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         percpu_free_rwsem(&sbi->cp_rwsem);
 711 }
 712
 713 static void f2fs_put_super(struct super_block *sb)
 714 {
 715         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 716
 717         if (sbi->s_proc) {
 718                 remove_proc_entry("segment_info", sbi->s_proc);
 719                 remove_proc_entry("segment_bits", sbi->s_proc);
 720                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 721         }
 722         kobject_del(&sbi->s_kobj);
 723
 724         stop_gc_thread(sbi);
 725
 726         /* prevent remaining shrinker jobs */
 727         mutex_lock(&sbi->umount_mutex);
 728
 729         /*
 730          * We don't need to do checkpoint when superblock is clean.
 731          * But, the previous checkpoint was not done by umount, it needs to do
 732          * clean checkpoint again.
 733          */
 734         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 735                         !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
 736                 struct cp_control cpc = {
 737                         .reason = CP_UMOUNT,
 738                 };
 739                 write_checkpoint(sbi, &cpc);
 740         }
 741
 742         /* write_checkpoint can update stat informaion */
 743         f2fs_destroy_stats(sbi);
 744
 745         /*
 746          * normally superblock is clean, so we need to release this.
 747          * In addition, EIO will skip do checkpoint, we need this as well.
 748          */
 749         release_ino_entry(sbi, true);
 750         release_discard_addrs(sbi);
 751
 752         f2fs_leave_shrinker(sbi);
 753         mutex_unlock(&sbi->umount_mutex);
 754
 755         /* our cp_error case, we can wait for any writeback page */
 756         f2fs_flush_merged_bios(sbi);
 757
 758         iput(sbi->node_inode);
 759         iput(sbi->meta_inode);
 760
 761         /* destroy f2fs internal modules */
 762         destroy_node_manager(sbi);
 763         destroy_segment_manager(sbi);
 764
 765         kfree(sbi->ckpt);
 766         kobject_put(&sbi->s_kobj);
 767         wait_for_completion(&sbi->s_kobj_unregister);
 768
 769         sb->s_fs_info = NULL;
 770         if (sbi->s_chksum_driver)
 771                 crypto_free_shash(sbi->s_chksum_driver);
 772         kfree(sbi->raw_super);
 773
 774         destroy_percpu_info(sbi);
 775         kfree(sbi);
 776 }
 777
 778 int f2fs_sync_fs(struct super_block *sb, int sync)
 779 {
 780         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 781         int err = 0;
 782
 783         trace_f2fs_sync_fs(sb, sync);
 784
 785         if (sync) {
 786                 struct cp_control cpc;
 787
 788                 cpc.reason = __get_cp_reason(sbi);
 789
 790                 mutex_lock(&sbi->gc_mutex);
 791                 err = write_checkpoint(sbi, &cpc);
 792                 mutex_unlock(&sbi->gc_mutex);
 793         }
 794         f2fs_trace_ios(NULL, 1);
 795
 796         return err;
 797 }
 798
 799 static int f2fs_freeze(struct super_block *sb)
 800 {
 801         int err;
 802
 803         if (f2fs_readonly(sb))
 804                 return 0;
 805
 806         err = f2fs_sync_fs(sb, 1);
 807         return err;
 808 }
 809
 810 static int f2fs_unfreeze(struct super_block *sb)
 811 {
 812         return 0;
 813 }
 814
 815 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 816 {
 817         struct super_block *sb = dentry->d_sb;
 818         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 819         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 820         block_t total_count, user_block_count, start_count, ovp_count;
 821
 822         total_count = le64_to_cpu(sbi->raw_super->block_count);
 823         user_block_count = sbi->user_block_count;
 824         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 825         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 826         buf->f_type = F2FS_SUPER_MAGIC;
 827         buf->f_bsize = sbi->blocksize;
 828
 829         buf->f_blocks = total_count - start_count;
 830         buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
 831         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 832
 833         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 834         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
 835
 836         buf->f_namelen = F2FS_NAME_LEN;
 837         buf->f_fsid.val[0] = (u32)id;
 838         buf->f_fsid.val[1] = (u32)(id >> 32);
 839
 840         return 0;
 841 }
 842
 843 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 844 {
 845         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 846
 847         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
 848                 if (test_opt(sbi, FORCE_FG_GC))
 849                         seq_printf(seq, ",background_gc=%s", "sync");
 850                 else
 851                         seq_printf(seq, ",background_gc=%s", "on");
 852         } else {
 853                 seq_printf(seq, ",background_gc=%s", "off");
 854         }
 855         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 856                 seq_puts(seq, ",disable_roll_forward");
 857         if (test_opt(sbi, DISCARD))
 858                 seq_puts(seq, ",discard");
 859         if (test_opt(sbi, NOHEAP))
 860                 seq_puts(seq, ",no_heap_alloc");
 861 #ifdef CONFIG_F2FS_FS_XATTR
 862         if (test_opt(sbi, XATTR_USER))
 863                 seq_puts(seq, ",user_xattr");
 864         else
 865                 seq_puts(seq, ",nouser_xattr");
 866         if (test_opt(sbi, INLINE_XATTR))
 867                 seq_puts(seq, ",inline_xattr");
 868 #endif
 869 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 870         if (test_opt(sbi, POSIX_ACL))
 871                 seq_puts(seq, ",acl");
 872         else
 873                 seq_puts(seq, ",noacl");
 874 #endif
 875         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 876                 seq_puts(seq, ",disable_ext_identify");
 877         if (test_opt(sbi, INLINE_DATA))
 878                 seq_puts(seq, ",inline_data");
 879         else
 880                 seq_puts(seq, ",noinline_data");
 881         if (test_opt(sbi, INLINE_DENTRY))
 882                 seq_puts(seq, ",inline_dentry");
 883         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 884                 seq_puts(seq, ",flush_merge");
 885         if (test_opt(sbi, NOBARRIER))
 886                 seq_puts(seq, ",nobarrier");
 887         if (test_opt(sbi, FASTBOOT))
 888                 seq_puts(seq, ",fastboot");
 889         if (test_opt(sbi, EXTENT_CACHE))
 890                 seq_puts(seq, ",extent_cache");
 891         else
 892                 seq_puts(seq, ",noextent_cache");
 893         if (test_opt(sbi, DATA_FLUSH))
 894                 seq_puts(seq, ",data_flush");
 895
 896         seq_puts(seq, ",mode=");
 897         if (test_opt(sbi, ADAPTIVE))
 898                 seq_puts(seq, "adaptive");
 899         else if (test_opt(sbi, LFS))
 900                 seq_puts(seq, "lfs");
 901         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 902
 903         return 0;
 904 }
 905
 906 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 907 {
 908         struct super_block *sb = seq->private;
 909         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 910         unsigned int total_segs =
 911                         le32_to_cpu(sbi->raw_super->segment_count_main);
 912         int i;
 913
 914         seq_puts(seq, "format: segment_type|valid_blocks\n"
 915                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 916
 917         for (i = 0; i < total_segs; i++) {
 918                 struct seg_entry *se = get_seg_entry(sbi, i);
 919
 920                 if ((i % 10) == 0)
 921                         seq_printf(seq, "%-10d", i);
 922                 seq_printf(seq, "%d|%-3u", se->type,
 923                                         get_valid_blocks(sbi, i, 1));
 924                 if ((i % 10) == 9 || i == (total_segs - 1))
 925                         seq_putc(seq, '\n');
 926                 else
 927                         seq_putc(seq, ' ');
 928         }
 929
 930         return 0;
 931 }
 932
 933 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
 934 {
 935         struct super_block *sb = seq->private;
 936         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 937         unsigned int total_segs =
 938                         le32_to_cpu(sbi->raw_super->segment_count_main);
 939         int i, j;
 940
 941         seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
 942                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 943
 944         for (i = 0; i < total_segs; i++) {
 945                 struct seg_entry *se = get_seg_entry(sbi, i);
 946
 947                 seq_printf(seq, "%-10d", i);
 948                 seq_printf(seq, "%d|%-3u|", se->type,
 949                                         get_valid_blocks(sbi, i, 1));
 950                 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
 951                         seq_printf(seq, "%x ", se->cur_valid_map[j]);
 952                 seq_putc(seq, '\n');
 953         }
 954         return 0;
 955 }
 956
 957 #define F2FS_PROC_FILE_DEF(_name)                                       \
 958 static int _name##_open_fs(struct inode *inode, struct file *file)      \
 959 {                                                                       \
 960         return single_open(file, _name##_seq_show, PDE_DATA(inode));    \
 961 }                                                                       \
 962                                                                         \
 963 static const struct file_operations f2fs_seq_##_name##_fops = {         \
 964         .open = _name##_open_fs,                                        \
 965         .read = seq_read,                                               \
 966         .llseek = seq_lseek,                                            \
 967         .release = single_release,                                      \
 968 };
 969
 970 F2FS_PROC_FILE_DEF(segment_info);
 971 F2FS_PROC_FILE_DEF(segment_bits);
 972
 973 static void default_options(struct f2fs_sb_info *sbi)
 974 {
 975         /* init some FS parameters */
 976         sbi->active_logs = NR_CURSEG_TYPE;
 977
 978         set_opt(sbi, BG_GC);
 979         set_opt(sbi, INLINE_DATA);
 980         set_opt(sbi, EXTENT_CACHE);
 981         sbi->sb->s_flags |= MS_LAZYTIME;
 982         set_opt(sbi, FLUSH_MERGE);
 983         if (f2fs_sb_mounted_hmsmr(sbi->sb)) {
 984                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 985                 set_opt(sbi, DISCARD);
 986         } else {
 987                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 988         }
 989
 990 #ifdef CONFIG_F2FS_FS_XATTR
 991         set_opt(sbi, XATTR_USER);
 992 #endif
 993 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 994         set_opt(sbi, POSIX_ACL);
 995 #endif
 996 }
 997
 998 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 999 {
1000         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1001         struct f2fs_mount_info org_mount_opt;
1002         int err, active_logs;
1003         bool need_restart_gc = false;
1004         bool need_stop_gc = false;
1005         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1006
1007         /*
1008          * Save the old mount options in case we
1009          * need to restore them.
1010          */
1011         org_mount_opt = sbi->mount_opt;
1012         active_logs = sbi->active_logs;
1013
1014         /* recover superblocks we couldn't write due to previous RO mount */
1015         if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1016                 err = f2fs_commit_super(sbi, false);
1017                 f2fs_msg(sb, KERN_INFO,
1018                         "Try to recover all the superblocks, ret: %d", err);
1019                 if (!err)
1020                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1021         }
1022
1023         sbi->mount_opt.opt = 0;
1024         default_options(sbi);
1025
1026         /* parse mount options */
1027         err = parse_options(sb, data);
1028         if (err)
1029                 goto restore_opts;
1030
1031         /*
1032          * Previous and new state of filesystem is RO,
1033          * so skip checking GC and FLUSH_MERGE conditions.
1034          */
1035         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1036                 goto skip;
1037
1038         /* disallow enable/disable extent_cache dynamically */
1039         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1040                 err = -EINVAL;
1041                 f2fs_msg(sbi->sb, KERN_WARNING,
1042                                 "switch extent_cache option is not allowed");
1043                 goto restore_opts;
1044         }
1045
1046         /*
1047          * We stop the GC thread if FS is mounted as RO
1048          * or if background_gc = off is passed in mount
1049          * option. Also sync the filesystem.
1050          */
1051         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1052                 if (sbi->gc_thread) {
1053                         stop_gc_thread(sbi);
1054                         need_restart_gc = true;
1055                 }
1056         } else if (!sbi->gc_thread) {
1057                 err = start_gc_thread(sbi);
1058                 if (err)
1059                         goto restore_opts;
1060                 need_stop_gc = true;
1061         }
1062
1063         if (*flags & MS_RDONLY) {
1064                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1065                 sync_inodes_sb(sb);
1066
1067                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1068                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1069                 f2fs_sync_fs(sb, 1);
1070                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1071         }
1072
1073         /*
1074          * We stop issue flush thread if FS is mounted as RO
1075          * or if flush_merge is not passed in mount option.
1076          */
1077         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1078                 destroy_flush_cmd_control(sbi);
1079         } else if (!SM_I(sbi)->cmd_control_info) {
1080                 err = create_flush_cmd_control(sbi);
1081                 if (err)
1082                         goto restore_gc;
1083         }
1084 skip:
1085         /* Update the POSIXACL Flag */
1086         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1087                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1088
1089         return 0;
1090 restore_gc:
1091         if (need_restart_gc) {
1092                 if (start_gc_thread(sbi))
1093                         f2fs_msg(sbi->sb, KERN_WARNING,
1094                                 "background gc thread has stopped");
1095         } else if (need_stop_gc) {
1096                 stop_gc_thread(sbi);
1097         }
1098 restore_opts:
1099         sbi->mount_opt = org_mount_opt;
1100         sbi->active_logs = active_logs;
1101         return err;
1102 }
1103
1104 static struct super_operations f2fs_sops = {
1105         .alloc_inode    = f2fs_alloc_inode,
1106         .drop_inode     = f2fs_drop_inode,
1107         .destroy_inode  = f2fs_destroy_inode,
1108         .write_inode    = f2fs_write_inode,
1109         .dirty_inode    = f2fs_dirty_inode,
1110         .show_options   = f2fs_show_options,
1111         .evict_inode    = f2fs_evict_inode,
1112         .put_super      = f2fs_put_super,
1113         .sync_fs        = f2fs_sync_fs,
1114         .freeze_fs      = f2fs_freeze,
1115         .unfreeze_fs    = f2fs_unfreeze,
1116         .statfs         = f2fs_statfs,
1117         .remount_fs     = f2fs_remount,
1118 };
1119
1120 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1121 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1122 {
1123         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1124                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1125                                 ctx, len, NULL);
1126 }
1127
1128 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1129 {
1130         *key = F2FS_I_SB(inode)->key_prefix;
1131         return F2FS_I_SB(inode)->key_prefix_size;
1132 }
1133
1134 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1135                                                         void *fs_data)
1136 {
1137         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1138                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1139                                 ctx, len, fs_data, XATTR_CREATE);
1140 }
1141
1142 static unsigned f2fs_max_namelen(struct inode *inode)
1143 {
1144         return S_ISLNK(inode->i_mode) ?
1145                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1146 }
1147
1148 static struct fscrypt_operations f2fs_cryptops = {
1149         .get_context    = f2fs_get_context,
1150         .key_prefix     = f2fs_key_prefix,
1151         .set_context    = f2fs_set_context,
1152         .is_encrypted   = f2fs_encrypted_inode,
1153         .empty_dir      = f2fs_empty_dir,
1154         .max_namelen    = f2fs_max_namelen,
1155 };
1156 #else
1157 static struct fscrypt_operations f2fs_cryptops = {
1158         .is_encrypted   = f2fs_encrypted_inode,
1159 };
1160 #endif
1161
1162 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1163                 u64 ino, u32 generation)
1164 {
1165         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1166         struct inode *inode;
1167
1168         if (check_nid_range(sbi, ino))
1169                 return ERR_PTR(-ESTALE);
1170
1171         /*
1172          * f2fs_iget isn't quite right if the inode is currently unallocated!
1173          * However f2fs_iget currently does appropriate checks to handle stale
1174          * inodes so everything is OK.
1175          */
1176         inode = f2fs_iget(sb, ino);
1177         if (IS_ERR(inode))
1178                 return ERR_CAST(inode);
1179         if (unlikely(generation && inode->i_generation != generation)) {
1180                 /* we didn't find the right inode.. */
1181                 iput(inode);
1182                 return ERR_PTR(-ESTALE);
1183         }
1184         return inode;
1185 }
1186
1187 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1188                 int fh_len, int fh_type)
1189 {
1190         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1191                                     f2fs_nfs_get_inode);
1192 }
1193
1194 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1195                 int fh_len, int fh_type)
1196 {
1197         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1198                                     f2fs_nfs_get_inode);
1199 }
1200
1201 static const struct export_operations f2fs_export_ops = {
1202         .fh_to_dentry = f2fs_fh_to_dentry,
1203         .fh_to_parent = f2fs_fh_to_parent,
1204         .get_parent = f2fs_get_parent,
1205 };
1206
1207 static loff_t max_file_blocks(void)
1208 {
1209         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1210         loff_t leaf_count = ADDRS_PER_BLOCK;
1211
1212         /* two direct node blocks */
1213         result += (leaf_count * 2);
1214
1215         /* two indirect node blocks */
1216         leaf_count *= NIDS_PER_BLOCK;
1217         result += (leaf_count * 2);
1218
1219         /* one double indirect node block */
1220         leaf_count *= NIDS_PER_BLOCK;
1221         result += leaf_count;
1222
1223         return result;
1224 }
1225
1226 static int __f2fs_commit_super(struct buffer_head *bh,
1227                         struct f2fs_super_block *super)
1228 {
1229         lock_buffer(bh);
1230         if (super)
1231                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1232         set_buffer_uptodate(bh);
1233         set_buffer_dirty(bh);
1234         unlock_buffer(bh);
1235
1236         /* it's rare case, we can do fua all the time */
1237         return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1238 }
1239
1240 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1241                                         struct buffer_head *bh)
1242 {
1243         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1244                                         (bh->b_data + F2FS_SUPER_OFFSET);
1245         struct super_block *sb = sbi->sb;
1246         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1247         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1248         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1249         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1250         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1251         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1252         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1253         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1254         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1255         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1256         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1257         u32 segment_count = le32_to_cpu(raw_super->segment_count);
1258         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1259         u64 main_end_blkaddr = main_blkaddr +
1260                                 (segment_count_main << log_blocks_per_seg);
1261         u64 seg_end_blkaddr = segment0_blkaddr +
1262                                 (segment_count << log_blocks_per_seg);
1263
1264         if (segment0_blkaddr != cp_blkaddr) {
1265                 f2fs_msg(sb, KERN_INFO,
1266                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1267                         segment0_blkaddr, cp_blkaddr);
1268                 return true;
1269         }
1270
1271         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1272                                                         sit_blkaddr) {
1273                 f2fs_msg(sb, KERN_INFO,
1274                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1275                         cp_blkaddr, sit_blkaddr,
1276                         segment_count_ckpt << log_blocks_per_seg);
1277                 return true;
1278         }
1279
1280         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1281                                                         nat_blkaddr) {
1282                 f2fs_msg(sb, KERN_INFO,
1283                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1284                         sit_blkaddr, nat_blkaddr,
1285                         segment_count_sit << log_blocks_per_seg);
1286                 return true;
1287         }
1288
1289         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1290                                                         ssa_blkaddr) {
1291                 f2fs_msg(sb, KERN_INFO,
1292                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1293                         nat_blkaddr, ssa_blkaddr,
1294                         segment_count_nat << log_blocks_per_seg);
1295                 return true;
1296         }
1297
1298         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1299                                                         main_blkaddr) {
1300                 f2fs_msg(sb, KERN_INFO,
1301                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1302                         ssa_blkaddr, main_blkaddr,
1303                         segment_count_ssa << log_blocks_per_seg);
1304                 return true;
1305         }
1306
1307         if (main_end_blkaddr > seg_end_blkaddr) {
1308                 f2fs_msg(sb, KERN_INFO,
1309                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1310                         main_blkaddr,
1311                         segment0_blkaddr +
1312                                 (segment_count << log_blocks_per_seg),
1313                         segment_count_main << log_blocks_per_seg);
1314                 return true;
1315         } else if (main_end_blkaddr < seg_end_blkaddr) {
1316                 int err = 0;
1317                 char *res;
1318
1319                 /* fix in-memory information all the time */
1320                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1321                                 segment0_blkaddr) >> log_blocks_per_seg);
1322
1323                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1324                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1325                         res = "internally";
1326                 } else {
1327                         err = __f2fs_commit_super(bh, NULL);
1328                         res = err ? "failed" : "done";
1329                 }
1330                 f2fs_msg(sb, KERN_INFO,
1331                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
1332                         res, main_blkaddr,
1333                         segment0_blkaddr +
1334                                 (segment_count << log_blocks_per_seg),
1335                         segment_count_main << log_blocks_per_seg);
1336                 if (err)
1337                         return true;
1338         }
1339         return false;
1340 }
1341
1342 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1343                                 struct buffer_head *bh)
1344 {
1345         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1346                                         (bh->b_data + F2FS_SUPER_OFFSET);
1347         struct super_block *sb = sbi->sb;
1348         unsigned int blocksize;
1349
1350         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1351                 f2fs_msg(sb, KERN_INFO,
1352                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1353                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1354                 return 1;
1355         }
1356
1357         /* Currently, support only 4KB page cache size */
1358         if (F2FS_BLKSIZE != PAGE_SIZE) {
1359                 f2fs_msg(sb, KERN_INFO,
1360                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1361                         PAGE_SIZE);
1362                 return 1;
1363         }
1364
1365         /* Currently, support only 4KB block size */
1366         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1367         if (blocksize != F2FS_BLKSIZE) {
1368                 f2fs_msg(sb, KERN_INFO,
1369                         "Invalid blocksize (%u), supports only 4KB\n",
1370                         blocksize);
1371                 return 1;
1372         }
1373
1374         /* check log blocks per segment */
1375         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1376                 f2fs_msg(sb, KERN_INFO,
1377                         "Invalid log blocks per segment (%u)\n",
1378                         le32_to_cpu(raw_super->log_blocks_per_seg));
1379                 return 1;
1380         }
1381
1382         /* Currently, support 512/1024/2048/4096 bytes sector size */
1383         if (le32_to_cpu(raw_super->log_sectorsize) >
1384                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1385                 le32_to_cpu(raw_super->log_sectorsize) <
1386                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1387                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1388                         le32_to_cpu(raw_super->log_sectorsize));
1389                 return 1;
1390         }
1391         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1392                 le32_to_cpu(raw_super->log_sectorsize) !=
1393                         F2FS_MAX_LOG_SECTOR_SIZE) {
1394                 f2fs_msg(sb, KERN_INFO,
1395                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1396                         le32_to_cpu(raw_super->log_sectors_per_block),
1397                         le32_to_cpu(raw_super->log_sectorsize));
1398                 return 1;
1399         }
1400
1401         /* check reserved ino info */
1402         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1403                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1404                 le32_to_cpu(raw_super->root_ino) != 3) {
1405                 f2fs_msg(sb, KERN_INFO,
1406                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1407                         le32_to_cpu(raw_super->node_ino),
1408                         le32_to_cpu(raw_super->meta_ino),
1409                         le32_to_cpu(raw_super->root_ino));
1410                 return 1;
1411         }
1412
1413         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1414         if (sanity_check_area_boundary(sbi, bh))
1415                 return 1;
1416
1417         return 0;
1418 }
1419
1420 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1421 {
1422         unsigned int total, fsmeta;
1423         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1424         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1425
1426         total = le32_to_cpu(raw_super->segment_count);
1427         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1428         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1429         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1430         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1431         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1432
1433         if (unlikely(fsmeta >= total))
1434                 return 1;
1435
1436         if (unlikely(f2fs_cp_error(sbi))) {
1437                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1438                 return 1;
1439         }
1440         return 0;
1441 }
1442
1443 static void init_sb_info(struct f2fs_sb_info *sbi)
1444 {
1445         struct f2fs_super_block *raw_super = sbi->raw_super;
1446
1447         sbi->log_sectors_per_block =
1448                 le32_to_cpu(raw_super->log_sectors_per_block);
1449         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1450         sbi->blocksize = 1 << sbi->log_blocksize;
1451         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1452         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1453         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1454         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1455         sbi->total_sections = le32_to_cpu(raw_super->section_count);
1456         sbi->total_node_count =
1457                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1458                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1459         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1460         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1461         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1462         sbi->cur_victim_sec = NULL_SECNO;
1463         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1464
1465         sbi->dir_level = DEF_DIR_LEVEL;
1466         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1467         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1468         clear_sbi_flag(sbi, SBI_NEED_FSCK);
1469
1470         INIT_LIST_HEAD(&sbi->s_list);
1471         mutex_init(&sbi->umount_mutex);
1472         mutex_init(&sbi->wio_mutex[NODE]);
1473         mutex_init(&sbi->wio_mutex[DATA]);
1474
1475 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1476         memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1477                                 F2FS_KEY_DESC_PREFIX_SIZE);
1478         sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1479 #endif
1480 }
1481
1482 static int init_percpu_info(struct f2fs_sb_info *sbi)
1483 {
1484         int i, err;
1485
1486         if (percpu_init_rwsem(&sbi->cp_rwsem))
1487                 return -ENOMEM;
1488
1489         for (i = 0; i < NR_COUNT_TYPE; i++) {
1490                 err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1491                 if (err)
1492                         return err;
1493         }
1494
1495         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1496         if (err)
1497                 return err;
1498
1499         return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1500                                                                 GFP_KERNEL);
1501 }
1502
1503 /*
1504  * Read f2fs raw super block.
1505  * Because we have two copies of super block, so read both of them
1506  * to get the first valid one. If any one of them is broken, we pass
1507  * them recovery flag back to the caller.
1508  */
1509 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1510                         struct f2fs_super_block **raw_super,
1511                         int *valid_super_block, int *recovery)
1512 {
1513         struct super_block *sb = sbi->sb;
1514         int block;
1515         struct buffer_head *bh;
1516         struct f2fs_super_block *super;
1517         int err = 0;
1518
1519         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1520         if (!super)
1521                 return -ENOMEM;
1522
1523         for (block = 0; block < 2; block++) {
1524                 bh = sb_bread(sb, block);
1525                 if (!bh) {
1526                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1527                                 block + 1);
1528                         err = -EIO;
1529                         continue;
1530                 }
1531
1532                 /* sanity checking of raw super */
1533                 if (sanity_check_raw_super(sbi, bh)) {
1534                         f2fs_msg(sb, KERN_ERR,
1535                                 "Can't find valid F2FS filesystem in %dth superblock",
1536                                 block + 1);
1537                         err = -EINVAL;
1538                         brelse(bh);
1539                         continue;
1540                 }
1541
1542                 if (!*raw_super) {
1543                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1544                                                         sizeof(*super));
1545                         *valid_super_block = block;
1546                         *raw_super = super;
1547                 }
1548                 brelse(bh);
1549         }
1550
1551         /* Fail to read any one of the superblocks*/
1552         if (err < 0)
1553                 *recovery = 1;
1554
1555         /* No valid superblock */
1556         if (!*raw_super)
1557                 kfree(super);
1558         else
1559                 err = 0;
1560
1561         return err;
1562 }
1563
1564 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1565 {
1566         struct buffer_head *bh;
1567         int err;
1568
1569         if ((recover && f2fs_readonly(sbi->sb)) ||
1570                                 bdev_read_only(sbi->sb->s_bdev)) {
1571                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1572                 return -EROFS;
1573         }
1574
1575         /* write back-up superblock first */
1576         bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1577         if (!bh)
1578                 return -EIO;
1579         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1580         brelse(bh);
1581
1582         /* if we are in recovery path, skip writing valid superblock */
1583         if (recover || err)
1584                 return err;
1585
1586         /* write current valid superblock */
1587         bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1588         if (!bh)
1589                 return -EIO;
1590         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1591         brelse(bh);
1592         return err;
1593 }
1594
1595 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1596 {
1597         struct f2fs_sb_info *sbi;
1598         struct f2fs_super_block *raw_super;
1599         struct inode *root;
1600         int err;
1601         bool retry = true, need_fsck = false;
1602         char *options = NULL;
1603         int recovery, i, valid_super_block;
1604         struct curseg_info *seg_i;
1605
1606 try_onemore:
1607         err = -EINVAL;
1608         raw_super = NULL;
1609         valid_super_block = -1;
1610         recovery = 0;
1611
1612         /* allocate memory for f2fs-specific super block info */
1613         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1614         if (!sbi)
1615                 return -ENOMEM;
1616
1617         sbi->sb = sb;
1618
1619         /* Load the checksum driver */
1620         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1621         if (IS_ERR(sbi->s_chksum_driver)) {
1622                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1623                 err = PTR_ERR(sbi->s_chksum_driver);
1624                 sbi->s_chksum_driver = NULL;
1625                 goto free_sbi;
1626         }
1627
1628         /* set a block size */
1629         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1630                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1631                 goto free_sbi;
1632         }
1633
1634         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1635                                                                 &recovery);
1636         if (err)
1637                 goto free_sbi;
1638
1639         sb->s_fs_info = sbi;
1640         sbi->raw_super = raw_super;
1641
1642         default_options(sbi);
1643         /* parse mount options */
1644         options = kstrdup((const char *)data, GFP_KERNEL);
1645         if (data && !options) {
1646                 err = -ENOMEM;
1647                 goto free_sb_buf;
1648         }
1649
1650         err = parse_options(sb, options);
1651         if (err)
1652                 goto free_options;
1653
1654         sbi->max_file_blocks = max_file_blocks();
1655         sb->s_maxbytes = sbi->max_file_blocks <<
1656                                 le32_to_cpu(raw_super->log_blocksize);
1657         sb->s_max_links = F2FS_LINK_MAX;
1658         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1659
1660         sb->s_op = &f2fs_sops;
1661         sb->s_cop = &f2fs_cryptops;
1662         sb->s_xattr = f2fs_xattr_handlers;
1663         sb->s_export_op = &f2fs_export_ops;
1664         sb->s_magic = F2FS_SUPER_MAGIC;
1665         sb->s_time_gran = 1;
1666         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1667                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1668         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1669
1670         /* init f2fs-specific super block info */
1671         sbi->valid_super_block = valid_super_block;
1672         mutex_init(&sbi->gc_mutex);
1673         mutex_init(&sbi->cp_mutex);
1674         init_rwsem(&sbi->node_write);
1675
1676         /* disallow all the data/node/meta page writes */
1677         set_sbi_flag(sbi, SBI_POR_DOING);
1678         spin_lock_init(&sbi->stat_lock);
1679
1680         init_rwsem(&sbi->read_io.io_rwsem);
1681         sbi->read_io.sbi = sbi;
1682         sbi->read_io.bio = NULL;
1683         for (i = 0; i < NR_PAGE_TYPE; i++) {
1684                 init_rwsem(&sbi->write_io[i].io_rwsem);
1685                 sbi->write_io[i].sbi = sbi;
1686                 sbi->write_io[i].bio = NULL;
1687         }
1688
1689         init_waitqueue_head(&sbi->cp_wait);
1690         init_sb_info(sbi);
1691
1692         err = init_percpu_info(sbi);
1693         if (err)
1694                 goto free_options;
1695
1696         /* get an inode for meta space */
1697         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1698         if (IS_ERR(sbi->meta_inode)) {
1699                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1700                 err = PTR_ERR(sbi->meta_inode);
1701                 goto free_options;
1702         }
1703
1704         err = get_valid_checkpoint(sbi);
1705         if (err) {
1706                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1707                 goto free_meta_inode;
1708         }
1709
1710         sbi->total_valid_node_count =
1711                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1712         percpu_counter_set(&sbi->total_valid_inode_count,
1713                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
1714         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1715         sbi->total_valid_block_count =
1716                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1717         sbi->last_valid_block_count = sbi->total_valid_block_count;
1718
1719         for (i = 0; i < NR_INODE_TYPE; i++) {
1720                 INIT_LIST_HEAD(&sbi->inode_list[i]);
1721                 spin_lock_init(&sbi->inode_lock[i]);
1722         }
1723
1724         init_extent_cache_info(sbi);
1725
1726         init_ino_entry_info(sbi);
1727
1728         /* setup f2fs internal modules */
1729         err = build_segment_manager(sbi);
1730         if (err) {
1731                 f2fs_msg(sb, KERN_ERR,
1732                         "Failed to initialize F2FS segment manager");
1733                 goto free_sm;
1734         }
1735         err = build_node_manager(sbi);
1736         if (err) {
1737                 f2fs_msg(sb, KERN_ERR,
1738                         "Failed to initialize F2FS node manager");
1739                 goto free_nm;
1740         }
1741
1742         /* For write statistics */
1743         if (sb->s_bdev->bd_part)
1744                 sbi->sectors_written_start =
1745                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1746
1747         /* Read accumulated write IO statistics if exists */
1748         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1749         if (__exist_node_summaries(sbi))
1750                 sbi->kbytes_written =
1751                         le64_to_cpu(seg_i->journal->info.kbytes_written);
1752
1753         build_gc_manager(sbi);
1754
1755         /* get an inode for node space */
1756         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1757         if (IS_ERR(sbi->node_inode)) {
1758                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1759                 err = PTR_ERR(sbi->node_inode);
1760                 goto free_nm;
1761         }
1762
1763         f2fs_join_shrinker(sbi);
1764
1765         /* if there are nt orphan nodes free them */
1766         err = recover_orphan_inodes(sbi);
1767         if (err)
1768                 goto free_node_inode;
1769
1770         /* read root inode and dentry */
1771         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1772         if (IS_ERR(root)) {
1773                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1774                 err = PTR_ERR(root);
1775                 goto free_node_inode;
1776         }
1777         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1778                 iput(root);
1779                 err = -EINVAL;
1780                 goto free_node_inode;
1781         }
1782
1783         sb->s_root = d_make_root(root); /* allocate root dentry */
1784         if (!sb->s_root) {
1785                 err = -ENOMEM;
1786                 goto free_root_inode;
1787         }
1788
1789         err = f2fs_build_stats(sbi);
1790         if (err)
1791                 goto free_root_inode;
1792
1793         if (f2fs_proc_root)
1794                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1795
1796         if (sbi->s_proc) {
1797                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1798                                  &f2fs_seq_segment_info_fops, sb);
1799                 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1800                                  &f2fs_seq_segment_bits_fops, sb);
1801         }
1802
1803         sbi->s_kobj.kset = f2fs_kset;
1804         init_completion(&sbi->s_kobj_unregister);
1805         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1806                                                         "%s", sb->s_id);
1807         if (err)
1808                 goto free_proc;
1809
1810         /* recover fsynced data */
1811         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1812                 /*
1813                  * mount should be failed, when device has readonly mode, and
1814                  * previous checkpoint was not done by clean system shutdown.
1815                  */
1816                 if (bdev_read_only(sb->s_bdev) &&
1817                                 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1818                         err = -EROFS;
1819                         goto free_kobj;
1820                 }
1821
1822                 if (need_fsck)
1823                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1824
1825                 err = recover_fsync_data(sbi, false);
1826                 if (err < 0) {
1827                         need_fsck = true;
1828                         f2fs_msg(sb, KERN_ERR,
1829                                 "Cannot recover all fsync data errno=%d", err);
1830                         goto free_kobj;
1831                 }
1832         } else {
1833                 err = recover_fsync_data(sbi, true);
1834
1835                 if (!f2fs_readonly(sb) && err > 0) {
1836                         err = -EINVAL;
1837                         f2fs_msg(sb, KERN_ERR,
1838                                 "Need to recover fsync data");
1839                         goto free_kobj;
1840                 }
1841         }
1842
1843         /* recover_fsync_data() cleared this already */
1844         clear_sbi_flag(sbi, SBI_POR_DOING);
1845
1846         /*
1847          * If filesystem is not mounted as read-only then
1848          * do start the gc_thread.
1849          */
1850         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1851                 /* After POR, we can run background GC thread.*/
1852                 err = start_gc_thread(sbi);
1853                 if (err)
1854                         goto free_kobj;
1855         }
1856         kfree(options);
1857
1858         /* recover broken superblock */
1859         if (recovery) {
1860                 err = f2fs_commit_super(sbi, true);
1861                 f2fs_msg(sb, KERN_INFO,
1862                         "Try to recover %dth superblock, ret: %d",
1863                         sbi->valid_super_block ? 1 : 2, err);
1864         }
1865
1866         f2fs_update_time(sbi, CP_TIME);
1867         f2fs_update_time(sbi, REQ_TIME);
1868         return 0;
1869
1870 free_kobj:
1871         f2fs_sync_inode_meta(sbi);
1872         kobject_del(&sbi->s_kobj);
1873         kobject_put(&sbi->s_kobj);
1874         wait_for_completion(&sbi->s_kobj_unregister);
1875 free_proc:
1876         if (sbi->s_proc) {
1877                 remove_proc_entry("segment_info", sbi->s_proc);
1878                 remove_proc_entry("segment_bits", sbi->s_proc);
1879                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1880         }
1881         f2fs_destroy_stats(sbi);
1882 free_root_inode:
1883         dput(sb->s_root);
1884         sb->s_root = NULL;
1885 free_node_inode:
1886         mutex_lock(&sbi->umount_mutex);
1887         f2fs_leave_shrinker(sbi);
1888         iput(sbi->node_inode);
1889         mutex_unlock(&sbi->umount_mutex);
1890 free_nm:
1891         destroy_node_manager(sbi);
1892 free_sm:
1893         destroy_segment_manager(sbi);
1894         kfree(sbi->ckpt);
1895 free_meta_inode:
1896         make_bad_inode(sbi->meta_inode);
1897         iput(sbi->meta_inode);
1898 free_options:
1899         destroy_percpu_info(sbi);
1900         kfree(options);
1901 free_sb_buf:
1902         kfree(raw_super);
1903 free_sbi:
1904         if (sbi->s_chksum_driver)
1905                 crypto_free_shash(sbi->s_chksum_driver);
1906         kfree(sbi);
1907
1908         /* give only one another chance */
1909         if (retry) {
1910                 retry = false;
1911                 shrink_dcache_sb(sb);
1912                 goto try_onemore;
1913         }
1914         return err;
1915 }
1916
1917 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1918                         const char *dev_name, void *data)
1919 {
1920         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1921 }
1922
1923 static void kill_f2fs_super(struct super_block *sb)
1924 {
1925         if (sb->s_root)
1926                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1927         kill_block_super(sb);
1928 }
1929
1930 static struct file_system_type f2fs_fs_type = {
1931         .owner          = THIS_MODULE,
1932         .name           = "f2fs",
1933         .mount          = f2fs_mount,
1934         .kill_sb        = kill_f2fs_super,
1935         .fs_flags       = FS_REQUIRES_DEV,
1936 };
1937 MODULE_ALIAS_FS("f2fs");
1938
1939 static int __init init_inodecache(void)
1940 {
1941         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1942                         sizeof(struct f2fs_inode_info), 0,
1943                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1944         if (!f2fs_inode_cachep)
1945                 return -ENOMEM;
1946         return 0;
1947 }
1948
1949 static void destroy_inodecache(void)
1950 {
1951         /*
1952          * Make sure all delayed rcu free inodes are flushed before we
1953          * destroy cache.
1954          */
1955         rcu_barrier();
1956         kmem_cache_destroy(f2fs_inode_cachep);
1957 }
1958
1959 static int __init init_f2fs_fs(void)
1960 {
1961         int err;
1962
1963         f2fs_build_trace_ios();
1964
1965         err = init_inodecache();
1966         if (err)
1967                 goto fail;
1968         err = create_node_manager_caches();
1969         if (err)
1970                 goto free_inodecache;
1971         err = create_segment_manager_caches();
1972         if (err)
1973                 goto free_node_manager_caches;
1974         err = create_checkpoint_caches();
1975         if (err)
1976                 goto free_segment_manager_caches;
1977         err = create_extent_cache();
1978         if (err)
1979                 goto free_checkpoint_caches;
1980         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1981         if (!f2fs_kset) {
1982                 err = -ENOMEM;
1983                 goto free_extent_cache;
1984         }
1985 #ifdef CONFIG_F2FS_FAULT_INJECTION
1986         f2fs_fault_inject.kset = f2fs_kset;
1987         f2fs_build_fault_attr(0);
1988         err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1989                                 NULL, "fault_injection");
1990         if (err) {
1991                 f2fs_fault_inject.kset = NULL;
1992                 goto free_kset;
1993         }
1994 #endif
1995         err = register_shrinker(&f2fs_shrinker_info);
1996         if (err)
1997                 goto free_kset;
1998
1999         err = register_filesystem(&f2fs_fs_type);
2000         if (err)
2001                 goto free_shrinker;
2002         err = f2fs_create_root_stats();
2003         if (err)
2004                 goto free_filesystem;
2005         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2006         return 0;
2007
2008 free_filesystem:
2009         unregister_filesystem(&f2fs_fs_type);
2010 free_shrinker:
2011         unregister_shrinker(&f2fs_shrinker_info);
2012 free_kset:
2013 #ifdef CONFIG_F2FS_FAULT_INJECTION
2014         if (f2fs_fault_inject.kset)
2015                 kobject_put(&f2fs_fault_inject);
2016 #endif
2017         kset_unregister(f2fs_kset);
2018 free_extent_cache:
2019         destroy_extent_cache();
2020 free_checkpoint_caches:
2021         destroy_checkpoint_caches();
2022 free_segment_manager_caches:
2023         destroy_segment_manager_caches();
2024 free_node_manager_caches:
2025         destroy_node_manager_caches();
2026 free_inodecache:
2027         destroy_inodecache();
2028 fail:
2029         return err;
2030 }
2031
2032 static void __exit exit_f2fs_fs(void)
2033 {
2034         remove_proc_entry("fs/f2fs", NULL);
2035         f2fs_destroy_root_stats();
2036         unregister_filesystem(&f2fs_fs_type);
2037         unregister_shrinker(&f2fs_shrinker_info);
2038 #ifdef CONFIG_F2FS_FAULT_INJECTION
2039         kobject_put(&f2fs_fault_inject);
2040 #endif
2041         kset_unregister(f2fs_kset);
2042         destroy_extent_cache();
2043         destroy_checkpoint_caches();
2044         destroy_segment_manager_caches();
2045         destroy_node_manager_caches();
2046         destroy_inodecache();
2047         f2fs_destroy_trace_ios();
2048 }
2049
2050 module_init(init_f2fs_fs)
2051 module_exit(exit_f2fs_fs)
2052
2053 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2054 MODULE_DESCRIPTION("Flash Friendly File System");
2055 MODULE_LICENSE("GPL");