f2aaa2cc6b3e01e88a7aeea026cfaa97f73ae1de
[muen/linux.git] / fs / f2fs / super.c
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
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
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
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
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])
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
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
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
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 |= SB_LAZYTIME;
624                         break;
625                 case Opt_nolazytime:
626                         sb->s_flags &= ~SB_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_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
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
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
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
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  */
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
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
1009 static void f2fs_destroy_inode(struct inode *inode)
1010 {
1011         call_rcu(&inode->i_rcu, f2fs_i_callback);
1012 }
1013
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
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
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
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
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
1165 static int f2fs_unfreeze(struct super_block *sb)
1166 {
1167         return 0;
1168 }
1169
1170 #ifdef CONFIG_QUOTA
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 = (dquot->dq_dqb.dqb_bsoftlimit ?
1186                  dquot->dq_dqb.dqb_bsoftlimit :
1187                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1188         if (limit && buf->f_blocks > limit) {
1189                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1190                 buf->f_blocks = limit;
1191                 buf->f_bfree = buf->f_bavail =
1192                         (buf->f_blocks > curblock) ?
1193                          (buf->f_blocks - curblock) : 0;
1194         }
1195
1196         limit = dquot->dq_dqb.dqb_isoftlimit ?
1197                 dquot->dq_dqb.dqb_isoftlimit :
1198                 dquot->dq_dqb.dqb_ihardlimit;
1199         if (limit && buf->f_files > limit) {
1200                 buf->f_files = limit;
1201                 buf->f_ffree =
1202                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1203                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1204         }
1205
1206         spin_unlock(&dquot->dq_dqb_lock);
1207         dqput(dquot);
1208         return 0;
1209 }
1210 #endif
1211
1212 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1213 {
1214         struct super_block *sb = dentry->d_sb;
1215         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1216         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1217         block_t total_count, user_block_count, start_count;
1218         u64 avail_node_count;
1219
1220         total_count = le64_to_cpu(sbi->raw_super->block_count);
1221         user_block_count = sbi->user_block_count;
1222         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1223         buf->f_type = F2FS_SUPER_MAGIC;
1224         buf->f_bsize = sbi->blocksize;
1225
1226         buf->f_blocks = total_count - start_count;
1227         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1228                                                 sbi->current_reserved_blocks;
1229         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1230                 buf->f_bfree = 0;
1231         else
1232                 buf->f_bfree -= sbi->unusable_block_count;
1233
1234         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1235                 buf->f_bavail = buf->f_bfree -
1236                                 F2FS_OPTION(sbi).root_reserved_blocks;
1237         else
1238                 buf->f_bavail = 0;
1239
1240         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1241                                                 F2FS_RESERVED_NODE_NUM;
1242
1243         if (avail_node_count > user_block_count) {
1244                 buf->f_files = user_block_count;
1245                 buf->f_ffree = buf->f_bavail;
1246         } else {
1247                 buf->f_files = avail_node_count;
1248                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1249                                         buf->f_bavail);
1250         }
1251
1252         buf->f_namelen = F2FS_NAME_LEN;
1253         buf->f_fsid.val[0] = (u32)id;
1254         buf->f_fsid.val[1] = (u32)(id >> 32);
1255
1256 #ifdef CONFIG_QUOTA
1257         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1258                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1259                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1260         }
1261 #endif
1262         return 0;
1263 }
1264
1265 static inline void f2fs_show_quota_options(struct seq_file *seq,
1266                                            struct super_block *sb)
1267 {
1268 #ifdef CONFIG_QUOTA
1269         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1270
1271         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1272                 char *fmtname = "";
1273
1274                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1275                 case QFMT_VFS_OLD:
1276                         fmtname = "vfsold";
1277                         break;
1278                 case QFMT_VFS_V0:
1279                         fmtname = "vfsv0";
1280                         break;
1281                 case QFMT_VFS_V1:
1282                         fmtname = "vfsv1";
1283                         break;
1284                 }
1285                 seq_printf(seq, ",jqfmt=%s", fmtname);
1286         }
1287
1288         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1289                 seq_show_option(seq, "usrjquota",
1290                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1291
1292         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1293                 seq_show_option(seq, "grpjquota",
1294                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1295
1296         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1297                 seq_show_option(seq, "prjjquota",
1298                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1299 #endif
1300 }
1301
1302 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1303 {
1304         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1305
1306         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1307                 if (test_opt(sbi, FORCE_FG_GC))
1308                         seq_printf(seq, ",background_gc=%s", "sync");
1309                 else
1310                         seq_printf(seq, ",background_gc=%s", "on");
1311         } else {
1312                 seq_printf(seq, ",background_gc=%s", "off");
1313         }
1314         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1315                 seq_puts(seq, ",disable_roll_forward");
1316         if (test_opt(sbi, DISCARD))
1317                 seq_puts(seq, ",discard");
1318         if (test_opt(sbi, NOHEAP))
1319                 seq_puts(seq, ",no_heap");
1320         else
1321                 seq_puts(seq, ",heap");
1322 #ifdef CONFIG_F2FS_FS_XATTR
1323         if (test_opt(sbi, XATTR_USER))
1324                 seq_puts(seq, ",user_xattr");
1325         else
1326                 seq_puts(seq, ",nouser_xattr");
1327         if (test_opt(sbi, INLINE_XATTR))
1328                 seq_puts(seq, ",inline_xattr");
1329         else
1330                 seq_puts(seq, ",noinline_xattr");
1331         if (test_opt(sbi, INLINE_XATTR_SIZE))
1332                 seq_printf(seq, ",inline_xattr_size=%u",
1333                                         F2FS_OPTION(sbi).inline_xattr_size);
1334 #endif
1335 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1336         if (test_opt(sbi, POSIX_ACL))
1337                 seq_puts(seq, ",acl");
1338         else
1339                 seq_puts(seq, ",noacl");
1340 #endif
1341         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1342                 seq_puts(seq, ",disable_ext_identify");
1343         if (test_opt(sbi, INLINE_DATA))
1344                 seq_puts(seq, ",inline_data");
1345         else
1346                 seq_puts(seq, ",noinline_data");
1347         if (test_opt(sbi, INLINE_DENTRY))
1348                 seq_puts(seq, ",inline_dentry");
1349         else
1350                 seq_puts(seq, ",noinline_dentry");
1351         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1352                 seq_puts(seq, ",flush_merge");
1353         if (test_opt(sbi, NOBARRIER))
1354                 seq_puts(seq, ",nobarrier");
1355         if (test_opt(sbi, FASTBOOT))
1356                 seq_puts(seq, ",fastboot");
1357         if (test_opt(sbi, EXTENT_CACHE))
1358                 seq_puts(seq, ",extent_cache");
1359         else
1360                 seq_puts(seq, ",noextent_cache");
1361         if (test_opt(sbi, DATA_FLUSH))
1362                 seq_puts(seq, ",data_flush");
1363
1364         seq_puts(seq, ",mode=");
1365         if (test_opt(sbi, ADAPTIVE))
1366                 seq_puts(seq, "adaptive");
1367         else if (test_opt(sbi, LFS))
1368                 seq_puts(seq, "lfs");
1369         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1370         if (test_opt(sbi, RESERVE_ROOT))
1371                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1372                                 F2FS_OPTION(sbi).root_reserved_blocks,
1373                                 from_kuid_munged(&init_user_ns,
1374                                         F2FS_OPTION(sbi).s_resuid),
1375                                 from_kgid_munged(&init_user_ns,
1376                                         F2FS_OPTION(sbi).s_resgid));
1377         if (F2FS_IO_SIZE_BITS(sbi))
1378                 seq_printf(seq, ",io_bits=%u",
1379                                 F2FS_OPTION(sbi).write_io_size_bits);
1380 #ifdef CONFIG_F2FS_FAULT_INJECTION
1381         if (test_opt(sbi, FAULT_INJECTION)) {
1382                 seq_printf(seq, ",fault_injection=%u",
1383                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1384                 seq_printf(seq, ",fault_type=%u",
1385                                 F2FS_OPTION(sbi).fault_info.inject_type);
1386         }
1387 #endif
1388 #ifdef CONFIG_QUOTA
1389         if (test_opt(sbi, QUOTA))
1390                 seq_puts(seq, ",quota");
1391         if (test_opt(sbi, USRQUOTA))
1392                 seq_puts(seq, ",usrquota");
1393         if (test_opt(sbi, GRPQUOTA))
1394                 seq_puts(seq, ",grpquota");
1395         if (test_opt(sbi, PRJQUOTA))
1396                 seq_puts(seq, ",prjquota");
1397 #endif
1398         f2fs_show_quota_options(seq, sbi->sb);
1399         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1400                 seq_printf(seq, ",whint_mode=%s", "user-based");
1401         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1402                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1403 #ifdef CONFIG_FS_ENCRYPTION
1404         if (F2FS_OPTION(sbi).test_dummy_encryption)
1405                 seq_puts(seq, ",test_dummy_encryption");
1406 #endif
1407
1408         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1409                 seq_printf(seq, ",alloc_mode=%s", "default");
1410         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1411                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1412
1413         if (test_opt(sbi, DISABLE_CHECKPOINT))
1414                 seq_puts(seq, ",checkpoint=disable");
1415
1416         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1417                 seq_printf(seq, ",fsync_mode=%s", "posix");
1418         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1419                 seq_printf(seq, ",fsync_mode=%s", "strict");
1420         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1421                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1422         return 0;
1423 }
1424
1425 static void default_options(struct f2fs_sb_info *sbi)
1426 {
1427         /* init some FS parameters */
1428         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1429         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1430         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1431         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1432         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1433         F2FS_OPTION(sbi).test_dummy_encryption = false;
1434         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1435         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1436
1437         set_opt(sbi, BG_GC);
1438         set_opt(sbi, INLINE_XATTR);
1439         set_opt(sbi, INLINE_DATA);
1440         set_opt(sbi, INLINE_DENTRY);
1441         set_opt(sbi, EXTENT_CACHE);
1442         set_opt(sbi, NOHEAP);
1443         clear_opt(sbi, DISABLE_CHECKPOINT);
1444         sbi->sb->s_flags |= SB_LAZYTIME;
1445         set_opt(sbi, FLUSH_MERGE);
1446         set_opt(sbi, DISCARD);
1447         if (f2fs_sb_has_blkzoned(sbi))
1448                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1449         else
1450                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1451
1452 #ifdef CONFIG_F2FS_FS_XATTR
1453         set_opt(sbi, XATTR_USER);
1454 #endif
1455 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1456         set_opt(sbi, POSIX_ACL);
1457 #endif
1458
1459         f2fs_build_fault_attr(sbi, 0, 0);
1460 }
1461
1462 #ifdef CONFIG_QUOTA
1463 static int f2fs_enable_quotas(struct super_block *sb);
1464 #endif
1465
1466 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1467 {
1468         unsigned int s_flags = sbi->sb->s_flags;
1469         struct cp_control cpc;
1470         int err = 0;
1471         int ret;
1472
1473         if (s_flags & SB_RDONLY) {
1474                 f2fs_msg(sbi->sb, KERN_ERR,
1475                                 "checkpoint=disable on readonly fs");
1476                 return -EINVAL;
1477         }
1478         sbi->sb->s_flags |= SB_ACTIVE;
1479
1480         f2fs_update_time(sbi, DISABLE_TIME);
1481
1482         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1483                 mutex_lock(&sbi->gc_mutex);
1484                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1485                 if (err == -ENODATA) {
1486                         err = 0;
1487                         break;
1488                 }
1489                 if (err && err != -EAGAIN)
1490                         break;
1491         }
1492
1493         ret = sync_filesystem(sbi->sb);
1494         if (ret || err) {
1495                 err = ret ? ret: err;
1496                 goto restore_flag;
1497         }
1498
1499         if (f2fs_disable_cp_again(sbi)) {
1500                 err = -EAGAIN;
1501                 goto restore_flag;
1502         }
1503
1504         mutex_lock(&sbi->gc_mutex);
1505         cpc.reason = CP_PAUSE;
1506         set_sbi_flag(sbi, SBI_CP_DISABLED);
1507         f2fs_write_checkpoint(sbi, &cpc);
1508
1509         sbi->unusable_block_count = 0;
1510         mutex_unlock(&sbi->gc_mutex);
1511 restore_flag:
1512         sbi->sb->s_flags = s_flags;     /* Restore MS_RDONLY status */
1513         return err;
1514 }
1515
1516 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1517 {
1518         mutex_lock(&sbi->gc_mutex);
1519         f2fs_dirty_to_prefree(sbi);
1520
1521         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1522         set_sbi_flag(sbi, SBI_IS_DIRTY);
1523         mutex_unlock(&sbi->gc_mutex);
1524
1525         f2fs_sync_fs(sbi->sb, 1);
1526 }
1527
1528 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1529 {
1530         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1531         struct f2fs_mount_info org_mount_opt;
1532         unsigned long old_sb_flags;
1533         int err;
1534         bool need_restart_gc = false;
1535         bool need_stop_gc = false;
1536         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1537         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1538         bool checkpoint_changed;
1539 #ifdef CONFIG_QUOTA
1540         int i, j;
1541 #endif
1542
1543         /*
1544          * Save the old mount options in case we
1545          * need to restore them.
1546          */
1547         org_mount_opt = sbi->mount_opt;
1548         old_sb_flags = sb->s_flags;
1549
1550 #ifdef CONFIG_QUOTA
1551         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1552         for (i = 0; i < MAXQUOTAS; i++) {
1553                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1554                         org_mount_opt.s_qf_names[i] =
1555                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1556                                 GFP_KERNEL);
1557                         if (!org_mount_opt.s_qf_names[i]) {
1558                                 for (j = 0; j < i; j++)
1559                                         kvfree(org_mount_opt.s_qf_names[j]);
1560                                 return -ENOMEM;
1561                         }
1562                 } else {
1563                         org_mount_opt.s_qf_names[i] = NULL;
1564                 }
1565         }
1566 #endif
1567
1568         /* recover superblocks we couldn't write due to previous RO mount */
1569         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1570                 err = f2fs_commit_super(sbi, false);
1571                 f2fs_msg(sb, KERN_INFO,
1572                         "Try to recover all the superblocks, ret: %d", err);
1573                 if (!err)
1574                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1575         }
1576
1577         default_options(sbi);
1578
1579         /* parse mount options */
1580         err = parse_options(sb, data);
1581         if (err)
1582                 goto restore_opts;
1583         checkpoint_changed =
1584                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1585
1586         /*
1587          * Previous and new state of filesystem is RO,
1588          * so skip checking GC and FLUSH_MERGE conditions.
1589          */
1590         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1591                 goto skip;
1592
1593 #ifdef CONFIG_QUOTA
1594         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1595                 err = dquot_suspend(sb, -1);
1596                 if (err < 0)
1597                         goto restore_opts;
1598         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1599                 /* dquot_resume needs RW */
1600                 sb->s_flags &= ~SB_RDONLY;
1601                 if (sb_any_quota_suspended(sb)) {
1602                         dquot_resume(sb, -1);
1603                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1604                         err = f2fs_enable_quotas(sb);
1605                         if (err)
1606                                 goto restore_opts;
1607                 }
1608         }
1609 #endif
1610         /* disallow enable/disable extent_cache dynamically */
1611         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1612                 err = -EINVAL;
1613                 f2fs_msg(sbi->sb, KERN_WARNING,
1614                                 "switch extent_cache option is not allowed");
1615                 goto restore_opts;
1616         }
1617
1618         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1619                 err = -EINVAL;
1620                 f2fs_msg(sbi->sb, KERN_WARNING,
1621                         "disabling checkpoint not compatible with read-only");
1622                 goto restore_opts;
1623         }
1624
1625         /*
1626          * We stop the GC thread if FS is mounted as RO
1627          * or if background_gc = off is passed in mount
1628          * option. Also sync the filesystem.
1629          */
1630         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1631                 if (sbi->gc_thread) {
1632                         f2fs_stop_gc_thread(sbi);
1633                         need_restart_gc = true;
1634                 }
1635         } else if (!sbi->gc_thread) {
1636                 err = f2fs_start_gc_thread(sbi);
1637                 if (err)
1638                         goto restore_opts;
1639                 need_stop_gc = true;
1640         }
1641
1642         if (*flags & SB_RDONLY ||
1643                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1644                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1645                 sync_inodes_sb(sb);
1646
1647                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1648                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1649                 f2fs_sync_fs(sb, 1);
1650                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1651         }
1652
1653         if (checkpoint_changed) {
1654                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1655                         err = f2fs_disable_checkpoint(sbi);
1656                         if (err)
1657                                 goto restore_gc;
1658                 } else {
1659                         f2fs_enable_checkpoint(sbi);
1660                 }
1661         }
1662
1663         /*
1664          * We stop issue flush thread if FS is mounted as RO
1665          * or if flush_merge is not passed in mount option.
1666          */
1667         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1668                 clear_opt(sbi, FLUSH_MERGE);
1669                 f2fs_destroy_flush_cmd_control(sbi, false);
1670         } else {
1671                 err = f2fs_create_flush_cmd_control(sbi);
1672                 if (err)
1673                         goto restore_gc;
1674         }
1675 skip:
1676 #ifdef CONFIG_QUOTA
1677         /* Release old quota file names */
1678         for (i = 0; i < MAXQUOTAS; i++)
1679                 kvfree(org_mount_opt.s_qf_names[i]);
1680 #endif
1681         /* Update the POSIXACL Flag */
1682         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1683                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1684
1685         limit_reserve_root(sbi);
1686         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1687         return 0;
1688 restore_gc:
1689         if (need_restart_gc) {
1690                 if (f2fs_start_gc_thread(sbi))
1691                         f2fs_msg(sbi->sb, KERN_WARNING,
1692                                 "background gc thread has stopped");
1693         } else if (need_stop_gc) {
1694                 f2fs_stop_gc_thread(sbi);
1695         }
1696 restore_opts:
1697 #ifdef CONFIG_QUOTA
1698         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1699         for (i = 0; i < MAXQUOTAS; i++) {
1700                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1701                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1702         }
1703 #endif
1704         sbi->mount_opt = org_mount_opt;
1705         sb->s_flags = old_sb_flags;
1706         return err;
1707 }
1708
1709 #ifdef CONFIG_QUOTA
1710 /* Read data from quotafile */
1711 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1712                                size_t len, loff_t off)
1713 {
1714         struct inode *inode = sb_dqopt(sb)->files[type];
1715         struct address_space *mapping = inode->i_mapping;
1716         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1717         int offset = off & (sb->s_blocksize - 1);
1718         int tocopy;
1719         size_t toread;
1720         loff_t i_size = i_size_read(inode);
1721         struct page *page;
1722         char *kaddr;
1723
1724         if (off > i_size)
1725                 return 0;
1726
1727         if (off + len > i_size)
1728                 len = i_size - off;
1729         toread = len;
1730         while (toread > 0) {
1731                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1732 repeat:
1733                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1734                 if (IS_ERR(page)) {
1735                         if (PTR_ERR(page) == -ENOMEM) {
1736                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1737                                 goto repeat;
1738                         }
1739                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1740                         return PTR_ERR(page);
1741                 }
1742
1743                 lock_page(page);
1744
1745                 if (unlikely(page->mapping != mapping)) {
1746                         f2fs_put_page(page, 1);
1747                         goto repeat;
1748                 }
1749                 if (unlikely(!PageUptodate(page))) {
1750                         f2fs_put_page(page, 1);
1751                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1752                         return -EIO;
1753                 }
1754
1755                 kaddr = kmap_atomic(page);
1756                 memcpy(data, kaddr + offset, tocopy);
1757                 kunmap_atomic(kaddr);
1758                 f2fs_put_page(page, 1);
1759
1760                 offset = 0;
1761                 toread -= tocopy;
1762                 data += tocopy;
1763                 blkidx++;
1764         }
1765         return len;
1766 }
1767
1768 /* Write to quotafile */
1769 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1770                                 const char *data, size_t len, loff_t off)
1771 {
1772         struct inode *inode = sb_dqopt(sb)->files[type];
1773         struct address_space *mapping = inode->i_mapping;
1774         const struct address_space_operations *a_ops = mapping->a_ops;
1775         int offset = off & (sb->s_blocksize - 1);
1776         size_t towrite = len;
1777         struct page *page;
1778         char *kaddr;
1779         int err = 0;
1780         int tocopy;
1781
1782         while (towrite > 0) {
1783                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1784                                                                 towrite);
1785 retry:
1786                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1787                                                         &page, NULL);
1788                 if (unlikely(err)) {
1789                         if (err == -ENOMEM) {
1790                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1791                                 goto retry;
1792                         }
1793                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1794                         break;
1795                 }
1796
1797                 kaddr = kmap_atomic(page);
1798                 memcpy(kaddr + offset, data, tocopy);
1799                 kunmap_atomic(kaddr);
1800                 flush_dcache_page(page);
1801
1802                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1803                                                 page, NULL);
1804                 offset = 0;
1805                 towrite -= tocopy;
1806                 off += tocopy;
1807                 data += tocopy;
1808                 cond_resched();
1809         }
1810
1811         if (len == towrite)
1812                 return err;
1813         inode->i_mtime = inode->i_ctime = current_time(inode);
1814         f2fs_mark_inode_dirty_sync(inode, false);
1815         return len - towrite;
1816 }
1817
1818 static struct dquot **f2fs_get_dquots(struct inode *inode)
1819 {
1820         return F2FS_I(inode)->i_dquot;
1821 }
1822
1823 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1824 {
1825         return &F2FS_I(inode)->i_reserved_quota;
1826 }
1827
1828 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1829 {
1830         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1831                 f2fs_msg(sbi->sb, KERN_ERR,
1832                         "quota sysfile may be corrupted, skip loading it");
1833                 return 0;
1834         }
1835
1836         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1837                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1838 }
1839
1840 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1841 {
1842         int enabled = 0;
1843         int i, err;
1844
1845         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1846                 err = f2fs_enable_quotas(sbi->sb);
1847                 if (err) {
1848                         f2fs_msg(sbi->sb, KERN_ERR,
1849                                         "Cannot turn on quota_ino: %d", err);
1850                         return 0;
1851                 }
1852                 return 1;
1853         }
1854
1855         for (i = 0; i < MAXQUOTAS; i++) {
1856                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1857                         err = f2fs_quota_on_mount(sbi, i);
1858                         if (!err) {
1859                                 enabled = 1;
1860                                 continue;
1861                         }
1862                         f2fs_msg(sbi->sb, KERN_ERR,
1863                                 "Cannot turn on quotas: %d on %d", err, i);
1864                 }
1865         }
1866         return enabled;
1867 }
1868
1869 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1870                              unsigned int flags)
1871 {
1872         struct inode *qf_inode;
1873         unsigned long qf_inum;
1874         int err;
1875
1876         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1877
1878         qf_inum = f2fs_qf_ino(sb, type);
1879         if (!qf_inum)
1880                 return -EPERM;
1881
1882         qf_inode = f2fs_iget(sb, qf_inum);
1883         if (IS_ERR(qf_inode)) {
1884                 f2fs_msg(sb, KERN_ERR,
1885                         "Bad quota inode %u:%lu", type, qf_inum);
1886                 return PTR_ERR(qf_inode);
1887         }
1888
1889         /* Don't account quota for quota files to avoid recursion */
1890         qf_inode->i_flags |= S_NOQUOTA;
1891         err = dquot_enable(qf_inode, type, format_id, flags);
1892         iput(qf_inode);
1893         return err;
1894 }
1895
1896 static int f2fs_enable_quotas(struct super_block *sb)
1897 {
1898         int type, err = 0;
1899         unsigned long qf_inum;
1900         bool quota_mopt[MAXQUOTAS] = {
1901                 test_opt(F2FS_SB(sb), USRQUOTA),
1902                 test_opt(F2FS_SB(sb), GRPQUOTA),
1903                 test_opt(F2FS_SB(sb), PRJQUOTA),
1904         };
1905
1906         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1907                 f2fs_msg(sb, KERN_ERR,
1908                         "quota file may be corrupted, skip loading it");
1909                 return 0;
1910         }
1911
1912         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1913
1914         for (type = 0; type < MAXQUOTAS; type++) {
1915                 qf_inum = f2fs_qf_ino(sb, type);
1916                 if (qf_inum) {
1917                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1918                                 DQUOT_USAGE_ENABLED |
1919                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1920                         if (err) {
1921                                 f2fs_msg(sb, KERN_ERR,
1922                                         "Failed to enable quota tracking "
1923                                         "(type=%d, err=%d). Please run "
1924                                         "fsck to fix.", type, err);
1925                                 for (type--; type >= 0; type--)
1926                                         dquot_quota_off(sb, type);
1927                                 set_sbi_flag(F2FS_SB(sb),
1928                                                 SBI_QUOTA_NEED_REPAIR);
1929                                 return err;
1930                         }
1931                 }
1932         }
1933         return 0;
1934 }
1935
1936 int f2fs_quota_sync(struct super_block *sb, int type)
1937 {
1938         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1939         struct quota_info *dqopt = sb_dqopt(sb);
1940         int cnt;
1941         int ret;
1942
1943         ret = dquot_writeback_dquots(sb, type);
1944         if (ret)
1945                 goto out;
1946
1947         /*
1948          * Now when everything is written we can discard the pagecache so
1949          * that userspace sees the changes.
1950          */
1951         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1952                 struct address_space *mapping;
1953
1954                 if (type != -1 && cnt != type)
1955                         continue;
1956                 if (!sb_has_quota_active(sb, cnt))
1957                         continue;
1958
1959                 mapping = dqopt->files[cnt]->i_mapping;
1960
1961                 ret = filemap_fdatawrite(mapping);
1962                 if (ret)
1963                         goto out;
1964
1965                 /* if we are using journalled quota */
1966                 if (is_journalled_quota(sbi))
1967                         continue;
1968
1969                 ret = filemap_fdatawait(mapping);
1970                 if (ret)
1971                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1972
1973                 inode_lock(dqopt->files[cnt]);
1974                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1975                 inode_unlock(dqopt->files[cnt]);
1976         }
1977 out:
1978         if (ret)
1979                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1980         return ret;
1981 }
1982
1983 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1984                                                         const struct path *path)
1985 {
1986         struct inode *inode;
1987         int err;
1988
1989         err = f2fs_quota_sync(sb, type);
1990         if (err)
1991                 return err;
1992
1993         err = dquot_quota_on(sb, type, format_id, path);
1994         if (err)
1995                 return err;
1996
1997         inode = d_inode(path->dentry);
1998
1999         inode_lock(inode);
2000         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2001         f2fs_set_inode_flags(inode);
2002         inode_unlock(inode);
2003         f2fs_mark_inode_dirty_sync(inode, false);
2004
2005         return 0;
2006 }
2007
2008 static int f2fs_quota_off(struct super_block *sb, int type)
2009 {
2010         struct inode *inode = sb_dqopt(sb)->files[type];
2011         int err;
2012
2013         if (!inode || !igrab(inode))
2014                 return dquot_quota_off(sb, type);
2015
2016         err = f2fs_quota_sync(sb, type);
2017         if (err)
2018                 goto out_put;
2019
2020         err = dquot_quota_off(sb, type);
2021         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2022                 goto out_put;
2023
2024         inode_lock(inode);
2025         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2026         f2fs_set_inode_flags(inode);
2027         inode_unlock(inode);
2028         f2fs_mark_inode_dirty_sync(inode, false);
2029 out_put:
2030         iput(inode);
2031         return err;
2032 }
2033
2034 void f2fs_quota_off_umount(struct super_block *sb)
2035 {
2036         int type;
2037         int err;
2038
2039         for (type = 0; type < MAXQUOTAS; type++) {
2040                 err = f2fs_quota_off(sb, type);
2041                 if (err) {
2042                         int ret = dquot_quota_off(sb, type);
2043
2044                         f2fs_msg(sb, KERN_ERR,
2045                                 "Fail to turn off disk quota "
2046                                 "(type: %d, err: %d, ret:%d), Please "
2047                                 "run fsck to fix it.", type, err, ret);
2048                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2049                 }
2050         }
2051         /*
2052          * In case of checkpoint=disable, we must flush quota blocks.
2053          * This can cause NULL exception for node_inode in end_io, since
2054          * put_super already dropped it.
2055          */
2056         sync_filesystem(sb);
2057 }
2058
2059 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2060 {
2061         struct quota_info *dqopt = sb_dqopt(sb);
2062         int type;
2063
2064         for (type = 0; type < MAXQUOTAS; type++) {
2065                 if (!dqopt->files[type])
2066                         continue;
2067                 f2fs_inode_synced(dqopt->files[type]);
2068         }
2069 }
2070
2071 static int f2fs_dquot_commit(struct dquot *dquot)
2072 {
2073         int ret;
2074
2075         ret = dquot_commit(dquot);
2076         if (ret < 0)
2077                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2078         return ret;
2079 }
2080
2081 static int f2fs_dquot_acquire(struct dquot *dquot)
2082 {
2083         int ret;
2084
2085         ret = dquot_acquire(dquot);
2086         if (ret < 0)
2087                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2088
2089         return ret;
2090 }
2091
2092 static int f2fs_dquot_release(struct dquot *dquot)
2093 {
2094         int ret;
2095
2096         ret = dquot_release(dquot);
2097         if (ret < 0)
2098                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2099         return ret;
2100 }
2101
2102 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2103 {
2104         struct super_block *sb = dquot->dq_sb;
2105         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2106         int ret;
2107
2108         ret = dquot_mark_dquot_dirty(dquot);
2109
2110         /* if we are using journalled quota */
2111         if (is_journalled_quota(sbi))
2112                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2113
2114         return ret;
2115 }
2116
2117 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2118 {
2119         int ret;
2120
2121         ret = dquot_commit_info(sb, type);
2122         if (ret < 0)
2123                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2124         return ret;
2125 }
2126
2127 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2128 {
2129         *projid = F2FS_I(inode)->i_projid;
2130         return 0;
2131 }
2132
2133 static const struct dquot_operations f2fs_quota_operations = {
2134         .get_reserved_space = f2fs_get_reserved_space,
2135         .write_dquot    = f2fs_dquot_commit,
2136         .acquire_dquot  = f2fs_dquot_acquire,
2137         .release_dquot  = f2fs_dquot_release,
2138         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2139         .write_info     = f2fs_dquot_commit_info,
2140         .alloc_dquot    = dquot_alloc,
2141         .destroy_dquot  = dquot_destroy,
2142         .get_projid     = f2fs_get_projid,
2143         .get_next_id    = dquot_get_next_id,
2144 };
2145
2146 static const struct quotactl_ops f2fs_quotactl_ops = {
2147         .quota_on       = f2fs_quota_on,
2148         .quota_off      = f2fs_quota_off,
2149         .quota_sync     = f2fs_quota_sync,
2150         .get_state      = dquot_get_state,
2151         .set_info       = dquot_set_dqinfo,
2152         .get_dqblk      = dquot_get_dqblk,
2153         .set_dqblk      = dquot_set_dqblk,
2154         .get_nextdqblk  = dquot_get_next_dqblk,
2155 };
2156 #else
2157 int f2fs_quota_sync(struct super_block *sb, int type)
2158 {
2159         return 0;
2160 }
2161
2162 void f2fs_quota_off_umount(struct super_block *sb)
2163 {
2164 }
2165 #endif
2166
2167 static const struct super_operations f2fs_sops = {
2168         .alloc_inode    = f2fs_alloc_inode,
2169         .drop_inode     = f2fs_drop_inode,
2170         .destroy_inode  = f2fs_destroy_inode,
2171         .write_inode    = f2fs_write_inode,
2172         .dirty_inode    = f2fs_dirty_inode,
2173         .show_options   = f2fs_show_options,
2174 #ifdef CONFIG_QUOTA
2175         .quota_read     = f2fs_quota_read,
2176         .quota_write    = f2fs_quota_write,
2177         .get_dquots     = f2fs_get_dquots,
2178 #endif
2179         .evict_inode    = f2fs_evict_inode,
2180         .put_super      = f2fs_put_super,
2181         .sync_fs        = f2fs_sync_fs,
2182         .freeze_fs      = f2fs_freeze,
2183         .unfreeze_fs    = f2fs_unfreeze,
2184         .statfs         = f2fs_statfs,
2185         .remount_fs     = f2fs_remount,
2186 };
2187
2188 #ifdef CONFIG_FS_ENCRYPTION
2189 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2190 {
2191         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2192                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2193                                 ctx, len, NULL);
2194 }
2195
2196 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2197                                                         void *fs_data)
2198 {
2199         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2200
2201         /*
2202          * Encrypting the root directory is not allowed because fsck
2203          * expects lost+found directory to exist and remain unencrypted
2204          * if LOST_FOUND feature is enabled.
2205          *
2206          */
2207         if (f2fs_sb_has_lost_found(sbi) &&
2208                         inode->i_ino == F2FS_ROOT_INO(sbi))
2209                 return -EPERM;
2210
2211         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2212                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2213                                 ctx, len, fs_data, XATTR_CREATE);
2214 }
2215
2216 static bool f2fs_dummy_context(struct inode *inode)
2217 {
2218         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2219 }
2220
2221 static const struct fscrypt_operations f2fs_cryptops = {
2222         .key_prefix     = "f2fs:",
2223         .get_context    = f2fs_get_context,
2224         .set_context    = f2fs_set_context,
2225         .dummy_context  = f2fs_dummy_context,
2226         .empty_dir      = f2fs_empty_dir,
2227         .max_namelen    = F2FS_NAME_LEN,
2228 };
2229 #endif
2230
2231 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2232                 u64 ino, u32 generation)
2233 {
2234         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2235         struct inode *inode;
2236
2237         if (f2fs_check_nid_range(sbi, ino))
2238                 return ERR_PTR(-ESTALE);
2239
2240         /*
2241          * f2fs_iget isn't quite right if the inode is currently unallocated!
2242          * However f2fs_iget currently does appropriate checks to handle stale
2243          * inodes so everything is OK.
2244          */
2245         inode = f2fs_iget(sb, ino);
2246         if (IS_ERR(inode))
2247                 return ERR_CAST(inode);
2248         if (unlikely(generation && inode->i_generation != generation)) {
2249                 /* we didn't find the right inode.. */
2250                 iput(inode);
2251                 return ERR_PTR(-ESTALE);
2252         }
2253         return inode;
2254 }
2255
2256 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2257                 int fh_len, int fh_type)
2258 {
2259         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2260                                     f2fs_nfs_get_inode);
2261 }
2262
2263 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2264                 int fh_len, int fh_type)
2265 {
2266         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2267                                     f2fs_nfs_get_inode);
2268 }
2269
2270 static const struct export_operations f2fs_export_ops = {
2271         .fh_to_dentry = f2fs_fh_to_dentry,
2272         .fh_to_parent = f2fs_fh_to_parent,
2273         .get_parent = f2fs_get_parent,
2274 };
2275
2276 static loff_t max_file_blocks(void)
2277 {
2278         loff_t result = 0;
2279         loff_t leaf_count = ADDRS_PER_BLOCK;
2280
2281         /*
2282          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2283          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2284          * space in inode.i_addr, it will be more safe to reassign
2285          * result as zero.
2286          */
2287
2288         /* two direct node blocks */
2289         result += (leaf_count * 2);
2290
2291         /* two indirect node blocks */
2292         leaf_count *= NIDS_PER_BLOCK;
2293         result += (leaf_count * 2);
2294
2295         /* one double indirect node block */
2296         leaf_count *= NIDS_PER_BLOCK;
2297         result += leaf_count;
2298
2299         return result;
2300 }
2301
2302 static int __f2fs_commit_super(struct buffer_head *bh,
2303                         struct f2fs_super_block *super)
2304 {
2305         lock_buffer(bh);
2306         if (super)
2307                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2308         set_buffer_dirty(bh);
2309         unlock_buffer(bh);
2310
2311         /* it's rare case, we can do fua all the time */
2312         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2313 }
2314
2315 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2316                                         struct buffer_head *bh)
2317 {
2318         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2319                                         (bh->b_data + F2FS_SUPER_OFFSET);
2320         struct super_block *sb = sbi->sb;
2321         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2322         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2323         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2324         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2325         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2326         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2327         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2328         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2329         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2330         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2331         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2332         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2333         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2334         u64 main_end_blkaddr = main_blkaddr +
2335                                 (segment_count_main << log_blocks_per_seg);
2336         u64 seg_end_blkaddr = segment0_blkaddr +
2337                                 (segment_count << log_blocks_per_seg);
2338
2339         if (segment0_blkaddr != cp_blkaddr) {
2340                 f2fs_msg(sb, KERN_INFO,
2341                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2342                         segment0_blkaddr, cp_blkaddr);
2343                 return true;
2344         }
2345
2346         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2347                                                         sit_blkaddr) {
2348                 f2fs_msg(sb, KERN_INFO,
2349                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2350                         cp_blkaddr, sit_blkaddr,
2351                         segment_count_ckpt << log_blocks_per_seg);
2352                 return true;
2353         }
2354
2355         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2356                                                         nat_blkaddr) {
2357                 f2fs_msg(sb, KERN_INFO,
2358                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2359                         sit_blkaddr, nat_blkaddr,
2360                         segment_count_sit << log_blocks_per_seg);
2361                 return true;
2362         }
2363
2364         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2365                                                         ssa_blkaddr) {
2366                 f2fs_msg(sb, KERN_INFO,
2367                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2368                         nat_blkaddr, ssa_blkaddr,
2369                         segment_count_nat << log_blocks_per_seg);
2370                 return true;
2371         }
2372
2373         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2374                                                         main_blkaddr) {
2375                 f2fs_msg(sb, KERN_INFO,
2376                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2377                         ssa_blkaddr, main_blkaddr,
2378                         segment_count_ssa << log_blocks_per_seg);
2379                 return true;
2380         }
2381
2382         if (main_end_blkaddr > seg_end_blkaddr) {
2383                 f2fs_msg(sb, KERN_INFO,
2384                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2385                         main_blkaddr,
2386                         segment0_blkaddr +
2387                                 (segment_count << log_blocks_per_seg),
2388                         segment_count_main << log_blocks_per_seg);
2389                 return true;
2390         } else if (main_end_blkaddr < seg_end_blkaddr) {
2391                 int err = 0;
2392                 char *res;
2393
2394                 /* fix in-memory information all the time */
2395                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2396                                 segment0_blkaddr) >> log_blocks_per_seg);
2397
2398                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2399                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2400                         res = "internally";
2401                 } else {
2402                         err = __f2fs_commit_super(bh, NULL);
2403                         res = err ? "failed" : "done";
2404                 }
2405                 f2fs_msg(sb, KERN_INFO,
2406                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
2407                         res, main_blkaddr,
2408                         segment0_blkaddr +
2409                                 (segment_count << log_blocks_per_seg),
2410                         segment_count_main << log_blocks_per_seg);
2411                 if (err)
2412                         return true;
2413         }
2414         return false;
2415 }
2416
2417 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2418                                 struct buffer_head *bh)
2419 {
2420         block_t segment_count, segs_per_sec, secs_per_zone;
2421         block_t total_sections, blocks_per_seg;
2422         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2423                                         (bh->b_data + F2FS_SUPER_OFFSET);
2424         struct super_block *sb = sbi->sb;
2425         unsigned int blocksize;
2426         size_t crc_offset = 0;
2427         __u32 crc = 0;
2428
2429         /* Check checksum_offset and crc in superblock */
2430         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2431                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2432                 if (crc_offset !=
2433                         offsetof(struct f2fs_super_block, crc)) {
2434                         f2fs_msg(sb, KERN_INFO,
2435                                 "Invalid SB checksum offset: %zu",
2436                                 crc_offset);
2437                         return 1;
2438                 }
2439                 crc = le32_to_cpu(raw_super->crc);
2440                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2441                         f2fs_msg(sb, KERN_INFO,
2442                                 "Invalid SB checksum value: %u", crc);
2443                         return 1;
2444                 }
2445         }
2446
2447         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2448                 f2fs_msg(sb, KERN_INFO,
2449                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
2450                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2451                 return 1;
2452         }
2453
2454         /* Currently, support only 4KB page cache size */
2455         if (F2FS_BLKSIZE != PAGE_SIZE) {
2456                 f2fs_msg(sb, KERN_INFO,
2457                         "Invalid page_cache_size (%lu), supports only 4KB\n",
2458                         PAGE_SIZE);
2459                 return 1;
2460         }
2461
2462         /* Currently, support only 4KB block size */
2463         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2464         if (blocksize != F2FS_BLKSIZE) {
2465                 f2fs_msg(sb, KERN_INFO,
2466                         "Invalid blocksize (%u), supports only 4KB\n",
2467                         blocksize);
2468                 return 1;
2469         }
2470
2471         /* check log blocks per segment */
2472         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2473                 f2fs_msg(sb, KERN_INFO,
2474                         "Invalid log blocks per segment (%u)\n",
2475                         le32_to_cpu(raw_super->log_blocks_per_seg));
2476                 return 1;
2477         }
2478
2479         /* Currently, support 512/1024/2048/4096 bytes sector size */
2480         if (le32_to_cpu(raw_super->log_sectorsize) >
2481                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2482                 le32_to_cpu(raw_super->log_sectorsize) <
2483                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2484                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2485                         le32_to_cpu(raw_super->log_sectorsize));
2486                 return 1;
2487         }
2488         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2489                 le32_to_cpu(raw_super->log_sectorsize) !=
2490                         F2FS_MAX_LOG_SECTOR_SIZE) {
2491                 f2fs_msg(sb, KERN_INFO,
2492                         "Invalid log sectors per block(%u) log sectorsize(%u)",
2493                         le32_to_cpu(raw_super->log_sectors_per_block),
2494                         le32_to_cpu(raw_super->log_sectorsize));
2495                 return 1;
2496         }
2497
2498         segment_count = le32_to_cpu(raw_super->segment_count);
2499         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2500         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2501         total_sections = le32_to_cpu(raw_super->section_count);
2502
2503         /* blocks_per_seg should be 512, given the above check */
2504         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2505
2506         if (segment_count > F2FS_MAX_SEGMENT ||
2507                                 segment_count < F2FS_MIN_SEGMENTS) {
2508                 f2fs_msg(sb, KERN_INFO,
2509                         "Invalid segment count (%u)",
2510                         segment_count);
2511                 return 1;
2512         }
2513
2514         if (total_sections > segment_count ||
2515                         total_sections < F2FS_MIN_SEGMENTS ||
2516                         segs_per_sec > segment_count || !segs_per_sec) {
2517                 f2fs_msg(sb, KERN_INFO,
2518                         "Invalid segment/section count (%u, %u x %u)",
2519                         segment_count, total_sections, segs_per_sec);
2520                 return 1;
2521         }
2522
2523         if ((segment_count / segs_per_sec) < total_sections) {
2524                 f2fs_msg(sb, KERN_INFO,
2525                         "Small segment_count (%u < %u * %u)",
2526                         segment_count, segs_per_sec, total_sections);
2527                 return 1;
2528         }
2529
2530         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2531                 f2fs_msg(sb, KERN_INFO,
2532                         "Wrong segment_count / block_count (%u > %llu)",
2533                         segment_count, le64_to_cpu(raw_super->block_count));
2534                 return 1;
2535         }
2536
2537         if (secs_per_zone > total_sections || !secs_per_zone) {
2538                 f2fs_msg(sb, KERN_INFO,
2539                         "Wrong secs_per_zone / total_sections (%u, %u)",
2540                         secs_per_zone, total_sections);
2541                 return 1;
2542         }
2543         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2544                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2545                         (le32_to_cpu(raw_super->extension_count) +
2546                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2547                 f2fs_msg(sb, KERN_INFO,
2548                         "Corrupted extension count (%u + %u > %u)",
2549                         le32_to_cpu(raw_super->extension_count),
2550                         raw_super->hot_ext_count,
2551                         F2FS_MAX_EXTENSION);
2552                 return 1;
2553         }
2554
2555         if (le32_to_cpu(raw_super->cp_payload) >
2556                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2557                 f2fs_msg(sb, KERN_INFO,
2558                         "Insane cp_payload (%u > %u)",
2559                         le32_to_cpu(raw_super->cp_payload),
2560                         blocks_per_seg - F2FS_CP_PACKS);
2561                 return 1;
2562         }
2563
2564         /* check reserved ino info */
2565         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2566                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2567                 le32_to_cpu(raw_super->root_ino) != 3) {
2568                 f2fs_msg(sb, KERN_INFO,
2569                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2570                         le32_to_cpu(raw_super->node_ino),
2571                         le32_to_cpu(raw_super->meta_ino),
2572                         le32_to_cpu(raw_super->root_ino));
2573                 return 1;
2574         }
2575
2576         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2577         if (sanity_check_area_boundary(sbi, bh))
2578                 return 1;
2579
2580         return 0;
2581 }
2582
2583 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2584 {
2585         unsigned int total, fsmeta;
2586         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2587         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2588         unsigned int ovp_segments, reserved_segments;
2589         unsigned int main_segs, blocks_per_seg;
2590         unsigned int sit_segs, nat_segs;
2591         unsigned int sit_bitmap_size, nat_bitmap_size;
2592         unsigned int log_blocks_per_seg;
2593         unsigned int segment_count_main;
2594         unsigned int cp_pack_start_sum, cp_payload;
2595         block_t user_block_count;
2596         int i, j;
2597
2598         total = le32_to_cpu(raw_super->segment_count);
2599         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2600         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2601         fsmeta += sit_segs;
2602         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2603         fsmeta += nat_segs;
2604         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2605         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2606
2607         if (unlikely(fsmeta >= total))
2608                 return 1;
2609
2610         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2611         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2612
2613         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2614                         ovp_segments == 0 || reserved_segments == 0)) {
2615                 f2fs_msg(sbi->sb, KERN_ERR,
2616                         "Wrong layout: check mkfs.f2fs version");
2617                 return 1;
2618         }
2619
2620         user_block_count = le64_to_cpu(ckpt->user_block_count);
2621         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2622         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2623         if (!user_block_count || user_block_count >=
2624                         segment_count_main << log_blocks_per_seg) {
2625                 f2fs_msg(sbi->sb, KERN_ERR,
2626                         "Wrong user_block_count: %u", user_block_count);
2627                 return 1;
2628         }
2629
2630         main_segs = le32_to_cpu(raw_super->segment_count_main);
2631         blocks_per_seg = sbi->blocks_per_seg;
2632
2633         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2634                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2635                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2636                         return 1;
2637                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2638                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2639                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2640                                 f2fs_msg(sbi->sb, KERN_ERR,
2641                                         "Node segment (%u, %u) has the same "
2642                                         "segno: %u", i, j,
2643                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2644                                 return 1;
2645                         }
2646                 }
2647         }
2648         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2649                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2650                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2651                         return 1;
2652                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2653                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2654                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2655                                 f2fs_msg(sbi->sb, KERN_ERR,
2656                                         "Data segment (%u, %u) has the same "
2657                                         "segno: %u", i, j,
2658                                         le32_to_cpu(ckpt->cur_data_segno[i]));
2659                                 return 1;
2660                         }
2661                 }
2662         }
2663         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2664                 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2665                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2666                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2667                                 f2fs_msg(sbi->sb, KERN_ERR,
2668                                         "Data segment (%u) and Data segment (%u)"
2669                                         " has the same segno: %u", i, j,
2670                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2671                                 return 1;
2672                         }
2673                 }
2674         }
2675
2676         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2677         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2678
2679         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2680                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2681                 f2fs_msg(sbi->sb, KERN_ERR,
2682                         "Wrong bitmap size: sit: %u, nat:%u",
2683                         sit_bitmap_size, nat_bitmap_size);
2684                 return 1;
2685         }
2686
2687         cp_pack_start_sum = __start_sum_addr(sbi);
2688         cp_payload = __cp_payload(sbi);
2689         if (cp_pack_start_sum < cp_payload + 1 ||
2690                 cp_pack_start_sum > blocks_per_seg - 1 -
2691                         NR_CURSEG_TYPE) {
2692                 f2fs_msg(sbi->sb, KERN_ERR,
2693                         "Wrong cp_pack_start_sum: %u",
2694                         cp_pack_start_sum);
2695                 return 1;
2696         }
2697
2698         if (unlikely(f2fs_cp_error(sbi))) {
2699                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2700                 return 1;
2701         }
2702         return 0;
2703 }
2704
2705 static void init_sb_info(struct f2fs_sb_info *sbi)
2706 {
2707         struct f2fs_super_block *raw_super = sbi->raw_super;
2708         int i;
2709
2710         sbi->log_sectors_per_block =
2711                 le32_to_cpu(raw_super->log_sectors_per_block);
2712         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2713         sbi->blocksize = 1 << sbi->log_blocksize;
2714         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2715         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2716         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2717         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2718         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2719         sbi->total_node_count =
2720                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2721                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2722         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2723         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2724         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2725         sbi->cur_victim_sec = NULL_SECNO;
2726         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2727         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2728         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2729         sbi->migration_granularity = sbi->segs_per_sec;
2730
2731         sbi->dir_level = DEF_DIR_LEVEL;
2732         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2733         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2734         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2735         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2736         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2737         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2738                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2739         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2740
2741         for (i = 0; i < NR_COUNT_TYPE; i++)
2742                 atomic_set(&sbi->nr_pages[i], 0);
2743
2744         for (i = 0; i < META; i++)
2745                 atomic_set(&sbi->wb_sync_req[i], 0);
2746
2747         INIT_LIST_HEAD(&sbi->s_list);
2748         mutex_init(&sbi->umount_mutex);
2749         init_rwsem(&sbi->io_order_lock);
2750         spin_lock_init(&sbi->cp_lock);
2751
2752         sbi->dirty_device = 0;
2753         spin_lock_init(&sbi->dev_lock);
2754
2755         init_rwsem(&sbi->sb_lock);
2756 }
2757
2758 static int init_percpu_info(struct f2fs_sb_info *sbi)
2759 {
2760         int err;
2761
2762         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2763         if (err)
2764                 return err;
2765
2766         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2767                                                                 GFP_KERNEL);
2768         if (err)
2769                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2770
2771         return err;
2772 }
2773
2774 #ifdef CONFIG_BLK_DEV_ZONED
2775 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2776 {
2777         struct block_device *bdev = FDEV(devi).bdev;
2778         sector_t nr_sectors = bdev->bd_part->nr_sects;
2779         sector_t sector = 0;
2780         struct blk_zone *zones;
2781         unsigned int i, nr_zones;
2782         unsigned int n = 0;
2783         int err = -EIO;
2784
2785         if (!f2fs_sb_has_blkzoned(sbi))
2786                 return 0;
2787
2788         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2789                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2790                 return -EINVAL;
2791         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2792         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2793                                 __ilog2_u32(sbi->blocks_per_blkz))
2794                 return -EINVAL;
2795         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2796         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2797                                         sbi->log_blocks_per_blkz;
2798         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2799                 FDEV(devi).nr_blkz++;
2800
2801         FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2802                                                                 GFP_KERNEL);
2803         if (!FDEV(devi).blkz_type)
2804                 return -ENOMEM;
2805
2806 #define F2FS_REPORT_NR_ZONES   4096
2807
2808         zones = f2fs_kzalloc(sbi,
2809                              array_size(F2FS_REPORT_NR_ZONES,
2810                                         sizeof(struct blk_zone)),
2811                              GFP_KERNEL);
2812         if (!zones)
2813                 return -ENOMEM;
2814
2815         /* Get block zones type */
2816         while (zones && sector < nr_sectors) {
2817
2818                 nr_zones = F2FS_REPORT_NR_ZONES;
2819                 err = blkdev_report_zones(bdev, sector,
2820                                           zones, &nr_zones,
2821                                           GFP_KERNEL);
2822                 if (err)
2823                         break;
2824                 if (!nr_zones) {
2825                         err = -EIO;
2826                         break;
2827                 }
2828
2829                 for (i = 0; i < nr_zones; i++) {
2830                         FDEV(devi).blkz_type[n] = zones[i].type;
2831                         sector += zones[i].len;
2832                         n++;
2833                 }
2834         }
2835
2836         kvfree(zones);
2837
2838         return err;
2839 }
2840 #endif
2841
2842 /*
2843  * Read f2fs raw super block.
2844  * Because we have two copies of super block, so read both of them
2845  * to get the first valid one. If any one of them is broken, we pass
2846  * them recovery flag back to the caller.
2847  */
2848 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2849                         struct f2fs_super_block **raw_super,
2850                         int *valid_super_block, int *recovery)
2851 {
2852         struct super_block *sb = sbi->sb;
2853         int block;
2854         struct buffer_head *bh;
2855         struct f2fs_super_block *super;
2856         int err = 0;
2857
2858         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2859         if (!super)
2860                 return -ENOMEM;
2861
2862         for (block = 0; block < 2; block++) {
2863                 bh = sb_bread(sb, block);
2864                 if (!bh) {
2865                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2866                                 block + 1);
2867                         err = -EIO;
2868                         continue;
2869                 }
2870
2871                 /* sanity checking of raw super */
2872                 if (sanity_check_raw_super(sbi, bh)) {
2873                         f2fs_msg(sb, KERN_ERR,
2874                                 "Can't find valid F2FS filesystem in %dth superblock",
2875                                 block + 1);
2876                         err = -EINVAL;
2877                         brelse(bh);
2878                         continue;
2879                 }
2880
2881                 if (!*raw_super) {
2882                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2883                                                         sizeof(*super));
2884                         *valid_super_block = block;
2885                         *raw_super = super;
2886                 }
2887                 brelse(bh);
2888         }
2889
2890         /* Fail to read any one of the superblocks*/
2891         if (err < 0)
2892                 *recovery = 1;
2893
2894         /* No valid superblock */
2895         if (!*raw_super)
2896                 kvfree(super);
2897         else
2898                 err = 0;
2899
2900         return err;
2901 }
2902
2903 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2904 {
2905         struct buffer_head *bh;
2906         __u32 crc = 0;
2907         int err;
2908
2909         if ((recover && f2fs_readonly(sbi->sb)) ||
2910                                 bdev_read_only(sbi->sb->s_bdev)) {
2911                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2912                 return -EROFS;
2913         }
2914
2915         /* we should update superblock crc here */
2916         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2917                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2918                                 offsetof(struct f2fs_super_block, crc));
2919                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2920         }
2921
2922         /* write back-up superblock first */
2923         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2924         if (!bh)
2925                 return -EIO;
2926         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2927         brelse(bh);
2928
2929         /* if we are in recovery path, skip writing valid superblock */
2930         if (recover || err)
2931                 return err;
2932
2933         /* write current valid superblock */
2934         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2935         if (!bh)
2936                 return -EIO;
2937         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2938         brelse(bh);
2939         return err;
2940 }
2941
2942 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2943 {
2944         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2945         unsigned int max_devices = MAX_DEVICES;
2946         int i;
2947
2948         /* Initialize single device information */
2949         if (!RDEV(0).path[0]) {
2950                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2951                         return 0;
2952                 max_devices = 1;
2953         }
2954
2955         /*
2956          * Initialize multiple devices information, or single
2957          * zoned block device information.
2958          */
2959         sbi->devs = f2fs_kzalloc(sbi,
2960                                  array_size(max_devices,
2961                                             sizeof(struct f2fs_dev_info)),
2962                                  GFP_KERNEL);
2963         if (!sbi->devs)
2964                 return -ENOMEM;
2965
2966         for (i = 0; i < max_devices; i++) {
2967
2968                 if (i > 0 && !RDEV(i).path[0])
2969                         break;
2970
2971                 if (max_devices == 1) {
2972                         /* Single zoned block device mount */
2973                         FDEV(0).bdev =
2974                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2975                                         sbi->sb->s_mode, sbi->sb->s_type);
2976                 } else {
2977                         /* Multi-device mount */
2978                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2979                         FDEV(i).total_segments =
2980                                 le32_to_cpu(RDEV(i).total_segments);
2981                         if (i == 0) {
2982                                 FDEV(i).start_blk = 0;
2983                                 FDEV(i).end_blk = FDEV(i).start_blk +
2984                                     (FDEV(i).total_segments <<
2985                                     sbi->log_blocks_per_seg) - 1 +
2986                                     le32_to_cpu(raw_super->segment0_blkaddr);
2987                         } else {
2988                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2989                                 FDEV(i).end_blk = FDEV(i).start_blk +
2990                                         (FDEV(i).total_segments <<
2991                                         sbi->log_blocks_per_seg) - 1;
2992                         }
2993                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2994                                         sbi->sb->s_mode, sbi->sb->s_type);
2995                 }
2996                 if (IS_ERR(FDEV(i).bdev))
2997                         return PTR_ERR(FDEV(i).bdev);
2998
2999                 /* to release errored devices */
3000                 sbi->s_ndevs = i + 1;
3001
3002 #ifdef CONFIG_BLK_DEV_ZONED
3003                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3004                                 !f2fs_sb_has_blkzoned(sbi)) {
3005                         f2fs_msg(sbi->sb, KERN_ERR,
3006                                 "Zoned block device feature not enabled\n");
3007                         return -EINVAL;
3008                 }
3009                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3010                         if (init_blkz_info(sbi, i)) {
3011                                 f2fs_msg(sbi->sb, KERN_ERR,
3012                                         "Failed to initialize F2FS blkzone information");
3013                                 return -EINVAL;
3014                         }
3015                         if (max_devices == 1)
3016                                 break;
3017                         f2fs_msg(sbi->sb, KERN_INFO,
3018                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3019                                 i, FDEV(i).path,
3020                                 FDEV(i).total_segments,
3021                                 FDEV(i).start_blk, FDEV(i).end_blk,
3022                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3023                                 "Host-aware" : "Host-managed");
3024                         continue;
3025                 }
3026 #endif
3027                 f2fs_msg(sbi->sb, KERN_INFO,
3028                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3029                                 i, FDEV(i).path,
3030                                 FDEV(i).total_segments,
3031                                 FDEV(i).start_blk, FDEV(i).end_blk);
3032         }
3033         f2fs_msg(sbi->sb, KERN_INFO,
3034                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3035         return 0;
3036 }
3037
3038 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3039 {
3040         struct f2fs_sm_info *sm_i = SM_I(sbi);
3041
3042         /* adjust parameters according to the volume size */
3043         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3044                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3045                 sm_i->dcc_info->discard_granularity = 1;
3046                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3047         }
3048
3049         sbi->readdir_ra = 1;
3050 }
3051
3052 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3053 {
3054         struct f2fs_sb_info *sbi;
3055         struct f2fs_super_block *raw_super;
3056         struct inode *root;
3057         int err;
3058         bool skip_recovery = false, need_fsck = false;
3059         char *options = NULL;
3060         int recovery, i, valid_super_block;
3061         struct curseg_info *seg_i;
3062         int retry_cnt = 1;
3063
3064 try_onemore:
3065         err = -EINVAL;
3066         raw_super = NULL;
3067         valid_super_block = -1;
3068         recovery = 0;
3069
3070         /* allocate memory for f2fs-specific super block info */
3071         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3072         if (!sbi)
3073                 return -ENOMEM;
3074
3075         sbi->sb = sb;
3076
3077         /* Load the checksum driver */
3078         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3079         if (IS_ERR(sbi->s_chksum_driver)) {
3080                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3081                 err = PTR_ERR(sbi->s_chksum_driver);
3082                 sbi->s_chksum_driver = NULL;
3083                 goto free_sbi;
3084         }
3085
3086         /* set a block size */
3087         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3088                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3089                 goto free_sbi;
3090         }
3091
3092         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3093                                                                 &recovery);
3094         if (err)
3095                 goto free_sbi;
3096
3097         sb->s_fs_info = sbi;
3098         sbi->raw_super = raw_super;
3099
3100         /* precompute checksum seed for metadata */
3101         if (f2fs_sb_has_inode_chksum(sbi))
3102                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3103                                                 sizeof(raw_super->uuid));
3104
3105         /*
3106          * The BLKZONED feature indicates that the drive was formatted with
3107          * zone alignment optimization. This is optional for host-aware
3108          * devices, but mandatory for host-managed zoned block devices.
3109          */
3110 #ifndef CONFIG_BLK_DEV_ZONED
3111         if (f2fs_sb_has_blkzoned(sbi)) {
3112                 f2fs_msg(sb, KERN_ERR,
3113                          "Zoned block device support is not enabled\n");
3114                 err = -EOPNOTSUPP;
3115                 goto free_sb_buf;
3116         }
3117 #endif
3118         default_options(sbi);
3119         /* parse mount options */
3120         options = kstrdup((const char *)data, GFP_KERNEL);
3121         if (data && !options) {
3122                 err = -ENOMEM;
3123                 goto free_sb_buf;
3124         }
3125
3126         err = parse_options(sb, options);
3127         if (err)
3128                 goto free_options;
3129
3130         sbi->max_file_blocks = max_file_blocks();
3131         sb->s_maxbytes = sbi->max_file_blocks <<
3132                                 le32_to_cpu(raw_super->log_blocksize);
3133         sb->s_max_links = F2FS_LINK_MAX;
3134
3135 #ifdef CONFIG_QUOTA
3136         sb->dq_op = &f2fs_quota_operations;
3137         if (f2fs_sb_has_quota_ino(sbi))
3138                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3139         else
3140                 sb->s_qcop = &f2fs_quotactl_ops;
3141         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3142
3143         if (f2fs_sb_has_quota_ino(sbi)) {
3144                 for (i = 0; i < MAXQUOTAS; i++) {
3145                         if (f2fs_qf_ino(sbi->sb, i))
3146                                 sbi->nquota_files++;
3147                 }
3148         }
3149 #endif
3150
3151         sb->s_op = &f2fs_sops;
3152 #ifdef CONFIG_FS_ENCRYPTION
3153         sb->s_cop = &f2fs_cryptops;
3154 #endif
3155         sb->s_xattr = f2fs_xattr_handlers;
3156         sb->s_export_op = &f2fs_export_ops;
3157         sb->s_magic = F2FS_SUPER_MAGIC;
3158         sb->s_time_gran = 1;
3159         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3160                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3161         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3162         sb->s_iflags |= SB_I_CGROUPWB;
3163
3164         /* init f2fs-specific super block info */
3165         sbi->valid_super_block = valid_super_block;
3166         mutex_init(&sbi->gc_mutex);
3167         mutex_init(&sbi->writepages);
3168         mutex_init(&sbi->cp_mutex);
3169         init_rwsem(&sbi->node_write);
3170         init_rwsem(&sbi->node_change);
3171
3172         /* disallow all the data/node/meta page writes */
3173         set_sbi_flag(sbi, SBI_POR_DOING);
3174         spin_lock_init(&sbi->stat_lock);
3175
3176         /* init iostat info */
3177         spin_lock_init(&sbi->iostat_lock);
3178         sbi->iostat_enable = false;
3179
3180         for (i = 0; i < NR_PAGE_TYPE; i++) {
3181                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3182                 int j;
3183
3184                 sbi->write_io[i] =
3185                         f2fs_kmalloc(sbi,
3186                                      array_size(n,
3187                                                 sizeof(struct f2fs_bio_info)),
3188                                      GFP_KERNEL);
3189                 if (!sbi->write_io[i]) {
3190                         err = -ENOMEM;
3191                         goto free_bio_info;
3192                 }
3193
3194                 for (j = HOT; j < n; j++) {
3195                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3196                         sbi->write_io[i][j].sbi = sbi;
3197                         sbi->write_io[i][j].bio = NULL;
3198                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3199                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3200                 }
3201         }
3202
3203         init_rwsem(&sbi->cp_rwsem);
3204         init_waitqueue_head(&sbi->cp_wait);
3205         init_sb_info(sbi);
3206
3207         err = init_percpu_info(sbi);
3208         if (err)
3209                 goto free_bio_info;
3210
3211         if (F2FS_IO_SIZE(sbi) > 1) {
3212                 sbi->write_io_dummy =
3213                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3214                 if (!sbi->write_io_dummy) {
3215                         err = -ENOMEM;
3216                         goto free_percpu;
3217                 }
3218         }
3219
3220         /* get an inode for meta space */
3221         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3222         if (IS_ERR(sbi->meta_inode)) {
3223                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3224                 err = PTR_ERR(sbi->meta_inode);
3225                 goto free_io_dummy;
3226         }
3227
3228         err = f2fs_get_valid_checkpoint(sbi);
3229         if (err) {
3230                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3231                 goto free_meta_inode;
3232         }
3233
3234         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3235                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3236         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3237                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3238                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3239         }
3240
3241         /* Initialize device list */
3242         err = f2fs_scan_devices(sbi);
3243         if (err) {
3244                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3245                 goto free_devices;
3246         }
3247
3248         sbi->total_valid_node_count =
3249                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3250         percpu_counter_set(&sbi->total_valid_inode_count,
3251                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3252         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3253         sbi->total_valid_block_count =
3254                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3255         sbi->last_valid_block_count = sbi->total_valid_block_count;
3256         sbi->reserved_blocks = 0;
3257         sbi->current_reserved_blocks = 0;
3258         limit_reserve_root(sbi);
3259
3260         for (i = 0; i < NR_INODE_TYPE; i++) {
3261                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3262                 spin_lock_init(&sbi->inode_lock[i]);
3263         }
3264
3265         f2fs_init_extent_cache_info(sbi);
3266
3267         f2fs_init_ino_entry_info(sbi);
3268
3269         f2fs_init_fsync_node_info(sbi);
3270
3271         /* setup f2fs internal modules */
3272         err = f2fs_build_segment_manager(sbi);
3273         if (err) {
3274                 f2fs_msg(sb, KERN_ERR,
3275                         "Failed to initialize F2FS segment manager");
3276                 goto free_sm;
3277         }
3278         err = f2fs_build_node_manager(sbi);
3279         if (err) {
3280                 f2fs_msg(sb, KERN_ERR,
3281                         "Failed to initialize F2FS node manager");
3282                 goto free_nm;
3283         }
3284
3285         /* For write statistics */
3286         if (sb->s_bdev->bd_part)
3287                 sbi->sectors_written_start =
3288                         (u64)part_stat_read(sb->s_bdev->bd_part,
3289                                             sectors[STAT_WRITE]);
3290
3291         /* Read accumulated write IO statistics if exists */
3292         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3293         if (__exist_node_summaries(sbi))
3294                 sbi->kbytes_written =
3295                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3296
3297         f2fs_build_gc_manager(sbi);
3298
3299         err = f2fs_build_stats(sbi);
3300         if (err)
3301                 goto free_nm;
3302
3303         /* get an inode for node space */
3304         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3305         if (IS_ERR(sbi->node_inode)) {
3306                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3307                 err = PTR_ERR(sbi->node_inode);
3308                 goto free_stats;
3309         }
3310
3311         /* read root inode and dentry */
3312         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3313         if (IS_ERR(root)) {
3314                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3315                 err = PTR_ERR(root);
3316                 goto free_node_inode;
3317         }
3318         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3319                         !root->i_size || !root->i_nlink) {
3320                 iput(root);
3321                 err = -EINVAL;
3322                 goto free_node_inode;
3323         }
3324
3325         sb->s_root = d_make_root(root); /* allocate root dentry */
3326         if (!sb->s_root) {
3327                 err = -ENOMEM;
3328                 goto free_node_inode;
3329         }
3330
3331         err = f2fs_register_sysfs(sbi);
3332         if (err)
3333                 goto free_root_inode;
3334
3335 #ifdef CONFIG_QUOTA
3336         /* Enable quota usage during mount */
3337         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3338                 err = f2fs_enable_quotas(sb);
3339                 if (err)
3340                         f2fs_msg(sb, KERN_ERR,
3341                                 "Cannot turn on quotas: error %d", err);
3342         }
3343 #endif
3344         /* if there are nt orphan nodes free them */
3345         err = f2fs_recover_orphan_inodes(sbi);
3346         if (err)
3347                 goto free_meta;
3348
3349         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3350                 goto reset_checkpoint;
3351
3352         /* recover fsynced data */
3353         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3354                 /*
3355                  * mount should be failed, when device has readonly mode, and
3356                  * previous checkpoint was not done by clean system shutdown.
3357                  */
3358                 if (bdev_read_only(sb->s_bdev) &&
3359                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3360                         err = -EROFS;
3361                         goto free_meta;
3362                 }
3363
3364                 if (need_fsck)
3365                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3366
3367                 if (skip_recovery)
3368                         goto reset_checkpoint;
3369
3370                 err = f2fs_recover_fsync_data(sbi, false);
3371                 if (err < 0) {
3372                         if (err != -ENOMEM)
3373                                 skip_recovery = true;
3374                         need_fsck = true;
3375                         f2fs_msg(sb, KERN_ERR,
3376                                 "Cannot recover all fsync data errno=%d", err);
3377                         goto free_meta;
3378                 }
3379         } else {
3380                 err = f2fs_recover_fsync_data(sbi, true);
3381
3382                 if (!f2fs_readonly(sb) && err > 0) {
3383                         err = -EINVAL;
3384                         f2fs_msg(sb, KERN_ERR,
3385                                 "Need to recover fsync data");
3386                         goto free_meta;
3387                 }
3388         }
3389 reset_checkpoint:
3390         /* f2fs_recover_fsync_data() cleared this already */
3391         clear_sbi_flag(sbi, SBI_POR_DOING);
3392
3393         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3394                 err = f2fs_disable_checkpoint(sbi);
3395                 if (err)
3396                         goto sync_free_meta;
3397         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3398                 f2fs_enable_checkpoint(sbi);
3399         }
3400
3401         /*
3402          * If filesystem is not mounted as read-only then
3403          * do start the gc_thread.
3404          */
3405         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3406                 /* After POR, we can run background GC thread.*/
3407                 err = f2fs_start_gc_thread(sbi);
3408                 if (err)
3409                         goto sync_free_meta;
3410         }
3411         kvfree(options);
3412
3413         /* recover broken superblock */
3414         if (recovery) {
3415                 err = f2fs_commit_super(sbi, true);
3416                 f2fs_msg(sb, KERN_INFO,
3417                         "Try to recover %dth superblock, ret: %d",
3418                         sbi->valid_super_block ? 1 : 2, err);
3419         }
3420
3421         f2fs_join_shrinker(sbi);
3422
3423         f2fs_tuning_parameters(sbi);
3424
3425         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3426                                 cur_cp_version(F2FS_CKPT(sbi)));
3427         f2fs_update_time(sbi, CP_TIME);
3428         f2fs_update_time(sbi, REQ_TIME);
3429         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3430         return 0;
3431
3432 sync_free_meta:
3433         /* safe to flush all the data */
3434         sync_filesystem(sbi->sb);
3435         retry_cnt = 0;
3436
3437 free_meta:
3438 #ifdef CONFIG_QUOTA
3439         f2fs_truncate_quota_inode_pages(sb);
3440         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3441                 f2fs_quota_off_umount(sbi->sb);
3442 #endif
3443         /*
3444          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3445          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3446          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3447          * falls into an infinite loop in f2fs_sync_meta_pages().
3448          */
3449         truncate_inode_pages_final(META_MAPPING(sbi));
3450         /* evict some inodes being cached by GC */
3451         evict_inodes(sb);
3452         f2fs_unregister_sysfs(sbi);
3453 free_root_inode:
3454         dput(sb->s_root);
3455         sb->s_root = NULL;
3456 free_node_inode:
3457         f2fs_release_ino_entry(sbi, true);
3458         truncate_inode_pages_final(NODE_MAPPING(sbi));
3459         iput(sbi->node_inode);
3460         sbi->node_inode = NULL;
3461 free_stats:
3462         f2fs_destroy_stats(sbi);
3463 free_nm:
3464         f2fs_destroy_node_manager(sbi);
3465 free_sm:
3466         f2fs_destroy_segment_manager(sbi);
3467 free_devices:
3468         destroy_device_list(sbi);
3469         kvfree(sbi->ckpt);
3470 free_meta_inode:
3471         make_bad_inode(sbi->meta_inode);
3472         iput(sbi->meta_inode);
3473         sbi->meta_inode = NULL;
3474 free_io_dummy:
3475         mempool_destroy(sbi->write_io_dummy);
3476 free_percpu:
3477         destroy_percpu_info(sbi);
3478 free_bio_info:
3479         for (i = 0; i < NR_PAGE_TYPE; i++)
3480                 kvfree(sbi->write_io[i]);
3481 free_options:
3482 #ifdef CONFIG_QUOTA
3483         for (i = 0; i < MAXQUOTAS; i++)
3484                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3485 #endif
3486         kvfree(options);
3487 free_sb_buf:
3488         kvfree(raw_super);
3489 free_sbi:
3490         if (sbi->s_chksum_driver)
3491                 crypto_free_shash(sbi->s_chksum_driver);
3492         kvfree(sbi);
3493
3494         /* give only one another chance */
3495         if (retry_cnt > 0 && skip_recovery) {
3496                 retry_cnt--;
3497                 shrink_dcache_sb(sb);
3498                 goto try_onemore;
3499         }
3500         return err;
3501 }
3502
3503 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3504                         const char *dev_name, void *data)
3505 {
3506         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3507 }
3508
3509 static void kill_f2fs_super(struct super_block *sb)
3510 {
3511         if (sb->s_root) {
3512                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3513
3514                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3515                 f2fs_stop_gc_thread(sbi);
3516                 f2fs_stop_discard_thread(sbi);
3517
3518                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3519                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3520                         struct cp_control cpc = {
3521                                 .reason = CP_UMOUNT,
3522                         };
3523                         f2fs_write_checkpoint(sbi, &cpc);
3524                 }
3525
3526                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3527                         sb->s_flags &= ~SB_RDONLY;
3528         }
3529         kill_block_super(sb);
3530 }
3531
3532 static struct file_system_type f2fs_fs_type = {
3533         .owner          = THIS_MODULE,
3534         .name           = "f2fs",
3535         .mount          = f2fs_mount,
3536         .kill_sb        = kill_f2fs_super,
3537         .fs_flags       = FS_REQUIRES_DEV,
3538 };
3539 MODULE_ALIAS_FS("f2fs");
3540
3541 static int __init init_inodecache(void)
3542 {
3543         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3544                         sizeof(struct f2fs_inode_info), 0,
3545                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3546         if (!f2fs_inode_cachep)
3547                 return -ENOMEM;
3548         return 0;
3549 }
3550
3551 static void destroy_inodecache(void)
3552 {
3553         /*
3554          * Make sure all delayed rcu free inodes are flushed before we
3555          * destroy cache.
3556          */
3557         rcu_barrier();
3558         kmem_cache_destroy(f2fs_inode_cachep);
3559 }
3560
3561 static int __init init_f2fs_fs(void)
3562 {
3563         int err;
3564
3565         if (PAGE_SIZE != F2FS_BLKSIZE) {
3566                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3567                                 PAGE_SIZE, F2FS_BLKSIZE);
3568                 return -EINVAL;
3569         }
3570
3571         f2fs_build_trace_ios();
3572
3573         err = init_inodecache();
3574         if (err)
3575                 goto fail;
3576         err = f2fs_create_node_manager_caches();
3577         if (err)
3578                 goto free_inodecache;
3579         err = f2fs_create_segment_manager_caches();
3580         if (err)
3581                 goto free_node_manager_caches;
3582         err = f2fs_create_checkpoint_caches();
3583         if (err)
3584                 goto free_segment_manager_caches;
3585         err = f2fs_create_extent_cache();
3586         if (err)
3587                 goto free_checkpoint_caches;
3588         err = f2fs_init_sysfs();
3589         if (err)
3590                 goto free_extent_cache;
3591         err = register_shrinker(&f2fs_shrinker_info);
3592         if (err)
3593                 goto free_sysfs;
3594         err = register_filesystem(&f2fs_fs_type);
3595         if (err)
3596                 goto free_shrinker;
3597         f2fs_create_root_stats();
3598         err = f2fs_init_post_read_processing();
3599         if (err)
3600                 goto free_root_stats;
3601         return 0;
3602
3603 free_root_stats:
3604         f2fs_destroy_root_stats();
3605         unregister_filesystem(&f2fs_fs_type);
3606 free_shrinker:
3607         unregister_shrinker(&f2fs_shrinker_info);
3608 free_sysfs:
3609         f2fs_exit_sysfs();
3610 free_extent_cache:
3611         f2fs_destroy_extent_cache();
3612 free_checkpoint_caches:
3613         f2fs_destroy_checkpoint_caches();
3614 free_segment_manager_caches:
3615         f2fs_destroy_segment_manager_caches();
3616 free_node_manager_caches:
3617         f2fs_destroy_node_manager_caches();
3618 free_inodecache:
3619         destroy_inodecache();
3620 fail:
3621         return err;
3622 }
3623
3624 static void __exit exit_f2fs_fs(void)
3625 {
3626         f2fs_destroy_post_read_processing();
3627         f2fs_destroy_root_stats();
3628         unregister_filesystem(&f2fs_fs_type);
3629         unregister_shrinker(&f2fs_shrinker_info);
3630         f2fs_exit_sysfs();
3631         f2fs_destroy_extent_cache();
3632         f2fs_destroy_checkpoint_caches();
3633         f2fs_destroy_segment_manager_caches();
3634         f2fs_destroy_node_manager_caches();
3635         destroy_inodecache();
3636         f2fs_destroy_trace_ios();
3637 }
3638
3639 module_init(init_f2fs_fs)
3640 module_exit(exit_f2fs_fs)
3641
3642 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3643 MODULE_DESCRIPTION("Flash Friendly File System");
3644 MODULE_LICENSE("GPL");
3645