dcdae3f932e2c24482ec3b26b15ebc5129728c86
[sfrench/cifs-2.6.git] / fs / f2fs / super.c
1 /*
2  * fs/f2fs/super.c
3  *
4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5  *             http://www.samsung.com/
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License version 2 as
9  * published by the Free Software Foundation.
10  */
11 #include <linux/module.h>
12 #include <linux/init.h>
13 #include <linux/fs.h>
14 #include <linux/statfs.h>
15 #include <linux/buffer_head.h>
16 #include <linux/backing-dev.h>
17 #include <linux/kthread.h>
18 #include <linux/parser.h>
19 #include <linux/mount.h>
20 #include <linux/seq_file.h>
21 #include <linux/proc_fs.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/quotaops.h>
26 #include <linux/f2fs_fs.h>
27 #include <linux/sysfs.h>
28 #include <linux/quota.h>
29
30 #include "f2fs.h"
31 #include "node.h"
32 #include "segment.h"
33 #include "xattr.h"
34 #include "gc.h"
35 #include "trace.h"
36
37 #define CREATE_TRACE_POINTS
38 #include <trace/events/f2fs.h>
39
40 static struct kmem_cache *f2fs_inode_cachep;
41
42 #ifdef CONFIG_F2FS_FAULT_INJECTION
43
44 char *fault_name[FAULT_MAX] = {
45         [FAULT_KMALLOC]         = "kmalloc",
46         [FAULT_KVMALLOC]        = "kvmalloc",
47         [FAULT_PAGE_ALLOC]      = "page alloc",
48         [FAULT_PAGE_GET]        = "page get",
49         [FAULT_ALLOC_BIO]       = "alloc bio",
50         [FAULT_ALLOC_NID]       = "alloc nid",
51         [FAULT_ORPHAN]          = "orphan",
52         [FAULT_BLOCK]           = "no more block",
53         [FAULT_DIR_DEPTH]       = "too big dir depth",
54         [FAULT_EVICT_INODE]     = "evict_inode fail",
55         [FAULT_TRUNCATE]        = "truncate fail",
56         [FAULT_IO]              = "IO error",
57         [FAULT_CHECKPOINT]      = "checkpoint error",
58 };
59
60 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
61                                                 unsigned int rate)
62 {
63         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
64
65         if (rate) {
66                 atomic_set(&ffi->inject_ops, 0);
67                 ffi->inject_rate = rate;
68                 ffi->inject_type = (1 << FAULT_MAX) - 1;
69         } else {
70                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
71         }
72 }
73 #endif
74
75 /* f2fs-wide shrinker description */
76 static struct shrinker f2fs_shrinker_info = {
77         .scan_objects = f2fs_shrink_scan,
78         .count_objects = f2fs_shrink_count,
79         .seeks = DEFAULT_SEEKS,
80 };
81
82 enum {
83         Opt_gc_background,
84         Opt_disable_roll_forward,
85         Opt_norecovery,
86         Opt_discard,
87         Opt_nodiscard,
88         Opt_noheap,
89         Opt_heap,
90         Opt_user_xattr,
91         Opt_nouser_xattr,
92         Opt_acl,
93         Opt_noacl,
94         Opt_active_logs,
95         Opt_disable_ext_identify,
96         Opt_inline_xattr,
97         Opt_noinline_xattr,
98         Opt_inline_xattr_size,
99         Opt_inline_data,
100         Opt_inline_dentry,
101         Opt_noinline_dentry,
102         Opt_flush_merge,
103         Opt_noflush_merge,
104         Opt_nobarrier,
105         Opt_fastboot,
106         Opt_extent_cache,
107         Opt_noextent_cache,
108         Opt_noinline_data,
109         Opt_data_flush,
110         Opt_reserve_root,
111         Opt_resgid,
112         Opt_resuid,
113         Opt_mode,
114         Opt_io_size_bits,
115         Opt_fault_injection,
116         Opt_lazytime,
117         Opt_nolazytime,
118         Opt_quota,
119         Opt_noquota,
120         Opt_usrquota,
121         Opt_grpquota,
122         Opt_prjquota,
123         Opt_usrjquota,
124         Opt_grpjquota,
125         Opt_prjjquota,
126         Opt_offusrjquota,
127         Opt_offgrpjquota,
128         Opt_offprjjquota,
129         Opt_jqfmt_vfsold,
130         Opt_jqfmt_vfsv0,
131         Opt_jqfmt_vfsv1,
132         Opt_whint,
133         Opt_alloc,
134         Opt_fsync,
135         Opt_err,
136 };
137
138 static match_table_t f2fs_tokens = {
139         {Opt_gc_background, "background_gc=%s"},
140         {Opt_disable_roll_forward, "disable_roll_forward"},
141         {Opt_norecovery, "norecovery"},
142         {Opt_discard, "discard"},
143         {Opt_nodiscard, "nodiscard"},
144         {Opt_noheap, "no_heap"},
145         {Opt_heap, "heap"},
146         {Opt_user_xattr, "user_xattr"},
147         {Opt_nouser_xattr, "nouser_xattr"},
148         {Opt_acl, "acl"},
149         {Opt_noacl, "noacl"},
150         {Opt_active_logs, "active_logs=%u"},
151         {Opt_disable_ext_identify, "disable_ext_identify"},
152         {Opt_inline_xattr, "inline_xattr"},
153         {Opt_noinline_xattr, "noinline_xattr"},
154         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
155         {Opt_inline_data, "inline_data"},
156         {Opt_inline_dentry, "inline_dentry"},
157         {Opt_noinline_dentry, "noinline_dentry"},
158         {Opt_flush_merge, "flush_merge"},
159         {Opt_noflush_merge, "noflush_merge"},
160         {Opt_nobarrier, "nobarrier"},
161         {Opt_fastboot, "fastboot"},
162         {Opt_extent_cache, "extent_cache"},
163         {Opt_noextent_cache, "noextent_cache"},
164         {Opt_noinline_data, "noinline_data"},
165         {Opt_data_flush, "data_flush"},
166         {Opt_reserve_root, "reserve_root=%u"},
167         {Opt_resgid, "resgid=%u"},
168         {Opt_resuid, "resuid=%u"},
169         {Opt_mode, "mode=%s"},
170         {Opt_io_size_bits, "io_bits=%u"},
171         {Opt_fault_injection, "fault_injection=%u"},
172         {Opt_lazytime, "lazytime"},
173         {Opt_nolazytime, "nolazytime"},
174         {Opt_quota, "quota"},
175         {Opt_noquota, "noquota"},
176         {Opt_usrquota, "usrquota"},
177         {Opt_grpquota, "grpquota"},
178         {Opt_prjquota, "prjquota"},
179         {Opt_usrjquota, "usrjquota=%s"},
180         {Opt_grpjquota, "grpjquota=%s"},
181         {Opt_prjjquota, "prjjquota=%s"},
182         {Opt_offusrjquota, "usrjquota="},
183         {Opt_offgrpjquota, "grpjquota="},
184         {Opt_offprjjquota, "prjjquota="},
185         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
186         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
187         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
188         {Opt_whint, "whint_mode=%s"},
189         {Opt_alloc, "alloc_mode=%s"},
190         {Opt_fsync, "fsync_mode=%s"},
191         {Opt_err, NULL},
192 };
193
194 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
195 {
196         struct va_format vaf;
197         va_list args;
198
199         va_start(args, fmt);
200         vaf.fmt = fmt;
201         vaf.va = &args;
202         printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
203         va_end(args);
204 }
205
206 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
207 {
208         block_t limit = (sbi->user_block_count << 1) / 1000;
209
210         /* limit is 0.2% */
211         if (test_opt(sbi, RESERVE_ROOT) &&
212                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
213                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
214                 f2fs_msg(sbi->sb, KERN_INFO,
215                         "Reduce reserved blocks for root = %u",
216                         F2FS_OPTION(sbi).root_reserved_blocks);
217         }
218         if (!test_opt(sbi, RESERVE_ROOT) &&
219                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
220                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
221                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
222                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
223                 f2fs_msg(sbi->sb, KERN_INFO,
224                         "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
225                                 from_kuid_munged(&init_user_ns,
226                                         F2FS_OPTION(sbi).s_resuid),
227                                 from_kgid_munged(&init_user_ns,
228                                         F2FS_OPTION(sbi).s_resgid));
229 }
230
231 static void init_once(void *foo)
232 {
233         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
234
235         inode_init_once(&fi->vfs_inode);
236 }
237
238 #ifdef CONFIG_QUOTA
239 static const char * const quotatypes[] = INITQFNAMES;
240 #define QTYPE2NAME(t) (quotatypes[t])
241 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
242                                                         substring_t *args)
243 {
244         struct f2fs_sb_info *sbi = F2FS_SB(sb);
245         char *qname;
246         int ret = -EINVAL;
247
248         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
249                 f2fs_msg(sb, KERN_ERR,
250                         "Cannot change journaled "
251                         "quota options when quota turned on");
252                 return -EINVAL;
253         }
254         if (f2fs_sb_has_quota_ino(sb)) {
255                 f2fs_msg(sb, KERN_INFO,
256                         "QUOTA feature is enabled, so ignore qf_name");
257                 return 0;
258         }
259
260         qname = match_strdup(args);
261         if (!qname) {
262                 f2fs_msg(sb, KERN_ERR,
263                         "Not enough memory for storing quotafile name");
264                 return -EINVAL;
265         }
266         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
267                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
268                         ret = 0;
269                 else
270                         f2fs_msg(sb, KERN_ERR,
271                                  "%s quota file already specified",
272                                  QTYPE2NAME(qtype));
273                 goto errout;
274         }
275         if (strchr(qname, '/')) {
276                 f2fs_msg(sb, KERN_ERR,
277                         "quotafile must be on filesystem root");
278                 goto errout;
279         }
280         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
281         set_opt(sbi, QUOTA);
282         return 0;
283 errout:
284         kfree(qname);
285         return ret;
286 }
287
288 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
289 {
290         struct f2fs_sb_info *sbi = F2FS_SB(sb);
291
292         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
293                 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
294                         " when quota turned on");
295                 return -EINVAL;
296         }
297         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
298         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
299         return 0;
300 }
301
302 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
303 {
304         /*
305          * We do the test below only for project quotas. 'usrquota' and
306          * 'grpquota' mount options are allowed even without quota feature
307          * to support legacy quotas in quota files.
308          */
309         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
310                 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
311                          "Cannot enable project quota enforcement.");
312                 return -1;
313         }
314         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
315                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
316                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
317                 if (test_opt(sbi, USRQUOTA) &&
318                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
319                         clear_opt(sbi, USRQUOTA);
320
321                 if (test_opt(sbi, GRPQUOTA) &&
322                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
323                         clear_opt(sbi, GRPQUOTA);
324
325                 if (test_opt(sbi, PRJQUOTA) &&
326                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
327                         clear_opt(sbi, PRJQUOTA);
328
329                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
330                                 test_opt(sbi, PRJQUOTA)) {
331                         f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
332                                         "format mixing");
333                         return -1;
334                 }
335
336                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
337                         f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
338                                         "not specified");
339                         return -1;
340                 }
341         }
342
343         if (f2fs_sb_has_quota_ino(sbi->sb) && F2FS_OPTION(sbi).s_jquota_fmt) {
344                 f2fs_msg(sbi->sb, KERN_INFO,
345                         "QUOTA feature is enabled, so ignore jquota_fmt");
346                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
347         }
348         if (f2fs_sb_has_quota_ino(sbi->sb) && f2fs_readonly(sbi->sb)) {
349                 f2fs_msg(sbi->sb, KERN_INFO,
350                          "Filesystem with quota feature cannot be mounted RDWR "
351                          "without CONFIG_QUOTA");
352                 return -1;
353         }
354         return 0;
355 }
356 #endif
357
358 static int parse_options(struct super_block *sb, char *options)
359 {
360         struct f2fs_sb_info *sbi = F2FS_SB(sb);
361         struct request_queue *q;
362         substring_t args[MAX_OPT_ARGS];
363         char *p, *name;
364         int arg = 0;
365         kuid_t uid;
366         kgid_t gid;
367 #ifdef CONFIG_QUOTA
368         int ret;
369 #endif
370
371         if (!options)
372                 return 0;
373
374         while ((p = strsep(&options, ",")) != NULL) {
375                 int token;
376                 if (!*p)
377                         continue;
378                 /*
379                  * Initialize args struct so we know whether arg was
380                  * found; some options take optional arguments.
381                  */
382                 args[0].to = args[0].from = NULL;
383                 token = match_token(p, f2fs_tokens, args);
384
385                 switch (token) {
386                 case Opt_gc_background:
387                         name = match_strdup(&args[0]);
388
389                         if (!name)
390                                 return -ENOMEM;
391                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
392                                 set_opt(sbi, BG_GC);
393                                 clear_opt(sbi, FORCE_FG_GC);
394                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
395                                 clear_opt(sbi, BG_GC);
396                                 clear_opt(sbi, FORCE_FG_GC);
397                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
398                                 set_opt(sbi, BG_GC);
399                                 set_opt(sbi, FORCE_FG_GC);
400                         } else {
401                                 kfree(name);
402                                 return -EINVAL;
403                         }
404                         kfree(name);
405                         break;
406                 case Opt_disable_roll_forward:
407                         set_opt(sbi, DISABLE_ROLL_FORWARD);
408                         break;
409                 case Opt_norecovery:
410                         /* this option mounts f2fs with ro */
411                         set_opt(sbi, DISABLE_ROLL_FORWARD);
412                         if (!f2fs_readonly(sb))
413                                 return -EINVAL;
414                         break;
415                 case Opt_discard:
416                         q = bdev_get_queue(sb->s_bdev);
417                         if (blk_queue_discard(q)) {
418                                 set_opt(sbi, DISCARD);
419                         } else if (!f2fs_sb_has_blkzoned(sb)) {
420                                 f2fs_msg(sb, KERN_WARNING,
421                                         "mounting with \"discard\" option, but "
422                                         "the device does not support discard");
423                         }
424                         break;
425                 case Opt_nodiscard:
426                         if (f2fs_sb_has_blkzoned(sb)) {
427                                 f2fs_msg(sb, KERN_WARNING,
428                                         "discard is required for zoned block devices");
429                                 return -EINVAL;
430                         }
431                         clear_opt(sbi, DISCARD);
432                         break;
433                 case Opt_noheap:
434                         set_opt(sbi, NOHEAP);
435                         break;
436                 case Opt_heap:
437                         clear_opt(sbi, NOHEAP);
438                         break;
439 #ifdef CONFIG_F2FS_FS_XATTR
440                 case Opt_user_xattr:
441                         set_opt(sbi, XATTR_USER);
442                         break;
443                 case Opt_nouser_xattr:
444                         clear_opt(sbi, XATTR_USER);
445                         break;
446                 case Opt_inline_xattr:
447                         set_opt(sbi, INLINE_XATTR);
448                         break;
449                 case Opt_noinline_xattr:
450                         clear_opt(sbi, INLINE_XATTR);
451                         break;
452                 case Opt_inline_xattr_size:
453                         if (args->from && match_int(args, &arg))
454                                 return -EINVAL;
455                         set_opt(sbi, INLINE_XATTR_SIZE);
456                         F2FS_OPTION(sbi).inline_xattr_size = arg;
457                         break;
458 #else
459                 case Opt_user_xattr:
460                         f2fs_msg(sb, KERN_INFO,
461                                 "user_xattr options not supported");
462                         break;
463                 case Opt_nouser_xattr:
464                         f2fs_msg(sb, KERN_INFO,
465                                 "nouser_xattr options not supported");
466                         break;
467                 case Opt_inline_xattr:
468                         f2fs_msg(sb, KERN_INFO,
469                                 "inline_xattr options not supported");
470                         break;
471                 case Opt_noinline_xattr:
472                         f2fs_msg(sb, KERN_INFO,
473                                 "noinline_xattr options not supported");
474                         break;
475 #endif
476 #ifdef CONFIG_F2FS_FS_POSIX_ACL
477                 case Opt_acl:
478                         set_opt(sbi, POSIX_ACL);
479                         break;
480                 case Opt_noacl:
481                         clear_opt(sbi, POSIX_ACL);
482                         break;
483 #else
484                 case Opt_acl:
485                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
486                         break;
487                 case Opt_noacl:
488                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
489                         break;
490 #endif
491                 case Opt_active_logs:
492                         if (args->from && match_int(args, &arg))
493                                 return -EINVAL;
494                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
495                                 return -EINVAL;
496                         F2FS_OPTION(sbi).active_logs = arg;
497                         break;
498                 case Opt_disable_ext_identify:
499                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
500                         break;
501                 case Opt_inline_data:
502                         set_opt(sbi, INLINE_DATA);
503                         break;
504                 case Opt_inline_dentry:
505                         set_opt(sbi, INLINE_DENTRY);
506                         break;
507                 case Opt_noinline_dentry:
508                         clear_opt(sbi, INLINE_DENTRY);
509                         break;
510                 case Opt_flush_merge:
511                         set_opt(sbi, FLUSH_MERGE);
512                         break;
513                 case Opt_noflush_merge:
514                         clear_opt(sbi, FLUSH_MERGE);
515                         break;
516                 case Opt_nobarrier:
517                         set_opt(sbi, NOBARRIER);
518                         break;
519                 case Opt_fastboot:
520                         set_opt(sbi, FASTBOOT);
521                         break;
522                 case Opt_extent_cache:
523                         set_opt(sbi, EXTENT_CACHE);
524                         break;
525                 case Opt_noextent_cache:
526                         clear_opt(sbi, EXTENT_CACHE);
527                         break;
528                 case Opt_noinline_data:
529                         clear_opt(sbi, INLINE_DATA);
530                         break;
531                 case Opt_data_flush:
532                         set_opt(sbi, DATA_FLUSH);
533                         break;
534                 case Opt_reserve_root:
535                         if (args->from && match_int(args, &arg))
536                                 return -EINVAL;
537                         if (test_opt(sbi, RESERVE_ROOT)) {
538                                 f2fs_msg(sb, KERN_INFO,
539                                         "Preserve previous reserve_root=%u",
540                                         F2FS_OPTION(sbi).root_reserved_blocks);
541                         } else {
542                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
543                                 set_opt(sbi, RESERVE_ROOT);
544                         }
545                         break;
546                 case Opt_resuid:
547                         if (args->from && match_int(args, &arg))
548                                 return -EINVAL;
549                         uid = make_kuid(current_user_ns(), arg);
550                         if (!uid_valid(uid)) {
551                                 f2fs_msg(sb, KERN_ERR,
552                                         "Invalid uid value %d", arg);
553                                 return -EINVAL;
554                         }
555                         F2FS_OPTION(sbi).s_resuid = uid;
556                         break;
557                 case Opt_resgid:
558                         if (args->from && match_int(args, &arg))
559                                 return -EINVAL;
560                         gid = make_kgid(current_user_ns(), arg);
561                         if (!gid_valid(gid)) {
562                                 f2fs_msg(sb, KERN_ERR,
563                                         "Invalid gid value %d", arg);
564                                 return -EINVAL;
565                         }
566                         F2FS_OPTION(sbi).s_resgid = gid;
567                         break;
568                 case Opt_mode:
569                         name = match_strdup(&args[0]);
570
571                         if (!name)
572                                 return -ENOMEM;
573                         if (strlen(name) == 8 &&
574                                         !strncmp(name, "adaptive", 8)) {
575                                 if (f2fs_sb_has_blkzoned(sb)) {
576                                         f2fs_msg(sb, KERN_WARNING,
577                                                  "adaptive mode is not allowed with "
578                                                  "zoned block device feature");
579                                         kfree(name);
580                                         return -EINVAL;
581                                 }
582                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
583                         } else if (strlen(name) == 3 &&
584                                         !strncmp(name, "lfs", 3)) {
585                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
586                         } else {
587                                 kfree(name);
588                                 return -EINVAL;
589                         }
590                         kfree(name);
591                         break;
592                 case Opt_io_size_bits:
593                         if (args->from && match_int(args, &arg))
594                                 return -EINVAL;
595                         if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
596                                 f2fs_msg(sb, KERN_WARNING,
597                                         "Not support %d, larger than %d",
598                                         1 << arg, BIO_MAX_PAGES);
599                                 return -EINVAL;
600                         }
601                         F2FS_OPTION(sbi).write_io_size_bits = arg;
602                         break;
603                 case Opt_fault_injection:
604                         if (args->from && match_int(args, &arg))
605                                 return -EINVAL;
606 #ifdef CONFIG_F2FS_FAULT_INJECTION
607                         f2fs_build_fault_attr(sbi, arg);
608                         set_opt(sbi, FAULT_INJECTION);
609 #else
610                         f2fs_msg(sb, KERN_INFO,
611                                 "FAULT_INJECTION was not selected");
612 #endif
613                         break;
614                 case Opt_lazytime:
615                         sb->s_flags |= SB_LAZYTIME;
616                         break;
617                 case Opt_nolazytime:
618                         sb->s_flags &= ~SB_LAZYTIME;
619                         break;
620 #ifdef CONFIG_QUOTA
621                 case Opt_quota:
622                 case Opt_usrquota:
623                         set_opt(sbi, USRQUOTA);
624                         break;
625                 case Opt_grpquota:
626                         set_opt(sbi, GRPQUOTA);
627                         break;
628                 case Opt_prjquota:
629                         set_opt(sbi, PRJQUOTA);
630                         break;
631                 case Opt_usrjquota:
632                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
633                         if (ret)
634                                 return ret;
635                         break;
636                 case Opt_grpjquota:
637                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
638                         if (ret)
639                                 return ret;
640                         break;
641                 case Opt_prjjquota:
642                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
643                         if (ret)
644                                 return ret;
645                         break;
646                 case Opt_offusrjquota:
647                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
648                         if (ret)
649                                 return ret;
650                         break;
651                 case Opt_offgrpjquota:
652                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
653                         if (ret)
654                                 return ret;
655                         break;
656                 case Opt_offprjjquota:
657                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
658                         if (ret)
659                                 return ret;
660                         break;
661                 case Opt_jqfmt_vfsold:
662                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
663                         break;
664                 case Opt_jqfmt_vfsv0:
665                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
666                         break;
667                 case Opt_jqfmt_vfsv1:
668                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
669                         break;
670                 case Opt_noquota:
671                         clear_opt(sbi, QUOTA);
672                         clear_opt(sbi, USRQUOTA);
673                         clear_opt(sbi, GRPQUOTA);
674                         clear_opt(sbi, PRJQUOTA);
675                         break;
676 #else
677                 case Opt_quota:
678                 case Opt_usrquota:
679                 case Opt_grpquota:
680                 case Opt_prjquota:
681                 case Opt_usrjquota:
682                 case Opt_grpjquota:
683                 case Opt_prjjquota:
684                 case Opt_offusrjquota:
685                 case Opt_offgrpjquota:
686                 case Opt_offprjjquota:
687                 case Opt_jqfmt_vfsold:
688                 case Opt_jqfmt_vfsv0:
689                 case Opt_jqfmt_vfsv1:
690                 case Opt_noquota:
691                         f2fs_msg(sb, KERN_INFO,
692                                         "quota operations not supported");
693                         break;
694 #endif
695                 case Opt_whint:
696                         name = match_strdup(&args[0]);
697                         if (!name)
698                                 return -ENOMEM;
699                         if (strlen(name) == 10 &&
700                                         !strncmp(name, "user-based", 10)) {
701                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
702                         } else if (strlen(name) == 3 &&
703                                         !strncmp(name, "off", 3)) {
704                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
705                         } else if (strlen(name) == 8 &&
706                                         !strncmp(name, "fs-based", 8)) {
707                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
708                         } else {
709                                 kfree(name);
710                                 return -EINVAL;
711                         }
712                         kfree(name);
713                         break;
714                 case Opt_alloc:
715                         name = match_strdup(&args[0]);
716                         if (!name)
717                                 return -ENOMEM;
718
719                         if (strlen(name) == 7 &&
720                                         !strncmp(name, "default", 7)) {
721                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
722                         } else if (strlen(name) == 5 &&
723                                         !strncmp(name, "reuse", 5)) {
724                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
725                         } else {
726                                 kfree(name);
727                                 return -EINVAL;
728                         }
729                         kfree(name);
730                         break;
731                 case Opt_fsync:
732                         name = match_strdup(&args[0]);
733                         if (!name)
734                                 return -ENOMEM;
735                         if (strlen(name) == 5 &&
736                                         !strncmp(name, "posix", 5)) {
737                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
738                         } else if (strlen(name) == 6 &&
739                                         !strncmp(name, "strict", 6)) {
740                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
741                         } else {
742                                 kfree(name);
743                                 return -EINVAL;
744                         }
745                         kfree(name);
746                         break;
747                 default:
748                         f2fs_msg(sb, KERN_ERR,
749                                 "Unrecognized mount option \"%s\" or missing value",
750                                 p);
751                         return -EINVAL;
752                 }
753         }
754 #ifdef CONFIG_QUOTA
755         if (f2fs_check_quota_options(sbi))
756                 return -EINVAL;
757 #endif
758
759         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
760                 f2fs_msg(sb, KERN_ERR,
761                                 "Should set mode=lfs with %uKB-sized IO",
762                                 F2FS_IO_SIZE_KB(sbi));
763                 return -EINVAL;
764         }
765
766         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
767                 if (!f2fs_sb_has_extra_attr(sb) ||
768                         !f2fs_sb_has_flexible_inline_xattr(sb)) {
769                         f2fs_msg(sb, KERN_ERR,
770                                         "extra_attr or flexible_inline_xattr "
771                                         "feature is off");
772                         return -EINVAL;
773                 }
774                 if (!test_opt(sbi, INLINE_XATTR)) {
775                         f2fs_msg(sb, KERN_ERR,
776                                         "inline_xattr_size option should be "
777                                         "set with inline_xattr option");
778                         return -EINVAL;
779                 }
780                 if (!F2FS_OPTION(sbi).inline_xattr_size ||
781                         F2FS_OPTION(sbi).inline_xattr_size >=
782                                         DEF_ADDRS_PER_INODE -
783                                         F2FS_TOTAL_EXTRA_ATTR_SIZE -
784                                         DEF_INLINE_RESERVED_SIZE -
785                                         DEF_MIN_INLINE_SIZE) {
786                         f2fs_msg(sb, KERN_ERR,
787                                         "inline xattr size is out of range");
788                         return -EINVAL;
789                 }
790         }
791
792         /* Not pass down write hints if the number of active logs is lesser
793          * than NR_CURSEG_TYPE.
794          */
795         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
796                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
797         return 0;
798 }
799
800 static struct inode *f2fs_alloc_inode(struct super_block *sb)
801 {
802         struct f2fs_inode_info *fi;
803
804         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
805         if (!fi)
806                 return NULL;
807
808         init_once((void *) fi);
809
810         /* Initialize f2fs-specific inode info */
811         atomic_set(&fi->dirty_pages, 0);
812         fi->i_current_depth = 1;
813         fi->i_advise = 0;
814         init_rwsem(&fi->i_sem);
815         INIT_LIST_HEAD(&fi->dirty_list);
816         INIT_LIST_HEAD(&fi->gdirty_list);
817         INIT_LIST_HEAD(&fi->inmem_ilist);
818         INIT_LIST_HEAD(&fi->inmem_pages);
819         mutex_init(&fi->inmem_lock);
820         init_rwsem(&fi->dio_rwsem[READ]);
821         init_rwsem(&fi->dio_rwsem[WRITE]);
822         init_rwsem(&fi->i_mmap_sem);
823         init_rwsem(&fi->i_xattr_sem);
824
825 #ifdef CONFIG_QUOTA
826         memset(&fi->i_dquot, 0, sizeof(fi->i_dquot));
827         fi->i_reserved_quota = 0;
828 #endif
829         /* Will be used by directory only */
830         fi->i_dir_level = F2FS_SB(sb)->dir_level;
831
832         return &fi->vfs_inode;
833 }
834
835 static int f2fs_drop_inode(struct inode *inode)
836 {
837         int ret;
838         /*
839          * This is to avoid a deadlock condition like below.
840          * writeback_single_inode(inode)
841          *  - f2fs_write_data_page
842          *    - f2fs_gc -> iput -> evict
843          *       - inode_wait_for_writeback(inode)
844          */
845         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
846                 if (!inode->i_nlink && !is_bad_inode(inode)) {
847                         /* to avoid evict_inode call simultaneously */
848                         atomic_inc(&inode->i_count);
849                         spin_unlock(&inode->i_lock);
850
851                         /* some remained atomic pages should discarded */
852                         if (f2fs_is_atomic_file(inode))
853                                 drop_inmem_pages(inode);
854
855                         /* should remain fi->extent_tree for writepage */
856                         f2fs_destroy_extent_node(inode);
857
858                         sb_start_intwrite(inode->i_sb);
859                         f2fs_i_size_write(inode, 0);
860
861                         if (F2FS_HAS_BLOCKS(inode))
862                                 f2fs_truncate(inode);
863
864                         sb_end_intwrite(inode->i_sb);
865
866                         spin_lock(&inode->i_lock);
867                         atomic_dec(&inode->i_count);
868                 }
869                 trace_f2fs_drop_inode(inode, 0);
870                 return 0;
871         }
872         ret = generic_drop_inode(inode);
873         trace_f2fs_drop_inode(inode, ret);
874         return ret;
875 }
876
877 int f2fs_inode_dirtied(struct inode *inode, bool sync)
878 {
879         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
880         int ret = 0;
881
882         spin_lock(&sbi->inode_lock[DIRTY_META]);
883         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
884                 ret = 1;
885         } else {
886                 set_inode_flag(inode, FI_DIRTY_INODE);
887                 stat_inc_dirty_inode(sbi, DIRTY_META);
888         }
889         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
890                 list_add_tail(&F2FS_I(inode)->gdirty_list,
891                                 &sbi->inode_list[DIRTY_META]);
892                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
893         }
894         spin_unlock(&sbi->inode_lock[DIRTY_META]);
895         return ret;
896 }
897
898 void f2fs_inode_synced(struct inode *inode)
899 {
900         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
901
902         spin_lock(&sbi->inode_lock[DIRTY_META]);
903         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
904                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
905                 return;
906         }
907         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
908                 list_del_init(&F2FS_I(inode)->gdirty_list);
909                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
910         }
911         clear_inode_flag(inode, FI_DIRTY_INODE);
912         clear_inode_flag(inode, FI_AUTO_RECOVER);
913         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
914         spin_unlock(&sbi->inode_lock[DIRTY_META]);
915 }
916
917 /*
918  * f2fs_dirty_inode() is called from __mark_inode_dirty()
919  *
920  * We should call set_dirty_inode to write the dirty inode through write_inode.
921  */
922 static void f2fs_dirty_inode(struct inode *inode, int flags)
923 {
924         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
925
926         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
927                         inode->i_ino == F2FS_META_INO(sbi))
928                 return;
929
930         if (flags == I_DIRTY_TIME)
931                 return;
932
933         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
934                 clear_inode_flag(inode, FI_AUTO_RECOVER);
935
936         f2fs_inode_dirtied(inode, false);
937 }
938
939 static void f2fs_i_callback(struct rcu_head *head)
940 {
941         struct inode *inode = container_of(head, struct inode, i_rcu);
942         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
943 }
944
945 static void f2fs_destroy_inode(struct inode *inode)
946 {
947         call_rcu(&inode->i_rcu, f2fs_i_callback);
948 }
949
950 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
951 {
952         percpu_counter_destroy(&sbi->alloc_valid_block_count);
953         percpu_counter_destroy(&sbi->total_valid_inode_count);
954 }
955
956 static void destroy_device_list(struct f2fs_sb_info *sbi)
957 {
958         int i;
959
960         for (i = 0; i < sbi->s_ndevs; i++) {
961                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
962 #ifdef CONFIG_BLK_DEV_ZONED
963                 kfree(FDEV(i).blkz_type);
964 #endif
965         }
966         kfree(sbi->devs);
967 }
968
969 static void f2fs_put_super(struct super_block *sb)
970 {
971         struct f2fs_sb_info *sbi = F2FS_SB(sb);
972         int i;
973         bool dropped;
974
975         f2fs_quota_off_umount(sb);
976
977         /* prevent remaining shrinker jobs */
978         mutex_lock(&sbi->umount_mutex);
979
980         /*
981          * We don't need to do checkpoint when superblock is clean.
982          * But, the previous checkpoint was not done by umount, it needs to do
983          * clean checkpoint again.
984          */
985         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
986                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
987                 struct cp_control cpc = {
988                         .reason = CP_UMOUNT,
989                 };
990                 write_checkpoint(sbi, &cpc);
991         }
992
993         /* be sure to wait for any on-going discard commands */
994         dropped = f2fs_wait_discard_bios(sbi);
995
996         if (f2fs_discard_en(sbi) && !sbi->discard_blks && !dropped) {
997                 struct cp_control cpc = {
998                         .reason = CP_UMOUNT | CP_TRIMMED,
999                 };
1000                 write_checkpoint(sbi, &cpc);
1001         }
1002
1003         /* write_checkpoint can update stat informaion */
1004         f2fs_destroy_stats(sbi);
1005
1006         /*
1007          * normally superblock is clean, so we need to release this.
1008          * In addition, EIO will skip do checkpoint, we need this as well.
1009          */
1010         release_ino_entry(sbi, true);
1011
1012         f2fs_leave_shrinker(sbi);
1013         mutex_unlock(&sbi->umount_mutex);
1014
1015         /* our cp_error case, we can wait for any writeback page */
1016         f2fs_flush_merged_writes(sbi);
1017
1018         iput(sbi->node_inode);
1019         iput(sbi->meta_inode);
1020
1021         /* destroy f2fs internal modules */
1022         destroy_node_manager(sbi);
1023         destroy_segment_manager(sbi);
1024
1025         kfree(sbi->ckpt);
1026
1027         f2fs_unregister_sysfs(sbi);
1028
1029         sb->s_fs_info = NULL;
1030         if (sbi->s_chksum_driver)
1031                 crypto_free_shash(sbi->s_chksum_driver);
1032         kfree(sbi->raw_super);
1033
1034         destroy_device_list(sbi);
1035         mempool_destroy(sbi->write_io_dummy);
1036 #ifdef CONFIG_QUOTA
1037         for (i = 0; i < MAXQUOTAS; i++)
1038                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1039 #endif
1040         destroy_percpu_info(sbi);
1041         for (i = 0; i < NR_PAGE_TYPE; i++)
1042                 kfree(sbi->write_io[i]);
1043         kfree(sbi);
1044 }
1045
1046 int f2fs_sync_fs(struct super_block *sb, int sync)
1047 {
1048         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1049         int err = 0;
1050
1051         if (unlikely(f2fs_cp_error(sbi)))
1052                 return 0;
1053
1054         trace_f2fs_sync_fs(sb, sync);
1055
1056         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1057                 return -EAGAIN;
1058
1059         if (sync) {
1060                 struct cp_control cpc;
1061
1062                 cpc.reason = __get_cp_reason(sbi);
1063
1064                 mutex_lock(&sbi->gc_mutex);
1065                 err = write_checkpoint(sbi, &cpc);
1066                 mutex_unlock(&sbi->gc_mutex);
1067         }
1068         f2fs_trace_ios(NULL, 1);
1069
1070         return err;
1071 }
1072
1073 static int f2fs_freeze(struct super_block *sb)
1074 {
1075         if (f2fs_readonly(sb))
1076                 return 0;
1077
1078         /* IO error happened before */
1079         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1080                 return -EIO;
1081
1082         /* must be clean, since sync_filesystem() was already called */
1083         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1084                 return -EINVAL;
1085         return 0;
1086 }
1087
1088 static int f2fs_unfreeze(struct super_block *sb)
1089 {
1090         return 0;
1091 }
1092
1093 #ifdef CONFIG_QUOTA
1094 static int f2fs_statfs_project(struct super_block *sb,
1095                                 kprojid_t projid, struct kstatfs *buf)
1096 {
1097         struct kqid qid;
1098         struct dquot *dquot;
1099         u64 limit;
1100         u64 curblock;
1101
1102         qid = make_kqid_projid(projid);
1103         dquot = dqget(sb, qid);
1104         if (IS_ERR(dquot))
1105                 return PTR_ERR(dquot);
1106         spin_lock(&dq_data_lock);
1107
1108         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1109                  dquot->dq_dqb.dqb_bsoftlimit :
1110                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1111         if (limit && buf->f_blocks > limit) {
1112                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1113                 buf->f_blocks = limit;
1114                 buf->f_bfree = buf->f_bavail =
1115                         (buf->f_blocks > curblock) ?
1116                          (buf->f_blocks - curblock) : 0;
1117         }
1118
1119         limit = dquot->dq_dqb.dqb_isoftlimit ?
1120                 dquot->dq_dqb.dqb_isoftlimit :
1121                 dquot->dq_dqb.dqb_ihardlimit;
1122         if (limit && buf->f_files > limit) {
1123                 buf->f_files = limit;
1124                 buf->f_ffree =
1125                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1126                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1127         }
1128
1129         spin_unlock(&dq_data_lock);
1130         dqput(dquot);
1131         return 0;
1132 }
1133 #endif
1134
1135 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1136 {
1137         struct super_block *sb = dentry->d_sb;
1138         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1139         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1140         block_t total_count, user_block_count, start_count;
1141         u64 avail_node_count;
1142
1143         total_count = le64_to_cpu(sbi->raw_super->block_count);
1144         user_block_count = sbi->user_block_count;
1145         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1146         buf->f_type = F2FS_SUPER_MAGIC;
1147         buf->f_bsize = sbi->blocksize;
1148
1149         buf->f_blocks = total_count - start_count;
1150         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1151                                                 sbi->current_reserved_blocks;
1152         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1153                 buf->f_bavail = buf->f_bfree -
1154                                 F2FS_OPTION(sbi).root_reserved_blocks;
1155         else
1156                 buf->f_bavail = 0;
1157
1158         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1159                                                 F2FS_RESERVED_NODE_NUM;
1160
1161         if (avail_node_count > user_block_count) {
1162                 buf->f_files = user_block_count;
1163                 buf->f_ffree = buf->f_bavail;
1164         } else {
1165                 buf->f_files = avail_node_count;
1166                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1167                                         buf->f_bavail);
1168         }
1169
1170         buf->f_namelen = F2FS_NAME_LEN;
1171         buf->f_fsid.val[0] = (u32)id;
1172         buf->f_fsid.val[1] = (u32)(id >> 32);
1173
1174 #ifdef CONFIG_QUOTA
1175         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1176                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1177                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1178         }
1179 #endif
1180         return 0;
1181 }
1182
1183 static inline void f2fs_show_quota_options(struct seq_file *seq,
1184                                            struct super_block *sb)
1185 {
1186 #ifdef CONFIG_QUOTA
1187         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1188
1189         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1190                 char *fmtname = "";
1191
1192                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1193                 case QFMT_VFS_OLD:
1194                         fmtname = "vfsold";
1195                         break;
1196                 case QFMT_VFS_V0:
1197                         fmtname = "vfsv0";
1198                         break;
1199                 case QFMT_VFS_V1:
1200                         fmtname = "vfsv1";
1201                         break;
1202                 }
1203                 seq_printf(seq, ",jqfmt=%s", fmtname);
1204         }
1205
1206         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1207                 seq_show_option(seq, "usrjquota",
1208                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1209
1210         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1211                 seq_show_option(seq, "grpjquota",
1212                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1213
1214         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1215                 seq_show_option(seq, "prjjquota",
1216                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1217 #endif
1218 }
1219
1220 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1221 {
1222         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1223
1224         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1225                 if (test_opt(sbi, FORCE_FG_GC))
1226                         seq_printf(seq, ",background_gc=%s", "sync");
1227                 else
1228                         seq_printf(seq, ",background_gc=%s", "on");
1229         } else {
1230                 seq_printf(seq, ",background_gc=%s", "off");
1231         }
1232         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1233                 seq_puts(seq, ",disable_roll_forward");
1234         if (test_opt(sbi, DISCARD))
1235                 seq_puts(seq, ",discard");
1236         if (test_opt(sbi, NOHEAP))
1237                 seq_puts(seq, ",no_heap");
1238         else
1239                 seq_puts(seq, ",heap");
1240 #ifdef CONFIG_F2FS_FS_XATTR
1241         if (test_opt(sbi, XATTR_USER))
1242                 seq_puts(seq, ",user_xattr");
1243         else
1244                 seq_puts(seq, ",nouser_xattr");
1245         if (test_opt(sbi, INLINE_XATTR))
1246                 seq_puts(seq, ",inline_xattr");
1247         else
1248                 seq_puts(seq, ",noinline_xattr");
1249         if (test_opt(sbi, INLINE_XATTR_SIZE))
1250                 seq_printf(seq, ",inline_xattr_size=%u",
1251                                         F2FS_OPTION(sbi).inline_xattr_size);
1252 #endif
1253 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1254         if (test_opt(sbi, POSIX_ACL))
1255                 seq_puts(seq, ",acl");
1256         else
1257                 seq_puts(seq, ",noacl");
1258 #endif
1259         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1260                 seq_puts(seq, ",disable_ext_identify");
1261         if (test_opt(sbi, INLINE_DATA))
1262                 seq_puts(seq, ",inline_data");
1263         else
1264                 seq_puts(seq, ",noinline_data");
1265         if (test_opt(sbi, INLINE_DENTRY))
1266                 seq_puts(seq, ",inline_dentry");
1267         else
1268                 seq_puts(seq, ",noinline_dentry");
1269         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1270                 seq_puts(seq, ",flush_merge");
1271         if (test_opt(sbi, NOBARRIER))
1272                 seq_puts(seq, ",nobarrier");
1273         if (test_opt(sbi, FASTBOOT))
1274                 seq_puts(seq, ",fastboot");
1275         if (test_opt(sbi, EXTENT_CACHE))
1276                 seq_puts(seq, ",extent_cache");
1277         else
1278                 seq_puts(seq, ",noextent_cache");
1279         if (test_opt(sbi, DATA_FLUSH))
1280                 seq_puts(seq, ",data_flush");
1281
1282         seq_puts(seq, ",mode=");
1283         if (test_opt(sbi, ADAPTIVE))
1284                 seq_puts(seq, "adaptive");
1285         else if (test_opt(sbi, LFS))
1286                 seq_puts(seq, "lfs");
1287         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1288         if (test_opt(sbi, RESERVE_ROOT))
1289                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1290                                 F2FS_OPTION(sbi).root_reserved_blocks,
1291                                 from_kuid_munged(&init_user_ns,
1292                                         F2FS_OPTION(sbi).s_resuid),
1293                                 from_kgid_munged(&init_user_ns,
1294                                         F2FS_OPTION(sbi).s_resgid));
1295         if (F2FS_IO_SIZE_BITS(sbi))
1296                 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1297 #ifdef CONFIG_F2FS_FAULT_INJECTION
1298         if (test_opt(sbi, FAULT_INJECTION))
1299                 seq_printf(seq, ",fault_injection=%u",
1300                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1301 #endif
1302 #ifdef CONFIG_QUOTA
1303         if (test_opt(sbi, QUOTA))
1304                 seq_puts(seq, ",quota");
1305         if (test_opt(sbi, USRQUOTA))
1306                 seq_puts(seq, ",usrquota");
1307         if (test_opt(sbi, GRPQUOTA))
1308                 seq_puts(seq, ",grpquota");
1309         if (test_opt(sbi, PRJQUOTA))
1310                 seq_puts(seq, ",prjquota");
1311 #endif
1312         f2fs_show_quota_options(seq, sbi->sb);
1313         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1314                 seq_printf(seq, ",whint_mode=%s", "user-based");
1315         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1316                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1317
1318         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1319                 seq_printf(seq, ",alloc_mode=%s", "default");
1320         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1321                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1322
1323         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1324                 seq_printf(seq, ",fsync_mode=%s", "posix");
1325         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1326                 seq_printf(seq, ",fsync_mode=%s", "strict");
1327         return 0;
1328 }
1329
1330 static void default_options(struct f2fs_sb_info *sbi)
1331 {
1332         /* init some FS parameters */
1333         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1334         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1335         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1336         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1337         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1338         sbi->readdir_ra = 1;
1339
1340         set_opt(sbi, BG_GC);
1341         set_opt(sbi, INLINE_XATTR);
1342         set_opt(sbi, INLINE_DATA);
1343         set_opt(sbi, INLINE_DENTRY);
1344         set_opt(sbi, EXTENT_CACHE);
1345         set_opt(sbi, NOHEAP);
1346         sbi->sb->s_flags |= SB_LAZYTIME;
1347         set_opt(sbi, FLUSH_MERGE);
1348         if (f2fs_sb_has_blkzoned(sbi->sb)) {
1349                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1350                 set_opt(sbi, DISCARD);
1351         } else {
1352                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1353         }
1354
1355 #ifdef CONFIG_F2FS_FS_XATTR
1356         set_opt(sbi, XATTR_USER);
1357 #endif
1358 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1359         set_opt(sbi, POSIX_ACL);
1360 #endif
1361
1362 #ifdef CONFIG_F2FS_FAULT_INJECTION
1363         f2fs_build_fault_attr(sbi, 0);
1364 #endif
1365 }
1366
1367 #ifdef CONFIG_QUOTA
1368 static int f2fs_enable_quotas(struct super_block *sb);
1369 #endif
1370 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1371 {
1372         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1373         struct f2fs_mount_info org_mount_opt;
1374         unsigned long old_sb_flags;
1375         int err;
1376         bool need_restart_gc = false;
1377         bool need_stop_gc = false;
1378         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1379 #ifdef CONFIG_QUOTA
1380         int i, j;
1381 #endif
1382
1383         /*
1384          * Save the old mount options in case we
1385          * need to restore them.
1386          */
1387         org_mount_opt = sbi->mount_opt;
1388         old_sb_flags = sb->s_flags;
1389
1390 #ifdef CONFIG_QUOTA
1391         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1392         for (i = 0; i < MAXQUOTAS; i++) {
1393                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1394                         org_mount_opt.s_qf_names[i] =
1395                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1396                                 GFP_KERNEL);
1397                         if (!org_mount_opt.s_qf_names[i]) {
1398                                 for (j = 0; j < i; j++)
1399                                         kfree(org_mount_opt.s_qf_names[j]);
1400                                 return -ENOMEM;
1401                         }
1402                 } else {
1403                         org_mount_opt.s_qf_names[i] = NULL;
1404                 }
1405         }
1406 #endif
1407
1408         /* recover superblocks we couldn't write due to previous RO mount */
1409         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1410                 err = f2fs_commit_super(sbi, false);
1411                 f2fs_msg(sb, KERN_INFO,
1412                         "Try to recover all the superblocks, ret: %d", err);
1413                 if (!err)
1414                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1415         }
1416
1417         default_options(sbi);
1418
1419         /* parse mount options */
1420         err = parse_options(sb, data);
1421         if (err)
1422                 goto restore_opts;
1423
1424         /*
1425          * Previous and new state of filesystem is RO,
1426          * so skip checking GC and FLUSH_MERGE conditions.
1427          */
1428         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1429                 goto skip;
1430
1431 #ifdef CONFIG_QUOTA
1432         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1433                 err = dquot_suspend(sb, -1);
1434                 if (err < 0)
1435                         goto restore_opts;
1436         } else if (f2fs_readonly(sb) && !(*flags & MS_RDONLY)) {
1437                 /* dquot_resume needs RW */
1438                 sb->s_flags &= ~SB_RDONLY;
1439                 if (sb_any_quota_suspended(sb)) {
1440                         dquot_resume(sb, -1);
1441                 } else if (f2fs_sb_has_quota_ino(sb)) {
1442                         err = f2fs_enable_quotas(sb);
1443                         if (err)
1444                                 goto restore_opts;
1445                 }
1446         }
1447 #endif
1448         /* disallow enable/disable extent_cache dynamically */
1449         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1450                 err = -EINVAL;
1451                 f2fs_msg(sbi->sb, KERN_WARNING,
1452                                 "switch extent_cache option is not allowed");
1453                 goto restore_opts;
1454         }
1455
1456         /*
1457          * We stop the GC thread if FS is mounted as RO
1458          * or if background_gc = off is passed in mount
1459          * option. Also sync the filesystem.
1460          */
1461         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1462                 if (sbi->gc_thread) {
1463                         stop_gc_thread(sbi);
1464                         need_restart_gc = true;
1465                 }
1466         } else if (!sbi->gc_thread) {
1467                 err = start_gc_thread(sbi);
1468                 if (err)
1469                         goto restore_opts;
1470                 need_stop_gc = true;
1471         }
1472
1473         if (*flags & SB_RDONLY ||
1474                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1475                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1476                 sync_inodes_sb(sb);
1477
1478                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1479                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1480                 f2fs_sync_fs(sb, 1);
1481                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1482         }
1483
1484         /*
1485          * We stop issue flush thread if FS is mounted as RO
1486          * or if flush_merge is not passed in mount option.
1487          */
1488         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1489                 clear_opt(sbi, FLUSH_MERGE);
1490                 destroy_flush_cmd_control(sbi, false);
1491         } else {
1492                 err = create_flush_cmd_control(sbi);
1493                 if (err)
1494                         goto restore_gc;
1495         }
1496 skip:
1497 #ifdef CONFIG_QUOTA
1498         /* Release old quota file names */
1499         for (i = 0; i < MAXQUOTAS; i++)
1500                 kfree(org_mount_opt.s_qf_names[i]);
1501 #endif
1502         /* Update the POSIXACL Flag */
1503         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1504                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1505
1506         limit_reserve_root(sbi);
1507         return 0;
1508 restore_gc:
1509         if (need_restart_gc) {
1510                 if (start_gc_thread(sbi))
1511                         f2fs_msg(sbi->sb, KERN_WARNING,
1512                                 "background gc thread has stopped");
1513         } else if (need_stop_gc) {
1514                 stop_gc_thread(sbi);
1515         }
1516 restore_opts:
1517 #ifdef CONFIG_QUOTA
1518         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1519         for (i = 0; i < MAXQUOTAS; i++) {
1520                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1521                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1522         }
1523 #endif
1524         sbi->mount_opt = org_mount_opt;
1525         sb->s_flags = old_sb_flags;
1526         return err;
1527 }
1528
1529 #ifdef CONFIG_QUOTA
1530 /* Read data from quotafile */
1531 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1532                                size_t len, loff_t off)
1533 {
1534         struct inode *inode = sb_dqopt(sb)->files[type];
1535         struct address_space *mapping = inode->i_mapping;
1536         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1537         int offset = off & (sb->s_blocksize - 1);
1538         int tocopy;
1539         size_t toread;
1540         loff_t i_size = i_size_read(inode);
1541         struct page *page;
1542         char *kaddr;
1543
1544         if (off > i_size)
1545                 return 0;
1546
1547         if (off + len > i_size)
1548                 len = i_size - off;
1549         toread = len;
1550         while (toread > 0) {
1551                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1552 repeat:
1553                 page = read_mapping_page(mapping, blkidx, NULL);
1554                 if (IS_ERR(page)) {
1555                         if (PTR_ERR(page) == -ENOMEM) {
1556                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1557                                 goto repeat;
1558                         }
1559                         return PTR_ERR(page);
1560                 }
1561
1562                 lock_page(page);
1563
1564                 if (unlikely(page->mapping != mapping)) {
1565                         f2fs_put_page(page, 1);
1566                         goto repeat;
1567                 }
1568                 if (unlikely(!PageUptodate(page))) {
1569                         f2fs_put_page(page, 1);
1570                         return -EIO;
1571                 }
1572
1573                 kaddr = kmap_atomic(page);
1574                 memcpy(data, kaddr + offset, tocopy);
1575                 kunmap_atomic(kaddr);
1576                 f2fs_put_page(page, 1);
1577
1578                 offset = 0;
1579                 toread -= tocopy;
1580                 data += tocopy;
1581                 blkidx++;
1582         }
1583         return len;
1584 }
1585
1586 /* Write to quotafile */
1587 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1588                                 const char *data, size_t len, loff_t off)
1589 {
1590         struct inode *inode = sb_dqopt(sb)->files[type];
1591         struct address_space *mapping = inode->i_mapping;
1592         const struct address_space_operations *a_ops = mapping->a_ops;
1593         int offset = off & (sb->s_blocksize - 1);
1594         size_t towrite = len;
1595         struct page *page;
1596         char *kaddr;
1597         int err = 0;
1598         int tocopy;
1599
1600         while (towrite > 0) {
1601                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1602                                                                 towrite);
1603 retry:
1604                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1605                                                         &page, NULL);
1606                 if (unlikely(err)) {
1607                         if (err == -ENOMEM) {
1608                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1609                                 goto retry;
1610                         }
1611                         break;
1612                 }
1613
1614                 kaddr = kmap_atomic(page);
1615                 memcpy(kaddr + offset, data, tocopy);
1616                 kunmap_atomic(kaddr);
1617                 flush_dcache_page(page);
1618
1619                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1620                                                 page, NULL);
1621                 offset = 0;
1622                 towrite -= tocopy;
1623                 off += tocopy;
1624                 data += tocopy;
1625                 cond_resched();
1626         }
1627
1628         if (len == towrite)
1629                 return err;
1630         inode->i_mtime = inode->i_ctime = current_time(inode);
1631         f2fs_mark_inode_dirty_sync(inode, false);
1632         return len - towrite;
1633 }
1634
1635 static struct dquot **f2fs_get_dquots(struct inode *inode)
1636 {
1637         return F2FS_I(inode)->i_dquot;
1638 }
1639
1640 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1641 {
1642         return &F2FS_I(inode)->i_reserved_quota;
1643 }
1644
1645 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1646 {
1647         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1648                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1649 }
1650
1651 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1652 {
1653         int enabled = 0;
1654         int i, err;
1655
1656         if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) {
1657                 err = f2fs_enable_quotas(sbi->sb);
1658                 if (err) {
1659                         f2fs_msg(sbi->sb, KERN_ERR,
1660                                         "Cannot turn on quota_ino: %d", err);
1661                         return 0;
1662                 }
1663                 return 1;
1664         }
1665
1666         for (i = 0; i < MAXQUOTAS; i++) {
1667                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1668                         err = f2fs_quota_on_mount(sbi, i);
1669                         if (!err) {
1670                                 enabled = 1;
1671                                 continue;
1672                         }
1673                         f2fs_msg(sbi->sb, KERN_ERR,
1674                                 "Cannot turn on quotas: %d on %d", err, i);
1675                 }
1676         }
1677         return enabled;
1678 }
1679
1680 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1681                              unsigned int flags)
1682 {
1683         struct inode *qf_inode;
1684         unsigned long qf_inum;
1685         int err;
1686
1687         BUG_ON(!f2fs_sb_has_quota_ino(sb));
1688
1689         qf_inum = f2fs_qf_ino(sb, type);
1690         if (!qf_inum)
1691                 return -EPERM;
1692
1693         qf_inode = f2fs_iget(sb, qf_inum);
1694         if (IS_ERR(qf_inode)) {
1695                 f2fs_msg(sb, KERN_ERR,
1696                         "Bad quota inode %u:%lu", type, qf_inum);
1697                 return PTR_ERR(qf_inode);
1698         }
1699
1700         /* Don't account quota for quota files to avoid recursion */
1701         qf_inode->i_flags |= S_NOQUOTA;
1702         err = dquot_enable(qf_inode, type, format_id, flags);
1703         iput(qf_inode);
1704         return err;
1705 }
1706
1707 static int f2fs_enable_quotas(struct super_block *sb)
1708 {
1709         int type, err = 0;
1710         unsigned long qf_inum;
1711         bool quota_mopt[MAXQUOTAS] = {
1712                 test_opt(F2FS_SB(sb), USRQUOTA),
1713                 test_opt(F2FS_SB(sb), GRPQUOTA),
1714                 test_opt(F2FS_SB(sb), PRJQUOTA),
1715         };
1716
1717         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
1718         for (type = 0; type < MAXQUOTAS; type++) {
1719                 qf_inum = f2fs_qf_ino(sb, type);
1720                 if (qf_inum) {
1721                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1722                                 DQUOT_USAGE_ENABLED |
1723                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1724                         if (err) {
1725                                 f2fs_msg(sb, KERN_ERR,
1726                                         "Failed to enable quota tracking "
1727                                         "(type=%d, err=%d). Please run "
1728                                         "fsck to fix.", type, err);
1729                                 for (type--; type >= 0; type--)
1730                                         dquot_quota_off(sb, type);
1731                                 return err;
1732                         }
1733                 }
1734         }
1735         return 0;
1736 }
1737
1738 static int f2fs_quota_sync(struct super_block *sb, int type)
1739 {
1740         struct quota_info *dqopt = sb_dqopt(sb);
1741         int cnt;
1742         int ret;
1743
1744         ret = dquot_writeback_dquots(sb, type);
1745         if (ret)
1746                 return ret;
1747
1748         /*
1749          * Now when everything is written we can discard the pagecache so
1750          * that userspace sees the changes.
1751          */
1752         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1753                 if (type != -1 && cnt != type)
1754                         continue;
1755                 if (!sb_has_quota_active(sb, cnt))
1756                         continue;
1757
1758                 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1759                 if (ret)
1760                         return ret;
1761
1762                 inode_lock(dqopt->files[cnt]);
1763                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1764                 inode_unlock(dqopt->files[cnt]);
1765         }
1766         return 0;
1767 }
1768
1769 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1770                                                         const struct path *path)
1771 {
1772         struct inode *inode;
1773         int err;
1774
1775         err = f2fs_quota_sync(sb, type);
1776         if (err)
1777                 return err;
1778
1779         err = dquot_quota_on(sb, type, format_id, path);
1780         if (err)
1781                 return err;
1782
1783         inode = d_inode(path->dentry);
1784
1785         inode_lock(inode);
1786         F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL;
1787         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1788                                         S_NOATIME | S_IMMUTABLE);
1789         inode_unlock(inode);
1790         f2fs_mark_inode_dirty_sync(inode, false);
1791
1792         return 0;
1793 }
1794
1795 static int f2fs_quota_off(struct super_block *sb, int type)
1796 {
1797         struct inode *inode = sb_dqopt(sb)->files[type];
1798         int err;
1799
1800         if (!inode || !igrab(inode))
1801                 return dquot_quota_off(sb, type);
1802
1803         f2fs_quota_sync(sb, type);
1804
1805         err = dquot_quota_off(sb, type);
1806         if (err || f2fs_sb_has_quota_ino(sb))
1807                 goto out_put;
1808
1809         inode_lock(inode);
1810         F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL);
1811         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1812         inode_unlock(inode);
1813         f2fs_mark_inode_dirty_sync(inode, false);
1814 out_put:
1815         iput(inode);
1816         return err;
1817 }
1818
1819 void f2fs_quota_off_umount(struct super_block *sb)
1820 {
1821         int type;
1822
1823         for (type = 0; type < MAXQUOTAS; type++)
1824                 f2fs_quota_off(sb, type);
1825 }
1826
1827 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1828 {
1829         *projid = F2FS_I(inode)->i_projid;
1830         return 0;
1831 }
1832
1833 static const struct dquot_operations f2fs_quota_operations = {
1834         .get_reserved_space = f2fs_get_reserved_space,
1835         .write_dquot    = dquot_commit,
1836         .acquire_dquot  = dquot_acquire,
1837         .release_dquot  = dquot_release,
1838         .mark_dirty     = dquot_mark_dquot_dirty,
1839         .write_info     = dquot_commit_info,
1840         .alloc_dquot    = dquot_alloc,
1841         .destroy_dquot  = dquot_destroy,
1842         .get_projid     = f2fs_get_projid,
1843         .get_next_id    = dquot_get_next_id,
1844 };
1845
1846 static const struct quotactl_ops f2fs_quotactl_ops = {
1847         .quota_on       = f2fs_quota_on,
1848         .quota_off      = f2fs_quota_off,
1849         .quota_sync     = f2fs_quota_sync,
1850         .get_state      = dquot_get_state,
1851         .set_info       = dquot_set_dqinfo,
1852         .get_dqblk      = dquot_get_dqblk,
1853         .set_dqblk      = dquot_set_dqblk,
1854         .get_nextdqblk  = dquot_get_next_dqblk,
1855 };
1856 #else
1857 void f2fs_quota_off_umount(struct super_block *sb)
1858 {
1859 }
1860 #endif
1861
1862 static const struct super_operations f2fs_sops = {
1863         .alloc_inode    = f2fs_alloc_inode,
1864         .drop_inode     = f2fs_drop_inode,
1865         .destroy_inode  = f2fs_destroy_inode,
1866         .write_inode    = f2fs_write_inode,
1867         .dirty_inode    = f2fs_dirty_inode,
1868         .show_options   = f2fs_show_options,
1869 #ifdef CONFIG_QUOTA
1870         .quota_read     = f2fs_quota_read,
1871         .quota_write    = f2fs_quota_write,
1872         .get_dquots     = f2fs_get_dquots,
1873 #endif
1874         .evict_inode    = f2fs_evict_inode,
1875         .put_super      = f2fs_put_super,
1876         .sync_fs        = f2fs_sync_fs,
1877         .freeze_fs      = f2fs_freeze,
1878         .unfreeze_fs    = f2fs_unfreeze,
1879         .statfs         = f2fs_statfs,
1880         .remount_fs     = f2fs_remount,
1881 };
1882
1883 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1884 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1885 {
1886         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1887                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1888                                 ctx, len, NULL);
1889 }
1890
1891 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1892                                                         void *fs_data)
1893 {
1894         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1895
1896         /*
1897          * Encrypting the root directory is not allowed because fsck
1898          * expects lost+found directory to exist and remain unencrypted
1899          * if LOST_FOUND feature is enabled.
1900          *
1901          */
1902         if (f2fs_sb_has_lost_found(sbi->sb) &&
1903                         inode->i_ino == F2FS_ROOT_INO(sbi))
1904                 return -EPERM;
1905
1906         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1907                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1908                                 ctx, len, fs_data, XATTR_CREATE);
1909 }
1910
1911 static unsigned f2fs_max_namelen(struct inode *inode)
1912 {
1913         return S_ISLNK(inode->i_mode) ?
1914                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1915 }
1916
1917 static const struct fscrypt_operations f2fs_cryptops = {
1918         .key_prefix     = "f2fs:",
1919         .get_context    = f2fs_get_context,
1920         .set_context    = f2fs_set_context,
1921         .empty_dir      = f2fs_empty_dir,
1922         .max_namelen    = f2fs_max_namelen,
1923 };
1924 #endif
1925
1926 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1927                 u64 ino, u32 generation)
1928 {
1929         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1930         struct inode *inode;
1931
1932         if (check_nid_range(sbi, ino))
1933                 return ERR_PTR(-ESTALE);
1934
1935         /*
1936          * f2fs_iget isn't quite right if the inode is currently unallocated!
1937          * However f2fs_iget currently does appropriate checks to handle stale
1938          * inodes so everything is OK.
1939          */
1940         inode = f2fs_iget(sb, ino);
1941         if (IS_ERR(inode))
1942                 return ERR_CAST(inode);
1943         if (unlikely(generation && inode->i_generation != generation)) {
1944                 /* we didn't find the right inode.. */
1945                 iput(inode);
1946                 return ERR_PTR(-ESTALE);
1947         }
1948         return inode;
1949 }
1950
1951 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1952                 int fh_len, int fh_type)
1953 {
1954         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1955                                     f2fs_nfs_get_inode);
1956 }
1957
1958 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1959                 int fh_len, int fh_type)
1960 {
1961         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1962                                     f2fs_nfs_get_inode);
1963 }
1964
1965 static const struct export_operations f2fs_export_ops = {
1966         .fh_to_dentry = f2fs_fh_to_dentry,
1967         .fh_to_parent = f2fs_fh_to_parent,
1968         .get_parent = f2fs_get_parent,
1969 };
1970
1971 static loff_t max_file_blocks(void)
1972 {
1973         loff_t result = 0;
1974         loff_t leaf_count = ADDRS_PER_BLOCK;
1975
1976         /*
1977          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1978          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1979          * space in inode.i_addr, it will be more safe to reassign
1980          * result as zero.
1981          */
1982
1983         /* two direct node blocks */
1984         result += (leaf_count * 2);
1985
1986         /* two indirect node blocks */
1987         leaf_count *= NIDS_PER_BLOCK;
1988         result += (leaf_count * 2);
1989
1990         /* one double indirect node block */
1991         leaf_count *= NIDS_PER_BLOCK;
1992         result += leaf_count;
1993
1994         return result;
1995 }
1996
1997 static int __f2fs_commit_super(struct buffer_head *bh,
1998                         struct f2fs_super_block *super)
1999 {
2000         lock_buffer(bh);
2001         if (super)
2002                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2003         set_buffer_dirty(bh);
2004         unlock_buffer(bh);
2005
2006         /* it's rare case, we can do fua all the time */
2007         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2008 }
2009
2010 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2011                                         struct buffer_head *bh)
2012 {
2013         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2014                                         (bh->b_data + F2FS_SUPER_OFFSET);
2015         struct super_block *sb = sbi->sb;
2016         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2017         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2018         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2019         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2020         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2021         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2022         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2023         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2024         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2025         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2026         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2027         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2028         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2029         u64 main_end_blkaddr = main_blkaddr +
2030                                 (segment_count_main << log_blocks_per_seg);
2031         u64 seg_end_blkaddr = segment0_blkaddr +
2032                                 (segment_count << log_blocks_per_seg);
2033
2034         if (segment0_blkaddr != cp_blkaddr) {
2035                 f2fs_msg(sb, KERN_INFO,
2036                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2037                         segment0_blkaddr, cp_blkaddr);
2038                 return true;
2039         }
2040
2041         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2042                                                         sit_blkaddr) {
2043                 f2fs_msg(sb, KERN_INFO,
2044                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2045                         cp_blkaddr, sit_blkaddr,
2046                         segment_count_ckpt << log_blocks_per_seg);
2047                 return true;
2048         }
2049
2050         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2051                                                         nat_blkaddr) {
2052                 f2fs_msg(sb, KERN_INFO,
2053                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2054                         sit_blkaddr, nat_blkaddr,
2055                         segment_count_sit << log_blocks_per_seg);
2056                 return true;
2057         }
2058
2059         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2060                                                         ssa_blkaddr) {
2061                 f2fs_msg(sb, KERN_INFO,
2062                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2063                         nat_blkaddr, ssa_blkaddr,
2064                         segment_count_nat << log_blocks_per_seg);
2065                 return true;
2066         }
2067
2068         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2069                                                         main_blkaddr) {
2070                 f2fs_msg(sb, KERN_INFO,
2071                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2072                         ssa_blkaddr, main_blkaddr,
2073                         segment_count_ssa << log_blocks_per_seg);
2074                 return true;
2075         }
2076
2077         if (main_end_blkaddr > seg_end_blkaddr) {
2078                 f2fs_msg(sb, KERN_INFO,
2079                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2080                         main_blkaddr,
2081                         segment0_blkaddr +
2082                                 (segment_count << log_blocks_per_seg),
2083                         segment_count_main << log_blocks_per_seg);
2084                 return true;
2085         } else if (main_end_blkaddr < seg_end_blkaddr) {
2086                 int err = 0;
2087                 char *res;
2088
2089                 /* fix in-memory information all the time */
2090                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2091                                 segment0_blkaddr) >> log_blocks_per_seg);
2092
2093                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2094                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2095                         res = "internally";
2096                 } else {
2097                         err = __f2fs_commit_super(bh, NULL);
2098                         res = err ? "failed" : "done";
2099                 }
2100                 f2fs_msg(sb, KERN_INFO,
2101                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
2102                         res, main_blkaddr,
2103                         segment0_blkaddr +
2104                                 (segment_count << log_blocks_per_seg),
2105                         segment_count_main << log_blocks_per_seg);
2106                 if (err)
2107                         return true;
2108         }
2109         return false;
2110 }
2111
2112 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2113                                 struct buffer_head *bh)
2114 {
2115         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2116                                         (bh->b_data + F2FS_SUPER_OFFSET);
2117         struct super_block *sb = sbi->sb;
2118         unsigned int blocksize;
2119
2120         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2121                 f2fs_msg(sb, KERN_INFO,
2122                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
2123                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2124                 return 1;
2125         }
2126
2127         /* Currently, support only 4KB page cache size */
2128         if (F2FS_BLKSIZE != PAGE_SIZE) {
2129                 f2fs_msg(sb, KERN_INFO,
2130                         "Invalid page_cache_size (%lu), supports only 4KB\n",
2131                         PAGE_SIZE);
2132                 return 1;
2133         }
2134
2135         /* Currently, support only 4KB block size */
2136         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2137         if (blocksize != F2FS_BLKSIZE) {
2138                 f2fs_msg(sb, KERN_INFO,
2139                         "Invalid blocksize (%u), supports only 4KB\n",
2140                         blocksize);
2141                 return 1;
2142         }
2143
2144         /* check log blocks per segment */
2145         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2146                 f2fs_msg(sb, KERN_INFO,
2147                         "Invalid log blocks per segment (%u)\n",
2148                         le32_to_cpu(raw_super->log_blocks_per_seg));
2149                 return 1;
2150         }
2151
2152         /* Currently, support 512/1024/2048/4096 bytes sector size */
2153         if (le32_to_cpu(raw_super->log_sectorsize) >
2154                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2155                 le32_to_cpu(raw_super->log_sectorsize) <
2156                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2157                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2158                         le32_to_cpu(raw_super->log_sectorsize));
2159                 return 1;
2160         }
2161         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2162                 le32_to_cpu(raw_super->log_sectorsize) !=
2163                         F2FS_MAX_LOG_SECTOR_SIZE) {
2164                 f2fs_msg(sb, KERN_INFO,
2165                         "Invalid log sectors per block(%u) log sectorsize(%u)",
2166                         le32_to_cpu(raw_super->log_sectors_per_block),
2167                         le32_to_cpu(raw_super->log_sectorsize));
2168                 return 1;
2169         }
2170
2171         /* check reserved ino info */
2172         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2173                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2174                 le32_to_cpu(raw_super->root_ino) != 3) {
2175                 f2fs_msg(sb, KERN_INFO,
2176                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2177                         le32_to_cpu(raw_super->node_ino),
2178                         le32_to_cpu(raw_super->meta_ino),
2179                         le32_to_cpu(raw_super->root_ino));
2180                 return 1;
2181         }
2182
2183         if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) {
2184                 f2fs_msg(sb, KERN_INFO,
2185                         "Invalid segment count (%u)",
2186                         le32_to_cpu(raw_super->segment_count));
2187                 return 1;
2188         }
2189
2190         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2191         if (sanity_check_area_boundary(sbi, bh))
2192                 return 1;
2193
2194         return 0;
2195 }
2196
2197 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
2198 {
2199         unsigned int total, fsmeta;
2200         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2201         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2202         unsigned int ovp_segments, reserved_segments;
2203         unsigned int main_segs, blocks_per_seg;
2204         int i;
2205
2206         total = le32_to_cpu(raw_super->segment_count);
2207         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2208         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
2209         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
2210         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2211         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2212
2213         if (unlikely(fsmeta >= total))
2214                 return 1;
2215
2216         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2217         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2218
2219         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2220                         ovp_segments == 0 || reserved_segments == 0)) {
2221                 f2fs_msg(sbi->sb, KERN_ERR,
2222                         "Wrong layout: check mkfs.f2fs version");
2223                 return 1;
2224         }
2225
2226         main_segs = le32_to_cpu(raw_super->segment_count_main);
2227         blocks_per_seg = sbi->blocks_per_seg;
2228
2229         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2230                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2231                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2232                         return 1;
2233         }
2234         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2235                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2236                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2237                         return 1;
2238         }
2239
2240         if (unlikely(f2fs_cp_error(sbi))) {
2241                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2242                 return 1;
2243         }
2244         return 0;
2245 }
2246
2247 static void init_sb_info(struct f2fs_sb_info *sbi)
2248 {
2249         struct f2fs_super_block *raw_super = sbi->raw_super;
2250         int i, j;
2251
2252         sbi->log_sectors_per_block =
2253                 le32_to_cpu(raw_super->log_sectors_per_block);
2254         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2255         sbi->blocksize = 1 << sbi->log_blocksize;
2256         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2257         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2258         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2259         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2260         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2261         sbi->total_node_count =
2262                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2263                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2264         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2265         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2266         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2267         sbi->cur_victim_sec = NULL_SECNO;
2268         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2269
2270         sbi->dir_level = DEF_DIR_LEVEL;
2271         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2272         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2273         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2274
2275         for (i = 0; i < NR_COUNT_TYPE; i++)
2276                 atomic_set(&sbi->nr_pages[i], 0);
2277
2278         atomic_set(&sbi->wb_sync_req, 0);
2279
2280         INIT_LIST_HEAD(&sbi->s_list);
2281         mutex_init(&sbi->umount_mutex);
2282         for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2283                 for (j = HOT; j < NR_TEMP_TYPE; j++)
2284                         mutex_init(&sbi->wio_mutex[i][j]);
2285         spin_lock_init(&sbi->cp_lock);
2286
2287         sbi->dirty_device = 0;
2288         spin_lock_init(&sbi->dev_lock);
2289
2290         init_rwsem(&sbi->sb_lock);
2291 }
2292
2293 static int init_percpu_info(struct f2fs_sb_info *sbi)
2294 {
2295         int err;
2296
2297         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2298         if (err)
2299                 return err;
2300
2301         return percpu_counter_init(&sbi->total_valid_inode_count, 0,
2302                                                                 GFP_KERNEL);
2303 }
2304
2305 #ifdef CONFIG_BLK_DEV_ZONED
2306 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2307 {
2308         struct block_device *bdev = FDEV(devi).bdev;
2309         sector_t nr_sectors = bdev->bd_part->nr_sects;
2310         sector_t sector = 0;
2311         struct blk_zone *zones;
2312         unsigned int i, nr_zones;
2313         unsigned int n = 0;
2314         int err = -EIO;
2315
2316         if (!f2fs_sb_has_blkzoned(sbi->sb))
2317                 return 0;
2318
2319         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2320                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2321                 return -EINVAL;
2322         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2323         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2324                                 __ilog2_u32(sbi->blocks_per_blkz))
2325                 return -EINVAL;
2326         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2327         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2328                                         sbi->log_blocks_per_blkz;
2329         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2330                 FDEV(devi).nr_blkz++;
2331
2332         FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2333                                                                 GFP_KERNEL);
2334         if (!FDEV(devi).blkz_type)
2335                 return -ENOMEM;
2336
2337 #define F2FS_REPORT_NR_ZONES   4096
2338
2339         zones = f2fs_kzalloc(sbi, sizeof(struct blk_zone) *
2340                                 F2FS_REPORT_NR_ZONES, GFP_KERNEL);
2341         if (!zones)
2342                 return -ENOMEM;
2343
2344         /* Get block zones type */
2345         while (zones && sector < nr_sectors) {
2346
2347                 nr_zones = F2FS_REPORT_NR_ZONES;
2348                 err = blkdev_report_zones(bdev, sector,
2349                                           zones, &nr_zones,
2350                                           GFP_KERNEL);
2351                 if (err)
2352                         break;
2353                 if (!nr_zones) {
2354                         err = -EIO;
2355                         break;
2356                 }
2357
2358                 for (i = 0; i < nr_zones; i++) {
2359                         FDEV(devi).blkz_type[n] = zones[i].type;
2360                         sector += zones[i].len;
2361                         n++;
2362                 }
2363         }
2364
2365         kfree(zones);
2366
2367         return err;
2368 }
2369 #endif
2370
2371 /*
2372  * Read f2fs raw super block.
2373  * Because we have two copies of super block, so read both of them
2374  * to get the first valid one. If any one of them is broken, we pass
2375  * them recovery flag back to the caller.
2376  */
2377 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2378                         struct f2fs_super_block **raw_super,
2379                         int *valid_super_block, int *recovery)
2380 {
2381         struct super_block *sb = sbi->sb;
2382         int block;
2383         struct buffer_head *bh;
2384         struct f2fs_super_block *super;
2385         int err = 0;
2386
2387         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2388         if (!super)
2389                 return -ENOMEM;
2390
2391         for (block = 0; block < 2; block++) {
2392                 bh = sb_bread(sb, block);
2393                 if (!bh) {
2394                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2395                                 block + 1);
2396                         err = -EIO;
2397                         continue;
2398                 }
2399
2400                 /* sanity checking of raw super */
2401                 if (sanity_check_raw_super(sbi, bh)) {
2402                         f2fs_msg(sb, KERN_ERR,
2403                                 "Can't find valid F2FS filesystem in %dth superblock",
2404                                 block + 1);
2405                         err = -EINVAL;
2406                         brelse(bh);
2407                         continue;
2408                 }
2409
2410                 if (!*raw_super) {
2411                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2412                                                         sizeof(*super));
2413                         *valid_super_block = block;
2414                         *raw_super = super;
2415                 }
2416                 brelse(bh);
2417         }
2418
2419         /* Fail to read any one of the superblocks*/
2420         if (err < 0)
2421                 *recovery = 1;
2422
2423         /* No valid superblock */
2424         if (!*raw_super)
2425                 kfree(super);
2426         else
2427                 err = 0;
2428
2429         return err;
2430 }
2431
2432 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2433 {
2434         struct buffer_head *bh;
2435         int err;
2436
2437         if ((recover && f2fs_readonly(sbi->sb)) ||
2438                                 bdev_read_only(sbi->sb->s_bdev)) {
2439                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2440                 return -EROFS;
2441         }
2442
2443         /* write back-up superblock first */
2444         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2445         if (!bh)
2446                 return -EIO;
2447         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2448         brelse(bh);
2449
2450         /* if we are in recovery path, skip writing valid superblock */
2451         if (recover || err)
2452                 return err;
2453
2454         /* write current valid superblock */
2455         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2456         if (!bh)
2457                 return -EIO;
2458         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2459         brelse(bh);
2460         return err;
2461 }
2462
2463 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2464 {
2465         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2466         unsigned int max_devices = MAX_DEVICES;
2467         int i;
2468
2469         /* Initialize single device information */
2470         if (!RDEV(0).path[0]) {
2471                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2472                         return 0;
2473                 max_devices = 1;
2474         }
2475
2476         /*
2477          * Initialize multiple devices information, or single
2478          * zoned block device information.
2479          */
2480         sbi->devs = f2fs_kzalloc(sbi, sizeof(struct f2fs_dev_info) *
2481                                                 max_devices, GFP_KERNEL);
2482         if (!sbi->devs)
2483                 return -ENOMEM;
2484
2485         for (i = 0; i < max_devices; i++) {
2486
2487                 if (i > 0 && !RDEV(i).path[0])
2488                         break;
2489
2490                 if (max_devices == 1) {
2491                         /* Single zoned block device mount */
2492                         FDEV(0).bdev =
2493                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2494                                         sbi->sb->s_mode, sbi->sb->s_type);
2495                 } else {
2496                         /* Multi-device mount */
2497                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2498                         FDEV(i).total_segments =
2499                                 le32_to_cpu(RDEV(i).total_segments);
2500                         if (i == 0) {
2501                                 FDEV(i).start_blk = 0;
2502                                 FDEV(i).end_blk = FDEV(i).start_blk +
2503                                     (FDEV(i).total_segments <<
2504                                     sbi->log_blocks_per_seg) - 1 +
2505                                     le32_to_cpu(raw_super->segment0_blkaddr);
2506                         } else {
2507                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2508                                 FDEV(i).end_blk = FDEV(i).start_blk +
2509                                         (FDEV(i).total_segments <<
2510                                         sbi->log_blocks_per_seg) - 1;
2511                         }
2512                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2513                                         sbi->sb->s_mode, sbi->sb->s_type);
2514                 }
2515                 if (IS_ERR(FDEV(i).bdev))
2516                         return PTR_ERR(FDEV(i).bdev);
2517
2518                 /* to release errored devices */
2519                 sbi->s_ndevs = i + 1;
2520
2521 #ifdef CONFIG_BLK_DEV_ZONED
2522                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2523                                 !f2fs_sb_has_blkzoned(sbi->sb)) {
2524                         f2fs_msg(sbi->sb, KERN_ERR,
2525                                 "Zoned block device feature not enabled\n");
2526                         return -EINVAL;
2527                 }
2528                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2529                         if (init_blkz_info(sbi, i)) {
2530                                 f2fs_msg(sbi->sb, KERN_ERR,
2531                                         "Failed to initialize F2FS blkzone information");
2532                                 return -EINVAL;
2533                         }
2534                         if (max_devices == 1)
2535                                 break;
2536                         f2fs_msg(sbi->sb, KERN_INFO,
2537                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2538                                 i, FDEV(i).path,
2539                                 FDEV(i).total_segments,
2540                                 FDEV(i).start_blk, FDEV(i).end_blk,
2541                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2542                                 "Host-aware" : "Host-managed");
2543                         continue;
2544                 }
2545 #endif
2546                 f2fs_msg(sbi->sb, KERN_INFO,
2547                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2548                                 i, FDEV(i).path,
2549                                 FDEV(i).total_segments,
2550                                 FDEV(i).start_blk, FDEV(i).end_blk);
2551         }
2552         f2fs_msg(sbi->sb, KERN_INFO,
2553                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2554         return 0;
2555 }
2556
2557 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
2558 {
2559         struct f2fs_sm_info *sm_i = SM_I(sbi);
2560
2561         /* adjust parameters according to the volume size */
2562         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
2563                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2564                 sm_i->dcc_info->discard_granularity = 1;
2565                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
2566         }
2567 }
2568
2569 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2570 {
2571         struct f2fs_sb_info *sbi;
2572         struct f2fs_super_block *raw_super;
2573         struct inode *root;
2574         int err;
2575         bool retry = true, need_fsck = false;
2576         char *options = NULL;
2577         int recovery, i, valid_super_block;
2578         struct curseg_info *seg_i;
2579
2580 try_onemore:
2581         err = -EINVAL;
2582         raw_super = NULL;
2583         valid_super_block = -1;
2584         recovery = 0;
2585
2586         /* allocate memory for f2fs-specific super block info */
2587         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2588         if (!sbi)
2589                 return -ENOMEM;
2590
2591         sbi->sb = sb;
2592
2593         /* Load the checksum driver */
2594         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2595         if (IS_ERR(sbi->s_chksum_driver)) {
2596                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2597                 err = PTR_ERR(sbi->s_chksum_driver);
2598                 sbi->s_chksum_driver = NULL;
2599                 goto free_sbi;
2600         }
2601
2602         /* set a block size */
2603         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2604                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2605                 goto free_sbi;
2606         }
2607
2608         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2609                                                                 &recovery);
2610         if (err)
2611                 goto free_sbi;
2612
2613         sb->s_fs_info = sbi;
2614         sbi->raw_super = raw_super;
2615
2616         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2617         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2618
2619         /* precompute checksum seed for metadata */
2620         if (f2fs_sb_has_inode_chksum(sb))
2621                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2622                                                 sizeof(raw_super->uuid));
2623
2624         /*
2625          * The BLKZONED feature indicates that the drive was formatted with
2626          * zone alignment optimization. This is optional for host-aware
2627          * devices, but mandatory for host-managed zoned block devices.
2628          */
2629 #ifndef CONFIG_BLK_DEV_ZONED
2630         if (f2fs_sb_has_blkzoned(sb)) {
2631                 f2fs_msg(sb, KERN_ERR,
2632                          "Zoned block device support is not enabled\n");
2633                 err = -EOPNOTSUPP;
2634                 goto free_sb_buf;
2635         }
2636 #endif
2637         default_options(sbi);
2638         /* parse mount options */
2639         options = kstrdup((const char *)data, GFP_KERNEL);
2640         if (data && !options) {
2641                 err = -ENOMEM;
2642                 goto free_sb_buf;
2643         }
2644
2645         err = parse_options(sb, options);
2646         if (err)
2647                 goto free_options;
2648
2649         sbi->max_file_blocks = max_file_blocks();
2650         sb->s_maxbytes = sbi->max_file_blocks <<
2651                                 le32_to_cpu(raw_super->log_blocksize);
2652         sb->s_max_links = F2FS_LINK_MAX;
2653         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2654
2655 #ifdef CONFIG_QUOTA
2656         sb->dq_op = &f2fs_quota_operations;
2657         if (f2fs_sb_has_quota_ino(sb))
2658                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2659         else
2660                 sb->s_qcop = &f2fs_quotactl_ops;
2661         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2662
2663         if (f2fs_sb_has_quota_ino(sbi->sb)) {
2664                 for (i = 0; i < MAXQUOTAS; i++) {
2665                         if (f2fs_qf_ino(sbi->sb, i))
2666                                 sbi->nquota_files++;
2667                 }
2668         }
2669 #endif
2670
2671         sb->s_op = &f2fs_sops;
2672 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2673         sb->s_cop = &f2fs_cryptops;
2674 #endif
2675         sb->s_xattr = f2fs_xattr_handlers;
2676         sb->s_export_op = &f2fs_export_ops;
2677         sb->s_magic = F2FS_SUPER_MAGIC;
2678         sb->s_time_gran = 1;
2679         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2680                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2681         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2682         sb->s_iflags |= SB_I_CGROUPWB;
2683
2684         /* init f2fs-specific super block info */
2685         sbi->valid_super_block = valid_super_block;
2686         mutex_init(&sbi->gc_mutex);
2687         mutex_init(&sbi->cp_mutex);
2688         init_rwsem(&sbi->node_write);
2689         init_rwsem(&sbi->node_change);
2690
2691         /* disallow all the data/node/meta page writes */
2692         set_sbi_flag(sbi, SBI_POR_DOING);
2693         spin_lock_init(&sbi->stat_lock);
2694
2695         /* init iostat info */
2696         spin_lock_init(&sbi->iostat_lock);
2697         sbi->iostat_enable = false;
2698
2699         for (i = 0; i < NR_PAGE_TYPE; i++) {
2700                 int n = (i == META) ? 1: NR_TEMP_TYPE;
2701                 int j;
2702
2703                 sbi->write_io[i] = f2fs_kmalloc(sbi,
2704                                         n * sizeof(struct f2fs_bio_info),
2705                                         GFP_KERNEL);
2706                 if (!sbi->write_io[i]) {
2707                         err = -ENOMEM;
2708                         goto free_options;
2709                 }
2710
2711                 for (j = HOT; j < n; j++) {
2712                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
2713                         sbi->write_io[i][j].sbi = sbi;
2714                         sbi->write_io[i][j].bio = NULL;
2715                         spin_lock_init(&sbi->write_io[i][j].io_lock);
2716                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2717                 }
2718         }
2719
2720         init_rwsem(&sbi->cp_rwsem);
2721         init_waitqueue_head(&sbi->cp_wait);
2722         init_sb_info(sbi);
2723
2724         err = init_percpu_info(sbi);
2725         if (err)
2726                 goto free_bio_info;
2727
2728         if (F2FS_IO_SIZE(sbi) > 1) {
2729                 sbi->write_io_dummy =
2730                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2731                 if (!sbi->write_io_dummy) {
2732                         err = -ENOMEM;
2733                         goto free_percpu;
2734                 }
2735         }
2736
2737         /* get an inode for meta space */
2738         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2739         if (IS_ERR(sbi->meta_inode)) {
2740                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2741                 err = PTR_ERR(sbi->meta_inode);
2742                 goto free_io_dummy;
2743         }
2744
2745         err = get_valid_checkpoint(sbi);
2746         if (err) {
2747                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2748                 goto free_meta_inode;
2749         }
2750
2751         /* Initialize device list */
2752         err = f2fs_scan_devices(sbi);
2753         if (err) {
2754                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2755                 goto free_devices;
2756         }
2757
2758         sbi->total_valid_node_count =
2759                                 le32_to_cpu(sbi->ckpt->valid_node_count);
2760         percpu_counter_set(&sbi->total_valid_inode_count,
2761                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
2762         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2763         sbi->total_valid_block_count =
2764                                 le64_to_cpu(sbi->ckpt->valid_block_count);
2765         sbi->last_valid_block_count = sbi->total_valid_block_count;
2766         sbi->reserved_blocks = 0;
2767         sbi->current_reserved_blocks = 0;
2768         limit_reserve_root(sbi);
2769
2770         for (i = 0; i < NR_INODE_TYPE; i++) {
2771                 INIT_LIST_HEAD(&sbi->inode_list[i]);
2772                 spin_lock_init(&sbi->inode_lock[i]);
2773         }
2774
2775         init_extent_cache_info(sbi);
2776
2777         init_ino_entry_info(sbi);
2778
2779         /* setup f2fs internal modules */
2780         err = build_segment_manager(sbi);
2781         if (err) {
2782                 f2fs_msg(sb, KERN_ERR,
2783                         "Failed to initialize F2FS segment manager");
2784                 goto free_sm;
2785         }
2786         err = build_node_manager(sbi);
2787         if (err) {
2788                 f2fs_msg(sb, KERN_ERR,
2789                         "Failed to initialize F2FS node manager");
2790                 goto free_nm;
2791         }
2792
2793         /* For write statistics */
2794         if (sb->s_bdev->bd_part)
2795                 sbi->sectors_written_start =
2796                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2797
2798         /* Read accumulated write IO statistics if exists */
2799         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2800         if (__exist_node_summaries(sbi))
2801                 sbi->kbytes_written =
2802                         le64_to_cpu(seg_i->journal->info.kbytes_written);
2803
2804         build_gc_manager(sbi);
2805
2806         /* get an inode for node space */
2807         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2808         if (IS_ERR(sbi->node_inode)) {
2809                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2810                 err = PTR_ERR(sbi->node_inode);
2811                 goto free_nm;
2812         }
2813
2814         err = f2fs_build_stats(sbi);
2815         if (err)
2816                 goto free_node_inode;
2817
2818         /* read root inode and dentry */
2819         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2820         if (IS_ERR(root)) {
2821                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2822                 err = PTR_ERR(root);
2823                 goto free_stats;
2824         }
2825         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2826                 iput(root);
2827                 err = -EINVAL;
2828                 goto free_node_inode;
2829         }
2830
2831         sb->s_root = d_make_root(root); /* allocate root dentry */
2832         if (!sb->s_root) {
2833                 err = -ENOMEM;
2834                 goto free_root_inode;
2835         }
2836
2837         err = f2fs_register_sysfs(sbi);
2838         if (err)
2839                 goto free_root_inode;
2840
2841 #ifdef CONFIG_QUOTA
2842         /*
2843          * Turn on quotas which were not enabled for read-only mounts if
2844          * filesystem has quota feature, so that they are updated correctly.
2845          */
2846         if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb)) {
2847                 err = f2fs_enable_quotas(sb);
2848                 if (err) {
2849                         f2fs_msg(sb, KERN_ERR,
2850                                 "Cannot turn on quotas: error %d", err);
2851                         goto free_sysfs;
2852                 }
2853         }
2854 #endif
2855         /* if there are nt orphan nodes free them */
2856         err = recover_orphan_inodes(sbi);
2857         if (err)
2858                 goto free_meta;
2859
2860         /* recover fsynced data */
2861         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2862                 /*
2863                  * mount should be failed, when device has readonly mode, and
2864                  * previous checkpoint was not done by clean system shutdown.
2865                  */
2866                 if (bdev_read_only(sb->s_bdev) &&
2867                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2868                         err = -EROFS;
2869                         goto free_meta;
2870                 }
2871
2872                 if (need_fsck)
2873                         set_sbi_flag(sbi, SBI_NEED_FSCK);
2874
2875                 if (!retry)
2876                         goto skip_recovery;
2877
2878                 err = recover_fsync_data(sbi, false);
2879                 if (err < 0) {
2880                         need_fsck = true;
2881                         f2fs_msg(sb, KERN_ERR,
2882                                 "Cannot recover all fsync data errno=%d", err);
2883                         goto free_meta;
2884                 }
2885         } else {
2886                 err = recover_fsync_data(sbi, true);
2887
2888                 if (!f2fs_readonly(sb) && err > 0) {
2889                         err = -EINVAL;
2890                         f2fs_msg(sb, KERN_ERR,
2891                                 "Need to recover fsync data");
2892                         goto free_meta;
2893                 }
2894         }
2895 skip_recovery:
2896         /* recover_fsync_data() cleared this already */
2897         clear_sbi_flag(sbi, SBI_POR_DOING);
2898
2899         /*
2900          * If filesystem is not mounted as read-only then
2901          * do start the gc_thread.
2902          */
2903         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2904                 /* After POR, we can run background GC thread.*/
2905                 err = start_gc_thread(sbi);
2906                 if (err)
2907                         goto free_meta;
2908         }
2909         kfree(options);
2910
2911         /* recover broken superblock */
2912         if (recovery) {
2913                 err = f2fs_commit_super(sbi, true);
2914                 f2fs_msg(sb, KERN_INFO,
2915                         "Try to recover %dth superblock, ret: %d",
2916                         sbi->valid_super_block ? 1 : 2, err);
2917         }
2918
2919         f2fs_join_shrinker(sbi);
2920
2921         f2fs_tuning_parameters(sbi);
2922
2923         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2924                                 cur_cp_version(F2FS_CKPT(sbi)));
2925         f2fs_update_time(sbi, CP_TIME);
2926         f2fs_update_time(sbi, REQ_TIME);
2927         return 0;
2928
2929 free_meta:
2930 #ifdef CONFIG_QUOTA
2931         if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb))
2932                 f2fs_quota_off_umount(sbi->sb);
2933 #endif
2934         f2fs_sync_inode_meta(sbi);
2935         /*
2936          * Some dirty meta pages can be produced by recover_orphan_inodes()
2937          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2938          * followed by write_checkpoint() through f2fs_write_node_pages(), which
2939          * falls into an infinite loop in sync_meta_pages().
2940          */
2941         truncate_inode_pages_final(META_MAPPING(sbi));
2942 #ifdef CONFIG_QUOTA
2943 free_sysfs:
2944 #endif
2945         f2fs_unregister_sysfs(sbi);
2946 free_root_inode:
2947         dput(sb->s_root);
2948         sb->s_root = NULL;
2949 free_stats:
2950         f2fs_destroy_stats(sbi);
2951 free_node_inode:
2952         release_ino_entry(sbi, true);
2953         truncate_inode_pages_final(NODE_MAPPING(sbi));
2954         iput(sbi->node_inode);
2955 free_nm:
2956         destroy_node_manager(sbi);
2957 free_sm:
2958         destroy_segment_manager(sbi);
2959 free_devices:
2960         destroy_device_list(sbi);
2961         kfree(sbi->ckpt);
2962 free_meta_inode:
2963         make_bad_inode(sbi->meta_inode);
2964         iput(sbi->meta_inode);
2965 free_io_dummy:
2966         mempool_destroy(sbi->write_io_dummy);
2967 free_percpu:
2968         destroy_percpu_info(sbi);
2969 free_bio_info:
2970         for (i = 0; i < NR_PAGE_TYPE; i++)
2971                 kfree(sbi->write_io[i]);
2972 free_options:
2973 #ifdef CONFIG_QUOTA
2974         for (i = 0; i < MAXQUOTAS; i++)
2975                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2976 #endif
2977         kfree(options);
2978 free_sb_buf:
2979         kfree(raw_super);
2980 free_sbi:
2981         if (sbi->s_chksum_driver)
2982                 crypto_free_shash(sbi->s_chksum_driver);
2983         kfree(sbi);
2984
2985         /* give only one another chance */
2986         if (retry) {
2987                 retry = false;
2988                 shrink_dcache_sb(sb);
2989                 goto try_onemore;
2990         }
2991         return err;
2992 }
2993
2994 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2995                         const char *dev_name, void *data)
2996 {
2997         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2998 }
2999
3000 static void kill_f2fs_super(struct super_block *sb)
3001 {
3002         if (sb->s_root) {
3003                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
3004                 stop_gc_thread(F2FS_SB(sb));
3005                 stop_discard_thread(F2FS_SB(sb));
3006         }
3007         kill_block_super(sb);
3008 }
3009
3010 static struct file_system_type f2fs_fs_type = {
3011         .owner          = THIS_MODULE,
3012         .name           = "f2fs",
3013         .mount          = f2fs_mount,
3014         .kill_sb        = kill_f2fs_super,
3015         .fs_flags       = FS_REQUIRES_DEV,
3016 };
3017 MODULE_ALIAS_FS("f2fs");
3018
3019 static int __init init_inodecache(void)
3020 {
3021         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3022                         sizeof(struct f2fs_inode_info), 0,
3023                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3024         if (!f2fs_inode_cachep)
3025                 return -ENOMEM;
3026         return 0;
3027 }
3028
3029 static void destroy_inodecache(void)
3030 {
3031         /*
3032          * Make sure all delayed rcu free inodes are flushed before we
3033          * destroy cache.
3034          */
3035         rcu_barrier();
3036         kmem_cache_destroy(f2fs_inode_cachep);
3037 }
3038
3039 static int __init init_f2fs_fs(void)
3040 {
3041         int err;
3042
3043         f2fs_build_trace_ios();
3044
3045         err = init_inodecache();
3046         if (err)
3047                 goto fail;
3048         err = create_node_manager_caches();
3049         if (err)
3050                 goto free_inodecache;
3051         err = create_segment_manager_caches();
3052         if (err)
3053                 goto free_node_manager_caches;
3054         err = create_checkpoint_caches();
3055         if (err)
3056                 goto free_segment_manager_caches;
3057         err = create_extent_cache();
3058         if (err)
3059                 goto free_checkpoint_caches;
3060         err = f2fs_init_sysfs();
3061         if (err)
3062                 goto free_extent_cache;
3063         err = register_shrinker(&f2fs_shrinker_info);
3064         if (err)
3065                 goto free_sysfs;
3066         err = register_filesystem(&f2fs_fs_type);
3067         if (err)
3068                 goto free_shrinker;
3069         err = f2fs_create_root_stats();
3070         if (err)
3071                 goto free_filesystem;
3072         return 0;
3073
3074 free_filesystem:
3075         unregister_filesystem(&f2fs_fs_type);
3076 free_shrinker:
3077         unregister_shrinker(&f2fs_shrinker_info);
3078 free_sysfs:
3079         f2fs_exit_sysfs();
3080 free_extent_cache:
3081         destroy_extent_cache();
3082 free_checkpoint_caches:
3083         destroy_checkpoint_caches();
3084 free_segment_manager_caches:
3085         destroy_segment_manager_caches();
3086 free_node_manager_caches:
3087         destroy_node_manager_caches();
3088 free_inodecache:
3089         destroy_inodecache();
3090 fail:
3091         return err;
3092 }
3093
3094 static void __exit exit_f2fs_fs(void)
3095 {
3096         f2fs_destroy_root_stats();
3097         unregister_filesystem(&f2fs_fs_type);
3098         unregister_shrinker(&f2fs_shrinker_info);
3099         f2fs_exit_sysfs();
3100         destroy_extent_cache();
3101         destroy_checkpoint_caches();
3102         destroy_segment_manager_caches();
3103         destroy_node_manager_caches();
3104         destroy_inodecache();
3105         f2fs_destroy_trace_ios();
3106 }
3107
3108 module_init(init_f2fs_fs)
3109 module_exit(exit_f2fs_fs)
3110
3111 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3112 MODULE_DESCRIPTION("Flash Friendly File System");
3113 MODULE_LICENSE("GPL");
3114