super.c

   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * Copyright (C) 2007 Oracle.  All rights reserved.
   4 */
   5
   6#include <linux/blkdev.h>
   7#include <linux/module.h>
   8#include <linux/fs.h>
   9#include <linux/pagemap.h>
  10#include <linux/highmem.h>
  11#include <linux/time.h>
  12#include <linux/init.h>
  13#include <linux/seq_file.h>
  14#include <linux/string.h>
  15#include <linux/backing-dev.h>
  16#include <linux/mount.h>
  17#include <linux/writeback.h>
  18#include <linux/statfs.h>
  19#include <linux/compat.h>
  20#include <linux/parser.h>
  21#include <linux/ctype.h>
  22#include <linux/namei.h>
  23#include <linux/miscdevice.h>
  24#include <linux/magic.h>
  25#include <linux/slab.h>
  26#include <linux/ratelimit.h>
  27#include <linux/crc32c.h>
  28#include <linux/btrfs.h>
  29#include <linux/security.h>
  30#include <linux/fs_parser.h>
  31#include "messages.h"
  32#include "delayed-inode.h"
  33#include "ctree.h"
  34#include "disk-io.h"
  35#include "transaction.h"
  36#include "btrfs_inode.h"
  37#include "direct-io.h"
  38#include "props.h"
  39#include "xattr.h"
  40#include "bio.h"
  41#include "export.h"
  42#include "compression.h"
  43#include "dev-replace.h"
  44#include "free-space-cache.h"
  45#include "backref.h"
  46#include "space-info.h"
  47#include "sysfs.h"
  48#include "zoned.h"
  49#include "tests/btrfs-tests.h"
  50#include "block-group.h"
  51#include "discard.h"
  52#include "qgroup.h"
  53#include "raid56.h"
  54#include "fs.h"
  55#include "accessors.h"
  56#include "defrag.h"
  57#include "dir-item.h"
  58#include "ioctl.h"
  59#include "scrub.h"
  60#include "verity.h"
  61#include "super.h"
  62#include "extent-tree.h"
  63#define CREATE_TRACE_POINTS
  64#include <trace/events/btrfs.h>
  65
  66static const struct super_operations btrfs_super_ops;
  67static struct file_system_type btrfs_fs_type;
  68
  69static void btrfs_put_super(struct super_block *sb)
  70{
  71	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
  72
  73	btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
  74	close_ctree(fs_info);
  75}
  76
  77/* Store the mount options related information. */
  78struct btrfs_fs_context {
  79	char *subvol_name;
  80	u64 subvol_objectid;
  81	u64 max_inline;
  82	u32 commit_interval;
  83	u32 metadata_ratio;
  84	u32 thread_pool_size;
  85	unsigned long long mount_opt;
  86	unsigned long compress_type:4;
  87	int compress_level;
  88	refcount_t refs;
  89};
  90
  91static void btrfs_emit_options(struct btrfs_fs_info *info,
  92			       struct btrfs_fs_context *old);
  93
  94enum {
  95	Opt_acl,
  96	Opt_clear_cache,
  97	Opt_commit_interval,
  98	Opt_compress,
  99	Opt_compress_force,
 100	Opt_compress_force_type,
 101	Opt_compress_type,
 102	Opt_degraded,
 103	Opt_device,
 104	Opt_fatal_errors,
 105	Opt_flushoncommit,
 106	Opt_max_inline,
 107	Opt_barrier,
 108	Opt_datacow,
 109	Opt_datasum,
 110	Opt_defrag,
 111	Opt_discard,
 112	Opt_discard_mode,
 113	Opt_ratio,
 114	Opt_rescan_uuid_tree,
 115	Opt_skip_balance,
 116	Opt_space_cache,
 117	Opt_space_cache_version,
 118	Opt_ssd,
 119	Opt_ssd_spread,
 120	Opt_subvol,
 121	Opt_subvol_empty,
 122	Opt_subvolid,
 123	Opt_thread_pool,
 124	Opt_treelog,
 125	Opt_user_subvol_rm_allowed,
 126	Opt_norecovery,
 127
 128	/* Rescue options */
 129	Opt_rescue,
 130	Opt_usebackuproot,
 131
 132	/* Debugging options */
 133	Opt_enospc_debug,
 134#ifdef CONFIG_BTRFS_DEBUG
 135	Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 136	Opt_ref_verify,
 137	Opt_ref_tracker,
 138#endif
 139	Opt_err,
 140};
 141
 142enum {
 143	Opt_fatal_errors_panic,
 144	Opt_fatal_errors_bug,
 145};
 146
 147static const struct constant_table btrfs_parameter_fatal_errors[] = {
 148	{ "panic", Opt_fatal_errors_panic },
 149	{ "bug", Opt_fatal_errors_bug },
 150	{}
 151};
 152
 153enum {
 154	Opt_discard_sync,
 155	Opt_discard_async,
 156};
 157
 158static const struct constant_table btrfs_parameter_discard[] = {
 159	{ "sync", Opt_discard_sync },
 160	{ "async", Opt_discard_async },
 161	{}
 162};
 163
 164enum {
 165	Opt_space_cache_v1,
 166	Opt_space_cache_v2,
 167};
 168
 169static const struct constant_table btrfs_parameter_space_cache[] = {
 170	{ "v1", Opt_space_cache_v1 },
 171	{ "v2", Opt_space_cache_v2 },
 172	{}
 173};
 174
 175enum {
 176	Opt_rescue_usebackuproot,
 177	Opt_rescue_nologreplay,
 178	Opt_rescue_ignorebadroots,
 179	Opt_rescue_ignoredatacsums,
 180	Opt_rescue_ignoremetacsums,
 181	Opt_rescue_ignoresuperflags,
 182	Opt_rescue_parameter_all,
 183};
 184
 185static const struct constant_table btrfs_parameter_rescue[] = {
 186	{ "usebackuproot", Opt_rescue_usebackuproot },
 187	{ "nologreplay", Opt_rescue_nologreplay },
 188	{ "ignorebadroots", Opt_rescue_ignorebadroots },
 189	{ "ibadroots", Opt_rescue_ignorebadroots },
 190	{ "ignoredatacsums", Opt_rescue_ignoredatacsums },
 191	{ "ignoremetacsums", Opt_rescue_ignoremetacsums},
 192	{ "ignoresuperflags", Opt_rescue_ignoresuperflags},
 193	{ "idatacsums", Opt_rescue_ignoredatacsums },
 194	{ "imetacsums", Opt_rescue_ignoremetacsums},
 195	{ "isuperflags", Opt_rescue_ignoresuperflags},
 196	{ "all", Opt_rescue_parameter_all },
 197	{}
 198};
 199
 200#ifdef CONFIG_BTRFS_DEBUG
 201enum {
 202	Opt_fragment_parameter_data,
 203	Opt_fragment_parameter_metadata,
 204	Opt_fragment_parameter_all,
 205};
 206
 207static const struct constant_table btrfs_parameter_fragment[] = {
 208	{ "data", Opt_fragment_parameter_data },
 209	{ "metadata", Opt_fragment_parameter_metadata },
 210	{ "all", Opt_fragment_parameter_all },
 211	{}
 212};
 213#endif
 214
 215static const struct fs_parameter_spec btrfs_fs_parameters[] = {
 216	fsparam_flag_no("acl", Opt_acl),
 217	fsparam_flag_no("autodefrag", Opt_defrag),
 218	fsparam_flag_no("barrier", Opt_barrier),
 219	fsparam_flag("clear_cache", Opt_clear_cache),
 220	fsparam_u32("commit", Opt_commit_interval),
 221	fsparam_flag("compress", Opt_compress),
 222	fsparam_string("compress", Opt_compress_type),
 223	fsparam_flag("compress-force", Opt_compress_force),
 224	fsparam_string("compress-force", Opt_compress_force_type),
 225	fsparam_flag_no("datacow", Opt_datacow),
 226	fsparam_flag_no("datasum", Opt_datasum),
 227	fsparam_flag("degraded", Opt_degraded),
 228	fsparam_string("device", Opt_device),
 229	fsparam_flag_no("discard", Opt_discard),
 230	fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
 231	fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
 232	fsparam_flag_no("flushoncommit", Opt_flushoncommit),
 233	fsparam_string("max_inline", Opt_max_inline),
 234	fsparam_u32("metadata_ratio", Opt_ratio),
 235	fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
 236	fsparam_flag("skip_balance", Opt_skip_balance),
 237	fsparam_flag_no("space_cache", Opt_space_cache),
 238	fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
 239	fsparam_flag_no("ssd", Opt_ssd),
 240	fsparam_flag_no("ssd_spread", Opt_ssd_spread),
 241	fsparam_string("subvol", Opt_subvol),
 242	fsparam_flag("subvol=", Opt_subvol_empty),
 243	fsparam_u64("subvolid", Opt_subvolid),
 244	fsparam_u32("thread_pool", Opt_thread_pool),
 245	fsparam_flag_no("treelog", Opt_treelog),
 246	fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
 247
 248	/* Rescue options. */
 249	fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
 250	/* Deprecated, with alias rescue=usebackuproot */
 251	__fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
 252	/* For compatibility only, alias for "rescue=nologreplay". */
 253	fsparam_flag("norecovery", Opt_norecovery),
 254
 255	/* Debugging options. */
 256	fsparam_flag_no("enospc_debug", Opt_enospc_debug),
 257#ifdef CONFIG_BTRFS_DEBUG
 258	fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
 259	fsparam_flag("ref_tracker", Opt_ref_tracker),
 260	fsparam_flag("ref_verify", Opt_ref_verify),
 261#endif
 262	{}
 263};
 264
 265static bool btrfs_match_compress_type(const char *string, const char *type, bool may_have_level)
 266{
 267	const int len = strlen(type);
 268
 269	return (strncmp(string, type, len) == 0) &&
 270		((may_have_level && string[len] == ':') || string[len] == '\0');
 271}
 272
 273static int btrfs_parse_compress(struct btrfs_fs_context *ctx,
 274				const struct fs_parameter *param, int opt)
 275{
 276	const char *string = param->string;
 277	int ret;
 278
 279	/*
 280	 * Provide the same semantics as older kernels that don't use fs
 281	 * context, specifying the "compress" option clears "force-compress"
 282	 * without the need to pass "compress-force=[no|none]" before
 283	 * specifying "compress".
 284	 */
 285	if (opt != Opt_compress_force && opt != Opt_compress_force_type)
 286		btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 287
 288	if (opt == Opt_compress || opt == Opt_compress_force) {
 289		ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 290		ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 291		btrfs_set_opt(ctx->mount_opt, COMPRESS);
 292		btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 293		btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 294	} else if (btrfs_match_compress_type(string, "zlib", true)) {
 295		ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 296		ret = btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB, string + 4,
 297					       &ctx->compress_level);
 298		if (ret < 0)
 299			goto error;
 300		btrfs_set_opt(ctx->mount_opt, COMPRESS);
 301		btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 302		btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 303	} else if (btrfs_match_compress_type(string, "lzo", true)) {
 304		ctx->compress_type = BTRFS_COMPRESS_LZO;
 305		ret = btrfs_compress_str2level(BTRFS_COMPRESS_LZO, string + 3,
 306					       &ctx->compress_level);
 307		if (ret < 0)
 308			goto error;
 309		if (string[3] == ':' && string[4])
 310			btrfs_warn(NULL, "Compression level ignored for LZO");
 311		btrfs_set_opt(ctx->mount_opt, COMPRESS);
 312		btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 313		btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 314	} else if (btrfs_match_compress_type(string, "zstd", true)) {
 315		ctx->compress_type = BTRFS_COMPRESS_ZSTD;
 316		ret = btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD, string + 4,
 317					       &ctx->compress_level);
 318		if (ret < 0)
 319			goto error;
 320		btrfs_set_opt(ctx->mount_opt, COMPRESS);
 321		btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 322		btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 323	} else if (btrfs_match_compress_type(string, "no", false) ||
 324		   btrfs_match_compress_type(string, "none", false)) {
 325		ctx->compress_level = 0;
 326		ctx->compress_type = 0;
 327		btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 328		btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 329	} else {
 330		ret = -EINVAL;
 331		goto error;
 332	}
 333	return 0;
 334error:
 335	btrfs_err(NULL, "failed to parse compression option '%s'", string);
 336	return ret;
 337
 338}
 339
 340static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
 341{
 342	struct btrfs_fs_context *ctx = fc->fs_private;
 343	struct fs_parse_result result;
 344	int opt;
 345
 346	opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
 347	if (opt < 0)
 348		return opt;
 349
 350	switch (opt) {
 351	case Opt_degraded:
 352		btrfs_set_opt(ctx->mount_opt, DEGRADED);
 353		break;
 354	case Opt_subvol_empty:
 355		/*
 356		 * This exists because we used to allow it on accident, so we're
 357		 * keeping it to maintain ABI.  See 37becec95ac3 ("Btrfs: allow
 358		 * empty subvol= again").
 359		 */
 360		break;
 361	case Opt_subvol:
 362		kfree(ctx->subvol_name);
 363		ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
 364		if (!ctx->subvol_name)
 365			return -ENOMEM;
 366		break;
 367	case Opt_subvolid:
 368		ctx->subvol_objectid = result.uint_64;
 369
 370		/* subvolid=0 means give me the original fs_tree. */
 371		if (!ctx->subvol_objectid)
 372			ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
 373		break;
 374	case Opt_device: {
 375		struct btrfs_device *device;
 376
 377		mutex_lock(&uuid_mutex);
 378		device = btrfs_scan_one_device(param->string, false);
 379		mutex_unlock(&uuid_mutex);
 380		if (IS_ERR(device))
 381			return PTR_ERR(device);
 382		break;
 383	}
 384	case Opt_datasum:
 385		if (result.negated) {
 386			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 387		} else {
 388			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 389			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 390		}
 391		break;
 392	case Opt_datacow:
 393		if (result.negated) {
 394			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 395			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 396			btrfs_set_opt(ctx->mount_opt, NODATACOW);
 397			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 398		} else {
 399			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 400		}
 401		break;
 402	case Opt_compress_force:
 403	case Opt_compress_force_type:
 404		btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
 405		fallthrough;
 406	case Opt_compress:
 407	case Opt_compress_type:
 408		if (btrfs_parse_compress(ctx, param, opt))
 409			return -EINVAL;
 410		break;
 411	case Opt_ssd:
 412		if (result.negated) {
 413			btrfs_set_opt(ctx->mount_opt, NOSSD);
 414			btrfs_clear_opt(ctx->mount_opt, SSD);
 415			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 416		} else {
 417			btrfs_set_opt(ctx->mount_opt, SSD);
 418			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 419		}
 420		break;
 421	case Opt_ssd_spread:
 422		if (result.negated) {
 423			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 424		} else {
 425			btrfs_set_opt(ctx->mount_opt, SSD);
 426			btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
 427			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 428		}
 429		break;
 430	case Opt_barrier:
 431		if (result.negated)
 432			btrfs_set_opt(ctx->mount_opt, NOBARRIER);
 433		else
 434			btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
 435		break;
 436	case Opt_thread_pool:
 437		if (result.uint_32 == 0) {
 438			btrfs_err(NULL, "invalid value 0 for thread_pool");
 439			return -EINVAL;
 440		}
 441		ctx->thread_pool_size = result.uint_32;
 442		break;
 443	case Opt_max_inline:
 444		ctx->max_inline = memparse(param->string, NULL);
 445		break;
 446	case Opt_acl:
 447		if (result.negated) {
 448			fc->sb_flags &= ~SB_POSIXACL;
 449		} else {
 450#ifdef CONFIG_BTRFS_FS_POSIX_ACL
 451			fc->sb_flags |= SB_POSIXACL;
 452#else
 453			btrfs_err(NULL, "support for ACL not compiled in");
 454			return -EINVAL;
 455#endif
 456		}
 457		/*
 458		 * VFS limits the ability to toggle ACL on and off via remount,
 459		 * despite every file system allowing this.  This seems to be
 460		 * an oversight since we all do, but it'll fail if we're
 461		 * remounting.  So don't set the mask here, we'll check it in
 462		 * btrfs_reconfigure and do the toggling ourselves.
 463		 */
 464		if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
 465			fc->sb_flags_mask |= SB_POSIXACL;
 466		break;
 467	case Opt_treelog:
 468		if (result.negated)
 469			btrfs_set_opt(ctx->mount_opt, NOTREELOG);
 470		else
 471			btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
 472		break;
 473	case Opt_norecovery:
 474		btrfs_info(NULL,
 475"'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
 476		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 477		break;
 478	case Opt_flushoncommit:
 479		if (result.negated)
 480			btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
 481		else
 482			btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
 483		break;
 484	case Opt_ratio:
 485		ctx->metadata_ratio = result.uint_32;
 486		break;
 487	case Opt_discard:
 488		if (result.negated) {
 489			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 490			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 491			btrfs_set_opt(ctx->mount_opt, NODISCARD);
 492		} else {
 493			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 494			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 495		}
 496		break;
 497	case Opt_discard_mode:
 498		switch (result.uint_32) {
 499		case Opt_discard_sync:
 500			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 501			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 502			break;
 503		case Opt_discard_async:
 504			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 505			btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
 506			break;
 507		default:
 508			btrfs_err(NULL, "unrecognized discard mode value %s",
 509				  param->key);
 510			return -EINVAL;
 511		}
 512		btrfs_clear_opt(ctx->mount_opt, NODISCARD);
 513		break;
 514	case Opt_space_cache:
 515		if (result.negated) {
 516			btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
 517			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 518			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 519		} else {
 520			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 521			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 522		}
 523		break;
 524	case Opt_space_cache_version:
 525		switch (result.uint_32) {
 526		case Opt_space_cache_v1:
 527			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 528			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 529			break;
 530		case Opt_space_cache_v2:
 531			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 532			btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
 533			break;
 534		default:
 535			btrfs_err(NULL, "unrecognized space_cache value %s",
 536				  param->key);
 537			return -EINVAL;
 538		}
 539		break;
 540	case Opt_rescan_uuid_tree:
 541		btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
 542		break;
 543	case Opt_clear_cache:
 544		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 545		break;
 546	case Opt_user_subvol_rm_allowed:
 547		btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
 548		break;
 549	case Opt_enospc_debug:
 550		if (result.negated)
 551			btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
 552		else
 553			btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
 554		break;
 555	case Opt_defrag:
 556		if (result.negated)
 557			btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
 558		else
 559			btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
 560		break;
 561	case Opt_usebackuproot:
 562		btrfs_warn(NULL,
 563			   "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
 564		btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 565
 566		/* If we're loading the backup roots we can't trust the space cache. */
 567		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 568		break;
 569	case Opt_skip_balance:
 570		btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
 571		break;
 572	case Opt_fatal_errors:
 573		switch (result.uint_32) {
 574		case Opt_fatal_errors_panic:
 575			btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 576			break;
 577		case Opt_fatal_errors_bug:
 578			btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 579			break;
 580		default:
 581			btrfs_err(NULL, "unrecognized fatal_errors value %s",
 582				  param->key);
 583			return -EINVAL;
 584		}
 585		break;
 586	case Opt_commit_interval:
 587		ctx->commit_interval = result.uint_32;
 588		if (ctx->commit_interval > BTRFS_WARNING_COMMIT_INTERVAL) {
 589			btrfs_warn(NULL, "excessive commit interval %u, use with care",
 590				   ctx->commit_interval);
 591		}
 592		if (ctx->commit_interval == 0)
 593			ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 594		break;
 595	case Opt_rescue:
 596		switch (result.uint_32) {
 597		case Opt_rescue_usebackuproot:
 598			btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 599			break;
 600		case Opt_rescue_nologreplay:
 601			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 602			break;
 603		case Opt_rescue_ignorebadroots:
 604			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 605			break;
 606		case Opt_rescue_ignoredatacsums:
 607			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 608			break;
 609		case Opt_rescue_ignoremetacsums:
 610			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
 611			break;
 612		case Opt_rescue_ignoresuperflags:
 613			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
 614			break;
 615		case Opt_rescue_parameter_all:
 616			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 617			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
 618			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
 619			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 620			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 621			break;
 622		default:
 623			btrfs_info(NULL, "unrecognized rescue option '%s'",
 624				   param->key);
 625			return -EINVAL;
 626		}
 627		break;
 628#ifdef CONFIG_BTRFS_DEBUG
 629	case Opt_fragment:
 630		switch (result.uint_32) {
 631		case Opt_fragment_parameter_all:
 632			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 633			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 634			break;
 635		case Opt_fragment_parameter_metadata:
 636			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 637			break;
 638		case Opt_fragment_parameter_data:
 639			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 640			break;
 641		default:
 642			btrfs_info(NULL, "unrecognized fragment option '%s'",
 643				   param->key);
 644			return -EINVAL;
 645		}
 646		break;
 647	case Opt_ref_verify:
 648		btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
 649		break;
 650	case Opt_ref_tracker:
 651		btrfs_set_opt(ctx->mount_opt, REF_TRACKER);
 652		break;
 653#endif
 654	default:
 655		btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
 656		return -EINVAL;
 657	}
 658
 659	return 0;
 660}
 661
 662/*
 663 * Some options only have meaning at mount time and shouldn't persist across
 664 * remounts, or be displayed. Clear these at the end of mount and remount code
 665 * paths.
 666 */
 667static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
 668{
 669	btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
 670	btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
 671	btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
 672}
 673
 674static bool check_ro_option(const struct btrfs_fs_info *fs_info,
 675			    unsigned long long mount_opt, unsigned long long opt,
 676			    const char *opt_name)
 677{
 678	if (mount_opt & opt) {
 679		btrfs_err(fs_info, "%s must be used with ro mount option",
 680			  opt_name);
 681		return true;
 682	}
 683	return false;
 684}
 685
 686bool btrfs_check_options(const struct btrfs_fs_info *info,
 687			 unsigned long long *mount_opt,
 688			 unsigned long flags)
 689{
 690	bool ret = true;
 691
 692	if (!(flags & SB_RDONLY) &&
 693	    (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
 694	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
 695	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
 696	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
 697	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
 698		ret = false;
 699
 700	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 701	    !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
 702	    !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
 703		btrfs_err(info, "cannot disable free-space-tree");
 704		ret = false;
 705	}
 706	if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
 707	     !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
 708		btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
 709		ret = false;
 710	}
 711
 712	if (btrfs_check_mountopts_zoned(info, mount_opt))
 713		ret = false;
 714
 715	if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
 716		if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
 717			btrfs_warn(info,
 718"space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
 719		}
 720	}
 721
 722	return ret;
 723}
 724
 725/*
 726 * This is subtle, we only call this during open_ctree().  We need to pre-load
 727 * the mount options with the on-disk settings.  Before the new mount API took
 728 * effect we would do this on mount and remount.  With the new mount API we'll
 729 * only do this on the initial mount.
 730 *
 731 * This isn't a change in behavior, because we're using the current state of the
 732 * file system to set the current mount options.  If you mounted with special
 733 * options to disable these features and then remounted we wouldn't revert the
 734 * settings, because mounting without these features cleared the on-disk
 735 * settings, so this being called on re-mount is not needed.
 736 */
 737void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
 738{
 739	if (fs_info->sectorsize != PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) {
 740		btrfs_info(fs_info,
 741			   "forcing free space tree for sector size %u with page size %lu",
 742			   fs_info->sectorsize, PAGE_SIZE);
 743		btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
 744		btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 745	}
 746
 747	/*
 748	 * At this point our mount options are populated, so we only mess with
 749	 * these settings if we don't have any settings already.
 750	 */
 751	if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
 752		return;
 753
 754	if (btrfs_is_zoned(fs_info) &&
 755	    btrfs_free_space_cache_v1_active(fs_info)) {
 756		btrfs_info(fs_info, "zoned: clearing existing space cache");
 757		btrfs_set_super_cache_generation(fs_info->super_copy, 0);
 758		return;
 759	}
 760
 761	if (btrfs_test_opt(fs_info, SPACE_CACHE))
 762		return;
 763
 764	if (btrfs_test_opt(fs_info, NOSPACECACHE))
 765		return;
 766
 767	/*
 768	 * At this point we don't have explicit options set by the user, set
 769	 * them ourselves based on the state of the file system.
 770	 */
 771	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
 772		btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 773	else if (btrfs_free_space_cache_v1_active(fs_info))
 774		btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
 775}
 776
 777static void set_device_specific_options(struct btrfs_fs_info *fs_info)
 778{
 779	if (!btrfs_test_opt(fs_info, NOSSD) &&
 780	    !fs_info->fs_devices->rotating)
 781		btrfs_set_opt(fs_info->mount_opt, SSD);
 782
 783	/*
 784	 * For devices supporting discard turn on discard=async automatically,
 785	 * unless it's already set or disabled. This could be turned off by
 786	 * nodiscard for the same mount.
 787	 *
 788	 * The zoned mode piggy backs on the discard functionality for
 789	 * resetting a zone. There is no reason to delay the zone reset as it is
 790	 * fast enough. So, do not enable async discard for zoned mode.
 791	 */
 792	if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
 793	      btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
 794	      btrfs_test_opt(fs_info, NODISCARD)) &&
 795	    fs_info->fs_devices->discardable &&
 796	    !btrfs_is_zoned(fs_info))
 797		btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
 798}
 799
 800char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 801					  u64 subvol_objectid)
 802{
 803	struct btrfs_root *root = fs_info->tree_root;
 804	struct btrfs_root *fs_root = NULL;
 805	struct btrfs_root_ref *root_ref;
 806	struct btrfs_inode_ref *inode_ref;
 807	struct btrfs_key key;
 808	BTRFS_PATH_AUTO_FREE(path);
 809	char *name = NULL, *ptr;
 810	u64 dirid;
 811	int len;
 812	int ret;
 813
 814	path = btrfs_alloc_path();
 815	if (!path)
 816		return ERR_PTR(-ENOMEM);
 817
 818	name = kmalloc(PATH_MAX, GFP_KERNEL);
 819	if (!name) {
 820		ret = -ENOMEM;
 821		goto err;
 822	}
 823	ptr = name + PATH_MAX - 1;
 824	ptr[0] = '\0';
 825
 826	/*
 827	 * Walk up the subvolume trees in the tree of tree roots by root
 828	 * backrefs until we hit the top-level subvolume.
 829	 */
 830	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
 831		key.objectid = subvol_objectid;
 832		key.type = BTRFS_ROOT_BACKREF_KEY;
 833		key.offset = (u64)-1;
 834
 835		ret = btrfs_search_backwards(root, &key, path);
 836		if (ret < 0) {
 837			goto err;
 838		} else if (ret > 0) {
 839			ret = -ENOENT;
 840			goto err;
 841		}
 842
 843		subvol_objectid = key.offset;
 844
 845		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
 846					  struct btrfs_root_ref);
 847		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
 848		ptr -= len + 1;
 849		if (ptr < name) {
 850			ret = -ENAMETOOLONG;
 851			goto err;
 852		}
 853		read_extent_buffer(path->nodes[0], ptr + 1,
 854				   (unsigned long)(root_ref + 1), len);
 855		ptr[0] = '/';
 856		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
 857		btrfs_release_path(path);
 858
 859		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
 860		if (IS_ERR(fs_root)) {
 861			ret = PTR_ERR(fs_root);
 862			fs_root = NULL;
 863			goto err;
 864		}
 865
 866		/*
 867		 * Walk up the filesystem tree by inode refs until we hit the
 868		 * root directory.
 869		 */
 870		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
 871			key.objectid = dirid;
 872			key.type = BTRFS_INODE_REF_KEY;
 873			key.offset = (u64)-1;
 874
 875			ret = btrfs_search_backwards(fs_root, &key, path);
 876			if (ret < 0) {
 877				goto err;
 878			} else if (ret > 0) {
 879				ret = -ENOENT;
 880				goto err;
 881			}
 882
 883			dirid = key.offset;
 884
 885			inode_ref = btrfs_item_ptr(path->nodes[0],
 886						   path->slots[0],
 887						   struct btrfs_inode_ref);
 888			len = btrfs_inode_ref_name_len(path->nodes[0],
 889						       inode_ref);
 890			ptr -= len + 1;
 891			if (ptr < name) {
 892				ret = -ENAMETOOLONG;
 893				goto err;
 894			}
 895			read_extent_buffer(path->nodes[0], ptr + 1,
 896					   (unsigned long)(inode_ref + 1), len);
 897			ptr[0] = '/';
 898			btrfs_release_path(path);
 899		}
 900		btrfs_put_root(fs_root);
 901		fs_root = NULL;
 902	}
 903
 904	if (ptr == name + PATH_MAX - 1) {
 905		name[0] = '/';
 906		name[1] = '\0';
 907	} else {
 908		memmove(name, ptr, name + PATH_MAX - ptr);
 909	}
 910	return name;
 911
 912err:
 913	btrfs_put_root(fs_root);
 914	kfree(name);
 915	return ERR_PTR(ret);
 916}
 917
 918static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
 919{
 920	struct btrfs_root *root = fs_info->tree_root;
 921	struct btrfs_dir_item *di;
 922	BTRFS_PATH_AUTO_FREE(path);
 923	struct btrfs_key location;
 924	struct fscrypt_str name = FSTR_INIT("default", 7);
 925	u64 dir_id;
 926
 927	path = btrfs_alloc_path();
 928	if (!path)
 929		return -ENOMEM;
 930
 931	/*
 932	 * Find the "default" dir item which points to the root item that we
 933	 * will mount by default if we haven't been given a specific subvolume
 934	 * to mount.
 935	 */
 936	dir_id = btrfs_super_root_dir(fs_info->super_copy);
 937	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
 938	if (IS_ERR(di)) {
 939		return PTR_ERR(di);
 940	}
 941	if (!di) {
 942		/*
 943		 * Ok the default dir item isn't there.  This is weird since
 944		 * it's always been there, but don't freak out, just try and
 945		 * mount the top-level subvolume.
 946		 */
 947		*objectid = BTRFS_FS_TREE_OBJECTID;
 948		return 0;
 949	}
 950
 951	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 952	*objectid = location.objectid;
 953	return 0;
 954}
 955
 956static int btrfs_fill_super(struct super_block *sb,
 957			    struct btrfs_fs_devices *fs_devices)
 958{
 959	struct btrfs_inode *inode;
 960	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 961	int ret;
 962
 963	sb->s_maxbytes = MAX_LFS_FILESIZE;
 964	sb->s_magic = BTRFS_SUPER_MAGIC;
 965	sb->s_op = &btrfs_super_ops;
 966	set_default_d_op(sb, &btrfs_dentry_operations);
 967	sb->s_export_op = &btrfs_export_ops;
 968#ifdef CONFIG_FS_VERITY
 969	sb->s_vop = &btrfs_verityops;
 970#endif
 971	sb->s_xattr = btrfs_xattr_handlers;
 972	sb->s_time_gran = 1;
 973	sb->s_iflags |= SB_I_CGROUPWB | SB_I_ALLOW_HSM;
 974
 975	ret = super_setup_bdi(sb);
 976	if (ret) {
 977		btrfs_err(fs_info, "super_setup_bdi failed");
 978		return ret;
 979	}
 980
 981	ret = open_ctree(sb, fs_devices);
 982	if (ret) {
 983		btrfs_err(fs_info, "open_ctree failed: %d", ret);
 984		return ret;
 985	}
 986
 987	btrfs_emit_options(fs_info, NULL);
 988
 989	inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
 990	if (IS_ERR(inode)) {
 991		ret = PTR_ERR(inode);
 992		btrfs_handle_fs_error(fs_info, ret, NULL);
 993		goto fail_close;
 994	}
 995
 996	sb->s_root = d_make_root(&inode->vfs_inode);
 997	if (!sb->s_root) {
 998		ret = -ENOMEM;
 999		goto fail_close;
1000	}
1001
1002	sb->s_flags |= SB_ACTIVE;
1003	return 0;
1004
1005fail_close:
1006	close_ctree(fs_info);
1007	return ret;
1008}
1009
1010int btrfs_sync_fs(struct super_block *sb, int wait)
1011{
1012	struct btrfs_trans_handle *trans;
1013	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1014	struct btrfs_root *root = fs_info->tree_root;
1015
1016	trace_btrfs_sync_fs(fs_info, wait);
1017
1018	if (!wait) {
1019		filemap_flush(fs_info->btree_inode->i_mapping);
1020		return 0;
1021	}
1022
1023	btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1024
1025	trans = btrfs_attach_transaction_barrier(root);
1026	if (IS_ERR(trans)) {
1027		/* no transaction, don't bother */
1028		if (PTR_ERR(trans) == -ENOENT) {
1029			/*
1030			 * Exit unless we have some pending changes
1031			 * that need to go through commit
1032			 */
1033			if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1034				      &fs_info->flags))
1035				return 0;
1036			/*
1037			 * A non-blocking test if the fs is frozen. We must not
1038			 * start a new transaction here otherwise a deadlock
1039			 * happens. The pending operations are delayed to the
1040			 * next commit after thawing.
1041			 */
1042			if (sb_start_write_trylock(sb))
1043				sb_end_write(sb);
1044			else
1045				return 0;
1046			trans = btrfs_start_transaction(root, 0);
1047		}
1048		if (IS_ERR(trans))
1049			return PTR_ERR(trans);
1050	}
1051	return btrfs_commit_transaction(trans);
1052}
1053
1054static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1055{
1056	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1057	*printed = true;
1058}
1059
1060static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1061{
1062	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1063	const char *compress_type;
1064	const char *subvol_name;
1065	bool printed = false;
1066
1067	if (btrfs_test_opt(info, DEGRADED))
1068		seq_puts(seq, ",degraded");
1069	if (btrfs_test_opt(info, NODATASUM))
1070		seq_puts(seq, ",nodatasum");
1071	if (btrfs_test_opt(info, NODATACOW))
1072		seq_puts(seq, ",nodatacow");
1073	if (btrfs_test_opt(info, NOBARRIER))
1074		seq_puts(seq, ",nobarrier");
1075	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1076		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1077	if (info->thread_pool_size !=  min_t(unsigned long,
1078					     num_online_cpus() + 2, 8))
1079		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1080	if (btrfs_test_opt(info, COMPRESS)) {
1081		compress_type = btrfs_compress_type2str(info->compress_type);
1082		if (btrfs_test_opt(info, FORCE_COMPRESS))
1083			seq_printf(seq, ",compress-force=%s", compress_type);
1084		else
1085			seq_printf(seq, ",compress=%s", compress_type);
1086		if (info->compress_level && info->compress_type != BTRFS_COMPRESS_LZO)
1087			seq_printf(seq, ":%d", info->compress_level);
1088	}
1089	if (btrfs_test_opt(info, NOSSD))
1090		seq_puts(seq, ",nossd");
1091	if (btrfs_test_opt(info, SSD_SPREAD))
1092		seq_puts(seq, ",ssd_spread");
1093	else if (btrfs_test_opt(info, SSD))
1094		seq_puts(seq, ",ssd");
1095	if (btrfs_test_opt(info, NOTREELOG))
1096		seq_puts(seq, ",notreelog");
1097	if (btrfs_test_opt(info, NOLOGREPLAY))
1098		print_rescue_option(seq, "nologreplay", &printed);
1099	if (btrfs_test_opt(info, USEBACKUPROOT))
1100		print_rescue_option(seq, "usebackuproot", &printed);
1101	if (btrfs_test_opt(info, IGNOREBADROOTS))
1102		print_rescue_option(seq, "ignorebadroots", &printed);
1103	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1104		print_rescue_option(seq, "ignoredatacsums", &printed);
1105	if (btrfs_test_opt(info, IGNOREMETACSUMS))
1106		print_rescue_option(seq, "ignoremetacsums", &printed);
1107	if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1108		print_rescue_option(seq, "ignoresuperflags", &printed);
1109	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1110		seq_puts(seq, ",flushoncommit");
1111	if (btrfs_test_opt(info, DISCARD_SYNC))
1112		seq_puts(seq, ",discard");
1113	if (btrfs_test_opt(info, DISCARD_ASYNC))
1114		seq_puts(seq, ",discard=async");
1115	if (!(info->sb->s_flags & SB_POSIXACL))
1116		seq_puts(seq, ",noacl");
1117	if (btrfs_free_space_cache_v1_active(info))
1118		seq_puts(seq, ",space_cache");
1119	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1120		seq_puts(seq, ",space_cache=v2");
1121	else
1122		seq_puts(seq, ",nospace_cache");
1123	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1124		seq_puts(seq, ",rescan_uuid_tree");
1125	if (btrfs_test_opt(info, CLEAR_CACHE))
1126		seq_puts(seq, ",clear_cache");
1127	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1128		seq_puts(seq, ",user_subvol_rm_allowed");
1129	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1130		seq_puts(seq, ",enospc_debug");
1131	if (btrfs_test_opt(info, AUTO_DEFRAG))
1132		seq_puts(seq, ",autodefrag");
1133	if (btrfs_test_opt(info, SKIP_BALANCE))
1134		seq_puts(seq, ",skip_balance");
1135	if (info->metadata_ratio)
1136		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1137	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1138		seq_puts(seq, ",fatal_errors=panic");
1139	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1140		seq_printf(seq, ",commit=%u", info->commit_interval);
1141#ifdef CONFIG_BTRFS_DEBUG
1142	if (btrfs_test_opt(info, FRAGMENT_DATA))
1143		seq_puts(seq, ",fragment=data");
1144	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1145		seq_puts(seq, ",fragment=metadata");
1146#endif
1147	if (btrfs_test_opt(info, REF_VERIFY))
1148		seq_puts(seq, ",ref_verify");
1149	if (btrfs_test_opt(info, REF_TRACKER))
1150		seq_puts(seq, ",ref_tracker");
1151	seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1152	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1153			btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1154	if (!IS_ERR(subvol_name)) {
1155		seq_show_option(seq, "subvol", subvol_name);
1156		kfree(subvol_name);
1157	}
1158	return 0;
1159}
1160
1161/*
1162 * subvolumes are identified by ino 256
1163 */
1164static inline bool is_subvolume_inode(struct inode *inode)
1165{
1166	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1167		return true;
1168	return false;
1169}
1170
1171static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1172				   struct vfsmount *mnt)
1173{
1174	struct dentry *root;
1175	int ret;
1176
1177	if (!subvol_name) {
1178		if (!subvol_objectid) {
1179			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1180							  &subvol_objectid);
1181			if (ret) {
1182				root = ERR_PTR(ret);
1183				goto out;
1184			}
1185		}
1186		subvol_name = btrfs_get_subvol_name_from_objectid(
1187					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1188		if (IS_ERR(subvol_name)) {
1189			root = ERR_CAST(subvol_name);
1190			subvol_name = NULL;
1191			goto out;
1192		}
1193
1194	}
1195
1196	root = mount_subtree(mnt, subvol_name);
1197	/* mount_subtree() drops our reference on the vfsmount. */
1198	mnt = NULL;
1199
1200	if (!IS_ERR(root)) {
1201		struct super_block *s = root->d_sb;
1202		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1203		struct inode *root_inode = d_inode(root);
1204		u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1205
1206		ret = 0;
1207		if (!is_subvolume_inode(root_inode)) {
1208			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1209			       subvol_name);
1210			ret = -EINVAL;
1211		}
1212		if (subvol_objectid && root_objectid != subvol_objectid) {
1213			/*
1214			 * This will also catch a race condition where a
1215			 * subvolume which was passed by ID is renamed and
1216			 * another subvolume is renamed over the old location.
1217			 */
1218			btrfs_err(fs_info,
1219				  "subvol '%s' does not match subvolid %llu",
1220				  subvol_name, subvol_objectid);
1221			ret = -EINVAL;
1222		}
1223		if (ret) {
1224			dput(root);
1225			root = ERR_PTR(ret);
1226			deactivate_locked_super(s);
1227		}
1228	}
1229
1230out:
1231	mntput(mnt);
1232	kfree(subvol_name);
1233	return root;
1234}
1235
1236static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1237				     u32 new_pool_size, u32 old_pool_size)
1238{
1239	if (new_pool_size == old_pool_size)
1240		return;
1241
1242	fs_info->thread_pool_size = new_pool_size;
1243
1244	btrfs_info(fs_info, "resize thread pool %d -> %d",
1245	       old_pool_size, new_pool_size);
1246
1247	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1248	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1249	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1250	workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1251	workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1252	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1253	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1254	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1255}
1256
1257static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1258				       unsigned long long old_opts, int flags)
1259{
1260	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1261	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1262	     (flags & SB_RDONLY))) {
1263		/* wait for any defraggers to finish */
1264		wait_event(fs_info->transaction_wait,
1265			   (atomic_read(&fs_info->defrag_running) == 0));
1266		if (flags & SB_RDONLY)
1267			sync_filesystem(fs_info->sb);
1268	}
1269}
1270
1271static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1272					 unsigned long long old_opts)
1273{
1274	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1275
1276	/*
1277	 * We need to cleanup all defraggable inodes if the autodefragment is
1278	 * close or the filesystem is read only.
1279	 */
1280	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1281	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1282		btrfs_cleanup_defrag_inodes(fs_info);
1283	}
1284
1285	/* If we toggled discard async */
1286	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1287	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1288		btrfs_discard_resume(fs_info);
1289	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1290		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1291		btrfs_discard_cleanup(fs_info);
1292
1293	/* If we toggled space cache */
1294	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1295		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1296}
1297
1298static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1299{
1300	int ret;
1301
1302	if (BTRFS_FS_ERROR(fs_info)) {
1303		btrfs_err(fs_info,
1304			  "remounting read-write after error is not allowed");
1305		return -EINVAL;
1306	}
1307
1308	if (fs_info->fs_devices->rw_devices == 0)
1309		return -EACCES;
1310
1311	if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1312		btrfs_warn(fs_info,
1313			   "too many missing devices, writable remount is not allowed");
1314		return -EACCES;
1315	}
1316
1317	if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1318		btrfs_warn(fs_info,
1319			   "mount required to replay tree-log, cannot remount read-write");
1320		return -EINVAL;
1321	}
1322
1323	/*
1324	 * NOTE: when remounting with a change that does writes, don't put it
1325	 * anywhere above this point, as we are not sure to be safe to write
1326	 * until we pass the above checks.
1327	 */
1328	ret = btrfs_start_pre_rw_mount(fs_info);
1329	if (ret)
1330		return ret;
1331
1332	btrfs_clear_sb_rdonly(fs_info->sb);
1333
1334	set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1335
1336	/*
1337	 * If we've gone from readonly -> read-write, we need to get our
1338	 * sync/async discard lists in the right state.
1339	 */
1340	btrfs_discard_resume(fs_info);
1341
1342	return 0;
1343}
1344
1345static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1346{
1347	/*
1348	 * This also happens on 'umount -rf' or on shutdown, when the
1349	 * filesystem is busy.
1350	 */
1351	cancel_work_sync(&fs_info->async_reclaim_work);
1352	cancel_work_sync(&fs_info->async_data_reclaim_work);
1353
1354	btrfs_discard_cleanup(fs_info);
1355
1356	/* Wait for the uuid_scan task to finish */
1357	down(&fs_info->uuid_tree_rescan_sem);
1358	/* Avoid complains from lockdep et al. */
1359	up(&fs_info->uuid_tree_rescan_sem);
1360
1361	btrfs_set_sb_rdonly(fs_info->sb);
1362
1363	/*
1364	 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1365	 * loop if it's already active.  If it's already asleep, we'll leave
1366	 * unused block groups on disk until we're mounted read-write again
1367	 * unless we clean them up here.
1368	 */
1369	btrfs_delete_unused_bgs(fs_info);
1370
1371	/*
1372	 * The cleaner task could be already running before we set the flag
1373	 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).  We must make
1374	 * sure that after we finish the remount, i.e. after we call
1375	 * btrfs_commit_super(), the cleaner can no longer start a transaction
1376	 * - either because it was dropping a dead root, running delayed iputs
1377	 *   or deleting an unused block group (the cleaner picked a block
1378	 *   group from the list of unused block groups before we were able to
1379	 *   in the previous call to btrfs_delete_unused_bgs()).
1380	 */
1381	wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1382
1383	/*
1384	 * We've set the superblock to RO mode, so we might have made the
1385	 * cleaner task sleep without running all pending delayed iputs. Go
1386	 * through all the delayed iputs here, so that if an unmount happens
1387	 * without remounting RW we don't end up at finishing close_ctree()
1388	 * with a non-empty list of delayed iputs.
1389	 */
1390	btrfs_run_delayed_iputs(fs_info);
1391
1392	btrfs_dev_replace_suspend_for_unmount(fs_info);
1393	btrfs_scrub_cancel(fs_info);
1394	btrfs_pause_balance(fs_info);
1395
1396	/*
1397	 * Pause the qgroup rescan worker if it is running. We don't want it to
1398	 * be still running after we are in RO mode, as after that, by the time
1399	 * we unmount, it might have left a transaction open, so we would leak
1400	 * the transaction and/or crash.
1401	 */
1402	btrfs_qgroup_wait_for_completion(fs_info, false);
1403
1404	return btrfs_commit_super(fs_info);
1405}
1406
1407static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1408{
1409	fs_info->max_inline = ctx->max_inline;
1410	fs_info->commit_interval = ctx->commit_interval;
1411	fs_info->metadata_ratio = ctx->metadata_ratio;
1412	fs_info->thread_pool_size = ctx->thread_pool_size;
1413	fs_info->mount_opt = ctx->mount_opt;
1414	fs_info->compress_type = ctx->compress_type;
1415	fs_info->compress_level = ctx->compress_level;
1416}
1417
1418static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1419{
1420	ctx->max_inline = fs_info->max_inline;
1421	ctx->commit_interval = fs_info->commit_interval;
1422	ctx->metadata_ratio = fs_info->metadata_ratio;
1423	ctx->thread_pool_size = fs_info->thread_pool_size;
1424	ctx->mount_opt = fs_info->mount_opt;
1425	ctx->compress_type = fs_info->compress_type;
1426	ctx->compress_level = fs_info->compress_level;
1427}
1428
1429#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...)			\
1430do {										\
1431	if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1432	    btrfs_raw_test_opt(fs_info->mount_opt, opt))			\
1433		btrfs_info(fs_info, fmt, ##args);				\
1434} while (0)
1435
1436#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...)	\
1437do {									\
1438	if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1439	    !btrfs_raw_test_opt(fs_info->mount_opt, opt))		\
1440		btrfs_info(fs_info, fmt, ##args);			\
1441} while (0)
1442
1443static void btrfs_emit_options(struct btrfs_fs_info *info,
1444			       struct btrfs_fs_context *old)
1445{
1446	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1447	btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1448	btrfs_info_if_set(info, old, NODATACOW, "setting nodatacow");
1449	btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1450	btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1451	btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1452	btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1453	btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1454	btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1455	btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1456	btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1457	btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1458	btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1459	btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1460	btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1461	btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1462	btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1463	btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1464	btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1465	btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1466	btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1467	btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1468	btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1469
1470	btrfs_info_if_unset(info, old, NODATASUM, "setting datasum");
1471	btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1472	btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1473	btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1474	btrfs_info_if_unset(info, old, NOBARRIER, "turning on barriers");
1475	btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1476	btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1477	btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1478	btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1479	btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1480
1481	/* Did the compression settings change? */
1482	if (btrfs_test_opt(info, COMPRESS) &&
1483	    (!old ||
1484	     old->compress_type != info->compress_type ||
1485	     old->compress_level != info->compress_level ||
1486	     (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1487	      btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1488		const char *compress_type = btrfs_compress_type2str(info->compress_type);
1489
1490		btrfs_info(info, "%s %s compression, level %d",
1491			   btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1492			   compress_type, info->compress_level);
1493	}
1494
1495	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1496		btrfs_info(info, "max_inline set to %llu", info->max_inline);
1497}
1498
1499static int btrfs_reconfigure(struct fs_context *fc)
1500{
1501	struct super_block *sb = fc->root->d_sb;
1502	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1503	struct btrfs_fs_context *ctx = fc->fs_private;
1504	struct btrfs_fs_context old_ctx;
1505	int ret = 0;
1506	bool mount_reconfigure = (fc->s_fs_info != NULL);
1507
1508	btrfs_info_to_ctx(fs_info, &old_ctx);
1509
1510	/*
1511	 * This is our "bind mount" trick, we don't want to allow the user to do
1512	 * anything other than mount a different ro/rw and a different subvol,
1513	 * all of the mount options should be maintained.
1514	 */
1515	if (mount_reconfigure)
1516		ctx->mount_opt = old_ctx.mount_opt;
1517
1518	sync_filesystem(sb);
1519	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1520
1521	if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1522		return -EINVAL;
1523
1524	ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1525	if (ret < 0)
1526		return ret;
1527
1528	btrfs_ctx_to_info(fs_info, ctx);
1529	btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1530	btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1531				 old_ctx.thread_pool_size);
1532
1533	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1534	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1535	    (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1536		btrfs_warn(fs_info,
1537		"remount supports changing free space tree only from RO to RW");
1538		/* Make sure free space cache options match the state on disk. */
1539		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1540			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1541			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1542		}
1543		if (btrfs_free_space_cache_v1_active(fs_info)) {
1544			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1545			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1546		}
1547	}
1548
1549	ret = 0;
1550	if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1551		ret = btrfs_remount_ro(fs_info);
1552	else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1553		ret = btrfs_remount_rw(fs_info);
1554	if (ret)
1555		goto restore;
1556
1557	/*
1558	 * If we set the mask during the parameter parsing VFS would reject the
1559	 * remount.  Here we can set the mask and the value will be updated
1560	 * appropriately.
1561	 */
1562	if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1563		fc->sb_flags_mask |= SB_POSIXACL;
1564
1565	btrfs_emit_options(fs_info, &old_ctx);
1566	wake_up_process(fs_info->transaction_kthread);
1567	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1568	btrfs_clear_oneshot_options(fs_info);
1569	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1570
1571	return 0;
1572restore:
1573	btrfs_ctx_to_info(fs_info, &old_ctx);
1574	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1575	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1576	return ret;
1577}
1578
1579/* Used to sort the devices by max_avail(descending sort) */
1580static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1581{
1582	const struct btrfs_device_info *dev_info1 = a;
1583	const struct btrfs_device_info *dev_info2 = b;
1584
1585	if (dev_info1->max_avail > dev_info2->max_avail)
1586		return -1;
1587	else if (dev_info1->max_avail < dev_info2->max_avail)
1588		return 1;
1589	return 0;
1590}
1591
1592/*
1593 * sort the devices by max_avail, in which max free extent size of each device
1594 * is stored.(Descending Sort)
1595 */
1596static inline void btrfs_descending_sort_devices(
1597					struct btrfs_device_info *devices,
1598					size_t nr_devices)
1599{
1600	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1601	     btrfs_cmp_device_free_bytes, NULL);
1602}
1603
1604/*
1605 * The helper to calc the free space on the devices that can be used to store
1606 * file data.
1607 */
1608static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1609					      u64 *free_bytes)
1610{
1611	struct btrfs_device_info AUTO_KFREE(devices_info);
1612	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1613	struct btrfs_device *device;
1614	u64 type;
1615	u64 avail_space;
1616	u64 min_stripe_size;
1617	int num_stripes = 1;
1618	int i = 0, nr_devices;
1619	const struct btrfs_raid_attr *rattr;
1620
1621	/*
1622	 * We aren't under the device list lock, so this is racy-ish, but good
1623	 * enough for our purposes.
1624	 */
1625	nr_devices = fs_info->fs_devices->open_devices;
1626	if (!nr_devices) {
1627		smp_mb();
1628		nr_devices = fs_info->fs_devices->open_devices;
1629		ASSERT(nr_devices);
1630		if (!nr_devices) {
1631			*free_bytes = 0;
1632			return 0;
1633		}
1634	}
1635
1636	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1637			       GFP_KERNEL);
1638	if (!devices_info)
1639		return -ENOMEM;
1640
1641	/* calc min stripe number for data space allocation */
1642	type = btrfs_data_alloc_profile(fs_info);
1643	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1644
1645	if (type & BTRFS_BLOCK_GROUP_RAID0)
1646		num_stripes = nr_devices;
1647	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1648		num_stripes = rattr->ncopies;
1649	else if (type & BTRFS_BLOCK_GROUP_RAID10)
1650		num_stripes = 4;
1651
1652	/* Adjust for more than 1 stripe per device */
1653	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1654
1655	rcu_read_lock();
1656	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1657		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1658						&device->dev_state) ||
1659		    !device->bdev ||
1660		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1661			continue;
1662
1663		if (i >= nr_devices)
1664			break;
1665
1666		avail_space = device->total_bytes - device->bytes_used;
1667
1668		/* align with stripe_len */
1669		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1670
1671		/*
1672		 * Ensure we have at least min_stripe_size on top of the
1673		 * reserved space on the device.
1674		 */
1675		if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1676			continue;
1677
1678		avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1679
1680		devices_info[i].dev = device;
1681		devices_info[i].max_avail = avail_space;
1682
1683		i++;
1684	}
1685	rcu_read_unlock();
1686
1687	nr_devices = i;
1688
1689	btrfs_descending_sort_devices(devices_info, nr_devices);
1690
1691	i = nr_devices - 1;
1692	avail_space = 0;
1693	while (nr_devices >= rattr->devs_min) {
1694		num_stripes = min(num_stripes, nr_devices);
1695
1696		if (devices_info[i].max_avail >= min_stripe_size) {
1697			int j;
1698			u64 alloc_size;
1699
1700			avail_space += devices_info[i].max_avail * num_stripes;
1701			alloc_size = devices_info[i].max_avail;
1702			for (j = i + 1 - num_stripes; j <= i; j++)
1703				devices_info[j].max_avail -= alloc_size;
1704		}
1705		i--;
1706		nr_devices--;
1707	}
1708
1709	*free_bytes = avail_space;
1710	return 0;
1711}
1712
1713/*
1714 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1715 *
1716 * If there's a redundant raid level at DATA block groups, use the respective
1717 * multiplier to scale the sizes.
1718 *
1719 * Unused device space usage is based on simulating the chunk allocator
1720 * algorithm that respects the device sizes and order of allocations.  This is
1721 * a close approximation of the actual use but there are other factors that may
1722 * change the result (like a new metadata chunk).
1723 *
1724 * If metadata is exhausted, f_bavail will be 0.
1725 */
1726static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1727{
1728	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1729	struct btrfs_super_block *disk_super = fs_info->super_copy;
1730	struct btrfs_space_info *found;
1731	u64 total_used = 0;
1732	u64 total_free_data = 0;
1733	u64 total_free_meta = 0;
1734	u32 bits = fs_info->sectorsize_bits;
1735	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1736	unsigned factor = 1;
1737	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1738	int ret;
1739	u64 thresh = 0;
1740	int mixed = 0;
1741
1742	list_for_each_entry(found, &fs_info->space_info, list) {
1743		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1744			int i;
1745
1746			total_free_data += found->disk_total - found->disk_used;
1747			total_free_data -=
1748				btrfs_account_ro_block_groups_free_space(found);
1749
1750			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1751				if (!list_empty(&found->block_groups[i]))
1752					factor = btrfs_bg_type_to_factor(
1753						btrfs_raid_array[i].bg_flag);
1754			}
1755		}
1756
1757		/*
1758		 * Metadata in mixed block group profiles are accounted in data
1759		 */
1760		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1761			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1762				mixed = 1;
1763			else
1764				total_free_meta += found->disk_total -
1765					found->disk_used;
1766		}
1767
1768		total_used += found->disk_used;
1769	}
1770
1771	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1772	buf->f_blocks >>= bits;
1773	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1774
1775	/* Account global block reserve as used, it's in logical size already */
1776	spin_lock(&block_rsv->lock);
1777	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
1778	if (buf->f_bfree >= block_rsv->size >> bits)
1779		buf->f_bfree -= block_rsv->size >> bits;
1780	else
1781		buf->f_bfree = 0;
1782	spin_unlock(&block_rsv->lock);
1783
1784	buf->f_bavail = div_u64(total_free_data, factor);
1785	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1786	if (ret)
1787		return ret;
1788	buf->f_bavail += div_u64(total_free_data, factor);
1789	buf->f_bavail = buf->f_bavail >> bits;
1790
1791	/*
1792	 * We calculate the remaining metadata space minus global reserve. If
1793	 * this is (supposedly) smaller than zero, there's no space. But this
1794	 * does not hold in practice, the exhausted state happens where's still
1795	 * some positive delta. So we apply some guesswork and compare the
1796	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
1797	 *
1798	 * We probably cannot calculate the exact threshold value because this
1799	 * depends on the internal reservations requested by various
1800	 * operations, so some operations that consume a few metadata will
1801	 * succeed even if the Avail is zero. But this is better than the other
1802	 * way around.
1803	 */
1804	thresh = SZ_4M;
1805
1806	/*
1807	 * We only want to claim there's no available space if we can no longer
1808	 * allocate chunks for our metadata profile and our global reserve will
1809	 * not fit in the free metadata space.  If we aren't ->full then we
1810	 * still can allocate chunks and thus are fine using the currently
1811	 * calculated f_bavail.
1812	 */
1813	if (!mixed && block_rsv->space_info->full &&
1814	    (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1815		buf->f_bavail = 0;
1816
1817	buf->f_type = BTRFS_SUPER_MAGIC;
1818	buf->f_bsize = fs_info->sectorsize;
1819	buf->f_namelen = BTRFS_NAME_LEN;
1820
1821	/* We treat it as constant endianness (it doesn't matter _which_)
1822	   because we want the fsid to come out the same whether mounted
1823	   on a big-endian or little-endian host */
1824	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1825	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1826	/* Mask in the root object ID too, to disambiguate subvols */
1827	buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1828	buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
1829
1830	return 0;
1831}
1832
1833static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1834{
1835	struct btrfs_fs_info *p = fc->s_fs_info;
1836	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1837
1838	return fs_info->fs_devices == p->fs_devices;
1839}
1840
1841static int btrfs_get_tree_super(struct fs_context *fc)
1842{
1843	struct btrfs_fs_info *fs_info = fc->s_fs_info;
1844	struct btrfs_fs_context *ctx = fc->fs_private;
1845	struct btrfs_fs_devices *fs_devices = NULL;
1846	struct btrfs_device *device;
1847	struct super_block *sb;
1848	blk_mode_t mode = sb_open_mode(fc->sb_flags);
1849	int ret;
1850
1851	btrfs_ctx_to_info(fs_info, ctx);
1852	mutex_lock(&uuid_mutex);
1853
1854	/*
1855	 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1856	 * either a valid device or an error.
1857	 */
1858	device = btrfs_scan_one_device(fc->source, true);
1859	ASSERT(device != NULL);
1860	if (IS_ERR(device)) {
1861		mutex_unlock(&uuid_mutex);
1862		return PTR_ERR(device);
1863	}
1864	fs_devices = device->fs_devices;
1865	/*
1866	 * We cannot hold uuid_mutex calling sget_fc(), it will lead to a
1867	 * locking order reversal with s_umount.
1868	 *
1869	 * So here we increase the holding number of fs_devices, this will ensure
1870	 * the fs_devices itself won't be freed.
1871	 */
1872	btrfs_fs_devices_inc_holding(fs_devices);
1873	fs_info->fs_devices = fs_devices;
1874	mutex_unlock(&uuid_mutex);
1875
1876
1877	sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1878	if (IS_ERR(sb)) {
1879		mutex_lock(&uuid_mutex);
1880		btrfs_fs_devices_dec_holding(fs_devices);
1881		/*
1882		 * Since the fs_devices is not opened, it can be freed at any
1883		 * time after unlocking uuid_mutex.  We need to avoid double
1884		 * free through put_fs_context()->btrfs_free_fs_info().
1885		 * So here we reset fs_info->fs_devices to NULL, and let the
1886		 * regular fs_devices reclaim path to handle it.
1887		 *
1888		 * This applies to all later branches where no fs_devices is
1889		 * opened.
1890		 */
1891		fs_info->fs_devices = NULL;
1892		mutex_unlock(&uuid_mutex);
1893		return PTR_ERR(sb);
1894	}
1895
1896	if (sb->s_root) {
1897		/*
1898		 * Not the first mount of the fs thus got an existing super block.
1899		 * Will reuse the returned super block, fs_info and fs_devices.
1900		 *
1901		 * fc->s_fs_info is not touched and will be later freed by
1902		 * put_fs_context() through btrfs_free_fs_context().
1903		 */
1904		ASSERT(fc->s_fs_info == fs_info);
1905
1906		mutex_lock(&uuid_mutex);
1907		btrfs_fs_devices_dec_holding(fs_devices);
1908		fs_info->fs_devices = NULL;
1909		mutex_unlock(&uuid_mutex);
1910		/*
1911		 * At this stage we may have RO flag mismatch between
1912		 * fc->sb_flags and sb->s_flags.  Caller should detect such
1913		 * mismatch and reconfigure with sb->s_umount rwsem held if
1914		 * needed.
1915		 */
1916	} else {
1917		struct block_device *bdev;
1918
1919		/*
1920		 * The first mount of the fs thus a new superblock, fc->s_fs_info
1921		 * must be NULL, and the ownership of our fs_info and fs_devices is
1922		 * transferred to the super block.
1923		 */
1924		ASSERT(fc->s_fs_info == NULL);
1925
1926		mutex_lock(&uuid_mutex);
1927		btrfs_fs_devices_dec_holding(fs_devices);
1928		ret = btrfs_open_devices(fs_devices, mode, sb);
1929		if (ret < 0)
1930			fs_info->fs_devices = NULL;
1931		mutex_unlock(&uuid_mutex);
1932		if (ret < 0) {
1933			deactivate_locked_super(sb);
1934			return ret;
1935		}
1936		if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1937			deactivate_locked_super(sb);
1938			return -EACCES;
1939		}
1940		set_device_specific_options(fs_info);
1941		bdev = fs_devices->latest_dev->bdev;
1942		snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1943		shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1944		ret = btrfs_fill_super(sb, fs_devices);
1945		if (ret) {
1946			deactivate_locked_super(sb);
1947			return ret;
1948		}
1949	}
1950
1951	btrfs_clear_oneshot_options(fs_info);
1952
1953	fc->root = dget(sb->s_root);
1954	return 0;
1955}
1956
1957/*
1958 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1959 * with different ro/rw options") the following works:
1960 *
1961 *        (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1962 *       (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1963 *
1964 * which looks nice and innocent but is actually pretty intricate and deserves
1965 * a long comment.
1966 *
1967 * On another filesystem a subvolume mount is close to something like:
1968 *
1969 *	(iii) # create rw superblock + initial mount
1970 *	      mount -t xfs /dev/sdb /opt/
1971 *
1972 *	      # create ro bind mount
1973 *	      mount --bind -o ro /opt/foo /mnt/foo
1974 *
1975 *	      # unmount initial mount
1976 *	      umount /opt
1977 *
1978 * Of course, there's some special subvolume sauce and there's the fact that the
1979 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1980 * it's very close and will help us understand the issue.
1981 *
1982 * The old mount API didn't cleanly distinguish between a mount being made ro
1983 * and a superblock being made ro.  The only way to change the ro state of
1984 * either object was by passing ms_rdonly. If a new mount was created via
1985 * mount(2) such as:
1986 *
1987 *      mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1988 *
1989 * the MS_RDONLY flag being specified had two effects:
1990 *
1991 * (1) MNT_READONLY was raised -> the resulting mount got
1992 *     @mnt->mnt_flags |= MNT_READONLY raised.
1993 *
1994 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1995 *     made the superblock ro. Note, how SB_RDONLY has the same value as
1996 *     ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1997 *
1998 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1999 * subtree mounted ro.
2000 *
2001 * But consider the effect on the old mount API on btrfs subvolume mounting
2002 * which combines the distinct step in (iii) into a single step.
2003 *
2004 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
2005 * is issued the superblock is ro and thus even if the mount created for (ii) is
2006 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
2007 * to rw for (ii) which it did using an internal remount call.
2008 *
2009 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
2010 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
2011 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
2012 * passed by mount(8) to mount(2).
2013 *
2014 * Enter the new mount API. The new mount API disambiguates making a mount ro
2015 * and making a superblock ro.
2016 *
2017 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
2018 *     fsmount() or mount_setattr() this is a pure VFS level change for a
2019 *     specific mount or mount tree that is never seen by the filesystem itself.
2020 *
2021 * (4) To turn a superblock ro the "ro" flag must be used with
2022 *     fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
2023 *     in fc->sb_flags.
2024 *
2025 * But, currently the util-linux mount command already utilizes the new mount
2026 * API and is still setting fsconfig(FSCONFIG_SET_FLAG, "ro") no matter if it's
2027 * btrfs or not, setting the whole super block RO.  To make per-subvolume mounting
2028 * work with different options work we need to keep backward compatibility.
2029 */
2030static int btrfs_reconfigure_for_mount(struct fs_context *fc)
2031{
2032	int ret = 0;
2033
2034	if (!(fc->sb_flags & SB_RDONLY) && (fc->root->d_sb->s_flags & SB_RDONLY))
2035		ret = btrfs_reconfigure(fc);
2036
2037	return ret;
2038}
2039
2040static int btrfs_get_tree_subvol(struct fs_context *fc)
2041{
2042	struct btrfs_fs_info *fs_info = NULL;
2043	struct btrfs_fs_context *ctx = fc->fs_private;
2044	struct fs_context *dup_fc;
2045	struct dentry *dentry;
2046	struct vfsmount *mnt;
2047	int ret = 0;
2048
2049	/*
2050	 * Setup a dummy root and fs_info for test/set super.  This is because
2051	 * we don't actually fill this stuff out until open_ctree, but we need
2052	 * then open_ctree will properly initialize the file system specific
2053	 * settings later.  btrfs_init_fs_info initializes the static elements
2054	 * of the fs_info (locks and such) to make cleanup easier if we find a
2055	 * superblock with our given fs_devices later on at sget() time.
2056	 */
2057	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2058	if (!fs_info)
2059		return -ENOMEM;
2060
2061	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2062	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2063	if (!fs_info->super_copy || !fs_info->super_for_commit) {
2064		/*
2065		 * Dont call btrfs_free_fs_info() to free it as it's still
2066		 * initialized partially.
2067		 */
2068		kfree(fs_info->super_copy);
2069		kfree(fs_info->super_for_commit);
2070		kvfree(fs_info);
2071		return -ENOMEM;
2072	}
2073	btrfs_init_fs_info(fs_info);
2074
2075	dup_fc = vfs_dup_fs_context(fc);
2076	if (IS_ERR(dup_fc)) {
2077		btrfs_free_fs_info(fs_info);
2078		return PTR_ERR(dup_fc);
2079	}
2080
2081	/*
2082	 * When we do the sget_fc this gets transferred to the sb, so we only
2083	 * need to set it on the dup_fc as this is what creates the super block.
2084	 */
2085	dup_fc->s_fs_info = fs_info;
2086
2087	ret = btrfs_get_tree_super(dup_fc);
2088	if (ret)
2089		goto error;
2090
2091	ret = btrfs_reconfigure_for_mount(dup_fc);
2092	up_write(&dup_fc->root->d_sb->s_umount);
2093	if (ret)
2094		goto error;
2095	mnt = vfs_create_mount(dup_fc);
2096	put_fs_context(dup_fc);
2097	if (IS_ERR(mnt))
2098		return PTR_ERR(mnt);
2099
2100	/*
2101	 * This free's ->subvol_name, because if it isn't set we have to
2102	 * allocate a buffer to hold the subvol_name, so we just drop our
2103	 * reference to it here.
2104	 */
2105	dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2106	ctx->subvol_name = NULL;
2107	if (IS_ERR(dentry))
2108		return PTR_ERR(dentry);
2109
2110	fc->root = dentry;
2111	return 0;
2112error:
2113	put_fs_context(dup_fc);
2114	return ret;
2115}
2116
2117static int btrfs_get_tree(struct fs_context *fc)
2118{
2119	ASSERT(fc->s_fs_info == NULL);
2120
2121	return btrfs_get_tree_subvol(fc);
2122}
2123
2124static void btrfs_kill_super(struct super_block *sb)
2125{
2126	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2127	kill_anon_super(sb);
2128	btrfs_free_fs_info(fs_info);
2129}
2130
2131static void btrfs_free_fs_context(struct fs_context *fc)
2132{
2133	struct btrfs_fs_context *ctx = fc->fs_private;
2134	struct btrfs_fs_info *fs_info = fc->s_fs_info;
2135
2136	if (fs_info)
2137		btrfs_free_fs_info(fs_info);
2138
2139	if (ctx && refcount_dec_and_test(&ctx->refs)) {
2140		kfree(ctx->subvol_name);
2141		kfree(ctx);
2142	}
2143}
2144
2145static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2146{
2147	struct btrfs_fs_context *ctx = src_fc->fs_private;
2148
2149	/*
2150	 * Give a ref to our ctx to this dup, as we want to keep it around for
2151	 * our original fc so we can have the subvolume name or objectid.
2152	 *
2153	 * We unset ->source in the original fc because the dup needs it for
2154	 * mounting, and then once we free the dup it'll free ->source, so we
2155	 * need to make sure we're only pointing to it in one fc.
2156	 */
2157	refcount_inc(&ctx->refs);
2158	fc->fs_private = ctx;
2159	fc->source = src_fc->source;
2160	src_fc->source = NULL;
2161	return 0;
2162}
2163
2164static const struct fs_context_operations btrfs_fs_context_ops = {
2165	.parse_param	= btrfs_parse_param,
2166	.reconfigure	= btrfs_reconfigure,
2167	.get_tree	= btrfs_get_tree,
2168	.dup		= btrfs_dup_fs_context,
2169	.free		= btrfs_free_fs_context,
2170};
2171
2172static int btrfs_init_fs_context(struct fs_context *fc)
2173{
2174	struct btrfs_fs_context *ctx;
2175
2176	ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2177	if (!ctx)
2178		return -ENOMEM;
2179
2180	refcount_set(&ctx->refs, 1);
2181	fc->fs_private = ctx;
2182	fc->ops = &btrfs_fs_context_ops;
2183
2184	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2185		btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2186	} else {
2187		ctx->thread_pool_size =
2188			min_t(unsigned long, num_online_cpus() + 2, 8);
2189		ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2190		ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2191	}
2192
2193#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2194	fc->sb_flags |= SB_POSIXACL;
2195#endif
2196	fc->sb_flags |= SB_I_VERSION;
2197
2198	return 0;
2199}
2200
2201static struct file_system_type btrfs_fs_type = {
2202	.owner			= THIS_MODULE,
2203	.name			= "btrfs",
2204	.init_fs_context	= btrfs_init_fs_context,
2205	.parameters		= btrfs_fs_parameters,
2206	.kill_sb		= btrfs_kill_super,
2207	.fs_flags		= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA |
2208				  FS_ALLOW_IDMAP | FS_MGTIME,
2209 };
2210
2211MODULE_ALIAS_FS("btrfs");
2212
2213static int btrfs_control_open(struct inode *inode, struct file *file)
2214{
2215	/*
2216	 * The control file's private_data is used to hold the
2217	 * transaction when it is started and is used to keep
2218	 * track of whether a transaction is already in progress.
2219	 */
2220	file->private_data = NULL;
2221	return 0;
2222}
2223
2224/*
2225 * Used by /dev/btrfs-control for devices ioctls.
2226 */
2227static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2228				unsigned long arg)
2229{
2230	struct btrfs_ioctl_vol_args *vol;
2231	struct btrfs_device *device = NULL;
2232	dev_t devt = 0;
2233	int ret = -ENOTTY;
2234
2235	if (!capable(CAP_SYS_ADMIN))
2236		return -EPERM;
2237
2238	vol = memdup_user((void __user *)arg, sizeof(*vol));
2239	if (IS_ERR(vol))
2240		return PTR_ERR(vol);
2241	ret = btrfs_check_ioctl_vol_args_path(vol);
2242	if (ret < 0)
2243		goto out;
2244
2245	switch (cmd) {
2246	case BTRFS_IOC_SCAN_DEV:
2247		mutex_lock(&uuid_mutex);
2248		/*
2249		 * Scanning outside of mount can return NULL which would turn
2250		 * into 0 error code.
2251		 */
2252		device = btrfs_scan_one_device(vol->name, false);
2253		ret = PTR_ERR_OR_ZERO(device);
2254		mutex_unlock(&uuid_mutex);
2255		break;
2256	case BTRFS_IOC_FORGET_DEV:
2257		if (vol->name[0] != 0) {
2258			ret = lookup_bdev(vol->name, &devt);
2259			if (ret)
2260				break;
2261		}
2262		ret = btrfs_forget_devices(devt);
2263		break;
2264	case BTRFS_IOC_DEVICES_READY:
2265		mutex_lock(&uuid_mutex);
2266		/*
2267		 * Scanning outside of mount can return NULL which would turn
2268		 * into 0 error code.
2269		 */
2270		device = btrfs_scan_one_device(vol->name, false);
2271		if (IS_ERR_OR_NULL(device)) {
2272			mutex_unlock(&uuid_mutex);
2273			ret = PTR_ERR_OR_ZERO(device);
2274			break;
2275		}
2276		ret = !(device->fs_devices->num_devices ==
2277			device->fs_devices->total_devices);
2278		mutex_unlock(&uuid_mutex);
2279		break;
2280	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2281		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2282		break;
2283	}
2284
2285out:
2286	kfree(vol);
2287	return ret;
2288}
2289
2290static int btrfs_freeze(struct super_block *sb)
2291{
2292	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2293
2294	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2295	/*
2296	 * We don't need a barrier here, we'll wait for any transaction that
2297	 * could be in progress on other threads (and do delayed iputs that
2298	 * we want to avoid on a frozen filesystem), or do the commit
2299	 * ourselves.
2300	 */
2301	return btrfs_commit_current_transaction(fs_info->tree_root);
2302}
2303
2304static int check_dev_super(struct btrfs_device *dev)
2305{
2306	struct btrfs_fs_info *fs_info = dev->fs_info;
2307	struct btrfs_super_block *sb;
2308	u64 last_trans;
2309	u16 csum_type;
2310	int ret = 0;
2311
2312	/* This should be called with fs still frozen. */
2313	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2314
2315	/* Missing dev, no need to check. */
2316	if (!dev->bdev)
2317		return 0;
2318
2319	/* Only need to check the primary super block. */
2320	sb = btrfs_read_disk_super(dev->bdev, 0, true);
2321	if (IS_ERR(sb))
2322		return PTR_ERR(sb);
2323
2324	/* Verify the checksum. */
2325	csum_type = btrfs_super_csum_type(sb);
2326	if (unlikely(csum_type != btrfs_super_csum_type(fs_info->super_copy))) {
2327		btrfs_err(fs_info, "csum type changed, has %u expect %u",
2328			  csum_type, btrfs_super_csum_type(fs_info->super_copy));
2329		ret = -EUCLEAN;
2330		goto out;
2331	}
2332
2333	if (unlikely(btrfs_check_super_csum(fs_info, sb))) {
2334		btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2335		ret = -EUCLEAN;
2336		goto out;
2337	}
2338
2339	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2340	ret = btrfs_validate_super(fs_info, sb, 0);
2341	if (ret < 0)
2342		goto out;
2343
2344	last_trans = btrfs_get_last_trans_committed(fs_info);
2345	if (unlikely(btrfs_super_generation(sb) != last_trans)) {
2346		btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2347			  btrfs_super_generation(sb), last_trans);
2348		ret = -EUCLEAN;
2349		goto out;
2350	}
2351out:
2352	btrfs_release_disk_super(sb);
2353	return ret;
2354}
2355
2356static int btrfs_unfreeze(struct super_block *sb)
2357{
2358	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2359	struct btrfs_device *device;
2360	int ret = 0;
2361
2362	/*
2363	 * Make sure the fs is not changed by accident (like hibernation then
2364	 * modified by other OS).
2365	 * If we found anything wrong, we mark the fs error immediately.
2366	 *
2367	 * And since the fs is frozen, no one can modify the fs yet, thus
2368	 * we don't need to hold device_list_mutex.
2369	 */
2370	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2371		ret = check_dev_super(device);
2372		if (ret < 0) {
2373			btrfs_handle_fs_error(fs_info, ret,
2374				"super block on devid %llu got modified unexpectedly",
2375				device->devid);
2376			break;
2377		}
2378	}
2379	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2380
2381	/*
2382	 * We still return 0, to allow VFS layer to unfreeze the fs even the
2383	 * above checks failed. Since the fs is either fine or read-only, we're
2384	 * safe to continue, without causing further damage.
2385	 */
2386	return 0;
2387}
2388
2389static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2390{
2391	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2392
2393	/*
2394	 * There should be always a valid pointer in latest_dev, it may be stale
2395	 * for a short moment in case it's being deleted but still valid until
2396	 * the end of RCU grace period.
2397	 */
2398	rcu_read_lock();
2399	seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2400	rcu_read_unlock();
2401
2402	return 0;
2403}
2404
2405static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2406{
2407	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2408	const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
2409
2410	trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2411
2412	return nr;
2413}
2414
2415static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)
2416{
2417	const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
2418	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2419
2420	btrfs_free_extent_maps(fs_info, nr_to_scan);
2421
2422	/* The extent map shrinker runs asynchronously, so always return 0. */
2423	return 0;
2424}
2425
2426#ifdef CONFIG_BTRFS_EXPERIMENTAL
2427static int btrfs_remove_bdev(struct super_block *sb, struct block_device *bdev)
2428{
2429	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2430	struct btrfs_device *device;
2431	struct btrfs_dev_lookup_args lookup_args = { .devt = bdev->bd_dev };
2432	bool can_rw;
2433
2434	mutex_lock(&fs_info->fs_devices->device_list_mutex);
2435	device = btrfs_find_device(fs_info->fs_devices, &lookup_args);
2436	if (!device) {
2437		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2438		/* Device not found, should not affect the running fs, just give a warning. */
2439		btrfs_warn(fs_info, "unable to find btrfs device for block device '%pg'", bdev);
2440		return 0;
2441	}
2442	/*
2443	 * The to-be-removed device is already missing?
2444	 *
2445	 * That's weird but no special handling needed and can exit right now.
2446	 */
2447	if (unlikely(test_and_set_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))) {
2448		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2449		btrfs_warn(fs_info, "btrfs device id %llu is already missing", device->devid);
2450		return 0;
2451	}
2452
2453	device->fs_devices->missing_devices++;
2454	if (test_and_clear_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
2455		list_del_init(&device->dev_alloc_list);
2456		WARN_ON(device->fs_devices->rw_devices < 1);
2457		device->fs_devices->rw_devices--;
2458	}
2459	can_rw = btrfs_check_rw_degradable(fs_info, device);
2460	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
2461	/*
2462	 * Now device is considered missing, btrfs_device_name() won't give a
2463	 * meaningful result anymore, so only output the devid.
2464	 */
2465	if (unlikely(!can_rw)) {
2466		btrfs_crit(fs_info,
2467		"btrfs device id %llu has gone missing, can not maintain read-write",
2468			   device->devid);
2469		return -EIO;
2470	}
2471	btrfs_warn(fs_info,
2472		   "btrfs device id %llu has gone missing, continue as degraded",
2473		   device->devid);
2474	btrfs_set_opt(fs_info->mount_opt, DEGRADED);
2475	return 0;
2476}
2477
2478static void btrfs_shutdown(struct super_block *sb)
2479{
2480	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2481
2482	btrfs_force_shutdown(fs_info);
2483}
2484#endif
2485
2486static const struct super_operations btrfs_super_ops = {
2487	.drop_inode	= btrfs_drop_inode,
2488	.evict_inode	= btrfs_evict_inode,
2489	.put_super	= btrfs_put_super,
2490	.sync_fs	= btrfs_sync_fs,
2491	.show_options	= btrfs_show_options,
2492	.show_devname	= btrfs_show_devname,
2493	.alloc_inode	= btrfs_alloc_inode,
2494	.destroy_inode	= btrfs_destroy_inode,
2495	.free_inode	= btrfs_free_inode,
2496	.statfs		= btrfs_statfs,
2497	.freeze_fs	= btrfs_freeze,
2498	.unfreeze_fs	= btrfs_unfreeze,
2499	.nr_cached_objects = btrfs_nr_cached_objects,
2500	.free_cached_objects = btrfs_free_cached_objects,
2501#ifdef CONFIG_BTRFS_EXPERIMENTAL
2502	.remove_bdev	= btrfs_remove_bdev,
2503	.shutdown	= btrfs_shutdown,
2504#endif
2505};
2506
2507static const struct file_operations btrfs_ctl_fops = {
2508	.open = btrfs_control_open,
2509	.unlocked_ioctl	 = btrfs_control_ioctl,
2510	.compat_ioctl = compat_ptr_ioctl,
2511	.owner	 = THIS_MODULE,
2512	.llseek = noop_llseek,
2513};
2514
2515static struct miscdevice btrfs_misc = {
2516	.minor		= BTRFS_MINOR,
2517	.name		= "btrfs-control",
2518	.fops		= &btrfs_ctl_fops
2519};
2520
2521MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2522MODULE_ALIAS("devname:btrfs-control");
2523
2524static int __init btrfs_interface_init(void)
2525{
2526	return misc_register(&btrfs_misc);
2527}
2528
2529static __cold void btrfs_interface_exit(void)
2530{
2531	misc_deregister(&btrfs_misc);
2532}
2533
2534static int __init btrfs_print_mod_info(void)
2535{
2536	static const char options[] = ""
2537#ifdef CONFIG_BTRFS_EXPERIMENTAL
2538			", experimental=on"
2539#endif
2540#ifdef CONFIG_BTRFS_DEBUG
2541			", debug=on"
2542#endif
2543#ifdef CONFIG_BTRFS_ASSERT
2544			", assert=on"
2545#endif
2546#ifdef CONFIG_BLK_DEV_ZONED
2547			", zoned=yes"
2548#else
2549			", zoned=no"
2550#endif
2551#ifdef CONFIG_FS_VERITY
2552			", fsverity=yes"
2553#else
2554			", fsverity=no"
2555#endif
2556			;
2557
2558#ifdef CONFIG_BTRFS_EXPERIMENTAL
2559	if (btrfs_get_mod_read_policy() == NULL)
2560		pr_info("Btrfs loaded%s\n", options);
2561	else
2562		pr_info("Btrfs loaded%s, read_policy=%s\n",
2563			 options, btrfs_get_mod_read_policy());
2564#else
2565	pr_info("Btrfs loaded%s\n", options);
2566#endif
2567
2568	return 0;
2569}
2570
2571static int register_btrfs(void)
2572{
2573	return register_filesystem(&btrfs_fs_type);
2574}
2575
2576static void unregister_btrfs(void)
2577{
2578	unregister_filesystem(&btrfs_fs_type);
2579}
2580
2581/* Helper structure for long init/exit functions. */
2582struct init_sequence {
2583	int (*init_func)(void);
2584	/* Can be NULL if the init_func doesn't need cleanup. */
2585	void (*exit_func)(void);
2586};
2587
2588static const struct init_sequence mod_init_seq[] = {
2589	{
2590		.init_func = btrfs_props_init,
2591		.exit_func = NULL,
2592	}, {
2593		.init_func = btrfs_init_sysfs,
2594		.exit_func = btrfs_exit_sysfs,
2595	}, {
2596		.init_func = btrfs_init_compress,
2597		.exit_func = btrfs_exit_compress,
2598	}, {
2599		.init_func = btrfs_init_cachep,
2600		.exit_func = btrfs_destroy_cachep,
2601	}, {
2602		.init_func = btrfs_init_dio,
2603		.exit_func = btrfs_destroy_dio,
2604	}, {
2605		.init_func = btrfs_transaction_init,
2606		.exit_func = btrfs_transaction_exit,
2607	}, {
2608		.init_func = btrfs_ctree_init,
2609		.exit_func = btrfs_ctree_exit,
2610	}, {
2611		.init_func = btrfs_free_space_init,
2612		.exit_func = btrfs_free_space_exit,
2613	}, {
2614		.init_func = btrfs_extent_state_init_cachep,
2615		.exit_func = btrfs_extent_state_free_cachep,
2616	}, {
2617		.init_func = extent_buffer_init_cachep,
2618		.exit_func = extent_buffer_free_cachep,
2619	}, {
2620		.init_func = btrfs_bioset_init,
2621		.exit_func = btrfs_bioset_exit,
2622	}, {
2623		.init_func = btrfs_extent_map_init,
2624		.exit_func = btrfs_extent_map_exit,
2625#ifdef CONFIG_BTRFS_EXPERIMENTAL
2626	}, {
2627		.init_func = btrfs_read_policy_init,
2628		.exit_func = NULL,
2629#endif
2630	}, {
2631		.init_func = ordered_data_init,
2632		.exit_func = ordered_data_exit,
2633	}, {
2634		.init_func = btrfs_delayed_inode_init,
2635		.exit_func = btrfs_delayed_inode_exit,
2636	}, {
2637		.init_func = btrfs_auto_defrag_init,
2638		.exit_func = btrfs_auto_defrag_exit,
2639	}, {
2640		.init_func = btrfs_delayed_ref_init,
2641		.exit_func = btrfs_delayed_ref_exit,
2642	}, {
2643		.init_func = btrfs_prelim_ref_init,
2644		.exit_func = btrfs_prelim_ref_exit,
2645	}, {
2646		.init_func = btrfs_interface_init,
2647		.exit_func = btrfs_interface_exit,
2648	}, {
2649		.init_func = btrfs_print_mod_info,
2650		.exit_func = NULL,
2651	}, {
2652		.init_func = btrfs_run_sanity_tests,
2653		.exit_func = NULL,
2654	}, {
2655		.init_func = register_btrfs,
2656		.exit_func = unregister_btrfs,
2657	}
2658};
2659
2660static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2661
2662static __always_inline void btrfs_exit_btrfs_fs(void)
2663{
2664	int i;
2665
2666	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2667		if (!mod_init_result[i])
2668			continue;
2669		if (mod_init_seq[i].exit_func)
2670			mod_init_seq[i].exit_func();
2671		mod_init_result[i] = false;
2672	}
2673}
2674
2675static void __exit exit_btrfs_fs(void)
2676{
2677	btrfs_exit_btrfs_fs();
2678	btrfs_cleanup_fs_uuids();
2679}
2680
2681static int __init init_btrfs_fs(void)
2682{
2683	int ret;
2684	int i;
2685
2686	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2687		ASSERT(!mod_init_result[i]);
2688		ret = mod_init_seq[i].init_func();
2689		if (ret < 0) {
2690			btrfs_exit_btrfs_fs();
2691			return ret;
2692		}
2693		mod_init_result[i] = true;
2694	}
2695	return 0;
2696}
2697
2698late_initcall(init_btrfs_fs);
2699module_exit(exit_btrfs_fs)
2700
2701MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2702MODULE_LICENSE("GPL");
2703MODULE_SOFTDEP("pre: crc32c");
2704MODULE_SOFTDEP("pre: xxhash64");
2705MODULE_SOFTDEP("pre: sha256");
2706MODULE_SOFTDEP("pre: blake2b-256");