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 /* f2fs-wide shrinker description */
  43 static struct shrinker f2fs_shrinker_info = {
  44         .scan_objects = f2fs_shrink_scan,
  45         .count_objects = f2fs_shrink_count,
  46         .seeks = DEFAULT_SEEKS,
  47 };
  48
  49 enum {
  50         Opt_gc_background,
  51         Opt_disable_roll_forward,
  52         Opt_norecovery,
  53         Opt_discard,
  54         Opt_noheap,
  55         Opt_user_xattr,
  56         Opt_nouser_xattr,
  57         Opt_acl,
  58         Opt_noacl,
  59         Opt_active_logs,
  60         Opt_disable_ext_identify,
  61         Opt_inline_xattr,
  62         Opt_inline_data,
  63         Opt_inline_dentry,
  64         Opt_flush_merge,
  65         Opt_nobarrier,
  66         Opt_fastboot,
  67         Opt_extent_cache,
  68         Opt_noextent_cache,
  69         Opt_noinline_data,
  70         Opt_data_flush,
  71         Opt_err,
  72 };
  73
  74 static match_table_t f2fs_tokens = {
  75         {Opt_gc_background, "background_gc=%s"},
  76         {Opt_disable_roll_forward, "disable_roll_forward"},
  77         {Opt_norecovery, "norecovery"},
  78         {Opt_discard, "discard"},
  79         {Opt_noheap, "no_heap"},
  80         {Opt_user_xattr, "user_xattr"},
  81         {Opt_nouser_xattr, "nouser_xattr"},
  82         {Opt_acl, "acl"},
  83         {Opt_noacl, "noacl"},
  84         {Opt_active_logs, "active_logs=%u"},
  85         {Opt_disable_ext_identify, "disable_ext_identify"},
  86         {Opt_inline_xattr, "inline_xattr"},
  87         {Opt_inline_data, "inline_data"},
  88         {Opt_inline_dentry, "inline_dentry"},
  89         {Opt_flush_merge, "flush_merge"},
  90         {Opt_nobarrier, "nobarrier"},
  91         {Opt_fastboot, "fastboot"},
  92         {Opt_extent_cache, "extent_cache"},
  93         {Opt_noextent_cache, "noextent_cache"},
  94         {Opt_noinline_data, "noinline_data"},
  95         {Opt_data_flush, "data_flush"},
  96         {Opt_err, NULL},
  97 };
  98
  99 /* Sysfs support for f2fs */
 100 enum {
 101         GC_THREAD,      /* struct f2fs_gc_thread */
 102         SM_INFO,        /* struct f2fs_sm_info */
 103         NM_INFO,        /* struct f2fs_nm_info */
 104         F2FS_SBI,       /* struct f2fs_sb_info */
 105 };
 106
 107 struct f2fs_attr {
 108         struct attribute attr;
 109         ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
 110         ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
 111                          const char *, size_t);
 112         int struct_type;
 113         int offset;
 114 };
 115
 116 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
 117 {
 118         if (struct_type == GC_THREAD)
 119                 return (unsigned char *)sbi->gc_thread;
 120         else if (struct_type == SM_INFO)
 121                 return (unsigned char *)SM_I(sbi);
 122         else if (struct_type == NM_INFO)
 123                 return (unsigned char *)NM_I(sbi);
 124         else if (struct_type == F2FS_SBI)
 125                 return (unsigned char *)sbi;
 126         return NULL;
 127 }
 128
 129 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
 130                 struct f2fs_sb_info *sbi, char *buf)
 131 {
 132         struct super_block *sb = sbi->sb;
 133
 134         if (!sb->s_bdev->bd_part)
 135                 return snprintf(buf, PAGE_SIZE, "0\n");
 136
 137         return snprintf(buf, PAGE_SIZE, "%llu\n",
 138                 (unsigned long long)(sbi->kbytes_written +
 139                         BD_PART_WRITTEN(sbi)));
 140 }
 141
 142 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
 143                         struct f2fs_sb_info *sbi, char *buf)
 144 {
 145         unsigned char *ptr = NULL;
 146         unsigned int *ui;
 147
 148         ptr = __struct_ptr(sbi, a->struct_type);
 149         if (!ptr)
 150                 return -EINVAL;
 151
 152         ui = (unsigned int *)(ptr + a->offset);
 153
 154         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
 155 }
 156
 157 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
 158                         struct f2fs_sb_info *sbi,
 159                         const char *buf, size_t count)
 160 {
 161         unsigned char *ptr;
 162         unsigned long t;
 163         unsigned int *ui;
 164         ssize_t ret;
 165
 166         ptr = __struct_ptr(sbi, a->struct_type);
 167         if (!ptr)
 168                 return -EINVAL;
 169
 170         ui = (unsigned int *)(ptr + a->offset);
 171
 172         ret = kstrtoul(skip_spaces(buf), 0, &t);
 173         if (ret < 0)
 174                 return ret;
 175         *ui = t;
 176         return count;
 177 }
 178
 179 static ssize_t f2fs_attr_show(struct kobject *kobj,
 180                                 struct attribute *attr, char *buf)
 181 {
 182         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 183                                                                 s_kobj);
 184         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 185
 186         return a->show ? a->show(a, sbi, buf) : 0;
 187 }
 188
 189 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
 190                                                 const char *buf, size_t len)
 191 {
 192         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 193                                                                         s_kobj);
 194         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 195
 196         return a->store ? a->store(a, sbi, buf, len) : 0;
 197 }
 198
 199 static void f2fs_sb_release(struct kobject *kobj)
 200 {
 201         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 202                                                                 s_kobj);
 203         complete(&sbi->s_kobj_unregister);
 204 }
 205
 206 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
 207 static struct f2fs_attr f2fs_attr_##_name = {                   \
 208         .attr = {.name = __stringify(_name), .mode = _mode },   \
 209         .show   = _show,                                        \
 210         .store  = _store,                                       \
 211         .struct_type = _struct_type,                            \
 212         .offset = _offset                                       \
 213 }
 214
 215 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)    \
 216         F2FS_ATTR_OFFSET(struct_type, name, 0644,               \
 217                 f2fs_sbi_show, f2fs_sbi_store,                  \
 218                 offsetof(struct struct_name, elname))
 219
 220 #define F2FS_GENERAL_RO_ATTR(name) \
 221 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
 222
 223 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
 224 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
 225 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
 226 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
 227 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
 228 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
 229 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
 230 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
 231 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
 232 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
 233 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
 234 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
 235 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
 236 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
 237 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
 238 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
 239 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
 240 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
 241
 242 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 243 static struct attribute *f2fs_attrs[] = {
 244         ATTR_LIST(gc_min_sleep_time),
 245         ATTR_LIST(gc_max_sleep_time),
 246         ATTR_LIST(gc_no_gc_sleep_time),
 247         ATTR_LIST(gc_idle),
 248         ATTR_LIST(reclaim_segments),
 249         ATTR_LIST(max_small_discards),
 250         ATTR_LIST(batched_trim_sections),
 251         ATTR_LIST(ipu_policy),
 252         ATTR_LIST(min_ipu_util),
 253         ATTR_LIST(min_fsync_blocks),
 254         ATTR_LIST(max_victim_search),
 255         ATTR_LIST(dir_level),
 256         ATTR_LIST(ram_thresh),
 257         ATTR_LIST(ra_nid_pages),
 258         ATTR_LIST(dirty_nats_ratio),
 259         ATTR_LIST(cp_interval),
 260         ATTR_LIST(idle_interval),
 261         ATTR_LIST(lifetime_write_kbytes),
 262         NULL,
 263 };
 264
 265 static const struct sysfs_ops f2fs_attr_ops = {
 266         .show   = f2fs_attr_show,
 267         .store  = f2fs_attr_store,
 268 };
 269
 270 static struct kobj_type f2fs_ktype = {
 271         .default_attrs  = f2fs_attrs,
 272         .sysfs_ops      = &f2fs_attr_ops,
 273         .release        = f2fs_sb_release,
 274 };
 275
 276 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 277 {
 278         struct va_format vaf;
 279         va_list args;
 280
 281         va_start(args, fmt);
 282         vaf.fmt = fmt;
 283         vaf.va = &args;
 284         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 285         va_end(args);
 286 }
 287
 288 static void init_once(void *foo)
 289 {
 290         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 291
 292         inode_init_once(&fi->vfs_inode);
 293 }
 294
 295 static int parse_options(struct super_block *sb, char *options)
 296 {
 297         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 298         struct request_queue *q;
 299         substring_t args[MAX_OPT_ARGS];
 300         char *p, *name;
 301         int arg = 0;
 302
 303         if (!options)
 304                 return 0;
 305
 306         while ((p = strsep(&options, ",")) != NULL) {
 307                 int token;
 308                 if (!*p)
 309                         continue;
 310                 /*
 311                  * Initialize args struct so we know whether arg was
 312                  * found; some options take optional arguments.
 313                  */
 314                 args[0].to = args[0].from = NULL;
 315                 token = match_token(p, f2fs_tokens, args);
 316
 317                 switch (token) {
 318                 case Opt_gc_background:
 319                         name = match_strdup(&args[0]);
 320
 321                         if (!name)
 322                                 return -ENOMEM;
 323                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 324                                 set_opt(sbi, BG_GC);
 325                                 clear_opt(sbi, FORCE_FG_GC);
 326                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 327                                 clear_opt(sbi, BG_GC);
 328                                 clear_opt(sbi, FORCE_FG_GC);
 329                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 330                                 set_opt(sbi, BG_GC);
 331                                 set_opt(sbi, FORCE_FG_GC);
 332                         } else {
 333                                 kfree(name);
 334                                 return -EINVAL;
 335                         }
 336                         kfree(name);
 337                         break;
 338                 case Opt_disable_roll_forward:
 339                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 340                         break;
 341                 case Opt_norecovery:
 342                         /* this option mounts f2fs with ro */
 343                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 344                         if (!f2fs_readonly(sb))
 345                                 return -EINVAL;
 346                         break;
 347                 case Opt_discard:
 348                         q = bdev_get_queue(sb->s_bdev);
 349                         if (blk_queue_discard(q)) {
 350                                 set_opt(sbi, DISCARD);
 351                         } else {
 352                                 f2fs_msg(sb, KERN_WARNING,
 353                                         "mounting with \"discard\" option, but "
 354                                         "the device does not support discard");
 355                         }
 356                         break;
 357                 case Opt_noheap:
 358                         set_opt(sbi, NOHEAP);
 359                         break;
 360 #ifdef CONFIG_F2FS_FS_XATTR
 361                 case Opt_user_xattr:
 362                         set_opt(sbi, XATTR_USER);
 363                         break;
 364                 case Opt_nouser_xattr:
 365                         clear_opt(sbi, XATTR_USER);
 366                         break;
 367                 case Opt_inline_xattr:
 368                         set_opt(sbi, INLINE_XATTR);
 369                         break;
 370 #else
 371                 case Opt_user_xattr:
 372                         f2fs_msg(sb, KERN_INFO,
 373                                 "user_xattr options not supported");
 374                         break;
 375                 case Opt_nouser_xattr:
 376                         f2fs_msg(sb, KERN_INFO,
 377                                 "nouser_xattr options not supported");
 378                         break;
 379                 case Opt_inline_xattr:
 380                         f2fs_msg(sb, KERN_INFO,
 381                                 "inline_xattr options not supported");
 382                         break;
 383 #endif
 384 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 385                 case Opt_acl:
 386                         set_opt(sbi, POSIX_ACL);
 387                         break;
 388                 case Opt_noacl:
 389                         clear_opt(sbi, POSIX_ACL);
 390                         break;
 391 #else
 392                 case Opt_acl:
 393                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 394                         break;
 395                 case Opt_noacl:
 396                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 397                         break;
 398 #endif
 399                 case Opt_active_logs:
 400                         if (args->from && match_int(args, &arg))
 401                                 return -EINVAL;
 402                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 403                                 return -EINVAL;
 404                         sbi->active_logs = arg;
 405                         break;
 406                 case Opt_disable_ext_identify:
 407                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 408                         break;
 409                 case Opt_inline_data:
 410                         set_opt(sbi, INLINE_DATA);
 411                         break;
 412                 case Opt_inline_dentry:
 413                         set_opt(sbi, INLINE_DENTRY);
 414                         break;
 415                 case Opt_flush_merge:
 416                         set_opt(sbi, FLUSH_MERGE);
 417                         break;
 418                 case Opt_nobarrier:
 419                         set_opt(sbi, NOBARRIER);
 420                         break;
 421                 case Opt_fastboot:
 422                         set_opt(sbi, FASTBOOT);
 423                         break;
 424                 case Opt_extent_cache:
 425                         set_opt(sbi, EXTENT_CACHE);
 426                         break;
 427                 case Opt_noextent_cache:
 428                         clear_opt(sbi, EXTENT_CACHE);
 429                         break;
 430                 case Opt_noinline_data:
 431                         clear_opt(sbi, INLINE_DATA);
 432                         break;
 433                 case Opt_data_flush:
 434                         set_opt(sbi, DATA_FLUSH);
 435                         break;
 436                 default:
 437                         f2fs_msg(sb, KERN_ERR,
 438                                 "Unrecognized mount option \"%s\" or missing value",
 439                                 p);
 440                         return -EINVAL;
 441                 }
 442         }
 443         return 0;
 444 }
 445
 446 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 447 {
 448         struct f2fs_inode_info *fi;
 449
 450         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 451         if (!fi)
 452                 return NULL;
 453
 454         init_once((void *) fi);
 455
 456         /* Initialize f2fs-specific inode info */
 457         fi->vfs_inode.i_version = 1;
 458         atomic_set(&fi->dirty_pages, 0);
 459         fi->i_current_depth = 1;
 460         fi->i_advise = 0;
 461         init_rwsem(&fi->i_sem);
 462         INIT_LIST_HEAD(&fi->dirty_list);
 463         INIT_LIST_HEAD(&fi->inmem_pages);
 464         mutex_init(&fi->inmem_lock);
 465
 466         set_inode_flag(fi, FI_NEW_INODE);
 467
 468         if (test_opt(F2FS_SB(sb), INLINE_XATTR))
 469                 set_inode_flag(fi, FI_INLINE_XATTR);
 470
 471         /* Will be used by directory only */
 472         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 473
 474 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 475         fi->i_crypt_info = NULL;
 476 #endif
 477         return &fi->vfs_inode;
 478 }
 479
 480 static int f2fs_drop_inode(struct inode *inode)
 481 {
 482         /*
 483          * This is to avoid a deadlock condition like below.
 484          * writeback_single_inode(inode)
 485          *  - f2fs_write_data_page
 486          *    - f2fs_gc -> iput -> evict
 487          *       - inode_wait_for_writeback(inode)
 488          */
 489         if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
 490                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 491                         /* to avoid evict_inode call simultaneously */
 492                         atomic_inc(&inode->i_count);
 493                         spin_unlock(&inode->i_lock);
 494
 495                         /* some remained atomic pages should discarded */
 496                         if (f2fs_is_atomic_file(inode))
 497                                 commit_inmem_pages(inode, true);
 498
 499                         /* should remain fi->extent_tree for writepage */
 500                         f2fs_destroy_extent_node(inode);
 501
 502                         sb_start_intwrite(inode->i_sb);
 503                         i_size_write(inode, 0);
 504
 505                         if (F2FS_HAS_BLOCKS(inode))
 506                                 f2fs_truncate(inode, true);
 507
 508                         sb_end_intwrite(inode->i_sb);
 509
 510 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 511                         if (F2FS_I(inode)->i_crypt_info)
 512                                 f2fs_free_encryption_info(inode,
 513                                         F2FS_I(inode)->i_crypt_info);
 514 #endif
 515                         spin_lock(&inode->i_lock);
 516                         atomic_dec(&inode->i_count);
 517                 }
 518                 return 0;
 519         }
 520         return generic_drop_inode(inode);
 521 }
 522
 523 /*
 524  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 525  *
 526  * We should call set_dirty_inode to write the dirty inode through write_inode.
 527  */
 528 static void f2fs_dirty_inode(struct inode *inode, int flags)
 529 {
 530         set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
 531 }
 532
 533 static void f2fs_i_callback(struct rcu_head *head)
 534 {
 535         struct inode *inode = container_of(head, struct inode, i_rcu);
 536         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 537 }
 538
 539 static void f2fs_destroy_inode(struct inode *inode)
 540 {
 541         call_rcu(&inode->i_rcu, f2fs_i_callback);
 542 }
 543
 544 static void f2fs_put_super(struct super_block *sb)
 545 {
 546         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 547
 548         if (sbi->s_proc) {
 549                 remove_proc_entry("segment_info", sbi->s_proc);
 550                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 551         }
 552         kobject_del(&sbi->s_kobj);
 553
 554         stop_gc_thread(sbi);
 555
 556         /* prevent remaining shrinker jobs */
 557         mutex_lock(&sbi->umount_mutex);
 558
 559         /*
 560          * We don't need to do checkpoint when superblock is clean.
 561          * But, the previous checkpoint was not done by umount, it needs to do
 562          * clean checkpoint again.
 563          */
 564         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 565                         !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
 566                 struct cp_control cpc = {
 567                         .reason = CP_UMOUNT,
 568                 };
 569                 write_checkpoint(sbi, &cpc);
 570         }
 571
 572         /* write_checkpoint can update stat informaion */
 573         f2fs_destroy_stats(sbi);
 574
 575         /*
 576          * normally superblock is clean, so we need to release this.
 577          * In addition, EIO will skip do checkpoint, we need this as well.
 578          */
 579         release_ino_entry(sbi);
 580         release_discard_addrs(sbi);
 581
 582         f2fs_leave_shrinker(sbi);
 583         mutex_unlock(&sbi->umount_mutex);
 584
 585         /* our cp_error case, we can wait for any writeback page */
 586         if (get_pages(sbi, F2FS_WRITEBACK)) {
 587                 f2fs_submit_merged_bio(sbi, DATA, WRITE);
 588                 f2fs_submit_merged_bio(sbi, NODE, WRITE);
 589                 f2fs_submit_merged_bio(sbi, META, WRITE);
 590         }
 591
 592         iput(sbi->node_inode);
 593         iput(sbi->meta_inode);
 594
 595         /* destroy f2fs internal modules */
 596         destroy_node_manager(sbi);
 597         destroy_segment_manager(sbi);
 598
 599         kfree(sbi->ckpt);
 600         kobject_put(&sbi->s_kobj);
 601         wait_for_completion(&sbi->s_kobj_unregister);
 602
 603         sb->s_fs_info = NULL;
 604         kfree(sbi->raw_super);
 605         kfree(sbi);
 606 }
 607
 608 int f2fs_sync_fs(struct super_block *sb, int sync)
 609 {
 610         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 611         int err = 0;
 612
 613         trace_f2fs_sync_fs(sb, sync);
 614
 615         if (sync) {
 616                 struct cp_control cpc;
 617
 618                 cpc.reason = __get_cp_reason(sbi);
 619
 620                 mutex_lock(&sbi->gc_mutex);
 621                 err = write_checkpoint(sbi, &cpc);
 622                 mutex_unlock(&sbi->gc_mutex);
 623         }
 624         f2fs_trace_ios(NULL, 1);
 625
 626         return err;
 627 }
 628
 629 static int f2fs_freeze(struct super_block *sb)
 630 {
 631         int err;
 632
 633         if (f2fs_readonly(sb))
 634                 return 0;
 635
 636         err = f2fs_sync_fs(sb, 1);
 637         return err;
 638 }
 639
 640 static int f2fs_unfreeze(struct super_block *sb)
 641 {
 642         return 0;
 643 }
 644
 645 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 646 {
 647         struct super_block *sb = dentry->d_sb;
 648         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 649         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 650         block_t total_count, user_block_count, start_count, ovp_count;
 651
 652         total_count = le64_to_cpu(sbi->raw_super->block_count);
 653         user_block_count = sbi->user_block_count;
 654         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 655         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 656         buf->f_type = F2FS_SUPER_MAGIC;
 657         buf->f_bsize = sbi->blocksize;
 658
 659         buf->f_blocks = total_count - start_count;
 660         buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
 661         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 662
 663         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 664         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
 665
 666         buf->f_namelen = F2FS_NAME_LEN;
 667         buf->f_fsid.val[0] = (u32)id;
 668         buf->f_fsid.val[1] = (u32)(id >> 32);
 669
 670         return 0;
 671 }
 672
 673 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 674 {
 675         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 676
 677         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
 678                 if (test_opt(sbi, FORCE_FG_GC))
 679                         seq_printf(seq, ",background_gc=%s", "sync");
 680                 else
 681                         seq_printf(seq, ",background_gc=%s", "on");
 682         } else {
 683                 seq_printf(seq, ",background_gc=%s", "off");
 684         }
 685         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 686                 seq_puts(seq, ",disable_roll_forward");
 687         if (test_opt(sbi, DISCARD))
 688                 seq_puts(seq, ",discard");
 689         if (test_opt(sbi, NOHEAP))
 690                 seq_puts(seq, ",no_heap_alloc");
 691 #ifdef CONFIG_F2FS_FS_XATTR
 692         if (test_opt(sbi, XATTR_USER))
 693                 seq_puts(seq, ",user_xattr");
 694         else
 695                 seq_puts(seq, ",nouser_xattr");
 696         if (test_opt(sbi, INLINE_XATTR))
 697                 seq_puts(seq, ",inline_xattr");
 698 #endif
 699 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 700         if (test_opt(sbi, POSIX_ACL))
 701                 seq_puts(seq, ",acl");
 702         else
 703                 seq_puts(seq, ",noacl");
 704 #endif
 705         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 706                 seq_puts(seq, ",disable_ext_identify");
 707         if (test_opt(sbi, INLINE_DATA))
 708                 seq_puts(seq, ",inline_data");
 709         else
 710                 seq_puts(seq, ",noinline_data");
 711         if (test_opt(sbi, INLINE_DENTRY))
 712                 seq_puts(seq, ",inline_dentry");
 713         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 714                 seq_puts(seq, ",flush_merge");
 715         if (test_opt(sbi, NOBARRIER))
 716                 seq_puts(seq, ",nobarrier");
 717         if (test_opt(sbi, FASTBOOT))
 718                 seq_puts(seq, ",fastboot");
 719         if (test_opt(sbi, EXTENT_CACHE))
 720                 seq_puts(seq, ",extent_cache");
 721         else
 722                 seq_puts(seq, ",noextent_cache");
 723         if (test_opt(sbi, DATA_FLUSH))
 724                 seq_puts(seq, ",data_flush");
 725         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 726
 727         return 0;
 728 }
 729
 730 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 731 {
 732         struct super_block *sb = seq->private;
 733         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 734         unsigned int total_segs =
 735                         le32_to_cpu(sbi->raw_super->segment_count_main);
 736         int i;
 737
 738         seq_puts(seq, "format: segment_type|valid_blocks\n"
 739                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 740
 741         for (i = 0; i < total_segs; i++) {
 742                 struct seg_entry *se = get_seg_entry(sbi, i);
 743
 744                 if ((i % 10) == 0)
 745                         seq_printf(seq, "%-10d", i);
 746                 seq_printf(seq, "%d|%-3u", se->type,
 747                                         get_valid_blocks(sbi, i, 1));
 748                 if ((i % 10) == 9 || i == (total_segs - 1))
 749                         seq_putc(seq, '\n');
 750                 else
 751                         seq_putc(seq, ' ');
 752         }
 753
 754         return 0;
 755 }
 756
 757 static int segment_info_open_fs(struct inode *inode, struct file *file)
 758 {
 759         return single_open(file, segment_info_seq_show, PDE_DATA(inode));
 760 }
 761
 762 static const struct file_operations f2fs_seq_segment_info_fops = {
 763         .owner = THIS_MODULE,
 764         .open = segment_info_open_fs,
 765         .read = seq_read,
 766         .llseek = seq_lseek,
 767         .release = single_release,
 768 };
 769
 770 static void default_options(struct f2fs_sb_info *sbi)
 771 {
 772         /* init some FS parameters */
 773         sbi->active_logs = NR_CURSEG_TYPE;
 774
 775         set_opt(sbi, BG_GC);
 776         set_opt(sbi, INLINE_DATA);
 777         set_opt(sbi, EXTENT_CACHE);
 778
 779 #ifdef CONFIG_F2FS_FS_XATTR
 780         set_opt(sbi, XATTR_USER);
 781 #endif
 782 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 783         set_opt(sbi, POSIX_ACL);
 784 #endif
 785 }
 786
 787 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 788 {
 789         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 790         struct f2fs_mount_info org_mount_opt;
 791         int err, active_logs;
 792         bool need_restart_gc = false;
 793         bool need_stop_gc = false;
 794         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
 795
 796         /*
 797          * Save the old mount options in case we
 798          * need to restore them.
 799          */
 800         org_mount_opt = sbi->mount_opt;
 801         active_logs = sbi->active_logs;
 802
 803         if (*flags & MS_RDONLY) {
 804                 set_opt(sbi, FASTBOOT);
 805                 set_sbi_flag(sbi, SBI_IS_DIRTY);
 806         }
 807
 808         sync_filesystem(sb);
 809
 810         sbi->mount_opt.opt = 0;
 811         default_options(sbi);
 812
 813         /* parse mount options */
 814         err = parse_options(sb, data);
 815         if (err)
 816                 goto restore_opts;
 817
 818         /*
 819          * Previous and new state of filesystem is RO,
 820          * so skip checking GC and FLUSH_MERGE conditions.
 821          */
 822         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
 823                 goto skip;
 824
 825         /* disallow enable/disable extent_cache dynamically */
 826         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
 827                 err = -EINVAL;
 828                 f2fs_msg(sbi->sb, KERN_WARNING,
 829                                 "switch extent_cache option is not allowed");
 830                 goto restore_opts;
 831         }
 832
 833         /*
 834          * We stop the GC thread if FS is mounted as RO
 835          * or if background_gc = off is passed in mount
 836          * option. Also sync the filesystem.
 837          */
 838         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
 839                 if (sbi->gc_thread) {
 840                         stop_gc_thread(sbi);
 841                         f2fs_sync_fs(sb, 1);
 842                         need_restart_gc = true;
 843                 }
 844         } else if (!sbi->gc_thread) {
 845                 err = start_gc_thread(sbi);
 846                 if (err)
 847                         goto restore_opts;
 848                 need_stop_gc = true;
 849         }
 850
 851         /*
 852          * We stop issue flush thread if FS is mounted as RO
 853          * or if flush_merge is not passed in mount option.
 854          */
 855         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
 856                 destroy_flush_cmd_control(sbi);
 857         } else if (!SM_I(sbi)->cmd_control_info) {
 858                 err = create_flush_cmd_control(sbi);
 859                 if (err)
 860                         goto restore_gc;
 861         }
 862 skip:
 863         /* Update the POSIXACL Flag */
 864          sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 865                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 866         return 0;
 867 restore_gc:
 868         if (need_restart_gc) {
 869                 if (start_gc_thread(sbi))
 870                         f2fs_msg(sbi->sb, KERN_WARNING,
 871                                 "background gc thread has stopped");
 872         } else if (need_stop_gc) {
 873                 stop_gc_thread(sbi);
 874         }
 875 restore_opts:
 876         sbi->mount_opt = org_mount_opt;
 877         sbi->active_logs = active_logs;
 878         return err;
 879 }
 880
 881 static struct super_operations f2fs_sops = {
 882         .alloc_inode    = f2fs_alloc_inode,
 883         .drop_inode     = f2fs_drop_inode,
 884         .destroy_inode  = f2fs_destroy_inode,
 885         .write_inode    = f2fs_write_inode,
 886         .dirty_inode    = f2fs_dirty_inode,
 887         .show_options   = f2fs_show_options,
 888         .evict_inode    = f2fs_evict_inode,
 889         .put_super      = f2fs_put_super,
 890         .sync_fs        = f2fs_sync_fs,
 891         .freeze_fs      = f2fs_freeze,
 892         .unfreeze_fs    = f2fs_unfreeze,
 893         .statfs         = f2fs_statfs,
 894         .remount_fs     = f2fs_remount,
 895 };
 896
 897 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
 898                 u64 ino, u32 generation)
 899 {
 900         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 901         struct inode *inode;
 902
 903         if (check_nid_range(sbi, ino))
 904                 return ERR_PTR(-ESTALE);
 905
 906         /*
 907          * f2fs_iget isn't quite right if the inode is currently unallocated!
 908          * However f2fs_iget currently does appropriate checks to handle stale
 909          * inodes so everything is OK.
 910          */
 911         inode = f2fs_iget(sb, ino);
 912         if (IS_ERR(inode))
 913                 return ERR_CAST(inode);
 914         if (unlikely(generation && inode->i_generation != generation)) {
 915                 /* we didn't find the right inode.. */
 916                 iput(inode);
 917                 return ERR_PTR(-ESTALE);
 918         }
 919         return inode;
 920 }
 921
 922 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
 923                 int fh_len, int fh_type)
 924 {
 925         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 926                                     f2fs_nfs_get_inode);
 927 }
 928
 929 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
 930                 int fh_len, int fh_type)
 931 {
 932         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 933                                     f2fs_nfs_get_inode);
 934 }
 935
 936 static const struct export_operations f2fs_export_ops = {
 937         .fh_to_dentry = f2fs_fh_to_dentry,
 938         .fh_to_parent = f2fs_fh_to_parent,
 939         .get_parent = f2fs_get_parent,
 940 };
 941
 942 static loff_t max_file_blocks(void)
 943 {
 944         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
 945         loff_t leaf_count = ADDRS_PER_BLOCK;
 946
 947         /* two direct node blocks */
 948         result += (leaf_count * 2);
 949
 950         /* two indirect node blocks */
 951         leaf_count *= NIDS_PER_BLOCK;
 952         result += (leaf_count * 2);
 953
 954         /* one double indirect node block */
 955         leaf_count *= NIDS_PER_BLOCK;
 956         result += leaf_count;
 957
 958         return result;
 959 }
 960
 961 static inline bool sanity_check_area_boundary(struct super_block *sb,
 962                                         struct f2fs_super_block *raw_super)
 963 {
 964         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
 965         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
 966         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
 967         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
 968         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
 969         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
 970         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
 971         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
 972         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
 973         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
 974         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
 975         u32 segment_count = le32_to_cpu(raw_super->segment_count);
 976         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
 977
 978         if (segment0_blkaddr != cp_blkaddr) {
 979                 f2fs_msg(sb, KERN_INFO,
 980                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
 981                         segment0_blkaddr, cp_blkaddr);
 982                 return true;
 983         }
 984
 985         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
 986                                                         sit_blkaddr) {
 987                 f2fs_msg(sb, KERN_INFO,
 988                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
 989                         cp_blkaddr, sit_blkaddr,
 990                         segment_count_ckpt << log_blocks_per_seg);
 991                 return true;
 992         }
 993
 994         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
 995                                                         nat_blkaddr) {
 996                 f2fs_msg(sb, KERN_INFO,
 997                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
 998                         sit_blkaddr, nat_blkaddr,
 999                         segment_count_sit << log_blocks_per_seg);
1000                 return true;
1001         }
1002
1003         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1004                                                         ssa_blkaddr) {
1005                 f2fs_msg(sb, KERN_INFO,
1006                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1007                         nat_blkaddr, ssa_blkaddr,
1008                         segment_count_nat << log_blocks_per_seg);
1009                 return true;
1010         }
1011
1012         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1013                                                         main_blkaddr) {
1014                 f2fs_msg(sb, KERN_INFO,
1015                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1016                         ssa_blkaddr, main_blkaddr,
1017                         segment_count_ssa << log_blocks_per_seg);
1018                 return true;
1019         }
1020
1021         if (main_blkaddr + (segment_count_main << log_blocks_per_seg) !=
1022                 segment0_blkaddr + (segment_count << log_blocks_per_seg)) {
1023                 f2fs_msg(sb, KERN_INFO,
1024                         "Wrong MAIN_AREA boundary, start(%u) end(%u) blocks(%u)",
1025                         main_blkaddr,
1026                         segment0_blkaddr + (segment_count << log_blocks_per_seg),
1027                         segment_count_main << log_blocks_per_seg);
1028                 return true;
1029         }
1030
1031         return false;
1032 }
1033
1034 static int sanity_check_raw_super(struct super_block *sb,
1035                         struct f2fs_super_block *raw_super)
1036 {
1037         unsigned int blocksize;
1038
1039         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1040                 f2fs_msg(sb, KERN_INFO,
1041                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1042                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1043                 return 1;
1044         }
1045
1046         /* Currently, support only 4KB page cache size */
1047         if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
1048                 f2fs_msg(sb, KERN_INFO,
1049                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1050                         PAGE_CACHE_SIZE);
1051                 return 1;
1052         }
1053
1054         /* Currently, support only 4KB block size */
1055         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1056         if (blocksize != F2FS_BLKSIZE) {
1057                 f2fs_msg(sb, KERN_INFO,
1058                         "Invalid blocksize (%u), supports only 4KB\n",
1059                         blocksize);
1060                 return 1;
1061         }
1062
1063         /* check log blocks per segment */
1064         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1065                 f2fs_msg(sb, KERN_INFO,
1066                         "Invalid log blocks per segment (%u)\n",
1067                         le32_to_cpu(raw_super->log_blocks_per_seg));
1068                 return 1;
1069         }
1070
1071         /* Currently, support 512/1024/2048/4096 bytes sector size */
1072         if (le32_to_cpu(raw_super->log_sectorsize) >
1073                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1074                 le32_to_cpu(raw_super->log_sectorsize) <
1075                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1076                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1077                         le32_to_cpu(raw_super->log_sectorsize));
1078                 return 1;
1079         }
1080         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1081                 le32_to_cpu(raw_super->log_sectorsize) !=
1082                         F2FS_MAX_LOG_SECTOR_SIZE) {
1083                 f2fs_msg(sb, KERN_INFO,
1084                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1085                         le32_to_cpu(raw_super->log_sectors_per_block),
1086                         le32_to_cpu(raw_super->log_sectorsize));
1087                 return 1;
1088         }
1089
1090         /* check reserved ino info */
1091         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1092                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1093                 le32_to_cpu(raw_super->root_ino) != 3) {
1094                 f2fs_msg(sb, KERN_INFO,
1095                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1096                         le32_to_cpu(raw_super->node_ino),
1097                         le32_to_cpu(raw_super->meta_ino),
1098                         le32_to_cpu(raw_super->root_ino));
1099                 return 1;
1100         }
1101
1102         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1103         if (sanity_check_area_boundary(sb, raw_super))
1104                 return 1;
1105
1106         return 0;
1107 }
1108
1109 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1110 {
1111         unsigned int total, fsmeta;
1112         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1113         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1114
1115         total = le32_to_cpu(raw_super->segment_count);
1116         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1117         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1118         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1119         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1120         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1121
1122         if (unlikely(fsmeta >= total))
1123                 return 1;
1124
1125         if (unlikely(f2fs_cp_error(sbi))) {
1126                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1127                 return 1;
1128         }
1129         return 0;
1130 }
1131
1132 static void init_sb_info(struct f2fs_sb_info *sbi)
1133 {
1134         struct f2fs_super_block *raw_super = sbi->raw_super;
1135         int i;
1136
1137         sbi->log_sectors_per_block =
1138                 le32_to_cpu(raw_super->log_sectors_per_block);
1139         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1140         sbi->blocksize = 1 << sbi->log_blocksize;
1141         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1142         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1143         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1144         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1145         sbi->total_sections = le32_to_cpu(raw_super->section_count);
1146         sbi->total_node_count =
1147                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1148                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1149         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1150         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1151         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1152         sbi->cur_victim_sec = NULL_SECNO;
1153         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1154
1155         for (i = 0; i < NR_COUNT_TYPE; i++)
1156                 atomic_set(&sbi->nr_pages[i], 0);
1157
1158         sbi->dir_level = DEF_DIR_LEVEL;
1159         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1160         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1161         clear_sbi_flag(sbi, SBI_NEED_FSCK);
1162
1163         INIT_LIST_HEAD(&sbi->s_list);
1164         mutex_init(&sbi->umount_mutex);
1165 }
1166
1167 /*
1168  * Read f2fs raw super block.
1169  * Because we have two copies of super block, so read the first one at first,
1170  * if the first one is invalid, move to read the second one.
1171  */
1172 static int read_raw_super_block(struct super_block *sb,
1173                         struct f2fs_super_block **raw_super,
1174                         int *valid_super_block, int *recovery)
1175 {
1176         int block = 0;
1177         struct buffer_head *bh;
1178         struct f2fs_super_block *super, *buf;
1179         int err = 0;
1180
1181         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1182         if (!super)
1183                 return -ENOMEM;
1184 retry:
1185         bh = sb_bread(sb, block);
1186         if (!bh) {
1187                 *recovery = 1;
1188                 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1189                                 block + 1);
1190                 err = -EIO;
1191                 goto next;
1192         }
1193
1194         buf = (struct f2fs_super_block *)(bh->b_data + F2FS_SUPER_OFFSET);
1195
1196         /* sanity checking of raw super */
1197         if (sanity_check_raw_super(sb, buf)) {
1198                 brelse(bh);
1199                 *recovery = 1;
1200                 f2fs_msg(sb, KERN_ERR,
1201                         "Can't find valid F2FS filesystem in %dth superblock",
1202                                                                 block + 1);
1203                 err = -EINVAL;
1204                 goto next;
1205         }
1206
1207         if (!*raw_super) {
1208                 memcpy(super, buf, sizeof(*super));
1209                 *valid_super_block = block;
1210                 *raw_super = super;
1211         }
1212         brelse(bh);
1213
1214 next:
1215         /* check the validity of the second superblock */
1216         if (block == 0) {
1217                 block++;
1218                 goto retry;
1219         }
1220
1221         /* No valid superblock */
1222         if (!*raw_super) {
1223                 kfree(super);
1224                 return err;
1225         }
1226
1227         return 0;
1228 }
1229
1230 static int __f2fs_commit_super(struct f2fs_sb_info *sbi, int block)
1231 {
1232         struct f2fs_super_block *super = F2FS_RAW_SUPER(sbi);
1233         struct buffer_head *bh;
1234         int err;
1235
1236         bh = sb_getblk(sbi->sb, block);
1237         if (!bh)
1238                 return -EIO;
1239
1240         lock_buffer(bh);
1241         memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1242         set_buffer_uptodate(bh);
1243         set_buffer_dirty(bh);
1244         unlock_buffer(bh);
1245
1246         /* it's rare case, we can do fua all the time */
1247         err = __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1248         brelse(bh);
1249
1250         return err;
1251 }
1252
1253 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1254 {
1255         int err;
1256
1257         /* write back-up superblock first */
1258         err = __f2fs_commit_super(sbi, sbi->valid_super_block ? 0 : 1);
1259
1260         /* if we are in recovery path, skip writing valid superblock */
1261         if (recover || err)
1262                 return err;
1263
1264         /* write current valid superblock */
1265         return __f2fs_commit_super(sbi, sbi->valid_super_block);
1266 }
1267
1268 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1269 {
1270         struct f2fs_sb_info *sbi;
1271         struct f2fs_super_block *raw_super;
1272         struct inode *root;
1273         long err;
1274         bool retry = true, need_fsck = false;
1275         char *options = NULL;
1276         int recovery, i, valid_super_block;
1277         struct curseg_info *seg_i;
1278
1279 try_onemore:
1280         err = -EINVAL;
1281         raw_super = NULL;
1282         valid_super_block = -1;
1283         recovery = 0;
1284
1285         /* allocate memory for f2fs-specific super block info */
1286         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1287         if (!sbi)
1288                 return -ENOMEM;
1289
1290         /* set a block size */
1291         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1292                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1293                 goto free_sbi;
1294         }
1295
1296         err = read_raw_super_block(sb, &raw_super, &valid_super_block,
1297                                                                 &recovery);
1298         if (err)
1299                 goto free_sbi;
1300
1301         sb->s_fs_info = sbi;
1302         default_options(sbi);
1303         /* parse mount options */
1304         options = kstrdup((const char *)data, GFP_KERNEL);
1305         if (data && !options) {
1306                 err = -ENOMEM;
1307                 goto free_sb_buf;
1308         }
1309
1310         err = parse_options(sb, options);
1311         if (err)
1312                 goto free_options;
1313
1314         sbi->max_file_blocks = max_file_blocks();
1315         sb->s_maxbytes = sbi->max_file_blocks <<
1316                                 le32_to_cpu(raw_super->log_blocksize);
1317         sb->s_max_links = F2FS_LINK_MAX;
1318         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1319
1320         sb->s_op = &f2fs_sops;
1321         sb->s_xattr = f2fs_xattr_handlers;
1322         sb->s_export_op = &f2fs_export_ops;
1323         sb->s_magic = F2FS_SUPER_MAGIC;
1324         sb->s_time_gran = 1;
1325         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1326                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1327         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1328
1329         /* init f2fs-specific super block info */
1330         sbi->sb = sb;
1331         sbi->raw_super = raw_super;
1332         sbi->valid_super_block = valid_super_block;
1333         mutex_init(&sbi->gc_mutex);
1334         mutex_init(&sbi->writepages);
1335         mutex_init(&sbi->cp_mutex);
1336         init_rwsem(&sbi->node_write);
1337
1338         /* disallow all the data/node/meta page writes */
1339         set_sbi_flag(sbi, SBI_POR_DOING);
1340         spin_lock_init(&sbi->stat_lock);
1341
1342         init_rwsem(&sbi->read_io.io_rwsem);
1343         sbi->read_io.sbi = sbi;
1344         sbi->read_io.bio = NULL;
1345         for (i = 0; i < NR_PAGE_TYPE; i++) {
1346                 init_rwsem(&sbi->write_io[i].io_rwsem);
1347                 sbi->write_io[i].sbi = sbi;
1348                 sbi->write_io[i].bio = NULL;
1349         }
1350
1351         init_rwsem(&sbi->cp_rwsem);
1352         init_waitqueue_head(&sbi->cp_wait);
1353         init_sb_info(sbi);
1354
1355         /* get an inode for meta space */
1356         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1357         if (IS_ERR(sbi->meta_inode)) {
1358                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1359                 err = PTR_ERR(sbi->meta_inode);
1360                 goto free_options;
1361         }
1362
1363         err = get_valid_checkpoint(sbi);
1364         if (err) {
1365                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1366                 goto free_meta_inode;
1367         }
1368
1369         /* sanity checking of checkpoint */
1370         err = -EINVAL;
1371         if (sanity_check_ckpt(sbi)) {
1372                 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1373                 goto free_cp;
1374         }
1375
1376         sbi->total_valid_node_count =
1377                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1378         sbi->total_valid_inode_count =
1379                                 le32_to_cpu(sbi->ckpt->valid_inode_count);
1380         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1381         sbi->total_valid_block_count =
1382                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1383         sbi->last_valid_block_count = sbi->total_valid_block_count;
1384         sbi->alloc_valid_block_count = 0;
1385         for (i = 0; i < NR_INODE_TYPE; i++) {
1386                 INIT_LIST_HEAD(&sbi->inode_list[i]);
1387                 spin_lock_init(&sbi->inode_lock[i]);
1388         }
1389
1390         init_extent_cache_info(sbi);
1391
1392         init_ino_entry_info(sbi);
1393
1394         /* setup f2fs internal modules */
1395         err = build_segment_manager(sbi);
1396         if (err) {
1397                 f2fs_msg(sb, KERN_ERR,
1398                         "Failed to initialize F2FS segment manager");
1399                 goto free_sm;
1400         }
1401         err = build_node_manager(sbi);
1402         if (err) {
1403                 f2fs_msg(sb, KERN_ERR,
1404                         "Failed to initialize F2FS node manager");
1405                 goto free_nm;
1406         }
1407
1408         /* For write statistics */
1409         if (sb->s_bdev->bd_part)
1410                 sbi->sectors_written_start =
1411                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1412
1413         /* Read accumulated write IO statistics if exists */
1414         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1415         if (__exist_node_summaries(sbi))
1416                 sbi->kbytes_written =
1417                         le64_to_cpu(seg_i->sum_blk->info.kbytes_written);
1418
1419         build_gc_manager(sbi);
1420
1421         /* get an inode for node space */
1422         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1423         if (IS_ERR(sbi->node_inode)) {
1424                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1425                 err = PTR_ERR(sbi->node_inode);
1426                 goto free_nm;
1427         }
1428
1429         f2fs_join_shrinker(sbi);
1430
1431         /* if there are nt orphan nodes free them */
1432         err = recover_orphan_inodes(sbi);
1433         if (err)
1434                 goto free_node_inode;
1435
1436         /* read root inode and dentry */
1437         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1438         if (IS_ERR(root)) {
1439                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1440                 err = PTR_ERR(root);
1441                 goto free_node_inode;
1442         }
1443         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1444                 iput(root);
1445                 err = -EINVAL;
1446                 goto free_node_inode;
1447         }
1448
1449         sb->s_root = d_make_root(root); /* allocate root dentry */
1450         if (!sb->s_root) {
1451                 err = -ENOMEM;
1452                 goto free_root_inode;
1453         }
1454
1455         err = f2fs_build_stats(sbi);
1456         if (err)
1457                 goto free_root_inode;
1458
1459         if (f2fs_proc_root)
1460                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1461
1462         if (sbi->s_proc)
1463                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1464                                  &f2fs_seq_segment_info_fops, sb);
1465
1466         sbi->s_kobj.kset = f2fs_kset;
1467         init_completion(&sbi->s_kobj_unregister);
1468         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1469                                                         "%s", sb->s_id);
1470         if (err)
1471                 goto free_proc;
1472
1473         /* recover fsynced data */
1474         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1475                 /*
1476                  * mount should be failed, when device has readonly mode, and
1477                  * previous checkpoint was not done by clean system shutdown.
1478                  */
1479                 if (bdev_read_only(sb->s_bdev) &&
1480                                 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1481                         err = -EROFS;
1482                         goto free_kobj;
1483                 }
1484
1485                 if (need_fsck)
1486                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1487
1488                 err = recover_fsync_data(sbi);
1489                 if (err) {
1490                         need_fsck = true;
1491                         f2fs_msg(sb, KERN_ERR,
1492                                 "Cannot recover all fsync data errno=%ld", err);
1493                         goto free_kobj;
1494                 }
1495         }
1496         /* recover_fsync_data() cleared this already */
1497         clear_sbi_flag(sbi, SBI_POR_DOING);
1498
1499         /*
1500          * If filesystem is not mounted as read-only then
1501          * do start the gc_thread.
1502          */
1503         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1504                 /* After POR, we can run background GC thread.*/
1505                 err = start_gc_thread(sbi);
1506                 if (err)
1507                         goto free_kobj;
1508         }
1509         kfree(options);
1510
1511         /* recover broken superblock */
1512         if (recovery && !f2fs_readonly(sb) && !bdev_read_only(sb->s_bdev)) {
1513                 f2fs_msg(sb, KERN_INFO, "Recover invalid superblock");
1514                 f2fs_commit_super(sbi, true);
1515         }
1516
1517         f2fs_update_time(sbi, CP_TIME);
1518         f2fs_update_time(sbi, REQ_TIME);
1519         return 0;
1520
1521 free_kobj:
1522         kobject_del(&sbi->s_kobj);
1523         kobject_put(&sbi->s_kobj);
1524         wait_for_completion(&sbi->s_kobj_unregister);
1525 free_proc:
1526         if (sbi->s_proc) {
1527                 remove_proc_entry("segment_info", sbi->s_proc);
1528                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1529         }
1530         f2fs_destroy_stats(sbi);
1531 free_root_inode:
1532         dput(sb->s_root);
1533         sb->s_root = NULL;
1534 free_node_inode:
1535         mutex_lock(&sbi->umount_mutex);
1536         f2fs_leave_shrinker(sbi);
1537         iput(sbi->node_inode);
1538         mutex_unlock(&sbi->umount_mutex);
1539 free_nm:
1540         destroy_node_manager(sbi);
1541 free_sm:
1542         destroy_segment_manager(sbi);
1543 free_cp:
1544         kfree(sbi->ckpt);
1545 free_meta_inode:
1546         make_bad_inode(sbi->meta_inode);
1547         iput(sbi->meta_inode);
1548 free_options:
1549         kfree(options);
1550 free_sb_buf:
1551         kfree(raw_super);
1552 free_sbi:
1553         kfree(sbi);
1554
1555         /* give only one another chance */
1556         if (retry) {
1557                 retry = false;
1558                 shrink_dcache_sb(sb);
1559                 goto try_onemore;
1560         }
1561         return err;
1562 }
1563
1564 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1565                         const char *dev_name, void *data)
1566 {
1567         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1568 }
1569
1570 static void kill_f2fs_super(struct super_block *sb)
1571 {
1572         if (sb->s_root)
1573                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1574         kill_block_super(sb);
1575 }
1576
1577 static struct file_system_type f2fs_fs_type = {
1578         .owner          = THIS_MODULE,
1579         .name           = "f2fs",
1580         .mount          = f2fs_mount,
1581         .kill_sb        = kill_f2fs_super,
1582         .fs_flags       = FS_REQUIRES_DEV,
1583 };
1584 MODULE_ALIAS_FS("f2fs");
1585
1586 static int __init init_inodecache(void)
1587 {
1588         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1589                         sizeof(struct f2fs_inode_info), 0,
1590                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1591         if (!f2fs_inode_cachep)
1592                 return -ENOMEM;
1593         return 0;
1594 }
1595
1596 static void destroy_inodecache(void)
1597 {
1598         /*
1599          * Make sure all delayed rcu free inodes are flushed before we
1600          * destroy cache.
1601          */
1602         rcu_barrier();
1603         kmem_cache_destroy(f2fs_inode_cachep);
1604 }
1605
1606 static int __init init_f2fs_fs(void)
1607 {
1608         int err;
1609
1610         f2fs_build_trace_ios();
1611
1612         err = init_inodecache();
1613         if (err)
1614                 goto fail;
1615         err = create_node_manager_caches();
1616         if (err)
1617                 goto free_inodecache;
1618         err = create_segment_manager_caches();
1619         if (err)
1620                 goto free_node_manager_caches;
1621         err = create_checkpoint_caches();
1622         if (err)
1623                 goto free_segment_manager_caches;
1624         err = create_extent_cache();
1625         if (err)
1626                 goto free_checkpoint_caches;
1627         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1628         if (!f2fs_kset) {
1629                 err = -ENOMEM;
1630                 goto free_extent_cache;
1631         }
1632         err = f2fs_init_crypto();
1633         if (err)
1634                 goto free_kset;
1635
1636         err = register_shrinker(&f2fs_shrinker_info);
1637         if (err)
1638                 goto free_crypto;
1639
1640         err = register_filesystem(&f2fs_fs_type);
1641         if (err)
1642                 goto free_shrinker;
1643         err = f2fs_create_root_stats();
1644         if (err)
1645                 goto free_filesystem;
1646         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1647         return 0;
1648
1649 free_filesystem:
1650         unregister_filesystem(&f2fs_fs_type);
1651 free_shrinker:
1652         unregister_shrinker(&f2fs_shrinker_info);
1653 free_crypto:
1654         f2fs_exit_crypto();
1655 free_kset:
1656         kset_unregister(f2fs_kset);
1657 free_extent_cache:
1658         destroy_extent_cache();
1659 free_checkpoint_caches:
1660         destroy_checkpoint_caches();
1661 free_segment_manager_caches:
1662         destroy_segment_manager_caches();
1663 free_node_manager_caches:
1664         destroy_node_manager_caches();
1665 free_inodecache:
1666         destroy_inodecache();
1667 fail:
1668         return err;
1669 }
1670
1671 static void __exit exit_f2fs_fs(void)
1672 {
1673         remove_proc_entry("fs/f2fs", NULL);
1674         f2fs_destroy_root_stats();
1675         unregister_shrinker(&f2fs_shrinker_info);
1676         unregister_filesystem(&f2fs_fs_type);
1677         f2fs_exit_crypto();
1678         destroy_extent_cache();
1679         destroy_checkpoint_caches();
1680         destroy_segment_manager_caches();
1681         destroy_node_manager_caches();
1682         destroy_inodecache();
1683         kset_unregister(f2fs_kset);
1684         f2fs_destroy_trace_ios();
1685 }
1686
1687 module_init(init_f2fs_fs)
1688 module_exit(exit_f2fs_fs)
1689
1690 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1691 MODULE_DESCRIPTION("Flash Friendly File System");
1692 MODULE_LICENSE("GPL");