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
 648         /*
 649          * f2fs_iget isn't quite right if the inode is currently unallocated!
 650          * However f2fs_iget currently does appropriate checks to handle stale
 651          * inodes so everything is OK.
 652          */
 653         inode = f2fs_iget(sb, ino);
 654         if (IS_ERR(inode))
 655                 return ERR_CAST(inode);
 656         if (unlikely(generation && inode->i_generation != generation)) {
 657                 /* we didn't find the right inode.. */
 658                 iput(inode);
 659                 return ERR_PTR(-ESTALE);
 660         }
 661         return inode;
 662 }
 663
 664 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
 665                 int fh_len, int fh_type)
 666 {
 667         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 668                                     f2fs_nfs_get_inode);
 669 }
 670
 671 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
 672                 int fh_len, int fh_type)
 673 {
 674         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 675                                     f2fs_nfs_get_inode);
 676 }
 677
 678 static const struct export_operations f2fs_export_ops = {
 679         .fh_to_dentry = f2fs_fh_to_dentry,
 680         .fh_to_parent = f2fs_fh_to_parent,
 681         .get_parent = f2fs_get_parent,
 682 };
 683
 684 static loff_t max_file_size(unsigned bits)
 685 {
 686         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
 687         loff_t leaf_count = ADDRS_PER_BLOCK;
 688
 689         /* two direct node blocks */
 690         result += (leaf_count * 2);
 691
 692         /* two indirect node blocks */
 693         leaf_count *= NIDS_PER_BLOCK;
 694         result += (leaf_count * 2);
 695
 696         /* one double indirect node block */
 697         leaf_count *= NIDS_PER_BLOCK;
 698         result += leaf_count;
 699
 700         result <<= bits;
 701         return result;
 702 }
 703
 704 static int sanity_check_raw_super(struct super_block *sb,
 705                         struct f2fs_super_block *raw_super)
 706 {
 707         unsigned int blocksize;
 708
 709         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
 710                 f2fs_msg(sb, KERN_INFO,
 711                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
 712                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
 713                 return 1;
 714         }
 715
 716         /* Currently, support only 4KB page cache size */
 717         if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
 718                 f2fs_msg(sb, KERN_INFO,
 719                         "Invalid page_cache_size (%lu), supports only 4KB\n",
 720                         PAGE_CACHE_SIZE);
 721                 return 1;
 722         }
 723
 724         /* Currently, support only 4KB block size */
 725         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
 726         if (blocksize != F2FS_BLKSIZE) {
 727                 f2fs_msg(sb, KERN_INFO,
 728                         "Invalid blocksize (%u), supports only 4KB\n",
 729                         blocksize);
 730                 return 1;
 731         }
 732
 733         if (le32_to_cpu(raw_super->log_sectorsize) !=
 734                                         F2FS_LOG_SECTOR_SIZE) {
 735                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
 736                 return 1;
 737         }
 738         if (le32_to_cpu(raw_super->log_sectors_per_block) !=
 739                                         F2FS_LOG_SECTORS_PER_BLOCK) {
 740                 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
 741                 return 1;
 742         }
 743         return 0;
 744 }
 745
 746 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
 747 {
 748         unsigned int total, fsmeta;
 749         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
 750         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 751
 752         total = le32_to_cpu(raw_super->segment_count);
 753         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
 754         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
 755         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
 756         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
 757         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
 758
 759         if (unlikely(fsmeta >= total))
 760                 return 1;
 761
 762         if (unlikely(is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
 763                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
 764                 return 1;
 765         }
 766         return 0;
 767 }
 768
 769 static void init_sb_info(struct f2fs_sb_info *sbi)
 770 {
 771         struct f2fs_super_block *raw_super = sbi->raw_super;
 772         int i;
 773
 774         sbi->log_sectors_per_block =
 775                 le32_to_cpu(raw_super->log_sectors_per_block);
 776         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
 777         sbi->blocksize = 1 << sbi->log_blocksize;
 778         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
 779         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
 780         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
 781         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
 782         sbi->total_sections = le32_to_cpu(raw_super->section_count);
 783         sbi->total_node_count =
 784                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
 785                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
 786         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
 787         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
 788         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
 789         sbi->cur_victim_sec = NULL_SECNO;
 790         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
 791
 792         for (i = 0; i < NR_COUNT_TYPE; i++)
 793                 atomic_set(&sbi->nr_pages[i], 0);
 794
 795         sbi->dir_level = DEF_DIR_LEVEL;
 796 }
 797
 798 /*
 799  * Read f2fs raw super block.
 800  * Because we have two copies of super block, so read the first one at first,
 801  * if the first one is invalid, move to read the second one.
 802  */
 803 static int read_raw_super_block(struct super_block *sb,
 804                         struct f2fs_super_block **raw_super,
 805                         struct buffer_head **raw_super_buf)
 806 {
 807         int block = 0;
 808
 809 retry:
 810         *raw_super_buf = sb_bread(sb, block);
 811         if (!*raw_super_buf) {
 812                 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
 813                                 block + 1);
 814                 if (block == 0) {
 815                         block++;
 816                         goto retry;
 817                 } else {
 818                         return -EIO;
 819                 }
 820         }
 821
 822         *raw_super = (struct f2fs_super_block *)
 823                 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
 824
 825         /* sanity checking of raw super */
 826         if (sanity_check_raw_super(sb, *raw_super)) {
 827                 brelse(*raw_super_buf);
 828                 f2fs_msg(sb, KERN_ERR,
 829                         "Can't find valid F2FS filesystem in %dth superblock",
 830                                                                 block + 1);
 831                 if (block == 0) {
 832                         block++;
 833                         goto retry;
 834                 } else {
 835                         return -EINVAL;
 836                 }
 837         }
 838
 839         return 0;
 840 }
 841
 842 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 843 {
 844         struct f2fs_sb_info *sbi;
 845         struct f2fs_super_block *raw_super;
 846         struct buffer_head *raw_super_buf;
 847         struct inode *root;
 848         long err = -EINVAL;
 849         int i;
 850
 851         /* allocate memory for f2fs-specific super block info */
 852         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
 853         if (!sbi)
 854                 return -ENOMEM;
 855
 856         /* set a block size */
 857         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
 858                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
 859                 goto free_sbi;
 860         }
 861
 862         err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
 863         if (err)
 864                 goto free_sbi;
 865
 866         sb->s_fs_info = sbi;
 867         /* init some FS parameters */
 868         sbi->active_logs = NR_CURSEG_TYPE;
 869
 870         set_opt(sbi, BG_GC);
 871
 872 #ifdef CONFIG_F2FS_FS_XATTR
 873         set_opt(sbi, XATTR_USER);
 874 #endif
 875 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 876         set_opt(sbi, POSIX_ACL);
 877 #endif
 878         /* parse mount options */
 879         err = parse_options(sb, (char *)data);
 880         if (err)
 881                 goto free_sb_buf;
 882
 883         sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
 884         sb->s_max_links = F2FS_LINK_MAX;
 885         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
 886
 887         sb->s_op = &f2fs_sops;
 888         sb->s_xattr = f2fs_xattr_handlers;
 889         sb->s_export_op = &f2fs_export_ops;
 890         sb->s_magic = F2FS_SUPER_MAGIC;
 891         sb->s_time_gran = 1;
 892         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 893                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 894         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
 895
 896         /* init f2fs-specific super block info */
 897         sbi->sb = sb;
 898         sbi->raw_super = raw_super;
 899         sbi->raw_super_buf = raw_super_buf;
 900         mutex_init(&sbi->gc_mutex);
 901         mutex_init(&sbi->writepages);
 902         mutex_init(&sbi->cp_mutex);
 903         mutex_init(&sbi->node_write);
 904         sbi->por_doing = false;
 905         spin_lock_init(&sbi->stat_lock);
 906
 907         mutex_init(&sbi->read_io.io_mutex);
 908         sbi->read_io.sbi = sbi;
 909         sbi->read_io.bio = NULL;
 910         for (i = 0; i < NR_PAGE_TYPE; i++) {
 911                 mutex_init(&sbi->write_io[i].io_mutex);
 912                 sbi->write_io[i].sbi = sbi;
 913                 sbi->write_io[i].bio = NULL;
 914         }
 915
 916         init_rwsem(&sbi->cp_rwsem);
 917         init_waitqueue_head(&sbi->cp_wait);
 918         init_sb_info(sbi);
 919
 920         /* get an inode for meta space */
 921         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
 922         if (IS_ERR(sbi->meta_inode)) {
 923                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
 924                 err = PTR_ERR(sbi->meta_inode);
 925                 goto free_sb_buf;
 926         }
 927
 928         err = get_valid_checkpoint(sbi);
 929         if (err) {
 930                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
 931                 goto free_meta_inode;
 932         }
 933
 934         /* sanity checking of checkpoint */
 935         err = -EINVAL;
 936         if (sanity_check_ckpt(sbi)) {
 937                 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
 938                 goto free_cp;
 939         }
 940
 941         sbi->total_valid_node_count =
 942                                 le32_to_cpu(sbi->ckpt->valid_node_count);
 943         sbi->total_valid_inode_count =
 944                                 le32_to_cpu(sbi->ckpt->valid_inode_count);
 945         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
 946         sbi->total_valid_block_count =
 947                                 le64_to_cpu(sbi->ckpt->valid_block_count);
 948         sbi->last_valid_block_count = sbi->total_valid_block_count;
 949         sbi->alloc_valid_block_count = 0;
 950         INIT_LIST_HEAD(&sbi->dir_inode_list);
 951         spin_lock_init(&sbi->dir_inode_lock);
 952
 953         init_orphan_info(sbi);
 954
 955         /* setup f2fs internal modules */
 956         err = build_segment_manager(sbi);
 957         if (err) {
 958                 f2fs_msg(sb, KERN_ERR,
 959                         "Failed to initialize F2FS segment manager");
 960                 goto free_sm;
 961         }
 962         err = build_node_manager(sbi);
 963         if (err) {
 964                 f2fs_msg(sb, KERN_ERR,
 965                         "Failed to initialize F2FS node manager");
 966                 goto free_nm;
 967         }
 968
 969         build_gc_manager(sbi);
 970
 971         /* get an inode for node space */
 972         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
 973         if (IS_ERR(sbi->node_inode)) {
 974                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
 975                 err = PTR_ERR(sbi->node_inode);
 976                 goto free_nm;
 977         }
 978
 979         /* if there are nt orphan nodes free them */
 980         recover_orphan_inodes(sbi);
 981
 982         /* read root inode and dentry */
 983         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
 984         if (IS_ERR(root)) {
 985                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
 986                 err = PTR_ERR(root);
 987                 goto free_node_inode;
 988         }
 989         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
 990                 err = -EINVAL;
 991                 goto free_root_inode;
 992         }
 993
 994         sb->s_root = d_make_root(root); /* allocate root dentry */
 995         if (!sb->s_root) {
 996                 err = -ENOMEM;
 997                 goto free_root_inode;
 998         }
 999
1000         err = f2fs_build_stats(sbi);
1001         if (err)
1002                 goto free_root_inode;
1003
1004         if (f2fs_proc_root)
1005                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1006
1007         if (sbi->s_proc)
1008                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1009                                  &f2fs_seq_segment_info_fops, sb);
1010
1011         if (test_opt(sbi, DISCARD)) {
1012                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1013                 if (!blk_queue_discard(q))
1014                         f2fs_msg(sb, KERN_WARNING,
1015                                         "mounting with \"discard\" option, but "
1016                                         "the device does not support discard");
1017         }
1018
1019         sbi->s_kobj.kset = f2fs_kset;
1020         init_completion(&sbi->s_kobj_unregister);
1021         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1022                                                         "%s", sb->s_id);
1023         if (err)
1024                 goto free_proc;
1025
1026         /* recover fsynced data */
1027         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1028                 err = recover_fsync_data(sbi);
1029                 if (err)
1030                         f2fs_msg(sb, KERN_ERR,
1031                                 "Cannot recover all fsync data errno=%ld", err);
1032         }
1033
1034         /*
1035          * If filesystem is not mounted as read-only then
1036          * do start the gc_thread.
1037          */
1038         if (!(sb->s_flags & MS_RDONLY)) {
1039                 /* After POR, we can run background GC thread.*/
1040                 err = start_gc_thread(sbi);
1041                 if (err)
1042                         goto free_kobj;
1043         }
1044         return 0;
1045
1046 free_kobj:
1047         kobject_del(&sbi->s_kobj);
1048 free_proc:
1049         if (sbi->s_proc) {
1050                 remove_proc_entry("segment_info", sbi->s_proc);
1051                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1052         }
1053         f2fs_destroy_stats(sbi);
1054 free_root_inode:
1055         dput(sb->s_root);
1056         sb->s_root = NULL;
1057 free_node_inode:
1058         iput(sbi->node_inode);
1059 free_nm:
1060         destroy_node_manager(sbi);
1061 free_sm:
1062         destroy_segment_manager(sbi);
1063 free_cp:
1064         kfree(sbi->ckpt);
1065 free_meta_inode:
1066         make_bad_inode(sbi->meta_inode);
1067         iput(sbi->meta_inode);
1068 free_sb_buf:
1069         brelse(raw_super_buf);
1070 free_sbi:
1071         kfree(sbi);
1072         return err;
1073 }
1074
1075 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1076                         const char *dev_name, void *data)
1077 {
1078         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1079 }
1080
1081 static struct file_system_type f2fs_fs_type = {
1082         .owner          = THIS_MODULE,
1083         .name           = "f2fs",
1084         .mount          = f2fs_mount,
1085         .kill_sb        = kill_block_super,
1086         .fs_flags       = FS_REQUIRES_DEV,
1087 };
1088 MODULE_ALIAS_FS("f2fs");
1089
1090 static int __init init_inodecache(void)
1091 {
1092         f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1093                         sizeof(struct f2fs_inode_info));
1094         if (!f2fs_inode_cachep)
1095                 return -ENOMEM;
1096         return 0;
1097 }
1098
1099 static void destroy_inodecache(void)
1100 {
1101         /*
1102          * Make sure all delayed rcu free inodes are flushed before we
1103          * destroy cache.
1104          */
1105         rcu_barrier();
1106         kmem_cache_destroy(f2fs_inode_cachep);
1107 }
1108
1109 static int __init init_f2fs_fs(void)
1110 {
1111         int err;
1112
1113         err = init_inodecache();
1114         if (err)
1115                 goto fail;
1116         err = create_node_manager_caches();
1117         if (err)
1118                 goto free_inodecache;
1119         err = create_segment_manager_caches();
1120         if (err)
1121                 goto free_node_manager_caches;
1122         err = create_gc_caches();
1123         if (err)
1124                 goto free_segment_manager_caches;
1125         err = create_checkpoint_caches();
1126         if (err)
1127                 goto free_gc_caches;
1128         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1129         if (!f2fs_kset) {
1130                 err = -ENOMEM;
1131                 goto free_checkpoint_caches;
1132         }
1133         err = register_filesystem(&f2fs_fs_type);
1134         if (err)
1135                 goto free_kset;
1136         f2fs_create_root_stats();
1137         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1138         return 0;
1139
1140 free_kset:
1141         kset_unregister(f2fs_kset);
1142 free_checkpoint_caches:
1143         destroy_checkpoint_caches();
1144 free_gc_caches:
1145         destroy_gc_caches();
1146 free_segment_manager_caches:
1147         destroy_segment_manager_caches();
1148 free_node_manager_caches:
1149         destroy_node_manager_caches();
1150 free_inodecache:
1151         destroy_inodecache();
1152 fail:
1153         return err;
1154 }
1155
1156 static void __exit exit_f2fs_fs(void)
1157 {
1158         remove_proc_entry("fs/f2fs", NULL);
1159         f2fs_destroy_root_stats();
1160         unregister_filesystem(&f2fs_fs_type);
1161         destroy_checkpoint_caches();
1162         destroy_gc_caches();
1163         destroy_segment_manager_caches();
1164         destroy_node_manager_caches();
1165         destroy_inodecache();
1166         kset_unregister(f2fs_kset);
1167 }
1168
1169 module_init(init_f2fs_fs)
1170 module_exit(exit_f2fs_fs)
1171
1172 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1173 MODULE_DESCRIPTION("Flash Friendly File System");
1174 MODULE_LICENSE("GPL");