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 "volumes.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 #define CREATE_TRACE_POINTS
  61 #include <trace/events/btrfs.h>
  62
  63 static const struct super_operations btrfs_super_ops;
  64
  65 /*
  66  * Types for mounting the default subvolume and a subvolume explicitly
  67  * requested by subvol=/path. That way the callchain is straightforward and we
  68  * don't have to play tricks with the mount options and recursive calls to
  69  * btrfs_mount.
  70  *
  71  * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
  72  */
  73 static struct file_system_type btrfs_fs_type;
  74 static struct file_system_type btrfs_root_fs_type;
  75
  76 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
  77
  78 static void btrfs_put_super(struct super_block *sb)
  79 {
  80         close_ctree(btrfs_sb(sb));
  81 }
  82
  83 enum {
  84         Opt_acl, Opt_noacl,
  85         Opt_clear_cache,
  86         Opt_commit_interval,
  87         Opt_compress,
  88         Opt_compress_force,
  89         Opt_compress_force_type,
  90         Opt_compress_type,
  91         Opt_degraded,
  92         Opt_device,
  93         Opt_fatal_errors,
  94         Opt_flushoncommit, Opt_noflushoncommit,
  95         Opt_max_inline,
  96         Opt_barrier, Opt_nobarrier,
  97         Opt_datacow, Opt_nodatacow,
  98         Opt_datasum, Opt_nodatasum,
  99         Opt_defrag, Opt_nodefrag,
 100         Opt_discard, Opt_nodiscard,
 101         Opt_discard_mode,
 102         Opt_norecovery,
 103         Opt_ratio,
 104         Opt_rescan_uuid_tree,
 105         Opt_skip_balance,
 106         Opt_space_cache, Opt_no_space_cache,
 107         Opt_space_cache_version,
 108         Opt_ssd, Opt_nossd,
 109         Opt_ssd_spread, Opt_nossd_spread,
 110         Opt_subvol,
 111         Opt_subvol_empty,
 112         Opt_subvolid,
 113         Opt_thread_pool,
 114         Opt_treelog, Opt_notreelog,
 115         Opt_user_subvol_rm_allowed,
 116
 117         /* Rescue options */
 118         Opt_rescue,
 119         Opt_usebackuproot,
 120         Opt_nologreplay,
 121         Opt_ignorebadroots,
 122         Opt_ignoredatacsums,
 123         Opt_rescue_all,
 124
 125         /* Deprecated options */
 126         Opt_recovery,
 127         Opt_inode_cache, Opt_noinode_cache,
 128
 129         /* Debugging options */
 130         Opt_check_integrity,
 131         Opt_check_integrity_including_extent_data,
 132         Opt_check_integrity_print_mask,
 133         Opt_enospc_debug, Opt_noenospc_debug,
 134 #ifdef CONFIG_BTRFS_DEBUG
 135         Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 136 #endif
 137 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 138         Opt_ref_verify,
 139 #endif
 140         Opt_err,
 141 };
 142
 143 static const match_table_t tokens = {
 144         {Opt_acl, "acl"},
 145         {Opt_noacl, "noacl"},
 146         {Opt_clear_cache, "clear_cache"},
 147         {Opt_commit_interval, "commit=%u"},
 148         {Opt_compress, "compress"},
 149         {Opt_compress_type, "compress=%s"},
 150         {Opt_compress_force, "compress-force"},
 151         {Opt_compress_force_type, "compress-force=%s"},
 152         {Opt_degraded, "degraded"},
 153         {Opt_device, "device=%s"},
 154         {Opt_fatal_errors, "fatal_errors=%s"},
 155         {Opt_flushoncommit, "flushoncommit"},
 156         {Opt_noflushoncommit, "noflushoncommit"},
 157         {Opt_inode_cache, "inode_cache"},
 158         {Opt_noinode_cache, "noinode_cache"},
 159         {Opt_max_inline, "max_inline=%s"},
 160         {Opt_barrier, "barrier"},
 161         {Opt_nobarrier, "nobarrier"},
 162         {Opt_datacow, "datacow"},
 163         {Opt_nodatacow, "nodatacow"},
 164         {Opt_datasum, "datasum"},
 165         {Opt_nodatasum, "nodatasum"},
 166         {Opt_defrag, "autodefrag"},
 167         {Opt_nodefrag, "noautodefrag"},
 168         {Opt_discard, "discard"},
 169         {Opt_discard_mode, "discard=%s"},
 170         {Opt_nodiscard, "nodiscard"},
 171         {Opt_norecovery, "norecovery"},
 172         {Opt_ratio, "metadata_ratio=%u"},
 173         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 174         {Opt_skip_balance, "skip_balance"},
 175         {Opt_space_cache, "space_cache"},
 176         {Opt_no_space_cache, "nospace_cache"},
 177         {Opt_space_cache_version, "space_cache=%s"},
 178         {Opt_ssd, "ssd"},
 179         {Opt_nossd, "nossd"},
 180         {Opt_ssd_spread, "ssd_spread"},
 181         {Opt_nossd_spread, "nossd_spread"},
 182         {Opt_subvol, "subvol=%s"},
 183         {Opt_subvol_empty, "subvol="},
 184         {Opt_subvolid, "subvolid=%s"},
 185         {Opt_thread_pool, "thread_pool=%u"},
 186         {Opt_treelog, "treelog"},
 187         {Opt_notreelog, "notreelog"},
 188         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 189
 190         /* Rescue options */
 191         {Opt_rescue, "rescue=%s"},
 192         /* Deprecated, with alias rescue=nologreplay */
 193         {Opt_nologreplay, "nologreplay"},
 194         /* Deprecated, with alias rescue=usebackuproot */
 195         {Opt_usebackuproot, "usebackuproot"},
 196
 197         /* Deprecated options */
 198         {Opt_recovery, "recovery"},
 199
 200         /* Debugging options */
 201         {Opt_check_integrity, "check_int"},
 202         {Opt_check_integrity_including_extent_data, "check_int_data"},
 203         {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
 204         {Opt_enospc_debug, "enospc_debug"},
 205         {Opt_noenospc_debug, "noenospc_debug"},
 206 #ifdef CONFIG_BTRFS_DEBUG
 207         {Opt_fragment_data, "fragment=data"},
 208         {Opt_fragment_metadata, "fragment=metadata"},
 209         {Opt_fragment_all, "fragment=all"},
 210 #endif
 211 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 212         {Opt_ref_verify, "ref_verify"},
 213 #endif
 214         {Opt_err, NULL},
 215 };
 216
 217 static const match_table_t rescue_tokens = {
 218         {Opt_usebackuproot, "usebackuproot"},
 219         {Opt_nologreplay, "nologreplay"},
 220         {Opt_ignorebadroots, "ignorebadroots"},
 221         {Opt_ignorebadroots, "ibadroots"},
 222         {Opt_ignoredatacsums, "ignoredatacsums"},
 223         {Opt_ignoredatacsums, "idatacsums"},
 224         {Opt_rescue_all, "all"},
 225         {Opt_err, NULL},
 226 };
 227
 228 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
 229                             const char *opt_name)
 230 {
 231         if (fs_info->mount_opt & opt) {
 232                 btrfs_err(fs_info, "%s must be used with ro mount option",
 233                           opt_name);
 234                 return true;
 235         }
 236         return false;
 237 }
 238
 239 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
 240 {
 241         char *opts;
 242         char *orig;
 243         char *p;
 244         substring_t args[MAX_OPT_ARGS];
 245         int ret = 0;
 246
 247         opts = kstrdup(options, GFP_KERNEL);
 248         if (!opts)
 249                 return -ENOMEM;
 250         orig = opts;
 251
 252         while ((p = strsep(&opts, ":")) != NULL) {
 253                 int token;
 254
 255                 if (!*p)
 256                         continue;
 257                 token = match_token(p, rescue_tokens, args);
 258                 switch (token){
 259                 case Opt_usebackuproot:
 260                         btrfs_info(info,
 261                                    "trying to use backup root at mount time");
 262                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 263                         break;
 264                 case Opt_nologreplay:
 265                         btrfs_set_and_info(info, NOLOGREPLAY,
 266                                            "disabling log replay at mount time");
 267                         break;
 268                 case Opt_ignorebadroots:
 269                         btrfs_set_and_info(info, IGNOREBADROOTS,
 270                                            "ignoring bad roots");
 271                         break;
 272                 case Opt_ignoredatacsums:
 273                         btrfs_set_and_info(info, IGNOREDATACSUMS,
 274                                            "ignoring data csums");
 275                         break;
 276                 case Opt_rescue_all:
 277                         btrfs_info(info, "enabling all of the rescue options");
 278                         btrfs_set_and_info(info, IGNOREDATACSUMS,
 279                                            "ignoring data csums");
 280                         btrfs_set_and_info(info, IGNOREBADROOTS,
 281                                            "ignoring bad roots");
 282                         btrfs_set_and_info(info, NOLOGREPLAY,
 283                                            "disabling log replay at mount time");
 284                         break;
 285                 case Opt_err:
 286                         btrfs_info(info, "unrecognized rescue option '%s'", p);
 287                         ret = -EINVAL;
 288                         goto out;
 289                 default:
 290                         break;
 291                 }
 292
 293         }
 294 out:
 295         kfree(orig);
 296         return ret;
 297 }
 298
 299 /*
 300  * Regular mount options parser.  Everything that is needed only when
 301  * reading in a new superblock is parsed here.
 302  * XXX JDM: This needs to be cleaned up for remount.
 303  */
 304 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 305                         unsigned long new_flags)
 306 {
 307         substring_t args[MAX_OPT_ARGS];
 308         char *p, *num;
 309         int intarg;
 310         int ret = 0;
 311         char *compress_type;
 312         bool compress_force = false;
 313         enum btrfs_compression_type saved_compress_type;
 314         int saved_compress_level;
 315         bool saved_compress_force;
 316         int no_compress = 0;
 317         const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
 318
 319         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
 320                 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
 321         else if (btrfs_free_space_cache_v1_active(info)) {
 322                 if (btrfs_is_zoned(info)) {
 323                         btrfs_info(info,
 324                         "zoned: clearing existing space cache");
 325                         btrfs_set_super_cache_generation(info->super_copy, 0);
 326                 } else {
 327                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 328                 }
 329         }
 330
 331         /*
 332          * Even the options are empty, we still need to do extra check
 333          * against new flags
 334          */
 335         if (!options)
 336                 goto check;
 337
 338         while ((p = strsep(&options, ",")) != NULL) {
 339                 int token;
 340                 if (!*p)
 341                         continue;
 342
 343                 token = match_token(p, tokens, args);
 344                 switch (token) {
 345                 case Opt_degraded:
 346                         btrfs_info(info, "allowing degraded mounts");
 347                         btrfs_set_opt(info->mount_opt, DEGRADED);
 348                         break;
 349                 case Opt_subvol:
 350                 case Opt_subvol_empty:
 351                 case Opt_subvolid:
 352                 case Opt_device:
 353                         /*
 354                          * These are parsed by btrfs_parse_subvol_options or
 355                          * btrfs_parse_device_options and can be ignored here.
 356                          */
 357                         break;
 358                 case Opt_nodatasum:
 359                         btrfs_set_and_info(info, NODATASUM,
 360                                            "setting nodatasum");
 361                         break;
 362                 case Opt_datasum:
 363                         if (btrfs_test_opt(info, NODATASUM)) {
 364                                 if (btrfs_test_opt(info, NODATACOW))
 365                                         btrfs_info(info,
 366                                                    "setting datasum, datacow enabled");
 367                                 else
 368                                         btrfs_info(info, "setting datasum");
 369                         }
 370                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 371                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 372                         break;
 373                 case Opt_nodatacow:
 374                         if (!btrfs_test_opt(info, NODATACOW)) {
 375                                 if (!btrfs_test_opt(info, COMPRESS) ||
 376                                     !btrfs_test_opt(info, FORCE_COMPRESS)) {
 377                                         btrfs_info(info,
 378                                                    "setting nodatacow, compression disabled");
 379                                 } else {
 380                                         btrfs_info(info, "setting nodatacow");
 381                                 }
 382                         }
 383                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 384                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 385                         btrfs_set_opt(info->mount_opt, NODATACOW);
 386                         btrfs_set_opt(info->mount_opt, NODATASUM);
 387                         break;
 388                 case Opt_datacow:
 389                         btrfs_clear_and_info(info, NODATACOW,
 390                                              "setting datacow");
 391                         break;
 392                 case Opt_compress_force:
 393                 case Opt_compress_force_type:
 394                         compress_force = true;
 395                         fallthrough;
 396                 case Opt_compress:
 397                 case Opt_compress_type:
 398                         saved_compress_type = btrfs_test_opt(info,
 399                                                              COMPRESS) ?
 400                                 info->compress_type : BTRFS_COMPRESS_NONE;
 401                         saved_compress_force =
 402                                 btrfs_test_opt(info, FORCE_COMPRESS);
 403                         saved_compress_level = info->compress_level;
 404                         if (token == Opt_compress ||
 405                             token == Opt_compress_force ||
 406                             strncmp(args[0].from, "zlib", 4) == 0) {
 407                                 compress_type = "zlib";
 408
 409                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 410                                 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 411                                 /*
 412                                  * args[0] contains uninitialized data since
 413                                  * for these tokens we don't expect any
 414                                  * parameter.
 415                                  */
 416                                 if (token != Opt_compress &&
 417                                     token != Opt_compress_force)
 418                                         info->compress_level =
 419                                           btrfs_compress_str2level(
 420                                                         BTRFS_COMPRESS_ZLIB,
 421                                                         args[0].from + 4);
 422                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 423                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 424                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 425                                 no_compress = 0;
 426                         } else if (strncmp(args[0].from, "lzo", 3) == 0) {
 427                                 compress_type = "lzo";
 428                                 info->compress_type = BTRFS_COMPRESS_LZO;
 429                                 info->compress_level = 0;
 430                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 431                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 432                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 433                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 434                                 no_compress = 0;
 435                         } else if (strncmp(args[0].from, "zstd", 4) == 0) {
 436                                 compress_type = "zstd";
 437                                 info->compress_type = BTRFS_COMPRESS_ZSTD;
 438                                 info->compress_level =
 439                                         btrfs_compress_str2level(
 440                                                          BTRFS_COMPRESS_ZSTD,
 441                                                          args[0].from + 4);
 442                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 443                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 444                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 445                                 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
 446                                 no_compress = 0;
 447                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 448                                 compress_type = "no";
 449                                 info->compress_level = 0;
 450                                 info->compress_type = 0;
 451                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 452                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 453                                 compress_force = false;
 454                                 no_compress++;
 455                         } else {
 456                                 btrfs_err(info, "unrecognized compression value %s",
 457                                           args[0].from);
 458                                 ret = -EINVAL;
 459                                 goto out;
 460                         }
 461
 462                         if (compress_force) {
 463                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 464                         } else {
 465                                 /*
 466                                  * If we remount from compress-force=xxx to
 467                                  * compress=xxx, we need clear FORCE_COMPRESS
 468                                  * flag, otherwise, there is no way for users
 469                                  * to disable forcible compression separately.
 470                                  */
 471                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 472                         }
 473                         if (no_compress == 1) {
 474                                 btrfs_info(info, "use no compression");
 475                         } else if ((info->compress_type != saved_compress_type) ||
 476                                    (compress_force != saved_compress_force) ||
 477                                    (info->compress_level != saved_compress_level)) {
 478                                 btrfs_info(info, "%s %s compression, level %d",
 479                                            (compress_force) ? "force" : "use",
 480                                            compress_type, info->compress_level);
 481                         }
 482                         compress_force = false;
 483                         break;
 484                 case Opt_ssd:
 485                         btrfs_set_and_info(info, SSD,
 486                                            "enabling ssd optimizations");
 487                         btrfs_clear_opt(info->mount_opt, NOSSD);
 488                         break;
 489                 case Opt_ssd_spread:
 490                         btrfs_set_and_info(info, SSD,
 491                                            "enabling ssd optimizations");
 492                         btrfs_set_and_info(info, SSD_SPREAD,
 493                                            "using spread ssd allocation scheme");
 494                         btrfs_clear_opt(info->mount_opt, NOSSD);
 495                         break;
 496                 case Opt_nossd:
 497                         btrfs_set_opt(info->mount_opt, NOSSD);
 498                         btrfs_clear_and_info(info, SSD,
 499                                              "not using ssd optimizations");
 500                         fallthrough;
 501                 case Opt_nossd_spread:
 502                         btrfs_clear_and_info(info, SSD_SPREAD,
 503                                              "not using spread ssd allocation scheme");
 504                         break;
 505                 case Opt_barrier:
 506                         btrfs_clear_and_info(info, NOBARRIER,
 507                                              "turning on barriers");
 508                         break;
 509                 case Opt_nobarrier:
 510                         btrfs_set_and_info(info, NOBARRIER,
 511                                            "turning off barriers");
 512                         break;
 513                 case Opt_thread_pool:
 514                         ret = match_int(&args[0], &intarg);
 515                         if (ret) {
 516                                 btrfs_err(info, "unrecognized thread_pool value %s",
 517                                           args[0].from);
 518                                 goto out;
 519                         } else if (intarg == 0) {
 520                                 btrfs_err(info, "invalid value 0 for thread_pool");
 521                                 ret = -EINVAL;
 522                                 goto out;
 523                         }
 524                         info->thread_pool_size = intarg;
 525                         break;
 526                 case Opt_max_inline:
 527                         num = match_strdup(&args[0]);
 528                         if (num) {
 529                                 info->max_inline = memparse(num, NULL);
 530                                 kfree(num);
 531
 532                                 if (info->max_inline) {
 533                                         info->max_inline = min_t(u64,
 534                                                 info->max_inline,
 535                                                 info->sectorsize);
 536                                 }
 537                                 btrfs_info(info, "max_inline at %llu",
 538                                            info->max_inline);
 539                         } else {
 540                                 ret = -ENOMEM;
 541                                 goto out;
 542                         }
 543                         break;
 544                 case Opt_acl:
 545 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 546                         info->sb->s_flags |= SB_POSIXACL;
 547                         break;
 548 #else
 549                         btrfs_err(info, "support for ACL not compiled in!");
 550                         ret = -EINVAL;
 551                         goto out;
 552 #endif
 553                 case Opt_noacl:
 554                         info->sb->s_flags &= ~SB_POSIXACL;
 555                         break;
 556                 case Opt_notreelog:
 557                         btrfs_set_and_info(info, NOTREELOG,
 558                                            "disabling tree log");
 559                         break;
 560                 case Opt_treelog:
 561                         btrfs_clear_and_info(info, NOTREELOG,
 562                                              "enabling tree log");
 563                         break;
 564                 case Opt_norecovery:
 565                 case Opt_nologreplay:
 566                         btrfs_warn(info,
 567                 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 568                         btrfs_set_and_info(info, NOLOGREPLAY,
 569                                            "disabling log replay at mount time");
 570                         break;
 571                 case Opt_flushoncommit:
 572                         btrfs_set_and_info(info, FLUSHONCOMMIT,
 573                                            "turning on flush-on-commit");
 574                         break;
 575                 case Opt_noflushoncommit:
 576                         btrfs_clear_and_info(info, FLUSHONCOMMIT,
 577                                              "turning off flush-on-commit");
 578                         break;
 579                 case Opt_ratio:
 580                         ret = match_int(&args[0], &intarg);
 581                         if (ret) {
 582                                 btrfs_err(info, "unrecognized metadata_ratio value %s",
 583                                           args[0].from);
 584                                 goto out;
 585                         }
 586                         info->metadata_ratio = intarg;
 587                         btrfs_info(info, "metadata ratio %u",
 588                                    info->metadata_ratio);
 589                         break;
 590                 case Opt_discard:
 591                 case Opt_discard_mode:
 592                         if (token == Opt_discard ||
 593                             strcmp(args[0].from, "sync") == 0) {
 594                                 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
 595                                 btrfs_set_and_info(info, DISCARD_SYNC,
 596                                                    "turning on sync discard");
 597                         } else if (strcmp(args[0].from, "async") == 0) {
 598                                 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
 599                                 btrfs_set_and_info(info, DISCARD_ASYNC,
 600                                                    "turning on async discard");
 601                         } else {
 602                                 btrfs_err(info, "unrecognized discard mode value %s",
 603                                           args[0].from);
 604                                 ret = -EINVAL;
 605                                 goto out;
 606                         }
 607                         btrfs_clear_opt(info->mount_opt, NODISCARD);
 608                         break;
 609                 case Opt_nodiscard:
 610                         btrfs_clear_and_info(info, DISCARD_SYNC,
 611                                              "turning off discard");
 612                         btrfs_clear_and_info(info, DISCARD_ASYNC,
 613                                              "turning off async discard");
 614                         btrfs_set_opt(info->mount_opt, NODISCARD);
 615                         break;
 616                 case Opt_space_cache:
 617                 case Opt_space_cache_version:
 618                         /*
 619                          * We already set FREE_SPACE_TREE above because we have
 620                          * compat_ro(FREE_SPACE_TREE) set, and we aren't going
 621                          * to allow v1 to be set for extent tree v2, simply
 622                          * ignore this setting if we're extent tree v2.
 623                          */
 624                         if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
 625                                 break;
 626                         if (token == Opt_space_cache ||
 627                             strcmp(args[0].from, "v1") == 0) {
 628                                 btrfs_clear_opt(info->mount_opt,
 629                                                 FREE_SPACE_TREE);
 630                                 btrfs_set_and_info(info, SPACE_CACHE,
 631                                            "enabling disk space caching");
 632                         } else if (strcmp(args[0].from, "v2") == 0) {
 633                                 btrfs_clear_opt(info->mount_opt,
 634                                                 SPACE_CACHE);
 635                                 btrfs_set_and_info(info, FREE_SPACE_TREE,
 636                                                    "enabling free space tree");
 637                         } else {
 638                                 btrfs_err(info, "unrecognized space_cache value %s",
 639                                           args[0].from);
 640                                 ret = -EINVAL;
 641                                 goto out;
 642                         }
 643                         break;
 644                 case Opt_rescan_uuid_tree:
 645                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 646                         break;
 647                 case Opt_no_space_cache:
 648                         /*
 649                          * We cannot operate without the free space tree with
 650                          * extent tree v2, ignore this option.
 651                          */
 652                         if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
 653                                 break;
 654                         if (btrfs_test_opt(info, SPACE_CACHE)) {
 655                                 btrfs_clear_and_info(info, SPACE_CACHE,
 656                                              "disabling disk space caching");
 657                         }
 658                         if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
 659                                 btrfs_clear_and_info(info, FREE_SPACE_TREE,
 660                                              "disabling free space tree");
 661                         }
 662                         break;
 663                 case Opt_inode_cache:
 664                 case Opt_noinode_cache:
 665                         btrfs_warn(info,
 666         "the 'inode_cache' option is deprecated and has no effect since 5.11");
 667                         break;
 668                 case Opt_clear_cache:
 669                         /*
 670                          * We cannot clear the free space tree with extent tree
 671                          * v2, ignore this option.
 672                          */
 673                         if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
 674                                 break;
 675                         btrfs_set_and_info(info, CLEAR_CACHE,
 676                                            "force clearing of disk cache");
 677                         break;
 678                 case Opt_user_subvol_rm_allowed:
 679                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 680                         break;
 681                 case Opt_enospc_debug:
 682                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 683                         break;
 684                 case Opt_noenospc_debug:
 685                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 686                         break;
 687                 case Opt_defrag:
 688                         btrfs_set_and_info(info, AUTO_DEFRAG,
 689                                            "enabling auto defrag");
 690                         break;
 691                 case Opt_nodefrag:
 692                         btrfs_clear_and_info(info, AUTO_DEFRAG,
 693                                              "disabling auto defrag");
 694                         break;
 695                 case Opt_recovery:
 696                 case Opt_usebackuproot:
 697                         btrfs_warn(info,
 698                         "'%s' is deprecated, use 'rescue=usebackuproot' instead",
 699                                    token == Opt_recovery ? "recovery" :
 700                                    "usebackuproot");
 701                         btrfs_info(info,
 702                                    "trying to use backup root at mount time");
 703                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 704                         break;
 705                 case Opt_skip_balance:
 706                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 707                         break;
 708 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 709                 case Opt_check_integrity_including_extent_data:
 710                         btrfs_info(info,
 711                                    "enabling check integrity including extent data");
 712                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
 713                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 714                         break;
 715                 case Opt_check_integrity:
 716                         btrfs_info(info, "enabling check integrity");
 717                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 718                         break;
 719                 case Opt_check_integrity_print_mask:
 720                         ret = match_int(&args[0], &intarg);
 721                         if (ret) {
 722                                 btrfs_err(info,
 723                                 "unrecognized check_integrity_print_mask value %s",
 724                                         args[0].from);
 725                                 goto out;
 726                         }
 727                         info->check_integrity_print_mask = intarg;
 728                         btrfs_info(info, "check_integrity_print_mask 0x%x",
 729                                    info->check_integrity_print_mask);
 730                         break;
 731 #else
 732                 case Opt_check_integrity_including_extent_data:
 733                 case Opt_check_integrity:
 734                 case Opt_check_integrity_print_mask:
 735                         btrfs_err(info,
 736                                   "support for check_integrity* not compiled in!");
 737                         ret = -EINVAL;
 738                         goto out;
 739 #endif
 740                 case Opt_fatal_errors:
 741                         if (strcmp(args[0].from, "panic") == 0) {
 742                                 btrfs_set_opt(info->mount_opt,
 743                                               PANIC_ON_FATAL_ERROR);
 744                         } else if (strcmp(args[0].from, "bug") == 0) {
 745                                 btrfs_clear_opt(info->mount_opt,
 746                                               PANIC_ON_FATAL_ERROR);
 747                         } else {
 748                                 btrfs_err(info, "unrecognized fatal_errors value %s",
 749                                           args[0].from);
 750                                 ret = -EINVAL;
 751                                 goto out;
 752                         }
 753                         break;
 754                 case Opt_commit_interval:
 755                         intarg = 0;
 756                         ret = match_int(&args[0], &intarg);
 757                         if (ret) {
 758                                 btrfs_err(info, "unrecognized commit_interval value %s",
 759                                           args[0].from);
 760                                 ret = -EINVAL;
 761                                 goto out;
 762                         }
 763                         if (intarg == 0) {
 764                                 btrfs_info(info,
 765                                            "using default commit interval %us",
 766                                            BTRFS_DEFAULT_COMMIT_INTERVAL);
 767                                 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
 768                         } else if (intarg > 300) {
 769                                 btrfs_warn(info, "excessive commit interval %d",
 770                                            intarg);
 771                         }
 772                         info->commit_interval = intarg;
 773                         break;
 774                 case Opt_rescue:
 775                         ret = parse_rescue_options(info, args[0].from);
 776                         if (ret < 0) {
 777                                 btrfs_err(info, "unrecognized rescue value %s",
 778                                           args[0].from);
 779                                 goto out;
 780                         }
 781                         break;
 782 #ifdef CONFIG_BTRFS_DEBUG
 783                 case Opt_fragment_all:
 784                         btrfs_info(info, "fragmenting all space");
 785                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 786                         btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
 787                         break;
 788                 case Opt_fragment_metadata:
 789                         btrfs_info(info, "fragmenting metadata");
 790                         btrfs_set_opt(info->mount_opt,
 791                                       FRAGMENT_METADATA);
 792                         break;
 793                 case Opt_fragment_data:
 794                         btrfs_info(info, "fragmenting data");
 795                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 796                         break;
 797 #endif
 798 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 799                 case Opt_ref_verify:
 800                         btrfs_info(info, "doing ref verification");
 801                         btrfs_set_opt(info->mount_opt, REF_VERIFY);
 802                         break;
 803 #endif
 804                 case Opt_err:
 805                         btrfs_err(info, "unrecognized mount option '%s'", p);
 806                         ret = -EINVAL;
 807                         goto out;
 808                 default:
 809                         break;
 810                 }
 811         }
 812 check:
 813         /* We're read-only, don't have to check. */
 814         if (new_flags & SB_RDONLY)
 815                 goto out;
 816
 817         if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
 818             check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
 819             check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
 820                 ret = -EINVAL;
 821 out:
 822         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 823             !btrfs_test_opt(info, FREE_SPACE_TREE) &&
 824             !btrfs_test_opt(info, CLEAR_CACHE)) {
 825                 btrfs_err(info, "cannot disable free space tree");
 826                 ret = -EINVAL;
 827
 828         }
 829         if (!ret)
 830                 ret = btrfs_check_mountopts_zoned(info);
 831         if (!ret && !remounting) {
 832                 if (btrfs_test_opt(info, SPACE_CACHE))
 833                         btrfs_info(info, "disk space caching is enabled");
 834                 if (btrfs_test_opt(info, FREE_SPACE_TREE))
 835                         btrfs_info(info, "using free space tree");
 836         }
 837         return ret;
 838 }
 839
 840 /*
 841  * Parse mount options that are required early in the mount process.
 842  *
 843  * All other options will be parsed on much later in the mount process and
 844  * only when we need to allocate a new super block.
 845  */
 846 static int btrfs_parse_device_options(const char *options, fmode_t flags,
 847                                       void *holder)
 848 {
 849         substring_t args[MAX_OPT_ARGS];
 850         char *device_name, *opts, *orig, *p;
 851         struct btrfs_device *device = NULL;
 852         int error = 0;
 853
 854         lockdep_assert_held(&uuid_mutex);
 855
 856         if (!options)
 857                 return 0;
 858
 859         /*
 860          * strsep changes the string, duplicate it because btrfs_parse_options
 861          * gets called later
 862          */
 863         opts = kstrdup(options, GFP_KERNEL);
 864         if (!opts)
 865                 return -ENOMEM;
 866         orig = opts;
 867
 868         while ((p = strsep(&opts, ",")) != NULL) {
 869                 int token;
 870
 871                 if (!*p)
 872                         continue;
 873
 874                 token = match_token(p, tokens, args);
 875                 if (token == Opt_device) {
 876                         device_name = match_strdup(&args[0]);
 877                         if (!device_name) {
 878                                 error = -ENOMEM;
 879                                 goto out;
 880                         }
 881                         device = btrfs_scan_one_device(device_name, flags,
 882                                         holder);
 883                         kfree(device_name);
 884                         if (IS_ERR(device)) {
 885                                 error = PTR_ERR(device);
 886                                 goto out;
 887                         }
 888                 }
 889         }
 890
 891 out:
 892         kfree(orig);
 893         return error;
 894 }
 895
 896 /*
 897  * Parse mount options that are related to subvolume id
 898  *
 899  * The value is later passed to mount_subvol()
 900  */
 901 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
 902                 u64 *subvol_objectid)
 903 {
 904         substring_t args[MAX_OPT_ARGS];
 905         char *opts, *orig, *p;
 906         int error = 0;
 907         u64 subvolid;
 908
 909         if (!options)
 910                 return 0;
 911
 912         /*
 913          * strsep changes the string, duplicate it because
 914          * btrfs_parse_device_options gets called later
 915          */
 916         opts = kstrdup(options, GFP_KERNEL);
 917         if (!opts)
 918                 return -ENOMEM;
 919         orig = opts;
 920
 921         while ((p = strsep(&opts, ",")) != NULL) {
 922                 int token;
 923                 if (!*p)
 924                         continue;
 925
 926                 token = match_token(p, tokens, args);
 927                 switch (token) {
 928                 case Opt_subvol:
 929                         kfree(*subvol_name);
 930                         *subvol_name = match_strdup(&args[0]);
 931                         if (!*subvol_name) {
 932                                 error = -ENOMEM;
 933                                 goto out;
 934                         }
 935                         break;
 936                 case Opt_subvolid:
 937                         error = match_u64(&args[0], &subvolid);
 938                         if (error)
 939                                 goto out;
 940
 941                         /* we want the original fs_tree */
 942                         if (subvolid == 0)
 943                                 subvolid = BTRFS_FS_TREE_OBJECTID;
 944
 945                         *subvol_objectid = subvolid;
 946                         break;
 947                 default:
 948                         break;
 949                 }
 950         }
 951
 952 out:
 953         kfree(orig);
 954         return error;
 955 }
 956
 957 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 958                                           u64 subvol_objectid)
 959 {
 960         struct btrfs_root *root = fs_info->tree_root;
 961         struct btrfs_root *fs_root = NULL;
 962         struct btrfs_root_ref *root_ref;
 963         struct btrfs_inode_ref *inode_ref;
 964         struct btrfs_key key;
 965         struct btrfs_path *path = NULL;
 966         char *name = NULL, *ptr;
 967         u64 dirid;
 968         int len;
 969         int ret;
 970
 971         path = btrfs_alloc_path();
 972         if (!path) {
 973                 ret = -ENOMEM;
 974                 goto err;
 975         }
 976
 977         name = kmalloc(PATH_MAX, GFP_KERNEL);
 978         if (!name) {
 979                 ret = -ENOMEM;
 980                 goto err;
 981         }
 982         ptr = name + PATH_MAX - 1;
 983         ptr[0] = '\0';
 984
 985         /*
 986          * Walk up the subvolume trees in the tree of tree roots by root
 987          * backrefs until we hit the top-level subvolume.
 988          */
 989         while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
 990                 key.objectid = subvol_objectid;
 991                 key.type = BTRFS_ROOT_BACKREF_KEY;
 992                 key.offset = (u64)-1;
 993
 994                 ret = btrfs_search_backwards(root, &key, path);
 995                 if (ret < 0) {
 996                         goto err;
 997                 } else if (ret > 0) {
 998                         ret = -ENOENT;
 999                         goto err;
1000                 }
1001
1002                 subvol_objectid = key.offset;
1003
1004                 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1005                                           struct btrfs_root_ref);
1006                 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1007                 ptr -= len + 1;
1008                 if (ptr < name) {
1009                         ret = -ENAMETOOLONG;
1010                         goto err;
1011                 }
1012                 read_extent_buffer(path->nodes[0], ptr + 1,
1013                                    (unsigned long)(root_ref + 1), len);
1014                 ptr[0] = '/';
1015                 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1016                 btrfs_release_path(path);
1017
1018                 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1019                 if (IS_ERR(fs_root)) {
1020                         ret = PTR_ERR(fs_root);
1021                         fs_root = NULL;
1022                         goto err;
1023                 }
1024
1025                 /*
1026                  * Walk up the filesystem tree by inode refs until we hit the
1027                  * root directory.
1028                  */
1029                 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1030                         key.objectid = dirid;
1031                         key.type = BTRFS_INODE_REF_KEY;
1032                         key.offset = (u64)-1;
1033
1034                         ret = btrfs_search_backwards(fs_root, &key, path);
1035                         if (ret < 0) {
1036                                 goto err;
1037                         } else if (ret > 0) {
1038                                 ret = -ENOENT;
1039                                 goto err;
1040                         }
1041
1042                         dirid = key.offset;
1043
1044                         inode_ref = btrfs_item_ptr(path->nodes[0],
1045                                                    path->slots[0],
1046                                                    struct btrfs_inode_ref);
1047                         len = btrfs_inode_ref_name_len(path->nodes[0],
1048                                                        inode_ref);
1049                         ptr -= len + 1;
1050                         if (ptr < name) {
1051                                 ret = -ENAMETOOLONG;
1052                                 goto err;
1053                         }
1054                         read_extent_buffer(path->nodes[0], ptr + 1,
1055                                            (unsigned long)(inode_ref + 1), len);
1056                         ptr[0] = '/';
1057                         btrfs_release_path(path);
1058                 }
1059                 btrfs_put_root(fs_root);
1060                 fs_root = NULL;
1061         }
1062
1063         btrfs_free_path(path);
1064         if (ptr == name + PATH_MAX - 1) {
1065                 name[0] = '/';
1066                 name[1] = '\0';
1067         } else {
1068                 memmove(name, ptr, name + PATH_MAX - ptr);
1069         }
1070         return name;
1071
1072 err:
1073         btrfs_put_root(fs_root);
1074         btrfs_free_path(path);
1075         kfree(name);
1076         return ERR_PTR(ret);
1077 }
1078
1079 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1080 {
1081         struct btrfs_root *root = fs_info->tree_root;
1082         struct btrfs_dir_item *di;
1083         struct btrfs_path *path;
1084         struct btrfs_key location;
1085         struct fscrypt_str name = FSTR_INIT("default", 7);
1086         u64 dir_id;
1087
1088         path = btrfs_alloc_path();
1089         if (!path)
1090                 return -ENOMEM;
1091
1092         /*
1093          * Find the "default" dir item which points to the root item that we
1094          * will mount by default if we haven't been given a specific subvolume
1095          * to mount.
1096          */
1097         dir_id = btrfs_super_root_dir(fs_info->super_copy);
1098         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
1099         if (IS_ERR(di)) {
1100                 btrfs_free_path(path);
1101                 return PTR_ERR(di);
1102         }
1103         if (!di) {
1104                 /*
1105                  * Ok the default dir item isn't there.  This is weird since
1106                  * it's always been there, but don't freak out, just try and
1107                  * mount the top-level subvolume.
1108                  */
1109                 btrfs_free_path(path);
1110                 *objectid = BTRFS_FS_TREE_OBJECTID;
1111                 return 0;
1112         }
1113
1114         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1115         btrfs_free_path(path);
1116         *objectid = location.objectid;
1117         return 0;
1118 }
1119
1120 static int btrfs_fill_super(struct super_block *sb,
1121                             struct btrfs_fs_devices *fs_devices,
1122                             void *data)
1123 {
1124         struct inode *inode;
1125         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1126         int err;
1127
1128         sb->s_maxbytes = MAX_LFS_FILESIZE;
1129         sb->s_magic = BTRFS_SUPER_MAGIC;
1130         sb->s_op = &btrfs_super_ops;
1131         sb->s_d_op = &btrfs_dentry_operations;
1132         sb->s_export_op = &btrfs_export_ops;
1133 #ifdef CONFIG_FS_VERITY
1134         sb->s_vop = &btrfs_verityops;
1135 #endif
1136         sb->s_xattr = btrfs_xattr_handlers;
1137         sb->s_time_gran = 1;
1138 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1139         sb->s_flags |= SB_POSIXACL;
1140 #endif
1141         sb->s_flags |= SB_I_VERSION;
1142         sb->s_iflags |= SB_I_CGROUPWB;
1143
1144         err = super_setup_bdi(sb);
1145         if (err) {
1146                 btrfs_err(fs_info, "super_setup_bdi failed");
1147                 return err;
1148         }
1149
1150         err = open_ctree(sb, fs_devices, (char *)data);
1151         if (err) {
1152                 btrfs_err(fs_info, "open_ctree failed");
1153                 return err;
1154         }
1155
1156         inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1157         if (IS_ERR(inode)) {
1158                 err = PTR_ERR(inode);
1159                 goto fail_close;
1160         }
1161
1162         sb->s_root = d_make_root(inode);
1163         if (!sb->s_root) {
1164                 err = -ENOMEM;
1165                 goto fail_close;
1166         }
1167
1168         sb->s_flags |= SB_ACTIVE;
1169         return 0;
1170
1171 fail_close:
1172         close_ctree(fs_info);
1173         return err;
1174 }
1175
1176 int btrfs_sync_fs(struct super_block *sb, int wait)
1177 {
1178         struct btrfs_trans_handle *trans;
1179         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1180         struct btrfs_root *root = fs_info->tree_root;
1181
1182         trace_btrfs_sync_fs(fs_info, wait);
1183
1184         if (!wait) {
1185                 filemap_flush(fs_info->btree_inode->i_mapping);
1186                 return 0;
1187         }
1188
1189         btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1190
1191         trans = btrfs_attach_transaction_barrier(root);
1192         if (IS_ERR(trans)) {
1193                 /* no transaction, don't bother */
1194                 if (PTR_ERR(trans) == -ENOENT) {
1195                         /*
1196                          * Exit unless we have some pending changes
1197                          * that need to go through commit
1198                          */
1199                         if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1200                                       &fs_info->flags))
1201                                 return 0;
1202                         /*
1203                          * A non-blocking test if the fs is frozen. We must not
1204                          * start a new transaction here otherwise a deadlock
1205                          * happens. The pending operations are delayed to the
1206                          * next commit after thawing.
1207                          */
1208                         if (sb_start_write_trylock(sb))
1209                                 sb_end_write(sb);
1210                         else
1211                                 return 0;
1212                         trans = btrfs_start_transaction(root, 0);
1213                 }
1214                 if (IS_ERR(trans))
1215                         return PTR_ERR(trans);
1216         }
1217         return btrfs_commit_transaction(trans);
1218 }
1219
1220 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1221 {
1222         seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1223         *printed = true;
1224 }
1225
1226 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1227 {
1228         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1229         const char *compress_type;
1230         const char *subvol_name;
1231         bool printed = false;
1232
1233         if (btrfs_test_opt(info, DEGRADED))
1234                 seq_puts(seq, ",degraded");
1235         if (btrfs_test_opt(info, NODATASUM))
1236                 seq_puts(seq, ",nodatasum");
1237         if (btrfs_test_opt(info, NODATACOW))
1238                 seq_puts(seq, ",nodatacow");
1239         if (btrfs_test_opt(info, NOBARRIER))
1240                 seq_puts(seq, ",nobarrier");
1241         if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1242                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1243         if (info->thread_pool_size !=  min_t(unsigned long,
1244                                              num_online_cpus() + 2, 8))
1245                 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1246         if (btrfs_test_opt(info, COMPRESS)) {
1247                 compress_type = btrfs_compress_type2str(info->compress_type);
1248                 if (btrfs_test_opt(info, FORCE_COMPRESS))
1249                         seq_printf(seq, ",compress-force=%s", compress_type);
1250                 else
1251                         seq_printf(seq, ",compress=%s", compress_type);
1252                 if (info->compress_level)
1253                         seq_printf(seq, ":%d", info->compress_level);
1254         }
1255         if (btrfs_test_opt(info, NOSSD))
1256                 seq_puts(seq, ",nossd");
1257         if (btrfs_test_opt(info, SSD_SPREAD))
1258                 seq_puts(seq, ",ssd_spread");
1259         else if (btrfs_test_opt(info, SSD))
1260                 seq_puts(seq, ",ssd");
1261         if (btrfs_test_opt(info, NOTREELOG))
1262                 seq_puts(seq, ",notreelog");
1263         if (btrfs_test_opt(info, NOLOGREPLAY))
1264                 print_rescue_option(seq, "nologreplay", &printed);
1265         if (btrfs_test_opt(info, USEBACKUPROOT))
1266                 print_rescue_option(seq, "usebackuproot", &printed);
1267         if (btrfs_test_opt(info, IGNOREBADROOTS))
1268                 print_rescue_option(seq, "ignorebadroots", &printed);
1269         if (btrfs_test_opt(info, IGNOREDATACSUMS))
1270                 print_rescue_option(seq, "ignoredatacsums", &printed);
1271         if (btrfs_test_opt(info, FLUSHONCOMMIT))
1272                 seq_puts(seq, ",flushoncommit");
1273         if (btrfs_test_opt(info, DISCARD_SYNC))
1274                 seq_puts(seq, ",discard");
1275         if (btrfs_test_opt(info, DISCARD_ASYNC))
1276                 seq_puts(seq, ",discard=async");
1277         if (!(info->sb->s_flags & SB_POSIXACL))
1278                 seq_puts(seq, ",noacl");
1279         if (btrfs_free_space_cache_v1_active(info))
1280                 seq_puts(seq, ",space_cache");
1281         else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1282                 seq_puts(seq, ",space_cache=v2");
1283         else
1284                 seq_puts(seq, ",nospace_cache");
1285         if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1286                 seq_puts(seq, ",rescan_uuid_tree");
1287         if (btrfs_test_opt(info, CLEAR_CACHE))
1288                 seq_puts(seq, ",clear_cache");
1289         if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1290                 seq_puts(seq, ",user_subvol_rm_allowed");
1291         if (btrfs_test_opt(info, ENOSPC_DEBUG))
1292                 seq_puts(seq, ",enospc_debug");
1293         if (btrfs_test_opt(info, AUTO_DEFRAG))
1294                 seq_puts(seq, ",autodefrag");
1295         if (btrfs_test_opt(info, SKIP_BALANCE))
1296                 seq_puts(seq, ",skip_balance");
1297 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1298         if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1299                 seq_puts(seq, ",check_int_data");
1300         else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1301                 seq_puts(seq, ",check_int");
1302         if (info->check_integrity_print_mask)
1303                 seq_printf(seq, ",check_int_print_mask=%d",
1304                                 info->check_integrity_print_mask);
1305 #endif
1306         if (info->metadata_ratio)
1307                 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1308         if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1309                 seq_puts(seq, ",fatal_errors=panic");
1310         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1311                 seq_printf(seq, ",commit=%u", info->commit_interval);
1312 #ifdef CONFIG_BTRFS_DEBUG
1313         if (btrfs_test_opt(info, FRAGMENT_DATA))
1314                 seq_puts(seq, ",fragment=data");
1315         if (btrfs_test_opt(info, FRAGMENT_METADATA))
1316                 seq_puts(seq, ",fragment=metadata");
1317 #endif
1318         if (btrfs_test_opt(info, REF_VERIFY))
1319                 seq_puts(seq, ",ref_verify");
1320         seq_printf(seq, ",subvolid=%llu",
1321                   BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1322         subvol_name = btrfs_get_subvol_name_from_objectid(info,
1323                         BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1324         if (!IS_ERR(subvol_name)) {
1325                 seq_puts(seq, ",subvol=");
1326                 seq_escape(seq, subvol_name, " \t\n\\");
1327                 kfree(subvol_name);
1328         }
1329         return 0;
1330 }
1331
1332 static int btrfs_test_super(struct super_block *s, void *data)
1333 {
1334         struct btrfs_fs_info *p = data;
1335         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1336
1337         return fs_info->fs_devices == p->fs_devices;
1338 }
1339
1340 static int btrfs_set_super(struct super_block *s, void *data)
1341 {
1342         int err = set_anon_super(s, data);
1343         if (!err)
1344                 s->s_fs_info = data;
1345         return err;
1346 }
1347
1348 /*
1349  * subvolumes are identified by ino 256
1350  */
1351 static inline int is_subvolume_inode(struct inode *inode)
1352 {
1353         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1354                 return 1;
1355         return 0;
1356 }
1357
1358 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1359                                    struct vfsmount *mnt)
1360 {
1361         struct dentry *root;
1362         int ret;
1363
1364         if (!subvol_name) {
1365                 if (!subvol_objectid) {
1366                         ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1367                                                           &subvol_objectid);
1368                         if (ret) {
1369                                 root = ERR_PTR(ret);
1370                                 goto out;
1371                         }
1372                 }
1373                 subvol_name = btrfs_get_subvol_name_from_objectid(
1374                                         btrfs_sb(mnt->mnt_sb), subvol_objectid);
1375                 if (IS_ERR(subvol_name)) {
1376                         root = ERR_CAST(subvol_name);
1377                         subvol_name = NULL;
1378                         goto out;
1379                 }
1380
1381         }
1382
1383         root = mount_subtree(mnt, subvol_name);
1384         /* mount_subtree() drops our reference on the vfsmount. */
1385         mnt = NULL;
1386
1387         if (!IS_ERR(root)) {
1388                 struct super_block *s = root->d_sb;
1389                 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1390                 struct inode *root_inode = d_inode(root);
1391                 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1392
1393                 ret = 0;
1394                 if (!is_subvolume_inode(root_inode)) {
1395                         btrfs_err(fs_info, "'%s' is not a valid subvolume",
1396                                subvol_name);
1397                         ret = -EINVAL;
1398                 }
1399                 if (subvol_objectid && root_objectid != subvol_objectid) {
1400                         /*
1401                          * This will also catch a race condition where a
1402                          * subvolume which was passed by ID is renamed and
1403                          * another subvolume is renamed over the old location.
1404                          */
1405                         btrfs_err(fs_info,
1406                                   "subvol '%s' does not match subvolid %llu",
1407                                   subvol_name, subvol_objectid);
1408                         ret = -EINVAL;
1409                 }
1410                 if (ret) {
1411                         dput(root);
1412                         root = ERR_PTR(ret);
1413                         deactivate_locked_super(s);
1414                 }
1415         }
1416
1417 out:
1418         mntput(mnt);
1419         kfree(subvol_name);
1420         return root;
1421 }
1422
1423 /*
1424  * Find a superblock for the given device / mount point.
1425  *
1426  * Note: This is based on mount_bdev from fs/super.c with a few additions
1427  *       for multiple device setup.  Make sure to keep it in sync.
1428  */
1429 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1430                 int flags, const char *device_name, void *data)
1431 {
1432         struct block_device *bdev = NULL;
1433         struct super_block *s;
1434         struct btrfs_device *device = NULL;
1435         struct btrfs_fs_devices *fs_devices = NULL;
1436         struct btrfs_fs_info *fs_info = NULL;
1437         void *new_sec_opts = NULL;
1438         fmode_t mode = FMODE_READ;
1439         int error = 0;
1440
1441         if (!(flags & SB_RDONLY))
1442                 mode |= FMODE_WRITE;
1443
1444         if (data) {
1445                 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1446                 if (error)
1447                         return ERR_PTR(error);
1448         }
1449
1450         /*
1451          * Setup a dummy root and fs_info for test/set super.  This is because
1452          * we don't actually fill this stuff out until open_ctree, but we need
1453          * then open_ctree will properly initialize the file system specific
1454          * settings later.  btrfs_init_fs_info initializes the static elements
1455          * of the fs_info (locks and such) to make cleanup easier if we find a
1456          * superblock with our given fs_devices later on at sget() time.
1457          */
1458         fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1459         if (!fs_info) {
1460                 error = -ENOMEM;
1461                 goto error_sec_opts;
1462         }
1463         btrfs_init_fs_info(fs_info);
1464
1465         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1466         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1467         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1468                 error = -ENOMEM;
1469                 goto error_fs_info;
1470         }
1471
1472         mutex_lock(&uuid_mutex);
1473         error = btrfs_parse_device_options(data, mode, fs_type);
1474         if (error) {
1475                 mutex_unlock(&uuid_mutex);
1476                 goto error_fs_info;
1477         }
1478
1479         device = btrfs_scan_one_device(device_name, mode, fs_type);
1480         if (IS_ERR(device)) {
1481                 mutex_unlock(&uuid_mutex);
1482                 error = PTR_ERR(device);
1483                 goto error_fs_info;
1484         }
1485
1486         fs_devices = device->fs_devices;
1487         fs_info->fs_devices = fs_devices;
1488
1489         error = btrfs_open_devices(fs_devices, mode, fs_type);
1490         mutex_unlock(&uuid_mutex);
1491         if (error)
1492                 goto error_fs_info;
1493
1494         if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1495                 error = -EACCES;
1496                 goto error_close_devices;
1497         }
1498
1499         bdev = fs_devices->latest_dev->bdev;
1500         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1501                  fs_info);
1502         if (IS_ERR(s)) {
1503                 error = PTR_ERR(s);
1504                 goto error_close_devices;
1505         }
1506
1507         if (s->s_root) {
1508                 btrfs_close_devices(fs_devices);
1509                 btrfs_free_fs_info(fs_info);
1510                 if ((flags ^ s->s_flags) & SB_RDONLY)
1511                         error = -EBUSY;
1512         } else {
1513                 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1514                 shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
1515                                         s->s_id);
1516                 btrfs_sb(s)->bdev_holder = fs_type;
1517                 if (!strstr(crc32c_impl(), "generic"))
1518                         set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1519                 error = btrfs_fill_super(s, fs_devices, data);
1520         }
1521         if (!error)
1522                 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1523         security_free_mnt_opts(&new_sec_opts);
1524         if (error) {
1525                 deactivate_locked_super(s);
1526                 return ERR_PTR(error);
1527         }
1528
1529         return dget(s->s_root);
1530
1531 error_close_devices:
1532         btrfs_close_devices(fs_devices);
1533 error_fs_info:
1534         btrfs_free_fs_info(fs_info);
1535 error_sec_opts:
1536         security_free_mnt_opts(&new_sec_opts);
1537         return ERR_PTR(error);
1538 }
1539
1540 /*
1541  * Mount function which is called by VFS layer.
1542  *
1543  * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1544  * which needs vfsmount* of device's root (/).  This means device's root has to
1545  * be mounted internally in any case.
1546  *
1547  * Operation flow:
1548  *   1. Parse subvol id related options for later use in mount_subvol().
1549  *
1550  *   2. Mount device's root (/) by calling vfs_kern_mount().
1551  *
1552  *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1553  *      first place. In order to avoid calling btrfs_mount() again, we use
1554  *      different file_system_type which is not registered to VFS by
1555  *      register_filesystem() (btrfs_root_fs_type). As a result,
1556  *      btrfs_mount_root() is called. The return value will be used by
1557  *      mount_subtree() in mount_subvol().
1558  *
1559  *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1560  *      "btrfs subvolume set-default", mount_subvol() is called always.
1561  */
1562 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1563                 const char *device_name, void *data)
1564 {
1565         struct vfsmount *mnt_root;
1566         struct dentry *root;
1567         char *subvol_name = NULL;
1568         u64 subvol_objectid = 0;
1569         int error = 0;
1570
1571         error = btrfs_parse_subvol_options(data, &subvol_name,
1572                                         &subvol_objectid);
1573         if (error) {
1574                 kfree(subvol_name);
1575                 return ERR_PTR(error);
1576         }
1577
1578         /* mount device's root (/) */
1579         mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1580         if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1581                 if (flags & SB_RDONLY) {
1582                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1583                                 flags & ~SB_RDONLY, device_name, data);
1584                 } else {
1585                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1586                                 flags | SB_RDONLY, device_name, data);
1587                         if (IS_ERR(mnt_root)) {
1588                                 root = ERR_CAST(mnt_root);
1589                                 kfree(subvol_name);
1590                                 goto out;
1591                         }
1592
1593                         down_write(&mnt_root->mnt_sb->s_umount);
1594                         error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1595                         up_write(&mnt_root->mnt_sb->s_umount);
1596                         if (error < 0) {
1597                                 root = ERR_PTR(error);
1598                                 mntput(mnt_root);
1599                                 kfree(subvol_name);
1600                                 goto out;
1601                         }
1602                 }
1603         }
1604         if (IS_ERR(mnt_root)) {
1605                 root = ERR_CAST(mnt_root);
1606                 kfree(subvol_name);
1607                 goto out;
1608         }
1609
1610         /* mount_subvol() will free subvol_name and mnt_root */
1611         root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1612
1613 out:
1614         return root;
1615 }
1616
1617 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1618                                      u32 new_pool_size, u32 old_pool_size)
1619 {
1620         if (new_pool_size == old_pool_size)
1621                 return;
1622
1623         fs_info->thread_pool_size = new_pool_size;
1624
1625         btrfs_info(fs_info, "resize thread pool %d -> %d",
1626                old_pool_size, new_pool_size);
1627
1628         btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1629         btrfs_workqueue_set_max(fs_info->hipri_workers, new_pool_size);
1630         btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1631         btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1632         btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1633         btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1634         btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1635 }
1636
1637 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1638                                        unsigned long old_opts, int flags)
1639 {
1640         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1641             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1642              (flags & SB_RDONLY))) {
1643                 /* wait for any defraggers to finish */
1644                 wait_event(fs_info->transaction_wait,
1645                            (atomic_read(&fs_info->defrag_running) == 0));
1646                 if (flags & SB_RDONLY)
1647                         sync_filesystem(fs_info->sb);
1648         }
1649 }
1650
1651 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1652                                          unsigned long old_opts)
1653 {
1654         const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1655
1656         /*
1657          * We need to cleanup all defragable inodes if the autodefragment is
1658          * close or the filesystem is read only.
1659          */
1660         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1661             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1662                 btrfs_cleanup_defrag_inodes(fs_info);
1663         }
1664
1665         /* If we toggled discard async */
1666         if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1667             btrfs_test_opt(fs_info, DISCARD_ASYNC))
1668                 btrfs_discard_resume(fs_info);
1669         else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1670                  !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1671                 btrfs_discard_cleanup(fs_info);
1672
1673         /* If we toggled space cache */
1674         if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1675                 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1676 }
1677
1678 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1679 {
1680         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1681         unsigned old_flags = sb->s_flags;
1682         unsigned long old_opts = fs_info->mount_opt;
1683         unsigned long old_compress_type = fs_info->compress_type;
1684         u64 old_max_inline = fs_info->max_inline;
1685         u32 old_thread_pool_size = fs_info->thread_pool_size;
1686         u32 old_metadata_ratio = fs_info->metadata_ratio;
1687         int ret;
1688
1689         sync_filesystem(sb);
1690         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1691
1692         if (data) {
1693                 void *new_sec_opts = NULL;
1694
1695                 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1696                 if (!ret)
1697                         ret = security_sb_remount(sb, new_sec_opts);
1698                 security_free_mnt_opts(&new_sec_opts);
1699                 if (ret)
1700                         goto restore;
1701         }
1702
1703         ret = btrfs_parse_options(fs_info, data, *flags);
1704         if (ret)
1705                 goto restore;
1706
1707         ret = btrfs_check_features(fs_info, sb);
1708         if (ret < 0)
1709                 goto restore;
1710
1711         btrfs_remount_begin(fs_info, old_opts, *flags);
1712         btrfs_resize_thread_pool(fs_info,
1713                 fs_info->thread_pool_size, old_thread_pool_size);
1714
1715         if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1716             (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1717             (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1718                 btrfs_warn(fs_info,
1719                 "remount supports changing free space tree only from ro to rw");
1720                 /* Make sure free space cache options match the state on disk */
1721                 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1722                         btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1723                         btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1724                 }
1725                 if (btrfs_free_space_cache_v1_active(fs_info)) {
1726                         btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1727                         btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1728                 }
1729         }
1730
1731         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1732                 goto out;
1733
1734         if (*flags & SB_RDONLY) {
1735                 /*
1736                  * this also happens on 'umount -rf' or on shutdown, when
1737                  * the filesystem is busy.
1738                  */
1739                 cancel_work_sync(&fs_info->async_reclaim_work);
1740                 cancel_work_sync(&fs_info->async_data_reclaim_work);
1741
1742                 btrfs_discard_cleanup(fs_info);
1743
1744                 /* wait for the uuid_scan task to finish */
1745                 down(&fs_info->uuid_tree_rescan_sem);
1746                 /* avoid complains from lockdep et al. */
1747                 up(&fs_info->uuid_tree_rescan_sem);
1748
1749                 btrfs_set_sb_rdonly(sb);
1750
1751                 /*
1752                  * Setting SB_RDONLY will put the cleaner thread to
1753                  * sleep at the next loop if it's already active.
1754                  * If it's already asleep, we'll leave unused block
1755                  * groups on disk until we're mounted read-write again
1756                  * unless we clean them up here.
1757                  */
1758                 btrfs_delete_unused_bgs(fs_info);
1759
1760                 /*
1761                  * The cleaner task could be already running before we set the
1762                  * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1763                  * We must make sure that after we finish the remount, i.e. after
1764                  * we call btrfs_commit_super(), the cleaner can no longer start
1765                  * a transaction - either because it was dropping a dead root,
1766                  * running delayed iputs or deleting an unused block group (the
1767                  * cleaner picked a block group from the list of unused block
1768                  * groups before we were able to in the previous call to
1769                  * btrfs_delete_unused_bgs()).
1770                  */
1771                 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1772                             TASK_UNINTERRUPTIBLE);
1773
1774                 /*
1775                  * We've set the superblock to RO mode, so we might have made
1776                  * the cleaner task sleep without running all pending delayed
1777                  * iputs. Go through all the delayed iputs here, so that if an
1778                  * unmount happens without remounting RW we don't end up at
1779                  * finishing close_ctree() with a non-empty list of delayed
1780                  * iputs.
1781                  */
1782                 btrfs_run_delayed_iputs(fs_info);
1783
1784                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1785                 btrfs_scrub_cancel(fs_info);
1786                 btrfs_pause_balance(fs_info);
1787
1788                 /*
1789                  * Pause the qgroup rescan worker if it is running. We don't want
1790                  * it to be still running after we are in RO mode, as after that,
1791                  * by the time we unmount, it might have left a transaction open,
1792                  * so we would leak the transaction and/or crash.
1793                  */
1794                 btrfs_qgroup_wait_for_completion(fs_info, false);
1795
1796                 ret = btrfs_commit_super(fs_info);
1797                 if (ret)
1798                         goto restore;
1799         } else {
1800                 if (BTRFS_FS_ERROR(fs_info)) {
1801                         btrfs_err(fs_info,
1802                                 "Remounting read-write after error is not allowed");
1803                         ret = -EINVAL;
1804                         goto restore;
1805                 }
1806                 if (fs_info->fs_devices->rw_devices == 0) {
1807                         ret = -EACCES;
1808                         goto restore;
1809                 }
1810
1811                 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1812                         btrfs_warn(fs_info,
1813                 "too many missing devices, writable remount is not allowed");
1814                         ret = -EACCES;
1815                         goto restore;
1816                 }
1817
1818                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1819                         btrfs_warn(fs_info,
1820                 "mount required to replay tree-log, cannot remount read-write");
1821                         ret = -EINVAL;
1822                         goto restore;
1823                 }
1824
1825                 /*
1826                  * NOTE: when remounting with a change that does writes, don't
1827                  * put it anywhere above this point, as we are not sure to be
1828                  * safe to write until we pass the above checks.
1829                  */
1830                 ret = btrfs_start_pre_rw_mount(fs_info);
1831                 if (ret)
1832                         goto restore;
1833
1834                 btrfs_clear_sb_rdonly(sb);
1835
1836                 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1837         }
1838 out:
1839         /*
1840          * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1841          * since the absence of the flag means it can be toggled off by remount.
1842          */
1843         *flags |= SB_I_VERSION;
1844
1845         wake_up_process(fs_info->transaction_kthread);
1846         btrfs_remount_cleanup(fs_info, old_opts);
1847         btrfs_clear_oneshot_options(fs_info);
1848         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1849
1850         return 0;
1851
1852 restore:
1853         /* We've hit an error - don't reset SB_RDONLY */
1854         if (sb_rdonly(sb))
1855                 old_flags |= SB_RDONLY;
1856         if (!(old_flags & SB_RDONLY))
1857                 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
1858         sb->s_flags = old_flags;
1859         fs_info->mount_opt = old_opts;
1860         fs_info->compress_type = old_compress_type;
1861         fs_info->max_inline = old_max_inline;
1862         btrfs_resize_thread_pool(fs_info,
1863                 old_thread_pool_size, fs_info->thread_pool_size);
1864         fs_info->metadata_ratio = old_metadata_ratio;
1865         btrfs_remount_cleanup(fs_info, old_opts);
1866         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1867
1868         return ret;
1869 }
1870
1871 /* Used to sort the devices by max_avail(descending sort) */
1872 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1873 {
1874         const struct btrfs_device_info *dev_info1 = a;
1875         const struct btrfs_device_info *dev_info2 = b;
1876
1877         if (dev_info1->max_avail > dev_info2->max_avail)
1878                 return -1;
1879         else if (dev_info1->max_avail < dev_info2->max_avail)
1880                 return 1;
1881         return 0;
1882 }
1883
1884 /*
1885  * sort the devices by max_avail, in which max free extent size of each device
1886  * is stored.(Descending Sort)
1887  */
1888 static inline void btrfs_descending_sort_devices(
1889                                         struct btrfs_device_info *devices,
1890                                         size_t nr_devices)
1891 {
1892         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1893              btrfs_cmp_device_free_bytes, NULL);
1894 }
1895
1896 /*
1897  * The helper to calc the free space on the devices that can be used to store
1898  * file data.
1899  */
1900 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1901                                               u64 *free_bytes)
1902 {
1903         struct btrfs_device_info *devices_info;
1904         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1905         struct btrfs_device *device;
1906         u64 type;
1907         u64 avail_space;
1908         u64 min_stripe_size;
1909         int num_stripes = 1;
1910         int i = 0, nr_devices;
1911         const struct btrfs_raid_attr *rattr;
1912
1913         /*
1914          * We aren't under the device list lock, so this is racy-ish, but good
1915          * enough for our purposes.
1916          */
1917         nr_devices = fs_info->fs_devices->open_devices;
1918         if (!nr_devices) {
1919                 smp_mb();
1920                 nr_devices = fs_info->fs_devices->open_devices;
1921                 ASSERT(nr_devices);
1922                 if (!nr_devices) {
1923                         *free_bytes = 0;
1924                         return 0;
1925                 }
1926         }
1927
1928         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1929                                GFP_KERNEL);
1930         if (!devices_info)
1931                 return -ENOMEM;
1932
1933         /* calc min stripe number for data space allocation */
1934         type = btrfs_data_alloc_profile(fs_info);
1935         rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1936
1937         if (type & BTRFS_BLOCK_GROUP_RAID0)
1938                 num_stripes = nr_devices;
1939         else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1940                 num_stripes = rattr->ncopies;
1941         else if (type & BTRFS_BLOCK_GROUP_RAID10)
1942                 num_stripes = 4;
1943
1944         /* Adjust for more than 1 stripe per device */
1945         min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1946
1947         rcu_read_lock();
1948         list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1949                 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1950                                                 &device->dev_state) ||
1951                     !device->bdev ||
1952                     test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1953                         continue;
1954
1955                 if (i >= nr_devices)
1956                         break;
1957
1958                 avail_space = device->total_bytes - device->bytes_used;
1959
1960                 /* align with stripe_len */
1961                 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1962
1963                 /*
1964                  * Ensure we have at least min_stripe_size on top of the
1965                  * reserved space on the device.
1966                  */
1967                 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1968                         continue;
1969
1970                 avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1971
1972                 devices_info[i].dev = device;
1973                 devices_info[i].max_avail = avail_space;
1974
1975                 i++;
1976         }
1977         rcu_read_unlock();
1978
1979         nr_devices = i;
1980
1981         btrfs_descending_sort_devices(devices_info, nr_devices);
1982
1983         i = nr_devices - 1;
1984         avail_space = 0;
1985         while (nr_devices >= rattr->devs_min) {
1986                 num_stripes = min(num_stripes, nr_devices);
1987
1988                 if (devices_info[i].max_avail >= min_stripe_size) {
1989                         int j;
1990                         u64 alloc_size;
1991
1992                         avail_space += devices_info[i].max_avail * num_stripes;
1993                         alloc_size = devices_info[i].max_avail;
1994                         for (j = i + 1 - num_stripes; j <= i; j++)
1995                                 devices_info[j].max_avail -= alloc_size;
1996                 }
1997                 i--;
1998                 nr_devices--;
1999         }
2000
2001         kfree(devices_info);
2002         *free_bytes = avail_space;
2003         return 0;
2004 }
2005
2006 /*
2007  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2008  *
2009  * If there's a redundant raid level at DATA block groups, use the respective
2010  * multiplier to scale the sizes.
2011  *
2012  * Unused device space usage is based on simulating the chunk allocator
2013  * algorithm that respects the device sizes and order of allocations.  This is
2014  * a close approximation of the actual use but there are other factors that may
2015  * change the result (like a new metadata chunk).
2016  *
2017  * If metadata is exhausted, f_bavail will be 0.
2018  */
2019 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2020 {
2021         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2022         struct btrfs_super_block *disk_super = fs_info->super_copy;
2023         struct btrfs_space_info *found;
2024         u64 total_used = 0;
2025         u64 total_free_data = 0;
2026         u64 total_free_meta = 0;
2027         u32 bits = fs_info->sectorsize_bits;
2028         __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2029         unsigned factor = 1;
2030         struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2031         int ret;
2032         u64 thresh = 0;
2033         int mixed = 0;
2034
2035         list_for_each_entry(found, &fs_info->space_info, list) {
2036                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2037                         int i;
2038
2039                         total_free_data += found->disk_total - found->disk_used;
2040                         total_free_data -=
2041                                 btrfs_account_ro_block_groups_free_space(found);
2042
2043                         for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2044                                 if (!list_empty(&found->block_groups[i]))
2045                                         factor = btrfs_bg_type_to_factor(
2046                                                 btrfs_raid_array[i].bg_flag);
2047                         }
2048                 }
2049
2050                 /*
2051                  * Metadata in mixed block goup profiles are accounted in data
2052                  */
2053                 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2054                         if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2055                                 mixed = 1;
2056                         else
2057                                 total_free_meta += found->disk_total -
2058                                         found->disk_used;
2059                 }
2060
2061                 total_used += found->disk_used;
2062         }
2063
2064         buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2065         buf->f_blocks >>= bits;
2066         buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2067
2068         /* Account global block reserve as used, it's in logical size already */
2069         spin_lock(&block_rsv->lock);
2070         /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2071         if (buf->f_bfree >= block_rsv->size >> bits)
2072                 buf->f_bfree -= block_rsv->size >> bits;
2073         else
2074                 buf->f_bfree = 0;
2075         spin_unlock(&block_rsv->lock);
2076
2077         buf->f_bavail = div_u64(total_free_data, factor);
2078         ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2079         if (ret)
2080                 return ret;
2081         buf->f_bavail += div_u64(total_free_data, factor);
2082         buf->f_bavail = buf->f_bavail >> bits;
2083
2084         /*
2085          * We calculate the remaining metadata space minus global reserve. If
2086          * this is (supposedly) smaller than zero, there's no space. But this
2087          * does not hold in practice, the exhausted state happens where's still
2088          * some positive delta. So we apply some guesswork and compare the
2089          * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2090          *
2091          * We probably cannot calculate the exact threshold value because this
2092          * depends on the internal reservations requested by various
2093          * operations, so some operations that consume a few metadata will
2094          * succeed even if the Avail is zero. But this is better than the other
2095          * way around.
2096          */
2097         thresh = SZ_4M;
2098
2099         /*
2100          * We only want to claim there's no available space if we can no longer
2101          * allocate chunks for our metadata profile and our global reserve will
2102          * not fit in the free metadata space.  If we aren't ->full then we
2103          * still can allocate chunks and thus are fine using the currently
2104          * calculated f_bavail.
2105          */
2106         if (!mixed && block_rsv->space_info->full &&
2107             total_free_meta - thresh < block_rsv->size)
2108                 buf->f_bavail = 0;
2109
2110         buf->f_type = BTRFS_SUPER_MAGIC;
2111         buf->f_bsize = dentry->d_sb->s_blocksize;
2112         buf->f_namelen = BTRFS_NAME_LEN;
2113
2114         /* We treat it as constant endianness (it doesn't matter _which_)
2115            because we want the fsid to come out the same whether mounted
2116            on a big-endian or little-endian host */
2117         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2118         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2119         /* Mask in the root object ID too, to disambiguate subvols */
2120         buf->f_fsid.val[0] ^=
2121                 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2122         buf->f_fsid.val[1] ^=
2123                 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2124
2125         return 0;
2126 }
2127
2128 static void btrfs_kill_super(struct super_block *sb)
2129 {
2130         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2131         kill_anon_super(sb);
2132         btrfs_free_fs_info(fs_info);
2133 }
2134
2135 static struct file_system_type btrfs_fs_type = {
2136         .owner          = THIS_MODULE,
2137         .name           = "btrfs",
2138         .mount          = btrfs_mount,
2139         .kill_sb        = btrfs_kill_super,
2140         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2141 };
2142
2143 static struct file_system_type btrfs_root_fs_type = {
2144         .owner          = THIS_MODULE,
2145         .name           = "btrfs",
2146         .mount          = btrfs_mount_root,
2147         .kill_sb        = btrfs_kill_super,
2148         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2149 };
2150
2151 MODULE_ALIAS_FS("btrfs");
2152
2153 static int btrfs_control_open(struct inode *inode, struct file *file)
2154 {
2155         /*
2156          * The control file's private_data is used to hold the
2157          * transaction when it is started and is used to keep
2158          * track of whether a transaction is already in progress.
2159          */
2160         file->private_data = NULL;
2161         return 0;
2162 }
2163
2164 /*
2165  * Used by /dev/btrfs-control for devices ioctls.
2166  */
2167 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2168                                 unsigned long arg)
2169 {
2170         struct btrfs_ioctl_vol_args *vol;
2171         struct btrfs_device *device = NULL;
2172         dev_t devt = 0;
2173         int ret = -ENOTTY;
2174
2175         if (!capable(CAP_SYS_ADMIN))
2176                 return -EPERM;
2177
2178         vol = memdup_user((void __user *)arg, sizeof(*vol));
2179         if (IS_ERR(vol))
2180                 return PTR_ERR(vol);
2181         vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2182
2183         switch (cmd) {
2184         case BTRFS_IOC_SCAN_DEV:
2185                 mutex_lock(&uuid_mutex);
2186                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2187                                                &btrfs_root_fs_type);
2188                 ret = PTR_ERR_OR_ZERO(device);
2189                 mutex_unlock(&uuid_mutex);
2190                 break;
2191         case BTRFS_IOC_FORGET_DEV:
2192                 if (vol->name[0] != 0) {
2193                         ret = lookup_bdev(vol->name, &devt);
2194                         if (ret)
2195                                 break;
2196                 }
2197                 ret = btrfs_forget_devices(devt);
2198                 break;
2199         case BTRFS_IOC_DEVICES_READY:
2200                 mutex_lock(&uuid_mutex);
2201                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2202                                                &btrfs_root_fs_type);
2203                 if (IS_ERR(device)) {
2204                         mutex_unlock(&uuid_mutex);
2205                         ret = PTR_ERR(device);
2206                         break;
2207                 }
2208                 ret = !(device->fs_devices->num_devices ==
2209                         device->fs_devices->total_devices);
2210                 mutex_unlock(&uuid_mutex);
2211                 break;
2212         case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2213                 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2214                 break;
2215         }
2216
2217         kfree(vol);
2218         return ret;
2219 }
2220
2221 static int btrfs_freeze(struct super_block *sb)
2222 {
2223         struct btrfs_trans_handle *trans;
2224         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2225         struct btrfs_root *root = fs_info->tree_root;
2226
2227         set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2228         /*
2229          * We don't need a barrier here, we'll wait for any transaction that
2230          * could be in progress on other threads (and do delayed iputs that
2231          * we want to avoid on a frozen filesystem), or do the commit
2232          * ourselves.
2233          */
2234         trans = btrfs_attach_transaction_barrier(root);
2235         if (IS_ERR(trans)) {
2236                 /* no transaction, don't bother */
2237                 if (PTR_ERR(trans) == -ENOENT)
2238                         return 0;
2239                 return PTR_ERR(trans);
2240         }
2241         return btrfs_commit_transaction(trans);
2242 }
2243
2244 static int check_dev_super(struct btrfs_device *dev)
2245 {
2246         struct btrfs_fs_info *fs_info = dev->fs_info;
2247         struct btrfs_super_block *sb;
2248         u16 csum_type;
2249         int ret = 0;
2250
2251         /* This should be called with fs still frozen. */
2252         ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2253
2254         /* Missing dev, no need to check. */
2255         if (!dev->bdev)
2256                 return 0;
2257
2258         /* Only need to check the primary super block. */
2259         sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2260         if (IS_ERR(sb))
2261                 return PTR_ERR(sb);
2262
2263         /* Verify the checksum. */
2264         csum_type = btrfs_super_csum_type(sb);
2265         if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2266                 btrfs_err(fs_info, "csum type changed, has %u expect %u",
2267                           csum_type, btrfs_super_csum_type(fs_info->super_copy));
2268                 ret = -EUCLEAN;
2269                 goto out;
2270         }
2271
2272         if (btrfs_check_super_csum(fs_info, sb)) {
2273                 btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2274                 ret = -EUCLEAN;
2275                 goto out;
2276         }
2277
2278         /* Btrfs_validate_super() includes fsid check against super->fsid. */
2279         ret = btrfs_validate_super(fs_info, sb, 0);
2280         if (ret < 0)
2281                 goto out;
2282
2283         if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
2284                 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2285                         btrfs_super_generation(sb),
2286                         fs_info->last_trans_committed);
2287                 ret = -EUCLEAN;
2288                 goto out;
2289         }
2290 out:
2291         btrfs_release_disk_super(sb);
2292         return ret;
2293 }
2294
2295 static int btrfs_unfreeze(struct super_block *sb)
2296 {
2297         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2298         struct btrfs_device *device;
2299         int ret = 0;
2300
2301         /*
2302          * Make sure the fs is not changed by accident (like hibernation then
2303          * modified by other OS).
2304          * If we found anything wrong, we mark the fs error immediately.
2305          *
2306          * And since the fs is frozen, no one can modify the fs yet, thus
2307          * we don't need to hold device_list_mutex.
2308          */
2309         list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2310                 ret = check_dev_super(device);
2311                 if (ret < 0) {
2312                         btrfs_handle_fs_error(fs_info, ret,
2313                                 "super block on devid %llu got modified unexpectedly",
2314                                 device->devid);
2315                         break;
2316                 }
2317         }
2318         clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2319
2320         /*
2321          * We still return 0, to allow VFS layer to unfreeze the fs even the
2322          * above checks failed. Since the fs is either fine or read-only, we're
2323          * safe to continue, without causing further damage.
2324          */
2325         return 0;
2326 }
2327
2328 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2329 {
2330         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2331
2332         /*
2333          * There should be always a valid pointer in latest_dev, it may be stale
2334          * for a short moment in case it's being deleted but still valid until
2335          * the end of RCU grace period.
2336          */
2337         rcu_read_lock();
2338         seq_escape(m, rcu_str_deref(fs_info->fs_devices->latest_dev->name), " \t\n\\");
2339         rcu_read_unlock();
2340
2341         return 0;
2342 }
2343
2344 static const struct super_operations btrfs_super_ops = {
2345         .drop_inode     = btrfs_drop_inode,
2346         .evict_inode    = btrfs_evict_inode,
2347         .put_super      = btrfs_put_super,
2348         .sync_fs        = btrfs_sync_fs,
2349         .show_options   = btrfs_show_options,
2350         .show_devname   = btrfs_show_devname,
2351         .alloc_inode    = btrfs_alloc_inode,
2352         .destroy_inode  = btrfs_destroy_inode,
2353         .free_inode     = btrfs_free_inode,
2354         .statfs         = btrfs_statfs,
2355         .remount_fs     = btrfs_remount,
2356         .freeze_fs      = btrfs_freeze,
2357         .unfreeze_fs    = btrfs_unfreeze,
2358 };
2359
2360 static const struct file_operations btrfs_ctl_fops = {
2361         .open = btrfs_control_open,
2362         .unlocked_ioctl  = btrfs_control_ioctl,
2363         .compat_ioctl = compat_ptr_ioctl,
2364         .owner   = THIS_MODULE,
2365         .llseek = noop_llseek,
2366 };
2367
2368 static struct miscdevice btrfs_misc = {
2369         .minor          = BTRFS_MINOR,
2370         .name           = "btrfs-control",
2371         .fops           = &btrfs_ctl_fops
2372 };
2373
2374 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2375 MODULE_ALIAS("devname:btrfs-control");
2376
2377 static int __init btrfs_interface_init(void)
2378 {
2379         return misc_register(&btrfs_misc);
2380 }
2381
2382 static __cold void btrfs_interface_exit(void)
2383 {
2384         misc_deregister(&btrfs_misc);
2385 }
2386
2387 static int __init btrfs_print_mod_info(void)
2388 {
2389         static const char options[] = ""
2390 #ifdef CONFIG_BTRFS_DEBUG
2391                         ", debug=on"
2392 #endif
2393 #ifdef CONFIG_BTRFS_ASSERT
2394                         ", assert=on"
2395 #endif
2396 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2397                         ", integrity-checker=on"
2398 #endif
2399 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2400                         ", ref-verify=on"
2401 #endif
2402 #ifdef CONFIG_BLK_DEV_ZONED
2403                         ", zoned=yes"
2404 #else
2405                         ", zoned=no"
2406 #endif
2407 #ifdef CONFIG_FS_VERITY
2408                         ", fsverity=yes"
2409 #else
2410                         ", fsverity=no"
2411 #endif
2412                         ;
2413         pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2414         return 0;
2415 }
2416
2417 static int register_btrfs(void)
2418 {
2419         return register_filesystem(&btrfs_fs_type);
2420 }
2421
2422 static void unregister_btrfs(void)
2423 {
2424         unregister_filesystem(&btrfs_fs_type);
2425 }
2426
2427 /* Helper structure for long init/exit functions. */
2428 struct init_sequence {
2429         int (*init_func)(void);
2430         /* Can be NULL if the init_func doesn't need cleanup. */
2431         void (*exit_func)(void);
2432 };
2433
2434 static const struct init_sequence mod_init_seq[] = {
2435         {
2436                 .init_func = btrfs_props_init,
2437                 .exit_func = NULL,
2438         }, {
2439                 .init_func = btrfs_init_sysfs,
2440                 .exit_func = btrfs_exit_sysfs,
2441         }, {
2442                 .init_func = btrfs_init_compress,
2443                 .exit_func = btrfs_exit_compress,
2444         }, {
2445                 .init_func = btrfs_init_cachep,
2446                 .exit_func = btrfs_destroy_cachep,
2447         }, {
2448                 .init_func = btrfs_transaction_init,
2449                 .exit_func = btrfs_transaction_exit,
2450         }, {
2451                 .init_func = btrfs_ctree_init,
2452                 .exit_func = btrfs_ctree_exit,
2453         }, {
2454                 .init_func = btrfs_free_space_init,
2455                 .exit_func = btrfs_free_space_exit,
2456         }, {
2457                 .init_func = extent_state_init_cachep,
2458                 .exit_func = extent_state_free_cachep,
2459         }, {
2460                 .init_func = extent_buffer_init_cachep,
2461                 .exit_func = extent_buffer_free_cachep,
2462         }, {
2463                 .init_func = btrfs_bioset_init,
2464                 .exit_func = btrfs_bioset_exit,
2465         }, {
2466                 .init_func = extent_map_init,
2467                 .exit_func = extent_map_exit,
2468         }, {
2469                 .init_func = ordered_data_init,
2470                 .exit_func = ordered_data_exit,
2471         }, {
2472                 .init_func = btrfs_delayed_inode_init,
2473                 .exit_func = btrfs_delayed_inode_exit,
2474         }, {
2475                 .init_func = btrfs_auto_defrag_init,
2476                 .exit_func = btrfs_auto_defrag_exit,
2477         }, {
2478                 .init_func = btrfs_delayed_ref_init,
2479                 .exit_func = btrfs_delayed_ref_exit,
2480         }, {
2481                 .init_func = btrfs_prelim_ref_init,
2482                 .exit_func = btrfs_prelim_ref_exit,
2483         }, {
2484                 .init_func = btrfs_interface_init,
2485                 .exit_func = btrfs_interface_exit,
2486         }, {
2487                 .init_func = btrfs_print_mod_info,
2488                 .exit_func = NULL,
2489         }, {
2490                 .init_func = btrfs_run_sanity_tests,
2491                 .exit_func = NULL,
2492         }, {
2493                 .init_func = register_btrfs,
2494                 .exit_func = unregister_btrfs,
2495         }
2496 };
2497
2498 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2499
2500 static __always_inline void btrfs_exit_btrfs_fs(void)
2501 {
2502         int i;
2503
2504         for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2505                 if (!mod_init_result[i])
2506                         continue;
2507                 if (mod_init_seq[i].exit_func)
2508                         mod_init_seq[i].exit_func();
2509                 mod_init_result[i] = false;
2510         }
2511 }
2512
2513 static void __exit exit_btrfs_fs(void)
2514 {
2515         btrfs_exit_btrfs_fs();
2516 }
2517
2518 static int __init init_btrfs_fs(void)
2519 {
2520         int ret;
2521         int i;
2522
2523         for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2524                 ASSERT(!mod_init_result[i]);
2525                 ret = mod_init_seq[i].init_func();
2526                 if (ret < 0) {
2527                         btrfs_exit_btrfs_fs();
2528                         return ret;
2529                 }
2530                 mod_init_result[i] = true;
2531         }
2532         return 0;
2533 }
2534
2535 late_initcall(init_btrfs_fs);
2536 module_exit(exit_btrfs_fs)
2537
2538 MODULE_LICENSE("GPL");
2539 MODULE_SOFTDEP("pre: crc32c");
2540 MODULE_SOFTDEP("pre: xxhash64");
2541 MODULE_SOFTDEP("pre: sha256");
2542 MODULE_SOFTDEP("pre: blake2b-256");