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 != 0 && (i % 10) == 0)
 550                         seq_puts(seq, "\n");
 551                 else
 552                         seq_puts(seq, " ");
 553         }
 554         return 0;
 555 }
 556
 557 static int segment_info_open_fs(struct inode *inode, struct file *file)
 558 {
 559         return single_open(file, segment_info_seq_show, PDE_DATA(inode));
 560 }
 561
 562 static const struct file_operations f2fs_seq_segment_info_fops = {
 563         .owner = THIS_MODULE,
 564         .open = segment_info_open_fs,
 565         .read = seq_read,
 566         .llseek = seq_lseek,
 567         .release = single_release,
 568 };
 569
 570 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 571 {
 572         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 573         struct f2fs_mount_info org_mount_opt;
 574         int err, active_logs;
 575
 576         /*
 577          * Save the old mount options in case we
 578          * need to restore them.
 579          */
 580         org_mount_opt = sbi->mount_opt;
 581         active_logs = sbi->active_logs;
 582
 583         /* parse mount options */
 584         err = parse_options(sb, data);
 585         if (err)
 586                 goto restore_opts;
 587
 588         /*
 589          * Previous and new state of filesystem is RO,
 590          * so no point in checking GC conditions.
 591          */
 592         if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
 593                 goto skip;
 594
 595         /*
 596          * We stop the GC thread if FS is mounted as RO
 597          * or if background_gc = off is passed in mount
 598          * option. Also sync the filesystem.
 599          */
 600         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
 601                 if (sbi->gc_thread) {
 602                         stop_gc_thread(sbi);
 603                         f2fs_sync_fs(sb, 1);
 604                 }
 605         } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
 606                 err = start_gc_thread(sbi);
 607                 if (err)
 608                         goto restore_opts;
 609         }
 610 skip:
 611         /* Update the POSIXACL Flag */
 612          sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 613                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 614         return 0;
 615
 616 restore_opts:
 617         sbi->mount_opt = org_mount_opt;
 618         sbi->active_logs = active_logs;
 619         return err;
 620 }
 621
 622 static struct super_operations f2fs_sops = {
 623         .alloc_inode    = f2fs_alloc_inode,
 624         .drop_inode     = f2fs_drop_inode,
 625         .destroy_inode  = f2fs_destroy_inode,
 626         .write_inode    = f2fs_write_inode,
 627         .dirty_inode    = f2fs_dirty_inode,
 628         .show_options   = f2fs_show_options,
 629         .evict_inode    = f2fs_evict_inode,
 630         .put_super      = f2fs_put_super,
 631         .sync_fs        = f2fs_sync_fs,
 632         .freeze_fs      = f2fs_freeze,
 633         .unfreeze_fs    = f2fs_unfreeze,
 634         .statfs         = f2fs_statfs,
 635         .remount_fs     = f2fs_remount,
 636 };
 637
 638 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
 639                 u64 ino, u32 generation)
 640 {
 641         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 642         struct inode *inode;
 643
 644         if (unlikely(ino < F2FS_ROOT_INO(sbi)))
 645                 return ERR_PTR(-ESTALE);
 646
 647         /*
 648          * f2fs_iget isn't quite right if the inode is currently unallocated!
 649          * However f2fs_iget currently does appropriate checks to handle stale
 650          * inodes so everything is OK.
 651          */
 652         inode = f2fs_iget(sb, ino);
 653         if (IS_ERR(inode))
 654                 return ERR_CAST(inode);
 655         if (unlikely(generation && inode->i_generation != generation)) {
 656                 /* we didn't find the right inode.. */
 657                 iput(inode);
 658                 return ERR_PTR(-ESTALE);
 659         }
 660         return inode;
 661 }
 662
 663 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
 664                 int fh_len, int fh_type)
 665 {
 666         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 667                                     f2fs_nfs_get_inode);
 668 }
 669
 670 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
 671                 int fh_len, int fh_type)
 672 {
 673         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 674                                     f2fs_nfs_get_inode);
 675 }
 676
 677 static const struct export_operations f2fs_export_ops = {
 678         .fh_to_dentry = f2fs_fh_to_dentry,
 679         .fh_to_parent = f2fs_fh_to_parent,
 680         .get_parent = f2fs_get_parent,
 681 };
 682
 683 static loff_t max_file_size(unsigned bits)
 684 {
 685         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
 686         loff_t leaf_count = ADDRS_PER_BLOCK;
 687
 688         /* two direct node blocks */
 689         result += (leaf_count * 2);
 690
 691         /* two indirect node blocks */
 692         leaf_count *= NIDS_PER_BLOCK;
 693         result += (leaf_count * 2);
 694
 695         /* one double indirect node block */
 696         leaf_count *= NIDS_PER_BLOCK;
 697         result += leaf_count;
 698
 699         result <<= bits;
 700         return result;
 701 }
 702
 703 static int sanity_check_raw_super(struct super_block *sb,
 704                         struct f2fs_super_block *raw_super)
 705 {
 706         unsigned int blocksize;
 707
 708         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
 709                 f2fs_msg(sb, KERN_INFO,
 710                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
 711                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
 712                 return 1;
 713         }
 714
 715         /* Currently, support only 4KB page cache size */
 716         if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
 717                 f2fs_msg(sb, KERN_INFO,
 718                         "Invalid page_cache_size (%lu), supports only 4KB\n",
 719                         PAGE_CACHE_SIZE);
 720                 return 1;
 721         }
 722
 723         /* Currently, support only 4KB block size */
 724         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
 725         if (blocksize != F2FS_BLKSIZE) {
 726                 f2fs_msg(sb, KERN_INFO,
 727                         "Invalid blocksize (%u), supports only 4KB\n",
 728                         blocksize);
 729                 return 1;
 730         }
 731
 732         if (le32_to_cpu(raw_super->log_sectorsize) !=
 733                                         F2FS_LOG_SECTOR_SIZE) {
 734                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
 735                 return 1;
 736         }
 737         if (le32_to_cpu(raw_super->log_sectors_per_block) !=
 738                                         F2FS_LOG_SECTORS_PER_BLOCK) {
 739                 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
 740                 return 1;
 741         }
 742         return 0;
 743 }
 744
 745 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
 746 {
 747         unsigned int total, fsmeta;
 748         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
 749         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 750
 751         total = le32_to_cpu(raw_super->segment_count);
 752         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
 753         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
 754         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
 755         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
 756         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
 757
 758         if (unlikely(fsmeta >= total))
 759                 return 1;
 760
 761         if (unlikely(is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
 762                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
 763                 return 1;
 764         }
 765         return 0;
 766 }
 767
 768 static void init_sb_info(struct f2fs_sb_info *sbi)
 769 {
 770         struct f2fs_super_block *raw_super = sbi->raw_super;
 771         int i;
 772
 773         sbi->log_sectors_per_block =
 774                 le32_to_cpu(raw_super->log_sectors_per_block);
 775         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
 776         sbi->blocksize = 1 << sbi->log_blocksize;
 777         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
 778         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
 779         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
 780         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
 781         sbi->total_sections = le32_to_cpu(raw_super->section_count);
 782         sbi->total_node_count =
 783                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
 784                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
 785         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
 786         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
 787         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
 788         sbi->cur_victim_sec = NULL_SECNO;
 789         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
 790
 791         for (i = 0; i < NR_COUNT_TYPE; i++)
 792                 atomic_set(&sbi->nr_pages[i], 0);
 793
 794         sbi->dir_level = DEF_DIR_LEVEL;
 795 }
 796
 797 /*
 798  * Read f2fs raw super block.
 799  * Because we have two copies of super block, so read the first one at first,
 800  * if the first one is invalid, move to read the second one.
 801  */
 802 static int read_raw_super_block(struct super_block *sb,
 803                         struct f2fs_super_block **raw_super,
 804                         struct buffer_head **raw_super_buf)
 805 {
 806         int block = 0;
 807
 808 retry:
 809         *raw_super_buf = sb_bread(sb, block);
 810         if (!*raw_super_buf) {
 811                 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
 812                                 block + 1);
 813                 if (block == 0) {
 814                         block++;
 815                         goto retry;
 816                 } else {
 817                         return -EIO;
 818                 }
 819         }
 820
 821         *raw_super = (struct f2fs_super_block *)
 822                 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
 823
 824         /* sanity checking of raw super */
 825         if (sanity_check_raw_super(sb, *raw_super)) {
 826                 brelse(*raw_super_buf);
 827                 f2fs_msg(sb, KERN_ERR,
 828                         "Can't find valid F2FS filesystem in %dth superblock",
 829                                                                 block + 1);
 830                 if (block == 0) {
 831                         block++;
 832                         goto retry;
 833                 } else {
 834                         return -EINVAL;
 835                 }
 836         }
 837
 838         return 0;
 839 }
 840
 841 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 842 {
 843         struct f2fs_sb_info *sbi;
 844         struct f2fs_super_block *raw_super;
 845         struct buffer_head *raw_super_buf;
 846         struct inode *root;
 847         long err = -EINVAL;
 848         int i;
 849
 850         /* allocate memory for f2fs-specific super block info */
 851         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
 852         if (!sbi)
 853                 return -ENOMEM;
 854
 855         /* set a block size */
 856         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
 857                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
 858                 goto free_sbi;
 859         }
 860
 861         err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
 862         if (err)
 863                 goto free_sbi;
 864
 865         sb->s_fs_info = sbi;
 866         /* init some FS parameters */
 867         sbi->active_logs = NR_CURSEG_TYPE;
 868
 869         set_opt(sbi, BG_GC);
 870
 871 #ifdef CONFIG_F2FS_FS_XATTR
 872         set_opt(sbi, XATTR_USER);
 873 #endif
 874 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 875         set_opt(sbi, POSIX_ACL);
 876 #endif
 877         /* parse mount options */
 878         err = parse_options(sb, (char *)data);
 879         if (err)
 880                 goto free_sb_buf;
 881
 882         sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
 883         sb->s_max_links = F2FS_LINK_MAX;
 884         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
 885
 886         sb->s_op = &f2fs_sops;
 887         sb->s_xattr = f2fs_xattr_handlers;
 888         sb->s_export_op = &f2fs_export_ops;
 889         sb->s_magic = F2FS_SUPER_MAGIC;
 890         sb->s_time_gran = 1;
 891         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 892                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 893         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
 894
 895         /* init f2fs-specific super block info */
 896         sbi->sb = sb;
 897         sbi->raw_super = raw_super;
 898         sbi->raw_super_buf = raw_super_buf;
 899         mutex_init(&sbi->gc_mutex);
 900         mutex_init(&sbi->writepages);
 901         mutex_init(&sbi->cp_mutex);
 902         mutex_init(&sbi->node_write);
 903         sbi->por_doing = false;
 904         spin_lock_init(&sbi->stat_lock);
 905
 906         mutex_init(&sbi->read_io.io_mutex);
 907         sbi->read_io.sbi = sbi;
 908         sbi->read_io.bio = NULL;
 909         for (i = 0; i < NR_PAGE_TYPE; i++) {
 910                 mutex_init(&sbi->write_io[i].io_mutex);
 911                 sbi->write_io[i].sbi = sbi;
 912                 sbi->write_io[i].bio = NULL;
 913         }
 914
 915         init_rwsem(&sbi->cp_rwsem);
 916         init_waitqueue_head(&sbi->cp_wait);
 917         init_sb_info(sbi);
 918
 919         /* get an inode for meta space */
 920         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
 921         if (IS_ERR(sbi->meta_inode)) {
 922                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
 923                 err = PTR_ERR(sbi->meta_inode);
 924                 goto free_sb_buf;
 925         }
 926
 927         err = get_valid_checkpoint(sbi);
 928         if (err) {
 929                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
 930                 goto free_meta_inode;
 931         }
 932
 933         /* sanity checking of checkpoint */
 934         err = -EINVAL;
 935         if (sanity_check_ckpt(sbi)) {
 936                 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
 937                 goto free_cp;
 938         }
 939
 940         sbi->total_valid_node_count =
 941                                 le32_to_cpu(sbi->ckpt->valid_node_count);
 942         sbi->total_valid_inode_count =
 943                                 le32_to_cpu(sbi->ckpt->valid_inode_count);
 944         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
 945         sbi->total_valid_block_count =
 946                                 le64_to_cpu(sbi->ckpt->valid_block_count);
 947         sbi->last_valid_block_count = sbi->total_valid_block_count;
 948         sbi->alloc_valid_block_count = 0;
 949         INIT_LIST_HEAD(&sbi->dir_inode_list);
 950         spin_lock_init(&sbi->dir_inode_lock);
 951
 952         init_orphan_info(sbi);
 953
 954         /* setup f2fs internal modules */
 955         err = build_segment_manager(sbi);
 956         if (err) {
 957                 f2fs_msg(sb, KERN_ERR,
 958                         "Failed to initialize F2FS segment manager");
 959                 goto free_sm;
 960         }
 961         err = build_node_manager(sbi);
 962         if (err) {
 963                 f2fs_msg(sb, KERN_ERR,
 964                         "Failed to initialize F2FS node manager");
 965                 goto free_nm;
 966         }
 967
 968         build_gc_manager(sbi);
 969
 970         /* get an inode for node space */
 971         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
 972         if (IS_ERR(sbi->node_inode)) {
 973                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
 974                 err = PTR_ERR(sbi->node_inode);
 975                 goto free_nm;
 976         }
 977
 978         /* if there are nt orphan nodes free them */
 979         recover_orphan_inodes(sbi);
 980
 981         /* read root inode and dentry */
 982         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
 983         if (IS_ERR(root)) {
 984                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
 985                 err = PTR_ERR(root);
 986                 goto free_node_inode;
 987         }
 988         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
 989                 err = -EINVAL;
 990                 goto free_root_inode;
 991         }
 992
 993         sb->s_root = d_make_root(root); /* allocate root dentry */
 994         if (!sb->s_root) {
 995                 err = -ENOMEM;
 996                 goto free_root_inode;
 997         }
 998
 999         err = f2fs_build_stats(sbi);
1000         if (err)
1001                 goto free_root_inode;
1002
1003         if (f2fs_proc_root)
1004                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1005
1006         if (sbi->s_proc)
1007                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1008                                  &f2fs_seq_segment_info_fops, sb);
1009
1010         if (test_opt(sbi, DISCARD)) {
1011                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1012                 if (!blk_queue_discard(q))
1013                         f2fs_msg(sb, KERN_WARNING,
1014                                         "mounting with \"discard\" option, but "
1015                                         "the device does not support discard");
1016         }
1017
1018         sbi->s_kobj.kset = f2fs_kset;
1019         init_completion(&sbi->s_kobj_unregister);
1020         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1021                                                         "%s", sb->s_id);
1022         if (err)
1023                 goto free_proc;
1024
1025         /* recover fsynced data */
1026         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1027                 err = recover_fsync_data(sbi);
1028                 if (err)
1029                         f2fs_msg(sb, KERN_ERR,
1030                                 "Cannot recover all fsync data errno=%ld", err);
1031         }
1032
1033         /*
1034          * If filesystem is not mounted as read-only then
1035          * do start the gc_thread.
1036          */
1037         if (!(sb->s_flags & MS_RDONLY)) {
1038                 /* After POR, we can run background GC thread.*/
1039                 err = start_gc_thread(sbi);
1040                 if (err)
1041                         goto free_kobj;
1042         }
1043         return 0;
1044
1045 free_kobj:
1046         kobject_del(&sbi->s_kobj);
1047 free_proc:
1048         if (sbi->s_proc) {
1049                 remove_proc_entry("segment_info", sbi->s_proc);
1050                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1051         }
1052         f2fs_destroy_stats(sbi);
1053 free_root_inode:
1054         dput(sb->s_root);
1055         sb->s_root = NULL;
1056 free_node_inode:
1057         iput(sbi->node_inode);
1058 free_nm:
1059         destroy_node_manager(sbi);
1060 free_sm:
1061         destroy_segment_manager(sbi);
1062 free_cp:
1063         kfree(sbi->ckpt);
1064 free_meta_inode:
1065         make_bad_inode(sbi->meta_inode);
1066         iput(sbi->meta_inode);
1067 free_sb_buf:
1068         brelse(raw_super_buf);
1069 free_sbi:
1070         kfree(sbi);
1071         return err;
1072 }
1073
1074 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1075                         const char *dev_name, void *data)
1076 {
1077         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1078 }
1079
1080 static struct file_system_type f2fs_fs_type = {
1081         .owner          = THIS_MODULE,
1082         .name           = "f2fs",
1083         .mount          = f2fs_mount,
1084         .kill_sb        = kill_block_super,
1085         .fs_flags       = FS_REQUIRES_DEV,
1086 };
1087 MODULE_ALIAS_FS("f2fs");
1088
1089 static int __init init_inodecache(void)
1090 {
1091         f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1092                         sizeof(struct f2fs_inode_info), NULL);
1093         if (!f2fs_inode_cachep)
1094                 return -ENOMEM;
1095         return 0;
1096 }
1097
1098 static void destroy_inodecache(void)
1099 {
1100         /*
1101          * Make sure all delayed rcu free inodes are flushed before we
1102          * destroy cache.
1103          */
1104         rcu_barrier();
1105         kmem_cache_destroy(f2fs_inode_cachep);
1106 }
1107
1108 static int __init init_f2fs_fs(void)
1109 {
1110         int err;
1111
1112         err = init_inodecache();
1113         if (err)
1114                 goto fail;
1115         err = create_node_manager_caches();
1116         if (err)
1117                 goto free_inodecache;
1118         err = create_segment_manager_caches();
1119         if (err)
1120                 goto free_node_manager_caches;
1121         err = create_gc_caches();
1122         if (err)
1123                 goto free_segment_manager_caches;
1124         err = create_checkpoint_caches();
1125         if (err)
1126                 goto free_gc_caches;
1127         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1128         if (!f2fs_kset) {
1129                 err = -ENOMEM;
1130                 goto free_checkpoint_caches;
1131         }
1132         err = register_filesystem(&f2fs_fs_type);
1133         if (err)
1134                 goto free_kset;
1135         f2fs_create_root_stats();
1136         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1137         return 0;
1138
1139 free_kset:
1140         kset_unregister(f2fs_kset);
1141 free_checkpoint_caches:
1142         destroy_checkpoint_caches();
1143 free_gc_caches:
1144         destroy_gc_caches();
1145 free_segment_manager_caches:
1146         destroy_segment_manager_caches();
1147 free_node_manager_caches:
1148         destroy_node_manager_caches();
1149 free_inodecache:
1150         destroy_inodecache();
1151 fail:
1152         return err;
1153 }
1154
1155 static void __exit exit_f2fs_fs(void)
1156 {
1157         remove_proc_entry("fs/f2fs", NULL);
1158         f2fs_destroy_root_stats();
1159         unregister_filesystem(&f2fs_fs_type);
1160         destroy_checkpoint_caches();
1161         destroy_gc_caches();
1162         destroy_segment_manager_caches();
1163         destroy_node_manager_caches();
1164         destroy_inodecache();
1165         kset_unregister(f2fs_kset);
1166 }
1167
1168 module_init(init_f2fs_fs)
1169 module_exit(exit_f2fs_fs)
1170
1171 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1172 MODULE_DESCRIPTION("Flash Friendly File System");
1173 MODULE_LICENSE("GPL");