fs/btrfs/super.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2007 Oracle.  All rights reserved.
   4  */
   5
   6 #include <linux/blkdev.h>
   7 #include <linux/module.h>
   8 #include <linux/fs.h>
   9 #include <linux/pagemap.h>
  10 #include <linux/highmem.h>
  11 #include <linux/time.h>
  12 #include <linux/init.h>
  13 #include <linux/seq_file.h>
  14 #include <linux/string.h>
  15 #include <linux/backing-dev.h>
  16 #include <linux/mount.h>
  17 #include <linux/writeback.h>
  18 #include <linux/statfs.h>
  19 #include <linux/compat.h>
  20 #include <linux/parser.h>
  21 #include <linux/ctype.h>
  22 #include <linux/namei.h>
  23 #include <linux/miscdevice.h>
  24 #include <linux/magic.h>
  25 #include <linux/slab.h>
  26 #include <linux/ratelimit.h>
  27 #include <linux/crc32c.h>
  28 #include <linux/btrfs.h>
  29 #include "messages.h"
  30 #include "delayed-inode.h"
  31 #include "ctree.h"
  32 #include "disk-io.h"
  33 #include "transaction.h"
  34 #include "btrfs_inode.h"
  35 #include "print-tree.h"
  36 #include "props.h"
  37 #include "xattr.h"
  38 #include "bio.h"
  39 #include "export.h"
  40 #include "compression.h"
  41 #include "rcu-string.h"
  42 #include "dev-replace.h"
  43 #include "free-space-cache.h"
  44 #include "backref.h"
  45 #include "space-info.h"
  46 #include "sysfs.h"
  47 #include "zoned.h"
  48 #include "tests/btrfs-tests.h"
  49 #include "block-group.h"
  50 #include "discard.h"
  51 #include "qgroup.h"
  52 #include "raid56.h"
  53 #include "fs.h"
  54 #include "accessors.h"
  55 #include "defrag.h"
  56 #include "dir-item.h"
  57 #include "ioctl.h"
  58 #include "scrub.h"
  59 #include "verity.h"
  60 #include "super.h"
  61 #include "extent-tree.h"
  62 #define CREATE_TRACE_POINTS
  63 #include <trace/events/btrfs.h>
  64
  65 static const struct super_operations btrfs_super_ops;
  66
  67 /*
  68  * Types for mounting the default subvolume and a subvolume explicitly
  69  * requested by subvol=/path. That way the callchain is straightforward and we
  70  * don't have to play tricks with the mount options and recursive calls to
  71  * btrfs_mount.
  72  *
  73  * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
  74  */
  75 static struct file_system_type btrfs_fs_type;
  76 static struct file_system_type btrfs_root_fs_type;
  77
  78 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
  79
  80 static void btrfs_put_super(struct super_block *sb)
  81 {
  82         close_ctree(btrfs_sb(sb));
  83 }
  84
  85 enum {
  86         Opt_acl, Opt_noacl,
  87         Opt_clear_cache,
  88         Opt_commit_interval,
  89         Opt_compress,
  90         Opt_compress_force,
  91         Opt_compress_force_type,
  92         Opt_compress_type,
  93         Opt_degraded,
  94         Opt_device,
  95         Opt_fatal_errors,
  96         Opt_flushoncommit, Opt_noflushoncommit,
  97         Opt_max_inline,
  98         Opt_barrier, Opt_nobarrier,
  99         Opt_datacow, Opt_nodatacow,
 100         Opt_datasum, Opt_nodatasum,
 101         Opt_defrag, Opt_nodefrag,
 102         Opt_discard, Opt_nodiscard,
 103         Opt_discard_mode,
 104         Opt_norecovery,
 105         Opt_ratio,
 106         Opt_rescan_uuid_tree,
 107         Opt_skip_balance,
 108         Opt_space_cache, Opt_no_space_cache,
 109         Opt_space_cache_version,
 110         Opt_ssd, Opt_nossd,
 111         Opt_ssd_spread, Opt_nossd_spread,
 112         Opt_subvol,
 113         Opt_subvol_empty,
 114         Opt_subvolid,
 115         Opt_thread_pool,
 116         Opt_treelog, Opt_notreelog,
 117         Opt_user_subvol_rm_allowed,
 118
 119         /* Rescue options */
 120         Opt_rescue,
 121         Opt_usebackuproot,
 122         Opt_nologreplay,
 123         Opt_ignorebadroots,
 124         Opt_ignoredatacsums,
 125         Opt_rescue_all,
 126
 127         /* Deprecated options */
 128         Opt_recovery,
 129         Opt_inode_cache, Opt_noinode_cache,
 130
 131         /* Debugging options */
 132         Opt_check_integrity,
 133         Opt_check_integrity_including_extent_data,
 134         Opt_check_integrity_print_mask,
 135         Opt_enospc_debug, Opt_noenospc_debug,
 136 #ifdef CONFIG_BTRFS_DEBUG
 137         Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 138 #endif
 139 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 140         Opt_ref_verify,
 141 #endif
 142         Opt_err,
 143 };
 144
 145 static const match_table_t tokens = {
 146         {Opt_acl, "acl"},
 147         {Opt_noacl, "noacl"},
 148         {Opt_clear_cache, "clear_cache"},
 149         {Opt_commit_interval, "commit=%u"},
 150         {Opt_compress, "compress"},
 151         {Opt_compress_type, "compress=%s"},
 152         {Opt_compress_force, "compress-force"},
 153         {Opt_compress_force_type, "compress-force=%s"},
 154         {Opt_degraded, "degraded"},
 155         {Opt_device, "device=%s"},
 156         {Opt_fatal_errors, "fatal_errors=%s"},
 157         {Opt_flushoncommit, "flushoncommit"},
 158         {Opt_noflushoncommit, "noflushoncommit"},
 159         {Opt_inode_cache, "inode_cache"},
 160         {Opt_noinode_cache, "noinode_cache"},
 161         {Opt_max_inline, "max_inline=%s"},
 162         {Opt_barrier, "barrier"},
 163         {Opt_nobarrier, "nobarrier"},
 164         {Opt_datacow, "datacow"},
 165         {Opt_nodatacow, "nodatacow"},
 166         {Opt_datasum, "datasum"},
 167         {Opt_nodatasum, "nodatasum"},
 168         {Opt_defrag, "autodefrag"},
 169         {Opt_nodefrag, "noautodefrag"},
 170         {Opt_discard, "discard"},
 171         {Opt_discard_mode, "discard=%s"},
 172         {Opt_nodiscard, "nodiscard"},
 173         {Opt_norecovery, "norecovery"},
 174         {Opt_ratio, "metadata_ratio=%u"},
 175         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 176         {Opt_skip_balance, "skip_balance"},
 177         {Opt_space_cache, "space_cache"},
 178         {Opt_no_space_cache, "nospace_cache"},
 179         {Opt_space_cache_version, "space_cache=%s"},
 180         {Opt_ssd, "ssd"},
 181         {Opt_nossd, "nossd"},
 182         {Opt_ssd_spread, "ssd_spread"},
 183         {Opt_nossd_spread, "nossd_spread"},
 184         {Opt_subvol, "subvol=%s"},
 185         {Opt_subvol_empty, "subvol="},
 186         {Opt_subvolid, "subvolid=%s"},
 187         {Opt_thread_pool, "thread_pool=%u"},
 188         {Opt_treelog, "treelog"},
 189         {Opt_notreelog, "notreelog"},
 190         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 191
 192         /* Rescue options */
 193         {Opt_rescue, "rescue=%s"},
 194         /* Deprecated, with alias rescue=nologreplay */
 195         {Opt_nologreplay, "nologreplay"},
 196         /* Deprecated, with alias rescue=usebackuproot */
 197         {Opt_usebackuproot, "usebackuproot"},
 198
 199         /* Deprecated options */
 200         {Opt_recovery, "recovery"},
 201
 202         /* Debugging options */
 203         {Opt_check_integrity, "check_int"},
 204         {Opt_check_integrity_including_extent_data, "check_int_data"},
 205         {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
 206         {Opt_enospc_debug, "enospc_debug"},
 207         {Opt_noenospc_debug, "noenospc_debug"},
 208 #ifdef CONFIG_BTRFS_DEBUG
 209         {Opt_fragment_data, "fragment=data"},
 210         {Opt_fragment_metadata, "fragment=metadata"},
 211         {Opt_fragment_all, "fragment=all"},
 212 #endif
 213 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 214         {Opt_ref_verify, "ref_verify"},
 215 #endif
 216         {Opt_err, NULL},
 217 };
 218
 219 static const match_table_t rescue_tokens = {
 220         {Opt_usebackuproot, "usebackuproot"},
 221         {Opt_nologreplay, "nologreplay"},
 222         {Opt_ignorebadroots, "ignorebadroots"},
 223         {Opt_ignorebadroots, "ibadroots"},
 224         {Opt_ignoredatacsums, "ignoredatacsums"},
 225         {Opt_ignoredatacsums, "idatacsums"},
 226         {Opt_rescue_all, "all"},
 227         {Opt_err, NULL},
 228 };
 229
 230 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
 231                             const char *opt_name)
 232 {
 233         if (fs_info->mount_opt & opt) {
 234                 btrfs_err(fs_info, "%s must be used with ro mount option",
 235                           opt_name);
 236                 return true;
 237         }
 238         return false;
 239 }
 240
 241 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
 242 {
 243         char *opts;
 244         char *orig;
 245         char *p;
 246         substring_t args[MAX_OPT_ARGS];
 247         int ret = 0;
 248
 249         opts = kstrdup(options, GFP_KERNEL);
 250         if (!opts)
 251                 return -ENOMEM;
 252         orig = opts;
 253
 254         while ((p = strsep(&opts, ":")) != NULL) {
 255                 int token;
 256
 257                 if (!*p)
 258                         continue;
 259                 token = match_token(p, rescue_tokens, args);
 260                 switch (token){
 261                 case Opt_usebackuproot:
 262                         btrfs_info(info,
 263                                    "trying to use backup root at mount time");
 264                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 265                         break;
 266                 case Opt_nologreplay:
 267                         btrfs_set_and_info(info, NOLOGREPLAY,
 268                                            "disabling log replay at mount time");
 269                         break;
 270                 case Opt_ignorebadroots:
 271                         btrfs_set_and_info(info, IGNOREBADROOTS,
 272                                            "ignoring bad roots");
 273                         break;
 274                 case Opt_ignoredatacsums:
 275                         btrfs_set_and_info(info, IGNOREDATACSUMS,
 276                                            "ignoring data csums");
 277                         break;
 278                 case Opt_rescue_all:
 279                         btrfs_info(info, "enabling all of the rescue options");
 280                         btrfs_set_and_info(info, IGNOREDATACSUMS,
 281                                            "ignoring data csums");
 282                         btrfs_set_and_info(info, IGNOREBADROOTS,
 283                                            "ignoring bad roots");
 284                         btrfs_set_and_info(info, NOLOGREPLAY,
 285                                            "disabling log replay at mount time");
 286                         break;
 287                 case Opt_err:
 288                         btrfs_info(info, "unrecognized rescue option '%s'", p);
 289                         ret = -EINVAL;
 290                         goto out;
 291                 default:
 292                         break;
 293                 }
 294
 295         }
 296 out:
 297         kfree(orig);
 298         return ret;
 299 }
 300
 301 /*
 302  * Regular mount options parser.  Everything that is needed only when
 303  * reading in a new superblock is parsed here.
 304  * XXX JDM: This needs to be cleaned up for remount.
 305  */
 306 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 307                         unsigned long new_flags)
 308 {
 309         substring_t args[MAX_OPT_ARGS];
 310         char *p, *num;
 311         int intarg;
 312         int ret = 0;
 313         char *compress_type;
 314         bool compress_force = false;
 315         enum btrfs_compression_type saved_compress_type;
 316         int saved_compress_level;
 317         bool saved_compress_force;
 318         int no_compress = 0;
 319         const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
 320
 321         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
 322                 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
 323         else if (btrfs_free_space_cache_v1_active(info)) {
 324                 if (btrfs_is_zoned(info)) {
 325                         btrfs_info(info,
 326                         "zoned: clearing existing space cache");
 327                         btrfs_set_super_cache_generation(info->super_copy, 0);
 328                 } else {
 329                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 330                 }
 331         }
 332
 333         /*
 334          * Even the options are empty, we still need to do extra check
 335          * against new flags
 336          */
 337         if (!options)
 338                 goto check;
 339
 340         while ((p = strsep(&options, ",")) != NULL) {
 341                 int token;
 342                 if (!*p)
 343                         continue;
 344
 345                 token = match_token(p, tokens, args);
 346                 switch (token) {
 347                 case Opt_degraded:
 348                         btrfs_info(info, "allowing degraded mounts");
 349                         btrfs_set_opt(info->mount_opt, DEGRADED);
 350                         break;
 351                 case Opt_subvol:
 352                 case Opt_subvol_empty:
 353                 case Opt_subvolid:
 354                 case Opt_device:
 355                         /*
 356                          * These are parsed by btrfs_parse_subvol_options or
 357                          * btrfs_parse_device_options and can be ignored here.
 358                          */
 359                         break;
 360                 case Opt_nodatasum:
 361                         btrfs_set_and_info(info, NODATASUM,
 362                                            "setting nodatasum");
 363                         break;
 364                 case Opt_datasum:
 365                         if (btrfs_test_opt(info, NODATASUM)) {
 366                                 if (btrfs_test_opt(info, NODATACOW))
 367                                         btrfs_info(info,
 368                                                    "setting datasum, datacow enabled");
 369                                 else
 370                                         btrfs_info(info, "setting datasum");
 371                         }
 372                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 373                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 374                         break;
 375                 case Opt_nodatacow:
 376                         if (!btrfs_test_opt(info, NODATACOW)) {
 377                                 if (!btrfs_test_opt(info, COMPRESS) ||
 378                                     !btrfs_test_opt(info, FORCE_COMPRESS)) {
 379                                         btrfs_info(info,
 380                                                    "setting nodatacow, compression disabled");
 381                                 } else {
 382                                         btrfs_info(info, "setting nodatacow");
 383                                 }
 384                         }
 385                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 386                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 387                         btrfs_set_opt(info->mount_opt, NODATACOW);
 388                         btrfs_set_opt(info->mount_opt, NODATASUM);
 389                         break;
 390                 case Opt_datacow:
 391                         btrfs_clear_and_info(info, NODATACOW,
 392                                              "setting datacow");
 393                         break;
 394                 case Opt_compress_force:
 395                 case Opt_compress_force_type:
 396                         compress_force = true;
 397                         fallthrough;
 398                 case Opt_compress:
 399                 case Opt_compress_type:
 400                         saved_compress_type = btrfs_test_opt(info,
 401                                                              COMPRESS) ?
 402                                 info->compress_type : BTRFS_COMPRESS_NONE;
 403                         saved_compress_force =
 404                                 btrfs_test_opt(info, FORCE_COMPRESS);
 405                         saved_compress_level = info->compress_level;
 406                         if (token == Opt_compress ||
 407                             token == Opt_compress_force ||
 408                             strncmp(args[0].from, "zlib", 4) == 0) {
 409                                 compress_type = "zlib";
 410
 411                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 412                                 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 413                                 /*
 414                                  * args[0] contains uninitialized data since
 415                                  * for these tokens we don't expect any
 416                                  * parameter.
 417                                  */
 418                                 if (token != Opt_compress &&
 419                                     token != Opt_compress_force)
 420                                         info->compress_level =
 421                                           btrfs_compress_str2level(
 422                                                         BTRFS_COMPRESS_ZLIB,
 423                                                         args[0].from + 4);
 424                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 425                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 426                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 427                                 no_compress = 0;
 428                         } else if (strncmp(args[0].from, "lzo", 3) == 0) {
 429                                 compress_type = "lzo";
 430                                 info->compress_type = BTRFS_COMPRESS_LZO;
 431                                 info->compress_level = 0;
 432                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 433                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 434                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 435                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 436                                 no_compress = 0;
 437                         } else if (strncmp(args[0].from, "zstd", 4) == 0) {
 438                                 compress_type = "zstd";
 439                                 info->compress_type = BTRFS_COMPRESS_ZSTD;
 440                                 info->compress_level =
 441                                         btrfs_compress_str2level(
 442                                                          BTRFS_COMPRESS_ZSTD,
 443                                                          args[0].from + 4);
 444                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 445                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 446                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 447                                 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
 448                                 no_compress = 0;
 449                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 450                                 compress_type = "no";
 451                                 info->compress_level = 0;
 452                                 info->compress_type = 0;
 453                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 454                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 455                                 compress_force = false;
 456                                 no_compress++;
 457                         } else {
 458                                 btrfs_err(info, "unrecognized compression value %s",
 459                                           args[0].from);
 460                                 ret = -EINVAL;
 461                                 goto out;
 462                         }
 463
 464                         if (compress_force) {
 465                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 466                         } else {
 467                                 /*
 468                                  * If we remount from compress-force=xxx to
 469                                  * compress=xxx, we need clear FORCE_COMPRESS
 470                                  * flag, otherwise, there is no way for users
 471                                  * to disable forcible compression separately.
 472                                  */
 473                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 474                         }
 475                         if (no_compress == 1) {
 476                                 btrfs_info(info, "use no compression");
 477                         } else if ((info->compress_type != saved_compress_type) ||
 478                                    (compress_force != saved_compress_force) ||
 479                                    (info->compress_level != saved_compress_level)) {
 480                                 btrfs_info(info, "%s %s compression, level %d",
 481                                            (compress_force) ? "force" : "use",
 482                                            compress_type, info->compress_level);
 483                         }
 484                         compress_force = false;
 485                         break;
 486                 case Opt_ssd:
 487                         btrfs_set_and_info(info, SSD,
 488                                            "enabling ssd optimizations");
 489                         btrfs_clear_opt(info->mount_opt, NOSSD);
 490                         break;
 491                 case Opt_ssd_spread:
 492                         btrfs_set_and_info(info, SSD,
 493                                            "enabling ssd optimizations");
 494                         btrfs_set_and_info(info, SSD_SPREAD,
 495                                            "using spread ssd allocation scheme");
 496                         btrfs_clear_opt(info->mount_opt, NOSSD);
 497                         break;
 498                 case Opt_nossd:
 499                         btrfs_set_opt(info->mount_opt, NOSSD);
 500                         btrfs_clear_and_info(info, SSD,
 501                                              "not using ssd optimizations");
 502                         fallthrough;
 503                 case Opt_nossd_spread:
 504                         btrfs_clear_and_info(info, SSD_SPREAD,
 505                                              "not using spread ssd allocation scheme");
 506                         break;
 507                 case Opt_barrier:
 508                         btrfs_clear_and_info(info, NOBARRIER,
 509                                              "turning on barriers");
 510                         break;
 511                 case Opt_nobarrier:
 512                         btrfs_set_and_info(info, NOBARRIER,
 513                                            "turning off barriers");
 514                         break;
 515                 case Opt_thread_pool:
 516                         ret = match_int(&args[0], &intarg);
 517                         if (ret) {
 518                                 btrfs_err(info, "unrecognized thread_pool value %s",
 519                                           args[0].from);
 520                                 goto out;
 521                         } else if (intarg == 0) {
 522                                 btrfs_err(info, "invalid value 0 for thread_pool");
 523                                 ret = -EINVAL;
 524                                 goto out;
 525                         }
 526                         info->thread_pool_size = intarg;
 527                         break;
 528                 case Opt_max_inline:
 529                         num = match_strdup(&args[0]);
 530                         if (num) {
 531                                 info->max_inline = memparse(num, NULL);
 532                                 kfree(num);
 533
 534                                 if (info->max_inline) {
 535                                         info->max_inline = min_t(u64,
 536                                                 info->max_inline,
 537                                                 info->sectorsize);
 538                                 }
 539                                 btrfs_info(info, "max_inline at %llu",
 540                                            info->max_inline);
 541                         } else {
 542                                 ret = -ENOMEM;
 543                                 goto out;
 544                         }
 545                         break;
 546                 case Opt_acl:
 547 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 548                         info->sb->s_flags |= SB_POSIXACL;
 549                         break;
 550 #else
 551                         btrfs_err(info, "support for ACL not compiled in!");
 552                         ret = -EINVAL;
 553                         goto out;
 554 #endif
 555                 case Opt_noacl:
 556                         info->sb->s_flags &= ~SB_POSIXACL;
 557                         break;
 558                 case Opt_notreelog:
 559                         btrfs_set_and_info(info, NOTREELOG,
 560                                            "disabling tree log");
 561                         break;
 562                 case Opt_treelog:
 563                         btrfs_clear_and_info(info, NOTREELOG,
 564                                              "enabling tree log");
 565                         break;
 566                 case Opt_norecovery:
 567                 case Opt_nologreplay:
 568                         btrfs_warn(info,
 569                 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 570                         btrfs_set_and_info(info, NOLOGREPLAY,
 571                                            "disabling log replay at mount time");
 572                         break;
 573                 case Opt_flushoncommit:
 574                         btrfs_set_and_info(info, FLUSHONCOMMIT,
 575                                            "turning on flush-on-commit");
 576                         break;
 577                 case Opt_noflushoncommit:
 578                         btrfs_clear_and_info(info, FLUSHONCOMMIT,
 579                                              "turning off flush-on-commit");
 580                         break;
 581                 case Opt_ratio:
 582                         ret = match_int(&args[0], &intarg);
 583                         if (ret) {
 584                                 btrfs_err(info, "unrecognized metadata_ratio value %s",
 585                                           args[0].from);
 586                                 goto out;
 587                         }
 588                         info->metadata_ratio = intarg;
 589                         btrfs_info(info, "metadata ratio %u",
 590                                    info->metadata_ratio);
 591                         break;
 592                 case Opt_discard:
 593                 case Opt_discard_mode:
 594                         if (token == Opt_discard ||
 595                             strcmp(args[0].from, "sync") == 0) {
 596                                 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
 597                                 btrfs_set_and_info(info, DISCARD_SYNC,
 598                                                    "turning on sync discard");
 599                         } else if (strcmp(args[0].from, "async") == 0) {
 600                                 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
 601                                 btrfs_set_and_info(info, DISCARD_ASYNC,
 602                                                    "turning on async discard");
 603                         } else {
 604                                 btrfs_err(info, "unrecognized discard mode value %s",
 605                                           args[0].from);
 606                                 ret = -EINVAL;
 607                                 goto out;
 608                         }
 609                         btrfs_clear_opt(info->mount_opt, NODISCARD);
 610                         break;
 611                 case Opt_nodiscard:
 612                         btrfs_clear_and_info(info, DISCARD_SYNC,
 613                                              "turning off discard");
 614                         btrfs_clear_and_info(info, DISCARD_ASYNC,
 615                                              "turning off async discard");
 616                         btrfs_set_opt(info->mount_opt, NODISCARD);
 617                         break;
 618                 case Opt_space_cache:
 619                 case Opt_space_cache_version:
 620                         /*
 621                          * We already set FREE_SPACE_TREE above because we have
 622                          * compat_ro(FREE_SPACE_TREE) set, and we aren't going
 623                          * to allow v1 to be set for extent tree v2, simply
 624                          * ignore this setting if we're extent tree v2.
 625                          */
 626                         if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
 627                                 break;
 628                         if (token == Opt_space_cache ||
 629                             strcmp(args[0].from, "v1") == 0) {
 630                                 btrfs_clear_opt(info->mount_opt,
 631                                                 FREE_SPACE_TREE);
 632                                 btrfs_set_and_info(info, SPACE_CACHE,
 633                                            "enabling disk space caching");
 634                         } else if (strcmp(args[0].from, "v2") == 0) {
 635                                 btrfs_clear_opt(info->mount_opt,
 636                                                 SPACE_CACHE);
 637                                 btrfs_set_and_info(info, FREE_SPACE_TREE,
 638                                                    "enabling free space tree");
 639                         } else {
 640                                 btrfs_err(info, "unrecognized space_cache value %s",
 641                                           args[0].from);
 642                                 ret = -EINVAL;
 643                                 goto out;
 644                         }
 645                         break;
 646                 case Opt_rescan_uuid_tree:
 647                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 648                         break;
 649                 case Opt_no_space_cache:
 650                         /*
 651                          * We cannot operate without the free space tree with
 652                          * extent tree v2, ignore this option.
 653                          */
 654                         if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
 655                                 break;
 656                         if (btrfs_test_opt(info, SPACE_CACHE)) {
 657                                 btrfs_clear_and_info(info, SPACE_CACHE,
 658                                              "disabling disk space caching");
 659                         }
 660                         if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
 661                                 btrfs_clear_and_info(info, FREE_SPACE_TREE,
 662                                              "disabling free space tree");
 663                         }
 664                         break;
 665                 case Opt_inode_cache:
 666                 case Opt_noinode_cache:
 667                         btrfs_warn(info,
 668         "the 'inode_cache' option is deprecated and has no effect since 5.11");
 669                         break;
 670                 case Opt_clear_cache:
 671                         /*
 672                          * We cannot clear the free space tree with extent tree
 673                          * v2, ignore this option.
 674                          */
 675                         if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
 676                                 break;
 677                         btrfs_set_and_info(info, CLEAR_CACHE,
 678                                            "force clearing of disk cache");
 679                         break;
 680                 case Opt_user_subvol_rm_allowed:
 681                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 682                         break;
 683                 case Opt_enospc_debug:
 684                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 685                         break;
 686                 case Opt_noenospc_debug:
 687                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 688                         break;
 689                 case Opt_defrag:
 690                         btrfs_set_and_info(info, AUTO_DEFRAG,
 691                                            "enabling auto defrag");
 692                         break;
 693                 case Opt_nodefrag:
 694                         btrfs_clear_and_info(info, AUTO_DEFRAG,
 695                                              "disabling auto defrag");
 696                         break;
 697                 case Opt_recovery:
 698                 case Opt_usebackuproot:
 699                         btrfs_warn(info,
 700                         "'%s' is deprecated, use 'rescue=usebackuproot' instead",
 701                                    token == Opt_recovery ? "recovery" :
 702                                    "usebackuproot");
 703                         btrfs_info(info,
 704                                    "trying to use backup root at mount time");
 705                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 706                         break;
 707                 case Opt_skip_balance:
 708                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 709                         break;
 710 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 711                 case Opt_check_integrity_including_extent_data:
 712                         btrfs_info(info,
 713                                    "enabling check integrity including extent data");
 714                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
 715                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 716                         break;
 717                 case Opt_check_integrity:
 718                         btrfs_info(info, "enabling check integrity");
 719                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 720                         break;
 721                 case Opt_check_integrity_print_mask:
 722                         ret = match_int(&args[0], &intarg);
 723                         if (ret) {
 724                                 btrfs_err(info,
 725                                 "unrecognized check_integrity_print_mask value %s",
 726                                         args[0].from);
 727                                 goto out;
 728                         }
 729                         info->check_integrity_print_mask = intarg;
 730                         btrfs_info(info, "check_integrity_print_mask 0x%x",
 731                                    info->check_integrity_print_mask);
 732                         break;
 733 #else
 734                 case Opt_check_integrity_including_extent_data:
 735                 case Opt_check_integrity:
 736                 case Opt_check_integrity_print_mask:
 737                         btrfs_err(info,
 738                                   "support for check_integrity* not compiled in!");
 739                         ret = -EINVAL;
 740                         goto out;
 741 #endif
 742                 case Opt_fatal_errors:
 743                         if (strcmp(args[0].from, "panic") == 0) {
 744                                 btrfs_set_opt(info->mount_opt,
 745                                               PANIC_ON_FATAL_ERROR);
 746                         } else if (strcmp(args[0].from, "bug") == 0) {
 747                                 btrfs_clear_opt(info->mount_opt,
 748                                               PANIC_ON_FATAL_ERROR);
 749                         } else {
 750                                 btrfs_err(info, "unrecognized fatal_errors value %s",
 751                                           args[0].from);
 752                                 ret = -EINVAL;
 753                                 goto out;
 754                         }
 755                         break;
 756                 case Opt_commit_interval:
 757                         intarg = 0;
 758                         ret = match_int(&args[0], &intarg);
 759                         if (ret) {
 760                                 btrfs_err(info, "unrecognized commit_interval value %s",
 761                                           args[0].from);
 762                                 ret = -EINVAL;
 763                                 goto out;
 764                         }
 765                         if (intarg == 0) {
 766                                 btrfs_info(info,
 767                                            "using default commit interval %us",
 768                                            BTRFS_DEFAULT_COMMIT_INTERVAL);
 769                                 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
 770                         } else if (intarg > 300) {
 771                                 btrfs_warn(info, "excessive commit interval %d",
 772                                            intarg);
 773                         }
 774                         info->commit_interval = intarg;
 775                         break;
 776                 case Opt_rescue:
 777                         ret = parse_rescue_options(info, args[0].from);
 778                         if (ret < 0) {
 779                                 btrfs_err(info, "unrecognized rescue value %s",
 780                                           args[0].from);
 781                                 goto out;
 782                         }
 783                         break;
 784 #ifdef CONFIG_BTRFS_DEBUG
 785                 case Opt_fragment_all:
 786                         btrfs_info(info, "fragmenting all space");
 787                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 788                         btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
 789                         break;
 790                 case Opt_fragment_metadata:
 791                         btrfs_info(info, "fragmenting metadata");
 792                         btrfs_set_opt(info->mount_opt,
 793                                       FRAGMENT_METADATA);
 794                         break;
 795                 case Opt_fragment_data:
 796                         btrfs_info(info, "fragmenting data");
 797                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 798                         break;
 799 #endif
 800 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 801                 case Opt_ref_verify:
 802                         btrfs_info(info, "doing ref verification");
 803                         btrfs_set_opt(info->mount_opt, REF_VERIFY);
 804                         break;
 805 #endif
 806                 case Opt_err:
 807                         btrfs_err(info, "unrecognized mount option '%s'", p);
 808                         ret = -EINVAL;
 809                         goto out;
 810                 default:
 811                         break;
 812                 }
 813         }
 814 check:
 815         /* We're read-only, don't have to check. */
 816         if (new_flags & SB_RDONLY)
 817                 goto out;
 818
 819         if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
 820             check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
 821             check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
 822                 ret = -EINVAL;
 823 out:
 824         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 825             !btrfs_test_opt(info, FREE_SPACE_TREE) &&
 826             !btrfs_test_opt(info, CLEAR_CACHE)) {
 827                 btrfs_err(info, "cannot disable free space tree");
 828                 ret = -EINVAL;
 829         }
 830         if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
 831              !btrfs_test_opt(info, FREE_SPACE_TREE)) {
 832                 btrfs_err(info, "cannot disable free space tree with block-group-tree feature");
 833                 ret = -EINVAL;
 834         }
 835         if (!ret)
 836                 ret = btrfs_check_mountopts_zoned(info);
 837         if (!ret && !remounting) {
 838                 if (btrfs_test_opt(info, SPACE_CACHE))
 839                         btrfs_info(info, "disk space caching is enabled");
 840                 if (btrfs_test_opt(info, FREE_SPACE_TREE))
 841                         btrfs_info(info, "using free space tree");
 842         }
 843         return ret;
 844 }
 845
 846 /*
 847  * Parse mount options that are required early in the mount process.
 848  *
 849  * All other options will be parsed on much later in the mount process and
 850  * only when we need to allocate a new super block.
 851  */
 852 static int btrfs_parse_device_options(const char *options, blk_mode_t flags)
 853 {
 854         substring_t args[MAX_OPT_ARGS];
 855         char *device_name, *opts, *orig, *p;
 856         struct btrfs_device *device = NULL;
 857         int error = 0;
 858
 859         lockdep_assert_held(&uuid_mutex);
 860
 861         if (!options)
 862                 return 0;
 863
 864         /*
 865          * strsep changes the string, duplicate it because btrfs_parse_options
 866          * gets called later
 867          */
 868         opts = kstrdup(options, GFP_KERNEL);
 869         if (!opts)
 870                 return -ENOMEM;
 871         orig = opts;
 872
 873         while ((p = strsep(&opts, ",")) != NULL) {
 874                 int token;
 875
 876                 if (!*p)
 877                         continue;
 878
 879                 token = match_token(p, tokens, args);
 880                 if (token == Opt_device) {
 881                         device_name = match_strdup(&args[0]);
 882                         if (!device_name) {
 883                                 error = -ENOMEM;
 884                                 goto out;
 885                         }
 886                         device = btrfs_scan_one_device(device_name, flags);
 887                         kfree(device_name);
 888                         if (IS_ERR(device)) {
 889                                 error = PTR_ERR(device);
 890                                 goto out;
 891                         }
 892                 }
 893         }
 894
 895 out:
 896         kfree(orig);
 897         return error;
 898 }
 899
 900 /*
 901  * Parse mount options that are related to subvolume id
 902  *
 903  * The value is later passed to mount_subvol()
 904  */
 905 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
 906                 u64 *subvol_objectid)
 907 {
 908         substring_t args[MAX_OPT_ARGS];
 909         char *opts, *orig, *p;
 910         int error = 0;
 911         u64 subvolid;
 912
 913         if (!options)
 914                 return 0;
 915
 916         /*
 917          * strsep changes the string, duplicate it because
 918          * btrfs_parse_device_options gets called later
 919          */
 920         opts = kstrdup(options, GFP_KERNEL);
 921         if (!opts)
 922                 return -ENOMEM;
 923         orig = opts;
 924
 925         while ((p = strsep(&opts, ",")) != NULL) {
 926                 int token;
 927                 if (!*p)
 928                         continue;
 929
 930                 token = match_token(p, tokens, args);
 931                 switch (token) {
 932                 case Opt_subvol:
 933                         kfree(*subvol_name);
 934                         *subvol_name = match_strdup(&args[0]);
 935                         if (!*subvol_name) {
 936                                 error = -ENOMEM;
 937                                 goto out;
 938                         }
 939                         break;
 940                 case Opt_subvolid:
 941                         error = match_u64(&args[0], &subvolid);
 942                         if (error)
 943                                 goto out;
 944
 945                         /* we want the original fs_tree */
 946                         if (subvolid == 0)
 947                                 subvolid = BTRFS_FS_TREE_OBJECTID;
 948
 949                         *subvol_objectid = subvolid;
 950                         break;
 951                 default:
 952                         break;
 953                 }
 954         }
 955
 956 out:
 957         kfree(orig);
 958         return error;
 959 }
 960
 961 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 962                                           u64 subvol_objectid)
 963 {
 964         struct btrfs_root *root = fs_info->tree_root;
 965         struct btrfs_root *fs_root = NULL;
 966         struct btrfs_root_ref *root_ref;
 967         struct btrfs_inode_ref *inode_ref;
 968         struct btrfs_key key;
 969         struct btrfs_path *path = NULL;
 970         char *name = NULL, *ptr;
 971         u64 dirid;
 972         int len;
 973         int ret;
 974
 975         path = btrfs_alloc_path();
 976         if (!path) {
 977                 ret = -ENOMEM;
 978                 goto err;
 979         }
 980
 981         name = kmalloc(PATH_MAX, GFP_KERNEL);
 982         if (!name) {
 983                 ret = -ENOMEM;
 984                 goto err;
 985         }
 986         ptr = name + PATH_MAX - 1;
 987         ptr[0] = '\0';
 988
 989         /*
 990          * Walk up the subvolume trees in the tree of tree roots by root
 991          * backrefs until we hit the top-level subvolume.
 992          */
 993         while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
 994                 key.objectid = subvol_objectid;
 995                 key.type = BTRFS_ROOT_BACKREF_KEY;
 996                 key.offset = (u64)-1;
 997
 998                 ret = btrfs_search_backwards(root, &key, path);
 999                 if (ret < 0) {
1000                         goto err;
1001                 } else if (ret > 0) {
1002                         ret = -ENOENT;
1003                         goto err;
1004                 }
1005
1006                 subvol_objectid = key.offset;
1007
1008                 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1009                                           struct btrfs_root_ref);
1010                 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1011                 ptr -= len + 1;
1012                 if (ptr < name) {
1013                         ret = -ENAMETOOLONG;
1014                         goto err;
1015                 }
1016                 read_extent_buffer(path->nodes[0], ptr + 1,
1017                                    (unsigned long)(root_ref + 1), len);
1018                 ptr[0] = '/';
1019                 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1020                 btrfs_release_path(path);
1021
1022                 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1023                 if (IS_ERR(fs_root)) {
1024                         ret = PTR_ERR(fs_root);
1025                         fs_root = NULL;
1026                         goto err;
1027                 }
1028
1029                 /*
1030                  * Walk up the filesystem tree by inode refs until we hit the
1031                  * root directory.
1032                  */
1033                 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1034                         key.objectid = dirid;
1035                         key.type = BTRFS_INODE_REF_KEY;
1036                         key.offset = (u64)-1;
1037
1038                         ret = btrfs_search_backwards(fs_root, &key, path);
1039                         if (ret < 0) {
1040                                 goto err;
1041                         } else if (ret > 0) {
1042                                 ret = -ENOENT;
1043                                 goto err;
1044                         }
1045
1046                         dirid = key.offset;
1047
1048                         inode_ref = btrfs_item_ptr(path->nodes[0],
1049                                                    path->slots[0],
1050                                                    struct btrfs_inode_ref);
1051                         len = btrfs_inode_ref_name_len(path->nodes[0],
1052                                                        inode_ref);
1053                         ptr -= len + 1;
1054                         if (ptr < name) {
1055                                 ret = -ENAMETOOLONG;
1056                                 goto err;
1057                         }
1058                         read_extent_buffer(path->nodes[0], ptr + 1,
1059                                            (unsigned long)(inode_ref + 1), len);
1060                         ptr[0] = '/';
1061                         btrfs_release_path(path);
1062                 }
1063                 btrfs_put_root(fs_root);
1064                 fs_root = NULL;
1065         }
1066
1067         btrfs_free_path(path);
1068         if (ptr == name + PATH_MAX - 1) {
1069                 name[0] = '/';
1070                 name[1] = '\0';
1071         } else {
1072                 memmove(name, ptr, name + PATH_MAX - ptr);
1073         }
1074         return name;
1075
1076 err:
1077         btrfs_put_root(fs_root);
1078         btrfs_free_path(path);
1079         kfree(name);
1080         return ERR_PTR(ret);
1081 }
1082
1083 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1084 {
1085         struct btrfs_root *root = fs_info->tree_root;
1086         struct btrfs_dir_item *di;
1087         struct btrfs_path *path;
1088         struct btrfs_key location;
1089         struct fscrypt_str name = FSTR_INIT("default", 7);
1090         u64 dir_id;
1091
1092         path = btrfs_alloc_path();
1093         if (!path)
1094                 return -ENOMEM;
1095
1096         /*
1097          * Find the "default" dir item which points to the root item that we
1098          * will mount by default if we haven't been given a specific subvolume
1099          * to mount.
1100          */
1101         dir_id = btrfs_super_root_dir(fs_info->super_copy);
1102         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
1103         if (IS_ERR(di)) {
1104                 btrfs_free_path(path);
1105                 return PTR_ERR(di);
1106         }
1107         if (!di) {
1108                 /*
1109                  * Ok the default dir item isn't there.  This is weird since
1110                  * it's always been there, but don't freak out, just try and
1111                  * mount the top-level subvolume.
1112                  */
1113                 btrfs_free_path(path);
1114                 *objectid = BTRFS_FS_TREE_OBJECTID;
1115                 return 0;
1116         }
1117
1118         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1119         btrfs_free_path(path);
1120         *objectid = location.objectid;
1121         return 0;
1122 }
1123
1124 static int btrfs_fill_super(struct super_block *sb,
1125                             struct btrfs_fs_devices *fs_devices,
1126                             void *data)
1127 {
1128         struct inode *inode;
1129         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1130         int err;
1131
1132         sb->s_maxbytes = MAX_LFS_FILESIZE;
1133         sb->s_magic = BTRFS_SUPER_MAGIC;
1134         sb->s_op = &btrfs_super_ops;
1135         sb->s_d_op = &btrfs_dentry_operations;
1136         sb->s_export_op = &btrfs_export_ops;
1137 #ifdef CONFIG_FS_VERITY
1138         sb->s_vop = &btrfs_verityops;
1139 #endif
1140         sb->s_xattr = btrfs_xattr_handlers;
1141         sb->s_time_gran = 1;
1142 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1143         sb->s_flags |= SB_POSIXACL;
1144 #endif
1145         sb->s_flags |= SB_I_VERSION;
1146         sb->s_iflags |= SB_I_CGROUPWB;
1147
1148         err = super_setup_bdi(sb);
1149         if (err) {
1150                 btrfs_err(fs_info, "super_setup_bdi failed");
1151                 return err;
1152         }
1153
1154         err = open_ctree(sb, fs_devices, (char *)data);
1155         if (err) {
1156                 btrfs_err(fs_info, "open_ctree failed");
1157                 return err;
1158         }
1159
1160         inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1161         if (IS_ERR(inode)) {
1162                 err = PTR_ERR(inode);
1163                 btrfs_handle_fs_error(fs_info, err, NULL);
1164                 goto fail_close;
1165         }
1166
1167         sb->s_root = d_make_root(inode);
1168         if (!sb->s_root) {
1169                 err = -ENOMEM;
1170                 goto fail_close;
1171         }
1172
1173         sb->s_flags |= SB_ACTIVE;
1174         return 0;
1175
1176 fail_close:
1177         close_ctree(fs_info);
1178         return err;
1179 }
1180
1181 int btrfs_sync_fs(struct super_block *sb, int wait)
1182 {
1183         struct btrfs_trans_handle *trans;
1184         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1185         struct btrfs_root *root = fs_info->tree_root;
1186
1187         trace_btrfs_sync_fs(fs_info, wait);
1188
1189         if (!wait) {
1190                 filemap_flush(fs_info->btree_inode->i_mapping);
1191                 return 0;
1192         }
1193
1194         btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1195
1196         trans = btrfs_attach_transaction_barrier(root);
1197         if (IS_ERR(trans)) {
1198                 /* no transaction, don't bother */
1199                 if (PTR_ERR(trans) == -ENOENT) {
1200                         /*
1201                          * Exit unless we have some pending changes
1202                          * that need to go through commit
1203                          */
1204                         if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1205                                       &fs_info->flags))
1206                                 return 0;
1207                         /*
1208                          * A non-blocking test if the fs is frozen. We must not
1209                          * start a new transaction here otherwise a deadlock
1210                          * happens. The pending operations are delayed to the
1211                          * next commit after thawing.
1212                          */
1213                         if (sb_start_write_trylock(sb))
1214                                 sb_end_write(sb);
1215                         else
1216                                 return 0;
1217                         trans = btrfs_start_transaction(root, 0);
1218                 }
1219                 if (IS_ERR(trans))
1220                         return PTR_ERR(trans);
1221         }
1222         return btrfs_commit_transaction(trans);
1223 }
1224
1225 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1226 {
1227         seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1228         *printed = true;
1229 }
1230
1231 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1232 {
1233         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1234         const char *compress_type;
1235         const char *subvol_name;
1236         bool printed = false;
1237
1238         if (btrfs_test_opt(info, DEGRADED))
1239                 seq_puts(seq, ",degraded");
1240         if (btrfs_test_opt(info, NODATASUM))
1241                 seq_puts(seq, ",nodatasum");
1242         if (btrfs_test_opt(info, NODATACOW))
1243                 seq_puts(seq, ",nodatacow");
1244         if (btrfs_test_opt(info, NOBARRIER))
1245                 seq_puts(seq, ",nobarrier");
1246         if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1247                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1248         if (info->thread_pool_size !=  min_t(unsigned long,
1249                                              num_online_cpus() + 2, 8))
1250                 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1251         if (btrfs_test_opt(info, COMPRESS)) {
1252                 compress_type = btrfs_compress_type2str(info->compress_type);
1253                 if (btrfs_test_opt(info, FORCE_COMPRESS))
1254                         seq_printf(seq, ",compress-force=%s", compress_type);
1255                 else
1256                         seq_printf(seq, ",compress=%s", compress_type);
1257                 if (info->compress_level)
1258                         seq_printf(seq, ":%d", info->compress_level);
1259         }
1260         if (btrfs_test_opt(info, NOSSD))
1261                 seq_puts(seq, ",nossd");
1262         if (btrfs_test_opt(info, SSD_SPREAD))
1263                 seq_puts(seq, ",ssd_spread");
1264         else if (btrfs_test_opt(info, SSD))
1265                 seq_puts(seq, ",ssd");
1266         if (btrfs_test_opt(info, NOTREELOG))
1267                 seq_puts(seq, ",notreelog");
1268         if (btrfs_test_opt(info, NOLOGREPLAY))
1269                 print_rescue_option(seq, "nologreplay", &printed);
1270         if (btrfs_test_opt(info, USEBACKUPROOT))
1271                 print_rescue_option(seq, "usebackuproot", &printed);
1272         if (btrfs_test_opt(info, IGNOREBADROOTS))
1273                 print_rescue_option(seq, "ignorebadroots", &printed);
1274         if (btrfs_test_opt(info, IGNOREDATACSUMS))
1275                 print_rescue_option(seq, "ignoredatacsums", &printed);
1276         if (btrfs_test_opt(info, FLUSHONCOMMIT))
1277                 seq_puts(seq, ",flushoncommit");
1278         if (btrfs_test_opt(info, DISCARD_SYNC))
1279                 seq_puts(seq, ",discard");
1280         if (btrfs_test_opt(info, DISCARD_ASYNC))
1281                 seq_puts(seq, ",discard=async");
1282         if (!(info->sb->s_flags & SB_POSIXACL))
1283                 seq_puts(seq, ",noacl");
1284         if (btrfs_free_space_cache_v1_active(info))
1285                 seq_puts(seq, ",space_cache");
1286         else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1287                 seq_puts(seq, ",space_cache=v2");
1288         else
1289                 seq_puts(seq, ",nospace_cache");
1290         if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1291                 seq_puts(seq, ",rescan_uuid_tree");
1292         if (btrfs_test_opt(info, CLEAR_CACHE))
1293                 seq_puts(seq, ",clear_cache");
1294         if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1295                 seq_puts(seq, ",user_subvol_rm_allowed");
1296         if (btrfs_test_opt(info, ENOSPC_DEBUG))
1297                 seq_puts(seq, ",enospc_debug");
1298         if (btrfs_test_opt(info, AUTO_DEFRAG))
1299                 seq_puts(seq, ",autodefrag");
1300         if (btrfs_test_opt(info, SKIP_BALANCE))
1301                 seq_puts(seq, ",skip_balance");
1302 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1303         if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1304                 seq_puts(seq, ",check_int_data");
1305         else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1306                 seq_puts(seq, ",check_int");
1307         if (info->check_integrity_print_mask)
1308                 seq_printf(seq, ",check_int_print_mask=%d",
1309                                 info->check_integrity_print_mask);
1310 #endif
1311         if (info->metadata_ratio)
1312                 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1313         if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1314                 seq_puts(seq, ",fatal_errors=panic");
1315         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1316                 seq_printf(seq, ",commit=%u", info->commit_interval);
1317 #ifdef CONFIG_BTRFS_DEBUG
1318         if (btrfs_test_opt(info, FRAGMENT_DATA))
1319                 seq_puts(seq, ",fragment=data");
1320         if (btrfs_test_opt(info, FRAGMENT_METADATA))
1321                 seq_puts(seq, ",fragment=metadata");
1322 #endif
1323         if (btrfs_test_opt(info, REF_VERIFY))
1324                 seq_puts(seq, ",ref_verify");
1325         seq_printf(seq, ",subvolid=%llu",
1326                   BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1327         subvol_name = btrfs_get_subvol_name_from_objectid(info,
1328                         BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1329         if (!IS_ERR(subvol_name)) {
1330                 seq_puts(seq, ",subvol=");
1331                 seq_escape(seq, subvol_name, " \t\n\\");
1332                 kfree(subvol_name);
1333         }
1334         return 0;
1335 }
1336
1337 static int btrfs_test_super(struct super_block *s, void *data)
1338 {
1339         struct btrfs_fs_info *p = data;
1340         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1341
1342         return fs_info->fs_devices == p->fs_devices;
1343 }
1344
1345 static int btrfs_set_super(struct super_block *s, void *data)
1346 {
1347         int err = set_anon_super(s, data);
1348         if (!err)
1349                 s->s_fs_info = data;
1350         return err;
1351 }
1352
1353 /*
1354  * subvolumes are identified by ino 256
1355  */
1356 static inline int is_subvolume_inode(struct inode *inode)
1357 {
1358         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1359                 return 1;
1360         return 0;
1361 }
1362
1363 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1364                                    struct vfsmount *mnt)
1365 {
1366         struct dentry *root;
1367         int ret;
1368
1369         if (!subvol_name) {
1370                 if (!subvol_objectid) {
1371                         ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1372                                                           &subvol_objectid);
1373                         if (ret) {
1374                                 root = ERR_PTR(ret);
1375                                 goto out;
1376                         }
1377                 }
1378                 subvol_name = btrfs_get_subvol_name_from_objectid(
1379                                         btrfs_sb(mnt->mnt_sb), subvol_objectid);
1380                 if (IS_ERR(subvol_name)) {
1381                         root = ERR_CAST(subvol_name);
1382                         subvol_name = NULL;
1383                         goto out;
1384                 }
1385
1386         }
1387
1388         root = mount_subtree(mnt, subvol_name);
1389         /* mount_subtree() drops our reference on the vfsmount. */
1390         mnt = NULL;
1391
1392         if (!IS_ERR(root)) {
1393                 struct super_block *s = root->d_sb;
1394                 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1395                 struct inode *root_inode = d_inode(root);
1396                 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1397
1398                 ret = 0;
1399                 if (!is_subvolume_inode(root_inode)) {
1400                         btrfs_err(fs_info, "'%s' is not a valid subvolume",
1401                                subvol_name);
1402                         ret = -EINVAL;
1403                 }
1404                 if (subvol_objectid && root_objectid != subvol_objectid) {
1405                         /*
1406                          * This will also catch a race condition where a
1407                          * subvolume which was passed by ID is renamed and
1408                          * another subvolume is renamed over the old location.
1409                          */
1410                         btrfs_err(fs_info,
1411                                   "subvol '%s' does not match subvolid %llu",
1412                                   subvol_name, subvol_objectid);
1413                         ret = -EINVAL;
1414                 }
1415                 if (ret) {
1416                         dput(root);
1417                         root = ERR_PTR(ret);
1418                         deactivate_locked_super(s);
1419                 }
1420         }
1421
1422 out:
1423         mntput(mnt);
1424         kfree(subvol_name);
1425         return root;
1426 }
1427
1428 /*
1429  * Find a superblock for the given device / mount point.
1430  *
1431  * Note: This is based on mount_bdev from fs/super.c with a few additions
1432  *       for multiple device setup.  Make sure to keep it in sync.
1433  */
1434 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1435                 int flags, const char *device_name, void *data)
1436 {
1437         struct block_device *bdev = NULL;
1438         struct super_block *s;
1439         struct btrfs_device *device = NULL;
1440         struct btrfs_fs_devices *fs_devices = NULL;
1441         struct btrfs_fs_info *fs_info = NULL;
1442         void *new_sec_opts = NULL;
1443         blk_mode_t mode = sb_open_mode(flags);
1444         int error = 0;
1445
1446         if (data) {
1447                 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1448                 if (error)
1449                         return ERR_PTR(error);
1450         }
1451
1452         /*
1453          * Setup a dummy root and fs_info for test/set super.  This is because
1454          * we don't actually fill this stuff out until open_ctree, but we need
1455          * then open_ctree will properly initialize the file system specific
1456          * settings later.  btrfs_init_fs_info initializes the static elements
1457          * of the fs_info (locks and such) to make cleanup easier if we find a
1458          * superblock with our given fs_devices later on at sget() time.
1459          */
1460         fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1461         if (!fs_info) {
1462                 error = -ENOMEM;
1463                 goto error_sec_opts;
1464         }
1465         btrfs_init_fs_info(fs_info);
1466
1467         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1468         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1469         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1470                 error = -ENOMEM;
1471                 goto error_fs_info;
1472         }
1473
1474         mutex_lock(&uuid_mutex);
1475         error = btrfs_parse_device_options(data, mode);
1476         if (error) {
1477                 mutex_unlock(&uuid_mutex);
1478                 goto error_fs_info;
1479         }
1480
1481         device = btrfs_scan_one_device(device_name, mode);
1482         if (IS_ERR(device)) {
1483                 mutex_unlock(&uuid_mutex);
1484                 error = PTR_ERR(device);
1485                 goto error_fs_info;
1486         }
1487
1488         fs_devices = device->fs_devices;
1489         fs_info->fs_devices = fs_devices;
1490
1491         error = btrfs_open_devices(fs_devices, mode, fs_type);
1492         mutex_unlock(&uuid_mutex);
1493         if (error)
1494                 goto error_fs_info;
1495
1496         if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1497                 error = -EACCES;
1498                 goto error_close_devices;
1499         }
1500
1501         bdev = fs_devices->latest_dev->bdev;
1502         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1503                  fs_info);
1504         if (IS_ERR(s)) {
1505                 error = PTR_ERR(s);
1506                 goto error_close_devices;
1507         }
1508
1509         if (s->s_root) {
1510                 btrfs_close_devices(fs_devices);
1511                 btrfs_free_fs_info(fs_info);
1512                 if ((flags ^ s->s_flags) & SB_RDONLY)
1513                         error = -EBUSY;
1514         } else {
1515                 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1516                 shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
1517                                         s->s_id);
1518                 btrfs_sb(s)->bdev_holder = fs_type;
1519                 error = btrfs_fill_super(s, fs_devices, data);
1520         }
1521         if (!error)
1522                 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1523         security_free_mnt_opts(&new_sec_opts);
1524         if (error) {
1525                 deactivate_locked_super(s);
1526                 return ERR_PTR(error);
1527         }
1528
1529         return dget(s->s_root);
1530
1531 error_close_devices:
1532         btrfs_close_devices(fs_devices);
1533 error_fs_info:
1534         btrfs_free_fs_info(fs_info);
1535 error_sec_opts:
1536         security_free_mnt_opts(&new_sec_opts);
1537         return ERR_PTR(error);
1538 }
1539
1540 /*
1541  * Mount function which is called by VFS layer.
1542  *
1543  * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1544  * which needs vfsmount* of device's root (/).  This means device's root has to
1545  * be mounted internally in any case.
1546  *
1547  * Operation flow:
1548  *   1. Parse subvol id related options for later use in mount_subvol().
1549  *
1550  *   2. Mount device's root (/) by calling vfs_kern_mount().
1551  *
1552  *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1553  *      first place. In order to avoid calling btrfs_mount() again, we use
1554  *      different file_system_type which is not registered to VFS by
1555  *      register_filesystem() (btrfs_root_fs_type). As a result,
1556  *      btrfs_mount_root() is called. The return value will be used by
1557  *      mount_subtree() in mount_subvol().
1558  *
1559  *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1560  *      "btrfs subvolume set-default", mount_subvol() is called always.
1561  */
1562 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1563                 const char *device_name, void *data)
1564 {
1565         struct vfsmount *mnt_root;
1566         struct dentry *root;
1567         char *subvol_name = NULL;
1568         u64 subvol_objectid = 0;
1569         int error = 0;
1570
1571         error = btrfs_parse_subvol_options(data, &subvol_name,
1572                                         &subvol_objectid);
1573         if (error) {
1574                 kfree(subvol_name);
1575                 return ERR_PTR(error);
1576         }
1577
1578         /* mount device's root (/) */
1579         mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1580         if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1581                 if (flags & SB_RDONLY) {
1582                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1583                                 flags & ~SB_RDONLY, device_name, data);
1584                 } else {
1585                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1586                                 flags | SB_RDONLY, device_name, data);
1587                         if (IS_ERR(mnt_root)) {
1588                                 root = ERR_CAST(mnt_root);
1589                                 kfree(subvol_name);
1590                                 goto out;
1591                         }
1592
1593                         down_write(&mnt_root->mnt_sb->s_umount);
1594                         error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1595                         up_write(&mnt_root->mnt_sb->s_umount);
1596                         if (error < 0) {
1597                                 root = ERR_PTR(error);
1598                                 mntput(mnt_root);
1599                                 kfree(subvol_name);
1600                                 goto out;
1601                         }
1602                 }
1603         }
1604         if (IS_ERR(mnt_root)) {
1605                 root = ERR_CAST(mnt_root);
1606                 kfree(subvol_name);
1607                 goto out;
1608         }
1609
1610         /* mount_subvol() will free subvol_name and mnt_root */
1611         root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1612
1613 out:
1614         return root;
1615 }
1616
1617 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1618                                      u32 new_pool_size, u32 old_pool_size)
1619 {
1620         if (new_pool_size == old_pool_size)
1621                 return;
1622
1623         fs_info->thread_pool_size = new_pool_size;
1624
1625         btrfs_info(fs_info, "resize thread pool %d -> %d",
1626                old_pool_size, new_pool_size);
1627
1628         btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1629         btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1630         btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1631         workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1632         workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1633         btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1634         btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1635         btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1636 }
1637
1638 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1639                                        unsigned long old_opts, int flags)
1640 {
1641         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1642             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1643              (flags & SB_RDONLY))) {
1644                 /* wait for any defraggers to finish */
1645                 wait_event(fs_info->transaction_wait,
1646                            (atomic_read(&fs_info->defrag_running) == 0));
1647                 if (flags & SB_RDONLY)
1648                         sync_filesystem(fs_info->sb);
1649         }
1650 }
1651
1652 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1653                                          unsigned long old_opts)
1654 {
1655         const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1656
1657         /*
1658          * We need to cleanup all defragable inodes if the autodefragment is
1659          * close or the filesystem is read only.
1660          */
1661         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1662             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1663                 btrfs_cleanup_defrag_inodes(fs_info);
1664         }
1665
1666         /* If we toggled discard async */
1667         if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1668             btrfs_test_opt(fs_info, DISCARD_ASYNC))
1669                 btrfs_discard_resume(fs_info);
1670         else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1671                  !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1672                 btrfs_discard_cleanup(fs_info);
1673
1674         /* If we toggled space cache */
1675         if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1676                 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1677 }
1678
1679 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1680 {
1681         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1682         unsigned old_flags = sb->s_flags;
1683         unsigned long old_opts = fs_info->mount_opt;
1684         unsigned long old_compress_type = fs_info->compress_type;
1685         u64 old_max_inline = fs_info->max_inline;
1686         u32 old_thread_pool_size = fs_info->thread_pool_size;
1687         u32 old_metadata_ratio = fs_info->metadata_ratio;
1688         int ret;
1689
1690         sync_filesystem(sb);
1691         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1692
1693         if (data) {
1694                 void *new_sec_opts = NULL;
1695
1696                 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1697                 if (!ret)
1698                         ret = security_sb_remount(sb, new_sec_opts);
1699                 security_free_mnt_opts(&new_sec_opts);
1700                 if (ret)
1701                         goto restore;
1702         }
1703
1704         ret = btrfs_parse_options(fs_info, data, *flags);
1705         if (ret)
1706                 goto restore;
1707
1708         ret = btrfs_check_features(fs_info, !(*flags & SB_RDONLY));
1709         if (ret < 0)
1710                 goto restore;
1711
1712         btrfs_remount_begin(fs_info, old_opts, *flags);
1713         btrfs_resize_thread_pool(fs_info,
1714                 fs_info->thread_pool_size, old_thread_pool_size);
1715
1716         if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1717             (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1718             (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1719                 btrfs_warn(fs_info,
1720                 "remount supports changing free space tree only from ro to rw");
1721                 /* Make sure free space cache options match the state on disk */
1722                 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1723                         btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1724                         btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1725                 }
1726                 if (btrfs_free_space_cache_v1_active(fs_info)) {
1727                         btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1728                         btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1729                 }
1730         }
1731
1732         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1733                 goto out;
1734
1735         if (*flags & SB_RDONLY) {
1736                 /*
1737                  * this also happens on 'umount -rf' or on shutdown, when
1738                  * the filesystem is busy.
1739                  */
1740                 cancel_work_sync(&fs_info->async_reclaim_work);
1741                 cancel_work_sync(&fs_info->async_data_reclaim_work);
1742
1743                 btrfs_discard_cleanup(fs_info);
1744
1745                 /* wait for the uuid_scan task to finish */
1746                 down(&fs_info->uuid_tree_rescan_sem);
1747                 /* avoid complains from lockdep et al. */
1748                 up(&fs_info->uuid_tree_rescan_sem);
1749
1750                 btrfs_set_sb_rdonly(sb);
1751
1752                 /*
1753                  * Setting SB_RDONLY will put the cleaner thread to
1754                  * sleep at the next loop if it's already active.
1755                  * If it's already asleep, we'll leave unused block
1756                  * groups on disk until we're mounted read-write again
1757                  * unless we clean them up here.
1758                  */
1759                 btrfs_delete_unused_bgs(fs_info);
1760
1761                 /*
1762                  * The cleaner task could be already running before we set the
1763                  * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1764                  * We must make sure that after we finish the remount, i.e. after
1765                  * we call btrfs_commit_super(), the cleaner can no longer start
1766                  * a transaction - either because it was dropping a dead root,
1767                  * running delayed iputs or deleting an unused block group (the
1768                  * cleaner picked a block group from the list of unused block
1769                  * groups before we were able to in the previous call to
1770                  * btrfs_delete_unused_bgs()).
1771                  */
1772                 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1773                             TASK_UNINTERRUPTIBLE);
1774
1775                 /*
1776                  * We've set the superblock to RO mode, so we might have made
1777                  * the cleaner task sleep without running all pending delayed
1778                  * iputs. Go through all the delayed iputs here, so that if an
1779                  * unmount happens without remounting RW we don't end up at
1780                  * finishing close_ctree() with a non-empty list of delayed
1781                  * iputs.
1782                  */
1783                 btrfs_run_delayed_iputs(fs_info);
1784
1785                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1786                 btrfs_scrub_cancel(fs_info);
1787                 btrfs_pause_balance(fs_info);
1788
1789                 /*
1790                  * Pause the qgroup rescan worker if it is running. We don't want
1791                  * it to be still running after we are in RO mode, as after that,
1792                  * by the time we unmount, it might have left a transaction open,
1793                  * so we would leak the transaction and/or crash.
1794                  */
1795                 btrfs_qgroup_wait_for_completion(fs_info, false);
1796
1797                 ret = btrfs_commit_super(fs_info);
1798                 if (ret)
1799                         goto restore;
1800         } else {
1801                 if (BTRFS_FS_ERROR(fs_info)) {
1802                         btrfs_err(fs_info,
1803                                 "Remounting read-write after error is not allowed");
1804                         ret = -EINVAL;
1805                         goto restore;
1806                 }
1807                 if (fs_info->fs_devices->rw_devices == 0) {
1808                         ret = -EACCES;
1809                         goto restore;
1810                 }
1811
1812                 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1813                         btrfs_warn(fs_info,
1814                 "too many missing devices, writable remount is not allowed");
1815                         ret = -EACCES;
1816                         goto restore;
1817                 }
1818
1819                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1820                         btrfs_warn(fs_info,
1821                 "mount required to replay tree-log, cannot remount read-write");
1822                         ret = -EINVAL;
1823                         goto restore;
1824                 }
1825
1826                 /*
1827                  * NOTE: when remounting with a change that does writes, don't
1828                  * put it anywhere above this point, as we are not sure to be
1829                  * safe to write until we pass the above checks.
1830                  */
1831                 ret = btrfs_start_pre_rw_mount(fs_info);
1832                 if (ret)
1833                         goto restore;
1834
1835                 btrfs_clear_sb_rdonly(sb);
1836
1837                 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1838
1839                 /*
1840                  * If we've gone from readonly -> read/write, we need to get
1841                  * our sync/async discard lists in the right state.
1842                  */
1843                 btrfs_discard_resume(fs_info);
1844         }
1845 out:
1846         /*
1847          * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1848          * since the absence of the flag means it can be toggled off by remount.
1849          */
1850         *flags |= SB_I_VERSION;
1851
1852         wake_up_process(fs_info->transaction_kthread);
1853         btrfs_remount_cleanup(fs_info, old_opts);
1854         btrfs_clear_oneshot_options(fs_info);
1855         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1856
1857         return 0;
1858
1859 restore:
1860         /* We've hit an error - don't reset SB_RDONLY */
1861         if (sb_rdonly(sb))
1862                 old_flags |= SB_RDONLY;
1863         if (!(old_flags & SB_RDONLY))
1864                 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
1865         sb->s_flags = old_flags;
1866         fs_info->mount_opt = old_opts;
1867         fs_info->compress_type = old_compress_type;
1868         fs_info->max_inline = old_max_inline;
1869         btrfs_resize_thread_pool(fs_info,
1870                 old_thread_pool_size, fs_info->thread_pool_size);
1871         fs_info->metadata_ratio = old_metadata_ratio;
1872         btrfs_remount_cleanup(fs_info, old_opts);
1873         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1874
1875         return ret;
1876 }
1877
1878 /* Used to sort the devices by max_avail(descending sort) */
1879 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1880 {
1881         const struct btrfs_device_info *dev_info1 = a;
1882         const struct btrfs_device_info *dev_info2 = b;
1883
1884         if (dev_info1->max_avail > dev_info2->max_avail)
1885                 return -1;
1886         else if (dev_info1->max_avail < dev_info2->max_avail)
1887                 return 1;
1888         return 0;
1889 }
1890
1891 /*
1892  * sort the devices by max_avail, in which max free extent size of each device
1893  * is stored.(Descending Sort)
1894  */
1895 static inline void btrfs_descending_sort_devices(
1896                                         struct btrfs_device_info *devices,
1897                                         size_t nr_devices)
1898 {
1899         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1900              btrfs_cmp_device_free_bytes, NULL);
1901 }
1902
1903 /*
1904  * The helper to calc the free space on the devices that can be used to store
1905  * file data.
1906  */
1907 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1908                                               u64 *free_bytes)
1909 {
1910         struct btrfs_device_info *devices_info;
1911         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1912         struct btrfs_device *device;
1913         u64 type;
1914         u64 avail_space;
1915         u64 min_stripe_size;
1916         int num_stripes = 1;
1917         int i = 0, nr_devices;
1918         const struct btrfs_raid_attr *rattr;
1919
1920         /*
1921          * We aren't under the device list lock, so this is racy-ish, but good
1922          * enough for our purposes.
1923          */
1924         nr_devices = fs_info->fs_devices->open_devices;
1925         if (!nr_devices) {
1926                 smp_mb();
1927                 nr_devices = fs_info->fs_devices->open_devices;
1928                 ASSERT(nr_devices);
1929                 if (!nr_devices) {
1930                         *free_bytes = 0;
1931                         return 0;
1932                 }
1933         }
1934
1935         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1936                                GFP_KERNEL);
1937         if (!devices_info)
1938                 return -ENOMEM;
1939
1940         /* calc min stripe number for data space allocation */
1941         type = btrfs_data_alloc_profile(fs_info);
1942         rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1943
1944         if (type & BTRFS_BLOCK_GROUP_RAID0)
1945                 num_stripes = nr_devices;
1946         else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1947                 num_stripes = rattr->ncopies;
1948         else if (type & BTRFS_BLOCK_GROUP_RAID10)
1949                 num_stripes = 4;
1950
1951         /* Adjust for more than 1 stripe per device */
1952         min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1953
1954         rcu_read_lock();
1955         list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1956                 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1957                                                 &device->dev_state) ||
1958                     !device->bdev ||
1959                     test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1960                         continue;
1961
1962                 if (i >= nr_devices)
1963                         break;
1964
1965                 avail_space = device->total_bytes - device->bytes_used;
1966
1967                 /* align with stripe_len */
1968                 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1969
1970                 /*
1971                  * Ensure we have at least min_stripe_size on top of the
1972                  * reserved space on the device.
1973                  */
1974                 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1975                         continue;
1976
1977                 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1978
1979                 devices_info[i].dev = device;
1980                 devices_info[i].max_avail = avail_space;
1981
1982                 i++;
1983         }
1984         rcu_read_unlock();
1985
1986         nr_devices = i;
1987
1988         btrfs_descending_sort_devices(devices_info, nr_devices);
1989
1990         i = nr_devices - 1;
1991         avail_space = 0;
1992         while (nr_devices >= rattr->devs_min) {
1993                 num_stripes = min(num_stripes, nr_devices);
1994
1995                 if (devices_info[i].max_avail >= min_stripe_size) {
1996                         int j;
1997                         u64 alloc_size;
1998
1999                         avail_space += devices_info[i].max_avail * num_stripes;
2000                         alloc_size = devices_info[i].max_avail;
2001                         for (j = i + 1 - num_stripes; j <= i; j++)
2002                                 devices_info[j].max_avail -= alloc_size;
2003                 }
2004                 i--;
2005                 nr_devices--;
2006         }
2007
2008         kfree(devices_info);
2009         *free_bytes = avail_space;
2010         return 0;
2011 }
2012
2013 /*
2014  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2015  *
2016  * If there's a redundant raid level at DATA block groups, use the respective
2017  * multiplier to scale the sizes.
2018  *
2019  * Unused device space usage is based on simulating the chunk allocator
2020  * algorithm that respects the device sizes and order of allocations.  This is
2021  * a close approximation of the actual use but there are other factors that may
2022  * change the result (like a new metadata chunk).
2023  *
2024  * If metadata is exhausted, f_bavail will be 0.
2025  */
2026 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2027 {
2028         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2029         struct btrfs_super_block *disk_super = fs_info->super_copy;
2030         struct btrfs_space_info *found;
2031         u64 total_used = 0;
2032         u64 total_free_data = 0;
2033         u64 total_free_meta = 0;
2034         u32 bits = fs_info->sectorsize_bits;
2035         __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2036         unsigned factor = 1;
2037         struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2038         int ret;
2039         u64 thresh = 0;
2040         int mixed = 0;
2041
2042         list_for_each_entry(found, &fs_info->space_info, list) {
2043                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2044                         int i;
2045
2046                         total_free_data += found->disk_total - found->disk_used;
2047                         total_free_data -=
2048                                 btrfs_account_ro_block_groups_free_space(found);
2049
2050                         for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2051                                 if (!list_empty(&found->block_groups[i]))
2052                                         factor = btrfs_bg_type_to_factor(
2053                                                 btrfs_raid_array[i].bg_flag);
2054                         }
2055                 }
2056
2057                 /*
2058                  * Metadata in mixed block group profiles are accounted in data
2059                  */
2060                 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2061                         if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2062                                 mixed = 1;
2063                         else
2064                                 total_free_meta += found->disk_total -
2065                                         found->disk_used;
2066                 }
2067
2068                 total_used += found->disk_used;
2069         }
2070
2071         buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2072         buf->f_blocks >>= bits;
2073         buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2074
2075         /* Account global block reserve as used, it's in logical size already */
2076         spin_lock(&block_rsv->lock);
2077         /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2078         if (buf->f_bfree >= block_rsv->size >> bits)
2079                 buf->f_bfree -= block_rsv->size >> bits;
2080         else
2081                 buf->f_bfree = 0;
2082         spin_unlock(&block_rsv->lock);
2083
2084         buf->f_bavail = div_u64(total_free_data, factor);
2085         ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2086         if (ret)
2087                 return ret;
2088         buf->f_bavail += div_u64(total_free_data, factor);
2089         buf->f_bavail = buf->f_bavail >> bits;
2090
2091         /*
2092          * We calculate the remaining metadata space minus global reserve. If
2093          * this is (supposedly) smaller than zero, there's no space. But this
2094          * does not hold in practice, the exhausted state happens where's still
2095          * some positive delta. So we apply some guesswork and compare the
2096          * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2097          *
2098          * We probably cannot calculate the exact threshold value because this
2099          * depends on the internal reservations requested by various
2100          * operations, so some operations that consume a few metadata will
2101          * succeed even if the Avail is zero. But this is better than the other
2102          * way around.
2103          */
2104         thresh = SZ_4M;
2105
2106         /*
2107          * We only want to claim there's no available space if we can no longer
2108          * allocate chunks for our metadata profile and our global reserve will
2109          * not fit in the free metadata space.  If we aren't ->full then we
2110          * still can allocate chunks and thus are fine using the currently
2111          * calculated f_bavail.
2112          */
2113         if (!mixed && block_rsv->space_info->full &&
2114             total_free_meta - thresh < block_rsv->size)
2115                 buf->f_bavail = 0;
2116
2117         buf->f_type = BTRFS_SUPER_MAGIC;
2118         buf->f_bsize = dentry->d_sb->s_blocksize;
2119         buf->f_namelen = BTRFS_NAME_LEN;
2120
2121         /* We treat it as constant endianness (it doesn't matter _which_)
2122            because we want the fsid to come out the same whether mounted
2123            on a big-endian or little-endian host */
2124         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2125         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2126         /* Mask in the root object ID too, to disambiguate subvols */
2127         buf->f_fsid.val[0] ^=
2128                 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2129         buf->f_fsid.val[1] ^=
2130                 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2131
2132         return 0;
2133 }
2134
2135 static void btrfs_kill_super(struct super_block *sb)
2136 {
2137         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2138         kill_anon_super(sb);
2139         btrfs_free_fs_info(fs_info);
2140 }
2141
2142 static struct file_system_type btrfs_fs_type = {
2143         .owner          = THIS_MODULE,
2144         .name           = "btrfs",
2145         .mount          = btrfs_mount,
2146         .kill_sb        = btrfs_kill_super,
2147         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2148 };
2149
2150 static struct file_system_type btrfs_root_fs_type = {
2151         .owner          = THIS_MODULE,
2152         .name           = "btrfs",
2153         .mount          = btrfs_mount_root,
2154         .kill_sb        = btrfs_kill_super,
2155         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2156 };
2157
2158 MODULE_ALIAS_FS("btrfs");
2159
2160 static int btrfs_control_open(struct inode *inode, struct file *file)
2161 {
2162         /*
2163          * The control file's private_data is used to hold the
2164          * transaction when it is started and is used to keep
2165          * track of whether a transaction is already in progress.
2166          */
2167         file->private_data = NULL;
2168         return 0;
2169 }
2170
2171 /*
2172  * Used by /dev/btrfs-control for devices ioctls.
2173  */
2174 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2175                                 unsigned long arg)
2176 {
2177         struct btrfs_ioctl_vol_args *vol;
2178         struct btrfs_device *device = NULL;
2179         dev_t devt = 0;
2180         int ret = -ENOTTY;
2181
2182         if (!capable(CAP_SYS_ADMIN))
2183                 return -EPERM;
2184
2185         vol = memdup_user((void __user *)arg, sizeof(*vol));
2186         if (IS_ERR(vol))
2187                 return PTR_ERR(vol);
2188         vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2189
2190         switch (cmd) {
2191         case BTRFS_IOC_SCAN_DEV:
2192                 mutex_lock(&uuid_mutex);
2193                 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ);
2194                 ret = PTR_ERR_OR_ZERO(device);
2195                 mutex_unlock(&uuid_mutex);
2196                 break;
2197         case BTRFS_IOC_FORGET_DEV:
2198                 if (vol->name[0] != 0) {
2199                         ret = lookup_bdev(vol->name, &devt);
2200                         if (ret)
2201                                 break;
2202                 }
2203                 ret = btrfs_forget_devices(devt);
2204                 break;
2205         case BTRFS_IOC_DEVICES_READY:
2206                 mutex_lock(&uuid_mutex);
2207                 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ);
2208                 if (IS_ERR(device)) {
2209                         mutex_unlock(&uuid_mutex);
2210                         ret = PTR_ERR(device);
2211                         break;
2212                 }
2213                 ret = !(device->fs_devices->num_devices ==
2214                         device->fs_devices->total_devices);
2215                 mutex_unlock(&uuid_mutex);
2216                 break;
2217         case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2218                 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2219                 break;
2220         }
2221
2222         kfree(vol);
2223         return ret;
2224 }
2225
2226 static int btrfs_freeze(struct super_block *sb)
2227 {
2228         struct btrfs_trans_handle *trans;
2229         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2230         struct btrfs_root *root = fs_info->tree_root;
2231
2232         set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2233         /*
2234          * We don't need a barrier here, we'll wait for any transaction that
2235          * could be in progress on other threads (and do delayed iputs that
2236          * we want to avoid on a frozen filesystem), or do the commit
2237          * ourselves.
2238          */
2239         trans = btrfs_attach_transaction_barrier(root);
2240         if (IS_ERR(trans)) {
2241                 /* no transaction, don't bother */
2242                 if (PTR_ERR(trans) == -ENOENT)
2243                         return 0;
2244                 return PTR_ERR(trans);
2245         }
2246         return btrfs_commit_transaction(trans);
2247 }
2248
2249 static int check_dev_super(struct btrfs_device *dev)
2250 {
2251         struct btrfs_fs_info *fs_info = dev->fs_info;
2252         struct btrfs_super_block *sb;
2253         u16 csum_type;
2254         int ret = 0;
2255
2256         /* This should be called with fs still frozen. */
2257         ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2258
2259         /* Missing dev, no need to check. */
2260         if (!dev->bdev)
2261                 return 0;
2262
2263         /* Only need to check the primary super block. */
2264         sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2265         if (IS_ERR(sb))
2266                 return PTR_ERR(sb);
2267
2268         /* Verify the checksum. */
2269         csum_type = btrfs_super_csum_type(sb);
2270         if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2271                 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2272                           csum_type, btrfs_super_csum_type(fs_info->super_copy));
2273                 ret = -EUCLEAN;
2274                 goto out;
2275         }
2276
2277         if (btrfs_check_super_csum(fs_info, sb)) {
2278                 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2279                 ret = -EUCLEAN;
2280                 goto out;
2281         }
2282
2283         /* Btrfs_validate_super() includes fsid check against super->fsid. */
2284         ret = btrfs_validate_super(fs_info, sb, 0);
2285         if (ret < 0)
2286                 goto out;
2287
2288         if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
2289                 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2290                         btrfs_super_generation(sb),
2291                         fs_info->last_trans_committed);
2292                 ret = -EUCLEAN;
2293                 goto out;
2294         }
2295 out:
2296         btrfs_release_disk_super(sb);
2297         return ret;
2298 }
2299
2300 static int btrfs_unfreeze(struct super_block *sb)
2301 {
2302         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2303         struct btrfs_device *device;
2304         int ret = 0;
2305
2306         /*
2307          * Make sure the fs is not changed by accident (like hibernation then
2308          * modified by other OS).
2309          * If we found anything wrong, we mark the fs error immediately.
2310          *
2311          * And since the fs is frozen, no one can modify the fs yet, thus
2312          * we don't need to hold device_list_mutex.
2313          */
2314         list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2315                 ret = check_dev_super(device);
2316                 if (ret < 0) {
2317                         btrfs_handle_fs_error(fs_info, ret,
2318                                 "super block on devid %llu got modified unexpectedly",
2319                                 device->devid);
2320                         break;
2321                 }
2322         }
2323         clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2324
2325         /*
2326          * We still return 0, to allow VFS layer to unfreeze the fs even the
2327          * above checks failed. Since the fs is either fine or read-only, we're
2328          * safe to continue, without causing further damage.
2329          */
2330         return 0;
2331 }
2332
2333 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2334 {
2335         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2336
2337         /*
2338          * There should be always a valid pointer in latest_dev, it may be stale
2339          * for a short moment in case it's being deleted but still valid until
2340          * the end of RCU grace period.
2341          */
2342         rcu_read_lock();
2343         seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2344         rcu_read_unlock();
2345
2346         return 0;
2347 }
2348
2349 static const struct super_operations btrfs_super_ops = {
2350         .drop_inode     = btrfs_drop_inode,
2351         .evict_inode    = btrfs_evict_inode,
2352         .put_super      = btrfs_put_super,
2353         .sync_fs        = btrfs_sync_fs,
2354         .show_options   = btrfs_show_options,
2355         .show_devname   = btrfs_show_devname,
2356         .alloc_inode    = btrfs_alloc_inode,
2357         .destroy_inode  = btrfs_destroy_inode,
2358         .free_inode     = btrfs_free_inode,
2359         .statfs         = btrfs_statfs,
2360         .remount_fs     = btrfs_remount,
2361         .freeze_fs      = btrfs_freeze,
2362         .unfreeze_fs    = btrfs_unfreeze,
2363 };
2364
2365 static const struct file_operations btrfs_ctl_fops = {
2366         .open = btrfs_control_open,
2367         .unlocked_ioctl  = btrfs_control_ioctl,
2368         .compat_ioctl = compat_ptr_ioctl,
2369         .owner   = THIS_MODULE,
2370         .llseek = noop_llseek,
2371 };
2372
2373 static struct miscdevice btrfs_misc = {
2374         .minor          = BTRFS_MINOR,
2375         .name           = "btrfs-control",
2376         .fops           = &btrfs_ctl_fops
2377 };
2378
2379 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2380 MODULE_ALIAS("devname:btrfs-control");
2381
2382 static int __init btrfs_interface_init(void)
2383 {
2384         return misc_register(&btrfs_misc);
2385 }
2386
2387 static __cold void btrfs_interface_exit(void)
2388 {
2389         misc_deregister(&btrfs_misc);
2390 }
2391
2392 static int __init btrfs_print_mod_info(void)
2393 {
2394         static const char options[] = ""
2395 #ifdef CONFIG_BTRFS_DEBUG
2396                         ", debug=on"
2397 #endif
2398 #ifdef CONFIG_BTRFS_ASSERT
2399                         ", assert=on"
2400 #endif
2401 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2402                         ", integrity-checker=on"
2403 #endif
2404 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2405                         ", ref-verify=on"
2406 #endif
2407 #ifdef CONFIG_BLK_DEV_ZONED
2408                         ", zoned=yes"
2409 #else
2410                         ", zoned=no"
2411 #endif
2412 #ifdef CONFIG_FS_VERITY
2413                         ", fsverity=yes"
2414 #else
2415                         ", fsverity=no"
2416 #endif
2417                         ;
2418         pr_info("Btrfs loaded%s\n", options);
2419         return 0;
2420 }
2421
2422 static int register_btrfs(void)
2423 {
2424         return register_filesystem(&btrfs_fs_type);
2425 }
2426
2427 static void unregister_btrfs(void)
2428 {
2429         unregister_filesystem(&btrfs_fs_type);
2430 }
2431
2432 /* Helper structure for long init/exit functions. */
2433 struct init_sequence {
2434         int (*init_func)(void);
2435         /* Can be NULL if the init_func doesn't need cleanup. */
2436         void (*exit_func)(void);
2437 };
2438
2439 static const struct init_sequence mod_init_seq[] = {
2440         {
2441                 .init_func = btrfs_props_init,
2442                 .exit_func = NULL,
2443         }, {
2444                 .init_func = btrfs_init_sysfs,
2445                 .exit_func = btrfs_exit_sysfs,
2446         }, {
2447                 .init_func = btrfs_init_compress,
2448                 .exit_func = btrfs_exit_compress,
2449         }, {
2450                 .init_func = btrfs_init_cachep,
2451                 .exit_func = btrfs_destroy_cachep,
2452         }, {
2453                 .init_func = btrfs_transaction_init,
2454                 .exit_func = btrfs_transaction_exit,
2455         }, {
2456                 .init_func = btrfs_ctree_init,
2457                 .exit_func = btrfs_ctree_exit,
2458         }, {
2459                 .init_func = btrfs_free_space_init,
2460                 .exit_func = btrfs_free_space_exit,
2461         }, {
2462                 .init_func = extent_state_init_cachep,
2463                 .exit_func = extent_state_free_cachep,
2464         }, {
2465                 .init_func = extent_buffer_init_cachep,
2466                 .exit_func = extent_buffer_free_cachep,
2467         }, {
2468                 .init_func = btrfs_bioset_init,
2469                 .exit_func = btrfs_bioset_exit,
2470         }, {
2471                 .init_func = extent_map_init,
2472                 .exit_func = extent_map_exit,
2473         }, {
2474                 .init_func = ordered_data_init,
2475                 .exit_func = ordered_data_exit,
2476         }, {
2477                 .init_func = btrfs_delayed_inode_init,
2478                 .exit_func = btrfs_delayed_inode_exit,
2479         }, {
2480                 .init_func = btrfs_auto_defrag_init,
2481                 .exit_func = btrfs_auto_defrag_exit,
2482         }, {
2483                 .init_func = btrfs_delayed_ref_init,
2484                 .exit_func = btrfs_delayed_ref_exit,
2485         }, {
2486                 .init_func = btrfs_prelim_ref_init,
2487                 .exit_func = btrfs_prelim_ref_exit,
2488         }, {
2489                 .init_func = btrfs_interface_init,
2490                 .exit_func = btrfs_interface_exit,
2491         }, {
2492                 .init_func = btrfs_print_mod_info,
2493                 .exit_func = NULL,
2494         }, {
2495                 .init_func = btrfs_run_sanity_tests,
2496                 .exit_func = NULL,
2497         }, {
2498                 .init_func = register_btrfs,
2499                 .exit_func = unregister_btrfs,
2500         }
2501 };
2502
2503 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2504
2505 static __always_inline void btrfs_exit_btrfs_fs(void)
2506 {
2507         int i;
2508
2509         for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2510                 if (!mod_init_result[i])
2511                         continue;
2512                 if (mod_init_seq[i].exit_func)
2513                         mod_init_seq[i].exit_func();
2514                 mod_init_result[i] = false;
2515         }
2516 }
2517
2518 static void __exit exit_btrfs_fs(void)
2519 {
2520         btrfs_exit_btrfs_fs();
2521         btrfs_cleanup_fs_uuids();
2522 }
2523
2524 static int __init init_btrfs_fs(void)
2525 {
2526         int ret;
2527         int i;
2528
2529         for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2530                 ASSERT(!mod_init_result[i]);
2531                 ret = mod_init_seq[i].init_func();
2532                 if (ret < 0) {
2533                         btrfs_exit_btrfs_fs();
2534                         return ret;
2535                 }
2536                 mod_init_result[i] = true;
2537         }
2538         return 0;
2539 }
2540
2541 late_initcall(init_btrfs_fs);
2542 module_exit(exit_btrfs_fs)
2543
2544 MODULE_LICENSE("GPL");
2545 MODULE_SOFTDEP("pre: crc32c");
2546 MODULE_SOFTDEP("pre: xxhash64");
2547 MODULE_SOFTDEP("pre: sha256");
2548 MODULE_SOFTDEP("pre: blake2b-256");