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/f2fs_fs.h>
  26 #include <linux/sysfs.h>
  27
  28 #include "f2fs.h"
  29 #include "node.h"
  30 #include "segment.h"
  31 #include "xattr.h"
  32 #include "gc.h"
  33 #include "trace.h"
  34
  35 #define CREATE_TRACE_POINTS
  36 #include <trace/events/f2fs.h>
  37
  38 static struct proc_dir_entry *f2fs_proc_root;
  39 static struct kmem_cache *f2fs_inode_cachep;
  40 static struct kset *f2fs_kset;
  41
  42 #ifdef CONFIG_F2FS_FAULT_INJECTION
  43 struct f2fs_fault_info f2fs_fault;
  44
  45 char *fault_name[FAULT_MAX] = {
  46         [FAULT_KMALLOC]         = "kmalloc",
  47         [FAULT_PAGE_ALLOC]      = "page alloc",
  48         [FAULT_ALLOC_NID]       = "alloc nid",
  49         [FAULT_ORPHAN]          = "orphan",
  50         [FAULT_BLOCK]           = "no more block",
  51         [FAULT_DIR_DEPTH]       = "too big dir depth",
  52         [FAULT_EVICT_INODE]     = "evict_inode fail",
  53 };
  54
  55 static void f2fs_build_fault_attr(unsigned int rate)
  56 {
  57         if (rate) {
  58                 atomic_set(&f2fs_fault.inject_ops, 0);
  59                 f2fs_fault.inject_rate = rate;
  60                 f2fs_fault.inject_type = (1 << FAULT_MAX) - 1;
  61         } else {
  62                 memset(&f2fs_fault, 0, sizeof(struct f2fs_fault_info));
  63         }
  64 }
  65 #endif
  66
  67 /* f2fs-wide shrinker description */
  68 static struct shrinker f2fs_shrinker_info = {
  69         .scan_objects = f2fs_shrink_scan,
  70         .count_objects = f2fs_shrink_count,
  71         .seeks = DEFAULT_SEEKS,
  72 };
  73
  74 enum {
  75         Opt_gc_background,
  76         Opt_disable_roll_forward,
  77         Opt_norecovery,
  78         Opt_discard,
  79         Opt_nodiscard,
  80         Opt_noheap,
  81         Opt_user_xattr,
  82         Opt_nouser_xattr,
  83         Opt_acl,
  84         Opt_noacl,
  85         Opt_active_logs,
  86         Opt_disable_ext_identify,
  87         Opt_inline_xattr,
  88         Opt_inline_data,
  89         Opt_inline_dentry,
  90         Opt_flush_merge,
  91         Opt_noflush_merge,
  92         Opt_nobarrier,
  93         Opt_fastboot,
  94         Opt_extent_cache,
  95         Opt_noextent_cache,
  96         Opt_noinline_data,
  97         Opt_data_flush,
  98         Opt_mode,
  99         Opt_fault_injection,
 100         Opt_lazytime,
 101         Opt_nolazytime,
 102         Opt_err,
 103 };
 104
 105 static match_table_t f2fs_tokens = {
 106         {Opt_gc_background, "background_gc=%s"},
 107         {Opt_disable_roll_forward, "disable_roll_forward"},
 108         {Opt_norecovery, "norecovery"},
 109         {Opt_discard, "discard"},
 110         {Opt_nodiscard, "nodiscard"},
 111         {Opt_noheap, "no_heap"},
 112         {Opt_user_xattr, "user_xattr"},
 113         {Opt_nouser_xattr, "nouser_xattr"},
 114         {Opt_acl, "acl"},
 115         {Opt_noacl, "noacl"},
 116         {Opt_active_logs, "active_logs=%u"},
 117         {Opt_disable_ext_identify, "disable_ext_identify"},
 118         {Opt_inline_xattr, "inline_xattr"},
 119         {Opt_inline_data, "inline_data"},
 120         {Opt_inline_dentry, "inline_dentry"},
 121         {Opt_flush_merge, "flush_merge"},
 122         {Opt_noflush_merge, "noflush_merge"},
 123         {Opt_nobarrier, "nobarrier"},
 124         {Opt_fastboot, "fastboot"},
 125         {Opt_extent_cache, "extent_cache"},
 126         {Opt_noextent_cache, "noextent_cache"},
 127         {Opt_noinline_data, "noinline_data"},
 128         {Opt_data_flush, "data_flush"},
 129         {Opt_mode, "mode=%s"},
 130         {Opt_fault_injection, "fault_injection=%u"},
 131         {Opt_lazytime, "lazytime"},
 132         {Opt_nolazytime, "nolazytime"},
 133         {Opt_err, NULL},
 134 };
 135
 136 /* Sysfs support for f2fs */
 137 enum {
 138         GC_THREAD,      /* struct f2fs_gc_thread */
 139         SM_INFO,        /* struct f2fs_sm_info */
 140         NM_INFO,        /* struct f2fs_nm_info */
 141         F2FS_SBI,       /* struct f2fs_sb_info */
 142 #ifdef CONFIG_F2FS_FAULT_INJECTION
 143         FAULT_INFO_RATE,        /* struct f2fs_fault_info */
 144         FAULT_INFO_TYPE,        /* struct f2fs_fault_info */
 145 #endif
 146 };
 147
 148 struct f2fs_attr {
 149         struct attribute attr;
 150         ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
 151         ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
 152                          const char *, size_t);
 153         int struct_type;
 154         int offset;
 155 };
 156
 157 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
 158 {
 159         if (struct_type == GC_THREAD)
 160                 return (unsigned char *)sbi->gc_thread;
 161         else if (struct_type == SM_INFO)
 162                 return (unsigned char *)SM_I(sbi);
 163         else if (struct_type == NM_INFO)
 164                 return (unsigned char *)NM_I(sbi);
 165         else if (struct_type == F2FS_SBI)
 166                 return (unsigned char *)sbi;
 167 #ifdef CONFIG_F2FS_FAULT_INJECTION
 168         else if (struct_type == FAULT_INFO_RATE ||
 169                                         struct_type == FAULT_INFO_TYPE)
 170                 return (unsigned char *)&f2fs_fault;
 171 #endif
 172         return NULL;
 173 }
 174
 175 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
 176                 struct f2fs_sb_info *sbi, char *buf)
 177 {
 178         struct super_block *sb = sbi->sb;
 179
 180         if (!sb->s_bdev->bd_part)
 181                 return snprintf(buf, PAGE_SIZE, "0\n");
 182
 183         return snprintf(buf, PAGE_SIZE, "%llu\n",
 184                 (unsigned long long)(sbi->kbytes_written +
 185                         BD_PART_WRITTEN(sbi)));
 186 }
 187
 188 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
 189                         struct f2fs_sb_info *sbi, char *buf)
 190 {
 191         unsigned char *ptr = NULL;
 192         unsigned int *ui;
 193
 194         ptr = __struct_ptr(sbi, a->struct_type);
 195         if (!ptr)
 196                 return -EINVAL;
 197
 198         ui = (unsigned int *)(ptr + a->offset);
 199
 200         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
 201 }
 202
 203 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
 204                         struct f2fs_sb_info *sbi,
 205                         const char *buf, size_t count)
 206 {
 207         unsigned char *ptr;
 208         unsigned long t;
 209         unsigned int *ui;
 210         ssize_t ret;
 211
 212         ptr = __struct_ptr(sbi, a->struct_type);
 213         if (!ptr)
 214                 return -EINVAL;
 215
 216         ui = (unsigned int *)(ptr + a->offset);
 217
 218         ret = kstrtoul(skip_spaces(buf), 0, &t);
 219         if (ret < 0)
 220                 return ret;
 221 #ifdef CONFIG_F2FS_FAULT_INJECTION
 222         if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
 223                 return -EINVAL;
 224 #endif
 225         *ui = t;
 226         return count;
 227 }
 228
 229 static ssize_t f2fs_attr_show(struct kobject *kobj,
 230                                 struct attribute *attr, char *buf)
 231 {
 232         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 233                                                                 s_kobj);
 234         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 235
 236         return a->show ? a->show(a, sbi, buf) : 0;
 237 }
 238
 239 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
 240                                                 const char *buf, size_t len)
 241 {
 242         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 243                                                                         s_kobj);
 244         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 245
 246         return a->store ? a->store(a, sbi, buf, len) : 0;
 247 }
 248
 249 static void f2fs_sb_release(struct kobject *kobj)
 250 {
 251         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 252                                                                 s_kobj);
 253         complete(&sbi->s_kobj_unregister);
 254 }
 255
 256 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
 257 static struct f2fs_attr f2fs_attr_##_name = {                   \
 258         .attr = {.name = __stringify(_name), .mode = _mode },   \
 259         .show   = _show,                                        \
 260         .store  = _store,                                       \
 261         .struct_type = _struct_type,                            \
 262         .offset = _offset                                       \
 263 }
 264
 265 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)    \
 266         F2FS_ATTR_OFFSET(struct_type, name, 0644,               \
 267                 f2fs_sbi_show, f2fs_sbi_store,                  \
 268                 offsetof(struct struct_name, elname))
 269
 270 #define F2FS_GENERAL_RO_ATTR(name) \
 271 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
 272
 273 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
 274 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
 275 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
 276 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
 277 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
 278 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
 279 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
 280 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
 281 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
 282 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
 283 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
 284 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
 285 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
 286 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
 287 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
 288 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
 289 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
 290 #ifdef CONFIG_F2FS_FAULT_INJECTION
 291 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
 292 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
 293 #endif
 294 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
 295
 296 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 297 static struct attribute *f2fs_attrs[] = {
 298         ATTR_LIST(gc_min_sleep_time),
 299         ATTR_LIST(gc_max_sleep_time),
 300         ATTR_LIST(gc_no_gc_sleep_time),
 301         ATTR_LIST(gc_idle),
 302         ATTR_LIST(reclaim_segments),
 303         ATTR_LIST(max_small_discards),
 304         ATTR_LIST(batched_trim_sections),
 305         ATTR_LIST(ipu_policy),
 306         ATTR_LIST(min_ipu_util),
 307         ATTR_LIST(min_fsync_blocks),
 308         ATTR_LIST(max_victim_search),
 309         ATTR_LIST(dir_level),
 310         ATTR_LIST(ram_thresh),
 311         ATTR_LIST(ra_nid_pages),
 312         ATTR_LIST(dirty_nats_ratio),
 313         ATTR_LIST(cp_interval),
 314         ATTR_LIST(idle_interval),
 315         ATTR_LIST(lifetime_write_kbytes),
 316         NULL,
 317 };
 318
 319 static const struct sysfs_ops f2fs_attr_ops = {
 320         .show   = f2fs_attr_show,
 321         .store  = f2fs_attr_store,
 322 };
 323
 324 static struct kobj_type f2fs_ktype = {
 325         .default_attrs  = f2fs_attrs,
 326         .sysfs_ops      = &f2fs_attr_ops,
 327         .release        = f2fs_sb_release,
 328 };
 329
 330 #ifdef CONFIG_F2FS_FAULT_INJECTION
 331 /* sysfs for f2fs fault injection */
 332 static struct kobject f2fs_fault_inject;
 333
 334 static struct attribute *f2fs_fault_attrs[] = {
 335         ATTR_LIST(inject_rate),
 336         ATTR_LIST(inject_type),
 337         NULL
 338 };
 339
 340 static struct kobj_type f2fs_fault_ktype = {
 341         .default_attrs  = f2fs_fault_attrs,
 342         .sysfs_ops      = &f2fs_attr_ops,
 343 };
 344 #endif
 345
 346 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 347 {
 348         struct va_format vaf;
 349         va_list args;
 350
 351         va_start(args, fmt);
 352         vaf.fmt = fmt;
 353         vaf.va = &args;
 354         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 355         va_end(args);
 356 }
 357
 358 static void init_once(void *foo)
 359 {
 360         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 361
 362         inode_init_once(&fi->vfs_inode);
 363 }
 364
 365 static int parse_options(struct super_block *sb, char *options)
 366 {
 367         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 368         struct request_queue *q;
 369         substring_t args[MAX_OPT_ARGS];
 370         char *p, *name;
 371         int arg = 0;
 372
 373 #ifdef CONFIG_F2FS_FAULT_INJECTION
 374         f2fs_build_fault_attr(0);
 375 #endif
 376
 377         if (!options)
 378                 return 0;
 379
 380         while ((p = strsep(&options, ",")) != NULL) {
 381                 int token;
 382                 if (!*p)
 383                         continue;
 384                 /*
 385                  * Initialize args struct so we know whether arg was
 386                  * found; some options take optional arguments.
 387                  */
 388                 args[0].to = args[0].from = NULL;
 389                 token = match_token(p, f2fs_tokens, args);
 390
 391                 switch (token) {
 392                 case Opt_gc_background:
 393                         name = match_strdup(&args[0]);
 394
 395                         if (!name)
 396                                 return -ENOMEM;
 397                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 398                                 set_opt(sbi, BG_GC);
 399                                 clear_opt(sbi, FORCE_FG_GC);
 400                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 401                                 clear_opt(sbi, BG_GC);
 402                                 clear_opt(sbi, FORCE_FG_GC);
 403                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 404                                 set_opt(sbi, BG_GC);
 405                                 set_opt(sbi, FORCE_FG_GC);
 406                         } else {
 407                                 kfree(name);
 408                                 return -EINVAL;
 409                         }
 410                         kfree(name);
 411                         break;
 412                 case Opt_disable_roll_forward:
 413                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 414                         break;
 415                 case Opt_norecovery:
 416                         /* this option mounts f2fs with ro */
 417                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 418                         if (!f2fs_readonly(sb))
 419                                 return -EINVAL;
 420                         break;
 421                 case Opt_discard:
 422                         q = bdev_get_queue(sb->s_bdev);
 423                         if (blk_queue_discard(q)) {
 424                                 set_opt(sbi, DISCARD);
 425                         } else {
 426                                 f2fs_msg(sb, KERN_WARNING,
 427                                         "mounting with \"discard\" option, but "
 428                                         "the device does not support discard");
 429                         }
 430                         break;
 431                 case Opt_nodiscard:
 432                         clear_opt(sbi, DISCARD);
 433                 case Opt_noheap:
 434                         set_opt(sbi, NOHEAP);
 435                         break;
 436 #ifdef CONFIG_F2FS_FS_XATTR
 437                 case Opt_user_xattr:
 438                         set_opt(sbi, XATTR_USER);
 439                         break;
 440                 case Opt_nouser_xattr:
 441                         clear_opt(sbi, XATTR_USER);
 442                         break;
 443                 case Opt_inline_xattr:
 444                         set_opt(sbi, INLINE_XATTR);
 445                         break;
 446 #else
 447                 case Opt_user_xattr:
 448                         f2fs_msg(sb, KERN_INFO,
 449                                 "user_xattr options not supported");
 450                         break;
 451                 case Opt_nouser_xattr:
 452                         f2fs_msg(sb, KERN_INFO,
 453                                 "nouser_xattr options not supported");
 454                         break;
 455                 case Opt_inline_xattr:
 456                         f2fs_msg(sb, KERN_INFO,
 457                                 "inline_xattr options not supported");
 458                         break;
 459 #endif
 460 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 461                 case Opt_acl:
 462                         set_opt(sbi, POSIX_ACL);
 463                         break;
 464                 case Opt_noacl:
 465                         clear_opt(sbi, POSIX_ACL);
 466                         break;
 467 #else
 468                 case Opt_acl:
 469                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 470                         break;
 471                 case Opt_noacl:
 472                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 473                         break;
 474 #endif
 475                 case Opt_active_logs:
 476                         if (args->from && match_int(args, &arg))
 477                                 return -EINVAL;
 478                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 479                                 return -EINVAL;
 480                         sbi->active_logs = arg;
 481                         break;
 482                 case Opt_disable_ext_identify:
 483                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 484                         break;
 485                 case Opt_inline_data:
 486                         set_opt(sbi, INLINE_DATA);
 487                         break;
 488                 case Opt_inline_dentry:
 489                         set_opt(sbi, INLINE_DENTRY);
 490                         break;
 491                 case Opt_flush_merge:
 492                         set_opt(sbi, FLUSH_MERGE);
 493                         break;
 494                 case Opt_noflush_merge:
 495                         clear_opt(sbi, FLUSH_MERGE);
 496                         break;
 497                 case Opt_nobarrier:
 498                         set_opt(sbi, NOBARRIER);
 499                         break;
 500                 case Opt_fastboot:
 501                         set_opt(sbi, FASTBOOT);
 502                         break;
 503                 case Opt_extent_cache:
 504                         set_opt(sbi, EXTENT_CACHE);
 505                         break;
 506                 case Opt_noextent_cache:
 507                         clear_opt(sbi, EXTENT_CACHE);
 508                         break;
 509                 case Opt_noinline_data:
 510                         clear_opt(sbi, INLINE_DATA);
 511                         break;
 512                 case Opt_data_flush:
 513                         set_opt(sbi, DATA_FLUSH);
 514                         break;
 515                 case Opt_mode:
 516                         name = match_strdup(&args[0]);
 517
 518                         if (!name)
 519                                 return -ENOMEM;
 520                         if (strlen(name) == 8 &&
 521                                         !strncmp(name, "adaptive", 8)) {
 522                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 523                         } else if (strlen(name) == 3 &&
 524                                         !strncmp(name, "lfs", 3)) {
 525                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 526                         } else {
 527                                 kfree(name);
 528                                 return -EINVAL;
 529                         }
 530                         kfree(name);
 531                         break;
 532                 case Opt_fault_injection:
 533                         if (args->from && match_int(args, &arg))
 534                                 return -EINVAL;
 535 #ifdef CONFIG_F2FS_FAULT_INJECTION
 536                         f2fs_build_fault_attr(arg);
 537 #else
 538                         f2fs_msg(sb, KERN_INFO,
 539                                 "FAULT_INJECTION was not selected");
 540 #endif
 541                         break;
 542                 case Opt_lazytime:
 543                         sb->s_flags |= MS_LAZYTIME;
 544                         break;
 545                 case Opt_nolazytime:
 546                         sb->s_flags &= ~MS_LAZYTIME;
 547                         break;
 548                 default:
 549                         f2fs_msg(sb, KERN_ERR,
 550                                 "Unrecognized mount option \"%s\" or missing value",
 551                                 p);
 552                         return -EINVAL;
 553                 }
 554         }
 555         return 0;
 556 }
 557
 558 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 559 {
 560         struct f2fs_inode_info *fi;
 561
 562         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 563         if (!fi)
 564                 return NULL;
 565
 566         init_once((void *) fi);
 567
 568         if (percpu_counter_init(&fi->dirty_pages, 0, GFP_NOFS)) {
 569                 kmem_cache_free(f2fs_inode_cachep, fi);
 570                 return NULL;
 571         }
 572
 573         /* Initialize f2fs-specific inode info */
 574         fi->vfs_inode.i_version = 1;
 575         fi->i_current_depth = 1;
 576         fi->i_advise = 0;
 577         init_rwsem(&fi->i_sem);
 578         INIT_LIST_HEAD(&fi->dirty_list);
 579         INIT_LIST_HEAD(&fi->gdirty_list);
 580         INIT_LIST_HEAD(&fi->inmem_pages);
 581         mutex_init(&fi->inmem_lock);
 582
 583         /* Will be used by directory only */
 584         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 585         return &fi->vfs_inode;
 586 }
 587
 588 static int f2fs_drop_inode(struct inode *inode)
 589 {
 590         /*
 591          * This is to avoid a deadlock condition like below.
 592          * writeback_single_inode(inode)
 593          *  - f2fs_write_data_page
 594          *    - f2fs_gc -> iput -> evict
 595          *       - inode_wait_for_writeback(inode)
 596          */
 597         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 598                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 599                         /* to avoid evict_inode call simultaneously */
 600                         atomic_inc(&inode->i_count);
 601                         spin_unlock(&inode->i_lock);
 602
 603                         /* some remained atomic pages should discarded */
 604                         if (f2fs_is_atomic_file(inode))
 605                                 drop_inmem_pages(inode);
 606
 607                         /* should remain fi->extent_tree for writepage */
 608                         f2fs_destroy_extent_node(inode);
 609
 610                         sb_start_intwrite(inode->i_sb);
 611                         f2fs_i_size_write(inode, 0);
 612
 613                         if (F2FS_HAS_BLOCKS(inode))
 614                                 f2fs_truncate(inode);
 615
 616                         sb_end_intwrite(inode->i_sb);
 617
 618                         fscrypt_put_encryption_info(inode, NULL);
 619                         spin_lock(&inode->i_lock);
 620                         atomic_dec(&inode->i_count);
 621                 }
 622                 return 0;
 623         }
 624
 625         return generic_drop_inode(inode);
 626 }
 627
 628 /*
 629  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 630  *
 631  * We should call set_dirty_inode to write the dirty inode through write_inode.
 632  */
 633 static void f2fs_dirty_inode(struct inode *inode, int flags)
 634 {
 635         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 636
 637         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 638                         inode->i_ino == F2FS_META_INO(sbi))
 639                 return;
 640
 641         if (flags == I_DIRTY_TIME)
 642                 return;
 643
 644         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 645                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 646
 647         spin_lock(&sbi->inode_lock[DIRTY_META]);
 648         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 649                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 650                 return;
 651         }
 652
 653         set_inode_flag(inode, FI_DIRTY_INODE);
 654         list_add_tail(&F2FS_I(inode)->gdirty_list,
 655                                 &sbi->inode_list[DIRTY_META]);
 656         inc_page_count(sbi, F2FS_DIRTY_IMETA);
 657         stat_inc_dirty_inode(sbi, DIRTY_META);
 658         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 659 }
 660
 661 void f2fs_inode_synced(struct inode *inode)
 662 {
 663         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 664
 665         spin_lock(&sbi->inode_lock[DIRTY_META]);
 666         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 667                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 668                 return;
 669         }
 670         list_del_init(&F2FS_I(inode)->gdirty_list);
 671         clear_inode_flag(inode, FI_DIRTY_INODE);
 672         clear_inode_flag(inode, FI_AUTO_RECOVER);
 673         dec_page_count(sbi, F2FS_DIRTY_IMETA);
 674         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 675         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 676 }
 677
 678 static void f2fs_i_callback(struct rcu_head *head)
 679 {
 680         struct inode *inode = container_of(head, struct inode, i_rcu);
 681         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 682 }
 683
 684 static void f2fs_destroy_inode(struct inode *inode)
 685 {
 686         percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
 687         call_rcu(&inode->i_rcu, f2fs_i_callback);
 688 }
 689
 690 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 691 {
 692         int i;
 693
 694         for (i = 0; i < NR_COUNT_TYPE; i++)
 695                 percpu_counter_destroy(&sbi->nr_pages[i]);
 696         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 697         percpu_counter_destroy(&sbi->total_valid_inode_count);
 698
 699         percpu_free_rwsem(&sbi->cp_rwsem);
 700 }
 701
 702 static void f2fs_put_super(struct super_block *sb)
 703 {
 704         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 705
 706         if (sbi->s_proc) {
 707                 remove_proc_entry("segment_info", sbi->s_proc);
 708                 remove_proc_entry("segment_bits", sbi->s_proc);
 709                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 710         }
 711         kobject_del(&sbi->s_kobj);
 712
 713         stop_gc_thread(sbi);
 714
 715         /* prevent remaining shrinker jobs */
 716         mutex_lock(&sbi->umount_mutex);
 717
 718         /*
 719          * We don't need to do checkpoint when superblock is clean.
 720          * But, the previous checkpoint was not done by umount, it needs to do
 721          * clean checkpoint again.
 722          */
 723         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 724                         !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
 725                 struct cp_control cpc = {
 726                         .reason = CP_UMOUNT,
 727                 };
 728                 write_checkpoint(sbi, &cpc);
 729         }
 730
 731         /* write_checkpoint can update stat informaion */
 732         f2fs_destroy_stats(sbi);
 733
 734         /*
 735          * normally superblock is clean, so we need to release this.
 736          * In addition, EIO will skip do checkpoint, we need this as well.
 737          */
 738         release_ino_entry(sbi, true);
 739         release_discard_addrs(sbi);
 740
 741         f2fs_leave_shrinker(sbi);
 742         mutex_unlock(&sbi->umount_mutex);
 743
 744         /* our cp_error case, we can wait for any writeback page */
 745         f2fs_flush_merged_bios(sbi);
 746
 747         iput(sbi->node_inode);
 748         iput(sbi->meta_inode);
 749
 750         /* destroy f2fs internal modules */
 751         destroy_node_manager(sbi);
 752         destroy_segment_manager(sbi);
 753
 754         kfree(sbi->ckpt);
 755         kobject_put(&sbi->s_kobj);
 756         wait_for_completion(&sbi->s_kobj_unregister);
 757
 758         sb->s_fs_info = NULL;
 759         if (sbi->s_chksum_driver)
 760                 crypto_free_shash(sbi->s_chksum_driver);
 761         kfree(sbi->raw_super);
 762
 763         destroy_percpu_info(sbi);
 764         kfree(sbi);
 765 }
 766
 767 int f2fs_sync_fs(struct super_block *sb, int sync)
 768 {
 769         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 770         int err = 0;
 771
 772         trace_f2fs_sync_fs(sb, sync);
 773
 774         if (sync) {
 775                 struct cp_control cpc;
 776
 777                 cpc.reason = __get_cp_reason(sbi);
 778
 779                 mutex_lock(&sbi->gc_mutex);
 780                 err = write_checkpoint(sbi, &cpc);
 781                 mutex_unlock(&sbi->gc_mutex);
 782         }
 783         f2fs_trace_ios(NULL, 1);
 784
 785         return err;
 786 }
 787
 788 static int f2fs_freeze(struct super_block *sb)
 789 {
 790         int err;
 791
 792         if (f2fs_readonly(sb))
 793                 return 0;
 794
 795         err = f2fs_sync_fs(sb, 1);
 796         return err;
 797 }
 798
 799 static int f2fs_unfreeze(struct super_block *sb)
 800 {
 801         return 0;
 802 }
 803
 804 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 805 {
 806         struct super_block *sb = dentry->d_sb;
 807         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 808         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 809         block_t total_count, user_block_count, start_count, ovp_count;
 810
 811         total_count = le64_to_cpu(sbi->raw_super->block_count);
 812         user_block_count = sbi->user_block_count;
 813         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 814         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 815         buf->f_type = F2FS_SUPER_MAGIC;
 816         buf->f_bsize = sbi->blocksize;
 817
 818         buf->f_blocks = total_count - start_count;
 819         buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
 820         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 821
 822         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 823         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
 824
 825         buf->f_namelen = F2FS_NAME_LEN;
 826         buf->f_fsid.val[0] = (u32)id;
 827         buf->f_fsid.val[1] = (u32)(id >> 32);
 828
 829         return 0;
 830 }
 831
 832 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 833 {
 834         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 835
 836         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
 837                 if (test_opt(sbi, FORCE_FG_GC))
 838                         seq_printf(seq, ",background_gc=%s", "sync");
 839                 else
 840                         seq_printf(seq, ",background_gc=%s", "on");
 841         } else {
 842                 seq_printf(seq, ",background_gc=%s", "off");
 843         }
 844         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 845                 seq_puts(seq, ",disable_roll_forward");
 846         if (test_opt(sbi, DISCARD))
 847                 seq_puts(seq, ",discard");
 848         if (test_opt(sbi, NOHEAP))
 849                 seq_puts(seq, ",no_heap_alloc");
 850 #ifdef CONFIG_F2FS_FS_XATTR
 851         if (test_opt(sbi, XATTR_USER))
 852                 seq_puts(seq, ",user_xattr");
 853         else
 854                 seq_puts(seq, ",nouser_xattr");
 855         if (test_opt(sbi, INLINE_XATTR))
 856                 seq_puts(seq, ",inline_xattr");
 857 #endif
 858 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 859         if (test_opt(sbi, POSIX_ACL))
 860                 seq_puts(seq, ",acl");
 861         else
 862                 seq_puts(seq, ",noacl");
 863 #endif
 864         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 865                 seq_puts(seq, ",disable_ext_identify");
 866         if (test_opt(sbi, INLINE_DATA))
 867                 seq_puts(seq, ",inline_data");
 868         else
 869                 seq_puts(seq, ",noinline_data");
 870         if (test_opt(sbi, INLINE_DENTRY))
 871                 seq_puts(seq, ",inline_dentry");
 872         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 873                 seq_puts(seq, ",flush_merge");
 874         if (test_opt(sbi, NOBARRIER))
 875                 seq_puts(seq, ",nobarrier");
 876         if (test_opt(sbi, FASTBOOT))
 877                 seq_puts(seq, ",fastboot");
 878         if (test_opt(sbi, EXTENT_CACHE))
 879                 seq_puts(seq, ",extent_cache");
 880         else
 881                 seq_puts(seq, ",noextent_cache");
 882         if (test_opt(sbi, DATA_FLUSH))
 883                 seq_puts(seq, ",data_flush");
 884
 885         seq_puts(seq, ",mode=");
 886         if (test_opt(sbi, ADAPTIVE))
 887                 seq_puts(seq, "adaptive");
 888         else if (test_opt(sbi, LFS))
 889                 seq_puts(seq, "lfs");
 890         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 891
 892         return 0;
 893 }
 894
 895 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 896 {
 897         struct super_block *sb = seq->private;
 898         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 899         unsigned int total_segs =
 900                         le32_to_cpu(sbi->raw_super->segment_count_main);
 901         int i;
 902
 903         seq_puts(seq, "format: segment_type|valid_blocks\n"
 904                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 905
 906         for (i = 0; i < total_segs; i++) {
 907                 struct seg_entry *se = get_seg_entry(sbi, i);
 908
 909                 if ((i % 10) == 0)
 910                         seq_printf(seq, "%-10d", i);
 911                 seq_printf(seq, "%d|%-3u", se->type,
 912                                         get_valid_blocks(sbi, i, 1));
 913                 if ((i % 10) == 9 || i == (total_segs - 1))
 914                         seq_putc(seq, '\n');
 915                 else
 916                         seq_putc(seq, ' ');
 917         }
 918
 919         return 0;
 920 }
 921
 922 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
 923 {
 924         struct super_block *sb = seq->private;
 925         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 926         unsigned int total_segs =
 927                         le32_to_cpu(sbi->raw_super->segment_count_main);
 928         int i, j;
 929
 930         seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
 931                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 932
 933         for (i = 0; i < total_segs; i++) {
 934                 struct seg_entry *se = get_seg_entry(sbi, i);
 935
 936                 seq_printf(seq, "%-10d", i);
 937                 seq_printf(seq, "%d|%-3u|", se->type,
 938                                         get_valid_blocks(sbi, i, 1));
 939                 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
 940                         seq_printf(seq, "%x ", se->cur_valid_map[j]);
 941                 seq_putc(seq, '\n');
 942         }
 943         return 0;
 944 }
 945
 946 #define F2FS_PROC_FILE_DEF(_name)                                       \
 947 static int _name##_open_fs(struct inode *inode, struct file *file)      \
 948 {                                                                       \
 949         return single_open(file, _name##_seq_show, PDE_DATA(inode));    \
 950 }                                                                       \
 951                                                                         \
 952 static const struct file_operations f2fs_seq_##_name##_fops = {         \
 953         .owner = THIS_MODULE,                                           \
 954         .open = _name##_open_fs,                                        \
 955         .read = seq_read,                                               \
 956         .llseek = seq_lseek,                                            \
 957         .release = single_release,                                      \
 958 };
 959
 960 F2FS_PROC_FILE_DEF(segment_info);
 961 F2FS_PROC_FILE_DEF(segment_bits);
 962
 963 static void default_options(struct f2fs_sb_info *sbi)
 964 {
 965         /* init some FS parameters */
 966         sbi->active_logs = NR_CURSEG_TYPE;
 967
 968         set_opt(sbi, BG_GC);
 969         set_opt(sbi, INLINE_DATA);
 970         set_opt(sbi, EXTENT_CACHE);
 971         sbi->sb->s_flags |= MS_LAZYTIME;
 972         set_opt(sbi, FLUSH_MERGE);
 973         if (f2fs_sb_mounted_hmsmr(sbi->sb)) {
 974                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 975                 set_opt(sbi, DISCARD);
 976         } else {
 977                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 978         }
 979
 980 #ifdef CONFIG_F2FS_FS_XATTR
 981         set_opt(sbi, XATTR_USER);
 982 #endif
 983 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 984         set_opt(sbi, POSIX_ACL);
 985 #endif
 986 }
 987
 988 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 989 {
 990         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 991         struct f2fs_mount_info org_mount_opt;
 992         int err, active_logs;
 993         bool need_restart_gc = false;
 994         bool need_stop_gc = false;
 995         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
 996
 997         /*
 998          * Save the old mount options in case we
 999          * need to restore them.
1000          */
1001         org_mount_opt = sbi->mount_opt;
1002         active_logs = sbi->active_logs;
1003
1004         /* recover superblocks we couldn't write due to previous RO mount */
1005         if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1006                 err = f2fs_commit_super(sbi, false);
1007                 f2fs_msg(sb, KERN_INFO,
1008                         "Try to recover all the superblocks, ret: %d", err);
1009                 if (!err)
1010                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1011         }
1012
1013         sbi->mount_opt.opt = 0;
1014         default_options(sbi);
1015
1016         /* parse mount options */
1017         err = parse_options(sb, data);
1018         if (err)
1019                 goto restore_opts;
1020
1021         /*
1022          * Previous and new state of filesystem is RO,
1023          * so skip checking GC and FLUSH_MERGE conditions.
1024          */
1025         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1026                 goto skip;
1027
1028         /* disallow enable/disable extent_cache dynamically */
1029         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1030                 err = -EINVAL;
1031                 f2fs_msg(sbi->sb, KERN_WARNING,
1032                                 "switch extent_cache option is not allowed");
1033                 goto restore_opts;
1034         }
1035
1036         /*
1037          * We stop the GC thread if FS is mounted as RO
1038          * or if background_gc = off is passed in mount
1039          * option. Also sync the filesystem.
1040          */
1041         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1042                 if (sbi->gc_thread) {
1043                         stop_gc_thread(sbi);
1044                         need_restart_gc = true;
1045                 }
1046         } else if (!sbi->gc_thread) {
1047                 err = start_gc_thread(sbi);
1048                 if (err)
1049                         goto restore_opts;
1050                 need_stop_gc = true;
1051         }
1052
1053         if (*flags & MS_RDONLY) {
1054                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1055                 sync_inodes_sb(sb);
1056
1057                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1058                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1059                 f2fs_sync_fs(sb, 1);
1060                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1061         }
1062
1063         /*
1064          * We stop issue flush thread if FS is mounted as RO
1065          * or if flush_merge is not passed in mount option.
1066          */
1067         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1068                 destroy_flush_cmd_control(sbi);
1069         } else if (!SM_I(sbi)->cmd_control_info) {
1070                 err = create_flush_cmd_control(sbi);
1071                 if (err)
1072                         goto restore_gc;
1073         }
1074 skip:
1075         /* Update the POSIXACL Flag */
1076         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1077                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1078
1079         return 0;
1080 restore_gc:
1081         if (need_restart_gc) {
1082                 if (start_gc_thread(sbi))
1083                         f2fs_msg(sbi->sb, KERN_WARNING,
1084                                 "background gc thread has stopped");
1085         } else if (need_stop_gc) {
1086                 stop_gc_thread(sbi);
1087         }
1088 restore_opts:
1089         sbi->mount_opt = org_mount_opt;
1090         sbi->active_logs = active_logs;
1091         return err;
1092 }
1093
1094 static struct super_operations f2fs_sops = {
1095         .alloc_inode    = f2fs_alloc_inode,
1096         .drop_inode     = f2fs_drop_inode,
1097         .destroy_inode  = f2fs_destroy_inode,
1098         .write_inode    = f2fs_write_inode,
1099         .dirty_inode    = f2fs_dirty_inode,
1100         .show_options   = f2fs_show_options,
1101         .evict_inode    = f2fs_evict_inode,
1102         .put_super      = f2fs_put_super,
1103         .sync_fs        = f2fs_sync_fs,
1104         .freeze_fs      = f2fs_freeze,
1105         .unfreeze_fs    = f2fs_unfreeze,
1106         .statfs         = f2fs_statfs,
1107         .remount_fs     = f2fs_remount,
1108 };
1109
1110 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1111 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1112 {
1113         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1114                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1115                                 ctx, len, NULL);
1116 }
1117
1118 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1119 {
1120         *key = F2FS_I_SB(inode)->key_prefix;
1121         return F2FS_I_SB(inode)->key_prefix_size;
1122 }
1123
1124 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1125                                                         void *fs_data)
1126 {
1127         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1128                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1129                                 ctx, len, fs_data, XATTR_CREATE);
1130 }
1131
1132 static unsigned f2fs_max_namelen(struct inode *inode)
1133 {
1134         return S_ISLNK(inode->i_mode) ?
1135                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1136 }
1137
1138 static struct fscrypt_operations f2fs_cryptops = {
1139         .get_context    = f2fs_get_context,
1140         .key_prefix     = f2fs_key_prefix,
1141         .set_context    = f2fs_set_context,
1142         .is_encrypted   = f2fs_encrypted_inode,
1143         .empty_dir      = f2fs_empty_dir,
1144         .max_namelen    = f2fs_max_namelen,
1145 };
1146 #else
1147 static struct fscrypt_operations f2fs_cryptops = {
1148         .is_encrypted   = f2fs_encrypted_inode,
1149 };
1150 #endif
1151
1152 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1153                 u64 ino, u32 generation)
1154 {
1155         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1156         struct inode *inode;
1157
1158         if (check_nid_range(sbi, ino))
1159                 return ERR_PTR(-ESTALE);
1160
1161         /*
1162          * f2fs_iget isn't quite right if the inode is currently unallocated!
1163          * However f2fs_iget currently does appropriate checks to handle stale
1164          * inodes so everything is OK.
1165          */
1166         inode = f2fs_iget(sb, ino);
1167         if (IS_ERR(inode))
1168                 return ERR_CAST(inode);
1169         if (unlikely(generation && inode->i_generation != generation)) {
1170                 /* we didn't find the right inode.. */
1171                 iput(inode);
1172                 return ERR_PTR(-ESTALE);
1173         }
1174         return inode;
1175 }
1176
1177 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1178                 int fh_len, int fh_type)
1179 {
1180         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1181                                     f2fs_nfs_get_inode);
1182 }
1183
1184 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1185                 int fh_len, int fh_type)
1186 {
1187         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1188                                     f2fs_nfs_get_inode);
1189 }
1190
1191 static const struct export_operations f2fs_export_ops = {
1192         .fh_to_dentry = f2fs_fh_to_dentry,
1193         .fh_to_parent = f2fs_fh_to_parent,
1194         .get_parent = f2fs_get_parent,
1195 };
1196
1197 static loff_t max_file_blocks(void)
1198 {
1199         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1200         loff_t leaf_count = ADDRS_PER_BLOCK;
1201
1202         /* two direct node blocks */
1203         result += (leaf_count * 2);
1204
1205         /* two indirect node blocks */
1206         leaf_count *= NIDS_PER_BLOCK;
1207         result += (leaf_count * 2);
1208
1209         /* one double indirect node block */
1210         leaf_count *= NIDS_PER_BLOCK;
1211         result += leaf_count;
1212
1213         return result;
1214 }
1215
1216 static int __f2fs_commit_super(struct buffer_head *bh,
1217                         struct f2fs_super_block *super)
1218 {
1219         lock_buffer(bh);
1220         if (super)
1221                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1222         set_buffer_uptodate(bh);
1223         set_buffer_dirty(bh);
1224         unlock_buffer(bh);
1225
1226         /* it's rare case, we can do fua all the time */
1227         return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1228 }
1229
1230 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1231                                         struct buffer_head *bh)
1232 {
1233         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1234                                         (bh->b_data + F2FS_SUPER_OFFSET);
1235         struct super_block *sb = sbi->sb;
1236         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1237         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1238         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1239         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1240         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1241         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1242         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1243         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1244         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1245         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1246         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1247         u32 segment_count = le32_to_cpu(raw_super->segment_count);
1248         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1249         u64 main_end_blkaddr = main_blkaddr +
1250                                 (segment_count_main << log_blocks_per_seg);
1251         u64 seg_end_blkaddr = segment0_blkaddr +
1252                                 (segment_count << log_blocks_per_seg);
1253
1254         if (segment0_blkaddr != cp_blkaddr) {
1255                 f2fs_msg(sb, KERN_INFO,
1256                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1257                         segment0_blkaddr, cp_blkaddr);
1258                 return true;
1259         }
1260
1261         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1262                                                         sit_blkaddr) {
1263                 f2fs_msg(sb, KERN_INFO,
1264                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1265                         cp_blkaddr, sit_blkaddr,
1266                         segment_count_ckpt << log_blocks_per_seg);
1267                 return true;
1268         }
1269
1270         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1271                                                         nat_blkaddr) {
1272                 f2fs_msg(sb, KERN_INFO,
1273                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1274                         sit_blkaddr, nat_blkaddr,
1275                         segment_count_sit << log_blocks_per_seg);
1276                 return true;
1277         }
1278
1279         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1280                                                         ssa_blkaddr) {
1281                 f2fs_msg(sb, KERN_INFO,
1282                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1283                         nat_blkaddr, ssa_blkaddr,
1284                         segment_count_nat << log_blocks_per_seg);
1285                 return true;
1286         }
1287
1288         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1289                                                         main_blkaddr) {
1290                 f2fs_msg(sb, KERN_INFO,
1291                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1292                         ssa_blkaddr, main_blkaddr,
1293                         segment_count_ssa << log_blocks_per_seg);
1294                 return true;
1295         }
1296
1297         if (main_end_blkaddr > seg_end_blkaddr) {
1298                 f2fs_msg(sb, KERN_INFO,
1299                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1300                         main_blkaddr,
1301                         segment0_blkaddr +
1302                                 (segment_count << log_blocks_per_seg),
1303                         segment_count_main << log_blocks_per_seg);
1304                 return true;
1305         } else if (main_end_blkaddr < seg_end_blkaddr) {
1306                 int err = 0;
1307                 char *res;
1308
1309                 /* fix in-memory information all the time */
1310                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1311                                 segment0_blkaddr) >> log_blocks_per_seg);
1312
1313                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1314                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1315                         res = "internally";
1316                 } else {
1317                         err = __f2fs_commit_super(bh, NULL);
1318                         res = err ? "failed" : "done";
1319                 }
1320                 f2fs_msg(sb, KERN_INFO,
1321                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
1322                         res, main_blkaddr,
1323                         segment0_blkaddr +
1324                                 (segment_count << log_blocks_per_seg),
1325                         segment_count_main << log_blocks_per_seg);
1326                 if (err)
1327                         return true;
1328         }
1329         return false;
1330 }
1331
1332 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1333                                 struct buffer_head *bh)
1334 {
1335         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1336                                         (bh->b_data + F2FS_SUPER_OFFSET);
1337         struct super_block *sb = sbi->sb;
1338         unsigned int blocksize;
1339
1340         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1341                 f2fs_msg(sb, KERN_INFO,
1342                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1343                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1344                 return 1;
1345         }
1346
1347         /* Currently, support only 4KB page cache size */
1348         if (F2FS_BLKSIZE != PAGE_SIZE) {
1349                 f2fs_msg(sb, KERN_INFO,
1350                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1351                         PAGE_SIZE);
1352                 return 1;
1353         }
1354
1355         /* Currently, support only 4KB block size */
1356         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1357         if (blocksize != F2FS_BLKSIZE) {
1358                 f2fs_msg(sb, KERN_INFO,
1359                         "Invalid blocksize (%u), supports only 4KB\n",
1360                         blocksize);
1361                 return 1;
1362         }
1363
1364         /* check log blocks per segment */
1365         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1366                 f2fs_msg(sb, KERN_INFO,
1367                         "Invalid log blocks per segment (%u)\n",
1368                         le32_to_cpu(raw_super->log_blocks_per_seg));
1369                 return 1;
1370         }
1371
1372         /* Currently, support 512/1024/2048/4096 bytes sector size */
1373         if (le32_to_cpu(raw_super->log_sectorsize) >
1374                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1375                 le32_to_cpu(raw_super->log_sectorsize) <
1376                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1377                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1378                         le32_to_cpu(raw_super->log_sectorsize));
1379                 return 1;
1380         }
1381         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1382                 le32_to_cpu(raw_super->log_sectorsize) !=
1383                         F2FS_MAX_LOG_SECTOR_SIZE) {
1384                 f2fs_msg(sb, KERN_INFO,
1385                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1386                         le32_to_cpu(raw_super->log_sectors_per_block),
1387                         le32_to_cpu(raw_super->log_sectorsize));
1388                 return 1;
1389         }
1390
1391         /* check reserved ino info */
1392         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1393                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1394                 le32_to_cpu(raw_super->root_ino) != 3) {
1395                 f2fs_msg(sb, KERN_INFO,
1396                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1397                         le32_to_cpu(raw_super->node_ino),
1398                         le32_to_cpu(raw_super->meta_ino),
1399                         le32_to_cpu(raw_super->root_ino));
1400                 return 1;
1401         }
1402
1403         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1404         if (sanity_check_area_boundary(sbi, bh))
1405                 return 1;
1406
1407         return 0;
1408 }
1409
1410 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1411 {
1412         unsigned int total, fsmeta;
1413         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1414         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1415
1416         total = le32_to_cpu(raw_super->segment_count);
1417         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1418         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1419         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1420         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1421         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1422
1423         if (unlikely(fsmeta >= total))
1424                 return 1;
1425
1426         if (unlikely(f2fs_cp_error(sbi))) {
1427                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1428                 return 1;
1429         }
1430         return 0;
1431 }
1432
1433 static void init_sb_info(struct f2fs_sb_info *sbi)
1434 {
1435         struct f2fs_super_block *raw_super = sbi->raw_super;
1436
1437         sbi->log_sectors_per_block =
1438                 le32_to_cpu(raw_super->log_sectors_per_block);
1439         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1440         sbi->blocksize = 1 << sbi->log_blocksize;
1441         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1442         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1443         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1444         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1445         sbi->total_sections = le32_to_cpu(raw_super->section_count);
1446         sbi->total_node_count =
1447                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1448                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1449         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1450         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1451         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1452         sbi->cur_victim_sec = NULL_SECNO;
1453         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1454
1455         sbi->dir_level = DEF_DIR_LEVEL;
1456         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1457         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1458         clear_sbi_flag(sbi, SBI_NEED_FSCK);
1459
1460         INIT_LIST_HEAD(&sbi->s_list);
1461         mutex_init(&sbi->umount_mutex);
1462         mutex_init(&sbi->wio_mutex[NODE]);
1463         mutex_init(&sbi->wio_mutex[DATA]);
1464
1465 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1466         memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1467                                 F2FS_KEY_DESC_PREFIX_SIZE);
1468         sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1469 #endif
1470 }
1471
1472 static int init_percpu_info(struct f2fs_sb_info *sbi)
1473 {
1474         int i, err;
1475
1476         if (percpu_init_rwsem(&sbi->cp_rwsem))
1477                 return -ENOMEM;
1478
1479         for (i = 0; i < NR_COUNT_TYPE; i++) {
1480                 err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1481                 if (err)
1482                         return err;
1483         }
1484
1485         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1486         if (err)
1487                 return err;
1488
1489         return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1490                                                                 GFP_KERNEL);
1491 }
1492
1493 /*
1494  * Read f2fs raw super block.
1495  * Because we have two copies of super block, so read both of them
1496  * to get the first valid one. If any one of them is broken, we pass
1497  * them recovery flag back to the caller.
1498  */
1499 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1500                         struct f2fs_super_block **raw_super,
1501                         int *valid_super_block, int *recovery)
1502 {
1503         struct super_block *sb = sbi->sb;
1504         int block;
1505         struct buffer_head *bh;
1506         struct f2fs_super_block *super;
1507         int err = 0;
1508
1509         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1510         if (!super)
1511                 return -ENOMEM;
1512
1513         for (block = 0; block < 2; block++) {
1514                 bh = sb_bread(sb, block);
1515                 if (!bh) {
1516                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1517                                 block + 1);
1518                         err = -EIO;
1519                         continue;
1520                 }
1521
1522                 /* sanity checking of raw super */
1523                 if (sanity_check_raw_super(sbi, bh)) {
1524                         f2fs_msg(sb, KERN_ERR,
1525                                 "Can't find valid F2FS filesystem in %dth superblock",
1526                                 block + 1);
1527                         err = -EINVAL;
1528                         brelse(bh);
1529                         continue;
1530                 }
1531
1532                 if (!*raw_super) {
1533                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1534                                                         sizeof(*super));
1535                         *valid_super_block = block;
1536                         *raw_super = super;
1537                 }
1538                 brelse(bh);
1539         }
1540
1541         /* Fail to read any one of the superblocks*/
1542         if (err < 0)
1543                 *recovery = 1;
1544
1545         /* No valid superblock */
1546         if (!*raw_super)
1547                 kfree(super);
1548         else
1549                 err = 0;
1550
1551         return err;
1552 }
1553
1554 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1555 {
1556         struct buffer_head *bh;
1557         int err;
1558
1559         if ((recover && f2fs_readonly(sbi->sb)) ||
1560                                 bdev_read_only(sbi->sb->s_bdev)) {
1561                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1562                 return -EROFS;
1563         }
1564
1565         /* write back-up superblock first */
1566         bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1567         if (!bh)
1568                 return -EIO;
1569         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1570         brelse(bh);
1571
1572         /* if we are in recovery path, skip writing valid superblock */
1573         if (recover || err)
1574                 return err;
1575
1576         /* write current valid superblock */
1577         bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1578         if (!bh)
1579                 return -EIO;
1580         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1581         brelse(bh);
1582         return err;
1583 }
1584
1585 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1586 {
1587         struct f2fs_sb_info *sbi;
1588         struct f2fs_super_block *raw_super;
1589         struct inode *root;
1590         int err;
1591         bool retry = true, need_fsck = false;
1592         char *options = NULL;
1593         int recovery, i, valid_super_block;
1594         struct curseg_info *seg_i;
1595
1596 try_onemore:
1597         err = -EINVAL;
1598         raw_super = NULL;
1599         valid_super_block = -1;
1600         recovery = 0;
1601
1602         /* allocate memory for f2fs-specific super block info */
1603         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1604         if (!sbi)
1605                 return -ENOMEM;
1606
1607         sbi->sb = sb;
1608
1609         /* Load the checksum driver */
1610         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1611         if (IS_ERR(sbi->s_chksum_driver)) {
1612                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1613                 err = PTR_ERR(sbi->s_chksum_driver);
1614                 sbi->s_chksum_driver = NULL;
1615                 goto free_sbi;
1616         }
1617
1618         /* set a block size */
1619         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1620                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1621                 goto free_sbi;
1622         }
1623
1624         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1625                                                                 &recovery);
1626         if (err)
1627                 goto free_sbi;
1628
1629         sb->s_fs_info = sbi;
1630         sbi->raw_super = raw_super;
1631
1632         default_options(sbi);
1633         /* parse mount options */
1634         options = kstrdup((const char *)data, GFP_KERNEL);
1635         if (data && !options) {
1636                 err = -ENOMEM;
1637                 goto free_sb_buf;
1638         }
1639
1640         err = parse_options(sb, options);
1641         if (err)
1642                 goto free_options;
1643
1644         sbi->max_file_blocks = max_file_blocks();
1645         sb->s_maxbytes = sbi->max_file_blocks <<
1646                                 le32_to_cpu(raw_super->log_blocksize);
1647         sb->s_max_links = F2FS_LINK_MAX;
1648         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1649
1650         sb->s_op = &f2fs_sops;
1651         sb->s_cop = &f2fs_cryptops;
1652         sb->s_xattr = f2fs_xattr_handlers;
1653         sb->s_export_op = &f2fs_export_ops;
1654         sb->s_magic = F2FS_SUPER_MAGIC;
1655         sb->s_time_gran = 1;
1656         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1657                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1658         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1659
1660         /* init f2fs-specific super block info */
1661         sbi->valid_super_block = valid_super_block;
1662         mutex_init(&sbi->gc_mutex);
1663         mutex_init(&sbi->cp_mutex);
1664         init_rwsem(&sbi->node_write);
1665
1666         /* disallow all the data/node/meta page writes */
1667         set_sbi_flag(sbi, SBI_POR_DOING);
1668         spin_lock_init(&sbi->stat_lock);
1669
1670         init_rwsem(&sbi->read_io.io_rwsem);
1671         sbi->read_io.sbi = sbi;
1672         sbi->read_io.bio = NULL;
1673         for (i = 0; i < NR_PAGE_TYPE; i++) {
1674                 init_rwsem(&sbi->write_io[i].io_rwsem);
1675                 sbi->write_io[i].sbi = sbi;
1676                 sbi->write_io[i].bio = NULL;
1677         }
1678
1679         init_waitqueue_head(&sbi->cp_wait);
1680         init_sb_info(sbi);
1681
1682         err = init_percpu_info(sbi);
1683         if (err)
1684                 goto free_options;
1685
1686         /* get an inode for meta space */
1687         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1688         if (IS_ERR(sbi->meta_inode)) {
1689                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1690                 err = PTR_ERR(sbi->meta_inode);
1691                 goto free_options;
1692         }
1693
1694         err = get_valid_checkpoint(sbi);
1695         if (err) {
1696                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1697                 goto free_meta_inode;
1698         }
1699
1700         sbi->total_valid_node_count =
1701                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1702         percpu_counter_set(&sbi->total_valid_inode_count,
1703                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
1704         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1705         sbi->total_valid_block_count =
1706                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1707         sbi->last_valid_block_count = sbi->total_valid_block_count;
1708
1709         for (i = 0; i < NR_INODE_TYPE; i++) {
1710                 INIT_LIST_HEAD(&sbi->inode_list[i]);
1711                 spin_lock_init(&sbi->inode_lock[i]);
1712         }
1713
1714         init_extent_cache_info(sbi);
1715
1716         init_ino_entry_info(sbi);
1717
1718         /* setup f2fs internal modules */
1719         err = build_segment_manager(sbi);
1720         if (err) {
1721                 f2fs_msg(sb, KERN_ERR,
1722                         "Failed to initialize F2FS segment manager");
1723                 goto free_sm;
1724         }
1725         err = build_node_manager(sbi);
1726         if (err) {
1727                 f2fs_msg(sb, KERN_ERR,
1728                         "Failed to initialize F2FS node manager");
1729                 goto free_nm;
1730         }
1731
1732         /* For write statistics */
1733         if (sb->s_bdev->bd_part)
1734                 sbi->sectors_written_start =
1735                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1736
1737         /* Read accumulated write IO statistics if exists */
1738         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1739         if (__exist_node_summaries(sbi))
1740                 sbi->kbytes_written =
1741                         le64_to_cpu(seg_i->journal->info.kbytes_written);
1742
1743         build_gc_manager(sbi);
1744
1745         /* get an inode for node space */
1746         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1747         if (IS_ERR(sbi->node_inode)) {
1748                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1749                 err = PTR_ERR(sbi->node_inode);
1750                 goto free_nm;
1751         }
1752
1753         f2fs_join_shrinker(sbi);
1754
1755         /* if there are nt orphan nodes free them */
1756         err = recover_orphan_inodes(sbi);
1757         if (err)
1758                 goto free_node_inode;
1759
1760         /* read root inode and dentry */
1761         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1762         if (IS_ERR(root)) {
1763                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1764                 err = PTR_ERR(root);
1765                 goto free_node_inode;
1766         }
1767         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1768                 iput(root);
1769                 err = -EINVAL;
1770                 goto free_node_inode;
1771         }
1772
1773         sb->s_root = d_make_root(root); /* allocate root dentry */
1774         if (!sb->s_root) {
1775                 err = -ENOMEM;
1776                 goto free_root_inode;
1777         }
1778
1779         err = f2fs_build_stats(sbi);
1780         if (err)
1781                 goto free_root_inode;
1782
1783         if (f2fs_proc_root)
1784                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1785
1786         if (sbi->s_proc) {
1787                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1788                                  &f2fs_seq_segment_info_fops, sb);
1789                 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1790                                  &f2fs_seq_segment_bits_fops, sb);
1791         }
1792
1793         sbi->s_kobj.kset = f2fs_kset;
1794         init_completion(&sbi->s_kobj_unregister);
1795         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1796                                                         "%s", sb->s_id);
1797         if (err)
1798                 goto free_proc;
1799
1800         /* recover fsynced data */
1801         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1802                 /*
1803                  * mount should be failed, when device has readonly mode, and
1804                  * previous checkpoint was not done by clean system shutdown.
1805                  */
1806                 if (bdev_read_only(sb->s_bdev) &&
1807                                 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1808                         err = -EROFS;
1809                         goto free_kobj;
1810                 }
1811
1812                 if (need_fsck)
1813                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1814
1815                 err = recover_fsync_data(sbi, false);
1816                 if (err < 0) {
1817                         need_fsck = true;
1818                         f2fs_msg(sb, KERN_ERR,
1819                                 "Cannot recover all fsync data errno=%d", err);
1820                         goto free_kobj;
1821                 }
1822         } else {
1823                 err = recover_fsync_data(sbi, true);
1824
1825                 if (!f2fs_readonly(sb) && err > 0) {
1826                         err = -EINVAL;
1827                         f2fs_msg(sb, KERN_ERR,
1828                                 "Need to recover fsync data");
1829                         goto free_kobj;
1830                 }
1831         }
1832
1833         /* recover_fsync_data() cleared this already */
1834         clear_sbi_flag(sbi, SBI_POR_DOING);
1835
1836         /*
1837          * If filesystem is not mounted as read-only then
1838          * do start the gc_thread.
1839          */
1840         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1841                 /* After POR, we can run background GC thread.*/
1842                 err = start_gc_thread(sbi);
1843                 if (err)
1844                         goto free_kobj;
1845         }
1846         kfree(options);
1847
1848         /* recover broken superblock */
1849         if (recovery) {
1850                 err = f2fs_commit_super(sbi, true);
1851                 f2fs_msg(sb, KERN_INFO,
1852                         "Try to recover %dth superblock, ret: %d",
1853                         sbi->valid_super_block ? 1 : 2, err);
1854         }
1855
1856         f2fs_update_time(sbi, CP_TIME);
1857         f2fs_update_time(sbi, REQ_TIME);
1858         return 0;
1859
1860 free_kobj:
1861         f2fs_sync_inode_meta(sbi);
1862         kobject_del(&sbi->s_kobj);
1863         kobject_put(&sbi->s_kobj);
1864         wait_for_completion(&sbi->s_kobj_unregister);
1865 free_proc:
1866         if (sbi->s_proc) {
1867                 remove_proc_entry("segment_info", sbi->s_proc);
1868                 remove_proc_entry("segment_bits", sbi->s_proc);
1869                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1870         }
1871         f2fs_destroy_stats(sbi);
1872 free_root_inode:
1873         dput(sb->s_root);
1874         sb->s_root = NULL;
1875 free_node_inode:
1876         mutex_lock(&sbi->umount_mutex);
1877         f2fs_leave_shrinker(sbi);
1878         iput(sbi->node_inode);
1879         mutex_unlock(&sbi->umount_mutex);
1880 free_nm:
1881         destroy_node_manager(sbi);
1882 free_sm:
1883         destroy_segment_manager(sbi);
1884         kfree(sbi->ckpt);
1885 free_meta_inode:
1886         make_bad_inode(sbi->meta_inode);
1887         iput(sbi->meta_inode);
1888 free_options:
1889         destroy_percpu_info(sbi);
1890         kfree(options);
1891 free_sb_buf:
1892         kfree(raw_super);
1893 free_sbi:
1894         if (sbi->s_chksum_driver)
1895                 crypto_free_shash(sbi->s_chksum_driver);
1896         kfree(sbi);
1897
1898         /* give only one another chance */
1899         if (retry) {
1900                 retry = false;
1901                 shrink_dcache_sb(sb);
1902                 goto try_onemore;
1903         }
1904         return err;
1905 }
1906
1907 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1908                         const char *dev_name, void *data)
1909 {
1910         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1911 }
1912
1913 static void kill_f2fs_super(struct super_block *sb)
1914 {
1915         if (sb->s_root)
1916                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1917         kill_block_super(sb);
1918 }
1919
1920 static struct file_system_type f2fs_fs_type = {
1921         .owner          = THIS_MODULE,
1922         .name           = "f2fs",
1923         .mount          = f2fs_mount,
1924         .kill_sb        = kill_f2fs_super,
1925         .fs_flags       = FS_REQUIRES_DEV,
1926 };
1927 MODULE_ALIAS_FS("f2fs");
1928
1929 static int __init init_inodecache(void)
1930 {
1931         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1932                         sizeof(struct f2fs_inode_info), 0,
1933                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1934         if (!f2fs_inode_cachep)
1935                 return -ENOMEM;
1936         return 0;
1937 }
1938
1939 static void destroy_inodecache(void)
1940 {
1941         /*
1942          * Make sure all delayed rcu free inodes are flushed before we
1943          * destroy cache.
1944          */
1945         rcu_barrier();
1946         kmem_cache_destroy(f2fs_inode_cachep);
1947 }
1948
1949 static int __init init_f2fs_fs(void)
1950 {
1951         int err;
1952
1953         f2fs_build_trace_ios();
1954
1955         err = init_inodecache();
1956         if (err)
1957                 goto fail;
1958         err = create_node_manager_caches();
1959         if (err)
1960                 goto free_inodecache;
1961         err = create_segment_manager_caches();
1962         if (err)
1963                 goto free_node_manager_caches;
1964         err = create_checkpoint_caches();
1965         if (err)
1966                 goto free_segment_manager_caches;
1967         err = create_extent_cache();
1968         if (err)
1969                 goto free_checkpoint_caches;
1970         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1971         if (!f2fs_kset) {
1972                 err = -ENOMEM;
1973                 goto free_extent_cache;
1974         }
1975 #ifdef CONFIG_F2FS_FAULT_INJECTION
1976         f2fs_fault_inject.kset = f2fs_kset;
1977         f2fs_build_fault_attr(0);
1978         err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1979                                 NULL, "fault_injection");
1980         if (err) {
1981                 f2fs_fault_inject.kset = NULL;
1982                 goto free_kset;
1983         }
1984 #endif
1985         err = register_shrinker(&f2fs_shrinker_info);
1986         if (err)
1987                 goto free_kset;
1988
1989         err = register_filesystem(&f2fs_fs_type);
1990         if (err)
1991                 goto free_shrinker;
1992         err = f2fs_create_root_stats();
1993         if (err)
1994                 goto free_filesystem;
1995         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1996         return 0;
1997
1998 free_filesystem:
1999         unregister_filesystem(&f2fs_fs_type);
2000 free_shrinker:
2001         unregister_shrinker(&f2fs_shrinker_info);
2002 free_kset:
2003 #ifdef CONFIG_F2FS_FAULT_INJECTION
2004         if (f2fs_fault_inject.kset)
2005                 kobject_put(&f2fs_fault_inject);
2006 #endif
2007         kset_unregister(f2fs_kset);
2008 free_extent_cache:
2009         destroy_extent_cache();
2010 free_checkpoint_caches:
2011         destroy_checkpoint_caches();
2012 free_segment_manager_caches:
2013         destroy_segment_manager_caches();
2014 free_node_manager_caches:
2015         destroy_node_manager_caches();
2016 free_inodecache:
2017         destroy_inodecache();
2018 fail:
2019         return err;
2020 }
2021
2022 static void __exit exit_f2fs_fs(void)
2023 {
2024         remove_proc_entry("fs/f2fs", NULL);
2025         f2fs_destroy_root_stats();
2026         unregister_filesystem(&f2fs_fs_type);
2027         unregister_shrinker(&f2fs_shrinker_info);
2028 #ifdef CONFIG_F2FS_FAULT_INJECTION
2029         kobject_put(&f2fs_fault_inject);
2030 #endif
2031         kset_unregister(f2fs_kset);
2032         destroy_extent_cache();
2033         destroy_checkpoint_caches();
2034         destroy_segment_manager_caches();
2035         destroy_node_manager_caches();
2036         destroy_inodecache();
2037         f2fs_destroy_trace_ios();
2038 }
2039
2040 module_init(init_f2fs_fs)
2041 module_exit(exit_f2fs_fs)
2042
2043 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2044 MODULE_DESCRIPTION("Flash Friendly File System");
2045 MODULE_LICENSE("GPL");