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] = "invalid blkaddr",
65 };
66
f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned long rate,unsigned long type)67 int f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned long rate,
68 unsigned long type)
69 {
70 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
71
72 if (rate) {
73 if (rate > INT_MAX)
74 return -EINVAL;
75 atomic_set(&ffi->inject_ops, 0);
76 ffi->inject_rate = (int)rate;
77 }
78
79 if (type) {
80 if (type >= BIT(FAULT_MAX))
81 return -EINVAL;
82 ffi->inject_type = (unsigned int)type;
83 }
84
85 if (!rate && !type)
86 memset(ffi, 0, sizeof(struct f2fs_fault_info));
87 else
88 f2fs_info(sbi,
89 "build fault injection attr: rate: %lu, type: 0x%lx",
90 rate, type);
91 return 0;
92 }
93 #endif
94
95 /* f2fs-wide shrinker description */
96 static struct shrinker f2fs_shrinker_info = {
97 .scan_objects = f2fs_shrink_scan,
98 .count_objects = f2fs_shrink_count,
99 .seeks = DEFAULT_SEEKS,
100 };
101
102 enum {
103 Opt_gc_background,
104 Opt_disable_roll_forward,
105 Opt_norecovery,
106 Opt_discard,
107 Opt_nodiscard,
108 Opt_noheap,
109 Opt_heap,
110 Opt_user_xattr,
111 Opt_nouser_xattr,
112 Opt_acl,
113 Opt_noacl,
114 Opt_active_logs,
115 Opt_disable_ext_identify,
116 Opt_inline_xattr,
117 Opt_noinline_xattr,
118 Opt_inline_xattr_size,
119 Opt_inline_data,
120 Opt_inline_dentry,
121 Opt_noinline_dentry,
122 Opt_flush_merge,
123 Opt_noflush_merge,
124 Opt_barrier,
125 Opt_nobarrier,
126 Opt_fastboot,
127 Opt_extent_cache,
128 Opt_noextent_cache,
129 Opt_noinline_data,
130 Opt_data_flush,
131 Opt_reserve_root,
132 Opt_resgid,
133 Opt_resuid,
134 Opt_mode,
135 Opt_fault_injection,
136 Opt_fault_type,
137 Opt_lazytime,
138 Opt_nolazytime,
139 Opt_quota,
140 Opt_noquota,
141 Opt_usrquota,
142 Opt_grpquota,
143 Opt_prjquota,
144 Opt_usrjquota,
145 Opt_grpjquota,
146 Opt_prjjquota,
147 Opt_offusrjquota,
148 Opt_offgrpjquota,
149 Opt_offprjjquota,
150 Opt_jqfmt_vfsold,
151 Opt_jqfmt_vfsv0,
152 Opt_jqfmt_vfsv1,
153 Opt_alloc,
154 Opt_fsync,
155 Opt_test_dummy_encryption,
156 Opt_inlinecrypt,
157 Opt_checkpoint_disable,
158 Opt_checkpoint_disable_cap,
159 Opt_checkpoint_disable_cap_perc,
160 Opt_checkpoint_enable,
161 Opt_checkpoint_merge,
162 Opt_nocheckpoint_merge,
163 Opt_compress_algorithm,
164 Opt_compress_log_size,
165 Opt_compress_extension,
166 Opt_nocompress_extension,
167 Opt_compress_chksum,
168 Opt_compress_mode,
169 Opt_compress_cache,
170 Opt_atgc,
171 Opt_gc_merge,
172 Opt_nogc_merge,
173 Opt_discard_unit,
174 Opt_memory_mode,
175 Opt_age_extent_cache,
176 Opt_errors,
177 Opt_err,
178 };
179
180 static match_table_t f2fs_tokens = {
181 {Opt_gc_background, "background_gc=%s"},
182 {Opt_disable_roll_forward, "disable_roll_forward"},
183 {Opt_norecovery, "norecovery"},
184 {Opt_discard, "discard"},
185 {Opt_nodiscard, "nodiscard"},
186 {Opt_noheap, "no_heap"},
187 {Opt_heap, "heap"},
188 {Opt_user_xattr, "user_xattr"},
189 {Opt_nouser_xattr, "nouser_xattr"},
190 {Opt_acl, "acl"},
191 {Opt_noacl, "noacl"},
192 {Opt_active_logs, "active_logs=%u"},
193 {Opt_disable_ext_identify, "disable_ext_identify"},
194 {Opt_inline_xattr, "inline_xattr"},
195 {Opt_noinline_xattr, "noinline_xattr"},
196 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
197 {Opt_inline_data, "inline_data"},
198 {Opt_inline_dentry, "inline_dentry"},
199 {Opt_noinline_dentry, "noinline_dentry"},
200 {Opt_flush_merge, "flush_merge"},
201 {Opt_noflush_merge, "noflush_merge"},
202 {Opt_barrier, "barrier"},
203 {Opt_nobarrier, "nobarrier"},
204 {Opt_fastboot, "fastboot"},
205 {Opt_extent_cache, "extent_cache"},
206 {Opt_noextent_cache, "noextent_cache"},
207 {Opt_noinline_data, "noinline_data"},
208 {Opt_data_flush, "data_flush"},
209 {Opt_reserve_root, "reserve_root=%u"},
210 {Opt_resgid, "resgid=%u"},
211 {Opt_resuid, "resuid=%u"},
212 {Opt_mode, "mode=%s"},
213 {Opt_fault_injection, "fault_injection=%u"},
214 {Opt_fault_type, "fault_type=%u"},
215 {Opt_lazytime, "lazytime"},
216 {Opt_nolazytime, "nolazytime"},
217 {Opt_quota, "quota"},
218 {Opt_noquota, "noquota"},
219 {Opt_usrquota, "usrquota"},
220 {Opt_grpquota, "grpquota"},
221 {Opt_prjquota, "prjquota"},
222 {Opt_usrjquota, "usrjquota=%s"},
223 {Opt_grpjquota, "grpjquota=%s"},
224 {Opt_prjjquota, "prjjquota=%s"},
225 {Opt_offusrjquota, "usrjquota="},
226 {Opt_offgrpjquota, "grpjquota="},
227 {Opt_offprjjquota, "prjjquota="},
228 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
229 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
230 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
231 {Opt_alloc, "alloc_mode=%s"},
232 {Opt_fsync, "fsync_mode=%s"},
233 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
234 {Opt_test_dummy_encryption, "test_dummy_encryption"},
235 {Opt_inlinecrypt, "inlinecrypt"},
236 {Opt_checkpoint_disable, "checkpoint=disable"},
237 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
238 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
239 {Opt_checkpoint_enable, "checkpoint=enable"},
240 {Opt_checkpoint_merge, "checkpoint_merge"},
241 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
242 {Opt_compress_algorithm, "compress_algorithm=%s"},
243 {Opt_compress_log_size, "compress_log_size=%u"},
244 {Opt_compress_extension, "compress_extension=%s"},
245 {Opt_nocompress_extension, "nocompress_extension=%s"},
246 {Opt_compress_chksum, "compress_chksum"},
247 {Opt_compress_mode, "compress_mode=%s"},
248 {Opt_compress_cache, "compress_cache"},
249 {Opt_atgc, "atgc"},
250 {Opt_gc_merge, "gc_merge"},
251 {Opt_nogc_merge, "nogc_merge"},
252 {Opt_discard_unit, "discard_unit=%s"},
253 {Opt_memory_mode, "memory=%s"},
254 {Opt_age_extent_cache, "age_extent_cache"},
255 {Opt_errors, "errors=%s"},
256 {Opt_err, NULL},
257 };
258
f2fs_printk(struct f2fs_sb_info * sbi,bool limit_rate,const char * fmt,...)259 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate,
260 const char *fmt, ...)
261 {
262 struct va_format vaf;
263 va_list args;
264 int level;
265
266 va_start(args, fmt);
267
268 level = printk_get_level(fmt);
269 vaf.fmt = printk_skip_level(fmt);
270 vaf.va = &args;
271 if (limit_rate)
272 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n",
273 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
274 else
275 printk("%c%cF2FS-fs (%s): %pV\n",
276 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
277
278 va_end(args);
279 }
280
281 #if IS_ENABLED(CONFIG_UNICODE)
282 static const struct f2fs_sb_encodings {
283 __u16 magic;
284 char *name;
285 unsigned int version;
286 } f2fs_sb_encoding_map[] = {
287 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
288 };
289
290 static const struct f2fs_sb_encodings *
f2fs_sb_read_encoding(const struct f2fs_super_block * sb)291 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
292 {
293 __u16 magic = le16_to_cpu(sb->s_encoding);
294 int i;
295
296 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
297 if (magic == f2fs_sb_encoding_map[i].magic)
298 return &f2fs_sb_encoding_map[i];
299
300 return NULL;
301 }
302
303 struct kmem_cache *f2fs_cf_name_slab;
f2fs_create_casefold_cache(void)304 static int __init f2fs_create_casefold_cache(void)
305 {
306 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
307 F2FS_NAME_LEN);
308 return f2fs_cf_name_slab ? 0 : -ENOMEM;
309 }
310
f2fs_destroy_casefold_cache(void)311 static void f2fs_destroy_casefold_cache(void)
312 {
313 kmem_cache_destroy(f2fs_cf_name_slab);
314 }
315 #else
f2fs_create_casefold_cache(void)316 static int __init f2fs_create_casefold_cache(void) { return 0; }
f2fs_destroy_casefold_cache(void)317 static void f2fs_destroy_casefold_cache(void) { }
318 #endif
319
limit_reserve_root(struct f2fs_sb_info * sbi)320 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
321 {
322 block_t limit = min((sbi->user_block_count >> 3),
323 sbi->user_block_count - sbi->reserved_blocks);
324
325 /* limit is 12.5% */
326 if (test_opt(sbi, RESERVE_ROOT) &&
327 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
328 F2FS_OPTION(sbi).root_reserved_blocks = limit;
329 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
330 F2FS_OPTION(sbi).root_reserved_blocks);
331 }
332 if (!test_opt(sbi, RESERVE_ROOT) &&
333 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
334 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
335 !gid_eq(F2FS_OPTION(sbi).s_resgid,
336 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
337 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
338 from_kuid_munged(&init_user_ns,
339 F2FS_OPTION(sbi).s_resuid),
340 from_kgid_munged(&init_user_ns,
341 F2FS_OPTION(sbi).s_resgid));
342 }
343
adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)344 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
345 {
346 if (!F2FS_OPTION(sbi).unusable_cap_perc)
347 return;
348
349 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
350 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
351 else
352 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
353 F2FS_OPTION(sbi).unusable_cap_perc;
354
355 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
356 F2FS_OPTION(sbi).unusable_cap,
357 F2FS_OPTION(sbi).unusable_cap_perc);
358 }
359
init_once(void * foo)360 static void init_once(void *foo)
361 {
362 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
363
364 inode_init_once(&fi->vfs_inode);
365 }
366
367 #ifdef CONFIG_QUOTA
368 static const char * const quotatypes[] = INITQFNAMES;
369 #define QTYPE2NAME(t) (quotatypes[t])
f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)370 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
371 substring_t *args)
372 {
373 struct f2fs_sb_info *sbi = F2FS_SB(sb);
374 char *qname;
375 int ret = -EINVAL;
376
377 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
378 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
379 return -EINVAL;
380 }
381 if (f2fs_sb_has_quota_ino(sbi)) {
382 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
383 return 0;
384 }
385
386 qname = match_strdup(args);
387 if (!qname) {
388 f2fs_err(sbi, "Not enough memory for storing quotafile name");
389 return -ENOMEM;
390 }
391 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
392 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
393 ret = 0;
394 else
395 f2fs_err(sbi, "%s quota file already specified",
396 QTYPE2NAME(qtype));
397 goto errout;
398 }
399 if (strchr(qname, '/')) {
400 f2fs_err(sbi, "quotafile must be on filesystem root");
401 goto errout;
402 }
403 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
404 set_opt(sbi, QUOTA);
405 return 0;
406 errout:
407 kfree(qname);
408 return ret;
409 }
410
f2fs_clear_qf_name(struct super_block * sb,int qtype)411 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
412 {
413 struct f2fs_sb_info *sbi = F2FS_SB(sb);
414
415 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
416 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
417 return -EINVAL;
418 }
419 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
420 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
421 return 0;
422 }
423
f2fs_check_quota_options(struct f2fs_sb_info * sbi)424 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
425 {
426 /*
427 * We do the test below only for project quotas. 'usrquota' and
428 * 'grpquota' mount options are allowed even without quota feature
429 * to support legacy quotas in quota files.
430 */
431 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
432 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
433 return -1;
434 }
435 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
436 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
437 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
438 if (test_opt(sbi, USRQUOTA) &&
439 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
440 clear_opt(sbi, USRQUOTA);
441
442 if (test_opt(sbi, GRPQUOTA) &&
443 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
444 clear_opt(sbi, GRPQUOTA);
445
446 if (test_opt(sbi, PRJQUOTA) &&
447 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
448 clear_opt(sbi, PRJQUOTA);
449
450 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
451 test_opt(sbi, PRJQUOTA)) {
452 f2fs_err(sbi, "old and new quota format mixing");
453 return -1;
454 }
455
456 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
457 f2fs_err(sbi, "journaled quota format not specified");
458 return -1;
459 }
460 }
461
462 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
463 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
464 F2FS_OPTION(sbi).s_jquota_fmt = 0;
465 }
466 return 0;
467 }
468 #endif
469
f2fs_set_test_dummy_encryption(struct super_block * sb,const char * opt,const substring_t * arg,bool is_remount)470 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
471 const char *opt,
472 const substring_t *arg,
473 bool is_remount)
474 {
475 struct f2fs_sb_info *sbi = F2FS_SB(sb);
476 struct fs_parameter param = {
477 .type = fs_value_is_string,
478 .string = arg->from ? arg->from : "",
479 };
480 struct fscrypt_dummy_policy *policy =
481 &F2FS_OPTION(sbi).dummy_enc_policy;
482 int err;
483
484 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
485 f2fs_warn(sbi, "test_dummy_encryption option not supported");
486 return -EINVAL;
487 }
488
489 if (!f2fs_sb_has_encrypt(sbi)) {
490 f2fs_err(sbi, "Encrypt feature is off");
491 return -EINVAL;
492 }
493
494 /*
495 * This mount option is just for testing, and it's not worthwhile to
496 * implement the extra complexity (e.g. RCU protection) that would be
497 * needed to allow it to be set or changed during remount. We do allow
498 * it to be specified during remount, but only if there is no change.
499 */
500 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
501 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
502 return -EINVAL;
503 }
504
505 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
506 if (err) {
507 if (err == -EEXIST)
508 f2fs_warn(sbi,
509 "Can't change test_dummy_encryption on remount");
510 else if (err == -EINVAL)
511 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
512 opt);
513 else
514 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
515 opt, err);
516 return -EINVAL;
517 }
518 f2fs_warn(sbi, "Test dummy encryption mode enabled");
519 return 0;
520 }
521
522 #ifdef CONFIG_F2FS_FS_COMPRESSION
is_compress_extension_exist(struct f2fs_sb_info * sbi,const char * new_ext,bool is_ext)523 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
524 const char *new_ext, bool is_ext)
525 {
526 unsigned char (*ext)[F2FS_EXTENSION_LEN];
527 int ext_cnt;
528 int i;
529
530 if (is_ext) {
531 ext = F2FS_OPTION(sbi).extensions;
532 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
533 } else {
534 ext = F2FS_OPTION(sbi).noextensions;
535 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
536 }
537
538 for (i = 0; i < ext_cnt; i++) {
539 if (!strcasecmp(new_ext, ext[i]))
540 return true;
541 }
542
543 return false;
544 }
545
546 /*
547 * 1. The same extension name cannot not appear in both compress and non-compress extension
548 * at the same time.
549 * 2. If the compress extension specifies all files, the types specified by the non-compress
550 * extension will be treated as special cases and will not be compressed.
551 * 3. Don't allow the non-compress extension specifies all files.
552 */
f2fs_test_compress_extension(struct f2fs_sb_info * sbi)553 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
554 {
555 unsigned char (*ext)[F2FS_EXTENSION_LEN];
556 unsigned char (*noext)[F2FS_EXTENSION_LEN];
557 int ext_cnt, noext_cnt, index = 0, no_index = 0;
558
559 ext = F2FS_OPTION(sbi).extensions;
560 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
561 noext = F2FS_OPTION(sbi).noextensions;
562 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
563
564 if (!noext_cnt)
565 return 0;
566
567 for (no_index = 0; no_index < noext_cnt; no_index++) {
568 if (!strcasecmp("*", noext[no_index])) {
569 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
570 return -EINVAL;
571 }
572 for (index = 0; index < ext_cnt; index++) {
573 if (!strcasecmp(ext[index], noext[no_index])) {
574 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
575 ext[index]);
576 return -EINVAL;
577 }
578 }
579 }
580 return 0;
581 }
582
583 #ifdef CONFIG_F2FS_FS_LZ4
f2fs_set_lz4hc_level(struct f2fs_sb_info * sbi,const char * str)584 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
585 {
586 #ifdef CONFIG_F2FS_FS_LZ4HC
587 unsigned int level;
588
589 if (strlen(str) == 3) {
590 F2FS_OPTION(sbi).compress_level = 0;
591 return 0;
592 }
593
594 str += 3;
595
596 if (str[0] != ':') {
597 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
598 return -EINVAL;
599 }
600 if (kstrtouint(str + 1, 10, &level))
601 return -EINVAL;
602
603 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
604 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
605 return -EINVAL;
606 }
607
608 F2FS_OPTION(sbi).compress_level = level;
609 return 0;
610 #else
611 if (strlen(str) == 3) {
612 F2FS_OPTION(sbi).compress_level = 0;
613 return 0;
614 }
615 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
616 return -EINVAL;
617 #endif
618 }
619 #endif
620
621 #ifdef CONFIG_F2FS_FS_ZSTD
f2fs_set_zstd_level(struct f2fs_sb_info * sbi,const char * str)622 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
623 {
624 int level;
625 int len = 4;
626
627 if (strlen(str) == len) {
628 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
629 return 0;
630 }
631
632 str += len;
633
634 if (str[0] != ':') {
635 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
636 return -EINVAL;
637 }
638 if (kstrtoint(str + 1, 10, &level))
639 return -EINVAL;
640
641 /* f2fs does not support negative compress level now */
642 if (level < 0) {
643 f2fs_info(sbi, "do not support negative compress level: %d", level);
644 return -ERANGE;
645 }
646
647 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
648 f2fs_info(sbi, "invalid zstd compress level: %d", level);
649 return -EINVAL;
650 }
651
652 F2FS_OPTION(sbi).compress_level = level;
653 return 0;
654 }
655 #endif
656 #endif
657
parse_options(struct super_block * sb,char * options,bool is_remount)658 static int parse_options(struct super_block *sb, char *options, bool is_remount)
659 {
660 struct f2fs_sb_info *sbi = F2FS_SB(sb);
661 substring_t args[MAX_OPT_ARGS];
662 #ifdef CONFIG_F2FS_FS_COMPRESSION
663 unsigned char (*ext)[F2FS_EXTENSION_LEN];
664 unsigned char (*noext)[F2FS_EXTENSION_LEN];
665 int ext_cnt, noext_cnt;
666 #endif
667 char *p, *name;
668 int arg = 0;
669 kuid_t uid;
670 kgid_t gid;
671 int ret;
672
673 if (!options)
674 goto default_check;
675
676 while ((p = strsep(&options, ",")) != NULL) {
677 int token;
678
679 if (!*p)
680 continue;
681 /*
682 * Initialize args struct so we know whether arg was
683 * found; some options take optional arguments.
684 */
685 args[0].to = args[0].from = NULL;
686 token = match_token(p, f2fs_tokens, args);
687
688 switch (token) {
689 case Opt_gc_background:
690 name = match_strdup(&args[0]);
691
692 if (!name)
693 return -ENOMEM;
694 if (!strcmp(name, "on")) {
695 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
696 } else if (!strcmp(name, "off")) {
697 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
698 } else if (!strcmp(name, "sync")) {
699 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
700 } else {
701 kfree(name);
702 return -EINVAL;
703 }
704 kfree(name);
705 break;
706 case Opt_disable_roll_forward:
707 set_opt(sbi, DISABLE_ROLL_FORWARD);
708 break;
709 case Opt_norecovery:
710 /* this option mounts f2fs with ro */
711 set_opt(sbi, NORECOVERY);
712 if (!f2fs_readonly(sb))
713 return -EINVAL;
714 break;
715 case Opt_discard:
716 if (!f2fs_hw_support_discard(sbi)) {
717 f2fs_warn(sbi, "device does not support discard");
718 break;
719 }
720 set_opt(sbi, DISCARD);
721 break;
722 case Opt_nodiscard:
723 if (f2fs_hw_should_discard(sbi)) {
724 f2fs_warn(sbi, "discard is required for zoned block devices");
725 return -EINVAL;
726 }
727 clear_opt(sbi, DISCARD);
728 break;
729 case Opt_noheap:
730 case Opt_heap:
731 f2fs_warn(sbi, "heap/no_heap options were deprecated");
732 break;
733 #ifdef CONFIG_F2FS_FS_XATTR
734 case Opt_user_xattr:
735 set_opt(sbi, XATTR_USER);
736 break;
737 case Opt_nouser_xattr:
738 clear_opt(sbi, XATTR_USER);
739 break;
740 case Opt_inline_xattr:
741 set_opt(sbi, INLINE_XATTR);
742 break;
743 case Opt_noinline_xattr:
744 clear_opt(sbi, INLINE_XATTR);
745 break;
746 case Opt_inline_xattr_size:
747 if (args->from && match_int(args, &arg))
748 return -EINVAL;
749 set_opt(sbi, INLINE_XATTR_SIZE);
750 F2FS_OPTION(sbi).inline_xattr_size = arg;
751 break;
752 #else
753 case Opt_user_xattr:
754 f2fs_info(sbi, "user_xattr options not supported");
755 break;
756 case Opt_nouser_xattr:
757 f2fs_info(sbi, "nouser_xattr options not supported");
758 break;
759 case Opt_inline_xattr:
760 f2fs_info(sbi, "inline_xattr options not supported");
761 break;
762 case Opt_noinline_xattr:
763 f2fs_info(sbi, "noinline_xattr options not supported");
764 break;
765 #endif
766 #ifdef CONFIG_F2FS_FS_POSIX_ACL
767 case Opt_acl:
768 set_opt(sbi, POSIX_ACL);
769 break;
770 case Opt_noacl:
771 clear_opt(sbi, POSIX_ACL);
772 break;
773 #else
774 case Opt_acl:
775 f2fs_info(sbi, "acl options not supported");
776 break;
777 case Opt_noacl:
778 f2fs_info(sbi, "noacl options not supported");
779 break;
780 #endif
781 case Opt_active_logs:
782 if (args->from && match_int(args, &arg))
783 return -EINVAL;
784 if (arg != 2 && arg != 4 &&
785 arg != NR_CURSEG_PERSIST_TYPE)
786 return -EINVAL;
787 F2FS_OPTION(sbi).active_logs = arg;
788 break;
789 case Opt_disable_ext_identify:
790 set_opt(sbi, DISABLE_EXT_IDENTIFY);
791 break;
792 case Opt_inline_data:
793 set_opt(sbi, INLINE_DATA);
794 break;
795 case Opt_inline_dentry:
796 set_opt(sbi, INLINE_DENTRY);
797 break;
798 case Opt_noinline_dentry:
799 clear_opt(sbi, INLINE_DENTRY);
800 break;
801 case Opt_flush_merge:
802 set_opt(sbi, FLUSH_MERGE);
803 break;
804 case Opt_noflush_merge:
805 clear_opt(sbi, FLUSH_MERGE);
806 break;
807 case Opt_nobarrier:
808 set_opt(sbi, NOBARRIER);
809 break;
810 case Opt_barrier:
811 clear_opt(sbi, NOBARRIER);
812 break;
813 case Opt_fastboot:
814 set_opt(sbi, FASTBOOT);
815 break;
816 case Opt_extent_cache:
817 set_opt(sbi, READ_EXTENT_CACHE);
818 break;
819 case Opt_noextent_cache:
820 clear_opt(sbi, READ_EXTENT_CACHE);
821 break;
822 case Opt_noinline_data:
823 clear_opt(sbi, INLINE_DATA);
824 break;
825 case Opt_data_flush:
826 set_opt(sbi, DATA_FLUSH);
827 break;
828 case Opt_reserve_root:
829 if (args->from && match_int(args, &arg))
830 return -EINVAL;
831 if (test_opt(sbi, RESERVE_ROOT)) {
832 f2fs_info(sbi, "Preserve previous reserve_root=%u",
833 F2FS_OPTION(sbi).root_reserved_blocks);
834 } else {
835 F2FS_OPTION(sbi).root_reserved_blocks = arg;
836 set_opt(sbi, RESERVE_ROOT);
837 }
838 break;
839 case Opt_resuid:
840 if (args->from && match_int(args, &arg))
841 return -EINVAL;
842 uid = make_kuid(current_user_ns(), arg);
843 if (!uid_valid(uid)) {
844 f2fs_err(sbi, "Invalid uid value %d", arg);
845 return -EINVAL;
846 }
847 F2FS_OPTION(sbi).s_resuid = uid;
848 break;
849 case Opt_resgid:
850 if (args->from && match_int(args, &arg))
851 return -EINVAL;
852 gid = make_kgid(current_user_ns(), arg);
853 if (!gid_valid(gid)) {
854 f2fs_err(sbi, "Invalid gid value %d", arg);
855 return -EINVAL;
856 }
857 F2FS_OPTION(sbi).s_resgid = gid;
858 break;
859 case Opt_mode:
860 name = match_strdup(&args[0]);
861
862 if (!name)
863 return -ENOMEM;
864 if (!strcmp(name, "adaptive")) {
865 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
866 } else if (!strcmp(name, "lfs")) {
867 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
868 } else if (!strcmp(name, "fragment:segment")) {
869 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
870 } else if (!strcmp(name, "fragment:block")) {
871 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
872 } else {
873 kfree(name);
874 return -EINVAL;
875 }
876 kfree(name);
877 break;
878 #ifdef CONFIG_F2FS_FAULT_INJECTION
879 case Opt_fault_injection:
880 if (args->from && match_int(args, &arg))
881 return -EINVAL;
882 if (f2fs_build_fault_attr(sbi, arg,
883 F2FS_ALL_FAULT_TYPE))
884 return -EINVAL;
885 set_opt(sbi, FAULT_INJECTION);
886 break;
887
888 case Opt_fault_type:
889 if (args->from && match_int(args, &arg))
890 return -EINVAL;
891 if (f2fs_build_fault_attr(sbi, 0, arg))
892 return -EINVAL;
893 set_opt(sbi, FAULT_INJECTION);
894 break;
895 #else
896 case Opt_fault_injection:
897 f2fs_info(sbi, "fault_injection options not supported");
898 break;
899
900 case Opt_fault_type:
901 f2fs_info(sbi, "fault_type options not supported");
902 break;
903 #endif
904 case Opt_lazytime:
905 sb->s_flags |= SB_LAZYTIME;
906 break;
907 case Opt_nolazytime:
908 sb->s_flags &= ~SB_LAZYTIME;
909 break;
910 #ifdef CONFIG_QUOTA
911 case Opt_quota:
912 case Opt_usrquota:
913 set_opt(sbi, USRQUOTA);
914 break;
915 case Opt_grpquota:
916 set_opt(sbi, GRPQUOTA);
917 break;
918 case Opt_prjquota:
919 set_opt(sbi, PRJQUOTA);
920 break;
921 case Opt_usrjquota:
922 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
923 if (ret)
924 return ret;
925 break;
926 case Opt_grpjquota:
927 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
928 if (ret)
929 return ret;
930 break;
931 case Opt_prjjquota:
932 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
933 if (ret)
934 return ret;
935 break;
936 case Opt_offusrjquota:
937 ret = f2fs_clear_qf_name(sb, USRQUOTA);
938 if (ret)
939 return ret;
940 break;
941 case Opt_offgrpjquota:
942 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
943 if (ret)
944 return ret;
945 break;
946 case Opt_offprjjquota:
947 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
948 if (ret)
949 return ret;
950 break;
951 case Opt_jqfmt_vfsold:
952 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
953 break;
954 case Opt_jqfmt_vfsv0:
955 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
956 break;
957 case Opt_jqfmt_vfsv1:
958 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
959 break;
960 case Opt_noquota:
961 clear_opt(sbi, QUOTA);
962 clear_opt(sbi, USRQUOTA);
963 clear_opt(sbi, GRPQUOTA);
964 clear_opt(sbi, PRJQUOTA);
965 break;
966 #else
967 case Opt_quota:
968 case Opt_usrquota:
969 case Opt_grpquota:
970 case Opt_prjquota:
971 case Opt_usrjquota:
972 case Opt_grpjquota:
973 case Opt_prjjquota:
974 case Opt_offusrjquota:
975 case Opt_offgrpjquota:
976 case Opt_offprjjquota:
977 case Opt_jqfmt_vfsold:
978 case Opt_jqfmt_vfsv0:
979 case Opt_jqfmt_vfsv1:
980 case Opt_noquota:
981 f2fs_info(sbi, "quota operations not supported");
982 break;
983 #endif
984 case Opt_alloc:
985 name = match_strdup(&args[0]);
986 if (!name)
987 return -ENOMEM;
988
989 if (!strcmp(name, "default")) {
990 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
991 } else if (!strcmp(name, "reuse")) {
992 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
993 } else {
994 kfree(name);
995 return -EINVAL;
996 }
997 kfree(name);
998 break;
999 case Opt_fsync:
1000 name = match_strdup(&args[0]);
1001 if (!name)
1002 return -ENOMEM;
1003 if (!strcmp(name, "posix")) {
1004 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1005 } else if (!strcmp(name, "strict")) {
1006 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1007 } else if (!strcmp(name, "nobarrier")) {
1008 F2FS_OPTION(sbi).fsync_mode =
1009 FSYNC_MODE_NOBARRIER;
1010 } else {
1011 kfree(name);
1012 return -EINVAL;
1013 }
1014 kfree(name);
1015 break;
1016 case Opt_test_dummy_encryption:
1017 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1018 is_remount);
1019 if (ret)
1020 return ret;
1021 break;
1022 case Opt_inlinecrypt:
1023 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1024 sb->s_flags |= SB_INLINECRYPT;
1025 #else
1026 f2fs_info(sbi, "inline encryption not supported");
1027 #endif
1028 break;
1029 case Opt_checkpoint_disable_cap_perc:
1030 if (args->from && match_int(args, &arg))
1031 return -EINVAL;
1032 if (arg < 0 || arg > 100)
1033 return -EINVAL;
1034 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1035 set_opt(sbi, DISABLE_CHECKPOINT);
1036 break;
1037 case Opt_checkpoint_disable_cap:
1038 if (args->from && match_int(args, &arg))
1039 return -EINVAL;
1040 F2FS_OPTION(sbi).unusable_cap = arg;
1041 set_opt(sbi, DISABLE_CHECKPOINT);
1042 break;
1043 case Opt_checkpoint_disable:
1044 set_opt(sbi, DISABLE_CHECKPOINT);
1045 break;
1046 case Opt_checkpoint_enable:
1047 clear_opt(sbi, DISABLE_CHECKPOINT);
1048 break;
1049 case Opt_checkpoint_merge:
1050 set_opt(sbi, MERGE_CHECKPOINT);
1051 break;
1052 case Opt_nocheckpoint_merge:
1053 clear_opt(sbi, MERGE_CHECKPOINT);
1054 break;
1055 #ifdef CONFIG_F2FS_FS_COMPRESSION
1056 case Opt_compress_algorithm:
1057 if (!f2fs_sb_has_compression(sbi)) {
1058 f2fs_info(sbi, "Image doesn't support compression");
1059 break;
1060 }
1061 name = match_strdup(&args[0]);
1062 if (!name)
1063 return -ENOMEM;
1064 if (!strcmp(name, "lzo")) {
1065 #ifdef CONFIG_F2FS_FS_LZO
1066 F2FS_OPTION(sbi).compress_level = 0;
1067 F2FS_OPTION(sbi).compress_algorithm =
1068 COMPRESS_LZO;
1069 #else
1070 f2fs_info(sbi, "kernel doesn't support lzo compression");
1071 #endif
1072 } else if (!strncmp(name, "lz4", 3)) {
1073 #ifdef CONFIG_F2FS_FS_LZ4
1074 ret = f2fs_set_lz4hc_level(sbi, name);
1075 if (ret) {
1076 kfree(name);
1077 return -EINVAL;
1078 }
1079 F2FS_OPTION(sbi).compress_algorithm =
1080 COMPRESS_LZ4;
1081 #else
1082 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1083 #endif
1084 } else if (!strncmp(name, "zstd", 4)) {
1085 #ifdef CONFIG_F2FS_FS_ZSTD
1086 ret = f2fs_set_zstd_level(sbi, name);
1087 if (ret) {
1088 kfree(name);
1089 return -EINVAL;
1090 }
1091 F2FS_OPTION(sbi).compress_algorithm =
1092 COMPRESS_ZSTD;
1093 #else
1094 f2fs_info(sbi, "kernel doesn't support zstd compression");
1095 #endif
1096 } else if (!strcmp(name, "lzo-rle")) {
1097 #ifdef CONFIG_F2FS_FS_LZORLE
1098 F2FS_OPTION(sbi).compress_level = 0;
1099 F2FS_OPTION(sbi).compress_algorithm =
1100 COMPRESS_LZORLE;
1101 #else
1102 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1103 #endif
1104 } else {
1105 kfree(name);
1106 return -EINVAL;
1107 }
1108 kfree(name);
1109 break;
1110 case Opt_compress_log_size:
1111 if (!f2fs_sb_has_compression(sbi)) {
1112 f2fs_info(sbi, "Image doesn't support compression");
1113 break;
1114 }
1115 if (args->from && match_int(args, &arg))
1116 return -EINVAL;
1117 if (arg < MIN_COMPRESS_LOG_SIZE ||
1118 arg > MAX_COMPRESS_LOG_SIZE) {
1119 f2fs_err(sbi,
1120 "Compress cluster log size is out of range");
1121 return -EINVAL;
1122 }
1123 F2FS_OPTION(sbi).compress_log_size = arg;
1124 break;
1125 case Opt_compress_extension:
1126 if (!f2fs_sb_has_compression(sbi)) {
1127 f2fs_info(sbi, "Image doesn't support compression");
1128 break;
1129 }
1130 name = match_strdup(&args[0]);
1131 if (!name)
1132 return -ENOMEM;
1133
1134 ext = F2FS_OPTION(sbi).extensions;
1135 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1136
1137 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1138 ext_cnt >= COMPRESS_EXT_NUM) {
1139 f2fs_err(sbi,
1140 "invalid extension length/number");
1141 kfree(name);
1142 return -EINVAL;
1143 }
1144
1145 if (is_compress_extension_exist(sbi, name, true)) {
1146 kfree(name);
1147 break;
1148 }
1149
1150 strcpy(ext[ext_cnt], name);
1151 F2FS_OPTION(sbi).compress_ext_cnt++;
1152 kfree(name);
1153 break;
1154 case Opt_nocompress_extension:
1155 if (!f2fs_sb_has_compression(sbi)) {
1156 f2fs_info(sbi, "Image doesn't support compression");
1157 break;
1158 }
1159 name = match_strdup(&args[0]);
1160 if (!name)
1161 return -ENOMEM;
1162
1163 noext = F2FS_OPTION(sbi).noextensions;
1164 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1165
1166 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1167 noext_cnt >= COMPRESS_EXT_NUM) {
1168 f2fs_err(sbi,
1169 "invalid extension length/number");
1170 kfree(name);
1171 return -EINVAL;
1172 }
1173
1174 if (is_compress_extension_exist(sbi, name, false)) {
1175 kfree(name);
1176 break;
1177 }
1178
1179 strcpy(noext[noext_cnt], name);
1180 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1181 kfree(name);
1182 break;
1183 case Opt_compress_chksum:
1184 if (!f2fs_sb_has_compression(sbi)) {
1185 f2fs_info(sbi, "Image doesn't support compression");
1186 break;
1187 }
1188 F2FS_OPTION(sbi).compress_chksum = true;
1189 break;
1190 case Opt_compress_mode:
1191 if (!f2fs_sb_has_compression(sbi)) {
1192 f2fs_info(sbi, "Image doesn't support compression");
1193 break;
1194 }
1195 name = match_strdup(&args[0]);
1196 if (!name)
1197 return -ENOMEM;
1198 if (!strcmp(name, "fs")) {
1199 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1200 } else if (!strcmp(name, "user")) {
1201 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1202 } else {
1203 kfree(name);
1204 return -EINVAL;
1205 }
1206 kfree(name);
1207 break;
1208 case Opt_compress_cache:
1209 if (!f2fs_sb_has_compression(sbi)) {
1210 f2fs_info(sbi, "Image doesn't support compression");
1211 break;
1212 }
1213 set_opt(sbi, COMPRESS_CACHE);
1214 break;
1215 #else
1216 case Opt_compress_algorithm:
1217 case Opt_compress_log_size:
1218 case Opt_compress_extension:
1219 case Opt_nocompress_extension:
1220 case Opt_compress_chksum:
1221 case Opt_compress_mode:
1222 case Opt_compress_cache:
1223 f2fs_info(sbi, "compression options not supported");
1224 break;
1225 #endif
1226 case Opt_atgc:
1227 set_opt(sbi, ATGC);
1228 break;
1229 case Opt_gc_merge:
1230 set_opt(sbi, GC_MERGE);
1231 break;
1232 case Opt_nogc_merge:
1233 clear_opt(sbi, GC_MERGE);
1234 break;
1235 case Opt_discard_unit:
1236 name = match_strdup(&args[0]);
1237 if (!name)
1238 return -ENOMEM;
1239 if (!strcmp(name, "block")) {
1240 F2FS_OPTION(sbi).discard_unit =
1241 DISCARD_UNIT_BLOCK;
1242 } else if (!strcmp(name, "segment")) {
1243 F2FS_OPTION(sbi).discard_unit =
1244 DISCARD_UNIT_SEGMENT;
1245 } else if (!strcmp(name, "section")) {
1246 F2FS_OPTION(sbi).discard_unit =
1247 DISCARD_UNIT_SECTION;
1248 } else {
1249 kfree(name);
1250 return -EINVAL;
1251 }
1252 kfree(name);
1253 break;
1254 case Opt_memory_mode:
1255 name = match_strdup(&args[0]);
1256 if (!name)
1257 return -ENOMEM;
1258 if (!strcmp(name, "normal")) {
1259 F2FS_OPTION(sbi).memory_mode =
1260 MEMORY_MODE_NORMAL;
1261 } else if (!strcmp(name, "low")) {
1262 F2FS_OPTION(sbi).memory_mode =
1263 MEMORY_MODE_LOW;
1264 } else {
1265 kfree(name);
1266 return -EINVAL;
1267 }
1268 kfree(name);
1269 break;
1270 case Opt_age_extent_cache:
1271 set_opt(sbi, AGE_EXTENT_CACHE);
1272 break;
1273 case Opt_errors:
1274 name = match_strdup(&args[0]);
1275 if (!name)
1276 return -ENOMEM;
1277 if (!strcmp(name, "remount-ro")) {
1278 F2FS_OPTION(sbi).errors =
1279 MOUNT_ERRORS_READONLY;
1280 } else if (!strcmp(name, "continue")) {
1281 F2FS_OPTION(sbi).errors =
1282 MOUNT_ERRORS_CONTINUE;
1283 } else if (!strcmp(name, "panic")) {
1284 F2FS_OPTION(sbi).errors =
1285 MOUNT_ERRORS_PANIC;
1286 } else {
1287 kfree(name);
1288 return -EINVAL;
1289 }
1290 kfree(name);
1291 break;
1292 default:
1293 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1294 p);
1295 return -EINVAL;
1296 }
1297 }
1298 default_check:
1299 #ifdef CONFIG_QUOTA
1300 if (f2fs_check_quota_options(sbi))
1301 return -EINVAL;
1302 #else
1303 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1304 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1305 return -EINVAL;
1306 }
1307 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1308 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1309 return -EINVAL;
1310 }
1311 #endif
1312 #if !IS_ENABLED(CONFIG_UNICODE)
1313 if (f2fs_sb_has_casefold(sbi)) {
1314 f2fs_err(sbi,
1315 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1316 return -EINVAL;
1317 }
1318 #endif
1319 /*
1320 * The BLKZONED feature indicates that the drive was formatted with
1321 * zone alignment optimization. This is optional for host-aware
1322 * devices, but mandatory for host-managed zoned block devices.
1323 */
1324 if (f2fs_sb_has_blkzoned(sbi)) {
1325 #ifdef CONFIG_BLK_DEV_ZONED
1326 if (F2FS_OPTION(sbi).discard_unit !=
1327 DISCARD_UNIT_SECTION) {
1328 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1329 F2FS_OPTION(sbi).discard_unit =
1330 DISCARD_UNIT_SECTION;
1331 }
1332
1333 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1334 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1335 return -EINVAL;
1336 }
1337 #else
1338 f2fs_err(sbi, "Zoned block device support is not enabled");
1339 return -EINVAL;
1340 #endif
1341 }
1342
1343 #ifdef CONFIG_F2FS_FS_COMPRESSION
1344 if (f2fs_test_compress_extension(sbi)) {
1345 f2fs_err(sbi, "invalid compress or nocompress extension");
1346 return -EINVAL;
1347 }
1348 #endif
1349
1350 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1351 int min_size, max_size;
1352
1353 if (!f2fs_sb_has_extra_attr(sbi) ||
1354 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1355 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1356 return -EINVAL;
1357 }
1358 if (!test_opt(sbi, INLINE_XATTR)) {
1359 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1360 return -EINVAL;
1361 }
1362
1363 min_size = MIN_INLINE_XATTR_SIZE;
1364 max_size = MAX_INLINE_XATTR_SIZE;
1365
1366 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1367 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1368 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1369 min_size, max_size);
1370 return -EINVAL;
1371 }
1372 }
1373
1374 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1375 f2fs_err(sbi, "LFS is not compatible with checkpoint=disable");
1376 return -EINVAL;
1377 }
1378
1379 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1380 f2fs_err(sbi, "LFS is not compatible with ATGC");
1381 return -EINVAL;
1382 }
1383
1384 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1385 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1386 return -EINVAL;
1387 }
1388
1389 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1390 f2fs_err(sbi, "Allow to mount readonly mode only");
1391 return -EROFS;
1392 }
1393 return 0;
1394 }
1395
f2fs_alloc_inode(struct super_block * sb)1396 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1397 {
1398 struct f2fs_inode_info *fi;
1399
1400 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1401 return NULL;
1402
1403 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1404 if (!fi)
1405 return NULL;
1406
1407 init_once((void *) fi);
1408
1409 /* Initialize f2fs-specific inode info */
1410 atomic_set(&fi->dirty_pages, 0);
1411 atomic_set(&fi->i_compr_blocks, 0);
1412 init_f2fs_rwsem(&fi->i_sem);
1413 spin_lock_init(&fi->i_size_lock);
1414 INIT_LIST_HEAD(&fi->dirty_list);
1415 INIT_LIST_HEAD(&fi->gdirty_list);
1416 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1417 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1418 init_f2fs_rwsem(&fi->i_xattr_sem);
1419
1420 /* Will be used by directory only */
1421 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1422
1423 return &fi->vfs_inode;
1424 }
1425
f2fs_drop_inode(struct inode * inode)1426 static int f2fs_drop_inode(struct inode *inode)
1427 {
1428 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1429 int ret;
1430
1431 /*
1432 * during filesystem shutdown, if checkpoint is disabled,
1433 * drop useless meta/node dirty pages.
1434 */
1435 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1436 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1437 inode->i_ino == F2FS_META_INO(sbi)) {
1438 trace_f2fs_drop_inode(inode, 1);
1439 return 1;
1440 }
1441 }
1442
1443 /*
1444 * This is to avoid a deadlock condition like below.
1445 * writeback_single_inode(inode)
1446 * - f2fs_write_data_page
1447 * - f2fs_gc -> iput -> evict
1448 * - inode_wait_for_writeback(inode)
1449 */
1450 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1451 if (!inode->i_nlink && !is_bad_inode(inode)) {
1452 /* to avoid evict_inode call simultaneously */
1453 atomic_inc(&inode->i_count);
1454 spin_unlock(&inode->i_lock);
1455
1456 /* should remain fi->extent_tree for writepage */
1457 f2fs_destroy_extent_node(inode);
1458
1459 sb_start_intwrite(inode->i_sb);
1460 f2fs_i_size_write(inode, 0);
1461
1462 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1463 inode, NULL, 0, DATA);
1464 truncate_inode_pages_final(inode->i_mapping);
1465
1466 if (F2FS_HAS_BLOCKS(inode))
1467 f2fs_truncate(inode);
1468
1469 sb_end_intwrite(inode->i_sb);
1470
1471 spin_lock(&inode->i_lock);
1472 atomic_dec(&inode->i_count);
1473 }
1474 trace_f2fs_drop_inode(inode, 0);
1475 return 0;
1476 }
1477 ret = generic_drop_inode(inode);
1478 if (!ret)
1479 ret = fscrypt_drop_inode(inode);
1480 trace_f2fs_drop_inode(inode, ret);
1481 return ret;
1482 }
1483
f2fs_inode_dirtied(struct inode * inode,bool sync)1484 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1485 {
1486 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1487 int ret = 0;
1488
1489 spin_lock(&sbi->inode_lock[DIRTY_META]);
1490 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1491 ret = 1;
1492 } else {
1493 set_inode_flag(inode, FI_DIRTY_INODE);
1494 stat_inc_dirty_inode(sbi, DIRTY_META);
1495 }
1496 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1497 list_add_tail(&F2FS_I(inode)->gdirty_list,
1498 &sbi->inode_list[DIRTY_META]);
1499 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1500 }
1501 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1502 return ret;
1503 }
1504
f2fs_inode_synced(struct inode * inode)1505 void f2fs_inode_synced(struct inode *inode)
1506 {
1507 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1508
1509 spin_lock(&sbi->inode_lock[DIRTY_META]);
1510 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1511 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1512 return;
1513 }
1514 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1515 list_del_init(&F2FS_I(inode)->gdirty_list);
1516 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1517 }
1518 clear_inode_flag(inode, FI_DIRTY_INODE);
1519 clear_inode_flag(inode, FI_AUTO_RECOVER);
1520 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1521 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1522 }
1523
1524 /*
1525 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1526 *
1527 * We should call set_dirty_inode to write the dirty inode through write_inode.
1528 */
f2fs_dirty_inode(struct inode * inode,int flags)1529 static void f2fs_dirty_inode(struct inode *inode, int flags)
1530 {
1531 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1532
1533 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1534 inode->i_ino == F2FS_META_INO(sbi))
1535 return;
1536
1537 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1538 clear_inode_flag(inode, FI_AUTO_RECOVER);
1539
1540 f2fs_inode_dirtied(inode, false);
1541 }
1542
f2fs_free_inode(struct inode * inode)1543 static void f2fs_free_inode(struct inode *inode)
1544 {
1545 fscrypt_free_inode(inode);
1546 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1547 }
1548
destroy_percpu_info(struct f2fs_sb_info * sbi)1549 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1550 {
1551 percpu_counter_destroy(&sbi->total_valid_inode_count);
1552 percpu_counter_destroy(&sbi->rf_node_block_count);
1553 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1554 }
1555
destroy_device_list(struct f2fs_sb_info * sbi)1556 static void destroy_device_list(struct f2fs_sb_info *sbi)
1557 {
1558 int i;
1559
1560 for (i = 0; i < sbi->s_ndevs; i++) {
1561 if (i > 0)
1562 blkdev_put(FDEV(i).bdev, sbi->sb);
1563 #ifdef CONFIG_BLK_DEV_ZONED
1564 kvfree(FDEV(i).blkz_seq);
1565 #endif
1566 }
1567 kvfree(sbi->devs);
1568 }
1569
f2fs_put_super(struct super_block * sb)1570 static void f2fs_put_super(struct super_block *sb)
1571 {
1572 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1573 int i;
1574 int err = 0;
1575 bool done;
1576
1577 /* unregister procfs/sysfs entries in advance to avoid race case */
1578 f2fs_unregister_sysfs(sbi);
1579
1580 f2fs_quota_off_umount(sb);
1581
1582 /* prevent remaining shrinker jobs */
1583 mutex_lock(&sbi->umount_mutex);
1584
1585 /*
1586 * flush all issued checkpoints and stop checkpoint issue thread.
1587 * after then, all checkpoints should be done by each process context.
1588 */
1589 f2fs_stop_ckpt_thread(sbi);
1590
1591 /*
1592 * We don't need to do checkpoint when superblock is clean.
1593 * But, the previous checkpoint was not done by umount, it needs to do
1594 * clean checkpoint again.
1595 */
1596 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1597 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1598 struct cp_control cpc = {
1599 .reason = CP_UMOUNT,
1600 };
1601 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1602 err = f2fs_write_checkpoint(sbi, &cpc);
1603 }
1604
1605 /* be sure to wait for any on-going discard commands */
1606 done = f2fs_issue_discard_timeout(sbi);
1607 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1608 struct cp_control cpc = {
1609 .reason = CP_UMOUNT | CP_TRIMMED,
1610 };
1611 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1612 err = f2fs_write_checkpoint(sbi, &cpc);
1613 }
1614
1615 /*
1616 * normally superblock is clean, so we need to release this.
1617 * In addition, EIO will skip do checkpoint, we need this as well.
1618 */
1619 f2fs_release_ino_entry(sbi, true);
1620
1621 f2fs_leave_shrinker(sbi);
1622 mutex_unlock(&sbi->umount_mutex);
1623
1624 /* our cp_error case, we can wait for any writeback page */
1625 f2fs_flush_merged_writes(sbi);
1626
1627 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1628
1629 if (err || f2fs_cp_error(sbi)) {
1630 truncate_inode_pages_final(NODE_MAPPING(sbi));
1631 truncate_inode_pages_final(META_MAPPING(sbi));
1632 }
1633
1634 for (i = 0; i < NR_COUNT_TYPE; i++) {
1635 if (!get_pages(sbi, i))
1636 continue;
1637 f2fs_err(sbi, "detect filesystem reference count leak during "
1638 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1639 f2fs_bug_on(sbi, 1);
1640 }
1641
1642 f2fs_bug_on(sbi, sbi->fsync_node_num);
1643
1644 f2fs_destroy_compress_inode(sbi);
1645
1646 iput(sbi->node_inode);
1647 sbi->node_inode = NULL;
1648
1649 iput(sbi->meta_inode);
1650 sbi->meta_inode = NULL;
1651
1652 /*
1653 * iput() can update stat information, if f2fs_write_checkpoint()
1654 * above failed with error.
1655 */
1656 f2fs_destroy_stats(sbi);
1657
1658 /* destroy f2fs internal modules */
1659 f2fs_destroy_node_manager(sbi);
1660 f2fs_destroy_segment_manager(sbi);
1661
1662 /* flush s_error_work before sbi destroy */
1663 flush_work(&sbi->s_error_work);
1664
1665 f2fs_destroy_post_read_wq(sbi);
1666
1667 kvfree(sbi->ckpt);
1668
1669 sb->s_fs_info = NULL;
1670 if (sbi->s_chksum_driver)
1671 crypto_free_shash(sbi->s_chksum_driver);
1672 kfree(sbi->raw_super);
1673
1674 destroy_device_list(sbi);
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 kfree(sbi);
1690 }
1691
f2fs_sync_fs(struct super_block * sb,int sync)1692 int f2fs_sync_fs(struct super_block *sb, int sync)
1693 {
1694 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1695 int err = 0;
1696
1697 if (unlikely(f2fs_cp_error(sbi)))
1698 return 0;
1699 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1700 return 0;
1701
1702 trace_f2fs_sync_fs(sb, sync);
1703
1704 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1705 return -EAGAIN;
1706
1707 if (sync) {
1708 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1709 err = f2fs_issue_checkpoint(sbi);
1710 }
1711
1712 return err;
1713 }
1714
f2fs_freeze(struct super_block * sb)1715 static int f2fs_freeze(struct super_block *sb)
1716 {
1717 if (f2fs_readonly(sb))
1718 return 0;
1719
1720 /* IO error happened before */
1721 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1722 return -EIO;
1723
1724 /* must be clean, since sync_filesystem() was already called */
1725 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1726 return -EINVAL;
1727
1728 /* Let's flush checkpoints and stop the thread. */
1729 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1730
1731 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1732 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1733 return 0;
1734 }
1735
f2fs_unfreeze(struct super_block * sb)1736 static int f2fs_unfreeze(struct super_block *sb)
1737 {
1738 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1739
1740 /*
1741 * It will update discard_max_bytes of mounted lvm device to zero
1742 * after creating snapshot on this lvm device, let's drop all
1743 * remained discards.
1744 * We don't need to disable real-time discard because discard_max_bytes
1745 * will recover after removal of snapshot.
1746 */
1747 if (test_opt(sbi, DISCARD) && !f2fs_hw_support_discard(sbi))
1748 f2fs_issue_discard_timeout(sbi);
1749
1750 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1751 return 0;
1752 }
1753
1754 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1755 static int f2fs_statfs_project(struct super_block *sb,
1756 kprojid_t projid, struct kstatfs *buf)
1757 {
1758 struct kqid qid;
1759 struct dquot *dquot;
1760 u64 limit;
1761 u64 curblock;
1762
1763 qid = make_kqid_projid(projid);
1764 dquot = dqget(sb, qid);
1765 if (IS_ERR(dquot))
1766 return PTR_ERR(dquot);
1767 spin_lock(&dquot->dq_dqb_lock);
1768
1769 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1770 dquot->dq_dqb.dqb_bhardlimit);
1771 if (limit)
1772 limit >>= sb->s_blocksize_bits;
1773
1774 if (limit && buf->f_blocks > limit) {
1775 curblock = (dquot->dq_dqb.dqb_curspace +
1776 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1777 buf->f_blocks = limit;
1778 buf->f_bfree = buf->f_bavail =
1779 (buf->f_blocks > curblock) ?
1780 (buf->f_blocks - curblock) : 0;
1781 }
1782
1783 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1784 dquot->dq_dqb.dqb_ihardlimit);
1785
1786 if (limit && buf->f_files > limit) {
1787 buf->f_files = limit;
1788 buf->f_ffree =
1789 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1790 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1791 }
1792
1793 spin_unlock(&dquot->dq_dqb_lock);
1794 dqput(dquot);
1795 return 0;
1796 }
1797 #endif
1798
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1799 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1800 {
1801 struct super_block *sb = dentry->d_sb;
1802 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1803 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1804 block_t total_count, user_block_count, start_count;
1805 u64 avail_node_count;
1806 unsigned int total_valid_node_count;
1807
1808 total_count = le64_to_cpu(sbi->raw_super->block_count);
1809 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1810 buf->f_type = F2FS_SUPER_MAGIC;
1811 buf->f_bsize = sbi->blocksize;
1812
1813 buf->f_blocks = total_count - start_count;
1814
1815 spin_lock(&sbi->stat_lock);
1816
1817 user_block_count = sbi->user_block_count;
1818 total_valid_node_count = valid_node_count(sbi);
1819 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1820 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1821 sbi->current_reserved_blocks;
1822
1823 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1824 buf->f_bfree = 0;
1825 else
1826 buf->f_bfree -= sbi->unusable_block_count;
1827 spin_unlock(&sbi->stat_lock);
1828
1829 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1830 buf->f_bavail = buf->f_bfree -
1831 F2FS_OPTION(sbi).root_reserved_blocks;
1832 else
1833 buf->f_bavail = 0;
1834
1835 if (avail_node_count > user_block_count) {
1836 buf->f_files = user_block_count;
1837 buf->f_ffree = buf->f_bavail;
1838 } else {
1839 buf->f_files = avail_node_count;
1840 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1841 buf->f_bavail);
1842 }
1843
1844 buf->f_namelen = F2FS_NAME_LEN;
1845 buf->f_fsid = u64_to_fsid(id);
1846
1847 #ifdef CONFIG_QUOTA
1848 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1849 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1850 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1851 }
1852 #endif
1853 return 0;
1854 }
1855
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1856 static inline void f2fs_show_quota_options(struct seq_file *seq,
1857 struct super_block *sb)
1858 {
1859 #ifdef CONFIG_QUOTA
1860 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1861
1862 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1863 char *fmtname = "";
1864
1865 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1866 case QFMT_VFS_OLD:
1867 fmtname = "vfsold";
1868 break;
1869 case QFMT_VFS_V0:
1870 fmtname = "vfsv0";
1871 break;
1872 case QFMT_VFS_V1:
1873 fmtname = "vfsv1";
1874 break;
1875 }
1876 seq_printf(seq, ",jqfmt=%s", fmtname);
1877 }
1878
1879 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1880 seq_show_option(seq, "usrjquota",
1881 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1882
1883 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1884 seq_show_option(seq, "grpjquota",
1885 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1886
1887 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1888 seq_show_option(seq, "prjjquota",
1889 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1890 #endif
1891 }
1892
1893 #ifdef CONFIG_F2FS_FS_COMPRESSION
f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1894 static inline void f2fs_show_compress_options(struct seq_file *seq,
1895 struct super_block *sb)
1896 {
1897 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1898 char *algtype = "";
1899 int i;
1900
1901 if (!f2fs_sb_has_compression(sbi))
1902 return;
1903
1904 switch (F2FS_OPTION(sbi).compress_algorithm) {
1905 case COMPRESS_LZO:
1906 algtype = "lzo";
1907 break;
1908 case COMPRESS_LZ4:
1909 algtype = "lz4";
1910 break;
1911 case COMPRESS_ZSTD:
1912 algtype = "zstd";
1913 break;
1914 case COMPRESS_LZORLE:
1915 algtype = "lzo-rle";
1916 break;
1917 }
1918 seq_printf(seq, ",compress_algorithm=%s", algtype);
1919
1920 if (F2FS_OPTION(sbi).compress_level)
1921 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1922
1923 seq_printf(seq, ",compress_log_size=%u",
1924 F2FS_OPTION(sbi).compress_log_size);
1925
1926 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1927 seq_printf(seq, ",compress_extension=%s",
1928 F2FS_OPTION(sbi).extensions[i]);
1929 }
1930
1931 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1932 seq_printf(seq, ",nocompress_extension=%s",
1933 F2FS_OPTION(sbi).noextensions[i]);
1934 }
1935
1936 if (F2FS_OPTION(sbi).compress_chksum)
1937 seq_puts(seq, ",compress_chksum");
1938
1939 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1940 seq_printf(seq, ",compress_mode=%s", "fs");
1941 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1942 seq_printf(seq, ",compress_mode=%s", "user");
1943
1944 if (test_opt(sbi, COMPRESS_CACHE))
1945 seq_puts(seq, ",compress_cache");
1946 }
1947 #endif
1948
f2fs_show_options(struct seq_file * seq,struct dentry * root)1949 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1950 {
1951 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1952
1953 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1954 seq_printf(seq, ",background_gc=%s", "sync");
1955 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1956 seq_printf(seq, ",background_gc=%s", "on");
1957 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1958 seq_printf(seq, ",background_gc=%s", "off");
1959
1960 if (test_opt(sbi, GC_MERGE))
1961 seq_puts(seq, ",gc_merge");
1962 else
1963 seq_puts(seq, ",nogc_merge");
1964
1965 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1966 seq_puts(seq, ",disable_roll_forward");
1967 if (test_opt(sbi, NORECOVERY))
1968 seq_puts(seq, ",norecovery");
1969 if (test_opt(sbi, DISCARD)) {
1970 seq_puts(seq, ",discard");
1971 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1972 seq_printf(seq, ",discard_unit=%s", "block");
1973 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1974 seq_printf(seq, ",discard_unit=%s", "segment");
1975 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1976 seq_printf(seq, ",discard_unit=%s", "section");
1977 } else {
1978 seq_puts(seq, ",nodiscard");
1979 }
1980 #ifdef CONFIG_F2FS_FS_XATTR
1981 if (test_opt(sbi, XATTR_USER))
1982 seq_puts(seq, ",user_xattr");
1983 else
1984 seq_puts(seq, ",nouser_xattr");
1985 if (test_opt(sbi, INLINE_XATTR))
1986 seq_puts(seq, ",inline_xattr");
1987 else
1988 seq_puts(seq, ",noinline_xattr");
1989 if (test_opt(sbi, INLINE_XATTR_SIZE))
1990 seq_printf(seq, ",inline_xattr_size=%u",
1991 F2FS_OPTION(sbi).inline_xattr_size);
1992 #endif
1993 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1994 if (test_opt(sbi, POSIX_ACL))
1995 seq_puts(seq, ",acl");
1996 else
1997 seq_puts(seq, ",noacl");
1998 #endif
1999 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
2000 seq_puts(seq, ",disable_ext_identify");
2001 if (test_opt(sbi, INLINE_DATA))
2002 seq_puts(seq, ",inline_data");
2003 else
2004 seq_puts(seq, ",noinline_data");
2005 if (test_opt(sbi, INLINE_DENTRY))
2006 seq_puts(seq, ",inline_dentry");
2007 else
2008 seq_puts(seq, ",noinline_dentry");
2009 if (test_opt(sbi, FLUSH_MERGE))
2010 seq_puts(seq, ",flush_merge");
2011 else
2012 seq_puts(seq, ",noflush_merge");
2013 if (test_opt(sbi, NOBARRIER))
2014 seq_puts(seq, ",nobarrier");
2015 else
2016 seq_puts(seq, ",barrier");
2017 if (test_opt(sbi, FASTBOOT))
2018 seq_puts(seq, ",fastboot");
2019 if (test_opt(sbi, READ_EXTENT_CACHE))
2020 seq_puts(seq, ",extent_cache");
2021 else
2022 seq_puts(seq, ",noextent_cache");
2023 if (test_opt(sbi, AGE_EXTENT_CACHE))
2024 seq_puts(seq, ",age_extent_cache");
2025 if (test_opt(sbi, DATA_FLUSH))
2026 seq_puts(seq, ",data_flush");
2027
2028 seq_puts(seq, ",mode=");
2029 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2030 seq_puts(seq, "adaptive");
2031 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2032 seq_puts(seq, "lfs");
2033 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2034 seq_puts(seq, "fragment:segment");
2035 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2036 seq_puts(seq, "fragment:block");
2037 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2038 if (test_opt(sbi, RESERVE_ROOT))
2039 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2040 F2FS_OPTION(sbi).root_reserved_blocks,
2041 from_kuid_munged(&init_user_ns,
2042 F2FS_OPTION(sbi).s_resuid),
2043 from_kgid_munged(&init_user_ns,
2044 F2FS_OPTION(sbi).s_resgid));
2045 #ifdef CONFIG_F2FS_FAULT_INJECTION
2046 if (test_opt(sbi, FAULT_INJECTION)) {
2047 seq_printf(seq, ",fault_injection=%u",
2048 F2FS_OPTION(sbi).fault_info.inject_rate);
2049 seq_printf(seq, ",fault_type=%u",
2050 F2FS_OPTION(sbi).fault_info.inject_type);
2051 }
2052 #endif
2053 #ifdef CONFIG_QUOTA
2054 if (test_opt(sbi, QUOTA))
2055 seq_puts(seq, ",quota");
2056 if (test_opt(sbi, USRQUOTA))
2057 seq_puts(seq, ",usrquota");
2058 if (test_opt(sbi, GRPQUOTA))
2059 seq_puts(seq, ",grpquota");
2060 if (test_opt(sbi, PRJQUOTA))
2061 seq_puts(seq, ",prjquota");
2062 #endif
2063 f2fs_show_quota_options(seq, sbi->sb);
2064
2065 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2066
2067 if (sbi->sb->s_flags & SB_INLINECRYPT)
2068 seq_puts(seq, ",inlinecrypt");
2069
2070 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2071 seq_printf(seq, ",alloc_mode=%s", "default");
2072 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2073 seq_printf(seq, ",alloc_mode=%s", "reuse");
2074
2075 if (test_opt(sbi, DISABLE_CHECKPOINT))
2076 seq_printf(seq, ",checkpoint=disable:%u",
2077 F2FS_OPTION(sbi).unusable_cap);
2078 if (test_opt(sbi, MERGE_CHECKPOINT))
2079 seq_puts(seq, ",checkpoint_merge");
2080 else
2081 seq_puts(seq, ",nocheckpoint_merge");
2082 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2083 seq_printf(seq, ",fsync_mode=%s", "posix");
2084 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2085 seq_printf(seq, ",fsync_mode=%s", "strict");
2086 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2087 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2088
2089 #ifdef CONFIG_F2FS_FS_COMPRESSION
2090 f2fs_show_compress_options(seq, sbi->sb);
2091 #endif
2092
2093 if (test_opt(sbi, ATGC))
2094 seq_puts(seq, ",atgc");
2095
2096 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2097 seq_printf(seq, ",memory=%s", "normal");
2098 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2099 seq_printf(seq, ",memory=%s", "low");
2100
2101 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2102 seq_printf(seq, ",errors=%s", "remount-ro");
2103 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2104 seq_printf(seq, ",errors=%s", "continue");
2105 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2106 seq_printf(seq, ",errors=%s", "panic");
2107
2108 return 0;
2109 }
2110
default_options(struct f2fs_sb_info * sbi,bool remount)2111 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2112 {
2113 /* init some FS parameters */
2114 if (!remount) {
2115 set_opt(sbi, READ_EXTENT_CACHE);
2116 clear_opt(sbi, DISABLE_CHECKPOINT);
2117
2118 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2119 set_opt(sbi, DISCARD);
2120
2121 if (f2fs_sb_has_blkzoned(sbi))
2122 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2123 else
2124 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2125 }
2126
2127 if (f2fs_sb_has_readonly(sbi))
2128 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2129 else
2130 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2131
2132 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2133 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2134 SMALL_VOLUME_SEGMENTS)
2135 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2136 else
2137 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2138 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2139 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2140 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2141 if (f2fs_sb_has_compression(sbi)) {
2142 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2143 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2144 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2145 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2146 }
2147 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2148 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2149 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2150
2151 set_opt(sbi, INLINE_XATTR);
2152 set_opt(sbi, INLINE_DATA);
2153 set_opt(sbi, INLINE_DENTRY);
2154 set_opt(sbi, MERGE_CHECKPOINT);
2155 F2FS_OPTION(sbi).unusable_cap = 0;
2156 sbi->sb->s_flags |= SB_LAZYTIME;
2157 if (!f2fs_is_readonly(sbi))
2158 set_opt(sbi, FLUSH_MERGE);
2159 if (f2fs_sb_has_blkzoned(sbi))
2160 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2161 else
2162 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2163
2164 #ifdef CONFIG_F2FS_FS_XATTR
2165 set_opt(sbi, XATTR_USER);
2166 #endif
2167 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2168 set_opt(sbi, POSIX_ACL);
2169 #endif
2170
2171 f2fs_build_fault_attr(sbi, 0, 0);
2172 }
2173
2174 #ifdef CONFIG_QUOTA
2175 static int f2fs_enable_quotas(struct super_block *sb);
2176 #endif
2177
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)2178 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2179 {
2180 unsigned int s_flags = sbi->sb->s_flags;
2181 struct cp_control cpc;
2182 unsigned int gc_mode = sbi->gc_mode;
2183 int err = 0;
2184 int ret;
2185 block_t unusable;
2186
2187 if (s_flags & SB_RDONLY) {
2188 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2189 return -EINVAL;
2190 }
2191 sbi->sb->s_flags |= SB_ACTIVE;
2192
2193 /* check if we need more GC first */
2194 unusable = f2fs_get_unusable_blocks(sbi);
2195 if (!f2fs_disable_cp_again(sbi, unusable))
2196 goto skip_gc;
2197
2198 f2fs_update_time(sbi, DISABLE_TIME);
2199
2200 sbi->gc_mode = GC_URGENT_HIGH;
2201
2202 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2203 struct f2fs_gc_control gc_control = {
2204 .victim_segno = NULL_SEGNO,
2205 .init_gc_type = FG_GC,
2206 .should_migrate_blocks = false,
2207 .err_gc_skipped = true,
2208 .nr_free_secs = 1 };
2209
2210 f2fs_down_write(&sbi->gc_lock);
2211 stat_inc_gc_call_count(sbi, FOREGROUND);
2212 err = f2fs_gc(sbi, &gc_control);
2213 if (err == -ENODATA) {
2214 err = 0;
2215 break;
2216 }
2217 if (err && err != -EAGAIN)
2218 break;
2219 }
2220
2221 ret = sync_filesystem(sbi->sb);
2222 if (ret || err) {
2223 err = ret ? ret : err;
2224 goto restore_flag;
2225 }
2226
2227 unusable = f2fs_get_unusable_blocks(sbi);
2228 if (f2fs_disable_cp_again(sbi, unusable)) {
2229 err = -EAGAIN;
2230 goto restore_flag;
2231 }
2232
2233 skip_gc:
2234 f2fs_down_write(&sbi->gc_lock);
2235 cpc.reason = CP_PAUSE;
2236 set_sbi_flag(sbi, SBI_CP_DISABLED);
2237 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2238 err = f2fs_write_checkpoint(sbi, &cpc);
2239 if (err)
2240 goto out_unlock;
2241
2242 spin_lock(&sbi->stat_lock);
2243 sbi->unusable_block_count = unusable;
2244 spin_unlock(&sbi->stat_lock);
2245
2246 out_unlock:
2247 f2fs_up_write(&sbi->gc_lock);
2248 restore_flag:
2249 sbi->gc_mode = gc_mode;
2250 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2251 return err;
2252 }
2253
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2254 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2255 {
2256 int retry = DEFAULT_RETRY_IO_COUNT;
2257
2258 /* we should flush all the data to keep data consistency */
2259 do {
2260 sync_inodes_sb(sbi->sb);
2261 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2262 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2263
2264 if (unlikely(retry < 0))
2265 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2266
2267 f2fs_down_write(&sbi->gc_lock);
2268 f2fs_dirty_to_prefree(sbi);
2269
2270 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2271 set_sbi_flag(sbi, SBI_IS_DIRTY);
2272 f2fs_up_write(&sbi->gc_lock);
2273
2274 f2fs_sync_fs(sbi->sb, 1);
2275
2276 /* Let's ensure there's no pending checkpoint anymore */
2277 f2fs_flush_ckpt_thread(sbi);
2278 }
2279
f2fs_remount(struct super_block * sb,int * flags,char * data)2280 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2281 {
2282 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2283 struct f2fs_mount_info org_mount_opt;
2284 unsigned long old_sb_flags;
2285 int err;
2286 bool need_restart_gc = false, need_stop_gc = false;
2287 bool need_restart_ckpt = false, need_stop_ckpt = false;
2288 bool need_restart_flush = false, need_stop_flush = false;
2289 bool need_restart_discard = false, need_stop_discard = false;
2290 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2291 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2292 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2293 bool no_atgc = !test_opt(sbi, ATGC);
2294 bool no_discard = !test_opt(sbi, DISCARD);
2295 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2296 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2297 #ifdef CONFIG_QUOTA
2298 int i, j;
2299 #endif
2300
2301 /*
2302 * Save the old mount options in case we
2303 * need to restore them.
2304 */
2305 org_mount_opt = sbi->mount_opt;
2306 old_sb_flags = sb->s_flags;
2307
2308 #ifdef CONFIG_QUOTA
2309 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2310 for (i = 0; i < MAXQUOTAS; i++) {
2311 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2312 org_mount_opt.s_qf_names[i] =
2313 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2314 GFP_KERNEL);
2315 if (!org_mount_opt.s_qf_names[i]) {
2316 for (j = 0; j < i; j++)
2317 kfree(org_mount_opt.s_qf_names[j]);
2318 return -ENOMEM;
2319 }
2320 } else {
2321 org_mount_opt.s_qf_names[i] = NULL;
2322 }
2323 }
2324 #endif
2325
2326 /* recover superblocks we couldn't write due to previous RO mount */
2327 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2328 err = f2fs_commit_super(sbi, false);
2329 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2330 err);
2331 if (!err)
2332 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2333 }
2334
2335 default_options(sbi, true);
2336
2337 /* parse mount options */
2338 err = parse_options(sb, data, true);
2339 if (err)
2340 goto restore_opts;
2341
2342 /* flush outstanding errors before changing fs state */
2343 flush_work(&sbi->s_error_work);
2344
2345 /*
2346 * Previous and new state of filesystem is RO,
2347 * so skip checking GC and FLUSH_MERGE conditions.
2348 */
2349 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2350 goto skip;
2351
2352 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2353 err = -EROFS;
2354 goto restore_opts;
2355 }
2356
2357 #ifdef CONFIG_QUOTA
2358 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2359 err = dquot_suspend(sb, -1);
2360 if (err < 0)
2361 goto restore_opts;
2362 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2363 /* dquot_resume needs RW */
2364 sb->s_flags &= ~SB_RDONLY;
2365 if (sb_any_quota_suspended(sb)) {
2366 dquot_resume(sb, -1);
2367 } else if (f2fs_sb_has_quota_ino(sbi)) {
2368 err = f2fs_enable_quotas(sb);
2369 if (err)
2370 goto restore_opts;
2371 }
2372 }
2373 #endif
2374 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2375 err = -EINVAL;
2376 f2fs_warn(sbi, "LFS is not compatible with IPU");
2377 goto restore_opts;
2378 }
2379
2380 /* disallow enable atgc dynamically */
2381 if (no_atgc == !!test_opt(sbi, ATGC)) {
2382 err = -EINVAL;
2383 f2fs_warn(sbi, "switch atgc option is not allowed");
2384 goto restore_opts;
2385 }
2386
2387 /* disallow enable/disable extent_cache dynamically */
2388 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2389 err = -EINVAL;
2390 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2391 goto restore_opts;
2392 }
2393 /* disallow enable/disable age extent_cache dynamically */
2394 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2395 err = -EINVAL;
2396 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2397 goto restore_opts;
2398 }
2399
2400 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2401 err = -EINVAL;
2402 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2403 goto restore_opts;
2404 }
2405
2406 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2407 err = -EINVAL;
2408 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2409 goto restore_opts;
2410 }
2411
2412 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2413 err = -EINVAL;
2414 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2415 goto restore_opts;
2416 }
2417
2418 /*
2419 * We stop the GC thread if FS is mounted as RO
2420 * or if background_gc = off is passed in mount
2421 * option. Also sync the filesystem.
2422 */
2423 if ((*flags & SB_RDONLY) ||
2424 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2425 !test_opt(sbi, GC_MERGE))) {
2426 if (sbi->gc_thread) {
2427 f2fs_stop_gc_thread(sbi);
2428 need_restart_gc = true;
2429 }
2430 } else if (!sbi->gc_thread) {
2431 err = f2fs_start_gc_thread(sbi);
2432 if (err)
2433 goto restore_opts;
2434 need_stop_gc = true;
2435 }
2436
2437 if (*flags & SB_RDONLY) {
2438 sync_inodes_sb(sb);
2439
2440 set_sbi_flag(sbi, SBI_IS_DIRTY);
2441 set_sbi_flag(sbi, SBI_IS_CLOSE);
2442 f2fs_sync_fs(sb, 1);
2443 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2444 }
2445
2446 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2447 !test_opt(sbi, MERGE_CHECKPOINT)) {
2448 f2fs_stop_ckpt_thread(sbi);
2449 need_restart_ckpt = true;
2450 } else {
2451 /* Flush if the prevous checkpoint, if exists. */
2452 f2fs_flush_ckpt_thread(sbi);
2453
2454 err = f2fs_start_ckpt_thread(sbi);
2455 if (err) {
2456 f2fs_err(sbi,
2457 "Failed to start F2FS issue_checkpoint_thread (%d)",
2458 err);
2459 goto restore_gc;
2460 }
2461 need_stop_ckpt = true;
2462 }
2463
2464 /*
2465 * We stop issue flush thread if FS is mounted as RO
2466 * or if flush_merge is not passed in mount option.
2467 */
2468 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2469 clear_opt(sbi, FLUSH_MERGE);
2470 f2fs_destroy_flush_cmd_control(sbi, false);
2471 need_restart_flush = true;
2472 } else {
2473 err = f2fs_create_flush_cmd_control(sbi);
2474 if (err)
2475 goto restore_ckpt;
2476 need_stop_flush = true;
2477 }
2478
2479 if (no_discard == !!test_opt(sbi, DISCARD)) {
2480 if (test_opt(sbi, DISCARD)) {
2481 err = f2fs_start_discard_thread(sbi);
2482 if (err)
2483 goto restore_flush;
2484 need_stop_discard = true;
2485 } else {
2486 f2fs_stop_discard_thread(sbi);
2487 f2fs_issue_discard_timeout(sbi);
2488 need_restart_discard = true;
2489 }
2490 }
2491
2492 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2493 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2494 err = f2fs_disable_checkpoint(sbi);
2495 if (err)
2496 goto restore_discard;
2497 } else {
2498 f2fs_enable_checkpoint(sbi);
2499 }
2500 }
2501
2502 skip:
2503 #ifdef CONFIG_QUOTA
2504 /* Release old quota file names */
2505 for (i = 0; i < MAXQUOTAS; i++)
2506 kfree(org_mount_opt.s_qf_names[i]);
2507 #endif
2508 /* Update the POSIXACL Flag */
2509 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2510 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2511
2512 limit_reserve_root(sbi);
2513 adjust_unusable_cap_perc(sbi);
2514 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2515 return 0;
2516 restore_discard:
2517 if (need_restart_discard) {
2518 if (f2fs_start_discard_thread(sbi))
2519 f2fs_warn(sbi, "discard has been stopped");
2520 } else if (need_stop_discard) {
2521 f2fs_stop_discard_thread(sbi);
2522 }
2523 restore_flush:
2524 if (need_restart_flush) {
2525 if (f2fs_create_flush_cmd_control(sbi))
2526 f2fs_warn(sbi, "background flush thread has stopped");
2527 } else if (need_stop_flush) {
2528 clear_opt(sbi, FLUSH_MERGE);
2529 f2fs_destroy_flush_cmd_control(sbi, false);
2530 }
2531 restore_ckpt:
2532 if (need_restart_ckpt) {
2533 if (f2fs_start_ckpt_thread(sbi))
2534 f2fs_warn(sbi, "background ckpt thread has stopped");
2535 } else if (need_stop_ckpt) {
2536 f2fs_stop_ckpt_thread(sbi);
2537 }
2538 restore_gc:
2539 if (need_restart_gc) {
2540 if (f2fs_start_gc_thread(sbi))
2541 f2fs_warn(sbi, "background gc thread has stopped");
2542 } else if (need_stop_gc) {
2543 f2fs_stop_gc_thread(sbi);
2544 }
2545 restore_opts:
2546 #ifdef CONFIG_QUOTA
2547 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2548 for (i = 0; i < MAXQUOTAS; i++) {
2549 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2550 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2551 }
2552 #endif
2553 sbi->mount_opt = org_mount_opt;
2554 sb->s_flags = old_sb_flags;
2555 return err;
2556 }
2557
f2fs_shutdown(struct super_block * sb)2558 static void f2fs_shutdown(struct super_block *sb)
2559 {
2560 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false, false);
2561 }
2562
2563 #ifdef CONFIG_QUOTA
f2fs_need_recovery(struct f2fs_sb_info * sbi)2564 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2565 {
2566 /* need to recovery orphan */
2567 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2568 return true;
2569 /* need to recovery data */
2570 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2571 return false;
2572 if (test_opt(sbi, NORECOVERY))
2573 return false;
2574 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2575 }
2576
f2fs_recover_quota_begin(struct f2fs_sb_info * sbi)2577 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2578 {
2579 bool readonly = f2fs_readonly(sbi->sb);
2580
2581 if (!f2fs_need_recovery(sbi))
2582 return false;
2583
2584 /* it doesn't need to check f2fs_sb_has_readonly() */
2585 if (f2fs_hw_is_readonly(sbi))
2586 return false;
2587
2588 if (readonly) {
2589 sbi->sb->s_flags &= ~SB_RDONLY;
2590 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2591 }
2592
2593 /*
2594 * Turn on quotas which were not enabled for read-only mounts if
2595 * filesystem has quota feature, so that they are updated correctly.
2596 */
2597 return f2fs_enable_quota_files(sbi, readonly);
2598 }
2599
f2fs_recover_quota_end(struct f2fs_sb_info * sbi,bool quota_enabled)2600 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2601 bool quota_enabled)
2602 {
2603 if (quota_enabled)
2604 f2fs_quota_off_umount(sbi->sb);
2605
2606 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2607 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2608 sbi->sb->s_flags |= SB_RDONLY;
2609 }
2610 }
2611
2612 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2613 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2614 size_t len, loff_t off)
2615 {
2616 struct inode *inode = sb_dqopt(sb)->files[type];
2617 struct address_space *mapping = inode->i_mapping;
2618 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2619 int offset = off & (sb->s_blocksize - 1);
2620 int tocopy;
2621 size_t toread;
2622 loff_t i_size = i_size_read(inode);
2623 struct page *page;
2624
2625 if (off > i_size)
2626 return 0;
2627
2628 if (off + len > i_size)
2629 len = i_size - off;
2630 toread = len;
2631 while (toread > 0) {
2632 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2633 repeat:
2634 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2635 if (IS_ERR(page)) {
2636 if (PTR_ERR(page) == -ENOMEM) {
2637 memalloc_retry_wait(GFP_NOFS);
2638 goto repeat;
2639 }
2640 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2641 return PTR_ERR(page);
2642 }
2643
2644 lock_page(page);
2645
2646 if (unlikely(page->mapping != mapping)) {
2647 f2fs_put_page(page, 1);
2648 goto repeat;
2649 }
2650 if (unlikely(!PageUptodate(page))) {
2651 f2fs_put_page(page, 1);
2652 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2653 return -EIO;
2654 }
2655
2656 memcpy_from_page(data, page, offset, tocopy);
2657 f2fs_put_page(page, 1);
2658
2659 offset = 0;
2660 toread -= tocopy;
2661 data += tocopy;
2662 blkidx++;
2663 }
2664 return len;
2665 }
2666
2667 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2668 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2669 const char *data, size_t len, loff_t off)
2670 {
2671 struct inode *inode = sb_dqopt(sb)->files[type];
2672 struct address_space *mapping = inode->i_mapping;
2673 const struct address_space_operations *a_ops = mapping->a_ops;
2674 int offset = off & (sb->s_blocksize - 1);
2675 size_t towrite = len;
2676 struct page *page;
2677 void *fsdata = NULL;
2678 int err = 0;
2679 int tocopy;
2680
2681 while (towrite > 0) {
2682 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2683 towrite);
2684 retry:
2685 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2686 &page, &fsdata);
2687 if (unlikely(err)) {
2688 if (err == -ENOMEM) {
2689 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2690 goto retry;
2691 }
2692 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2693 break;
2694 }
2695
2696 memcpy_to_page(page, offset, data, tocopy);
2697
2698 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2699 page, fsdata);
2700 offset = 0;
2701 towrite -= tocopy;
2702 off += tocopy;
2703 data += tocopy;
2704 cond_resched();
2705 }
2706
2707 if (len == towrite)
2708 return err;
2709 inode->i_mtime = inode_set_ctime_current(inode);
2710 f2fs_mark_inode_dirty_sync(inode, false);
2711 return len - towrite;
2712 }
2713
f2fs_dquot_initialize(struct inode * inode)2714 int f2fs_dquot_initialize(struct inode *inode)
2715 {
2716 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2717 return -ESRCH;
2718
2719 return dquot_initialize(inode);
2720 }
2721
f2fs_get_dquots(struct inode * inode)2722 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2723 {
2724 return F2FS_I(inode)->i_dquot;
2725 }
2726
f2fs_get_reserved_space(struct inode * inode)2727 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2728 {
2729 return &F2FS_I(inode)->i_reserved_quota;
2730 }
2731
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2732 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2733 {
2734 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2735 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2736 return 0;
2737 }
2738
2739 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2740 F2FS_OPTION(sbi).s_jquota_fmt, type);
2741 }
2742
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2743 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2744 {
2745 int enabled = 0;
2746 int i, err;
2747
2748 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2749 err = f2fs_enable_quotas(sbi->sb);
2750 if (err) {
2751 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2752 return 0;
2753 }
2754 return 1;
2755 }
2756
2757 for (i = 0; i < MAXQUOTAS; i++) {
2758 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2759 err = f2fs_quota_on_mount(sbi, i);
2760 if (!err) {
2761 enabled = 1;
2762 continue;
2763 }
2764 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2765 err, i);
2766 }
2767 }
2768 return enabled;
2769 }
2770
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2771 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2772 unsigned int flags)
2773 {
2774 struct inode *qf_inode;
2775 unsigned long qf_inum;
2776 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2777 int err;
2778
2779 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2780
2781 qf_inum = f2fs_qf_ino(sb, type);
2782 if (!qf_inum)
2783 return -EPERM;
2784
2785 qf_inode = f2fs_iget(sb, qf_inum);
2786 if (IS_ERR(qf_inode)) {
2787 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2788 return PTR_ERR(qf_inode);
2789 }
2790
2791 /* Don't account quota for quota files to avoid recursion */
2792 inode_lock(qf_inode);
2793 qf_inode->i_flags |= S_NOQUOTA;
2794
2795 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2796 F2FS_I(qf_inode)->i_flags |= qf_flag;
2797 f2fs_set_inode_flags(qf_inode);
2798 }
2799 inode_unlock(qf_inode);
2800
2801 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2802 iput(qf_inode);
2803 return err;
2804 }
2805
f2fs_enable_quotas(struct super_block * sb)2806 static int f2fs_enable_quotas(struct super_block *sb)
2807 {
2808 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2809 int type, err = 0;
2810 unsigned long qf_inum;
2811 bool quota_mopt[MAXQUOTAS] = {
2812 test_opt(sbi, USRQUOTA),
2813 test_opt(sbi, GRPQUOTA),
2814 test_opt(sbi, PRJQUOTA),
2815 };
2816
2817 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2818 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2819 return 0;
2820 }
2821
2822 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2823
2824 for (type = 0; type < MAXQUOTAS; type++) {
2825 qf_inum = f2fs_qf_ino(sb, type);
2826 if (qf_inum) {
2827 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2828 DQUOT_USAGE_ENABLED |
2829 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2830 if (err) {
2831 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2832 type, err);
2833 for (type--; type >= 0; type--)
2834 dquot_quota_off(sb, type);
2835 set_sbi_flag(F2FS_SB(sb),
2836 SBI_QUOTA_NEED_REPAIR);
2837 return err;
2838 }
2839 }
2840 }
2841 return 0;
2842 }
2843
f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2844 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2845 {
2846 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2847 struct address_space *mapping = dqopt->files[type]->i_mapping;
2848 int ret = 0;
2849
2850 ret = dquot_writeback_dquots(sbi->sb, type);
2851 if (ret)
2852 goto out;
2853
2854 ret = filemap_fdatawrite(mapping);
2855 if (ret)
2856 goto out;
2857
2858 /* if we are using journalled quota */
2859 if (is_journalled_quota(sbi))
2860 goto out;
2861
2862 ret = filemap_fdatawait(mapping);
2863
2864 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2865 out:
2866 if (ret)
2867 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2868 return ret;
2869 }
2870
f2fs_quota_sync(struct super_block * sb,int type)2871 int f2fs_quota_sync(struct super_block *sb, int type)
2872 {
2873 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2874 struct quota_info *dqopt = sb_dqopt(sb);
2875 int cnt;
2876 int ret = 0;
2877
2878 /*
2879 * Now when everything is written we can discard the pagecache so
2880 * that userspace sees the changes.
2881 */
2882 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2883
2884 if (type != -1 && cnt != type)
2885 continue;
2886
2887 if (!sb_has_quota_active(sb, cnt))
2888 continue;
2889
2890 if (!f2fs_sb_has_quota_ino(sbi))
2891 inode_lock(dqopt->files[cnt]);
2892
2893 /*
2894 * do_quotactl
2895 * f2fs_quota_sync
2896 * f2fs_down_read(quota_sem)
2897 * dquot_writeback_dquots()
2898 * f2fs_dquot_commit
2899 * block_operation
2900 * f2fs_down_read(quota_sem)
2901 */
2902 f2fs_lock_op(sbi);
2903 f2fs_down_read(&sbi->quota_sem);
2904
2905 ret = f2fs_quota_sync_file(sbi, cnt);
2906
2907 f2fs_up_read(&sbi->quota_sem);
2908 f2fs_unlock_op(sbi);
2909
2910 if (!f2fs_sb_has_quota_ino(sbi))
2911 inode_unlock(dqopt->files[cnt]);
2912
2913 if (ret)
2914 break;
2915 }
2916 return ret;
2917 }
2918
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2919 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2920 const struct path *path)
2921 {
2922 struct inode *inode;
2923 int err;
2924
2925 /* if quota sysfile exists, deny enabling quota with specific file */
2926 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2927 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2928 return -EBUSY;
2929 }
2930
2931 if (path->dentry->d_sb != sb)
2932 return -EXDEV;
2933
2934 err = f2fs_quota_sync(sb, type);
2935 if (err)
2936 return err;
2937
2938 inode = d_inode(path->dentry);
2939
2940 err = filemap_fdatawrite(inode->i_mapping);
2941 if (err)
2942 return err;
2943
2944 err = filemap_fdatawait(inode->i_mapping);
2945 if (err)
2946 return err;
2947
2948 err = dquot_quota_on(sb, type, format_id, path);
2949 if (err)
2950 return err;
2951
2952 inode_lock(inode);
2953 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2954 f2fs_set_inode_flags(inode);
2955 inode_unlock(inode);
2956 f2fs_mark_inode_dirty_sync(inode, false);
2957
2958 return 0;
2959 }
2960
__f2fs_quota_off(struct super_block * sb,int type)2961 static int __f2fs_quota_off(struct super_block *sb, int type)
2962 {
2963 struct inode *inode = sb_dqopt(sb)->files[type];
2964 int err;
2965
2966 if (!inode || !igrab(inode))
2967 return dquot_quota_off(sb, type);
2968
2969 err = f2fs_quota_sync(sb, type);
2970 if (err)
2971 goto out_put;
2972
2973 err = dquot_quota_off(sb, type);
2974 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2975 goto out_put;
2976
2977 inode_lock(inode);
2978 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
2979 f2fs_set_inode_flags(inode);
2980 inode_unlock(inode);
2981 f2fs_mark_inode_dirty_sync(inode, false);
2982 out_put:
2983 iput(inode);
2984 return err;
2985 }
2986
f2fs_quota_off(struct super_block * sb,int type)2987 static int f2fs_quota_off(struct super_block *sb, int type)
2988 {
2989 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2990 int err;
2991
2992 err = __f2fs_quota_off(sb, type);
2993
2994 /*
2995 * quotactl can shutdown journalled quota, result in inconsistence
2996 * between quota record and fs data by following updates, tag the
2997 * flag to let fsck be aware of it.
2998 */
2999 if (is_journalled_quota(sbi))
3000 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3001 return err;
3002 }
3003
f2fs_quota_off_umount(struct super_block * sb)3004 void f2fs_quota_off_umount(struct super_block *sb)
3005 {
3006 int type;
3007 int err;
3008
3009 for (type = 0; type < MAXQUOTAS; type++) {
3010 err = __f2fs_quota_off(sb, type);
3011 if (err) {
3012 int ret = dquot_quota_off(sb, type);
3013
3014 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3015 type, err, ret);
3016 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3017 }
3018 }
3019 /*
3020 * In case of checkpoint=disable, we must flush quota blocks.
3021 * This can cause NULL exception for node_inode in end_io, since
3022 * put_super already dropped it.
3023 */
3024 sync_filesystem(sb);
3025 }
3026
f2fs_truncate_quota_inode_pages(struct super_block * sb)3027 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3028 {
3029 struct quota_info *dqopt = sb_dqopt(sb);
3030 int type;
3031
3032 for (type = 0; type < MAXQUOTAS; type++) {
3033 if (!dqopt->files[type])
3034 continue;
3035 f2fs_inode_synced(dqopt->files[type]);
3036 }
3037 }
3038
f2fs_dquot_commit(struct dquot * dquot)3039 static int f2fs_dquot_commit(struct dquot *dquot)
3040 {
3041 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3042 int ret;
3043
3044 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3045 ret = dquot_commit(dquot);
3046 if (ret < 0)
3047 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3048 f2fs_up_read(&sbi->quota_sem);
3049 return ret;
3050 }
3051
f2fs_dquot_acquire(struct dquot * dquot)3052 static int f2fs_dquot_acquire(struct dquot *dquot)
3053 {
3054 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3055 int ret;
3056
3057 f2fs_down_read(&sbi->quota_sem);
3058 ret = dquot_acquire(dquot);
3059 if (ret < 0)
3060 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3061 f2fs_up_read(&sbi->quota_sem);
3062 return ret;
3063 }
3064
f2fs_dquot_release(struct dquot * dquot)3065 static int f2fs_dquot_release(struct dquot *dquot)
3066 {
3067 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3068 int ret = dquot_release(dquot);
3069
3070 if (ret < 0)
3071 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3072 return ret;
3073 }
3074
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)3075 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3076 {
3077 struct super_block *sb = dquot->dq_sb;
3078 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3079 int ret = dquot_mark_dquot_dirty(dquot);
3080
3081 /* if we are using journalled quota */
3082 if (is_journalled_quota(sbi))
3083 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3084
3085 return ret;
3086 }
3087
f2fs_dquot_commit_info(struct super_block * sb,int type)3088 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3089 {
3090 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3091 int ret = dquot_commit_info(sb, type);
3092
3093 if (ret < 0)
3094 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3095 return ret;
3096 }
3097
f2fs_get_projid(struct inode * inode,kprojid_t * projid)3098 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3099 {
3100 *projid = F2FS_I(inode)->i_projid;
3101 return 0;
3102 }
3103
3104 static const struct dquot_operations f2fs_quota_operations = {
3105 .get_reserved_space = f2fs_get_reserved_space,
3106 .write_dquot = f2fs_dquot_commit,
3107 .acquire_dquot = f2fs_dquot_acquire,
3108 .release_dquot = f2fs_dquot_release,
3109 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3110 .write_info = f2fs_dquot_commit_info,
3111 .alloc_dquot = dquot_alloc,
3112 .destroy_dquot = dquot_destroy,
3113 .get_projid = f2fs_get_projid,
3114 .get_next_id = dquot_get_next_id,
3115 };
3116
3117 static const struct quotactl_ops f2fs_quotactl_ops = {
3118 .quota_on = f2fs_quota_on,
3119 .quota_off = f2fs_quota_off,
3120 .quota_sync = f2fs_quota_sync,
3121 .get_state = dquot_get_state,
3122 .set_info = dquot_set_dqinfo,
3123 .get_dqblk = dquot_get_dqblk,
3124 .set_dqblk = dquot_set_dqblk,
3125 .get_nextdqblk = dquot_get_next_dqblk,
3126 };
3127 #else
f2fs_dquot_initialize(struct inode * inode)3128 int f2fs_dquot_initialize(struct inode *inode)
3129 {
3130 return 0;
3131 }
3132
f2fs_quota_sync(struct super_block * sb,int type)3133 int f2fs_quota_sync(struct super_block *sb, int type)
3134 {
3135 return 0;
3136 }
3137
f2fs_quota_off_umount(struct super_block * sb)3138 void f2fs_quota_off_umount(struct super_block *sb)
3139 {
3140 }
3141 #endif
3142
3143 static const struct super_operations f2fs_sops = {
3144 .alloc_inode = f2fs_alloc_inode,
3145 .free_inode = f2fs_free_inode,
3146 .drop_inode = f2fs_drop_inode,
3147 .write_inode = f2fs_write_inode,
3148 .dirty_inode = f2fs_dirty_inode,
3149 .show_options = f2fs_show_options,
3150 #ifdef CONFIG_QUOTA
3151 .quota_read = f2fs_quota_read,
3152 .quota_write = f2fs_quota_write,
3153 .get_dquots = f2fs_get_dquots,
3154 #endif
3155 .evict_inode = f2fs_evict_inode,
3156 .put_super = f2fs_put_super,
3157 .sync_fs = f2fs_sync_fs,
3158 .freeze_fs = f2fs_freeze,
3159 .unfreeze_fs = f2fs_unfreeze,
3160 .statfs = f2fs_statfs,
3161 .remount_fs = f2fs_remount,
3162 .shutdown = f2fs_shutdown,
3163 };
3164
3165 #ifdef CONFIG_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)3166 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3167 {
3168 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3169 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3170 ctx, len, NULL);
3171 }
3172
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)3173 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3174 void *fs_data)
3175 {
3176 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3177
3178 /*
3179 * Encrypting the root directory is not allowed because fsck
3180 * expects lost+found directory to exist and remain unencrypted
3181 * if LOST_FOUND feature is enabled.
3182 *
3183 */
3184 if (f2fs_sb_has_lost_found(sbi) &&
3185 inode->i_ino == F2FS_ROOT_INO(sbi))
3186 return -EPERM;
3187
3188 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3189 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3190 ctx, len, fs_data, XATTR_CREATE);
3191 }
3192
f2fs_get_dummy_policy(struct super_block * sb)3193 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3194 {
3195 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3196 }
3197
f2fs_has_stable_inodes(struct super_block * sb)3198 static bool f2fs_has_stable_inodes(struct super_block *sb)
3199 {
3200 return true;
3201 }
3202
f2fs_get_ino_and_lblk_bits(struct super_block * sb,int * ino_bits_ret,int * lblk_bits_ret)3203 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3204 int *ino_bits_ret, int *lblk_bits_ret)
3205 {
3206 *ino_bits_ret = 8 * sizeof(nid_t);
3207 *lblk_bits_ret = 8 * sizeof(block_t);
3208 }
3209
f2fs_get_devices(struct super_block * sb,unsigned int * num_devs)3210 static struct block_device **f2fs_get_devices(struct super_block *sb,
3211 unsigned int *num_devs)
3212 {
3213 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3214 struct block_device **devs;
3215 int i;
3216
3217 if (!f2fs_is_multi_device(sbi))
3218 return NULL;
3219
3220 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3221 if (!devs)
3222 return ERR_PTR(-ENOMEM);
3223
3224 for (i = 0; i < sbi->s_ndevs; i++)
3225 devs[i] = FDEV(i).bdev;
3226 *num_devs = sbi->s_ndevs;
3227 return devs;
3228 }
3229
3230 static const struct fscrypt_operations f2fs_cryptops = {
3231 .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_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
3238 .get_devices = f2fs_get_devices,
3239 };
3240 #endif
3241
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)3242 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3243 u64 ino, u32 generation)
3244 {
3245 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3246 struct inode *inode;
3247
3248 if (f2fs_check_nid_range(sbi, ino))
3249 return ERR_PTR(-ESTALE);
3250
3251 /*
3252 * f2fs_iget isn't quite right if the inode is currently unallocated!
3253 * However f2fs_iget currently does appropriate checks to handle stale
3254 * inodes so everything is OK.
3255 */
3256 inode = f2fs_iget(sb, ino);
3257 if (IS_ERR(inode))
3258 return ERR_CAST(inode);
3259 if (unlikely(generation && inode->i_generation != generation)) {
3260 /* we didn't find the right inode.. */
3261 iput(inode);
3262 return ERR_PTR(-ESTALE);
3263 }
3264 return inode;
3265 }
3266
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3267 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3268 int fh_len, int fh_type)
3269 {
3270 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3271 f2fs_nfs_get_inode);
3272 }
3273
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3274 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3275 int fh_len, int fh_type)
3276 {
3277 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3278 f2fs_nfs_get_inode);
3279 }
3280
3281 static const struct export_operations f2fs_export_ops = {
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 return result;
3316 }
3317
__f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)3318 static int __f2fs_commit_super(struct buffer_head *bh,
3319 struct f2fs_super_block *super)
3320 {
3321 lock_buffer(bh);
3322 if (super)
3323 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3324 set_buffer_dirty(bh);
3325 unlock_buffer(bh);
3326
3327 /* it's rare case, we can do fua all the time */
3328 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3329 }
3330
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)3331 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3332 struct buffer_head *bh)
3333 {
3334 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3335 (bh->b_data + F2FS_SUPER_OFFSET);
3336 struct super_block *sb = sbi->sb;
3337 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3338 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3339 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3340 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3341 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3342 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3343 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3344 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3345 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3346 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3347 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3348 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3349 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3350 u64 main_end_blkaddr = main_blkaddr +
3351 ((u64)segment_count_main << log_blocks_per_seg);
3352 u64 seg_end_blkaddr = segment0_blkaddr +
3353 ((u64)segment_count << log_blocks_per_seg);
3354
3355 if (segment0_blkaddr != cp_blkaddr) {
3356 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3357 segment0_blkaddr, cp_blkaddr);
3358 return true;
3359 }
3360
3361 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3362 sit_blkaddr) {
3363 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3364 cp_blkaddr, sit_blkaddr,
3365 segment_count_ckpt << log_blocks_per_seg);
3366 return true;
3367 }
3368
3369 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3370 nat_blkaddr) {
3371 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3372 sit_blkaddr, nat_blkaddr,
3373 segment_count_sit << log_blocks_per_seg);
3374 return true;
3375 }
3376
3377 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3378 ssa_blkaddr) {
3379 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3380 nat_blkaddr, ssa_blkaddr,
3381 segment_count_nat << log_blocks_per_seg);
3382 return true;
3383 }
3384
3385 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3386 main_blkaddr) {
3387 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3388 ssa_blkaddr, main_blkaddr,
3389 segment_count_ssa << log_blocks_per_seg);
3390 return true;
3391 }
3392
3393 if (main_end_blkaddr > seg_end_blkaddr) {
3394 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3395 main_blkaddr, seg_end_blkaddr,
3396 segment_count_main << log_blocks_per_seg);
3397 return true;
3398 } else if (main_end_blkaddr < seg_end_blkaddr) {
3399 int err = 0;
3400 char *res;
3401
3402 /* fix in-memory information all the time */
3403 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3404 segment0_blkaddr) >> log_blocks_per_seg);
3405
3406 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3407 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3408 res = "internally";
3409 } else {
3410 err = __f2fs_commit_super(bh, NULL);
3411 res = err ? "failed" : "done";
3412 }
3413 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3414 res, main_blkaddr, seg_end_blkaddr,
3415 segment_count_main << log_blocks_per_seg);
3416 if (err)
3417 return true;
3418 }
3419 return false;
3420 }
3421
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)3422 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3423 struct buffer_head *bh)
3424 {
3425 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3426 block_t total_sections, blocks_per_seg;
3427 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3428 (bh->b_data + F2FS_SUPER_OFFSET);
3429 size_t crc_offset = 0;
3430 __u32 crc = 0;
3431
3432 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3433 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3434 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3435 return -EINVAL;
3436 }
3437
3438 /* Check checksum_offset and crc in superblock */
3439 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3440 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3441 if (crc_offset !=
3442 offsetof(struct f2fs_super_block, crc)) {
3443 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3444 crc_offset);
3445 return -EFSCORRUPTED;
3446 }
3447 crc = le32_to_cpu(raw_super->crc);
3448 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3449 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3450 return -EFSCORRUPTED;
3451 }
3452 }
3453
3454 /* Currently, support only 4KB block size */
3455 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3456 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3457 le32_to_cpu(raw_super->log_blocksize),
3458 F2FS_BLKSIZE_BITS);
3459 return -EFSCORRUPTED;
3460 }
3461
3462 /* check log blocks per segment */
3463 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3464 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3465 le32_to_cpu(raw_super->log_blocks_per_seg));
3466 return -EFSCORRUPTED;
3467 }
3468
3469 /* Currently, support 512/1024/2048/4096 bytes sector size */
3470 if (le32_to_cpu(raw_super->log_sectorsize) >
3471 F2FS_MAX_LOG_SECTOR_SIZE ||
3472 le32_to_cpu(raw_super->log_sectorsize) <
3473 F2FS_MIN_LOG_SECTOR_SIZE) {
3474 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3475 le32_to_cpu(raw_super->log_sectorsize));
3476 return -EFSCORRUPTED;
3477 }
3478 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3479 le32_to_cpu(raw_super->log_sectorsize) !=
3480 F2FS_MAX_LOG_SECTOR_SIZE) {
3481 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3482 le32_to_cpu(raw_super->log_sectors_per_block),
3483 le32_to_cpu(raw_super->log_sectorsize));
3484 return -EFSCORRUPTED;
3485 }
3486
3487 segment_count = le32_to_cpu(raw_super->segment_count);
3488 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3489 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3490 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3491 total_sections = le32_to_cpu(raw_super->section_count);
3492
3493 /* blocks_per_seg should be 512, given the above check */
3494 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3495
3496 if (segment_count > F2FS_MAX_SEGMENT ||
3497 segment_count < F2FS_MIN_SEGMENTS) {
3498 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3499 return -EFSCORRUPTED;
3500 }
3501
3502 if (total_sections > segment_count_main || total_sections < 1 ||
3503 segs_per_sec > segment_count || !segs_per_sec) {
3504 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3505 segment_count, total_sections, segs_per_sec);
3506 return -EFSCORRUPTED;
3507 }
3508
3509 if (segment_count_main != total_sections * segs_per_sec) {
3510 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3511 segment_count_main, total_sections, segs_per_sec);
3512 return -EFSCORRUPTED;
3513 }
3514
3515 if ((segment_count / segs_per_sec) < total_sections) {
3516 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3517 segment_count, segs_per_sec, total_sections);
3518 return -EFSCORRUPTED;
3519 }
3520
3521 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3522 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3523 segment_count, le64_to_cpu(raw_super->block_count));
3524 return -EFSCORRUPTED;
3525 }
3526
3527 if (RDEV(0).path[0]) {
3528 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3529 int i = 1;
3530
3531 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3532 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3533 i++;
3534 }
3535 if (segment_count != dev_seg_count) {
3536 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3537 segment_count, dev_seg_count);
3538 return -EFSCORRUPTED;
3539 }
3540 } else {
3541 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3542 !bdev_is_zoned(sbi->sb->s_bdev)) {
3543 f2fs_info(sbi, "Zoned block device path is missing");
3544 return -EFSCORRUPTED;
3545 }
3546 }
3547
3548 if (secs_per_zone > total_sections || !secs_per_zone) {
3549 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3550 secs_per_zone, total_sections);
3551 return -EFSCORRUPTED;
3552 }
3553 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3554 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3555 (le32_to_cpu(raw_super->extension_count) +
3556 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3557 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3558 le32_to_cpu(raw_super->extension_count),
3559 raw_super->hot_ext_count,
3560 F2FS_MAX_EXTENSION);
3561 return -EFSCORRUPTED;
3562 }
3563
3564 if (le32_to_cpu(raw_super->cp_payload) >=
3565 (blocks_per_seg - F2FS_CP_PACKS -
3566 NR_CURSEG_PERSIST_TYPE)) {
3567 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3568 le32_to_cpu(raw_super->cp_payload),
3569 blocks_per_seg - F2FS_CP_PACKS -
3570 NR_CURSEG_PERSIST_TYPE);
3571 return -EFSCORRUPTED;
3572 }
3573
3574 /* check reserved ino info */
3575 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3576 le32_to_cpu(raw_super->meta_ino) != 2 ||
3577 le32_to_cpu(raw_super->root_ino) != 3) {
3578 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3579 le32_to_cpu(raw_super->node_ino),
3580 le32_to_cpu(raw_super->meta_ino),
3581 le32_to_cpu(raw_super->root_ino));
3582 return -EFSCORRUPTED;
3583 }
3584
3585 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3586 if (sanity_check_area_boundary(sbi, bh))
3587 return -EFSCORRUPTED;
3588
3589 return 0;
3590 }
3591
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3592 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3593 {
3594 unsigned int total, fsmeta;
3595 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3596 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3597 unsigned int ovp_segments, reserved_segments;
3598 unsigned int main_segs, blocks_per_seg;
3599 unsigned int sit_segs, nat_segs;
3600 unsigned int sit_bitmap_size, nat_bitmap_size;
3601 unsigned int log_blocks_per_seg;
3602 unsigned int segment_count_main;
3603 unsigned int cp_pack_start_sum, cp_payload;
3604 block_t user_block_count, valid_user_blocks;
3605 block_t avail_node_count, valid_node_count;
3606 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3607 int i, j;
3608
3609 total = le32_to_cpu(raw_super->segment_count);
3610 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3611 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3612 fsmeta += sit_segs;
3613 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3614 fsmeta += nat_segs;
3615 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3616 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3617
3618 if (unlikely(fsmeta >= total))
3619 return 1;
3620
3621 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3622 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3623
3624 if (!f2fs_sb_has_readonly(sbi) &&
3625 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3626 ovp_segments == 0 || reserved_segments == 0)) {
3627 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3628 return 1;
3629 }
3630 user_block_count = le64_to_cpu(ckpt->user_block_count);
3631 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3632 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3633 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3634 if (!user_block_count || user_block_count >=
3635 segment_count_main << log_blocks_per_seg) {
3636 f2fs_err(sbi, "Wrong user_block_count: %u",
3637 user_block_count);
3638 return 1;
3639 }
3640
3641 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3642 if (valid_user_blocks > user_block_count) {
3643 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3644 valid_user_blocks, user_block_count);
3645 return 1;
3646 }
3647
3648 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3649 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3650 if (valid_node_count > avail_node_count) {
3651 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3652 valid_node_count, avail_node_count);
3653 return 1;
3654 }
3655
3656 main_segs = le32_to_cpu(raw_super->segment_count_main);
3657 blocks_per_seg = BLKS_PER_SEG(sbi);
3658
3659 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3660 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3661 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3662 return 1;
3663
3664 if (f2fs_sb_has_readonly(sbi))
3665 goto check_data;
3666
3667 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3668 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3669 le32_to_cpu(ckpt->cur_node_segno[j])) {
3670 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3671 i, j,
3672 le32_to_cpu(ckpt->cur_node_segno[i]));
3673 return 1;
3674 }
3675 }
3676 }
3677 check_data:
3678 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3679 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3680 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3681 return 1;
3682
3683 if (f2fs_sb_has_readonly(sbi))
3684 goto skip_cross;
3685
3686 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3687 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3688 le32_to_cpu(ckpt->cur_data_segno[j])) {
3689 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3690 i, j,
3691 le32_to_cpu(ckpt->cur_data_segno[i]));
3692 return 1;
3693 }
3694 }
3695 }
3696 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3697 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3698 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3699 le32_to_cpu(ckpt->cur_data_segno[j])) {
3700 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3701 i, j,
3702 le32_to_cpu(ckpt->cur_node_segno[i]));
3703 return 1;
3704 }
3705 }
3706 }
3707 skip_cross:
3708 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3709 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3710
3711 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3712 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3713 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3714 sit_bitmap_size, nat_bitmap_size);
3715 return 1;
3716 }
3717
3718 cp_pack_start_sum = __start_sum_addr(sbi);
3719 cp_payload = __cp_payload(sbi);
3720 if (cp_pack_start_sum < cp_payload + 1 ||
3721 cp_pack_start_sum > blocks_per_seg - 1 -
3722 NR_CURSEG_PERSIST_TYPE) {
3723 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3724 cp_pack_start_sum);
3725 return 1;
3726 }
3727
3728 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3729 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3730 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3731 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3732 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3733 le32_to_cpu(ckpt->checksum_offset));
3734 return 1;
3735 }
3736
3737 nat_blocks = nat_segs << log_blocks_per_seg;
3738 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3739 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3740 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3741 (cp_payload + F2FS_CP_PACKS +
3742 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3743 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3744 cp_payload, nat_bits_blocks);
3745 return 1;
3746 }
3747
3748 if (unlikely(f2fs_cp_error(sbi))) {
3749 f2fs_err(sbi, "A bug case: need to run fsck");
3750 return 1;
3751 }
3752 return 0;
3753 }
3754
init_sb_info(struct f2fs_sb_info * sbi)3755 static void init_sb_info(struct f2fs_sb_info *sbi)
3756 {
3757 struct f2fs_super_block *raw_super = sbi->raw_super;
3758 int i;
3759
3760 sbi->log_sectors_per_block =
3761 le32_to_cpu(raw_super->log_sectors_per_block);
3762 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3763 sbi->blocksize = BIT(sbi->log_blocksize);
3764 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3765 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3766 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3767 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3768 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3769 sbi->total_node_count =
3770 ((le32_to_cpu(raw_super->segment_count_nat) / 2) *
3771 NAT_ENTRY_PER_BLOCK) << sbi->log_blocks_per_seg;
3772 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3773 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3774 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3775 sbi->cur_victim_sec = NULL_SECNO;
3776 sbi->gc_mode = GC_NORMAL;
3777 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3778 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3779 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3780 sbi->migration_granularity = SEGS_PER_SEC(sbi);
3781 sbi->seq_file_ra_mul = MIN_RA_MUL;
3782 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3783 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3784 spin_lock_init(&sbi->gc_remaining_trials_lock);
3785 atomic64_set(&sbi->current_atomic_write, 0);
3786
3787 sbi->dir_level = DEF_DIR_LEVEL;
3788 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3789 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3790 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3791 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3792 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3793 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3794 DEF_UMOUNT_DISCARD_TIMEOUT;
3795 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3796
3797 for (i = 0; i < NR_COUNT_TYPE; i++)
3798 atomic_set(&sbi->nr_pages[i], 0);
3799
3800 for (i = 0; i < META; i++)
3801 atomic_set(&sbi->wb_sync_req[i], 0);
3802
3803 INIT_LIST_HEAD(&sbi->s_list);
3804 mutex_init(&sbi->umount_mutex);
3805 init_f2fs_rwsem(&sbi->io_order_lock);
3806 spin_lock_init(&sbi->cp_lock);
3807
3808 sbi->dirty_device = 0;
3809 spin_lock_init(&sbi->dev_lock);
3810
3811 init_f2fs_rwsem(&sbi->sb_lock);
3812 init_f2fs_rwsem(&sbi->pin_sem);
3813 }
3814
init_percpu_info(struct f2fs_sb_info * sbi)3815 static int init_percpu_info(struct f2fs_sb_info *sbi)
3816 {
3817 int err;
3818
3819 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3820 if (err)
3821 return err;
3822
3823 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3824 if (err)
3825 goto err_valid_block;
3826
3827 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3828 GFP_KERNEL);
3829 if (err)
3830 goto err_node_block;
3831 return 0;
3832
3833 err_node_block:
3834 percpu_counter_destroy(&sbi->rf_node_block_count);
3835 err_valid_block:
3836 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3837 return err;
3838 }
3839
3840 #ifdef CONFIG_BLK_DEV_ZONED
3841
3842 struct f2fs_report_zones_args {
3843 struct f2fs_sb_info *sbi;
3844 struct f2fs_dev_info *dev;
3845 };
3846
f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3847 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3848 void *data)
3849 {
3850 struct f2fs_report_zones_args *rz_args = data;
3851 block_t unusable_blocks = (zone->len - zone->capacity) >>
3852 F2FS_LOG_SECTORS_PER_BLOCK;
3853
3854 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3855 return 0;
3856
3857 set_bit(idx, rz_args->dev->blkz_seq);
3858 if (!rz_args->sbi->unusable_blocks_per_sec) {
3859 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3860 return 0;
3861 }
3862 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3863 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3864 return -EINVAL;
3865 }
3866 return 0;
3867 }
3868
init_blkz_info(struct f2fs_sb_info * sbi,int devi)3869 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3870 {
3871 struct block_device *bdev = FDEV(devi).bdev;
3872 sector_t nr_sectors = bdev_nr_sectors(bdev);
3873 struct f2fs_report_zones_args rep_zone_arg;
3874 u64 zone_sectors;
3875 int ret;
3876
3877 if (!f2fs_sb_has_blkzoned(sbi))
3878 return 0;
3879
3880 zone_sectors = bdev_zone_sectors(bdev);
3881 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3882 SECTOR_TO_BLOCK(zone_sectors))
3883 return -EINVAL;
3884 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3885 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3886 sbi->blocks_per_blkz);
3887 if (nr_sectors & (zone_sectors - 1))
3888 FDEV(devi).nr_blkz++;
3889
3890 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3891 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3892 * sizeof(unsigned long),
3893 GFP_KERNEL);
3894 if (!FDEV(devi).blkz_seq)
3895 return -ENOMEM;
3896
3897 rep_zone_arg.sbi = sbi;
3898 rep_zone_arg.dev = &FDEV(devi);
3899
3900 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3901 &rep_zone_arg);
3902 if (ret < 0)
3903 return ret;
3904 return 0;
3905 }
3906 #endif
3907
3908 /*
3909 * Read f2fs raw super block.
3910 * Because we have two copies of super block, so read both of them
3911 * to get the first valid one. If any one of them is broken, we pass
3912 * them recovery flag back to the caller.
3913 */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3914 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3915 struct f2fs_super_block **raw_super,
3916 int *valid_super_block, int *recovery)
3917 {
3918 struct super_block *sb = sbi->sb;
3919 int block;
3920 struct buffer_head *bh;
3921 struct f2fs_super_block *super;
3922 int err = 0;
3923
3924 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3925 if (!super)
3926 return -ENOMEM;
3927
3928 for (block = 0; block < 2; block++) {
3929 bh = sb_bread(sb, block);
3930 if (!bh) {
3931 f2fs_err(sbi, "Unable to read %dth superblock",
3932 block + 1);
3933 err = -EIO;
3934 *recovery = 1;
3935 continue;
3936 }
3937
3938 /* sanity checking of raw super */
3939 err = sanity_check_raw_super(sbi, bh);
3940 if (err) {
3941 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3942 block + 1);
3943 brelse(bh);
3944 *recovery = 1;
3945 continue;
3946 }
3947
3948 if (!*raw_super) {
3949 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3950 sizeof(*super));
3951 *valid_super_block = block;
3952 *raw_super = super;
3953 }
3954 brelse(bh);
3955 }
3956
3957 /* No valid superblock */
3958 if (!*raw_super)
3959 kfree(super);
3960 else
3961 err = 0;
3962
3963 return err;
3964 }
3965
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)3966 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3967 {
3968 struct buffer_head *bh;
3969 __u32 crc = 0;
3970 int err;
3971
3972 if ((recover && f2fs_readonly(sbi->sb)) ||
3973 f2fs_hw_is_readonly(sbi)) {
3974 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3975 return -EROFS;
3976 }
3977
3978 /* we should update superblock crc here */
3979 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3980 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3981 offsetof(struct f2fs_super_block, crc));
3982 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3983 }
3984
3985 /* write back-up superblock first */
3986 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3987 if (!bh)
3988 return -EIO;
3989 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3990 brelse(bh);
3991
3992 /* if we are in recovery path, skip writing valid superblock */
3993 if (recover || err)
3994 return err;
3995
3996 /* write current valid superblock */
3997 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3998 if (!bh)
3999 return -EIO;
4000 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4001 brelse(bh);
4002 return err;
4003 }
4004
save_stop_reason(struct f2fs_sb_info * sbi,unsigned char reason)4005 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4006 {
4007 unsigned long flags;
4008
4009 spin_lock_irqsave(&sbi->error_lock, flags);
4010 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4011 sbi->stop_reason[reason]++;
4012 spin_unlock_irqrestore(&sbi->error_lock, flags);
4013 }
4014
f2fs_record_stop_reason(struct f2fs_sb_info * sbi)4015 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4016 {
4017 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4018 unsigned long flags;
4019 int err;
4020
4021 f2fs_down_write(&sbi->sb_lock);
4022
4023 spin_lock_irqsave(&sbi->error_lock, flags);
4024 if (sbi->error_dirty) {
4025 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4026 MAX_F2FS_ERRORS);
4027 sbi->error_dirty = false;
4028 }
4029 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4030 spin_unlock_irqrestore(&sbi->error_lock, flags);
4031
4032 err = f2fs_commit_super(sbi, false);
4033
4034 f2fs_up_write(&sbi->sb_lock);
4035 if (err)
4036 f2fs_err(sbi, "f2fs_commit_super fails to record err:%d", err);
4037 }
4038
f2fs_save_errors(struct f2fs_sb_info * sbi,unsigned char flag)4039 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4040 {
4041 unsigned long flags;
4042
4043 spin_lock_irqsave(&sbi->error_lock, flags);
4044 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4045 set_bit(flag, (unsigned long *)sbi->errors);
4046 sbi->error_dirty = true;
4047 }
4048 spin_unlock_irqrestore(&sbi->error_lock, flags);
4049 }
4050
f2fs_update_errors(struct f2fs_sb_info * sbi)4051 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4052 {
4053 unsigned long flags;
4054 bool need_update = false;
4055
4056 spin_lock_irqsave(&sbi->error_lock, flags);
4057 if (sbi->error_dirty) {
4058 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4059 MAX_F2FS_ERRORS);
4060 sbi->error_dirty = false;
4061 need_update = true;
4062 }
4063 spin_unlock_irqrestore(&sbi->error_lock, flags);
4064
4065 return need_update;
4066 }
4067
f2fs_record_errors(struct f2fs_sb_info * sbi,unsigned char error)4068 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4069 {
4070 int err;
4071
4072 f2fs_down_write(&sbi->sb_lock);
4073
4074 if (!f2fs_update_errors(sbi))
4075 goto out_unlock;
4076
4077 err = f2fs_commit_super(sbi, false);
4078 if (err)
4079 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
4080 error, err);
4081 out_unlock:
4082 f2fs_up_write(&sbi->sb_lock);
4083 }
4084
f2fs_handle_error(struct f2fs_sb_info * sbi,unsigned char error)4085 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4086 {
4087 f2fs_save_errors(sbi, error);
4088 f2fs_record_errors(sbi, error);
4089 }
4090
f2fs_handle_error_async(struct f2fs_sb_info * sbi,unsigned char error)4091 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4092 {
4093 f2fs_save_errors(sbi, error);
4094
4095 if (!sbi->error_dirty)
4096 return;
4097 if (!test_bit(error, (unsigned long *)sbi->errors))
4098 return;
4099 schedule_work(&sbi->s_error_work);
4100 }
4101
system_going_down(void)4102 static bool system_going_down(void)
4103 {
4104 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4105 || system_state == SYSTEM_RESTART;
4106 }
4107
f2fs_handle_critical_error(struct f2fs_sb_info * sbi,unsigned char reason)4108 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason)
4109 {
4110 struct super_block *sb = sbi->sb;
4111 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4112 bool continue_fs = !shutdown &&
4113 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4114
4115 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4116
4117 if (!f2fs_hw_is_readonly(sbi)) {
4118 save_stop_reason(sbi, reason);
4119
4120 /*
4121 * always create an asynchronous task to record stop_reason
4122 * in order to avoid potential deadlock when running into
4123 * f2fs_record_stop_reason() synchronously.
4124 */
4125 schedule_work(&sbi->s_error_work);
4126 }
4127
4128 /*
4129 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4130 * could panic during 'reboot -f' as the underlying device got already
4131 * disabled.
4132 */
4133 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4134 !shutdown && !system_going_down() &&
4135 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4136 panic("F2FS-fs (device %s): panic forced after error\n",
4137 sb->s_id);
4138
4139 if (shutdown)
4140 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4141
4142 /*
4143 * Continue filesystem operators if errors=continue. Should not set
4144 * RO by shutdown, since RO bypasses thaw_super which can hang the
4145 * system.
4146 */
4147 if (continue_fs || f2fs_readonly(sb) || shutdown) {
4148 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason);
4149 return;
4150 }
4151
4152 f2fs_warn(sbi, "Remounting filesystem read-only");
4153
4154 /*
4155 * We have already set CP_ERROR_FLAG flag to stop all updates
4156 * to filesystem, so it doesn't need to set SB_RDONLY flag here
4157 * because the flag should be set covered w/ sb->s_umount semaphore
4158 * via remount procedure, otherwise, it will confuse code like
4159 * freeze_super() which will lead to deadlocks and other problems.
4160 */
4161 }
4162
f2fs_record_error_work(struct work_struct * work)4163 static void f2fs_record_error_work(struct work_struct *work)
4164 {
4165 struct f2fs_sb_info *sbi = container_of(work,
4166 struct f2fs_sb_info, s_error_work);
4167
4168 f2fs_record_stop_reason(sbi);
4169 }
4170
f2fs_scan_devices(struct f2fs_sb_info * sbi)4171 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4172 {
4173 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4174 unsigned int max_devices = MAX_DEVICES;
4175 unsigned int logical_blksize;
4176 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4177 int i;
4178
4179 /* Initialize single device information */
4180 if (!RDEV(0).path[0]) {
4181 if (!bdev_is_zoned(sbi->sb->s_bdev))
4182 return 0;
4183 max_devices = 1;
4184 }
4185
4186 /*
4187 * Initialize multiple devices information, or single
4188 * zoned block device information.
4189 */
4190 sbi->devs = f2fs_kzalloc(sbi,
4191 array_size(max_devices,
4192 sizeof(struct f2fs_dev_info)),
4193 GFP_KERNEL);
4194 if (!sbi->devs)
4195 return -ENOMEM;
4196
4197 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4198 sbi->aligned_blksize = true;
4199
4200 for (i = 0; i < max_devices; i++) {
4201 if (i == 0)
4202 FDEV(0).bdev = sbi->sb->s_bdev;
4203 else if (!RDEV(i).path[0])
4204 break;
4205
4206 if (max_devices > 1) {
4207 /* Multi-device mount */
4208 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4209 FDEV(i).total_segments =
4210 le32_to_cpu(RDEV(i).total_segments);
4211 if (i == 0) {
4212 FDEV(i).start_blk = 0;
4213 FDEV(i).end_blk = FDEV(i).start_blk +
4214 (FDEV(i).total_segments <<
4215 sbi->log_blocks_per_seg) - 1 +
4216 le32_to_cpu(raw_super->segment0_blkaddr);
4217 } else {
4218 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4219 FDEV(i).end_blk = FDEV(i).start_blk +
4220 (FDEV(i).total_segments <<
4221 sbi->log_blocks_per_seg) - 1;
4222 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
4223 mode, sbi->sb, NULL);
4224 }
4225 }
4226 if (IS_ERR(FDEV(i).bdev))
4227 return PTR_ERR(FDEV(i).bdev);
4228
4229 /* to release errored devices */
4230 sbi->s_ndevs = i + 1;
4231
4232 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4233 sbi->aligned_blksize = false;
4234
4235 #ifdef CONFIG_BLK_DEV_ZONED
4236 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
4237 !f2fs_sb_has_blkzoned(sbi)) {
4238 f2fs_err(sbi, "Zoned block device feature not enabled");
4239 return -EINVAL;
4240 }
4241 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
4242 if (init_blkz_info(sbi, i)) {
4243 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4244 return -EINVAL;
4245 }
4246 if (max_devices == 1)
4247 break;
4248 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
4249 i, FDEV(i).path,
4250 FDEV(i).total_segments,
4251 FDEV(i).start_blk, FDEV(i).end_blk,
4252 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
4253 "Host-aware" : "Host-managed");
4254 continue;
4255 }
4256 #endif
4257 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4258 i, FDEV(i).path,
4259 FDEV(i).total_segments,
4260 FDEV(i).start_blk, FDEV(i).end_blk);
4261 }
4262 return 0;
4263 }
4264
f2fs_setup_casefold(struct f2fs_sb_info * sbi)4265 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4266 {
4267 #if IS_ENABLED(CONFIG_UNICODE)
4268 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4269 const struct f2fs_sb_encodings *encoding_info;
4270 struct unicode_map *encoding;
4271 __u16 encoding_flags;
4272
4273 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4274 if (!encoding_info) {
4275 f2fs_err(sbi,
4276 "Encoding requested by superblock is unknown");
4277 return -EINVAL;
4278 }
4279
4280 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4281 encoding = utf8_load(encoding_info->version);
4282 if (IS_ERR(encoding)) {
4283 f2fs_err(sbi,
4284 "can't mount with superblock charset: %s-%u.%u.%u "
4285 "not supported by the kernel. flags: 0x%x.",
4286 encoding_info->name,
4287 unicode_major(encoding_info->version),
4288 unicode_minor(encoding_info->version),
4289 unicode_rev(encoding_info->version),
4290 encoding_flags);
4291 return PTR_ERR(encoding);
4292 }
4293 f2fs_info(sbi, "Using encoding defined by superblock: "
4294 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4295 unicode_major(encoding_info->version),
4296 unicode_minor(encoding_info->version),
4297 unicode_rev(encoding_info->version),
4298 encoding_flags);
4299
4300 sbi->sb->s_encoding = encoding;
4301 sbi->sb->s_encoding_flags = encoding_flags;
4302 }
4303 #else
4304 if (f2fs_sb_has_casefold(sbi)) {
4305 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4306 return -EINVAL;
4307 }
4308 #endif
4309 return 0;
4310 }
4311
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)4312 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4313 {
4314 /* adjust parameters according to the volume size */
4315 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4316 if (f2fs_block_unit_discard(sbi))
4317 SM_I(sbi)->dcc_info->discard_granularity =
4318 MIN_DISCARD_GRANULARITY;
4319 if (!f2fs_lfs_mode(sbi))
4320 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4321 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4322 }
4323
4324 sbi->readdir_ra = true;
4325 }
4326
f2fs_fill_super(struct super_block * sb,void * data,int silent)4327 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4328 {
4329 struct f2fs_sb_info *sbi;
4330 struct f2fs_super_block *raw_super;
4331 struct inode *root;
4332 int err;
4333 bool skip_recovery = false, need_fsck = false;
4334 char *options = NULL;
4335 int recovery, i, valid_super_block;
4336 struct curseg_info *seg_i;
4337 int retry_cnt = 1;
4338 #ifdef CONFIG_QUOTA
4339 bool quota_enabled = false;
4340 #endif
4341
4342 try_onemore:
4343 err = -EINVAL;
4344 raw_super = NULL;
4345 valid_super_block = -1;
4346 recovery = 0;
4347
4348 /* allocate memory for f2fs-specific super block info */
4349 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4350 if (!sbi)
4351 return -ENOMEM;
4352
4353 sbi->sb = sb;
4354
4355 /* initialize locks within allocated memory */
4356 init_f2fs_rwsem(&sbi->gc_lock);
4357 mutex_init(&sbi->writepages);
4358 init_f2fs_rwsem(&sbi->cp_global_sem);
4359 init_f2fs_rwsem(&sbi->node_write);
4360 init_f2fs_rwsem(&sbi->node_change);
4361 spin_lock_init(&sbi->stat_lock);
4362 init_f2fs_rwsem(&sbi->cp_rwsem);
4363 init_f2fs_rwsem(&sbi->quota_sem);
4364 init_waitqueue_head(&sbi->cp_wait);
4365 spin_lock_init(&sbi->error_lock);
4366
4367 for (i = 0; i < NR_INODE_TYPE; i++) {
4368 INIT_LIST_HEAD(&sbi->inode_list[i]);
4369 spin_lock_init(&sbi->inode_lock[i]);
4370 }
4371 mutex_init(&sbi->flush_lock);
4372
4373 /* Load the checksum driver */
4374 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4375 if (IS_ERR(sbi->s_chksum_driver)) {
4376 f2fs_err(sbi, "Cannot load crc32 driver.");
4377 err = PTR_ERR(sbi->s_chksum_driver);
4378 sbi->s_chksum_driver = NULL;
4379 goto free_sbi;
4380 }
4381
4382 /* set a block size */
4383 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4384 f2fs_err(sbi, "unable to set blocksize");
4385 goto free_sbi;
4386 }
4387
4388 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4389 &recovery);
4390 if (err)
4391 goto free_sbi;
4392
4393 sb->s_fs_info = sbi;
4394 sbi->raw_super = raw_super;
4395
4396 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4397 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4398 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4399
4400 /* precompute checksum seed for metadata */
4401 if (f2fs_sb_has_inode_chksum(sbi))
4402 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4403 sizeof(raw_super->uuid));
4404
4405 default_options(sbi, false);
4406 /* parse mount options */
4407 options = kstrdup((const char *)data, GFP_KERNEL);
4408 if (data && !options) {
4409 err = -ENOMEM;
4410 goto free_sb_buf;
4411 }
4412
4413 err = parse_options(sb, options, false);
4414 if (err)
4415 goto free_options;
4416
4417 sb->s_maxbytes = max_file_blocks(NULL) <<
4418 le32_to_cpu(raw_super->log_blocksize);
4419 sb->s_max_links = F2FS_LINK_MAX;
4420
4421 err = f2fs_setup_casefold(sbi);
4422 if (err)
4423 goto free_options;
4424
4425 #ifdef CONFIG_QUOTA
4426 sb->dq_op = &f2fs_quota_operations;
4427 sb->s_qcop = &f2fs_quotactl_ops;
4428 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4429
4430 if (f2fs_sb_has_quota_ino(sbi)) {
4431 for (i = 0; i < MAXQUOTAS; i++) {
4432 if (f2fs_qf_ino(sbi->sb, i))
4433 sbi->nquota_files++;
4434 }
4435 }
4436 #endif
4437
4438 sb->s_op = &f2fs_sops;
4439 #ifdef CONFIG_FS_ENCRYPTION
4440 sb->s_cop = &f2fs_cryptops;
4441 #endif
4442 #ifdef CONFIG_FS_VERITY
4443 sb->s_vop = &f2fs_verityops;
4444 #endif
4445 sb->s_xattr = f2fs_xattr_handlers;
4446 sb->s_export_op = &f2fs_export_ops;
4447 sb->s_magic = F2FS_SUPER_MAGIC;
4448 sb->s_time_gran = 1;
4449 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4450 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4451 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4452 sb->s_iflags |= SB_I_CGROUPWB;
4453
4454 /* init f2fs-specific super block info */
4455 sbi->valid_super_block = valid_super_block;
4456
4457 /* disallow all the data/node/meta page writes */
4458 set_sbi_flag(sbi, SBI_POR_DOING);
4459
4460 err = f2fs_init_write_merge_io(sbi);
4461 if (err)
4462 goto free_bio_info;
4463
4464 init_sb_info(sbi);
4465
4466 err = f2fs_init_iostat(sbi);
4467 if (err)
4468 goto free_bio_info;
4469
4470 err = init_percpu_info(sbi);
4471 if (err)
4472 goto free_iostat;
4473
4474 /* init per sbi slab cache */
4475 err = f2fs_init_xattr_caches(sbi);
4476 if (err)
4477 goto free_percpu;
4478 err = f2fs_init_page_array_cache(sbi);
4479 if (err)
4480 goto free_xattr_cache;
4481
4482 /* get an inode for meta space */
4483 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4484 if (IS_ERR(sbi->meta_inode)) {
4485 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4486 err = PTR_ERR(sbi->meta_inode);
4487 goto free_page_array_cache;
4488 }
4489
4490 err = f2fs_get_valid_checkpoint(sbi);
4491 if (err) {
4492 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4493 goto free_meta_inode;
4494 }
4495
4496 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4497 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4498 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4499 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4500 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4501 }
4502
4503 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4504 set_sbi_flag(sbi, SBI_NEED_FSCK);
4505
4506 /* Initialize device list */
4507 err = f2fs_scan_devices(sbi);
4508 if (err) {
4509 f2fs_err(sbi, "Failed to find devices");
4510 goto free_devices;
4511 }
4512
4513 err = f2fs_init_post_read_wq(sbi);
4514 if (err) {
4515 f2fs_err(sbi, "Failed to initialize post read workqueue");
4516 goto free_devices;
4517 }
4518
4519 sbi->total_valid_node_count =
4520 le32_to_cpu(sbi->ckpt->valid_node_count);
4521 percpu_counter_set(&sbi->total_valid_inode_count,
4522 le32_to_cpu(sbi->ckpt->valid_inode_count));
4523 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4524 sbi->total_valid_block_count =
4525 le64_to_cpu(sbi->ckpt->valid_block_count);
4526 sbi->last_valid_block_count = sbi->total_valid_block_count;
4527 sbi->reserved_blocks = 0;
4528 sbi->current_reserved_blocks = 0;
4529 limit_reserve_root(sbi);
4530 adjust_unusable_cap_perc(sbi);
4531
4532 f2fs_init_extent_cache_info(sbi);
4533
4534 f2fs_init_ino_entry_info(sbi);
4535
4536 f2fs_init_fsync_node_info(sbi);
4537
4538 /* setup checkpoint request control and start checkpoint issue thread */
4539 f2fs_init_ckpt_req_control(sbi);
4540 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4541 test_opt(sbi, MERGE_CHECKPOINT)) {
4542 err = f2fs_start_ckpt_thread(sbi);
4543 if (err) {
4544 f2fs_err(sbi,
4545 "Failed to start F2FS issue_checkpoint_thread (%d)",
4546 err);
4547 goto stop_ckpt_thread;
4548 }
4549 }
4550
4551 /* setup f2fs internal modules */
4552 err = f2fs_build_segment_manager(sbi);
4553 if (err) {
4554 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4555 err);
4556 goto free_sm;
4557 }
4558 err = f2fs_build_node_manager(sbi);
4559 if (err) {
4560 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4561 err);
4562 goto free_nm;
4563 }
4564
4565 /* For write statistics */
4566 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4567
4568 /* Read accumulated write IO statistics if exists */
4569 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4570 if (__exist_node_summaries(sbi))
4571 sbi->kbytes_written =
4572 le64_to_cpu(seg_i->journal->info.kbytes_written);
4573
4574 f2fs_build_gc_manager(sbi);
4575
4576 err = f2fs_build_stats(sbi);
4577 if (err)
4578 goto free_nm;
4579
4580 /* get an inode for node space */
4581 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4582 if (IS_ERR(sbi->node_inode)) {
4583 f2fs_err(sbi, "Failed to read node inode");
4584 err = PTR_ERR(sbi->node_inode);
4585 goto free_stats;
4586 }
4587
4588 /* read root inode and dentry */
4589 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4590 if (IS_ERR(root)) {
4591 f2fs_err(sbi, "Failed to read root inode");
4592 err = PTR_ERR(root);
4593 goto free_node_inode;
4594 }
4595 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4596 !root->i_size || !root->i_nlink) {
4597 iput(root);
4598 err = -EINVAL;
4599 goto free_node_inode;
4600 }
4601
4602 sb->s_root = d_make_root(root); /* allocate root dentry */
4603 if (!sb->s_root) {
4604 err = -ENOMEM;
4605 goto free_node_inode;
4606 }
4607
4608 err = f2fs_init_compress_inode(sbi);
4609 if (err)
4610 goto free_root_inode;
4611
4612 err = f2fs_register_sysfs(sbi);
4613 if (err)
4614 goto free_compress_inode;
4615
4616 #ifdef CONFIG_QUOTA
4617 /* Enable quota usage during mount */
4618 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4619 err = f2fs_enable_quotas(sb);
4620 if (err)
4621 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4622 }
4623
4624 quota_enabled = f2fs_recover_quota_begin(sbi);
4625 #endif
4626 /* if there are any orphan inodes, free them */
4627 err = f2fs_recover_orphan_inodes(sbi);
4628 if (err)
4629 goto free_meta;
4630
4631 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4632 goto reset_checkpoint;
4633
4634 /* recover fsynced data */
4635 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4636 !test_opt(sbi, NORECOVERY)) {
4637 /*
4638 * mount should be failed, when device has readonly mode, and
4639 * previous checkpoint was not done by clean system shutdown.
4640 */
4641 if (f2fs_hw_is_readonly(sbi)) {
4642 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4643 err = f2fs_recover_fsync_data(sbi, true);
4644 if (err > 0) {
4645 err = -EROFS;
4646 f2fs_err(sbi, "Need to recover fsync data, but "
4647 "write access unavailable, please try "
4648 "mount w/ disable_roll_forward or norecovery");
4649 }
4650 if (err < 0)
4651 goto free_meta;
4652 }
4653 f2fs_info(sbi, "write access unavailable, skipping recovery");
4654 goto reset_checkpoint;
4655 }
4656
4657 if (need_fsck)
4658 set_sbi_flag(sbi, SBI_NEED_FSCK);
4659
4660 if (skip_recovery)
4661 goto reset_checkpoint;
4662
4663 err = f2fs_recover_fsync_data(sbi, false);
4664 if (err < 0) {
4665 if (err != -ENOMEM)
4666 skip_recovery = true;
4667 need_fsck = true;
4668 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4669 err);
4670 goto free_meta;
4671 }
4672 } else {
4673 err = f2fs_recover_fsync_data(sbi, true);
4674
4675 if (!f2fs_readonly(sb) && err > 0) {
4676 err = -EINVAL;
4677 f2fs_err(sbi, "Need to recover fsync data");
4678 goto free_meta;
4679 }
4680 }
4681
4682 #ifdef CONFIG_QUOTA
4683 f2fs_recover_quota_end(sbi, quota_enabled);
4684 #endif
4685
4686 /*
4687 * If the f2fs is not readonly and fsync data recovery succeeds,
4688 * check zoned block devices' write pointer consistency.
4689 */
4690 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4691 err = f2fs_check_write_pointer(sbi);
4692 if (err)
4693 goto free_meta;
4694 }
4695
4696 reset_checkpoint:
4697 f2fs_init_inmem_curseg(sbi);
4698
4699 /* f2fs_recover_fsync_data() cleared this already */
4700 clear_sbi_flag(sbi, SBI_POR_DOING);
4701
4702 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4703 err = f2fs_disable_checkpoint(sbi);
4704 if (err)
4705 goto sync_free_meta;
4706 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4707 f2fs_enable_checkpoint(sbi);
4708 }
4709
4710 /*
4711 * If filesystem is not mounted as read-only then
4712 * do start the gc_thread.
4713 */
4714 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4715 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4716 /* After POR, we can run background GC thread.*/
4717 err = f2fs_start_gc_thread(sbi);
4718 if (err)
4719 goto sync_free_meta;
4720 }
4721 kvfree(options);
4722
4723 /* recover broken superblock */
4724 if (recovery) {
4725 err = f2fs_commit_super(sbi, true);
4726 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4727 sbi->valid_super_block ? 1 : 2, err);
4728 }
4729
4730 f2fs_join_shrinker(sbi);
4731
4732 f2fs_tuning_parameters(sbi);
4733
4734 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4735 cur_cp_version(F2FS_CKPT(sbi)));
4736 f2fs_update_time(sbi, CP_TIME);
4737 f2fs_update_time(sbi, REQ_TIME);
4738 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4739 return 0;
4740
4741 sync_free_meta:
4742 /* safe to flush all the data */
4743 sync_filesystem(sbi->sb);
4744 retry_cnt = 0;
4745
4746 free_meta:
4747 #ifdef CONFIG_QUOTA
4748 f2fs_truncate_quota_inode_pages(sb);
4749 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4750 f2fs_quota_off_umount(sbi->sb);
4751 #endif
4752 /*
4753 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4754 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4755 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4756 * falls into an infinite loop in f2fs_sync_meta_pages().
4757 */
4758 truncate_inode_pages_final(META_MAPPING(sbi));
4759 /* evict some inodes being cached by GC */
4760 evict_inodes(sb);
4761 f2fs_unregister_sysfs(sbi);
4762 free_compress_inode:
4763 f2fs_destroy_compress_inode(sbi);
4764 free_root_inode:
4765 dput(sb->s_root);
4766 sb->s_root = NULL;
4767 free_node_inode:
4768 f2fs_release_ino_entry(sbi, true);
4769 truncate_inode_pages_final(NODE_MAPPING(sbi));
4770 iput(sbi->node_inode);
4771 sbi->node_inode = NULL;
4772 free_stats:
4773 f2fs_destroy_stats(sbi);
4774 free_nm:
4775 /* stop discard thread before destroying node manager */
4776 f2fs_stop_discard_thread(sbi);
4777 f2fs_destroy_node_manager(sbi);
4778 free_sm:
4779 f2fs_destroy_segment_manager(sbi);
4780 stop_ckpt_thread:
4781 f2fs_stop_ckpt_thread(sbi);
4782 /* flush s_error_work before sbi destroy */
4783 flush_work(&sbi->s_error_work);
4784 f2fs_destroy_post_read_wq(sbi);
4785 free_devices:
4786 destroy_device_list(sbi);
4787 kvfree(sbi->ckpt);
4788 free_meta_inode:
4789 make_bad_inode(sbi->meta_inode);
4790 iput(sbi->meta_inode);
4791 sbi->meta_inode = NULL;
4792 free_page_array_cache:
4793 f2fs_destroy_page_array_cache(sbi);
4794 free_xattr_cache:
4795 f2fs_destroy_xattr_caches(sbi);
4796 free_percpu:
4797 destroy_percpu_info(sbi);
4798 free_iostat:
4799 f2fs_destroy_iostat(sbi);
4800 free_bio_info:
4801 for (i = 0; i < NR_PAGE_TYPE; i++)
4802 kvfree(sbi->write_io[i]);
4803
4804 #if IS_ENABLED(CONFIG_UNICODE)
4805 utf8_unload(sb->s_encoding);
4806 sb->s_encoding = NULL;
4807 #endif
4808 free_options:
4809 #ifdef CONFIG_QUOTA
4810 for (i = 0; i < MAXQUOTAS; i++)
4811 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4812 #endif
4813 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4814 kvfree(options);
4815 free_sb_buf:
4816 kfree(raw_super);
4817 free_sbi:
4818 if (sbi->s_chksum_driver)
4819 crypto_free_shash(sbi->s_chksum_driver);
4820 kfree(sbi);
4821
4822 /* give only one another chance */
4823 if (retry_cnt > 0 && skip_recovery) {
4824 retry_cnt--;
4825 shrink_dcache_sb(sb);
4826 goto try_onemore;
4827 }
4828 return err;
4829 }
4830
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4831 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4832 const char *dev_name, void *data)
4833 {
4834 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4835 }
4836
kill_f2fs_super(struct super_block * sb)4837 static void kill_f2fs_super(struct super_block *sb)
4838 {
4839 if (sb->s_root) {
4840 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4841
4842 set_sbi_flag(sbi, SBI_IS_CLOSE);
4843 f2fs_stop_gc_thread(sbi);
4844 f2fs_stop_discard_thread(sbi);
4845
4846 #ifdef CONFIG_F2FS_FS_COMPRESSION
4847 /*
4848 * latter evict_inode() can bypass checking and invalidating
4849 * compress inode cache.
4850 */
4851 if (test_opt(sbi, COMPRESS_CACHE))
4852 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4853 #endif
4854
4855 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4856 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4857 struct cp_control cpc = {
4858 .reason = CP_UMOUNT,
4859 };
4860 stat_inc_cp_call_count(sbi, TOTAL_CALL);
4861 f2fs_write_checkpoint(sbi, &cpc);
4862 }
4863
4864 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4865 sb->s_flags &= ~SB_RDONLY;
4866 }
4867 kill_block_super(sb);
4868 }
4869
4870 static struct file_system_type f2fs_fs_type = {
4871 .owner = THIS_MODULE,
4872 .name = "f2fs",
4873 .mount = f2fs_mount,
4874 .kill_sb = kill_f2fs_super,
4875 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4876 };
4877 MODULE_ALIAS_FS("f2fs");
4878
init_inodecache(void)4879 static int __init init_inodecache(void)
4880 {
4881 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4882 sizeof(struct f2fs_inode_info), 0,
4883 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4884 return f2fs_inode_cachep ? 0 : -ENOMEM;
4885 }
4886
destroy_inodecache(void)4887 static void destroy_inodecache(void)
4888 {
4889 /*
4890 * Make sure all delayed rcu free inodes are flushed before we
4891 * destroy cache.
4892 */
4893 rcu_barrier();
4894 kmem_cache_destroy(f2fs_inode_cachep);
4895 }
4896
init_f2fs_fs(void)4897 static int __init init_f2fs_fs(void)
4898 {
4899 int err;
4900
4901 if (PAGE_SIZE != F2FS_BLKSIZE) {
4902 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4903 PAGE_SIZE, F2FS_BLKSIZE);
4904 return -EINVAL;
4905 }
4906
4907 err = init_inodecache();
4908 if (err)
4909 goto fail;
4910 err = f2fs_create_node_manager_caches();
4911 if (err)
4912 goto free_inodecache;
4913 err = f2fs_create_segment_manager_caches();
4914 if (err)
4915 goto free_node_manager_caches;
4916 err = f2fs_create_checkpoint_caches();
4917 if (err)
4918 goto free_segment_manager_caches;
4919 err = f2fs_create_recovery_cache();
4920 if (err)
4921 goto free_checkpoint_caches;
4922 err = f2fs_create_extent_cache();
4923 if (err)
4924 goto free_recovery_cache;
4925 err = f2fs_create_garbage_collection_cache();
4926 if (err)
4927 goto free_extent_cache;
4928 err = f2fs_init_sysfs();
4929 if (err)
4930 goto free_garbage_collection_cache;
4931 err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4932 if (err)
4933 goto free_sysfs;
4934 err = register_filesystem(&f2fs_fs_type);
4935 if (err)
4936 goto free_shrinker;
4937 f2fs_create_root_stats();
4938 err = f2fs_init_post_read_processing();
4939 if (err)
4940 goto free_root_stats;
4941 err = f2fs_init_iostat_processing();
4942 if (err)
4943 goto free_post_read;
4944 err = f2fs_init_bio_entry_cache();
4945 if (err)
4946 goto free_iostat;
4947 err = f2fs_init_bioset();
4948 if (err)
4949 goto free_bio_entry_cache;
4950 err = f2fs_init_compress_mempool();
4951 if (err)
4952 goto free_bioset;
4953 err = f2fs_init_compress_cache();
4954 if (err)
4955 goto free_compress_mempool;
4956 err = f2fs_create_casefold_cache();
4957 if (err)
4958 goto free_compress_cache;
4959 return 0;
4960 free_compress_cache:
4961 f2fs_destroy_compress_cache();
4962 free_compress_mempool:
4963 f2fs_destroy_compress_mempool();
4964 free_bioset:
4965 f2fs_destroy_bioset();
4966 free_bio_entry_cache:
4967 f2fs_destroy_bio_entry_cache();
4968 free_iostat:
4969 f2fs_destroy_iostat_processing();
4970 free_post_read:
4971 f2fs_destroy_post_read_processing();
4972 free_root_stats:
4973 f2fs_destroy_root_stats();
4974 unregister_filesystem(&f2fs_fs_type);
4975 free_shrinker:
4976 unregister_shrinker(&f2fs_shrinker_info);
4977 free_sysfs:
4978 f2fs_exit_sysfs();
4979 free_garbage_collection_cache:
4980 f2fs_destroy_garbage_collection_cache();
4981 free_extent_cache:
4982 f2fs_destroy_extent_cache();
4983 free_recovery_cache:
4984 f2fs_destroy_recovery_cache();
4985 free_checkpoint_caches:
4986 f2fs_destroy_checkpoint_caches();
4987 free_segment_manager_caches:
4988 f2fs_destroy_segment_manager_caches();
4989 free_node_manager_caches:
4990 f2fs_destroy_node_manager_caches();
4991 free_inodecache:
4992 destroy_inodecache();
4993 fail:
4994 return err;
4995 }
4996
exit_f2fs_fs(void)4997 static void __exit exit_f2fs_fs(void)
4998 {
4999 f2fs_destroy_casefold_cache();
5000 f2fs_destroy_compress_cache();
5001 f2fs_destroy_compress_mempool();
5002 f2fs_destroy_bioset();
5003 f2fs_destroy_bio_entry_cache();
5004 f2fs_destroy_iostat_processing();
5005 f2fs_destroy_post_read_processing();
5006 f2fs_destroy_root_stats();
5007 unregister_filesystem(&f2fs_fs_type);
5008 unregister_shrinker(&f2fs_shrinker_info);
5009 f2fs_exit_sysfs();
5010 f2fs_destroy_garbage_collection_cache();
5011 f2fs_destroy_extent_cache();
5012 f2fs_destroy_recovery_cache();
5013 f2fs_destroy_checkpoint_caches();
5014 f2fs_destroy_segment_manager_caches();
5015 f2fs_destroy_node_manager_caches();
5016 destroy_inodecache();
5017 }
5018
5019 module_init(init_f2fs_fs)
5020 module_exit(exit_f2fs_fs)
5021
5022 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5023 MODULE_DESCRIPTION("Flash Friendly File System");
5024 MODULE_LICENSE("GPL");
5025 MODULE_SOFTDEP("pre: crc32");
5026
5027