fs/btrfs/super.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2007 Oracle.  All rights reserved.
   4  */
   5
   6 #include <linux/blkdev.h>
   7 #include <linux/module.h>
   8 #include <linux/fs.h>
   9 #include <linux/pagemap.h>
  10 #include <linux/highmem.h>
  11 #include <linux/time.h>
  12 #include <linux/init.h>
  13 #include <linux/seq_file.h>
  14 #include <linux/string.h>
  15 #include <linux/backing-dev.h>
  16 #include <linux/mount.h>
  17 #include <linux/writeback.h>
  18 #include <linux/statfs.h>
  19 #include <linux/compat.h>
  20 #include <linux/parser.h>
  21 #include <linux/ctype.h>
  22 #include <linux/namei.h>
  23 #include <linux/miscdevice.h>
  24 #include <linux/magic.h>
  25 #include <linux/slab.h>
  26 #include <linux/cleancache.h>
  27 #include <linux/ratelimit.h>
  28 #include <linux/crc32c.h>
  29 #include <linux/btrfs.h>
  30 #include "delayed-inode.h"
  31 #include "ctree.h"
  32 #include "disk-io.h"
  33 #include "transaction.h"
  34 #include "btrfs_inode.h"
  35 #include "print-tree.h"
  36 #include "props.h"
  37 #include "xattr.h"
  38 #include "volumes.h"
  39 #include "export.h"
  40 #include "compression.h"
  41 #include "rcu-string.h"
  42 #include "dev-replace.h"
  43 #include "free-space-cache.h"
  44 #include "backref.h"
  45 #include "space-info.h"
  46 #include "sysfs.h"
  47 #include "zoned.h"
  48 #include "tests/btrfs-tests.h"
  49 #include "block-group.h"
  50 #include "discard.h"
  51 #include "qgroup.h"
  52 #define CREATE_TRACE_POINTS
  53 #include <trace/events/btrfs.h>
  54
  55 static const struct super_operations btrfs_super_ops;
  56
  57 /*
  58  * Types for mounting the default subvolume and a subvolume explicitly
  59  * requested by subvol=/path. That way the callchain is straightforward and we
  60  * don't have to play tricks with the mount options and recursive calls to
  61  * btrfs_mount.
  62  *
  63  * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
  64  */
  65 static struct file_system_type btrfs_fs_type;
  66 static struct file_system_type btrfs_root_fs_type;
  67
  68 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
  69
  70 /*
  71  * Generally the error codes correspond to their respective errors, but there
  72  * are a few special cases.
  73  *
  74  * EUCLEAN: Any sort of corruption that we encounter.  The tree-checker for
  75  *          instance will return EUCLEAN if any of the blocks are corrupted in
  76  *          a way that is problematic.  We want to reserve EUCLEAN for these
  77  *          sort of corruptions.
  78  *
  79  * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
  80  *        need to use EROFS for this case.  We will have no idea of the
  81  *        original failure, that will have been reported at the time we tripped
  82  *        over the error.  Each subsequent error that doesn't have any context
  83  *        of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
  84  */
  85 const char * __attribute_const__ btrfs_decode_error(int errno)
  86 {
  87         char *errstr = "unknown";
  88
  89         switch (errno) {
  90         case -ENOENT:           /* -2 */
  91                 errstr = "No such entry";
  92                 break;
  93         case -EIO:              /* -5 */
  94                 errstr = "IO failure";
  95                 break;
  96         case -ENOMEM:           /* -12*/
  97                 errstr = "Out of memory";
  98                 break;
  99         case -EEXIST:           /* -17 */
 100                 errstr = "Object already exists";
 101                 break;
 102         case -ENOSPC:           /* -28 */
 103                 errstr = "No space left";
 104                 break;
 105         case -EROFS:            /* -30 */
 106                 errstr = "Readonly filesystem";
 107                 break;
 108         case -EOPNOTSUPP:       /* -95 */
 109                 errstr = "Operation not supported";
 110                 break;
 111         case -EUCLEAN:          /* -117 */
 112                 errstr = "Filesystem corrupted";
 113                 break;
 114         case -EDQUOT:           /* -122 */
 115                 errstr = "Quota exceeded";
 116                 break;
 117         }
 118
 119         return errstr;
 120 }
 121
 122 /*
 123  * __btrfs_handle_fs_error decodes expected errors from the caller and
 124  * invokes the appropriate error response.
 125  */
 126 __cold
 127 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
 128                        unsigned int line, int errno, const char *fmt, ...)
 129 {
 130         struct super_block *sb = fs_info->sb;
 131 #ifdef CONFIG_PRINTK
 132         const char *errstr;
 133 #endif
 134
 135         /*
 136          * Special case: if the error is EROFS, and we're already
 137          * under SB_RDONLY, then it is safe here.
 138          */
 139         if (errno == -EROFS && sb_rdonly(sb))
 140                 return;
 141
 142 #ifdef CONFIG_PRINTK
 143         errstr = btrfs_decode_error(errno);
 144         if (fmt) {
 145                 struct va_format vaf;
 146                 va_list args;
 147
 148                 va_start(args, fmt);
 149                 vaf.fmt = fmt;
 150                 vaf.va = &args;
 151
 152                 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
 153                         sb->s_id, function, line, errno, errstr, &vaf);
 154                 va_end(args);
 155         } else {
 156                 pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
 157                         sb->s_id, function, line, errno, errstr);
 158         }
 159 #endif
 160
 161         /*
 162          * Today we only save the error info to memory.  Long term we'll
 163          * also send it down to the disk
 164          */
 165         set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
 166
 167         /* Don't go through full error handling during mount */
 168         if (!(sb->s_flags & SB_BORN))
 169                 return;
 170
 171         if (sb_rdonly(sb))
 172                 return;
 173
 174         btrfs_discard_stop(fs_info);
 175
 176         /* btrfs handle error by forcing the filesystem readonly */
 177         btrfs_set_sb_rdonly(sb);
 178         btrfs_info(fs_info, "forced readonly");
 179         /*
 180          * Note that a running device replace operation is not canceled here
 181          * although there is no way to update the progress. It would add the
 182          * risk of a deadlock, therefore the canceling is omitted. The only
 183          * penalty is that some I/O remains active until the procedure
 184          * completes. The next time when the filesystem is mounted writable
 185          * again, the device replace operation continues.
 186          */
 187 }
 188
 189 #ifdef CONFIG_PRINTK
 190 static const char * const logtypes[] = {
 191         "emergency",
 192         "alert",
 193         "critical",
 194         "error",
 195         "warning",
 196         "notice",
 197         "info",
 198         "debug",
 199 };
 200
 201
 202 /*
 203  * Use one ratelimit state per log level so that a flood of less important
 204  * messages doesn't cause more important ones to be dropped.
 205  */
 206 static struct ratelimit_state printk_limits[] = {
 207         RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
 208         RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
 209         RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
 210         RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
 211         RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
 212         RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
 213         RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
 214         RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
 215 };
 216
 217 void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 218 {
 219         char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 220         struct va_format vaf;
 221         va_list args;
 222         int kern_level;
 223         const char *type = logtypes[4];
 224         struct ratelimit_state *ratelimit = &printk_limits[4];
 225
 226         va_start(args, fmt);
 227
 228         while ((kern_level = printk_get_level(fmt)) != 0) {
 229                 size_t size = printk_skip_level(fmt) - fmt;
 230
 231                 if (kern_level >= '0' && kern_level <= '7') {
 232                         memcpy(lvl, fmt,  size);
 233                         lvl[size] = '\0';
 234                         type = logtypes[kern_level - '0'];
 235                         ratelimit = &printk_limits[kern_level - '0'];
 236                 }
 237                 fmt += size;
 238         }
 239
 240         vaf.fmt = fmt;
 241         vaf.va = &args;
 242
 243         if (__ratelimit(ratelimit)) {
 244                 if (fs_info)
 245                         printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
 246                                 fs_info->sb->s_id, &vaf);
 247                 else
 248                         printk("%sBTRFS %s: %pV\n", lvl, type, &vaf);
 249         }
 250
 251         va_end(args);
 252 }
 253 #endif
 254
 255 #if BITS_PER_LONG == 32
 256 void __cold btrfs_warn_32bit_limit(struct btrfs_fs_info *fs_info)
 257 {
 258         if (!test_and_set_bit(BTRFS_FS_32BIT_WARN, &fs_info->flags)) {
 259                 btrfs_warn(fs_info, "reaching 32bit limit for logical addresses");
 260                 btrfs_warn(fs_info,
 261 "due to page cache limit on 32bit systems, btrfs can't access metadata at or beyond %lluT",
 262                            BTRFS_32BIT_MAX_FILE_SIZE >> 40);
 263                 btrfs_warn(fs_info,
 264                            "please consider upgrading to 64bit kernel/hardware");
 265         }
 266 }
 267
 268 void __cold btrfs_err_32bit_limit(struct btrfs_fs_info *fs_info)
 269 {
 270         if (!test_and_set_bit(BTRFS_FS_32BIT_ERROR, &fs_info->flags)) {
 271                 btrfs_err(fs_info, "reached 32bit limit for logical addresses");
 272                 btrfs_err(fs_info,
 273 "due to page cache limit on 32bit systems, metadata beyond %lluT can't be accessed",
 274                           BTRFS_32BIT_MAX_FILE_SIZE >> 40);
 275                 btrfs_err(fs_info,
 276                            "please consider upgrading to 64bit kernel/hardware");
 277         }
 278 }
 279 #endif
 280
 281 /*
 282  * We only mark the transaction aborted and then set the file system read-only.
 283  * This will prevent new transactions from starting or trying to join this
 284  * one.
 285  *
 286  * This means that error recovery at the call site is limited to freeing
 287  * any local memory allocations and passing the error code up without
 288  * further cleanup. The transaction should complete as it normally would
 289  * in the call path but will return -EIO.
 290  *
 291  * We'll complete the cleanup in btrfs_end_transaction and
 292  * btrfs_commit_transaction.
 293  */
 294 __cold
 295 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 296                                const char *function,
 297                                unsigned int line, int errno)
 298 {
 299         struct btrfs_fs_info *fs_info = trans->fs_info;
 300
 301         WRITE_ONCE(trans->aborted, errno);
 302         WRITE_ONCE(trans->transaction->aborted, errno);
 303         /* Wake up anybody who may be waiting on this transaction */
 304         wake_up(&fs_info->transaction_wait);
 305         wake_up(&fs_info->transaction_blocked_wait);
 306         __btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
 307 }
 308 /*
 309  * __btrfs_panic decodes unexpected, fatal errors from the caller,
 310  * issues an alert, and either panics or BUGs, depending on mount options.
 311  */
 312 __cold
 313 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 314                    unsigned int line, int errno, const char *fmt, ...)
 315 {
 316         char *s_id = "<unknown>";
 317         const char *errstr;
 318         struct va_format vaf = { .fmt = fmt };
 319         va_list args;
 320
 321         if (fs_info)
 322                 s_id = fs_info->sb->s_id;
 323
 324         va_start(args, fmt);
 325         vaf.va = &args;
 326
 327         errstr = btrfs_decode_error(errno);
 328         if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
 329                 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
 330                         s_id, function, line, &vaf, errno, errstr);
 331
 332         btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
 333                    function, line, &vaf, errno, errstr);
 334         va_end(args);
 335         /* Caller calls BUG() */
 336 }
 337
 338 static void btrfs_put_super(struct super_block *sb)
 339 {
 340         close_ctree(btrfs_sb(sb));
 341 }
 342
 343 enum {
 344         Opt_acl, Opt_noacl,
 345         Opt_clear_cache,
 346         Opt_commit_interval,
 347         Opt_compress,
 348         Opt_compress_force,
 349         Opt_compress_force_type,
 350         Opt_compress_type,
 351         Opt_degraded,
 352         Opt_device,
 353         Opt_fatal_errors,
 354         Opt_flushoncommit, Opt_noflushoncommit,
 355         Opt_max_inline,
 356         Opt_barrier, Opt_nobarrier,
 357         Opt_datacow, Opt_nodatacow,
 358         Opt_datasum, Opt_nodatasum,
 359         Opt_defrag, Opt_nodefrag,
 360         Opt_discard, Opt_nodiscard,
 361         Opt_discard_mode,
 362         Opt_norecovery,
 363         Opt_ratio,
 364         Opt_rescan_uuid_tree,
 365         Opt_skip_balance,
 366         Opt_space_cache, Opt_no_space_cache,
 367         Opt_space_cache_version,
 368         Opt_ssd, Opt_nossd,
 369         Opt_ssd_spread, Opt_nossd_spread,
 370         Opt_subvol,
 371         Opt_subvol_empty,
 372         Opt_subvolid,
 373         Opt_thread_pool,
 374         Opt_treelog, Opt_notreelog,
 375         Opt_user_subvol_rm_allowed,
 376
 377         /* Rescue options */
 378         Opt_rescue,
 379         Opt_usebackuproot,
 380         Opt_nologreplay,
 381         Opt_ignorebadroots,
 382         Opt_ignoredatacsums,
 383         Opt_rescue_all,
 384
 385         /* Deprecated options */
 386         Opt_recovery,
 387         Opt_inode_cache, Opt_noinode_cache,
 388
 389         /* Debugging options */
 390         Opt_check_integrity,
 391         Opt_check_integrity_including_extent_data,
 392         Opt_check_integrity_print_mask,
 393         Opt_enospc_debug, Opt_noenospc_debug,
 394 #ifdef CONFIG_BTRFS_DEBUG
 395         Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 396 #endif
 397 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 398         Opt_ref_verify,
 399 #endif
 400         Opt_err,
 401 };
 402
 403 static const match_table_t tokens = {
 404         {Opt_acl, "acl"},
 405         {Opt_noacl, "noacl"},
 406         {Opt_clear_cache, "clear_cache"},
 407         {Opt_commit_interval, "commit=%u"},
 408         {Opt_compress, "compress"},
 409         {Opt_compress_type, "compress=%s"},
 410         {Opt_compress_force, "compress-force"},
 411         {Opt_compress_force_type, "compress-force=%s"},
 412         {Opt_degraded, "degraded"},
 413         {Opt_device, "device=%s"},
 414         {Opt_fatal_errors, "fatal_errors=%s"},
 415         {Opt_flushoncommit, "flushoncommit"},
 416         {Opt_noflushoncommit, "noflushoncommit"},
 417         {Opt_inode_cache, "inode_cache"},
 418         {Opt_noinode_cache, "noinode_cache"},
 419         {Opt_max_inline, "max_inline=%s"},
 420         {Opt_barrier, "barrier"},
 421         {Opt_nobarrier, "nobarrier"},
 422         {Opt_datacow, "datacow"},
 423         {Opt_nodatacow, "nodatacow"},
 424         {Opt_datasum, "datasum"},
 425         {Opt_nodatasum, "nodatasum"},
 426         {Opt_defrag, "autodefrag"},
 427         {Opt_nodefrag, "noautodefrag"},
 428         {Opt_discard, "discard"},
 429         {Opt_discard_mode, "discard=%s"},
 430         {Opt_nodiscard, "nodiscard"},
 431         {Opt_norecovery, "norecovery"},
 432         {Opt_ratio, "metadata_ratio=%u"},
 433         {Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 434         {Opt_skip_balance, "skip_balance"},
 435         {Opt_space_cache, "space_cache"},
 436         {Opt_no_space_cache, "nospace_cache"},
 437         {Opt_space_cache_version, "space_cache=%s"},
 438         {Opt_ssd, "ssd"},
 439         {Opt_nossd, "nossd"},
 440         {Opt_ssd_spread, "ssd_spread"},
 441         {Opt_nossd_spread, "nossd_spread"},
 442         {Opt_subvol, "subvol=%s"},
 443         {Opt_subvol_empty, "subvol="},
 444         {Opt_subvolid, "subvolid=%s"},
 445         {Opt_thread_pool, "thread_pool=%u"},
 446         {Opt_treelog, "treelog"},
 447         {Opt_notreelog, "notreelog"},
 448         {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 449
 450         /* Rescue options */
 451         {Opt_rescue, "rescue=%s"},
 452         /* Deprecated, with alias rescue=nologreplay */
 453         {Opt_nologreplay, "nologreplay"},
 454         /* Deprecated, with alias rescue=usebackuproot */
 455         {Opt_usebackuproot, "usebackuproot"},
 456
 457         /* Deprecated options */
 458         {Opt_recovery, "recovery"},
 459
 460         /* Debugging options */
 461         {Opt_check_integrity, "check_int"},
 462         {Opt_check_integrity_including_extent_data, "check_int_data"},
 463         {Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
 464         {Opt_enospc_debug, "enospc_debug"},
 465         {Opt_noenospc_debug, "noenospc_debug"},
 466 #ifdef CONFIG_BTRFS_DEBUG
 467         {Opt_fragment_data, "fragment=data"},
 468         {Opt_fragment_metadata, "fragment=metadata"},
 469         {Opt_fragment_all, "fragment=all"},
 470 #endif
 471 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
 472         {Opt_ref_verify, "ref_verify"},
 473 #endif
 474         {Opt_err, NULL},
 475 };
 476
 477 static const match_table_t rescue_tokens = {
 478         {Opt_usebackuproot, "usebackuproot"},
 479         {Opt_nologreplay, "nologreplay"},
 480         {Opt_ignorebadroots, "ignorebadroots"},
 481         {Opt_ignorebadroots, "ibadroots"},
 482         {Opt_ignoredatacsums, "ignoredatacsums"},
 483         {Opt_ignoredatacsums, "idatacsums"},
 484         {Opt_rescue_all, "all"},
 485         {Opt_err, NULL},
 486 };
 487
 488 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
 489                             const char *opt_name)
 490 {
 491         if (fs_info->mount_opt & opt) {
 492                 btrfs_err(fs_info, "%s must be used with ro mount option",
 493                           opt_name);
 494                 return true;
 495         }
 496         return false;
 497 }
 498
 499 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
 500 {
 501         char *opts;
 502         char *orig;
 503         char *p;
 504         substring_t args[MAX_OPT_ARGS];
 505         int ret = 0;
 506
 507         opts = kstrdup(options, GFP_KERNEL);
 508         if (!opts)
 509                 return -ENOMEM;
 510         orig = opts;
 511
 512         while ((p = strsep(&opts, ":")) != NULL) {
 513                 int token;
 514
 515                 if (!*p)
 516                         continue;
 517                 token = match_token(p, rescue_tokens, args);
 518                 switch (token){
 519                 case Opt_usebackuproot:
 520                         btrfs_info(info,
 521                                    "trying to use backup root at mount time");
 522                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 523                         break;
 524                 case Opt_nologreplay:
 525                         btrfs_set_and_info(info, NOLOGREPLAY,
 526                                            "disabling log replay at mount time");
 527                         break;
 528                 case Opt_ignorebadroots:
 529                         btrfs_set_and_info(info, IGNOREBADROOTS,
 530                                            "ignoring bad roots");
 531                         break;
 532                 case Opt_ignoredatacsums:
 533                         btrfs_set_and_info(info, IGNOREDATACSUMS,
 534                                            "ignoring data csums");
 535                         break;
 536                 case Opt_rescue_all:
 537                         btrfs_info(info, "enabling all of the rescue options");
 538                         btrfs_set_and_info(info, IGNOREDATACSUMS,
 539                                            "ignoring data csums");
 540                         btrfs_set_and_info(info, IGNOREBADROOTS,
 541                                            "ignoring bad roots");
 542                         btrfs_set_and_info(info, NOLOGREPLAY,
 543                                            "disabling log replay at mount time");
 544                         break;
 545                 case Opt_err:
 546                         btrfs_info(info, "unrecognized rescue option '%s'", p);
 547                         ret = -EINVAL;
 548                         goto out;
 549                 default:
 550                         break;
 551                 }
 552
 553         }
 554 out:
 555         kfree(orig);
 556         return ret;
 557 }
 558
 559 /*
 560  * Regular mount options parser.  Everything that is needed only when
 561  * reading in a new superblock is parsed here.
 562  * XXX JDM: This needs to be cleaned up for remount.
 563  */
 564 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 565                         unsigned long new_flags)
 566 {
 567         substring_t args[MAX_OPT_ARGS];
 568         char *p, *num;
 569         int intarg;
 570         int ret = 0;
 571         char *compress_type;
 572         bool compress_force = false;
 573         enum btrfs_compression_type saved_compress_type;
 574         int saved_compress_level;
 575         bool saved_compress_force;
 576         int no_compress = 0;
 577
 578         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
 579                 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
 580         else if (btrfs_free_space_cache_v1_active(info)) {
 581                 if (btrfs_is_zoned(info)) {
 582                         btrfs_info(info,
 583                         "zoned: clearing existing space cache");
 584                         btrfs_set_super_cache_generation(info->super_copy, 0);
 585                 } else {
 586                         btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 587                 }
 588         }
 589
 590         /*
 591          * Even the options are empty, we still need to do extra check
 592          * against new flags
 593          */
 594         if (!options)
 595                 goto check;
 596
 597         while ((p = strsep(&options, ",")) != NULL) {
 598                 int token;
 599                 if (!*p)
 600                         continue;
 601
 602                 token = match_token(p, tokens, args);
 603                 switch (token) {
 604                 case Opt_degraded:
 605                         btrfs_info(info, "allowing degraded mounts");
 606                         btrfs_set_opt(info->mount_opt, DEGRADED);
 607                         break;
 608                 case Opt_subvol:
 609                 case Opt_subvol_empty:
 610                 case Opt_subvolid:
 611                 case Opt_device:
 612                         /*
 613                          * These are parsed by btrfs_parse_subvol_options or
 614                          * btrfs_parse_device_options and can be ignored here.
 615                          */
 616                         break;
 617                 case Opt_nodatasum:
 618                         btrfs_set_and_info(info, NODATASUM,
 619                                            "setting nodatasum");
 620                         break;
 621                 case Opt_datasum:
 622                         if (btrfs_test_opt(info, NODATASUM)) {
 623                                 if (btrfs_test_opt(info, NODATACOW))
 624                                         btrfs_info(info,
 625                                                    "setting datasum, datacow enabled");
 626                                 else
 627                                         btrfs_info(info, "setting datasum");
 628                         }
 629                         btrfs_clear_opt(info->mount_opt, NODATACOW);
 630                         btrfs_clear_opt(info->mount_opt, NODATASUM);
 631                         break;
 632                 case Opt_nodatacow:
 633                         if (!btrfs_test_opt(info, NODATACOW)) {
 634                                 if (!btrfs_test_opt(info, COMPRESS) ||
 635                                     !btrfs_test_opt(info, FORCE_COMPRESS)) {
 636                                         btrfs_info(info,
 637                                                    "setting nodatacow, compression disabled");
 638                                 } else {
 639                                         btrfs_info(info, "setting nodatacow");
 640                                 }
 641                         }
 642                         btrfs_clear_opt(info->mount_opt, COMPRESS);
 643                         btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 644                         btrfs_set_opt(info->mount_opt, NODATACOW);
 645                         btrfs_set_opt(info->mount_opt, NODATASUM);
 646                         break;
 647                 case Opt_datacow:
 648                         btrfs_clear_and_info(info, NODATACOW,
 649                                              "setting datacow");
 650                         break;
 651                 case Opt_compress_force:
 652                 case Opt_compress_force_type:
 653                         compress_force = true;
 654                         fallthrough;
 655                 case Opt_compress:
 656                 case Opt_compress_type:
 657                         saved_compress_type = btrfs_test_opt(info,
 658                                                              COMPRESS) ?
 659                                 info->compress_type : BTRFS_COMPRESS_NONE;
 660                         saved_compress_force =
 661                                 btrfs_test_opt(info, FORCE_COMPRESS);
 662                         saved_compress_level = info->compress_level;
 663                         if (token == Opt_compress ||
 664                             token == Opt_compress_force ||
 665                             strncmp(args[0].from, "zlib", 4) == 0) {
 666                                 compress_type = "zlib";
 667
 668                                 info->compress_type = BTRFS_COMPRESS_ZLIB;
 669                                 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 670                                 /*
 671                                  * args[0] contains uninitialized data since
 672                                  * for these tokens we don't expect any
 673                                  * parameter.
 674                                  */
 675                                 if (token != Opt_compress &&
 676                                     token != Opt_compress_force)
 677                                         info->compress_level =
 678                                           btrfs_compress_str2level(
 679                                                         BTRFS_COMPRESS_ZLIB,
 680                                                         args[0].from + 4);
 681                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 682                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 683                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 684                                 no_compress = 0;
 685                         } else if (strncmp(args[0].from, "lzo", 3) == 0) {
 686                                 compress_type = "lzo";
 687                                 info->compress_type = BTRFS_COMPRESS_LZO;
 688                                 info->compress_level = 0;
 689                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 690                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 691                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 692                                 btrfs_set_fs_incompat(info, COMPRESS_LZO);
 693                                 no_compress = 0;
 694                         } else if (strncmp(args[0].from, "zstd", 4) == 0) {
 695                                 compress_type = "zstd";
 696                                 info->compress_type = BTRFS_COMPRESS_ZSTD;
 697                                 info->compress_level =
 698                                         btrfs_compress_str2level(
 699                                                          BTRFS_COMPRESS_ZSTD,
 700                                                          args[0].from + 4);
 701                                 btrfs_set_opt(info->mount_opt, COMPRESS);
 702                                 btrfs_clear_opt(info->mount_opt, NODATACOW);
 703                                 btrfs_clear_opt(info->mount_opt, NODATASUM);
 704                                 btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
 705                                 no_compress = 0;
 706                         } else if (strncmp(args[0].from, "no", 2) == 0) {
 707                                 compress_type = "no";
 708                                 info->compress_level = 0;
 709                                 info->compress_type = 0;
 710                                 btrfs_clear_opt(info->mount_opt, COMPRESS);
 711                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 712                                 compress_force = false;
 713                                 no_compress++;
 714                         } else {
 715                                 btrfs_err(info, "unrecognized compression value %s",
 716                                           args[0].from);
 717                                 ret = -EINVAL;
 718                                 goto out;
 719                         }
 720
 721                         if (compress_force) {
 722                                 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 723                         } else {
 724                                 /*
 725                                  * If we remount from compress-force=xxx to
 726                                  * compress=xxx, we need clear FORCE_COMPRESS
 727                                  * flag, otherwise, there is no way for users
 728                                  * to disable forcible compression separately.
 729                                  */
 730                                 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 731                         }
 732                         if (no_compress == 1) {
 733                                 btrfs_info(info, "use no compression");
 734                         } else if ((info->compress_type != saved_compress_type) ||
 735                                    (compress_force != saved_compress_force) ||
 736                                    (info->compress_level != saved_compress_level)) {
 737                                 btrfs_info(info, "%s %s compression, level %d",
 738                                            (compress_force) ? "force" : "use",
 739                                            compress_type, info->compress_level);
 740                         }
 741                         compress_force = false;
 742                         break;
 743                 case Opt_ssd:
 744                         btrfs_set_and_info(info, SSD,
 745                                            "enabling ssd optimizations");
 746                         btrfs_clear_opt(info->mount_opt, NOSSD);
 747                         break;
 748                 case Opt_ssd_spread:
 749                         btrfs_set_and_info(info, SSD,
 750                                            "enabling ssd optimizations");
 751                         btrfs_set_and_info(info, SSD_SPREAD,
 752                                            "using spread ssd allocation scheme");
 753                         btrfs_clear_opt(info->mount_opt, NOSSD);
 754                         break;
 755                 case Opt_nossd:
 756                         btrfs_set_opt(info->mount_opt, NOSSD);
 757                         btrfs_clear_and_info(info, SSD,
 758                                              "not using ssd optimizations");
 759                         fallthrough;
 760                 case Opt_nossd_spread:
 761                         btrfs_clear_and_info(info, SSD_SPREAD,
 762                                              "not using spread ssd allocation scheme");
 763                         break;
 764                 case Opt_barrier:
 765                         btrfs_clear_and_info(info, NOBARRIER,
 766                                              "turning on barriers");
 767                         break;
 768                 case Opt_nobarrier:
 769                         btrfs_set_and_info(info, NOBARRIER,
 770                                            "turning off barriers");
 771                         break;
 772                 case Opt_thread_pool:
 773                         ret = match_int(&args[0], &intarg);
 774                         if (ret) {
 775                                 btrfs_err(info, "unrecognized thread_pool value %s",
 776                                           args[0].from);
 777                                 goto out;
 778                         } else if (intarg == 0) {
 779                                 btrfs_err(info, "invalid value 0 for thread_pool");
 780                                 ret = -EINVAL;
 781                                 goto out;
 782                         }
 783                         info->thread_pool_size = intarg;
 784                         break;
 785                 case Opt_max_inline:
 786                         num = match_strdup(&args[0]);
 787                         if (num) {
 788                                 info->max_inline = memparse(num, NULL);
 789                                 kfree(num);
 790
 791                                 if (info->max_inline) {
 792                                         info->max_inline = min_t(u64,
 793                                                 info->max_inline,
 794                                                 info->sectorsize);
 795                                 }
 796                                 btrfs_info(info, "max_inline at %llu",
 797                                            info->max_inline);
 798                         } else {
 799                                 ret = -ENOMEM;
 800                                 goto out;
 801                         }
 802                         break;
 803                 case Opt_acl:
 804 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 805                         info->sb->s_flags |= SB_POSIXACL;
 806                         break;
 807 #else
 808                         btrfs_err(info, "support for ACL not compiled in!");
 809                         ret = -EINVAL;
 810                         goto out;
 811 #endif
 812                 case Opt_noacl:
 813                         info->sb->s_flags &= ~SB_POSIXACL;
 814                         break;
 815                 case Opt_notreelog:
 816                         btrfs_set_and_info(info, NOTREELOG,
 817                                            "disabling tree log");
 818                         break;
 819                 case Opt_treelog:
 820                         btrfs_clear_and_info(info, NOTREELOG,
 821                                              "enabling tree log");
 822                         break;
 823                 case Opt_norecovery:
 824                 case Opt_nologreplay:
 825                         btrfs_warn(info,
 826                 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 827                         btrfs_set_and_info(info, NOLOGREPLAY,
 828                                            "disabling log replay at mount time");
 829                         break;
 830                 case Opt_flushoncommit:
 831                         btrfs_set_and_info(info, FLUSHONCOMMIT,
 832                                            "turning on flush-on-commit");
 833                         break;
 834                 case Opt_noflushoncommit:
 835                         btrfs_clear_and_info(info, FLUSHONCOMMIT,
 836                                              "turning off flush-on-commit");
 837                         break;
 838                 case Opt_ratio:
 839                         ret = match_int(&args[0], &intarg);
 840                         if (ret) {
 841                                 btrfs_err(info, "unrecognized metadata_ratio value %s",
 842                                           args[0].from);
 843                                 goto out;
 844                         }
 845                         info->metadata_ratio = intarg;
 846                         btrfs_info(info, "metadata ratio %u",
 847                                    info->metadata_ratio);
 848                         break;
 849                 case Opt_discard:
 850                 case Opt_discard_mode:
 851                         if (token == Opt_discard ||
 852                             strcmp(args[0].from, "sync") == 0) {
 853                                 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
 854                                 btrfs_set_and_info(info, DISCARD_SYNC,
 855                                                    "turning on sync discard");
 856                         } else if (strcmp(args[0].from, "async") == 0) {
 857                                 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
 858                                 btrfs_set_and_info(info, DISCARD_ASYNC,
 859                                                    "turning on async discard");
 860                         } else {
 861                                 btrfs_err(info, "unrecognized discard mode value %s",
 862                                           args[0].from);
 863                                 ret = -EINVAL;
 864                                 goto out;
 865                         }
 866                         break;
 867                 case Opt_nodiscard:
 868                         btrfs_clear_and_info(info, DISCARD_SYNC,
 869                                              "turning off discard");
 870                         btrfs_clear_and_info(info, DISCARD_ASYNC,
 871                                              "turning off async discard");
 872                         break;
 873                 case Opt_space_cache:
 874                 case Opt_space_cache_version:
 875                         if (token == Opt_space_cache ||
 876                             strcmp(args[0].from, "v1") == 0) {
 877                                 btrfs_clear_opt(info->mount_opt,
 878                                                 FREE_SPACE_TREE);
 879                                 btrfs_set_and_info(info, SPACE_CACHE,
 880                                            "enabling disk space caching");
 881                         } else if (strcmp(args[0].from, "v2") == 0) {
 882                                 btrfs_clear_opt(info->mount_opt,
 883                                                 SPACE_CACHE);
 884                                 btrfs_set_and_info(info, FREE_SPACE_TREE,
 885                                                    "enabling free space tree");
 886                         } else {
 887                                 btrfs_err(info, "unrecognized space_cache value %s",
 888                                           args[0].from);
 889                                 ret = -EINVAL;
 890                                 goto out;
 891                         }
 892                         break;
 893                 case Opt_rescan_uuid_tree:
 894                         btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 895                         break;
 896                 case Opt_no_space_cache:
 897                         if (btrfs_test_opt(info, SPACE_CACHE)) {
 898                                 btrfs_clear_and_info(info, SPACE_CACHE,
 899                                              "disabling disk space caching");
 900                         }
 901                         if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
 902                                 btrfs_clear_and_info(info, FREE_SPACE_TREE,
 903                                              "disabling free space tree");
 904                         }
 905                         break;
 906                 case Opt_inode_cache:
 907                 case Opt_noinode_cache:
 908                         btrfs_warn(info,
 909         "the 'inode_cache' option is deprecated and has no effect since 5.11");
 910                         break;
 911                 case Opt_clear_cache:
 912                         btrfs_set_and_info(info, CLEAR_CACHE,
 913                                            "force clearing of disk cache");
 914                         break;
 915                 case Opt_user_subvol_rm_allowed:
 916                         btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 917                         break;
 918                 case Opt_enospc_debug:
 919                         btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 920                         break;
 921                 case Opt_noenospc_debug:
 922                         btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 923                         break;
 924                 case Opt_defrag:
 925                         btrfs_set_and_info(info, AUTO_DEFRAG,
 926                                            "enabling auto defrag");
 927                         break;
 928                 case Opt_nodefrag:
 929                         btrfs_clear_and_info(info, AUTO_DEFRAG,
 930                                              "disabling auto defrag");
 931                         break;
 932                 case Opt_recovery:
 933                 case Opt_usebackuproot:
 934                         btrfs_warn(info,
 935                         "'%s' is deprecated, use 'rescue=usebackuproot' instead",
 936                                    token == Opt_recovery ? "recovery" :
 937                                    "usebackuproot");
 938                         btrfs_info(info,
 939                                    "trying to use backup root at mount time");
 940                         btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 941                         break;
 942                 case Opt_skip_balance:
 943                         btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 944                         break;
 945 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 946                 case Opt_check_integrity_including_extent_data:
 947                         btrfs_info(info,
 948                                    "enabling check integrity including extent data");
 949                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
 950                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 951                         break;
 952                 case Opt_check_integrity:
 953                         btrfs_info(info, "enabling check integrity");
 954                         btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 955                         break;
 956                 case Opt_check_integrity_print_mask:
 957                         ret = match_int(&args[0], &intarg);
 958                         if (ret) {
 959                                 btrfs_err(info,
 960                                 "unrecognized check_integrity_print_mask value %s",
 961                                         args[0].from);
 962                                 goto out;
 963                         }
 964                         info->check_integrity_print_mask = intarg;
 965                         btrfs_info(info, "check_integrity_print_mask 0x%x",
 966                                    info->check_integrity_print_mask);
 967                         break;
 968 #else
 969                 case Opt_check_integrity_including_extent_data:
 970                 case Opt_check_integrity:
 971                 case Opt_check_integrity_print_mask:
 972                         btrfs_err(info,
 973                                   "support for check_integrity* not compiled in!");
 974                         ret = -EINVAL;
 975                         goto out;
 976 #endif
 977                 case Opt_fatal_errors:
 978                         if (strcmp(args[0].from, "panic") == 0) {
 979                                 btrfs_set_opt(info->mount_opt,
 980                                               PANIC_ON_FATAL_ERROR);
 981                         } else if (strcmp(args[0].from, "bug") == 0) {
 982                                 btrfs_clear_opt(info->mount_opt,
 983                                               PANIC_ON_FATAL_ERROR);
 984                         } else {
 985                                 btrfs_err(info, "unrecognized fatal_errors value %s",
 986                                           args[0].from);
 987                                 ret = -EINVAL;
 988                                 goto out;
 989                         }
 990                         break;
 991                 case Opt_commit_interval:
 992                         intarg = 0;
 993                         ret = match_int(&args[0], &intarg);
 994                         if (ret) {
 995                                 btrfs_err(info, "unrecognized commit_interval value %s",
 996                                           args[0].from);
 997                                 ret = -EINVAL;
 998                                 goto out;
 999                         }
1000                         if (intarg == 0) {
1001                                 btrfs_info(info,
1002                                            "using default commit interval %us",
1003                                            BTRFS_DEFAULT_COMMIT_INTERVAL);
1004                                 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
1005                         } else if (intarg > 300) {
1006                                 btrfs_warn(info, "excessive commit interval %d",
1007                                            intarg);
1008                         }
1009                         info->commit_interval = intarg;
1010                         break;
1011                 case Opt_rescue:
1012                         ret = parse_rescue_options(info, args[0].from);
1013                         if (ret < 0) {
1014                                 btrfs_err(info, "unrecognized rescue value %s",
1015                                           args[0].from);
1016                                 goto out;
1017                         }
1018                         break;
1019 #ifdef CONFIG_BTRFS_DEBUG
1020                 case Opt_fragment_all:
1021                         btrfs_info(info, "fragmenting all space");
1022                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1023                         btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
1024                         break;
1025                 case Opt_fragment_metadata:
1026                         btrfs_info(info, "fragmenting metadata");
1027                         btrfs_set_opt(info->mount_opt,
1028                                       FRAGMENT_METADATA);
1029                         break;
1030                 case Opt_fragment_data:
1031                         btrfs_info(info, "fragmenting data");
1032                         btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1033                         break;
1034 #endif
1035 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
1036                 case Opt_ref_verify:
1037                         btrfs_info(info, "doing ref verification");
1038                         btrfs_set_opt(info->mount_opt, REF_VERIFY);
1039                         break;
1040 #endif
1041                 case Opt_err:
1042                         btrfs_err(info, "unrecognized mount option '%s'", p);
1043                         ret = -EINVAL;
1044                         goto out;
1045                 default:
1046                         break;
1047                 }
1048         }
1049 check:
1050         /* We're read-only, don't have to check. */
1051         if (new_flags & SB_RDONLY)
1052                 goto out;
1053
1054         if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
1055             check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
1056             check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
1057                 ret = -EINVAL;
1058 out:
1059         if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1060             !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1061             !btrfs_test_opt(info, CLEAR_CACHE)) {
1062                 btrfs_err(info, "cannot disable free space tree");
1063                 ret = -EINVAL;
1064
1065         }
1066         if (!ret)
1067                 ret = btrfs_check_mountopts_zoned(info);
1068         if (!ret && btrfs_test_opt(info, SPACE_CACHE))
1069                 btrfs_info(info, "disk space caching is enabled");
1070         if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
1071                 btrfs_info(info, "using free space tree");
1072         return ret;
1073 }
1074
1075 /*
1076  * Parse mount options that are required early in the mount process.
1077  *
1078  * All other options will be parsed on much later in the mount process and
1079  * only when we need to allocate a new super block.
1080  */
1081 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1082                                       void *holder)
1083 {
1084         substring_t args[MAX_OPT_ARGS];
1085         char *device_name, *opts, *orig, *p;
1086         struct btrfs_device *device = NULL;
1087         int error = 0;
1088
1089         lockdep_assert_held(&uuid_mutex);
1090
1091         if (!options)
1092                 return 0;
1093
1094         /*
1095          * strsep changes the string, duplicate it because btrfs_parse_options
1096          * gets called later
1097          */
1098         opts = kstrdup(options, GFP_KERNEL);
1099         if (!opts)
1100                 return -ENOMEM;
1101         orig = opts;
1102
1103         while ((p = strsep(&opts, ",")) != NULL) {
1104                 int token;
1105
1106                 if (!*p)
1107                         continue;
1108
1109                 token = match_token(p, tokens, args);
1110                 if (token == Opt_device) {
1111                         device_name = match_strdup(&args[0]);
1112                         if (!device_name) {
1113                                 error = -ENOMEM;
1114                                 goto out;
1115                         }
1116                         device = btrfs_scan_one_device(device_name, flags,
1117                                         holder);
1118                         kfree(device_name);
1119                         if (IS_ERR(device)) {
1120                                 error = PTR_ERR(device);
1121                                 goto out;
1122                         }
1123                 }
1124         }
1125
1126 out:
1127         kfree(orig);
1128         return error;
1129 }
1130
1131 /*
1132  * Parse mount options that are related to subvolume id
1133  *
1134  * The value is later passed to mount_subvol()
1135  */
1136 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1137                 u64 *subvol_objectid)
1138 {
1139         substring_t args[MAX_OPT_ARGS];
1140         char *opts, *orig, *p;
1141         int error = 0;
1142         u64 subvolid;
1143
1144         if (!options)
1145                 return 0;
1146
1147         /*
1148          * strsep changes the string, duplicate it because
1149          * btrfs_parse_device_options gets called later
1150          */
1151         opts = kstrdup(options, GFP_KERNEL);
1152         if (!opts)
1153                 return -ENOMEM;
1154         orig = opts;
1155
1156         while ((p = strsep(&opts, ",")) != NULL) {
1157                 int token;
1158                 if (!*p)
1159                         continue;
1160
1161                 token = match_token(p, tokens, args);
1162                 switch (token) {
1163                 case Opt_subvol:
1164                         kfree(*subvol_name);
1165                         *subvol_name = match_strdup(&args[0]);
1166                         if (!*subvol_name) {
1167                                 error = -ENOMEM;
1168                                 goto out;
1169                         }
1170                         break;
1171                 case Opt_subvolid:
1172                         error = match_u64(&args[0], &subvolid);
1173                         if (error)
1174                                 goto out;
1175
1176                         /* we want the original fs_tree */
1177                         if (subvolid == 0)
1178                                 subvolid = BTRFS_FS_TREE_OBJECTID;
1179
1180                         *subvol_objectid = subvolid;
1181                         break;
1182                 default:
1183                         break;
1184                 }
1185         }
1186
1187 out:
1188         kfree(orig);
1189         return error;
1190 }
1191
1192 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1193                                           u64 subvol_objectid)
1194 {
1195         struct btrfs_root *root = fs_info->tree_root;
1196         struct btrfs_root *fs_root = NULL;
1197         struct btrfs_root_ref *root_ref;
1198         struct btrfs_inode_ref *inode_ref;
1199         struct btrfs_key key;
1200         struct btrfs_path *path = NULL;
1201         char *name = NULL, *ptr;
1202         u64 dirid;
1203         int len;
1204         int ret;
1205
1206         path = btrfs_alloc_path();
1207         if (!path) {
1208                 ret = -ENOMEM;
1209                 goto err;
1210         }
1211
1212         name = kmalloc(PATH_MAX, GFP_KERNEL);
1213         if (!name) {
1214                 ret = -ENOMEM;
1215                 goto err;
1216         }
1217         ptr = name + PATH_MAX - 1;
1218         ptr[0] = '\0';
1219
1220         /*
1221          * Walk up the subvolume trees in the tree of tree roots by root
1222          * backrefs until we hit the top-level subvolume.
1223          */
1224         while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1225                 key.objectid = subvol_objectid;
1226                 key.type = BTRFS_ROOT_BACKREF_KEY;
1227                 key.offset = (u64)-1;
1228
1229                 ret = btrfs_search_backwards(root, &key, path);
1230                 if (ret < 0) {
1231                         goto err;
1232                 } else if (ret > 0) {
1233                         ret = -ENOENT;
1234                         goto err;
1235                 }
1236
1237                 subvol_objectid = key.offset;
1238
1239                 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1240                                           struct btrfs_root_ref);
1241                 len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1242                 ptr -= len + 1;
1243                 if (ptr < name) {
1244                         ret = -ENAMETOOLONG;
1245                         goto err;
1246                 }
1247                 read_extent_buffer(path->nodes[0], ptr + 1,
1248                                    (unsigned long)(root_ref + 1), len);
1249                 ptr[0] = '/';
1250                 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1251                 btrfs_release_path(path);
1252
1253                 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1254                 if (IS_ERR(fs_root)) {
1255                         ret = PTR_ERR(fs_root);
1256                         fs_root = NULL;
1257                         goto err;
1258                 }
1259
1260                 /*
1261                  * Walk up the filesystem tree by inode refs until we hit the
1262                  * root directory.
1263                  */
1264                 while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1265                         key.objectid = dirid;
1266                         key.type = BTRFS_INODE_REF_KEY;
1267                         key.offset = (u64)-1;
1268
1269                         ret = btrfs_search_backwards(fs_root, &key, path);
1270                         if (ret < 0) {
1271                                 goto err;
1272                         } else if (ret > 0) {
1273                                 ret = -ENOENT;
1274                                 goto err;
1275                         }
1276
1277                         dirid = key.offset;
1278
1279                         inode_ref = btrfs_item_ptr(path->nodes[0],
1280                                                    path->slots[0],
1281                                                    struct btrfs_inode_ref);
1282                         len = btrfs_inode_ref_name_len(path->nodes[0],
1283                                                        inode_ref);
1284                         ptr -= len + 1;
1285                         if (ptr < name) {
1286                                 ret = -ENAMETOOLONG;
1287                                 goto err;
1288                         }
1289                         read_extent_buffer(path->nodes[0], ptr + 1,
1290                                            (unsigned long)(inode_ref + 1), len);
1291                         ptr[0] = '/';
1292                         btrfs_release_path(path);
1293                 }
1294                 btrfs_put_root(fs_root);
1295                 fs_root = NULL;
1296         }
1297
1298         btrfs_free_path(path);
1299         if (ptr == name + PATH_MAX - 1) {
1300                 name[0] = '/';
1301                 name[1] = '\0';
1302         } else {
1303                 memmove(name, ptr, name + PATH_MAX - ptr);
1304         }
1305         return name;
1306
1307 err:
1308         btrfs_put_root(fs_root);
1309         btrfs_free_path(path);
1310         kfree(name);
1311         return ERR_PTR(ret);
1312 }
1313
1314 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1315 {
1316         struct btrfs_root *root = fs_info->tree_root;
1317         struct btrfs_dir_item *di;
1318         struct btrfs_path *path;
1319         struct btrfs_key location;
1320         u64 dir_id;
1321
1322         path = btrfs_alloc_path();
1323         if (!path)
1324                 return -ENOMEM;
1325
1326         /*
1327          * Find the "default" dir item which points to the root item that we
1328          * will mount by default if we haven't been given a specific subvolume
1329          * to mount.
1330          */
1331         dir_id = btrfs_super_root_dir(fs_info->super_copy);
1332         di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1333         if (IS_ERR(di)) {
1334                 btrfs_free_path(path);
1335                 return PTR_ERR(di);
1336         }
1337         if (!di) {
1338                 /*
1339                  * Ok the default dir item isn't there.  This is weird since
1340                  * it's always been there, but don't freak out, just try and
1341                  * mount the top-level subvolume.
1342                  */
1343                 btrfs_free_path(path);
1344                 *objectid = BTRFS_FS_TREE_OBJECTID;
1345                 return 0;
1346         }
1347
1348         btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1349         btrfs_free_path(path);
1350         *objectid = location.objectid;
1351         return 0;
1352 }
1353
1354 static int btrfs_fill_super(struct super_block *sb,
1355                             struct btrfs_fs_devices *fs_devices,
1356                             void *data)
1357 {
1358         struct inode *inode;
1359         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1360         int err;
1361
1362         sb->s_maxbytes = MAX_LFS_FILESIZE;
1363         sb->s_magic = BTRFS_SUPER_MAGIC;
1364         sb->s_op = &btrfs_super_ops;
1365         sb->s_d_op = &btrfs_dentry_operations;
1366         sb->s_export_op = &btrfs_export_ops;
1367 #ifdef CONFIG_FS_VERITY
1368         sb->s_vop = &btrfs_verityops;
1369 #endif
1370         sb->s_xattr = btrfs_xattr_handlers;
1371         sb->s_time_gran = 1;
1372 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1373         sb->s_flags |= SB_POSIXACL;
1374 #endif
1375         sb->s_flags |= SB_I_VERSION;
1376         sb->s_iflags |= SB_I_CGROUPWB;
1377
1378         err = super_setup_bdi(sb);
1379         if (err) {
1380                 btrfs_err(fs_info, "super_setup_bdi failed");
1381                 return err;
1382         }
1383
1384         err = open_ctree(sb, fs_devices, (char *)data);
1385         if (err) {
1386                 btrfs_err(fs_info, "open_ctree failed");
1387                 return err;
1388         }
1389
1390         inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1391         if (IS_ERR(inode)) {
1392                 err = PTR_ERR(inode);
1393                 goto fail_close;
1394         }
1395
1396         sb->s_root = d_make_root(inode);
1397         if (!sb->s_root) {
1398                 err = -ENOMEM;
1399                 goto fail_close;
1400         }
1401
1402         cleancache_init_fs(sb);
1403         sb->s_flags |= SB_ACTIVE;
1404         return 0;
1405
1406 fail_close:
1407         close_ctree(fs_info);
1408         return err;
1409 }
1410
1411 int btrfs_sync_fs(struct super_block *sb, int wait)
1412 {
1413         struct btrfs_trans_handle *trans;
1414         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1415         struct btrfs_root *root = fs_info->tree_root;
1416
1417         trace_btrfs_sync_fs(fs_info, wait);
1418
1419         if (!wait) {
1420                 filemap_flush(fs_info->btree_inode->i_mapping);
1421                 return 0;
1422         }
1423
1424         btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1425
1426         trans = btrfs_attach_transaction_barrier(root);
1427         if (IS_ERR(trans)) {
1428                 /* no transaction, don't bother */
1429                 if (PTR_ERR(trans) == -ENOENT) {
1430                         /*
1431                          * Exit unless we have some pending changes
1432                          * that need to go through commit
1433                          */
1434                         if (fs_info->pending_changes == 0)
1435                                 return 0;
1436                         /*
1437                          * A non-blocking test if the fs is frozen. We must not
1438                          * start a new transaction here otherwise a deadlock
1439                          * happens. The pending operations are delayed to the
1440                          * next commit after thawing.
1441                          */
1442                         if (sb_start_write_trylock(sb))
1443                                 sb_end_write(sb);
1444                         else
1445                                 return 0;
1446                         trans = btrfs_start_transaction(root, 0);
1447                 }
1448                 if (IS_ERR(trans))
1449                         return PTR_ERR(trans);
1450         }
1451         return btrfs_commit_transaction(trans);
1452 }
1453
1454 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1455 {
1456         seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1457         *printed = true;
1458 }
1459
1460 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1461 {
1462         struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1463         const char *compress_type;
1464         const char *subvol_name;
1465         bool printed = false;
1466
1467         if (btrfs_test_opt(info, DEGRADED))
1468                 seq_puts(seq, ",degraded");
1469         if (btrfs_test_opt(info, NODATASUM))
1470                 seq_puts(seq, ",nodatasum");
1471         if (btrfs_test_opt(info, NODATACOW))
1472                 seq_puts(seq, ",nodatacow");
1473         if (btrfs_test_opt(info, NOBARRIER))
1474                 seq_puts(seq, ",nobarrier");
1475         if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1476                 seq_printf(seq, ",max_inline=%llu", info->max_inline);
1477         if (info->thread_pool_size !=  min_t(unsigned long,
1478                                              num_online_cpus() + 2, 8))
1479                 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1480         if (btrfs_test_opt(info, COMPRESS)) {
1481                 compress_type = btrfs_compress_type2str(info->compress_type);
1482                 if (btrfs_test_opt(info, FORCE_COMPRESS))
1483                         seq_printf(seq, ",compress-force=%s", compress_type);
1484                 else
1485                         seq_printf(seq, ",compress=%s", compress_type);
1486                 if (info->compress_level)
1487                         seq_printf(seq, ":%d", info->compress_level);
1488         }
1489         if (btrfs_test_opt(info, NOSSD))
1490                 seq_puts(seq, ",nossd");
1491         if (btrfs_test_opt(info, SSD_SPREAD))
1492                 seq_puts(seq, ",ssd_spread");
1493         else if (btrfs_test_opt(info, SSD))
1494                 seq_puts(seq, ",ssd");
1495         if (btrfs_test_opt(info, NOTREELOG))
1496                 seq_puts(seq, ",notreelog");
1497         if (btrfs_test_opt(info, NOLOGREPLAY))
1498                 print_rescue_option(seq, "nologreplay", &printed);
1499         if (btrfs_test_opt(info, USEBACKUPROOT))
1500                 print_rescue_option(seq, "usebackuproot", &printed);
1501         if (btrfs_test_opt(info, IGNOREBADROOTS))
1502                 print_rescue_option(seq, "ignorebadroots", &printed);
1503         if (btrfs_test_opt(info, IGNOREDATACSUMS))
1504                 print_rescue_option(seq, "ignoredatacsums", &printed);
1505         if (btrfs_test_opt(info, FLUSHONCOMMIT))
1506                 seq_puts(seq, ",flushoncommit");
1507         if (btrfs_test_opt(info, DISCARD_SYNC))
1508                 seq_puts(seq, ",discard");
1509         if (btrfs_test_opt(info, DISCARD_ASYNC))
1510                 seq_puts(seq, ",discard=async");
1511         if (!(info->sb->s_flags & SB_POSIXACL))
1512                 seq_puts(seq, ",noacl");
1513         if (btrfs_free_space_cache_v1_active(info))
1514                 seq_puts(seq, ",space_cache");
1515         else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1516                 seq_puts(seq, ",space_cache=v2");
1517         else
1518                 seq_puts(seq, ",nospace_cache");
1519         if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1520                 seq_puts(seq, ",rescan_uuid_tree");
1521         if (btrfs_test_opt(info, CLEAR_CACHE))
1522                 seq_puts(seq, ",clear_cache");
1523         if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1524                 seq_puts(seq, ",user_subvol_rm_allowed");
1525         if (btrfs_test_opt(info, ENOSPC_DEBUG))
1526                 seq_puts(seq, ",enospc_debug");
1527         if (btrfs_test_opt(info, AUTO_DEFRAG))
1528                 seq_puts(seq, ",autodefrag");
1529         if (btrfs_test_opt(info, SKIP_BALANCE))
1530                 seq_puts(seq, ",skip_balance");
1531 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1532         if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1533                 seq_puts(seq, ",check_int_data");
1534         else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1535                 seq_puts(seq, ",check_int");
1536         if (info->check_integrity_print_mask)
1537                 seq_printf(seq, ",check_int_print_mask=%d",
1538                                 info->check_integrity_print_mask);
1539 #endif
1540         if (info->metadata_ratio)
1541                 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1542         if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1543                 seq_puts(seq, ",fatal_errors=panic");
1544         if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1545                 seq_printf(seq, ",commit=%u", info->commit_interval);
1546 #ifdef CONFIG_BTRFS_DEBUG
1547         if (btrfs_test_opt(info, FRAGMENT_DATA))
1548                 seq_puts(seq, ",fragment=data");
1549         if (btrfs_test_opt(info, FRAGMENT_METADATA))
1550                 seq_puts(seq, ",fragment=metadata");
1551 #endif
1552         if (btrfs_test_opt(info, REF_VERIFY))
1553                 seq_puts(seq, ",ref_verify");
1554         seq_printf(seq, ",subvolid=%llu",
1555                   BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1556         subvol_name = btrfs_get_subvol_name_from_objectid(info,
1557                         BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1558         if (!IS_ERR(subvol_name)) {
1559                 seq_puts(seq, ",subvol=");
1560                 seq_escape(seq, subvol_name, " \t\n\\");
1561                 kfree(subvol_name);
1562         }
1563         return 0;
1564 }
1565
1566 static int btrfs_test_super(struct super_block *s, void *data)
1567 {
1568         struct btrfs_fs_info *p = data;
1569         struct btrfs_fs_info *fs_info = btrfs_sb(s);
1570
1571         return fs_info->fs_devices == p->fs_devices;
1572 }
1573
1574 static int btrfs_set_super(struct super_block *s, void *data)
1575 {
1576         int err = set_anon_super(s, data);
1577         if (!err)
1578                 s->s_fs_info = data;
1579         return err;
1580 }
1581
1582 /*
1583  * subvolumes are identified by ino 256
1584  */
1585 static inline int is_subvolume_inode(struct inode *inode)
1586 {
1587         if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1588                 return 1;
1589         return 0;
1590 }
1591
1592 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1593                                    struct vfsmount *mnt)
1594 {
1595         struct dentry *root;
1596         int ret;
1597
1598         if (!subvol_name) {
1599                 if (!subvol_objectid) {
1600                         ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1601                                                           &subvol_objectid);
1602                         if (ret) {
1603                                 root = ERR_PTR(ret);
1604                                 goto out;
1605                         }
1606                 }
1607                 subvol_name = btrfs_get_subvol_name_from_objectid(
1608                                         btrfs_sb(mnt->mnt_sb), subvol_objectid);
1609                 if (IS_ERR(subvol_name)) {
1610                         root = ERR_CAST(subvol_name);
1611                         subvol_name = NULL;
1612                         goto out;
1613                 }
1614
1615         }
1616
1617         root = mount_subtree(mnt, subvol_name);
1618         /* mount_subtree() drops our reference on the vfsmount. */
1619         mnt = NULL;
1620
1621         if (!IS_ERR(root)) {
1622                 struct super_block *s = root->d_sb;
1623                 struct btrfs_fs_info *fs_info = btrfs_sb(s);
1624                 struct inode *root_inode = d_inode(root);
1625                 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1626
1627                 ret = 0;
1628                 if (!is_subvolume_inode(root_inode)) {
1629                         btrfs_err(fs_info, "'%s' is not a valid subvolume",
1630                                subvol_name);
1631                         ret = -EINVAL;
1632                 }
1633                 if (subvol_objectid && root_objectid != subvol_objectid) {
1634                         /*
1635                          * This will also catch a race condition where a
1636                          * subvolume which was passed by ID is renamed and
1637                          * another subvolume is renamed over the old location.
1638                          */
1639                         btrfs_err(fs_info,
1640                                   "subvol '%s' does not match subvolid %llu",
1641                                   subvol_name, subvol_objectid);
1642                         ret = -EINVAL;
1643                 }
1644                 if (ret) {
1645                         dput(root);
1646                         root = ERR_PTR(ret);
1647                         deactivate_locked_super(s);
1648                 }
1649         }
1650
1651 out:
1652         mntput(mnt);
1653         kfree(subvol_name);
1654         return root;
1655 }
1656
1657 /*
1658  * Find a superblock for the given device / mount point.
1659  *
1660  * Note: This is based on mount_bdev from fs/super.c with a few additions
1661  *       for multiple device setup.  Make sure to keep it in sync.
1662  */
1663 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1664                 int flags, const char *device_name, void *data)
1665 {
1666         struct block_device *bdev = NULL;
1667         struct super_block *s;
1668         struct btrfs_device *device = NULL;
1669         struct btrfs_fs_devices *fs_devices = NULL;
1670         struct btrfs_fs_info *fs_info = NULL;
1671         void *new_sec_opts = NULL;
1672         fmode_t mode = FMODE_READ;
1673         int error = 0;
1674
1675         if (!(flags & SB_RDONLY))
1676                 mode |= FMODE_WRITE;
1677
1678         if (data) {
1679                 error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1680                 if (error)
1681                         return ERR_PTR(error);
1682         }
1683
1684         /*
1685          * Setup a dummy root and fs_info for test/set super.  This is because
1686          * we don't actually fill this stuff out until open_ctree, but we need
1687          * then open_ctree will properly initialize the file system specific
1688          * settings later.  btrfs_init_fs_info initializes the static elements
1689          * of the fs_info (locks and such) to make cleanup easier if we find a
1690          * superblock with our given fs_devices later on at sget() time.
1691          */
1692         fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1693         if (!fs_info) {
1694                 error = -ENOMEM;
1695                 goto error_sec_opts;
1696         }
1697         btrfs_init_fs_info(fs_info);
1698
1699         fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1700         fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1701         if (!fs_info->super_copy || !fs_info->super_for_commit) {
1702                 error = -ENOMEM;
1703                 goto error_fs_info;
1704         }
1705
1706         mutex_lock(&uuid_mutex);
1707         error = btrfs_parse_device_options(data, mode, fs_type);
1708         if (error) {
1709                 mutex_unlock(&uuid_mutex);
1710                 goto error_fs_info;
1711         }
1712
1713         device = btrfs_scan_one_device(device_name, mode, fs_type);
1714         if (IS_ERR(device)) {
1715                 mutex_unlock(&uuid_mutex);
1716                 error = PTR_ERR(device);
1717                 goto error_fs_info;
1718         }
1719
1720         fs_devices = device->fs_devices;
1721         fs_info->fs_devices = fs_devices;
1722
1723         error = btrfs_open_devices(fs_devices, mode, fs_type);
1724         mutex_unlock(&uuid_mutex);
1725         if (error)
1726                 goto error_fs_info;
1727
1728         if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1729                 error = -EACCES;
1730                 goto error_close_devices;
1731         }
1732
1733         bdev = fs_devices->latest_dev->bdev;
1734         s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1735                  fs_info);
1736         if (IS_ERR(s)) {
1737                 error = PTR_ERR(s);
1738                 goto error_close_devices;
1739         }
1740
1741         if (s->s_root) {
1742                 btrfs_close_devices(fs_devices);
1743                 btrfs_free_fs_info(fs_info);
1744                 if ((flags ^ s->s_flags) & SB_RDONLY)
1745                         error = -EBUSY;
1746         } else {
1747                 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1748                 btrfs_sb(s)->bdev_holder = fs_type;
1749                 if (!strstr(crc32c_impl(), "generic"))
1750                         set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1751                 error = btrfs_fill_super(s, fs_devices, data);
1752         }
1753         if (!error)
1754                 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1755         security_free_mnt_opts(&new_sec_opts);
1756         if (error) {
1757                 deactivate_locked_super(s);
1758                 return ERR_PTR(error);
1759         }
1760
1761         return dget(s->s_root);
1762
1763 error_close_devices:
1764         btrfs_close_devices(fs_devices);
1765 error_fs_info:
1766         btrfs_free_fs_info(fs_info);
1767 error_sec_opts:
1768         security_free_mnt_opts(&new_sec_opts);
1769         return ERR_PTR(error);
1770 }
1771
1772 /*
1773  * Mount function which is called by VFS layer.
1774  *
1775  * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1776  * which needs vfsmount* of device's root (/).  This means device's root has to
1777  * be mounted internally in any case.
1778  *
1779  * Operation flow:
1780  *   1. Parse subvol id related options for later use in mount_subvol().
1781  *
1782  *   2. Mount device's root (/) by calling vfs_kern_mount().
1783  *
1784  *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1785  *      first place. In order to avoid calling btrfs_mount() again, we use
1786  *      different file_system_type which is not registered to VFS by
1787  *      register_filesystem() (btrfs_root_fs_type). As a result,
1788  *      btrfs_mount_root() is called. The return value will be used by
1789  *      mount_subtree() in mount_subvol().
1790  *
1791  *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1792  *      "btrfs subvolume set-default", mount_subvol() is called always.
1793  */
1794 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1795                 const char *device_name, void *data)
1796 {
1797         struct vfsmount *mnt_root;
1798         struct dentry *root;
1799         char *subvol_name = NULL;
1800         u64 subvol_objectid = 0;
1801         int error = 0;
1802
1803         error = btrfs_parse_subvol_options(data, &subvol_name,
1804                                         &subvol_objectid);
1805         if (error) {
1806                 kfree(subvol_name);
1807                 return ERR_PTR(error);
1808         }
1809
1810         /* mount device's root (/) */
1811         mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1812         if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1813                 if (flags & SB_RDONLY) {
1814                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1815                                 flags & ~SB_RDONLY, device_name, data);
1816                 } else {
1817                         mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1818                                 flags | SB_RDONLY, device_name, data);
1819                         if (IS_ERR(mnt_root)) {
1820                                 root = ERR_CAST(mnt_root);
1821                                 kfree(subvol_name);
1822                                 goto out;
1823                         }
1824
1825                         down_write(&mnt_root->mnt_sb->s_umount);
1826                         error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1827                         up_write(&mnt_root->mnt_sb->s_umount);
1828                         if (error < 0) {
1829                                 root = ERR_PTR(error);
1830                                 mntput(mnt_root);
1831                                 kfree(subvol_name);
1832                                 goto out;
1833                         }
1834                 }
1835         }
1836         if (IS_ERR(mnt_root)) {
1837                 root = ERR_CAST(mnt_root);
1838                 kfree(subvol_name);
1839                 goto out;
1840         }
1841
1842         /* mount_subvol() will free subvol_name and mnt_root */
1843         root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1844
1845 out:
1846         return root;
1847 }
1848
1849 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1850                                      u32 new_pool_size, u32 old_pool_size)
1851 {
1852         if (new_pool_size == old_pool_size)
1853                 return;
1854
1855         fs_info->thread_pool_size = new_pool_size;
1856
1857         btrfs_info(fs_info, "resize thread pool %d -> %d",
1858                old_pool_size, new_pool_size);
1859
1860         btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1861         btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1862         btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1863         btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1864         btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1865         btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1866                                 new_pool_size);
1867         btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1868         btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1869         btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1870         btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1871         btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1872                                 new_pool_size);
1873 }
1874
1875 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1876                                        unsigned long old_opts, int flags)
1877 {
1878         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1879             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1880              (flags & SB_RDONLY))) {
1881                 /* wait for any defraggers to finish */
1882                 wait_event(fs_info->transaction_wait,
1883                            (atomic_read(&fs_info->defrag_running) == 0));
1884                 if (flags & SB_RDONLY)
1885                         sync_filesystem(fs_info->sb);
1886         }
1887 }
1888
1889 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1890                                          unsigned long old_opts)
1891 {
1892         const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1893
1894         /*
1895          * We need to cleanup all defragable inodes if the autodefragment is
1896          * close or the filesystem is read only.
1897          */
1898         if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1899             (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1900                 btrfs_cleanup_defrag_inodes(fs_info);
1901         }
1902
1903         /* If we toggled discard async */
1904         if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1905             btrfs_test_opt(fs_info, DISCARD_ASYNC))
1906                 btrfs_discard_resume(fs_info);
1907         else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1908                  !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1909                 btrfs_discard_cleanup(fs_info);
1910
1911         /* If we toggled space cache */
1912         if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1913                 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1914 }
1915
1916 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1917 {
1918         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1919         unsigned old_flags = sb->s_flags;
1920         unsigned long old_opts = fs_info->mount_opt;
1921         unsigned long old_compress_type = fs_info->compress_type;
1922         u64 old_max_inline = fs_info->max_inline;
1923         u32 old_thread_pool_size = fs_info->thread_pool_size;
1924         u32 old_metadata_ratio = fs_info->metadata_ratio;
1925         int ret;
1926
1927         sync_filesystem(sb);
1928         set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1929
1930         if (data) {
1931                 void *new_sec_opts = NULL;
1932
1933                 ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1934                 if (!ret)
1935                         ret = security_sb_remount(sb, new_sec_opts);
1936                 security_free_mnt_opts(&new_sec_opts);
1937                 if (ret)
1938                         goto restore;
1939         }
1940
1941         ret = btrfs_parse_options(fs_info, data, *flags);
1942         if (ret)
1943                 goto restore;
1944
1945         /* V1 cache is not supported for subpage mount. */
1946         if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) {
1947                 btrfs_warn(fs_info,
1948         "v1 space cache is not supported for page size %lu with sectorsize %u",
1949                            PAGE_SIZE, fs_info->sectorsize);
1950                 ret = -EINVAL;
1951                 goto restore;
1952         }
1953         btrfs_remount_begin(fs_info, old_opts, *flags);
1954         btrfs_resize_thread_pool(fs_info,
1955                 fs_info->thread_pool_size, old_thread_pool_size);
1956
1957         if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1958             (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1959             (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1960                 btrfs_warn(fs_info,
1961                 "remount supports changing free space tree only from ro to rw");
1962                 /* Make sure free space cache options match the state on disk */
1963                 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1964                         btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1965                         btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1966                 }
1967                 if (btrfs_free_space_cache_v1_active(fs_info)) {
1968                         btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1969                         btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1970                 }
1971         }
1972
1973         if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1974                 goto out;
1975
1976         if (*flags & SB_RDONLY) {
1977                 /*
1978                  * this also happens on 'umount -rf' or on shutdown, when
1979                  * the filesystem is busy.
1980                  */
1981                 cancel_work_sync(&fs_info->async_reclaim_work);
1982                 cancel_work_sync(&fs_info->async_data_reclaim_work);
1983
1984                 btrfs_discard_cleanup(fs_info);
1985
1986                 /* wait for the uuid_scan task to finish */
1987                 down(&fs_info->uuid_tree_rescan_sem);
1988                 /* avoid complains from lockdep et al. */
1989                 up(&fs_info->uuid_tree_rescan_sem);
1990
1991                 btrfs_set_sb_rdonly(sb);
1992
1993                 /*
1994                  * Setting SB_RDONLY will put the cleaner thread to
1995                  * sleep at the next loop if it's already active.
1996                  * If it's already asleep, we'll leave unused block
1997                  * groups on disk until we're mounted read-write again
1998                  * unless we clean them up here.
1999                  */
2000                 btrfs_delete_unused_bgs(fs_info);
2001
2002                 /*
2003                  * The cleaner task could be already running before we set the
2004                  * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
2005                  * We must make sure that after we finish the remount, i.e. after
2006                  * we call btrfs_commit_super(), the cleaner can no longer start
2007                  * a transaction - either because it was dropping a dead root,
2008                  * running delayed iputs or deleting an unused block group (the
2009                  * cleaner picked a block group from the list of unused block
2010                  * groups before we were able to in the previous call to
2011                  * btrfs_delete_unused_bgs()).
2012                  */
2013                 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
2014                             TASK_UNINTERRUPTIBLE);
2015
2016                 /*
2017                  * We've set the superblock to RO mode, so we might have made
2018                  * the cleaner task sleep without running all pending delayed
2019                  * iputs. Go through all the delayed iputs here, so that if an
2020                  * unmount happens without remounting RW we don't end up at
2021                  * finishing close_ctree() with a non-empty list of delayed
2022                  * iputs.
2023                  */
2024                 btrfs_run_delayed_iputs(fs_info);
2025
2026                 btrfs_dev_replace_suspend_for_unmount(fs_info);
2027                 btrfs_scrub_cancel(fs_info);
2028                 btrfs_pause_balance(fs_info);
2029
2030                 /*
2031                  * Pause the qgroup rescan worker if it is running. We don't want
2032                  * it to be still running after we are in RO mode, as after that,
2033                  * by the time we unmount, it might have left a transaction open,
2034                  * so we would leak the transaction and/or crash.
2035                  */
2036                 btrfs_qgroup_wait_for_completion(fs_info, false);
2037
2038                 ret = btrfs_commit_super(fs_info);
2039                 if (ret)
2040                         goto restore;
2041         } else {
2042                 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2043                         btrfs_err(fs_info,
2044                                 "Remounting read-write after error is not allowed");
2045                         ret = -EINVAL;
2046                         goto restore;
2047                 }
2048                 if (fs_info->fs_devices->rw_devices == 0) {
2049                         ret = -EACCES;
2050                         goto restore;
2051                 }
2052
2053                 if (!btrfs_check_rw_degradable(fs_info, NULL)) {
2054                         btrfs_warn(fs_info,
2055                 "too many missing devices, writable remount is not allowed");
2056                         ret = -EACCES;
2057                         goto restore;
2058                 }
2059
2060                 if (btrfs_super_log_root(fs_info->super_copy) != 0) {
2061                         btrfs_warn(fs_info,
2062                 "mount required to replay tree-log, cannot remount read-write");
2063                         ret = -EINVAL;
2064                         goto restore;
2065                 }
2066
2067                 /*
2068                  * NOTE: when remounting with a change that does writes, don't
2069                  * put it anywhere above this point, as we are not sure to be
2070                  * safe to write until we pass the above checks.
2071                  */
2072                 ret = btrfs_start_pre_rw_mount(fs_info);
2073                 if (ret)
2074                         goto restore;
2075
2076                 btrfs_clear_sb_rdonly(sb);
2077
2078                 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
2079         }
2080 out:
2081         /*
2082          * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2083          * since the absence of the flag means it can be toggled off by remount.
2084          */
2085         *flags |= SB_I_VERSION;
2086
2087         wake_up_process(fs_info->transaction_kthread);
2088         btrfs_remount_cleanup(fs_info, old_opts);
2089         btrfs_clear_oneshot_options(fs_info);
2090         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2091
2092         return 0;
2093
2094 restore:
2095         /* We've hit an error - don't reset SB_RDONLY */
2096         if (sb_rdonly(sb))
2097                 old_flags |= SB_RDONLY;
2098         if (!(old_flags & SB_RDONLY))
2099                 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
2100         sb->s_flags = old_flags;
2101         fs_info->mount_opt = old_opts;
2102         fs_info->compress_type = old_compress_type;
2103         fs_info->max_inline = old_max_inline;
2104         btrfs_resize_thread_pool(fs_info,
2105                 old_thread_pool_size, fs_info->thread_pool_size);
2106         fs_info->metadata_ratio = old_metadata_ratio;
2107         btrfs_remount_cleanup(fs_info, old_opts);
2108         clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2109
2110         return ret;
2111 }
2112
2113 /* Used to sort the devices by max_avail(descending sort) */
2114 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
2115 {
2116         const struct btrfs_device_info *dev_info1 = a;
2117         const struct btrfs_device_info *dev_info2 = b;
2118
2119         if (dev_info1->max_avail > dev_info2->max_avail)
2120                 return -1;
2121         else if (dev_info1->max_avail < dev_info2->max_avail)
2122                 return 1;
2123         return 0;
2124 }
2125
2126 /*
2127  * sort the devices by max_avail, in which max free extent size of each device
2128  * is stored.(Descending Sort)
2129  */
2130 static inline void btrfs_descending_sort_devices(
2131                                         struct btrfs_device_info *devices,
2132                                         size_t nr_devices)
2133 {
2134         sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2135              btrfs_cmp_device_free_bytes, NULL);
2136 }
2137
2138 /*
2139  * The helper to calc the free space on the devices that can be used to store
2140  * file data.
2141  */
2142 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2143                                               u64 *free_bytes)
2144 {
2145         struct btrfs_device_info *devices_info;
2146         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2147         struct btrfs_device *device;
2148         u64 type;
2149         u64 avail_space;
2150         u64 min_stripe_size;
2151         int num_stripes = 1;
2152         int i = 0, nr_devices;
2153         const struct btrfs_raid_attr *rattr;
2154
2155         /*
2156          * We aren't under the device list lock, so this is racy-ish, but good
2157          * enough for our purposes.
2158          */
2159         nr_devices = fs_info->fs_devices->open_devices;
2160         if (!nr_devices) {
2161                 smp_mb();
2162                 nr_devices = fs_info->fs_devices->open_devices;
2163                 ASSERT(nr_devices);
2164                 if (!nr_devices) {
2165                         *free_bytes = 0;
2166                         return 0;
2167                 }
2168         }
2169
2170         devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2171                                GFP_KERNEL);
2172         if (!devices_info)
2173                 return -ENOMEM;
2174
2175         /* calc min stripe number for data space allocation */
2176         type = btrfs_data_alloc_profile(fs_info);
2177         rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2178
2179         if (type & BTRFS_BLOCK_GROUP_RAID0)
2180                 num_stripes = nr_devices;
2181         else if (type & BTRFS_BLOCK_GROUP_RAID1)
2182                 num_stripes = 2;
2183         else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2184                 num_stripes = 3;
2185         else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2186                 num_stripes = 4;
2187         else if (type & BTRFS_BLOCK_GROUP_RAID10)
2188                 num_stripes = 4;
2189
2190         /* Adjust for more than 1 stripe per device */
2191         min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2192
2193         rcu_read_lock();
2194         list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2195                 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2196                                                 &device->dev_state) ||
2197                     !device->bdev ||
2198                     test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2199                         continue;
2200
2201                 if (i >= nr_devices)
2202                         break;
2203
2204                 avail_space = device->total_bytes - device->bytes_used;
2205
2206                 /* align with stripe_len */
2207                 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2208
2209                 /*
2210                  * In order to avoid overwriting the superblock on the drive,
2211                  * btrfs starts at an offset of at least 1MB when doing chunk
2212                  * allocation.
2213                  *
2214                  * This ensures we have at least min_stripe_size free space
2215                  * after excluding 1MB.
2216                  */
2217                 if (avail_space <= SZ_1M + min_stripe_size)
2218                         continue;
2219
2220                 avail_space -= SZ_1M;
2221
2222                 devices_info[i].dev = device;
2223                 devices_info[i].max_avail = avail_space;
2224
2225                 i++;
2226         }
2227         rcu_read_unlock();
2228
2229         nr_devices = i;
2230
2231         btrfs_descending_sort_devices(devices_info, nr_devices);
2232
2233         i = nr_devices - 1;
2234         avail_space = 0;
2235         while (nr_devices >= rattr->devs_min) {
2236                 num_stripes = min(num_stripes, nr_devices);
2237
2238                 if (devices_info[i].max_avail >= min_stripe_size) {
2239                         int j;
2240                         u64 alloc_size;
2241
2242                         avail_space += devices_info[i].max_avail * num_stripes;
2243                         alloc_size = devices_info[i].max_avail;
2244                         for (j = i + 1 - num_stripes; j <= i; j++)
2245                                 devices_info[j].max_avail -= alloc_size;
2246                 }
2247                 i--;
2248                 nr_devices--;
2249         }
2250
2251         kfree(devices_info);
2252         *free_bytes = avail_space;
2253         return 0;
2254 }
2255
2256 /*
2257  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2258  *
2259  * If there's a redundant raid level at DATA block groups, use the respective
2260  * multiplier to scale the sizes.
2261  *
2262  * Unused device space usage is based on simulating the chunk allocator
2263  * algorithm that respects the device sizes and order of allocations.  This is
2264  * a close approximation of the actual use but there are other factors that may
2265  * change the result (like a new metadata chunk).
2266  *
2267  * If metadata is exhausted, f_bavail will be 0.
2268  */
2269 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2270 {
2271         struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2272         struct btrfs_super_block *disk_super = fs_info->super_copy;
2273         struct btrfs_space_info *found;
2274         u64 total_used = 0;
2275         u64 total_free_data = 0;
2276         u64 total_free_meta = 0;
2277         u32 bits = fs_info->sectorsize_bits;
2278         __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2279         unsigned factor = 1;
2280         struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2281         int ret;
2282         u64 thresh = 0;
2283         int mixed = 0;
2284
2285         list_for_each_entry(found, &fs_info->space_info, list) {
2286                 if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2287                         int i;
2288
2289                         total_free_data += found->disk_total - found->disk_used;
2290                         total_free_data -=
2291                                 btrfs_account_ro_block_groups_free_space(found);
2292
2293                         for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2294                                 if (!list_empty(&found->block_groups[i]))
2295                                         factor = btrfs_bg_type_to_factor(
2296                                                 btrfs_raid_array[i].bg_flag);
2297                         }
2298                 }
2299
2300                 /*
2301                  * Metadata in mixed block goup profiles are accounted in data
2302                  */
2303                 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2304                         if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2305                                 mixed = 1;
2306                         else
2307                                 total_free_meta += found->disk_total -
2308                                         found->disk_used;
2309                 }
2310
2311                 total_used += found->disk_used;
2312         }
2313
2314         buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2315         buf->f_blocks >>= bits;
2316         buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2317
2318         /* Account global block reserve as used, it's in logical size already */
2319         spin_lock(&block_rsv->lock);
2320         /* Mixed block groups accounting is not byte-accurate, avoid overflow */
2321         if (buf->f_bfree >= block_rsv->size >> bits)
2322                 buf->f_bfree -= block_rsv->size >> bits;
2323         else
2324                 buf->f_bfree = 0;
2325         spin_unlock(&block_rsv->lock);
2326
2327         buf->f_bavail = div_u64(total_free_data, factor);
2328         ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2329         if (ret)
2330                 return ret;
2331         buf->f_bavail += div_u64(total_free_data, factor);
2332         buf->f_bavail = buf->f_bavail >> bits;
2333
2334         /*
2335          * We calculate the remaining metadata space minus global reserve. If
2336          * this is (supposedly) smaller than zero, there's no space. But this
2337          * does not hold in practice, the exhausted state happens where's still
2338          * some positive delta. So we apply some guesswork and compare the
2339          * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2340          *
2341          * We probably cannot calculate the exact threshold value because this
2342          * depends on the internal reservations requested by various
2343          * operations, so some operations that consume a few metadata will
2344          * succeed even if the Avail is zero. But this is better than the other
2345          * way around.
2346          */
2347         thresh = SZ_4M;
2348
2349         /*
2350          * We only want to claim there's no available space if we can no longer
2351          * allocate chunks for our metadata profile and our global reserve will
2352          * not fit in the free metadata space.  If we aren't ->full then we
2353          * still can allocate chunks and thus are fine using the currently
2354          * calculated f_bavail.
2355          */
2356         if (!mixed && block_rsv->space_info->full &&
2357             total_free_meta - thresh < block_rsv->size)
2358                 buf->f_bavail = 0;
2359
2360         buf->f_type = BTRFS_SUPER_MAGIC;
2361         buf->f_bsize = dentry->d_sb->s_blocksize;
2362         buf->f_namelen = BTRFS_NAME_LEN;
2363
2364         /* We treat it as constant endianness (it doesn't matter _which_)
2365            because we want the fsid to come out the same whether mounted
2366            on a big-endian or little-endian host */
2367         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2368         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2369         /* Mask in the root object ID too, to disambiguate subvols */
2370         buf->f_fsid.val[0] ^=
2371                 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2372         buf->f_fsid.val[1] ^=
2373                 BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2374
2375         return 0;
2376 }
2377
2378 static void btrfs_kill_super(struct super_block *sb)
2379 {
2380         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2381         kill_anon_super(sb);
2382         btrfs_free_fs_info(fs_info);
2383 }
2384
2385 static struct file_system_type btrfs_fs_type = {
2386         .owner          = THIS_MODULE,
2387         .name           = "btrfs",
2388         .mount          = btrfs_mount,
2389         .kill_sb        = btrfs_kill_super,
2390         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2391 };
2392
2393 static struct file_system_type btrfs_root_fs_type = {
2394         .owner          = THIS_MODULE,
2395         .name           = "btrfs",
2396         .mount          = btrfs_mount_root,
2397         .kill_sb        = btrfs_kill_super,
2398         .fs_flags       = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2399 };
2400
2401 MODULE_ALIAS_FS("btrfs");
2402
2403 static int btrfs_control_open(struct inode *inode, struct file *file)
2404 {
2405         /*
2406          * The control file's private_data is used to hold the
2407          * transaction when it is started and is used to keep
2408          * track of whether a transaction is already in progress.
2409          */
2410         file->private_data = NULL;
2411         return 0;
2412 }
2413
2414 /*
2415  * Used by /dev/btrfs-control for devices ioctls.
2416  */
2417 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2418                                 unsigned long arg)
2419 {
2420         struct btrfs_ioctl_vol_args *vol;
2421         struct btrfs_device *device = NULL;
2422         int ret = -ENOTTY;
2423
2424         if (!capable(CAP_SYS_ADMIN))
2425                 return -EPERM;
2426
2427         vol = memdup_user((void __user *)arg, sizeof(*vol));
2428         if (IS_ERR(vol))
2429                 return PTR_ERR(vol);
2430         vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2431
2432         switch (cmd) {
2433         case BTRFS_IOC_SCAN_DEV:
2434                 mutex_lock(&uuid_mutex);
2435                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2436                                                &btrfs_root_fs_type);
2437                 ret = PTR_ERR_OR_ZERO(device);
2438                 mutex_unlock(&uuid_mutex);
2439                 break;
2440         case BTRFS_IOC_FORGET_DEV:
2441                 ret = btrfs_forget_devices(vol->name);
2442                 break;
2443         case BTRFS_IOC_DEVICES_READY:
2444                 mutex_lock(&uuid_mutex);
2445                 device = btrfs_scan_one_device(vol->name, FMODE_READ,
2446                                                &btrfs_root_fs_type);
2447                 if (IS_ERR(device)) {
2448                         mutex_unlock(&uuid_mutex);
2449                         ret = PTR_ERR(device);
2450                         break;
2451                 }
2452                 ret = !(device->fs_devices->num_devices ==
2453                         device->fs_devices->total_devices);
2454                 mutex_unlock(&uuid_mutex);
2455                 break;
2456         case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2457                 ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2458                 break;
2459         }
2460
2461         kfree(vol);
2462         return ret;
2463 }
2464
2465 static int btrfs_freeze(struct super_block *sb)
2466 {
2467         struct btrfs_trans_handle *trans;
2468         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2469         struct btrfs_root *root = fs_info->tree_root;
2470
2471         set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2472         /*
2473          * We don't need a barrier here, we'll wait for any transaction that
2474          * could be in progress on other threads (and do delayed iputs that
2475          * we want to avoid on a frozen filesystem), or do the commit
2476          * ourselves.
2477          */
2478         trans = btrfs_attach_transaction_barrier(root);
2479         if (IS_ERR(trans)) {
2480                 /* no transaction, don't bother */
2481                 if (PTR_ERR(trans) == -ENOENT)
2482                         return 0;
2483                 return PTR_ERR(trans);
2484         }
2485         return btrfs_commit_transaction(trans);
2486 }
2487
2488 static int btrfs_unfreeze(struct super_block *sb)
2489 {
2490         struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2491
2492         clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2493         return 0;
2494 }
2495
2496 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2497 {
2498         struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2499
2500         /*
2501          * There should be always a valid pointer in latest_dev, it may be stale
2502          * for a short moment in case it's being deleted but still valid until
2503          * the end of RCU grace period.
2504          */
2505         rcu_read_lock();
2506         seq_escape(m, rcu_str_deref(fs_info->fs_devices->latest_dev->name), " \t\n\\");
2507         rcu_read_unlock();
2508
2509         return 0;
2510 }
2511
2512 static const struct super_operations btrfs_super_ops = {
2513         .drop_inode     = btrfs_drop_inode,
2514         .evict_inode    = btrfs_evict_inode,
2515         .put_super      = btrfs_put_super,
2516         .sync_fs        = btrfs_sync_fs,
2517         .show_options   = btrfs_show_options,
2518         .show_devname   = btrfs_show_devname,
2519         .alloc_inode    = btrfs_alloc_inode,
2520         .destroy_inode  = btrfs_destroy_inode,
2521         .free_inode     = btrfs_free_inode,
2522         .statfs         = btrfs_statfs,
2523         .remount_fs     = btrfs_remount,
2524         .freeze_fs      = btrfs_freeze,
2525         .unfreeze_fs    = btrfs_unfreeze,
2526 };
2527
2528 static const struct file_operations btrfs_ctl_fops = {
2529         .open = btrfs_control_open,
2530         .unlocked_ioctl  = btrfs_control_ioctl,
2531         .compat_ioctl = compat_ptr_ioctl,
2532         .owner   = THIS_MODULE,
2533         .llseek = noop_llseek,
2534 };
2535
2536 static struct miscdevice btrfs_misc = {
2537         .minor          = BTRFS_MINOR,
2538         .name           = "btrfs-control",
2539         .fops           = &btrfs_ctl_fops
2540 };
2541
2542 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2543 MODULE_ALIAS("devname:btrfs-control");
2544
2545 static int __init btrfs_interface_init(void)
2546 {
2547         return misc_register(&btrfs_misc);
2548 }
2549
2550 static __cold void btrfs_interface_exit(void)
2551 {
2552         misc_deregister(&btrfs_misc);
2553 }
2554
2555 static void __init btrfs_print_mod_info(void)
2556 {
2557         static const char options[] = ""
2558 #ifdef CONFIG_BTRFS_DEBUG
2559                         ", debug=on"
2560 #endif
2561 #ifdef CONFIG_BTRFS_ASSERT
2562                         ", assert=on"
2563 #endif
2564 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2565                         ", integrity-checker=on"
2566 #endif
2567 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2568                         ", ref-verify=on"
2569 #endif
2570 #ifdef CONFIG_BLK_DEV_ZONED
2571                         ", zoned=yes"
2572 #else
2573                         ", zoned=no"
2574 #endif
2575 #ifdef CONFIG_FS_VERITY
2576                         ", fsverity=yes"
2577 #else
2578                         ", fsverity=no"
2579 #endif
2580                         ;
2581         pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2582 }
2583
2584 static int __init init_btrfs_fs(void)
2585 {
2586         int err;
2587
2588         btrfs_props_init();
2589
2590         err = btrfs_init_sysfs();
2591         if (err)
2592                 return err;
2593
2594         btrfs_init_compress();
2595
2596         err = btrfs_init_cachep();
2597         if (err)
2598                 goto free_compress;
2599
2600         err = extent_io_init();
2601         if (err)
2602                 goto free_cachep;
2603
2604         err = extent_state_cache_init();
2605         if (err)
2606                 goto free_extent_io;
2607
2608         err = extent_map_init();
2609         if (err)
2610                 goto free_extent_state_cache;
2611
2612         err = ordered_data_init();
2613         if (err)
2614                 goto free_extent_map;
2615
2616         err = btrfs_delayed_inode_init();
2617         if (err)
2618                 goto free_ordered_data;
2619
2620         err = btrfs_auto_defrag_init();
2621         if (err)
2622                 goto free_delayed_inode;
2623
2624         err = btrfs_delayed_ref_init();
2625         if (err)
2626                 goto free_auto_defrag;
2627
2628         err = btrfs_prelim_ref_init();
2629         if (err)
2630                 goto free_delayed_ref;
2631
2632         err = btrfs_end_io_wq_init();
2633         if (err)
2634                 goto free_prelim_ref;
2635
2636         err = btrfs_interface_init();
2637         if (err)
2638                 goto free_end_io_wq;
2639
2640         btrfs_print_mod_info();
2641
2642         err = btrfs_run_sanity_tests();
2643         if (err)
2644                 goto unregister_ioctl;
2645
2646         err = register_filesystem(&btrfs_fs_type);
2647         if (err)
2648                 goto unregister_ioctl;
2649
2650         return 0;
2651
2652 unregister_ioctl:
2653         btrfs_interface_exit();
2654 free_end_io_wq:
2655         btrfs_end_io_wq_exit();
2656 free_prelim_ref:
2657         btrfs_prelim_ref_exit();
2658 free_delayed_ref:
2659         btrfs_delayed_ref_exit();
2660 free_auto_defrag:
2661         btrfs_auto_defrag_exit();
2662 free_delayed_inode:
2663         btrfs_delayed_inode_exit();
2664 free_ordered_data:
2665         ordered_data_exit();
2666 free_extent_map:
2667         extent_map_exit();
2668 free_extent_state_cache:
2669         extent_state_cache_exit();
2670 free_extent_io:
2671         extent_io_exit();
2672 free_cachep:
2673         btrfs_destroy_cachep();
2674 free_compress:
2675         btrfs_exit_compress();
2676         btrfs_exit_sysfs();
2677
2678         return err;
2679 }
2680
2681 static void __exit exit_btrfs_fs(void)
2682 {
2683         btrfs_destroy_cachep();
2684         btrfs_delayed_ref_exit();
2685         btrfs_auto_defrag_exit();
2686         btrfs_delayed_inode_exit();
2687         btrfs_prelim_ref_exit();
2688         ordered_data_exit();
2689         extent_map_exit();
2690         extent_state_cache_exit();
2691         extent_io_exit();
2692         btrfs_interface_exit();
2693         btrfs_end_io_wq_exit();
2694         unregister_filesystem(&btrfs_fs_type);
2695         btrfs_exit_sysfs();
2696         btrfs_cleanup_fs_uuids();
2697         btrfs_exit_compress();
2698 }
2699
2700 late_initcall(init_btrfs_fs);
2701 module_exit(exit_btrfs_fs)
2702
2703 MODULE_LICENSE("GPL");
2704 MODULE_SOFTDEP("pre: crc32c");
2705 MODULE_SOFTDEP("pre: xxhash64");
2706 MODULE_SOFTDEP("pre: sha256");
2707 MODULE_SOFTDEP("pre: blake2b-256");