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/cleancache.h>
  27 #include <linux/ratelimit.h>
  28 #include <linux/crc32c.h>
  29 #include <linux/btrfs.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 "tests/btrfs-tests.h"
  48 #include "block-group.h"
  49 #include "discard.h"
  50
  51 #include "qgroup.h"
  52 #define CREATE_TRACE_POINTS
  53 #include <trace/events/btrfs.h>
  54
  55 static const struct super_operations btrfs_super_ops;
  56
  57 /*
  58  * Types for mounting the default subvolume and a subvolume explicitly
  59  * requested by subvol=/path. That way the callchain is straightforward and we
  60  * don't have to play tricks with the mount options and recursive calls to
  61  * btrfs_mount.
  62  *
  63  * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
  64  */
  65 static struct file_system_type btrfs_fs_type;
  66 static struct file_system_type btrfs_root_fs_type;
  67
  68 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
  69
  70 /*
  71  * Generally the error codes correspond to their respective errors, but there
  72  * are a few special cases.
  73  *
  74  * EUCLEAN: Any sort of corruption that we encounter.  The tree-checker for
  75  *          instance will return EUCLEAN if any of the blocks are corrupted in
  76  *          a way that is problematic.  We want to reserve EUCLEAN for these
  77  *          sort of corruptions.
  78  *
  79  * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
  80  *        need to use EROFS for this case.  We will have no idea of the
  81  *        original failure, that will have been reported at the time we tripped
  82  *        over the error.  Each subsequent error that doesn't have any context
  83  *        of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
  84  */
  85 const char * __attribute_const__ btrfs_decode_error(int errno)
  86 {
  87         char *errstr = "unknown";
  88
  89         switch (errno) {
  90         case -ENOENT:           /* -2 */
  91                 errstr = "No such entry";
  92                 break;
  93         case -EIO:              /* -5 */
  94                 errstr = "IO failure";
  95                 break;
  96         case -ENOMEM:           /* -12*/
  97                 errstr = "Out of memory";
  98                 break;
  99         case -EEXIST:           /* -17 */
 100                 errstr = "Object already exists";
 101                 break;
 102         case -ENOSPC:           /* -28 */
 103                 errstr = "No space left";
 104                 break;
 105         case -EROFS:            /* -30 */
 106                 errstr = "Readonly filesystem";
 107                 break;
 108         case -EOPNOTSUPP:       /* -95 */
 109                 errstr = "Operation not supported";
 110                 break;
 111         case -EUCLEAN:          /* -117 */
 112                 errstr = "Filesystem corrupted";
 113                 break;
 114         case -EDQUOT:           /* -122 */
 115                 errstr = "Quota exceeded";
 116                 break;
 117         }
 118
 119         return errstr;
 120 }
 121
 122 /*
 123  * __btrfs_handle_fs_error decodes expected errors from the caller and
 124  * invokes the appropriate error response.
 125  */
 126 __cold
 127 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
 128                        unsigned int line, int errno, const char *fmt, ...)
 129 {
 130         struct super_block *sb = fs_info->sb;
 131 #ifdef CONFIG_PRINTK
 132         const char *errstr;
 133 #endif
 134
 135         /*
 136          * Special case: if the error is EROFS, and we're already
 137          * under SB_RDONLY, then it is safe here.
 138          */
 139         if (errno == -EROFS && sb_rdonly(sb))
 140                 return;
 141
 142 #ifdef CONFIG_PRINTK
 143         errstr = btrfs_decode_error(errno);
 144         if (fmt) {
 145                 struct va_format vaf;
 146                 va_list args;
 147
 148                 va_start(args, fmt);
 149                 vaf.fmt = fmt;
 150                 vaf.va = &args;
 151
 152                 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
 153                         sb->s_id, function, line, errno, errstr, &vaf);
 154                 va_end(args);
 155         } else {
 156                 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
 157                         sb->s_id, function, line, errno, errstr);
 158         }
 159 #endif
 160
 161         /*
 162          * Today we only save the error info to memory.  Long term we'll
 163          * also send it down to the disk
 164          */
 165         set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
 166
 167         /* Don't go through full error handling during mount */
 168         if (!(sb->s_flags & SB_BORN))
 169                 return;
 170
 171         if (sb_rdonly(sb))
 172                 return;
 173
 174         btrfs_discard_stop(fs_info);
 175
 176         /* btrfs handle error by forcing the filesystem readonly */
 177         sb->s_flags |= SB_RDONLY;
 178         btrfs_info(fs_info, "forced readonly");
 179         /*
 180          * Note that a running device replace operation is not canceled here
 181          * although there is no way to update the progress. It would add the
 182          * risk of a deadlock, therefore the canceling is omitted. The only
 183          * penalty is that some I/O remains active until the procedure
 184          * completes. The next time when the filesystem is mounted writable
 185          * again, the device replace operation continues.
 186          */
 187 }
 188
 189 #ifdef CONFIG_PRINTK
 190 static const char * const logtypes[] = {
 191         "emergency",
 192         "alert",
 193         "critical",
 194         "error",
 195         "warning",
 196         "notice",
 197         "info",
 198         "debug",
 199 };
 200
 201
 202 /*
 203  * Use one ratelimit state per log level so that a flood of less important
 204  * messages doesn't cause more important ones to be dropped.
 205  */
 206 static struct ratelimit_state printk_limits[] = {
 207         RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
 208         RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
 209         RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
 210         RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
 211         RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
 212         RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
 213         RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
 214         RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
 215 };
 216
 217 void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 218 {
 219         char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 220         struct va_format vaf;
 221         va_list args;
 222         int kern_level;
 223         const char *type = logtypes[4];
 224         struct ratelimit_state *ratelimit = &printk_limits[4];
 225
 226         va_start(args, fmt);
 227
 228         while ((kern_level = printk_get_level(fmt)) != 0) {
 229                 size_t size = printk_skip_level(fmt) - fmt;
 230
 231                 if (kern_level >= '0' && kern_level <= '7') {
 232                         memcpy(lvl, fmt,  size);
 233                         lvl[size] = '\0';
 234                         type = logtypes[kern_level - '0'];
 235                         ratelimit = &printk_limits[kern_level - '0'];
 236                 }
 237                 fmt += size;
 238         }
 239
 240         vaf.fmt = fmt;
 241         vaf.va = &args;
 242
 243         if (__ratelimit(ratelimit))
 244                 printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
 245                         fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
 246
 247         va_end(args);
 248 }
 249 #endif
 250
 251 /*
 252  * We only mark the transaction aborted and then set the file system read-only.
 253  * This will prevent new transactions from starting or trying to join this
 254  * one.
 255  *
 256  * This means that error recovery at the call site is limited to freeing
 257  * any local memory allocations and passing the error code up without
 258  * further cleanup. The transaction should complete as it normally would
 259  * in the call path but will return -EIO.
 260  *
 261  * We'll complete the cleanup in btrfs_end_transaction and
 262  * btrfs_commit_transaction.
 263  */
 264 __cold
 265 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 266                                const char *function,
 267                                unsigned int line, int errno)
 268 {
 269         struct btrfs_fs_info *fs_info = trans->fs_info;
 270
 271         WRITE_ONCE(trans->aborted, errno);
 272         /* Nothing used. The other threads that have joined this
 273          * transaction may be able to continue. */
 274         if (!trans->dirty && list_empty(&trans->new_bgs)) {
 275                 const char *errstr;
 276
 277                 errstr = btrfs_decode_error(errno);
 278                 btrfs_warn(fs_info,
 279                            "%s:%d: Aborting unused transaction(%s).",
 280                            function, line, errstr);
 281                 return;
 282         }
 283         WRITE_ONCE(trans->transaction->aborted, errno);
 284         /* Wake up anybody who may be waiting on this transaction */
 285         wake_up(&fs_info->transaction_wait);
 286         wake_up(&fs_info->transaction_blocked_wait);
 287         __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
 288 }
 289 /*
 290  * __btrfs_panic decodes unexpected, fatal errors from the caller,
 291  * issues an alert, and either panics or BUGs, depending on mount options.
 292  */
 293 __cold
 294 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 295                    unsigned int line, int errno, const char *fmt, ...)
 296 {
 297         char *s_id = "<unknown>";
 298         const char *errstr;
 299         struct va_format vaf = { .fmt = fmt };
 300         va_list args;
 301
 302         if (fs_info)
 303                 s_id = fs_info->sb->s_id;
 304
 305         va_start(args, fmt);
 306         vaf.va = &args;
 307
 308         errstr = btrfs_decode_error(errno);
 309         if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
 310                 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
 311                         s_id, function, line, &vaf, errno, errstr);
 312
 313         btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
 314                    function, line, &vaf, errno, errstr);
 315         va_end(args);
 316         /* Caller calls BUG() */
 317 }
 318
 319 static void btrfs_put_super(struct super_block *sb)
 320 {
 321         close_ctree(btrfs_sb(sb));
 322 }
 323
 324 enum {
 325         Opt_acl, Opt_noacl,
 326         Opt_clear_cache,
 327         Opt_commit_interval,
 328         Opt_compress,
 329         Opt_compress_force,
 330         Opt_compress_force_type,
 331         Opt_compress_type,
 332         Opt_degraded,
 333         Opt_device,
 334         Opt_fatal_errors,
 335         Opt_flushoncommit, Opt_noflushoncommit,
 336         Opt_inode_cache, Opt_noinode_cache,
 337         Opt_max_inline,
 338         Opt_barrier, Opt_nobarrier,
 339         Opt_datacow, Opt_nodatacow,
 340         Opt_datasum, Opt_nodatasum,
 341         Opt_defrag, Opt_nodefrag,
 342         Opt_discard, Opt_nodiscard,
 343         Opt_discard_mode,
 344         Opt_norecovery,
 345         Opt_ratio,
 346         Opt_rescan_uuid_tree,
 347         Opt_skip_balance,
 348         Opt_space_cache, Opt_no_space_cache,
 349         Opt_space_cache_version,
 350         Opt_ssd, Opt_nossd,
 351         Opt_ssd_spread, Opt_nossd_spread,
 352         Opt_subvol,
 353         Opt_subvol_empty,
 354         Opt_subvolid,
 355         Opt_thread_pool,
 356         Opt_treelog, Opt_notreelog,
 357         Opt_user_subvol_rm_allowed,
 358
 359         /* Rescue options */
 360         Opt_rescue,
 361         Opt_usebackuproot,
 362         Opt_nologreplay,
 363
 364         /* Deprecated options */
 365         Opt_recovery,
 366
 367         /* Debugging options */
 368         Opt_check_integrity,
 369         Opt_check_integrity_including_extent_data,
 370         Opt_check_integrity_print_mask,
 371         Opt_enospc_debug, Opt_noenospc_debug,
 372 #ifdef CONFIG_BTRFS_DEBUG
 373         Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 374 #endif
 375 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 376         Opt_ref_verify,
 377 #endif
 378         Opt_err,
 379 };
 380
 381 static const match_table_t tokens = {
 382         {Opt_acl, "acl"},
 383         {Opt_noacl, "noacl"},
 384         {Opt_clear_cache, "clear_cache"},
 385         {Opt_commit_interval, "commit=%u"},
 386         {Opt_compress, "compress"},
 387         {Opt_compress_type, "compress=%s"},
 388         {Opt_compress_force, "compress-force"},
 389         {Opt_compress_force_type, "compress-force=%s"},
 390         {Opt_degraded, "degraded"},
 391         {Opt_device, "device=%s"},
 392         {Opt_fatal_errors, "fatal_errors=%s"},
 393         {Opt_flushoncommit, "flushoncommit"},
 394         {Opt_noflushoncommit, "noflushoncommit"},
 395         {Opt_inode_cache, "inode_cache"},
 396         {Opt_noinode_cache, "noinode_cache"},
 397         {Opt_max_inline, "max_inline=%s"},
 398         {Opt_barrier, "barrier"},
 399         {Opt_nobarrier, "nobarrier"},
 400         {Opt_datacow, "datacow"},
 401         {Opt_nodatacow, "nodatacow"},
 402         {Opt_datasum, "datasum"},
 403         {Opt_nodatasum, "nodatasum"},
 404         {Opt_defrag, "autodefrag"},
 405         {Opt_nodefrag, "noautodefrag"},
 406         {Opt_discard, "discard"},
 407         {Opt_discard_mode, "discard=%s"},
 408         {Opt_nodiscard, "nodiscard"},
 409         {Opt_norecovery, "norecovery"},
 410         {Opt_ratio, "metadata_ratio=%u"},
 411         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 412         {Opt_skip_balance, "skip_balance"},
 413         {Opt_space_cache, "space_cache"},
 414         {Opt_no_space_cache, "nospace_cache"},
 415         {Opt_space_cache_version, "space_cache=%s"},
 416         {Opt_ssd, "ssd"},
 417         {Opt_nossd, "nossd"},
 418         {Opt_ssd_spread, "ssd_spread"},
 419         {Opt_nossd_spread, "nossd_spread"},
 420         {Opt_subvol, "subvol=%s"},
 421         {Opt_subvol_empty, "subvol="},
 422         {Opt_subvolid, "subvolid=%s"},
 423         {Opt_thread_pool, "thread_pool=%u"},
 424         {Opt_treelog, "treelog"},
 425         {Opt_notreelog, "notreelog"},
 426         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 427
 428         /* Rescue options */
 429         {Opt_rescue, "rescue=%s"},
 430         /* Deprecated, with alias rescue=nologreplay */
 431         {Opt_nologreplay, "nologreplay"},
 432         /* Deprecated, with alias rescue=usebackuproot */
 433         {Opt_usebackuproot, "usebackuproot"},
 434
 435         /* Deprecated options */
 436         {Opt_recovery, "recovery"},
 437
 438         /* Debugging options */
 439         {Opt_check_integrity, "check_int"},
 440         {Opt_check_integrity_including_extent_data, "check_int_data"},
 441         {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
 442         {Opt_enospc_debug, "enospc_debug"},
 443         {Opt_noenospc_debug, "noenospc_debug"},
 444 #ifdef CONFIG_BTRFS_DEBUG
 445         {Opt_fragment_data, "fragment=data"},
 446         {Opt_fragment_metadata, "fragment=metadata"},
 447         {Opt_fragment_all, "fragment=all"},
 448 #endif
 449 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 450         {Opt_ref_verify, "ref_verify"},
 451 #endif
 452         {Opt_err, NULL},
 453 };
 454
 455 static const match_table_t rescue_tokens = {
 456         {Opt_usebackuproot, "usebackuproot"},
 457         {Opt_nologreplay, "nologreplay"},
 458         {Opt_err, NULL},
 459 };
 460
 461 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
 462 {
 463         char *opts;
 464         char *orig;
 465         char *p;
 466         substring_t args[MAX_OPT_ARGS];
 467         int ret = 0;
 468
 469         opts = kstrdup(options, GFP_KERNEL);
 470         if (!opts)
 471                 return -ENOMEM;
 472         orig = opts;
 473
 474         while ((p = strsep(&opts, ":")) != NULL) {
 475                 int token;
 476
 477                 if (!*p)
 478                         continue;
 479                 token = match_token(p, rescue_tokens, args);
 480                 switch (token){
 481                 case Opt_usebackuproot:
 482                         btrfs_info(info,
 483                                    "trying to use backup root at mount time");
 484                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 485                         break;
 486                 case Opt_nologreplay:
 487                         btrfs_set_and_info(info, NOLOGREPLAY,
 488                                            "disabling log replay at mount time");
 489                         break;
 490                 case Opt_err:
 491                         btrfs_info(info, "unrecognized rescue option '%s'", p);
 492                         ret = -EINVAL;
 493                         goto out;
 494                 default:
 495                         break;
 496                 }
 497
 498         }
 499 out:
 500         kfree(orig);
 501         return ret;
 502 }
 503
 504 /*
 505  * Regular mount options parser.  Everything that is needed only when
 506  * reading in a new superblock is parsed here.
 507  * XXX JDM: This needs to be cleaned up for remount.
 508  */
 509 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 510                         unsigned long new_flags)
 511 {
 512         substring_t args[MAX_OPT_ARGS];
 513         char *p, *num;
 514         u64 cache_gen;
 515         int intarg;
 516         int ret = 0;
 517         char *compress_type;
 518         bool compress_force = false;
 519         enum btrfs_compression_type saved_compress_type;
 520         int saved_compress_level;
 521         bool saved_compress_force;
 522         int no_compress = 0;
 523
 524         cache_gen = btrfs_super_cache_generation(info->super_copy);
 525         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
 526                 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
 527         else if (cache_gen)
 528                 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 529
 530         /*
 531          * Even the options are empty, we still need to do extra check
 532          * against new flags
 533          */
 534         if (!options)
 535                 goto check;
 536
 537         while ((p = strsep(&options, ",")) != NULL) {
 538                 int token;
 539                 if (!*p)
 540                         continue;
 541
 542                 token = match_token(p, tokens, args);
 543                 switch (token) {
 544                 case Opt_degraded:
 545                         btrfs_info(info, "allowing degraded mounts");
 546                         btrfs_set_opt(info->mount_opt, DEGRADED);
 547                         break;
 548                 case Opt_subvol:
 549                 case Opt_subvol_empty:
 550                 case Opt_subvolid:
 551                 case Opt_device:
 552                         /*
 553                          * These are parsed by btrfs_parse_subvol_options or
 554                          * btrfs_parse_device_options and can be ignored here.
 555                          */
 556                         break;
 557                 case Opt_nodatasum:
 558                         btrfs_set_and_info(info, NODATASUM,
 559                                            "setting nodatasum");
 560                         break;
 561                 case Opt_datasum:
 562                         if (btrfs_test_opt(info, NODATASUM)) {
 563                                 if (btrfs_test_opt(info, NODATACOW))
 564                                         btrfs_info(info,
 565                                                    "setting datasum, datacow enabled");
 566                                 else
 567                                         btrfs_info(info, "setting datasum");
 568                         }
 569                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 570                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 571                         break;
 572                 case Opt_nodatacow:
 573                         if (!btrfs_test_opt(info, NODATACOW)) {
 574                                 if (!btrfs_test_opt(info, COMPRESS) ||
 575                                     !btrfs_test_opt(info, FORCE_COMPRESS)) {
 576                                         btrfs_info(info,
 577                                                    "setting nodatacow, compression disabled");
 578                                 } else {
 579                                         btrfs_info(info, "setting nodatacow");
 580                                 }
 581                         }
 582                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 583                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 584                         btrfs_set_opt(info->mount_opt, NODATACOW);
 585                         btrfs_set_opt(info->mount_opt, NODATASUM);
 586                         break;
 587                 case Opt_datacow:
 588                         btrfs_clear_and_info(info, NODATACOW,
 589                                              "setting datacow");
 590                         break;
 591                 case Opt_compress_force:
 592                 case Opt_compress_force_type:
 593                         compress_force = true;
 594                         fallthrough;
 595                 case Opt_compress:
 596                 case Opt_compress_type:
 597                         saved_compress_type = btrfs_test_opt(info,
 598                                                              COMPRESS) ?
 599                                 info->compress_type : BTRFS_COMPRESS_NONE;
 600                         saved_compress_force =
 601                                 btrfs_test_opt(info, FORCE_COMPRESS);
 602                         saved_compress_level = info->compress_level;
 603                         if (token == Opt_compress ||
 604                             token == Opt_compress_force ||
 605                             strncmp(args[0].from, "zlib", 4) == 0) {
 606                                 compress_type = "zlib";
 607
 608                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 609                                 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 610                                 /*
 611                                  * args[0] contains uninitialized data since
 612                                  * for these tokens we don't expect any
 613                                  * parameter.
 614                                  */
 615                                 if (token != Opt_compress &&
 616                                     token != Opt_compress_force)
 617                                         info->compress_level =
 618                                           btrfs_compress_str2level(
 619                                                         BTRFS_COMPRESS_ZLIB,
 620                                                         args[0].from + 4);
 621                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 622                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 623                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 624                                 no_compress = 0;
 625                         } else if (strncmp(args[0].from, "lzo", 3) == 0) {
 626                                 compress_type = "lzo";
 627                                 info->compress_type = BTRFS_COMPRESS_LZO;
 628                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 629                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 630                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 631                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 632                                 no_compress = 0;
 633                         } else if (strncmp(args[0].from, "zstd", 4) == 0) {
 634                                 compress_type = "zstd";
 635                                 info->compress_type = BTRFS_COMPRESS_ZSTD;
 636                                 info->compress_level =
 637                                         btrfs_compress_str2level(
 638                                                          BTRFS_COMPRESS_ZSTD,
 639                                                          args[0].from + 4);
 640                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 641                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 642                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 643                                 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
 644                                 no_compress = 0;
 645                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 646                                 compress_type = "no";
 647                                 info->compress_level = 0;
 648                                 info->compress_type = 0;
 649                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 650                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 651                                 compress_force = false;
 652                                 no_compress++;
 653                         } else {
 654                                 ret = -EINVAL;
 655                                 goto out;
 656                         }
 657
 658                         if (compress_force) {
 659                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 660                         } else {
 661                                 /*
 662                                  * If we remount from compress-force=xxx to
 663                                  * compress=xxx, we need clear FORCE_COMPRESS
 664                                  * flag, otherwise, there is no way for users
 665                                  * to disable forcible compression separately.
 666                                  */
 667                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 668                         }
 669                         if (no_compress == 1) {
 670                                 btrfs_info(info, "use no compression");
 671                         } else if ((info->compress_type != saved_compress_type) ||
 672                                    (compress_force != saved_compress_force) ||
 673                                    (info->compress_level != saved_compress_level)) {
 674                                 btrfs_info(info, "%s %s compression, level %d",
 675                                            (compress_force) ? "force" : "use",
 676                                            compress_type, info->compress_level);
 677                         }
 678                         compress_force = false;
 679                         break;
 680                 case Opt_ssd:
 681                         btrfs_set_and_info(info, SSD,
 682                                            "enabling ssd optimizations");
 683                         btrfs_clear_opt(info->mount_opt, NOSSD);
 684                         break;
 685                 case Opt_ssd_spread:
 686                         btrfs_set_and_info(info, SSD,
 687                                            "enabling ssd optimizations");
 688                         btrfs_set_and_info(info, SSD_SPREAD,
 689                                            "using spread ssd allocation scheme");
 690                         btrfs_clear_opt(info->mount_opt, NOSSD);
 691                         break;
 692                 case Opt_nossd:
 693                         btrfs_set_opt(info->mount_opt, NOSSD);
 694                         btrfs_clear_and_info(info, SSD,
 695                                              "not using ssd optimizations");
 696                         fallthrough;
 697                 case Opt_nossd_spread:
 698                         btrfs_clear_and_info(info, SSD_SPREAD,
 699                                              "not using spread ssd allocation scheme");
 700                         break;
 701                 case Opt_barrier:
 702                         btrfs_clear_and_info(info, NOBARRIER,
 703                                              "turning on barriers");
 704                         break;
 705                 case Opt_nobarrier:
 706                         btrfs_set_and_info(info, NOBARRIER,
 707                                            "turning off barriers");
 708                         break;
 709                 case Opt_thread_pool:
 710                         ret = match_int(&args[0], &intarg);
 711                         if (ret) {
 712                                 goto out;
 713                         } else if (intarg == 0) {
 714                                 ret = -EINVAL;
 715                                 goto out;
 716                         }
 717                         info->thread_pool_size = intarg;
 718                         break;
 719                 case Opt_max_inline:
 720                         num = match_strdup(&args[0]);
 721                         if (num) {
 722                                 info->max_inline = memparse(num, NULL);
 723                                 kfree(num);
 724
 725                                 if (info->max_inline) {
 726                                         info->max_inline = min_t(u64,
 727                                                 info->max_inline,
 728                                                 info->sectorsize);
 729                                 }
 730                                 btrfs_info(info, "max_inline at %llu",
 731                                            info->max_inline);
 732                         } else {
 733                                 ret = -ENOMEM;
 734                                 goto out;
 735                         }
 736                         break;
 737                 case Opt_acl:
 738 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 739                         info->sb->s_flags |= SB_POSIXACL;
 740                         break;
 741 #else
 742                         btrfs_err(info, "support for ACL not compiled in!");
 743                         ret = -EINVAL;
 744                         goto out;
 745 #endif
 746                 case Opt_noacl:
 747                         info->sb->s_flags &= ~SB_POSIXACL;
 748                         break;
 749                 case Opt_notreelog:
 750                         btrfs_set_and_info(info, NOTREELOG,
 751                                            "disabling tree log");
 752                         break;
 753                 case Opt_treelog:
 754                         btrfs_clear_and_info(info, NOTREELOG,
 755                                              "enabling tree log");
 756                         break;
 757                 case Opt_norecovery:
 758                 case Opt_nologreplay:
 759                         btrfs_warn(info,
 760                 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 761                         btrfs_set_and_info(info, NOLOGREPLAY,
 762                                            "disabling log replay at mount time");
 763                         break;
 764                 case Opt_flushoncommit:
 765                         btrfs_set_and_info(info, FLUSHONCOMMIT,
 766                                            "turning on flush-on-commit");
 767                         break;
 768                 case Opt_noflushoncommit:
 769                         btrfs_clear_and_info(info, FLUSHONCOMMIT,
 770                                              "turning off flush-on-commit");
 771                         break;
 772                 case Opt_ratio:
 773                         ret = match_int(&args[0], &intarg);
 774                         if (ret)
 775                                 goto out;
 776                         info->metadata_ratio = intarg;
 777                         btrfs_info(info, "metadata ratio %u",
 778                                    info->metadata_ratio);
 779                         break;
 780                 case Opt_discard:
 781                 case Opt_discard_mode:
 782                         if (token == Opt_discard ||
 783                             strcmp(args[0].from, "sync") == 0) {
 784                                 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
 785                                 btrfs_set_and_info(info, DISCARD_SYNC,
 786                                                    "turning on sync discard");
 787                         } else if (strcmp(args[0].from, "async") == 0) {
 788                                 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
 789                                 btrfs_set_and_info(info, DISCARD_ASYNC,
 790                                                    "turning on async discard");
 791                         } else {
 792                                 ret = -EINVAL;
 793                                 goto out;
 794                         }
 795                         break;
 796                 case Opt_nodiscard:
 797                         btrfs_clear_and_info(info, DISCARD_SYNC,
 798                                              "turning off discard");
 799                         btrfs_clear_and_info(info, DISCARD_ASYNC,
 800                                              "turning off async discard");
 801                         break;
 802                 case Opt_space_cache:
 803                 case Opt_space_cache_version:
 804                         if (token == Opt_space_cache ||
 805                             strcmp(args[0].from, "v1") == 0) {
 806                                 btrfs_clear_opt(info->mount_opt,
 807                                                 FREE_SPACE_TREE);
 808                                 btrfs_set_and_info(info, SPACE_CACHE,
 809                                            "enabling disk space caching");
 810                         } else if (strcmp(args[0].from, "v2") == 0) {
 811                                 btrfs_clear_opt(info->mount_opt,
 812                                                 SPACE_CACHE);
 813                                 btrfs_set_and_info(info, FREE_SPACE_TREE,
 814                                                    "enabling free space tree");
 815                         } else {
 816                                 ret = -EINVAL;
 817                                 goto out;
 818                         }
 819                         break;
 820                 case Opt_rescan_uuid_tree:
 821                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 822                         break;
 823                 case Opt_no_space_cache:
 824                         if (btrfs_test_opt(info, SPACE_CACHE)) {
 825                                 btrfs_clear_and_info(info, SPACE_CACHE,
 826                                              "disabling disk space caching");
 827                         }
 828                         if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
 829                                 btrfs_clear_and_info(info, FREE_SPACE_TREE,
 830                                              "disabling free space tree");
 831                         }
 832                         break;
 833                 case Opt_inode_cache:
 834                         btrfs_warn(info,
 835         "the 'inode_cache' option is deprecated and will have no effect from 5.11");
 836                         btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
 837                                            "enabling inode map caching");
 838                         break;
 839                 case Opt_noinode_cache:
 840                         btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
 841                                              "disabling inode map caching");
 842                         break;
 843                 case Opt_clear_cache:
 844                         btrfs_set_and_info(info, CLEAR_CACHE,
 845                                            "force clearing of disk cache");
 846                         break;
 847                 case Opt_user_subvol_rm_allowed:
 848                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 849                         break;
 850                 case Opt_enospc_debug:
 851                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 852                         break;
 853                 case Opt_noenospc_debug:
 854                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 855                         break;
 856                 case Opt_defrag:
 857                         btrfs_set_and_info(info, AUTO_DEFRAG,
 858                                            "enabling auto defrag");
 859                         break;
 860                 case Opt_nodefrag:
 861                         btrfs_clear_and_info(info, AUTO_DEFRAG,
 862                                              "disabling auto defrag");
 863                         break;
 864                 case Opt_recovery:
 865                 case Opt_usebackuproot:
 866                         btrfs_warn(info,
 867                         "'%s' is deprecated, use 'rescue=usebackuproot' instead",
 868                                    token == Opt_recovery ? "recovery" :
 869                                    "usebackuproot");
 870                         btrfs_info(info,
 871                                    "trying to use backup root at mount time");
 872                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 873                         break;
 874                 case Opt_skip_balance:
 875                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 876                         break;
 877 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 878                 case Opt_check_integrity_including_extent_data:
 879                         btrfs_info(info,
 880                                    "enabling check integrity including extent data");
 881                         btrfs_set_opt(info->mount_opt,
 882                                       CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 883                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 884                         break;
 885                 case Opt_check_integrity:
 886                         btrfs_info(info, "enabling check integrity");
 887                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 888                         break;
 889                 case Opt_check_integrity_print_mask:
 890                         ret = match_int(&args[0], &intarg);
 891                         if (ret)
 892                                 goto out;
 893                         info->check_integrity_print_mask = intarg;
 894                         btrfs_info(info, "check_integrity_print_mask 0x%x",
 895                                    info->check_integrity_print_mask);
 896                         break;
 897 #else
 898                 case Opt_check_integrity_including_extent_data:
 899                 case Opt_check_integrity:
 900                 case Opt_check_integrity_print_mask:
 901                         btrfs_err(info,
 902                                   "support for check_integrity* not compiled in!");
 903                         ret = -EINVAL;
 904                         goto out;
 905 #endif
 906                 case Opt_fatal_errors:
 907                         if (strcmp(args[0].from, "panic") == 0)
 908                                 btrfs_set_opt(info->mount_opt,
 909                                               PANIC_ON_FATAL_ERROR);
 910                         else if (strcmp(args[0].from, "bug") == 0)
 911                                 btrfs_clear_opt(info->mount_opt,
 912                                               PANIC_ON_FATAL_ERROR);
 913                         else {
 914                                 ret = -EINVAL;
 915                                 goto out;
 916                         }
 917                         break;
 918                 case Opt_commit_interval:
 919                         intarg = 0;
 920                         ret = match_int(&args[0], &intarg);
 921                         if (ret)
 922                                 goto out;
 923                         if (intarg == 0) {
 924                                 btrfs_info(info,
 925                                            "using default commit interval %us",
 926                                            BTRFS_DEFAULT_COMMIT_INTERVAL);
 927                                 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
 928                         } else if (intarg > 300) {
 929                                 btrfs_warn(info, "excessive commit interval %d",
 930                                            intarg);
 931                         }
 932                         info->commit_interval = intarg;
 933                         break;
 934                 case Opt_rescue:
 935                         ret = parse_rescue_options(info, args[0].from);
 936                         if (ret < 0)
 937                                 goto out;
 938                         break;
 939 #ifdef CONFIG_BTRFS_DEBUG
 940                 case Opt_fragment_all:
 941                         btrfs_info(info, "fragmenting all space");
 942                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 943                         btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
 944                         break;
 945                 case Opt_fragment_metadata:
 946                         btrfs_info(info, "fragmenting metadata");
 947                         btrfs_set_opt(info->mount_opt,
 948                                       FRAGMENT_METADATA);
 949                         break;
 950                 case Opt_fragment_data:
 951                         btrfs_info(info, "fragmenting data");
 952                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 953                         break;
 954 #endif
 955 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 956                 case Opt_ref_verify:
 957                         btrfs_info(info, "doing ref verification");
 958                         btrfs_set_opt(info->mount_opt, REF_VERIFY);
 959                         break;
 960 #endif
 961                 case Opt_err:
 962                         btrfs_err(info, "unrecognized mount option '%s'", p);
 963                         ret = -EINVAL;
 964                         goto out;
 965                 default:
 966                         break;
 967                 }
 968         }
 969 check:
 970         /*
 971          * Extra check for current option against current flag
 972          */
 973         if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
 974                 btrfs_err(info,
 975                           "nologreplay must be used with ro mount option");
 976                 ret = -EINVAL;
 977         }
 978 out:
 979         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 980             !btrfs_test_opt(info, FREE_SPACE_TREE) &&
 981             !btrfs_test_opt(info, CLEAR_CACHE)) {
 982                 btrfs_err(info, "cannot disable free space tree");
 983                 ret = -EINVAL;
 984
 985         }
 986         if (!ret && btrfs_test_opt(info, SPACE_CACHE))
 987                 btrfs_info(info, "disk space caching is enabled");
 988         if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
 989                 btrfs_info(info, "using free space tree");
 990         return ret;
 991 }
 992
 993 /*
 994  * Parse mount options that are required early in the mount process.
 995  *
 996  * All other options will be parsed on much later in the mount process and
 997  * only when we need to allocate a new super block.
 998  */
 999 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1000                                       void *holder)
1001 {
1002         substring_t args[MAX_OPT_ARGS];
1003         char *device_name, *opts, *orig, *p;
1004         struct btrfs_device *device = NULL;
1005         int error = 0;
1006
1007         lockdep_assert_held(&uuid_mutex);
1008
1009         if (!options)
1010                 return 0;
1011
1012         /*
1013          * strsep changes the string, duplicate it because btrfs_parse_options
1014          * gets called later
1015          */
1016         opts = kstrdup(options, GFP_KERNEL);
1017         if (!opts)
1018                 return -ENOMEM;
1019         orig = opts;
1020
1021         while ((p = strsep(&opts, ",")) != NULL) {
1022                 int token;
1023
1024                 if (!*p)
1025                         continue;
1026
1027                 token = match_token(p, tokens, args);
1028                 if (token == Opt_device) {
1029                         device_name = match_strdup(&args[0]);
1030                         if (!device_name) {
1031                                 error = -ENOMEM;
1032                                 goto out;
1033                         }
1034                         device = btrfs_scan_one_device(device_name, flags,
1035                                         holder);
1036                         kfree(device_name);
1037                         if (IS_ERR(device)) {
1038                                 error = PTR_ERR(device);
1039                                 goto out;
1040                         }
1041                 }
1042         }
1043
1044 out:
1045         kfree(orig);
1046         return error;
1047 }
1048
1049 /*
1050  * Parse mount options that are related to subvolume id
1051  *
1052  * The value is later passed to mount_subvol()
1053  */
1054 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1055                 u64 *subvol_objectid)
1056 {
1057         substring_t args[MAX_OPT_ARGS];
1058         char *opts, *orig, *p;
1059         int error = 0;
1060         u64 subvolid;
1061
1062         if (!options)
1063                 return 0;
1064
1065         /*
1066          * strsep changes the string, duplicate it because
1067          * btrfs_parse_device_options gets called later
1068          */
1069         opts = kstrdup(options, GFP_KERNEL);
1070         if (!opts)
1071                 return -ENOMEM;
1072         orig = opts;
1073
1074         while ((p = strsep(&opts, ",")) != NULL) {
1075                 int token;
1076                 if (!*p)
1077                         continue;
1078
1079                 token = match_token(p, tokens, args);
1080                 switch (token) {
1081                 case Opt_subvol:
1082                         kfree(*subvol_name);
1083                         *subvol_name = match_strdup(&args[0]);
1084                         if (!*subvol_name) {
1085                                 error = -ENOMEM;
1086                                 goto out;
1087                         }
1088                         break;
1089                 case Opt_subvolid:
1090                         error = match_u64(&args[0], &subvolid);
1091                         if (error)
1092                                 goto out;
1093
1094                         /* we want the original fs_tree */
1095                         if (subvolid == 0)
1096                                 subvolid = BTRFS_FS_TREE_OBJECTID;
1097
1098                         *subvol_objectid = subvolid;
1099                         break;
1100                 default:
1101                         break;
1102                 }
1103         }
1104
1105 out:
1106         kfree(orig);
1107         return error;
1108 }
1109
1110 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1111                                           u64 subvol_objectid)
1112 {
1113         struct btrfs_root *root = fs_info->tree_root;
1114         struct btrfs_root *fs_root = NULL;
1115         struct btrfs_root_ref *root_ref;
1116         struct btrfs_inode_ref *inode_ref;
1117         struct btrfs_key key;
1118         struct btrfs_path *path = NULL;
1119         char *name = NULL, *ptr;
1120         u64 dirid;
1121         int len;
1122         int ret;
1123
1124         path = btrfs_alloc_path();
1125         if (!path) {
1126                 ret = -ENOMEM;
1127                 goto err;
1128         }
1129         path->leave_spinning = 1;
1130
1131         name = kmalloc(PATH_MAX, GFP_KERNEL);
1132         if (!name) {
1133                 ret = -ENOMEM;
1134                 goto err;
1135         }
1136         ptr = name + PATH_MAX - 1;
1137         ptr[0] = '\0';
1138
1139         /*
1140          * Walk up the subvolume trees in the tree of tree roots by root
1141          * backrefs until we hit the top-level subvolume.
1142          */
1143         while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1144                 key.objectid = subvol_objectid;
1145                 key.type = BTRFS_ROOT_BACKREF_KEY;
1146                 key.offset = (u64)-1;
1147
1148                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1149                 if (ret < 0) {
1150                         goto err;
1151                 } else if (ret > 0) {
1152                         ret = btrfs_previous_item(root, path, subvol_objectid,
1153                                                   BTRFS_ROOT_BACKREF_KEY);
1154                         if (ret < 0) {
1155                                 goto err;
1156                         } else if (ret > 0) {
1157                                 ret = -ENOENT;
1158                                 goto err;
1159                         }
1160                 }
1161
1162                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1163                 subvol_objectid = key.offset;
1164
1165                 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1166                                           struct btrfs_root_ref);
1167                 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1168                 ptr -= len + 1;
1169                 if (ptr < name) {
1170                         ret = -ENAMETOOLONG;
1171                         goto err;
1172                 }
1173                 read_extent_buffer(path->nodes[0], ptr + 1,
1174                                    (unsigned long)(root_ref + 1), len);
1175                 ptr[0] = '/';
1176                 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1177                 btrfs_release_path(path);
1178
1179                 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1180                 if (IS_ERR(fs_root)) {
1181                         ret = PTR_ERR(fs_root);
1182                         fs_root = NULL;
1183                         goto err;
1184                 }
1185
1186                 /*
1187                  * Walk up the filesystem tree by inode refs until we hit the
1188                  * root directory.
1189                  */
1190                 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1191                         key.objectid = dirid;
1192                         key.type = BTRFS_INODE_REF_KEY;
1193                         key.offset = (u64)-1;
1194
1195                         ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1196                         if (ret < 0) {
1197                                 goto err;
1198                         } else if (ret > 0) {
1199                                 ret = btrfs_previous_item(fs_root, path, dirid,
1200                                                           BTRFS_INODE_REF_KEY);
1201                                 if (ret < 0) {
1202                                         goto err;
1203                                 } else if (ret > 0) {
1204                                         ret = -ENOENT;
1205                                         goto err;
1206                                 }
1207                         }
1208
1209                         btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1210                         dirid = key.offset;
1211
1212                         inode_ref = btrfs_item_ptr(path->nodes[0],
1213                                                    path->slots[0],
1214                                                    struct btrfs_inode_ref);
1215                         len = btrfs_inode_ref_name_len(path->nodes[0],
1216                                                        inode_ref);
1217                         ptr -= len + 1;
1218                         if (ptr < name) {
1219                                 ret = -ENAMETOOLONG;
1220                                 goto err;
1221                         }
1222                         read_extent_buffer(path->nodes[0], ptr + 1,
1223                                            (unsigned long)(inode_ref + 1), len);
1224                         ptr[0] = '/';
1225                         btrfs_release_path(path);
1226                 }
1227                 btrfs_put_root(fs_root);
1228                 fs_root = NULL;
1229         }
1230
1231         btrfs_free_path(path);
1232         if (ptr == name + PATH_MAX - 1) {
1233                 name[0] = '/';
1234                 name[1] = '\0';
1235         } else {
1236                 memmove(name, ptr, name + PATH_MAX - ptr);
1237         }
1238         return name;
1239
1240 err:
1241         btrfs_put_root(fs_root);
1242         btrfs_free_path(path);
1243         kfree(name);
1244         return ERR_PTR(ret);
1245 }
1246
1247 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1248 {
1249         struct btrfs_root *root = fs_info->tree_root;
1250         struct btrfs_dir_item *di;
1251         struct btrfs_path *path;
1252         struct btrfs_key location;
1253         u64 dir_id;
1254
1255         path = btrfs_alloc_path();
1256         if (!path)
1257                 return -ENOMEM;
1258         path->leave_spinning = 1;
1259
1260         /*
1261          * Find the "default" dir item which points to the root item that we
1262          * will mount by default if we haven't been given a specific subvolume
1263          * to mount.
1264          */
1265         dir_id = btrfs_super_root_dir(fs_info->super_copy);
1266         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1267         if (IS_ERR(di)) {
1268                 btrfs_free_path(path);
1269                 return PTR_ERR(di);
1270         }
1271         if (!di) {
1272                 /*
1273                  * Ok the default dir item isn't there.  This is weird since
1274                  * it's always been there, but don't freak out, just try and
1275                  * mount the top-level subvolume.
1276                  */
1277                 btrfs_free_path(path);
1278                 *objectid = BTRFS_FS_TREE_OBJECTID;
1279                 return 0;
1280         }
1281
1282         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1283         btrfs_free_path(path);
1284         *objectid = location.objectid;
1285         return 0;
1286 }
1287
1288 static int btrfs_fill_super(struct super_block *sb,
1289                             struct btrfs_fs_devices *fs_devices,
1290                             void *data)
1291 {
1292         struct inode *inode;
1293         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1294         int err;
1295
1296         sb->s_maxbytes = MAX_LFS_FILESIZE;
1297         sb->s_magic = BTRFS_SUPER_MAGIC;
1298         sb->s_op = &btrfs_super_ops;
1299         sb->s_d_op = &btrfs_dentry_operations;
1300         sb->s_export_op = &btrfs_export_ops;
1301         sb->s_xattr = btrfs_xattr_handlers;
1302         sb->s_time_gran = 1;
1303 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1304         sb->s_flags |= SB_POSIXACL;
1305 #endif
1306         sb->s_flags |= SB_I_VERSION;
1307         sb->s_iflags |= SB_I_CGROUPWB;
1308
1309         err = super_setup_bdi(sb);
1310         if (err) {
1311                 btrfs_err(fs_info, "super_setup_bdi failed");
1312                 return err;
1313         }
1314
1315         err = open_ctree(sb, fs_devices, (char *)data);
1316         if (err) {
1317                 btrfs_err(fs_info, "open_ctree failed");
1318                 return err;
1319         }
1320
1321         inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1322         if (IS_ERR(inode)) {
1323                 err = PTR_ERR(inode);
1324                 goto fail_close;
1325         }
1326
1327         sb->s_root = d_make_root(inode);
1328         if (!sb->s_root) {
1329                 err = -ENOMEM;
1330                 goto fail_close;
1331         }
1332
1333         cleancache_init_fs(sb);
1334         sb->s_flags |= SB_ACTIVE;
1335         return 0;
1336
1337 fail_close:
1338         close_ctree(fs_info);
1339         return err;
1340 }
1341
1342 int btrfs_sync_fs(struct super_block *sb, int wait)
1343 {
1344         struct btrfs_trans_handle *trans;
1345         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1346         struct btrfs_root *root = fs_info->tree_root;
1347
1348         trace_btrfs_sync_fs(fs_info, wait);
1349
1350         if (!wait) {
1351                 filemap_flush(fs_info->btree_inode->i_mapping);
1352                 return 0;
1353         }
1354
1355         btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1356
1357         trans = btrfs_attach_transaction_barrier(root);
1358         if (IS_ERR(trans)) {
1359                 /* no transaction, don't bother */
1360                 if (PTR_ERR(trans) == -ENOENT) {
1361                         /*
1362                          * Exit unless we have some pending changes
1363                          * that need to go through commit
1364                          */
1365                         if (fs_info->pending_changes == 0)
1366                                 return 0;
1367                         /*
1368                          * A non-blocking test if the fs is frozen. We must not
1369                          * start a new transaction here otherwise a deadlock
1370                          * happens. The pending operations are delayed to the
1371                          * next commit after thawing.
1372                          */
1373                         if (sb_start_write_trylock(sb))
1374                                 sb_end_write(sb);
1375                         else
1376                                 return 0;
1377                         trans = btrfs_start_transaction(root, 0);
1378                 }
1379                 if (IS_ERR(trans))
1380                         return PTR_ERR(trans);
1381         }
1382         return btrfs_commit_transaction(trans);
1383 }
1384
1385 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1386 {
1387         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1388         const char *compress_type;
1389         const char *subvol_name;
1390
1391         if (btrfs_test_opt(info, DEGRADED))
1392                 seq_puts(seq, ",degraded");
1393         if (btrfs_test_opt(info, NODATASUM))
1394                 seq_puts(seq, ",nodatasum");
1395         if (btrfs_test_opt(info, NODATACOW))
1396                 seq_puts(seq, ",nodatacow");
1397         if (btrfs_test_opt(info, NOBARRIER))
1398                 seq_puts(seq, ",nobarrier");
1399         if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1400                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1401         if (info->thread_pool_size !=  min_t(unsigned long,
1402                                              num_online_cpus() + 2, 8))
1403                 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1404         if (btrfs_test_opt(info, COMPRESS)) {
1405                 compress_type = btrfs_compress_type2str(info->compress_type);
1406                 if (btrfs_test_opt(info, FORCE_COMPRESS))
1407                         seq_printf(seq, ",compress-force=%s", compress_type);
1408                 else
1409                         seq_printf(seq, ",compress=%s", compress_type);
1410                 if (info->compress_level)
1411                         seq_printf(seq, ":%d", info->compress_level);
1412         }
1413         if (btrfs_test_opt(info, NOSSD))
1414                 seq_puts(seq, ",nossd");
1415         if (btrfs_test_opt(info, SSD_SPREAD))
1416                 seq_puts(seq, ",ssd_spread");
1417         else if (btrfs_test_opt(info, SSD))
1418                 seq_puts(seq, ",ssd");
1419         if (btrfs_test_opt(info, NOTREELOG))
1420                 seq_puts(seq, ",notreelog");
1421         if (btrfs_test_opt(info, NOLOGREPLAY))
1422                 seq_puts(seq, ",rescue=nologreplay");
1423         if (btrfs_test_opt(info, FLUSHONCOMMIT))
1424                 seq_puts(seq, ",flushoncommit");
1425         if (btrfs_test_opt(info, DISCARD_SYNC))
1426                 seq_puts(seq, ",discard");
1427         if (btrfs_test_opt(info, DISCARD_ASYNC))
1428                 seq_puts(seq, ",discard=async");
1429         if (!(info->sb->s_flags & SB_POSIXACL))
1430                 seq_puts(seq, ",noacl");
1431         if (btrfs_test_opt(info, SPACE_CACHE))
1432                 seq_puts(seq, ",space_cache");
1433         else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1434                 seq_puts(seq, ",space_cache=v2");
1435         else
1436                 seq_puts(seq, ",nospace_cache");
1437         if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1438                 seq_puts(seq, ",rescan_uuid_tree");
1439         if (btrfs_test_opt(info, CLEAR_CACHE))
1440                 seq_puts(seq, ",clear_cache");
1441         if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1442                 seq_puts(seq, ",user_subvol_rm_allowed");
1443         if (btrfs_test_opt(info, ENOSPC_DEBUG))
1444                 seq_puts(seq, ",enospc_debug");
1445         if (btrfs_test_opt(info, AUTO_DEFRAG))
1446                 seq_puts(seq, ",autodefrag");
1447         if (btrfs_test_opt(info, INODE_MAP_CACHE))
1448                 seq_puts(seq, ",inode_cache");
1449         if (btrfs_test_opt(info, SKIP_BALANCE))
1450                 seq_puts(seq, ",skip_balance");
1451 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1452         if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1453                 seq_puts(seq, ",check_int_data");
1454         else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1455                 seq_puts(seq, ",check_int");
1456         if (info->check_integrity_print_mask)
1457                 seq_printf(seq, ",check_int_print_mask=%d",
1458                                 info->check_integrity_print_mask);
1459 #endif
1460         if (info->metadata_ratio)
1461                 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1462         if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1463                 seq_puts(seq, ",fatal_errors=panic");
1464         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1465                 seq_printf(seq, ",commit=%u", info->commit_interval);
1466 #ifdef CONFIG_BTRFS_DEBUG
1467         if (btrfs_test_opt(info, FRAGMENT_DATA))
1468                 seq_puts(seq, ",fragment=data");
1469         if (btrfs_test_opt(info, FRAGMENT_METADATA))
1470                 seq_puts(seq, ",fragment=metadata");
1471 #endif
1472         if (btrfs_test_opt(info, REF_VERIFY))
1473                 seq_puts(seq, ",ref_verify");
1474         seq_printf(seq, ",subvolid=%llu",
1475                   BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1476         subvol_name = btrfs_get_subvol_name_from_objectid(info,
1477                         BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1478         if (!IS_ERR(subvol_name)) {
1479                 seq_puts(seq, ",subvol=");
1480                 seq_escape(seq, subvol_name, " \t\n\\");
1481                 kfree(subvol_name);
1482         }
1483         return 0;
1484 }
1485
1486 static int btrfs_test_super(struct super_block *s, void *data)
1487 {
1488         struct btrfs_fs_info *p = data;
1489         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1490
1491         return fs_info->fs_devices == p->fs_devices;
1492 }
1493
1494 static int btrfs_set_super(struct super_block *s, void *data)
1495 {
1496         int err = set_anon_super(s, data);
1497         if (!err)
1498                 s->s_fs_info = data;
1499         return err;
1500 }
1501
1502 /*
1503  * subvolumes are identified by ino 256
1504  */
1505 static inline int is_subvolume_inode(struct inode *inode)
1506 {
1507         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1508                 return 1;
1509         return 0;
1510 }
1511
1512 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1513                                    struct vfsmount *mnt)
1514 {
1515         struct dentry *root;
1516         int ret;
1517
1518         if (!subvol_name) {
1519                 if (!subvol_objectid) {
1520                         ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1521                                                           &subvol_objectid);
1522                         if (ret) {
1523                                 root = ERR_PTR(ret);
1524                                 goto out;
1525                         }
1526                 }
1527                 subvol_name = btrfs_get_subvol_name_from_objectid(
1528                                         btrfs_sb(mnt->mnt_sb), subvol_objectid);
1529                 if (IS_ERR(subvol_name)) {
1530                         root = ERR_CAST(subvol_name);
1531                         subvol_name = NULL;
1532                         goto out;
1533                 }
1534
1535         }
1536
1537         root = mount_subtree(mnt, subvol_name);
1538         /* mount_subtree() drops our reference on the vfsmount. */
1539         mnt = NULL;
1540
1541         if (!IS_ERR(root)) {
1542                 struct super_block *s = root->d_sb;
1543                 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1544                 struct inode *root_inode = d_inode(root);
1545                 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1546
1547                 ret = 0;
1548                 if (!is_subvolume_inode(root_inode)) {
1549                         btrfs_err(fs_info, "'%s' is not a valid subvolume",
1550                                subvol_name);
1551                         ret = -EINVAL;
1552                 }
1553                 if (subvol_objectid && root_objectid != subvol_objectid) {
1554                         /*
1555                          * This will also catch a race condition where a
1556                          * subvolume which was passed by ID is renamed and
1557                          * another subvolume is renamed over the old location.
1558                          */
1559                         btrfs_err(fs_info,
1560                                   "subvol '%s' does not match subvolid %llu",
1561                                   subvol_name, subvol_objectid);
1562                         ret = -EINVAL;
1563                 }
1564                 if (ret) {
1565                         dput(root);
1566                         root = ERR_PTR(ret);
1567                         deactivate_locked_super(s);
1568                 }
1569         }
1570
1571 out:
1572         mntput(mnt);
1573         kfree(subvol_name);
1574         return root;
1575 }
1576
1577 /*
1578  * Find a superblock for the given device / mount point.
1579  *
1580  * Note: This is based on mount_bdev from fs/super.c with a few additions
1581  *       for multiple device setup.  Make sure to keep it in sync.
1582  */
1583 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1584                 int flags, const char *device_name, void *data)
1585 {
1586         struct block_device *bdev = NULL;
1587         struct super_block *s;
1588         struct btrfs_device *device = NULL;
1589         struct btrfs_fs_devices *fs_devices = NULL;
1590         struct btrfs_fs_info *fs_info = NULL;
1591         void *new_sec_opts = NULL;
1592         fmode_t mode = FMODE_READ;
1593         int error = 0;
1594
1595         if (!(flags & SB_RDONLY))
1596                 mode |= FMODE_WRITE;
1597
1598         if (data) {
1599                 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1600                 if (error)
1601                         return ERR_PTR(error);
1602         }
1603
1604         /*
1605          * Setup a dummy root and fs_info for test/set super.  This is because
1606          * we don't actually fill this stuff out until open_ctree, but we need
1607          * then open_ctree will properly initialize the file system specific
1608          * settings later.  btrfs_init_fs_info initializes the static elements
1609          * of the fs_info (locks and such) to make cleanup easier if we find a
1610          * superblock with our given fs_devices later on at sget() time.
1611          */
1612         fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1613         if (!fs_info) {
1614                 error = -ENOMEM;
1615                 goto error_sec_opts;
1616         }
1617         btrfs_init_fs_info(fs_info);
1618
1619         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1620         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1621         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1622                 error = -ENOMEM;
1623                 goto error_fs_info;
1624         }
1625
1626         mutex_lock(&uuid_mutex);
1627         error = btrfs_parse_device_options(data, mode, fs_type);
1628         if (error) {
1629                 mutex_unlock(&uuid_mutex);
1630                 goto error_fs_info;
1631         }
1632
1633         device = btrfs_scan_one_device(device_name, mode, fs_type);
1634         if (IS_ERR(device)) {
1635                 mutex_unlock(&uuid_mutex);
1636                 error = PTR_ERR(device);
1637                 goto error_fs_info;
1638         }
1639
1640         fs_devices = device->fs_devices;
1641         fs_info->fs_devices = fs_devices;
1642
1643         error = btrfs_open_devices(fs_devices, mode, fs_type);
1644         mutex_unlock(&uuid_mutex);
1645         if (error)
1646                 goto error_fs_info;
1647
1648         if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1649                 error = -EACCES;
1650                 goto error_close_devices;
1651         }
1652
1653         bdev = fs_devices->latest_bdev;
1654         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1655                  fs_info);
1656         if (IS_ERR(s)) {
1657                 error = PTR_ERR(s);
1658                 goto error_close_devices;
1659         }
1660
1661         if (s->s_root) {
1662                 btrfs_close_devices(fs_devices);
1663                 btrfs_free_fs_info(fs_info);
1664                 if ((flags ^ s->s_flags) & SB_RDONLY)
1665                         error = -EBUSY;
1666         } else {
1667                 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1668                 btrfs_sb(s)->bdev_holder = fs_type;
1669                 if (!strstr(crc32c_impl(), "generic"))
1670                         set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1671                 error = btrfs_fill_super(s, fs_devices, data);
1672         }
1673         if (!error)
1674                 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1675         security_free_mnt_opts(&new_sec_opts);
1676         if (error) {
1677                 deactivate_locked_super(s);
1678                 return ERR_PTR(error);
1679         }
1680
1681         return dget(s->s_root);
1682
1683 error_close_devices:
1684         btrfs_close_devices(fs_devices);
1685 error_fs_info:
1686         btrfs_free_fs_info(fs_info);
1687 error_sec_opts:
1688         security_free_mnt_opts(&new_sec_opts);
1689         return ERR_PTR(error);
1690 }
1691
1692 /*
1693  * Mount function which is called by VFS layer.
1694  *
1695  * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1696  * which needs vfsmount* of device's root (/).  This means device's root has to
1697  * be mounted internally in any case.
1698  *
1699  * Operation flow:
1700  *   1. Parse subvol id related options for later use in mount_subvol().
1701  *
1702  *   2. Mount device's root (/) by calling vfs_kern_mount().
1703  *
1704  *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1705  *      first place. In order to avoid calling btrfs_mount() again, we use
1706  *      different file_system_type which is not registered to VFS by
1707  *      register_filesystem() (btrfs_root_fs_type). As a result,
1708  *      btrfs_mount_root() is called. The return value will be used by
1709  *      mount_subtree() in mount_subvol().
1710  *
1711  *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1712  *      "btrfs subvolume set-default", mount_subvol() is called always.
1713  */
1714 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1715                 const char *device_name, void *data)
1716 {
1717         struct vfsmount *mnt_root;
1718         struct dentry *root;
1719         char *subvol_name = NULL;
1720         u64 subvol_objectid = 0;
1721         int error = 0;
1722
1723         error = btrfs_parse_subvol_options(data, &subvol_name,
1724                                         &subvol_objectid);
1725         if (error) {
1726                 kfree(subvol_name);
1727                 return ERR_PTR(error);
1728         }
1729
1730         /* mount device's root (/) */
1731         mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1732         if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1733                 if (flags & SB_RDONLY) {
1734                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1735                                 flags & ~SB_RDONLY, device_name, data);
1736                 } else {
1737                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1738                                 flags | SB_RDONLY, device_name, data);
1739                         if (IS_ERR(mnt_root)) {
1740                                 root = ERR_CAST(mnt_root);
1741                                 kfree(subvol_name);
1742                                 goto out;
1743                         }
1744
1745                         down_write(&mnt_root->mnt_sb->s_umount);
1746                         error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1747                         up_write(&mnt_root->mnt_sb->s_umount);
1748                         if (error < 0) {
1749                                 root = ERR_PTR(error);
1750                                 mntput(mnt_root);
1751                                 kfree(subvol_name);
1752                                 goto out;
1753                         }
1754                 }
1755         }
1756         if (IS_ERR(mnt_root)) {
1757                 root = ERR_CAST(mnt_root);
1758                 kfree(subvol_name);
1759                 goto out;
1760         }
1761
1762         /* mount_subvol() will free subvol_name and mnt_root */
1763         root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1764
1765 out:
1766         return root;
1767 }
1768
1769 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1770                                      u32 new_pool_size, u32 old_pool_size)
1771 {
1772         if (new_pool_size == old_pool_size)
1773                 return;
1774
1775         fs_info->thread_pool_size = new_pool_size;
1776
1777         btrfs_info(fs_info, "resize thread pool %d -> %d",
1778                old_pool_size, new_pool_size);
1779
1780         btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1781         btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1782         btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1783         btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1784         btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1785         btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1786                                 new_pool_size);
1787         btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1788         btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1789         btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1790         btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1791         btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1792                                 new_pool_size);
1793 }
1794
1795 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1796                                        unsigned long old_opts, int flags)
1797 {
1798         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1799             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1800              (flags & SB_RDONLY))) {
1801                 /* wait for any defraggers to finish */
1802                 wait_event(fs_info->transaction_wait,
1803                            (atomic_read(&fs_info->defrag_running) == 0));
1804                 if (flags & SB_RDONLY)
1805                         sync_filesystem(fs_info->sb);
1806         }
1807 }
1808
1809 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1810                                          unsigned long old_opts)
1811 {
1812         /*
1813          * We need to cleanup all defragable inodes if the autodefragment is
1814          * close or the filesystem is read only.
1815          */
1816         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1817             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1818                 btrfs_cleanup_defrag_inodes(fs_info);
1819         }
1820
1821         /* If we toggled discard async */
1822         if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1823             btrfs_test_opt(fs_info, DISCARD_ASYNC))
1824                 btrfs_discard_resume(fs_info);
1825         else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1826                  !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1827                 btrfs_discard_cleanup(fs_info);
1828 }
1829
1830 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1831 {
1832         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1833         struct btrfs_root *root = fs_info->tree_root;
1834         unsigned old_flags = sb->s_flags;
1835         unsigned long old_opts = fs_info->mount_opt;
1836         unsigned long old_compress_type = fs_info->compress_type;
1837         u64 old_max_inline = fs_info->max_inline;
1838         u32 old_thread_pool_size = fs_info->thread_pool_size;
1839         u32 old_metadata_ratio = fs_info->metadata_ratio;
1840         int ret;
1841
1842         sync_filesystem(sb);
1843         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1844
1845         if (data) {
1846                 void *new_sec_opts = NULL;
1847
1848                 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1849                 if (!ret)
1850                         ret = security_sb_remount(sb, new_sec_opts);
1851                 security_free_mnt_opts(&new_sec_opts);
1852                 if (ret)
1853                         goto restore;
1854         }
1855
1856         ret = btrfs_parse_options(fs_info, data, *flags);
1857         if (ret)
1858                 goto restore;
1859
1860         btrfs_remount_begin(fs_info, old_opts, *flags);
1861         btrfs_resize_thread_pool(fs_info,
1862                 fs_info->thread_pool_size, old_thread_pool_size);
1863
1864         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1865                 goto out;
1866
1867         if (*flags & SB_RDONLY) {
1868                 /*
1869                  * this also happens on 'umount -rf' or on shutdown, when
1870                  * the filesystem is busy.
1871                  */
1872                 cancel_work_sync(&fs_info->async_reclaim_work);
1873
1874                 btrfs_discard_cleanup(fs_info);
1875
1876                 /* wait for the uuid_scan task to finish */
1877                 down(&fs_info->uuid_tree_rescan_sem);
1878                 /* avoid complains from lockdep et al. */
1879                 up(&fs_info->uuid_tree_rescan_sem);
1880
1881                 sb->s_flags |= SB_RDONLY;
1882
1883                 /*
1884                  * Setting SB_RDONLY will put the cleaner thread to
1885                  * sleep at the next loop if it's already active.
1886                  * If it's already asleep, we'll leave unused block
1887                  * groups on disk until we're mounted read-write again
1888                  * unless we clean them up here.
1889                  */
1890                 btrfs_delete_unused_bgs(fs_info);
1891
1892                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1893                 btrfs_scrub_cancel(fs_info);
1894                 btrfs_pause_balance(fs_info);
1895
1896                 ret = btrfs_commit_super(fs_info);
1897                 if (ret)
1898                         goto restore;
1899         } else {
1900                 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1901                         btrfs_err(fs_info,
1902                                 "Remounting read-write after error is not allowed");
1903                         ret = -EINVAL;
1904                         goto restore;
1905                 }
1906                 if (fs_info->fs_devices->rw_devices == 0) {
1907                         ret = -EACCES;
1908                         goto restore;
1909                 }
1910
1911                 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1912                         btrfs_warn(fs_info,
1913                 "too many missing devices, writable remount is not allowed");
1914                         ret = -EACCES;
1915                         goto restore;
1916                 }
1917
1918                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1919                         btrfs_warn(fs_info,
1920                 "mount required to replay tree-log, cannot remount read-write");
1921                         ret = -EINVAL;
1922                         goto restore;
1923                 }
1924
1925                 ret = btrfs_cleanup_fs_roots(fs_info);
1926                 if (ret)
1927                         goto restore;
1928
1929                 /* recover relocation */
1930                 mutex_lock(&fs_info->cleaner_mutex);
1931                 ret = btrfs_recover_relocation(root);
1932                 mutex_unlock(&fs_info->cleaner_mutex);
1933                 if (ret)
1934                         goto restore;
1935
1936                 ret = btrfs_resume_balance_async(fs_info);
1937                 if (ret)
1938                         goto restore;
1939
1940                 ret = btrfs_resume_dev_replace_async(fs_info);
1941                 if (ret) {
1942                         btrfs_warn(fs_info, "failed to resume dev_replace");
1943                         goto restore;
1944                 }
1945
1946                 btrfs_qgroup_rescan_resume(fs_info);
1947
1948                 if (!fs_info->uuid_root) {
1949                         btrfs_info(fs_info, "creating UUID tree");
1950                         ret = btrfs_create_uuid_tree(fs_info);
1951                         if (ret) {
1952                                 btrfs_warn(fs_info,
1953                                            "failed to create the UUID tree %d",
1954                                            ret);
1955                                 goto restore;
1956                         }
1957                 }
1958                 sb->s_flags &= ~SB_RDONLY;
1959
1960                 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1961         }
1962 out:
1963         /*
1964          * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1965          * since the absence of the flag means it can be toggled off by remount.
1966          */
1967         *flags |= SB_I_VERSION;
1968
1969         wake_up_process(fs_info->transaction_kthread);
1970         btrfs_remount_cleanup(fs_info, old_opts);
1971         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1972
1973         return 0;
1974
1975 restore:
1976         /* We've hit an error - don't reset SB_RDONLY */
1977         if (sb_rdonly(sb))
1978                 old_flags |= SB_RDONLY;
1979         sb->s_flags = old_flags;
1980         fs_info->mount_opt = old_opts;
1981         fs_info->compress_type = old_compress_type;
1982         fs_info->max_inline = old_max_inline;
1983         btrfs_resize_thread_pool(fs_info,
1984                 old_thread_pool_size, fs_info->thread_pool_size);
1985         fs_info->metadata_ratio = old_metadata_ratio;
1986         btrfs_remount_cleanup(fs_info, old_opts);
1987         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1988
1989         return ret;
1990 }
1991
1992 /* Used to sort the devices by max_avail(descending sort) */
1993 static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
1994                                        const void *dev_info2)
1995 {
1996         if (((struct btrfs_device_info *)dev_info1)->max_avail >
1997             ((struct btrfs_device_info *)dev_info2)->max_avail)
1998                 return -1;
1999         else if (((struct btrfs_device_info *)dev_info1)->max_avail <
2000                  ((struct btrfs_device_info *)dev_info2)->max_avail)
2001                 return 1;
2002         else
2003         return 0;
2004 }
2005
2006 /*
2007  * sort the devices by max_avail, in which max free extent size of each device
2008  * is stored.(Descending Sort)
2009  */
2010 static inline void btrfs_descending_sort_devices(
2011                                         struct btrfs_device_info *devices,
2012                                         size_t nr_devices)
2013 {
2014         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2015              btrfs_cmp_device_free_bytes, NULL);
2016 }
2017
2018 /*
2019  * The helper to calc the free space on the devices that can be used to store
2020  * file data.
2021  */
2022 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2023                                               u64 *free_bytes)
2024 {
2025         struct btrfs_device_info *devices_info;
2026         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2027         struct btrfs_device *device;
2028         u64 type;
2029         u64 avail_space;
2030         u64 min_stripe_size;
2031         int num_stripes = 1;
2032         int i = 0, nr_devices;
2033         const struct btrfs_raid_attr *rattr;
2034
2035         /*
2036          * We aren't under the device list lock, so this is racy-ish, but good
2037          * enough for our purposes.
2038          */
2039         nr_devices = fs_info->fs_devices->open_devices;
2040         if (!nr_devices) {
2041                 smp_mb();
2042                 nr_devices = fs_info->fs_devices->open_devices;
2043                 ASSERT(nr_devices);
2044                 if (!nr_devices) {
2045                         *free_bytes = 0;
2046                         return 0;
2047                 }
2048         }
2049
2050         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2051                                GFP_KERNEL);
2052         if (!devices_info)
2053                 return -ENOMEM;
2054
2055         /* calc min stripe number for data space allocation */
2056         type = btrfs_data_alloc_profile(fs_info);
2057         rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2058
2059         if (type & BTRFS_BLOCK_GROUP_RAID0)
2060                 num_stripes = nr_devices;
2061         else if (type & BTRFS_BLOCK_GROUP_RAID1)
2062                 num_stripes = 2;
2063         else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2064                 num_stripes = 3;
2065         else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2066                 num_stripes = 4;
2067         else if (type & BTRFS_BLOCK_GROUP_RAID10)
2068                 num_stripes = 4;
2069
2070         /* Adjust for more than 1 stripe per device */
2071         min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2072
2073         rcu_read_lock();
2074         list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2075                 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2076                                                 &device->dev_state) ||
2077                     !device->bdev ||
2078                     test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2079                         continue;
2080
2081                 if (i >= nr_devices)
2082                         break;
2083
2084                 avail_space = device->total_bytes - device->bytes_used;
2085
2086                 /* align with stripe_len */
2087                 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2088
2089                 /*
2090                  * In order to avoid overwriting the superblock on the drive,
2091                  * btrfs starts at an offset of at least 1MB when doing chunk
2092                  * allocation.
2093                  *
2094                  * This ensures we have at least min_stripe_size free space
2095                  * after excluding 1MB.
2096                  */
2097                 if (avail_space <= SZ_1M + min_stripe_size)
2098                         continue;
2099
2100                 avail_space -= SZ_1M;
2101
2102                 devices_info[i].dev = device;
2103                 devices_info[i].max_avail = avail_space;
2104
2105                 i++;
2106         }
2107         rcu_read_unlock();
2108
2109         nr_devices = i;
2110
2111         btrfs_descending_sort_devices(devices_info, nr_devices);
2112
2113         i = nr_devices - 1;
2114         avail_space = 0;
2115         while (nr_devices >= rattr->devs_min) {
2116                 num_stripes = min(num_stripes, nr_devices);
2117
2118                 if (devices_info[i].max_avail >= min_stripe_size) {
2119                         int j;
2120                         u64 alloc_size;
2121
2122                         avail_space += devices_info[i].max_avail * num_stripes;
2123                         alloc_size = devices_info[i].max_avail;
2124                         for (j = i + 1 - num_stripes; j <= i; j++)
2125                                 devices_info[j].max_avail -= alloc_size;
2126                 }
2127                 i--;
2128                 nr_devices--;
2129         }
2130
2131         kfree(devices_info);
2132         *free_bytes = avail_space;
2133         return 0;
2134 }
2135
2136 /*
2137  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2138  *
2139  * If there's a redundant raid level at DATA block groups, use the respective
2140  * multiplier to scale the sizes.
2141  *
2142  * Unused device space usage is based on simulating the chunk allocator
2143  * algorithm that respects the device sizes and order of allocations.  This is
2144  * a close approximation of the actual use but there are other factors that may
2145  * change the result (like a new metadata chunk).
2146  *
2147  * If metadata is exhausted, f_bavail will be 0.
2148  */
2149 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2150 {
2151         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2152         struct btrfs_super_block *disk_super = fs_info->super_copy;
2153         struct btrfs_space_info *found;
2154         u64 total_used = 0;
2155         u64 total_free_data = 0;
2156         u64 total_free_meta = 0;
2157         int bits = dentry->d_sb->s_blocksize_bits;
2158         __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2159         unsigned factor = 1;
2160         struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2161         int ret;
2162         u64 thresh = 0;
2163         int mixed = 0;
2164
2165         rcu_read_lock();
2166         list_for_each_entry_rcu(found, &fs_info->space_info, list) {
2167                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2168                         int i;
2169
2170                         total_free_data += found->disk_total - found->disk_used;
2171                         total_free_data -=
2172                                 btrfs_account_ro_block_groups_free_space(found);
2173
2174                         for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2175                                 if (!list_empty(&found->block_groups[i]))
2176                                         factor = btrfs_bg_type_to_factor(
2177                                                 btrfs_raid_array[i].bg_flag);
2178                         }
2179                 }
2180
2181                 /*
2182                  * Metadata in mixed block goup profiles are accounted in data
2183                  */
2184                 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2185                         if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2186                                 mixed = 1;
2187                         else
2188                                 total_free_meta += found->disk_total -
2189                                         found->disk_used;
2190                 }
2191
2192                 total_used += found->disk_used;
2193         }
2194
2195         rcu_read_unlock();
2196
2197         buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2198         buf->f_blocks >>= bits;
2199         buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2200
2201         /* Account global block reserve as used, it's in logical size already */
2202         spin_lock(&block_rsv->lock);
2203         /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2204         if (buf->f_bfree >= block_rsv->size >> bits)
2205                 buf->f_bfree -= block_rsv->size >> bits;
2206         else
2207                 buf->f_bfree = 0;
2208         spin_unlock(&block_rsv->lock);
2209
2210         buf->f_bavail = div_u64(total_free_data, factor);
2211         ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2212         if (ret)
2213                 return ret;
2214         buf->f_bavail += div_u64(total_free_data, factor);
2215         buf->f_bavail = buf->f_bavail >> bits;
2216
2217         /*
2218          * We calculate the remaining metadata space minus global reserve. If
2219          * this is (supposedly) smaller than zero, there's no space. But this
2220          * does not hold in practice, the exhausted state happens where's still
2221          * some positive delta. So we apply some guesswork and compare the
2222          * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2223          *
2224          * We probably cannot calculate the exact threshold value because this
2225          * depends on the internal reservations requested by various
2226          * operations, so some operations that consume a few metadata will
2227          * succeed even if the Avail is zero. But this is better than the other
2228          * way around.
2229          */
2230         thresh = SZ_4M;
2231
2232         /*
2233          * We only want to claim there's no available space if we can no longer
2234          * allocate chunks for our metadata profile and our global reserve will
2235          * not fit in the free metadata space.  If we aren't ->full then we
2236          * still can allocate chunks and thus are fine using the currently
2237          * calculated f_bavail.
2238          */
2239         if (!mixed && block_rsv->space_info->full &&
2240             total_free_meta - thresh < block_rsv->size)
2241                 buf->f_bavail = 0;
2242
2243         buf->f_type = BTRFS_SUPER_MAGIC;
2244         buf->f_bsize = dentry->d_sb->s_blocksize;
2245         buf->f_namelen = BTRFS_NAME_LEN;
2246
2247         /* We treat it as constant endianness (it doesn't matter _which_)
2248            because we want the fsid to come out the same whether mounted
2249            on a big-endian or little-endian host */
2250         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2251         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2252         /* Mask in the root object ID too, to disambiguate subvols */
2253         buf->f_fsid.val[0] ^=
2254                 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2255         buf->f_fsid.val[1] ^=
2256                 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2257
2258         return 0;
2259 }
2260
2261 static void btrfs_kill_super(struct super_block *sb)
2262 {
2263         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2264         kill_anon_super(sb);
2265         btrfs_free_fs_info(fs_info);
2266 }
2267
2268 static struct file_system_type btrfs_fs_type = {
2269         .owner          = THIS_MODULE,
2270         .name           = "btrfs",
2271         .mount          = btrfs_mount,
2272         .kill_sb        = btrfs_kill_super,
2273         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2274 };
2275
2276 static struct file_system_type btrfs_root_fs_type = {
2277         .owner          = THIS_MODULE,
2278         .name           = "btrfs",
2279         .mount          = btrfs_mount_root,
2280         .kill_sb        = btrfs_kill_super,
2281         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2282 };
2283
2284 MODULE_ALIAS_FS("btrfs");
2285
2286 static int btrfs_control_open(struct inode *inode, struct file *file)
2287 {
2288         /*
2289          * The control file's private_data is used to hold the
2290          * transaction when it is started and is used to keep
2291          * track of whether a transaction is already in progress.
2292          */
2293         file->private_data = NULL;
2294         return 0;
2295 }
2296
2297 /*
2298  * Used by /dev/btrfs-control for devices ioctls.
2299  */
2300 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2301                                 unsigned long arg)
2302 {
2303         struct btrfs_ioctl_vol_args *vol;
2304         struct btrfs_device *device = NULL;
2305         int ret = -ENOTTY;
2306
2307         if (!capable(CAP_SYS_ADMIN))
2308                 return -EPERM;
2309
2310         vol = memdup_user((void __user *)arg, sizeof(*vol));
2311         if (IS_ERR(vol))
2312                 return PTR_ERR(vol);
2313         vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2314
2315         switch (cmd) {
2316         case BTRFS_IOC_SCAN_DEV:
2317                 mutex_lock(&uuid_mutex);
2318                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2319                                                &btrfs_root_fs_type);
2320                 ret = PTR_ERR_OR_ZERO(device);
2321                 mutex_unlock(&uuid_mutex);
2322                 break;
2323         case BTRFS_IOC_FORGET_DEV:
2324                 ret = btrfs_forget_devices(vol->name);
2325                 break;
2326         case BTRFS_IOC_DEVICES_READY:
2327                 mutex_lock(&uuid_mutex);
2328                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2329                                                &btrfs_root_fs_type);
2330                 if (IS_ERR(device)) {
2331                         mutex_unlock(&uuid_mutex);
2332                         ret = PTR_ERR(device);
2333                         break;
2334                 }
2335                 ret = !(device->fs_devices->num_devices ==
2336                         device->fs_devices->total_devices);
2337                 mutex_unlock(&uuid_mutex);
2338                 break;
2339         case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2340                 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2341                 break;
2342         }
2343
2344         kfree(vol);
2345         return ret;
2346 }
2347
2348 static int btrfs_freeze(struct super_block *sb)
2349 {
2350         struct btrfs_trans_handle *trans;
2351         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2352         struct btrfs_root *root = fs_info->tree_root;
2353
2354         set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2355         /*
2356          * We don't need a barrier here, we'll wait for any transaction that
2357          * could be in progress on other threads (and do delayed iputs that
2358          * we want to avoid on a frozen filesystem), or do the commit
2359          * ourselves.
2360          */
2361         trans = btrfs_attach_transaction_barrier(root);
2362         if (IS_ERR(trans)) {
2363                 /* no transaction, don't bother */
2364                 if (PTR_ERR(trans) == -ENOENT)
2365                         return 0;
2366                 return PTR_ERR(trans);
2367         }
2368         return btrfs_commit_transaction(trans);
2369 }
2370
2371 static int btrfs_unfreeze(struct super_block *sb)
2372 {
2373         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2374
2375         clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2376         return 0;
2377 }
2378
2379 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2380 {
2381         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2382         struct btrfs_device *dev, *first_dev = NULL;
2383
2384         /*
2385          * Lightweight locking of the devices. We should not need
2386          * device_list_mutex here as we only read the device data and the list
2387          * is protected by RCU.  Even if a device is deleted during the list
2388          * traversals, we'll get valid data, the freeing callback will wait at
2389          * least until the rcu_read_unlock.
2390          */
2391         rcu_read_lock();
2392         list_for_each_entry_rcu(dev, &fs_info->fs_devices->devices, dev_list) {
2393                 if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2394                         continue;
2395                 if (!dev->name)
2396                         continue;
2397                 if (!first_dev || dev->devid < first_dev->devid)
2398                         first_dev = dev;
2399         }
2400
2401         if (first_dev)
2402                 seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2403         else
2404                 WARN_ON(1);
2405         rcu_read_unlock();
2406         return 0;
2407 }
2408
2409 static const struct super_operations btrfs_super_ops = {
2410         .drop_inode     = btrfs_drop_inode,
2411         .evict_inode    = btrfs_evict_inode,
2412         .put_super      = btrfs_put_super,
2413         .sync_fs        = btrfs_sync_fs,
2414         .show_options   = btrfs_show_options,
2415         .show_devname   = btrfs_show_devname,
2416         .alloc_inode    = btrfs_alloc_inode,
2417         .destroy_inode  = btrfs_destroy_inode,
2418         .free_inode     = btrfs_free_inode,
2419         .statfs         = btrfs_statfs,
2420         .remount_fs     = btrfs_remount,
2421         .freeze_fs      = btrfs_freeze,
2422         .unfreeze_fs    = btrfs_unfreeze,
2423 };
2424
2425 static const struct file_operations btrfs_ctl_fops = {
2426         .open = btrfs_control_open,
2427         .unlocked_ioctl  = btrfs_control_ioctl,
2428         .compat_ioctl = compat_ptr_ioctl,
2429         .owner   = THIS_MODULE,
2430         .llseek = noop_llseek,
2431 };
2432
2433 static struct miscdevice btrfs_misc = {
2434         .minor          = BTRFS_MINOR,
2435         .name           = "btrfs-control",
2436         .fops           = &btrfs_ctl_fops
2437 };
2438
2439 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2440 MODULE_ALIAS("devname:btrfs-control");
2441
2442 static int __init btrfs_interface_init(void)
2443 {
2444         return misc_register(&btrfs_misc);
2445 }
2446
2447 static __cold void btrfs_interface_exit(void)
2448 {
2449         misc_deregister(&btrfs_misc);
2450 }
2451
2452 static void __init btrfs_print_mod_info(void)
2453 {
2454         static const char options[] = ""
2455 #ifdef CONFIG_BTRFS_DEBUG
2456                         ", debug=on"
2457 #endif
2458 #ifdef CONFIG_BTRFS_ASSERT
2459                         ", assert=on"
2460 #endif
2461 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2462                         ", integrity-checker=on"
2463 #endif
2464 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2465                         ", ref-verify=on"
2466 #endif
2467                         ;
2468         pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2469 }
2470
2471 static int __init init_btrfs_fs(void)
2472 {
2473         int err;
2474
2475         btrfs_props_init();
2476
2477         err = btrfs_init_sysfs();
2478         if (err)
2479                 return err;
2480
2481         btrfs_init_compress();
2482
2483         err = btrfs_init_cachep();
2484         if (err)
2485                 goto free_compress;
2486
2487         err = extent_io_init();
2488         if (err)
2489                 goto free_cachep;
2490
2491         err = extent_state_cache_init();
2492         if (err)
2493                 goto free_extent_io;
2494
2495         err = extent_map_init();
2496         if (err)
2497                 goto free_extent_state_cache;
2498
2499         err = ordered_data_init();
2500         if (err)
2501                 goto free_extent_map;
2502
2503         err = btrfs_delayed_inode_init();
2504         if (err)
2505                 goto free_ordered_data;
2506
2507         err = btrfs_auto_defrag_init();
2508         if (err)
2509                 goto free_delayed_inode;
2510
2511         err = btrfs_delayed_ref_init();
2512         if (err)
2513                 goto free_auto_defrag;
2514
2515         err = btrfs_prelim_ref_init();
2516         if (err)
2517                 goto free_delayed_ref;
2518
2519         err = btrfs_end_io_wq_init();
2520         if (err)
2521                 goto free_prelim_ref;
2522
2523         err = btrfs_interface_init();
2524         if (err)
2525                 goto free_end_io_wq;
2526
2527         btrfs_init_lockdep();
2528
2529         btrfs_print_mod_info();
2530
2531         err = btrfs_run_sanity_tests();
2532         if (err)
2533                 goto unregister_ioctl;
2534
2535         err = register_filesystem(&btrfs_fs_type);
2536         if (err)
2537                 goto unregister_ioctl;
2538
2539         return 0;
2540
2541 unregister_ioctl:
2542         btrfs_interface_exit();
2543 free_end_io_wq:
2544         btrfs_end_io_wq_exit();
2545 free_prelim_ref:
2546         btrfs_prelim_ref_exit();
2547 free_delayed_ref:
2548         btrfs_delayed_ref_exit();
2549 free_auto_defrag:
2550         btrfs_auto_defrag_exit();
2551 free_delayed_inode:
2552         btrfs_delayed_inode_exit();
2553 free_ordered_data:
2554         ordered_data_exit();
2555 free_extent_map:
2556         extent_map_exit();
2557 free_extent_state_cache:
2558         extent_state_cache_exit();
2559 free_extent_io:
2560         extent_io_exit();
2561 free_cachep:
2562         btrfs_destroy_cachep();
2563 free_compress:
2564         btrfs_exit_compress();
2565         btrfs_exit_sysfs();
2566
2567         return err;
2568 }
2569
2570 static void __exit exit_btrfs_fs(void)
2571 {
2572         btrfs_destroy_cachep();
2573         btrfs_delayed_ref_exit();
2574         btrfs_auto_defrag_exit();
2575         btrfs_delayed_inode_exit();
2576         btrfs_prelim_ref_exit();
2577         ordered_data_exit();
2578         extent_map_exit();
2579         extent_state_cache_exit();
2580         extent_io_exit();
2581         btrfs_interface_exit();
2582         btrfs_end_io_wq_exit();
2583         unregister_filesystem(&btrfs_fs_type);
2584         btrfs_exit_sysfs();
2585         btrfs_cleanup_fs_uuids();
2586         btrfs_exit_compress();
2587 }
2588
2589 late_initcall(init_btrfs_fs);
2590 module_exit(exit_btrfs_fs)
2591
2592 MODULE_LICENSE("GPL");
2593 MODULE_SOFTDEP("pre: crc32c");
2594 MODULE_SOFTDEP("pre: xxhash64");
2595 MODULE_SOFTDEP("pre: sha256");
2596 MODULE_SOFTDEP("pre: blake2b-256");