?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
nvme-rdma: cancel async events before freeing event struct
[linux/fpc-iii.git] / fs / f2fs / super.c
blobfa461db696e79fc0a54ea194b6141387c51b5beb
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
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "gc.h"
33 #include "trace.h"
34
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
37
38 static struct kmem_cache *f2fs_inode_cachep;
39
40 #ifdef CONFIG_F2FS_FAULT_INJECTION
41
42 const char *f2fs_fault_name[FAULT_MAX] = {
43 [FAULT_KMALLOC] = "kmalloc",
44 [FAULT_KVMALLOC] = "kvmalloc",
45 [FAULT_PAGE_ALLOC] = "page alloc",
46 [FAULT_PAGE_GET] = "page get",
47 [FAULT_ALLOC_BIO] = "alloc bio",
48 [FAULT_ALLOC_NID] = "alloc nid",
49 [FAULT_ORPHAN] = "orphan",
50 [FAULT_BLOCK] = "no more block",
51 [FAULT_DIR_DEPTH] = "too big dir depth",
52 [FAULT_EVICT_INODE] = "evict_inode fail",
53 [FAULT_TRUNCATE] = "truncate fail",
54 [FAULT_READ_IO] = "read IO error",
55 [FAULT_CHECKPOINT] = "checkpoint error",
56 [FAULT_DISCARD] = "discard error",
57 [FAULT_WRITE_IO] = "write IO error",
58 };
59
60 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
61 unsigned int type)
62 {
63 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
64
65 if (rate) {
66 atomic_set(&ffi->inject_ops, 0);
67 ffi->inject_rate = rate;
68 }
69
70 if (type)
71 ffi->inject_type = type;
72
73 if (!rate && !type)
74 memset(ffi, 0, sizeof(struct f2fs_fault_info));
75 }
76 #endif
77
78 /* f2fs-wide shrinker description */
79 static struct shrinker f2fs_shrinker_info = {
80 .scan_objects = f2fs_shrink_scan,
81 .count_objects = f2fs_shrink_count,
82 .seeks = DEFAULT_SEEKS,
83 };
84
85 enum {
86 Opt_gc_background,
87 Opt_disable_roll_forward,
88 Opt_norecovery,
89 Opt_discard,
90 Opt_nodiscard,
91 Opt_noheap,
92 Opt_heap,
93 Opt_user_xattr,
94 Opt_nouser_xattr,
95 Opt_acl,
96 Opt_noacl,
97 Opt_active_logs,
98 Opt_disable_ext_identify,
99 Opt_inline_xattr,
100 Opt_noinline_xattr,
101 Opt_inline_xattr_size,
102 Opt_inline_data,
103 Opt_inline_dentry,
104 Opt_noinline_dentry,
105 Opt_flush_merge,
106 Opt_noflush_merge,
107 Opt_nobarrier,
108 Opt_fastboot,
109 Opt_extent_cache,
110 Opt_noextent_cache,
111 Opt_noinline_data,
112 Opt_data_flush,
113 Opt_reserve_root,
114 Opt_resgid,
115 Opt_resuid,
116 Opt_mode,
117 Opt_io_size_bits,
118 Opt_fault_injection,
119 Opt_fault_type,
120 Opt_lazytime,
121 Opt_nolazytime,
122 Opt_quota,
123 Opt_noquota,
124 Opt_usrquota,
125 Opt_grpquota,
126 Opt_prjquota,
127 Opt_usrjquota,
128 Opt_grpjquota,
129 Opt_prjjquota,
130 Opt_offusrjquota,
131 Opt_offgrpjquota,
132 Opt_offprjjquota,
133 Opt_jqfmt_vfsold,
134 Opt_jqfmt_vfsv0,
135 Opt_jqfmt_vfsv1,
136 Opt_whint,
137 Opt_alloc,
138 Opt_fsync,
139 Opt_test_dummy_encryption,
140 Opt_checkpoint_disable,
141 Opt_checkpoint_disable_cap,
142 Opt_checkpoint_disable_cap_perc,
143 Opt_checkpoint_enable,
144 Opt_err,
145 };
146
147 static match_table_t f2fs_tokens = {
148 {Opt_gc_background, "background_gc=%s"},
149 {Opt_disable_roll_forward, "disable_roll_forward"},
150 {Opt_norecovery, "norecovery"},
151 {Opt_discard, "discard"},
152 {Opt_nodiscard, "nodiscard"},
153 {Opt_noheap, "no_heap"},
154 {Opt_heap, "heap"},
155 {Opt_user_xattr, "user_xattr"},
156 {Opt_nouser_xattr, "nouser_xattr"},
157 {Opt_acl, "acl"},
158 {Opt_noacl, "noacl"},
159 {Opt_active_logs, "active_logs=%u"},
160 {Opt_disable_ext_identify, "disable_ext_identify"},
161 {Opt_inline_xattr, "inline_xattr"},
162 {Opt_noinline_xattr, "noinline_xattr"},
163 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
164 {Opt_inline_data, "inline_data"},
165 {Opt_inline_dentry, "inline_dentry"},
166 {Opt_noinline_dentry, "noinline_dentry"},
167 {Opt_flush_merge, "flush_merge"},
168 {Opt_noflush_merge, "noflush_merge"},
169 {Opt_nobarrier, "nobarrier"},
170 {Opt_fastboot, "fastboot"},
171 {Opt_extent_cache, "extent_cache"},
172 {Opt_noextent_cache, "noextent_cache"},
173 {Opt_noinline_data, "noinline_data"},
174 {Opt_data_flush, "data_flush"},
175 {Opt_reserve_root, "reserve_root=%u"},
176 {Opt_resgid, "resgid=%u"},
177 {Opt_resuid, "resuid=%u"},
178 {Opt_mode, "mode=%s"},
179 {Opt_io_size_bits, "io_bits=%u"},
180 {Opt_fault_injection, "fault_injection=%u"},
181 {Opt_fault_type, "fault_type=%u"},
182 {Opt_lazytime, "lazytime"},
183 {Opt_nolazytime, "nolazytime"},
184 {Opt_quota, "quota"},
185 {Opt_noquota, "noquota"},
186 {Opt_usrquota, "usrquota"},
187 {Opt_grpquota, "grpquota"},
188 {Opt_prjquota, "prjquota"},
189 {Opt_usrjquota, "usrjquota=%s"},
190 {Opt_grpjquota, "grpjquota=%s"},
191 {Opt_prjjquota, "prjjquota=%s"},
192 {Opt_offusrjquota, "usrjquota="},
193 {Opt_offgrpjquota, "grpjquota="},
194 {Opt_offprjjquota, "prjjquota="},
195 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
196 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
197 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
198 {Opt_whint, "whint_mode=%s"},
199 {Opt_alloc, "alloc_mode=%s"},
200 {Opt_fsync, "fsync_mode=%s"},
201 {Opt_test_dummy_encryption, "test_dummy_encryption"},
202 {Opt_checkpoint_disable, "checkpoint=disable"},
203 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
204 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
205 {Opt_checkpoint_enable, "checkpoint=enable"},
206 {Opt_err, NULL},
207 };
208
209 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
210 {
211 struct va_format vaf;
212 va_list args;
213 int level;
214
215 va_start(args, fmt);
216
217 level = printk_get_level(fmt);
218 vaf.fmt = printk_skip_level(fmt);
219 vaf.va = &args;
220 printk("%c%cF2FS-fs (%s): %pV\n",
221 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
222
223 va_end(args);
224 }
225
226 #ifdef CONFIG_UNICODE
227 static const struct f2fs_sb_encodings {
228 __u16 magic;
229 char *name;
230 char *version;
231 } f2fs_sb_encoding_map[] = {
232 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
233 };
234
235 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
236 const struct f2fs_sb_encodings **encoding,
237 __u16 *flags)
238 {
239 __u16 magic = le16_to_cpu(sb->s_encoding);
240 int i;
241
242 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
243 if (magic == f2fs_sb_encoding_map[i].magic)
244 break;
245
246 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
247 return -EINVAL;
248
249 *encoding = &f2fs_sb_encoding_map[i];
250 *flags = le16_to_cpu(sb->s_encoding_flags);
251
252 return 0;
253 }
254 #endif
255
256 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
257 {
258 block_t limit = min((sbi->user_block_count << 1) / 1000,
259 sbi->user_block_count - sbi->reserved_blocks);
260
261 /* limit is 0.2% */
262 if (test_opt(sbi, RESERVE_ROOT) &&
263 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
264 F2FS_OPTION(sbi).root_reserved_blocks = limit;
265 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
266 F2FS_OPTION(sbi).root_reserved_blocks);
267 }
268 if (!test_opt(sbi, RESERVE_ROOT) &&
269 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
270 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
271 !gid_eq(F2FS_OPTION(sbi).s_resgid,
272 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
273 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
274 from_kuid_munged(&init_user_ns,
275 F2FS_OPTION(sbi).s_resuid),
276 from_kgid_munged(&init_user_ns,
277 F2FS_OPTION(sbi).s_resgid));
278 }
279
280 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
281 {
282 if (!F2FS_OPTION(sbi).unusable_cap_perc)
283 return;
284
285 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
286 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
287 else
288 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
289 F2FS_OPTION(sbi).unusable_cap_perc;
290
291 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
292 F2FS_OPTION(sbi).unusable_cap,
293 F2FS_OPTION(sbi).unusable_cap_perc);
294 }
295
296 static void init_once(void *foo)
297 {
298 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
299
300 inode_init_once(&fi->vfs_inode);
301 }
302
303 #ifdef CONFIG_QUOTA
304 static const char * const quotatypes[] = INITQFNAMES;
305 #define QTYPE2NAME(t) (quotatypes[t])
306 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
307 substring_t *args)
308 {
309 struct f2fs_sb_info *sbi = F2FS_SB(sb);
310 char *qname;
311 int ret = -EINVAL;
312
313 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
314 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
315 return -EINVAL;
316 }
317 if (f2fs_sb_has_quota_ino(sbi)) {
318 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
319 return 0;
320 }
321
322 qname = match_strdup(args);
323 if (!qname) {
324 f2fs_err(sbi, "Not enough memory for storing quotafile name");
325 return -ENOMEM;
326 }
327 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
328 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
329 ret = 0;
330 else
331 f2fs_err(sbi, "%s quota file already specified",
332 QTYPE2NAME(qtype));
333 goto errout;
334 }
335 if (strchr(qname, '/')) {
336 f2fs_err(sbi, "quotafile must be on filesystem root");
337 goto errout;
338 }
339 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
340 set_opt(sbi, QUOTA);
341 return 0;
342 errout:
343 kvfree(qname);
344 return ret;
345 }
346
347 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
348 {
349 struct f2fs_sb_info *sbi = F2FS_SB(sb);
350
351 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
352 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
353 return -EINVAL;
354 }
355 kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
356 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
357 return 0;
358 }
359
360 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
361 {
362 /*
363 * We do the test below only for project quotas. 'usrquota' and
364 * 'grpquota' mount options are allowed even without quota feature
365 * to support legacy quotas in quota files.
366 */
367 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
368 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
369 return -1;
370 }
371 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
372 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
373 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
374 if (test_opt(sbi, USRQUOTA) &&
375 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
376 clear_opt(sbi, USRQUOTA);
377
378 if (test_opt(sbi, GRPQUOTA) &&
379 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
380 clear_opt(sbi, GRPQUOTA);
381
382 if (test_opt(sbi, PRJQUOTA) &&
383 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
384 clear_opt(sbi, PRJQUOTA);
385
386 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
387 test_opt(sbi, PRJQUOTA)) {
388 f2fs_err(sbi, "old and new quota format mixing");
389 return -1;
390 }
391
392 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
393 f2fs_err(sbi, "journaled quota format not specified");
394 return -1;
395 }
396 }
397
398 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
399 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
400 F2FS_OPTION(sbi).s_jquota_fmt = 0;
401 }
402 return 0;
403 }
404 #endif
405
406 static int parse_options(struct super_block *sb, char *options)
407 {
408 struct f2fs_sb_info *sbi = F2FS_SB(sb);
409 substring_t args[MAX_OPT_ARGS];
410 char *p, *name;
411 int arg = 0;
412 kuid_t uid;
413 kgid_t gid;
414 #ifdef CONFIG_QUOTA
415 int ret;
416 #endif
417
418 if (!options)
419 return 0;
420
421 while ((p = strsep(&options, ",")) != NULL) {
422 int token;
423 if (!*p)
424 continue;
425 /*
426 * Initialize args struct so we know whether arg was
427 * found; some options take optional arguments.
428 */
429 args[0].to = args[0].from = NULL;
430 token = match_token(p, f2fs_tokens, args);
431
432 switch (token) {
433 case Opt_gc_background:
434 name = match_strdup(&args[0]);
435
436 if (!name)
437 return -ENOMEM;
438 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
439 set_opt(sbi, BG_GC);
440 clear_opt(sbi, FORCE_FG_GC);
441 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
442 clear_opt(sbi, BG_GC);
443 clear_opt(sbi, FORCE_FG_GC);
444 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
445 set_opt(sbi, BG_GC);
446 set_opt(sbi, FORCE_FG_GC);
447 } else {
448 kvfree(name);
449 return -EINVAL;
450 }
451 kvfree(name);
452 break;
453 case Opt_disable_roll_forward:
454 set_opt(sbi, DISABLE_ROLL_FORWARD);
455 break;
456 case Opt_norecovery:
457 /* this option mounts f2fs with ro */
458 set_opt(sbi, NORECOVERY);
459 if (!f2fs_readonly(sb))
460 return -EINVAL;
461 break;
462 case Opt_discard:
463 set_opt(sbi, DISCARD);
464 break;
465 case Opt_nodiscard:
466 if (f2fs_sb_has_blkzoned(sbi)) {
467 f2fs_warn(sbi, "discard is required for zoned block devices");
468 return -EINVAL;
469 }
470 clear_opt(sbi, DISCARD);
471 break;
472 case Opt_noheap:
473 set_opt(sbi, NOHEAP);
474 break;
475 case Opt_heap:
476 clear_opt(sbi, NOHEAP);
477 break;
478 #ifdef CONFIG_F2FS_FS_XATTR
479 case Opt_user_xattr:
480 set_opt(sbi, XATTR_USER);
481 break;
482 case Opt_nouser_xattr:
483 clear_opt(sbi, XATTR_USER);
484 break;
485 case Opt_inline_xattr:
486 set_opt(sbi, INLINE_XATTR);
487 break;
488 case Opt_noinline_xattr:
489 clear_opt(sbi, INLINE_XATTR);
490 break;
491 case Opt_inline_xattr_size:
492 if (args->from && match_int(args, &arg))
493 return -EINVAL;
494 set_opt(sbi, INLINE_XATTR_SIZE);
495 F2FS_OPTION(sbi).inline_xattr_size = arg;
496 break;
497 #else
498 case Opt_user_xattr:
499 f2fs_info(sbi, "user_xattr options not supported");
500 break;
501 case Opt_nouser_xattr:
502 f2fs_info(sbi, "nouser_xattr options not supported");
503 break;
504 case Opt_inline_xattr:
505 f2fs_info(sbi, "inline_xattr options not supported");
506 break;
507 case Opt_noinline_xattr:
508 f2fs_info(sbi, "noinline_xattr options not supported");
509 break;
510 #endif
511 #ifdef CONFIG_F2FS_FS_POSIX_ACL
512 case Opt_acl:
513 set_opt(sbi, POSIX_ACL);
514 break;
515 case Opt_noacl:
516 clear_opt(sbi, POSIX_ACL);
517 break;
518 #else
519 case Opt_acl:
520 f2fs_info(sbi, "acl options not supported");
521 break;
522 case Opt_noacl:
523 f2fs_info(sbi, "noacl options not supported");
524 break;
525 #endif
526 case Opt_active_logs:
527 if (args->from && match_int(args, &arg))
528 return -EINVAL;
529 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
530 return -EINVAL;
531 F2FS_OPTION(sbi).active_logs = arg;
532 break;
533 case Opt_disable_ext_identify:
534 set_opt(sbi, DISABLE_EXT_IDENTIFY);
535 break;
536 case Opt_inline_data:
537 set_opt(sbi, INLINE_DATA);
538 break;
539 case Opt_inline_dentry:
540 set_opt(sbi, INLINE_DENTRY);
541 break;
542 case Opt_noinline_dentry:
543 clear_opt(sbi, INLINE_DENTRY);
544 break;
545 case Opt_flush_merge:
546 set_opt(sbi, FLUSH_MERGE);
547 break;
548 case Opt_noflush_merge:
549 clear_opt(sbi, FLUSH_MERGE);
550 break;
551 case Opt_nobarrier:
552 set_opt(sbi, NOBARRIER);
553 break;
554 case Opt_fastboot:
555 set_opt(sbi, FASTBOOT);
556 break;
557 case Opt_extent_cache:
558 set_opt(sbi, EXTENT_CACHE);
559 break;
560 case Opt_noextent_cache:
561 clear_opt(sbi, EXTENT_CACHE);
562 break;
563 case Opt_noinline_data:
564 clear_opt(sbi, INLINE_DATA);
565 break;
566 case Opt_data_flush:
567 set_opt(sbi, DATA_FLUSH);
568 break;
569 case Opt_reserve_root:
570 if (args->from && match_int(args, &arg))
571 return -EINVAL;
572 if (test_opt(sbi, RESERVE_ROOT)) {
573 f2fs_info(sbi, "Preserve previous reserve_root=%u",
574 F2FS_OPTION(sbi).root_reserved_blocks);
575 } else {
576 F2FS_OPTION(sbi).root_reserved_blocks = arg;
577 set_opt(sbi, RESERVE_ROOT);
578 }
579 break;
580 case Opt_resuid:
581 if (args->from && match_int(args, &arg))
582 return -EINVAL;
583 uid = make_kuid(current_user_ns(), arg);
584 if (!uid_valid(uid)) {
585 f2fs_err(sbi, "Invalid uid value %d", arg);
586 return -EINVAL;
587 }
588 F2FS_OPTION(sbi).s_resuid = uid;
589 break;
590 case Opt_resgid:
591 if (args->from && match_int(args, &arg))
592 return -EINVAL;
593 gid = make_kgid(current_user_ns(), arg);
594 if (!gid_valid(gid)) {
595 f2fs_err(sbi, "Invalid gid value %d", arg);
596 return -EINVAL;
597 }
598 F2FS_OPTION(sbi).s_resgid = gid;
599 break;
600 case Opt_mode:
601 name = match_strdup(&args[0]);
602
603 if (!name)
604 return -ENOMEM;
605 if (strlen(name) == 8 &&
606 !strncmp(name, "adaptive", 8)) {
607 if (f2fs_sb_has_blkzoned(sbi)) {
608 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
609 kvfree(name);
610 return -EINVAL;
611 }
612 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
613 } else if (strlen(name) == 3 &&
614 !strncmp(name, "lfs", 3)) {
615 set_opt_mode(sbi, F2FS_MOUNT_LFS);
616 } else {
617 kvfree(name);
618 return -EINVAL;
619 }
620 kvfree(name);
621 break;
622 case Opt_io_size_bits:
623 if (args->from && match_int(args, &arg))
624 return -EINVAL;
625 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
626 f2fs_warn(sbi, "Not support %d, larger than %d",
627 1 << arg, BIO_MAX_PAGES);
628 return -EINVAL;
629 }
630 F2FS_OPTION(sbi).write_io_size_bits = arg;
631 break;
632 #ifdef CONFIG_F2FS_FAULT_INJECTION
633 case Opt_fault_injection:
634 if (args->from && match_int(args, &arg))
635 return -EINVAL;
636 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
637 set_opt(sbi, FAULT_INJECTION);
638 break;
639
640 case Opt_fault_type:
641 if (args->from && match_int(args, &arg))
642 return -EINVAL;
643 f2fs_build_fault_attr(sbi, 0, arg);
644 set_opt(sbi, FAULT_INJECTION);
645 break;
646 #else
647 case Opt_fault_injection:
648 f2fs_info(sbi, "fault_injection options not supported");
649 break;
650
651 case Opt_fault_type:
652 f2fs_info(sbi, "fault_type options not supported");
653 break;
654 #endif
655 case Opt_lazytime:
656 sb->s_flags |= SB_LAZYTIME;
657 break;
658 case Opt_nolazytime:
659 sb->s_flags &= ~SB_LAZYTIME;
660 break;
661 #ifdef CONFIG_QUOTA
662 case Opt_quota:
663 case Opt_usrquota:
664 set_opt(sbi, USRQUOTA);
665 break;
666 case Opt_grpquota:
667 set_opt(sbi, GRPQUOTA);
668 break;
669 case Opt_prjquota:
670 set_opt(sbi, PRJQUOTA);
671 break;
672 case Opt_usrjquota:
673 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
674 if (ret)
675 return ret;
676 break;
677 case Opt_grpjquota:
678 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
679 if (ret)
680 return ret;
681 break;
682 case Opt_prjjquota:
683 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
684 if (ret)
685 return ret;
686 break;
687 case Opt_offusrjquota:
688 ret = f2fs_clear_qf_name(sb, USRQUOTA);
689 if (ret)
690 return ret;
691 break;
692 case Opt_offgrpjquota:
693 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
694 if (ret)
695 return ret;
696 break;
697 case Opt_offprjjquota:
698 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
699 if (ret)
700 return ret;
701 break;
702 case Opt_jqfmt_vfsold:
703 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
704 break;
705 case Opt_jqfmt_vfsv0:
706 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
707 break;
708 case Opt_jqfmt_vfsv1:
709 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
710 break;
711 case Opt_noquota:
712 clear_opt(sbi, QUOTA);
713 clear_opt(sbi, USRQUOTA);
714 clear_opt(sbi, GRPQUOTA);
715 clear_opt(sbi, PRJQUOTA);
716 break;
717 #else
718 case Opt_quota:
719 case Opt_usrquota:
720 case Opt_grpquota:
721 case Opt_prjquota:
722 case Opt_usrjquota:
723 case Opt_grpjquota:
724 case Opt_prjjquota:
725 case Opt_offusrjquota:
726 case Opt_offgrpjquota:
727 case Opt_offprjjquota:
728 case Opt_jqfmt_vfsold:
729 case Opt_jqfmt_vfsv0:
730 case Opt_jqfmt_vfsv1:
731 case Opt_noquota:
732 f2fs_info(sbi, "quota operations not supported");
733 break;
734 #endif
735 case Opt_whint:
736 name = match_strdup(&args[0]);
737 if (!name)
738 return -ENOMEM;
739 if (strlen(name) == 10 &&
740 !strncmp(name, "user-based", 10)) {
741 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
742 } else if (strlen(name) == 3 &&
743 !strncmp(name, "off", 3)) {
744 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
745 } else if (strlen(name) == 8 &&
746 !strncmp(name, "fs-based", 8)) {
747 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
748 } else {
749 kvfree(name);
750 return -EINVAL;
751 }
752 kvfree(name);
753 break;
754 case Opt_alloc:
755 name = match_strdup(&args[0]);
756 if (!name)
757 return -ENOMEM;
758
759 if (strlen(name) == 7 &&
760 !strncmp(name, "default", 7)) {
761 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
762 } else if (strlen(name) == 5 &&
763 !strncmp(name, "reuse", 5)) {
764 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
765 } else {
766 kvfree(name);
767 return -EINVAL;
768 }
769 kvfree(name);
770 break;
771 case Opt_fsync:
772 name = match_strdup(&args[0]);
773 if (!name)
774 return -ENOMEM;
775 if (strlen(name) == 5 &&
776 !strncmp(name, "posix", 5)) {
777 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
778 } else if (strlen(name) == 6 &&
779 !strncmp(name, "strict", 6)) {
780 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
781 } else if (strlen(name) == 9 &&
782 !strncmp(name, "nobarrier", 9)) {
783 F2FS_OPTION(sbi).fsync_mode =
784 FSYNC_MODE_NOBARRIER;
785 } else {
786 kvfree(name);
787 return -EINVAL;
788 }
789 kvfree(name);
790 break;
791 case Opt_test_dummy_encryption:
792 #ifdef CONFIG_FS_ENCRYPTION
793 if (!f2fs_sb_has_encrypt(sbi)) {
794 f2fs_err(sbi, "Encrypt feature is off");
795 return -EINVAL;
796 }
797
798 F2FS_OPTION(sbi).test_dummy_encryption = true;
799 f2fs_info(sbi, "Test dummy encryption mode enabled");
800 #else
801 f2fs_info(sbi, "Test dummy encryption mount option ignored");
802 #endif
803 break;
804 case Opt_checkpoint_disable_cap_perc:
805 if (args->from && match_int(args, &arg))
806 return -EINVAL;
807 if (arg < 0 || arg > 100)
808 return -EINVAL;
809 F2FS_OPTION(sbi).unusable_cap_perc = arg;
810 set_opt(sbi, DISABLE_CHECKPOINT);
811 break;
812 case Opt_checkpoint_disable_cap:
813 if (args->from && match_int(args, &arg))
814 return -EINVAL;
815 F2FS_OPTION(sbi).unusable_cap = arg;
816 set_opt(sbi, DISABLE_CHECKPOINT);
817 break;
818 case Opt_checkpoint_disable:
819 set_opt(sbi, DISABLE_CHECKPOINT);
820 break;
821 case Opt_checkpoint_enable:
822 clear_opt(sbi, DISABLE_CHECKPOINT);
823 break;
824 default:
825 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
826 p);
827 return -EINVAL;
828 }
829 }
830 #ifdef CONFIG_QUOTA
831 if (f2fs_check_quota_options(sbi))
832 return -EINVAL;
833 #else
834 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
835 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
836 return -EINVAL;
837 }
838 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
839 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
840 return -EINVAL;
841 }
842 #endif
843 #ifndef CONFIG_UNICODE
844 if (f2fs_sb_has_casefold(sbi)) {
845 f2fs_err(sbi,
846 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
847 return -EINVAL;
848 }
849 #endif
850
851 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
852 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
853 F2FS_IO_SIZE_KB(sbi));
854 return -EINVAL;
855 }
856
857 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
858 int min_size, max_size;
859
860 if (!f2fs_sb_has_extra_attr(sbi) ||
861 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
862 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
863 return -EINVAL;
864 }
865 if (!test_opt(sbi, INLINE_XATTR)) {
866 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
867 return -EINVAL;
868 }
869
870 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
871 max_size = MAX_INLINE_XATTR_SIZE;
872
873 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
874 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
875 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
876 min_size, max_size);
877 return -EINVAL;
878 }
879 }
880
881 if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
882 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
883 return -EINVAL;
884 }
885
886 /* Not pass down write hints if the number of active logs is lesser
887 * than NR_CURSEG_TYPE.
888 */
889 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
890 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
891 return 0;
892 }
893
894 static struct inode *f2fs_alloc_inode(struct super_block *sb)
895 {
896 struct f2fs_inode_info *fi;
897
898 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
899 if (!fi)
900 return NULL;
901
902 init_once((void *) fi);
903
904 /* Initialize f2fs-specific inode info */
905 atomic_set(&fi->dirty_pages, 0);
906 init_rwsem(&fi->i_sem);
907 INIT_LIST_HEAD(&fi->dirty_list);
908 INIT_LIST_HEAD(&fi->gdirty_list);
909 INIT_LIST_HEAD(&fi->inmem_ilist);
910 INIT_LIST_HEAD(&fi->inmem_pages);
911 mutex_init(&fi->inmem_lock);
912 init_rwsem(&fi->i_gc_rwsem[READ]);
913 init_rwsem(&fi->i_gc_rwsem[WRITE]);
914 init_rwsem(&fi->i_mmap_sem);
915 init_rwsem(&fi->i_xattr_sem);
916
917 /* Will be used by directory only */
918 fi->i_dir_level = F2FS_SB(sb)->dir_level;
919
920 return &fi->vfs_inode;
921 }
922
923 static int f2fs_drop_inode(struct inode *inode)
924 {
925 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
926 int ret;
927
928 /*
929 * during filesystem shutdown, if checkpoint is disabled,
930 * drop useless meta/node dirty pages.
931 */
932 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
933 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
934 inode->i_ino == F2FS_META_INO(sbi)) {
935 trace_f2fs_drop_inode(inode, 1);
936 return 1;
937 }
938 }
939
940 /*
941 * This is to avoid a deadlock condition like below.
942 * writeback_single_inode(inode)
943 * - f2fs_write_data_page
944 * - f2fs_gc -> iput -> evict
945 * - inode_wait_for_writeback(inode)
946 */
947 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
948 if (!inode->i_nlink && !is_bad_inode(inode)) {
949 /* to avoid evict_inode call simultaneously */
950 atomic_inc(&inode->i_count);
951 spin_unlock(&inode->i_lock);
952
953 /* some remained atomic pages should discarded */
954 if (f2fs_is_atomic_file(inode))
955 f2fs_drop_inmem_pages(inode);
956
957 /* should remain fi->extent_tree for writepage */
958 f2fs_destroy_extent_node(inode);
959
960 sb_start_intwrite(inode->i_sb);
961 f2fs_i_size_write(inode, 0);
962
963 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
964 inode, NULL, 0, DATA);
965 truncate_inode_pages_final(inode->i_mapping);
966
967 if (F2FS_HAS_BLOCKS(inode))
968 f2fs_truncate(inode);
969
970 sb_end_intwrite(inode->i_sb);
971
972 spin_lock(&inode->i_lock);
973 atomic_dec(&inode->i_count);
974 }
975 trace_f2fs_drop_inode(inode, 0);
976 return 0;
977 }
978 ret = generic_drop_inode(inode);
979 if (!ret)
980 ret = fscrypt_drop_inode(inode);
981 trace_f2fs_drop_inode(inode, ret);
982 return ret;
983 }
984
985 int f2fs_inode_dirtied(struct inode *inode, bool sync)
986 {
987 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
988 int ret = 0;
989
990 spin_lock(&sbi->inode_lock[DIRTY_META]);
991 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
992 ret = 1;
993 } else {
994 set_inode_flag(inode, FI_DIRTY_INODE);
995 stat_inc_dirty_inode(sbi, DIRTY_META);
996 }
997 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
998 list_add_tail(&F2FS_I(inode)->gdirty_list,
999 &sbi->inode_list[DIRTY_META]);
1000 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1001 }
1002 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1003 return ret;
1004 }
1005
1006 void f2fs_inode_synced(struct inode *inode)
1007 {
1008 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1009
1010 spin_lock(&sbi->inode_lock[DIRTY_META]);
1011 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1012 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1013 return;
1014 }
1015 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1016 list_del_init(&F2FS_I(inode)->gdirty_list);
1017 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1018 }
1019 clear_inode_flag(inode, FI_DIRTY_INODE);
1020 clear_inode_flag(inode, FI_AUTO_RECOVER);
1021 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1022 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1023 }
1024
1025 /*
1026 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1027 *
1028 * We should call set_dirty_inode to write the dirty inode through write_inode.
1029 */
1030 static void f2fs_dirty_inode(struct inode *inode, int flags)
1031 {
1032 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1033
1034 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1035 inode->i_ino == F2FS_META_INO(sbi))
1036 return;
1037
1038 if (flags == I_DIRTY_TIME)
1039 return;
1040
1041 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1042 clear_inode_flag(inode, FI_AUTO_RECOVER);
1043
1044 f2fs_inode_dirtied(inode, false);
1045 }
1046
1047 static void f2fs_free_inode(struct inode *inode)
1048 {
1049 fscrypt_free_inode(inode);
1050 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1051 }
1052
1053 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1054 {
1055 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1056 percpu_counter_destroy(&sbi->total_valid_inode_count);
1057 }
1058
1059 static void destroy_device_list(struct f2fs_sb_info *sbi)
1060 {
1061 int i;
1062
1063 for (i = 0; i < sbi->s_ndevs; i++) {
1064 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1065 #ifdef CONFIG_BLK_DEV_ZONED
1066 kvfree(FDEV(i).blkz_seq);
1067 #endif
1068 }
1069 kvfree(sbi->devs);
1070 }
1071
1072 static void f2fs_put_super(struct super_block *sb)
1073 {
1074 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1075 int i;
1076 bool dropped;
1077
1078 /* unregister procfs/sysfs entries in advance to avoid race case */
1079 f2fs_unregister_sysfs(sbi);
1080
1081 f2fs_quota_off_umount(sb);
1082
1083 /* prevent remaining shrinker jobs */
1084 mutex_lock(&sbi->umount_mutex);
1085
1086 /*
1087 * We don't need to do checkpoint when superblock is clean.
1088 * But, the previous checkpoint was not done by umount, it needs to do
1089 * clean checkpoint again.
1090 */
1091 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1092 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1093 struct cp_control cpc = {
1094 .reason = CP_UMOUNT,
1095 };
1096 f2fs_write_checkpoint(sbi, &cpc);
1097 }
1098
1099 /* be sure to wait for any on-going discard commands */
1100 dropped = f2fs_issue_discard_timeout(sbi);
1101
1102 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1103 !sbi->discard_blks && !dropped) {
1104 struct cp_control cpc = {
1105 .reason = CP_UMOUNT | CP_TRIMMED,
1106 };
1107 f2fs_write_checkpoint(sbi, &cpc);
1108 }
1109
1110 /*
1111 * normally superblock is clean, so we need to release this.
1112 * In addition, EIO will skip do checkpoint, we need this as well.
1113 */
1114 f2fs_release_ino_entry(sbi, true);
1115
1116 f2fs_leave_shrinker(sbi);
1117 mutex_unlock(&sbi->umount_mutex);
1118
1119 /* our cp_error case, we can wait for any writeback page */
1120 f2fs_flush_merged_writes(sbi);
1121
1122 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1123
1124 f2fs_bug_on(sbi, sbi->fsync_node_num);
1125
1126 iput(sbi->node_inode);
1127 sbi->node_inode = NULL;
1128
1129 iput(sbi->meta_inode);
1130 sbi->meta_inode = NULL;
1131
1132 /*
1133 * iput() can update stat information, if f2fs_write_checkpoint()
1134 * above failed with error.
1135 */
1136 f2fs_destroy_stats(sbi);
1137
1138 /* destroy f2fs internal modules */
1139 f2fs_destroy_node_manager(sbi);
1140 f2fs_destroy_segment_manager(sbi);
1141
1142 kvfree(sbi->ckpt);
1143
1144 sb->s_fs_info = NULL;
1145 if (sbi->s_chksum_driver)
1146 crypto_free_shash(sbi->s_chksum_driver);
1147 kvfree(sbi->raw_super);
1148
1149 destroy_device_list(sbi);
1150 mempool_destroy(sbi->write_io_dummy);
1151 #ifdef CONFIG_QUOTA
1152 for (i = 0; i < MAXQUOTAS; i++)
1153 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1154 #endif
1155 destroy_percpu_info(sbi);
1156 for (i = 0; i < NR_PAGE_TYPE; i++)
1157 kvfree(sbi->write_io[i]);
1158 #ifdef CONFIG_UNICODE
1159 utf8_unload(sbi->s_encoding);
1160 #endif
1161 kvfree(sbi);
1162 }
1163
1164 int f2fs_sync_fs(struct super_block *sb, int sync)
1165 {
1166 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1167 int err = 0;
1168
1169 if (unlikely(f2fs_cp_error(sbi)))
1170 return 0;
1171 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1172 return 0;
1173
1174 trace_f2fs_sync_fs(sb, sync);
1175
1176 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1177 return -EAGAIN;
1178
1179 if (sync) {
1180 struct cp_control cpc;
1181
1182 cpc.reason = __get_cp_reason(sbi);
1183
1184 mutex_lock(&sbi->gc_mutex);
1185 err = f2fs_write_checkpoint(sbi, &cpc);
1186 mutex_unlock(&sbi->gc_mutex);
1187 }
1188 f2fs_trace_ios(NULL, 1);
1189
1190 return err;
1191 }
1192
1193 static int f2fs_freeze(struct super_block *sb)
1194 {
1195 if (f2fs_readonly(sb))
1196 return 0;
1197
1198 /* IO error happened before */
1199 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1200 return -EIO;
1201
1202 /* must be clean, since sync_filesystem() was already called */
1203 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1204 return -EINVAL;
1205 return 0;
1206 }
1207
1208 static int f2fs_unfreeze(struct super_block *sb)
1209 {
1210 return 0;
1211 }
1212
1213 #ifdef CONFIG_QUOTA
1214 static int f2fs_statfs_project(struct super_block *sb,
1215 kprojid_t projid, struct kstatfs *buf)
1216 {
1217 struct kqid qid;
1218 struct dquot *dquot;
1219 u64 limit;
1220 u64 curblock;
1221
1222 qid = make_kqid_projid(projid);
1223 dquot = dqget(sb, qid);
1224 if (IS_ERR(dquot))
1225 return PTR_ERR(dquot);
1226 spin_lock(&dquot->dq_dqb_lock);
1227
1228 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1229 dquot->dq_dqb.dqb_bhardlimit);
1230 if (limit)
1231 limit >>= sb->s_blocksize_bits;
1232
1233 if (limit && buf->f_blocks > limit) {
1234 curblock = (dquot->dq_dqb.dqb_curspace +
1235 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1236 buf->f_blocks = limit;
1237 buf->f_bfree = buf->f_bavail =
1238 (buf->f_blocks > curblock) ?
1239 (buf->f_blocks - curblock) : 0;
1240 }
1241
1242 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1243 dquot->dq_dqb.dqb_ihardlimit);
1244
1245 if (limit && buf->f_files > limit) {
1246 buf->f_files = limit;
1247 buf->f_ffree =
1248 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1249 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1250 }
1251
1252 spin_unlock(&dquot->dq_dqb_lock);
1253 dqput(dquot);
1254 return 0;
1255 }
1256 #endif
1257
1258 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1259 {
1260 struct super_block *sb = dentry->d_sb;
1261 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1262 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1263 block_t total_count, user_block_count, start_count;
1264 u64 avail_node_count;
1265
1266 total_count = le64_to_cpu(sbi->raw_super->block_count);
1267 user_block_count = sbi->user_block_count;
1268 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1269 buf->f_type = F2FS_SUPER_MAGIC;
1270 buf->f_bsize = sbi->blocksize;
1271
1272 buf->f_blocks = total_count - start_count;
1273 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1274 sbi->current_reserved_blocks;
1275
1276 spin_lock(&sbi->stat_lock);
1277 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1278 buf->f_bfree = 0;
1279 else
1280 buf->f_bfree -= sbi->unusable_block_count;
1281 spin_unlock(&sbi->stat_lock);
1282
1283 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1284 buf->f_bavail = buf->f_bfree -
1285 F2FS_OPTION(sbi).root_reserved_blocks;
1286 else
1287 buf->f_bavail = 0;
1288
1289 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1290
1291 if (avail_node_count > user_block_count) {
1292 buf->f_files = user_block_count;
1293 buf->f_ffree = buf->f_bavail;
1294 } else {
1295 buf->f_files = avail_node_count;
1296 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1297 buf->f_bavail);
1298 }
1299
1300 buf->f_namelen = F2FS_NAME_LEN;
1301 buf->f_fsid.val[0] = (u32)id;
1302 buf->f_fsid.val[1] = (u32)(id >> 32);
1303
1304 #ifdef CONFIG_QUOTA
1305 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1306 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1307 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1308 }
1309 #endif
1310 return 0;
1311 }
1312
1313 static inline void f2fs_show_quota_options(struct seq_file *seq,
1314 struct super_block *sb)
1315 {
1316 #ifdef CONFIG_QUOTA
1317 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1318
1319 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1320 char *fmtname = "";
1321
1322 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1323 case QFMT_VFS_OLD:
1324 fmtname = "vfsold";
1325 break;
1326 case QFMT_VFS_V0:
1327 fmtname = "vfsv0";
1328 break;
1329 case QFMT_VFS_V1:
1330 fmtname = "vfsv1";
1331 break;
1332 }
1333 seq_printf(seq, ",jqfmt=%s", fmtname);
1334 }
1335
1336 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1337 seq_show_option(seq, "usrjquota",
1338 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1339
1340 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1341 seq_show_option(seq, "grpjquota",
1342 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1343
1344 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1345 seq_show_option(seq, "prjjquota",
1346 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1347 #endif
1348 }
1349
1350 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1351 {
1352 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1353
1354 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1355 if (test_opt(sbi, FORCE_FG_GC))
1356 seq_printf(seq, ",background_gc=%s", "sync");
1357 else
1358 seq_printf(seq, ",background_gc=%s", "on");
1359 } else {
1360 seq_printf(seq, ",background_gc=%s", "off");
1361 }
1362 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1363 seq_puts(seq, ",disable_roll_forward");
1364 if (test_opt(sbi, NORECOVERY))
1365 seq_puts(seq, ",norecovery");
1366 if (test_opt(sbi, DISCARD))
1367 seq_puts(seq, ",discard");
1368 else
1369 seq_puts(seq, ",nodiscard");
1370 if (test_opt(sbi, NOHEAP))
1371 seq_puts(seq, ",no_heap");
1372 else
1373 seq_puts(seq, ",heap");
1374 #ifdef CONFIG_F2FS_FS_XATTR
1375 if (test_opt(sbi, XATTR_USER))
1376 seq_puts(seq, ",user_xattr");
1377 else
1378 seq_puts(seq, ",nouser_xattr");
1379 if (test_opt(sbi, INLINE_XATTR))
1380 seq_puts(seq, ",inline_xattr");
1381 else
1382 seq_puts(seq, ",noinline_xattr");
1383 if (test_opt(sbi, INLINE_XATTR_SIZE))
1384 seq_printf(seq, ",inline_xattr_size=%u",
1385 F2FS_OPTION(sbi).inline_xattr_size);
1386 #endif
1387 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1388 if (test_opt(sbi, POSIX_ACL))
1389 seq_puts(seq, ",acl");
1390 else
1391 seq_puts(seq, ",noacl");
1392 #endif
1393 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1394 seq_puts(seq, ",disable_ext_identify");
1395 if (test_opt(sbi, INLINE_DATA))
1396 seq_puts(seq, ",inline_data");
1397 else
1398 seq_puts(seq, ",noinline_data");
1399 if (test_opt(sbi, INLINE_DENTRY))
1400 seq_puts(seq, ",inline_dentry");
1401 else
1402 seq_puts(seq, ",noinline_dentry");
1403 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1404 seq_puts(seq, ",flush_merge");
1405 if (test_opt(sbi, NOBARRIER))
1406 seq_puts(seq, ",nobarrier");
1407 if (test_opt(sbi, FASTBOOT))
1408 seq_puts(seq, ",fastboot");
1409 if (test_opt(sbi, EXTENT_CACHE))
1410 seq_puts(seq, ",extent_cache");
1411 else
1412 seq_puts(seq, ",noextent_cache");
1413 if (test_opt(sbi, DATA_FLUSH))
1414 seq_puts(seq, ",data_flush");
1415
1416 seq_puts(seq, ",mode=");
1417 if (test_opt(sbi, ADAPTIVE))
1418 seq_puts(seq, "adaptive");
1419 else if (test_opt(sbi, LFS))
1420 seq_puts(seq, "lfs");
1421 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1422 if (test_opt(sbi, RESERVE_ROOT))
1423 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1424 F2FS_OPTION(sbi).root_reserved_blocks,
1425 from_kuid_munged(&init_user_ns,
1426 F2FS_OPTION(sbi).s_resuid),
1427 from_kgid_munged(&init_user_ns,
1428 F2FS_OPTION(sbi).s_resgid));
1429 if (F2FS_IO_SIZE_BITS(sbi))
1430 seq_printf(seq, ",io_bits=%u",
1431 F2FS_OPTION(sbi).write_io_size_bits);
1432 #ifdef CONFIG_F2FS_FAULT_INJECTION
1433 if (test_opt(sbi, FAULT_INJECTION)) {
1434 seq_printf(seq, ",fault_injection=%u",
1435 F2FS_OPTION(sbi).fault_info.inject_rate);
1436 seq_printf(seq, ",fault_type=%u",
1437 F2FS_OPTION(sbi).fault_info.inject_type);
1438 }
1439 #endif
1440 #ifdef CONFIG_QUOTA
1441 if (test_opt(sbi, QUOTA))
1442 seq_puts(seq, ",quota");
1443 if (test_opt(sbi, USRQUOTA))
1444 seq_puts(seq, ",usrquota");
1445 if (test_opt(sbi, GRPQUOTA))
1446 seq_puts(seq, ",grpquota");
1447 if (test_opt(sbi, PRJQUOTA))
1448 seq_puts(seq, ",prjquota");
1449 #endif
1450 f2fs_show_quota_options(seq, sbi->sb);
1451 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1452 seq_printf(seq, ",whint_mode=%s", "user-based");
1453 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1454 seq_printf(seq, ",whint_mode=%s", "fs-based");
1455 #ifdef CONFIG_FS_ENCRYPTION
1456 if (F2FS_OPTION(sbi).test_dummy_encryption)
1457 seq_puts(seq, ",test_dummy_encryption");
1458 #endif
1459
1460 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1461 seq_printf(seq, ",alloc_mode=%s", "default");
1462 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1463 seq_printf(seq, ",alloc_mode=%s", "reuse");
1464
1465 if (test_opt(sbi, DISABLE_CHECKPOINT))
1466 seq_printf(seq, ",checkpoint=disable:%u",
1467 F2FS_OPTION(sbi).unusable_cap);
1468 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1469 seq_printf(seq, ",fsync_mode=%s", "posix");
1470 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1471 seq_printf(seq, ",fsync_mode=%s", "strict");
1472 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1473 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1474 return 0;
1475 }
1476
1477 static void default_options(struct f2fs_sb_info *sbi)
1478 {
1479 /* init some FS parameters */
1480 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1481 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1482 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1483 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1484 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1485 F2FS_OPTION(sbi).test_dummy_encryption = false;
1486 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1487 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1488
1489 set_opt(sbi, BG_GC);
1490 set_opt(sbi, INLINE_XATTR);
1491 set_opt(sbi, INLINE_DATA);
1492 set_opt(sbi, INLINE_DENTRY);
1493 set_opt(sbi, EXTENT_CACHE);
1494 set_opt(sbi, NOHEAP);
1495 clear_opt(sbi, DISABLE_CHECKPOINT);
1496 F2FS_OPTION(sbi).unusable_cap = 0;
1497 sbi->sb->s_flags |= SB_LAZYTIME;
1498 set_opt(sbi, FLUSH_MERGE);
1499 set_opt(sbi, DISCARD);
1500 if (f2fs_sb_has_blkzoned(sbi))
1501 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1502 else
1503 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1504
1505 #ifdef CONFIG_F2FS_FS_XATTR
1506 set_opt(sbi, XATTR_USER);
1507 #endif
1508 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1509 set_opt(sbi, POSIX_ACL);
1510 #endif
1511
1512 f2fs_build_fault_attr(sbi, 0, 0);
1513 }
1514
1515 #ifdef CONFIG_QUOTA
1516 static int f2fs_enable_quotas(struct super_block *sb);
1517 #endif
1518
1519 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1520 {
1521 unsigned int s_flags = sbi->sb->s_flags;
1522 struct cp_control cpc;
1523 int err = 0;
1524 int ret;
1525 block_t unusable;
1526
1527 if (s_flags & SB_RDONLY) {
1528 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1529 return -EINVAL;
1530 }
1531 sbi->sb->s_flags |= SB_ACTIVE;
1532
1533 f2fs_update_time(sbi, DISABLE_TIME);
1534
1535 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1536 mutex_lock(&sbi->gc_mutex);
1537 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1538 if (err == -ENODATA) {
1539 err = 0;
1540 break;
1541 }
1542 if (err && err != -EAGAIN)
1543 break;
1544 }
1545
1546 ret = sync_filesystem(sbi->sb);
1547 if (ret || err) {
1548 err = ret ? ret: err;
1549 goto restore_flag;
1550 }
1551
1552 unusable = f2fs_get_unusable_blocks(sbi);
1553 if (f2fs_disable_cp_again(sbi, unusable)) {
1554 err = -EAGAIN;
1555 goto restore_flag;
1556 }
1557
1558 mutex_lock(&sbi->gc_mutex);
1559 cpc.reason = CP_PAUSE;
1560 set_sbi_flag(sbi, SBI_CP_DISABLED);
1561 err = f2fs_write_checkpoint(sbi, &cpc);
1562 if (err)
1563 goto out_unlock;
1564
1565 spin_lock(&sbi->stat_lock);
1566 sbi->unusable_block_count = unusable;
1567 spin_unlock(&sbi->stat_lock);
1568
1569 out_unlock:
1570 mutex_unlock(&sbi->gc_mutex);
1571 restore_flag:
1572 sbi->sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1573 return err;
1574 }
1575
1576 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1577 {
1578 mutex_lock(&sbi->gc_mutex);
1579 f2fs_dirty_to_prefree(sbi);
1580
1581 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1582 set_sbi_flag(sbi, SBI_IS_DIRTY);
1583 mutex_unlock(&sbi->gc_mutex);
1584
1585 f2fs_sync_fs(sbi->sb, 1);
1586 }
1587
1588 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1589 {
1590 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1591 struct f2fs_mount_info org_mount_opt;
1592 unsigned long old_sb_flags;
1593 int err;
1594 bool need_restart_gc = false;
1595 bool need_stop_gc = false;
1596 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1597 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1598 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1599 bool checkpoint_changed;
1600 #ifdef CONFIG_QUOTA
1601 int i, j;
1602 #endif
1603
1604 /*
1605 * Save the old mount options in case we
1606 * need to restore them.
1607 */
1608 org_mount_opt = sbi->mount_opt;
1609 old_sb_flags = sb->s_flags;
1610
1611 #ifdef CONFIG_QUOTA
1612 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1613 for (i = 0; i < MAXQUOTAS; i++) {
1614 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1615 org_mount_opt.s_qf_names[i] =
1616 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1617 GFP_KERNEL);
1618 if (!org_mount_opt.s_qf_names[i]) {
1619 for (j = 0; j < i; j++)
1620 kvfree(org_mount_opt.s_qf_names[j]);
1621 return -ENOMEM;
1622 }
1623 } else {
1624 org_mount_opt.s_qf_names[i] = NULL;
1625 }
1626 }
1627 #endif
1628
1629 /* recover superblocks we couldn't write due to previous RO mount */
1630 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1631 err = f2fs_commit_super(sbi, false);
1632 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1633 err);
1634 if (!err)
1635 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1636 }
1637
1638 default_options(sbi);
1639
1640 /* parse mount options */
1641 err = parse_options(sb, data);
1642 if (err)
1643 goto restore_opts;
1644 checkpoint_changed =
1645 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1646
1647 /*
1648 * Previous and new state of filesystem is RO,
1649 * so skip checking GC and FLUSH_MERGE conditions.
1650 */
1651 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1652 goto skip;
1653
1654 #ifdef CONFIG_QUOTA
1655 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1656 err = dquot_suspend(sb, -1);
1657 if (err < 0)
1658 goto restore_opts;
1659 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1660 /* dquot_resume needs RW */
1661 sb->s_flags &= ~SB_RDONLY;
1662 if (sb_any_quota_suspended(sb)) {
1663 dquot_resume(sb, -1);
1664 } else if (f2fs_sb_has_quota_ino(sbi)) {
1665 err = f2fs_enable_quotas(sb);
1666 if (err)
1667 goto restore_opts;
1668 }
1669 }
1670 #endif
1671 /* disallow enable/disable extent_cache dynamically */
1672 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1673 err = -EINVAL;
1674 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1675 goto restore_opts;
1676 }
1677
1678 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1679 err = -EINVAL;
1680 f2fs_warn(sbi, "switch io_bits option is not allowed");
1681 goto restore_opts;
1682 }
1683
1684 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1685 err = -EINVAL;
1686 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1687 goto restore_opts;
1688 }
1689
1690 /*
1691 * We stop the GC thread if FS is mounted as RO
1692 * or if background_gc = off is passed in mount
1693 * option. Also sync the filesystem.
1694 */
1695 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1696 if (sbi->gc_thread) {
1697 f2fs_stop_gc_thread(sbi);
1698 need_restart_gc = true;
1699 }
1700 } else if (!sbi->gc_thread) {
1701 err = f2fs_start_gc_thread(sbi);
1702 if (err)
1703 goto restore_opts;
1704 need_stop_gc = true;
1705 }
1706
1707 if (*flags & SB_RDONLY ||
1708 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1709 writeback_inodes_sb(sb, WB_REASON_SYNC);
1710 sync_inodes_sb(sb);
1711
1712 set_sbi_flag(sbi, SBI_IS_DIRTY);
1713 set_sbi_flag(sbi, SBI_IS_CLOSE);
1714 f2fs_sync_fs(sb, 1);
1715 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1716 }
1717
1718 if (checkpoint_changed) {
1719 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1720 err = f2fs_disable_checkpoint(sbi);
1721 if (err)
1722 goto restore_gc;
1723 } else {
1724 f2fs_enable_checkpoint(sbi);
1725 }
1726 }
1727
1728 /*
1729 * We stop issue flush thread if FS is mounted as RO
1730 * or if flush_merge is not passed in mount option.
1731 */
1732 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1733 clear_opt(sbi, FLUSH_MERGE);
1734 f2fs_destroy_flush_cmd_control(sbi, false);
1735 } else {
1736 err = f2fs_create_flush_cmd_control(sbi);
1737 if (err)
1738 goto restore_gc;
1739 }
1740 skip:
1741 #ifdef CONFIG_QUOTA
1742 /* Release old quota file names */
1743 for (i = 0; i < MAXQUOTAS; i++)
1744 kvfree(org_mount_opt.s_qf_names[i]);
1745 #endif
1746 /* Update the POSIXACL Flag */
1747 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1748 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1749
1750 limit_reserve_root(sbi);
1751 adjust_unusable_cap_perc(sbi);
1752 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1753 return 0;
1754 restore_gc:
1755 if (need_restart_gc) {
1756 if (f2fs_start_gc_thread(sbi))
1757 f2fs_warn(sbi, "background gc thread has stopped");
1758 } else if (need_stop_gc) {
1759 f2fs_stop_gc_thread(sbi);
1760 }
1761 restore_opts:
1762 #ifdef CONFIG_QUOTA
1763 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1764 for (i = 0; i < MAXQUOTAS; i++) {
1765 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1766 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1767 }
1768 #endif
1769 sbi->mount_opt = org_mount_opt;
1770 sb->s_flags = old_sb_flags;
1771 return err;
1772 }
1773
1774 #ifdef CONFIG_QUOTA
1775 /* Read data from quotafile */
1776 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1777 size_t len, loff_t off)
1778 {
1779 struct inode *inode = sb_dqopt(sb)->files[type];
1780 struct address_space *mapping = inode->i_mapping;
1781 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1782 int offset = off & (sb->s_blocksize - 1);
1783 int tocopy;
1784 size_t toread;
1785 loff_t i_size = i_size_read(inode);
1786 struct page *page;
1787 char *kaddr;
1788
1789 if (off > i_size)
1790 return 0;
1791
1792 if (off + len > i_size)
1793 len = i_size - off;
1794 toread = len;
1795 while (toread > 0) {
1796 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1797 repeat:
1798 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1799 if (IS_ERR(page)) {
1800 if (PTR_ERR(page) == -ENOMEM) {
1801 congestion_wait(BLK_RW_ASYNC, HZ/50);
1802 goto repeat;
1803 }
1804 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1805 return PTR_ERR(page);
1806 }
1807
1808 lock_page(page);
1809
1810 if (unlikely(page->mapping != mapping)) {
1811 f2fs_put_page(page, 1);
1812 goto repeat;
1813 }
1814 if (unlikely(!PageUptodate(page))) {
1815 f2fs_put_page(page, 1);
1816 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1817 return -EIO;
1818 }
1819
1820 kaddr = kmap_atomic(page);
1821 memcpy(data, kaddr + offset, tocopy);
1822 kunmap_atomic(kaddr);
1823 f2fs_put_page(page, 1);
1824
1825 offset = 0;
1826 toread -= tocopy;
1827 data += tocopy;
1828 blkidx++;
1829 }
1830 return len;
1831 }
1832
1833 /* Write to quotafile */
1834 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1835 const char *data, size_t len, loff_t off)
1836 {
1837 struct inode *inode = sb_dqopt(sb)->files[type];
1838 struct address_space *mapping = inode->i_mapping;
1839 const struct address_space_operations *a_ops = mapping->a_ops;
1840 int offset = off & (sb->s_blocksize - 1);
1841 size_t towrite = len;
1842 struct page *page;
1843 void *fsdata = NULL;
1844 char *kaddr;
1845 int err = 0;
1846 int tocopy;
1847
1848 while (towrite > 0) {
1849 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1850 towrite);
1851 retry:
1852 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1853 &page, &fsdata);
1854 if (unlikely(err)) {
1855 if (err == -ENOMEM) {
1856 congestion_wait(BLK_RW_ASYNC, HZ/50);
1857 goto retry;
1858 }
1859 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1860 break;
1861 }
1862
1863 kaddr = kmap_atomic(page);
1864 memcpy(kaddr + offset, data, tocopy);
1865 kunmap_atomic(kaddr);
1866 flush_dcache_page(page);
1867
1868 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1869 page, fsdata);
1870 offset = 0;
1871 towrite -= tocopy;
1872 off += tocopy;
1873 data += tocopy;
1874 cond_resched();
1875 }
1876
1877 if (len == towrite)
1878 return err;
1879 inode->i_mtime = inode->i_ctime = current_time(inode);
1880 f2fs_mark_inode_dirty_sync(inode, false);
1881 return len - towrite;
1882 }
1883
1884 static struct dquot **f2fs_get_dquots(struct inode *inode)
1885 {
1886 return F2FS_I(inode)->i_dquot;
1887 }
1888
1889 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1890 {
1891 return &F2FS_I(inode)->i_reserved_quota;
1892 }
1893
1894 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1895 {
1896 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1897 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1898 return 0;
1899 }
1900
1901 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1902 F2FS_OPTION(sbi).s_jquota_fmt, type);
1903 }
1904
1905 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1906 {
1907 int enabled = 0;
1908 int i, err;
1909
1910 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1911 err = f2fs_enable_quotas(sbi->sb);
1912 if (err) {
1913 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
1914 return 0;
1915 }
1916 return 1;
1917 }
1918
1919 for (i = 0; i < MAXQUOTAS; i++) {
1920 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1921 err = f2fs_quota_on_mount(sbi, i);
1922 if (!err) {
1923 enabled = 1;
1924 continue;
1925 }
1926 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
1927 err, i);
1928 }
1929 }
1930 return enabled;
1931 }
1932
1933 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1934 unsigned int flags)
1935 {
1936 struct inode *qf_inode;
1937 unsigned long qf_inum;
1938 int err;
1939
1940 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1941
1942 qf_inum = f2fs_qf_ino(sb, type);
1943 if (!qf_inum)
1944 return -EPERM;
1945
1946 qf_inode = f2fs_iget(sb, qf_inum);
1947 if (IS_ERR(qf_inode)) {
1948 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
1949 return PTR_ERR(qf_inode);
1950 }
1951
1952 /* Don't account quota for quota files to avoid recursion */
1953 qf_inode->i_flags |= S_NOQUOTA;
1954 err = dquot_enable(qf_inode, type, format_id, flags);
1955 iput(qf_inode);
1956 return err;
1957 }
1958
1959 static int f2fs_enable_quotas(struct super_block *sb)
1960 {
1961 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1962 int type, err = 0;
1963 unsigned long qf_inum;
1964 bool quota_mopt[MAXQUOTAS] = {
1965 test_opt(sbi, USRQUOTA),
1966 test_opt(sbi, GRPQUOTA),
1967 test_opt(sbi, PRJQUOTA),
1968 };
1969
1970 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1971 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
1972 return 0;
1973 }
1974
1975 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1976
1977 for (type = 0; type < MAXQUOTAS; type++) {
1978 qf_inum = f2fs_qf_ino(sb, type);
1979 if (qf_inum) {
1980 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1981 DQUOT_USAGE_ENABLED |
1982 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1983 if (err) {
1984 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
1985 type, err);
1986 for (type--; type >= 0; type--)
1987 dquot_quota_off(sb, type);
1988 set_sbi_flag(F2FS_SB(sb),
1989 SBI_QUOTA_NEED_REPAIR);
1990 return err;
1991 }
1992 }
1993 }
1994 return 0;
1995 }
1996
1997 int f2fs_quota_sync(struct super_block *sb, int type)
1998 {
1999 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2000 struct quota_info *dqopt = sb_dqopt(sb);
2001 int cnt;
2002 int ret;
2003
2004 /*
2005 * do_quotactl
2006 * f2fs_quota_sync
2007 * down_read(quota_sem)
2008 * dquot_writeback_dquots()
2009 * f2fs_dquot_commit
2010 * block_operation
2011 * down_read(quota_sem)
2012 */
2013 f2fs_lock_op(sbi);
2014
2015 down_read(&sbi->quota_sem);
2016 ret = dquot_writeback_dquots(sb, type);
2017 if (ret)
2018 goto out;
2019
2020 /*
2021 * Now when everything is written we can discard the pagecache so
2022 * that userspace sees the changes.
2023 */
2024 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2025 struct address_space *mapping;
2026
2027 if (type != -1 && cnt != type)
2028 continue;
2029 if (!sb_has_quota_active(sb, cnt))
2030 continue;
2031
2032 mapping = dqopt->files[cnt]->i_mapping;
2033
2034 ret = filemap_fdatawrite(mapping);
2035 if (ret)
2036 goto out;
2037
2038 /* if we are using journalled quota */
2039 if (is_journalled_quota(sbi))
2040 continue;
2041
2042 ret = filemap_fdatawait(mapping);
2043 if (ret)
2044 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2045
2046 inode_lock(dqopt->files[cnt]);
2047 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2048 inode_unlock(dqopt->files[cnt]);
2049 }
2050 out:
2051 if (ret)
2052 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2053 up_read(&sbi->quota_sem);
2054 f2fs_unlock_op(sbi);
2055 return ret;
2056 }
2057
2058 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2059 const struct path *path)
2060 {
2061 struct inode *inode;
2062 int err;
2063
2064 /* if quota sysfile exists, deny enabling quota with specific file */
2065 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2066 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2067 return -EBUSY;
2068 }
2069
2070 err = f2fs_quota_sync(sb, type);
2071 if (err)
2072 return err;
2073
2074 err = dquot_quota_on(sb, type, format_id, path);
2075 if (err)
2076 return err;
2077
2078 inode = d_inode(path->dentry);
2079
2080 inode_lock(inode);
2081 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2082 f2fs_set_inode_flags(inode);
2083 inode_unlock(inode);
2084 f2fs_mark_inode_dirty_sync(inode, false);
2085
2086 return 0;
2087 }
2088
2089 static int __f2fs_quota_off(struct super_block *sb, int type)
2090 {
2091 struct inode *inode = sb_dqopt(sb)->files[type];
2092 int err;
2093
2094 if (!inode || !igrab(inode))
2095 return dquot_quota_off(sb, type);
2096
2097 err = f2fs_quota_sync(sb, type);
2098 if (err)
2099 goto out_put;
2100
2101 err = dquot_quota_off(sb, type);
2102 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2103 goto out_put;
2104
2105 inode_lock(inode);
2106 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2107 f2fs_set_inode_flags(inode);
2108 inode_unlock(inode);
2109 f2fs_mark_inode_dirty_sync(inode, false);
2110 out_put:
2111 iput(inode);
2112 return err;
2113 }
2114
2115 static int f2fs_quota_off(struct super_block *sb, int type)
2116 {
2117 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2118 int err;
2119
2120 err = __f2fs_quota_off(sb, type);
2121
2122 /*
2123 * quotactl can shutdown journalled quota, result in inconsistence
2124 * between quota record and fs data by following updates, tag the
2125 * flag to let fsck be aware of it.
2126 */
2127 if (is_journalled_quota(sbi))
2128 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2129 return err;
2130 }
2131
2132 void f2fs_quota_off_umount(struct super_block *sb)
2133 {
2134 int type;
2135 int err;
2136
2137 for (type = 0; type < MAXQUOTAS; type++) {
2138 err = __f2fs_quota_off(sb, type);
2139 if (err) {
2140 int ret = dquot_quota_off(sb, type);
2141
2142 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2143 type, err, ret);
2144 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2145 }
2146 }
2147 /*
2148 * In case of checkpoint=disable, we must flush quota blocks.
2149 * This can cause NULL exception for node_inode in end_io, since
2150 * put_super already dropped it.
2151 */
2152 sync_filesystem(sb);
2153 }
2154
2155 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2156 {
2157 struct quota_info *dqopt = sb_dqopt(sb);
2158 int type;
2159
2160 for (type = 0; type < MAXQUOTAS; type++) {
2161 if (!dqopt->files[type])
2162 continue;
2163 f2fs_inode_synced(dqopt->files[type]);
2164 }
2165 }
2166
2167 static int f2fs_dquot_commit(struct dquot *dquot)
2168 {
2169 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2170 int ret;
2171
2172 down_read(&sbi->quota_sem);
2173 ret = dquot_commit(dquot);
2174 if (ret < 0)
2175 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2176 up_read(&sbi->quota_sem);
2177 return ret;
2178 }
2179
2180 static int f2fs_dquot_acquire(struct dquot *dquot)
2181 {
2182 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2183 int ret;
2184
2185 down_read(&sbi->quota_sem);
2186 ret = dquot_acquire(dquot);
2187 if (ret < 0)
2188 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2189 up_read(&sbi->quota_sem);
2190 return ret;
2191 }
2192
2193 static int f2fs_dquot_release(struct dquot *dquot)
2194 {
2195 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2196 int ret;
2197
2198 down_read(&sbi->quota_sem);
2199 ret = dquot_release(dquot);
2200 if (ret < 0)
2201 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2202 up_read(&sbi->quota_sem);
2203 return ret;
2204 }
2205
2206 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2207 {
2208 struct super_block *sb = dquot->dq_sb;
2209 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2210 int ret;
2211
2212 down_read(&sbi->quota_sem);
2213 ret = dquot_mark_dquot_dirty(dquot);
2214
2215 /* if we are using journalled quota */
2216 if (is_journalled_quota(sbi))
2217 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2218
2219 up_read(&sbi->quota_sem);
2220 return ret;
2221 }
2222
2223 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2224 {
2225 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2226 int ret;
2227
2228 down_read(&sbi->quota_sem);
2229 ret = dquot_commit_info(sb, type);
2230 if (ret < 0)
2231 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2232 up_read(&sbi->quota_sem);
2233 return ret;
2234 }
2235
2236 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2237 {
2238 *projid = F2FS_I(inode)->i_projid;
2239 return 0;
2240 }
2241
2242 static const struct dquot_operations f2fs_quota_operations = {
2243 .get_reserved_space = f2fs_get_reserved_space,
2244 .write_dquot = f2fs_dquot_commit,
2245 .acquire_dquot = f2fs_dquot_acquire,
2246 .release_dquot = f2fs_dquot_release,
2247 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2248 .write_info = f2fs_dquot_commit_info,
2249 .alloc_dquot = dquot_alloc,
2250 .destroy_dquot = dquot_destroy,
2251 .get_projid = f2fs_get_projid,
2252 .get_next_id = dquot_get_next_id,
2253 };
2254
2255 static const struct quotactl_ops f2fs_quotactl_ops = {
2256 .quota_on = f2fs_quota_on,
2257 .quota_off = f2fs_quota_off,
2258 .quota_sync = f2fs_quota_sync,
2259 .get_state = dquot_get_state,
2260 .set_info = dquot_set_dqinfo,
2261 .get_dqblk = dquot_get_dqblk,
2262 .set_dqblk = dquot_set_dqblk,
2263 .get_nextdqblk = dquot_get_next_dqblk,
2264 };
2265 #else
2266 int f2fs_quota_sync(struct super_block *sb, int type)
2267 {
2268 return 0;
2269 }
2270
2271 void f2fs_quota_off_umount(struct super_block *sb)
2272 {
2273 }
2274 #endif
2275
2276 static const struct super_operations f2fs_sops = {
2277 .alloc_inode = f2fs_alloc_inode,
2278 .free_inode = f2fs_free_inode,
2279 .drop_inode = f2fs_drop_inode,
2280 .write_inode = f2fs_write_inode,
2281 .dirty_inode = f2fs_dirty_inode,
2282 .show_options = f2fs_show_options,
2283 #ifdef CONFIG_QUOTA
2284 .quota_read = f2fs_quota_read,
2285 .quota_write = f2fs_quota_write,
2286 .get_dquots = f2fs_get_dquots,
2287 #endif
2288 .evict_inode = f2fs_evict_inode,
2289 .put_super = f2fs_put_super,
2290 .sync_fs = f2fs_sync_fs,
2291 .freeze_fs = f2fs_freeze,
2292 .unfreeze_fs = f2fs_unfreeze,
2293 .statfs = f2fs_statfs,
2294 .remount_fs = f2fs_remount,
2295 };
2296
2297 #ifdef CONFIG_FS_ENCRYPTION
2298 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2299 {
2300 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2301 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2302 ctx, len, NULL);
2303 }
2304
2305 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2306 void *fs_data)
2307 {
2308 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2309
2310 /*
2311 * Encrypting the root directory is not allowed because fsck
2312 * expects lost+found directory to exist and remain unencrypted
2313 * if LOST_FOUND feature is enabled.
2314 *
2315 */
2316 if (f2fs_sb_has_lost_found(sbi) &&
2317 inode->i_ino == F2FS_ROOT_INO(sbi))
2318 return -EPERM;
2319
2320 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2321 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2322 ctx, len, fs_data, XATTR_CREATE);
2323 }
2324
2325 static bool f2fs_dummy_context(struct inode *inode)
2326 {
2327 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2328 }
2329
2330 static const struct fscrypt_operations f2fs_cryptops = {
2331 .key_prefix = "f2fs:",
2332 .get_context = f2fs_get_context,
2333 .set_context = f2fs_set_context,
2334 .dummy_context = f2fs_dummy_context,
2335 .empty_dir = f2fs_empty_dir,
2336 .max_namelen = F2FS_NAME_LEN,
2337 };
2338 #endif
2339
2340 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2341 u64 ino, u32 generation)
2342 {
2343 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2344 struct inode *inode;
2345
2346 if (f2fs_check_nid_range(sbi, ino))
2347 return ERR_PTR(-ESTALE);
2348
2349 /*
2350 * f2fs_iget isn't quite right if the inode is currently unallocated!
2351 * However f2fs_iget currently does appropriate checks to handle stale
2352 * inodes so everything is OK.
2353 */
2354 inode = f2fs_iget(sb, ino);
2355 if (IS_ERR(inode))
2356 return ERR_CAST(inode);
2357 if (unlikely(generation && inode->i_generation != generation)) {
2358 /* we didn't find the right inode.. */
2359 iput(inode);
2360 return ERR_PTR(-ESTALE);
2361 }
2362 return inode;
2363 }
2364
2365 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2366 int fh_len, int fh_type)
2367 {
2368 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2369 f2fs_nfs_get_inode);
2370 }
2371
2372 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2373 int fh_len, int fh_type)
2374 {
2375 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2376 f2fs_nfs_get_inode);
2377 }
2378
2379 static const struct export_operations f2fs_export_ops = {
2380 .fh_to_dentry = f2fs_fh_to_dentry,
2381 .fh_to_parent = f2fs_fh_to_parent,
2382 .get_parent = f2fs_get_parent,
2383 };
2384
2385 static loff_t max_file_blocks(void)
2386 {
2387 loff_t result = 0;
2388 loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2389
2390 /*
2391 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2392 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2393 * space in inode.i_addr, it will be more safe to reassign
2394 * result as zero.
2395 */
2396
2397 /* two direct node blocks */
2398 result += (leaf_count * 2);
2399
2400 /* two indirect node blocks */
2401 leaf_count *= NIDS_PER_BLOCK;
2402 result += (leaf_count * 2);
2403
2404 /* one double indirect node block */
2405 leaf_count *= NIDS_PER_BLOCK;
2406 result += leaf_count;
2407
2408 return result;
2409 }
2410
2411 static int __f2fs_commit_super(struct buffer_head *bh,
2412 struct f2fs_super_block *super)
2413 {
2414 lock_buffer(bh);
2415 if (super)
2416 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2417 set_buffer_dirty(bh);
2418 unlock_buffer(bh);
2419
2420 /* it's rare case, we can do fua all the time */
2421 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2422 }
2423
2424 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2425 struct buffer_head *bh)
2426 {
2427 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2428 (bh->b_data + F2FS_SUPER_OFFSET);
2429 struct super_block *sb = sbi->sb;
2430 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2431 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2432 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2433 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2434 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2435 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2436 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2437 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2438 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2439 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2440 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2441 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2442 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2443 u64 main_end_blkaddr = main_blkaddr +
2444 (segment_count_main << log_blocks_per_seg);
2445 u64 seg_end_blkaddr = segment0_blkaddr +
2446 (segment_count << log_blocks_per_seg);
2447
2448 if (segment0_blkaddr != cp_blkaddr) {
2449 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2450 segment0_blkaddr, cp_blkaddr);
2451 return true;
2452 }
2453
2454 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2455 sit_blkaddr) {
2456 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2457 cp_blkaddr, sit_blkaddr,
2458 segment_count_ckpt << log_blocks_per_seg);
2459 return true;
2460 }
2461
2462 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2463 nat_blkaddr) {
2464 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2465 sit_blkaddr, nat_blkaddr,
2466 segment_count_sit << log_blocks_per_seg);
2467 return true;
2468 }
2469
2470 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2471 ssa_blkaddr) {
2472 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2473 nat_blkaddr, ssa_blkaddr,
2474 segment_count_nat << log_blocks_per_seg);
2475 return true;
2476 }
2477
2478 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2479 main_blkaddr) {
2480 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2481 ssa_blkaddr, main_blkaddr,
2482 segment_count_ssa << log_blocks_per_seg);
2483 return true;
2484 }
2485
2486 if (main_end_blkaddr > seg_end_blkaddr) {
2487 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2488 main_blkaddr,
2489 segment0_blkaddr +
2490 (segment_count << log_blocks_per_seg),
2491 segment_count_main << log_blocks_per_seg);
2492 return true;
2493 } else if (main_end_blkaddr < seg_end_blkaddr) {
2494 int err = 0;
2495 char *res;
2496
2497 /* fix in-memory information all the time */
2498 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2499 segment0_blkaddr) >> log_blocks_per_seg);
2500
2501 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2502 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2503 res = "internally";
2504 } else {
2505 err = __f2fs_commit_super(bh, NULL);
2506 res = err ? "failed" : "done";
2507 }
2508 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2509 res, main_blkaddr,
2510 segment0_blkaddr +
2511 (segment_count << log_blocks_per_seg),
2512 segment_count_main << log_blocks_per_seg);
2513 if (err)
2514 return true;
2515 }
2516 return false;
2517 }
2518
2519 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2520 struct buffer_head *bh)
2521 {
2522 block_t segment_count, segs_per_sec, secs_per_zone;
2523 block_t total_sections, blocks_per_seg;
2524 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2525 (bh->b_data + F2FS_SUPER_OFFSET);
2526 unsigned int blocksize;
2527 size_t crc_offset = 0;
2528 __u32 crc = 0;
2529
2530 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2531 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2532 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2533 return -EINVAL;
2534 }
2535
2536 /* Check checksum_offset and crc in superblock */
2537 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2538 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2539 if (crc_offset !=
2540 offsetof(struct f2fs_super_block, crc)) {
2541 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2542 crc_offset);
2543 return -EFSCORRUPTED;
2544 }
2545 crc = le32_to_cpu(raw_super->crc);
2546 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2547 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2548 return -EFSCORRUPTED;
2549 }
2550 }
2551
2552 /* Currently, support only 4KB page cache size */
2553 if (F2FS_BLKSIZE != PAGE_SIZE) {
2554 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2555 PAGE_SIZE);
2556 return -EFSCORRUPTED;
2557 }
2558
2559 /* Currently, support only 4KB block size */
2560 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2561 if (blocksize != F2FS_BLKSIZE) {
2562 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2563 blocksize);
2564 return -EFSCORRUPTED;
2565 }
2566
2567 /* check log blocks per segment */
2568 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2569 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2570 le32_to_cpu(raw_super->log_blocks_per_seg));
2571 return -EFSCORRUPTED;
2572 }
2573
2574 /* Currently, support 512/1024/2048/4096 bytes sector size */
2575 if (le32_to_cpu(raw_super->log_sectorsize) >
2576 F2FS_MAX_LOG_SECTOR_SIZE ||
2577 le32_to_cpu(raw_super->log_sectorsize) <
2578 F2FS_MIN_LOG_SECTOR_SIZE) {
2579 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2580 le32_to_cpu(raw_super->log_sectorsize));
2581 return -EFSCORRUPTED;
2582 }
2583 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2584 le32_to_cpu(raw_super->log_sectorsize) !=
2585 F2FS_MAX_LOG_SECTOR_SIZE) {
2586 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2587 le32_to_cpu(raw_super->log_sectors_per_block),
2588 le32_to_cpu(raw_super->log_sectorsize));
2589 return -EFSCORRUPTED;
2590 }
2591
2592 segment_count = le32_to_cpu(raw_super->segment_count);
2593 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2594 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2595 total_sections = le32_to_cpu(raw_super->section_count);
2596
2597 /* blocks_per_seg should be 512, given the above check */
2598 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2599
2600 if (segment_count > F2FS_MAX_SEGMENT ||
2601 segment_count < F2FS_MIN_SEGMENTS) {
2602 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2603 return -EFSCORRUPTED;
2604 }
2605
2606 if (total_sections > segment_count ||
2607 total_sections < F2FS_MIN_SEGMENTS ||
2608 segs_per_sec > segment_count || !segs_per_sec) {
2609 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2610 segment_count, total_sections, segs_per_sec);
2611 return -EFSCORRUPTED;
2612 }
2613
2614 if ((segment_count / segs_per_sec) < total_sections) {
2615 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2616 segment_count, segs_per_sec, total_sections);
2617 return -EFSCORRUPTED;
2618 }
2619
2620 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2621 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2622 segment_count, le64_to_cpu(raw_super->block_count));
2623 return -EFSCORRUPTED;
2624 }
2625
2626 if (secs_per_zone > total_sections || !secs_per_zone) {
2627 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2628 secs_per_zone, total_sections);
2629 return -EFSCORRUPTED;
2630 }
2631 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2632 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2633 (le32_to_cpu(raw_super->extension_count) +
2634 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2635 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2636 le32_to_cpu(raw_super->extension_count),
2637 raw_super->hot_ext_count,
2638 F2FS_MAX_EXTENSION);
2639 return -EFSCORRUPTED;
2640 }
2641
2642 if (le32_to_cpu(raw_super->cp_payload) >
2643 (blocks_per_seg - F2FS_CP_PACKS)) {
2644 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2645 le32_to_cpu(raw_super->cp_payload),
2646 blocks_per_seg - F2FS_CP_PACKS);
2647 return -EFSCORRUPTED;
2648 }
2649
2650 /* check reserved ino info */
2651 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2652 le32_to_cpu(raw_super->meta_ino) != 2 ||
2653 le32_to_cpu(raw_super->root_ino) != 3) {
2654 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2655 le32_to_cpu(raw_super->node_ino),
2656 le32_to_cpu(raw_super->meta_ino),
2657 le32_to_cpu(raw_super->root_ino));
2658 return -EFSCORRUPTED;
2659 }
2660
2661 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2662 if (sanity_check_area_boundary(sbi, bh))
2663 return -EFSCORRUPTED;
2664
2665 return 0;
2666 }
2667
2668 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2669 {
2670 unsigned int total, fsmeta;
2671 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2672 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2673 unsigned int ovp_segments, reserved_segments;
2674 unsigned int main_segs, blocks_per_seg;
2675 unsigned int sit_segs, nat_segs;
2676 unsigned int sit_bitmap_size, nat_bitmap_size;
2677 unsigned int log_blocks_per_seg;
2678 unsigned int segment_count_main;
2679 unsigned int cp_pack_start_sum, cp_payload;
2680 block_t user_block_count, valid_user_blocks;
2681 block_t avail_node_count, valid_node_count;
2682 int i, j;
2683
2684 total = le32_to_cpu(raw_super->segment_count);
2685 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2686 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2687 fsmeta += sit_segs;
2688 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2689 fsmeta += nat_segs;
2690 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2691 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2692
2693 if (unlikely(fsmeta >= total))
2694 return 1;
2695
2696 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2697 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2698
2699 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2700 ovp_segments == 0 || reserved_segments == 0)) {
2701 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2702 return 1;
2703 }
2704
2705 user_block_count = le64_to_cpu(ckpt->user_block_count);
2706 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2707 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2708 if (!user_block_count || user_block_count >=
2709 segment_count_main << log_blocks_per_seg) {
2710 f2fs_err(sbi, "Wrong user_block_count: %u",
2711 user_block_count);
2712 return 1;
2713 }
2714
2715 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2716 if (valid_user_blocks > user_block_count) {
2717 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2718 valid_user_blocks, user_block_count);
2719 return 1;
2720 }
2721
2722 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2723 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2724 if (valid_node_count > avail_node_count) {
2725 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2726 valid_node_count, avail_node_count);
2727 return 1;
2728 }
2729
2730 main_segs = le32_to_cpu(raw_super->segment_count_main);
2731 blocks_per_seg = sbi->blocks_per_seg;
2732
2733 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2734 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2735 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2736 return 1;
2737 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2738 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2739 le32_to_cpu(ckpt->cur_node_segno[j])) {
2740 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2741 i, j,
2742 le32_to_cpu(ckpt->cur_node_segno[i]));
2743 return 1;
2744 }
2745 }
2746 }
2747 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2748 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2749 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2750 return 1;
2751 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2752 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2753 le32_to_cpu(ckpt->cur_data_segno[j])) {
2754 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2755 i, j,
2756 le32_to_cpu(ckpt->cur_data_segno[i]));
2757 return 1;
2758 }
2759 }
2760 }
2761 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2762 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2763 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2764 le32_to_cpu(ckpt->cur_data_segno[j])) {
2765 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2766 i, j,
2767 le32_to_cpu(ckpt->cur_node_segno[i]));
2768 return 1;
2769 }
2770 }
2771 }
2772
2773 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2774 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2775
2776 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2777 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2778 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2779 sit_bitmap_size, nat_bitmap_size);
2780 return 1;
2781 }
2782
2783 cp_pack_start_sum = __start_sum_addr(sbi);
2784 cp_payload = __cp_payload(sbi);
2785 if (cp_pack_start_sum < cp_payload + 1 ||
2786 cp_pack_start_sum > blocks_per_seg - 1 -
2787 NR_CURSEG_TYPE) {
2788 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2789 cp_pack_start_sum);
2790 return 1;
2791 }
2792
2793 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2794 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2795 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2796 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2797 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2798 le32_to_cpu(ckpt->checksum_offset));
2799 return 1;
2800 }
2801
2802 if (unlikely(f2fs_cp_error(sbi))) {
2803 f2fs_err(sbi, "A bug case: need to run fsck");
2804 return 1;
2805 }
2806 return 0;
2807 }
2808
2809 static void init_sb_info(struct f2fs_sb_info *sbi)
2810 {
2811 struct f2fs_super_block *raw_super = sbi->raw_super;
2812 int i;
2813
2814 sbi->log_sectors_per_block =
2815 le32_to_cpu(raw_super->log_sectors_per_block);
2816 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2817 sbi->blocksize = 1 << sbi->log_blocksize;
2818 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2819 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2820 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2821 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2822 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2823 sbi->total_node_count =
2824 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2825 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2826 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2827 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2828 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2829 sbi->cur_victim_sec = NULL_SECNO;
2830 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2831 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2832 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2833 sbi->migration_granularity = sbi->segs_per_sec;
2834
2835 sbi->dir_level = DEF_DIR_LEVEL;
2836 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2837 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2838 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2839 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2840 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2841 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2842 DEF_UMOUNT_DISCARD_TIMEOUT;
2843 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2844
2845 for (i = 0; i < NR_COUNT_TYPE; i++)
2846 atomic_set(&sbi->nr_pages[i], 0);
2847
2848 for (i = 0; i < META; i++)
2849 atomic_set(&sbi->wb_sync_req[i], 0);
2850
2851 INIT_LIST_HEAD(&sbi->s_list);
2852 mutex_init(&sbi->umount_mutex);
2853 init_rwsem(&sbi->io_order_lock);
2854 spin_lock_init(&sbi->cp_lock);
2855
2856 sbi->dirty_device = 0;
2857 spin_lock_init(&sbi->dev_lock);
2858
2859 init_rwsem(&sbi->sb_lock);
2860 }
2861
2862 static int init_percpu_info(struct f2fs_sb_info *sbi)
2863 {
2864 int err;
2865
2866 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2867 if (err)
2868 return err;
2869
2870 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2871 GFP_KERNEL);
2872 if (err)
2873 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2874
2875 return err;
2876 }
2877
2878 #ifdef CONFIG_BLK_DEV_ZONED
2879 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2880 {
2881 struct block_device *bdev = FDEV(devi).bdev;
2882 sector_t nr_sectors = bdev->bd_part->nr_sects;
2883 sector_t sector = 0;
2884 struct blk_zone *zones;
2885 unsigned int i, nr_zones;
2886 unsigned int n = 0;
2887 int err = -EIO;
2888
2889 if (!f2fs_sb_has_blkzoned(sbi))
2890 return 0;
2891
2892 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2893 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2894 return -EINVAL;
2895 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2896 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2897 __ilog2_u32(sbi->blocks_per_blkz))
2898 return -EINVAL;
2899 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2900 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2901 sbi->log_blocks_per_blkz;
2902 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2903 FDEV(devi).nr_blkz++;
2904
2905 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
2906 BITS_TO_LONGS(FDEV(devi).nr_blkz)
2907 * sizeof(unsigned long),
2908 GFP_KERNEL);
2909 if (!FDEV(devi).blkz_seq)
2910 return -ENOMEM;
2911
2912 #define F2FS_REPORT_NR_ZONES 4096
2913
2914 zones = f2fs_kzalloc(sbi,
2915 array_size(F2FS_REPORT_NR_ZONES,
2916 sizeof(struct blk_zone)),
2917 GFP_KERNEL);
2918 if (!zones)
2919 return -ENOMEM;
2920
2921 /* Get block zones type */
2922 while (zones && sector < nr_sectors) {
2923
2924 nr_zones = F2FS_REPORT_NR_ZONES;
2925 err = blkdev_report_zones(bdev, sector, zones, &nr_zones);
2926 if (err)
2927 break;
2928 if (!nr_zones) {
2929 err = -EIO;
2930 break;
2931 }
2932
2933 for (i = 0; i < nr_zones; i++) {
2934 if (zones[i].type != BLK_ZONE_TYPE_CONVENTIONAL)
2935 set_bit(n, FDEV(devi).blkz_seq);
2936 sector += zones[i].len;
2937 n++;
2938 }
2939 }
2940
2941 kvfree(zones);
2942
2943 return err;
2944 }
2945 #endif
2946
2947 /*
2948 * Read f2fs raw super block.
2949 * Because we have two copies of super block, so read both of them
2950 * to get the first valid one. If any one of them is broken, we pass
2951 * them recovery flag back to the caller.
2952 */
2953 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2954 struct f2fs_super_block **raw_super,
2955 int *valid_super_block, int *recovery)
2956 {
2957 struct super_block *sb = sbi->sb;
2958 int block;
2959 struct buffer_head *bh;
2960 struct f2fs_super_block *super;
2961 int err = 0;
2962
2963 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2964 if (!super)
2965 return -ENOMEM;
2966
2967 for (block = 0; block < 2; block++) {
2968 bh = sb_bread(sb, block);
2969 if (!bh) {
2970 f2fs_err(sbi, "Unable to read %dth superblock",
2971 block + 1);
2972 err = -EIO;
2973 continue;
2974 }
2975
2976 /* sanity checking of raw super */
2977 err = sanity_check_raw_super(sbi, bh);
2978 if (err) {
2979 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
2980 block + 1);
2981 brelse(bh);
2982 continue;
2983 }
2984
2985 if (!*raw_super) {
2986 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2987 sizeof(*super));
2988 *valid_super_block = block;
2989 *raw_super = super;
2990 }
2991 brelse(bh);
2992 }
2993
2994 /* Fail to read any one of the superblocks*/
2995 if (err < 0)
2996 *recovery = 1;
2997
2998 /* No valid superblock */
2999 if (!*raw_super)
3000 kvfree(super);
3001 else
3002 err = 0;
3003
3004 return err;
3005 }
3006
3007 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3008 {
3009 struct buffer_head *bh;
3010 __u32 crc = 0;
3011 int err;
3012
3013 if ((recover && f2fs_readonly(sbi->sb)) ||
3014 bdev_read_only(sbi->sb->s_bdev)) {
3015 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3016 return -EROFS;
3017 }
3018
3019 /* we should update superblock crc here */
3020 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3021 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3022 offsetof(struct f2fs_super_block, crc));
3023 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3024 }
3025
3026 /* write back-up superblock first */
3027 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3028 if (!bh)
3029 return -EIO;
3030 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3031 brelse(bh);
3032
3033 /* if we are in recovery path, skip writing valid superblock */
3034 if (recover || err)
3035 return err;
3036
3037 /* write current valid superblock */
3038 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3039 if (!bh)
3040 return -EIO;
3041 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3042 brelse(bh);
3043 return err;
3044 }
3045
3046 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3047 {
3048 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3049 unsigned int max_devices = MAX_DEVICES;
3050 int i;
3051
3052 /* Initialize single device information */
3053 if (!RDEV(0).path[0]) {
3054 if (!bdev_is_zoned(sbi->sb->s_bdev))
3055 return 0;
3056 max_devices = 1;
3057 }
3058
3059 /*
3060 * Initialize multiple devices information, or single
3061 * zoned block device information.
3062 */
3063 sbi->devs = f2fs_kzalloc(sbi,
3064 array_size(max_devices,
3065 sizeof(struct f2fs_dev_info)),
3066 GFP_KERNEL);
3067 if (!sbi->devs)
3068 return -ENOMEM;
3069
3070 for (i = 0; i < max_devices; i++) {
3071
3072 if (i > 0 && !RDEV(i).path[0])
3073 break;
3074
3075 if (max_devices == 1) {
3076 /* Single zoned block device mount */
3077 FDEV(0).bdev =
3078 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3079 sbi->sb->s_mode, sbi->sb->s_type);
3080 } else {
3081 /* Multi-device mount */
3082 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3083 FDEV(i).total_segments =
3084 le32_to_cpu(RDEV(i).total_segments);
3085 if (i == 0) {
3086 FDEV(i).start_blk = 0;
3087 FDEV(i).end_blk = FDEV(i).start_blk +
3088 (FDEV(i).total_segments <<
3089 sbi->log_blocks_per_seg) - 1 +
3090 le32_to_cpu(raw_super->segment0_blkaddr);
3091 } else {
3092 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3093 FDEV(i).end_blk = FDEV(i).start_blk +
3094 (FDEV(i).total_segments <<
3095 sbi->log_blocks_per_seg) - 1;
3096 }
3097 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3098 sbi->sb->s_mode, sbi->sb->s_type);
3099 }
3100 if (IS_ERR(FDEV(i).bdev))
3101 return PTR_ERR(FDEV(i).bdev);
3102
3103 /* to release errored devices */
3104 sbi->s_ndevs = i + 1;
3105
3106 #ifdef CONFIG_BLK_DEV_ZONED
3107 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3108 !f2fs_sb_has_blkzoned(sbi)) {
3109 f2fs_err(sbi, "Zoned block device feature not enabled\n");
3110 return -EINVAL;
3111 }
3112 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3113 if (init_blkz_info(sbi, i)) {
3114 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3115 return -EINVAL;
3116 }
3117 if (max_devices == 1)
3118 break;
3119 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3120 i, FDEV(i).path,
3121 FDEV(i).total_segments,
3122 FDEV(i).start_blk, FDEV(i).end_blk,
3123 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3124 "Host-aware" : "Host-managed");
3125 continue;
3126 }
3127 #endif
3128 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3129 i, FDEV(i).path,
3130 FDEV(i).total_segments,
3131 FDEV(i).start_blk, FDEV(i).end_blk);
3132 }
3133 f2fs_info(sbi,
3134 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3135 return 0;
3136 }
3137
3138 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3139 {
3140 #ifdef CONFIG_UNICODE
3141 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3142 const struct f2fs_sb_encodings *encoding_info;
3143 struct unicode_map *encoding;
3144 __u16 encoding_flags;
3145
3146 if (f2fs_sb_has_encrypt(sbi)) {
3147 f2fs_err(sbi,
3148 "Can't mount with encoding and encryption");
3149 return -EINVAL;
3150 }
3151
3152 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3153 &encoding_flags)) {
3154 f2fs_err(sbi,
3155 "Encoding requested by superblock is unknown");
3156 return -EINVAL;
3157 }
3158
3159 encoding = utf8_load(encoding_info->version);
3160 if (IS_ERR(encoding)) {
3161 f2fs_err(sbi,
3162 "can't mount with superblock charset: %s-%s "
3163 "not supported by the kernel. flags: 0x%x.",
3164 encoding_info->name, encoding_info->version,
3165 encoding_flags);
3166 return PTR_ERR(encoding);
3167 }
3168 f2fs_info(sbi, "Using encoding defined by superblock: "
3169 "%s-%s with flags 0x%hx", encoding_info->name,
3170 encoding_info->version?:"\b", encoding_flags);
3171
3172 sbi->s_encoding = encoding;
3173 sbi->s_encoding_flags = encoding_flags;
3174 sbi->sb->s_d_op = &f2fs_dentry_ops;
3175 }
3176 #else
3177 if (f2fs_sb_has_casefold(sbi)) {
3178 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3179 return -EINVAL;
3180 }
3181 #endif
3182 return 0;
3183 }
3184
3185 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3186 {
3187 struct f2fs_sm_info *sm_i = SM_I(sbi);
3188
3189 /* adjust parameters according to the volume size */
3190 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3191 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3192 sm_i->dcc_info->discard_granularity = 1;
3193 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3194 }
3195
3196 sbi->readdir_ra = 1;
3197 }
3198
3199 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3200 {
3201 struct f2fs_sb_info *sbi;
3202 struct f2fs_super_block *raw_super;
3203 struct inode *root;
3204 int err;
3205 bool skip_recovery = false, need_fsck = false;
3206 char *options = NULL;
3207 int recovery, i, valid_super_block;
3208 struct curseg_info *seg_i;
3209 int retry_cnt = 1;
3210
3211 try_onemore:
3212 err = -EINVAL;
3213 raw_super = NULL;
3214 valid_super_block = -1;
3215 recovery = 0;
3216
3217 /* allocate memory for f2fs-specific super block info */
3218 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3219 if (!sbi)
3220 return -ENOMEM;
3221
3222 sbi->sb = sb;
3223
3224 /* Load the checksum driver */
3225 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3226 if (IS_ERR(sbi->s_chksum_driver)) {
3227 f2fs_err(sbi, "Cannot load crc32 driver.");
3228 err = PTR_ERR(sbi->s_chksum_driver);
3229 sbi->s_chksum_driver = NULL;
3230 goto free_sbi;
3231 }
3232
3233 /* set a block size */
3234 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3235 f2fs_err(sbi, "unable to set blocksize");
3236 goto free_sbi;
3237 }
3238
3239 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3240 &recovery);
3241 if (err)
3242 goto free_sbi;
3243
3244 sb->s_fs_info = sbi;
3245 sbi->raw_super = raw_super;
3246
3247 /* precompute checksum seed for metadata */
3248 if (f2fs_sb_has_inode_chksum(sbi))
3249 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3250 sizeof(raw_super->uuid));
3251
3252 /*
3253 * The BLKZONED feature indicates that the drive was formatted with
3254 * zone alignment optimization. This is optional for host-aware
3255 * devices, but mandatory for host-managed zoned block devices.
3256 */
3257 #ifndef CONFIG_BLK_DEV_ZONED
3258 if (f2fs_sb_has_blkzoned(sbi)) {
3259 f2fs_err(sbi, "Zoned block device support is not enabled");
3260 err = -EOPNOTSUPP;
3261 goto free_sb_buf;
3262 }
3263 #endif
3264 default_options(sbi);
3265 /* parse mount options */
3266 options = kstrdup((const char *)data, GFP_KERNEL);
3267 if (data && !options) {
3268 err = -ENOMEM;
3269 goto free_sb_buf;
3270 }
3271
3272 err = parse_options(sb, options);
3273 if (err)
3274 goto free_options;
3275
3276 sbi->max_file_blocks = max_file_blocks();
3277 sb->s_maxbytes = sbi->max_file_blocks <<
3278 le32_to_cpu(raw_super->log_blocksize);
3279 sb->s_max_links = F2FS_LINK_MAX;
3280
3281 err = f2fs_setup_casefold(sbi);
3282 if (err)
3283 goto free_options;
3284
3285 #ifdef CONFIG_QUOTA
3286 sb->dq_op = &f2fs_quota_operations;
3287 sb->s_qcop = &f2fs_quotactl_ops;
3288 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3289
3290 if (f2fs_sb_has_quota_ino(sbi)) {
3291 for (i = 0; i < MAXQUOTAS; i++) {
3292 if (f2fs_qf_ino(sbi->sb, i))
3293 sbi->nquota_files++;
3294 }
3295 }
3296 #endif
3297
3298 sb->s_op = &f2fs_sops;
3299 #ifdef CONFIG_FS_ENCRYPTION
3300 sb->s_cop = &f2fs_cryptops;
3301 #endif
3302 #ifdef CONFIG_FS_VERITY
3303 sb->s_vop = &f2fs_verityops;
3304 #endif
3305 sb->s_xattr = f2fs_xattr_handlers;
3306 sb->s_export_op = &f2fs_export_ops;
3307 sb->s_magic = F2FS_SUPER_MAGIC;
3308 sb->s_time_gran = 1;
3309 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3310 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3311 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3312 sb->s_iflags |= SB_I_CGROUPWB;
3313
3314 /* init f2fs-specific super block info */
3315 sbi->valid_super_block = valid_super_block;
3316 mutex_init(&sbi->gc_mutex);
3317 mutex_init(&sbi->writepages);
3318 mutex_init(&sbi->cp_mutex);
3319 mutex_init(&sbi->resize_mutex);
3320 init_rwsem(&sbi->node_write);
3321 init_rwsem(&sbi->node_change);
3322
3323 /* disallow all the data/node/meta page writes */
3324 set_sbi_flag(sbi, SBI_POR_DOING);
3325 spin_lock_init(&sbi->stat_lock);
3326
3327 /* init iostat info */
3328 spin_lock_init(&sbi->iostat_lock);
3329 sbi->iostat_enable = false;
3330
3331 for (i = 0; i < NR_PAGE_TYPE; i++) {
3332 int n = (i == META) ? 1: NR_TEMP_TYPE;
3333 int j;
3334
3335 sbi->write_io[i] =
3336 f2fs_kmalloc(sbi,
3337 array_size(n,
3338 sizeof(struct f2fs_bio_info)),
3339 GFP_KERNEL);
3340 if (!sbi->write_io[i]) {
3341 err = -ENOMEM;
3342 goto free_bio_info;
3343 }
3344
3345 for (j = HOT; j < n; j++) {
3346 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3347 sbi->write_io[i][j].sbi = sbi;
3348 sbi->write_io[i][j].bio = NULL;
3349 spin_lock_init(&sbi->write_io[i][j].io_lock);
3350 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3351 }
3352 }
3353
3354 init_rwsem(&sbi->cp_rwsem);
3355 init_rwsem(&sbi->quota_sem);
3356 init_waitqueue_head(&sbi->cp_wait);
3357 init_sb_info(sbi);
3358
3359 err = init_percpu_info(sbi);
3360 if (err)
3361 goto free_bio_info;
3362
3363 if (F2FS_IO_ALIGNED(sbi)) {
3364 sbi->write_io_dummy =
3365 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3366 if (!sbi->write_io_dummy) {
3367 err = -ENOMEM;
3368 goto free_percpu;
3369 }
3370 }
3371
3372 /* get an inode for meta space */
3373 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3374 if (IS_ERR(sbi->meta_inode)) {
3375 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3376 err = PTR_ERR(sbi->meta_inode);
3377 goto free_io_dummy;
3378 }
3379
3380 err = f2fs_get_valid_checkpoint(sbi);
3381 if (err) {
3382 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3383 goto free_meta_inode;
3384 }
3385
3386 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3387 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3388 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3389 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3390 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3391 }
3392
3393 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3394 set_sbi_flag(sbi, SBI_NEED_FSCK);
3395
3396 /* Initialize device list */
3397 err = f2fs_scan_devices(sbi);
3398 if (err) {
3399 f2fs_err(sbi, "Failed to find devices");
3400 goto free_devices;
3401 }
3402
3403 sbi->total_valid_node_count =
3404 le32_to_cpu(sbi->ckpt->valid_node_count);
3405 percpu_counter_set(&sbi->total_valid_inode_count,
3406 le32_to_cpu(sbi->ckpt->valid_inode_count));
3407 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3408 sbi->total_valid_block_count =
3409 le64_to_cpu(sbi->ckpt->valid_block_count);
3410 sbi->last_valid_block_count = sbi->total_valid_block_count;
3411 sbi->reserved_blocks = 0;
3412 sbi->current_reserved_blocks = 0;
3413 limit_reserve_root(sbi);
3414 adjust_unusable_cap_perc(sbi);
3415
3416 for (i = 0; i < NR_INODE_TYPE; i++) {
3417 INIT_LIST_HEAD(&sbi->inode_list[i]);
3418 spin_lock_init(&sbi->inode_lock[i]);
3419 }
3420 mutex_init(&sbi->flush_lock);
3421
3422 f2fs_init_extent_cache_info(sbi);
3423
3424 f2fs_init_ino_entry_info(sbi);
3425
3426 f2fs_init_fsync_node_info(sbi);
3427
3428 /* setup f2fs internal modules */
3429 err = f2fs_build_segment_manager(sbi);
3430 if (err) {
3431 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3432 err);
3433 goto free_sm;
3434 }
3435 err = f2fs_build_node_manager(sbi);
3436 if (err) {
3437 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3438 err);
3439 goto free_nm;
3440 }
3441
3442 /* For write statistics */
3443 if (sb->s_bdev->bd_part)
3444 sbi->sectors_written_start =
3445 (u64)part_stat_read(sb->s_bdev->bd_part,
3446 sectors[STAT_WRITE]);
3447
3448 /* Read accumulated write IO statistics if exists */
3449 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3450 if (__exist_node_summaries(sbi))
3451 sbi->kbytes_written =
3452 le64_to_cpu(seg_i->journal->info.kbytes_written);
3453
3454 f2fs_build_gc_manager(sbi);
3455
3456 err = f2fs_build_stats(sbi);
3457 if (err)
3458 goto free_nm;
3459
3460 /* get an inode for node space */
3461 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3462 if (IS_ERR(sbi->node_inode)) {
3463 f2fs_err(sbi, "Failed to read node inode");
3464 err = PTR_ERR(sbi->node_inode);
3465 goto free_stats;
3466 }
3467
3468 /* read root inode and dentry */
3469 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3470 if (IS_ERR(root)) {
3471 f2fs_err(sbi, "Failed to read root inode");
3472 err = PTR_ERR(root);
3473 goto free_node_inode;
3474 }
3475 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3476 !root->i_size || !root->i_nlink) {
3477 iput(root);
3478 err = -EINVAL;
3479 goto free_node_inode;
3480 }
3481
3482 sb->s_root = d_make_root(root); /* allocate root dentry */
3483 if (!sb->s_root) {
3484 err = -ENOMEM;
3485 goto free_node_inode;
3486 }
3487
3488 err = f2fs_register_sysfs(sbi);
3489 if (err)
3490 goto free_root_inode;
3491
3492 #ifdef CONFIG_QUOTA
3493 /* Enable quota usage during mount */
3494 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3495 err = f2fs_enable_quotas(sb);
3496 if (err)
3497 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3498 }
3499 #endif
3500 /* if there are nt orphan nodes free them */
3501 err = f2fs_recover_orphan_inodes(sbi);
3502 if (err)
3503 goto free_meta;
3504
3505 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3506 goto reset_checkpoint;
3507
3508 /* recover fsynced data */
3509 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3510 !test_opt(sbi, NORECOVERY)) {
3511 /*
3512 * mount should be failed, when device has readonly mode, and
3513 * previous checkpoint was not done by clean system shutdown.
3514 */
3515 if (f2fs_hw_is_readonly(sbi)) {
3516 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3517 err = -EROFS;
3518 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3519 goto free_meta;
3520 }
3521 f2fs_info(sbi, "write access unavailable, skipping recovery");
3522 goto reset_checkpoint;
3523 }
3524
3525 if (need_fsck)
3526 set_sbi_flag(sbi, SBI_NEED_FSCK);
3527
3528 if (skip_recovery)
3529 goto reset_checkpoint;
3530
3531 err = f2fs_recover_fsync_data(sbi, false);
3532 if (err < 0) {
3533 if (err != -ENOMEM)
3534 skip_recovery = true;
3535 need_fsck = true;
3536 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3537 err);
3538 goto free_meta;
3539 }
3540 } else {
3541 err = f2fs_recover_fsync_data(sbi, true);
3542
3543 if (!f2fs_readonly(sb) && err > 0) {
3544 err = -EINVAL;
3545 f2fs_err(sbi, "Need to recover fsync data");
3546 goto free_meta;
3547 }
3548 }
3549 reset_checkpoint:
3550 /* f2fs_recover_fsync_data() cleared this already */
3551 clear_sbi_flag(sbi, SBI_POR_DOING);
3552
3553 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3554 err = f2fs_disable_checkpoint(sbi);
3555 if (err)
3556 goto sync_free_meta;
3557 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3558 f2fs_enable_checkpoint(sbi);
3559 }
3560
3561 /*
3562 * If filesystem is not mounted as read-only then
3563 * do start the gc_thread.
3564 */
3565 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3566 /* After POR, we can run background GC thread.*/
3567 err = f2fs_start_gc_thread(sbi);
3568 if (err)
3569 goto sync_free_meta;
3570 }
3571 kvfree(options);
3572
3573 /* recover broken superblock */
3574 if (recovery) {
3575 err = f2fs_commit_super(sbi, true);
3576 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3577 sbi->valid_super_block ? 1 : 2, err);
3578 }
3579
3580 f2fs_join_shrinker(sbi);
3581
3582 f2fs_tuning_parameters(sbi);
3583
3584 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3585 cur_cp_version(F2FS_CKPT(sbi)));
3586 f2fs_update_time(sbi, CP_TIME);
3587 f2fs_update_time(sbi, REQ_TIME);
3588 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3589 return 0;
3590
3591 sync_free_meta:
3592 /* safe to flush all the data */
3593 sync_filesystem(sbi->sb);
3594 retry_cnt = 0;
3595
3596 free_meta:
3597 #ifdef CONFIG_QUOTA
3598 f2fs_truncate_quota_inode_pages(sb);
3599 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3600 f2fs_quota_off_umount(sbi->sb);
3601 #endif
3602 /*
3603 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3604 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3605 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3606 * falls into an infinite loop in f2fs_sync_meta_pages().
3607 */
3608 truncate_inode_pages_final(META_MAPPING(sbi));
3609 /* evict some inodes being cached by GC */
3610 evict_inodes(sb);
3611 f2fs_unregister_sysfs(sbi);
3612 free_root_inode:
3613 dput(sb->s_root);
3614 sb->s_root = NULL;
3615 free_node_inode:
3616 f2fs_release_ino_entry(sbi, true);
3617 truncate_inode_pages_final(NODE_MAPPING(sbi));
3618 iput(sbi->node_inode);
3619 sbi->node_inode = NULL;
3620 free_stats:
3621 f2fs_destroy_stats(sbi);
3622 free_nm:
3623 f2fs_destroy_node_manager(sbi);
3624 free_sm:
3625 f2fs_destroy_segment_manager(sbi);
3626 free_devices:
3627 destroy_device_list(sbi);
3628 kvfree(sbi->ckpt);
3629 free_meta_inode:
3630 make_bad_inode(sbi->meta_inode);
3631 iput(sbi->meta_inode);
3632 sbi->meta_inode = NULL;
3633 free_io_dummy:
3634 mempool_destroy(sbi->write_io_dummy);
3635 free_percpu:
3636 destroy_percpu_info(sbi);
3637 free_bio_info:
3638 for (i = 0; i < NR_PAGE_TYPE; i++)
3639 kvfree(sbi->write_io[i]);
3640
3641 #ifdef CONFIG_UNICODE
3642 utf8_unload(sbi->s_encoding);
3643 #endif
3644 free_options:
3645 #ifdef CONFIG_QUOTA
3646 for (i = 0; i < MAXQUOTAS; i++)
3647 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3648 #endif
3649 kvfree(options);
3650 free_sb_buf:
3651 kvfree(raw_super);
3652 free_sbi:
3653 if (sbi->s_chksum_driver)
3654 crypto_free_shash(sbi->s_chksum_driver);
3655 kvfree(sbi);
3656
3657 /* give only one another chance */
3658 if (retry_cnt > 0 && skip_recovery) {
3659 retry_cnt--;
3660 shrink_dcache_sb(sb);
3661 goto try_onemore;
3662 }
3663 return err;
3664 }
3665
3666 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3667 const char *dev_name, void *data)
3668 {
3669 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3670 }
3671
3672 static void kill_f2fs_super(struct super_block *sb)
3673 {
3674 if (sb->s_root) {
3675 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3676
3677 set_sbi_flag(sbi, SBI_IS_CLOSE);
3678 f2fs_stop_gc_thread(sbi);
3679 f2fs_stop_discard_thread(sbi);
3680
3681 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3682 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3683 struct cp_control cpc = {
3684 .reason = CP_UMOUNT,
3685 };
3686 f2fs_write_checkpoint(sbi, &cpc);
3687 }
3688
3689 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3690 sb->s_flags &= ~SB_RDONLY;
3691 }
3692 kill_block_super(sb);
3693 }
3694
3695 static struct file_system_type f2fs_fs_type = {
3696 .owner = THIS_MODULE,
3697 .name = "f2fs",
3698 .mount = f2fs_mount,
3699 .kill_sb = kill_f2fs_super,
3700 .fs_flags = FS_REQUIRES_DEV,
3701 };
3702 MODULE_ALIAS_FS("f2fs");
3703
3704 static int __init init_inodecache(void)
3705 {
3706 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3707 sizeof(struct f2fs_inode_info), 0,
3708 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3709 if (!f2fs_inode_cachep)
3710 return -ENOMEM;
3711 return 0;
3712 }
3713
3714 static void destroy_inodecache(void)
3715 {
3716 /*
3717 * Make sure all delayed rcu free inodes are flushed before we
3718 * destroy cache.
3719 */
3720 rcu_barrier();
3721 kmem_cache_destroy(f2fs_inode_cachep);
3722 }
3723
3724 static int __init init_f2fs_fs(void)
3725 {
3726 int err;
3727
3728 if (PAGE_SIZE != F2FS_BLKSIZE) {
3729 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3730 PAGE_SIZE, F2FS_BLKSIZE);
3731 return -EINVAL;
3732 }
3733
3734 f2fs_build_trace_ios();
3735
3736 err = init_inodecache();
3737 if (err)
3738 goto fail;
3739 err = f2fs_create_node_manager_caches();
3740 if (err)
3741 goto free_inodecache;
3742 err = f2fs_create_segment_manager_caches();
3743 if (err)
3744 goto free_node_manager_caches;
3745 err = f2fs_create_checkpoint_caches();
3746 if (err)
3747 goto free_segment_manager_caches;
3748 err = f2fs_create_extent_cache();
3749 if (err)
3750 goto free_checkpoint_caches;
3751 err = f2fs_init_sysfs();
3752 if (err)
3753 goto free_extent_cache;
3754 err = register_shrinker(&f2fs_shrinker_info);
3755 if (err)
3756 goto free_sysfs;
3757 err = register_filesystem(&f2fs_fs_type);
3758 if (err)
3759 goto free_shrinker;
3760 f2fs_create_root_stats();
3761 err = f2fs_init_post_read_processing();
3762 if (err)
3763 goto free_root_stats;
3764 return 0;
3765
3766 free_root_stats:
3767 f2fs_destroy_root_stats();
3768 unregister_filesystem(&f2fs_fs_type);
3769 free_shrinker:
3770 unregister_shrinker(&f2fs_shrinker_info);
3771 free_sysfs:
3772 f2fs_exit_sysfs();
3773 free_extent_cache:
3774 f2fs_destroy_extent_cache();
3775 free_checkpoint_caches:
3776 f2fs_destroy_checkpoint_caches();
3777 free_segment_manager_caches:
3778 f2fs_destroy_segment_manager_caches();
3779 free_node_manager_caches:
3780 f2fs_destroy_node_manager_caches();
3781 free_inodecache:
3782 destroy_inodecache();
3783 fail:
3784 return err;
3785 }
3786
3787 static void __exit exit_f2fs_fs(void)
3788 {
3789 f2fs_destroy_post_read_processing();
3790 f2fs_destroy_root_stats();
3791 unregister_filesystem(&f2fs_fs_type);
3792 unregister_shrinker(&f2fs_shrinker_info);
3793 f2fs_exit_sysfs();
3794 f2fs_destroy_extent_cache();
3795 f2fs_destroy_checkpoint_caches();
3796 f2fs_destroy_segment_manager_caches();
3797 f2fs_destroy_node_manager_caches();
3798 destroy_inodecache();
3799 f2fs_destroy_trace_ios();
3800 }
3801
3802 module_init(init_f2fs_fs)
3803 module_exit(exit_f2fs_fs)
3804
3805 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3806 MODULE_DESCRIPTION("Flash Friendly File System");
3807 MODULE_LICENSE("GPL");
3808
Linux with added FPC-III support