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 #include "trace.h"
  34
  35 #define CREATE_TRACE_POINTS
  36 #include <trace/events/f2fs.h>
  37
  38 static struct proc_dir_entry *f2fs_proc_root;
  39 static struct kmem_cache *f2fs_inode_cachep;
  40 static struct kset *f2fs_kset;
  41
  42 #ifdef CONFIG_F2FS_FAULT_INJECTION
  43 struct f2fs_fault_info f2fs_fault;
  44
  45 char *fault_name[FAULT_MAX] = {
  46         [FAULT_KMALLOC]         = "kmalloc",
  47         [FAULT_PAGE_ALLOC]      = "page alloc",
  48         [FAULT_ALLOC_NID]       = "alloc nid",
  49         [FAULT_ORPHAN]          = "orphan",
  50         [FAULT_BLOCK]           = "no more block",
  51         [FAULT_DIR_DEPTH]       = "too big dir depth",
  52 };
  53
  54 static void f2fs_build_fault_attr(unsigned int rate)
  55 {
  56         if (rate) {
  57                 atomic_set(&f2fs_fault.inject_ops, 0);
  58                 f2fs_fault.inject_rate = rate;
  59                 f2fs_fault.inject_type = (1 << FAULT_MAX) - 1;
  60         } else {
  61                 memset(&f2fs_fault, 0, sizeof(struct f2fs_fault_info));
  62         }
  63 }
  64 #endif
  65
  66 /* f2fs-wide shrinker description */
  67 static struct shrinker f2fs_shrinker_info = {
  68         .scan_objects = f2fs_shrink_scan,
  69         .count_objects = f2fs_shrink_count,
  70         .seeks = DEFAULT_SEEKS,
  71 };
  72
  73 enum {
  74         Opt_gc_background,
  75         Opt_disable_roll_forward,
  76         Opt_norecovery,
  77         Opt_discard,
  78         Opt_noheap,
  79         Opt_user_xattr,
  80         Opt_nouser_xattr,
  81         Opt_acl,
  82         Opt_noacl,
  83         Opt_active_logs,
  84         Opt_disable_ext_identify,
  85         Opt_inline_xattr,
  86         Opt_inline_data,
  87         Opt_inline_dentry,
  88         Opt_flush_merge,
  89         Opt_nobarrier,
  90         Opt_fastboot,
  91         Opt_extent_cache,
  92         Opt_noextent_cache,
  93         Opt_noinline_data,
  94         Opt_data_flush,
  95         Opt_fault_injection,
  96         Opt_err,
  97 };
  98
  99 static match_table_t f2fs_tokens = {
 100         {Opt_gc_background, "background_gc=%s"},
 101         {Opt_disable_roll_forward, "disable_roll_forward"},
 102         {Opt_norecovery, "norecovery"},
 103         {Opt_discard, "discard"},
 104         {Opt_noheap, "no_heap"},
 105         {Opt_user_xattr, "user_xattr"},
 106         {Opt_nouser_xattr, "nouser_xattr"},
 107         {Opt_acl, "acl"},
 108         {Opt_noacl, "noacl"},
 109         {Opt_active_logs, "active_logs=%u"},
 110         {Opt_disable_ext_identify, "disable_ext_identify"},
 111         {Opt_inline_xattr, "inline_xattr"},
 112         {Opt_inline_data, "inline_data"},
 113         {Opt_inline_dentry, "inline_dentry"},
 114         {Opt_flush_merge, "flush_merge"},
 115         {Opt_nobarrier, "nobarrier"},
 116         {Opt_fastboot, "fastboot"},
 117         {Opt_extent_cache, "extent_cache"},
 118         {Opt_noextent_cache, "noextent_cache"},
 119         {Opt_noinline_data, "noinline_data"},
 120         {Opt_data_flush, "data_flush"},
 121         {Opt_fault_injection, "fault_injection=%u"},
 122         {Opt_err, NULL},
 123 };
 124
 125 /* Sysfs support for f2fs */
 126 enum {
 127         GC_THREAD,      /* struct f2fs_gc_thread */
 128         SM_INFO,        /* struct f2fs_sm_info */
 129         NM_INFO,        /* struct f2fs_nm_info */
 130         F2FS_SBI,       /* struct f2fs_sb_info */
 131 #ifdef CONFIG_F2FS_FAULT_INJECTION
 132         FAULT_INFO_RATE,        /* struct f2fs_fault_info */
 133         FAULT_INFO_TYPE,        /* struct f2fs_fault_info */
 134 #endif
 135 };
 136
 137 struct f2fs_attr {
 138         struct attribute attr;
 139         ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
 140         ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
 141                          const char *, size_t);
 142         int struct_type;
 143         int offset;
 144 };
 145
 146 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
 147 {
 148         if (struct_type == GC_THREAD)
 149                 return (unsigned char *)sbi->gc_thread;
 150         else if (struct_type == SM_INFO)
 151                 return (unsigned char *)SM_I(sbi);
 152         else if (struct_type == NM_INFO)
 153                 return (unsigned char *)NM_I(sbi);
 154         else if (struct_type == F2FS_SBI)
 155                 return (unsigned char *)sbi;
 156 #ifdef CONFIG_F2FS_FAULT_INJECTION
 157         else if (struct_type == FAULT_INFO_RATE ||
 158                                         struct_type == FAULT_INFO_TYPE)
 159                 return (unsigned char *)&f2fs_fault;
 160 #endif
 161         return NULL;
 162 }
 163
 164 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
 165                 struct f2fs_sb_info *sbi, char *buf)
 166 {
 167         struct super_block *sb = sbi->sb;
 168
 169         if (!sb->s_bdev->bd_part)
 170                 return snprintf(buf, PAGE_SIZE, "0\n");
 171
 172         return snprintf(buf, PAGE_SIZE, "%llu\n",
 173                 (unsigned long long)(sbi->kbytes_written +
 174                         BD_PART_WRITTEN(sbi)));
 175 }
 176
 177 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
 178                         struct f2fs_sb_info *sbi, char *buf)
 179 {
 180         unsigned char *ptr = NULL;
 181         unsigned int *ui;
 182
 183         ptr = __struct_ptr(sbi, a->struct_type);
 184         if (!ptr)
 185                 return -EINVAL;
 186
 187         ui = (unsigned int *)(ptr + a->offset);
 188
 189         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
 190 }
 191
 192 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
 193                         struct f2fs_sb_info *sbi,
 194                         const char *buf, size_t count)
 195 {
 196         unsigned char *ptr;
 197         unsigned long t;
 198         unsigned int *ui;
 199         ssize_t ret;
 200
 201         ptr = __struct_ptr(sbi, a->struct_type);
 202         if (!ptr)
 203                 return -EINVAL;
 204
 205         ui = (unsigned int *)(ptr + a->offset);
 206
 207         ret = kstrtoul(skip_spaces(buf), 0, &t);
 208         if (ret < 0)
 209                 return ret;
 210 #ifdef CONFIG_F2FS_FAULT_INJECTION
 211         if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
 212                 return -EINVAL;
 213 #endif
 214         *ui = t;
 215         return count;
 216 }
 217
 218 static ssize_t f2fs_attr_show(struct kobject *kobj,
 219                                 struct attribute *attr, char *buf)
 220 {
 221         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 222                                                                 s_kobj);
 223         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 224
 225         return a->show ? a->show(a, sbi, buf) : 0;
 226 }
 227
 228 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
 229                                                 const char *buf, size_t len)
 230 {
 231         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 232                                                                         s_kobj);
 233         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 234
 235         return a->store ? a->store(a, sbi, buf, len) : 0;
 236 }
 237
 238 static void f2fs_sb_release(struct kobject *kobj)
 239 {
 240         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 241                                                                 s_kobj);
 242         complete(&sbi->s_kobj_unregister);
 243 }
 244
 245 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
 246 static struct f2fs_attr f2fs_attr_##_name = {                   \
 247         .attr = {.name = __stringify(_name), .mode = _mode },   \
 248         .show   = _show,                                        \
 249         .store  = _store,                                       \
 250         .struct_type = _struct_type,                            \
 251         .offset = _offset                                       \
 252 }
 253
 254 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)    \
 255         F2FS_ATTR_OFFSET(struct_type, name, 0644,               \
 256                 f2fs_sbi_show, f2fs_sbi_store,                  \
 257                 offsetof(struct struct_name, elname))
 258
 259 #define F2FS_GENERAL_RO_ATTR(name) \
 260 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
 261
 262 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
 263 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
 264 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
 265 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
 266 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
 267 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
 268 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
 269 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
 270 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
 271 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
 272 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
 273 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
 274 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
 275 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
 276 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
 277 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
 278 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
 279 #ifdef CONFIG_F2FS_FAULT_INJECTION
 280 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
 281 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
 282 #endif
 283 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
 284
 285 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 286 static struct attribute *f2fs_attrs[] = {
 287         ATTR_LIST(gc_min_sleep_time),
 288         ATTR_LIST(gc_max_sleep_time),
 289         ATTR_LIST(gc_no_gc_sleep_time),
 290         ATTR_LIST(gc_idle),
 291         ATTR_LIST(reclaim_segments),
 292         ATTR_LIST(max_small_discards),
 293         ATTR_LIST(batched_trim_sections),
 294         ATTR_LIST(ipu_policy),
 295         ATTR_LIST(min_ipu_util),
 296         ATTR_LIST(min_fsync_blocks),
 297         ATTR_LIST(max_victim_search),
 298         ATTR_LIST(dir_level),
 299         ATTR_LIST(ram_thresh),
 300         ATTR_LIST(ra_nid_pages),
 301         ATTR_LIST(dirty_nats_ratio),
 302         ATTR_LIST(cp_interval),
 303         ATTR_LIST(idle_interval),
 304         ATTR_LIST(lifetime_write_kbytes),
 305         NULL,
 306 };
 307
 308 static const struct sysfs_ops f2fs_attr_ops = {
 309         .show   = f2fs_attr_show,
 310         .store  = f2fs_attr_store,
 311 };
 312
 313 static struct kobj_type f2fs_ktype = {
 314         .default_attrs  = f2fs_attrs,
 315         .sysfs_ops      = &f2fs_attr_ops,
 316         .release        = f2fs_sb_release,
 317 };
 318
 319 #ifdef CONFIG_F2FS_FAULT_INJECTION
 320 /* sysfs for f2fs fault injection */
 321 static struct kobject f2fs_fault_inject;
 322
 323 static struct attribute *f2fs_fault_attrs[] = {
 324         ATTR_LIST(inject_rate),
 325         ATTR_LIST(inject_type),
 326         NULL
 327 };
 328
 329 static struct kobj_type f2fs_fault_ktype = {
 330         .default_attrs  = f2fs_fault_attrs,
 331         .sysfs_ops      = &f2fs_attr_ops,
 332 };
 333 #endif
 334
 335 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 336 {
 337         struct va_format vaf;
 338         va_list args;
 339
 340         va_start(args, fmt);
 341         vaf.fmt = fmt;
 342         vaf.va = &args;
 343         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 344         va_end(args);
 345 }
 346
 347 static void init_once(void *foo)
 348 {
 349         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 350
 351         inode_init_once(&fi->vfs_inode);
 352 }
 353
 354 static int parse_options(struct super_block *sb, char *options)
 355 {
 356         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 357         struct request_queue *q;
 358         substring_t args[MAX_OPT_ARGS];
 359         char *p, *name;
 360         int arg = 0;
 361
 362 #ifdef CONFIG_F2FS_FAULT_INJECTION
 363         f2fs_build_fault_attr(0);
 364 #endif
 365
 366         if (!options)
 367                 return 0;
 368
 369         while ((p = strsep(&options, ",")) != NULL) {
 370                 int token;
 371                 if (!*p)
 372                         continue;
 373                 /*
 374                  * Initialize args struct so we know whether arg was
 375                  * found; some options take optional arguments.
 376                  */
 377                 args[0].to = args[0].from = NULL;
 378                 token = match_token(p, f2fs_tokens, args);
 379
 380                 switch (token) {
 381                 case Opt_gc_background:
 382                         name = match_strdup(&args[0]);
 383
 384                         if (!name)
 385                                 return -ENOMEM;
 386                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 387                                 set_opt(sbi, BG_GC);
 388                                 clear_opt(sbi, FORCE_FG_GC);
 389                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 390                                 clear_opt(sbi, BG_GC);
 391                                 clear_opt(sbi, FORCE_FG_GC);
 392                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 393                                 set_opt(sbi, BG_GC);
 394                                 set_opt(sbi, FORCE_FG_GC);
 395                         } else {
 396                                 kfree(name);
 397                                 return -EINVAL;
 398                         }
 399                         kfree(name);
 400                         break;
 401                 case Opt_disable_roll_forward:
 402                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 403                         break;
 404                 case Opt_norecovery:
 405                         /* this option mounts f2fs with ro */
 406                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 407                         if (!f2fs_readonly(sb))
 408                                 return -EINVAL;
 409                         break;
 410                 case Opt_discard:
 411                         q = bdev_get_queue(sb->s_bdev);
 412                         if (blk_queue_discard(q)) {
 413                                 set_opt(sbi, DISCARD);
 414                         } else {
 415                                 f2fs_msg(sb, KERN_WARNING,
 416                                         "mounting with \"discard\" option, but "
 417                                         "the device does not support discard");
 418                         }
 419                         break;
 420                 case Opt_noheap:
 421                         set_opt(sbi, NOHEAP);
 422                         break;
 423 #ifdef CONFIG_F2FS_FS_XATTR
 424                 case Opt_user_xattr:
 425                         set_opt(sbi, XATTR_USER);
 426                         break;
 427                 case Opt_nouser_xattr:
 428                         clear_opt(sbi, XATTR_USER);
 429                         break;
 430                 case Opt_inline_xattr:
 431                         set_opt(sbi, INLINE_XATTR);
 432                         break;
 433 #else
 434                 case Opt_user_xattr:
 435                         f2fs_msg(sb, KERN_INFO,
 436                                 "user_xattr options not supported");
 437                         break;
 438                 case Opt_nouser_xattr:
 439                         f2fs_msg(sb, KERN_INFO,
 440                                 "nouser_xattr options not supported");
 441                         break;
 442                 case Opt_inline_xattr:
 443                         f2fs_msg(sb, KERN_INFO,
 444                                 "inline_xattr options not supported");
 445                         break;
 446 #endif
 447 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 448                 case Opt_acl:
 449                         set_opt(sbi, POSIX_ACL);
 450                         break;
 451                 case Opt_noacl:
 452                         clear_opt(sbi, POSIX_ACL);
 453                         break;
 454 #else
 455                 case Opt_acl:
 456                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 457                         break;
 458                 case Opt_noacl:
 459                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 460                         break;
 461 #endif
 462                 case Opt_active_logs:
 463                         if (args->from && match_int(args, &arg))
 464                                 return -EINVAL;
 465                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 466                                 return -EINVAL;
 467                         sbi->active_logs = arg;
 468                         break;
 469                 case Opt_disable_ext_identify:
 470                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 471                         break;
 472                 case Opt_inline_data:
 473                         set_opt(sbi, INLINE_DATA);
 474                         break;
 475                 case Opt_inline_dentry:
 476                         set_opt(sbi, INLINE_DENTRY);
 477                         break;
 478                 case Opt_flush_merge:
 479                         set_opt(sbi, FLUSH_MERGE);
 480                         break;
 481                 case Opt_nobarrier:
 482                         set_opt(sbi, NOBARRIER);
 483                         break;
 484                 case Opt_fastboot:
 485                         set_opt(sbi, FASTBOOT);
 486                         break;
 487                 case Opt_extent_cache:
 488                         set_opt(sbi, EXTENT_CACHE);
 489                         break;
 490                 case Opt_noextent_cache:
 491                         clear_opt(sbi, EXTENT_CACHE);
 492                         break;
 493                 case Opt_noinline_data:
 494                         clear_opt(sbi, INLINE_DATA);
 495                         break;
 496                 case Opt_data_flush:
 497                         set_opt(sbi, DATA_FLUSH);
 498                         break;
 499                 case Opt_fault_injection:
 500                         if (args->from && match_int(args, &arg))
 501                                 return -EINVAL;
 502 #ifdef CONFIG_F2FS_FAULT_INJECTION
 503                         f2fs_build_fault_attr(arg);
 504 #else
 505                         f2fs_msg(sb, KERN_INFO,
 506                                 "FAULT_INJECTION was not selected");
 507 #endif
 508                         break;
 509                 default:
 510                         f2fs_msg(sb, KERN_ERR,
 511                                 "Unrecognized mount option \"%s\" or missing value",
 512                                 p);
 513                         return -EINVAL;
 514                 }
 515         }
 516         return 0;
 517 }
 518
 519 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 520 {
 521         struct f2fs_inode_info *fi;
 522
 523         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 524         if (!fi)
 525                 return NULL;
 526
 527         init_once((void *) fi);
 528
 529         if (percpu_counter_init(&fi->dirty_pages, 0, GFP_NOFS)) {
 530                 kmem_cache_free(f2fs_inode_cachep, fi);
 531                 return NULL;
 532         }
 533
 534         /* Initialize f2fs-specific inode info */
 535         fi->vfs_inode.i_version = 1;
 536         fi->i_current_depth = 1;
 537         fi->i_advise = 0;
 538         init_rwsem(&fi->i_sem);
 539         INIT_LIST_HEAD(&fi->dirty_list);
 540         INIT_LIST_HEAD(&fi->inmem_pages);
 541         mutex_init(&fi->inmem_lock);
 542
 543         set_inode_flag(fi, FI_NEW_INODE);
 544
 545         if (test_opt(F2FS_SB(sb), INLINE_XATTR))
 546                 set_inode_flag(fi, FI_INLINE_XATTR);
 547
 548         /* Will be used by directory only */
 549         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 550         return &fi->vfs_inode;
 551 }
 552
 553 static int f2fs_drop_inode(struct inode *inode)
 554 {
 555         /*
 556          * This is to avoid a deadlock condition like below.
 557          * writeback_single_inode(inode)
 558          *  - f2fs_write_data_page
 559          *    - f2fs_gc -> iput -> evict
 560          *       - inode_wait_for_writeback(inode)
 561          */
 562         if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
 563                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 564                         /* to avoid evict_inode call simultaneously */
 565                         atomic_inc(&inode->i_count);
 566                         spin_unlock(&inode->i_lock);
 567
 568                         /* some remained atomic pages should discarded */
 569                         if (f2fs_is_atomic_file(inode))
 570                                 drop_inmem_pages(inode);
 571
 572                         /* should remain fi->extent_tree for writepage */
 573                         f2fs_destroy_extent_node(inode);
 574
 575                         sb_start_intwrite(inode->i_sb);
 576                         i_size_write(inode, 0);
 577
 578                         if (F2FS_HAS_BLOCKS(inode))
 579                                 f2fs_truncate(inode, true);
 580
 581                         sb_end_intwrite(inode->i_sb);
 582
 583                         fscrypt_put_encryption_info(inode, NULL);
 584                         spin_lock(&inode->i_lock);
 585                         atomic_dec(&inode->i_count);
 586                 }
 587                 return 0;
 588         }
 589         return generic_drop_inode(inode);
 590 }
 591
 592 /*
 593  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 594  *
 595  * We should call set_dirty_inode to write the dirty inode through write_inode.
 596  */
 597 static void f2fs_dirty_inode(struct inode *inode, int flags)
 598 {
 599         set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
 600 }
 601
 602 static void f2fs_i_callback(struct rcu_head *head)
 603 {
 604         struct inode *inode = container_of(head, struct inode, i_rcu);
 605         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 606 }
 607
 608 static void f2fs_destroy_inode(struct inode *inode)
 609 {
 610         percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
 611         call_rcu(&inode->i_rcu, f2fs_i_callback);
 612 }
 613
 614 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 615 {
 616         int i;
 617
 618         for (i = 0; i < NR_COUNT_TYPE; i++)
 619                 percpu_counter_destroy(&sbi->nr_pages[i]);
 620         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 621         percpu_counter_destroy(&sbi->total_valid_inode_count);
 622 }
 623
 624 static void f2fs_put_super(struct super_block *sb)
 625 {
 626         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 627
 628         if (sbi->s_proc) {
 629                 remove_proc_entry("segment_info", sbi->s_proc);
 630                 remove_proc_entry("segment_bits", sbi->s_proc);
 631                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 632         }
 633         kobject_del(&sbi->s_kobj);
 634
 635         stop_gc_thread(sbi);
 636
 637         /* prevent remaining shrinker jobs */
 638         mutex_lock(&sbi->umount_mutex);
 639
 640         /*
 641          * We don't need to do checkpoint when superblock is clean.
 642          * But, the previous checkpoint was not done by umount, it needs to do
 643          * clean checkpoint again.
 644          */
 645         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 646                         !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
 647                 struct cp_control cpc = {
 648                         .reason = CP_UMOUNT,
 649                 };
 650                 write_checkpoint(sbi, &cpc);
 651         }
 652
 653         /* write_checkpoint can update stat informaion */
 654         f2fs_destroy_stats(sbi);
 655
 656         /*
 657          * normally superblock is clean, so we need to release this.
 658          * In addition, EIO will skip do checkpoint, we need this as well.
 659          */
 660         release_ino_entry(sbi, true);
 661         release_discard_addrs(sbi);
 662
 663         f2fs_leave_shrinker(sbi);
 664         mutex_unlock(&sbi->umount_mutex);
 665
 666         /* our cp_error case, we can wait for any writeback page */
 667         f2fs_flush_merged_bios(sbi);
 668
 669         iput(sbi->node_inode);
 670         iput(sbi->meta_inode);
 671
 672         /* destroy f2fs internal modules */
 673         destroy_node_manager(sbi);
 674         destroy_segment_manager(sbi);
 675
 676         kfree(sbi->ckpt);
 677         kobject_put(&sbi->s_kobj);
 678         wait_for_completion(&sbi->s_kobj_unregister);
 679
 680         sb->s_fs_info = NULL;
 681         if (sbi->s_chksum_driver)
 682                 crypto_free_shash(sbi->s_chksum_driver);
 683         kfree(sbi->raw_super);
 684
 685         destroy_percpu_info(sbi);
 686         kfree(sbi);
 687 }
 688
 689 int f2fs_sync_fs(struct super_block *sb, int sync)
 690 {
 691         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 692         int err = 0;
 693
 694         trace_f2fs_sync_fs(sb, sync);
 695
 696         if (sync) {
 697                 struct cp_control cpc;
 698
 699                 cpc.reason = __get_cp_reason(sbi);
 700
 701                 mutex_lock(&sbi->gc_mutex);
 702                 err = write_checkpoint(sbi, &cpc);
 703                 mutex_unlock(&sbi->gc_mutex);
 704         }
 705         f2fs_trace_ios(NULL, 1);
 706
 707         return err;
 708 }
 709
 710 static int f2fs_freeze(struct super_block *sb)
 711 {
 712         int err;
 713
 714         if (f2fs_readonly(sb))
 715                 return 0;
 716
 717         err = f2fs_sync_fs(sb, 1);
 718         return err;
 719 }
 720
 721 static int f2fs_unfreeze(struct super_block *sb)
 722 {
 723         return 0;
 724 }
 725
 726 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 727 {
 728         struct super_block *sb = dentry->d_sb;
 729         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 730         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 731         block_t total_count, user_block_count, start_count, ovp_count;
 732
 733         total_count = le64_to_cpu(sbi->raw_super->block_count);
 734         user_block_count = sbi->user_block_count;
 735         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 736         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 737         buf->f_type = F2FS_SUPER_MAGIC;
 738         buf->f_bsize = sbi->blocksize;
 739
 740         buf->f_blocks = total_count - start_count;
 741         buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
 742         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 743
 744         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 745         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
 746
 747         buf->f_namelen = F2FS_NAME_LEN;
 748         buf->f_fsid.val[0] = (u32)id;
 749         buf->f_fsid.val[1] = (u32)(id >> 32);
 750
 751         return 0;
 752 }
 753
 754 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 755 {
 756         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 757
 758         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
 759                 if (test_opt(sbi, FORCE_FG_GC))
 760                         seq_printf(seq, ",background_gc=%s", "sync");
 761                 else
 762                         seq_printf(seq, ",background_gc=%s", "on");
 763         } else {
 764                 seq_printf(seq, ",background_gc=%s", "off");
 765         }
 766         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 767                 seq_puts(seq, ",disable_roll_forward");
 768         if (test_opt(sbi, DISCARD))
 769                 seq_puts(seq, ",discard");
 770         if (test_opt(sbi, NOHEAP))
 771                 seq_puts(seq, ",no_heap_alloc");
 772 #ifdef CONFIG_F2FS_FS_XATTR
 773         if (test_opt(sbi, XATTR_USER))
 774                 seq_puts(seq, ",user_xattr");
 775         else
 776                 seq_puts(seq, ",nouser_xattr");
 777         if (test_opt(sbi, INLINE_XATTR))
 778                 seq_puts(seq, ",inline_xattr");
 779 #endif
 780 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 781         if (test_opt(sbi, POSIX_ACL))
 782                 seq_puts(seq, ",acl");
 783         else
 784                 seq_puts(seq, ",noacl");
 785 #endif
 786         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 787                 seq_puts(seq, ",disable_ext_identify");
 788         if (test_opt(sbi, INLINE_DATA))
 789                 seq_puts(seq, ",inline_data");
 790         else
 791                 seq_puts(seq, ",noinline_data");
 792         if (test_opt(sbi, INLINE_DENTRY))
 793                 seq_puts(seq, ",inline_dentry");
 794         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 795                 seq_puts(seq, ",flush_merge");
 796         if (test_opt(sbi, NOBARRIER))
 797                 seq_puts(seq, ",nobarrier");
 798         if (test_opt(sbi, FASTBOOT))
 799                 seq_puts(seq, ",fastboot");
 800         if (test_opt(sbi, EXTENT_CACHE))
 801                 seq_puts(seq, ",extent_cache");
 802         else
 803                 seq_puts(seq, ",noextent_cache");
 804         if (test_opt(sbi, DATA_FLUSH))
 805                 seq_puts(seq, ",data_flush");
 806         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 807
 808         return 0;
 809 }
 810
 811 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 812 {
 813         struct super_block *sb = seq->private;
 814         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 815         unsigned int total_segs =
 816                         le32_to_cpu(sbi->raw_super->segment_count_main);
 817         int i;
 818
 819         seq_puts(seq, "format: segment_type|valid_blocks\n"
 820                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 821
 822         for (i = 0; i < total_segs; i++) {
 823                 struct seg_entry *se = get_seg_entry(sbi, i);
 824
 825                 if ((i % 10) == 0)
 826                         seq_printf(seq, "%-10d", i);
 827                 seq_printf(seq, "%d|%-3u", se->type,
 828                                         get_valid_blocks(sbi, i, 1));
 829                 if ((i % 10) == 9 || i == (total_segs - 1))
 830                         seq_putc(seq, '\n');
 831                 else
 832                         seq_putc(seq, ' ');
 833         }
 834
 835         return 0;
 836 }
 837
 838 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
 839 {
 840         struct super_block *sb = seq->private;
 841         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 842         unsigned int total_segs =
 843                         le32_to_cpu(sbi->raw_super->segment_count_main);
 844         int i, j;
 845
 846         seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
 847                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 848
 849         for (i = 0; i < total_segs; i++) {
 850                 struct seg_entry *se = get_seg_entry(sbi, i);
 851
 852                 seq_printf(seq, "%-10d", i);
 853                 seq_printf(seq, "%d|%-3u|", se->type,
 854                                         get_valid_blocks(sbi, i, 1));
 855                 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
 856                         seq_printf(seq, "%x ", se->cur_valid_map[j]);
 857                 seq_putc(seq, '\n');
 858         }
 859         return 0;
 860 }
 861
 862 #define F2FS_PROC_FILE_DEF(_name)                                       \
 863 static int _name##_open_fs(struct inode *inode, struct file *file)      \
 864 {                                                                       \
 865         return single_open(file, _name##_seq_show, PDE_DATA(inode));    \
 866 }                                                                       \
 867                                                                         \
 868 static const struct file_operations f2fs_seq_##_name##_fops = {         \
 869         .open = _name##_open_fs,                                        \
 870         .read = seq_read,                                               \
 871         .llseek = seq_lseek,                                            \
 872         .release = single_release,                                      \
 873 };
 874
 875 F2FS_PROC_FILE_DEF(segment_info);
 876 F2FS_PROC_FILE_DEF(segment_bits);
 877
 878 static void default_options(struct f2fs_sb_info *sbi)
 879 {
 880         /* init some FS parameters */
 881         sbi->active_logs = NR_CURSEG_TYPE;
 882
 883         set_opt(sbi, BG_GC);
 884         set_opt(sbi, INLINE_DATA);
 885         set_opt(sbi, EXTENT_CACHE);
 886
 887 #ifdef CONFIG_F2FS_FS_XATTR
 888         set_opt(sbi, XATTR_USER);
 889 #endif
 890 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 891         set_opt(sbi, POSIX_ACL);
 892 #endif
 893 }
 894
 895 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 896 {
 897         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 898         struct f2fs_mount_info org_mount_opt;
 899         int err, active_logs;
 900         bool need_restart_gc = false;
 901         bool need_stop_gc = false;
 902         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
 903
 904         /*
 905          * Save the old mount options in case we
 906          * need to restore them.
 907          */
 908         org_mount_opt = sbi->mount_opt;
 909         active_logs = sbi->active_logs;
 910
 911         /* recover superblocks we couldn't write due to previous RO mount */
 912         if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
 913                 err = f2fs_commit_super(sbi, false);
 914                 f2fs_msg(sb, KERN_INFO,
 915                         "Try to recover all the superblocks, ret: %d", err);
 916                 if (!err)
 917                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
 918         }
 919
 920         sbi->mount_opt.opt = 0;
 921         default_options(sbi);
 922
 923         /* parse mount options */
 924         err = parse_options(sb, data);
 925         if (err)
 926                 goto restore_opts;
 927
 928         /*
 929          * Previous and new state of filesystem is RO,
 930          * so skip checking GC and FLUSH_MERGE conditions.
 931          */
 932         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
 933                 goto skip;
 934
 935         /* disallow enable/disable extent_cache dynamically */
 936         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
 937                 err = -EINVAL;
 938                 f2fs_msg(sbi->sb, KERN_WARNING,
 939                                 "switch extent_cache option is not allowed");
 940                 goto restore_opts;
 941         }
 942
 943         /*
 944          * We stop the GC thread if FS is mounted as RO
 945          * or if background_gc = off is passed in mount
 946          * option. Also sync the filesystem.
 947          */
 948         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
 949                 if (sbi->gc_thread) {
 950                         stop_gc_thread(sbi);
 951                         need_restart_gc = true;
 952                 }
 953         } else if (!sbi->gc_thread) {
 954                 err = start_gc_thread(sbi);
 955                 if (err)
 956                         goto restore_opts;
 957                 need_stop_gc = true;
 958         }
 959
 960         if (*flags & MS_RDONLY) {
 961                 writeback_inodes_sb(sb, WB_REASON_SYNC);
 962                 sync_inodes_sb(sb);
 963
 964                 set_sbi_flag(sbi, SBI_IS_DIRTY);
 965                 set_sbi_flag(sbi, SBI_IS_CLOSE);
 966                 f2fs_sync_fs(sb, 1);
 967                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
 968         }
 969
 970         /*
 971          * We stop issue flush thread if FS is mounted as RO
 972          * or if flush_merge is not passed in mount option.
 973          */
 974         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
 975                 destroy_flush_cmd_control(sbi);
 976         } else if (!SM_I(sbi)->cmd_control_info) {
 977                 err = create_flush_cmd_control(sbi);
 978                 if (err)
 979                         goto restore_gc;
 980         }
 981 skip:
 982         /* Update the POSIXACL Flag */
 983         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 984                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
 985
 986         return 0;
 987 restore_gc:
 988         if (need_restart_gc) {
 989                 if (start_gc_thread(sbi))
 990                         f2fs_msg(sbi->sb, KERN_WARNING,
 991                                 "background gc thread has stopped");
 992         } else if (need_stop_gc) {
 993                 stop_gc_thread(sbi);
 994         }
 995 restore_opts:
 996         sbi->mount_opt = org_mount_opt;
 997         sbi->active_logs = active_logs;
 998         return err;
 999 }
1000
1001 static struct super_operations f2fs_sops = {
1002         .alloc_inode    = f2fs_alloc_inode,
1003         .drop_inode     = f2fs_drop_inode,
1004         .destroy_inode  = f2fs_destroy_inode,
1005         .write_inode    = f2fs_write_inode,
1006         .dirty_inode    = f2fs_dirty_inode,
1007         .show_options   = f2fs_show_options,
1008         .evict_inode    = f2fs_evict_inode,
1009         .put_super      = f2fs_put_super,
1010         .sync_fs        = f2fs_sync_fs,
1011         .freeze_fs      = f2fs_freeze,
1012         .unfreeze_fs    = f2fs_unfreeze,
1013         .statfs         = f2fs_statfs,
1014         .remount_fs     = f2fs_remount,
1015 };
1016
1017 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1018 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1019 {
1020         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1021                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1022                                 ctx, len, NULL);
1023 }
1024
1025 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1026 {
1027         *key = F2FS_I_SB(inode)->key_prefix;
1028         return F2FS_I_SB(inode)->key_prefix_size;
1029 }
1030
1031 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1032                                                         void *fs_data)
1033 {
1034         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1035                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1036                                 ctx, len, fs_data, XATTR_CREATE);
1037 }
1038
1039 static unsigned f2fs_max_namelen(struct inode *inode)
1040 {
1041         return S_ISLNK(inode->i_mode) ?
1042                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1043 }
1044
1045 static struct fscrypt_operations f2fs_cryptops = {
1046         .get_context    = f2fs_get_context,
1047         .key_prefix     = f2fs_key_prefix,
1048         .set_context    = f2fs_set_context,
1049         .is_encrypted   = f2fs_encrypted_inode,
1050         .empty_dir      = f2fs_empty_dir,
1051         .max_namelen    = f2fs_max_namelen,
1052 };
1053 #else
1054 static struct fscrypt_operations f2fs_cryptops = {
1055         .is_encrypted   = f2fs_encrypted_inode,
1056 };
1057 #endif
1058
1059 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1060                 u64 ino, u32 generation)
1061 {
1062         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1063         struct inode *inode;
1064
1065         if (check_nid_range(sbi, ino))
1066                 return ERR_PTR(-ESTALE);
1067
1068         /*
1069          * f2fs_iget isn't quite right if the inode is currently unallocated!
1070          * However f2fs_iget currently does appropriate checks to handle stale
1071          * inodes so everything is OK.
1072          */
1073         inode = f2fs_iget(sb, ino);
1074         if (IS_ERR(inode))
1075                 return ERR_CAST(inode);
1076         if (unlikely(generation && inode->i_generation != generation)) {
1077                 /* we didn't find the right inode.. */
1078                 iput(inode);
1079                 return ERR_PTR(-ESTALE);
1080         }
1081         return inode;
1082 }
1083
1084 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1085                 int fh_len, int fh_type)
1086 {
1087         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1088                                     f2fs_nfs_get_inode);
1089 }
1090
1091 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1092                 int fh_len, int fh_type)
1093 {
1094         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1095                                     f2fs_nfs_get_inode);
1096 }
1097
1098 static const struct export_operations f2fs_export_ops = {
1099         .fh_to_dentry = f2fs_fh_to_dentry,
1100         .fh_to_parent = f2fs_fh_to_parent,
1101         .get_parent = f2fs_get_parent,
1102 };
1103
1104 static loff_t max_file_blocks(void)
1105 {
1106         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1107         loff_t leaf_count = ADDRS_PER_BLOCK;
1108
1109         /* two direct node blocks */
1110         result += (leaf_count * 2);
1111
1112         /* two indirect node blocks */
1113         leaf_count *= NIDS_PER_BLOCK;
1114         result += (leaf_count * 2);
1115
1116         /* one double indirect node block */
1117         leaf_count *= NIDS_PER_BLOCK;
1118         result += leaf_count;
1119
1120         return result;
1121 }
1122
1123 static int __f2fs_commit_super(struct buffer_head *bh,
1124                         struct f2fs_super_block *super)
1125 {
1126         lock_buffer(bh);
1127         if (super)
1128                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1129         set_buffer_uptodate(bh);
1130         set_buffer_dirty(bh);
1131         unlock_buffer(bh);
1132
1133         /* it's rare case, we can do fua all the time */
1134         return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1135 }
1136
1137 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1138                                         struct buffer_head *bh)
1139 {
1140         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1141                                         (bh->b_data + F2FS_SUPER_OFFSET);
1142         struct super_block *sb = sbi->sb;
1143         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1144         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1145         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1146         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1147         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1148         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1149         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1150         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1151         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1152         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1153         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1154         u32 segment_count = le32_to_cpu(raw_super->segment_count);
1155         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1156         u64 main_end_blkaddr = main_blkaddr +
1157                                 (segment_count_main << log_blocks_per_seg);
1158         u64 seg_end_blkaddr = segment0_blkaddr +
1159                                 (segment_count << log_blocks_per_seg);
1160
1161         if (segment0_blkaddr != cp_blkaddr) {
1162                 f2fs_msg(sb, KERN_INFO,
1163                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1164                         segment0_blkaddr, cp_blkaddr);
1165                 return true;
1166         }
1167
1168         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1169                                                         sit_blkaddr) {
1170                 f2fs_msg(sb, KERN_INFO,
1171                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1172                         cp_blkaddr, sit_blkaddr,
1173                         segment_count_ckpt << log_blocks_per_seg);
1174                 return true;
1175         }
1176
1177         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1178                                                         nat_blkaddr) {
1179                 f2fs_msg(sb, KERN_INFO,
1180                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1181                         sit_blkaddr, nat_blkaddr,
1182                         segment_count_sit << log_blocks_per_seg);
1183                 return true;
1184         }
1185
1186         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1187                                                         ssa_blkaddr) {
1188                 f2fs_msg(sb, KERN_INFO,
1189                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1190                         nat_blkaddr, ssa_blkaddr,
1191                         segment_count_nat << log_blocks_per_seg);
1192                 return true;
1193         }
1194
1195         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1196                                                         main_blkaddr) {
1197                 f2fs_msg(sb, KERN_INFO,
1198                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1199                         ssa_blkaddr, main_blkaddr,
1200                         segment_count_ssa << log_blocks_per_seg);
1201                 return true;
1202         }
1203
1204         if (main_end_blkaddr > seg_end_blkaddr) {
1205                 f2fs_msg(sb, KERN_INFO,
1206                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1207                         main_blkaddr,
1208                         segment0_blkaddr +
1209                                 (segment_count << log_blocks_per_seg),
1210                         segment_count_main << log_blocks_per_seg);
1211                 return true;
1212         } else if (main_end_blkaddr < seg_end_blkaddr) {
1213                 int err = 0;
1214                 char *res;
1215
1216                 /* fix in-memory information all the time */
1217                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1218                                 segment0_blkaddr) >> log_blocks_per_seg);
1219
1220                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1221                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1222                         res = "internally";
1223                 } else {
1224                         err = __f2fs_commit_super(bh, NULL);
1225                         res = err ? "failed" : "done";
1226                 }
1227                 f2fs_msg(sb, KERN_INFO,
1228                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
1229                         res, main_blkaddr,
1230                         segment0_blkaddr +
1231                                 (segment_count << log_blocks_per_seg),
1232                         segment_count_main << log_blocks_per_seg);
1233                 if (err)
1234                         return true;
1235         }
1236         return false;
1237 }
1238
1239 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1240                                 struct buffer_head *bh)
1241 {
1242         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1243                                         (bh->b_data + F2FS_SUPER_OFFSET);
1244         struct super_block *sb = sbi->sb;
1245         unsigned int blocksize;
1246
1247         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1248                 f2fs_msg(sb, KERN_INFO,
1249                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1250                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1251                 return 1;
1252         }
1253
1254         /* Currently, support only 4KB page cache size */
1255         if (F2FS_BLKSIZE != PAGE_SIZE) {
1256                 f2fs_msg(sb, KERN_INFO,
1257                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1258                         PAGE_SIZE);
1259                 return 1;
1260         }
1261
1262         /* Currently, support only 4KB block size */
1263         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1264         if (blocksize != F2FS_BLKSIZE) {
1265                 f2fs_msg(sb, KERN_INFO,
1266                         "Invalid blocksize (%u), supports only 4KB\n",
1267                         blocksize);
1268                 return 1;
1269         }
1270
1271         /* check log blocks per segment */
1272         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1273                 f2fs_msg(sb, KERN_INFO,
1274                         "Invalid log blocks per segment (%u)\n",
1275                         le32_to_cpu(raw_super->log_blocks_per_seg));
1276                 return 1;
1277         }
1278
1279         /* Currently, support 512/1024/2048/4096 bytes sector size */
1280         if (le32_to_cpu(raw_super->log_sectorsize) >
1281                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1282                 le32_to_cpu(raw_super->log_sectorsize) <
1283                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1284                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1285                         le32_to_cpu(raw_super->log_sectorsize));
1286                 return 1;
1287         }
1288         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1289                 le32_to_cpu(raw_super->log_sectorsize) !=
1290                         F2FS_MAX_LOG_SECTOR_SIZE) {
1291                 f2fs_msg(sb, KERN_INFO,
1292                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1293                         le32_to_cpu(raw_super->log_sectors_per_block),
1294                         le32_to_cpu(raw_super->log_sectorsize));
1295                 return 1;
1296         }
1297
1298         /* check reserved ino info */
1299         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1300                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1301                 le32_to_cpu(raw_super->root_ino) != 3) {
1302                 f2fs_msg(sb, KERN_INFO,
1303                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1304                         le32_to_cpu(raw_super->node_ino),
1305                         le32_to_cpu(raw_super->meta_ino),
1306                         le32_to_cpu(raw_super->root_ino));
1307                 return 1;
1308         }
1309
1310         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1311         if (sanity_check_area_boundary(sbi, bh))
1312                 return 1;
1313
1314         return 0;
1315 }
1316
1317 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1318 {
1319         unsigned int total, fsmeta;
1320         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1321         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1322
1323         total = le32_to_cpu(raw_super->segment_count);
1324         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1325         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1326         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1327         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1328         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1329
1330         if (unlikely(fsmeta >= total))
1331                 return 1;
1332
1333         if (unlikely(f2fs_cp_error(sbi))) {
1334                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1335                 return 1;
1336         }
1337         return 0;
1338 }
1339
1340 static void init_sb_info(struct f2fs_sb_info *sbi)
1341 {
1342         struct f2fs_super_block *raw_super = sbi->raw_super;
1343
1344         sbi->log_sectors_per_block =
1345                 le32_to_cpu(raw_super->log_sectors_per_block);
1346         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1347         sbi->blocksize = 1 << sbi->log_blocksize;
1348         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1349         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1350         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1351         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1352         sbi->total_sections = le32_to_cpu(raw_super->section_count);
1353         sbi->total_node_count =
1354                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1355                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1356         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1357         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1358         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1359         sbi->cur_victim_sec = NULL_SECNO;
1360         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1361
1362         sbi->dir_level = DEF_DIR_LEVEL;
1363         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1364         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1365         clear_sbi_flag(sbi, SBI_NEED_FSCK);
1366
1367         INIT_LIST_HEAD(&sbi->s_list);
1368         mutex_init(&sbi->umount_mutex);
1369
1370 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1371         memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1372                                 F2FS_KEY_DESC_PREFIX_SIZE);
1373         sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1374 #endif
1375 }
1376
1377 static int init_percpu_info(struct f2fs_sb_info *sbi)
1378 {
1379         int i, err;
1380
1381         for (i = 0; i < NR_COUNT_TYPE; i++) {
1382                 err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1383                 if (err)
1384                         return err;
1385         }
1386
1387         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1388         if (err)
1389                 return err;
1390
1391         return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1392                                                                 GFP_KERNEL);
1393 }
1394
1395 /*
1396  * Read f2fs raw super block.
1397  * Because we have two copies of super block, so read both of them
1398  * to get the first valid one. If any one of them is broken, we pass
1399  * them recovery flag back to the caller.
1400  */
1401 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1402                         struct f2fs_super_block **raw_super,
1403                         int *valid_super_block, int *recovery)
1404 {
1405         struct super_block *sb = sbi->sb;
1406         int block;
1407         struct buffer_head *bh;
1408         struct f2fs_super_block *super;
1409         int err = 0;
1410
1411         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1412         if (!super)
1413                 return -ENOMEM;
1414
1415         for (block = 0; block < 2; block++) {
1416                 bh = sb_bread(sb, block);
1417                 if (!bh) {
1418                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1419                                 block + 1);
1420                         err = -EIO;
1421                         continue;
1422                 }
1423
1424                 /* sanity checking of raw super */
1425                 if (sanity_check_raw_super(sbi, bh)) {
1426                         f2fs_msg(sb, KERN_ERR,
1427                                 "Can't find valid F2FS filesystem in %dth superblock",
1428                                 block + 1);
1429                         err = -EINVAL;
1430                         brelse(bh);
1431                         continue;
1432                 }
1433
1434                 if (!*raw_super) {
1435                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1436                                                         sizeof(*super));
1437                         *valid_super_block = block;
1438                         *raw_super = super;
1439                 }
1440                 brelse(bh);
1441         }
1442
1443         /* Fail to read any one of the superblocks*/
1444         if (err < 0)
1445                 *recovery = 1;
1446
1447         /* No valid superblock */
1448         if (!*raw_super)
1449                 kfree(super);
1450         else
1451                 err = 0;
1452
1453         return err;
1454 }
1455
1456 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1457 {
1458         struct buffer_head *bh;
1459         int err;
1460
1461         if ((recover && f2fs_readonly(sbi->sb)) ||
1462                                 bdev_read_only(sbi->sb->s_bdev)) {
1463                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1464                 return -EROFS;
1465         }
1466
1467         /* write back-up superblock first */
1468         bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1469         if (!bh)
1470                 return -EIO;
1471         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1472         brelse(bh);
1473
1474         /* if we are in recovery path, skip writing valid superblock */
1475         if (recover || err)
1476                 return err;
1477
1478         /* write current valid superblock */
1479         bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1480         if (!bh)
1481                 return -EIO;
1482         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1483         brelse(bh);
1484         return err;
1485 }
1486
1487 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1488 {
1489         struct f2fs_sb_info *sbi;
1490         struct f2fs_super_block *raw_super;
1491         struct inode *root;
1492         int err;
1493         bool retry = true, need_fsck = false;
1494         char *options = NULL;
1495         int recovery, i, valid_super_block;
1496         struct curseg_info *seg_i;
1497
1498 try_onemore:
1499         err = -EINVAL;
1500         raw_super = NULL;
1501         valid_super_block = -1;
1502         recovery = 0;
1503
1504         /* allocate memory for f2fs-specific super block info */
1505         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1506         if (!sbi)
1507                 return -ENOMEM;
1508
1509         sbi->sb = sb;
1510
1511         /* Load the checksum driver */
1512         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1513         if (IS_ERR(sbi->s_chksum_driver)) {
1514                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1515                 err = PTR_ERR(sbi->s_chksum_driver);
1516                 sbi->s_chksum_driver = NULL;
1517                 goto free_sbi;
1518         }
1519
1520         /* set a block size */
1521         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1522                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1523                 goto free_sbi;
1524         }
1525
1526         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1527                                                                 &recovery);
1528         if (err)
1529                 goto free_sbi;
1530
1531         sb->s_fs_info = sbi;
1532         default_options(sbi);
1533         /* parse mount options */
1534         options = kstrdup((const char *)data, GFP_KERNEL);
1535         if (data && !options) {
1536                 err = -ENOMEM;
1537                 goto free_sb_buf;
1538         }
1539
1540         err = parse_options(sb, options);
1541         if (err)
1542                 goto free_options;
1543
1544         sbi->max_file_blocks = max_file_blocks();
1545         sb->s_maxbytes = sbi->max_file_blocks <<
1546                                 le32_to_cpu(raw_super->log_blocksize);
1547         sb->s_max_links = F2FS_LINK_MAX;
1548         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1549
1550         sb->s_op = &f2fs_sops;
1551         sb->s_cop = &f2fs_cryptops;
1552         sb->s_xattr = f2fs_xattr_handlers;
1553         sb->s_export_op = &f2fs_export_ops;
1554         sb->s_magic = F2FS_SUPER_MAGIC;
1555         sb->s_time_gran = 1;
1556         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1557                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1558         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1559
1560         /* init f2fs-specific super block info */
1561         sbi->raw_super = raw_super;
1562         sbi->valid_super_block = valid_super_block;
1563         mutex_init(&sbi->gc_mutex);
1564         mutex_init(&sbi->writepages);
1565         mutex_init(&sbi->cp_mutex);
1566         init_rwsem(&sbi->node_write);
1567
1568         /* disallow all the data/node/meta page writes */
1569         set_sbi_flag(sbi, SBI_POR_DOING);
1570         spin_lock_init(&sbi->stat_lock);
1571
1572         init_rwsem(&sbi->read_io.io_rwsem);
1573         sbi->read_io.sbi = sbi;
1574         sbi->read_io.bio = NULL;
1575         for (i = 0; i < NR_PAGE_TYPE; i++) {
1576                 init_rwsem(&sbi->write_io[i].io_rwsem);
1577                 sbi->write_io[i].sbi = sbi;
1578                 sbi->write_io[i].bio = NULL;
1579         }
1580
1581         init_rwsem(&sbi->cp_rwsem);
1582         init_waitqueue_head(&sbi->cp_wait);
1583         init_sb_info(sbi);
1584
1585         err = init_percpu_info(sbi);
1586         if (err)
1587                 goto free_options;
1588
1589         /* get an inode for meta space */
1590         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1591         if (IS_ERR(sbi->meta_inode)) {
1592                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1593                 err = PTR_ERR(sbi->meta_inode);
1594                 goto free_options;
1595         }
1596
1597         err = get_valid_checkpoint(sbi);
1598         if (err) {
1599                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1600                 goto free_meta_inode;
1601         }
1602
1603         sbi->total_valid_node_count =
1604                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1605         percpu_counter_set(&sbi->total_valid_inode_count,
1606                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
1607         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1608         sbi->total_valid_block_count =
1609                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1610         sbi->last_valid_block_count = sbi->total_valid_block_count;
1611
1612         for (i = 0; i < NR_INODE_TYPE; i++) {
1613                 INIT_LIST_HEAD(&sbi->inode_list[i]);
1614                 spin_lock_init(&sbi->inode_lock[i]);
1615         }
1616
1617         init_extent_cache_info(sbi);
1618
1619         init_ino_entry_info(sbi);
1620
1621         /* setup f2fs internal modules */
1622         err = build_segment_manager(sbi);
1623         if (err) {
1624                 f2fs_msg(sb, KERN_ERR,
1625                         "Failed to initialize F2FS segment manager");
1626                 goto free_sm;
1627         }
1628         err = build_node_manager(sbi);
1629         if (err) {
1630                 f2fs_msg(sb, KERN_ERR,
1631                         "Failed to initialize F2FS node manager");
1632                 goto free_nm;
1633         }
1634
1635         /* For write statistics */
1636         if (sb->s_bdev->bd_part)
1637                 sbi->sectors_written_start =
1638                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1639
1640         /* Read accumulated write IO statistics if exists */
1641         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1642         if (__exist_node_summaries(sbi))
1643                 sbi->kbytes_written =
1644                         le64_to_cpu(seg_i->journal->info.kbytes_written);
1645
1646         build_gc_manager(sbi);
1647
1648         /* get an inode for node space */
1649         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1650         if (IS_ERR(sbi->node_inode)) {
1651                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1652                 err = PTR_ERR(sbi->node_inode);
1653                 goto free_nm;
1654         }
1655
1656         f2fs_join_shrinker(sbi);
1657
1658         /* if there are nt orphan nodes free them */
1659         err = recover_orphan_inodes(sbi);
1660         if (err)
1661                 goto free_node_inode;
1662
1663         /* read root inode and dentry */
1664         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1665         if (IS_ERR(root)) {
1666                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1667                 err = PTR_ERR(root);
1668                 goto free_node_inode;
1669         }
1670         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1671                 iput(root);
1672                 err = -EINVAL;
1673                 goto free_node_inode;
1674         }
1675
1676         sb->s_root = d_make_root(root); /* allocate root dentry */
1677         if (!sb->s_root) {
1678                 err = -ENOMEM;
1679                 goto free_root_inode;
1680         }
1681
1682         err = f2fs_build_stats(sbi);
1683         if (err)
1684                 goto free_root_inode;
1685
1686         if (f2fs_proc_root)
1687                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1688
1689         if (sbi->s_proc) {
1690                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1691                                  &f2fs_seq_segment_info_fops, sb);
1692                 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1693                                  &f2fs_seq_segment_bits_fops, sb);
1694         }
1695
1696         sbi->s_kobj.kset = f2fs_kset;
1697         init_completion(&sbi->s_kobj_unregister);
1698         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1699                                                         "%s", sb->s_id);
1700         if (err)
1701                 goto free_proc;
1702
1703         /* recover fsynced data */
1704         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1705                 /*
1706                  * mount should be failed, when device has readonly mode, and
1707                  * previous checkpoint was not done by clean system shutdown.
1708                  */
1709                 if (bdev_read_only(sb->s_bdev) &&
1710                                 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1711                         err = -EROFS;
1712                         goto free_kobj;
1713                 }
1714
1715                 if (need_fsck)
1716                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1717
1718                 err = recover_fsync_data(sbi, false);
1719                 if (err < 0) {
1720                         need_fsck = true;
1721                         f2fs_msg(sb, KERN_ERR,
1722                                 "Cannot recover all fsync data errno=%d", err);
1723                         goto free_kobj;
1724                 }
1725         } else {
1726                 err = recover_fsync_data(sbi, true);
1727
1728                 if (!f2fs_readonly(sb) && err > 0) {
1729                         err = -EINVAL;
1730                         f2fs_msg(sb, KERN_ERR,
1731                                 "Need to recover fsync data");
1732                         goto free_kobj;
1733                 }
1734         }
1735
1736         /* recover_fsync_data() cleared this already */
1737         clear_sbi_flag(sbi, SBI_POR_DOING);
1738
1739         /*
1740          * If filesystem is not mounted as read-only then
1741          * do start the gc_thread.
1742          */
1743         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1744                 /* After POR, we can run background GC thread.*/
1745                 err = start_gc_thread(sbi);
1746                 if (err)
1747                         goto free_kobj;
1748         }
1749         kfree(options);
1750
1751         /* recover broken superblock */
1752         if (recovery) {
1753                 err = f2fs_commit_super(sbi, true);
1754                 f2fs_msg(sb, KERN_INFO,
1755                         "Try to recover %dth superblock, ret: %d",
1756                         sbi->valid_super_block ? 1 : 2, err);
1757         }
1758
1759         f2fs_update_time(sbi, CP_TIME);
1760         f2fs_update_time(sbi, REQ_TIME);
1761         return 0;
1762
1763 free_kobj:
1764         kobject_del(&sbi->s_kobj);
1765         kobject_put(&sbi->s_kobj);
1766         wait_for_completion(&sbi->s_kobj_unregister);
1767 free_proc:
1768         if (sbi->s_proc) {
1769                 remove_proc_entry("segment_info", sbi->s_proc);
1770                 remove_proc_entry("segment_bits", sbi->s_proc);
1771                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1772         }
1773         f2fs_destroy_stats(sbi);
1774 free_root_inode:
1775         dput(sb->s_root);
1776         sb->s_root = NULL;
1777 free_node_inode:
1778         mutex_lock(&sbi->umount_mutex);
1779         f2fs_leave_shrinker(sbi);
1780         iput(sbi->node_inode);
1781         mutex_unlock(&sbi->umount_mutex);
1782 free_nm:
1783         destroy_node_manager(sbi);
1784 free_sm:
1785         destroy_segment_manager(sbi);
1786         kfree(sbi->ckpt);
1787 free_meta_inode:
1788         make_bad_inode(sbi->meta_inode);
1789         iput(sbi->meta_inode);
1790 free_options:
1791         destroy_percpu_info(sbi);
1792         kfree(options);
1793 free_sb_buf:
1794         kfree(raw_super);
1795 free_sbi:
1796         if (sbi->s_chksum_driver)
1797                 crypto_free_shash(sbi->s_chksum_driver);
1798         kfree(sbi);
1799
1800         /* give only one another chance */
1801         if (retry) {
1802                 retry = false;
1803                 shrink_dcache_sb(sb);
1804                 goto try_onemore;
1805         }
1806         return err;
1807 }
1808
1809 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1810                         const char *dev_name, void *data)
1811 {
1812         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1813 }
1814
1815 static void kill_f2fs_super(struct super_block *sb)
1816 {
1817         if (sb->s_root)
1818                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1819         kill_block_super(sb);
1820 }
1821
1822 static struct file_system_type f2fs_fs_type = {
1823         .owner          = THIS_MODULE,
1824         .name           = "f2fs",
1825         .mount          = f2fs_mount,
1826         .kill_sb        = kill_f2fs_super,
1827         .fs_flags       = FS_REQUIRES_DEV,
1828 };
1829 MODULE_ALIAS_FS("f2fs");
1830
1831 static int __init init_inodecache(void)
1832 {
1833         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1834                         sizeof(struct f2fs_inode_info), 0,
1835                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1836         if (!f2fs_inode_cachep)
1837                 return -ENOMEM;
1838         return 0;
1839 }
1840
1841 static void destroy_inodecache(void)
1842 {
1843         /*
1844          * Make sure all delayed rcu free inodes are flushed before we
1845          * destroy cache.
1846          */
1847         rcu_barrier();
1848         kmem_cache_destroy(f2fs_inode_cachep);
1849 }
1850
1851 static int __init init_f2fs_fs(void)
1852 {
1853         int err;
1854
1855         f2fs_build_trace_ios();
1856
1857         err = init_inodecache();
1858         if (err)
1859                 goto fail;
1860         err = create_node_manager_caches();
1861         if (err)
1862                 goto free_inodecache;
1863         err = create_segment_manager_caches();
1864         if (err)
1865                 goto free_node_manager_caches;
1866         err = create_checkpoint_caches();
1867         if (err)
1868                 goto free_segment_manager_caches;
1869         err = create_extent_cache();
1870         if (err)
1871                 goto free_checkpoint_caches;
1872         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1873         if (!f2fs_kset) {
1874                 err = -ENOMEM;
1875                 goto free_extent_cache;
1876         }
1877 #ifdef CONFIG_F2FS_FAULT_INJECTION
1878         f2fs_fault_inject.kset = f2fs_kset;
1879         f2fs_build_fault_attr(0);
1880         err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1881                                 NULL, "fault_injection");
1882         if (err) {
1883                 f2fs_fault_inject.kset = NULL;
1884                 goto free_kset;
1885         }
1886 #endif
1887         err = register_shrinker(&f2fs_shrinker_info);
1888         if (err)
1889                 goto free_kset;
1890
1891         err = register_filesystem(&f2fs_fs_type);
1892         if (err)
1893                 goto free_shrinker;
1894         err = f2fs_create_root_stats();
1895         if (err)
1896                 goto free_filesystem;
1897         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1898         return 0;
1899
1900 free_filesystem:
1901         unregister_filesystem(&f2fs_fs_type);
1902 free_shrinker:
1903         unregister_shrinker(&f2fs_shrinker_info);
1904 free_kset:
1905 #ifdef CONFIG_F2FS_FAULT_INJECTION
1906         if (f2fs_fault_inject.kset)
1907                 kobject_put(&f2fs_fault_inject);
1908 #endif
1909         kset_unregister(f2fs_kset);
1910 free_extent_cache:
1911         destroy_extent_cache();
1912 free_checkpoint_caches:
1913         destroy_checkpoint_caches();
1914 free_segment_manager_caches:
1915         destroy_segment_manager_caches();
1916 free_node_manager_caches:
1917         destroy_node_manager_caches();
1918 free_inodecache:
1919         destroy_inodecache();
1920 fail:
1921         return err;
1922 }
1923
1924 static void __exit exit_f2fs_fs(void)
1925 {
1926         remove_proc_entry("fs/f2fs", NULL);
1927         f2fs_destroy_root_stats();
1928         unregister_filesystem(&f2fs_fs_type);
1929         unregister_shrinker(&f2fs_shrinker_info);
1930 #ifdef CONFIG_F2FS_FAULT_INJECTION
1931         kobject_put(&f2fs_fault_inject);
1932 #endif
1933         kset_unregister(f2fs_kset);
1934         destroy_extent_cache();
1935         destroy_checkpoint_caches();
1936         destroy_segment_manager_caches();
1937         destroy_node_manager_caches();
1938         destroy_inodecache();
1939         f2fs_destroy_trace_ios();
1940 }
1941
1942 module_init(init_f2fs_fs)
1943 module_exit(exit_f2fs_fs)
1944
1945 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1946 MODULE_DESCRIPTION("Flash Friendly File System");
1947 MODULE_LICENSE("GPL");