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/statfs.h>
12 #include <linux/buffer_head.h>
13 #include <linux/backing-dev.h>
14 #include <linux/kthread.h>
15 #include <linux/parser.h>
16 #include <linux/mount.h>
17 #include <linux/seq_file.h>
18 #include <linux/proc_fs.h>
19 #include <linux/random.h>
20 #include <linux/exportfs.h>
21 #include <linux/blkdev.h>
22 #include <linux/quotaops.h>
23 #include <linux/f2fs_fs.h>
24 #include <linux/sysfs.h>
25 #include <linux/quota.h>
26 #include <linux/unicode.h>
27 #include <linux/part_stat.h>
28 #include <linux/zstd.h>
29 #include <linux/lz4.h>
30
31 #include "f2fs.h"
32 #include "node.h"
33 #include "segment.h"
34 #include "xattr.h"
35 #include "gc.h"
36
37 #define CREATE_TRACE_POINTS
38 #include <trace/events/f2fs.h>
39
40 static struct kmem_cache *f2fs_inode_cachep;
41
42 #ifdef CONFIG_F2FS_FAULT_INJECTION
43
44 const char *f2fs_fault_name[FAULT_MAX] = {
45 [FAULT_KMALLOC] = "kmalloc",
46 [FAULT_KVMALLOC] = "kvmalloc",
47 [FAULT_PAGE_ALLOC] = "page alloc",
48 [FAULT_PAGE_GET] = "page get",
49 [FAULT_ALLOC_NID] = "alloc nid",
50 [FAULT_ORPHAN] = "orphan",
51 [FAULT_BLOCK] = "no more block",
52 [FAULT_DIR_DEPTH] = "too big dir depth",
53 [FAULT_EVICT_INODE] = "evict_inode fail",
54 [FAULT_TRUNCATE] = "truncate fail",
55 [FAULT_READ_IO] = "read IO error",
56 [FAULT_CHECKPOINT] = "checkpoint error",
57 [FAULT_DISCARD] = "discard error",
58 [FAULT_WRITE_IO] = "write IO error",
59 };
60
f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned int rate,unsigned int type)61 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
62 unsigned int type)
63 {
64 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
65
66 if (rate) {
67 atomic_set(&ffi->inject_ops, 0);
68 ffi->inject_rate = rate;
69 }
70
71 if (type)
72 ffi->inject_type = type;
73
74 if (!rate && !type)
75 memset(ffi, 0, sizeof(struct f2fs_fault_info));
76 }
77 #endif
78
79 /* f2fs-wide shrinker description */
80 static struct shrinker f2fs_shrinker_info = {
81 .scan_objects = f2fs_shrink_scan,
82 .count_objects = f2fs_shrink_count,
83 .seeks = DEFAULT_SEEKS,
84 };
85
86 enum {
87 Opt_gc_background,
88 Opt_disable_roll_forward,
89 Opt_norecovery,
90 Opt_discard,
91 Opt_nodiscard,
92 Opt_noheap,
93 Opt_heap,
94 Opt_user_xattr,
95 Opt_nouser_xattr,
96 Opt_acl,
97 Opt_noacl,
98 Opt_active_logs,
99 Opt_disable_ext_identify,
100 Opt_inline_xattr,
101 Opt_noinline_xattr,
102 Opt_inline_xattr_size,
103 Opt_inline_data,
104 Opt_inline_dentry,
105 Opt_noinline_dentry,
106 Opt_flush_merge,
107 Opt_noflush_merge,
108 Opt_nobarrier,
109 Opt_fastboot,
110 Opt_extent_cache,
111 Opt_noextent_cache,
112 Opt_noinline_data,
113 Opt_data_flush,
114 Opt_reserve_root,
115 Opt_resgid,
116 Opt_resuid,
117 Opt_mode,
118 Opt_io_size_bits,
119 Opt_fault_injection,
120 Opt_fault_type,
121 Opt_lazytime,
122 Opt_nolazytime,
123 Opt_quota,
124 Opt_noquota,
125 Opt_usrquota,
126 Opt_grpquota,
127 Opt_prjquota,
128 Opt_usrjquota,
129 Opt_grpjquota,
130 Opt_prjjquota,
131 Opt_offusrjquota,
132 Opt_offgrpjquota,
133 Opt_offprjjquota,
134 Opt_jqfmt_vfsold,
135 Opt_jqfmt_vfsv0,
136 Opt_jqfmt_vfsv1,
137 Opt_whint,
138 Opt_alloc,
139 Opt_fsync,
140 Opt_test_dummy_encryption,
141 Opt_inlinecrypt,
142 Opt_checkpoint_disable,
143 Opt_checkpoint_disable_cap,
144 Opt_checkpoint_disable_cap_perc,
145 Opt_checkpoint_enable,
146 Opt_checkpoint_merge,
147 Opt_nocheckpoint_merge,
148 Opt_compress_algorithm,
149 Opt_compress_log_size,
150 Opt_compress_extension,
151 Opt_compress_chksum,
152 Opt_compress_mode,
153 Opt_compress_cache,
154 Opt_atgc,
155 Opt_gc_merge,
156 Opt_nogc_merge,
157 Opt_memory_mode,
158 Opt_age_extent_cache,
159 Opt_err,
160 };
161
162 static match_table_t f2fs_tokens = {
163 {Opt_gc_background, "background_gc=%s"},
164 {Opt_disable_roll_forward, "disable_roll_forward"},
165 {Opt_norecovery, "norecovery"},
166 {Opt_discard, "discard"},
167 {Opt_nodiscard, "nodiscard"},
168 {Opt_noheap, "no_heap"},
169 {Opt_heap, "heap"},
170 {Opt_user_xattr, "user_xattr"},
171 {Opt_nouser_xattr, "nouser_xattr"},
172 {Opt_acl, "acl"},
173 {Opt_noacl, "noacl"},
174 {Opt_active_logs, "active_logs=%u"},
175 {Opt_disable_ext_identify, "disable_ext_identify"},
176 {Opt_inline_xattr, "inline_xattr"},
177 {Opt_noinline_xattr, "noinline_xattr"},
178 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
179 {Opt_inline_data, "inline_data"},
180 {Opt_inline_dentry, "inline_dentry"},
181 {Opt_noinline_dentry, "noinline_dentry"},
182 {Opt_flush_merge, "flush_merge"},
183 {Opt_noflush_merge, "noflush_merge"},
184 {Opt_nobarrier, "nobarrier"},
185 {Opt_fastboot, "fastboot"},
186 {Opt_extent_cache, "extent_cache"},
187 {Opt_noextent_cache, "noextent_cache"},
188 {Opt_noinline_data, "noinline_data"},
189 {Opt_data_flush, "data_flush"},
190 {Opt_reserve_root, "reserve_root=%u"},
191 {Opt_resgid, "resgid=%u"},
192 {Opt_resuid, "resuid=%u"},
193 {Opt_mode, "mode=%s"},
194 {Opt_io_size_bits, "io_bits=%u"},
195 {Opt_fault_injection, "fault_injection=%u"},
196 {Opt_fault_type, "fault_type=%u"},
197 {Opt_lazytime, "lazytime"},
198 {Opt_nolazytime, "nolazytime"},
199 {Opt_quota, "quota"},
200 {Opt_noquota, "noquota"},
201 {Opt_usrquota, "usrquota"},
202 {Opt_grpquota, "grpquota"},
203 {Opt_prjquota, "prjquota"},
204 {Opt_usrjquota, "usrjquota=%s"},
205 {Opt_grpjquota, "grpjquota=%s"},
206 {Opt_prjjquota, "prjjquota=%s"},
207 {Opt_offusrjquota, "usrjquota="},
208 {Opt_offgrpjquota, "grpjquota="},
209 {Opt_offprjjquota, "prjjquota="},
210 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
211 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
212 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
213 {Opt_whint, "whint_mode=%s"},
214 {Opt_alloc, "alloc_mode=%s"},
215 {Opt_fsync, "fsync_mode=%s"},
216 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
217 {Opt_test_dummy_encryption, "test_dummy_encryption"},
218 {Opt_inlinecrypt, "inlinecrypt"},
219 {Opt_checkpoint_disable, "checkpoint=disable"},
220 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
221 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
222 {Opt_checkpoint_enable, "checkpoint=enable"},
223 {Opt_checkpoint_merge, "checkpoint_merge"},
224 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
225 {Opt_compress_algorithm, "compress_algorithm=%s"},
226 {Opt_compress_log_size, "compress_log_size=%u"},
227 {Opt_compress_extension, "compress_extension=%s"},
228 {Opt_compress_chksum, "compress_chksum"},
229 {Opt_compress_mode, "compress_mode=%s"},
230 {Opt_compress_cache, "compress_cache"},
231 {Opt_atgc, "atgc"},
232 {Opt_gc_merge, "gc_merge"},
233 {Opt_nogc_merge, "nogc_merge"},
234 {Opt_memory_mode, "memory=%s"},
235 {Opt_age_extent_cache, "age_extent_cache"},
236 {Opt_err, NULL},
237 };
238
f2fs_printk(struct f2fs_sb_info * sbi,const char * fmt,...)239 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
240 {
241 struct va_format vaf;
242 va_list args;
243 int level;
244
245 va_start(args, fmt);
246
247 level = printk_get_level(fmt);
248 vaf.fmt = printk_skip_level(fmt);
249 vaf.va = &args;
250 printk("%c%cF2FS-fs (%s): %pV\n",
251 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
252
253 va_end(args);
254 }
255
256 #ifdef CONFIG_UNICODE
257 static const struct f2fs_sb_encodings {
258 __u16 magic;
259 char *name;
260 char *version;
261 } f2fs_sb_encoding_map[] = {
262 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
263 };
264
f2fs_sb_read_encoding(const struct f2fs_super_block * sb,const struct f2fs_sb_encodings ** encoding,__u16 * flags)265 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
266 const struct f2fs_sb_encodings **encoding,
267 __u16 *flags)
268 {
269 __u16 magic = le16_to_cpu(sb->s_encoding);
270 int i;
271
272 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
273 if (magic == f2fs_sb_encoding_map[i].magic)
274 break;
275
276 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
277 return -EINVAL;
278
279 *encoding = &f2fs_sb_encoding_map[i];
280 *flags = le16_to_cpu(sb->s_encoding_flags);
281
282 return 0;
283 }
284
285 struct kmem_cache *f2fs_cf_name_slab;
f2fs_create_casefold_cache(void)286 static int __init f2fs_create_casefold_cache(void)
287 {
288 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
289 F2FS_NAME_LEN);
290 if (!f2fs_cf_name_slab)
291 return -ENOMEM;
292 return 0;
293 }
294
f2fs_destroy_casefold_cache(void)295 static void f2fs_destroy_casefold_cache(void)
296 {
297 kmem_cache_destroy(f2fs_cf_name_slab);
298 }
299 #else
f2fs_create_casefold_cache(void)300 static int __init f2fs_create_casefold_cache(void) { return 0; }
f2fs_destroy_casefold_cache(void)301 static void f2fs_destroy_casefold_cache(void) { }
302 #endif
303
limit_reserve_root(struct f2fs_sb_info * sbi)304 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
305 {
306 block_t limit = min((sbi->user_block_count >> 3),
307 sbi->user_block_count - sbi->reserved_blocks);
308
309 /* limit is 12.5% */
310 if (test_opt(sbi, RESERVE_ROOT) &&
311 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
312 F2FS_OPTION(sbi).root_reserved_blocks = limit;
313 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
314 F2FS_OPTION(sbi).root_reserved_blocks);
315 }
316 if (!test_opt(sbi, RESERVE_ROOT) &&
317 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
318 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
319 !gid_eq(F2FS_OPTION(sbi).s_resgid,
320 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
321 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
322 from_kuid_munged(&init_user_ns,
323 F2FS_OPTION(sbi).s_resuid),
324 from_kgid_munged(&init_user_ns,
325 F2FS_OPTION(sbi).s_resgid));
326 }
327
adjust_reserved_segment(struct f2fs_sb_info * sbi)328 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
329 {
330 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
331 unsigned int avg_vblocks;
332 unsigned int wanted_reserved_segments;
333 block_t avail_user_block_count;
334
335 if (!F2FS_IO_ALIGNED(sbi))
336 return 0;
337
338 /* average valid block count in section in worst case */
339 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
340
341 /*
342 * we need enough free space when migrating one section in worst case
343 */
344 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
345 reserved_segments(sbi);
346 wanted_reserved_segments -= reserved_segments(sbi);
347
348 avail_user_block_count = sbi->user_block_count -
349 sbi->current_reserved_blocks -
350 F2FS_OPTION(sbi).root_reserved_blocks;
351
352 if (wanted_reserved_segments * sbi->blocks_per_seg >
353 avail_user_block_count) {
354 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
355 wanted_reserved_segments,
356 avail_user_block_count >> sbi->log_blocks_per_seg);
357 return -ENOSPC;
358 }
359
360 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
361
362 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
363 wanted_reserved_segments);
364
365 return 0;
366 }
367
adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)368 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
369 {
370 if (!F2FS_OPTION(sbi).unusable_cap_perc)
371 return;
372
373 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
374 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
375 else
376 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
377 F2FS_OPTION(sbi).unusable_cap_perc;
378
379 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
380 F2FS_OPTION(sbi).unusable_cap,
381 F2FS_OPTION(sbi).unusable_cap_perc);
382 }
383
init_once(void * foo)384 static void init_once(void *foo)
385 {
386 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
387
388 inode_init_once(&fi->vfs_inode);
389 }
390
391 #ifdef CONFIG_QUOTA
392 static const char * const quotatypes[] = INITQFNAMES;
393 #define QTYPE2NAME(t) (quotatypes[t])
f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)394 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
395 substring_t *args)
396 {
397 struct f2fs_sb_info *sbi = F2FS_SB(sb);
398 char *qname;
399 int ret = -EINVAL;
400
401 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
402 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
403 return -EINVAL;
404 }
405 if (f2fs_sb_has_quota_ino(sbi)) {
406 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
407 return 0;
408 }
409
410 qname = match_strdup(args);
411 if (!qname) {
412 f2fs_err(sbi, "Not enough memory for storing quotafile name");
413 return -ENOMEM;
414 }
415 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
416 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
417 ret = 0;
418 else
419 f2fs_err(sbi, "%s quota file already specified",
420 QTYPE2NAME(qtype));
421 goto errout;
422 }
423 if (strchr(qname, '/')) {
424 f2fs_err(sbi, "quotafile must be on filesystem root");
425 goto errout;
426 }
427 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
428 set_opt(sbi, QUOTA);
429 return 0;
430 errout:
431 kfree(qname);
432 return ret;
433 }
434
f2fs_clear_qf_name(struct super_block * sb,int qtype)435 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
436 {
437 struct f2fs_sb_info *sbi = F2FS_SB(sb);
438
439 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
440 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
441 return -EINVAL;
442 }
443 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
444 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
445 return 0;
446 }
447
f2fs_check_quota_options(struct f2fs_sb_info * sbi)448 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
449 {
450 /*
451 * We do the test below only for project quotas. 'usrquota' and
452 * 'grpquota' mount options are allowed even without quota feature
453 * to support legacy quotas in quota files.
454 */
455 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
456 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
457 return -1;
458 }
459 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
460 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
461 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
462 if (test_opt(sbi, USRQUOTA) &&
463 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
464 clear_opt(sbi, USRQUOTA);
465
466 if (test_opt(sbi, GRPQUOTA) &&
467 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
468 clear_opt(sbi, GRPQUOTA);
469
470 if (test_opt(sbi, PRJQUOTA) &&
471 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
472 clear_opt(sbi, PRJQUOTA);
473
474 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
475 test_opt(sbi, PRJQUOTA)) {
476 f2fs_err(sbi, "old and new quota format mixing");
477 return -1;
478 }
479
480 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
481 f2fs_err(sbi, "journaled quota format not specified");
482 return -1;
483 }
484 }
485
486 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
487 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
488 F2FS_OPTION(sbi).s_jquota_fmt = 0;
489 }
490 return 0;
491 }
492 #endif
493
f2fs_set_test_dummy_encryption(struct super_block * sb,const char * opt,const substring_t * arg,bool is_remount)494 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
495 const char *opt,
496 const substring_t *arg,
497 bool is_remount)
498 {
499 struct f2fs_sb_info *sbi = F2FS_SB(sb);
500 #ifdef CONFIG_FS_ENCRYPTION
501 int err;
502
503 if (!f2fs_sb_has_encrypt(sbi)) {
504 f2fs_err(sbi, "Encrypt feature is off");
505 return -EINVAL;
506 }
507
508 /*
509 * This mount option is just for testing, and it's not worthwhile to
510 * implement the extra complexity (e.g. RCU protection) that would be
511 * needed to allow it to be set or changed during remount. We do allow
512 * it to be specified during remount, but only if there is no change.
513 */
514 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
515 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
516 return -EINVAL;
517 }
518 err = fscrypt_set_test_dummy_encryption(
519 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
520 if (err) {
521 if (err == -EEXIST)
522 f2fs_warn(sbi,
523 "Can't change test_dummy_encryption on remount");
524 else if (err == -EINVAL)
525 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
526 opt);
527 else
528 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
529 opt, err);
530 return -EINVAL;
531 }
532 f2fs_warn(sbi, "Test dummy encryption mode enabled");
533 #else
534 f2fs_warn(sbi, "Test dummy encryption mount option ignored");
535 #endif
536 return 0;
537 }
538
539 #ifdef CONFIG_F2FS_FS_COMPRESSION
540 #ifdef CONFIG_F2FS_FS_LZ4
f2fs_set_lz4hc_level(struct f2fs_sb_info * sbi,const char * str)541 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
542 {
543 #ifdef CONFIG_F2FS_FS_LZ4HC
544 unsigned int level;
545 #endif
546
547 if (strlen(str) == 3) {
548 F2FS_OPTION(sbi).compress_level = 0;
549 return 0;
550 }
551
552 #ifdef CONFIG_F2FS_FS_LZ4HC
553 str += 3;
554
555 if (str[0] != ':') {
556 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
557 return -EINVAL;
558 }
559 if (kstrtouint(str + 1, 10, &level))
560 return -EINVAL;
561
562 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
563 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
564 return -EINVAL;
565 }
566
567 F2FS_OPTION(sbi).compress_level = level;
568 return 0;
569 #else
570 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
571 return -EINVAL;
572 #endif
573 }
574 #endif
575
576 #ifdef CONFIG_F2FS_FS_ZSTD
f2fs_set_zstd_level(struct f2fs_sb_info * sbi,const char * str)577 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
578 {
579 unsigned int level;
580 int len = 4;
581
582 if (strlen(str) == len) {
583 F2FS_OPTION(sbi).compress_level = 0;
584 return 0;
585 }
586
587 str += len;
588
589 if (str[0] != ':') {
590 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
591 return -EINVAL;
592 }
593 if (kstrtouint(str + 1, 10, &level))
594 return -EINVAL;
595
596 if (!level || level > ZSTD_maxCLevel()) {
597 f2fs_info(sbi, "invalid zstd compress level: %d", level);
598 return -EINVAL;
599 }
600
601 F2FS_OPTION(sbi).compress_level = level;
602 return 0;
603 }
604 #endif
605 #endif
606
parse_options(struct super_block * sb,char * options,bool is_remount)607 static int parse_options(struct super_block *sb, char *options, bool is_remount)
608 {
609 struct f2fs_sb_info *sbi = F2FS_SB(sb);
610 substring_t args[MAX_OPT_ARGS];
611 #ifdef CONFIG_F2FS_FS_COMPRESSION
612 unsigned char (*ext)[F2FS_EXTENSION_LEN];
613 int ext_cnt;
614 #endif
615 char *p, *name;
616 int arg = 0;
617 kuid_t uid;
618 kgid_t gid;
619 int ret;
620
621 if (!options)
622 goto default_check;
623
624 while ((p = strsep(&options, ",")) != NULL) {
625 int token;
626
627 if (!*p)
628 continue;
629 /*
630 * Initialize args struct so we know whether arg was
631 * found; some options take optional arguments.
632 */
633 args[0].to = args[0].from = NULL;
634 token = match_token(p, f2fs_tokens, args);
635
636 switch (token) {
637 case Opt_gc_background:
638 name = match_strdup(&args[0]);
639
640 if (!name)
641 return -ENOMEM;
642 if (!strcmp(name, "on")) {
643 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
644 } else if (!strcmp(name, "off")) {
645 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
646 } else if (!strcmp(name, "sync")) {
647 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
648 } else {
649 kfree(name);
650 return -EINVAL;
651 }
652 kfree(name);
653 break;
654 case Opt_disable_roll_forward:
655 set_opt(sbi, DISABLE_ROLL_FORWARD);
656 break;
657 case Opt_norecovery:
658 /* this option mounts f2fs with ro */
659 set_opt(sbi, NORECOVERY);
660 if (!f2fs_readonly(sb))
661 return -EINVAL;
662 break;
663 case Opt_discard:
664 set_opt(sbi, DISCARD);
665 break;
666 case Opt_nodiscard:
667 if (f2fs_sb_has_blkzoned(sbi)) {
668 f2fs_warn(sbi, "discard is required for zoned block devices");
669 return -EINVAL;
670 }
671 clear_opt(sbi, DISCARD);
672 break;
673 case Opt_noheap:
674 set_opt(sbi, NOHEAP);
675 break;
676 case Opt_heap:
677 clear_opt(sbi, NOHEAP);
678 break;
679 #ifdef CONFIG_F2FS_FS_XATTR
680 case Opt_user_xattr:
681 set_opt(sbi, XATTR_USER);
682 break;
683 case Opt_nouser_xattr:
684 clear_opt(sbi, XATTR_USER);
685 break;
686 case Opt_inline_xattr:
687 set_opt(sbi, INLINE_XATTR);
688 break;
689 case Opt_noinline_xattr:
690 clear_opt(sbi, INLINE_XATTR);
691 break;
692 case Opt_inline_xattr_size:
693 if (args->from && match_int(args, &arg))
694 return -EINVAL;
695 set_opt(sbi, INLINE_XATTR_SIZE);
696 F2FS_OPTION(sbi).inline_xattr_size = arg;
697 break;
698 #else
699 case Opt_user_xattr:
700 f2fs_info(sbi, "user_xattr options not supported");
701 break;
702 case Opt_nouser_xattr:
703 f2fs_info(sbi, "nouser_xattr options not supported");
704 break;
705 case Opt_inline_xattr:
706 f2fs_info(sbi, "inline_xattr options not supported");
707 break;
708 case Opt_noinline_xattr:
709 f2fs_info(sbi, "noinline_xattr options not supported");
710 break;
711 #endif
712 #ifdef CONFIG_F2FS_FS_POSIX_ACL
713 case Opt_acl:
714 set_opt(sbi, POSIX_ACL);
715 break;
716 case Opt_noacl:
717 clear_opt(sbi, POSIX_ACL);
718 break;
719 #else
720 case Opt_acl:
721 f2fs_info(sbi, "acl options not supported");
722 break;
723 case Opt_noacl:
724 f2fs_info(sbi, "noacl options not supported");
725 break;
726 #endif
727 case Opt_active_logs:
728 if (args->from && match_int(args, &arg))
729 return -EINVAL;
730 if (arg != 2 && arg != 4 &&
731 arg != NR_CURSEG_PERSIST_TYPE)
732 return -EINVAL;
733 F2FS_OPTION(sbi).active_logs = arg;
734 break;
735 case Opt_disable_ext_identify:
736 set_opt(sbi, DISABLE_EXT_IDENTIFY);
737 break;
738 case Opt_inline_data:
739 set_opt(sbi, INLINE_DATA);
740 break;
741 case Opt_inline_dentry:
742 set_opt(sbi, INLINE_DENTRY);
743 break;
744 case Opt_noinline_dentry:
745 clear_opt(sbi, INLINE_DENTRY);
746 break;
747 case Opt_flush_merge:
748 set_opt(sbi, FLUSH_MERGE);
749 break;
750 case Opt_noflush_merge:
751 clear_opt(sbi, FLUSH_MERGE);
752 break;
753 case Opt_nobarrier:
754 set_opt(sbi, NOBARRIER);
755 break;
756 case Opt_fastboot:
757 set_opt(sbi, FASTBOOT);
758 break;
759 case Opt_extent_cache:
760 set_opt(sbi, READ_EXTENT_CACHE);
761 break;
762 case Opt_noextent_cache:
763 clear_opt(sbi, READ_EXTENT_CACHE);
764 break;
765 case Opt_noinline_data:
766 clear_opt(sbi, INLINE_DATA);
767 break;
768 case Opt_data_flush:
769 set_opt(sbi, DATA_FLUSH);
770 break;
771 case Opt_reserve_root:
772 if (args->from && match_int(args, &arg))
773 return -EINVAL;
774 if (test_opt(sbi, RESERVE_ROOT)) {
775 f2fs_info(sbi, "Preserve previous reserve_root=%u",
776 F2FS_OPTION(sbi).root_reserved_blocks);
777 } else {
778 F2FS_OPTION(sbi).root_reserved_blocks = arg;
779 set_opt(sbi, RESERVE_ROOT);
780 }
781 break;
782 case Opt_resuid:
783 if (args->from && match_int(args, &arg))
784 return -EINVAL;
785 uid = make_kuid(current_user_ns(), arg);
786 if (!uid_valid(uid)) {
787 f2fs_err(sbi, "Invalid uid value %d", arg);
788 return -EINVAL;
789 }
790 F2FS_OPTION(sbi).s_resuid = uid;
791 break;
792 case Opt_resgid:
793 if (args->from && match_int(args, &arg))
794 return -EINVAL;
795 gid = make_kgid(current_user_ns(), arg);
796 if (!gid_valid(gid)) {
797 f2fs_err(sbi, "Invalid gid value %d", arg);
798 return -EINVAL;
799 }
800 F2FS_OPTION(sbi).s_resgid = gid;
801 break;
802 case Opt_mode:
803 name = match_strdup(&args[0]);
804
805 if (!name)
806 return -ENOMEM;
807 if (!strcmp(name, "adaptive")) {
808 if (f2fs_sb_has_blkzoned(sbi)) {
809 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
810 kfree(name);
811 return -EINVAL;
812 }
813 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
814 } else if (!strcmp(name, "lfs")) {
815 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
816 } else {
817 kfree(name);
818 return -EINVAL;
819 }
820 kfree(name);
821 break;
822 case Opt_io_size_bits:
823 if (args->from && match_int(args, &arg))
824 return -EINVAL;
825 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
826 f2fs_warn(sbi, "Not support %d, larger than %d",
827 1 << arg, BIO_MAX_PAGES);
828 return -EINVAL;
829 }
830 F2FS_OPTION(sbi).write_io_size_bits = arg;
831 break;
832 #ifdef CONFIG_F2FS_FAULT_INJECTION
833 case Opt_fault_injection:
834 if (args->from && match_int(args, &arg))
835 return -EINVAL;
836 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
837 set_opt(sbi, FAULT_INJECTION);
838 break;
839
840 case Opt_fault_type:
841 if (args->from && match_int(args, &arg))
842 return -EINVAL;
843 f2fs_build_fault_attr(sbi, 0, arg);
844 set_opt(sbi, FAULT_INJECTION);
845 break;
846 #else
847 case Opt_fault_injection:
848 f2fs_info(sbi, "fault_injection options not supported");
849 break;
850
851 case Opt_fault_type:
852 f2fs_info(sbi, "fault_type options not supported");
853 break;
854 #endif
855 case Opt_lazytime:
856 sb->s_flags |= SB_LAZYTIME;
857 break;
858 case Opt_nolazytime:
859 sb->s_flags &= ~SB_LAZYTIME;
860 break;
861 #ifdef CONFIG_QUOTA
862 case Opt_quota:
863 case Opt_usrquota:
864 set_opt(sbi, USRQUOTA);
865 break;
866 case Opt_grpquota:
867 set_opt(sbi, GRPQUOTA);
868 break;
869 case Opt_prjquota:
870 set_opt(sbi, PRJQUOTA);
871 break;
872 case Opt_usrjquota:
873 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
874 if (ret)
875 return ret;
876 break;
877 case Opt_grpjquota:
878 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
879 if (ret)
880 return ret;
881 break;
882 case Opt_prjjquota:
883 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
884 if (ret)
885 return ret;
886 break;
887 case Opt_offusrjquota:
888 ret = f2fs_clear_qf_name(sb, USRQUOTA);
889 if (ret)
890 return ret;
891 break;
892 case Opt_offgrpjquota:
893 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
894 if (ret)
895 return ret;
896 break;
897 case Opt_offprjjquota:
898 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
899 if (ret)
900 return ret;
901 break;
902 case Opt_jqfmt_vfsold:
903 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
904 break;
905 case Opt_jqfmt_vfsv0:
906 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
907 break;
908 case Opt_jqfmt_vfsv1:
909 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
910 break;
911 case Opt_noquota:
912 clear_opt(sbi, QUOTA);
913 clear_opt(sbi, USRQUOTA);
914 clear_opt(sbi, GRPQUOTA);
915 clear_opt(sbi, PRJQUOTA);
916 break;
917 #else
918 case Opt_quota:
919 case Opt_usrquota:
920 case Opt_grpquota:
921 case Opt_prjquota:
922 case Opt_usrjquota:
923 case Opt_grpjquota:
924 case Opt_prjjquota:
925 case Opt_offusrjquota:
926 case Opt_offgrpjquota:
927 case Opt_offprjjquota:
928 case Opt_jqfmt_vfsold:
929 case Opt_jqfmt_vfsv0:
930 case Opt_jqfmt_vfsv1:
931 case Opt_noquota:
932 f2fs_info(sbi, "quota operations not supported");
933 break;
934 #endif
935 case Opt_whint:
936 name = match_strdup(&args[0]);
937 if (!name)
938 return -ENOMEM;
939 if (!strcmp(name, "user-based")) {
940 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
941 } else if (!strcmp(name, "off")) {
942 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
943 } else if (!strcmp(name, "fs-based")) {
944 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
945 } else {
946 kfree(name);
947 return -EINVAL;
948 }
949 kfree(name);
950 break;
951 case Opt_alloc:
952 name = match_strdup(&args[0]);
953 if (!name)
954 return -ENOMEM;
955
956 if (!strcmp(name, "default")) {
957 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
958 } else if (!strcmp(name, "reuse")) {
959 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
960 } else {
961 kfree(name);
962 return -EINVAL;
963 }
964 kfree(name);
965 break;
966 case Opt_fsync:
967 name = match_strdup(&args[0]);
968 if (!name)
969 return -ENOMEM;
970 if (!strcmp(name, "posix")) {
971 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
972 } else if (!strcmp(name, "strict")) {
973 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
974 } else if (!strcmp(name, "nobarrier")) {
975 F2FS_OPTION(sbi).fsync_mode =
976 FSYNC_MODE_NOBARRIER;
977 } else {
978 kfree(name);
979 return -EINVAL;
980 }
981 kfree(name);
982 break;
983 case Opt_test_dummy_encryption:
984 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
985 is_remount);
986 if (ret)
987 return ret;
988 break;
989 case Opt_inlinecrypt:
990 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
991 sb->s_flags |= SB_INLINECRYPT;
992 #else
993 f2fs_info(sbi, "inline encryption not supported");
994 #endif
995 break;
996 case Opt_checkpoint_disable_cap_perc:
997 if (args->from && match_int(args, &arg))
998 return -EINVAL;
999 if (arg < 0 || arg > 100)
1000 return -EINVAL;
1001 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1002 set_opt(sbi, DISABLE_CHECKPOINT);
1003 break;
1004 case Opt_checkpoint_disable_cap:
1005 if (args->from && match_int(args, &arg))
1006 return -EINVAL;
1007 F2FS_OPTION(sbi).unusable_cap = arg;
1008 set_opt(sbi, DISABLE_CHECKPOINT);
1009 break;
1010 case Opt_checkpoint_disable:
1011 set_opt(sbi, DISABLE_CHECKPOINT);
1012 break;
1013 case Opt_checkpoint_enable:
1014 clear_opt(sbi, DISABLE_CHECKPOINT);
1015 break;
1016 case Opt_checkpoint_merge:
1017 set_opt(sbi, MERGE_CHECKPOINT);
1018 break;
1019 case Opt_nocheckpoint_merge:
1020 clear_opt(sbi, MERGE_CHECKPOINT);
1021 break;
1022 #ifdef CONFIG_F2FS_FS_COMPRESSION
1023 case Opt_compress_algorithm:
1024 if (!f2fs_sb_has_compression(sbi)) {
1025 f2fs_info(sbi, "Image doesn't support compression");
1026 break;
1027 }
1028 name = match_strdup(&args[0]);
1029 if (!name)
1030 return -ENOMEM;
1031 if (!strcmp(name, "lzo")) {
1032 #ifdef CONFIG_F2FS_FS_LZO
1033 F2FS_OPTION(sbi).compress_level = 0;
1034 F2FS_OPTION(sbi).compress_algorithm =
1035 COMPRESS_LZO;
1036 #else
1037 f2fs_info(sbi, "kernel doesn't support lzo compression");
1038 #endif
1039 } else if (!strncmp(name, "lz4", 3)) {
1040 #ifdef CONFIG_F2FS_FS_LZ4
1041 ret = f2fs_set_lz4hc_level(sbi, name);
1042 if (ret) {
1043 kfree(name);
1044 return -EINVAL;
1045 }
1046 F2FS_OPTION(sbi).compress_algorithm =
1047 COMPRESS_LZ4;
1048 #else
1049 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1050 #endif
1051 } else if (!strncmp(name, "zstd", 4)) {
1052 #ifdef CONFIG_F2FS_FS_ZSTD
1053 ret = f2fs_set_zstd_level(sbi, name);
1054 if (ret) {
1055 kfree(name);
1056 return -EINVAL;
1057 }
1058 F2FS_OPTION(sbi).compress_algorithm =
1059 COMPRESS_ZSTD;
1060 #else
1061 f2fs_info(sbi, "kernel doesn't support zstd compression");
1062 #endif
1063 } else if (!strcmp(name, "lzo-rle")) {
1064 #ifdef CONFIG_F2FS_FS_LZORLE
1065 F2FS_OPTION(sbi).compress_level = 0;
1066 F2FS_OPTION(sbi).compress_algorithm =
1067 COMPRESS_LZORLE;
1068 #else
1069 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1070 #endif
1071 } else {
1072 kfree(name);
1073 return -EINVAL;
1074 }
1075 kfree(name);
1076 break;
1077 case Opt_compress_log_size:
1078 if (!f2fs_sb_has_compression(sbi)) {
1079 f2fs_info(sbi, "Image doesn't support compression");
1080 break;
1081 }
1082 if (args->from && match_int(args, &arg))
1083 return -EINVAL;
1084 if (arg < MIN_COMPRESS_LOG_SIZE ||
1085 arg > MAX_COMPRESS_LOG_SIZE) {
1086 f2fs_err(sbi,
1087 "Compress cluster log size is out of range");
1088 return -EINVAL;
1089 }
1090 F2FS_OPTION(sbi).compress_log_size = arg;
1091 break;
1092 case Opt_compress_extension:
1093 if (!f2fs_sb_has_compression(sbi)) {
1094 f2fs_info(sbi, "Image doesn't support compression");
1095 break;
1096 }
1097 name = match_strdup(&args[0]);
1098 if (!name)
1099 return -ENOMEM;
1100
1101 ext = F2FS_OPTION(sbi).extensions;
1102 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1103
1104 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1105 ext_cnt >= COMPRESS_EXT_NUM) {
1106 f2fs_err(sbi,
1107 "invalid extension length/number");
1108 kfree(name);
1109 return -EINVAL;
1110 }
1111
1112 strcpy(ext[ext_cnt], name);
1113 F2FS_OPTION(sbi).compress_ext_cnt++;
1114 kfree(name);
1115 break;
1116 case Opt_compress_chksum:
1117 F2FS_OPTION(sbi).compress_chksum = true;
1118 break;
1119 case Opt_compress_mode:
1120 name = match_strdup(&args[0]);
1121 if (!name)
1122 return -ENOMEM;
1123 if (!strcmp(name, "fs")) {
1124 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1125 } else if (!strcmp(name, "user")) {
1126 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1127 } else {
1128 kfree(name);
1129 return -EINVAL;
1130 }
1131 kfree(name);
1132 break;
1133 case Opt_compress_cache:
1134 set_opt(sbi, COMPRESS_CACHE);
1135 break;
1136 #else
1137 case Opt_compress_algorithm:
1138 case Opt_compress_log_size:
1139 case Opt_compress_extension:
1140 case Opt_compress_chksum:
1141 case Opt_compress_mode:
1142 case Opt_compress_cache:
1143 f2fs_info(sbi, "compression options not supported");
1144 break;
1145 #endif
1146 case Opt_atgc:
1147 set_opt(sbi, ATGC);
1148 break;
1149 case Opt_gc_merge:
1150 set_opt(sbi, GC_MERGE);
1151 break;
1152 case Opt_nogc_merge:
1153 clear_opt(sbi, GC_MERGE);
1154 break;
1155 case Opt_age_extent_cache:
1156 set_opt(sbi, AGE_EXTENT_CACHE);
1157 break;
1158 case Opt_memory_mode:
1159 name = match_strdup(&args[0]);
1160 if (!name)
1161 return -ENOMEM;
1162 if (!strcmp(name, "normal")) {
1163 F2FS_OPTION(sbi).memory_mode =
1164 MEMORY_MODE_NORMAL;
1165 } else if (!strcmp(name, "low")) {
1166 F2FS_OPTION(sbi).memory_mode =
1167 MEMORY_MODE_LOW;
1168 } else {
1169 kfree(name);
1170 return -EINVAL;
1171 }
1172 kfree(name);
1173 break;
1174 default:
1175 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1176 p);
1177 return -EINVAL;
1178 }
1179 }
1180 default_check:
1181 #ifdef CONFIG_QUOTA
1182 if (f2fs_check_quota_options(sbi))
1183 return -EINVAL;
1184 #else
1185 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1186 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1187 return -EINVAL;
1188 }
1189 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1190 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1191 return -EINVAL;
1192 }
1193 #endif
1194 #ifndef CONFIG_UNICODE
1195 if (f2fs_sb_has_casefold(sbi)) {
1196 f2fs_err(sbi,
1197 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1198 return -EINVAL;
1199 }
1200 #endif
1201 /*
1202 * The BLKZONED feature indicates that the drive was formatted with
1203 * zone alignment optimization. This is optional for host-aware
1204 * devices, but mandatory for host-managed zoned block devices.
1205 */
1206 #ifndef CONFIG_BLK_DEV_ZONED
1207 if (f2fs_sb_has_blkzoned(sbi)) {
1208 f2fs_err(sbi, "Zoned block device support is not enabled");
1209 return -EINVAL;
1210 }
1211 #endif
1212
1213 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1214 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1215 F2FS_IO_SIZE_KB(sbi));
1216 return -EINVAL;
1217 }
1218
1219 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1220 int min_size, max_size;
1221
1222 if (!f2fs_sb_has_extra_attr(sbi) ||
1223 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1224 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1225 return -EINVAL;
1226 }
1227 if (!test_opt(sbi, INLINE_XATTR)) {
1228 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1229 return -EINVAL;
1230 }
1231
1232 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1233 max_size = MAX_INLINE_XATTR_SIZE;
1234
1235 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1236 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1237 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1238 min_size, max_size);
1239 return -EINVAL;
1240 }
1241 }
1242
1243 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1244 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1245 return -EINVAL;
1246 }
1247
1248 /* Not pass down write hints if the number of active logs is lesser
1249 * than NR_CURSEG_PERSIST_TYPE.
1250 */
1251 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE)
1252 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1253
1254 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1255 f2fs_err(sbi, "Allow to mount readonly mode only");
1256 return -EROFS;
1257 }
1258 return 0;
1259 }
1260
f2fs_alloc_inode(struct super_block * sb)1261 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1262 {
1263 struct f2fs_inode_info *fi;
1264
1265 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
1266 if (!fi)
1267 return NULL;
1268
1269 init_once((void *) fi);
1270
1271 /* Initialize f2fs-specific inode info */
1272 atomic_set(&fi->dirty_pages, 0);
1273 atomic_set(&fi->i_compr_blocks, 0);
1274 init_f2fs_rwsem(&fi->i_sem);
1275 spin_lock_init(&fi->i_size_lock);
1276 INIT_LIST_HEAD(&fi->dirty_list);
1277 INIT_LIST_HEAD(&fi->gdirty_list);
1278 INIT_LIST_HEAD(&fi->inmem_ilist);
1279 INIT_LIST_HEAD(&fi->inmem_pages);
1280 mutex_init(&fi->inmem_lock);
1281 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1282 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1283 init_f2fs_rwsem(&fi->i_mmap_sem);
1284 init_f2fs_rwsem(&fi->i_xattr_sem);
1285
1286 /* Will be used by directory only */
1287 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1288
1289 return &fi->vfs_inode;
1290 }
1291
f2fs_drop_inode(struct inode * inode)1292 static int f2fs_drop_inode(struct inode *inode)
1293 {
1294 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1295 int ret;
1296
1297 /*
1298 * during filesystem shutdown, if checkpoint is disabled,
1299 * drop useless meta/node dirty pages.
1300 */
1301 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1302 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1303 inode->i_ino == F2FS_META_INO(sbi)) {
1304 trace_f2fs_drop_inode(inode, 1);
1305 return 1;
1306 }
1307 }
1308
1309 /*
1310 * This is to avoid a deadlock condition like below.
1311 * writeback_single_inode(inode)
1312 * - f2fs_write_data_page
1313 * - f2fs_gc -> iput -> evict
1314 * - inode_wait_for_writeback(inode)
1315 */
1316 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1317 if (!inode->i_nlink && !is_bad_inode(inode)) {
1318 /* to avoid evict_inode call simultaneously */
1319 atomic_inc(&inode->i_count);
1320 spin_unlock(&inode->i_lock);
1321
1322 /* some remained atomic pages should discarded */
1323 if (f2fs_is_atomic_file(inode))
1324 f2fs_drop_inmem_pages(inode);
1325
1326 /* should remain fi->extent_tree for writepage */
1327 f2fs_destroy_extent_node(inode);
1328
1329 sb_start_intwrite(inode->i_sb);
1330 f2fs_i_size_write(inode, 0);
1331
1332 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1333 inode, NULL, 0, DATA);
1334 truncate_inode_pages_final(inode->i_mapping);
1335
1336 if (F2FS_HAS_BLOCKS(inode))
1337 f2fs_truncate(inode);
1338
1339 sb_end_intwrite(inode->i_sb);
1340
1341 spin_lock(&inode->i_lock);
1342 atomic_dec(&inode->i_count);
1343 }
1344 trace_f2fs_drop_inode(inode, 0);
1345 return 0;
1346 }
1347 ret = generic_drop_inode(inode);
1348 if (!ret)
1349 ret = fscrypt_drop_inode(inode);
1350 trace_f2fs_drop_inode(inode, ret);
1351 return ret;
1352 }
1353
f2fs_inode_dirtied(struct inode * inode,bool sync)1354 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1355 {
1356 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1357 int ret = 0;
1358
1359 spin_lock(&sbi->inode_lock[DIRTY_META]);
1360 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1361 ret = 1;
1362 } else {
1363 set_inode_flag(inode, FI_DIRTY_INODE);
1364 stat_inc_dirty_inode(sbi, DIRTY_META);
1365 }
1366 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1367 list_add_tail(&F2FS_I(inode)->gdirty_list,
1368 &sbi->inode_list[DIRTY_META]);
1369 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1370 }
1371 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1372 return ret;
1373 }
1374
f2fs_inode_synced(struct inode * inode)1375 void f2fs_inode_synced(struct inode *inode)
1376 {
1377 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1378
1379 spin_lock(&sbi->inode_lock[DIRTY_META]);
1380 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1381 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1382 return;
1383 }
1384 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1385 list_del_init(&F2FS_I(inode)->gdirty_list);
1386 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1387 }
1388 clear_inode_flag(inode, FI_DIRTY_INODE);
1389 clear_inode_flag(inode, FI_AUTO_RECOVER);
1390 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1391 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1392 }
1393
1394 /*
1395 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1396 *
1397 * We should call set_dirty_inode to write the dirty inode through write_inode.
1398 */
f2fs_dirty_inode(struct inode * inode,int flags)1399 static void f2fs_dirty_inode(struct inode *inode, int flags)
1400 {
1401 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1402
1403 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1404 inode->i_ino == F2FS_META_INO(sbi))
1405 return;
1406
1407 if (flags == I_DIRTY_TIME)
1408 return;
1409
1410 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1411 clear_inode_flag(inode, FI_AUTO_RECOVER);
1412
1413 f2fs_inode_dirtied(inode, false);
1414 }
1415
f2fs_free_inode(struct inode * inode)1416 static void f2fs_free_inode(struct inode *inode)
1417 {
1418 fscrypt_free_inode(inode);
1419 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1420 }
1421
destroy_percpu_info(struct f2fs_sb_info * sbi)1422 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1423 {
1424 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1425 percpu_counter_destroy(&sbi->total_valid_inode_count);
1426 }
1427
destroy_device_list(struct f2fs_sb_info * sbi)1428 static void destroy_device_list(struct f2fs_sb_info *sbi)
1429 {
1430 int i;
1431
1432 for (i = 0; i < sbi->s_ndevs; i++) {
1433 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1434 #ifdef CONFIG_BLK_DEV_ZONED
1435 kvfree(FDEV(i).blkz_seq);
1436 #endif
1437 }
1438 kvfree(sbi->devs);
1439 }
1440
f2fs_put_super(struct super_block * sb)1441 static void f2fs_put_super(struct super_block *sb)
1442 {
1443 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1444 int i;
1445 bool dropped;
1446
1447 /* unregister procfs/sysfs entries in advance to avoid race case */
1448 f2fs_unregister_sysfs(sbi);
1449
1450 f2fs_quota_off_umount(sb);
1451
1452 /* prevent remaining shrinker jobs */
1453 mutex_lock(&sbi->umount_mutex);
1454
1455 /*
1456 * flush all issued checkpoints and stop checkpoint issue thread.
1457 * after then, all checkpoints should be done by each process context.
1458 */
1459 f2fs_stop_ckpt_thread(sbi);
1460
1461 /*
1462 * We don't need to do checkpoint when superblock is clean.
1463 * But, the previous checkpoint was not done by umount, it needs to do
1464 * clean checkpoint again.
1465 */
1466 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1467 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1468 struct cp_control cpc = {
1469 .reason = CP_UMOUNT,
1470 };
1471 f2fs_write_checkpoint(sbi, &cpc);
1472 }
1473
1474 /* be sure to wait for any on-going discard commands */
1475 dropped = f2fs_issue_discard_timeout(sbi);
1476
1477 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1478 !sbi->discard_blks && !dropped) {
1479 struct cp_control cpc = {
1480 .reason = CP_UMOUNT | CP_TRIMMED,
1481 };
1482 f2fs_write_checkpoint(sbi, &cpc);
1483 }
1484
1485 /*
1486 * normally superblock is clean, so we need to release this.
1487 * In addition, EIO will skip do checkpoint, we need this as well.
1488 */
1489 f2fs_release_ino_entry(sbi, true);
1490
1491 f2fs_leave_shrinker(sbi);
1492 mutex_unlock(&sbi->umount_mutex);
1493
1494 /* our cp_error case, we can wait for any writeback page */
1495 f2fs_flush_merged_writes(sbi);
1496
1497 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1498
1499 f2fs_bug_on(sbi, sbi->fsync_node_num);
1500
1501 f2fs_destroy_compress_inode(sbi);
1502
1503 iput(sbi->node_inode);
1504 sbi->node_inode = NULL;
1505
1506 iput(sbi->meta_inode);
1507 sbi->meta_inode = NULL;
1508
1509 /*
1510 * iput() can update stat information, if f2fs_write_checkpoint()
1511 * above failed with error.
1512 */
1513 f2fs_destroy_stats(sbi);
1514
1515 /* destroy f2fs internal modules */
1516 f2fs_destroy_node_manager(sbi);
1517 f2fs_destroy_segment_manager(sbi);
1518
1519 f2fs_destroy_post_read_wq(sbi);
1520
1521 kvfree(sbi->ckpt);
1522
1523 sb->s_fs_info = NULL;
1524 if (sbi->s_chksum_driver)
1525 crypto_free_shash(sbi->s_chksum_driver);
1526 kfree(sbi->raw_super);
1527
1528 destroy_device_list(sbi);
1529 f2fs_destroy_page_array_cache(sbi);
1530 f2fs_destroy_xattr_caches(sbi);
1531 mempool_destroy(sbi->write_io_dummy);
1532 #ifdef CONFIG_QUOTA
1533 for (i = 0; i < MAXQUOTAS; i++)
1534 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1535 #endif
1536 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1537 destroy_percpu_info(sbi);
1538 for (i = 0; i < NR_PAGE_TYPE; i++)
1539 kvfree(sbi->write_io[i]);
1540 #ifdef CONFIG_UNICODE
1541 utf8_unload(sb->s_encoding);
1542 #endif
1543 kfree(sbi);
1544 }
1545
f2fs_sync_fs(struct super_block * sb,int sync)1546 int f2fs_sync_fs(struct super_block *sb, int sync)
1547 {
1548 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1549 int err = 0;
1550
1551 if (unlikely(f2fs_cp_error(sbi)))
1552 return 0;
1553 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1554 return 0;
1555
1556 trace_f2fs_sync_fs(sb, sync);
1557
1558 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1559 return -EAGAIN;
1560
1561 if (sync)
1562 err = f2fs_issue_checkpoint(sbi);
1563
1564 return err;
1565 }
1566
f2fs_freeze(struct super_block * sb)1567 static int f2fs_freeze(struct super_block *sb)
1568 {
1569 if (f2fs_readonly(sb))
1570 return 0;
1571
1572 /* IO error happened before */
1573 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1574 return -EIO;
1575
1576 /* must be clean, since sync_filesystem() was already called */
1577 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1578 return -EINVAL;
1579
1580 /* Let's flush checkpoints and stop the thread. */
1581 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1582
1583 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1584 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1585 return 0;
1586 }
1587
f2fs_unfreeze(struct super_block * sb)1588 static int f2fs_unfreeze(struct super_block *sb)
1589 {
1590 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1591 return 0;
1592 }
1593
1594 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1595 static int f2fs_statfs_project(struct super_block *sb,
1596 kprojid_t projid, struct kstatfs *buf)
1597 {
1598 struct kqid qid;
1599 struct dquot *dquot;
1600 u64 limit;
1601 u64 curblock;
1602
1603 qid = make_kqid_projid(projid);
1604 dquot = dqget(sb, qid);
1605 if (IS_ERR(dquot))
1606 return PTR_ERR(dquot);
1607 spin_lock(&dquot->dq_dqb_lock);
1608
1609 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1610 dquot->dq_dqb.dqb_bhardlimit);
1611 if (limit)
1612 limit >>= sb->s_blocksize_bits;
1613
1614 if (limit && buf->f_blocks > limit) {
1615 curblock = (dquot->dq_dqb.dqb_curspace +
1616 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1617 buf->f_blocks = limit;
1618 buf->f_bfree = buf->f_bavail =
1619 (buf->f_blocks > curblock) ?
1620 (buf->f_blocks - curblock) : 0;
1621 }
1622
1623 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1624 dquot->dq_dqb.dqb_ihardlimit);
1625
1626 if (limit && buf->f_files > limit) {
1627 buf->f_files = limit;
1628 buf->f_ffree =
1629 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1630 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1631 }
1632
1633 spin_unlock(&dquot->dq_dqb_lock);
1634 dqput(dquot);
1635 return 0;
1636 }
1637 #endif
1638
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1639 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1640 {
1641 struct super_block *sb = dentry->d_sb;
1642 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1643 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1644 block_t total_count, user_block_count, start_count;
1645 u64 avail_node_count;
1646
1647 total_count = le64_to_cpu(sbi->raw_super->block_count);
1648 user_block_count = sbi->user_block_count;
1649 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1650 buf->f_type = F2FS_SUPER_MAGIC;
1651 buf->f_bsize = sbi->blocksize;
1652
1653 buf->f_blocks = total_count - start_count;
1654 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1655 sbi->current_reserved_blocks;
1656
1657 spin_lock(&sbi->stat_lock);
1658 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1659 buf->f_bfree = 0;
1660 else
1661 buf->f_bfree -= sbi->unusable_block_count;
1662 spin_unlock(&sbi->stat_lock);
1663
1664 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1665 buf->f_bavail = buf->f_bfree -
1666 F2FS_OPTION(sbi).root_reserved_blocks;
1667 else
1668 buf->f_bavail = 0;
1669
1670 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1671
1672 if (avail_node_count > user_block_count) {
1673 buf->f_files = user_block_count;
1674 buf->f_ffree = buf->f_bavail;
1675 } else {
1676 buf->f_files = avail_node_count;
1677 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1678 buf->f_bavail);
1679 }
1680
1681 buf->f_namelen = F2FS_NAME_LEN;
1682 buf->f_fsid = u64_to_fsid(id);
1683
1684 #ifdef CONFIG_QUOTA
1685 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1686 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1687 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1688 }
1689 #endif
1690 return 0;
1691 }
1692
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1693 static inline void f2fs_show_quota_options(struct seq_file *seq,
1694 struct super_block *sb)
1695 {
1696 #ifdef CONFIG_QUOTA
1697 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1698
1699 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1700 char *fmtname = "";
1701
1702 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1703 case QFMT_VFS_OLD:
1704 fmtname = "vfsold";
1705 break;
1706 case QFMT_VFS_V0:
1707 fmtname = "vfsv0";
1708 break;
1709 case QFMT_VFS_V1:
1710 fmtname = "vfsv1";
1711 break;
1712 }
1713 seq_printf(seq, ",jqfmt=%s", fmtname);
1714 }
1715
1716 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1717 seq_show_option(seq, "usrjquota",
1718 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1719
1720 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1721 seq_show_option(seq, "grpjquota",
1722 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1723
1724 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1725 seq_show_option(seq, "prjjquota",
1726 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1727 #endif
1728 }
1729
1730 #ifdef CONFIG_F2FS_FS_COMPRESSION
f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1731 static inline void f2fs_show_compress_options(struct seq_file *seq,
1732 struct super_block *sb)
1733 {
1734 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1735 char *algtype = "";
1736 int i;
1737
1738 if (!f2fs_sb_has_compression(sbi))
1739 return;
1740
1741 switch (F2FS_OPTION(sbi).compress_algorithm) {
1742 case COMPRESS_LZO:
1743 algtype = "lzo";
1744 break;
1745 case COMPRESS_LZ4:
1746 algtype = "lz4";
1747 break;
1748 case COMPRESS_ZSTD:
1749 algtype = "zstd";
1750 break;
1751 case COMPRESS_LZORLE:
1752 algtype = "lzo-rle";
1753 break;
1754 }
1755 seq_printf(seq, ",compress_algorithm=%s", algtype);
1756
1757 if (F2FS_OPTION(sbi).compress_level)
1758 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1759
1760 seq_printf(seq, ",compress_log_size=%u",
1761 F2FS_OPTION(sbi).compress_log_size);
1762
1763 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1764 seq_printf(seq, ",compress_extension=%s",
1765 F2FS_OPTION(sbi).extensions[i]);
1766 }
1767
1768 if (F2FS_OPTION(sbi).compress_chksum)
1769 seq_puts(seq, ",compress_chksum");
1770
1771 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1772 seq_printf(seq, ",compress_mode=%s", "fs");
1773 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1774 seq_printf(seq, ",compress_mode=%s", "user");
1775
1776 if (test_opt(sbi, COMPRESS_CACHE))
1777 seq_puts(seq, ",compress_cache");
1778 }
1779 #endif
1780
f2fs_show_options(struct seq_file * seq,struct dentry * root)1781 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1782 {
1783 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1784
1785 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1786 seq_printf(seq, ",background_gc=%s", "sync");
1787 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1788 seq_printf(seq, ",background_gc=%s", "on");
1789 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1790 seq_printf(seq, ",background_gc=%s", "off");
1791
1792 if (test_opt(sbi, GC_MERGE))
1793 seq_puts(seq, ",gc_merge");
1794
1795 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1796 seq_puts(seq, ",disable_roll_forward");
1797 if (test_opt(sbi, NORECOVERY))
1798 seq_puts(seq, ",norecovery");
1799 if (test_opt(sbi, DISCARD))
1800 seq_puts(seq, ",discard");
1801 else
1802 seq_puts(seq, ",nodiscard");
1803 if (test_opt(sbi, NOHEAP))
1804 seq_puts(seq, ",no_heap");
1805 else
1806 seq_puts(seq, ",heap");
1807 #ifdef CONFIG_F2FS_FS_XATTR
1808 if (test_opt(sbi, XATTR_USER))
1809 seq_puts(seq, ",user_xattr");
1810 else
1811 seq_puts(seq, ",nouser_xattr");
1812 if (test_opt(sbi, INLINE_XATTR))
1813 seq_puts(seq, ",inline_xattr");
1814 else
1815 seq_puts(seq, ",noinline_xattr");
1816 if (test_opt(sbi, INLINE_XATTR_SIZE))
1817 seq_printf(seq, ",inline_xattr_size=%u",
1818 F2FS_OPTION(sbi).inline_xattr_size);
1819 #endif
1820 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1821 if (test_opt(sbi, POSIX_ACL))
1822 seq_puts(seq, ",acl");
1823 else
1824 seq_puts(seq, ",noacl");
1825 #endif
1826 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1827 seq_puts(seq, ",disable_ext_identify");
1828 if (test_opt(sbi, INLINE_DATA))
1829 seq_puts(seq, ",inline_data");
1830 else
1831 seq_puts(seq, ",noinline_data");
1832 if (test_opt(sbi, INLINE_DENTRY))
1833 seq_puts(seq, ",inline_dentry");
1834 else
1835 seq_puts(seq, ",noinline_dentry");
1836 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1837 seq_puts(seq, ",flush_merge");
1838 if (test_opt(sbi, NOBARRIER))
1839 seq_puts(seq, ",nobarrier");
1840 if (test_opt(sbi, FASTBOOT))
1841 seq_puts(seq, ",fastboot");
1842 if (test_opt(sbi, READ_EXTENT_CACHE))
1843 seq_puts(seq, ",extent_cache");
1844 else
1845 seq_puts(seq, ",noextent_cache");
1846 if (test_opt(sbi, AGE_EXTENT_CACHE))
1847 seq_puts(seq, ",age_extent_cache");
1848 if (test_opt(sbi, DATA_FLUSH))
1849 seq_puts(seq, ",data_flush");
1850
1851 seq_puts(seq, ",mode=");
1852 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1853 seq_puts(seq, "adaptive");
1854 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1855 seq_puts(seq, "lfs");
1856 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1857 if (test_opt(sbi, RESERVE_ROOT))
1858 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1859 F2FS_OPTION(sbi).root_reserved_blocks,
1860 from_kuid_munged(&init_user_ns,
1861 F2FS_OPTION(sbi).s_resuid),
1862 from_kgid_munged(&init_user_ns,
1863 F2FS_OPTION(sbi).s_resgid));
1864 if (F2FS_IO_SIZE_BITS(sbi))
1865 seq_printf(seq, ",io_bits=%u",
1866 F2FS_OPTION(sbi).write_io_size_bits);
1867 #ifdef CONFIG_F2FS_FAULT_INJECTION
1868 if (test_opt(sbi, FAULT_INJECTION)) {
1869 seq_printf(seq, ",fault_injection=%u",
1870 F2FS_OPTION(sbi).fault_info.inject_rate);
1871 seq_printf(seq, ",fault_type=%u",
1872 F2FS_OPTION(sbi).fault_info.inject_type);
1873 }
1874 #endif
1875 #ifdef CONFIG_QUOTA
1876 if (test_opt(sbi, QUOTA))
1877 seq_puts(seq, ",quota");
1878 if (test_opt(sbi, USRQUOTA))
1879 seq_puts(seq, ",usrquota");
1880 if (test_opt(sbi, GRPQUOTA))
1881 seq_puts(seq, ",grpquota");
1882 if (test_opt(sbi, PRJQUOTA))
1883 seq_puts(seq, ",prjquota");
1884 #endif
1885 f2fs_show_quota_options(seq, sbi->sb);
1886 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1887 seq_printf(seq, ",whint_mode=%s", "user-based");
1888 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1889 seq_printf(seq, ",whint_mode=%s", "fs-based");
1890
1891 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1892
1893 if (sbi->sb->s_flags & SB_INLINECRYPT)
1894 seq_puts(seq, ",inlinecrypt");
1895
1896 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1897 seq_printf(seq, ",alloc_mode=%s", "default");
1898 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1899 seq_printf(seq, ",alloc_mode=%s", "reuse");
1900
1901 if (test_opt(sbi, DISABLE_CHECKPOINT))
1902 seq_printf(seq, ",checkpoint=disable:%u",
1903 F2FS_OPTION(sbi).unusable_cap);
1904 if (test_opt(sbi, MERGE_CHECKPOINT))
1905 seq_puts(seq, ",checkpoint_merge");
1906 else
1907 seq_puts(seq, ",nocheckpoint_merge");
1908 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1909 seq_printf(seq, ",fsync_mode=%s", "posix");
1910 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1911 seq_printf(seq, ",fsync_mode=%s", "strict");
1912 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1913 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1914
1915 #ifdef CONFIG_F2FS_FS_COMPRESSION
1916 f2fs_show_compress_options(seq, sbi->sb);
1917 #endif
1918
1919 if (test_opt(sbi, ATGC))
1920 seq_puts(seq, ",atgc");
1921
1922 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
1923 seq_printf(seq, ",memory=%s", "normal");
1924 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
1925 seq_printf(seq, ",memory=%s", "low");
1926
1927 return 0;
1928 }
1929
default_options(struct f2fs_sb_info * sbi,bool remount)1930 static void default_options(struct f2fs_sb_info *sbi, bool remount)
1931 {
1932 /* init some FS parameters */
1933 if (!remount) {
1934 set_opt(sbi, READ_EXTENT_CACHE);
1935 clear_opt(sbi, DISABLE_CHECKPOINT);
1936
1937 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
1938 set_opt(sbi, DISCARD);
1939 }
1940
1941 if (f2fs_sb_has_readonly(sbi))
1942 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
1943 else
1944 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
1945
1946 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1947 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1948 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1949 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1950 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1951 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1952 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1953 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1954 F2FS_OPTION(sbi).compress_ext_cnt = 0;
1955 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1956 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1957 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
1958
1959 sbi->sb->s_flags &= ~SB_INLINECRYPT;
1960
1961 set_opt(sbi, INLINE_XATTR);
1962 set_opt(sbi, INLINE_DATA);
1963 set_opt(sbi, INLINE_DENTRY);
1964 set_opt(sbi, NOHEAP);
1965 set_opt(sbi, MERGE_CHECKPOINT);
1966 F2FS_OPTION(sbi).unusable_cap = 0;
1967 sbi->sb->s_flags |= SB_LAZYTIME;
1968 set_opt(sbi, FLUSH_MERGE);
1969 if (f2fs_sb_has_blkzoned(sbi))
1970 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1971 else
1972 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1973
1974 #ifdef CONFIG_F2FS_FS_XATTR
1975 set_opt(sbi, XATTR_USER);
1976 #endif
1977 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1978 set_opt(sbi, POSIX_ACL);
1979 #endif
1980
1981 f2fs_build_fault_attr(sbi, 0, 0);
1982 }
1983
1984 #ifdef CONFIG_QUOTA
1985 static int f2fs_enable_quotas(struct super_block *sb);
1986 #endif
1987
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)1988 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1989 {
1990 unsigned int s_flags = sbi->sb->s_flags;
1991 struct cp_control cpc;
1992 int err = 0;
1993 int ret;
1994 block_t unusable;
1995
1996 if (s_flags & SB_RDONLY) {
1997 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1998 return -EINVAL;
1999 }
2000 sbi->sb->s_flags |= SB_ACTIVE;
2001
2002 /* check if we need more GC first */
2003 unusable = f2fs_get_unusable_blocks(sbi);
2004 if (!f2fs_disable_cp_again(sbi, unusable))
2005 goto skip_gc;
2006
2007 f2fs_update_time(sbi, DISABLE_TIME);
2008
2009 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2010 f2fs_down_write(&sbi->gc_lock);
2011 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
2012 if (err == -ENODATA) {
2013 err = 0;
2014 break;
2015 }
2016 if (err && err != -EAGAIN)
2017 break;
2018 }
2019
2020 ret = sync_filesystem(sbi->sb);
2021 if (ret || err) {
2022 err = ret ? ret : err;
2023 goto restore_flag;
2024 }
2025
2026 unusable = f2fs_get_unusable_blocks(sbi);
2027 if (f2fs_disable_cp_again(sbi, unusable)) {
2028 err = -EAGAIN;
2029 goto restore_flag;
2030 }
2031
2032 skip_gc:
2033 f2fs_down_write(&sbi->gc_lock);
2034 cpc.reason = CP_PAUSE;
2035 set_sbi_flag(sbi, SBI_CP_DISABLED);
2036 err = f2fs_write_checkpoint(sbi, &cpc);
2037 if (err)
2038 goto out_unlock;
2039
2040 spin_lock(&sbi->stat_lock);
2041 sbi->unusable_block_count = unusable;
2042 spin_unlock(&sbi->stat_lock);
2043
2044 out_unlock:
2045 f2fs_up_write(&sbi->gc_lock);
2046 restore_flag:
2047 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2048 return err;
2049 }
2050
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2051 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2052 {
2053 int retry = DEFAULT_RETRY_IO_COUNT;
2054
2055 /* we should flush all the data to keep data consistency */
2056 do {
2057 sync_inodes_sb(sbi->sb);
2058 cond_resched();
2059 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
2060 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2061
2062 if (unlikely(retry < 0))
2063 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2064
2065 f2fs_down_write(&sbi->gc_lock);
2066 f2fs_dirty_to_prefree(sbi);
2067
2068 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2069 set_sbi_flag(sbi, SBI_IS_DIRTY);
2070 f2fs_up_write(&sbi->gc_lock);
2071
2072 f2fs_sync_fs(sbi->sb, 1);
2073
2074 /* Let's ensure there's no pending checkpoint anymore */
2075 f2fs_flush_ckpt_thread(sbi);
2076 }
2077
f2fs_remount(struct super_block * sb,int * flags,char * data)2078 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2079 {
2080 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2081 struct f2fs_mount_info org_mount_opt;
2082 unsigned long old_sb_flags;
2083 int err;
2084 bool need_restart_gc = false, need_stop_gc = false;
2085 bool need_restart_ckpt = false, need_stop_ckpt = false;
2086 bool need_restart_flush = false, need_stop_flush = false;
2087 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2088 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2089 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
2090 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2091 bool no_atgc = !test_opt(sbi, ATGC);
2092 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2093 bool checkpoint_changed;
2094 #ifdef CONFIG_QUOTA
2095 int i, j;
2096 #endif
2097
2098 /*
2099 * Save the old mount options in case we
2100 * need to restore them.
2101 */
2102 org_mount_opt = sbi->mount_opt;
2103 old_sb_flags = sb->s_flags;
2104
2105 #ifdef CONFIG_QUOTA
2106 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2107 for (i = 0; i < MAXQUOTAS; i++) {
2108 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2109 org_mount_opt.s_qf_names[i] =
2110 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2111 GFP_KERNEL);
2112 if (!org_mount_opt.s_qf_names[i]) {
2113 for (j = 0; j < i; j++)
2114 kfree(org_mount_opt.s_qf_names[j]);
2115 return -ENOMEM;
2116 }
2117 } else {
2118 org_mount_opt.s_qf_names[i] = NULL;
2119 }
2120 }
2121 #endif
2122
2123 /* recover superblocks we couldn't write due to previous RO mount */
2124 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2125 err = f2fs_commit_super(sbi, false);
2126 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2127 err);
2128 if (!err)
2129 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2130 }
2131
2132 default_options(sbi, true);
2133
2134 /* parse mount options */
2135 err = parse_options(sb, data, true);
2136 if (err)
2137 goto restore_opts;
2138 checkpoint_changed =
2139 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
2140
2141 /*
2142 * Previous and new state of filesystem is RO,
2143 * so skip checking GC and FLUSH_MERGE conditions.
2144 */
2145 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2146 goto skip;
2147
2148 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2149 err = -EROFS;
2150 goto restore_opts;
2151 }
2152
2153 #ifdef CONFIG_QUOTA
2154 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2155 err = dquot_suspend(sb, -1);
2156 if (err < 0)
2157 goto restore_opts;
2158 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2159 /* dquot_resume needs RW */
2160 sb->s_flags &= ~SB_RDONLY;
2161 if (sb_any_quota_suspended(sb)) {
2162 dquot_resume(sb, -1);
2163 } else if (f2fs_sb_has_quota_ino(sbi)) {
2164 err = f2fs_enable_quotas(sb);
2165 if (err)
2166 goto restore_opts;
2167 }
2168 }
2169 #endif
2170 /* disallow enable atgc dynamically */
2171 if (no_atgc == !!test_opt(sbi, ATGC)) {
2172 err = -EINVAL;
2173 f2fs_warn(sbi, "switch atgc option is not allowed");
2174 goto restore_opts;
2175 }
2176
2177 /* disallow enable/disable extent_cache dynamically */
2178 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2179 err = -EINVAL;
2180 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2181 goto restore_opts;
2182 }
2183 /* disallow enable/disable age extent_cache dynamically */
2184 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2185 err = -EINVAL;
2186 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2187 goto restore_opts;
2188 }
2189
2190 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2191 err = -EINVAL;
2192 f2fs_warn(sbi, "switch io_bits option is not allowed");
2193 goto restore_opts;
2194 }
2195
2196 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2197 err = -EINVAL;
2198 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2199 goto restore_opts;
2200 }
2201
2202 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2203 err = -EINVAL;
2204 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2205 goto restore_opts;
2206 }
2207
2208 /*
2209 * We stop the GC thread if FS is mounted as RO
2210 * or if background_gc = off is passed in mount
2211 * option. Also sync the filesystem.
2212 */
2213 if ((*flags & SB_RDONLY) ||
2214 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2215 !test_opt(sbi, GC_MERGE))) {
2216 if (sbi->gc_thread) {
2217 f2fs_stop_gc_thread(sbi);
2218 need_restart_gc = true;
2219 }
2220 } else if (!sbi->gc_thread) {
2221 err = f2fs_start_gc_thread(sbi);
2222 if (err)
2223 goto restore_opts;
2224 need_stop_gc = true;
2225 }
2226
2227 if (*flags & SB_RDONLY ||
2228 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
2229 sync_inodes_sb(sb);
2230
2231 set_sbi_flag(sbi, SBI_IS_DIRTY);
2232 set_sbi_flag(sbi, SBI_IS_CLOSE);
2233 f2fs_sync_fs(sb, 1);
2234 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2235 }
2236
2237 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2238 !test_opt(sbi, MERGE_CHECKPOINT)) {
2239 f2fs_stop_ckpt_thread(sbi);
2240 need_restart_ckpt = true;
2241 } else {
2242 /* Flush if the prevous checkpoint, if exists. */
2243 f2fs_flush_ckpt_thread(sbi);
2244
2245 err = f2fs_start_ckpt_thread(sbi);
2246 if (err) {
2247 f2fs_err(sbi,
2248 "Failed to start F2FS issue_checkpoint_thread (%d)",
2249 err);
2250 goto restore_gc;
2251 }
2252 need_stop_ckpt = true;
2253 }
2254
2255 /*
2256 * We stop issue flush thread if FS is mounted as RO
2257 * or if flush_merge is not passed in mount option.
2258 */
2259 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2260 clear_opt(sbi, FLUSH_MERGE);
2261 f2fs_destroy_flush_cmd_control(sbi, false);
2262 need_restart_flush = true;
2263 } else {
2264 err = f2fs_create_flush_cmd_control(sbi);
2265 if (err)
2266 goto restore_ckpt;
2267 need_stop_flush = true;
2268 }
2269
2270 if (checkpoint_changed) {
2271 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2272 err = f2fs_disable_checkpoint(sbi);
2273 if (err)
2274 goto restore_flush;
2275 } else {
2276 f2fs_enable_checkpoint(sbi);
2277 }
2278 }
2279
2280 skip:
2281 #ifdef CONFIG_QUOTA
2282 /* Release old quota file names */
2283 for (i = 0; i < MAXQUOTAS; i++)
2284 kfree(org_mount_opt.s_qf_names[i]);
2285 #endif
2286 /* Update the POSIXACL Flag */
2287 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2288 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2289
2290 limit_reserve_root(sbi);
2291 adjust_unusable_cap_perc(sbi);
2292 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2293 return 0;
2294 restore_flush:
2295 if (need_restart_flush) {
2296 if (f2fs_create_flush_cmd_control(sbi))
2297 f2fs_warn(sbi, "background flush thread has stopped");
2298 } else if (need_stop_flush) {
2299 clear_opt(sbi, FLUSH_MERGE);
2300 f2fs_destroy_flush_cmd_control(sbi, false);
2301 }
2302 restore_ckpt:
2303 if (need_restart_ckpt) {
2304 if (f2fs_start_ckpt_thread(sbi))
2305 f2fs_warn(sbi, "background ckpt thread has stopped");
2306 } else if (need_stop_ckpt) {
2307 f2fs_stop_ckpt_thread(sbi);
2308 }
2309 restore_gc:
2310 if (need_restart_gc) {
2311 if (f2fs_start_gc_thread(sbi))
2312 f2fs_warn(sbi, "background gc thread has stopped");
2313 } else if (need_stop_gc) {
2314 f2fs_stop_gc_thread(sbi);
2315 }
2316 restore_opts:
2317 #ifdef CONFIG_QUOTA
2318 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2319 for (i = 0; i < MAXQUOTAS; i++) {
2320 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2321 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2322 }
2323 #endif
2324 sbi->mount_opt = org_mount_opt;
2325 sb->s_flags = old_sb_flags;
2326 return err;
2327 }
2328
2329 #ifdef CONFIG_QUOTA
2330 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2331 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2332 size_t len, loff_t off)
2333 {
2334 struct inode *inode = sb_dqopt(sb)->files[type];
2335 struct address_space *mapping = inode->i_mapping;
2336 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2337 int offset = off & (sb->s_blocksize - 1);
2338 int tocopy;
2339 size_t toread;
2340 loff_t i_size = i_size_read(inode);
2341 struct page *page;
2342
2343 if (off > i_size)
2344 return 0;
2345
2346 if (off + len > i_size)
2347 len = i_size - off;
2348 toread = len;
2349 while (toread > 0) {
2350 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2351 repeat:
2352 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2353 if (IS_ERR(page)) {
2354 if (PTR_ERR(page) == -ENOMEM) {
2355 congestion_wait(BLK_RW_ASYNC,
2356 DEFAULT_IO_TIMEOUT);
2357 goto repeat;
2358 }
2359 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2360 return PTR_ERR(page);
2361 }
2362
2363 lock_page(page);
2364
2365 if (unlikely(page->mapping != mapping)) {
2366 f2fs_put_page(page, 1);
2367 goto repeat;
2368 }
2369 if (unlikely(!PageUptodate(page))) {
2370 f2fs_put_page(page, 1);
2371 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2372 return -EIO;
2373 }
2374
2375 memcpy_from_page(data, page, offset, tocopy);
2376 f2fs_put_page(page, 1);
2377
2378 offset = 0;
2379 toread -= tocopy;
2380 data += tocopy;
2381 blkidx++;
2382 }
2383 return len;
2384 }
2385
2386 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2387 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2388 const char *data, size_t len, loff_t off)
2389 {
2390 struct inode *inode = sb_dqopt(sb)->files[type];
2391 struct address_space *mapping = inode->i_mapping;
2392 const struct address_space_operations *a_ops = mapping->a_ops;
2393 int offset = off & (sb->s_blocksize - 1);
2394 size_t towrite = len;
2395 struct page *page;
2396 void *fsdata = NULL;
2397 int err = 0;
2398 int tocopy;
2399
2400 while (towrite > 0) {
2401 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2402 towrite);
2403 retry:
2404 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2405 &page, &fsdata);
2406 if (unlikely(err)) {
2407 if (err == -ENOMEM) {
2408 congestion_wait(BLK_RW_ASYNC,
2409 DEFAULT_IO_TIMEOUT);
2410 goto retry;
2411 }
2412 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2413 break;
2414 }
2415
2416 memcpy_to_page(page, offset, data, tocopy);
2417
2418 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2419 page, fsdata);
2420 offset = 0;
2421 towrite -= tocopy;
2422 off += tocopy;
2423 data += tocopy;
2424 cond_resched();
2425 }
2426
2427 if (len == towrite)
2428 return err;
2429 inode->i_mtime = inode->i_ctime = current_time(inode);
2430 f2fs_mark_inode_dirty_sync(inode, false);
2431 return len - towrite;
2432 }
2433
f2fs_get_dquots(struct inode * inode)2434 static struct dquot **f2fs_get_dquots(struct inode *inode)
2435 {
2436 return F2FS_I(inode)->i_dquot;
2437 }
2438
f2fs_get_reserved_space(struct inode * inode)2439 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2440 {
2441 return &F2FS_I(inode)->i_reserved_quota;
2442 }
2443
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2444 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2445 {
2446 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2447 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2448 return 0;
2449 }
2450
2451 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2452 F2FS_OPTION(sbi).s_jquota_fmt, type);
2453 }
2454
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2455 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2456 {
2457 int enabled = 0;
2458 int i, err;
2459
2460 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2461 err = f2fs_enable_quotas(sbi->sb);
2462 if (err) {
2463 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2464 return 0;
2465 }
2466 return 1;
2467 }
2468
2469 for (i = 0; i < MAXQUOTAS; i++) {
2470 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2471 err = f2fs_quota_on_mount(sbi, i);
2472 if (!err) {
2473 enabled = 1;
2474 continue;
2475 }
2476 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2477 err, i);
2478 }
2479 }
2480 return enabled;
2481 }
2482
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2483 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2484 unsigned int flags)
2485 {
2486 struct inode *qf_inode;
2487 unsigned long qf_inum;
2488 int err;
2489
2490 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2491
2492 qf_inum = f2fs_qf_ino(sb, type);
2493 if (!qf_inum)
2494 return -EPERM;
2495
2496 qf_inode = f2fs_iget(sb, qf_inum);
2497 if (IS_ERR(qf_inode)) {
2498 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2499 return PTR_ERR(qf_inode);
2500 }
2501
2502 /* Don't account quota for quota files to avoid recursion */
2503 qf_inode->i_flags |= S_NOQUOTA;
2504 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2505 iput(qf_inode);
2506 return err;
2507 }
2508
f2fs_enable_quotas(struct super_block * sb)2509 static int f2fs_enable_quotas(struct super_block *sb)
2510 {
2511 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2512 int type, err = 0;
2513 unsigned long qf_inum;
2514 bool quota_mopt[MAXQUOTAS] = {
2515 test_opt(sbi, USRQUOTA),
2516 test_opt(sbi, GRPQUOTA),
2517 test_opt(sbi, PRJQUOTA),
2518 };
2519
2520 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2521 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2522 return 0;
2523 }
2524
2525 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2526
2527 for (type = 0; type < MAXQUOTAS; type++) {
2528 qf_inum = f2fs_qf_ino(sb, type);
2529 if (qf_inum) {
2530 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2531 DQUOT_USAGE_ENABLED |
2532 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2533 if (err) {
2534 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2535 type, err);
2536 for (type--; type >= 0; type--)
2537 dquot_quota_off(sb, type);
2538 set_sbi_flag(F2FS_SB(sb),
2539 SBI_QUOTA_NEED_REPAIR);
2540 return err;
2541 }
2542 }
2543 }
2544 return 0;
2545 }
2546
f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2547 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2548 {
2549 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2550 struct address_space *mapping = dqopt->files[type]->i_mapping;
2551 int ret = 0;
2552
2553 ret = dquot_writeback_dquots(sbi->sb, type);
2554 if (ret)
2555 goto out;
2556
2557 ret = filemap_fdatawrite(mapping);
2558 if (ret)
2559 goto out;
2560
2561 /* if we are using journalled quota */
2562 if (is_journalled_quota(sbi))
2563 goto out;
2564
2565 ret = filemap_fdatawait(mapping);
2566
2567 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2568 out:
2569 if (ret)
2570 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2571 return ret;
2572 }
2573
f2fs_quota_sync(struct super_block * sb,int type)2574 int f2fs_quota_sync(struct super_block *sb, int type)
2575 {
2576 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2577 struct quota_info *dqopt = sb_dqopt(sb);
2578 int cnt;
2579 int ret = 0;
2580
2581 /*
2582 * Now when everything is written we can discard the pagecache so
2583 * that userspace sees the changes.
2584 */
2585 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2586
2587 if (type != -1 && cnt != type)
2588 continue;
2589
2590 if (!sb_has_quota_active(sb, cnt))
2591 continue;
2592
2593 if (!f2fs_sb_has_quota_ino(sbi))
2594 inode_lock(dqopt->files[cnt]);
2595
2596 /*
2597 * do_quotactl
2598 * f2fs_quota_sync
2599 * f2fs_down_read(quota_sem)
2600 * dquot_writeback_dquots()
2601 * f2fs_dquot_commit
2602 * block_operation
2603 * f2fs_down_read(quota_sem)
2604 */
2605 f2fs_lock_op(sbi);
2606 f2fs_down_read(&sbi->quota_sem);
2607
2608 ret = f2fs_quota_sync_file(sbi, cnt);
2609
2610 f2fs_up_read(&sbi->quota_sem);
2611 f2fs_unlock_op(sbi);
2612
2613 if (!f2fs_sb_has_quota_ino(sbi))
2614 inode_unlock(dqopt->files[cnt]);
2615
2616 if (ret)
2617 break;
2618 }
2619 return ret;
2620 }
2621
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2622 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2623 const struct path *path)
2624 {
2625 struct inode *inode;
2626 int err;
2627
2628 /* if quota sysfile exists, deny enabling quota with specific file */
2629 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2630 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2631 return -EBUSY;
2632 }
2633
2634 err = f2fs_quota_sync(sb, type);
2635 if (err)
2636 return err;
2637
2638 err = dquot_quota_on(sb, type, format_id, path);
2639 if (err)
2640 return err;
2641
2642 inode = d_inode(path->dentry);
2643
2644 inode_lock(inode);
2645 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2646 f2fs_set_inode_flags(inode);
2647 inode_unlock(inode);
2648 f2fs_mark_inode_dirty_sync(inode, false);
2649
2650 return 0;
2651 }
2652
__f2fs_quota_off(struct super_block * sb,int type)2653 static int __f2fs_quota_off(struct super_block *sb, int type)
2654 {
2655 struct inode *inode = sb_dqopt(sb)->files[type];
2656 int err;
2657
2658 if (!inode || !igrab(inode))
2659 return dquot_quota_off(sb, type);
2660
2661 err = f2fs_quota_sync(sb, type);
2662 if (err)
2663 goto out_put;
2664
2665 err = dquot_quota_off(sb, type);
2666 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2667 goto out_put;
2668
2669 inode_lock(inode);
2670 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2671 f2fs_set_inode_flags(inode);
2672 inode_unlock(inode);
2673 f2fs_mark_inode_dirty_sync(inode, false);
2674 out_put:
2675 iput(inode);
2676 return err;
2677 }
2678
f2fs_quota_off(struct super_block * sb,int type)2679 static int f2fs_quota_off(struct super_block *sb, int type)
2680 {
2681 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2682 int err;
2683
2684 err = __f2fs_quota_off(sb, type);
2685
2686 /*
2687 * quotactl can shutdown journalled quota, result in inconsistence
2688 * between quota record and fs data by following updates, tag the
2689 * flag to let fsck be aware of it.
2690 */
2691 if (is_journalled_quota(sbi))
2692 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2693 return err;
2694 }
2695
f2fs_quota_off_umount(struct super_block * sb)2696 void f2fs_quota_off_umount(struct super_block *sb)
2697 {
2698 int type;
2699 int err;
2700
2701 for (type = 0; type < MAXQUOTAS; type++) {
2702 err = __f2fs_quota_off(sb, type);
2703 if (err) {
2704 int ret = dquot_quota_off(sb, type);
2705
2706 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2707 type, err, ret);
2708 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2709 }
2710 }
2711 /*
2712 * In case of checkpoint=disable, we must flush quota blocks.
2713 * This can cause NULL exception for node_inode in end_io, since
2714 * put_super already dropped it.
2715 */
2716 sync_filesystem(sb);
2717 }
2718
f2fs_truncate_quota_inode_pages(struct super_block * sb)2719 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2720 {
2721 struct quota_info *dqopt = sb_dqopt(sb);
2722 int type;
2723
2724 for (type = 0; type < MAXQUOTAS; type++) {
2725 if (!dqopt->files[type])
2726 continue;
2727 f2fs_inode_synced(dqopt->files[type]);
2728 }
2729 }
2730
f2fs_dquot_commit(struct dquot * dquot)2731 static int f2fs_dquot_commit(struct dquot *dquot)
2732 {
2733 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2734 int ret;
2735
2736 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2737 ret = dquot_commit(dquot);
2738 if (ret < 0)
2739 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2740 f2fs_up_read(&sbi->quota_sem);
2741 return ret;
2742 }
2743
f2fs_dquot_acquire(struct dquot * dquot)2744 static int f2fs_dquot_acquire(struct dquot *dquot)
2745 {
2746 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2747 int ret;
2748
2749 f2fs_down_read(&sbi->quota_sem);
2750 ret = dquot_acquire(dquot);
2751 if (ret < 0)
2752 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2753 f2fs_up_read(&sbi->quota_sem);
2754 return ret;
2755 }
2756
f2fs_dquot_release(struct dquot * dquot)2757 static int f2fs_dquot_release(struct dquot *dquot)
2758 {
2759 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2760 int ret = dquot_release(dquot);
2761
2762 if (ret < 0)
2763 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2764 return ret;
2765 }
2766
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)2767 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2768 {
2769 struct super_block *sb = dquot->dq_sb;
2770 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2771 int ret = dquot_mark_dquot_dirty(dquot);
2772
2773 /* if we are using journalled quota */
2774 if (is_journalled_quota(sbi))
2775 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2776
2777 return ret;
2778 }
2779
f2fs_dquot_commit_info(struct super_block * sb,int type)2780 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2781 {
2782 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2783 int ret = dquot_commit_info(sb, type);
2784
2785 if (ret < 0)
2786 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2787 return ret;
2788 }
2789
f2fs_get_projid(struct inode * inode,kprojid_t * projid)2790 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2791 {
2792 *projid = F2FS_I(inode)->i_projid;
2793 return 0;
2794 }
2795
2796 static const struct dquot_operations f2fs_quota_operations = {
2797 .get_reserved_space = f2fs_get_reserved_space,
2798 .write_dquot = f2fs_dquot_commit,
2799 .acquire_dquot = f2fs_dquot_acquire,
2800 .release_dquot = f2fs_dquot_release,
2801 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2802 .write_info = f2fs_dquot_commit_info,
2803 .alloc_dquot = dquot_alloc,
2804 .destroy_dquot = dquot_destroy,
2805 .get_projid = f2fs_get_projid,
2806 .get_next_id = dquot_get_next_id,
2807 };
2808
2809 static const struct quotactl_ops f2fs_quotactl_ops = {
2810 .quota_on = f2fs_quota_on,
2811 .quota_off = f2fs_quota_off,
2812 .quota_sync = f2fs_quota_sync,
2813 .get_state = dquot_get_state,
2814 .set_info = dquot_set_dqinfo,
2815 .get_dqblk = dquot_get_dqblk,
2816 .set_dqblk = dquot_set_dqblk,
2817 .get_nextdqblk = dquot_get_next_dqblk,
2818 };
2819 #else
f2fs_quota_sync(struct super_block * sb,int type)2820 int f2fs_quota_sync(struct super_block *sb, int type)
2821 {
2822 return 0;
2823 }
2824
f2fs_quota_off_umount(struct super_block * sb)2825 void f2fs_quota_off_umount(struct super_block *sb)
2826 {
2827 }
2828 #endif
2829
2830 static const struct super_operations f2fs_sops = {
2831 .alloc_inode = f2fs_alloc_inode,
2832 .free_inode = f2fs_free_inode,
2833 .drop_inode = f2fs_drop_inode,
2834 .write_inode = f2fs_write_inode,
2835 .dirty_inode = f2fs_dirty_inode,
2836 .show_options = f2fs_show_options,
2837 #ifdef CONFIG_QUOTA
2838 .quota_read = f2fs_quota_read,
2839 .quota_write = f2fs_quota_write,
2840 .get_dquots = f2fs_get_dquots,
2841 #endif
2842 .evict_inode = f2fs_evict_inode,
2843 .put_super = f2fs_put_super,
2844 .sync_fs = f2fs_sync_fs,
2845 .freeze_fs = f2fs_freeze,
2846 .unfreeze_fs = f2fs_unfreeze,
2847 .statfs = f2fs_statfs,
2848 .remount_fs = f2fs_remount,
2849 };
2850
2851 #ifdef CONFIG_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)2852 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2853 {
2854 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2855 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2856 ctx, len, NULL);
2857 }
2858
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)2859 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2860 void *fs_data)
2861 {
2862 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2863
2864 /*
2865 * Encrypting the root directory is not allowed because fsck
2866 * expects lost+found directory to exist and remain unencrypted
2867 * if LOST_FOUND feature is enabled.
2868 *
2869 */
2870 if (f2fs_sb_has_lost_found(sbi) &&
2871 inode->i_ino == F2FS_ROOT_INO(sbi))
2872 return -EPERM;
2873
2874 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2875 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2876 ctx, len, fs_data, XATTR_CREATE);
2877 }
2878
f2fs_get_dummy_policy(struct super_block * sb)2879 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
2880 {
2881 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
2882 }
2883
f2fs_has_stable_inodes(struct super_block * sb)2884 static bool f2fs_has_stable_inodes(struct super_block *sb)
2885 {
2886 return true;
2887 }
2888
f2fs_get_ino_and_lblk_bits(struct super_block * sb,int * ino_bits_ret,int * lblk_bits_ret)2889 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2890 int *ino_bits_ret, int *lblk_bits_ret)
2891 {
2892 *ino_bits_ret = 8 * sizeof(nid_t);
2893 *lblk_bits_ret = 8 * sizeof(block_t);
2894 }
2895
f2fs_get_num_devices(struct super_block * sb)2896 static int f2fs_get_num_devices(struct super_block *sb)
2897 {
2898 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2899
2900 if (f2fs_is_multi_device(sbi))
2901 return sbi->s_ndevs;
2902 return 1;
2903 }
2904
f2fs_get_devices(struct super_block * sb,struct request_queue ** devs)2905 static void f2fs_get_devices(struct super_block *sb,
2906 struct request_queue **devs)
2907 {
2908 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2909 int i;
2910
2911 for (i = 0; i < sbi->s_ndevs; i++)
2912 devs[i] = bdev_get_queue(FDEV(i).bdev);
2913 }
2914
2915 static const struct fscrypt_operations f2fs_cryptops = {
2916 .key_prefix = "f2fs:",
2917 .get_context = f2fs_get_context,
2918 .set_context = f2fs_set_context,
2919 .get_dummy_policy = f2fs_get_dummy_policy,
2920 .empty_dir = f2fs_empty_dir,
2921 .max_namelen = F2FS_NAME_LEN,
2922 .has_stable_inodes = f2fs_has_stable_inodes,
2923 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2924 .get_num_devices = f2fs_get_num_devices,
2925 .get_devices = f2fs_get_devices,
2926 };
2927 #endif
2928
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)2929 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2930 u64 ino, u32 generation)
2931 {
2932 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2933 struct inode *inode;
2934
2935 if (f2fs_check_nid_range(sbi, ino))
2936 return ERR_PTR(-ESTALE);
2937
2938 /*
2939 * f2fs_iget isn't quite right if the inode is currently unallocated!
2940 * However f2fs_iget currently does appropriate checks to handle stale
2941 * inodes so everything is OK.
2942 */
2943 inode = f2fs_iget(sb, ino);
2944 if (IS_ERR(inode))
2945 return ERR_CAST(inode);
2946 if (unlikely(generation && inode->i_generation != generation)) {
2947 /* we didn't find the right inode.. */
2948 iput(inode);
2949 return ERR_PTR(-ESTALE);
2950 }
2951 return inode;
2952 }
2953
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)2954 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2955 int fh_len, int fh_type)
2956 {
2957 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2958 f2fs_nfs_get_inode);
2959 }
2960
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)2961 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2962 int fh_len, int fh_type)
2963 {
2964 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2965 f2fs_nfs_get_inode);
2966 }
2967
2968 static const struct export_operations f2fs_export_ops = {
2969 .fh_to_dentry = f2fs_fh_to_dentry,
2970 .fh_to_parent = f2fs_fh_to_parent,
2971 .get_parent = f2fs_get_parent,
2972 };
2973
max_file_blocks(struct inode * inode)2974 loff_t max_file_blocks(struct inode *inode)
2975 {
2976 loff_t result = 0;
2977 loff_t leaf_count;
2978
2979 /*
2980 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2981 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2982 * space in inode.i_addr, it will be more safe to reassign
2983 * result as zero.
2984 */
2985
2986 if (inode && f2fs_compressed_file(inode))
2987 leaf_count = ADDRS_PER_BLOCK(inode);
2988 else
2989 leaf_count = DEF_ADDRS_PER_BLOCK;
2990
2991 /* two direct node blocks */
2992 result += (leaf_count * 2);
2993
2994 /* two indirect node blocks */
2995 leaf_count *= NIDS_PER_BLOCK;
2996 result += (leaf_count * 2);
2997
2998 /* one double indirect node block */
2999 leaf_count *= NIDS_PER_BLOCK;
3000 result += leaf_count;
3001
3002 return result;
3003 }
3004
__f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)3005 static int __f2fs_commit_super(struct buffer_head *bh,
3006 struct f2fs_super_block *super)
3007 {
3008 lock_buffer(bh);
3009 if (super)
3010 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3011 set_buffer_dirty(bh);
3012 unlock_buffer(bh);
3013
3014 /* it's rare case, we can do fua all the time */
3015 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3016 }
3017
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)3018 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3019 struct buffer_head *bh)
3020 {
3021 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3022 (bh->b_data + F2FS_SUPER_OFFSET);
3023 struct super_block *sb = sbi->sb;
3024 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3025 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3026 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3027 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3028 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3029 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3030 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3031 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3032 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3033 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3034 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3035 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3036 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3037 u64 main_end_blkaddr = main_blkaddr +
3038 (segment_count_main << log_blocks_per_seg);
3039 u64 seg_end_blkaddr = segment0_blkaddr +
3040 (segment_count << log_blocks_per_seg);
3041
3042 if (segment0_blkaddr != cp_blkaddr) {
3043 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3044 segment0_blkaddr, cp_blkaddr);
3045 return true;
3046 }
3047
3048 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3049 sit_blkaddr) {
3050 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3051 cp_blkaddr, sit_blkaddr,
3052 segment_count_ckpt << log_blocks_per_seg);
3053 return true;
3054 }
3055
3056 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3057 nat_blkaddr) {
3058 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3059 sit_blkaddr, nat_blkaddr,
3060 segment_count_sit << log_blocks_per_seg);
3061 return true;
3062 }
3063
3064 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3065 ssa_blkaddr) {
3066 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3067 nat_blkaddr, ssa_blkaddr,
3068 segment_count_nat << log_blocks_per_seg);
3069 return true;
3070 }
3071
3072 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3073 main_blkaddr) {
3074 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3075 ssa_blkaddr, main_blkaddr,
3076 segment_count_ssa << log_blocks_per_seg);
3077 return true;
3078 }
3079
3080 if (main_end_blkaddr > seg_end_blkaddr) {
3081 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3082 main_blkaddr, seg_end_blkaddr,
3083 segment_count_main << log_blocks_per_seg);
3084 return true;
3085 } else if (main_end_blkaddr < seg_end_blkaddr) {
3086 int err = 0;
3087 char *res;
3088
3089 /* fix in-memory information all the time */
3090 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3091 segment0_blkaddr) >> log_blocks_per_seg);
3092
3093 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3094 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3095 res = "internally";
3096 } else {
3097 err = __f2fs_commit_super(bh, NULL);
3098 res = err ? "failed" : "done";
3099 }
3100 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3101 res, main_blkaddr, seg_end_blkaddr,
3102 segment_count_main << log_blocks_per_seg);
3103 if (err)
3104 return true;
3105 }
3106 return false;
3107 }
3108
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)3109 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3110 struct buffer_head *bh)
3111 {
3112 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3113 block_t total_sections, blocks_per_seg;
3114 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3115 (bh->b_data + F2FS_SUPER_OFFSET);
3116 size_t crc_offset = 0;
3117 __u32 crc = 0;
3118
3119 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3120 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3121 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3122 return -EINVAL;
3123 }
3124
3125 /* Check checksum_offset and crc in superblock */
3126 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3127 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3128 if (crc_offset !=
3129 offsetof(struct f2fs_super_block, crc)) {
3130 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3131 crc_offset);
3132 return -EFSCORRUPTED;
3133 }
3134 crc = le32_to_cpu(raw_super->crc);
3135 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3136 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3137 return -EFSCORRUPTED;
3138 }
3139 }
3140
3141 /* Currently, support only 4KB block size */
3142 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3143 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3144 le32_to_cpu(raw_super->log_blocksize),
3145 F2FS_BLKSIZE_BITS);
3146 return -EFSCORRUPTED;
3147 }
3148
3149 /* check log blocks per segment */
3150 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3151 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3152 le32_to_cpu(raw_super->log_blocks_per_seg));
3153 return -EFSCORRUPTED;
3154 }
3155
3156 /* Currently, support 512/1024/2048/4096 bytes sector size */
3157 if (le32_to_cpu(raw_super->log_sectorsize) >
3158 F2FS_MAX_LOG_SECTOR_SIZE ||
3159 le32_to_cpu(raw_super->log_sectorsize) <
3160 F2FS_MIN_LOG_SECTOR_SIZE) {
3161 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3162 le32_to_cpu(raw_super->log_sectorsize));
3163 return -EFSCORRUPTED;
3164 }
3165 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3166 le32_to_cpu(raw_super->log_sectorsize) !=
3167 F2FS_MAX_LOG_SECTOR_SIZE) {
3168 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3169 le32_to_cpu(raw_super->log_sectors_per_block),
3170 le32_to_cpu(raw_super->log_sectorsize));
3171 return -EFSCORRUPTED;
3172 }
3173
3174 segment_count = le32_to_cpu(raw_super->segment_count);
3175 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3176 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3177 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3178 total_sections = le32_to_cpu(raw_super->section_count);
3179
3180 /* blocks_per_seg should be 512, given the above check */
3181 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3182
3183 if (segment_count > F2FS_MAX_SEGMENT ||
3184 segment_count < F2FS_MIN_SEGMENTS) {
3185 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3186 return -EFSCORRUPTED;
3187 }
3188
3189 if (total_sections > segment_count_main || total_sections < 1 ||
3190 segs_per_sec > segment_count || !segs_per_sec) {
3191 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3192 segment_count, total_sections, segs_per_sec);
3193 return -EFSCORRUPTED;
3194 }
3195
3196 if (segment_count_main != total_sections * segs_per_sec) {
3197 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3198 segment_count_main, total_sections, segs_per_sec);
3199 return -EFSCORRUPTED;
3200 }
3201
3202 if ((segment_count / segs_per_sec) < total_sections) {
3203 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3204 segment_count, segs_per_sec, total_sections);
3205 return -EFSCORRUPTED;
3206 }
3207
3208 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3209 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3210 segment_count, le64_to_cpu(raw_super->block_count));
3211 return -EFSCORRUPTED;
3212 }
3213
3214 if (RDEV(0).path[0]) {
3215 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3216 int i = 1;
3217
3218 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3219 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3220 i++;
3221 }
3222 if (segment_count != dev_seg_count) {
3223 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3224 segment_count, dev_seg_count);
3225 return -EFSCORRUPTED;
3226 }
3227 } else {
3228 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3229 !bdev_is_zoned(sbi->sb->s_bdev)) {
3230 f2fs_info(sbi, "Zoned block device path is missing");
3231 return -EFSCORRUPTED;
3232 }
3233 }
3234
3235 if (secs_per_zone > total_sections || !secs_per_zone) {
3236 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3237 secs_per_zone, total_sections);
3238 return -EFSCORRUPTED;
3239 }
3240 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3241 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3242 (le32_to_cpu(raw_super->extension_count) +
3243 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3244 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3245 le32_to_cpu(raw_super->extension_count),
3246 raw_super->hot_ext_count,
3247 F2FS_MAX_EXTENSION);
3248 return -EFSCORRUPTED;
3249 }
3250
3251 if (le32_to_cpu(raw_super->cp_payload) >=
3252 (blocks_per_seg - F2FS_CP_PACKS -
3253 NR_CURSEG_PERSIST_TYPE)) {
3254 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3255 le32_to_cpu(raw_super->cp_payload),
3256 blocks_per_seg - F2FS_CP_PACKS -
3257 NR_CURSEG_PERSIST_TYPE);
3258 return -EFSCORRUPTED;
3259 }
3260
3261 /* check reserved ino info */
3262 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3263 le32_to_cpu(raw_super->meta_ino) != 2 ||
3264 le32_to_cpu(raw_super->root_ino) != 3) {
3265 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3266 le32_to_cpu(raw_super->node_ino),
3267 le32_to_cpu(raw_super->meta_ino),
3268 le32_to_cpu(raw_super->root_ino));
3269 return -EFSCORRUPTED;
3270 }
3271
3272 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3273 if (sanity_check_area_boundary(sbi, bh))
3274 return -EFSCORRUPTED;
3275
3276 return 0;
3277 }
3278
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3279 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3280 {
3281 unsigned int total, fsmeta;
3282 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3283 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3284 unsigned int ovp_segments, reserved_segments;
3285 unsigned int main_segs, blocks_per_seg;
3286 unsigned int sit_segs, nat_segs;
3287 unsigned int sit_bitmap_size, nat_bitmap_size;
3288 unsigned int log_blocks_per_seg;
3289 unsigned int segment_count_main;
3290 unsigned int cp_pack_start_sum, cp_payload;
3291 block_t user_block_count, valid_user_blocks;
3292 block_t avail_node_count, valid_node_count;
3293 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3294 int i, j;
3295
3296 total = le32_to_cpu(raw_super->segment_count);
3297 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3298 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3299 fsmeta += sit_segs;
3300 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3301 fsmeta += nat_segs;
3302 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3303 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3304
3305 if (unlikely(fsmeta >= total))
3306 return 1;
3307
3308 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3309 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3310
3311 if (!f2fs_sb_has_readonly(sbi) &&
3312 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3313 ovp_segments == 0 || reserved_segments == 0)) {
3314 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3315 return 1;
3316 }
3317 user_block_count = le64_to_cpu(ckpt->user_block_count);
3318 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3319 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3320 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3321 if (!user_block_count || user_block_count >=
3322 segment_count_main << log_blocks_per_seg) {
3323 f2fs_err(sbi, "Wrong user_block_count: %u",
3324 user_block_count);
3325 return 1;
3326 }
3327
3328 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3329 if (valid_user_blocks > user_block_count) {
3330 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3331 valid_user_blocks, user_block_count);
3332 return 1;
3333 }
3334
3335 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3336 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3337 if (valid_node_count > avail_node_count) {
3338 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3339 valid_node_count, avail_node_count);
3340 return 1;
3341 }
3342
3343 main_segs = le32_to_cpu(raw_super->segment_count_main);
3344 blocks_per_seg = sbi->blocks_per_seg;
3345
3346 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3347 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3348 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3349 return 1;
3350
3351 if (f2fs_sb_has_readonly(sbi))
3352 goto check_data;
3353
3354 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3355 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3356 le32_to_cpu(ckpt->cur_node_segno[j])) {
3357 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3358 i, j,
3359 le32_to_cpu(ckpt->cur_node_segno[i]));
3360 return 1;
3361 }
3362 }
3363 }
3364 check_data:
3365 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3366 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3367 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3368 return 1;
3369
3370 if (f2fs_sb_has_readonly(sbi))
3371 goto skip_cross;
3372
3373 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3374 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3375 le32_to_cpu(ckpt->cur_data_segno[j])) {
3376 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3377 i, j,
3378 le32_to_cpu(ckpt->cur_data_segno[i]));
3379 return 1;
3380 }
3381 }
3382 }
3383 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3384 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3385 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3386 le32_to_cpu(ckpt->cur_data_segno[j])) {
3387 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3388 i, j,
3389 le32_to_cpu(ckpt->cur_node_segno[i]));
3390 return 1;
3391 }
3392 }
3393 }
3394 skip_cross:
3395 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3396 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3397
3398 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3399 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3400 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3401 sit_bitmap_size, nat_bitmap_size);
3402 return 1;
3403 }
3404
3405 cp_pack_start_sum = __start_sum_addr(sbi);
3406 cp_payload = __cp_payload(sbi);
3407 if (cp_pack_start_sum < cp_payload + 1 ||
3408 cp_pack_start_sum > blocks_per_seg - 1 -
3409 NR_CURSEG_PERSIST_TYPE) {
3410 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3411 cp_pack_start_sum);
3412 return 1;
3413 }
3414
3415 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3416 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3417 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3418 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3419 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3420 le32_to_cpu(ckpt->checksum_offset));
3421 return 1;
3422 }
3423
3424 nat_blocks = nat_segs << log_blocks_per_seg;
3425 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3426 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3427 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3428 (cp_payload + F2FS_CP_PACKS +
3429 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3430 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3431 cp_payload, nat_bits_blocks);
3432 return 1;
3433 }
3434
3435 if (unlikely(f2fs_cp_error(sbi))) {
3436 f2fs_err(sbi, "A bug case: need to run fsck");
3437 return 1;
3438 }
3439 return 0;
3440 }
3441
init_sb_info(struct f2fs_sb_info * sbi)3442 static void init_sb_info(struct f2fs_sb_info *sbi)
3443 {
3444 struct f2fs_super_block *raw_super = sbi->raw_super;
3445 int i;
3446
3447 sbi->log_sectors_per_block =
3448 le32_to_cpu(raw_super->log_sectors_per_block);
3449 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3450 sbi->blocksize = 1 << sbi->log_blocksize;
3451 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3452 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3453 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3454 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3455 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3456 sbi->total_node_count =
3457 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3458 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3459 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3460 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3461 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3462 sbi->cur_victim_sec = NULL_SECNO;
3463 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3464 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3465 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3466 sbi->migration_granularity = sbi->segs_per_sec;
3467
3468 sbi->dir_level = DEF_DIR_LEVEL;
3469 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3470 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3471 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3472 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3473 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3474 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3475 DEF_UMOUNT_DISCARD_TIMEOUT;
3476 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3477
3478 for (i = 0; i < NR_COUNT_TYPE; i++)
3479 atomic_set(&sbi->nr_pages[i], 0);
3480
3481 for (i = 0; i < META; i++)
3482 atomic_set(&sbi->wb_sync_req[i], 0);
3483
3484 INIT_LIST_HEAD(&sbi->s_list);
3485 mutex_init(&sbi->umount_mutex);
3486 init_f2fs_rwsem(&sbi->io_order_lock);
3487 spin_lock_init(&sbi->cp_lock);
3488
3489 sbi->dirty_device = 0;
3490 spin_lock_init(&sbi->dev_lock);
3491
3492 init_f2fs_rwsem(&sbi->sb_lock);
3493 init_f2fs_rwsem(&sbi->pin_sem);
3494 }
3495
init_percpu_info(struct f2fs_sb_info * sbi)3496 static int init_percpu_info(struct f2fs_sb_info *sbi)
3497 {
3498 int err;
3499
3500 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3501 if (err)
3502 return err;
3503
3504 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3505 GFP_KERNEL);
3506 if (err)
3507 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3508
3509 return err;
3510 }
3511
3512 #ifdef CONFIG_BLK_DEV_ZONED
3513
3514 struct f2fs_report_zones_args {
3515 struct f2fs_sb_info *sbi;
3516 struct f2fs_dev_info *dev;
3517 };
3518
f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3519 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3520 void *data)
3521 {
3522 struct f2fs_report_zones_args *rz_args = data;
3523 block_t unusable_blocks = (zone->len - zone->capacity) >>
3524 F2FS_LOG_SECTORS_PER_BLOCK;
3525
3526 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3527 return 0;
3528
3529 set_bit(idx, rz_args->dev->blkz_seq);
3530 if (!rz_args->sbi->unusable_blocks_per_sec) {
3531 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3532 return 0;
3533 }
3534 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3535 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3536 return -EINVAL;
3537 }
3538 return 0;
3539 }
3540
init_blkz_info(struct f2fs_sb_info * sbi,int devi)3541 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3542 {
3543 struct block_device *bdev = FDEV(devi).bdev;
3544 sector_t nr_sectors = bdev->bd_part->nr_sects;
3545 struct f2fs_report_zones_args rep_zone_arg;
3546 int ret;
3547
3548 if (!f2fs_sb_has_blkzoned(sbi))
3549 return 0;
3550
3551 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3552 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3553 return -EINVAL;
3554 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3555 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3556 __ilog2_u32(sbi->blocks_per_blkz))
3557 return -EINVAL;
3558 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3559 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3560 sbi->log_blocks_per_blkz;
3561 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3562 FDEV(devi).nr_blkz++;
3563
3564 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3565 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3566 * sizeof(unsigned long),
3567 GFP_KERNEL);
3568 if (!FDEV(devi).blkz_seq)
3569 return -ENOMEM;
3570
3571 rep_zone_arg.sbi = sbi;
3572 rep_zone_arg.dev = &FDEV(devi);
3573
3574 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3575 &rep_zone_arg);
3576 if (ret < 0)
3577 return ret;
3578 return 0;
3579 }
3580 #endif
3581
3582 /*
3583 * Read f2fs raw super block.
3584 * Because we have two copies of super block, so read both of them
3585 * to get the first valid one. If any one of them is broken, we pass
3586 * them recovery flag back to the caller.
3587 */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3588 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3589 struct f2fs_super_block **raw_super,
3590 int *valid_super_block, int *recovery)
3591 {
3592 struct super_block *sb = sbi->sb;
3593 int block;
3594 struct buffer_head *bh;
3595 struct f2fs_super_block *super;
3596 int err = 0;
3597
3598 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3599 if (!super)
3600 return -ENOMEM;
3601
3602 for (block = 0; block < 2; block++) {
3603 bh = sb_bread(sb, block);
3604 if (!bh) {
3605 f2fs_err(sbi, "Unable to read %dth superblock",
3606 block + 1);
3607 err = -EIO;
3608 *recovery = 1;
3609 continue;
3610 }
3611
3612 /* sanity checking of raw super */
3613 err = sanity_check_raw_super(sbi, bh);
3614 if (err) {
3615 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3616 block + 1);
3617 brelse(bh);
3618 *recovery = 1;
3619 continue;
3620 }
3621
3622 if (!*raw_super) {
3623 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3624 sizeof(*super));
3625 *valid_super_block = block;
3626 *raw_super = super;
3627 }
3628 brelse(bh);
3629 }
3630
3631 /* No valid superblock */
3632 if (!*raw_super)
3633 kfree(super);
3634 else
3635 err = 0;
3636
3637 return err;
3638 }
3639
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)3640 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3641 {
3642 struct buffer_head *bh;
3643 __u32 crc = 0;
3644 int err;
3645
3646 if ((recover && f2fs_readonly(sbi->sb)) ||
3647 bdev_read_only(sbi->sb->s_bdev)) {
3648 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3649 return -EROFS;
3650 }
3651
3652 /* we should update superblock crc here */
3653 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3654 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3655 offsetof(struct f2fs_super_block, crc));
3656 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3657 }
3658
3659 /* write back-up superblock first */
3660 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3661 if (!bh)
3662 return -EIO;
3663 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3664 brelse(bh);
3665
3666 /* if we are in recovery path, skip writing valid superblock */
3667 if (recover || err)
3668 return err;
3669
3670 /* write current valid superblock */
3671 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3672 if (!bh)
3673 return -EIO;
3674 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3675 brelse(bh);
3676 return err;
3677 }
3678
f2fs_handle_stop(struct f2fs_sb_info * sbi,unsigned char reason)3679 void f2fs_handle_stop(struct f2fs_sb_info *sbi, unsigned char reason)
3680 {
3681 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3682 int err;
3683
3684 f2fs_bug_on(sbi, reason >= MAX_STOP_REASON);
3685
3686 f2fs_down_write(&sbi->sb_lock);
3687
3688 if (raw_super->s_stop_reason[reason] < ((1 << BITS_PER_BYTE) - 1))
3689 raw_super->s_stop_reason[reason]++;
3690
3691 err = f2fs_commit_super(sbi, false);
3692 if (err)
3693 f2fs_err(sbi, "f2fs_commit_super fails to record reason:%u err:%d",
3694 reason, err);
3695
3696 f2fs_up_write(&sbi->sb_lock);
3697 }
3698
f2fs_scan_devices(struct f2fs_sb_info * sbi)3699 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3700 {
3701 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3702 unsigned int max_devices = MAX_DEVICES;
3703 int i;
3704
3705 /* Initialize single device information */
3706 if (!RDEV(0).path[0]) {
3707 if (!bdev_is_zoned(sbi->sb->s_bdev))
3708 return 0;
3709 max_devices = 1;
3710 }
3711
3712 /*
3713 * Initialize multiple devices information, or single
3714 * zoned block device information.
3715 */
3716 sbi->devs = f2fs_kzalloc(sbi,
3717 array_size(max_devices,
3718 sizeof(struct f2fs_dev_info)),
3719 GFP_KERNEL);
3720 if (!sbi->devs)
3721 return -ENOMEM;
3722
3723 for (i = 0; i < max_devices; i++) {
3724
3725 if (i > 0 && !RDEV(i).path[0])
3726 break;
3727
3728 if (max_devices == 1) {
3729 /* Single zoned block device mount */
3730 FDEV(0).bdev =
3731 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3732 sbi->sb->s_mode, sbi->sb->s_type);
3733 } else {
3734 /* Multi-device mount */
3735 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3736 FDEV(i).total_segments =
3737 le32_to_cpu(RDEV(i).total_segments);
3738 if (i == 0) {
3739 FDEV(i).start_blk = 0;
3740 FDEV(i).end_blk = FDEV(i).start_blk +
3741 (FDEV(i).total_segments <<
3742 sbi->log_blocks_per_seg) - 1 +
3743 le32_to_cpu(raw_super->segment0_blkaddr);
3744 } else {
3745 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3746 FDEV(i).end_blk = FDEV(i).start_blk +
3747 (FDEV(i).total_segments <<
3748 sbi->log_blocks_per_seg) - 1;
3749 }
3750 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3751 sbi->sb->s_mode, sbi->sb->s_type);
3752 }
3753 if (IS_ERR(FDEV(i).bdev))
3754 return PTR_ERR(FDEV(i).bdev);
3755
3756 /* to release errored devices */
3757 sbi->s_ndevs = i + 1;
3758
3759 #ifdef CONFIG_BLK_DEV_ZONED
3760 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3761 !f2fs_sb_has_blkzoned(sbi)) {
3762 f2fs_err(sbi, "Zoned block device feature not enabled");
3763 return -EINVAL;
3764 }
3765 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3766 if (init_blkz_info(sbi, i)) {
3767 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3768 return -EINVAL;
3769 }
3770 if (max_devices == 1)
3771 break;
3772 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3773 i, FDEV(i).path,
3774 FDEV(i).total_segments,
3775 FDEV(i).start_blk, FDEV(i).end_blk,
3776 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3777 "Host-aware" : "Host-managed");
3778 continue;
3779 }
3780 #endif
3781 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3782 i, FDEV(i).path,
3783 FDEV(i).total_segments,
3784 FDEV(i).start_blk, FDEV(i).end_blk);
3785 }
3786 f2fs_info(sbi,
3787 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3788 return 0;
3789 }
3790
f2fs_setup_casefold(struct f2fs_sb_info * sbi)3791 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3792 {
3793 #ifdef CONFIG_UNICODE
3794 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3795 const struct f2fs_sb_encodings *encoding_info;
3796 struct unicode_map *encoding;
3797 __u16 encoding_flags;
3798
3799 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3800 &encoding_flags)) {
3801 f2fs_err(sbi,
3802 "Encoding requested by superblock is unknown");
3803 return -EINVAL;
3804 }
3805
3806 encoding = utf8_load(encoding_info->version);
3807 if (IS_ERR(encoding)) {
3808 f2fs_err(sbi,
3809 "can't mount with superblock charset: %s-%s "
3810 "not supported by the kernel. flags: 0x%x.",
3811 encoding_info->name, encoding_info->version,
3812 encoding_flags);
3813 return PTR_ERR(encoding);
3814 }
3815 f2fs_info(sbi, "Using encoding defined by superblock: "
3816 "%s-%s with flags 0x%hx", encoding_info->name,
3817 encoding_info->version?:"\b", encoding_flags);
3818
3819 sbi->sb->s_encoding = encoding;
3820 sbi->sb->s_encoding_flags = encoding_flags;
3821 }
3822 #else
3823 if (f2fs_sb_has_casefold(sbi)) {
3824 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3825 return -EINVAL;
3826 }
3827 #endif
3828 return 0;
3829 }
3830
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)3831 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3832 {
3833 struct f2fs_sm_info *sm_i = SM_I(sbi);
3834
3835 /* adjust parameters according to the volume size */
3836 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3837 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3838 sm_i->dcc_info->discard_granularity = 1;
3839 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE |
3840 1 << F2FS_IPU_HONOR_OPU_WRITE;
3841 }
3842
3843 sbi->readdir_ra = 1;
3844 }
3845
f2fs_fill_super(struct super_block * sb,void * data,int silent)3846 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3847 {
3848 struct f2fs_sb_info *sbi;
3849 struct f2fs_super_block *raw_super;
3850 struct inode *root;
3851 int err;
3852 bool skip_recovery = false, need_fsck = false;
3853 char *options = NULL;
3854 int recovery, i, valid_super_block;
3855 struct curseg_info *seg_i;
3856 int retry_cnt = 1;
3857
3858 try_onemore:
3859 err = -EINVAL;
3860 raw_super = NULL;
3861 valid_super_block = -1;
3862 recovery = 0;
3863
3864 /* allocate memory for f2fs-specific super block info */
3865 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3866 if (!sbi)
3867 return -ENOMEM;
3868
3869 sbi->sb = sb;
3870
3871 /* Load the checksum driver */
3872 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3873 if (IS_ERR(sbi->s_chksum_driver)) {
3874 f2fs_err(sbi, "Cannot load crc32 driver.");
3875 err = PTR_ERR(sbi->s_chksum_driver);
3876 sbi->s_chksum_driver = NULL;
3877 goto free_sbi;
3878 }
3879
3880 /* set a block size */
3881 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3882 f2fs_err(sbi, "unable to set blocksize");
3883 goto free_sbi;
3884 }
3885
3886 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3887 &recovery);
3888 if (err)
3889 goto free_sbi;
3890
3891 sb->s_fs_info = sbi;
3892 sbi->raw_super = raw_super;
3893
3894 /* precompute checksum seed for metadata */
3895 if (f2fs_sb_has_inode_chksum(sbi))
3896 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3897 sizeof(raw_super->uuid));
3898
3899 default_options(sbi, false);
3900 /* parse mount options */
3901 options = kstrdup((const char *)data, GFP_KERNEL);
3902 if (data && !options) {
3903 err = -ENOMEM;
3904 goto free_sb_buf;
3905 }
3906
3907 err = parse_options(sb, options, false);
3908 if (err)
3909 goto free_options;
3910
3911 sb->s_maxbytes = max_file_blocks(NULL) <<
3912 le32_to_cpu(raw_super->log_blocksize);
3913 sb->s_max_links = F2FS_LINK_MAX;
3914
3915 err = f2fs_setup_casefold(sbi);
3916 if (err)
3917 goto free_options;
3918
3919 #ifdef CONFIG_QUOTA
3920 sb->dq_op = &f2fs_quota_operations;
3921 sb->s_qcop = &f2fs_quotactl_ops;
3922 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3923
3924 if (f2fs_sb_has_quota_ino(sbi)) {
3925 for (i = 0; i < MAXQUOTAS; i++) {
3926 if (f2fs_qf_ino(sbi->sb, i))
3927 sbi->nquota_files++;
3928 }
3929 }
3930 #endif
3931
3932 sb->s_op = &f2fs_sops;
3933 #ifdef CONFIG_FS_ENCRYPTION
3934 sb->s_cop = &f2fs_cryptops;
3935 #endif
3936 #ifdef CONFIG_FS_VERITY
3937 sb->s_vop = &f2fs_verityops;
3938 #endif
3939 sb->s_xattr = f2fs_xattr_handlers;
3940 sb->s_export_op = &f2fs_export_ops;
3941 sb->s_magic = F2FS_SUPER_MAGIC;
3942 sb->s_time_gran = 1;
3943 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3944 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3945 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3946 sb->s_iflags |= SB_I_CGROUPWB;
3947
3948 /* init f2fs-specific super block info */
3949 sbi->valid_super_block = valid_super_block;
3950 init_f2fs_rwsem(&sbi->gc_lock);
3951 mutex_init(&sbi->writepages);
3952 init_f2fs_rwsem(&sbi->cp_global_sem);
3953 init_f2fs_rwsem(&sbi->node_write);
3954 init_f2fs_rwsem(&sbi->node_change);
3955
3956 /* disallow all the data/node/meta page writes */
3957 set_sbi_flag(sbi, SBI_POR_DOING);
3958 spin_lock_init(&sbi->stat_lock);
3959
3960 /* init iostat info */
3961 spin_lock_init(&sbi->iostat_lock);
3962 sbi->iostat_enable = false;
3963 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3964
3965 for (i = 0; i < NR_PAGE_TYPE; i++) {
3966 int n = (i == META) ? 1 : NR_TEMP_TYPE;
3967 int j;
3968
3969 sbi->write_io[i] =
3970 f2fs_kmalloc(sbi,
3971 array_size(n,
3972 sizeof(struct f2fs_bio_info)),
3973 GFP_KERNEL);
3974 if (!sbi->write_io[i]) {
3975 err = -ENOMEM;
3976 goto free_bio_info;
3977 }
3978
3979 for (j = HOT; j < n; j++) {
3980 init_f2fs_rwsem(&sbi->write_io[i][j].io_rwsem);
3981 sbi->write_io[i][j].sbi = sbi;
3982 sbi->write_io[i][j].bio = NULL;
3983 spin_lock_init(&sbi->write_io[i][j].io_lock);
3984 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3985 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3986 init_f2fs_rwsem(&sbi->write_io[i][j].bio_list_lock);
3987 }
3988 }
3989
3990 init_f2fs_rwsem(&sbi->cp_rwsem);
3991 init_f2fs_rwsem(&sbi->quota_sem);
3992 init_waitqueue_head(&sbi->cp_wait);
3993 init_sb_info(sbi);
3994
3995 err = init_percpu_info(sbi);
3996 if (err)
3997 goto free_bio_info;
3998
3999 if (F2FS_IO_ALIGNED(sbi)) {
4000 sbi->write_io_dummy =
4001 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4002 if (!sbi->write_io_dummy) {
4003 err = -ENOMEM;
4004 goto free_percpu;
4005 }
4006 }
4007
4008 /* init per sbi slab cache */
4009 err = f2fs_init_xattr_caches(sbi);
4010 if (err)
4011 goto free_io_dummy;
4012 err = f2fs_init_page_array_cache(sbi);
4013 if (err)
4014 goto free_xattr_cache;
4015
4016 /* get an inode for meta space */
4017 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4018 if (IS_ERR(sbi->meta_inode)) {
4019 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4020 err = PTR_ERR(sbi->meta_inode);
4021 goto free_page_array_cache;
4022 }
4023
4024 err = f2fs_get_valid_checkpoint(sbi);
4025 if (err) {
4026 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4027 goto free_meta_inode;
4028 }
4029
4030 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4031 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4032 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4033 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4034 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4035 }
4036
4037 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4038 set_sbi_flag(sbi, SBI_NEED_FSCK);
4039
4040 /* Initialize device list */
4041 err = f2fs_scan_devices(sbi);
4042 if (err) {
4043 f2fs_err(sbi, "Failed to find devices");
4044 goto free_devices;
4045 }
4046
4047 err = f2fs_init_post_read_wq(sbi);
4048 if (err) {
4049 f2fs_err(sbi, "Failed to initialize post read workqueue");
4050 goto free_devices;
4051 }
4052
4053 sbi->total_valid_node_count =
4054 le32_to_cpu(sbi->ckpt->valid_node_count);
4055 percpu_counter_set(&sbi->total_valid_inode_count,
4056 le32_to_cpu(sbi->ckpt->valid_inode_count));
4057 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4058 sbi->total_valid_block_count =
4059 le64_to_cpu(sbi->ckpt->valid_block_count);
4060 sbi->last_valid_block_count = sbi->total_valid_block_count;
4061 sbi->reserved_blocks = 0;
4062 sbi->current_reserved_blocks = 0;
4063 limit_reserve_root(sbi);
4064 adjust_unusable_cap_perc(sbi);
4065
4066 for (i = 0; i < NR_INODE_TYPE; i++) {
4067 INIT_LIST_HEAD(&sbi->inode_list[i]);
4068 spin_lock_init(&sbi->inode_lock[i]);
4069 }
4070 mutex_init(&sbi->flush_lock);
4071
4072 f2fs_init_extent_cache_info(sbi);
4073
4074 f2fs_init_ino_entry_info(sbi);
4075
4076 f2fs_init_fsync_node_info(sbi);
4077
4078 /* setup checkpoint request control and start checkpoint issue thread */
4079 f2fs_init_ckpt_req_control(sbi);
4080 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4081 test_opt(sbi, MERGE_CHECKPOINT)) {
4082 err = f2fs_start_ckpt_thread(sbi);
4083 if (err) {
4084 f2fs_err(sbi,
4085 "Failed to start F2FS issue_checkpoint_thread (%d)",
4086 err);
4087 goto stop_ckpt_thread;
4088 }
4089 }
4090
4091 /* setup f2fs internal modules */
4092 err = f2fs_build_segment_manager(sbi);
4093 if (err) {
4094 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4095 err);
4096 goto free_sm;
4097 }
4098 err = f2fs_build_node_manager(sbi);
4099 if (err) {
4100 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4101 err);
4102 goto free_nm;
4103 }
4104
4105 err = adjust_reserved_segment(sbi);
4106 if (err)
4107 goto free_nm;
4108
4109 /* For write statistics */
4110 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4111
4112 /* Read accumulated write IO statistics if exists */
4113 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4114 if (__exist_node_summaries(sbi))
4115 sbi->kbytes_written =
4116 le64_to_cpu(seg_i->journal->info.kbytes_written);
4117
4118 f2fs_build_gc_manager(sbi);
4119
4120 err = f2fs_build_stats(sbi);
4121 if (err)
4122 goto free_nm;
4123
4124 /* get an inode for node space */
4125 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4126 if (IS_ERR(sbi->node_inode)) {
4127 f2fs_err(sbi, "Failed to read node inode");
4128 err = PTR_ERR(sbi->node_inode);
4129 goto free_stats;
4130 }
4131
4132 /* read root inode and dentry */
4133 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4134 if (IS_ERR(root)) {
4135 f2fs_err(sbi, "Failed to read root inode");
4136 err = PTR_ERR(root);
4137 goto free_node_inode;
4138 }
4139 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4140 !root->i_size || !root->i_nlink) {
4141 iput(root);
4142 err = -EINVAL;
4143 goto free_node_inode;
4144 }
4145
4146 sb->s_root = d_make_root(root); /* allocate root dentry */
4147 if (!sb->s_root) {
4148 err = -ENOMEM;
4149 goto free_node_inode;
4150 }
4151
4152 err = f2fs_init_compress_inode(sbi);
4153 if (err)
4154 goto free_root_inode;
4155
4156 err = f2fs_register_sysfs(sbi);
4157 if (err)
4158 goto free_compress_inode;
4159
4160 #ifdef CONFIG_QUOTA
4161 /* Enable quota usage during mount */
4162 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4163 err = f2fs_enable_quotas(sb);
4164 if (err)
4165 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4166 }
4167 #endif
4168 /* if there are any orphan inodes, free them */
4169 err = f2fs_recover_orphan_inodes(sbi);
4170 if (err)
4171 goto free_meta;
4172
4173 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4174 goto reset_checkpoint;
4175
4176 /* recover fsynced data */
4177 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4178 !test_opt(sbi, NORECOVERY)) {
4179 /*
4180 * mount should be failed, when device has readonly mode, and
4181 * previous checkpoint was not done by clean system shutdown.
4182 */
4183 if (f2fs_hw_is_readonly(sbi)) {
4184 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4185 err = f2fs_recover_fsync_data(sbi, true);
4186 if (err > 0) {
4187 err = -EROFS;
4188 f2fs_err(sbi, "Need to recover fsync data, but "
4189 "write access unavailable, please try "
4190 "mount w/ disable_roll_forward or norecovery");
4191 }
4192 if (err < 0)
4193 goto free_meta;
4194 }
4195 f2fs_info(sbi, "write access unavailable, skipping recovery");
4196 goto reset_checkpoint;
4197 }
4198
4199 if (need_fsck)
4200 set_sbi_flag(sbi, SBI_NEED_FSCK);
4201
4202 if (skip_recovery)
4203 goto reset_checkpoint;
4204
4205 err = f2fs_recover_fsync_data(sbi, false);
4206 if (err < 0) {
4207 if (err != -ENOMEM)
4208 skip_recovery = true;
4209 need_fsck = true;
4210 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4211 err);
4212 goto free_meta;
4213 }
4214 } else {
4215 err = f2fs_recover_fsync_data(sbi, true);
4216
4217 if (!f2fs_readonly(sb) && err > 0) {
4218 err = -EINVAL;
4219 f2fs_err(sbi, "Need to recover fsync data");
4220 goto free_meta;
4221 }
4222 }
4223
4224 /*
4225 * If the f2fs is not readonly and fsync data recovery succeeds,
4226 * check zoned block devices' write pointer consistency.
4227 */
4228 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4229 err = f2fs_check_write_pointer(sbi);
4230 if (err)
4231 goto free_meta;
4232 }
4233
4234 reset_checkpoint:
4235 f2fs_init_inmem_curseg(sbi);
4236
4237 /* f2fs_recover_fsync_data() cleared this already */
4238 clear_sbi_flag(sbi, SBI_POR_DOING);
4239
4240 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4241 err = f2fs_disable_checkpoint(sbi);
4242 if (err)
4243 goto sync_free_meta;
4244 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4245 f2fs_enable_checkpoint(sbi);
4246 }
4247
4248 /*
4249 * If filesystem is not mounted as read-only then
4250 * do start the gc_thread.
4251 */
4252 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4253 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4254 /* After POR, we can run background GC thread.*/
4255 err = f2fs_start_gc_thread(sbi);
4256 if (err)
4257 goto sync_free_meta;
4258 }
4259 kvfree(options);
4260
4261 /* recover broken superblock */
4262 if (recovery) {
4263 err = f2fs_commit_super(sbi, true);
4264 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4265 sbi->valid_super_block ? 1 : 2, err);
4266 }
4267
4268 f2fs_join_shrinker(sbi);
4269
4270 f2fs_tuning_parameters(sbi);
4271
4272 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4273 cur_cp_version(F2FS_CKPT(sbi)));
4274 f2fs_update_time(sbi, CP_TIME);
4275 f2fs_update_time(sbi, REQ_TIME);
4276 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4277 return 0;
4278
4279 sync_free_meta:
4280 /* safe to flush all the data */
4281 sync_filesystem(sbi->sb);
4282 retry_cnt = 0;
4283
4284 free_meta:
4285 #ifdef CONFIG_QUOTA
4286 f2fs_truncate_quota_inode_pages(sb);
4287 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4288 f2fs_quota_off_umount(sbi->sb);
4289 #endif
4290 /*
4291 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4292 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4293 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4294 * falls into an infinite loop in f2fs_sync_meta_pages().
4295 */
4296 truncate_inode_pages_final(META_MAPPING(sbi));
4297 /* evict some inodes being cached by GC */
4298 evict_inodes(sb);
4299 f2fs_unregister_sysfs(sbi);
4300 free_compress_inode:
4301 f2fs_destroy_compress_inode(sbi);
4302 free_root_inode:
4303 dput(sb->s_root);
4304 sb->s_root = NULL;
4305 free_node_inode:
4306 f2fs_release_ino_entry(sbi, true);
4307 truncate_inode_pages_final(NODE_MAPPING(sbi));
4308 iput(sbi->node_inode);
4309 sbi->node_inode = NULL;
4310 free_stats:
4311 f2fs_destroy_stats(sbi);
4312 free_nm:
4313 /* stop discard thread before destroying node manager */
4314 f2fs_stop_discard_thread(sbi);
4315 f2fs_destroy_node_manager(sbi);
4316 free_sm:
4317 f2fs_destroy_segment_manager(sbi);
4318 f2fs_destroy_post_read_wq(sbi);
4319 stop_ckpt_thread:
4320 f2fs_stop_ckpt_thread(sbi);
4321 free_devices:
4322 destroy_device_list(sbi);
4323 kvfree(sbi->ckpt);
4324 free_meta_inode:
4325 make_bad_inode(sbi->meta_inode);
4326 iput(sbi->meta_inode);
4327 sbi->meta_inode = NULL;
4328 free_page_array_cache:
4329 f2fs_destroy_page_array_cache(sbi);
4330 free_xattr_cache:
4331 f2fs_destroy_xattr_caches(sbi);
4332 free_io_dummy:
4333 mempool_destroy(sbi->write_io_dummy);
4334 free_percpu:
4335 destroy_percpu_info(sbi);
4336 free_bio_info:
4337 for (i = 0; i < NR_PAGE_TYPE; i++)
4338 kvfree(sbi->write_io[i]);
4339
4340 #ifdef CONFIG_UNICODE
4341 utf8_unload(sb->s_encoding);
4342 sb->s_encoding = NULL;
4343 #endif
4344 free_options:
4345 #ifdef CONFIG_QUOTA
4346 for (i = 0; i < MAXQUOTAS; i++)
4347 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4348 #endif
4349 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4350 kvfree(options);
4351 free_sb_buf:
4352 kfree(raw_super);
4353 free_sbi:
4354 if (sbi->s_chksum_driver)
4355 crypto_free_shash(sbi->s_chksum_driver);
4356 kfree(sbi);
4357
4358 /* give only one another chance */
4359 if (retry_cnt > 0 && skip_recovery) {
4360 retry_cnt--;
4361 shrink_dcache_sb(sb);
4362 goto try_onemore;
4363 }
4364 return err;
4365 }
4366
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4367 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4368 const char *dev_name, void *data)
4369 {
4370 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4371 }
4372
kill_f2fs_super(struct super_block * sb)4373 static void kill_f2fs_super(struct super_block *sb)
4374 {
4375 if (sb->s_root) {
4376 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4377
4378 set_sbi_flag(sbi, SBI_IS_CLOSE);
4379 f2fs_stop_gc_thread(sbi);
4380 f2fs_stop_discard_thread(sbi);
4381
4382 #ifdef CONFIG_F2FS_FS_COMPRESSION
4383 /*
4384 * latter evict_inode() can bypass checking and invalidating
4385 * compress inode cache.
4386 */
4387 if (test_opt(sbi, COMPRESS_CACHE))
4388 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4389 #endif
4390
4391 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4392 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4393 struct cp_control cpc = {
4394 .reason = CP_UMOUNT,
4395 };
4396 f2fs_write_checkpoint(sbi, &cpc);
4397 }
4398
4399 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4400 sb->s_flags &= ~SB_RDONLY;
4401 }
4402 kill_block_super(sb);
4403 }
4404
4405 static struct file_system_type f2fs_fs_type = {
4406 .owner = THIS_MODULE,
4407 .name = "f2fs",
4408 .mount = f2fs_mount,
4409 .kill_sb = kill_f2fs_super,
4410 .fs_flags = FS_REQUIRES_DEV,
4411 };
4412 MODULE_ALIAS_FS("f2fs");
4413
init_inodecache(void)4414 static int __init init_inodecache(void)
4415 {
4416 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4417 sizeof(struct f2fs_inode_info), 0,
4418 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4419 if (!f2fs_inode_cachep)
4420 return -ENOMEM;
4421 return 0;
4422 }
4423
destroy_inodecache(void)4424 static void destroy_inodecache(void)
4425 {
4426 /*
4427 * Make sure all delayed rcu free inodes are flushed before we
4428 * destroy cache.
4429 */
4430 rcu_barrier();
4431 kmem_cache_destroy(f2fs_inode_cachep);
4432 }
4433
init_f2fs_fs(void)4434 static int __init init_f2fs_fs(void)
4435 {
4436 int err;
4437
4438 if (PAGE_SIZE != F2FS_BLKSIZE) {
4439 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4440 PAGE_SIZE, F2FS_BLKSIZE);
4441 return -EINVAL;
4442 }
4443
4444 err = init_inodecache();
4445 if (err)
4446 goto fail;
4447 err = f2fs_create_node_manager_caches();
4448 if (err)
4449 goto free_inodecache;
4450 err = f2fs_create_segment_manager_caches();
4451 if (err)
4452 goto free_node_manager_caches;
4453 err = f2fs_create_checkpoint_caches();
4454 if (err)
4455 goto free_segment_manager_caches;
4456 err = f2fs_create_recovery_cache();
4457 if (err)
4458 goto free_checkpoint_caches;
4459 err = f2fs_create_extent_cache();
4460 if (err)
4461 goto free_recovery_cache;
4462 err = f2fs_create_garbage_collection_cache();
4463 if (err)
4464 goto free_extent_cache;
4465 err = f2fs_init_sysfs();
4466 if (err)
4467 goto free_garbage_collection_cache;
4468 err = register_shrinker(&f2fs_shrinker_info);
4469 if (err)
4470 goto free_sysfs;
4471 err = register_filesystem(&f2fs_fs_type);
4472 if (err)
4473 goto free_shrinker;
4474 f2fs_create_root_stats();
4475 err = f2fs_init_post_read_processing();
4476 if (err)
4477 goto free_root_stats;
4478 err = f2fs_init_bio_entry_cache();
4479 if (err)
4480 goto free_post_read;
4481 err = f2fs_init_bioset();
4482 if (err)
4483 goto free_bio_enrty_cache;
4484 err = f2fs_init_compress_mempool();
4485 if (err)
4486 goto free_bioset;
4487 err = f2fs_init_compress_cache();
4488 if (err)
4489 goto free_compress_mempool;
4490 err = f2fs_create_casefold_cache();
4491 if (err)
4492 goto free_compress_cache;
4493 return 0;
4494 free_compress_cache:
4495 f2fs_destroy_compress_cache();
4496 free_compress_mempool:
4497 f2fs_destroy_compress_mempool();
4498 free_bioset:
4499 f2fs_destroy_bioset();
4500 free_bio_enrty_cache:
4501 f2fs_destroy_bio_entry_cache();
4502 free_post_read:
4503 f2fs_destroy_post_read_processing();
4504 free_root_stats:
4505 f2fs_destroy_root_stats();
4506 unregister_filesystem(&f2fs_fs_type);
4507 free_shrinker:
4508 unregister_shrinker(&f2fs_shrinker_info);
4509 free_sysfs:
4510 f2fs_exit_sysfs();
4511 free_garbage_collection_cache:
4512 f2fs_destroy_garbage_collection_cache();
4513 free_extent_cache:
4514 f2fs_destroy_extent_cache();
4515 free_recovery_cache:
4516 f2fs_destroy_recovery_cache();
4517 free_checkpoint_caches:
4518 f2fs_destroy_checkpoint_caches();
4519 free_segment_manager_caches:
4520 f2fs_destroy_segment_manager_caches();
4521 free_node_manager_caches:
4522 f2fs_destroy_node_manager_caches();
4523 free_inodecache:
4524 destroy_inodecache();
4525 fail:
4526 return err;
4527 }
4528
exit_f2fs_fs(void)4529 static void __exit exit_f2fs_fs(void)
4530 {
4531 f2fs_destroy_casefold_cache();
4532 f2fs_destroy_compress_cache();
4533 f2fs_destroy_compress_mempool();
4534 f2fs_destroy_bioset();
4535 f2fs_destroy_bio_entry_cache();
4536 f2fs_destroy_post_read_processing();
4537 f2fs_destroy_root_stats();
4538 unregister_filesystem(&f2fs_fs_type);
4539 unregister_shrinker(&f2fs_shrinker_info);
4540 f2fs_exit_sysfs();
4541 f2fs_destroy_garbage_collection_cache();
4542 f2fs_destroy_extent_cache();
4543 f2fs_destroy_recovery_cache();
4544 f2fs_destroy_checkpoint_caches();
4545 f2fs_destroy_segment_manager_caches();
4546 f2fs_destroy_node_manager_caches();
4547 destroy_inodecache();
4548 }
4549
4550 module_init(init_f2fs_fs)
4551 module_exit(exit_f2fs_fs)
4552
4553 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4554 MODULE_DESCRIPTION("Flash Friendly File System");
4555 MODULE_LICENSE("GPL");
4556 MODULE_IMPORT_NS(ANDROID_GKI_VFS_EXPORT_ONLY);
4557 MODULE_SOFTDEP("pre: crc32");
4558
4559