fs/btrfs/super.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2007 Oracle.  All rights reserved.
   4  */
   5
   6 #include <linux/blkdev.h>
   7 #include <linux/module.h>
   8 #include <linux/fs.h>
   9 #include <linux/pagemap.h>
  10 #include <linux/highmem.h>
  11 #include <linux/time.h>
  12 #include <linux/init.h>
  13 #include <linux/seq_file.h>
  14 #include <linux/string.h>
  15 #include <linux/backing-dev.h>
  16 #include <linux/mount.h>
  17 #include <linux/writeback.h>
  18 #include <linux/statfs.h>
  19 #include <linux/compat.h>
  20 #include <linux/parser.h>
  21 #include <linux/ctype.h>
  22 #include <linux/namei.h>
  23 #include <linux/miscdevice.h>
  24 #include <linux/magic.h>
  25 #include <linux/slab.h>
  26 #include <linux/ratelimit.h>
  27 #include <linux/crc32c.h>
  28 #include <linux/btrfs.h>
  29 #include "delayed-inode.h"
  30 #include "ctree.h"
  31 #include "disk-io.h"
  32 #include "transaction.h"
  33 #include "btrfs_inode.h"
  34 #include "print-tree.h"
  35 #include "props.h"
  36 #include "xattr.h"
  37 #include "volumes.h"
  38 #include "export.h"
  39 #include "compression.h"
  40 #include "rcu-string.h"
  41 #include "dev-replace.h"
  42 #include "free-space-cache.h"
  43 #include "backref.h"
  44 #include "space-info.h"
  45 #include "sysfs.h"
  46 #include "zoned.h"
  47 #include "tests/btrfs-tests.h"
  48 #include "block-group.h"
  49 #include "discard.h"
  50 #include "qgroup.h"
  51 #define CREATE_TRACE_POINTS
  52 #include <trace/events/btrfs.h>
  53
  54 static const struct super_operations btrfs_super_ops;
  55
  56 /*
  57  * Types for mounting the default subvolume and a subvolume explicitly
  58  * requested by subvol=/path. That way the callchain is straightforward and we
  59  * don't have to play tricks with the mount options and recursive calls to
  60  * btrfs_mount.
  61  *
  62  * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
  63  */
  64 static struct file_system_type btrfs_fs_type;
  65 static struct file_system_type btrfs_root_fs_type;
  66
  67 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
  68
  69 /*
  70  * Generally the error codes correspond to their respective errors, but there
  71  * are a few special cases.
  72  *
  73  * EUCLEAN: Any sort of corruption that we encounter.  The tree-checker for
  74  *          instance will return EUCLEAN if any of the blocks are corrupted in
  75  *          a way that is problematic.  We want to reserve EUCLEAN for these
  76  *          sort of corruptions.
  77  *
  78  * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
  79  *        need to use EROFS for this case.  We will have no idea of the
  80  *        original failure, that will have been reported at the time we tripped
  81  *        over the error.  Each subsequent error that doesn't have any context
  82  *        of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
  83  */
  84 const char * __attribute_const__ btrfs_decode_error(int errno)
  85 {
  86         char *errstr = "unknown";
  87
  88         switch (errno) {
  89         case -ENOENT:           /* -2 */
  90                 errstr = "No such entry";
  91                 break;
  92         case -EIO:              /* -5 */
  93                 errstr = "IO failure";
  94                 break;
  95         case -ENOMEM:           /* -12*/
  96                 errstr = "Out of memory";
  97                 break;
  98         case -EEXIST:           /* -17 */
  99                 errstr = "Object already exists";
 100                 break;
 101         case -ENOSPC:           /* -28 */
 102                 errstr = "No space left";
 103                 break;
 104         case -EROFS:            /* -30 */
 105                 errstr = "Readonly filesystem";
 106                 break;
 107         case -EOPNOTSUPP:       /* -95 */
 108                 errstr = "Operation not supported";
 109                 break;
 110         case -EUCLEAN:          /* -117 */
 111                 errstr = "Filesystem corrupted";
 112                 break;
 113         case -EDQUOT:           /* -122 */
 114                 errstr = "Quota exceeded";
 115                 break;
 116         }
 117
 118         return errstr;
 119 }
 120
 121 /*
 122  * __btrfs_handle_fs_error decodes expected errors from the caller and
 123  * invokes the appropriate error response.
 124  */
 125 __cold
 126 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
 127                        unsigned int line, int errno, const char *fmt, ...)
 128 {
 129         struct super_block *sb = fs_info->sb;
 130 #ifdef CONFIG_PRINTK
 131         const char *errstr;
 132 #endif
 133
 134         /*
 135          * Special case: if the error is EROFS, and we're already
 136          * under SB_RDONLY, then it is safe here.
 137          */
 138         if (errno == -EROFS && sb_rdonly(sb))
 139                 return;
 140
 141 #ifdef CONFIG_PRINTK
 142         errstr = btrfs_decode_error(errno);
 143         if (fmt) {
 144                 struct va_format vaf;
 145                 va_list args;
 146
 147                 va_start(args, fmt);
 148                 vaf.fmt = fmt;
 149                 vaf.va = &args;
 150
 151                 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
 152                         sb->s_id, function, line, errno, errstr, &vaf);
 153                 va_end(args);
 154         } else {
 155                 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
 156                         sb->s_id, function, line, errno, errstr);
 157         }
 158 #endif
 159
 160         /*
 161          * Today we only save the error info to memory.  Long term we'll
 162          * also send it down to the disk
 163          */
 164         set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
 165
 166         /* Don't go through full error handling during mount */
 167         if (!(sb->s_flags & SB_BORN))
 168                 return;
 169
 170         if (sb_rdonly(sb))
 171                 return;
 172
 173         btrfs_discard_stop(fs_info);
 174
 175         /* btrfs handle error by forcing the filesystem readonly */
 176         btrfs_set_sb_rdonly(sb);
 177         btrfs_info(fs_info, "forced readonly");
 178         /*
 179          * Note that a running device replace operation is not canceled here
 180          * although there is no way to update the progress. It would add the
 181          * risk of a deadlock, therefore the canceling is omitted. The only
 182          * penalty is that some I/O remains active until the procedure
 183          * completes. The next time when the filesystem is mounted writable
 184          * again, the device replace operation continues.
 185          */
 186 }
 187
 188 #ifdef CONFIG_PRINTK
 189 static const char * const logtypes[] = {
 190         "emergency",
 191         "alert",
 192         "critical",
 193         "error",
 194         "warning",
 195         "notice",
 196         "info",
 197         "debug",
 198 };
 199
 200
 201 /*
 202  * Use one ratelimit state per log level so that a flood of less important
 203  * messages doesn't cause more important ones to be dropped.
 204  */
 205 static struct ratelimit_state printk_limits[] = {
 206         RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
 207         RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
 208         RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
 209         RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
 210         RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
 211         RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
 212         RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
 213         RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
 214 };
 215
 216 void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 217 {
 218         char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 219         struct va_format vaf;
 220         va_list args;
 221         int kern_level;
 222         const char *type = logtypes[4];
 223         struct ratelimit_state *ratelimit = &printk_limits[4];
 224
 225         va_start(args, fmt);
 226
 227         while ((kern_level = printk_get_level(fmt)) != 0) {
 228                 size_t size = printk_skip_level(fmt) - fmt;
 229
 230                 if (kern_level >= '0' && kern_level <= '7') {
 231                         memcpy(lvl, fmt,  size);
 232                         lvl[size] = '\0';
 233                         type = logtypes[kern_level - '0'];
 234                         ratelimit = &printk_limits[kern_level - '0'];
 235                 }
 236                 fmt += size;
 237         }
 238
 239         vaf.fmt = fmt;
 240         vaf.va = &args;
 241
 242         if (__ratelimit(ratelimit)) {
 243                 if (fs_info)
 244                         printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
 245                                 fs_info->sb->s_id, &vaf);
 246                 else
 247                         printk("%sBTRFS %s: %pV\n", lvl, type, &vaf);
 248         }
 249
 250         va_end(args);
 251 }
 252 #endif
 253
 254 #if BITS_PER_LONG == 32
 255 void __cold btrfs_warn_32bit_limit(struct btrfs_fs_info *fs_info)
 256 {
 257         if (!test_and_set_bit(BTRFS_FS_32BIT_WARN, &fs_info->flags)) {
 258                 btrfs_warn(fs_info, "reaching 32bit limit for logical addresses");
 259                 btrfs_warn(fs_info,
 260 "due to page cache limit on 32bit systems, btrfs can't access metadata at or beyond %lluT",
 261                            BTRFS_32BIT_MAX_FILE_SIZE >> 40);
 262                 btrfs_warn(fs_info,
 263                            "please consider upgrading to 64bit kernel/hardware");
 264         }
 265 }
 266
 267 void __cold btrfs_err_32bit_limit(struct btrfs_fs_info *fs_info)
 268 {
 269         if (!test_and_set_bit(BTRFS_FS_32BIT_ERROR, &fs_info->flags)) {
 270                 btrfs_err(fs_info, "reached 32bit limit for logical addresses");
 271                 btrfs_err(fs_info,
 272 "due to page cache limit on 32bit systems, metadata beyond %lluT can't be accessed",
 273                           BTRFS_32BIT_MAX_FILE_SIZE >> 40);
 274                 btrfs_err(fs_info,
 275                            "please consider upgrading to 64bit kernel/hardware");
 276         }
 277 }
 278 #endif
 279
 280 /*
 281  * We only mark the transaction aborted and then set the file system read-only.
 282  * This will prevent new transactions from starting or trying to join this
 283  * one.
 284  *
 285  * This means that error recovery at the call site is limited to freeing
 286  * any local memory allocations and passing the error code up without
 287  * further cleanup. The transaction should complete as it normally would
 288  * in the call path but will return -EIO.
 289  *
 290  * We'll complete the cleanup in btrfs_end_transaction and
 291  * btrfs_commit_transaction.
 292  */
 293 __cold
 294 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 295                                const char *function,
 296                                unsigned int line, int errno)
 297 {
 298         struct btrfs_fs_info *fs_info = trans->fs_info;
 299
 300         WRITE_ONCE(trans->aborted, errno);
 301         WRITE_ONCE(trans->transaction->aborted, errno);
 302         /* Wake up anybody who may be waiting on this transaction */
 303         wake_up(&fs_info->transaction_wait);
 304         wake_up(&fs_info->transaction_blocked_wait);
 305         __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
 306 }
 307 /*
 308  * __btrfs_panic decodes unexpected, fatal errors from the caller,
 309  * issues an alert, and either panics or BUGs, depending on mount options.
 310  */
 311 __cold
 312 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 313                    unsigned int line, int errno, const char *fmt, ...)
 314 {
 315         char *s_id = "<unknown>";
 316         const char *errstr;
 317         struct va_format vaf = { .fmt = fmt };
 318         va_list args;
 319
 320         if (fs_info)
 321                 s_id = fs_info->sb->s_id;
 322
 323         va_start(args, fmt);
 324         vaf.va = &args;
 325
 326         errstr = btrfs_decode_error(errno);
 327         if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
 328                 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
 329                         s_id, function, line, &vaf, errno, errstr);
 330
 331         btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
 332                    function, line, &vaf, errno, errstr);
 333         va_end(args);
 334         /* Caller calls BUG() */
 335 }
 336
 337 static void btrfs_put_super(struct super_block *sb)
 338 {
 339         close_ctree(btrfs_sb(sb));
 340 }
 341
 342 enum {
 343         Opt_acl, Opt_noacl,
 344         Opt_clear_cache,
 345         Opt_commit_interval,
 346         Opt_compress,
 347         Opt_compress_force,
 348         Opt_compress_force_type,
 349         Opt_compress_type,
 350         Opt_degraded,
 351         Opt_device,
 352         Opt_fatal_errors,
 353         Opt_flushoncommit, Opt_noflushoncommit,
 354         Opt_max_inline,
 355         Opt_barrier, Opt_nobarrier,
 356         Opt_datacow, Opt_nodatacow,
 357         Opt_datasum, Opt_nodatasum,
 358         Opt_defrag, Opt_nodefrag,
 359         Opt_discard, Opt_nodiscard,
 360         Opt_discard_mode,
 361         Opt_norecovery,
 362         Opt_ratio,
 363         Opt_rescan_uuid_tree,
 364         Opt_skip_balance,
 365         Opt_space_cache, Opt_no_space_cache,
 366         Opt_space_cache_version,
 367         Opt_ssd, Opt_nossd,
 368         Opt_ssd_spread, Opt_nossd_spread,
 369         Opt_subvol,
 370         Opt_subvol_empty,
 371         Opt_subvolid,
 372         Opt_thread_pool,
 373         Opt_treelog, Opt_notreelog,
 374         Opt_user_subvol_rm_allowed,
 375
 376         /* Rescue options */
 377         Opt_rescue,
 378         Opt_usebackuproot,
 379         Opt_nologreplay,
 380         Opt_ignorebadroots,
 381         Opt_ignoredatacsums,
 382         Opt_rescue_all,
 383
 384         /* Deprecated options */
 385         Opt_recovery,
 386         Opt_inode_cache, Opt_noinode_cache,
 387
 388         /* Debugging options */
 389         Opt_check_integrity,
 390         Opt_check_integrity_including_extent_data,
 391         Opt_check_integrity_print_mask,
 392         Opt_enospc_debug, Opt_noenospc_debug,
 393 #ifdef CONFIG_BTRFS_DEBUG
 394         Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 395 #endif
 396 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 397         Opt_ref_verify,
 398 #endif
 399         Opt_err,
 400 };
 401
 402 static const match_table_t tokens = {
 403         {Opt_acl, "acl"},
 404         {Opt_noacl, "noacl"},
 405         {Opt_clear_cache, "clear_cache"},
 406         {Opt_commit_interval, "commit=%u"},
 407         {Opt_compress, "compress"},
 408         {Opt_compress_type, "compress=%s"},
 409         {Opt_compress_force, "compress-force"},
 410         {Opt_compress_force_type, "compress-force=%s"},
 411         {Opt_degraded, "degraded"},
 412         {Opt_device, "device=%s"},
 413         {Opt_fatal_errors, "fatal_errors=%s"},
 414         {Opt_flushoncommit, "flushoncommit"},
 415         {Opt_noflushoncommit, "noflushoncommit"},
 416         {Opt_inode_cache, "inode_cache"},
 417         {Opt_noinode_cache, "noinode_cache"},
 418         {Opt_max_inline, "max_inline=%s"},
 419         {Opt_barrier, "barrier"},
 420         {Opt_nobarrier, "nobarrier"},
 421         {Opt_datacow, "datacow"},
 422         {Opt_nodatacow, "nodatacow"},
 423         {Opt_datasum, "datasum"},
 424         {Opt_nodatasum, "nodatasum"},
 425         {Opt_defrag, "autodefrag"},
 426         {Opt_nodefrag, "noautodefrag"},
 427         {Opt_discard, "discard"},
 428         {Opt_discard_mode, "discard=%s"},
 429         {Opt_nodiscard, "nodiscard"},
 430         {Opt_norecovery, "norecovery"},
 431         {Opt_ratio, "metadata_ratio=%u"},
 432         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 433         {Opt_skip_balance, "skip_balance"},
 434         {Opt_space_cache, "space_cache"},
 435         {Opt_no_space_cache, "nospace_cache"},
 436         {Opt_space_cache_version, "space_cache=%s"},
 437         {Opt_ssd, "ssd"},
 438         {Opt_nossd, "nossd"},
 439         {Opt_ssd_spread, "ssd_spread"},
 440         {Opt_nossd_spread, "nossd_spread"},
 441         {Opt_subvol, "subvol=%s"},
 442         {Opt_subvol_empty, "subvol="},
 443         {Opt_subvolid, "subvolid=%s"},
 444         {Opt_thread_pool, "thread_pool=%u"},
 445         {Opt_treelog, "treelog"},
 446         {Opt_notreelog, "notreelog"},
 447         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 448
 449         /* Rescue options */
 450         {Opt_rescue, "rescue=%s"},
 451         /* Deprecated, with alias rescue=nologreplay */
 452         {Opt_nologreplay, "nologreplay"},
 453         /* Deprecated, with alias rescue=usebackuproot */
 454         {Opt_usebackuproot, "usebackuproot"},
 455
 456         /* Deprecated options */
 457         {Opt_recovery, "recovery"},
 458
 459         /* Debugging options */
 460         {Opt_check_integrity, "check_int"},
 461         {Opt_check_integrity_including_extent_data, "check_int_data"},
 462         {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
 463         {Opt_enospc_debug, "enospc_debug"},
 464         {Opt_noenospc_debug, "noenospc_debug"},
 465 #ifdef CONFIG_BTRFS_DEBUG
 466         {Opt_fragment_data, "fragment=data"},
 467         {Opt_fragment_metadata, "fragment=metadata"},
 468         {Opt_fragment_all, "fragment=all"},
 469 #endif
 470 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 471         {Opt_ref_verify, "ref_verify"},
 472 #endif
 473         {Opt_err, NULL},
 474 };
 475
 476 static const match_table_t rescue_tokens = {
 477         {Opt_usebackuproot, "usebackuproot"},
 478         {Opt_nologreplay, "nologreplay"},
 479         {Opt_ignorebadroots, "ignorebadroots"},
 480         {Opt_ignorebadroots, "ibadroots"},
 481         {Opt_ignoredatacsums, "ignoredatacsums"},
 482         {Opt_ignoredatacsums, "idatacsums"},
 483         {Opt_rescue_all, "all"},
 484         {Opt_err, NULL},
 485 };
 486
 487 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
 488                             const char *opt_name)
 489 {
 490         if (fs_info->mount_opt & opt) {
 491                 btrfs_err(fs_info, "%s must be used with ro mount option",
 492                           opt_name);
 493                 return true;
 494         }
 495         return false;
 496 }
 497
 498 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
 499 {
 500         char *opts;
 501         char *orig;
 502         char *p;
 503         substring_t args[MAX_OPT_ARGS];
 504         int ret = 0;
 505
 506         opts = kstrdup(options, GFP_KERNEL);
 507         if (!opts)
 508                 return -ENOMEM;
 509         orig = opts;
 510
 511         while ((p = strsep(&opts, ":")) != NULL) {
 512                 int token;
 513
 514                 if (!*p)
 515                         continue;
 516                 token = match_token(p, rescue_tokens, args);
 517                 switch (token){
 518                 case Opt_usebackuproot:
 519                         btrfs_info(info,
 520                                    "trying to use backup root at mount time");
 521                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 522                         break;
 523                 case Opt_nologreplay:
 524                         btrfs_set_and_info(info, NOLOGREPLAY,
 525                                            "disabling log replay at mount time");
 526                         break;
 527                 case Opt_ignorebadroots:
 528                         btrfs_set_and_info(info, IGNOREBADROOTS,
 529                                            "ignoring bad roots");
 530                         break;
 531                 case Opt_ignoredatacsums:
 532                         btrfs_set_and_info(info, IGNOREDATACSUMS,
 533                                            "ignoring data csums");
 534                         break;
 535                 case Opt_rescue_all:
 536                         btrfs_info(info, "enabling all of the rescue options");
 537                         btrfs_set_and_info(info, IGNOREDATACSUMS,
 538                                            "ignoring data csums");
 539                         btrfs_set_and_info(info, IGNOREBADROOTS,
 540                                            "ignoring bad roots");
 541                         btrfs_set_and_info(info, NOLOGREPLAY,
 542                                            "disabling log replay at mount time");
 543                         break;
 544                 case Opt_err:
 545                         btrfs_info(info, "unrecognized rescue option '%s'", p);
 546                         ret = -EINVAL;
 547                         goto out;
 548                 default:
 549                         break;
 550                 }
 551
 552         }
 553 out:
 554         kfree(orig);
 555         return ret;
 556 }
 557
 558 /*
 559  * Regular mount options parser.  Everything that is needed only when
 560  * reading in a new superblock is parsed here.
 561  * XXX JDM: This needs to be cleaned up for remount.
 562  */
 563 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 564                         unsigned long new_flags)
 565 {
 566         substring_t args[MAX_OPT_ARGS];
 567         char *p, *num;
 568         int intarg;
 569         int ret = 0;
 570         char *compress_type;
 571         bool compress_force = false;
 572         enum btrfs_compression_type saved_compress_type;
 573         int saved_compress_level;
 574         bool saved_compress_force;
 575         int no_compress = 0;
 576
 577         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
 578                 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
 579         else if (btrfs_free_space_cache_v1_active(info)) {
 580                 if (btrfs_is_zoned(info)) {
 581                         btrfs_info(info,
 582                         "zoned: clearing existing space cache");
 583                         btrfs_set_super_cache_generation(info->super_copy, 0);
 584                 } else {
 585                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 586                 }
 587         }
 588
 589         /*
 590          * Even the options are empty, we still need to do extra check
 591          * against new flags
 592          */
 593         if (!options)
 594                 goto check;
 595
 596         while ((p = strsep(&options, ",")) != NULL) {
 597                 int token;
 598                 if (!*p)
 599                         continue;
 600
 601                 token = match_token(p, tokens, args);
 602                 switch (token) {
 603                 case Opt_degraded:
 604                         btrfs_info(info, "allowing degraded mounts");
 605                         btrfs_set_opt(info->mount_opt, DEGRADED);
 606                         break;
 607                 case Opt_subvol:
 608                 case Opt_subvol_empty:
 609                 case Opt_subvolid:
 610                 case Opt_device:
 611                         /*
 612                          * These are parsed by btrfs_parse_subvol_options or
 613                          * btrfs_parse_device_options and can be ignored here.
 614                          */
 615                         break;
 616                 case Opt_nodatasum:
 617                         btrfs_set_and_info(info, NODATASUM,
 618                                            "setting nodatasum");
 619                         break;
 620                 case Opt_datasum:
 621                         if (btrfs_test_opt(info, NODATASUM)) {
 622                                 if (btrfs_test_opt(info, NODATACOW))
 623                                         btrfs_info(info,
 624                                                    "setting datasum, datacow enabled");
 625                                 else
 626                                         btrfs_info(info, "setting datasum");
 627                         }
 628                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 629                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 630                         break;
 631                 case Opt_nodatacow:
 632                         if (!btrfs_test_opt(info, NODATACOW)) {
 633                                 if (!btrfs_test_opt(info, COMPRESS) ||
 634                                     !btrfs_test_opt(info, FORCE_COMPRESS)) {
 635                                         btrfs_info(info,
 636                                                    "setting nodatacow, compression disabled");
 637                                 } else {
 638                                         btrfs_info(info, "setting nodatacow");
 639                                 }
 640                         }
 641                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 642                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 643                         btrfs_set_opt(info->mount_opt, NODATACOW);
 644                         btrfs_set_opt(info->mount_opt, NODATASUM);
 645                         break;
 646                 case Opt_datacow:
 647                         btrfs_clear_and_info(info, NODATACOW,
 648                                              "setting datacow");
 649                         break;
 650                 case Opt_compress_force:
 651                 case Opt_compress_force_type:
 652                         compress_force = true;
 653                         fallthrough;
 654                 case Opt_compress:
 655                 case Opt_compress_type:
 656                         saved_compress_type = btrfs_test_opt(info,
 657                                                              COMPRESS) ?
 658                                 info->compress_type : BTRFS_COMPRESS_NONE;
 659                         saved_compress_force =
 660                                 btrfs_test_opt(info, FORCE_COMPRESS);
 661                         saved_compress_level = info->compress_level;
 662                         if (token == Opt_compress ||
 663                             token == Opt_compress_force ||
 664                             strncmp(args[0].from, "zlib", 4) == 0) {
 665                                 compress_type = "zlib";
 666
 667                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 668                                 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 669                                 /*
 670                                  * args[0] contains uninitialized data since
 671                                  * for these tokens we don't expect any
 672                                  * parameter.
 673                                  */
 674                                 if (token != Opt_compress &&
 675                                     token != Opt_compress_force)
 676                                         info->compress_level =
 677                                           btrfs_compress_str2level(
 678                                                         BTRFS_COMPRESS_ZLIB,
 679                                                         args[0].from + 4);
 680                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 681                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 682                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 683                                 no_compress = 0;
 684                         } else if (strncmp(args[0].from, "lzo", 3) == 0) {
 685                                 compress_type = "lzo";
 686                                 info->compress_type = BTRFS_COMPRESS_LZO;
 687                                 info->compress_level = 0;
 688                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 689                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 690                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 691                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 692                                 no_compress = 0;
 693                         } else if (strncmp(args[0].from, "zstd", 4) == 0) {
 694                                 compress_type = "zstd";
 695                                 info->compress_type = BTRFS_COMPRESS_ZSTD;
 696                                 info->compress_level =
 697                                         btrfs_compress_str2level(
 698                                                          BTRFS_COMPRESS_ZSTD,
 699                                                          args[0].from + 4);
 700                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 701                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 702                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 703                                 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
 704                                 no_compress = 0;
 705                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 706                                 compress_type = "no";
 707                                 info->compress_level = 0;
 708                                 info->compress_type = 0;
 709                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 710                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 711                                 compress_force = false;
 712                                 no_compress++;
 713                         } else {
 714                                 ret = -EINVAL;
 715                                 goto out;
 716                         }
 717
 718                         if (compress_force) {
 719                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 720                         } else {
 721                                 /*
 722                                  * If we remount from compress-force=xxx to
 723                                  * compress=xxx, we need clear FORCE_COMPRESS
 724                                  * flag, otherwise, there is no way for users
 725                                  * to disable forcible compression separately.
 726                                  */
 727                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 728                         }
 729                         if (no_compress == 1) {
 730                                 btrfs_info(info, "use no compression");
 731                         } else if ((info->compress_type != saved_compress_type) ||
 732                                    (compress_force != saved_compress_force) ||
 733                                    (info->compress_level != saved_compress_level)) {
 734                                 btrfs_info(info, "%s %s compression, level %d",
 735                                            (compress_force) ? "force" : "use",
 736                                            compress_type, info->compress_level);
 737                         }
 738                         compress_force = false;
 739                         break;
 740                 case Opt_ssd:
 741                         btrfs_set_and_info(info, SSD,
 742                                            "enabling ssd optimizations");
 743                         btrfs_clear_opt(info->mount_opt, NOSSD);
 744                         break;
 745                 case Opt_ssd_spread:
 746                         btrfs_set_and_info(info, SSD,
 747                                            "enabling ssd optimizations");
 748                         btrfs_set_and_info(info, SSD_SPREAD,
 749                                            "using spread ssd allocation scheme");
 750                         btrfs_clear_opt(info->mount_opt, NOSSD);
 751                         break;
 752                 case Opt_nossd:
 753                         btrfs_set_opt(info->mount_opt, NOSSD);
 754                         btrfs_clear_and_info(info, SSD,
 755                                              "not using ssd optimizations");
 756                         fallthrough;
 757                 case Opt_nossd_spread:
 758                         btrfs_clear_and_info(info, SSD_SPREAD,
 759                                              "not using spread ssd allocation scheme");
 760                         break;
 761                 case Opt_barrier:
 762                         btrfs_clear_and_info(info, NOBARRIER,
 763                                              "turning on barriers");
 764                         break;
 765                 case Opt_nobarrier:
 766                         btrfs_set_and_info(info, NOBARRIER,
 767                                            "turning off barriers");
 768                         break;
 769                 case Opt_thread_pool:
 770                         ret = match_int(&args[0], &intarg);
 771                         if (ret) {
 772                                 goto out;
 773                         } else if (intarg == 0) {
 774                                 ret = -EINVAL;
 775                                 goto out;
 776                         }
 777                         info->thread_pool_size = intarg;
 778                         break;
 779                 case Opt_max_inline:
 780                         num = match_strdup(&args[0]);
 781                         if (num) {
 782                                 info->max_inline = memparse(num, NULL);
 783                                 kfree(num);
 784
 785                                 if (info->max_inline) {
 786                                         info->max_inline = min_t(u64,
 787                                                 info->max_inline,
 788                                                 info->sectorsize);
 789                                 }
 790                                 btrfs_info(info, "max_inline at %llu",
 791                                            info->max_inline);
 792                         } else {
 793                                 ret = -ENOMEM;
 794                                 goto out;
 795                         }
 796                         break;
 797                 case Opt_acl:
 798 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 799                         info->sb->s_flags |= SB_POSIXACL;
 800                         break;
 801 #else
 802                         btrfs_err(info, "support for ACL not compiled in!");
 803                         ret = -EINVAL;
 804                         goto out;
 805 #endif
 806                 case Opt_noacl:
 807                         info->sb->s_flags &= ~SB_POSIXACL;
 808                         break;
 809                 case Opt_notreelog:
 810                         btrfs_set_and_info(info, NOTREELOG,
 811                                            "disabling tree log");
 812                         break;
 813                 case Opt_treelog:
 814                         btrfs_clear_and_info(info, NOTREELOG,
 815                                              "enabling tree log");
 816                         break;
 817                 case Opt_norecovery:
 818                 case Opt_nologreplay:
 819                         btrfs_warn(info,
 820                 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 821                         btrfs_set_and_info(info, NOLOGREPLAY,
 822                                            "disabling log replay at mount time");
 823                         break;
 824                 case Opt_flushoncommit:
 825                         btrfs_set_and_info(info, FLUSHONCOMMIT,
 826                                            "turning on flush-on-commit");
 827                         break;
 828                 case Opt_noflushoncommit:
 829                         btrfs_clear_and_info(info, FLUSHONCOMMIT,
 830                                              "turning off flush-on-commit");
 831                         break;
 832                 case Opt_ratio:
 833                         ret = match_int(&args[0], &intarg);
 834                         if (ret)
 835                                 goto out;
 836                         info->metadata_ratio = intarg;
 837                         btrfs_info(info, "metadata ratio %u",
 838                                    info->metadata_ratio);
 839                         break;
 840                 case Opt_discard:
 841                 case Opt_discard_mode:
 842                         if (token == Opt_discard ||
 843                             strcmp(args[0].from, "sync") == 0) {
 844                                 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
 845                                 btrfs_set_and_info(info, DISCARD_SYNC,
 846                                                    "turning on sync discard");
 847                         } else if (strcmp(args[0].from, "async") == 0) {
 848                                 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
 849                                 btrfs_set_and_info(info, DISCARD_ASYNC,
 850                                                    "turning on async discard");
 851                         } else {
 852                                 ret = -EINVAL;
 853                                 goto out;
 854                         }
 855                         break;
 856                 case Opt_nodiscard:
 857                         btrfs_clear_and_info(info, DISCARD_SYNC,
 858                                              "turning off discard");
 859                         btrfs_clear_and_info(info, DISCARD_ASYNC,
 860                                              "turning off async discard");
 861                         break;
 862                 case Opt_space_cache:
 863                 case Opt_space_cache_version:
 864                         if (token == Opt_space_cache ||
 865                             strcmp(args[0].from, "v1") == 0) {
 866                                 btrfs_clear_opt(info->mount_opt,
 867                                                 FREE_SPACE_TREE);
 868                                 btrfs_set_and_info(info, SPACE_CACHE,
 869                                            "enabling disk space caching");
 870                         } else if (strcmp(args[0].from, "v2") == 0) {
 871                                 btrfs_clear_opt(info->mount_opt,
 872                                                 SPACE_CACHE);
 873                                 btrfs_set_and_info(info, FREE_SPACE_TREE,
 874                                                    "enabling free space tree");
 875                         } else {
 876                                 ret = -EINVAL;
 877                                 goto out;
 878                         }
 879                         break;
 880                 case Opt_rescan_uuid_tree:
 881                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 882                         break;
 883                 case Opt_no_space_cache:
 884                         if (btrfs_test_opt(info, SPACE_CACHE)) {
 885                                 btrfs_clear_and_info(info, SPACE_CACHE,
 886                                              "disabling disk space caching");
 887                         }
 888                         if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
 889                                 btrfs_clear_and_info(info, FREE_SPACE_TREE,
 890                                              "disabling free space tree");
 891                         }
 892                         break;
 893                 case Opt_inode_cache:
 894                 case Opt_noinode_cache:
 895                         btrfs_warn(info,
 896         "the 'inode_cache' option is deprecated and has no effect since 5.11");
 897                         break;
 898                 case Opt_clear_cache:
 899                         btrfs_set_and_info(info, CLEAR_CACHE,
 900                                            "force clearing of disk cache");
 901                         break;
 902                 case Opt_user_subvol_rm_allowed:
 903                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 904                         break;
 905                 case Opt_enospc_debug:
 906                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 907                         break;
 908                 case Opt_noenospc_debug:
 909                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 910                         break;
 911                 case Opt_defrag:
 912                         btrfs_set_and_info(info, AUTO_DEFRAG,
 913                                            "enabling auto defrag");
 914                         break;
 915                 case Opt_nodefrag:
 916                         btrfs_clear_and_info(info, AUTO_DEFRAG,
 917                                              "disabling auto defrag");
 918                         break;
 919                 case Opt_recovery:
 920                 case Opt_usebackuproot:
 921                         btrfs_warn(info,
 922                         "'%s' is deprecated, use 'rescue=usebackuproot' instead",
 923                                    token == Opt_recovery ? "recovery" :
 924                                    "usebackuproot");
 925                         btrfs_info(info,
 926                                    "trying to use backup root at mount time");
 927                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 928                         break;
 929                 case Opt_skip_balance:
 930                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 931                         break;
 932 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 933                 case Opt_check_integrity_including_extent_data:
 934                         btrfs_info(info,
 935                                    "enabling check integrity including extent data");
 936                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
 937                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 938                         break;
 939                 case Opt_check_integrity:
 940                         btrfs_info(info, "enabling check integrity");
 941                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 942                         break;
 943                 case Opt_check_integrity_print_mask:
 944                         ret = match_int(&args[0], &intarg);
 945                         if (ret)
 946                                 goto out;
 947                         info->check_integrity_print_mask = intarg;
 948                         btrfs_info(info, "check_integrity_print_mask 0x%x",
 949                                    info->check_integrity_print_mask);
 950                         break;
 951 #else
 952                 case Opt_check_integrity_including_extent_data:
 953                 case Opt_check_integrity:
 954                 case Opt_check_integrity_print_mask:
 955                         btrfs_err(info,
 956                                   "support for check_integrity* not compiled in!");
 957                         ret = -EINVAL;
 958                         goto out;
 959 #endif
 960                 case Opt_fatal_errors:
 961                         if (strcmp(args[0].from, "panic") == 0)
 962                                 btrfs_set_opt(info->mount_opt,
 963                                               PANIC_ON_FATAL_ERROR);
 964                         else if (strcmp(args[0].from, "bug") == 0)
 965                                 btrfs_clear_opt(info->mount_opt,
 966                                               PANIC_ON_FATAL_ERROR);
 967                         else {
 968                                 ret = -EINVAL;
 969                                 goto out;
 970                         }
 971                         break;
 972                 case Opt_commit_interval:
 973                         intarg = 0;
 974                         ret = match_int(&args[0], &intarg);
 975                         if (ret)
 976                                 goto out;
 977                         if (intarg == 0) {
 978                                 btrfs_info(info,
 979                                            "using default commit interval %us",
 980                                            BTRFS_DEFAULT_COMMIT_INTERVAL);
 981                                 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
 982                         } else if (intarg > 300) {
 983                                 btrfs_warn(info, "excessive commit interval %d",
 984                                            intarg);
 985                         }
 986                         info->commit_interval = intarg;
 987                         break;
 988                 case Opt_rescue:
 989                         ret = parse_rescue_options(info, args[0].from);
 990                         if (ret < 0)
 991                                 goto out;
 992                         break;
 993 #ifdef CONFIG_BTRFS_DEBUG
 994                 case Opt_fragment_all:
 995                         btrfs_info(info, "fragmenting all space");
 996                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 997                         btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
 998                         break;
 999                 case Opt_fragment_metadata:
1000                         btrfs_info(info, "fragmenting metadata");
1001                         btrfs_set_opt(info->mount_opt,
1002                                       FRAGMENT_METADATA);
1003                         break;
1004                 case Opt_fragment_data:
1005                         btrfs_info(info, "fragmenting data");
1006                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1007                         break;
1008 #endif
1009 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
1010                 case Opt_ref_verify:
1011                         btrfs_info(info, "doing ref verification");
1012                         btrfs_set_opt(info->mount_opt, REF_VERIFY);
1013                         break;
1014 #endif
1015                 case Opt_err:
1016                         btrfs_err(info, "unrecognized mount option '%s'", p);
1017                         ret = -EINVAL;
1018                         goto out;
1019                 default:
1020                         break;
1021                 }
1022         }
1023 check:
1024         /* We're read-only, don't have to check. */
1025         if (new_flags & SB_RDONLY)
1026                 goto out;
1027
1028         if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
1029             check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
1030             check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
1031                 ret = -EINVAL;
1032 out:
1033         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1034             !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1035             !btrfs_test_opt(info, CLEAR_CACHE)) {
1036                 btrfs_err(info, "cannot disable free space tree");
1037                 ret = -EINVAL;
1038
1039         }
1040         if (!ret)
1041                 ret = btrfs_check_mountopts_zoned(info);
1042         if (!ret && btrfs_test_opt(info, SPACE_CACHE))
1043                 btrfs_info(info, "disk space caching is enabled");
1044         if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
1045                 btrfs_info(info, "using free space tree");
1046         return ret;
1047 }
1048
1049 /*
1050  * Parse mount options that are required early in the mount process.
1051  *
1052  * All other options will be parsed on much later in the mount process and
1053  * only when we need to allocate a new super block.
1054  */
1055 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1056                                       void *holder)
1057 {
1058         substring_t args[MAX_OPT_ARGS];
1059         char *device_name, *opts, *orig, *p;
1060         struct btrfs_device *device = NULL;
1061         int error = 0;
1062
1063         lockdep_assert_held(&uuid_mutex);
1064
1065         if (!options)
1066                 return 0;
1067
1068         /*
1069          * strsep changes the string, duplicate it because btrfs_parse_options
1070          * gets called later
1071          */
1072         opts = kstrdup(options, GFP_KERNEL);
1073         if (!opts)
1074                 return -ENOMEM;
1075         orig = opts;
1076
1077         while ((p = strsep(&opts, ",")) != NULL) {
1078                 int token;
1079
1080                 if (!*p)
1081                         continue;
1082
1083                 token = match_token(p, tokens, args);
1084                 if (token == Opt_device) {
1085                         device_name = match_strdup(&args[0]);
1086                         if (!device_name) {
1087                                 error = -ENOMEM;
1088                                 goto out;
1089                         }
1090                         device = btrfs_scan_one_device(device_name, flags,
1091                                         holder);
1092                         kfree(device_name);
1093                         if (IS_ERR(device)) {
1094                                 error = PTR_ERR(device);
1095                                 goto out;
1096                         }
1097                 }
1098         }
1099
1100 out:
1101         kfree(orig);
1102         return error;
1103 }
1104
1105 /*
1106  * Parse mount options that are related to subvolume id
1107  *
1108  * The value is later passed to mount_subvol()
1109  */
1110 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1111                 u64 *subvol_objectid)
1112 {
1113         substring_t args[MAX_OPT_ARGS];
1114         char *opts, *orig, *p;
1115         int error = 0;
1116         u64 subvolid;
1117
1118         if (!options)
1119                 return 0;
1120
1121         /*
1122          * strsep changes the string, duplicate it because
1123          * btrfs_parse_device_options gets called later
1124          */
1125         opts = kstrdup(options, GFP_KERNEL);
1126         if (!opts)
1127                 return -ENOMEM;
1128         orig = opts;
1129
1130         while ((p = strsep(&opts, ",")) != NULL) {
1131                 int token;
1132                 if (!*p)
1133                         continue;
1134
1135                 token = match_token(p, tokens, args);
1136                 switch (token) {
1137                 case Opt_subvol:
1138                         kfree(*subvol_name);
1139                         *subvol_name = match_strdup(&args[0]);
1140                         if (!*subvol_name) {
1141                                 error = -ENOMEM;
1142                                 goto out;
1143                         }
1144                         break;
1145                 case Opt_subvolid:
1146                         error = match_u64(&args[0], &subvolid);
1147                         if (error)
1148                                 goto out;
1149
1150                         /* we want the original fs_tree */
1151                         if (subvolid == 0)
1152                                 subvolid = BTRFS_FS_TREE_OBJECTID;
1153
1154                         *subvol_objectid = subvolid;
1155                         break;
1156                 default:
1157                         break;
1158                 }
1159         }
1160
1161 out:
1162         kfree(orig);
1163         return error;
1164 }
1165
1166 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1167                                           u64 subvol_objectid)
1168 {
1169         struct btrfs_root *root = fs_info->tree_root;
1170         struct btrfs_root *fs_root = NULL;
1171         struct btrfs_root_ref *root_ref;
1172         struct btrfs_inode_ref *inode_ref;
1173         struct btrfs_key key;
1174         struct btrfs_path *path = NULL;
1175         char *name = NULL, *ptr;
1176         u64 dirid;
1177         int len;
1178         int ret;
1179
1180         path = btrfs_alloc_path();
1181         if (!path) {
1182                 ret = -ENOMEM;
1183                 goto err;
1184         }
1185
1186         name = kmalloc(PATH_MAX, GFP_KERNEL);
1187         if (!name) {
1188                 ret = -ENOMEM;
1189                 goto err;
1190         }
1191         ptr = name + PATH_MAX - 1;
1192         ptr[0] = '\0';
1193
1194         /*
1195          * Walk up the subvolume trees in the tree of tree roots by root
1196          * backrefs until we hit the top-level subvolume.
1197          */
1198         while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1199                 key.objectid = subvol_objectid;
1200                 key.type = BTRFS_ROOT_BACKREF_KEY;
1201                 key.offset = (u64)-1;
1202
1203                 ret = btrfs_search_backwards(root, &key, path);
1204                 if (ret < 0) {
1205                         goto err;
1206                 } else if (ret > 0) {
1207                         ret = -ENOENT;
1208                         goto err;
1209                 }
1210
1211                 subvol_objectid = key.offset;
1212
1213                 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1214                                           struct btrfs_root_ref);
1215                 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1216                 ptr -= len + 1;
1217                 if (ptr < name) {
1218                         ret = -ENAMETOOLONG;
1219                         goto err;
1220                 }
1221                 read_extent_buffer(path->nodes[0], ptr + 1,
1222                                    (unsigned long)(root_ref + 1), len);
1223                 ptr[0] = '/';
1224                 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1225                 btrfs_release_path(path);
1226
1227                 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1228                 if (IS_ERR(fs_root)) {
1229                         ret = PTR_ERR(fs_root);
1230                         fs_root = NULL;
1231                         goto err;
1232                 }
1233
1234                 /*
1235                  * Walk up the filesystem tree by inode refs until we hit the
1236                  * root directory.
1237                  */
1238                 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1239                         key.objectid = dirid;
1240                         key.type = BTRFS_INODE_REF_KEY;
1241                         key.offset = (u64)-1;
1242
1243                         ret = btrfs_search_backwards(fs_root, &key, path);
1244                         if (ret < 0) {
1245                                 goto err;
1246                         } else if (ret > 0) {
1247                                 ret = -ENOENT;
1248                                 goto err;
1249                         }
1250
1251                         dirid = key.offset;
1252
1253                         inode_ref = btrfs_item_ptr(path->nodes[0],
1254                                                    path->slots[0],
1255                                                    struct btrfs_inode_ref);
1256                         len = btrfs_inode_ref_name_len(path->nodes[0],
1257                                                        inode_ref);
1258                         ptr -= len + 1;
1259                         if (ptr < name) {
1260                                 ret = -ENAMETOOLONG;
1261                                 goto err;
1262                         }
1263                         read_extent_buffer(path->nodes[0], ptr + 1,
1264                                            (unsigned long)(inode_ref + 1), len);
1265                         ptr[0] = '/';
1266                         btrfs_release_path(path);
1267                 }
1268                 btrfs_put_root(fs_root);
1269                 fs_root = NULL;
1270         }
1271
1272         btrfs_free_path(path);
1273         if (ptr == name + PATH_MAX - 1) {
1274                 name[0] = '/';
1275                 name[1] = '\0';
1276         } else {
1277                 memmove(name, ptr, name + PATH_MAX - ptr);
1278         }
1279         return name;
1280
1281 err:
1282         btrfs_put_root(fs_root);
1283         btrfs_free_path(path);
1284         kfree(name);
1285         return ERR_PTR(ret);
1286 }
1287
1288 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1289 {
1290         struct btrfs_root *root = fs_info->tree_root;
1291         struct btrfs_dir_item *di;
1292         struct btrfs_path *path;
1293         struct btrfs_key location;
1294         u64 dir_id;
1295
1296         path = btrfs_alloc_path();
1297         if (!path)
1298                 return -ENOMEM;
1299
1300         /*
1301          * Find the "default" dir item which points to the root item that we
1302          * will mount by default if we haven't been given a specific subvolume
1303          * to mount.
1304          */
1305         dir_id = btrfs_super_root_dir(fs_info->super_copy);
1306         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1307         if (IS_ERR(di)) {
1308                 btrfs_free_path(path);
1309                 return PTR_ERR(di);
1310         }
1311         if (!di) {
1312                 /*
1313                  * Ok the default dir item isn't there.  This is weird since
1314                  * it's always been there, but don't freak out, just try and
1315                  * mount the top-level subvolume.
1316                  */
1317                 btrfs_free_path(path);
1318                 *objectid = BTRFS_FS_TREE_OBJECTID;
1319                 return 0;
1320         }
1321
1322         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1323         btrfs_free_path(path);
1324         *objectid = location.objectid;
1325         return 0;
1326 }
1327
1328 static int btrfs_fill_super(struct super_block *sb,
1329                             struct btrfs_fs_devices *fs_devices,
1330                             void *data)
1331 {
1332         struct inode *inode;
1333         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1334         int err;
1335
1336         sb->s_maxbytes = MAX_LFS_FILESIZE;
1337         sb->s_magic = BTRFS_SUPER_MAGIC;
1338         sb->s_op = &btrfs_super_ops;
1339         sb->s_d_op = &btrfs_dentry_operations;
1340         sb->s_export_op = &btrfs_export_ops;
1341 #ifdef CONFIG_FS_VERITY
1342         sb->s_vop = &btrfs_verityops;
1343 #endif
1344         sb->s_xattr = btrfs_xattr_handlers;
1345         sb->s_time_gran = 1;
1346 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1347         sb->s_flags |= SB_POSIXACL;
1348 #endif
1349         sb->s_flags |= SB_I_VERSION;
1350         sb->s_iflags |= SB_I_CGROUPWB;
1351
1352         err = super_setup_bdi(sb);
1353         if (err) {
1354                 btrfs_err(fs_info, "super_setup_bdi failed");
1355                 return err;
1356         }
1357
1358         err = open_ctree(sb, fs_devices, (char *)data);
1359         if (err) {
1360                 btrfs_err(fs_info, "open_ctree failed");
1361                 return err;
1362         }
1363
1364         inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1365         if (IS_ERR(inode)) {
1366                 err = PTR_ERR(inode);
1367                 goto fail_close;
1368         }
1369
1370         sb->s_root = d_make_root(inode);
1371         if (!sb->s_root) {
1372                 err = -ENOMEM;
1373                 goto fail_close;
1374         }
1375
1376         sb->s_flags |= SB_ACTIVE;
1377         return 0;
1378
1379 fail_close:
1380         close_ctree(fs_info);
1381         return err;
1382 }
1383
1384 int btrfs_sync_fs(struct super_block *sb, int wait)
1385 {
1386         struct btrfs_trans_handle *trans;
1387         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1388         struct btrfs_root *root = fs_info->tree_root;
1389
1390         trace_btrfs_sync_fs(fs_info, wait);
1391
1392         if (!wait) {
1393                 filemap_flush(fs_info->btree_inode->i_mapping);
1394                 return 0;
1395         }
1396
1397         btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1398
1399         trans = btrfs_attach_transaction_barrier(root);
1400         if (IS_ERR(trans)) {
1401                 /* no transaction, don't bother */
1402                 if (PTR_ERR(trans) == -ENOENT) {
1403                         /*
1404                          * Exit unless we have some pending changes
1405                          * that need to go through commit
1406                          */
1407                         if (fs_info->pending_changes == 0)
1408                                 return 0;
1409                         /*
1410                          * A non-blocking test if the fs is frozen. We must not
1411                          * start a new transaction here otherwise a deadlock
1412                          * happens. The pending operations are delayed to the
1413                          * next commit after thawing.
1414                          */
1415                         if (sb_start_write_trylock(sb))
1416                                 sb_end_write(sb);
1417                         else
1418                                 return 0;
1419                         trans = btrfs_start_transaction(root, 0);
1420                 }
1421                 if (IS_ERR(trans))
1422                         return PTR_ERR(trans);
1423         }
1424         return btrfs_commit_transaction(trans);
1425 }
1426
1427 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1428 {
1429         seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1430         *printed = true;
1431 }
1432
1433 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1434 {
1435         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1436         const char *compress_type;
1437         const char *subvol_name;
1438         bool printed = false;
1439
1440         if (btrfs_test_opt(info, DEGRADED))
1441                 seq_puts(seq, ",degraded");
1442         if (btrfs_test_opt(info, NODATASUM))
1443                 seq_puts(seq, ",nodatasum");
1444         if (btrfs_test_opt(info, NODATACOW))
1445                 seq_puts(seq, ",nodatacow");
1446         if (btrfs_test_opt(info, NOBARRIER))
1447                 seq_puts(seq, ",nobarrier");
1448         if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1449                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1450         if (info->thread_pool_size !=  min_t(unsigned long,
1451                                              num_online_cpus() + 2, 8))
1452                 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1453         if (btrfs_test_opt(info, COMPRESS)) {
1454                 compress_type = btrfs_compress_type2str(info->compress_type);
1455                 if (btrfs_test_opt(info, FORCE_COMPRESS))
1456                         seq_printf(seq, ",compress-force=%s", compress_type);
1457                 else
1458                         seq_printf(seq, ",compress=%s", compress_type);
1459                 if (info->compress_level)
1460                         seq_printf(seq, ":%d", info->compress_level);
1461         }
1462         if (btrfs_test_opt(info, NOSSD))
1463                 seq_puts(seq, ",nossd");
1464         if (btrfs_test_opt(info, SSD_SPREAD))
1465                 seq_puts(seq, ",ssd_spread");
1466         else if (btrfs_test_opt(info, SSD))
1467                 seq_puts(seq, ",ssd");
1468         if (btrfs_test_opt(info, NOTREELOG))
1469                 seq_puts(seq, ",notreelog");
1470         if (btrfs_test_opt(info, NOLOGREPLAY))
1471                 print_rescue_option(seq, "nologreplay", &printed);
1472         if (btrfs_test_opt(info, USEBACKUPROOT))
1473                 print_rescue_option(seq, "usebackuproot", &printed);
1474         if (btrfs_test_opt(info, IGNOREBADROOTS))
1475                 print_rescue_option(seq, "ignorebadroots", &printed);
1476         if (btrfs_test_opt(info, IGNOREDATACSUMS))
1477                 print_rescue_option(seq, "ignoredatacsums", &printed);
1478         if (btrfs_test_opt(info, FLUSHONCOMMIT))
1479                 seq_puts(seq, ",flushoncommit");
1480         if (btrfs_test_opt(info, DISCARD_SYNC))
1481                 seq_puts(seq, ",discard");
1482         if (btrfs_test_opt(info, DISCARD_ASYNC))
1483                 seq_puts(seq, ",discard=async");
1484         if (!(info->sb->s_flags & SB_POSIXACL))
1485                 seq_puts(seq, ",noacl");
1486         if (btrfs_free_space_cache_v1_active(info))
1487                 seq_puts(seq, ",space_cache");
1488         else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1489                 seq_puts(seq, ",space_cache=v2");
1490         else
1491                 seq_puts(seq, ",nospace_cache");
1492         if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1493                 seq_puts(seq, ",rescan_uuid_tree");
1494         if (btrfs_test_opt(info, CLEAR_CACHE))
1495                 seq_puts(seq, ",clear_cache");
1496         if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1497                 seq_puts(seq, ",user_subvol_rm_allowed");
1498         if (btrfs_test_opt(info, ENOSPC_DEBUG))
1499                 seq_puts(seq, ",enospc_debug");
1500         if (btrfs_test_opt(info, AUTO_DEFRAG))
1501                 seq_puts(seq, ",autodefrag");
1502         if (btrfs_test_opt(info, SKIP_BALANCE))
1503                 seq_puts(seq, ",skip_balance");
1504 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1505         if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1506                 seq_puts(seq, ",check_int_data");
1507         else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1508                 seq_puts(seq, ",check_int");
1509         if (info->check_integrity_print_mask)
1510                 seq_printf(seq, ",check_int_print_mask=%d",
1511                                 info->check_integrity_print_mask);
1512 #endif
1513         if (info->metadata_ratio)
1514                 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1515         if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1516                 seq_puts(seq, ",fatal_errors=panic");
1517         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1518                 seq_printf(seq, ",commit=%u", info->commit_interval);
1519 #ifdef CONFIG_BTRFS_DEBUG
1520         if (btrfs_test_opt(info, FRAGMENT_DATA))
1521                 seq_puts(seq, ",fragment=data");
1522         if (btrfs_test_opt(info, FRAGMENT_METADATA))
1523                 seq_puts(seq, ",fragment=metadata");
1524 #endif
1525         if (btrfs_test_opt(info, REF_VERIFY))
1526                 seq_puts(seq, ",ref_verify");
1527         seq_printf(seq, ",subvolid=%llu",
1528                   BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1529         subvol_name = btrfs_get_subvol_name_from_objectid(info,
1530                         BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1531         if (!IS_ERR(subvol_name)) {
1532                 seq_puts(seq, ",subvol=");
1533                 seq_escape(seq, subvol_name, " \t\n\\");
1534                 kfree(subvol_name);
1535         }
1536         return 0;
1537 }
1538
1539 static int btrfs_test_super(struct super_block *s, void *data)
1540 {
1541         struct btrfs_fs_info *p = data;
1542         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1543
1544         return fs_info->fs_devices == p->fs_devices;
1545 }
1546
1547 static int btrfs_set_super(struct super_block *s, void *data)
1548 {
1549         int err = set_anon_super(s, data);
1550         if (!err)
1551                 s->s_fs_info = data;
1552         return err;
1553 }
1554
1555 /*
1556  * subvolumes are identified by ino 256
1557  */
1558 static inline int is_subvolume_inode(struct inode *inode)
1559 {
1560         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1561                 return 1;
1562         return 0;
1563 }
1564
1565 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1566                                    struct vfsmount *mnt)
1567 {
1568         struct dentry *root;
1569         int ret;
1570
1571         if (!subvol_name) {
1572                 if (!subvol_objectid) {
1573                         ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1574                                                           &subvol_objectid);
1575                         if (ret) {
1576                                 root = ERR_PTR(ret);
1577                                 goto out;
1578                         }
1579                 }
1580                 subvol_name = btrfs_get_subvol_name_from_objectid(
1581                                         btrfs_sb(mnt->mnt_sb), subvol_objectid);
1582                 if (IS_ERR(subvol_name)) {
1583                         root = ERR_CAST(subvol_name);
1584                         subvol_name = NULL;
1585                         goto out;
1586                 }
1587
1588         }
1589
1590         root = mount_subtree(mnt, subvol_name);
1591         /* mount_subtree() drops our reference on the vfsmount. */
1592         mnt = NULL;
1593
1594         if (!IS_ERR(root)) {
1595                 struct super_block *s = root->d_sb;
1596                 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1597                 struct inode *root_inode = d_inode(root);
1598                 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1599
1600                 ret = 0;
1601                 if (!is_subvolume_inode(root_inode)) {
1602                         btrfs_err(fs_info, "'%s' is not a valid subvolume",
1603                                subvol_name);
1604                         ret = -EINVAL;
1605                 }
1606                 if (subvol_objectid && root_objectid != subvol_objectid) {
1607                         /*
1608                          * This will also catch a race condition where a
1609                          * subvolume which was passed by ID is renamed and
1610                          * another subvolume is renamed over the old location.
1611                          */
1612                         btrfs_err(fs_info,
1613                                   "subvol '%s' does not match subvolid %llu",
1614                                   subvol_name, subvol_objectid);
1615                         ret = -EINVAL;
1616                 }
1617                 if (ret) {
1618                         dput(root);
1619                         root = ERR_PTR(ret);
1620                         deactivate_locked_super(s);
1621                 }
1622         }
1623
1624 out:
1625         mntput(mnt);
1626         kfree(subvol_name);
1627         return root;
1628 }
1629
1630 /*
1631  * Find a superblock for the given device / mount point.
1632  *
1633  * Note: This is based on mount_bdev from fs/super.c with a few additions
1634  *       for multiple device setup.  Make sure to keep it in sync.
1635  */
1636 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1637                 int flags, const char *device_name, void *data)
1638 {
1639         struct block_device *bdev = NULL;
1640         struct super_block *s;
1641         struct btrfs_device *device = NULL;
1642         struct btrfs_fs_devices *fs_devices = NULL;
1643         struct btrfs_fs_info *fs_info = NULL;
1644         void *new_sec_opts = NULL;
1645         fmode_t mode = FMODE_READ;
1646         int error = 0;
1647
1648         if (!(flags & SB_RDONLY))
1649                 mode |= FMODE_WRITE;
1650
1651         if (data) {
1652                 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1653                 if (error)
1654                         return ERR_PTR(error);
1655         }
1656
1657         /*
1658          * Setup a dummy root and fs_info for test/set super.  This is because
1659          * we don't actually fill this stuff out until open_ctree, but we need
1660          * then open_ctree will properly initialize the file system specific
1661          * settings later.  btrfs_init_fs_info initializes the static elements
1662          * of the fs_info (locks and such) to make cleanup easier if we find a
1663          * superblock with our given fs_devices later on at sget() time.
1664          */
1665         fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1666         if (!fs_info) {
1667                 error = -ENOMEM;
1668                 goto error_sec_opts;
1669         }
1670         btrfs_init_fs_info(fs_info);
1671
1672         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1673         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1674         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1675                 error = -ENOMEM;
1676                 goto error_fs_info;
1677         }
1678
1679         mutex_lock(&uuid_mutex);
1680         error = btrfs_parse_device_options(data, mode, fs_type);
1681         if (error) {
1682                 mutex_unlock(&uuid_mutex);
1683                 goto error_fs_info;
1684         }
1685
1686         device = btrfs_scan_one_device(device_name, mode, fs_type);
1687         if (IS_ERR(device)) {
1688                 mutex_unlock(&uuid_mutex);
1689                 error = PTR_ERR(device);
1690                 goto error_fs_info;
1691         }
1692
1693         fs_devices = device->fs_devices;
1694         fs_info->fs_devices = fs_devices;
1695
1696         error = btrfs_open_devices(fs_devices, mode, fs_type);
1697         mutex_unlock(&uuid_mutex);
1698         if (error)
1699                 goto error_fs_info;
1700
1701         if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1702                 error = -EACCES;
1703                 goto error_close_devices;
1704         }
1705
1706         bdev = fs_devices->latest_dev->bdev;
1707         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1708                  fs_info);
1709         if (IS_ERR(s)) {
1710                 error = PTR_ERR(s);
1711                 goto error_close_devices;
1712         }
1713
1714         if (s->s_root) {
1715                 btrfs_close_devices(fs_devices);
1716                 btrfs_free_fs_info(fs_info);
1717                 if ((flags ^ s->s_flags) & SB_RDONLY)
1718                         error = -EBUSY;
1719         } else {
1720                 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1721                 btrfs_sb(s)->bdev_holder = fs_type;
1722                 if (!strstr(crc32c_impl(), "generic"))
1723                         set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1724                 error = btrfs_fill_super(s, fs_devices, data);
1725         }
1726         if (!error)
1727                 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1728         security_free_mnt_opts(&new_sec_opts);
1729         if (error) {
1730                 deactivate_locked_super(s);
1731                 return ERR_PTR(error);
1732         }
1733
1734         return dget(s->s_root);
1735
1736 error_close_devices:
1737         btrfs_close_devices(fs_devices);
1738 error_fs_info:
1739         btrfs_free_fs_info(fs_info);
1740 error_sec_opts:
1741         security_free_mnt_opts(&new_sec_opts);
1742         return ERR_PTR(error);
1743 }
1744
1745 /*
1746  * Mount function which is called by VFS layer.
1747  *
1748  * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1749  * which needs vfsmount* of device's root (/).  This means device's root has to
1750  * be mounted internally in any case.
1751  *
1752  * Operation flow:
1753  *   1. Parse subvol id related options for later use in mount_subvol().
1754  *
1755  *   2. Mount device's root (/) by calling vfs_kern_mount().
1756  *
1757  *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1758  *      first place. In order to avoid calling btrfs_mount() again, we use
1759  *      different file_system_type which is not registered to VFS by
1760  *      register_filesystem() (btrfs_root_fs_type). As a result,
1761  *      btrfs_mount_root() is called. The return value will be used by
1762  *      mount_subtree() in mount_subvol().
1763  *
1764  *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1765  *      "btrfs subvolume set-default", mount_subvol() is called always.
1766  */
1767 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1768                 const char *device_name, void *data)
1769 {
1770         struct vfsmount *mnt_root;
1771         struct dentry *root;
1772         char *subvol_name = NULL;
1773         u64 subvol_objectid = 0;
1774         int error = 0;
1775
1776         error = btrfs_parse_subvol_options(data, &subvol_name,
1777                                         &subvol_objectid);
1778         if (error) {
1779                 kfree(subvol_name);
1780                 return ERR_PTR(error);
1781         }
1782
1783         /* mount device's root (/) */
1784         mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1785         if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1786                 if (flags & SB_RDONLY) {
1787                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1788                                 flags & ~SB_RDONLY, device_name, data);
1789                 } else {
1790                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1791                                 flags | SB_RDONLY, device_name, data);
1792                         if (IS_ERR(mnt_root)) {
1793                                 root = ERR_CAST(mnt_root);
1794                                 kfree(subvol_name);
1795                                 goto out;
1796                         }
1797
1798                         down_write(&mnt_root->mnt_sb->s_umount);
1799                         error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1800                         up_write(&mnt_root->mnt_sb->s_umount);
1801                         if (error < 0) {
1802                                 root = ERR_PTR(error);
1803                                 mntput(mnt_root);
1804                                 kfree(subvol_name);
1805                                 goto out;
1806                         }
1807                 }
1808         }
1809         if (IS_ERR(mnt_root)) {
1810                 root = ERR_CAST(mnt_root);
1811                 kfree(subvol_name);
1812                 goto out;
1813         }
1814
1815         /* mount_subvol() will free subvol_name and mnt_root */
1816         root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1817
1818 out:
1819         return root;
1820 }
1821
1822 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1823                                      u32 new_pool_size, u32 old_pool_size)
1824 {
1825         if (new_pool_size == old_pool_size)
1826                 return;
1827
1828         fs_info->thread_pool_size = new_pool_size;
1829
1830         btrfs_info(fs_info, "resize thread pool %d -> %d",
1831                old_pool_size, new_pool_size);
1832
1833         btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1834         btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1835         btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1836         btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1837         btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1838         btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1839                                 new_pool_size);
1840         btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1841         btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1842         btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1843         btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1844                                 new_pool_size);
1845 }
1846
1847 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1848                                        unsigned long old_opts, int flags)
1849 {
1850         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1851             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1852              (flags & SB_RDONLY))) {
1853                 /* wait for any defraggers to finish */
1854                 wait_event(fs_info->transaction_wait,
1855                            (atomic_read(&fs_info->defrag_running) == 0));
1856                 if (flags & SB_RDONLY)
1857                         sync_filesystem(fs_info->sb);
1858         }
1859 }
1860
1861 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1862                                          unsigned long old_opts)
1863 {
1864         const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1865
1866         /*
1867          * We need to cleanup all defragable inodes if the autodefragment is
1868          * close or the filesystem is read only.
1869          */
1870         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1871             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1872                 btrfs_cleanup_defrag_inodes(fs_info);
1873         }
1874
1875         /* If we toggled discard async */
1876         if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1877             btrfs_test_opt(fs_info, DISCARD_ASYNC))
1878                 btrfs_discard_resume(fs_info);
1879         else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1880                  !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1881                 btrfs_discard_cleanup(fs_info);
1882
1883         /* If we toggled space cache */
1884         if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1885                 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1886 }
1887
1888 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1889 {
1890         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1891         unsigned old_flags = sb->s_flags;
1892         unsigned long old_opts = fs_info->mount_opt;
1893         unsigned long old_compress_type = fs_info->compress_type;
1894         u64 old_max_inline = fs_info->max_inline;
1895         u32 old_thread_pool_size = fs_info->thread_pool_size;
1896         u32 old_metadata_ratio = fs_info->metadata_ratio;
1897         int ret;
1898
1899         sync_filesystem(sb);
1900         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1901
1902         if (data) {
1903                 void *new_sec_opts = NULL;
1904
1905                 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1906                 if (!ret)
1907                         ret = security_sb_remount(sb, new_sec_opts);
1908                 security_free_mnt_opts(&new_sec_opts);
1909                 if (ret)
1910                         goto restore;
1911         }
1912
1913         ret = btrfs_parse_options(fs_info, data, *flags);
1914         if (ret)
1915                 goto restore;
1916
1917         btrfs_remount_begin(fs_info, old_opts, *flags);
1918         btrfs_resize_thread_pool(fs_info,
1919                 fs_info->thread_pool_size, old_thread_pool_size);
1920
1921         if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1922             (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1923             (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1924                 btrfs_warn(fs_info,
1925                 "remount supports changing free space tree only from ro to rw");
1926                 /* Make sure free space cache options match the state on disk */
1927                 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1928                         btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1929                         btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1930                 }
1931                 if (btrfs_free_space_cache_v1_active(fs_info)) {
1932                         btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1933                         btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1934                 }
1935         }
1936
1937         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1938                 goto out;
1939
1940         if (*flags & SB_RDONLY) {
1941                 /*
1942                  * this also happens on 'umount -rf' or on shutdown, when
1943                  * the filesystem is busy.
1944                  */
1945                 cancel_work_sync(&fs_info->async_reclaim_work);
1946                 cancel_work_sync(&fs_info->async_data_reclaim_work);
1947
1948                 btrfs_discard_cleanup(fs_info);
1949
1950                 /* wait for the uuid_scan task to finish */
1951                 down(&fs_info->uuid_tree_rescan_sem);
1952                 /* avoid complains from lockdep et al. */
1953                 up(&fs_info->uuid_tree_rescan_sem);
1954
1955                 btrfs_set_sb_rdonly(sb);
1956
1957                 /*
1958                  * Setting SB_RDONLY will put the cleaner thread to
1959                  * sleep at the next loop if it's already active.
1960                  * If it's already asleep, we'll leave unused block
1961                  * groups on disk until we're mounted read-write again
1962                  * unless we clean them up here.
1963                  */
1964                 btrfs_delete_unused_bgs(fs_info);
1965
1966                 /*
1967                  * The cleaner task could be already running before we set the
1968                  * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1969                  * We must make sure that after we finish the remount, i.e. after
1970                  * we call btrfs_commit_super(), the cleaner can no longer start
1971                  * a transaction - either because it was dropping a dead root,
1972                  * running delayed iputs or deleting an unused block group (the
1973                  * cleaner picked a block group from the list of unused block
1974                  * groups before we were able to in the previous call to
1975                  * btrfs_delete_unused_bgs()).
1976                  */
1977                 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1978                             TASK_UNINTERRUPTIBLE);
1979
1980                 /*
1981                  * We've set the superblock to RO mode, so we might have made
1982                  * the cleaner task sleep without running all pending delayed
1983                  * iputs. Go through all the delayed iputs here, so that if an
1984                  * unmount happens without remounting RW we don't end up at
1985                  * finishing close_ctree() with a non-empty list of delayed
1986                  * iputs.
1987                  */
1988                 btrfs_run_delayed_iputs(fs_info);
1989
1990                 btrfs_dev_replace_suspend_for_unmount(fs_info);
1991                 btrfs_scrub_cancel(fs_info);
1992                 btrfs_pause_balance(fs_info);
1993
1994                 /*
1995                  * Pause the qgroup rescan worker if it is running. We don't want
1996                  * it to be still running after we are in RO mode, as after that,
1997                  * by the time we unmount, it might have left a transaction open,
1998                  * so we would leak the transaction and/or crash.
1999                  */
2000                 btrfs_qgroup_wait_for_completion(fs_info, false);
2001
2002                 ret = btrfs_commit_super(fs_info);
2003                 if (ret)
2004                         goto restore;
2005         } else {
2006                 if (BTRFS_FS_ERROR(fs_info)) {
2007                         btrfs_err(fs_info,
2008                                 "Remounting read-write after error is not allowed");
2009                         ret = -EINVAL;
2010                         goto restore;
2011                 }
2012                 if (fs_info->fs_devices->rw_devices == 0) {
2013                         ret = -EACCES;
2014                         goto restore;
2015                 }
2016
2017                 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
2018                         btrfs_warn(fs_info,
2019                 "too many missing devices, writable remount is not allowed");
2020                         ret = -EACCES;
2021                         goto restore;
2022                 }
2023
2024                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
2025                         btrfs_warn(fs_info,
2026                 "mount required to replay tree-log, cannot remount read-write");
2027                         ret = -EINVAL;
2028                         goto restore;
2029                 }
2030
2031                 /*
2032                  * NOTE: when remounting with a change that does writes, don't
2033                  * put it anywhere above this point, as we are not sure to be
2034                  * safe to write until we pass the above checks.
2035                  */
2036                 ret = btrfs_start_pre_rw_mount(fs_info);
2037                 if (ret)
2038                         goto restore;
2039
2040                 btrfs_clear_sb_rdonly(sb);
2041
2042                 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
2043         }
2044 out:
2045         /*
2046          * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2047          * since the absence of the flag means it can be toggled off by remount.
2048          */
2049         *flags |= SB_I_VERSION;
2050
2051         wake_up_process(fs_info->transaction_kthread);
2052         btrfs_remount_cleanup(fs_info, old_opts);
2053         btrfs_clear_oneshot_options(fs_info);
2054         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2055
2056         return 0;
2057
2058 restore:
2059         /* We've hit an error - don't reset SB_RDONLY */
2060         if (sb_rdonly(sb))
2061                 old_flags |= SB_RDONLY;
2062         if (!(old_flags & SB_RDONLY))
2063                 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
2064         sb->s_flags = old_flags;
2065         fs_info->mount_opt = old_opts;
2066         fs_info->compress_type = old_compress_type;
2067         fs_info->max_inline = old_max_inline;
2068         btrfs_resize_thread_pool(fs_info,
2069                 old_thread_pool_size, fs_info->thread_pool_size);
2070         fs_info->metadata_ratio = old_metadata_ratio;
2071         btrfs_remount_cleanup(fs_info, old_opts);
2072         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2073
2074         return ret;
2075 }
2076
2077 /* Used to sort the devices by max_avail(descending sort) */
2078 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
2079 {
2080         const struct btrfs_device_info *dev_info1 = a;
2081         const struct btrfs_device_info *dev_info2 = b;
2082
2083         if (dev_info1->max_avail > dev_info2->max_avail)
2084                 return -1;
2085         else if (dev_info1->max_avail < dev_info2->max_avail)
2086                 return 1;
2087         return 0;
2088 }
2089
2090 /*
2091  * sort the devices by max_avail, in which max free extent size of each device
2092  * is stored.(Descending Sort)
2093  */
2094 static inline void btrfs_descending_sort_devices(
2095                                         struct btrfs_device_info *devices,
2096                                         size_t nr_devices)
2097 {
2098         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2099              btrfs_cmp_device_free_bytes, NULL);
2100 }
2101
2102 /*
2103  * The helper to calc the free space on the devices that can be used to store
2104  * file data.
2105  */
2106 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2107                                               u64 *free_bytes)
2108 {
2109         struct btrfs_device_info *devices_info;
2110         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2111         struct btrfs_device *device;
2112         u64 type;
2113         u64 avail_space;
2114         u64 min_stripe_size;
2115         int num_stripes = 1;
2116         int i = 0, nr_devices;
2117         const struct btrfs_raid_attr *rattr;
2118
2119         /*
2120          * We aren't under the device list lock, so this is racy-ish, but good
2121          * enough for our purposes.
2122          */
2123         nr_devices = fs_info->fs_devices->open_devices;
2124         if (!nr_devices) {
2125                 smp_mb();
2126                 nr_devices = fs_info->fs_devices->open_devices;
2127                 ASSERT(nr_devices);
2128                 if (!nr_devices) {
2129                         *free_bytes = 0;
2130                         return 0;
2131                 }
2132         }
2133
2134         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2135                                GFP_KERNEL);
2136         if (!devices_info)
2137                 return -ENOMEM;
2138
2139         /* calc min stripe number for data space allocation */
2140         type = btrfs_data_alloc_profile(fs_info);
2141         rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2142
2143         if (type & BTRFS_BLOCK_GROUP_RAID0)
2144                 num_stripes = nr_devices;
2145         else if (type & BTRFS_BLOCK_GROUP_RAID1)
2146                 num_stripes = 2;
2147         else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2148                 num_stripes = 3;
2149         else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2150                 num_stripes = 4;
2151         else if (type & BTRFS_BLOCK_GROUP_RAID10)
2152                 num_stripes = 4;
2153
2154         /* Adjust for more than 1 stripe per device */
2155         min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2156
2157         rcu_read_lock();
2158         list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2159                 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2160                                                 &device->dev_state) ||
2161                     !device->bdev ||
2162                     test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2163                         continue;
2164
2165                 if (i >= nr_devices)
2166                         break;
2167
2168                 avail_space = device->total_bytes - device->bytes_used;
2169
2170                 /* align with stripe_len */
2171                 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2172
2173                 /*
2174                  * In order to avoid overwriting the superblock on the drive,
2175                  * btrfs starts at an offset of at least 1MB when doing chunk
2176                  * allocation.
2177                  *
2178                  * This ensures we have at least min_stripe_size free space
2179                  * after excluding 1MB.
2180                  */
2181                 if (avail_space <= SZ_1M + min_stripe_size)
2182                         continue;
2183
2184                 avail_space -= SZ_1M;
2185
2186                 devices_info[i].dev = device;
2187                 devices_info[i].max_avail = avail_space;
2188
2189                 i++;
2190         }
2191         rcu_read_unlock();
2192
2193         nr_devices = i;
2194
2195         btrfs_descending_sort_devices(devices_info, nr_devices);
2196
2197         i = nr_devices - 1;
2198         avail_space = 0;
2199         while (nr_devices >= rattr->devs_min) {
2200                 num_stripes = min(num_stripes, nr_devices);
2201
2202                 if (devices_info[i].max_avail >= min_stripe_size) {
2203                         int j;
2204                         u64 alloc_size;
2205
2206                         avail_space += devices_info[i].max_avail * num_stripes;
2207                         alloc_size = devices_info[i].max_avail;
2208                         for (j = i + 1 - num_stripes; j <= i; j++)
2209                                 devices_info[j].max_avail -= alloc_size;
2210                 }
2211                 i--;
2212                 nr_devices--;
2213         }
2214
2215         kfree(devices_info);
2216         *free_bytes = avail_space;
2217         return 0;
2218 }
2219
2220 /*
2221  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2222  *
2223  * If there's a redundant raid level at DATA block groups, use the respective
2224  * multiplier to scale the sizes.
2225  *
2226  * Unused device space usage is based on simulating the chunk allocator
2227  * algorithm that respects the device sizes and order of allocations.  This is
2228  * a close approximation of the actual use but there are other factors that may
2229  * change the result (like a new metadata chunk).
2230  *
2231  * If metadata is exhausted, f_bavail will be 0.
2232  */
2233 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2234 {
2235         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2236         struct btrfs_super_block *disk_super = fs_info->super_copy;
2237         struct btrfs_space_info *found;
2238         u64 total_used = 0;
2239         u64 total_free_data = 0;
2240         u64 total_free_meta = 0;
2241         u32 bits = fs_info->sectorsize_bits;
2242         __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2243         unsigned factor = 1;
2244         struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2245         int ret;
2246         u64 thresh = 0;
2247         int mixed = 0;
2248
2249         list_for_each_entry(found, &fs_info->space_info, list) {
2250                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2251                         int i;
2252
2253                         total_free_data += found->disk_total - found->disk_used;
2254                         total_free_data -=
2255                                 btrfs_account_ro_block_groups_free_space(found);
2256
2257                         for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2258                                 if (!list_empty(&found->block_groups[i]))
2259                                         factor = btrfs_bg_type_to_factor(
2260                                                 btrfs_raid_array[i].bg_flag);
2261                         }
2262                 }
2263
2264                 /*
2265                  * Metadata in mixed block goup profiles are accounted in data
2266                  */
2267                 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2268                         if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2269                                 mixed = 1;
2270                         else
2271                                 total_free_meta += found->disk_total -
2272                                         found->disk_used;
2273                 }
2274
2275                 total_used += found->disk_used;
2276         }
2277
2278         buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2279         buf->f_blocks >>= bits;
2280         buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2281
2282         /* Account global block reserve as used, it's in logical size already */
2283         spin_lock(&block_rsv->lock);
2284         /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2285         if (buf->f_bfree >= block_rsv->size >> bits)
2286                 buf->f_bfree -= block_rsv->size >> bits;
2287         else
2288                 buf->f_bfree = 0;
2289         spin_unlock(&block_rsv->lock);
2290
2291         buf->f_bavail = div_u64(total_free_data, factor);
2292         ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2293         if (ret)
2294                 return ret;
2295         buf->f_bavail += div_u64(total_free_data, factor);
2296         buf->f_bavail = buf->f_bavail >> bits;
2297
2298         /*
2299          * We calculate the remaining metadata space minus global reserve. If
2300          * this is (supposedly) smaller than zero, there's no space. But this
2301          * does not hold in practice, the exhausted state happens where's still
2302          * some positive delta. So we apply some guesswork and compare the
2303          * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2304          *
2305          * We probably cannot calculate the exact threshold value because this
2306          * depends on the internal reservations requested by various
2307          * operations, so some operations that consume a few metadata will
2308          * succeed even if the Avail is zero. But this is better than the other
2309          * way around.
2310          */
2311         thresh = SZ_4M;
2312
2313         /*
2314          * We only want to claim there's no available space if we can no longer
2315          * allocate chunks for our metadata profile and our global reserve will
2316          * not fit in the free metadata space.  If we aren't ->full then we
2317          * still can allocate chunks and thus are fine using the currently
2318          * calculated f_bavail.
2319          */
2320         if (!mixed && block_rsv->space_info->full &&
2321             total_free_meta - thresh < block_rsv->size)
2322                 buf->f_bavail = 0;
2323
2324         buf->f_type = BTRFS_SUPER_MAGIC;
2325         buf->f_bsize = dentry->d_sb->s_blocksize;
2326         buf->f_namelen = BTRFS_NAME_LEN;
2327
2328         /* We treat it as constant endianness (it doesn't matter _which_)
2329            because we want the fsid to come out the same whether mounted
2330            on a big-endian or little-endian host */
2331         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2332         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2333         /* Mask in the root object ID too, to disambiguate subvols */
2334         buf->f_fsid.val[0] ^=
2335                 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2336         buf->f_fsid.val[1] ^=
2337                 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2338
2339         return 0;
2340 }
2341
2342 static void btrfs_kill_super(struct super_block *sb)
2343 {
2344         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2345         kill_anon_super(sb);
2346         btrfs_free_fs_info(fs_info);
2347 }
2348
2349 static struct file_system_type btrfs_fs_type = {
2350         .owner          = THIS_MODULE,
2351         .name           = "btrfs",
2352         .mount          = btrfs_mount,
2353         .kill_sb        = btrfs_kill_super,
2354         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2355 };
2356
2357 static struct file_system_type btrfs_root_fs_type = {
2358         .owner          = THIS_MODULE,
2359         .name           = "btrfs",
2360         .mount          = btrfs_mount_root,
2361         .kill_sb        = btrfs_kill_super,
2362         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2363 };
2364
2365 MODULE_ALIAS_FS("btrfs");
2366
2367 static int btrfs_control_open(struct inode *inode, struct file *file)
2368 {
2369         /*
2370          * The control file's private_data is used to hold the
2371          * transaction when it is started and is used to keep
2372          * track of whether a transaction is already in progress.
2373          */
2374         file->private_data = NULL;
2375         return 0;
2376 }
2377
2378 /*
2379  * Used by /dev/btrfs-control for devices ioctls.
2380  */
2381 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2382                                 unsigned long arg)
2383 {
2384         struct btrfs_ioctl_vol_args *vol;
2385         struct btrfs_device *device = NULL;
2386         int ret = -ENOTTY;
2387
2388         if (!capable(CAP_SYS_ADMIN))
2389                 return -EPERM;
2390
2391         vol = memdup_user((void __user *)arg, sizeof(*vol));
2392         if (IS_ERR(vol))
2393                 return PTR_ERR(vol);
2394         vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2395
2396         switch (cmd) {
2397         case BTRFS_IOC_SCAN_DEV:
2398                 mutex_lock(&uuid_mutex);
2399                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2400                                                &btrfs_root_fs_type);
2401                 ret = PTR_ERR_OR_ZERO(device);
2402                 mutex_unlock(&uuid_mutex);
2403                 break;
2404         case BTRFS_IOC_FORGET_DEV:
2405                 ret = btrfs_forget_devices(vol->name);
2406                 break;
2407         case BTRFS_IOC_DEVICES_READY:
2408                 mutex_lock(&uuid_mutex);
2409                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2410                                                &btrfs_root_fs_type);
2411                 if (IS_ERR(device)) {
2412                         mutex_unlock(&uuid_mutex);
2413                         ret = PTR_ERR(device);
2414                         break;
2415                 }
2416                 ret = !(device->fs_devices->num_devices ==
2417                         device->fs_devices->total_devices);
2418                 mutex_unlock(&uuid_mutex);
2419                 break;
2420         case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2421                 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2422                 break;
2423         }
2424
2425         kfree(vol);
2426         return ret;
2427 }
2428
2429 static int btrfs_freeze(struct super_block *sb)
2430 {
2431         struct btrfs_trans_handle *trans;
2432         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2433         struct btrfs_root *root = fs_info->tree_root;
2434
2435         set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2436         /*
2437          * We don't need a barrier here, we'll wait for any transaction that
2438          * could be in progress on other threads (and do delayed iputs that
2439          * we want to avoid on a frozen filesystem), or do the commit
2440          * ourselves.
2441          */
2442         trans = btrfs_attach_transaction_barrier(root);
2443         if (IS_ERR(trans)) {
2444                 /* no transaction, don't bother */
2445                 if (PTR_ERR(trans) == -ENOENT)
2446                         return 0;
2447                 return PTR_ERR(trans);
2448         }
2449         return btrfs_commit_transaction(trans);
2450 }
2451
2452 static int btrfs_unfreeze(struct super_block *sb)
2453 {
2454         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2455
2456         clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2457         return 0;
2458 }
2459
2460 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2461 {
2462         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2463
2464         /*
2465          * There should be always a valid pointer in latest_dev, it may be stale
2466          * for a short moment in case it's being deleted but still valid until
2467          * the end of RCU grace period.
2468          */
2469         rcu_read_lock();
2470         seq_escape(m, rcu_str_deref(fs_info->fs_devices->latest_dev->name), " \t\n\\");
2471         rcu_read_unlock();
2472
2473         return 0;
2474 }
2475
2476 static const struct super_operations btrfs_super_ops = {
2477         .drop_inode     = btrfs_drop_inode,
2478         .evict_inode    = btrfs_evict_inode,
2479         .put_super      = btrfs_put_super,
2480         .sync_fs        = btrfs_sync_fs,
2481         .show_options   = btrfs_show_options,
2482         .show_devname   = btrfs_show_devname,
2483         .alloc_inode    = btrfs_alloc_inode,
2484         .destroy_inode  = btrfs_destroy_inode,
2485         .free_inode     = btrfs_free_inode,
2486         .statfs         = btrfs_statfs,
2487         .remount_fs     = btrfs_remount,
2488         .freeze_fs      = btrfs_freeze,
2489         .unfreeze_fs    = btrfs_unfreeze,
2490 };
2491
2492 static const struct file_operations btrfs_ctl_fops = {
2493         .open = btrfs_control_open,
2494         .unlocked_ioctl  = btrfs_control_ioctl,
2495         .compat_ioctl = compat_ptr_ioctl,
2496         .owner   = THIS_MODULE,
2497         .llseek = noop_llseek,
2498 };
2499
2500 static struct miscdevice btrfs_misc = {
2501         .minor          = BTRFS_MINOR,
2502         .name           = "btrfs-control",
2503         .fops           = &btrfs_ctl_fops
2504 };
2505
2506 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2507 MODULE_ALIAS("devname:btrfs-control");
2508
2509 static int __init btrfs_interface_init(void)
2510 {
2511         return misc_register(&btrfs_misc);
2512 }
2513
2514 static __cold void btrfs_interface_exit(void)
2515 {
2516         misc_deregister(&btrfs_misc);
2517 }
2518
2519 static void __init btrfs_print_mod_info(void)
2520 {
2521         static const char options[] = ""
2522 #ifdef CONFIG_BTRFS_DEBUG
2523                         ", debug=on"
2524 #endif
2525 #ifdef CONFIG_BTRFS_ASSERT
2526                         ", assert=on"
2527 #endif
2528 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2529                         ", integrity-checker=on"
2530 #endif
2531 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2532                         ", ref-verify=on"
2533 #endif
2534 #ifdef CONFIG_BLK_DEV_ZONED
2535                         ", zoned=yes"
2536 #else
2537                         ", zoned=no"
2538 #endif
2539 #ifdef CONFIG_FS_VERITY
2540                         ", fsverity=yes"
2541 #else
2542                         ", fsverity=no"
2543 #endif
2544                         ;
2545         pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2546 }
2547
2548 static int __init init_btrfs_fs(void)
2549 {
2550         int err;
2551
2552         btrfs_props_init();
2553
2554         err = btrfs_init_sysfs();
2555         if (err)
2556                 return err;
2557
2558         btrfs_init_compress();
2559
2560         err = btrfs_init_cachep();
2561         if (err)
2562                 goto free_compress;
2563
2564         err = extent_io_init();
2565         if (err)
2566                 goto free_cachep;
2567
2568         err = extent_state_cache_init();
2569         if (err)
2570                 goto free_extent_io;
2571
2572         err = extent_map_init();
2573         if (err)
2574                 goto free_extent_state_cache;
2575
2576         err = ordered_data_init();
2577         if (err)
2578                 goto free_extent_map;
2579
2580         err = btrfs_delayed_inode_init();
2581         if (err)
2582                 goto free_ordered_data;
2583
2584         err = btrfs_auto_defrag_init();
2585         if (err)
2586                 goto free_delayed_inode;
2587
2588         err = btrfs_delayed_ref_init();
2589         if (err)
2590                 goto free_auto_defrag;
2591
2592         err = btrfs_prelim_ref_init();
2593         if (err)
2594                 goto free_delayed_ref;
2595
2596         err = btrfs_end_io_wq_init();
2597         if (err)
2598                 goto free_prelim_ref;
2599
2600         err = btrfs_interface_init();
2601         if (err)
2602                 goto free_end_io_wq;
2603
2604         btrfs_print_mod_info();
2605
2606         err = btrfs_run_sanity_tests();
2607         if (err)
2608                 goto unregister_ioctl;
2609
2610         err = register_filesystem(&btrfs_fs_type);
2611         if (err)
2612                 goto unregister_ioctl;
2613
2614         return 0;
2615
2616 unregister_ioctl:
2617         btrfs_interface_exit();
2618 free_end_io_wq:
2619         btrfs_end_io_wq_exit();
2620 free_prelim_ref:
2621         btrfs_prelim_ref_exit();
2622 free_delayed_ref:
2623         btrfs_delayed_ref_exit();
2624 free_auto_defrag:
2625         btrfs_auto_defrag_exit();
2626 free_delayed_inode:
2627         btrfs_delayed_inode_exit();
2628 free_ordered_data:
2629         ordered_data_exit();
2630 free_extent_map:
2631         extent_map_exit();
2632 free_extent_state_cache:
2633         extent_state_cache_exit();
2634 free_extent_io:
2635         extent_io_exit();
2636 free_cachep:
2637         btrfs_destroy_cachep();
2638 free_compress:
2639         btrfs_exit_compress();
2640         btrfs_exit_sysfs();
2641
2642         return err;
2643 }
2644
2645 static void __exit exit_btrfs_fs(void)
2646 {
2647         btrfs_destroy_cachep();
2648         btrfs_delayed_ref_exit();
2649         btrfs_auto_defrag_exit();
2650         btrfs_delayed_inode_exit();
2651         btrfs_prelim_ref_exit();
2652         ordered_data_exit();
2653         extent_map_exit();
2654         extent_state_cache_exit();
2655         extent_io_exit();
2656         btrfs_interface_exit();
2657         btrfs_end_io_wq_exit();
2658         unregister_filesystem(&btrfs_fs_type);
2659         btrfs_exit_sysfs();
2660         btrfs_cleanup_fs_uuids();
2661         btrfs_exit_compress();
2662 }
2663
2664 late_initcall(init_btrfs_fs);
2665 module_exit(exit_btrfs_fs)
2666
2667 MODULE_LICENSE("GPL");
2668 MODULE_SOFTDEP("pre: crc32c");
2669 MODULE_SOFTDEP("pre: xxhash64");
2670 MODULE_SOFTDEP("pre: sha256");
2671 MODULE_SOFTDEP("pre: blake2b-256");