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 enum {
  43         Opt_gc_background,
  44         Opt_disable_roll_forward,
  45         Opt_norecovery,
  46         Opt_discard,
  47         Opt_noheap,
  48         Opt_user_xattr,
  49         Opt_nouser_xattr,
  50         Opt_acl,
  51         Opt_noacl,
  52         Opt_active_logs,
  53         Opt_disable_ext_identify,
  54         Opt_inline_xattr,
  55         Opt_inline_data,
  56         Opt_inline_dentry,
  57         Opt_flush_merge,
  58         Opt_nobarrier,
  59         Opt_fastboot,
  60         Opt_extent_cache,
  61         Opt_noinline_data,
  62         Opt_err,
  63 };
  64
  65 static match_table_t f2fs_tokens = {
  66         {Opt_gc_background, "background_gc=%s"},
  67         {Opt_disable_roll_forward, "disable_roll_forward"},
  68         {Opt_norecovery, "norecovery"},
  69         {Opt_discard, "discard"},
  70         {Opt_noheap, "no_heap"},
  71         {Opt_user_xattr, "user_xattr"},
  72         {Opt_nouser_xattr, "nouser_xattr"},
  73         {Opt_acl, "acl"},
  74         {Opt_noacl, "noacl"},
  75         {Opt_active_logs, "active_logs=%u"},
  76         {Opt_disable_ext_identify, "disable_ext_identify"},
  77         {Opt_inline_xattr, "inline_xattr"},
  78         {Opt_inline_data, "inline_data"},
  79         {Opt_inline_dentry, "inline_dentry"},
  80         {Opt_flush_merge, "flush_merge"},
  81         {Opt_nobarrier, "nobarrier"},
  82         {Opt_fastboot, "fastboot"},
  83         {Opt_extent_cache, "extent_cache"},
  84         {Opt_noinline_data, "noinline_data"},
  85         {Opt_err, NULL},
  86 };
  87
  88 /* Sysfs support for f2fs */
  89 enum {
  90         GC_THREAD,      /* struct f2fs_gc_thread */
  91         SM_INFO,        /* struct f2fs_sm_info */
  92         NM_INFO,        /* struct f2fs_nm_info */
  93         F2FS_SBI,       /* struct f2fs_sb_info */
  94 };
  95
  96 struct f2fs_attr {
  97         struct attribute attr;
  98         ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
  99         ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
 100                          const char *, size_t);
 101         int struct_type;
 102         int offset;
 103 };
 104
 105 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
 106 {
 107         if (struct_type == GC_THREAD)
 108                 return (unsigned char *)sbi->gc_thread;
 109         else if (struct_type == SM_INFO)
 110                 return (unsigned char *)SM_I(sbi);
 111         else if (struct_type == NM_INFO)
 112                 return (unsigned char *)NM_I(sbi);
 113         else if (struct_type == F2FS_SBI)
 114                 return (unsigned char *)sbi;
 115         return NULL;
 116 }
 117
 118 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
 119                         struct f2fs_sb_info *sbi, char *buf)
 120 {
 121         unsigned char *ptr = NULL;
 122         unsigned int *ui;
 123
 124         ptr = __struct_ptr(sbi, a->struct_type);
 125         if (!ptr)
 126                 return -EINVAL;
 127
 128         ui = (unsigned int *)(ptr + a->offset);
 129
 130         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
 131 }
 132
 133 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
 134                         struct f2fs_sb_info *sbi,
 135                         const char *buf, size_t count)
 136 {
 137         unsigned char *ptr;
 138         unsigned long t;
 139         unsigned int *ui;
 140         ssize_t ret;
 141
 142         ptr = __struct_ptr(sbi, a->struct_type);
 143         if (!ptr)
 144                 return -EINVAL;
 145
 146         ui = (unsigned int *)(ptr + a->offset);
 147
 148         ret = kstrtoul(skip_spaces(buf), 0, &t);
 149         if (ret < 0)
 150                 return ret;
 151         *ui = t;
 152         return count;
 153 }
 154
 155 static ssize_t f2fs_attr_show(struct kobject *kobj,
 156                                 struct attribute *attr, char *buf)
 157 {
 158         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 159                                                                 s_kobj);
 160         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 161
 162         return a->show ? a->show(a, sbi, buf) : 0;
 163 }
 164
 165 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
 166                                                 const char *buf, size_t len)
 167 {
 168         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 169                                                                         s_kobj);
 170         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 171
 172         return a->store ? a->store(a, sbi, buf, len) : 0;
 173 }
 174
 175 static void f2fs_sb_release(struct kobject *kobj)
 176 {
 177         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 178                                                                 s_kobj);
 179         complete(&sbi->s_kobj_unregister);
 180 }
 181
 182 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
 183 static struct f2fs_attr f2fs_attr_##_name = {                   \
 184         .attr = {.name = __stringify(_name), .mode = _mode },   \
 185         .show   = _show,                                        \
 186         .store  = _store,                                       \
 187         .struct_type = _struct_type,                            \
 188         .offset = _offset                                       \
 189 }
 190
 191 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)    \
 192         F2FS_ATTR_OFFSET(struct_type, name, 0644,               \
 193                 f2fs_sbi_show, f2fs_sbi_store,                  \
 194                 offsetof(struct struct_name, elname))
 195
 196 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
 197 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
 198 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
 199 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
 200 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
 201 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
 202 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
 203 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
 204 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
 205 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
 206 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
 207 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
 208 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
 209
 210 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 211 static struct attribute *f2fs_attrs[] = {
 212         ATTR_LIST(gc_min_sleep_time),
 213         ATTR_LIST(gc_max_sleep_time),
 214         ATTR_LIST(gc_no_gc_sleep_time),
 215         ATTR_LIST(gc_idle),
 216         ATTR_LIST(reclaim_segments),
 217         ATTR_LIST(max_small_discards),
 218         ATTR_LIST(batched_trim_sections),
 219         ATTR_LIST(ipu_policy),
 220         ATTR_LIST(min_ipu_util),
 221         ATTR_LIST(min_fsync_blocks),
 222         ATTR_LIST(max_victim_search),
 223         ATTR_LIST(dir_level),
 224         ATTR_LIST(ram_thresh),
 225         NULL,
 226 };
 227
 228 static const struct sysfs_ops f2fs_attr_ops = {
 229         .show   = f2fs_attr_show,
 230         .store  = f2fs_attr_store,
 231 };
 232
 233 static struct kobj_type f2fs_ktype = {
 234         .default_attrs  = f2fs_attrs,
 235         .sysfs_ops      = &f2fs_attr_ops,
 236         .release        = f2fs_sb_release,
 237 };
 238
 239 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 240 {
 241         struct va_format vaf;
 242         va_list args;
 243
 244         va_start(args, fmt);
 245         vaf.fmt = fmt;
 246         vaf.va = &args;
 247         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 248         va_end(args);
 249 }
 250
 251 static void init_once(void *foo)
 252 {
 253         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 254
 255         inode_init_once(&fi->vfs_inode);
 256 }
 257
 258 static int parse_options(struct super_block *sb, char *options)
 259 {
 260         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 261         struct request_queue *q;
 262         substring_t args[MAX_OPT_ARGS];
 263         char *p, *name;
 264         int arg = 0;
 265
 266         if (!options)
 267                 return 0;
 268
 269         while ((p = strsep(&options, ",")) != NULL) {
 270                 int token;
 271                 if (!*p)
 272                         continue;
 273                 /*
 274                  * Initialize args struct so we know whether arg was
 275                  * found; some options take optional arguments.
 276                  */
 277                 args[0].to = args[0].from = NULL;
 278                 token = match_token(p, f2fs_tokens, args);
 279
 280                 switch (token) {
 281                 case Opt_gc_background:
 282                         name = match_strdup(&args[0]);
 283
 284                         if (!name)
 285                                 return -ENOMEM;
 286                         if (strlen(name) == 2 && !strncmp(name, "on", 2))
 287                                 set_opt(sbi, BG_GC);
 288                         else if (strlen(name) == 3 && !strncmp(name, "off", 3))
 289                                 clear_opt(sbi, BG_GC);
 290                         else {
 291                                 kfree(name);
 292                                 return -EINVAL;
 293                         }
 294                         kfree(name);
 295                         break;
 296                 case Opt_disable_roll_forward:
 297                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 298                         break;
 299                 case Opt_norecovery:
 300                         /* this option mounts f2fs with ro */
 301                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 302                         if (!f2fs_readonly(sb))
 303                                 return -EINVAL;
 304                         break;
 305                 case Opt_discard:
 306                         q = bdev_get_queue(sb->s_bdev);
 307                         if (blk_queue_discard(q)) {
 308                                 set_opt(sbi, DISCARD);
 309                         } else {
 310                                 f2fs_msg(sb, KERN_WARNING,
 311                                         "mounting with \"discard\" option, but "
 312                                         "the device does not support discard");
 313                         }
 314                         break;
 315                 case Opt_noheap:
 316                         set_opt(sbi, NOHEAP);
 317                         break;
 318 #ifdef CONFIG_F2FS_FS_XATTR
 319                 case Opt_user_xattr:
 320                         set_opt(sbi, XATTR_USER);
 321                         break;
 322                 case Opt_nouser_xattr:
 323                         clear_opt(sbi, XATTR_USER);
 324                         break;
 325                 case Opt_inline_xattr:
 326                         set_opt(sbi, INLINE_XATTR);
 327                         break;
 328 #else
 329                 case Opt_user_xattr:
 330                         f2fs_msg(sb, KERN_INFO,
 331                                 "user_xattr options not supported");
 332                         break;
 333                 case Opt_nouser_xattr:
 334                         f2fs_msg(sb, KERN_INFO,
 335                                 "nouser_xattr options not supported");
 336                         break;
 337                 case Opt_inline_xattr:
 338                         f2fs_msg(sb, KERN_INFO,
 339                                 "inline_xattr options not supported");
 340                         break;
 341 #endif
 342 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 343                 case Opt_acl:
 344                         set_opt(sbi, POSIX_ACL);
 345                         break;
 346                 case Opt_noacl:
 347                         clear_opt(sbi, POSIX_ACL);
 348                         break;
 349 #else
 350                 case Opt_acl:
 351                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 352                         break;
 353                 case Opt_noacl:
 354                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 355                         break;
 356 #endif
 357                 case Opt_active_logs:
 358                         if (args->from && match_int(args, &arg))
 359                                 return -EINVAL;
 360                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 361                                 return -EINVAL;
 362                         sbi->active_logs = arg;
 363                         break;
 364                 case Opt_disable_ext_identify:
 365                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 366                         break;
 367                 case Opt_inline_data:
 368                         set_opt(sbi, INLINE_DATA);
 369                         break;
 370                 case Opt_inline_dentry:
 371                         set_opt(sbi, INLINE_DENTRY);
 372                         break;
 373                 case Opt_flush_merge:
 374                         set_opt(sbi, FLUSH_MERGE);
 375                         break;
 376                 case Opt_nobarrier:
 377                         set_opt(sbi, NOBARRIER);
 378                         break;
 379                 case Opt_fastboot:
 380                         set_opt(sbi, FASTBOOT);
 381                         break;
 382                 case Opt_extent_cache:
 383                         set_opt(sbi, EXTENT_CACHE);
 384                         break;
 385                 case Opt_noinline_data:
 386                         clear_opt(sbi, INLINE_DATA);
 387                         break;
 388                 default:
 389                         f2fs_msg(sb, KERN_ERR,
 390                                 "Unrecognized mount option \"%s\" or missing value",
 391                                 p);
 392                         return -EINVAL;
 393                 }
 394         }
 395         return 0;
 396 }
 397
 398 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 399 {
 400         struct f2fs_inode_info *fi;
 401
 402         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 403         if (!fi)
 404                 return NULL;
 405
 406         init_once((void *) fi);
 407
 408         /* Initialize f2fs-specific inode info */
 409         fi->vfs_inode.i_version = 1;
 410         atomic_set(&fi->dirty_pages, 0);
 411         fi->i_current_depth = 1;
 412         fi->i_advise = 0;
 413         rwlock_init(&fi->ext_lock);
 414         init_rwsem(&fi->i_sem);
 415         INIT_RADIX_TREE(&fi->inmem_root, GFP_NOFS);
 416         INIT_LIST_HEAD(&fi->inmem_pages);
 417         mutex_init(&fi->inmem_lock);
 418
 419         set_inode_flag(fi, FI_NEW_INODE);
 420
 421         if (test_opt(F2FS_SB(sb), INLINE_XATTR))
 422                 set_inode_flag(fi, FI_INLINE_XATTR);
 423
 424         /* Will be used by directory only */
 425         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 426
 427 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 428         fi->i_crypt_info = NULL;
 429 #endif
 430         return &fi->vfs_inode;
 431 }
 432
 433 static int f2fs_drop_inode(struct inode *inode)
 434 {
 435         /*
 436          * This is to avoid a deadlock condition like below.
 437          * writeback_single_inode(inode)
 438          *  - f2fs_write_data_page
 439          *    - f2fs_gc -> iput -> evict
 440          *       - inode_wait_for_writeback(inode)
 441          */
 442         if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
 443                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 444                         spin_unlock(&inode->i_lock);
 445
 446                         /* some remained atomic pages should discarded */
 447                         if (f2fs_is_atomic_file(inode))
 448                                 commit_inmem_pages(inode, true);
 449
 450                         sb_start_intwrite(inode->i_sb);
 451                         i_size_write(inode, 0);
 452
 453                         if (F2FS_HAS_BLOCKS(inode))
 454                                 f2fs_truncate(inode);
 455
 456                         sb_end_intwrite(inode->i_sb);
 457
 458 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 459                         if (F2FS_I(inode)->i_crypt_info)
 460                                 f2fs_free_encryption_info(inode,
 461                                         F2FS_I(inode)->i_crypt_info);
 462 #endif
 463                         spin_lock(&inode->i_lock);
 464                 }
 465                 return 0;
 466         }
 467         return generic_drop_inode(inode);
 468 }
 469
 470 /*
 471  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 472  *
 473  * We should call set_dirty_inode to write the dirty inode through write_inode.
 474  */
 475 static void f2fs_dirty_inode(struct inode *inode, int flags)
 476 {
 477         set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
 478 }
 479
 480 static void f2fs_i_callback(struct rcu_head *head)
 481 {
 482         struct inode *inode = container_of(head, struct inode, i_rcu);
 483         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 484 }
 485
 486 static void f2fs_destroy_inode(struct inode *inode)
 487 {
 488         call_rcu(&inode->i_rcu, f2fs_i_callback);
 489 }
 490
 491 static void f2fs_put_super(struct super_block *sb)
 492 {
 493         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 494
 495         if (sbi->s_proc) {
 496                 remove_proc_entry("segment_info", sbi->s_proc);
 497                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 498         }
 499         kobject_del(&sbi->s_kobj);
 500
 501         f2fs_destroy_stats(sbi);
 502         stop_gc_thread(sbi);
 503
 504         /*
 505          * We don't need to do checkpoint when superblock is clean.
 506          * But, the previous checkpoint was not done by umount, it needs to do
 507          * clean checkpoint again.
 508          */
 509         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 510                         !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
 511                 struct cp_control cpc = {
 512                         .reason = CP_UMOUNT,
 513                 };
 514                 write_checkpoint(sbi, &cpc);
 515         }
 516
 517         /*
 518          * normally superblock is clean, so we need to release this.
 519          * In addition, EIO will skip do checkpoint, we need this as well.
 520          */
 521         release_dirty_inode(sbi);
 522         release_discard_addrs(sbi);
 523
 524         iput(sbi->node_inode);
 525         iput(sbi->meta_inode);
 526
 527         /* destroy f2fs internal modules */
 528         destroy_node_manager(sbi);
 529         destroy_segment_manager(sbi);
 530
 531         kfree(sbi->ckpt);
 532         kobject_put(&sbi->s_kobj);
 533         wait_for_completion(&sbi->s_kobj_unregister);
 534
 535         sb->s_fs_info = NULL;
 536         brelse(sbi->raw_super_buf);
 537         kfree(sbi);
 538 }
 539
 540 int f2fs_sync_fs(struct super_block *sb, int sync)
 541 {
 542         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 543
 544         trace_f2fs_sync_fs(sb, sync);
 545
 546         if (sync) {
 547                 struct cp_control cpc;
 548
 549                 cpc.reason = __get_cp_reason(sbi);
 550
 551                 mutex_lock(&sbi->gc_mutex);
 552                 write_checkpoint(sbi, &cpc);
 553                 mutex_unlock(&sbi->gc_mutex);
 554         } else {
 555                 f2fs_balance_fs(sbi);
 556         }
 557         f2fs_trace_ios(NULL, 1);
 558
 559         return 0;
 560 }
 561
 562 static int f2fs_freeze(struct super_block *sb)
 563 {
 564         int err;
 565
 566         if (f2fs_readonly(sb))
 567                 return 0;
 568
 569         err = f2fs_sync_fs(sb, 1);
 570         return err;
 571 }
 572
 573 static int f2fs_unfreeze(struct super_block *sb)
 574 {
 575         return 0;
 576 }
 577
 578 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 579 {
 580         struct super_block *sb = dentry->d_sb;
 581         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 582         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 583         block_t total_count, user_block_count, start_count, ovp_count;
 584
 585         total_count = le64_to_cpu(sbi->raw_super->block_count);
 586         user_block_count = sbi->user_block_count;
 587         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 588         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 589         buf->f_type = F2FS_SUPER_MAGIC;
 590         buf->f_bsize = sbi->blocksize;
 591
 592         buf->f_blocks = total_count - start_count;
 593         buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
 594         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 595
 596         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 597         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
 598
 599         buf->f_namelen = F2FS_NAME_LEN;
 600         buf->f_fsid.val[0] = (u32)id;
 601         buf->f_fsid.val[1] = (u32)(id >> 32);
 602
 603         return 0;
 604 }
 605
 606 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 607 {
 608         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 609
 610         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
 611                 seq_printf(seq, ",background_gc=%s", "on");
 612         else
 613                 seq_printf(seq, ",background_gc=%s", "off");
 614         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 615                 seq_puts(seq, ",disable_roll_forward");
 616         if (test_opt(sbi, DISCARD))
 617                 seq_puts(seq, ",discard");
 618         if (test_opt(sbi, NOHEAP))
 619                 seq_puts(seq, ",no_heap_alloc");
 620 #ifdef CONFIG_F2FS_FS_XATTR
 621         if (test_opt(sbi, XATTR_USER))
 622                 seq_puts(seq, ",user_xattr");
 623         else
 624                 seq_puts(seq, ",nouser_xattr");
 625         if (test_opt(sbi, INLINE_XATTR))
 626                 seq_puts(seq, ",inline_xattr");
 627 #endif
 628 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 629         if (test_opt(sbi, POSIX_ACL))
 630                 seq_puts(seq, ",acl");
 631         else
 632                 seq_puts(seq, ",noacl");
 633 #endif
 634         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 635                 seq_puts(seq, ",disable_ext_identify");
 636         if (test_opt(sbi, INLINE_DATA))
 637                 seq_puts(seq, ",inline_data");
 638         else
 639                 seq_puts(seq, ",noinline_data");
 640         if (test_opt(sbi, INLINE_DENTRY))
 641                 seq_puts(seq, ",inline_dentry");
 642         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 643                 seq_puts(seq, ",flush_merge");
 644         if (test_opt(sbi, NOBARRIER))
 645                 seq_puts(seq, ",nobarrier");
 646         if (test_opt(sbi, FASTBOOT))
 647                 seq_puts(seq, ",fastboot");
 648         if (test_opt(sbi, EXTENT_CACHE))
 649                 seq_puts(seq, ",extent_cache");
 650         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 651
 652         return 0;
 653 }
 654
 655 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 656 {
 657         struct super_block *sb = seq->private;
 658         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 659         unsigned int total_segs =
 660                         le32_to_cpu(sbi->raw_super->segment_count_main);
 661         int i;
 662
 663         seq_puts(seq, "format: segment_type|valid_blocks\n"
 664                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 665
 666         for (i = 0; i < total_segs; i++) {
 667                 struct seg_entry *se = get_seg_entry(sbi, i);
 668
 669                 if ((i % 10) == 0)
 670                         seq_printf(seq, "%-5d", i);
 671                 seq_printf(seq, "%d|%-3u", se->type,
 672                                         get_valid_blocks(sbi, i, 1));
 673                 if ((i % 10) == 9 || i == (total_segs - 1))
 674                         seq_putc(seq, '\n');
 675                 else
 676                         seq_putc(seq, ' ');
 677         }
 678
 679         return 0;
 680 }
 681
 682 static int segment_info_open_fs(struct inode *inode, struct file *file)
 683 {
 684         return single_open(file, segment_info_seq_show, PDE_DATA(inode));
 685 }
 686
 687 static const struct file_operations f2fs_seq_segment_info_fops = {
 688         .owner = THIS_MODULE,
 689         .open = segment_info_open_fs,
 690         .read = seq_read,
 691         .llseek = seq_lseek,
 692         .release = single_release,
 693 };
 694
 695 static void default_options(struct f2fs_sb_info *sbi)
 696 {
 697         /* init some FS parameters */
 698         sbi->active_logs = NR_CURSEG_TYPE;
 699
 700         set_opt(sbi, BG_GC);
 701         set_opt(sbi, INLINE_DATA);
 702
 703 #ifdef CONFIG_F2FS_FS_XATTR
 704         set_opt(sbi, XATTR_USER);
 705 #endif
 706 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 707         set_opt(sbi, POSIX_ACL);
 708 #endif
 709 }
 710
 711 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 712 {
 713         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 714         struct f2fs_mount_info org_mount_opt;
 715         int err, active_logs;
 716         bool need_restart_gc = false;
 717         bool need_stop_gc = false;
 718
 719         sync_filesystem(sb);
 720
 721         /*
 722          * Save the old mount options in case we
 723          * need to restore them.
 724          */
 725         org_mount_opt = sbi->mount_opt;
 726         active_logs = sbi->active_logs;
 727
 728         sbi->mount_opt.opt = 0;
 729         default_options(sbi);
 730
 731         /* parse mount options */
 732         err = parse_options(sb, data);
 733         if (err)
 734                 goto restore_opts;
 735
 736         /*
 737          * Previous and new state of filesystem is RO,
 738          * so skip checking GC and FLUSH_MERGE conditions.
 739          */
 740         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
 741                 goto skip;
 742
 743         /*
 744          * We stop the GC thread if FS is mounted as RO
 745          * or if background_gc = off is passed in mount
 746          * option. Also sync the filesystem.
 747          */
 748         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
 749                 if (sbi->gc_thread) {
 750                         stop_gc_thread(sbi);
 751                         f2fs_sync_fs(sb, 1);
 752                         need_restart_gc = true;
 753                 }
 754         } else if (!sbi->gc_thread) {
 755                 err = start_gc_thread(sbi);
 756                 if (err)
 757                         goto restore_opts;
 758                 need_stop_gc = true;
 759         }
 760
 761         /*
 762          * We stop issue flush thread if FS is mounted as RO
 763          * or if flush_merge is not passed in mount option.
 764          */
 765         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
 766                 destroy_flush_cmd_control(sbi);
 767         } else if (!SM_I(sbi)->cmd_control_info) {
 768                 err = create_flush_cmd_control(sbi);
 769                 if (err)
 770                         goto restore_gc;
 771         }
 772 skip:
 773         /* Update the POSIXACL Flag */
 774          sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 775                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 776         return 0;
 777 restore_gc:
 778         if (need_restart_gc) {
 779                 if (start_gc_thread(sbi))
 780                         f2fs_msg(sbi->sb, KERN_WARNING,
 781                                 "background gc thread has stopped");
 782         } else if (need_stop_gc) {
 783                 stop_gc_thread(sbi);
 784         }
 785 restore_opts:
 786         sbi->mount_opt = org_mount_opt;
 787         sbi->active_logs = active_logs;
 788         return err;
 789 }
 790
 791 static struct super_operations f2fs_sops = {
 792         .alloc_inode    = f2fs_alloc_inode,
 793         .drop_inode     = f2fs_drop_inode,
 794         .destroy_inode  = f2fs_destroy_inode,
 795         .write_inode    = f2fs_write_inode,
 796         .dirty_inode    = f2fs_dirty_inode,
 797         .show_options   = f2fs_show_options,
 798         .evict_inode    = f2fs_evict_inode,
 799         .put_super      = f2fs_put_super,
 800         .sync_fs        = f2fs_sync_fs,
 801         .freeze_fs      = f2fs_freeze,
 802         .unfreeze_fs    = f2fs_unfreeze,
 803         .statfs         = f2fs_statfs,
 804         .remount_fs     = f2fs_remount,
 805 };
 806
 807 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
 808                 u64 ino, u32 generation)
 809 {
 810         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 811         struct inode *inode;
 812
 813         if (check_nid_range(sbi, ino))
 814                 return ERR_PTR(-ESTALE);
 815
 816         /*
 817          * f2fs_iget isn't quite right if the inode is currently unallocated!
 818          * However f2fs_iget currently does appropriate checks to handle stale
 819          * inodes so everything is OK.
 820          */
 821         inode = f2fs_iget(sb, ino);
 822         if (IS_ERR(inode))
 823                 return ERR_CAST(inode);
 824         if (unlikely(generation && inode->i_generation != generation)) {
 825                 /* we didn't find the right inode.. */
 826                 iput(inode);
 827                 return ERR_PTR(-ESTALE);
 828         }
 829         return inode;
 830 }
 831
 832 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
 833                 int fh_len, int fh_type)
 834 {
 835         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 836                                     f2fs_nfs_get_inode);
 837 }
 838
 839 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
 840                 int fh_len, int fh_type)
 841 {
 842         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 843                                     f2fs_nfs_get_inode);
 844 }
 845
 846 static const struct export_operations f2fs_export_ops = {
 847         .fh_to_dentry = f2fs_fh_to_dentry,
 848         .fh_to_parent = f2fs_fh_to_parent,
 849         .get_parent = f2fs_get_parent,
 850 };
 851
 852 static loff_t max_file_size(unsigned bits)
 853 {
 854         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
 855         loff_t leaf_count = ADDRS_PER_BLOCK;
 856
 857         /* two direct node blocks */
 858         result += (leaf_count * 2);
 859
 860         /* two indirect node blocks */
 861         leaf_count *= NIDS_PER_BLOCK;
 862         result += (leaf_count * 2);
 863
 864         /* one double indirect node block */
 865         leaf_count *= NIDS_PER_BLOCK;
 866         result += leaf_count;
 867
 868         result <<= bits;
 869         return result;
 870 }
 871
 872 static int sanity_check_raw_super(struct super_block *sb,
 873                         struct f2fs_super_block *raw_super)
 874 {
 875         unsigned int blocksize;
 876
 877         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
 878                 f2fs_msg(sb, KERN_INFO,
 879                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
 880                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
 881                 return 1;
 882         }
 883
 884         /* Currently, support only 4KB page cache size */
 885         if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
 886                 f2fs_msg(sb, KERN_INFO,
 887                         "Invalid page_cache_size (%lu), supports only 4KB\n",
 888                         PAGE_CACHE_SIZE);
 889                 return 1;
 890         }
 891
 892         /* Currently, support only 4KB block size */
 893         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
 894         if (blocksize != F2FS_BLKSIZE) {
 895                 f2fs_msg(sb, KERN_INFO,
 896                         "Invalid blocksize (%u), supports only 4KB\n",
 897                         blocksize);
 898                 return 1;
 899         }
 900
 901         /* Currently, support 512/1024/2048/4096 bytes sector size */
 902         if (le32_to_cpu(raw_super->log_sectorsize) >
 903                                 F2FS_MAX_LOG_SECTOR_SIZE ||
 904                 le32_to_cpu(raw_super->log_sectorsize) <
 905                                 F2FS_MIN_LOG_SECTOR_SIZE) {
 906                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
 907                         le32_to_cpu(raw_super->log_sectorsize));
 908                 return 1;
 909         }
 910         if (le32_to_cpu(raw_super->log_sectors_per_block) +
 911                 le32_to_cpu(raw_super->log_sectorsize) !=
 912                         F2FS_MAX_LOG_SECTOR_SIZE) {
 913                 f2fs_msg(sb, KERN_INFO,
 914                         "Invalid log sectors per block(%u) log sectorsize(%u)",
 915                         le32_to_cpu(raw_super->log_sectors_per_block),
 916                         le32_to_cpu(raw_super->log_sectorsize));
 917                 return 1;
 918         }
 919         return 0;
 920 }
 921
 922 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
 923 {
 924         unsigned int total, fsmeta;
 925         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
 926         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 927
 928         total = le32_to_cpu(raw_super->segment_count);
 929         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
 930         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
 931         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
 932         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
 933         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
 934
 935         if (unlikely(fsmeta >= total))
 936                 return 1;
 937
 938         if (unlikely(f2fs_cp_error(sbi))) {
 939                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
 940                 return 1;
 941         }
 942         return 0;
 943 }
 944
 945 static void init_sb_info(struct f2fs_sb_info *sbi)
 946 {
 947         struct f2fs_super_block *raw_super = sbi->raw_super;
 948         int i;
 949
 950         sbi->log_sectors_per_block =
 951                 le32_to_cpu(raw_super->log_sectors_per_block);
 952         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
 953         sbi->blocksize = 1 << sbi->log_blocksize;
 954         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
 955         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
 956         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
 957         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
 958         sbi->total_sections = le32_to_cpu(raw_super->section_count);
 959         sbi->total_node_count =
 960                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
 961                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
 962         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
 963         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
 964         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
 965         sbi->cur_victim_sec = NULL_SECNO;
 966         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
 967
 968         for (i = 0; i < NR_COUNT_TYPE; i++)
 969                 atomic_set(&sbi->nr_pages[i], 0);
 970
 971         sbi->dir_level = DEF_DIR_LEVEL;
 972         clear_sbi_flag(sbi, SBI_NEED_FSCK);
 973 }
 974
 975 /*
 976  * Read f2fs raw super block.
 977  * Because we have two copies of super block, so read the first one at first,
 978  * if the first one is invalid, move to read the second one.
 979  */
 980 static int read_raw_super_block(struct super_block *sb,
 981                         struct f2fs_super_block **raw_super,
 982                         struct buffer_head **raw_super_buf,
 983                         int *recovery)
 984 {
 985         int block = 0;
 986         struct buffer_head *buffer;
 987         struct f2fs_super_block *super;
 988         int err = 0;
 989
 990 retry:
 991         buffer = sb_bread(sb, block);
 992         if (!buffer) {
 993                 *recovery = 1;
 994                 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
 995                                 block + 1);
 996                 if (block == 0) {
 997                         block++;
 998                         goto retry;
 999                 } else {
1000                         err = -EIO;
1001                         goto out;
1002                 }
1003         }
1004
1005         super = (struct f2fs_super_block *)
1006                 ((char *)(buffer)->b_data + F2FS_SUPER_OFFSET);
1007
1008         /* sanity checking of raw super */
1009         if (sanity_check_raw_super(sb, super)) {
1010                 brelse(buffer);
1011                 *recovery = 1;
1012                 f2fs_msg(sb, KERN_ERR,
1013                         "Can't find valid F2FS filesystem in %dth superblock",
1014                                                                 block + 1);
1015                 if (block == 0) {
1016                         block++;
1017                         goto retry;
1018                 } else {
1019                         err = -EINVAL;
1020                         goto out;
1021                 }
1022         }
1023
1024         if (!*raw_super) {
1025                 *raw_super_buf = buffer;
1026                 *raw_super = super;
1027         } else {
1028                 /* already have a valid superblock */
1029                 brelse(buffer);
1030         }
1031
1032         /* check the validity of the second superblock */
1033         if (block == 0) {
1034                 block++;
1035                 goto retry;
1036         }
1037
1038 out:
1039         /* No valid superblock */
1040         if (!*raw_super)
1041                 return err;
1042
1043         return 0;
1044 }
1045
1046 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1047 {
1048         struct buffer_head *sbh = sbi->raw_super_buf;
1049         sector_t block = sbh->b_blocknr;
1050         int err;
1051
1052         /* write back-up superblock first */
1053         sbh->b_blocknr = block ? 0 : 1;
1054         mark_buffer_dirty(sbh);
1055         err = sync_dirty_buffer(sbh);
1056
1057         sbh->b_blocknr = block;
1058
1059         /* if we are in recovery path, skip writing valid superblock */
1060         if (recover || err)
1061                 goto out;
1062
1063         /* write current valid superblock */
1064         mark_buffer_dirty(sbh);
1065         err = sync_dirty_buffer(sbh);
1066 out:
1067         clear_buffer_write_io_error(sbh);
1068         set_buffer_uptodate(sbh);
1069         return err;
1070 }
1071
1072 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1073 {
1074         struct f2fs_sb_info *sbi;
1075         struct f2fs_super_block *raw_super;
1076         struct buffer_head *raw_super_buf;
1077         struct inode *root;
1078         long err;
1079         bool retry = true, need_fsck = false;
1080         char *options = NULL;
1081         int recovery, i;
1082
1083 try_onemore:
1084         err = -EINVAL;
1085         raw_super = NULL;
1086         raw_super_buf = NULL;
1087         recovery = 0;
1088
1089         /* allocate memory for f2fs-specific super block info */
1090         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1091         if (!sbi)
1092                 return -ENOMEM;
1093
1094         /* set a block size */
1095         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1096                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1097                 goto free_sbi;
1098         }
1099
1100         err = read_raw_super_block(sb, &raw_super, &raw_super_buf, &recovery);
1101         if (err)
1102                 goto free_sbi;
1103
1104         sb->s_fs_info = sbi;
1105         default_options(sbi);
1106         /* parse mount options */
1107         options = kstrdup((const char *)data, GFP_KERNEL);
1108         if (data && !options) {
1109                 err = -ENOMEM;
1110                 goto free_sb_buf;
1111         }
1112
1113         err = parse_options(sb, options);
1114         if (err)
1115                 goto free_options;
1116
1117         sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
1118         sb->s_max_links = F2FS_LINK_MAX;
1119         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1120
1121         sb->s_op = &f2fs_sops;
1122         sb->s_xattr = f2fs_xattr_handlers;
1123         sb->s_export_op = &f2fs_export_ops;
1124         sb->s_magic = F2FS_SUPER_MAGIC;
1125         sb->s_time_gran = 1;
1126         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1127                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1128         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1129
1130         /* init f2fs-specific super block info */
1131         sbi->sb = sb;
1132         sbi->raw_super = raw_super;
1133         sbi->raw_super_buf = raw_super_buf;
1134         mutex_init(&sbi->gc_mutex);
1135         mutex_init(&sbi->writepages);
1136         mutex_init(&sbi->cp_mutex);
1137         init_rwsem(&sbi->node_write);
1138         clear_sbi_flag(sbi, SBI_POR_DOING);
1139         spin_lock_init(&sbi->stat_lock);
1140
1141         init_rwsem(&sbi->read_io.io_rwsem);
1142         sbi->read_io.sbi = sbi;
1143         sbi->read_io.bio = NULL;
1144         for (i = 0; i < NR_PAGE_TYPE; i++) {
1145                 init_rwsem(&sbi->write_io[i].io_rwsem);
1146                 sbi->write_io[i].sbi = sbi;
1147                 sbi->write_io[i].bio = NULL;
1148         }
1149
1150         init_rwsem(&sbi->cp_rwsem);
1151         init_waitqueue_head(&sbi->cp_wait);
1152         init_sb_info(sbi);
1153
1154         /* get an inode for meta space */
1155         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1156         if (IS_ERR(sbi->meta_inode)) {
1157                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1158                 err = PTR_ERR(sbi->meta_inode);
1159                 goto free_options;
1160         }
1161
1162         err = get_valid_checkpoint(sbi);
1163         if (err) {
1164                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1165                 goto free_meta_inode;
1166         }
1167
1168         /* sanity checking of checkpoint */
1169         err = -EINVAL;
1170         if (sanity_check_ckpt(sbi)) {
1171                 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1172                 goto free_cp;
1173         }
1174
1175         sbi->total_valid_node_count =
1176                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1177         sbi->total_valid_inode_count =
1178                                 le32_to_cpu(sbi->ckpt->valid_inode_count);
1179         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1180         sbi->total_valid_block_count =
1181                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1182         sbi->last_valid_block_count = sbi->total_valid_block_count;
1183         sbi->alloc_valid_block_count = 0;
1184         INIT_LIST_HEAD(&sbi->dir_inode_list);
1185         spin_lock_init(&sbi->dir_inode_lock);
1186
1187         init_extent_cache_info(sbi);
1188
1189         init_ino_entry_info(sbi);
1190
1191         /* setup f2fs internal modules */
1192         err = build_segment_manager(sbi);
1193         if (err) {
1194                 f2fs_msg(sb, KERN_ERR,
1195                         "Failed to initialize F2FS segment manager");
1196                 goto free_sm;
1197         }
1198         err = build_node_manager(sbi);
1199         if (err) {
1200                 f2fs_msg(sb, KERN_ERR,
1201                         "Failed to initialize F2FS node manager");
1202                 goto free_nm;
1203         }
1204
1205         build_gc_manager(sbi);
1206
1207         /* get an inode for node space */
1208         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1209         if (IS_ERR(sbi->node_inode)) {
1210                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1211                 err = PTR_ERR(sbi->node_inode);
1212                 goto free_nm;
1213         }
1214
1215         /* if there are nt orphan nodes free them */
1216         recover_orphan_inodes(sbi);
1217
1218         /* read root inode and dentry */
1219         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1220         if (IS_ERR(root)) {
1221                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1222                 err = PTR_ERR(root);
1223                 goto free_node_inode;
1224         }
1225         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1226                 iput(root);
1227                 err = -EINVAL;
1228                 goto free_node_inode;
1229         }
1230
1231         sb->s_root = d_make_root(root); /* allocate root dentry */
1232         if (!sb->s_root) {
1233                 err = -ENOMEM;
1234                 goto free_root_inode;
1235         }
1236
1237         err = f2fs_build_stats(sbi);
1238         if (err)
1239                 goto free_root_inode;
1240
1241         if (f2fs_proc_root)
1242                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1243
1244         if (sbi->s_proc)
1245                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1246                                  &f2fs_seq_segment_info_fops, sb);
1247
1248         sbi->s_kobj.kset = f2fs_kset;
1249         init_completion(&sbi->s_kobj_unregister);
1250         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1251                                                         "%s", sb->s_id);
1252         if (err)
1253                 goto free_proc;
1254
1255         /* recover fsynced data */
1256         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1257                 /*
1258                  * mount should be failed, when device has readonly mode, and
1259                  * previous checkpoint was not done by clean system shutdown.
1260                  */
1261                 if (bdev_read_only(sb->s_bdev) &&
1262                                 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1263                         err = -EROFS;
1264                         goto free_kobj;
1265                 }
1266
1267                 if (need_fsck)
1268                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1269
1270                 err = recover_fsync_data(sbi);
1271                 if (err) {
1272                         need_fsck = true;
1273                         f2fs_msg(sb, KERN_ERR,
1274                                 "Cannot recover all fsync data errno=%ld", err);
1275                         goto free_kobj;
1276                 }
1277         }
1278
1279         /*
1280          * If filesystem is not mounted as read-only then
1281          * do start the gc_thread.
1282          */
1283         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1284                 /* After POR, we can run background GC thread.*/
1285                 err = start_gc_thread(sbi);
1286                 if (err)
1287                         goto free_kobj;
1288         }
1289         kfree(options);
1290
1291         /* recover broken superblock */
1292         if (recovery && !f2fs_readonly(sb) && !bdev_read_only(sb->s_bdev)) {
1293                 f2fs_msg(sb, KERN_INFO, "Recover invalid superblock");
1294                 f2fs_commit_super(sbi, true);
1295         }
1296
1297         return 0;
1298
1299 free_kobj:
1300         kobject_del(&sbi->s_kobj);
1301 free_proc:
1302         if (sbi->s_proc) {
1303                 remove_proc_entry("segment_info", sbi->s_proc);
1304                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1305         }
1306         f2fs_destroy_stats(sbi);
1307 free_root_inode:
1308         dput(sb->s_root);
1309         sb->s_root = NULL;
1310 free_node_inode:
1311         iput(sbi->node_inode);
1312 free_nm:
1313         destroy_node_manager(sbi);
1314 free_sm:
1315         destroy_segment_manager(sbi);
1316 free_cp:
1317         kfree(sbi->ckpt);
1318 free_meta_inode:
1319         make_bad_inode(sbi->meta_inode);
1320         iput(sbi->meta_inode);
1321 free_options:
1322         kfree(options);
1323 free_sb_buf:
1324         brelse(raw_super_buf);
1325 free_sbi:
1326         kfree(sbi);
1327
1328         /* give only one another chance */
1329         if (retry) {
1330                 retry = false;
1331                 shrink_dcache_sb(sb);
1332                 goto try_onemore;
1333         }
1334         return err;
1335 }
1336
1337 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1338                         const char *dev_name, void *data)
1339 {
1340         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1341 }
1342
1343 static void kill_f2fs_super(struct super_block *sb)
1344 {
1345         if (sb->s_root)
1346                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1347         kill_block_super(sb);
1348 }
1349
1350 static struct file_system_type f2fs_fs_type = {
1351         .owner          = THIS_MODULE,
1352         .name           = "f2fs",
1353         .mount          = f2fs_mount,
1354         .kill_sb        = kill_f2fs_super,
1355         .fs_flags       = FS_REQUIRES_DEV,
1356 };
1357 MODULE_ALIAS_FS("f2fs");
1358
1359 static int __init init_inodecache(void)
1360 {
1361         f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1362                         sizeof(struct f2fs_inode_info));
1363         if (!f2fs_inode_cachep)
1364                 return -ENOMEM;
1365         return 0;
1366 }
1367
1368 static void destroy_inodecache(void)
1369 {
1370         /*
1371          * Make sure all delayed rcu free inodes are flushed before we
1372          * destroy cache.
1373          */
1374         rcu_barrier();
1375         kmem_cache_destroy(f2fs_inode_cachep);
1376 }
1377
1378 static int __init init_f2fs_fs(void)
1379 {
1380         int err;
1381
1382         f2fs_build_trace_ios();
1383
1384         err = init_inodecache();
1385         if (err)
1386                 goto fail;
1387         err = create_node_manager_caches();
1388         if (err)
1389                 goto free_inodecache;
1390         err = create_segment_manager_caches();
1391         if (err)
1392                 goto free_node_manager_caches;
1393         err = create_checkpoint_caches();
1394         if (err)
1395                 goto free_segment_manager_caches;
1396         err = create_extent_cache();
1397         if (err)
1398                 goto free_checkpoint_caches;
1399         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1400         if (!f2fs_kset) {
1401                 err = -ENOMEM;
1402                 goto free_extent_cache;
1403         }
1404         err = f2fs_init_crypto();
1405         if (err)
1406                 goto free_kset;
1407         err = register_filesystem(&f2fs_fs_type);
1408         if (err)
1409                 goto free_crypto;
1410         f2fs_create_root_stats();
1411         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1412         return 0;
1413
1414 free_crypto:
1415         f2fs_exit_crypto();
1416 free_kset:
1417         kset_unregister(f2fs_kset);
1418 free_extent_cache:
1419         destroy_extent_cache();
1420 free_checkpoint_caches:
1421         destroy_checkpoint_caches();
1422 free_segment_manager_caches:
1423         destroy_segment_manager_caches();
1424 free_node_manager_caches:
1425         destroy_node_manager_caches();
1426 free_inodecache:
1427         destroy_inodecache();
1428 fail:
1429         return err;
1430 }
1431
1432 static void __exit exit_f2fs_fs(void)
1433 {
1434         remove_proc_entry("fs/f2fs", NULL);
1435         f2fs_destroy_root_stats();
1436         unregister_filesystem(&f2fs_fs_type);
1437         f2fs_exit_crypto();
1438         destroy_extent_cache();
1439         destroy_checkpoint_caches();
1440         destroy_segment_manager_caches();
1441         destroy_node_manager_caches();
1442         destroy_inodecache();
1443         kset_unregister(f2fs_kset);
1444         f2fs_destroy_trace_ios();
1445 }
1446
1447 module_init(init_f2fs_fs)
1448 module_exit(exit_f2fs_fs)
1449
1450 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1451 MODULE_DESCRIPTION("Flash Friendly File System");
1452 MODULE_LICENSE("GPL");