Linux Audio

Check our new training course

Open-source upstreaming

Need help get the support for your hardware in upstream Linux?
Loading...
v6.8
   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 "print-tree.h"
  38#include "props.h"
  39#include "xattr.h"
  40#include "bio.h"
  41#include "export.h"
  42#include "compression.h"
  43#include "rcu-string.h"
  44#include "dev-replace.h"
  45#include "free-space-cache.h"
  46#include "backref.h"
  47#include "space-info.h"
  48#include "sysfs.h"
  49#include "zoned.h"
  50#include "tests/btrfs-tests.h"
  51#include "block-group.h"
  52#include "discard.h"
  53#include "qgroup.h"
  54#include "raid56.h"
  55#include "fs.h"
  56#include "accessors.h"
  57#include "defrag.h"
  58#include "dir-item.h"
  59#include "ioctl.h"
  60#include "scrub.h"
  61#include "verity.h"
  62#include "super.h"
  63#include "extent-tree.h"
  64#define CREATE_TRACE_POINTS
  65#include <trace/events/btrfs.h>
  66
  67static const struct super_operations btrfs_super_ops;
 
 
 
 
 
 
 
 
 
  68static struct file_system_type btrfs_fs_type;
 
 
 
  69
  70static void btrfs_put_super(struct super_block *sb)
  71{
  72	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
  73
  74	btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
  75	close_ctree(fs_info);
  76}
  77
  78/* Store the mount options related information. */
  79struct btrfs_fs_context {
  80	char *subvol_name;
  81	u64 subvol_objectid;
  82	u64 max_inline;
  83	u32 commit_interval;
  84	u32 metadata_ratio;
  85	u32 thread_pool_size;
  86	unsigned long mount_opt;
  87	unsigned long compress_type:4;
  88	unsigned int compress_level;
  89	refcount_t refs;
  90};
  91
  92enum {
  93	Opt_acl,
  94	Opt_clear_cache,
  95	Opt_commit_interval,
  96	Opt_compress,
  97	Opt_compress_force,
  98	Opt_compress_force_type,
  99	Opt_compress_type,
 100	Opt_degraded,
 101	Opt_device,
 102	Opt_fatal_errors,
 103	Opt_flushoncommit,
 104	Opt_max_inline,
 105	Opt_barrier,
 106	Opt_datacow,
 107	Opt_datasum,
 108	Opt_defrag,
 109	Opt_discard,
 110	Opt_discard_mode,
 
 111	Opt_ratio,
 112	Opt_rescan_uuid_tree,
 113	Opt_skip_balance,
 114	Opt_space_cache,
 115	Opt_space_cache_version,
 116	Opt_ssd,
 117	Opt_ssd_spread,
 118	Opt_subvol,
 119	Opt_subvol_empty,
 120	Opt_subvolid,
 121	Opt_thread_pool,
 122	Opt_treelog,
 123	Opt_user_subvol_rm_allowed,
 124
 125	/* Rescue options */
 126	Opt_rescue,
 127	Opt_usebackuproot,
 128	Opt_nologreplay,
 129	Opt_ignorebadroots,
 130	Opt_ignoredatacsums,
 131	Opt_rescue_all,
 132
 
 
 
 
 133	/* Debugging options */
 134	Opt_enospc_debug,
 
 
 
 135#ifdef CONFIG_BTRFS_DEBUG
 136	Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 137#endif
 138#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 139	Opt_ref_verify,
 140#endif
 141	Opt_err,
 142};
 143
 144enum {
 145	Opt_fatal_errors_panic,
 146	Opt_fatal_errors_bug,
 147};
 148
 149static const struct constant_table btrfs_parameter_fatal_errors[] = {
 150	{ "panic", Opt_fatal_errors_panic },
 151	{ "bug", Opt_fatal_errors_bug },
 152	{}
 153};
 154
 155enum {
 156	Opt_discard_sync,
 157	Opt_discard_async,
 158};
 159
 160static const struct constant_table btrfs_parameter_discard[] = {
 161	{ "sync", Opt_discard_sync },
 162	{ "async", Opt_discard_async },
 163	{}
 164};
 165
 166enum {
 167	Opt_space_cache_v1,
 168	Opt_space_cache_v2,
 169};
 170
 171static const struct constant_table btrfs_parameter_space_cache[] = {
 172	{ "v1", Opt_space_cache_v1 },
 173	{ "v2", Opt_space_cache_v2 },
 174	{}
 175};
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 176
 177enum {
 178	Opt_rescue_usebackuproot,
 179	Opt_rescue_nologreplay,
 180	Opt_rescue_ignorebadroots,
 181	Opt_rescue_ignoredatacsums,
 182	Opt_rescue_parameter_all,
 183};
 184
 185static const struct constant_table btrfs_parameter_rescue[] = {
 186	{ "usebackuproot", Opt_rescue_usebackuproot },
 187	{ "nologreplay", Opt_rescue_nologreplay },
 188	{ "ignorebadroots", Opt_rescue_ignorebadroots },
 189	{ "ibadroots", Opt_rescue_ignorebadroots },
 190	{ "ignoredatacsums", Opt_rescue_ignoredatacsums },
 191	{ "idatacsums", Opt_rescue_ignoredatacsums },
 192	{ "all", Opt_rescue_parameter_all },
 193	{}
 194};
 195
 196#ifdef CONFIG_BTRFS_DEBUG
 197enum {
 198	Opt_fragment_parameter_data,
 199	Opt_fragment_parameter_metadata,
 200	Opt_fragment_parameter_all,
 201};
 202
 203static const struct constant_table btrfs_parameter_fragment[] = {
 204	{ "data", Opt_fragment_parameter_data },
 205	{ "metadata", Opt_fragment_parameter_metadata },
 206	{ "all", Opt_fragment_parameter_all },
 207	{}
 208};
 209#endif
 210
 211static const struct fs_parameter_spec btrfs_fs_parameters[] = {
 212	fsparam_flag_no("acl", Opt_acl),
 213	fsparam_flag_no("autodefrag", Opt_defrag),
 214	fsparam_flag_no("barrier", Opt_barrier),
 215	fsparam_flag("clear_cache", Opt_clear_cache),
 216	fsparam_u32("commit", Opt_commit_interval),
 217	fsparam_flag("compress", Opt_compress),
 218	fsparam_string("compress", Opt_compress_type),
 219	fsparam_flag("compress-force", Opt_compress_force),
 220	fsparam_string("compress-force", Opt_compress_force_type),
 221	fsparam_flag_no("datacow", Opt_datacow),
 222	fsparam_flag_no("datasum", Opt_datasum),
 223	fsparam_flag("degraded", Opt_degraded),
 224	fsparam_string("device", Opt_device),
 225	fsparam_flag_no("discard", Opt_discard),
 226	fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
 227	fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
 228	fsparam_flag_no("flushoncommit", Opt_flushoncommit),
 229	fsparam_string("max_inline", Opt_max_inline),
 230	fsparam_u32("metadata_ratio", Opt_ratio),
 231	fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
 232	fsparam_flag("skip_balance", Opt_skip_balance),
 233	fsparam_flag_no("space_cache", Opt_space_cache),
 234	fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
 235	fsparam_flag_no("ssd", Opt_ssd),
 236	fsparam_flag_no("ssd_spread", Opt_ssd_spread),
 237	fsparam_string("subvol", Opt_subvol),
 238	fsparam_flag("subvol=", Opt_subvol_empty),
 239	fsparam_u64("subvolid", Opt_subvolid),
 240	fsparam_u32("thread_pool", Opt_thread_pool),
 241	fsparam_flag_no("treelog", Opt_treelog),
 242	fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
 243
 244	/* Rescue options. */
 245	fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
 246	/* Deprecated, with alias rescue=nologreplay */
 247	__fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
 248	/* Deprecated, with alias rescue=usebackuproot */
 249	__fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
 
 
 
 250
 251	/* Debugging options. */
 252	fsparam_flag_no("enospc_debug", Opt_enospc_debug),
 
 
 
 
 253#ifdef CONFIG_BTRFS_DEBUG
 254	fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
 
 
 255#endif
 256#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 257	fsparam_flag("ref_verify", Opt_ref_verify),
 258#endif
 259	{}
 
 
 
 
 
 
 
 
 
 
 
 260};
 261
 262/* No support for restricting writes to btrfs devices yet... */
 263static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
 264{
 265	return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
 
 
 
 
 
 266}
 267
 268static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
 269{
 270	struct btrfs_fs_context *ctx = fc->fs_private;
 271	struct fs_parse_result result;
 272	int opt;
 
 
 273
 274	opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
 275	if (opt < 0)
 276		return opt;
 
 277
 278	switch (opt) {
 279	case Opt_degraded:
 280		btrfs_set_opt(ctx->mount_opt, DEGRADED);
 281		break;
 282	case Opt_subvol_empty:
 283		/*
 284		 * This exists because we used to allow it on accident, so we're
 285		 * keeping it to maintain ABI.  See 37becec95ac3 ("Btrfs: allow
 286		 * empty subvol= again").
 287		 */
 288		break;
 289	case Opt_subvol:
 290		kfree(ctx->subvol_name);
 291		ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
 292		if (!ctx->subvol_name)
 293			return -ENOMEM;
 294		break;
 295	case Opt_subvolid:
 296		ctx->subvol_objectid = result.uint_64;
 297
 298		/* subvolid=0 means give me the original fs_tree. */
 299		if (!ctx->subvol_objectid)
 300			ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
 301		break;
 302	case Opt_device: {
 303		struct btrfs_device *device;
 304		blk_mode_t mode = btrfs_open_mode(fc);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 305
 306		mutex_lock(&uuid_mutex);
 307		device = btrfs_scan_one_device(param->string, mode, false);
 308		mutex_unlock(&uuid_mutex);
 309		if (IS_ERR(device))
 310			return PTR_ERR(device);
 311		break;
 312	}
 313	case Opt_datasum:
 314		if (result.negated) {
 315			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 316		} else {
 317			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 318			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 319		}
 320		break;
 321	case Opt_datacow:
 322		if (result.negated) {
 323			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 324			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 325			btrfs_set_opt(ctx->mount_opt, NODATACOW);
 326			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 327		} else {
 328			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 329		}
 330		break;
 331	case Opt_compress_force:
 332	case Opt_compress_force_type:
 333		btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
 334		fallthrough;
 335	case Opt_compress:
 336	case Opt_compress_type:
 337		if (opt == Opt_compress || opt == Opt_compress_force) {
 338			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 339			ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 340			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 341			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 342			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 343		} else if (strncmp(param->string, "zlib", 4) == 0) {
 344			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 345			ctx->compress_level =
 346				btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
 347							 param->string + 4);
 348			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 349			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 350			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 351		} else if (strncmp(param->string, "lzo", 3) == 0) {
 352			ctx->compress_type = BTRFS_COMPRESS_LZO;
 353			ctx->compress_level = 0;
 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, "zstd", 4) == 0) {
 358			ctx->compress_type = BTRFS_COMPRESS_ZSTD;
 359			ctx->compress_level =
 360				btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
 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, "no", 2) == 0) {
 366			ctx->compress_level = 0;
 367			ctx->compress_type = 0;
 368			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 369			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 370		} else {
 371			btrfs_err(NULL, "unrecognized compression value %s",
 372				  param->string);
 373			return -EINVAL;
 374		}
 375		break;
 376	case Opt_ssd:
 377		if (result.negated) {
 378			btrfs_set_opt(ctx->mount_opt, NOSSD);
 379			btrfs_clear_opt(ctx->mount_opt, SSD);
 380			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 381		} else {
 382			btrfs_set_opt(ctx->mount_opt, SSD);
 383			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 384		}
 385		break;
 386	case Opt_ssd_spread:
 387		if (result.negated) {
 388			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 389		} else {
 390			btrfs_set_opt(ctx->mount_opt, SSD);
 391			btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
 392			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 393		}
 394		break;
 395	case Opt_barrier:
 396		if (result.negated)
 397			btrfs_set_opt(ctx->mount_opt, NOBARRIER);
 398		else
 399			btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
 400		break;
 401	case Opt_thread_pool:
 402		if (result.uint_32 == 0) {
 403			btrfs_err(NULL, "invalid value 0 for thread_pool");
 404			return -EINVAL;
 405		}
 406		ctx->thread_pool_size = result.uint_32;
 407		break;
 408	case Opt_max_inline:
 409		ctx->max_inline = memparse(param->string, NULL);
 410		break;
 411	case Opt_acl:
 412		if (result.negated) {
 413			fc->sb_flags &= ~SB_POSIXACL;
 414		} else {
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 415#ifdef CONFIG_BTRFS_FS_POSIX_ACL
 416			fc->sb_flags |= SB_POSIXACL;
 
 417#else
 418			btrfs_err(NULL, "support for ACL not compiled in");
 419			return -EINVAL;
 
 420#endif
 421		}
 422		/*
 423		 * VFS limits the ability to toggle ACL on and off via remount,
 424		 * despite every file system allowing this.  This seems to be
 425		 * an oversight since we all do, but it'll fail if we're
 426		 * remounting.  So don't set the mask here, we'll check it in
 427		 * btrfs_reconfigure and do the toggling ourselves.
 428		 */
 429		if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
 430			fc->sb_flags_mask |= SB_POSIXACL;
 431		break;
 432	case Opt_treelog:
 433		if (result.negated)
 434			btrfs_set_opt(ctx->mount_opt, NOTREELOG);
 435		else
 436			btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
 437		break;
 438	case Opt_nologreplay:
 439		btrfs_warn(NULL,
 440		"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 441		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 442		break;
 443	case Opt_flushoncommit:
 444		if (result.negated)
 445			btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
 446		else
 447			btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
 448		break;
 449	case Opt_ratio:
 450		ctx->metadata_ratio = result.uint_32;
 451		break;
 452	case Opt_discard:
 453		if (result.negated) {
 454			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 455			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 456			btrfs_set_opt(ctx->mount_opt, NODISCARD);
 457		} else {
 458			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 459			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 460		}
 461		break;
 462	case Opt_discard_mode:
 463		switch (result.uint_32) {
 464		case Opt_discard_sync:
 465			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 466			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 467			break;
 468		case Opt_discard_async:
 469			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 470			btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
 471			break;
 472		default:
 473			btrfs_err(NULL, "unrecognized discard mode value %s",
 474				  param->key);
 475			return -EINVAL;
 476		}
 477		btrfs_clear_opt(ctx->mount_opt, NODISCARD);
 478		break;
 479	case Opt_space_cache:
 480		if (result.negated) {
 481			btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
 482			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 483			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 484		} else {
 485			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 486			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 487		}
 488		break;
 489	case Opt_space_cache_version:
 490		switch (result.uint_32) {
 491		case Opt_space_cache_v1:
 492			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 493			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 494			break;
 495		case Opt_space_cache_v2:
 496			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 497			btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
 498			break;
 499		default:
 500			btrfs_err(NULL, "unrecognized space_cache value %s",
 501				  param->key);
 502			return -EINVAL;
 503		}
 504		break;
 505	case Opt_rescan_uuid_tree:
 506		btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
 507		break;
 508	case Opt_clear_cache:
 509		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 510		break;
 511	case Opt_user_subvol_rm_allowed:
 512		btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
 513		break;
 514	case Opt_enospc_debug:
 515		if (result.negated)
 516			btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
 517		else
 518			btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
 519		break;
 520	case Opt_defrag:
 521		if (result.negated)
 522			btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
 523		else
 524			btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
 525		break;
 526	case Opt_usebackuproot:
 527		btrfs_warn(NULL,
 528			   "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
 529		btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 530
 531		/* If we're loading the backup roots we can't trust the space cache. */
 532		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 533		break;
 534	case Opt_skip_balance:
 535		btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
 536		break;
 537	case Opt_fatal_errors:
 538		switch (result.uint_32) {
 539		case Opt_fatal_errors_panic:
 540			btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 541			break;
 542		case Opt_fatal_errors_bug:
 543			btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 
 
 
 
 
 
 
 
 544			break;
 545		default:
 546			btrfs_err(NULL, "unrecognized fatal_errors value %s",
 547				  param->key);
 548			return -EINVAL;
 549		}
 550		break;
 551	case Opt_commit_interval:
 552		ctx->commit_interval = result.uint_32;
 553		if (ctx->commit_interval == 0)
 554			ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 555		break;
 556	case Opt_rescue:
 557		switch (result.uint_32) {
 558		case Opt_rescue_usebackuproot:
 559			btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 560			break;
 561		case Opt_rescue_nologreplay:
 562			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 563			break;
 564		case Opt_rescue_ignorebadroots:
 565			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 566			break;
 567		case Opt_rescue_ignoredatacsums:
 568			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 569			break;
 570		case Opt_rescue_parameter_all:
 571			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 572			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 573			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 574			break;
 575		default:
 576			btrfs_info(NULL, "unrecognized rescue option '%s'",
 577				   param->key);
 578			return -EINVAL;
 579		}
 580		break;
 581#ifdef CONFIG_BTRFS_DEBUG
 582	case Opt_fragment:
 583		switch (result.uint_32) {
 584		case Opt_fragment_parameter_all:
 585			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 586			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 587			break;
 588		case Opt_fragment_parameter_metadata:
 589			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 
 
 590			break;
 591		case Opt_fragment_parameter_data:
 592			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 
 593			break;
 594		default:
 595			btrfs_info(NULL, "unrecognized fragment option '%s'",
 596				   param->key);
 597			return -EINVAL;
 598		}
 599		break;
 600#endif
 601#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 602	case Opt_ref_verify:
 603		btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
 604		break;
 
 605#endif
 606	default:
 607		btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
 608		return -EINVAL;
 
 
 
 
 609	}
 
 
 
 
 610
 611	return 0;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 612}
 613
 614/*
 615 * Some options only have meaning at mount time and shouldn't persist across
 616 * remounts, or be displayed. Clear these at the end of mount and remount code
 617 * paths.
 
 618 */
 619static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
 
 620{
 621	btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
 622	btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
 623	btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
 624}
 625
 626static bool check_ro_option(struct btrfs_fs_info *fs_info,
 627			    unsigned long mount_opt, unsigned long opt,
 628			    const char *opt_name)
 629{
 630	if (mount_opt & opt) {
 631		btrfs_err(fs_info, "%s must be used with ro mount option",
 632			  opt_name);
 633		return true;
 634	}
 635	return false;
 636}
 637
 638bool btrfs_check_options(struct btrfs_fs_info *info, unsigned long *mount_opt,
 639			 unsigned long flags)
 640{
 641	bool ret = true;
 642
 643	if (!(flags & SB_RDONLY) &&
 644	    (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
 645	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
 646	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums")))
 647		ret = false;
 
 
 
 648
 649	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 650	    !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
 651	    !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
 652		btrfs_err(info, "cannot disable free-space-tree");
 653		ret = false;
 654	}
 655	if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
 656	     !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
 657		btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
 658		ret = false;
 659	}
 660
 661	if (btrfs_check_mountopts_zoned(info, mount_opt))
 662		ret = false;
 663
 664	if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
 665		if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE))
 666			btrfs_info(info, "disk space caching is enabled");
 667		if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
 668			btrfs_info(info, "using free-space-tree");
 
 
 
 
 
 
 
 
 
 
 669	}
 670
 671	return ret;
 
 
 672}
 673
 674/*
 675 * This is subtle, we only call this during open_ctree().  We need to pre-load
 676 * the mount options with the on-disk settings.  Before the new mount API took
 677 * effect we would do this on mount and remount.  With the new mount API we'll
 678 * only do this on the initial mount.
 679 *
 680 * This isn't a change in behavior, because we're using the current state of the
 681 * file system to set the current mount options.  If you mounted with special
 682 * options to disable these features and then remounted we wouldn't revert the
 683 * settings, because mounting without these features cleared the on-disk
 684 * settings, so this being called on re-mount is not needed.
 685 */
 686void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
 
 687{
 688	if (fs_info->sectorsize < PAGE_SIZE) {
 689		btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
 690		if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
 691			btrfs_info(fs_info,
 692				   "forcing free space tree for sector size %u with page size %lu",
 693				   fs_info->sectorsize, PAGE_SIZE);
 694			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 695		}
 696	}
 697
 698	/*
 699	 * At this point our mount options are populated, so we only mess with
 700	 * these settings if we don't have any settings already.
 701	 */
 702	if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
 703		return;
 704
 705	if (btrfs_is_zoned(fs_info) &&
 706	    btrfs_free_space_cache_v1_active(fs_info)) {
 707		btrfs_info(fs_info, "zoned: clearing existing space cache");
 708		btrfs_set_super_cache_generation(fs_info->super_copy, 0);
 709		return;
 710	}
 711
 712	if (btrfs_test_opt(fs_info, SPACE_CACHE))
 713		return;
 
 
 714
 715	if (btrfs_test_opt(fs_info, NOSPACECACHE))
 716		return;
 
 
 
 
 
 
 
 
 
 
 
 
 717
 718	/*
 719	 * At this point we don't have explicit options set by the user, set
 720	 * them ourselves based on the state of the file system.
 721	 */
 722	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
 723		btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 724	else if (btrfs_free_space_cache_v1_active(fs_info))
 725		btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
 726}
 727
 728static void set_device_specific_options(struct btrfs_fs_info *fs_info)
 729{
 730	if (!btrfs_test_opt(fs_info, NOSSD) &&
 731	    !fs_info->fs_devices->rotating)
 732		btrfs_set_opt(fs_info->mount_opt, SSD);
 
 733
 734	/*
 735	 * For devices supporting discard turn on discard=async automatically,
 736	 * unless it's already set or disabled. This could be turned off by
 737	 * nodiscard for the same mount.
 738	 *
 739	 * The zoned mode piggy backs on the discard functionality for
 740	 * resetting a zone. There is no reason to delay the zone reset as it is
 741	 * fast enough. So, do not enable async discard for zoned mode.
 742	 */
 743	if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
 744	      btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
 745	      btrfs_test_opt(fs_info, NODISCARD)) &&
 746	    fs_info->fs_devices->discardable &&
 747	    !btrfs_is_zoned(fs_info))
 748		btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
 749}
 750
 751char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 752					  u64 subvol_objectid)
 753{
 754	struct btrfs_root *root = fs_info->tree_root;
 755	struct btrfs_root *fs_root = NULL;
 756	struct btrfs_root_ref *root_ref;
 757	struct btrfs_inode_ref *inode_ref;
 758	struct btrfs_key key;
 759	struct btrfs_path *path = NULL;
 760	char *name = NULL, *ptr;
 761	u64 dirid;
 762	int len;
 763	int ret;
 764
 765	path = btrfs_alloc_path();
 766	if (!path) {
 767		ret = -ENOMEM;
 768		goto err;
 769	}
 770
 771	name = kmalloc(PATH_MAX, GFP_KERNEL);
 772	if (!name) {
 773		ret = -ENOMEM;
 774		goto err;
 775	}
 776	ptr = name + PATH_MAX - 1;
 777	ptr[0] = '\0';
 778
 779	/*
 780	 * Walk up the subvolume trees in the tree of tree roots by root
 781	 * backrefs until we hit the top-level subvolume.
 782	 */
 783	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
 784		key.objectid = subvol_objectid;
 785		key.type = BTRFS_ROOT_BACKREF_KEY;
 786		key.offset = (u64)-1;
 787
 788		ret = btrfs_search_backwards(root, &key, path);
 789		if (ret < 0) {
 790			goto err;
 791		} else if (ret > 0) {
 792			ret = -ENOENT;
 793			goto err;
 794		}
 795
 796		subvol_objectid = key.offset;
 797
 798		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
 799					  struct btrfs_root_ref);
 800		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
 801		ptr -= len + 1;
 802		if (ptr < name) {
 803			ret = -ENAMETOOLONG;
 804			goto err;
 805		}
 806		read_extent_buffer(path->nodes[0], ptr + 1,
 807				   (unsigned long)(root_ref + 1), len);
 808		ptr[0] = '/';
 809		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
 810		btrfs_release_path(path);
 811
 812		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
 813		if (IS_ERR(fs_root)) {
 814			ret = PTR_ERR(fs_root);
 815			fs_root = NULL;
 816			goto err;
 817		}
 818
 819		/*
 820		 * Walk up the filesystem tree by inode refs until we hit the
 821		 * root directory.
 822		 */
 823		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
 824			key.objectid = dirid;
 825			key.type = BTRFS_INODE_REF_KEY;
 826			key.offset = (u64)-1;
 827
 828			ret = btrfs_search_backwards(fs_root, &key, path);
 829			if (ret < 0) {
 830				goto err;
 831			} else if (ret > 0) {
 832				ret = -ENOENT;
 833				goto err;
 834			}
 835
 836			dirid = key.offset;
 837
 838			inode_ref = btrfs_item_ptr(path->nodes[0],
 839						   path->slots[0],
 840						   struct btrfs_inode_ref);
 841			len = btrfs_inode_ref_name_len(path->nodes[0],
 842						       inode_ref);
 843			ptr -= len + 1;
 844			if (ptr < name) {
 845				ret = -ENAMETOOLONG;
 846				goto err;
 847			}
 848			read_extent_buffer(path->nodes[0], ptr + 1,
 849					   (unsigned long)(inode_ref + 1), len);
 850			ptr[0] = '/';
 851			btrfs_release_path(path);
 852		}
 853		btrfs_put_root(fs_root);
 854		fs_root = NULL;
 855	}
 856
 857	btrfs_free_path(path);
 858	if (ptr == name + PATH_MAX - 1) {
 859		name[0] = '/';
 860		name[1] = '\0';
 861	} else {
 862		memmove(name, ptr, name + PATH_MAX - ptr);
 863	}
 864	return name;
 865
 866err:
 867	btrfs_put_root(fs_root);
 868	btrfs_free_path(path);
 869	kfree(name);
 870	return ERR_PTR(ret);
 871}
 872
 873static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
 874{
 875	struct btrfs_root *root = fs_info->tree_root;
 876	struct btrfs_dir_item *di;
 877	struct btrfs_path *path;
 878	struct btrfs_key location;
 879	struct fscrypt_str name = FSTR_INIT("default", 7);
 880	u64 dir_id;
 881
 882	path = btrfs_alloc_path();
 883	if (!path)
 884		return -ENOMEM;
 885
 886	/*
 887	 * Find the "default" dir item which points to the root item that we
 888	 * will mount by default if we haven't been given a specific subvolume
 889	 * to mount.
 890	 */
 891	dir_id = btrfs_super_root_dir(fs_info->super_copy);
 892	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
 893	if (IS_ERR(di)) {
 894		btrfs_free_path(path);
 895		return PTR_ERR(di);
 896	}
 897	if (!di) {
 898		/*
 899		 * Ok the default dir item isn't there.  This is weird since
 900		 * it's always been there, but don't freak out, just try and
 901		 * mount the top-level subvolume.
 902		 */
 903		btrfs_free_path(path);
 904		*objectid = BTRFS_FS_TREE_OBJECTID;
 905		return 0;
 906	}
 907
 908	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 909	btrfs_free_path(path);
 910	*objectid = location.objectid;
 911	return 0;
 912}
 913
 914static int btrfs_fill_super(struct super_block *sb,
 915			    struct btrfs_fs_devices *fs_devices,
 916			    void *data)
 917{
 918	struct inode *inode;
 919	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 920	int err;
 921
 922	sb->s_maxbytes = MAX_LFS_FILESIZE;
 923	sb->s_magic = BTRFS_SUPER_MAGIC;
 924	sb->s_op = &btrfs_super_ops;
 925	sb->s_d_op = &btrfs_dentry_operations;
 926	sb->s_export_op = &btrfs_export_ops;
 927#ifdef CONFIG_FS_VERITY
 928	sb->s_vop = &btrfs_verityops;
 929#endif
 930	sb->s_xattr = btrfs_xattr_handlers;
 931	sb->s_time_gran = 1;
 
 
 
 
 932	sb->s_iflags |= SB_I_CGROUPWB;
 933
 934	err = super_setup_bdi(sb);
 935	if (err) {
 936		btrfs_err(fs_info, "super_setup_bdi failed");
 937		return err;
 938	}
 939
 940	err = open_ctree(sb, fs_devices, (char *)data);
 941	if (err) {
 942		btrfs_err(fs_info, "open_ctree failed");
 943		return err;
 944	}
 945
 946	inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
 947	if (IS_ERR(inode)) {
 948		err = PTR_ERR(inode);
 949		btrfs_handle_fs_error(fs_info, err, NULL);
 950		goto fail_close;
 951	}
 952
 953	sb->s_root = d_make_root(inode);
 954	if (!sb->s_root) {
 955		err = -ENOMEM;
 956		goto fail_close;
 957	}
 958
 959	sb->s_flags |= SB_ACTIVE;
 960	return 0;
 961
 962fail_close:
 963	close_ctree(fs_info);
 964	return err;
 965}
 966
 967int btrfs_sync_fs(struct super_block *sb, int wait)
 968{
 969	struct btrfs_trans_handle *trans;
 970	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 971	struct btrfs_root *root = fs_info->tree_root;
 972
 973	trace_btrfs_sync_fs(fs_info, wait);
 974
 975	if (!wait) {
 976		filemap_flush(fs_info->btree_inode->i_mapping);
 977		return 0;
 978	}
 979
 980	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
 981
 982	trans = btrfs_attach_transaction_barrier(root);
 983	if (IS_ERR(trans)) {
 984		/* no transaction, don't bother */
 985		if (PTR_ERR(trans) == -ENOENT) {
 986			/*
 987			 * Exit unless we have some pending changes
 988			 * that need to go through commit
 989			 */
 990			if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
 991				      &fs_info->flags))
 992				return 0;
 993			/*
 994			 * A non-blocking test if the fs is frozen. We must not
 995			 * start a new transaction here otherwise a deadlock
 996			 * happens. The pending operations are delayed to the
 997			 * next commit after thawing.
 998			 */
 999			if (sb_start_write_trylock(sb))
1000				sb_end_write(sb);
1001			else
1002				return 0;
1003			trans = btrfs_start_transaction(root, 0);
1004		}
1005		if (IS_ERR(trans))
1006			return PTR_ERR(trans);
1007	}
1008	return btrfs_commit_transaction(trans);
1009}
1010
1011static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1012{
1013	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1014	*printed = true;
1015}
1016
1017static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1018{
1019	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1020	const char *compress_type;
1021	const char *subvol_name;
1022	bool printed = false;
1023
1024	if (btrfs_test_opt(info, DEGRADED))
1025		seq_puts(seq, ",degraded");
1026	if (btrfs_test_opt(info, NODATASUM))
1027		seq_puts(seq, ",nodatasum");
1028	if (btrfs_test_opt(info, NODATACOW))
1029		seq_puts(seq, ",nodatacow");
1030	if (btrfs_test_opt(info, NOBARRIER))
1031		seq_puts(seq, ",nobarrier");
1032	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1033		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1034	if (info->thread_pool_size !=  min_t(unsigned long,
1035					     num_online_cpus() + 2, 8))
1036		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1037	if (btrfs_test_opt(info, COMPRESS)) {
1038		compress_type = btrfs_compress_type2str(info->compress_type);
1039		if (btrfs_test_opt(info, FORCE_COMPRESS))
1040			seq_printf(seq, ",compress-force=%s", compress_type);
1041		else
1042			seq_printf(seq, ",compress=%s", compress_type);
1043		if (info->compress_level)
1044			seq_printf(seq, ":%d", info->compress_level);
1045	}
1046	if (btrfs_test_opt(info, NOSSD))
1047		seq_puts(seq, ",nossd");
1048	if (btrfs_test_opt(info, SSD_SPREAD))
1049		seq_puts(seq, ",ssd_spread");
1050	else if (btrfs_test_opt(info, SSD))
1051		seq_puts(seq, ",ssd");
1052	if (btrfs_test_opt(info, NOTREELOG))
1053		seq_puts(seq, ",notreelog");
1054	if (btrfs_test_opt(info, NOLOGREPLAY))
1055		print_rescue_option(seq, "nologreplay", &printed);
1056	if (btrfs_test_opt(info, USEBACKUPROOT))
1057		print_rescue_option(seq, "usebackuproot", &printed);
1058	if (btrfs_test_opt(info, IGNOREBADROOTS))
1059		print_rescue_option(seq, "ignorebadroots", &printed);
1060	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1061		print_rescue_option(seq, "ignoredatacsums", &printed);
1062	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1063		seq_puts(seq, ",flushoncommit");
1064	if (btrfs_test_opt(info, DISCARD_SYNC))
1065		seq_puts(seq, ",discard");
1066	if (btrfs_test_opt(info, DISCARD_ASYNC))
1067		seq_puts(seq, ",discard=async");
1068	if (!(info->sb->s_flags & SB_POSIXACL))
1069		seq_puts(seq, ",noacl");
1070	if (btrfs_free_space_cache_v1_active(info))
1071		seq_puts(seq, ",space_cache");
1072	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1073		seq_puts(seq, ",space_cache=v2");
1074	else
1075		seq_puts(seq, ",nospace_cache");
1076	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1077		seq_puts(seq, ",rescan_uuid_tree");
1078	if (btrfs_test_opt(info, CLEAR_CACHE))
1079		seq_puts(seq, ",clear_cache");
1080	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1081		seq_puts(seq, ",user_subvol_rm_allowed");
1082	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1083		seq_puts(seq, ",enospc_debug");
1084	if (btrfs_test_opt(info, AUTO_DEFRAG))
1085		seq_puts(seq, ",autodefrag");
1086	if (btrfs_test_opt(info, SKIP_BALANCE))
1087		seq_puts(seq, ",skip_balance");
 
 
 
 
 
 
 
 
 
1088	if (info->metadata_ratio)
1089		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1090	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1091		seq_puts(seq, ",fatal_errors=panic");
1092	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1093		seq_printf(seq, ",commit=%u", info->commit_interval);
1094#ifdef CONFIG_BTRFS_DEBUG
1095	if (btrfs_test_opt(info, FRAGMENT_DATA))
1096		seq_puts(seq, ",fragment=data");
1097	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1098		seq_puts(seq, ",fragment=metadata");
1099#endif
1100	if (btrfs_test_opt(info, REF_VERIFY))
1101		seq_puts(seq, ",ref_verify");
1102	seq_printf(seq, ",subvolid=%llu",
1103		  BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1104	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1105			BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1106	if (!IS_ERR(subvol_name)) {
1107		seq_puts(seq, ",subvol=");
1108		seq_escape(seq, subvol_name, " \t\n\\");
1109		kfree(subvol_name);
1110	}
1111	return 0;
1112}
1113
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1114/*
1115 * subvolumes are identified by ino 256
1116 */
1117static inline int is_subvolume_inode(struct inode *inode)
1118{
1119	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1120		return 1;
1121	return 0;
1122}
1123
1124static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1125				   struct vfsmount *mnt)
1126{
1127	struct dentry *root;
1128	int ret;
1129
1130	if (!subvol_name) {
1131		if (!subvol_objectid) {
1132			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1133							  &subvol_objectid);
1134			if (ret) {
1135				root = ERR_PTR(ret);
1136				goto out;
1137			}
1138		}
1139		subvol_name = btrfs_get_subvol_name_from_objectid(
1140					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1141		if (IS_ERR(subvol_name)) {
1142			root = ERR_CAST(subvol_name);
1143			subvol_name = NULL;
1144			goto out;
1145		}
1146
1147	}
1148
1149	root = mount_subtree(mnt, subvol_name);
1150	/* mount_subtree() drops our reference on the vfsmount. */
1151	mnt = NULL;
1152
1153	if (!IS_ERR(root)) {
1154		struct super_block *s = root->d_sb;
1155		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1156		struct inode *root_inode = d_inode(root);
1157		u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1158
1159		ret = 0;
1160		if (!is_subvolume_inode(root_inode)) {
1161			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1162			       subvol_name);
1163			ret = -EINVAL;
1164		}
1165		if (subvol_objectid && root_objectid != subvol_objectid) {
1166			/*
1167			 * This will also catch a race condition where a
1168			 * subvolume which was passed by ID is renamed and
1169			 * another subvolume is renamed over the old location.
1170			 */
1171			btrfs_err(fs_info,
1172				  "subvol '%s' does not match subvolid %llu",
1173				  subvol_name, subvol_objectid);
1174			ret = -EINVAL;
1175		}
1176		if (ret) {
1177			dput(root);
1178			root = ERR_PTR(ret);
1179			deactivate_locked_super(s);
1180		}
1181	}
1182
1183out:
1184	mntput(mnt);
1185	kfree(subvol_name);
1186	return root;
1187}
1188
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1189static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1190				     u32 new_pool_size, u32 old_pool_size)
1191{
1192	if (new_pool_size == old_pool_size)
1193		return;
1194
1195	fs_info->thread_pool_size = new_pool_size;
1196
1197	btrfs_info(fs_info, "resize thread pool %d -> %d",
1198	       old_pool_size, new_pool_size);
1199
1200	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
 
1201	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1202	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1203	workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1204	workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1205	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1206	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1207	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1208}
1209
1210static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1211				       unsigned long old_opts, int flags)
1212{
1213	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1214	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1215	     (flags & SB_RDONLY))) {
1216		/* wait for any defraggers to finish */
1217		wait_event(fs_info->transaction_wait,
1218			   (atomic_read(&fs_info->defrag_running) == 0));
1219		if (flags & SB_RDONLY)
1220			sync_filesystem(fs_info->sb);
1221	}
1222}
1223
1224static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1225					 unsigned long old_opts)
1226{
1227	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1228
1229	/*
1230	 * We need to cleanup all defragable inodes if the autodefragment is
1231	 * close or the filesystem is read only.
1232	 */
1233	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1234	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1235		btrfs_cleanup_defrag_inodes(fs_info);
1236	}
1237
1238	/* If we toggled discard async */
1239	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1240	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1241		btrfs_discard_resume(fs_info);
1242	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1243		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1244		btrfs_discard_cleanup(fs_info);
1245
1246	/* If we toggled space cache */
1247	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1248		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1249}
1250
1251static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1252{
 
 
 
 
 
 
 
1253	int ret;
1254
1255	if (BTRFS_FS_ERROR(fs_info)) {
1256		btrfs_err(fs_info,
1257			  "remounting read-write after error is not allowed");
1258		return -EINVAL;
1259	}
1260
1261	if (fs_info->fs_devices->rw_devices == 0)
1262		return -EACCES;
1263
1264	if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1265		btrfs_warn(fs_info,
1266			   "too many missing devices, writable remount is not allowed");
1267		return -EACCES;
 
 
1268	}
1269
1270	if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1271		btrfs_warn(fs_info,
1272			   "mount required to replay tree-log, cannot remount read-write");
1273		return -EINVAL;
1274	}
1275
1276	/*
1277	 * NOTE: when remounting with a change that does writes, don't put it
1278	 * anywhere above this point, as we are not sure to be safe to write
1279	 * until we pass the above checks.
1280	 */
1281	ret = btrfs_start_pre_rw_mount(fs_info);
1282	if (ret)
1283		return ret;
1284
1285	btrfs_clear_sb_rdonly(fs_info->sb);
1286
1287	set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1288
1289	/*
1290	 * If we've gone from readonly -> read-write, we need to get our
1291	 * sync/async discard lists in the right state.
1292	 */
1293	btrfs_discard_resume(fs_info);
1294
1295	return 0;
1296}
1297
1298static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1299{
1300	/*
1301	 * This also happens on 'umount -rf' or on shutdown, when the
1302	 * filesystem is busy.
1303	 */
1304	cancel_work_sync(&fs_info->async_reclaim_work);
1305	cancel_work_sync(&fs_info->async_data_reclaim_work);
1306
1307	btrfs_discard_cleanup(fs_info);
 
 
1308
1309	/* Wait for the uuid_scan task to finish */
1310	down(&fs_info->uuid_tree_rescan_sem);
1311	/* Avoid complains from lockdep et al. */
1312	up(&fs_info->uuid_tree_rescan_sem);
1313
1314	btrfs_set_sb_rdonly(fs_info->sb);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1315
1316	/*
1317	 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1318	 * loop if it's already active.  If it's already asleep, we'll leave
1319	 * unused block groups on disk until we're mounted read-write again
1320	 * unless we clean them up here.
1321	 */
1322	btrfs_delete_unused_bgs(fs_info);
1323
1324	/*
1325	 * The cleaner task could be already running before we set the flag
1326	 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).  We must make
1327	 * sure that after we finish the remount, i.e. after we call
1328	 * btrfs_commit_super(), the cleaner can no longer start a transaction
1329	 * - either because it was dropping a dead root, running delayed iputs
1330	 *   or deleting an unused block group (the cleaner picked a block
1331	 *   group from the list of unused block groups before we were able to
1332	 *   in the previous call to btrfs_delete_unused_bgs()).
1333	 */
1334	wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1335
1336	/*
1337	 * We've set the superblock to RO mode, so we might have made the
1338	 * cleaner task sleep without running all pending delayed iputs. Go
1339	 * through all the delayed iputs here, so that if an unmount happens
1340	 * without remounting RW we don't end up at finishing close_ctree()
1341	 * with a non-empty list of delayed iputs.
1342	 */
1343	btrfs_run_delayed_iputs(fs_info);
1344
1345	btrfs_dev_replace_suspend_for_unmount(fs_info);
1346	btrfs_scrub_cancel(fs_info);
1347	btrfs_pause_balance(fs_info);
 
1348
1349	/*
1350	 * Pause the qgroup rescan worker if it is running. We don't want it to
1351	 * be still running after we are in RO mode, as after that, by the time
1352	 * we unmount, it might have left a transaction open, so we would leak
1353	 * the transaction and/or crash.
1354	 */
1355	btrfs_qgroup_wait_for_completion(fs_info, false);
1356
1357	return btrfs_commit_super(fs_info);
1358}
 
 
 
 
 
 
1359
1360static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1361{
1362	fs_info->max_inline = ctx->max_inline;
1363	fs_info->commit_interval = ctx->commit_interval;
1364	fs_info->metadata_ratio = ctx->metadata_ratio;
1365	fs_info->thread_pool_size = ctx->thread_pool_size;
1366	fs_info->mount_opt = ctx->mount_opt;
1367	fs_info->compress_type = ctx->compress_type;
1368	fs_info->compress_level = ctx->compress_level;
1369}
1370
1371static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1372{
1373	ctx->max_inline = fs_info->max_inline;
1374	ctx->commit_interval = fs_info->commit_interval;
1375	ctx->metadata_ratio = fs_info->metadata_ratio;
1376	ctx->thread_pool_size = fs_info->thread_pool_size;
1377	ctx->mount_opt = fs_info->mount_opt;
1378	ctx->compress_type = fs_info->compress_type;
1379	ctx->compress_level = fs_info->compress_level;
1380}
1381
1382#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...)			\
1383do {										\
1384	if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1385	    btrfs_raw_test_opt(fs_info->mount_opt, opt))			\
1386		btrfs_info(fs_info, fmt, ##args);				\
1387} while (0)
1388
1389#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...)	\
1390do {									\
1391	if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1392	    !btrfs_raw_test_opt(fs_info->mount_opt, opt))		\
1393		btrfs_info(fs_info, fmt, ##args);			\
1394} while (0)
1395
1396static void btrfs_emit_options(struct btrfs_fs_info *info,
1397			       struct btrfs_fs_context *old)
1398{
1399	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1400	btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1401	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1402	btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1403	btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1404	btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1405	btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1406	btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1407	btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1408	btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1409	btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1410	btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1411	btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1412	btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1413	btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1414	btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1415	btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1416	btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1417	btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1418	btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1419	btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1420
1421	btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1422	btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1423	btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1424	btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1425	btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1426	btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1427	btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1428	btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1429	btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1430
1431	/* Did the compression settings change? */
1432	if (btrfs_test_opt(info, COMPRESS) &&
1433	    (!old ||
1434	     old->compress_type != info->compress_type ||
1435	     old->compress_level != info->compress_level ||
1436	     (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1437	      btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1438		const char *compress_type = btrfs_compress_type2str(info->compress_type);
1439
1440		btrfs_info(info, "%s %s compression, level %d",
1441			   btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1442			   compress_type, info->compress_level);
1443	}
1444
1445	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1446		btrfs_info(info, "max_inline set to %llu", info->max_inline);
1447}
 
 
 
 
 
 
1448
1449static int btrfs_reconfigure(struct fs_context *fc)
1450{
1451	struct super_block *sb = fc->root->d_sb;
1452	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1453	struct btrfs_fs_context *ctx = fc->fs_private;
1454	struct btrfs_fs_context old_ctx;
1455	int ret = 0;
1456	bool mount_reconfigure = (fc->s_fs_info != NULL);
1457
1458	btrfs_info_to_ctx(fs_info, &old_ctx);
 
 
 
 
 
 
1459
1460	/*
1461	 * This is our "bind mount" trick, we don't want to allow the user to do
1462	 * anything other than mount a different ro/rw and a different subvol,
1463	 * all of the mount options should be maintained.
1464	 */
1465	if (mount_reconfigure)
1466		ctx->mount_opt = old_ctx.mount_opt;
 
 
 
 
 
 
 
1467
1468	sync_filesystem(sb);
1469	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
 
 
 
 
1470
1471	if (!mount_reconfigure &&
1472	    !btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1473		return -EINVAL;
 
 
 
1474
1475	ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1476	if (ret < 0)
1477		return ret;
 
 
 
 
 
1478
1479	btrfs_ctx_to_info(fs_info, ctx);
1480	btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1481	btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1482				 old_ctx.thread_pool_size);
1483
1484	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1485	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1486	    (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1487		btrfs_warn(fs_info,
1488		"remount supports changing free space tree only from RO to RW");
1489		/* Make sure free space cache options match the state on disk. */
1490		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1491			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1492			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1493		}
1494		if (btrfs_free_space_cache_v1_active(fs_info)) {
1495			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1496			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1497		}
1498	}
1499
1500	ret = 0;
1501	if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1502		ret = btrfs_remount_ro(fs_info);
1503	else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1504		ret = btrfs_remount_rw(fs_info);
1505	if (ret)
1506		goto restore;
1507
1508	/*
1509	 * If we set the mask during the parameter parsing VFS would reject the
1510	 * remount.  Here we can set the mask and the value will be updated
1511	 * appropriately.
1512	 */
1513	if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1514		fc->sb_flags_mask |= SB_POSIXACL;
1515
1516	btrfs_emit_options(fs_info, &old_ctx);
1517	wake_up_process(fs_info->transaction_kthread);
1518	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1519	btrfs_clear_oneshot_options(fs_info);
1520	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1521
1522	return 0;
 
1523restore:
1524	btrfs_ctx_to_info(fs_info, &old_ctx);
1525	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
 
 
 
 
 
 
 
 
 
 
 
1526	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
 
1527	return ret;
1528}
1529
1530/* Used to sort the devices by max_avail(descending sort) */
1531static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1532{
1533	const struct btrfs_device_info *dev_info1 = a;
1534	const struct btrfs_device_info *dev_info2 = b;
1535
1536	if (dev_info1->max_avail > dev_info2->max_avail)
1537		return -1;
1538	else if (dev_info1->max_avail < dev_info2->max_avail)
1539		return 1;
1540	return 0;
1541}
1542
1543/*
1544 * sort the devices by max_avail, in which max free extent size of each device
1545 * is stored.(Descending Sort)
1546 */
1547static inline void btrfs_descending_sort_devices(
1548					struct btrfs_device_info *devices,
1549					size_t nr_devices)
1550{
1551	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1552	     btrfs_cmp_device_free_bytes, NULL);
1553}
1554
1555/*
1556 * The helper to calc the free space on the devices that can be used to store
1557 * file data.
1558 */
1559static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1560					      u64 *free_bytes)
1561{
1562	struct btrfs_device_info *devices_info;
1563	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1564	struct btrfs_device *device;
1565	u64 type;
1566	u64 avail_space;
1567	u64 min_stripe_size;
1568	int num_stripes = 1;
1569	int i = 0, nr_devices;
1570	const struct btrfs_raid_attr *rattr;
1571
1572	/*
1573	 * We aren't under the device list lock, so this is racy-ish, but good
1574	 * enough for our purposes.
1575	 */
1576	nr_devices = fs_info->fs_devices->open_devices;
1577	if (!nr_devices) {
1578		smp_mb();
1579		nr_devices = fs_info->fs_devices->open_devices;
1580		ASSERT(nr_devices);
1581		if (!nr_devices) {
1582			*free_bytes = 0;
1583			return 0;
1584		}
1585	}
1586
1587	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1588			       GFP_KERNEL);
1589	if (!devices_info)
1590		return -ENOMEM;
1591
1592	/* calc min stripe number for data space allocation */
1593	type = btrfs_data_alloc_profile(fs_info);
1594	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1595
1596	if (type & BTRFS_BLOCK_GROUP_RAID0)
1597		num_stripes = nr_devices;
1598	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1599		num_stripes = rattr->ncopies;
1600	else if (type & BTRFS_BLOCK_GROUP_RAID10)
1601		num_stripes = 4;
1602
1603	/* Adjust for more than 1 stripe per device */
1604	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1605
1606	rcu_read_lock();
1607	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1608		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1609						&device->dev_state) ||
1610		    !device->bdev ||
1611		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1612			continue;
1613
1614		if (i >= nr_devices)
1615			break;
1616
1617		avail_space = device->total_bytes - device->bytes_used;
1618
1619		/* align with stripe_len */
1620		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1621
1622		/*
1623		 * Ensure we have at least min_stripe_size on top of the
1624		 * reserved space on the device.
1625		 */
1626		if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1627			continue;
1628
1629		avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1630
1631		devices_info[i].dev = device;
1632		devices_info[i].max_avail = avail_space;
1633
1634		i++;
1635	}
1636	rcu_read_unlock();
1637
1638	nr_devices = i;
1639
1640	btrfs_descending_sort_devices(devices_info, nr_devices);
1641
1642	i = nr_devices - 1;
1643	avail_space = 0;
1644	while (nr_devices >= rattr->devs_min) {
1645		num_stripes = min(num_stripes, nr_devices);
1646
1647		if (devices_info[i].max_avail >= min_stripe_size) {
1648			int j;
1649			u64 alloc_size;
1650
1651			avail_space += devices_info[i].max_avail * num_stripes;
1652			alloc_size = devices_info[i].max_avail;
1653			for (j = i + 1 - num_stripes; j <= i; j++)
1654				devices_info[j].max_avail -= alloc_size;
1655		}
1656		i--;
1657		nr_devices--;
1658	}
1659
1660	kfree(devices_info);
1661	*free_bytes = avail_space;
1662	return 0;
1663}
1664
1665/*
1666 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1667 *
1668 * If there's a redundant raid level at DATA block groups, use the respective
1669 * multiplier to scale the sizes.
1670 *
1671 * Unused device space usage is based on simulating the chunk allocator
1672 * algorithm that respects the device sizes and order of allocations.  This is
1673 * a close approximation of the actual use but there are other factors that may
1674 * change the result (like a new metadata chunk).
1675 *
1676 * If metadata is exhausted, f_bavail will be 0.
1677 */
1678static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1679{
1680	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1681	struct btrfs_super_block *disk_super = fs_info->super_copy;
1682	struct btrfs_space_info *found;
1683	u64 total_used = 0;
1684	u64 total_free_data = 0;
1685	u64 total_free_meta = 0;
1686	u32 bits = fs_info->sectorsize_bits;
1687	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1688	unsigned factor = 1;
1689	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1690	int ret;
1691	u64 thresh = 0;
1692	int mixed = 0;
1693
1694	list_for_each_entry(found, &fs_info->space_info, list) {
1695		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1696			int i;
1697
1698			total_free_data += found->disk_total - found->disk_used;
1699			total_free_data -=
1700				btrfs_account_ro_block_groups_free_space(found);
1701
1702			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1703				if (!list_empty(&found->block_groups[i]))
1704					factor = btrfs_bg_type_to_factor(
1705						btrfs_raid_array[i].bg_flag);
1706			}
1707		}
1708
1709		/*
1710		 * Metadata in mixed block group profiles are accounted in data
1711		 */
1712		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1713			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1714				mixed = 1;
1715			else
1716				total_free_meta += found->disk_total -
1717					found->disk_used;
1718		}
1719
1720		total_used += found->disk_used;
1721	}
1722
1723	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1724	buf->f_blocks >>= bits;
1725	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1726
1727	/* Account global block reserve as used, it's in logical size already */
1728	spin_lock(&block_rsv->lock);
1729	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
1730	if (buf->f_bfree >= block_rsv->size >> bits)
1731		buf->f_bfree -= block_rsv->size >> bits;
1732	else
1733		buf->f_bfree = 0;
1734	spin_unlock(&block_rsv->lock);
1735
1736	buf->f_bavail = div_u64(total_free_data, factor);
1737	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1738	if (ret)
1739		return ret;
1740	buf->f_bavail += div_u64(total_free_data, factor);
1741	buf->f_bavail = buf->f_bavail >> bits;
1742
1743	/*
1744	 * We calculate the remaining metadata space minus global reserve. If
1745	 * this is (supposedly) smaller than zero, there's no space. But this
1746	 * does not hold in practice, the exhausted state happens where's still
1747	 * some positive delta. So we apply some guesswork and compare the
1748	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
1749	 *
1750	 * We probably cannot calculate the exact threshold value because this
1751	 * depends on the internal reservations requested by various
1752	 * operations, so some operations that consume a few metadata will
1753	 * succeed even if the Avail is zero. But this is better than the other
1754	 * way around.
1755	 */
1756	thresh = SZ_4M;
1757
1758	/*
1759	 * We only want to claim there's no available space if we can no longer
1760	 * allocate chunks for our metadata profile and our global reserve will
1761	 * not fit in the free metadata space.  If we aren't ->full then we
1762	 * still can allocate chunks and thus are fine using the currently
1763	 * calculated f_bavail.
1764	 */
1765	if (!mixed && block_rsv->space_info->full &&
1766	    (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1767		buf->f_bavail = 0;
1768
1769	buf->f_type = BTRFS_SUPER_MAGIC;
1770	buf->f_bsize = dentry->d_sb->s_blocksize;
1771	buf->f_namelen = BTRFS_NAME_LEN;
1772
1773	/* We treat it as constant endianness (it doesn't matter _which_)
1774	   because we want the fsid to come out the same whether mounted
1775	   on a big-endian or little-endian host */
1776	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1777	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1778	/* Mask in the root object ID too, to disambiguate subvols */
1779	buf->f_fsid.val[0] ^=
1780		BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
1781	buf->f_fsid.val[1] ^=
1782		BTRFS_I(d_inode(dentry))->root->root_key.objectid;
1783
1784	return 0;
1785}
1786
1787static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1788{
1789	struct btrfs_fs_info *p = fc->s_fs_info;
1790	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1791
1792	return fs_info->fs_devices == p->fs_devices;
1793}
1794
1795static int btrfs_get_tree_super(struct fs_context *fc)
1796{
1797	struct btrfs_fs_info *fs_info = fc->s_fs_info;
1798	struct btrfs_fs_context *ctx = fc->fs_private;
1799	struct btrfs_fs_devices *fs_devices = NULL;
1800	struct block_device *bdev;
1801	struct btrfs_device *device;
1802	struct super_block *sb;
1803	blk_mode_t mode = btrfs_open_mode(fc);
1804	int ret;
1805
1806	btrfs_ctx_to_info(fs_info, ctx);
1807	mutex_lock(&uuid_mutex);
1808
1809	/*
1810	 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1811	 * either a valid device or an error.
1812	 */
1813	device = btrfs_scan_one_device(fc->source, mode, true);
1814	ASSERT(device != NULL);
1815	if (IS_ERR(device)) {
1816		mutex_unlock(&uuid_mutex);
1817		return PTR_ERR(device);
1818	}
1819
1820	fs_devices = device->fs_devices;
1821	fs_info->fs_devices = fs_devices;
1822
1823	ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1824	mutex_unlock(&uuid_mutex);
1825	if (ret)
1826		return ret;
1827
1828	if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1829		ret = -EACCES;
1830		goto error;
1831	}
1832
1833	bdev = fs_devices->latest_dev->bdev;
1834
1835	/*
1836	 * From now on the error handling is not straightforward.
1837	 *
1838	 * If successful, this will transfer the fs_info into the super block,
1839	 * and fc->s_fs_info will be NULL.  However if there's an existing
1840	 * super, we'll still have fc->s_fs_info populated.  If we error
1841	 * completely out it'll be cleaned up when we drop the fs_context,
1842	 * otherwise it's tied to the lifetime of the super_block.
1843	 */
1844	sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1845	if (IS_ERR(sb)) {
1846		ret = PTR_ERR(sb);
1847		goto error;
1848	}
1849
1850	set_device_specific_options(fs_info);
1851
1852	if (sb->s_root) {
1853		btrfs_close_devices(fs_devices);
1854		if ((fc->sb_flags ^ sb->s_flags) & SB_RDONLY)
1855			ret = -EBUSY;
1856	} else {
1857		snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1858		shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1859		btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1860		ret = btrfs_fill_super(sb, fs_devices, NULL);
1861	}
1862
1863	if (ret) {
1864		deactivate_locked_super(sb);
1865		return ret;
1866	}
1867
1868	btrfs_clear_oneshot_options(fs_info);
1869
1870	fc->root = dget(sb->s_root);
1871	return 0;
1872
1873error:
1874	btrfs_close_devices(fs_devices);
1875	return ret;
1876}
1877
1878/*
1879 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1880 * with different ro/rw options") the following works:
1881 *
1882 *        (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1883 *       (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1884 *
1885 * which looks nice and innocent but is actually pretty intricate and deserves
1886 * a long comment.
1887 *
1888 * On another filesystem a subvolume mount is close to something like:
1889 *
1890 *	(iii) # create rw superblock + initial mount
1891 *	      mount -t xfs /dev/sdb /opt/
1892 *
1893 *	      # create ro bind mount
1894 *	      mount --bind -o ro /opt/foo /mnt/foo
1895 *
1896 *	      # unmount initial mount
1897 *	      umount /opt
1898 *
1899 * Of course, there's some special subvolume sauce and there's the fact that the
1900 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1901 * it's very close and will help us understand the issue.
1902 *
1903 * The old mount API didn't cleanly distinguish between a mount being made ro
1904 * and a superblock being made ro.  The only way to change the ro state of
1905 * either object was by passing ms_rdonly. If a new mount was created via
1906 * mount(2) such as:
1907 *
1908 *      mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1909 *
1910 * the MS_RDONLY flag being specified had two effects:
1911 *
1912 * (1) MNT_READONLY was raised -> the resulting mount got
1913 *     @mnt->mnt_flags |= MNT_READONLY raised.
1914 *
1915 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1916 *     made the superblock ro. Note, how SB_RDONLY has the same value as
1917 *     ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1918 *
1919 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1920 * subtree mounted ro.
1921 *
1922 * But consider the effect on the old mount API on btrfs subvolume mounting
1923 * which combines the distinct step in (iii) into a single step.
1924 *
1925 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1926 * is issued the superblock is ro and thus even if the mount created for (ii) is
1927 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1928 * to rw for (ii) which it did using an internal remount call.
1929 *
1930 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1931 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1932 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1933 * passed by mount(8) to mount(2).
1934 *
1935 * Enter the new mount API. The new mount API disambiguates making a mount ro
1936 * and making a superblock ro.
1937 *
1938 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1939 *     fsmount() or mount_setattr() this is a pure VFS level change for a
1940 *     specific mount or mount tree that is never seen by the filesystem itself.
1941 *
1942 * (4) To turn a superblock ro the "ro" flag must be used with
1943 *     fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1944 *     in fc->sb_flags.
1945 *
1946 * This disambiguation has rather positive consequences.  Mounting a subvolume
1947 * ro will not also turn the superblock ro. Only the mount for the subvolume
1948 * will become ro.
1949 *
1950 * So, if the superblock creation request comes from the new mount API the
1951 * caller must have explicitly done:
1952 *
1953 *      fsconfig(FSCONFIG_SET_FLAG, "ro")
1954 *      fsmount/mount_setattr(MOUNT_ATTR_RDONLY)
1955 *
1956 * IOW, at some point the caller must have explicitly turned the whole
1957 * superblock ro and we shouldn't just undo it like we did for the old mount
1958 * API. In any case, it lets us avoid the hack in the new mount API.
1959 *
1960 * Consequently, the remounting hack must only be used for requests originating
1961 * from the old mount API and should be marked for full deprecation so it can be
1962 * turned off in a couple of years.
1963 *
1964 * The new mount API has no reason to support this hack.
1965 */
1966static struct vfsmount *btrfs_reconfigure_for_mount(struct fs_context *fc)
1967{
1968	struct vfsmount *mnt;
1969	int ret;
1970	const bool ro2rw = !(fc->sb_flags & SB_RDONLY);
1971
1972	/*
1973	 * We got an EBUSY because our SB_RDONLY flag didn't match the existing
1974	 * super block, so invert our setting here and retry the mount so we
1975	 * can get our vfsmount.
1976	 */
1977	if (ro2rw)
1978		fc->sb_flags |= SB_RDONLY;
1979	else
1980		fc->sb_flags &= ~SB_RDONLY;
1981
1982	mnt = fc_mount(fc);
1983	if (IS_ERR(mnt))
1984		return mnt;
1985
1986	if (!fc->oldapi || !ro2rw)
1987		return mnt;
1988
1989	/* We need to convert to rw, call reconfigure. */
1990	fc->sb_flags &= ~SB_RDONLY;
1991	down_write(&mnt->mnt_sb->s_umount);
1992	ret = btrfs_reconfigure(fc);
1993	up_write(&mnt->mnt_sb->s_umount);
1994	if (ret) {
1995		mntput(mnt);
1996		return ERR_PTR(ret);
1997	}
1998	return mnt;
1999}
2000
2001static int btrfs_get_tree_subvol(struct fs_context *fc)
2002{
2003	struct btrfs_fs_info *fs_info = NULL;
2004	struct btrfs_fs_context *ctx = fc->fs_private;
2005	struct fs_context *dup_fc;
2006	struct dentry *dentry;
2007	struct vfsmount *mnt;
2008
2009	/*
2010	 * Setup a dummy root and fs_info for test/set super.  This is because
2011	 * we don't actually fill this stuff out until open_ctree, but we need
2012	 * then open_ctree will properly initialize the file system specific
2013	 * settings later.  btrfs_init_fs_info initializes the static elements
2014	 * of the fs_info (locks and such) to make cleanup easier if we find a
2015	 * superblock with our given fs_devices later on at sget() time.
2016	 */
2017	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2018	if (!fs_info)
2019		return -ENOMEM;
2020
2021	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2022	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2023	if (!fs_info->super_copy || !fs_info->super_for_commit) {
2024		btrfs_free_fs_info(fs_info);
2025		return -ENOMEM;
2026	}
2027	btrfs_init_fs_info(fs_info);
2028
2029	dup_fc = vfs_dup_fs_context(fc);
2030	if (IS_ERR(dup_fc)) {
2031		btrfs_free_fs_info(fs_info);
2032		return PTR_ERR(dup_fc);
2033	}
2034
2035	/*
2036	 * When we do the sget_fc this gets transferred to the sb, so we only
2037	 * need to set it on the dup_fc as this is what creates the super block.
2038	 */
2039	dup_fc->s_fs_info = fs_info;
2040
2041	/*
2042	 * We'll do the security settings in our btrfs_get_tree_super() mount
2043	 * loop, they were duplicated into dup_fc, we can drop the originals
2044	 * here.
2045	 */
2046	security_free_mnt_opts(&fc->security);
2047	fc->security = NULL;
2048
2049	mnt = fc_mount(dup_fc);
2050	if (PTR_ERR_OR_ZERO(mnt) == -EBUSY)
2051		mnt = btrfs_reconfigure_for_mount(dup_fc);
2052	put_fs_context(dup_fc);
2053	if (IS_ERR(mnt))
2054		return PTR_ERR(mnt);
2055
2056	/*
2057	 * This free's ->subvol_name, because if it isn't set we have to
2058	 * allocate a buffer to hold the subvol_name, so we just drop our
2059	 * reference to it here.
2060	 */
2061	dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2062	ctx->subvol_name = NULL;
2063	if (IS_ERR(dentry))
2064		return PTR_ERR(dentry);
2065
2066	fc->root = dentry;
2067	return 0;
2068}
2069
2070static int btrfs_get_tree(struct fs_context *fc)
2071{
2072	/*
2073	 * Since we use mount_subtree to mount the default/specified subvol, we
2074	 * have to do mounts in two steps.
2075	 *
2076	 * First pass through we call btrfs_get_tree_subvol(), this is just a
2077	 * wrapper around fc_mount() to call back into here again, and this time
2078	 * we'll call btrfs_get_tree_super().  This will do the open_ctree() and
2079	 * everything to open the devices and file system.  Then we return back
2080	 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2081	 * from there we can do our mount_subvol() call, which will lookup
2082	 * whichever subvol we're mounting and setup this fc with the
2083	 * appropriate dentry for the subvol.
2084	 */
2085	if (fc->s_fs_info)
2086		return btrfs_get_tree_super(fc);
2087	return btrfs_get_tree_subvol(fc);
2088}
2089
2090static void btrfs_kill_super(struct super_block *sb)
2091{
2092	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2093	kill_anon_super(sb);
2094	btrfs_free_fs_info(fs_info);
2095}
2096
2097static void btrfs_free_fs_context(struct fs_context *fc)
2098{
2099	struct btrfs_fs_context *ctx = fc->fs_private;
2100	struct btrfs_fs_info *fs_info = fc->s_fs_info;
2101
2102	if (fs_info)
2103		btrfs_free_fs_info(fs_info);
2104
2105	if (ctx && refcount_dec_and_test(&ctx->refs)) {
2106		kfree(ctx->subvol_name);
2107		kfree(ctx);
2108	}
2109}
2110
2111static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2112{
2113	struct btrfs_fs_context *ctx = src_fc->fs_private;
2114
2115	/*
2116	 * Give a ref to our ctx to this dup, as we want to keep it around for
2117	 * our original fc so we can have the subvolume name or objectid.
2118	 *
2119	 * We unset ->source in the original fc because the dup needs it for
2120	 * mounting, and then once we free the dup it'll free ->source, so we
2121	 * need to make sure we're only pointing to it in one fc.
2122	 */
2123	refcount_inc(&ctx->refs);
2124	fc->fs_private = ctx;
2125	fc->source = src_fc->source;
2126	src_fc->source = NULL;
2127	return 0;
2128}
2129
2130static const struct fs_context_operations btrfs_fs_context_ops = {
2131	.parse_param	= btrfs_parse_param,
2132	.reconfigure	= btrfs_reconfigure,
2133	.get_tree	= btrfs_get_tree,
2134	.dup		= btrfs_dup_fs_context,
2135	.free		= btrfs_free_fs_context,
2136};
2137
2138static int btrfs_init_fs_context(struct fs_context *fc)
2139{
2140	struct btrfs_fs_context *ctx;
2141
2142	ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2143	if (!ctx)
2144		return -ENOMEM;
2145
2146	refcount_set(&ctx->refs, 1);
2147	fc->fs_private = ctx;
2148	fc->ops = &btrfs_fs_context_ops;
2149
2150	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2151		btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2152	} else {
2153		ctx->thread_pool_size =
2154			min_t(unsigned long, num_online_cpus() + 2, 8);
2155		ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2156		ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2157	}
2158
2159#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2160	fc->sb_flags |= SB_POSIXACL;
2161#endif
2162	fc->sb_flags |= SB_I_VERSION;
2163
2164	return 0;
2165}
2166
2167static struct file_system_type btrfs_fs_type = {
2168	.owner			= THIS_MODULE,
2169	.name			= "btrfs",
2170	.init_fs_context	= btrfs_init_fs_context,
2171	.parameters		= btrfs_fs_parameters,
2172	.kill_sb		= btrfs_kill_super,
2173	.fs_flags		= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2174 };
2175
2176MODULE_ALIAS_FS("btrfs");
2177
2178static int btrfs_control_open(struct inode *inode, struct file *file)
2179{
2180	/*
2181	 * The control file's private_data is used to hold the
2182	 * transaction when it is started and is used to keep
2183	 * track of whether a transaction is already in progress.
2184	 */
2185	file->private_data = NULL;
2186	return 0;
2187}
2188
2189/*
2190 * Used by /dev/btrfs-control for devices ioctls.
2191 */
2192static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2193				unsigned long arg)
2194{
2195	struct btrfs_ioctl_vol_args *vol;
2196	struct btrfs_device *device = NULL;
2197	dev_t devt = 0;
2198	int ret = -ENOTTY;
2199
2200	if (!capable(CAP_SYS_ADMIN))
2201		return -EPERM;
2202
2203	vol = memdup_user((void __user *)arg, sizeof(*vol));
2204	if (IS_ERR(vol))
2205		return PTR_ERR(vol);
2206	vol->name[BTRFS_PATH_NAME_MAX] = '\0';
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
2248	kfree(vol);
2249	return ret;
2250}
2251
2252static int btrfs_freeze(struct super_block *sb)
2253{
2254	struct btrfs_trans_handle *trans;
2255	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2256	struct btrfs_root *root = fs_info->tree_root;
2257
2258	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2259	/*
2260	 * We don't need a barrier here, we'll wait for any transaction that
2261	 * could be in progress on other threads (and do delayed iputs that
2262	 * we want to avoid on a frozen filesystem), or do the commit
2263	 * ourselves.
2264	 */
2265	trans = btrfs_attach_transaction_barrier(root);
2266	if (IS_ERR(trans)) {
2267		/* no transaction, don't bother */
2268		if (PTR_ERR(trans) == -ENOENT)
2269			return 0;
2270		return PTR_ERR(trans);
2271	}
2272	return btrfs_commit_transaction(trans);
2273}
2274
2275static int check_dev_super(struct btrfs_device *dev)
2276{
2277	struct btrfs_fs_info *fs_info = dev->fs_info;
2278	struct btrfs_super_block *sb;
2279	u64 last_trans;
2280	u16 csum_type;
2281	int ret = 0;
2282
2283	/* This should be called with fs still frozen. */
2284	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2285
2286	/* Missing dev, no need to check. */
2287	if (!dev->bdev)
2288		return 0;
2289
2290	/* Only need to check the primary super block. */
2291	sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2292	if (IS_ERR(sb))
2293		return PTR_ERR(sb);
2294
2295	/* Verify the checksum. */
2296	csum_type = btrfs_super_csum_type(sb);
2297	if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2298		btrfs_err(fs_info, "csum type changed, has %u expect %u",
2299			  csum_type, btrfs_super_csum_type(fs_info->super_copy));
2300		ret = -EUCLEAN;
2301		goto out;
2302	}
2303
2304	if (btrfs_check_super_csum(fs_info, sb)) {
2305		btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2306		ret = -EUCLEAN;
2307		goto out;
2308	}
2309
2310	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2311	ret = btrfs_validate_super(fs_info, sb, 0);
2312	if (ret < 0)
2313		goto out;
2314
2315	last_trans = btrfs_get_last_trans_committed(fs_info);
2316	if (btrfs_super_generation(sb) != last_trans) {
2317		btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2318			  btrfs_super_generation(sb), last_trans);
 
2319		ret = -EUCLEAN;
2320		goto out;
2321	}
2322out:
2323	btrfs_release_disk_super(sb);
2324	return ret;
2325}
2326
2327static int btrfs_unfreeze(struct super_block *sb)
2328{
2329	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2330	struct btrfs_device *device;
2331	int ret = 0;
2332
2333	/*
2334	 * Make sure the fs is not changed by accident (like hibernation then
2335	 * modified by other OS).
2336	 * If we found anything wrong, we mark the fs error immediately.
2337	 *
2338	 * And since the fs is frozen, no one can modify the fs yet, thus
2339	 * we don't need to hold device_list_mutex.
2340	 */
2341	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2342		ret = check_dev_super(device);
2343		if (ret < 0) {
2344			btrfs_handle_fs_error(fs_info, ret,
2345				"super block on devid %llu got modified unexpectedly",
2346				device->devid);
2347			break;
2348		}
2349	}
2350	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2351
2352	/*
2353	 * We still return 0, to allow VFS layer to unfreeze the fs even the
2354	 * above checks failed. Since the fs is either fine or read-only, we're
2355	 * safe to continue, without causing further damage.
2356	 */
2357	return 0;
2358}
2359
2360static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2361{
2362	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2363
2364	/*
2365	 * There should be always a valid pointer in latest_dev, it may be stale
2366	 * for a short moment in case it's being deleted but still valid until
2367	 * the end of RCU grace period.
2368	 */
2369	rcu_read_lock();
2370	seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2371	rcu_read_unlock();
2372
2373	return 0;
2374}
2375
2376static const struct super_operations btrfs_super_ops = {
2377	.drop_inode	= btrfs_drop_inode,
2378	.evict_inode	= btrfs_evict_inode,
2379	.put_super	= btrfs_put_super,
2380	.sync_fs	= btrfs_sync_fs,
2381	.show_options	= btrfs_show_options,
2382	.show_devname	= btrfs_show_devname,
2383	.alloc_inode	= btrfs_alloc_inode,
2384	.destroy_inode	= btrfs_destroy_inode,
2385	.free_inode	= btrfs_free_inode,
2386	.statfs		= btrfs_statfs,
 
2387	.freeze_fs	= btrfs_freeze,
2388	.unfreeze_fs	= btrfs_unfreeze,
2389};
2390
2391static const struct file_operations btrfs_ctl_fops = {
2392	.open = btrfs_control_open,
2393	.unlocked_ioctl	 = btrfs_control_ioctl,
2394	.compat_ioctl = compat_ptr_ioctl,
2395	.owner	 = THIS_MODULE,
2396	.llseek = noop_llseek,
2397};
2398
2399static struct miscdevice btrfs_misc = {
2400	.minor		= BTRFS_MINOR,
2401	.name		= "btrfs-control",
2402	.fops		= &btrfs_ctl_fops
2403};
2404
2405MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2406MODULE_ALIAS("devname:btrfs-control");
2407
2408static int __init btrfs_interface_init(void)
2409{
2410	return misc_register(&btrfs_misc);
2411}
2412
2413static __cold void btrfs_interface_exit(void)
2414{
2415	misc_deregister(&btrfs_misc);
2416}
2417
2418static int __init btrfs_print_mod_info(void)
2419{
2420	static const char options[] = ""
2421#ifdef CONFIG_BTRFS_DEBUG
2422			", debug=on"
2423#endif
2424#ifdef CONFIG_BTRFS_ASSERT
2425			", assert=on"
2426#endif
 
 
 
2427#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2428			", ref-verify=on"
2429#endif
2430#ifdef CONFIG_BLK_DEV_ZONED
2431			", zoned=yes"
2432#else
2433			", zoned=no"
2434#endif
2435#ifdef CONFIG_FS_VERITY
2436			", fsverity=yes"
2437#else
2438			", fsverity=no"
2439#endif
2440			;
2441	pr_info("Btrfs loaded%s\n", options);
2442	return 0;
2443}
2444
2445static int register_btrfs(void)
2446{
2447	return register_filesystem(&btrfs_fs_type);
2448}
2449
2450static void unregister_btrfs(void)
2451{
2452	unregister_filesystem(&btrfs_fs_type);
2453}
2454
2455/* Helper structure for long init/exit functions. */
2456struct init_sequence {
2457	int (*init_func)(void);
2458	/* Can be NULL if the init_func doesn't need cleanup. */
2459	void (*exit_func)(void);
2460};
2461
2462static const struct init_sequence mod_init_seq[] = {
2463	{
2464		.init_func = btrfs_props_init,
2465		.exit_func = NULL,
2466	}, {
2467		.init_func = btrfs_init_sysfs,
2468		.exit_func = btrfs_exit_sysfs,
2469	}, {
2470		.init_func = btrfs_init_compress,
2471		.exit_func = btrfs_exit_compress,
2472	}, {
2473		.init_func = btrfs_init_cachep,
2474		.exit_func = btrfs_destroy_cachep,
2475	}, {
2476		.init_func = btrfs_transaction_init,
2477		.exit_func = btrfs_transaction_exit,
2478	}, {
2479		.init_func = btrfs_ctree_init,
2480		.exit_func = btrfs_ctree_exit,
2481	}, {
2482		.init_func = btrfs_free_space_init,
2483		.exit_func = btrfs_free_space_exit,
2484	}, {
2485		.init_func = extent_state_init_cachep,
2486		.exit_func = extent_state_free_cachep,
2487	}, {
2488		.init_func = extent_buffer_init_cachep,
2489		.exit_func = extent_buffer_free_cachep,
2490	}, {
2491		.init_func = btrfs_bioset_init,
2492		.exit_func = btrfs_bioset_exit,
2493	}, {
2494		.init_func = extent_map_init,
2495		.exit_func = extent_map_exit,
2496	}, {
2497		.init_func = ordered_data_init,
2498		.exit_func = ordered_data_exit,
2499	}, {
2500		.init_func = btrfs_delayed_inode_init,
2501		.exit_func = btrfs_delayed_inode_exit,
2502	}, {
2503		.init_func = btrfs_auto_defrag_init,
2504		.exit_func = btrfs_auto_defrag_exit,
2505	}, {
2506		.init_func = btrfs_delayed_ref_init,
2507		.exit_func = btrfs_delayed_ref_exit,
2508	}, {
2509		.init_func = btrfs_prelim_ref_init,
2510		.exit_func = btrfs_prelim_ref_exit,
2511	}, {
2512		.init_func = btrfs_interface_init,
2513		.exit_func = btrfs_interface_exit,
2514	}, {
2515		.init_func = btrfs_print_mod_info,
2516		.exit_func = NULL,
2517	}, {
2518		.init_func = btrfs_run_sanity_tests,
2519		.exit_func = NULL,
2520	}, {
2521		.init_func = register_btrfs,
2522		.exit_func = unregister_btrfs,
2523	}
2524};
2525
2526static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2527
2528static __always_inline void btrfs_exit_btrfs_fs(void)
2529{
2530	int i;
2531
2532	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2533		if (!mod_init_result[i])
2534			continue;
2535		if (mod_init_seq[i].exit_func)
2536			mod_init_seq[i].exit_func();
2537		mod_init_result[i] = false;
2538	}
2539}
2540
2541static void __exit exit_btrfs_fs(void)
2542{
2543	btrfs_exit_btrfs_fs();
2544	btrfs_cleanup_fs_uuids();
2545}
2546
2547static int __init init_btrfs_fs(void)
2548{
2549	int ret;
2550	int i;
2551
2552	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2553		ASSERT(!mod_init_result[i]);
2554		ret = mod_init_seq[i].init_func();
2555		if (ret < 0) {
2556			btrfs_exit_btrfs_fs();
2557			return ret;
2558		}
2559		mod_init_result[i] = true;
2560	}
2561	return 0;
2562}
2563
2564late_initcall(init_btrfs_fs);
2565module_exit(exit_btrfs_fs)
2566
2567MODULE_LICENSE("GPL");
2568MODULE_SOFTDEP("pre: crc32c");
2569MODULE_SOFTDEP("pre: xxhash64");
2570MODULE_SOFTDEP("pre: sha256");
2571MODULE_SOFTDEP("pre: blake2b-256");
v6.2
   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 "messages.h"
  30#include "delayed-inode.h"
  31#include "ctree.h"
  32#include "disk-io.h"
  33#include "transaction.h"
  34#include "btrfs_inode.h"
  35#include "print-tree.h"
  36#include "props.h"
  37#include "xattr.h"
  38#include "bio.h"
  39#include "export.h"
  40#include "compression.h"
  41#include "rcu-string.h"
  42#include "dev-replace.h"
  43#include "free-space-cache.h"
  44#include "backref.h"
  45#include "space-info.h"
  46#include "sysfs.h"
  47#include "zoned.h"
  48#include "tests/btrfs-tests.h"
  49#include "block-group.h"
  50#include "discard.h"
  51#include "qgroup.h"
  52#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#define CREATE_TRACE_POINTS
  62#include <trace/events/btrfs.h>
  63
  64static const struct super_operations btrfs_super_ops;
  65
  66/*
  67 * Types for mounting the default subvolume and a subvolume explicitly
  68 * requested by subvol=/path. That way the callchain is straightforward and we
  69 * don't have to play tricks with the mount options and recursive calls to
  70 * btrfs_mount.
  71 *
  72 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
  73 */
  74static struct file_system_type btrfs_fs_type;
  75static struct file_system_type btrfs_root_fs_type;
  76
  77static int btrfs_remount(struct super_block *sb, int *flags, char *data);
  78
  79static void btrfs_put_super(struct super_block *sb)
  80{
  81	close_ctree(btrfs_sb(sb));
 
 
 
  82}
  83
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  84enum {
  85	Opt_acl, Opt_noacl,
  86	Opt_clear_cache,
  87	Opt_commit_interval,
  88	Opt_compress,
  89	Opt_compress_force,
  90	Opt_compress_force_type,
  91	Opt_compress_type,
  92	Opt_degraded,
  93	Opt_device,
  94	Opt_fatal_errors,
  95	Opt_flushoncommit, Opt_noflushoncommit,
  96	Opt_max_inline,
  97	Opt_barrier, Opt_nobarrier,
  98	Opt_datacow, Opt_nodatacow,
  99	Opt_datasum, Opt_nodatasum,
 100	Opt_defrag, Opt_nodefrag,
 101	Opt_discard, Opt_nodiscard,
 102	Opt_discard_mode,
 103	Opt_norecovery,
 104	Opt_ratio,
 105	Opt_rescan_uuid_tree,
 106	Opt_skip_balance,
 107	Opt_space_cache, Opt_no_space_cache,
 108	Opt_space_cache_version,
 109	Opt_ssd, Opt_nossd,
 110	Opt_ssd_spread, Opt_nossd_spread,
 111	Opt_subvol,
 112	Opt_subvol_empty,
 113	Opt_subvolid,
 114	Opt_thread_pool,
 115	Opt_treelog, Opt_notreelog,
 116	Opt_user_subvol_rm_allowed,
 117
 118	/* Rescue options */
 119	Opt_rescue,
 120	Opt_usebackuproot,
 121	Opt_nologreplay,
 122	Opt_ignorebadroots,
 123	Opt_ignoredatacsums,
 124	Opt_rescue_all,
 125
 126	/* Deprecated options */
 127	Opt_recovery,
 128	Opt_inode_cache, Opt_noinode_cache,
 129
 130	/* Debugging options */
 131	Opt_check_integrity,
 132	Opt_check_integrity_including_extent_data,
 133	Opt_check_integrity_print_mask,
 134	Opt_enospc_debug, Opt_noenospc_debug,
 135#ifdef CONFIG_BTRFS_DEBUG
 136	Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 137#endif
 138#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 139	Opt_ref_verify,
 140#endif
 141	Opt_err,
 142};
 143
 144static const match_table_t tokens = {
 145	{Opt_acl, "acl"},
 146	{Opt_noacl, "noacl"},
 147	{Opt_clear_cache, "clear_cache"},
 148	{Opt_commit_interval, "commit=%u"},
 149	{Opt_compress, "compress"},
 150	{Opt_compress_type, "compress=%s"},
 151	{Opt_compress_force, "compress-force"},
 152	{Opt_compress_force_type, "compress-force=%s"},
 153	{Opt_degraded, "degraded"},
 154	{Opt_device, "device=%s"},
 155	{Opt_fatal_errors, "fatal_errors=%s"},
 156	{Opt_flushoncommit, "flushoncommit"},
 157	{Opt_noflushoncommit, "noflushoncommit"},
 158	{Opt_inode_cache, "inode_cache"},
 159	{Opt_noinode_cache, "noinode_cache"},
 160	{Opt_max_inline, "max_inline=%s"},
 161	{Opt_barrier, "barrier"},
 162	{Opt_nobarrier, "nobarrier"},
 163	{Opt_datacow, "datacow"},
 164	{Opt_nodatacow, "nodatacow"},
 165	{Opt_datasum, "datasum"},
 166	{Opt_nodatasum, "nodatasum"},
 167	{Opt_defrag, "autodefrag"},
 168	{Opt_nodefrag, "noautodefrag"},
 169	{Opt_discard, "discard"},
 170	{Opt_discard_mode, "discard=%s"},
 171	{Opt_nodiscard, "nodiscard"},
 172	{Opt_norecovery, "norecovery"},
 173	{Opt_ratio, "metadata_ratio=%u"},
 174	{Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 175	{Opt_skip_balance, "skip_balance"},
 176	{Opt_space_cache, "space_cache"},
 177	{Opt_no_space_cache, "nospace_cache"},
 178	{Opt_space_cache_version, "space_cache=%s"},
 179	{Opt_ssd, "ssd"},
 180	{Opt_nossd, "nossd"},
 181	{Opt_ssd_spread, "ssd_spread"},
 182	{Opt_nossd_spread, "nossd_spread"},
 183	{Opt_subvol, "subvol=%s"},
 184	{Opt_subvol_empty, "subvol="},
 185	{Opt_subvolid, "subvolid=%s"},
 186	{Opt_thread_pool, "thread_pool=%u"},
 187	{Opt_treelog, "treelog"},
 188	{Opt_notreelog, "notreelog"},
 189	{Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 190
 191	/* Rescue options */
 192	{Opt_rescue, "rescue=%s"},
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 193	/* Deprecated, with alias rescue=nologreplay */
 194	{Opt_nologreplay, "nologreplay"},
 195	/* Deprecated, with alias rescue=usebackuproot */
 196	{Opt_usebackuproot, "usebackuproot"},
 197
 198	/* Deprecated options */
 199	{Opt_recovery, "recovery"},
 200
 201	/* Debugging options */
 202	{Opt_check_integrity, "check_int"},
 203	{Opt_check_integrity_including_extent_data, "check_int_data"},
 204	{Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
 205	{Opt_enospc_debug, "enospc_debug"},
 206	{Opt_noenospc_debug, "noenospc_debug"},
 207#ifdef CONFIG_BTRFS_DEBUG
 208	{Opt_fragment_data, "fragment=data"},
 209	{Opt_fragment_metadata, "fragment=metadata"},
 210	{Opt_fragment_all, "fragment=all"},
 211#endif
 212#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 213	{Opt_ref_verify, "ref_verify"},
 214#endif
 215	{Opt_err, NULL},
 216};
 217
 218static const match_table_t rescue_tokens = {
 219	{Opt_usebackuproot, "usebackuproot"},
 220	{Opt_nologreplay, "nologreplay"},
 221	{Opt_ignorebadroots, "ignorebadroots"},
 222	{Opt_ignorebadroots, "ibadroots"},
 223	{Opt_ignoredatacsums, "ignoredatacsums"},
 224	{Opt_ignoredatacsums, "idatacsums"},
 225	{Opt_rescue_all, "all"},
 226	{Opt_err, NULL},
 227};
 228
 229static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
 230			    const char *opt_name)
 231{
 232	if (fs_info->mount_opt & opt) {
 233		btrfs_err(fs_info, "%s must be used with ro mount option",
 234			  opt_name);
 235		return true;
 236	}
 237	return false;
 238}
 239
 240static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
 241{
 242	char *opts;
 243	char *orig;
 244	char *p;
 245	substring_t args[MAX_OPT_ARGS];
 246	int ret = 0;
 247
 248	opts = kstrdup(options, GFP_KERNEL);
 249	if (!opts)
 250		return -ENOMEM;
 251	orig = opts;
 252
 253	while ((p = strsep(&opts, ":")) != NULL) {
 254		int token;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 255
 256		if (!*p)
 257			continue;
 258		token = match_token(p, rescue_tokens, args);
 259		switch (token){
 260		case Opt_usebackuproot:
 261			btrfs_info(info,
 262				   "trying to use backup root at mount time");
 263			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 264			break;
 265		case Opt_nologreplay:
 266			btrfs_set_and_info(info, NOLOGREPLAY,
 267					   "disabling log replay at mount time");
 268			break;
 269		case Opt_ignorebadroots:
 270			btrfs_set_and_info(info, IGNOREBADROOTS,
 271					   "ignoring bad roots");
 272			break;
 273		case Opt_ignoredatacsums:
 274			btrfs_set_and_info(info, IGNOREDATACSUMS,
 275					   "ignoring data csums");
 276			break;
 277		case Opt_rescue_all:
 278			btrfs_info(info, "enabling all of the rescue options");
 279			btrfs_set_and_info(info, IGNOREDATACSUMS,
 280					   "ignoring data csums");
 281			btrfs_set_and_info(info, IGNOREBADROOTS,
 282					   "ignoring bad roots");
 283			btrfs_set_and_info(info, NOLOGREPLAY,
 284					   "disabling log replay at mount time");
 285			break;
 286		case Opt_err:
 287			btrfs_info(info, "unrecognized rescue option '%s'", p);
 288			ret = -EINVAL;
 289			goto out;
 290		default:
 291			break;
 292		}
 293
 
 
 
 
 
 
 294	}
 295out:
 296	kfree(orig);
 297	return ret;
 298}
 299
 300/*
 301 * Regular mount options parser.  Everything that is needed only when
 302 * reading in a new superblock is parsed here.
 303 * XXX JDM: This needs to be cleaned up for remount.
 304 */
 305int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 306			unsigned long new_flags)
 307{
 308	substring_t args[MAX_OPT_ARGS];
 309	char *p, *num;
 310	int intarg;
 311	int ret = 0;
 312	char *compress_type;
 313	bool compress_force = false;
 314	enum btrfs_compression_type saved_compress_type;
 315	int saved_compress_level;
 316	bool saved_compress_force;
 317	int no_compress = 0;
 318	const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
 319
 320	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
 321		btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
 322	else if (btrfs_free_space_cache_v1_active(info)) {
 323		if (btrfs_is_zoned(info)) {
 324			btrfs_info(info,
 325			"zoned: clearing existing space cache");
 326			btrfs_set_super_cache_generation(info->super_copy, 0);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 327		} else {
 328			btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 329		}
 330	}
 331
 332	/*
 333	 * Even the options are empty, we still need to do extra check
 334	 * against new flags
 335	 */
 336	if (!options)
 337		goto check;
 338
 339	while ((p = strsep(&options, ",")) != NULL) {
 340		int token;
 341		if (!*p)
 342			continue;
 343
 344		token = match_token(p, tokens, args);
 345		switch (token) {
 346		case Opt_degraded:
 347			btrfs_info(info, "allowing degraded mounts");
 348			btrfs_set_opt(info->mount_opt, DEGRADED);
 349			break;
 350		case Opt_subvol:
 351		case Opt_subvol_empty:
 352		case Opt_subvolid:
 353		case Opt_device:
 354			/*
 355			 * These are parsed by btrfs_parse_subvol_options or
 356			 * btrfs_parse_device_options and can be ignored here.
 357			 */
 358			break;
 359		case Opt_nodatasum:
 360			btrfs_set_and_info(info, NODATASUM,
 361					   "setting nodatasum");
 362			break;
 363		case Opt_datasum:
 364			if (btrfs_test_opt(info, NODATASUM)) {
 365				if (btrfs_test_opt(info, NODATACOW))
 366					btrfs_info(info,
 367						   "setting datasum, datacow enabled");
 368				else
 369					btrfs_info(info, "setting datasum");
 370			}
 371			btrfs_clear_opt(info->mount_opt, NODATACOW);
 372			btrfs_clear_opt(info->mount_opt, NODATASUM);
 373			break;
 374		case Opt_nodatacow:
 375			if (!btrfs_test_opt(info, NODATACOW)) {
 376				if (!btrfs_test_opt(info, COMPRESS) ||
 377				    !btrfs_test_opt(info, FORCE_COMPRESS)) {
 378					btrfs_info(info,
 379						   "setting nodatacow, compression disabled");
 380				} else {
 381					btrfs_info(info, "setting nodatacow");
 382				}
 383			}
 384			btrfs_clear_opt(info->mount_opt, COMPRESS);
 385			btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 386			btrfs_set_opt(info->mount_opt, NODATACOW);
 387			btrfs_set_opt(info->mount_opt, NODATASUM);
 388			break;
 389		case Opt_datacow:
 390			btrfs_clear_and_info(info, NODATACOW,
 391					     "setting datacow");
 392			break;
 393		case Opt_compress_force:
 394		case Opt_compress_force_type:
 395			compress_force = true;
 396			fallthrough;
 397		case Opt_compress:
 398		case Opt_compress_type:
 399			saved_compress_type = btrfs_test_opt(info,
 400							     COMPRESS) ?
 401				info->compress_type : BTRFS_COMPRESS_NONE;
 402			saved_compress_force =
 403				btrfs_test_opt(info, FORCE_COMPRESS);
 404			saved_compress_level = info->compress_level;
 405			if (token == Opt_compress ||
 406			    token == Opt_compress_force ||
 407			    strncmp(args[0].from, "zlib", 4) == 0) {
 408				compress_type = "zlib";
 409
 410				info->compress_type = BTRFS_COMPRESS_ZLIB;
 411				info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 412				/*
 413				 * args[0] contains uninitialized data since
 414				 * for these tokens we don't expect any
 415				 * parameter.
 416				 */
 417				if (token != Opt_compress &&
 418				    token != Opt_compress_force)
 419					info->compress_level =
 420					  btrfs_compress_str2level(
 421							BTRFS_COMPRESS_ZLIB,
 422							args[0].from + 4);
 423				btrfs_set_opt(info->mount_opt, COMPRESS);
 424				btrfs_clear_opt(info->mount_opt, NODATACOW);
 425				btrfs_clear_opt(info->mount_opt, NODATASUM);
 426				no_compress = 0;
 427			} else if (strncmp(args[0].from, "lzo", 3) == 0) {
 428				compress_type = "lzo";
 429				info->compress_type = BTRFS_COMPRESS_LZO;
 430				info->compress_level = 0;
 431				btrfs_set_opt(info->mount_opt, COMPRESS);
 432				btrfs_clear_opt(info->mount_opt, NODATACOW);
 433				btrfs_clear_opt(info->mount_opt, NODATASUM);
 434				btrfs_set_fs_incompat(info, COMPRESS_LZO);
 435				no_compress = 0;
 436			} else if (strncmp(args[0].from, "zstd", 4) == 0) {
 437				compress_type = "zstd";
 438				info->compress_type = BTRFS_COMPRESS_ZSTD;
 439				info->compress_level =
 440					btrfs_compress_str2level(
 441							 BTRFS_COMPRESS_ZSTD,
 442							 args[0].from + 4);
 443				btrfs_set_opt(info->mount_opt, COMPRESS);
 444				btrfs_clear_opt(info->mount_opt, NODATACOW);
 445				btrfs_clear_opt(info->mount_opt, NODATASUM);
 446				btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
 447				no_compress = 0;
 448			} else if (strncmp(args[0].from, "no", 2) == 0) {
 449				compress_type = "no";
 450				info->compress_level = 0;
 451				info->compress_type = 0;
 452				btrfs_clear_opt(info->mount_opt, COMPRESS);
 453				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 454				compress_force = false;
 455				no_compress++;
 456			} else {
 457				btrfs_err(info, "unrecognized compression value %s",
 458					  args[0].from);
 459				ret = -EINVAL;
 460				goto out;
 461			}
 462
 463			if (compress_force) {
 464				btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 465			} else {
 466				/*
 467				 * If we remount from compress-force=xxx to
 468				 * compress=xxx, we need clear FORCE_COMPRESS
 469				 * flag, otherwise, there is no way for users
 470				 * to disable forcible compression separately.
 471				 */
 472				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 473			}
 474			if (no_compress == 1) {
 475				btrfs_info(info, "use no compression");
 476			} else if ((info->compress_type != saved_compress_type) ||
 477				   (compress_force != saved_compress_force) ||
 478				   (info->compress_level != saved_compress_level)) {
 479				btrfs_info(info, "%s %s compression, level %d",
 480					   (compress_force) ? "force" : "use",
 481					   compress_type, info->compress_level);
 482			}
 483			compress_force = false;
 484			break;
 485		case Opt_ssd:
 486			btrfs_set_and_info(info, SSD,
 487					   "enabling ssd optimizations");
 488			btrfs_clear_opt(info->mount_opt, NOSSD);
 489			break;
 490		case Opt_ssd_spread:
 491			btrfs_set_and_info(info, SSD,
 492					   "enabling ssd optimizations");
 493			btrfs_set_and_info(info, SSD_SPREAD,
 494					   "using spread ssd allocation scheme");
 495			btrfs_clear_opt(info->mount_opt, NOSSD);
 496			break;
 497		case Opt_nossd:
 498			btrfs_set_opt(info->mount_opt, NOSSD);
 499			btrfs_clear_and_info(info, SSD,
 500					     "not using ssd optimizations");
 501			fallthrough;
 502		case Opt_nossd_spread:
 503			btrfs_clear_and_info(info, SSD_SPREAD,
 504					     "not using spread ssd allocation scheme");
 505			break;
 506		case Opt_barrier:
 507			btrfs_clear_and_info(info, NOBARRIER,
 508					     "turning on barriers");
 509			break;
 510		case Opt_nobarrier:
 511			btrfs_set_and_info(info, NOBARRIER,
 512					   "turning off barriers");
 513			break;
 514		case Opt_thread_pool:
 515			ret = match_int(&args[0], &intarg);
 516			if (ret) {
 517				btrfs_err(info, "unrecognized thread_pool value %s",
 518					  args[0].from);
 519				goto out;
 520			} else if (intarg == 0) {
 521				btrfs_err(info, "invalid value 0 for thread_pool");
 522				ret = -EINVAL;
 523				goto out;
 524			}
 525			info->thread_pool_size = intarg;
 526			break;
 527		case Opt_max_inline:
 528			num = match_strdup(&args[0]);
 529			if (num) {
 530				info->max_inline = memparse(num, NULL);
 531				kfree(num);
 532
 533				if (info->max_inline) {
 534					info->max_inline = min_t(u64,
 535						info->max_inline,
 536						info->sectorsize);
 537				}
 538				btrfs_info(info, "max_inline at %llu",
 539					   info->max_inline);
 540			} else {
 541				ret = -ENOMEM;
 542				goto out;
 543			}
 544			break;
 545		case Opt_acl:
 546#ifdef CONFIG_BTRFS_FS_POSIX_ACL
 547			info->sb->s_flags |= SB_POSIXACL;
 548			break;
 549#else
 550			btrfs_err(info, "support for ACL not compiled in!");
 551			ret = -EINVAL;
 552			goto out;
 553#endif
 554		case Opt_noacl:
 555			info->sb->s_flags &= ~SB_POSIXACL;
 556			break;
 557		case Opt_notreelog:
 558			btrfs_set_and_info(info, NOTREELOG,
 559					   "disabling tree log");
 560			break;
 561		case Opt_treelog:
 562			btrfs_clear_and_info(info, NOTREELOG,
 563					     "enabling tree log");
 564			break;
 565		case Opt_norecovery:
 566		case Opt_nologreplay:
 567			btrfs_warn(info,
 
 
 
 
 
 568		"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 569			btrfs_set_and_info(info, NOLOGREPLAY,
 570					   "disabling log replay at mount time");
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 571			break;
 572		case Opt_flushoncommit:
 573			btrfs_set_and_info(info, FLUSHONCOMMIT,
 574					   "turning on flush-on-commit");
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 575			break;
 576		case Opt_noflushoncommit:
 577			btrfs_clear_and_info(info, FLUSHONCOMMIT,
 578					     "turning off flush-on-commit");
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 579			break;
 580		case Opt_ratio:
 581			ret = match_int(&args[0], &intarg);
 582			if (ret) {
 583				btrfs_err(info, "unrecognized metadata_ratio value %s",
 584					  args[0].from);
 585				goto out;
 586			}
 587			info->metadata_ratio = intarg;
 588			btrfs_info(info, "metadata ratio %u",
 589				   info->metadata_ratio);
 590			break;
 591		case Opt_discard:
 592		case Opt_discard_mode:
 593			if (token == Opt_discard ||
 594			    strcmp(args[0].from, "sync") == 0) {
 595				btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
 596				btrfs_set_and_info(info, DISCARD_SYNC,
 597						   "turning on sync discard");
 598			} else if (strcmp(args[0].from, "async") == 0) {
 599				btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
 600				btrfs_set_and_info(info, DISCARD_ASYNC,
 601						   "turning on async discard");
 602			} else {
 603				btrfs_err(info, "unrecognized discard mode value %s",
 604					  args[0].from);
 605				ret = -EINVAL;
 606				goto out;
 607			}
 608			btrfs_clear_opt(info->mount_opt, NODISCARD);
 609			break;
 610		case Opt_nodiscard:
 611			btrfs_clear_and_info(info, DISCARD_SYNC,
 612					     "turning off discard");
 613			btrfs_clear_and_info(info, DISCARD_ASYNC,
 614					     "turning off async discard");
 615			btrfs_set_opt(info->mount_opt, NODISCARD);
 616			break;
 617		case Opt_space_cache:
 618		case Opt_space_cache_version:
 619			/*
 620			 * We already set FREE_SPACE_TREE above because we have
 621			 * compat_ro(FREE_SPACE_TREE) set, and we aren't going
 622			 * to allow v1 to be set for extent tree v2, simply
 623			 * ignore this setting if we're extent tree v2.
 624			 */
 625			if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
 626				break;
 627			if (token == Opt_space_cache ||
 628			    strcmp(args[0].from, "v1") == 0) {
 629				btrfs_clear_opt(info->mount_opt,
 630						FREE_SPACE_TREE);
 631				btrfs_set_and_info(info, SPACE_CACHE,
 632					   "enabling disk space caching");
 633			} else if (strcmp(args[0].from, "v2") == 0) {
 634				btrfs_clear_opt(info->mount_opt,
 635						SPACE_CACHE);
 636				btrfs_set_and_info(info, FREE_SPACE_TREE,
 637						   "enabling free space tree");
 638			} else {
 639				btrfs_err(info, "unrecognized space_cache value %s",
 640					  args[0].from);
 641				ret = -EINVAL;
 642				goto out;
 643			}
 644			break;
 645		case Opt_rescan_uuid_tree:
 646			btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 647			break;
 648		case Opt_no_space_cache:
 649			/*
 650			 * We cannot operate without the free space tree with
 651			 * extent tree v2, ignore this option.
 652			 */
 653			if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
 654				break;
 655			if (btrfs_test_opt(info, SPACE_CACHE)) {
 656				btrfs_clear_and_info(info, SPACE_CACHE,
 657					     "disabling disk space caching");
 658			}
 659			if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
 660				btrfs_clear_and_info(info, FREE_SPACE_TREE,
 661					     "disabling free space tree");
 662			}
 663			break;
 664		case Opt_inode_cache:
 665		case Opt_noinode_cache:
 666			btrfs_warn(info,
 667	"the 'inode_cache' option is deprecated and has no effect since 5.11");
 668			break;
 669		case Opt_clear_cache:
 670			/*
 671			 * We cannot clear the free space tree with extent tree
 672			 * v2, ignore this option.
 673			 */
 674			if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
 675				break;
 676			btrfs_set_and_info(info, CLEAR_CACHE,
 677					   "force clearing of disk cache");
 678			break;
 679		case Opt_user_subvol_rm_allowed:
 680			btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 681			break;
 682		case Opt_enospc_debug:
 683			btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 684			break;
 685		case Opt_noenospc_debug:
 686			btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 687			break;
 688		case Opt_defrag:
 689			btrfs_set_and_info(info, AUTO_DEFRAG,
 690					   "enabling auto defrag");
 691			break;
 692		case Opt_nodefrag:
 693			btrfs_clear_and_info(info, AUTO_DEFRAG,
 694					     "disabling auto defrag");
 695			break;
 696		case Opt_recovery:
 697		case Opt_usebackuproot:
 698			btrfs_warn(info,
 699			"'%s' is deprecated, use 'rescue=usebackuproot' instead",
 700				   token == Opt_recovery ? "recovery" :
 701				   "usebackuproot");
 702			btrfs_info(info,
 703				   "trying to use backup root at mount time");
 704			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 705			break;
 706		case Opt_skip_balance:
 707			btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 708			break;
 709#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 710		case Opt_check_integrity_including_extent_data:
 711			btrfs_info(info,
 712				   "enabling check integrity including extent data");
 713			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
 714			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 715			break;
 716		case Opt_check_integrity:
 717			btrfs_info(info, "enabling check integrity");
 718			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 719			break;
 720		case Opt_check_integrity_print_mask:
 721			ret = match_int(&args[0], &intarg);
 722			if (ret) {
 723				btrfs_err(info,
 724				"unrecognized check_integrity_print_mask value %s",
 725					args[0].from);
 726				goto out;
 727			}
 728			info->check_integrity_print_mask = intarg;
 729			btrfs_info(info, "check_integrity_print_mask 0x%x",
 730				   info->check_integrity_print_mask);
 731			break;
 732#else
 733		case Opt_check_integrity_including_extent_data:
 734		case Opt_check_integrity:
 735		case Opt_check_integrity_print_mask:
 736			btrfs_err(info,
 737				  "support for check_integrity* not compiled in!");
 738			ret = -EINVAL;
 739			goto out;
 740#endif
 741		case Opt_fatal_errors:
 742			if (strcmp(args[0].from, "panic") == 0) {
 743				btrfs_set_opt(info->mount_opt,
 744					      PANIC_ON_FATAL_ERROR);
 745			} else if (strcmp(args[0].from, "bug") == 0) {
 746				btrfs_clear_opt(info->mount_opt,
 747					      PANIC_ON_FATAL_ERROR);
 748			} else {
 749				btrfs_err(info, "unrecognized fatal_errors value %s",
 750					  args[0].from);
 751				ret = -EINVAL;
 752				goto out;
 753			}
 754			break;
 755		case Opt_commit_interval:
 756			intarg = 0;
 757			ret = match_int(&args[0], &intarg);
 758			if (ret) {
 759				btrfs_err(info, "unrecognized commit_interval value %s",
 760					  args[0].from);
 761				ret = -EINVAL;
 762				goto out;
 763			}
 764			if (intarg == 0) {
 765				btrfs_info(info,
 766					   "using default commit interval %us",
 767					   BTRFS_DEFAULT_COMMIT_INTERVAL);
 768				intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
 769			} else if (intarg > 300) {
 770				btrfs_warn(info, "excessive commit interval %d",
 771					   intarg);
 772			}
 773			info->commit_interval = intarg;
 774			break;
 775		case Opt_rescue:
 776			ret = parse_rescue_options(info, args[0].from);
 777			if (ret < 0) {
 778				btrfs_err(info, "unrecognized rescue value %s",
 779					  args[0].from);
 780				goto out;
 781			}
 782			break;
 
 
 
 
 
 
 783#ifdef CONFIG_BTRFS_DEBUG
 784		case Opt_fragment_all:
 785			btrfs_info(info, "fragmenting all space");
 786			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 787			btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
 
 788			break;
 789		case Opt_fragment_metadata:
 790			btrfs_info(info, "fragmenting metadata");
 791			btrfs_set_opt(info->mount_opt,
 792				      FRAGMENT_METADATA);
 793			break;
 794		case Opt_fragment_data:
 795			btrfs_info(info, "fragmenting data");
 796			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 797			break;
 
 
 
 
 
 
 798#endif
 799#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 800		case Opt_ref_verify:
 801			btrfs_info(info, "doing ref verification");
 802			btrfs_set_opt(info->mount_opt, REF_VERIFY);
 803			break;
 804#endif
 805		case Opt_err:
 806			btrfs_err(info, "unrecognized mount option '%s'", p);
 807			ret = -EINVAL;
 808			goto out;
 809		default:
 810			break;
 811		}
 812	}
 813check:
 814	/* We're read-only, don't have to check. */
 815	if (new_flags & SB_RDONLY)
 816		goto out;
 817
 818	if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
 819	    check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
 820	    check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
 821		ret = -EINVAL;
 822out:
 823	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 824	    !btrfs_test_opt(info, FREE_SPACE_TREE) &&
 825	    !btrfs_test_opt(info, CLEAR_CACHE)) {
 826		btrfs_err(info, "cannot disable free space tree");
 827		ret = -EINVAL;
 828
 829	}
 830	if (!ret)
 831		ret = btrfs_check_mountopts_zoned(info);
 832	if (!ret && !remounting) {
 833		if (btrfs_test_opt(info, SPACE_CACHE))
 834			btrfs_info(info, "disk space caching is enabled");
 835		if (btrfs_test_opt(info, FREE_SPACE_TREE))
 836			btrfs_info(info, "using free space tree");
 837	}
 838	return ret;
 839}
 840
 841/*
 842 * Parse mount options that are required early in the mount process.
 843 *
 844 * All other options will be parsed on much later in the mount process and
 845 * only when we need to allocate a new super block.
 846 */
 847static int btrfs_parse_device_options(const char *options, fmode_t flags,
 848				      void *holder)
 849{
 850	substring_t args[MAX_OPT_ARGS];
 851	char *device_name, *opts, *orig, *p;
 852	struct btrfs_device *device = NULL;
 853	int error = 0;
 854
 855	lockdep_assert_held(&uuid_mutex);
 
 
 
 
 
 
 
 
 
 
 856
 857	if (!options)
 858		return 0;
 
 
 859
 860	/*
 861	 * strsep changes the string, duplicate it because btrfs_parse_options
 862	 * gets called later
 863	 */
 864	opts = kstrdup(options, GFP_KERNEL);
 865	if (!opts)
 866		return -ENOMEM;
 867	orig = opts;
 868
 869	while ((p = strsep(&opts, ",")) != NULL) {
 870		int token;
 
 
 
 
 
 
 
 
 
 871
 872		if (!*p)
 873			continue;
 874
 875		token = match_token(p, tokens, args);
 876		if (token == Opt_device) {
 877			device_name = match_strdup(&args[0]);
 878			if (!device_name) {
 879				error = -ENOMEM;
 880				goto out;
 881			}
 882			device = btrfs_scan_one_device(device_name, flags,
 883					holder);
 884			kfree(device_name);
 885			if (IS_ERR(device)) {
 886				error = PTR_ERR(device);
 887				goto out;
 888			}
 889		}
 890	}
 891
 892out:
 893	kfree(orig);
 894	return error;
 895}
 896
 897/*
 898 * Parse mount options that are related to subvolume id
 
 
 
 899 *
 900 * The value is later passed to mount_subvol()
 
 
 
 
 901 */
 902static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
 903		u64 *subvol_objectid)
 904{
 905	substring_t args[MAX_OPT_ARGS];
 906	char *opts, *orig, *p;
 907	int error = 0;
 908	u64 subvolid;
 909
 910	if (!options)
 911		return 0;
 
 
 912
 913	/*
 914	 * strsep changes the string, duplicate it because
 915	 * btrfs_parse_device_options gets called later
 916	 */
 917	opts = kstrdup(options, GFP_KERNEL);
 918	if (!opts)
 919		return -ENOMEM;
 920	orig = opts;
 
 
 
 
 
 921
 922	while ((p = strsep(&opts, ",")) != NULL) {
 923		int token;
 924		if (!*p)
 925			continue;
 926
 927		token = match_token(p, tokens, args);
 928		switch (token) {
 929		case Opt_subvol:
 930			kfree(*subvol_name);
 931			*subvol_name = match_strdup(&args[0]);
 932			if (!*subvol_name) {
 933				error = -ENOMEM;
 934				goto out;
 935			}
 936			break;
 937		case Opt_subvolid:
 938			error = match_u64(&args[0], &subvolid);
 939			if (error)
 940				goto out;
 941
 942			/* we want the original fs_tree */
 943			if (subvolid == 0)
 944				subvolid = BTRFS_FS_TREE_OBJECTID;
 
 
 
 
 
 
 945
 946			*subvol_objectid = subvolid;
 947			break;
 948		default:
 949			break;
 950		}
 951	}
 952
 953out:
 954	kfree(orig);
 955	return error;
 
 
 
 
 
 
 
 
 
 
 
 
 956}
 957
 958char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 959					  u64 subvol_objectid)
 960{
 961	struct btrfs_root *root = fs_info->tree_root;
 962	struct btrfs_root *fs_root = NULL;
 963	struct btrfs_root_ref *root_ref;
 964	struct btrfs_inode_ref *inode_ref;
 965	struct btrfs_key key;
 966	struct btrfs_path *path = NULL;
 967	char *name = NULL, *ptr;
 968	u64 dirid;
 969	int len;
 970	int ret;
 971
 972	path = btrfs_alloc_path();
 973	if (!path) {
 974		ret = -ENOMEM;
 975		goto err;
 976	}
 977
 978	name = kmalloc(PATH_MAX, GFP_KERNEL);
 979	if (!name) {
 980		ret = -ENOMEM;
 981		goto err;
 982	}
 983	ptr = name + PATH_MAX - 1;
 984	ptr[0] = '\0';
 985
 986	/*
 987	 * Walk up the subvolume trees in the tree of tree roots by root
 988	 * backrefs until we hit the top-level subvolume.
 989	 */
 990	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
 991		key.objectid = subvol_objectid;
 992		key.type = BTRFS_ROOT_BACKREF_KEY;
 993		key.offset = (u64)-1;
 994
 995		ret = btrfs_search_backwards(root, &key, path);
 996		if (ret < 0) {
 997			goto err;
 998		} else if (ret > 0) {
 999			ret = -ENOENT;
1000			goto err;
1001		}
1002
1003		subvol_objectid = key.offset;
1004
1005		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1006					  struct btrfs_root_ref);
1007		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1008		ptr -= len + 1;
1009		if (ptr < name) {
1010			ret = -ENAMETOOLONG;
1011			goto err;
1012		}
1013		read_extent_buffer(path->nodes[0], ptr + 1,
1014				   (unsigned long)(root_ref + 1), len);
1015		ptr[0] = '/';
1016		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1017		btrfs_release_path(path);
1018
1019		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1020		if (IS_ERR(fs_root)) {
1021			ret = PTR_ERR(fs_root);
1022			fs_root = NULL;
1023			goto err;
1024		}
1025
1026		/*
1027		 * Walk up the filesystem tree by inode refs until we hit the
1028		 * root directory.
1029		 */
1030		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1031			key.objectid = dirid;
1032			key.type = BTRFS_INODE_REF_KEY;
1033			key.offset = (u64)-1;
1034
1035			ret = btrfs_search_backwards(fs_root, &key, path);
1036			if (ret < 0) {
1037				goto err;
1038			} else if (ret > 0) {
1039				ret = -ENOENT;
1040				goto err;
1041			}
1042
1043			dirid = key.offset;
1044
1045			inode_ref = btrfs_item_ptr(path->nodes[0],
1046						   path->slots[0],
1047						   struct btrfs_inode_ref);
1048			len = btrfs_inode_ref_name_len(path->nodes[0],
1049						       inode_ref);
1050			ptr -= len + 1;
1051			if (ptr < name) {
1052				ret = -ENAMETOOLONG;
1053				goto err;
1054			}
1055			read_extent_buffer(path->nodes[0], ptr + 1,
1056					   (unsigned long)(inode_ref + 1), len);
1057			ptr[0] = '/';
1058			btrfs_release_path(path);
1059		}
1060		btrfs_put_root(fs_root);
1061		fs_root = NULL;
1062	}
1063
1064	btrfs_free_path(path);
1065	if (ptr == name + PATH_MAX - 1) {
1066		name[0] = '/';
1067		name[1] = '\0';
1068	} else {
1069		memmove(name, ptr, name + PATH_MAX - ptr);
1070	}
1071	return name;
1072
1073err:
1074	btrfs_put_root(fs_root);
1075	btrfs_free_path(path);
1076	kfree(name);
1077	return ERR_PTR(ret);
1078}
1079
1080static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1081{
1082	struct btrfs_root *root = fs_info->tree_root;
1083	struct btrfs_dir_item *di;
1084	struct btrfs_path *path;
1085	struct btrfs_key location;
1086	struct fscrypt_str name = FSTR_INIT("default", 7);
1087	u64 dir_id;
1088
1089	path = btrfs_alloc_path();
1090	if (!path)
1091		return -ENOMEM;
1092
1093	/*
1094	 * Find the "default" dir item which points to the root item that we
1095	 * will mount by default if we haven't been given a specific subvolume
1096	 * to mount.
1097	 */
1098	dir_id = btrfs_super_root_dir(fs_info->super_copy);
1099	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
1100	if (IS_ERR(di)) {
1101		btrfs_free_path(path);
1102		return PTR_ERR(di);
1103	}
1104	if (!di) {
1105		/*
1106		 * Ok the default dir item isn't there.  This is weird since
1107		 * it's always been there, but don't freak out, just try and
1108		 * mount the top-level subvolume.
1109		 */
1110		btrfs_free_path(path);
1111		*objectid = BTRFS_FS_TREE_OBJECTID;
1112		return 0;
1113	}
1114
1115	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1116	btrfs_free_path(path);
1117	*objectid = location.objectid;
1118	return 0;
1119}
1120
1121static int btrfs_fill_super(struct super_block *sb,
1122			    struct btrfs_fs_devices *fs_devices,
1123			    void *data)
1124{
1125	struct inode *inode;
1126	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1127	int err;
1128
1129	sb->s_maxbytes = MAX_LFS_FILESIZE;
1130	sb->s_magic = BTRFS_SUPER_MAGIC;
1131	sb->s_op = &btrfs_super_ops;
1132	sb->s_d_op = &btrfs_dentry_operations;
1133	sb->s_export_op = &btrfs_export_ops;
1134#ifdef CONFIG_FS_VERITY
1135	sb->s_vop = &btrfs_verityops;
1136#endif
1137	sb->s_xattr = btrfs_xattr_handlers;
1138	sb->s_time_gran = 1;
1139#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1140	sb->s_flags |= SB_POSIXACL;
1141#endif
1142	sb->s_flags |= SB_I_VERSION;
1143	sb->s_iflags |= SB_I_CGROUPWB;
1144
1145	err = super_setup_bdi(sb);
1146	if (err) {
1147		btrfs_err(fs_info, "super_setup_bdi failed");
1148		return err;
1149	}
1150
1151	err = open_ctree(sb, fs_devices, (char *)data);
1152	if (err) {
1153		btrfs_err(fs_info, "open_ctree failed");
1154		return err;
1155	}
1156
1157	inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1158	if (IS_ERR(inode)) {
1159		err = PTR_ERR(inode);
 
1160		goto fail_close;
1161	}
1162
1163	sb->s_root = d_make_root(inode);
1164	if (!sb->s_root) {
1165		err = -ENOMEM;
1166		goto fail_close;
1167	}
1168
1169	sb->s_flags |= SB_ACTIVE;
1170	return 0;
1171
1172fail_close:
1173	close_ctree(fs_info);
1174	return err;
1175}
1176
1177int btrfs_sync_fs(struct super_block *sb, int wait)
1178{
1179	struct btrfs_trans_handle *trans;
1180	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1181	struct btrfs_root *root = fs_info->tree_root;
1182
1183	trace_btrfs_sync_fs(fs_info, wait);
1184
1185	if (!wait) {
1186		filemap_flush(fs_info->btree_inode->i_mapping);
1187		return 0;
1188	}
1189
1190	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1191
1192	trans = btrfs_attach_transaction_barrier(root);
1193	if (IS_ERR(trans)) {
1194		/* no transaction, don't bother */
1195		if (PTR_ERR(trans) == -ENOENT) {
1196			/*
1197			 * Exit unless we have some pending changes
1198			 * that need to go through commit
1199			 */
1200			if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1201				      &fs_info->flags))
1202				return 0;
1203			/*
1204			 * A non-blocking test if the fs is frozen. We must not
1205			 * start a new transaction here otherwise a deadlock
1206			 * happens. The pending operations are delayed to the
1207			 * next commit after thawing.
1208			 */
1209			if (sb_start_write_trylock(sb))
1210				sb_end_write(sb);
1211			else
1212				return 0;
1213			trans = btrfs_start_transaction(root, 0);
1214		}
1215		if (IS_ERR(trans))
1216			return PTR_ERR(trans);
1217	}
1218	return btrfs_commit_transaction(trans);
1219}
1220
1221static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1222{
1223	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1224	*printed = true;
1225}
1226
1227static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1228{
1229	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1230	const char *compress_type;
1231	const char *subvol_name;
1232	bool printed = false;
1233
1234	if (btrfs_test_opt(info, DEGRADED))
1235		seq_puts(seq, ",degraded");
1236	if (btrfs_test_opt(info, NODATASUM))
1237		seq_puts(seq, ",nodatasum");
1238	if (btrfs_test_opt(info, NODATACOW))
1239		seq_puts(seq, ",nodatacow");
1240	if (btrfs_test_opt(info, NOBARRIER))
1241		seq_puts(seq, ",nobarrier");
1242	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1243		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1244	if (info->thread_pool_size !=  min_t(unsigned long,
1245					     num_online_cpus() + 2, 8))
1246		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1247	if (btrfs_test_opt(info, COMPRESS)) {
1248		compress_type = btrfs_compress_type2str(info->compress_type);
1249		if (btrfs_test_opt(info, FORCE_COMPRESS))
1250			seq_printf(seq, ",compress-force=%s", compress_type);
1251		else
1252			seq_printf(seq, ",compress=%s", compress_type);
1253		if (info->compress_level)
1254			seq_printf(seq, ":%d", info->compress_level);
1255	}
1256	if (btrfs_test_opt(info, NOSSD))
1257		seq_puts(seq, ",nossd");
1258	if (btrfs_test_opt(info, SSD_SPREAD))
1259		seq_puts(seq, ",ssd_spread");
1260	else if (btrfs_test_opt(info, SSD))
1261		seq_puts(seq, ",ssd");
1262	if (btrfs_test_opt(info, NOTREELOG))
1263		seq_puts(seq, ",notreelog");
1264	if (btrfs_test_opt(info, NOLOGREPLAY))
1265		print_rescue_option(seq, "nologreplay", &printed);
1266	if (btrfs_test_opt(info, USEBACKUPROOT))
1267		print_rescue_option(seq, "usebackuproot", &printed);
1268	if (btrfs_test_opt(info, IGNOREBADROOTS))
1269		print_rescue_option(seq, "ignorebadroots", &printed);
1270	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1271		print_rescue_option(seq, "ignoredatacsums", &printed);
1272	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1273		seq_puts(seq, ",flushoncommit");
1274	if (btrfs_test_opt(info, DISCARD_SYNC))
1275		seq_puts(seq, ",discard");
1276	if (btrfs_test_opt(info, DISCARD_ASYNC))
1277		seq_puts(seq, ",discard=async");
1278	if (!(info->sb->s_flags & SB_POSIXACL))
1279		seq_puts(seq, ",noacl");
1280	if (btrfs_free_space_cache_v1_active(info))
1281		seq_puts(seq, ",space_cache");
1282	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1283		seq_puts(seq, ",space_cache=v2");
1284	else
1285		seq_puts(seq, ",nospace_cache");
1286	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1287		seq_puts(seq, ",rescan_uuid_tree");
1288	if (btrfs_test_opt(info, CLEAR_CACHE))
1289		seq_puts(seq, ",clear_cache");
1290	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1291		seq_puts(seq, ",user_subvol_rm_allowed");
1292	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1293		seq_puts(seq, ",enospc_debug");
1294	if (btrfs_test_opt(info, AUTO_DEFRAG))
1295		seq_puts(seq, ",autodefrag");
1296	if (btrfs_test_opt(info, SKIP_BALANCE))
1297		seq_puts(seq, ",skip_balance");
1298#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1299	if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1300		seq_puts(seq, ",check_int_data");
1301	else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1302		seq_puts(seq, ",check_int");
1303	if (info->check_integrity_print_mask)
1304		seq_printf(seq, ",check_int_print_mask=%d",
1305				info->check_integrity_print_mask);
1306#endif
1307	if (info->metadata_ratio)
1308		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1309	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1310		seq_puts(seq, ",fatal_errors=panic");
1311	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1312		seq_printf(seq, ",commit=%u", info->commit_interval);
1313#ifdef CONFIG_BTRFS_DEBUG
1314	if (btrfs_test_opt(info, FRAGMENT_DATA))
1315		seq_puts(seq, ",fragment=data");
1316	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1317		seq_puts(seq, ",fragment=metadata");
1318#endif
1319	if (btrfs_test_opt(info, REF_VERIFY))
1320		seq_puts(seq, ",ref_verify");
1321	seq_printf(seq, ",subvolid=%llu",
1322		  BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1323	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1324			BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1325	if (!IS_ERR(subvol_name)) {
1326		seq_puts(seq, ",subvol=");
1327		seq_escape(seq, subvol_name, " \t\n\\");
1328		kfree(subvol_name);
1329	}
1330	return 0;
1331}
1332
1333static int btrfs_test_super(struct super_block *s, void *data)
1334{
1335	struct btrfs_fs_info *p = data;
1336	struct btrfs_fs_info *fs_info = btrfs_sb(s);
1337
1338	return fs_info->fs_devices == p->fs_devices;
1339}
1340
1341static int btrfs_set_super(struct super_block *s, void *data)
1342{
1343	int err = set_anon_super(s, data);
1344	if (!err)
1345		s->s_fs_info = data;
1346	return err;
1347}
1348
1349/*
1350 * subvolumes are identified by ino 256
1351 */
1352static inline int is_subvolume_inode(struct inode *inode)
1353{
1354	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1355		return 1;
1356	return 0;
1357}
1358
1359static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1360				   struct vfsmount *mnt)
1361{
1362	struct dentry *root;
1363	int ret;
1364
1365	if (!subvol_name) {
1366		if (!subvol_objectid) {
1367			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1368							  &subvol_objectid);
1369			if (ret) {
1370				root = ERR_PTR(ret);
1371				goto out;
1372			}
1373		}
1374		subvol_name = btrfs_get_subvol_name_from_objectid(
1375					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1376		if (IS_ERR(subvol_name)) {
1377			root = ERR_CAST(subvol_name);
1378			subvol_name = NULL;
1379			goto out;
1380		}
1381
1382	}
1383
1384	root = mount_subtree(mnt, subvol_name);
1385	/* mount_subtree() drops our reference on the vfsmount. */
1386	mnt = NULL;
1387
1388	if (!IS_ERR(root)) {
1389		struct super_block *s = root->d_sb;
1390		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1391		struct inode *root_inode = d_inode(root);
1392		u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1393
1394		ret = 0;
1395		if (!is_subvolume_inode(root_inode)) {
1396			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1397			       subvol_name);
1398			ret = -EINVAL;
1399		}
1400		if (subvol_objectid && root_objectid != subvol_objectid) {
1401			/*
1402			 * This will also catch a race condition where a
1403			 * subvolume which was passed by ID is renamed and
1404			 * another subvolume is renamed over the old location.
1405			 */
1406			btrfs_err(fs_info,
1407				  "subvol '%s' does not match subvolid %llu",
1408				  subvol_name, subvol_objectid);
1409			ret = -EINVAL;
1410		}
1411		if (ret) {
1412			dput(root);
1413			root = ERR_PTR(ret);
1414			deactivate_locked_super(s);
1415		}
1416	}
1417
1418out:
1419	mntput(mnt);
1420	kfree(subvol_name);
1421	return root;
1422}
1423
1424/*
1425 * Find a superblock for the given device / mount point.
1426 *
1427 * Note: This is based on mount_bdev from fs/super.c with a few additions
1428 *       for multiple device setup.  Make sure to keep it in sync.
1429 */
1430static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1431		int flags, const char *device_name, void *data)
1432{
1433	struct block_device *bdev = NULL;
1434	struct super_block *s;
1435	struct btrfs_device *device = NULL;
1436	struct btrfs_fs_devices *fs_devices = NULL;
1437	struct btrfs_fs_info *fs_info = NULL;
1438	void *new_sec_opts = NULL;
1439	fmode_t mode = FMODE_READ;
1440	int error = 0;
1441
1442	if (!(flags & SB_RDONLY))
1443		mode |= FMODE_WRITE;
1444
1445	if (data) {
1446		error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1447		if (error)
1448			return ERR_PTR(error);
1449	}
1450
1451	/*
1452	 * Setup a dummy root and fs_info for test/set super.  This is because
1453	 * we don't actually fill this stuff out until open_ctree, but we need
1454	 * then open_ctree will properly initialize the file system specific
1455	 * settings later.  btrfs_init_fs_info initializes the static elements
1456	 * of the fs_info (locks and such) to make cleanup easier if we find a
1457	 * superblock with our given fs_devices later on at sget() time.
1458	 */
1459	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1460	if (!fs_info) {
1461		error = -ENOMEM;
1462		goto error_sec_opts;
1463	}
1464	btrfs_init_fs_info(fs_info);
1465
1466	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1467	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1468	if (!fs_info->super_copy || !fs_info->super_for_commit) {
1469		error = -ENOMEM;
1470		goto error_fs_info;
1471	}
1472
1473	mutex_lock(&uuid_mutex);
1474	error = btrfs_parse_device_options(data, mode, fs_type);
1475	if (error) {
1476		mutex_unlock(&uuid_mutex);
1477		goto error_fs_info;
1478	}
1479
1480	device = btrfs_scan_one_device(device_name, mode, fs_type);
1481	if (IS_ERR(device)) {
1482		mutex_unlock(&uuid_mutex);
1483		error = PTR_ERR(device);
1484		goto error_fs_info;
1485	}
1486
1487	fs_devices = device->fs_devices;
1488	fs_info->fs_devices = fs_devices;
1489
1490	error = btrfs_open_devices(fs_devices, mode, fs_type);
1491	mutex_unlock(&uuid_mutex);
1492	if (error)
1493		goto error_fs_info;
1494
1495	if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1496		error = -EACCES;
1497		goto error_close_devices;
1498	}
1499
1500	bdev = fs_devices->latest_dev->bdev;
1501	s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1502		 fs_info);
1503	if (IS_ERR(s)) {
1504		error = PTR_ERR(s);
1505		goto error_close_devices;
1506	}
1507
1508	if (s->s_root) {
1509		btrfs_close_devices(fs_devices);
1510		btrfs_free_fs_info(fs_info);
1511		if ((flags ^ s->s_flags) & SB_RDONLY)
1512			error = -EBUSY;
1513	} else {
1514		snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1515		shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
1516					s->s_id);
1517		btrfs_sb(s)->bdev_holder = fs_type;
1518		if (!strstr(crc32c_impl(), "generic"))
1519			set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1520		error = btrfs_fill_super(s, fs_devices, data);
1521	}
1522	if (!error)
1523		error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1524	security_free_mnt_opts(&new_sec_opts);
1525	if (error) {
1526		deactivate_locked_super(s);
1527		return ERR_PTR(error);
1528	}
1529
1530	return dget(s->s_root);
1531
1532error_close_devices:
1533	btrfs_close_devices(fs_devices);
1534error_fs_info:
1535	btrfs_free_fs_info(fs_info);
1536error_sec_opts:
1537	security_free_mnt_opts(&new_sec_opts);
1538	return ERR_PTR(error);
1539}
1540
1541/*
1542 * Mount function which is called by VFS layer.
1543 *
1544 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1545 * which needs vfsmount* of device's root (/).  This means device's root has to
1546 * be mounted internally in any case.
1547 *
1548 * Operation flow:
1549 *   1. Parse subvol id related options for later use in mount_subvol().
1550 *
1551 *   2. Mount device's root (/) by calling vfs_kern_mount().
1552 *
1553 *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1554 *      first place. In order to avoid calling btrfs_mount() again, we use
1555 *      different file_system_type which is not registered to VFS by
1556 *      register_filesystem() (btrfs_root_fs_type). As a result,
1557 *      btrfs_mount_root() is called. The return value will be used by
1558 *      mount_subtree() in mount_subvol().
1559 *
1560 *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1561 *      "btrfs subvolume set-default", mount_subvol() is called always.
1562 */
1563static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1564		const char *device_name, void *data)
1565{
1566	struct vfsmount *mnt_root;
1567	struct dentry *root;
1568	char *subvol_name = NULL;
1569	u64 subvol_objectid = 0;
1570	int error = 0;
1571
1572	error = btrfs_parse_subvol_options(data, &subvol_name,
1573					&subvol_objectid);
1574	if (error) {
1575		kfree(subvol_name);
1576		return ERR_PTR(error);
1577	}
1578
1579	/* mount device's root (/) */
1580	mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1581	if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1582		if (flags & SB_RDONLY) {
1583			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1584				flags & ~SB_RDONLY, device_name, data);
1585		} else {
1586			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1587				flags | SB_RDONLY, device_name, data);
1588			if (IS_ERR(mnt_root)) {
1589				root = ERR_CAST(mnt_root);
1590				kfree(subvol_name);
1591				goto out;
1592			}
1593
1594			down_write(&mnt_root->mnt_sb->s_umount);
1595			error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1596			up_write(&mnt_root->mnt_sb->s_umount);
1597			if (error < 0) {
1598				root = ERR_PTR(error);
1599				mntput(mnt_root);
1600				kfree(subvol_name);
1601				goto out;
1602			}
1603		}
1604	}
1605	if (IS_ERR(mnt_root)) {
1606		root = ERR_CAST(mnt_root);
1607		kfree(subvol_name);
1608		goto out;
1609	}
1610
1611	/* mount_subvol() will free subvol_name and mnt_root */
1612	root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1613
1614out:
1615	return root;
1616}
1617
1618static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1619				     u32 new_pool_size, u32 old_pool_size)
1620{
1621	if (new_pool_size == old_pool_size)
1622		return;
1623
1624	fs_info->thread_pool_size = new_pool_size;
1625
1626	btrfs_info(fs_info, "resize thread pool %d -> %d",
1627	       old_pool_size, new_pool_size);
1628
1629	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1630	btrfs_workqueue_set_max(fs_info->hipri_workers, new_pool_size);
1631	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1632	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
 
 
1633	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1634	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1635	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1636}
1637
1638static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1639				       unsigned long old_opts, int flags)
1640{
1641	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1642	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1643	     (flags & SB_RDONLY))) {
1644		/* wait for any defraggers to finish */
1645		wait_event(fs_info->transaction_wait,
1646			   (atomic_read(&fs_info->defrag_running) == 0));
1647		if (flags & SB_RDONLY)
1648			sync_filesystem(fs_info->sb);
1649	}
1650}
1651
1652static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1653					 unsigned long old_opts)
1654{
1655	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1656
1657	/*
1658	 * We need to cleanup all defragable inodes if the autodefragment is
1659	 * close or the filesystem is read only.
1660	 */
1661	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1662	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1663		btrfs_cleanup_defrag_inodes(fs_info);
1664	}
1665
1666	/* If we toggled discard async */
1667	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1668	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1669		btrfs_discard_resume(fs_info);
1670	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1671		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1672		btrfs_discard_cleanup(fs_info);
1673
1674	/* If we toggled space cache */
1675	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1676		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1677}
1678
1679static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1680{
1681	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1682	unsigned old_flags = sb->s_flags;
1683	unsigned long old_opts = fs_info->mount_opt;
1684	unsigned long old_compress_type = fs_info->compress_type;
1685	u64 old_max_inline = fs_info->max_inline;
1686	u32 old_thread_pool_size = fs_info->thread_pool_size;
1687	u32 old_metadata_ratio = fs_info->metadata_ratio;
1688	int ret;
1689
1690	sync_filesystem(sb);
1691	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
 
 
 
1692
1693	if (data) {
1694		void *new_sec_opts = NULL;
1695
1696		ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1697		if (!ret)
1698			ret = security_sb_remount(sb, new_sec_opts);
1699		security_free_mnt_opts(&new_sec_opts);
1700		if (ret)
1701			goto restore;
1702	}
1703
1704	ret = btrfs_parse_options(fs_info, data, *flags);
 
 
 
 
 
 
 
 
 
 
 
1705	if (ret)
1706		goto restore;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1707
1708	ret = btrfs_check_features(fs_info, !(*flags & SB_RDONLY));
1709	if (ret < 0)
1710		goto restore;
1711
1712	btrfs_remount_begin(fs_info, old_opts, *flags);
1713	btrfs_resize_thread_pool(fs_info,
1714		fs_info->thread_pool_size, old_thread_pool_size);
 
1715
1716	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1717	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1718	    (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1719		btrfs_warn(fs_info,
1720		"remount supports changing free space tree only from ro to rw");
1721		/* Make sure free space cache options match the state on disk */
1722		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1723			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1724			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1725		}
1726		if (btrfs_free_space_cache_v1_active(fs_info)) {
1727			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1728			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1729		}
1730	}
1731
1732	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1733		goto out;
 
 
 
 
 
1734
1735	if (*flags & SB_RDONLY) {
1736		/*
1737		 * this also happens on 'umount -rf' or on shutdown, when
1738		 * the filesystem is busy.
1739		 */
1740		cancel_work_sync(&fs_info->async_reclaim_work);
1741		cancel_work_sync(&fs_info->async_data_reclaim_work);
 
 
 
 
1742
1743		btrfs_discard_cleanup(fs_info);
 
 
 
 
 
 
 
1744
1745		/* wait for the uuid_scan task to finish */
1746		down(&fs_info->uuid_tree_rescan_sem);
1747		/* avoid complains from lockdep et al. */
1748		up(&fs_info->uuid_tree_rescan_sem);
1749
1750		btrfs_set_sb_rdonly(sb);
 
 
 
 
 
 
1751
1752		/*
1753		 * Setting SB_RDONLY will put the cleaner thread to
1754		 * sleep at the next loop if it's already active.
1755		 * If it's already asleep, we'll leave unused block
1756		 * groups on disk until we're mounted read-write again
1757		 * unless we clean them up here.
1758		 */
1759		btrfs_delete_unused_bgs(fs_info);
1760
1761		/*
1762		 * The cleaner task could be already running before we set the
1763		 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1764		 * We must make sure that after we finish the remount, i.e. after
1765		 * we call btrfs_commit_super(), the cleaner can no longer start
1766		 * a transaction - either because it was dropping a dead root,
1767		 * running delayed iputs or deleting an unused block group (the
1768		 * cleaner picked a block group from the list of unused block
1769		 * groups before we were able to in the previous call to
1770		 * btrfs_delete_unused_bgs()).
1771		 */
1772		wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1773			    TASK_UNINTERRUPTIBLE);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1774
1775		/*
1776		 * We've set the superblock to RO mode, so we might have made
1777		 * the cleaner task sleep without running all pending delayed
1778		 * iputs. Go through all the delayed iputs here, so that if an
1779		 * unmount happens without remounting RW we don't end up at
1780		 * finishing close_ctree() with a non-empty list of delayed
1781		 * iputs.
1782		 */
1783		btrfs_run_delayed_iputs(fs_info);
1784
1785		btrfs_dev_replace_suspend_for_unmount(fs_info);
1786		btrfs_scrub_cancel(fs_info);
1787		btrfs_pause_balance(fs_info);
 
 
 
 
 
1788
1789		/*
1790		 * Pause the qgroup rescan worker if it is running. We don't want
1791		 * it to be still running after we are in RO mode, as after that,
1792		 * by the time we unmount, it might have left a transaction open,
1793		 * so we would leak the transaction and/or crash.
1794		 */
1795		btrfs_qgroup_wait_for_completion(fs_info, false);
1796
1797		ret = btrfs_commit_super(fs_info);
1798		if (ret)
1799			goto restore;
1800	} else {
1801		if (BTRFS_FS_ERROR(fs_info)) {
1802			btrfs_err(fs_info,
1803				"Remounting read-write after error is not allowed");
1804			ret = -EINVAL;
1805			goto restore;
1806		}
1807		if (fs_info->fs_devices->rw_devices == 0) {
1808			ret = -EACCES;
1809			goto restore;
1810		}
1811
1812		if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1813			btrfs_warn(fs_info,
1814		"too many missing devices, writable remount is not allowed");
1815			ret = -EACCES;
1816			goto restore;
1817		}
1818
1819		if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1820			btrfs_warn(fs_info,
1821		"mount required to replay tree-log, cannot remount read-write");
1822			ret = -EINVAL;
1823			goto restore;
1824		}
1825
1826		/*
1827		 * NOTE: when remounting with a change that does writes, don't
1828		 * put it anywhere above this point, as we are not sure to be
1829		 * safe to write until we pass the above checks.
1830		 */
1831		ret = btrfs_start_pre_rw_mount(fs_info);
1832		if (ret)
1833			goto restore;
1834
1835		btrfs_clear_sb_rdonly(sb);
 
 
 
1836
1837		set_bit(BTRFS_FS_OPEN, &fs_info->flags);
 
 
 
 
 
 
 
 
 
 
 
 
 
1838	}
1839out:
 
 
 
 
 
 
 
 
1840	/*
1841	 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
1842	 * since the absence of the flag means it can be toggled off by remount.
 
1843	 */
1844	*flags |= SB_I_VERSION;
 
1845
 
1846	wake_up_process(fs_info->transaction_kthread);
1847	btrfs_remount_cleanup(fs_info, old_opts);
1848	btrfs_clear_oneshot_options(fs_info);
1849	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1850
1851	return 0;
1852
1853restore:
1854	/* We've hit an error - don't reset SB_RDONLY */
1855	if (sb_rdonly(sb))
1856		old_flags |= SB_RDONLY;
1857	if (!(old_flags & SB_RDONLY))
1858		clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
1859	sb->s_flags = old_flags;
1860	fs_info->mount_opt = old_opts;
1861	fs_info->compress_type = old_compress_type;
1862	fs_info->max_inline = old_max_inline;
1863	btrfs_resize_thread_pool(fs_info,
1864		old_thread_pool_size, fs_info->thread_pool_size);
1865	fs_info->metadata_ratio = old_metadata_ratio;
1866	btrfs_remount_cleanup(fs_info, old_opts);
1867	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1868
1869	return ret;
1870}
1871
1872/* Used to sort the devices by max_avail(descending sort) */
1873static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1874{
1875	const struct btrfs_device_info *dev_info1 = a;
1876	const struct btrfs_device_info *dev_info2 = b;
1877
1878	if (dev_info1->max_avail > dev_info2->max_avail)
1879		return -1;
1880	else if (dev_info1->max_avail < dev_info2->max_avail)
1881		return 1;
1882	return 0;
1883}
1884
1885/*
1886 * sort the devices by max_avail, in which max free extent size of each device
1887 * is stored.(Descending Sort)
1888 */
1889static inline void btrfs_descending_sort_devices(
1890					struct btrfs_device_info *devices,
1891					size_t nr_devices)
1892{
1893	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1894	     btrfs_cmp_device_free_bytes, NULL);
1895}
1896
1897/*
1898 * The helper to calc the free space on the devices that can be used to store
1899 * file data.
1900 */
1901static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1902					      u64 *free_bytes)
1903{
1904	struct btrfs_device_info *devices_info;
1905	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1906	struct btrfs_device *device;
1907	u64 type;
1908	u64 avail_space;
1909	u64 min_stripe_size;
1910	int num_stripes = 1;
1911	int i = 0, nr_devices;
1912	const struct btrfs_raid_attr *rattr;
1913
1914	/*
1915	 * We aren't under the device list lock, so this is racy-ish, but good
1916	 * enough for our purposes.
1917	 */
1918	nr_devices = fs_info->fs_devices->open_devices;
1919	if (!nr_devices) {
1920		smp_mb();
1921		nr_devices = fs_info->fs_devices->open_devices;
1922		ASSERT(nr_devices);
1923		if (!nr_devices) {
1924			*free_bytes = 0;
1925			return 0;
1926		}
1927	}
1928
1929	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1930			       GFP_KERNEL);
1931	if (!devices_info)
1932		return -ENOMEM;
1933
1934	/* calc min stripe number for data space allocation */
1935	type = btrfs_data_alloc_profile(fs_info);
1936	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1937
1938	if (type & BTRFS_BLOCK_GROUP_RAID0)
1939		num_stripes = nr_devices;
1940	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1941		num_stripes = rattr->ncopies;
1942	else if (type & BTRFS_BLOCK_GROUP_RAID10)
1943		num_stripes = 4;
1944
1945	/* Adjust for more than 1 stripe per device */
1946	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1947
1948	rcu_read_lock();
1949	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1950		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1951						&device->dev_state) ||
1952		    !device->bdev ||
1953		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1954			continue;
1955
1956		if (i >= nr_devices)
1957			break;
1958
1959		avail_space = device->total_bytes - device->bytes_used;
1960
1961		/* align with stripe_len */
1962		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1963
1964		/*
1965		 * Ensure we have at least min_stripe_size on top of the
1966		 * reserved space on the device.
1967		 */
1968		if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1969			continue;
1970
1971		avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1972
1973		devices_info[i].dev = device;
1974		devices_info[i].max_avail = avail_space;
1975
1976		i++;
1977	}
1978	rcu_read_unlock();
1979
1980	nr_devices = i;
1981
1982	btrfs_descending_sort_devices(devices_info, nr_devices);
1983
1984	i = nr_devices - 1;
1985	avail_space = 0;
1986	while (nr_devices >= rattr->devs_min) {
1987		num_stripes = min(num_stripes, nr_devices);
1988
1989		if (devices_info[i].max_avail >= min_stripe_size) {
1990			int j;
1991			u64 alloc_size;
1992
1993			avail_space += devices_info[i].max_avail * num_stripes;
1994			alloc_size = devices_info[i].max_avail;
1995			for (j = i + 1 - num_stripes; j <= i; j++)
1996				devices_info[j].max_avail -= alloc_size;
1997		}
1998		i--;
1999		nr_devices--;
2000	}
2001
2002	kfree(devices_info);
2003	*free_bytes = avail_space;
2004	return 0;
2005}
2006
2007/*
2008 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2009 *
2010 * If there's a redundant raid level at DATA block groups, use the respective
2011 * multiplier to scale the sizes.
2012 *
2013 * Unused device space usage is based on simulating the chunk allocator
2014 * algorithm that respects the device sizes and order of allocations.  This is
2015 * a close approximation of the actual use but there are other factors that may
2016 * change the result (like a new metadata chunk).
2017 *
2018 * If metadata is exhausted, f_bavail will be 0.
2019 */
2020static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2021{
2022	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2023	struct btrfs_super_block *disk_super = fs_info->super_copy;
2024	struct btrfs_space_info *found;
2025	u64 total_used = 0;
2026	u64 total_free_data = 0;
2027	u64 total_free_meta = 0;
2028	u32 bits = fs_info->sectorsize_bits;
2029	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2030	unsigned factor = 1;
2031	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2032	int ret;
2033	u64 thresh = 0;
2034	int mixed = 0;
2035
2036	list_for_each_entry(found, &fs_info->space_info, list) {
2037		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2038			int i;
2039
2040			total_free_data += found->disk_total - found->disk_used;
2041			total_free_data -=
2042				btrfs_account_ro_block_groups_free_space(found);
2043
2044			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2045				if (!list_empty(&found->block_groups[i]))
2046					factor = btrfs_bg_type_to_factor(
2047						btrfs_raid_array[i].bg_flag);
2048			}
2049		}
2050
2051		/*
2052		 * Metadata in mixed block goup profiles are accounted in data
2053		 */
2054		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2055			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2056				mixed = 1;
2057			else
2058				total_free_meta += found->disk_total -
2059					found->disk_used;
2060		}
2061
2062		total_used += found->disk_used;
2063	}
2064
2065	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2066	buf->f_blocks >>= bits;
2067	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2068
2069	/* Account global block reserve as used, it's in logical size already */
2070	spin_lock(&block_rsv->lock);
2071	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
2072	if (buf->f_bfree >= block_rsv->size >> bits)
2073		buf->f_bfree -= block_rsv->size >> bits;
2074	else
2075		buf->f_bfree = 0;
2076	spin_unlock(&block_rsv->lock);
2077
2078	buf->f_bavail = div_u64(total_free_data, factor);
2079	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2080	if (ret)
2081		return ret;
2082	buf->f_bavail += div_u64(total_free_data, factor);
2083	buf->f_bavail = buf->f_bavail >> bits;
2084
2085	/*
2086	 * We calculate the remaining metadata space minus global reserve. If
2087	 * this is (supposedly) smaller than zero, there's no space. But this
2088	 * does not hold in practice, the exhausted state happens where's still
2089	 * some positive delta. So we apply some guesswork and compare the
2090	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2091	 *
2092	 * We probably cannot calculate the exact threshold value because this
2093	 * depends on the internal reservations requested by various
2094	 * operations, so some operations that consume a few metadata will
2095	 * succeed even if the Avail is zero. But this is better than the other
2096	 * way around.
2097	 */
2098	thresh = SZ_4M;
2099
2100	/*
2101	 * We only want to claim there's no available space if we can no longer
2102	 * allocate chunks for our metadata profile and our global reserve will
2103	 * not fit in the free metadata space.  If we aren't ->full then we
2104	 * still can allocate chunks and thus are fine using the currently
2105	 * calculated f_bavail.
2106	 */
2107	if (!mixed && block_rsv->space_info->full &&
2108	    total_free_meta - thresh < block_rsv->size)
2109		buf->f_bavail = 0;
2110
2111	buf->f_type = BTRFS_SUPER_MAGIC;
2112	buf->f_bsize = dentry->d_sb->s_blocksize;
2113	buf->f_namelen = BTRFS_NAME_LEN;
2114
2115	/* We treat it as constant endianness (it doesn't matter _which_)
2116	   because we want the fsid to come out the same whether mounted
2117	   on a big-endian or little-endian host */
2118	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2119	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2120	/* Mask in the root object ID too, to disambiguate subvols */
2121	buf->f_fsid.val[0] ^=
2122		BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2123	buf->f_fsid.val[1] ^=
2124		BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2125
2126	return 0;
2127}
2128
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2129static void btrfs_kill_super(struct super_block *sb)
2130{
2131	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2132	kill_anon_super(sb);
2133	btrfs_free_fs_info(fs_info);
2134}
2135
2136static struct file_system_type btrfs_fs_type = {
2137	.owner		= THIS_MODULE,
2138	.name		= "btrfs",
2139	.mount		= btrfs_mount,
2140	.kill_sb	= btrfs_kill_super,
2141	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2142};
2143
2144static struct file_system_type btrfs_root_fs_type = {
2145	.owner		= THIS_MODULE,
2146	.name		= "btrfs",
2147	.mount		= btrfs_mount_root,
2148	.kill_sb	= btrfs_kill_super,
2149	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2150};
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2151
2152MODULE_ALIAS_FS("btrfs");
2153
2154static int btrfs_control_open(struct inode *inode, struct file *file)
2155{
2156	/*
2157	 * The control file's private_data is used to hold the
2158	 * transaction when it is started and is used to keep
2159	 * track of whether a transaction is already in progress.
2160	 */
2161	file->private_data = NULL;
2162	return 0;
2163}
2164
2165/*
2166 * Used by /dev/btrfs-control for devices ioctls.
2167 */
2168static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2169				unsigned long arg)
2170{
2171	struct btrfs_ioctl_vol_args *vol;
2172	struct btrfs_device *device = NULL;
2173	dev_t devt = 0;
2174	int ret = -ENOTTY;
2175
2176	if (!capable(CAP_SYS_ADMIN))
2177		return -EPERM;
2178
2179	vol = memdup_user((void __user *)arg, sizeof(*vol));
2180	if (IS_ERR(vol))
2181		return PTR_ERR(vol);
2182	vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2183
2184	switch (cmd) {
2185	case BTRFS_IOC_SCAN_DEV:
2186		mutex_lock(&uuid_mutex);
2187		device = btrfs_scan_one_device(vol->name, FMODE_READ,
2188					       &btrfs_root_fs_type);
 
 
 
2189		ret = PTR_ERR_OR_ZERO(device);
2190		mutex_unlock(&uuid_mutex);
2191		break;
2192	case BTRFS_IOC_FORGET_DEV:
2193		if (vol->name[0] != 0) {
2194			ret = lookup_bdev(vol->name, &devt);
2195			if (ret)
2196				break;
2197		}
2198		ret = btrfs_forget_devices(devt);
2199		break;
2200	case BTRFS_IOC_DEVICES_READY:
2201		mutex_lock(&uuid_mutex);
2202		device = btrfs_scan_one_device(vol->name, FMODE_READ,
2203					       &btrfs_root_fs_type);
2204		if (IS_ERR(device)) {
 
 
 
2205			mutex_unlock(&uuid_mutex);
2206			ret = PTR_ERR(device);
2207			break;
2208		}
2209		ret = !(device->fs_devices->num_devices ==
2210			device->fs_devices->total_devices);
2211		mutex_unlock(&uuid_mutex);
2212		break;
2213	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2214		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2215		break;
2216	}
2217
2218	kfree(vol);
2219	return ret;
2220}
2221
2222static int btrfs_freeze(struct super_block *sb)
2223{
2224	struct btrfs_trans_handle *trans;
2225	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2226	struct btrfs_root *root = fs_info->tree_root;
2227
2228	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2229	/*
2230	 * We don't need a barrier here, we'll wait for any transaction that
2231	 * could be in progress on other threads (and do delayed iputs that
2232	 * we want to avoid on a frozen filesystem), or do the commit
2233	 * ourselves.
2234	 */
2235	trans = btrfs_attach_transaction_barrier(root);
2236	if (IS_ERR(trans)) {
2237		/* no transaction, don't bother */
2238		if (PTR_ERR(trans) == -ENOENT)
2239			return 0;
2240		return PTR_ERR(trans);
2241	}
2242	return btrfs_commit_transaction(trans);
2243}
2244
2245static int check_dev_super(struct btrfs_device *dev)
2246{
2247	struct btrfs_fs_info *fs_info = dev->fs_info;
2248	struct btrfs_super_block *sb;
 
2249	u16 csum_type;
2250	int ret = 0;
2251
2252	/* This should be called with fs still frozen. */
2253	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2254
2255	/* Missing dev, no need to check. */
2256	if (!dev->bdev)
2257		return 0;
2258
2259	/* Only need to check the primary super block. */
2260	sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2261	if (IS_ERR(sb))
2262		return PTR_ERR(sb);
2263
2264	/* Verify the checksum. */
2265	csum_type = btrfs_super_csum_type(sb);
2266	if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2267		btrfs_err(fs_info, "csum type changed, has %u expect %u",
2268			  csum_type, btrfs_super_csum_type(fs_info->super_copy));
2269		ret = -EUCLEAN;
2270		goto out;
2271	}
2272
2273	if (btrfs_check_super_csum(fs_info, sb)) {
2274		btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2275		ret = -EUCLEAN;
2276		goto out;
2277	}
2278
2279	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2280	ret = btrfs_validate_super(fs_info, sb, 0);
2281	if (ret < 0)
2282		goto out;
2283
2284	if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
 
2285		btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2286			btrfs_super_generation(sb),
2287			fs_info->last_trans_committed);
2288		ret = -EUCLEAN;
2289		goto out;
2290	}
2291out:
2292	btrfs_release_disk_super(sb);
2293	return ret;
2294}
2295
2296static int btrfs_unfreeze(struct super_block *sb)
2297{
2298	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2299	struct btrfs_device *device;
2300	int ret = 0;
2301
2302	/*
2303	 * Make sure the fs is not changed by accident (like hibernation then
2304	 * modified by other OS).
2305	 * If we found anything wrong, we mark the fs error immediately.
2306	 *
2307	 * And since the fs is frozen, no one can modify the fs yet, thus
2308	 * we don't need to hold device_list_mutex.
2309	 */
2310	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2311		ret = check_dev_super(device);
2312		if (ret < 0) {
2313			btrfs_handle_fs_error(fs_info, ret,
2314				"super block on devid %llu got modified unexpectedly",
2315				device->devid);
2316			break;
2317		}
2318	}
2319	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2320
2321	/*
2322	 * We still return 0, to allow VFS layer to unfreeze the fs even the
2323	 * above checks failed. Since the fs is either fine or read-only, we're
2324	 * safe to continue, without causing further damage.
2325	 */
2326	return 0;
2327}
2328
2329static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2330{
2331	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2332
2333	/*
2334	 * There should be always a valid pointer in latest_dev, it may be stale
2335	 * for a short moment in case it's being deleted but still valid until
2336	 * the end of RCU grace period.
2337	 */
2338	rcu_read_lock();
2339	seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2340	rcu_read_unlock();
2341
2342	return 0;
2343}
2344
2345static const struct super_operations btrfs_super_ops = {
2346	.drop_inode	= btrfs_drop_inode,
2347	.evict_inode	= btrfs_evict_inode,
2348	.put_super	= btrfs_put_super,
2349	.sync_fs	= btrfs_sync_fs,
2350	.show_options	= btrfs_show_options,
2351	.show_devname	= btrfs_show_devname,
2352	.alloc_inode	= btrfs_alloc_inode,
2353	.destroy_inode	= btrfs_destroy_inode,
2354	.free_inode	= btrfs_free_inode,
2355	.statfs		= btrfs_statfs,
2356	.remount_fs	= btrfs_remount,
2357	.freeze_fs	= btrfs_freeze,
2358	.unfreeze_fs	= btrfs_unfreeze,
2359};
2360
2361static const struct file_operations btrfs_ctl_fops = {
2362	.open = btrfs_control_open,
2363	.unlocked_ioctl	 = btrfs_control_ioctl,
2364	.compat_ioctl = compat_ptr_ioctl,
2365	.owner	 = THIS_MODULE,
2366	.llseek = noop_llseek,
2367};
2368
2369static struct miscdevice btrfs_misc = {
2370	.minor		= BTRFS_MINOR,
2371	.name		= "btrfs-control",
2372	.fops		= &btrfs_ctl_fops
2373};
2374
2375MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2376MODULE_ALIAS("devname:btrfs-control");
2377
2378static int __init btrfs_interface_init(void)
2379{
2380	return misc_register(&btrfs_misc);
2381}
2382
2383static __cold void btrfs_interface_exit(void)
2384{
2385	misc_deregister(&btrfs_misc);
2386}
2387
2388static int __init btrfs_print_mod_info(void)
2389{
2390	static const char options[] = ""
2391#ifdef CONFIG_BTRFS_DEBUG
2392			", debug=on"
2393#endif
2394#ifdef CONFIG_BTRFS_ASSERT
2395			", assert=on"
2396#endif
2397#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2398			", integrity-checker=on"
2399#endif
2400#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2401			", ref-verify=on"
2402#endif
2403#ifdef CONFIG_BLK_DEV_ZONED
2404			", zoned=yes"
2405#else
2406			", zoned=no"
2407#endif
2408#ifdef CONFIG_FS_VERITY
2409			", fsverity=yes"
2410#else
2411			", fsverity=no"
2412#endif
2413			;
2414	pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2415	return 0;
2416}
2417
2418static int register_btrfs(void)
2419{
2420	return register_filesystem(&btrfs_fs_type);
2421}
2422
2423static void unregister_btrfs(void)
2424{
2425	unregister_filesystem(&btrfs_fs_type);
2426}
2427
2428/* Helper structure for long init/exit functions. */
2429struct init_sequence {
2430	int (*init_func)(void);
2431	/* Can be NULL if the init_func doesn't need cleanup. */
2432	void (*exit_func)(void);
2433};
2434
2435static const struct init_sequence mod_init_seq[] = {
2436	{
2437		.init_func = btrfs_props_init,
2438		.exit_func = NULL,
2439	}, {
2440		.init_func = btrfs_init_sysfs,
2441		.exit_func = btrfs_exit_sysfs,
2442	}, {
2443		.init_func = btrfs_init_compress,
2444		.exit_func = btrfs_exit_compress,
2445	}, {
2446		.init_func = btrfs_init_cachep,
2447		.exit_func = btrfs_destroy_cachep,
2448	}, {
2449		.init_func = btrfs_transaction_init,
2450		.exit_func = btrfs_transaction_exit,
2451	}, {
2452		.init_func = btrfs_ctree_init,
2453		.exit_func = btrfs_ctree_exit,
2454	}, {
2455		.init_func = btrfs_free_space_init,
2456		.exit_func = btrfs_free_space_exit,
2457	}, {
2458		.init_func = extent_state_init_cachep,
2459		.exit_func = extent_state_free_cachep,
2460	}, {
2461		.init_func = extent_buffer_init_cachep,
2462		.exit_func = extent_buffer_free_cachep,
2463	}, {
2464		.init_func = btrfs_bioset_init,
2465		.exit_func = btrfs_bioset_exit,
2466	}, {
2467		.init_func = extent_map_init,
2468		.exit_func = extent_map_exit,
2469	}, {
2470		.init_func = ordered_data_init,
2471		.exit_func = ordered_data_exit,
2472	}, {
2473		.init_func = btrfs_delayed_inode_init,
2474		.exit_func = btrfs_delayed_inode_exit,
2475	}, {
2476		.init_func = btrfs_auto_defrag_init,
2477		.exit_func = btrfs_auto_defrag_exit,
2478	}, {
2479		.init_func = btrfs_delayed_ref_init,
2480		.exit_func = btrfs_delayed_ref_exit,
2481	}, {
2482		.init_func = btrfs_prelim_ref_init,
2483		.exit_func = btrfs_prelim_ref_exit,
2484	}, {
2485		.init_func = btrfs_interface_init,
2486		.exit_func = btrfs_interface_exit,
2487	}, {
2488		.init_func = btrfs_print_mod_info,
2489		.exit_func = NULL,
2490	}, {
2491		.init_func = btrfs_run_sanity_tests,
2492		.exit_func = NULL,
2493	}, {
2494		.init_func = register_btrfs,
2495		.exit_func = unregister_btrfs,
2496	}
2497};
2498
2499static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2500
2501static __always_inline void btrfs_exit_btrfs_fs(void)
2502{
2503	int i;
2504
2505	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2506		if (!mod_init_result[i])
2507			continue;
2508		if (mod_init_seq[i].exit_func)
2509			mod_init_seq[i].exit_func();
2510		mod_init_result[i] = false;
2511	}
2512}
2513
2514static void __exit exit_btrfs_fs(void)
2515{
2516	btrfs_exit_btrfs_fs();
2517	btrfs_cleanup_fs_uuids();
2518}
2519
2520static int __init init_btrfs_fs(void)
2521{
2522	int ret;
2523	int i;
2524
2525	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2526		ASSERT(!mod_init_result[i]);
2527		ret = mod_init_seq[i].init_func();
2528		if (ret < 0) {
2529			btrfs_exit_btrfs_fs();
2530			return ret;
2531		}
2532		mod_init_result[i] = true;
2533	}
2534	return 0;
2535}
2536
2537late_initcall(init_btrfs_fs);
2538module_exit(exit_btrfs_fs)
2539
2540MODULE_LICENSE("GPL");
2541MODULE_SOFTDEP("pre: crc32c");
2542MODULE_SOFTDEP("pre: xxhash64");
2543MODULE_SOFTDEP("pre: sha256");
2544MODULE_SOFTDEP("pre: blake2b-256");