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         NM_INFO,        /* struct f2fs_nm_info */
  78         F2FS_SBI,       /* struct f2fs_sb_info */
  79 };
  80
  81 struct f2fs_attr {
  82         struct attribute attr;
  83         ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
  84         ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
  85                          const char *, size_t);
  86         int struct_type;
  87         int offset;
  88 };
  89
  90 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
  91 {
  92         if (struct_type == GC_THREAD)
  93                 return (unsigned char *)sbi->gc_thread;
  94         else if (struct_type == SM_INFO)
  95                 return (unsigned char *)SM_I(sbi);
  96         else if (struct_type == NM_INFO)
  97                 return (unsigned char *)NM_I(sbi);
  98         else if (struct_type == F2FS_SBI)
  99                 return (unsigned char *)sbi;
 100         return NULL;
 101 }
 102
 103 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
 104                         struct f2fs_sb_info *sbi, char *buf)
 105 {
 106         unsigned char *ptr = NULL;
 107         unsigned int *ui;
 108
 109         ptr = __struct_ptr(sbi, a->struct_type);
 110         if (!ptr)
 111                 return -EINVAL;
 112
 113         ui = (unsigned int *)(ptr + a->offset);
 114
 115         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
 116 }
 117
 118 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
 119                         struct f2fs_sb_info *sbi,
 120                         const char *buf, size_t count)
 121 {
 122         unsigned char *ptr;
 123         unsigned long t;
 124         unsigned int *ui;
 125         ssize_t ret;
 126
 127         ptr = __struct_ptr(sbi, a->struct_type);
 128         if (!ptr)
 129                 return -EINVAL;
 130
 131         ui = (unsigned int *)(ptr + a->offset);
 132
 133         ret = kstrtoul(skip_spaces(buf), 0, &t);
 134         if (ret < 0)
 135                 return ret;
 136         *ui = t;
 137         return count;
 138 }
 139
 140 static ssize_t f2fs_attr_show(struct kobject *kobj,
 141                                 struct attribute *attr, char *buf)
 142 {
 143         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 144                                                                 s_kobj);
 145         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 146
 147         return a->show ? a->show(a, sbi, buf) : 0;
 148 }
 149
 150 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
 151                                                 const char *buf, size_t len)
 152 {
 153         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 154                                                                         s_kobj);
 155         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 156
 157         return a->store ? a->store(a, sbi, buf, len) : 0;
 158 }
 159
 160 static void f2fs_sb_release(struct kobject *kobj)
 161 {
 162         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 163                                                                 s_kobj);
 164         complete(&sbi->s_kobj_unregister);
 165 }
 166
 167 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
 168 static struct f2fs_attr f2fs_attr_##_name = {                   \
 169         .attr = {.name = __stringify(_name), .mode = _mode },   \
 170         .show   = _show,                                        \
 171         .store  = _store,                                       \
 172         .struct_type = _struct_type,                            \
 173         .offset = _offset                                       \
 174 }
 175
 176 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)    \
 177         F2FS_ATTR_OFFSET(struct_type, name, 0644,               \
 178                 f2fs_sbi_show, f2fs_sbi_store,                  \
 179                 offsetof(struct struct_name, elname))
 180
 181 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
 182 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
 183 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
 184 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
 185 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
 186 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
 187 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
 188 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
 189 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
 190 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
 191 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
 192
 193 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 194 static struct attribute *f2fs_attrs[] = {
 195         ATTR_LIST(gc_min_sleep_time),
 196         ATTR_LIST(gc_max_sleep_time),
 197         ATTR_LIST(gc_no_gc_sleep_time),
 198         ATTR_LIST(gc_idle),
 199         ATTR_LIST(reclaim_segments),
 200         ATTR_LIST(max_small_discards),
 201         ATTR_LIST(ipu_policy),
 202         ATTR_LIST(min_ipu_util),
 203         ATTR_LIST(max_victim_search),
 204         ATTR_LIST(dir_level),
 205         ATTR_LIST(ram_thresh),
 206         NULL,
 207 };
 208
 209 static const struct sysfs_ops f2fs_attr_ops = {
 210         .show   = f2fs_attr_show,
 211         .store  = f2fs_attr_store,
 212 };
 213
 214 static struct kobj_type f2fs_ktype = {
 215         .default_attrs  = f2fs_attrs,
 216         .sysfs_ops      = &f2fs_attr_ops,
 217         .release        = f2fs_sb_release,
 218 };
 219
 220 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 221 {
 222         struct va_format vaf;
 223         va_list args;
 224
 225         va_start(args, fmt);
 226         vaf.fmt = fmt;
 227         vaf.va = &args;
 228         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 229         va_end(args);
 230 }
 231
 232 static void init_once(void *foo)
 233 {
 234         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 235
 236         inode_init_once(&fi->vfs_inode);
 237 }
 238
 239 static int parse_options(struct super_block *sb, char *options)
 240 {
 241         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 242         substring_t args[MAX_OPT_ARGS];
 243         char *p, *name;
 244         int arg = 0;
 245
 246         if (!options)
 247                 return 0;
 248
 249         while ((p = strsep(&options, ",")) != NULL) {
 250                 int token;
 251                 if (!*p)
 252                         continue;
 253                 /*
 254                  * Initialize args struct so we know whether arg was
 255                  * found; some options take optional arguments.
 256                  */
 257                 args[0].to = args[0].from = NULL;
 258                 token = match_token(p, f2fs_tokens, args);
 259
 260                 switch (token) {
 261                 case Opt_gc_background:
 262                         name = match_strdup(&args[0]);
 263
 264                         if (!name)
 265                                 return -ENOMEM;
 266                         if (strlen(name) == 2 && !strncmp(name, "on", 2))
 267                                 set_opt(sbi, BG_GC);
 268                         else if (strlen(name) == 3 && !strncmp(name, "off", 3))
 269                                 clear_opt(sbi, BG_GC);
 270                         else {
 271                                 kfree(name);
 272                                 return -EINVAL;
 273                         }
 274                         kfree(name);
 275                         break;
 276                 case Opt_disable_roll_forward:
 277                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 278                         break;
 279                 case Opt_discard:
 280                         set_opt(sbi, DISCARD);
 281                         break;
 282                 case Opt_noheap:
 283                         set_opt(sbi, NOHEAP);
 284                         break;
 285 #ifdef CONFIG_F2FS_FS_XATTR
 286                 case Opt_user_xattr:
 287                         set_opt(sbi, XATTR_USER);
 288                         break;
 289                 case Opt_nouser_xattr:
 290                         clear_opt(sbi, XATTR_USER);
 291                         break;
 292                 case Opt_inline_xattr:
 293                         set_opt(sbi, INLINE_XATTR);
 294                         break;
 295 #else
 296                 case Opt_user_xattr:
 297                         f2fs_msg(sb, KERN_INFO,
 298                                 "user_xattr options not supported");
 299                         break;
 300                 case Opt_nouser_xattr:
 301                         f2fs_msg(sb, KERN_INFO,
 302                                 "nouser_xattr options not supported");
 303                         break;
 304                 case Opt_inline_xattr:
 305                         f2fs_msg(sb, KERN_INFO,
 306                                 "inline_xattr options not supported");
 307                         break;
 308 #endif
 309 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 310                 case Opt_acl:
 311                         set_opt(sbi, POSIX_ACL);
 312                         break;
 313                 case Opt_noacl:
 314                         clear_opt(sbi, POSIX_ACL);
 315                         break;
 316 #else
 317                 case Opt_acl:
 318                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 319                         break;
 320                 case Opt_noacl:
 321                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 322                         break;
 323 #endif
 324                 case Opt_active_logs:
 325                         if (args->from && match_int(args, &arg))
 326                                 return -EINVAL;
 327                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 328                                 return -EINVAL;
 329                         sbi->active_logs = arg;
 330                         break;
 331                 case Opt_disable_ext_identify:
 332                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 333                         break;
 334                 case Opt_inline_data:
 335                         set_opt(sbi, INLINE_DATA);
 336                         break;
 337                 default:
 338                         f2fs_msg(sb, KERN_ERR,
 339                                 "Unrecognized mount option \"%s\" or missing value",
 340                                 p);
 341                         return -EINVAL;
 342                 }
 343         }
 344         return 0;
 345 }
 346
 347 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 348 {
 349         struct f2fs_inode_info *fi;
 350
 351         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 352         if (!fi)
 353                 return NULL;
 354
 355         init_once((void *) fi);
 356
 357         /* Initialize f2fs-specific inode info */
 358         fi->vfs_inode.i_version = 1;
 359         atomic_set(&fi->dirty_dents, 0);
 360         fi->i_current_depth = 1;
 361         fi->i_advise = 0;
 362         rwlock_init(&fi->ext.ext_lock);
 363         init_rwsem(&fi->i_sem);
 364
 365         set_inode_flag(fi, FI_NEW_INODE);
 366
 367         if (test_opt(F2FS_SB(sb), INLINE_XATTR))
 368                 set_inode_flag(fi, FI_INLINE_XATTR);
 369
 370         /* Will be used by directory only */
 371         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 372
 373         return &fi->vfs_inode;
 374 }
 375
 376 static int f2fs_drop_inode(struct inode *inode)
 377 {
 378         /*
 379          * This is to avoid a deadlock condition like below.
 380          * writeback_single_inode(inode)
 381          *  - f2fs_write_data_page
 382          *    - f2fs_gc -> iput -> evict
 383          *       - inode_wait_for_writeback(inode)
 384          */
 385         if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
 386                 return 0;
 387         return generic_drop_inode(inode);
 388 }
 389
 390 /*
 391  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 392  *
 393  * We should call set_dirty_inode to write the dirty inode through write_inode.
 394  */
 395 static void f2fs_dirty_inode(struct inode *inode, int flags)
 396 {
 397         set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
 398 }
 399
 400 static void f2fs_i_callback(struct rcu_head *head)
 401 {
 402         struct inode *inode = container_of(head, struct inode, i_rcu);
 403         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 404 }
 405
 406 static void f2fs_destroy_inode(struct inode *inode)
 407 {
 408         call_rcu(&inode->i_rcu, f2fs_i_callback);
 409 }
 410
 411 static void f2fs_put_super(struct super_block *sb)
 412 {
 413         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 414
 415         if (sbi->s_proc) {
 416                 remove_proc_entry("segment_info", sbi->s_proc);
 417                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 418         }
 419         kobject_del(&sbi->s_kobj);
 420
 421         f2fs_destroy_stats(sbi);
 422         stop_gc_thread(sbi);
 423
 424         /* We don't need to do checkpoint when it's clean */
 425         if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
 426                 write_checkpoint(sbi, true);
 427
 428         iput(sbi->node_inode);
 429         iput(sbi->meta_inode);
 430
 431         /* destroy f2fs internal modules */
 432         destroy_node_manager(sbi);
 433         destroy_segment_manager(sbi);
 434
 435         kfree(sbi->ckpt);
 436         kobject_put(&sbi->s_kobj);
 437         wait_for_completion(&sbi->s_kobj_unregister);
 438
 439         sb->s_fs_info = NULL;
 440         brelse(sbi->raw_super_buf);
 441         kfree(sbi);
 442 }
 443
 444 int f2fs_sync_fs(struct super_block *sb, int sync)
 445 {
 446         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 447
 448         trace_f2fs_sync_fs(sb, sync);
 449
 450         if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
 451                 return 0;
 452
 453         if (sync) {
 454                 mutex_lock(&sbi->gc_mutex);
 455                 write_checkpoint(sbi, false);
 456                 mutex_unlock(&sbi->gc_mutex);
 457         } else {
 458                 f2fs_balance_fs(sbi);
 459         }
 460
 461         return 0;
 462 }
 463
 464 static int f2fs_freeze(struct super_block *sb)
 465 {
 466         int err;
 467
 468         if (f2fs_readonly(sb))
 469                 return 0;
 470
 471         err = f2fs_sync_fs(sb, 1);
 472         return err;
 473 }
 474
 475 static int f2fs_unfreeze(struct super_block *sb)
 476 {
 477         return 0;
 478 }
 479
 480 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 481 {
 482         struct super_block *sb = dentry->d_sb;
 483         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 484         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 485         block_t total_count, user_block_count, start_count, ovp_count;
 486
 487         total_count = le64_to_cpu(sbi->raw_super->block_count);
 488         user_block_count = sbi->user_block_count;
 489         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 490         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 491         buf->f_type = F2FS_SUPER_MAGIC;
 492         buf->f_bsize = sbi->blocksize;
 493
 494         buf->f_blocks = total_count - start_count;
 495         buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
 496         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 497
 498         buf->f_files = sbi->total_node_count;
 499         buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
 500
 501         buf->f_namelen = F2FS_NAME_LEN;
 502         buf->f_fsid.val[0] = (u32)id;
 503         buf->f_fsid.val[1] = (u32)(id >> 32);
 504
 505         return 0;
 506 }
 507
 508 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 509 {
 510         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 511
 512         if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
 513                 seq_printf(seq, ",background_gc=%s", "on");
 514         else
 515                 seq_printf(seq, ",background_gc=%s", "off");
 516         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 517                 seq_puts(seq, ",disable_roll_forward");
 518         if (test_opt(sbi, DISCARD))
 519                 seq_puts(seq, ",discard");
 520         if (test_opt(sbi, NOHEAP))
 521                 seq_puts(seq, ",no_heap_alloc");
 522 #ifdef CONFIG_F2FS_FS_XATTR
 523         if (test_opt(sbi, XATTR_USER))
 524                 seq_puts(seq, ",user_xattr");
 525         else
 526                 seq_puts(seq, ",nouser_xattr");
 527         if (test_opt(sbi, INLINE_XATTR))
 528                 seq_puts(seq, ",inline_xattr");
 529 #endif
 530 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 531         if (test_opt(sbi, POSIX_ACL))
 532                 seq_puts(seq, ",acl");
 533         else
 534                 seq_puts(seq, ",noacl");
 535 #endif
 536         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 537                 seq_puts(seq, ",disable_ext_identify");
 538         if (test_opt(sbi, INLINE_DATA))
 539                 seq_puts(seq, ",inline_data");
 540         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 541
 542         return 0;
 543 }
 544
 545 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 546 {
 547         struct super_block *sb = seq->private;
 548         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 549         unsigned int total_segs =
 550                         le32_to_cpu(sbi->raw_super->segment_count_main);
 551         int i;
 552
 553         seq_puts(seq, "format: segment_type|valid_blocks\n"
 554                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 555
 556         for (i = 0; i < total_segs; i++) {
 557                 struct seg_entry *se = get_seg_entry(sbi, i);
 558
 559                 if ((i % 10) == 0)
 560                         seq_printf(seq, "%-5d", i);
 561                 seq_printf(seq, "%d|%-3u", se->type,
 562                                         get_valid_blocks(sbi, i, 1));
 563                 if ((i % 10) == 9 || i == (total_segs - 1))
 564                         seq_putc(seq, '\n');
 565                 else
 566                         seq_putc(seq, ' ');
 567         }
 568
 569         return 0;
 570 }
 571
 572 static int segment_info_open_fs(struct inode *inode, struct file *file)
 573 {
 574         return single_open(file, segment_info_seq_show, PDE_DATA(inode));
 575 }
 576
 577 static const struct file_operations f2fs_seq_segment_info_fops = {
 578         .owner = THIS_MODULE,
 579         .open = segment_info_open_fs,
 580         .read = seq_read,
 581         .llseek = seq_lseek,
 582         .release = single_release,
 583 };
 584
 585 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 586 {
 587         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 588         struct f2fs_mount_info org_mount_opt;
 589         int err, active_logs;
 590
 591         /*
 592          * Save the old mount options in case we
 593          * need to restore them.
 594          */
 595         org_mount_opt = sbi->mount_opt;
 596         active_logs = sbi->active_logs;
 597
 598         /* parse mount options */
 599         err = parse_options(sb, data);
 600         if (err)
 601                 goto restore_opts;
 602
 603         /*
 604          * Previous and new state of filesystem is RO,
 605          * so no point in checking GC conditions.
 606          */
 607         if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
 608                 goto skip;
 609
 610         /*
 611          * We stop the GC thread if FS is mounted as RO
 612          * or if background_gc = off is passed in mount
 613          * option. Also sync the filesystem.
 614          */
 615         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
 616                 if (sbi->gc_thread) {
 617                         stop_gc_thread(sbi);
 618                         f2fs_sync_fs(sb, 1);
 619                 }
 620         } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
 621                 err = start_gc_thread(sbi);
 622                 if (err)
 623                         goto restore_opts;
 624         }
 625 skip:
 626         /* Update the POSIXACL Flag */
 627          sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 628                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 629         return 0;
 630
 631 restore_opts:
 632         sbi->mount_opt = org_mount_opt;
 633         sbi->active_logs = active_logs;
 634         return err;
 635 }
 636
 637 static struct super_operations f2fs_sops = {
 638         .alloc_inode    = f2fs_alloc_inode,
 639         .drop_inode     = f2fs_drop_inode,
 640         .destroy_inode  = f2fs_destroy_inode,
 641         .write_inode    = f2fs_write_inode,
 642         .dirty_inode    = f2fs_dirty_inode,
 643         .show_options   = f2fs_show_options,
 644         .evict_inode    = f2fs_evict_inode,
 645         .put_super      = f2fs_put_super,
 646         .sync_fs        = f2fs_sync_fs,
 647         .freeze_fs      = f2fs_freeze,
 648         .unfreeze_fs    = f2fs_unfreeze,
 649         .statfs         = f2fs_statfs,
 650         .remount_fs     = f2fs_remount,
 651 };
 652
 653 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
 654                 u64 ino, u32 generation)
 655 {
 656         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 657         struct inode *inode;
 658
 659         if (unlikely(ino < F2FS_ROOT_INO(sbi)))
 660                 return ERR_PTR(-ESTALE);
 661         if (unlikely(ino >= NM_I(sbi)->max_nid))
 662                 return ERR_PTR(-ESTALE);
 663
 664         /*
 665          * f2fs_iget isn't quite right if the inode is currently unallocated!
 666          * However f2fs_iget currently does appropriate checks to handle stale
 667          * inodes so everything is OK.
 668          */
 669         inode = f2fs_iget(sb, ino);
 670         if (IS_ERR(inode))
 671                 return ERR_CAST(inode);
 672         if (unlikely(generation && inode->i_generation != generation)) {
 673                 /* we didn't find the right inode.. */
 674                 iput(inode);
 675                 return ERR_PTR(-ESTALE);
 676         }
 677         return inode;
 678 }
 679
 680 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
 681                 int fh_len, int fh_type)
 682 {
 683         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 684                                     f2fs_nfs_get_inode);
 685 }
 686
 687 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
 688                 int fh_len, int fh_type)
 689 {
 690         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 691                                     f2fs_nfs_get_inode);
 692 }
 693
 694 static const struct export_operations f2fs_export_ops = {
 695         .fh_to_dentry = f2fs_fh_to_dentry,
 696         .fh_to_parent = f2fs_fh_to_parent,
 697         .get_parent = f2fs_get_parent,
 698 };
 699
 700 static loff_t max_file_size(unsigned bits)
 701 {
 702         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
 703         loff_t leaf_count = ADDRS_PER_BLOCK;
 704
 705         /* two direct node blocks */
 706         result += (leaf_count * 2);
 707
 708         /* two indirect node blocks */
 709         leaf_count *= NIDS_PER_BLOCK;
 710         result += (leaf_count * 2);
 711
 712         /* one double indirect node block */
 713         leaf_count *= NIDS_PER_BLOCK;
 714         result += leaf_count;
 715
 716         result <<= bits;
 717         return result;
 718 }
 719
 720 static int sanity_check_raw_super(struct super_block *sb,
 721                         struct f2fs_super_block *raw_super)
 722 {
 723         unsigned int blocksize;
 724
 725         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
 726                 f2fs_msg(sb, KERN_INFO,
 727                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
 728                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
 729                 return 1;
 730         }
 731
 732         /* Currently, support only 4KB page cache size */
 733         if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
 734                 f2fs_msg(sb, KERN_INFO,
 735                         "Invalid page_cache_size (%lu), supports only 4KB\n",
 736                         PAGE_CACHE_SIZE);
 737                 return 1;
 738         }
 739
 740         /* Currently, support only 4KB block size */
 741         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
 742         if (blocksize != F2FS_BLKSIZE) {
 743                 f2fs_msg(sb, KERN_INFO,
 744                         "Invalid blocksize (%u), supports only 4KB\n",
 745                         blocksize);
 746                 return 1;
 747         }
 748
 749         if (le32_to_cpu(raw_super->log_sectorsize) !=
 750                                         F2FS_LOG_SECTOR_SIZE) {
 751                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
 752                 return 1;
 753         }
 754         if (le32_to_cpu(raw_super->log_sectors_per_block) !=
 755                                         F2FS_LOG_SECTORS_PER_BLOCK) {
 756                 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
 757                 return 1;
 758         }
 759         return 0;
 760 }
 761
 762 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
 763 {
 764         unsigned int total, fsmeta;
 765         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
 766         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 767
 768         total = le32_to_cpu(raw_super->segment_count);
 769         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
 770         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
 771         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
 772         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
 773         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
 774
 775         if (unlikely(fsmeta >= total))
 776                 return 1;
 777
 778         if (unlikely(is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
 779                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
 780                 return 1;
 781         }
 782         return 0;
 783 }
 784
 785 static void init_sb_info(struct f2fs_sb_info *sbi)
 786 {
 787         struct f2fs_super_block *raw_super = sbi->raw_super;
 788         int i;
 789
 790         sbi->log_sectors_per_block =
 791                 le32_to_cpu(raw_super->log_sectors_per_block);
 792         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
 793         sbi->blocksize = 1 << sbi->log_blocksize;
 794         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
 795         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
 796         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
 797         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
 798         sbi->total_sections = le32_to_cpu(raw_super->section_count);
 799         sbi->total_node_count =
 800                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
 801                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
 802         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
 803         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
 804         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
 805         sbi->cur_victim_sec = NULL_SECNO;
 806         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
 807
 808         for (i = 0; i < NR_COUNT_TYPE; i++)
 809                 atomic_set(&sbi->nr_pages[i], 0);
 810
 811         sbi->dir_level = DEF_DIR_LEVEL;
 812 }
 813
 814 /*
 815  * Read f2fs raw super block.
 816  * Because we have two copies of super block, so read the first one at first,
 817  * if the first one is invalid, move to read the second one.
 818  */
 819 static int read_raw_super_block(struct super_block *sb,
 820                         struct f2fs_super_block **raw_super,
 821                         struct buffer_head **raw_super_buf)
 822 {
 823         int block = 0;
 824
 825 retry:
 826         *raw_super_buf = sb_bread(sb, block);
 827         if (!*raw_super_buf) {
 828                 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
 829                                 block + 1);
 830                 if (block == 0) {
 831                         block++;
 832                         goto retry;
 833                 } else {
 834                         return -EIO;
 835                 }
 836         }
 837
 838         *raw_super = (struct f2fs_super_block *)
 839                 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
 840
 841         /* sanity checking of raw super */
 842         if (sanity_check_raw_super(sb, *raw_super)) {
 843                 brelse(*raw_super_buf);
 844                 f2fs_msg(sb, KERN_ERR,
 845                         "Can't find valid F2FS filesystem in %dth superblock",
 846                                                                 block + 1);
 847                 if (block == 0) {
 848                         block++;
 849                         goto retry;
 850                 } else {
 851                         return -EINVAL;
 852                 }
 853         }
 854
 855         return 0;
 856 }
 857
 858 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 859 {
 860         struct f2fs_sb_info *sbi;
 861         struct f2fs_super_block *raw_super;
 862         struct buffer_head *raw_super_buf;
 863         struct inode *root;
 864         long err = -EINVAL;
 865         int i;
 866
 867         /* allocate memory for f2fs-specific super block info */
 868         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
 869         if (!sbi)
 870                 return -ENOMEM;
 871
 872         /* set a block size */
 873         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
 874                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
 875                 goto free_sbi;
 876         }
 877
 878         err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
 879         if (err)
 880                 goto free_sbi;
 881
 882         sb->s_fs_info = sbi;
 883         /* init some FS parameters */
 884         sbi->active_logs = NR_CURSEG_TYPE;
 885
 886         set_opt(sbi, BG_GC);
 887
 888 #ifdef CONFIG_F2FS_FS_XATTR
 889         set_opt(sbi, XATTR_USER);
 890 #endif
 891 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 892         set_opt(sbi, POSIX_ACL);
 893 #endif
 894         /* parse mount options */
 895         err = parse_options(sb, (char *)data);
 896         if (err)
 897                 goto free_sb_buf;
 898
 899         sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
 900         sb->s_max_links = F2FS_LINK_MAX;
 901         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
 902
 903         sb->s_op = &f2fs_sops;
 904         sb->s_xattr = f2fs_xattr_handlers;
 905         sb->s_export_op = &f2fs_export_ops;
 906         sb->s_magic = F2FS_SUPER_MAGIC;
 907         sb->s_time_gran = 1;
 908         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 909                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 910         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
 911
 912         /* init f2fs-specific super block info */
 913         sbi->sb = sb;
 914         sbi->raw_super = raw_super;
 915         sbi->raw_super_buf = raw_super_buf;
 916         mutex_init(&sbi->gc_mutex);
 917         mutex_init(&sbi->writepages);
 918         mutex_init(&sbi->cp_mutex);
 919         mutex_init(&sbi->node_write);
 920         sbi->por_doing = false;
 921         spin_lock_init(&sbi->stat_lock);
 922
 923         init_rwsem(&sbi->read_io.io_rwsem);
 924         sbi->read_io.sbi = sbi;
 925         sbi->read_io.bio = NULL;
 926         for (i = 0; i < NR_PAGE_TYPE; i++) {
 927                 init_rwsem(&sbi->write_io[i].io_rwsem);
 928                 sbi->write_io[i].sbi = sbi;
 929                 sbi->write_io[i].bio = NULL;
 930         }
 931
 932         init_rwsem(&sbi->cp_rwsem);
 933         init_waitqueue_head(&sbi->cp_wait);
 934         init_sb_info(sbi);
 935
 936         /* get an inode for meta space */
 937         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
 938         if (IS_ERR(sbi->meta_inode)) {
 939                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
 940                 err = PTR_ERR(sbi->meta_inode);
 941                 goto free_sb_buf;
 942         }
 943
 944         err = get_valid_checkpoint(sbi);
 945         if (err) {
 946                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
 947                 goto free_meta_inode;
 948         }
 949
 950         /* sanity checking of checkpoint */
 951         err = -EINVAL;
 952         if (sanity_check_ckpt(sbi)) {
 953                 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
 954                 goto free_cp;
 955         }
 956
 957         sbi->total_valid_node_count =
 958                                 le32_to_cpu(sbi->ckpt->valid_node_count);
 959         sbi->total_valid_inode_count =
 960                                 le32_to_cpu(sbi->ckpt->valid_inode_count);
 961         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
 962         sbi->total_valid_block_count =
 963                                 le64_to_cpu(sbi->ckpt->valid_block_count);
 964         sbi->last_valid_block_count = sbi->total_valid_block_count;
 965         sbi->alloc_valid_block_count = 0;
 966         INIT_LIST_HEAD(&sbi->dir_inode_list);
 967         spin_lock_init(&sbi->dir_inode_lock);
 968
 969         init_orphan_info(sbi);
 970
 971         /* setup f2fs internal modules */
 972         err = build_segment_manager(sbi);
 973         if (err) {
 974                 f2fs_msg(sb, KERN_ERR,
 975                         "Failed to initialize F2FS segment manager");
 976                 goto free_sm;
 977         }
 978         err = build_node_manager(sbi);
 979         if (err) {
 980                 f2fs_msg(sb, KERN_ERR,
 981                         "Failed to initialize F2FS node manager");
 982                 goto free_nm;
 983         }
 984
 985         build_gc_manager(sbi);
 986
 987         /* get an inode for node space */
 988         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
 989         if (IS_ERR(sbi->node_inode)) {
 990                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
 991                 err = PTR_ERR(sbi->node_inode);
 992                 goto free_nm;
 993         }
 994
 995         /* if there are nt orphan nodes free them */
 996         recover_orphan_inodes(sbi);
 997
 998         /* read root inode and dentry */
 999         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1000         if (IS_ERR(root)) {
1001                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1002                 err = PTR_ERR(root);
1003                 goto free_node_inode;
1004         }
1005         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1006                 err = -EINVAL;
1007                 goto free_root_inode;
1008         }
1009
1010         sb->s_root = d_make_root(root); /* allocate root dentry */
1011         if (!sb->s_root) {
1012                 err = -ENOMEM;
1013                 goto free_root_inode;
1014         }
1015
1016         err = f2fs_build_stats(sbi);
1017         if (err)
1018                 goto free_root_inode;
1019
1020         if (f2fs_proc_root)
1021                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1022
1023         if (sbi->s_proc)
1024                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1025                                  &f2fs_seq_segment_info_fops, sb);
1026
1027         if (test_opt(sbi, DISCARD)) {
1028                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1029                 if (!blk_queue_discard(q))
1030                         f2fs_msg(sb, KERN_WARNING,
1031                                         "mounting with \"discard\" option, but "
1032                                         "the device does not support discard");
1033         }
1034
1035         sbi->s_kobj.kset = f2fs_kset;
1036         init_completion(&sbi->s_kobj_unregister);
1037         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1038                                                         "%s", sb->s_id);
1039         if (err)
1040                 goto free_proc;
1041
1042         /* recover fsynced data */
1043         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1044                 err = recover_fsync_data(sbi);
1045                 if (err)
1046                         f2fs_msg(sb, KERN_ERR,
1047                                 "Cannot recover all fsync data errno=%ld", err);
1048         }
1049
1050         /*
1051          * If filesystem is not mounted as read-only then
1052          * do start the gc_thread.
1053          */
1054         if (!(sb->s_flags & MS_RDONLY)) {
1055                 /* After POR, we can run background GC thread.*/
1056                 err = start_gc_thread(sbi);
1057                 if (err)
1058                         goto free_kobj;
1059         }
1060         return 0;
1061
1062 free_kobj:
1063         kobject_del(&sbi->s_kobj);
1064 free_proc:
1065         if (sbi->s_proc) {
1066                 remove_proc_entry("segment_info", sbi->s_proc);
1067                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1068         }
1069         f2fs_destroy_stats(sbi);
1070 free_root_inode:
1071         dput(sb->s_root);
1072         sb->s_root = NULL;
1073 free_node_inode:
1074         iput(sbi->node_inode);
1075 free_nm:
1076         destroy_node_manager(sbi);
1077 free_sm:
1078         destroy_segment_manager(sbi);
1079 free_cp:
1080         kfree(sbi->ckpt);
1081 free_meta_inode:
1082         make_bad_inode(sbi->meta_inode);
1083         iput(sbi->meta_inode);
1084 free_sb_buf:
1085         brelse(raw_super_buf);
1086 free_sbi:
1087         kfree(sbi);
1088         return err;
1089 }
1090
1091 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1092                         const char *dev_name, void *data)
1093 {
1094         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1095 }
1096
1097 static struct file_system_type f2fs_fs_type = {
1098         .owner          = THIS_MODULE,
1099         .name           = "f2fs",
1100         .mount          = f2fs_mount,
1101         .kill_sb        = kill_block_super,
1102         .fs_flags       = FS_REQUIRES_DEV,
1103 };
1104 MODULE_ALIAS_FS("f2fs");
1105
1106 static int __init init_inodecache(void)
1107 {
1108         f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1109                         sizeof(struct f2fs_inode_info));
1110         if (!f2fs_inode_cachep)
1111                 return -ENOMEM;
1112         return 0;
1113 }
1114
1115 static void destroy_inodecache(void)
1116 {
1117         /*
1118          * Make sure all delayed rcu free inodes are flushed before we
1119          * destroy cache.
1120          */
1121         rcu_barrier();
1122         kmem_cache_destroy(f2fs_inode_cachep);
1123 }
1124
1125 static int __init init_f2fs_fs(void)
1126 {
1127         int err;
1128
1129         err = init_inodecache();
1130         if (err)
1131                 goto fail;
1132         err = create_node_manager_caches();
1133         if (err)
1134                 goto free_inodecache;
1135         err = create_segment_manager_caches();
1136         if (err)
1137                 goto free_node_manager_caches;
1138         err = create_gc_caches();
1139         if (err)
1140                 goto free_segment_manager_caches;
1141         err = create_checkpoint_caches();
1142         if (err)
1143                 goto free_gc_caches;
1144         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1145         if (!f2fs_kset) {
1146                 err = -ENOMEM;
1147                 goto free_checkpoint_caches;
1148         }
1149         err = register_filesystem(&f2fs_fs_type);
1150         if (err)
1151                 goto free_kset;
1152         f2fs_create_root_stats();
1153         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1154         return 0;
1155
1156 free_kset:
1157         kset_unregister(f2fs_kset);
1158 free_checkpoint_caches:
1159         destroy_checkpoint_caches();
1160 free_gc_caches:
1161         destroy_gc_caches();
1162 free_segment_manager_caches:
1163         destroy_segment_manager_caches();
1164 free_node_manager_caches:
1165         destroy_node_manager_caches();
1166 free_inodecache:
1167         destroy_inodecache();
1168 fail:
1169         return err;
1170 }
1171
1172 static void __exit exit_f2fs_fs(void)
1173 {
1174         remove_proc_entry("fs/f2fs", NULL);
1175         f2fs_destroy_root_stats();
1176         unregister_filesystem(&f2fs_fs_type);
1177         destroy_checkpoint_caches();
1178         destroy_gc_caches();
1179         destroy_segment_manager_caches();
1180         destroy_node_manager_caches();
1181         destroy_inodecache();
1182         kset_unregister(f2fs_kset);
1183 }
1184
1185 module_init(init_f2fs_fs)
1186 module_exit(exit_f2fs_fs)
1187
1188 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1189 MODULE_DESCRIPTION("Flash Friendly File System");
1190 MODULE_LICENSE("GPL");