Linux Audio

Check our new training course

Loading...
v6.13.7
   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	unsigned int compress_level;
  88	refcount_t refs;
  89};
  90
  91enum {
  92	Opt_acl,
  93	Opt_clear_cache,
  94	Opt_commit_interval,
  95	Opt_compress,
  96	Opt_compress_force,
  97	Opt_compress_force_type,
  98	Opt_compress_type,
  99	Opt_degraded,
 100	Opt_device,
 101	Opt_fatal_errors,
 102	Opt_flushoncommit,
 103	Opt_max_inline,
 104	Opt_barrier,
 105	Opt_datacow,
 106	Opt_datasum,
 107	Opt_defrag,
 108	Opt_discard,
 109	Opt_discard_mode,
 110	Opt_ratio,
 111	Opt_rescan_uuid_tree,
 112	Opt_skip_balance,
 113	Opt_space_cache,
 114	Opt_space_cache_version,
 115	Opt_ssd,
 116	Opt_ssd_spread,
 117	Opt_subvol,
 118	Opt_subvol_empty,
 119	Opt_subvolid,
 120	Opt_thread_pool,
 121	Opt_treelog,
 122	Opt_user_subvol_rm_allowed,
 123	Opt_norecovery,
 124
 125	/* Rescue options */
 126	Opt_rescue,
 127	Opt_usebackuproot,
 128	Opt_nologreplay,
 
 
 
 129
 130	/* Debugging options */
 131	Opt_enospc_debug,
 132#ifdef CONFIG_BTRFS_DEBUG
 133	Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 134#endif
 135#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 136	Opt_ref_verify,
 137#endif
 138	Opt_err,
 139};
 140
 141enum {
 142	Opt_fatal_errors_panic,
 143	Opt_fatal_errors_bug,
 144};
 145
 146static const struct constant_table btrfs_parameter_fatal_errors[] = {
 147	{ "panic", Opt_fatal_errors_panic },
 148	{ "bug", Opt_fatal_errors_bug },
 149	{}
 150};
 151
 152enum {
 153	Opt_discard_sync,
 154	Opt_discard_async,
 155};
 156
 157static const struct constant_table btrfs_parameter_discard[] = {
 158	{ "sync", Opt_discard_sync },
 159	{ "async", Opt_discard_async },
 160	{}
 161};
 162
 163enum {
 164	Opt_space_cache_v1,
 165	Opt_space_cache_v2,
 166};
 167
 168static const struct constant_table btrfs_parameter_space_cache[] = {
 169	{ "v1", Opt_space_cache_v1 },
 170	{ "v2", Opt_space_cache_v2 },
 171	{}
 172};
 173
 174enum {
 175	Opt_rescue_usebackuproot,
 176	Opt_rescue_nologreplay,
 177	Opt_rescue_ignorebadroots,
 178	Opt_rescue_ignoredatacsums,
 179	Opt_rescue_ignoremetacsums,
 180	Opt_rescue_ignoresuperflags,
 181	Opt_rescue_parameter_all,
 182};
 183
 184static const struct constant_table btrfs_parameter_rescue[] = {
 185	{ "usebackuproot", Opt_rescue_usebackuproot },
 186	{ "nologreplay", Opt_rescue_nologreplay },
 187	{ "ignorebadroots", Opt_rescue_ignorebadroots },
 188	{ "ibadroots", Opt_rescue_ignorebadroots },
 189	{ "ignoredatacsums", Opt_rescue_ignoredatacsums },
 190	{ "ignoremetacsums", Opt_rescue_ignoremetacsums},
 191	{ "ignoresuperflags", Opt_rescue_ignoresuperflags},
 192	{ "idatacsums", Opt_rescue_ignoredatacsums },
 193	{ "imetacsums", Opt_rescue_ignoremetacsums},
 194	{ "isuperflags", Opt_rescue_ignoresuperflags},
 195	{ "all", Opt_rescue_parameter_all },
 196	{}
 197};
 198
 199#ifdef CONFIG_BTRFS_DEBUG
 200enum {
 201	Opt_fragment_parameter_data,
 202	Opt_fragment_parameter_metadata,
 203	Opt_fragment_parameter_all,
 204};
 205
 206static const struct constant_table btrfs_parameter_fragment[] = {
 207	{ "data", Opt_fragment_parameter_data },
 208	{ "metadata", Opt_fragment_parameter_metadata },
 209	{ "all", Opt_fragment_parameter_all },
 210	{}
 211};
 212#endif
 213
 214static const struct fs_parameter_spec btrfs_fs_parameters[] = {
 215	fsparam_flag_no("acl", Opt_acl),
 216	fsparam_flag_no("autodefrag", Opt_defrag),
 217	fsparam_flag_no("barrier", Opt_barrier),
 218	fsparam_flag("clear_cache", Opt_clear_cache),
 219	fsparam_u32("commit", Opt_commit_interval),
 220	fsparam_flag("compress", Opt_compress),
 221	fsparam_string("compress", Opt_compress_type),
 222	fsparam_flag("compress-force", Opt_compress_force),
 223	fsparam_string("compress-force", Opt_compress_force_type),
 224	fsparam_flag_no("datacow", Opt_datacow),
 225	fsparam_flag_no("datasum", Opt_datasum),
 226	fsparam_flag("degraded", Opt_degraded),
 227	fsparam_string("device", Opt_device),
 228	fsparam_flag_no("discard", Opt_discard),
 229	fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
 230	fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
 231	fsparam_flag_no("flushoncommit", Opt_flushoncommit),
 232	fsparam_string("max_inline", Opt_max_inline),
 233	fsparam_u32("metadata_ratio", Opt_ratio),
 234	fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
 235	fsparam_flag("skip_balance", Opt_skip_balance),
 236	fsparam_flag_no("space_cache", Opt_space_cache),
 237	fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
 238	fsparam_flag_no("ssd", Opt_ssd),
 239	fsparam_flag_no("ssd_spread", Opt_ssd_spread),
 240	fsparam_string("subvol", Opt_subvol),
 241	fsparam_flag("subvol=", Opt_subvol_empty),
 242	fsparam_u64("subvolid", Opt_subvolid),
 243	fsparam_u32("thread_pool", Opt_thread_pool),
 244	fsparam_flag_no("treelog", Opt_treelog),
 245	fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
 246
 247	/* Rescue options. */
 248	fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
 249	/* Deprecated, with alias rescue=nologreplay */
 250	__fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
 251	/* Deprecated, with alias rescue=usebackuproot */
 252	__fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
 253	/* For compatibility only, alias for "rescue=nologreplay". */
 254	fsparam_flag("norecovery", Opt_norecovery),
 255
 256	/* Debugging options. */
 257	fsparam_flag_no("enospc_debug", Opt_enospc_debug),
 258#ifdef CONFIG_BTRFS_DEBUG
 259	fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
 260#endif
 261#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 262	fsparam_flag("ref_verify", Opt_ref_verify),
 263#endif
 264	{}
 265};
 266
 267/* No support for restricting writes to btrfs devices yet... */
 268static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
 269{
 270	return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
 271}
 272
 273static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
 274{
 275	struct btrfs_fs_context *ctx = fc->fs_private;
 276	struct fs_parse_result result;
 277	int opt;
 278
 279	opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
 280	if (opt < 0)
 281		return opt;
 282
 283	switch (opt) {
 284	case Opt_degraded:
 285		btrfs_set_opt(ctx->mount_opt, DEGRADED);
 286		break;
 287	case Opt_subvol_empty:
 288		/*
 289		 * This exists because we used to allow it on accident, so we're
 290		 * keeping it to maintain ABI.  See 37becec95ac3 ("Btrfs: allow
 291		 * empty subvol= again").
 292		 */
 293		break;
 294	case Opt_subvol:
 295		kfree(ctx->subvol_name);
 296		ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
 297		if (!ctx->subvol_name)
 298			return -ENOMEM;
 299		break;
 300	case Opt_subvolid:
 301		ctx->subvol_objectid = result.uint_64;
 302
 303		/* subvolid=0 means give me the original fs_tree. */
 304		if (!ctx->subvol_objectid)
 305			ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
 306		break;
 307	case Opt_device: {
 308		struct btrfs_device *device;
 309		blk_mode_t mode = btrfs_open_mode(fc);
 310
 311		mutex_lock(&uuid_mutex);
 312		device = btrfs_scan_one_device(param->string, mode, false);
 313		mutex_unlock(&uuid_mutex);
 314		if (IS_ERR(device))
 315			return PTR_ERR(device);
 316		break;
 317	}
 318	case Opt_datasum:
 319		if (result.negated) {
 320			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 321		} else {
 322			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 323			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 324		}
 325		break;
 326	case Opt_datacow:
 327		if (result.negated) {
 328			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 329			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 330			btrfs_set_opt(ctx->mount_opt, NODATACOW);
 331			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 332		} else {
 333			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 334		}
 335		break;
 336	case Opt_compress_force:
 337	case Opt_compress_force_type:
 338		btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
 339		fallthrough;
 340	case Opt_compress:
 341	case Opt_compress_type:
 342		/*
 343		 * Provide the same semantics as older kernels that don't use fs
 344		 * context, specifying the "compress" option clears
 345		 * "force-compress" without the need to pass
 346		 * "compress-force=[no|none]" before specifying "compress".
 347		 */
 348		if (opt != Opt_compress_force && opt != Opt_compress_force_type)
 349			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 350
 351		if (opt == Opt_compress || opt == Opt_compress_force) {
 352			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 353			ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 354			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 355			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 356			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 357		} else if (strncmp(param->string, "zlib", 4) == 0) {
 358			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 359			ctx->compress_level =
 360				btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
 361							 param->string + 4);
 362			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 363			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 364			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 365		} else if (strncmp(param->string, "lzo", 3) == 0) {
 366			ctx->compress_type = BTRFS_COMPRESS_LZO;
 367			ctx->compress_level = 0;
 368			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 369			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 370			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 371		} else if (strncmp(param->string, "zstd", 4) == 0) {
 372			ctx->compress_type = BTRFS_COMPRESS_ZSTD;
 373			ctx->compress_level =
 374				btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
 375							 param->string + 4);
 376			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 377			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 378			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 379		} else if (strncmp(param->string, "no", 2) == 0) {
 380			ctx->compress_level = 0;
 381			ctx->compress_type = 0;
 382			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 383			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 384		} else {
 385			btrfs_err(NULL, "unrecognized compression value %s",
 386				  param->string);
 387			return -EINVAL;
 388		}
 389		break;
 390	case Opt_ssd:
 391		if (result.negated) {
 392			btrfs_set_opt(ctx->mount_opt, NOSSD);
 393			btrfs_clear_opt(ctx->mount_opt, SSD);
 394			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 395		} else {
 396			btrfs_set_opt(ctx->mount_opt, SSD);
 397			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 398		}
 399		break;
 400	case Opt_ssd_spread:
 401		if (result.negated) {
 402			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 403		} else {
 404			btrfs_set_opt(ctx->mount_opt, SSD);
 405			btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
 406			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 407		}
 408		break;
 409	case Opt_barrier:
 410		if (result.negated)
 411			btrfs_set_opt(ctx->mount_opt, NOBARRIER);
 412		else
 413			btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
 414		break;
 415	case Opt_thread_pool:
 416		if (result.uint_32 == 0) {
 417			btrfs_err(NULL, "invalid value 0 for thread_pool");
 418			return -EINVAL;
 419		}
 420		ctx->thread_pool_size = result.uint_32;
 421		break;
 422	case Opt_max_inline:
 423		ctx->max_inline = memparse(param->string, NULL);
 424		break;
 425	case Opt_acl:
 426		if (result.negated) {
 427			fc->sb_flags &= ~SB_POSIXACL;
 428		} else {
 429#ifdef CONFIG_BTRFS_FS_POSIX_ACL
 430			fc->sb_flags |= SB_POSIXACL;
 431#else
 432			btrfs_err(NULL, "support for ACL not compiled in");
 433			return -EINVAL;
 434#endif
 435		}
 436		/*
 437		 * VFS limits the ability to toggle ACL on and off via remount,
 438		 * despite every file system allowing this.  This seems to be
 439		 * an oversight since we all do, but it'll fail if we're
 440		 * remounting.  So don't set the mask here, we'll check it in
 441		 * btrfs_reconfigure and do the toggling ourselves.
 442		 */
 443		if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
 444			fc->sb_flags_mask |= SB_POSIXACL;
 445		break;
 446	case Opt_treelog:
 447		if (result.negated)
 448			btrfs_set_opt(ctx->mount_opt, NOTREELOG);
 449		else
 450			btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
 451		break;
 452	case Opt_nologreplay:
 453		btrfs_warn(NULL,
 454		"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 455		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 456		break;
 457	case Opt_norecovery:
 458		btrfs_info(NULL,
 459"'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
 460		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 461		break;
 462	case Opt_flushoncommit:
 463		if (result.negated)
 464			btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
 465		else
 466			btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
 467		break;
 468	case Opt_ratio:
 469		ctx->metadata_ratio = result.uint_32;
 470		break;
 471	case Opt_discard:
 472		if (result.negated) {
 473			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 474			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 475			btrfs_set_opt(ctx->mount_opt, NODISCARD);
 476		} else {
 477			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 478			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 479		}
 480		break;
 481	case Opt_discard_mode:
 482		switch (result.uint_32) {
 483		case Opt_discard_sync:
 484			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 485			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 486			break;
 487		case Opt_discard_async:
 488			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 489			btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
 490			break;
 491		default:
 492			btrfs_err(NULL, "unrecognized discard mode value %s",
 493				  param->key);
 494			return -EINVAL;
 495		}
 496		btrfs_clear_opt(ctx->mount_opt, NODISCARD);
 497		break;
 498	case Opt_space_cache:
 499		if (result.negated) {
 500			btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
 501			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 502			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 503		} else {
 504			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 505			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 506		}
 507		break;
 508	case Opt_space_cache_version:
 509		switch (result.uint_32) {
 510		case Opt_space_cache_v1:
 511			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 512			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 513			break;
 514		case Opt_space_cache_v2:
 515			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 516			btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
 517			break;
 518		default:
 519			btrfs_err(NULL, "unrecognized space_cache value %s",
 520				  param->key);
 521			return -EINVAL;
 522		}
 523		break;
 524	case Opt_rescan_uuid_tree:
 525		btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
 526		break;
 527	case Opt_clear_cache:
 528		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 529		break;
 530	case Opt_user_subvol_rm_allowed:
 531		btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
 532		break;
 533	case Opt_enospc_debug:
 534		if (result.negated)
 535			btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
 536		else
 537			btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
 538		break;
 539	case Opt_defrag:
 540		if (result.negated)
 541			btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
 542		else
 543			btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
 544		break;
 545	case Opt_usebackuproot:
 546		btrfs_warn(NULL,
 547			   "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
 548		btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 549
 550		/* If we're loading the backup roots we can't trust the space cache. */
 551		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 552		break;
 553	case Opt_skip_balance:
 554		btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
 555		break;
 556	case Opt_fatal_errors:
 557		switch (result.uint_32) {
 558		case Opt_fatal_errors_panic:
 559			btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 560			break;
 561		case Opt_fatal_errors_bug:
 562			btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 563			break;
 564		default:
 565			btrfs_err(NULL, "unrecognized fatal_errors value %s",
 566				  param->key);
 567			return -EINVAL;
 568		}
 569		break;
 570	case Opt_commit_interval:
 571		ctx->commit_interval = result.uint_32;
 572		if (ctx->commit_interval == 0)
 573			ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 574		break;
 575	case Opt_rescue:
 576		switch (result.uint_32) {
 577		case Opt_rescue_usebackuproot:
 578			btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 579			break;
 580		case Opt_rescue_nologreplay:
 581			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 582			break;
 583		case Opt_rescue_ignorebadroots:
 584			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 585			break;
 586		case Opt_rescue_ignoredatacsums:
 587			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 588			break;
 589		case Opt_rescue_ignoremetacsums:
 590			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
 591			break;
 592		case Opt_rescue_ignoresuperflags:
 593			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
 594			break;
 595		case Opt_rescue_parameter_all:
 596			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 597			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
 598			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
 599			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 600			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 601			break;
 602		default:
 603			btrfs_info(NULL, "unrecognized rescue option '%s'",
 604				   param->key);
 605			return -EINVAL;
 606		}
 607		break;
 608#ifdef CONFIG_BTRFS_DEBUG
 609	case Opt_fragment:
 610		switch (result.uint_32) {
 611		case Opt_fragment_parameter_all:
 612			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 613			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 614			break;
 615		case Opt_fragment_parameter_metadata:
 616			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 617			break;
 618		case Opt_fragment_parameter_data:
 619			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 620			break;
 621		default:
 622			btrfs_info(NULL, "unrecognized fragment option '%s'",
 623				   param->key);
 624			return -EINVAL;
 625		}
 626		break;
 627#endif
 628#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 629	case Opt_ref_verify:
 630		btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
 631		break;
 632#endif
 633	default:
 634		btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
 635		return -EINVAL;
 636	}
 637
 638	return 0;
 639}
 640
 641/*
 642 * Some options only have meaning at mount time and shouldn't persist across
 643 * remounts, or be displayed. Clear these at the end of mount and remount code
 644 * paths.
 645 */
 646static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
 647{
 648	btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
 649	btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
 650	btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
 651}
 652
 653static bool check_ro_option(const struct btrfs_fs_info *fs_info,
 654			    unsigned long long mount_opt, unsigned long long opt,
 655			    const char *opt_name)
 656{
 657	if (mount_opt & opt) {
 658		btrfs_err(fs_info, "%s must be used with ro mount option",
 659			  opt_name);
 660		return true;
 661	}
 662	return false;
 663}
 664
 665bool btrfs_check_options(const struct btrfs_fs_info *info,
 666			 unsigned long long *mount_opt,
 667			 unsigned long flags)
 668{
 669	bool ret = true;
 670
 671	if (!(flags & SB_RDONLY) &&
 672	    (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
 673	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
 674	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
 675	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
 676	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
 677		ret = false;
 678
 679	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 680	    !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
 681	    !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
 682		btrfs_err(info, "cannot disable free-space-tree");
 683		ret = false;
 684	}
 685	if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
 686	     !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
 687		btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
 688		ret = false;
 689	}
 690
 691	if (btrfs_check_mountopts_zoned(info, mount_opt))
 692		ret = false;
 693
 694	if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
 695		if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
 696			btrfs_info(info, "disk space caching is enabled");
 697			btrfs_warn(info,
 698"space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
 699		}
 700		if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
 701			btrfs_info(info, "using free-space-tree");
 702	}
 703
 704	return ret;
 705}
 706
 707/*
 708 * This is subtle, we only call this during open_ctree().  We need to pre-load
 709 * the mount options with the on-disk settings.  Before the new mount API took
 710 * effect we would do this on mount and remount.  With the new mount API we'll
 711 * only do this on the initial mount.
 712 *
 713 * This isn't a change in behavior, because we're using the current state of the
 714 * file system to set the current mount options.  If you mounted with special
 715 * options to disable these features and then remounted we wouldn't revert the
 716 * settings, because mounting without these features cleared the on-disk
 717 * settings, so this being called on re-mount is not needed.
 718 */
 719void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
 720{
 721	if (fs_info->sectorsize < PAGE_SIZE) {
 722		btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
 723		if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
 724			btrfs_info(fs_info,
 725				   "forcing free space tree for sector size %u with page size %lu",
 726				   fs_info->sectorsize, PAGE_SIZE);
 727			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 728		}
 729	}
 730
 731	/*
 732	 * At this point our mount options are populated, so we only mess with
 733	 * these settings if we don't have any settings already.
 734	 */
 735	if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
 736		return;
 737
 738	if (btrfs_is_zoned(fs_info) &&
 739	    btrfs_free_space_cache_v1_active(fs_info)) {
 740		btrfs_info(fs_info, "zoned: clearing existing space cache");
 741		btrfs_set_super_cache_generation(fs_info->super_copy, 0);
 742		return;
 743	}
 744
 745	if (btrfs_test_opt(fs_info, SPACE_CACHE))
 746		return;
 747
 748	if (btrfs_test_opt(fs_info, NOSPACECACHE))
 749		return;
 750
 751	/*
 752	 * At this point we don't have explicit options set by the user, set
 753	 * them ourselves based on the state of the file system.
 754	 */
 755	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
 756		btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 757	else if (btrfs_free_space_cache_v1_active(fs_info))
 758		btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
 759}
 760
 761static void set_device_specific_options(struct btrfs_fs_info *fs_info)
 762{
 763	if (!btrfs_test_opt(fs_info, NOSSD) &&
 764	    !fs_info->fs_devices->rotating)
 765		btrfs_set_opt(fs_info->mount_opt, SSD);
 766
 767	/*
 768	 * For devices supporting discard turn on discard=async automatically,
 769	 * unless it's already set or disabled. This could be turned off by
 770	 * nodiscard for the same mount.
 771	 *
 772	 * The zoned mode piggy backs on the discard functionality for
 773	 * resetting a zone. There is no reason to delay the zone reset as it is
 774	 * fast enough. So, do not enable async discard for zoned mode.
 775	 */
 776	if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
 777	      btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
 778	      btrfs_test_opt(fs_info, NODISCARD)) &&
 779	    fs_info->fs_devices->discardable &&
 780	    !btrfs_is_zoned(fs_info))
 781		btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
 782}
 783
 784char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 785					  u64 subvol_objectid)
 786{
 787	struct btrfs_root *root = fs_info->tree_root;
 788	struct btrfs_root *fs_root = NULL;
 789	struct btrfs_root_ref *root_ref;
 790	struct btrfs_inode_ref *inode_ref;
 791	struct btrfs_key key;
 792	struct btrfs_path *path = NULL;
 793	char *name = NULL, *ptr;
 794	u64 dirid;
 795	int len;
 796	int ret;
 797
 798	path = btrfs_alloc_path();
 799	if (!path) {
 800		ret = -ENOMEM;
 801		goto err;
 802	}
 803
 804	name = kmalloc(PATH_MAX, GFP_KERNEL);
 805	if (!name) {
 806		ret = -ENOMEM;
 807		goto err;
 808	}
 809	ptr = name + PATH_MAX - 1;
 810	ptr[0] = '\0';
 811
 812	/*
 813	 * Walk up the subvolume trees in the tree of tree roots by root
 814	 * backrefs until we hit the top-level subvolume.
 815	 */
 816	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
 817		key.objectid = subvol_objectid;
 818		key.type = BTRFS_ROOT_BACKREF_KEY;
 819		key.offset = (u64)-1;
 820
 821		ret = btrfs_search_backwards(root, &key, path);
 822		if (ret < 0) {
 823			goto err;
 824		} else if (ret > 0) {
 825			ret = -ENOENT;
 826			goto err;
 827		}
 828
 829		subvol_objectid = key.offset;
 830
 831		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
 832					  struct btrfs_root_ref);
 833		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
 834		ptr -= len + 1;
 835		if (ptr < name) {
 836			ret = -ENAMETOOLONG;
 837			goto err;
 838		}
 839		read_extent_buffer(path->nodes[0], ptr + 1,
 840				   (unsigned long)(root_ref + 1), len);
 841		ptr[0] = '/';
 842		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
 843		btrfs_release_path(path);
 844
 845		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
 846		if (IS_ERR(fs_root)) {
 847			ret = PTR_ERR(fs_root);
 848			fs_root = NULL;
 849			goto err;
 850		}
 851
 852		/*
 853		 * Walk up the filesystem tree by inode refs until we hit the
 854		 * root directory.
 855		 */
 856		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
 857			key.objectid = dirid;
 858			key.type = BTRFS_INODE_REF_KEY;
 859			key.offset = (u64)-1;
 860
 861			ret = btrfs_search_backwards(fs_root, &key, path);
 862			if (ret < 0) {
 863				goto err;
 864			} else if (ret > 0) {
 865				ret = -ENOENT;
 866				goto err;
 867			}
 868
 869			dirid = key.offset;
 870
 871			inode_ref = btrfs_item_ptr(path->nodes[0],
 872						   path->slots[0],
 873						   struct btrfs_inode_ref);
 874			len = btrfs_inode_ref_name_len(path->nodes[0],
 875						       inode_ref);
 876			ptr -= len + 1;
 877			if (ptr < name) {
 878				ret = -ENAMETOOLONG;
 879				goto err;
 880			}
 881			read_extent_buffer(path->nodes[0], ptr + 1,
 882					   (unsigned long)(inode_ref + 1), len);
 883			ptr[0] = '/';
 884			btrfs_release_path(path);
 885		}
 886		btrfs_put_root(fs_root);
 887		fs_root = NULL;
 888	}
 889
 890	btrfs_free_path(path);
 891	if (ptr == name + PATH_MAX - 1) {
 892		name[0] = '/';
 893		name[1] = '\0';
 894	} else {
 895		memmove(name, ptr, name + PATH_MAX - ptr);
 896	}
 897	return name;
 898
 899err:
 900	btrfs_put_root(fs_root);
 901	btrfs_free_path(path);
 902	kfree(name);
 903	return ERR_PTR(ret);
 904}
 905
 906static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
 907{
 908	struct btrfs_root *root = fs_info->tree_root;
 909	struct btrfs_dir_item *di;
 910	struct btrfs_path *path;
 911	struct btrfs_key location;
 912	struct fscrypt_str name = FSTR_INIT("default", 7);
 913	u64 dir_id;
 914
 915	path = btrfs_alloc_path();
 916	if (!path)
 917		return -ENOMEM;
 918
 919	/*
 920	 * Find the "default" dir item which points to the root item that we
 921	 * will mount by default if we haven't been given a specific subvolume
 922	 * to mount.
 923	 */
 924	dir_id = btrfs_super_root_dir(fs_info->super_copy);
 925	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
 926	if (IS_ERR(di)) {
 927		btrfs_free_path(path);
 928		return PTR_ERR(di);
 929	}
 930	if (!di) {
 931		/*
 932		 * Ok the default dir item isn't there.  This is weird since
 933		 * it's always been there, but don't freak out, just try and
 934		 * mount the top-level subvolume.
 935		 */
 936		btrfs_free_path(path);
 937		*objectid = BTRFS_FS_TREE_OBJECTID;
 938		return 0;
 939	}
 940
 941	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 942	btrfs_free_path(path);
 943	*objectid = location.objectid;
 944	return 0;
 945}
 946
 947static int btrfs_fill_super(struct super_block *sb,
 948			    struct btrfs_fs_devices *fs_devices)
 
 949{
 950	struct inode *inode;
 951	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 952	int err;
 953
 954	sb->s_maxbytes = MAX_LFS_FILESIZE;
 955	sb->s_magic = BTRFS_SUPER_MAGIC;
 956	sb->s_op = &btrfs_super_ops;
 957	sb->s_d_op = &btrfs_dentry_operations;
 958	sb->s_export_op = &btrfs_export_ops;
 959#ifdef CONFIG_FS_VERITY
 960	sb->s_vop = &btrfs_verityops;
 961#endif
 962	sb->s_xattr = btrfs_xattr_handlers;
 963	sb->s_time_gran = 1;
 964	sb->s_iflags |= SB_I_CGROUPWB;
 965
 966	err = super_setup_bdi(sb);
 967	if (err) {
 968		btrfs_err(fs_info, "super_setup_bdi failed");
 969		return err;
 970	}
 971
 972	err = open_ctree(sb, fs_devices);
 973	if (err) {
 974		btrfs_err(fs_info, "open_ctree failed: %d", err);
 975		return err;
 976	}
 977
 978	inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
 979	if (IS_ERR(inode)) {
 980		err = PTR_ERR(inode);
 981		btrfs_handle_fs_error(fs_info, err, NULL);
 982		goto fail_close;
 983	}
 984
 985	sb->s_root = d_make_root(inode);
 986	if (!sb->s_root) {
 987		err = -ENOMEM;
 988		goto fail_close;
 989	}
 990
 991	sb->s_flags |= SB_ACTIVE;
 992	return 0;
 993
 994fail_close:
 995	close_ctree(fs_info);
 996	return err;
 997}
 998
 999int btrfs_sync_fs(struct super_block *sb, int wait)
1000{
1001	struct btrfs_trans_handle *trans;
1002	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1003	struct btrfs_root *root = fs_info->tree_root;
1004
1005	trace_btrfs_sync_fs(fs_info, wait);
1006
1007	if (!wait) {
1008		filemap_flush(fs_info->btree_inode->i_mapping);
1009		return 0;
1010	}
1011
1012	btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1013
1014	trans = btrfs_attach_transaction_barrier(root);
1015	if (IS_ERR(trans)) {
1016		/* no transaction, don't bother */
1017		if (PTR_ERR(trans) == -ENOENT) {
1018			/*
1019			 * Exit unless we have some pending changes
1020			 * that need to go through commit
1021			 */
1022			if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1023				      &fs_info->flags))
1024				return 0;
1025			/*
1026			 * A non-blocking test if the fs is frozen. We must not
1027			 * start a new transaction here otherwise a deadlock
1028			 * happens. The pending operations are delayed to the
1029			 * next commit after thawing.
1030			 */
1031			if (sb_start_write_trylock(sb))
1032				sb_end_write(sb);
1033			else
1034				return 0;
1035			trans = btrfs_start_transaction(root, 0);
1036		}
1037		if (IS_ERR(trans))
1038			return PTR_ERR(trans);
1039	}
1040	return btrfs_commit_transaction(trans);
1041}
1042
1043static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1044{
1045	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1046	*printed = true;
1047}
1048
1049static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1050{
1051	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1052	const char *compress_type;
1053	const char *subvol_name;
1054	bool printed = false;
1055
1056	if (btrfs_test_opt(info, DEGRADED))
1057		seq_puts(seq, ",degraded");
1058	if (btrfs_test_opt(info, NODATASUM))
1059		seq_puts(seq, ",nodatasum");
1060	if (btrfs_test_opt(info, NODATACOW))
1061		seq_puts(seq, ",nodatacow");
1062	if (btrfs_test_opt(info, NOBARRIER))
1063		seq_puts(seq, ",nobarrier");
1064	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1065		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1066	if (info->thread_pool_size !=  min_t(unsigned long,
1067					     num_online_cpus() + 2, 8))
1068		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1069	if (btrfs_test_opt(info, COMPRESS)) {
1070		compress_type = btrfs_compress_type2str(info->compress_type);
1071		if (btrfs_test_opt(info, FORCE_COMPRESS))
1072			seq_printf(seq, ",compress-force=%s", compress_type);
1073		else
1074			seq_printf(seq, ",compress=%s", compress_type);
1075		if (info->compress_level)
1076			seq_printf(seq, ":%d", info->compress_level);
1077	}
1078	if (btrfs_test_opt(info, NOSSD))
1079		seq_puts(seq, ",nossd");
1080	if (btrfs_test_opt(info, SSD_SPREAD))
1081		seq_puts(seq, ",ssd_spread");
1082	else if (btrfs_test_opt(info, SSD))
1083		seq_puts(seq, ",ssd");
1084	if (btrfs_test_opt(info, NOTREELOG))
1085		seq_puts(seq, ",notreelog");
1086	if (btrfs_test_opt(info, NOLOGREPLAY))
1087		print_rescue_option(seq, "nologreplay", &printed);
1088	if (btrfs_test_opt(info, USEBACKUPROOT))
1089		print_rescue_option(seq, "usebackuproot", &printed);
1090	if (btrfs_test_opt(info, IGNOREBADROOTS))
1091		print_rescue_option(seq, "ignorebadroots", &printed);
1092	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1093		print_rescue_option(seq, "ignoredatacsums", &printed);
1094	if (btrfs_test_opt(info, IGNOREMETACSUMS))
1095		print_rescue_option(seq, "ignoremetacsums", &printed);
1096	if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1097		print_rescue_option(seq, "ignoresuperflags", &printed);
1098	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1099		seq_puts(seq, ",flushoncommit");
1100	if (btrfs_test_opt(info, DISCARD_SYNC))
1101		seq_puts(seq, ",discard");
1102	if (btrfs_test_opt(info, DISCARD_ASYNC))
1103		seq_puts(seq, ",discard=async");
1104	if (!(info->sb->s_flags & SB_POSIXACL))
1105		seq_puts(seq, ",noacl");
1106	if (btrfs_free_space_cache_v1_active(info))
1107		seq_puts(seq, ",space_cache");
1108	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1109		seq_puts(seq, ",space_cache=v2");
1110	else
1111		seq_puts(seq, ",nospace_cache");
1112	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1113		seq_puts(seq, ",rescan_uuid_tree");
1114	if (btrfs_test_opt(info, CLEAR_CACHE))
1115		seq_puts(seq, ",clear_cache");
1116	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1117		seq_puts(seq, ",user_subvol_rm_allowed");
1118	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1119		seq_puts(seq, ",enospc_debug");
1120	if (btrfs_test_opt(info, AUTO_DEFRAG))
1121		seq_puts(seq, ",autodefrag");
1122	if (btrfs_test_opt(info, SKIP_BALANCE))
1123		seq_puts(seq, ",skip_balance");
1124	if (info->metadata_ratio)
1125		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1126	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1127		seq_puts(seq, ",fatal_errors=panic");
1128	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1129		seq_printf(seq, ",commit=%u", info->commit_interval);
1130#ifdef CONFIG_BTRFS_DEBUG
1131	if (btrfs_test_opt(info, FRAGMENT_DATA))
1132		seq_puts(seq, ",fragment=data");
1133	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1134		seq_puts(seq, ",fragment=metadata");
1135#endif
1136	if (btrfs_test_opt(info, REF_VERIFY))
1137		seq_puts(seq, ",ref_verify");
1138	seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
 
1139	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1140			btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1141	if (!IS_ERR(subvol_name)) {
1142		seq_puts(seq, ",subvol=");
1143		seq_escape(seq, subvol_name, " \t\n\\");
1144		kfree(subvol_name);
1145	}
1146	return 0;
1147}
1148
1149/*
1150 * subvolumes are identified by ino 256
1151 */
1152static inline int is_subvolume_inode(struct inode *inode)
1153{
1154	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1155		return 1;
1156	return 0;
1157}
1158
1159static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1160				   struct vfsmount *mnt)
1161{
1162	struct dentry *root;
1163	int ret;
1164
1165	if (!subvol_name) {
1166		if (!subvol_objectid) {
1167			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1168							  &subvol_objectid);
1169			if (ret) {
1170				root = ERR_PTR(ret);
1171				goto out;
1172			}
1173		}
1174		subvol_name = btrfs_get_subvol_name_from_objectid(
1175					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1176		if (IS_ERR(subvol_name)) {
1177			root = ERR_CAST(subvol_name);
1178			subvol_name = NULL;
1179			goto out;
1180		}
1181
1182	}
1183
1184	root = mount_subtree(mnt, subvol_name);
1185	/* mount_subtree() drops our reference on the vfsmount. */
1186	mnt = NULL;
1187
1188	if (!IS_ERR(root)) {
1189		struct super_block *s = root->d_sb;
1190		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1191		struct inode *root_inode = d_inode(root);
1192		u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1193
1194		ret = 0;
1195		if (!is_subvolume_inode(root_inode)) {
1196			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1197			       subvol_name);
1198			ret = -EINVAL;
1199		}
1200		if (subvol_objectid && root_objectid != subvol_objectid) {
1201			/*
1202			 * This will also catch a race condition where a
1203			 * subvolume which was passed by ID is renamed and
1204			 * another subvolume is renamed over the old location.
1205			 */
1206			btrfs_err(fs_info,
1207				  "subvol '%s' does not match subvolid %llu",
1208				  subvol_name, subvol_objectid);
1209			ret = -EINVAL;
1210		}
1211		if (ret) {
1212			dput(root);
1213			root = ERR_PTR(ret);
1214			deactivate_locked_super(s);
1215		}
1216	}
1217
1218out:
1219	mntput(mnt);
1220	kfree(subvol_name);
1221	return root;
1222}
1223
1224static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1225				     u32 new_pool_size, u32 old_pool_size)
1226{
1227	if (new_pool_size == old_pool_size)
1228		return;
1229
1230	fs_info->thread_pool_size = new_pool_size;
1231
1232	btrfs_info(fs_info, "resize thread pool %d -> %d",
1233	       old_pool_size, new_pool_size);
1234
1235	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1236	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1237	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1238	workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1239	workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1240	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1241	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1242	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1243}
1244
1245static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1246				       unsigned long long old_opts, int flags)
1247{
1248	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1249	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1250	     (flags & SB_RDONLY))) {
1251		/* wait for any defraggers to finish */
1252		wait_event(fs_info->transaction_wait,
1253			   (atomic_read(&fs_info->defrag_running) == 0));
1254		if (flags & SB_RDONLY)
1255			sync_filesystem(fs_info->sb);
1256	}
1257}
1258
1259static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1260					 unsigned long long old_opts)
1261{
1262	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1263
1264	/*
1265	 * We need to cleanup all defragable inodes if the autodefragment is
1266	 * close or the filesystem is read only.
1267	 */
1268	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1269	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1270		btrfs_cleanup_defrag_inodes(fs_info);
1271	}
1272
1273	/* If we toggled discard async */
1274	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1275	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1276		btrfs_discard_resume(fs_info);
1277	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1278		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1279		btrfs_discard_cleanup(fs_info);
1280
1281	/* If we toggled space cache */
1282	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1283		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1284}
1285
1286static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1287{
1288	int ret;
1289
1290	if (BTRFS_FS_ERROR(fs_info)) {
1291		btrfs_err(fs_info,
1292			  "remounting read-write after error is not allowed");
1293		return -EINVAL;
1294	}
1295
1296	if (fs_info->fs_devices->rw_devices == 0)
1297		return -EACCES;
1298
1299	if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1300		btrfs_warn(fs_info,
1301			   "too many missing devices, writable remount is not allowed");
1302		return -EACCES;
1303	}
1304
1305	if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1306		btrfs_warn(fs_info,
1307			   "mount required to replay tree-log, cannot remount read-write");
1308		return -EINVAL;
1309	}
1310
1311	/*
1312	 * NOTE: when remounting with a change that does writes, don't put it
1313	 * anywhere above this point, as we are not sure to be safe to write
1314	 * until we pass the above checks.
1315	 */
1316	ret = btrfs_start_pre_rw_mount(fs_info);
1317	if (ret)
1318		return ret;
1319
1320	btrfs_clear_sb_rdonly(fs_info->sb);
1321
1322	set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1323
1324	/*
1325	 * If we've gone from readonly -> read-write, we need to get our
1326	 * sync/async discard lists in the right state.
1327	 */
1328	btrfs_discard_resume(fs_info);
1329
1330	return 0;
1331}
1332
1333static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1334{
1335	/*
1336	 * This also happens on 'umount -rf' or on shutdown, when the
1337	 * filesystem is busy.
1338	 */
1339	cancel_work_sync(&fs_info->async_reclaim_work);
1340	cancel_work_sync(&fs_info->async_data_reclaim_work);
1341
1342	btrfs_discard_cleanup(fs_info);
1343
1344	/* Wait for the uuid_scan task to finish */
1345	down(&fs_info->uuid_tree_rescan_sem);
1346	/* Avoid complains from lockdep et al. */
1347	up(&fs_info->uuid_tree_rescan_sem);
1348
1349	btrfs_set_sb_rdonly(fs_info->sb);
1350
1351	/*
1352	 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1353	 * loop if it's already active.  If it's already asleep, we'll leave
1354	 * unused block groups on disk until we're mounted read-write again
1355	 * unless we clean them up here.
1356	 */
1357	btrfs_delete_unused_bgs(fs_info);
1358
1359	/*
1360	 * The cleaner task could be already running before we set the flag
1361	 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).  We must make
1362	 * sure that after we finish the remount, i.e. after we call
1363	 * btrfs_commit_super(), the cleaner can no longer start a transaction
1364	 * - either because it was dropping a dead root, running delayed iputs
1365	 *   or deleting an unused block group (the cleaner picked a block
1366	 *   group from the list of unused block groups before we were able to
1367	 *   in the previous call to btrfs_delete_unused_bgs()).
1368	 */
1369	wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1370
1371	/*
1372	 * We've set the superblock to RO mode, so we might have made the
1373	 * cleaner task sleep without running all pending delayed iputs. Go
1374	 * through all the delayed iputs here, so that if an unmount happens
1375	 * without remounting RW we don't end up at finishing close_ctree()
1376	 * with a non-empty list of delayed iputs.
1377	 */
1378	btrfs_run_delayed_iputs(fs_info);
1379
1380	btrfs_dev_replace_suspend_for_unmount(fs_info);
1381	btrfs_scrub_cancel(fs_info);
1382	btrfs_pause_balance(fs_info);
1383
1384	/*
1385	 * Pause the qgroup rescan worker if it is running. We don't want it to
1386	 * be still running after we are in RO mode, as after that, by the time
1387	 * we unmount, it might have left a transaction open, so we would leak
1388	 * the transaction and/or crash.
1389	 */
1390	btrfs_qgroup_wait_for_completion(fs_info, false);
1391
1392	return btrfs_commit_super(fs_info);
1393}
1394
1395static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1396{
1397	fs_info->max_inline = ctx->max_inline;
1398	fs_info->commit_interval = ctx->commit_interval;
1399	fs_info->metadata_ratio = ctx->metadata_ratio;
1400	fs_info->thread_pool_size = ctx->thread_pool_size;
1401	fs_info->mount_opt = ctx->mount_opt;
1402	fs_info->compress_type = ctx->compress_type;
1403	fs_info->compress_level = ctx->compress_level;
1404}
1405
1406static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1407{
1408	ctx->max_inline = fs_info->max_inline;
1409	ctx->commit_interval = fs_info->commit_interval;
1410	ctx->metadata_ratio = fs_info->metadata_ratio;
1411	ctx->thread_pool_size = fs_info->thread_pool_size;
1412	ctx->mount_opt = fs_info->mount_opt;
1413	ctx->compress_type = fs_info->compress_type;
1414	ctx->compress_level = fs_info->compress_level;
1415}
1416
1417#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...)			\
1418do {										\
1419	if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1420	    btrfs_raw_test_opt(fs_info->mount_opt, opt))			\
1421		btrfs_info(fs_info, fmt, ##args);				\
1422} while (0)
1423
1424#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...)	\
1425do {									\
1426	if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1427	    !btrfs_raw_test_opt(fs_info->mount_opt, opt))		\
1428		btrfs_info(fs_info, fmt, ##args);			\
1429} while (0)
1430
1431static void btrfs_emit_options(struct btrfs_fs_info *info,
1432			       struct btrfs_fs_context *old)
1433{
1434	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1435	btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1436	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1437	btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1438	btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1439	btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1440	btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1441	btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1442	btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1443	btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1444	btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1445	btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1446	btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1447	btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1448	btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1449	btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1450	btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1451	btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1452	btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1453	btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1454	btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1455	btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1456	btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1457
1458	btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1459	btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1460	btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1461	btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1462	btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1463	btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1464	btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1465	btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1466	btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1467
1468	/* Did the compression settings change? */
1469	if (btrfs_test_opt(info, COMPRESS) &&
1470	    (!old ||
1471	     old->compress_type != info->compress_type ||
1472	     old->compress_level != info->compress_level ||
1473	     (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1474	      btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1475		const char *compress_type = btrfs_compress_type2str(info->compress_type);
1476
1477		btrfs_info(info, "%s %s compression, level %d",
1478			   btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1479			   compress_type, info->compress_level);
1480	}
1481
1482	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1483		btrfs_info(info, "max_inline set to %llu", info->max_inline);
1484}
1485
1486static int btrfs_reconfigure(struct fs_context *fc)
1487{
1488	struct super_block *sb = fc->root->d_sb;
1489	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1490	struct btrfs_fs_context *ctx = fc->fs_private;
1491	struct btrfs_fs_context old_ctx;
1492	int ret = 0;
1493	bool mount_reconfigure = (fc->s_fs_info != NULL);
1494
1495	btrfs_info_to_ctx(fs_info, &old_ctx);
1496
1497	/*
1498	 * This is our "bind mount" trick, we don't want to allow the user to do
1499	 * anything other than mount a different ro/rw and a different subvol,
1500	 * all of the mount options should be maintained.
1501	 */
1502	if (mount_reconfigure)
1503		ctx->mount_opt = old_ctx.mount_opt;
1504
1505	sync_filesystem(sb);
1506	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1507
1508	if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
 
1509		return -EINVAL;
1510
1511	ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1512	if (ret < 0)
1513		return ret;
1514
1515	btrfs_ctx_to_info(fs_info, ctx);
1516	btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1517	btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1518				 old_ctx.thread_pool_size);
1519
1520	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1521	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1522	    (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1523		btrfs_warn(fs_info,
1524		"remount supports changing free space tree only from RO to RW");
1525		/* Make sure free space cache options match the state on disk. */
1526		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1527			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1528			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1529		}
1530		if (btrfs_free_space_cache_v1_active(fs_info)) {
1531			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1532			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1533		}
1534	}
1535
1536	ret = 0;
1537	if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1538		ret = btrfs_remount_ro(fs_info);
1539	else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1540		ret = btrfs_remount_rw(fs_info);
1541	if (ret)
1542		goto restore;
1543
1544	/*
1545	 * If we set the mask during the parameter parsing VFS would reject the
1546	 * remount.  Here we can set the mask and the value will be updated
1547	 * appropriately.
1548	 */
1549	if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1550		fc->sb_flags_mask |= SB_POSIXACL;
1551
1552	btrfs_emit_options(fs_info, &old_ctx);
1553	wake_up_process(fs_info->transaction_kthread);
1554	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1555	btrfs_clear_oneshot_options(fs_info);
1556	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1557
1558	return 0;
1559restore:
1560	btrfs_ctx_to_info(fs_info, &old_ctx);
1561	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1562	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1563	return ret;
1564}
1565
1566/* Used to sort the devices by max_avail(descending sort) */
1567static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1568{
1569	const struct btrfs_device_info *dev_info1 = a;
1570	const struct btrfs_device_info *dev_info2 = b;
1571
1572	if (dev_info1->max_avail > dev_info2->max_avail)
1573		return -1;
1574	else if (dev_info1->max_avail < dev_info2->max_avail)
1575		return 1;
1576	return 0;
1577}
1578
1579/*
1580 * sort the devices by max_avail, in which max free extent size of each device
1581 * is stored.(Descending Sort)
1582 */
1583static inline void btrfs_descending_sort_devices(
1584					struct btrfs_device_info *devices,
1585					size_t nr_devices)
1586{
1587	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1588	     btrfs_cmp_device_free_bytes, NULL);
1589}
1590
1591/*
1592 * The helper to calc the free space on the devices that can be used to store
1593 * file data.
1594 */
1595static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1596					      u64 *free_bytes)
1597{
1598	struct btrfs_device_info *devices_info;
1599	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1600	struct btrfs_device *device;
1601	u64 type;
1602	u64 avail_space;
1603	u64 min_stripe_size;
1604	int num_stripes = 1;
1605	int i = 0, nr_devices;
1606	const struct btrfs_raid_attr *rattr;
1607
1608	/*
1609	 * We aren't under the device list lock, so this is racy-ish, but good
1610	 * enough for our purposes.
1611	 */
1612	nr_devices = fs_info->fs_devices->open_devices;
1613	if (!nr_devices) {
1614		smp_mb();
1615		nr_devices = fs_info->fs_devices->open_devices;
1616		ASSERT(nr_devices);
1617		if (!nr_devices) {
1618			*free_bytes = 0;
1619			return 0;
1620		}
1621	}
1622
1623	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1624			       GFP_KERNEL);
1625	if (!devices_info)
1626		return -ENOMEM;
1627
1628	/* calc min stripe number for data space allocation */
1629	type = btrfs_data_alloc_profile(fs_info);
1630	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1631
1632	if (type & BTRFS_BLOCK_GROUP_RAID0)
1633		num_stripes = nr_devices;
1634	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1635		num_stripes = rattr->ncopies;
1636	else if (type & BTRFS_BLOCK_GROUP_RAID10)
1637		num_stripes = 4;
1638
1639	/* Adjust for more than 1 stripe per device */
1640	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1641
1642	rcu_read_lock();
1643	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1644		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1645						&device->dev_state) ||
1646		    !device->bdev ||
1647		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1648			continue;
1649
1650		if (i >= nr_devices)
1651			break;
1652
1653		avail_space = device->total_bytes - device->bytes_used;
1654
1655		/* align with stripe_len */
1656		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1657
1658		/*
1659		 * Ensure we have at least min_stripe_size on top of the
1660		 * reserved space on the device.
1661		 */
1662		if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1663			continue;
1664
1665		avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1666
1667		devices_info[i].dev = device;
1668		devices_info[i].max_avail = avail_space;
1669
1670		i++;
1671	}
1672	rcu_read_unlock();
1673
1674	nr_devices = i;
1675
1676	btrfs_descending_sort_devices(devices_info, nr_devices);
1677
1678	i = nr_devices - 1;
1679	avail_space = 0;
1680	while (nr_devices >= rattr->devs_min) {
1681		num_stripes = min(num_stripes, nr_devices);
1682
1683		if (devices_info[i].max_avail >= min_stripe_size) {
1684			int j;
1685			u64 alloc_size;
1686
1687			avail_space += devices_info[i].max_avail * num_stripes;
1688			alloc_size = devices_info[i].max_avail;
1689			for (j = i + 1 - num_stripes; j <= i; j++)
1690				devices_info[j].max_avail -= alloc_size;
1691		}
1692		i--;
1693		nr_devices--;
1694	}
1695
1696	kfree(devices_info);
1697	*free_bytes = avail_space;
1698	return 0;
1699}
1700
1701/*
1702 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1703 *
1704 * If there's a redundant raid level at DATA block groups, use the respective
1705 * multiplier to scale the sizes.
1706 *
1707 * Unused device space usage is based on simulating the chunk allocator
1708 * algorithm that respects the device sizes and order of allocations.  This is
1709 * a close approximation of the actual use but there are other factors that may
1710 * change the result (like a new metadata chunk).
1711 *
1712 * If metadata is exhausted, f_bavail will be 0.
1713 */
1714static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1715{
1716	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1717	struct btrfs_super_block *disk_super = fs_info->super_copy;
1718	struct btrfs_space_info *found;
1719	u64 total_used = 0;
1720	u64 total_free_data = 0;
1721	u64 total_free_meta = 0;
1722	u32 bits = fs_info->sectorsize_bits;
1723	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1724	unsigned factor = 1;
1725	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1726	int ret;
1727	u64 thresh = 0;
1728	int mixed = 0;
1729
1730	list_for_each_entry(found, &fs_info->space_info, list) {
1731		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1732			int i;
1733
1734			total_free_data += found->disk_total - found->disk_used;
1735			total_free_data -=
1736				btrfs_account_ro_block_groups_free_space(found);
1737
1738			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1739				if (!list_empty(&found->block_groups[i]))
1740					factor = btrfs_bg_type_to_factor(
1741						btrfs_raid_array[i].bg_flag);
1742			}
1743		}
1744
1745		/*
1746		 * Metadata in mixed block group profiles are accounted in data
1747		 */
1748		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1749			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1750				mixed = 1;
1751			else
1752				total_free_meta += found->disk_total -
1753					found->disk_used;
1754		}
1755
1756		total_used += found->disk_used;
1757	}
1758
1759	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1760	buf->f_blocks >>= bits;
1761	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1762
1763	/* Account global block reserve as used, it's in logical size already */
1764	spin_lock(&block_rsv->lock);
1765	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
1766	if (buf->f_bfree >= block_rsv->size >> bits)
1767		buf->f_bfree -= block_rsv->size >> bits;
1768	else
1769		buf->f_bfree = 0;
1770	spin_unlock(&block_rsv->lock);
1771
1772	buf->f_bavail = div_u64(total_free_data, factor);
1773	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1774	if (ret)
1775		return ret;
1776	buf->f_bavail += div_u64(total_free_data, factor);
1777	buf->f_bavail = buf->f_bavail >> bits;
1778
1779	/*
1780	 * We calculate the remaining metadata space minus global reserve. If
1781	 * this is (supposedly) smaller than zero, there's no space. But this
1782	 * does not hold in practice, the exhausted state happens where's still
1783	 * some positive delta. So we apply some guesswork and compare the
1784	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
1785	 *
1786	 * We probably cannot calculate the exact threshold value because this
1787	 * depends on the internal reservations requested by various
1788	 * operations, so some operations that consume a few metadata will
1789	 * succeed even if the Avail is zero. But this is better than the other
1790	 * way around.
1791	 */
1792	thresh = SZ_4M;
1793
1794	/*
1795	 * We only want to claim there's no available space if we can no longer
1796	 * allocate chunks for our metadata profile and our global reserve will
1797	 * not fit in the free metadata space.  If we aren't ->full then we
1798	 * still can allocate chunks and thus are fine using the currently
1799	 * calculated f_bavail.
1800	 */
1801	if (!mixed && block_rsv->space_info->full &&
1802	    (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1803		buf->f_bavail = 0;
1804
1805	buf->f_type = BTRFS_SUPER_MAGIC;
1806	buf->f_bsize = fs_info->sectorsize;
1807	buf->f_namelen = BTRFS_NAME_LEN;
1808
1809	/* We treat it as constant endianness (it doesn't matter _which_)
1810	   because we want the fsid to come out the same whether mounted
1811	   on a big-endian or little-endian host */
1812	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1813	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1814	/* Mask in the root object ID too, to disambiguate subvols */
1815	buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1816	buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
 
 
1817
1818	return 0;
1819}
1820
1821static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1822{
1823	struct btrfs_fs_info *p = fc->s_fs_info;
1824	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1825
1826	return fs_info->fs_devices == p->fs_devices;
1827}
1828
1829static int btrfs_get_tree_super(struct fs_context *fc)
1830{
1831	struct btrfs_fs_info *fs_info = fc->s_fs_info;
1832	struct btrfs_fs_context *ctx = fc->fs_private;
1833	struct btrfs_fs_devices *fs_devices = NULL;
1834	struct block_device *bdev;
1835	struct btrfs_device *device;
1836	struct super_block *sb;
1837	blk_mode_t mode = btrfs_open_mode(fc);
1838	int ret;
1839
1840	btrfs_ctx_to_info(fs_info, ctx);
1841	mutex_lock(&uuid_mutex);
1842
1843	/*
1844	 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1845	 * either a valid device or an error.
1846	 */
1847	device = btrfs_scan_one_device(fc->source, mode, true);
1848	ASSERT(device != NULL);
1849	if (IS_ERR(device)) {
1850		mutex_unlock(&uuid_mutex);
1851		return PTR_ERR(device);
1852	}
1853
1854	fs_devices = device->fs_devices;
1855	fs_info->fs_devices = fs_devices;
1856
1857	ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1858	mutex_unlock(&uuid_mutex);
1859	if (ret)
1860		return ret;
1861
1862	if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1863		ret = -EACCES;
1864		goto error;
1865	}
1866
1867	bdev = fs_devices->latest_dev->bdev;
1868
1869	/*
1870	 * From now on the error handling is not straightforward.
1871	 *
1872	 * If successful, this will transfer the fs_info into the super block,
1873	 * and fc->s_fs_info will be NULL.  However if there's an existing
1874	 * super, we'll still have fc->s_fs_info populated.  If we error
1875	 * completely out it'll be cleaned up when we drop the fs_context,
1876	 * otherwise it's tied to the lifetime of the super_block.
1877	 */
1878	sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1879	if (IS_ERR(sb)) {
1880		ret = PTR_ERR(sb);
1881		goto error;
1882	}
1883
1884	set_device_specific_options(fs_info);
1885
1886	if (sb->s_root) {
1887		btrfs_close_devices(fs_devices);
1888		/*
1889		 * At this stage we may have RO flag mismatch between
1890		 * fc->sb_flags and sb->s_flags.  Caller should detect such
1891		 * mismatch and reconfigure with sb->s_umount rwsem held if
1892		 * needed.
1893		 */
1894	} else {
1895		snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1896		shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1897		btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1898		ret = btrfs_fill_super(sb, fs_devices);
1899		if (ret) {
1900			deactivate_locked_super(sb);
1901			return ret;
1902		}
 
1903	}
1904
1905	btrfs_clear_oneshot_options(fs_info);
1906
1907	fc->root = dget(sb->s_root);
1908	return 0;
1909
1910error:
1911	btrfs_close_devices(fs_devices);
1912	return ret;
1913}
1914
1915/*
1916 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1917 * with different ro/rw options") the following works:
1918 *
1919 *        (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1920 *       (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1921 *
1922 * which looks nice and innocent but is actually pretty intricate and deserves
1923 * a long comment.
1924 *
1925 * On another filesystem a subvolume mount is close to something like:
1926 *
1927 *	(iii) # create rw superblock + initial mount
1928 *	      mount -t xfs /dev/sdb /opt/
1929 *
1930 *	      # create ro bind mount
1931 *	      mount --bind -o ro /opt/foo /mnt/foo
1932 *
1933 *	      # unmount initial mount
1934 *	      umount /opt
1935 *
1936 * Of course, there's some special subvolume sauce and there's the fact that the
1937 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1938 * it's very close and will help us understand the issue.
1939 *
1940 * The old mount API didn't cleanly distinguish between a mount being made ro
1941 * and a superblock being made ro.  The only way to change the ro state of
1942 * either object was by passing ms_rdonly. If a new mount was created via
1943 * mount(2) such as:
1944 *
1945 *      mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1946 *
1947 * the MS_RDONLY flag being specified had two effects:
1948 *
1949 * (1) MNT_READONLY was raised -> the resulting mount got
1950 *     @mnt->mnt_flags |= MNT_READONLY raised.
1951 *
1952 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1953 *     made the superblock ro. Note, how SB_RDONLY has the same value as
1954 *     ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1955 *
1956 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1957 * subtree mounted ro.
1958 *
1959 * But consider the effect on the old mount API on btrfs subvolume mounting
1960 * which combines the distinct step in (iii) into a single step.
1961 *
1962 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1963 * is issued the superblock is ro and thus even if the mount created for (ii) is
1964 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1965 * to rw for (ii) which it did using an internal remount call.
1966 *
1967 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1968 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1969 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1970 * passed by mount(8) to mount(2).
1971 *
1972 * Enter the new mount API. The new mount API disambiguates making a mount ro
1973 * and making a superblock ro.
1974 *
1975 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1976 *     fsmount() or mount_setattr() this is a pure VFS level change for a
1977 *     specific mount or mount tree that is never seen by the filesystem itself.
1978 *
1979 * (4) To turn a superblock ro the "ro" flag must be used with
1980 *     fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1981 *     in fc->sb_flags.
1982 *
1983 * But, currently the util-linux mount command already utilizes the new mount
1984 * API and is still setting fsconfig(FSCONFIG_SET_FLAG, "ro") no matter if it's
1985 * btrfs or not, setting the whole super block RO.  To make per-subvolume mounting
1986 * work with different options work we need to keep backward compatibility.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1987 */
1988static int btrfs_reconfigure_for_mount(struct fs_context *fc, struct vfsmount *mnt)
1989{
1990	int ret = 0;
 
 
1991
1992	if (fc->sb_flags & SB_RDONLY)
1993		return ret;
 
 
 
 
 
 
 
1994
 
 
 
 
 
 
 
 
 
1995	down_write(&mnt->mnt_sb->s_umount);
1996	if (!(fc->sb_flags & SB_RDONLY) && (mnt->mnt_sb->s_flags & SB_RDONLY))
1997		ret = btrfs_reconfigure(fc);
1998	up_write(&mnt->mnt_sb->s_umount);
1999	return ret;
 
 
 
 
2000}
2001
2002static int btrfs_get_tree_subvol(struct fs_context *fc)
2003{
2004	struct btrfs_fs_info *fs_info = NULL;
2005	struct btrfs_fs_context *ctx = fc->fs_private;
2006	struct fs_context *dup_fc;
2007	struct dentry *dentry;
2008	struct vfsmount *mnt;
2009	int ret = 0;
2010
2011	/*
2012	 * Setup a dummy root and fs_info for test/set super.  This is because
2013	 * we don't actually fill this stuff out until open_ctree, but we need
2014	 * then open_ctree will properly initialize the file system specific
2015	 * settings later.  btrfs_init_fs_info initializes the static elements
2016	 * of the fs_info (locks and such) to make cleanup easier if we find a
2017	 * superblock with our given fs_devices later on at sget() time.
2018	 */
2019	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2020	if (!fs_info)
2021		return -ENOMEM;
2022
2023	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2024	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2025	if (!fs_info->super_copy || !fs_info->super_for_commit) {
2026		btrfs_free_fs_info(fs_info);
2027		return -ENOMEM;
2028	}
2029	btrfs_init_fs_info(fs_info);
2030
2031	dup_fc = vfs_dup_fs_context(fc);
2032	if (IS_ERR(dup_fc)) {
2033		btrfs_free_fs_info(fs_info);
2034		return PTR_ERR(dup_fc);
2035	}
2036
2037	/*
2038	 * When we do the sget_fc this gets transferred to the sb, so we only
2039	 * need to set it on the dup_fc as this is what creates the super block.
2040	 */
2041	dup_fc->s_fs_info = fs_info;
2042
2043	/*
2044	 * We'll do the security settings in our btrfs_get_tree_super() mount
2045	 * loop, they were duplicated into dup_fc, we can drop the originals
2046	 * here.
2047	 */
2048	security_free_mnt_opts(&fc->security);
2049	fc->security = NULL;
2050
2051	mnt = fc_mount(dup_fc);
2052	if (IS_ERR(mnt)) {
2053		put_fs_context(dup_fc);
2054		return PTR_ERR(mnt);
2055	}
2056	ret = btrfs_reconfigure_for_mount(dup_fc, mnt);
2057	put_fs_context(dup_fc);
2058	if (ret) {
2059		mntput(mnt);
2060		return ret;
2061	}
2062
2063	/*
2064	 * This free's ->subvol_name, because if it isn't set we have to
2065	 * allocate a buffer to hold the subvol_name, so we just drop our
2066	 * reference to it here.
2067	 */
2068	dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2069	ctx->subvol_name = NULL;
2070	if (IS_ERR(dentry))
2071		return PTR_ERR(dentry);
2072
2073	fc->root = dentry;
2074	return 0;
2075}
2076
2077static int btrfs_get_tree(struct fs_context *fc)
2078{
2079	/*
2080	 * Since we use mount_subtree to mount the default/specified subvol, we
2081	 * have to do mounts in two steps.
2082	 *
2083	 * First pass through we call btrfs_get_tree_subvol(), this is just a
2084	 * wrapper around fc_mount() to call back into here again, and this time
2085	 * we'll call btrfs_get_tree_super().  This will do the open_ctree() and
2086	 * everything to open the devices and file system.  Then we return back
2087	 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2088	 * from there we can do our mount_subvol() call, which will lookup
2089	 * whichever subvol we're mounting and setup this fc with the
2090	 * appropriate dentry for the subvol.
2091	 */
2092	if (fc->s_fs_info)
2093		return btrfs_get_tree_super(fc);
2094	return btrfs_get_tree_subvol(fc);
2095}
2096
2097static void btrfs_kill_super(struct super_block *sb)
2098{
2099	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2100	kill_anon_super(sb);
2101	btrfs_free_fs_info(fs_info);
2102}
2103
2104static void btrfs_free_fs_context(struct fs_context *fc)
2105{
2106	struct btrfs_fs_context *ctx = fc->fs_private;
2107	struct btrfs_fs_info *fs_info = fc->s_fs_info;
2108
2109	if (fs_info)
2110		btrfs_free_fs_info(fs_info);
2111
2112	if (ctx && refcount_dec_and_test(&ctx->refs)) {
2113		kfree(ctx->subvol_name);
2114		kfree(ctx);
2115	}
2116}
2117
2118static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2119{
2120	struct btrfs_fs_context *ctx = src_fc->fs_private;
2121
2122	/*
2123	 * Give a ref to our ctx to this dup, as we want to keep it around for
2124	 * our original fc so we can have the subvolume name or objectid.
2125	 *
2126	 * We unset ->source in the original fc because the dup needs it for
2127	 * mounting, and then once we free the dup it'll free ->source, so we
2128	 * need to make sure we're only pointing to it in one fc.
2129	 */
2130	refcount_inc(&ctx->refs);
2131	fc->fs_private = ctx;
2132	fc->source = src_fc->source;
2133	src_fc->source = NULL;
2134	return 0;
2135}
2136
2137static const struct fs_context_operations btrfs_fs_context_ops = {
2138	.parse_param	= btrfs_parse_param,
2139	.reconfigure	= btrfs_reconfigure,
2140	.get_tree	= btrfs_get_tree,
2141	.dup		= btrfs_dup_fs_context,
2142	.free		= btrfs_free_fs_context,
2143};
2144
2145static int btrfs_init_fs_context(struct fs_context *fc)
2146{
2147	struct btrfs_fs_context *ctx;
2148
2149	ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2150	if (!ctx)
2151		return -ENOMEM;
2152
2153	refcount_set(&ctx->refs, 1);
2154	fc->fs_private = ctx;
2155	fc->ops = &btrfs_fs_context_ops;
2156
2157	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2158		btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2159	} else {
2160		ctx->thread_pool_size =
2161			min_t(unsigned long, num_online_cpus() + 2, 8);
2162		ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2163		ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2164	}
2165
2166#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2167	fc->sb_flags |= SB_POSIXACL;
2168#endif
2169	fc->sb_flags |= SB_I_VERSION;
2170
2171	return 0;
2172}
2173
2174static struct file_system_type btrfs_fs_type = {
2175	.owner			= THIS_MODULE,
2176	.name			= "btrfs",
2177	.init_fs_context	= btrfs_init_fs_context,
2178	.parameters		= btrfs_fs_parameters,
2179	.kill_sb		= btrfs_kill_super,
2180	.fs_flags		= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA |
2181				  FS_ALLOW_IDMAP | FS_MGTIME,
2182 };
2183
2184MODULE_ALIAS_FS("btrfs");
2185
2186static int btrfs_control_open(struct inode *inode, struct file *file)
2187{
2188	/*
2189	 * The control file's private_data is used to hold the
2190	 * transaction when it is started and is used to keep
2191	 * track of whether a transaction is already in progress.
2192	 */
2193	file->private_data = NULL;
2194	return 0;
2195}
2196
2197/*
2198 * Used by /dev/btrfs-control for devices ioctls.
2199 */
2200static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2201				unsigned long arg)
2202{
2203	struct btrfs_ioctl_vol_args *vol;
2204	struct btrfs_device *device = NULL;
2205	dev_t devt = 0;
2206	int ret = -ENOTTY;
2207
2208	if (!capable(CAP_SYS_ADMIN))
2209		return -EPERM;
2210
2211	vol = memdup_user((void __user *)arg, sizeof(*vol));
2212	if (IS_ERR(vol))
2213		return PTR_ERR(vol);
2214	ret = btrfs_check_ioctl_vol_args_path(vol);
2215	if (ret < 0)
2216		goto out;
2217
2218	switch (cmd) {
2219	case BTRFS_IOC_SCAN_DEV:
2220		mutex_lock(&uuid_mutex);
2221		/*
2222		 * Scanning outside of mount can return NULL which would turn
2223		 * into 0 error code.
2224		 */
2225		device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2226		ret = PTR_ERR_OR_ZERO(device);
2227		mutex_unlock(&uuid_mutex);
2228		break;
2229	case BTRFS_IOC_FORGET_DEV:
2230		if (vol->name[0] != 0) {
2231			ret = lookup_bdev(vol->name, &devt);
2232			if (ret)
2233				break;
2234		}
2235		ret = btrfs_forget_devices(devt);
2236		break;
2237	case BTRFS_IOC_DEVICES_READY:
2238		mutex_lock(&uuid_mutex);
2239		/*
2240		 * Scanning outside of mount can return NULL which would turn
2241		 * into 0 error code.
2242		 */
2243		device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2244		if (IS_ERR_OR_NULL(device)) {
2245			mutex_unlock(&uuid_mutex);
2246			if (IS_ERR(device))
2247				ret = PTR_ERR(device);
2248			else
2249				ret = 0;
2250			break;
2251		}
2252		ret = !(device->fs_devices->num_devices ==
2253			device->fs_devices->total_devices);
2254		mutex_unlock(&uuid_mutex);
2255		break;
2256	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2257		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2258		break;
2259	}
2260
2261out:
2262	kfree(vol);
2263	return ret;
2264}
2265
2266static int btrfs_freeze(struct super_block *sb)
2267{
 
2268	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 
2269
2270	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2271	/*
2272	 * We don't need a barrier here, we'll wait for any transaction that
2273	 * could be in progress on other threads (and do delayed iputs that
2274	 * we want to avoid on a frozen filesystem), or do the commit
2275	 * ourselves.
2276	 */
2277	return btrfs_commit_current_transaction(fs_info->tree_root);
 
 
 
 
 
 
 
2278}
2279
2280static int check_dev_super(struct btrfs_device *dev)
2281{
2282	struct btrfs_fs_info *fs_info = dev->fs_info;
2283	struct btrfs_super_block *sb;
2284	u64 last_trans;
2285	u16 csum_type;
2286	int ret = 0;
2287
2288	/* This should be called with fs still frozen. */
2289	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2290
2291	/* Missing dev, no need to check. */
2292	if (!dev->bdev)
2293		return 0;
2294
2295	/* Only need to check the primary super block. */
2296	sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2297	if (IS_ERR(sb))
2298		return PTR_ERR(sb);
2299
2300	/* Verify the checksum. */
2301	csum_type = btrfs_super_csum_type(sb);
2302	if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2303		btrfs_err(fs_info, "csum type changed, has %u expect %u",
2304			  csum_type, btrfs_super_csum_type(fs_info->super_copy));
2305		ret = -EUCLEAN;
2306		goto out;
2307	}
2308
2309	if (btrfs_check_super_csum(fs_info, sb)) {
2310		btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2311		ret = -EUCLEAN;
2312		goto out;
2313	}
2314
2315	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2316	ret = btrfs_validate_super(fs_info, sb, 0);
2317	if (ret < 0)
2318		goto out;
2319
2320	last_trans = btrfs_get_last_trans_committed(fs_info);
2321	if (btrfs_super_generation(sb) != last_trans) {
2322		btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2323			  btrfs_super_generation(sb), last_trans);
2324		ret = -EUCLEAN;
2325		goto out;
2326	}
2327out:
2328	btrfs_release_disk_super(sb);
2329	return ret;
2330}
2331
2332static int btrfs_unfreeze(struct super_block *sb)
2333{
2334	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2335	struct btrfs_device *device;
2336	int ret = 0;
2337
2338	/*
2339	 * Make sure the fs is not changed by accident (like hibernation then
2340	 * modified by other OS).
2341	 * If we found anything wrong, we mark the fs error immediately.
2342	 *
2343	 * And since the fs is frozen, no one can modify the fs yet, thus
2344	 * we don't need to hold device_list_mutex.
2345	 */
2346	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2347		ret = check_dev_super(device);
2348		if (ret < 0) {
2349			btrfs_handle_fs_error(fs_info, ret,
2350				"super block on devid %llu got modified unexpectedly",
2351				device->devid);
2352			break;
2353		}
2354	}
2355	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2356
2357	/*
2358	 * We still return 0, to allow VFS layer to unfreeze the fs even the
2359	 * above checks failed. Since the fs is either fine or read-only, we're
2360	 * safe to continue, without causing further damage.
2361	 */
2362	return 0;
2363}
2364
2365static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2366{
2367	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2368
2369	/*
2370	 * There should be always a valid pointer in latest_dev, it may be stale
2371	 * for a short moment in case it's being deleted but still valid until
2372	 * the end of RCU grace period.
2373	 */
2374	rcu_read_lock();
2375	seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2376	rcu_read_unlock();
2377
2378	return 0;
2379}
2380
2381static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2382{
2383	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2384	const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
2385
2386	trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2387
2388	return nr;
2389}
2390
2391static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)
2392{
2393	const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
2394	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2395
2396	btrfs_free_extent_maps(fs_info, nr_to_scan);
2397
2398	/* The extent map shrinker runs asynchronously, so always return 0. */
2399	return 0;
2400}
2401
2402static const struct super_operations btrfs_super_ops = {
2403	.drop_inode	= btrfs_drop_inode,
2404	.evict_inode	= btrfs_evict_inode,
2405	.put_super	= btrfs_put_super,
2406	.sync_fs	= btrfs_sync_fs,
2407	.show_options	= btrfs_show_options,
2408	.show_devname	= btrfs_show_devname,
2409	.alloc_inode	= btrfs_alloc_inode,
2410	.destroy_inode	= btrfs_destroy_inode,
2411	.free_inode	= btrfs_free_inode,
2412	.statfs		= btrfs_statfs,
2413	.freeze_fs	= btrfs_freeze,
2414	.unfreeze_fs	= btrfs_unfreeze,
2415	.nr_cached_objects = btrfs_nr_cached_objects,
2416	.free_cached_objects = btrfs_free_cached_objects,
2417};
2418
2419static const struct file_operations btrfs_ctl_fops = {
2420	.open = btrfs_control_open,
2421	.unlocked_ioctl	 = btrfs_control_ioctl,
2422	.compat_ioctl = compat_ptr_ioctl,
2423	.owner	 = THIS_MODULE,
2424	.llseek = noop_llseek,
2425};
2426
2427static struct miscdevice btrfs_misc = {
2428	.minor		= BTRFS_MINOR,
2429	.name		= "btrfs-control",
2430	.fops		= &btrfs_ctl_fops
2431};
2432
2433MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2434MODULE_ALIAS("devname:btrfs-control");
2435
2436static int __init btrfs_interface_init(void)
2437{
2438	return misc_register(&btrfs_misc);
2439}
2440
2441static __cold void btrfs_interface_exit(void)
2442{
2443	misc_deregister(&btrfs_misc);
2444}
2445
2446static int __init btrfs_print_mod_info(void)
2447{
2448	static const char options[] = ""
2449#ifdef CONFIG_BTRFS_DEBUG
2450			", debug=on"
2451#endif
2452#ifdef CONFIG_BTRFS_ASSERT
2453			", assert=on"
2454#endif
2455#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2456			", ref-verify=on"
2457#endif
2458#ifdef CONFIG_BLK_DEV_ZONED
2459			", zoned=yes"
2460#else
2461			", zoned=no"
2462#endif
2463#ifdef CONFIG_FS_VERITY
2464			", fsverity=yes"
2465#else
2466			", fsverity=no"
2467#endif
2468			;
2469	pr_info("Btrfs loaded%s\n", options);
2470	return 0;
2471}
2472
2473static int register_btrfs(void)
2474{
2475	return register_filesystem(&btrfs_fs_type);
2476}
2477
2478static void unregister_btrfs(void)
2479{
2480	unregister_filesystem(&btrfs_fs_type);
2481}
2482
2483/* Helper structure for long init/exit functions. */
2484struct init_sequence {
2485	int (*init_func)(void);
2486	/* Can be NULL if the init_func doesn't need cleanup. */
2487	void (*exit_func)(void);
2488};
2489
2490static const struct init_sequence mod_init_seq[] = {
2491	{
2492		.init_func = btrfs_props_init,
2493		.exit_func = NULL,
2494	}, {
2495		.init_func = btrfs_init_sysfs,
2496		.exit_func = btrfs_exit_sysfs,
2497	}, {
2498		.init_func = btrfs_init_compress,
2499		.exit_func = btrfs_exit_compress,
2500	}, {
2501		.init_func = btrfs_init_cachep,
2502		.exit_func = btrfs_destroy_cachep,
2503	}, {
2504		.init_func = btrfs_init_dio,
2505		.exit_func = btrfs_destroy_dio,
2506	}, {
2507		.init_func = btrfs_transaction_init,
2508		.exit_func = btrfs_transaction_exit,
2509	}, {
2510		.init_func = btrfs_ctree_init,
2511		.exit_func = btrfs_ctree_exit,
2512	}, {
2513		.init_func = btrfs_free_space_init,
2514		.exit_func = btrfs_free_space_exit,
2515	}, {
2516		.init_func = extent_state_init_cachep,
2517		.exit_func = extent_state_free_cachep,
2518	}, {
2519		.init_func = extent_buffer_init_cachep,
2520		.exit_func = extent_buffer_free_cachep,
2521	}, {
2522		.init_func = btrfs_bioset_init,
2523		.exit_func = btrfs_bioset_exit,
2524	}, {
2525		.init_func = extent_map_init,
2526		.exit_func = extent_map_exit,
2527	}, {
2528		.init_func = ordered_data_init,
2529		.exit_func = ordered_data_exit,
2530	}, {
2531		.init_func = btrfs_delayed_inode_init,
2532		.exit_func = btrfs_delayed_inode_exit,
2533	}, {
2534		.init_func = btrfs_auto_defrag_init,
2535		.exit_func = btrfs_auto_defrag_exit,
2536	}, {
2537		.init_func = btrfs_delayed_ref_init,
2538		.exit_func = btrfs_delayed_ref_exit,
2539	}, {
2540		.init_func = btrfs_prelim_ref_init,
2541		.exit_func = btrfs_prelim_ref_exit,
2542	}, {
2543		.init_func = btrfs_interface_init,
2544		.exit_func = btrfs_interface_exit,
2545	}, {
2546		.init_func = btrfs_print_mod_info,
2547		.exit_func = NULL,
2548	}, {
2549		.init_func = btrfs_run_sanity_tests,
2550		.exit_func = NULL,
2551	}, {
2552		.init_func = register_btrfs,
2553		.exit_func = unregister_btrfs,
2554	}
2555};
2556
2557static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2558
2559static __always_inline void btrfs_exit_btrfs_fs(void)
2560{
2561	int i;
2562
2563	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2564		if (!mod_init_result[i])
2565			continue;
2566		if (mod_init_seq[i].exit_func)
2567			mod_init_seq[i].exit_func();
2568		mod_init_result[i] = false;
2569	}
2570}
2571
2572static void __exit exit_btrfs_fs(void)
2573{
2574	btrfs_exit_btrfs_fs();
2575	btrfs_cleanup_fs_uuids();
2576}
2577
2578static int __init init_btrfs_fs(void)
2579{
2580	int ret;
2581	int i;
2582
2583	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2584		ASSERT(!mod_init_result[i]);
2585		ret = mod_init_seq[i].init_func();
2586		if (ret < 0) {
2587			btrfs_exit_btrfs_fs();
2588			return ret;
2589		}
2590		mod_init_result[i] = true;
2591	}
2592	return 0;
2593}
2594
2595late_initcall(init_btrfs_fs);
2596module_exit(exit_btrfs_fs)
2597
2598MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2599MODULE_LICENSE("GPL");
2600MODULE_SOFTDEP("pre: crc32c");
2601MODULE_SOFTDEP("pre: xxhash64");
2602MODULE_SOFTDEP("pre: sha256");
2603MODULE_SOFTDEP("pre: blake2b-256");
v6.9.4
   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 "props.h"
  38#include "xattr.h"
  39#include "bio.h"
  40#include "export.h"
  41#include "compression.h"
  42#include "dev-replace.h"
  43#include "free-space-cache.h"
  44#include "backref.h"
  45#include "space-info.h"
  46#include "sysfs.h"
  47#include "zoned.h"
  48#include "tests/btrfs-tests.h"
  49#include "block-group.h"
  50#include "discard.h"
  51#include "qgroup.h"
  52#include "raid56.h"
  53#include "fs.h"
  54#include "accessors.h"
  55#include "defrag.h"
  56#include "dir-item.h"
  57#include "ioctl.h"
  58#include "scrub.h"
  59#include "verity.h"
  60#include "super.h"
  61#include "extent-tree.h"
  62#define CREATE_TRACE_POINTS
  63#include <trace/events/btrfs.h>
  64
  65static const struct super_operations btrfs_super_ops;
  66static struct file_system_type btrfs_fs_type;
  67
  68static void btrfs_put_super(struct super_block *sb)
  69{
  70	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
  71
  72	btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
  73	close_ctree(fs_info);
  74}
  75
  76/* Store the mount options related information. */
  77struct btrfs_fs_context {
  78	char *subvol_name;
  79	u64 subvol_objectid;
  80	u64 max_inline;
  81	u32 commit_interval;
  82	u32 metadata_ratio;
  83	u32 thread_pool_size;
  84	unsigned long mount_opt;
  85	unsigned long compress_type:4;
  86	unsigned int compress_level;
  87	refcount_t refs;
  88};
  89
  90enum {
  91	Opt_acl,
  92	Opt_clear_cache,
  93	Opt_commit_interval,
  94	Opt_compress,
  95	Opt_compress_force,
  96	Opt_compress_force_type,
  97	Opt_compress_type,
  98	Opt_degraded,
  99	Opt_device,
 100	Opt_fatal_errors,
 101	Opt_flushoncommit,
 102	Opt_max_inline,
 103	Opt_barrier,
 104	Opt_datacow,
 105	Opt_datasum,
 106	Opt_defrag,
 107	Opt_discard,
 108	Opt_discard_mode,
 109	Opt_ratio,
 110	Opt_rescan_uuid_tree,
 111	Opt_skip_balance,
 112	Opt_space_cache,
 113	Opt_space_cache_version,
 114	Opt_ssd,
 115	Opt_ssd_spread,
 116	Opt_subvol,
 117	Opt_subvol_empty,
 118	Opt_subvolid,
 119	Opt_thread_pool,
 120	Opt_treelog,
 121	Opt_user_subvol_rm_allowed,
 
 122
 123	/* Rescue options */
 124	Opt_rescue,
 125	Opt_usebackuproot,
 126	Opt_nologreplay,
 127	Opt_ignorebadroots,
 128	Opt_ignoredatacsums,
 129	Opt_rescue_all,
 130
 131	/* Debugging options */
 132	Opt_enospc_debug,
 133#ifdef CONFIG_BTRFS_DEBUG
 134	Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 135#endif
 136#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 137	Opt_ref_verify,
 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_parameter_all,
 181};
 182
 183static const struct constant_table btrfs_parameter_rescue[] = {
 184	{ "usebackuproot", Opt_rescue_usebackuproot },
 185	{ "nologreplay", Opt_rescue_nologreplay },
 186	{ "ignorebadroots", Opt_rescue_ignorebadroots },
 187	{ "ibadroots", Opt_rescue_ignorebadroots },
 188	{ "ignoredatacsums", Opt_rescue_ignoredatacsums },
 
 
 189	{ "idatacsums", Opt_rescue_ignoredatacsums },
 
 
 190	{ "all", Opt_rescue_parameter_all },
 191	{}
 192};
 193
 194#ifdef CONFIG_BTRFS_DEBUG
 195enum {
 196	Opt_fragment_parameter_data,
 197	Opt_fragment_parameter_metadata,
 198	Opt_fragment_parameter_all,
 199};
 200
 201static const struct constant_table btrfs_parameter_fragment[] = {
 202	{ "data", Opt_fragment_parameter_data },
 203	{ "metadata", Opt_fragment_parameter_metadata },
 204	{ "all", Opt_fragment_parameter_all },
 205	{}
 206};
 207#endif
 208
 209static const struct fs_parameter_spec btrfs_fs_parameters[] = {
 210	fsparam_flag_no("acl", Opt_acl),
 211	fsparam_flag_no("autodefrag", Opt_defrag),
 212	fsparam_flag_no("barrier", Opt_barrier),
 213	fsparam_flag("clear_cache", Opt_clear_cache),
 214	fsparam_u32("commit", Opt_commit_interval),
 215	fsparam_flag("compress", Opt_compress),
 216	fsparam_string("compress", Opt_compress_type),
 217	fsparam_flag("compress-force", Opt_compress_force),
 218	fsparam_string("compress-force", Opt_compress_force_type),
 219	fsparam_flag_no("datacow", Opt_datacow),
 220	fsparam_flag_no("datasum", Opt_datasum),
 221	fsparam_flag("degraded", Opt_degraded),
 222	fsparam_string("device", Opt_device),
 223	fsparam_flag_no("discard", Opt_discard),
 224	fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
 225	fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
 226	fsparam_flag_no("flushoncommit", Opt_flushoncommit),
 227	fsparam_string("max_inline", Opt_max_inline),
 228	fsparam_u32("metadata_ratio", Opt_ratio),
 229	fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
 230	fsparam_flag("skip_balance", Opt_skip_balance),
 231	fsparam_flag_no("space_cache", Opt_space_cache),
 232	fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
 233	fsparam_flag_no("ssd", Opt_ssd),
 234	fsparam_flag_no("ssd_spread", Opt_ssd_spread),
 235	fsparam_string("subvol", Opt_subvol),
 236	fsparam_flag("subvol=", Opt_subvol_empty),
 237	fsparam_u64("subvolid", Opt_subvolid),
 238	fsparam_u32("thread_pool", Opt_thread_pool),
 239	fsparam_flag_no("treelog", Opt_treelog),
 240	fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
 241
 242	/* Rescue options. */
 243	fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
 244	/* Deprecated, with alias rescue=nologreplay */
 245	__fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
 246	/* Deprecated, with alias rescue=usebackuproot */
 247	__fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
 
 
 248
 249	/* Debugging options. */
 250	fsparam_flag_no("enospc_debug", Opt_enospc_debug),
 251#ifdef CONFIG_BTRFS_DEBUG
 252	fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
 253#endif
 254#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 255	fsparam_flag("ref_verify", Opt_ref_verify),
 256#endif
 257	{}
 258};
 259
 260/* No support for restricting writes to btrfs devices yet... */
 261static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
 262{
 263	return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
 264}
 265
 266static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
 267{
 268	struct btrfs_fs_context *ctx = fc->fs_private;
 269	struct fs_parse_result result;
 270	int opt;
 271
 272	opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
 273	if (opt < 0)
 274		return opt;
 275
 276	switch (opt) {
 277	case Opt_degraded:
 278		btrfs_set_opt(ctx->mount_opt, DEGRADED);
 279		break;
 280	case Opt_subvol_empty:
 281		/*
 282		 * This exists because we used to allow it on accident, so we're
 283		 * keeping it to maintain ABI.  See 37becec95ac3 ("Btrfs: allow
 284		 * empty subvol= again").
 285		 */
 286		break;
 287	case Opt_subvol:
 288		kfree(ctx->subvol_name);
 289		ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
 290		if (!ctx->subvol_name)
 291			return -ENOMEM;
 292		break;
 293	case Opt_subvolid:
 294		ctx->subvol_objectid = result.uint_64;
 295
 296		/* subvolid=0 means give me the original fs_tree. */
 297		if (!ctx->subvol_objectid)
 298			ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
 299		break;
 300	case Opt_device: {
 301		struct btrfs_device *device;
 302		blk_mode_t mode = btrfs_open_mode(fc);
 303
 304		mutex_lock(&uuid_mutex);
 305		device = btrfs_scan_one_device(param->string, mode, false);
 306		mutex_unlock(&uuid_mutex);
 307		if (IS_ERR(device))
 308			return PTR_ERR(device);
 309		break;
 310	}
 311	case Opt_datasum:
 312		if (result.negated) {
 313			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 314		} else {
 315			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 316			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 317		}
 318		break;
 319	case Opt_datacow:
 320		if (result.negated) {
 321			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 322			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 323			btrfs_set_opt(ctx->mount_opt, NODATACOW);
 324			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 325		} else {
 326			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 327		}
 328		break;
 329	case Opt_compress_force:
 330	case Opt_compress_force_type:
 331		btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
 332		fallthrough;
 333	case Opt_compress:
 334	case Opt_compress_type:
 
 
 
 
 
 
 
 
 
 335		if (opt == Opt_compress || opt == Opt_compress_force) {
 336			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 337			ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 338			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 339			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 340			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 341		} else if (strncmp(param->string, "zlib", 4) == 0) {
 342			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 343			ctx->compress_level =
 344				btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
 345							 param->string + 4);
 346			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 347			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 348			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 349		} else if (strncmp(param->string, "lzo", 3) == 0) {
 350			ctx->compress_type = BTRFS_COMPRESS_LZO;
 351			ctx->compress_level = 0;
 352			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 353			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 354			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 355		} else if (strncmp(param->string, "zstd", 4) == 0) {
 356			ctx->compress_type = BTRFS_COMPRESS_ZSTD;
 357			ctx->compress_level =
 358				btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
 359							 param->string + 4);
 360			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 361			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 362			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 363		} else if (strncmp(param->string, "no", 2) == 0) {
 364			ctx->compress_level = 0;
 365			ctx->compress_type = 0;
 366			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 367			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 368		} else {
 369			btrfs_err(NULL, "unrecognized compression value %s",
 370				  param->string);
 371			return -EINVAL;
 372		}
 373		break;
 374	case Opt_ssd:
 375		if (result.negated) {
 376			btrfs_set_opt(ctx->mount_opt, NOSSD);
 377			btrfs_clear_opt(ctx->mount_opt, SSD);
 378			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 379		} else {
 380			btrfs_set_opt(ctx->mount_opt, SSD);
 381			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 382		}
 383		break;
 384	case Opt_ssd_spread:
 385		if (result.negated) {
 386			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 387		} else {
 388			btrfs_set_opt(ctx->mount_opt, SSD);
 389			btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
 390			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 391		}
 392		break;
 393	case Opt_barrier:
 394		if (result.negated)
 395			btrfs_set_opt(ctx->mount_opt, NOBARRIER);
 396		else
 397			btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
 398		break;
 399	case Opt_thread_pool:
 400		if (result.uint_32 == 0) {
 401			btrfs_err(NULL, "invalid value 0 for thread_pool");
 402			return -EINVAL;
 403		}
 404		ctx->thread_pool_size = result.uint_32;
 405		break;
 406	case Opt_max_inline:
 407		ctx->max_inline = memparse(param->string, NULL);
 408		break;
 409	case Opt_acl:
 410		if (result.negated) {
 411			fc->sb_flags &= ~SB_POSIXACL;
 412		} else {
 413#ifdef CONFIG_BTRFS_FS_POSIX_ACL
 414			fc->sb_flags |= SB_POSIXACL;
 415#else
 416			btrfs_err(NULL, "support for ACL not compiled in");
 417			return -EINVAL;
 418#endif
 419		}
 420		/*
 421		 * VFS limits the ability to toggle ACL on and off via remount,
 422		 * despite every file system allowing this.  This seems to be
 423		 * an oversight since we all do, but it'll fail if we're
 424		 * remounting.  So don't set the mask here, we'll check it in
 425		 * btrfs_reconfigure and do the toggling ourselves.
 426		 */
 427		if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
 428			fc->sb_flags_mask |= SB_POSIXACL;
 429		break;
 430	case Opt_treelog:
 431		if (result.negated)
 432			btrfs_set_opt(ctx->mount_opt, NOTREELOG);
 433		else
 434			btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
 435		break;
 436	case Opt_nologreplay:
 437		btrfs_warn(NULL,
 438		"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 439		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 440		break;
 
 
 
 
 
 441	case Opt_flushoncommit:
 442		if (result.negated)
 443			btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
 444		else
 445			btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
 446		break;
 447	case Opt_ratio:
 448		ctx->metadata_ratio = result.uint_32;
 449		break;
 450	case Opt_discard:
 451		if (result.negated) {
 452			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 453			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 454			btrfs_set_opt(ctx->mount_opt, NODISCARD);
 455		} else {
 456			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 457			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 458		}
 459		break;
 460	case Opt_discard_mode:
 461		switch (result.uint_32) {
 462		case Opt_discard_sync:
 463			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 464			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 465			break;
 466		case Opt_discard_async:
 467			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 468			btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
 469			break;
 470		default:
 471			btrfs_err(NULL, "unrecognized discard mode value %s",
 472				  param->key);
 473			return -EINVAL;
 474		}
 475		btrfs_clear_opt(ctx->mount_opt, NODISCARD);
 476		break;
 477	case Opt_space_cache:
 478		if (result.negated) {
 479			btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
 480			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 481			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 482		} else {
 483			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 484			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 485		}
 486		break;
 487	case Opt_space_cache_version:
 488		switch (result.uint_32) {
 489		case Opt_space_cache_v1:
 490			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 491			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 492			break;
 493		case Opt_space_cache_v2:
 494			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 495			btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
 496			break;
 497		default:
 498			btrfs_err(NULL, "unrecognized space_cache value %s",
 499				  param->key);
 500			return -EINVAL;
 501		}
 502		break;
 503	case Opt_rescan_uuid_tree:
 504		btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
 505		break;
 506	case Opt_clear_cache:
 507		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 508		break;
 509	case Opt_user_subvol_rm_allowed:
 510		btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
 511		break;
 512	case Opt_enospc_debug:
 513		if (result.negated)
 514			btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
 515		else
 516			btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
 517		break;
 518	case Opt_defrag:
 519		if (result.negated)
 520			btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
 521		else
 522			btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
 523		break;
 524	case Opt_usebackuproot:
 525		btrfs_warn(NULL,
 526			   "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
 527		btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 528
 529		/* If we're loading the backup roots we can't trust the space cache. */
 530		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 531		break;
 532	case Opt_skip_balance:
 533		btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
 534		break;
 535	case Opt_fatal_errors:
 536		switch (result.uint_32) {
 537		case Opt_fatal_errors_panic:
 538			btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 539			break;
 540		case Opt_fatal_errors_bug:
 541			btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 542			break;
 543		default:
 544			btrfs_err(NULL, "unrecognized fatal_errors value %s",
 545				  param->key);
 546			return -EINVAL;
 547		}
 548		break;
 549	case Opt_commit_interval:
 550		ctx->commit_interval = result.uint_32;
 551		if (ctx->commit_interval == 0)
 552			ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 553		break;
 554	case Opt_rescue:
 555		switch (result.uint_32) {
 556		case Opt_rescue_usebackuproot:
 557			btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 558			break;
 559		case Opt_rescue_nologreplay:
 560			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 561			break;
 562		case Opt_rescue_ignorebadroots:
 563			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 564			break;
 565		case Opt_rescue_ignoredatacsums:
 566			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 567			break;
 
 
 
 
 
 
 568		case Opt_rescue_parameter_all:
 569			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 
 
 570			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 571			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 572			break;
 573		default:
 574			btrfs_info(NULL, "unrecognized rescue option '%s'",
 575				   param->key);
 576			return -EINVAL;
 577		}
 578		break;
 579#ifdef CONFIG_BTRFS_DEBUG
 580	case Opt_fragment:
 581		switch (result.uint_32) {
 582		case Opt_fragment_parameter_all:
 583			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 584			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 585			break;
 586		case Opt_fragment_parameter_metadata:
 587			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 588			break;
 589		case Opt_fragment_parameter_data:
 590			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 591			break;
 592		default:
 593			btrfs_info(NULL, "unrecognized fragment option '%s'",
 594				   param->key);
 595			return -EINVAL;
 596		}
 597		break;
 598#endif
 599#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 600	case Opt_ref_verify:
 601		btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
 602		break;
 603#endif
 604	default:
 605		btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
 606		return -EINVAL;
 607	}
 608
 609	return 0;
 610}
 611
 612/*
 613 * Some options only have meaning at mount time and shouldn't persist across
 614 * remounts, or be displayed. Clear these at the end of mount and remount code
 615 * paths.
 616 */
 617static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
 618{
 619	btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
 620	btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
 621	btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
 622}
 623
 624static bool check_ro_option(struct btrfs_fs_info *fs_info,
 625			    unsigned long mount_opt, unsigned long opt,
 626			    const char *opt_name)
 627{
 628	if (mount_opt & opt) {
 629		btrfs_err(fs_info, "%s must be used with ro mount option",
 630			  opt_name);
 631		return true;
 632	}
 633	return false;
 634}
 635
 636bool btrfs_check_options(struct btrfs_fs_info *info, unsigned long *mount_opt,
 
 637			 unsigned long flags)
 638{
 639	bool ret = true;
 640
 641	if (!(flags & SB_RDONLY) &&
 642	    (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
 643	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
 644	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums")))
 
 
 645		ret = false;
 646
 647	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 648	    !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
 649	    !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
 650		btrfs_err(info, "cannot disable free-space-tree");
 651		ret = false;
 652	}
 653	if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
 654	     !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
 655		btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
 656		ret = false;
 657	}
 658
 659	if (btrfs_check_mountopts_zoned(info, mount_opt))
 660		ret = false;
 661
 662	if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
 663		if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE))
 664			btrfs_info(info, "disk space caching is enabled");
 
 
 
 665		if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
 666			btrfs_info(info, "using free-space-tree");
 667	}
 668
 669	return ret;
 670}
 671
 672/*
 673 * This is subtle, we only call this during open_ctree().  We need to pre-load
 674 * the mount options with the on-disk settings.  Before the new mount API took
 675 * effect we would do this on mount and remount.  With the new mount API we'll
 676 * only do this on the initial mount.
 677 *
 678 * This isn't a change in behavior, because we're using the current state of the
 679 * file system to set the current mount options.  If you mounted with special
 680 * options to disable these features and then remounted we wouldn't revert the
 681 * settings, because mounting without these features cleared the on-disk
 682 * settings, so this being called on re-mount is not needed.
 683 */
 684void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
 685{
 686	if (fs_info->sectorsize < PAGE_SIZE) {
 687		btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
 688		if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
 689			btrfs_info(fs_info,
 690				   "forcing free space tree for sector size %u with page size %lu",
 691				   fs_info->sectorsize, PAGE_SIZE);
 692			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 693		}
 694	}
 695
 696	/*
 697	 * At this point our mount options are populated, so we only mess with
 698	 * these settings if we don't have any settings already.
 699	 */
 700	if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
 701		return;
 702
 703	if (btrfs_is_zoned(fs_info) &&
 704	    btrfs_free_space_cache_v1_active(fs_info)) {
 705		btrfs_info(fs_info, "zoned: clearing existing space cache");
 706		btrfs_set_super_cache_generation(fs_info->super_copy, 0);
 707		return;
 708	}
 709
 710	if (btrfs_test_opt(fs_info, SPACE_CACHE))
 711		return;
 712
 713	if (btrfs_test_opt(fs_info, NOSPACECACHE))
 714		return;
 715
 716	/*
 717	 * At this point we don't have explicit options set by the user, set
 718	 * them ourselves based on the state of the file system.
 719	 */
 720	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
 721		btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 722	else if (btrfs_free_space_cache_v1_active(fs_info))
 723		btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
 724}
 725
 726static void set_device_specific_options(struct btrfs_fs_info *fs_info)
 727{
 728	if (!btrfs_test_opt(fs_info, NOSSD) &&
 729	    !fs_info->fs_devices->rotating)
 730		btrfs_set_opt(fs_info->mount_opt, SSD);
 731
 732	/*
 733	 * For devices supporting discard turn on discard=async automatically,
 734	 * unless it's already set or disabled. This could be turned off by
 735	 * nodiscard for the same mount.
 736	 *
 737	 * The zoned mode piggy backs on the discard functionality for
 738	 * resetting a zone. There is no reason to delay the zone reset as it is
 739	 * fast enough. So, do not enable async discard for zoned mode.
 740	 */
 741	if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
 742	      btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
 743	      btrfs_test_opt(fs_info, NODISCARD)) &&
 744	    fs_info->fs_devices->discardable &&
 745	    !btrfs_is_zoned(fs_info))
 746		btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
 747}
 748
 749char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 750					  u64 subvol_objectid)
 751{
 752	struct btrfs_root *root = fs_info->tree_root;
 753	struct btrfs_root *fs_root = NULL;
 754	struct btrfs_root_ref *root_ref;
 755	struct btrfs_inode_ref *inode_ref;
 756	struct btrfs_key key;
 757	struct btrfs_path *path = NULL;
 758	char *name = NULL, *ptr;
 759	u64 dirid;
 760	int len;
 761	int ret;
 762
 763	path = btrfs_alloc_path();
 764	if (!path) {
 765		ret = -ENOMEM;
 766		goto err;
 767	}
 768
 769	name = kmalloc(PATH_MAX, GFP_KERNEL);
 770	if (!name) {
 771		ret = -ENOMEM;
 772		goto err;
 773	}
 774	ptr = name + PATH_MAX - 1;
 775	ptr[0] = '\0';
 776
 777	/*
 778	 * Walk up the subvolume trees in the tree of tree roots by root
 779	 * backrefs until we hit the top-level subvolume.
 780	 */
 781	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
 782		key.objectid = subvol_objectid;
 783		key.type = BTRFS_ROOT_BACKREF_KEY;
 784		key.offset = (u64)-1;
 785
 786		ret = btrfs_search_backwards(root, &key, path);
 787		if (ret < 0) {
 788			goto err;
 789		} else if (ret > 0) {
 790			ret = -ENOENT;
 791			goto err;
 792		}
 793
 794		subvol_objectid = key.offset;
 795
 796		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
 797					  struct btrfs_root_ref);
 798		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
 799		ptr -= len + 1;
 800		if (ptr < name) {
 801			ret = -ENAMETOOLONG;
 802			goto err;
 803		}
 804		read_extent_buffer(path->nodes[0], ptr + 1,
 805				   (unsigned long)(root_ref + 1), len);
 806		ptr[0] = '/';
 807		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
 808		btrfs_release_path(path);
 809
 810		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
 811		if (IS_ERR(fs_root)) {
 812			ret = PTR_ERR(fs_root);
 813			fs_root = NULL;
 814			goto err;
 815		}
 816
 817		/*
 818		 * Walk up the filesystem tree by inode refs until we hit the
 819		 * root directory.
 820		 */
 821		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
 822			key.objectid = dirid;
 823			key.type = BTRFS_INODE_REF_KEY;
 824			key.offset = (u64)-1;
 825
 826			ret = btrfs_search_backwards(fs_root, &key, path);
 827			if (ret < 0) {
 828				goto err;
 829			} else if (ret > 0) {
 830				ret = -ENOENT;
 831				goto err;
 832			}
 833
 834			dirid = key.offset;
 835
 836			inode_ref = btrfs_item_ptr(path->nodes[0],
 837						   path->slots[0],
 838						   struct btrfs_inode_ref);
 839			len = btrfs_inode_ref_name_len(path->nodes[0],
 840						       inode_ref);
 841			ptr -= len + 1;
 842			if (ptr < name) {
 843				ret = -ENAMETOOLONG;
 844				goto err;
 845			}
 846			read_extent_buffer(path->nodes[0], ptr + 1,
 847					   (unsigned long)(inode_ref + 1), len);
 848			ptr[0] = '/';
 849			btrfs_release_path(path);
 850		}
 851		btrfs_put_root(fs_root);
 852		fs_root = NULL;
 853	}
 854
 855	btrfs_free_path(path);
 856	if (ptr == name + PATH_MAX - 1) {
 857		name[0] = '/';
 858		name[1] = '\0';
 859	} else {
 860		memmove(name, ptr, name + PATH_MAX - ptr);
 861	}
 862	return name;
 863
 864err:
 865	btrfs_put_root(fs_root);
 866	btrfs_free_path(path);
 867	kfree(name);
 868	return ERR_PTR(ret);
 869}
 870
 871static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
 872{
 873	struct btrfs_root *root = fs_info->tree_root;
 874	struct btrfs_dir_item *di;
 875	struct btrfs_path *path;
 876	struct btrfs_key location;
 877	struct fscrypt_str name = FSTR_INIT("default", 7);
 878	u64 dir_id;
 879
 880	path = btrfs_alloc_path();
 881	if (!path)
 882		return -ENOMEM;
 883
 884	/*
 885	 * Find the "default" dir item which points to the root item that we
 886	 * will mount by default if we haven't been given a specific subvolume
 887	 * to mount.
 888	 */
 889	dir_id = btrfs_super_root_dir(fs_info->super_copy);
 890	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
 891	if (IS_ERR(di)) {
 892		btrfs_free_path(path);
 893		return PTR_ERR(di);
 894	}
 895	if (!di) {
 896		/*
 897		 * Ok the default dir item isn't there.  This is weird since
 898		 * it's always been there, but don't freak out, just try and
 899		 * mount the top-level subvolume.
 900		 */
 901		btrfs_free_path(path);
 902		*objectid = BTRFS_FS_TREE_OBJECTID;
 903		return 0;
 904	}
 905
 906	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 907	btrfs_free_path(path);
 908	*objectid = location.objectid;
 909	return 0;
 910}
 911
 912static int btrfs_fill_super(struct super_block *sb,
 913			    struct btrfs_fs_devices *fs_devices,
 914			    void *data)
 915{
 916	struct inode *inode;
 917	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 918	int err;
 919
 920	sb->s_maxbytes = MAX_LFS_FILESIZE;
 921	sb->s_magic = BTRFS_SUPER_MAGIC;
 922	sb->s_op = &btrfs_super_ops;
 923	sb->s_d_op = &btrfs_dentry_operations;
 924	sb->s_export_op = &btrfs_export_ops;
 925#ifdef CONFIG_FS_VERITY
 926	sb->s_vop = &btrfs_verityops;
 927#endif
 928	sb->s_xattr = btrfs_xattr_handlers;
 929	sb->s_time_gran = 1;
 930	sb->s_iflags |= SB_I_CGROUPWB;
 931
 932	err = super_setup_bdi(sb);
 933	if (err) {
 934		btrfs_err(fs_info, "super_setup_bdi failed");
 935		return err;
 936	}
 937
 938	err = open_ctree(sb, fs_devices, (char *)data);
 939	if (err) {
 940		btrfs_err(fs_info, "open_ctree failed");
 941		return err;
 942	}
 943
 944	inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
 945	if (IS_ERR(inode)) {
 946		err = PTR_ERR(inode);
 947		btrfs_handle_fs_error(fs_info, err, NULL);
 948		goto fail_close;
 949	}
 950
 951	sb->s_root = d_make_root(inode);
 952	if (!sb->s_root) {
 953		err = -ENOMEM;
 954		goto fail_close;
 955	}
 956
 957	sb->s_flags |= SB_ACTIVE;
 958	return 0;
 959
 960fail_close:
 961	close_ctree(fs_info);
 962	return err;
 963}
 964
 965int btrfs_sync_fs(struct super_block *sb, int wait)
 966{
 967	struct btrfs_trans_handle *trans;
 968	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 969	struct btrfs_root *root = fs_info->tree_root;
 970
 971	trace_btrfs_sync_fs(fs_info, wait);
 972
 973	if (!wait) {
 974		filemap_flush(fs_info->btree_inode->i_mapping);
 975		return 0;
 976	}
 977
 978	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
 979
 980	trans = btrfs_attach_transaction_barrier(root);
 981	if (IS_ERR(trans)) {
 982		/* no transaction, don't bother */
 983		if (PTR_ERR(trans) == -ENOENT) {
 984			/*
 985			 * Exit unless we have some pending changes
 986			 * that need to go through commit
 987			 */
 988			if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
 989				      &fs_info->flags))
 990				return 0;
 991			/*
 992			 * A non-blocking test if the fs is frozen. We must not
 993			 * start a new transaction here otherwise a deadlock
 994			 * happens. The pending operations are delayed to the
 995			 * next commit after thawing.
 996			 */
 997			if (sb_start_write_trylock(sb))
 998				sb_end_write(sb);
 999			else
1000				return 0;
1001			trans = btrfs_start_transaction(root, 0);
1002		}
1003		if (IS_ERR(trans))
1004			return PTR_ERR(trans);
1005	}
1006	return btrfs_commit_transaction(trans);
1007}
1008
1009static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1010{
1011	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1012	*printed = true;
1013}
1014
1015static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1016{
1017	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1018	const char *compress_type;
1019	const char *subvol_name;
1020	bool printed = false;
1021
1022	if (btrfs_test_opt(info, DEGRADED))
1023		seq_puts(seq, ",degraded");
1024	if (btrfs_test_opt(info, NODATASUM))
1025		seq_puts(seq, ",nodatasum");
1026	if (btrfs_test_opt(info, NODATACOW))
1027		seq_puts(seq, ",nodatacow");
1028	if (btrfs_test_opt(info, NOBARRIER))
1029		seq_puts(seq, ",nobarrier");
1030	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1031		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1032	if (info->thread_pool_size !=  min_t(unsigned long,
1033					     num_online_cpus() + 2, 8))
1034		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1035	if (btrfs_test_opt(info, COMPRESS)) {
1036		compress_type = btrfs_compress_type2str(info->compress_type);
1037		if (btrfs_test_opt(info, FORCE_COMPRESS))
1038			seq_printf(seq, ",compress-force=%s", compress_type);
1039		else
1040			seq_printf(seq, ",compress=%s", compress_type);
1041		if (info->compress_level)
1042			seq_printf(seq, ":%d", info->compress_level);
1043	}
1044	if (btrfs_test_opt(info, NOSSD))
1045		seq_puts(seq, ",nossd");
1046	if (btrfs_test_opt(info, SSD_SPREAD))
1047		seq_puts(seq, ",ssd_spread");
1048	else if (btrfs_test_opt(info, SSD))
1049		seq_puts(seq, ",ssd");
1050	if (btrfs_test_opt(info, NOTREELOG))
1051		seq_puts(seq, ",notreelog");
1052	if (btrfs_test_opt(info, NOLOGREPLAY))
1053		print_rescue_option(seq, "nologreplay", &printed);
1054	if (btrfs_test_opt(info, USEBACKUPROOT))
1055		print_rescue_option(seq, "usebackuproot", &printed);
1056	if (btrfs_test_opt(info, IGNOREBADROOTS))
1057		print_rescue_option(seq, "ignorebadroots", &printed);
1058	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1059		print_rescue_option(seq, "ignoredatacsums", &printed);
 
 
 
 
1060	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1061		seq_puts(seq, ",flushoncommit");
1062	if (btrfs_test_opt(info, DISCARD_SYNC))
1063		seq_puts(seq, ",discard");
1064	if (btrfs_test_opt(info, DISCARD_ASYNC))
1065		seq_puts(seq, ",discard=async");
1066	if (!(info->sb->s_flags & SB_POSIXACL))
1067		seq_puts(seq, ",noacl");
1068	if (btrfs_free_space_cache_v1_active(info))
1069		seq_puts(seq, ",space_cache");
1070	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1071		seq_puts(seq, ",space_cache=v2");
1072	else
1073		seq_puts(seq, ",nospace_cache");
1074	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1075		seq_puts(seq, ",rescan_uuid_tree");
1076	if (btrfs_test_opt(info, CLEAR_CACHE))
1077		seq_puts(seq, ",clear_cache");
1078	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1079		seq_puts(seq, ",user_subvol_rm_allowed");
1080	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1081		seq_puts(seq, ",enospc_debug");
1082	if (btrfs_test_opt(info, AUTO_DEFRAG))
1083		seq_puts(seq, ",autodefrag");
1084	if (btrfs_test_opt(info, SKIP_BALANCE))
1085		seq_puts(seq, ",skip_balance");
1086	if (info->metadata_ratio)
1087		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1088	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1089		seq_puts(seq, ",fatal_errors=panic");
1090	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1091		seq_printf(seq, ",commit=%u", info->commit_interval);
1092#ifdef CONFIG_BTRFS_DEBUG
1093	if (btrfs_test_opt(info, FRAGMENT_DATA))
1094		seq_puts(seq, ",fragment=data");
1095	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1096		seq_puts(seq, ",fragment=metadata");
1097#endif
1098	if (btrfs_test_opt(info, REF_VERIFY))
1099		seq_puts(seq, ",ref_verify");
1100	seq_printf(seq, ",subvolid=%llu",
1101		  BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1102	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1103			BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1104	if (!IS_ERR(subvol_name)) {
1105		seq_puts(seq, ",subvol=");
1106		seq_escape(seq, subvol_name, " \t\n\\");
1107		kfree(subvol_name);
1108	}
1109	return 0;
1110}
1111
1112/*
1113 * subvolumes are identified by ino 256
1114 */
1115static inline int is_subvolume_inode(struct inode *inode)
1116{
1117	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1118		return 1;
1119	return 0;
1120}
1121
1122static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1123				   struct vfsmount *mnt)
1124{
1125	struct dentry *root;
1126	int ret;
1127
1128	if (!subvol_name) {
1129		if (!subvol_objectid) {
1130			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1131							  &subvol_objectid);
1132			if (ret) {
1133				root = ERR_PTR(ret);
1134				goto out;
1135			}
1136		}
1137		subvol_name = btrfs_get_subvol_name_from_objectid(
1138					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1139		if (IS_ERR(subvol_name)) {
1140			root = ERR_CAST(subvol_name);
1141			subvol_name = NULL;
1142			goto out;
1143		}
1144
1145	}
1146
1147	root = mount_subtree(mnt, subvol_name);
1148	/* mount_subtree() drops our reference on the vfsmount. */
1149	mnt = NULL;
1150
1151	if (!IS_ERR(root)) {
1152		struct super_block *s = root->d_sb;
1153		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1154		struct inode *root_inode = d_inode(root);
1155		u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1156
1157		ret = 0;
1158		if (!is_subvolume_inode(root_inode)) {
1159			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1160			       subvol_name);
1161			ret = -EINVAL;
1162		}
1163		if (subvol_objectid && root_objectid != subvol_objectid) {
1164			/*
1165			 * This will also catch a race condition where a
1166			 * subvolume which was passed by ID is renamed and
1167			 * another subvolume is renamed over the old location.
1168			 */
1169			btrfs_err(fs_info,
1170				  "subvol '%s' does not match subvolid %llu",
1171				  subvol_name, subvol_objectid);
1172			ret = -EINVAL;
1173		}
1174		if (ret) {
1175			dput(root);
1176			root = ERR_PTR(ret);
1177			deactivate_locked_super(s);
1178		}
1179	}
1180
1181out:
1182	mntput(mnt);
1183	kfree(subvol_name);
1184	return root;
1185}
1186
1187static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1188				     u32 new_pool_size, u32 old_pool_size)
1189{
1190	if (new_pool_size == old_pool_size)
1191		return;
1192
1193	fs_info->thread_pool_size = new_pool_size;
1194
1195	btrfs_info(fs_info, "resize thread pool %d -> %d",
1196	       old_pool_size, new_pool_size);
1197
1198	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1199	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1200	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1201	workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1202	workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1203	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1204	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1205	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1206}
1207
1208static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1209				       unsigned long old_opts, int flags)
1210{
1211	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1212	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1213	     (flags & SB_RDONLY))) {
1214		/* wait for any defraggers to finish */
1215		wait_event(fs_info->transaction_wait,
1216			   (atomic_read(&fs_info->defrag_running) == 0));
1217		if (flags & SB_RDONLY)
1218			sync_filesystem(fs_info->sb);
1219	}
1220}
1221
1222static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1223					 unsigned long old_opts)
1224{
1225	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1226
1227	/*
1228	 * We need to cleanup all defragable inodes if the autodefragment is
1229	 * close or the filesystem is read only.
1230	 */
1231	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1232	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1233		btrfs_cleanup_defrag_inodes(fs_info);
1234	}
1235
1236	/* If we toggled discard async */
1237	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1238	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1239		btrfs_discard_resume(fs_info);
1240	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1241		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1242		btrfs_discard_cleanup(fs_info);
1243
1244	/* If we toggled space cache */
1245	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1246		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1247}
1248
1249static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1250{
1251	int ret;
1252
1253	if (BTRFS_FS_ERROR(fs_info)) {
1254		btrfs_err(fs_info,
1255			  "remounting read-write after error is not allowed");
1256		return -EINVAL;
1257	}
1258
1259	if (fs_info->fs_devices->rw_devices == 0)
1260		return -EACCES;
1261
1262	if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1263		btrfs_warn(fs_info,
1264			   "too many missing devices, writable remount is not allowed");
1265		return -EACCES;
1266	}
1267
1268	if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1269		btrfs_warn(fs_info,
1270			   "mount required to replay tree-log, cannot remount read-write");
1271		return -EINVAL;
1272	}
1273
1274	/*
1275	 * NOTE: when remounting with a change that does writes, don't put it
1276	 * anywhere above this point, as we are not sure to be safe to write
1277	 * until we pass the above checks.
1278	 */
1279	ret = btrfs_start_pre_rw_mount(fs_info);
1280	if (ret)
1281		return ret;
1282
1283	btrfs_clear_sb_rdonly(fs_info->sb);
1284
1285	set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1286
1287	/*
1288	 * If we've gone from readonly -> read-write, we need to get our
1289	 * sync/async discard lists in the right state.
1290	 */
1291	btrfs_discard_resume(fs_info);
1292
1293	return 0;
1294}
1295
1296static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1297{
1298	/*
1299	 * This also happens on 'umount -rf' or on shutdown, when the
1300	 * filesystem is busy.
1301	 */
1302	cancel_work_sync(&fs_info->async_reclaim_work);
1303	cancel_work_sync(&fs_info->async_data_reclaim_work);
1304
1305	btrfs_discard_cleanup(fs_info);
1306
1307	/* Wait for the uuid_scan task to finish */
1308	down(&fs_info->uuid_tree_rescan_sem);
1309	/* Avoid complains from lockdep et al. */
1310	up(&fs_info->uuid_tree_rescan_sem);
1311
1312	btrfs_set_sb_rdonly(fs_info->sb);
1313
1314	/*
1315	 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1316	 * loop if it's already active.  If it's already asleep, we'll leave
1317	 * unused block groups on disk until we're mounted read-write again
1318	 * unless we clean them up here.
1319	 */
1320	btrfs_delete_unused_bgs(fs_info);
1321
1322	/*
1323	 * The cleaner task could be already running before we set the flag
1324	 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).  We must make
1325	 * sure that after we finish the remount, i.e. after we call
1326	 * btrfs_commit_super(), the cleaner can no longer start a transaction
1327	 * - either because it was dropping a dead root, running delayed iputs
1328	 *   or deleting an unused block group (the cleaner picked a block
1329	 *   group from the list of unused block groups before we were able to
1330	 *   in the previous call to btrfs_delete_unused_bgs()).
1331	 */
1332	wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1333
1334	/*
1335	 * We've set the superblock to RO mode, so we might have made the
1336	 * cleaner task sleep without running all pending delayed iputs. Go
1337	 * through all the delayed iputs here, so that if an unmount happens
1338	 * without remounting RW we don't end up at finishing close_ctree()
1339	 * with a non-empty list of delayed iputs.
1340	 */
1341	btrfs_run_delayed_iputs(fs_info);
1342
1343	btrfs_dev_replace_suspend_for_unmount(fs_info);
1344	btrfs_scrub_cancel(fs_info);
1345	btrfs_pause_balance(fs_info);
1346
1347	/*
1348	 * Pause the qgroup rescan worker if it is running. We don't want it to
1349	 * be still running after we are in RO mode, as after that, by the time
1350	 * we unmount, it might have left a transaction open, so we would leak
1351	 * the transaction and/or crash.
1352	 */
1353	btrfs_qgroup_wait_for_completion(fs_info, false);
1354
1355	return btrfs_commit_super(fs_info);
1356}
1357
1358static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1359{
1360	fs_info->max_inline = ctx->max_inline;
1361	fs_info->commit_interval = ctx->commit_interval;
1362	fs_info->metadata_ratio = ctx->metadata_ratio;
1363	fs_info->thread_pool_size = ctx->thread_pool_size;
1364	fs_info->mount_opt = ctx->mount_opt;
1365	fs_info->compress_type = ctx->compress_type;
1366	fs_info->compress_level = ctx->compress_level;
1367}
1368
1369static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1370{
1371	ctx->max_inline = fs_info->max_inline;
1372	ctx->commit_interval = fs_info->commit_interval;
1373	ctx->metadata_ratio = fs_info->metadata_ratio;
1374	ctx->thread_pool_size = fs_info->thread_pool_size;
1375	ctx->mount_opt = fs_info->mount_opt;
1376	ctx->compress_type = fs_info->compress_type;
1377	ctx->compress_level = fs_info->compress_level;
1378}
1379
1380#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...)			\
1381do {										\
1382	if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1383	    btrfs_raw_test_opt(fs_info->mount_opt, opt))			\
1384		btrfs_info(fs_info, fmt, ##args);				\
1385} while (0)
1386
1387#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...)	\
1388do {									\
1389	if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1390	    !btrfs_raw_test_opt(fs_info->mount_opt, opt))		\
1391		btrfs_info(fs_info, fmt, ##args);			\
1392} while (0)
1393
1394static void btrfs_emit_options(struct btrfs_fs_info *info,
1395			       struct btrfs_fs_context *old)
1396{
1397	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1398	btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1399	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1400	btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1401	btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1402	btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1403	btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1404	btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1405	btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1406	btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1407	btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1408	btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1409	btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1410	btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1411	btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1412	btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1413	btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1414	btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1415	btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1416	btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1417	btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
 
 
1418
1419	btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1420	btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1421	btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1422	btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1423	btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1424	btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1425	btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1426	btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1427	btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1428
1429	/* Did the compression settings change? */
1430	if (btrfs_test_opt(info, COMPRESS) &&
1431	    (!old ||
1432	     old->compress_type != info->compress_type ||
1433	     old->compress_level != info->compress_level ||
1434	     (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1435	      btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1436		const char *compress_type = btrfs_compress_type2str(info->compress_type);
1437
1438		btrfs_info(info, "%s %s compression, level %d",
1439			   btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1440			   compress_type, info->compress_level);
1441	}
1442
1443	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1444		btrfs_info(info, "max_inline set to %llu", info->max_inline);
1445}
1446
1447static int btrfs_reconfigure(struct fs_context *fc)
1448{
1449	struct super_block *sb = fc->root->d_sb;
1450	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1451	struct btrfs_fs_context *ctx = fc->fs_private;
1452	struct btrfs_fs_context old_ctx;
1453	int ret = 0;
1454	bool mount_reconfigure = (fc->s_fs_info != NULL);
1455
1456	btrfs_info_to_ctx(fs_info, &old_ctx);
1457
1458	/*
1459	 * This is our "bind mount" trick, we don't want to allow the user to do
1460	 * anything other than mount a different ro/rw and a different subvol,
1461	 * all of the mount options should be maintained.
1462	 */
1463	if (mount_reconfigure)
1464		ctx->mount_opt = old_ctx.mount_opt;
1465
1466	sync_filesystem(sb);
1467	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1468
1469	if (!mount_reconfigure &&
1470	    !btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1471		return -EINVAL;
1472
1473	ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1474	if (ret < 0)
1475		return ret;
1476
1477	btrfs_ctx_to_info(fs_info, ctx);
1478	btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1479	btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1480				 old_ctx.thread_pool_size);
1481
1482	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1483	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1484	    (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1485		btrfs_warn(fs_info,
1486		"remount supports changing free space tree only from RO to RW");
1487		/* Make sure free space cache options match the state on disk. */
1488		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1489			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1490			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1491		}
1492		if (btrfs_free_space_cache_v1_active(fs_info)) {
1493			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1494			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1495		}
1496	}
1497
1498	ret = 0;
1499	if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1500		ret = btrfs_remount_ro(fs_info);
1501	else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1502		ret = btrfs_remount_rw(fs_info);
1503	if (ret)
1504		goto restore;
1505
1506	/*
1507	 * If we set the mask during the parameter parsing VFS would reject the
1508	 * remount.  Here we can set the mask and the value will be updated
1509	 * appropriately.
1510	 */
1511	if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1512		fc->sb_flags_mask |= SB_POSIXACL;
1513
1514	btrfs_emit_options(fs_info, &old_ctx);
1515	wake_up_process(fs_info->transaction_kthread);
1516	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1517	btrfs_clear_oneshot_options(fs_info);
1518	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1519
1520	return 0;
1521restore:
1522	btrfs_ctx_to_info(fs_info, &old_ctx);
1523	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1524	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1525	return ret;
1526}
1527
1528/* Used to sort the devices by max_avail(descending sort) */
1529static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1530{
1531	const struct btrfs_device_info *dev_info1 = a;
1532	const struct btrfs_device_info *dev_info2 = b;
1533
1534	if (dev_info1->max_avail > dev_info2->max_avail)
1535		return -1;
1536	else if (dev_info1->max_avail < dev_info2->max_avail)
1537		return 1;
1538	return 0;
1539}
1540
1541/*
1542 * sort the devices by max_avail, in which max free extent size of each device
1543 * is stored.(Descending Sort)
1544 */
1545static inline void btrfs_descending_sort_devices(
1546					struct btrfs_device_info *devices,
1547					size_t nr_devices)
1548{
1549	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1550	     btrfs_cmp_device_free_bytes, NULL);
1551}
1552
1553/*
1554 * The helper to calc the free space on the devices that can be used to store
1555 * file data.
1556 */
1557static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1558					      u64 *free_bytes)
1559{
1560	struct btrfs_device_info *devices_info;
1561	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1562	struct btrfs_device *device;
1563	u64 type;
1564	u64 avail_space;
1565	u64 min_stripe_size;
1566	int num_stripes = 1;
1567	int i = 0, nr_devices;
1568	const struct btrfs_raid_attr *rattr;
1569
1570	/*
1571	 * We aren't under the device list lock, so this is racy-ish, but good
1572	 * enough for our purposes.
1573	 */
1574	nr_devices = fs_info->fs_devices->open_devices;
1575	if (!nr_devices) {
1576		smp_mb();
1577		nr_devices = fs_info->fs_devices->open_devices;
1578		ASSERT(nr_devices);
1579		if (!nr_devices) {
1580			*free_bytes = 0;
1581			return 0;
1582		}
1583	}
1584
1585	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1586			       GFP_KERNEL);
1587	if (!devices_info)
1588		return -ENOMEM;
1589
1590	/* calc min stripe number for data space allocation */
1591	type = btrfs_data_alloc_profile(fs_info);
1592	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1593
1594	if (type & BTRFS_BLOCK_GROUP_RAID0)
1595		num_stripes = nr_devices;
1596	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1597		num_stripes = rattr->ncopies;
1598	else if (type & BTRFS_BLOCK_GROUP_RAID10)
1599		num_stripes = 4;
1600
1601	/* Adjust for more than 1 stripe per device */
1602	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1603
1604	rcu_read_lock();
1605	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1606		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1607						&device->dev_state) ||
1608		    !device->bdev ||
1609		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1610			continue;
1611
1612		if (i >= nr_devices)
1613			break;
1614
1615		avail_space = device->total_bytes - device->bytes_used;
1616
1617		/* align with stripe_len */
1618		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1619
1620		/*
1621		 * Ensure we have at least min_stripe_size on top of the
1622		 * reserved space on the device.
1623		 */
1624		if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1625			continue;
1626
1627		avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1628
1629		devices_info[i].dev = device;
1630		devices_info[i].max_avail = avail_space;
1631
1632		i++;
1633	}
1634	rcu_read_unlock();
1635
1636	nr_devices = i;
1637
1638	btrfs_descending_sort_devices(devices_info, nr_devices);
1639
1640	i = nr_devices - 1;
1641	avail_space = 0;
1642	while (nr_devices >= rattr->devs_min) {
1643		num_stripes = min(num_stripes, nr_devices);
1644
1645		if (devices_info[i].max_avail >= min_stripe_size) {
1646			int j;
1647			u64 alloc_size;
1648
1649			avail_space += devices_info[i].max_avail * num_stripes;
1650			alloc_size = devices_info[i].max_avail;
1651			for (j = i + 1 - num_stripes; j <= i; j++)
1652				devices_info[j].max_avail -= alloc_size;
1653		}
1654		i--;
1655		nr_devices--;
1656	}
1657
1658	kfree(devices_info);
1659	*free_bytes = avail_space;
1660	return 0;
1661}
1662
1663/*
1664 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1665 *
1666 * If there's a redundant raid level at DATA block groups, use the respective
1667 * multiplier to scale the sizes.
1668 *
1669 * Unused device space usage is based on simulating the chunk allocator
1670 * algorithm that respects the device sizes and order of allocations.  This is
1671 * a close approximation of the actual use but there are other factors that may
1672 * change the result (like a new metadata chunk).
1673 *
1674 * If metadata is exhausted, f_bavail will be 0.
1675 */
1676static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1677{
1678	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1679	struct btrfs_super_block *disk_super = fs_info->super_copy;
1680	struct btrfs_space_info *found;
1681	u64 total_used = 0;
1682	u64 total_free_data = 0;
1683	u64 total_free_meta = 0;
1684	u32 bits = fs_info->sectorsize_bits;
1685	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1686	unsigned factor = 1;
1687	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1688	int ret;
1689	u64 thresh = 0;
1690	int mixed = 0;
1691
1692	list_for_each_entry(found, &fs_info->space_info, list) {
1693		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1694			int i;
1695
1696			total_free_data += found->disk_total - found->disk_used;
1697			total_free_data -=
1698				btrfs_account_ro_block_groups_free_space(found);
1699
1700			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1701				if (!list_empty(&found->block_groups[i]))
1702					factor = btrfs_bg_type_to_factor(
1703						btrfs_raid_array[i].bg_flag);
1704			}
1705		}
1706
1707		/*
1708		 * Metadata in mixed block group profiles are accounted in data
1709		 */
1710		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1711			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1712				mixed = 1;
1713			else
1714				total_free_meta += found->disk_total -
1715					found->disk_used;
1716		}
1717
1718		total_used += found->disk_used;
1719	}
1720
1721	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1722	buf->f_blocks >>= bits;
1723	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1724
1725	/* Account global block reserve as used, it's in logical size already */
1726	spin_lock(&block_rsv->lock);
1727	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
1728	if (buf->f_bfree >= block_rsv->size >> bits)
1729		buf->f_bfree -= block_rsv->size >> bits;
1730	else
1731		buf->f_bfree = 0;
1732	spin_unlock(&block_rsv->lock);
1733
1734	buf->f_bavail = div_u64(total_free_data, factor);
1735	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1736	if (ret)
1737		return ret;
1738	buf->f_bavail += div_u64(total_free_data, factor);
1739	buf->f_bavail = buf->f_bavail >> bits;
1740
1741	/*
1742	 * We calculate the remaining metadata space minus global reserve. If
1743	 * this is (supposedly) smaller than zero, there's no space. But this
1744	 * does not hold in practice, the exhausted state happens where's still
1745	 * some positive delta. So we apply some guesswork and compare the
1746	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
1747	 *
1748	 * We probably cannot calculate the exact threshold value because this
1749	 * depends on the internal reservations requested by various
1750	 * operations, so some operations that consume a few metadata will
1751	 * succeed even if the Avail is zero. But this is better than the other
1752	 * way around.
1753	 */
1754	thresh = SZ_4M;
1755
1756	/*
1757	 * We only want to claim there's no available space if we can no longer
1758	 * allocate chunks for our metadata profile and our global reserve will
1759	 * not fit in the free metadata space.  If we aren't ->full then we
1760	 * still can allocate chunks and thus are fine using the currently
1761	 * calculated f_bavail.
1762	 */
1763	if (!mixed && block_rsv->space_info->full &&
1764	    (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1765		buf->f_bavail = 0;
1766
1767	buf->f_type = BTRFS_SUPER_MAGIC;
1768	buf->f_bsize = fs_info->sectorsize;
1769	buf->f_namelen = BTRFS_NAME_LEN;
1770
1771	/* We treat it as constant endianness (it doesn't matter _which_)
1772	   because we want the fsid to come out the same whether mounted
1773	   on a big-endian or little-endian host */
1774	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1775	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1776	/* Mask in the root object ID too, to disambiguate subvols */
1777	buf->f_fsid.val[0] ^=
1778		BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
1779	buf->f_fsid.val[1] ^=
1780		BTRFS_I(d_inode(dentry))->root->root_key.objectid;
1781
1782	return 0;
1783}
1784
1785static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1786{
1787	struct btrfs_fs_info *p = fc->s_fs_info;
1788	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1789
1790	return fs_info->fs_devices == p->fs_devices;
1791}
1792
1793static int btrfs_get_tree_super(struct fs_context *fc)
1794{
1795	struct btrfs_fs_info *fs_info = fc->s_fs_info;
1796	struct btrfs_fs_context *ctx = fc->fs_private;
1797	struct btrfs_fs_devices *fs_devices = NULL;
1798	struct block_device *bdev;
1799	struct btrfs_device *device;
1800	struct super_block *sb;
1801	blk_mode_t mode = btrfs_open_mode(fc);
1802	int ret;
1803
1804	btrfs_ctx_to_info(fs_info, ctx);
1805	mutex_lock(&uuid_mutex);
1806
1807	/*
1808	 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1809	 * either a valid device or an error.
1810	 */
1811	device = btrfs_scan_one_device(fc->source, mode, true);
1812	ASSERT(device != NULL);
1813	if (IS_ERR(device)) {
1814		mutex_unlock(&uuid_mutex);
1815		return PTR_ERR(device);
1816	}
1817
1818	fs_devices = device->fs_devices;
1819	fs_info->fs_devices = fs_devices;
1820
1821	ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1822	mutex_unlock(&uuid_mutex);
1823	if (ret)
1824		return ret;
1825
1826	if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1827		ret = -EACCES;
1828		goto error;
1829	}
1830
1831	bdev = fs_devices->latest_dev->bdev;
1832
1833	/*
1834	 * From now on the error handling is not straightforward.
1835	 *
1836	 * If successful, this will transfer the fs_info into the super block,
1837	 * and fc->s_fs_info will be NULL.  However if there's an existing
1838	 * super, we'll still have fc->s_fs_info populated.  If we error
1839	 * completely out it'll be cleaned up when we drop the fs_context,
1840	 * otherwise it's tied to the lifetime of the super_block.
1841	 */
1842	sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1843	if (IS_ERR(sb)) {
1844		ret = PTR_ERR(sb);
1845		goto error;
1846	}
1847
1848	set_device_specific_options(fs_info);
1849
1850	if (sb->s_root) {
1851		btrfs_close_devices(fs_devices);
1852		if ((fc->sb_flags ^ sb->s_flags) & SB_RDONLY)
1853			ret = -EBUSY;
 
 
 
 
1854	} else {
1855		snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1856		shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1857		btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1858		ret = btrfs_fill_super(sb, fs_devices, NULL);
1859	}
1860
1861	if (ret) {
1862		deactivate_locked_super(sb);
1863		return ret;
1864	}
1865
1866	btrfs_clear_oneshot_options(fs_info);
1867
1868	fc->root = dget(sb->s_root);
1869	return 0;
1870
1871error:
1872	btrfs_close_devices(fs_devices);
1873	return ret;
1874}
1875
1876/*
1877 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1878 * with different ro/rw options") the following works:
1879 *
1880 *        (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1881 *       (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1882 *
1883 * which looks nice and innocent but is actually pretty intricate and deserves
1884 * a long comment.
1885 *
1886 * On another filesystem a subvolume mount is close to something like:
1887 *
1888 *	(iii) # create rw superblock + initial mount
1889 *	      mount -t xfs /dev/sdb /opt/
1890 *
1891 *	      # create ro bind mount
1892 *	      mount --bind -o ro /opt/foo /mnt/foo
1893 *
1894 *	      # unmount initial mount
1895 *	      umount /opt
1896 *
1897 * Of course, there's some special subvolume sauce and there's the fact that the
1898 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1899 * it's very close and will help us understand the issue.
1900 *
1901 * The old mount API didn't cleanly distinguish between a mount being made ro
1902 * and a superblock being made ro.  The only way to change the ro state of
1903 * either object was by passing ms_rdonly. If a new mount was created via
1904 * mount(2) such as:
1905 *
1906 *      mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1907 *
1908 * the MS_RDONLY flag being specified had two effects:
1909 *
1910 * (1) MNT_READONLY was raised -> the resulting mount got
1911 *     @mnt->mnt_flags |= MNT_READONLY raised.
1912 *
1913 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1914 *     made the superblock ro. Note, how SB_RDONLY has the same value as
1915 *     ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1916 *
1917 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1918 * subtree mounted ro.
1919 *
1920 * But consider the effect on the old mount API on btrfs subvolume mounting
1921 * which combines the distinct step in (iii) into a single step.
1922 *
1923 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1924 * is issued the superblock is ro and thus even if the mount created for (ii) is
1925 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1926 * to rw for (ii) which it did using an internal remount call.
1927 *
1928 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1929 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1930 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1931 * passed by mount(8) to mount(2).
1932 *
1933 * Enter the new mount API. The new mount API disambiguates making a mount ro
1934 * and making a superblock ro.
1935 *
1936 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1937 *     fsmount() or mount_setattr() this is a pure VFS level change for a
1938 *     specific mount or mount tree that is never seen by the filesystem itself.
1939 *
1940 * (4) To turn a superblock ro the "ro" flag must be used with
1941 *     fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1942 *     in fc->sb_flags.
1943 *
1944 * This disambiguation has rather positive consequences.  Mounting a subvolume
1945 * ro will not also turn the superblock ro. Only the mount for the subvolume
1946 * will become ro.
1947 *
1948 * So, if the superblock creation request comes from the new mount API the
1949 * caller must have explicitly done:
1950 *
1951 *      fsconfig(FSCONFIG_SET_FLAG, "ro")
1952 *      fsmount/mount_setattr(MOUNT_ATTR_RDONLY)
1953 *
1954 * IOW, at some point the caller must have explicitly turned the whole
1955 * superblock ro and we shouldn't just undo it like we did for the old mount
1956 * API. In any case, it lets us avoid the hack in the new mount API.
1957 *
1958 * Consequently, the remounting hack must only be used for requests originating
1959 * from the old mount API and should be marked for full deprecation so it can be
1960 * turned off in a couple of years.
1961 *
1962 * The new mount API has no reason to support this hack.
1963 */
1964static struct vfsmount *btrfs_reconfigure_for_mount(struct fs_context *fc)
1965{
1966	struct vfsmount *mnt;
1967	int ret;
1968	const bool ro2rw = !(fc->sb_flags & SB_RDONLY);
1969
1970	/*
1971	 * We got an EBUSY because our SB_RDONLY flag didn't match the existing
1972	 * super block, so invert our setting here and retry the mount so we
1973	 * can get our vfsmount.
1974	 */
1975	if (ro2rw)
1976		fc->sb_flags |= SB_RDONLY;
1977	else
1978		fc->sb_flags &= ~SB_RDONLY;
1979
1980	mnt = fc_mount(fc);
1981	if (IS_ERR(mnt))
1982		return mnt;
1983
1984	if (!fc->oldapi || !ro2rw)
1985		return mnt;
1986
1987	/* We need to convert to rw, call reconfigure. */
1988	fc->sb_flags &= ~SB_RDONLY;
1989	down_write(&mnt->mnt_sb->s_umount);
1990	ret = btrfs_reconfigure(fc);
 
1991	up_write(&mnt->mnt_sb->s_umount);
1992	if (ret) {
1993		mntput(mnt);
1994		return ERR_PTR(ret);
1995	}
1996	return mnt;
1997}
1998
1999static int btrfs_get_tree_subvol(struct fs_context *fc)
2000{
2001	struct btrfs_fs_info *fs_info = NULL;
2002	struct btrfs_fs_context *ctx = fc->fs_private;
2003	struct fs_context *dup_fc;
2004	struct dentry *dentry;
2005	struct vfsmount *mnt;
 
2006
2007	/*
2008	 * Setup a dummy root and fs_info for test/set super.  This is because
2009	 * we don't actually fill this stuff out until open_ctree, but we need
2010	 * then open_ctree will properly initialize the file system specific
2011	 * settings later.  btrfs_init_fs_info initializes the static elements
2012	 * of the fs_info (locks and such) to make cleanup easier if we find a
2013	 * superblock with our given fs_devices later on at sget() time.
2014	 */
2015	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2016	if (!fs_info)
2017		return -ENOMEM;
2018
2019	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2020	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2021	if (!fs_info->super_copy || !fs_info->super_for_commit) {
2022		btrfs_free_fs_info(fs_info);
2023		return -ENOMEM;
2024	}
2025	btrfs_init_fs_info(fs_info);
2026
2027	dup_fc = vfs_dup_fs_context(fc);
2028	if (IS_ERR(dup_fc)) {
2029		btrfs_free_fs_info(fs_info);
2030		return PTR_ERR(dup_fc);
2031	}
2032
2033	/*
2034	 * When we do the sget_fc this gets transferred to the sb, so we only
2035	 * need to set it on the dup_fc as this is what creates the super block.
2036	 */
2037	dup_fc->s_fs_info = fs_info;
2038
2039	/*
2040	 * We'll do the security settings in our btrfs_get_tree_super() mount
2041	 * loop, they were duplicated into dup_fc, we can drop the originals
2042	 * here.
2043	 */
2044	security_free_mnt_opts(&fc->security);
2045	fc->security = NULL;
2046
2047	mnt = fc_mount(dup_fc);
2048	if (PTR_ERR_OR_ZERO(mnt) == -EBUSY)
2049		mnt = btrfs_reconfigure_for_mount(dup_fc);
 
 
 
2050	put_fs_context(dup_fc);
2051	if (IS_ERR(mnt))
2052		return PTR_ERR(mnt);
 
 
2053
2054	/*
2055	 * This free's ->subvol_name, because if it isn't set we have to
2056	 * allocate a buffer to hold the subvol_name, so we just drop our
2057	 * reference to it here.
2058	 */
2059	dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2060	ctx->subvol_name = NULL;
2061	if (IS_ERR(dentry))
2062		return PTR_ERR(dentry);
2063
2064	fc->root = dentry;
2065	return 0;
2066}
2067
2068static int btrfs_get_tree(struct fs_context *fc)
2069{
2070	/*
2071	 * Since we use mount_subtree to mount the default/specified subvol, we
2072	 * have to do mounts in two steps.
2073	 *
2074	 * First pass through we call btrfs_get_tree_subvol(), this is just a
2075	 * wrapper around fc_mount() to call back into here again, and this time
2076	 * we'll call btrfs_get_tree_super().  This will do the open_ctree() and
2077	 * everything to open the devices and file system.  Then we return back
2078	 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2079	 * from there we can do our mount_subvol() call, which will lookup
2080	 * whichever subvol we're mounting and setup this fc with the
2081	 * appropriate dentry for the subvol.
2082	 */
2083	if (fc->s_fs_info)
2084		return btrfs_get_tree_super(fc);
2085	return btrfs_get_tree_subvol(fc);
2086}
2087
2088static void btrfs_kill_super(struct super_block *sb)
2089{
2090	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2091	kill_anon_super(sb);
2092	btrfs_free_fs_info(fs_info);
2093}
2094
2095static void btrfs_free_fs_context(struct fs_context *fc)
2096{
2097	struct btrfs_fs_context *ctx = fc->fs_private;
2098	struct btrfs_fs_info *fs_info = fc->s_fs_info;
2099
2100	if (fs_info)
2101		btrfs_free_fs_info(fs_info);
2102
2103	if (ctx && refcount_dec_and_test(&ctx->refs)) {
2104		kfree(ctx->subvol_name);
2105		kfree(ctx);
2106	}
2107}
2108
2109static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2110{
2111	struct btrfs_fs_context *ctx = src_fc->fs_private;
2112
2113	/*
2114	 * Give a ref to our ctx to this dup, as we want to keep it around for
2115	 * our original fc so we can have the subvolume name or objectid.
2116	 *
2117	 * We unset ->source in the original fc because the dup needs it for
2118	 * mounting, and then once we free the dup it'll free ->source, so we
2119	 * need to make sure we're only pointing to it in one fc.
2120	 */
2121	refcount_inc(&ctx->refs);
2122	fc->fs_private = ctx;
2123	fc->source = src_fc->source;
2124	src_fc->source = NULL;
2125	return 0;
2126}
2127
2128static const struct fs_context_operations btrfs_fs_context_ops = {
2129	.parse_param	= btrfs_parse_param,
2130	.reconfigure	= btrfs_reconfigure,
2131	.get_tree	= btrfs_get_tree,
2132	.dup		= btrfs_dup_fs_context,
2133	.free		= btrfs_free_fs_context,
2134};
2135
2136static int btrfs_init_fs_context(struct fs_context *fc)
2137{
2138	struct btrfs_fs_context *ctx;
2139
2140	ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2141	if (!ctx)
2142		return -ENOMEM;
2143
2144	refcount_set(&ctx->refs, 1);
2145	fc->fs_private = ctx;
2146	fc->ops = &btrfs_fs_context_ops;
2147
2148	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2149		btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2150	} else {
2151		ctx->thread_pool_size =
2152			min_t(unsigned long, num_online_cpus() + 2, 8);
2153		ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2154		ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2155	}
2156
2157#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2158	fc->sb_flags |= SB_POSIXACL;
2159#endif
2160	fc->sb_flags |= SB_I_VERSION;
2161
2162	return 0;
2163}
2164
2165static struct file_system_type btrfs_fs_type = {
2166	.owner			= THIS_MODULE,
2167	.name			= "btrfs",
2168	.init_fs_context	= btrfs_init_fs_context,
2169	.parameters		= btrfs_fs_parameters,
2170	.kill_sb		= btrfs_kill_super,
2171	.fs_flags		= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
 
2172 };
2173
2174MODULE_ALIAS_FS("btrfs");
2175
2176static int btrfs_control_open(struct inode *inode, struct file *file)
2177{
2178	/*
2179	 * The control file's private_data is used to hold the
2180	 * transaction when it is started and is used to keep
2181	 * track of whether a transaction is already in progress.
2182	 */
2183	file->private_data = NULL;
2184	return 0;
2185}
2186
2187/*
2188 * Used by /dev/btrfs-control for devices ioctls.
2189 */
2190static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2191				unsigned long arg)
2192{
2193	struct btrfs_ioctl_vol_args *vol;
2194	struct btrfs_device *device = NULL;
2195	dev_t devt = 0;
2196	int ret = -ENOTTY;
2197
2198	if (!capable(CAP_SYS_ADMIN))
2199		return -EPERM;
2200
2201	vol = memdup_user((void __user *)arg, sizeof(*vol));
2202	if (IS_ERR(vol))
2203		return PTR_ERR(vol);
2204	ret = btrfs_check_ioctl_vol_args_path(vol);
2205	if (ret < 0)
2206		goto out;
2207
2208	switch (cmd) {
2209	case BTRFS_IOC_SCAN_DEV:
2210		mutex_lock(&uuid_mutex);
2211		/*
2212		 * Scanning outside of mount can return NULL which would turn
2213		 * into 0 error code.
2214		 */
2215		device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2216		ret = PTR_ERR_OR_ZERO(device);
2217		mutex_unlock(&uuid_mutex);
2218		break;
2219	case BTRFS_IOC_FORGET_DEV:
2220		if (vol->name[0] != 0) {
2221			ret = lookup_bdev(vol->name, &devt);
2222			if (ret)
2223				break;
2224		}
2225		ret = btrfs_forget_devices(devt);
2226		break;
2227	case BTRFS_IOC_DEVICES_READY:
2228		mutex_lock(&uuid_mutex);
2229		/*
2230		 * Scanning outside of mount can return NULL which would turn
2231		 * into 0 error code.
2232		 */
2233		device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2234		if (IS_ERR_OR_NULL(device)) {
2235			mutex_unlock(&uuid_mutex);
2236			ret = PTR_ERR(device);
 
 
 
2237			break;
2238		}
2239		ret = !(device->fs_devices->num_devices ==
2240			device->fs_devices->total_devices);
2241		mutex_unlock(&uuid_mutex);
2242		break;
2243	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2244		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2245		break;
2246	}
2247
2248out:
2249	kfree(vol);
2250	return ret;
2251}
2252
2253static int btrfs_freeze(struct super_block *sb)
2254{
2255	struct btrfs_trans_handle *trans;
2256	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2257	struct btrfs_root *root = fs_info->tree_root;
2258
2259	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2260	/*
2261	 * We don't need a barrier here, we'll wait for any transaction that
2262	 * could be in progress on other threads (and do delayed iputs that
2263	 * we want to avoid on a frozen filesystem), or do the commit
2264	 * ourselves.
2265	 */
2266	trans = btrfs_attach_transaction_barrier(root);
2267	if (IS_ERR(trans)) {
2268		/* no transaction, don't bother */
2269		if (PTR_ERR(trans) == -ENOENT)
2270			return 0;
2271		return PTR_ERR(trans);
2272	}
2273	return btrfs_commit_transaction(trans);
2274}
2275
2276static int check_dev_super(struct btrfs_device *dev)
2277{
2278	struct btrfs_fs_info *fs_info = dev->fs_info;
2279	struct btrfs_super_block *sb;
2280	u64 last_trans;
2281	u16 csum_type;
2282	int ret = 0;
2283
2284	/* This should be called with fs still frozen. */
2285	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2286
2287	/* Missing dev, no need to check. */
2288	if (!dev->bdev)
2289		return 0;
2290
2291	/* Only need to check the primary super block. */
2292	sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2293	if (IS_ERR(sb))
2294		return PTR_ERR(sb);
2295
2296	/* Verify the checksum. */
2297	csum_type = btrfs_super_csum_type(sb);
2298	if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2299		btrfs_err(fs_info, "csum type changed, has %u expect %u",
2300			  csum_type, btrfs_super_csum_type(fs_info->super_copy));
2301		ret = -EUCLEAN;
2302		goto out;
2303	}
2304
2305	if (btrfs_check_super_csum(fs_info, sb)) {
2306		btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2307		ret = -EUCLEAN;
2308		goto out;
2309	}
2310
2311	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2312	ret = btrfs_validate_super(fs_info, sb, 0);
2313	if (ret < 0)
2314		goto out;
2315
2316	last_trans = btrfs_get_last_trans_committed(fs_info);
2317	if (btrfs_super_generation(sb) != last_trans) {
2318		btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2319			  btrfs_super_generation(sb), last_trans);
2320		ret = -EUCLEAN;
2321		goto out;
2322	}
2323out:
2324	btrfs_release_disk_super(sb);
2325	return ret;
2326}
2327
2328static int btrfs_unfreeze(struct super_block *sb)
2329{
2330	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2331	struct btrfs_device *device;
2332	int ret = 0;
2333
2334	/*
2335	 * Make sure the fs is not changed by accident (like hibernation then
2336	 * modified by other OS).
2337	 * If we found anything wrong, we mark the fs error immediately.
2338	 *
2339	 * And since the fs is frozen, no one can modify the fs yet, thus
2340	 * we don't need to hold device_list_mutex.
2341	 */
2342	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2343		ret = check_dev_super(device);
2344		if (ret < 0) {
2345			btrfs_handle_fs_error(fs_info, ret,
2346				"super block on devid %llu got modified unexpectedly",
2347				device->devid);
2348			break;
2349		}
2350	}
2351	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2352
2353	/*
2354	 * We still return 0, to allow VFS layer to unfreeze the fs even the
2355	 * above checks failed. Since the fs is either fine or read-only, we're
2356	 * safe to continue, without causing further damage.
2357	 */
2358	return 0;
2359}
2360
2361static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2362{
2363	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2364
2365	/*
2366	 * There should be always a valid pointer in latest_dev, it may be stale
2367	 * for a short moment in case it's being deleted but still valid until
2368	 * the end of RCU grace period.
2369	 */
2370	rcu_read_lock();
2371	seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2372	rcu_read_unlock();
2373
2374	return 0;
2375}
2376
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2377static const struct super_operations btrfs_super_ops = {
2378	.drop_inode	= btrfs_drop_inode,
2379	.evict_inode	= btrfs_evict_inode,
2380	.put_super	= btrfs_put_super,
2381	.sync_fs	= btrfs_sync_fs,
2382	.show_options	= btrfs_show_options,
2383	.show_devname	= btrfs_show_devname,
2384	.alloc_inode	= btrfs_alloc_inode,
2385	.destroy_inode	= btrfs_destroy_inode,
2386	.free_inode	= btrfs_free_inode,
2387	.statfs		= btrfs_statfs,
2388	.freeze_fs	= btrfs_freeze,
2389	.unfreeze_fs	= btrfs_unfreeze,
 
 
2390};
2391
2392static const struct file_operations btrfs_ctl_fops = {
2393	.open = btrfs_control_open,
2394	.unlocked_ioctl	 = btrfs_control_ioctl,
2395	.compat_ioctl = compat_ptr_ioctl,
2396	.owner	 = THIS_MODULE,
2397	.llseek = noop_llseek,
2398};
2399
2400static struct miscdevice btrfs_misc = {
2401	.minor		= BTRFS_MINOR,
2402	.name		= "btrfs-control",
2403	.fops		= &btrfs_ctl_fops
2404};
2405
2406MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2407MODULE_ALIAS("devname:btrfs-control");
2408
2409static int __init btrfs_interface_init(void)
2410{
2411	return misc_register(&btrfs_misc);
2412}
2413
2414static __cold void btrfs_interface_exit(void)
2415{
2416	misc_deregister(&btrfs_misc);
2417}
2418
2419static int __init btrfs_print_mod_info(void)
2420{
2421	static const char options[] = ""
2422#ifdef CONFIG_BTRFS_DEBUG
2423			", debug=on"
2424#endif
2425#ifdef CONFIG_BTRFS_ASSERT
2426			", assert=on"
2427#endif
2428#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2429			", ref-verify=on"
2430#endif
2431#ifdef CONFIG_BLK_DEV_ZONED
2432			", zoned=yes"
2433#else
2434			", zoned=no"
2435#endif
2436#ifdef CONFIG_FS_VERITY
2437			", fsverity=yes"
2438#else
2439			", fsverity=no"
2440#endif
2441			;
2442	pr_info("Btrfs loaded%s\n", options);
2443	return 0;
2444}
2445
2446static int register_btrfs(void)
2447{
2448	return register_filesystem(&btrfs_fs_type);
2449}
2450
2451static void unregister_btrfs(void)
2452{
2453	unregister_filesystem(&btrfs_fs_type);
2454}
2455
2456/* Helper structure for long init/exit functions. */
2457struct init_sequence {
2458	int (*init_func)(void);
2459	/* Can be NULL if the init_func doesn't need cleanup. */
2460	void (*exit_func)(void);
2461};
2462
2463static const struct init_sequence mod_init_seq[] = {
2464	{
2465		.init_func = btrfs_props_init,
2466		.exit_func = NULL,
2467	}, {
2468		.init_func = btrfs_init_sysfs,
2469		.exit_func = btrfs_exit_sysfs,
2470	}, {
2471		.init_func = btrfs_init_compress,
2472		.exit_func = btrfs_exit_compress,
2473	}, {
2474		.init_func = btrfs_init_cachep,
2475		.exit_func = btrfs_destroy_cachep,
2476	}, {
 
 
 
2477		.init_func = btrfs_transaction_init,
2478		.exit_func = btrfs_transaction_exit,
2479	}, {
2480		.init_func = btrfs_ctree_init,
2481		.exit_func = btrfs_ctree_exit,
2482	}, {
2483		.init_func = btrfs_free_space_init,
2484		.exit_func = btrfs_free_space_exit,
2485	}, {
2486		.init_func = extent_state_init_cachep,
2487		.exit_func = extent_state_free_cachep,
2488	}, {
2489		.init_func = extent_buffer_init_cachep,
2490		.exit_func = extent_buffer_free_cachep,
2491	}, {
2492		.init_func = btrfs_bioset_init,
2493		.exit_func = btrfs_bioset_exit,
2494	}, {
2495		.init_func = extent_map_init,
2496		.exit_func = extent_map_exit,
2497	}, {
2498		.init_func = ordered_data_init,
2499		.exit_func = ordered_data_exit,
2500	}, {
2501		.init_func = btrfs_delayed_inode_init,
2502		.exit_func = btrfs_delayed_inode_exit,
2503	}, {
2504		.init_func = btrfs_auto_defrag_init,
2505		.exit_func = btrfs_auto_defrag_exit,
2506	}, {
2507		.init_func = btrfs_delayed_ref_init,
2508		.exit_func = btrfs_delayed_ref_exit,
2509	}, {
2510		.init_func = btrfs_prelim_ref_init,
2511		.exit_func = btrfs_prelim_ref_exit,
2512	}, {
2513		.init_func = btrfs_interface_init,
2514		.exit_func = btrfs_interface_exit,
2515	}, {
2516		.init_func = btrfs_print_mod_info,
2517		.exit_func = NULL,
2518	}, {
2519		.init_func = btrfs_run_sanity_tests,
2520		.exit_func = NULL,
2521	}, {
2522		.init_func = register_btrfs,
2523		.exit_func = unregister_btrfs,
2524	}
2525};
2526
2527static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2528
2529static __always_inline void btrfs_exit_btrfs_fs(void)
2530{
2531	int i;
2532
2533	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2534		if (!mod_init_result[i])
2535			continue;
2536		if (mod_init_seq[i].exit_func)
2537			mod_init_seq[i].exit_func();
2538		mod_init_result[i] = false;
2539	}
2540}
2541
2542static void __exit exit_btrfs_fs(void)
2543{
2544	btrfs_exit_btrfs_fs();
2545	btrfs_cleanup_fs_uuids();
2546}
2547
2548static int __init init_btrfs_fs(void)
2549{
2550	int ret;
2551	int i;
2552
2553	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2554		ASSERT(!mod_init_result[i]);
2555		ret = mod_init_seq[i].init_func();
2556		if (ret < 0) {
2557			btrfs_exit_btrfs_fs();
2558			return ret;
2559		}
2560		mod_init_result[i] = true;
2561	}
2562	return 0;
2563}
2564
2565late_initcall(init_btrfs_fs);
2566module_exit(exit_btrfs_fs)
2567
 
2568MODULE_LICENSE("GPL");
2569MODULE_SOFTDEP("pre: crc32c");
2570MODULE_SOFTDEP("pre: xxhash64");
2571MODULE_SOFTDEP("pre: sha256");
2572MODULE_SOFTDEP("pre: blake2b-256");