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
  44 char *fault_name[FAULT_MAX] = {
  45         [FAULT_KMALLOC]         = "kmalloc",
  46         [FAULT_PAGE_ALLOC]      = "page alloc",
  47         [FAULT_ALLOC_NID]       = "alloc nid",
  48         [FAULT_ORPHAN]          = "orphan",
  49         [FAULT_BLOCK]           = "no more block",
  50         [FAULT_DIR_DEPTH]       = "too big dir depth",
  51         [FAULT_EVICT_INODE]     = "evict_inode fail",
  52         [FAULT_IO]              = "IO error",
  53         [FAULT_CHECKPOINT]      = "checkpoint error",
  54 };
  55
  56 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
  57                                                 unsigned int rate)
  58 {
  59         struct f2fs_fault_info *ffi = &sbi->fault_info;
  60
  61         if (rate) {
  62                 atomic_set(&ffi->inject_ops, 0);
  63                 ffi->inject_rate = rate;
  64                 ffi->inject_type = (1 << FAULT_MAX) - 1;
  65         } else {
  66                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  67         }
  68 }
  69 #endif
  70
  71 /* f2fs-wide shrinker description */
  72 static struct shrinker f2fs_shrinker_info = {
  73         .scan_objects = f2fs_shrink_scan,
  74         .count_objects = f2fs_shrink_count,
  75         .seeks = DEFAULT_SEEKS,
  76 };
  77
  78 enum {
  79         Opt_gc_background,
  80         Opt_disable_roll_forward,
  81         Opt_norecovery,
  82         Opt_discard,
  83         Opt_nodiscard,
  84         Opt_noheap,
  85         Opt_user_xattr,
  86         Opt_nouser_xattr,
  87         Opt_acl,
  88         Opt_noacl,
  89         Opt_active_logs,
  90         Opt_disable_ext_identify,
  91         Opt_inline_xattr,
  92         Opt_inline_data,
  93         Opt_inline_dentry,
  94         Opt_noinline_dentry,
  95         Opt_flush_merge,
  96         Opt_noflush_merge,
  97         Opt_nobarrier,
  98         Opt_fastboot,
  99         Opt_extent_cache,
 100         Opt_noextent_cache,
 101         Opt_noinline_data,
 102         Opt_data_flush,
 103         Opt_mode,
 104         Opt_fault_injection,
 105         Opt_lazytime,
 106         Opt_nolazytime,
 107         Opt_err,
 108 };
 109
 110 static match_table_t f2fs_tokens = {
 111         {Opt_gc_background, "background_gc=%s"},
 112         {Opt_disable_roll_forward, "disable_roll_forward"},
 113         {Opt_norecovery, "norecovery"},
 114         {Opt_discard, "discard"},
 115         {Opt_nodiscard, "nodiscard"},
 116         {Opt_noheap, "no_heap"},
 117         {Opt_user_xattr, "user_xattr"},
 118         {Opt_nouser_xattr, "nouser_xattr"},
 119         {Opt_acl, "acl"},
 120         {Opt_noacl, "noacl"},
 121         {Opt_active_logs, "active_logs=%u"},
 122         {Opt_disable_ext_identify, "disable_ext_identify"},
 123         {Opt_inline_xattr, "inline_xattr"},
 124         {Opt_inline_data, "inline_data"},
 125         {Opt_inline_dentry, "inline_dentry"},
 126         {Opt_noinline_dentry, "noinline_dentry"},
 127         {Opt_flush_merge, "flush_merge"},
 128         {Opt_noflush_merge, "noflush_merge"},
 129         {Opt_nobarrier, "nobarrier"},
 130         {Opt_fastboot, "fastboot"},
 131         {Opt_extent_cache, "extent_cache"},
 132         {Opt_noextent_cache, "noextent_cache"},
 133         {Opt_noinline_data, "noinline_data"},
 134         {Opt_data_flush, "data_flush"},
 135         {Opt_mode, "mode=%s"},
 136         {Opt_fault_injection, "fault_injection=%u"},
 137         {Opt_lazytime, "lazytime"},
 138         {Opt_nolazytime, "nolazytime"},
 139         {Opt_err, NULL},
 140 };
 141
 142 /* Sysfs support for f2fs */
 143 enum {
 144         GC_THREAD,      /* struct f2fs_gc_thread */
 145         SM_INFO,        /* struct f2fs_sm_info */
 146         NM_INFO,        /* struct f2fs_nm_info */
 147         F2FS_SBI,       /* struct f2fs_sb_info */
 148 #ifdef CONFIG_F2FS_FAULT_INJECTION
 149         FAULT_INFO_RATE,        /* struct f2fs_fault_info */
 150         FAULT_INFO_TYPE,        /* struct f2fs_fault_info */
 151 #endif
 152 };
 153
 154 struct f2fs_attr {
 155         struct attribute attr;
 156         ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
 157         ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
 158                          const char *, size_t);
 159         int struct_type;
 160         int offset;
 161 };
 162
 163 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
 164 {
 165         if (struct_type == GC_THREAD)
 166                 return (unsigned char *)sbi->gc_thread;
 167         else if (struct_type == SM_INFO)
 168                 return (unsigned char *)SM_I(sbi);
 169         else if (struct_type == NM_INFO)
 170                 return (unsigned char *)NM_I(sbi);
 171         else if (struct_type == F2FS_SBI)
 172                 return (unsigned char *)sbi;
 173 #ifdef CONFIG_F2FS_FAULT_INJECTION
 174         else if (struct_type == FAULT_INFO_RATE ||
 175                                         struct_type == FAULT_INFO_TYPE)
 176                 return (unsigned char *)&sbi->fault_info;
 177 #endif
 178         return NULL;
 179 }
 180
 181 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
 182                 struct f2fs_sb_info *sbi, char *buf)
 183 {
 184         struct super_block *sb = sbi->sb;
 185
 186         if (!sb->s_bdev->bd_part)
 187                 return snprintf(buf, PAGE_SIZE, "0\n");
 188
 189         return snprintf(buf, PAGE_SIZE, "%llu\n",
 190                 (unsigned long long)(sbi->kbytes_written +
 191                         BD_PART_WRITTEN(sbi)));
 192 }
 193
 194 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
 195                         struct f2fs_sb_info *sbi, char *buf)
 196 {
 197         unsigned char *ptr = NULL;
 198         unsigned int *ui;
 199
 200         ptr = __struct_ptr(sbi, a->struct_type);
 201         if (!ptr)
 202                 return -EINVAL;
 203
 204         ui = (unsigned int *)(ptr + a->offset);
 205
 206         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
 207 }
 208
 209 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
 210                         struct f2fs_sb_info *sbi,
 211                         const char *buf, size_t count)
 212 {
 213         unsigned char *ptr;
 214         unsigned long t;
 215         unsigned int *ui;
 216         ssize_t ret;
 217
 218         ptr = __struct_ptr(sbi, a->struct_type);
 219         if (!ptr)
 220                 return -EINVAL;
 221
 222         ui = (unsigned int *)(ptr + a->offset);
 223
 224         ret = kstrtoul(skip_spaces(buf), 0, &t);
 225         if (ret < 0)
 226                 return ret;
 227 #ifdef CONFIG_F2FS_FAULT_INJECTION
 228         if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
 229                 return -EINVAL;
 230 #endif
 231         *ui = t;
 232         return count;
 233 }
 234
 235 static ssize_t f2fs_attr_show(struct kobject *kobj,
 236                                 struct attribute *attr, char *buf)
 237 {
 238         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 239                                                                 s_kobj);
 240         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 241
 242         return a->show ? a->show(a, sbi, buf) : 0;
 243 }
 244
 245 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
 246                                                 const char *buf, size_t len)
 247 {
 248         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 249                                                                         s_kobj);
 250         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
 251
 252         return a->store ? a->store(a, sbi, buf, len) : 0;
 253 }
 254
 255 static void f2fs_sb_release(struct kobject *kobj)
 256 {
 257         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
 258                                                                 s_kobj);
 259         complete(&sbi->s_kobj_unregister);
 260 }
 261
 262 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
 263 static struct f2fs_attr f2fs_attr_##_name = {                   \
 264         .attr = {.name = __stringify(_name), .mode = _mode },   \
 265         .show   = _show,                                        \
 266         .store  = _store,                                       \
 267         .struct_type = _struct_type,                            \
 268         .offset = _offset                                       \
 269 }
 270
 271 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)    \
 272         F2FS_ATTR_OFFSET(struct_type, name, 0644,               \
 273                 f2fs_sbi_show, f2fs_sbi_store,                  \
 274                 offsetof(struct struct_name, elname))
 275
 276 #define F2FS_GENERAL_RO_ATTR(name) \
 277 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
 278
 279 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
 280 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
 281 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
 282 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
 283 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
 284 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
 285 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
 286 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
 287 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
 288 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
 289 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
 290 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
 291 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
 292 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
 293 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
 294 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
 295 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
 296 #ifdef CONFIG_F2FS_FAULT_INJECTION
 297 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
 298 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
 299 #endif
 300 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
 301
 302 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 303 static struct attribute *f2fs_attrs[] = {
 304         ATTR_LIST(gc_min_sleep_time),
 305         ATTR_LIST(gc_max_sleep_time),
 306         ATTR_LIST(gc_no_gc_sleep_time),
 307         ATTR_LIST(gc_idle),
 308         ATTR_LIST(reclaim_segments),
 309         ATTR_LIST(max_small_discards),
 310         ATTR_LIST(batched_trim_sections),
 311         ATTR_LIST(ipu_policy),
 312         ATTR_LIST(min_ipu_util),
 313         ATTR_LIST(min_fsync_blocks),
 314         ATTR_LIST(max_victim_search),
 315         ATTR_LIST(dir_level),
 316         ATTR_LIST(ram_thresh),
 317         ATTR_LIST(ra_nid_pages),
 318         ATTR_LIST(dirty_nats_ratio),
 319         ATTR_LIST(cp_interval),
 320         ATTR_LIST(idle_interval),
 321 #ifdef CONFIG_F2FS_FAULT_INJECTION
 322         ATTR_LIST(inject_rate),
 323         ATTR_LIST(inject_type),
 324 #endif
 325         ATTR_LIST(lifetime_write_kbytes),
 326         NULL,
 327 };
 328
 329 static const struct sysfs_ops f2fs_attr_ops = {
 330         .show   = f2fs_attr_show,
 331         .store  = f2fs_attr_store,
 332 };
 333
 334 static struct kobj_type f2fs_ktype = {
 335         .default_attrs  = f2fs_attrs,
 336         .sysfs_ops      = &f2fs_attr_ops,
 337         .release        = f2fs_sb_release,
 338 };
 339
 340 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 341 {
 342         struct va_format vaf;
 343         va_list args;
 344
 345         va_start(args, fmt);
 346         vaf.fmt = fmt;
 347         vaf.va = &args;
 348         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 349         va_end(args);
 350 }
 351
 352 static void init_once(void *foo)
 353 {
 354         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 355
 356         inode_init_once(&fi->vfs_inode);
 357 }
 358
 359 static int parse_options(struct super_block *sb, char *options)
 360 {
 361         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 362         struct request_queue *q;
 363         substring_t args[MAX_OPT_ARGS];
 364         char *p, *name;
 365         int arg = 0;
 366
 367         if (!options)
 368                 return 0;
 369
 370         while ((p = strsep(&options, ",")) != NULL) {
 371                 int token;
 372                 if (!*p)
 373                         continue;
 374                 /*
 375                  * Initialize args struct so we know whether arg was
 376                  * found; some options take optional arguments.
 377                  */
 378                 args[0].to = args[0].from = NULL;
 379                 token = match_token(p, f2fs_tokens, args);
 380
 381                 switch (token) {
 382                 case Opt_gc_background:
 383                         name = match_strdup(&args[0]);
 384
 385                         if (!name)
 386                                 return -ENOMEM;
 387                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 388                                 set_opt(sbi, BG_GC);
 389                                 clear_opt(sbi, FORCE_FG_GC);
 390                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 391                                 clear_opt(sbi, BG_GC);
 392                                 clear_opt(sbi, FORCE_FG_GC);
 393                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 394                                 set_opt(sbi, BG_GC);
 395                                 set_opt(sbi, FORCE_FG_GC);
 396                         } else {
 397                                 kfree(name);
 398                                 return -EINVAL;
 399                         }
 400                         kfree(name);
 401                         break;
 402                 case Opt_disable_roll_forward:
 403                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 404                         break;
 405                 case Opt_norecovery:
 406                         /* this option mounts f2fs with ro */
 407                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 408                         if (!f2fs_readonly(sb))
 409                                 return -EINVAL;
 410                         break;
 411                 case Opt_discard:
 412                         q = bdev_get_queue(sb->s_bdev);
 413                         if (blk_queue_discard(q)) {
 414                                 set_opt(sbi, DISCARD);
 415                         } else {
 416                                 f2fs_msg(sb, KERN_WARNING,
 417                                         "mounting with \"discard\" option, but "
 418                                         "the device does not support discard");
 419                         }
 420                         break;
 421                 case Opt_nodiscard:
 422                         clear_opt(sbi, DISCARD);
 423                 case Opt_noheap:
 424                         set_opt(sbi, NOHEAP);
 425                         break;
 426 #ifdef CONFIG_F2FS_FS_XATTR
 427                 case Opt_user_xattr:
 428                         set_opt(sbi, XATTR_USER);
 429                         break;
 430                 case Opt_nouser_xattr:
 431                         clear_opt(sbi, XATTR_USER);
 432                         break;
 433                 case Opt_inline_xattr:
 434                         set_opt(sbi, INLINE_XATTR);
 435                         break;
 436 #else
 437                 case Opt_user_xattr:
 438                         f2fs_msg(sb, KERN_INFO,
 439                                 "user_xattr options not supported");
 440                         break;
 441                 case Opt_nouser_xattr:
 442                         f2fs_msg(sb, KERN_INFO,
 443                                 "nouser_xattr options not supported");
 444                         break;
 445                 case Opt_inline_xattr:
 446                         f2fs_msg(sb, KERN_INFO,
 447                                 "inline_xattr options not supported");
 448                         break;
 449 #endif
 450 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 451                 case Opt_acl:
 452                         set_opt(sbi, POSIX_ACL);
 453                         break;
 454                 case Opt_noacl:
 455                         clear_opt(sbi, POSIX_ACL);
 456                         break;
 457 #else
 458                 case Opt_acl:
 459                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 460                         break;
 461                 case Opt_noacl:
 462                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 463                         break;
 464 #endif
 465                 case Opt_active_logs:
 466                         if (args->from && match_int(args, &arg))
 467                                 return -EINVAL;
 468                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 469                                 return -EINVAL;
 470                         sbi->active_logs = arg;
 471                         break;
 472                 case Opt_disable_ext_identify:
 473                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 474                         break;
 475                 case Opt_inline_data:
 476                         set_opt(sbi, INLINE_DATA);
 477                         break;
 478                 case Opt_inline_dentry:
 479                         set_opt(sbi, INLINE_DENTRY);
 480                         break;
 481                 case Opt_noinline_dentry:
 482                         clear_opt(sbi, INLINE_DENTRY);
 483                         break;
 484                 case Opt_flush_merge:
 485                         set_opt(sbi, FLUSH_MERGE);
 486                         break;
 487                 case Opt_noflush_merge:
 488                         clear_opt(sbi, FLUSH_MERGE);
 489                         break;
 490                 case Opt_nobarrier:
 491                         set_opt(sbi, NOBARRIER);
 492                         break;
 493                 case Opt_fastboot:
 494                         set_opt(sbi, FASTBOOT);
 495                         break;
 496                 case Opt_extent_cache:
 497                         set_opt(sbi, EXTENT_CACHE);
 498                         break;
 499                 case Opt_noextent_cache:
 500                         clear_opt(sbi, EXTENT_CACHE);
 501                         break;
 502                 case Opt_noinline_data:
 503                         clear_opt(sbi, INLINE_DATA);
 504                         break;
 505                 case Opt_data_flush:
 506                         set_opt(sbi, DATA_FLUSH);
 507                         break;
 508                 case Opt_mode:
 509                         name = match_strdup(&args[0]);
 510
 511                         if (!name)
 512                                 return -ENOMEM;
 513                         if (strlen(name) == 8 &&
 514                                         !strncmp(name, "adaptive", 8)) {
 515                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 516                         } else if (strlen(name) == 3 &&
 517                                         !strncmp(name, "lfs", 3)) {
 518                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 519                         } else {
 520                                 kfree(name);
 521                                 return -EINVAL;
 522                         }
 523                         kfree(name);
 524                         break;
 525                 case Opt_fault_injection:
 526                         if (args->from && match_int(args, &arg))
 527                                 return -EINVAL;
 528 #ifdef CONFIG_F2FS_FAULT_INJECTION
 529                         f2fs_build_fault_attr(sbi, arg);
 530 #else
 531                         f2fs_msg(sb, KERN_INFO,
 532                                 "FAULT_INJECTION was not selected");
 533 #endif
 534                         break;
 535                 case Opt_lazytime:
 536                         sb->s_flags |= MS_LAZYTIME;
 537                         break;
 538                 case Opt_nolazytime:
 539                         sb->s_flags &= ~MS_LAZYTIME;
 540                         break;
 541                 default:
 542                         f2fs_msg(sb, KERN_ERR,
 543                                 "Unrecognized mount option \"%s\" or missing value",
 544                                 p);
 545                         return -EINVAL;
 546                 }
 547         }
 548         return 0;
 549 }
 550
 551 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 552 {
 553         struct f2fs_inode_info *fi;
 554
 555         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 556         if (!fi)
 557                 return NULL;
 558
 559         init_once((void *) fi);
 560
 561         /* Initialize f2fs-specific inode info */
 562         fi->vfs_inode.i_version = 1;
 563         atomic_set(&fi->dirty_pages, 0);
 564         fi->i_current_depth = 1;
 565         fi->i_advise = 0;
 566         init_rwsem(&fi->i_sem);
 567         INIT_LIST_HEAD(&fi->dirty_list);
 568         INIT_LIST_HEAD(&fi->gdirty_list);
 569         INIT_LIST_HEAD(&fi->inmem_pages);
 570         mutex_init(&fi->inmem_lock);
 571         init_rwsem(&fi->dio_rwsem[READ]);
 572         init_rwsem(&fi->dio_rwsem[WRITE]);
 573
 574         /* Will be used by directory only */
 575         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 576         return &fi->vfs_inode;
 577 }
 578
 579 static int f2fs_drop_inode(struct inode *inode)
 580 {
 581         /*
 582          * This is to avoid a deadlock condition like below.
 583          * writeback_single_inode(inode)
 584          *  - f2fs_write_data_page
 585          *    - f2fs_gc -> iput -> evict
 586          *       - inode_wait_for_writeback(inode)
 587          */
 588         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 589                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 590                         /* to avoid evict_inode call simultaneously */
 591                         atomic_inc(&inode->i_count);
 592                         spin_unlock(&inode->i_lock);
 593
 594                         /* some remained atomic pages should discarded */
 595                         if (f2fs_is_atomic_file(inode))
 596                                 drop_inmem_pages(inode);
 597
 598                         /* should remain fi->extent_tree for writepage */
 599                         f2fs_destroy_extent_node(inode);
 600
 601                         sb_start_intwrite(inode->i_sb);
 602                         f2fs_i_size_write(inode, 0);
 603
 604                         if (F2FS_HAS_BLOCKS(inode))
 605                                 f2fs_truncate(inode);
 606
 607                         sb_end_intwrite(inode->i_sb);
 608
 609                         fscrypt_put_encryption_info(inode, NULL);
 610                         spin_lock(&inode->i_lock);
 611                         atomic_dec(&inode->i_count);
 612                 }
 613                 return 0;
 614         }
 615
 616         return generic_drop_inode(inode);
 617 }
 618
 619 int f2fs_inode_dirtied(struct inode *inode)
 620 {
 621         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 622
 623         spin_lock(&sbi->inode_lock[DIRTY_META]);
 624         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 625                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 626                 return 1;
 627         }
 628
 629         set_inode_flag(inode, FI_DIRTY_INODE);
 630         list_add_tail(&F2FS_I(inode)->gdirty_list,
 631                                 &sbi->inode_list[DIRTY_META]);
 632         inc_page_count(sbi, F2FS_DIRTY_IMETA);
 633         stat_inc_dirty_inode(sbi, DIRTY_META);
 634         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 635
 636         return 0;
 637 }
 638
 639 void f2fs_inode_synced(struct inode *inode)
 640 {
 641         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 642
 643         spin_lock(&sbi->inode_lock[DIRTY_META]);
 644         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 645                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 646                 return;
 647         }
 648         list_del_init(&F2FS_I(inode)->gdirty_list);
 649         clear_inode_flag(inode, FI_DIRTY_INODE);
 650         clear_inode_flag(inode, FI_AUTO_RECOVER);
 651         dec_page_count(sbi, F2FS_DIRTY_IMETA);
 652         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 653         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 654 }
 655
 656 /*
 657  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 658  *
 659  * We should call set_dirty_inode to write the dirty inode through write_inode.
 660  */
 661 static void f2fs_dirty_inode(struct inode *inode, int flags)
 662 {
 663         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 664
 665         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 666                         inode->i_ino == F2FS_META_INO(sbi))
 667                 return;
 668
 669         if (flags == I_DIRTY_TIME)
 670                 return;
 671
 672         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 673                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 674
 675         f2fs_inode_dirtied(inode);
 676 }
 677
 678 static void f2fs_i_callback(struct rcu_head *head)
 679 {
 680         struct inode *inode = container_of(head, struct inode, i_rcu);
 681         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 682 }
 683
 684 static void f2fs_destroy_inode(struct inode *inode)
 685 {
 686         call_rcu(&inode->i_rcu, f2fs_i_callback);
 687 }
 688
 689 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 690 {
 691         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 692         percpu_counter_destroy(&sbi->total_valid_inode_count);
 693 }
 694
 695 static void f2fs_put_super(struct super_block *sb)
 696 {
 697         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 698
 699         if (sbi->s_proc) {
 700                 remove_proc_entry("segment_info", sbi->s_proc);
 701                 remove_proc_entry("segment_bits", sbi->s_proc);
 702                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 703         }
 704         kobject_del(&sbi->s_kobj);
 705
 706         stop_gc_thread(sbi);
 707
 708         /* prevent remaining shrinker jobs */
 709         mutex_lock(&sbi->umount_mutex);
 710
 711         /*
 712          * We don't need to do checkpoint when superblock is clean.
 713          * But, the previous checkpoint was not done by umount, it needs to do
 714          * clean checkpoint again.
 715          */
 716         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 717                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
 718                 struct cp_control cpc = {
 719                         .reason = CP_UMOUNT,
 720                 };
 721                 write_checkpoint(sbi, &cpc);
 722         }
 723
 724         /* write_checkpoint can update stat informaion */
 725         f2fs_destroy_stats(sbi);
 726
 727         /*
 728          * normally superblock is clean, so we need to release this.
 729          * In addition, EIO will skip do checkpoint, we need this as well.
 730          */
 731         release_ino_entry(sbi, true);
 732         release_discard_addrs(sbi);
 733
 734         f2fs_leave_shrinker(sbi);
 735         mutex_unlock(&sbi->umount_mutex);
 736
 737         /* our cp_error case, we can wait for any writeback page */
 738         f2fs_flush_merged_bios(sbi);
 739
 740         iput(sbi->node_inode);
 741         iput(sbi->meta_inode);
 742
 743         /* destroy f2fs internal modules */
 744         destroy_node_manager(sbi);
 745         destroy_segment_manager(sbi);
 746
 747         kfree(sbi->ckpt);
 748         kobject_put(&sbi->s_kobj);
 749         wait_for_completion(&sbi->s_kobj_unregister);
 750
 751         sb->s_fs_info = NULL;
 752         if (sbi->s_chksum_driver)
 753                 crypto_free_shash(sbi->s_chksum_driver);
 754         kfree(sbi->raw_super);
 755
 756         destroy_percpu_info(sbi);
 757         kfree(sbi);
 758 }
 759
 760 int f2fs_sync_fs(struct super_block *sb, int sync)
 761 {
 762         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 763         int err = 0;
 764
 765         trace_f2fs_sync_fs(sb, sync);
 766
 767         if (sync) {
 768                 struct cp_control cpc;
 769
 770                 cpc.reason = __get_cp_reason(sbi);
 771
 772                 mutex_lock(&sbi->gc_mutex);
 773                 err = write_checkpoint(sbi, &cpc);
 774                 mutex_unlock(&sbi->gc_mutex);
 775         }
 776         f2fs_trace_ios(NULL, 1);
 777
 778         return err;
 779 }
 780
 781 static int f2fs_freeze(struct super_block *sb)
 782 {
 783         int err;
 784
 785         if (f2fs_readonly(sb))
 786                 return 0;
 787
 788         err = f2fs_sync_fs(sb, 1);
 789         return err;
 790 }
 791
 792 static int f2fs_unfreeze(struct super_block *sb)
 793 {
 794         return 0;
 795 }
 796
 797 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 798 {
 799         struct super_block *sb = dentry->d_sb;
 800         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 801         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 802         block_t total_count, user_block_count, start_count, ovp_count;
 803
 804         total_count = le64_to_cpu(sbi->raw_super->block_count);
 805         user_block_count = sbi->user_block_count;
 806         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 807         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 808         buf->f_type = F2FS_SUPER_MAGIC;
 809         buf->f_bsize = sbi->blocksize;
 810
 811         buf->f_blocks = total_count - start_count;
 812         buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
 813         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 814
 815         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 816         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
 817
 818         buf->f_namelen = F2FS_NAME_LEN;
 819         buf->f_fsid.val[0] = (u32)id;
 820         buf->f_fsid.val[1] = (u32)(id >> 32);
 821
 822         return 0;
 823 }
 824
 825 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 826 {
 827         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 828
 829         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
 830                 if (test_opt(sbi, FORCE_FG_GC))
 831                         seq_printf(seq, ",background_gc=%s", "sync");
 832                 else
 833                         seq_printf(seq, ",background_gc=%s", "on");
 834         } else {
 835                 seq_printf(seq, ",background_gc=%s", "off");
 836         }
 837         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 838                 seq_puts(seq, ",disable_roll_forward");
 839         if (test_opt(sbi, DISCARD))
 840                 seq_puts(seq, ",discard");
 841         if (test_opt(sbi, NOHEAP))
 842                 seq_puts(seq, ",no_heap_alloc");
 843 #ifdef CONFIG_F2FS_FS_XATTR
 844         if (test_opt(sbi, XATTR_USER))
 845                 seq_puts(seq, ",user_xattr");
 846         else
 847                 seq_puts(seq, ",nouser_xattr");
 848         if (test_opt(sbi, INLINE_XATTR))
 849                 seq_puts(seq, ",inline_xattr");
 850 #endif
 851 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 852         if (test_opt(sbi, POSIX_ACL))
 853                 seq_puts(seq, ",acl");
 854         else
 855                 seq_puts(seq, ",noacl");
 856 #endif
 857         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 858                 seq_puts(seq, ",disable_ext_identify");
 859         if (test_opt(sbi, INLINE_DATA))
 860                 seq_puts(seq, ",inline_data");
 861         else
 862                 seq_puts(seq, ",noinline_data");
 863         if (test_opt(sbi, INLINE_DENTRY))
 864                 seq_puts(seq, ",inline_dentry");
 865         else
 866                 seq_puts(seq, ",noinline_dentry");
 867         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 868                 seq_puts(seq, ",flush_merge");
 869         if (test_opt(sbi, NOBARRIER))
 870                 seq_puts(seq, ",nobarrier");
 871         if (test_opt(sbi, FASTBOOT))
 872                 seq_puts(seq, ",fastboot");
 873         if (test_opt(sbi, EXTENT_CACHE))
 874                 seq_puts(seq, ",extent_cache");
 875         else
 876                 seq_puts(seq, ",noextent_cache");
 877         if (test_opt(sbi, DATA_FLUSH))
 878                 seq_puts(seq, ",data_flush");
 879
 880         seq_puts(seq, ",mode=");
 881         if (test_opt(sbi, ADAPTIVE))
 882                 seq_puts(seq, "adaptive");
 883         else if (test_opt(sbi, LFS))
 884                 seq_puts(seq, "lfs");
 885         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 886
 887         return 0;
 888 }
 889
 890 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 891 {
 892         struct super_block *sb = seq->private;
 893         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 894         unsigned int total_segs =
 895                         le32_to_cpu(sbi->raw_super->segment_count_main);
 896         int i;
 897
 898         seq_puts(seq, "format: segment_type|valid_blocks\n"
 899                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 900
 901         for (i = 0; i < total_segs; i++) {
 902                 struct seg_entry *se = get_seg_entry(sbi, i);
 903
 904                 if ((i % 10) == 0)
 905                         seq_printf(seq, "%-10d", i);
 906                 seq_printf(seq, "%d|%-3u", se->type,
 907                                         get_valid_blocks(sbi, i, 1));
 908                 if ((i % 10) == 9 || i == (total_segs - 1))
 909                         seq_putc(seq, '\n');
 910                 else
 911                         seq_putc(seq, ' ');
 912         }
 913
 914         return 0;
 915 }
 916
 917 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
 918 {
 919         struct super_block *sb = seq->private;
 920         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 921         unsigned int total_segs =
 922                         le32_to_cpu(sbi->raw_super->segment_count_main);
 923         int i, j;
 924
 925         seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
 926                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 927
 928         for (i = 0; i < total_segs; i++) {
 929                 struct seg_entry *se = get_seg_entry(sbi, i);
 930
 931                 seq_printf(seq, "%-10d", i);
 932                 seq_printf(seq, "%d|%-3u|", se->type,
 933                                         get_valid_blocks(sbi, i, 1));
 934                 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
 935                         seq_printf(seq, " %.2x", se->cur_valid_map[j]);
 936                 seq_putc(seq, '\n');
 937         }
 938         return 0;
 939 }
 940
 941 #define F2FS_PROC_FILE_DEF(_name)                                       \
 942 static int _name##_open_fs(struct inode *inode, struct file *file)      \
 943 {                                                                       \
 944         return single_open(file, _name##_seq_show, PDE_DATA(inode));    \
 945 }                                                                       \
 946                                                                         \
 947 static const struct file_operations f2fs_seq_##_name##_fops = {         \
 948         .open = _name##_open_fs,                                        \
 949         .read = seq_read,                                               \
 950         .llseek = seq_lseek,                                            \
 951         .release = single_release,                                      \
 952 };
 953
 954 F2FS_PROC_FILE_DEF(segment_info);
 955 F2FS_PROC_FILE_DEF(segment_bits);
 956
 957 static void default_options(struct f2fs_sb_info *sbi)
 958 {
 959         /* init some FS parameters */
 960         sbi->active_logs = NR_CURSEG_TYPE;
 961
 962         set_opt(sbi, BG_GC);
 963         set_opt(sbi, INLINE_DATA);
 964         set_opt(sbi, INLINE_DENTRY);
 965         set_opt(sbi, EXTENT_CACHE);
 966         sbi->sb->s_flags |= MS_LAZYTIME;
 967         set_opt(sbi, FLUSH_MERGE);
 968         if (f2fs_sb_mounted_hmsmr(sbi->sb)) {
 969                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 970                 set_opt(sbi, DISCARD);
 971         } else {
 972                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 973         }
 974
 975 #ifdef CONFIG_F2FS_FS_XATTR
 976         set_opt(sbi, XATTR_USER);
 977 #endif
 978 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 979         set_opt(sbi, POSIX_ACL);
 980 #endif
 981
 982 #ifdef CONFIG_F2FS_FAULT_INJECTION
 983         f2fs_build_fault_attr(sbi, 0);
 984 #endif
 985 }
 986
 987 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 988 {
 989         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 990         struct f2fs_mount_info org_mount_opt;
 991         int err, active_logs;
 992         bool need_restart_gc = false;
 993         bool need_stop_gc = false;
 994         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
 995 #ifdef CONFIG_F2FS_FAULT_INJECTION
 996         struct f2fs_fault_info ffi = sbi->fault_info;
 997 #endif
 998
 999         /*
1000          * Save the old mount options in case we
1001          * need to restore them.
1002          */
1003         org_mount_opt = sbi->mount_opt;
1004         active_logs = sbi->active_logs;
1005
1006         /* recover superblocks we couldn't write due to previous RO mount */
1007         if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1008                 err = f2fs_commit_super(sbi, false);
1009                 f2fs_msg(sb, KERN_INFO,
1010                         "Try to recover all the superblocks, ret: %d", err);
1011                 if (!err)
1012                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1013         }
1014
1015         sbi->mount_opt.opt = 0;
1016         default_options(sbi);
1017
1018         /* parse mount options */
1019         err = parse_options(sb, data);
1020         if (err)
1021                 goto restore_opts;
1022
1023         /*
1024          * Previous and new state of filesystem is RO,
1025          * so skip checking GC and FLUSH_MERGE conditions.
1026          */
1027         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1028                 goto skip;
1029
1030         /* disallow enable/disable extent_cache dynamically */
1031         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1032                 err = -EINVAL;
1033                 f2fs_msg(sbi->sb, KERN_WARNING,
1034                                 "switch extent_cache option is not allowed");
1035                 goto restore_opts;
1036         }
1037
1038         /*
1039          * We stop the GC thread if FS is mounted as RO
1040          * or if background_gc = off is passed in mount
1041          * option. Also sync the filesystem.
1042          */
1043         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1044                 if (sbi->gc_thread) {
1045                         stop_gc_thread(sbi);
1046                         need_restart_gc = true;
1047                 }
1048         } else if (!sbi->gc_thread) {
1049                 err = start_gc_thread(sbi);
1050                 if (err)
1051                         goto restore_opts;
1052                 need_stop_gc = true;
1053         }
1054
1055         if (*flags & MS_RDONLY) {
1056                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1057                 sync_inodes_sb(sb);
1058
1059                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1060                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1061                 f2fs_sync_fs(sb, 1);
1062                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1063         }
1064
1065         /*
1066          * We stop issue flush thread if FS is mounted as RO
1067          * or if flush_merge is not passed in mount option.
1068          */
1069         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1070                 destroy_flush_cmd_control(sbi);
1071         } else if (!SM_I(sbi)->cmd_control_info) {
1072                 err = create_flush_cmd_control(sbi);
1073                 if (err)
1074                         goto restore_gc;
1075         }
1076 skip:
1077         /* Update the POSIXACL Flag */
1078         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1079                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1080
1081         return 0;
1082 restore_gc:
1083         if (need_restart_gc) {
1084                 if (start_gc_thread(sbi))
1085                         f2fs_msg(sbi->sb, KERN_WARNING,
1086                                 "background gc thread has stopped");
1087         } else if (need_stop_gc) {
1088                 stop_gc_thread(sbi);
1089         }
1090 restore_opts:
1091         sbi->mount_opt = org_mount_opt;
1092         sbi->active_logs = active_logs;
1093 #ifdef CONFIG_F2FS_FAULT_INJECTION
1094         sbi->fault_info = ffi;
1095 #endif
1096         return err;
1097 }
1098
1099 static struct super_operations f2fs_sops = {
1100         .alloc_inode    = f2fs_alloc_inode,
1101         .drop_inode     = f2fs_drop_inode,
1102         .destroy_inode  = f2fs_destroy_inode,
1103         .write_inode    = f2fs_write_inode,
1104         .dirty_inode    = f2fs_dirty_inode,
1105         .show_options   = f2fs_show_options,
1106         .evict_inode    = f2fs_evict_inode,
1107         .put_super      = f2fs_put_super,
1108         .sync_fs        = f2fs_sync_fs,
1109         .freeze_fs      = f2fs_freeze,
1110         .unfreeze_fs    = f2fs_unfreeze,
1111         .statfs         = f2fs_statfs,
1112         .remount_fs     = f2fs_remount,
1113 };
1114
1115 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1116 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1117 {
1118         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1119                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1120                                 ctx, len, NULL);
1121 }
1122
1123 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1124 {
1125         *key = F2FS_I_SB(inode)->key_prefix;
1126         return F2FS_I_SB(inode)->key_prefix_size;
1127 }
1128
1129 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1130                                                         void *fs_data)
1131 {
1132         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1133                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1134                                 ctx, len, fs_data, XATTR_CREATE);
1135 }
1136
1137 static unsigned f2fs_max_namelen(struct inode *inode)
1138 {
1139         return S_ISLNK(inode->i_mode) ?
1140                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1141 }
1142
1143 static struct fscrypt_operations f2fs_cryptops = {
1144         .get_context    = f2fs_get_context,
1145         .key_prefix     = f2fs_key_prefix,
1146         .set_context    = f2fs_set_context,
1147         .is_encrypted   = f2fs_encrypted_inode,
1148         .empty_dir      = f2fs_empty_dir,
1149         .max_namelen    = f2fs_max_namelen,
1150 };
1151 #else
1152 static struct fscrypt_operations f2fs_cryptops = {
1153         .is_encrypted   = f2fs_encrypted_inode,
1154 };
1155 #endif
1156
1157 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1158                 u64 ino, u32 generation)
1159 {
1160         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1161         struct inode *inode;
1162
1163         if (check_nid_range(sbi, ino))
1164                 return ERR_PTR(-ESTALE);
1165
1166         /*
1167          * f2fs_iget isn't quite right if the inode is currently unallocated!
1168          * However f2fs_iget currently does appropriate checks to handle stale
1169          * inodes so everything is OK.
1170          */
1171         inode = f2fs_iget(sb, ino);
1172         if (IS_ERR(inode))
1173                 return ERR_CAST(inode);
1174         if (unlikely(generation && inode->i_generation != generation)) {
1175                 /* we didn't find the right inode.. */
1176                 iput(inode);
1177                 return ERR_PTR(-ESTALE);
1178         }
1179         return inode;
1180 }
1181
1182 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1183                 int fh_len, int fh_type)
1184 {
1185         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1186                                     f2fs_nfs_get_inode);
1187 }
1188
1189 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1190                 int fh_len, int fh_type)
1191 {
1192         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1193                                     f2fs_nfs_get_inode);
1194 }
1195
1196 static const struct export_operations f2fs_export_ops = {
1197         .fh_to_dentry = f2fs_fh_to_dentry,
1198         .fh_to_parent = f2fs_fh_to_parent,
1199         .get_parent = f2fs_get_parent,
1200 };
1201
1202 static loff_t max_file_blocks(void)
1203 {
1204         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1205         loff_t leaf_count = ADDRS_PER_BLOCK;
1206
1207         /* two direct node blocks */
1208         result += (leaf_count * 2);
1209
1210         /* two indirect node blocks */
1211         leaf_count *= NIDS_PER_BLOCK;
1212         result += (leaf_count * 2);
1213
1214         /* one double indirect node block */
1215         leaf_count *= NIDS_PER_BLOCK;
1216         result += leaf_count;
1217
1218         return result;
1219 }
1220
1221 static int __f2fs_commit_super(struct buffer_head *bh,
1222                         struct f2fs_super_block *super)
1223 {
1224         lock_buffer(bh);
1225         if (super)
1226                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1227         set_buffer_uptodate(bh);
1228         set_buffer_dirty(bh);
1229         unlock_buffer(bh);
1230
1231         /* it's rare case, we can do fua all the time */
1232         return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1233 }
1234
1235 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1236                                         struct buffer_head *bh)
1237 {
1238         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1239                                         (bh->b_data + F2FS_SUPER_OFFSET);
1240         struct super_block *sb = sbi->sb;
1241         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1242         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1243         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1244         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1245         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1246         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1247         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1248         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1249         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1250         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1251         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1252         u32 segment_count = le32_to_cpu(raw_super->segment_count);
1253         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1254         u64 main_end_blkaddr = main_blkaddr +
1255                                 (segment_count_main << log_blocks_per_seg);
1256         u64 seg_end_blkaddr = segment0_blkaddr +
1257                                 (segment_count << log_blocks_per_seg);
1258
1259         if (segment0_blkaddr != cp_blkaddr) {
1260                 f2fs_msg(sb, KERN_INFO,
1261                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1262                         segment0_blkaddr, cp_blkaddr);
1263                 return true;
1264         }
1265
1266         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1267                                                         sit_blkaddr) {
1268                 f2fs_msg(sb, KERN_INFO,
1269                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1270                         cp_blkaddr, sit_blkaddr,
1271                         segment_count_ckpt << log_blocks_per_seg);
1272                 return true;
1273         }
1274
1275         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1276                                                         nat_blkaddr) {
1277                 f2fs_msg(sb, KERN_INFO,
1278                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1279                         sit_blkaddr, nat_blkaddr,
1280                         segment_count_sit << log_blocks_per_seg);
1281                 return true;
1282         }
1283
1284         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1285                                                         ssa_blkaddr) {
1286                 f2fs_msg(sb, KERN_INFO,
1287                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1288                         nat_blkaddr, ssa_blkaddr,
1289                         segment_count_nat << log_blocks_per_seg);
1290                 return true;
1291         }
1292
1293         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1294                                                         main_blkaddr) {
1295                 f2fs_msg(sb, KERN_INFO,
1296                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1297                         ssa_blkaddr, main_blkaddr,
1298                         segment_count_ssa << log_blocks_per_seg);
1299                 return true;
1300         }
1301
1302         if (main_end_blkaddr > seg_end_blkaddr) {
1303                 f2fs_msg(sb, KERN_INFO,
1304                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1305                         main_blkaddr,
1306                         segment0_blkaddr +
1307                                 (segment_count << log_blocks_per_seg),
1308                         segment_count_main << log_blocks_per_seg);
1309                 return true;
1310         } else if (main_end_blkaddr < seg_end_blkaddr) {
1311                 int err = 0;
1312                 char *res;
1313
1314                 /* fix in-memory information all the time */
1315                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1316                                 segment0_blkaddr) >> log_blocks_per_seg);
1317
1318                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1319                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1320                         res = "internally";
1321                 } else {
1322                         err = __f2fs_commit_super(bh, NULL);
1323                         res = err ? "failed" : "done";
1324                 }
1325                 f2fs_msg(sb, KERN_INFO,
1326                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
1327                         res, main_blkaddr,
1328                         segment0_blkaddr +
1329                                 (segment_count << log_blocks_per_seg),
1330                         segment_count_main << log_blocks_per_seg);
1331                 if (err)
1332                         return true;
1333         }
1334         return false;
1335 }
1336
1337 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1338                                 struct buffer_head *bh)
1339 {
1340         block_t segment_count, segs_per_sec, secs_per_zone;
1341         block_t total_sections, blocks_per_seg;
1342         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1343                                         (bh->b_data + F2FS_SUPER_OFFSET);
1344         struct super_block *sb = sbi->sb;
1345         unsigned int blocksize;
1346
1347         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1348                 f2fs_msg(sb, KERN_INFO,
1349                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1350                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1351                 return 1;
1352         }
1353
1354         /* Currently, support only 4KB page cache size */
1355         if (F2FS_BLKSIZE != PAGE_SIZE) {
1356                 f2fs_msg(sb, KERN_INFO,
1357                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1358                         PAGE_SIZE);
1359                 return 1;
1360         }
1361
1362         /* Currently, support only 4KB block size */
1363         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1364         if (blocksize != F2FS_BLKSIZE) {
1365                 f2fs_msg(sb, KERN_INFO,
1366                         "Invalid blocksize (%u), supports only 4KB\n",
1367                         blocksize);
1368                 return 1;
1369         }
1370
1371         /* check log blocks per segment */
1372         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1373                 f2fs_msg(sb, KERN_INFO,
1374                         "Invalid log blocks per segment (%u)\n",
1375                         le32_to_cpu(raw_super->log_blocks_per_seg));
1376                 return 1;
1377         }
1378
1379         /* Currently, support 512/1024/2048/4096 bytes sector size */
1380         if (le32_to_cpu(raw_super->log_sectorsize) >
1381                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1382                 le32_to_cpu(raw_super->log_sectorsize) <
1383                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1384                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1385                         le32_to_cpu(raw_super->log_sectorsize));
1386                 return 1;
1387         }
1388         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1389                 le32_to_cpu(raw_super->log_sectorsize) !=
1390                         F2FS_MAX_LOG_SECTOR_SIZE) {
1391                 f2fs_msg(sb, KERN_INFO,
1392                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1393                         le32_to_cpu(raw_super->log_sectors_per_block),
1394                         le32_to_cpu(raw_super->log_sectorsize));
1395                 return 1;
1396         }
1397
1398         segment_count = le32_to_cpu(raw_super->segment_count);
1399         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1400         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1401         total_sections = le32_to_cpu(raw_super->section_count);
1402
1403         /* blocks_per_seg should be 512, given the above check */
1404         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
1405
1406         if (segment_count > F2FS_MAX_SEGMENT ||
1407                                 segment_count < F2FS_MIN_SEGMENTS) {
1408                 f2fs_msg(sb, KERN_INFO,
1409                         "Invalid segment count (%u)",
1410                         segment_count);
1411                 return 1;
1412         }
1413
1414         if (total_sections > segment_count ||
1415                         total_sections < F2FS_MIN_SEGMENTS ||
1416                         segs_per_sec > segment_count || !segs_per_sec) {
1417                 f2fs_msg(sb, KERN_INFO,
1418                         "Invalid segment/section count (%u, %u x %u)",
1419                         segment_count, total_sections, segs_per_sec);
1420                 return 1;
1421         }
1422
1423         if ((segment_count / segs_per_sec) < total_sections) {
1424                 f2fs_msg(sb, KERN_INFO,
1425                         "Small segment_count (%u < %u * %u)",
1426                         segment_count, segs_per_sec, total_sections);
1427                 return 1;
1428         }
1429
1430         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
1431                 f2fs_msg(sb, KERN_INFO,
1432                         "Wrong segment_count / block_count (%u > %llu)",
1433                         segment_count, le64_to_cpu(raw_super->block_count));
1434                 return 1;
1435         }
1436
1437         if (secs_per_zone > total_sections || !secs_per_zone) {
1438                 f2fs_msg(sb, KERN_INFO,
1439                         "Wrong secs_per_zone / total_sections (%u, %u)",
1440                         secs_per_zone, total_sections);
1441                 return 1;
1442         }
1443         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION) {
1444                 f2fs_msg(sb, KERN_INFO,
1445                         "Corrupted extension count (%u > %u)",
1446                         le32_to_cpu(raw_super->extension_count),
1447                         F2FS_MAX_EXTENSION);
1448                 return 1;
1449         }
1450
1451         if (le32_to_cpu(raw_super->cp_payload) >
1452                                 (blocks_per_seg - F2FS_CP_PACKS)) {
1453                 f2fs_msg(sb, KERN_INFO,
1454                         "Insane cp_payload (%u > %u)",
1455                         le32_to_cpu(raw_super->cp_payload),
1456                         blocks_per_seg - F2FS_CP_PACKS);
1457                 return 1;
1458         }
1459
1460         /* check reserved ino info */
1461         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1462                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1463                 le32_to_cpu(raw_super->root_ino) != 3) {
1464                 f2fs_msg(sb, KERN_INFO,
1465                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1466                         le32_to_cpu(raw_super->node_ino),
1467                         le32_to_cpu(raw_super->meta_ino),
1468                         le32_to_cpu(raw_super->root_ino));
1469                 return 1;
1470         }
1471
1472         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1473         if (sanity_check_area_boundary(sbi, bh))
1474                 return 1;
1475
1476         return 0;
1477 }
1478
1479 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1480 {
1481         unsigned int total, fsmeta;
1482         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1483         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1484         unsigned int ovp_segments, reserved_segments;
1485         unsigned int main_segs, blocks_per_seg;
1486         unsigned int sit_segs, nat_segs;
1487         unsigned int sit_bitmap_size, nat_bitmap_size;
1488         unsigned int log_blocks_per_seg;
1489         unsigned int segment_count_main;
1490         unsigned int cp_pack_start_sum, cp_payload;
1491         block_t user_block_count;
1492         int i, j;
1493
1494         total = le32_to_cpu(raw_super->segment_count);
1495         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1496         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
1497         fsmeta += sit_segs;
1498         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
1499         fsmeta += nat_segs;
1500         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1501         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1502
1503         if (unlikely(fsmeta >= total))
1504                 return 1;
1505
1506         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1507         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1508
1509         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1510                         ovp_segments == 0 || reserved_segments == 0)) {
1511                 f2fs_msg(sbi->sb, KERN_ERR,
1512                         "Wrong layout: check mkfs.f2fs version");
1513                 return 1;
1514         }
1515
1516         user_block_count = le64_to_cpu(ckpt->user_block_count);
1517         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1518         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1519         if (!user_block_count || user_block_count >=
1520                         segment_count_main << log_blocks_per_seg) {
1521                 f2fs_msg(sbi->sb, KERN_ERR,
1522                         "Wrong user_block_count: %u", user_block_count);
1523                 return 1;
1524         }
1525
1526         main_segs = le32_to_cpu(raw_super->segment_count_main);
1527         blocks_per_seg = sbi->blocks_per_seg;
1528
1529         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
1530                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
1531                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
1532                         return 1;
1533                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
1534                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
1535                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
1536                                 f2fs_msg(sbi->sb, KERN_ERR,
1537                                         "Node segment (%u, %u) has the same "
1538                                         "segno: %u", i, j,
1539                                         le32_to_cpu(ckpt->cur_node_segno[i]));
1540                                 return 1;
1541                         }
1542                 }
1543         }
1544         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
1545                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
1546                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
1547                         return 1;
1548                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
1549                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
1550                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
1551                                 f2fs_msg(sbi->sb, KERN_ERR,
1552                                         "Data segment (%u, %u) has the same "
1553                                         "segno: %u", i, j,
1554                                         le32_to_cpu(ckpt->cur_data_segno[i]));
1555                                 return 1;
1556                         }
1557                 }
1558         }
1559         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
1560                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
1561                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
1562                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
1563                                 f2fs_msg(sbi->sb, KERN_ERR,
1564                                         "Node segment (%u) and Data segment (%u)"
1565                                         " has the same segno: %u", i, j,
1566                                         le32_to_cpu(ckpt->cur_node_segno[i]));
1567                                 return 1;
1568                         }
1569                 }
1570         }
1571
1572         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
1573         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
1574
1575         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
1576                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
1577                 f2fs_msg(sbi->sb, KERN_ERR,
1578                         "Wrong bitmap size: sit: %u, nat:%u",
1579                         sit_bitmap_size, nat_bitmap_size);
1580                 return 1;
1581         }
1582
1583         cp_pack_start_sum = __start_sum_addr(sbi);
1584         cp_payload = __cp_payload(sbi);
1585         if (cp_pack_start_sum < cp_payload + 1 ||
1586                 cp_pack_start_sum > blocks_per_seg - 1 -
1587                         NR_CURSEG_TYPE) {
1588                 f2fs_msg(sbi->sb, KERN_ERR,
1589                         "Wrong cp_pack_start_sum: %u",
1590                         cp_pack_start_sum);
1591                 return 1;
1592         }
1593
1594         if (unlikely(f2fs_cp_error(sbi))) {
1595                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1596                 return 1;
1597         }
1598         return 0;
1599 }
1600
1601 static void init_sb_info(struct f2fs_sb_info *sbi)
1602 {
1603         struct f2fs_super_block *raw_super = sbi->raw_super;
1604         int i;
1605
1606         sbi->log_sectors_per_block =
1607                 le32_to_cpu(raw_super->log_sectors_per_block);
1608         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1609         sbi->blocksize = 1 << sbi->log_blocksize;
1610         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1611         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1612         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1613         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1614         sbi->total_sections = le32_to_cpu(raw_super->section_count);
1615         sbi->total_node_count =
1616                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1617                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1618         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1619         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1620         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1621         sbi->cur_victim_sec = NULL_SECNO;
1622         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1623
1624         sbi->dir_level = DEF_DIR_LEVEL;
1625         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1626         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1627         clear_sbi_flag(sbi, SBI_NEED_FSCK);
1628
1629         for (i = 0; i < NR_COUNT_TYPE; i++)
1630                 atomic_set(&sbi->nr_pages[i], 0);
1631
1632         INIT_LIST_HEAD(&sbi->s_list);
1633         mutex_init(&sbi->umount_mutex);
1634         mutex_init(&sbi->wio_mutex[NODE]);
1635         mutex_init(&sbi->wio_mutex[DATA]);
1636         spin_lock_init(&sbi->cp_lock);
1637
1638 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1639         memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1640                                 F2FS_KEY_DESC_PREFIX_SIZE);
1641         sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1642 #endif
1643 }
1644
1645 static int init_percpu_info(struct f2fs_sb_info *sbi)
1646 {
1647         int err;
1648
1649         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1650         if (err)
1651                 return err;
1652
1653         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
1654                                                                 GFP_KERNEL);
1655         if (err)
1656                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1657
1658         return err;
1659 }
1660
1661 /*
1662  * Read f2fs raw super block.
1663  * Because we have two copies of super block, so read both of them
1664  * to get the first valid one. If any one of them is broken, we pass
1665  * them recovery flag back to the caller.
1666  */
1667 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1668                         struct f2fs_super_block **raw_super,
1669                         int *valid_super_block, int *recovery)
1670 {
1671         struct super_block *sb = sbi->sb;
1672         int block;
1673         struct buffer_head *bh;
1674         struct f2fs_super_block *super;
1675         int err = 0;
1676
1677         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1678         if (!super)
1679                 return -ENOMEM;
1680
1681         for (block = 0; block < 2; block++) {
1682                 bh = sb_bread(sb, block);
1683                 if (!bh) {
1684                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1685                                 block + 1);
1686                         err = -EIO;
1687                         continue;
1688                 }
1689
1690                 /* sanity checking of raw super */
1691                 if (sanity_check_raw_super(sbi, bh)) {
1692                         f2fs_msg(sb, KERN_ERR,
1693                                 "Can't find valid F2FS filesystem in %dth superblock",
1694                                 block + 1);
1695                         err = -EINVAL;
1696                         brelse(bh);
1697                         continue;
1698                 }
1699
1700                 if (!*raw_super) {
1701                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1702                                                         sizeof(*super));
1703                         *valid_super_block = block;
1704                         *raw_super = super;
1705                 }
1706                 brelse(bh);
1707         }
1708
1709         /* Fail to read any one of the superblocks*/
1710         if (err < 0)
1711                 *recovery = 1;
1712
1713         /* No valid superblock */
1714         if (!*raw_super)
1715                 kfree(super);
1716         else
1717                 err = 0;
1718
1719         return err;
1720 }
1721
1722 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1723 {
1724         struct buffer_head *bh;
1725         int err;
1726
1727         if ((recover && f2fs_readonly(sbi->sb)) ||
1728                                 bdev_read_only(sbi->sb->s_bdev)) {
1729                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1730                 return -EROFS;
1731         }
1732
1733         /* write back-up superblock first */
1734         bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1735         if (!bh)
1736                 return -EIO;
1737         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1738         brelse(bh);
1739
1740         /* if we are in recovery path, skip writing valid superblock */
1741         if (recover || err)
1742                 return err;
1743
1744         /* write current valid superblock */
1745         bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1746         if (!bh)
1747                 return -EIO;
1748         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1749         brelse(bh);
1750         return err;
1751 }
1752
1753 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1754 {
1755         struct f2fs_sb_info *sbi;
1756         struct f2fs_super_block *raw_super;
1757         struct inode *root;
1758         int err;
1759         bool retry = true, need_fsck = false;
1760         char *options = NULL;
1761         int recovery, i, valid_super_block;
1762         struct curseg_info *seg_i;
1763
1764 try_onemore:
1765         err = -EINVAL;
1766         raw_super = NULL;
1767         valid_super_block = -1;
1768         recovery = 0;
1769
1770         /* allocate memory for f2fs-specific super block info */
1771         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1772         if (!sbi)
1773                 return -ENOMEM;
1774
1775         sbi->sb = sb;
1776
1777         /* Load the checksum driver */
1778         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1779         if (IS_ERR(sbi->s_chksum_driver)) {
1780                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1781                 err = PTR_ERR(sbi->s_chksum_driver);
1782                 sbi->s_chksum_driver = NULL;
1783                 goto free_sbi;
1784         }
1785
1786         /* set a block size */
1787         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1788                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1789                 goto free_sbi;
1790         }
1791
1792         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1793                                                                 &recovery);
1794         if (err)
1795                 goto free_sbi;
1796
1797         sb->s_fs_info = sbi;
1798         sbi->raw_super = raw_super;
1799
1800         default_options(sbi);
1801         /* parse mount options */
1802         options = kstrdup((const char *)data, GFP_KERNEL);
1803         if (data && !options) {
1804                 err = -ENOMEM;
1805                 goto free_sb_buf;
1806         }
1807
1808         err = parse_options(sb, options);
1809         if (err)
1810                 goto free_options;
1811
1812         sbi->max_file_blocks = max_file_blocks();
1813         sb->s_maxbytes = sbi->max_file_blocks <<
1814                                 le32_to_cpu(raw_super->log_blocksize);
1815         sb->s_max_links = F2FS_LINK_MAX;
1816         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1817
1818         sb->s_op = &f2fs_sops;
1819         sb->s_cop = &f2fs_cryptops;
1820         sb->s_xattr = f2fs_xattr_handlers;
1821         sb->s_export_op = &f2fs_export_ops;
1822         sb->s_magic = F2FS_SUPER_MAGIC;
1823         sb->s_time_gran = 1;
1824         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1825                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1826         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1827
1828         /* init f2fs-specific super block info */
1829         sbi->valid_super_block = valid_super_block;
1830         mutex_init(&sbi->gc_mutex);
1831         mutex_init(&sbi->cp_mutex);
1832         init_rwsem(&sbi->node_write);
1833
1834         /* disallow all the data/node/meta page writes */
1835         set_sbi_flag(sbi, SBI_POR_DOING);
1836         spin_lock_init(&sbi->stat_lock);
1837
1838         init_rwsem(&sbi->read_io.io_rwsem);
1839         sbi->read_io.sbi = sbi;
1840         sbi->read_io.bio = NULL;
1841         for (i = 0; i < NR_PAGE_TYPE; i++) {
1842                 init_rwsem(&sbi->write_io[i].io_rwsem);
1843                 sbi->write_io[i].sbi = sbi;
1844                 sbi->write_io[i].bio = NULL;
1845         }
1846
1847         init_rwsem(&sbi->cp_rwsem);
1848         init_waitqueue_head(&sbi->cp_wait);
1849         init_sb_info(sbi);
1850
1851         err = init_percpu_info(sbi);
1852         if (err)
1853                 goto free_options;
1854
1855         /* get an inode for meta space */
1856         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1857         if (IS_ERR(sbi->meta_inode)) {
1858                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1859                 err = PTR_ERR(sbi->meta_inode);
1860                 goto free_options;
1861         }
1862
1863         err = get_valid_checkpoint(sbi);
1864         if (err) {
1865                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1866                 goto free_meta_inode;
1867         }
1868
1869         sbi->total_valid_node_count =
1870                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1871         percpu_counter_set(&sbi->total_valid_inode_count,
1872                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
1873         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1874         sbi->total_valid_block_count =
1875                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1876         sbi->last_valid_block_count = sbi->total_valid_block_count;
1877
1878         for (i = 0; i < NR_INODE_TYPE; i++) {
1879                 INIT_LIST_HEAD(&sbi->inode_list[i]);
1880                 spin_lock_init(&sbi->inode_lock[i]);
1881         }
1882
1883         init_extent_cache_info(sbi);
1884
1885         init_ino_entry_info(sbi);
1886
1887         /* setup f2fs internal modules */
1888         err = build_segment_manager(sbi);
1889         if (err) {
1890                 f2fs_msg(sb, KERN_ERR,
1891                         "Failed to initialize F2FS segment manager");
1892                 goto free_sm;
1893         }
1894         err = build_node_manager(sbi);
1895         if (err) {
1896                 f2fs_msg(sb, KERN_ERR,
1897                         "Failed to initialize F2FS node manager");
1898                 goto free_nm;
1899         }
1900
1901         /* For write statistics */
1902         if (sb->s_bdev->bd_part)
1903                 sbi->sectors_written_start =
1904                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1905
1906         /* Read accumulated write IO statistics if exists */
1907         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1908         if (__exist_node_summaries(sbi))
1909                 sbi->kbytes_written =
1910                         le64_to_cpu(seg_i->journal->info.kbytes_written);
1911
1912         build_gc_manager(sbi);
1913
1914         /* get an inode for node space */
1915         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1916         if (IS_ERR(sbi->node_inode)) {
1917                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1918                 err = PTR_ERR(sbi->node_inode);
1919                 goto free_nm;
1920         }
1921
1922         f2fs_join_shrinker(sbi);
1923
1924         /* if there are nt orphan nodes free them */
1925         err = recover_orphan_inodes(sbi);
1926         if (err)
1927                 goto free_node_inode;
1928
1929         /* read root inode and dentry */
1930         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1931         if (IS_ERR(root)) {
1932                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1933                 err = PTR_ERR(root);
1934                 goto free_node_inode;
1935         }
1936         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1937                 iput(root);
1938                 err = -EINVAL;
1939                 goto free_node_inode;
1940         }
1941
1942         sb->s_root = d_make_root(root); /* allocate root dentry */
1943         if (!sb->s_root) {
1944                 err = -ENOMEM;
1945                 goto free_root_inode;
1946         }
1947
1948         err = f2fs_build_stats(sbi);
1949         if (err)
1950                 goto free_root_inode;
1951
1952         if (f2fs_proc_root)
1953                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1954
1955         if (sbi->s_proc) {
1956                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1957                                  &f2fs_seq_segment_info_fops, sb);
1958                 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1959                                  &f2fs_seq_segment_bits_fops, sb);
1960         }
1961
1962         sbi->s_kobj.kset = f2fs_kset;
1963         init_completion(&sbi->s_kobj_unregister);
1964         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1965                                                         "%s", sb->s_id);
1966         if (err)
1967                 goto free_proc;
1968
1969         /* recover fsynced data */
1970         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1971                 /*
1972                  * mount should be failed, when device has readonly mode, and
1973                  * previous checkpoint was not done by clean system shutdown.
1974                  */
1975                 if (bdev_read_only(sb->s_bdev) &&
1976                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
1977                         err = -EROFS;
1978                         goto free_kobj;
1979                 }
1980
1981                 if (need_fsck)
1982                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1983
1984                 if (!retry)
1985                         goto skip_recovery;
1986
1987                 err = recover_fsync_data(sbi, false);
1988                 if (err < 0) {
1989                         need_fsck = true;
1990                         f2fs_msg(sb, KERN_ERR,
1991                                 "Cannot recover all fsync data errno=%d", err);
1992                         goto free_kobj;
1993                 }
1994         } else {
1995                 err = recover_fsync_data(sbi, true);
1996
1997                 if (!f2fs_readonly(sb) && err > 0) {
1998                         err = -EINVAL;
1999                         f2fs_msg(sb, KERN_ERR,
2000                                 "Need to recover fsync data");
2001                         goto free_kobj;
2002                 }
2003         }
2004 skip_recovery:
2005         /* recover_fsync_data() cleared this already */
2006         clear_sbi_flag(sbi, SBI_POR_DOING);
2007
2008         /*
2009          * If filesystem is not mounted as read-only then
2010          * do start the gc_thread.
2011          */
2012         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2013                 /* After POR, we can run background GC thread.*/
2014                 err = start_gc_thread(sbi);
2015                 if (err)
2016                         goto free_kobj;
2017         }
2018         kfree(options);
2019
2020         /* recover broken superblock */
2021         if (recovery) {
2022                 err = f2fs_commit_super(sbi, true);
2023                 f2fs_msg(sb, KERN_INFO,
2024                         "Try to recover %dth superblock, ret: %d",
2025                         sbi->valid_super_block ? 1 : 2, err);
2026         }
2027
2028         f2fs_update_time(sbi, CP_TIME);
2029         f2fs_update_time(sbi, REQ_TIME);
2030         return 0;
2031
2032 free_kobj:
2033         f2fs_sync_inode_meta(sbi);
2034         kobject_del(&sbi->s_kobj);
2035         kobject_put(&sbi->s_kobj);
2036         wait_for_completion(&sbi->s_kobj_unregister);
2037 free_proc:
2038         if (sbi->s_proc) {
2039                 remove_proc_entry("segment_info", sbi->s_proc);
2040                 remove_proc_entry("segment_bits", sbi->s_proc);
2041                 remove_proc_entry(sb->s_id, f2fs_proc_root);
2042         }
2043         f2fs_destroy_stats(sbi);
2044 free_root_inode:
2045         dput(sb->s_root);
2046         sb->s_root = NULL;
2047 free_node_inode:
2048         truncate_inode_pages_final(NODE_MAPPING(sbi));
2049         mutex_lock(&sbi->umount_mutex);
2050         release_ino_entry(sbi, true);
2051         f2fs_leave_shrinker(sbi);
2052         iput(sbi->node_inode);
2053         mutex_unlock(&sbi->umount_mutex);
2054 free_nm:
2055         destroy_node_manager(sbi);
2056 free_sm:
2057         destroy_segment_manager(sbi);
2058         kfree(sbi->ckpt);
2059 free_meta_inode:
2060         make_bad_inode(sbi->meta_inode);
2061         iput(sbi->meta_inode);
2062 free_options:
2063         destroy_percpu_info(sbi);
2064         kfree(options);
2065 free_sb_buf:
2066         kfree(raw_super);
2067 free_sbi:
2068         if (sbi->s_chksum_driver)
2069                 crypto_free_shash(sbi->s_chksum_driver);
2070         kfree(sbi);
2071
2072         /* give only one another chance */
2073         if (retry) {
2074                 retry = false;
2075                 shrink_dcache_sb(sb);
2076                 goto try_onemore;
2077         }
2078         return err;
2079 }
2080
2081 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2082                         const char *dev_name, void *data)
2083 {
2084         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2085 }
2086
2087 static void kill_f2fs_super(struct super_block *sb)
2088 {
2089         if (sb->s_root)
2090                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2091         kill_block_super(sb);
2092 }
2093
2094 static struct file_system_type f2fs_fs_type = {
2095         .owner          = THIS_MODULE,
2096         .name           = "f2fs",
2097         .mount          = f2fs_mount,
2098         .kill_sb        = kill_f2fs_super,
2099         .fs_flags       = FS_REQUIRES_DEV,
2100 };
2101 MODULE_ALIAS_FS("f2fs");
2102
2103 static int __init init_inodecache(void)
2104 {
2105         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2106                         sizeof(struct f2fs_inode_info), 0,
2107                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2108         if (!f2fs_inode_cachep)
2109                 return -ENOMEM;
2110         return 0;
2111 }
2112
2113 static void destroy_inodecache(void)
2114 {
2115         /*
2116          * Make sure all delayed rcu free inodes are flushed before we
2117          * destroy cache.
2118          */
2119         rcu_barrier();
2120         kmem_cache_destroy(f2fs_inode_cachep);
2121 }
2122
2123 static int __init init_f2fs_fs(void)
2124 {
2125         int err;
2126
2127         if (PAGE_SIZE != F2FS_BLKSIZE) {
2128                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
2129                                 PAGE_SIZE, F2FS_BLKSIZE);
2130                 return -EINVAL;
2131         }
2132
2133         f2fs_build_trace_ios();
2134
2135         err = init_inodecache();
2136         if (err)
2137                 goto fail;
2138         err = create_node_manager_caches();
2139         if (err)
2140                 goto free_inodecache;
2141         err = create_segment_manager_caches();
2142         if (err)
2143                 goto free_node_manager_caches;
2144         err = create_checkpoint_caches();
2145         if (err)
2146                 goto free_segment_manager_caches;
2147         err = create_extent_cache();
2148         if (err)
2149                 goto free_checkpoint_caches;
2150         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
2151         if (!f2fs_kset) {
2152                 err = -ENOMEM;
2153                 goto free_extent_cache;
2154         }
2155         err = register_shrinker(&f2fs_shrinker_info);
2156         if (err)
2157                 goto free_kset;
2158
2159         err = register_filesystem(&f2fs_fs_type);
2160         if (err)
2161                 goto free_shrinker;
2162         err = f2fs_create_root_stats();
2163         if (err)
2164                 goto free_filesystem;
2165         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2166         return 0;
2167
2168 free_filesystem:
2169         unregister_filesystem(&f2fs_fs_type);
2170 free_shrinker:
2171         unregister_shrinker(&f2fs_shrinker_info);
2172 free_kset:
2173         kset_unregister(f2fs_kset);
2174 free_extent_cache:
2175         destroy_extent_cache();
2176 free_checkpoint_caches:
2177         destroy_checkpoint_caches();
2178 free_segment_manager_caches:
2179         destroy_segment_manager_caches();
2180 free_node_manager_caches:
2181         destroy_node_manager_caches();
2182 free_inodecache:
2183         destroy_inodecache();
2184 fail:
2185         return err;
2186 }
2187
2188 static void __exit exit_f2fs_fs(void)
2189 {
2190         remove_proc_entry("fs/f2fs", NULL);
2191         f2fs_destroy_root_stats();
2192         unregister_filesystem(&f2fs_fs_type);
2193         unregister_shrinker(&f2fs_shrinker_info);
2194         kset_unregister(f2fs_kset);
2195         destroy_extent_cache();
2196         destroy_checkpoint_caches();
2197         destroy_segment_manager_caches();
2198         destroy_node_manager_caches();
2199         destroy_inodecache();
2200         f2fs_destroy_trace_ios();
2201 }
2202
2203 module_init(init_f2fs_fs)
2204 module_exit(exit_f2fs_fs)
2205
2206 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2207 MODULE_DESCRIPTION("Flash Friendly File System");
2208 MODULE_LICENSE("GPL");