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 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
 188
 189 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 190 static struct attribute *f2fs_attrs[] = {
 191         ATTR_LIST(gc_min_sleep_time),
 192         ATTR_LIST(gc_max_sleep_time),
 193         ATTR_LIST(gc_no_gc_sleep_time),
 194         ATTR_LIST(gc_idle),
 195         ATTR_LIST(reclaim_segments),
 196         ATTR_LIST(max_small_discards),
 197         ATTR_LIST(ipu_policy),
 198         ATTR_LIST(min_ipu_util),
 199         ATTR_LIST(max_victim_search),
 200         ATTR_LIST(dir_level),
 201         NULL,
 202 };
 203
 204 static const struct sysfs_ops f2fs_attr_ops = {
 205         .show   = f2fs_attr_show,
 206         .store  = f2fs_attr_store,
 207 };
 208
 209 static struct kobj_type f2fs_ktype = {
 210         .default_attrs  = f2fs_attrs,
 211         .sysfs_ops      = &f2fs_attr_ops,
 212         .release        = f2fs_sb_release,
 213 };
 214
 215 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 216 {
 217         struct va_format vaf;
 218         va_list args;
 219
 220         va_start(args, fmt);
 221         vaf.fmt = fmt;
 222         vaf.va = &args;
 223         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 224         va_end(args);
 225 }
 226
 227 static void init_once(void *foo)
 228 {
 229         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 230
 231         inode_init_once(&fi->vfs_inode);
 232 }
 233
 234 static int parse_options(struct super_block *sb, char *options)
 235 {
 236         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 237         substring_t args[MAX_OPT_ARGS];
 238         char *p, *name;
 239         int arg = 0;
 240
 241         if (!options)
 242                 return 0;
 243
 244         while ((p = strsep(&options, ",")) != NULL) {
 245                 int token;
 246                 if (!*p)
 247                         continue;
 248                 /*
 249                  * Initialize args struct so we know whether arg was
 250                  * found; some options take optional arguments.
 251                  */
 252                 args[0].to = args[0].from = NULL;
 253                 token = match_token(p, f2fs_tokens, args);
 254
 255                 switch (token) {
 256                 case Opt_gc_background:
 257                         name = match_strdup(&args[0]);
 258
 259                         if (!name)
 260                                 return -ENOMEM;
 261                         if (!strncmp(name, "on", 2))
 262                                 set_opt(sbi, BG_GC);
 263                         else if (!strncmp(name, "off", 3))
 264                                 clear_opt(sbi, BG_GC);
 265                         else {
 266                                 kfree(name);
 267                                 return -EINVAL;
 268                         }
 269                         kfree(name);
 270                         break;
 271                 case Opt_disable_roll_forward:
 272                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 273                         break;
 274                 case Opt_discard:
 275                         set_opt(sbi, DISCARD);
 276                         break;
 277                 case Opt_noheap:
 278                         set_opt(sbi, NOHEAP);
 279                         break;
 280 #ifdef CONFIG_F2FS_FS_XATTR
 281                 case Opt_user_xattr:
 282                         set_opt(sbi, XATTR_USER);
 283                         break;
 284                 case Opt_nouser_xattr:
 285                         clear_opt(sbi, XATTR_USER);
 286                         break;
 287                 case Opt_inline_xattr:
 288                         set_opt(sbi, INLINE_XATTR);
 289                         break;
 290 #else
 291                 case Opt_user_xattr:
 292                         f2fs_msg(sb, KERN_INFO,
 293                                 "user_xattr options not supported");
 294                         break;
 295                 case Opt_nouser_xattr:
 296                         f2fs_msg(sb, KERN_INFO,
 297                                 "nouser_xattr options not supported");
 298                         break;
 299                 case Opt_inline_xattr:
 300                         f2fs_msg(sb, KERN_INFO,
 301                                 "inline_xattr options not supported");
 302                         break;
 303 #endif
 304 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 305                 case Opt_acl:
 306                         set_opt(sbi, POSIX_ACL);
 307                         break;
 308                 case Opt_noacl:
 309                         clear_opt(sbi, POSIX_ACL);
 310                         break;
 311 #else
 312                 case Opt_acl:
 313                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 314                         break;
 315                 case Opt_noacl:
 316                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 317                         break;
 318 #endif
 319                 case Opt_active_logs:
 320                         if (args->from && match_int(args, &arg))
 321                                 return -EINVAL;
 322                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 323                                 return -EINVAL;
 324                         sbi->active_logs = arg;
 325                         break;
 326                 case Opt_disable_ext_identify:
 327                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 328                         break;
 329                 case Opt_inline_data:
 330                         set_opt(sbi, INLINE_DATA);
 331                         break;
 332                 default:
 333                         f2fs_msg(sb, KERN_ERR,
 334                                 "Unrecognized mount option \"%s\" or missing value",
 335                                 p);
 336                         return -EINVAL;
 337                 }
 338         }
 339         return 0;
 340 }
 341
 342 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 343 {
 344         struct f2fs_inode_info *fi;
 345
 346         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 347         if (!fi)
 348                 return NULL;
 349
 350         init_once((void *) fi);
 351
 352         /* Initialize f2fs-specific inode info */
 353         fi->vfs_inode.i_version = 1;
 354         atomic_set(&fi->dirty_dents, 0);
 355         fi->i_current_depth = 1;
 356         fi->i_advise = 0;
 357         rwlock_init(&fi->ext.ext_lock);
 358
 359         set_inode_flag(fi, FI_NEW_INODE);
 360
 361         if (test_opt(F2FS_SB(sb), INLINE_XATTR))
 362                 set_inode_flag(fi, FI_INLINE_XATTR);
 363
 364         /* Will be used by directory only */
 365         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 366
 367         return &fi->vfs_inode;
 368 }
 369
 370 static int f2fs_drop_inode(struct inode *inode)
 371 {
 372         /*
 373          * This is to avoid a deadlock condition like below.
 374          * writeback_single_inode(inode)
 375          *  - f2fs_write_data_page
 376          *    - f2fs_gc -> iput -> evict
 377          *       - inode_wait_for_writeback(inode)
 378          */
 379         if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
 380                 return 0;
 381         return generic_drop_inode(inode);
 382 }
 383
 384 /*
 385  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 386  *
 387  * We should call set_dirty_inode to write the dirty inode through write_inode.
 388  */
 389 static void f2fs_dirty_inode(struct inode *inode, int flags)
 390 {
 391         set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
 392 }
 393
 394 static void f2fs_i_callback(struct rcu_head *head)
 395 {
 396         struct inode *inode = container_of(head, struct inode, i_rcu);
 397         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 398 }
 399
 400 static void f2fs_destroy_inode(struct inode *inode)
 401 {
 402         call_rcu(&inode->i_rcu, f2fs_i_callback);
 403 }
 404
 405 static void f2fs_put_super(struct super_block *sb)
 406 {
 407         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 408
 409         if (sbi->s_proc) {
 410                 remove_proc_entry("segment_info", sbi->s_proc);
 411                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 412         }
 413         kobject_del(&sbi->s_kobj);
 414
 415         f2fs_destroy_stats(sbi);
 416         stop_gc_thread(sbi);
 417
 418         /* We don't need to do checkpoint when it's clean */
 419         if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
 420                 write_checkpoint(sbi, true);
 421
 422         iput(sbi->node_inode);
 423         iput(sbi->meta_inode);
 424
 425         /* destroy f2fs internal modules */
 426         destroy_node_manager(sbi);
 427         destroy_segment_manager(sbi);
 428
 429         kfree(sbi->ckpt);
 430         kobject_put(&sbi->s_kobj);
 431         wait_for_completion(&sbi->s_kobj_unregister);
 432
 433         sb->s_fs_info = NULL;
 434         brelse(sbi->raw_super_buf);
 435         kfree(sbi);
 436 }
 437
 438 int f2fs_sync_fs(struct super_block *sb, int sync)
 439 {
 440         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 441
 442         trace_f2fs_sync_fs(sb, sync);
 443
 444         if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
 445                 return 0;
 446
 447         if (sync) {
 448                 mutex_lock(&sbi->gc_mutex);
 449                 write_checkpoint(sbi, false);
 450                 mutex_unlock(&sbi->gc_mutex);
 451         } else {
 452                 f2fs_balance_fs(sbi);
 453         }
 454
 455         return 0;
 456 }
 457
 458 static int f2fs_freeze(struct super_block *sb)
 459 {
 460         int err;
 461
 462         if (f2fs_readonly(sb))
 463                 return 0;
 464
 465         err = f2fs_sync_fs(sb, 1);
 466         return err;
 467 }
 468
 469 static int f2fs_unfreeze(struct super_block *sb)
 470 {
 471         return 0;
 472 }
 473
 474 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 475 {
 476         struct super_block *sb = dentry->d_sb;
 477         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 478         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 479         block_t total_count, user_block_count, start_count, ovp_count;
 480
 481         total_count = le64_to_cpu(sbi->raw_super->block_count);
 482         user_block_count = sbi->user_block_count;
 483         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 484         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 485         buf->f_type = F2FS_SUPER_MAGIC;
 486         buf->f_bsize = sbi->blocksize;
 487
 488         buf->f_blocks = total_count - start_count;
 489         buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
 490         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 491
 492         buf->f_files = sbi->total_node_count;
 493         buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
 494
 495         buf->f_namelen = F2FS_NAME_LEN;
 496         buf->f_fsid.val[0] = (u32)id;
 497         buf->f_fsid.val[1] = (u32)(id >> 32);
 498
 499         return 0;
 500 }
 501
 502 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 503 {
 504         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 505
 506         if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
 507                 seq_printf(seq, ",background_gc=%s", "on");
 508         else
 509                 seq_printf(seq, ",background_gc=%s", "off");
 510         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 511                 seq_puts(seq, ",disable_roll_forward");
 512         if (test_opt(sbi, DISCARD))
 513                 seq_puts(seq, ",discard");
 514         if (test_opt(sbi, NOHEAP))
 515                 seq_puts(seq, ",no_heap_alloc");
 516 #ifdef CONFIG_F2FS_FS_XATTR
 517         if (test_opt(sbi, XATTR_USER))
 518                 seq_puts(seq, ",user_xattr");
 519         else
 520                 seq_puts(seq, ",nouser_xattr");
 521         if (test_opt(sbi, INLINE_XATTR))
 522                 seq_puts(seq, ",inline_xattr");
 523 #endif
 524 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 525         if (test_opt(sbi, POSIX_ACL))
 526                 seq_puts(seq, ",acl");
 527         else
 528                 seq_puts(seq, ",noacl");
 529 #endif
 530         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 531                 seq_puts(seq, ",disable_ext_identify");
 532         if (test_opt(sbi, INLINE_DATA))
 533                 seq_puts(seq, ",inline_data");
 534         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 535
 536         return 0;
 537 }
 538
 539 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 540 {
 541         struct super_block *sb = seq->private;
 542         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 543         unsigned int total_segs =
 544                         le32_to_cpu(sbi->raw_super->segment_count_main);
 545         int i;
 546
 547         for (i = 0; i < total_segs; i++) {
 548                 seq_printf(seq, "%u", get_valid_blocks(sbi, i, 1));
 549                 if ((i % 10) == 9 || i == (total_segs - 1))
 550                         seq_putc(seq, '\n');
 551                 else
 552                         seq_putc(seq, ' ');
 553         }
 554
 555         return 0;
 556 }
 557
 558 static int segment_info_open_fs(struct inode *inode, struct file *file)
 559 {
 560         return single_open(file, segment_info_seq_show, PDE_DATA(inode));
 561 }
 562
 563 static const struct file_operations f2fs_seq_segment_info_fops = {
 564         .owner = THIS_MODULE,
 565         .open = segment_info_open_fs,
 566         .read = seq_read,
 567         .llseek = seq_lseek,
 568         .release = single_release,
 569 };
 570
 571 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 572 {
 573         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 574         struct f2fs_mount_info org_mount_opt;
 575         int err, active_logs;
 576
 577         /*
 578          * Save the old mount options in case we
 579          * need to restore them.
 580          */
 581         org_mount_opt = sbi->mount_opt;
 582         active_logs = sbi->active_logs;
 583
 584         /* parse mount options */
 585         err = parse_options(sb, data);
 586         if (err)
 587                 goto restore_opts;
 588
 589         /*
 590          * Previous and new state of filesystem is RO,
 591          * so no point in checking GC conditions.
 592          */
 593         if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
 594                 goto skip;
 595
 596         /*
 597          * We stop the GC thread if FS is mounted as RO
 598          * or if background_gc = off is passed in mount
 599          * option. Also sync the filesystem.
 600          */
 601         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
 602                 if (sbi->gc_thread) {
 603                         stop_gc_thread(sbi);
 604                         f2fs_sync_fs(sb, 1);
 605                 }
 606         } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
 607                 err = start_gc_thread(sbi);
 608                 if (err)
 609                         goto restore_opts;
 610         }
 611 skip:
 612         /* Update the POSIXACL Flag */
 613          sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 614                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 615         return 0;
 616
 617 restore_opts:
 618         sbi->mount_opt = org_mount_opt;
 619         sbi->active_logs = active_logs;
 620         return err;
 621 }
 622
 623 static struct super_operations f2fs_sops = {
 624         .alloc_inode    = f2fs_alloc_inode,
 625         .drop_inode     = f2fs_drop_inode,
 626         .destroy_inode  = f2fs_destroy_inode,
 627         .write_inode    = f2fs_write_inode,
 628         .dirty_inode    = f2fs_dirty_inode,
 629         .show_options   = f2fs_show_options,
 630         .evict_inode    = f2fs_evict_inode,
 631         .put_super      = f2fs_put_super,
 632         .sync_fs        = f2fs_sync_fs,
 633         .freeze_fs      = f2fs_freeze,
 634         .unfreeze_fs    = f2fs_unfreeze,
 635         .statfs         = f2fs_statfs,
 636         .remount_fs     = f2fs_remount,
 637 };
 638
 639 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
 640                 u64 ino, u32 generation)
 641 {
 642         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 643         struct inode *inode;
 644
 645         if (unlikely(ino < F2FS_ROOT_INO(sbi)))
 646                 return ERR_PTR(-ESTALE);
 647         if (unlikely(ino >= NM_I(sbi)->max_nid))
 648                 return ERR_PTR(-ESTALE);
 649
 650         /*
 651          * f2fs_iget isn't quite right if the inode is currently unallocated!
 652          * However f2fs_iget currently does appropriate checks to handle stale
 653          * inodes so everything is OK.
 654          */
 655         inode = f2fs_iget(sb, ino);
 656         if (IS_ERR(inode))
 657                 return ERR_CAST(inode);
 658         if (unlikely(generation && inode->i_generation != generation)) {
 659                 /* we didn't find the right inode.. */
 660                 iput(inode);
 661                 return ERR_PTR(-ESTALE);
 662         }
 663         return inode;
 664 }
 665
 666 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
 667                 int fh_len, int fh_type)
 668 {
 669         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 670                                     f2fs_nfs_get_inode);
 671 }
 672
 673 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
 674                 int fh_len, int fh_type)
 675 {
 676         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 677                                     f2fs_nfs_get_inode);
 678 }
 679
 680 static const struct export_operations f2fs_export_ops = {
 681         .fh_to_dentry = f2fs_fh_to_dentry,
 682         .fh_to_parent = f2fs_fh_to_parent,
 683         .get_parent = f2fs_get_parent,
 684 };
 685
 686 static loff_t max_file_size(unsigned bits)
 687 {
 688         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
 689         loff_t leaf_count = ADDRS_PER_BLOCK;
 690
 691         /* two direct node blocks */
 692         result += (leaf_count * 2);
 693
 694         /* two indirect node blocks */
 695         leaf_count *= NIDS_PER_BLOCK;
 696         result += (leaf_count * 2);
 697
 698         /* one double indirect node block */
 699         leaf_count *= NIDS_PER_BLOCK;
 700         result += leaf_count;
 701
 702         result <<= bits;
 703         return result;
 704 }
 705
 706 static int sanity_check_raw_super(struct super_block *sb,
 707                         struct f2fs_super_block *raw_super)
 708 {
 709         unsigned int blocksize;
 710
 711         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
 712                 f2fs_msg(sb, KERN_INFO,
 713                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
 714                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
 715                 return 1;
 716         }
 717
 718         /* Currently, support only 4KB page cache size */
 719         if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
 720                 f2fs_msg(sb, KERN_INFO,
 721                         "Invalid page_cache_size (%lu), supports only 4KB\n",
 722                         PAGE_CACHE_SIZE);
 723                 return 1;
 724         }
 725
 726         /* Currently, support only 4KB block size */
 727         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
 728         if (blocksize != F2FS_BLKSIZE) {
 729                 f2fs_msg(sb, KERN_INFO,
 730                         "Invalid blocksize (%u), supports only 4KB\n",
 731                         blocksize);
 732                 return 1;
 733         }
 734
 735         if (le32_to_cpu(raw_super->log_sectorsize) !=
 736                                         F2FS_LOG_SECTOR_SIZE) {
 737                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
 738                 return 1;
 739         }
 740         if (le32_to_cpu(raw_super->log_sectors_per_block) !=
 741                                         F2FS_LOG_SECTORS_PER_BLOCK) {
 742                 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
 743                 return 1;
 744         }
 745         return 0;
 746 }
 747
 748 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
 749 {
 750         unsigned int total, fsmeta;
 751         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
 752         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 753
 754         total = le32_to_cpu(raw_super->segment_count);
 755         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
 756         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
 757         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
 758         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
 759         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
 760
 761         if (unlikely(fsmeta >= total))
 762                 return 1;
 763
 764         if (unlikely(is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
 765                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
 766                 return 1;
 767         }
 768         return 0;
 769 }
 770
 771 static void init_sb_info(struct f2fs_sb_info *sbi)
 772 {
 773         struct f2fs_super_block *raw_super = sbi->raw_super;
 774         int i;
 775
 776         sbi->log_sectors_per_block =
 777                 le32_to_cpu(raw_super->log_sectors_per_block);
 778         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
 779         sbi->blocksize = 1 << sbi->log_blocksize;
 780         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
 781         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
 782         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
 783         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
 784         sbi->total_sections = le32_to_cpu(raw_super->section_count);
 785         sbi->total_node_count =
 786                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
 787                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
 788         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
 789         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
 790         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
 791         sbi->cur_victim_sec = NULL_SECNO;
 792         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
 793
 794         for (i = 0; i < NR_COUNT_TYPE; i++)
 795                 atomic_set(&sbi->nr_pages[i], 0);
 796
 797         sbi->dir_level = DEF_DIR_LEVEL;
 798 }
 799
 800 /*
 801  * Read f2fs raw super block.
 802  * Because we have two copies of super block, so read the first one at first,
 803  * if the first one is invalid, move to read the second one.
 804  */
 805 static int read_raw_super_block(struct super_block *sb,
 806                         struct f2fs_super_block **raw_super,
 807                         struct buffer_head **raw_super_buf)
 808 {
 809         int block = 0;
 810
 811 retry:
 812         *raw_super_buf = sb_bread(sb, block);
 813         if (!*raw_super_buf) {
 814                 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
 815                                 block + 1);
 816                 if (block == 0) {
 817                         block++;
 818                         goto retry;
 819                 } else {
 820                         return -EIO;
 821                 }
 822         }
 823
 824         *raw_super = (struct f2fs_super_block *)
 825                 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
 826
 827         /* sanity checking of raw super */
 828         if (sanity_check_raw_super(sb, *raw_super)) {
 829                 brelse(*raw_super_buf);
 830                 f2fs_msg(sb, KERN_ERR,
 831                         "Can't find valid F2FS filesystem in %dth superblock",
 832                                                                 block + 1);
 833                 if (block == 0) {
 834                         block++;
 835                         goto retry;
 836                 } else {
 837                         return -EINVAL;
 838                 }
 839         }
 840
 841         return 0;
 842 }
 843
 844 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 845 {
 846         struct f2fs_sb_info *sbi;
 847         struct f2fs_super_block *raw_super;
 848         struct buffer_head *raw_super_buf;
 849         struct inode *root;
 850         long err = -EINVAL;
 851         int i;
 852
 853         /* allocate memory for f2fs-specific super block info */
 854         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
 855         if (!sbi)
 856                 return -ENOMEM;
 857
 858         /* set a block size */
 859         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
 860                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
 861                 goto free_sbi;
 862         }
 863
 864         err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
 865         if (err)
 866                 goto free_sbi;
 867
 868         sb->s_fs_info = sbi;
 869         /* init some FS parameters */
 870         sbi->active_logs = NR_CURSEG_TYPE;
 871
 872         set_opt(sbi, BG_GC);
 873
 874 #ifdef CONFIG_F2FS_FS_XATTR
 875         set_opt(sbi, XATTR_USER);
 876 #endif
 877 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 878         set_opt(sbi, POSIX_ACL);
 879 #endif
 880         /* parse mount options */
 881         err = parse_options(sb, (char *)data);
 882         if (err)
 883                 goto free_sb_buf;
 884
 885         sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
 886         sb->s_max_links = F2FS_LINK_MAX;
 887         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
 888
 889         sb->s_op = &f2fs_sops;
 890         sb->s_xattr = f2fs_xattr_handlers;
 891         sb->s_export_op = &f2fs_export_ops;
 892         sb->s_magic = F2FS_SUPER_MAGIC;
 893         sb->s_time_gran = 1;
 894         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 895                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 896         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
 897
 898         /* init f2fs-specific super block info */
 899         sbi->sb = sb;
 900         sbi->raw_super = raw_super;
 901         sbi->raw_super_buf = raw_super_buf;
 902         mutex_init(&sbi->gc_mutex);
 903         mutex_init(&sbi->writepages);
 904         mutex_init(&sbi->cp_mutex);
 905         mutex_init(&sbi->node_write);
 906         sbi->por_doing = false;
 907         spin_lock_init(&sbi->stat_lock);
 908
 909         mutex_init(&sbi->read_io.io_mutex);
 910         sbi->read_io.sbi = sbi;
 911         sbi->read_io.bio = NULL;
 912         for (i = 0; i < NR_PAGE_TYPE; i++) {
 913                 mutex_init(&sbi->write_io[i].io_mutex);
 914                 sbi->write_io[i].sbi = sbi;
 915                 sbi->write_io[i].bio = NULL;
 916         }
 917
 918         init_rwsem(&sbi->cp_rwsem);
 919         init_waitqueue_head(&sbi->cp_wait);
 920         init_sb_info(sbi);
 921
 922         /* get an inode for meta space */
 923         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
 924         if (IS_ERR(sbi->meta_inode)) {
 925                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
 926                 err = PTR_ERR(sbi->meta_inode);
 927                 goto free_sb_buf;
 928         }
 929
 930         err = get_valid_checkpoint(sbi);
 931         if (err) {
 932                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
 933                 goto free_meta_inode;
 934         }
 935
 936         /* sanity checking of checkpoint */
 937         err = -EINVAL;
 938         if (sanity_check_ckpt(sbi)) {
 939                 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
 940                 goto free_cp;
 941         }
 942
 943         sbi->total_valid_node_count =
 944                                 le32_to_cpu(sbi->ckpt->valid_node_count);
 945         sbi->total_valid_inode_count =
 946                                 le32_to_cpu(sbi->ckpt->valid_inode_count);
 947         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
 948         sbi->total_valid_block_count =
 949                                 le64_to_cpu(sbi->ckpt->valid_block_count);
 950         sbi->last_valid_block_count = sbi->total_valid_block_count;
 951         sbi->alloc_valid_block_count = 0;
 952         INIT_LIST_HEAD(&sbi->dir_inode_list);
 953         spin_lock_init(&sbi->dir_inode_lock);
 954
 955         init_orphan_info(sbi);
 956
 957         /* setup f2fs internal modules */
 958         err = build_segment_manager(sbi);
 959         if (err) {
 960                 f2fs_msg(sb, KERN_ERR,
 961                         "Failed to initialize F2FS segment manager");
 962                 goto free_sm;
 963         }
 964         err = build_node_manager(sbi);
 965         if (err) {
 966                 f2fs_msg(sb, KERN_ERR,
 967                         "Failed to initialize F2FS node manager");
 968                 goto free_nm;
 969         }
 970
 971         build_gc_manager(sbi);
 972
 973         /* get an inode for node space */
 974         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
 975         if (IS_ERR(sbi->node_inode)) {
 976                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
 977                 err = PTR_ERR(sbi->node_inode);
 978                 goto free_nm;
 979         }
 980
 981         /* if there are nt orphan nodes free them */
 982         recover_orphan_inodes(sbi);
 983
 984         /* read root inode and dentry */
 985         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
 986         if (IS_ERR(root)) {
 987                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
 988                 err = PTR_ERR(root);
 989                 goto free_node_inode;
 990         }
 991         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
 992                 err = -EINVAL;
 993                 goto free_root_inode;
 994         }
 995
 996         sb->s_root = d_make_root(root); /* allocate root dentry */
 997         if (!sb->s_root) {
 998                 err = -ENOMEM;
 999                 goto free_root_inode;
1000         }
1001
1002         err = f2fs_build_stats(sbi);
1003         if (err)
1004                 goto free_root_inode;
1005
1006         if (f2fs_proc_root)
1007                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1008
1009         if (sbi->s_proc)
1010                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1011                                  &f2fs_seq_segment_info_fops, sb);
1012
1013         if (test_opt(sbi, DISCARD)) {
1014                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1015                 if (!blk_queue_discard(q))
1016                         f2fs_msg(sb, KERN_WARNING,
1017                                         "mounting with \"discard\" option, but "
1018                                         "the device does not support discard");
1019         }
1020
1021         sbi->s_kobj.kset = f2fs_kset;
1022         init_completion(&sbi->s_kobj_unregister);
1023         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1024                                                         "%s", sb->s_id);
1025         if (err)
1026                 goto free_proc;
1027
1028         /* recover fsynced data */
1029         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1030                 err = recover_fsync_data(sbi);
1031                 if (err)
1032                         f2fs_msg(sb, KERN_ERR,
1033                                 "Cannot recover all fsync data errno=%ld", err);
1034         }
1035
1036         /*
1037          * If filesystem is not mounted as read-only then
1038          * do start the gc_thread.
1039          */
1040         if (!(sb->s_flags & MS_RDONLY)) {
1041                 /* After POR, we can run background GC thread.*/
1042                 err = start_gc_thread(sbi);
1043                 if (err)
1044                         goto free_kobj;
1045         }
1046         return 0;
1047
1048 free_kobj:
1049         kobject_del(&sbi->s_kobj);
1050 free_proc:
1051         if (sbi->s_proc) {
1052                 remove_proc_entry("segment_info", sbi->s_proc);
1053                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1054         }
1055         f2fs_destroy_stats(sbi);
1056 free_root_inode:
1057         dput(sb->s_root);
1058         sb->s_root = NULL;
1059 free_node_inode:
1060         iput(sbi->node_inode);
1061 free_nm:
1062         destroy_node_manager(sbi);
1063 free_sm:
1064         destroy_segment_manager(sbi);
1065 free_cp:
1066         kfree(sbi->ckpt);
1067 free_meta_inode:
1068         make_bad_inode(sbi->meta_inode);
1069         iput(sbi->meta_inode);
1070 free_sb_buf:
1071         brelse(raw_super_buf);
1072 free_sbi:
1073         kfree(sbi);
1074         return err;
1075 }
1076
1077 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1078                         const char *dev_name, void *data)
1079 {
1080         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1081 }
1082
1083 static struct file_system_type f2fs_fs_type = {
1084         .owner          = THIS_MODULE,
1085         .name           = "f2fs",
1086         .mount          = f2fs_mount,
1087         .kill_sb        = kill_block_super,
1088         .fs_flags       = FS_REQUIRES_DEV,
1089 };
1090 MODULE_ALIAS_FS("f2fs");
1091
1092 static int __init init_inodecache(void)
1093 {
1094         f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1095                         sizeof(struct f2fs_inode_info));
1096         if (!f2fs_inode_cachep)
1097                 return -ENOMEM;
1098         return 0;
1099 }
1100
1101 static void destroy_inodecache(void)
1102 {
1103         /*
1104          * Make sure all delayed rcu free inodes are flushed before we
1105          * destroy cache.
1106          */
1107         rcu_barrier();
1108         kmem_cache_destroy(f2fs_inode_cachep);
1109 }
1110
1111 static int __init init_f2fs_fs(void)
1112 {
1113         int err;
1114
1115         err = init_inodecache();
1116         if (err)
1117                 goto fail;
1118         err = create_node_manager_caches();
1119         if (err)
1120                 goto free_inodecache;
1121         err = create_segment_manager_caches();
1122         if (err)
1123                 goto free_node_manager_caches;
1124         err = create_gc_caches();
1125         if (err)
1126                 goto free_segment_manager_caches;
1127         err = create_checkpoint_caches();
1128         if (err)
1129                 goto free_gc_caches;
1130         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1131         if (!f2fs_kset) {
1132                 err = -ENOMEM;
1133                 goto free_checkpoint_caches;
1134         }
1135         err = register_filesystem(&f2fs_fs_type);
1136         if (err)
1137                 goto free_kset;
1138         f2fs_create_root_stats();
1139         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1140         return 0;
1141
1142 free_kset:
1143         kset_unregister(f2fs_kset);
1144 free_checkpoint_caches:
1145         destroy_checkpoint_caches();
1146 free_gc_caches:
1147         destroy_gc_caches();
1148 free_segment_manager_caches:
1149         destroy_segment_manager_caches();
1150 free_node_manager_caches:
1151         destroy_node_manager_caches();
1152 free_inodecache:
1153         destroy_inodecache();
1154 fail:
1155         return err;
1156 }
1157
1158 static void __exit exit_f2fs_fs(void)
1159 {
1160         remove_proc_entry("fs/f2fs", NULL);
1161         f2fs_destroy_root_stats();
1162         unregister_filesystem(&f2fs_fs_type);
1163         destroy_checkpoint_caches();
1164         destroy_gc_caches();
1165         destroy_segment_manager_caches();
1166         destroy_node_manager_caches();
1167         destroy_inodecache();
1168         kset_unregister(f2fs_kset);
1169 }
1170
1171 module_init(init_f2fs_fs)
1172 module_exit(exit_f2fs_fs)
1173
1174 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1175 MODULE_DESCRIPTION("Flash Friendly File System");
1176 MODULE_LICENSE("GPL");