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
  34 #define CREATE_TRACE_POINTS
  35 #include <trace/events/f2fs.h>
  36
  37 static struct proc_dir_entry *f2fs_proc_root;
  38 static struct kmem_cache *f2fs_inode_cachep;
  39 static struct kset *f2fs_kset;
  40
  41 enum {
  42         Opt_gc_background,
  43         Opt_disable_roll_forward,
  44         Opt_discard,
  45         Opt_noheap,
  46         Opt_user_xattr,
  47         Opt_nouser_xattr,
  48         Opt_acl,
  49         Opt_noacl,
  50         Opt_active_logs,
  51         Opt_disable_ext_identify,
  52         Opt_inline_xattr,
  53         Opt_inline_data,
  54         Opt_err,
  55 };
  56
  57 static match_table_t f2fs_tokens = {
  58         {Opt_gc_background, "background_gc=%s"},
  59         {Opt_disable_roll_forward, "disable_roll_forward"},
  60         {Opt_discard, "discard"},
  61         {Opt_noheap, "no_heap"},
  62         {Opt_user_xattr, "user_xattr"},
  63         {Opt_nouser_xattr, "nouser_xattr"},
  64         {Opt_acl, "acl"},
  65         {Opt_noacl, "noacl"},
  66         {Opt_active_logs, "active_logs=%u"},
  67         {Opt_disable_ext_identify, "disable_ext_identify"},
  68         {Opt_inline_xattr, "inline_xattr"},
  69         {Opt_inline_data, "inline_data"},
  70         {Opt_err, NULL},
  71 };
  72
  73 /* Sysfs support for f2fs */
  74 enum {
  75         GC_THREAD,      /* struct f2fs_gc_thread */
  76         SM_INFO,        /* struct f2fs_sm_info */
  77         F2FS_SBI,       /* struct f2fs_sb_info */
  78 };
  79
  80 struct f2fs_attr {
  81         struct attribute attr;
  82         ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
  83         ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
  84                          const char *, size_t);
  85         int struct_type;
  86         int offset;
  87 };
  88
  89 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
  90 {
  91         if (struct_type == GC_THREAD)
  92                 return (unsigned char *)sbi->gc_thread;
  93         else if (struct_type == SM_INFO)
  94                 return (unsigned char *)SM_I(sbi);
  95         else if (struct_type == F2FS_SBI)
  96                 return (unsigned char *)sbi;
  97         return NULL;
  98 }
  99
 100 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
 101                         struct f2fs_sb_info *sbi, char *buf)
 102 {
 103         unsigned char *ptr = NULL;
 104         unsigned int *ui;
 105
 106         ptr = __struct_ptr(sbi, a->struct_type);
 107         if (!ptr)
 108                 return -EINVAL;
 109
 110         ui = (unsigned int *)(ptr + a->offset);
 111
 112         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
 113 }
 114
 115 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
 116                         struct f2fs_sb_info *sbi,
 117                         const char *buf, size_t count)
 118 {
 119         unsigned char *ptr;
 120         unsigned long t;
 121         unsigned int *ui;
 122         ssize_t ret;
 123
 124         ptr = __struct_ptr(sbi, a->struct_type);
 125         if (!ptr)
 126                 return -EINVAL;
 127
 128         ui = (unsigned int *)(ptr + a->offset);
 129
 130         ret = kstrtoul(skip_spaces(buf), 0, &t);
 131         if (ret < 0)
 132                 return ret;
 133         *ui = t;
 134         return count;
 135 }
 136
 137 static ssize_t f2fs_attr_show(struct kobject *kobj,
 138                                 struct attribute *attr, char *buf)
 139 {
 140         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 141                                                                 s_kobj);
 142         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 143
 144         return a->show ? a->show(a, sbi, buf) : 0;
 145 }
 146
 147 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
 148                                                 const char *buf, size_t len)
 149 {
 150         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 151                                                                         s_kobj);
 152         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 153
 154         return a->store ? a->store(a, sbi, buf, len) : 0;
 155 }
 156
 157 static void f2fs_sb_release(struct kobject *kobj)
 158 {
 159         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 160                                                                 s_kobj);
 161         complete(&sbi->s_kobj_unregister);
 162 }
 163
 164 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
 165 static struct f2fs_attr f2fs_attr_##_name = {                   \
 166         .attr = {.name = __stringify(_name), .mode = _mode },   \
 167         .show   = _show,                                        \
 168         .store  = _store,                                       \
 169         .struct_type = _struct_type,                            \
 170         .offset = _offset                                       \
 171 }
 172
 173 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)    \
 174         F2FS_ATTR_OFFSET(struct_type, name, 0644,               \
 175                 f2fs_sbi_show, f2fs_sbi_store,                  \
 176                 offsetof(struct struct_name, elname))
 177
 178 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
 179 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
 180 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
 181 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
 182 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
 183 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
 184 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
 185 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
 186 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
 187
 188 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 189 static struct attribute *f2fs_attrs[] = {
 190         ATTR_LIST(gc_min_sleep_time),
 191         ATTR_LIST(gc_max_sleep_time),
 192         ATTR_LIST(gc_no_gc_sleep_time),
 193         ATTR_LIST(gc_idle),
 194         ATTR_LIST(reclaim_segments),
 195         ATTR_LIST(max_small_discards),
 196         ATTR_LIST(ipu_policy),
 197         ATTR_LIST(min_ipu_util),
 198         ATTR_LIST(max_victim_search),
 199         NULL,
 200 };
 201
 202 static const struct sysfs_ops f2fs_attr_ops = {
 203         .show   = f2fs_attr_show,
 204         .store  = f2fs_attr_store,
 205 };
 206
 207 static struct kobj_type f2fs_ktype = {
 208         .default_attrs  = f2fs_attrs,
 209         .sysfs_ops      = &f2fs_attr_ops,
 210         .release        = f2fs_sb_release,
 211 };
 212
 213 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 214 {
 215         struct va_format vaf;
 216         va_list args;
 217
 218         va_start(args, fmt);
 219         vaf.fmt = fmt;
 220         vaf.va = &args;
 221         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 222         va_end(args);
 223 }
 224
 225 static void init_once(void *foo)
 226 {
 227         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 228
 229         inode_init_once(&fi->vfs_inode);
 230 }
 231
 232 static int parse_options(struct super_block *sb, char *options)
 233 {
 234         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 235         substring_t args[MAX_OPT_ARGS];
 236         char *p, *name;
 237         int arg = 0;
 238
 239         if (!options)
 240                 return 0;
 241
 242         while ((p = strsep(&options, ",")) != NULL) {
 243                 int token;
 244                 if (!*p)
 245                         continue;
 246                 /*
 247                  * Initialize args struct so we know whether arg was
 248                  * found; some options take optional arguments.
 249                  */
 250                 args[0].to = args[0].from = NULL;
 251                 token = match_token(p, f2fs_tokens, args);
 252
 253                 switch (token) {
 254                 case Opt_gc_background:
 255                         name = match_strdup(&args[0]);
 256
 257                         if (!name)
 258                                 return -ENOMEM;
 259                         if (!strncmp(name, "on", 2))
 260                                 set_opt(sbi, BG_GC);
 261                         else if (!strncmp(name, "off", 3))
 262                                 clear_opt(sbi, BG_GC);
 263                         else {
 264                                 kfree(name);
 265                                 return -EINVAL;
 266                         }
 267                         kfree(name);
 268                         break;
 269                 case Opt_disable_roll_forward:
 270                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 271                         break;
 272                 case Opt_discard:
 273                         set_opt(sbi, DISCARD);
 274                         break;
 275                 case Opt_noheap:
 276                         set_opt(sbi, NOHEAP);
 277                         break;
 278 #ifdef CONFIG_F2FS_FS_XATTR
 279                 case Opt_user_xattr:
 280                         set_opt(sbi, XATTR_USER);
 281                         break;
 282                 case Opt_nouser_xattr:
 283                         clear_opt(sbi, XATTR_USER);
 284                         break;
 285                 case Opt_inline_xattr:
 286                         set_opt(sbi, INLINE_XATTR);
 287                         break;
 288 #else
 289                 case Opt_user_xattr:
 290                         f2fs_msg(sb, KERN_INFO,
 291                                 "user_xattr options not supported");
 292                         break;
 293                 case Opt_nouser_xattr:
 294                         f2fs_msg(sb, KERN_INFO,
 295                                 "nouser_xattr options not supported");
 296                         break;
 297                 case Opt_inline_xattr:
 298                         f2fs_msg(sb, KERN_INFO,
 299                                 "inline_xattr options not supported");
 300                         break;
 301 #endif
 302 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 303                 case Opt_acl:
 304                         set_opt(sbi, POSIX_ACL);
 305                         break;
 306                 case Opt_noacl:
 307                         clear_opt(sbi, POSIX_ACL);
 308                         break;
 309 #else
 310                 case Opt_acl:
 311                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 312                         break;
 313                 case Opt_noacl:
 314                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 315                         break;
 316 #endif
 317                 case Opt_active_logs:
 318                         if (args->from && match_int(args, &arg))
 319                                 return -EINVAL;
 320                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 321                                 return -EINVAL;
 322                         sbi->active_logs = arg;
 323                         break;
 324                 case Opt_disable_ext_identify:
 325                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 326                         break;
 327                 case Opt_inline_data:
 328                         set_opt(sbi, INLINE_DATA);
 329                         break;
 330                 default:
 331                         f2fs_msg(sb, KERN_ERR,
 332                                 "Unrecognized mount option \"%s\" or missing value",
 333                                 p);
 334                         return -EINVAL;
 335                 }
 336         }
 337         return 0;
 338 }
 339
 340 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 341 {
 342         struct f2fs_inode_info *fi;
 343
 344         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 345         if (!fi)
 346                 return NULL;
 347
 348         init_once((void *) fi);
 349
 350         /* Initialize f2fs-specific inode info */
 351         fi->vfs_inode.i_version = 1;
 352         atomic_set(&fi->dirty_dents, 0);
 353         fi->i_current_depth = 1;
 354         fi->i_advise = 0;
 355         rwlock_init(&fi->ext.ext_lock);
 356
 357         set_inode_flag(fi, FI_NEW_INODE);
 358
 359         if (test_opt(F2FS_SB(sb), INLINE_XATTR))
 360                 set_inode_flag(fi, FI_INLINE_XATTR);
 361
 362         return &fi->vfs_inode;
 363 }
 364
 365 static int f2fs_drop_inode(struct inode *inode)
 366 {
 367         /*
 368          * This is to avoid a deadlock condition like below.
 369          * writeback_single_inode(inode)
 370          *  - f2fs_write_data_page
 371          *    - f2fs_gc -> iput -> evict
 372          *       - inode_wait_for_writeback(inode)
 373          */
 374         if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
 375                 return 0;
 376         return generic_drop_inode(inode);
 377 }
 378
 379 /*
 380  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 381  *
 382  * We should call set_dirty_inode to write the dirty inode through write_inode.
 383  */
 384 static void f2fs_dirty_inode(struct inode *inode, int flags)
 385 {
 386         set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
 387 }
 388
 389 static void f2fs_i_callback(struct rcu_head *head)
 390 {
 391         struct inode *inode = container_of(head, struct inode, i_rcu);
 392         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 393 }
 394
 395 static void f2fs_destroy_inode(struct inode *inode)
 396 {
 397         call_rcu(&inode->i_rcu, f2fs_i_callback);
 398 }
 399
 400 static void f2fs_put_super(struct super_block *sb)
 401 {
 402         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 403
 404         if (sbi->s_proc) {
 405                 remove_proc_entry("segment_info", sbi->s_proc);
 406                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 407         }
 408         kobject_del(&sbi->s_kobj);
 409
 410         f2fs_destroy_stats(sbi);
 411         stop_gc_thread(sbi);
 412
 413         /* We don't need to do checkpoint when it's clean */
 414         if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
 415                 write_checkpoint(sbi, true);
 416
 417         iput(sbi->node_inode);
 418         iput(sbi->meta_inode);
 419
 420         /* destroy f2fs internal modules */
 421         destroy_node_manager(sbi);
 422         destroy_segment_manager(sbi);
 423
 424         kfree(sbi->ckpt);
 425         kobject_put(&sbi->s_kobj);
 426         wait_for_completion(&sbi->s_kobj_unregister);
 427
 428         sb->s_fs_info = NULL;
 429         brelse(sbi->raw_super_buf);
 430         kfree(sbi);
 431 }
 432
 433 int f2fs_sync_fs(struct super_block *sb, int sync)
 434 {
 435         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 436
 437         trace_f2fs_sync_fs(sb, sync);
 438
 439         if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
 440                 return 0;
 441
 442         if (sync) {
 443                 mutex_lock(&sbi->gc_mutex);
 444                 write_checkpoint(sbi, false);
 445                 mutex_unlock(&sbi->gc_mutex);
 446         } else {
 447                 f2fs_balance_fs(sbi);
 448         }
 449
 450         return 0;
 451 }
 452
 453 static int f2fs_freeze(struct super_block *sb)
 454 {
 455         int err;
 456
 457         if (f2fs_readonly(sb))
 458                 return 0;
 459
 460         err = f2fs_sync_fs(sb, 1);
 461         return err;
 462 }
 463
 464 static int f2fs_unfreeze(struct super_block *sb)
 465 {
 466         return 0;
 467 }
 468
 469 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 470 {
 471         struct super_block *sb = dentry->d_sb;
 472         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 473         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 474         block_t total_count, user_block_count, start_count, ovp_count;
 475
 476         total_count = le64_to_cpu(sbi->raw_super->block_count);
 477         user_block_count = sbi->user_block_count;
 478         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 479         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 480         buf->f_type = F2FS_SUPER_MAGIC;
 481         buf->f_bsize = sbi->blocksize;
 482
 483         buf->f_blocks = total_count - start_count;
 484         buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
 485         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 486
 487         buf->f_files = sbi->total_node_count;
 488         buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
 489
 490         buf->f_namelen = F2FS_NAME_LEN;
 491         buf->f_fsid.val[0] = (u32)id;
 492         buf->f_fsid.val[1] = (u32)(id >> 32);
 493
 494         return 0;
 495 }
 496
 497 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 498 {
 499         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 500
 501         if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
 502                 seq_printf(seq, ",background_gc=%s", "on");
 503         else
 504                 seq_printf(seq, ",background_gc=%s", "off");
 505         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 506                 seq_puts(seq, ",disable_roll_forward");
 507         if (test_opt(sbi, DISCARD))
 508                 seq_puts(seq, ",discard");
 509         if (test_opt(sbi, NOHEAP))
 510                 seq_puts(seq, ",no_heap_alloc");
 511 #ifdef CONFIG_F2FS_FS_XATTR
 512         if (test_opt(sbi, XATTR_USER))
 513                 seq_puts(seq, ",user_xattr");
 514         else
 515                 seq_puts(seq, ",nouser_xattr");
 516         if (test_opt(sbi, INLINE_XATTR))
 517                 seq_puts(seq, ",inline_xattr");
 518 #endif
 519 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 520         if (test_opt(sbi, POSIX_ACL))
 521                 seq_puts(seq, ",acl");
 522         else
 523                 seq_puts(seq, ",noacl");
 524 #endif
 525         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 526                 seq_puts(seq, ",disable_ext_identify");
 527         if (test_opt(sbi, INLINE_DATA))
 528                 seq_puts(seq, ",inline_data");
 529         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 530
 531         return 0;
 532 }
 533
 534 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 535 {
 536         struct super_block *sb = seq->private;
 537         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 538         unsigned int total_segs =
 539                         le32_to_cpu(sbi->raw_super->segment_count_main);
 540         int i;
 541
 542         for (i = 0; i < total_segs; i++) {
 543                 seq_printf(seq, "%u", get_valid_blocks(sbi, i, 1));
 544                 if (i != 0 && (i % 10) == 0)
 545                         seq_puts(seq, "\n");
 546                 else
 547                         seq_puts(seq, " ");
 548         }
 549         return 0;
 550 }
 551
 552 static int segment_info_open_fs(struct inode *inode, struct file *file)
 553 {
 554         return single_open(file, segment_info_seq_show, PDE_DATA(inode));
 555 }
 556
 557 static const struct file_operations f2fs_seq_segment_info_fops = {
 558         .owner = THIS_MODULE,
 559         .open = segment_info_open_fs,
 560         .read = seq_read,
 561         .llseek = seq_lseek,
 562         .release = single_release,
 563 };
 564
 565 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 566 {
 567         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 568         struct f2fs_mount_info org_mount_opt;
 569         int err, active_logs;
 570
 571         sync_filesystem(sb);
 572
 573         /*
 574          * Save the old mount options in case we
 575          * need to restore them.
 576          */
 577         org_mount_opt = sbi->mount_opt;
 578         active_logs = sbi->active_logs;
 579
 580         /* parse mount options */
 581         err = parse_options(sb, data);
 582         if (err)
 583                 goto restore_opts;
 584
 585         /*
 586          * Previous and new state of filesystem is RO,
 587          * so no point in checking GC conditions.
 588          */
 589         if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
 590                 goto skip;
 591
 592         /*
 593          * We stop the GC thread if FS is mounted as RO
 594          * or if background_gc = off is passed in mount
 595          * option. Also sync the filesystem.
 596          */
 597         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
 598                 if (sbi->gc_thread) {
 599                         stop_gc_thread(sbi);
 600                         f2fs_sync_fs(sb, 1);
 601                 }
 602         } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
 603                 err = start_gc_thread(sbi);
 604                 if (err)
 605                         goto restore_opts;
 606         }
 607 skip:
 608         /* Update the POSIXACL Flag */
 609          sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 610                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 611         return 0;
 612
 613 restore_opts:
 614         sbi->mount_opt = org_mount_opt;
 615         sbi->active_logs = active_logs;
 616         return err;
 617 }
 618
 619 static struct super_operations f2fs_sops = {
 620         .alloc_inode    = f2fs_alloc_inode,
 621         .drop_inode     = f2fs_drop_inode,
 622         .destroy_inode  = f2fs_destroy_inode,
 623         .write_inode    = f2fs_write_inode,
 624         .dirty_inode    = f2fs_dirty_inode,
 625         .show_options   = f2fs_show_options,
 626         .evict_inode    = f2fs_evict_inode,
 627         .put_super      = f2fs_put_super,
 628         .sync_fs        = f2fs_sync_fs,
 629         .freeze_fs      = f2fs_freeze,
 630         .unfreeze_fs    = f2fs_unfreeze,
 631         .statfs         = f2fs_statfs,
 632         .remount_fs     = f2fs_remount,
 633 };
 634
 635 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
 636                 u64 ino, u32 generation)
 637 {
 638         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 639         struct inode *inode;
 640
 641         if (unlikely(ino < F2FS_ROOT_INO(sbi)))
 642                 return ERR_PTR(-ESTALE);
 643
 644         /*
 645          * f2fs_iget isn't quite right if the inode is currently unallocated!
 646          * However f2fs_iget currently does appropriate checks to handle stale
 647          * inodes so everything is OK.
 648          */
 649         inode = f2fs_iget(sb, ino);
 650         if (IS_ERR(inode))
 651                 return ERR_CAST(inode);
 652         if (unlikely(generation && inode->i_generation != generation)) {
 653                 /* we didn't find the right inode.. */
 654                 iput(inode);
 655                 return ERR_PTR(-ESTALE);
 656         }
 657         return inode;
 658 }
 659
 660 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
 661                 int fh_len, int fh_type)
 662 {
 663         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 664                                     f2fs_nfs_get_inode);
 665 }
 666
 667 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
 668                 int fh_len, int fh_type)
 669 {
 670         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 671                                     f2fs_nfs_get_inode);
 672 }
 673
 674 static const struct export_operations f2fs_export_ops = {
 675         .fh_to_dentry = f2fs_fh_to_dentry,
 676         .fh_to_parent = f2fs_fh_to_parent,
 677         .get_parent = f2fs_get_parent,
 678 };
 679
 680 static loff_t max_file_size(unsigned bits)
 681 {
 682         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
 683         loff_t leaf_count = ADDRS_PER_BLOCK;
 684
 685         /* two direct node blocks */
 686         result += (leaf_count * 2);
 687
 688         /* two indirect node blocks */
 689         leaf_count *= NIDS_PER_BLOCK;
 690         result += (leaf_count * 2);
 691
 692         /* one double indirect node block */
 693         leaf_count *= NIDS_PER_BLOCK;
 694         result += leaf_count;
 695
 696         result <<= bits;
 697         return result;
 698 }
 699
 700 static int sanity_check_raw_super(struct super_block *sb,
 701                         struct f2fs_super_block *raw_super)
 702 {
 703         unsigned int blocksize;
 704
 705         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
 706                 f2fs_msg(sb, KERN_INFO,
 707                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
 708                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
 709                 return 1;
 710         }
 711
 712         /* Currently, support only 4KB page cache size */
 713         if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
 714                 f2fs_msg(sb, KERN_INFO,
 715                         "Invalid page_cache_size (%lu), supports only 4KB\n",
 716                         PAGE_CACHE_SIZE);
 717                 return 1;
 718         }
 719
 720         /* Currently, support only 4KB block size */
 721         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
 722         if (blocksize != F2FS_BLKSIZE) {
 723                 f2fs_msg(sb, KERN_INFO,
 724                         "Invalid blocksize (%u), supports only 4KB\n",
 725                         blocksize);
 726                 return 1;
 727         }
 728
 729         if (le32_to_cpu(raw_super->log_sectorsize) !=
 730                                         F2FS_LOG_SECTOR_SIZE) {
 731                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
 732                 return 1;
 733         }
 734         if (le32_to_cpu(raw_super->log_sectors_per_block) !=
 735                                         F2FS_LOG_SECTORS_PER_BLOCK) {
 736                 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
 737                 return 1;
 738         }
 739         return 0;
 740 }
 741
 742 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
 743 {
 744         unsigned int total, fsmeta;
 745         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
 746         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 747
 748         total = le32_to_cpu(raw_super->segment_count);
 749         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
 750         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
 751         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
 752         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
 753         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
 754
 755         if (unlikely(fsmeta >= total))
 756                 return 1;
 757
 758         if (unlikely(is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
 759                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
 760                 return 1;
 761         }
 762         return 0;
 763 }
 764
 765 static void init_sb_info(struct f2fs_sb_info *sbi)
 766 {
 767         struct f2fs_super_block *raw_super = sbi->raw_super;
 768         int i;
 769
 770         sbi->log_sectors_per_block =
 771                 le32_to_cpu(raw_super->log_sectors_per_block);
 772         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
 773         sbi->blocksize = 1 << sbi->log_blocksize;
 774         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
 775         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
 776         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
 777         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
 778         sbi->total_sections = le32_to_cpu(raw_super->section_count);
 779         sbi->total_node_count =
 780                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
 781                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
 782         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
 783         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
 784         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
 785         sbi->cur_victim_sec = NULL_SECNO;
 786         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
 787
 788         for (i = 0; i < NR_COUNT_TYPE; i++)
 789                 atomic_set(&sbi->nr_pages[i], 0);
 790 }
 791
 792 /*
 793  * Read f2fs raw super block.
 794  * Because we have two copies of super block, so read the first one at first,
 795  * if the first one is invalid, move to read the second one.
 796  */
 797 static int read_raw_super_block(struct super_block *sb,
 798                         struct f2fs_super_block **raw_super,
 799                         struct buffer_head **raw_super_buf)
 800 {
 801         int block = 0;
 802
 803 retry:
 804         *raw_super_buf = sb_bread(sb, block);
 805         if (!*raw_super_buf) {
 806                 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
 807                                 block + 1);
 808                 if (block == 0) {
 809                         block++;
 810                         goto retry;
 811                 } else {
 812                         return -EIO;
 813                 }
 814         }
 815
 816         *raw_super = (struct f2fs_super_block *)
 817                 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
 818
 819         /* sanity checking of raw super */
 820         if (sanity_check_raw_super(sb, *raw_super)) {
 821                 brelse(*raw_super_buf);
 822                 f2fs_msg(sb, KERN_ERR,
 823                         "Can't find valid F2FS filesystem in %dth superblock",
 824                                                                 block + 1);
 825                 if (block == 0) {
 826                         block++;
 827                         goto retry;
 828                 } else {
 829                         return -EINVAL;
 830                 }
 831         }
 832
 833         return 0;
 834 }
 835
 836 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 837 {
 838         struct f2fs_sb_info *sbi;
 839         struct f2fs_super_block *raw_super;
 840         struct buffer_head *raw_super_buf;
 841         struct inode *root;
 842         long err = -EINVAL;
 843         int i;
 844
 845         /* allocate memory for f2fs-specific super block info */
 846         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
 847         if (!sbi)
 848                 return -ENOMEM;
 849
 850         /* set a block size */
 851         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
 852                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
 853                 goto free_sbi;
 854         }
 855
 856         err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
 857         if (err)
 858                 goto free_sbi;
 859
 860         sb->s_fs_info = sbi;
 861         /* init some FS parameters */
 862         sbi->active_logs = NR_CURSEG_TYPE;
 863
 864         set_opt(sbi, BG_GC);
 865
 866 #ifdef CONFIG_F2FS_FS_XATTR
 867         set_opt(sbi, XATTR_USER);
 868 #endif
 869 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 870         set_opt(sbi, POSIX_ACL);
 871 #endif
 872         /* parse mount options */
 873         err = parse_options(sb, (char *)data);
 874         if (err)
 875                 goto free_sb_buf;
 876
 877         sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
 878         sb->s_max_links = F2FS_LINK_MAX;
 879         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
 880
 881         sb->s_op = &f2fs_sops;
 882         sb->s_xattr = f2fs_xattr_handlers;
 883         sb->s_export_op = &f2fs_export_ops;
 884         sb->s_magic = F2FS_SUPER_MAGIC;
 885         sb->s_time_gran = 1;
 886         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 887                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 888         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
 889
 890         /* init f2fs-specific super block info */
 891         sbi->sb = sb;
 892         sbi->raw_super = raw_super;
 893         sbi->raw_super_buf = raw_super_buf;
 894         mutex_init(&sbi->gc_mutex);
 895         mutex_init(&sbi->writepages);
 896         mutex_init(&sbi->cp_mutex);
 897         mutex_init(&sbi->node_write);
 898         sbi->por_doing = false;
 899         spin_lock_init(&sbi->stat_lock);
 900
 901         mutex_init(&sbi->read_io.io_mutex);
 902         sbi->read_io.sbi = sbi;
 903         sbi->read_io.bio = NULL;
 904         for (i = 0; i < NR_PAGE_TYPE; i++) {
 905                 mutex_init(&sbi->write_io[i].io_mutex);
 906                 sbi->write_io[i].sbi = sbi;
 907                 sbi->write_io[i].bio = NULL;
 908         }
 909
 910         init_rwsem(&sbi->cp_rwsem);
 911         init_waitqueue_head(&sbi->cp_wait);
 912         init_sb_info(sbi);
 913
 914         /* get an inode for meta space */
 915         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
 916         if (IS_ERR(sbi->meta_inode)) {
 917                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
 918                 err = PTR_ERR(sbi->meta_inode);
 919                 goto free_sb_buf;
 920         }
 921
 922         err = get_valid_checkpoint(sbi);
 923         if (err) {
 924                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
 925                 goto free_meta_inode;
 926         }
 927
 928         /* sanity checking of checkpoint */
 929         err = -EINVAL;
 930         if (sanity_check_ckpt(sbi)) {
 931                 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
 932                 goto free_cp;
 933         }
 934
 935         sbi->total_valid_node_count =
 936                                 le32_to_cpu(sbi->ckpt->valid_node_count);
 937         sbi->total_valid_inode_count =
 938                                 le32_to_cpu(sbi->ckpt->valid_inode_count);
 939         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
 940         sbi->total_valid_block_count =
 941                                 le64_to_cpu(sbi->ckpt->valid_block_count);
 942         sbi->last_valid_block_count = sbi->total_valid_block_count;
 943         sbi->alloc_valid_block_count = 0;
 944         INIT_LIST_HEAD(&sbi->dir_inode_list);
 945         spin_lock_init(&sbi->dir_inode_lock);
 946
 947         init_orphan_info(sbi);
 948
 949         /* setup f2fs internal modules */
 950         err = build_segment_manager(sbi);
 951         if (err) {
 952                 f2fs_msg(sb, KERN_ERR,
 953                         "Failed to initialize F2FS segment manager");
 954                 goto free_sm;
 955         }
 956         err = build_node_manager(sbi);
 957         if (err) {
 958                 f2fs_msg(sb, KERN_ERR,
 959                         "Failed to initialize F2FS node manager");
 960                 goto free_nm;
 961         }
 962
 963         build_gc_manager(sbi);
 964
 965         /* get an inode for node space */
 966         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
 967         if (IS_ERR(sbi->node_inode)) {
 968                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
 969                 err = PTR_ERR(sbi->node_inode);
 970                 goto free_nm;
 971         }
 972
 973         /* if there are nt orphan nodes free them */
 974         recover_orphan_inodes(sbi);
 975
 976         /* read root inode and dentry */
 977         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
 978         if (IS_ERR(root)) {
 979                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
 980                 err = PTR_ERR(root);
 981                 goto free_node_inode;
 982         }
 983         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
 984                 err = -EINVAL;
 985                 goto free_root_inode;
 986         }
 987
 988         sb->s_root = d_make_root(root); /* allocate root dentry */
 989         if (!sb->s_root) {
 990                 err = -ENOMEM;
 991                 goto free_root_inode;
 992         }
 993
 994         /* recover fsynced data */
 995         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
 996                 err = recover_fsync_data(sbi);
 997                 if (err)
 998                         f2fs_msg(sb, KERN_ERR,
 999                                 "Cannot recover all fsync data errno=%ld", err);
1000         }
1001
1002         /*
1003          * If filesystem is not mounted as read-only then
1004          * do start the gc_thread.
1005          */
1006         if (!(sb->s_flags & MS_RDONLY)) {
1007                 /* After POR, we can run background GC thread.*/
1008                 err = start_gc_thread(sbi);
1009                 if (err)
1010                         goto free_gc;
1011         }
1012
1013         err = f2fs_build_stats(sbi);
1014         if (err)
1015                 goto free_gc;
1016
1017         if (f2fs_proc_root)
1018                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1019
1020         if (sbi->s_proc)
1021                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1022                                  &f2fs_seq_segment_info_fops, sb);
1023
1024         if (test_opt(sbi, DISCARD)) {
1025                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1026                 if (!blk_queue_discard(q))
1027                         f2fs_msg(sb, KERN_WARNING,
1028                                         "mounting with \"discard\" option, but "
1029                                         "the device does not support discard");
1030         }
1031
1032         sbi->s_kobj.kset = f2fs_kset;
1033         init_completion(&sbi->s_kobj_unregister);
1034         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1035                                                         "%s", sb->s_id);
1036         if (err)
1037                 goto fail;
1038
1039         return 0;
1040 fail:
1041         if (sbi->s_proc) {
1042                 remove_proc_entry("segment_info", sbi->s_proc);
1043                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1044         }
1045         f2fs_destroy_stats(sbi);
1046 free_gc:
1047         stop_gc_thread(sbi);
1048 free_root_inode:
1049         dput(sb->s_root);
1050         sb->s_root = NULL;
1051 free_node_inode:
1052         iput(sbi->node_inode);
1053 free_nm:
1054         destroy_node_manager(sbi);
1055 free_sm:
1056         destroy_segment_manager(sbi);
1057 free_cp:
1058         kfree(sbi->ckpt);
1059 free_meta_inode:
1060         make_bad_inode(sbi->meta_inode);
1061         iput(sbi->meta_inode);
1062 free_sb_buf:
1063         brelse(raw_super_buf);
1064 free_sbi:
1065         kfree(sbi);
1066         return err;
1067 }
1068
1069 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1070                         const char *dev_name, void *data)
1071 {
1072         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1073 }
1074
1075 static struct file_system_type f2fs_fs_type = {
1076         .owner          = THIS_MODULE,
1077         .name           = "f2fs",
1078         .mount          = f2fs_mount,
1079         .kill_sb        = kill_block_super,
1080         .fs_flags       = FS_REQUIRES_DEV,
1081 };
1082 MODULE_ALIAS_FS("f2fs");
1083
1084 static int __init init_inodecache(void)
1085 {
1086         f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1087                         sizeof(struct f2fs_inode_info), NULL);
1088         if (!f2fs_inode_cachep)
1089                 return -ENOMEM;
1090         return 0;
1091 }
1092
1093 static void destroy_inodecache(void)
1094 {
1095         /*
1096          * Make sure all delayed rcu free inodes are flushed before we
1097          * destroy cache.
1098          */
1099         rcu_barrier();
1100         kmem_cache_destroy(f2fs_inode_cachep);
1101 }
1102
1103 static int __init init_f2fs_fs(void)
1104 {
1105         int err;
1106
1107         err = init_inodecache();
1108         if (err)
1109                 goto fail;
1110         err = create_node_manager_caches();
1111         if (err)
1112                 goto free_inodecache;
1113         err = create_segment_manager_caches();
1114         if (err)
1115                 goto free_node_manager_caches;
1116         err = create_gc_caches();
1117         if (err)
1118                 goto free_segment_manager_caches;
1119         err = create_checkpoint_caches();
1120         if (err)
1121                 goto free_gc_caches;
1122         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1123         if (!f2fs_kset) {
1124                 err = -ENOMEM;
1125                 goto free_checkpoint_caches;
1126         }
1127         err = register_filesystem(&f2fs_fs_type);
1128         if (err)
1129                 goto free_kset;
1130         f2fs_create_root_stats();
1131         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1132         return 0;
1133
1134 free_kset:
1135         kset_unregister(f2fs_kset);
1136 free_checkpoint_caches:
1137         destroy_checkpoint_caches();
1138 free_gc_caches:
1139         destroy_gc_caches();
1140 free_segment_manager_caches:
1141         destroy_segment_manager_caches();
1142 free_node_manager_caches:
1143         destroy_node_manager_caches();
1144 free_inodecache:
1145         destroy_inodecache();
1146 fail:
1147         return err;
1148 }
1149
1150 static void __exit exit_f2fs_fs(void)
1151 {
1152         remove_proc_entry("fs/f2fs", NULL);
1153         f2fs_destroy_root_stats();
1154         unregister_filesystem(&f2fs_fs_type);
1155         destroy_checkpoint_caches();
1156         destroy_gc_caches();
1157         destroy_segment_manager_caches();
1158         destroy_node_manager_caches();
1159         destroy_inodecache();
1160         kset_unregister(f2fs_kset);
1161 }
1162
1163 module_init(init_f2fs_fs)
1164 module_exit(exit_f2fs_fs)
1165
1166 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1167 MODULE_DESCRIPTION("Flash Friendly File System");
1168 MODULE_LICENSE("GPL");