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         }
 831         if (!ret)
 832                 ret = btrfs_check_mountopts_zoned(info);
 833         if (!ret && !remounting) {
 834                 if (btrfs_test_opt(info, SPACE_CACHE))
 835                         btrfs_info(info, "disk space caching is enabled");
 836                 if (btrfs_test_opt(info, FREE_SPACE_TREE))
 837                         btrfs_info(info, "using free space tree");
 838         }
 839         return ret;
 840 }
 841
 842 /*
 843  * Parse mount options that are required early in the mount process.
 844  *
 845  * All other options will be parsed on much later in the mount process and
 846  * only when we need to allocate a new super block.
 847  */
 848 static int btrfs_parse_device_options(const char *options, fmode_t flags,
 849                                       void *holder)
 850 {
 851         substring_t args[MAX_OPT_ARGS];
 852         char *device_name, *opts, *orig, *p;
 853         struct btrfs_device *device = NULL;
 854         int error = 0;
 855
 856         lockdep_assert_held(&uuid_mutex);
 857
 858         if (!options)
 859                 return 0;
 860
 861         /*
 862          * strsep changes the string, duplicate it because btrfs_parse_options
 863          * gets called later
 864          */
 865         opts = kstrdup(options, GFP_KERNEL);
 866         if (!opts)
 867                 return -ENOMEM;
 868         orig = opts;
 869
 870         while ((p = strsep(&opts, ",")) != NULL) {
 871                 int token;
 872
 873                 if (!*p)
 874                         continue;
 875
 876                 token = match_token(p, tokens, args);
 877                 if (token == Opt_device) {
 878                         device_name = match_strdup(&args[0]);
 879                         if (!device_name) {
 880                                 error = -ENOMEM;
 881                                 goto out;
 882                         }
 883                         device = btrfs_scan_one_device(device_name, flags,
 884                                         holder);
 885                         kfree(device_name);
 886                         if (IS_ERR(device)) {
 887                                 error = PTR_ERR(device);
 888                                 goto out;
 889                         }
 890                 }
 891         }
 892
 893 out:
 894         kfree(orig);
 895         return error;
 896 }
 897
 898 /*
 899  * Parse mount options that are related to subvolume id
 900  *
 901  * The value is later passed to mount_subvol()
 902  */
 903 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
 904                 u64 *subvol_objectid)
 905 {
 906         substring_t args[MAX_OPT_ARGS];
 907         char *opts, *orig, *p;
 908         int error = 0;
 909         u64 subvolid;
 910
 911         if (!options)
 912                 return 0;
 913
 914         /*
 915          * strsep changes the string, duplicate it because
 916          * btrfs_parse_device_options gets called later
 917          */
 918         opts = kstrdup(options, GFP_KERNEL);
 919         if (!opts)
 920                 return -ENOMEM;
 921         orig = opts;
 922
 923         while ((p = strsep(&opts, ",")) != NULL) {
 924                 int token;
 925                 if (!*p)
 926                         continue;
 927
 928                 token = match_token(p, tokens, args);
 929                 switch (token) {
 930                 case Opt_subvol:
 931                         kfree(*subvol_name);
 932                         *subvol_name = match_strdup(&args[0]);
 933                         if (!*subvol_name) {
 934                                 error = -ENOMEM;
 935                                 goto out;
 936                         }
 937                         break;
 938                 case Opt_subvolid:
 939                         error = match_u64(&args[0], &subvolid);
 940                         if (error)
 941                                 goto out;
 942
 943                         /* we want the original fs_tree */
 944                         if (subvolid == 0)
 945                                 subvolid = BTRFS_FS_TREE_OBJECTID;
 946
 947                         *subvol_objectid = subvolid;
 948                         break;
 949                 default:
 950                         break;
 951                 }
 952         }
 953
 954 out:
 955         kfree(orig);
 956         return error;
 957 }
 958
 959 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 960                                           u64 subvol_objectid)
 961 {
 962         struct btrfs_root *root = fs_info->tree_root;
 963         struct btrfs_root *fs_root = NULL;
 964         struct btrfs_root_ref *root_ref;
 965         struct btrfs_inode_ref *inode_ref;
 966         struct btrfs_key key;
 967         struct btrfs_path *path = NULL;
 968         char *name = NULL, *ptr;
 969         u64 dirid;
 970         int len;
 971         int ret;
 972
 973         path = btrfs_alloc_path();
 974         if (!path) {
 975                 ret = -ENOMEM;
 976                 goto err;
 977         }
 978
 979         name = kmalloc(PATH_MAX, GFP_KERNEL);
 980         if (!name) {
 981                 ret = -ENOMEM;
 982                 goto err;
 983         }
 984         ptr = name + PATH_MAX - 1;
 985         ptr[0] = '\0';
 986
 987         /*
 988          * Walk up the subvolume trees in the tree of tree roots by root
 989          * backrefs until we hit the top-level subvolume.
 990          */
 991         while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
 992                 key.objectid = subvol_objectid;
 993                 key.type = BTRFS_ROOT_BACKREF_KEY;
 994                 key.offset = (u64)-1;
 995
 996                 ret = btrfs_search_backwards(root, &key, path);
 997                 if (ret < 0) {
 998                         goto err;
 999                 } else if (ret > 0) {
1000                         ret = -ENOENT;
1001                         goto err;
1002                 }
1003
1004                 subvol_objectid = key.offset;
1005
1006                 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1007                                           struct btrfs_root_ref);
1008                 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1009                 ptr -= len + 1;
1010                 if (ptr < name) {
1011                         ret = -ENAMETOOLONG;
1012                         goto err;
1013                 }
1014                 read_extent_buffer(path->nodes[0], ptr + 1,
1015                                    (unsigned long)(root_ref + 1), len);
1016                 ptr[0] = '/';
1017                 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1018                 btrfs_release_path(path);
1019
1020                 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1021                 if (IS_ERR(fs_root)) {
1022                         ret = PTR_ERR(fs_root);
1023                         fs_root = NULL;
1024                         goto err;
1025                 }
1026
1027                 /*
1028                  * Walk up the filesystem tree by inode refs until we hit the
1029                  * root directory.
1030                  */
1031                 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1032                         key.objectid = dirid;
1033                         key.type = BTRFS_INODE_REF_KEY;
1034                         key.offset = (u64)-1;
1035
1036                         ret = btrfs_search_backwards(fs_root, &key, path);
1037                         if (ret < 0) {
1038                                 goto err;
1039                         } else if (ret > 0) {
1040                                 ret = -ENOENT;
1041                                 goto err;
1042                         }
1043
1044                         dirid = key.offset;
1045
1046                         inode_ref = btrfs_item_ptr(path->nodes[0],
1047                                                    path->slots[0],
1048                                                    struct btrfs_inode_ref);
1049                         len = btrfs_inode_ref_name_len(path->nodes[0],
1050                                                        inode_ref);
1051                         ptr -= len + 1;
1052                         if (ptr < name) {
1053                                 ret = -ENAMETOOLONG;
1054                                 goto err;
1055                         }
1056                         read_extent_buffer(path->nodes[0], ptr + 1,
1057                                            (unsigned long)(inode_ref + 1), len);
1058                         ptr[0] = '/';
1059                         btrfs_release_path(path);
1060                 }
1061                 btrfs_put_root(fs_root);
1062                 fs_root = NULL;
1063         }
1064
1065         btrfs_free_path(path);
1066         if (ptr == name + PATH_MAX - 1) {
1067                 name[0] = '/';
1068                 name[1] = '\0';
1069         } else {
1070                 memmove(name, ptr, name + PATH_MAX - ptr);
1071         }
1072         return name;
1073
1074 err:
1075         btrfs_put_root(fs_root);
1076         btrfs_free_path(path);
1077         kfree(name);
1078         return ERR_PTR(ret);
1079 }
1080
1081 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1082 {
1083         struct btrfs_root *root = fs_info->tree_root;
1084         struct btrfs_dir_item *di;
1085         struct btrfs_path *path;
1086         struct btrfs_key location;
1087         struct fscrypt_str name = FSTR_INIT("default", 7);
1088         u64 dir_id;
1089
1090         path = btrfs_alloc_path();
1091         if (!path)
1092                 return -ENOMEM;
1093
1094         /*
1095          * Find the "default" dir item which points to the root item that we
1096          * will mount by default if we haven't been given a specific subvolume
1097          * to mount.
1098          */
1099         dir_id = btrfs_super_root_dir(fs_info->super_copy);
1100         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
1101         if (IS_ERR(di)) {
1102                 btrfs_free_path(path);
1103                 return PTR_ERR(di);
1104         }
1105         if (!di) {
1106                 /*
1107                  * Ok the default dir item isn't there.  This is weird since
1108                  * it's always been there, but don't freak out, just try and
1109                  * mount the top-level subvolume.
1110                  */
1111                 btrfs_free_path(path);
1112                 *objectid = BTRFS_FS_TREE_OBJECTID;
1113                 return 0;
1114         }
1115
1116         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1117         btrfs_free_path(path);
1118         *objectid = location.objectid;
1119         return 0;
1120 }
1121
1122 static int btrfs_fill_super(struct super_block *sb,
1123                             struct btrfs_fs_devices *fs_devices,
1124                             void *data)
1125 {
1126         struct inode *inode;
1127         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1128         int err;
1129
1130         sb->s_maxbytes = MAX_LFS_FILESIZE;
1131         sb->s_magic = BTRFS_SUPER_MAGIC;
1132         sb->s_op = &btrfs_super_ops;
1133         sb->s_d_op = &btrfs_dentry_operations;
1134         sb->s_export_op = &btrfs_export_ops;
1135 #ifdef CONFIG_FS_VERITY
1136         sb->s_vop = &btrfs_verityops;
1137 #endif
1138         sb->s_xattr = btrfs_xattr_handlers;
1139         sb->s_time_gran = 1;
1140 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1141         sb->s_flags |= SB_POSIXACL;
1142 #endif
1143         sb->s_flags |= SB_I_VERSION;
1144         sb->s_iflags |= SB_I_CGROUPWB;
1145
1146         err = super_setup_bdi(sb);
1147         if (err) {
1148                 btrfs_err(fs_info, "super_setup_bdi failed");
1149                 return err;
1150         }
1151
1152         err = open_ctree(sb, fs_devices, (char *)data);
1153         if (err) {
1154                 btrfs_err(fs_info, "open_ctree failed");
1155                 return err;
1156         }
1157
1158         inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1159         if (IS_ERR(inode)) {
1160                 err = PTR_ERR(inode);
1161                 btrfs_handle_fs_error(fs_info, err, NULL);
1162                 goto fail_close;
1163         }
1164
1165         sb->s_root = d_make_root(inode);
1166         if (!sb->s_root) {
1167                 err = -ENOMEM;
1168                 goto fail_close;
1169         }
1170
1171         sb->s_flags |= SB_ACTIVE;
1172         return 0;
1173
1174 fail_close:
1175         close_ctree(fs_info);
1176         return err;
1177 }
1178
1179 int btrfs_sync_fs(struct super_block *sb, int wait)
1180 {
1181         struct btrfs_trans_handle *trans;
1182         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1183         struct btrfs_root *root = fs_info->tree_root;
1184
1185         trace_btrfs_sync_fs(fs_info, wait);
1186
1187         if (!wait) {
1188                 filemap_flush(fs_info->btree_inode->i_mapping);
1189                 return 0;
1190         }
1191
1192         btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1193
1194         trans = btrfs_attach_transaction_barrier(root);
1195         if (IS_ERR(trans)) {
1196                 /* no transaction, don't bother */
1197                 if (PTR_ERR(trans) == -ENOENT) {
1198                         /*
1199                          * Exit unless we have some pending changes
1200                          * that need to go through commit
1201                          */
1202                         if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1203                                       &fs_info->flags))
1204                                 return 0;
1205                         /*
1206                          * A non-blocking test if the fs is frozen. We must not
1207                          * start a new transaction here otherwise a deadlock
1208                          * happens. The pending operations are delayed to the
1209                          * next commit after thawing.
1210                          */
1211                         if (sb_start_write_trylock(sb))
1212                                 sb_end_write(sb);
1213                         else
1214                                 return 0;
1215                         trans = btrfs_start_transaction(root, 0);
1216                 }
1217                 if (IS_ERR(trans))
1218                         return PTR_ERR(trans);
1219         }
1220         return btrfs_commit_transaction(trans);
1221 }
1222
1223 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1224 {
1225         seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1226         *printed = true;
1227 }
1228
1229 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1230 {
1231         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1232         const char *compress_type;
1233         const char *subvol_name;
1234         bool printed = false;
1235
1236         if (btrfs_test_opt(info, DEGRADED))
1237                 seq_puts(seq, ",degraded");
1238         if (btrfs_test_opt(info, NODATASUM))
1239                 seq_puts(seq, ",nodatasum");
1240         if (btrfs_test_opt(info, NODATACOW))
1241                 seq_puts(seq, ",nodatacow");
1242         if (btrfs_test_opt(info, NOBARRIER))
1243                 seq_puts(seq, ",nobarrier");
1244         if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1245                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1246         if (info->thread_pool_size !=  min_t(unsigned long,
1247                                              num_online_cpus() + 2, 8))
1248                 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1249         if (btrfs_test_opt(info, COMPRESS)) {
1250                 compress_type = btrfs_compress_type2str(info->compress_type);
1251                 if (btrfs_test_opt(info, FORCE_COMPRESS))
1252                         seq_printf(seq, ",compress-force=%s", compress_type);
1253                 else
1254                         seq_printf(seq, ",compress=%s", compress_type);
1255                 if (info->compress_level)
1256                         seq_printf(seq, ":%d", info->compress_level);
1257         }
1258         if (btrfs_test_opt(info, NOSSD))
1259                 seq_puts(seq, ",nossd");
1260         if (btrfs_test_opt(info, SSD_SPREAD))
1261                 seq_puts(seq, ",ssd_spread");
1262         else if (btrfs_test_opt(info, SSD))
1263                 seq_puts(seq, ",ssd");
1264         if (btrfs_test_opt(info, NOTREELOG))
1265                 seq_puts(seq, ",notreelog");
1266         if (btrfs_test_opt(info, NOLOGREPLAY))
1267                 print_rescue_option(seq, "nologreplay", &printed);
1268         if (btrfs_test_opt(info, USEBACKUPROOT))
1269                 print_rescue_option(seq, "usebackuproot", &printed);
1270         if (btrfs_test_opt(info, IGNOREBADROOTS))
1271                 print_rescue_option(seq, "ignorebadroots", &printed);
1272         if (btrfs_test_opt(info, IGNOREDATACSUMS))
1273                 print_rescue_option(seq, "ignoredatacsums", &printed);
1274         if (btrfs_test_opt(info, FLUSHONCOMMIT))
1275                 seq_puts(seq, ",flushoncommit");
1276         if (btrfs_test_opt(info, DISCARD_SYNC))
1277                 seq_puts(seq, ",discard");
1278         if (btrfs_test_opt(info, DISCARD_ASYNC))
1279                 seq_puts(seq, ",discard=async");
1280         if (!(info->sb->s_flags & SB_POSIXACL))
1281                 seq_puts(seq, ",noacl");
1282         if (btrfs_free_space_cache_v1_active(info))
1283                 seq_puts(seq, ",space_cache");
1284         else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1285                 seq_puts(seq, ",space_cache=v2");
1286         else
1287                 seq_puts(seq, ",nospace_cache");
1288         if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1289                 seq_puts(seq, ",rescan_uuid_tree");
1290         if (btrfs_test_opt(info, CLEAR_CACHE))
1291                 seq_puts(seq, ",clear_cache");
1292         if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1293                 seq_puts(seq, ",user_subvol_rm_allowed");
1294         if (btrfs_test_opt(info, ENOSPC_DEBUG))
1295                 seq_puts(seq, ",enospc_debug");
1296         if (btrfs_test_opt(info, AUTO_DEFRAG))
1297                 seq_puts(seq, ",autodefrag");
1298         if (btrfs_test_opt(info, SKIP_BALANCE))
1299                 seq_puts(seq, ",skip_balance");
1300 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1301         if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1302                 seq_puts(seq, ",check_int_data");
1303         else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1304                 seq_puts(seq, ",check_int");
1305         if (info->check_integrity_print_mask)
1306                 seq_printf(seq, ",check_int_print_mask=%d",
1307                                 info->check_integrity_print_mask);
1308 #endif
1309         if (info->metadata_ratio)
1310                 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1311         if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1312                 seq_puts(seq, ",fatal_errors=panic");
1313         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1314                 seq_printf(seq, ",commit=%u", info->commit_interval);
1315 #ifdef CONFIG_BTRFS_DEBUG
1316         if (btrfs_test_opt(info, FRAGMENT_DATA))
1317                 seq_puts(seq, ",fragment=data");
1318         if (btrfs_test_opt(info, FRAGMENT_METADATA))
1319                 seq_puts(seq, ",fragment=metadata");
1320 #endif
1321         if (btrfs_test_opt(info, REF_VERIFY))
1322                 seq_puts(seq, ",ref_verify");
1323         seq_printf(seq, ",subvolid=%llu",
1324                   BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1325         subvol_name = btrfs_get_subvol_name_from_objectid(info,
1326                         BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1327         if (!IS_ERR(subvol_name)) {
1328                 seq_puts(seq, ",subvol=");
1329                 seq_escape(seq, subvol_name, " \t\n\\");
1330                 kfree(subvol_name);
1331         }
1332         return 0;
1333 }
1334
1335 static int btrfs_test_super(struct super_block *s, void *data)
1336 {
1337         struct btrfs_fs_info *p = data;
1338         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1339
1340         return fs_info->fs_devices == p->fs_devices;
1341 }
1342
1343 static int btrfs_set_super(struct super_block *s, void *data)
1344 {
1345         int err = set_anon_super(s, data);
1346         if (!err)
1347                 s->s_fs_info = data;
1348         return err;
1349 }
1350
1351 /*
1352  * subvolumes are identified by ino 256
1353  */
1354 static inline int is_subvolume_inode(struct inode *inode)
1355 {
1356         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1357                 return 1;
1358         return 0;
1359 }
1360
1361 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1362                                    struct vfsmount *mnt)
1363 {
1364         struct dentry *root;
1365         int ret;
1366
1367         if (!subvol_name) {
1368                 if (!subvol_objectid) {
1369                         ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1370                                                           &subvol_objectid);
1371                         if (ret) {
1372                                 root = ERR_PTR(ret);
1373                                 goto out;
1374                         }
1375                 }
1376                 subvol_name = btrfs_get_subvol_name_from_objectid(
1377                                         btrfs_sb(mnt->mnt_sb), subvol_objectid);
1378                 if (IS_ERR(subvol_name)) {
1379                         root = ERR_CAST(subvol_name);
1380                         subvol_name = NULL;
1381                         goto out;
1382                 }
1383
1384         }
1385
1386         root = mount_subtree(mnt, subvol_name);
1387         /* mount_subtree() drops our reference on the vfsmount. */
1388         mnt = NULL;
1389
1390         if (!IS_ERR(root)) {
1391                 struct super_block *s = root->d_sb;
1392                 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1393                 struct inode *root_inode = d_inode(root);
1394                 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1395
1396                 ret = 0;
1397                 if (!is_subvolume_inode(root_inode)) {
1398                         btrfs_err(fs_info, "'%s' is not a valid subvolume",
1399                                subvol_name);
1400                         ret = -EINVAL;
1401                 }
1402                 if (subvol_objectid && root_objectid != subvol_objectid) {
1403                         /*
1404                          * This will also catch a race condition where a
1405                          * subvolume which was passed by ID is renamed and
1406                          * another subvolume is renamed over the old location.
1407                          */
1408                         btrfs_err(fs_info,
1409                                   "subvol '%s' does not match subvolid %llu",
1410                                   subvol_name, subvol_objectid);
1411                         ret = -EINVAL;
1412                 }
1413                 if (ret) {
1414                         dput(root);
1415                         root = ERR_PTR(ret);
1416                         deactivate_locked_super(s);
1417                 }
1418         }
1419
1420 out:
1421         mntput(mnt);
1422         kfree(subvol_name);
1423         return root;
1424 }
1425
1426 /*
1427  * Find a superblock for the given device / mount point.
1428  *
1429  * Note: This is based on mount_bdev from fs/super.c with a few additions
1430  *       for multiple device setup.  Make sure to keep it in sync.
1431  */
1432 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1433                 int flags, const char *device_name, void *data)
1434 {
1435         struct block_device *bdev = NULL;
1436         struct super_block *s;
1437         struct btrfs_device *device = NULL;
1438         struct btrfs_fs_devices *fs_devices = NULL;
1439         struct btrfs_fs_info *fs_info = NULL;
1440         void *new_sec_opts = NULL;
1441         fmode_t mode = FMODE_READ;
1442         int error = 0;
1443
1444         if (!(flags & SB_RDONLY))
1445                 mode |= FMODE_WRITE;
1446
1447         if (data) {
1448                 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1449                 if (error)
1450                         return ERR_PTR(error);
1451         }
1452
1453         /*
1454          * Setup a dummy root and fs_info for test/set super.  This is because
1455          * we don't actually fill this stuff out until open_ctree, but we need
1456          * then open_ctree will properly initialize the file system specific
1457          * settings later.  btrfs_init_fs_info initializes the static elements
1458          * of the fs_info (locks and such) to make cleanup easier if we find a
1459          * superblock with our given fs_devices later on at sget() time.
1460          */
1461         fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1462         if (!fs_info) {
1463                 error = -ENOMEM;
1464                 goto error_sec_opts;
1465         }
1466         btrfs_init_fs_info(fs_info);
1467
1468         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1469         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1470         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1471                 error = -ENOMEM;
1472                 goto error_fs_info;
1473         }
1474
1475         mutex_lock(&uuid_mutex);
1476         error = btrfs_parse_device_options(data, mode, fs_type);
1477         if (error) {
1478                 mutex_unlock(&uuid_mutex);
1479                 goto error_fs_info;
1480         }
1481
1482         device = btrfs_scan_one_device(device_name, mode, fs_type);
1483         if (IS_ERR(device)) {
1484                 mutex_unlock(&uuid_mutex);
1485                 error = PTR_ERR(device);
1486                 goto error_fs_info;
1487         }
1488
1489         fs_devices = device->fs_devices;
1490         fs_info->fs_devices = fs_devices;
1491
1492         error = btrfs_open_devices(fs_devices, mode, fs_type);
1493         mutex_unlock(&uuid_mutex);
1494         if (error)
1495                 goto error_fs_info;
1496
1497         if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1498                 error = -EACCES;
1499                 goto error_close_devices;
1500         }
1501
1502         bdev = fs_devices->latest_dev->bdev;
1503         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1504                  fs_info);
1505         if (IS_ERR(s)) {
1506                 error = PTR_ERR(s);
1507                 goto error_close_devices;
1508         }
1509
1510         if (s->s_root) {
1511                 btrfs_close_devices(fs_devices);
1512                 btrfs_free_fs_info(fs_info);
1513                 if ((flags ^ s->s_flags) & SB_RDONLY)
1514                         error = -EBUSY;
1515         } else {
1516                 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1517                 shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
1518                                         s->s_id);
1519                 btrfs_sb(s)->bdev_holder = fs_type;
1520                 error = btrfs_fill_super(s, fs_devices, data);
1521         }
1522         if (!error)
1523                 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1524         security_free_mnt_opts(&new_sec_opts);
1525         if (error) {
1526                 deactivate_locked_super(s);
1527                 return ERR_PTR(error);
1528         }
1529
1530         return dget(s->s_root);
1531
1532 error_close_devices:
1533         btrfs_close_devices(fs_devices);
1534 error_fs_info:
1535         btrfs_free_fs_info(fs_info);
1536 error_sec_opts:
1537         security_free_mnt_opts(&new_sec_opts);
1538         return ERR_PTR(error);
1539 }
1540
1541 /*
1542  * Mount function which is called by VFS layer.
1543  *
1544  * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1545  * which needs vfsmount* of device's root (/).  This means device's root has to
1546  * be mounted internally in any case.
1547  *
1548  * Operation flow:
1549  *   1. Parse subvol id related options for later use in mount_subvol().
1550  *
1551  *   2. Mount device's root (/) by calling vfs_kern_mount().
1552  *
1553  *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1554  *      first place. In order to avoid calling btrfs_mount() again, we use
1555  *      different file_system_type which is not registered to VFS by
1556  *      register_filesystem() (btrfs_root_fs_type). As a result,
1557  *      btrfs_mount_root() is called. The return value will be used by
1558  *      mount_subtree() in mount_subvol().
1559  *
1560  *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1561  *      "btrfs subvolume set-default", mount_subvol() is called always.
1562  */
1563 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1564                 const char *device_name, void *data)
1565 {
1566         struct vfsmount *mnt_root;
1567         struct dentry *root;
1568         char *subvol_name = NULL;
1569         u64 subvol_objectid = 0;
1570         int error = 0;
1571
1572         error = btrfs_parse_subvol_options(data, &subvol_name,
1573                                         &subvol_objectid);
1574         if (error) {
1575                 kfree(subvol_name);
1576                 return ERR_PTR(error);
1577         }
1578
1579         /* mount device's root (/) */
1580         mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1581         if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1582                 if (flags & SB_RDONLY) {
1583                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1584                                 flags & ~SB_RDONLY, device_name, data);
1585                 } else {
1586                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1587                                 flags | SB_RDONLY, device_name, data);
1588                         if (IS_ERR(mnt_root)) {
1589                                 root = ERR_CAST(mnt_root);
1590                                 kfree(subvol_name);
1591                                 goto out;
1592                         }
1593
1594                         down_write(&mnt_root->mnt_sb->s_umount);
1595                         error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1596                         up_write(&mnt_root->mnt_sb->s_umount);
1597                         if (error < 0) {
1598                                 root = ERR_PTR(error);
1599                                 mntput(mnt_root);
1600                                 kfree(subvol_name);
1601                                 goto out;
1602                         }
1603                 }
1604         }
1605         if (IS_ERR(mnt_root)) {
1606                 root = ERR_CAST(mnt_root);
1607                 kfree(subvol_name);
1608                 goto out;
1609         }
1610
1611         /* mount_subvol() will free subvol_name and mnt_root */
1612         root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1613
1614 out:
1615         return root;
1616 }
1617
1618 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1619                                      u32 new_pool_size, u32 old_pool_size)
1620 {
1621         if (new_pool_size == old_pool_size)
1622                 return;
1623
1624         fs_info->thread_pool_size = new_pool_size;
1625
1626         btrfs_info(fs_info, "resize thread pool %d -> %d",
1627                old_pool_size, new_pool_size);
1628
1629         btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1630         btrfs_workqueue_set_max(fs_info->hipri_workers, new_pool_size);
1631         btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1632         btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1633         workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1634         workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1635         btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1636         btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1637         btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1638 }
1639
1640 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1641                                        unsigned long old_opts, int flags)
1642 {
1643         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1644             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1645              (flags & SB_RDONLY))) {
1646                 /* wait for any defraggers to finish */
1647                 wait_event(fs_info->transaction_wait,
1648                            (atomic_read(&fs_info->defrag_running) == 0));
1649                 if (flags & SB_RDONLY)
1650                         sync_filesystem(fs_info->sb);
1651         }
1652 }
1653
1654 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1655                                          unsigned long old_opts)
1656 {
1657         const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1658
1659         /*
1660          * We need to cleanup all defragable inodes if the autodefragment is
1661          * close or the filesystem is read only.
1662          */
1663         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1664             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1665                 btrfs_cleanup_defrag_inodes(fs_info);
1666         }
1667
1668         /* If we toggled discard async */
1669         if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1670             btrfs_test_opt(fs_info, DISCARD_ASYNC))
1671                 btrfs_discard_resume(fs_info);
1672         else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1673                  !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1674                 btrfs_discard_cleanup(fs_info);
1675
1676         /* If we toggled space cache */
1677         if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1678                 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1679 }
1680
1681 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1682 {
1683         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1684         unsigned old_flags = sb->s_flags;
1685         unsigned long old_opts = fs_info->mount_opt;
1686         unsigned long old_compress_type = fs_info->compress_type;
1687         u64 old_max_inline = fs_info->max_inline;
1688         u32 old_thread_pool_size = fs_info->thread_pool_size;
1689         u32 old_metadata_ratio = fs_info->metadata_ratio;
1690         int ret;
1691
1692         sync_filesystem(sb);
1693         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1694
1695         if (data) {
1696                 void *new_sec_opts = NULL;
1697
1698                 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1699                 if (!ret)
1700                         ret = security_sb_remount(sb, new_sec_opts);
1701                 security_free_mnt_opts(&new_sec_opts);
1702                 if (ret)
1703                         goto restore;
1704         }
1705
1706         ret = btrfs_parse_options(fs_info, data, *flags);
1707         if (ret)
1708                 goto restore;
1709
1710         ret = btrfs_check_features(fs_info, !(*flags & SB_RDONLY));
1711         if (ret < 0)
1712                 goto restore;
1713
1714         btrfs_remount_begin(fs_info, old_opts, *flags);
1715         btrfs_resize_thread_pool(fs_info,
1716                 fs_info->thread_pool_size, old_thread_pool_size);
1717
1718         if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1719             (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1720             (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1721                 btrfs_warn(fs_info,
1722                 "remount supports changing free space tree only from ro to rw");
1723                 /* Make sure free space cache options match the state on disk */
1724                 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1725                         btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1726                         btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1727                 }
1728                 if (btrfs_free_space_cache_v1_active(fs_info)) {
1729                         btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1730                         btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1731                 }
1732         }
1733
1734         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1735                 goto out;
1736
1737         if (*flags & SB_RDONLY) {
1738                 /*
1739                  * this also happens on 'umount -rf' or on shutdown, when
1740                  * the filesystem is busy.
1741                  */
1742                 cancel_work_sync(&fs_info->async_reclaim_work);
1743                 cancel_work_sync(&fs_info->async_data_reclaim_work);
1744
1745                 btrfs_discard_cleanup(fs_info);
1746
1747                 /* wait for the uuid_scan task to finish */
1748                 down(&fs_info->uuid_tree_rescan_sem);
1749                 /* avoid complains from lockdep et al. */
1750                 up(&fs_info->uuid_tree_rescan_sem);
1751
1752                 btrfs_set_sb_rdonly(sb);
1753
1754                 /*
1755                  * Setting SB_RDONLY will put the cleaner thread to
1756                  * sleep at the next loop if it's already active.
1757                  * If it's already asleep, we'll leave unused block
1758                  * groups on disk until we're mounted read-write again
1759                  * unless we clean them up here.
1760                  */
1761                 btrfs_delete_unused_bgs(fs_info);
1762
1763                 /*
1764                  * The cleaner task could be already running before we set the
1765                  * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1766                  * We must make sure that after we finish the remount, i.e. after
1767                  * we call btrfs_commit_super(), the cleaner can no longer start
1768                  * a transaction - either because it was dropping a dead root,
1769                  * running delayed iputs or deleting an unused block group (the
1770                  * cleaner picked a block group from the list of unused block
1771                  * groups before we were able to in the previous call to
1772                  * btrfs_delete_unused_bgs()).
1773                  */
1774                 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1775                             TASK_UNINTERRUPTIBLE);
1776
1777                 /*
1778                  * We've set the superblock to RO mode, so we might have made
1779                  * the cleaner task sleep without running all pending delayed
1780                  * iputs. Go through all the delayed iputs here, so that if an
1781                  * unmount happens without remounting RW we don't end up at
1782                  * finishing close_ctree() with a non-empty list of delayed
1783                  * iputs.
1784                  */
1785                 btrfs_run_delayed_iputs(fs_info);
1786
1787                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1788                 btrfs_scrub_cancel(fs_info);
1789                 btrfs_pause_balance(fs_info);
1790
1791                 /*
1792                  * Pause the qgroup rescan worker if it is running. We don't want
1793                  * it to be still running after we are in RO mode, as after that,
1794                  * by the time we unmount, it might have left a transaction open,
1795                  * so we would leak the transaction and/or crash.
1796                  */
1797                 btrfs_qgroup_wait_for_completion(fs_info, false);
1798
1799                 ret = btrfs_commit_super(fs_info);
1800                 if (ret)
1801                         goto restore;
1802         } else {
1803                 if (BTRFS_FS_ERROR(fs_info)) {
1804                         btrfs_err(fs_info,
1805                                 "Remounting read-write after error is not allowed");
1806                         ret = -EINVAL;
1807                         goto restore;
1808                 }
1809                 if (fs_info->fs_devices->rw_devices == 0) {
1810                         ret = -EACCES;
1811                         goto restore;
1812                 }
1813
1814                 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1815                         btrfs_warn(fs_info,
1816                 "too many missing devices, writable remount is not allowed");
1817                         ret = -EACCES;
1818                         goto restore;
1819                 }
1820
1821                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1822                         btrfs_warn(fs_info,
1823                 "mount required to replay tree-log, cannot remount read-write");
1824                         ret = -EINVAL;
1825                         goto restore;
1826                 }
1827
1828                 /*
1829                  * NOTE: when remounting with a change that does writes, don't
1830                  * put it anywhere above this point, as we are not sure to be
1831                  * safe to write until we pass the above checks.
1832                  */
1833                 ret = btrfs_start_pre_rw_mount(fs_info);
1834                 if (ret)
1835                         goto restore;
1836
1837                 btrfs_clear_sb_rdonly(sb);
1838
1839                 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1840         }
1841 out:
1842         /*
1843          * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1844          * since the absence of the flag means it can be toggled off by remount.
1845          */
1846         *flags |= SB_I_VERSION;
1847
1848         wake_up_process(fs_info->transaction_kthread);
1849         btrfs_remount_cleanup(fs_info, old_opts);
1850         btrfs_clear_oneshot_options(fs_info);
1851         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1852
1853         return 0;
1854
1855 restore:
1856         /* We've hit an error - don't reset SB_RDONLY */
1857         if (sb_rdonly(sb))
1858                 old_flags |= SB_RDONLY;
1859         if (!(old_flags & SB_RDONLY))
1860                 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
1861         sb->s_flags = old_flags;
1862         fs_info->mount_opt = old_opts;
1863         fs_info->compress_type = old_compress_type;
1864         fs_info->max_inline = old_max_inline;
1865         btrfs_resize_thread_pool(fs_info,
1866                 old_thread_pool_size, fs_info->thread_pool_size);
1867         fs_info->metadata_ratio = old_metadata_ratio;
1868         btrfs_remount_cleanup(fs_info, old_opts);
1869         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1870
1871         return ret;
1872 }
1873
1874 /* Used to sort the devices by max_avail(descending sort) */
1875 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1876 {
1877         const struct btrfs_device_info *dev_info1 = a;
1878         const struct btrfs_device_info *dev_info2 = b;
1879
1880         if (dev_info1->max_avail > dev_info2->max_avail)
1881                 return -1;
1882         else if (dev_info1->max_avail < dev_info2->max_avail)
1883                 return 1;
1884         return 0;
1885 }
1886
1887 /*
1888  * sort the devices by max_avail, in which max free extent size of each device
1889  * is stored.(Descending Sort)
1890  */
1891 static inline void btrfs_descending_sort_devices(
1892                                         struct btrfs_device_info *devices,
1893                                         size_t nr_devices)
1894 {
1895         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1896              btrfs_cmp_device_free_bytes, NULL);
1897 }
1898
1899 /*
1900  * The helper to calc the free space on the devices that can be used to store
1901  * file data.
1902  */
1903 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1904                                               u64 *free_bytes)
1905 {
1906         struct btrfs_device_info *devices_info;
1907         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1908         struct btrfs_device *device;
1909         u64 type;
1910         u64 avail_space;
1911         u64 min_stripe_size;
1912         int num_stripes = 1;
1913         int i = 0, nr_devices;
1914         const struct btrfs_raid_attr *rattr;
1915
1916         /*
1917          * We aren't under the device list lock, so this is racy-ish, but good
1918          * enough for our purposes.
1919          */
1920         nr_devices = fs_info->fs_devices->open_devices;
1921         if (!nr_devices) {
1922                 smp_mb();
1923                 nr_devices = fs_info->fs_devices->open_devices;
1924                 ASSERT(nr_devices);
1925                 if (!nr_devices) {
1926                         *free_bytes = 0;
1927                         return 0;
1928                 }
1929         }
1930
1931         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1932                                GFP_KERNEL);
1933         if (!devices_info)
1934                 return -ENOMEM;
1935
1936         /* calc min stripe number for data space allocation */
1937         type = btrfs_data_alloc_profile(fs_info);
1938         rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1939
1940         if (type & BTRFS_BLOCK_GROUP_RAID0)
1941                 num_stripes = nr_devices;
1942         else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1943                 num_stripes = rattr->ncopies;
1944         else if (type & BTRFS_BLOCK_GROUP_RAID10)
1945                 num_stripes = 4;
1946
1947         /* Adjust for more than 1 stripe per device */
1948         min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1949
1950         rcu_read_lock();
1951         list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1952                 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1953                                                 &device->dev_state) ||
1954                     !device->bdev ||
1955                     test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1956                         continue;
1957
1958                 if (i >= nr_devices)
1959                         break;
1960
1961                 avail_space = device->total_bytes - device->bytes_used;
1962
1963                 /* align with stripe_len */
1964                 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1965
1966                 /*
1967                  * Ensure we have at least min_stripe_size on top of the
1968                  * reserved space on the device.
1969                  */
1970                 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1971                         continue;
1972
1973                 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1974
1975                 devices_info[i].dev = device;
1976                 devices_info[i].max_avail = avail_space;
1977
1978                 i++;
1979         }
1980         rcu_read_unlock();
1981
1982         nr_devices = i;
1983
1984         btrfs_descending_sort_devices(devices_info, nr_devices);
1985
1986         i = nr_devices - 1;
1987         avail_space = 0;
1988         while (nr_devices >= rattr->devs_min) {
1989                 num_stripes = min(num_stripes, nr_devices);
1990
1991                 if (devices_info[i].max_avail >= min_stripe_size) {
1992                         int j;
1993                         u64 alloc_size;
1994
1995                         avail_space += devices_info[i].max_avail * num_stripes;
1996                         alloc_size = devices_info[i].max_avail;
1997                         for (j = i + 1 - num_stripes; j <= i; j++)
1998                                 devices_info[j].max_avail -= alloc_size;
1999                 }
2000                 i--;
2001                 nr_devices--;
2002         }
2003
2004         kfree(devices_info);
2005         *free_bytes = avail_space;
2006         return 0;
2007 }
2008
2009 /*
2010  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2011  *
2012  * If there's a redundant raid level at DATA block groups, use the respective
2013  * multiplier to scale the sizes.
2014  *
2015  * Unused device space usage is based on simulating the chunk allocator
2016  * algorithm that respects the device sizes and order of allocations.  This is
2017  * a close approximation of the actual use but there are other factors that may
2018  * change the result (like a new metadata chunk).
2019  *
2020  * If metadata is exhausted, f_bavail will be 0.
2021  */
2022 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2023 {
2024         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2025         struct btrfs_super_block *disk_super = fs_info->super_copy;
2026         struct btrfs_space_info *found;
2027         u64 total_used = 0;
2028         u64 total_free_data = 0;
2029         u64 total_free_meta = 0;
2030         u32 bits = fs_info->sectorsize_bits;
2031         __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2032         unsigned factor = 1;
2033         struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2034         int ret;
2035         u64 thresh = 0;
2036         int mixed = 0;
2037
2038         list_for_each_entry(found, &fs_info->space_info, list) {
2039                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2040                         int i;
2041
2042                         total_free_data += found->disk_total - found->disk_used;
2043                         total_free_data -=
2044                                 btrfs_account_ro_block_groups_free_space(found);
2045
2046                         for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2047                                 if (!list_empty(&found->block_groups[i]))
2048                                         factor = btrfs_bg_type_to_factor(
2049                                                 btrfs_raid_array[i].bg_flag);
2050                         }
2051                 }
2052
2053                 /*
2054                  * Metadata in mixed block group profiles are accounted in data
2055                  */
2056                 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2057                         if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2058                                 mixed = 1;
2059                         else
2060                                 total_free_meta += found->disk_total -
2061                                         found->disk_used;
2062                 }
2063
2064                 total_used += found->disk_used;
2065         }
2066
2067         buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2068         buf->f_blocks >>= bits;
2069         buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2070
2071         /* Account global block reserve as used, it's in logical size already */
2072         spin_lock(&block_rsv->lock);
2073         /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2074         if (buf->f_bfree >= block_rsv->size >> bits)
2075                 buf->f_bfree -= block_rsv->size >> bits;
2076         else
2077                 buf->f_bfree = 0;
2078         spin_unlock(&block_rsv->lock);
2079
2080         buf->f_bavail = div_u64(total_free_data, factor);
2081         ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2082         if (ret)
2083                 return ret;
2084         buf->f_bavail += div_u64(total_free_data, factor);
2085         buf->f_bavail = buf->f_bavail >> bits;
2086
2087         /*
2088          * We calculate the remaining metadata space minus global reserve. If
2089          * this is (supposedly) smaller than zero, there's no space. But this
2090          * does not hold in practice, the exhausted state happens where's still
2091          * some positive delta. So we apply some guesswork and compare the
2092          * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2093          *
2094          * We probably cannot calculate the exact threshold value because this
2095          * depends on the internal reservations requested by various
2096          * operations, so some operations that consume a few metadata will
2097          * succeed even if the Avail is zero. But this is better than the other
2098          * way around.
2099          */
2100         thresh = SZ_4M;
2101
2102         /*
2103          * We only want to claim there's no available space if we can no longer
2104          * allocate chunks for our metadata profile and our global reserve will
2105          * not fit in the free metadata space.  If we aren't ->full then we
2106          * still can allocate chunks and thus are fine using the currently
2107          * calculated f_bavail.
2108          */
2109         if (!mixed && block_rsv->space_info->full &&
2110             total_free_meta - thresh < block_rsv->size)
2111                 buf->f_bavail = 0;
2112
2113         buf->f_type = BTRFS_SUPER_MAGIC;
2114         buf->f_bsize = dentry->d_sb->s_blocksize;
2115         buf->f_namelen = BTRFS_NAME_LEN;
2116
2117         /* We treat it as constant endianness (it doesn't matter _which_)
2118            because we want the fsid to come out the same whether mounted
2119            on a big-endian or little-endian host */
2120         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2121         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2122         /* Mask in the root object ID too, to disambiguate subvols */
2123         buf->f_fsid.val[0] ^=
2124                 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2125         buf->f_fsid.val[1] ^=
2126                 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2127
2128         return 0;
2129 }
2130
2131 static void btrfs_kill_super(struct super_block *sb)
2132 {
2133         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2134         kill_anon_super(sb);
2135         btrfs_free_fs_info(fs_info);
2136 }
2137
2138 static struct file_system_type btrfs_fs_type = {
2139         .owner          = THIS_MODULE,
2140         .name           = "btrfs",
2141         .mount          = btrfs_mount,
2142         .kill_sb        = btrfs_kill_super,
2143         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2144 };
2145
2146 static struct file_system_type btrfs_root_fs_type = {
2147         .owner          = THIS_MODULE,
2148         .name           = "btrfs",
2149         .mount          = btrfs_mount_root,
2150         .kill_sb        = btrfs_kill_super,
2151         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2152 };
2153
2154 MODULE_ALIAS_FS("btrfs");
2155
2156 static int btrfs_control_open(struct inode *inode, struct file *file)
2157 {
2158         /*
2159          * The control file's private_data is used to hold the
2160          * transaction when it is started and is used to keep
2161          * track of whether a transaction is already in progress.
2162          */
2163         file->private_data = NULL;
2164         return 0;
2165 }
2166
2167 /*
2168  * Used by /dev/btrfs-control for devices ioctls.
2169  */
2170 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2171                                 unsigned long arg)
2172 {
2173         struct btrfs_ioctl_vol_args *vol;
2174         struct btrfs_device *device = NULL;
2175         dev_t devt = 0;
2176         int ret = -ENOTTY;
2177
2178         if (!capable(CAP_SYS_ADMIN))
2179                 return -EPERM;
2180
2181         vol = memdup_user((void __user *)arg, sizeof(*vol));
2182         if (IS_ERR(vol))
2183                 return PTR_ERR(vol);
2184         vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2185
2186         switch (cmd) {
2187         case BTRFS_IOC_SCAN_DEV:
2188                 mutex_lock(&uuid_mutex);
2189                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2190                                                &btrfs_root_fs_type);
2191                 ret = PTR_ERR_OR_ZERO(device);
2192                 mutex_unlock(&uuid_mutex);
2193                 break;
2194         case BTRFS_IOC_FORGET_DEV:
2195                 if (vol->name[0] != 0) {
2196                         ret = lookup_bdev(vol->name, &devt);
2197                         if (ret)
2198                                 break;
2199                 }
2200                 ret = btrfs_forget_devices(devt);
2201                 break;
2202         case BTRFS_IOC_DEVICES_READY:
2203                 mutex_lock(&uuid_mutex);
2204                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2205                                                &btrfs_root_fs_type);
2206                 if (IS_ERR(device)) {
2207                         mutex_unlock(&uuid_mutex);
2208                         ret = PTR_ERR(device);
2209                         break;
2210                 }
2211                 ret = !(device->fs_devices->num_devices ==
2212                         device->fs_devices->total_devices);
2213                 mutex_unlock(&uuid_mutex);
2214                 break;
2215         case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2216                 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2217                 break;
2218         }
2219
2220         kfree(vol);
2221         return ret;
2222 }
2223
2224 static int btrfs_freeze(struct super_block *sb)
2225 {
2226         struct btrfs_trans_handle *trans;
2227         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2228         struct btrfs_root *root = fs_info->tree_root;
2229
2230         set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2231         /*
2232          * We don't need a barrier here, we'll wait for any transaction that
2233          * could be in progress on other threads (and do delayed iputs that
2234          * we want to avoid on a frozen filesystem), or do the commit
2235          * ourselves.
2236          */
2237         trans = btrfs_attach_transaction_barrier(root);
2238         if (IS_ERR(trans)) {
2239                 /* no transaction, don't bother */
2240                 if (PTR_ERR(trans) == -ENOENT)
2241                         return 0;
2242                 return PTR_ERR(trans);
2243         }
2244         return btrfs_commit_transaction(trans);
2245 }
2246
2247 static int check_dev_super(struct btrfs_device *dev)
2248 {
2249         struct btrfs_fs_info *fs_info = dev->fs_info;
2250         struct btrfs_super_block *sb;
2251         u16 csum_type;
2252         int ret = 0;
2253
2254         /* This should be called with fs still frozen. */
2255         ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2256
2257         /* Missing dev, no need to check. */
2258         if (!dev->bdev)
2259                 return 0;
2260
2261         /* Only need to check the primary super block. */
2262         sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2263         if (IS_ERR(sb))
2264                 return PTR_ERR(sb);
2265
2266         /* Verify the checksum. */
2267         csum_type = btrfs_super_csum_type(sb);
2268         if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2269                 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2270                           csum_type, btrfs_super_csum_type(fs_info->super_copy));
2271                 ret = -EUCLEAN;
2272                 goto out;
2273         }
2274
2275         if (btrfs_check_super_csum(fs_info, sb)) {
2276                 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2277                 ret = -EUCLEAN;
2278                 goto out;
2279         }
2280
2281         /* Btrfs_validate_super() includes fsid check against super->fsid. */
2282         ret = btrfs_validate_super(fs_info, sb, 0);
2283         if (ret < 0)
2284                 goto out;
2285
2286         if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
2287                 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2288                         btrfs_super_generation(sb),
2289                         fs_info->last_trans_committed);
2290                 ret = -EUCLEAN;
2291                 goto out;
2292         }
2293 out:
2294         btrfs_release_disk_super(sb);
2295         return ret;
2296 }
2297
2298 static int btrfs_unfreeze(struct super_block *sb)
2299 {
2300         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2301         struct btrfs_device *device;
2302         int ret = 0;
2303
2304         /*
2305          * Make sure the fs is not changed by accident (like hibernation then
2306          * modified by other OS).
2307          * If we found anything wrong, we mark the fs error immediately.
2308          *
2309          * And since the fs is frozen, no one can modify the fs yet, thus
2310          * we don't need to hold device_list_mutex.
2311          */
2312         list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2313                 ret = check_dev_super(device);
2314                 if (ret < 0) {
2315                         btrfs_handle_fs_error(fs_info, ret,
2316                                 "super block on devid %llu got modified unexpectedly",
2317                                 device->devid);
2318                         break;
2319                 }
2320         }
2321         clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2322
2323         /*
2324          * We still return 0, to allow VFS layer to unfreeze the fs even the
2325          * above checks failed. Since the fs is either fine or read-only, we're
2326          * safe to continue, without causing further damage.
2327          */
2328         return 0;
2329 }
2330
2331 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2332 {
2333         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2334
2335         /*
2336          * There should be always a valid pointer in latest_dev, it may be stale
2337          * for a short moment in case it's being deleted but still valid until
2338          * the end of RCU grace period.
2339          */
2340         rcu_read_lock();
2341         seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2342         rcu_read_unlock();
2343
2344         return 0;
2345 }
2346
2347 static const struct super_operations btrfs_super_ops = {
2348         .drop_inode     = btrfs_drop_inode,
2349         .evict_inode    = btrfs_evict_inode,
2350         .put_super      = btrfs_put_super,
2351         .sync_fs        = btrfs_sync_fs,
2352         .show_options   = btrfs_show_options,
2353         .show_devname   = btrfs_show_devname,
2354         .alloc_inode    = btrfs_alloc_inode,
2355         .destroy_inode  = btrfs_destroy_inode,
2356         .free_inode     = btrfs_free_inode,
2357         .statfs         = btrfs_statfs,
2358         .remount_fs     = btrfs_remount,
2359         .freeze_fs      = btrfs_freeze,
2360         .unfreeze_fs    = btrfs_unfreeze,
2361 };
2362
2363 static const struct file_operations btrfs_ctl_fops = {
2364         .open = btrfs_control_open,
2365         .unlocked_ioctl  = btrfs_control_ioctl,
2366         .compat_ioctl = compat_ptr_ioctl,
2367         .owner   = THIS_MODULE,
2368         .llseek = noop_llseek,
2369 };
2370
2371 static struct miscdevice btrfs_misc = {
2372         .minor          = BTRFS_MINOR,
2373         .name           = "btrfs-control",
2374         .fops           = &btrfs_ctl_fops
2375 };
2376
2377 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2378 MODULE_ALIAS("devname:btrfs-control");
2379
2380 static int __init btrfs_interface_init(void)
2381 {
2382         return misc_register(&btrfs_misc);
2383 }
2384
2385 static __cold void btrfs_interface_exit(void)
2386 {
2387         misc_deregister(&btrfs_misc);
2388 }
2389
2390 static int __init btrfs_print_mod_info(void)
2391 {
2392         static const char options[] = ""
2393 #ifdef CONFIG_BTRFS_DEBUG
2394                         ", debug=on"
2395 #endif
2396 #ifdef CONFIG_BTRFS_ASSERT
2397                         ", assert=on"
2398 #endif
2399 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2400                         ", integrity-checker=on"
2401 #endif
2402 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2403                         ", ref-verify=on"
2404 #endif
2405 #ifdef CONFIG_BLK_DEV_ZONED
2406                         ", zoned=yes"
2407 #else
2408                         ", zoned=no"
2409 #endif
2410 #ifdef CONFIG_FS_VERITY
2411                         ", fsverity=yes"
2412 #else
2413                         ", fsverity=no"
2414 #endif
2415                         ;
2416         pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2417         return 0;
2418 }
2419
2420 static int register_btrfs(void)
2421 {
2422         return register_filesystem(&btrfs_fs_type);
2423 }
2424
2425 static void unregister_btrfs(void)
2426 {
2427         unregister_filesystem(&btrfs_fs_type);
2428 }
2429
2430 /* Helper structure for long init/exit functions. */
2431 struct init_sequence {
2432         int (*init_func)(void);
2433         /* Can be NULL if the init_func doesn't need cleanup. */
2434         void (*exit_func)(void);
2435 };
2436
2437 static const struct init_sequence mod_init_seq[] = {
2438         {
2439                 .init_func = btrfs_props_init,
2440                 .exit_func = NULL,
2441         }, {
2442                 .init_func = btrfs_init_sysfs,
2443                 .exit_func = btrfs_exit_sysfs,
2444         }, {
2445                 .init_func = btrfs_init_compress,
2446                 .exit_func = btrfs_exit_compress,
2447         }, {
2448                 .init_func = btrfs_init_cachep,
2449                 .exit_func = btrfs_destroy_cachep,
2450         }, {
2451                 .init_func = btrfs_transaction_init,
2452                 .exit_func = btrfs_transaction_exit,
2453         }, {
2454                 .init_func = btrfs_ctree_init,
2455                 .exit_func = btrfs_ctree_exit,
2456         }, {
2457                 .init_func = btrfs_free_space_init,
2458                 .exit_func = btrfs_free_space_exit,
2459         }, {
2460                 .init_func = extent_state_init_cachep,
2461                 .exit_func = extent_state_free_cachep,
2462         }, {
2463                 .init_func = extent_buffer_init_cachep,
2464                 .exit_func = extent_buffer_free_cachep,
2465         }, {
2466                 .init_func = btrfs_bioset_init,
2467                 .exit_func = btrfs_bioset_exit,
2468         }, {
2469                 .init_func = extent_map_init,
2470                 .exit_func = extent_map_exit,
2471         }, {
2472                 .init_func = ordered_data_init,
2473                 .exit_func = ordered_data_exit,
2474         }, {
2475                 .init_func = btrfs_delayed_inode_init,
2476                 .exit_func = btrfs_delayed_inode_exit,
2477         }, {
2478                 .init_func = btrfs_auto_defrag_init,
2479                 .exit_func = btrfs_auto_defrag_exit,
2480         }, {
2481                 .init_func = btrfs_delayed_ref_init,
2482                 .exit_func = btrfs_delayed_ref_exit,
2483         }, {
2484                 .init_func = btrfs_prelim_ref_init,
2485                 .exit_func = btrfs_prelim_ref_exit,
2486         }, {
2487                 .init_func = btrfs_interface_init,
2488                 .exit_func = btrfs_interface_exit,
2489         }, {
2490                 .init_func = btrfs_print_mod_info,
2491                 .exit_func = NULL,
2492         }, {
2493                 .init_func = btrfs_run_sanity_tests,
2494                 .exit_func = NULL,
2495         }, {
2496                 .init_func = register_btrfs,
2497                 .exit_func = unregister_btrfs,
2498         }
2499 };
2500
2501 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2502
2503 static __always_inline void btrfs_exit_btrfs_fs(void)
2504 {
2505         int i;
2506
2507         for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2508                 if (!mod_init_result[i])
2509                         continue;
2510                 if (mod_init_seq[i].exit_func)
2511                         mod_init_seq[i].exit_func();
2512                 mod_init_result[i] = false;
2513         }
2514 }
2515
2516 static void __exit exit_btrfs_fs(void)
2517 {
2518         btrfs_exit_btrfs_fs();
2519         btrfs_cleanup_fs_uuids();
2520 }
2521
2522 static int __init init_btrfs_fs(void)
2523 {
2524         int ret;
2525         int i;
2526
2527         for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2528                 ASSERT(!mod_init_result[i]);
2529                 ret = mod_init_seq[i].init_func();
2530                 if (ret < 0) {
2531                         btrfs_exit_btrfs_fs();
2532                         return ret;
2533                 }
2534                 mod_init_result[i] = true;
2535         }
2536         return 0;
2537 }
2538
2539 late_initcall(init_btrfs_fs);
2540 module_exit(exit_btrfs_fs)
2541
2542 MODULE_LICENSE("GPL");
2543 MODULE_SOFTDEP("pre: crc32c");
2544 MODULE_SOFTDEP("pre: xxhash64");
2545 MODULE_SOFTDEP("pre: sha256");
2546 MODULE_SOFTDEP("pre: blake2b-256");