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
27 #include "f2fs.h"
28 #include "node.h"
29 #include "segment.h"
30 #include "xattr.h"
31 #include "gc.h"
32 #include "trace.h"
33
34 #define CREATE_TRACE_POINTS
35 #include <trace/events/f2fs.h>
36
37 static struct kmem_cache *f2fs_inode_cachep;
38
39 #ifdef CONFIG_F2FS_FAULT_INJECTION
40
41 const char *f2fs_fault_name[FAULT_MAX] = {
42 [FAULT_KMALLOC] = "kmalloc",
43 [FAULT_KVMALLOC] = "kvmalloc",
44 [FAULT_PAGE_ALLOC] = "page alloc",
45 [FAULT_PAGE_GET] = "page get",
46 [FAULT_ALLOC_BIO] = "alloc bio",
47 [FAULT_ALLOC_NID] = "alloc nid",
48 [FAULT_ORPHAN] = "orphan",
49 [FAULT_BLOCK] = "no more block",
50 [FAULT_DIR_DEPTH] = "too big dir depth",
51 [FAULT_EVICT_INODE] = "evict_inode fail",
52 [FAULT_TRUNCATE] = "truncate fail",
53 [FAULT_READ_IO] = "read IO error",
54 [FAULT_CHECKPOINT] = "checkpoint error",
55 [FAULT_DISCARD] = "discard error",
56 [FAULT_WRITE_IO] = "write IO error",
57 };
58
f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned int rate,unsigned int type)59 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
60 unsigned int type)
61 {
62 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
63
64 if (rate) {
65 atomic_set(&ffi->inject_ops, 0);
66 ffi->inject_rate = rate;
67 }
68
69 if (type)
70 ffi->inject_type = type;
71
72 if (!rate && !type)
73 memset(ffi, 0, sizeof(struct f2fs_fault_info));
74 }
75 #endif
76
77 /* f2fs-wide shrinker description */
78 static struct shrinker f2fs_shrinker_info = {
79 .scan_objects = f2fs_shrink_scan,
80 .count_objects = f2fs_shrink_count,
81 .seeks = DEFAULT_SEEKS,
82 };
83
84 enum {
85 Opt_gc_background,
86 Opt_disable_roll_forward,
87 Opt_norecovery,
88 Opt_discard,
89 Opt_nodiscard,
90 Opt_noheap,
91 Opt_heap,
92 Opt_user_xattr,
93 Opt_nouser_xattr,
94 Opt_acl,
95 Opt_noacl,
96 Opt_active_logs,
97 Opt_disable_ext_identify,
98 Opt_inline_xattr,
99 Opt_noinline_xattr,
100 Opt_inline_xattr_size,
101 Opt_inline_data,
102 Opt_inline_dentry,
103 Opt_noinline_dentry,
104 Opt_flush_merge,
105 Opt_noflush_merge,
106 Opt_nobarrier,
107 Opt_fastboot,
108 Opt_extent_cache,
109 Opt_noextent_cache,
110 Opt_noinline_data,
111 Opt_data_flush,
112 Opt_reserve_root,
113 Opt_resgid,
114 Opt_resuid,
115 Opt_mode,
116 Opt_io_size_bits,
117 Opt_fault_injection,
118 Opt_fault_type,
119 Opt_lazytime,
120 Opt_nolazytime,
121 Opt_quota,
122 Opt_noquota,
123 Opt_usrquota,
124 Opt_grpquota,
125 Opt_prjquota,
126 Opt_usrjquota,
127 Opt_grpjquota,
128 Opt_prjjquota,
129 Opt_offusrjquota,
130 Opt_offgrpjquota,
131 Opt_offprjjquota,
132 Opt_jqfmt_vfsold,
133 Opt_jqfmt_vfsv0,
134 Opt_jqfmt_vfsv1,
135 Opt_whint,
136 Opt_alloc,
137 Opt_fsync,
138 Opt_test_dummy_encryption,
139 Opt_checkpoint,
140 Opt_err,
141 };
142
143 static match_table_t f2fs_tokens = {
144 {Opt_gc_background, "background_gc=%s"},
145 {Opt_disable_roll_forward, "disable_roll_forward"},
146 {Opt_norecovery, "norecovery"},
147 {Opt_discard, "discard"},
148 {Opt_nodiscard, "nodiscard"},
149 {Opt_noheap, "no_heap"},
150 {Opt_heap, "heap"},
151 {Opt_user_xattr, "user_xattr"},
152 {Opt_nouser_xattr, "nouser_xattr"},
153 {Opt_acl, "acl"},
154 {Opt_noacl, "noacl"},
155 {Opt_active_logs, "active_logs=%u"},
156 {Opt_disable_ext_identify, "disable_ext_identify"},
157 {Opt_inline_xattr, "inline_xattr"},
158 {Opt_noinline_xattr, "noinline_xattr"},
159 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
160 {Opt_inline_data, "inline_data"},
161 {Opt_inline_dentry, "inline_dentry"},
162 {Opt_noinline_dentry, "noinline_dentry"},
163 {Opt_flush_merge, "flush_merge"},
164 {Opt_noflush_merge, "noflush_merge"},
165 {Opt_nobarrier, "nobarrier"},
166 {Opt_fastboot, "fastboot"},
167 {Opt_extent_cache, "extent_cache"},
168 {Opt_noextent_cache, "noextent_cache"},
169 {Opt_noinline_data, "noinline_data"},
170 {Opt_data_flush, "data_flush"},
171 {Opt_reserve_root, "reserve_root=%u"},
172 {Opt_resgid, "resgid=%u"},
173 {Opt_resuid, "resuid=%u"},
174 {Opt_mode, "mode=%s"},
175 {Opt_io_size_bits, "io_bits=%u"},
176 {Opt_fault_injection, "fault_injection=%u"},
177 {Opt_fault_type, "fault_type=%u"},
178 {Opt_lazytime, "lazytime"},
179 {Opt_nolazytime, "nolazytime"},
180 {Opt_quota, "quota"},
181 {Opt_noquota, "noquota"},
182 {Opt_usrquota, "usrquota"},
183 {Opt_grpquota, "grpquota"},
184 {Opt_prjquota, "prjquota"},
185 {Opt_usrjquota, "usrjquota=%s"},
186 {Opt_grpjquota, "grpjquota=%s"},
187 {Opt_prjjquota, "prjjquota=%s"},
188 {Opt_offusrjquota, "usrjquota="},
189 {Opt_offgrpjquota, "grpjquota="},
190 {Opt_offprjjquota, "prjjquota="},
191 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
192 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
193 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
194 {Opt_whint, "whint_mode=%s"},
195 {Opt_alloc, "alloc_mode=%s"},
196 {Opt_fsync, "fsync_mode=%s"},
197 {Opt_test_dummy_encryption, "test_dummy_encryption"},
198 {Opt_checkpoint, "checkpoint=%s"},
199 {Opt_err, NULL},
200 };
201
f2fs_msg(struct super_block * sb,const char * level,const char * fmt,...)202 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
203 {
204 struct va_format vaf;
205 va_list args;
206
207 va_start(args, fmt);
208 vaf.fmt = fmt;
209 vaf.va = &args;
210 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
211 va_end(args);
212 }
213
limit_reserve_root(struct f2fs_sb_info * sbi)214 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
215 {
216 block_t limit = (sbi->user_block_count << 1) / 1000;
217
218 /* limit is 0.2% */
219 if (test_opt(sbi, RESERVE_ROOT) &&
220 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
221 F2FS_OPTION(sbi).root_reserved_blocks = limit;
222 f2fs_msg(sbi->sb, KERN_INFO,
223 "Reduce reserved blocks for root = %u",
224 F2FS_OPTION(sbi).root_reserved_blocks);
225 }
226 if (!test_opt(sbi, RESERVE_ROOT) &&
227 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
228 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
229 !gid_eq(F2FS_OPTION(sbi).s_resgid,
230 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
231 f2fs_msg(sbi->sb, KERN_INFO,
232 "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
233 from_kuid_munged(&init_user_ns,
234 F2FS_OPTION(sbi).s_resuid),
235 from_kgid_munged(&init_user_ns,
236 F2FS_OPTION(sbi).s_resgid));
237 }
238
init_once(void * foo)239 static void init_once(void *foo)
240 {
241 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
242
243 inode_init_once(&fi->vfs_inode);
244 }
245
246 #ifdef CONFIG_QUOTA
247 static const char * const quotatypes[] = INITQFNAMES;
248 #define QTYPE2NAME(t) (quotatypes[t])
f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)249 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
250 substring_t *args)
251 {
252 struct f2fs_sb_info *sbi = F2FS_SB(sb);
253 char *qname;
254 int ret = -EINVAL;
255
256 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
257 f2fs_msg(sb, KERN_ERR,
258 "Cannot change journaled "
259 "quota options when quota turned on");
260 return -EINVAL;
261 }
262 if (f2fs_sb_has_quota_ino(sbi)) {
263 f2fs_msg(sb, KERN_INFO,
264 "QUOTA feature is enabled, so ignore qf_name");
265 return 0;
266 }
267
268 qname = match_strdup(args);
269 if (!qname) {
270 f2fs_msg(sb, KERN_ERR,
271 "Not enough memory for storing quotafile name");
272 return -ENOMEM;
273 }
274 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
275 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
276 ret = 0;
277 else
278 f2fs_msg(sb, KERN_ERR,
279 "%s quota file already specified",
280 QTYPE2NAME(qtype));
281 goto errout;
282 }
283 if (strchr(qname, '/')) {
284 f2fs_msg(sb, KERN_ERR,
285 "quotafile must be on filesystem root");
286 goto errout;
287 }
288 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
289 set_opt(sbi, QUOTA);
290 return 0;
291 errout:
292 kvfree(qname);
293 return ret;
294 }
295
f2fs_clear_qf_name(struct super_block * sb,int qtype)296 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
297 {
298 struct f2fs_sb_info *sbi = F2FS_SB(sb);
299
300 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
301 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
302 " when quota turned on");
303 return -EINVAL;
304 }
305 kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
306 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
307 return 0;
308 }
309
f2fs_check_quota_options(struct f2fs_sb_info * sbi)310 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
311 {
312 /*
313 * We do the test below only for project quotas. 'usrquota' and
314 * 'grpquota' mount options are allowed even without quota feature
315 * to support legacy quotas in quota files.
316 */
317 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
318 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
319 "Cannot enable project quota enforcement.");
320 return -1;
321 }
322 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
323 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
324 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
325 if (test_opt(sbi, USRQUOTA) &&
326 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
327 clear_opt(sbi, USRQUOTA);
328
329 if (test_opt(sbi, GRPQUOTA) &&
330 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
331 clear_opt(sbi, GRPQUOTA);
332
333 if (test_opt(sbi, PRJQUOTA) &&
334 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
335 clear_opt(sbi, PRJQUOTA);
336
337 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
338 test_opt(sbi, PRJQUOTA)) {
339 f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
340 "format mixing");
341 return -1;
342 }
343
344 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
345 f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
346 "not specified");
347 return -1;
348 }
349 }
350
351 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
352 f2fs_msg(sbi->sb, KERN_INFO,
353 "QUOTA feature is enabled, so ignore jquota_fmt");
354 F2FS_OPTION(sbi).s_jquota_fmt = 0;
355 }
356 return 0;
357 }
358 #endif
359
parse_options(struct super_block * sb,char * options)360 static int parse_options(struct super_block *sb, char *options)
361 {
362 struct f2fs_sb_info *sbi = F2FS_SB(sb);
363 substring_t args[MAX_OPT_ARGS];
364 char *p, *name;
365 int arg = 0;
366 kuid_t uid;
367 kgid_t gid;
368 #ifdef CONFIG_QUOTA
369 int ret;
370 #endif
371
372 if (!options)
373 return 0;
374
375 while ((p = strsep(&options, ",")) != NULL) {
376 int token;
377 if (!*p)
378 continue;
379 /*
380 * Initialize args struct so we know whether arg was
381 * found; some options take optional arguments.
382 */
383 args[0].to = args[0].from = NULL;
384 token = match_token(p, f2fs_tokens, args);
385
386 switch (token) {
387 case Opt_gc_background:
388 name = match_strdup(&args[0]);
389
390 if (!name)
391 return -ENOMEM;
392 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
393 set_opt(sbi, BG_GC);
394 clear_opt(sbi, FORCE_FG_GC);
395 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
396 clear_opt(sbi, BG_GC);
397 clear_opt(sbi, FORCE_FG_GC);
398 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
399 set_opt(sbi, BG_GC);
400 set_opt(sbi, FORCE_FG_GC);
401 } else {
402 kvfree(name);
403 return -EINVAL;
404 }
405 kvfree(name);
406 break;
407 case Opt_disable_roll_forward:
408 set_opt(sbi, DISABLE_ROLL_FORWARD);
409 break;
410 case Opt_norecovery:
411 /* this option mounts f2fs with ro */
412 set_opt(sbi, DISABLE_ROLL_FORWARD);
413 if (!f2fs_readonly(sb))
414 return -EINVAL;
415 break;
416 case Opt_discard:
417 set_opt(sbi, DISCARD);
418 break;
419 case Opt_nodiscard:
420 if (f2fs_sb_has_blkzoned(sbi)) {
421 f2fs_msg(sb, KERN_WARNING,
422 "discard is required for zoned block devices");
423 return -EINVAL;
424 }
425 clear_opt(sbi, DISCARD);
426 break;
427 case Opt_noheap:
428 set_opt(sbi, NOHEAP);
429 break;
430 case Opt_heap:
431 clear_opt(sbi, NOHEAP);
432 break;
433 #ifdef CONFIG_F2FS_FS_XATTR
434 case Opt_user_xattr:
435 set_opt(sbi, XATTR_USER);
436 break;
437 case Opt_nouser_xattr:
438 clear_opt(sbi, XATTR_USER);
439 break;
440 case Opt_inline_xattr:
441 set_opt(sbi, INLINE_XATTR);
442 break;
443 case Opt_noinline_xattr:
444 clear_opt(sbi, INLINE_XATTR);
445 break;
446 case Opt_inline_xattr_size:
447 if (args->from && match_int(args, &arg))
448 return -EINVAL;
449 set_opt(sbi, INLINE_XATTR_SIZE);
450 F2FS_OPTION(sbi).inline_xattr_size = arg;
451 break;
452 #else
453 case Opt_user_xattr:
454 f2fs_msg(sb, KERN_INFO,
455 "user_xattr options not supported");
456 break;
457 case Opt_nouser_xattr:
458 f2fs_msg(sb, KERN_INFO,
459 "nouser_xattr options not supported");
460 break;
461 case Opt_inline_xattr:
462 f2fs_msg(sb, KERN_INFO,
463 "inline_xattr options not supported");
464 break;
465 case Opt_noinline_xattr:
466 f2fs_msg(sb, KERN_INFO,
467 "noinline_xattr options not supported");
468 break;
469 #endif
470 #ifdef CONFIG_F2FS_FS_POSIX_ACL
471 case Opt_acl:
472 set_opt(sbi, POSIX_ACL);
473 break;
474 case Opt_noacl:
475 clear_opt(sbi, POSIX_ACL);
476 break;
477 #else
478 case Opt_acl:
479 f2fs_msg(sb, KERN_INFO, "acl options not supported");
480 break;
481 case Opt_noacl:
482 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
483 break;
484 #endif
485 case Opt_active_logs:
486 if (args->from && match_int(args, &arg))
487 return -EINVAL;
488 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
489 return -EINVAL;
490 F2FS_OPTION(sbi).active_logs = arg;
491 break;
492 case Opt_disable_ext_identify:
493 set_opt(sbi, DISABLE_EXT_IDENTIFY);
494 break;
495 case Opt_inline_data:
496 set_opt(sbi, INLINE_DATA);
497 break;
498 case Opt_inline_dentry:
499 set_opt(sbi, INLINE_DENTRY);
500 break;
501 case Opt_noinline_dentry:
502 clear_opt(sbi, INLINE_DENTRY);
503 break;
504 case Opt_flush_merge:
505 set_opt(sbi, FLUSH_MERGE);
506 break;
507 case Opt_noflush_merge:
508 clear_opt(sbi, FLUSH_MERGE);
509 break;
510 case Opt_nobarrier:
511 set_opt(sbi, NOBARRIER);
512 break;
513 case Opt_fastboot:
514 set_opt(sbi, FASTBOOT);
515 break;
516 case Opt_extent_cache:
517 set_opt(sbi, EXTENT_CACHE);
518 break;
519 case Opt_noextent_cache:
520 clear_opt(sbi, EXTENT_CACHE);
521 break;
522 case Opt_noinline_data:
523 clear_opt(sbi, INLINE_DATA);
524 break;
525 case Opt_data_flush:
526 set_opt(sbi, DATA_FLUSH);
527 break;
528 case Opt_reserve_root:
529 if (args->from && match_int(args, &arg))
530 return -EINVAL;
531 if (test_opt(sbi, RESERVE_ROOT)) {
532 f2fs_msg(sb, KERN_INFO,
533 "Preserve previous reserve_root=%u",
534 F2FS_OPTION(sbi).root_reserved_blocks);
535 } else {
536 F2FS_OPTION(sbi).root_reserved_blocks = arg;
537 set_opt(sbi, RESERVE_ROOT);
538 }
539 break;
540 case Opt_resuid:
541 if (args->from && match_int(args, &arg))
542 return -EINVAL;
543 uid = make_kuid(current_user_ns(), arg);
544 if (!uid_valid(uid)) {
545 f2fs_msg(sb, KERN_ERR,
546 "Invalid uid value %d", arg);
547 return -EINVAL;
548 }
549 F2FS_OPTION(sbi).s_resuid = uid;
550 break;
551 case Opt_resgid:
552 if (args->from && match_int(args, &arg))
553 return -EINVAL;
554 gid = make_kgid(current_user_ns(), arg);
555 if (!gid_valid(gid)) {
556 f2fs_msg(sb, KERN_ERR,
557 "Invalid gid value %d", arg);
558 return -EINVAL;
559 }
560 F2FS_OPTION(sbi).s_resgid = gid;
561 break;
562 case Opt_mode:
563 name = match_strdup(&args[0]);
564
565 if (!name)
566 return -ENOMEM;
567 if (strlen(name) == 8 &&
568 !strncmp(name, "adaptive", 8)) {
569 if (f2fs_sb_has_blkzoned(sbi)) {
570 f2fs_msg(sb, KERN_WARNING,
571 "adaptive mode is not allowed with "
572 "zoned block device feature");
573 kvfree(name);
574 return -EINVAL;
575 }
576 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
577 } else if (strlen(name) == 3 &&
578 !strncmp(name, "lfs", 3)) {
579 set_opt_mode(sbi, F2FS_MOUNT_LFS);
580 } else {
581 kvfree(name);
582 return -EINVAL;
583 }
584 kvfree(name);
585 break;
586 case Opt_io_size_bits:
587 if (args->from && match_int(args, &arg))
588 return -EINVAL;
589 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
590 f2fs_msg(sb, KERN_WARNING,
591 "Not support %d, larger than %d",
592 1 << arg, BIO_MAX_PAGES);
593 return -EINVAL;
594 }
595 F2FS_OPTION(sbi).write_io_size_bits = arg;
596 break;
597 #ifdef CONFIG_F2FS_FAULT_INJECTION
598 case Opt_fault_injection:
599 if (args->from && match_int(args, &arg))
600 return -EINVAL;
601 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
602 set_opt(sbi, FAULT_INJECTION);
603 break;
604
605 case Opt_fault_type:
606 if (args->from && match_int(args, &arg))
607 return -EINVAL;
608 f2fs_build_fault_attr(sbi, 0, arg);
609 set_opt(sbi, FAULT_INJECTION);
610 break;
611 #else
612 case Opt_fault_injection:
613 f2fs_msg(sb, KERN_INFO,
614 "fault_injection options not supported");
615 break;
616
617 case Opt_fault_type:
618 f2fs_msg(sb, KERN_INFO,
619 "fault_type options not supported");
620 break;
621 #endif
622 case Opt_lazytime:
623 sb->s_flags |= MS_LAZYTIME;
624 break;
625 case Opt_nolazytime:
626 sb->s_flags &= ~MS_LAZYTIME;
627 break;
628 #ifdef CONFIG_QUOTA
629 case Opt_quota:
630 case Opt_usrquota:
631 set_opt(sbi, USRQUOTA);
632 break;
633 case Opt_grpquota:
634 set_opt(sbi, GRPQUOTA);
635 break;
636 case Opt_prjquota:
637 set_opt(sbi, PRJQUOTA);
638 break;
639 case Opt_usrjquota:
640 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
641 if (ret)
642 return ret;
643 break;
644 case Opt_grpjquota:
645 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
646 if (ret)
647 return ret;
648 break;
649 case Opt_prjjquota:
650 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
651 if (ret)
652 return ret;
653 break;
654 case Opt_offusrjquota:
655 ret = f2fs_clear_qf_name(sb, USRQUOTA);
656 if (ret)
657 return ret;
658 break;
659 case Opt_offgrpjquota:
660 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
661 if (ret)
662 return ret;
663 break;
664 case Opt_offprjjquota:
665 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
666 if (ret)
667 return ret;
668 break;
669 case Opt_jqfmt_vfsold:
670 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
671 break;
672 case Opt_jqfmt_vfsv0:
673 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
674 break;
675 case Opt_jqfmt_vfsv1:
676 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
677 break;
678 case Opt_noquota:
679 clear_opt(sbi, QUOTA);
680 clear_opt(sbi, USRQUOTA);
681 clear_opt(sbi, GRPQUOTA);
682 clear_opt(sbi, PRJQUOTA);
683 break;
684 #else
685 case Opt_quota:
686 case Opt_usrquota:
687 case Opt_grpquota:
688 case Opt_prjquota:
689 case Opt_usrjquota:
690 case Opt_grpjquota:
691 case Opt_prjjquota:
692 case Opt_offusrjquota:
693 case Opt_offgrpjquota:
694 case Opt_offprjjquota:
695 case Opt_jqfmt_vfsold:
696 case Opt_jqfmt_vfsv0:
697 case Opt_jqfmt_vfsv1:
698 case Opt_noquota:
699 f2fs_msg(sb, KERN_INFO,
700 "quota operations not supported");
701 break;
702 #endif
703 case Opt_whint:
704 name = match_strdup(&args[0]);
705 if (!name)
706 return -ENOMEM;
707 if (strlen(name) == 10 &&
708 !strncmp(name, "user-based", 10)) {
709 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
710 } else if (strlen(name) == 3 &&
711 !strncmp(name, "off", 3)) {
712 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
713 } else if (strlen(name) == 8 &&
714 !strncmp(name, "fs-based", 8)) {
715 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
716 } else {
717 kvfree(name);
718 return -EINVAL;
719 }
720 kvfree(name);
721 break;
722 case Opt_alloc:
723 name = match_strdup(&args[0]);
724 if (!name)
725 return -ENOMEM;
726
727 if (strlen(name) == 7 &&
728 !strncmp(name, "default", 7)) {
729 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
730 } else if (strlen(name) == 5 &&
731 !strncmp(name, "reuse", 5)) {
732 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
733 } else {
734 kvfree(name);
735 return -EINVAL;
736 }
737 kvfree(name);
738 break;
739 case Opt_fsync:
740 name = match_strdup(&args[0]);
741 if (!name)
742 return -ENOMEM;
743 if (strlen(name) == 5 &&
744 !strncmp(name, "posix", 5)) {
745 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
746 } else if (strlen(name) == 6 &&
747 !strncmp(name, "strict", 6)) {
748 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
749 } else if (strlen(name) == 9 &&
750 !strncmp(name, "nobarrier", 9)) {
751 F2FS_OPTION(sbi).fsync_mode =
752 FSYNC_MODE_NOBARRIER;
753 } else {
754 kvfree(name);
755 return -EINVAL;
756 }
757 kvfree(name);
758 break;
759 case Opt_test_dummy_encryption:
760 #ifdef CONFIG_F2FS_FS_ENCRYPTION
761 if (!f2fs_sb_has_encrypt(sbi)) {
762 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
763 return -EINVAL;
764 }
765
766 F2FS_OPTION(sbi).test_dummy_encryption = true;
767 f2fs_msg(sb, KERN_INFO,
768 "Test dummy encryption mode enabled");
769 #else
770 f2fs_msg(sb, KERN_INFO,
771 "Test dummy encryption mount option ignored");
772 #endif
773 break;
774 case Opt_checkpoint:
775 name = match_strdup(&args[0]);
776 if (!name)
777 return -ENOMEM;
778
779 if (strlen(name) == 6 &&
780 !strncmp(name, "enable", 6)) {
781 clear_opt(sbi, DISABLE_CHECKPOINT);
782 } else if (strlen(name) == 7 &&
783 !strncmp(name, "disable", 7)) {
784 set_opt(sbi, DISABLE_CHECKPOINT);
785 } else {
786 kvfree(name);
787 return -EINVAL;
788 }
789 kvfree(name);
790 break;
791 default:
792 f2fs_msg(sb, KERN_ERR,
793 "Unrecognized mount option \"%s\" or missing value",
794 p);
795 return -EINVAL;
796 }
797 }
798 #ifdef CONFIG_QUOTA
799 if (f2fs_check_quota_options(sbi))
800 return -EINVAL;
801 #else
802 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
803 f2fs_msg(sbi->sb, KERN_INFO,
804 "Filesystem with quota feature cannot be mounted RDWR "
805 "without CONFIG_QUOTA");
806 return -EINVAL;
807 }
808 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
809 f2fs_msg(sb, KERN_ERR,
810 "Filesystem with project quota feature cannot be "
811 "mounted RDWR without CONFIG_QUOTA");
812 return -EINVAL;
813 }
814 #endif
815
816 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
817 f2fs_msg(sb, KERN_ERR,
818 "Should set mode=lfs with %uKB-sized IO",
819 F2FS_IO_SIZE_KB(sbi));
820 return -EINVAL;
821 }
822
823 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
824 int min_size, max_size;
825
826 if (!f2fs_sb_has_extra_attr(sbi) ||
827 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
828 f2fs_msg(sb, KERN_ERR,
829 "extra_attr or flexible_inline_xattr "
830 "feature is off");
831 return -EINVAL;
832 }
833 if (!test_opt(sbi, INLINE_XATTR)) {
834 f2fs_msg(sb, KERN_ERR,
835 "inline_xattr_size option should be "
836 "set with inline_xattr option");
837 return -EINVAL;
838 }
839
840 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
841 max_size = MAX_INLINE_XATTR_SIZE;
842
843 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
844 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
845 f2fs_msg(sb, KERN_ERR,
846 "inline xattr size is out of range: %d ~ %d",
847 min_size, max_size);
848 return -EINVAL;
849 }
850 }
851
852 if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
853 f2fs_msg(sb, KERN_ERR,
854 "LFS not compatible with checkpoint=disable\n");
855 return -EINVAL;
856 }
857
858 /* Not pass down write hints if the number of active logs is lesser
859 * than NR_CURSEG_TYPE.
860 */
861 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
862 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
863 return 0;
864 }
865
f2fs_alloc_inode(struct super_block * sb)866 static struct inode *f2fs_alloc_inode(struct super_block *sb)
867 {
868 struct f2fs_inode_info *fi;
869
870 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
871 if (!fi)
872 return NULL;
873
874 init_once((void *) fi);
875
876 /* Initialize f2fs-specific inode info */
877 atomic_set(&fi->dirty_pages, 0);
878 init_rwsem(&fi->i_sem);
879 INIT_LIST_HEAD(&fi->dirty_list);
880 INIT_LIST_HEAD(&fi->gdirty_list);
881 INIT_LIST_HEAD(&fi->inmem_ilist);
882 INIT_LIST_HEAD(&fi->inmem_pages);
883 mutex_init(&fi->inmem_lock);
884 init_rwsem(&fi->i_gc_rwsem[READ]);
885 init_rwsem(&fi->i_gc_rwsem[WRITE]);
886 init_rwsem(&fi->i_mmap_sem);
887 init_rwsem(&fi->i_xattr_sem);
888
889 /* Will be used by directory only */
890 fi->i_dir_level = F2FS_SB(sb)->dir_level;
891
892 return &fi->vfs_inode;
893 }
894
f2fs_drop_inode(struct inode * inode)895 static int f2fs_drop_inode(struct inode *inode)
896 {
897 int ret;
898 /*
899 * This is to avoid a deadlock condition like below.
900 * writeback_single_inode(inode)
901 * - f2fs_write_data_page
902 * - f2fs_gc -> iput -> evict
903 * - inode_wait_for_writeback(inode)
904 */
905 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
906 if (!inode->i_nlink && !is_bad_inode(inode)) {
907 /* to avoid evict_inode call simultaneously */
908 atomic_inc(&inode->i_count);
909 spin_unlock(&inode->i_lock);
910
911 /* some remained atomic pages should discarded */
912 if (f2fs_is_atomic_file(inode))
913 f2fs_drop_inmem_pages(inode);
914
915 /* should remain fi->extent_tree for writepage */
916 f2fs_destroy_extent_node(inode);
917
918 sb_start_intwrite(inode->i_sb);
919 f2fs_i_size_write(inode, 0);
920
921 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
922 inode, NULL, 0, DATA);
923 truncate_inode_pages_final(inode->i_mapping);
924
925 if (F2FS_HAS_BLOCKS(inode))
926 f2fs_truncate(inode);
927
928 sb_end_intwrite(inode->i_sb);
929
930 spin_lock(&inode->i_lock);
931 atomic_dec(&inode->i_count);
932 }
933 trace_f2fs_drop_inode(inode, 0);
934 return 0;
935 }
936 ret = generic_drop_inode(inode);
937 trace_f2fs_drop_inode(inode, ret);
938 return ret;
939 }
940
f2fs_inode_dirtied(struct inode * inode,bool sync)941 int f2fs_inode_dirtied(struct inode *inode, bool sync)
942 {
943 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
944 int ret = 0;
945
946 spin_lock(&sbi->inode_lock[DIRTY_META]);
947 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
948 ret = 1;
949 } else {
950 set_inode_flag(inode, FI_DIRTY_INODE);
951 stat_inc_dirty_inode(sbi, DIRTY_META);
952 }
953 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
954 list_add_tail(&F2FS_I(inode)->gdirty_list,
955 &sbi->inode_list[DIRTY_META]);
956 inc_page_count(sbi, F2FS_DIRTY_IMETA);
957 }
958 spin_unlock(&sbi->inode_lock[DIRTY_META]);
959 return ret;
960 }
961
f2fs_inode_synced(struct inode * inode)962 void f2fs_inode_synced(struct inode *inode)
963 {
964 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
965
966 spin_lock(&sbi->inode_lock[DIRTY_META]);
967 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
968 spin_unlock(&sbi->inode_lock[DIRTY_META]);
969 return;
970 }
971 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
972 list_del_init(&F2FS_I(inode)->gdirty_list);
973 dec_page_count(sbi, F2FS_DIRTY_IMETA);
974 }
975 clear_inode_flag(inode, FI_DIRTY_INODE);
976 clear_inode_flag(inode, FI_AUTO_RECOVER);
977 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
978 spin_unlock(&sbi->inode_lock[DIRTY_META]);
979 }
980
981 /*
982 * f2fs_dirty_inode() is called from __mark_inode_dirty()
983 *
984 * We should call set_dirty_inode to write the dirty inode through write_inode.
985 */
f2fs_dirty_inode(struct inode * inode,int flags)986 static void f2fs_dirty_inode(struct inode *inode, int flags)
987 {
988 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
989
990 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
991 inode->i_ino == F2FS_META_INO(sbi))
992 return;
993
994 if (flags == I_DIRTY_TIME)
995 return;
996
997 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
998 clear_inode_flag(inode, FI_AUTO_RECOVER);
999
1000 f2fs_inode_dirtied(inode, false);
1001 }
1002
f2fs_i_callback(struct rcu_head * head)1003 static void f2fs_i_callback(struct rcu_head *head)
1004 {
1005 struct inode *inode = container_of(head, struct inode, i_rcu);
1006 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1007 }
1008
f2fs_destroy_inode(struct inode * inode)1009 static void f2fs_destroy_inode(struct inode *inode)
1010 {
1011 call_rcu(&inode->i_rcu, f2fs_i_callback);
1012 }
1013
destroy_percpu_info(struct f2fs_sb_info * sbi)1014 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1015 {
1016 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1017 percpu_counter_destroy(&sbi->total_valid_inode_count);
1018 }
1019
destroy_device_list(struct f2fs_sb_info * sbi)1020 static void destroy_device_list(struct f2fs_sb_info *sbi)
1021 {
1022 int i;
1023
1024 for (i = 0; i < sbi->s_ndevs; i++) {
1025 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1026 #ifdef CONFIG_BLK_DEV_ZONED
1027 kvfree(FDEV(i).blkz_type);
1028 #endif
1029 }
1030 kvfree(sbi->devs);
1031 }
1032
f2fs_put_super(struct super_block * sb)1033 static void f2fs_put_super(struct super_block *sb)
1034 {
1035 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1036 int i;
1037 bool dropped;
1038
1039 f2fs_quota_off_umount(sb);
1040
1041 /* prevent remaining shrinker jobs */
1042 mutex_lock(&sbi->umount_mutex);
1043
1044 /*
1045 * We don't need to do checkpoint when superblock is clean.
1046 * But, the previous checkpoint was not done by umount, it needs to do
1047 * clean checkpoint again.
1048 */
1049 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1050 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1051 struct cp_control cpc = {
1052 .reason = CP_UMOUNT,
1053 };
1054 f2fs_write_checkpoint(sbi, &cpc);
1055 }
1056
1057 /* be sure to wait for any on-going discard commands */
1058 dropped = f2fs_issue_discard_timeout(sbi);
1059
1060 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1061 !sbi->discard_blks && !dropped) {
1062 struct cp_control cpc = {
1063 .reason = CP_UMOUNT | CP_TRIMMED,
1064 };
1065 f2fs_write_checkpoint(sbi, &cpc);
1066 }
1067
1068 /*
1069 * normally superblock is clean, so we need to release this.
1070 * In addition, EIO will skip do checkpoint, we need this as well.
1071 */
1072 f2fs_release_ino_entry(sbi, true);
1073
1074 f2fs_leave_shrinker(sbi);
1075 mutex_unlock(&sbi->umount_mutex);
1076
1077 /* our cp_error case, we can wait for any writeback page */
1078 f2fs_flush_merged_writes(sbi);
1079
1080 f2fs_wait_on_all_pages_writeback(sbi);
1081
1082 f2fs_bug_on(sbi, sbi->fsync_node_num);
1083
1084 iput(sbi->node_inode);
1085 sbi->node_inode = NULL;
1086
1087 iput(sbi->meta_inode);
1088 sbi->meta_inode = NULL;
1089
1090 /*
1091 * iput() can update stat information, if f2fs_write_checkpoint()
1092 * above failed with error.
1093 */
1094 f2fs_destroy_stats(sbi);
1095
1096 /* destroy f2fs internal modules */
1097 f2fs_destroy_node_manager(sbi);
1098 f2fs_destroy_segment_manager(sbi);
1099
1100 kvfree(sbi->ckpt);
1101
1102 f2fs_unregister_sysfs(sbi);
1103
1104 sb->s_fs_info = NULL;
1105 if (sbi->s_chksum_driver)
1106 crypto_free_shash(sbi->s_chksum_driver);
1107 kvfree(sbi->raw_super);
1108
1109 destroy_device_list(sbi);
1110 mempool_destroy(sbi->write_io_dummy);
1111 #ifdef CONFIG_QUOTA
1112 for (i = 0; i < MAXQUOTAS; i++)
1113 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1114 #endif
1115 destroy_percpu_info(sbi);
1116 for (i = 0; i < NR_PAGE_TYPE; i++)
1117 kvfree(sbi->write_io[i]);
1118 kvfree(sbi);
1119 }
1120
f2fs_sync_fs(struct super_block * sb,int sync)1121 int f2fs_sync_fs(struct super_block *sb, int sync)
1122 {
1123 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1124 int err = 0;
1125
1126 if (unlikely(f2fs_cp_error(sbi)))
1127 return 0;
1128 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1129 return 0;
1130
1131 trace_f2fs_sync_fs(sb, sync);
1132
1133 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1134 return -EAGAIN;
1135
1136 if (sync) {
1137 struct cp_control cpc;
1138
1139 cpc.reason = __get_cp_reason(sbi);
1140
1141 mutex_lock(&sbi->gc_mutex);
1142 err = f2fs_write_checkpoint(sbi, &cpc);
1143 mutex_unlock(&sbi->gc_mutex);
1144 }
1145 f2fs_trace_ios(NULL, 1);
1146
1147 return err;
1148 }
1149
f2fs_freeze(struct super_block * sb)1150 static int f2fs_freeze(struct super_block *sb)
1151 {
1152 if (f2fs_readonly(sb))
1153 return 0;
1154
1155 /* IO error happened before */
1156 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1157 return -EIO;
1158
1159 /* must be clean, since sync_filesystem() was already called */
1160 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1161 return -EINVAL;
1162 return 0;
1163 }
1164
f2fs_unfreeze(struct super_block * sb)1165 static int f2fs_unfreeze(struct super_block *sb)
1166 {
1167 return 0;
1168 }
1169
1170 #ifdef CONFIG_QUOTA
f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1171 static int f2fs_statfs_project(struct super_block *sb,
1172 kprojid_t projid, struct kstatfs *buf)
1173 {
1174 struct kqid qid;
1175 struct dquot *dquot;
1176 u64 limit;
1177 u64 curblock;
1178
1179 qid = make_kqid_projid(projid);
1180 dquot = dqget(sb, qid);
1181 if (IS_ERR(dquot))
1182 return PTR_ERR(dquot);
1183 spin_lock(&dquot->dq_dqb_lock);
1184
1185 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1186 dquot->dq_dqb.dqb_bhardlimit);
1187 if (limit)
1188 limit >>= sb->s_blocksize_bits;
1189
1190 if (limit && buf->f_blocks > limit) {
1191 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1192 buf->f_blocks = limit;
1193 buf->f_bfree = buf->f_bavail =
1194 (buf->f_blocks > curblock) ?
1195 (buf->f_blocks - curblock) : 0;
1196 }
1197
1198 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1199 dquot->dq_dqb.dqb_ihardlimit);
1200
1201 if (limit && buf->f_files > limit) {
1202 buf->f_files = limit;
1203 buf->f_ffree =
1204 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1205 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1206 }
1207
1208 spin_unlock(&dquot->dq_dqb_lock);
1209 dqput(dquot);
1210 return 0;
1211 }
1212 #endif
1213
f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1214 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1215 {
1216 struct super_block *sb = dentry->d_sb;
1217 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1218 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1219 block_t total_count, user_block_count, start_count;
1220 u64 avail_node_count;
1221
1222 total_count = le64_to_cpu(sbi->raw_super->block_count);
1223 user_block_count = sbi->user_block_count;
1224 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1225 buf->f_type = F2FS_SUPER_MAGIC;
1226 buf->f_bsize = sbi->blocksize;
1227
1228 buf->f_blocks = total_count - start_count;
1229 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1230 sbi->current_reserved_blocks;
1231 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1232 buf->f_bfree = 0;
1233 else
1234 buf->f_bfree -= sbi->unusable_block_count;
1235
1236 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1237 buf->f_bavail = buf->f_bfree -
1238 F2FS_OPTION(sbi).root_reserved_blocks;
1239 else
1240 buf->f_bavail = 0;
1241
1242 avail_node_count = sbi->total_node_count - sbi->nquota_files -
1243 F2FS_RESERVED_NODE_NUM;
1244
1245 if (avail_node_count > user_block_count) {
1246 buf->f_files = user_block_count;
1247 buf->f_ffree = buf->f_bavail;
1248 } else {
1249 buf->f_files = avail_node_count;
1250 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1251 buf->f_bavail);
1252 }
1253
1254 buf->f_namelen = F2FS_NAME_LEN;
1255 buf->f_fsid.val[0] = (u32)id;
1256 buf->f_fsid.val[1] = (u32)(id >> 32);
1257
1258 #ifdef CONFIG_QUOTA
1259 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1260 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1261 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1262 }
1263 #endif
1264 return 0;
1265 }
1266
f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1267 static inline void f2fs_show_quota_options(struct seq_file *seq,
1268 struct super_block *sb)
1269 {
1270 #ifdef CONFIG_QUOTA
1271 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1272
1273 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1274 char *fmtname = "";
1275
1276 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1277 case QFMT_VFS_OLD:
1278 fmtname = "vfsold";
1279 break;
1280 case QFMT_VFS_V0:
1281 fmtname = "vfsv0";
1282 break;
1283 case QFMT_VFS_V1:
1284 fmtname = "vfsv1";
1285 break;
1286 }
1287 seq_printf(seq, ",jqfmt=%s", fmtname);
1288 }
1289
1290 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1291 seq_show_option(seq, "usrjquota",
1292 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1293
1294 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1295 seq_show_option(seq, "grpjquota",
1296 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1297
1298 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1299 seq_show_option(seq, "prjjquota",
1300 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1301 #endif
1302 }
1303
f2fs_show_options(struct seq_file * seq,struct dentry * root)1304 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1305 {
1306 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1307
1308 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1309 if (test_opt(sbi, FORCE_FG_GC))
1310 seq_printf(seq, ",background_gc=%s", "sync");
1311 else
1312 seq_printf(seq, ",background_gc=%s", "on");
1313 } else {
1314 seq_printf(seq, ",background_gc=%s", "off");
1315 }
1316 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1317 seq_puts(seq, ",disable_roll_forward");
1318 if (test_opt(sbi, DISCARD))
1319 seq_puts(seq, ",discard");
1320 if (test_opt(sbi, NOHEAP))
1321 seq_puts(seq, ",no_heap");
1322 else
1323 seq_puts(seq, ",heap");
1324 #ifdef CONFIG_F2FS_FS_XATTR
1325 if (test_opt(sbi, XATTR_USER))
1326 seq_puts(seq, ",user_xattr");
1327 else
1328 seq_puts(seq, ",nouser_xattr");
1329 if (test_opt(sbi, INLINE_XATTR))
1330 seq_puts(seq, ",inline_xattr");
1331 else
1332 seq_puts(seq, ",noinline_xattr");
1333 if (test_opt(sbi, INLINE_XATTR_SIZE))
1334 seq_printf(seq, ",inline_xattr_size=%u",
1335 F2FS_OPTION(sbi).inline_xattr_size);
1336 #endif
1337 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1338 if (test_opt(sbi, POSIX_ACL))
1339 seq_puts(seq, ",acl");
1340 else
1341 seq_puts(seq, ",noacl");
1342 #endif
1343 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1344 seq_puts(seq, ",disable_ext_identify");
1345 if (test_opt(sbi, INLINE_DATA))
1346 seq_puts(seq, ",inline_data");
1347 else
1348 seq_puts(seq, ",noinline_data");
1349 if (test_opt(sbi, INLINE_DENTRY))
1350 seq_puts(seq, ",inline_dentry");
1351 else
1352 seq_puts(seq, ",noinline_dentry");
1353 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1354 seq_puts(seq, ",flush_merge");
1355 if (test_opt(sbi, NOBARRIER))
1356 seq_puts(seq, ",nobarrier");
1357 if (test_opt(sbi, FASTBOOT))
1358 seq_puts(seq, ",fastboot");
1359 if (test_opt(sbi, EXTENT_CACHE))
1360 seq_puts(seq, ",extent_cache");
1361 else
1362 seq_puts(seq, ",noextent_cache");
1363 if (test_opt(sbi, DATA_FLUSH))
1364 seq_puts(seq, ",data_flush");
1365
1366 seq_puts(seq, ",mode=");
1367 if (test_opt(sbi, ADAPTIVE))
1368 seq_puts(seq, "adaptive");
1369 else if (test_opt(sbi, LFS))
1370 seq_puts(seq, "lfs");
1371 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1372 if (test_opt(sbi, RESERVE_ROOT))
1373 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1374 F2FS_OPTION(sbi).root_reserved_blocks,
1375 from_kuid_munged(&init_user_ns,
1376 F2FS_OPTION(sbi).s_resuid),
1377 from_kgid_munged(&init_user_ns,
1378 F2FS_OPTION(sbi).s_resgid));
1379 if (F2FS_IO_SIZE_BITS(sbi))
1380 seq_printf(seq, ",io_bits=%u",
1381 F2FS_OPTION(sbi).write_io_size_bits);
1382 #ifdef CONFIG_F2FS_FAULT_INJECTION
1383 if (test_opt(sbi, FAULT_INJECTION)) {
1384 seq_printf(seq, ",fault_injection=%u",
1385 F2FS_OPTION(sbi).fault_info.inject_rate);
1386 seq_printf(seq, ",fault_type=%u",
1387 F2FS_OPTION(sbi).fault_info.inject_type);
1388 }
1389 #endif
1390 #ifdef CONFIG_QUOTA
1391 if (test_opt(sbi, QUOTA))
1392 seq_puts(seq, ",quota");
1393 if (test_opt(sbi, USRQUOTA))
1394 seq_puts(seq, ",usrquota");
1395 if (test_opt(sbi, GRPQUOTA))
1396 seq_puts(seq, ",grpquota");
1397 if (test_opt(sbi, PRJQUOTA))
1398 seq_puts(seq, ",prjquota");
1399 #endif
1400 f2fs_show_quota_options(seq, sbi->sb);
1401 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1402 seq_printf(seq, ",whint_mode=%s", "user-based");
1403 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1404 seq_printf(seq, ",whint_mode=%s", "fs-based");
1405 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1406 if (F2FS_OPTION(sbi).test_dummy_encryption)
1407 seq_puts(seq, ",test_dummy_encryption");
1408 #endif
1409
1410 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1411 seq_printf(seq, ",alloc_mode=%s", "default");
1412 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1413 seq_printf(seq, ",alloc_mode=%s", "reuse");
1414
1415 if (test_opt(sbi, DISABLE_CHECKPOINT))
1416 seq_puts(seq, ",checkpoint=disable");
1417
1418 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1419 seq_printf(seq, ",fsync_mode=%s", "posix");
1420 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1421 seq_printf(seq, ",fsync_mode=%s", "strict");
1422 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1423 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1424 return 0;
1425 }
1426
default_options(struct f2fs_sb_info * sbi)1427 static void default_options(struct f2fs_sb_info *sbi)
1428 {
1429 /* init some FS parameters */
1430 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1431 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1432 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1433 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1434 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1435 F2FS_OPTION(sbi).test_dummy_encryption = false;
1436 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1437 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1438
1439 set_opt(sbi, BG_GC);
1440 set_opt(sbi, INLINE_XATTR);
1441 set_opt(sbi, INLINE_DATA);
1442 set_opt(sbi, INLINE_DENTRY);
1443 set_opt(sbi, EXTENT_CACHE);
1444 set_opt(sbi, NOHEAP);
1445 sbi->sb->s_flags |= MS_LAZYTIME;
1446 clear_opt(sbi, DISABLE_CHECKPOINT);
1447 set_opt(sbi, FLUSH_MERGE);
1448 set_opt(sbi, DISCARD);
1449 if (f2fs_sb_has_blkzoned(sbi))
1450 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1451 else
1452 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1453
1454 #ifdef CONFIG_F2FS_FS_XATTR
1455 set_opt(sbi, XATTR_USER);
1456 #endif
1457 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1458 set_opt(sbi, POSIX_ACL);
1459 #endif
1460
1461 f2fs_build_fault_attr(sbi, 0, 0);
1462 }
1463
1464 #ifdef CONFIG_QUOTA
1465 static int f2fs_enable_quotas(struct super_block *sb);
1466 #endif
1467
f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)1468 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1469 {
1470 unsigned int s_flags = sbi->sb->s_flags;
1471 struct cp_control cpc;
1472 int err = 0;
1473 int ret;
1474
1475 if (s_flags & SB_RDONLY) {
1476 f2fs_msg(sbi->sb, KERN_ERR,
1477 "checkpoint=disable on readonly fs");
1478 return -EINVAL;
1479 }
1480 sbi->sb->s_flags |= SB_ACTIVE;
1481
1482 f2fs_update_time(sbi, DISABLE_TIME);
1483
1484 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1485 mutex_lock(&sbi->gc_mutex);
1486 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1487 if (err == -ENODATA) {
1488 err = 0;
1489 break;
1490 }
1491 if (err && err != -EAGAIN)
1492 break;
1493 }
1494
1495 ret = sync_filesystem(sbi->sb);
1496 if (ret || err) {
1497 err = ret ? ret: err;
1498 goto restore_flag;
1499 }
1500
1501 if (f2fs_disable_cp_again(sbi)) {
1502 err = -EAGAIN;
1503 goto restore_flag;
1504 }
1505
1506 mutex_lock(&sbi->gc_mutex);
1507 cpc.reason = CP_PAUSE;
1508 set_sbi_flag(sbi, SBI_CP_DISABLED);
1509 f2fs_write_checkpoint(sbi, &cpc);
1510
1511 sbi->unusable_block_count = 0;
1512 mutex_unlock(&sbi->gc_mutex);
1513 restore_flag:
1514 sbi->sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1515 return err;
1516 }
1517
f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)1518 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1519 {
1520 mutex_lock(&sbi->gc_mutex);
1521 f2fs_dirty_to_prefree(sbi);
1522
1523 clear_sbi_flag(sbi, SBI_CP_DISABLED);
1524 set_sbi_flag(sbi, SBI_IS_DIRTY);
1525 mutex_unlock(&sbi->gc_mutex);
1526
1527 f2fs_sync_fs(sbi->sb, 1);
1528 }
1529
f2fs_remount(struct super_block * sb,int * flags,char * data)1530 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1531 {
1532 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1533 struct f2fs_mount_info org_mount_opt;
1534 unsigned long old_sb_flags;
1535 int err;
1536 bool need_restart_gc = false;
1537 bool need_stop_gc = false;
1538 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1539 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1540 bool checkpoint_changed;
1541 #ifdef CONFIG_QUOTA
1542 int i, j;
1543 #endif
1544
1545 /*
1546 * Save the old mount options in case we
1547 * need to restore them.
1548 */
1549 org_mount_opt = sbi->mount_opt;
1550 old_sb_flags = sb->s_flags;
1551
1552 #ifdef CONFIG_QUOTA
1553 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1554 for (i = 0; i < MAXQUOTAS; i++) {
1555 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1556 org_mount_opt.s_qf_names[i] =
1557 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1558 GFP_KERNEL);
1559 if (!org_mount_opt.s_qf_names[i]) {
1560 for (j = 0; j < i; j++)
1561 kvfree(org_mount_opt.s_qf_names[j]);
1562 return -ENOMEM;
1563 }
1564 } else {
1565 org_mount_opt.s_qf_names[i] = NULL;
1566 }
1567 }
1568 #endif
1569
1570 /* recover superblocks we couldn't write due to previous RO mount */
1571 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1572 err = f2fs_commit_super(sbi, false);
1573 f2fs_msg(sb, KERN_INFO,
1574 "Try to recover all the superblocks, ret: %d", err);
1575 if (!err)
1576 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1577 }
1578
1579 default_options(sbi);
1580
1581 /* parse mount options */
1582 err = parse_options(sb, data);
1583 if (err)
1584 goto restore_opts;
1585 checkpoint_changed =
1586 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1587
1588 /*
1589 * Previous and new state of filesystem is RO,
1590 * so skip checking GC and FLUSH_MERGE conditions.
1591 */
1592 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1593 goto skip;
1594
1595 #ifdef CONFIG_QUOTA
1596 if (!f2fs_readonly(sb) && (*flags & MS_RDONLY)) {
1597 err = dquot_suspend(sb, -1);
1598 if (err < 0)
1599 goto restore_opts;
1600 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1601 /* dquot_resume needs RW */
1602 sb->s_flags &= ~MS_RDONLY;
1603 if (sb_any_quota_suspended(sb)) {
1604 dquot_resume(sb, -1);
1605 } else if (f2fs_sb_has_quota_ino(sbi)) {
1606 err = f2fs_enable_quotas(sb);
1607 if (err)
1608 goto restore_opts;
1609 }
1610 }
1611 #endif
1612 /* disallow enable/disable extent_cache dynamically */
1613 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1614 err = -EINVAL;
1615 f2fs_msg(sbi->sb, KERN_WARNING,
1616 "switch extent_cache option is not allowed");
1617 goto restore_opts;
1618 }
1619
1620 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1621 err = -EINVAL;
1622 f2fs_msg(sbi->sb, KERN_WARNING,
1623 "disabling checkpoint not compatible with read-only");
1624 goto restore_opts;
1625 }
1626
1627 /*
1628 * We stop the GC thread if FS is mounted as RO
1629 * or if background_gc = off is passed in mount
1630 * option. Also sync the filesystem.
1631 */
1632 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1633 if (sbi->gc_thread) {
1634 f2fs_stop_gc_thread(sbi);
1635 need_restart_gc = true;
1636 }
1637 } else if (!sbi->gc_thread) {
1638 err = f2fs_start_gc_thread(sbi);
1639 if (err)
1640 goto restore_opts;
1641 need_stop_gc = true;
1642 }
1643
1644 if (*flags & MS_RDONLY ||
1645 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1646 writeback_inodes_sb(sb, WB_REASON_SYNC);
1647 sync_inodes_sb(sb);
1648
1649 set_sbi_flag(sbi, SBI_IS_DIRTY);
1650 set_sbi_flag(sbi, SBI_IS_CLOSE);
1651 f2fs_sync_fs(sb, 1);
1652 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1653 }
1654
1655 if (checkpoint_changed) {
1656 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1657 err = f2fs_disable_checkpoint(sbi);
1658 if (err)
1659 goto restore_gc;
1660 } else {
1661 f2fs_enable_checkpoint(sbi);
1662 }
1663 }
1664
1665 /*
1666 * We stop issue flush thread if FS is mounted as RO
1667 * or if flush_merge is not passed in mount option.
1668 */
1669 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1670 clear_opt(sbi, FLUSH_MERGE);
1671 f2fs_destroy_flush_cmd_control(sbi, false);
1672 } else {
1673 err = f2fs_create_flush_cmd_control(sbi);
1674 if (err)
1675 goto restore_gc;
1676 }
1677 skip:
1678 #ifdef CONFIG_QUOTA
1679 /* Release old quota file names */
1680 for (i = 0; i < MAXQUOTAS; i++)
1681 kvfree(org_mount_opt.s_qf_names[i]);
1682 #endif
1683 /* Update the POSIXACL Flag */
1684 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1685 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1686
1687 limit_reserve_root(sbi);
1688 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1689 return 0;
1690 restore_gc:
1691 if (need_restart_gc) {
1692 if (f2fs_start_gc_thread(sbi))
1693 f2fs_msg(sbi->sb, KERN_WARNING,
1694 "background gc thread has stopped");
1695 } else if (need_stop_gc) {
1696 f2fs_stop_gc_thread(sbi);
1697 }
1698 restore_opts:
1699 #ifdef CONFIG_QUOTA
1700 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1701 for (i = 0; i < MAXQUOTAS; i++) {
1702 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1703 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1704 }
1705 #endif
1706 sbi->mount_opt = org_mount_opt;
1707 sb->s_flags = old_sb_flags;
1708 return err;
1709 }
1710
1711 #ifdef CONFIG_QUOTA
1712 /* Read data from quotafile */
f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)1713 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1714 size_t len, loff_t off)
1715 {
1716 struct inode *inode = sb_dqopt(sb)->files[type];
1717 struct address_space *mapping = inode->i_mapping;
1718 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1719 int offset = off & (sb->s_blocksize - 1);
1720 int tocopy;
1721 size_t toread;
1722 loff_t i_size = i_size_read(inode);
1723 struct page *page;
1724 char *kaddr;
1725
1726 if (off > i_size)
1727 return 0;
1728
1729 if (off + len > i_size)
1730 len = i_size - off;
1731 toread = len;
1732 while (toread > 0) {
1733 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1734 repeat:
1735 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1736 if (IS_ERR(page)) {
1737 if (PTR_ERR(page) == -ENOMEM) {
1738 congestion_wait(BLK_RW_ASYNC, HZ/50);
1739 goto repeat;
1740 }
1741 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1742 return PTR_ERR(page);
1743 }
1744
1745 lock_page(page);
1746
1747 if (unlikely(page->mapping != mapping)) {
1748 f2fs_put_page(page, 1);
1749 goto repeat;
1750 }
1751 if (unlikely(!PageUptodate(page))) {
1752 f2fs_put_page(page, 1);
1753 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1754 return -EIO;
1755 }
1756
1757 kaddr = kmap_atomic(page);
1758 memcpy(data, kaddr + offset, tocopy);
1759 kunmap_atomic(kaddr);
1760 f2fs_put_page(page, 1);
1761
1762 offset = 0;
1763 toread -= tocopy;
1764 data += tocopy;
1765 blkidx++;
1766 }
1767 return len;
1768 }
1769
1770 /* Write to quotafile */
f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)1771 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1772 const char *data, size_t len, loff_t off)
1773 {
1774 struct inode *inode = sb_dqopt(sb)->files[type];
1775 struct address_space *mapping = inode->i_mapping;
1776 const struct address_space_operations *a_ops = mapping->a_ops;
1777 int offset = off & (sb->s_blocksize - 1);
1778 size_t towrite = len;
1779 struct page *page;
1780 char *kaddr;
1781 int err = 0;
1782 int tocopy;
1783
1784 while (towrite > 0) {
1785 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1786 towrite);
1787 retry:
1788 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1789 &page, NULL);
1790 if (unlikely(err)) {
1791 if (err == -ENOMEM) {
1792 congestion_wait(BLK_RW_ASYNC, HZ/50);
1793 goto retry;
1794 }
1795 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1796 break;
1797 }
1798
1799 kaddr = kmap_atomic(page);
1800 memcpy(kaddr + offset, data, tocopy);
1801 kunmap_atomic(kaddr);
1802 flush_dcache_page(page);
1803
1804 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1805 page, NULL);
1806 offset = 0;
1807 towrite -= tocopy;
1808 off += tocopy;
1809 data += tocopy;
1810 cond_resched();
1811 }
1812
1813 if (len == towrite)
1814 return err;
1815 inode->i_mtime = inode->i_ctime = current_time(inode);
1816 f2fs_mark_inode_dirty_sync(inode, false);
1817 return len - towrite;
1818 }
1819
f2fs_get_dquots(struct inode * inode)1820 static struct dquot **f2fs_get_dquots(struct inode *inode)
1821 {
1822 return F2FS_I(inode)->i_dquot;
1823 }
1824
f2fs_get_reserved_space(struct inode * inode)1825 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1826 {
1827 return &F2FS_I(inode)->i_reserved_quota;
1828 }
1829
f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)1830 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1831 {
1832 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1833 f2fs_msg(sbi->sb, KERN_ERR,
1834 "quota sysfile may be corrupted, skip loading it");
1835 return 0;
1836 }
1837
1838 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1839 F2FS_OPTION(sbi).s_jquota_fmt, type);
1840 }
1841
f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)1842 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1843 {
1844 int enabled = 0;
1845 int i, err;
1846
1847 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1848 err = f2fs_enable_quotas(sbi->sb);
1849 if (err) {
1850 f2fs_msg(sbi->sb, KERN_ERR,
1851 "Cannot turn on quota_ino: %d", err);
1852 return 0;
1853 }
1854 return 1;
1855 }
1856
1857 for (i = 0; i < MAXQUOTAS; i++) {
1858 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1859 err = f2fs_quota_on_mount(sbi, i);
1860 if (!err) {
1861 enabled = 1;
1862 continue;
1863 }
1864 f2fs_msg(sbi->sb, KERN_ERR,
1865 "Cannot turn on quotas: %d on %d", err, i);
1866 }
1867 }
1868 return enabled;
1869 }
1870
f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)1871 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1872 unsigned int flags)
1873 {
1874 struct inode *qf_inode;
1875 unsigned long qf_inum;
1876 int err;
1877
1878 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1879
1880 qf_inum = f2fs_qf_ino(sb, type);
1881 if (!qf_inum)
1882 return -EPERM;
1883
1884 qf_inode = f2fs_iget(sb, qf_inum);
1885 if (IS_ERR(qf_inode)) {
1886 f2fs_msg(sb, KERN_ERR,
1887 "Bad quota inode %u:%lu", type, qf_inum);
1888 return PTR_ERR(qf_inode);
1889 }
1890
1891 /* Don't account quota for quota files to avoid recursion */
1892 qf_inode->i_flags |= S_NOQUOTA;
1893 err = dquot_enable(qf_inode, type, format_id, flags);
1894 iput(qf_inode);
1895 return err;
1896 }
1897
f2fs_enable_quotas(struct super_block * sb)1898 static int f2fs_enable_quotas(struct super_block *sb)
1899 {
1900 int type, err = 0;
1901 unsigned long qf_inum;
1902 bool quota_mopt[MAXQUOTAS] = {
1903 test_opt(F2FS_SB(sb), USRQUOTA),
1904 test_opt(F2FS_SB(sb), GRPQUOTA),
1905 test_opt(F2FS_SB(sb), PRJQUOTA),
1906 };
1907
1908 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1909 f2fs_msg(sb, KERN_ERR,
1910 "quota file may be corrupted, skip loading it");
1911 return 0;
1912 }
1913
1914 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1915
1916 for (type = 0; type < MAXQUOTAS; type++) {
1917 qf_inum = f2fs_qf_ino(sb, type);
1918 if (qf_inum) {
1919 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1920 DQUOT_USAGE_ENABLED |
1921 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1922 if (err) {
1923 f2fs_msg(sb, KERN_ERR,
1924 "Failed to enable quota tracking "
1925 "(type=%d, err=%d). Please run "
1926 "fsck to fix.", type, err);
1927 for (type--; type >= 0; type--)
1928 dquot_quota_off(sb, type);
1929 set_sbi_flag(F2FS_SB(sb),
1930 SBI_QUOTA_NEED_REPAIR);
1931 return err;
1932 }
1933 }
1934 }
1935 return 0;
1936 }
1937
f2fs_quota_sync(struct super_block * sb,int type)1938 int f2fs_quota_sync(struct super_block *sb, int type)
1939 {
1940 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1941 struct quota_info *dqopt = sb_dqopt(sb);
1942 int cnt;
1943 int ret;
1944
1945 ret = dquot_writeback_dquots(sb, type);
1946 if (ret)
1947 goto out;
1948
1949 /*
1950 * Now when everything is written we can discard the pagecache so
1951 * that userspace sees the changes.
1952 */
1953 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1954 struct address_space *mapping;
1955
1956 if (type != -1 && cnt != type)
1957 continue;
1958 if (!sb_has_quota_active(sb, cnt))
1959 continue;
1960
1961 mapping = dqopt->files[cnt]->i_mapping;
1962
1963 ret = filemap_fdatawrite(mapping);
1964 if (ret)
1965 goto out;
1966
1967 /* if we are using journalled quota */
1968 if (is_journalled_quota(sbi))
1969 continue;
1970
1971 ret = filemap_fdatawait(mapping);
1972 if (ret)
1973 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1974
1975 inode_lock(dqopt->files[cnt]);
1976 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1977 inode_unlock(dqopt->files[cnt]);
1978 }
1979 out:
1980 if (ret)
1981 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1982 return ret;
1983 }
1984
f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)1985 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1986 const struct path *path)
1987 {
1988 struct inode *inode;
1989 int err;
1990
1991 err = f2fs_quota_sync(sb, type);
1992 if (err)
1993 return err;
1994
1995 err = dquot_quota_on(sb, type, format_id, path);
1996 if (err)
1997 return err;
1998
1999 inode = d_inode(path->dentry);
2000
2001 inode_lock(inode);
2002 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2003 f2fs_set_inode_flags(inode);
2004 inode_unlock(inode);
2005 f2fs_mark_inode_dirty_sync(inode, false);
2006
2007 return 0;
2008 }
2009
f2fs_quota_off(struct super_block * sb,int type)2010 static int f2fs_quota_off(struct super_block *sb, int type)
2011 {
2012 struct inode *inode = sb_dqopt(sb)->files[type];
2013 int err;
2014
2015 if (!inode || !igrab(inode))
2016 return dquot_quota_off(sb, type);
2017
2018 err = f2fs_quota_sync(sb, type);
2019 if (err)
2020 goto out_put;
2021
2022 err = dquot_quota_off(sb, type);
2023 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2024 goto out_put;
2025
2026 inode_lock(inode);
2027 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2028 f2fs_set_inode_flags(inode);
2029 inode_unlock(inode);
2030 f2fs_mark_inode_dirty_sync(inode, false);
2031 out_put:
2032 iput(inode);
2033 return err;
2034 }
2035
f2fs_quota_off_umount(struct super_block * sb)2036 void f2fs_quota_off_umount(struct super_block *sb)
2037 {
2038 int type;
2039 int err;
2040
2041 for (type = 0; type < MAXQUOTAS; type++) {
2042 err = f2fs_quota_off(sb, type);
2043 if (err) {
2044 int ret = dquot_quota_off(sb, type);
2045
2046 f2fs_msg(sb, KERN_ERR,
2047 "Fail to turn off disk quota "
2048 "(type: %d, err: %d, ret:%d), Please "
2049 "run fsck to fix it.", type, err, ret);
2050 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2051 }
2052 }
2053 /*
2054 * In case of checkpoint=disable, we must flush quota blocks.
2055 * This can cause NULL exception for node_inode in end_io, since
2056 * put_super already dropped it.
2057 */
2058 sync_filesystem(sb);
2059 }
2060
f2fs_truncate_quota_inode_pages(struct super_block * sb)2061 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2062 {
2063 struct quota_info *dqopt = sb_dqopt(sb);
2064 int type;
2065
2066 for (type = 0; type < MAXQUOTAS; type++) {
2067 if (!dqopt->files[type])
2068 continue;
2069 f2fs_inode_synced(dqopt->files[type]);
2070 }
2071 }
2072
f2fs_dquot_commit(struct dquot * dquot)2073 static int f2fs_dquot_commit(struct dquot *dquot)
2074 {
2075 int ret;
2076
2077 ret = dquot_commit(dquot);
2078 if (ret < 0)
2079 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2080 return ret;
2081 }
2082
f2fs_dquot_acquire(struct dquot * dquot)2083 static int f2fs_dquot_acquire(struct dquot *dquot)
2084 {
2085 int ret;
2086
2087 ret = dquot_acquire(dquot);
2088 if (ret < 0)
2089 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2090
2091 return ret;
2092 }
2093
f2fs_dquot_release(struct dquot * dquot)2094 static int f2fs_dquot_release(struct dquot *dquot)
2095 {
2096 int ret;
2097
2098 ret = dquot_release(dquot);
2099 if (ret < 0)
2100 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2101 return ret;
2102 }
2103
f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)2104 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2105 {
2106 struct super_block *sb = dquot->dq_sb;
2107 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2108 int ret;
2109
2110 ret = dquot_mark_dquot_dirty(dquot);
2111
2112 /* if we are using journalled quota */
2113 if (is_journalled_quota(sbi))
2114 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2115
2116 return ret;
2117 }
2118
f2fs_dquot_commit_info(struct super_block * sb,int type)2119 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2120 {
2121 int ret;
2122
2123 ret = dquot_commit_info(sb, type);
2124 if (ret < 0)
2125 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2126 return ret;
2127 }
2128
f2fs_get_projid(struct inode * inode,kprojid_t * projid)2129 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2130 {
2131 *projid = F2FS_I(inode)->i_projid;
2132 return 0;
2133 }
2134
2135 static const struct dquot_operations f2fs_quota_operations = {
2136 .get_reserved_space = f2fs_get_reserved_space,
2137 .write_dquot = f2fs_dquot_commit,
2138 .acquire_dquot = f2fs_dquot_acquire,
2139 .release_dquot = f2fs_dquot_release,
2140 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2141 .write_info = f2fs_dquot_commit_info,
2142 .alloc_dquot = dquot_alloc,
2143 .destroy_dquot = dquot_destroy,
2144 .get_projid = f2fs_get_projid,
2145 .get_next_id = dquot_get_next_id,
2146 };
2147
2148 static const struct quotactl_ops f2fs_quotactl_ops = {
2149 .quota_on = f2fs_quota_on,
2150 .quota_off = f2fs_quota_off,
2151 .quota_sync = f2fs_quota_sync,
2152 .get_state = dquot_get_state,
2153 .set_info = dquot_set_dqinfo,
2154 .get_dqblk = dquot_get_dqblk,
2155 .set_dqblk = dquot_set_dqblk,
2156 .get_nextdqblk = dquot_get_next_dqblk,
2157 };
2158 #else
f2fs_quota_sync(struct super_block * sb,int type)2159 int f2fs_quota_sync(struct super_block *sb, int type)
2160 {
2161 return 0;
2162 }
2163
f2fs_quota_off_umount(struct super_block * sb)2164 void f2fs_quota_off_umount(struct super_block *sb)
2165 {
2166 }
2167 #endif
2168
2169 static const struct super_operations f2fs_sops = {
2170 .alloc_inode = f2fs_alloc_inode,
2171 .drop_inode = f2fs_drop_inode,
2172 .destroy_inode = f2fs_destroy_inode,
2173 .write_inode = f2fs_write_inode,
2174 .dirty_inode = f2fs_dirty_inode,
2175 .show_options = f2fs_show_options,
2176 #ifdef CONFIG_QUOTA
2177 .quota_read = f2fs_quota_read,
2178 .quota_write = f2fs_quota_write,
2179 .get_dquots = f2fs_get_dquots,
2180 #endif
2181 .evict_inode = f2fs_evict_inode,
2182 .put_super = f2fs_put_super,
2183 .sync_fs = f2fs_sync_fs,
2184 .freeze_fs = f2fs_freeze,
2185 .unfreeze_fs = f2fs_unfreeze,
2186 .statfs = f2fs_statfs,
2187 .remount_fs = f2fs_remount,
2188 };
2189
2190 #ifdef CONFIG_F2FS_FS_ENCRYPTION
f2fs_get_context(struct inode * inode,void * ctx,size_t len)2191 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2192 {
2193 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2194 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2195 ctx, len, NULL);
2196 }
2197
f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)2198 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2199 void *fs_data)
2200 {
2201 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2202
2203 /*
2204 * Encrypting the root directory is not allowed because fsck
2205 * expects lost+found directory to exist and remain unencrypted
2206 * if LOST_FOUND feature is enabled.
2207 *
2208 */
2209 if (f2fs_sb_has_lost_found(sbi) &&
2210 inode->i_ino == F2FS_ROOT_INO(sbi))
2211 return -EPERM;
2212
2213 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2214 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2215 ctx, len, fs_data, XATTR_CREATE);
2216 }
2217
f2fs_dummy_context(struct inode * inode)2218 static bool f2fs_dummy_context(struct inode *inode)
2219 {
2220 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2221 }
2222
2223 static const struct fscrypt_operations f2fs_cryptops = {
2224 .key_prefix = "f2fs:",
2225 .get_context = f2fs_get_context,
2226 .set_context = f2fs_set_context,
2227 .dummy_context = f2fs_dummy_context,
2228 .empty_dir = f2fs_empty_dir,
2229 .max_namelen = F2FS_NAME_LEN,
2230 };
2231 #endif
2232
f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)2233 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2234 u64 ino, u32 generation)
2235 {
2236 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2237 struct inode *inode;
2238
2239 if (f2fs_check_nid_range(sbi, ino))
2240 return ERR_PTR(-ESTALE);
2241
2242 /*
2243 * f2fs_iget isn't quite right if the inode is currently unallocated!
2244 * However f2fs_iget currently does appropriate checks to handle stale
2245 * inodes so everything is OK.
2246 */
2247 inode = f2fs_iget(sb, ino);
2248 if (IS_ERR(inode))
2249 return ERR_CAST(inode);
2250 if (unlikely(generation && inode->i_generation != generation)) {
2251 /* we didn't find the right inode.. */
2252 iput(inode);
2253 return ERR_PTR(-ESTALE);
2254 }
2255 return inode;
2256 }
2257
f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)2258 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2259 int fh_len, int fh_type)
2260 {
2261 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2262 f2fs_nfs_get_inode);
2263 }
2264
f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)2265 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2266 int fh_len, int fh_type)
2267 {
2268 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2269 f2fs_nfs_get_inode);
2270 }
2271
2272 static const struct export_operations f2fs_export_ops = {
2273 .fh_to_dentry = f2fs_fh_to_dentry,
2274 .fh_to_parent = f2fs_fh_to_parent,
2275 .get_parent = f2fs_get_parent,
2276 };
2277
max_file_blocks(void)2278 static loff_t max_file_blocks(void)
2279 {
2280 loff_t result = 0;
2281 loff_t leaf_count = ADDRS_PER_BLOCK;
2282
2283 /*
2284 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2285 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2286 * space in inode.i_addr, it will be more safe to reassign
2287 * result as zero.
2288 */
2289
2290 /* two direct node blocks */
2291 result += (leaf_count * 2);
2292
2293 /* two indirect node blocks */
2294 leaf_count *= NIDS_PER_BLOCK;
2295 result += (leaf_count * 2);
2296
2297 /* one double indirect node block */
2298 leaf_count *= NIDS_PER_BLOCK;
2299 result += leaf_count;
2300
2301 return result;
2302 }
2303
__f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)2304 static int __f2fs_commit_super(struct buffer_head *bh,
2305 struct f2fs_super_block *super)
2306 {
2307 lock_buffer(bh);
2308 if (super)
2309 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2310 set_buffer_dirty(bh);
2311 unlock_buffer(bh);
2312
2313 /* it's rare case, we can do fua all the time */
2314 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2315 }
2316
sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)2317 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2318 struct buffer_head *bh)
2319 {
2320 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2321 (bh->b_data + F2FS_SUPER_OFFSET);
2322 struct super_block *sb = sbi->sb;
2323 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2324 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2325 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2326 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2327 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2328 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2329 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2330 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2331 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2332 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2333 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2334 u32 segment_count = le32_to_cpu(raw_super->segment_count);
2335 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2336 u64 main_end_blkaddr = main_blkaddr +
2337 (segment_count_main << log_blocks_per_seg);
2338 u64 seg_end_blkaddr = segment0_blkaddr +
2339 (segment_count << log_blocks_per_seg);
2340
2341 if (segment0_blkaddr != cp_blkaddr) {
2342 f2fs_msg(sb, KERN_INFO,
2343 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2344 segment0_blkaddr, cp_blkaddr);
2345 return true;
2346 }
2347
2348 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2349 sit_blkaddr) {
2350 f2fs_msg(sb, KERN_INFO,
2351 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2352 cp_blkaddr, sit_blkaddr,
2353 segment_count_ckpt << log_blocks_per_seg);
2354 return true;
2355 }
2356
2357 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2358 nat_blkaddr) {
2359 f2fs_msg(sb, KERN_INFO,
2360 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2361 sit_blkaddr, nat_blkaddr,
2362 segment_count_sit << log_blocks_per_seg);
2363 return true;
2364 }
2365
2366 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2367 ssa_blkaddr) {
2368 f2fs_msg(sb, KERN_INFO,
2369 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2370 nat_blkaddr, ssa_blkaddr,
2371 segment_count_nat << log_blocks_per_seg);
2372 return true;
2373 }
2374
2375 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2376 main_blkaddr) {
2377 f2fs_msg(sb, KERN_INFO,
2378 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2379 ssa_blkaddr, main_blkaddr,
2380 segment_count_ssa << log_blocks_per_seg);
2381 return true;
2382 }
2383
2384 if (main_end_blkaddr > seg_end_blkaddr) {
2385 f2fs_msg(sb, KERN_INFO,
2386 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2387 main_blkaddr,
2388 segment0_blkaddr +
2389 (segment_count << log_blocks_per_seg),
2390 segment_count_main << log_blocks_per_seg);
2391 return true;
2392 } else if (main_end_blkaddr < seg_end_blkaddr) {
2393 int err = 0;
2394 char *res;
2395
2396 /* fix in-memory information all the time */
2397 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2398 segment0_blkaddr) >> log_blocks_per_seg);
2399
2400 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2401 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2402 res = "internally";
2403 } else {
2404 err = __f2fs_commit_super(bh, NULL);
2405 res = err ? "failed" : "done";
2406 }
2407 f2fs_msg(sb, KERN_INFO,
2408 "Fix alignment : %s, start(%u) end(%u) block(%u)",
2409 res, main_blkaddr,
2410 segment0_blkaddr +
2411 (segment_count << log_blocks_per_seg),
2412 segment_count_main << log_blocks_per_seg);
2413 if (err)
2414 return true;
2415 }
2416 return false;
2417 }
2418
sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)2419 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2420 struct buffer_head *bh)
2421 {
2422 block_t segment_count, segs_per_sec, secs_per_zone;
2423 block_t total_sections, blocks_per_seg;
2424 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2425 (bh->b_data + F2FS_SUPER_OFFSET);
2426 struct super_block *sb = sbi->sb;
2427 unsigned int blocksize;
2428 size_t crc_offset = 0;
2429 __u32 crc = 0;
2430
2431 /* Check checksum_offset and crc in superblock */
2432 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2433 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2434 if (crc_offset !=
2435 offsetof(struct f2fs_super_block, crc)) {
2436 f2fs_msg(sb, KERN_INFO,
2437 "Invalid SB checksum offset: %zu",
2438 crc_offset);
2439 return 1;
2440 }
2441 crc = le32_to_cpu(raw_super->crc);
2442 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2443 f2fs_msg(sb, KERN_INFO,
2444 "Invalid SB checksum value: %u", crc);
2445 return 1;
2446 }
2447 }
2448
2449 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2450 f2fs_msg(sb, KERN_INFO,
2451 "Magic Mismatch, valid(0x%x) - read(0x%x)",
2452 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2453 return -EINVAL;
2454 }
2455
2456 /* Currently, support only 4KB page cache size */
2457 if (F2FS_BLKSIZE != PAGE_SIZE) {
2458 f2fs_msg(sb, KERN_INFO,
2459 "Invalid page_cache_size (%lu), supports only 4KB\n",
2460 PAGE_SIZE);
2461 return -EFSCORRUPTED;
2462 }
2463
2464 /* Currently, support only 4KB block size */
2465 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2466 if (blocksize != F2FS_BLKSIZE) {
2467 f2fs_msg(sb, KERN_INFO,
2468 "Invalid blocksize (%u), supports only 4KB\n",
2469 blocksize);
2470 return -EFSCORRUPTED;
2471 }
2472
2473 /* check log blocks per segment */
2474 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2475 f2fs_msg(sb, KERN_INFO,
2476 "Invalid log blocks per segment (%u)\n",
2477 le32_to_cpu(raw_super->log_blocks_per_seg));
2478 return -EFSCORRUPTED;
2479 }
2480
2481 /* Currently, support 512/1024/2048/4096 bytes sector size */
2482 if (le32_to_cpu(raw_super->log_sectorsize) >
2483 F2FS_MAX_LOG_SECTOR_SIZE ||
2484 le32_to_cpu(raw_super->log_sectorsize) <
2485 F2FS_MIN_LOG_SECTOR_SIZE) {
2486 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2487 le32_to_cpu(raw_super->log_sectorsize));
2488 return -EFSCORRUPTED;
2489 }
2490 if (le32_to_cpu(raw_super->log_sectors_per_block) +
2491 le32_to_cpu(raw_super->log_sectorsize) !=
2492 F2FS_MAX_LOG_SECTOR_SIZE) {
2493 f2fs_msg(sb, KERN_INFO,
2494 "Invalid log sectors per block(%u) log sectorsize(%u)",
2495 le32_to_cpu(raw_super->log_sectors_per_block),
2496 le32_to_cpu(raw_super->log_sectorsize));
2497 return -EFSCORRUPTED;
2498 }
2499
2500 segment_count = le32_to_cpu(raw_super->segment_count);
2501 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2502 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2503 total_sections = le32_to_cpu(raw_super->section_count);
2504
2505 /* blocks_per_seg should be 512, given the above check */
2506 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2507
2508 if (segment_count > F2FS_MAX_SEGMENT ||
2509 segment_count < F2FS_MIN_SEGMENTS) {
2510 f2fs_msg(sb, KERN_INFO,
2511 "Invalid segment count (%u)",
2512 segment_count);
2513 return -EFSCORRUPTED;
2514 }
2515
2516 if (total_sections > segment_count ||
2517 total_sections < F2FS_MIN_SEGMENTS ||
2518 segs_per_sec > segment_count || !segs_per_sec) {
2519 f2fs_msg(sb, KERN_INFO,
2520 "Invalid segment/section count (%u, %u x %u)",
2521 segment_count, total_sections, segs_per_sec);
2522 return -EFSCORRUPTED;
2523 }
2524
2525 if ((segment_count / segs_per_sec) < total_sections) {
2526 f2fs_msg(sb, KERN_INFO,
2527 "Small segment_count (%u < %u * %u)",
2528 segment_count, segs_per_sec, total_sections);
2529 return -EFSCORRUPTED;
2530 }
2531
2532 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2533 f2fs_msg(sb, KERN_INFO,
2534 "Wrong segment_count / block_count (%u > %llu)",
2535 segment_count, le64_to_cpu(raw_super->block_count));
2536 return -EFSCORRUPTED;
2537 }
2538
2539 if (secs_per_zone > total_sections || !secs_per_zone) {
2540 f2fs_msg(sb, KERN_INFO,
2541 "Wrong secs_per_zone / total_sections (%u, %u)",
2542 secs_per_zone, total_sections);
2543 return -EFSCORRUPTED;
2544 }
2545 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2546 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2547 (le32_to_cpu(raw_super->extension_count) +
2548 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2549 f2fs_msg(sb, KERN_INFO,
2550 "Corrupted extension count (%u + %u > %u)",
2551 le32_to_cpu(raw_super->extension_count),
2552 raw_super->hot_ext_count,
2553 F2FS_MAX_EXTENSION);
2554 return -EFSCORRUPTED;
2555 }
2556
2557 if (le32_to_cpu(raw_super->cp_payload) >
2558 (blocks_per_seg - F2FS_CP_PACKS)) {
2559 f2fs_msg(sb, KERN_INFO,
2560 "Insane cp_payload (%u > %u)",
2561 le32_to_cpu(raw_super->cp_payload),
2562 blocks_per_seg - F2FS_CP_PACKS);
2563 return -EFSCORRUPTED;
2564 }
2565
2566 /* check reserved ino info */
2567 if (le32_to_cpu(raw_super->node_ino) != 1 ||
2568 le32_to_cpu(raw_super->meta_ino) != 2 ||
2569 le32_to_cpu(raw_super->root_ino) != 3) {
2570 f2fs_msg(sb, KERN_INFO,
2571 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2572 le32_to_cpu(raw_super->node_ino),
2573 le32_to_cpu(raw_super->meta_ino),
2574 le32_to_cpu(raw_super->root_ino));
2575 return -EFSCORRUPTED;
2576 }
2577
2578 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2579 if (sanity_check_area_boundary(sbi, bh))
2580 return -EFSCORRUPTED;
2581
2582 return 0;
2583 }
2584
f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)2585 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2586 {
2587 unsigned int total, fsmeta;
2588 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2589 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2590 unsigned int ovp_segments, reserved_segments;
2591 unsigned int main_segs, blocks_per_seg;
2592 unsigned int sit_segs, nat_segs;
2593 unsigned int sit_bitmap_size, nat_bitmap_size;
2594 unsigned int log_blocks_per_seg;
2595 unsigned int segment_count_main;
2596 unsigned int cp_pack_start_sum, cp_payload;
2597 block_t user_block_count;
2598 int i, j;
2599
2600 total = le32_to_cpu(raw_super->segment_count);
2601 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2602 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2603 fsmeta += sit_segs;
2604 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2605 fsmeta += nat_segs;
2606 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2607 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2608
2609 if (unlikely(fsmeta >= total))
2610 return 1;
2611
2612 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2613 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2614
2615 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2616 ovp_segments == 0 || reserved_segments == 0)) {
2617 f2fs_msg(sbi->sb, KERN_ERR,
2618 "Wrong layout: check mkfs.f2fs version");
2619 return 1;
2620 }
2621
2622 user_block_count = le64_to_cpu(ckpt->user_block_count);
2623 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2624 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2625 if (!user_block_count || user_block_count >=
2626 segment_count_main << log_blocks_per_seg) {
2627 f2fs_msg(sbi->sb, KERN_ERR,
2628 "Wrong user_block_count: %u", user_block_count);
2629 return 1;
2630 }
2631
2632 main_segs = le32_to_cpu(raw_super->segment_count_main);
2633 blocks_per_seg = sbi->blocks_per_seg;
2634
2635 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2636 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2637 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2638 return 1;
2639 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2640 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2641 le32_to_cpu(ckpt->cur_node_segno[j])) {
2642 f2fs_msg(sbi->sb, KERN_ERR,
2643 "Node segment (%u, %u) has the same "
2644 "segno: %u", i, j,
2645 le32_to_cpu(ckpt->cur_node_segno[i]));
2646 return 1;
2647 }
2648 }
2649 }
2650 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2651 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2652 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2653 return 1;
2654 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2655 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2656 le32_to_cpu(ckpt->cur_data_segno[j])) {
2657 f2fs_msg(sbi->sb, KERN_ERR,
2658 "Data segment (%u, %u) has the same "
2659 "segno: %u", i, j,
2660 le32_to_cpu(ckpt->cur_data_segno[i]));
2661 return 1;
2662 }
2663 }
2664 }
2665 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2666 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2667 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2668 le32_to_cpu(ckpt->cur_data_segno[j])) {
2669 f2fs_msg(sbi->sb, KERN_ERR,
2670 "Node segment (%u) and Data segment (%u)"
2671 " has the same segno: %u", i, j,
2672 le32_to_cpu(ckpt->cur_node_segno[i]));
2673 return 1;
2674 }
2675 }
2676 }
2677
2678 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2679 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2680
2681 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2682 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2683 f2fs_msg(sbi->sb, KERN_ERR,
2684 "Wrong bitmap size: sit: %u, nat:%u",
2685 sit_bitmap_size, nat_bitmap_size);
2686 return 1;
2687 }
2688
2689 cp_pack_start_sum = __start_sum_addr(sbi);
2690 cp_payload = __cp_payload(sbi);
2691 if (cp_pack_start_sum < cp_payload + 1 ||
2692 cp_pack_start_sum > blocks_per_seg - 1 -
2693 NR_CURSEG_TYPE) {
2694 f2fs_msg(sbi->sb, KERN_ERR,
2695 "Wrong cp_pack_start_sum: %u",
2696 cp_pack_start_sum);
2697 return 1;
2698 }
2699
2700 if (unlikely(f2fs_cp_error(sbi))) {
2701 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2702 return 1;
2703 }
2704 return 0;
2705 }
2706
init_sb_info(struct f2fs_sb_info * sbi)2707 static void init_sb_info(struct f2fs_sb_info *sbi)
2708 {
2709 struct f2fs_super_block *raw_super = sbi->raw_super;
2710 int i;
2711
2712 sbi->log_sectors_per_block =
2713 le32_to_cpu(raw_super->log_sectors_per_block);
2714 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2715 sbi->blocksize = 1 << sbi->log_blocksize;
2716 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2717 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2718 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2719 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2720 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2721 sbi->total_node_count =
2722 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2723 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2724 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2725 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2726 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2727 sbi->cur_victim_sec = NULL_SECNO;
2728 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2729 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2730 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2731 sbi->migration_granularity = sbi->segs_per_sec;
2732
2733 sbi->dir_level = DEF_DIR_LEVEL;
2734 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2735 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2736 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2737 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2738 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2739 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2740 DEF_UMOUNT_DISCARD_TIMEOUT;
2741 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2742
2743 for (i = 0; i < NR_COUNT_TYPE; i++)
2744 atomic_set(&sbi->nr_pages[i], 0);
2745
2746 for (i = 0; i < META; i++)
2747 atomic_set(&sbi->wb_sync_req[i], 0);
2748
2749 INIT_LIST_HEAD(&sbi->s_list);
2750 mutex_init(&sbi->umount_mutex);
2751 init_rwsem(&sbi->io_order_lock);
2752 spin_lock_init(&sbi->cp_lock);
2753
2754 sbi->dirty_device = 0;
2755 spin_lock_init(&sbi->dev_lock);
2756
2757 init_rwsem(&sbi->sb_lock);
2758 }
2759
init_percpu_info(struct f2fs_sb_info * sbi)2760 static int init_percpu_info(struct f2fs_sb_info *sbi)
2761 {
2762 int err;
2763
2764 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2765 if (err)
2766 return err;
2767
2768 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2769 GFP_KERNEL);
2770 if (err)
2771 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2772
2773 return err;
2774 }
2775
2776 #ifdef CONFIG_BLK_DEV_ZONED
init_blkz_info(struct f2fs_sb_info * sbi,int devi)2777 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2778 {
2779 struct block_device *bdev = FDEV(devi).bdev;
2780 sector_t nr_sectors = bdev->bd_part->nr_sects;
2781 sector_t sector = 0;
2782 struct blk_zone *zones;
2783 unsigned int i, nr_zones;
2784 unsigned int n = 0;
2785 int err = -EIO;
2786
2787 if (!f2fs_sb_has_blkzoned(sbi))
2788 return 0;
2789
2790 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2791 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2792 return -EINVAL;
2793 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2794 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2795 __ilog2_u32(sbi->blocks_per_blkz))
2796 return -EINVAL;
2797 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2798 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2799 sbi->log_blocks_per_blkz;
2800 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2801 FDEV(devi).nr_blkz++;
2802
2803 FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2804 GFP_KERNEL);
2805 if (!FDEV(devi).blkz_type)
2806 return -ENOMEM;
2807
2808 #define F2FS_REPORT_NR_ZONES 4096
2809
2810 zones = f2fs_kzalloc(sbi,
2811 array_size(F2FS_REPORT_NR_ZONES,
2812 sizeof(struct blk_zone)),
2813 GFP_KERNEL);
2814 if (!zones)
2815 return -ENOMEM;
2816
2817 /* Get block zones type */
2818 while (zones && sector < nr_sectors) {
2819
2820 nr_zones = F2FS_REPORT_NR_ZONES;
2821 err = blkdev_report_zones(bdev, sector,
2822 zones, &nr_zones,
2823 GFP_KERNEL);
2824 if (err)
2825 break;
2826 if (!nr_zones) {
2827 err = -EIO;
2828 break;
2829 }
2830
2831 for (i = 0; i < nr_zones; i++) {
2832 FDEV(devi).blkz_type[n] = zones[i].type;
2833 sector += zones[i].len;
2834 n++;
2835 }
2836 }
2837
2838 kvfree(zones);
2839
2840 return err;
2841 }
2842 #endif
2843
2844 /*
2845 * Read f2fs raw super block.
2846 * Because we have two copies of super block, so read both of them
2847 * to get the first valid one. If any one of them is broken, we pass
2848 * them recovery flag back to the caller.
2849 */
read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)2850 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2851 struct f2fs_super_block **raw_super,
2852 int *valid_super_block, int *recovery)
2853 {
2854 struct super_block *sb = sbi->sb;
2855 int block;
2856 struct buffer_head *bh;
2857 struct f2fs_super_block *super;
2858 int err = 0;
2859
2860 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2861 if (!super)
2862 return -ENOMEM;
2863
2864 for (block = 0; block < 2; block++) {
2865 bh = sb_bread(sb, block);
2866 if (!bh) {
2867 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2868 block + 1);
2869 err = -EIO;
2870 continue;
2871 }
2872
2873 /* sanity checking of raw super */
2874 err = sanity_check_raw_super(sbi, bh);
2875 if (err) {
2876 f2fs_msg(sb, KERN_ERR,
2877 "Can't find valid F2FS filesystem in %dth superblock",
2878 block + 1);
2879 brelse(bh);
2880 continue;
2881 }
2882
2883 if (!*raw_super) {
2884 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2885 sizeof(*super));
2886 *valid_super_block = block;
2887 *raw_super = super;
2888 }
2889 brelse(bh);
2890 }
2891
2892 /* Fail to read any one of the superblocks*/
2893 if (err < 0)
2894 *recovery = 1;
2895
2896 /* No valid superblock */
2897 if (!*raw_super)
2898 kvfree(super);
2899 else
2900 err = 0;
2901
2902 return err;
2903 }
2904
f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)2905 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2906 {
2907 struct buffer_head *bh;
2908 __u32 crc = 0;
2909 int err;
2910
2911 if ((recover && f2fs_readonly(sbi->sb)) ||
2912 bdev_read_only(sbi->sb->s_bdev)) {
2913 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2914 return -EROFS;
2915 }
2916
2917 /* we should update superblock crc here */
2918 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2919 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2920 offsetof(struct f2fs_super_block, crc));
2921 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2922 }
2923
2924 /* write back-up superblock first */
2925 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2926 if (!bh)
2927 return -EIO;
2928 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2929 brelse(bh);
2930
2931 /* if we are in recovery path, skip writing valid superblock */
2932 if (recover || err)
2933 return err;
2934
2935 /* write current valid superblock */
2936 bh = sb_bread(sbi->sb, sbi->valid_super_block);
2937 if (!bh)
2938 return -EIO;
2939 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2940 brelse(bh);
2941 return err;
2942 }
2943
f2fs_scan_devices(struct f2fs_sb_info * sbi)2944 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2945 {
2946 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2947 unsigned int max_devices = MAX_DEVICES;
2948 int i;
2949
2950 /* Initialize single device information */
2951 if (!RDEV(0).path[0]) {
2952 if (!bdev_is_zoned(sbi->sb->s_bdev))
2953 return 0;
2954 max_devices = 1;
2955 }
2956
2957 /*
2958 * Initialize multiple devices information, or single
2959 * zoned block device information.
2960 */
2961 sbi->devs = f2fs_kzalloc(sbi,
2962 array_size(max_devices,
2963 sizeof(struct f2fs_dev_info)),
2964 GFP_KERNEL);
2965 if (!sbi->devs)
2966 return -ENOMEM;
2967
2968 for (i = 0; i < max_devices; i++) {
2969
2970 if (i > 0 && !RDEV(i).path[0])
2971 break;
2972
2973 if (max_devices == 1) {
2974 /* Single zoned block device mount */
2975 FDEV(0).bdev =
2976 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2977 sbi->sb->s_mode, sbi->sb->s_type);
2978 } else {
2979 /* Multi-device mount */
2980 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2981 FDEV(i).total_segments =
2982 le32_to_cpu(RDEV(i).total_segments);
2983 if (i == 0) {
2984 FDEV(i).start_blk = 0;
2985 FDEV(i).end_blk = FDEV(i).start_blk +
2986 (FDEV(i).total_segments <<
2987 sbi->log_blocks_per_seg) - 1 +
2988 le32_to_cpu(raw_super->segment0_blkaddr);
2989 } else {
2990 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2991 FDEV(i).end_blk = FDEV(i).start_blk +
2992 (FDEV(i).total_segments <<
2993 sbi->log_blocks_per_seg) - 1;
2994 }
2995 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2996 sbi->sb->s_mode, sbi->sb->s_type);
2997 }
2998 if (IS_ERR(FDEV(i).bdev))
2999 return PTR_ERR(FDEV(i).bdev);
3000
3001 /* to release errored devices */
3002 sbi->s_ndevs = i + 1;
3003
3004 #ifdef CONFIG_BLK_DEV_ZONED
3005 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3006 !f2fs_sb_has_blkzoned(sbi)) {
3007 f2fs_msg(sbi->sb, KERN_ERR,
3008 "Zoned block device feature not enabled\n");
3009 return -EINVAL;
3010 }
3011 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3012 if (init_blkz_info(sbi, i)) {
3013 f2fs_msg(sbi->sb, KERN_ERR,
3014 "Failed to initialize F2FS blkzone information");
3015 return -EINVAL;
3016 }
3017 if (max_devices == 1)
3018 break;
3019 f2fs_msg(sbi->sb, KERN_INFO,
3020 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3021 i, FDEV(i).path,
3022 FDEV(i).total_segments,
3023 FDEV(i).start_blk, FDEV(i).end_blk,
3024 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3025 "Host-aware" : "Host-managed");
3026 continue;
3027 }
3028 #endif
3029 f2fs_msg(sbi->sb, KERN_INFO,
3030 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3031 i, FDEV(i).path,
3032 FDEV(i).total_segments,
3033 FDEV(i).start_blk, FDEV(i).end_blk);
3034 }
3035 f2fs_msg(sbi->sb, KERN_INFO,
3036 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3037 return 0;
3038 }
3039
f2fs_tuning_parameters(struct f2fs_sb_info * sbi)3040 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3041 {
3042 struct f2fs_sm_info *sm_i = SM_I(sbi);
3043
3044 /* adjust parameters according to the volume size */
3045 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3046 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3047 sm_i->dcc_info->discard_granularity = 1;
3048 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3049 }
3050
3051 sbi->readdir_ra = 1;
3052 }
3053
f2fs_fill_super(struct super_block * sb,void * data,int silent)3054 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3055 {
3056 struct f2fs_sb_info *sbi;
3057 struct f2fs_super_block *raw_super;
3058 struct inode *root;
3059 int err;
3060 bool skip_recovery = false, need_fsck = false;
3061 char *options = NULL;
3062 int recovery, i, valid_super_block;
3063 struct curseg_info *seg_i;
3064 int retry_cnt = 1;
3065
3066 try_onemore:
3067 err = -EINVAL;
3068 raw_super = NULL;
3069 valid_super_block = -1;
3070 recovery = 0;
3071
3072 /* allocate memory for f2fs-specific super block info */
3073 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3074 if (!sbi)
3075 return -ENOMEM;
3076
3077 sbi->sb = sb;
3078
3079 /* Load the checksum driver */
3080 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3081 if (IS_ERR(sbi->s_chksum_driver)) {
3082 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3083 err = PTR_ERR(sbi->s_chksum_driver);
3084 sbi->s_chksum_driver = NULL;
3085 goto free_sbi;
3086 }
3087
3088 /* set a block size */
3089 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3090 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3091 goto free_sbi;
3092 }
3093
3094 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3095 &recovery);
3096 if (err)
3097 goto free_sbi;
3098
3099 sb->s_fs_info = sbi;
3100 sbi->raw_super = raw_super;
3101
3102 /* precompute checksum seed for metadata */
3103 if (f2fs_sb_has_inode_chksum(sbi))
3104 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3105 sizeof(raw_super->uuid));
3106
3107 /*
3108 * The BLKZONED feature indicates that the drive was formatted with
3109 * zone alignment optimization. This is optional for host-aware
3110 * devices, but mandatory for host-managed zoned block devices.
3111 */
3112 #ifndef CONFIG_BLK_DEV_ZONED
3113 if (f2fs_sb_has_blkzoned(sbi)) {
3114 f2fs_msg(sb, KERN_ERR,
3115 "Zoned block device support is not enabled\n");
3116 err = -EOPNOTSUPP;
3117 goto free_sb_buf;
3118 }
3119 #endif
3120 default_options(sbi);
3121 /* parse mount options */
3122 options = kstrdup((const char *)data, GFP_KERNEL);
3123 if (data && !options) {
3124 err = -ENOMEM;
3125 goto free_sb_buf;
3126 }
3127
3128 err = parse_options(sb, options);
3129 if (err)
3130 goto free_options;
3131
3132 sbi->max_file_blocks = max_file_blocks();
3133 sb->s_maxbytes = sbi->max_file_blocks <<
3134 le32_to_cpu(raw_super->log_blocksize);
3135 sb->s_max_links = F2FS_LINK_MAX;
3136
3137 #ifdef CONFIG_QUOTA
3138 sb->dq_op = &f2fs_quota_operations;
3139 if (f2fs_sb_has_quota_ino(sbi))
3140 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3141 else
3142 sb->s_qcop = &f2fs_quotactl_ops;
3143 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3144
3145 if (f2fs_sb_has_quota_ino(sbi)) {
3146 for (i = 0; i < MAXQUOTAS; i++) {
3147 if (f2fs_qf_ino(sbi->sb, i))
3148 sbi->nquota_files++;
3149 }
3150 }
3151 #endif
3152
3153 sb->s_op = &f2fs_sops;
3154 #ifdef CONFIG_F2FS_FS_ENCRYPTION
3155 sb->s_cop = &f2fs_cryptops;
3156 #endif
3157 sb->s_xattr = f2fs_xattr_handlers;
3158 sb->s_export_op = &f2fs_export_ops;
3159 sb->s_magic = F2FS_SUPER_MAGIC;
3160 sb->s_time_gran = 1;
3161 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3162 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
3163 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3164 sb->s_iflags |= SB_I_CGROUPWB;
3165
3166 /* init f2fs-specific super block info */
3167 sbi->valid_super_block = valid_super_block;
3168 mutex_init(&sbi->gc_mutex);
3169 mutex_init(&sbi->writepages);
3170 mutex_init(&sbi->cp_mutex);
3171 init_rwsem(&sbi->node_write);
3172 init_rwsem(&sbi->node_change);
3173
3174 /* disallow all the data/node/meta page writes */
3175 set_sbi_flag(sbi, SBI_POR_DOING);
3176 spin_lock_init(&sbi->stat_lock);
3177
3178 /* init iostat info */
3179 spin_lock_init(&sbi->iostat_lock);
3180 sbi->iostat_enable = false;
3181
3182 for (i = 0; i < NR_PAGE_TYPE; i++) {
3183 int n = (i == META) ? 1: NR_TEMP_TYPE;
3184 int j;
3185
3186 sbi->write_io[i] =
3187 f2fs_kmalloc(sbi,
3188 array_size(n,
3189 sizeof(struct f2fs_bio_info)),
3190 GFP_KERNEL);
3191 if (!sbi->write_io[i]) {
3192 err = -ENOMEM;
3193 goto free_bio_info;
3194 }
3195
3196 for (j = HOT; j < n; j++) {
3197 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3198 sbi->write_io[i][j].sbi = sbi;
3199 sbi->write_io[i][j].bio = NULL;
3200 spin_lock_init(&sbi->write_io[i][j].io_lock);
3201 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3202 }
3203 }
3204
3205 init_rwsem(&sbi->cp_rwsem);
3206 init_waitqueue_head(&sbi->cp_wait);
3207 init_sb_info(sbi);
3208
3209 err = init_percpu_info(sbi);
3210 if (err)
3211 goto free_bio_info;
3212
3213 if (F2FS_IO_SIZE(sbi) > 1) {
3214 sbi->write_io_dummy =
3215 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3216 if (!sbi->write_io_dummy) {
3217 err = -ENOMEM;
3218 goto free_percpu;
3219 }
3220 }
3221
3222 /* get an inode for meta space */
3223 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3224 if (IS_ERR(sbi->meta_inode)) {
3225 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3226 err = PTR_ERR(sbi->meta_inode);
3227 goto free_io_dummy;
3228 }
3229
3230 err = f2fs_get_valid_checkpoint(sbi);
3231 if (err) {
3232 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3233 goto free_meta_inode;
3234 }
3235
3236 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3237 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3238 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3239 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3240 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3241 }
3242
3243 /* Initialize device list */
3244 err = f2fs_scan_devices(sbi);
3245 if (err) {
3246 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3247 goto free_devices;
3248 }
3249
3250 sbi->total_valid_node_count =
3251 le32_to_cpu(sbi->ckpt->valid_node_count);
3252 percpu_counter_set(&sbi->total_valid_inode_count,
3253 le32_to_cpu(sbi->ckpt->valid_inode_count));
3254 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3255 sbi->total_valid_block_count =
3256 le64_to_cpu(sbi->ckpt->valid_block_count);
3257 sbi->last_valid_block_count = sbi->total_valid_block_count;
3258 sbi->reserved_blocks = 0;
3259 sbi->current_reserved_blocks = 0;
3260 limit_reserve_root(sbi);
3261
3262 for (i = 0; i < NR_INODE_TYPE; i++) {
3263 INIT_LIST_HEAD(&sbi->inode_list[i]);
3264 spin_lock_init(&sbi->inode_lock[i]);
3265 }
3266
3267 f2fs_init_extent_cache_info(sbi);
3268
3269 f2fs_init_ino_entry_info(sbi);
3270
3271 f2fs_init_fsync_node_info(sbi);
3272
3273 /* setup f2fs internal modules */
3274 err = f2fs_build_segment_manager(sbi);
3275 if (err) {
3276 f2fs_msg(sb, KERN_ERR,
3277 "Failed to initialize F2FS segment manager");
3278 goto free_sm;
3279 }
3280 err = f2fs_build_node_manager(sbi);
3281 if (err) {
3282 f2fs_msg(sb, KERN_ERR,
3283 "Failed to initialize F2FS node manager");
3284 goto free_nm;
3285 }
3286
3287 /* For write statistics */
3288 if (sb->s_bdev->bd_part)
3289 sbi->sectors_written_start =
3290 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3291
3292 /* Read accumulated write IO statistics if exists */
3293 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3294 if (__exist_node_summaries(sbi))
3295 sbi->kbytes_written =
3296 le64_to_cpu(seg_i->journal->info.kbytes_written);
3297
3298 f2fs_build_gc_manager(sbi);
3299
3300 err = f2fs_build_stats(sbi);
3301 if (err)
3302 goto free_nm;
3303
3304 /* get an inode for node space */
3305 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3306 if (IS_ERR(sbi->node_inode)) {
3307 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3308 err = PTR_ERR(sbi->node_inode);
3309 goto free_stats;
3310 }
3311
3312 /* read root inode and dentry */
3313 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3314 if (IS_ERR(root)) {
3315 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3316 err = PTR_ERR(root);
3317 goto free_node_inode;
3318 }
3319 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3320 !root->i_size || !root->i_nlink) {
3321 iput(root);
3322 err = -EINVAL;
3323 goto free_node_inode;
3324 }
3325
3326 sb->s_root = d_make_root(root); /* allocate root dentry */
3327 if (!sb->s_root) {
3328 err = -ENOMEM;
3329 goto free_node_inode;
3330 }
3331
3332 err = f2fs_register_sysfs(sbi);
3333 if (err)
3334 goto free_root_inode;
3335
3336 #ifdef CONFIG_QUOTA
3337 /* Enable quota usage during mount */
3338 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3339 err = f2fs_enable_quotas(sb);
3340 if (err)
3341 f2fs_msg(sb, KERN_ERR,
3342 "Cannot turn on quotas: error %d", err);
3343 }
3344 #endif
3345 /* if there are nt orphan nodes free them */
3346 err = f2fs_recover_orphan_inodes(sbi);
3347 if (err)
3348 goto free_meta;
3349
3350 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3351 goto reset_checkpoint;
3352
3353 /* recover fsynced data */
3354 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3355 /*
3356 * mount should be failed, when device has readonly mode, and
3357 * previous checkpoint was not done by clean system shutdown.
3358 */
3359 if (bdev_read_only(sb->s_bdev) &&
3360 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3361 err = -EROFS;
3362 goto free_meta;
3363 }
3364
3365 if (need_fsck)
3366 set_sbi_flag(sbi, SBI_NEED_FSCK);
3367
3368 if (skip_recovery)
3369 goto reset_checkpoint;
3370
3371 err = f2fs_recover_fsync_data(sbi, false);
3372 if (err < 0) {
3373 if (err != -ENOMEM)
3374 skip_recovery = true;
3375 need_fsck = true;
3376 f2fs_msg(sb, KERN_ERR,
3377 "Cannot recover all fsync data errno=%d", err);
3378 goto free_meta;
3379 }
3380 } else {
3381 err = f2fs_recover_fsync_data(sbi, true);
3382
3383 if (!f2fs_readonly(sb) && err > 0) {
3384 err = -EINVAL;
3385 f2fs_msg(sb, KERN_ERR,
3386 "Need to recover fsync data");
3387 goto free_meta;
3388 }
3389 }
3390 reset_checkpoint:
3391 /* f2fs_recover_fsync_data() cleared this already */
3392 clear_sbi_flag(sbi, SBI_POR_DOING);
3393
3394 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3395 err = f2fs_disable_checkpoint(sbi);
3396 if (err)
3397 goto sync_free_meta;
3398 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3399 f2fs_enable_checkpoint(sbi);
3400 }
3401
3402 /*
3403 * If filesystem is not mounted as read-only then
3404 * do start the gc_thread.
3405 */
3406 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3407 /* After POR, we can run background GC thread.*/
3408 err = f2fs_start_gc_thread(sbi);
3409 if (err)
3410 goto sync_free_meta;
3411 }
3412 kvfree(options);
3413
3414 /* recover broken superblock */
3415 if (recovery) {
3416 err = f2fs_commit_super(sbi, true);
3417 f2fs_msg(sb, KERN_INFO,
3418 "Try to recover %dth superblock, ret: %d",
3419 sbi->valid_super_block ? 1 : 2, err);
3420 }
3421
3422 f2fs_join_shrinker(sbi);
3423
3424 f2fs_tuning_parameters(sbi);
3425
3426 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3427 cur_cp_version(F2FS_CKPT(sbi)));
3428 f2fs_update_time(sbi, CP_TIME);
3429 f2fs_update_time(sbi, REQ_TIME);
3430 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3431 return 0;
3432
3433 sync_free_meta:
3434 /* safe to flush all the data */
3435 sync_filesystem(sbi->sb);
3436 retry_cnt = 0;
3437
3438 free_meta:
3439 #ifdef CONFIG_QUOTA
3440 f2fs_truncate_quota_inode_pages(sb);
3441 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3442 f2fs_quota_off_umount(sbi->sb);
3443 #endif
3444 /*
3445 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3446 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3447 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3448 * falls into an infinite loop in f2fs_sync_meta_pages().
3449 */
3450 truncate_inode_pages_final(META_MAPPING(sbi));
3451 /* evict some inodes being cached by GC */
3452 evict_inodes(sb);
3453 f2fs_unregister_sysfs(sbi);
3454 free_root_inode:
3455 dput(sb->s_root);
3456 sb->s_root = NULL;
3457 free_node_inode:
3458 f2fs_release_ino_entry(sbi, true);
3459 truncate_inode_pages_final(NODE_MAPPING(sbi));
3460 iput(sbi->node_inode);
3461 sbi->node_inode = NULL;
3462 free_stats:
3463 f2fs_destroy_stats(sbi);
3464 free_nm:
3465 f2fs_destroy_node_manager(sbi);
3466 free_sm:
3467 f2fs_destroy_segment_manager(sbi);
3468 free_devices:
3469 destroy_device_list(sbi);
3470 kvfree(sbi->ckpt);
3471 free_meta_inode:
3472 make_bad_inode(sbi->meta_inode);
3473 iput(sbi->meta_inode);
3474 sbi->meta_inode = NULL;
3475 free_io_dummy:
3476 mempool_destroy(sbi->write_io_dummy);
3477 free_percpu:
3478 destroy_percpu_info(sbi);
3479 free_bio_info:
3480 for (i = 0; i < NR_PAGE_TYPE; i++)
3481 kvfree(sbi->write_io[i]);
3482 free_options:
3483 #ifdef CONFIG_QUOTA
3484 for (i = 0; i < MAXQUOTAS; i++)
3485 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3486 #endif
3487 kvfree(options);
3488 free_sb_buf:
3489 kvfree(raw_super);
3490 free_sbi:
3491 if (sbi->s_chksum_driver)
3492 crypto_free_shash(sbi->s_chksum_driver);
3493 kvfree(sbi);
3494
3495 /* give only one another chance */
3496 if (retry_cnt > 0 && skip_recovery) {
3497 retry_cnt--;
3498 shrink_dcache_sb(sb);
3499 goto try_onemore;
3500 }
3501 return err;
3502 }
3503
f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)3504 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3505 const char *dev_name, void *data)
3506 {
3507 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3508 }
3509
kill_f2fs_super(struct super_block * sb)3510 static void kill_f2fs_super(struct super_block *sb)
3511 {
3512 if (sb->s_root) {
3513 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3514
3515 set_sbi_flag(sbi, SBI_IS_CLOSE);
3516 f2fs_stop_gc_thread(sbi);
3517 f2fs_stop_discard_thread(sbi);
3518
3519 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3520 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3521 struct cp_control cpc = {
3522 .reason = CP_UMOUNT,
3523 };
3524 f2fs_write_checkpoint(sbi, &cpc);
3525 }
3526
3527 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3528 sb->s_flags &= ~SB_RDONLY;
3529 }
3530 kill_block_super(sb);
3531 }
3532
3533 static struct file_system_type f2fs_fs_type = {
3534 .owner = THIS_MODULE,
3535 .name = "f2fs",
3536 .mount = f2fs_mount,
3537 .kill_sb = kill_f2fs_super,
3538 .fs_flags = FS_REQUIRES_DEV,
3539 };
3540 MODULE_ALIAS_FS("f2fs");
3541
init_inodecache(void)3542 static int __init init_inodecache(void)
3543 {
3544 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3545 sizeof(struct f2fs_inode_info), 0,
3546 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3547 if (!f2fs_inode_cachep)
3548 return -ENOMEM;
3549 return 0;
3550 }
3551
destroy_inodecache(void)3552 static void destroy_inodecache(void)
3553 {
3554 /*
3555 * Make sure all delayed rcu free inodes are flushed before we
3556 * destroy cache.
3557 */
3558 rcu_barrier();
3559 kmem_cache_destroy(f2fs_inode_cachep);
3560 }
3561
init_f2fs_fs(void)3562 static int __init init_f2fs_fs(void)
3563 {
3564 int err;
3565
3566 if (PAGE_SIZE != F2FS_BLKSIZE) {
3567 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3568 PAGE_SIZE, F2FS_BLKSIZE);
3569 return -EINVAL;
3570 }
3571
3572 f2fs_build_trace_ios();
3573
3574 err = init_inodecache();
3575 if (err)
3576 goto fail;
3577 err = f2fs_create_node_manager_caches();
3578 if (err)
3579 goto free_inodecache;
3580 err = f2fs_create_segment_manager_caches();
3581 if (err)
3582 goto free_node_manager_caches;
3583 err = f2fs_create_checkpoint_caches();
3584 if (err)
3585 goto free_segment_manager_caches;
3586 err = f2fs_create_extent_cache();
3587 if (err)
3588 goto free_checkpoint_caches;
3589 err = f2fs_init_sysfs();
3590 if (err)
3591 goto free_extent_cache;
3592 err = register_shrinker(&f2fs_shrinker_info);
3593 if (err)
3594 goto free_sysfs;
3595 err = register_filesystem(&f2fs_fs_type);
3596 if (err)
3597 goto free_shrinker;
3598 f2fs_create_root_stats();
3599 err = f2fs_init_post_read_processing();
3600 if (err)
3601 goto free_root_stats;
3602 return 0;
3603
3604 free_root_stats:
3605 f2fs_destroy_root_stats();
3606 unregister_filesystem(&f2fs_fs_type);
3607 free_shrinker:
3608 unregister_shrinker(&f2fs_shrinker_info);
3609 free_sysfs:
3610 f2fs_exit_sysfs();
3611 free_extent_cache:
3612 f2fs_destroy_extent_cache();
3613 free_checkpoint_caches:
3614 f2fs_destroy_checkpoint_caches();
3615 free_segment_manager_caches:
3616 f2fs_destroy_segment_manager_caches();
3617 free_node_manager_caches:
3618 f2fs_destroy_node_manager_caches();
3619 free_inodecache:
3620 destroy_inodecache();
3621 fail:
3622 return err;
3623 }
3624
exit_f2fs_fs(void)3625 static void __exit exit_f2fs_fs(void)
3626 {
3627 f2fs_destroy_post_read_processing();
3628 f2fs_destroy_root_stats();
3629 unregister_filesystem(&f2fs_fs_type);
3630 unregister_shrinker(&f2fs_shrinker_info);
3631 f2fs_exit_sysfs();
3632 f2fs_destroy_extent_cache();
3633 f2fs_destroy_checkpoint_caches();
3634 f2fs_destroy_segment_manager_caches();
3635 f2fs_destroy_node_manager_caches();
3636 destroy_inodecache();
3637 f2fs_destroy_trace_ios();
3638 }
3639
3640 module_init(init_f2fs_fs)
3641 module_exit(exit_f2fs_fs)
3642
3643 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3644 MODULE_DESCRIPTION("Flash Friendly File System");
3645 MODULE_LICENSE("GPL");
3646
3647