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         init_rwsem(&fi->dio_rwsem[READ]);
 583         init_rwsem(&fi->dio_rwsem[WRITE]);
 584
 585         /* Will be used by directory only */
 586         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 587         return &fi->vfs_inode;
 588 }
 589
 590 static int f2fs_drop_inode(struct inode *inode)
 591 {
 592         /*
 593          * This is to avoid a deadlock condition like below.
 594          * writeback_single_inode(inode)
 595          *  - f2fs_write_data_page
 596          *    - f2fs_gc -> iput -> evict
 597          *       - inode_wait_for_writeback(inode)
 598          */
 599         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 600                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 601                         /* to avoid evict_inode call simultaneously */
 602                         atomic_inc(&inode->i_count);
 603                         spin_unlock(&inode->i_lock);
 604
 605                         /* some remained atomic pages should discarded */
 606                         if (f2fs_is_atomic_file(inode))
 607                                 drop_inmem_pages(inode);
 608
 609                         /* should remain fi->extent_tree for writepage */
 610                         f2fs_destroy_extent_node(inode);
 611
 612                         sb_start_intwrite(inode->i_sb);
 613                         f2fs_i_size_write(inode, 0);
 614
 615                         if (F2FS_HAS_BLOCKS(inode))
 616                                 f2fs_truncate(inode);
 617
 618                         sb_end_intwrite(inode->i_sb);
 619
 620                         fscrypt_put_encryption_info(inode, NULL);
 621                         spin_lock(&inode->i_lock);
 622                         atomic_dec(&inode->i_count);
 623                 }
 624                 return 0;
 625         }
 626
 627         return generic_drop_inode(inode);
 628 }
 629
 630 int f2fs_inode_dirtied(struct inode *inode)
 631 {
 632         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 633
 634         spin_lock(&sbi->inode_lock[DIRTY_META]);
 635         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 636                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 637                 return 1;
 638         }
 639
 640         set_inode_flag(inode, FI_DIRTY_INODE);
 641         list_add_tail(&F2FS_I(inode)->gdirty_list,
 642                                 &sbi->inode_list[DIRTY_META]);
 643         inc_page_count(sbi, F2FS_DIRTY_IMETA);
 644         stat_inc_dirty_inode(sbi, DIRTY_META);
 645         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 646
 647         return 0;
 648 }
 649
 650 void f2fs_inode_synced(struct inode *inode)
 651 {
 652         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 653
 654         spin_lock(&sbi->inode_lock[DIRTY_META]);
 655         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 656                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 657                 return;
 658         }
 659         list_del_init(&F2FS_I(inode)->gdirty_list);
 660         clear_inode_flag(inode, FI_DIRTY_INODE);
 661         clear_inode_flag(inode, FI_AUTO_RECOVER);
 662         dec_page_count(sbi, F2FS_DIRTY_IMETA);
 663         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 664         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 665 }
 666
 667 /*
 668  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 669  *
 670  * We should call set_dirty_inode to write the dirty inode through write_inode.
 671  */
 672 static void f2fs_dirty_inode(struct inode *inode, int flags)
 673 {
 674         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 675
 676         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 677                         inode->i_ino == F2FS_META_INO(sbi))
 678                 return;
 679
 680         if (flags == I_DIRTY_TIME)
 681                 return;
 682
 683         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 684                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 685
 686         f2fs_inode_dirtied(inode);
 687 }
 688
 689 static void f2fs_i_callback(struct rcu_head *head)
 690 {
 691         struct inode *inode = container_of(head, struct inode, i_rcu);
 692         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 693 }
 694
 695 static void f2fs_destroy_inode(struct inode *inode)
 696 {
 697         percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
 698         call_rcu(&inode->i_rcu, f2fs_i_callback);
 699 }
 700
 701 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 702 {
 703         int i;
 704
 705         for (i = 0; i < NR_COUNT_TYPE; i++)
 706                 percpu_counter_destroy(&sbi->nr_pages[i]);
 707         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 708         percpu_counter_destroy(&sbi->total_valid_inode_count);
 709 }
 710
 711 static void f2fs_put_super(struct super_block *sb)
 712 {
 713         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 714
 715         if (sbi->s_proc) {
 716                 remove_proc_entry("segment_info", sbi->s_proc);
 717                 remove_proc_entry("segment_bits", sbi->s_proc);
 718                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 719         }
 720         kobject_del(&sbi->s_kobj);
 721
 722         stop_gc_thread(sbi);
 723
 724         /* prevent remaining shrinker jobs */
 725         mutex_lock(&sbi->umount_mutex);
 726
 727         /*
 728          * We don't need to do checkpoint when superblock is clean.
 729          * But, the previous checkpoint was not done by umount, it needs to do
 730          * clean checkpoint again.
 731          */
 732         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 733                         !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
 734                 struct cp_control cpc = {
 735                         .reason = CP_UMOUNT,
 736                 };
 737                 write_checkpoint(sbi, &cpc);
 738         }
 739
 740         /* write_checkpoint can update stat informaion */
 741         f2fs_destroy_stats(sbi);
 742
 743         /*
 744          * normally superblock is clean, so we need to release this.
 745          * In addition, EIO will skip do checkpoint, we need this as well.
 746          */
 747         release_ino_entry(sbi, true);
 748         release_discard_addrs(sbi);
 749
 750         f2fs_leave_shrinker(sbi);
 751         mutex_unlock(&sbi->umount_mutex);
 752
 753         /* our cp_error case, we can wait for any writeback page */
 754         f2fs_flush_merged_bios(sbi);
 755
 756         iput(sbi->node_inode);
 757         iput(sbi->meta_inode);
 758
 759         /* destroy f2fs internal modules */
 760         destroy_node_manager(sbi);
 761         destroy_segment_manager(sbi);
 762
 763         kfree(sbi->ckpt);
 764         kobject_put(&sbi->s_kobj);
 765         wait_for_completion(&sbi->s_kobj_unregister);
 766
 767         sb->s_fs_info = NULL;
 768         if (sbi->s_chksum_driver)
 769                 crypto_free_shash(sbi->s_chksum_driver);
 770         kfree(sbi->raw_super);
 771
 772         destroy_percpu_info(sbi);
 773         kfree(sbi);
 774 }
 775
 776 int f2fs_sync_fs(struct super_block *sb, int sync)
 777 {
 778         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 779         int err = 0;
 780
 781         trace_f2fs_sync_fs(sb, sync);
 782
 783         if (sync) {
 784                 struct cp_control cpc;
 785
 786                 cpc.reason = __get_cp_reason(sbi);
 787
 788                 mutex_lock(&sbi->gc_mutex);
 789                 err = write_checkpoint(sbi, &cpc);
 790                 mutex_unlock(&sbi->gc_mutex);
 791         }
 792         f2fs_trace_ios(NULL, 1);
 793
 794         return err;
 795 }
 796
 797 static int f2fs_freeze(struct super_block *sb)
 798 {
 799         int err;
 800
 801         if (f2fs_readonly(sb))
 802                 return 0;
 803
 804         err = f2fs_sync_fs(sb, 1);
 805         return err;
 806 }
 807
 808 static int f2fs_unfreeze(struct super_block *sb)
 809 {
 810         return 0;
 811 }
 812
 813 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 814 {
 815         struct super_block *sb = dentry->d_sb;
 816         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 817         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 818         block_t total_count, user_block_count, start_count, ovp_count;
 819
 820         total_count = le64_to_cpu(sbi->raw_super->block_count);
 821         user_block_count = sbi->user_block_count;
 822         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 823         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 824         buf->f_type = F2FS_SUPER_MAGIC;
 825         buf->f_bsize = sbi->blocksize;
 826
 827         buf->f_blocks = total_count - start_count;
 828         buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
 829         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 830
 831         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 832         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
 833
 834         buf->f_namelen = F2FS_NAME_LEN;
 835         buf->f_fsid.val[0] = (u32)id;
 836         buf->f_fsid.val[1] = (u32)(id >> 32);
 837
 838         return 0;
 839 }
 840
 841 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 842 {
 843         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 844
 845         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
 846                 if (test_opt(sbi, FORCE_FG_GC))
 847                         seq_printf(seq, ",background_gc=%s", "sync");
 848                 else
 849                         seq_printf(seq, ",background_gc=%s", "on");
 850         } else {
 851                 seq_printf(seq, ",background_gc=%s", "off");
 852         }
 853         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 854                 seq_puts(seq, ",disable_roll_forward");
 855         if (test_opt(sbi, DISCARD))
 856                 seq_puts(seq, ",discard");
 857         if (test_opt(sbi, NOHEAP))
 858                 seq_puts(seq, ",no_heap_alloc");
 859 #ifdef CONFIG_F2FS_FS_XATTR
 860         if (test_opt(sbi, XATTR_USER))
 861                 seq_puts(seq, ",user_xattr");
 862         else
 863                 seq_puts(seq, ",nouser_xattr");
 864         if (test_opt(sbi, INLINE_XATTR))
 865                 seq_puts(seq, ",inline_xattr");
 866 #endif
 867 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 868         if (test_opt(sbi, POSIX_ACL))
 869                 seq_puts(seq, ",acl");
 870         else
 871                 seq_puts(seq, ",noacl");
 872 #endif
 873         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 874                 seq_puts(seq, ",disable_ext_identify");
 875         if (test_opt(sbi, INLINE_DATA))
 876                 seq_puts(seq, ",inline_data");
 877         else
 878                 seq_puts(seq, ",noinline_data");
 879         if (test_opt(sbi, INLINE_DENTRY))
 880                 seq_puts(seq, ",inline_dentry");
 881         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 882                 seq_puts(seq, ",flush_merge");
 883         if (test_opt(sbi, NOBARRIER))
 884                 seq_puts(seq, ",nobarrier");
 885         if (test_opt(sbi, FASTBOOT))
 886                 seq_puts(seq, ",fastboot");
 887         if (test_opt(sbi, EXTENT_CACHE))
 888                 seq_puts(seq, ",extent_cache");
 889         else
 890                 seq_puts(seq, ",noextent_cache");
 891         if (test_opt(sbi, DATA_FLUSH))
 892                 seq_puts(seq, ",data_flush");
 893
 894         seq_puts(seq, ",mode=");
 895         if (test_opt(sbi, ADAPTIVE))
 896                 seq_puts(seq, "adaptive");
 897         else if (test_opt(sbi, LFS))
 898                 seq_puts(seq, "lfs");
 899         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 900
 901         return 0;
 902 }
 903
 904 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 905 {
 906         struct super_block *sb = seq->private;
 907         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 908         unsigned int total_segs =
 909                         le32_to_cpu(sbi->raw_super->segment_count_main);
 910         int i;
 911
 912         seq_puts(seq, "format: segment_type|valid_blocks\n"
 913                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 914
 915         for (i = 0; i < total_segs; i++) {
 916                 struct seg_entry *se = get_seg_entry(sbi, i);
 917
 918                 if ((i % 10) == 0)
 919                         seq_printf(seq, "%-10d", i);
 920                 seq_printf(seq, "%d|%-3u", se->type,
 921                                         get_valid_blocks(sbi, i, 1));
 922                 if ((i % 10) == 9 || i == (total_segs - 1))
 923                         seq_putc(seq, '\n');
 924                 else
 925                         seq_putc(seq, ' ');
 926         }
 927
 928         return 0;
 929 }
 930
 931 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
 932 {
 933         struct super_block *sb = seq->private;
 934         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 935         unsigned int total_segs =
 936                         le32_to_cpu(sbi->raw_super->segment_count_main);
 937         int i, j;
 938
 939         seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
 940                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 941
 942         for (i = 0; i < total_segs; i++) {
 943                 struct seg_entry *se = get_seg_entry(sbi, i);
 944
 945                 seq_printf(seq, "%-10d", i);
 946                 seq_printf(seq, "%d|%-3u|", se->type,
 947                                         get_valid_blocks(sbi, i, 1));
 948                 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
 949                         seq_printf(seq, "%x ", se->cur_valid_map[j]);
 950                 seq_putc(seq, '\n');
 951         }
 952         return 0;
 953 }
 954
 955 #define F2FS_PROC_FILE_DEF(_name)                                       \
 956 static int _name##_open_fs(struct inode *inode, struct file *file)      \
 957 {                                                                       \
 958         return single_open(file, _name##_seq_show, PDE_DATA(inode));    \
 959 }                                                                       \
 960                                                                         \
 961 static const struct file_operations f2fs_seq_##_name##_fops = {         \
 962         .open = _name##_open_fs,                                        \
 963         .read = seq_read,                                               \
 964         .llseek = seq_lseek,                                            \
 965         .release = single_release,                                      \
 966 };
 967
 968 F2FS_PROC_FILE_DEF(segment_info);
 969 F2FS_PROC_FILE_DEF(segment_bits);
 970
 971 static void default_options(struct f2fs_sb_info *sbi)
 972 {
 973         /* init some FS parameters */
 974         sbi->active_logs = NR_CURSEG_TYPE;
 975
 976         set_opt(sbi, BG_GC);
 977         set_opt(sbi, INLINE_DATA);
 978         set_opt(sbi, EXTENT_CACHE);
 979         sbi->sb->s_flags |= MS_LAZYTIME;
 980         set_opt(sbi, FLUSH_MERGE);
 981         if (f2fs_sb_mounted_hmsmr(sbi->sb)) {
 982                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 983                 set_opt(sbi, DISCARD);
 984         } else {
 985                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 986         }
 987
 988 #ifdef CONFIG_F2FS_FS_XATTR
 989         set_opt(sbi, XATTR_USER);
 990 #endif
 991 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 992         set_opt(sbi, POSIX_ACL);
 993 #endif
 994 }
 995
 996 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 997 {
 998         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 999         struct f2fs_mount_info org_mount_opt;
1000         int err, active_logs;
1001         bool need_restart_gc = false;
1002         bool need_stop_gc = false;
1003         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1004
1005         /*
1006          * Save the old mount options in case we
1007          * need to restore them.
1008          */
1009         org_mount_opt = sbi->mount_opt;
1010         active_logs = sbi->active_logs;
1011
1012         /* recover superblocks we couldn't write due to previous RO mount */
1013         if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1014                 err = f2fs_commit_super(sbi, false);
1015                 f2fs_msg(sb, KERN_INFO,
1016                         "Try to recover all the superblocks, ret: %d", err);
1017                 if (!err)
1018                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1019         }
1020
1021         sbi->mount_opt.opt = 0;
1022         default_options(sbi);
1023
1024         /* parse mount options */
1025         err = parse_options(sb, data);
1026         if (err)
1027                 goto restore_opts;
1028
1029         /*
1030          * Previous and new state of filesystem is RO,
1031          * so skip checking GC and FLUSH_MERGE conditions.
1032          */
1033         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1034                 goto skip;
1035
1036         /* disallow enable/disable extent_cache dynamically */
1037         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1038                 err = -EINVAL;
1039                 f2fs_msg(sbi->sb, KERN_WARNING,
1040                                 "switch extent_cache option is not allowed");
1041                 goto restore_opts;
1042         }
1043
1044         /*
1045          * We stop the GC thread if FS is mounted as RO
1046          * or if background_gc = off is passed in mount
1047          * option. Also sync the filesystem.
1048          */
1049         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1050                 if (sbi->gc_thread) {
1051                         stop_gc_thread(sbi);
1052                         need_restart_gc = true;
1053                 }
1054         } else if (!sbi->gc_thread) {
1055                 err = start_gc_thread(sbi);
1056                 if (err)
1057                         goto restore_opts;
1058                 need_stop_gc = true;
1059         }
1060
1061         if (*flags & MS_RDONLY) {
1062                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1063                 sync_inodes_sb(sb);
1064
1065                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1066                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1067                 f2fs_sync_fs(sb, 1);
1068                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1069         }
1070
1071         /*
1072          * We stop issue flush thread if FS is mounted as RO
1073          * or if flush_merge is not passed in mount option.
1074          */
1075         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1076                 destroy_flush_cmd_control(sbi);
1077         } else if (!SM_I(sbi)->cmd_control_info) {
1078                 err = create_flush_cmd_control(sbi);
1079                 if (err)
1080                         goto restore_gc;
1081         }
1082 skip:
1083         /* Update the POSIXACL Flag */
1084         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1085                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1086
1087         return 0;
1088 restore_gc:
1089         if (need_restart_gc) {
1090                 if (start_gc_thread(sbi))
1091                         f2fs_msg(sbi->sb, KERN_WARNING,
1092                                 "background gc thread has stopped");
1093         } else if (need_stop_gc) {
1094                 stop_gc_thread(sbi);
1095         }
1096 restore_opts:
1097         sbi->mount_opt = org_mount_opt;
1098         sbi->active_logs = active_logs;
1099         return err;
1100 }
1101
1102 static struct super_operations f2fs_sops = {
1103         .alloc_inode    = f2fs_alloc_inode,
1104         .drop_inode     = f2fs_drop_inode,
1105         .destroy_inode  = f2fs_destroy_inode,
1106         .write_inode    = f2fs_write_inode,
1107         .dirty_inode    = f2fs_dirty_inode,
1108         .show_options   = f2fs_show_options,
1109         .evict_inode    = f2fs_evict_inode,
1110         .put_super      = f2fs_put_super,
1111         .sync_fs        = f2fs_sync_fs,
1112         .freeze_fs      = f2fs_freeze,
1113         .unfreeze_fs    = f2fs_unfreeze,
1114         .statfs         = f2fs_statfs,
1115         .remount_fs     = f2fs_remount,
1116 };
1117
1118 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1119 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1120 {
1121         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1122                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1123                                 ctx, len, NULL);
1124 }
1125
1126 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1127 {
1128         *key = F2FS_I_SB(inode)->key_prefix;
1129         return F2FS_I_SB(inode)->key_prefix_size;
1130 }
1131
1132 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1133                                                         void *fs_data)
1134 {
1135         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1136                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1137                                 ctx, len, fs_data, XATTR_CREATE);
1138 }
1139
1140 static unsigned f2fs_max_namelen(struct inode *inode)
1141 {
1142         return S_ISLNK(inode->i_mode) ?
1143                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1144 }
1145
1146 static struct fscrypt_operations f2fs_cryptops = {
1147         .get_context    = f2fs_get_context,
1148         .key_prefix     = f2fs_key_prefix,
1149         .set_context    = f2fs_set_context,
1150         .is_encrypted   = f2fs_encrypted_inode,
1151         .empty_dir      = f2fs_empty_dir,
1152         .max_namelen    = f2fs_max_namelen,
1153 };
1154 #else
1155 static struct fscrypt_operations f2fs_cryptops = {
1156         .is_encrypted   = f2fs_encrypted_inode,
1157 };
1158 #endif
1159
1160 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1161                 u64 ino, u32 generation)
1162 {
1163         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1164         struct inode *inode;
1165
1166         if (check_nid_range(sbi, ino))
1167                 return ERR_PTR(-ESTALE);
1168
1169         /*
1170          * f2fs_iget isn't quite right if the inode is currently unallocated!
1171          * However f2fs_iget currently does appropriate checks to handle stale
1172          * inodes so everything is OK.
1173          */
1174         inode = f2fs_iget(sb, ino);
1175         if (IS_ERR(inode))
1176                 return ERR_CAST(inode);
1177         if (unlikely(generation && inode->i_generation != generation)) {
1178                 /* we didn't find the right inode.. */
1179                 iput(inode);
1180                 return ERR_PTR(-ESTALE);
1181         }
1182         return inode;
1183 }
1184
1185 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1186                 int fh_len, int fh_type)
1187 {
1188         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1189                                     f2fs_nfs_get_inode);
1190 }
1191
1192 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1193                 int fh_len, int fh_type)
1194 {
1195         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1196                                     f2fs_nfs_get_inode);
1197 }
1198
1199 static const struct export_operations f2fs_export_ops = {
1200         .fh_to_dentry = f2fs_fh_to_dentry,
1201         .fh_to_parent = f2fs_fh_to_parent,
1202         .get_parent = f2fs_get_parent,
1203 };
1204
1205 static loff_t max_file_blocks(void)
1206 {
1207         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1208         loff_t leaf_count = ADDRS_PER_BLOCK;
1209
1210         /* two direct node blocks */
1211         result += (leaf_count * 2);
1212
1213         /* two indirect node blocks */
1214         leaf_count *= NIDS_PER_BLOCK;
1215         result += (leaf_count * 2);
1216
1217         /* one double indirect node block */
1218         leaf_count *= NIDS_PER_BLOCK;
1219         result += leaf_count;
1220
1221         return result;
1222 }
1223
1224 static int __f2fs_commit_super(struct buffer_head *bh,
1225                         struct f2fs_super_block *super)
1226 {
1227         lock_buffer(bh);
1228         if (super)
1229                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1230         set_buffer_uptodate(bh);
1231         set_buffer_dirty(bh);
1232         unlock_buffer(bh);
1233
1234         /* it's rare case, we can do fua all the time */
1235         return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1236 }
1237
1238 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1239                                         struct buffer_head *bh)
1240 {
1241         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1242                                         (bh->b_data + F2FS_SUPER_OFFSET);
1243         struct super_block *sb = sbi->sb;
1244         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1245         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1246         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1247         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1248         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1249         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1250         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1251         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1252         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1253         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1254         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1255         u32 segment_count = le32_to_cpu(raw_super->segment_count);
1256         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1257         u64 main_end_blkaddr = main_blkaddr +
1258                                 (segment_count_main << log_blocks_per_seg);
1259         u64 seg_end_blkaddr = segment0_blkaddr +
1260                                 (segment_count << log_blocks_per_seg);
1261
1262         if (segment0_blkaddr != cp_blkaddr) {
1263                 f2fs_msg(sb, KERN_INFO,
1264                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1265                         segment0_blkaddr, cp_blkaddr);
1266                 return true;
1267         }
1268
1269         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1270                                                         sit_blkaddr) {
1271                 f2fs_msg(sb, KERN_INFO,
1272                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1273                         cp_blkaddr, sit_blkaddr,
1274                         segment_count_ckpt << log_blocks_per_seg);
1275                 return true;
1276         }
1277
1278         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1279                                                         nat_blkaddr) {
1280                 f2fs_msg(sb, KERN_INFO,
1281                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1282                         sit_blkaddr, nat_blkaddr,
1283                         segment_count_sit << log_blocks_per_seg);
1284                 return true;
1285         }
1286
1287         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1288                                                         ssa_blkaddr) {
1289                 f2fs_msg(sb, KERN_INFO,
1290                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1291                         nat_blkaddr, ssa_blkaddr,
1292                         segment_count_nat << log_blocks_per_seg);
1293                 return true;
1294         }
1295
1296         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1297                                                         main_blkaddr) {
1298                 f2fs_msg(sb, KERN_INFO,
1299                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1300                         ssa_blkaddr, main_blkaddr,
1301                         segment_count_ssa << log_blocks_per_seg);
1302                 return true;
1303         }
1304
1305         if (main_end_blkaddr > seg_end_blkaddr) {
1306                 f2fs_msg(sb, KERN_INFO,
1307                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1308                         main_blkaddr,
1309                         segment0_blkaddr +
1310                                 (segment_count << log_blocks_per_seg),
1311                         segment_count_main << log_blocks_per_seg);
1312                 return true;
1313         } else if (main_end_blkaddr < seg_end_blkaddr) {
1314                 int err = 0;
1315                 char *res;
1316
1317                 /* fix in-memory information all the time */
1318                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1319                                 segment0_blkaddr) >> log_blocks_per_seg);
1320
1321                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1322                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1323                         res = "internally";
1324                 } else {
1325                         err = __f2fs_commit_super(bh, NULL);
1326                         res = err ? "failed" : "done";
1327                 }
1328                 f2fs_msg(sb, KERN_INFO,
1329                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
1330                         res, main_blkaddr,
1331                         segment0_blkaddr +
1332                                 (segment_count << log_blocks_per_seg),
1333                         segment_count_main << log_blocks_per_seg);
1334                 if (err)
1335                         return true;
1336         }
1337         return false;
1338 }
1339
1340 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1341                                 struct buffer_head *bh)
1342 {
1343         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1344                                         (bh->b_data + F2FS_SUPER_OFFSET);
1345         struct super_block *sb = sbi->sb;
1346         unsigned int blocksize;
1347
1348         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1349                 f2fs_msg(sb, KERN_INFO,
1350                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1351                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1352                 return 1;
1353         }
1354
1355         /* Currently, support only 4KB page cache size */
1356         if (F2FS_BLKSIZE != PAGE_SIZE) {
1357                 f2fs_msg(sb, KERN_INFO,
1358                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1359                         PAGE_SIZE);
1360                 return 1;
1361         }
1362
1363         /* Currently, support only 4KB block size */
1364         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1365         if (blocksize != F2FS_BLKSIZE) {
1366                 f2fs_msg(sb, KERN_INFO,
1367                         "Invalid blocksize (%u), supports only 4KB\n",
1368                         blocksize);
1369                 return 1;
1370         }
1371
1372         /* check log blocks per segment */
1373         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1374                 f2fs_msg(sb, KERN_INFO,
1375                         "Invalid log blocks per segment (%u)\n",
1376                         le32_to_cpu(raw_super->log_blocks_per_seg));
1377                 return 1;
1378         }
1379
1380         /* Currently, support 512/1024/2048/4096 bytes sector size */
1381         if (le32_to_cpu(raw_super->log_sectorsize) >
1382                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1383                 le32_to_cpu(raw_super->log_sectorsize) <
1384                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1385                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1386                         le32_to_cpu(raw_super->log_sectorsize));
1387                 return 1;
1388         }
1389         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1390                 le32_to_cpu(raw_super->log_sectorsize) !=
1391                         F2FS_MAX_LOG_SECTOR_SIZE) {
1392                 f2fs_msg(sb, KERN_INFO,
1393                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1394                         le32_to_cpu(raw_super->log_sectors_per_block),
1395                         le32_to_cpu(raw_super->log_sectorsize));
1396                 return 1;
1397         }
1398
1399         /* check reserved ino info */
1400         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1401                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1402                 le32_to_cpu(raw_super->root_ino) != 3) {
1403                 f2fs_msg(sb, KERN_INFO,
1404                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1405                         le32_to_cpu(raw_super->node_ino),
1406                         le32_to_cpu(raw_super->meta_ino),
1407                         le32_to_cpu(raw_super->root_ino));
1408                 return 1;
1409         }
1410
1411         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1412         if (sanity_check_area_boundary(sbi, bh))
1413                 return 1;
1414
1415         return 0;
1416 }
1417
1418 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1419 {
1420         unsigned int total, fsmeta;
1421         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1422         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1423
1424         total = le32_to_cpu(raw_super->segment_count);
1425         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1426         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1427         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1428         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1429         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1430
1431         if (unlikely(fsmeta >= total))
1432                 return 1;
1433
1434         if (unlikely(f2fs_cp_error(sbi))) {
1435                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1436                 return 1;
1437         }
1438         return 0;
1439 }
1440
1441 static void init_sb_info(struct f2fs_sb_info *sbi)
1442 {
1443         struct f2fs_super_block *raw_super = sbi->raw_super;
1444
1445         sbi->log_sectors_per_block =
1446                 le32_to_cpu(raw_super->log_sectors_per_block);
1447         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1448         sbi->blocksize = 1 << sbi->log_blocksize;
1449         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1450         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1451         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1452         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1453         sbi->total_sections = le32_to_cpu(raw_super->section_count);
1454         sbi->total_node_count =
1455                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1456                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1457         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1458         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1459         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1460         sbi->cur_victim_sec = NULL_SECNO;
1461         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1462
1463         sbi->dir_level = DEF_DIR_LEVEL;
1464         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1465         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1466         clear_sbi_flag(sbi, SBI_NEED_FSCK);
1467
1468         INIT_LIST_HEAD(&sbi->s_list);
1469         mutex_init(&sbi->umount_mutex);
1470         mutex_init(&sbi->wio_mutex[NODE]);
1471         mutex_init(&sbi->wio_mutex[DATA]);
1472
1473 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1474         memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1475                                 F2FS_KEY_DESC_PREFIX_SIZE);
1476         sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1477 #endif
1478 }
1479
1480 static int init_percpu_info(struct f2fs_sb_info *sbi)
1481 {
1482         int i, err;
1483
1484         for (i = 0; i < NR_COUNT_TYPE; i++) {
1485                 err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1486                 if (err)
1487                         return err;
1488         }
1489
1490         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1491         if (err)
1492                 return err;
1493
1494         return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1495                                                                 GFP_KERNEL);
1496 }
1497
1498 /*
1499  * Read f2fs raw super block.
1500  * Because we have two copies of super block, so read both of them
1501  * to get the first valid one. If any one of them is broken, we pass
1502  * them recovery flag back to the caller.
1503  */
1504 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1505                         struct f2fs_super_block **raw_super,
1506                         int *valid_super_block, int *recovery)
1507 {
1508         struct super_block *sb = sbi->sb;
1509         int block;
1510         struct buffer_head *bh;
1511         struct f2fs_super_block *super;
1512         int err = 0;
1513
1514         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1515         if (!super)
1516                 return -ENOMEM;
1517
1518         for (block = 0; block < 2; block++) {
1519                 bh = sb_bread(sb, block);
1520                 if (!bh) {
1521                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1522                                 block + 1);
1523                         err = -EIO;
1524                         continue;
1525                 }
1526
1527                 /* sanity checking of raw super */
1528                 if (sanity_check_raw_super(sbi, bh)) {
1529                         f2fs_msg(sb, KERN_ERR,
1530                                 "Can't find valid F2FS filesystem in %dth superblock",
1531                                 block + 1);
1532                         err = -EINVAL;
1533                         brelse(bh);
1534                         continue;
1535                 }
1536
1537                 if (!*raw_super) {
1538                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1539                                                         sizeof(*super));
1540                         *valid_super_block = block;
1541                         *raw_super = super;
1542                 }
1543                 brelse(bh);
1544         }
1545
1546         /* Fail to read any one of the superblocks*/
1547         if (err < 0)
1548                 *recovery = 1;
1549
1550         /* No valid superblock */
1551         if (!*raw_super)
1552                 kfree(super);
1553         else
1554                 err = 0;
1555
1556         return err;
1557 }
1558
1559 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1560 {
1561         struct buffer_head *bh;
1562         int err;
1563
1564         if ((recover && f2fs_readonly(sbi->sb)) ||
1565                                 bdev_read_only(sbi->sb->s_bdev)) {
1566                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1567                 return -EROFS;
1568         }
1569
1570         /* write back-up superblock first */
1571         bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1572         if (!bh)
1573                 return -EIO;
1574         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1575         brelse(bh);
1576
1577         /* if we are in recovery path, skip writing valid superblock */
1578         if (recover || err)
1579                 return err;
1580
1581         /* write current valid superblock */
1582         bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1583         if (!bh)
1584                 return -EIO;
1585         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1586         brelse(bh);
1587         return err;
1588 }
1589
1590 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1591 {
1592         struct f2fs_sb_info *sbi;
1593         struct f2fs_super_block *raw_super;
1594         struct inode *root;
1595         int err;
1596         bool retry = true, need_fsck = false;
1597         char *options = NULL;
1598         int recovery, i, valid_super_block;
1599         struct curseg_info *seg_i;
1600
1601 try_onemore:
1602         err = -EINVAL;
1603         raw_super = NULL;
1604         valid_super_block = -1;
1605         recovery = 0;
1606
1607         /* allocate memory for f2fs-specific super block info */
1608         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1609         if (!sbi)
1610                 return -ENOMEM;
1611
1612         sbi->sb = sb;
1613
1614         /* Load the checksum driver */
1615         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1616         if (IS_ERR(sbi->s_chksum_driver)) {
1617                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1618                 err = PTR_ERR(sbi->s_chksum_driver);
1619                 sbi->s_chksum_driver = NULL;
1620                 goto free_sbi;
1621         }
1622
1623         /* set a block size */
1624         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1625                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1626                 goto free_sbi;
1627         }
1628
1629         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1630                                                                 &recovery);
1631         if (err)
1632                 goto free_sbi;
1633
1634         sb->s_fs_info = sbi;
1635         sbi->raw_super = raw_super;
1636
1637         default_options(sbi);
1638         /* parse mount options */
1639         options = kstrdup((const char *)data, GFP_KERNEL);
1640         if (data && !options) {
1641                 err = -ENOMEM;
1642                 goto free_sb_buf;
1643         }
1644
1645         err = parse_options(sb, options);
1646         if (err)
1647                 goto free_options;
1648
1649         sbi->max_file_blocks = max_file_blocks();
1650         sb->s_maxbytes = sbi->max_file_blocks <<
1651                                 le32_to_cpu(raw_super->log_blocksize);
1652         sb->s_max_links = F2FS_LINK_MAX;
1653         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1654
1655         sb->s_op = &f2fs_sops;
1656         sb->s_cop = &f2fs_cryptops;
1657         sb->s_xattr = f2fs_xattr_handlers;
1658         sb->s_export_op = &f2fs_export_ops;
1659         sb->s_magic = F2FS_SUPER_MAGIC;
1660         sb->s_time_gran = 1;
1661         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1662                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1663         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1664
1665         /* init f2fs-specific super block info */
1666         sbi->valid_super_block = valid_super_block;
1667         mutex_init(&sbi->gc_mutex);
1668         mutex_init(&sbi->cp_mutex);
1669         init_rwsem(&sbi->node_write);
1670
1671         /* disallow all the data/node/meta page writes */
1672         set_sbi_flag(sbi, SBI_POR_DOING);
1673         spin_lock_init(&sbi->stat_lock);
1674
1675         init_rwsem(&sbi->read_io.io_rwsem);
1676         sbi->read_io.sbi = sbi;
1677         sbi->read_io.bio = NULL;
1678         for (i = 0; i < NR_PAGE_TYPE; i++) {
1679                 init_rwsem(&sbi->write_io[i].io_rwsem);
1680                 sbi->write_io[i].sbi = sbi;
1681                 sbi->write_io[i].bio = NULL;
1682         }
1683
1684         init_rwsem(&sbi->cp_rwsem);
1685         init_waitqueue_head(&sbi->cp_wait);
1686         init_sb_info(sbi);
1687
1688         err = init_percpu_info(sbi);
1689         if (err)
1690                 goto free_options;
1691
1692         /* get an inode for meta space */
1693         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1694         if (IS_ERR(sbi->meta_inode)) {
1695                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1696                 err = PTR_ERR(sbi->meta_inode);
1697                 goto free_options;
1698         }
1699
1700         err = get_valid_checkpoint(sbi);
1701         if (err) {
1702                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1703                 goto free_meta_inode;
1704         }
1705
1706         sbi->total_valid_node_count =
1707                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1708         percpu_counter_set(&sbi->total_valid_inode_count,
1709                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
1710         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1711         sbi->total_valid_block_count =
1712                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1713         sbi->last_valid_block_count = sbi->total_valid_block_count;
1714
1715         for (i = 0; i < NR_INODE_TYPE; i++) {
1716                 INIT_LIST_HEAD(&sbi->inode_list[i]);
1717                 spin_lock_init(&sbi->inode_lock[i]);
1718         }
1719
1720         init_extent_cache_info(sbi);
1721
1722         init_ino_entry_info(sbi);
1723
1724         /* setup f2fs internal modules */
1725         err = build_segment_manager(sbi);
1726         if (err) {
1727                 f2fs_msg(sb, KERN_ERR,
1728                         "Failed to initialize F2FS segment manager");
1729                 goto free_sm;
1730         }
1731         err = build_node_manager(sbi);
1732         if (err) {
1733                 f2fs_msg(sb, KERN_ERR,
1734                         "Failed to initialize F2FS node manager");
1735                 goto free_nm;
1736         }
1737
1738         /* For write statistics */
1739         if (sb->s_bdev->bd_part)
1740                 sbi->sectors_written_start =
1741                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1742
1743         /* Read accumulated write IO statistics if exists */
1744         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1745         if (__exist_node_summaries(sbi))
1746                 sbi->kbytes_written =
1747                         le64_to_cpu(seg_i->journal->info.kbytes_written);
1748
1749         build_gc_manager(sbi);
1750
1751         /* get an inode for node space */
1752         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1753         if (IS_ERR(sbi->node_inode)) {
1754                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1755                 err = PTR_ERR(sbi->node_inode);
1756                 goto free_nm;
1757         }
1758
1759         f2fs_join_shrinker(sbi);
1760
1761         /* if there are nt orphan nodes free them */
1762         err = recover_orphan_inodes(sbi);
1763         if (err)
1764                 goto free_node_inode;
1765
1766         /* read root inode and dentry */
1767         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1768         if (IS_ERR(root)) {
1769                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1770                 err = PTR_ERR(root);
1771                 goto free_node_inode;
1772         }
1773         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1774                 iput(root);
1775                 err = -EINVAL;
1776                 goto free_node_inode;
1777         }
1778
1779         sb->s_root = d_make_root(root); /* allocate root dentry */
1780         if (!sb->s_root) {
1781                 err = -ENOMEM;
1782                 goto free_root_inode;
1783         }
1784
1785         err = f2fs_build_stats(sbi);
1786         if (err)
1787                 goto free_root_inode;
1788
1789         if (f2fs_proc_root)
1790                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1791
1792         if (sbi->s_proc) {
1793                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1794                                  &f2fs_seq_segment_info_fops, sb);
1795                 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1796                                  &f2fs_seq_segment_bits_fops, sb);
1797         }
1798
1799         sbi->s_kobj.kset = f2fs_kset;
1800         init_completion(&sbi->s_kobj_unregister);
1801         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1802                                                         "%s", sb->s_id);
1803         if (err)
1804                 goto free_proc;
1805
1806         /* recover fsynced data */
1807         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1808                 /*
1809                  * mount should be failed, when device has readonly mode, and
1810                  * previous checkpoint was not done by clean system shutdown.
1811                  */
1812                 if (bdev_read_only(sb->s_bdev) &&
1813                                 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1814                         err = -EROFS;
1815                         goto free_kobj;
1816                 }
1817
1818                 if (need_fsck)
1819                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1820
1821                 err = recover_fsync_data(sbi, false);
1822                 if (err < 0) {
1823                         need_fsck = true;
1824                         f2fs_msg(sb, KERN_ERR,
1825                                 "Cannot recover all fsync data errno=%d", err);
1826                         goto free_kobj;
1827                 }
1828         } else {
1829                 err = recover_fsync_data(sbi, true);
1830
1831                 if (!f2fs_readonly(sb) && err > 0) {
1832                         err = -EINVAL;
1833                         f2fs_msg(sb, KERN_ERR,
1834                                 "Need to recover fsync data");
1835                         goto free_kobj;
1836                 }
1837         }
1838
1839         /* recover_fsync_data() cleared this already */
1840         clear_sbi_flag(sbi, SBI_POR_DOING);
1841
1842         /*
1843          * If filesystem is not mounted as read-only then
1844          * do start the gc_thread.
1845          */
1846         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1847                 /* After POR, we can run background GC thread.*/
1848                 err = start_gc_thread(sbi);
1849                 if (err)
1850                         goto free_kobj;
1851         }
1852         kfree(options);
1853
1854         /* recover broken superblock */
1855         if (recovery) {
1856                 err = f2fs_commit_super(sbi, true);
1857                 f2fs_msg(sb, KERN_INFO,
1858                         "Try to recover %dth superblock, ret: %d",
1859                         sbi->valid_super_block ? 1 : 2, err);
1860         }
1861
1862         f2fs_update_time(sbi, CP_TIME);
1863         f2fs_update_time(sbi, REQ_TIME);
1864         return 0;
1865
1866 free_kobj:
1867         f2fs_sync_inode_meta(sbi);
1868         kobject_del(&sbi->s_kobj);
1869         kobject_put(&sbi->s_kobj);
1870         wait_for_completion(&sbi->s_kobj_unregister);
1871 free_proc:
1872         if (sbi->s_proc) {
1873                 remove_proc_entry("segment_info", sbi->s_proc);
1874                 remove_proc_entry("segment_bits", sbi->s_proc);
1875                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1876         }
1877         f2fs_destroy_stats(sbi);
1878 free_root_inode:
1879         dput(sb->s_root);
1880         sb->s_root = NULL;
1881 free_node_inode:
1882         mutex_lock(&sbi->umount_mutex);
1883         f2fs_leave_shrinker(sbi);
1884         iput(sbi->node_inode);
1885         mutex_unlock(&sbi->umount_mutex);
1886 free_nm:
1887         destroy_node_manager(sbi);
1888 free_sm:
1889         destroy_segment_manager(sbi);
1890         kfree(sbi->ckpt);
1891 free_meta_inode:
1892         make_bad_inode(sbi->meta_inode);
1893         iput(sbi->meta_inode);
1894 free_options:
1895         destroy_percpu_info(sbi);
1896         kfree(options);
1897 free_sb_buf:
1898         kfree(raw_super);
1899 free_sbi:
1900         if (sbi->s_chksum_driver)
1901                 crypto_free_shash(sbi->s_chksum_driver);
1902         kfree(sbi);
1903
1904         /* give only one another chance */
1905         if (retry) {
1906                 retry = false;
1907                 shrink_dcache_sb(sb);
1908                 goto try_onemore;
1909         }
1910         return err;
1911 }
1912
1913 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1914                         const char *dev_name, void *data)
1915 {
1916         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1917 }
1918
1919 static void kill_f2fs_super(struct super_block *sb)
1920 {
1921         if (sb->s_root)
1922                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1923         kill_block_super(sb);
1924 }
1925
1926 static struct file_system_type f2fs_fs_type = {
1927         .owner          = THIS_MODULE,
1928         .name           = "f2fs",
1929         .mount          = f2fs_mount,
1930         .kill_sb        = kill_f2fs_super,
1931         .fs_flags       = FS_REQUIRES_DEV,
1932 };
1933 MODULE_ALIAS_FS("f2fs");
1934
1935 static int __init init_inodecache(void)
1936 {
1937         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1938                         sizeof(struct f2fs_inode_info), 0,
1939                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1940         if (!f2fs_inode_cachep)
1941                 return -ENOMEM;
1942         return 0;
1943 }
1944
1945 static void destroy_inodecache(void)
1946 {
1947         /*
1948          * Make sure all delayed rcu free inodes are flushed before we
1949          * destroy cache.
1950          */
1951         rcu_barrier();
1952         kmem_cache_destroy(f2fs_inode_cachep);
1953 }
1954
1955 static int __init init_f2fs_fs(void)
1956 {
1957         int err;
1958
1959         f2fs_build_trace_ios();
1960
1961         err = init_inodecache();
1962         if (err)
1963                 goto fail;
1964         err = create_node_manager_caches();
1965         if (err)
1966                 goto free_inodecache;
1967         err = create_segment_manager_caches();
1968         if (err)
1969                 goto free_node_manager_caches;
1970         err = create_checkpoint_caches();
1971         if (err)
1972                 goto free_segment_manager_caches;
1973         err = create_extent_cache();
1974         if (err)
1975                 goto free_checkpoint_caches;
1976         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1977         if (!f2fs_kset) {
1978                 err = -ENOMEM;
1979                 goto free_extent_cache;
1980         }
1981 #ifdef CONFIG_F2FS_FAULT_INJECTION
1982         f2fs_fault_inject.kset = f2fs_kset;
1983         f2fs_build_fault_attr(0);
1984         err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1985                                 NULL, "fault_injection");
1986         if (err) {
1987                 f2fs_fault_inject.kset = NULL;
1988                 goto free_kset;
1989         }
1990 #endif
1991         err = register_shrinker(&f2fs_shrinker_info);
1992         if (err)
1993                 goto free_kset;
1994
1995         err = register_filesystem(&f2fs_fs_type);
1996         if (err)
1997                 goto free_shrinker;
1998         err = f2fs_create_root_stats();
1999         if (err)
2000                 goto free_filesystem;
2001         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2002         return 0;
2003
2004 free_filesystem:
2005         unregister_filesystem(&f2fs_fs_type);
2006 free_shrinker:
2007         unregister_shrinker(&f2fs_shrinker_info);
2008 free_kset:
2009 #ifdef CONFIG_F2FS_FAULT_INJECTION
2010         if (f2fs_fault_inject.kset)
2011                 kobject_put(&f2fs_fault_inject);
2012 #endif
2013         kset_unregister(f2fs_kset);
2014 free_extent_cache:
2015         destroy_extent_cache();
2016 free_checkpoint_caches:
2017         destroy_checkpoint_caches();
2018 free_segment_manager_caches:
2019         destroy_segment_manager_caches();
2020 free_node_manager_caches:
2021         destroy_node_manager_caches();
2022 free_inodecache:
2023         destroy_inodecache();
2024 fail:
2025         return err;
2026 }
2027
2028 static void __exit exit_f2fs_fs(void)
2029 {
2030         remove_proc_entry("fs/f2fs", NULL);
2031         f2fs_destroy_root_stats();
2032         unregister_filesystem(&f2fs_fs_type);
2033         unregister_shrinker(&f2fs_shrinker_info);
2034 #ifdef CONFIG_F2FS_FAULT_INJECTION
2035         kobject_put(&f2fs_fault_inject);
2036 #endif
2037         kset_unregister(f2fs_kset);
2038         destroy_extent_cache();
2039         destroy_checkpoint_caches();
2040         destroy_segment_manager_caches();
2041         destroy_node_manager_caches();
2042         destroy_inodecache();
2043         f2fs_destroy_trace_ios();
2044 }
2045
2046 module_init(init_f2fs_fs)
2047 module_exit(exit_f2fs_fs)
2048
2049 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2050 MODULE_DESCRIPTION("Flash Friendly File System");
2051 MODULE_LICENSE("GPL");