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 if (!f2fs_sb_mounted_blkzoned(sb)) {
 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                         break;
 424                 case Opt_noheap:
 425                         set_opt(sbi, NOHEAP);
 426                         break;
 427 #ifdef CONFIG_F2FS_FS_XATTR
 428                 case Opt_user_xattr:
 429                         set_opt(sbi, XATTR_USER);
 430                         break;
 431                 case Opt_nouser_xattr:
 432                         clear_opt(sbi, XATTR_USER);
 433                         break;
 434                 case Opt_inline_xattr:
 435                         set_opt(sbi, INLINE_XATTR);
 436                         break;
 437 #else
 438                 case Opt_user_xattr:
 439                         f2fs_msg(sb, KERN_INFO,
 440                                 "user_xattr options not supported");
 441                         break;
 442                 case Opt_nouser_xattr:
 443                         f2fs_msg(sb, KERN_INFO,
 444                                 "nouser_xattr options not supported");
 445                         break;
 446                 case Opt_inline_xattr:
 447                         f2fs_msg(sb, KERN_INFO,
 448                                 "inline_xattr options not supported");
 449                         break;
 450 #endif
 451 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 452                 case Opt_acl:
 453                         set_opt(sbi, POSIX_ACL);
 454                         break;
 455                 case Opt_noacl:
 456                         clear_opt(sbi, POSIX_ACL);
 457                         break;
 458 #else
 459                 case Opt_acl:
 460                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 461                         break;
 462                 case Opt_noacl:
 463                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 464                         break;
 465 #endif
 466                 case Opt_active_logs:
 467                         if (args->from && match_int(args, &arg))
 468                                 return -EINVAL;
 469                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 470                                 return -EINVAL;
 471                         sbi->active_logs = arg;
 472                         break;
 473                 case Opt_disable_ext_identify:
 474                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 475                         break;
 476                 case Opt_inline_data:
 477                         set_opt(sbi, INLINE_DATA);
 478                         break;
 479                 case Opt_inline_dentry:
 480                         set_opt(sbi, INLINE_DENTRY);
 481                         break;
 482                 case Opt_noinline_dentry:
 483                         clear_opt(sbi, INLINE_DENTRY);
 484                         break;
 485                 case Opt_flush_merge:
 486                         set_opt(sbi, FLUSH_MERGE);
 487                         break;
 488                 case Opt_noflush_merge:
 489                         clear_opt(sbi, FLUSH_MERGE);
 490                         break;
 491                 case Opt_nobarrier:
 492                         set_opt(sbi, NOBARRIER);
 493                         break;
 494                 case Opt_fastboot:
 495                         set_opt(sbi, FASTBOOT);
 496                         break;
 497                 case Opt_extent_cache:
 498                         set_opt(sbi, EXTENT_CACHE);
 499                         break;
 500                 case Opt_noextent_cache:
 501                         clear_opt(sbi, EXTENT_CACHE);
 502                         break;
 503                 case Opt_noinline_data:
 504                         clear_opt(sbi, INLINE_DATA);
 505                         break;
 506                 case Opt_data_flush:
 507                         set_opt(sbi, DATA_FLUSH);
 508                         break;
 509                 case Opt_mode:
 510                         name = match_strdup(&args[0]);
 511
 512                         if (!name)
 513                                 return -ENOMEM;
 514                         if (strlen(name) == 8 &&
 515                                         !strncmp(name, "adaptive", 8)) {
 516                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 517                         } else if (strlen(name) == 3 &&
 518                                         !strncmp(name, "lfs", 3)) {
 519                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 520                         } else {
 521                                 kfree(name);
 522                                 return -EINVAL;
 523                         }
 524                         kfree(name);
 525                         break;
 526                 case Opt_fault_injection:
 527                         if (args->from && match_int(args, &arg))
 528                                 return -EINVAL;
 529 #ifdef CONFIG_F2FS_FAULT_INJECTION
 530                         f2fs_build_fault_attr(sbi, arg);
 531 #else
 532                         f2fs_msg(sb, KERN_INFO,
 533                                 "FAULT_INJECTION was not selected");
 534 #endif
 535                         break;
 536                 case Opt_lazytime:
 537                         sb->s_flags |= MS_LAZYTIME;
 538                         break;
 539                 case Opt_nolazytime:
 540                         sb->s_flags &= ~MS_LAZYTIME;
 541                         break;
 542                 default:
 543                         f2fs_msg(sb, KERN_ERR,
 544                                 "Unrecognized mount option \"%s\" or missing value",
 545                                 p);
 546                         return -EINVAL;
 547                 }
 548         }
 549         return 0;
 550 }
 551
 552 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 553 {
 554         struct f2fs_inode_info *fi;
 555
 556         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 557         if (!fi)
 558                 return NULL;
 559
 560         init_once((void *) fi);
 561
 562         if (percpu_counter_init(&fi->dirty_pages, 0, GFP_NOFS)) {
 563                 kmem_cache_free(f2fs_inode_cachep, fi);
 564                 return NULL;
 565         }
 566
 567         /* Initialize f2fs-specific inode info */
 568         fi->vfs_inode.i_version = 1;
 569         fi->i_current_depth = 1;
 570         fi->i_advise = 0;
 571         init_rwsem(&fi->i_sem);
 572         INIT_LIST_HEAD(&fi->dirty_list);
 573         INIT_LIST_HEAD(&fi->gdirty_list);
 574         INIT_LIST_HEAD(&fi->inmem_pages);
 575         mutex_init(&fi->inmem_lock);
 576         init_rwsem(&fi->dio_rwsem[READ]);
 577         init_rwsem(&fi->dio_rwsem[WRITE]);
 578
 579         /* Will be used by directory only */
 580         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 581         return &fi->vfs_inode;
 582 }
 583
 584 static int f2fs_drop_inode(struct inode *inode)
 585 {
 586         /*
 587          * This is to avoid a deadlock condition like below.
 588          * writeback_single_inode(inode)
 589          *  - f2fs_write_data_page
 590          *    - f2fs_gc -> iput -> evict
 591          *       - inode_wait_for_writeback(inode)
 592          */
 593         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 594                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 595                         /* to avoid evict_inode call simultaneously */
 596                         atomic_inc(&inode->i_count);
 597                         spin_unlock(&inode->i_lock);
 598
 599                         /* some remained atomic pages should discarded */
 600                         if (f2fs_is_atomic_file(inode))
 601                                 drop_inmem_pages(inode);
 602
 603                         /* should remain fi->extent_tree for writepage */
 604                         f2fs_destroy_extent_node(inode);
 605
 606                         sb_start_intwrite(inode->i_sb);
 607                         f2fs_i_size_write(inode, 0);
 608
 609                         if (F2FS_HAS_BLOCKS(inode))
 610                                 f2fs_truncate(inode);
 611
 612                         sb_end_intwrite(inode->i_sb);
 613
 614                         fscrypt_put_encryption_info(inode, NULL);
 615                         spin_lock(&inode->i_lock);
 616                         atomic_dec(&inode->i_count);
 617                 }
 618                 return 0;
 619         }
 620
 621         return generic_drop_inode(inode);
 622 }
 623
 624 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 625 {
 626         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 627         int ret = 0;
 628
 629         spin_lock(&sbi->inode_lock[DIRTY_META]);
 630         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 631                 ret = 1;
 632         } else {
 633                 set_inode_flag(inode, FI_DIRTY_INODE);
 634                 stat_inc_dirty_inode(sbi, DIRTY_META);
 635         }
 636         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 637                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 638                                 &sbi->inode_list[DIRTY_META]);
 639                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 640         }
 641         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 642         return ret;
 643 }
 644
 645 void f2fs_inode_synced(struct inode *inode)
 646 {
 647         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 648
 649         spin_lock(&sbi->inode_lock[DIRTY_META]);
 650         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 651                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 652                 return;
 653         }
 654         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 655                 list_del_init(&F2FS_I(inode)->gdirty_list);
 656                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 657         }
 658         clear_inode_flag(inode, FI_DIRTY_INODE);
 659         clear_inode_flag(inode, FI_AUTO_RECOVER);
 660         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 661         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 662 }
 663
 664 /*
 665  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 666  *
 667  * We should call set_dirty_inode to write the dirty inode through write_inode.
 668  */
 669 static void f2fs_dirty_inode(struct inode *inode, int flags)
 670 {
 671         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 672
 673         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 674                         inode->i_ino == F2FS_META_INO(sbi))
 675                 return;
 676
 677         if (flags == I_DIRTY_TIME)
 678                 return;
 679
 680         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 681                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 682
 683         f2fs_inode_dirtied(inode, false);
 684 }
 685
 686 static void f2fs_i_callback(struct rcu_head *head)
 687 {
 688         struct inode *inode = container_of(head, struct inode, i_rcu);
 689         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 690 }
 691
 692 static void f2fs_destroy_inode(struct inode *inode)
 693 {
 694         percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
 695         call_rcu(&inode->i_rcu, f2fs_i_callback);
 696 }
 697
 698 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 699 {
 700         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 701         percpu_counter_destroy(&sbi->total_valid_inode_count);
 702 }
 703
 704 static void f2fs_put_super(struct super_block *sb)
 705 {
 706         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 707
 708         if (sbi->s_proc) {
 709                 remove_proc_entry("segment_info", sbi->s_proc);
 710                 remove_proc_entry("segment_bits", sbi->s_proc);
 711                 remove_proc_entry(sb->s_id, f2fs_proc_root);
 712         }
 713         kobject_del(&sbi->s_kobj);
 714
 715         stop_gc_thread(sbi);
 716
 717         /* prevent remaining shrinker jobs */
 718         mutex_lock(&sbi->umount_mutex);
 719
 720         /*
 721          * We don't need to do checkpoint when superblock is clean.
 722          * But, the previous checkpoint was not done by umount, it needs to do
 723          * clean checkpoint again.
 724          */
 725         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 726                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
 727                 struct cp_control cpc = {
 728                         .reason = CP_UMOUNT,
 729                 };
 730                 write_checkpoint(sbi, &cpc);
 731         }
 732
 733         /* write_checkpoint can update stat informaion */
 734         f2fs_destroy_stats(sbi);
 735
 736         /*
 737          * normally superblock is clean, so we need to release this.
 738          * In addition, EIO will skip do checkpoint, we need this as well.
 739          */
 740         release_ino_entry(sbi, true);
 741
 742         f2fs_leave_shrinker(sbi);
 743         mutex_unlock(&sbi->umount_mutex);
 744
 745         /* our cp_error case, we can wait for any writeback page */
 746         f2fs_flush_merged_bios(sbi);
 747
 748         iput(sbi->node_inode);
 749         iput(sbi->meta_inode);
 750
 751         /* destroy f2fs internal modules */
 752         destroy_node_manager(sbi);
 753         destroy_segment_manager(sbi);
 754
 755         kfree(sbi->ckpt);
 756         kobject_put(&sbi->s_kobj);
 757         wait_for_completion(&sbi->s_kobj_unregister);
 758
 759         sb->s_fs_info = NULL;
 760         if (sbi->s_chksum_driver)
 761                 crypto_free_shash(sbi->s_chksum_driver);
 762         kfree(sbi->raw_super);
 763
 764         destroy_percpu_info(sbi);
 765         kfree(sbi);
 766 }
 767
 768 int f2fs_sync_fs(struct super_block *sb, int sync)
 769 {
 770         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 771         int err = 0;
 772
 773         trace_f2fs_sync_fs(sb, sync);
 774
 775         if (sync) {
 776                 struct cp_control cpc;
 777
 778                 cpc.reason = __get_cp_reason(sbi);
 779
 780                 mutex_lock(&sbi->gc_mutex);
 781                 err = write_checkpoint(sbi, &cpc);
 782                 mutex_unlock(&sbi->gc_mutex);
 783         }
 784         f2fs_trace_ios(NULL, 1);
 785
 786         return err;
 787 }
 788
 789 static int f2fs_freeze(struct super_block *sb)
 790 {
 791         int err;
 792
 793         if (f2fs_readonly(sb))
 794                 return 0;
 795
 796         err = f2fs_sync_fs(sb, 1);
 797         return err;
 798 }
 799
 800 static int f2fs_unfreeze(struct super_block *sb)
 801 {
 802         return 0;
 803 }
 804
 805 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 806 {
 807         struct super_block *sb = dentry->d_sb;
 808         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 809         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 810         block_t total_count, user_block_count, start_count, ovp_count;
 811
 812         total_count = le64_to_cpu(sbi->raw_super->block_count);
 813         user_block_count = sbi->user_block_count;
 814         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 815         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 816         buf->f_type = F2FS_SUPER_MAGIC;
 817         buf->f_bsize = sbi->blocksize;
 818
 819         buf->f_blocks = total_count - start_count;
 820         buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
 821         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
 822
 823         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 824         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
 825
 826         buf->f_namelen = F2FS_NAME_LEN;
 827         buf->f_fsid.val[0] = (u32)id;
 828         buf->f_fsid.val[1] = (u32)(id >> 32);
 829
 830         return 0;
 831 }
 832
 833 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 834 {
 835         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
 836
 837         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
 838                 if (test_opt(sbi, FORCE_FG_GC))
 839                         seq_printf(seq, ",background_gc=%s", "sync");
 840                 else
 841                         seq_printf(seq, ",background_gc=%s", "on");
 842         } else {
 843                 seq_printf(seq, ",background_gc=%s", "off");
 844         }
 845         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
 846                 seq_puts(seq, ",disable_roll_forward");
 847         if (test_opt(sbi, DISCARD))
 848                 seq_puts(seq, ",discard");
 849         if (test_opt(sbi, NOHEAP))
 850                 seq_puts(seq, ",no_heap_alloc");
 851 #ifdef CONFIG_F2FS_FS_XATTR
 852         if (test_opt(sbi, XATTR_USER))
 853                 seq_puts(seq, ",user_xattr");
 854         else
 855                 seq_puts(seq, ",nouser_xattr");
 856         if (test_opt(sbi, INLINE_XATTR))
 857                 seq_puts(seq, ",inline_xattr");
 858 #endif
 859 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 860         if (test_opt(sbi, POSIX_ACL))
 861                 seq_puts(seq, ",acl");
 862         else
 863                 seq_puts(seq, ",noacl");
 864 #endif
 865         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
 866                 seq_puts(seq, ",disable_ext_identify");
 867         if (test_opt(sbi, INLINE_DATA))
 868                 seq_puts(seq, ",inline_data");
 869         else
 870                 seq_puts(seq, ",noinline_data");
 871         if (test_opt(sbi, INLINE_DENTRY))
 872                 seq_puts(seq, ",inline_dentry");
 873         else
 874                 seq_puts(seq, ",noinline_dentry");
 875         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 876                 seq_puts(seq, ",flush_merge");
 877         if (test_opt(sbi, NOBARRIER))
 878                 seq_puts(seq, ",nobarrier");
 879         if (test_opt(sbi, FASTBOOT))
 880                 seq_puts(seq, ",fastboot");
 881         if (test_opt(sbi, EXTENT_CACHE))
 882                 seq_puts(seq, ",extent_cache");
 883         else
 884                 seq_puts(seq, ",noextent_cache");
 885         if (test_opt(sbi, DATA_FLUSH))
 886                 seq_puts(seq, ",data_flush");
 887
 888         seq_puts(seq, ",mode=");
 889         if (test_opt(sbi, ADAPTIVE))
 890                 seq_puts(seq, "adaptive");
 891         else if (test_opt(sbi, LFS))
 892                 seq_puts(seq, "lfs");
 893         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 894
 895         return 0;
 896 }
 897
 898 static int segment_info_seq_show(struct seq_file *seq, void *offset)
 899 {
 900         struct super_block *sb = seq->private;
 901         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 902         unsigned int total_segs =
 903                         le32_to_cpu(sbi->raw_super->segment_count_main);
 904         int i;
 905
 906         seq_puts(seq, "format: segment_type|valid_blocks\n"
 907                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 908
 909         for (i = 0; i < total_segs; i++) {
 910                 struct seg_entry *se = get_seg_entry(sbi, i);
 911
 912                 if ((i % 10) == 0)
 913                         seq_printf(seq, "%-10d", i);
 914                 seq_printf(seq, "%d|%-3u", se->type,
 915                                         get_valid_blocks(sbi, i, 1));
 916                 if ((i % 10) == 9 || i == (total_segs - 1))
 917                         seq_putc(seq, '\n');
 918                 else
 919                         seq_putc(seq, ' ');
 920         }
 921
 922         return 0;
 923 }
 924
 925 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
 926 {
 927         struct super_block *sb = seq->private;
 928         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 929         unsigned int total_segs =
 930                         le32_to_cpu(sbi->raw_super->segment_count_main);
 931         int i, j;
 932
 933         seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
 934                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
 935
 936         for (i = 0; i < total_segs; i++) {
 937                 struct seg_entry *se = get_seg_entry(sbi, i);
 938
 939                 seq_printf(seq, "%-10d", i);
 940                 seq_printf(seq, "%d|%-3u|", se->type,
 941                                         get_valid_blocks(sbi, i, 1));
 942                 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
 943                         seq_printf(seq, " %.2x", se->cur_valid_map[j]);
 944                 seq_putc(seq, '\n');
 945         }
 946         return 0;
 947 }
 948
 949 #define F2FS_PROC_FILE_DEF(_name)                                       \
 950 static int _name##_open_fs(struct inode *inode, struct file *file)      \
 951 {                                                                       \
 952         return single_open(file, _name##_seq_show, PDE_DATA(inode));    \
 953 }                                                                       \
 954                                                                         \
 955 static const struct file_operations f2fs_seq_##_name##_fops = {         \
 956         .open = _name##_open_fs,                                        \
 957         .read = seq_read,                                               \
 958         .llseek = seq_lseek,                                            \
 959         .release = single_release,                                      \
 960 };
 961
 962 F2FS_PROC_FILE_DEF(segment_info);
 963 F2FS_PROC_FILE_DEF(segment_bits);
 964
 965 static void default_options(struct f2fs_sb_info *sbi)
 966 {
 967         /* init some FS parameters */
 968         sbi->active_logs = NR_CURSEG_TYPE;
 969
 970         set_opt(sbi, BG_GC);
 971         set_opt(sbi, INLINE_DATA);
 972         set_opt(sbi, INLINE_DENTRY);
 973         set_opt(sbi, EXTENT_CACHE);
 974         sbi->sb->s_flags |= MS_LAZYTIME;
 975         set_opt(sbi, FLUSH_MERGE);
 976         if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
 977                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 978                 set_opt(sbi, DISCARD);
 979         } else {
 980                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 981         }
 982
 983 #ifdef CONFIG_F2FS_FS_XATTR
 984         set_opt(sbi, XATTR_USER);
 985 #endif
 986 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 987         set_opt(sbi, POSIX_ACL);
 988 #endif
 989
 990 #ifdef CONFIG_F2FS_FAULT_INJECTION
 991         f2fs_build_fault_attr(sbi, 0);
 992 #endif
 993 }
 994
 995 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 996 {
 997         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 998         struct f2fs_mount_info org_mount_opt;
 999         int err, active_logs;
1000         bool need_restart_gc = false;
1001         bool need_stop_gc = false;
1002         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1003 #ifdef CONFIG_F2FS_FAULT_INJECTION
1004         struct f2fs_fault_info ffi = sbi->fault_info;
1005 #endif
1006
1007         /*
1008          * Save the old mount options in case we
1009          * need to restore them.
1010          */
1011         org_mount_opt = sbi->mount_opt;
1012         active_logs = sbi->active_logs;
1013
1014         /* recover superblocks we couldn't write due to previous RO mount */
1015         if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1016                 err = f2fs_commit_super(sbi, false);
1017                 f2fs_msg(sb, KERN_INFO,
1018                         "Try to recover all the superblocks, ret: %d", err);
1019                 if (!err)
1020                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1021         }
1022
1023         sbi->mount_opt.opt = 0;
1024         default_options(sbi);
1025
1026         /* parse mount options */
1027         err = parse_options(sb, data);
1028         if (err)
1029                 goto restore_opts;
1030
1031         /*
1032          * Previous and new state of filesystem is RO,
1033          * so skip checking GC and FLUSH_MERGE conditions.
1034          */
1035         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1036                 goto skip;
1037
1038         /* disallow enable/disable extent_cache dynamically */
1039         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1040                 err = -EINVAL;
1041                 f2fs_msg(sbi->sb, KERN_WARNING,
1042                                 "switch extent_cache option is not allowed");
1043                 goto restore_opts;
1044         }
1045
1046         /*
1047          * We stop the GC thread if FS is mounted as RO
1048          * or if background_gc = off is passed in mount
1049          * option. Also sync the filesystem.
1050          */
1051         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1052                 if (sbi->gc_thread) {
1053                         stop_gc_thread(sbi);
1054                         need_restart_gc = true;
1055                 }
1056         } else if (!sbi->gc_thread) {
1057                 err = start_gc_thread(sbi);
1058                 if (err)
1059                         goto restore_opts;
1060                 need_stop_gc = true;
1061         }
1062
1063         if (*flags & MS_RDONLY) {
1064                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1065                 sync_inodes_sb(sb);
1066
1067                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1068                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1069                 f2fs_sync_fs(sb, 1);
1070                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1071         }
1072
1073         /*
1074          * We stop issue flush thread if FS is mounted as RO
1075          * or if flush_merge is not passed in mount option.
1076          */
1077         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1078                 destroy_flush_cmd_control(sbi);
1079         } else if (!SM_I(sbi)->cmd_control_info) {
1080                 err = create_flush_cmd_control(sbi);
1081                 if (err)
1082                         goto restore_gc;
1083         }
1084 skip:
1085         /* Update the POSIXACL Flag */
1086         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1087                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1088
1089         return 0;
1090 restore_gc:
1091         if (need_restart_gc) {
1092                 if (start_gc_thread(sbi))
1093                         f2fs_msg(sbi->sb, KERN_WARNING,
1094                                 "background gc thread has stopped");
1095         } else if (need_stop_gc) {
1096                 stop_gc_thread(sbi);
1097         }
1098 restore_opts:
1099         sbi->mount_opt = org_mount_opt;
1100         sbi->active_logs = active_logs;
1101 #ifdef CONFIG_F2FS_FAULT_INJECTION
1102         sbi->fault_info = ffi;
1103 #endif
1104         return err;
1105 }
1106
1107 static struct super_operations f2fs_sops = {
1108         .alloc_inode    = f2fs_alloc_inode,
1109         .drop_inode     = f2fs_drop_inode,
1110         .destroy_inode  = f2fs_destroy_inode,
1111         .write_inode    = f2fs_write_inode,
1112         .dirty_inode    = f2fs_dirty_inode,
1113         .show_options   = f2fs_show_options,
1114         .evict_inode    = f2fs_evict_inode,
1115         .put_super      = f2fs_put_super,
1116         .sync_fs        = f2fs_sync_fs,
1117         .freeze_fs      = f2fs_freeze,
1118         .unfreeze_fs    = f2fs_unfreeze,
1119         .statfs         = f2fs_statfs,
1120         .remount_fs     = f2fs_remount,
1121 };
1122
1123 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1124 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1125 {
1126         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1127                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1128                                 ctx, len, NULL);
1129 }
1130
1131 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1132 {
1133         *key = F2FS_I_SB(inode)->key_prefix;
1134         return F2FS_I_SB(inode)->key_prefix_size;
1135 }
1136
1137 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1138                                                         void *fs_data)
1139 {
1140         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1141                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1142                                 ctx, len, fs_data, XATTR_CREATE);
1143 }
1144
1145 static unsigned f2fs_max_namelen(struct inode *inode)
1146 {
1147         return S_ISLNK(inode->i_mode) ?
1148                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1149 }
1150
1151 static struct fscrypt_operations f2fs_cryptops = {
1152         .get_context    = f2fs_get_context,
1153         .key_prefix     = f2fs_key_prefix,
1154         .set_context    = f2fs_set_context,
1155         .is_encrypted   = f2fs_encrypted_inode,
1156         .empty_dir      = f2fs_empty_dir,
1157         .max_namelen    = f2fs_max_namelen,
1158 };
1159 #else
1160 static struct fscrypt_operations f2fs_cryptops = {
1161         .is_encrypted   = f2fs_encrypted_inode,
1162 };
1163 #endif
1164
1165 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1166                 u64 ino, u32 generation)
1167 {
1168         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1169         struct inode *inode;
1170
1171         if (check_nid_range(sbi, ino))
1172                 return ERR_PTR(-ESTALE);
1173
1174         /*
1175          * f2fs_iget isn't quite right if the inode is currently unallocated!
1176          * However f2fs_iget currently does appropriate checks to handle stale
1177          * inodes so everything is OK.
1178          */
1179         inode = f2fs_iget(sb, ino);
1180         if (IS_ERR(inode))
1181                 return ERR_CAST(inode);
1182         if (unlikely(generation && inode->i_generation != generation)) {
1183                 /* we didn't find the right inode.. */
1184                 iput(inode);
1185                 return ERR_PTR(-ESTALE);
1186         }
1187         return inode;
1188 }
1189
1190 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1191                 int fh_len, int fh_type)
1192 {
1193         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1194                                     f2fs_nfs_get_inode);
1195 }
1196
1197 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1198                 int fh_len, int fh_type)
1199 {
1200         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1201                                     f2fs_nfs_get_inode);
1202 }
1203
1204 static const struct export_operations f2fs_export_ops = {
1205         .fh_to_dentry = f2fs_fh_to_dentry,
1206         .fh_to_parent = f2fs_fh_to_parent,
1207         .get_parent = f2fs_get_parent,
1208 };
1209
1210 static loff_t max_file_blocks(void)
1211 {
1212         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1213         loff_t leaf_count = ADDRS_PER_BLOCK;
1214
1215         /* two direct node blocks */
1216         result += (leaf_count * 2);
1217
1218         /* two indirect node blocks */
1219         leaf_count *= NIDS_PER_BLOCK;
1220         result += (leaf_count * 2);
1221
1222         /* one double indirect node block */
1223         leaf_count *= NIDS_PER_BLOCK;
1224         result += leaf_count;
1225
1226         return result;
1227 }
1228
1229 static int __f2fs_commit_super(struct buffer_head *bh,
1230                         struct f2fs_super_block *super)
1231 {
1232         lock_buffer(bh);
1233         if (super)
1234                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1235         set_buffer_uptodate(bh);
1236         set_buffer_dirty(bh);
1237         unlock_buffer(bh);
1238
1239         /* it's rare case, we can do fua all the time */
1240         return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1241 }
1242
1243 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1244                                         struct buffer_head *bh)
1245 {
1246         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1247                                         (bh->b_data + F2FS_SUPER_OFFSET);
1248         struct super_block *sb = sbi->sb;
1249         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1250         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1251         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1252         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1253         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1254         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1255         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1256         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1257         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1258         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1259         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1260         u32 segment_count = le32_to_cpu(raw_super->segment_count);
1261         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1262         u64 main_end_blkaddr = main_blkaddr +
1263                                 (segment_count_main << log_blocks_per_seg);
1264         u64 seg_end_blkaddr = segment0_blkaddr +
1265                                 (segment_count << log_blocks_per_seg);
1266
1267         if (segment0_blkaddr != cp_blkaddr) {
1268                 f2fs_msg(sb, KERN_INFO,
1269                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1270                         segment0_blkaddr, cp_blkaddr);
1271                 return true;
1272         }
1273
1274         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1275                                                         sit_blkaddr) {
1276                 f2fs_msg(sb, KERN_INFO,
1277                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1278                         cp_blkaddr, sit_blkaddr,
1279                         segment_count_ckpt << log_blocks_per_seg);
1280                 return true;
1281         }
1282
1283         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1284                                                         nat_blkaddr) {
1285                 f2fs_msg(sb, KERN_INFO,
1286                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1287                         sit_blkaddr, nat_blkaddr,
1288                         segment_count_sit << log_blocks_per_seg);
1289                 return true;
1290         }
1291
1292         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1293                                                         ssa_blkaddr) {
1294                 f2fs_msg(sb, KERN_INFO,
1295                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1296                         nat_blkaddr, ssa_blkaddr,
1297                         segment_count_nat << log_blocks_per_seg);
1298                 return true;
1299         }
1300
1301         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1302                                                         main_blkaddr) {
1303                 f2fs_msg(sb, KERN_INFO,
1304                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1305                         ssa_blkaddr, main_blkaddr,
1306                         segment_count_ssa << log_blocks_per_seg);
1307                 return true;
1308         }
1309
1310         if (main_end_blkaddr > seg_end_blkaddr) {
1311                 f2fs_msg(sb, KERN_INFO,
1312                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1313                         main_blkaddr,
1314                         segment0_blkaddr +
1315                                 (segment_count << log_blocks_per_seg),
1316                         segment_count_main << log_blocks_per_seg);
1317                 return true;
1318         } else if (main_end_blkaddr < seg_end_blkaddr) {
1319                 int err = 0;
1320                 char *res;
1321
1322                 /* fix in-memory information all the time */
1323                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1324                                 segment0_blkaddr) >> log_blocks_per_seg);
1325
1326                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1327                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1328                         res = "internally";
1329                 } else {
1330                         err = __f2fs_commit_super(bh, NULL);
1331                         res = err ? "failed" : "done";
1332                 }
1333                 f2fs_msg(sb, KERN_INFO,
1334                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
1335                         res, main_blkaddr,
1336                         segment0_blkaddr +
1337                                 (segment_count << log_blocks_per_seg),
1338                         segment_count_main << log_blocks_per_seg);
1339                 if (err)
1340                         return true;
1341         }
1342         return false;
1343 }
1344
1345 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1346                                 struct buffer_head *bh)
1347 {
1348         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1349                                         (bh->b_data + F2FS_SUPER_OFFSET);
1350         struct super_block *sb = sbi->sb;
1351         unsigned int blocksize;
1352
1353         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1354                 f2fs_msg(sb, KERN_INFO,
1355                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1356                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1357                 return 1;
1358         }
1359
1360         /* Currently, support only 4KB page cache size */
1361         if (F2FS_BLKSIZE != PAGE_SIZE) {
1362                 f2fs_msg(sb, KERN_INFO,
1363                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1364                         PAGE_SIZE);
1365                 return 1;
1366         }
1367
1368         /* Currently, support only 4KB block size */
1369         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1370         if (blocksize != F2FS_BLKSIZE) {
1371                 f2fs_msg(sb, KERN_INFO,
1372                         "Invalid blocksize (%u), supports only 4KB\n",
1373                         blocksize);
1374                 return 1;
1375         }
1376
1377         /* check log blocks per segment */
1378         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1379                 f2fs_msg(sb, KERN_INFO,
1380                         "Invalid log blocks per segment (%u)\n",
1381                         le32_to_cpu(raw_super->log_blocks_per_seg));
1382                 return 1;
1383         }
1384
1385         /* Currently, support 512/1024/2048/4096 bytes sector size */
1386         if (le32_to_cpu(raw_super->log_sectorsize) >
1387                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1388                 le32_to_cpu(raw_super->log_sectorsize) <
1389                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1390                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1391                         le32_to_cpu(raw_super->log_sectorsize));
1392                 return 1;
1393         }
1394         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1395                 le32_to_cpu(raw_super->log_sectorsize) !=
1396                         F2FS_MAX_LOG_SECTOR_SIZE) {
1397                 f2fs_msg(sb, KERN_INFO,
1398                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1399                         le32_to_cpu(raw_super->log_sectors_per_block),
1400                         le32_to_cpu(raw_super->log_sectorsize));
1401                 return 1;
1402         }
1403
1404         /* check reserved ino info */
1405         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1406                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1407                 le32_to_cpu(raw_super->root_ino) != 3) {
1408                 f2fs_msg(sb, KERN_INFO,
1409                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1410                         le32_to_cpu(raw_super->node_ino),
1411                         le32_to_cpu(raw_super->meta_ino),
1412                         le32_to_cpu(raw_super->root_ino));
1413                 return 1;
1414         }
1415
1416         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1417         if (sanity_check_area_boundary(sbi, bh))
1418                 return 1;
1419
1420         return 0;
1421 }
1422
1423 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1424 {
1425         unsigned int total, fsmeta;
1426         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1427         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1428
1429         total = le32_to_cpu(raw_super->segment_count);
1430         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1431         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1432         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1433         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1434         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1435
1436         if (unlikely(fsmeta >= total))
1437                 return 1;
1438
1439         if (unlikely(f2fs_cp_error(sbi))) {
1440                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1441                 return 1;
1442         }
1443         return 0;
1444 }
1445
1446 static void init_sb_info(struct f2fs_sb_info *sbi)
1447 {
1448         struct f2fs_super_block *raw_super = sbi->raw_super;
1449         int i;
1450
1451         sbi->log_sectors_per_block =
1452                 le32_to_cpu(raw_super->log_sectors_per_block);
1453         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1454         sbi->blocksize = 1 << sbi->log_blocksize;
1455         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1456         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1457         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1458         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1459         sbi->total_sections = le32_to_cpu(raw_super->section_count);
1460         sbi->total_node_count =
1461                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1462                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1463         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1464         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1465         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1466         sbi->cur_victim_sec = NULL_SECNO;
1467         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1468
1469         sbi->dir_level = DEF_DIR_LEVEL;
1470         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1471         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1472         clear_sbi_flag(sbi, SBI_NEED_FSCK);
1473
1474         for (i = 0; i < NR_COUNT_TYPE; i++)
1475                 atomic_set(&sbi->nr_pages[i], 0);
1476
1477         INIT_LIST_HEAD(&sbi->s_list);
1478         mutex_init(&sbi->umount_mutex);
1479         mutex_init(&sbi->wio_mutex[NODE]);
1480         mutex_init(&sbi->wio_mutex[DATA]);
1481         spin_lock_init(&sbi->cp_lock);
1482
1483 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1484         memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1485                                 F2FS_KEY_DESC_PREFIX_SIZE);
1486         sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1487 #endif
1488 }
1489
1490 static int init_percpu_info(struct f2fs_sb_info *sbi)
1491 {
1492         int err;
1493
1494         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1495         if (err)
1496                 return err;
1497
1498         return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1499                                                                 GFP_KERNEL);
1500 }
1501
1502 /*
1503  * Read f2fs raw super block.
1504  * Because we have two copies of super block, so read both of them
1505  * to get the first valid one. If any one of them is broken, we pass
1506  * them recovery flag back to the caller.
1507  */
1508 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1509                         struct f2fs_super_block **raw_super,
1510                         int *valid_super_block, int *recovery)
1511 {
1512         struct super_block *sb = sbi->sb;
1513         int block;
1514         struct buffer_head *bh;
1515         struct f2fs_super_block *super;
1516         int err = 0;
1517
1518         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1519         if (!super)
1520                 return -ENOMEM;
1521
1522         for (block = 0; block < 2; block++) {
1523                 bh = sb_bread(sb, block);
1524                 if (!bh) {
1525                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1526                                 block + 1);
1527                         err = -EIO;
1528                         continue;
1529                 }
1530
1531                 /* sanity checking of raw super */
1532                 if (sanity_check_raw_super(sbi, bh)) {
1533                         f2fs_msg(sb, KERN_ERR,
1534                                 "Can't find valid F2FS filesystem in %dth superblock",
1535                                 block + 1);
1536                         err = -EINVAL;
1537                         brelse(bh);
1538                         continue;
1539                 }
1540
1541                 if (!*raw_super) {
1542                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1543                                                         sizeof(*super));
1544                         *valid_super_block = block;
1545                         *raw_super = super;
1546                 }
1547                 brelse(bh);
1548         }
1549
1550         /* Fail to read any one of the superblocks*/
1551         if (err < 0)
1552                 *recovery = 1;
1553
1554         /* No valid superblock */
1555         if (!*raw_super)
1556                 kfree(super);
1557         else
1558                 err = 0;
1559
1560         return err;
1561 }
1562
1563 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1564 {
1565         struct buffer_head *bh;
1566         int err;
1567
1568         if ((recover && f2fs_readonly(sbi->sb)) ||
1569                                 bdev_read_only(sbi->sb->s_bdev)) {
1570                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1571                 return -EROFS;
1572         }
1573
1574         /* write back-up superblock first */
1575         bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1576         if (!bh)
1577                 return -EIO;
1578         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1579         brelse(bh);
1580
1581         /* if we are in recovery path, skip writing valid superblock */
1582         if (recover || err)
1583                 return err;
1584
1585         /* write current valid superblock */
1586         bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1587         if (!bh)
1588                 return -EIO;
1589         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1590         brelse(bh);
1591         return err;
1592 }
1593
1594 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1595 {
1596         struct f2fs_sb_info *sbi;
1597         struct f2fs_super_block *raw_super;
1598         struct inode *root;
1599         int err;
1600         bool retry = true, need_fsck = false;
1601         char *options = NULL;
1602         int recovery, i, valid_super_block;
1603         struct curseg_info *seg_i;
1604
1605 try_onemore:
1606         err = -EINVAL;
1607         raw_super = NULL;
1608         valid_super_block = -1;
1609         recovery = 0;
1610
1611         /* allocate memory for f2fs-specific super block info */
1612         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1613         if (!sbi)
1614                 return -ENOMEM;
1615
1616         sbi->sb = sb;
1617
1618         /* Load the checksum driver */
1619         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1620         if (IS_ERR(sbi->s_chksum_driver)) {
1621                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1622                 err = PTR_ERR(sbi->s_chksum_driver);
1623                 sbi->s_chksum_driver = NULL;
1624                 goto free_sbi;
1625         }
1626
1627         /* set a block size */
1628         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1629                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1630                 goto free_sbi;
1631         }
1632
1633         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1634                                                                 &recovery);
1635         if (err)
1636                 goto free_sbi;
1637
1638         sb->s_fs_info = sbi;
1639         sbi->raw_super = raw_super;
1640
1641         /*
1642          * The BLKZONED feature indicates that the drive was formatted with
1643          * zone alignment optimization. This is optional for host-aware
1644          * devices, but mandatory for host-managed zoned block devices.
1645          */
1646 #ifndef CONFIG_BLK_DEV_ZONED
1647         if (f2fs_sb_mounted_blkzoned(sb)) {
1648                 f2fs_msg(sb, KERN_ERR,
1649                          "Zoned block device support is not enabled\n");
1650                 goto free_sb_buf;
1651         }
1652 #else
1653         if (bdev_zoned_model(sb->s_bdev) == BLK_ZONED_HM &&
1654             !f2fs_sb_mounted_blkzoned(sb)) {
1655                 f2fs_msg(sb, KERN_ERR,
1656                          "Zoned block device feature not enabled\n");
1657                 goto free_sb_buf;
1658         }
1659 #endif
1660
1661         default_options(sbi);
1662         /* parse mount options */
1663         options = kstrdup((const char *)data, GFP_KERNEL);
1664         if (data && !options) {
1665                 err = -ENOMEM;
1666                 goto free_sb_buf;
1667         }
1668
1669         err = parse_options(sb, options);
1670         if (err)
1671                 goto free_options;
1672
1673         sbi->max_file_blocks = max_file_blocks();
1674         sb->s_maxbytes = sbi->max_file_blocks <<
1675                                 le32_to_cpu(raw_super->log_blocksize);
1676         sb->s_max_links = F2FS_LINK_MAX;
1677         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1678
1679         sb->s_op = &f2fs_sops;
1680         sb->s_cop = &f2fs_cryptops;
1681         sb->s_xattr = f2fs_xattr_handlers;
1682         sb->s_export_op = &f2fs_export_ops;
1683         sb->s_magic = F2FS_SUPER_MAGIC;
1684         sb->s_time_gran = 1;
1685         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1686                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1687         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1688
1689         /* init f2fs-specific super block info */
1690         sbi->valid_super_block = valid_super_block;
1691         mutex_init(&sbi->gc_mutex);
1692         mutex_init(&sbi->cp_mutex);
1693         init_rwsem(&sbi->node_write);
1694
1695         /* disallow all the data/node/meta page writes */
1696         set_sbi_flag(sbi, SBI_POR_DOING);
1697         spin_lock_init(&sbi->stat_lock);
1698
1699         init_rwsem(&sbi->read_io.io_rwsem);
1700         sbi->read_io.sbi = sbi;
1701         sbi->read_io.bio = NULL;
1702         for (i = 0; i < NR_PAGE_TYPE; i++) {
1703                 init_rwsem(&sbi->write_io[i].io_rwsem);
1704                 sbi->write_io[i].sbi = sbi;
1705                 sbi->write_io[i].bio = NULL;
1706         }
1707
1708         init_rwsem(&sbi->cp_rwsem);
1709         init_waitqueue_head(&sbi->cp_wait);
1710         init_sb_info(sbi);
1711
1712         err = init_percpu_info(sbi);
1713         if (err)
1714                 goto free_options;
1715
1716         /* get an inode for meta space */
1717         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1718         if (IS_ERR(sbi->meta_inode)) {
1719                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1720                 err = PTR_ERR(sbi->meta_inode);
1721                 goto free_options;
1722         }
1723
1724         err = get_valid_checkpoint(sbi);
1725         if (err) {
1726                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1727                 goto free_meta_inode;
1728         }
1729
1730         sbi->total_valid_node_count =
1731                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1732         percpu_counter_set(&sbi->total_valid_inode_count,
1733                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
1734         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1735         sbi->total_valid_block_count =
1736                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1737         sbi->last_valid_block_count = sbi->total_valid_block_count;
1738
1739         for (i = 0; i < NR_INODE_TYPE; i++) {
1740                 INIT_LIST_HEAD(&sbi->inode_list[i]);
1741                 spin_lock_init(&sbi->inode_lock[i]);
1742         }
1743
1744         init_extent_cache_info(sbi);
1745
1746         init_ino_entry_info(sbi);
1747
1748         /* setup f2fs internal modules */
1749         err = build_segment_manager(sbi);
1750         if (err) {
1751                 f2fs_msg(sb, KERN_ERR,
1752                         "Failed to initialize F2FS segment manager");
1753                 goto free_sm;
1754         }
1755         err = build_node_manager(sbi);
1756         if (err) {
1757                 f2fs_msg(sb, KERN_ERR,
1758                         "Failed to initialize F2FS node manager");
1759                 goto free_nm;
1760         }
1761
1762         /* For write statistics */
1763         if (sb->s_bdev->bd_part)
1764                 sbi->sectors_written_start =
1765                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1766
1767         /* Read accumulated write IO statistics if exists */
1768         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1769         if (__exist_node_summaries(sbi))
1770                 sbi->kbytes_written =
1771                         le64_to_cpu(seg_i->journal->info.kbytes_written);
1772
1773         build_gc_manager(sbi);
1774
1775         /* get an inode for node space */
1776         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1777         if (IS_ERR(sbi->node_inode)) {
1778                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1779                 err = PTR_ERR(sbi->node_inode);
1780                 goto free_nm;
1781         }
1782
1783         f2fs_join_shrinker(sbi);
1784
1785         /* if there are nt orphan nodes free them */
1786         err = recover_orphan_inodes(sbi);
1787         if (err)
1788                 goto free_node_inode;
1789
1790         /* read root inode and dentry */
1791         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1792         if (IS_ERR(root)) {
1793                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1794                 err = PTR_ERR(root);
1795                 goto free_node_inode;
1796         }
1797         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1798                 iput(root);
1799                 err = -EINVAL;
1800                 goto free_node_inode;
1801         }
1802
1803         sb->s_root = d_make_root(root); /* allocate root dentry */
1804         if (!sb->s_root) {
1805                 err = -ENOMEM;
1806                 goto free_root_inode;
1807         }
1808
1809         err = f2fs_build_stats(sbi);
1810         if (err)
1811                 goto free_root_inode;
1812
1813         if (f2fs_proc_root)
1814                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1815
1816         if (sbi->s_proc) {
1817                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1818                                  &f2fs_seq_segment_info_fops, sb);
1819                 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1820                                  &f2fs_seq_segment_bits_fops, sb);
1821         }
1822
1823         sbi->s_kobj.kset = f2fs_kset;
1824         init_completion(&sbi->s_kobj_unregister);
1825         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1826                                                         "%s", sb->s_id);
1827         if (err)
1828                 goto free_proc;
1829
1830         /* recover fsynced data */
1831         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1832                 /*
1833                  * mount should be failed, when device has readonly mode, and
1834                  * previous checkpoint was not done by clean system shutdown.
1835                  */
1836                 if (bdev_read_only(sb->s_bdev) &&
1837                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
1838                         err = -EROFS;
1839                         goto free_kobj;
1840                 }
1841
1842                 if (need_fsck)
1843                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1844
1845                 if (!retry)
1846                         goto skip_recovery;
1847
1848                 err = recover_fsync_data(sbi, false);
1849                 if (err < 0) {
1850                         need_fsck = true;
1851                         f2fs_msg(sb, KERN_ERR,
1852                                 "Cannot recover all fsync data errno=%d", err);
1853                         goto free_kobj;
1854                 }
1855         } else {
1856                 err = recover_fsync_data(sbi, true);
1857
1858                 if (!f2fs_readonly(sb) && err > 0) {
1859                         err = -EINVAL;
1860                         f2fs_msg(sb, KERN_ERR,
1861                                 "Need to recover fsync data");
1862                         goto free_kobj;
1863                 }
1864         }
1865 skip_recovery:
1866         /* recover_fsync_data() cleared this already */
1867         clear_sbi_flag(sbi, SBI_POR_DOING);
1868
1869         /*
1870          * If filesystem is not mounted as read-only then
1871          * do start the gc_thread.
1872          */
1873         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1874                 /* After POR, we can run background GC thread.*/
1875                 err = start_gc_thread(sbi);
1876                 if (err)
1877                         goto free_kobj;
1878         }
1879         kfree(options);
1880
1881         /* recover broken superblock */
1882         if (recovery) {
1883                 err = f2fs_commit_super(sbi, true);
1884                 f2fs_msg(sb, KERN_INFO,
1885                         "Try to recover %dth superblock, ret: %d",
1886                         sbi->valid_super_block ? 1 : 2, err);
1887         }
1888
1889         f2fs_update_time(sbi, CP_TIME);
1890         f2fs_update_time(sbi, REQ_TIME);
1891         return 0;
1892
1893 free_kobj:
1894         f2fs_sync_inode_meta(sbi);
1895         kobject_del(&sbi->s_kobj);
1896         kobject_put(&sbi->s_kobj);
1897         wait_for_completion(&sbi->s_kobj_unregister);
1898 free_proc:
1899         if (sbi->s_proc) {
1900                 remove_proc_entry("segment_info", sbi->s_proc);
1901                 remove_proc_entry("segment_bits", sbi->s_proc);
1902                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1903         }
1904         f2fs_destroy_stats(sbi);
1905 free_root_inode:
1906         dput(sb->s_root);
1907         sb->s_root = NULL;
1908 free_node_inode:
1909         truncate_inode_pages_final(NODE_MAPPING(sbi));
1910         mutex_lock(&sbi->umount_mutex);
1911         release_ino_entry(sbi, true);
1912         f2fs_leave_shrinker(sbi);
1913         /*
1914          * Some dirty meta pages can be produced by recover_orphan_inodes()
1915          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
1916          * followed by write_checkpoint() through f2fs_write_node_pages(), which
1917          * falls into an infinite loop in sync_meta_pages().
1918          */
1919         truncate_inode_pages_final(META_MAPPING(sbi));
1920         iput(sbi->node_inode);
1921         mutex_unlock(&sbi->umount_mutex);
1922 free_nm:
1923         destroy_node_manager(sbi);
1924 free_sm:
1925         destroy_segment_manager(sbi);
1926         kfree(sbi->ckpt);
1927 free_meta_inode:
1928         make_bad_inode(sbi->meta_inode);
1929         iput(sbi->meta_inode);
1930 free_options:
1931         destroy_percpu_info(sbi);
1932         kfree(options);
1933 free_sb_buf:
1934         kfree(raw_super);
1935 free_sbi:
1936         if (sbi->s_chksum_driver)
1937                 crypto_free_shash(sbi->s_chksum_driver);
1938         kfree(sbi);
1939
1940         /* give only one another chance */
1941         if (retry) {
1942                 retry = false;
1943                 shrink_dcache_sb(sb);
1944                 goto try_onemore;
1945         }
1946         return err;
1947 }
1948
1949 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1950                         const char *dev_name, void *data)
1951 {
1952         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1953 }
1954
1955 static void kill_f2fs_super(struct super_block *sb)
1956 {
1957         if (sb->s_root)
1958                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1959         kill_block_super(sb);
1960 }
1961
1962 static struct file_system_type f2fs_fs_type = {
1963         .owner          = THIS_MODULE,
1964         .name           = "f2fs",
1965         .mount          = f2fs_mount,
1966         .kill_sb        = kill_f2fs_super,
1967         .fs_flags       = FS_REQUIRES_DEV,
1968 };
1969 MODULE_ALIAS_FS("f2fs");
1970
1971 static int __init init_inodecache(void)
1972 {
1973         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1974                         sizeof(struct f2fs_inode_info), 0,
1975                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1976         if (!f2fs_inode_cachep)
1977                 return -ENOMEM;
1978         return 0;
1979 }
1980
1981 static void destroy_inodecache(void)
1982 {
1983         /*
1984          * Make sure all delayed rcu free inodes are flushed before we
1985          * destroy cache.
1986          */
1987         rcu_barrier();
1988         kmem_cache_destroy(f2fs_inode_cachep);
1989 }
1990
1991 static int __init init_f2fs_fs(void)
1992 {
1993         int err;
1994
1995         f2fs_build_trace_ios();
1996
1997         err = init_inodecache();
1998         if (err)
1999                 goto fail;
2000         err = create_node_manager_caches();
2001         if (err)
2002                 goto free_inodecache;
2003         err = create_segment_manager_caches();
2004         if (err)
2005                 goto free_node_manager_caches;
2006         err = create_checkpoint_caches();
2007         if (err)
2008                 goto free_segment_manager_caches;
2009         err = create_extent_cache();
2010         if (err)
2011                 goto free_checkpoint_caches;
2012         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
2013         if (!f2fs_kset) {
2014                 err = -ENOMEM;
2015                 goto free_extent_cache;
2016         }
2017         err = register_shrinker(&f2fs_shrinker_info);
2018         if (err)
2019                 goto free_kset;
2020
2021         err = register_filesystem(&f2fs_fs_type);
2022         if (err)
2023                 goto free_shrinker;
2024         err = f2fs_create_root_stats();
2025         if (err)
2026                 goto free_filesystem;
2027         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2028         return 0;
2029
2030 free_filesystem:
2031         unregister_filesystem(&f2fs_fs_type);
2032 free_shrinker:
2033         unregister_shrinker(&f2fs_shrinker_info);
2034 free_kset:
2035         kset_unregister(f2fs_kset);
2036 free_extent_cache:
2037         destroy_extent_cache();
2038 free_checkpoint_caches:
2039         destroy_checkpoint_caches();
2040 free_segment_manager_caches:
2041         destroy_segment_manager_caches();
2042 free_node_manager_caches:
2043         destroy_node_manager_caches();
2044 free_inodecache:
2045         destroy_inodecache();
2046 fail:
2047         return err;
2048 }
2049
2050 static void __exit exit_f2fs_fs(void)
2051 {
2052         remove_proc_entry("fs/f2fs", NULL);
2053         f2fs_destroy_root_stats();
2054         unregister_filesystem(&f2fs_fs_type);
2055         unregister_shrinker(&f2fs_shrinker_info);
2056         kset_unregister(f2fs_kset);
2057         destroy_extent_cache();
2058         destroy_checkpoint_caches();
2059         destroy_segment_manager_caches();
2060         destroy_node_manager_caches();
2061         destroy_inodecache();
2062         f2fs_destroy_trace_ios();
2063 }
2064
2065 module_init(init_f2fs_fs)
2066 module_exit(exit_f2fs_fs)
2067
2068 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2069 MODULE_DESCRIPTION("Flash Friendly File System");
2070 MODULE_LICENSE("GPL");