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
1503 /* if atomic write is not committed, set inode w/ atomic dirty */
1504 if (!ret && f2fs_is_atomic_file(inode) &&
1505 !is_inode_flag_set(inode, FI_ATOMIC_COMMITTED))
1506 set_inode_flag(inode, FI_ATOMIC_DIRTIED);
1507
1508 return ret;
1509 }
1510
f2fs_inode_synced(struct inode * inode)1511 void f2fs_inode_synced(struct inode *inode)
1512 {
1513 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1514
1515 spin_lock(&sbi->inode_lock[DIRTY_META]);
1516 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1517 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1518 return;
1519 }
1520 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1521 list_del_init(&F2FS_I(inode)->gdirty_list);
1522 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1523 }
1524 clear_inode_flag(inode, FI_DIRTY_INODE);
1525 clear_inode_flag(inode, FI_AUTO_RECOVER);
1526 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1527 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1528 }
1529
1530 /*
1531 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1532 *
1533 * We should call set_dirty_inode to write the dirty inode through write_inode.
1534 */
f2fs_dirty_inode(struct inode * inode,int flags)1535 static void f2fs_dirty_inode(struct inode *inode, int flags)
1536 {
1537 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1538
1539 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1540 inode->i_ino == F2FS_META_INO(sbi))
1541 return;
1542
1543 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1544 clear_inode_flag(inode, FI_AUTO_RECOVER);
1545
1546 f2fs_inode_dirtied(inode, false);
1547 }
1548
f2fs_free_inode(struct inode * inode)1549 static void f2fs_free_inode(struct inode *inode)
1550 {
1551 fscrypt_free_inode(inode);
1552 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1553 }
1554
destroy_percpu_info(struct f2fs_sb_info * sbi)1555 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1556 {
1557 percpu_counter_destroy(&sbi->total_valid_inode_count);
1558 percpu_counter_destroy(&sbi->rf_node_block_count);
1559 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1560 }
1561
destroy_device_list(struct f2fs_sb_info * sbi)1562 static void destroy_device_list(struct f2fs_sb_info *sbi)
1563 {
1564 int i;
1565
1566 for (i = 0; i < sbi->s_ndevs; i++) {
1567 if (i > 0)
1568 blkdev_put(FDEV(i).bdev, sbi->sb);
1569 #ifdef CONFIG_BLK_DEV_ZONED
1570 kvfree(FDEV(i).blkz_seq);
1571 #endif
1572 }
1573 kvfree(sbi->devs);
1574 }
1575
f2fs_put_super(struct super_block * sb)1576 static void f2fs_put_super(struct super_block *sb)
1577 {
1578 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1579 int i;
1580 int err = 0;
1581 bool done;
1582
1583 /* unregister procfs/sysfs entries in advance to avoid race case */
1584 f2fs_unregister_sysfs(sbi);
1585
1586 f2fs_quota_off_umount(sb);
1587
1588 /* prevent remaining shrinker jobs */
1589 mutex_lock(&sbi->umount_mutex);
1590
1591 /*
1592 * flush all issued checkpoints and stop checkpoint issue thread.
1593 * after then, all checkpoints should be done by each process context.
1594 */
1595 f2fs_stop_ckpt_thread(sbi);
1596
1597 /*
1598 * We don't need to do checkpoint when superblock is clean.
1599 * But, the previous checkpoint was not done by umount, it needs to do
1600 * clean checkpoint again.
1601 */
1602 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1603 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1604 struct cp_control cpc = {
1605 .reason = CP_UMOUNT,
1606 };
1607 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1608 err = f2fs_write_checkpoint(sbi, &cpc);
1609 }
1610
1611 /* be sure to wait for any on-going discard commands */
1612 done = f2fs_issue_discard_timeout(sbi);
1613 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1614 struct cp_control cpc = {
1615 .reason = CP_UMOUNT | CP_TRIMMED,
1616 };
1617 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1618 err = f2fs_write_checkpoint(sbi, &cpc);
1619 }
1620
1621 /*
1622 * normally superblock is clean, so we need to release this.
1623 * In addition, EIO will skip do checkpoint, we need this as well.
1624 */
1625 f2fs_release_ino_entry(sbi, true);
1626
1627 f2fs_leave_shrinker(sbi);
1628 mutex_unlock(&sbi->umount_mutex);
1629
1630 /* our cp_error case, we can wait for any writeback page */
1631 f2fs_flush_merged_writes(sbi);
1632
1633 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1634
1635 if (err || f2fs_cp_error(sbi)) {
1636 truncate_inode_pages_final(NODE_MAPPING(sbi));
1637 truncate_inode_pages_final(META_MAPPING(sbi));
1638 }
1639
1640 for (i = 0; i < NR_COUNT_TYPE; i++) {
1641 if (!get_pages(sbi, i))
1642 continue;
1643 f2fs_err(sbi, "detect filesystem reference count leak during "
1644 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1645 f2fs_bug_on(sbi, 1);
1646 }
1647
1648 f2fs_bug_on(sbi, sbi->fsync_node_num);
1649
1650 f2fs_destroy_compress_inode(sbi);
1651
1652 iput(sbi->node_inode);
1653 sbi->node_inode = NULL;
1654
1655 iput(sbi->meta_inode);
1656 sbi->meta_inode = NULL;
1657
1658 /*
1659 * iput() can update stat information, if f2fs_write_checkpoint()
1660 * above failed with error.
1661 */
1662 f2fs_destroy_stats(sbi);
1663
1664 /* destroy f2fs internal modules */
1665 f2fs_destroy_node_manager(sbi);
1666 f2fs_destroy_segment_manager(sbi);
1667
1668 /* flush s_error_work before sbi destroy */
1669 flush_work(&sbi->s_error_work);
1670
1671 f2fs_destroy_post_read_wq(sbi);
1672
1673 kvfree(sbi->ckpt);
1674
1675 sb->s_fs_info = NULL;
1676 if (sbi->s_chksum_driver)
1677 crypto_free_shash(sbi->s_chksum_driver);
1678 kfree(sbi->raw_super);
1679
1680 destroy_device_list(sbi);
1681 f2fs_destroy_page_array_cache(sbi);
1682 f2fs_destroy_xattr_caches(sbi);
1683 #ifdef CONFIG_QUOTA
1684 for (i = 0; i < MAXQUOTAS; i++)
1685 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1686 #endif
1687 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1688 destroy_percpu_info(sbi);
1689 f2fs_destroy_iostat(sbi);
1690 for (i = 0; i < NR_PAGE_TYPE; i++)
1691 kvfree(sbi->write_io[i]);
1692 #if IS_ENABLED(CONFIG_UNICODE)
1693 utf8_unload(sb->s_encoding);
1694 #endif
1695 kfree(sbi);
1696 }
1697
f2fs_sync_fs(struct super_block * sb,int sync)1698 int f2fs_sync_fs(struct super_block *sb, int sync)
1699 {
1700 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1701 int err = 0;
1702
1703 if (unlikely(f2fs_cp_error(sbi)))
1704 return 0;
1705 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1706 return 0;
1707
1708 trace_f2fs_sync_fs(sb, sync);
1709
1710 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1711 return -EAGAIN;
1712
1713 if (sync) {
1714 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1715 err = f2fs_issue_checkpoint(sbi);
1716 }
1717
1718 return err;
1719 }
1720
f2fs_freeze(struct super_block * sb)1721 static int f2fs_freeze(struct super_block *sb)
1722 {
1723 if (f2fs_readonly(sb))
1724 return 0;
1725
1726 /* IO error happened before */
1727 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1728 return -EIO;
1729
1730 /* must be clean, since sync_filesystem() was already called */
1731 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1732 return -EINVAL;
1733
1734 /* Let's flush checkpoints and stop the thread. */
1735 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1736
1737 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1738 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1739 return 0;
1740 }
1741
f2fs_unfreeze(struct super_block * sb)1742 static int f2fs_unfreeze(struct super_block *sb)
1743 {
1744 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1745
1746 /*
1747 * It will update discard_max_bytes of mounted lvm device to zero
1748 * after creating snapshot on this lvm device, let's drop all
1749 * remained discards.
1750 * We don't need to disable real-time discard because discard_max_bytes
1751 * will recover after removal of snapshot.
1752 */
1753 if (test_opt(sbi, DISCARD) && !f2fs_hw_support_discard(sbi))
1754 f2fs_issue_discard_timeout(sbi);
1755
1756 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1757 return 0;
1758 }
1759
1760 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1761 static int f2fs_statfs_project(struct super_block *sb,
1762 kprojid_t projid, struct kstatfs *buf)
1763 {
1764 struct kqid qid;
1765 struct dquot *dquot;
1766 u64 limit;
1767 u64 curblock;
1768
1769 qid = make_kqid_projid(projid);
1770 dquot = dqget(sb, qid);
1771 if (IS_ERR(dquot))
1772 return PTR_ERR(dquot);
1773 spin_lock(&dquot->dq_dqb_lock);
1774
1775 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1776 dquot->dq_dqb.dqb_bhardlimit);
1777 limit >>= sb->s_blocksize_bits;
1778
1779 if (limit) {
1780 uint64_t remaining = 0;
1781
1782 curblock = (dquot->dq_dqb.dqb_curspace +
1783 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1784 if (limit > curblock)
1785 remaining = limit - curblock;
1786
1787 buf->f_blocks = min(buf->f_blocks, limit);
1788 buf->f_bfree = min(buf->f_bfree, remaining);
1789 buf->f_bavail = min(buf->f_bavail, remaining);
1790 }
1791
1792 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1793 dquot->dq_dqb.dqb_ihardlimit);
1794
1795 if (limit) {
1796 uint64_t remaining = 0;
1797
1798 if (limit > dquot->dq_dqb.dqb_curinodes)
1799 remaining = limit - dquot->dq_dqb.dqb_curinodes;
1800
1801 buf->f_files = min(buf->f_files, limit);
1802 buf->f_ffree = min(buf->f_ffree, remaining);
1803 }
1804
1805 spin_unlock(&dquot->dq_dqb_lock);
1806 dqput(dquot);
1807 return 0;
1808 }
1809 #endif
1810
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1811 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1812 {
1813 struct super_block *sb = dentry->d_sb;
1814 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1815 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1816 block_t total_count, user_block_count, start_count;
1817 u64 avail_node_count;
1818 unsigned int total_valid_node_count;
1819
1820 total_count = le64_to_cpu(sbi->raw_super->block_count);
1821 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1822 buf->f_type = F2FS_SUPER_MAGIC;
1823 buf->f_bsize = sbi->blocksize;
1824
1825 buf->f_blocks = total_count - start_count;
1826
1827 spin_lock(&sbi->stat_lock);
1828
1829 user_block_count = sbi->user_block_count;
1830 total_valid_node_count = valid_node_count(sbi);
1831 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1832 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1833 sbi->current_reserved_blocks;
1834
1835 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1836 buf->f_bfree = 0;
1837 else
1838 buf->f_bfree -= sbi->unusable_block_count;
1839 spin_unlock(&sbi->stat_lock);
1840
1841 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1842 buf->f_bavail = buf->f_bfree -
1843 F2FS_OPTION(sbi).root_reserved_blocks;
1844 else
1845 buf->f_bavail = 0;
1846
1847 if (avail_node_count > user_block_count) {
1848 buf->f_files = user_block_count;
1849 buf->f_ffree = buf->f_bavail;
1850 } else {
1851 buf->f_files = avail_node_count;
1852 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1853 buf->f_bavail);
1854 }
1855
1856 buf->f_namelen = F2FS_NAME_LEN;
1857 buf->f_fsid = u64_to_fsid(id);
1858
1859 #ifdef CONFIG_QUOTA
1860 if (is_inode_flag_set(d_inode(dentry), FI_PROJ_INHERIT) &&
1861 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1862 f2fs_statfs_project(sb, F2FS_I(d_inode(dentry))->i_projid, buf);
1863 }
1864 #endif
1865 return 0;
1866 }
1867
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1868 static inline void f2fs_show_quota_options(struct seq_file *seq,
1869 struct super_block *sb)
1870 {
1871 #ifdef CONFIG_QUOTA
1872 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1873
1874 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1875 char *fmtname = "";
1876
1877 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1878 case QFMT_VFS_OLD:
1879 fmtname = "vfsold";
1880 break;
1881 case QFMT_VFS_V0:
1882 fmtname = "vfsv0";
1883 break;
1884 case QFMT_VFS_V1:
1885 fmtname = "vfsv1";
1886 break;
1887 }
1888 seq_printf(seq, ",jqfmt=%s", fmtname);
1889 }
1890
1891 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1892 seq_show_option(seq, "usrjquota",
1893 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1894
1895 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1896 seq_show_option(seq, "grpjquota",
1897 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1898
1899 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1900 seq_show_option(seq, "prjjquota",
1901 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1902 #endif
1903 }
1904
1905 #ifdef CONFIG_F2FS_FS_COMPRESSION
f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1906 static inline void f2fs_show_compress_options(struct seq_file *seq,
1907 struct super_block *sb)
1908 {
1909 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1910 char *algtype = "";
1911 int i;
1912
1913 if (!f2fs_sb_has_compression(sbi))
1914 return;
1915
1916 switch (F2FS_OPTION(sbi).compress_algorithm) {
1917 case COMPRESS_LZO:
1918 algtype = "lzo";
1919 break;
1920 case COMPRESS_LZ4:
1921 algtype = "lz4";
1922 break;
1923 case COMPRESS_ZSTD:
1924 algtype = "zstd";
1925 break;
1926 case COMPRESS_LZORLE:
1927 algtype = "lzo-rle";
1928 break;
1929 }
1930 seq_printf(seq, ",compress_algorithm=%s", algtype);
1931
1932 if (F2FS_OPTION(sbi).compress_level)
1933 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1934
1935 seq_printf(seq, ",compress_log_size=%u",
1936 F2FS_OPTION(sbi).compress_log_size);
1937
1938 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1939 seq_printf(seq, ",compress_extension=%s",
1940 F2FS_OPTION(sbi).extensions[i]);
1941 }
1942
1943 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1944 seq_printf(seq, ",nocompress_extension=%s",
1945 F2FS_OPTION(sbi).noextensions[i]);
1946 }
1947
1948 if (F2FS_OPTION(sbi).compress_chksum)
1949 seq_puts(seq, ",compress_chksum");
1950
1951 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1952 seq_printf(seq, ",compress_mode=%s", "fs");
1953 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1954 seq_printf(seq, ",compress_mode=%s", "user");
1955
1956 if (test_opt(sbi, COMPRESS_CACHE))
1957 seq_puts(seq, ",compress_cache");
1958 }
1959 #endif
1960
f2fs_show_options(struct seq_file * seq,struct dentry * root)1961 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1962 {
1963 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1964
1965 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1966 seq_printf(seq, ",background_gc=%s", "sync");
1967 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1968 seq_printf(seq, ",background_gc=%s", "on");
1969 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1970 seq_printf(seq, ",background_gc=%s", "off");
1971
1972 if (test_opt(sbi, GC_MERGE))
1973 seq_puts(seq, ",gc_merge");
1974 else
1975 seq_puts(seq, ",nogc_merge");
1976
1977 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1978 seq_puts(seq, ",disable_roll_forward");
1979 if (test_opt(sbi, NORECOVERY))
1980 seq_puts(seq, ",norecovery");
1981 if (test_opt(sbi, DISCARD)) {
1982 seq_puts(seq, ",discard");
1983 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1984 seq_printf(seq, ",discard_unit=%s", "block");
1985 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1986 seq_printf(seq, ",discard_unit=%s", "segment");
1987 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1988 seq_printf(seq, ",discard_unit=%s", "section");
1989 } else {
1990 seq_puts(seq, ",nodiscard");
1991 }
1992 #ifdef CONFIG_F2FS_FS_XATTR
1993 if (test_opt(sbi, XATTR_USER))
1994 seq_puts(seq, ",user_xattr");
1995 else
1996 seq_puts(seq, ",nouser_xattr");
1997 if (test_opt(sbi, INLINE_XATTR))
1998 seq_puts(seq, ",inline_xattr");
1999 else
2000 seq_puts(seq, ",noinline_xattr");
2001 if (test_opt(sbi, INLINE_XATTR_SIZE))
2002 seq_printf(seq, ",inline_xattr_size=%u",
2003 F2FS_OPTION(sbi).inline_xattr_size);
2004 #endif
2005 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2006 if (test_opt(sbi, POSIX_ACL))
2007 seq_puts(seq, ",acl");
2008 else
2009 seq_puts(seq, ",noacl");
2010 #endif
2011 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
2012 seq_puts(seq, ",disable_ext_identify");
2013 if (test_opt(sbi, INLINE_DATA))
2014 seq_puts(seq, ",inline_data");
2015 else
2016 seq_puts(seq, ",noinline_data");
2017 if (test_opt(sbi, INLINE_DENTRY))
2018 seq_puts(seq, ",inline_dentry");
2019 else
2020 seq_puts(seq, ",noinline_dentry");
2021 if (test_opt(sbi, FLUSH_MERGE))
2022 seq_puts(seq, ",flush_merge");
2023 else
2024 seq_puts(seq, ",noflush_merge");
2025 if (test_opt(sbi, NOBARRIER))
2026 seq_puts(seq, ",nobarrier");
2027 else
2028 seq_puts(seq, ",barrier");
2029 if (test_opt(sbi, FASTBOOT))
2030 seq_puts(seq, ",fastboot");
2031 if (test_opt(sbi, READ_EXTENT_CACHE))
2032 seq_puts(seq, ",extent_cache");
2033 else
2034 seq_puts(seq, ",noextent_cache");
2035 if (test_opt(sbi, AGE_EXTENT_CACHE))
2036 seq_puts(seq, ",age_extent_cache");
2037 if (test_opt(sbi, DATA_FLUSH))
2038 seq_puts(seq, ",data_flush");
2039
2040 seq_puts(seq, ",mode=");
2041 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2042 seq_puts(seq, "adaptive");
2043 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2044 seq_puts(seq, "lfs");
2045 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2046 seq_puts(seq, "fragment:segment");
2047 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2048 seq_puts(seq, "fragment:block");
2049 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2050 if (test_opt(sbi, RESERVE_ROOT))
2051 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2052 F2FS_OPTION(sbi).root_reserved_blocks,
2053 from_kuid_munged(&init_user_ns,
2054 F2FS_OPTION(sbi).s_resuid),
2055 from_kgid_munged(&init_user_ns,
2056 F2FS_OPTION(sbi).s_resgid));
2057 #ifdef CONFIG_F2FS_FAULT_INJECTION
2058 if (test_opt(sbi, FAULT_INJECTION)) {
2059 seq_printf(seq, ",fault_injection=%u",
2060 F2FS_OPTION(sbi).fault_info.inject_rate);
2061 seq_printf(seq, ",fault_type=%u",
2062 F2FS_OPTION(sbi).fault_info.inject_type);
2063 }
2064 #endif
2065 #ifdef CONFIG_QUOTA
2066 if (test_opt(sbi, QUOTA))
2067 seq_puts(seq, ",quota");
2068 if (test_opt(sbi, USRQUOTA))
2069 seq_puts(seq, ",usrquota");
2070 if (test_opt(sbi, GRPQUOTA))
2071 seq_puts(seq, ",grpquota");
2072 if (test_opt(sbi, PRJQUOTA))
2073 seq_puts(seq, ",prjquota");
2074 #endif
2075 f2fs_show_quota_options(seq, sbi->sb);
2076
2077 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2078
2079 if (sbi->sb->s_flags & SB_INLINECRYPT)
2080 seq_puts(seq, ",inlinecrypt");
2081
2082 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2083 seq_printf(seq, ",alloc_mode=%s", "default");
2084 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2085 seq_printf(seq, ",alloc_mode=%s", "reuse");
2086
2087 if (test_opt(sbi, DISABLE_CHECKPOINT))
2088 seq_printf(seq, ",checkpoint=disable:%u",
2089 F2FS_OPTION(sbi).unusable_cap);
2090 if (test_opt(sbi, MERGE_CHECKPOINT))
2091 seq_puts(seq, ",checkpoint_merge");
2092 else
2093 seq_puts(seq, ",nocheckpoint_merge");
2094 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2095 seq_printf(seq, ",fsync_mode=%s", "posix");
2096 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2097 seq_printf(seq, ",fsync_mode=%s", "strict");
2098 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2099 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2100
2101 #ifdef CONFIG_F2FS_FS_COMPRESSION
2102 f2fs_show_compress_options(seq, sbi->sb);
2103 #endif
2104
2105 if (test_opt(sbi, ATGC))
2106 seq_puts(seq, ",atgc");
2107
2108 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2109 seq_printf(seq, ",memory=%s", "normal");
2110 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2111 seq_printf(seq, ",memory=%s", "low");
2112
2113 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2114 seq_printf(seq, ",errors=%s", "remount-ro");
2115 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2116 seq_printf(seq, ",errors=%s", "continue");
2117 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2118 seq_printf(seq, ",errors=%s", "panic");
2119
2120 return 0;
2121 }
2122
default_options(struct f2fs_sb_info * sbi,bool remount)2123 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2124 {
2125 /* init some FS parameters */
2126 if (!remount) {
2127 set_opt(sbi, READ_EXTENT_CACHE);
2128 clear_opt(sbi, DISABLE_CHECKPOINT);
2129
2130 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2131 set_opt(sbi, DISCARD);
2132
2133 if (f2fs_sb_has_blkzoned(sbi))
2134 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2135 else
2136 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2137 }
2138
2139 if (f2fs_sb_has_readonly(sbi))
2140 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2141 else
2142 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2143
2144 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2145 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2146 SMALL_VOLUME_SEGMENTS)
2147 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2148 else
2149 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2150 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2151 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2152 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2153 if (f2fs_sb_has_compression(sbi)) {
2154 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2155 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2156 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2157 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2158 }
2159 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2160 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2161 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2162
2163 set_opt(sbi, INLINE_XATTR);
2164 set_opt(sbi, INLINE_DATA);
2165 set_opt(sbi, INLINE_DENTRY);
2166 set_opt(sbi, MERGE_CHECKPOINT);
2167 F2FS_OPTION(sbi).unusable_cap = 0;
2168 sbi->sb->s_flags |= SB_LAZYTIME;
2169 if (!f2fs_is_readonly(sbi))
2170 set_opt(sbi, FLUSH_MERGE);
2171 if (f2fs_sb_has_blkzoned(sbi))
2172 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2173 else
2174 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2175
2176 #ifdef CONFIG_F2FS_FS_XATTR
2177 set_opt(sbi, XATTR_USER);
2178 #endif
2179 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2180 set_opt(sbi, POSIX_ACL);
2181 #endif
2182
2183 f2fs_build_fault_attr(sbi, 0, 0);
2184 }
2185
2186 #ifdef CONFIG_QUOTA
2187 static int f2fs_enable_quotas(struct super_block *sb);
2188 #endif
2189
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)2190 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2191 {
2192 unsigned int s_flags = sbi->sb->s_flags;
2193 struct cp_control cpc;
2194 unsigned int gc_mode = sbi->gc_mode;
2195 int err = 0;
2196 int ret;
2197 block_t unusable;
2198
2199 if (s_flags & SB_RDONLY) {
2200 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2201 return -EINVAL;
2202 }
2203 sbi->sb->s_flags |= SB_ACTIVE;
2204
2205 /* check if we need more GC first */
2206 unusable = f2fs_get_unusable_blocks(sbi);
2207 if (!f2fs_disable_cp_again(sbi, unusable))
2208 goto skip_gc;
2209
2210 f2fs_update_time(sbi, DISABLE_TIME);
2211
2212 sbi->gc_mode = GC_URGENT_HIGH;
2213
2214 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2215 struct f2fs_gc_control gc_control = {
2216 .victim_segno = NULL_SEGNO,
2217 .init_gc_type = FG_GC,
2218 .should_migrate_blocks = false,
2219 .err_gc_skipped = true,
2220 .nr_free_secs = 1 };
2221
2222 f2fs_down_write(&sbi->gc_lock);
2223 stat_inc_gc_call_count(sbi, FOREGROUND);
2224 err = f2fs_gc(sbi, &gc_control);
2225 if (err == -ENODATA) {
2226 err = 0;
2227 break;
2228 }
2229 if (err && err != -EAGAIN)
2230 break;
2231 }
2232
2233 ret = sync_filesystem(sbi->sb);
2234 if (ret || err) {
2235 err = ret ? ret : err;
2236 goto restore_flag;
2237 }
2238
2239 unusable = f2fs_get_unusable_blocks(sbi);
2240 if (f2fs_disable_cp_again(sbi, unusable)) {
2241 err = -EAGAIN;
2242 goto restore_flag;
2243 }
2244
2245 skip_gc:
2246 f2fs_down_write(&sbi->gc_lock);
2247 cpc.reason = CP_PAUSE;
2248 set_sbi_flag(sbi, SBI_CP_DISABLED);
2249 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2250 err = f2fs_write_checkpoint(sbi, &cpc);
2251 if (err)
2252 goto out_unlock;
2253
2254 spin_lock(&sbi->stat_lock);
2255 sbi->unusable_block_count = unusable;
2256 spin_unlock(&sbi->stat_lock);
2257
2258 out_unlock:
2259 f2fs_up_write(&sbi->gc_lock);
2260 restore_flag:
2261 sbi->gc_mode = gc_mode;
2262 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2263 return err;
2264 }
2265
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2266 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2267 {
2268 int retry = DEFAULT_RETRY_IO_COUNT;
2269
2270 /* we should flush all the data to keep data consistency */
2271 do {
2272 sync_inodes_sb(sbi->sb);
2273 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2274 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2275
2276 if (unlikely(retry < 0))
2277 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2278
2279 f2fs_down_write(&sbi->gc_lock);
2280 f2fs_dirty_to_prefree(sbi);
2281
2282 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2283 set_sbi_flag(sbi, SBI_IS_DIRTY);
2284 f2fs_up_write(&sbi->gc_lock);
2285
2286 f2fs_sync_fs(sbi->sb, 1);
2287
2288 /* Let's ensure there's no pending checkpoint anymore */
2289 f2fs_flush_ckpt_thread(sbi);
2290 }
2291
f2fs_remount(struct super_block * sb,int * flags,char * data)2292 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2293 {
2294 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2295 struct f2fs_mount_info org_mount_opt;
2296 unsigned long old_sb_flags;
2297 int err;
2298 bool need_restart_gc = false, need_stop_gc = false;
2299 bool need_restart_ckpt = false, need_stop_ckpt = false;
2300 bool need_restart_flush = false, need_stop_flush = false;
2301 bool need_restart_discard = false, need_stop_discard = false;
2302 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2303 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2304 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2305 bool no_atgc = !test_opt(sbi, ATGC);
2306 bool no_discard = !test_opt(sbi, DISCARD);
2307 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2308 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2309 #ifdef CONFIG_QUOTA
2310 int i, j;
2311 #endif
2312
2313 /*
2314 * Save the old mount options in case we
2315 * need to restore them.
2316 */
2317 org_mount_opt = sbi->mount_opt;
2318 old_sb_flags = sb->s_flags;
2319
2320 #ifdef CONFIG_QUOTA
2321 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2322 for (i = 0; i < MAXQUOTAS; i++) {
2323 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2324 org_mount_opt.s_qf_names[i] =
2325 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2326 GFP_KERNEL);
2327 if (!org_mount_opt.s_qf_names[i]) {
2328 for (j = 0; j < i; j++)
2329 kfree(org_mount_opt.s_qf_names[j]);
2330 return -ENOMEM;
2331 }
2332 } else {
2333 org_mount_opt.s_qf_names[i] = NULL;
2334 }
2335 }
2336 #endif
2337
2338 /* recover superblocks we couldn't write due to previous RO mount */
2339 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2340 err = f2fs_commit_super(sbi, false);
2341 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2342 err);
2343 if (!err)
2344 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2345 }
2346
2347 default_options(sbi, true);
2348
2349 /* parse mount options */
2350 err = parse_options(sb, data, true);
2351 if (err)
2352 goto restore_opts;
2353
2354 /* flush outstanding errors before changing fs state */
2355 flush_work(&sbi->s_error_work);
2356
2357 /*
2358 * Previous and new state of filesystem is RO,
2359 * so skip checking GC and FLUSH_MERGE conditions.
2360 */
2361 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2362 goto skip;
2363
2364 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2365 err = -EROFS;
2366 goto restore_opts;
2367 }
2368
2369 #ifdef CONFIG_QUOTA
2370 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2371 err = dquot_suspend(sb, -1);
2372 if (err < 0)
2373 goto restore_opts;
2374 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2375 /* dquot_resume needs RW */
2376 sb->s_flags &= ~SB_RDONLY;
2377 if (sb_any_quota_suspended(sb)) {
2378 dquot_resume(sb, -1);
2379 } else if (f2fs_sb_has_quota_ino(sbi)) {
2380 err = f2fs_enable_quotas(sb);
2381 if (err)
2382 goto restore_opts;
2383 }
2384 }
2385 #endif
2386 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2387 err = -EINVAL;
2388 f2fs_warn(sbi, "LFS is not compatible with IPU");
2389 goto restore_opts;
2390 }
2391
2392 /* disallow enable atgc dynamically */
2393 if (no_atgc == !!test_opt(sbi, ATGC)) {
2394 err = -EINVAL;
2395 f2fs_warn(sbi, "switch atgc option is not allowed");
2396 goto restore_opts;
2397 }
2398
2399 /* disallow enable/disable extent_cache dynamically */
2400 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2401 err = -EINVAL;
2402 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2403 goto restore_opts;
2404 }
2405 /* disallow enable/disable age extent_cache dynamically */
2406 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2407 err = -EINVAL;
2408 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2409 goto restore_opts;
2410 }
2411
2412 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2413 err = -EINVAL;
2414 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2415 goto restore_opts;
2416 }
2417
2418 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2419 err = -EINVAL;
2420 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2421 goto restore_opts;
2422 }
2423
2424 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2425 err = -EINVAL;
2426 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2427 goto restore_opts;
2428 }
2429
2430 /*
2431 * We stop the GC thread if FS is mounted as RO
2432 * or if background_gc = off is passed in mount
2433 * option. Also sync the filesystem.
2434 */
2435 if ((*flags & SB_RDONLY) ||
2436 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2437 !test_opt(sbi, GC_MERGE))) {
2438 if (sbi->gc_thread) {
2439 f2fs_stop_gc_thread(sbi);
2440 need_restart_gc = true;
2441 }
2442 } else if (!sbi->gc_thread) {
2443 err = f2fs_start_gc_thread(sbi);
2444 if (err)
2445 goto restore_opts;
2446 need_stop_gc = true;
2447 }
2448
2449 if (*flags & SB_RDONLY) {
2450 sync_inodes_sb(sb);
2451
2452 set_sbi_flag(sbi, SBI_IS_DIRTY);
2453 set_sbi_flag(sbi, SBI_IS_CLOSE);
2454 f2fs_sync_fs(sb, 1);
2455 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2456 }
2457
2458 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2459 !test_opt(sbi, MERGE_CHECKPOINT)) {
2460 f2fs_stop_ckpt_thread(sbi);
2461 need_restart_ckpt = true;
2462 } else {
2463 /* Flush if the prevous checkpoint, if exists. */
2464 f2fs_flush_ckpt_thread(sbi);
2465
2466 err = f2fs_start_ckpt_thread(sbi);
2467 if (err) {
2468 f2fs_err(sbi,
2469 "Failed to start F2FS issue_checkpoint_thread (%d)",
2470 err);
2471 goto restore_gc;
2472 }
2473 need_stop_ckpt = true;
2474 }
2475
2476 /*
2477 * We stop issue flush thread if FS is mounted as RO
2478 * or if flush_merge is not passed in mount option.
2479 */
2480 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2481 clear_opt(sbi, FLUSH_MERGE);
2482 f2fs_destroy_flush_cmd_control(sbi, false);
2483 need_restart_flush = true;
2484 } else {
2485 err = f2fs_create_flush_cmd_control(sbi);
2486 if (err)
2487 goto restore_ckpt;
2488 need_stop_flush = true;
2489 }
2490
2491 if (no_discard == !!test_opt(sbi, DISCARD)) {
2492 if (test_opt(sbi, DISCARD)) {
2493 err = f2fs_start_discard_thread(sbi);
2494 if (err)
2495 goto restore_flush;
2496 need_stop_discard = true;
2497 } else {
2498 f2fs_stop_discard_thread(sbi);
2499 f2fs_issue_discard_timeout(sbi);
2500 need_restart_discard = true;
2501 }
2502 }
2503
2504 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2505 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2506 err = f2fs_disable_checkpoint(sbi);
2507 if (err)
2508 goto restore_discard;
2509 } else {
2510 f2fs_enable_checkpoint(sbi);
2511 }
2512 }
2513
2514 skip:
2515 #ifdef CONFIG_QUOTA
2516 /* Release old quota file names */
2517 for (i = 0; i < MAXQUOTAS; i++)
2518 kfree(org_mount_opt.s_qf_names[i]);
2519 #endif
2520 /* Update the POSIXACL Flag */
2521 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2522 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2523
2524 limit_reserve_root(sbi);
2525 adjust_unusable_cap_perc(sbi);
2526 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2527 return 0;
2528 restore_discard:
2529 if (need_restart_discard) {
2530 if (f2fs_start_discard_thread(sbi))
2531 f2fs_warn(sbi, "discard has been stopped");
2532 } else if (need_stop_discard) {
2533 f2fs_stop_discard_thread(sbi);
2534 }
2535 restore_flush:
2536 if (need_restart_flush) {
2537 if (f2fs_create_flush_cmd_control(sbi))
2538 f2fs_warn(sbi, "background flush thread has stopped");
2539 } else if (need_stop_flush) {
2540 clear_opt(sbi, FLUSH_MERGE);
2541 f2fs_destroy_flush_cmd_control(sbi, false);
2542 }
2543 restore_ckpt:
2544 if (need_restart_ckpt) {
2545 if (f2fs_start_ckpt_thread(sbi))
2546 f2fs_warn(sbi, "background ckpt thread has stopped");
2547 } else if (need_stop_ckpt) {
2548 f2fs_stop_ckpt_thread(sbi);
2549 }
2550 restore_gc:
2551 if (need_restart_gc) {
2552 if (f2fs_start_gc_thread(sbi))
2553 f2fs_warn(sbi, "background gc thread has stopped");
2554 } else if (need_stop_gc) {
2555 f2fs_stop_gc_thread(sbi);
2556 }
2557 restore_opts:
2558 #ifdef CONFIG_QUOTA
2559 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2560 for (i = 0; i < MAXQUOTAS; i++) {
2561 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2562 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2563 }
2564 #endif
2565 sbi->mount_opt = org_mount_opt;
2566 sb->s_flags = old_sb_flags;
2567 return err;
2568 }
2569
f2fs_shutdown(struct super_block * sb)2570 static void f2fs_shutdown(struct super_block *sb)
2571 {
2572 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false, false);
2573 }
2574
2575 #ifdef CONFIG_QUOTA
f2fs_need_recovery(struct f2fs_sb_info * sbi)2576 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2577 {
2578 /* need to recovery orphan */
2579 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2580 return true;
2581 /* need to recovery data */
2582 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2583 return false;
2584 if (test_opt(sbi, NORECOVERY))
2585 return false;
2586 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2587 }
2588
f2fs_recover_quota_begin(struct f2fs_sb_info * sbi)2589 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2590 {
2591 bool readonly = f2fs_readonly(sbi->sb);
2592
2593 if (!f2fs_need_recovery(sbi))
2594 return false;
2595
2596 /* it doesn't need to check f2fs_sb_has_readonly() */
2597 if (f2fs_hw_is_readonly(sbi))
2598 return false;
2599
2600 if (readonly) {
2601 sbi->sb->s_flags &= ~SB_RDONLY;
2602 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2603 }
2604
2605 /*
2606 * Turn on quotas which were not enabled for read-only mounts if
2607 * filesystem has quota feature, so that they are updated correctly.
2608 */
2609 return f2fs_enable_quota_files(sbi, readonly);
2610 }
2611
f2fs_recover_quota_end(struct f2fs_sb_info * sbi,bool quota_enabled)2612 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2613 bool quota_enabled)
2614 {
2615 if (quota_enabled)
2616 f2fs_quota_off_umount(sbi->sb);
2617
2618 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2619 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2620 sbi->sb->s_flags |= SB_RDONLY;
2621 }
2622 }
2623
2624 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2625 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2626 size_t len, loff_t off)
2627 {
2628 struct inode *inode = sb_dqopt(sb)->files[type];
2629 struct address_space *mapping = inode->i_mapping;
2630 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2631 int offset = off & (sb->s_blocksize - 1);
2632 int tocopy;
2633 size_t toread;
2634 loff_t i_size = i_size_read(inode);
2635 struct page *page;
2636
2637 if (off > i_size)
2638 return 0;
2639
2640 if (off + len > i_size)
2641 len = i_size - off;
2642 toread = len;
2643 while (toread > 0) {
2644 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2645 repeat:
2646 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2647 if (IS_ERR(page)) {
2648 if (PTR_ERR(page) == -ENOMEM) {
2649 memalloc_retry_wait(GFP_NOFS);
2650 goto repeat;
2651 }
2652 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2653 return PTR_ERR(page);
2654 }
2655
2656 lock_page(page);
2657
2658 if (unlikely(page->mapping != mapping)) {
2659 f2fs_put_page(page, 1);
2660 goto repeat;
2661 }
2662 if (unlikely(!PageUptodate(page))) {
2663 f2fs_put_page(page, 1);
2664 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2665 return -EIO;
2666 }
2667
2668 memcpy_from_page(data, page, offset, tocopy);
2669 f2fs_put_page(page, 1);
2670
2671 offset = 0;
2672 toread -= tocopy;
2673 data += tocopy;
2674 blkidx++;
2675 }
2676 return len;
2677 }
2678
2679 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2680 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2681 const char *data, size_t len, loff_t off)
2682 {
2683 struct inode *inode = sb_dqopt(sb)->files[type];
2684 struct address_space *mapping = inode->i_mapping;
2685 const struct address_space_operations *a_ops = mapping->a_ops;
2686 int offset = off & (sb->s_blocksize - 1);
2687 size_t towrite = len;
2688 struct page *page;
2689 void *fsdata = NULL;
2690 int err = 0;
2691 int tocopy;
2692
2693 while (towrite > 0) {
2694 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2695 towrite);
2696 retry:
2697 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2698 &page, &fsdata);
2699 if (unlikely(err)) {
2700 if (err == -ENOMEM) {
2701 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2702 goto retry;
2703 }
2704 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2705 break;
2706 }
2707
2708 memcpy_to_page(page, offset, data, tocopy);
2709
2710 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2711 page, fsdata);
2712 offset = 0;
2713 towrite -= tocopy;
2714 off += tocopy;
2715 data += tocopy;
2716 cond_resched();
2717 }
2718
2719 if (len == towrite)
2720 return err;
2721 inode->i_mtime = inode_set_ctime_current(inode);
2722 f2fs_mark_inode_dirty_sync(inode, false);
2723 return len - towrite;
2724 }
2725
f2fs_dquot_initialize(struct inode * inode)2726 int f2fs_dquot_initialize(struct inode *inode)
2727 {
2728 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2729 return -ESRCH;
2730
2731 return dquot_initialize(inode);
2732 }
2733
f2fs_get_dquots(struct inode * inode)2734 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2735 {
2736 return F2FS_I(inode)->i_dquot;
2737 }
2738
f2fs_get_reserved_space(struct inode * inode)2739 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2740 {
2741 return &F2FS_I(inode)->i_reserved_quota;
2742 }
2743
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2744 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2745 {
2746 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2747 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2748 return 0;
2749 }
2750
2751 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2752 F2FS_OPTION(sbi).s_jquota_fmt, type);
2753 }
2754
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2755 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2756 {
2757 int enabled = 0;
2758 int i, err;
2759
2760 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2761 err = f2fs_enable_quotas(sbi->sb);
2762 if (err) {
2763 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2764 return 0;
2765 }
2766 return 1;
2767 }
2768
2769 for (i = 0; i < MAXQUOTAS; i++) {
2770 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2771 err = f2fs_quota_on_mount(sbi, i);
2772 if (!err) {
2773 enabled = 1;
2774 continue;
2775 }
2776 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2777 err, i);
2778 }
2779 }
2780 return enabled;
2781 }
2782
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2783 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2784 unsigned int flags)
2785 {
2786 struct inode *qf_inode;
2787 unsigned long qf_inum;
2788 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2789 int err;
2790
2791 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2792
2793 qf_inum = f2fs_qf_ino(sb, type);
2794 if (!qf_inum)
2795 return -EPERM;
2796
2797 qf_inode = f2fs_iget(sb, qf_inum);
2798 if (IS_ERR(qf_inode)) {
2799 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2800 return PTR_ERR(qf_inode);
2801 }
2802
2803 /* Don't account quota for quota files to avoid recursion */
2804 inode_lock(qf_inode);
2805 qf_inode->i_flags |= S_NOQUOTA;
2806
2807 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2808 F2FS_I(qf_inode)->i_flags |= qf_flag;
2809 f2fs_set_inode_flags(qf_inode);
2810 }
2811 inode_unlock(qf_inode);
2812
2813 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2814 iput(qf_inode);
2815 return err;
2816 }
2817
f2fs_enable_quotas(struct super_block * sb)2818 static int f2fs_enable_quotas(struct super_block *sb)
2819 {
2820 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2821 int type, err = 0;
2822 unsigned long qf_inum;
2823 bool quota_mopt[MAXQUOTAS] = {
2824 test_opt(sbi, USRQUOTA),
2825 test_opt(sbi, GRPQUOTA),
2826 test_opt(sbi, PRJQUOTA),
2827 };
2828
2829 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2830 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2831 return 0;
2832 }
2833
2834 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2835
2836 for (type = 0; type < MAXQUOTAS; type++) {
2837 qf_inum = f2fs_qf_ino(sb, type);
2838 if (qf_inum) {
2839 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2840 DQUOT_USAGE_ENABLED |
2841 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2842 if (err) {
2843 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2844 type, err);
2845 for (type--; type >= 0; type--)
2846 dquot_quota_off(sb, type);
2847 set_sbi_flag(F2FS_SB(sb),
2848 SBI_QUOTA_NEED_REPAIR);
2849 return err;
2850 }
2851 }
2852 }
2853 return 0;
2854 }
2855
f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2856 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2857 {
2858 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2859 struct address_space *mapping = dqopt->files[type]->i_mapping;
2860 int ret = 0;
2861
2862 ret = dquot_writeback_dquots(sbi->sb, type);
2863 if (ret)
2864 goto out;
2865
2866 ret = filemap_fdatawrite(mapping);
2867 if (ret)
2868 goto out;
2869
2870 /* if we are using journalled quota */
2871 if (is_journalled_quota(sbi))
2872 goto out;
2873
2874 ret = filemap_fdatawait(mapping);
2875
2876 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2877 out:
2878 if (ret)
2879 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2880 return ret;
2881 }
2882
f2fs_quota_sync(struct super_block * sb,int type)2883 int f2fs_quota_sync(struct super_block *sb, int type)
2884 {
2885 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2886 struct quota_info *dqopt = sb_dqopt(sb);
2887 int cnt;
2888 int ret = 0;
2889
2890 /*
2891 * Now when everything is written we can discard the pagecache so
2892 * that userspace sees the changes.
2893 */
2894 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2895
2896 if (type != -1 && cnt != type)
2897 continue;
2898
2899 if (!sb_has_quota_active(sb, cnt))
2900 continue;
2901
2902 if (!f2fs_sb_has_quota_ino(sbi))
2903 inode_lock(dqopt->files[cnt]);
2904
2905 /*
2906 * do_quotactl
2907 * f2fs_quota_sync
2908 * f2fs_down_read(quota_sem)
2909 * dquot_writeback_dquots()
2910 * f2fs_dquot_commit
2911 * block_operation
2912 * f2fs_down_read(quota_sem)
2913 */
2914 f2fs_lock_op(sbi);
2915 f2fs_down_read(&sbi->quota_sem);
2916
2917 ret = f2fs_quota_sync_file(sbi, cnt);
2918
2919 f2fs_up_read(&sbi->quota_sem);
2920 f2fs_unlock_op(sbi);
2921
2922 if (!f2fs_sb_has_quota_ino(sbi))
2923 inode_unlock(dqopt->files[cnt]);
2924
2925 if (ret)
2926 break;
2927 }
2928 return ret;
2929 }
2930
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2931 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2932 const struct path *path)
2933 {
2934 struct inode *inode;
2935 int err;
2936
2937 /* if quota sysfile exists, deny enabling quota with specific file */
2938 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2939 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2940 return -EBUSY;
2941 }
2942
2943 if (path->dentry->d_sb != sb)
2944 return -EXDEV;
2945
2946 err = f2fs_quota_sync(sb, type);
2947 if (err)
2948 return err;
2949
2950 inode = d_inode(path->dentry);
2951
2952 err = filemap_fdatawrite(inode->i_mapping);
2953 if (err)
2954 return err;
2955
2956 err = filemap_fdatawait(inode->i_mapping);
2957 if (err)
2958 return err;
2959
2960 err = dquot_quota_on(sb, type, format_id, path);
2961 if (err)
2962 return err;
2963
2964 inode_lock(inode);
2965 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2966 f2fs_set_inode_flags(inode);
2967 inode_unlock(inode);
2968 f2fs_mark_inode_dirty_sync(inode, false);
2969
2970 return 0;
2971 }
2972
__f2fs_quota_off(struct super_block * sb,int type)2973 static int __f2fs_quota_off(struct super_block *sb, int type)
2974 {
2975 struct inode *inode = sb_dqopt(sb)->files[type];
2976 int err;
2977
2978 if (!inode || !igrab(inode))
2979 return dquot_quota_off(sb, type);
2980
2981 err = f2fs_quota_sync(sb, type);
2982 if (err)
2983 goto out_put;
2984
2985 err = dquot_quota_off(sb, type);
2986 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2987 goto out_put;
2988
2989 inode_lock(inode);
2990 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
2991 f2fs_set_inode_flags(inode);
2992 inode_unlock(inode);
2993 f2fs_mark_inode_dirty_sync(inode, false);
2994 out_put:
2995 iput(inode);
2996 return err;
2997 }
2998
f2fs_quota_off(struct super_block * sb,int type)2999 static int f2fs_quota_off(struct super_block *sb, int type)
3000 {
3001 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3002 int err;
3003
3004 err = __f2fs_quota_off(sb, type);
3005
3006 /*
3007 * quotactl can shutdown journalled quota, result in inconsistence
3008 * between quota record and fs data by following updates, tag the
3009 * flag to let fsck be aware of it.
3010 */
3011 if (is_journalled_quota(sbi))
3012 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3013 return err;
3014 }
3015
f2fs_quota_off_umount(struct super_block * sb)3016 void f2fs_quota_off_umount(struct super_block *sb)
3017 {
3018 int type;
3019 int err;
3020
3021 for (type = 0; type < MAXQUOTAS; type++) {
3022 err = __f2fs_quota_off(sb, type);
3023 if (err) {
3024 int ret = dquot_quota_off(sb, type);
3025
3026 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3027 type, err, ret);
3028 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3029 }
3030 }
3031 /*
3032 * In case of checkpoint=disable, we must flush quota blocks.
3033 * This can cause NULL exception for node_inode in end_io, since
3034 * put_super already dropped it.
3035 */
3036 sync_filesystem(sb);
3037 }
3038
f2fs_truncate_quota_inode_pages(struct super_block * sb)3039 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3040 {
3041 struct quota_info *dqopt = sb_dqopt(sb);
3042 int type;
3043
3044 for (type = 0; type < MAXQUOTAS; type++) {
3045 if (!dqopt->files[type])
3046 continue;
3047 f2fs_inode_synced(dqopt->files[type]);
3048 }
3049 }
3050
f2fs_dquot_commit(struct dquot * dquot)3051 static int f2fs_dquot_commit(struct dquot *dquot)
3052 {
3053 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3054 int ret;
3055
3056 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3057 ret = dquot_commit(dquot);
3058 if (ret < 0)
3059 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3060 f2fs_up_read(&sbi->quota_sem);
3061 return ret;
3062 }
3063
f2fs_dquot_acquire(struct dquot * dquot)3064 static int f2fs_dquot_acquire(struct dquot *dquot)
3065 {
3066 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3067 int ret;
3068
3069 f2fs_down_read(&sbi->quota_sem);
3070 ret = dquot_acquire(dquot);
3071 if (ret < 0)
3072 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3073 f2fs_up_read(&sbi->quota_sem);
3074 return ret;
3075 }
3076
f2fs_dquot_release(struct dquot * dquot)3077 static int f2fs_dquot_release(struct dquot *dquot)
3078 {
3079 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3080 int ret = dquot_release(dquot);
3081
3082 if (ret < 0)
3083 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3084 return ret;
3085 }
3086
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)3087 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3088 {
3089 struct super_block *sb = dquot->dq_sb;
3090 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3091 int ret = dquot_mark_dquot_dirty(dquot);
3092
3093 /* if we are using journalled quota */
3094 if (is_journalled_quota(sbi))
3095 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3096
3097 return ret;
3098 }
3099
f2fs_dquot_commit_info(struct super_block * sb,int type)3100 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3101 {
3102 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3103 int ret = dquot_commit_info(sb, type);
3104
3105 if (ret < 0)
3106 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3107 return ret;
3108 }
3109
f2fs_get_projid(struct inode * inode,kprojid_t * projid)3110 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3111 {
3112 *projid = F2FS_I(inode)->i_projid;
3113 return 0;
3114 }
3115
3116 static const struct dquot_operations f2fs_quota_operations = {
3117 .get_reserved_space = f2fs_get_reserved_space,
3118 .write_dquot = f2fs_dquot_commit,
3119 .acquire_dquot = f2fs_dquot_acquire,
3120 .release_dquot = f2fs_dquot_release,
3121 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3122 .write_info = f2fs_dquot_commit_info,
3123 .alloc_dquot = dquot_alloc,
3124 .destroy_dquot = dquot_destroy,
3125 .get_projid = f2fs_get_projid,
3126 .get_next_id = dquot_get_next_id,
3127 };
3128
3129 static const struct quotactl_ops f2fs_quotactl_ops = {
3130 .quota_on = f2fs_quota_on,
3131 .quota_off = f2fs_quota_off,
3132 .quota_sync = f2fs_quota_sync,
3133 .get_state = dquot_get_state,
3134 .set_info = dquot_set_dqinfo,
3135 .get_dqblk = dquot_get_dqblk,
3136 .set_dqblk = dquot_set_dqblk,
3137 .get_nextdqblk = dquot_get_next_dqblk,
3138 };
3139 #else
f2fs_dquot_initialize(struct inode * inode)3140 int f2fs_dquot_initialize(struct inode *inode)
3141 {
3142 return 0;
3143 }
3144
f2fs_quota_sync(struct super_block * sb,int type)3145 int f2fs_quota_sync(struct super_block *sb, int type)
3146 {
3147 return 0;
3148 }
3149
f2fs_quota_off_umount(struct super_block * sb)3150 void f2fs_quota_off_umount(struct super_block *sb)
3151 {
3152 }
3153 #endif
3154
3155 static const struct super_operations f2fs_sops = {
3156 .alloc_inode = f2fs_alloc_inode,
3157 .free_inode = f2fs_free_inode,
3158 .drop_inode = f2fs_drop_inode,
3159 .write_inode = f2fs_write_inode,
3160 .dirty_inode = f2fs_dirty_inode,
3161 .show_options = f2fs_show_options,
3162 #ifdef CONFIG_QUOTA
3163 .quota_read = f2fs_quota_read,
3164 .quota_write = f2fs_quota_write,
3165 .get_dquots = f2fs_get_dquots,
3166 #endif
3167 .evict_inode = f2fs_evict_inode,
3168 .put_super = f2fs_put_super,
3169 .sync_fs = f2fs_sync_fs,
3170 .freeze_fs = f2fs_freeze,
3171 .unfreeze_fs = f2fs_unfreeze,
3172 .statfs = f2fs_statfs,
3173 .remount_fs = f2fs_remount,
3174 .shutdown = f2fs_shutdown,
3175 };
3176
3177 #ifdef CONFIG_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)3178 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3179 {
3180 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3181 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3182 ctx, len, NULL);
3183 }
3184
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)3185 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3186 void *fs_data)
3187 {
3188 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3189
3190 /*
3191 * Encrypting the root directory is not allowed because fsck
3192 * expects lost+found directory to exist and remain unencrypted
3193 * if LOST_FOUND feature is enabled.
3194 *
3195 */
3196 if (f2fs_sb_has_lost_found(sbi) &&
3197 inode->i_ino == F2FS_ROOT_INO(sbi))
3198 return -EPERM;
3199
3200 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3201 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3202 ctx, len, fs_data, XATTR_CREATE);
3203 }
3204
f2fs_get_dummy_policy(struct super_block * sb)3205 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3206 {
3207 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3208 }
3209
f2fs_has_stable_inodes(struct super_block * sb)3210 static bool f2fs_has_stable_inodes(struct super_block *sb)
3211 {
3212 return true;
3213 }
3214
f2fs_get_ino_and_lblk_bits(struct super_block * sb,int * ino_bits_ret,int * lblk_bits_ret)3215 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3216 int *ino_bits_ret, int *lblk_bits_ret)
3217 {
3218 *ino_bits_ret = 8 * sizeof(nid_t);
3219 *lblk_bits_ret = 8 * sizeof(block_t);
3220 }
3221
f2fs_get_devices(struct super_block * sb,unsigned int * num_devs)3222 static struct block_device **f2fs_get_devices(struct super_block *sb,
3223 unsigned int *num_devs)
3224 {
3225 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3226 struct block_device **devs;
3227 int i;
3228
3229 if (!f2fs_is_multi_device(sbi))
3230 return NULL;
3231
3232 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3233 if (!devs)
3234 return ERR_PTR(-ENOMEM);
3235
3236 for (i = 0; i < sbi->s_ndevs; i++)
3237 devs[i] = FDEV(i).bdev;
3238 *num_devs = sbi->s_ndevs;
3239 return devs;
3240 }
3241
3242 static const struct fscrypt_operations f2fs_cryptops = {
3243 .key_prefix = "f2fs:",
3244 .get_context = f2fs_get_context,
3245 .set_context = f2fs_set_context,
3246 .get_dummy_policy = f2fs_get_dummy_policy,
3247 .empty_dir = f2fs_empty_dir,
3248 .has_stable_inodes = f2fs_has_stable_inodes,
3249 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
3250 .get_devices = f2fs_get_devices,
3251 };
3252 #endif
3253
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)3254 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3255 u64 ino, u32 generation)
3256 {
3257 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3258 struct inode *inode;
3259
3260 if (f2fs_check_nid_range(sbi, ino))
3261 return ERR_PTR(-ESTALE);
3262
3263 /*
3264 * f2fs_iget isn't quite right if the inode is currently unallocated!
3265 * However f2fs_iget currently does appropriate checks to handle stale
3266 * inodes so everything is OK.
3267 */
3268 inode = f2fs_iget(sb, ino);
3269 if (IS_ERR(inode))
3270 return ERR_CAST(inode);
3271 if (unlikely(generation && inode->i_generation != generation)) {
3272 /* we didn't find the right inode.. */
3273 iput(inode);
3274 return ERR_PTR(-ESTALE);
3275 }
3276 return inode;
3277 }
3278
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3279 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3280 int fh_len, int fh_type)
3281 {
3282 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3283 f2fs_nfs_get_inode);
3284 }
3285
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3286 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3287 int fh_len, int fh_type)
3288 {
3289 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3290 f2fs_nfs_get_inode);
3291 }
3292
3293 static const struct export_operations f2fs_export_ops = {
3294 .fh_to_dentry = f2fs_fh_to_dentry,
3295 .fh_to_parent = f2fs_fh_to_parent,
3296 .get_parent = f2fs_get_parent,
3297 };
3298
max_file_blocks(struct inode * inode)3299 loff_t max_file_blocks(struct inode *inode)
3300 {
3301 loff_t result = 0;
3302 loff_t leaf_count;
3303
3304 /*
3305 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3306 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3307 * space in inode.i_addr, it will be more safe to reassign
3308 * result as zero.
3309 */
3310
3311 if (inode && f2fs_compressed_file(inode))
3312 leaf_count = ADDRS_PER_BLOCK(inode);
3313 else
3314 leaf_count = DEF_ADDRS_PER_BLOCK;
3315
3316 /* two direct node blocks */
3317 result += (leaf_count * 2);
3318
3319 /* two indirect node blocks */
3320 leaf_count *= NIDS_PER_BLOCK;
3321 result += (leaf_count * 2);
3322
3323 /* one double indirect node block */
3324 leaf_count *= NIDS_PER_BLOCK;
3325 result += leaf_count;
3326
3327 return result;
3328 }
3329
__f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)3330 static int __f2fs_commit_super(struct buffer_head *bh,
3331 struct f2fs_super_block *super)
3332 {
3333 lock_buffer(bh);
3334 if (super)
3335 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3336 set_buffer_dirty(bh);
3337 unlock_buffer(bh);
3338
3339 /* it's rare case, we can do fua all the time */
3340 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3341 }
3342
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)3343 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3344 struct buffer_head *bh)
3345 {
3346 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3347 (bh->b_data + F2FS_SUPER_OFFSET);
3348 struct super_block *sb = sbi->sb;
3349 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3350 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3351 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3352 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3353 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3354 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3355 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3356 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3357 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3358 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3359 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3360 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3361 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3362 u64 main_end_blkaddr = main_blkaddr +
3363 ((u64)segment_count_main << log_blocks_per_seg);
3364 u64 seg_end_blkaddr = segment0_blkaddr +
3365 ((u64)segment_count << log_blocks_per_seg);
3366
3367 if (segment0_blkaddr != cp_blkaddr) {
3368 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3369 segment0_blkaddr, cp_blkaddr);
3370 return true;
3371 }
3372
3373 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3374 sit_blkaddr) {
3375 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3376 cp_blkaddr, sit_blkaddr,
3377 segment_count_ckpt << log_blocks_per_seg);
3378 return true;
3379 }
3380
3381 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3382 nat_blkaddr) {
3383 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3384 sit_blkaddr, nat_blkaddr,
3385 segment_count_sit << log_blocks_per_seg);
3386 return true;
3387 }
3388
3389 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3390 ssa_blkaddr) {
3391 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3392 nat_blkaddr, ssa_blkaddr,
3393 segment_count_nat << log_blocks_per_seg);
3394 return true;
3395 }
3396
3397 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3398 main_blkaddr) {
3399 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3400 ssa_blkaddr, main_blkaddr,
3401 segment_count_ssa << log_blocks_per_seg);
3402 return true;
3403 }
3404
3405 if (main_end_blkaddr > seg_end_blkaddr) {
3406 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3407 main_blkaddr, seg_end_blkaddr,
3408 segment_count_main << log_blocks_per_seg);
3409 return true;
3410 } else if (main_end_blkaddr < seg_end_blkaddr) {
3411 int err = 0;
3412 char *res;
3413
3414 /* fix in-memory information all the time */
3415 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3416 segment0_blkaddr) >> log_blocks_per_seg);
3417
3418 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3419 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3420 res = "internally";
3421 } else {
3422 err = __f2fs_commit_super(bh, NULL);
3423 res = err ? "failed" : "done";
3424 }
3425 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3426 res, main_blkaddr, seg_end_blkaddr,
3427 segment_count_main << log_blocks_per_seg);
3428 if (err)
3429 return true;
3430 }
3431 return false;
3432 }
3433
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)3434 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3435 struct buffer_head *bh)
3436 {
3437 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3438 block_t total_sections, blocks_per_seg;
3439 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3440 (bh->b_data + F2FS_SUPER_OFFSET);
3441 size_t crc_offset = 0;
3442 __u32 crc = 0;
3443
3444 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3445 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3446 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3447 return -EINVAL;
3448 }
3449
3450 /* Check checksum_offset and crc in superblock */
3451 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3452 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3453 if (crc_offset !=
3454 offsetof(struct f2fs_super_block, crc)) {
3455 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3456 crc_offset);
3457 return -EFSCORRUPTED;
3458 }
3459 crc = le32_to_cpu(raw_super->crc);
3460 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3461 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3462 return -EFSCORRUPTED;
3463 }
3464 }
3465
3466 /* Currently, support only 4KB block size */
3467 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3468 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3469 le32_to_cpu(raw_super->log_blocksize),
3470 F2FS_BLKSIZE_BITS);
3471 return -EFSCORRUPTED;
3472 }
3473
3474 /* check log blocks per segment */
3475 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3476 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3477 le32_to_cpu(raw_super->log_blocks_per_seg));
3478 return -EFSCORRUPTED;
3479 }
3480
3481 /* Currently, support 512/1024/2048/4096 bytes sector size */
3482 if (le32_to_cpu(raw_super->log_sectorsize) >
3483 F2FS_MAX_LOG_SECTOR_SIZE ||
3484 le32_to_cpu(raw_super->log_sectorsize) <
3485 F2FS_MIN_LOG_SECTOR_SIZE) {
3486 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3487 le32_to_cpu(raw_super->log_sectorsize));
3488 return -EFSCORRUPTED;
3489 }
3490 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3491 le32_to_cpu(raw_super->log_sectorsize) !=
3492 F2FS_MAX_LOG_SECTOR_SIZE) {
3493 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3494 le32_to_cpu(raw_super->log_sectors_per_block),
3495 le32_to_cpu(raw_super->log_sectorsize));
3496 return -EFSCORRUPTED;
3497 }
3498
3499 segment_count = le32_to_cpu(raw_super->segment_count);
3500 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3501 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3502 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3503 total_sections = le32_to_cpu(raw_super->section_count);
3504
3505 /* blocks_per_seg should be 512, given the above check */
3506 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3507
3508 if (segment_count > F2FS_MAX_SEGMENT ||
3509 segment_count < F2FS_MIN_SEGMENTS) {
3510 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3511 return -EFSCORRUPTED;
3512 }
3513
3514 if (total_sections > segment_count_main || total_sections < 1 ||
3515 segs_per_sec > segment_count || !segs_per_sec) {
3516 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3517 segment_count, total_sections, segs_per_sec);
3518 return -EFSCORRUPTED;
3519 }
3520
3521 if (segment_count_main != total_sections * segs_per_sec) {
3522 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3523 segment_count_main, total_sections, segs_per_sec);
3524 return -EFSCORRUPTED;
3525 }
3526
3527 if ((segment_count / segs_per_sec) < total_sections) {
3528 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3529 segment_count, segs_per_sec, total_sections);
3530 return -EFSCORRUPTED;
3531 }
3532
3533 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3534 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3535 segment_count, le64_to_cpu(raw_super->block_count));
3536 return -EFSCORRUPTED;
3537 }
3538
3539 if (RDEV(0).path[0]) {
3540 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3541 int i = 1;
3542
3543 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3544 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3545 i++;
3546 }
3547 if (segment_count != dev_seg_count) {
3548 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3549 segment_count, dev_seg_count);
3550 return -EFSCORRUPTED;
3551 }
3552 } else {
3553 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3554 !bdev_is_zoned(sbi->sb->s_bdev)) {
3555 f2fs_info(sbi, "Zoned block device path is missing");
3556 return -EFSCORRUPTED;
3557 }
3558 }
3559
3560 if (secs_per_zone > total_sections || !secs_per_zone) {
3561 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3562 secs_per_zone, total_sections);
3563 return -EFSCORRUPTED;
3564 }
3565 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3566 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3567 (le32_to_cpu(raw_super->extension_count) +
3568 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3569 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3570 le32_to_cpu(raw_super->extension_count),
3571 raw_super->hot_ext_count,
3572 F2FS_MAX_EXTENSION);
3573 return -EFSCORRUPTED;
3574 }
3575
3576 if (le32_to_cpu(raw_super->cp_payload) >=
3577 (blocks_per_seg - F2FS_CP_PACKS -
3578 NR_CURSEG_PERSIST_TYPE)) {
3579 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3580 le32_to_cpu(raw_super->cp_payload),
3581 blocks_per_seg - F2FS_CP_PACKS -
3582 NR_CURSEG_PERSIST_TYPE);
3583 return -EFSCORRUPTED;
3584 }
3585
3586 /* check reserved ino info */
3587 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3588 le32_to_cpu(raw_super->meta_ino) != 2 ||
3589 le32_to_cpu(raw_super->root_ino) != 3) {
3590 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3591 le32_to_cpu(raw_super->node_ino),
3592 le32_to_cpu(raw_super->meta_ino),
3593 le32_to_cpu(raw_super->root_ino));
3594 return -EFSCORRUPTED;
3595 }
3596
3597 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3598 if (sanity_check_area_boundary(sbi, bh))
3599 return -EFSCORRUPTED;
3600
3601 return 0;
3602 }
3603
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3604 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3605 {
3606 unsigned int total, fsmeta;
3607 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3608 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3609 unsigned int ovp_segments, reserved_segments;
3610 unsigned int main_segs, blocks_per_seg;
3611 unsigned int sit_segs, nat_segs;
3612 unsigned int sit_bitmap_size, nat_bitmap_size;
3613 unsigned int log_blocks_per_seg;
3614 unsigned int segment_count_main;
3615 unsigned int cp_pack_start_sum, cp_payload;
3616 block_t user_block_count, valid_user_blocks;
3617 block_t avail_node_count, valid_node_count;
3618 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3619 unsigned int sit_blk_cnt;
3620 int i, j;
3621
3622 total = le32_to_cpu(raw_super->segment_count);
3623 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3624 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3625 fsmeta += sit_segs;
3626 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3627 fsmeta += nat_segs;
3628 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3629 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3630
3631 if (unlikely(fsmeta >= total))
3632 return 1;
3633
3634 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3635 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3636
3637 if (!f2fs_sb_has_readonly(sbi) &&
3638 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3639 ovp_segments == 0 || reserved_segments == 0)) {
3640 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3641 return 1;
3642 }
3643 user_block_count = le64_to_cpu(ckpt->user_block_count);
3644 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3645 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3646 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3647 if (!user_block_count || user_block_count >=
3648 segment_count_main << log_blocks_per_seg) {
3649 f2fs_err(sbi, "Wrong user_block_count: %u",
3650 user_block_count);
3651 return 1;
3652 }
3653
3654 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3655 if (valid_user_blocks > user_block_count) {
3656 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3657 valid_user_blocks, user_block_count);
3658 return 1;
3659 }
3660
3661 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3662 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3663 if (valid_node_count > avail_node_count) {
3664 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3665 valid_node_count, avail_node_count);
3666 return 1;
3667 }
3668
3669 main_segs = le32_to_cpu(raw_super->segment_count_main);
3670 blocks_per_seg = BLKS_PER_SEG(sbi);
3671
3672 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3673 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3674 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3675 return 1;
3676
3677 if (f2fs_sb_has_readonly(sbi))
3678 goto check_data;
3679
3680 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3681 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3682 le32_to_cpu(ckpt->cur_node_segno[j])) {
3683 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3684 i, j,
3685 le32_to_cpu(ckpt->cur_node_segno[i]));
3686 return 1;
3687 }
3688 }
3689 }
3690 check_data:
3691 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3692 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3693 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3694 return 1;
3695
3696 if (f2fs_sb_has_readonly(sbi))
3697 goto skip_cross;
3698
3699 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3700 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3701 le32_to_cpu(ckpt->cur_data_segno[j])) {
3702 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3703 i, j,
3704 le32_to_cpu(ckpt->cur_data_segno[i]));
3705 return 1;
3706 }
3707 }
3708 }
3709 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3710 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3711 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3712 le32_to_cpu(ckpt->cur_data_segno[j])) {
3713 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3714 i, j,
3715 le32_to_cpu(ckpt->cur_node_segno[i]));
3716 return 1;
3717 }
3718 }
3719 }
3720 skip_cross:
3721 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3722 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3723
3724 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3725 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3726 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3727 sit_bitmap_size, nat_bitmap_size);
3728 return 1;
3729 }
3730
3731 sit_blk_cnt = DIV_ROUND_UP(main_segs, SIT_ENTRY_PER_BLOCK);
3732 if (sit_bitmap_size * 8 < sit_blk_cnt) {
3733 f2fs_err(sbi, "Wrong bitmap size: sit: %u, sit_blk_cnt:%u",
3734 sit_bitmap_size, sit_blk_cnt);
3735 return 1;
3736 }
3737
3738 cp_pack_start_sum = __start_sum_addr(sbi);
3739 cp_payload = __cp_payload(sbi);
3740 if (cp_pack_start_sum < cp_payload + 1 ||
3741 cp_pack_start_sum > blocks_per_seg - 1 -
3742 NR_CURSEG_PERSIST_TYPE) {
3743 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3744 cp_pack_start_sum);
3745 return 1;
3746 }
3747
3748 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3749 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3750 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3751 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3752 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3753 le32_to_cpu(ckpt->checksum_offset));
3754 return 1;
3755 }
3756
3757 nat_blocks = nat_segs << log_blocks_per_seg;
3758 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3759 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3760 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3761 (cp_payload + F2FS_CP_PACKS +
3762 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3763 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3764 cp_payload, nat_bits_blocks);
3765 return 1;
3766 }
3767
3768 if (unlikely(f2fs_cp_error(sbi))) {
3769 f2fs_err(sbi, "A bug case: need to run fsck");
3770 return 1;
3771 }
3772 return 0;
3773 }
3774
init_sb_info(struct f2fs_sb_info * sbi)3775 static void init_sb_info(struct f2fs_sb_info *sbi)
3776 {
3777 struct f2fs_super_block *raw_super = sbi->raw_super;
3778 int i;
3779
3780 sbi->log_sectors_per_block =
3781 le32_to_cpu(raw_super->log_sectors_per_block);
3782 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3783 sbi->blocksize = BIT(sbi->log_blocksize);
3784 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3785 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3786 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3787 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3788 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3789 sbi->total_node_count =
3790 ((le32_to_cpu(raw_super->segment_count_nat) / 2) *
3791 NAT_ENTRY_PER_BLOCK) << sbi->log_blocks_per_seg;
3792 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3793 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3794 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3795 sbi->cur_victim_sec = NULL_SECNO;
3796 sbi->gc_mode = GC_NORMAL;
3797 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3798 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3799 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3800 sbi->migration_granularity = SEGS_PER_SEC(sbi);
3801 sbi->seq_file_ra_mul = MIN_RA_MUL;
3802 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3803 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3804 spin_lock_init(&sbi->gc_remaining_trials_lock);
3805 atomic64_set(&sbi->current_atomic_write, 0);
3806
3807 sbi->dir_level = DEF_DIR_LEVEL;
3808 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3809 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3810 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3811 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3812 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3813 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3814 DEF_UMOUNT_DISCARD_TIMEOUT;
3815 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3816
3817 for (i = 0; i < NR_COUNT_TYPE; i++)
3818 atomic_set(&sbi->nr_pages[i], 0);
3819
3820 for (i = 0; i < META; i++)
3821 atomic_set(&sbi->wb_sync_req[i], 0);
3822
3823 INIT_LIST_HEAD(&sbi->s_list);
3824 mutex_init(&sbi->umount_mutex);
3825 init_f2fs_rwsem(&sbi->io_order_lock);
3826 spin_lock_init(&sbi->cp_lock);
3827
3828 sbi->dirty_device = 0;
3829 spin_lock_init(&sbi->dev_lock);
3830
3831 init_f2fs_rwsem(&sbi->sb_lock);
3832 init_f2fs_rwsem(&sbi->pin_sem);
3833 }
3834
init_percpu_info(struct f2fs_sb_info * sbi)3835 static int init_percpu_info(struct f2fs_sb_info *sbi)
3836 {
3837 int err;
3838
3839 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3840 if (err)
3841 return err;
3842
3843 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3844 if (err)
3845 goto err_valid_block;
3846
3847 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3848 GFP_KERNEL);
3849 if (err)
3850 goto err_node_block;
3851 return 0;
3852
3853 err_node_block:
3854 percpu_counter_destroy(&sbi->rf_node_block_count);
3855 err_valid_block:
3856 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3857 return err;
3858 }
3859
3860 #ifdef CONFIG_BLK_DEV_ZONED
3861
3862 struct f2fs_report_zones_args {
3863 struct f2fs_sb_info *sbi;
3864 struct f2fs_dev_info *dev;
3865 };
3866
f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3867 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3868 void *data)
3869 {
3870 struct f2fs_report_zones_args *rz_args = data;
3871 block_t unusable_blocks = (zone->len - zone->capacity) >>
3872 F2FS_LOG_SECTORS_PER_BLOCK;
3873
3874 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3875 return 0;
3876
3877 set_bit(idx, rz_args->dev->blkz_seq);
3878 if (!rz_args->sbi->unusable_blocks_per_sec) {
3879 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3880 return 0;
3881 }
3882 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3883 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3884 return -EINVAL;
3885 }
3886 return 0;
3887 }
3888
init_blkz_info(struct f2fs_sb_info * sbi,int devi)3889 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3890 {
3891 struct block_device *bdev = FDEV(devi).bdev;
3892 sector_t nr_sectors = bdev_nr_sectors(bdev);
3893 struct f2fs_report_zones_args rep_zone_arg;
3894 u64 zone_sectors;
3895 int ret;
3896
3897 if (!f2fs_sb_has_blkzoned(sbi))
3898 return 0;
3899
3900 zone_sectors = bdev_zone_sectors(bdev);
3901 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3902 SECTOR_TO_BLOCK(zone_sectors))
3903 return -EINVAL;
3904 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3905 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3906 sbi->blocks_per_blkz);
3907 if (nr_sectors & (zone_sectors - 1))
3908 FDEV(devi).nr_blkz++;
3909
3910 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3911 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3912 * sizeof(unsigned long),
3913 GFP_KERNEL);
3914 if (!FDEV(devi).blkz_seq)
3915 return -ENOMEM;
3916
3917 rep_zone_arg.sbi = sbi;
3918 rep_zone_arg.dev = &FDEV(devi);
3919
3920 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3921 &rep_zone_arg);
3922 if (ret < 0)
3923 return ret;
3924 return 0;
3925 }
3926 #endif
3927
3928 /*
3929 * Read f2fs raw super block.
3930 * Because we have two copies of super block, so read both of them
3931 * to get the first valid one. If any one of them is broken, we pass
3932 * them recovery flag back to the caller.
3933 */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3934 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3935 struct f2fs_super_block **raw_super,
3936 int *valid_super_block, int *recovery)
3937 {
3938 struct super_block *sb = sbi->sb;
3939 int block;
3940 struct buffer_head *bh;
3941 struct f2fs_super_block *super;
3942 int err = 0;
3943
3944 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3945 if (!super)
3946 return -ENOMEM;
3947
3948 for (block = 0; block < 2; block++) {
3949 bh = sb_bread(sb, block);
3950 if (!bh) {
3951 f2fs_err(sbi, "Unable to read %dth superblock",
3952 block + 1);
3953 err = -EIO;
3954 *recovery = 1;
3955 continue;
3956 }
3957
3958 /* sanity checking of raw super */
3959 err = sanity_check_raw_super(sbi, bh);
3960 if (err) {
3961 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3962 block + 1);
3963 brelse(bh);
3964 *recovery = 1;
3965 continue;
3966 }
3967
3968 if (!*raw_super) {
3969 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3970 sizeof(*super));
3971 *valid_super_block = block;
3972 *raw_super = super;
3973 }
3974 brelse(bh);
3975 }
3976
3977 /* No valid superblock */
3978 if (!*raw_super)
3979 kfree(super);
3980 else
3981 err = 0;
3982
3983 return err;
3984 }
3985
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)3986 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3987 {
3988 struct buffer_head *bh;
3989 __u32 crc = 0;
3990 int err;
3991
3992 if ((recover && f2fs_readonly(sbi->sb)) ||
3993 f2fs_hw_is_readonly(sbi)) {
3994 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3995 return -EROFS;
3996 }
3997
3998 /* we should update superblock crc here */
3999 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
4000 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
4001 offsetof(struct f2fs_super_block, crc));
4002 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
4003 }
4004
4005 /* write back-up superblock first */
4006 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
4007 if (!bh)
4008 return -EIO;
4009 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4010 brelse(bh);
4011
4012 /* if we are in recovery path, skip writing valid superblock */
4013 if (recover || err)
4014 return err;
4015
4016 /* write current valid superblock */
4017 bh = sb_bread(sbi->sb, sbi->valid_super_block);
4018 if (!bh)
4019 return -EIO;
4020 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4021 brelse(bh);
4022 return err;
4023 }
4024
save_stop_reason(struct f2fs_sb_info * sbi,unsigned char reason)4025 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4026 {
4027 unsigned long flags;
4028
4029 spin_lock_irqsave(&sbi->error_lock, flags);
4030 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4031 sbi->stop_reason[reason]++;
4032 spin_unlock_irqrestore(&sbi->error_lock, flags);
4033 }
4034
f2fs_record_stop_reason(struct f2fs_sb_info * sbi)4035 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4036 {
4037 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4038 unsigned long flags;
4039 int err;
4040
4041 f2fs_down_write(&sbi->sb_lock);
4042
4043 spin_lock_irqsave(&sbi->error_lock, flags);
4044 if (sbi->error_dirty) {
4045 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4046 MAX_F2FS_ERRORS);
4047 sbi->error_dirty = false;
4048 }
4049 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4050 spin_unlock_irqrestore(&sbi->error_lock, flags);
4051
4052 err = f2fs_commit_super(sbi, false);
4053
4054 f2fs_up_write(&sbi->sb_lock);
4055 if (err)
4056 f2fs_err(sbi, "f2fs_commit_super fails to record err:%d", err);
4057 }
4058
f2fs_save_errors(struct f2fs_sb_info * sbi,unsigned char flag)4059 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4060 {
4061 unsigned long flags;
4062
4063 spin_lock_irqsave(&sbi->error_lock, flags);
4064 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4065 set_bit(flag, (unsigned long *)sbi->errors);
4066 sbi->error_dirty = true;
4067 }
4068 spin_unlock_irqrestore(&sbi->error_lock, flags);
4069 }
4070
f2fs_update_errors(struct f2fs_sb_info * sbi)4071 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4072 {
4073 unsigned long flags;
4074 bool need_update = false;
4075
4076 spin_lock_irqsave(&sbi->error_lock, flags);
4077 if (sbi->error_dirty) {
4078 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4079 MAX_F2FS_ERRORS);
4080 sbi->error_dirty = false;
4081 need_update = true;
4082 }
4083 spin_unlock_irqrestore(&sbi->error_lock, flags);
4084
4085 return need_update;
4086 }
4087
f2fs_record_errors(struct f2fs_sb_info * sbi,unsigned char error)4088 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4089 {
4090 int err;
4091
4092 f2fs_down_write(&sbi->sb_lock);
4093
4094 if (!f2fs_update_errors(sbi))
4095 goto out_unlock;
4096
4097 err = f2fs_commit_super(sbi, false);
4098 if (err)
4099 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
4100 error, err);
4101 out_unlock:
4102 f2fs_up_write(&sbi->sb_lock);
4103 }
4104
f2fs_handle_error(struct f2fs_sb_info * sbi,unsigned char error)4105 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4106 {
4107 f2fs_save_errors(sbi, error);
4108 f2fs_record_errors(sbi, error);
4109 }
4110
f2fs_handle_error_async(struct f2fs_sb_info * sbi,unsigned char error)4111 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4112 {
4113 f2fs_save_errors(sbi, error);
4114
4115 if (!sbi->error_dirty)
4116 return;
4117 if (!test_bit(error, (unsigned long *)sbi->errors))
4118 return;
4119 schedule_work(&sbi->s_error_work);
4120 }
4121
system_going_down(void)4122 static bool system_going_down(void)
4123 {
4124 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4125 || system_state == SYSTEM_RESTART;
4126 }
4127
f2fs_handle_critical_error(struct f2fs_sb_info * sbi,unsigned char reason)4128 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason)
4129 {
4130 struct super_block *sb = sbi->sb;
4131 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4132 bool continue_fs = !shutdown &&
4133 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4134
4135 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4136
4137 if (!f2fs_hw_is_readonly(sbi)) {
4138 save_stop_reason(sbi, reason);
4139
4140 /*
4141 * always create an asynchronous task to record stop_reason
4142 * in order to avoid potential deadlock when running into
4143 * f2fs_record_stop_reason() synchronously.
4144 */
4145 schedule_work(&sbi->s_error_work);
4146 }
4147
4148 /*
4149 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4150 * could panic during 'reboot -f' as the underlying device got already
4151 * disabled.
4152 */
4153 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4154 !shutdown && !system_going_down() &&
4155 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4156 panic("F2FS-fs (device %s): panic forced after error\n",
4157 sb->s_id);
4158
4159 if (shutdown)
4160 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4161
4162 /*
4163 * Continue filesystem operators if errors=continue. Should not set
4164 * RO by shutdown, since RO bypasses thaw_super which can hang the
4165 * system.
4166 */
4167 if (continue_fs || f2fs_readonly(sb) || shutdown) {
4168 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason);
4169 return;
4170 }
4171
4172 f2fs_warn(sbi, "Remounting filesystem read-only");
4173
4174 /*
4175 * We have already set CP_ERROR_FLAG flag to stop all updates
4176 * to filesystem, so it doesn't need to set SB_RDONLY flag here
4177 * because the flag should be set covered w/ sb->s_umount semaphore
4178 * via remount procedure, otherwise, it will confuse code like
4179 * freeze_super() which will lead to deadlocks and other problems.
4180 */
4181 }
4182
f2fs_record_error_work(struct work_struct * work)4183 static void f2fs_record_error_work(struct work_struct *work)
4184 {
4185 struct f2fs_sb_info *sbi = container_of(work,
4186 struct f2fs_sb_info, s_error_work);
4187
4188 f2fs_record_stop_reason(sbi);
4189 }
4190
f2fs_scan_devices(struct f2fs_sb_info * sbi)4191 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4192 {
4193 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4194 unsigned int max_devices = MAX_DEVICES;
4195 unsigned int logical_blksize;
4196 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4197 int i;
4198
4199 /* Initialize single device information */
4200 if (!RDEV(0).path[0]) {
4201 if (!bdev_is_zoned(sbi->sb->s_bdev))
4202 return 0;
4203 max_devices = 1;
4204 }
4205
4206 /*
4207 * Initialize multiple devices information, or single
4208 * zoned block device information.
4209 */
4210 sbi->devs = f2fs_kzalloc(sbi,
4211 array_size(max_devices,
4212 sizeof(struct f2fs_dev_info)),
4213 GFP_KERNEL);
4214 if (!sbi->devs)
4215 return -ENOMEM;
4216
4217 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4218 sbi->aligned_blksize = true;
4219
4220 for (i = 0; i < max_devices; i++) {
4221 if (i == 0)
4222 FDEV(0).bdev = sbi->sb->s_bdev;
4223 else if (!RDEV(i).path[0])
4224 break;
4225
4226 if (max_devices > 1) {
4227 /* Multi-device mount */
4228 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4229 FDEV(i).total_segments =
4230 le32_to_cpu(RDEV(i).total_segments);
4231 if (i == 0) {
4232 FDEV(i).start_blk = 0;
4233 FDEV(i).end_blk = FDEV(i).start_blk +
4234 (FDEV(i).total_segments <<
4235 sbi->log_blocks_per_seg) - 1 +
4236 le32_to_cpu(raw_super->segment0_blkaddr);
4237 } else {
4238 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4239 FDEV(i).end_blk = FDEV(i).start_blk +
4240 (FDEV(i).total_segments <<
4241 sbi->log_blocks_per_seg) - 1;
4242 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
4243 mode, sbi->sb, NULL);
4244 }
4245 }
4246 if (IS_ERR(FDEV(i).bdev))
4247 return PTR_ERR(FDEV(i).bdev);
4248
4249 /* to release errored devices */
4250 sbi->s_ndevs = i + 1;
4251
4252 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4253 sbi->aligned_blksize = false;
4254
4255 #ifdef CONFIG_BLK_DEV_ZONED
4256 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
4257 !f2fs_sb_has_blkzoned(sbi)) {
4258 f2fs_err(sbi, "Zoned block device feature not enabled");
4259 return -EINVAL;
4260 }
4261 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
4262 if (init_blkz_info(sbi, i)) {
4263 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4264 return -EINVAL;
4265 }
4266 if (max_devices == 1)
4267 break;
4268 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
4269 i, FDEV(i).path,
4270 FDEV(i).total_segments,
4271 FDEV(i).start_blk, FDEV(i).end_blk,
4272 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
4273 "Host-aware" : "Host-managed");
4274 continue;
4275 }
4276 #endif
4277 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4278 i, FDEV(i).path,
4279 FDEV(i).total_segments,
4280 FDEV(i).start_blk, FDEV(i).end_blk);
4281 }
4282 return 0;
4283 }
4284
f2fs_setup_casefold(struct f2fs_sb_info * sbi)4285 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4286 {
4287 #if IS_ENABLED(CONFIG_UNICODE)
4288 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4289 const struct f2fs_sb_encodings *encoding_info;
4290 struct unicode_map *encoding;
4291 __u16 encoding_flags;
4292
4293 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4294 if (!encoding_info) {
4295 f2fs_err(sbi,
4296 "Encoding requested by superblock is unknown");
4297 return -EINVAL;
4298 }
4299
4300 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4301 encoding = utf8_load(encoding_info->version);
4302 if (IS_ERR(encoding)) {
4303 f2fs_err(sbi,
4304 "can't mount with superblock charset: %s-%u.%u.%u "
4305 "not supported by the kernel. flags: 0x%x.",
4306 encoding_info->name,
4307 unicode_major(encoding_info->version),
4308 unicode_minor(encoding_info->version),
4309 unicode_rev(encoding_info->version),
4310 encoding_flags);
4311 return PTR_ERR(encoding);
4312 }
4313 f2fs_info(sbi, "Using encoding defined by superblock: "
4314 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4315 unicode_major(encoding_info->version),
4316 unicode_minor(encoding_info->version),
4317 unicode_rev(encoding_info->version),
4318 encoding_flags);
4319
4320 sbi->sb->s_encoding = encoding;
4321 sbi->sb->s_encoding_flags = encoding_flags;
4322 }
4323 #else
4324 if (f2fs_sb_has_casefold(sbi)) {
4325 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4326 return -EINVAL;
4327 }
4328 #endif
4329 return 0;
4330 }
4331
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)4332 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4333 {
4334 /* adjust parameters according to the volume size */
4335 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4336 if (f2fs_block_unit_discard(sbi))
4337 SM_I(sbi)->dcc_info->discard_granularity =
4338 MIN_DISCARD_GRANULARITY;
4339 if (!f2fs_lfs_mode(sbi))
4340 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4341 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4342 }
4343
4344 sbi->readdir_ra = true;
4345 }
4346
f2fs_fill_super(struct super_block * sb,void * data,int silent)4347 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4348 {
4349 struct f2fs_sb_info *sbi;
4350 struct f2fs_super_block *raw_super;
4351 struct inode *root;
4352 int err;
4353 bool skip_recovery = false, need_fsck = false;
4354 char *options = NULL;
4355 int recovery, i, valid_super_block;
4356 struct curseg_info *seg_i;
4357 int retry_cnt = 1;
4358 #ifdef CONFIG_QUOTA
4359 bool quota_enabled = false;
4360 #endif
4361
4362 try_onemore:
4363 err = -EINVAL;
4364 raw_super = NULL;
4365 valid_super_block = -1;
4366 recovery = 0;
4367
4368 /* allocate memory for f2fs-specific super block info */
4369 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4370 if (!sbi)
4371 return -ENOMEM;
4372
4373 sbi->sb = sb;
4374
4375 /* initialize locks within allocated memory */
4376 init_f2fs_rwsem(&sbi->gc_lock);
4377 mutex_init(&sbi->writepages);
4378 init_f2fs_rwsem(&sbi->cp_global_sem);
4379 init_f2fs_rwsem(&sbi->node_write);
4380 init_f2fs_rwsem(&sbi->node_change);
4381 spin_lock_init(&sbi->stat_lock);
4382 init_f2fs_rwsem(&sbi->cp_rwsem);
4383 init_f2fs_rwsem(&sbi->quota_sem);
4384 init_waitqueue_head(&sbi->cp_wait);
4385 spin_lock_init(&sbi->error_lock);
4386
4387 for (i = 0; i < NR_INODE_TYPE; i++) {
4388 INIT_LIST_HEAD(&sbi->inode_list[i]);
4389 spin_lock_init(&sbi->inode_lock[i]);
4390 }
4391 mutex_init(&sbi->flush_lock);
4392
4393 /* Load the checksum driver */
4394 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4395 if (IS_ERR(sbi->s_chksum_driver)) {
4396 f2fs_err(sbi, "Cannot load crc32 driver.");
4397 err = PTR_ERR(sbi->s_chksum_driver);
4398 sbi->s_chksum_driver = NULL;
4399 goto free_sbi;
4400 }
4401
4402 /* set a block size */
4403 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4404 f2fs_err(sbi, "unable to set blocksize");
4405 goto free_sbi;
4406 }
4407
4408 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4409 &recovery);
4410 if (err)
4411 goto free_sbi;
4412
4413 sb->s_fs_info = sbi;
4414 sbi->raw_super = raw_super;
4415
4416 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4417 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4418 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4419
4420 /* precompute checksum seed for metadata */
4421 if (f2fs_sb_has_inode_chksum(sbi))
4422 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4423 sizeof(raw_super->uuid));
4424
4425 default_options(sbi, false);
4426 /* parse mount options */
4427 options = kstrdup((const char *)data, GFP_KERNEL);
4428 if (data && !options) {
4429 err = -ENOMEM;
4430 goto free_sb_buf;
4431 }
4432
4433 err = parse_options(sb, options, false);
4434 if (err)
4435 goto free_options;
4436
4437 sb->s_maxbytes = max_file_blocks(NULL) <<
4438 le32_to_cpu(raw_super->log_blocksize);
4439 sb->s_max_links = F2FS_LINK_MAX;
4440
4441 err = f2fs_setup_casefold(sbi);
4442 if (err)
4443 goto free_options;
4444
4445 #ifdef CONFIG_QUOTA
4446 sb->dq_op = &f2fs_quota_operations;
4447 sb->s_qcop = &f2fs_quotactl_ops;
4448 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4449
4450 if (f2fs_sb_has_quota_ino(sbi)) {
4451 for (i = 0; i < MAXQUOTAS; i++) {
4452 if (f2fs_qf_ino(sbi->sb, i))
4453 sbi->nquota_files++;
4454 }
4455 }
4456 #endif
4457
4458 sb->s_op = &f2fs_sops;
4459 #ifdef CONFIG_FS_ENCRYPTION
4460 sb->s_cop = &f2fs_cryptops;
4461 #endif
4462 #ifdef CONFIG_FS_VERITY
4463 sb->s_vop = &f2fs_verityops;
4464 #endif
4465 sb->s_xattr = f2fs_xattr_handlers;
4466 sb->s_export_op = &f2fs_export_ops;
4467 sb->s_magic = F2FS_SUPER_MAGIC;
4468 sb->s_time_gran = 1;
4469 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4470 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4471 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4472 sb->s_iflags |= SB_I_CGROUPWB;
4473
4474 /* init f2fs-specific super block info */
4475 sbi->valid_super_block = valid_super_block;
4476
4477 /* disallow all the data/node/meta page writes */
4478 set_sbi_flag(sbi, SBI_POR_DOING);
4479
4480 err = f2fs_init_write_merge_io(sbi);
4481 if (err)
4482 goto free_bio_info;
4483
4484 init_sb_info(sbi);
4485
4486 err = f2fs_init_iostat(sbi);
4487 if (err)
4488 goto free_bio_info;
4489
4490 err = init_percpu_info(sbi);
4491 if (err)
4492 goto free_iostat;
4493
4494 /* init per sbi slab cache */
4495 err = f2fs_init_xattr_caches(sbi);
4496 if (err)
4497 goto free_percpu;
4498 err = f2fs_init_page_array_cache(sbi);
4499 if (err)
4500 goto free_xattr_cache;
4501
4502 /* get an inode for meta space */
4503 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4504 if (IS_ERR(sbi->meta_inode)) {
4505 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4506 err = PTR_ERR(sbi->meta_inode);
4507 goto free_page_array_cache;
4508 }
4509
4510 err = f2fs_get_valid_checkpoint(sbi);
4511 if (err) {
4512 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4513 goto free_meta_inode;
4514 }
4515
4516 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4517 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4518 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4519 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4520 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4521 }
4522
4523 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4524 set_sbi_flag(sbi, SBI_NEED_FSCK);
4525
4526 /* Initialize device list */
4527 err = f2fs_scan_devices(sbi);
4528 if (err) {
4529 f2fs_err(sbi, "Failed to find devices");
4530 goto free_devices;
4531 }
4532
4533 err = f2fs_init_post_read_wq(sbi);
4534 if (err) {
4535 f2fs_err(sbi, "Failed to initialize post read workqueue");
4536 goto free_devices;
4537 }
4538
4539 sbi->total_valid_node_count =
4540 le32_to_cpu(sbi->ckpt->valid_node_count);
4541 percpu_counter_set(&sbi->total_valid_inode_count,
4542 le32_to_cpu(sbi->ckpt->valid_inode_count));
4543 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4544 sbi->total_valid_block_count =
4545 le64_to_cpu(sbi->ckpt->valid_block_count);
4546 sbi->last_valid_block_count = sbi->total_valid_block_count;
4547 sbi->reserved_blocks = 0;
4548 sbi->current_reserved_blocks = 0;
4549 limit_reserve_root(sbi);
4550 adjust_unusable_cap_perc(sbi);
4551
4552 f2fs_init_extent_cache_info(sbi);
4553
4554 f2fs_init_ino_entry_info(sbi);
4555
4556 f2fs_init_fsync_node_info(sbi);
4557
4558 /* setup checkpoint request control and start checkpoint issue thread */
4559 f2fs_init_ckpt_req_control(sbi);
4560 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4561 test_opt(sbi, MERGE_CHECKPOINT)) {
4562 err = f2fs_start_ckpt_thread(sbi);
4563 if (err) {
4564 f2fs_err(sbi,
4565 "Failed to start F2FS issue_checkpoint_thread (%d)",
4566 err);
4567 goto stop_ckpt_thread;
4568 }
4569 }
4570
4571 /* setup f2fs internal modules */
4572 err = f2fs_build_segment_manager(sbi);
4573 if (err) {
4574 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4575 err);
4576 goto free_sm;
4577 }
4578 err = f2fs_build_node_manager(sbi);
4579 if (err) {
4580 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4581 err);
4582 goto free_nm;
4583 }
4584
4585 /* For write statistics */
4586 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4587
4588 /* Read accumulated write IO statistics if exists */
4589 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4590 if (__exist_node_summaries(sbi))
4591 sbi->kbytes_written =
4592 le64_to_cpu(seg_i->journal->info.kbytes_written);
4593
4594 f2fs_build_gc_manager(sbi);
4595
4596 err = f2fs_build_stats(sbi);
4597 if (err)
4598 goto free_nm;
4599
4600 /* get an inode for node space */
4601 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4602 if (IS_ERR(sbi->node_inode)) {
4603 f2fs_err(sbi, "Failed to read node inode");
4604 err = PTR_ERR(sbi->node_inode);
4605 goto free_stats;
4606 }
4607
4608 /* read root inode and dentry */
4609 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4610 if (IS_ERR(root)) {
4611 f2fs_err(sbi, "Failed to read root inode");
4612 err = PTR_ERR(root);
4613 goto free_node_inode;
4614 }
4615 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4616 !root->i_size || !root->i_nlink) {
4617 iput(root);
4618 err = -EINVAL;
4619 goto free_node_inode;
4620 }
4621
4622 sb->s_root = d_make_root(root); /* allocate root dentry */
4623 if (!sb->s_root) {
4624 err = -ENOMEM;
4625 goto free_node_inode;
4626 }
4627
4628 err = f2fs_init_compress_inode(sbi);
4629 if (err)
4630 goto free_root_inode;
4631
4632 err = f2fs_register_sysfs(sbi);
4633 if (err)
4634 goto free_compress_inode;
4635
4636 #ifdef CONFIG_QUOTA
4637 /* Enable quota usage during mount */
4638 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4639 err = f2fs_enable_quotas(sb);
4640 if (err)
4641 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4642 }
4643
4644 quota_enabled = f2fs_recover_quota_begin(sbi);
4645 #endif
4646 /* if there are any orphan inodes, free them */
4647 err = f2fs_recover_orphan_inodes(sbi);
4648 if (err)
4649 goto free_meta;
4650
4651 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4652 goto reset_checkpoint;
4653
4654 /* recover fsynced data */
4655 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4656 !test_opt(sbi, NORECOVERY)) {
4657 /*
4658 * mount should be failed, when device has readonly mode, and
4659 * previous checkpoint was not done by clean system shutdown.
4660 */
4661 if (f2fs_hw_is_readonly(sbi)) {
4662 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4663 err = f2fs_recover_fsync_data(sbi, true);
4664 if (err > 0) {
4665 err = -EROFS;
4666 f2fs_err(sbi, "Need to recover fsync data, but "
4667 "write access unavailable, please try "
4668 "mount w/ disable_roll_forward or norecovery");
4669 }
4670 if (err < 0)
4671 goto free_meta;
4672 }
4673 f2fs_info(sbi, "write access unavailable, skipping recovery");
4674 goto reset_checkpoint;
4675 }
4676
4677 if (need_fsck)
4678 set_sbi_flag(sbi, SBI_NEED_FSCK);
4679
4680 if (skip_recovery)
4681 goto reset_checkpoint;
4682
4683 err = f2fs_recover_fsync_data(sbi, false);
4684 if (err < 0) {
4685 if (err != -ENOMEM)
4686 skip_recovery = true;
4687 need_fsck = true;
4688 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4689 err);
4690 goto free_meta;
4691 }
4692 } else {
4693 err = f2fs_recover_fsync_data(sbi, true);
4694
4695 if (!f2fs_readonly(sb) && err > 0) {
4696 err = -EINVAL;
4697 f2fs_err(sbi, "Need to recover fsync data");
4698 goto free_meta;
4699 }
4700 }
4701
4702 #ifdef CONFIG_QUOTA
4703 f2fs_recover_quota_end(sbi, quota_enabled);
4704 #endif
4705
4706 /*
4707 * If the f2fs is not readonly and fsync data recovery succeeds,
4708 * check zoned block devices' write pointer consistency.
4709 */
4710 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4711 err = f2fs_check_write_pointer(sbi);
4712 if (err)
4713 goto free_meta;
4714 }
4715
4716 reset_checkpoint:
4717 f2fs_init_inmem_curseg(sbi);
4718
4719 /* f2fs_recover_fsync_data() cleared this already */
4720 clear_sbi_flag(sbi, SBI_POR_DOING);
4721
4722 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4723 err = f2fs_disable_checkpoint(sbi);
4724 if (err)
4725 goto sync_free_meta;
4726 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4727 f2fs_enable_checkpoint(sbi);
4728 }
4729
4730 /*
4731 * If filesystem is not mounted as read-only then
4732 * do start the gc_thread.
4733 */
4734 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4735 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4736 /* After POR, we can run background GC thread.*/
4737 err = f2fs_start_gc_thread(sbi);
4738 if (err)
4739 goto sync_free_meta;
4740 }
4741 kvfree(options);
4742
4743 /* recover broken superblock */
4744 if (recovery) {
4745 err = f2fs_commit_super(sbi, true);
4746 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4747 sbi->valid_super_block ? 1 : 2, err);
4748 }
4749
4750 f2fs_join_shrinker(sbi);
4751
4752 f2fs_tuning_parameters(sbi);
4753
4754 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4755 cur_cp_version(F2FS_CKPT(sbi)));
4756 f2fs_update_time(sbi, CP_TIME);
4757 f2fs_update_time(sbi, REQ_TIME);
4758 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4759 return 0;
4760
4761 sync_free_meta:
4762 /* safe to flush all the data */
4763 sync_filesystem(sbi->sb);
4764 retry_cnt = 0;
4765
4766 free_meta:
4767 #ifdef CONFIG_QUOTA
4768 f2fs_truncate_quota_inode_pages(sb);
4769 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4770 f2fs_quota_off_umount(sbi->sb);
4771 #endif
4772 /*
4773 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4774 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4775 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4776 * falls into an infinite loop in f2fs_sync_meta_pages().
4777 */
4778 truncate_inode_pages_final(META_MAPPING(sbi));
4779 /* evict some inodes being cached by GC */
4780 evict_inodes(sb);
4781 f2fs_unregister_sysfs(sbi);
4782 free_compress_inode:
4783 f2fs_destroy_compress_inode(sbi);
4784 free_root_inode:
4785 dput(sb->s_root);
4786 sb->s_root = NULL;
4787 free_node_inode:
4788 f2fs_release_ino_entry(sbi, true);
4789 truncate_inode_pages_final(NODE_MAPPING(sbi));
4790 iput(sbi->node_inode);
4791 sbi->node_inode = NULL;
4792 free_stats:
4793 f2fs_destroy_stats(sbi);
4794 free_nm:
4795 /* stop discard thread before destroying node manager */
4796 f2fs_stop_discard_thread(sbi);
4797 f2fs_destroy_node_manager(sbi);
4798 free_sm:
4799 f2fs_destroy_segment_manager(sbi);
4800 stop_ckpt_thread:
4801 f2fs_stop_ckpt_thread(sbi);
4802 /* flush s_error_work before sbi destroy */
4803 flush_work(&sbi->s_error_work);
4804 f2fs_destroy_post_read_wq(sbi);
4805 free_devices:
4806 destroy_device_list(sbi);
4807 kvfree(sbi->ckpt);
4808 free_meta_inode:
4809 make_bad_inode(sbi->meta_inode);
4810 iput(sbi->meta_inode);
4811 sbi->meta_inode = NULL;
4812 free_page_array_cache:
4813 f2fs_destroy_page_array_cache(sbi);
4814 free_xattr_cache:
4815 f2fs_destroy_xattr_caches(sbi);
4816 free_percpu:
4817 destroy_percpu_info(sbi);
4818 free_iostat:
4819 f2fs_destroy_iostat(sbi);
4820 free_bio_info:
4821 for (i = 0; i < NR_PAGE_TYPE; i++)
4822 kvfree(sbi->write_io[i]);
4823
4824 #if IS_ENABLED(CONFIG_UNICODE)
4825 utf8_unload(sb->s_encoding);
4826 sb->s_encoding = NULL;
4827 #endif
4828 free_options:
4829 #ifdef CONFIG_QUOTA
4830 for (i = 0; i < MAXQUOTAS; i++)
4831 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4832 #endif
4833 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4834 kvfree(options);
4835 free_sb_buf:
4836 kfree(raw_super);
4837 free_sbi:
4838 if (sbi->s_chksum_driver)
4839 crypto_free_shash(sbi->s_chksum_driver);
4840 kfree(sbi);
4841
4842 /* give only one another chance */
4843 if (retry_cnt > 0 && skip_recovery) {
4844 retry_cnt--;
4845 shrink_dcache_sb(sb);
4846 goto try_onemore;
4847 }
4848 return err;
4849 }
4850
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4851 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4852 const char *dev_name, void *data)
4853 {
4854 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4855 }
4856
kill_f2fs_super(struct super_block * sb)4857 static void kill_f2fs_super(struct super_block *sb)
4858 {
4859 if (sb->s_root) {
4860 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4861
4862 set_sbi_flag(sbi, SBI_IS_CLOSE);
4863 f2fs_stop_gc_thread(sbi);
4864 f2fs_stop_discard_thread(sbi);
4865
4866 #ifdef CONFIG_F2FS_FS_COMPRESSION
4867 /*
4868 * latter evict_inode() can bypass checking and invalidating
4869 * compress inode cache.
4870 */
4871 if (test_opt(sbi, COMPRESS_CACHE))
4872 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4873 #endif
4874
4875 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4876 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4877 struct cp_control cpc = {
4878 .reason = CP_UMOUNT,
4879 };
4880 stat_inc_cp_call_count(sbi, TOTAL_CALL);
4881 f2fs_write_checkpoint(sbi, &cpc);
4882 }
4883
4884 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4885 sb->s_flags &= ~SB_RDONLY;
4886 }
4887 kill_block_super(sb);
4888 }
4889
4890 static struct file_system_type f2fs_fs_type = {
4891 .owner = THIS_MODULE,
4892 .name = "f2fs",
4893 .mount = f2fs_mount,
4894 .kill_sb = kill_f2fs_super,
4895 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4896 };
4897 MODULE_ALIAS_FS("f2fs");
4898
init_inodecache(void)4899 static int __init init_inodecache(void)
4900 {
4901 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4902 sizeof(struct f2fs_inode_info), 0,
4903 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4904 return f2fs_inode_cachep ? 0 : -ENOMEM;
4905 }
4906
destroy_inodecache(void)4907 static void destroy_inodecache(void)
4908 {
4909 /*
4910 * Make sure all delayed rcu free inodes are flushed before we
4911 * destroy cache.
4912 */
4913 rcu_barrier();
4914 kmem_cache_destroy(f2fs_inode_cachep);
4915 }
4916
init_f2fs_fs(void)4917 static int __init init_f2fs_fs(void)
4918 {
4919 int err;
4920
4921 if (PAGE_SIZE != F2FS_BLKSIZE) {
4922 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4923 PAGE_SIZE, F2FS_BLKSIZE);
4924 return -EINVAL;
4925 }
4926
4927 err = init_inodecache();
4928 if (err)
4929 goto fail;
4930 err = f2fs_create_node_manager_caches();
4931 if (err)
4932 goto free_inodecache;
4933 err = f2fs_create_segment_manager_caches();
4934 if (err)
4935 goto free_node_manager_caches;
4936 err = f2fs_create_checkpoint_caches();
4937 if (err)
4938 goto free_segment_manager_caches;
4939 err = f2fs_create_recovery_cache();
4940 if (err)
4941 goto free_checkpoint_caches;
4942 err = f2fs_create_extent_cache();
4943 if (err)
4944 goto free_recovery_cache;
4945 err = f2fs_create_garbage_collection_cache();
4946 if (err)
4947 goto free_extent_cache;
4948 err = f2fs_init_sysfs();
4949 if (err)
4950 goto free_garbage_collection_cache;
4951 err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4952 if (err)
4953 goto free_sysfs;
4954 f2fs_create_root_stats();
4955 err = f2fs_init_post_read_processing();
4956 if (err)
4957 goto free_root_stats;
4958 err = f2fs_init_iostat_processing();
4959 if (err)
4960 goto free_post_read;
4961 err = f2fs_init_bio_entry_cache();
4962 if (err)
4963 goto free_iostat;
4964 err = f2fs_init_bioset();
4965 if (err)
4966 goto free_bio_entry_cache;
4967 err = f2fs_init_compress_mempool();
4968 if (err)
4969 goto free_bioset;
4970 err = f2fs_init_compress_cache();
4971 if (err)
4972 goto free_compress_mempool;
4973 err = f2fs_create_casefold_cache();
4974 if (err)
4975 goto free_compress_cache;
4976 err = register_filesystem(&f2fs_fs_type);
4977 if (err)
4978 goto free_casefold_cache;
4979 return 0;
4980 free_casefold_cache:
4981 f2fs_destroy_casefold_cache();
4982 free_compress_cache:
4983 f2fs_destroy_compress_cache();
4984 free_compress_mempool:
4985 f2fs_destroy_compress_mempool();
4986 free_bioset:
4987 f2fs_destroy_bioset();
4988 free_bio_entry_cache:
4989 f2fs_destroy_bio_entry_cache();
4990 free_iostat:
4991 f2fs_destroy_iostat_processing();
4992 free_post_read:
4993 f2fs_destroy_post_read_processing();
4994 free_root_stats:
4995 f2fs_destroy_root_stats();
4996 unregister_shrinker(&f2fs_shrinker_info);
4997 free_sysfs:
4998 f2fs_exit_sysfs();
4999 free_garbage_collection_cache:
5000 f2fs_destroy_garbage_collection_cache();
5001 free_extent_cache:
5002 f2fs_destroy_extent_cache();
5003 free_recovery_cache:
5004 f2fs_destroy_recovery_cache();
5005 free_checkpoint_caches:
5006 f2fs_destroy_checkpoint_caches();
5007 free_segment_manager_caches:
5008 f2fs_destroy_segment_manager_caches();
5009 free_node_manager_caches:
5010 f2fs_destroy_node_manager_caches();
5011 free_inodecache:
5012 destroy_inodecache();
5013 fail:
5014 return err;
5015 }
5016
exit_f2fs_fs(void)5017 static void __exit exit_f2fs_fs(void)
5018 {
5019 unregister_filesystem(&f2fs_fs_type);
5020 f2fs_destroy_casefold_cache();
5021 f2fs_destroy_compress_cache();
5022 f2fs_destroy_compress_mempool();
5023 f2fs_destroy_bioset();
5024 f2fs_destroy_bio_entry_cache();
5025 f2fs_destroy_iostat_processing();
5026 f2fs_destroy_post_read_processing();
5027 f2fs_destroy_root_stats();
5028 unregister_shrinker(&f2fs_shrinker_info);
5029 f2fs_exit_sysfs();
5030 f2fs_destroy_garbage_collection_cache();
5031 f2fs_destroy_extent_cache();
5032 f2fs_destroy_recovery_cache();
5033 f2fs_destroy_checkpoint_caches();
5034 f2fs_destroy_segment_manager_caches();
5035 f2fs_destroy_node_manager_caches();
5036 destroy_inodecache();
5037 }
5038
5039 module_init(init_f2fs_fs)
5040 module_exit(exit_f2fs_fs)
5041
5042 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5043 MODULE_DESCRIPTION("Flash Friendly File System");
5044 MODULE_LICENSE("GPL");
5045 MODULE_SOFTDEP("pre: crc32");
5046
5047