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         return &fi->vfs_inode;
 474 }
 475
 476 static int f2fs_drop_inode(struct inode *inode)
 477 {
 478         /*
 479          * This is to avoid a deadlock condition like below.
 480          * writeback_single_inode(inode)
 481          *  - f2fs_write_data_page
 482          *    - f2fs_gc -> iput -> evict
 483          *       - inode_wait_for_writeback(inode)
 484          */
 485         if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
 486                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 487                         /* to avoid evict_inode call simultaneously */
 488                         atomic_inc(&inode->i_count);
 489                         spin_unlock(&inode->i_lock);
 490
 491                         /* some remained atomic pages should discarded */
 492                         if (f2fs_is_atomic_file(inode))
 493                                 drop_inmem_pages(inode);
 494
 495                         /* should remain fi->extent_tree for writepage */
 496                         f2fs_destroy_extent_node(inode);
 497
 498                         sb_start_intwrite(inode->i_sb);
 499                         i_size_write(inode, 0);
 500
 501                         if (F2FS_HAS_BLOCKS(inode))
 502                                 f2fs_truncate(inode, true);
 503
 504                         sb_end_intwrite(inode->i_sb);
 505
 506                         fscrypt_put_encryption_info(inode, NULL);
 507                         spin_lock(&inode->i_lock);
 508                         atomic_dec(&inode->i_count);
 509                 }
 510                 return 0;
 511         }
 512         return generic_drop_inode(inode);
 513 }
 514
 515 /*
 516  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 517  *
 518  * We should call set_dirty_inode to write the dirty inode through write_inode.
 519  */
 520 static void f2fs_dirty_inode(struct inode *inode, int flags)
 521 {
 522         set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
 523 }
 524
 525 static void f2fs_i_callback(struct rcu_head *head)
 526 {
 527         struct inode *inode = container_of(head, struct inode, i_rcu);
 528         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 529 }
 530
 531 static void f2fs_destroy_inode(struct inode *inode)
 532 {
 533         call_rcu(&inode->i_rcu, f2fs_i_callback);
 534 }
 535
 536 static void f2fs_put_super(struct super_block *sb)
 537 {
 538         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 539
 540         if (sbi->s_proc) {
 541                 remove_proc_entry("segment_info", sbi->s_proc);
 542                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 543         }
 544         kobject_del(&sbi->s_kobj);
 545
 546         stop_gc_thread(sbi);
 547
 548         /* prevent remaining shrinker jobs */
 549         mutex_lock(&sbi->umount_mutex);
 550
 551         /*
 552          * We don't need to do checkpoint when superblock is clean.
 553          * But, the previous checkpoint was not done by umount, it needs to do
 554          * clean checkpoint again.
 555          */
 556         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 557                         !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
 558                 struct cp_control cpc = {
 559                         .reason = CP_UMOUNT,
 560                 };
 561                 write_checkpoint(sbi, &cpc);
 562         }
 563
 564         /* write_checkpoint can update stat informaion */
 565         f2fs_destroy_stats(sbi);
 566
 567         /*
 568          * normally superblock is clean, so we need to release this.
 569          * In addition, EIO will skip do checkpoint, we need this as well.
 570          */
 571         release_ino_entry(sbi);
 572         release_discard_addrs(sbi);
 573
 574         f2fs_leave_shrinker(sbi);
 575         mutex_unlock(&sbi->umount_mutex);
 576
 577         /* our cp_error case, we can wait for any writeback page */
 578         if (get_pages(sbi, F2FS_WRITEBACK))
 579                 f2fs_flush_merged_bios(sbi);
 580
 581         iput(sbi->node_inode);
 582         iput(sbi->meta_inode);
 583
 584         /* destroy f2fs internal modules */
 585         destroy_node_manager(sbi);
 586         destroy_segment_manager(sbi);
 587
 588         kfree(sbi->ckpt);
 589         kobject_put(&sbi->s_kobj);
 590         wait_for_completion(&sbi->s_kobj_unregister);
 591
 592         sb->s_fs_info = NULL;
 593         if (sbi->s_chksum_driver)
 594                 crypto_free_shash(sbi->s_chksum_driver);
 595         kfree(sbi->raw_super);
 596         kfree(sbi);
 597 }
 598
 599 int f2fs_sync_fs(struct super_block *sb, int sync)
 600 {
 601         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 602         int err = 0;
 603
 604         trace_f2fs_sync_fs(sb, sync);
 605
 606         if (sync) {
 607                 struct cp_control cpc;
 608
 609                 cpc.reason = __get_cp_reason(sbi);
 610
 611                 mutex_lock(&sbi->gc_mutex);
 612                 err = write_checkpoint(sbi, &cpc);
 613                 mutex_unlock(&sbi->gc_mutex);
 614         }
 615         f2fs_trace_ios(NULL, 1);
 616
 617         return err;
 618 }
 619
 620 static int f2fs_freeze(struct super_block *sb)
 621 {
 622         int err;
 623
 624         if (f2fs_readonly(sb))
 625                 return 0;
 626
 627         err = f2fs_sync_fs(sb, 1);
 628         return err;
 629 }
 630
 631 static int f2fs_unfreeze(struct super_block *sb)
 632 {
 633         return 0;
 634 }
 635
 636 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 637 {
 638         struct super_block *sb = dentry->d_sb;
 639         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 640         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 641         block_t total_count, user_block_count, start_count, ovp_count;
 642
 643         total_count = le64_to_cpu(sbi->raw_super->block_count);
 644         user_block_count = sbi->user_block_count;
 645         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 646         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 647         buf->f_type = F2FS_SUPER_MAGIC;
 648         buf->f_bsize = sbi->blocksize;
 649
 650         buf->f_blocks = total_count - start_count;
 651         buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
 652         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 653
 654         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 655         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
 656
 657         buf->f_namelen = F2FS_NAME_LEN;
 658         buf->f_fsid.val[0] = (u32)id;
 659         buf->f_fsid.val[1] = (u32)(id >> 32);
 660
 661         return 0;
 662 }
 663
 664 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 665 {
 666         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 667
 668         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
 669                 if (test_opt(sbi, FORCE_FG_GC))
 670                         seq_printf(seq, ",background_gc=%s", "sync");
 671                 else
 672                         seq_printf(seq, ",background_gc=%s", "on");
 673         } else {
 674                 seq_printf(seq, ",background_gc=%s", "off");
 675         }
 676         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 677                 seq_puts(seq, ",disable_roll_forward");
 678         if (test_opt(sbi, DISCARD))
 679                 seq_puts(seq, ",discard");
 680         if (test_opt(sbi, NOHEAP))
 681                 seq_puts(seq, ",no_heap_alloc");
 682 #ifdef CONFIG_F2FS_FS_XATTR
 683         if (test_opt(sbi, XATTR_USER))
 684                 seq_puts(seq, ",user_xattr");
 685         else
 686                 seq_puts(seq, ",nouser_xattr");
 687         if (test_opt(sbi, INLINE_XATTR))
 688                 seq_puts(seq, ",inline_xattr");
 689 #endif
 690 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 691         if (test_opt(sbi, POSIX_ACL))
 692                 seq_puts(seq, ",acl");
 693         else
 694                 seq_puts(seq, ",noacl");
 695 #endif
 696         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 697                 seq_puts(seq, ",disable_ext_identify");
 698         if (test_opt(sbi, INLINE_DATA))
 699                 seq_puts(seq, ",inline_data");
 700         else
 701                 seq_puts(seq, ",noinline_data");
 702         if (test_opt(sbi, INLINE_DENTRY))
 703                 seq_puts(seq, ",inline_dentry");
 704         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 705                 seq_puts(seq, ",flush_merge");
 706         if (test_opt(sbi, NOBARRIER))
 707                 seq_puts(seq, ",nobarrier");
 708         if (test_opt(sbi, FASTBOOT))
 709                 seq_puts(seq, ",fastboot");
 710         if (test_opt(sbi, EXTENT_CACHE))
 711                 seq_puts(seq, ",extent_cache");
 712         else
 713                 seq_puts(seq, ",noextent_cache");
 714         if (test_opt(sbi, DATA_FLUSH))
 715                 seq_puts(seq, ",data_flush");
 716         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 717
 718         return 0;
 719 }
 720
 721 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 722 {
 723         struct super_block *sb = seq->private;
 724         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 725         unsigned int total_segs =
 726                         le32_to_cpu(sbi->raw_super->segment_count_main);
 727         int i;
 728
 729         seq_puts(seq, "format: segment_type|valid_blocks\n"
 730                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 731
 732         for (i = 0; i < total_segs; i++) {
 733                 struct seg_entry *se = get_seg_entry(sbi, i);
 734
 735                 if ((i % 10) == 0)
 736                         seq_printf(seq, "%-10d", i);
 737                 seq_printf(seq, "%d|%-3u", se->type,
 738                                         get_valid_blocks(sbi, i, 1));
 739                 if ((i % 10) == 9 || i == (total_segs - 1))
 740                         seq_putc(seq, '\n');
 741                 else
 742                         seq_putc(seq, ' ');
 743         }
 744
 745         return 0;
 746 }
 747
 748 static int segment_info_open_fs(struct inode *inode, struct file *file)
 749 {
 750         return single_open(file, segment_info_seq_show, PDE_DATA(inode));
 751 }
 752
 753 static const struct file_operations f2fs_seq_segment_info_fops = {
 754         .owner = THIS_MODULE,
 755         .open = segment_info_open_fs,
 756         .read = seq_read,
 757         .llseek = seq_lseek,
 758         .release = single_release,
 759 };
 760
 761 static void default_options(struct f2fs_sb_info *sbi)
 762 {
 763         /* init some FS parameters */
 764         sbi->active_logs = NR_CURSEG_TYPE;
 765
 766         set_opt(sbi, BG_GC);
 767         set_opt(sbi, INLINE_DATA);
 768         set_opt(sbi, EXTENT_CACHE);
 769
 770 #ifdef CONFIG_F2FS_FS_XATTR
 771         set_opt(sbi, XATTR_USER);
 772 #endif
 773 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 774         set_opt(sbi, POSIX_ACL);
 775 #endif
 776 }
 777
 778 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 779 {
 780         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 781         struct f2fs_mount_info org_mount_opt;
 782         int err, active_logs;
 783         bool need_restart_gc = false;
 784         bool need_stop_gc = false;
 785         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
 786
 787         /*
 788          * Save the old mount options in case we
 789          * need to restore them.
 790          */
 791         org_mount_opt = sbi->mount_opt;
 792         active_logs = sbi->active_logs;
 793
 794         if (*flags & MS_RDONLY) {
 795                 set_opt(sbi, FASTBOOT);
 796                 set_sbi_flag(sbi, SBI_IS_DIRTY);
 797         }
 798
 799         sync_filesystem(sb);
 800
 801         sbi->mount_opt.opt = 0;
 802         default_options(sbi);
 803
 804         /* parse mount options */
 805         err = parse_options(sb, data);
 806         if (err)
 807                 goto restore_opts;
 808
 809         /*
 810          * Previous and new state of filesystem is RO,
 811          * so skip checking GC and FLUSH_MERGE conditions.
 812          */
 813         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
 814                 goto skip;
 815
 816         /* disallow enable/disable extent_cache dynamically */
 817         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
 818                 err = -EINVAL;
 819                 f2fs_msg(sbi->sb, KERN_WARNING,
 820                                 "switch extent_cache option is not allowed");
 821                 goto restore_opts;
 822         }
 823
 824         /*
 825          * We stop the GC thread if FS is mounted as RO
 826          * or if background_gc = off is passed in mount
 827          * option. Also sync the filesystem.
 828          */
 829         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
 830                 if (sbi->gc_thread) {
 831                         stop_gc_thread(sbi);
 832                         f2fs_sync_fs(sb, 1);
 833                         need_restart_gc = true;
 834                 }
 835         } else if (!sbi->gc_thread) {
 836                 err = start_gc_thread(sbi);
 837                 if (err)
 838                         goto restore_opts;
 839                 need_stop_gc = true;
 840         }
 841
 842         /*
 843          * We stop issue flush thread if FS is mounted as RO
 844          * or if flush_merge is not passed in mount option.
 845          */
 846         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
 847                 destroy_flush_cmd_control(sbi);
 848         } else if (!SM_I(sbi)->cmd_control_info) {
 849                 err = create_flush_cmd_control(sbi);
 850                 if (err)
 851                         goto restore_gc;
 852         }
 853 skip:
 854         /* Update the POSIXACL Flag */
 855          sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 856                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 857         return 0;
 858 restore_gc:
 859         if (need_restart_gc) {
 860                 if (start_gc_thread(sbi))
 861                         f2fs_msg(sbi->sb, KERN_WARNING,
 862                                 "background gc thread has stopped");
 863         } else if (need_stop_gc) {
 864                 stop_gc_thread(sbi);
 865         }
 866 restore_opts:
 867         sbi->mount_opt = org_mount_opt;
 868         sbi->active_logs = active_logs;
 869         return err;
 870 }
 871
 872 static struct super_operations f2fs_sops = {
 873         .alloc_inode    = f2fs_alloc_inode,
 874         .drop_inode     = f2fs_drop_inode,
 875         .destroy_inode  = f2fs_destroy_inode,
 876         .write_inode    = f2fs_write_inode,
 877         .dirty_inode    = f2fs_dirty_inode,
 878         .show_options   = f2fs_show_options,
 879         .evict_inode    = f2fs_evict_inode,
 880         .put_super      = f2fs_put_super,
 881         .sync_fs        = f2fs_sync_fs,
 882         .freeze_fs      = f2fs_freeze,
 883         .unfreeze_fs    = f2fs_unfreeze,
 884         .statfs         = f2fs_statfs,
 885         .remount_fs     = f2fs_remount,
 886 };
 887
 888 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 889 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
 890 {
 891         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
 892                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
 893                                 ctx, len, NULL);
 894 }
 895
 896 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
 897                                                         void *fs_data)
 898 {
 899         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
 900                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
 901                                 ctx, len, fs_data, XATTR_CREATE);
 902 }
 903
 904 static unsigned f2fs_max_namelen(struct inode *inode)
 905 {
 906         return S_ISLNK(inode->i_mode) ?
 907                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
 908 }
 909
 910 static struct fscrypt_operations f2fs_cryptops = {
 911         .get_context    = f2fs_get_context,
 912         .set_context    = f2fs_set_context,
 913         .is_encrypted   = f2fs_encrypted_inode,
 914         .empty_dir      = f2fs_empty_dir,
 915         .max_namelen    = f2fs_max_namelen,
 916 };
 917 #else
 918 static struct fscrypt_operations f2fs_cryptops = {
 919         .is_encrypted   = f2fs_encrypted_inode,
 920 };
 921 #endif
 922
 923 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
 924                 u64 ino, u32 generation)
 925 {
 926         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 927         struct inode *inode;
 928
 929         if (check_nid_range(sbi, ino))
 930                 return ERR_PTR(-ESTALE);
 931
 932         /*
 933          * f2fs_iget isn't quite right if the inode is currently unallocated!
 934          * However f2fs_iget currently does appropriate checks to handle stale
 935          * inodes so everything is OK.
 936          */
 937         inode = f2fs_iget(sb, ino);
 938         if (IS_ERR(inode))
 939                 return ERR_CAST(inode);
 940         if (unlikely(generation && inode->i_generation != generation)) {
 941                 /* we didn't find the right inode.. */
 942                 iput(inode);
 943                 return ERR_PTR(-ESTALE);
 944         }
 945         return inode;
 946 }
 947
 948 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
 949                 int fh_len, int fh_type)
 950 {
 951         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 952                                     f2fs_nfs_get_inode);
 953 }
 954
 955 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
 956                 int fh_len, int fh_type)
 957 {
 958         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 959                                     f2fs_nfs_get_inode);
 960 }
 961
 962 static const struct export_operations f2fs_export_ops = {
 963         .fh_to_dentry = f2fs_fh_to_dentry,
 964         .fh_to_parent = f2fs_fh_to_parent,
 965         .get_parent = f2fs_get_parent,
 966 };
 967
 968 static loff_t max_file_blocks(void)
 969 {
 970         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
 971         loff_t leaf_count = ADDRS_PER_BLOCK;
 972
 973         /* two direct node blocks */
 974         result += (leaf_count * 2);
 975
 976         /* two indirect node blocks */
 977         leaf_count *= NIDS_PER_BLOCK;
 978         result += (leaf_count * 2);
 979
 980         /* one double indirect node block */
 981         leaf_count *= NIDS_PER_BLOCK;
 982         result += leaf_count;
 983
 984         return result;
 985 }
 986
 987 static int __f2fs_commit_super(struct buffer_head *bh,
 988                         struct f2fs_super_block *super)
 989 {
 990         lock_buffer(bh);
 991         if (super)
 992                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
 993         set_buffer_uptodate(bh);
 994         set_buffer_dirty(bh);
 995         unlock_buffer(bh);
 996
 997         /* it's rare case, we can do fua all the time */
 998         return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
 999 }
1000
1001 static inline bool sanity_check_area_boundary(struct super_block *sb,
1002                                         struct buffer_head *bh)
1003 {
1004         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1005                                         (bh->b_data + F2FS_SUPER_OFFSET);
1006         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1007         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1008         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1009         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1010         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1011         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1012         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1013         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1014         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1015         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1016         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1017         u32 segment_count = le32_to_cpu(raw_super->segment_count);
1018         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1019         u64 main_end_blkaddr = main_blkaddr +
1020                                 (segment_count_main << log_blocks_per_seg);
1021         u64 seg_end_blkaddr = segment0_blkaddr +
1022                                 (segment_count << log_blocks_per_seg);
1023
1024         if (segment0_blkaddr != cp_blkaddr) {
1025                 f2fs_msg(sb, KERN_INFO,
1026                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1027                         segment0_blkaddr, cp_blkaddr);
1028                 return true;
1029         }
1030
1031         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1032                                                         sit_blkaddr) {
1033                 f2fs_msg(sb, KERN_INFO,
1034                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1035                         cp_blkaddr, sit_blkaddr,
1036                         segment_count_ckpt << log_blocks_per_seg);
1037                 return true;
1038         }
1039
1040         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1041                                                         nat_blkaddr) {
1042                 f2fs_msg(sb, KERN_INFO,
1043                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1044                         sit_blkaddr, nat_blkaddr,
1045                         segment_count_sit << log_blocks_per_seg);
1046                 return true;
1047         }
1048
1049         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1050                                                         ssa_blkaddr) {
1051                 f2fs_msg(sb, KERN_INFO,
1052                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1053                         nat_blkaddr, ssa_blkaddr,
1054                         segment_count_nat << log_blocks_per_seg);
1055                 return true;
1056         }
1057
1058         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1059                                                         main_blkaddr) {
1060                 f2fs_msg(sb, KERN_INFO,
1061                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1062                         ssa_blkaddr, main_blkaddr,
1063                         segment_count_ssa << log_blocks_per_seg);
1064                 return true;
1065         }
1066
1067         if (main_end_blkaddr > seg_end_blkaddr) {
1068                 f2fs_msg(sb, KERN_INFO,
1069                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1070                         main_blkaddr,
1071                         segment0_blkaddr +
1072                                 (segment_count << log_blocks_per_seg),
1073                         segment_count_main << log_blocks_per_seg);
1074                 return true;
1075         } else if (main_end_blkaddr < seg_end_blkaddr) {
1076                 int err = 0;
1077                 char *res;
1078
1079                 /* fix in-memory information all the time */
1080                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1081                                 segment0_blkaddr) >> log_blocks_per_seg);
1082
1083                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1084                         res = "internally";
1085                 } else {
1086                         err = __f2fs_commit_super(bh, NULL);
1087                         res = err ? "failed" : "done";
1088                 }
1089                 f2fs_msg(sb, KERN_INFO,
1090                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
1091                         res, main_blkaddr,
1092                         segment0_blkaddr +
1093                                 (segment_count << log_blocks_per_seg),
1094                         segment_count_main << log_blocks_per_seg);
1095                 if (err)
1096                         return true;
1097         }
1098         return false;
1099 }
1100
1101 static int sanity_check_raw_super(struct super_block *sb,
1102                                 struct buffer_head *bh)
1103 {
1104         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1105                                         (bh->b_data + F2FS_SUPER_OFFSET);
1106         unsigned int blocksize;
1107
1108         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1109                 f2fs_msg(sb, KERN_INFO,
1110                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1111                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1112                 return 1;
1113         }
1114
1115         /* Currently, support only 4KB page cache size */
1116         if (F2FS_BLKSIZE != PAGE_SIZE) {
1117                 f2fs_msg(sb, KERN_INFO,
1118                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1119                         PAGE_SIZE);
1120                 return 1;
1121         }
1122
1123         /* Currently, support only 4KB block size */
1124         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1125         if (blocksize != F2FS_BLKSIZE) {
1126                 f2fs_msg(sb, KERN_INFO,
1127                         "Invalid blocksize (%u), supports only 4KB\n",
1128                         blocksize);
1129                 return 1;
1130         }
1131
1132         /* check log blocks per segment */
1133         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1134                 f2fs_msg(sb, KERN_INFO,
1135                         "Invalid log blocks per segment (%u)\n",
1136                         le32_to_cpu(raw_super->log_blocks_per_seg));
1137                 return 1;
1138         }
1139
1140         /* Currently, support 512/1024/2048/4096 bytes sector size */
1141         if (le32_to_cpu(raw_super->log_sectorsize) >
1142                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1143                 le32_to_cpu(raw_super->log_sectorsize) <
1144                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1145                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1146                         le32_to_cpu(raw_super->log_sectorsize));
1147                 return 1;
1148         }
1149         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1150                 le32_to_cpu(raw_super->log_sectorsize) !=
1151                         F2FS_MAX_LOG_SECTOR_SIZE) {
1152                 f2fs_msg(sb, KERN_INFO,
1153                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1154                         le32_to_cpu(raw_super->log_sectors_per_block),
1155                         le32_to_cpu(raw_super->log_sectorsize));
1156                 return 1;
1157         }
1158
1159         /* check reserved ino info */
1160         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1161                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1162                 le32_to_cpu(raw_super->root_ino) != 3) {
1163                 f2fs_msg(sb, KERN_INFO,
1164                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1165                         le32_to_cpu(raw_super->node_ino),
1166                         le32_to_cpu(raw_super->meta_ino),
1167                         le32_to_cpu(raw_super->root_ino));
1168                 return 1;
1169         }
1170
1171         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1172         if (sanity_check_area_boundary(sb, bh))
1173                 return 1;
1174
1175         return 0;
1176 }
1177
1178 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1179 {
1180         unsigned int total, fsmeta;
1181         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1182         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1183
1184         total = le32_to_cpu(raw_super->segment_count);
1185         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1186         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1187         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1188         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1189         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1190
1191         if (unlikely(fsmeta >= total))
1192                 return 1;
1193
1194         if (unlikely(f2fs_cp_error(sbi))) {
1195                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1196                 return 1;
1197         }
1198         return 0;
1199 }
1200
1201 static void init_sb_info(struct f2fs_sb_info *sbi)
1202 {
1203         struct f2fs_super_block *raw_super = sbi->raw_super;
1204         int i;
1205
1206         sbi->log_sectors_per_block =
1207                 le32_to_cpu(raw_super->log_sectors_per_block);
1208         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1209         sbi->blocksize = 1 << sbi->log_blocksize;
1210         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1211         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1212         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1213         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1214         sbi->total_sections = le32_to_cpu(raw_super->section_count);
1215         sbi->total_node_count =
1216                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1217                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1218         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1219         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1220         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1221         sbi->cur_victim_sec = NULL_SECNO;
1222         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1223
1224         for (i = 0; i < NR_COUNT_TYPE; i++)
1225                 atomic_set(&sbi->nr_pages[i], 0);
1226
1227         sbi->dir_level = DEF_DIR_LEVEL;
1228         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1229         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1230         clear_sbi_flag(sbi, SBI_NEED_FSCK);
1231
1232         INIT_LIST_HEAD(&sbi->s_list);
1233         mutex_init(&sbi->umount_mutex);
1234 }
1235
1236 /*
1237  * Read f2fs raw super block.
1238  * Because we have two copies of super block, so read both of them
1239  * to get the first valid one. If any one of them is broken, we pass
1240  * them recovery flag back to the caller.
1241  */
1242 static int read_raw_super_block(struct super_block *sb,
1243                         struct f2fs_super_block **raw_super,
1244                         int *valid_super_block, int *recovery)
1245 {
1246         int block;
1247         struct buffer_head *bh;
1248         struct f2fs_super_block *super;
1249         int err = 0;
1250
1251         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1252         if (!super)
1253                 return -ENOMEM;
1254
1255         for (block = 0; block < 2; block++) {
1256                 bh = sb_bread(sb, block);
1257                 if (!bh) {
1258                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1259                                 block + 1);
1260                         err = -EIO;
1261                         continue;
1262                 }
1263
1264                 /* sanity checking of raw super */
1265                 if (sanity_check_raw_super(sb, bh)) {
1266                         f2fs_msg(sb, KERN_ERR,
1267                                 "Can't find valid F2FS filesystem in %dth superblock",
1268                                 block + 1);
1269                         err = -EINVAL;
1270                         brelse(bh);
1271                         continue;
1272                 }
1273
1274                 if (!*raw_super) {
1275                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1276                                                         sizeof(*super));
1277                         *valid_super_block = block;
1278                         *raw_super = super;
1279                 }
1280                 brelse(bh);
1281         }
1282
1283         /* Fail to read any one of the superblocks*/
1284         if (err < 0)
1285                 *recovery = 1;
1286
1287         /* No valid superblock */
1288         if (!*raw_super)
1289                 kfree(super);
1290         else
1291                 err = 0;
1292
1293         return err;
1294 }
1295
1296 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1297 {
1298         struct buffer_head *bh;
1299         int err;
1300
1301         /* write back-up superblock first */
1302         bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1303         if (!bh)
1304                 return -EIO;
1305         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1306         brelse(bh);
1307
1308         /* if we are in recovery path, skip writing valid superblock */
1309         if (recover || err)
1310                 return err;
1311
1312         /* write current valid superblock */
1313         bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1314         if (!bh)
1315                 return -EIO;
1316         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1317         brelse(bh);
1318         return err;
1319 }
1320
1321 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1322 {
1323         struct f2fs_sb_info *sbi;
1324         struct f2fs_super_block *raw_super;
1325         struct inode *root;
1326         long err;
1327         bool retry = true, need_fsck = false;
1328         char *options = NULL;
1329         int recovery, i, valid_super_block;
1330         struct curseg_info *seg_i;
1331
1332 try_onemore:
1333         err = -EINVAL;
1334         raw_super = NULL;
1335         valid_super_block = -1;
1336         recovery = 0;
1337
1338         /* allocate memory for f2fs-specific super block info */
1339         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1340         if (!sbi)
1341                 return -ENOMEM;
1342
1343         /* Load the checksum driver */
1344         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1345         if (IS_ERR(sbi->s_chksum_driver)) {
1346                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1347                 err = PTR_ERR(sbi->s_chksum_driver);
1348                 sbi->s_chksum_driver = NULL;
1349                 goto free_sbi;
1350         }
1351
1352         /* set a block size */
1353         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1354                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1355                 goto free_sbi;
1356         }
1357
1358         err = read_raw_super_block(sb, &raw_super, &valid_super_block,
1359                                                                 &recovery);
1360         if (err)
1361                 goto free_sbi;
1362
1363         sb->s_fs_info = sbi;
1364         default_options(sbi);
1365         /* parse mount options */
1366         options = kstrdup((const char *)data, GFP_KERNEL);
1367         if (data && !options) {
1368                 err = -ENOMEM;
1369                 goto free_sb_buf;
1370         }
1371
1372         err = parse_options(sb, options);
1373         if (err)
1374                 goto free_options;
1375
1376         sbi->max_file_blocks = max_file_blocks();
1377         sb->s_maxbytes = sbi->max_file_blocks <<
1378                                 le32_to_cpu(raw_super->log_blocksize);
1379         sb->s_max_links = F2FS_LINK_MAX;
1380         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1381
1382         sb->s_op = &f2fs_sops;
1383         sb->s_cop = &f2fs_cryptops;
1384         sb->s_xattr = f2fs_xattr_handlers;
1385         sb->s_export_op = &f2fs_export_ops;
1386         sb->s_magic = F2FS_SUPER_MAGIC;
1387         sb->s_time_gran = 1;
1388         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1389                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1390         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1391
1392         /* init f2fs-specific super block info */
1393         sbi->sb = sb;
1394         sbi->raw_super = raw_super;
1395         sbi->valid_super_block = valid_super_block;
1396         mutex_init(&sbi->gc_mutex);
1397         mutex_init(&sbi->writepages);
1398         mutex_init(&sbi->cp_mutex);
1399         init_rwsem(&sbi->node_write);
1400
1401         /* disallow all the data/node/meta page writes */
1402         set_sbi_flag(sbi, SBI_POR_DOING);
1403         spin_lock_init(&sbi->stat_lock);
1404
1405         init_rwsem(&sbi->read_io.io_rwsem);
1406         sbi->read_io.sbi = sbi;
1407         sbi->read_io.bio = NULL;
1408         for (i = 0; i < NR_PAGE_TYPE; i++) {
1409                 init_rwsem(&sbi->write_io[i].io_rwsem);
1410                 sbi->write_io[i].sbi = sbi;
1411                 sbi->write_io[i].bio = NULL;
1412         }
1413
1414         init_rwsem(&sbi->cp_rwsem);
1415         init_waitqueue_head(&sbi->cp_wait);
1416         init_sb_info(sbi);
1417
1418         /* get an inode for meta space */
1419         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1420         if (IS_ERR(sbi->meta_inode)) {
1421                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1422                 err = PTR_ERR(sbi->meta_inode);
1423                 goto free_options;
1424         }
1425
1426         err = get_valid_checkpoint(sbi);
1427         if (err) {
1428                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1429                 goto free_meta_inode;
1430         }
1431
1432         sbi->total_valid_node_count =
1433                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1434         sbi->total_valid_inode_count =
1435                                 le32_to_cpu(sbi->ckpt->valid_inode_count);
1436         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1437         sbi->total_valid_block_count =
1438                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1439         sbi->last_valid_block_count = sbi->total_valid_block_count;
1440         sbi->alloc_valid_block_count = 0;
1441         for (i = 0; i < NR_INODE_TYPE; i++) {
1442                 INIT_LIST_HEAD(&sbi->inode_list[i]);
1443                 spin_lock_init(&sbi->inode_lock[i]);
1444         }
1445
1446         init_extent_cache_info(sbi);
1447
1448         init_ino_entry_info(sbi);
1449
1450         /* setup f2fs internal modules */
1451         err = build_segment_manager(sbi);
1452         if (err) {
1453                 f2fs_msg(sb, KERN_ERR,
1454                         "Failed to initialize F2FS segment manager");
1455                 goto free_sm;
1456         }
1457         err = build_node_manager(sbi);
1458         if (err) {
1459                 f2fs_msg(sb, KERN_ERR,
1460                         "Failed to initialize F2FS node manager");
1461                 goto free_nm;
1462         }
1463
1464         /* For write statistics */
1465         if (sb->s_bdev->bd_part)
1466                 sbi->sectors_written_start =
1467                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1468
1469         /* Read accumulated write IO statistics if exists */
1470         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1471         if (__exist_node_summaries(sbi))
1472                 sbi->kbytes_written =
1473                         le64_to_cpu(seg_i->journal->info.kbytes_written);
1474
1475         build_gc_manager(sbi);
1476
1477         /* get an inode for node space */
1478         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1479         if (IS_ERR(sbi->node_inode)) {
1480                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1481                 err = PTR_ERR(sbi->node_inode);
1482                 goto free_nm;
1483         }
1484
1485         f2fs_join_shrinker(sbi);
1486
1487         /* if there are nt orphan nodes free them */
1488         err = recover_orphan_inodes(sbi);
1489         if (err)
1490                 goto free_node_inode;
1491
1492         /* read root inode and dentry */
1493         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1494         if (IS_ERR(root)) {
1495                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1496                 err = PTR_ERR(root);
1497                 goto free_node_inode;
1498         }
1499         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1500                 iput(root);
1501                 err = -EINVAL;
1502                 goto free_node_inode;
1503         }
1504
1505         sb->s_root = d_make_root(root); /* allocate root dentry */
1506         if (!sb->s_root) {
1507                 err = -ENOMEM;
1508                 goto free_root_inode;
1509         }
1510
1511         err = f2fs_build_stats(sbi);
1512         if (err)
1513                 goto free_root_inode;
1514
1515         if (f2fs_proc_root)
1516                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1517
1518         if (sbi->s_proc)
1519                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1520                                  &f2fs_seq_segment_info_fops, sb);
1521
1522         sbi->s_kobj.kset = f2fs_kset;
1523         init_completion(&sbi->s_kobj_unregister);
1524         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1525                                                         "%s", sb->s_id);
1526         if (err)
1527                 goto free_proc;
1528
1529         /* recover fsynced data */
1530         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1531                 /*
1532                  * mount should be failed, when device has readonly mode, and
1533                  * previous checkpoint was not done by clean system shutdown.
1534                  */
1535                 if (bdev_read_only(sb->s_bdev) &&
1536                                 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1537                         err = -EROFS;
1538                         goto free_kobj;
1539                 }
1540
1541                 if (need_fsck)
1542                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1543
1544                 err = recover_fsync_data(sbi);
1545                 if (err) {
1546                         need_fsck = true;
1547                         f2fs_msg(sb, KERN_ERR,
1548                                 "Cannot recover all fsync data errno=%ld", err);
1549                         goto free_kobj;
1550                 }
1551         }
1552         /* recover_fsync_data() cleared this already */
1553         clear_sbi_flag(sbi, SBI_POR_DOING);
1554
1555         /*
1556          * If filesystem is not mounted as read-only then
1557          * do start the gc_thread.
1558          */
1559         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1560                 /* After POR, we can run background GC thread.*/
1561                 err = start_gc_thread(sbi);
1562                 if (err)
1563                         goto free_kobj;
1564         }
1565         kfree(options);
1566
1567         /* recover broken superblock */
1568         if (recovery && !f2fs_readonly(sb) && !bdev_read_only(sb->s_bdev)) {
1569                 err = f2fs_commit_super(sbi, true);
1570                 f2fs_msg(sb, KERN_INFO,
1571                         "Try to recover %dth superblock, ret: %ld",
1572                         sbi->valid_super_block ? 1 : 2, err);
1573         }
1574
1575         f2fs_update_time(sbi, CP_TIME);
1576         f2fs_update_time(sbi, REQ_TIME);
1577         return 0;
1578
1579 free_kobj:
1580         kobject_del(&sbi->s_kobj);
1581         kobject_put(&sbi->s_kobj);
1582         wait_for_completion(&sbi->s_kobj_unregister);
1583 free_proc:
1584         if (sbi->s_proc) {
1585                 remove_proc_entry("segment_info", sbi->s_proc);
1586                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1587         }
1588         f2fs_destroy_stats(sbi);
1589 free_root_inode:
1590         dput(sb->s_root);
1591         sb->s_root = NULL;
1592 free_node_inode:
1593         mutex_lock(&sbi->umount_mutex);
1594         f2fs_leave_shrinker(sbi);
1595         iput(sbi->node_inode);
1596         mutex_unlock(&sbi->umount_mutex);
1597 free_nm:
1598         destroy_node_manager(sbi);
1599 free_sm:
1600         destroy_segment_manager(sbi);
1601         kfree(sbi->ckpt);
1602 free_meta_inode:
1603         make_bad_inode(sbi->meta_inode);
1604         iput(sbi->meta_inode);
1605 free_options:
1606         kfree(options);
1607 free_sb_buf:
1608         kfree(raw_super);
1609 free_sbi:
1610         if (sbi->s_chksum_driver)
1611                 crypto_free_shash(sbi->s_chksum_driver);
1612         kfree(sbi);
1613
1614         /* give only one another chance */
1615         if (retry) {
1616                 retry = false;
1617                 shrink_dcache_sb(sb);
1618                 goto try_onemore;
1619         }
1620         return err;
1621 }
1622
1623 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1624                         const char *dev_name, void *data)
1625 {
1626         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1627 }
1628
1629 static void kill_f2fs_super(struct super_block *sb)
1630 {
1631         if (sb->s_root)
1632                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1633         kill_block_super(sb);
1634 }
1635
1636 static struct file_system_type f2fs_fs_type = {
1637         .owner          = THIS_MODULE,
1638         .name           = "f2fs",
1639         .mount          = f2fs_mount,
1640         .kill_sb        = kill_f2fs_super,
1641         .fs_flags       = FS_REQUIRES_DEV,
1642 };
1643 MODULE_ALIAS_FS("f2fs");
1644
1645 static int __init init_inodecache(void)
1646 {
1647         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1648                         sizeof(struct f2fs_inode_info), 0,
1649                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1650         if (!f2fs_inode_cachep)
1651                 return -ENOMEM;
1652         return 0;
1653 }
1654
1655 static void destroy_inodecache(void)
1656 {
1657         /*
1658          * Make sure all delayed rcu free inodes are flushed before we
1659          * destroy cache.
1660          */
1661         rcu_barrier();
1662         kmem_cache_destroy(f2fs_inode_cachep);
1663 }
1664
1665 static int __init init_f2fs_fs(void)
1666 {
1667         int err;
1668
1669         f2fs_build_trace_ios();
1670
1671         err = init_inodecache();
1672         if (err)
1673                 goto fail;
1674         err = create_node_manager_caches();
1675         if (err)
1676                 goto free_inodecache;
1677         err = create_segment_manager_caches();
1678         if (err)
1679                 goto free_node_manager_caches;
1680         err = create_checkpoint_caches();
1681         if (err)
1682                 goto free_segment_manager_caches;
1683         err = create_extent_cache();
1684         if (err)
1685                 goto free_checkpoint_caches;
1686         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1687         if (!f2fs_kset) {
1688                 err = -ENOMEM;
1689                 goto free_extent_cache;
1690         }
1691         err = register_shrinker(&f2fs_shrinker_info);
1692         if (err)
1693                 goto free_kset;
1694
1695         err = register_filesystem(&f2fs_fs_type);
1696         if (err)
1697                 goto free_shrinker;
1698         err = f2fs_create_root_stats();
1699         if (err)
1700                 goto free_filesystem;
1701         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1702         return 0;
1703
1704 free_filesystem:
1705         unregister_filesystem(&f2fs_fs_type);
1706 free_shrinker:
1707         unregister_shrinker(&f2fs_shrinker_info);
1708 free_kset:
1709         kset_unregister(f2fs_kset);
1710 free_extent_cache:
1711         destroy_extent_cache();
1712 free_checkpoint_caches:
1713         destroy_checkpoint_caches();
1714 free_segment_manager_caches:
1715         destroy_segment_manager_caches();
1716 free_node_manager_caches:
1717         destroy_node_manager_caches();
1718 free_inodecache:
1719         destroy_inodecache();
1720 fail:
1721         return err;
1722 }
1723
1724 static void __exit exit_f2fs_fs(void)
1725 {
1726         remove_proc_entry("fs/f2fs", NULL);
1727         f2fs_destroy_root_stats();
1728         unregister_shrinker(&f2fs_shrinker_info);
1729         unregister_filesystem(&f2fs_fs_type);
1730         destroy_extent_cache();
1731         destroy_checkpoint_caches();
1732         destroy_segment_manager_caches();
1733         destroy_node_manager_caches();
1734         destroy_inodecache();
1735         kset_unregister(f2fs_kset);
1736         f2fs_destroy_trace_ios();
1737 }
1738
1739 module_init(init_f2fs_fs)
1740 module_exit(exit_f2fs_fs)
1741
1742 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1743 MODULE_DESCRIPTION("Flash Friendly File System");
1744 MODULE_LICENSE("GPL");