0;
4770
4771 sync_free_meta:
4772 /* safe to flush all the data */
4773 sync_filesystem(sbi->sb);
4774 retry_cnt = 0;
4775
4776 free_meta:
4777 #ifdef CONFIG_QUOTA
4778 f2fs_truncate_quota_inode_pages(sb);
4779 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4780 f2fs_quota_off_umount(sbi->sb);
4781 #endif
4782 /*
4783 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4784 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4785 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4786 * falls into an infinite loop in f2fs_sync_meta_pages().
4787 */
4788 truncate_inode_pages_final(META_MAPPING(sbi));
4789 /* evict some inodes being cached by GC */
4790 evict_inodes(sb);
4791 f2fs_unregister_sysfs(sbi);
4792 free_compress_inode:
4793 f2fs_destroy_compress_inode(sbi);
4794 free_root_inode:
4795 dput(sb->s_root);
4796 sb->s_root = NULL;
4797 free_node_inode:
4798 f2fs_release_ino_entry(sbi, true);
4799 truncate_inode_pages_final(NODE_MAPPING(sbi));
4800 iput(sbi->node_inode);
4801 sbi->node_inode = NULL;
4802 free_stats:
4803 f2fs_destroy_stats(sbi);
4804 free_nm:
4805 /* stop discard thread before destroying node manager */
4806 f2fs_stop_discard_thread(sbi);
4807 f2fs_destroy_node_manager(sbi);
4808 free_sm:
4809 f2fs_destroy_segment_manager(sbi);
4810 stop_ckpt_thread:
4811 f2fs_stop_ckpt_thread(sbi);
4812 /* flush s_error_work before sbi destroy */
4813 flush_work(&sbi->s_error_work);
4814 f2fs_destroy_post_read_wq(sbi);
4815 free_devices:
4816 destroy_device_list(sbi);
4817 kvfree(sbi->ckpt);
4818 free_meta_inode:
4819 make_bad_inode(sbi->meta_inode);
4820 iput(sbi->meta_inode);
4821 sbi->meta_inode = NULL;
4822 free_page_array_cache:
4823 f2fs_destroy_page_array_cache(sbi);
4824 free_xattr_cache:
4825 f2fs_destroy_xattr_caches(sbi);
4826 free_percpu:
4827 destroy_percpu_info(sbi);
4828 free_iostat:
4829 f2fs_destroy_iostat(sbi);
4830 free_bio_info:
4831 for (i = 0; i < NR_PAGE_TYPE; i++)
4832 kvfree(sbi->write_io[i]);
4833
4834 #if IS_ENABLED(CONFIG_UNICODE)
4835 utf8_unload(sb->s_encoding);
4836 sb->s_encoding = NULL;
4837 #endif
4838 free_options:
4839 #ifdef CONFIG_QUOTA
4840 for (i = 0; i < MAXQUOTAS; i++)
4841 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4842 #endif
4843 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4844 kvfree(options);
4845 free_sb_buf:
4846 kfree(raw_super);
4847 free_sbi:
4848 if (sbi->s_chksum_driver)
4849 crypto_free_shash(sbi->s_chksum_driver);
4850 kfree(sbi);
4851 sb->s_fs_info = NULL;
4852
4853 /* give only one another chance */
4854 if (retry_cnt > 0 && skip_recovery) {
4855 retry_cnt--;
4856 shrink_dcache_sb(sb);
4857 goto try_onemore;
4858 }
4859 return err;
4860 }
4861
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4862 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4863 const char *dev_name, void *data)
4864 {
4865 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4866 }
4867
kill_f2fs_super(struct super_block * sb)4868 static void kill_f2fs_super(struct super_block *sb)
4869 {
4870 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4871
4872 if (sb->s_root) {
4873 set_sbi_flag(sbi, SBI_IS_CLOSE);
4874 f2fs_stop_gc_thread(sbi);
4875 f2fs_stop_discard_thread(sbi);
4876
4877 #ifdef CONFIG_F2FS_FS_COMPRESSION
4878 /*
4879 * latter evict_inode() can bypass checking and invalidating
4880 * compress inode cache.
4881 */
4882 if (test_opt(sbi, COMPRESS_CACHE))
4883 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4884 #endif
4885
4886 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4887 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4888 struct cp_control cpc = {
4889 .reason = CP_UMOUNT,
4890 };
4891 stat_inc_cp_call_count(sbi, TOTAL_CALL);
4892 f2fs_write_checkpoint(sbi, &cpc);
4893 }
4894
4895 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4896 sb->s_flags &= ~SB_RDONLY;
4897 }
4898 kill_block_super(sb);
4899 /* Release block devices last, after fscrypt_destroy_keyring(). */
4900 if (sbi) {
4901 destroy_device_list(sbi);
4902 kfree(sbi);
4903 sb->s_fs_info = NULL;
4904 }
4905 }
4906
4907 static struct file_system_type f2fs_fs_type = {
4908 .owner = THIS_MODULE,
4909 .name = "f2fs",
4910 .mount = f2fs_mount,
4911 .kill_sb = kill_f2fs_super,
4912 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4913 };
4914 MODULE_ALIAS_FS("f2fs");
4915
init_inodecache(void)4916 static int __init init_inodecache(void)
4917 {
4918 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4919 sizeof(struct f2fs_inode_info), 0,
4920 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4921 return f2fs_inode_cachep ? 0 : -ENOMEM;
4922 }
4923
destroy_inodecache(void)4924 static void destroy_inodecache(void)
4925 {
4926 /*
4927 * Make sure all delayed rcu free inodes are flushed before we
4928 * destroy cache.
4929 */
4930 rcu_barrier();
4931 kmem_cache_destroy(f2fs_inode_cachep);
4932 }
4933
init_f2fs_fs(void)4934 static int __init init_f2fs_fs(void)
4935 {
4936 int err;
4937
4938 err = init_inodecache();
4939 if (err)
4940 goto fail;
4941 err = f2fs_create_node_manager_caches();
4942 if (err)
4943 goto free_inodecache;
4944 err = f2fs_create_segment_manager_caches();
4945 if (err)
4946 goto free_node_manager_caches;
4947 err = f2fs_create_checkpoint_caches();
4948 if (err)
4949 goto free_segment_manager_caches;
4950 err = f2fs_create_recovery_cache();
4951 if (err)
4952 goto free_checkpoint_caches;
4953 err = f2fs_create_extent_cache();
4954 if (err)
4955 goto free_recovery_cache;
4956 err = f2fs_create_garbage_collection_cache();
4957 if (err)
4958 goto free_extent_cache;
4959 err = f2fs_init_sysfs();
4960 if (err)
4961 goto free_garbage_collection_cache;
4962 err = f2fs_init_shrinker();
4963 if (err)
4964 goto free_sysfs;
4965 err = register_filesystem(&f2fs_fs_type);
4966 if (err)
4967 goto free_shrinker;
4968 f2fs_create_root_stats();
4969 err = f2fs_init_post_read_processing();
4970 if (err)
4971 goto free_root_stats;
4972 err = f2fs_init_iostat_processing();
4973 if (err)
4974 goto free_post_read;
4975 err = f2fs_init_bio_entry_cache();
4976 if (err)
4977 goto free_iostat;
4978 err = f2fs_init_bioset();
4979 if (err)
4980 goto free_bio_entry_cache;
4981 err = f2fs_init_compress_mempool();
4982 if (err)
4983 goto free_bioset;
4984 err = f2fs_init_compress_cache();
4985 if (err)
4986 goto free_compress_mempool;
4987 err = f2fs_create_casefold_cache();
4988 if (err)
4989 goto free_compress_cache;
4990 return 0;
4991 free_compress_cache:
4992 f2fs_destroy_compress_cache();
4993 free_compress_mempool:
4994 f2fs_destroy_compress_mempool();
4995 free_bioset:
4996 f2fs_destroy_bioset();
4997 free_bio_entry_cache:
4998 f2fs_destroy_bio_entry_cache();
4999 free_iostat:
5000 f2fs_destroy_iostat_processing();
5001 free_post_read:
5002 f2fs_destroy_post_read_processing();
5003 free_root_stats:
5004 f2fs_destroy_root_stats();
5005 unregister_filesystem(&f2fs_fs_type);
5006 free_shrinker:
5007 f2fs_exit_shrinker();
5008 free_sysfs:
5009 f2fs_exit_sysfs();
5010 free_garbage_collection_cache:
5011 f2fs_destroy_garbage_collection_cache();
5012 free_extent_cache:
5013 f2fs_destroy_extent_cache();
5014 free_recovery_cache:
5015 f2fs_destroy_recovery_cache();
5016 free_checkpoint_caches:
5017 f2fs_destroy_checkpoint_caches();
5018 free_segment_manager_caches:
5019 f2fs_destroy_segment_manager_caches();
5020 free_node_manager_caches:
5021 f2fs_destroy_node_manager_caches();
5022 free_inodecache:
5023 destroy_inodecache();
5024 fail:
5025 return err;
5026 }
5027
exit_f2fs_fs(void)5028 static void __exit exit_f2fs_fs(void)
5029 {
5030 f2fs_destroy_casefold_cache();
5031 f2fs_destroy_compress_cache();
5032 f2fs_destroy_compress_mempool();
5033 f2fs_destroy_bioset();
5034 f2fs_destroy_bio_entry_cache();
5035 f2fs_destroy_iostat_processing();
5036 f2fs_destroy_post_read_processing();
5037 f2fs_destroy_root_stats();
5038 unregister_filesystem(&f2fs_fs_type);
5039 f2fs_exit_shrinker();
5040 f2fs_exit_sysfs();
5041 f2fs_destroy_garbage_collection_cache();
5042 f2fs_destroy_extent_cache();
5043 f2fs_destroy_recovery_cache();
5044 f2fs_destroy_checkpoint_caches();
5045 f2fs_destroy_segment_manager_caches();
5046 f2fs_destroy_node_manager_caches();
5047 destroy_inodecache();
5048 }
5049
5050 module_init(init_f2fs_fs)
5051 module_exit(exit_f2fs_fs)
5052
5053 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5054 MODULE_DESCRIPTION("Flash Friendly File System");
5055 MODULE_LICENSE("GPL");
5056 MODULE_SOFTDEP("pre: crc32");
5057
5058