1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * fs/f2fs/super.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8 #include <linux/module.h>
9 #include <linux/init.h>
10 #include <linux/fs.h>
11 #include <linux/fs_context.h>
12 #include <linux/sched/mm.h>
13 #include <linux/statfs.h>
14 #include <linux/buffer_head.h>
15 #include <linux/kthread.h>
16 #include <linux/parser.h>
17 #include <linux/mount.h>
18 #include <linux/seq_file.h>
19 #include <linux/proc_fs.h>
20 #include <linux/random.h>
21 #include <linux/exportfs.h>
22 #include <linux/blkdev.h>
23 #include <linux/quotaops.h>
24 #include <linux/f2fs_fs.h>
25 #include <linux/sysfs.h>
26 #include <linux/quota.h>
27 #include <linux/unicode.h>
28 #include <linux/part_stat.h>
29 #include <linux/zstd.h>
30 #include <linux/lz4.h>
31
32 #include "f2fs.h"
33 #include "node.h"
34 #include "segment.h"
35 #include "xattr.h"
36 #include "gc.h"
37 #include "iostat.h"
38
39 #define CREATE_TRACE_POINTS
40 #include <trace/events/f2fs.h>
41
42 static struct kmem_cache *f2fs_inode_cachep;
43
44 #ifdef CONFIG_F2FS_FAULT_INJECTION
45
46 const char *f2fs_fault_name[FAULT_MAX] = {
47 [FAULT_KMALLOC] = "kmalloc",
48 [FAULT_KVMALLOC] = "kvmalloc",
49 [FAULT_PAGE_ALLOC] = "page alloc",
50 [FAULT_PAGE_GET] = "page get",
51 [FAULT_ALLOC_NID] = "alloc nid",
52 [FAULT_ORPHAN] = "orphan",
53 [FAULT_BLOCK] = "no more block",
54 [FAULT_DIR_DEPTH] = "too big dir depth",
55 [FAULT_EVICT_INODE] = "evict_inode fail",
56 [FAULT_TRUNCATE] = "truncate fail",
57 [FAULT_READ_IO] = "read IO error",
58 [FAULT_CHECKPOINT] = "checkpoint error",
59 [FAULT_DISCARD] = "discard error",
60 [FAULT_WRITE_IO] = "write IO error",
61 [FAULT_SLAB_ALLOC] = "slab alloc",
62 [FAULT_DQUOT_INIT] = "dquot initialize",
63 [FAULT_LOCK_OP] = "lock_op",
64 [FAULT_BLKADDR] = "invalid blkaddr",
65 };
66
f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned int rate,unsigned int type)67 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
68 unsigned int type)
69 {
70 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
71
72 if (rate) {
73 atomic_set(&ffi->inject_ops, 0);
74 ffi->inject_rate = rate;
75 }
76
77 if (type)
78 ffi->inject_type = type;
79
80 if (!rate && !type)
81 memset(ffi, 0, sizeof(struct f2fs_fault_info));
82 }
83 #endif
84
85 /* f2fs-wide shrinker description */
86 static struct shrinker f2fs_shrinker_info = {
87 .scan_objects = f2fs_shrink_scan,
88 .count_objects = f2fs_shrink_count,
89 .seeks = DEFAULT_SEEKS,
90 };
91
92 enum {
93 Opt_gc_background,
94 Opt_disable_roll_forward,
95 Opt_norecovery,
96 Opt_discard,
97 Opt_nodiscard,
98 Opt_noheap,
99 Opt_heap,
100 Opt_user_xattr,
101 Opt_nouser_xattr,
102 Opt_acl,
103 Opt_noacl,
104 Opt_active_logs,
105 Opt_disable_ext_identify,
106 Opt_inline_xattr,
107 Opt_noinline_xattr,
108 Opt_inline_xattr_size,
109 Opt_inline_data,
110 Opt_inline_dentry,
111 Opt_noinline_dentry,
112 Opt_flush_merge,
113 Opt_noflush_merge,
114 Opt_barrier,
115 Opt_nobarrier,
116 Opt_fastboot,
117 Opt_extent_cache,
118 Opt_noextent_cache,
119 Opt_noinline_data,
120 Opt_data_flush,
121 Opt_reserve_root,
122 Opt_resgid,
123 Opt_resuid,
124 Opt_mode,
125 Opt_io_size_bits,
126 Opt_fault_injection,
127 Opt_fault_type,
128 Opt_lazytime,
129 Opt_nolazytime,
130 Opt_quota,
131 Opt_noquota,
132 Opt_usrquota,
133 Opt_grpquota,
134 Opt_prjquota,
135 Opt_usrjquota,
136 Opt_grpjquota,
137 Opt_prjjquota,
138 Opt_offusrjquota,
139 Opt_offgrpjquota,
140 Opt_offprjjquota,
141 Opt_jqfmt_vfsold,
142 Opt_jqfmt_vfsv0,
143 Opt_jqfmt_vfsv1,
144 Opt_alloc,
145 Opt_fsync,
146 Opt_test_dummy_encryption,
147 Opt_inlinecrypt,
148 Opt_checkpoint_disable,
149 Opt_checkpoint_disable_cap,
150 Opt_checkpoint_disable_cap_perc,
151 Opt_checkpoint_enable,
152 Opt_checkpoint_merge,
153 Opt_nocheckpoint_merge,
154 Opt_compress_algorithm,
155 Opt_compress_log_size,
156 Opt_compress_extension,
157 Opt_nocompress_extension,
158 Opt_compress_chksum,
159 Opt_compress_mode,
160 Opt_compress_cache,
161 Opt_atgc,
162 Opt_gc_merge,
163 Opt_nogc_merge,
164 Opt_discard_unit,
165 Opt_memory_mode,
166 Opt_age_extent_cache,
167 Opt_errors,
168 Opt_err,
169 };
170
171 static match_table_t f2fs_tokens = {
172 {Opt_gc_background, "background_gc=%s"},
173 {Opt_disable_roll_forward, "disable_roll_forward"},
174 {Opt_norecovery, "norecovery"},
175 {Opt_discard, "discard"},
176 {Opt_nodiscard, "nodiscard"},
177 {Opt_noheap, "no_heap"},
178 {Opt_heap, "heap"},
179 {Opt_user_xattr, "user_xattr"},
180 {Opt_nouser_xattr, "nouser_xattr"},
181 {Opt_acl, "acl"},
182 {Opt_noacl, "noacl"},
183 {Opt_active_logs, "active_logs=%u"},
184 {Opt_disable_ext_identify, "disable_ext_identify"},
185 {Opt_inline_xattr, "inline_xattr"},
186 {Opt_noinline_xattr, "noinline_xattr"},
187 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
188 {Opt_inline_data, "inline_data"},
189 {Opt_inline_dentry, "inline_dentry"},
190 {Opt_noinline_dentry, "noinline_dentry"},
191 {Opt_flush_merge, "flush_merge"},
192 {Opt_noflush_merge, "noflush_merge"},
193 {Opt_barrier, "barrier"},
194 {Opt_nobarrier, "nobarrier"},
195 {Opt_fastboot, "fastboot"},
196 {Opt_extent_cache, "extent_cache"},
197 {Opt_noextent_cache, "noextent_cache"},
198 {Opt_noinline_data, "noinline_data"},
199 {Opt_data_flush, "data_flush"},
200 {Opt_reserve_root, "reserve_root=%u"},
201 {Opt_resgid, "resgid=%u"},
202 {Opt_resuid, "resuid=%u"},
203 {Opt_mode, "mode=%s"},
204 {Opt_io_size_bits, "io_bits=%u"},
205 {Opt_fault_injection, "fault_injection=%u"},
206 {Opt_fault_type, "fault_type=%u"},
207 {Opt_lazytime, "lazytime"},
208 {Opt_nolazytime, "nolazytime"},
209 {Opt_quota, "quota"},
210 {Opt_noquota, "noquota"},
211 {Opt_usrquota, "usrquota"},
212 {Opt_grpquota, "grpquota"},
213 {Opt_prjquota, "prjquota"},
214 {Opt_usrjquota, "usrjquota=%s"},
215 {Opt_grpjquota, "grpjquota=%s"},
216 {Opt_prjjquota, "prjjquota=%s"},
217 {Opt_offusrjquota, "usrjquota="},
218 {Opt_offgrpjquota, "grpjquota="},
219 {Opt_offprjjquota, "prjjquota="},
220 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
221 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
222 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
223 {Opt_alloc, "alloc_mode=%s"},
224 {Opt_fsync, "fsync_mode=%s"},
225 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
226 {Opt_test_dummy_encryption, "test_dummy_encryption"},
227 {Opt_inlinecrypt, "inlinecrypt"},
228 {Opt_checkpoint_disable, "checkpoint=disable"},
229 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
230 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
231 {Opt_checkpoint_enable, "checkpoint=enable"},
232 {Opt_checkpoint_merge, "checkpoint_merge"},
233 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
234 {Opt_compress_algorithm, "compress_algorithm=%s"},
235 {Opt_compress_log_size, "compress_log_size=%u"},
236 {Opt_compress_extension, "compress_extension=%s"},
237 {Opt_nocompress_extension, "nocompress_extension=%s"},
238 {Opt_compress_chksum, "compress_chksum"},
239 {Opt_compress_mode, "compress_mode=%s"},
240 {Opt_compress_cache, "compress_cache"},
241 {Opt_atgc, "atgc"},
242 {Opt_gc_merge, "gc_merge"},
243 {Opt_nogc_merge, "nogc_merge"},
244 {Opt_discard_unit, "discard_unit=%s"},
245 {Opt_memory_mode, "memory=%s"},
246 {Opt_age_extent_cache, "age_extent_cache"},
247 {Opt_errors, "errors=%s"},
248 {Opt_err, NULL},
249 };
250
f2fs_printk(struct f2fs_sb_info * sbi,const char * fmt,...)251 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
252 {
253 struct va_format vaf;
254 va_list args;
255 int level;
256
257 va_start(args, fmt);
258
259 level = printk_get_level(fmt);
260 vaf.fmt = printk_skip_level(fmt);
261 vaf.va = &args;
262 printk("%c%cF2FS-fs (%s): %pV\n",
263 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
264
265 va_end(args);
266 }
267
268 #if IS_ENABLED(CONFIG_UNICODE)
269 static const struct f2fs_sb_encodings {
270 __u16 magic;
271 char *name;
272 unsigned int version;
273 } f2fs_sb_encoding_map[] = {
274 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
275 };
276
277 static const struct f2fs_sb_encodings *
f2fs_sb_read_encoding(const struct f2fs_super_block * sb)278 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
279 {
280 __u16 magic = le16_to_cpu(sb->s_encoding);
281 int i;
282
283 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
284 if (magic == f2fs_sb_encoding_map[i].magic)
285 return &f2fs_sb_encoding_map[i];
286
287 return NULL;
288 }
289
290 struct kmem_cache *f2fs_cf_name_slab;
f2fs_create_casefold_cache(void)291 static int __init f2fs_create_casefold_cache(void)
292 {
293 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
294 F2FS_NAME_LEN);
295 return f2fs_cf_name_slab ? 0 : -ENOMEM;
296 }
297
f2fs_destroy_casefold_cache(void)298 static void f2fs_destroy_casefold_cache(void)
299 {
300 kmem_cache_destroy(f2fs_cf_name_slab);
301 }
302 #else
f2fs_create_casefold_cache(void)303 static int __init f2fs_create_casefold_cache(void) { return 0; }
f2fs_destroy_casefold_cache(void)304 static void f2fs_destroy_casefold_cache(void) { }
305 #endif
306
limit_reserve_root(struct f2fs_sb_info * sbi)307 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
308 {
309 block_t limit = min((sbi->user_block_count >> 3),
310 sbi->user_block_count - sbi->reserved_blocks);
311
312 /* limit is 12.5% */
313 if (test_opt(sbi, RESERVE_ROOT) &&
314 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
315 F2FS_OPTION(sbi).root_reserved_blocks = limit;
316 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
317 F2FS_OPTION(sbi).root_reserved_blocks);
318 }
319 if (!test_opt(sbi, RESERVE_ROOT) &&
320 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
321 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
322 !gid_eq(F2FS_OPTION(sbi).s_resgid,
323 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
324 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
325 from_kuid_munged(&init_user_ns,
326 F2FS_OPTION(sbi).s_resuid),
327 from_kgid_munged(&init_user_ns,
328 F2FS_OPTION(sbi).s_resgid));
329 }
330
adjust_reserved_segment(struct f2fs_sb_info * sbi)331 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
332 {
333 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
334 unsigned int avg_vblocks;
335 unsigned int wanted_reserved_segments;
336 block_t avail_user_block_count;
337
338 if (!F2FS_IO_ALIGNED(sbi))
339 return 0;
340
341 /* average valid block count in section in worst case */
342 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
343
344 /*
345 * we need enough free space when migrating one section in worst case
346 */
347 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
348 reserved_segments(sbi);
349 wanted_reserved_segments -= reserved_segments(sbi);
350
351 avail_user_block_count = sbi->user_block_count -
352 sbi->current_reserved_blocks -
353 F2FS_OPTION(sbi).root_reserved_blocks;
354
355 if (wanted_reserved_segments * sbi->blocks_per_seg >
356 avail_user_block_count) {
357 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
358 wanted_reserved_segments,
359 avail_user_block_count >> sbi->log_blocks_per_seg);
360 return -ENOSPC;
361 }
362
363 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
364
365 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
366 wanted_reserved_segments);
367
368 return 0;
369 }
370
adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)371 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
372 {
373 if (!F2FS_OPTION(sbi).unusable_cap_perc)
374 return;
375
376 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
377 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
378 else
379 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
380 F2FS_OPTION(sbi).unusable_cap_perc;
381
382 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
383 F2FS_OPTION(sbi).unusable_cap,
384 F2FS_OPTION(sbi).unusable_cap_perc);
385 }
386
init_once(void * foo)387 static void init_once(void *foo)
388 {
389 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
390
391 inode_init_once(&fi->vfs_inode);
392 }
393
394 #ifdef CONFIG_QUOTA
395 static const char * const quotatypes[] = INITQFNAMES;
396 #define QTYPE2NAME(t) (quotatypes[t])
f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)397 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
398 substring_t *args)
399 {
400 struct f2fs_sb_info *sbi = F2FS_SB(sb);
401 char *qname;
402 int ret = -EINVAL;
403
404 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
405 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
406 return -EINVAL;
407 }
408 if (f2fs_sb_has_quota_ino(sbi)) {
409 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
410 return 0;
411 }
412
413 qname = match_strdup(args);
414 if (!qname) {
415 f2fs_err(sbi, "Not enough memory for storing quotafile name");
416 return -ENOMEM;
417 }
418 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
419 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
420 ret = 0;
421 else
422 f2fs_err(sbi, "%s quota file already specified",
423 QTYPE2NAME(qtype));
424 goto errout;
425 }
426 if (strchr(qname, '/')) {
427 f2fs_err(sbi, "quotafile must be on filesystem root");
428 goto errout;
429 }
430 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
431 set_opt(sbi, QUOTA);
432 return 0;
433 errout:
434 kfree(qname);
435 return ret;
436 }
437
f2fs_clear_qf_name(struct super_block * sb,int qtype)438 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
439 {
440 struct f2fs_sb_info *sbi = F2FS_SB(sb);
441
442 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
443 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
444 return -EINVAL;
445 }
446 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
447 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
448 return 0;
449 }
450
f2fs_check_quota_options(struct f2fs_sb_info * sbi)451 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
452 {
453 /*
454 * We do the test below only for project quotas. 'usrquota' and
455 * 'grpquota' mount options are allowed even without quota feature
456 * to support legacy quotas in quota files.
457 */
458 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
459 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
460 return -1;
461 }
462 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
463 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
464 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
465 if (test_opt(sbi, USRQUOTA) &&
466 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
467 clear_opt(sbi, USRQUOTA);
468
469 if (test_opt(sbi, GRPQUOTA) &&
470 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
471 clear_opt(sbi, GRPQUOTA);
472
473 if (test_opt(sbi, PRJQUOTA) &&
474 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
475 clear_opt(sbi, PRJQUOTA);
476
477 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
478 test_opt(sbi, PRJQUOTA)) {
479 f2fs_err(sbi, "old and new quota format mixing");
480 return -1;
481 }
482
483 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
484 f2fs_err(sbi, "journaled quota format not specified");
485 return -1;
486 }
487 }
488
489 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
490 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
491 F2FS_OPTION(sbi).s_jquota_fmt = 0;
492 }
493 return 0;
494 }
495 #endif
496
f2fs_set_test_dummy_encryption(struct super_block * sb,const char * opt,const substring_t * arg,bool is_remount)497 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
498 const char *opt,
499 const substring_t *arg,
500 bool is_remount)
501 {
502 struct f2fs_sb_info *sbi = F2FS_SB(sb);
503 struct fs_parameter param = {
504 .type = fs_value_is_string,
505 .string = arg->from ? arg->from : "",
506 };
507 struct fscrypt_dummy_policy *policy =
508 &F2FS_OPTION(sbi).dummy_enc_policy;
509 int err;
510
511 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
512 f2fs_warn(sbi, "test_dummy_encryption option not supported");
513 return -EINVAL;
514 }
515
516 if (!f2fs_sb_has_encrypt(sbi)) {
517 f2fs_err(sbi, "Encrypt feature is off");
518 return -EINVAL;
519 }
520
521 /*
522 * This mount option is just for testing, and it's not worthwhile to
523 * implement the extra complexity (e.g. RCU protection) that would be
524 * needed to allow it to be set or changed during remount. We do allow
525 * it to be specified during remount, but only if there is no change.
526 */
527 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
528 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
529 return -EINVAL;
530 }
531
532 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
533 if (err) {
534 if (err == -EEXIST)
535 f2fs_warn(sbi,
536 "Can't change test_dummy_encryption on remount");
537 else if (err == -EINVAL)
538 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
539 opt);
540 else
541 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
542 opt, err);
543 return -EINVAL;
544 }
545 f2fs_warn(sbi, "Test dummy encryption mode enabled");
546 return 0;
547 }
548
549 #ifdef CONFIG_F2FS_FS_COMPRESSION
is_compress_extension_exist(struct f2fs_sb_info * sbi,const char * new_ext,bool is_ext)550 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
551 const char *new_ext, bool is_ext)
552 {
553 unsigned char (*ext)[F2FS_EXTENSION_LEN];
554 int ext_cnt;
555 int i;
556
557 if (is_ext) {
558 ext = F2FS_OPTION(sbi).extensions;
559 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
560 } else {
561 ext = F2FS_OPTION(sbi).noextensions;
562 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
563 }
564
565 for (i = 0; i < ext_cnt; i++) {
566 if (!strcasecmp(new_ext, ext[i]))
567 return true;
568 }
569
570 return false;
571 }
572
573 /*
574 * 1. The same extension name cannot not appear in both compress and non-compress extension
575 * at the same time.
576 * 2. If the compress extension specifies all files, the types specified by the non-compress
577 * extension will be treated as special cases and will not be compressed.
578 * 3. Don't allow the non-compress extension specifies all files.
579 */
f2fs_test_compress_extension(struct f2fs_sb_info * sbi)580 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
581 {
582 unsigned char (*ext)[F2FS_EXTENSION_LEN];
583 unsigned char (*noext)[F2FS_EXTENSION_LEN];
584 int ext_cnt, noext_cnt, index = 0, no_index = 0;
585
586 ext = F2FS_OPTION(sbi).extensions;
587 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
588 noext = F2FS_OPTION(sbi).noextensions;
589 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
590
591 if (!noext_cnt)
592 return 0;
593
594 for (no_index = 0; no_index < noext_cnt; no_index++) {
595 if (!strcasecmp("*", noext[no_index])) {
596 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
597 return -EINVAL;
598 }
599 for (index = 0; index < ext_cnt; index++) {
600 if (!strcasecmp(ext[index], noext[no_index])) {
601 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
602 ext[index]);
603 return -EINVAL;
604 }
605 }
606 }
607 return 0;
608 }
609
610 #ifdef CONFIG_F2FS_FS_LZ4
f2fs_set_lz4hc_level(struct f2fs_sb_info * sbi,const char * str)611 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
612 {
613 #ifdef CONFIG_F2FS_FS_LZ4HC
614 unsigned int level;
615
616 if (strlen(str) == 3) {
617 F2FS_OPTION(sbi).compress_level = 0;
618 return 0;
619 }
620
621 str += 3;
622
623 if (str[0] != ':') {
624 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
625 return -EINVAL;
626 }
627 if (kstrtouint(str + 1, 10, &level))
628 return -EINVAL;
629
630 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
631 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
632 return -EINVAL;
633 }
634
635 F2FS_OPTION(sbi).compress_level = level;
636 return 0;
637 #else
638 if (strlen(str) == 3) {
639 F2FS_OPTION(sbi).compress_level = 0;
640 return 0;
641 }
642 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
643 return -EINVAL;
644 #endif
645 }
646 #endif
647
648 #ifdef CONFIG_F2FS_FS_ZSTD
f2fs_set_zstd_level(struct f2fs_sb_info * sbi,const char * str)649 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
650 {
651 int level;
652 int len = 4;
653
654 if (strlen(str) == len) {
655 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
656 return 0;
657 }
658
659 str += len;
660
661 if (str[0] != ':') {
662 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
663 return -EINVAL;
664 }
665 if (kstrtoint(str + 1, 10, &level))
666 return -EINVAL;
667
668 /* f2fs does not support negative compress level now */
669 if (level < 0) {
670 f2fs_info(sbi, "do not support negative compress level: %d", level);
671 return -ERANGE;
672 }
673
674 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
675 f2fs_info(sbi, "invalid zstd compress level: %d", level);
676 return -EINVAL;
677 }
678
679 F2FS_OPTION(sbi).compress_level = level;
680 return 0;
681 }
682 #endif
683 #endif
684
parse_options(struct super_block * sb,char * options,bool is_remount)685 static int parse_options(struct super_block *sb, char *options, bool is_remount)
686 {
687 struct f2fs_sb_info *sbi = F2FS_SB(sb);
688 substring_t args[MAX_OPT_ARGS];
689 #ifdef CONFIG_F2FS_FS_COMPRESSION
690 unsigned char (*ext)[F2FS_EXTENSION_LEN];
691 unsigned char (*noext)[F2FS_EXTENSION_LEN];
692 int ext_cnt, noext_cnt;
693 #endif
694 char *p, *name;
695 int arg = 0;
696 kuid_t uid;
697 kgid_t gid;
698 int ret;
699
700 if (!options)
701 goto default_check;
702
703 while ((p = strsep(&options, ",")) != NULL) {
704 int token;
705
706 if (!*p)
707 continue;
708 /*
709 * Initialize args struct so we know whether arg was
710 * found; some options take optional arguments.
711 */
712 args[0].to = args[0].from = NULL;
713 token = match_token(p, f2fs_tokens, args);
714
715 switch (token) {
716 case Opt_gc_background:
717 name = match_strdup(&args[0]);
718
719 if (!name)
720 return -ENOMEM;
721 if (!strcmp(name, "on")) {
722 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
723 } else if (!strcmp(name, "off")) {
724 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
725 } else if (!strcmp(name, "sync")) {
726 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
727 } else {
728 kfree(name);
729 return -EINVAL;
730 }
731 kfree(name);
732 break;
733 case Opt_disable_roll_forward:
734 set_opt(sbi, DISABLE_ROLL_FORWARD);
735 break;
736 case Opt_norecovery:
737 /* this option mounts f2fs with ro */
738 set_opt(sbi, NORECOVERY);
739 if (!f2fs_readonly(sb))
740 return -EINVAL;
741 break;
742 case Opt_discard:
743 if (!f2fs_hw_support_discard(sbi)) {
744 f2fs_warn(sbi, "device does not support discard");
745 break;
746 }
747 set_opt(sbi, DISCARD);
748 break;
749 case Opt_nodiscard:
750 if (f2fs_hw_should_discard(sbi)) {
751 f2fs_warn(sbi, "discard is required for zoned block devices");
752 return -EINVAL;
753 }
754 clear_opt(sbi, DISCARD);
755 break;
756 case Opt_noheap:
757 set_opt(sbi, NOHEAP);
758 break;
759 case Opt_heap:
760 clear_opt(sbi, NOHEAP);
761 break;
762 #ifdef CONFIG_F2FS_FS_XATTR
763 case Opt_user_xattr:
764 set_opt(sbi, XATTR_USER);
765 break;
766 case Opt_nouser_xattr:
767 clear_opt(sbi, XATTR_USER);
768 break;
769 case Opt_inline_xattr:
770 set_opt(sbi, INLINE_XATTR);
771 break;
772 case Opt_noinline_xattr:
773 clear_opt(sbi, INLINE_XATTR);
774 break;
775 case Opt_inline_xattr_size:
776 if (args->from && match_int(args, &arg))
777 return -EINVAL;
778 set_opt(sbi, INLINE_XATTR_SIZE);
779 F2FS_OPTION(sbi).inline_xattr_size = arg;
780 break;
781 #else
782 case Opt_user_xattr:
783 f2fs_info(sbi, "user_xattr options not supported");
784 break;
785 case Opt_nouser_xattr:
786 f2fs_info(sbi, "nouser_xattr options not supported");
787 break;
788 case Opt_inline_xattr:
789 f2fs_info(sbi, "inline_xattr options not supported");
790 break;
791 case Opt_noinline_xattr:
792 f2fs_info(sbi, "noinline_xattr options not supported");
793 break;
794 #endif
795 #ifdef CONFIG_F2FS_FS_POSIX_ACL
796 case Opt_acl:
797 set_opt(sbi, POSIX_ACL);
798 break;
799 case Opt_noacl:
800 clear_opt(sbi, POSIX_ACL);
801 break;
802 #else
803 case Opt_acl:
804 f2fs_info(sbi, "acl options not supported");
805 break;
806 case Opt_noacl:
807 f2fs_info(sbi, "noacl options not supported");
808 break;
809 #endif
810 case Opt_active_logs:
811 if (args->from && match_int(args, &arg))
812 return -EINVAL;
813 if (arg != 2 && arg != 4 &&
814 arg != NR_CURSEG_PERSIST_TYPE)
815 return -EINVAL;
816 F2FS_OPTION(sbi).active_logs = arg;
817 break;
818 case Opt_disable_ext_identify:
819 set_opt(sbi, DISABLE_EXT_IDENTIFY);
820 break;
821 case Opt_inline_data:
822 set_opt(sbi, INLINE_DATA);
823 break;
824 case Opt_inline_dentry:
825 set_opt(sbi, INLINE_DENTRY);
826 break;
827 case Opt_noinline_dentry:
828 clear_opt(sbi, INLINE_DENTRY);
829 break;
830 case Opt_flush_merge:
831 set_opt(sbi, FLUSH_MERGE);
832 break;
833 case Opt_noflush_merge:
834 clear_opt(sbi, FLUSH_MERGE);
835 break;
836 case Opt_nobarrier:
837 set_opt(sbi, NOBARRIER);
838 break;
839 case Opt_barrier:
840 clear_opt(sbi, NOBARRIER);
841 break;
842 case Opt_fastboot:
843 set_opt(sbi, FASTBOOT);
844 break;
845 case Opt_extent_cache:
846 set_opt(sbi, READ_EXTENT_CACHE);
847 break;
848 case Opt_noextent_cache:
849 clear_opt(sbi, READ_EXTENT_CACHE);
850 break;
851 case Opt_noinline_data:
852 clear_opt(sbi, INLINE_DATA);
853 break;
854 case Opt_data_flush:
855 set_opt(sbi, DATA_FLUSH);
856 break;
857 case Opt_reserve_root:
858 if (args->from && match_int(args, &arg))
859 return -EINVAL;
860 if (test_opt(sbi, RESERVE_ROOT)) {
861 f2fs_info(sbi, "Preserve previous reserve_root=%u",
862 F2FS_OPTION(sbi).root_reserved_blocks);
863 } else {
864 F2FS_OPTION(sbi).root_reserved_blocks = arg;
865 set_opt(sbi, RESERVE_ROOT);
866 }
867 break;
868 case Opt_resuid:
869 if (args->from && match_int(args, &arg))
870 return -EINVAL;
871 uid = make_kuid(current_user_ns(), arg);
872 if (!uid_valid(uid)) {
873 f2fs_err(sbi, "Invalid uid value %d", arg);
874 return -EINVAL;
875 }
876 F2FS_OPTION(sbi).s_resuid = uid;
877 break;
878 case Opt_resgid:
879 if (args->from && match_int(args, &arg))
880 return -EINVAL;
881 gid = make_kgid(current_user_ns(), arg);
882 if (!gid_valid(gid)) {
883 f2fs_err(sbi, "Invalid gid value %d", arg);
884 return -EINVAL;
885 }
886 F2FS_OPTION(sbi).s_resgid = gid;
887 break;
888 case Opt_mode:
889 name = match_strdup(&args[0]);
890
891 if (!name)
892 return -ENOMEM;
893 if (!strcmp(name, "adaptive")) {
894 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
895 } else if (!strcmp(name, "lfs")) {
896 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
897 } else if (!strcmp(name, "fragment:segment")) {
898 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
899 } else if (!strcmp(name, "fragment:block")) {
900 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
901 } else {
902 kfree(name);
903 return -EINVAL;
904 }
905 kfree(name);
906 break;
907 case Opt_io_size_bits:
908 if (args->from && match_int(args, &arg))
909 return -EINVAL;
910 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
911 f2fs_warn(sbi, "Not support %ld, larger than %d",
912 BIT(arg), BIO_MAX_VECS);
913 return -EINVAL;
914 }
915 F2FS_OPTION(sbi).write_io_size_bits = arg;
916 break;
917 #ifdef CONFIG_F2FS_FAULT_INJECTION
918 case Opt_fault_injection:
919 if (args->from && match_int(args, &arg))
920 return -EINVAL;
921 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
922 set_opt(sbi, FAULT_INJECTION);
923 break;
924
925 case Opt_fault_type:
926 if (args->from && match_int(args, &arg))
927 return -EINVAL;
928 f2fs_build_fault_attr(sbi, 0, arg);
929 set_opt(sbi, FAULT_INJECTION);
930 break;
931 #else
932 case Opt_fault_injection:
933 f2fs_info(sbi, "fault_injection options not supported");
934 break;
935
936 case Opt_fault_type:
937 f2fs_info(sbi, "fault_type options not supported");
938 break;
939 #endif
940 case Opt_lazytime:
941 sb->s_flags |= SB_LAZYTIME;
942 break;
943 case Opt_nolazytime:
944 sb->s_flags &= ~SB_LAZYTIME;
945 break;
946 #ifdef CONFIG_QUOTA
947 case Opt_quota:
948 case Opt_usrquota:
949 set_opt(sbi, USRQUOTA);
950 break;
951 case Opt_grpquota:
952 set_opt(sbi, GRPQUOTA);
953 break;
954 case Opt_prjquota:
955 set_opt(sbi, PRJQUOTA);
956 break;
957 case Opt_usrjquota:
958 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
959 if (ret)
960 return ret;
961 break;
962 case Opt_grpjquota:
963 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
964 if (ret)
965 return ret;
966 break;
967 case Opt_prjjquota:
968 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
969 if (ret)
970 return ret;
971 break;
972 case Opt_offusrjquota:
973 ret = f2fs_clear_qf_name(sb, USRQUOTA);
974 if (ret)
975 return ret;
976 break;
977 case Opt_offgrpjquota:
978 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
979 if (ret)
980 return ret;
981 break;
982 case Opt_offprjjquota:
983 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
984 if (ret)
985 return ret;
986 break;
987 case Opt_jqfmt_vfsold:
988 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
989 break;
990 case Opt_jqfmt_vfsv0:
991 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
992 break;
993 case Opt_jqfmt_vfsv1:
994 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
995 break;
996 case Opt_noquota:
997 clear_opt(sbi, QUOTA);
998 clear_opt(sbi, USRQUOTA);
999 clear_opt(sbi, GRPQUOTA);
1000 clear_opt(sbi, PRJQUOTA);
1001 break;
1002 #else
1003 case Opt_quota:
1004 case Opt_usrquota:
1005 case Opt_grpquota:
1006 case Opt_prjquota:
1007 case Opt_usrjquota:
1008 case Opt_grpjquota:
1009 case Opt_prjjquota:
1010 case Opt_offusrjquota:
1011 case Opt_offgrpjquota:
1012 case Opt_offprjjquota:
1013 case Opt_jqfmt_vfsold:
1014 case Opt_jqfmt_vfsv0:
1015 case Opt_jqfmt_vfsv1:
1016 case Opt_noquota:
1017 f2fs_info(sbi, "quota operations not supported");
1018 break;
1019 #endif
1020 case Opt_alloc:
1021 name = match_strdup(&args[0]);
1022 if (!name)
1023 return -ENOMEM;
1024
1025 if (!strcmp(name, "default")) {
1026 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1027 } else if (!strcmp(name, "reuse")) {
1028 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1029 } else {
1030 kfree(name);
1031 return -EINVAL;
1032 }
1033 kfree(name);
1034 break;
1035 case Opt_fsync:
1036 name = match_strdup(&args[0]);
1037 if (!name)
1038 return -ENOMEM;
1039 if (!strcmp(name, "posix")) {
1040 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1041 } else if (!strcmp(name, "strict")) {
1042 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1043 } else if (!strcmp(name, "nobarrier")) {
1044 F2FS_OPTION(sbi).fsync_mode =
1045 FSYNC_MODE_NOBARRIER;
1046 } else {
1047 kfree(name);
1048 return -EINVAL;
1049 }
1050 kfree(name);
1051 break;
1052 case Opt_test_dummy_encryption:
1053 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1054 is_remount);
1055 if (ret)
1056 return ret;
1057 break;
1058 case Opt_inlinecrypt:
1059 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1060 sb->s_flags |= SB_INLINECRYPT;
1061 #else
1062 f2fs_info(sbi, "inline encryption not supported");
1063 #endif
1064 break;
1065 case Opt_checkpoint_disable_cap_perc:
1066 if (args->from && match_int(args, &arg))
1067 return -EINVAL;
1068 if (arg < 0 || arg > 100)
1069 return -EINVAL;
1070 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1071 set_opt(sbi, DISABLE_CHECKPOINT);
1072 break;
1073 case Opt_checkpoint_disable_cap:
1074 if (args->from && match_int(args, &arg))
1075 return -EINVAL;
1076 F2FS_OPTION(sbi).unusable_cap = arg;
1077 set_opt(sbi, DISABLE_CHECKPOINT);
1078 break;
1079 case Opt_checkpoint_disable:
1080 set_opt(sbi, DISABLE_CHECKPOINT);
1081 break;
1082 case Opt_checkpoint_enable:
1083 clear_opt(sbi, DISABLE_CHECKPOINT);
1084 break;
1085 case Opt_checkpoint_merge:
1086 set_opt(sbi, MERGE_CHECKPOINT);
1087 break;
1088 case Opt_nocheckpoint_merge:
1089 clear_opt(sbi, MERGE_CHECKPOINT);
1090 break;
1091 #ifdef CONFIG_F2FS_FS_COMPRESSION
1092 case Opt_compress_algorithm:
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 if (!strcmp(name, "lzo")) {
1101 #ifdef CONFIG_F2FS_FS_LZO
1102 F2FS_OPTION(sbi).compress_level = 0;
1103 F2FS_OPTION(sbi).compress_algorithm =
1104 COMPRESS_LZO;
1105 #else
1106 f2fs_info(sbi, "kernel doesn't support lzo compression");
1107 #endif
1108 } else if (!strncmp(name, "lz4", 3)) {
1109 #ifdef CONFIG_F2FS_FS_LZ4
1110 ret = f2fs_set_lz4hc_level(sbi, name);
1111 if (ret) {
1112 kfree(name);
1113 return -EINVAL;
1114 }
1115 F2FS_OPTION(sbi).compress_algorithm =
1116 COMPRESS_LZ4;
1117 #else
1118 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1119 #endif
1120 } else if (!strncmp(name, "zstd", 4)) {
1121 #ifdef CONFIG_F2FS_FS_ZSTD
1122 ret = f2fs_set_zstd_level(sbi, name);
1123 if (ret) {
1124 kfree(name);
1125 return -EINVAL;
1126 }
1127 F2FS_OPTION(sbi).compress_algorithm =
1128 COMPRESS_ZSTD;
1129 #else
1130 f2fs_info(sbi, "kernel doesn't support zstd compression");
1131 #endif
1132 } else if (!strcmp(name, "lzo-rle")) {
1133 #ifdef CONFIG_F2FS_FS_LZORLE
1134 F2FS_OPTION(sbi).compress_level = 0;
1135 F2FS_OPTION(sbi).compress_algorithm =
1136 COMPRESS_LZORLE;
1137 #else
1138 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1139 #endif
1140 } else {
1141 kfree(name);
1142 return -EINVAL;
1143 }
1144 kfree(name);
1145 break;
1146 case Opt_compress_log_size:
1147 if (!f2fs_sb_has_compression(sbi)) {
1148 f2fs_info(sbi, "Image doesn't support compression");
1149 break;
1150 }
1151 if (args->from && match_int(args, &arg))
1152 return -EINVAL;
1153 if (arg < MIN_COMPRESS_LOG_SIZE ||
1154 arg > MAX_COMPRESS_LOG_SIZE) {
1155 f2fs_err(sbi,
1156 "Compress cluster log size is out of range");
1157 return -EINVAL;
1158 }
1159 F2FS_OPTION(sbi).compress_log_size = arg;
1160 break;
1161 case Opt_compress_extension:
1162 if (!f2fs_sb_has_compression(sbi)) {
1163 f2fs_info(sbi, "Image doesn't support compression");
1164 break;
1165 }
1166 name = match_strdup(&args[0]);
1167 if (!name)
1168 return -ENOMEM;
1169
1170 ext = F2FS_OPTION(sbi).extensions;
1171 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1172
1173 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1174 ext_cnt >= COMPRESS_EXT_NUM) {
1175 f2fs_err(sbi,
1176 "invalid extension length/number");
1177 kfree(name);
1178 return -EINVAL;
1179 }
1180
1181 if (is_compress_extension_exist(sbi, name, true)) {
1182 kfree(name);
1183 break;
1184 }
1185
1186 strcpy(ext[ext_cnt], name);
1187 F2FS_OPTION(sbi).compress_ext_cnt++;
1188 kfree(name);
1189 break;
1190 case Opt_nocompress_extension:
1191 if (!f2fs_sb_has_compression(sbi)) {
1192 f2fs_info(sbi, "Image doesn't support compression");
1193 break;
1194 }
1195 name = match_strdup(&args[0]);
1196 if (!name)
1197 return -ENOMEM;
1198
1199 noext = F2FS_OPTION(sbi).noextensions;
1200 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1201
1202 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1203 noext_cnt >= COMPRESS_EXT_NUM) {
1204 f2fs_err(sbi,
1205 "invalid extension length/number");
1206 kfree(name);
1207 return -EINVAL;
1208 }
1209
1210 if (is_compress_extension_exist(sbi, name, false)) {
1211 kfree(name);
1212 break;
1213 }
1214
1215 strcpy(noext[noext_cnt], name);
1216 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1217 kfree(name);
1218 break;
1219 case Opt_compress_chksum:
1220 if (!f2fs_sb_has_compression(sbi)) {
1221 f2fs_info(sbi, "Image doesn't support compression");
1222 break;
1223 }
1224 F2FS_OPTION(sbi).compress_chksum = true;
1225 break;
1226 case Opt_compress_mode:
1227 if (!f2fs_sb_has_compression(sbi)) {
1228 f2fs_info(sbi, "Image doesn't support compression");
1229 break;
1230 }
1231 name = match_strdup(&args[0]);
1232 if (!name)
1233 return -ENOMEM;
1234 if (!strcmp(name, "fs")) {
1235 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1236 } else if (!strcmp(name, "user")) {
1237 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1238 } else {
1239 kfree(name);
1240 return -EINVAL;
1241 }
1242 kfree(name);
1243 break;
1244 case Opt_compress_cache:
1245 if (!f2fs_sb_has_compression(sbi)) {
1246 f2fs_info(sbi, "Image doesn't support compression");
1247 break;
1248 }
1249 set_opt(sbi, COMPRESS_CACHE);
1250 break;
1251 #else
1252 case Opt_compress_algorithm:
1253 case Opt_compress_log_size:
1254 case Opt_compress_extension:
1255 case Opt_nocompress_extension:
1256 case Opt_compress_chksum:
1257 case Opt_compress_mode:
1258 case Opt_compress_cache:
1259 f2fs_info(sbi, "compression options not supported");
1260 break;
1261 #endif
1262 case Opt_atgc:
1263 set_opt(sbi, ATGC);
1264 break;
1265 case Opt_gc_merge:
1266 set_opt(sbi, GC_MERGE);
1267 break;
1268 case Opt_nogc_merge:
1269 clear_opt(sbi, GC_MERGE);
1270 break;
1271 case Opt_discard_unit:
1272 name = match_strdup(&args[0]);
1273 if (!name)
1274 return -ENOMEM;
1275 if (!strcmp(name, "block")) {
1276 F2FS_OPTION(sbi).discard_unit =
1277 DISCARD_UNIT_BLOCK;
1278 } else if (!strcmp(name, "segment")) {
1279 F2FS_OPTION(sbi).discard_unit =
1280 DISCARD_UNIT_SEGMENT;
1281 } else if (!strcmp(name, "section")) {
1282 F2FS_OPTION(sbi).discard_unit =
1283 DISCARD_UNIT_SECTION;
1284 } else {
1285 kfree(name);
1286 return -EINVAL;
1287 }
1288 kfree(name);
1289 break;
1290 case Opt_memory_mode:
1291 name = match_strdup(&args[0]);
1292 if (!name)
1293 return -ENOMEM;
1294 if (!strcmp(name, "normal")) {
1295 F2FS_OPTION(sbi).memory_mode =
1296 MEMORY_MODE_NORMAL;
1297 } else if (!strcmp(name, "low")) {
1298 F2FS_OPTION(sbi).memory_mode =
1299 MEMORY_MODE_LOW;
1300 } else {
1301 kfree(name);
1302 return -EINVAL;
1303 }
1304 kfree(name);
1305 break;
1306 case Opt_age_extent_cache:
1307 set_opt(sbi, AGE_EXTENT_CACHE);
1308 break;
1309 case Opt_errors:
1310 name = match_strdup(&args[0]);
1311 if (!name)
1312 return -ENOMEM;
1313 if (!strcmp(name, "remount-ro")) {
1314 F2FS_OPTION(sbi).errors =
1315 MOUNT_ERRORS_READONLY;
1316 } else if (!strcmp(name, "continue")) {
1317 F2FS_OPTION(sbi).errors =
1318 MOUNT_ERRORS_CONTINUE;
1319 } else if (!strcmp(name, "panic")) {
1320 F2FS_OPTION(sbi).errors =
1321 MOUNT_ERRORS_PANIC;
1322 } else {
1323 kfree(name);
1324 return -EINVAL;
1325 }
1326 kfree(name);
1327 break;
1328 default:
1329 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1330 p);
1331 return -EINVAL;
1332 }
1333 }
1334 default_check:
1335 #ifdef CONFIG_QUOTA
1336 if (f2fs_check_quota_options(sbi))
1337 return -EINVAL;
1338 #else
1339 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1340 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1341 return -EINVAL;
1342 }
1343 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1344 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1345 return -EINVAL;
1346 }
1347 #endif
1348 #if !IS_ENABLED(CONFIG_UNICODE)
1349 if (f2fs_sb_has_casefold(sbi)) {
1350 f2fs_err(sbi,
1351 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1352 return -EINVAL;
1353 }
1354 #endif
1355 /*
1356 * The BLKZONED feature indicates that the drive was formatted with
1357 * zone alignment optimization. This is optional for host-aware
1358 * devices, but mandatory for host-managed zoned block devices.
1359 */
1360 if (f2fs_sb_has_blkzoned(sbi)) {
1361 #ifdef CONFIG_BLK_DEV_ZONED
1362 if (F2FS_OPTION(sbi).discard_unit !=
1363 DISCARD_UNIT_SECTION) {
1364 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1365 F2FS_OPTION(sbi).discard_unit =
1366 DISCARD_UNIT_SECTION;
1367 }
1368
1369 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1370 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1371 return -EINVAL;
1372 }
1373 #else
1374 f2fs_err(sbi, "Zoned block device support is not enabled");
1375 return -EINVAL;
1376 #endif
1377 }
1378
1379 #ifdef CONFIG_F2FS_FS_COMPRESSION
1380 if (f2fs_test_compress_extension(sbi)) {
1381 f2fs_err(sbi, "invalid compress or nocompress extension");
1382 return -EINVAL;
1383 }
1384 #endif
1385
1386 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1387 f2fs_err(sbi, "Should set mode=lfs with %luKB-sized IO",
1388 F2FS_IO_SIZE_KB(sbi));
1389 return -EINVAL;
1390 }
1391
1392 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1393 int min_size, max_size;
1394
1395 if (!f2fs_sb_has_extra_attr(sbi) ||
1396 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1397 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1398 return -EINVAL;
1399 }
1400 if (!test_opt(sbi, INLINE_XATTR)) {
1401 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1402 return -EINVAL;
1403 }
1404
1405 min_size = MIN_INLINE_XATTR_SIZE;
1406 max_size = MAX_INLINE_XATTR_SIZE;
1407
1408 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1409 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1410 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1411 min_size, max_size);
1412 return -EINVAL;
1413 }
1414 }
1415
1416 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1417 f2fs_err(sbi, "LFS is not compatible with checkpoint=disable");
1418 return -EINVAL;
1419 }
1420
1421 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1422 f2fs_err(sbi, "LFS is not compatible with ATGC");
1423 return -EINVAL;
1424 }
1425
1426 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1427 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1428 return -EINVAL;
1429 }
1430
1431 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1432 f2fs_err(sbi, "Allow to mount readonly mode only");
1433 return -EROFS;
1434 }
1435 return 0;
1436 }
1437
f2fs_alloc_inode(struct super_block * sb)1438 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1439 {
1440 struct f2fs_inode_info *fi;
1441
1442 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1443 return NULL;
1444
1445 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1446 if (!fi)
1447 return NULL;
1448
1449 init_once((void *) fi);
1450
1451 /* Initialize f2fs-specific inode info */
1452 atomic_set(&fi->dirty_pages, 0);
1453 atomic_set(&fi->i_compr_blocks, 0);
1454 init_f2fs_rwsem(&fi->i_sem);
1455 spin_lock_init(&fi->i_size_lock);
1456 INIT_LIST_HEAD(&fi->dirty_list);
1457 INIT_LIST_HEAD(&fi->gdirty_list);
1458 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1459 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1460 init_f2fs_rwsem(&fi->i_xattr_sem);
1461
1462 /* Will be used by directory only */
1463 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1464
1465 return &fi->vfs_inode;
1466 }
1467
f2fs_drop_inode(struct inode * inode)1468 static int f2fs_drop_inode(struct inode *inode)
1469 {
1470 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1471 int ret;
1472
1473 /*
1474 * during filesystem shutdown, if checkpoint is disabled,
1475 * drop useless meta/node dirty pages.
1476 */
1477 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1478 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1479 inode->i_ino == F2FS_META_INO(sbi)) {
1480 trace_f2fs_drop_inode(inode, 1);
1481 return 1;
1482 }
1483 }
1484
1485 /*
1486 * This is to avoid a deadlock condition like below.
1487 * writeback_single_inode(inode)
1488 * - f2fs_write_data_page
1489 * - f2fs_gc -> iput -> evict
1490 * - inode_wait_for_writeback(inode)
1491 */
1492 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1493 if (!inode->i_nlink && !is_bad_inode(inode)) {
1494 /* to avoid evict_inode call simultaneously */
1495 atomic_inc(&inode->i_count);
1496 spin_unlock(&inode->i_lock);
1497
1498 /* should remain fi->extent_tree for writepage */
1499 f2fs_destroy_extent_node(inode);
1500
1501 sb_start_intwrite(inode->i_sb);
1502 f2fs_i_size_write(inode, 0);
1503
1504 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1505 inode, NULL, 0, DATA);
1506 truncate_inode_pages_final(inode->i_mapping);
1507
1508 if (F2FS_HAS_BLOCKS(inode))
1509 f2fs_truncate(inode);
1510
1511 sb_end_intwrite(inode->i_sb);
1512
1513 spin_lock(&inode->i_lock);
1514 atomic_dec(&inode->i_count);
1515 }
1516 trace_f2fs_drop_inode(inode, 0);
1517 return 0;
1518 }
1519 ret = generic_drop_inode(inode);
1520 if (!ret)
1521 ret = fscrypt_drop_inode(inode);
1522 trace_f2fs_drop_inode(inode, ret);
1523 return ret;
1524 }
1525
f2fs_inode_dirtied(struct inode * inode,bool sync)1526 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1527 {
1528 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1529 int ret = 0;
1530
1531 spin_lock(&sbi->inode_lock[DIRTY_META]);
1532 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1533 ret = 1;
1534 } else {
1535 set_inode_flag(inode, FI_DIRTY_INODE);
1536 stat_inc_dirty_inode(sbi, DIRTY_META);
1537 }
1538 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1539 list_add_tail(&F2FS_I(inode)->gdirty_list,
1540 &sbi->inode_list[DIRTY_META]);
1541 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1542 }
1543 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1544 return ret;
1545 }
1546
f2fs_inode_synced(struct inode * inode)1547 void f2fs_inode_synced(struct inode *inode)
1548 {
1549 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1550
1551 spin_lock(&sbi->inode_lock[DIRTY_META]);
1552 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1553 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1554 return;
1555 }
1556 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1557 list_del_init(&F2FS_I(inode)->gdirty_list);
1558 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1559 }
1560 clear_inode_flag(inode, FI_DIRTY_INODE);
1561 clear_inode_flag(inode, FI_AUTO_RECOVER);
1562 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1563 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1564 }
1565
1566 /*
1567 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1568 *
1569 * We should call set_dirty_inode to write the dirty inode through write_inode.
1570 */
f2fs_dirty_inode(struct inode * inode,int flags)1571 static void f2fs_dirty_inode(struct inode *inode, int flags)
1572 {
1573 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1574
1575 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1576 inode->i_ino == F2FS_META_INO(sbi))
1577 return;
1578
1579 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1580 clear_inode_flag(inode, FI_AUTO_RECOVER);
1581
1582 f2fs_inode_dirtied(inode, false);
1583 }
1584
f2fs_free_inode(struct inode * inode)1585 static void f2fs_free_inode(struct inode *inode)
1586 {
1587 fscrypt_free_inode(inode);
1588 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1589 }
1590
destroy_percpu_info(struct f2fs_sb_info * sbi)1591 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1592 {
1593 percpu_counter_destroy(&sbi->total_valid_inode_count);
1594 percpu_counter_destroy(&sbi->rf_node_block_count);
1595 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1596 }
1597
destroy_device_list(struct f2fs_sb_info * sbi)1598 static void destroy_device_list(struct f2fs_sb_info *sbi)
1599 {
1600 int i;
1601
1602 for (i = 0; i < sbi->s_ndevs; i++) {
1603 if (i > 0)
1604 blkdev_put(FDEV(i).bdev, sbi->sb);
1605 #ifdef CONFIG_BLK_DEV_ZONED
1606 kvfree(FDEV(i).blkz_seq);
1607 #endif
1608 }
1609 kvfree(sbi->devs);
1610 }
1611
f2fs_put_super(struct super_block * sb)1612 static void f2fs_put_super(struct super_block *sb)
1613 {
1614 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1615 int i;
1616 int err = 0;
1617 bool done;
1618
1619 /* unregister procfs/sysfs entries in advance to avoid race case */
1620 f2fs_unregister_sysfs(sbi);
1621
1622 f2fs_quota_off_umount(sb);
1623
1624 /* prevent remaining shrinker jobs */
1625 mutex_lock(&sbi->umount_mutex);
1626
1627 /*
1628 * flush all issued checkpoints and stop checkpoint issue thread.
1629 * after then, all checkpoints should be done by each process context.
1630 */
1631 f2fs_stop_ckpt_thread(sbi);
1632
1633 /*
1634 * We don't need to do checkpoint when superblock is clean.
1635 * But, the previous checkpoint was not done by umount, it needs to do
1636 * clean checkpoint again.
1637 */
1638 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1639 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1640 struct cp_control cpc = {
1641 .reason = CP_UMOUNT,
1642 };
1643 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1644 err = f2fs_write_checkpoint(sbi, &cpc);
1645 }
1646
1647 /* be sure to wait for any on-going discard commands */
1648 done = f2fs_issue_discard_timeout(sbi);
1649 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1650 struct cp_control cpc = {
1651 .reason = CP_UMOUNT | CP_TRIMMED,
1652 };
1653 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1654 err = f2fs_write_checkpoint(sbi, &cpc);
1655 }
1656
1657 /*
1658 * normally superblock is clean, so we need to release this.
1659 * In addition, EIO will skip do checkpoint, we need this as well.
1660 */
1661 f2fs_release_ino_entry(sbi, true);
1662
1663 f2fs_leave_shrinker(sbi);
1664 mutex_unlock(&sbi->umount_mutex);
1665
1666 /* our cp_error case, we can wait for any writeback page */
1667 f2fs_flush_merged_writes(sbi);
1668
1669 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1670
1671 if (err || f2fs_cp_error(sbi)) {
1672 truncate_inode_pages_final(NODE_MAPPING(sbi));
1673 truncate_inode_pages_final(META_MAPPING(sbi));
1674 }
1675
1676 for (i = 0; i < NR_COUNT_TYPE; i++) {
1677 if (!get_pages(sbi, i))
1678 continue;
1679 f2fs_err(sbi, "detect filesystem reference count leak during "
1680 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1681 f2fs_bug_on(sbi, 1);
1682 }
1683
1684 f2fs_bug_on(sbi, sbi->fsync_node_num);
1685
1686 f2fs_destroy_compress_inode(sbi);
1687
1688 iput(sbi->node_inode);
1689 sbi->node_inode = NULL;
1690
1691 iput(sbi->meta_inode);
1692 sbi->meta_inode = NULL;
1693
1694 /*
1695 * iput() can update stat information, if f2fs_write_checkpoint()
1696 * above failed with error.
1697 */
1698 f2fs_destroy_stats(sbi);
1699
1700 /* destroy f2fs internal modules */
1701 f2fs_destroy_node_manager(sbi);
1702 f2fs_destroy_segment_manager(sbi);
1703
1704 /* flush s_error_work before sbi destroy */
1705 flush_work(&sbi->s_error_work);
1706
1707 f2fs_destroy_post_read_wq(sbi);
1708
1709 kvfree(sbi->ckpt);
1710
1711 sb->s_fs_info = NULL;
1712 if (sbi->s_chksum_driver)
1713 crypto_free_shash(sbi->s_chksum_driver);
1714 kfree(sbi->raw_super);
1715
1716 destroy_device_list(sbi);
1717 f2fs_destroy_page_array_cache(sbi);
1718 f2fs_destroy_xattr_caches(sbi);
1719 mempool_destroy(sbi->write_io_dummy);
1720 #ifdef CONFIG_QUOTA
1721 for (i = 0; i < MAXQUOTAS; i++)
1722 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1723 #endif
1724 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1725 destroy_percpu_info(sbi);
1726 f2fs_destroy_iostat(sbi);
1727 for (i = 0; i < NR_PAGE_TYPE; i++)
1728 kvfree(sbi->write_io[i]);
1729 #if IS_ENABLED(CONFIG_UNICODE)
1730 utf8_unload(sb->s_encoding);
1731 #endif
1732 kfree(sbi);
1733 }
1734
f2fs_sync_fs(struct super_block * sb,int sync)1735 int f2fs_sync_fs(struct super_block *sb, int sync)
1736 {
1737 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1738 int err = 0;
1739
1740 if (unlikely(f2fs_cp_error(sbi)))
1741 return 0;
1742 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1743 return 0;
1744
1745 trace_f2fs_sync_fs(sb, sync);
1746
1747 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1748 return -EAGAIN;
1749
1750 if (sync) {
1751 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1752 err = f2fs_issue_checkpoint(sbi);
1753 }
1754
1755 return err;
1756 }
1757
f2fs_freeze(struct super_block * sb)1758 static int f2fs_freeze(struct super_block *sb)
1759 {
1760 if (f2fs_readonly(sb))
1761 return 0;
1762
1763 /* IO error happened before */
1764 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1765 return -EIO;
1766
1767 /* must be clean, since sync_filesystem() was already called */
1768 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1769 return -EINVAL;
1770
1771 /* Let's flush checkpoints and stop the thread. */
1772 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1773
1774 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1775 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1776 return 0;
1777 }
1778
f2fs_unfreeze(struct super_block * sb)1779 static int f2fs_unfreeze(struct super_block *sb)
1780 {
1781 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1782 return 0;
1783 }
1784
1785 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1786 static int f2fs_statfs_project(struct super_block *sb,
1787 kprojid_t projid, struct kstatfs *buf)
1788 {
1789 struct kqid qid;
1790 struct dquot *dquot;
1791 u64 limit;
1792 u64 curblock;
1793
1794 qid = make_kqid_projid(projid);
1795 dquot = dqget(sb, qid);
1796 if (IS_ERR(dquot))
1797 return PTR_ERR(dquot);
1798 spin_lock(&dquot->dq_dqb_lock);
1799
1800 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1801 dquot->dq_dqb.dqb_bhardlimit);
1802 if (limit)
1803 limit >>= sb->s_blocksize_bits;
1804
1805 if (limit && buf->f_blocks > limit) {
1806 curblock = (dquot->dq_dqb.dqb_curspace +
1807 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1808 buf->f_blocks = limit;
1809 buf->f_bfree = buf->f_bavail =
1810 (buf->f_blocks > curblock) ?
1811 (buf->f_blocks - curblock) : 0;
1812 }
1813
1814 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1815 dquot->dq_dqb.dqb_ihardlimit);
1816
1817 if (limit && buf->f_files > limit) {
1818 buf->f_files = limit;
1819 buf->f_ffree =
1820 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1821 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1822 }
1823
1824 spin_unlock(&dquot->dq_dqb_lock);
1825 dqput(dquot);
1826 return 0;
1827 }
1828 #endif
1829
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1830 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1831 {
1832 struct super_block *sb = dentry->d_sb;
1833 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1834 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1835 block_t total_count, user_block_count, start_count;
1836 u64 avail_node_count;
1837 unsigned int total_valid_node_count;
1838
1839 total_count = le64_to_cpu(sbi->raw_super->block_count);
1840 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1841 buf->f_type = F2FS_SUPER_MAGIC;
1842 buf->f_bsize = sbi->blocksize;
1843
1844 buf->f_blocks = total_count - start_count;
1845
1846 spin_lock(&sbi->stat_lock);
1847
1848 user_block_count = sbi->user_block_count;
1849 total_valid_node_count = valid_node_count(sbi);
1850 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1851 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1852 sbi->current_reserved_blocks;
1853
1854 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1855 buf->f_bfree = 0;
1856 else
1857 buf->f_bfree -= sbi->unusable_block_count;
1858 spin_unlock(&sbi->stat_lock);
1859
1860 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1861 buf->f_bavail = buf->f_bfree -
1862 F2FS_OPTION(sbi).root_reserved_blocks;
1863 else
1864 buf->f_bavail = 0;
1865
1866 if (avail_node_count > user_block_count) {
1867 buf->f_files = user_block_count;
1868 buf->f_ffree = buf->f_bavail;
1869 } else {
1870 buf->f_files = avail_node_count;
1871 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1872 buf->f_bavail);
1873 }
1874
1875 buf->f_namelen = F2FS_NAME_LEN;
1876 buf->f_fsid = u64_to_fsid(id);
1877
1878 #ifdef CONFIG_QUOTA
1879 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1880 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1881 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1882 }
1883 #endif
1884 return 0;
1885 }
1886
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1887 static inline void f2fs_show_quota_options(struct seq_file *seq,
1888 struct super_block *sb)
1889 {
1890 #ifdef CONFIG_QUOTA
1891 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1892
1893 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1894 char *fmtname = "";
1895
1896 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1897 case QFMT_VFS_OLD:
1898 fmtname = "vfsold";
1899 break;
1900 case QFMT_VFS_V0:
1901 fmtname = "vfsv0";
1902 break;
1903 case QFMT_VFS_V1:
1904 fmtname = "vfsv1";
1905 break;
1906 }
1907 seq_printf(seq, ",jqfmt=%s", fmtname);
1908 }
1909
1910 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1911 seq_show_option(seq, "usrjquota",
1912 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1913
1914 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1915 seq_show_option(seq, "grpjquota",
1916 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1917
1918 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1919 seq_show_option(seq, "prjjquota",
1920 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1921 #endif
1922 }
1923
1924 #ifdef CONFIG_F2FS_FS_COMPRESSION
f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1925 static inline void f2fs_show_compress_options(struct seq_file *seq,
1926 struct super_block *sb)
1927 {
1928 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1929 char *algtype = "";
1930 int i;
1931
1932 if (!f2fs_sb_has_compression(sbi))
1933 return;
1934
1935 switch (F2FS_OPTION(sbi).compress_algorithm) {
1936 case COMPRESS_LZO:
1937 algtype = "lzo";
1938 break;
1939 case COMPRESS_LZ4:
1940 algtype = "lz4";
1941 break;
1942 case COMPRESS_ZSTD:
1943 algtype = "zstd";
1944 break;
1945 case COMPRESS_LZORLE:
1946 algtype = "lzo-rle";
1947 break;
1948 }
1949 seq_printf(seq, ",compress_algorithm=%s", algtype);
1950
1951 if (F2FS_OPTION(sbi).compress_level)
1952 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1953
1954 seq_printf(seq, ",compress_log_size=%u",
1955 F2FS_OPTION(sbi).compress_log_size);
1956
1957 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1958 seq_printf(seq, ",compress_extension=%s",
1959 F2FS_OPTION(sbi).extensions[i]);
1960 }
1961
1962 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1963 seq_printf(seq, ",nocompress_extension=%s",
1964 F2FS_OPTION(sbi).noextensions[i]);
1965 }
1966
1967 if (F2FS_OPTION(sbi).compress_chksum)
1968 seq_puts(seq, ",compress_chksum");
1969
1970 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1971 seq_printf(seq, ",compress_mode=%s", "fs");
1972 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1973 seq_printf(seq, ",compress_mode=%s", "user");
1974
1975 if (test_opt(sbi, COMPRESS_CACHE))
1976 seq_puts(seq, ",compress_cache");
1977 }
1978 #endif
1979
f2fs_show_options(struct seq_file * seq,struct dentry * root)1980 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1981 {
1982 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1983
1984 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1985 seq_printf(seq, ",background_gc=%s", "sync");
1986 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1987 seq_printf(seq, ",background_gc=%s", "on");
1988 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1989 seq_printf(seq, ",background_gc=%s", "off");
1990
1991 if (test_opt(sbi, GC_MERGE))
1992 seq_puts(seq, ",gc_merge");
1993 else
1994 seq_puts(seq, ",nogc_merge");
1995
1996 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1997 seq_puts(seq, ",disable_roll_forward");
1998 if (test_opt(sbi, NORECOVERY))
1999 seq_puts(seq, ",norecovery");
2000 if (test_opt(sbi, DISCARD)) {
2001 seq_puts(seq, ",discard");
2002 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2003 seq_printf(seq, ",discard_unit=%s", "block");
2004 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2005 seq_printf(seq, ",discard_unit=%s", "segment");
2006 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2007 seq_printf(seq, ",discard_unit=%s", "section");
2008 } else {
2009 seq_puts(seq, ",nodiscard");
2010 }
2011 if (test_opt(sbi, NOHEAP))
2012 seq_puts(seq, ",no_heap");
2013 else
2014 seq_puts(seq, ",heap");
2015 #ifdef CONFIG_F2FS_FS_XATTR
2016 if (test_opt(sbi, XATTR_USER))
2017 seq_puts(seq, ",user_xattr");
2018 else
2019 seq_puts(seq, ",nouser_xattr");
2020 if (test_opt(sbi, INLINE_XATTR))
2021 seq_puts(seq, ",inline_xattr");
2022 else
2023 seq_puts(seq, ",noinline_xattr");
2024 if (test_opt(sbi, INLINE_XATTR_SIZE))
2025 seq_printf(seq, ",inline_xattr_size=%u",
2026 F2FS_OPTION(sbi).inline_xattr_size);
2027 #endif
2028 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2029 if (test_opt(sbi, POSIX_ACL))
2030 seq_puts(seq, ",acl");
2031 else
2032 seq_puts(seq, ",noacl");
2033 #endif
2034 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
2035 seq_puts(seq, ",disable_ext_identify");
2036 if (test_opt(sbi, INLINE_DATA))
2037 seq_puts(seq, ",inline_data");
2038 else
2039 seq_puts(seq, ",noinline_data");
2040 if (test_opt(sbi, INLINE_DENTRY))
2041 seq_puts(seq, ",inline_dentry");
2042 else
2043 seq_puts(seq, ",noinline_dentry");
2044 if (test_opt(sbi, FLUSH_MERGE))
2045 seq_puts(seq, ",flush_merge");
2046 else
2047 seq_puts(seq, ",noflush_merge");
2048 if (test_opt(sbi, NOBARRIER))
2049 seq_puts(seq, ",nobarrier");
2050 else
2051 seq_puts(seq, ",barrier");
2052 if (test_opt(sbi, FASTBOOT))
2053 seq_puts(seq, ",fastboot");
2054 if (test_opt(sbi, READ_EXTENT_CACHE))
2055 seq_puts(seq, ",extent_cache");
2056 else
2057 seq_puts(seq, ",noextent_cache");
2058 if (test_opt(sbi, AGE_EXTENT_CACHE))
2059 seq_puts(seq, ",age_extent_cache");
2060 if (test_opt(sbi, DATA_FLUSH))
2061 seq_puts(seq, ",data_flush");
2062
2063 seq_puts(seq, ",mode=");
2064 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2065 seq_puts(seq, "adaptive");
2066 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2067 seq_puts(seq, "lfs");
2068 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2069 seq_puts(seq, "fragment:segment");
2070 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2071 seq_puts(seq, "fragment:block");
2072 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2073 if (test_opt(sbi, RESERVE_ROOT))
2074 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2075 F2FS_OPTION(sbi).root_reserved_blocks,
2076 from_kuid_munged(&init_user_ns,
2077 F2FS_OPTION(sbi).s_resuid),
2078 from_kgid_munged(&init_user_ns,
2079 F2FS_OPTION(sbi).s_resgid));
2080 if (F2FS_IO_SIZE_BITS(sbi))
2081 seq_printf(seq, ",io_bits=%u",
2082 F2FS_OPTION(sbi).write_io_size_bits);
2083 #ifdef CONFIG_F2FS_FAULT_INJECTION
2084 if (test_opt(sbi, FAULT_INJECTION)) {
2085 seq_printf(seq, ",fault_injection=%u",
2086 F2FS_OPTION(sbi).fault_info.inject_rate);
2087 seq_printf(seq, ",fault_type=%u",
2088 F2FS_OPTION(sbi).fault_info.inject_type);
2089 }
2090 #endif
2091 #ifdef CONFIG_QUOTA
2092 if (test_opt(sbi, QUOTA))
2093 seq_puts(seq, ",quota");
2094 if (test_opt(sbi, USRQUOTA))
2095 seq_puts(seq, ",usrquota");
2096 if (test_opt(sbi, GRPQUOTA))
2097 seq_puts(seq, ",grpquota");
2098 if (test_opt(sbi, PRJQUOTA))
2099 seq_puts(seq, ",prjquota");
2100 #endif
2101 f2fs_show_quota_options(seq, sbi->sb);
2102
2103 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2104
2105 if (sbi->sb->s_flags & SB_INLINECRYPT)
2106 seq_puts(seq, ",inlinecrypt");
2107
2108 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2109 seq_printf(seq, ",alloc_mode=%s", "default");
2110 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2111 seq_printf(seq, ",alloc_mode=%s", "reuse");
2112
2113 if (test_opt(sbi, DISABLE_CHECKPOINT))
2114 seq_printf(seq, ",checkpoint=disable:%u",
2115 F2FS_OPTION(sbi).unusable_cap);
2116 if (test_opt(sbi, MERGE_CHECKPOINT))
2117 seq_puts(seq, ",checkpoint_merge");
2118 else
2119 seq_puts(seq, ",nocheckpoint_merge");
2120 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2121 seq_printf(seq, ",fsync_mode=%s", "posix");
2122 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2123 seq_printf(seq, ",fsync_mode=%s", "strict");
2124 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2125 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2126
2127 #ifdef CONFIG_F2FS_FS_COMPRESSION
2128 f2fs_show_compress_options(seq, sbi->sb);
2129 #endif
2130
2131 if (test_opt(sbi, ATGC))
2132 seq_puts(seq, ",atgc");
2133
2134 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2135 seq_printf(seq, ",memory=%s", "normal");
2136 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2137 seq_printf(seq, ",memory=%s", "low");
2138
2139 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2140 seq_printf(seq, ",errors=%s", "remount-ro");
2141 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2142 seq_printf(seq, ",errors=%s", "continue");
2143 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2144 seq_printf(seq, ",errors=%s", "panic");
2145
2146 return 0;
2147 }
2148
default_options(struct f2fs_sb_info * sbi,bool remount)2149 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2150 {
2151 /* init some FS parameters */
2152 if (!remount) {
2153 set_opt(sbi, READ_EXTENT_CACHE);
2154 clear_opt(sbi, DISABLE_CHECKPOINT);
2155
2156 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2157 set_opt(sbi, DISCARD);
2158
2159 if (f2fs_sb_has_blkzoned(sbi))
2160 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2161 else
2162 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2163 }
2164
2165 if (f2fs_sb_has_readonly(sbi))
2166 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2167 else
2168 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2169
2170 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2171 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2172 SMALL_VOLUME_SEGMENTS)
2173 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2174 else
2175 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2176 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2177 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2178 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2179 if (f2fs_sb_has_compression(sbi)) {
2180 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2181 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2182 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2183 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2184 }
2185 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2186 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2187 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2188
2189 sbi->sb->s_flags &= ~SB_INLINECRYPT;
2190
2191 set_opt(sbi, INLINE_XATTR);
2192 set_opt(sbi, INLINE_DATA);
2193 set_opt(sbi, INLINE_DENTRY);
2194 set_opt(sbi, NOHEAP);
2195 set_opt(sbi, MERGE_CHECKPOINT);
2196 F2FS_OPTION(sbi).unusable_cap = 0;
2197 sbi->sb->s_flags |= SB_LAZYTIME;
2198 if (!f2fs_is_readonly(sbi))
2199 set_opt(sbi, FLUSH_MERGE);
2200 if (f2fs_sb_has_blkzoned(sbi))
2201 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2202 else
2203 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2204
2205 #ifdef CONFIG_F2FS_FS_XATTR
2206 set_opt(sbi, XATTR_USER);
2207 #endif
2208 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2209 set_opt(sbi, POSIX_ACL);
2210 #endif
2211
2212 f2fs_build_fault_attr(sbi, 0, 0);
2213 }
2214
2215 #ifdef CONFIG_QUOTA
2216 static int f2fs_enable_quotas(struct super_block *sb);
2217 #endif
2218
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)2219 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2220 {
2221 unsigned int s_flags = sbi->sb->s_flags;
2222 struct cp_control cpc;
2223 unsigned int gc_mode = sbi->gc_mode;
2224 int err = 0;
2225 int ret;
2226 block_t unusable;
2227
2228 if (s_flags & SB_RDONLY) {
2229 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2230 return -EINVAL;
2231 }
2232 sbi->sb->s_flags |= SB_ACTIVE;
2233
2234 /* check if we need more GC first */
2235 unusable = f2fs_get_unusable_blocks(sbi);
2236 if (!f2fs_disable_cp_again(sbi, unusable))
2237 goto skip_gc;
2238
2239 f2fs_update_time(sbi, DISABLE_TIME);
2240
2241 sbi->gc_mode = GC_URGENT_HIGH;
2242
2243 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2244 struct f2fs_gc_control gc_control = {
2245 .victim_segno = NULL_SEGNO,
2246 .init_gc_type = FG_GC,
2247 .should_migrate_blocks = false,
2248 .err_gc_skipped = true,
2249 .nr_free_secs = 1 };
2250
2251 f2fs_down_write(&sbi->gc_lock);
2252 stat_inc_gc_call_count(sbi, FOREGROUND);
2253 err = f2fs_gc(sbi, &gc_control);
2254 if (err == -ENODATA) {
2255 err = 0;
2256 break;
2257 }
2258 if (err && err != -EAGAIN)
2259 break;
2260 }
2261
2262 ret = sync_filesystem(sbi->sb);
2263 if (ret || err) {
2264 err = ret ? ret : err;
2265 goto restore_flag;
2266 }
2267
2268 unusable = f2fs_get_unusable_blocks(sbi);
2269 if (f2fs_disable_cp_again(sbi, unusable)) {
2270 err = -EAGAIN;
2271 goto restore_flag;
2272 }
2273
2274 skip_gc:
2275 f2fs_down_write(&sbi->gc_lock);
2276 cpc.reason = CP_PAUSE;
2277 set_sbi_flag(sbi, SBI_CP_DISABLED);
2278 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2279 err = f2fs_write_checkpoint(sbi, &cpc);
2280 if (err)
2281 goto out_unlock;
2282
2283 spin_lock(&sbi->stat_lock);
2284 sbi->unusable_block_count = unusable;
2285 spin_unlock(&sbi->stat_lock);
2286
2287 out_unlock:
2288 f2fs_up_write(&sbi->gc_lock);
2289 restore_flag:
2290 sbi->gc_mode = gc_mode;
2291 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2292 return err;
2293 }
2294
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2295 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2296 {
2297 int retry = DEFAULT_RETRY_IO_COUNT;
2298
2299 /* we should flush all the data to keep data consistency */
2300 do {
2301 sync_inodes_sb(sbi->sb);
2302 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2303 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2304
2305 if (unlikely(retry < 0))
2306 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2307
2308 f2fs_down_write(&sbi->gc_lock);
2309 f2fs_dirty_to_prefree(sbi);
2310
2311 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2312 set_sbi_flag(sbi, SBI_IS_DIRTY);
2313 f2fs_up_write(&sbi->gc_lock);
2314
2315 f2fs_sync_fs(sbi->sb, 1);
2316
2317 /* Let's ensure there's no pending checkpoint anymore */
2318 f2fs_flush_ckpt_thread(sbi);
2319 }
2320
f2fs_remount(struct super_block * sb,int * flags,char * data)2321 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2322 {
2323 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2324 struct f2fs_mount_info org_mount_opt;
2325 unsigned long old_sb_flags;
2326 int err;
2327 bool need_restart_gc = false, need_stop_gc = false;
2328 bool need_restart_ckpt = false, need_stop_ckpt = false;
2329 bool need_restart_flush = false, need_stop_flush = false;
2330 bool need_restart_discard = false, need_stop_discard = false;
2331 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2332 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2333 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2334 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2335 bool no_atgc = !test_opt(sbi, ATGC);
2336 bool no_discard = !test_opt(sbi, DISCARD);
2337 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2338 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2339 #ifdef CONFIG_QUOTA
2340 int i, j;
2341 #endif
2342
2343 /*
2344 * Save the old mount options in case we
2345 * need to restore them.
2346 */
2347 org_mount_opt = sbi->mount_opt;
2348 old_sb_flags = sb->s_flags;
2349
2350 #ifdef CONFIG_QUOTA
2351 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2352 for (i = 0; i < MAXQUOTAS; i++) {
2353 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2354 org_mount_opt.s_qf_names[i] =
2355 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2356 GFP_KERNEL);
2357 if (!org_mount_opt.s_qf_names[i]) {
2358 for (j = 0; j < i; j++)
2359 kfree(org_mount_opt.s_qf_names[j]);
2360 return -ENOMEM;
2361 }
2362 } else {
2363 org_mount_opt.s_qf_names[i] = NULL;
2364 }
2365 }
2366 #endif
2367
2368 /* recover superblocks we couldn't write due to previous RO mount */
2369 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2370 err = f2fs_commit_super(sbi, false);
2371 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2372 err);
2373 if (!err)
2374 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2375 }
2376
2377 default_options(sbi, true);
2378
2379 /* parse mount options */
2380 err = parse_options(sb, data, true);
2381 if (err)
2382 goto restore_opts;
2383
2384 /* flush outstanding errors before changing fs state */
2385 flush_work(&sbi->s_error_work);
2386
2387 /*
2388 * Previous and new state of filesystem is RO,
2389 * so skip checking GC and FLUSH_MERGE conditions.
2390 */
2391 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2392 goto skip;
2393
2394 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2395 err = -EROFS;
2396 goto restore_opts;
2397 }
2398
2399 #ifdef CONFIG_QUOTA
2400 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2401 err = dquot_suspend(sb, -1);
2402 if (err < 0)
2403 goto restore_opts;
2404 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2405 /* dquot_resume needs RW */
2406 sb->s_flags &= ~SB_RDONLY;
2407 if (sb_any_quota_suspended(sb)) {
2408 dquot_resume(sb, -1);
2409 } else if (f2fs_sb_has_quota_ino(sbi)) {
2410 err = f2fs_enable_quotas(sb);
2411 if (err)
2412 goto restore_opts;
2413 }
2414 }
2415 #endif
2416 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2417 err = -EINVAL;
2418 f2fs_warn(sbi, "LFS is not compatible with IPU");
2419 goto restore_opts;
2420 }
2421
2422 /* disallow enable atgc dynamically */
2423 if (no_atgc == !!test_opt(sbi, ATGC)) {
2424 err = -EINVAL;
2425 f2fs_warn(sbi, "switch atgc option is not allowed");
2426 goto restore_opts;
2427 }
2428
2429 /* disallow enable/disable extent_cache dynamically */
2430 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2431 err = -EINVAL;
2432 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2433 goto restore_opts;
2434 }
2435 /* disallow enable/disable age extent_cache dynamically */
2436 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2437 err = -EINVAL;
2438 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2439 goto restore_opts;
2440 }
2441
2442 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2443 err = -EINVAL;
2444 f2fs_warn(sbi, "switch io_bits option is not allowed");
2445 goto restore_opts;
2446 }
2447
2448 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2449 err = -EINVAL;
2450 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2451 goto restore_opts;
2452 }
2453
2454 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2455 err = -EINVAL;
2456 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2457 goto restore_opts;
2458 }
2459
2460 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2461 err = -EINVAL;
2462 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2463 goto restore_opts;
2464 }
2465
2466 /*
2467 * We stop the GC thread if FS is mounted as RO
2468 * or if background_gc = off is passed in mount
2469 * option. Also sync the filesystem.
2470 */
2471 if ((*flags & SB_RDONLY) ||
2472 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2473 !test_opt(sbi, GC_MERGE))) {
2474 if (sbi->gc_thread) {
2475 f2fs_stop_gc_thread(sbi);
2476 need_restart_gc = true;
2477 }
2478 } else if (!sbi->gc_thread) {
2479 err = f2fs_start_gc_thread(sbi);
2480 if (err)
2481 goto restore_opts;
2482 need_stop_gc = true;
2483 }
2484
2485 if (*flags & SB_RDONLY) {
2486 sync_inodes_sb(sb);
2487
2488 set_sbi_flag(sbi, SBI_IS_DIRTY);
2489 set_sbi_flag(sbi, SBI_IS_CLOSE);
2490 f2fs_sync_fs(sb, 1);
2491 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2492 }
2493
2494 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2495 !test_opt(sbi, MERGE_CHECKPOINT)) {
2496 f2fs_stop_ckpt_thread(sbi);
2497 need_restart_ckpt = true;
2498 } else {
2499 /* Flush if the prevous checkpoint, if exists. */
2500 f2fs_flush_ckpt_thread(sbi);
2501
2502 err = f2fs_start_ckpt_thread(sbi);
2503 if (err) {
2504 f2fs_err(sbi,
2505 "Failed to start F2FS issue_checkpoint_thread (%d)",
2506 err);
2507 goto restore_gc;
2508 }
2509 need_stop_ckpt = true;
2510 }
2511
2512 /*
2513 * We stop issue flush thread if FS is mounted as RO
2514 * or if flush_merge is not passed in mount option.
2515 */
2516 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2517 clear_opt(sbi, FLUSH_MERGE);
2518 f2fs_destroy_flush_cmd_control(sbi, false);
2519 need_restart_flush = true;
2520 } else {
2521 err = f2fs_create_flush_cmd_control(sbi);
2522 if (err)
2523 goto restore_ckpt;
2524 need_stop_flush = true;
2525 }
2526
2527 if (no_discard == !!test_opt(sbi, DISCARD)) {
2528 if (test_opt(sbi, DISCARD)) {
2529 err = f2fs_start_discard_thread(sbi);
2530 if (err)
2531 goto restore_flush;
2532 need_stop_discard = true;
2533 } else {
2534 f2fs_stop_discard_thread(sbi);
2535 f2fs_issue_discard_timeout(sbi);
2536 need_restart_discard = true;
2537 }
2538 }
2539
2540 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2541 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2542 err = f2fs_disable_checkpoint(sbi);
2543 if (err)
2544 goto restore_discard;
2545 } else {
2546 f2fs_enable_checkpoint(sbi);
2547 }
2548 }
2549
2550 skip:
2551 #ifdef CONFIG_QUOTA
2552 /* Release old quota file names */
2553 for (i = 0; i < MAXQUOTAS; i++)
2554 kfree(org_mount_opt.s_qf_names[i]);
2555 #endif
2556 /* Update the POSIXACL Flag */
2557 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2558 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2559
2560 limit_reserve_root(sbi);
2561 adjust_unusable_cap_perc(sbi);
2562 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2563 return 0;
2564 restore_discard:
2565 if (need_restart_discard) {
2566 if (f2fs_start_discard_thread(sbi))
2567 f2fs_warn(sbi, "discard has been stopped");
2568 } else if (need_stop_discard) {
2569 f2fs_stop_discard_thread(sbi);
2570 }
2571 restore_flush:
2572 if (need_restart_flush) {
2573 if (f2fs_create_flush_cmd_control(sbi))
2574 f2fs_warn(sbi, "background flush thread has stopped");
2575 } else if (need_stop_flush) {
2576 clear_opt(sbi, FLUSH_MERGE);
2577 f2fs_destroy_flush_cmd_control(sbi, false);
2578 }
2579 restore_ckpt:
2580 if (need_restart_ckpt) {
2581 if (f2fs_start_ckpt_thread(sbi))
2582 f2fs_warn(sbi, "background ckpt thread has stopped");
2583 } else if (need_stop_ckpt) {
2584 f2fs_stop_ckpt_thread(sbi);
2585 }
2586 restore_gc:
2587 if (need_restart_gc) {
2588 if (f2fs_start_gc_thread(sbi))
2589 f2fs_warn(sbi, "background gc thread has stopped");
2590 } else if (need_stop_gc) {
2591 f2fs_stop_gc_thread(sbi);
2592 }
2593 restore_opts:
2594 #ifdef CONFIG_QUOTA
2595 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2596 for (i = 0; i < MAXQUOTAS; i++) {
2597 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2598 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2599 }
2600 #endif
2601 sbi->mount_opt = org_mount_opt;
2602 sb->s_flags = old_sb_flags;
2603 return err;
2604 }
2605
2606 #ifdef CONFIG_QUOTA
f2fs_need_recovery(struct f2fs_sb_info * sbi)2607 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2608 {
2609 /* need to recovery orphan */
2610 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2611 return true;
2612 /* need to recovery data */
2613 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2614 return false;
2615 if (test_opt(sbi, NORECOVERY))
2616 return false;
2617 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2618 }
2619
f2fs_recover_quota_begin(struct f2fs_sb_info * sbi)2620 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2621 {
2622 bool readonly = f2fs_readonly(sbi->sb);
2623
2624 if (!f2fs_need_recovery(sbi))
2625 return false;
2626
2627 /* it doesn't need to check f2fs_sb_has_readonly() */
2628 if (f2fs_hw_is_readonly(sbi))
2629 return false;
2630
2631 if (readonly) {
2632 sbi->sb->s_flags &= ~SB_RDONLY;
2633 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2634 }
2635
2636 /*
2637 * Turn on quotas which were not enabled for read-only mounts if
2638 * filesystem has quota feature, so that they are updated correctly.
2639 */
2640 return f2fs_enable_quota_files(sbi, readonly);
2641 }
2642
f2fs_recover_quota_end(struct f2fs_sb_info * sbi,bool quota_enabled)2643 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2644 bool quota_enabled)
2645 {
2646 if (quota_enabled)
2647 f2fs_quota_off_umount(sbi->sb);
2648
2649 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2650 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2651 sbi->sb->s_flags |= SB_RDONLY;
2652 }
2653 }
2654
2655 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2656 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2657 size_t len, loff_t off)
2658 {
2659 struct inode *inode = sb_dqopt(sb)->files[type];
2660 struct address_space *mapping = inode->i_mapping;
2661 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2662 int offset = off & (sb->s_blocksize - 1);
2663 int tocopy;
2664 size_t toread;
2665 loff_t i_size = i_size_read(inode);
2666 struct page *page;
2667
2668 if (off > i_size)
2669 return 0;
2670
2671 if (off + len > i_size)
2672 len = i_size - off;
2673 toread = len;
2674 while (toread > 0) {
2675 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2676 repeat:
2677 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2678 if (IS_ERR(page)) {
2679 if (PTR_ERR(page) == -ENOMEM) {
2680 memalloc_retry_wait(GFP_NOFS);
2681 goto repeat;
2682 }
2683 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2684 return PTR_ERR(page);
2685 }
2686
2687 lock_page(page);
2688
2689 if (unlikely(page->mapping != mapping)) {
2690 f2fs_put_page(page, 1);
2691 goto repeat;
2692 }
2693 if (unlikely(!PageUptodate(page))) {
2694 f2fs_put_page(page, 1);
2695 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2696 return -EIO;
2697 }
2698
2699 memcpy_from_page(data, page, offset, tocopy);
2700 f2fs_put_page(page, 1);
2701
2702 offset = 0;
2703 toread -= tocopy;
2704 data += tocopy;
2705 blkidx++;
2706 }
2707 return len;
2708 }
2709
2710 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2711 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2712 const char *data, size_t len, loff_t off)
2713 {
2714 struct inode *inode = sb_dqopt(sb)->files[type];
2715 struct address_space *mapping = inode->i_mapping;
2716 const struct address_space_operations *a_ops = mapping->a_ops;
2717 int offset = off & (sb->s_blocksize - 1);
2718 size_t towrite = len;
2719 struct page *page;
2720 void *fsdata = NULL;
2721 int err = 0;
2722 int tocopy;
2723
2724 while (towrite > 0) {
2725 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2726 towrite);
2727 retry:
2728 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2729 &page, &fsdata);
2730 if (unlikely(err)) {
2731 if (err == -ENOMEM) {
2732 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2733 goto retry;
2734 }
2735 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2736 break;
2737 }
2738
2739 memcpy_to_page(page, offset, data, tocopy);
2740
2741 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2742 page, fsdata);
2743 offset = 0;
2744 towrite -= tocopy;
2745 off += tocopy;
2746 data += tocopy;
2747 cond_resched();
2748 }
2749
2750 if (len == towrite)
2751 return err;
2752 inode->i_mtime = inode_set_ctime_current(inode);
2753 f2fs_mark_inode_dirty_sync(inode, false);
2754 return len - towrite;
2755 }
2756
f2fs_dquot_initialize(struct inode * inode)2757 int f2fs_dquot_initialize(struct inode *inode)
2758 {
2759 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2760 return -ESRCH;
2761
2762 return dquot_initialize(inode);
2763 }
2764
f2fs_get_dquots(struct inode * inode)2765 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2766 {
2767 return F2FS_I(inode)->i_dquot;
2768 }
2769
f2fs_get_reserved_space(struct inode * inode)2770 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2771 {
2772 return &F2FS_I(inode)->i_reserved_quota;
2773 }
2774
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2775 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2776 {
2777 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2778 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2779 return 0;
2780 }
2781
2782 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2783 F2FS_OPTION(sbi).s_jquota_fmt, type);
2784 }
2785
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2786 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2787 {
2788 int enabled = 0;
2789 int i, err;
2790
2791 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2792 err = f2fs_enable_quotas(sbi->sb);
2793 if (err) {
2794 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2795 return 0;
2796 }
2797 return 1;
2798 }
2799
2800 for (i = 0; i < MAXQUOTAS; i++) {
2801 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2802 err = f2fs_quota_on_mount(sbi, i);
2803 if (!err) {
2804 enabled = 1;
2805 continue;
2806 }
2807 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2808 err, i);
2809 }
2810 }
2811 return enabled;
2812 }
2813
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2814 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2815 unsigned int flags)
2816 {
2817 struct inode *qf_inode;
2818 unsigned long qf_inum;
2819 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2820 int err;
2821
2822 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2823
2824 qf_inum = f2fs_qf_ino(sb, type);
2825 if (!qf_inum)
2826 return -EPERM;
2827
2828 qf_inode = f2fs_iget(sb, qf_inum);
2829 if (IS_ERR(qf_inode)) {
2830 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2831 return PTR_ERR(qf_inode);
2832 }
2833
2834 /* Don't account quota for quota files to avoid recursion */
2835 inode_lock(qf_inode);
2836 qf_inode->i_flags |= S_NOQUOTA;
2837
2838 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2839 F2FS_I(qf_inode)->i_flags |= qf_flag;
2840 f2fs_set_inode_flags(qf_inode);
2841 }
2842 inode_unlock(qf_inode);
2843
2844 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2845 iput(qf_inode);
2846 return err;
2847 }
2848
f2fs_enable_quotas(struct super_block * sb)2849 static int f2fs_enable_quotas(struct super_block *sb)
2850 {
2851 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2852 int type, err = 0;
2853 unsigned long qf_inum;
2854 bool quota_mopt[MAXQUOTAS] = {
2855 test_opt(sbi, USRQUOTA),
2856 test_opt(sbi, GRPQUOTA),
2857 test_opt(sbi, PRJQUOTA),
2858 };
2859
2860 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2861 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2862 return 0;
2863 }
2864
2865 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2866
2867 for (type = 0; type < MAXQUOTAS; type++) {
2868 qf_inum = f2fs_qf_ino(sb, type);
2869 if (qf_inum) {
2870 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2871 DQUOT_USAGE_ENABLED |
2872 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2873 if (err) {
2874 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2875 type, err);
2876 for (type--; type >= 0; type--)
2877 dquot_quota_off(sb, type);
2878 set_sbi_flag(F2FS_SB(sb),
2879 SBI_QUOTA_NEED_REPAIR);
2880 return err;
2881 }
2882 }
2883 }
2884 return 0;
2885 }
2886
f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2887 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2888 {
2889 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2890 struct address_space *mapping = dqopt->files[type]->i_mapping;
2891 int ret = 0;
2892
2893 ret = dquot_writeback_dquots(sbi->sb, type);
2894 if (ret)
2895 goto out;
2896
2897 ret = filemap_fdatawrite(mapping);
2898 if (ret)
2899 goto out;
2900
2901 /* if we are using journalled quota */
2902 if (is_journalled_quota(sbi))
2903 goto out;
2904
2905 ret = filemap_fdatawait(mapping);
2906
2907 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2908 out:
2909 if (ret)
2910 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2911 return ret;
2912 }
2913
f2fs_quota_sync(struct super_block * sb,int type)2914 int f2fs_quota_sync(struct super_block *sb, int type)
2915 {
2916 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2917 struct quota_info *dqopt = sb_dqopt(sb);
2918 int cnt;
2919 int ret = 0;
2920
2921 /*
2922 * Now when everything is written we can discard the pagecache so
2923 * that userspace sees the changes.
2924 */
2925 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2926
2927 if (type != -1 && cnt != type)
2928 continue;
2929
2930 if (!sb_has_quota_active(sb, cnt))
2931 continue;
2932
2933 if (!f2fs_sb_has_quota_ino(sbi))
2934 inode_lock(dqopt->files[cnt]);
2935
2936 /*
2937 * do_quotactl
2938 * f2fs_quota_sync
2939 * f2fs_down_read(quota_sem)
2940 * dquot_writeback_dquots()
2941 * f2fs_dquot_commit
2942 * block_operation
2943 * f2fs_down_read(quota_sem)
2944 */
2945 f2fs_lock_op(sbi);
2946 f2fs_down_read(&sbi->quota_sem);
2947
2948 ret = f2fs_quota_sync_file(sbi, cnt);
2949
2950 f2fs_up_read(&sbi->quota_sem);
2951 f2fs_unlock_op(sbi);
2952
2953 if (!f2fs_sb_has_quota_ino(sbi))
2954 inode_unlock(dqopt->files[cnt]);
2955
2956 if (ret)
2957 break;
2958 }
2959 return ret;
2960 }
2961
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2962 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2963 const struct path *path)
2964 {
2965 struct inode *inode;
2966 int err;
2967
2968 /* if quota sysfile exists, deny enabling quota with specific file */
2969 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2970 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2971 return -EBUSY;
2972 }
2973
2974 if (path->dentry->d_sb != sb)
2975 return -EXDEV;
2976
2977 err = f2fs_quota_sync(sb, type);
2978 if (err)
2979 return err;
2980
2981 inode = d_inode(path->dentry);
2982
2983 err = filemap_fdatawrite(inode->i_mapping);
2984 if (err)
2985 return err;
2986
2987 err = filemap_fdatawait(inode->i_mapping);
2988 if (err)
2989 return err;
2990
2991 err = dquot_quota_on(sb, type, format_id, path);
2992 if (err)
2993 return err;
2994
2995 inode_lock(inode);
2996 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2997 f2fs_set_inode_flags(inode);
2998 inode_unlock(inode);
2999 f2fs_mark_inode_dirty_sync(inode, false);
3000
3001 return 0;
3002 }
3003
__f2fs_quota_off(struct super_block * sb,int type)3004 static int __f2fs_quota_off(struct super_block *sb, int type)
3005 {
3006 struct inode *inode = sb_dqopt(sb)->files[type];
3007 int err;
3008
3009 if (!inode || !igrab(inode))
3010 return dquot_quota_off(sb, type);
3011
3012 err = f2fs_quota_sync(sb, type);
3013 if (err)
3014 goto out_put;
3015
3016 err = dquot_quota_off(sb, type);
3017 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
3018 goto out_put;
3019
3020 inode_lock(inode);
3021 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
3022 f2fs_set_inode_flags(inode);
3023 inode_unlock(inode);
3024 f2fs_mark_inode_dirty_sync(inode, false);
3025 out_put:
3026 iput(inode);
3027 return err;
3028 }
3029
f2fs_quota_off(struct super_block * sb,int type)3030 static int f2fs_quota_off(struct super_block *sb, int type)
3031 {
3032 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3033 int err;
3034
3035 err = __f2fs_quota_off(sb, type);
3036
3037 /*
3038 * quotactl can shutdown journalled quota, result in inconsistence
3039 * between quota record and fs data by following updates, tag the
3040 * flag to let fsck be aware of it.
3041 */
3042 if (is_journalled_quota(sbi))
3043 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3044 return err;
3045 }
3046
f2fs_quota_off_umount(struct super_block * sb)3047 void f2fs_quota_off_umount(struct super_block *sb)
3048 {
3049 int type;
3050 int err;
3051
3052 for (type = 0; type < MAXQUOTAS; type++) {
3053 err = __f2fs_quota_off(sb, type);
3054 if (err) {
3055 int ret = dquot_quota_off(sb, type);
3056
3057 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3058 type, err, ret);
3059 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3060 }
3061 }
3062 /*
3063 * In case of checkpoint=disable, we must flush quota blocks.
3064 * This can cause NULL exception for node_inode in end_io, since
3065 * put_super already dropped it.
3066 */
3067 sync_filesystem(sb);
3068 }
3069
f2fs_truncate_quota_inode_pages(struct super_block * sb)3070 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3071 {
3072 struct quota_info *dqopt = sb_dqopt(sb);
3073 int type;
3074
3075 for (type = 0; type < MAXQUOTAS; type++) {
3076 if (!dqopt->files[type])
3077 continue;
3078 f2fs_inode_synced(dqopt->files[type]);
3079 }
3080 }
3081
f2fs_dquot_commit(struct dquot * dquot)3082 static int f2fs_dquot_commit(struct dquot *dquot)
3083 {
3084 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3085 int ret;
3086
3087 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3088 ret = dquot_commit(dquot);
3089 if (ret < 0)
3090 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3091 f2fs_up_read(&sbi->quota_sem);
3092 return ret;
3093 }
3094
f2fs_dquot_acquire(struct dquot * dquot)3095 static int f2fs_dquot_acquire(struct dquot *dquot)
3096 {
3097 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3098 int ret;
3099
3100 f2fs_down_read(&sbi->quota_sem);
3101 ret = dquot_acquire(dquot);
3102 if (ret < 0)
3103 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3104 f2fs_up_read(&sbi->quota_sem);
3105 return ret;
3106 }
3107
f2fs_dquot_release(struct dquot * dquot)3108 static int f2fs_dquot_release(struct dquot *dquot)
3109 {
3110 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3111 int ret = dquot_release(dquot);
3112
3113 if (ret < 0)
3114 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3115 return ret;
3116 }
3117
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)3118 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3119 {
3120 struct super_block *sb = dquot->dq_sb;
3121 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3122 int ret = dquot_mark_dquot_dirty(dquot);
3123
3124 /* if we are using journalled quota */
3125 if (is_journalled_quota(sbi))
3126 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3127
3128 return ret;
3129 }
3130
f2fs_dquot_commit_info(struct super_block * sb,int type)3131 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3132 {
3133 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3134 int ret = dquot_commit_info(sb, type);
3135
3136 if (ret < 0)
3137 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3138 return ret;
3139 }
3140
f2fs_get_projid(struct inode * inode,kprojid_t * projid)3141 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3142 {
3143 *projid = F2FS_I(inode)->i_projid;
3144 return 0;
3145 }
3146
3147 static const struct dquot_operations f2fs_quota_operations = {
3148 .get_reserved_space = f2fs_get_reserved_space,
3149 .write_dquot = f2fs_dquot_commit,
3150 .acquire_dquot = f2fs_dquot_acquire,
3151 .release_dquot = f2fs_dquot_release,
3152 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3153 .write_info = f2fs_dquot_commit_info,
3154 .alloc_dquot = dquot_alloc,
3155 .destroy_dquot = dquot_destroy,
3156 .get_projid = f2fs_get_projid,
3157 .get_next_id = dquot_get_next_id,
3158 };
3159
3160 static const struct quotactl_ops f2fs_quotactl_ops = {
3161 .quota_on = f2fs_quota_on,
3162 .quota_off = f2fs_quota_off,
3163 .quota_sync = f2fs_quota_sync,
3164 .get_state = dquot_get_state,
3165 .set_info = dquot_set_dqinfo,
3166 .get_dqblk = dquot_get_dqblk,
3167 .set_dqblk = dquot_set_dqblk,
3168 .get_nextdqblk = dquot_get_next_dqblk,
3169 };
3170 #else
f2fs_dquot_initialize(struct inode * inode)3171 int f2fs_dquot_initialize(struct inode *inode)
3172 {
3173 return 0;
3174 }
3175
f2fs_quota_sync(struct super_block * sb,int type)3176 int f2fs_quota_sync(struct super_block *sb, int type)
3177 {
3178 return 0;
3179 }
3180
f2fs_quota_off_umount(struct super_block * sb)3181 void f2fs_quota_off_umount(struct super_block *sb)
3182 {
3183 }
3184 #endif
3185
3186 static const struct super_operations f2fs_sops = {
3187 .alloc_inode = f2fs_alloc_inode,
3188 .free_inode = f2fs_free_inode,
3189 .drop_inode = f2fs_drop_inode,
3190 .write_inode = f2fs_write_inode,
3191 .dirty_inode = f2fs_dirty_inode,
3192 .show_options = f2fs_show_options,
3193 #ifdef CONFIG_QUOTA
3194 .quota_read = f2fs_quota_read,
3195 .quota_write = f2fs_quota_write,
3196 .get_dquots = f2fs_get_dquots,
3197 #endif
3198 .evict_inode = f2fs_evict_inode,
3199 .put_super = f2fs_put_super,
3200 .sync_fs = f2fs_sync_fs,
3201 .freeze_fs = f2fs_freeze,
3202 .unfreeze_fs = f2fs_unfreeze,
3203 .statfs = f2fs_statfs,
3204 .remount_fs = f2fs_remount,
3205 };
3206
3207 #ifdef CONFIG_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)3208 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3209 {
3210 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3211 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3212 ctx, len, NULL);
3213 }
3214
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)3215 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3216 void *fs_data)
3217 {
3218 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3219
3220 /*
3221 * Encrypting the root directory is not allowed because fsck
3222 * expects lost+found directory to exist and remain unencrypted
3223 * if LOST_FOUND feature is enabled.
3224 *
3225 */
3226 if (f2fs_sb_has_lost_found(sbi) &&
3227 inode->i_ino == F2FS_ROOT_INO(sbi))
3228 return -EPERM;
3229
3230 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3231 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3232 ctx, len, fs_data, XATTR_CREATE);
3233 }
3234
f2fs_get_dummy_policy(struct super_block * sb)3235 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3236 {
3237 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3238 }
3239
f2fs_has_stable_inodes(struct super_block * sb)3240 static bool f2fs_has_stable_inodes(struct super_block *sb)
3241 {
3242 return true;
3243 }
3244
f2fs_get_ino_and_lblk_bits(struct super_block * sb,int * ino_bits_ret,int * lblk_bits_ret)3245 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3246 int *ino_bits_ret, int *lblk_bits_ret)
3247 {
3248 *ino_bits_ret = 8 * sizeof(nid_t);
3249 *lblk_bits_ret = 8 * sizeof(block_t);
3250 }
3251
f2fs_get_devices(struct super_block * sb,unsigned int * num_devs)3252 static struct block_device **f2fs_get_devices(struct super_block *sb,
3253 unsigned int *num_devs)
3254 {
3255 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3256 struct block_device **devs;
3257 int i;
3258
3259 if (!f2fs_is_multi_device(sbi))
3260 return NULL;
3261
3262 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3263 if (!devs)
3264 return ERR_PTR(-ENOMEM);
3265
3266 for (i = 0; i < sbi->s_ndevs; i++)
3267 devs[i] = FDEV(i).bdev;
3268 *num_devs = sbi->s_ndevs;
3269 return devs;
3270 }
3271
3272 static const struct fscrypt_operations f2fs_cryptops = {
3273 .key_prefix = "f2fs:",
3274 .get_context = f2fs_get_context,
3275 .set_context = f2fs_set_context,
3276 .get_dummy_policy = f2fs_get_dummy_policy,
3277 .empty_dir = f2fs_empty_dir,
3278 .has_stable_inodes = f2fs_has_stable_inodes,
3279 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
3280 .get_devices = f2fs_get_devices,
3281 };
3282 #endif
3283
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)3284 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3285 u64 ino, u32 generation)
3286 {
3287 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3288 struct inode *inode;
3289
3290 if (f2fs_check_nid_range(sbi, ino))
3291 return ERR_PTR(-ESTALE);
3292
3293 /*
3294 * f2fs_iget isn't quite right if the inode is currently unallocated!
3295 * However f2fs_iget currently does appropriate checks to handle stale
3296 * inodes so everything is OK.
3297 */
3298 inode = f2fs_iget(sb, ino);
3299 if (IS_ERR(inode))
3300 return ERR_CAST(inode);
3301 if (unlikely(generation && inode->i_generation != generation)) {
3302 /* we didn't find the right inode.. */
3303 iput(inode);
3304 return ERR_PTR(-ESTALE);
3305 }
3306 return inode;
3307 }
3308
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3309 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3310 int fh_len, int fh_type)
3311 {
3312 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3313 f2fs_nfs_get_inode);
3314 }
3315
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3316 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3317 int fh_len, int fh_type)
3318 {
3319 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3320 f2fs_nfs_get_inode);
3321 }
3322
3323 static const struct export_operations f2fs_export_ops = {
3324 .fh_to_dentry = f2fs_fh_to_dentry,
3325 .fh_to_parent = f2fs_fh_to_parent,
3326 .get_parent = f2fs_get_parent,
3327 };
3328
max_file_blocks(struct inode * inode)3329 loff_t max_file_blocks(struct inode *inode)
3330 {
3331 loff_t result = 0;
3332 loff_t leaf_count;
3333
3334 /*
3335 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3336 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3337 * space in inode.i_addr, it will be more safe to reassign
3338 * result as zero.
3339 */
3340
3341 if (inode && f2fs_compressed_file(inode))
3342 leaf_count = ADDRS_PER_BLOCK(inode);
3343 else
3344 leaf_count = DEF_ADDRS_PER_BLOCK;
3345
3346 /* two direct node blocks */
3347 result += (leaf_count * 2);
3348
3349 /* two indirect node blocks */
3350 leaf_count *= NIDS_PER_BLOCK;
3351 result += (leaf_count * 2);
3352
3353 /* one double indirect node block */
3354 leaf_count *= NIDS_PER_BLOCK;
3355 result += leaf_count;
3356
3357 return result;
3358 }
3359
__f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)3360 static int __f2fs_commit_super(struct buffer_head *bh,
3361 struct f2fs_super_block *super)
3362 {
3363 lock_buffer(bh);
3364 if (super)
3365 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3366 set_buffer_dirty(bh);
3367 unlock_buffer(bh);
3368
3369 /* it's rare case, we can do fua all the time */
3370 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3371 }
3372
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)3373 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3374 struct buffer_head *bh)
3375 {
3376 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3377 (bh->b_data + F2FS_SUPER_OFFSET);
3378 struct super_block *sb = sbi->sb;
3379 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3380 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3381 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3382 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3383 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3384 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3385 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3386 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3387 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3388 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3389 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3390 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3391 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3392 u64 main_end_blkaddr = main_blkaddr +
3393 (segment_count_main << log_blocks_per_seg);
3394 u64 seg_end_blkaddr = segment0_blkaddr +
3395 (segment_count << log_blocks_per_seg);
3396
3397 if (segment0_blkaddr != cp_blkaddr) {
3398 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3399 segment0_blkaddr, cp_blkaddr);
3400 return true;
3401 }
3402
3403 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3404 sit_blkaddr) {
3405 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3406 cp_blkaddr, sit_blkaddr,
3407 segment_count_ckpt << log_blocks_per_seg);
3408 return true;
3409 }
3410
3411 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3412 nat_blkaddr) {
3413 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3414 sit_blkaddr, nat_blkaddr,
3415 segment_count_sit << log_blocks_per_seg);
3416 return true;
3417 }
3418
3419 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3420 ssa_blkaddr) {
3421 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3422 nat_blkaddr, ssa_blkaddr,
3423 segment_count_nat << log_blocks_per_seg);
3424 return true;
3425 }
3426
3427 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3428 main_blkaddr) {
3429 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3430 ssa_blkaddr, main_blkaddr,
3431 segment_count_ssa << log_blocks_per_seg);
3432 return true;
3433 }
3434
3435 if (main_end_blkaddr > seg_end_blkaddr) {
3436 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3437 main_blkaddr, seg_end_blkaddr,
3438 segment_count_main << log_blocks_per_seg);
3439 return true;
3440 } else if (main_end_blkaddr < seg_end_blkaddr) {
3441 int err = 0;
3442 char *res;
3443
3444 /* fix in-memory information all the time */
3445 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3446 segment0_blkaddr) >> log_blocks_per_seg);
3447
3448 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3449 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3450 res = "internally";
3451 } else {
3452 err = __f2fs_commit_super(bh, NULL);
3453 res = err ? "failed" : "done";
3454 }
3455 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3456 res, main_blkaddr, seg_end_blkaddr,
3457 segment_count_main << log_blocks_per_seg);
3458 if (err)
3459 return true;
3460 }
3461 return false;
3462 }
3463
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)3464 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3465 struct buffer_head *bh)
3466 {
3467 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3468 block_t total_sections, blocks_per_seg;
3469 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3470 (bh->b_data + F2FS_SUPER_OFFSET);
3471 size_t crc_offset = 0;
3472 __u32 crc = 0;
3473
3474 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3475 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3476 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3477 return -EINVAL;
3478 }
3479
3480 /* Check checksum_offset and crc in superblock */
3481 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3482 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3483 if (crc_offset !=
3484 offsetof(struct f2fs_super_block, crc)) {
3485 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3486 crc_offset);
3487 return -EFSCORRUPTED;
3488 }
3489 crc = le32_to_cpu(raw_super->crc);
3490 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3491 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3492 return -EFSCORRUPTED;
3493 }
3494 }
3495
3496 /* Currently, support only 4KB block size */
3497 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3498 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3499 le32_to_cpu(raw_super->log_blocksize),
3500 F2FS_BLKSIZE_BITS);
3501 return -EFSCORRUPTED;
3502 }
3503
3504 /* check log blocks per segment */
3505 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3506 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3507 le32_to_cpu(raw_super->log_blocks_per_seg));
3508 return -EFSCORRUPTED;
3509 }
3510
3511 /* Currently, support 512/1024/2048/4096 bytes sector size */
3512 if (le32_to_cpu(raw_super->log_sectorsize) >
3513 F2FS_MAX_LOG_SECTOR_SIZE ||
3514 le32_to_cpu(raw_super->log_sectorsize) <
3515 F2FS_MIN_LOG_SECTOR_SIZE) {
3516 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3517 le32_to_cpu(raw_super->log_sectorsize));
3518 return -EFSCORRUPTED;
3519 }
3520 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3521 le32_to_cpu(raw_super->log_sectorsize) !=
3522 F2FS_MAX_LOG_SECTOR_SIZE) {
3523 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3524 le32_to_cpu(raw_super->log_sectors_per_block),
3525 le32_to_cpu(raw_super->log_sectorsize));
3526 return -EFSCORRUPTED;
3527 }
3528
3529 segment_count = le32_to_cpu(raw_super->segment_count);
3530 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3531 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3532 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3533 total_sections = le32_to_cpu(raw_super->section_count);
3534
3535 /* blocks_per_seg should be 512, given the above check */
3536 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3537
3538 if (segment_count > F2FS_MAX_SEGMENT ||
3539 segment_count < F2FS_MIN_SEGMENTS) {
3540 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3541 return -EFSCORRUPTED;
3542 }
3543
3544 if (total_sections > segment_count_main || total_sections < 1 ||
3545 segs_per_sec > segment_count || !segs_per_sec) {
3546 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3547 segment_count, total_sections, segs_per_sec);
3548 return -EFSCORRUPTED;
3549 }
3550
3551 if (segment_count_main != total_sections * segs_per_sec) {
3552 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3553 segment_count_main, total_sections, segs_per_sec);
3554 return -EFSCORRUPTED;
3555 }
3556
3557 if ((segment_count / segs_per_sec) < total_sections) {
3558 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3559 segment_count, segs_per_sec, total_sections);
3560 return -EFSCORRUPTED;
3561 }
3562
3563 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3564 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3565 segment_count, le64_to_cpu(raw_super->block_count));
3566 return -EFSCORRUPTED;
3567 }
3568
3569 if (RDEV(0).path[0]) {
3570 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3571 int i = 1;
3572
3573 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3574 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3575 i++;
3576 }
3577 if (segment_count != dev_seg_count) {
3578 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3579 segment_count, dev_seg_count);
3580 return -EFSCORRUPTED;
3581 }
3582 } else {
3583 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3584 !bdev_is_zoned(sbi->sb->s_bdev)) {
3585 f2fs_info(sbi, "Zoned block device path is missing");
3586 return -EFSCORRUPTED;
3587 }
3588 }
3589
3590 if (secs_per_zone > total_sections || !secs_per_zone) {
3591 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3592 secs_per_zone, total_sections);
3593 return -EFSCORRUPTED;
3594 }
3595 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3596 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3597 (le32_to_cpu(raw_super->extension_count) +
3598 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3599 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3600 le32_to_cpu(raw_super->extension_count),
3601 raw_super->hot_ext_count,
3602 F2FS_MAX_EXTENSION);
3603 return -EFSCORRUPTED;
3604 }
3605
3606 if (le32_to_cpu(raw_super->cp_payload) >=
3607 (blocks_per_seg - F2FS_CP_PACKS -
3608 NR_CURSEG_PERSIST_TYPE)) {
3609 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3610 le32_to_cpu(raw_super->cp_payload),
3611 blocks_per_seg - F2FS_CP_PACKS -
3612 NR_CURSEG_PERSIST_TYPE);
3613 return -EFSCORRUPTED;
3614 }
3615
3616 /* check reserved ino info */
3617 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3618 le32_to_cpu(raw_super->meta_ino) != 2 ||
3619 le32_to_cpu(raw_super->root_ino) != 3) {
3620 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3621 le32_to_cpu(raw_super->node_ino),
3622 le32_to_cpu(raw_super->meta_ino),
3623 le32_to_cpu(raw_super->root_ino));
3624 return -EFSCORRUPTED;
3625 }
3626
3627 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3628 if (sanity_check_area_boundary(sbi, bh))
3629 return -EFSCORRUPTED;
3630
3631 return 0;
3632 }
3633
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3634 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3635 {
3636 unsigned int total, fsmeta;
3637 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3638 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3639 unsigned int ovp_segments, reserved_segments;
3640 unsigned int main_segs, blocks_per_seg;
3641 unsigned int sit_segs, nat_segs;
3642 unsigned int sit_bitmap_size, nat_bitmap_size;
3643 unsigned int log_blocks_per_seg;
3644 unsigned int segment_count_main;
3645 unsigned int cp_pack_start_sum, cp_payload;
3646 block_t user_block_count, valid_user_blocks;
3647 block_t avail_node_count, valid_node_count;
3648 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3649 int i, j;
3650
3651 total = le32_to_cpu(raw_super->segment_count);
3652 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3653 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3654 fsmeta += sit_segs;
3655 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3656 fsmeta += nat_segs;
3657 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3658 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3659
3660 if (unlikely(fsmeta >= total))
3661 return 1;
3662
3663 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3664 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3665
3666 if (!f2fs_sb_has_readonly(sbi) &&
3667 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3668 ovp_segments == 0 || reserved_segments == 0)) {
3669 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3670 return 1;
3671 }
3672 user_block_count = le64_to_cpu(ckpt->user_block_count);
3673 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3674 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3675 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3676 if (!user_block_count || user_block_count >=
3677 segment_count_main << log_blocks_per_seg) {
3678 f2fs_err(sbi, "Wrong user_block_count: %u",
3679 user_block_count);
3680 return 1;
3681 }
3682
3683 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3684 if (valid_user_blocks > user_block_count) {
3685 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3686 valid_user_blocks, user_block_count);
3687 return 1;
3688 }
3689
3690 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3691 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3692 if (valid_node_count > avail_node_count) {
3693 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3694 valid_node_count, avail_node_count);
3695 return 1;
3696 }
3697
3698 main_segs = le32_to_cpu(raw_super->segment_count_main);
3699 blocks_per_seg = sbi->blocks_per_seg;
3700
3701 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3702 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3703 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3704 return 1;
3705
3706 if (f2fs_sb_has_readonly(sbi))
3707 goto check_data;
3708
3709 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3710 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3711 le32_to_cpu(ckpt->cur_node_segno[j])) {
3712 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3713 i, j,
3714 le32_to_cpu(ckpt->cur_node_segno[i]));
3715 return 1;
3716 }
3717 }
3718 }
3719 check_data:
3720 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3721 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3722 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3723 return 1;
3724
3725 if (f2fs_sb_has_readonly(sbi))
3726 goto skip_cross;
3727
3728 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3729 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3730 le32_to_cpu(ckpt->cur_data_segno[j])) {
3731 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3732 i, j,
3733 le32_to_cpu(ckpt->cur_data_segno[i]));
3734 return 1;
3735 }
3736 }
3737 }
3738 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3739 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3740 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3741 le32_to_cpu(ckpt->cur_data_segno[j])) {
3742 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3743 i, j,
3744 le32_to_cpu(ckpt->cur_node_segno[i]));
3745 return 1;
3746 }
3747 }
3748 }
3749 skip_cross:
3750 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3751 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3752
3753 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3754 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3755 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3756 sit_bitmap_size, nat_bitmap_size);
3757 return 1;
3758 }
3759
3760 cp_pack_start_sum = __start_sum_addr(sbi);
3761 cp_payload = __cp_payload(sbi);
3762 if (cp_pack_start_sum < cp_payload + 1 ||
3763 cp_pack_start_sum > blocks_per_seg - 1 -
3764 NR_CURSEG_PERSIST_TYPE) {
3765 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3766 cp_pack_start_sum);
3767 return 1;
3768 }
3769
3770 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3771 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3772 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3773 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3774 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3775 le32_to_cpu(ckpt->checksum_offset));
3776 return 1;
3777 }
3778
3779 nat_blocks = nat_segs << log_blocks_per_seg;
3780 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3781 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3782 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3783 (cp_payload + F2FS_CP_PACKS +
3784 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3785 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3786 cp_payload, nat_bits_blocks);
3787 return 1;
3788 }
3789
3790 if (unlikely(f2fs_cp_error(sbi))) {
3791 f2fs_err(sbi, "A bug case: need to run fsck");
3792 return 1;
3793 }
3794 return 0;
3795 }
3796
init_sb_info(struct f2fs_sb_info * sbi)3797 static void init_sb_info(struct f2fs_sb_info *sbi)
3798 {
3799 struct f2fs_super_block *raw_super = sbi->raw_super;
3800 int i;
3801
3802 sbi->log_sectors_per_block =
3803 le32_to_cpu(raw_super->log_sectors_per_block);
3804 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3805 sbi->blocksize = BIT(sbi->log_blocksize);
3806 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3807 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3808 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3809 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3810 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3811 sbi->total_node_count =
3812 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3813 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3814 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3815 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3816 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3817 sbi->cur_victim_sec = NULL_SECNO;
3818 sbi->gc_mode = GC_NORMAL;
3819 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3820 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3821 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3822 sbi->migration_granularity = sbi->segs_per_sec;
3823 sbi->seq_file_ra_mul = MIN_RA_MUL;
3824 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3825 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3826 spin_lock_init(&sbi->gc_remaining_trials_lock);
3827 atomic64_set(&sbi->current_atomic_write, 0);
3828
3829 sbi->dir_level = DEF_DIR_LEVEL;
3830 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3831 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3832 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3833 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3834 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3835 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3836 DEF_UMOUNT_DISCARD_TIMEOUT;
3837 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3838
3839 for (i = 0; i < NR_COUNT_TYPE; i++)
3840 atomic_set(&sbi->nr_pages[i], 0);
3841
3842 for (i = 0; i < META; i++)
3843 atomic_set(&sbi->wb_sync_req[i], 0);
3844
3845 INIT_LIST_HEAD(&sbi->s_list);
3846 mutex_init(&sbi->umount_mutex);
3847 init_f2fs_rwsem(&sbi->io_order_lock);
3848 spin_lock_init(&sbi->cp_lock);
3849
3850 sbi->dirty_device = 0;
3851 spin_lock_init(&sbi->dev_lock);
3852
3853 init_f2fs_rwsem(&sbi->sb_lock);
3854 init_f2fs_rwsem(&sbi->pin_sem);
3855 }
3856
init_percpu_info(struct f2fs_sb_info * sbi)3857 static int init_percpu_info(struct f2fs_sb_info *sbi)
3858 {
3859 int err;
3860
3861 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3862 if (err)
3863 return err;
3864
3865 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3866 if (err)
3867 goto err_valid_block;
3868
3869 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3870 GFP_KERNEL);
3871 if (err)
3872 goto err_node_block;
3873 return 0;
3874
3875 err_node_block:
3876 percpu_counter_destroy(&sbi->rf_node_block_count);
3877 err_valid_block:
3878 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3879 return err;
3880 }
3881
3882 #ifdef CONFIG_BLK_DEV_ZONED
3883
3884 struct f2fs_report_zones_args {
3885 struct f2fs_sb_info *sbi;
3886 struct f2fs_dev_info *dev;
3887 };
3888
f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3889 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3890 void *data)
3891 {
3892 struct f2fs_report_zones_args *rz_args = data;
3893 block_t unusable_blocks = (zone->len - zone->capacity) >>
3894 F2FS_LOG_SECTORS_PER_BLOCK;
3895
3896 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3897 return 0;
3898
3899 set_bit(idx, rz_args->dev->blkz_seq);
3900 if (!rz_args->sbi->unusable_blocks_per_sec) {
3901 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3902 return 0;
3903 }
3904 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3905 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3906 return -EINVAL;
3907 }
3908 return 0;
3909 }
3910
init_blkz_info(struct f2fs_sb_info * sbi,int devi)3911 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3912 {
3913 struct block_device *bdev = FDEV(devi).bdev;
3914 sector_t nr_sectors = bdev_nr_sectors(bdev);
3915 struct f2fs_report_zones_args rep_zone_arg;
3916 u64 zone_sectors;
3917 int ret;
3918
3919 if (!f2fs_sb_has_blkzoned(sbi))
3920 return 0;
3921
3922 zone_sectors = bdev_zone_sectors(bdev);
3923 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3924 SECTOR_TO_BLOCK(zone_sectors))
3925 return -EINVAL;
3926 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3927 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3928 sbi->blocks_per_blkz);
3929 if (nr_sectors & (zone_sectors - 1))
3930 FDEV(devi).nr_blkz++;
3931
3932 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3933 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3934 * sizeof(unsigned long),
3935 GFP_KERNEL);
3936 if (!FDEV(devi).blkz_seq)
3937 return -ENOMEM;
3938
3939 rep_zone_arg.sbi = sbi;
3940 rep_zone_arg.dev = &FDEV(devi);
3941
3942 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3943 &rep_zone_arg);
3944 if (ret < 0)
3945 return ret;
3946 return 0;
3947 }
3948 #endif
3949
3950 /*
3951 * Read f2fs raw super block.
3952 * Because we have two copies of super block, so read both of them
3953 * to get the first valid one. If any one of them is broken, we pass
3954 * them recovery flag back to the caller.
3955 */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3956 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3957 struct f2fs_super_block **raw_super,
3958 int *valid_super_block, int *recovery)
3959 {
3960 struct super_block *sb = sbi->sb;
3961 int block;
3962 struct buffer_head *bh;
3963 struct f2fs_super_block *super;
3964 int err = 0;
3965
3966 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3967 if (!super)
3968 return -ENOMEM;
3969
3970 for (block = 0; block < 2; block++) {
3971 bh = sb_bread(sb, block);
3972 if (!bh) {
3973 f2fs_err(sbi, "Unable to read %dth superblock",
3974 block + 1);
3975 err = -EIO;
3976 *recovery = 1;
3977 continue;
3978 }
3979
3980 /* sanity checking of raw super */
3981 err = sanity_check_raw_super(sbi, bh);
3982 if (err) {
3983 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3984 block + 1);
3985 brelse(bh);
3986 *recovery = 1;
3987 continue;
3988 }
3989
3990 if (!*raw_super) {
3991 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3992 sizeof(*super));
3993 *valid_super_block = block;
3994 *raw_super = super;
3995 }
3996 brelse(bh);
3997 }
3998
3999 /* No valid superblock */
4000 if (!*raw_super)
4001 kfree(super);
4002 else
4003 err = 0;
4004
4005 return err;
4006 }
4007
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)4008 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
4009 {
4010 struct buffer_head *bh;
4011 __u32 crc = 0;
4012 int err;
4013
4014 if ((recover && f2fs_readonly(sbi->sb)) ||
4015 f2fs_hw_is_readonly(sbi)) {
4016 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
4017 return -EROFS;
4018 }
4019
4020 /* we should update superblock crc here */
4021 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
4022 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
4023 offsetof(struct f2fs_super_block, crc));
4024 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
4025 }
4026
4027 /* write back-up superblock first */
4028 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
4029 if (!bh)
4030 return -EIO;
4031 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4032 brelse(bh);
4033
4034 /* if we are in recovery path, skip writing valid superblock */
4035 if (recover || err)
4036 return err;
4037
4038 /* write current valid superblock */
4039 bh = sb_bread(sbi->sb, sbi->valid_super_block);
4040 if (!bh)
4041 return -EIO;
4042 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4043 brelse(bh);
4044 return err;
4045 }
4046
save_stop_reason(struct f2fs_sb_info * sbi,unsigned char reason)4047 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4048 {
4049 unsigned long flags;
4050
4051 spin_lock_irqsave(&sbi->error_lock, flags);
4052 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4053 sbi->stop_reason[reason]++;
4054 spin_unlock_irqrestore(&sbi->error_lock, flags);
4055 }
4056
f2fs_record_stop_reason(struct f2fs_sb_info * sbi)4057 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4058 {
4059 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4060 unsigned long flags;
4061 int err;
4062
4063 f2fs_down_write(&sbi->sb_lock);
4064
4065 spin_lock_irqsave(&sbi->error_lock, flags);
4066 if (sbi->error_dirty) {
4067 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4068 MAX_F2FS_ERRORS);
4069 sbi->error_dirty = false;
4070 }
4071 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4072 spin_unlock_irqrestore(&sbi->error_lock, flags);
4073
4074 err = f2fs_commit_super(sbi, false);
4075
4076 f2fs_up_write(&sbi->sb_lock);
4077 if (err)
4078 f2fs_err(sbi, "f2fs_commit_super fails to record err:%d", err);
4079 }
4080
f2fs_save_errors(struct f2fs_sb_info * sbi,unsigned char flag)4081 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4082 {
4083 unsigned long flags;
4084
4085 spin_lock_irqsave(&sbi->error_lock, flags);
4086 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4087 set_bit(flag, (unsigned long *)sbi->errors);
4088 sbi->error_dirty = true;
4089 }
4090 spin_unlock_irqrestore(&sbi->error_lock, flags);
4091 }
4092
f2fs_update_errors(struct f2fs_sb_info * sbi)4093 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4094 {
4095 unsigned long flags;
4096 bool need_update = false;
4097
4098 spin_lock_irqsave(&sbi->error_lock, flags);
4099 if (sbi->error_dirty) {
4100 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4101 MAX_F2FS_ERRORS);
4102 sbi->error_dirty = false;
4103 need_update = true;
4104 }
4105 spin_unlock_irqrestore(&sbi->error_lock, flags);
4106
4107 return need_update;
4108 }
4109
f2fs_record_errors(struct f2fs_sb_info * sbi,unsigned char error)4110 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4111 {
4112 int err;
4113
4114 f2fs_down_write(&sbi->sb_lock);
4115
4116 if (!f2fs_update_errors(sbi))
4117 goto out_unlock;
4118
4119 err = f2fs_commit_super(sbi, false);
4120 if (err)
4121 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
4122 error, err);
4123 out_unlock:
4124 f2fs_up_write(&sbi->sb_lock);
4125 }
4126
f2fs_handle_error(struct f2fs_sb_info * sbi,unsigned char error)4127 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4128 {
4129 f2fs_save_errors(sbi, error);
4130 f2fs_record_errors(sbi, error);
4131 }
4132
f2fs_handle_error_async(struct f2fs_sb_info * sbi,unsigned char error)4133 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4134 {
4135 f2fs_save_errors(sbi, error);
4136
4137 if (!sbi->error_dirty)
4138 return;
4139 if (!test_bit(error, (unsigned long *)sbi->errors))
4140 return;
4141 schedule_work(&sbi->s_error_work);
4142 }
4143
system_going_down(void)4144 static bool system_going_down(void)
4145 {
4146 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4147 || system_state == SYSTEM_RESTART;
4148 }
4149
f2fs_handle_critical_error(struct f2fs_sb_info * sbi,unsigned char reason,bool irq_context)4150 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason,
4151 bool irq_context)
4152 {
4153 struct super_block *sb = sbi->sb;
4154 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4155 bool continue_fs = !shutdown &&
4156 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4157
4158 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4159
4160 if (!f2fs_hw_is_readonly(sbi)) {
4161 save_stop_reason(sbi, reason);
4162
4163 if (irq_context && !shutdown)
4164 schedule_work(&sbi->s_error_work);
4165 else
4166 f2fs_record_stop_reason(sbi);
4167 }
4168
4169 /*
4170 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4171 * could panic during 'reboot -f' as the underlying device got already
4172 * disabled.
4173 */
4174 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4175 !shutdown && !system_going_down() &&
4176 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4177 panic("F2FS-fs (device %s): panic forced after error\n",
4178 sb->s_id);
4179
4180 if (shutdown)
4181 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4182
4183 /* continue filesystem operators if errors=continue */
4184 if (continue_fs || f2fs_readonly(sb))
4185 return;
4186
4187 f2fs_warn(sbi, "Remounting filesystem read-only");
4188 /*
4189 * Make sure updated value of ->s_mount_flags will be visible before
4190 * ->s_flags update
4191 */
4192 smp_wmb();
4193 sb->s_flags |= SB_RDONLY;
4194 }
4195
f2fs_record_error_work(struct work_struct * work)4196 static void f2fs_record_error_work(struct work_struct *work)
4197 {
4198 struct f2fs_sb_info *sbi = container_of(work,
4199 struct f2fs_sb_info, s_error_work);
4200
4201 f2fs_record_stop_reason(sbi);
4202 }
4203
f2fs_scan_devices(struct f2fs_sb_info * sbi)4204 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4205 {
4206 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4207 unsigned int max_devices = MAX_DEVICES;
4208 unsigned int logical_blksize;
4209 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4210 int i;
4211
4212 /* Initialize single device information */
4213 if (!RDEV(0).path[0]) {
4214 if (!bdev_is_zoned(sbi->sb->s_bdev))
4215 return 0;
4216 max_devices = 1;
4217 }
4218
4219 /*
4220 * Initialize multiple devices information, or single
4221 * zoned block device information.
4222 */
4223 sbi->devs = f2fs_kzalloc(sbi,
4224 array_size(max_devices,
4225 sizeof(struct f2fs_dev_info)),
4226 GFP_KERNEL);
4227 if (!sbi->devs)
4228 return -ENOMEM;
4229
4230 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4231 sbi->aligned_blksize = true;
4232
4233 for (i = 0; i < max_devices; i++) {
4234 if (i == 0)
4235 FDEV(0).bdev = sbi->sb->s_bdev;
4236 else if (!RDEV(i).path[0])
4237 break;
4238
4239 if (max_devices > 1) {
4240 /* Multi-device mount */
4241 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4242 FDEV(i).total_segments =
4243 le32_to_cpu(RDEV(i).total_segments);
4244 if (i == 0) {
4245 FDEV(i).start_blk = 0;
4246 FDEV(i).end_blk = FDEV(i).start_blk +
4247 (FDEV(i).total_segments <<
4248 sbi->log_blocks_per_seg) - 1 +
4249 le32_to_cpu(raw_super->segment0_blkaddr);
4250 } else {
4251 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4252 FDEV(i).end_blk = FDEV(i).start_blk +
4253 (FDEV(i).total_segments <<
4254 sbi->log_blocks_per_seg) - 1;
4255 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
4256 mode, sbi->sb, NULL);
4257 }
4258 }
4259 if (IS_ERR(FDEV(i).bdev))
4260 return PTR_ERR(FDEV(i).bdev);
4261
4262 /* to release errored devices */
4263 sbi->s_ndevs = i + 1;
4264
4265 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4266 sbi->aligned_blksize = false;
4267
4268 #ifdef CONFIG_BLK_DEV_ZONED
4269 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
4270 !f2fs_sb_has_blkzoned(sbi)) {
4271 f2fs_err(sbi, "Zoned block device feature not enabled");
4272 return -EINVAL;
4273 }
4274 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
4275 if (init_blkz_info(sbi, i)) {
4276 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4277 return -EINVAL;
4278 }
4279 if (max_devices == 1)
4280 break;
4281 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
4282 i, FDEV(i).path,
4283 FDEV(i).total_segments,
4284 FDEV(i).start_blk, FDEV(i).end_blk,
4285 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
4286 "Host-aware" : "Host-managed");
4287 continue;
4288 }
4289 #endif
4290 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4291 i, FDEV(i).path,
4292 FDEV(i).total_segments,
4293 FDEV(i).start_blk, FDEV(i).end_blk);
4294 }
4295 f2fs_info(sbi,
4296 "IO Block Size: %8ld KB", F2FS_IO_SIZE_KB(sbi));
4297 return 0;
4298 }
4299
f2fs_setup_casefold(struct f2fs_sb_info * sbi)4300 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4301 {
4302 #if IS_ENABLED(CONFIG_UNICODE)
4303 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4304 const struct f2fs_sb_encodings *encoding_info;
4305 struct unicode_map *encoding;
4306 __u16 encoding_flags;
4307
4308 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4309 if (!encoding_info) {
4310 f2fs_err(sbi,
4311 "Encoding requested by superblock is unknown");
4312 return -EINVAL;
4313 }
4314
4315 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4316 encoding = utf8_load(encoding_info->version);
4317 if (IS_ERR(encoding)) {
4318 f2fs_err(sbi,
4319 "can't mount with superblock charset: %s-%u.%u.%u "
4320 "not supported by the kernel. flags: 0x%x.",
4321 encoding_info->name,
4322 unicode_major(encoding_info->version),
4323 unicode_minor(encoding_info->version),
4324 unicode_rev(encoding_info->version),
4325 encoding_flags);
4326 return PTR_ERR(encoding);
4327 }
4328 f2fs_info(sbi, "Using encoding defined by superblock: "
4329 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4330 unicode_major(encoding_info->version),
4331 unicode_minor(encoding_info->version),
4332 unicode_rev(encoding_info->version),
4333 encoding_flags);
4334
4335 sbi->sb->s_encoding = encoding;
4336 sbi->sb->s_encoding_flags = encoding_flags;
4337 }
4338 #else
4339 if (f2fs_sb_has_casefold(sbi)) {
4340 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4341 return -EINVAL;
4342 }
4343 #endif
4344 return 0;
4345 }
4346
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)4347 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4348 {
4349 /* adjust parameters according to the volume size */
4350 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4351 if (f2fs_block_unit_discard(sbi))
4352 SM_I(sbi)->dcc_info->discard_granularity =
4353 MIN_DISCARD_GRANULARITY;
4354 if (!f2fs_lfs_mode(sbi))
4355 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4356 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4357 }
4358
4359 sbi->readdir_ra = true;
4360 }
4361
f2fs_fill_super(struct super_block * sb,void * data,int silent)4362 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4363 {
4364 struct f2fs_sb_info *sbi;
4365 struct f2fs_super_block *raw_super;
4366 struct inode *root;
4367 int err;
4368 bool skip_recovery = false, need_fsck = false;
4369 char *options = NULL;
4370 int recovery, i, valid_super_block;
4371 struct curseg_info *seg_i;
4372 int retry_cnt = 1;
4373 #ifdef CONFIG_QUOTA
4374 bool quota_enabled = false;
4375 #endif
4376
4377 try_onemore:
4378 err = -EINVAL;
4379 raw_super = NULL;
4380 valid_super_block = -1;
4381 recovery = 0;
4382
4383 /* allocate memory for f2fs-specific super block info */
4384 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4385 if (!sbi)
4386 return -ENOMEM;
4387
4388 sbi->sb = sb;
4389
4390 /* initialize locks within allocated memory */
4391 init_f2fs_rwsem(&sbi->gc_lock);
4392 mutex_init(&sbi->writepages);
4393 init_f2fs_rwsem(&sbi->cp_global_sem);
4394 init_f2fs_rwsem(&sbi->node_write);
4395 init_f2fs_rwsem(&sbi->node_change);
4396 spin_lock_init(&sbi->stat_lock);
4397 init_f2fs_rwsem(&sbi->cp_rwsem);
4398 init_f2fs_rwsem(&sbi->quota_sem);
4399 init_waitqueue_head(&sbi->cp_wait);
4400 spin_lock_init(&sbi->error_lock);
4401
4402 for (i = 0; i < NR_INODE_TYPE; i++) {
4403 INIT_LIST_HEAD(&sbi->inode_list[i]);
4404 spin_lock_init(&sbi->inode_lock[i]);
4405 }
4406 mutex_init(&sbi->flush_lock);
4407
4408 /* Load the checksum driver */
4409 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4410 if (IS_ERR(sbi->s_chksum_driver)) {
4411 f2fs_err(sbi, "Cannot load crc32 driver.");
4412 err = PTR_ERR(sbi->s_chksum_driver);
4413 sbi->s_chksum_driver = NULL;
4414 goto free_sbi;
4415 }
4416
4417 /* set a block size */
4418 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4419 f2fs_err(sbi, "unable to set blocksize");
4420 goto free_sbi;
4421 }
4422
4423 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4424 &recovery);
4425 if (err)
4426 goto free_sbi;
4427
4428 sb->s_fs_info = sbi;
4429 sbi->raw_super = raw_super;
4430
4431 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4432 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4433 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4434
4435 /* precompute checksum seed for metadata */
4436 if (f2fs_sb_has_inode_chksum(sbi))
4437 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4438 sizeof(raw_super->uuid));
4439
4440 default_options(sbi, false);
4441 /* parse mount options */
4442 options = kstrdup((const char *)data, GFP_KERNEL);
4443 if (data && !options) {
4444 err = -ENOMEM;
4445 goto free_sb_buf;
4446 }
4447
4448 err = parse_options(sb, options, false);
4449 if (err)
4450 goto free_options;
4451
4452 sb->s_maxbytes = max_file_blocks(NULL) <<
4453 le32_to_cpu(raw_super->log_blocksize);
4454 sb->s_max_links = F2FS_LINK_MAX;
4455
4456 err = f2fs_setup_casefold(sbi);
4457 if (err)
4458 goto free_options;
4459
4460 #ifdef CONFIG_QUOTA
4461 sb->dq_op = &f2fs_quota_operations;
4462 sb->s_qcop = &f2fs_quotactl_ops;
4463 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4464
4465 if (f2fs_sb_has_quota_ino(sbi)) {
4466 for (i = 0; i < MAXQUOTAS; i++) {
4467 if (f2fs_qf_ino(sbi->sb, i))
4468 sbi->nquota_files++;
4469 }
4470 }
4471 #endif
4472
4473 sb->s_op = &f2fs_sops;
4474 #ifdef CONFIG_FS_ENCRYPTION
4475 sb->s_cop = &f2fs_cryptops;
4476 #endif
4477 #ifdef CONFIG_FS_VERITY
4478 sb->s_vop = &f2fs_verityops;
4479 #endif
4480 sb->s_xattr = f2fs_xattr_handlers;
4481 sb->s_export_op = &f2fs_export_ops;
4482 sb->s_magic = F2FS_SUPER_MAGIC;
4483 sb->s_time_gran = 1;
4484 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4485 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4486 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4487 sb->s_iflags |= SB_I_CGROUPWB;
4488
4489 /* init f2fs-specific super block info */
4490 sbi->valid_super_block = valid_super_block;
4491
4492 /* disallow all the data/node/meta page writes */
4493 set_sbi_flag(sbi, SBI_POR_DOING);
4494
4495 err = f2fs_init_write_merge_io(sbi);
4496 if (err)
4497 goto free_bio_info;
4498
4499 init_sb_info(sbi);
4500
4501 err = f2fs_init_iostat(sbi);
4502 if (err)
4503 goto free_bio_info;
4504
4505 err = init_percpu_info(sbi);
4506 if (err)
4507 goto free_iostat;
4508
4509 if (F2FS_IO_ALIGNED(sbi)) {
4510 sbi->write_io_dummy =
4511 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4512 if (!sbi->write_io_dummy) {
4513 err = -ENOMEM;
4514 goto free_percpu;
4515 }
4516 }
4517
4518 /* init per sbi slab cache */
4519 err = f2fs_init_xattr_caches(sbi);
4520 if (err)
4521 goto free_io_dummy;
4522 err = f2fs_init_page_array_cache(sbi);
4523 if (err)
4524 goto free_xattr_cache;
4525
4526 /* get an inode for meta space */
4527 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4528 if (IS_ERR(sbi->meta_inode)) {
4529 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4530 err = PTR_ERR(sbi->meta_inode);
4531 goto free_page_array_cache;
4532 }
4533
4534 err = f2fs_get_valid_checkpoint(sbi);
4535 if (err) {
4536 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4537 goto free_meta_inode;
4538 }
4539
4540 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4541 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4542 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4543 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4544 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4545 }
4546
4547 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4548 set_sbi_flag(sbi, SBI_NEED_FSCK);
4549
4550 /* Initialize device list */
4551 err = f2fs_scan_devices(sbi);
4552 if (err) {
4553 f2fs_err(sbi, "Failed to find devices");
4554 goto free_devices;
4555 }
4556
4557 err = f2fs_init_post_read_wq(sbi);
4558 if (err) {
4559 f2fs_err(sbi, "Failed to initialize post read workqueue");
4560 goto free_devices;
4561 }
4562
4563 sbi->total_valid_node_count =
4564 le32_to_cpu(sbi->ckpt->valid_node_count);
4565 percpu_counter_set(&sbi->total_valid_inode_count,
4566 le32_to_cpu(sbi->ckpt->valid_inode_count));
4567 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4568 sbi->total_valid_block_count =
4569 le64_to_cpu(sbi->ckpt->valid_block_count);
4570 sbi->last_valid_block_count = sbi->total_valid_block_count;
4571 sbi->reserved_blocks = 0;
4572 sbi->current_reserved_blocks = 0;
4573 limit_reserve_root(sbi);
4574 adjust_unusable_cap_perc(sbi);
4575
4576 f2fs_init_extent_cache_info(sbi);
4577
4578 f2fs_init_ino_entry_info(sbi);
4579
4580 f2fs_init_fsync_node_info(sbi);
4581
4582 /* setup checkpoint request control and start checkpoint issue thread */
4583 f2fs_init_ckpt_req_control(sbi);
4584 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4585 test_opt(sbi, MERGE_CHECKPOINT)) {
4586 err = f2fs_start_ckpt_thread(sbi);
4587 if (err) {
4588 f2fs_err(sbi,
4589 "Failed to start F2FS issue_checkpoint_thread (%d)",
4590 err);
4591 goto stop_ckpt_thread;
4592 }
4593 }
4594
4595 /* setup f2fs internal modules */
4596 err = f2fs_build_segment_manager(sbi);
4597 if (err) {
4598 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4599 err);
4600 goto free_sm;
4601 }
4602 err = f2fs_build_node_manager(sbi);
4603 if (err) {
4604 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4605 err);
4606 goto free_nm;
4607 }
4608
4609 err = adjust_reserved_segment(sbi);
4610 if (err)
4611 goto free_nm;
4612
4613 /* For write statistics */
4614 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4615
4616 /* Read accumulated write IO statistics if exists */
4617 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4618 if (__exist_node_summaries(sbi))
4619 sbi->kbytes_written =
4620 le64_to_cpu(seg_i->journal->info.kbytes_written);
4621
4622 f2fs_build_gc_manager(sbi);
4623
4624 err = f2fs_build_stats(sbi);
4625 if (err)
4626 goto free_nm;
4627
4628 /* get an inode for node space */
4629 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4630 if (IS_ERR(sbi->node_inode)) {
4631 f2fs_err(sbi, "Failed to read node inode");
4632 err = PTR_ERR(sbi->node_inode);
4633 goto free_stats;
4634 }
4635
4636 /* read root inode and dentry */
4637 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4638 if (IS_ERR(root)) {
4639 f2fs_err(sbi, "Failed to read root inode");
4640 err = PTR_ERR(root);
4641 goto free_node_inode;
4642 }
4643 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4644 !root->i_size || !root->i_nlink) {
4645 iput(root);
4646 err = -EINVAL;
4647 goto free_node_inode;
4648 }
4649
4650 sb->s_root = d_make_root(root); /* allocate root dentry */
4651 if (!sb->s_root) {
4652 err = -ENOMEM;
4653 goto free_node_inode;
4654 }
4655
4656 err = f2fs_init_compress_inode(sbi);
4657 if (err)
4658 goto free_root_inode;
4659
4660 err = f2fs_register_sysfs(sbi);
4661 if (err)
4662 goto free_compress_inode;
4663
4664 #ifdef CONFIG_QUOTA
4665 /* Enable quota usage during mount */
4666 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4667 err = f2fs_enable_quotas(sb);
4668 if (err)
4669 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4670 }
4671
4672 quota_enabled = f2fs_recover_quota_begin(sbi);
4673 #endif
4674 /* if there are any orphan inodes, free them */
4675 err = f2fs_recover_orphan_inodes(sbi);
4676 if (err)
4677 goto free_meta;
4678
4679 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4680 goto reset_checkpoint;
4681
4682 /* recover fsynced data */
4683 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4684 !test_opt(sbi, NORECOVERY)) {
4685 /*
4686 * mount should be failed, when device has readonly mode, and
4687 * previous checkpoint was not done by clean system shutdown.
4688 */
4689 if (f2fs_hw_is_readonly(sbi)) {
4690 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4691 err = f2fs_recover_fsync_data(sbi, true);
4692 if (err > 0) {
4693 err = -EROFS;
4694 f2fs_err(sbi, "Need to recover fsync data, but "
4695 "write access unavailable, please try "
4696 "mount w/ disable_roll_forward or norecovery");
4697 }
4698 if (err < 0)
4699 goto free_meta;
4700 }
4701 f2fs_info(sbi, "write access unavailable, skipping recovery");
4702 goto reset_checkpoint;
4703 }
4704
4705 if (need_fsck)
4706 set_sbi_flag(sbi, SBI_NEED_FSCK);
4707
4708 if (skip_recovery)
4709 goto reset_checkpoint;
4710
4711 err = f2fs_recover_fsync_data(sbi, false);
4712 if (err < 0) {
4713 if (err != -ENOMEM)
4714 skip_recovery = true;
4715 need_fsck = true;
4716 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4717 err);
4718 goto free_meta;
4719 }
4720 } else {
4721 err = f2fs_recover_fsync_data(sbi, true);
4722
4723 if (!f2fs_readonly(sb) && err > 0) {
4724 err = -EINVAL;
4725 f2fs_err(sbi, "Need to recover fsync data");
4726 goto free_meta;
4727 }
4728 }
4729
4730 #ifdef CONFIG_QUOTA
4731 f2fs_recover_quota_end(sbi, quota_enabled);
4732 #endif
4733
4734 /*
4735 * If the f2fs is not readonly and fsync data recovery succeeds,
4736 * check zoned block devices' write pointer consistency.
4737 */
4738 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4739 err = f2fs_check_write_pointer(sbi);
4740 if (err)
4741 goto free_meta;
4742 }
4743
4744 reset_checkpoint:
4745 f2fs_init_inmem_curseg(sbi);
4746
4747 /* f2fs_recover_fsync_data() cleared this already */
4748 clear_sbi_flag(sbi, SBI_POR_DOING);
4749
4750 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4751 err = f2fs_disable_checkpoint(sbi);
4752 if (err)
4753 goto sync_free_meta;
4754 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4755 f2fs_enable_checkpoint(sbi);
4756 }
4757
4758 /*
4759 * If filesystem is not mounted as read-only then
4760 * do start the gc_thread.
4761 */
4762 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4763 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4764 /* After POR, we can run background GC thread.*/
4765 err = f2fs_start_gc_thread(sbi);
4766 if (err)
4767 goto sync_free_meta;
4768 }
4769 kvfree(options);
4770
4771 /* recover broken superblock */
4772 if (recovery) {
4773 err = f2fs_commit_super(sbi, true);
4774 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4775 sbi->valid_super_block ? 1 : 2, err);
4776 }
4777
4778 f2fs_join_shrinker(sbi);
4779
4780 f2fs_tuning_parameters(sbi);
4781
4782 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4783 cur_cp_version(F2FS_CKPT(sbi)));
4784 f2fs_update_time(sbi, CP_TIME);
4785 f2fs_update_time(sbi, REQ_TIME);
4786 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4787 return 0;
4788
4789 sync_free_meta:
4790 /* safe to flush all the data */
4791 sync_filesystem(sbi->sb);
4792 retry_cnt = 0;
4793
4794 free_meta:
4795 #ifdef CONFIG_QUOTA
4796 f2fs_truncate_quota_inode_pages(sb);
4797 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4798 f2fs_quota_off_umount(sbi->sb);
4799 #endif
4800 /*
4801 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4802 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4803 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4804 * falls into an infinite loop in f2fs_sync_meta_pages().
4805 */
4806 truncate_inode_pages_final(META_MAPPING(sbi));
4807 /* evict some inodes being cached by GC */
4808 evict_inodes(sb);
4809 f2fs_unregister_sysfs(sbi);
4810 free_compress_inode:
4811 f2fs_destroy_compress_inode(sbi);
4812 free_root_inode:
4813 dput(sb->s_root);
4814 sb->s_root = NULL;
4815 free_node_inode:
4816 f2fs_release_ino_entry(sbi, true);
4817 truncate_inode_pages_final(NODE_MAPPING(sbi));
4818 iput(sbi->node_inode);
4819 sbi->node_inode = NULL;
4820 free_stats:
4821 f2fs_destroy_stats(sbi);
4822 free_nm:
4823 /* stop discard thread before destroying node manager */
4824 f2fs_stop_discard_thread(sbi);
4825 f2fs_destroy_node_manager(sbi);
4826 free_sm:
4827 f2fs_destroy_segment_manager(sbi);
4828 stop_ckpt_thread:
4829 f2fs_stop_ckpt_thread(sbi);
4830 /* flush s_error_work before sbi destroy */
4831 flush_work(&sbi->s_error_work);
4832 f2fs_destroy_post_read_wq(sbi);
4833 free_devices:
4834 destroy_device_list(sbi);
4835 kvfree(sbi->ckpt);
4836 free_meta_inode:
4837 make_bad_inode(sbi->meta_inode);
4838 iput(sbi->meta_inode);
4839 sbi->meta_inode = NULL;
4840 free_page_array_cache:
4841 f2fs_destroy_page_array_cache(sbi);
4842 free_xattr_cache:
4843 f2fs_destroy_xattr_caches(sbi);
4844 free_io_dummy:
4845 mempool_destroy(sbi->write_io_dummy);
4846 free_percpu:
4847 destroy_percpu_info(sbi);
4848 free_iostat:
4849 f2fs_destroy_iostat(sbi);
4850 free_bio_info:
4851 for (i = 0; i < NR_PAGE_TYPE; i++)
4852 kvfree(sbi->write_io[i]);
4853
4854 #if IS_ENABLED(CONFIG_UNICODE)
4855 utf8_unload(sb->s_encoding);
4856 sb->s_encoding = NULL;
4857 #endif
4858 free_options:
4859 #ifdef CONFIG_QUOTA
4860 for (i = 0; i < MAXQUOTAS; i++)
4861 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4862 #endif
4863 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4864 kvfree(options);
4865 free_sb_buf:
4866 kfree(raw_super);
4867 free_sbi:
4868 if (sbi->s_chksum_driver)
4869 crypto_free_shash(sbi->s_chksum_driver);
4870 kfree(sbi);
4871
4872 /* give only one another chance */
4873 if (retry_cnt > 0 && skip_recovery) {
4874 retry_cnt--;
4875 shrink_dcache_sb(sb);
4876 goto try_onemore;
4877 }
4878 return err;
4879 }
4880
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4881 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4882 const char *dev_name, void *data)
4883 {
4884 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4885 }
4886
kill_f2fs_super(struct super_block * sb)4887 static void kill_f2fs_super(struct super_block *sb)
4888 {
4889 if (sb->s_root) {
4890 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4891
4892 set_sbi_flag(sbi, SBI_IS_CLOSE);
4893 f2fs_stop_gc_thread(sbi);
4894 f2fs_stop_discard_thread(sbi);
4895
4896 #ifdef CONFIG_F2FS_FS_COMPRESSION
4897 /*
4898 * latter evict_inode() can bypass checking and invalidating
4899 * compress inode cache.
4900 */
4901 if (test_opt(sbi, COMPRESS_CACHE))
4902 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4903 #endif
4904
4905 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4906 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4907 struct cp_control cpc = {
4908 .reason = CP_UMOUNT,
4909 };
4910 stat_inc_cp_call_count(sbi, TOTAL_CALL);
4911 f2fs_write_checkpoint(sbi, &cpc);
4912 }
4913
4914 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4915 sb->s_flags &= ~SB_RDONLY;
4916 }
4917 kill_block_super(sb);
4918 }
4919
4920 static struct file_system_type f2fs_fs_type = {
4921 .owner = THIS_MODULE,
4922 .name = "f2fs",
4923 .mount = f2fs_mount,
4924 .kill_sb = kill_f2fs_super,
4925 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4926 };
4927 MODULE_ALIAS_FS("f2fs");
4928
init_inodecache(void)4929 static int __init init_inodecache(void)
4930 {
4931 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4932 sizeof(struct f2fs_inode_info), 0,
4933 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4934 return f2fs_inode_cachep ? 0 : -ENOMEM;
4935 }
4936
destroy_inodecache(void)4937 static void destroy_inodecache(void)
4938 {
4939 /*
4940 * Make sure all delayed rcu free inodes are flushed before we
4941 * destroy cache.
4942 */
4943 rcu_barrier();
4944 kmem_cache_destroy(f2fs_inode_cachep);
4945 }
4946
init_f2fs_fs(void)4947 static int __init init_f2fs_fs(void)
4948 {
4949 int err;
4950
4951 if (PAGE_SIZE != F2FS_BLKSIZE) {
4952 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4953 PAGE_SIZE, F2FS_BLKSIZE);
4954 return -EINVAL;
4955 }
4956
4957 err = init_inodecache();
4958 if (err)
4959 goto fail;
4960 err = f2fs_create_node_manager_caches();
4961 if (err)
4962 goto free_inodecache;
4963 err = f2fs_create_segment_manager_caches();
4964 if (err)
4965 goto free_node_manager_caches;
4966 err = f2fs_create_checkpoint_caches();
4967 if (err)
4968 goto free_segment_manager_caches;
4969 err = f2fs_create_recovery_cache();
4970 if (err)
4971 goto free_checkpoint_caches;
4972 err = f2fs_create_extent_cache();
4973 if (err)
4974 goto free_recovery_cache;
4975 err = f2fs_create_garbage_collection_cache();
4976 if (err)
4977 goto free_extent_cache;
4978 err = f2fs_init_sysfs();
4979 if (err)
4980 goto free_garbage_collection_cache;
4981 err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4982 if (err)
4983 goto free_sysfs;
4984 err = register_filesystem(&f2fs_fs_type);
4985 if (err)
4986 goto free_shrinker;
4987 f2fs_create_root_stats();
4988 err = f2fs_init_post_read_processing();
4989 if (err)
4990 goto free_root_stats;
4991 err = f2fs_init_iostat_processing();
4992 if (err)
4993 goto free_post_read;
4994 err = f2fs_init_bio_entry_cache();
4995 if (err)
4996 goto free_iostat;
4997 err = f2fs_init_bioset();
4998 if (err)
4999 goto free_bio_entry_cache;
5000 err = f2fs_init_compress_mempool();
5001 if (err)
5002 goto free_bioset;
5003 err = f2fs_init_compress_cache();
5004 if (err)
5005 goto free_compress_mempool;
5006 err = f2fs_create_casefold_cache();
5007 if (err)
5008 goto free_compress_cache;
5009 return 0;
5010 free_compress_cache:
5011 f2fs_destroy_compress_cache();
5012 free_compress_mempool:
5013 f2fs_destroy_compress_mempool();
5014 free_bioset:
5015 f2fs_destroy_bioset();
5016 free_bio_entry_cache:
5017 f2fs_destroy_bio_entry_cache();
5018 free_iostat:
5019 f2fs_destroy_iostat_processing();
5020 free_post_read:
5021 f2fs_destroy_post_read_processing();
5022 free_root_stats:
5023 f2fs_destroy_root_stats();
5024 unregister_filesystem(&f2fs_fs_type);
5025 free_shrinker:
5026 unregister_shrinker(&f2fs_shrinker_info);
5027 free_sysfs:
5028 f2fs_exit_sysfs();
5029 free_garbage_collection_cache:
5030 f2fs_destroy_garbage_collection_cache();
5031 free_extent_cache:
5032 f2fs_destroy_extent_cache();
5033 free_recovery_cache:
5034 f2fs_destroy_recovery_cache();
5035 free_checkpoint_caches:
5036 f2fs_destroy_checkpoint_caches();
5037 free_segment_manager_caches:
5038 f2fs_destroy_segment_manager_caches();
5039 free_node_manager_caches:
5040 f2fs_destroy_node_manager_caches();
5041 free_inodecache:
5042 destroy_inodecache();
5043 fail:
5044 return err;
5045 }
5046
exit_f2fs_fs(void)5047 static void __exit exit_f2fs_fs(void)
5048 {
5049 f2fs_destroy_casefold_cache();
5050 f2fs_destroy_compress_cache();
5051 f2fs_destroy_compress_mempool();
5052 f2fs_destroy_bioset();
5053 f2fs_destroy_bio_entry_cache();
5054 f2fs_destroy_iostat_processing();
5055 f2fs_destroy_post_read_processing();
5056 f2fs_destroy_root_stats();
5057 unregister_filesystem(&f2fs_fs_type);
5058 unregister_shrinker(&f2fs_shrinker_info);
5059 f2fs_exit_sysfs();
5060 f2fs_destroy_garbage_collection_cache();
5061 f2fs_destroy_extent_cache();
5062 f2fs_destroy_recovery_cache();
5063 f2fs_destroy_checkpoint_caches();
5064 f2fs_destroy_segment_manager_caches();
5065 f2fs_destroy_node_manager_caches();
5066 destroy_inodecache();
5067 }
5068
5069 module_init(init_f2fs_fs)
5070 module_exit(exit_f2fs_fs)
5071
5072 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5073 MODULE_DESCRIPTION("Flash Friendly File System");
5074 MODULE_LICENSE("GPL");
5075 MODULE_SOFTDEP("pre: crc32");
5076
5077