?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
fanotify: merge duplicate events on parent and child
[linux/fpc-iii.git] / fs / f2fs / super.c
blob8deb0a260d928b7d65966a0795f792b549f7ebde
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_compress_algorithm,
145 Opt_compress_log_size,
146 Opt_compress_extension,
147 Opt_err,
148 };
149
150 static match_table_t f2fs_tokens = {
151 {Opt_gc_background, "background_gc=%s"},
152 {Opt_disable_roll_forward, "disable_roll_forward"},
153 {Opt_norecovery, "norecovery"},
154 {Opt_discard, "discard"},
155 {Opt_nodiscard, "nodiscard"},
156 {Opt_noheap, "no_heap"},
157 {Opt_heap, "heap"},
158 {Opt_user_xattr, "user_xattr"},
159 {Opt_nouser_xattr, "nouser_xattr"},
160 {Opt_acl, "acl"},
161 {Opt_noacl, "noacl"},
162 {Opt_active_logs, "active_logs=%u"},
163 {Opt_disable_ext_identify, "disable_ext_identify"},
164 {Opt_inline_xattr, "inline_xattr"},
165 {Opt_noinline_xattr, "noinline_xattr"},
166 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
167 {Opt_inline_data, "inline_data"},
168 {Opt_inline_dentry, "inline_dentry"},
169 {Opt_noinline_dentry, "noinline_dentry"},
170 {Opt_flush_merge, "flush_merge"},
171 {Opt_noflush_merge, "noflush_merge"},
172 {Opt_nobarrier, "nobarrier"},
173 {Opt_fastboot, "fastboot"},
174 {Opt_extent_cache, "extent_cache"},
175 {Opt_noextent_cache, "noextent_cache"},
176 {Opt_noinline_data, "noinline_data"},
177 {Opt_data_flush, "data_flush"},
178 {Opt_reserve_root, "reserve_root=%u"},
179 {Opt_resgid, "resgid=%u"},
180 {Opt_resuid, "resuid=%u"},
181 {Opt_mode, "mode=%s"},
182 {Opt_io_size_bits, "io_bits=%u"},
183 {Opt_fault_injection, "fault_injection=%u"},
184 {Opt_fault_type, "fault_type=%u"},
185 {Opt_lazytime, "lazytime"},
186 {Opt_nolazytime, "nolazytime"},
187 {Opt_quota, "quota"},
188 {Opt_noquota, "noquota"},
189 {Opt_usrquota, "usrquota"},
190 {Opt_grpquota, "grpquota"},
191 {Opt_prjquota, "prjquota"},
192 {Opt_usrjquota, "usrjquota=%s"},
193 {Opt_grpjquota, "grpjquota=%s"},
194 {Opt_prjjquota, "prjjquota=%s"},
195 {Opt_offusrjquota, "usrjquota="},
196 {Opt_offgrpjquota, "grpjquota="},
197 {Opt_offprjjquota, "prjjquota="},
198 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
199 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
200 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
201 {Opt_whint, "whint_mode=%s"},
202 {Opt_alloc, "alloc_mode=%s"},
203 {Opt_fsync, "fsync_mode=%s"},
204 {Opt_test_dummy_encryption, "test_dummy_encryption"},
205 {Opt_checkpoint_disable, "checkpoint=disable"},
206 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
207 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
208 {Opt_checkpoint_enable, "checkpoint=enable"},
209 {Opt_compress_algorithm, "compress_algorithm=%s"},
210 {Opt_compress_log_size, "compress_log_size=%u"},
211 {Opt_compress_extension, "compress_extension=%s"},
212 {Opt_err, NULL},
213 };
214
215 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
216 {
217 struct va_format vaf;
218 va_list args;
219 int level;
220
221 va_start(args, fmt);
222
223 level = printk_get_level(fmt);
224 vaf.fmt = printk_skip_level(fmt);
225 vaf.va = &args;
226 printk("%c%cF2FS-fs (%s): %pV\n",
227 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
228
229 va_end(args);
230 }
231
232 #ifdef CONFIG_UNICODE
233 static const struct f2fs_sb_encodings {
234 __u16 magic;
235 char *name;
236 char *version;
237 } f2fs_sb_encoding_map[] = {
238 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
239 };
240
241 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
242 const struct f2fs_sb_encodings **encoding,
243 __u16 *flags)
244 {
245 __u16 magic = le16_to_cpu(sb->s_encoding);
246 int i;
247
248 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
249 if (magic == f2fs_sb_encoding_map[i].magic)
250 break;
251
252 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
253 return -EINVAL;
254
255 *encoding = &f2fs_sb_encoding_map[i];
256 *flags = le16_to_cpu(sb->s_encoding_flags);
257
258 return 0;
259 }
260 #endif
261
262 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
263 {
264 block_t limit = min((sbi->user_block_count << 1) / 1000,
265 sbi->user_block_count - sbi->reserved_blocks);
266
267 /* limit is 0.2% */
268 if (test_opt(sbi, RESERVE_ROOT) &&
269 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
270 F2FS_OPTION(sbi).root_reserved_blocks = limit;
271 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
272 F2FS_OPTION(sbi).root_reserved_blocks);
273 }
274 if (!test_opt(sbi, RESERVE_ROOT) &&
275 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
276 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
277 !gid_eq(F2FS_OPTION(sbi).s_resgid,
278 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
279 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
280 from_kuid_munged(&init_user_ns,
281 F2FS_OPTION(sbi).s_resuid),
282 from_kgid_munged(&init_user_ns,
283 F2FS_OPTION(sbi).s_resgid));
284 }
285
286 static void init_once(void *foo)
287 {
288 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
289
290 inode_init_once(&fi->vfs_inode);
291 }
292
293 #ifdef CONFIG_QUOTA
294 static const char * const quotatypes[] = INITQFNAMES;
295 #define QTYPE2NAME(t) (quotatypes[t])
296 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
297 substring_t *args)
298 {
299 struct f2fs_sb_info *sbi = F2FS_SB(sb);
300 char *qname;
301 int ret = -EINVAL;
302
303 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
304 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
305 return -EINVAL;
306 }
307 if (f2fs_sb_has_quota_ino(sbi)) {
308 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
309 return 0;
310 }
311
312 qname = match_strdup(args);
313 if (!qname) {
314 f2fs_err(sbi, "Not enough memory for storing quotafile name");
315 return -ENOMEM;
316 }
317 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
318 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
319 ret = 0;
320 else
321 f2fs_err(sbi, "%s quota file already specified",
322 QTYPE2NAME(qtype));
323 goto errout;
324 }
325 if (strchr(qname, '/')) {
326 f2fs_err(sbi, "quotafile must be on filesystem root");
327 goto errout;
328 }
329 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
330 set_opt(sbi, QUOTA);
331 return 0;
332 errout:
333 kvfree(qname);
334 return ret;
335 }
336
337 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
338 {
339 struct f2fs_sb_info *sbi = F2FS_SB(sb);
340
341 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
342 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
343 return -EINVAL;
344 }
345 kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
346 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
347 return 0;
348 }
349
350 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
351 {
352 /*
353 * We do the test below only for project quotas. 'usrquota' and
354 * 'grpquota' mount options are allowed even without quota feature
355 * to support legacy quotas in quota files.
356 */
357 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
358 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
359 return -1;
360 }
361 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
362 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
363 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
364 if (test_opt(sbi, USRQUOTA) &&
365 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
366 clear_opt(sbi, USRQUOTA);
367
368 if (test_opt(sbi, GRPQUOTA) &&
369 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
370 clear_opt(sbi, GRPQUOTA);
371
372 if (test_opt(sbi, PRJQUOTA) &&
373 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
374 clear_opt(sbi, PRJQUOTA);
375
376 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
377 test_opt(sbi, PRJQUOTA)) {
378 f2fs_err(sbi, "old and new quota format mixing");
379 return -1;
380 }
381
382 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
383 f2fs_err(sbi, "journaled quota format not specified");
384 return -1;
385 }
386 }
387
388 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
389 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
390 F2FS_OPTION(sbi).s_jquota_fmt = 0;
391 }
392 return 0;
393 }
394 #endif
395
396 static int parse_options(struct super_block *sb, char *options)
397 {
398 struct f2fs_sb_info *sbi = F2FS_SB(sb);
399 substring_t args[MAX_OPT_ARGS];
400 unsigned char (*ext)[F2FS_EXTENSION_LEN];
401 char *p, *name;
402 int arg = 0, ext_cnt;
403 kuid_t uid;
404 kgid_t gid;
405 #ifdef CONFIG_QUOTA
406 int ret;
407 #endif
408
409 if (!options)
410 return 0;
411
412 while ((p = strsep(&options, ",")) != NULL) {
413 int token;
414 if (!*p)
415 continue;
416 /*
417 * Initialize args struct so we know whether arg was
418 * found; some options take optional arguments.
419 */
420 args[0].to = args[0].from = NULL;
421 token = match_token(p, f2fs_tokens, args);
422
423 switch (token) {
424 case Opt_gc_background:
425 name = match_strdup(&args[0]);
426
427 if (!name)
428 return -ENOMEM;
429 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
430 set_opt(sbi, BG_GC);
431 clear_opt(sbi, FORCE_FG_GC);
432 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
433 clear_opt(sbi, BG_GC);
434 clear_opt(sbi, FORCE_FG_GC);
435 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
436 set_opt(sbi, BG_GC);
437 set_opt(sbi, FORCE_FG_GC);
438 } else {
439 kvfree(name);
440 return -EINVAL;
441 }
442 kvfree(name);
443 break;
444 case Opt_disable_roll_forward:
445 set_opt(sbi, DISABLE_ROLL_FORWARD);
446 break;
447 case Opt_norecovery:
448 /* this option mounts f2fs with ro */
449 set_opt(sbi, NORECOVERY);
450 if (!f2fs_readonly(sb))
451 return -EINVAL;
452 break;
453 case Opt_discard:
454 set_opt(sbi, DISCARD);
455 break;
456 case Opt_nodiscard:
457 if (f2fs_sb_has_blkzoned(sbi)) {
458 f2fs_warn(sbi, "discard is required for zoned block devices");
459 return -EINVAL;
460 }
461 clear_opt(sbi, DISCARD);
462 break;
463 case Opt_noheap:
464 set_opt(sbi, NOHEAP);
465 break;
466 case Opt_heap:
467 clear_opt(sbi, NOHEAP);
468 break;
469 #ifdef CONFIG_F2FS_FS_XATTR
470 case Opt_user_xattr:
471 set_opt(sbi, XATTR_USER);
472 break;
473 case Opt_nouser_xattr:
474 clear_opt(sbi, XATTR_USER);
475 break;
476 case Opt_inline_xattr:
477 set_opt(sbi, INLINE_XATTR);
478 break;
479 case Opt_noinline_xattr:
480 clear_opt(sbi, INLINE_XATTR);
481 break;
482 case Opt_inline_xattr_size:
483 if (args->from && match_int(args, &arg))
484 return -EINVAL;
485 set_opt(sbi, INLINE_XATTR_SIZE);
486 F2FS_OPTION(sbi).inline_xattr_size = arg;
487 break;
488 #else
489 case Opt_user_xattr:
490 f2fs_info(sbi, "user_xattr options not supported");
491 break;
492 case Opt_nouser_xattr:
493 f2fs_info(sbi, "nouser_xattr options not supported");
494 break;
495 case Opt_inline_xattr:
496 f2fs_info(sbi, "inline_xattr options not supported");
497 break;
498 case Opt_noinline_xattr:
499 f2fs_info(sbi, "noinline_xattr options not supported");
500 break;
501 #endif
502 #ifdef CONFIG_F2FS_FS_POSIX_ACL
503 case Opt_acl:
504 set_opt(sbi, POSIX_ACL);
505 break;
506 case Opt_noacl:
507 clear_opt(sbi, POSIX_ACL);
508 break;
509 #else
510 case Opt_acl:
511 f2fs_info(sbi, "acl options not supported");
512 break;
513 case Opt_noacl:
514 f2fs_info(sbi, "noacl options not supported");
515 break;
516 #endif
517 case Opt_active_logs:
518 if (args->from && match_int(args, &arg))
519 return -EINVAL;
520 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
521 return -EINVAL;
522 F2FS_OPTION(sbi).active_logs = arg;
523 break;
524 case Opt_disable_ext_identify:
525 set_opt(sbi, DISABLE_EXT_IDENTIFY);
526 break;
527 case Opt_inline_data:
528 set_opt(sbi, INLINE_DATA);
529 break;
530 case Opt_inline_dentry:
531 set_opt(sbi, INLINE_DENTRY);
532 break;
533 case Opt_noinline_dentry:
534 clear_opt(sbi, INLINE_DENTRY);
535 break;
536 case Opt_flush_merge:
537 set_opt(sbi, FLUSH_MERGE);
538 break;
539 case Opt_noflush_merge:
540 clear_opt(sbi, FLUSH_MERGE);
541 break;
542 case Opt_nobarrier:
543 set_opt(sbi, NOBARRIER);
544 break;
545 case Opt_fastboot:
546 set_opt(sbi, FASTBOOT);
547 break;
548 case Opt_extent_cache:
549 set_opt(sbi, EXTENT_CACHE);
550 break;
551 case Opt_noextent_cache:
552 clear_opt(sbi, EXTENT_CACHE);
553 break;
554 case Opt_noinline_data:
555 clear_opt(sbi, INLINE_DATA);
556 break;
557 case Opt_data_flush:
558 set_opt(sbi, DATA_FLUSH);
559 break;
560 case Opt_reserve_root:
561 if (args->from && match_int(args, &arg))
562 return -EINVAL;
563 if (test_opt(sbi, RESERVE_ROOT)) {
564 f2fs_info(sbi, "Preserve previous reserve_root=%u",
565 F2FS_OPTION(sbi).root_reserved_blocks);
566 } else {
567 F2FS_OPTION(sbi).root_reserved_blocks = arg;
568 set_opt(sbi, RESERVE_ROOT);
569 }
570 break;
571 case Opt_resuid:
572 if (args->from && match_int(args, &arg))
573 return -EINVAL;
574 uid = make_kuid(current_user_ns(), arg);
575 if (!uid_valid(uid)) {
576 f2fs_err(sbi, "Invalid uid value %d", arg);
577 return -EINVAL;
578 }
579 F2FS_OPTION(sbi).s_resuid = uid;
580 break;
581 case Opt_resgid:
582 if (args->from && match_int(args, &arg))
583 return -EINVAL;
584 gid = make_kgid(current_user_ns(), arg);
585 if (!gid_valid(gid)) {
586 f2fs_err(sbi, "Invalid gid value %d", arg);
587 return -EINVAL;
588 }
589 F2FS_OPTION(sbi).s_resgid = gid;
590 break;
591 case Opt_mode:
592 name = match_strdup(&args[0]);
593
594 if (!name)
595 return -ENOMEM;
596 if (strlen(name) == 8 &&
597 !strncmp(name, "adaptive", 8)) {
598 if (f2fs_sb_has_blkzoned(sbi)) {
599 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
600 kvfree(name);
601 return -EINVAL;
602 }
603 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
604 } else if (strlen(name) == 3 &&
605 !strncmp(name, "lfs", 3)) {
606 set_opt_mode(sbi, F2FS_MOUNT_LFS);
607 } else {
608 kvfree(name);
609 return -EINVAL;
610 }
611 kvfree(name);
612 break;
613 case Opt_io_size_bits:
614 if (args->from && match_int(args, &arg))
615 return -EINVAL;
616 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
617 f2fs_warn(sbi, "Not support %d, larger than %d",
618 1 << arg, BIO_MAX_PAGES);
619 return -EINVAL;
620 }
621 F2FS_OPTION(sbi).write_io_size_bits = arg;
622 break;
623 #ifdef CONFIG_F2FS_FAULT_INJECTION
624 case Opt_fault_injection:
625 if (args->from && match_int(args, &arg))
626 return -EINVAL;
627 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
628 set_opt(sbi, FAULT_INJECTION);
629 break;
630
631 case Opt_fault_type:
632 if (args->from && match_int(args, &arg))
633 return -EINVAL;
634 f2fs_build_fault_attr(sbi, 0, arg);
635 set_opt(sbi, FAULT_INJECTION);
636 break;
637 #else
638 case Opt_fault_injection:
639 f2fs_info(sbi, "fault_injection options not supported");
640 break;
641
642 case Opt_fault_type:
643 f2fs_info(sbi, "fault_type options not supported");
644 break;
645 #endif
646 case Opt_lazytime:
647 sb->s_flags |= SB_LAZYTIME;
648 break;
649 case Opt_nolazytime:
650 sb->s_flags &= ~SB_LAZYTIME;
651 break;
652 #ifdef CONFIG_QUOTA
653 case Opt_quota:
654 case Opt_usrquota:
655 set_opt(sbi, USRQUOTA);
656 break;
657 case Opt_grpquota:
658 set_opt(sbi, GRPQUOTA);
659 break;
660 case Opt_prjquota:
661 set_opt(sbi, PRJQUOTA);
662 break;
663 case Opt_usrjquota:
664 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
665 if (ret)
666 return ret;
667 break;
668 case Opt_grpjquota:
669 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
670 if (ret)
671 return ret;
672 break;
673 case Opt_prjjquota:
674 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
675 if (ret)
676 return ret;
677 break;
678 case Opt_offusrjquota:
679 ret = f2fs_clear_qf_name(sb, USRQUOTA);
680 if (ret)
681 return ret;
682 break;
683 case Opt_offgrpjquota:
684 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
685 if (ret)
686 return ret;
687 break;
688 case Opt_offprjjquota:
689 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
690 if (ret)
691 return ret;
692 break;
693 case Opt_jqfmt_vfsold:
694 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
695 break;
696 case Opt_jqfmt_vfsv0:
697 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
698 break;
699 case Opt_jqfmt_vfsv1:
700 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
701 break;
702 case Opt_noquota:
703 clear_opt(sbi, QUOTA);
704 clear_opt(sbi, USRQUOTA);
705 clear_opt(sbi, GRPQUOTA);
706 clear_opt(sbi, PRJQUOTA);
707 break;
708 #else
709 case Opt_quota:
710 case Opt_usrquota:
711 case Opt_grpquota:
712 case Opt_prjquota:
713 case Opt_usrjquota:
714 case Opt_grpjquota:
715 case Opt_prjjquota:
716 case Opt_offusrjquota:
717 case Opt_offgrpjquota:
718 case Opt_offprjjquota:
719 case Opt_jqfmt_vfsold:
720 case Opt_jqfmt_vfsv0:
721 case Opt_jqfmt_vfsv1:
722 case Opt_noquota:
723 f2fs_info(sbi, "quota operations not supported");
724 break;
725 #endif
726 case Opt_whint:
727 name = match_strdup(&args[0]);
728 if (!name)
729 return -ENOMEM;
730 if (strlen(name) == 10 &&
731 !strncmp(name, "user-based", 10)) {
732 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
733 } else if (strlen(name) == 3 &&
734 !strncmp(name, "off", 3)) {
735 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
736 } else if (strlen(name) == 8 &&
737 !strncmp(name, "fs-based", 8)) {
738 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
739 } else {
740 kvfree(name);
741 return -EINVAL;
742 }
743 kvfree(name);
744 break;
745 case Opt_alloc:
746 name = match_strdup(&args[0]);
747 if (!name)
748 return -ENOMEM;
749
750 if (strlen(name) == 7 &&
751 !strncmp(name, "default", 7)) {
752 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
753 } else if (strlen(name) == 5 &&
754 !strncmp(name, "reuse", 5)) {
755 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
756 } else {
757 kvfree(name);
758 return -EINVAL;
759 }
760 kvfree(name);
761 break;
762 case Opt_fsync:
763 name = match_strdup(&args[0]);
764 if (!name)
765 return -ENOMEM;
766 if (strlen(name) == 5 &&
767 !strncmp(name, "posix", 5)) {
768 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
769 } else if (strlen(name) == 6 &&
770 !strncmp(name, "strict", 6)) {
771 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
772 } else if (strlen(name) == 9 &&
773 !strncmp(name, "nobarrier", 9)) {
774 F2FS_OPTION(sbi).fsync_mode =
775 FSYNC_MODE_NOBARRIER;
776 } else {
777 kvfree(name);
778 return -EINVAL;
779 }
780 kvfree(name);
781 break;
782 case Opt_test_dummy_encryption:
783 #ifdef CONFIG_FS_ENCRYPTION
784 if (!f2fs_sb_has_encrypt(sbi)) {
785 f2fs_err(sbi, "Encrypt feature is off");
786 return -EINVAL;
787 }
788
789 F2FS_OPTION(sbi).test_dummy_encryption = true;
790 f2fs_info(sbi, "Test dummy encryption mode enabled");
791 #else
792 f2fs_info(sbi, "Test dummy encryption mount option ignored");
793 #endif
794 break;
795 case Opt_checkpoint_disable_cap_perc:
796 if (args->from && match_int(args, &arg))
797 return -EINVAL;
798 if (arg < 0 || arg > 100)
799 return -EINVAL;
800 if (arg == 100)
801 F2FS_OPTION(sbi).unusable_cap =
802 sbi->user_block_count;
803 else
804 F2FS_OPTION(sbi).unusable_cap =
805 (sbi->user_block_count / 100) * arg;
806 set_opt(sbi, DISABLE_CHECKPOINT);
807 break;
808 case Opt_checkpoint_disable_cap:
809 if (args->from && match_int(args, &arg))
810 return -EINVAL;
811 F2FS_OPTION(sbi).unusable_cap = arg;
812 set_opt(sbi, DISABLE_CHECKPOINT);
813 break;
814 case Opt_checkpoint_disable:
815 set_opt(sbi, DISABLE_CHECKPOINT);
816 break;
817 case Opt_checkpoint_enable:
818 clear_opt(sbi, DISABLE_CHECKPOINT);
819 break;
820 case Opt_compress_algorithm:
821 if (!f2fs_sb_has_compression(sbi)) {
822 f2fs_err(sbi, "Compression feature if off");
823 return -EINVAL;
824 }
825 name = match_strdup(&args[0]);
826 if (!name)
827 return -ENOMEM;
828 if (strlen(name) == 3 && !strcmp(name, "lzo")) {
829 F2FS_OPTION(sbi).compress_algorithm =
830 COMPRESS_LZO;
831 } else if (strlen(name) == 3 &&
832 !strcmp(name, "lz4")) {
833 F2FS_OPTION(sbi).compress_algorithm =
834 COMPRESS_LZ4;
835 } else {
836 kfree(name);
837 return -EINVAL;
838 }
839 kfree(name);
840 break;
841 case Opt_compress_log_size:
842 if (!f2fs_sb_has_compression(sbi)) {
843 f2fs_err(sbi, "Compression feature is off");
844 return -EINVAL;
845 }
846 if (args->from && match_int(args, &arg))
847 return -EINVAL;
848 if (arg < MIN_COMPRESS_LOG_SIZE ||
849 arg > MAX_COMPRESS_LOG_SIZE) {
850 f2fs_err(sbi,
851 "Compress cluster log size is out of range");
852 return -EINVAL;
853 }
854 F2FS_OPTION(sbi).compress_log_size = arg;
855 break;
856 case Opt_compress_extension:
857 if (!f2fs_sb_has_compression(sbi)) {
858 f2fs_err(sbi, "Compression feature is off");
859 return -EINVAL;
860 }
861 name = match_strdup(&args[0]);
862 if (!name)
863 return -ENOMEM;
864
865 ext = F2FS_OPTION(sbi).extensions;
866 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
867
868 if (strlen(name) >= F2FS_EXTENSION_LEN ||
869 ext_cnt >= COMPRESS_EXT_NUM) {
870 f2fs_err(sbi,
871 "invalid extension length/number");
872 kfree(name);
873 return -EINVAL;
874 }
875
876 strcpy(ext[ext_cnt], name);
877 F2FS_OPTION(sbi).compress_ext_cnt++;
878 kfree(name);
879 break;
880 default:
881 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
882 p);
883 return -EINVAL;
884 }
885 }
886 #ifdef CONFIG_QUOTA
887 if (f2fs_check_quota_options(sbi))
888 return -EINVAL;
889 #else
890 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
891 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
892 return -EINVAL;
893 }
894 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
895 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
896 return -EINVAL;
897 }
898 #endif
899 #ifndef CONFIG_UNICODE
900 if (f2fs_sb_has_casefold(sbi)) {
901 f2fs_err(sbi,
902 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
903 return -EINVAL;
904 }
905 #endif
906
907 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
908 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
909 F2FS_IO_SIZE_KB(sbi));
910 return -EINVAL;
911 }
912
913 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
914 int min_size, max_size;
915
916 if (!f2fs_sb_has_extra_attr(sbi) ||
917 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
918 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
919 return -EINVAL;
920 }
921 if (!test_opt(sbi, INLINE_XATTR)) {
922 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
923 return -EINVAL;
924 }
925
926 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
927 max_size = MAX_INLINE_XATTR_SIZE;
928
929 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
930 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
931 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
932 min_size, max_size);
933 return -EINVAL;
934 }
935 }
936
937 if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
938 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
939 return -EINVAL;
940 }
941
942 /* Not pass down write hints if the number of active logs is lesser
943 * than NR_CURSEG_TYPE.
944 */
945 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
946 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
947 return 0;
948 }
949
950 static struct inode *f2fs_alloc_inode(struct super_block *sb)
951 {
952 struct f2fs_inode_info *fi;
953
954 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
955 if (!fi)
956 return NULL;
957
958 init_once((void *) fi);
959
960 /* Initialize f2fs-specific inode info */
961 atomic_set(&fi->dirty_pages, 0);
962 init_rwsem(&fi->i_sem);
963 INIT_LIST_HEAD(&fi->dirty_list);
964 INIT_LIST_HEAD(&fi->gdirty_list);
965 INIT_LIST_HEAD(&fi->inmem_ilist);
966 INIT_LIST_HEAD(&fi->inmem_pages);
967 mutex_init(&fi->inmem_lock);
968 init_rwsem(&fi->i_gc_rwsem[READ]);
969 init_rwsem(&fi->i_gc_rwsem[WRITE]);
970 init_rwsem(&fi->i_mmap_sem);
971 init_rwsem(&fi->i_xattr_sem);
972
973 /* Will be used by directory only */
974 fi->i_dir_level = F2FS_SB(sb)->dir_level;
975
976 return &fi->vfs_inode;
977 }
978
979 static int f2fs_drop_inode(struct inode *inode)
980 {
981 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
982 int ret;
983
984 /*
985 * during filesystem shutdown, if checkpoint is disabled,
986 * drop useless meta/node dirty pages.
987 */
988 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
989 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
990 inode->i_ino == F2FS_META_INO(sbi)) {
991 trace_f2fs_drop_inode(inode, 1);
992 return 1;
993 }
994 }
995
996 /*
997 * This is to avoid a deadlock condition like below.
998 * writeback_single_inode(inode)
999 * - f2fs_write_data_page
1000 * - f2fs_gc -> iput -> evict
1001 * - inode_wait_for_writeback(inode)
1002 */
1003 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1004 if (!inode->i_nlink && !is_bad_inode(inode)) {
1005 /* to avoid evict_inode call simultaneously */
1006 atomic_inc(&inode->i_count);
1007 spin_unlock(&inode->i_lock);
1008
1009 /* some remained atomic pages should discarded */
1010 if (f2fs_is_atomic_file(inode))
1011 f2fs_drop_inmem_pages(inode);
1012
1013 /* should remain fi->extent_tree for writepage */
1014 f2fs_destroy_extent_node(inode);
1015
1016 sb_start_intwrite(inode->i_sb);
1017 f2fs_i_size_write(inode, 0);
1018
1019 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1020 inode, NULL, 0, DATA);
1021 truncate_inode_pages_final(inode->i_mapping);
1022
1023 if (F2FS_HAS_BLOCKS(inode))
1024 f2fs_truncate(inode);
1025
1026 sb_end_intwrite(inode->i_sb);
1027
1028 spin_lock(&inode->i_lock);
1029 atomic_dec(&inode->i_count);
1030 }
1031 trace_f2fs_drop_inode(inode, 0);
1032 return 0;
1033 }
1034 ret = generic_drop_inode(inode);
1035 if (!ret)
1036 ret = fscrypt_drop_inode(inode);
1037 trace_f2fs_drop_inode(inode, ret);
1038 return ret;
1039 }
1040
1041 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1042 {
1043 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1044 int ret = 0;
1045
1046 spin_lock(&sbi->inode_lock[DIRTY_META]);
1047 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1048 ret = 1;
1049 } else {
1050 set_inode_flag(inode, FI_DIRTY_INODE);
1051 stat_inc_dirty_inode(sbi, DIRTY_META);
1052 }
1053 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1054 list_add_tail(&F2FS_I(inode)->gdirty_list,
1055 &sbi->inode_list[DIRTY_META]);
1056 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1057 }
1058 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1059 return ret;
1060 }
1061
1062 void f2fs_inode_synced(struct inode *inode)
1063 {
1064 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1065
1066 spin_lock(&sbi->inode_lock[DIRTY_META]);
1067 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1068 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1069 return;
1070 }
1071 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1072 list_del_init(&F2FS_I(inode)->gdirty_list);
1073 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1074 }
1075 clear_inode_flag(inode, FI_DIRTY_INODE);
1076 clear_inode_flag(inode, FI_AUTO_RECOVER);
1077 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1078 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1079 }
1080
1081 /*
1082 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1083 *
1084 * We should call set_dirty_inode to write the dirty inode through write_inode.
1085 */
1086 static void f2fs_dirty_inode(struct inode *inode, int flags)
1087 {
1088 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1089
1090 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1091 inode->i_ino == F2FS_META_INO(sbi))
1092 return;
1093
1094 if (flags == I_DIRTY_TIME)
1095 return;
1096
1097 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1098 clear_inode_flag(inode, FI_AUTO_RECOVER);
1099
1100 f2fs_inode_dirtied(inode, false);
1101 }
1102
1103 static void f2fs_free_inode(struct inode *inode)
1104 {
1105 fscrypt_free_inode(inode);
1106 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1107 }
1108
1109 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1110 {
1111 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1112 percpu_counter_destroy(&sbi->total_valid_inode_count);
1113 }
1114
1115 static void destroy_device_list(struct f2fs_sb_info *sbi)
1116 {
1117 int i;
1118
1119 for (i = 0; i < sbi->s_ndevs; i++) {
1120 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1121 #ifdef CONFIG_BLK_DEV_ZONED
1122 kvfree(FDEV(i).blkz_seq);
1123 #endif
1124 }
1125 kvfree(sbi->devs);
1126 }
1127
1128 static void f2fs_put_super(struct super_block *sb)
1129 {
1130 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1131 int i;
1132 bool dropped;
1133
1134 f2fs_quota_off_umount(sb);
1135
1136 /* prevent remaining shrinker jobs */
1137 mutex_lock(&sbi->umount_mutex);
1138
1139 /*
1140 * We don't need to do checkpoint when superblock is clean.
1141 * But, the previous checkpoint was not done by umount, it needs to do
1142 * clean checkpoint again.
1143 */
1144 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1145 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1146 struct cp_control cpc = {
1147 .reason = CP_UMOUNT,
1148 };
1149 f2fs_write_checkpoint(sbi, &cpc);
1150 }
1151
1152 /* be sure to wait for any on-going discard commands */
1153 dropped = f2fs_issue_discard_timeout(sbi);
1154
1155 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1156 !sbi->discard_blks && !dropped) {
1157 struct cp_control cpc = {
1158 .reason = CP_UMOUNT | CP_TRIMMED,
1159 };
1160 f2fs_write_checkpoint(sbi, &cpc);
1161 }
1162
1163 /*
1164 * normally superblock is clean, so we need to release this.
1165 * In addition, EIO will skip do checkpoint, we need this as well.
1166 */
1167 f2fs_release_ino_entry(sbi, true);
1168
1169 f2fs_leave_shrinker(sbi);
1170 mutex_unlock(&sbi->umount_mutex);
1171
1172 /* our cp_error case, we can wait for any writeback page */
1173 f2fs_flush_merged_writes(sbi);
1174
1175 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1176
1177 f2fs_bug_on(sbi, sbi->fsync_node_num);
1178
1179 iput(sbi->node_inode);
1180 sbi->node_inode = NULL;
1181
1182 iput(sbi->meta_inode);
1183 sbi->meta_inode = NULL;
1184
1185 /*
1186 * iput() can update stat information, if f2fs_write_checkpoint()
1187 * above failed with error.
1188 */
1189 f2fs_destroy_stats(sbi);
1190
1191 /* destroy f2fs internal modules */
1192 f2fs_destroy_node_manager(sbi);
1193 f2fs_destroy_segment_manager(sbi);
1194
1195 f2fs_destroy_post_read_wq(sbi);
1196
1197 kvfree(sbi->ckpt);
1198
1199 f2fs_unregister_sysfs(sbi);
1200
1201 sb->s_fs_info = NULL;
1202 if (sbi->s_chksum_driver)
1203 crypto_free_shash(sbi->s_chksum_driver);
1204 kvfree(sbi->raw_super);
1205
1206 destroy_device_list(sbi);
1207 mempool_destroy(sbi->write_io_dummy);
1208 #ifdef CONFIG_QUOTA
1209 for (i = 0; i < MAXQUOTAS; i++)
1210 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1211 #endif
1212 destroy_percpu_info(sbi);
1213 for (i = 0; i < NR_PAGE_TYPE; i++)
1214 kvfree(sbi->write_io[i]);
1215 #ifdef CONFIG_UNICODE
1216 utf8_unload(sbi->s_encoding);
1217 #endif
1218 kvfree(sbi);
1219 }
1220
1221 int f2fs_sync_fs(struct super_block *sb, int sync)
1222 {
1223 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1224 int err = 0;
1225
1226 if (unlikely(f2fs_cp_error(sbi)))
1227 return 0;
1228 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1229 return 0;
1230
1231 trace_f2fs_sync_fs(sb, sync);
1232
1233 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1234 return -EAGAIN;
1235
1236 if (sync) {
1237 struct cp_control cpc;
1238
1239 cpc.reason = __get_cp_reason(sbi);
1240
1241 down_write(&sbi->gc_lock);
1242 err = f2fs_write_checkpoint(sbi, &cpc);
1243 up_write(&sbi->gc_lock);
1244 }
1245 f2fs_trace_ios(NULL, 1);
1246
1247 return err;
1248 }
1249
1250 static int f2fs_freeze(struct super_block *sb)
1251 {
1252 if (f2fs_readonly(sb))
1253 return 0;
1254
1255 /* IO error happened before */
1256 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1257 return -EIO;
1258
1259 /* must be clean, since sync_filesystem() was already called */
1260 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1261 return -EINVAL;
1262 return 0;
1263 }
1264
1265 static int f2fs_unfreeze(struct super_block *sb)
1266 {
1267 return 0;
1268 }
1269
1270 #ifdef CONFIG_QUOTA
1271 static int f2fs_statfs_project(struct super_block *sb,
1272 kprojid_t projid, struct kstatfs *buf)
1273 {
1274 struct kqid qid;
1275 struct dquot *dquot;
1276 u64 limit;
1277 u64 curblock;
1278
1279 qid = make_kqid_projid(projid);
1280 dquot = dqget(sb, qid);
1281 if (IS_ERR(dquot))
1282 return PTR_ERR(dquot);
1283 spin_lock(&dquot->dq_dqb_lock);
1284
1285 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1286 dquot->dq_dqb.dqb_bhardlimit);
1287 if (limit)
1288 limit >>= sb->s_blocksize_bits;
1289
1290 if (limit && buf->f_blocks > limit) {
1291 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1292 buf->f_blocks = limit;
1293 buf->f_bfree = buf->f_bavail =
1294 (buf->f_blocks > curblock) ?
1295 (buf->f_blocks - curblock) : 0;
1296 }
1297
1298 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1299 dquot->dq_dqb.dqb_ihardlimit);
1300
1301 if (limit && buf->f_files > limit) {
1302 buf->f_files = limit;
1303 buf->f_ffree =
1304 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1305 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1306 }
1307
1308 spin_unlock(&dquot->dq_dqb_lock);
1309 dqput(dquot);
1310 return 0;
1311 }
1312 #endif
1313
1314 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1315 {
1316 struct super_block *sb = dentry->d_sb;
1317 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1318 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1319 block_t total_count, user_block_count, start_count;
1320 u64 avail_node_count;
1321
1322 total_count = le64_to_cpu(sbi->raw_super->block_count);
1323 user_block_count = sbi->user_block_count;
1324 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1325 buf->f_type = F2FS_SUPER_MAGIC;
1326 buf->f_bsize = sbi->blocksize;
1327
1328 buf->f_blocks = total_count - start_count;
1329 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1330 sbi->current_reserved_blocks;
1331
1332 spin_lock(&sbi->stat_lock);
1333 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1334 buf->f_bfree = 0;
1335 else
1336 buf->f_bfree -= sbi->unusable_block_count;
1337 spin_unlock(&sbi->stat_lock);
1338
1339 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1340 buf->f_bavail = buf->f_bfree -
1341 F2FS_OPTION(sbi).root_reserved_blocks;
1342 else
1343 buf->f_bavail = 0;
1344
1345 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1346
1347 if (avail_node_count > user_block_count) {
1348 buf->f_files = user_block_count;
1349 buf->f_ffree = buf->f_bavail;
1350 } else {
1351 buf->f_files = avail_node_count;
1352 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1353 buf->f_bavail);
1354 }
1355
1356 buf->f_namelen = F2FS_NAME_LEN;
1357 buf->f_fsid.val[0] = (u32)id;
1358 buf->f_fsid.val[1] = (u32)(id >> 32);
1359
1360 #ifdef CONFIG_QUOTA
1361 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1362 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1363 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1364 }
1365 #endif
1366 return 0;
1367 }
1368
1369 static inline void f2fs_show_quota_options(struct seq_file *seq,
1370 struct super_block *sb)
1371 {
1372 #ifdef CONFIG_QUOTA
1373 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1374
1375 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1376 char *fmtname = "";
1377
1378 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1379 case QFMT_VFS_OLD:
1380 fmtname = "vfsold";
1381 break;
1382 case QFMT_VFS_V0:
1383 fmtname = "vfsv0";
1384 break;
1385 case QFMT_VFS_V1:
1386 fmtname = "vfsv1";
1387 break;
1388 }
1389 seq_printf(seq, ",jqfmt=%s", fmtname);
1390 }
1391
1392 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1393 seq_show_option(seq, "usrjquota",
1394 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1395
1396 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1397 seq_show_option(seq, "grpjquota",
1398 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1399
1400 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1401 seq_show_option(seq, "prjjquota",
1402 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1403 #endif
1404 }
1405
1406 static inline void f2fs_show_compress_options(struct seq_file *seq,
1407 struct super_block *sb)
1408 {
1409 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1410 char *algtype = "";
1411 int i;
1412
1413 if (!f2fs_sb_has_compression(sbi))
1414 return;
1415
1416 switch (F2FS_OPTION(sbi).compress_algorithm) {
1417 case COMPRESS_LZO:
1418 algtype = "lzo";
1419 break;
1420 case COMPRESS_LZ4:
1421 algtype = "lz4";
1422 break;
1423 }
1424 seq_printf(seq, ",compress_algorithm=%s", algtype);
1425
1426 seq_printf(seq, ",compress_log_size=%u",
1427 F2FS_OPTION(sbi).compress_log_size);
1428
1429 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1430 seq_printf(seq, ",compress_extension=%s",
1431 F2FS_OPTION(sbi).extensions[i]);
1432 }
1433 }
1434
1435 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1436 {
1437 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1438
1439 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1440 if (test_opt(sbi, FORCE_FG_GC))
1441 seq_printf(seq, ",background_gc=%s", "sync");
1442 else
1443 seq_printf(seq, ",background_gc=%s", "on");
1444 } else {
1445 seq_printf(seq, ",background_gc=%s", "off");
1446 }
1447 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1448 seq_puts(seq, ",disable_roll_forward");
1449 if (test_opt(sbi, NORECOVERY))
1450 seq_puts(seq, ",norecovery");
1451 if (test_opt(sbi, DISCARD))
1452 seq_puts(seq, ",discard");
1453 else
1454 seq_puts(seq, ",nodiscard");
1455 if (test_opt(sbi, NOHEAP))
1456 seq_puts(seq, ",no_heap");
1457 else
1458 seq_puts(seq, ",heap");
1459 #ifdef CONFIG_F2FS_FS_XATTR
1460 if (test_opt(sbi, XATTR_USER))
1461 seq_puts(seq, ",user_xattr");
1462 else
1463 seq_puts(seq, ",nouser_xattr");
1464 if (test_opt(sbi, INLINE_XATTR))
1465 seq_puts(seq, ",inline_xattr");
1466 else
1467 seq_puts(seq, ",noinline_xattr");
1468 if (test_opt(sbi, INLINE_XATTR_SIZE))
1469 seq_printf(seq, ",inline_xattr_size=%u",
1470 F2FS_OPTION(sbi).inline_xattr_size);
1471 #endif
1472 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1473 if (test_opt(sbi, POSIX_ACL))
1474 seq_puts(seq, ",acl");
1475 else
1476 seq_puts(seq, ",noacl");
1477 #endif
1478 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1479 seq_puts(seq, ",disable_ext_identify");
1480 if (test_opt(sbi, INLINE_DATA))
1481 seq_puts(seq, ",inline_data");
1482 else
1483 seq_puts(seq, ",noinline_data");
1484 if (test_opt(sbi, INLINE_DENTRY))
1485 seq_puts(seq, ",inline_dentry");
1486 else
1487 seq_puts(seq, ",noinline_dentry");
1488 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1489 seq_puts(seq, ",flush_merge");
1490 if (test_opt(sbi, NOBARRIER))
1491 seq_puts(seq, ",nobarrier");
1492 if (test_opt(sbi, FASTBOOT))
1493 seq_puts(seq, ",fastboot");
1494 if (test_opt(sbi, EXTENT_CACHE))
1495 seq_puts(seq, ",extent_cache");
1496 else
1497 seq_puts(seq, ",noextent_cache");
1498 if (test_opt(sbi, DATA_FLUSH))
1499 seq_puts(seq, ",data_flush");
1500
1501 seq_puts(seq, ",mode=");
1502 if (test_opt(sbi, ADAPTIVE))
1503 seq_puts(seq, "adaptive");
1504 else if (test_opt(sbi, LFS))
1505 seq_puts(seq, "lfs");
1506 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1507 if (test_opt(sbi, RESERVE_ROOT))
1508 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1509 F2FS_OPTION(sbi).root_reserved_blocks,
1510 from_kuid_munged(&init_user_ns,
1511 F2FS_OPTION(sbi).s_resuid),
1512 from_kgid_munged(&init_user_ns,
1513 F2FS_OPTION(sbi).s_resgid));
1514 if (F2FS_IO_SIZE_BITS(sbi))
1515 seq_printf(seq, ",io_bits=%u",
1516 F2FS_OPTION(sbi).write_io_size_bits);
1517 #ifdef CONFIG_F2FS_FAULT_INJECTION
1518 if (test_opt(sbi, FAULT_INJECTION)) {
1519 seq_printf(seq, ",fault_injection=%u",
1520 F2FS_OPTION(sbi).fault_info.inject_rate);
1521 seq_printf(seq, ",fault_type=%u",
1522 F2FS_OPTION(sbi).fault_info.inject_type);
1523 }
1524 #endif
1525 #ifdef CONFIG_QUOTA
1526 if (test_opt(sbi, QUOTA))
1527 seq_puts(seq, ",quota");
1528 if (test_opt(sbi, USRQUOTA))
1529 seq_puts(seq, ",usrquota");
1530 if (test_opt(sbi, GRPQUOTA))
1531 seq_puts(seq, ",grpquota");
1532 if (test_opt(sbi, PRJQUOTA))
1533 seq_puts(seq, ",prjquota");
1534 #endif
1535 f2fs_show_quota_options(seq, sbi->sb);
1536 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1537 seq_printf(seq, ",whint_mode=%s", "user-based");
1538 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1539 seq_printf(seq, ",whint_mode=%s", "fs-based");
1540 #ifdef CONFIG_FS_ENCRYPTION
1541 if (F2FS_OPTION(sbi).test_dummy_encryption)
1542 seq_puts(seq, ",test_dummy_encryption");
1543 #endif
1544
1545 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1546 seq_printf(seq, ",alloc_mode=%s", "default");
1547 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1548 seq_printf(seq, ",alloc_mode=%s", "reuse");
1549
1550 if (test_opt(sbi, DISABLE_CHECKPOINT))
1551 seq_printf(seq, ",checkpoint=disable:%u",
1552 F2FS_OPTION(sbi).unusable_cap);
1553 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1554 seq_printf(seq, ",fsync_mode=%s", "posix");
1555 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1556 seq_printf(seq, ",fsync_mode=%s", "strict");
1557 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1558 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1559
1560 f2fs_show_compress_options(seq, sbi->sb);
1561 return 0;
1562 }
1563
1564 static void default_options(struct f2fs_sb_info *sbi)
1565 {
1566 /* init some FS parameters */
1567 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1568 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1569 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1570 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1571 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1572 F2FS_OPTION(sbi).test_dummy_encryption = false;
1573 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1574 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1575 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZO;
1576 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1577 F2FS_OPTION(sbi).compress_ext_cnt = 0;
1578
1579 set_opt(sbi, BG_GC);
1580 set_opt(sbi, INLINE_XATTR);
1581 set_opt(sbi, INLINE_DATA);
1582 set_opt(sbi, INLINE_DENTRY);
1583 set_opt(sbi, EXTENT_CACHE);
1584 set_opt(sbi, NOHEAP);
1585 clear_opt(sbi, DISABLE_CHECKPOINT);
1586 F2FS_OPTION(sbi).unusable_cap = 0;
1587 sbi->sb->s_flags |= SB_LAZYTIME;
1588 set_opt(sbi, FLUSH_MERGE);
1589 set_opt(sbi, DISCARD);
1590 if (f2fs_sb_has_blkzoned(sbi))
1591 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1592 else
1593 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1594
1595 #ifdef CONFIG_F2FS_FS_XATTR
1596 set_opt(sbi, XATTR_USER);
1597 #endif
1598 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1599 set_opt(sbi, POSIX_ACL);
1600 #endif
1601
1602 f2fs_build_fault_attr(sbi, 0, 0);
1603 }
1604
1605 #ifdef CONFIG_QUOTA
1606 static int f2fs_enable_quotas(struct super_block *sb);
1607 #endif
1608
1609 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1610 {
1611 unsigned int s_flags = sbi->sb->s_flags;
1612 struct cp_control cpc;
1613 int err = 0;
1614 int ret;
1615 block_t unusable;
1616
1617 if (s_flags & SB_RDONLY) {
1618 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1619 return -EINVAL;
1620 }
1621 sbi->sb->s_flags |= SB_ACTIVE;
1622
1623 f2fs_update_time(sbi, DISABLE_TIME);
1624
1625 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1626 down_write(&sbi->gc_lock);
1627 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1628 if (err == -ENODATA) {
1629 err = 0;
1630 break;
1631 }
1632 if (err && err != -EAGAIN)
1633 break;
1634 }
1635
1636 ret = sync_filesystem(sbi->sb);
1637 if (ret || err) {
1638 err = ret ? ret: err;
1639 goto restore_flag;
1640 }
1641
1642 unusable = f2fs_get_unusable_blocks(sbi);
1643 if (f2fs_disable_cp_again(sbi, unusable)) {
1644 err = -EAGAIN;
1645 goto restore_flag;
1646 }
1647
1648 down_write(&sbi->gc_lock);
1649 cpc.reason = CP_PAUSE;
1650 set_sbi_flag(sbi, SBI_CP_DISABLED);
1651 err = f2fs_write_checkpoint(sbi, &cpc);
1652 if (err)
1653 goto out_unlock;
1654
1655 spin_lock(&sbi->stat_lock);
1656 sbi->unusable_block_count = unusable;
1657 spin_unlock(&sbi->stat_lock);
1658
1659 out_unlock:
1660 up_write(&sbi->gc_lock);
1661 restore_flag:
1662 sbi->sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1663 return err;
1664 }
1665
1666 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1667 {
1668 down_write(&sbi->gc_lock);
1669 f2fs_dirty_to_prefree(sbi);
1670
1671 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1672 set_sbi_flag(sbi, SBI_IS_DIRTY);
1673 up_write(&sbi->gc_lock);
1674
1675 f2fs_sync_fs(sbi->sb, 1);
1676 }
1677
1678 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1679 {
1680 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1681 struct f2fs_mount_info org_mount_opt;
1682 unsigned long old_sb_flags;
1683 int err;
1684 bool need_restart_gc = false;
1685 bool need_stop_gc = false;
1686 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1687 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1688 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1689 bool checkpoint_changed;
1690 #ifdef CONFIG_QUOTA
1691 int i, j;
1692 #endif
1693
1694 /*
1695 * Save the old mount options in case we
1696 * need to restore them.
1697 */
1698 org_mount_opt = sbi->mount_opt;
1699 old_sb_flags = sb->s_flags;
1700
1701 #ifdef CONFIG_QUOTA
1702 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1703 for (i = 0; i < MAXQUOTAS; i++) {
1704 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1705 org_mount_opt.s_qf_names[i] =
1706 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1707 GFP_KERNEL);
1708 if (!org_mount_opt.s_qf_names[i]) {
1709 for (j = 0; j < i; j++)
1710 kvfree(org_mount_opt.s_qf_names[j]);
1711 return -ENOMEM;
1712 }
1713 } else {
1714 org_mount_opt.s_qf_names[i] = NULL;
1715 }
1716 }
1717 #endif
1718
1719 /* recover superblocks we couldn't write due to previous RO mount */
1720 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1721 err = f2fs_commit_super(sbi, false);
1722 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1723 err);
1724 if (!err)
1725 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1726 }
1727
1728 default_options(sbi);
1729
1730 /* parse mount options */
1731 err = parse_options(sb, data);
1732 if (err)
1733 goto restore_opts;
1734 checkpoint_changed =
1735 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1736
1737 /*
1738 * Previous and new state of filesystem is RO,
1739 * so skip checking GC and FLUSH_MERGE conditions.
1740 */
1741 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1742 goto skip;
1743
1744 #ifdef CONFIG_QUOTA
1745 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1746 err = dquot_suspend(sb, -1);
1747 if (err < 0)
1748 goto restore_opts;
1749 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1750 /* dquot_resume needs RW */
1751 sb->s_flags &= ~SB_RDONLY;
1752 if (sb_any_quota_suspended(sb)) {
1753 dquot_resume(sb, -1);
1754 } else if (f2fs_sb_has_quota_ino(sbi)) {
1755 err = f2fs_enable_quotas(sb);
1756 if (err)
1757 goto restore_opts;
1758 }
1759 }
1760 #endif
1761 /* disallow enable/disable extent_cache dynamically */
1762 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1763 err = -EINVAL;
1764 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1765 goto restore_opts;
1766 }
1767
1768 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1769 err = -EINVAL;
1770 f2fs_warn(sbi, "switch io_bits option is not allowed");
1771 goto restore_opts;
1772 }
1773
1774 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1775 err = -EINVAL;
1776 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1777 goto restore_opts;
1778 }
1779
1780 /*
1781 * We stop the GC thread if FS is mounted as RO
1782 * or if background_gc = off is passed in mount
1783 * option. Also sync the filesystem.
1784 */
1785 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1786 if (sbi->gc_thread) {
1787 f2fs_stop_gc_thread(sbi);
1788 need_restart_gc = true;
1789 }
1790 } else if (!sbi->gc_thread) {
1791 err = f2fs_start_gc_thread(sbi);
1792 if (err)
1793 goto restore_opts;
1794 need_stop_gc = true;
1795 }
1796
1797 if (*flags & SB_RDONLY ||
1798 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1799 writeback_inodes_sb(sb, WB_REASON_SYNC);
1800 sync_inodes_sb(sb);
1801
1802 set_sbi_flag(sbi, SBI_IS_DIRTY);
1803 set_sbi_flag(sbi, SBI_IS_CLOSE);
1804 f2fs_sync_fs(sb, 1);
1805 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1806 }
1807
1808 if (checkpoint_changed) {
1809 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1810 err = f2fs_disable_checkpoint(sbi);
1811 if (err)
1812 goto restore_gc;
1813 } else {
1814 f2fs_enable_checkpoint(sbi);
1815 }
1816 }
1817
1818 /*
1819 * We stop issue flush thread if FS is mounted as RO
1820 * or if flush_merge is not passed in mount option.
1821 */
1822 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1823 clear_opt(sbi, FLUSH_MERGE);
1824 f2fs_destroy_flush_cmd_control(sbi, false);
1825 } else {
1826 err = f2fs_create_flush_cmd_control(sbi);
1827 if (err)
1828 goto restore_gc;
1829 }
1830 skip:
1831 #ifdef CONFIG_QUOTA
1832 /* Release old quota file names */
1833 for (i = 0; i < MAXQUOTAS; i++)
1834 kvfree(org_mount_opt.s_qf_names[i]);
1835 #endif
1836 /* Update the POSIXACL Flag */
1837 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1838 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1839
1840 limit_reserve_root(sbi);
1841 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1842 return 0;
1843 restore_gc:
1844 if (need_restart_gc) {
1845 if (f2fs_start_gc_thread(sbi))
1846 f2fs_warn(sbi, "background gc thread has stopped");
1847 } else if (need_stop_gc) {
1848 f2fs_stop_gc_thread(sbi);
1849 }
1850 restore_opts:
1851 #ifdef CONFIG_QUOTA
1852 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1853 for (i = 0; i < MAXQUOTAS; i++) {
1854 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1855 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1856 }
1857 #endif
1858 sbi->mount_opt = org_mount_opt;
1859 sb->s_flags = old_sb_flags;
1860 return err;
1861 }
1862
1863 #ifdef CONFIG_QUOTA
1864 /* Read data from quotafile */
1865 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1866 size_t len, loff_t off)
1867 {
1868 struct inode *inode = sb_dqopt(sb)->files[type];
1869 struct address_space *mapping = inode->i_mapping;
1870 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1871 int offset = off & (sb->s_blocksize - 1);
1872 int tocopy;
1873 size_t toread;
1874 loff_t i_size = i_size_read(inode);
1875 struct page *page;
1876 char *kaddr;
1877
1878 if (off > i_size)
1879 return 0;
1880
1881 if (off + len > i_size)
1882 len = i_size - off;
1883 toread = len;
1884 while (toread > 0) {
1885 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1886 repeat:
1887 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1888 if (IS_ERR(page)) {
1889 if (PTR_ERR(page) == -ENOMEM) {
1890 congestion_wait(BLK_RW_ASYNC, HZ/50);
1891 goto repeat;
1892 }
1893 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1894 return PTR_ERR(page);
1895 }
1896
1897 lock_page(page);
1898
1899 if (unlikely(page->mapping != mapping)) {
1900 f2fs_put_page(page, 1);
1901 goto repeat;
1902 }
1903 if (unlikely(!PageUptodate(page))) {
1904 f2fs_put_page(page, 1);
1905 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1906 return -EIO;
1907 }
1908
1909 kaddr = kmap_atomic(page);
1910 memcpy(data, kaddr + offset, tocopy);
1911 kunmap_atomic(kaddr);
1912 f2fs_put_page(page, 1);
1913
1914 offset = 0;
1915 toread -= tocopy;
1916 data += tocopy;
1917 blkidx++;
1918 }
1919 return len;
1920 }
1921
1922 /* Write to quotafile */
1923 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1924 const char *data, size_t len, loff_t off)
1925 {
1926 struct inode *inode = sb_dqopt(sb)->files[type];
1927 struct address_space *mapping = inode->i_mapping;
1928 const struct address_space_operations *a_ops = mapping->a_ops;
1929 int offset = off & (sb->s_blocksize - 1);
1930 size_t towrite = len;
1931 struct page *page;
1932 void *fsdata = NULL;
1933 char *kaddr;
1934 int err = 0;
1935 int tocopy;
1936
1937 while (towrite > 0) {
1938 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1939 towrite);
1940 retry:
1941 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1942 &page, &fsdata);
1943 if (unlikely(err)) {
1944 if (err == -ENOMEM) {
1945 congestion_wait(BLK_RW_ASYNC, HZ/50);
1946 goto retry;
1947 }
1948 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1949 break;
1950 }
1951
1952 kaddr = kmap_atomic(page);
1953 memcpy(kaddr + offset, data, tocopy);
1954 kunmap_atomic(kaddr);
1955 flush_dcache_page(page);
1956
1957 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1958 page, fsdata);
1959 offset = 0;
1960 towrite -= tocopy;
1961 off += tocopy;
1962 data += tocopy;
1963 cond_resched();
1964 }
1965
1966 if (len == towrite)
1967 return err;
1968 inode->i_mtime = inode->i_ctime = current_time(inode);
1969 f2fs_mark_inode_dirty_sync(inode, false);
1970 return len - towrite;
1971 }
1972
1973 static struct dquot **f2fs_get_dquots(struct inode *inode)
1974 {
1975 return F2FS_I(inode)->i_dquot;
1976 }
1977
1978 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1979 {
1980 return &F2FS_I(inode)->i_reserved_quota;
1981 }
1982
1983 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1984 {
1985 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1986 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1987 return 0;
1988 }
1989
1990 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1991 F2FS_OPTION(sbi).s_jquota_fmt, type);
1992 }
1993
1994 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1995 {
1996 int enabled = 0;
1997 int i, err;
1998
1999 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2000 err = f2fs_enable_quotas(sbi->sb);
2001 if (err) {
2002 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2003 return 0;
2004 }
2005 return 1;
2006 }
2007
2008 for (i = 0; i < MAXQUOTAS; i++) {
2009 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2010 err = f2fs_quota_on_mount(sbi, i);
2011 if (!err) {
2012 enabled = 1;
2013 continue;
2014 }
2015 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2016 err, i);
2017 }
2018 }
2019 return enabled;
2020 }
2021
2022 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2023 unsigned int flags)
2024 {
2025 struct inode *qf_inode;
2026 unsigned long qf_inum;
2027 int err;
2028
2029 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2030
2031 qf_inum = f2fs_qf_ino(sb, type);
2032 if (!qf_inum)
2033 return -EPERM;
2034
2035 qf_inode = f2fs_iget(sb, qf_inum);
2036 if (IS_ERR(qf_inode)) {
2037 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2038 return PTR_ERR(qf_inode);
2039 }
2040
2041 /* Don't account quota for quota files to avoid recursion */
2042 qf_inode->i_flags |= S_NOQUOTA;
2043 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2044 iput(qf_inode);
2045 return err;
2046 }
2047
2048 static int f2fs_enable_quotas(struct super_block *sb)
2049 {
2050 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2051 int type, err = 0;
2052 unsigned long qf_inum;
2053 bool quota_mopt[MAXQUOTAS] = {
2054 test_opt(sbi, USRQUOTA),
2055 test_opt(sbi, GRPQUOTA),
2056 test_opt(sbi, PRJQUOTA),
2057 };
2058
2059 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2060 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2061 return 0;
2062 }
2063
2064 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2065
2066 for (type = 0; type < MAXQUOTAS; type++) {
2067 qf_inum = f2fs_qf_ino(sb, type);
2068 if (qf_inum) {
2069 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2070 DQUOT_USAGE_ENABLED |
2071 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2072 if (err) {
2073 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2074 type, err);
2075 for (type--; type >= 0; type--)
2076 dquot_quota_off(sb, type);
2077 set_sbi_flag(F2FS_SB(sb),
2078 SBI_QUOTA_NEED_REPAIR);
2079 return err;
2080 }
2081 }
2082 }
2083 return 0;
2084 }
2085
2086 int f2fs_quota_sync(struct super_block *sb, int type)
2087 {
2088 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2089 struct quota_info *dqopt = sb_dqopt(sb);
2090 int cnt;
2091 int ret;
2092
2093 /*
2094 * do_quotactl
2095 * f2fs_quota_sync
2096 * down_read(quota_sem)
2097 * dquot_writeback_dquots()
2098 * f2fs_dquot_commit
2099 * block_operation
2100 * down_read(quota_sem)
2101 */
2102 f2fs_lock_op(sbi);
2103
2104 down_read(&sbi->quota_sem);
2105 ret = dquot_writeback_dquots(sb, type);
2106 if (ret)
2107 goto out;
2108
2109 /*
2110 * Now when everything is written we can discard the pagecache so
2111 * that userspace sees the changes.
2112 */
2113 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2114 struct address_space *mapping;
2115
2116 if (type != -1 && cnt != type)
2117 continue;
2118 if (!sb_has_quota_active(sb, cnt))
2119 continue;
2120
2121 mapping = dqopt->files[cnt]->i_mapping;
2122
2123 ret = filemap_fdatawrite(mapping);
2124 if (ret)
2125 goto out;
2126
2127 /* if we are using journalled quota */
2128 if (is_journalled_quota(sbi))
2129 continue;
2130
2131 ret = filemap_fdatawait(mapping);
2132 if (ret)
2133 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2134
2135 inode_lock(dqopt->files[cnt]);
2136 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2137 inode_unlock(dqopt->files[cnt]);
2138 }
2139 out:
2140 if (ret)
2141 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2142 up_read(&sbi->quota_sem);
2143 f2fs_unlock_op(sbi);
2144 return ret;
2145 }
2146
2147 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2148 const struct path *path)
2149 {
2150 struct inode *inode;
2151 int err;
2152
2153 /* if quota sysfile exists, deny enabling quota with specific file */
2154 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2155 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2156 return -EBUSY;
2157 }
2158
2159 err = f2fs_quota_sync(sb, type);
2160 if (err)
2161 return err;
2162
2163 err = dquot_quota_on(sb, type, format_id, path);
2164 if (err)
2165 return err;
2166
2167 inode = d_inode(path->dentry);
2168
2169 inode_lock(inode);
2170 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2171 f2fs_set_inode_flags(inode);
2172 inode_unlock(inode);
2173 f2fs_mark_inode_dirty_sync(inode, false);
2174
2175 return 0;
2176 }
2177
2178 static int __f2fs_quota_off(struct super_block *sb, int type)
2179 {
2180 struct inode *inode = sb_dqopt(sb)->files[type];
2181 int err;
2182
2183 if (!inode || !igrab(inode))
2184 return dquot_quota_off(sb, type);
2185
2186 err = f2fs_quota_sync(sb, type);
2187 if (err)
2188 goto out_put;
2189
2190 err = dquot_quota_off(sb, type);
2191 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2192 goto out_put;
2193
2194 inode_lock(inode);
2195 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2196 f2fs_set_inode_flags(inode);
2197 inode_unlock(inode);
2198 f2fs_mark_inode_dirty_sync(inode, false);
2199 out_put:
2200 iput(inode);
2201 return err;
2202 }
2203
2204 static int f2fs_quota_off(struct super_block *sb, int type)
2205 {
2206 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2207 int err;
2208
2209 err = __f2fs_quota_off(sb, type);
2210
2211 /*
2212 * quotactl can shutdown journalled quota, result in inconsistence
2213 * between quota record and fs data by following updates, tag the
2214 * flag to let fsck be aware of it.
2215 */
2216 if (is_journalled_quota(sbi))
2217 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2218 return err;
2219 }
2220
2221 void f2fs_quota_off_umount(struct super_block *sb)
2222 {
2223 int type;
2224 int err;
2225
2226 for (type = 0; type < MAXQUOTAS; type++) {
2227 err = __f2fs_quota_off(sb, type);
2228 if (err) {
2229 int ret = dquot_quota_off(sb, type);
2230
2231 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2232 type, err, ret);
2233 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2234 }
2235 }
2236 /*
2237 * In case of checkpoint=disable, we must flush quota blocks.
2238 * This can cause NULL exception for node_inode in end_io, since
2239 * put_super already dropped it.
2240 */
2241 sync_filesystem(sb);
2242 }
2243
2244 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2245 {
2246 struct quota_info *dqopt = sb_dqopt(sb);
2247 int type;
2248
2249 for (type = 0; type < MAXQUOTAS; type++) {
2250 if (!dqopt->files[type])
2251 continue;
2252 f2fs_inode_synced(dqopt->files[type]);
2253 }
2254 }
2255
2256 static int f2fs_dquot_commit(struct dquot *dquot)
2257 {
2258 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2259 int ret;
2260
2261 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2262 ret = dquot_commit(dquot);
2263 if (ret < 0)
2264 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2265 up_read(&sbi->quota_sem);
2266 return ret;
2267 }
2268
2269 static int f2fs_dquot_acquire(struct dquot *dquot)
2270 {
2271 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2272 int ret;
2273
2274 down_read(&sbi->quota_sem);
2275 ret = dquot_acquire(dquot);
2276 if (ret < 0)
2277 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2278 up_read(&sbi->quota_sem);
2279 return ret;
2280 }
2281
2282 static int f2fs_dquot_release(struct dquot *dquot)
2283 {
2284 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2285 int ret = dquot_release(dquot);
2286
2287 if (ret < 0)
2288 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2289 return ret;
2290 }
2291
2292 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2293 {
2294 struct super_block *sb = dquot->dq_sb;
2295 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2296 int ret = dquot_mark_dquot_dirty(dquot);
2297
2298 /* if we are using journalled quota */
2299 if (is_journalled_quota(sbi))
2300 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2301
2302 return ret;
2303 }
2304
2305 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2306 {
2307 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2308 int ret = dquot_commit_info(sb, type);
2309
2310 if (ret < 0)
2311 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2312 return ret;
2313 }
2314
2315 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2316 {
2317 *projid = F2FS_I(inode)->i_projid;
2318 return 0;
2319 }
2320
2321 static const struct dquot_operations f2fs_quota_operations = {
2322 .get_reserved_space = f2fs_get_reserved_space,
2323 .write_dquot = f2fs_dquot_commit,
2324 .acquire_dquot = f2fs_dquot_acquire,
2325 .release_dquot = f2fs_dquot_release,
2326 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2327 .write_info = f2fs_dquot_commit_info,
2328 .alloc_dquot = dquot_alloc,
2329 .destroy_dquot = dquot_destroy,
2330 .get_projid = f2fs_get_projid,
2331 .get_next_id = dquot_get_next_id,
2332 };
2333
2334 static const struct quotactl_ops f2fs_quotactl_ops = {
2335 .quota_on = f2fs_quota_on,
2336 .quota_off = f2fs_quota_off,
2337 .quota_sync = f2fs_quota_sync,
2338 .get_state = dquot_get_state,
2339 .set_info = dquot_set_dqinfo,
2340 .get_dqblk = dquot_get_dqblk,
2341 .set_dqblk = dquot_set_dqblk,
2342 .get_nextdqblk = dquot_get_next_dqblk,
2343 };
2344 #else
2345 int f2fs_quota_sync(struct super_block *sb, int type)
2346 {
2347 return 0;
2348 }
2349
2350 void f2fs_quota_off_umount(struct super_block *sb)
2351 {
2352 }
2353 #endif
2354
2355 static const struct super_operations f2fs_sops = {
2356 .alloc_inode = f2fs_alloc_inode,
2357 .free_inode = f2fs_free_inode,
2358 .drop_inode = f2fs_drop_inode,
2359 .write_inode = f2fs_write_inode,
2360 .dirty_inode = f2fs_dirty_inode,
2361 .show_options = f2fs_show_options,
2362 #ifdef CONFIG_QUOTA
2363 .quota_read = f2fs_quota_read,
2364 .quota_write = f2fs_quota_write,
2365 .get_dquots = f2fs_get_dquots,
2366 #endif
2367 .evict_inode = f2fs_evict_inode,
2368 .put_super = f2fs_put_super,
2369 .sync_fs = f2fs_sync_fs,
2370 .freeze_fs = f2fs_freeze,
2371 .unfreeze_fs = f2fs_unfreeze,
2372 .statfs = f2fs_statfs,
2373 .remount_fs = f2fs_remount,
2374 };
2375
2376 #ifdef CONFIG_FS_ENCRYPTION
2377 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2378 {
2379 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2380 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2381 ctx, len, NULL);
2382 }
2383
2384 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2385 void *fs_data)
2386 {
2387 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2388
2389 /*
2390 * Encrypting the root directory is not allowed because fsck
2391 * expects lost+found directory to exist and remain unencrypted
2392 * if LOST_FOUND feature is enabled.
2393 *
2394 */
2395 if (f2fs_sb_has_lost_found(sbi) &&
2396 inode->i_ino == F2FS_ROOT_INO(sbi))
2397 return -EPERM;
2398
2399 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2400 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2401 ctx, len, fs_data, XATTR_CREATE);
2402 }
2403
2404 static bool f2fs_dummy_context(struct inode *inode)
2405 {
2406 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2407 }
2408
2409 static bool f2fs_has_stable_inodes(struct super_block *sb)
2410 {
2411 return true;
2412 }
2413
2414 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2415 int *ino_bits_ret, int *lblk_bits_ret)
2416 {
2417 *ino_bits_ret = 8 * sizeof(nid_t);
2418 *lblk_bits_ret = 8 * sizeof(block_t);
2419 }
2420
2421 static const struct fscrypt_operations f2fs_cryptops = {
2422 .key_prefix = "f2fs:",
2423 .get_context = f2fs_get_context,
2424 .set_context = f2fs_set_context,
2425 .dummy_context = f2fs_dummy_context,
2426 .empty_dir = f2fs_empty_dir,
2427 .max_namelen = F2FS_NAME_LEN,
2428 .has_stable_inodes = f2fs_has_stable_inodes,
2429 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2430 };
2431 #endif
2432
2433 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2434 u64 ino, u32 generation)
2435 {
2436 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2437 struct inode *inode;
2438
2439 if (f2fs_check_nid_range(sbi, ino))
2440 return ERR_PTR(-ESTALE);
2441
2442 /*
2443 * f2fs_iget isn't quite right if the inode is currently unallocated!
2444 * However f2fs_iget currently does appropriate checks to handle stale
2445 * inodes so everything is OK.
2446 */
2447 inode = f2fs_iget(sb, ino);
2448 if (IS_ERR(inode))
2449 return ERR_CAST(inode);
2450 if (unlikely(generation && inode->i_generation != generation)) {
2451 /* we didn't find the right inode.. */
2452 iput(inode);
2453 return ERR_PTR(-ESTALE);
2454 }
2455 return inode;
2456 }
2457
2458 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2459 int fh_len, int fh_type)
2460 {
2461 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2462 f2fs_nfs_get_inode);
2463 }
2464
2465 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2466 int fh_len, int fh_type)
2467 {
2468 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2469 f2fs_nfs_get_inode);
2470 }
2471
2472 static const struct export_operations f2fs_export_ops = {
2473 .fh_to_dentry = f2fs_fh_to_dentry,
2474 .fh_to_parent = f2fs_fh_to_parent,
2475 .get_parent = f2fs_get_parent,
2476 };
2477
2478 static loff_t max_file_blocks(void)
2479 {
2480 loff_t result = 0;
2481 loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2482
2483 /*
2484 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2485 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2486 * space in inode.i_addr, it will be more safe to reassign
2487 * result as zero.
2488 */
2489
2490 /* two direct node blocks */
2491 result += (leaf_count * 2);
2492
2493 /* two indirect node blocks */
2494 leaf_count *= NIDS_PER_BLOCK;
2495 result += (leaf_count * 2);
2496
2497 /* one double indirect node block */
2498 leaf_count *= NIDS_PER_BLOCK;
2499 result += leaf_count;
2500
2501 return result;
2502 }
2503
2504 static int __f2fs_commit_super(struct buffer_head *bh,
2505 struct f2fs_super_block *super)
2506 {
2507 lock_buffer(bh);
2508 if (super)
2509 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2510 set_buffer_dirty(bh);
2511 unlock_buffer(bh);
2512
2513 /* it's rare case, we can do fua all the time */
2514 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2515 }
2516
2517 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2518 struct buffer_head *bh)
2519 {
2520 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2521 (bh->b_data + F2FS_SUPER_OFFSET);
2522 struct super_block *sb = sbi->sb;
2523 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2524 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2525 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2526 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2527 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2528 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2529 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2530 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2531 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2532 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2533 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2534 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2535 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2536 u64 main_end_blkaddr = main_blkaddr +
2537 (segment_count_main << log_blocks_per_seg);
2538 u64 seg_end_blkaddr = segment0_blkaddr +
2539 (segment_count << log_blocks_per_seg);
2540
2541 if (segment0_blkaddr != cp_blkaddr) {
2542 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2543 segment0_blkaddr, cp_blkaddr);
2544 return true;
2545 }
2546
2547 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2548 sit_blkaddr) {
2549 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2550 cp_blkaddr, sit_blkaddr,
2551 segment_count_ckpt << log_blocks_per_seg);
2552 return true;
2553 }
2554
2555 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2556 nat_blkaddr) {
2557 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2558 sit_blkaddr, nat_blkaddr,
2559 segment_count_sit << log_blocks_per_seg);
2560 return true;
2561 }
2562
2563 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2564 ssa_blkaddr) {
2565 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2566 nat_blkaddr, ssa_blkaddr,
2567 segment_count_nat << log_blocks_per_seg);
2568 return true;
2569 }
2570
2571 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2572 main_blkaddr) {
2573 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2574 ssa_blkaddr, main_blkaddr,
2575 segment_count_ssa << log_blocks_per_seg);
2576 return true;
2577 }
2578
2579 if (main_end_blkaddr > seg_end_blkaddr) {
2580 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2581 main_blkaddr,
2582 segment0_blkaddr +
2583 (segment_count << log_blocks_per_seg),
2584 segment_count_main << log_blocks_per_seg);
2585 return true;
2586 } else if (main_end_blkaddr < seg_end_blkaddr) {
2587 int err = 0;
2588 char *res;
2589
2590 /* fix in-memory information all the time */
2591 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2592 segment0_blkaddr) >> log_blocks_per_seg);
2593
2594 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2595 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2596 res = "internally";
2597 } else {
2598 err = __f2fs_commit_super(bh, NULL);
2599 res = err ? "failed" : "done";
2600 }
2601 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2602 res, main_blkaddr,
2603 segment0_blkaddr +
2604 (segment_count << log_blocks_per_seg),
2605 segment_count_main << log_blocks_per_seg);
2606 if (err)
2607 return true;
2608 }
2609 return false;
2610 }
2611
2612 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2613 struct buffer_head *bh)
2614 {
2615 block_t segment_count, segs_per_sec, secs_per_zone;
2616 block_t total_sections, blocks_per_seg;
2617 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2618 (bh->b_data + F2FS_SUPER_OFFSET);
2619 unsigned int blocksize;
2620 size_t crc_offset = 0;
2621 __u32 crc = 0;
2622
2623 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2624 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2625 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2626 return -EINVAL;
2627 }
2628
2629 /* Check checksum_offset and crc in superblock */
2630 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2631 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2632 if (crc_offset !=
2633 offsetof(struct f2fs_super_block, crc)) {
2634 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2635 crc_offset);
2636 return -EFSCORRUPTED;
2637 }
2638 crc = le32_to_cpu(raw_super->crc);
2639 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2640 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2641 return -EFSCORRUPTED;
2642 }
2643 }
2644
2645 /* Currently, support only 4KB page cache size */
2646 if (F2FS_BLKSIZE != PAGE_SIZE) {
2647 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2648 PAGE_SIZE);
2649 return -EFSCORRUPTED;
2650 }
2651
2652 /* Currently, support only 4KB block size */
2653 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2654 if (blocksize != F2FS_BLKSIZE) {
2655 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2656 blocksize);
2657 return -EFSCORRUPTED;
2658 }
2659
2660 /* check log blocks per segment */
2661 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2662 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2663 le32_to_cpu(raw_super->log_blocks_per_seg));
2664 return -EFSCORRUPTED;
2665 }
2666
2667 /* Currently, support 512/1024/2048/4096 bytes sector size */
2668 if (le32_to_cpu(raw_super->log_sectorsize) >
2669 F2FS_MAX_LOG_SECTOR_SIZE ||
2670 le32_to_cpu(raw_super->log_sectorsize) <
2671 F2FS_MIN_LOG_SECTOR_SIZE) {
2672 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2673 le32_to_cpu(raw_super->log_sectorsize));
2674 return -EFSCORRUPTED;
2675 }
2676 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2677 le32_to_cpu(raw_super->log_sectorsize) !=
2678 F2FS_MAX_LOG_SECTOR_SIZE) {
2679 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2680 le32_to_cpu(raw_super->log_sectors_per_block),
2681 le32_to_cpu(raw_super->log_sectorsize));
2682 return -EFSCORRUPTED;
2683 }
2684
2685 segment_count = le32_to_cpu(raw_super->segment_count);
2686 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2687 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2688 total_sections = le32_to_cpu(raw_super->section_count);
2689
2690 /* blocks_per_seg should be 512, given the above check */
2691 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2692
2693 if (segment_count > F2FS_MAX_SEGMENT ||
2694 segment_count < F2FS_MIN_SEGMENTS) {
2695 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2696 return -EFSCORRUPTED;
2697 }
2698
2699 if (total_sections > segment_count ||
2700 total_sections < F2FS_MIN_SEGMENTS ||
2701 segs_per_sec > segment_count || !segs_per_sec) {
2702 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2703 segment_count, total_sections, segs_per_sec);
2704 return -EFSCORRUPTED;
2705 }
2706
2707 if ((segment_count / segs_per_sec) < total_sections) {
2708 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2709 segment_count, segs_per_sec, total_sections);
2710 return -EFSCORRUPTED;
2711 }
2712
2713 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2714 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2715 segment_count, le64_to_cpu(raw_super->block_count));
2716 return -EFSCORRUPTED;
2717 }
2718
2719 if (RDEV(0).path[0]) {
2720 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2721 int i = 1;
2722
2723 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2724 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2725 i++;
2726 }
2727 if (segment_count != dev_seg_count) {
2728 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2729 segment_count, dev_seg_count);
2730 return -EFSCORRUPTED;
2731 }
2732 }
2733
2734 if (secs_per_zone > total_sections || !secs_per_zone) {
2735 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2736 secs_per_zone, total_sections);
2737 return -EFSCORRUPTED;
2738 }
2739 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2740 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2741 (le32_to_cpu(raw_super->extension_count) +
2742 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2743 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2744 le32_to_cpu(raw_super->extension_count),
2745 raw_super->hot_ext_count,
2746 F2FS_MAX_EXTENSION);
2747 return -EFSCORRUPTED;
2748 }
2749
2750 if (le32_to_cpu(raw_super->cp_payload) >
2751 (blocks_per_seg - F2FS_CP_PACKS)) {
2752 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2753 le32_to_cpu(raw_super->cp_payload),
2754 blocks_per_seg - F2FS_CP_PACKS);
2755 return -EFSCORRUPTED;
2756 }
2757
2758 /* check reserved ino info */
2759 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2760 le32_to_cpu(raw_super->meta_ino) != 2 ||
2761 le32_to_cpu(raw_super->root_ino) != 3) {
2762 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2763 le32_to_cpu(raw_super->node_ino),
2764 le32_to_cpu(raw_super->meta_ino),
2765 le32_to_cpu(raw_super->root_ino));
2766 return -EFSCORRUPTED;
2767 }
2768
2769 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2770 if (sanity_check_area_boundary(sbi, bh))
2771 return -EFSCORRUPTED;
2772
2773 return 0;
2774 }
2775
2776 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2777 {
2778 unsigned int total, fsmeta;
2779 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2780 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2781 unsigned int ovp_segments, reserved_segments;
2782 unsigned int main_segs, blocks_per_seg;
2783 unsigned int sit_segs, nat_segs;
2784 unsigned int sit_bitmap_size, nat_bitmap_size;
2785 unsigned int log_blocks_per_seg;
2786 unsigned int segment_count_main;
2787 unsigned int cp_pack_start_sum, cp_payload;
2788 block_t user_block_count, valid_user_blocks;
2789 block_t avail_node_count, valid_node_count;
2790 int i, j;
2791
2792 total = le32_to_cpu(raw_super->segment_count);
2793 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2794 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2795 fsmeta += sit_segs;
2796 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2797 fsmeta += nat_segs;
2798 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2799 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2800
2801 if (unlikely(fsmeta >= total))
2802 return 1;
2803
2804 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2805 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2806
2807 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2808 ovp_segments == 0 || reserved_segments == 0)) {
2809 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2810 return 1;
2811 }
2812
2813 user_block_count = le64_to_cpu(ckpt->user_block_count);
2814 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2815 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2816 if (!user_block_count || user_block_count >=
2817 segment_count_main << log_blocks_per_seg) {
2818 f2fs_err(sbi, "Wrong user_block_count: %u",
2819 user_block_count);
2820 return 1;
2821 }
2822
2823 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2824 if (valid_user_blocks > user_block_count) {
2825 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2826 valid_user_blocks, user_block_count);
2827 return 1;
2828 }
2829
2830 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2831 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2832 if (valid_node_count > avail_node_count) {
2833 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2834 valid_node_count, avail_node_count);
2835 return 1;
2836 }
2837
2838 main_segs = le32_to_cpu(raw_super->segment_count_main);
2839 blocks_per_seg = sbi->blocks_per_seg;
2840
2841 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2842 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2843 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2844 return 1;
2845 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2846 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2847 le32_to_cpu(ckpt->cur_node_segno[j])) {
2848 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2849 i, j,
2850 le32_to_cpu(ckpt->cur_node_segno[i]));
2851 return 1;
2852 }
2853 }
2854 }
2855 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2856 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2857 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2858 return 1;
2859 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2860 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2861 le32_to_cpu(ckpt->cur_data_segno[j])) {
2862 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2863 i, j,
2864 le32_to_cpu(ckpt->cur_data_segno[i]));
2865 return 1;
2866 }
2867 }
2868 }
2869 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2870 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2871 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2872 le32_to_cpu(ckpt->cur_data_segno[j])) {
2873 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2874 i, j,
2875 le32_to_cpu(ckpt->cur_node_segno[i]));
2876 return 1;
2877 }
2878 }
2879 }
2880
2881 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2882 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2883
2884 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2885 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2886 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2887 sit_bitmap_size, nat_bitmap_size);
2888 return 1;
2889 }
2890
2891 cp_pack_start_sum = __start_sum_addr(sbi);
2892 cp_payload = __cp_payload(sbi);
2893 if (cp_pack_start_sum < cp_payload + 1 ||
2894 cp_pack_start_sum > blocks_per_seg - 1 -
2895 NR_CURSEG_TYPE) {
2896 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2897 cp_pack_start_sum);
2898 return 1;
2899 }
2900
2901 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2902 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2903 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2904 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2905 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2906 le32_to_cpu(ckpt->checksum_offset));
2907 return 1;
2908 }
2909
2910 if (unlikely(f2fs_cp_error(sbi))) {
2911 f2fs_err(sbi, "A bug case: need to run fsck");
2912 return 1;
2913 }
2914 return 0;
2915 }
2916
2917 static void init_sb_info(struct f2fs_sb_info *sbi)
2918 {
2919 struct f2fs_super_block *raw_super = sbi->raw_super;
2920 int i;
2921
2922 sbi->log_sectors_per_block =
2923 le32_to_cpu(raw_super->log_sectors_per_block);
2924 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2925 sbi->blocksize = 1 << sbi->log_blocksize;
2926 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2927 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2928 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2929 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2930 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2931 sbi->total_node_count =
2932 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2933 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2934 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2935 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2936 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2937 sbi->cur_victim_sec = NULL_SECNO;
2938 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2939 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2940 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2941 sbi->migration_granularity = sbi->segs_per_sec;
2942
2943 sbi->dir_level = DEF_DIR_LEVEL;
2944 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2945 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2946 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2947 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2948 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2949 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2950 DEF_UMOUNT_DISCARD_TIMEOUT;
2951 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2952
2953 for (i = 0; i < NR_COUNT_TYPE; i++)
2954 atomic_set(&sbi->nr_pages[i], 0);
2955
2956 for (i = 0; i < META; i++)
2957 atomic_set(&sbi->wb_sync_req[i], 0);
2958
2959 INIT_LIST_HEAD(&sbi->s_list);
2960 mutex_init(&sbi->umount_mutex);
2961 init_rwsem(&sbi->io_order_lock);
2962 spin_lock_init(&sbi->cp_lock);
2963
2964 sbi->dirty_device = 0;
2965 spin_lock_init(&sbi->dev_lock);
2966
2967 init_rwsem(&sbi->sb_lock);
2968 init_rwsem(&sbi->pin_sem);
2969 }
2970
2971 static int init_percpu_info(struct f2fs_sb_info *sbi)
2972 {
2973 int err;
2974
2975 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2976 if (err)
2977 return err;
2978
2979 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2980 GFP_KERNEL);
2981 if (err)
2982 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2983
2984 return err;
2985 }
2986
2987 #ifdef CONFIG_BLK_DEV_ZONED
2988 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
2989 void *data)
2990 {
2991 struct f2fs_dev_info *dev = data;
2992
2993 if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
2994 set_bit(idx, dev->blkz_seq);
2995 return 0;
2996 }
2997
2998 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2999 {
3000 struct block_device *bdev = FDEV(devi).bdev;
3001 sector_t nr_sectors = bdev->bd_part->nr_sects;
3002 int ret;
3003
3004 if (!f2fs_sb_has_blkzoned(sbi))
3005 return 0;
3006
3007 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3008 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3009 return -EINVAL;
3010 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3011 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3012 __ilog2_u32(sbi->blocks_per_blkz))
3013 return -EINVAL;
3014 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3015 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3016 sbi->log_blocks_per_blkz;
3017 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3018 FDEV(devi).nr_blkz++;
3019
3020 FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
3021 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3022 * sizeof(unsigned long),
3023 GFP_KERNEL);
3024 if (!FDEV(devi).blkz_seq)
3025 return -ENOMEM;
3026
3027 /* Get block zones type */
3028 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3029 &FDEV(devi));
3030 if (ret < 0)
3031 return ret;
3032
3033 return 0;
3034 }
3035 #endif
3036
3037 /*
3038 * Read f2fs raw super block.
3039 * Because we have two copies of super block, so read both of them
3040 * to get the first valid one. If any one of them is broken, we pass
3041 * them recovery flag back to the caller.
3042 */
3043 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3044 struct f2fs_super_block **raw_super,
3045 int *valid_super_block, int *recovery)
3046 {
3047 struct super_block *sb = sbi->sb;
3048 int block;
3049 struct buffer_head *bh;
3050 struct f2fs_super_block *super;
3051 int err = 0;
3052
3053 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3054 if (!super)
3055 return -ENOMEM;
3056
3057 for (block = 0; block < 2; block++) {
3058 bh = sb_bread(sb, block);
3059 if (!bh) {
3060 f2fs_err(sbi, "Unable to read %dth superblock",
3061 block + 1);
3062 err = -EIO;
3063 *recovery = 1;
3064 continue;
3065 }
3066
3067 /* sanity checking of raw super */
3068 err = sanity_check_raw_super(sbi, bh);
3069 if (err) {
3070 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3071 block + 1);
3072 brelse(bh);
3073 *recovery = 1;
3074 continue;
3075 }
3076
3077 if (!*raw_super) {
3078 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3079 sizeof(*super));
3080 *valid_super_block = block;
3081 *raw_super = super;
3082 }
3083 brelse(bh);
3084 }
3085
3086 /* No valid superblock */
3087 if (!*raw_super)
3088 kvfree(super);
3089 else
3090 err = 0;
3091
3092 return err;
3093 }
3094
3095 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3096 {
3097 struct buffer_head *bh;
3098 __u32 crc = 0;
3099 int err;
3100
3101 if ((recover && f2fs_readonly(sbi->sb)) ||
3102 bdev_read_only(sbi->sb->s_bdev)) {
3103 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3104 return -EROFS;
3105 }
3106
3107 /* we should update superblock crc here */
3108 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3109 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3110 offsetof(struct f2fs_super_block, crc));
3111 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3112 }
3113
3114 /* write back-up superblock first */
3115 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3116 if (!bh)
3117 return -EIO;
3118 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3119 brelse(bh);
3120
3121 /* if we are in recovery path, skip writing valid superblock */
3122 if (recover || err)
3123 return err;
3124
3125 /* write current valid superblock */
3126 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3127 if (!bh)
3128 return -EIO;
3129 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3130 brelse(bh);
3131 return err;
3132 }
3133
3134 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3135 {
3136 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3137 unsigned int max_devices = MAX_DEVICES;
3138 int i;
3139
3140 /* Initialize single device information */
3141 if (!RDEV(0).path[0]) {
3142 if (!bdev_is_zoned(sbi->sb->s_bdev))
3143 return 0;
3144 max_devices = 1;
3145 }
3146
3147 /*
3148 * Initialize multiple devices information, or single
3149 * zoned block device information.
3150 */
3151 sbi->devs = f2fs_kzalloc(sbi,
3152 array_size(max_devices,
3153 sizeof(struct f2fs_dev_info)),
3154 GFP_KERNEL);
3155 if (!sbi->devs)
3156 return -ENOMEM;
3157
3158 for (i = 0; i < max_devices; i++) {
3159
3160 if (i > 0 && !RDEV(i).path[0])
3161 break;
3162
3163 if (max_devices == 1) {
3164 /* Single zoned block device mount */
3165 FDEV(0).bdev =
3166 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3167 sbi->sb->s_mode, sbi->sb->s_type);
3168 } else {
3169 /* Multi-device mount */
3170 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3171 FDEV(i).total_segments =
3172 le32_to_cpu(RDEV(i).total_segments);
3173 if (i == 0) {
3174 FDEV(i).start_blk = 0;
3175 FDEV(i).end_blk = FDEV(i).start_blk +
3176 (FDEV(i).total_segments <<
3177 sbi->log_blocks_per_seg) - 1 +
3178 le32_to_cpu(raw_super->segment0_blkaddr);
3179 } else {
3180 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3181 FDEV(i).end_blk = FDEV(i).start_blk +
3182 (FDEV(i).total_segments <<
3183 sbi->log_blocks_per_seg) - 1;
3184 }
3185 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3186 sbi->sb->s_mode, sbi->sb->s_type);
3187 }
3188 if (IS_ERR(FDEV(i).bdev))
3189 return PTR_ERR(FDEV(i).bdev);
3190
3191 /* to release errored devices */
3192 sbi->s_ndevs = i + 1;
3193
3194 #ifdef CONFIG_BLK_DEV_ZONED
3195 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3196 !f2fs_sb_has_blkzoned(sbi)) {
3197 f2fs_err(sbi, "Zoned block device feature not enabled\n");
3198 return -EINVAL;
3199 }
3200 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3201 if (init_blkz_info(sbi, i)) {
3202 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3203 return -EINVAL;
3204 }
3205 if (max_devices == 1)
3206 break;
3207 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3208 i, FDEV(i).path,
3209 FDEV(i).total_segments,
3210 FDEV(i).start_blk, FDEV(i).end_blk,
3211 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3212 "Host-aware" : "Host-managed");
3213 continue;
3214 }
3215 #endif
3216 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3217 i, FDEV(i).path,
3218 FDEV(i).total_segments,
3219 FDEV(i).start_blk, FDEV(i).end_blk);
3220 }
3221 f2fs_info(sbi,
3222 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3223 return 0;
3224 }
3225
3226 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3227 {
3228 #ifdef CONFIG_UNICODE
3229 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3230 const struct f2fs_sb_encodings *encoding_info;
3231 struct unicode_map *encoding;
3232 __u16 encoding_flags;
3233
3234 if (f2fs_sb_has_encrypt(sbi)) {
3235 f2fs_err(sbi,
3236 "Can't mount with encoding and encryption");
3237 return -EINVAL;
3238 }
3239
3240 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3241 &encoding_flags)) {
3242 f2fs_err(sbi,
3243 "Encoding requested by superblock is unknown");
3244 return -EINVAL;
3245 }
3246
3247 encoding = utf8_load(encoding_info->version);
3248 if (IS_ERR(encoding)) {
3249 f2fs_err(sbi,
3250 "can't mount with superblock charset: %s-%s "
3251 "not supported by the kernel. flags: 0x%x.",
3252 encoding_info->name, encoding_info->version,
3253 encoding_flags);
3254 return PTR_ERR(encoding);
3255 }
3256 f2fs_info(sbi, "Using encoding defined by superblock: "
3257 "%s-%s with flags 0x%hx", encoding_info->name,
3258 encoding_info->version?:"\b", encoding_flags);
3259
3260 sbi->s_encoding = encoding;
3261 sbi->s_encoding_flags = encoding_flags;
3262 sbi->sb->s_d_op = &f2fs_dentry_ops;
3263 }
3264 #else
3265 if (f2fs_sb_has_casefold(sbi)) {
3266 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3267 return -EINVAL;
3268 }
3269 #endif
3270 return 0;
3271 }
3272
3273 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3274 {
3275 struct f2fs_sm_info *sm_i = SM_I(sbi);
3276
3277 /* adjust parameters according to the volume size */
3278 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3279 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3280 sm_i->dcc_info->discard_granularity = 1;
3281 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3282 }
3283
3284 sbi->readdir_ra = 1;
3285 }
3286
3287 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3288 {
3289 struct f2fs_sb_info *sbi;
3290 struct f2fs_super_block *raw_super;
3291 struct inode *root;
3292 int err;
3293 bool skip_recovery = false, need_fsck = false;
3294 char *options = NULL;
3295 int recovery, i, valid_super_block;
3296 struct curseg_info *seg_i;
3297 int retry_cnt = 1;
3298
3299 try_onemore:
3300 err = -EINVAL;
3301 raw_super = NULL;
3302 valid_super_block = -1;
3303 recovery = 0;
3304
3305 /* allocate memory for f2fs-specific super block info */
3306 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3307 if (!sbi)
3308 return -ENOMEM;
3309
3310 sbi->sb = sb;
3311
3312 /* Load the checksum driver */
3313 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3314 if (IS_ERR(sbi->s_chksum_driver)) {
3315 f2fs_err(sbi, "Cannot load crc32 driver.");
3316 err = PTR_ERR(sbi->s_chksum_driver);
3317 sbi->s_chksum_driver = NULL;
3318 goto free_sbi;
3319 }
3320
3321 /* set a block size */
3322 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3323 f2fs_err(sbi, "unable to set blocksize");
3324 goto free_sbi;
3325 }
3326
3327 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3328 &recovery);
3329 if (err)
3330 goto free_sbi;
3331
3332 sb->s_fs_info = sbi;
3333 sbi->raw_super = raw_super;
3334
3335 /* precompute checksum seed for metadata */
3336 if (f2fs_sb_has_inode_chksum(sbi))
3337 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3338 sizeof(raw_super->uuid));
3339
3340 /*
3341 * The BLKZONED feature indicates that the drive was formatted with
3342 * zone alignment optimization. This is optional for host-aware
3343 * devices, but mandatory for host-managed zoned block devices.
3344 */
3345 #ifndef CONFIG_BLK_DEV_ZONED
3346 if (f2fs_sb_has_blkzoned(sbi)) {
3347 f2fs_err(sbi, "Zoned block device support is not enabled");
3348 err = -EOPNOTSUPP;
3349 goto free_sb_buf;
3350 }
3351 #endif
3352 default_options(sbi);
3353 /* parse mount options */
3354 options = kstrdup((const char *)data, GFP_KERNEL);
3355 if (data && !options) {
3356 err = -ENOMEM;
3357 goto free_sb_buf;
3358 }
3359
3360 err = parse_options(sb, options);
3361 if (err)
3362 goto free_options;
3363
3364 sbi->max_file_blocks = max_file_blocks();
3365 sb->s_maxbytes = sbi->max_file_blocks <<
3366 le32_to_cpu(raw_super->log_blocksize);
3367 sb->s_max_links = F2FS_LINK_MAX;
3368
3369 err = f2fs_setup_casefold(sbi);
3370 if (err)
3371 goto free_options;
3372
3373 #ifdef CONFIG_QUOTA
3374 sb->dq_op = &f2fs_quota_operations;
3375 sb->s_qcop = &f2fs_quotactl_ops;
3376 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3377
3378 if (f2fs_sb_has_quota_ino(sbi)) {
3379 for (i = 0; i < MAXQUOTAS; i++) {
3380 if (f2fs_qf_ino(sbi->sb, i))
3381 sbi->nquota_files++;
3382 }
3383 }
3384 #endif
3385
3386 sb->s_op = &f2fs_sops;
3387 #ifdef CONFIG_FS_ENCRYPTION
3388 sb->s_cop = &f2fs_cryptops;
3389 #endif
3390 #ifdef CONFIG_FS_VERITY
3391 sb->s_vop = &f2fs_verityops;
3392 #endif
3393 sb->s_xattr = f2fs_xattr_handlers;
3394 sb->s_export_op = &f2fs_export_ops;
3395 sb->s_magic = F2FS_SUPER_MAGIC;
3396 sb->s_time_gran = 1;
3397 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3398 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3399 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3400 sb->s_iflags |= SB_I_CGROUPWB;
3401
3402 /* init f2fs-specific super block info */
3403 sbi->valid_super_block = valid_super_block;
3404 init_rwsem(&sbi->gc_lock);
3405 mutex_init(&sbi->writepages);
3406 mutex_init(&sbi->cp_mutex);
3407 mutex_init(&sbi->resize_mutex);
3408 init_rwsem(&sbi->node_write);
3409 init_rwsem(&sbi->node_change);
3410
3411 /* disallow all the data/node/meta page writes */
3412 set_sbi_flag(sbi, SBI_POR_DOING);
3413 spin_lock_init(&sbi->stat_lock);
3414
3415 /* init iostat info */
3416 spin_lock_init(&sbi->iostat_lock);
3417 sbi->iostat_enable = false;
3418
3419 for (i = 0; i < NR_PAGE_TYPE; i++) {
3420 int n = (i == META) ? 1: NR_TEMP_TYPE;
3421 int j;
3422
3423 sbi->write_io[i] =
3424 f2fs_kmalloc(sbi,
3425 array_size(n,
3426 sizeof(struct f2fs_bio_info)),
3427 GFP_KERNEL);
3428 if (!sbi->write_io[i]) {
3429 err = -ENOMEM;
3430 goto free_bio_info;
3431 }
3432
3433 for (j = HOT; j < n; j++) {
3434 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3435 sbi->write_io[i][j].sbi = sbi;
3436 sbi->write_io[i][j].bio = NULL;
3437 spin_lock_init(&sbi->write_io[i][j].io_lock);
3438 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3439 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3440 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3441 }
3442 }
3443
3444 init_rwsem(&sbi->cp_rwsem);
3445 init_rwsem(&sbi->quota_sem);
3446 init_waitqueue_head(&sbi->cp_wait);
3447 init_sb_info(sbi);
3448
3449 err = init_percpu_info(sbi);
3450 if (err)
3451 goto free_bio_info;
3452
3453 if (F2FS_IO_ALIGNED(sbi)) {
3454 sbi->write_io_dummy =
3455 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3456 if (!sbi->write_io_dummy) {
3457 err = -ENOMEM;
3458 goto free_percpu;
3459 }
3460 }
3461
3462 /* get an inode for meta space */
3463 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3464 if (IS_ERR(sbi->meta_inode)) {
3465 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3466 err = PTR_ERR(sbi->meta_inode);
3467 goto free_io_dummy;
3468 }
3469
3470 err = f2fs_get_valid_checkpoint(sbi);
3471 if (err) {
3472 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3473 goto free_meta_inode;
3474 }
3475
3476 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3477 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3478 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3479 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3480 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3481 }
3482
3483 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3484 set_sbi_flag(sbi, SBI_NEED_FSCK);
3485
3486 /* Initialize device list */
3487 err = f2fs_scan_devices(sbi);
3488 if (err) {
3489 f2fs_err(sbi, "Failed to find devices");
3490 goto free_devices;
3491 }
3492
3493 err = f2fs_init_post_read_wq(sbi);
3494 if (err) {
3495 f2fs_err(sbi, "Failed to initialize post read workqueue");
3496 goto free_devices;
3497 }
3498
3499 sbi->total_valid_node_count =
3500 le32_to_cpu(sbi->ckpt->valid_node_count);
3501 percpu_counter_set(&sbi->total_valid_inode_count,
3502 le32_to_cpu(sbi->ckpt->valid_inode_count));
3503 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3504 sbi->total_valid_block_count =
3505 le64_to_cpu(sbi->ckpt->valid_block_count);
3506 sbi->last_valid_block_count = sbi->total_valid_block_count;
3507 sbi->reserved_blocks = 0;
3508 sbi->current_reserved_blocks = 0;
3509 limit_reserve_root(sbi);
3510
3511 for (i = 0; i < NR_INODE_TYPE; i++) {
3512 INIT_LIST_HEAD(&sbi->inode_list[i]);
3513 spin_lock_init(&sbi->inode_lock[i]);
3514 }
3515 mutex_init(&sbi->flush_lock);
3516
3517 f2fs_init_extent_cache_info(sbi);
3518
3519 f2fs_init_ino_entry_info(sbi);
3520
3521 f2fs_init_fsync_node_info(sbi);
3522
3523 /* setup f2fs internal modules */
3524 err = f2fs_build_segment_manager(sbi);
3525 if (err) {
3526 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3527 err);
3528 goto free_sm;
3529 }
3530 err = f2fs_build_node_manager(sbi);
3531 if (err) {
3532 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3533 err);
3534 goto free_nm;
3535 }
3536
3537 /* For write statistics */
3538 if (sb->s_bdev->bd_part)
3539 sbi->sectors_written_start =
3540 (u64)part_stat_read(sb->s_bdev->bd_part,
3541 sectors[STAT_WRITE]);
3542
3543 /* Read accumulated write IO statistics if exists */
3544 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3545 if (__exist_node_summaries(sbi))
3546 sbi->kbytes_written =
3547 le64_to_cpu(seg_i->journal->info.kbytes_written);
3548
3549 f2fs_build_gc_manager(sbi);
3550
3551 err = f2fs_build_stats(sbi);
3552 if (err)
3553 goto free_nm;
3554
3555 /* get an inode for node space */
3556 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3557 if (IS_ERR(sbi->node_inode)) {
3558 f2fs_err(sbi, "Failed to read node inode");
3559 err = PTR_ERR(sbi->node_inode);
3560 goto free_stats;
3561 }
3562
3563 /* read root inode and dentry */
3564 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3565 if (IS_ERR(root)) {
3566 f2fs_err(sbi, "Failed to read root inode");
3567 err = PTR_ERR(root);
3568 goto free_node_inode;
3569 }
3570 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3571 !root->i_size || !root->i_nlink) {
3572 iput(root);
3573 err = -EINVAL;
3574 goto free_node_inode;
3575 }
3576
3577 sb->s_root = d_make_root(root); /* allocate root dentry */
3578 if (!sb->s_root) {
3579 err = -ENOMEM;
3580 goto free_node_inode;
3581 }
3582
3583 err = f2fs_register_sysfs(sbi);
3584 if (err)
3585 goto free_root_inode;
3586
3587 #ifdef CONFIG_QUOTA
3588 /* Enable quota usage during mount */
3589 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3590 err = f2fs_enable_quotas(sb);
3591 if (err)
3592 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3593 }
3594 #endif
3595 /* if there are nt orphan nodes free them */
3596 err = f2fs_recover_orphan_inodes(sbi);
3597 if (err)
3598 goto free_meta;
3599
3600 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3601 goto reset_checkpoint;
3602
3603 /* recover fsynced data */
3604 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3605 !test_opt(sbi, NORECOVERY)) {
3606 /*
3607 * mount should be failed, when device has readonly mode, and
3608 * previous checkpoint was not done by clean system shutdown.
3609 */
3610 if (f2fs_hw_is_readonly(sbi)) {
3611 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3612 err = -EROFS;
3613 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3614 goto free_meta;
3615 }
3616 f2fs_info(sbi, "write access unavailable, skipping recovery");
3617 goto reset_checkpoint;
3618 }
3619
3620 if (need_fsck)
3621 set_sbi_flag(sbi, SBI_NEED_FSCK);
3622
3623 if (skip_recovery)
3624 goto reset_checkpoint;
3625
3626 err = f2fs_recover_fsync_data(sbi, false);
3627 if (err < 0) {
3628 if (err != -ENOMEM)
3629 skip_recovery = true;
3630 need_fsck = true;
3631 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3632 err);
3633 goto free_meta;
3634 }
3635 } else {
3636 err = f2fs_recover_fsync_data(sbi, true);
3637
3638 if (!f2fs_readonly(sb) && err > 0) {
3639 err = -EINVAL;
3640 f2fs_err(sbi, "Need to recover fsync data");
3641 goto free_meta;
3642 }
3643 }
3644
3645 /*
3646 * If the f2fs is not readonly and fsync data recovery succeeds,
3647 * check zoned block devices' write pointer consistency.
3648 */
3649 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3650 err = f2fs_check_write_pointer(sbi);
3651 if (err)
3652 goto free_meta;
3653 }
3654
3655 reset_checkpoint:
3656 /* f2fs_recover_fsync_data() cleared this already */
3657 clear_sbi_flag(sbi, SBI_POR_DOING);
3658
3659 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3660 err = f2fs_disable_checkpoint(sbi);
3661 if (err)
3662 goto sync_free_meta;
3663 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3664 f2fs_enable_checkpoint(sbi);
3665 }
3666
3667 /*
3668 * If filesystem is not mounted as read-only then
3669 * do start the gc_thread.
3670 */
3671 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3672 /* After POR, we can run background GC thread.*/
3673 err = f2fs_start_gc_thread(sbi);
3674 if (err)
3675 goto sync_free_meta;
3676 }
3677 kvfree(options);
3678
3679 /* recover broken superblock */
3680 if (recovery) {
3681 err = f2fs_commit_super(sbi, true);
3682 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3683 sbi->valid_super_block ? 1 : 2, err);
3684 }
3685
3686 f2fs_join_shrinker(sbi);
3687
3688 f2fs_tuning_parameters(sbi);
3689
3690 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3691 cur_cp_version(F2FS_CKPT(sbi)));
3692 f2fs_update_time(sbi, CP_TIME);
3693 f2fs_update_time(sbi, REQ_TIME);
3694 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3695 return 0;
3696
3697 sync_free_meta:
3698 /* safe to flush all the data */
3699 sync_filesystem(sbi->sb);
3700 retry_cnt = 0;
3701
3702 free_meta:
3703 #ifdef CONFIG_QUOTA
3704 f2fs_truncate_quota_inode_pages(sb);
3705 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3706 f2fs_quota_off_umount(sbi->sb);
3707 #endif
3708 /*
3709 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3710 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3711 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3712 * falls into an infinite loop in f2fs_sync_meta_pages().
3713 */
3714 truncate_inode_pages_final(META_MAPPING(sbi));
3715 /* evict some inodes being cached by GC */
3716 evict_inodes(sb);
3717 f2fs_unregister_sysfs(sbi);
3718 free_root_inode:
3719 dput(sb->s_root);
3720 sb->s_root = NULL;
3721 free_node_inode:
3722 f2fs_release_ino_entry(sbi, true);
3723 truncate_inode_pages_final(NODE_MAPPING(sbi));
3724 iput(sbi->node_inode);
3725 sbi->node_inode = NULL;
3726 free_stats:
3727 f2fs_destroy_stats(sbi);
3728 free_nm:
3729 f2fs_destroy_node_manager(sbi);
3730 free_sm:
3731 f2fs_destroy_segment_manager(sbi);
3732 f2fs_destroy_post_read_wq(sbi);
3733 free_devices:
3734 destroy_device_list(sbi);
3735 kvfree(sbi->ckpt);
3736 free_meta_inode:
3737 make_bad_inode(sbi->meta_inode);
3738 iput(sbi->meta_inode);
3739 sbi->meta_inode = NULL;
3740 free_io_dummy:
3741 mempool_destroy(sbi->write_io_dummy);
3742 free_percpu:
3743 destroy_percpu_info(sbi);
3744 free_bio_info:
3745 for (i = 0; i < NR_PAGE_TYPE; i++)
3746 kvfree(sbi->write_io[i]);
3747
3748 #ifdef CONFIG_UNICODE
3749 utf8_unload(sbi->s_encoding);
3750 #endif
3751 free_options:
3752 #ifdef CONFIG_QUOTA
3753 for (i = 0; i < MAXQUOTAS; i++)
3754 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3755 #endif
3756 kvfree(options);
3757 free_sb_buf:
3758 kvfree(raw_super);
3759 free_sbi:
3760 if (sbi->s_chksum_driver)
3761 crypto_free_shash(sbi->s_chksum_driver);
3762 kvfree(sbi);
3763
3764 /* give only one another chance */
3765 if (retry_cnt > 0 && skip_recovery) {
3766 retry_cnt--;
3767 shrink_dcache_sb(sb);
3768 goto try_onemore;
3769 }
3770 return err;
3771 }
3772
3773 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3774 const char *dev_name, void *data)
3775 {
3776 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3777 }
3778
3779 static void kill_f2fs_super(struct super_block *sb)
3780 {
3781 if (sb->s_root) {
3782 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3783
3784 set_sbi_flag(sbi, SBI_IS_CLOSE);
3785 f2fs_stop_gc_thread(sbi);
3786 f2fs_stop_discard_thread(sbi);
3787
3788 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3789 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3790 struct cp_control cpc = {
3791 .reason = CP_UMOUNT,
3792 };
3793 f2fs_write_checkpoint(sbi, &cpc);
3794 }
3795
3796 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3797 sb->s_flags &= ~SB_RDONLY;
3798 }
3799 kill_block_super(sb);
3800 }
3801
3802 static struct file_system_type f2fs_fs_type = {
3803 .owner = THIS_MODULE,
3804 .name = "f2fs",
3805 .mount = f2fs_mount,
3806 .kill_sb = kill_f2fs_super,
3807 .fs_flags = FS_REQUIRES_DEV,
3808 };
3809 MODULE_ALIAS_FS("f2fs");
3810
3811 static int __init init_inodecache(void)
3812 {
3813 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3814 sizeof(struct f2fs_inode_info), 0,
3815 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3816 if (!f2fs_inode_cachep)
3817 return -ENOMEM;
3818 return 0;
3819 }
3820
3821 static void destroy_inodecache(void)
3822 {
3823 /*
3824 * Make sure all delayed rcu free inodes are flushed before we
3825 * destroy cache.
3826 */
3827 rcu_barrier();
3828 kmem_cache_destroy(f2fs_inode_cachep);
3829 }
3830
3831 static int __init init_f2fs_fs(void)
3832 {
3833 int err;
3834
3835 if (PAGE_SIZE != F2FS_BLKSIZE) {
3836 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3837 PAGE_SIZE, F2FS_BLKSIZE);
3838 return -EINVAL;
3839 }
3840
3841 f2fs_build_trace_ios();
3842
3843 err = init_inodecache();
3844 if (err)
3845 goto fail;
3846 err = f2fs_create_node_manager_caches();
3847 if (err)
3848 goto free_inodecache;
3849 err = f2fs_create_segment_manager_caches();
3850 if (err)
3851 goto free_node_manager_caches;
3852 err = f2fs_create_checkpoint_caches();
3853 if (err)
3854 goto free_segment_manager_caches;
3855 err = f2fs_create_extent_cache();
3856 if (err)
3857 goto free_checkpoint_caches;
3858 err = f2fs_init_sysfs();
3859 if (err)
3860 goto free_extent_cache;
3861 err = register_shrinker(&f2fs_shrinker_info);
3862 if (err)
3863 goto free_sysfs;
3864 err = register_filesystem(&f2fs_fs_type);
3865 if (err)
3866 goto free_shrinker;
3867 f2fs_create_root_stats();
3868 err = f2fs_init_post_read_processing();
3869 if (err)
3870 goto free_root_stats;
3871 err = f2fs_init_bio_entry_cache();
3872 if (err)
3873 goto free_post_read;
3874 err = f2fs_init_bioset();
3875 if (err)
3876 goto free_bio_enrty_cache;
3877 return 0;
3878 free_bio_enrty_cache:
3879 f2fs_destroy_bio_entry_cache();
3880 free_post_read:
3881 f2fs_destroy_post_read_processing();
3882 free_root_stats:
3883 f2fs_destroy_root_stats();
3884 unregister_filesystem(&f2fs_fs_type);
3885 free_shrinker:
3886 unregister_shrinker(&f2fs_shrinker_info);
3887 free_sysfs:
3888 f2fs_exit_sysfs();
3889 free_extent_cache:
3890 f2fs_destroy_extent_cache();
3891 free_checkpoint_caches:
3892 f2fs_destroy_checkpoint_caches();
3893 free_segment_manager_caches:
3894 f2fs_destroy_segment_manager_caches();
3895 free_node_manager_caches:
3896 f2fs_destroy_node_manager_caches();
3897 free_inodecache:
3898 destroy_inodecache();
3899 fail:
3900 return err;
3901 }
3902
3903 static void __exit exit_f2fs_fs(void)
3904 {
3905 f2fs_destroy_bioset();
3906 f2fs_destroy_bio_entry_cache();
3907 f2fs_destroy_post_read_processing();
3908 f2fs_destroy_root_stats();
3909 unregister_filesystem(&f2fs_fs_type);
3910 unregister_shrinker(&f2fs_shrinker_info);
3911 f2fs_exit_sysfs();
3912 f2fs_destroy_extent_cache();
3913 f2fs_destroy_checkpoint_caches();
3914 f2fs_destroy_segment_manager_caches();
3915 f2fs_destroy_node_manager_caches();
3916 destroy_inodecache();
3917 f2fs_destroy_trace_ios();
3918 }
3919
3920 module_init(init_f2fs_fs)
3921 module_exit(exit_f2fs_fs)
3922
3923 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3924 MODULE_DESCRIPTION("Flash Friendly File System");
3925 MODULE_LICENSE("GPL");
3926
Linux with added FPC-III support