Loading...
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
42static struct kmem_cache *f2fs_inode_cachep;
43
44#ifdef CONFIG_F2FS_FAULT_INJECTION
45
46const 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
67void 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 */
86static struct shrinker f2fs_shrinker_info = {
87 .scan_objects = f2fs_shrink_scan,
88 .count_objects = f2fs_shrink_count,
89 .seeks = DEFAULT_SEEKS,
90};
91
92enum {
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_err,
168};
169
170static match_table_t f2fs_tokens = {
171 {Opt_gc_background, "background_gc=%s"},
172 {Opt_disable_roll_forward, "disable_roll_forward"},
173 {Opt_norecovery, "norecovery"},
174 {Opt_discard, "discard"},
175 {Opt_nodiscard, "nodiscard"},
176 {Opt_noheap, "no_heap"},
177 {Opt_heap, "heap"},
178 {Opt_user_xattr, "user_xattr"},
179 {Opt_nouser_xattr, "nouser_xattr"},
180 {Opt_acl, "acl"},
181 {Opt_noacl, "noacl"},
182 {Opt_active_logs, "active_logs=%u"},
183 {Opt_disable_ext_identify, "disable_ext_identify"},
184 {Opt_inline_xattr, "inline_xattr"},
185 {Opt_noinline_xattr, "noinline_xattr"},
186 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
187 {Opt_inline_data, "inline_data"},
188 {Opt_inline_dentry, "inline_dentry"},
189 {Opt_noinline_dentry, "noinline_dentry"},
190 {Opt_flush_merge, "flush_merge"},
191 {Opt_noflush_merge, "noflush_merge"},
192 {Opt_barrier, "barrier"},
193 {Opt_nobarrier, "nobarrier"},
194 {Opt_fastboot, "fastboot"},
195 {Opt_extent_cache, "extent_cache"},
196 {Opt_noextent_cache, "noextent_cache"},
197 {Opt_noinline_data, "noinline_data"},
198 {Opt_data_flush, "data_flush"},
199 {Opt_reserve_root, "reserve_root=%u"},
200 {Opt_resgid, "resgid=%u"},
201 {Opt_resuid, "resuid=%u"},
202 {Opt_mode, "mode=%s"},
203 {Opt_io_size_bits, "io_bits=%u"},
204 {Opt_fault_injection, "fault_injection=%u"},
205 {Opt_fault_type, "fault_type=%u"},
206 {Opt_lazytime, "lazytime"},
207 {Opt_nolazytime, "nolazytime"},
208 {Opt_quota, "quota"},
209 {Opt_noquota, "noquota"},
210 {Opt_usrquota, "usrquota"},
211 {Opt_grpquota, "grpquota"},
212 {Opt_prjquota, "prjquota"},
213 {Opt_usrjquota, "usrjquota=%s"},
214 {Opt_grpjquota, "grpjquota=%s"},
215 {Opt_prjjquota, "prjjquota=%s"},
216 {Opt_offusrjquota, "usrjquota="},
217 {Opt_offgrpjquota, "grpjquota="},
218 {Opt_offprjjquota, "prjjquota="},
219 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
220 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
221 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
222 {Opt_alloc, "alloc_mode=%s"},
223 {Opt_fsync, "fsync_mode=%s"},
224 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
225 {Opt_test_dummy_encryption, "test_dummy_encryption"},
226 {Opt_inlinecrypt, "inlinecrypt"},
227 {Opt_checkpoint_disable, "checkpoint=disable"},
228 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
229 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
230 {Opt_checkpoint_enable, "checkpoint=enable"},
231 {Opt_checkpoint_merge, "checkpoint_merge"},
232 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
233 {Opt_compress_algorithm, "compress_algorithm=%s"},
234 {Opt_compress_log_size, "compress_log_size=%u"},
235 {Opt_compress_extension, "compress_extension=%s"},
236 {Opt_nocompress_extension, "nocompress_extension=%s"},
237 {Opt_compress_chksum, "compress_chksum"},
238 {Opt_compress_mode, "compress_mode=%s"},
239 {Opt_compress_cache, "compress_cache"},
240 {Opt_atgc, "atgc"},
241 {Opt_gc_merge, "gc_merge"},
242 {Opt_nogc_merge, "nogc_merge"},
243 {Opt_discard_unit, "discard_unit=%s"},
244 {Opt_memory_mode, "memory=%s"},
245 {Opt_age_extent_cache, "age_extent_cache"},
246 {Opt_err, NULL},
247};
248
249void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
250{
251 struct va_format vaf;
252 va_list args;
253 int level;
254
255 va_start(args, fmt);
256
257 level = printk_get_level(fmt);
258 vaf.fmt = printk_skip_level(fmt);
259 vaf.va = &args;
260 printk("%c%cF2FS-fs (%s): %pV\n",
261 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
262
263 va_end(args);
264}
265
266#if IS_ENABLED(CONFIG_UNICODE)
267static const struct f2fs_sb_encodings {
268 __u16 magic;
269 char *name;
270 unsigned int version;
271} f2fs_sb_encoding_map[] = {
272 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
273};
274
275static const struct f2fs_sb_encodings *
276f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
277{
278 __u16 magic = le16_to_cpu(sb->s_encoding);
279 int i;
280
281 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
282 if (magic == f2fs_sb_encoding_map[i].magic)
283 return &f2fs_sb_encoding_map[i];
284
285 return NULL;
286}
287
288struct kmem_cache *f2fs_cf_name_slab;
289static int __init f2fs_create_casefold_cache(void)
290{
291 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
292 F2FS_NAME_LEN);
293 return f2fs_cf_name_slab ? 0 : -ENOMEM;
294}
295
296static void f2fs_destroy_casefold_cache(void)
297{
298 kmem_cache_destroy(f2fs_cf_name_slab);
299}
300#else
301static int __init f2fs_create_casefold_cache(void) { return 0; }
302static void f2fs_destroy_casefold_cache(void) { }
303#endif
304
305static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
306{
307 block_t limit = min((sbi->user_block_count >> 3),
308 sbi->user_block_count - sbi->reserved_blocks);
309
310 /* limit is 12.5% */
311 if (test_opt(sbi, RESERVE_ROOT) &&
312 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
313 F2FS_OPTION(sbi).root_reserved_blocks = limit;
314 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
315 F2FS_OPTION(sbi).root_reserved_blocks);
316 }
317 if (!test_opt(sbi, RESERVE_ROOT) &&
318 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
319 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
320 !gid_eq(F2FS_OPTION(sbi).s_resgid,
321 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
322 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
323 from_kuid_munged(&init_user_ns,
324 F2FS_OPTION(sbi).s_resuid),
325 from_kgid_munged(&init_user_ns,
326 F2FS_OPTION(sbi).s_resgid));
327}
328
329static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
330{
331 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
332 unsigned int avg_vblocks;
333 unsigned int wanted_reserved_segments;
334 block_t avail_user_block_count;
335
336 if (!F2FS_IO_ALIGNED(sbi))
337 return 0;
338
339 /* average valid block count in section in worst case */
340 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
341
342 /*
343 * we need enough free space when migrating one section in worst case
344 */
345 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
346 reserved_segments(sbi);
347 wanted_reserved_segments -= reserved_segments(sbi);
348
349 avail_user_block_count = sbi->user_block_count -
350 sbi->current_reserved_blocks -
351 F2FS_OPTION(sbi).root_reserved_blocks;
352
353 if (wanted_reserved_segments * sbi->blocks_per_seg >
354 avail_user_block_count) {
355 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
356 wanted_reserved_segments,
357 avail_user_block_count >> sbi->log_blocks_per_seg);
358 return -ENOSPC;
359 }
360
361 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
362
363 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
364 wanted_reserved_segments);
365
366 return 0;
367}
368
369static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
370{
371 if (!F2FS_OPTION(sbi).unusable_cap_perc)
372 return;
373
374 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
375 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
376 else
377 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
378 F2FS_OPTION(sbi).unusable_cap_perc;
379
380 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
381 F2FS_OPTION(sbi).unusable_cap,
382 F2FS_OPTION(sbi).unusable_cap_perc);
383}
384
385static void init_once(void *foo)
386{
387 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
388
389 inode_init_once(&fi->vfs_inode);
390}
391
392#ifdef CONFIG_QUOTA
393static const char * const quotatypes[] = INITQFNAMES;
394#define QTYPE2NAME(t) (quotatypes[t])
395static int f2fs_set_qf_name(struct super_block *sb, int qtype,
396 substring_t *args)
397{
398 struct f2fs_sb_info *sbi = F2FS_SB(sb);
399 char *qname;
400 int ret = -EINVAL;
401
402 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
403 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
404 return -EINVAL;
405 }
406 if (f2fs_sb_has_quota_ino(sbi)) {
407 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
408 return 0;
409 }
410
411 qname = match_strdup(args);
412 if (!qname) {
413 f2fs_err(sbi, "Not enough memory for storing quotafile name");
414 return -ENOMEM;
415 }
416 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
417 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
418 ret = 0;
419 else
420 f2fs_err(sbi, "%s quota file already specified",
421 QTYPE2NAME(qtype));
422 goto errout;
423 }
424 if (strchr(qname, '/')) {
425 f2fs_err(sbi, "quotafile must be on filesystem root");
426 goto errout;
427 }
428 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
429 set_opt(sbi, QUOTA);
430 return 0;
431errout:
432 kfree(qname);
433 return ret;
434}
435
436static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
437{
438 struct f2fs_sb_info *sbi = F2FS_SB(sb);
439
440 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
441 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
442 return -EINVAL;
443 }
444 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
445 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
446 return 0;
447}
448
449static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
450{
451 /*
452 * We do the test below only for project quotas. 'usrquota' and
453 * 'grpquota' mount options are allowed even without quota feature
454 * to support legacy quotas in quota files.
455 */
456 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
457 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
458 return -1;
459 }
460 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
461 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
462 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
463 if (test_opt(sbi, USRQUOTA) &&
464 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
465 clear_opt(sbi, USRQUOTA);
466
467 if (test_opt(sbi, GRPQUOTA) &&
468 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
469 clear_opt(sbi, GRPQUOTA);
470
471 if (test_opt(sbi, PRJQUOTA) &&
472 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
473 clear_opt(sbi, PRJQUOTA);
474
475 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
476 test_opt(sbi, PRJQUOTA)) {
477 f2fs_err(sbi, "old and new quota format mixing");
478 return -1;
479 }
480
481 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
482 f2fs_err(sbi, "journaled quota format not specified");
483 return -1;
484 }
485 }
486
487 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
488 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
489 F2FS_OPTION(sbi).s_jquota_fmt = 0;
490 }
491 return 0;
492}
493#endif
494
495static int f2fs_set_test_dummy_encryption(struct super_block *sb,
496 const char *opt,
497 const substring_t *arg,
498 bool is_remount)
499{
500 struct f2fs_sb_info *sbi = F2FS_SB(sb);
501 struct fs_parameter param = {
502 .type = fs_value_is_string,
503 .string = arg->from ? arg->from : "",
504 };
505 struct fscrypt_dummy_policy *policy =
506 &F2FS_OPTION(sbi).dummy_enc_policy;
507 int err;
508
509 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
510 f2fs_warn(sbi, "test_dummy_encryption option not supported");
511 return -EINVAL;
512 }
513
514 if (!f2fs_sb_has_encrypt(sbi)) {
515 f2fs_err(sbi, "Encrypt feature is off");
516 return -EINVAL;
517 }
518
519 /*
520 * This mount option is just for testing, and it's not worthwhile to
521 * implement the extra complexity (e.g. RCU protection) that would be
522 * needed to allow it to be set or changed during remount. We do allow
523 * it to be specified during remount, but only if there is no change.
524 */
525 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
526 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
527 return -EINVAL;
528 }
529
530 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
531 if (err) {
532 if (err == -EEXIST)
533 f2fs_warn(sbi,
534 "Can't change test_dummy_encryption on remount");
535 else if (err == -EINVAL)
536 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
537 opt);
538 else
539 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
540 opt, err);
541 return -EINVAL;
542 }
543 err = fscrypt_add_test_dummy_key(sb, policy);
544 if (err) {
545 f2fs_warn(sbi, "Error adding test dummy encryption key [%d]",
546 err);
547 return err;
548 }
549 f2fs_warn(sbi, "Test dummy encryption mode enabled");
550 return 0;
551}
552
553#ifdef CONFIG_F2FS_FS_COMPRESSION
554/*
555 * 1. The same extension name cannot not appear in both compress and non-compress extension
556 * at the same time.
557 * 2. If the compress extension specifies all files, the types specified by the non-compress
558 * extension will be treated as special cases and will not be compressed.
559 * 3. Don't allow the non-compress extension specifies all files.
560 */
561static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
562{
563 unsigned char (*ext)[F2FS_EXTENSION_LEN];
564 unsigned char (*noext)[F2FS_EXTENSION_LEN];
565 int ext_cnt, noext_cnt, index = 0, no_index = 0;
566
567 ext = F2FS_OPTION(sbi).extensions;
568 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
569 noext = F2FS_OPTION(sbi).noextensions;
570 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
571
572 if (!noext_cnt)
573 return 0;
574
575 for (no_index = 0; no_index < noext_cnt; no_index++) {
576 if (!strcasecmp("*", noext[no_index])) {
577 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
578 return -EINVAL;
579 }
580 for (index = 0; index < ext_cnt; index++) {
581 if (!strcasecmp(ext[index], noext[no_index])) {
582 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
583 ext[index]);
584 return -EINVAL;
585 }
586 }
587 }
588 return 0;
589}
590
591#ifdef CONFIG_F2FS_FS_LZ4
592static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
593{
594#ifdef CONFIG_F2FS_FS_LZ4HC
595 unsigned int level;
596#endif
597
598 if (strlen(str) == 3) {
599 F2FS_OPTION(sbi).compress_level = 0;
600 return 0;
601 }
602
603#ifdef CONFIG_F2FS_FS_LZ4HC
604 str += 3;
605
606 if (str[0] != ':') {
607 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
608 return -EINVAL;
609 }
610 if (kstrtouint(str + 1, 10, &level))
611 return -EINVAL;
612
613 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
614 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
615 return -EINVAL;
616 }
617
618 F2FS_OPTION(sbi).compress_level = level;
619 return 0;
620#else
621 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
622 return -EINVAL;
623#endif
624}
625#endif
626
627#ifdef CONFIG_F2FS_FS_ZSTD
628static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
629{
630 unsigned int level;
631 int len = 4;
632
633 if (strlen(str) == len) {
634 F2FS_OPTION(sbi).compress_level = 0;
635 return 0;
636 }
637
638 str += len;
639
640 if (str[0] != ':') {
641 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
642 return -EINVAL;
643 }
644 if (kstrtouint(str + 1, 10, &level))
645 return -EINVAL;
646
647 if (!level || level > zstd_max_clevel()) {
648 f2fs_info(sbi, "invalid zstd compress level: %d", level);
649 return -EINVAL;
650 }
651
652 F2FS_OPTION(sbi).compress_level = level;
653 return 0;
654}
655#endif
656#endif
657
658static int parse_options(struct super_block *sb, char *options, bool is_remount)
659{
660 struct f2fs_sb_info *sbi = F2FS_SB(sb);
661 substring_t args[MAX_OPT_ARGS];
662#ifdef CONFIG_F2FS_FS_COMPRESSION
663 unsigned char (*ext)[F2FS_EXTENSION_LEN];
664 unsigned char (*noext)[F2FS_EXTENSION_LEN];
665 int ext_cnt, noext_cnt;
666#endif
667 char *p, *name;
668 int arg = 0;
669 kuid_t uid;
670 kgid_t gid;
671 int ret;
672
673 if (!options)
674 goto default_check;
675
676 while ((p = strsep(&options, ",")) != NULL) {
677 int token;
678
679 if (!*p)
680 continue;
681 /*
682 * Initialize args struct so we know whether arg was
683 * found; some options take optional arguments.
684 */
685 args[0].to = args[0].from = NULL;
686 token = match_token(p, f2fs_tokens, args);
687
688 switch (token) {
689 case Opt_gc_background:
690 name = match_strdup(&args[0]);
691
692 if (!name)
693 return -ENOMEM;
694 if (!strcmp(name, "on")) {
695 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
696 } else if (!strcmp(name, "off")) {
697 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
698 } else if (!strcmp(name, "sync")) {
699 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
700 } else {
701 kfree(name);
702 return -EINVAL;
703 }
704 kfree(name);
705 break;
706 case Opt_disable_roll_forward:
707 set_opt(sbi, DISABLE_ROLL_FORWARD);
708 break;
709 case Opt_norecovery:
710 /* this option mounts f2fs with ro */
711 set_opt(sbi, NORECOVERY);
712 if (!f2fs_readonly(sb))
713 return -EINVAL;
714 break;
715 case Opt_discard:
716 if (!f2fs_hw_support_discard(sbi)) {
717 f2fs_warn(sbi, "device does not support discard");
718 break;
719 }
720 set_opt(sbi, DISCARD);
721 break;
722 case Opt_nodiscard:
723 if (f2fs_hw_should_discard(sbi)) {
724 f2fs_warn(sbi, "discard is required for zoned block devices");
725 return -EINVAL;
726 }
727 clear_opt(sbi, DISCARD);
728 break;
729 case Opt_noheap:
730 set_opt(sbi, NOHEAP);
731 break;
732 case Opt_heap:
733 clear_opt(sbi, NOHEAP);
734 break;
735#ifdef CONFIG_F2FS_FS_XATTR
736 case Opt_user_xattr:
737 set_opt(sbi, XATTR_USER);
738 break;
739 case Opt_nouser_xattr:
740 clear_opt(sbi, XATTR_USER);
741 break;
742 case Opt_inline_xattr:
743 set_opt(sbi, INLINE_XATTR);
744 break;
745 case Opt_noinline_xattr:
746 clear_opt(sbi, INLINE_XATTR);
747 break;
748 case Opt_inline_xattr_size:
749 if (args->from && match_int(args, &arg))
750 return -EINVAL;
751 set_opt(sbi, INLINE_XATTR_SIZE);
752 F2FS_OPTION(sbi).inline_xattr_size = arg;
753 break;
754#else
755 case Opt_user_xattr:
756 f2fs_info(sbi, "user_xattr options not supported");
757 break;
758 case Opt_nouser_xattr:
759 f2fs_info(sbi, "nouser_xattr options not supported");
760 break;
761 case Opt_inline_xattr:
762 f2fs_info(sbi, "inline_xattr options not supported");
763 break;
764 case Opt_noinline_xattr:
765 f2fs_info(sbi, "noinline_xattr options not supported");
766 break;
767#endif
768#ifdef CONFIG_F2FS_FS_POSIX_ACL
769 case Opt_acl:
770 set_opt(sbi, POSIX_ACL);
771 break;
772 case Opt_noacl:
773 clear_opt(sbi, POSIX_ACL);
774 break;
775#else
776 case Opt_acl:
777 f2fs_info(sbi, "acl options not supported");
778 break;
779 case Opt_noacl:
780 f2fs_info(sbi, "noacl options not supported");
781 break;
782#endif
783 case Opt_active_logs:
784 if (args->from && match_int(args, &arg))
785 return -EINVAL;
786 if (arg != 2 && arg != 4 &&
787 arg != NR_CURSEG_PERSIST_TYPE)
788 return -EINVAL;
789 F2FS_OPTION(sbi).active_logs = arg;
790 break;
791 case Opt_disable_ext_identify:
792 set_opt(sbi, DISABLE_EXT_IDENTIFY);
793 break;
794 case Opt_inline_data:
795 set_opt(sbi, INLINE_DATA);
796 break;
797 case Opt_inline_dentry:
798 set_opt(sbi, INLINE_DENTRY);
799 break;
800 case Opt_noinline_dentry:
801 clear_opt(sbi, INLINE_DENTRY);
802 break;
803 case Opt_flush_merge:
804 set_opt(sbi, FLUSH_MERGE);
805 break;
806 case Opt_noflush_merge:
807 clear_opt(sbi, FLUSH_MERGE);
808 break;
809 case Opt_nobarrier:
810 set_opt(sbi, NOBARRIER);
811 break;
812 case Opt_barrier:
813 clear_opt(sbi, NOBARRIER);
814 break;
815 case Opt_fastboot:
816 set_opt(sbi, FASTBOOT);
817 break;
818 case Opt_extent_cache:
819 set_opt(sbi, READ_EXTENT_CACHE);
820 break;
821 case Opt_noextent_cache:
822 clear_opt(sbi, READ_EXTENT_CACHE);
823 break;
824 case Opt_noinline_data:
825 clear_opt(sbi, INLINE_DATA);
826 break;
827 case Opt_data_flush:
828 set_opt(sbi, DATA_FLUSH);
829 break;
830 case Opt_reserve_root:
831 if (args->from && match_int(args, &arg))
832 return -EINVAL;
833 if (test_opt(sbi, RESERVE_ROOT)) {
834 f2fs_info(sbi, "Preserve previous reserve_root=%u",
835 F2FS_OPTION(sbi).root_reserved_blocks);
836 } else {
837 F2FS_OPTION(sbi).root_reserved_blocks = arg;
838 set_opt(sbi, RESERVE_ROOT);
839 }
840 break;
841 case Opt_resuid:
842 if (args->from && match_int(args, &arg))
843 return -EINVAL;
844 uid = make_kuid(current_user_ns(), arg);
845 if (!uid_valid(uid)) {
846 f2fs_err(sbi, "Invalid uid value %d", arg);
847 return -EINVAL;
848 }
849 F2FS_OPTION(sbi).s_resuid = uid;
850 break;
851 case Opt_resgid:
852 if (args->from && match_int(args, &arg))
853 return -EINVAL;
854 gid = make_kgid(current_user_ns(), arg);
855 if (!gid_valid(gid)) {
856 f2fs_err(sbi, "Invalid gid value %d", arg);
857 return -EINVAL;
858 }
859 F2FS_OPTION(sbi).s_resgid = gid;
860 break;
861 case Opt_mode:
862 name = match_strdup(&args[0]);
863
864 if (!name)
865 return -ENOMEM;
866 if (!strcmp(name, "adaptive")) {
867 if (f2fs_sb_has_blkzoned(sbi)) {
868 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
869 kfree(name);
870 return -EINVAL;
871 }
872 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
873 } else if (!strcmp(name, "lfs")) {
874 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
875 } else if (!strcmp(name, "fragment:segment")) {
876 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
877 } else if (!strcmp(name, "fragment:block")) {
878 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
879 } else {
880 kfree(name);
881 return -EINVAL;
882 }
883 kfree(name);
884 break;
885 case Opt_io_size_bits:
886 if (args->from && match_int(args, &arg))
887 return -EINVAL;
888 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
889 f2fs_warn(sbi, "Not support %d, larger than %d",
890 1 << arg, BIO_MAX_VECS);
891 return -EINVAL;
892 }
893 F2FS_OPTION(sbi).write_io_size_bits = arg;
894 break;
895#ifdef CONFIG_F2FS_FAULT_INJECTION
896 case Opt_fault_injection:
897 if (args->from && match_int(args, &arg))
898 return -EINVAL;
899 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
900 set_opt(sbi, FAULT_INJECTION);
901 break;
902
903 case Opt_fault_type:
904 if (args->from && match_int(args, &arg))
905 return -EINVAL;
906 f2fs_build_fault_attr(sbi, 0, arg);
907 set_opt(sbi, FAULT_INJECTION);
908 break;
909#else
910 case Opt_fault_injection:
911 f2fs_info(sbi, "fault_injection options not supported");
912 break;
913
914 case Opt_fault_type:
915 f2fs_info(sbi, "fault_type options not supported");
916 break;
917#endif
918 case Opt_lazytime:
919 sb->s_flags |= SB_LAZYTIME;
920 break;
921 case Opt_nolazytime:
922 sb->s_flags &= ~SB_LAZYTIME;
923 break;
924#ifdef CONFIG_QUOTA
925 case Opt_quota:
926 case Opt_usrquota:
927 set_opt(sbi, USRQUOTA);
928 break;
929 case Opt_grpquota:
930 set_opt(sbi, GRPQUOTA);
931 break;
932 case Opt_prjquota:
933 set_opt(sbi, PRJQUOTA);
934 break;
935 case Opt_usrjquota:
936 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
937 if (ret)
938 return ret;
939 break;
940 case Opt_grpjquota:
941 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
942 if (ret)
943 return ret;
944 break;
945 case Opt_prjjquota:
946 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
947 if (ret)
948 return ret;
949 break;
950 case Opt_offusrjquota:
951 ret = f2fs_clear_qf_name(sb, USRQUOTA);
952 if (ret)
953 return ret;
954 break;
955 case Opt_offgrpjquota:
956 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
957 if (ret)
958 return ret;
959 break;
960 case Opt_offprjjquota:
961 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
962 if (ret)
963 return ret;
964 break;
965 case Opt_jqfmt_vfsold:
966 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
967 break;
968 case Opt_jqfmt_vfsv0:
969 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
970 break;
971 case Opt_jqfmt_vfsv1:
972 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
973 break;
974 case Opt_noquota:
975 clear_opt(sbi, QUOTA);
976 clear_opt(sbi, USRQUOTA);
977 clear_opt(sbi, GRPQUOTA);
978 clear_opt(sbi, PRJQUOTA);
979 break;
980#else
981 case Opt_quota:
982 case Opt_usrquota:
983 case Opt_grpquota:
984 case Opt_prjquota:
985 case Opt_usrjquota:
986 case Opt_grpjquota:
987 case Opt_prjjquota:
988 case Opt_offusrjquota:
989 case Opt_offgrpjquota:
990 case Opt_offprjjquota:
991 case Opt_jqfmt_vfsold:
992 case Opt_jqfmt_vfsv0:
993 case Opt_jqfmt_vfsv1:
994 case Opt_noquota:
995 f2fs_info(sbi, "quota operations not supported");
996 break;
997#endif
998 case Opt_alloc:
999 name = match_strdup(&args[0]);
1000 if (!name)
1001 return -ENOMEM;
1002
1003 if (!strcmp(name, "default")) {
1004 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1005 } else if (!strcmp(name, "reuse")) {
1006 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1007 } else {
1008 kfree(name);
1009 return -EINVAL;
1010 }
1011 kfree(name);
1012 break;
1013 case Opt_fsync:
1014 name = match_strdup(&args[0]);
1015 if (!name)
1016 return -ENOMEM;
1017 if (!strcmp(name, "posix")) {
1018 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1019 } else if (!strcmp(name, "strict")) {
1020 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1021 } else if (!strcmp(name, "nobarrier")) {
1022 F2FS_OPTION(sbi).fsync_mode =
1023 FSYNC_MODE_NOBARRIER;
1024 } else {
1025 kfree(name);
1026 return -EINVAL;
1027 }
1028 kfree(name);
1029 break;
1030 case Opt_test_dummy_encryption:
1031 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1032 is_remount);
1033 if (ret)
1034 return ret;
1035 break;
1036 case Opt_inlinecrypt:
1037#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1038 sb->s_flags |= SB_INLINECRYPT;
1039#else
1040 f2fs_info(sbi, "inline encryption not supported");
1041#endif
1042 break;
1043 case Opt_checkpoint_disable_cap_perc:
1044 if (args->from && match_int(args, &arg))
1045 return -EINVAL;
1046 if (arg < 0 || arg > 100)
1047 return -EINVAL;
1048 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1049 set_opt(sbi, DISABLE_CHECKPOINT);
1050 break;
1051 case Opt_checkpoint_disable_cap:
1052 if (args->from && match_int(args, &arg))
1053 return -EINVAL;
1054 F2FS_OPTION(sbi).unusable_cap = arg;
1055 set_opt(sbi, DISABLE_CHECKPOINT);
1056 break;
1057 case Opt_checkpoint_disable:
1058 set_opt(sbi, DISABLE_CHECKPOINT);
1059 break;
1060 case Opt_checkpoint_enable:
1061 clear_opt(sbi, DISABLE_CHECKPOINT);
1062 break;
1063 case Opt_checkpoint_merge:
1064 set_opt(sbi, MERGE_CHECKPOINT);
1065 break;
1066 case Opt_nocheckpoint_merge:
1067 clear_opt(sbi, MERGE_CHECKPOINT);
1068 break;
1069#ifdef CONFIG_F2FS_FS_COMPRESSION
1070 case Opt_compress_algorithm:
1071 if (!f2fs_sb_has_compression(sbi)) {
1072 f2fs_info(sbi, "Image doesn't support compression");
1073 break;
1074 }
1075 name = match_strdup(&args[0]);
1076 if (!name)
1077 return -ENOMEM;
1078 if (!strcmp(name, "lzo")) {
1079#ifdef CONFIG_F2FS_FS_LZO
1080 F2FS_OPTION(sbi).compress_level = 0;
1081 F2FS_OPTION(sbi).compress_algorithm =
1082 COMPRESS_LZO;
1083#else
1084 f2fs_info(sbi, "kernel doesn't support lzo compression");
1085#endif
1086 } else if (!strncmp(name, "lz4", 3)) {
1087#ifdef CONFIG_F2FS_FS_LZ4
1088 ret = f2fs_set_lz4hc_level(sbi, name);
1089 if (ret) {
1090 kfree(name);
1091 return -EINVAL;
1092 }
1093 F2FS_OPTION(sbi).compress_algorithm =
1094 COMPRESS_LZ4;
1095#else
1096 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1097#endif
1098 } else if (!strncmp(name, "zstd", 4)) {
1099#ifdef CONFIG_F2FS_FS_ZSTD
1100 ret = f2fs_set_zstd_level(sbi, name);
1101 if (ret) {
1102 kfree(name);
1103 return -EINVAL;
1104 }
1105 F2FS_OPTION(sbi).compress_algorithm =
1106 COMPRESS_ZSTD;
1107#else
1108 f2fs_info(sbi, "kernel doesn't support zstd compression");
1109#endif
1110 } else if (!strcmp(name, "lzo-rle")) {
1111#ifdef CONFIG_F2FS_FS_LZORLE
1112 F2FS_OPTION(sbi).compress_level = 0;
1113 F2FS_OPTION(sbi).compress_algorithm =
1114 COMPRESS_LZORLE;
1115#else
1116 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1117#endif
1118 } else {
1119 kfree(name);
1120 return -EINVAL;
1121 }
1122 kfree(name);
1123 break;
1124 case Opt_compress_log_size:
1125 if (!f2fs_sb_has_compression(sbi)) {
1126 f2fs_info(sbi, "Image doesn't support compression");
1127 break;
1128 }
1129 if (args->from && match_int(args, &arg))
1130 return -EINVAL;
1131 if (arg < MIN_COMPRESS_LOG_SIZE ||
1132 arg > MAX_COMPRESS_LOG_SIZE) {
1133 f2fs_err(sbi,
1134 "Compress cluster log size is out of range");
1135 return -EINVAL;
1136 }
1137 F2FS_OPTION(sbi).compress_log_size = arg;
1138 break;
1139 case Opt_compress_extension:
1140 if (!f2fs_sb_has_compression(sbi)) {
1141 f2fs_info(sbi, "Image doesn't support compression");
1142 break;
1143 }
1144 name = match_strdup(&args[0]);
1145 if (!name)
1146 return -ENOMEM;
1147
1148 ext = F2FS_OPTION(sbi).extensions;
1149 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1150
1151 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1152 ext_cnt >= COMPRESS_EXT_NUM) {
1153 f2fs_err(sbi,
1154 "invalid extension length/number");
1155 kfree(name);
1156 return -EINVAL;
1157 }
1158
1159 strcpy(ext[ext_cnt], name);
1160 F2FS_OPTION(sbi).compress_ext_cnt++;
1161 kfree(name);
1162 break;
1163 case Opt_nocompress_extension:
1164 if (!f2fs_sb_has_compression(sbi)) {
1165 f2fs_info(sbi, "Image doesn't support compression");
1166 break;
1167 }
1168 name = match_strdup(&args[0]);
1169 if (!name)
1170 return -ENOMEM;
1171
1172 noext = F2FS_OPTION(sbi).noextensions;
1173 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1174
1175 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1176 noext_cnt >= COMPRESS_EXT_NUM) {
1177 f2fs_err(sbi,
1178 "invalid extension length/number");
1179 kfree(name);
1180 return -EINVAL;
1181 }
1182
1183 strcpy(noext[noext_cnt], name);
1184 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1185 kfree(name);
1186 break;
1187 case Opt_compress_chksum:
1188 F2FS_OPTION(sbi).compress_chksum = true;
1189 break;
1190 case Opt_compress_mode:
1191 name = match_strdup(&args[0]);
1192 if (!name)
1193 return -ENOMEM;
1194 if (!strcmp(name, "fs")) {
1195 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1196 } else if (!strcmp(name, "user")) {
1197 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1198 } else {
1199 kfree(name);
1200 return -EINVAL;
1201 }
1202 kfree(name);
1203 break;
1204 case Opt_compress_cache:
1205 set_opt(sbi, COMPRESS_CACHE);
1206 break;
1207#else
1208 case Opt_compress_algorithm:
1209 case Opt_compress_log_size:
1210 case Opt_compress_extension:
1211 case Opt_nocompress_extension:
1212 case Opt_compress_chksum:
1213 case Opt_compress_mode:
1214 case Opt_compress_cache:
1215 f2fs_info(sbi, "compression options not supported");
1216 break;
1217#endif
1218 case Opt_atgc:
1219 set_opt(sbi, ATGC);
1220 break;
1221 case Opt_gc_merge:
1222 set_opt(sbi, GC_MERGE);
1223 break;
1224 case Opt_nogc_merge:
1225 clear_opt(sbi, GC_MERGE);
1226 break;
1227 case Opt_discard_unit:
1228 name = match_strdup(&args[0]);
1229 if (!name)
1230 return -ENOMEM;
1231 if (!strcmp(name, "block")) {
1232 F2FS_OPTION(sbi).discard_unit =
1233 DISCARD_UNIT_BLOCK;
1234 } else if (!strcmp(name, "segment")) {
1235 F2FS_OPTION(sbi).discard_unit =
1236 DISCARD_UNIT_SEGMENT;
1237 } else if (!strcmp(name, "section")) {
1238 F2FS_OPTION(sbi).discard_unit =
1239 DISCARD_UNIT_SECTION;
1240 } else {
1241 kfree(name);
1242 return -EINVAL;
1243 }
1244 kfree(name);
1245 break;
1246 case Opt_memory_mode:
1247 name = match_strdup(&args[0]);
1248 if (!name)
1249 return -ENOMEM;
1250 if (!strcmp(name, "normal")) {
1251 F2FS_OPTION(sbi).memory_mode =
1252 MEMORY_MODE_NORMAL;
1253 } else if (!strcmp(name, "low")) {
1254 F2FS_OPTION(sbi).memory_mode =
1255 MEMORY_MODE_LOW;
1256 } else {
1257 kfree(name);
1258 return -EINVAL;
1259 }
1260 kfree(name);
1261 break;
1262 case Opt_age_extent_cache:
1263 set_opt(sbi, AGE_EXTENT_CACHE);
1264 break;
1265 default:
1266 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1267 p);
1268 return -EINVAL;
1269 }
1270 }
1271default_check:
1272#ifdef CONFIG_QUOTA
1273 if (f2fs_check_quota_options(sbi))
1274 return -EINVAL;
1275#else
1276 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1277 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1278 return -EINVAL;
1279 }
1280 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1281 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1282 return -EINVAL;
1283 }
1284#endif
1285#if !IS_ENABLED(CONFIG_UNICODE)
1286 if (f2fs_sb_has_casefold(sbi)) {
1287 f2fs_err(sbi,
1288 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1289 return -EINVAL;
1290 }
1291#endif
1292 /*
1293 * The BLKZONED feature indicates that the drive was formatted with
1294 * zone alignment optimization. This is optional for host-aware
1295 * devices, but mandatory for host-managed zoned block devices.
1296 */
1297#ifndef CONFIG_BLK_DEV_ZONED
1298 if (f2fs_sb_has_blkzoned(sbi)) {
1299 f2fs_err(sbi, "Zoned block device support is not enabled");
1300 return -EINVAL;
1301 }
1302#endif
1303 if (f2fs_sb_has_blkzoned(sbi)) {
1304 if (F2FS_OPTION(sbi).discard_unit !=
1305 DISCARD_UNIT_SECTION) {
1306 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1307 F2FS_OPTION(sbi).discard_unit =
1308 DISCARD_UNIT_SECTION;
1309 }
1310 }
1311
1312#ifdef CONFIG_F2FS_FS_COMPRESSION
1313 if (f2fs_test_compress_extension(sbi)) {
1314 f2fs_err(sbi, "invalid compress or nocompress extension");
1315 return -EINVAL;
1316 }
1317#endif
1318
1319 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1320 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1321 F2FS_IO_SIZE_KB(sbi));
1322 return -EINVAL;
1323 }
1324
1325 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1326 int min_size, max_size;
1327
1328 if (!f2fs_sb_has_extra_attr(sbi) ||
1329 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1330 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1331 return -EINVAL;
1332 }
1333 if (!test_opt(sbi, INLINE_XATTR)) {
1334 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1335 return -EINVAL;
1336 }
1337
1338 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1339 max_size = MAX_INLINE_XATTR_SIZE;
1340
1341 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1342 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1343 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1344 min_size, max_size);
1345 return -EINVAL;
1346 }
1347 }
1348
1349 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1350 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1351 return -EINVAL;
1352 }
1353
1354 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1355 f2fs_err(sbi, "LFS not compatible with ATGC");
1356 return -EINVAL;
1357 }
1358
1359 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1360 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1361 return -EINVAL;
1362 }
1363
1364 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1365 f2fs_err(sbi, "Allow to mount readonly mode only");
1366 return -EROFS;
1367 }
1368 return 0;
1369}
1370
1371static struct inode *f2fs_alloc_inode(struct super_block *sb)
1372{
1373 struct f2fs_inode_info *fi;
1374
1375 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) {
1376 f2fs_show_injection_info(F2FS_SB(sb), FAULT_SLAB_ALLOC);
1377 return NULL;
1378 }
1379
1380 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1381 if (!fi)
1382 return NULL;
1383
1384 init_once((void *) fi);
1385
1386 /* Initialize f2fs-specific inode info */
1387 atomic_set(&fi->dirty_pages, 0);
1388 atomic_set(&fi->i_compr_blocks, 0);
1389 init_f2fs_rwsem(&fi->i_sem);
1390 spin_lock_init(&fi->i_size_lock);
1391 INIT_LIST_HEAD(&fi->dirty_list);
1392 INIT_LIST_HEAD(&fi->gdirty_list);
1393 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1394 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1395 init_f2fs_rwsem(&fi->i_xattr_sem);
1396
1397 /* Will be used by directory only */
1398 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1399
1400 return &fi->vfs_inode;
1401}
1402
1403static int f2fs_drop_inode(struct inode *inode)
1404{
1405 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1406 int ret;
1407
1408 /*
1409 * during filesystem shutdown, if checkpoint is disabled,
1410 * drop useless meta/node dirty pages.
1411 */
1412 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1413 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1414 inode->i_ino == F2FS_META_INO(sbi)) {
1415 trace_f2fs_drop_inode(inode, 1);
1416 return 1;
1417 }
1418 }
1419
1420 /*
1421 * This is to avoid a deadlock condition like below.
1422 * writeback_single_inode(inode)
1423 * - f2fs_write_data_page
1424 * - f2fs_gc -> iput -> evict
1425 * - inode_wait_for_writeback(inode)
1426 */
1427 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1428 if (!inode->i_nlink && !is_bad_inode(inode)) {
1429 /* to avoid evict_inode call simultaneously */
1430 atomic_inc(&inode->i_count);
1431 spin_unlock(&inode->i_lock);
1432
1433 f2fs_abort_atomic_write(inode, true);
1434
1435 /* should remain fi->extent_tree for writepage */
1436 f2fs_destroy_extent_node(inode);
1437
1438 sb_start_intwrite(inode->i_sb);
1439 f2fs_i_size_write(inode, 0);
1440
1441 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1442 inode, NULL, 0, DATA);
1443 truncate_inode_pages_final(inode->i_mapping);
1444
1445 if (F2FS_HAS_BLOCKS(inode))
1446 f2fs_truncate(inode);
1447
1448 sb_end_intwrite(inode->i_sb);
1449
1450 spin_lock(&inode->i_lock);
1451 atomic_dec(&inode->i_count);
1452 }
1453 trace_f2fs_drop_inode(inode, 0);
1454 return 0;
1455 }
1456 ret = generic_drop_inode(inode);
1457 if (!ret)
1458 ret = fscrypt_drop_inode(inode);
1459 trace_f2fs_drop_inode(inode, ret);
1460 return ret;
1461}
1462
1463int f2fs_inode_dirtied(struct inode *inode, bool sync)
1464{
1465 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1466 int ret = 0;
1467
1468 spin_lock(&sbi->inode_lock[DIRTY_META]);
1469 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1470 ret = 1;
1471 } else {
1472 set_inode_flag(inode, FI_DIRTY_INODE);
1473 stat_inc_dirty_inode(sbi, DIRTY_META);
1474 }
1475 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1476 list_add_tail(&F2FS_I(inode)->gdirty_list,
1477 &sbi->inode_list[DIRTY_META]);
1478 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1479 }
1480 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1481 return ret;
1482}
1483
1484void f2fs_inode_synced(struct inode *inode)
1485{
1486 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1487
1488 spin_lock(&sbi->inode_lock[DIRTY_META]);
1489 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1490 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1491 return;
1492 }
1493 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1494 list_del_init(&F2FS_I(inode)->gdirty_list);
1495 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1496 }
1497 clear_inode_flag(inode, FI_DIRTY_INODE);
1498 clear_inode_flag(inode, FI_AUTO_RECOVER);
1499 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1500 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1501}
1502
1503/*
1504 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1505 *
1506 * We should call set_dirty_inode to write the dirty inode through write_inode.
1507 */
1508static void f2fs_dirty_inode(struct inode *inode, int flags)
1509{
1510 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1511
1512 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1513 inode->i_ino == F2FS_META_INO(sbi))
1514 return;
1515
1516 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1517 clear_inode_flag(inode, FI_AUTO_RECOVER);
1518
1519 f2fs_inode_dirtied(inode, false);
1520}
1521
1522static void f2fs_free_inode(struct inode *inode)
1523{
1524 fscrypt_free_inode(inode);
1525 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1526}
1527
1528static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1529{
1530 percpu_counter_destroy(&sbi->total_valid_inode_count);
1531 percpu_counter_destroy(&sbi->rf_node_block_count);
1532 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1533}
1534
1535static void destroy_device_list(struct f2fs_sb_info *sbi)
1536{
1537 int i;
1538
1539 for (i = 0; i < sbi->s_ndevs; i++) {
1540 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1541#ifdef CONFIG_BLK_DEV_ZONED
1542 kvfree(FDEV(i).blkz_seq);
1543#endif
1544 }
1545 kvfree(sbi->devs);
1546}
1547
1548static void f2fs_put_super(struct super_block *sb)
1549{
1550 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1551 int i;
1552 bool dropped;
1553
1554 /* unregister procfs/sysfs entries in advance to avoid race case */
1555 f2fs_unregister_sysfs(sbi);
1556
1557 f2fs_quota_off_umount(sb);
1558
1559 /* prevent remaining shrinker jobs */
1560 mutex_lock(&sbi->umount_mutex);
1561
1562 /*
1563 * flush all issued checkpoints and stop checkpoint issue thread.
1564 * after then, all checkpoints should be done by each process context.
1565 */
1566 f2fs_stop_ckpt_thread(sbi);
1567
1568 /*
1569 * We don't need to do checkpoint when superblock is clean.
1570 * But, the previous checkpoint was not done by umount, it needs to do
1571 * clean checkpoint again.
1572 */
1573 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1574 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1575 struct cp_control cpc = {
1576 .reason = CP_UMOUNT,
1577 };
1578 f2fs_write_checkpoint(sbi, &cpc);
1579 }
1580
1581 /* be sure to wait for any on-going discard commands */
1582 dropped = f2fs_issue_discard_timeout(sbi);
1583
1584 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && !dropped) {
1585 struct cp_control cpc = {
1586 .reason = CP_UMOUNT | CP_TRIMMED,
1587 };
1588 f2fs_write_checkpoint(sbi, &cpc);
1589 }
1590
1591 /*
1592 * normally superblock is clean, so we need to release this.
1593 * In addition, EIO will skip do checkpoint, we need this as well.
1594 */
1595 f2fs_release_ino_entry(sbi, true);
1596
1597 f2fs_leave_shrinker(sbi);
1598 mutex_unlock(&sbi->umount_mutex);
1599
1600 /* our cp_error case, we can wait for any writeback page */
1601 f2fs_flush_merged_writes(sbi);
1602
1603 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1604
1605 f2fs_bug_on(sbi, sbi->fsync_node_num);
1606
1607 f2fs_destroy_compress_inode(sbi);
1608
1609 iput(sbi->node_inode);
1610 sbi->node_inode = NULL;
1611
1612 iput(sbi->meta_inode);
1613 sbi->meta_inode = NULL;
1614
1615 /*
1616 * iput() can update stat information, if f2fs_write_checkpoint()
1617 * above failed with error.
1618 */
1619 f2fs_destroy_stats(sbi);
1620
1621 /* destroy f2fs internal modules */
1622 f2fs_destroy_node_manager(sbi);
1623 f2fs_destroy_segment_manager(sbi);
1624
1625 f2fs_destroy_post_read_wq(sbi);
1626
1627 kvfree(sbi->ckpt);
1628
1629 sb->s_fs_info = NULL;
1630 if (sbi->s_chksum_driver)
1631 crypto_free_shash(sbi->s_chksum_driver);
1632 kfree(sbi->raw_super);
1633
1634 destroy_device_list(sbi);
1635 f2fs_destroy_page_array_cache(sbi);
1636 f2fs_destroy_xattr_caches(sbi);
1637 mempool_destroy(sbi->write_io_dummy);
1638#ifdef CONFIG_QUOTA
1639 for (i = 0; i < MAXQUOTAS; i++)
1640 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1641#endif
1642 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1643 destroy_percpu_info(sbi);
1644 f2fs_destroy_iostat(sbi);
1645 for (i = 0; i < NR_PAGE_TYPE; i++)
1646 kvfree(sbi->write_io[i]);
1647#if IS_ENABLED(CONFIG_UNICODE)
1648 utf8_unload(sb->s_encoding);
1649#endif
1650 kfree(sbi);
1651}
1652
1653int f2fs_sync_fs(struct super_block *sb, int sync)
1654{
1655 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1656 int err = 0;
1657
1658 if (unlikely(f2fs_cp_error(sbi)))
1659 return 0;
1660 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1661 return 0;
1662
1663 trace_f2fs_sync_fs(sb, sync);
1664
1665 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1666 return -EAGAIN;
1667
1668 if (sync)
1669 err = f2fs_issue_checkpoint(sbi);
1670
1671 return err;
1672}
1673
1674static int f2fs_freeze(struct super_block *sb)
1675{
1676 if (f2fs_readonly(sb))
1677 return 0;
1678
1679 /* IO error happened before */
1680 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1681 return -EIO;
1682
1683 /* must be clean, since sync_filesystem() was already called */
1684 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1685 return -EINVAL;
1686
1687 /* Let's flush checkpoints and stop the thread. */
1688 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1689
1690 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1691 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1692 return 0;
1693}
1694
1695static int f2fs_unfreeze(struct super_block *sb)
1696{
1697 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1698 return 0;
1699}
1700
1701#ifdef CONFIG_QUOTA
1702static int f2fs_statfs_project(struct super_block *sb,
1703 kprojid_t projid, struct kstatfs *buf)
1704{
1705 struct kqid qid;
1706 struct dquot *dquot;
1707 u64 limit;
1708 u64 curblock;
1709
1710 qid = make_kqid_projid(projid);
1711 dquot = dqget(sb, qid);
1712 if (IS_ERR(dquot))
1713 return PTR_ERR(dquot);
1714 spin_lock(&dquot->dq_dqb_lock);
1715
1716 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1717 dquot->dq_dqb.dqb_bhardlimit);
1718 if (limit)
1719 limit >>= sb->s_blocksize_bits;
1720
1721 if (limit && buf->f_blocks > limit) {
1722 curblock = (dquot->dq_dqb.dqb_curspace +
1723 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1724 buf->f_blocks = limit;
1725 buf->f_bfree = buf->f_bavail =
1726 (buf->f_blocks > curblock) ?
1727 (buf->f_blocks - curblock) : 0;
1728 }
1729
1730 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1731 dquot->dq_dqb.dqb_ihardlimit);
1732
1733 if (limit && buf->f_files > limit) {
1734 buf->f_files = limit;
1735 buf->f_ffree =
1736 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1737 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1738 }
1739
1740 spin_unlock(&dquot->dq_dqb_lock);
1741 dqput(dquot);
1742 return 0;
1743}
1744#endif
1745
1746static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1747{
1748 struct super_block *sb = dentry->d_sb;
1749 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1750 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1751 block_t total_count, user_block_count, start_count;
1752 u64 avail_node_count;
1753 unsigned int total_valid_node_count;
1754
1755 total_count = le64_to_cpu(sbi->raw_super->block_count);
1756 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1757 buf->f_type = F2FS_SUPER_MAGIC;
1758 buf->f_bsize = sbi->blocksize;
1759
1760 buf->f_blocks = total_count - start_count;
1761
1762 spin_lock(&sbi->stat_lock);
1763
1764 user_block_count = sbi->user_block_count;
1765 total_valid_node_count = valid_node_count(sbi);
1766 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1767 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1768 sbi->current_reserved_blocks;
1769
1770 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1771 buf->f_bfree = 0;
1772 else
1773 buf->f_bfree -= sbi->unusable_block_count;
1774 spin_unlock(&sbi->stat_lock);
1775
1776 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1777 buf->f_bavail = buf->f_bfree -
1778 F2FS_OPTION(sbi).root_reserved_blocks;
1779 else
1780 buf->f_bavail = 0;
1781
1782 if (avail_node_count > user_block_count) {
1783 buf->f_files = user_block_count;
1784 buf->f_ffree = buf->f_bavail;
1785 } else {
1786 buf->f_files = avail_node_count;
1787 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1788 buf->f_bavail);
1789 }
1790
1791 buf->f_namelen = F2FS_NAME_LEN;
1792 buf->f_fsid = u64_to_fsid(id);
1793
1794#ifdef CONFIG_QUOTA
1795 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1796 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1797 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1798 }
1799#endif
1800 return 0;
1801}
1802
1803static inline void f2fs_show_quota_options(struct seq_file *seq,
1804 struct super_block *sb)
1805{
1806#ifdef CONFIG_QUOTA
1807 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1808
1809 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1810 char *fmtname = "";
1811
1812 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1813 case QFMT_VFS_OLD:
1814 fmtname = "vfsold";
1815 break;
1816 case QFMT_VFS_V0:
1817 fmtname = "vfsv0";
1818 break;
1819 case QFMT_VFS_V1:
1820 fmtname = "vfsv1";
1821 break;
1822 }
1823 seq_printf(seq, ",jqfmt=%s", fmtname);
1824 }
1825
1826 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1827 seq_show_option(seq, "usrjquota",
1828 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1829
1830 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1831 seq_show_option(seq, "grpjquota",
1832 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1833
1834 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1835 seq_show_option(seq, "prjjquota",
1836 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1837#endif
1838}
1839
1840#ifdef CONFIG_F2FS_FS_COMPRESSION
1841static inline void f2fs_show_compress_options(struct seq_file *seq,
1842 struct super_block *sb)
1843{
1844 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1845 char *algtype = "";
1846 int i;
1847
1848 if (!f2fs_sb_has_compression(sbi))
1849 return;
1850
1851 switch (F2FS_OPTION(sbi).compress_algorithm) {
1852 case COMPRESS_LZO:
1853 algtype = "lzo";
1854 break;
1855 case COMPRESS_LZ4:
1856 algtype = "lz4";
1857 break;
1858 case COMPRESS_ZSTD:
1859 algtype = "zstd";
1860 break;
1861 case COMPRESS_LZORLE:
1862 algtype = "lzo-rle";
1863 break;
1864 }
1865 seq_printf(seq, ",compress_algorithm=%s", algtype);
1866
1867 if (F2FS_OPTION(sbi).compress_level)
1868 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1869
1870 seq_printf(seq, ",compress_log_size=%u",
1871 F2FS_OPTION(sbi).compress_log_size);
1872
1873 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1874 seq_printf(seq, ",compress_extension=%s",
1875 F2FS_OPTION(sbi).extensions[i]);
1876 }
1877
1878 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1879 seq_printf(seq, ",nocompress_extension=%s",
1880 F2FS_OPTION(sbi).noextensions[i]);
1881 }
1882
1883 if (F2FS_OPTION(sbi).compress_chksum)
1884 seq_puts(seq, ",compress_chksum");
1885
1886 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1887 seq_printf(seq, ",compress_mode=%s", "fs");
1888 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1889 seq_printf(seq, ",compress_mode=%s", "user");
1890
1891 if (test_opt(sbi, COMPRESS_CACHE))
1892 seq_puts(seq, ",compress_cache");
1893}
1894#endif
1895
1896static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1897{
1898 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1899
1900 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1901 seq_printf(seq, ",background_gc=%s", "sync");
1902 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1903 seq_printf(seq, ",background_gc=%s", "on");
1904 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1905 seq_printf(seq, ",background_gc=%s", "off");
1906
1907 if (test_opt(sbi, GC_MERGE))
1908 seq_puts(seq, ",gc_merge");
1909
1910 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1911 seq_puts(seq, ",disable_roll_forward");
1912 if (test_opt(sbi, NORECOVERY))
1913 seq_puts(seq, ",norecovery");
1914 if (test_opt(sbi, DISCARD))
1915 seq_puts(seq, ",discard");
1916 else
1917 seq_puts(seq, ",nodiscard");
1918 if (test_opt(sbi, NOHEAP))
1919 seq_puts(seq, ",no_heap");
1920 else
1921 seq_puts(seq, ",heap");
1922#ifdef CONFIG_F2FS_FS_XATTR
1923 if (test_opt(sbi, XATTR_USER))
1924 seq_puts(seq, ",user_xattr");
1925 else
1926 seq_puts(seq, ",nouser_xattr");
1927 if (test_opt(sbi, INLINE_XATTR))
1928 seq_puts(seq, ",inline_xattr");
1929 else
1930 seq_puts(seq, ",noinline_xattr");
1931 if (test_opt(sbi, INLINE_XATTR_SIZE))
1932 seq_printf(seq, ",inline_xattr_size=%u",
1933 F2FS_OPTION(sbi).inline_xattr_size);
1934#endif
1935#ifdef CONFIG_F2FS_FS_POSIX_ACL
1936 if (test_opt(sbi, POSIX_ACL))
1937 seq_puts(seq, ",acl");
1938 else
1939 seq_puts(seq, ",noacl");
1940#endif
1941 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1942 seq_puts(seq, ",disable_ext_identify");
1943 if (test_opt(sbi, INLINE_DATA))
1944 seq_puts(seq, ",inline_data");
1945 else
1946 seq_puts(seq, ",noinline_data");
1947 if (test_opt(sbi, INLINE_DENTRY))
1948 seq_puts(seq, ",inline_dentry");
1949 else
1950 seq_puts(seq, ",noinline_dentry");
1951 if (test_opt(sbi, FLUSH_MERGE))
1952 seq_puts(seq, ",flush_merge");
1953 else
1954 seq_puts(seq, ",noflush_merge");
1955 if (test_opt(sbi, NOBARRIER))
1956 seq_puts(seq, ",nobarrier");
1957 else
1958 seq_puts(seq, ",barrier");
1959 if (test_opt(sbi, FASTBOOT))
1960 seq_puts(seq, ",fastboot");
1961 if (test_opt(sbi, READ_EXTENT_CACHE))
1962 seq_puts(seq, ",extent_cache");
1963 else
1964 seq_puts(seq, ",noextent_cache");
1965 if (test_opt(sbi, AGE_EXTENT_CACHE))
1966 seq_puts(seq, ",age_extent_cache");
1967 if (test_opt(sbi, DATA_FLUSH))
1968 seq_puts(seq, ",data_flush");
1969
1970 seq_puts(seq, ",mode=");
1971 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1972 seq_puts(seq, "adaptive");
1973 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1974 seq_puts(seq, "lfs");
1975 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1976 seq_puts(seq, "fragment:segment");
1977 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1978 seq_puts(seq, "fragment:block");
1979 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1980 if (test_opt(sbi, RESERVE_ROOT))
1981 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1982 F2FS_OPTION(sbi).root_reserved_blocks,
1983 from_kuid_munged(&init_user_ns,
1984 F2FS_OPTION(sbi).s_resuid),
1985 from_kgid_munged(&init_user_ns,
1986 F2FS_OPTION(sbi).s_resgid));
1987 if (F2FS_IO_SIZE_BITS(sbi))
1988 seq_printf(seq, ",io_bits=%u",
1989 F2FS_OPTION(sbi).write_io_size_bits);
1990#ifdef CONFIG_F2FS_FAULT_INJECTION
1991 if (test_opt(sbi, FAULT_INJECTION)) {
1992 seq_printf(seq, ",fault_injection=%u",
1993 F2FS_OPTION(sbi).fault_info.inject_rate);
1994 seq_printf(seq, ",fault_type=%u",
1995 F2FS_OPTION(sbi).fault_info.inject_type);
1996 }
1997#endif
1998#ifdef CONFIG_QUOTA
1999 if (test_opt(sbi, QUOTA))
2000 seq_puts(seq, ",quota");
2001 if (test_opt(sbi, USRQUOTA))
2002 seq_puts(seq, ",usrquota");
2003 if (test_opt(sbi, GRPQUOTA))
2004 seq_puts(seq, ",grpquota");
2005 if (test_opt(sbi, PRJQUOTA))
2006 seq_puts(seq, ",prjquota");
2007#endif
2008 f2fs_show_quota_options(seq, sbi->sb);
2009
2010 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2011
2012 if (sbi->sb->s_flags & SB_INLINECRYPT)
2013 seq_puts(seq, ",inlinecrypt");
2014
2015 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2016 seq_printf(seq, ",alloc_mode=%s", "default");
2017 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2018 seq_printf(seq, ",alloc_mode=%s", "reuse");
2019
2020 if (test_opt(sbi, DISABLE_CHECKPOINT))
2021 seq_printf(seq, ",checkpoint=disable:%u",
2022 F2FS_OPTION(sbi).unusable_cap);
2023 if (test_opt(sbi, MERGE_CHECKPOINT))
2024 seq_puts(seq, ",checkpoint_merge");
2025 else
2026 seq_puts(seq, ",nocheckpoint_merge");
2027 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2028 seq_printf(seq, ",fsync_mode=%s", "posix");
2029 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2030 seq_printf(seq, ",fsync_mode=%s", "strict");
2031 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2032 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2033
2034#ifdef CONFIG_F2FS_FS_COMPRESSION
2035 f2fs_show_compress_options(seq, sbi->sb);
2036#endif
2037
2038 if (test_opt(sbi, ATGC))
2039 seq_puts(seq, ",atgc");
2040
2041 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2042 seq_printf(seq, ",discard_unit=%s", "block");
2043 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2044 seq_printf(seq, ",discard_unit=%s", "segment");
2045 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2046 seq_printf(seq, ",discard_unit=%s", "section");
2047
2048 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2049 seq_printf(seq, ",memory=%s", "normal");
2050 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2051 seq_printf(seq, ",memory=%s", "low");
2052
2053 return 0;
2054}
2055
2056static void default_options(struct f2fs_sb_info *sbi)
2057{
2058 /* init some FS parameters */
2059 if (f2fs_sb_has_readonly(sbi))
2060 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2061 else
2062 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2063
2064 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2065 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2066 SMALL_VOLUME_SEGMENTS)
2067 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2068 else
2069 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2070 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2071 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2072 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2073 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2074 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2075 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2076 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2077 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2078 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2079
2080 sbi->sb->s_flags &= ~SB_INLINECRYPT;
2081
2082 set_opt(sbi, INLINE_XATTR);
2083 set_opt(sbi, INLINE_DATA);
2084 set_opt(sbi, INLINE_DENTRY);
2085 set_opt(sbi, READ_EXTENT_CACHE);
2086 set_opt(sbi, NOHEAP);
2087 clear_opt(sbi, DISABLE_CHECKPOINT);
2088 set_opt(sbi, MERGE_CHECKPOINT);
2089 F2FS_OPTION(sbi).unusable_cap = 0;
2090 sbi->sb->s_flags |= SB_LAZYTIME;
2091 if (!f2fs_is_readonly(sbi))
2092 set_opt(sbi, FLUSH_MERGE);
2093 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2094 set_opt(sbi, DISCARD);
2095 if (f2fs_sb_has_blkzoned(sbi)) {
2096 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2097 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2098 } else {
2099 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2100 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2101 }
2102
2103#ifdef CONFIG_F2FS_FS_XATTR
2104 set_opt(sbi, XATTR_USER);
2105#endif
2106#ifdef CONFIG_F2FS_FS_POSIX_ACL
2107 set_opt(sbi, POSIX_ACL);
2108#endif
2109
2110 f2fs_build_fault_attr(sbi, 0, 0);
2111}
2112
2113#ifdef CONFIG_QUOTA
2114static int f2fs_enable_quotas(struct super_block *sb);
2115#endif
2116
2117static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2118{
2119 unsigned int s_flags = sbi->sb->s_flags;
2120 struct cp_control cpc;
2121 unsigned int gc_mode = sbi->gc_mode;
2122 int err = 0;
2123 int ret;
2124 block_t unusable;
2125
2126 if (s_flags & SB_RDONLY) {
2127 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2128 return -EINVAL;
2129 }
2130 sbi->sb->s_flags |= SB_ACTIVE;
2131
2132 /* check if we need more GC first */
2133 unusable = f2fs_get_unusable_blocks(sbi);
2134 if (!f2fs_disable_cp_again(sbi, unusable))
2135 goto skip_gc;
2136
2137 f2fs_update_time(sbi, DISABLE_TIME);
2138
2139 sbi->gc_mode = GC_URGENT_HIGH;
2140
2141 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2142 struct f2fs_gc_control gc_control = {
2143 .victim_segno = NULL_SEGNO,
2144 .init_gc_type = FG_GC,
2145 .should_migrate_blocks = false,
2146 .err_gc_skipped = true,
2147 .nr_free_secs = 1 };
2148
2149 f2fs_down_write(&sbi->gc_lock);
2150 err = f2fs_gc(sbi, &gc_control);
2151 if (err == -ENODATA) {
2152 err = 0;
2153 break;
2154 }
2155 if (err && err != -EAGAIN)
2156 break;
2157 }
2158
2159 ret = sync_filesystem(sbi->sb);
2160 if (ret || err) {
2161 err = ret ? ret : err;
2162 goto restore_flag;
2163 }
2164
2165 unusable = f2fs_get_unusable_blocks(sbi);
2166 if (f2fs_disable_cp_again(sbi, unusable)) {
2167 err = -EAGAIN;
2168 goto restore_flag;
2169 }
2170
2171skip_gc:
2172 f2fs_down_write(&sbi->gc_lock);
2173 cpc.reason = CP_PAUSE;
2174 set_sbi_flag(sbi, SBI_CP_DISABLED);
2175 err = f2fs_write_checkpoint(sbi, &cpc);
2176 if (err)
2177 goto out_unlock;
2178
2179 spin_lock(&sbi->stat_lock);
2180 sbi->unusable_block_count = unusable;
2181 spin_unlock(&sbi->stat_lock);
2182
2183out_unlock:
2184 f2fs_up_write(&sbi->gc_lock);
2185restore_flag:
2186 sbi->gc_mode = gc_mode;
2187 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2188 return err;
2189}
2190
2191static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2192{
2193 int retry = DEFAULT_RETRY_IO_COUNT;
2194
2195 /* we should flush all the data to keep data consistency */
2196 do {
2197 sync_inodes_sb(sbi->sb);
2198 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2199 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2200
2201 if (unlikely(retry < 0))
2202 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2203
2204 f2fs_down_write(&sbi->gc_lock);
2205 f2fs_dirty_to_prefree(sbi);
2206
2207 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2208 set_sbi_flag(sbi, SBI_IS_DIRTY);
2209 f2fs_up_write(&sbi->gc_lock);
2210
2211 f2fs_sync_fs(sbi->sb, 1);
2212
2213 /* Let's ensure there's no pending checkpoint anymore */
2214 f2fs_flush_ckpt_thread(sbi);
2215}
2216
2217static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2218{
2219 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2220 struct f2fs_mount_info org_mount_opt;
2221 unsigned long old_sb_flags;
2222 int err;
2223 bool need_restart_gc = false, need_stop_gc = false;
2224 bool need_restart_ckpt = false, need_stop_ckpt = false;
2225 bool need_restart_flush = false, need_stop_flush = false;
2226 bool need_restart_discard = false, need_stop_discard = false;
2227 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2228 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2229 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2230 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2231 bool no_atgc = !test_opt(sbi, ATGC);
2232 bool no_discard = !test_opt(sbi, DISCARD);
2233 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2234 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2235#ifdef CONFIG_QUOTA
2236 int i, j;
2237#endif
2238
2239 /*
2240 * Save the old mount options in case we
2241 * need to restore them.
2242 */
2243 org_mount_opt = sbi->mount_opt;
2244 old_sb_flags = sb->s_flags;
2245
2246#ifdef CONFIG_QUOTA
2247 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2248 for (i = 0; i < MAXQUOTAS; i++) {
2249 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2250 org_mount_opt.s_qf_names[i] =
2251 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2252 GFP_KERNEL);
2253 if (!org_mount_opt.s_qf_names[i]) {
2254 for (j = 0; j < i; j++)
2255 kfree(org_mount_opt.s_qf_names[j]);
2256 return -ENOMEM;
2257 }
2258 } else {
2259 org_mount_opt.s_qf_names[i] = NULL;
2260 }
2261 }
2262#endif
2263
2264 /* recover superblocks we couldn't write due to previous RO mount */
2265 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2266 err = f2fs_commit_super(sbi, false);
2267 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2268 err);
2269 if (!err)
2270 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2271 }
2272
2273 default_options(sbi);
2274
2275 /* parse mount options */
2276 err = parse_options(sb, data, true);
2277 if (err)
2278 goto restore_opts;
2279
2280 /*
2281 * Previous and new state of filesystem is RO,
2282 * so skip checking GC and FLUSH_MERGE conditions.
2283 */
2284 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2285 goto skip;
2286
2287 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2288 err = -EROFS;
2289 goto restore_opts;
2290 }
2291
2292#ifdef CONFIG_QUOTA
2293 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2294 err = dquot_suspend(sb, -1);
2295 if (err < 0)
2296 goto restore_opts;
2297 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2298 /* dquot_resume needs RW */
2299 sb->s_flags &= ~SB_RDONLY;
2300 if (sb_any_quota_suspended(sb)) {
2301 dquot_resume(sb, -1);
2302 } else if (f2fs_sb_has_quota_ino(sbi)) {
2303 err = f2fs_enable_quotas(sb);
2304 if (err)
2305 goto restore_opts;
2306 }
2307 }
2308#endif
2309 /* disallow enable atgc dynamically */
2310 if (no_atgc == !!test_opt(sbi, ATGC)) {
2311 err = -EINVAL;
2312 f2fs_warn(sbi, "switch atgc option is not allowed");
2313 goto restore_opts;
2314 }
2315
2316 /* disallow enable/disable extent_cache dynamically */
2317 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2318 err = -EINVAL;
2319 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2320 goto restore_opts;
2321 }
2322 /* disallow enable/disable age extent_cache dynamically */
2323 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2324 err = -EINVAL;
2325 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2326 goto restore_opts;
2327 }
2328
2329 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2330 err = -EINVAL;
2331 f2fs_warn(sbi, "switch io_bits option is not allowed");
2332 goto restore_opts;
2333 }
2334
2335 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2336 err = -EINVAL;
2337 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2338 goto restore_opts;
2339 }
2340
2341 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2342 err = -EINVAL;
2343 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2344 goto restore_opts;
2345 }
2346
2347 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2348 err = -EINVAL;
2349 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2350 goto restore_opts;
2351 }
2352
2353 /*
2354 * We stop the GC thread if FS is mounted as RO
2355 * or if background_gc = off is passed in mount
2356 * option. Also sync the filesystem.
2357 */
2358 if ((*flags & SB_RDONLY) ||
2359 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2360 !test_opt(sbi, GC_MERGE))) {
2361 if (sbi->gc_thread) {
2362 f2fs_stop_gc_thread(sbi);
2363 need_restart_gc = true;
2364 }
2365 } else if (!sbi->gc_thread) {
2366 err = f2fs_start_gc_thread(sbi);
2367 if (err)
2368 goto restore_opts;
2369 need_stop_gc = true;
2370 }
2371
2372 if (*flags & SB_RDONLY) {
2373 sync_inodes_sb(sb);
2374
2375 set_sbi_flag(sbi, SBI_IS_DIRTY);
2376 set_sbi_flag(sbi, SBI_IS_CLOSE);
2377 f2fs_sync_fs(sb, 1);
2378 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2379 }
2380
2381 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2382 !test_opt(sbi, MERGE_CHECKPOINT)) {
2383 f2fs_stop_ckpt_thread(sbi);
2384 need_restart_ckpt = true;
2385 } else {
2386 /* Flush if the prevous checkpoint, if exists. */
2387 f2fs_flush_ckpt_thread(sbi);
2388
2389 err = f2fs_start_ckpt_thread(sbi);
2390 if (err) {
2391 f2fs_err(sbi,
2392 "Failed to start F2FS issue_checkpoint_thread (%d)",
2393 err);
2394 goto restore_gc;
2395 }
2396 need_stop_ckpt = true;
2397 }
2398
2399 /*
2400 * We stop issue flush thread if FS is mounted as RO
2401 * or if flush_merge is not passed in mount option.
2402 */
2403 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2404 clear_opt(sbi, FLUSH_MERGE);
2405 f2fs_destroy_flush_cmd_control(sbi, false);
2406 need_restart_flush = true;
2407 } else {
2408 err = f2fs_create_flush_cmd_control(sbi);
2409 if (err)
2410 goto restore_ckpt;
2411 need_stop_flush = true;
2412 }
2413
2414 if (no_discard == !!test_opt(sbi, DISCARD)) {
2415 if (test_opt(sbi, DISCARD)) {
2416 err = f2fs_start_discard_thread(sbi);
2417 if (err)
2418 goto restore_flush;
2419 need_stop_discard = true;
2420 } else {
2421 f2fs_stop_discard_thread(sbi);
2422 f2fs_issue_discard_timeout(sbi);
2423 need_restart_discard = true;
2424 }
2425 }
2426
2427 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2428 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2429 err = f2fs_disable_checkpoint(sbi);
2430 if (err)
2431 goto restore_discard;
2432 } else {
2433 f2fs_enable_checkpoint(sbi);
2434 }
2435 }
2436
2437skip:
2438#ifdef CONFIG_QUOTA
2439 /* Release old quota file names */
2440 for (i = 0; i < MAXQUOTAS; i++)
2441 kfree(org_mount_opt.s_qf_names[i]);
2442#endif
2443 /* Update the POSIXACL Flag */
2444 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2445 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2446
2447 limit_reserve_root(sbi);
2448 adjust_unusable_cap_perc(sbi);
2449 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2450 return 0;
2451restore_discard:
2452 if (need_restart_discard) {
2453 if (f2fs_start_discard_thread(sbi))
2454 f2fs_warn(sbi, "discard has been stopped");
2455 } else if (need_stop_discard) {
2456 f2fs_stop_discard_thread(sbi);
2457 }
2458restore_flush:
2459 if (need_restart_flush) {
2460 if (f2fs_create_flush_cmd_control(sbi))
2461 f2fs_warn(sbi, "background flush thread has stopped");
2462 } else if (need_stop_flush) {
2463 clear_opt(sbi, FLUSH_MERGE);
2464 f2fs_destroy_flush_cmd_control(sbi, false);
2465 }
2466restore_ckpt:
2467 if (need_restart_ckpt) {
2468 if (f2fs_start_ckpt_thread(sbi))
2469 f2fs_warn(sbi, "background ckpt thread has stopped");
2470 } else if (need_stop_ckpt) {
2471 f2fs_stop_ckpt_thread(sbi);
2472 }
2473restore_gc:
2474 if (need_restart_gc) {
2475 if (f2fs_start_gc_thread(sbi))
2476 f2fs_warn(sbi, "background gc thread has stopped");
2477 } else if (need_stop_gc) {
2478 f2fs_stop_gc_thread(sbi);
2479 }
2480restore_opts:
2481#ifdef CONFIG_QUOTA
2482 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2483 for (i = 0; i < MAXQUOTAS; i++) {
2484 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2485 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2486 }
2487#endif
2488 sbi->mount_opt = org_mount_opt;
2489 sb->s_flags = old_sb_flags;
2490 return err;
2491}
2492
2493#ifdef CONFIG_QUOTA
2494/* Read data from quotafile */
2495static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2496 size_t len, loff_t off)
2497{
2498 struct inode *inode = sb_dqopt(sb)->files[type];
2499 struct address_space *mapping = inode->i_mapping;
2500 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2501 int offset = off & (sb->s_blocksize - 1);
2502 int tocopy;
2503 size_t toread;
2504 loff_t i_size = i_size_read(inode);
2505 struct page *page;
2506
2507 if (off > i_size)
2508 return 0;
2509
2510 if (off + len > i_size)
2511 len = i_size - off;
2512 toread = len;
2513 while (toread > 0) {
2514 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2515repeat:
2516 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2517 if (IS_ERR(page)) {
2518 if (PTR_ERR(page) == -ENOMEM) {
2519 memalloc_retry_wait(GFP_NOFS);
2520 goto repeat;
2521 }
2522 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2523 return PTR_ERR(page);
2524 }
2525
2526 lock_page(page);
2527
2528 if (unlikely(page->mapping != mapping)) {
2529 f2fs_put_page(page, 1);
2530 goto repeat;
2531 }
2532 if (unlikely(!PageUptodate(page))) {
2533 f2fs_put_page(page, 1);
2534 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2535 return -EIO;
2536 }
2537
2538 memcpy_from_page(data, page, offset, tocopy);
2539 f2fs_put_page(page, 1);
2540
2541 offset = 0;
2542 toread -= tocopy;
2543 data += tocopy;
2544 blkidx++;
2545 }
2546 return len;
2547}
2548
2549/* Write to quotafile */
2550static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2551 const char *data, size_t len, loff_t off)
2552{
2553 struct inode *inode = sb_dqopt(sb)->files[type];
2554 struct address_space *mapping = inode->i_mapping;
2555 const struct address_space_operations *a_ops = mapping->a_ops;
2556 int offset = off & (sb->s_blocksize - 1);
2557 size_t towrite = len;
2558 struct page *page;
2559 void *fsdata = NULL;
2560 int err = 0;
2561 int tocopy;
2562
2563 while (towrite > 0) {
2564 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2565 towrite);
2566retry:
2567 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2568 &page, &fsdata);
2569 if (unlikely(err)) {
2570 if (err == -ENOMEM) {
2571 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2572 goto retry;
2573 }
2574 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2575 break;
2576 }
2577
2578 memcpy_to_page(page, offset, data, tocopy);
2579
2580 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2581 page, fsdata);
2582 offset = 0;
2583 towrite -= tocopy;
2584 off += tocopy;
2585 data += tocopy;
2586 cond_resched();
2587 }
2588
2589 if (len == towrite)
2590 return err;
2591 inode->i_mtime = inode->i_ctime = current_time(inode);
2592 f2fs_mark_inode_dirty_sync(inode, false);
2593 return len - towrite;
2594}
2595
2596int f2fs_dquot_initialize(struct inode *inode)
2597{
2598 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) {
2599 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT);
2600 return -ESRCH;
2601 }
2602
2603 return dquot_initialize(inode);
2604}
2605
2606static struct dquot **f2fs_get_dquots(struct inode *inode)
2607{
2608 return F2FS_I(inode)->i_dquot;
2609}
2610
2611static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2612{
2613 return &F2FS_I(inode)->i_reserved_quota;
2614}
2615
2616static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2617{
2618 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2619 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2620 return 0;
2621 }
2622
2623 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2624 F2FS_OPTION(sbi).s_jquota_fmt, type);
2625}
2626
2627int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2628{
2629 int enabled = 0;
2630 int i, err;
2631
2632 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2633 err = f2fs_enable_quotas(sbi->sb);
2634 if (err) {
2635 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2636 return 0;
2637 }
2638 return 1;
2639 }
2640
2641 for (i = 0; i < MAXQUOTAS; i++) {
2642 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2643 err = f2fs_quota_on_mount(sbi, i);
2644 if (!err) {
2645 enabled = 1;
2646 continue;
2647 }
2648 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2649 err, i);
2650 }
2651 }
2652 return enabled;
2653}
2654
2655static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2656 unsigned int flags)
2657{
2658 struct inode *qf_inode;
2659 unsigned long qf_inum;
2660 int err;
2661
2662 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2663
2664 qf_inum = f2fs_qf_ino(sb, type);
2665 if (!qf_inum)
2666 return -EPERM;
2667
2668 qf_inode = f2fs_iget(sb, qf_inum);
2669 if (IS_ERR(qf_inode)) {
2670 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2671 return PTR_ERR(qf_inode);
2672 }
2673
2674 /* Don't account quota for quota files to avoid recursion */
2675 qf_inode->i_flags |= S_NOQUOTA;
2676 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2677 iput(qf_inode);
2678 return err;
2679}
2680
2681static int f2fs_enable_quotas(struct super_block *sb)
2682{
2683 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2684 int type, err = 0;
2685 unsigned long qf_inum;
2686 bool quota_mopt[MAXQUOTAS] = {
2687 test_opt(sbi, USRQUOTA),
2688 test_opt(sbi, GRPQUOTA),
2689 test_opt(sbi, PRJQUOTA),
2690 };
2691
2692 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2693 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2694 return 0;
2695 }
2696
2697 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2698
2699 for (type = 0; type < MAXQUOTAS; type++) {
2700 qf_inum = f2fs_qf_ino(sb, type);
2701 if (qf_inum) {
2702 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2703 DQUOT_USAGE_ENABLED |
2704 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2705 if (err) {
2706 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2707 type, err);
2708 for (type--; type >= 0; type--)
2709 dquot_quota_off(sb, type);
2710 set_sbi_flag(F2FS_SB(sb),
2711 SBI_QUOTA_NEED_REPAIR);
2712 return err;
2713 }
2714 }
2715 }
2716 return 0;
2717}
2718
2719static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2720{
2721 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2722 struct address_space *mapping = dqopt->files[type]->i_mapping;
2723 int ret = 0;
2724
2725 ret = dquot_writeback_dquots(sbi->sb, type);
2726 if (ret)
2727 goto out;
2728
2729 ret = filemap_fdatawrite(mapping);
2730 if (ret)
2731 goto out;
2732
2733 /* if we are using journalled quota */
2734 if (is_journalled_quota(sbi))
2735 goto out;
2736
2737 ret = filemap_fdatawait(mapping);
2738
2739 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2740out:
2741 if (ret)
2742 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2743 return ret;
2744}
2745
2746int f2fs_quota_sync(struct super_block *sb, int type)
2747{
2748 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2749 struct quota_info *dqopt = sb_dqopt(sb);
2750 int cnt;
2751 int ret = 0;
2752
2753 /*
2754 * Now when everything is written we can discard the pagecache so
2755 * that userspace sees the changes.
2756 */
2757 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2758
2759 if (type != -1 && cnt != type)
2760 continue;
2761
2762 if (!sb_has_quota_active(sb, cnt))
2763 continue;
2764
2765 if (!f2fs_sb_has_quota_ino(sbi))
2766 inode_lock(dqopt->files[cnt]);
2767
2768 /*
2769 * do_quotactl
2770 * f2fs_quota_sync
2771 * f2fs_down_read(quota_sem)
2772 * dquot_writeback_dquots()
2773 * f2fs_dquot_commit
2774 * block_operation
2775 * f2fs_down_read(quota_sem)
2776 */
2777 f2fs_lock_op(sbi);
2778 f2fs_down_read(&sbi->quota_sem);
2779
2780 ret = f2fs_quota_sync_file(sbi, cnt);
2781
2782 f2fs_up_read(&sbi->quota_sem);
2783 f2fs_unlock_op(sbi);
2784
2785 if (!f2fs_sb_has_quota_ino(sbi))
2786 inode_unlock(dqopt->files[cnt]);
2787
2788 if (ret)
2789 break;
2790 }
2791 return ret;
2792}
2793
2794static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2795 const struct path *path)
2796{
2797 struct inode *inode;
2798 int err;
2799
2800 /* if quota sysfile exists, deny enabling quota with specific file */
2801 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2802 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2803 return -EBUSY;
2804 }
2805
2806 err = f2fs_quota_sync(sb, type);
2807 if (err)
2808 return err;
2809
2810 err = dquot_quota_on(sb, type, format_id, path);
2811 if (err)
2812 return err;
2813
2814 inode = d_inode(path->dentry);
2815
2816 inode_lock(inode);
2817 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2818 f2fs_set_inode_flags(inode);
2819 inode_unlock(inode);
2820 f2fs_mark_inode_dirty_sync(inode, false);
2821
2822 return 0;
2823}
2824
2825static int __f2fs_quota_off(struct super_block *sb, int type)
2826{
2827 struct inode *inode = sb_dqopt(sb)->files[type];
2828 int err;
2829
2830 if (!inode || !igrab(inode))
2831 return dquot_quota_off(sb, type);
2832
2833 err = f2fs_quota_sync(sb, type);
2834 if (err)
2835 goto out_put;
2836
2837 err = dquot_quota_off(sb, type);
2838 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2839 goto out_put;
2840
2841 inode_lock(inode);
2842 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2843 f2fs_set_inode_flags(inode);
2844 inode_unlock(inode);
2845 f2fs_mark_inode_dirty_sync(inode, false);
2846out_put:
2847 iput(inode);
2848 return err;
2849}
2850
2851static int f2fs_quota_off(struct super_block *sb, int type)
2852{
2853 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2854 int err;
2855
2856 err = __f2fs_quota_off(sb, type);
2857
2858 /*
2859 * quotactl can shutdown journalled quota, result in inconsistence
2860 * between quota record and fs data by following updates, tag the
2861 * flag to let fsck be aware of it.
2862 */
2863 if (is_journalled_quota(sbi))
2864 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2865 return err;
2866}
2867
2868void f2fs_quota_off_umount(struct super_block *sb)
2869{
2870 int type;
2871 int err;
2872
2873 for (type = 0; type < MAXQUOTAS; type++) {
2874 err = __f2fs_quota_off(sb, type);
2875 if (err) {
2876 int ret = dquot_quota_off(sb, type);
2877
2878 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2879 type, err, ret);
2880 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2881 }
2882 }
2883 /*
2884 * In case of checkpoint=disable, we must flush quota blocks.
2885 * This can cause NULL exception for node_inode in end_io, since
2886 * put_super already dropped it.
2887 */
2888 sync_filesystem(sb);
2889}
2890
2891static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2892{
2893 struct quota_info *dqopt = sb_dqopt(sb);
2894 int type;
2895
2896 for (type = 0; type < MAXQUOTAS; type++) {
2897 if (!dqopt->files[type])
2898 continue;
2899 f2fs_inode_synced(dqopt->files[type]);
2900 }
2901}
2902
2903static int f2fs_dquot_commit(struct dquot *dquot)
2904{
2905 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2906 int ret;
2907
2908 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2909 ret = dquot_commit(dquot);
2910 if (ret < 0)
2911 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2912 f2fs_up_read(&sbi->quota_sem);
2913 return ret;
2914}
2915
2916static int f2fs_dquot_acquire(struct dquot *dquot)
2917{
2918 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2919 int ret;
2920
2921 f2fs_down_read(&sbi->quota_sem);
2922 ret = dquot_acquire(dquot);
2923 if (ret < 0)
2924 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2925 f2fs_up_read(&sbi->quota_sem);
2926 return ret;
2927}
2928
2929static int f2fs_dquot_release(struct dquot *dquot)
2930{
2931 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2932 int ret = dquot_release(dquot);
2933
2934 if (ret < 0)
2935 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2936 return ret;
2937}
2938
2939static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2940{
2941 struct super_block *sb = dquot->dq_sb;
2942 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2943 int ret = dquot_mark_dquot_dirty(dquot);
2944
2945 /* if we are using journalled quota */
2946 if (is_journalled_quota(sbi))
2947 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2948
2949 return ret;
2950}
2951
2952static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2953{
2954 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2955 int ret = dquot_commit_info(sb, type);
2956
2957 if (ret < 0)
2958 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2959 return ret;
2960}
2961
2962static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2963{
2964 *projid = F2FS_I(inode)->i_projid;
2965 return 0;
2966}
2967
2968static const struct dquot_operations f2fs_quota_operations = {
2969 .get_reserved_space = f2fs_get_reserved_space,
2970 .write_dquot = f2fs_dquot_commit,
2971 .acquire_dquot = f2fs_dquot_acquire,
2972 .release_dquot = f2fs_dquot_release,
2973 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2974 .write_info = f2fs_dquot_commit_info,
2975 .alloc_dquot = dquot_alloc,
2976 .destroy_dquot = dquot_destroy,
2977 .get_projid = f2fs_get_projid,
2978 .get_next_id = dquot_get_next_id,
2979};
2980
2981static const struct quotactl_ops f2fs_quotactl_ops = {
2982 .quota_on = f2fs_quota_on,
2983 .quota_off = f2fs_quota_off,
2984 .quota_sync = f2fs_quota_sync,
2985 .get_state = dquot_get_state,
2986 .set_info = dquot_set_dqinfo,
2987 .get_dqblk = dquot_get_dqblk,
2988 .set_dqblk = dquot_set_dqblk,
2989 .get_nextdqblk = dquot_get_next_dqblk,
2990};
2991#else
2992int f2fs_dquot_initialize(struct inode *inode)
2993{
2994 return 0;
2995}
2996
2997int f2fs_quota_sync(struct super_block *sb, int type)
2998{
2999 return 0;
3000}
3001
3002void f2fs_quota_off_umount(struct super_block *sb)
3003{
3004}
3005#endif
3006
3007static const struct super_operations f2fs_sops = {
3008 .alloc_inode = f2fs_alloc_inode,
3009 .free_inode = f2fs_free_inode,
3010 .drop_inode = f2fs_drop_inode,
3011 .write_inode = f2fs_write_inode,
3012 .dirty_inode = f2fs_dirty_inode,
3013 .show_options = f2fs_show_options,
3014#ifdef CONFIG_QUOTA
3015 .quota_read = f2fs_quota_read,
3016 .quota_write = f2fs_quota_write,
3017 .get_dquots = f2fs_get_dquots,
3018#endif
3019 .evict_inode = f2fs_evict_inode,
3020 .put_super = f2fs_put_super,
3021 .sync_fs = f2fs_sync_fs,
3022 .freeze_fs = f2fs_freeze,
3023 .unfreeze_fs = f2fs_unfreeze,
3024 .statfs = f2fs_statfs,
3025 .remount_fs = f2fs_remount,
3026};
3027
3028#ifdef CONFIG_FS_ENCRYPTION
3029static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3030{
3031 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3032 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3033 ctx, len, NULL);
3034}
3035
3036static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3037 void *fs_data)
3038{
3039 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3040
3041 /*
3042 * Encrypting the root directory is not allowed because fsck
3043 * expects lost+found directory to exist and remain unencrypted
3044 * if LOST_FOUND feature is enabled.
3045 *
3046 */
3047 if (f2fs_sb_has_lost_found(sbi) &&
3048 inode->i_ino == F2FS_ROOT_INO(sbi))
3049 return -EPERM;
3050
3051 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3052 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3053 ctx, len, fs_data, XATTR_CREATE);
3054}
3055
3056static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3057{
3058 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3059}
3060
3061static bool f2fs_has_stable_inodes(struct super_block *sb)
3062{
3063 return true;
3064}
3065
3066static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3067 int *ino_bits_ret, int *lblk_bits_ret)
3068{
3069 *ino_bits_ret = 8 * sizeof(nid_t);
3070 *lblk_bits_ret = 8 * sizeof(block_t);
3071}
3072
3073static struct block_device **f2fs_get_devices(struct super_block *sb,
3074 unsigned int *num_devs)
3075{
3076 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3077 struct block_device **devs;
3078 int i;
3079
3080 if (!f2fs_is_multi_device(sbi))
3081 return NULL;
3082
3083 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3084 if (!devs)
3085 return ERR_PTR(-ENOMEM);
3086
3087 for (i = 0; i < sbi->s_ndevs; i++)
3088 devs[i] = FDEV(i).bdev;
3089 *num_devs = sbi->s_ndevs;
3090 return devs;
3091}
3092
3093static const struct fscrypt_operations f2fs_cryptops = {
3094 .key_prefix = "f2fs:",
3095 .get_context = f2fs_get_context,
3096 .set_context = f2fs_set_context,
3097 .get_dummy_policy = f2fs_get_dummy_policy,
3098 .empty_dir = f2fs_empty_dir,
3099 .has_stable_inodes = f2fs_has_stable_inodes,
3100 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
3101 .get_devices = f2fs_get_devices,
3102};
3103#endif
3104
3105static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3106 u64 ino, u32 generation)
3107{
3108 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3109 struct inode *inode;
3110
3111 if (f2fs_check_nid_range(sbi, ino))
3112 return ERR_PTR(-ESTALE);
3113
3114 /*
3115 * f2fs_iget isn't quite right if the inode is currently unallocated!
3116 * However f2fs_iget currently does appropriate checks to handle stale
3117 * inodes so everything is OK.
3118 */
3119 inode = f2fs_iget(sb, ino);
3120 if (IS_ERR(inode))
3121 return ERR_CAST(inode);
3122 if (unlikely(generation && inode->i_generation != generation)) {
3123 /* we didn't find the right inode.. */
3124 iput(inode);
3125 return ERR_PTR(-ESTALE);
3126 }
3127 return inode;
3128}
3129
3130static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3131 int fh_len, int fh_type)
3132{
3133 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3134 f2fs_nfs_get_inode);
3135}
3136
3137static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3138 int fh_len, int fh_type)
3139{
3140 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3141 f2fs_nfs_get_inode);
3142}
3143
3144static const struct export_operations f2fs_export_ops = {
3145 .fh_to_dentry = f2fs_fh_to_dentry,
3146 .fh_to_parent = f2fs_fh_to_parent,
3147 .get_parent = f2fs_get_parent,
3148};
3149
3150loff_t max_file_blocks(struct inode *inode)
3151{
3152 loff_t result = 0;
3153 loff_t leaf_count;
3154
3155 /*
3156 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3157 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3158 * space in inode.i_addr, it will be more safe to reassign
3159 * result as zero.
3160 */
3161
3162 if (inode && f2fs_compressed_file(inode))
3163 leaf_count = ADDRS_PER_BLOCK(inode);
3164 else
3165 leaf_count = DEF_ADDRS_PER_BLOCK;
3166
3167 /* two direct node blocks */
3168 result += (leaf_count * 2);
3169
3170 /* two indirect node blocks */
3171 leaf_count *= NIDS_PER_BLOCK;
3172 result += (leaf_count * 2);
3173
3174 /* one double indirect node block */
3175 leaf_count *= NIDS_PER_BLOCK;
3176 result += leaf_count;
3177
3178 return result;
3179}
3180
3181static int __f2fs_commit_super(struct buffer_head *bh,
3182 struct f2fs_super_block *super)
3183{
3184 lock_buffer(bh);
3185 if (super)
3186 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3187 set_buffer_dirty(bh);
3188 unlock_buffer(bh);
3189
3190 /* it's rare case, we can do fua all the time */
3191 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3192}
3193
3194static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3195 struct buffer_head *bh)
3196{
3197 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3198 (bh->b_data + F2FS_SUPER_OFFSET);
3199 struct super_block *sb = sbi->sb;
3200 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3201 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3202 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3203 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3204 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3205 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3206 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3207 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3208 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3209 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3210 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3211 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3212 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3213 u64 main_end_blkaddr = main_blkaddr +
3214 (segment_count_main << log_blocks_per_seg);
3215 u64 seg_end_blkaddr = segment0_blkaddr +
3216 (segment_count << log_blocks_per_seg);
3217
3218 if (segment0_blkaddr != cp_blkaddr) {
3219 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3220 segment0_blkaddr, cp_blkaddr);
3221 return true;
3222 }
3223
3224 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3225 sit_blkaddr) {
3226 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3227 cp_blkaddr, sit_blkaddr,
3228 segment_count_ckpt << log_blocks_per_seg);
3229 return true;
3230 }
3231
3232 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3233 nat_blkaddr) {
3234 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3235 sit_blkaddr, nat_blkaddr,
3236 segment_count_sit << log_blocks_per_seg);
3237 return true;
3238 }
3239
3240 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3241 ssa_blkaddr) {
3242 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3243 nat_blkaddr, ssa_blkaddr,
3244 segment_count_nat << log_blocks_per_seg);
3245 return true;
3246 }
3247
3248 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3249 main_blkaddr) {
3250 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3251 ssa_blkaddr, main_blkaddr,
3252 segment_count_ssa << log_blocks_per_seg);
3253 return true;
3254 }
3255
3256 if (main_end_blkaddr > seg_end_blkaddr) {
3257 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3258 main_blkaddr, seg_end_blkaddr,
3259 segment_count_main << log_blocks_per_seg);
3260 return true;
3261 } else if (main_end_blkaddr < seg_end_blkaddr) {
3262 int err = 0;
3263 char *res;
3264
3265 /* fix in-memory information all the time */
3266 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3267 segment0_blkaddr) >> log_blocks_per_seg);
3268
3269 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3270 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3271 res = "internally";
3272 } else {
3273 err = __f2fs_commit_super(bh, NULL);
3274 res = err ? "failed" : "done";
3275 }
3276 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3277 res, main_blkaddr, seg_end_blkaddr,
3278 segment_count_main << log_blocks_per_seg);
3279 if (err)
3280 return true;
3281 }
3282 return false;
3283}
3284
3285static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3286 struct buffer_head *bh)
3287{
3288 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3289 block_t total_sections, blocks_per_seg;
3290 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3291 (bh->b_data + F2FS_SUPER_OFFSET);
3292 size_t crc_offset = 0;
3293 __u32 crc = 0;
3294
3295 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3296 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3297 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3298 return -EINVAL;
3299 }
3300
3301 /* Check checksum_offset and crc in superblock */
3302 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3303 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3304 if (crc_offset !=
3305 offsetof(struct f2fs_super_block, crc)) {
3306 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3307 crc_offset);
3308 return -EFSCORRUPTED;
3309 }
3310 crc = le32_to_cpu(raw_super->crc);
3311 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3312 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3313 return -EFSCORRUPTED;
3314 }
3315 }
3316
3317 /* Currently, support only 4KB block size */
3318 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3319 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3320 le32_to_cpu(raw_super->log_blocksize),
3321 F2FS_BLKSIZE_BITS);
3322 return -EFSCORRUPTED;
3323 }
3324
3325 /* check log blocks per segment */
3326 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3327 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3328 le32_to_cpu(raw_super->log_blocks_per_seg));
3329 return -EFSCORRUPTED;
3330 }
3331
3332 /* Currently, support 512/1024/2048/4096 bytes sector size */
3333 if (le32_to_cpu(raw_super->log_sectorsize) >
3334 F2FS_MAX_LOG_SECTOR_SIZE ||
3335 le32_to_cpu(raw_super->log_sectorsize) <
3336 F2FS_MIN_LOG_SECTOR_SIZE) {
3337 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3338 le32_to_cpu(raw_super->log_sectorsize));
3339 return -EFSCORRUPTED;
3340 }
3341 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3342 le32_to_cpu(raw_super->log_sectorsize) !=
3343 F2FS_MAX_LOG_SECTOR_SIZE) {
3344 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3345 le32_to_cpu(raw_super->log_sectors_per_block),
3346 le32_to_cpu(raw_super->log_sectorsize));
3347 return -EFSCORRUPTED;
3348 }
3349
3350 segment_count = le32_to_cpu(raw_super->segment_count);
3351 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3352 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3353 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3354 total_sections = le32_to_cpu(raw_super->section_count);
3355
3356 /* blocks_per_seg should be 512, given the above check */
3357 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3358
3359 if (segment_count > F2FS_MAX_SEGMENT ||
3360 segment_count < F2FS_MIN_SEGMENTS) {
3361 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3362 return -EFSCORRUPTED;
3363 }
3364
3365 if (total_sections > segment_count_main || total_sections < 1 ||
3366 segs_per_sec > segment_count || !segs_per_sec) {
3367 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3368 segment_count, total_sections, segs_per_sec);
3369 return -EFSCORRUPTED;
3370 }
3371
3372 if (segment_count_main != total_sections * segs_per_sec) {
3373 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3374 segment_count_main, total_sections, segs_per_sec);
3375 return -EFSCORRUPTED;
3376 }
3377
3378 if ((segment_count / segs_per_sec) < total_sections) {
3379 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3380 segment_count, segs_per_sec, total_sections);
3381 return -EFSCORRUPTED;
3382 }
3383
3384 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3385 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3386 segment_count, le64_to_cpu(raw_super->block_count));
3387 return -EFSCORRUPTED;
3388 }
3389
3390 if (RDEV(0).path[0]) {
3391 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3392 int i = 1;
3393
3394 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3395 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3396 i++;
3397 }
3398 if (segment_count != dev_seg_count) {
3399 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3400 segment_count, dev_seg_count);
3401 return -EFSCORRUPTED;
3402 }
3403 } else {
3404 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3405 !bdev_is_zoned(sbi->sb->s_bdev)) {
3406 f2fs_info(sbi, "Zoned block device path is missing");
3407 return -EFSCORRUPTED;
3408 }
3409 }
3410
3411 if (secs_per_zone > total_sections || !secs_per_zone) {
3412 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3413 secs_per_zone, total_sections);
3414 return -EFSCORRUPTED;
3415 }
3416 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3417 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3418 (le32_to_cpu(raw_super->extension_count) +
3419 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3420 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3421 le32_to_cpu(raw_super->extension_count),
3422 raw_super->hot_ext_count,
3423 F2FS_MAX_EXTENSION);
3424 return -EFSCORRUPTED;
3425 }
3426
3427 if (le32_to_cpu(raw_super->cp_payload) >=
3428 (blocks_per_seg - F2FS_CP_PACKS -
3429 NR_CURSEG_PERSIST_TYPE)) {
3430 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3431 le32_to_cpu(raw_super->cp_payload),
3432 blocks_per_seg - F2FS_CP_PACKS -
3433 NR_CURSEG_PERSIST_TYPE);
3434 return -EFSCORRUPTED;
3435 }
3436
3437 /* check reserved ino info */
3438 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3439 le32_to_cpu(raw_super->meta_ino) != 2 ||
3440 le32_to_cpu(raw_super->root_ino) != 3) {
3441 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3442 le32_to_cpu(raw_super->node_ino),
3443 le32_to_cpu(raw_super->meta_ino),
3444 le32_to_cpu(raw_super->root_ino));
3445 return -EFSCORRUPTED;
3446 }
3447
3448 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3449 if (sanity_check_area_boundary(sbi, bh))
3450 return -EFSCORRUPTED;
3451
3452 return 0;
3453}
3454
3455int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3456{
3457 unsigned int total, fsmeta;
3458 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3459 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3460 unsigned int ovp_segments, reserved_segments;
3461 unsigned int main_segs, blocks_per_seg;
3462 unsigned int sit_segs, nat_segs;
3463 unsigned int sit_bitmap_size, nat_bitmap_size;
3464 unsigned int log_blocks_per_seg;
3465 unsigned int segment_count_main;
3466 unsigned int cp_pack_start_sum, cp_payload;
3467 block_t user_block_count, valid_user_blocks;
3468 block_t avail_node_count, valid_node_count;
3469 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3470 int i, j;
3471
3472 total = le32_to_cpu(raw_super->segment_count);
3473 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3474 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3475 fsmeta += sit_segs;
3476 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3477 fsmeta += nat_segs;
3478 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3479 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3480
3481 if (unlikely(fsmeta >= total))
3482 return 1;
3483
3484 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3485 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3486
3487 if (!f2fs_sb_has_readonly(sbi) &&
3488 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3489 ovp_segments == 0 || reserved_segments == 0)) {
3490 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3491 return 1;
3492 }
3493 user_block_count = le64_to_cpu(ckpt->user_block_count);
3494 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3495 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3496 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3497 if (!user_block_count || user_block_count >=
3498 segment_count_main << log_blocks_per_seg) {
3499 f2fs_err(sbi, "Wrong user_block_count: %u",
3500 user_block_count);
3501 return 1;
3502 }
3503
3504 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3505 if (valid_user_blocks > user_block_count) {
3506 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3507 valid_user_blocks, user_block_count);
3508 return 1;
3509 }
3510
3511 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3512 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3513 if (valid_node_count > avail_node_count) {
3514 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3515 valid_node_count, avail_node_count);
3516 return 1;
3517 }
3518
3519 main_segs = le32_to_cpu(raw_super->segment_count_main);
3520 blocks_per_seg = sbi->blocks_per_seg;
3521
3522 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3523 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3524 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3525 return 1;
3526
3527 if (f2fs_sb_has_readonly(sbi))
3528 goto check_data;
3529
3530 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3531 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3532 le32_to_cpu(ckpt->cur_node_segno[j])) {
3533 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3534 i, j,
3535 le32_to_cpu(ckpt->cur_node_segno[i]));
3536 return 1;
3537 }
3538 }
3539 }
3540check_data:
3541 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3542 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3543 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3544 return 1;
3545
3546 if (f2fs_sb_has_readonly(sbi))
3547 goto skip_cross;
3548
3549 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3550 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3551 le32_to_cpu(ckpt->cur_data_segno[j])) {
3552 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3553 i, j,
3554 le32_to_cpu(ckpt->cur_data_segno[i]));
3555 return 1;
3556 }
3557 }
3558 }
3559 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3560 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3561 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3562 le32_to_cpu(ckpt->cur_data_segno[j])) {
3563 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3564 i, j,
3565 le32_to_cpu(ckpt->cur_node_segno[i]));
3566 return 1;
3567 }
3568 }
3569 }
3570skip_cross:
3571 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3572 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3573
3574 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3575 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3576 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3577 sit_bitmap_size, nat_bitmap_size);
3578 return 1;
3579 }
3580
3581 cp_pack_start_sum = __start_sum_addr(sbi);
3582 cp_payload = __cp_payload(sbi);
3583 if (cp_pack_start_sum < cp_payload + 1 ||
3584 cp_pack_start_sum > blocks_per_seg - 1 -
3585 NR_CURSEG_PERSIST_TYPE) {
3586 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3587 cp_pack_start_sum);
3588 return 1;
3589 }
3590
3591 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3592 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3593 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3594 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3595 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3596 le32_to_cpu(ckpt->checksum_offset));
3597 return 1;
3598 }
3599
3600 nat_blocks = nat_segs << log_blocks_per_seg;
3601 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3602 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3603 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3604 (cp_payload + F2FS_CP_PACKS +
3605 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3606 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3607 cp_payload, nat_bits_blocks);
3608 return 1;
3609 }
3610
3611 if (unlikely(f2fs_cp_error(sbi))) {
3612 f2fs_err(sbi, "A bug case: need to run fsck");
3613 return 1;
3614 }
3615 return 0;
3616}
3617
3618static void init_sb_info(struct f2fs_sb_info *sbi)
3619{
3620 struct f2fs_super_block *raw_super = sbi->raw_super;
3621 int i;
3622
3623 sbi->log_sectors_per_block =
3624 le32_to_cpu(raw_super->log_sectors_per_block);
3625 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3626 sbi->blocksize = 1 << sbi->log_blocksize;
3627 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3628 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3629 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3630 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3631 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3632 sbi->total_node_count =
3633 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3634 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3635 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3636 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3637 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3638 sbi->cur_victim_sec = NULL_SECNO;
3639 sbi->gc_mode = GC_NORMAL;
3640 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3641 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3642 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3643 sbi->migration_granularity = sbi->segs_per_sec;
3644 sbi->seq_file_ra_mul = MIN_RA_MUL;
3645 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3646 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3647 spin_lock_init(&sbi->gc_remaining_trials_lock);
3648 atomic64_set(&sbi->current_atomic_write, 0);
3649
3650 sbi->dir_level = DEF_DIR_LEVEL;
3651 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3652 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3653 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3654 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3655 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3656 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3657 DEF_UMOUNT_DISCARD_TIMEOUT;
3658 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3659
3660 for (i = 0; i < NR_COUNT_TYPE; i++)
3661 atomic_set(&sbi->nr_pages[i], 0);
3662
3663 for (i = 0; i < META; i++)
3664 atomic_set(&sbi->wb_sync_req[i], 0);
3665
3666 INIT_LIST_HEAD(&sbi->s_list);
3667 mutex_init(&sbi->umount_mutex);
3668 init_f2fs_rwsem(&sbi->io_order_lock);
3669 spin_lock_init(&sbi->cp_lock);
3670
3671 sbi->dirty_device = 0;
3672 spin_lock_init(&sbi->dev_lock);
3673
3674 init_f2fs_rwsem(&sbi->sb_lock);
3675 init_f2fs_rwsem(&sbi->pin_sem);
3676}
3677
3678static int init_percpu_info(struct f2fs_sb_info *sbi)
3679{
3680 int err;
3681
3682 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3683 if (err)
3684 return err;
3685
3686 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3687 if (err)
3688 goto err_valid_block;
3689
3690 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3691 GFP_KERNEL);
3692 if (err)
3693 goto err_node_block;
3694 return 0;
3695
3696err_node_block:
3697 percpu_counter_destroy(&sbi->rf_node_block_count);
3698err_valid_block:
3699 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3700 return err;
3701}
3702
3703#ifdef CONFIG_BLK_DEV_ZONED
3704
3705struct f2fs_report_zones_args {
3706 struct f2fs_sb_info *sbi;
3707 struct f2fs_dev_info *dev;
3708};
3709
3710static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3711 void *data)
3712{
3713 struct f2fs_report_zones_args *rz_args = data;
3714 block_t unusable_blocks = (zone->len - zone->capacity) >>
3715 F2FS_LOG_SECTORS_PER_BLOCK;
3716
3717 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3718 return 0;
3719
3720 set_bit(idx, rz_args->dev->blkz_seq);
3721 if (!rz_args->sbi->unusable_blocks_per_sec) {
3722 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3723 return 0;
3724 }
3725 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3726 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3727 return -EINVAL;
3728 }
3729 return 0;
3730}
3731
3732static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3733{
3734 struct block_device *bdev = FDEV(devi).bdev;
3735 sector_t nr_sectors = bdev_nr_sectors(bdev);
3736 struct f2fs_report_zones_args rep_zone_arg;
3737 u64 zone_sectors;
3738 int ret;
3739
3740 if (!f2fs_sb_has_blkzoned(sbi))
3741 return 0;
3742
3743 zone_sectors = bdev_zone_sectors(bdev);
3744 if (!is_power_of_2(zone_sectors)) {
3745 f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n");
3746 return -EINVAL;
3747 }
3748
3749 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3750 SECTOR_TO_BLOCK(zone_sectors))
3751 return -EINVAL;
3752 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3753 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3754 __ilog2_u32(sbi->blocks_per_blkz))
3755 return -EINVAL;
3756 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3757 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3758 sbi->log_blocks_per_blkz;
3759 if (nr_sectors & (zone_sectors - 1))
3760 FDEV(devi).nr_blkz++;
3761
3762 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3763 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3764 * sizeof(unsigned long),
3765 GFP_KERNEL);
3766 if (!FDEV(devi).blkz_seq)
3767 return -ENOMEM;
3768
3769 rep_zone_arg.sbi = sbi;
3770 rep_zone_arg.dev = &FDEV(devi);
3771
3772 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3773 &rep_zone_arg);
3774 if (ret < 0)
3775 return ret;
3776 return 0;
3777}
3778#endif
3779
3780/*
3781 * Read f2fs raw super block.
3782 * Because we have two copies of super block, so read both of them
3783 * to get the first valid one. If any one of them is broken, we pass
3784 * them recovery flag back to the caller.
3785 */
3786static int read_raw_super_block(struct f2fs_sb_info *sbi,
3787 struct f2fs_super_block **raw_super,
3788 int *valid_super_block, int *recovery)
3789{
3790 struct super_block *sb = sbi->sb;
3791 int block;
3792 struct buffer_head *bh;
3793 struct f2fs_super_block *super;
3794 int err = 0;
3795
3796 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3797 if (!super)
3798 return -ENOMEM;
3799
3800 for (block = 0; block < 2; block++) {
3801 bh = sb_bread(sb, block);
3802 if (!bh) {
3803 f2fs_err(sbi, "Unable to read %dth superblock",
3804 block + 1);
3805 err = -EIO;
3806 *recovery = 1;
3807 continue;
3808 }
3809
3810 /* sanity checking of raw super */
3811 err = sanity_check_raw_super(sbi, bh);
3812 if (err) {
3813 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3814 block + 1);
3815 brelse(bh);
3816 *recovery = 1;
3817 continue;
3818 }
3819
3820 if (!*raw_super) {
3821 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3822 sizeof(*super));
3823 *valid_super_block = block;
3824 *raw_super = super;
3825 }
3826 brelse(bh);
3827 }
3828
3829 /* No valid superblock */
3830 if (!*raw_super)
3831 kfree(super);
3832 else
3833 err = 0;
3834
3835 return err;
3836}
3837
3838int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3839{
3840 struct buffer_head *bh;
3841 __u32 crc = 0;
3842 int err;
3843
3844 if ((recover && f2fs_readonly(sbi->sb)) ||
3845 bdev_read_only(sbi->sb->s_bdev)) {
3846 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3847 return -EROFS;
3848 }
3849
3850 /* we should update superblock crc here */
3851 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3852 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3853 offsetof(struct f2fs_super_block, crc));
3854 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3855 }
3856
3857 /* write back-up superblock first */
3858 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3859 if (!bh)
3860 return -EIO;
3861 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3862 brelse(bh);
3863
3864 /* if we are in recovery path, skip writing valid superblock */
3865 if (recover || err)
3866 return err;
3867
3868 /* write current valid superblock */
3869 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3870 if (!bh)
3871 return -EIO;
3872 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3873 brelse(bh);
3874 return err;
3875}
3876
3877void f2fs_handle_stop(struct f2fs_sb_info *sbi, unsigned char reason)
3878{
3879 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3880 int err;
3881
3882 f2fs_down_write(&sbi->sb_lock);
3883
3884 if (raw_super->s_stop_reason[reason] < ((1 << BITS_PER_BYTE) - 1))
3885 raw_super->s_stop_reason[reason]++;
3886
3887 err = f2fs_commit_super(sbi, false);
3888 if (err)
3889 f2fs_err(sbi, "f2fs_commit_super fails to record reason:%u err:%d",
3890 reason, err);
3891 f2fs_up_write(&sbi->sb_lock);
3892}
3893
3894static void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
3895{
3896 spin_lock(&sbi->error_lock);
3897 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
3898 set_bit(flag, (unsigned long *)sbi->errors);
3899 sbi->error_dirty = true;
3900 }
3901 spin_unlock(&sbi->error_lock);
3902}
3903
3904static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
3905{
3906 bool need_update = false;
3907
3908 spin_lock(&sbi->error_lock);
3909 if (sbi->error_dirty) {
3910 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
3911 MAX_F2FS_ERRORS);
3912 sbi->error_dirty = false;
3913 need_update = true;
3914 }
3915 spin_unlock(&sbi->error_lock);
3916
3917 return need_update;
3918}
3919
3920void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
3921{
3922 int err;
3923
3924 f2fs_save_errors(sbi, error);
3925
3926 f2fs_down_write(&sbi->sb_lock);
3927
3928 if (!f2fs_update_errors(sbi))
3929 goto out_unlock;
3930
3931 err = f2fs_commit_super(sbi, false);
3932 if (err)
3933 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
3934 error, err);
3935out_unlock:
3936 f2fs_up_write(&sbi->sb_lock);
3937}
3938
3939static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3940{
3941 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3942 unsigned int max_devices = MAX_DEVICES;
3943 unsigned int logical_blksize;
3944 int i;
3945
3946 /* Initialize single device information */
3947 if (!RDEV(0).path[0]) {
3948 if (!bdev_is_zoned(sbi->sb->s_bdev))
3949 return 0;
3950 max_devices = 1;
3951 }
3952
3953 /*
3954 * Initialize multiple devices information, or single
3955 * zoned block device information.
3956 */
3957 sbi->devs = f2fs_kzalloc(sbi,
3958 array_size(max_devices,
3959 sizeof(struct f2fs_dev_info)),
3960 GFP_KERNEL);
3961 if (!sbi->devs)
3962 return -ENOMEM;
3963
3964 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3965 sbi->aligned_blksize = true;
3966
3967 for (i = 0; i < max_devices; i++) {
3968
3969 if (i > 0 && !RDEV(i).path[0])
3970 break;
3971
3972 if (max_devices == 1) {
3973 /* Single zoned block device mount */
3974 FDEV(0).bdev =
3975 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3976 sbi->sb->s_mode, sbi->sb->s_type);
3977 } else {
3978 /* Multi-device mount */
3979 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3980 FDEV(i).total_segments =
3981 le32_to_cpu(RDEV(i).total_segments);
3982 if (i == 0) {
3983 FDEV(i).start_blk = 0;
3984 FDEV(i).end_blk = FDEV(i).start_blk +
3985 (FDEV(i).total_segments <<
3986 sbi->log_blocks_per_seg) - 1 +
3987 le32_to_cpu(raw_super->segment0_blkaddr);
3988 } else {
3989 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3990 FDEV(i).end_blk = FDEV(i).start_blk +
3991 (FDEV(i).total_segments <<
3992 sbi->log_blocks_per_seg) - 1;
3993 }
3994 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3995 sbi->sb->s_mode, sbi->sb->s_type);
3996 }
3997 if (IS_ERR(FDEV(i).bdev))
3998 return PTR_ERR(FDEV(i).bdev);
3999
4000 /* to release errored devices */
4001 sbi->s_ndevs = i + 1;
4002
4003 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4004 sbi->aligned_blksize = false;
4005
4006#ifdef CONFIG_BLK_DEV_ZONED
4007 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
4008 !f2fs_sb_has_blkzoned(sbi)) {
4009 f2fs_err(sbi, "Zoned block device feature not enabled");
4010 return -EINVAL;
4011 }
4012 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
4013 if (init_blkz_info(sbi, i)) {
4014 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4015 return -EINVAL;
4016 }
4017 if (max_devices == 1)
4018 break;
4019 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
4020 i, FDEV(i).path,
4021 FDEV(i).total_segments,
4022 FDEV(i).start_blk, FDEV(i).end_blk,
4023 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
4024 "Host-aware" : "Host-managed");
4025 continue;
4026 }
4027#endif
4028 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4029 i, FDEV(i).path,
4030 FDEV(i).total_segments,
4031 FDEV(i).start_blk, FDEV(i).end_blk);
4032 }
4033 f2fs_info(sbi,
4034 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
4035 return 0;
4036}
4037
4038static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4039{
4040#if IS_ENABLED(CONFIG_UNICODE)
4041 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4042 const struct f2fs_sb_encodings *encoding_info;
4043 struct unicode_map *encoding;
4044 __u16 encoding_flags;
4045
4046 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4047 if (!encoding_info) {
4048 f2fs_err(sbi,
4049 "Encoding requested by superblock is unknown");
4050 return -EINVAL;
4051 }
4052
4053 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4054 encoding = utf8_load(encoding_info->version);
4055 if (IS_ERR(encoding)) {
4056 f2fs_err(sbi,
4057 "can't mount with superblock charset: %s-%u.%u.%u "
4058 "not supported by the kernel. flags: 0x%x.",
4059 encoding_info->name,
4060 unicode_major(encoding_info->version),
4061 unicode_minor(encoding_info->version),
4062 unicode_rev(encoding_info->version),
4063 encoding_flags);
4064 return PTR_ERR(encoding);
4065 }
4066 f2fs_info(sbi, "Using encoding defined by superblock: "
4067 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4068 unicode_major(encoding_info->version),
4069 unicode_minor(encoding_info->version),
4070 unicode_rev(encoding_info->version),
4071 encoding_flags);
4072
4073 sbi->sb->s_encoding = encoding;
4074 sbi->sb->s_encoding_flags = encoding_flags;
4075 }
4076#else
4077 if (f2fs_sb_has_casefold(sbi)) {
4078 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4079 return -EINVAL;
4080 }
4081#endif
4082 return 0;
4083}
4084
4085static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4086{
4087 /* adjust parameters according to the volume size */
4088 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4089 if (f2fs_block_unit_discard(sbi))
4090 SM_I(sbi)->dcc_info->discard_granularity =
4091 MIN_DISCARD_GRANULARITY;
4092 SM_I(sbi)->ipu_policy = 1 << F2FS_IPU_FORCE |
4093 1 << F2FS_IPU_HONOR_OPU_WRITE;
4094 }
4095
4096 sbi->readdir_ra = true;
4097}
4098
4099static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4100{
4101 struct f2fs_sb_info *sbi;
4102 struct f2fs_super_block *raw_super;
4103 struct inode *root;
4104 int err;
4105 bool skip_recovery = false, need_fsck = false;
4106 char *options = NULL;
4107 int recovery, i, valid_super_block;
4108 struct curseg_info *seg_i;
4109 int retry_cnt = 1;
4110
4111try_onemore:
4112 err = -EINVAL;
4113 raw_super = NULL;
4114 valid_super_block = -1;
4115 recovery = 0;
4116
4117 /* allocate memory for f2fs-specific super block info */
4118 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4119 if (!sbi)
4120 return -ENOMEM;
4121
4122 sbi->sb = sb;
4123
4124 /* initialize locks within allocated memory */
4125 init_f2fs_rwsem(&sbi->gc_lock);
4126 mutex_init(&sbi->writepages);
4127 init_f2fs_rwsem(&sbi->cp_global_sem);
4128 init_f2fs_rwsem(&sbi->node_write);
4129 init_f2fs_rwsem(&sbi->node_change);
4130 spin_lock_init(&sbi->stat_lock);
4131 init_f2fs_rwsem(&sbi->cp_rwsem);
4132 init_f2fs_rwsem(&sbi->quota_sem);
4133 init_waitqueue_head(&sbi->cp_wait);
4134 spin_lock_init(&sbi->error_lock);
4135
4136 for (i = 0; i < NR_INODE_TYPE; i++) {
4137 INIT_LIST_HEAD(&sbi->inode_list[i]);
4138 spin_lock_init(&sbi->inode_lock[i]);
4139 }
4140 mutex_init(&sbi->flush_lock);
4141
4142 /* Load the checksum driver */
4143 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4144 if (IS_ERR(sbi->s_chksum_driver)) {
4145 f2fs_err(sbi, "Cannot load crc32 driver.");
4146 err = PTR_ERR(sbi->s_chksum_driver);
4147 sbi->s_chksum_driver = NULL;
4148 goto free_sbi;
4149 }
4150
4151 /* set a block size */
4152 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4153 f2fs_err(sbi, "unable to set blocksize");
4154 goto free_sbi;
4155 }
4156
4157 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4158 &recovery);
4159 if (err)
4160 goto free_sbi;
4161
4162 sb->s_fs_info = sbi;
4163 sbi->raw_super = raw_super;
4164
4165 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4166
4167 /* precompute checksum seed for metadata */
4168 if (f2fs_sb_has_inode_chksum(sbi))
4169 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4170 sizeof(raw_super->uuid));
4171
4172 default_options(sbi);
4173 /* parse mount options */
4174 options = kstrdup((const char *)data, GFP_KERNEL);
4175 if (data && !options) {
4176 err = -ENOMEM;
4177 goto free_sb_buf;
4178 }
4179
4180 err = parse_options(sb, options, false);
4181 if (err)
4182 goto free_options;
4183
4184 sb->s_maxbytes = max_file_blocks(NULL) <<
4185 le32_to_cpu(raw_super->log_blocksize);
4186 sb->s_max_links = F2FS_LINK_MAX;
4187
4188 err = f2fs_setup_casefold(sbi);
4189 if (err)
4190 goto free_options;
4191
4192#ifdef CONFIG_QUOTA
4193 sb->dq_op = &f2fs_quota_operations;
4194 sb->s_qcop = &f2fs_quotactl_ops;
4195 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4196
4197 if (f2fs_sb_has_quota_ino(sbi)) {
4198 for (i = 0; i < MAXQUOTAS; i++) {
4199 if (f2fs_qf_ino(sbi->sb, i))
4200 sbi->nquota_files++;
4201 }
4202 }
4203#endif
4204
4205 sb->s_op = &f2fs_sops;
4206#ifdef CONFIG_FS_ENCRYPTION
4207 sb->s_cop = &f2fs_cryptops;
4208#endif
4209#ifdef CONFIG_FS_VERITY
4210 sb->s_vop = &f2fs_verityops;
4211#endif
4212 sb->s_xattr = f2fs_xattr_handlers;
4213 sb->s_export_op = &f2fs_export_ops;
4214 sb->s_magic = F2FS_SUPER_MAGIC;
4215 sb->s_time_gran = 1;
4216 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4217 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4218 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4219 sb->s_iflags |= SB_I_CGROUPWB;
4220
4221 /* init f2fs-specific super block info */
4222 sbi->valid_super_block = valid_super_block;
4223
4224 /* disallow all the data/node/meta page writes */
4225 set_sbi_flag(sbi, SBI_POR_DOING);
4226
4227 err = f2fs_init_write_merge_io(sbi);
4228 if (err)
4229 goto free_bio_info;
4230
4231 init_sb_info(sbi);
4232
4233 err = f2fs_init_iostat(sbi);
4234 if (err)
4235 goto free_bio_info;
4236
4237 err = init_percpu_info(sbi);
4238 if (err)
4239 goto free_iostat;
4240
4241 if (F2FS_IO_ALIGNED(sbi)) {
4242 sbi->write_io_dummy =
4243 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4244 if (!sbi->write_io_dummy) {
4245 err = -ENOMEM;
4246 goto free_percpu;
4247 }
4248 }
4249
4250 /* init per sbi slab cache */
4251 err = f2fs_init_xattr_caches(sbi);
4252 if (err)
4253 goto free_io_dummy;
4254 err = f2fs_init_page_array_cache(sbi);
4255 if (err)
4256 goto free_xattr_cache;
4257
4258 /* get an inode for meta space */
4259 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4260 if (IS_ERR(sbi->meta_inode)) {
4261 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4262 err = PTR_ERR(sbi->meta_inode);
4263 goto free_page_array_cache;
4264 }
4265
4266 err = f2fs_get_valid_checkpoint(sbi);
4267 if (err) {
4268 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4269 goto free_meta_inode;
4270 }
4271
4272 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4273 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4274 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4275 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4276 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4277 }
4278
4279 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4280 set_sbi_flag(sbi, SBI_NEED_FSCK);
4281
4282 /* Initialize device list */
4283 err = f2fs_scan_devices(sbi);
4284 if (err) {
4285 f2fs_err(sbi, "Failed to find devices");
4286 goto free_devices;
4287 }
4288
4289 err = f2fs_init_post_read_wq(sbi);
4290 if (err) {
4291 f2fs_err(sbi, "Failed to initialize post read workqueue");
4292 goto free_devices;
4293 }
4294
4295 sbi->total_valid_node_count =
4296 le32_to_cpu(sbi->ckpt->valid_node_count);
4297 percpu_counter_set(&sbi->total_valid_inode_count,
4298 le32_to_cpu(sbi->ckpt->valid_inode_count));
4299 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4300 sbi->total_valid_block_count =
4301 le64_to_cpu(sbi->ckpt->valid_block_count);
4302 sbi->last_valid_block_count = sbi->total_valid_block_count;
4303 sbi->reserved_blocks = 0;
4304 sbi->current_reserved_blocks = 0;
4305 limit_reserve_root(sbi);
4306 adjust_unusable_cap_perc(sbi);
4307
4308 f2fs_init_extent_cache_info(sbi);
4309
4310 f2fs_init_ino_entry_info(sbi);
4311
4312 f2fs_init_fsync_node_info(sbi);
4313
4314 /* setup checkpoint request control and start checkpoint issue thread */
4315 f2fs_init_ckpt_req_control(sbi);
4316 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4317 test_opt(sbi, MERGE_CHECKPOINT)) {
4318 err = f2fs_start_ckpt_thread(sbi);
4319 if (err) {
4320 f2fs_err(sbi,
4321 "Failed to start F2FS issue_checkpoint_thread (%d)",
4322 err);
4323 goto stop_ckpt_thread;
4324 }
4325 }
4326
4327 /* setup f2fs internal modules */
4328 err = f2fs_build_segment_manager(sbi);
4329 if (err) {
4330 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4331 err);
4332 goto free_sm;
4333 }
4334 err = f2fs_build_node_manager(sbi);
4335 if (err) {
4336 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4337 err);
4338 goto free_nm;
4339 }
4340
4341 err = adjust_reserved_segment(sbi);
4342 if (err)
4343 goto free_nm;
4344
4345 /* For write statistics */
4346 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4347
4348 /* Read accumulated write IO statistics if exists */
4349 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4350 if (__exist_node_summaries(sbi))
4351 sbi->kbytes_written =
4352 le64_to_cpu(seg_i->journal->info.kbytes_written);
4353
4354 f2fs_build_gc_manager(sbi);
4355
4356 err = f2fs_build_stats(sbi);
4357 if (err)
4358 goto free_nm;
4359
4360 /* get an inode for node space */
4361 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4362 if (IS_ERR(sbi->node_inode)) {
4363 f2fs_err(sbi, "Failed to read node inode");
4364 err = PTR_ERR(sbi->node_inode);
4365 goto free_stats;
4366 }
4367
4368 /* read root inode and dentry */
4369 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4370 if (IS_ERR(root)) {
4371 f2fs_err(sbi, "Failed to read root inode");
4372 err = PTR_ERR(root);
4373 goto free_node_inode;
4374 }
4375 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4376 !root->i_size || !root->i_nlink) {
4377 iput(root);
4378 err = -EINVAL;
4379 goto free_node_inode;
4380 }
4381
4382 sb->s_root = d_make_root(root); /* allocate root dentry */
4383 if (!sb->s_root) {
4384 err = -ENOMEM;
4385 goto free_node_inode;
4386 }
4387
4388 err = f2fs_init_compress_inode(sbi);
4389 if (err)
4390 goto free_root_inode;
4391
4392 err = f2fs_register_sysfs(sbi);
4393 if (err)
4394 goto free_compress_inode;
4395
4396#ifdef CONFIG_QUOTA
4397 /* Enable quota usage during mount */
4398 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4399 err = f2fs_enable_quotas(sb);
4400 if (err)
4401 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4402 }
4403#endif
4404 /* if there are any orphan inodes, free them */
4405 err = f2fs_recover_orphan_inodes(sbi);
4406 if (err)
4407 goto free_meta;
4408
4409 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4410 goto reset_checkpoint;
4411
4412 /* recover fsynced data */
4413 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4414 !test_opt(sbi, NORECOVERY)) {
4415 /*
4416 * mount should be failed, when device has readonly mode, and
4417 * previous checkpoint was not done by clean system shutdown.
4418 */
4419 if (f2fs_hw_is_readonly(sbi)) {
4420 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4421 err = f2fs_recover_fsync_data(sbi, true);
4422 if (err > 0) {
4423 err = -EROFS;
4424 f2fs_err(sbi, "Need to recover fsync data, but "
4425 "write access unavailable, please try "
4426 "mount w/ disable_roll_forward or norecovery");
4427 }
4428 if (err < 0)
4429 goto free_meta;
4430 }
4431 f2fs_info(sbi, "write access unavailable, skipping recovery");
4432 goto reset_checkpoint;
4433 }
4434
4435 if (need_fsck)
4436 set_sbi_flag(sbi, SBI_NEED_FSCK);
4437
4438 if (skip_recovery)
4439 goto reset_checkpoint;
4440
4441 err = f2fs_recover_fsync_data(sbi, false);
4442 if (err < 0) {
4443 if (err != -ENOMEM)
4444 skip_recovery = true;
4445 need_fsck = true;
4446 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4447 err);
4448 goto free_meta;
4449 }
4450 } else {
4451 err = f2fs_recover_fsync_data(sbi, true);
4452
4453 if (!f2fs_readonly(sb) && err > 0) {
4454 err = -EINVAL;
4455 f2fs_err(sbi, "Need to recover fsync data");
4456 goto free_meta;
4457 }
4458 }
4459
4460 /*
4461 * If the f2fs is not readonly and fsync data recovery succeeds,
4462 * check zoned block devices' write pointer consistency.
4463 */
4464 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4465 err = f2fs_check_write_pointer(sbi);
4466 if (err)
4467 goto free_meta;
4468 }
4469
4470reset_checkpoint:
4471 f2fs_init_inmem_curseg(sbi);
4472
4473 /* f2fs_recover_fsync_data() cleared this already */
4474 clear_sbi_flag(sbi, SBI_POR_DOING);
4475
4476 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4477 err = f2fs_disable_checkpoint(sbi);
4478 if (err)
4479 goto sync_free_meta;
4480 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4481 f2fs_enable_checkpoint(sbi);
4482 }
4483
4484 /*
4485 * If filesystem is not mounted as read-only then
4486 * do start the gc_thread.
4487 */
4488 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4489 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4490 /* After POR, we can run background GC thread.*/
4491 err = f2fs_start_gc_thread(sbi);
4492 if (err)
4493 goto sync_free_meta;
4494 }
4495 kvfree(options);
4496
4497 /* recover broken superblock */
4498 if (recovery) {
4499 err = f2fs_commit_super(sbi, true);
4500 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4501 sbi->valid_super_block ? 1 : 2, err);
4502 }
4503
4504 f2fs_join_shrinker(sbi);
4505
4506 f2fs_tuning_parameters(sbi);
4507
4508 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4509 cur_cp_version(F2FS_CKPT(sbi)));
4510 f2fs_update_time(sbi, CP_TIME);
4511 f2fs_update_time(sbi, REQ_TIME);
4512 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4513 return 0;
4514
4515sync_free_meta:
4516 /* safe to flush all the data */
4517 sync_filesystem(sbi->sb);
4518 retry_cnt = 0;
4519
4520free_meta:
4521#ifdef CONFIG_QUOTA
4522 f2fs_truncate_quota_inode_pages(sb);
4523 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4524 f2fs_quota_off_umount(sbi->sb);
4525#endif
4526 /*
4527 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4528 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4529 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4530 * falls into an infinite loop in f2fs_sync_meta_pages().
4531 */
4532 truncate_inode_pages_final(META_MAPPING(sbi));
4533 /* evict some inodes being cached by GC */
4534 evict_inodes(sb);
4535 f2fs_unregister_sysfs(sbi);
4536free_compress_inode:
4537 f2fs_destroy_compress_inode(sbi);
4538free_root_inode:
4539 dput(sb->s_root);
4540 sb->s_root = NULL;
4541free_node_inode:
4542 f2fs_release_ino_entry(sbi, true);
4543 truncate_inode_pages_final(NODE_MAPPING(sbi));
4544 iput(sbi->node_inode);
4545 sbi->node_inode = NULL;
4546free_stats:
4547 f2fs_destroy_stats(sbi);
4548free_nm:
4549 /* stop discard thread before destroying node manager */
4550 f2fs_stop_discard_thread(sbi);
4551 f2fs_destroy_node_manager(sbi);
4552free_sm:
4553 f2fs_destroy_segment_manager(sbi);
4554stop_ckpt_thread:
4555 f2fs_stop_ckpt_thread(sbi);
4556 f2fs_destroy_post_read_wq(sbi);
4557free_devices:
4558 destroy_device_list(sbi);
4559 kvfree(sbi->ckpt);
4560free_meta_inode:
4561 make_bad_inode(sbi->meta_inode);
4562 iput(sbi->meta_inode);
4563 sbi->meta_inode = NULL;
4564free_page_array_cache:
4565 f2fs_destroy_page_array_cache(sbi);
4566free_xattr_cache:
4567 f2fs_destroy_xattr_caches(sbi);
4568free_io_dummy:
4569 mempool_destroy(sbi->write_io_dummy);
4570free_percpu:
4571 destroy_percpu_info(sbi);
4572free_iostat:
4573 f2fs_destroy_iostat(sbi);
4574free_bio_info:
4575 for (i = 0; i < NR_PAGE_TYPE; i++)
4576 kvfree(sbi->write_io[i]);
4577
4578#if IS_ENABLED(CONFIG_UNICODE)
4579 utf8_unload(sb->s_encoding);
4580 sb->s_encoding = NULL;
4581#endif
4582free_options:
4583#ifdef CONFIG_QUOTA
4584 for (i = 0; i < MAXQUOTAS; i++)
4585 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4586#endif
4587 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4588 kvfree(options);
4589free_sb_buf:
4590 kfree(raw_super);
4591free_sbi:
4592 if (sbi->s_chksum_driver)
4593 crypto_free_shash(sbi->s_chksum_driver);
4594 kfree(sbi);
4595
4596 /* give only one another chance */
4597 if (retry_cnt > 0 && skip_recovery) {
4598 retry_cnt--;
4599 shrink_dcache_sb(sb);
4600 goto try_onemore;
4601 }
4602 return err;
4603}
4604
4605static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4606 const char *dev_name, void *data)
4607{
4608 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4609}
4610
4611static void kill_f2fs_super(struct super_block *sb)
4612{
4613 if (sb->s_root) {
4614 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4615
4616 set_sbi_flag(sbi, SBI_IS_CLOSE);
4617 f2fs_stop_gc_thread(sbi);
4618 f2fs_stop_discard_thread(sbi);
4619
4620#ifdef CONFIG_F2FS_FS_COMPRESSION
4621 /*
4622 * latter evict_inode() can bypass checking and invalidating
4623 * compress inode cache.
4624 */
4625 if (test_opt(sbi, COMPRESS_CACHE))
4626 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4627#endif
4628
4629 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4630 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4631 struct cp_control cpc = {
4632 .reason = CP_UMOUNT,
4633 };
4634 f2fs_write_checkpoint(sbi, &cpc);
4635 }
4636
4637 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4638 sb->s_flags &= ~SB_RDONLY;
4639 }
4640 kill_block_super(sb);
4641}
4642
4643static struct file_system_type f2fs_fs_type = {
4644 .owner = THIS_MODULE,
4645 .name = "f2fs",
4646 .mount = f2fs_mount,
4647 .kill_sb = kill_f2fs_super,
4648 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4649};
4650MODULE_ALIAS_FS("f2fs");
4651
4652static int __init init_inodecache(void)
4653{
4654 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4655 sizeof(struct f2fs_inode_info), 0,
4656 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4657 return f2fs_inode_cachep ? 0 : -ENOMEM;
4658}
4659
4660static void destroy_inodecache(void)
4661{
4662 /*
4663 * Make sure all delayed rcu free inodes are flushed before we
4664 * destroy cache.
4665 */
4666 rcu_barrier();
4667 kmem_cache_destroy(f2fs_inode_cachep);
4668}
4669
4670static int __init init_f2fs_fs(void)
4671{
4672 int err;
4673
4674 if (PAGE_SIZE != F2FS_BLKSIZE) {
4675 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4676 PAGE_SIZE, F2FS_BLKSIZE);
4677 return -EINVAL;
4678 }
4679
4680 err = init_inodecache();
4681 if (err)
4682 goto fail;
4683 err = f2fs_create_node_manager_caches();
4684 if (err)
4685 goto free_inodecache;
4686 err = f2fs_create_segment_manager_caches();
4687 if (err)
4688 goto free_node_manager_caches;
4689 err = f2fs_create_checkpoint_caches();
4690 if (err)
4691 goto free_segment_manager_caches;
4692 err = f2fs_create_recovery_cache();
4693 if (err)
4694 goto free_checkpoint_caches;
4695 err = f2fs_create_extent_cache();
4696 if (err)
4697 goto free_recovery_cache;
4698 err = f2fs_create_garbage_collection_cache();
4699 if (err)
4700 goto free_extent_cache;
4701 err = f2fs_init_sysfs();
4702 if (err)
4703 goto free_garbage_collection_cache;
4704 err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4705 if (err)
4706 goto free_sysfs;
4707 err = register_filesystem(&f2fs_fs_type);
4708 if (err)
4709 goto free_shrinker;
4710 f2fs_create_root_stats();
4711 err = f2fs_init_post_read_processing();
4712 if (err)
4713 goto free_root_stats;
4714 err = f2fs_init_iostat_processing();
4715 if (err)
4716 goto free_post_read;
4717 err = f2fs_init_bio_entry_cache();
4718 if (err)
4719 goto free_iostat;
4720 err = f2fs_init_bioset();
4721 if (err)
4722 goto free_bio_entry_cache;
4723 err = f2fs_init_compress_mempool();
4724 if (err)
4725 goto free_bioset;
4726 err = f2fs_init_compress_cache();
4727 if (err)
4728 goto free_compress_mempool;
4729 err = f2fs_create_casefold_cache();
4730 if (err)
4731 goto free_compress_cache;
4732 return 0;
4733free_compress_cache:
4734 f2fs_destroy_compress_cache();
4735free_compress_mempool:
4736 f2fs_destroy_compress_mempool();
4737free_bioset:
4738 f2fs_destroy_bioset();
4739free_bio_entry_cache:
4740 f2fs_destroy_bio_entry_cache();
4741free_iostat:
4742 f2fs_destroy_iostat_processing();
4743free_post_read:
4744 f2fs_destroy_post_read_processing();
4745free_root_stats:
4746 f2fs_destroy_root_stats();
4747 unregister_filesystem(&f2fs_fs_type);
4748free_shrinker:
4749 unregister_shrinker(&f2fs_shrinker_info);
4750free_sysfs:
4751 f2fs_exit_sysfs();
4752free_garbage_collection_cache:
4753 f2fs_destroy_garbage_collection_cache();
4754free_extent_cache:
4755 f2fs_destroy_extent_cache();
4756free_recovery_cache:
4757 f2fs_destroy_recovery_cache();
4758free_checkpoint_caches:
4759 f2fs_destroy_checkpoint_caches();
4760free_segment_manager_caches:
4761 f2fs_destroy_segment_manager_caches();
4762free_node_manager_caches:
4763 f2fs_destroy_node_manager_caches();
4764free_inodecache:
4765 destroy_inodecache();
4766fail:
4767 return err;
4768}
4769
4770static void __exit exit_f2fs_fs(void)
4771{
4772 f2fs_destroy_casefold_cache();
4773 f2fs_destroy_compress_cache();
4774 f2fs_destroy_compress_mempool();
4775 f2fs_destroy_bioset();
4776 f2fs_destroy_bio_entry_cache();
4777 f2fs_destroy_iostat_processing();
4778 f2fs_destroy_post_read_processing();
4779 f2fs_destroy_root_stats();
4780 unregister_filesystem(&f2fs_fs_type);
4781 unregister_shrinker(&f2fs_shrinker_info);
4782 f2fs_exit_sysfs();
4783 f2fs_destroy_garbage_collection_cache();
4784 f2fs_destroy_extent_cache();
4785 f2fs_destroy_recovery_cache();
4786 f2fs_destroy_checkpoint_caches();
4787 f2fs_destroy_segment_manager_caches();
4788 f2fs_destroy_node_manager_caches();
4789 destroy_inodecache();
4790}
4791
4792module_init(init_f2fs_fs)
4793module_exit(exit_f2fs_fs)
4794
4795MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4796MODULE_DESCRIPTION("Flash Friendly File System");
4797MODULE_LICENSE("GPL");
4798MODULE_SOFTDEP("pre: crc32");
4799
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * fs/f2fs/super.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8#include <linux/module.h>
9#include <linux/init.h>
10#include <linux/fs.h>
11#include <linux/statfs.h>
12#include <linux/buffer_head.h>
13#include <linux/backing-dev.h>
14#include <linux/kthread.h>
15#include <linux/parser.h>
16#include <linux/mount.h>
17#include <linux/seq_file.h>
18#include <linux/proc_fs.h>
19#include <linux/random.h>
20#include <linux/exportfs.h>
21#include <linux/blkdev.h>
22#include <linux/quotaops.h>
23#include <linux/f2fs_fs.h>
24#include <linux/sysfs.h>
25#include <linux/quota.h>
26#include <linux/unicode.h>
27#include <linux/part_stat.h>
28
29#include "f2fs.h"
30#include "node.h"
31#include "segment.h"
32#include "xattr.h"
33#include "gc.h"
34#include "trace.h"
35
36#define CREATE_TRACE_POINTS
37#include <trace/events/f2fs.h>
38
39static struct kmem_cache *f2fs_inode_cachep;
40
41#ifdef CONFIG_F2FS_FAULT_INJECTION
42
43const char *f2fs_fault_name[FAULT_MAX] = {
44 [FAULT_KMALLOC] = "kmalloc",
45 [FAULT_KVMALLOC] = "kvmalloc",
46 [FAULT_PAGE_ALLOC] = "page alloc",
47 [FAULT_PAGE_GET] = "page get",
48 [FAULT_ALLOC_BIO] = "alloc bio",
49 [FAULT_ALLOC_NID] = "alloc nid",
50 [FAULT_ORPHAN] = "orphan",
51 [FAULT_BLOCK] = "no more block",
52 [FAULT_DIR_DEPTH] = "too big dir depth",
53 [FAULT_EVICT_INODE] = "evict_inode fail",
54 [FAULT_TRUNCATE] = "truncate fail",
55 [FAULT_READ_IO] = "read IO error",
56 [FAULT_CHECKPOINT] = "checkpoint error",
57 [FAULT_DISCARD] = "discard error",
58 [FAULT_WRITE_IO] = "write IO error",
59};
60
61void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
62 unsigned int type)
63{
64 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
65
66 if (rate) {
67 atomic_set(&ffi->inject_ops, 0);
68 ffi->inject_rate = rate;
69 }
70
71 if (type)
72 ffi->inject_type = type;
73
74 if (!rate && !type)
75 memset(ffi, 0, sizeof(struct f2fs_fault_info));
76}
77#endif
78
79/* f2fs-wide shrinker description */
80static struct shrinker f2fs_shrinker_info = {
81 .scan_objects = f2fs_shrink_scan,
82 .count_objects = f2fs_shrink_count,
83 .seeks = DEFAULT_SEEKS,
84};
85
86enum {
87 Opt_gc_background,
88 Opt_disable_roll_forward,
89 Opt_norecovery,
90 Opt_discard,
91 Opt_nodiscard,
92 Opt_noheap,
93 Opt_heap,
94 Opt_user_xattr,
95 Opt_nouser_xattr,
96 Opt_acl,
97 Opt_noacl,
98 Opt_active_logs,
99 Opt_disable_ext_identify,
100 Opt_inline_xattr,
101 Opt_noinline_xattr,
102 Opt_inline_xattr_size,
103 Opt_inline_data,
104 Opt_inline_dentry,
105 Opt_noinline_dentry,
106 Opt_flush_merge,
107 Opt_noflush_merge,
108 Opt_nobarrier,
109 Opt_fastboot,
110 Opt_extent_cache,
111 Opt_noextent_cache,
112 Opt_noinline_data,
113 Opt_data_flush,
114 Opt_reserve_root,
115 Opt_resgid,
116 Opt_resuid,
117 Opt_mode,
118 Opt_io_size_bits,
119 Opt_fault_injection,
120 Opt_fault_type,
121 Opt_lazytime,
122 Opt_nolazytime,
123 Opt_quota,
124 Opt_noquota,
125 Opt_usrquota,
126 Opt_grpquota,
127 Opt_prjquota,
128 Opt_usrjquota,
129 Opt_grpjquota,
130 Opt_prjjquota,
131 Opt_offusrjquota,
132 Opt_offgrpjquota,
133 Opt_offprjjquota,
134 Opt_jqfmt_vfsold,
135 Opt_jqfmt_vfsv0,
136 Opt_jqfmt_vfsv1,
137 Opt_whint,
138 Opt_alloc,
139 Opt_fsync,
140 Opt_test_dummy_encryption,
141 Opt_inlinecrypt,
142 Opt_checkpoint_disable,
143 Opt_checkpoint_disable_cap,
144 Opt_checkpoint_disable_cap_perc,
145 Opt_checkpoint_enable,
146 Opt_compress_algorithm,
147 Opt_compress_log_size,
148 Opt_compress_extension,
149 Opt_err,
150};
151
152static match_table_t f2fs_tokens = {
153 {Opt_gc_background, "background_gc=%s"},
154 {Opt_disable_roll_forward, "disable_roll_forward"},
155 {Opt_norecovery, "norecovery"},
156 {Opt_discard, "discard"},
157 {Opt_nodiscard, "nodiscard"},
158 {Opt_noheap, "no_heap"},
159 {Opt_heap, "heap"},
160 {Opt_user_xattr, "user_xattr"},
161 {Opt_nouser_xattr, "nouser_xattr"},
162 {Opt_acl, "acl"},
163 {Opt_noacl, "noacl"},
164 {Opt_active_logs, "active_logs=%u"},
165 {Opt_disable_ext_identify, "disable_ext_identify"},
166 {Opt_inline_xattr, "inline_xattr"},
167 {Opt_noinline_xattr, "noinline_xattr"},
168 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
169 {Opt_inline_data, "inline_data"},
170 {Opt_inline_dentry, "inline_dentry"},
171 {Opt_noinline_dentry, "noinline_dentry"},
172 {Opt_flush_merge, "flush_merge"},
173 {Opt_noflush_merge, "noflush_merge"},
174 {Opt_nobarrier, "nobarrier"},
175 {Opt_fastboot, "fastboot"},
176 {Opt_extent_cache, "extent_cache"},
177 {Opt_noextent_cache, "noextent_cache"},
178 {Opt_noinline_data, "noinline_data"},
179 {Opt_data_flush, "data_flush"},
180 {Opt_reserve_root, "reserve_root=%u"},
181 {Opt_resgid, "resgid=%u"},
182 {Opt_resuid, "resuid=%u"},
183 {Opt_mode, "mode=%s"},
184 {Opt_io_size_bits, "io_bits=%u"},
185 {Opt_fault_injection, "fault_injection=%u"},
186 {Opt_fault_type, "fault_type=%u"},
187 {Opt_lazytime, "lazytime"},
188 {Opt_nolazytime, "nolazytime"},
189 {Opt_quota, "quota"},
190 {Opt_noquota, "noquota"},
191 {Opt_usrquota, "usrquota"},
192 {Opt_grpquota, "grpquota"},
193 {Opt_prjquota, "prjquota"},
194 {Opt_usrjquota, "usrjquota=%s"},
195 {Opt_grpjquota, "grpjquota=%s"},
196 {Opt_prjjquota, "prjjquota=%s"},
197 {Opt_offusrjquota, "usrjquota="},
198 {Opt_offgrpjquota, "grpjquota="},
199 {Opt_offprjjquota, "prjjquota="},
200 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
201 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
202 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
203 {Opt_whint, "whint_mode=%s"},
204 {Opt_alloc, "alloc_mode=%s"},
205 {Opt_fsync, "fsync_mode=%s"},
206 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
207 {Opt_test_dummy_encryption, "test_dummy_encryption"},
208 {Opt_inlinecrypt, "inlinecrypt"},
209 {Opt_checkpoint_disable, "checkpoint=disable"},
210 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
211 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
212 {Opt_checkpoint_enable, "checkpoint=enable"},
213 {Opt_compress_algorithm, "compress_algorithm=%s"},
214 {Opt_compress_log_size, "compress_log_size=%u"},
215 {Opt_compress_extension, "compress_extension=%s"},
216 {Opt_err, NULL},
217};
218
219void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
220{
221 struct va_format vaf;
222 va_list args;
223 int level;
224
225 va_start(args, fmt);
226
227 level = printk_get_level(fmt);
228 vaf.fmt = printk_skip_level(fmt);
229 vaf.va = &args;
230 printk("%c%cF2FS-fs (%s): %pV\n",
231 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
232
233 va_end(args);
234}
235
236#ifdef CONFIG_UNICODE
237static const struct f2fs_sb_encodings {
238 __u16 magic;
239 char *name;
240 char *version;
241} f2fs_sb_encoding_map[] = {
242 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
243};
244
245static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
246 const struct f2fs_sb_encodings **encoding,
247 __u16 *flags)
248{
249 __u16 magic = le16_to_cpu(sb->s_encoding);
250 int i;
251
252 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
253 if (magic == f2fs_sb_encoding_map[i].magic)
254 break;
255
256 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
257 return -EINVAL;
258
259 *encoding = &f2fs_sb_encoding_map[i];
260 *flags = le16_to_cpu(sb->s_encoding_flags);
261
262 return 0;
263}
264#endif
265
266static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
267{
268 block_t limit = min((sbi->user_block_count << 1) / 1000,
269 sbi->user_block_count - sbi->reserved_blocks);
270
271 /* limit is 0.2% */
272 if (test_opt(sbi, RESERVE_ROOT) &&
273 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
274 F2FS_OPTION(sbi).root_reserved_blocks = limit;
275 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
276 F2FS_OPTION(sbi).root_reserved_blocks);
277 }
278 if (!test_opt(sbi, RESERVE_ROOT) &&
279 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
280 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
281 !gid_eq(F2FS_OPTION(sbi).s_resgid,
282 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
283 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
284 from_kuid_munged(&init_user_ns,
285 F2FS_OPTION(sbi).s_resuid),
286 from_kgid_munged(&init_user_ns,
287 F2FS_OPTION(sbi).s_resgid));
288}
289
290static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
291{
292 if (!F2FS_OPTION(sbi).unusable_cap_perc)
293 return;
294
295 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
296 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
297 else
298 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
299 F2FS_OPTION(sbi).unusable_cap_perc;
300
301 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
302 F2FS_OPTION(sbi).unusable_cap,
303 F2FS_OPTION(sbi).unusable_cap_perc);
304}
305
306static void init_once(void *foo)
307{
308 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
309
310 inode_init_once(&fi->vfs_inode);
311}
312
313#ifdef CONFIG_QUOTA
314static const char * const quotatypes[] = INITQFNAMES;
315#define QTYPE2NAME(t) (quotatypes[t])
316static int f2fs_set_qf_name(struct super_block *sb, int qtype,
317 substring_t *args)
318{
319 struct f2fs_sb_info *sbi = F2FS_SB(sb);
320 char *qname;
321 int ret = -EINVAL;
322
323 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
324 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
325 return -EINVAL;
326 }
327 if (f2fs_sb_has_quota_ino(sbi)) {
328 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
329 return 0;
330 }
331
332 qname = match_strdup(args);
333 if (!qname) {
334 f2fs_err(sbi, "Not enough memory for storing quotafile name");
335 return -ENOMEM;
336 }
337 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
338 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
339 ret = 0;
340 else
341 f2fs_err(sbi, "%s quota file already specified",
342 QTYPE2NAME(qtype));
343 goto errout;
344 }
345 if (strchr(qname, '/')) {
346 f2fs_err(sbi, "quotafile must be on filesystem root");
347 goto errout;
348 }
349 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
350 set_opt(sbi, QUOTA);
351 return 0;
352errout:
353 kfree(qname);
354 return ret;
355}
356
357static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
358{
359 struct f2fs_sb_info *sbi = F2FS_SB(sb);
360
361 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
362 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
363 return -EINVAL;
364 }
365 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
366 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
367 return 0;
368}
369
370static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
371{
372 /*
373 * We do the test below only for project quotas. 'usrquota' and
374 * 'grpquota' mount options are allowed even without quota feature
375 * to support legacy quotas in quota files.
376 */
377 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
378 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
379 return -1;
380 }
381 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
382 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
383 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
384 if (test_opt(sbi, USRQUOTA) &&
385 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
386 clear_opt(sbi, USRQUOTA);
387
388 if (test_opt(sbi, GRPQUOTA) &&
389 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
390 clear_opt(sbi, GRPQUOTA);
391
392 if (test_opt(sbi, PRJQUOTA) &&
393 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
394 clear_opt(sbi, PRJQUOTA);
395
396 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
397 test_opt(sbi, PRJQUOTA)) {
398 f2fs_err(sbi, "old and new quota format mixing");
399 return -1;
400 }
401
402 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
403 f2fs_err(sbi, "journaled quota format not specified");
404 return -1;
405 }
406 }
407
408 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
409 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
410 F2FS_OPTION(sbi).s_jquota_fmt = 0;
411 }
412 return 0;
413}
414#endif
415
416static int f2fs_set_test_dummy_encryption(struct super_block *sb,
417 const char *opt,
418 const substring_t *arg,
419 bool is_remount)
420{
421 struct f2fs_sb_info *sbi = F2FS_SB(sb);
422#ifdef CONFIG_FS_ENCRYPTION
423 int err;
424
425 if (!f2fs_sb_has_encrypt(sbi)) {
426 f2fs_err(sbi, "Encrypt feature is off");
427 return -EINVAL;
428 }
429
430 /*
431 * This mount option is just for testing, and it's not worthwhile to
432 * implement the extra complexity (e.g. RCU protection) that would be
433 * needed to allow it to be set or changed during remount. We do allow
434 * it to be specified during remount, but only if there is no change.
435 */
436 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_ctx.ctx) {
437 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
438 return -EINVAL;
439 }
440 err = fscrypt_set_test_dummy_encryption(
441 sb, arg, &F2FS_OPTION(sbi).dummy_enc_ctx);
442 if (err) {
443 if (err == -EEXIST)
444 f2fs_warn(sbi,
445 "Can't change test_dummy_encryption on remount");
446 else if (err == -EINVAL)
447 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
448 opt);
449 else
450 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
451 opt, err);
452 return -EINVAL;
453 }
454 f2fs_warn(sbi, "Test dummy encryption mode enabled");
455#else
456 f2fs_warn(sbi, "Test dummy encryption mount option ignored");
457#endif
458 return 0;
459}
460
461static int parse_options(struct super_block *sb, char *options, bool is_remount)
462{
463 struct f2fs_sb_info *sbi = F2FS_SB(sb);
464 substring_t args[MAX_OPT_ARGS];
465#ifdef CONFIG_F2FS_FS_COMPRESSION
466 unsigned char (*ext)[F2FS_EXTENSION_LEN];
467 int ext_cnt;
468#endif
469 char *p, *name;
470 int arg = 0;
471 kuid_t uid;
472 kgid_t gid;
473 int ret;
474
475 if (!options)
476 return 0;
477
478 while ((p = strsep(&options, ",")) != NULL) {
479 int token;
480 if (!*p)
481 continue;
482 /*
483 * Initialize args struct so we know whether arg was
484 * found; some options take optional arguments.
485 */
486 args[0].to = args[0].from = NULL;
487 token = match_token(p, f2fs_tokens, args);
488
489 switch (token) {
490 case Opt_gc_background:
491 name = match_strdup(&args[0]);
492
493 if (!name)
494 return -ENOMEM;
495 if (!strcmp(name, "on")) {
496 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
497 } else if (!strcmp(name, "off")) {
498 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
499 } else if (!strcmp(name, "sync")) {
500 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
501 } else {
502 kfree(name);
503 return -EINVAL;
504 }
505 kfree(name);
506 break;
507 case Opt_disable_roll_forward:
508 set_opt(sbi, DISABLE_ROLL_FORWARD);
509 break;
510 case Opt_norecovery:
511 /* this option mounts f2fs with ro */
512 set_opt(sbi, NORECOVERY);
513 if (!f2fs_readonly(sb))
514 return -EINVAL;
515 break;
516 case Opt_discard:
517 set_opt(sbi, DISCARD);
518 break;
519 case Opt_nodiscard:
520 if (f2fs_sb_has_blkzoned(sbi)) {
521 f2fs_warn(sbi, "discard is required for zoned block devices");
522 return -EINVAL;
523 }
524 clear_opt(sbi, DISCARD);
525 break;
526 case Opt_noheap:
527 set_opt(sbi, NOHEAP);
528 break;
529 case Opt_heap:
530 clear_opt(sbi, NOHEAP);
531 break;
532#ifdef CONFIG_F2FS_FS_XATTR
533 case Opt_user_xattr:
534 set_opt(sbi, XATTR_USER);
535 break;
536 case Opt_nouser_xattr:
537 clear_opt(sbi, XATTR_USER);
538 break;
539 case Opt_inline_xattr:
540 set_opt(sbi, INLINE_XATTR);
541 break;
542 case Opt_noinline_xattr:
543 clear_opt(sbi, INLINE_XATTR);
544 break;
545 case Opt_inline_xattr_size:
546 if (args->from && match_int(args, &arg))
547 return -EINVAL;
548 set_opt(sbi, INLINE_XATTR_SIZE);
549 F2FS_OPTION(sbi).inline_xattr_size = arg;
550 break;
551#else
552 case Opt_user_xattr:
553 f2fs_info(sbi, "user_xattr options not supported");
554 break;
555 case Opt_nouser_xattr:
556 f2fs_info(sbi, "nouser_xattr options not supported");
557 break;
558 case Opt_inline_xattr:
559 f2fs_info(sbi, "inline_xattr options not supported");
560 break;
561 case Opt_noinline_xattr:
562 f2fs_info(sbi, "noinline_xattr options not supported");
563 break;
564#endif
565#ifdef CONFIG_F2FS_FS_POSIX_ACL
566 case Opt_acl:
567 set_opt(sbi, POSIX_ACL);
568 break;
569 case Opt_noacl:
570 clear_opt(sbi, POSIX_ACL);
571 break;
572#else
573 case Opt_acl:
574 f2fs_info(sbi, "acl options not supported");
575 break;
576 case Opt_noacl:
577 f2fs_info(sbi, "noacl options not supported");
578 break;
579#endif
580 case Opt_active_logs:
581 if (args->from && match_int(args, &arg))
582 return -EINVAL;
583 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
584 return -EINVAL;
585 F2FS_OPTION(sbi).active_logs = arg;
586 break;
587 case Opt_disable_ext_identify:
588 set_opt(sbi, DISABLE_EXT_IDENTIFY);
589 break;
590 case Opt_inline_data:
591 set_opt(sbi, INLINE_DATA);
592 break;
593 case Opt_inline_dentry:
594 set_opt(sbi, INLINE_DENTRY);
595 break;
596 case Opt_noinline_dentry:
597 clear_opt(sbi, INLINE_DENTRY);
598 break;
599 case Opt_flush_merge:
600 set_opt(sbi, FLUSH_MERGE);
601 break;
602 case Opt_noflush_merge:
603 clear_opt(sbi, FLUSH_MERGE);
604 break;
605 case Opt_nobarrier:
606 set_opt(sbi, NOBARRIER);
607 break;
608 case Opt_fastboot:
609 set_opt(sbi, FASTBOOT);
610 break;
611 case Opt_extent_cache:
612 set_opt(sbi, EXTENT_CACHE);
613 break;
614 case Opt_noextent_cache:
615 clear_opt(sbi, EXTENT_CACHE);
616 break;
617 case Opt_noinline_data:
618 clear_opt(sbi, INLINE_DATA);
619 break;
620 case Opt_data_flush:
621 set_opt(sbi, DATA_FLUSH);
622 break;
623 case Opt_reserve_root:
624 if (args->from && match_int(args, &arg))
625 return -EINVAL;
626 if (test_opt(sbi, RESERVE_ROOT)) {
627 f2fs_info(sbi, "Preserve previous reserve_root=%u",
628 F2FS_OPTION(sbi).root_reserved_blocks);
629 } else {
630 F2FS_OPTION(sbi).root_reserved_blocks = arg;
631 set_opt(sbi, RESERVE_ROOT);
632 }
633 break;
634 case Opt_resuid:
635 if (args->from && match_int(args, &arg))
636 return -EINVAL;
637 uid = make_kuid(current_user_ns(), arg);
638 if (!uid_valid(uid)) {
639 f2fs_err(sbi, "Invalid uid value %d", arg);
640 return -EINVAL;
641 }
642 F2FS_OPTION(sbi).s_resuid = uid;
643 break;
644 case Opt_resgid:
645 if (args->from && match_int(args, &arg))
646 return -EINVAL;
647 gid = make_kgid(current_user_ns(), arg);
648 if (!gid_valid(gid)) {
649 f2fs_err(sbi, "Invalid gid value %d", arg);
650 return -EINVAL;
651 }
652 F2FS_OPTION(sbi).s_resgid = gid;
653 break;
654 case Opt_mode:
655 name = match_strdup(&args[0]);
656
657 if (!name)
658 return -ENOMEM;
659 if (!strcmp(name, "adaptive")) {
660 if (f2fs_sb_has_blkzoned(sbi)) {
661 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
662 kfree(name);
663 return -EINVAL;
664 }
665 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
666 } else if (!strcmp(name, "lfs")) {
667 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
668 } else {
669 kfree(name);
670 return -EINVAL;
671 }
672 kfree(name);
673 break;
674 case Opt_io_size_bits:
675 if (args->from && match_int(args, &arg))
676 return -EINVAL;
677 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
678 f2fs_warn(sbi, "Not support %d, larger than %d",
679 1 << arg, BIO_MAX_PAGES);
680 return -EINVAL;
681 }
682 F2FS_OPTION(sbi).write_io_size_bits = arg;
683 break;
684#ifdef CONFIG_F2FS_FAULT_INJECTION
685 case Opt_fault_injection:
686 if (args->from && match_int(args, &arg))
687 return -EINVAL;
688 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
689 set_opt(sbi, FAULT_INJECTION);
690 break;
691
692 case Opt_fault_type:
693 if (args->from && match_int(args, &arg))
694 return -EINVAL;
695 f2fs_build_fault_attr(sbi, 0, arg);
696 set_opt(sbi, FAULT_INJECTION);
697 break;
698#else
699 case Opt_fault_injection:
700 f2fs_info(sbi, "fault_injection options not supported");
701 break;
702
703 case Opt_fault_type:
704 f2fs_info(sbi, "fault_type options not supported");
705 break;
706#endif
707 case Opt_lazytime:
708 sb->s_flags |= SB_LAZYTIME;
709 break;
710 case Opt_nolazytime:
711 sb->s_flags &= ~SB_LAZYTIME;
712 break;
713#ifdef CONFIG_QUOTA
714 case Opt_quota:
715 case Opt_usrquota:
716 set_opt(sbi, USRQUOTA);
717 break;
718 case Opt_grpquota:
719 set_opt(sbi, GRPQUOTA);
720 break;
721 case Opt_prjquota:
722 set_opt(sbi, PRJQUOTA);
723 break;
724 case Opt_usrjquota:
725 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
726 if (ret)
727 return ret;
728 break;
729 case Opt_grpjquota:
730 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
731 if (ret)
732 return ret;
733 break;
734 case Opt_prjjquota:
735 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
736 if (ret)
737 return ret;
738 break;
739 case Opt_offusrjquota:
740 ret = f2fs_clear_qf_name(sb, USRQUOTA);
741 if (ret)
742 return ret;
743 break;
744 case Opt_offgrpjquota:
745 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
746 if (ret)
747 return ret;
748 break;
749 case Opt_offprjjquota:
750 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
751 if (ret)
752 return ret;
753 break;
754 case Opt_jqfmt_vfsold:
755 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
756 break;
757 case Opt_jqfmt_vfsv0:
758 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
759 break;
760 case Opt_jqfmt_vfsv1:
761 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
762 break;
763 case Opt_noquota:
764 clear_opt(sbi, QUOTA);
765 clear_opt(sbi, USRQUOTA);
766 clear_opt(sbi, GRPQUOTA);
767 clear_opt(sbi, PRJQUOTA);
768 break;
769#else
770 case Opt_quota:
771 case Opt_usrquota:
772 case Opt_grpquota:
773 case Opt_prjquota:
774 case Opt_usrjquota:
775 case Opt_grpjquota:
776 case Opt_prjjquota:
777 case Opt_offusrjquota:
778 case Opt_offgrpjquota:
779 case Opt_offprjjquota:
780 case Opt_jqfmt_vfsold:
781 case Opt_jqfmt_vfsv0:
782 case Opt_jqfmt_vfsv1:
783 case Opt_noquota:
784 f2fs_info(sbi, "quota operations not supported");
785 break;
786#endif
787 case Opt_whint:
788 name = match_strdup(&args[0]);
789 if (!name)
790 return -ENOMEM;
791 if (!strcmp(name, "user-based")) {
792 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
793 } else if (!strcmp(name, "off")) {
794 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
795 } else if (!strcmp(name, "fs-based")) {
796 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
797 } else {
798 kfree(name);
799 return -EINVAL;
800 }
801 kfree(name);
802 break;
803 case Opt_alloc:
804 name = match_strdup(&args[0]);
805 if (!name)
806 return -ENOMEM;
807
808 if (!strcmp(name, "default")) {
809 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
810 } else if (!strcmp(name, "reuse")) {
811 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
812 } else {
813 kfree(name);
814 return -EINVAL;
815 }
816 kfree(name);
817 break;
818 case Opt_fsync:
819 name = match_strdup(&args[0]);
820 if (!name)
821 return -ENOMEM;
822 if (!strcmp(name, "posix")) {
823 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
824 } else if (!strcmp(name, "strict")) {
825 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
826 } else if (!strcmp(name, "nobarrier")) {
827 F2FS_OPTION(sbi).fsync_mode =
828 FSYNC_MODE_NOBARRIER;
829 } else {
830 kfree(name);
831 return -EINVAL;
832 }
833 kfree(name);
834 break;
835 case Opt_test_dummy_encryption:
836 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
837 is_remount);
838 if (ret)
839 return ret;
840 break;
841 case Opt_inlinecrypt:
842#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
843 sb->s_flags |= SB_INLINECRYPT;
844#else
845 f2fs_info(sbi, "inline encryption not supported");
846#endif
847 break;
848 case Opt_checkpoint_disable_cap_perc:
849 if (args->from && match_int(args, &arg))
850 return -EINVAL;
851 if (arg < 0 || arg > 100)
852 return -EINVAL;
853 F2FS_OPTION(sbi).unusable_cap_perc = arg;
854 set_opt(sbi, DISABLE_CHECKPOINT);
855 break;
856 case Opt_checkpoint_disable_cap:
857 if (args->from && match_int(args, &arg))
858 return -EINVAL;
859 F2FS_OPTION(sbi).unusable_cap = arg;
860 set_opt(sbi, DISABLE_CHECKPOINT);
861 break;
862 case Opt_checkpoint_disable:
863 set_opt(sbi, DISABLE_CHECKPOINT);
864 break;
865 case Opt_checkpoint_enable:
866 clear_opt(sbi, DISABLE_CHECKPOINT);
867 break;
868#ifdef CONFIG_F2FS_FS_COMPRESSION
869 case Opt_compress_algorithm:
870 if (!f2fs_sb_has_compression(sbi)) {
871 f2fs_err(sbi, "Compression feature if off");
872 return -EINVAL;
873 }
874 name = match_strdup(&args[0]);
875 if (!name)
876 return -ENOMEM;
877 if (!strcmp(name, "lzo")) {
878 F2FS_OPTION(sbi).compress_algorithm =
879 COMPRESS_LZO;
880 } else if (!strcmp(name, "lz4")) {
881 F2FS_OPTION(sbi).compress_algorithm =
882 COMPRESS_LZ4;
883 } else if (!strcmp(name, "zstd")) {
884 F2FS_OPTION(sbi).compress_algorithm =
885 COMPRESS_ZSTD;
886 } else if (!strcmp(name, "lzo-rle")) {
887 F2FS_OPTION(sbi).compress_algorithm =
888 COMPRESS_LZORLE;
889 } else {
890 kfree(name);
891 return -EINVAL;
892 }
893 kfree(name);
894 break;
895 case Opt_compress_log_size:
896 if (!f2fs_sb_has_compression(sbi)) {
897 f2fs_err(sbi, "Compression feature is off");
898 return -EINVAL;
899 }
900 if (args->from && match_int(args, &arg))
901 return -EINVAL;
902 if (arg < MIN_COMPRESS_LOG_SIZE ||
903 arg > MAX_COMPRESS_LOG_SIZE) {
904 f2fs_err(sbi,
905 "Compress cluster log size is out of range");
906 return -EINVAL;
907 }
908 F2FS_OPTION(sbi).compress_log_size = arg;
909 break;
910 case Opt_compress_extension:
911 if (!f2fs_sb_has_compression(sbi)) {
912 f2fs_err(sbi, "Compression feature is off");
913 return -EINVAL;
914 }
915 name = match_strdup(&args[0]);
916 if (!name)
917 return -ENOMEM;
918
919 ext = F2FS_OPTION(sbi).extensions;
920 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
921
922 if (strlen(name) >= F2FS_EXTENSION_LEN ||
923 ext_cnt >= COMPRESS_EXT_NUM) {
924 f2fs_err(sbi,
925 "invalid extension length/number");
926 kfree(name);
927 return -EINVAL;
928 }
929
930 strcpy(ext[ext_cnt], name);
931 F2FS_OPTION(sbi).compress_ext_cnt++;
932 kfree(name);
933 break;
934#else
935 case Opt_compress_algorithm:
936 case Opt_compress_log_size:
937 case Opt_compress_extension:
938 f2fs_info(sbi, "compression options not supported");
939 break;
940#endif
941 default:
942 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
943 p);
944 return -EINVAL;
945 }
946 }
947#ifdef CONFIG_QUOTA
948 if (f2fs_check_quota_options(sbi))
949 return -EINVAL;
950#else
951 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
952 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
953 return -EINVAL;
954 }
955 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
956 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
957 return -EINVAL;
958 }
959#endif
960#ifndef CONFIG_UNICODE
961 if (f2fs_sb_has_casefold(sbi)) {
962 f2fs_err(sbi,
963 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
964 return -EINVAL;
965 }
966#endif
967
968 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
969 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
970 F2FS_IO_SIZE_KB(sbi));
971 return -EINVAL;
972 }
973
974 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
975 int min_size, max_size;
976
977 if (!f2fs_sb_has_extra_attr(sbi) ||
978 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
979 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
980 return -EINVAL;
981 }
982 if (!test_opt(sbi, INLINE_XATTR)) {
983 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
984 return -EINVAL;
985 }
986
987 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
988 max_size = MAX_INLINE_XATTR_SIZE;
989
990 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
991 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
992 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
993 min_size, max_size);
994 return -EINVAL;
995 }
996 }
997
998 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
999 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
1000 return -EINVAL;
1001 }
1002
1003 /* Not pass down write hints if the number of active logs is lesser
1004 * than NR_CURSEG_TYPE.
1005 */
1006 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
1007 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1008 return 0;
1009}
1010
1011static struct inode *f2fs_alloc_inode(struct super_block *sb)
1012{
1013 struct f2fs_inode_info *fi;
1014
1015 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
1016 if (!fi)
1017 return NULL;
1018
1019 init_once((void *) fi);
1020
1021 /* Initialize f2fs-specific inode info */
1022 atomic_set(&fi->dirty_pages, 0);
1023 init_rwsem(&fi->i_sem);
1024 spin_lock_init(&fi->i_size_lock);
1025 INIT_LIST_HEAD(&fi->dirty_list);
1026 INIT_LIST_HEAD(&fi->gdirty_list);
1027 INIT_LIST_HEAD(&fi->inmem_ilist);
1028 INIT_LIST_HEAD(&fi->inmem_pages);
1029 mutex_init(&fi->inmem_lock);
1030 init_rwsem(&fi->i_gc_rwsem[READ]);
1031 init_rwsem(&fi->i_gc_rwsem[WRITE]);
1032 init_rwsem(&fi->i_mmap_sem);
1033 init_rwsem(&fi->i_xattr_sem);
1034
1035 /* Will be used by directory only */
1036 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1037
1038 fi->ra_offset = -1;
1039
1040 return &fi->vfs_inode;
1041}
1042
1043static int f2fs_drop_inode(struct inode *inode)
1044{
1045 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1046 int ret;
1047
1048 /*
1049 * during filesystem shutdown, if checkpoint is disabled,
1050 * drop useless meta/node dirty pages.
1051 */
1052 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1053 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1054 inode->i_ino == F2FS_META_INO(sbi)) {
1055 trace_f2fs_drop_inode(inode, 1);
1056 return 1;
1057 }
1058 }
1059
1060 /*
1061 * This is to avoid a deadlock condition like below.
1062 * writeback_single_inode(inode)
1063 * - f2fs_write_data_page
1064 * - f2fs_gc -> iput -> evict
1065 * - inode_wait_for_writeback(inode)
1066 */
1067 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1068 if (!inode->i_nlink && !is_bad_inode(inode)) {
1069 /* to avoid evict_inode call simultaneously */
1070 atomic_inc(&inode->i_count);
1071 spin_unlock(&inode->i_lock);
1072
1073 /* some remained atomic pages should discarded */
1074 if (f2fs_is_atomic_file(inode))
1075 f2fs_drop_inmem_pages(inode);
1076
1077 /* should remain fi->extent_tree for writepage */
1078 f2fs_destroy_extent_node(inode);
1079
1080 sb_start_intwrite(inode->i_sb);
1081 f2fs_i_size_write(inode, 0);
1082
1083 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1084 inode, NULL, 0, DATA);
1085 truncate_inode_pages_final(inode->i_mapping);
1086
1087 if (F2FS_HAS_BLOCKS(inode))
1088 f2fs_truncate(inode);
1089
1090 sb_end_intwrite(inode->i_sb);
1091
1092 spin_lock(&inode->i_lock);
1093 atomic_dec(&inode->i_count);
1094 }
1095 trace_f2fs_drop_inode(inode, 0);
1096 return 0;
1097 }
1098 ret = generic_drop_inode(inode);
1099 if (!ret)
1100 ret = fscrypt_drop_inode(inode);
1101 trace_f2fs_drop_inode(inode, ret);
1102 return ret;
1103}
1104
1105int f2fs_inode_dirtied(struct inode *inode, bool sync)
1106{
1107 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1108 int ret = 0;
1109
1110 spin_lock(&sbi->inode_lock[DIRTY_META]);
1111 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1112 ret = 1;
1113 } else {
1114 set_inode_flag(inode, FI_DIRTY_INODE);
1115 stat_inc_dirty_inode(sbi, DIRTY_META);
1116 }
1117 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1118 list_add_tail(&F2FS_I(inode)->gdirty_list,
1119 &sbi->inode_list[DIRTY_META]);
1120 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1121 }
1122 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1123 return ret;
1124}
1125
1126void f2fs_inode_synced(struct inode *inode)
1127{
1128 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1129
1130 spin_lock(&sbi->inode_lock[DIRTY_META]);
1131 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1132 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1133 return;
1134 }
1135 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1136 list_del_init(&F2FS_I(inode)->gdirty_list);
1137 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1138 }
1139 clear_inode_flag(inode, FI_DIRTY_INODE);
1140 clear_inode_flag(inode, FI_AUTO_RECOVER);
1141 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1142 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1143}
1144
1145/*
1146 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1147 *
1148 * We should call set_dirty_inode to write the dirty inode through write_inode.
1149 */
1150static void f2fs_dirty_inode(struct inode *inode, int flags)
1151{
1152 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1153
1154 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1155 inode->i_ino == F2FS_META_INO(sbi))
1156 return;
1157
1158 if (flags == I_DIRTY_TIME)
1159 return;
1160
1161 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1162 clear_inode_flag(inode, FI_AUTO_RECOVER);
1163
1164 f2fs_inode_dirtied(inode, false);
1165}
1166
1167static void f2fs_free_inode(struct inode *inode)
1168{
1169 fscrypt_free_inode(inode);
1170 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1171}
1172
1173static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1174{
1175 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1176 percpu_counter_destroy(&sbi->total_valid_inode_count);
1177}
1178
1179static void destroy_device_list(struct f2fs_sb_info *sbi)
1180{
1181 int i;
1182
1183 for (i = 0; i < sbi->s_ndevs; i++) {
1184 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1185#ifdef CONFIG_BLK_DEV_ZONED
1186 kvfree(FDEV(i).blkz_seq);
1187#endif
1188 }
1189 kvfree(sbi->devs);
1190}
1191
1192static void f2fs_put_super(struct super_block *sb)
1193{
1194 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1195 int i;
1196 bool dropped;
1197
1198 /* unregister procfs/sysfs entries in advance to avoid race case */
1199 f2fs_unregister_sysfs(sbi);
1200
1201 f2fs_quota_off_umount(sb);
1202
1203 /* prevent remaining shrinker jobs */
1204 mutex_lock(&sbi->umount_mutex);
1205
1206 /*
1207 * We don't need to do checkpoint when superblock is clean.
1208 * But, the previous checkpoint was not done by umount, it needs to do
1209 * clean checkpoint again.
1210 */
1211 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1212 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1213 struct cp_control cpc = {
1214 .reason = CP_UMOUNT,
1215 };
1216 f2fs_write_checkpoint(sbi, &cpc);
1217 }
1218
1219 /* be sure to wait for any on-going discard commands */
1220 dropped = f2fs_issue_discard_timeout(sbi);
1221
1222 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1223 !sbi->discard_blks && !dropped) {
1224 struct cp_control cpc = {
1225 .reason = CP_UMOUNT | CP_TRIMMED,
1226 };
1227 f2fs_write_checkpoint(sbi, &cpc);
1228 }
1229
1230 /*
1231 * normally superblock is clean, so we need to release this.
1232 * In addition, EIO will skip do checkpoint, we need this as well.
1233 */
1234 f2fs_release_ino_entry(sbi, true);
1235
1236 f2fs_leave_shrinker(sbi);
1237 mutex_unlock(&sbi->umount_mutex);
1238
1239 /* our cp_error case, we can wait for any writeback page */
1240 f2fs_flush_merged_writes(sbi);
1241
1242 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1243
1244 f2fs_bug_on(sbi, sbi->fsync_node_num);
1245
1246 iput(sbi->node_inode);
1247 sbi->node_inode = NULL;
1248
1249 iput(sbi->meta_inode);
1250 sbi->meta_inode = NULL;
1251
1252 /*
1253 * iput() can update stat information, if f2fs_write_checkpoint()
1254 * above failed with error.
1255 */
1256 f2fs_destroy_stats(sbi);
1257
1258 /* destroy f2fs internal modules */
1259 f2fs_destroy_node_manager(sbi);
1260 f2fs_destroy_segment_manager(sbi);
1261
1262 f2fs_destroy_post_read_wq(sbi);
1263
1264 kvfree(sbi->ckpt);
1265
1266 sb->s_fs_info = NULL;
1267 if (sbi->s_chksum_driver)
1268 crypto_free_shash(sbi->s_chksum_driver);
1269 kfree(sbi->raw_super);
1270
1271 destroy_device_list(sbi);
1272 f2fs_destroy_xattr_caches(sbi);
1273 mempool_destroy(sbi->write_io_dummy);
1274#ifdef CONFIG_QUOTA
1275 for (i = 0; i < MAXQUOTAS; i++)
1276 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1277#endif
1278 fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx);
1279 destroy_percpu_info(sbi);
1280 for (i = 0; i < NR_PAGE_TYPE; i++)
1281 kvfree(sbi->write_io[i]);
1282#ifdef CONFIG_UNICODE
1283 utf8_unload(sbi->s_encoding);
1284#endif
1285 kfree(sbi);
1286}
1287
1288int f2fs_sync_fs(struct super_block *sb, int sync)
1289{
1290 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1291 int err = 0;
1292
1293 if (unlikely(f2fs_cp_error(sbi)))
1294 return 0;
1295 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1296 return 0;
1297
1298 trace_f2fs_sync_fs(sb, sync);
1299
1300 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1301 return -EAGAIN;
1302
1303 if (sync) {
1304 struct cp_control cpc;
1305
1306 cpc.reason = __get_cp_reason(sbi);
1307
1308 down_write(&sbi->gc_lock);
1309 err = f2fs_write_checkpoint(sbi, &cpc);
1310 up_write(&sbi->gc_lock);
1311 }
1312 f2fs_trace_ios(NULL, 1);
1313
1314 return err;
1315}
1316
1317static int f2fs_freeze(struct super_block *sb)
1318{
1319 if (f2fs_readonly(sb))
1320 return 0;
1321
1322 /* IO error happened before */
1323 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1324 return -EIO;
1325
1326 /* must be clean, since sync_filesystem() was already called */
1327 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1328 return -EINVAL;
1329 return 0;
1330}
1331
1332static int f2fs_unfreeze(struct super_block *sb)
1333{
1334 return 0;
1335}
1336
1337#ifdef CONFIG_QUOTA
1338static int f2fs_statfs_project(struct super_block *sb,
1339 kprojid_t projid, struct kstatfs *buf)
1340{
1341 struct kqid qid;
1342 struct dquot *dquot;
1343 u64 limit;
1344 u64 curblock;
1345
1346 qid = make_kqid_projid(projid);
1347 dquot = dqget(sb, qid);
1348 if (IS_ERR(dquot))
1349 return PTR_ERR(dquot);
1350 spin_lock(&dquot->dq_dqb_lock);
1351
1352 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1353 dquot->dq_dqb.dqb_bhardlimit);
1354 if (limit)
1355 limit >>= sb->s_blocksize_bits;
1356
1357 if (limit && buf->f_blocks > limit) {
1358 curblock = (dquot->dq_dqb.dqb_curspace +
1359 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1360 buf->f_blocks = limit;
1361 buf->f_bfree = buf->f_bavail =
1362 (buf->f_blocks > curblock) ?
1363 (buf->f_blocks - curblock) : 0;
1364 }
1365
1366 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1367 dquot->dq_dqb.dqb_ihardlimit);
1368
1369 if (limit && buf->f_files > limit) {
1370 buf->f_files = limit;
1371 buf->f_ffree =
1372 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1373 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1374 }
1375
1376 spin_unlock(&dquot->dq_dqb_lock);
1377 dqput(dquot);
1378 return 0;
1379}
1380#endif
1381
1382static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1383{
1384 struct super_block *sb = dentry->d_sb;
1385 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1386 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1387 block_t total_count, user_block_count, start_count;
1388 u64 avail_node_count;
1389
1390 total_count = le64_to_cpu(sbi->raw_super->block_count);
1391 user_block_count = sbi->user_block_count;
1392 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1393 buf->f_type = F2FS_SUPER_MAGIC;
1394 buf->f_bsize = sbi->blocksize;
1395
1396 buf->f_blocks = total_count - start_count;
1397 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1398 sbi->current_reserved_blocks;
1399
1400 spin_lock(&sbi->stat_lock);
1401 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1402 buf->f_bfree = 0;
1403 else
1404 buf->f_bfree -= sbi->unusable_block_count;
1405 spin_unlock(&sbi->stat_lock);
1406
1407 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1408 buf->f_bavail = buf->f_bfree -
1409 F2FS_OPTION(sbi).root_reserved_blocks;
1410 else
1411 buf->f_bavail = 0;
1412
1413 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1414
1415 if (avail_node_count > user_block_count) {
1416 buf->f_files = user_block_count;
1417 buf->f_ffree = buf->f_bavail;
1418 } else {
1419 buf->f_files = avail_node_count;
1420 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1421 buf->f_bavail);
1422 }
1423
1424 buf->f_namelen = F2FS_NAME_LEN;
1425 buf->f_fsid.val[0] = (u32)id;
1426 buf->f_fsid.val[1] = (u32)(id >> 32);
1427
1428#ifdef CONFIG_QUOTA
1429 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1430 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1431 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1432 }
1433#endif
1434 return 0;
1435}
1436
1437static inline void f2fs_show_quota_options(struct seq_file *seq,
1438 struct super_block *sb)
1439{
1440#ifdef CONFIG_QUOTA
1441 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1442
1443 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1444 char *fmtname = "";
1445
1446 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1447 case QFMT_VFS_OLD:
1448 fmtname = "vfsold";
1449 break;
1450 case QFMT_VFS_V0:
1451 fmtname = "vfsv0";
1452 break;
1453 case QFMT_VFS_V1:
1454 fmtname = "vfsv1";
1455 break;
1456 }
1457 seq_printf(seq, ",jqfmt=%s", fmtname);
1458 }
1459
1460 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1461 seq_show_option(seq, "usrjquota",
1462 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1463
1464 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1465 seq_show_option(seq, "grpjquota",
1466 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1467
1468 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1469 seq_show_option(seq, "prjjquota",
1470 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1471#endif
1472}
1473
1474static inline void f2fs_show_compress_options(struct seq_file *seq,
1475 struct super_block *sb)
1476{
1477 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1478 char *algtype = "";
1479 int i;
1480
1481 if (!f2fs_sb_has_compression(sbi))
1482 return;
1483
1484 switch (F2FS_OPTION(sbi).compress_algorithm) {
1485 case COMPRESS_LZO:
1486 algtype = "lzo";
1487 break;
1488 case COMPRESS_LZ4:
1489 algtype = "lz4";
1490 break;
1491 case COMPRESS_ZSTD:
1492 algtype = "zstd";
1493 break;
1494 case COMPRESS_LZORLE:
1495 algtype = "lzo-rle";
1496 break;
1497 }
1498 seq_printf(seq, ",compress_algorithm=%s", algtype);
1499
1500 seq_printf(seq, ",compress_log_size=%u",
1501 F2FS_OPTION(sbi).compress_log_size);
1502
1503 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1504 seq_printf(seq, ",compress_extension=%s",
1505 F2FS_OPTION(sbi).extensions[i]);
1506 }
1507}
1508
1509static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1510{
1511 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1512
1513 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1514 seq_printf(seq, ",background_gc=%s", "sync");
1515 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1516 seq_printf(seq, ",background_gc=%s", "on");
1517 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1518 seq_printf(seq, ",background_gc=%s", "off");
1519
1520 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1521 seq_puts(seq, ",disable_roll_forward");
1522 if (test_opt(sbi, NORECOVERY))
1523 seq_puts(seq, ",norecovery");
1524 if (test_opt(sbi, DISCARD))
1525 seq_puts(seq, ",discard");
1526 else
1527 seq_puts(seq, ",nodiscard");
1528 if (test_opt(sbi, NOHEAP))
1529 seq_puts(seq, ",no_heap");
1530 else
1531 seq_puts(seq, ",heap");
1532#ifdef CONFIG_F2FS_FS_XATTR
1533 if (test_opt(sbi, XATTR_USER))
1534 seq_puts(seq, ",user_xattr");
1535 else
1536 seq_puts(seq, ",nouser_xattr");
1537 if (test_opt(sbi, INLINE_XATTR))
1538 seq_puts(seq, ",inline_xattr");
1539 else
1540 seq_puts(seq, ",noinline_xattr");
1541 if (test_opt(sbi, INLINE_XATTR_SIZE))
1542 seq_printf(seq, ",inline_xattr_size=%u",
1543 F2FS_OPTION(sbi).inline_xattr_size);
1544#endif
1545#ifdef CONFIG_F2FS_FS_POSIX_ACL
1546 if (test_opt(sbi, POSIX_ACL))
1547 seq_puts(seq, ",acl");
1548 else
1549 seq_puts(seq, ",noacl");
1550#endif
1551 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1552 seq_puts(seq, ",disable_ext_identify");
1553 if (test_opt(sbi, INLINE_DATA))
1554 seq_puts(seq, ",inline_data");
1555 else
1556 seq_puts(seq, ",noinline_data");
1557 if (test_opt(sbi, INLINE_DENTRY))
1558 seq_puts(seq, ",inline_dentry");
1559 else
1560 seq_puts(seq, ",noinline_dentry");
1561 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1562 seq_puts(seq, ",flush_merge");
1563 if (test_opt(sbi, NOBARRIER))
1564 seq_puts(seq, ",nobarrier");
1565 if (test_opt(sbi, FASTBOOT))
1566 seq_puts(seq, ",fastboot");
1567 if (test_opt(sbi, EXTENT_CACHE))
1568 seq_puts(seq, ",extent_cache");
1569 else
1570 seq_puts(seq, ",noextent_cache");
1571 if (test_opt(sbi, DATA_FLUSH))
1572 seq_puts(seq, ",data_flush");
1573
1574 seq_puts(seq, ",mode=");
1575 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1576 seq_puts(seq, "adaptive");
1577 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1578 seq_puts(seq, "lfs");
1579 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1580 if (test_opt(sbi, RESERVE_ROOT))
1581 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1582 F2FS_OPTION(sbi).root_reserved_blocks,
1583 from_kuid_munged(&init_user_ns,
1584 F2FS_OPTION(sbi).s_resuid),
1585 from_kgid_munged(&init_user_ns,
1586 F2FS_OPTION(sbi).s_resgid));
1587 if (F2FS_IO_SIZE_BITS(sbi))
1588 seq_printf(seq, ",io_bits=%u",
1589 F2FS_OPTION(sbi).write_io_size_bits);
1590#ifdef CONFIG_F2FS_FAULT_INJECTION
1591 if (test_opt(sbi, FAULT_INJECTION)) {
1592 seq_printf(seq, ",fault_injection=%u",
1593 F2FS_OPTION(sbi).fault_info.inject_rate);
1594 seq_printf(seq, ",fault_type=%u",
1595 F2FS_OPTION(sbi).fault_info.inject_type);
1596 }
1597#endif
1598#ifdef CONFIG_QUOTA
1599 if (test_opt(sbi, QUOTA))
1600 seq_puts(seq, ",quota");
1601 if (test_opt(sbi, USRQUOTA))
1602 seq_puts(seq, ",usrquota");
1603 if (test_opt(sbi, GRPQUOTA))
1604 seq_puts(seq, ",grpquota");
1605 if (test_opt(sbi, PRJQUOTA))
1606 seq_puts(seq, ",prjquota");
1607#endif
1608 f2fs_show_quota_options(seq, sbi->sb);
1609 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1610 seq_printf(seq, ",whint_mode=%s", "user-based");
1611 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1612 seq_printf(seq, ",whint_mode=%s", "fs-based");
1613
1614 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1615
1616 if (sbi->sb->s_flags & SB_INLINECRYPT)
1617 seq_puts(seq, ",inlinecrypt");
1618
1619 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1620 seq_printf(seq, ",alloc_mode=%s", "default");
1621 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1622 seq_printf(seq, ",alloc_mode=%s", "reuse");
1623
1624 if (test_opt(sbi, DISABLE_CHECKPOINT))
1625 seq_printf(seq, ",checkpoint=disable:%u",
1626 F2FS_OPTION(sbi).unusable_cap);
1627 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1628 seq_printf(seq, ",fsync_mode=%s", "posix");
1629 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1630 seq_printf(seq, ",fsync_mode=%s", "strict");
1631 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1632 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1633
1634#ifdef CONFIG_F2FS_FS_COMPRESSION
1635 f2fs_show_compress_options(seq, sbi->sb);
1636#endif
1637 return 0;
1638}
1639
1640static void default_options(struct f2fs_sb_info *sbi)
1641{
1642 /* init some FS parameters */
1643 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1644 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1645 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1646 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1647 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1648 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1649 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1650 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1651 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1652 F2FS_OPTION(sbi).compress_ext_cnt = 0;
1653 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1654
1655 sbi->sb->s_flags &= ~SB_INLINECRYPT;
1656
1657 set_opt(sbi, INLINE_XATTR);
1658 set_opt(sbi, INLINE_DATA);
1659 set_opt(sbi, INLINE_DENTRY);
1660 set_opt(sbi, EXTENT_CACHE);
1661 set_opt(sbi, NOHEAP);
1662 clear_opt(sbi, DISABLE_CHECKPOINT);
1663 F2FS_OPTION(sbi).unusable_cap = 0;
1664 sbi->sb->s_flags |= SB_LAZYTIME;
1665 set_opt(sbi, FLUSH_MERGE);
1666 set_opt(sbi, DISCARD);
1667 if (f2fs_sb_has_blkzoned(sbi))
1668 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1669 else
1670 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1671
1672#ifdef CONFIG_F2FS_FS_XATTR
1673 set_opt(sbi, XATTR_USER);
1674#endif
1675#ifdef CONFIG_F2FS_FS_POSIX_ACL
1676 set_opt(sbi, POSIX_ACL);
1677#endif
1678
1679 f2fs_build_fault_attr(sbi, 0, 0);
1680}
1681
1682#ifdef CONFIG_QUOTA
1683static int f2fs_enable_quotas(struct super_block *sb);
1684#endif
1685
1686static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1687{
1688 unsigned int s_flags = sbi->sb->s_flags;
1689 struct cp_control cpc;
1690 int err = 0;
1691 int ret;
1692 block_t unusable;
1693
1694 if (s_flags & SB_RDONLY) {
1695 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1696 return -EINVAL;
1697 }
1698 sbi->sb->s_flags |= SB_ACTIVE;
1699
1700 f2fs_update_time(sbi, DISABLE_TIME);
1701
1702 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1703 down_write(&sbi->gc_lock);
1704 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1705 if (err == -ENODATA) {
1706 err = 0;
1707 break;
1708 }
1709 if (err && err != -EAGAIN)
1710 break;
1711 }
1712
1713 ret = sync_filesystem(sbi->sb);
1714 if (ret || err) {
1715 err = ret ? ret: err;
1716 goto restore_flag;
1717 }
1718
1719 unusable = f2fs_get_unusable_blocks(sbi);
1720 if (f2fs_disable_cp_again(sbi, unusable)) {
1721 err = -EAGAIN;
1722 goto restore_flag;
1723 }
1724
1725 down_write(&sbi->gc_lock);
1726 cpc.reason = CP_PAUSE;
1727 set_sbi_flag(sbi, SBI_CP_DISABLED);
1728 err = f2fs_write_checkpoint(sbi, &cpc);
1729 if (err)
1730 goto out_unlock;
1731
1732 spin_lock(&sbi->stat_lock);
1733 sbi->unusable_block_count = unusable;
1734 spin_unlock(&sbi->stat_lock);
1735
1736out_unlock:
1737 up_write(&sbi->gc_lock);
1738restore_flag:
1739 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
1740 return err;
1741}
1742
1743static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1744{
1745 down_write(&sbi->gc_lock);
1746 f2fs_dirty_to_prefree(sbi);
1747
1748 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1749 set_sbi_flag(sbi, SBI_IS_DIRTY);
1750 up_write(&sbi->gc_lock);
1751
1752 f2fs_sync_fs(sbi->sb, 1);
1753}
1754
1755static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1756{
1757 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1758 struct f2fs_mount_info org_mount_opt;
1759 unsigned long old_sb_flags;
1760 int err;
1761 bool need_restart_gc = false;
1762 bool need_stop_gc = false;
1763 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1764 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1765 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1766 bool checkpoint_changed;
1767#ifdef CONFIG_QUOTA
1768 int i, j;
1769#endif
1770
1771 /*
1772 * Save the old mount options in case we
1773 * need to restore them.
1774 */
1775 org_mount_opt = sbi->mount_opt;
1776 old_sb_flags = sb->s_flags;
1777
1778#ifdef CONFIG_QUOTA
1779 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1780 for (i = 0; i < MAXQUOTAS; i++) {
1781 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1782 org_mount_opt.s_qf_names[i] =
1783 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1784 GFP_KERNEL);
1785 if (!org_mount_opt.s_qf_names[i]) {
1786 for (j = 0; j < i; j++)
1787 kfree(org_mount_opt.s_qf_names[j]);
1788 return -ENOMEM;
1789 }
1790 } else {
1791 org_mount_opt.s_qf_names[i] = NULL;
1792 }
1793 }
1794#endif
1795
1796 /* recover superblocks we couldn't write due to previous RO mount */
1797 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1798 err = f2fs_commit_super(sbi, false);
1799 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1800 err);
1801 if (!err)
1802 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1803 }
1804
1805 default_options(sbi);
1806
1807 /* parse mount options */
1808 err = parse_options(sb, data, true);
1809 if (err)
1810 goto restore_opts;
1811 checkpoint_changed =
1812 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1813
1814 /*
1815 * Previous and new state of filesystem is RO,
1816 * so skip checking GC and FLUSH_MERGE conditions.
1817 */
1818 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1819 goto skip;
1820
1821#ifdef CONFIG_QUOTA
1822 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1823 err = dquot_suspend(sb, -1);
1824 if (err < 0)
1825 goto restore_opts;
1826 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1827 /* dquot_resume needs RW */
1828 sb->s_flags &= ~SB_RDONLY;
1829 if (sb_any_quota_suspended(sb)) {
1830 dquot_resume(sb, -1);
1831 } else if (f2fs_sb_has_quota_ino(sbi)) {
1832 err = f2fs_enable_quotas(sb);
1833 if (err)
1834 goto restore_opts;
1835 }
1836 }
1837#endif
1838 /* disallow enable/disable extent_cache dynamically */
1839 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1840 err = -EINVAL;
1841 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1842 goto restore_opts;
1843 }
1844
1845 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1846 err = -EINVAL;
1847 f2fs_warn(sbi, "switch io_bits option is not allowed");
1848 goto restore_opts;
1849 }
1850
1851 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1852 err = -EINVAL;
1853 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1854 goto restore_opts;
1855 }
1856
1857 /*
1858 * We stop the GC thread if FS is mounted as RO
1859 * or if background_gc = off is passed in mount
1860 * option. Also sync the filesystem.
1861 */
1862 if ((*flags & SB_RDONLY) ||
1863 F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) {
1864 if (sbi->gc_thread) {
1865 f2fs_stop_gc_thread(sbi);
1866 need_restart_gc = true;
1867 }
1868 } else if (!sbi->gc_thread) {
1869 err = f2fs_start_gc_thread(sbi);
1870 if (err)
1871 goto restore_opts;
1872 need_stop_gc = true;
1873 }
1874
1875 if (*flags & SB_RDONLY ||
1876 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1877 writeback_inodes_sb(sb, WB_REASON_SYNC);
1878 sync_inodes_sb(sb);
1879
1880 set_sbi_flag(sbi, SBI_IS_DIRTY);
1881 set_sbi_flag(sbi, SBI_IS_CLOSE);
1882 f2fs_sync_fs(sb, 1);
1883 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1884 }
1885
1886 if (checkpoint_changed) {
1887 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1888 err = f2fs_disable_checkpoint(sbi);
1889 if (err)
1890 goto restore_gc;
1891 } else {
1892 f2fs_enable_checkpoint(sbi);
1893 }
1894 }
1895
1896 /*
1897 * We stop issue flush thread if FS is mounted as RO
1898 * or if flush_merge is not passed in mount option.
1899 */
1900 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1901 clear_opt(sbi, FLUSH_MERGE);
1902 f2fs_destroy_flush_cmd_control(sbi, false);
1903 } else {
1904 err = f2fs_create_flush_cmd_control(sbi);
1905 if (err)
1906 goto restore_gc;
1907 }
1908skip:
1909#ifdef CONFIG_QUOTA
1910 /* Release old quota file names */
1911 for (i = 0; i < MAXQUOTAS; i++)
1912 kfree(org_mount_opt.s_qf_names[i]);
1913#endif
1914 /* Update the POSIXACL Flag */
1915 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1916 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1917
1918 limit_reserve_root(sbi);
1919 adjust_unusable_cap_perc(sbi);
1920 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1921 return 0;
1922restore_gc:
1923 if (need_restart_gc) {
1924 if (f2fs_start_gc_thread(sbi))
1925 f2fs_warn(sbi, "background gc thread has stopped");
1926 } else if (need_stop_gc) {
1927 f2fs_stop_gc_thread(sbi);
1928 }
1929restore_opts:
1930#ifdef CONFIG_QUOTA
1931 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1932 for (i = 0; i < MAXQUOTAS; i++) {
1933 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1934 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1935 }
1936#endif
1937 sbi->mount_opt = org_mount_opt;
1938 sb->s_flags = old_sb_flags;
1939 return err;
1940}
1941
1942#ifdef CONFIG_QUOTA
1943/* Read data from quotafile */
1944static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1945 size_t len, loff_t off)
1946{
1947 struct inode *inode = sb_dqopt(sb)->files[type];
1948 struct address_space *mapping = inode->i_mapping;
1949 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1950 int offset = off & (sb->s_blocksize - 1);
1951 int tocopy;
1952 size_t toread;
1953 loff_t i_size = i_size_read(inode);
1954 struct page *page;
1955 char *kaddr;
1956
1957 if (off > i_size)
1958 return 0;
1959
1960 if (off + len > i_size)
1961 len = i_size - off;
1962 toread = len;
1963 while (toread > 0) {
1964 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1965repeat:
1966 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1967 if (IS_ERR(page)) {
1968 if (PTR_ERR(page) == -ENOMEM) {
1969 congestion_wait(BLK_RW_ASYNC,
1970 DEFAULT_IO_TIMEOUT);
1971 goto repeat;
1972 }
1973 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1974 return PTR_ERR(page);
1975 }
1976
1977 lock_page(page);
1978
1979 if (unlikely(page->mapping != mapping)) {
1980 f2fs_put_page(page, 1);
1981 goto repeat;
1982 }
1983 if (unlikely(!PageUptodate(page))) {
1984 f2fs_put_page(page, 1);
1985 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1986 return -EIO;
1987 }
1988
1989 kaddr = kmap_atomic(page);
1990 memcpy(data, kaddr + offset, tocopy);
1991 kunmap_atomic(kaddr);
1992 f2fs_put_page(page, 1);
1993
1994 offset = 0;
1995 toread -= tocopy;
1996 data += tocopy;
1997 blkidx++;
1998 }
1999 return len;
2000}
2001
2002/* Write to quotafile */
2003static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2004 const char *data, size_t len, loff_t off)
2005{
2006 struct inode *inode = sb_dqopt(sb)->files[type];
2007 struct address_space *mapping = inode->i_mapping;
2008 const struct address_space_operations *a_ops = mapping->a_ops;
2009 int offset = off & (sb->s_blocksize - 1);
2010 size_t towrite = len;
2011 struct page *page;
2012 void *fsdata = NULL;
2013 char *kaddr;
2014 int err = 0;
2015 int tocopy;
2016
2017 while (towrite > 0) {
2018 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2019 towrite);
2020retry:
2021 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2022 &page, &fsdata);
2023 if (unlikely(err)) {
2024 if (err == -ENOMEM) {
2025 congestion_wait(BLK_RW_ASYNC,
2026 DEFAULT_IO_TIMEOUT);
2027 goto retry;
2028 }
2029 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2030 break;
2031 }
2032
2033 kaddr = kmap_atomic(page);
2034 memcpy(kaddr + offset, data, tocopy);
2035 kunmap_atomic(kaddr);
2036 flush_dcache_page(page);
2037
2038 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2039 page, fsdata);
2040 offset = 0;
2041 towrite -= tocopy;
2042 off += tocopy;
2043 data += tocopy;
2044 cond_resched();
2045 }
2046
2047 if (len == towrite)
2048 return err;
2049 inode->i_mtime = inode->i_ctime = current_time(inode);
2050 f2fs_mark_inode_dirty_sync(inode, false);
2051 return len - towrite;
2052}
2053
2054static struct dquot **f2fs_get_dquots(struct inode *inode)
2055{
2056 return F2FS_I(inode)->i_dquot;
2057}
2058
2059static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2060{
2061 return &F2FS_I(inode)->i_reserved_quota;
2062}
2063
2064static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2065{
2066 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2067 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2068 return 0;
2069 }
2070
2071 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2072 F2FS_OPTION(sbi).s_jquota_fmt, type);
2073}
2074
2075int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2076{
2077 int enabled = 0;
2078 int i, err;
2079
2080 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2081 err = f2fs_enable_quotas(sbi->sb);
2082 if (err) {
2083 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2084 return 0;
2085 }
2086 return 1;
2087 }
2088
2089 for (i = 0; i < MAXQUOTAS; i++) {
2090 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2091 err = f2fs_quota_on_mount(sbi, i);
2092 if (!err) {
2093 enabled = 1;
2094 continue;
2095 }
2096 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2097 err, i);
2098 }
2099 }
2100 return enabled;
2101}
2102
2103static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2104 unsigned int flags)
2105{
2106 struct inode *qf_inode;
2107 unsigned long qf_inum;
2108 int err;
2109
2110 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2111
2112 qf_inum = f2fs_qf_ino(sb, type);
2113 if (!qf_inum)
2114 return -EPERM;
2115
2116 qf_inode = f2fs_iget(sb, qf_inum);
2117 if (IS_ERR(qf_inode)) {
2118 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2119 return PTR_ERR(qf_inode);
2120 }
2121
2122 /* Don't account quota for quota files to avoid recursion */
2123 qf_inode->i_flags |= S_NOQUOTA;
2124 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2125 iput(qf_inode);
2126 return err;
2127}
2128
2129static int f2fs_enable_quotas(struct super_block *sb)
2130{
2131 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2132 int type, err = 0;
2133 unsigned long qf_inum;
2134 bool quota_mopt[MAXQUOTAS] = {
2135 test_opt(sbi, USRQUOTA),
2136 test_opt(sbi, GRPQUOTA),
2137 test_opt(sbi, PRJQUOTA),
2138 };
2139
2140 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2141 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2142 return 0;
2143 }
2144
2145 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2146
2147 for (type = 0; type < MAXQUOTAS; type++) {
2148 qf_inum = f2fs_qf_ino(sb, type);
2149 if (qf_inum) {
2150 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2151 DQUOT_USAGE_ENABLED |
2152 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2153 if (err) {
2154 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2155 type, err);
2156 for (type--; type >= 0; type--)
2157 dquot_quota_off(sb, type);
2158 set_sbi_flag(F2FS_SB(sb),
2159 SBI_QUOTA_NEED_REPAIR);
2160 return err;
2161 }
2162 }
2163 }
2164 return 0;
2165}
2166
2167int f2fs_quota_sync(struct super_block *sb, int type)
2168{
2169 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2170 struct quota_info *dqopt = sb_dqopt(sb);
2171 int cnt;
2172 int ret;
2173
2174 /*
2175 * do_quotactl
2176 * f2fs_quota_sync
2177 * down_read(quota_sem)
2178 * dquot_writeback_dquots()
2179 * f2fs_dquot_commit
2180 * block_operation
2181 * down_read(quota_sem)
2182 */
2183 f2fs_lock_op(sbi);
2184
2185 down_read(&sbi->quota_sem);
2186 ret = dquot_writeback_dquots(sb, type);
2187 if (ret)
2188 goto out;
2189
2190 /*
2191 * Now when everything is written we can discard the pagecache so
2192 * that userspace sees the changes.
2193 */
2194 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2195 struct address_space *mapping;
2196
2197 if (type != -1 && cnt != type)
2198 continue;
2199 if (!sb_has_quota_active(sb, cnt))
2200 continue;
2201
2202 mapping = dqopt->files[cnt]->i_mapping;
2203
2204 ret = filemap_fdatawrite(mapping);
2205 if (ret)
2206 goto out;
2207
2208 /* if we are using journalled quota */
2209 if (is_journalled_quota(sbi))
2210 continue;
2211
2212 ret = filemap_fdatawait(mapping);
2213 if (ret)
2214 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2215
2216 inode_lock(dqopt->files[cnt]);
2217 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2218 inode_unlock(dqopt->files[cnt]);
2219 }
2220out:
2221 if (ret)
2222 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2223 up_read(&sbi->quota_sem);
2224 f2fs_unlock_op(sbi);
2225 return ret;
2226}
2227
2228static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2229 const struct path *path)
2230{
2231 struct inode *inode;
2232 int err;
2233
2234 /* if quota sysfile exists, deny enabling quota with specific file */
2235 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2236 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2237 return -EBUSY;
2238 }
2239
2240 err = f2fs_quota_sync(sb, type);
2241 if (err)
2242 return err;
2243
2244 err = dquot_quota_on(sb, type, format_id, path);
2245 if (err)
2246 return err;
2247
2248 inode = d_inode(path->dentry);
2249
2250 inode_lock(inode);
2251 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2252 f2fs_set_inode_flags(inode);
2253 inode_unlock(inode);
2254 f2fs_mark_inode_dirty_sync(inode, false);
2255
2256 return 0;
2257}
2258
2259static int __f2fs_quota_off(struct super_block *sb, int type)
2260{
2261 struct inode *inode = sb_dqopt(sb)->files[type];
2262 int err;
2263
2264 if (!inode || !igrab(inode))
2265 return dquot_quota_off(sb, type);
2266
2267 err = f2fs_quota_sync(sb, type);
2268 if (err)
2269 goto out_put;
2270
2271 err = dquot_quota_off(sb, type);
2272 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2273 goto out_put;
2274
2275 inode_lock(inode);
2276 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2277 f2fs_set_inode_flags(inode);
2278 inode_unlock(inode);
2279 f2fs_mark_inode_dirty_sync(inode, false);
2280out_put:
2281 iput(inode);
2282 return err;
2283}
2284
2285static int f2fs_quota_off(struct super_block *sb, int type)
2286{
2287 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2288 int err;
2289
2290 err = __f2fs_quota_off(sb, type);
2291
2292 /*
2293 * quotactl can shutdown journalled quota, result in inconsistence
2294 * between quota record and fs data by following updates, tag the
2295 * flag to let fsck be aware of it.
2296 */
2297 if (is_journalled_quota(sbi))
2298 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2299 return err;
2300}
2301
2302void f2fs_quota_off_umount(struct super_block *sb)
2303{
2304 int type;
2305 int err;
2306
2307 for (type = 0; type < MAXQUOTAS; type++) {
2308 err = __f2fs_quota_off(sb, type);
2309 if (err) {
2310 int ret = dquot_quota_off(sb, type);
2311
2312 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2313 type, err, ret);
2314 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2315 }
2316 }
2317 /*
2318 * In case of checkpoint=disable, we must flush quota blocks.
2319 * This can cause NULL exception for node_inode in end_io, since
2320 * put_super already dropped it.
2321 */
2322 sync_filesystem(sb);
2323}
2324
2325static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2326{
2327 struct quota_info *dqopt = sb_dqopt(sb);
2328 int type;
2329
2330 for (type = 0; type < MAXQUOTAS; type++) {
2331 if (!dqopt->files[type])
2332 continue;
2333 f2fs_inode_synced(dqopt->files[type]);
2334 }
2335}
2336
2337static int f2fs_dquot_commit(struct dquot *dquot)
2338{
2339 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2340 int ret;
2341
2342 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2343 ret = dquot_commit(dquot);
2344 if (ret < 0)
2345 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2346 up_read(&sbi->quota_sem);
2347 return ret;
2348}
2349
2350static int f2fs_dquot_acquire(struct dquot *dquot)
2351{
2352 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2353 int ret;
2354
2355 down_read(&sbi->quota_sem);
2356 ret = dquot_acquire(dquot);
2357 if (ret < 0)
2358 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2359 up_read(&sbi->quota_sem);
2360 return ret;
2361}
2362
2363static int f2fs_dquot_release(struct dquot *dquot)
2364{
2365 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2366 int ret = dquot_release(dquot);
2367
2368 if (ret < 0)
2369 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2370 return ret;
2371}
2372
2373static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2374{
2375 struct super_block *sb = dquot->dq_sb;
2376 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2377 int ret = dquot_mark_dquot_dirty(dquot);
2378
2379 /* if we are using journalled quota */
2380 if (is_journalled_quota(sbi))
2381 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2382
2383 return ret;
2384}
2385
2386static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2387{
2388 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2389 int ret = dquot_commit_info(sb, type);
2390
2391 if (ret < 0)
2392 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2393 return ret;
2394}
2395
2396static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2397{
2398 *projid = F2FS_I(inode)->i_projid;
2399 return 0;
2400}
2401
2402static const struct dquot_operations f2fs_quota_operations = {
2403 .get_reserved_space = f2fs_get_reserved_space,
2404 .write_dquot = f2fs_dquot_commit,
2405 .acquire_dquot = f2fs_dquot_acquire,
2406 .release_dquot = f2fs_dquot_release,
2407 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2408 .write_info = f2fs_dquot_commit_info,
2409 .alloc_dquot = dquot_alloc,
2410 .destroy_dquot = dquot_destroy,
2411 .get_projid = f2fs_get_projid,
2412 .get_next_id = dquot_get_next_id,
2413};
2414
2415static const struct quotactl_ops f2fs_quotactl_ops = {
2416 .quota_on = f2fs_quota_on,
2417 .quota_off = f2fs_quota_off,
2418 .quota_sync = f2fs_quota_sync,
2419 .get_state = dquot_get_state,
2420 .set_info = dquot_set_dqinfo,
2421 .get_dqblk = dquot_get_dqblk,
2422 .set_dqblk = dquot_set_dqblk,
2423 .get_nextdqblk = dquot_get_next_dqblk,
2424};
2425#else
2426int f2fs_quota_sync(struct super_block *sb, int type)
2427{
2428 return 0;
2429}
2430
2431void f2fs_quota_off_umount(struct super_block *sb)
2432{
2433}
2434#endif
2435
2436static const struct super_operations f2fs_sops = {
2437 .alloc_inode = f2fs_alloc_inode,
2438 .free_inode = f2fs_free_inode,
2439 .drop_inode = f2fs_drop_inode,
2440 .write_inode = f2fs_write_inode,
2441 .dirty_inode = f2fs_dirty_inode,
2442 .show_options = f2fs_show_options,
2443#ifdef CONFIG_QUOTA
2444 .quota_read = f2fs_quota_read,
2445 .quota_write = f2fs_quota_write,
2446 .get_dquots = f2fs_get_dquots,
2447#endif
2448 .evict_inode = f2fs_evict_inode,
2449 .put_super = f2fs_put_super,
2450 .sync_fs = f2fs_sync_fs,
2451 .freeze_fs = f2fs_freeze,
2452 .unfreeze_fs = f2fs_unfreeze,
2453 .statfs = f2fs_statfs,
2454 .remount_fs = f2fs_remount,
2455};
2456
2457#ifdef CONFIG_FS_ENCRYPTION
2458static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2459{
2460 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2461 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2462 ctx, len, NULL);
2463}
2464
2465static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2466 void *fs_data)
2467{
2468 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2469
2470 /*
2471 * Encrypting the root directory is not allowed because fsck
2472 * expects lost+found directory to exist and remain unencrypted
2473 * if LOST_FOUND feature is enabled.
2474 *
2475 */
2476 if (f2fs_sb_has_lost_found(sbi) &&
2477 inode->i_ino == F2FS_ROOT_INO(sbi))
2478 return -EPERM;
2479
2480 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2481 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2482 ctx, len, fs_data, XATTR_CREATE);
2483}
2484
2485static const union fscrypt_context *
2486f2fs_get_dummy_context(struct super_block *sb)
2487{
2488 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_ctx.ctx;
2489}
2490
2491static bool f2fs_has_stable_inodes(struct super_block *sb)
2492{
2493 return true;
2494}
2495
2496static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2497 int *ino_bits_ret, int *lblk_bits_ret)
2498{
2499 *ino_bits_ret = 8 * sizeof(nid_t);
2500 *lblk_bits_ret = 8 * sizeof(block_t);
2501}
2502
2503static int f2fs_get_num_devices(struct super_block *sb)
2504{
2505 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2506
2507 if (f2fs_is_multi_device(sbi))
2508 return sbi->s_ndevs;
2509 return 1;
2510}
2511
2512static void f2fs_get_devices(struct super_block *sb,
2513 struct request_queue **devs)
2514{
2515 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2516 int i;
2517
2518 for (i = 0; i < sbi->s_ndevs; i++)
2519 devs[i] = bdev_get_queue(FDEV(i).bdev);
2520}
2521
2522static const struct fscrypt_operations f2fs_cryptops = {
2523 .key_prefix = "f2fs:",
2524 .get_context = f2fs_get_context,
2525 .set_context = f2fs_set_context,
2526 .get_dummy_context = f2fs_get_dummy_context,
2527 .empty_dir = f2fs_empty_dir,
2528 .max_namelen = F2FS_NAME_LEN,
2529 .has_stable_inodes = f2fs_has_stable_inodes,
2530 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2531 .get_num_devices = f2fs_get_num_devices,
2532 .get_devices = f2fs_get_devices,
2533};
2534#endif
2535
2536static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2537 u64 ino, u32 generation)
2538{
2539 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2540 struct inode *inode;
2541
2542 if (f2fs_check_nid_range(sbi, ino))
2543 return ERR_PTR(-ESTALE);
2544
2545 /*
2546 * f2fs_iget isn't quite right if the inode is currently unallocated!
2547 * However f2fs_iget currently does appropriate checks to handle stale
2548 * inodes so everything is OK.
2549 */
2550 inode = f2fs_iget(sb, ino);
2551 if (IS_ERR(inode))
2552 return ERR_CAST(inode);
2553 if (unlikely(generation && inode->i_generation != generation)) {
2554 /* we didn't find the right inode.. */
2555 iput(inode);
2556 return ERR_PTR(-ESTALE);
2557 }
2558 return inode;
2559}
2560
2561static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2562 int fh_len, int fh_type)
2563{
2564 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2565 f2fs_nfs_get_inode);
2566}
2567
2568static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2569 int fh_len, int fh_type)
2570{
2571 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2572 f2fs_nfs_get_inode);
2573}
2574
2575static const struct export_operations f2fs_export_ops = {
2576 .fh_to_dentry = f2fs_fh_to_dentry,
2577 .fh_to_parent = f2fs_fh_to_parent,
2578 .get_parent = f2fs_get_parent,
2579};
2580
2581static loff_t max_file_blocks(void)
2582{
2583 loff_t result = 0;
2584 loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2585
2586 /*
2587 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2588 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2589 * space in inode.i_addr, it will be more safe to reassign
2590 * result as zero.
2591 */
2592
2593 /* two direct node blocks */
2594 result += (leaf_count * 2);
2595
2596 /* two indirect node blocks */
2597 leaf_count *= NIDS_PER_BLOCK;
2598 result += (leaf_count * 2);
2599
2600 /* one double indirect node block */
2601 leaf_count *= NIDS_PER_BLOCK;
2602 result += leaf_count;
2603
2604 return result;
2605}
2606
2607static int __f2fs_commit_super(struct buffer_head *bh,
2608 struct f2fs_super_block *super)
2609{
2610 lock_buffer(bh);
2611 if (super)
2612 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2613 set_buffer_dirty(bh);
2614 unlock_buffer(bh);
2615
2616 /* it's rare case, we can do fua all the time */
2617 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2618}
2619
2620static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2621 struct buffer_head *bh)
2622{
2623 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2624 (bh->b_data + F2FS_SUPER_OFFSET);
2625 struct super_block *sb = sbi->sb;
2626 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2627 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2628 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2629 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2630 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2631 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2632 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2633 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2634 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2635 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2636 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2637 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2638 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2639 u64 main_end_blkaddr = main_blkaddr +
2640 (segment_count_main << log_blocks_per_seg);
2641 u64 seg_end_blkaddr = segment0_blkaddr +
2642 (segment_count << log_blocks_per_seg);
2643
2644 if (segment0_blkaddr != cp_blkaddr) {
2645 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2646 segment0_blkaddr, cp_blkaddr);
2647 return true;
2648 }
2649
2650 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2651 sit_blkaddr) {
2652 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2653 cp_blkaddr, sit_blkaddr,
2654 segment_count_ckpt << log_blocks_per_seg);
2655 return true;
2656 }
2657
2658 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2659 nat_blkaddr) {
2660 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2661 sit_blkaddr, nat_blkaddr,
2662 segment_count_sit << log_blocks_per_seg);
2663 return true;
2664 }
2665
2666 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2667 ssa_blkaddr) {
2668 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2669 nat_blkaddr, ssa_blkaddr,
2670 segment_count_nat << log_blocks_per_seg);
2671 return true;
2672 }
2673
2674 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2675 main_blkaddr) {
2676 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2677 ssa_blkaddr, main_blkaddr,
2678 segment_count_ssa << log_blocks_per_seg);
2679 return true;
2680 }
2681
2682 if (main_end_blkaddr > seg_end_blkaddr) {
2683 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2684 main_blkaddr,
2685 segment0_blkaddr +
2686 (segment_count << log_blocks_per_seg),
2687 segment_count_main << log_blocks_per_seg);
2688 return true;
2689 } else if (main_end_blkaddr < seg_end_blkaddr) {
2690 int err = 0;
2691 char *res;
2692
2693 /* fix in-memory information all the time */
2694 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2695 segment0_blkaddr) >> log_blocks_per_seg);
2696
2697 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2698 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2699 res = "internally";
2700 } else {
2701 err = __f2fs_commit_super(bh, NULL);
2702 res = err ? "failed" : "done";
2703 }
2704 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2705 res, main_blkaddr,
2706 segment0_blkaddr +
2707 (segment_count << log_blocks_per_seg),
2708 segment_count_main << log_blocks_per_seg);
2709 if (err)
2710 return true;
2711 }
2712 return false;
2713}
2714
2715static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2716 struct buffer_head *bh)
2717{
2718 block_t segment_count, segs_per_sec, secs_per_zone;
2719 block_t total_sections, blocks_per_seg;
2720 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2721 (bh->b_data + F2FS_SUPER_OFFSET);
2722 unsigned int blocksize;
2723 size_t crc_offset = 0;
2724 __u32 crc = 0;
2725
2726 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2727 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2728 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2729 return -EINVAL;
2730 }
2731
2732 /* Check checksum_offset and crc in superblock */
2733 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2734 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2735 if (crc_offset !=
2736 offsetof(struct f2fs_super_block, crc)) {
2737 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2738 crc_offset);
2739 return -EFSCORRUPTED;
2740 }
2741 crc = le32_to_cpu(raw_super->crc);
2742 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2743 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2744 return -EFSCORRUPTED;
2745 }
2746 }
2747
2748 /* Currently, support only 4KB page cache size */
2749 if (F2FS_BLKSIZE != PAGE_SIZE) {
2750 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2751 PAGE_SIZE);
2752 return -EFSCORRUPTED;
2753 }
2754
2755 /* Currently, support only 4KB block size */
2756 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2757 if (blocksize != F2FS_BLKSIZE) {
2758 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2759 blocksize);
2760 return -EFSCORRUPTED;
2761 }
2762
2763 /* check log blocks per segment */
2764 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2765 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2766 le32_to_cpu(raw_super->log_blocks_per_seg));
2767 return -EFSCORRUPTED;
2768 }
2769
2770 /* Currently, support 512/1024/2048/4096 bytes sector size */
2771 if (le32_to_cpu(raw_super->log_sectorsize) >
2772 F2FS_MAX_LOG_SECTOR_SIZE ||
2773 le32_to_cpu(raw_super->log_sectorsize) <
2774 F2FS_MIN_LOG_SECTOR_SIZE) {
2775 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2776 le32_to_cpu(raw_super->log_sectorsize));
2777 return -EFSCORRUPTED;
2778 }
2779 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2780 le32_to_cpu(raw_super->log_sectorsize) !=
2781 F2FS_MAX_LOG_SECTOR_SIZE) {
2782 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2783 le32_to_cpu(raw_super->log_sectors_per_block),
2784 le32_to_cpu(raw_super->log_sectorsize));
2785 return -EFSCORRUPTED;
2786 }
2787
2788 segment_count = le32_to_cpu(raw_super->segment_count);
2789 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2790 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2791 total_sections = le32_to_cpu(raw_super->section_count);
2792
2793 /* blocks_per_seg should be 512, given the above check */
2794 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2795
2796 if (segment_count > F2FS_MAX_SEGMENT ||
2797 segment_count < F2FS_MIN_SEGMENTS) {
2798 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2799 return -EFSCORRUPTED;
2800 }
2801
2802 if (total_sections > segment_count ||
2803 total_sections < F2FS_MIN_SEGMENTS ||
2804 segs_per_sec > segment_count || !segs_per_sec) {
2805 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2806 segment_count, total_sections, segs_per_sec);
2807 return -EFSCORRUPTED;
2808 }
2809
2810 if ((segment_count / segs_per_sec) < total_sections) {
2811 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2812 segment_count, segs_per_sec, total_sections);
2813 return -EFSCORRUPTED;
2814 }
2815
2816 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2817 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2818 segment_count, le64_to_cpu(raw_super->block_count));
2819 return -EFSCORRUPTED;
2820 }
2821
2822 if (RDEV(0).path[0]) {
2823 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2824 int i = 1;
2825
2826 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2827 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2828 i++;
2829 }
2830 if (segment_count != dev_seg_count) {
2831 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2832 segment_count, dev_seg_count);
2833 return -EFSCORRUPTED;
2834 }
2835 }
2836
2837 if (secs_per_zone > total_sections || !secs_per_zone) {
2838 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2839 secs_per_zone, total_sections);
2840 return -EFSCORRUPTED;
2841 }
2842 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2843 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2844 (le32_to_cpu(raw_super->extension_count) +
2845 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2846 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2847 le32_to_cpu(raw_super->extension_count),
2848 raw_super->hot_ext_count,
2849 F2FS_MAX_EXTENSION);
2850 return -EFSCORRUPTED;
2851 }
2852
2853 if (le32_to_cpu(raw_super->cp_payload) >
2854 (blocks_per_seg - F2FS_CP_PACKS)) {
2855 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2856 le32_to_cpu(raw_super->cp_payload),
2857 blocks_per_seg - F2FS_CP_PACKS);
2858 return -EFSCORRUPTED;
2859 }
2860
2861 /* check reserved ino info */
2862 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2863 le32_to_cpu(raw_super->meta_ino) != 2 ||
2864 le32_to_cpu(raw_super->root_ino) != 3) {
2865 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2866 le32_to_cpu(raw_super->node_ino),
2867 le32_to_cpu(raw_super->meta_ino),
2868 le32_to_cpu(raw_super->root_ino));
2869 return -EFSCORRUPTED;
2870 }
2871
2872 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2873 if (sanity_check_area_boundary(sbi, bh))
2874 return -EFSCORRUPTED;
2875
2876 return 0;
2877}
2878
2879int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2880{
2881 unsigned int total, fsmeta;
2882 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2883 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2884 unsigned int ovp_segments, reserved_segments;
2885 unsigned int main_segs, blocks_per_seg;
2886 unsigned int sit_segs, nat_segs;
2887 unsigned int sit_bitmap_size, nat_bitmap_size;
2888 unsigned int log_blocks_per_seg;
2889 unsigned int segment_count_main;
2890 unsigned int cp_pack_start_sum, cp_payload;
2891 block_t user_block_count, valid_user_blocks;
2892 block_t avail_node_count, valid_node_count;
2893 int i, j;
2894
2895 total = le32_to_cpu(raw_super->segment_count);
2896 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2897 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2898 fsmeta += sit_segs;
2899 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2900 fsmeta += nat_segs;
2901 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2902 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2903
2904 if (unlikely(fsmeta >= total))
2905 return 1;
2906
2907 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2908 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2909
2910 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2911 ovp_segments == 0 || reserved_segments == 0)) {
2912 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2913 return 1;
2914 }
2915
2916 user_block_count = le64_to_cpu(ckpt->user_block_count);
2917 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2918 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2919 if (!user_block_count || user_block_count >=
2920 segment_count_main << log_blocks_per_seg) {
2921 f2fs_err(sbi, "Wrong user_block_count: %u",
2922 user_block_count);
2923 return 1;
2924 }
2925
2926 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2927 if (valid_user_blocks > user_block_count) {
2928 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2929 valid_user_blocks, user_block_count);
2930 return 1;
2931 }
2932
2933 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2934 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2935 if (valid_node_count > avail_node_count) {
2936 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2937 valid_node_count, avail_node_count);
2938 return 1;
2939 }
2940
2941 main_segs = le32_to_cpu(raw_super->segment_count_main);
2942 blocks_per_seg = sbi->blocks_per_seg;
2943
2944 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2945 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2946 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2947 return 1;
2948 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2949 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2950 le32_to_cpu(ckpt->cur_node_segno[j])) {
2951 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2952 i, j,
2953 le32_to_cpu(ckpt->cur_node_segno[i]));
2954 return 1;
2955 }
2956 }
2957 }
2958 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2959 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2960 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2961 return 1;
2962 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2963 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2964 le32_to_cpu(ckpt->cur_data_segno[j])) {
2965 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2966 i, j,
2967 le32_to_cpu(ckpt->cur_data_segno[i]));
2968 return 1;
2969 }
2970 }
2971 }
2972 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2973 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2974 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2975 le32_to_cpu(ckpt->cur_data_segno[j])) {
2976 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2977 i, j,
2978 le32_to_cpu(ckpt->cur_node_segno[i]));
2979 return 1;
2980 }
2981 }
2982 }
2983
2984 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2985 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2986
2987 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2988 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2989 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2990 sit_bitmap_size, nat_bitmap_size);
2991 return 1;
2992 }
2993
2994 cp_pack_start_sum = __start_sum_addr(sbi);
2995 cp_payload = __cp_payload(sbi);
2996 if (cp_pack_start_sum < cp_payload + 1 ||
2997 cp_pack_start_sum > blocks_per_seg - 1 -
2998 NR_CURSEG_TYPE) {
2999 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3000 cp_pack_start_sum);
3001 return 1;
3002 }
3003
3004 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3005 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3006 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3007 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3008 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3009 le32_to_cpu(ckpt->checksum_offset));
3010 return 1;
3011 }
3012
3013 if (unlikely(f2fs_cp_error(sbi))) {
3014 f2fs_err(sbi, "A bug case: need to run fsck");
3015 return 1;
3016 }
3017 return 0;
3018}
3019
3020static void init_sb_info(struct f2fs_sb_info *sbi)
3021{
3022 struct f2fs_super_block *raw_super = sbi->raw_super;
3023 int i;
3024
3025 sbi->log_sectors_per_block =
3026 le32_to_cpu(raw_super->log_sectors_per_block);
3027 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3028 sbi->blocksize = 1 << sbi->log_blocksize;
3029 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3030 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3031 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3032 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3033 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3034 sbi->total_node_count =
3035 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3036 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3037 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
3038 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
3039 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
3040 sbi->cur_victim_sec = NULL_SECNO;
3041 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3042 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3043 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3044 sbi->migration_granularity = sbi->segs_per_sec;
3045
3046 sbi->dir_level = DEF_DIR_LEVEL;
3047 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3048 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3049 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3050 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3051 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3052 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3053 DEF_UMOUNT_DISCARD_TIMEOUT;
3054 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3055
3056 for (i = 0; i < NR_COUNT_TYPE; i++)
3057 atomic_set(&sbi->nr_pages[i], 0);
3058
3059 for (i = 0; i < META; i++)
3060 atomic_set(&sbi->wb_sync_req[i], 0);
3061
3062 INIT_LIST_HEAD(&sbi->s_list);
3063 mutex_init(&sbi->umount_mutex);
3064 init_rwsem(&sbi->io_order_lock);
3065 spin_lock_init(&sbi->cp_lock);
3066
3067 sbi->dirty_device = 0;
3068 spin_lock_init(&sbi->dev_lock);
3069
3070 init_rwsem(&sbi->sb_lock);
3071 init_rwsem(&sbi->pin_sem);
3072}
3073
3074static int init_percpu_info(struct f2fs_sb_info *sbi)
3075{
3076 int err;
3077
3078 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3079 if (err)
3080 return err;
3081
3082 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3083 GFP_KERNEL);
3084 if (err)
3085 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3086
3087 return err;
3088}
3089
3090#ifdef CONFIG_BLK_DEV_ZONED
3091static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3092 void *data)
3093{
3094 struct f2fs_dev_info *dev = data;
3095
3096 if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
3097 set_bit(idx, dev->blkz_seq);
3098 return 0;
3099}
3100
3101static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3102{
3103 struct block_device *bdev = FDEV(devi).bdev;
3104 sector_t nr_sectors = bdev->bd_part->nr_sects;
3105 int ret;
3106
3107 if (!f2fs_sb_has_blkzoned(sbi))
3108 return 0;
3109
3110 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3111 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3112 return -EINVAL;
3113 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3114 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3115 __ilog2_u32(sbi->blocks_per_blkz))
3116 return -EINVAL;
3117 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3118 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3119 sbi->log_blocks_per_blkz;
3120 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3121 FDEV(devi).nr_blkz++;
3122
3123 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3124 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3125 * sizeof(unsigned long),
3126 GFP_KERNEL);
3127 if (!FDEV(devi).blkz_seq)
3128 return -ENOMEM;
3129
3130 /* Get block zones type */
3131 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3132 &FDEV(devi));
3133 if (ret < 0)
3134 return ret;
3135
3136 return 0;
3137}
3138#endif
3139
3140/*
3141 * Read f2fs raw super block.
3142 * Because we have two copies of super block, so read both of them
3143 * to get the first valid one. If any one of them is broken, we pass
3144 * them recovery flag back to the caller.
3145 */
3146static int read_raw_super_block(struct f2fs_sb_info *sbi,
3147 struct f2fs_super_block **raw_super,
3148 int *valid_super_block, int *recovery)
3149{
3150 struct super_block *sb = sbi->sb;
3151 int block;
3152 struct buffer_head *bh;
3153 struct f2fs_super_block *super;
3154 int err = 0;
3155
3156 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3157 if (!super)
3158 return -ENOMEM;
3159
3160 for (block = 0; block < 2; block++) {
3161 bh = sb_bread(sb, block);
3162 if (!bh) {
3163 f2fs_err(sbi, "Unable to read %dth superblock",
3164 block + 1);
3165 err = -EIO;
3166 *recovery = 1;
3167 continue;
3168 }
3169
3170 /* sanity checking of raw super */
3171 err = sanity_check_raw_super(sbi, bh);
3172 if (err) {
3173 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3174 block + 1);
3175 brelse(bh);
3176 *recovery = 1;
3177 continue;
3178 }
3179
3180 if (!*raw_super) {
3181 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3182 sizeof(*super));
3183 *valid_super_block = block;
3184 *raw_super = super;
3185 }
3186 brelse(bh);
3187 }
3188
3189 /* No valid superblock */
3190 if (!*raw_super)
3191 kfree(super);
3192 else
3193 err = 0;
3194
3195 return err;
3196}
3197
3198int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3199{
3200 struct buffer_head *bh;
3201 __u32 crc = 0;
3202 int err;
3203
3204 if ((recover && f2fs_readonly(sbi->sb)) ||
3205 bdev_read_only(sbi->sb->s_bdev)) {
3206 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3207 return -EROFS;
3208 }
3209
3210 /* we should update superblock crc here */
3211 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3212 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3213 offsetof(struct f2fs_super_block, crc));
3214 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3215 }
3216
3217 /* write back-up superblock first */
3218 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3219 if (!bh)
3220 return -EIO;
3221 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3222 brelse(bh);
3223
3224 /* if we are in recovery path, skip writing valid superblock */
3225 if (recover || err)
3226 return err;
3227
3228 /* write current valid superblock */
3229 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3230 if (!bh)
3231 return -EIO;
3232 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3233 brelse(bh);
3234 return err;
3235}
3236
3237static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3238{
3239 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3240 unsigned int max_devices = MAX_DEVICES;
3241 int i;
3242
3243 /* Initialize single device information */
3244 if (!RDEV(0).path[0]) {
3245 if (!bdev_is_zoned(sbi->sb->s_bdev))
3246 return 0;
3247 max_devices = 1;
3248 }
3249
3250 /*
3251 * Initialize multiple devices information, or single
3252 * zoned block device information.
3253 */
3254 sbi->devs = f2fs_kzalloc(sbi,
3255 array_size(max_devices,
3256 sizeof(struct f2fs_dev_info)),
3257 GFP_KERNEL);
3258 if (!sbi->devs)
3259 return -ENOMEM;
3260
3261 for (i = 0; i < max_devices; i++) {
3262
3263 if (i > 0 && !RDEV(i).path[0])
3264 break;
3265
3266 if (max_devices == 1) {
3267 /* Single zoned block device mount */
3268 FDEV(0).bdev =
3269 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3270 sbi->sb->s_mode, sbi->sb->s_type);
3271 } else {
3272 /* Multi-device mount */
3273 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3274 FDEV(i).total_segments =
3275 le32_to_cpu(RDEV(i).total_segments);
3276 if (i == 0) {
3277 FDEV(i).start_blk = 0;
3278 FDEV(i).end_blk = FDEV(i).start_blk +
3279 (FDEV(i).total_segments <<
3280 sbi->log_blocks_per_seg) - 1 +
3281 le32_to_cpu(raw_super->segment0_blkaddr);
3282 } else {
3283 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3284 FDEV(i).end_blk = FDEV(i).start_blk +
3285 (FDEV(i).total_segments <<
3286 sbi->log_blocks_per_seg) - 1;
3287 }
3288 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3289 sbi->sb->s_mode, sbi->sb->s_type);
3290 }
3291 if (IS_ERR(FDEV(i).bdev))
3292 return PTR_ERR(FDEV(i).bdev);
3293
3294 /* to release errored devices */
3295 sbi->s_ndevs = i + 1;
3296
3297#ifdef CONFIG_BLK_DEV_ZONED
3298 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3299 !f2fs_sb_has_blkzoned(sbi)) {
3300 f2fs_err(sbi, "Zoned block device feature not enabled\n");
3301 return -EINVAL;
3302 }
3303 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3304 if (init_blkz_info(sbi, i)) {
3305 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3306 return -EINVAL;
3307 }
3308 if (max_devices == 1)
3309 break;
3310 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3311 i, FDEV(i).path,
3312 FDEV(i).total_segments,
3313 FDEV(i).start_blk, FDEV(i).end_blk,
3314 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3315 "Host-aware" : "Host-managed");
3316 continue;
3317 }
3318#endif
3319 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3320 i, FDEV(i).path,
3321 FDEV(i).total_segments,
3322 FDEV(i).start_blk, FDEV(i).end_blk);
3323 }
3324 f2fs_info(sbi,
3325 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3326 return 0;
3327}
3328
3329static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3330{
3331#ifdef CONFIG_UNICODE
3332 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3333 const struct f2fs_sb_encodings *encoding_info;
3334 struct unicode_map *encoding;
3335 __u16 encoding_flags;
3336
3337 if (f2fs_sb_has_encrypt(sbi)) {
3338 f2fs_err(sbi,
3339 "Can't mount with encoding and encryption");
3340 return -EINVAL;
3341 }
3342
3343 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3344 &encoding_flags)) {
3345 f2fs_err(sbi,
3346 "Encoding requested by superblock is unknown");
3347 return -EINVAL;
3348 }
3349
3350 encoding = utf8_load(encoding_info->version);
3351 if (IS_ERR(encoding)) {
3352 f2fs_err(sbi,
3353 "can't mount with superblock charset: %s-%s "
3354 "not supported by the kernel. flags: 0x%x.",
3355 encoding_info->name, encoding_info->version,
3356 encoding_flags);
3357 return PTR_ERR(encoding);
3358 }
3359 f2fs_info(sbi, "Using encoding defined by superblock: "
3360 "%s-%s with flags 0x%hx", encoding_info->name,
3361 encoding_info->version?:"\b", encoding_flags);
3362
3363 sbi->s_encoding = encoding;
3364 sbi->s_encoding_flags = encoding_flags;
3365 sbi->sb->s_d_op = &f2fs_dentry_ops;
3366 }
3367#else
3368 if (f2fs_sb_has_casefold(sbi)) {
3369 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3370 return -EINVAL;
3371 }
3372#endif
3373 return 0;
3374}
3375
3376static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3377{
3378 struct f2fs_sm_info *sm_i = SM_I(sbi);
3379
3380 /* adjust parameters according to the volume size */
3381 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3382 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3383 sm_i->dcc_info->discard_granularity = 1;
3384 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3385 }
3386
3387 sbi->readdir_ra = 1;
3388}
3389
3390static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3391{
3392 struct f2fs_sb_info *sbi;
3393 struct f2fs_super_block *raw_super;
3394 struct inode *root;
3395 int err;
3396 bool skip_recovery = false, need_fsck = false;
3397 char *options = NULL;
3398 int recovery, i, valid_super_block;
3399 struct curseg_info *seg_i;
3400 int retry_cnt = 1;
3401
3402try_onemore:
3403 err = -EINVAL;
3404 raw_super = NULL;
3405 valid_super_block = -1;
3406 recovery = 0;
3407
3408 /* allocate memory for f2fs-specific super block info */
3409 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3410 if (!sbi)
3411 return -ENOMEM;
3412
3413 sbi->sb = sb;
3414
3415 /* Load the checksum driver */
3416 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3417 if (IS_ERR(sbi->s_chksum_driver)) {
3418 f2fs_err(sbi, "Cannot load crc32 driver.");
3419 err = PTR_ERR(sbi->s_chksum_driver);
3420 sbi->s_chksum_driver = NULL;
3421 goto free_sbi;
3422 }
3423
3424 /* set a block size */
3425 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3426 f2fs_err(sbi, "unable to set blocksize");
3427 goto free_sbi;
3428 }
3429
3430 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3431 &recovery);
3432 if (err)
3433 goto free_sbi;
3434
3435 sb->s_fs_info = sbi;
3436 sbi->raw_super = raw_super;
3437
3438 /* precompute checksum seed for metadata */
3439 if (f2fs_sb_has_inode_chksum(sbi))
3440 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3441 sizeof(raw_super->uuid));
3442
3443 /*
3444 * The BLKZONED feature indicates that the drive was formatted with
3445 * zone alignment optimization. This is optional for host-aware
3446 * devices, but mandatory for host-managed zoned block devices.
3447 */
3448#ifndef CONFIG_BLK_DEV_ZONED
3449 if (f2fs_sb_has_blkzoned(sbi)) {
3450 f2fs_err(sbi, "Zoned block device support is not enabled");
3451 err = -EOPNOTSUPP;
3452 goto free_sb_buf;
3453 }
3454#endif
3455 default_options(sbi);
3456 /* parse mount options */
3457 options = kstrdup((const char *)data, GFP_KERNEL);
3458 if (data && !options) {
3459 err = -ENOMEM;
3460 goto free_sb_buf;
3461 }
3462
3463 err = parse_options(sb, options, false);
3464 if (err)
3465 goto free_options;
3466
3467 sbi->max_file_blocks = max_file_blocks();
3468 sb->s_maxbytes = sbi->max_file_blocks <<
3469 le32_to_cpu(raw_super->log_blocksize);
3470 sb->s_max_links = F2FS_LINK_MAX;
3471
3472 err = f2fs_setup_casefold(sbi);
3473 if (err)
3474 goto free_options;
3475
3476#ifdef CONFIG_QUOTA
3477 sb->dq_op = &f2fs_quota_operations;
3478 sb->s_qcop = &f2fs_quotactl_ops;
3479 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3480
3481 if (f2fs_sb_has_quota_ino(sbi)) {
3482 for (i = 0; i < MAXQUOTAS; i++) {
3483 if (f2fs_qf_ino(sbi->sb, i))
3484 sbi->nquota_files++;
3485 }
3486 }
3487#endif
3488
3489 sb->s_op = &f2fs_sops;
3490#ifdef CONFIG_FS_ENCRYPTION
3491 sb->s_cop = &f2fs_cryptops;
3492#endif
3493#ifdef CONFIG_FS_VERITY
3494 sb->s_vop = &f2fs_verityops;
3495#endif
3496 sb->s_xattr = f2fs_xattr_handlers;
3497 sb->s_export_op = &f2fs_export_ops;
3498 sb->s_magic = F2FS_SUPER_MAGIC;
3499 sb->s_time_gran = 1;
3500 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3501 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3502 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3503 sb->s_iflags |= SB_I_CGROUPWB;
3504
3505 /* init f2fs-specific super block info */
3506 sbi->valid_super_block = valid_super_block;
3507 init_rwsem(&sbi->gc_lock);
3508 mutex_init(&sbi->writepages);
3509 mutex_init(&sbi->cp_mutex);
3510 init_rwsem(&sbi->node_write);
3511 init_rwsem(&sbi->node_change);
3512
3513 /* disallow all the data/node/meta page writes */
3514 set_sbi_flag(sbi, SBI_POR_DOING);
3515 spin_lock_init(&sbi->stat_lock);
3516
3517 /* init iostat info */
3518 spin_lock_init(&sbi->iostat_lock);
3519 sbi->iostat_enable = false;
3520 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3521
3522 for (i = 0; i < NR_PAGE_TYPE; i++) {
3523 int n = (i == META) ? 1: NR_TEMP_TYPE;
3524 int j;
3525
3526 sbi->write_io[i] =
3527 f2fs_kmalloc(sbi,
3528 array_size(n,
3529 sizeof(struct f2fs_bio_info)),
3530 GFP_KERNEL);
3531 if (!sbi->write_io[i]) {
3532 err = -ENOMEM;
3533 goto free_bio_info;
3534 }
3535
3536 for (j = HOT; j < n; j++) {
3537 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3538 sbi->write_io[i][j].sbi = sbi;
3539 sbi->write_io[i][j].bio = NULL;
3540 spin_lock_init(&sbi->write_io[i][j].io_lock);
3541 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3542 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3543 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3544 }
3545 }
3546
3547 init_rwsem(&sbi->cp_rwsem);
3548 init_rwsem(&sbi->quota_sem);
3549 init_waitqueue_head(&sbi->cp_wait);
3550 init_sb_info(sbi);
3551
3552 err = init_percpu_info(sbi);
3553 if (err)
3554 goto free_bio_info;
3555
3556 if (F2FS_IO_ALIGNED(sbi)) {
3557 sbi->write_io_dummy =
3558 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3559 if (!sbi->write_io_dummy) {
3560 err = -ENOMEM;
3561 goto free_percpu;
3562 }
3563 }
3564
3565 /* init per sbi slab cache */
3566 err = f2fs_init_xattr_caches(sbi);
3567 if (err)
3568 goto free_io_dummy;
3569
3570 /* get an inode for meta space */
3571 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3572 if (IS_ERR(sbi->meta_inode)) {
3573 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3574 err = PTR_ERR(sbi->meta_inode);
3575 goto free_xattr_cache;
3576 }
3577
3578 err = f2fs_get_valid_checkpoint(sbi);
3579 if (err) {
3580 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3581 goto free_meta_inode;
3582 }
3583
3584 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3585 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3586 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3587 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3588 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3589 }
3590
3591 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3592 set_sbi_flag(sbi, SBI_NEED_FSCK);
3593
3594 /* Initialize device list */
3595 err = f2fs_scan_devices(sbi);
3596 if (err) {
3597 f2fs_err(sbi, "Failed to find devices");
3598 goto free_devices;
3599 }
3600
3601 err = f2fs_init_post_read_wq(sbi);
3602 if (err) {
3603 f2fs_err(sbi, "Failed to initialize post read workqueue");
3604 goto free_devices;
3605 }
3606
3607 sbi->total_valid_node_count =
3608 le32_to_cpu(sbi->ckpt->valid_node_count);
3609 percpu_counter_set(&sbi->total_valid_inode_count,
3610 le32_to_cpu(sbi->ckpt->valid_inode_count));
3611 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3612 sbi->total_valid_block_count =
3613 le64_to_cpu(sbi->ckpt->valid_block_count);
3614 sbi->last_valid_block_count = sbi->total_valid_block_count;
3615 sbi->reserved_blocks = 0;
3616 sbi->current_reserved_blocks = 0;
3617 limit_reserve_root(sbi);
3618 adjust_unusable_cap_perc(sbi);
3619
3620 for (i = 0; i < NR_INODE_TYPE; i++) {
3621 INIT_LIST_HEAD(&sbi->inode_list[i]);
3622 spin_lock_init(&sbi->inode_lock[i]);
3623 }
3624 mutex_init(&sbi->flush_lock);
3625
3626 f2fs_init_extent_cache_info(sbi);
3627
3628 f2fs_init_ino_entry_info(sbi);
3629
3630 f2fs_init_fsync_node_info(sbi);
3631
3632 /* setup f2fs internal modules */
3633 err = f2fs_build_segment_manager(sbi);
3634 if (err) {
3635 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3636 err);
3637 goto free_sm;
3638 }
3639 err = f2fs_build_node_manager(sbi);
3640 if (err) {
3641 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3642 err);
3643 goto free_nm;
3644 }
3645
3646 /* For write statistics */
3647 if (sb->s_bdev->bd_part)
3648 sbi->sectors_written_start =
3649 (u64)part_stat_read(sb->s_bdev->bd_part,
3650 sectors[STAT_WRITE]);
3651
3652 /* Read accumulated write IO statistics if exists */
3653 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3654 if (__exist_node_summaries(sbi))
3655 sbi->kbytes_written =
3656 le64_to_cpu(seg_i->journal->info.kbytes_written);
3657
3658 f2fs_build_gc_manager(sbi);
3659
3660 err = f2fs_build_stats(sbi);
3661 if (err)
3662 goto free_nm;
3663
3664 /* get an inode for node space */
3665 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3666 if (IS_ERR(sbi->node_inode)) {
3667 f2fs_err(sbi, "Failed to read node inode");
3668 err = PTR_ERR(sbi->node_inode);
3669 goto free_stats;
3670 }
3671
3672 /* read root inode and dentry */
3673 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3674 if (IS_ERR(root)) {
3675 f2fs_err(sbi, "Failed to read root inode");
3676 err = PTR_ERR(root);
3677 goto free_node_inode;
3678 }
3679 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3680 !root->i_size || !root->i_nlink) {
3681 iput(root);
3682 err = -EINVAL;
3683 goto free_node_inode;
3684 }
3685
3686 sb->s_root = d_make_root(root); /* allocate root dentry */
3687 if (!sb->s_root) {
3688 err = -ENOMEM;
3689 goto free_node_inode;
3690 }
3691
3692 err = f2fs_register_sysfs(sbi);
3693 if (err)
3694 goto free_root_inode;
3695
3696#ifdef CONFIG_QUOTA
3697 /* Enable quota usage during mount */
3698 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3699 err = f2fs_enable_quotas(sb);
3700 if (err)
3701 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3702 }
3703#endif
3704 /* if there are any orphan inodes, free them */
3705 err = f2fs_recover_orphan_inodes(sbi);
3706 if (err)
3707 goto free_meta;
3708
3709 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3710 goto reset_checkpoint;
3711
3712 /* recover fsynced data */
3713 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3714 !test_opt(sbi, NORECOVERY)) {
3715 /*
3716 * mount should be failed, when device has readonly mode, and
3717 * previous checkpoint was not done by clean system shutdown.
3718 */
3719 if (f2fs_hw_is_readonly(sbi)) {
3720 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3721 err = -EROFS;
3722 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3723 goto free_meta;
3724 }
3725 f2fs_info(sbi, "write access unavailable, skipping recovery");
3726 goto reset_checkpoint;
3727 }
3728
3729 if (need_fsck)
3730 set_sbi_flag(sbi, SBI_NEED_FSCK);
3731
3732 if (skip_recovery)
3733 goto reset_checkpoint;
3734
3735 err = f2fs_recover_fsync_data(sbi, false);
3736 if (err < 0) {
3737 if (err != -ENOMEM)
3738 skip_recovery = true;
3739 need_fsck = true;
3740 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3741 err);
3742 goto free_meta;
3743 }
3744 } else {
3745 err = f2fs_recover_fsync_data(sbi, true);
3746
3747 if (!f2fs_readonly(sb) && err > 0) {
3748 err = -EINVAL;
3749 f2fs_err(sbi, "Need to recover fsync data");
3750 goto free_meta;
3751 }
3752 }
3753
3754 /*
3755 * If the f2fs is not readonly and fsync data recovery succeeds,
3756 * check zoned block devices' write pointer consistency.
3757 */
3758 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3759 err = f2fs_check_write_pointer(sbi);
3760 if (err)
3761 goto free_meta;
3762 }
3763
3764reset_checkpoint:
3765 /* f2fs_recover_fsync_data() cleared this already */
3766 clear_sbi_flag(sbi, SBI_POR_DOING);
3767
3768 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3769 err = f2fs_disable_checkpoint(sbi);
3770 if (err)
3771 goto sync_free_meta;
3772 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3773 f2fs_enable_checkpoint(sbi);
3774 }
3775
3776 /*
3777 * If filesystem is not mounted as read-only then
3778 * do start the gc_thread.
3779 */
3780 if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) {
3781 /* After POR, we can run background GC thread.*/
3782 err = f2fs_start_gc_thread(sbi);
3783 if (err)
3784 goto sync_free_meta;
3785 }
3786 kvfree(options);
3787
3788 /* recover broken superblock */
3789 if (recovery) {
3790 err = f2fs_commit_super(sbi, true);
3791 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3792 sbi->valid_super_block ? 1 : 2, err);
3793 }
3794
3795 f2fs_join_shrinker(sbi);
3796
3797 f2fs_tuning_parameters(sbi);
3798
3799 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3800 cur_cp_version(F2FS_CKPT(sbi)));
3801 f2fs_update_time(sbi, CP_TIME);
3802 f2fs_update_time(sbi, REQ_TIME);
3803 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3804 return 0;
3805
3806sync_free_meta:
3807 /* safe to flush all the data */
3808 sync_filesystem(sbi->sb);
3809 retry_cnt = 0;
3810
3811free_meta:
3812#ifdef CONFIG_QUOTA
3813 f2fs_truncate_quota_inode_pages(sb);
3814 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3815 f2fs_quota_off_umount(sbi->sb);
3816#endif
3817 /*
3818 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3819 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3820 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3821 * falls into an infinite loop in f2fs_sync_meta_pages().
3822 */
3823 truncate_inode_pages_final(META_MAPPING(sbi));
3824 /* evict some inodes being cached by GC */
3825 evict_inodes(sb);
3826 f2fs_unregister_sysfs(sbi);
3827free_root_inode:
3828 dput(sb->s_root);
3829 sb->s_root = NULL;
3830free_node_inode:
3831 f2fs_release_ino_entry(sbi, true);
3832 truncate_inode_pages_final(NODE_MAPPING(sbi));
3833 iput(sbi->node_inode);
3834 sbi->node_inode = NULL;
3835free_stats:
3836 f2fs_destroy_stats(sbi);
3837free_nm:
3838 f2fs_destroy_node_manager(sbi);
3839free_sm:
3840 f2fs_destroy_segment_manager(sbi);
3841 f2fs_destroy_post_read_wq(sbi);
3842free_devices:
3843 destroy_device_list(sbi);
3844 kvfree(sbi->ckpt);
3845free_meta_inode:
3846 make_bad_inode(sbi->meta_inode);
3847 iput(sbi->meta_inode);
3848 sbi->meta_inode = NULL;
3849free_xattr_cache:
3850 f2fs_destroy_xattr_caches(sbi);
3851free_io_dummy:
3852 mempool_destroy(sbi->write_io_dummy);
3853free_percpu:
3854 destroy_percpu_info(sbi);
3855free_bio_info:
3856 for (i = 0; i < NR_PAGE_TYPE; i++)
3857 kvfree(sbi->write_io[i]);
3858
3859#ifdef CONFIG_UNICODE
3860 utf8_unload(sbi->s_encoding);
3861#endif
3862free_options:
3863#ifdef CONFIG_QUOTA
3864 for (i = 0; i < MAXQUOTAS; i++)
3865 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
3866#endif
3867 fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx);
3868 kvfree(options);
3869free_sb_buf:
3870 kfree(raw_super);
3871free_sbi:
3872 if (sbi->s_chksum_driver)
3873 crypto_free_shash(sbi->s_chksum_driver);
3874 kfree(sbi);
3875
3876 /* give only one another chance */
3877 if (retry_cnt > 0 && skip_recovery) {
3878 retry_cnt--;
3879 shrink_dcache_sb(sb);
3880 goto try_onemore;
3881 }
3882 return err;
3883}
3884
3885static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3886 const char *dev_name, void *data)
3887{
3888 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3889}
3890
3891static void kill_f2fs_super(struct super_block *sb)
3892{
3893 if (sb->s_root) {
3894 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3895
3896 set_sbi_flag(sbi, SBI_IS_CLOSE);
3897 f2fs_stop_gc_thread(sbi);
3898 f2fs_stop_discard_thread(sbi);
3899
3900 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3901 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3902 struct cp_control cpc = {
3903 .reason = CP_UMOUNT,
3904 };
3905 f2fs_write_checkpoint(sbi, &cpc);
3906 }
3907
3908 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3909 sb->s_flags &= ~SB_RDONLY;
3910 }
3911 kill_block_super(sb);
3912}
3913
3914static struct file_system_type f2fs_fs_type = {
3915 .owner = THIS_MODULE,
3916 .name = "f2fs",
3917 .mount = f2fs_mount,
3918 .kill_sb = kill_f2fs_super,
3919 .fs_flags = FS_REQUIRES_DEV,
3920};
3921MODULE_ALIAS_FS("f2fs");
3922
3923static int __init init_inodecache(void)
3924{
3925 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3926 sizeof(struct f2fs_inode_info), 0,
3927 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3928 if (!f2fs_inode_cachep)
3929 return -ENOMEM;
3930 return 0;
3931}
3932
3933static void destroy_inodecache(void)
3934{
3935 /*
3936 * Make sure all delayed rcu free inodes are flushed before we
3937 * destroy cache.
3938 */
3939 rcu_barrier();
3940 kmem_cache_destroy(f2fs_inode_cachep);
3941}
3942
3943static int __init init_f2fs_fs(void)
3944{
3945 int err;
3946
3947 if (PAGE_SIZE != F2FS_BLKSIZE) {
3948 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3949 PAGE_SIZE, F2FS_BLKSIZE);
3950 return -EINVAL;
3951 }
3952
3953 f2fs_build_trace_ios();
3954
3955 err = init_inodecache();
3956 if (err)
3957 goto fail;
3958 err = f2fs_create_node_manager_caches();
3959 if (err)
3960 goto free_inodecache;
3961 err = f2fs_create_segment_manager_caches();
3962 if (err)
3963 goto free_node_manager_caches;
3964 err = f2fs_create_checkpoint_caches();
3965 if (err)
3966 goto free_segment_manager_caches;
3967 err = f2fs_create_extent_cache();
3968 if (err)
3969 goto free_checkpoint_caches;
3970 err = f2fs_init_sysfs();
3971 if (err)
3972 goto free_extent_cache;
3973 err = register_shrinker(&f2fs_shrinker_info);
3974 if (err)
3975 goto free_sysfs;
3976 err = register_filesystem(&f2fs_fs_type);
3977 if (err)
3978 goto free_shrinker;
3979 f2fs_create_root_stats();
3980 err = f2fs_init_post_read_processing();
3981 if (err)
3982 goto free_root_stats;
3983 err = f2fs_init_bio_entry_cache();
3984 if (err)
3985 goto free_post_read;
3986 err = f2fs_init_bioset();
3987 if (err)
3988 goto free_bio_enrty_cache;
3989 err = f2fs_init_compress_mempool();
3990 if (err)
3991 goto free_bioset;
3992 return 0;
3993free_bioset:
3994 f2fs_destroy_bioset();
3995free_bio_enrty_cache:
3996 f2fs_destroy_bio_entry_cache();
3997free_post_read:
3998 f2fs_destroy_post_read_processing();
3999free_root_stats:
4000 f2fs_destroy_root_stats();
4001 unregister_filesystem(&f2fs_fs_type);
4002free_shrinker:
4003 unregister_shrinker(&f2fs_shrinker_info);
4004free_sysfs:
4005 f2fs_exit_sysfs();
4006free_extent_cache:
4007 f2fs_destroy_extent_cache();
4008free_checkpoint_caches:
4009 f2fs_destroy_checkpoint_caches();
4010free_segment_manager_caches:
4011 f2fs_destroy_segment_manager_caches();
4012free_node_manager_caches:
4013 f2fs_destroy_node_manager_caches();
4014free_inodecache:
4015 destroy_inodecache();
4016fail:
4017 return err;
4018}
4019
4020static void __exit exit_f2fs_fs(void)
4021{
4022 f2fs_destroy_compress_mempool();
4023 f2fs_destroy_bioset();
4024 f2fs_destroy_bio_entry_cache();
4025 f2fs_destroy_post_read_processing();
4026 f2fs_destroy_root_stats();
4027 unregister_filesystem(&f2fs_fs_type);
4028 unregister_shrinker(&f2fs_shrinker_info);
4029 f2fs_exit_sysfs();
4030 f2fs_destroy_extent_cache();
4031 f2fs_destroy_checkpoint_caches();
4032 f2fs_destroy_segment_manager_caches();
4033 f2fs_destroy_node_manager_caches();
4034 destroy_inodecache();
4035 f2fs_destroy_trace_ios();
4036}
4037
4038module_init(init_f2fs_fs)
4039module_exit(exit_f2fs_fs)
4040
4041MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4042MODULE_DESCRIPTION("Flash Friendly File System");
4043MODULE_LICENSE("GPL");
4044