Linux Audio

Check our new training course

Yocto / OpenEmbedded training

Feb 10-13, 2025
Register
Loading...
v6.13.7
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * Copyright (C) 2007 Oracle.  All rights reserved.
   4 */
   5
   6#include <linux/blkdev.h>
   7#include <linux/module.h>
 
   8#include <linux/fs.h>
   9#include <linux/pagemap.h>
  10#include <linux/highmem.h>
  11#include <linux/time.h>
  12#include <linux/init.h>
  13#include <linux/seq_file.h>
  14#include <linux/string.h>
  15#include <linux/backing-dev.h>
  16#include <linux/mount.h>
 
 
  17#include <linux/writeback.h>
  18#include <linux/statfs.h>
  19#include <linux/compat.h>
  20#include <linux/parser.h>
  21#include <linux/ctype.h>
  22#include <linux/namei.h>
  23#include <linux/miscdevice.h>
  24#include <linux/magic.h>
  25#include <linux/slab.h>
 
  26#include <linux/ratelimit.h>
  27#include <linux/crc32c.h>
  28#include <linux/btrfs.h>
  29#include <linux/security.h>
  30#include <linux/fs_parser.h>
  31#include "messages.h"
  32#include "delayed-inode.h"
  33#include "ctree.h"
  34#include "disk-io.h"
  35#include "transaction.h"
  36#include "btrfs_inode.h"
  37#include "direct-io.h"
  38#include "props.h"
  39#include "xattr.h"
  40#include "bio.h"
  41#include "export.h"
  42#include "compression.h"
 
  43#include "dev-replace.h"
  44#include "free-space-cache.h"
  45#include "backref.h"
  46#include "space-info.h"
  47#include "sysfs.h"
  48#include "zoned.h"
  49#include "tests/btrfs-tests.h"
  50#include "block-group.h"
  51#include "discard.h"
  52#include "qgroup.h"
  53#include "raid56.h"
  54#include "fs.h"
  55#include "accessors.h"
  56#include "defrag.h"
  57#include "dir-item.h"
  58#include "ioctl.h"
  59#include "scrub.h"
  60#include "verity.h"
  61#include "super.h"
  62#include "extent-tree.h"
  63#define CREATE_TRACE_POINTS
  64#include <trace/events/btrfs.h>
  65
  66static const struct super_operations btrfs_super_ops;
 
 
 
 
 
 
 
 
 
  67static struct file_system_type btrfs_fs_type;
 
  68
  69static void btrfs_put_super(struct super_block *sb)
 
 
  70{
  71	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
  72
  73	btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
  74	close_ctree(fs_info);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  75}
  76
  77/* Store the mount options related information. */
  78struct btrfs_fs_context {
  79	char *subvol_name;
  80	u64 subvol_objectid;
  81	u64 max_inline;
  82	u32 commit_interval;
  83	u32 metadata_ratio;
  84	u32 thread_pool_size;
  85	unsigned long long mount_opt;
  86	unsigned long compress_type:4;
  87	unsigned int compress_level;
  88	refcount_t refs;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  89};
  90
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  91enum {
  92	Opt_acl,
  93	Opt_clear_cache,
  94	Opt_commit_interval,
  95	Opt_compress,
  96	Opt_compress_force,
  97	Opt_compress_force_type,
  98	Opt_compress_type,
  99	Opt_degraded,
 100	Opt_device,
 101	Opt_fatal_errors,
 102	Opt_flushoncommit,
 
 103	Opt_max_inline,
 104	Opt_barrier,
 105	Opt_datacow,
 106	Opt_datasum,
 107	Opt_defrag,
 108	Opt_discard,
 109	Opt_discard_mode,
 
 110	Opt_ratio,
 111	Opt_rescan_uuid_tree,
 112	Opt_skip_balance,
 113	Opt_space_cache,
 114	Opt_space_cache_version,
 115	Opt_ssd,
 116	Opt_ssd_spread,
 117	Opt_subvol,
 118	Opt_subvol_empty,
 119	Opt_subvolid,
 120	Opt_thread_pool,
 121	Opt_treelog,
 
 122	Opt_user_subvol_rm_allowed,
 123	Opt_norecovery,
 124
 125	/* Rescue options */
 126	Opt_rescue,
 127	Opt_usebackuproot,
 128	Opt_nologreplay,
 129
 130	/* Debugging options */
 131	Opt_enospc_debug,
 
 
 
 132#ifdef CONFIG_BTRFS_DEBUG
 133	Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 134#endif
 135#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 136	Opt_ref_verify,
 137#endif
 138	Opt_err,
 139};
 140
 141enum {
 142	Opt_fatal_errors_panic,
 143	Opt_fatal_errors_bug,
 144};
 145
 146static const struct constant_table btrfs_parameter_fatal_errors[] = {
 147	{ "panic", Opt_fatal_errors_panic },
 148	{ "bug", Opt_fatal_errors_bug },
 149	{}
 150};
 151
 152enum {
 153	Opt_discard_sync,
 154	Opt_discard_async,
 155};
 156
 157static const struct constant_table btrfs_parameter_discard[] = {
 158	{ "sync", Opt_discard_sync },
 159	{ "async", Opt_discard_async },
 160	{}
 161};
 162
 163enum {
 164	Opt_space_cache_v1,
 165	Opt_space_cache_v2,
 166};
 167
 168static const struct constant_table btrfs_parameter_space_cache[] = {
 169	{ "v1", Opt_space_cache_v1 },
 170	{ "v2", Opt_space_cache_v2 },
 171	{}
 172};
 173
 174enum {
 175	Opt_rescue_usebackuproot,
 176	Opt_rescue_nologreplay,
 177	Opt_rescue_ignorebadroots,
 178	Opt_rescue_ignoredatacsums,
 179	Opt_rescue_ignoremetacsums,
 180	Opt_rescue_ignoresuperflags,
 181	Opt_rescue_parameter_all,
 182};
 183
 184static const struct constant_table btrfs_parameter_rescue[] = {
 185	{ "usebackuproot", Opt_rescue_usebackuproot },
 186	{ "nologreplay", Opt_rescue_nologreplay },
 187	{ "ignorebadroots", Opt_rescue_ignorebadroots },
 188	{ "ibadroots", Opt_rescue_ignorebadroots },
 189	{ "ignoredatacsums", Opt_rescue_ignoredatacsums },
 190	{ "ignoremetacsums", Opt_rescue_ignoremetacsums},
 191	{ "ignoresuperflags", Opt_rescue_ignoresuperflags},
 192	{ "idatacsums", Opt_rescue_ignoredatacsums },
 193	{ "imetacsums", Opt_rescue_ignoremetacsums},
 194	{ "isuperflags", Opt_rescue_ignoresuperflags},
 195	{ "all", Opt_rescue_parameter_all },
 196	{}
 197};
 198
 199#ifdef CONFIG_BTRFS_DEBUG
 200enum {
 201	Opt_fragment_parameter_data,
 202	Opt_fragment_parameter_metadata,
 203	Opt_fragment_parameter_all,
 204};
 205
 206static const struct constant_table btrfs_parameter_fragment[] = {
 207	{ "data", Opt_fragment_parameter_data },
 208	{ "metadata", Opt_fragment_parameter_metadata },
 209	{ "all", Opt_fragment_parameter_all },
 210	{}
 211};
 212#endif
 213
 214static const struct fs_parameter_spec btrfs_fs_parameters[] = {
 215	fsparam_flag_no("acl", Opt_acl),
 216	fsparam_flag_no("autodefrag", Opt_defrag),
 217	fsparam_flag_no("barrier", Opt_barrier),
 218	fsparam_flag("clear_cache", Opt_clear_cache),
 219	fsparam_u32("commit", Opt_commit_interval),
 220	fsparam_flag("compress", Opt_compress),
 221	fsparam_string("compress", Opt_compress_type),
 222	fsparam_flag("compress-force", Opt_compress_force),
 223	fsparam_string("compress-force", Opt_compress_force_type),
 224	fsparam_flag_no("datacow", Opt_datacow),
 225	fsparam_flag_no("datasum", Opt_datasum),
 226	fsparam_flag("degraded", Opt_degraded),
 227	fsparam_string("device", Opt_device),
 228	fsparam_flag_no("discard", Opt_discard),
 229	fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
 230	fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
 231	fsparam_flag_no("flushoncommit", Opt_flushoncommit),
 232	fsparam_string("max_inline", Opt_max_inline),
 233	fsparam_u32("metadata_ratio", Opt_ratio),
 234	fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
 235	fsparam_flag("skip_balance", Opt_skip_balance),
 236	fsparam_flag_no("space_cache", Opt_space_cache),
 237	fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
 238	fsparam_flag_no("ssd", Opt_ssd),
 239	fsparam_flag_no("ssd_spread", Opt_ssd_spread),
 240	fsparam_string("subvol", Opt_subvol),
 241	fsparam_flag("subvol=", Opt_subvol_empty),
 242	fsparam_u64("subvolid", Opt_subvolid),
 243	fsparam_u32("thread_pool", Opt_thread_pool),
 244	fsparam_flag_no("treelog", Opt_treelog),
 245	fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
 246
 247	/* Rescue options. */
 248	fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
 249	/* Deprecated, with alias rescue=nologreplay */
 250	__fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
 251	/* Deprecated, with alias rescue=usebackuproot */
 252	__fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
 253	/* For compatibility only, alias for "rescue=nologreplay". */
 254	fsparam_flag("norecovery", Opt_norecovery),
 255
 256	/* Debugging options. */
 257	fsparam_flag_no("enospc_debug", Opt_enospc_debug),
 
 
 
 
 258#ifdef CONFIG_BTRFS_DEBUG
 259	fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
 
 
 260#endif
 261#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 262	fsparam_flag("ref_verify", Opt_ref_verify),
 263#endif
 264	{}
 265};
 266
 267/* No support for restricting writes to btrfs devices yet... */
 268static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
 269{
 270	return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
 271}
 272
 273static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
 274{
 275	struct btrfs_fs_context *ctx = fc->fs_private;
 276	struct fs_parse_result result;
 277	int opt;
 278
 279	opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
 280	if (opt < 0)
 281		return opt;
 282
 283	switch (opt) {
 284	case Opt_degraded:
 285		btrfs_set_opt(ctx->mount_opt, DEGRADED);
 286		break;
 287	case Opt_subvol_empty:
 288		/*
 289		 * This exists because we used to allow it on accident, so we're
 290		 * keeping it to maintain ABI.  See 37becec95ac3 ("Btrfs: allow
 291		 * empty subvol= again").
 292		 */
 293		break;
 294	case Opt_subvol:
 295		kfree(ctx->subvol_name);
 296		ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
 297		if (!ctx->subvol_name)
 298			return -ENOMEM;
 299		break;
 300	case Opt_subvolid:
 301		ctx->subvol_objectid = result.uint_64;
 
 302
 303		/* subvolid=0 means give me the original fs_tree. */
 304		if (!ctx->subvol_objectid)
 305			ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
 306		break;
 307	case Opt_device: {
 308		struct btrfs_device *device;
 309		blk_mode_t mode = btrfs_open_mode(fc);
 310
 311		mutex_lock(&uuid_mutex);
 312		device = btrfs_scan_one_device(param->string, mode, false);
 313		mutex_unlock(&uuid_mutex);
 314		if (IS_ERR(device))
 315			return PTR_ERR(device);
 316		break;
 317	}
 318	case Opt_datasum:
 319		if (result.negated) {
 320			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 321		} else {
 322			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 323			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 324		}
 325		break;
 326	case Opt_datacow:
 327		if (result.negated) {
 328			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 329			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 330			btrfs_set_opt(ctx->mount_opt, NODATACOW);
 331			btrfs_set_opt(ctx->mount_opt, NODATASUM);
 332		} else {
 333			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 334		}
 335		break;
 336	case Opt_compress_force:
 337	case Opt_compress_force_type:
 338		btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
 339		fallthrough;
 340	case Opt_compress:
 341	case Opt_compress_type:
 342		/*
 343		 * Provide the same semantics as older kernels that don't use fs
 344		 * context, specifying the "compress" option clears
 345		 * "force-compress" without the need to pass
 346		 * "compress-force=[no|none]" before specifying "compress".
 347		 */
 348		if (opt != Opt_compress_force && opt != Opt_compress_force_type)
 349			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 350
 351		if (opt == Opt_compress || opt == Opt_compress_force) {
 352			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 353			ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 354			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 355			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 356			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 357		} else if (strncmp(param->string, "zlib", 4) == 0) {
 358			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
 359			ctx->compress_level =
 360				btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
 361							 param->string + 4);
 362			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 363			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 364			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 365		} else if (strncmp(param->string, "lzo", 3) == 0) {
 366			ctx->compress_type = BTRFS_COMPRESS_LZO;
 367			ctx->compress_level = 0;
 368			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 369			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 370			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 371		} else if (strncmp(param->string, "zstd", 4) == 0) {
 372			ctx->compress_type = BTRFS_COMPRESS_ZSTD;
 373			ctx->compress_level =
 374				btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
 375							 param->string + 4);
 376			btrfs_set_opt(ctx->mount_opt, COMPRESS);
 377			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
 378			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
 379		} else if (strncmp(param->string, "no", 2) == 0) {
 380			ctx->compress_level = 0;
 381			ctx->compress_type = 0;
 382			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
 383			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
 384		} else {
 385			btrfs_err(NULL, "unrecognized compression value %s",
 386				  param->string);
 387			return -EINVAL;
 388		}
 389		break;
 390	case Opt_ssd:
 391		if (result.negated) {
 392			btrfs_set_opt(ctx->mount_opt, NOSSD);
 393			btrfs_clear_opt(ctx->mount_opt, SSD);
 394			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 395		} else {
 396			btrfs_set_opt(ctx->mount_opt, SSD);
 397			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 398		}
 399		break;
 400	case Opt_ssd_spread:
 401		if (result.negated) {
 402			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
 403		} else {
 404			btrfs_set_opt(ctx->mount_opt, SSD);
 405			btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
 406			btrfs_clear_opt(ctx->mount_opt, NOSSD);
 407		}
 408		break;
 409	case Opt_barrier:
 410		if (result.negated)
 411			btrfs_set_opt(ctx->mount_opt, NOBARRIER);
 412		else
 413			btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
 414		break;
 415	case Opt_thread_pool:
 416		if (result.uint_32 == 0) {
 417			btrfs_err(NULL, "invalid value 0 for thread_pool");
 418			return -EINVAL;
 419		}
 420		ctx->thread_pool_size = result.uint_32;
 421		break;
 422	case Opt_max_inline:
 423		ctx->max_inline = memparse(param->string, NULL);
 424		break;
 425	case Opt_acl:
 426		if (result.negated) {
 427			fc->sb_flags &= ~SB_POSIXACL;
 428		} else {
 
 
 
 
 
 
 429#ifdef CONFIG_BTRFS_FS_POSIX_ACL
 430			fc->sb_flags |= SB_POSIXACL;
 
 431#else
 432			btrfs_err(NULL, "support for ACL not compiled in");
 433			return -EINVAL;
 
 434#endif
 435		}
 436		/*
 437		 * VFS limits the ability to toggle ACL on and off via remount,
 438		 * despite every file system allowing this.  This seems to be
 439		 * an oversight since we all do, but it'll fail if we're
 440		 * remounting.  So don't set the mask here, we'll check it in
 441		 * btrfs_reconfigure and do the toggling ourselves.
 442		 */
 443		if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
 444			fc->sb_flags_mask |= SB_POSIXACL;
 445		break;
 446	case Opt_treelog:
 447		if (result.negated)
 448			btrfs_set_opt(ctx->mount_opt, NOTREELOG);
 449		else
 450			btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
 451		break;
 452	case Opt_nologreplay:
 453		btrfs_warn(NULL,
 454		"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
 455		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 456		break;
 457	case Opt_norecovery:
 458		btrfs_info(NULL,
 459"'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
 460		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 461		break;
 462	case Opt_flushoncommit:
 463		if (result.negated)
 464			btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
 465		else
 466			btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
 467		break;
 468	case Opt_ratio:
 469		ctx->metadata_ratio = result.uint_32;
 470		break;
 471	case Opt_discard:
 472		if (result.negated) {
 473			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 474			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 475			btrfs_set_opt(ctx->mount_opt, NODISCARD);
 476		} else {
 477			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 478			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 479		}
 480		break;
 481	case Opt_discard_mode:
 482		switch (result.uint_32) {
 483		case Opt_discard_sync:
 484			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
 485			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
 486			break;
 487		case Opt_discard_async:
 488			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
 489			btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
 490			break;
 491		default:
 492			btrfs_err(NULL, "unrecognized discard mode value %s",
 493				  param->key);
 494			return -EINVAL;
 495		}
 496		btrfs_clear_opt(ctx->mount_opt, NODISCARD);
 497		break;
 498	case Opt_space_cache:
 499		if (result.negated) {
 500			btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
 501			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 502			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 503		} else {
 504			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 505			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 506		}
 507		break;
 508	case Opt_space_cache_version:
 509		switch (result.uint_32) {
 510		case Opt_space_cache_v1:
 511			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
 512			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
 513			break;
 514		case Opt_space_cache_v2:
 515			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
 516			btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
 517			break;
 518		default:
 519			btrfs_err(NULL, "unrecognized space_cache value %s",
 520				  param->key);
 521			return -EINVAL;
 522		}
 523		break;
 524	case Opt_rescan_uuid_tree:
 525		btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
 526		break;
 527	case Opt_clear_cache:
 528		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 529		break;
 530	case Opt_user_subvol_rm_allowed:
 531		btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
 532		break;
 533	case Opt_enospc_debug:
 534		if (result.negated)
 535			btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
 536		else
 537			btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
 538		break;
 539	case Opt_defrag:
 540		if (result.negated)
 541			btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
 542		else
 543			btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
 544		break;
 545	case Opt_usebackuproot:
 546		btrfs_warn(NULL,
 547			   "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
 548		btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 549
 550		/* If we're loading the backup roots we can't trust the space cache. */
 551		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
 552		break;
 553	case Opt_skip_balance:
 554		btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
 555		break;
 556	case Opt_fatal_errors:
 557		switch (result.uint_32) {
 558		case Opt_fatal_errors_panic:
 559			btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 560			break;
 561		case Opt_fatal_errors_bug:
 562			btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
 
 
 563			break;
 564		default:
 565			btrfs_err(NULL, "unrecognized fatal_errors value %s",
 566				  param->key);
 567			return -EINVAL;
 568		}
 569		break;
 570	case Opt_commit_interval:
 571		ctx->commit_interval = result.uint_32;
 572		if (ctx->commit_interval == 0)
 573			ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
 574		break;
 575	case Opt_rescue:
 576		switch (result.uint_32) {
 577		case Opt_rescue_usebackuproot:
 578			btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
 579			break;
 580		case Opt_rescue_nologreplay:
 581			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 582			break;
 583		case Opt_rescue_ignorebadroots:
 584			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 585			break;
 586		case Opt_rescue_ignoredatacsums:
 587			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 588			break;
 589		case Opt_rescue_ignoremetacsums:
 590			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
 591			break;
 592		case Opt_rescue_ignoresuperflags:
 593			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
 594			break;
 595		case Opt_rescue_parameter_all:
 596			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
 597			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
 598			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
 599			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
 600			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 601			break;
 602		default:
 603			btrfs_info(NULL, "unrecognized rescue option '%s'",
 604				   param->key);
 605			return -EINVAL;
 606		}
 607		break;
 608#ifdef CONFIG_BTRFS_DEBUG
 609	case Opt_fragment:
 610		switch (result.uint_32) {
 611		case Opt_fragment_parameter_all:
 612			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 613			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 614			break;
 615		case Opt_fragment_parameter_metadata:
 616			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
 
 
 617			break;
 618		case Opt_fragment_parameter_data:
 619			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
 
 620			break;
 621		default:
 622			btrfs_info(NULL, "unrecognized fragment option '%s'",
 623				   param->key);
 624			return -EINVAL;
 625		}
 626		break;
 627#endif
 628#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 629	case Opt_ref_verify:
 630		btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
 631		break;
 
 632#endif
 633	default:
 634		btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
 635		return -EINVAL;
 
 
 
 
 636	}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 637
 638	return 0;
 
 
 
 
 
 639}
 640
 641/*
 642 * Some options only have meaning at mount time and shouldn't persist across
 643 * remounts, or be displayed. Clear these at the end of mount and remount code
 644 * paths.
 
 645 */
 646static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
 
 647{
 648	btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
 649	btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
 650	btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
 651}
 652
 653static bool check_ro_option(const struct btrfs_fs_info *fs_info,
 654			    unsigned long long mount_opt, unsigned long long opt,
 655			    const char *opt_name)
 656{
 657	if (mount_opt & opt) {
 658		btrfs_err(fs_info, "%s must be used with ro mount option",
 659			  opt_name);
 660		return true;
 661	}
 662	return false;
 663}
 664
 665bool btrfs_check_options(const struct btrfs_fs_info *info,
 666			 unsigned long long *mount_opt,
 667			 unsigned long flags)
 668{
 669	bool ret = true;
 
 
 
 670
 671	if (!(flags & SB_RDONLY) &&
 672	    (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
 673	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
 674	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
 675	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
 676	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
 677		ret = false;
 678
 679	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 680	    !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
 681	    !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
 682		btrfs_err(info, "cannot disable free-space-tree");
 683		ret = false;
 684	}
 685	if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
 686	     !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
 687		btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
 688		ret = false;
 689	}
 690
 691	if (btrfs_check_mountopts_zoned(info, mount_opt))
 692		ret = false;
 693
 694	if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
 695		if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
 696			btrfs_info(info, "disk space caching is enabled");
 697			btrfs_warn(info,
 698"space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
 699		}
 700		if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
 701			btrfs_info(info, "using free-space-tree");
 702	}
 703
 704	return ret;
 
 
 705}
 706
 707/*
 708 * This is subtle, we only call this during open_ctree().  We need to pre-load
 709 * the mount options with the on-disk settings.  Before the new mount API took
 710 * effect we would do this on mount and remount.  With the new mount API we'll
 711 * only do this on the initial mount.
 712 *
 713 * This isn't a change in behavior, because we're using the current state of the
 714 * file system to set the current mount options.  If you mounted with special
 715 * options to disable these features and then remounted we wouldn't revert the
 716 * settings, because mounting without these features cleared the on-disk
 717 * settings, so this being called on re-mount is not needed.
 718 */
 719void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
 
 720{
 721	if (fs_info->sectorsize < PAGE_SIZE) {
 722		btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
 723		if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
 724			btrfs_info(fs_info,
 725				   "forcing free space tree for sector size %u with page size %lu",
 726				   fs_info->sectorsize, PAGE_SIZE);
 727			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 728		}
 729	}
 730
 731	/*
 732	 * At this point our mount options are populated, so we only mess with
 733	 * these settings if we don't have any settings already.
 734	 */
 735	if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
 736		return;
 737
 738	if (btrfs_is_zoned(fs_info) &&
 739	    btrfs_free_space_cache_v1_active(fs_info)) {
 740		btrfs_info(fs_info, "zoned: clearing existing space cache");
 741		btrfs_set_super_cache_generation(fs_info->super_copy, 0);
 742		return;
 743	}
 744
 745	if (btrfs_test_opt(fs_info, SPACE_CACHE))
 746		return;
 
 
 747
 748	if (btrfs_test_opt(fs_info, NOSPACECACHE))
 749		return;
 
 
 
 
 
 
 
 
 
 
 
 
 750
 751	/*
 752	 * At this point we don't have explicit options set by the user, set
 753	 * them ourselves based on the state of the file system.
 754	 */
 755	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
 756		btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
 757	else if (btrfs_free_space_cache_v1_active(fs_info))
 758		btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
 759}
 760
 761static void set_device_specific_options(struct btrfs_fs_info *fs_info)
 762{
 763	if (!btrfs_test_opt(fs_info, NOSSD) &&
 764	    !fs_info->fs_devices->rotating)
 765		btrfs_set_opt(fs_info->mount_opt, SSD);
 
 
 
 
 766
 767	/*
 768	 * For devices supporting discard turn on discard=async automatically,
 769	 * unless it's already set or disabled. This could be turned off by
 770	 * nodiscard for the same mount.
 771	 *
 772	 * The zoned mode piggy backs on the discard functionality for
 773	 * resetting a zone. There is no reason to delay the zone reset as it is
 774	 * fast enough. So, do not enable async discard for zoned mode.
 775	 */
 776	if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
 777	      btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
 778	      btrfs_test_opt(fs_info, NODISCARD)) &&
 779	    fs_info->fs_devices->discardable &&
 780	    !btrfs_is_zoned(fs_info))
 781		btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
 782}
 783
 784char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 785					  u64 subvol_objectid)
 786{
 787	struct btrfs_root *root = fs_info->tree_root;
 788	struct btrfs_root *fs_root = NULL;
 789	struct btrfs_root_ref *root_ref;
 790	struct btrfs_inode_ref *inode_ref;
 791	struct btrfs_key key;
 792	struct btrfs_path *path = NULL;
 793	char *name = NULL, *ptr;
 794	u64 dirid;
 795	int len;
 796	int ret;
 797
 798	path = btrfs_alloc_path();
 799	if (!path) {
 800		ret = -ENOMEM;
 801		goto err;
 802	}
 
 803
 804	name = kmalloc(PATH_MAX, GFP_KERNEL);
 805	if (!name) {
 806		ret = -ENOMEM;
 807		goto err;
 808	}
 809	ptr = name + PATH_MAX - 1;
 810	ptr[0] = '\0';
 811
 812	/*
 813	 * Walk up the subvolume trees in the tree of tree roots by root
 814	 * backrefs until we hit the top-level subvolume.
 815	 */
 816	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
 817		key.objectid = subvol_objectid;
 818		key.type = BTRFS_ROOT_BACKREF_KEY;
 819		key.offset = (u64)-1;
 820
 821		ret = btrfs_search_backwards(root, &key, path);
 822		if (ret < 0) {
 823			goto err;
 824		} else if (ret > 0) {
 825			ret = -ENOENT;
 826			goto err;
 
 
 
 
 
 
 827		}
 828
 
 829		subvol_objectid = key.offset;
 830
 831		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
 832					  struct btrfs_root_ref);
 833		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
 834		ptr -= len + 1;
 835		if (ptr < name) {
 836			ret = -ENAMETOOLONG;
 837			goto err;
 838		}
 839		read_extent_buffer(path->nodes[0], ptr + 1,
 840				   (unsigned long)(root_ref + 1), len);
 841		ptr[0] = '/';
 842		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
 843		btrfs_release_path(path);
 844
 845		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
 
 
 
 846		if (IS_ERR(fs_root)) {
 847			ret = PTR_ERR(fs_root);
 848			fs_root = NULL;
 849			goto err;
 850		}
 851
 852		/*
 853		 * Walk up the filesystem tree by inode refs until we hit the
 854		 * root directory.
 855		 */
 856		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
 857			key.objectid = dirid;
 858			key.type = BTRFS_INODE_REF_KEY;
 859			key.offset = (u64)-1;
 860
 861			ret = btrfs_search_backwards(fs_root, &key, path);
 862			if (ret < 0) {
 863				goto err;
 864			} else if (ret > 0) {
 865				ret = -ENOENT;
 866				goto err;
 
 
 
 
 
 
 867			}
 868
 
 869			dirid = key.offset;
 870
 871			inode_ref = btrfs_item_ptr(path->nodes[0],
 872						   path->slots[0],
 873						   struct btrfs_inode_ref);
 874			len = btrfs_inode_ref_name_len(path->nodes[0],
 875						       inode_ref);
 876			ptr -= len + 1;
 877			if (ptr < name) {
 878				ret = -ENAMETOOLONG;
 879				goto err;
 880			}
 881			read_extent_buffer(path->nodes[0], ptr + 1,
 882					   (unsigned long)(inode_ref + 1), len);
 883			ptr[0] = '/';
 884			btrfs_release_path(path);
 885		}
 886		btrfs_put_root(fs_root);
 887		fs_root = NULL;
 888	}
 889
 890	btrfs_free_path(path);
 891	if (ptr == name + PATH_MAX - 1) {
 892		name[0] = '/';
 893		name[1] = '\0';
 894	} else {
 895		memmove(name, ptr, name + PATH_MAX - ptr);
 896	}
 897	return name;
 898
 899err:
 900	btrfs_put_root(fs_root);
 901	btrfs_free_path(path);
 902	kfree(name);
 903	return ERR_PTR(ret);
 904}
 905
 906static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
 907{
 908	struct btrfs_root *root = fs_info->tree_root;
 909	struct btrfs_dir_item *di;
 910	struct btrfs_path *path;
 911	struct btrfs_key location;
 912	struct fscrypt_str name = FSTR_INIT("default", 7);
 913	u64 dir_id;
 914
 915	path = btrfs_alloc_path();
 916	if (!path)
 917		return -ENOMEM;
 
 918
 919	/*
 920	 * Find the "default" dir item which points to the root item that we
 921	 * will mount by default if we haven't been given a specific subvolume
 922	 * to mount.
 923	 */
 924	dir_id = btrfs_super_root_dir(fs_info->super_copy);
 925	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
 926	if (IS_ERR(di)) {
 927		btrfs_free_path(path);
 928		return PTR_ERR(di);
 929	}
 930	if (!di) {
 931		/*
 932		 * Ok the default dir item isn't there.  This is weird since
 933		 * it's always been there, but don't freak out, just try and
 934		 * mount the top-level subvolume.
 935		 */
 936		btrfs_free_path(path);
 937		*objectid = BTRFS_FS_TREE_OBJECTID;
 938		return 0;
 939	}
 940
 941	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 942	btrfs_free_path(path);
 943	*objectid = location.objectid;
 944	return 0;
 945}
 946
 947static int btrfs_fill_super(struct super_block *sb,
 948			    struct btrfs_fs_devices *fs_devices)
 
 949{
 950	struct inode *inode;
 951	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 
 952	int err;
 953
 954	sb->s_maxbytes = MAX_LFS_FILESIZE;
 955	sb->s_magic = BTRFS_SUPER_MAGIC;
 956	sb->s_op = &btrfs_super_ops;
 957	sb->s_d_op = &btrfs_dentry_operations;
 958	sb->s_export_op = &btrfs_export_ops;
 959#ifdef CONFIG_FS_VERITY
 960	sb->s_vop = &btrfs_verityops;
 961#endif
 962	sb->s_xattr = btrfs_xattr_handlers;
 963	sb->s_time_gran = 1;
 
 
 
 
 964	sb->s_iflags |= SB_I_CGROUPWB;
 965
 966	err = super_setup_bdi(sb);
 967	if (err) {
 968		btrfs_err(fs_info, "super_setup_bdi failed");
 969		return err;
 970	}
 971
 972	err = open_ctree(sb, fs_devices);
 973	if (err) {
 974		btrfs_err(fs_info, "open_ctree failed: %d", err);
 975		return err;
 976	}
 977
 978	inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
 
 
 
 979	if (IS_ERR(inode)) {
 980		err = PTR_ERR(inode);
 981		btrfs_handle_fs_error(fs_info, err, NULL);
 982		goto fail_close;
 983	}
 984
 985	sb->s_root = d_make_root(inode);
 986	if (!sb->s_root) {
 987		err = -ENOMEM;
 988		goto fail_close;
 989	}
 990
 
 991	sb->s_flags |= SB_ACTIVE;
 992	return 0;
 993
 994fail_close:
 995	close_ctree(fs_info);
 996	return err;
 997}
 998
 999int btrfs_sync_fs(struct super_block *sb, int wait)
1000{
1001	struct btrfs_trans_handle *trans;
1002	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1003	struct btrfs_root *root = fs_info->tree_root;
1004
1005	trace_btrfs_sync_fs(fs_info, wait);
1006
1007	if (!wait) {
1008		filemap_flush(fs_info->btree_inode->i_mapping);
1009		return 0;
1010	}
1011
1012	btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1013
1014	trans = btrfs_attach_transaction_barrier(root);
1015	if (IS_ERR(trans)) {
1016		/* no transaction, don't bother */
1017		if (PTR_ERR(trans) == -ENOENT) {
1018			/*
1019			 * Exit unless we have some pending changes
1020			 * that need to go through commit
1021			 */
1022			if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1023				      &fs_info->flags))
1024				return 0;
1025			/*
1026			 * A non-blocking test if the fs is frozen. We must not
1027			 * start a new transaction here otherwise a deadlock
1028			 * happens. The pending operations are delayed to the
1029			 * next commit after thawing.
1030			 */
1031			if (sb_start_write_trylock(sb))
1032				sb_end_write(sb);
1033			else
1034				return 0;
1035			trans = btrfs_start_transaction(root, 0);
1036		}
1037		if (IS_ERR(trans))
1038			return PTR_ERR(trans);
1039	}
1040	return btrfs_commit_transaction(trans);
1041}
1042
1043static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1044{
1045	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1046	*printed = true;
1047}
1048
1049static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1050{
1051	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1052	const char *compress_type;
1053	const char *subvol_name;
1054	bool printed = false;
1055
1056	if (btrfs_test_opt(info, DEGRADED))
1057		seq_puts(seq, ",degraded");
1058	if (btrfs_test_opt(info, NODATASUM))
1059		seq_puts(seq, ",nodatasum");
1060	if (btrfs_test_opt(info, NODATACOW))
1061		seq_puts(seq, ",nodatacow");
1062	if (btrfs_test_opt(info, NOBARRIER))
1063		seq_puts(seq, ",nobarrier");
1064	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1065		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1066	if (info->thread_pool_size !=  min_t(unsigned long,
1067					     num_online_cpus() + 2, 8))
1068		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1069	if (btrfs_test_opt(info, COMPRESS)) {
1070		compress_type = btrfs_compress_type2str(info->compress_type);
1071		if (btrfs_test_opt(info, FORCE_COMPRESS))
1072			seq_printf(seq, ",compress-force=%s", compress_type);
1073		else
1074			seq_printf(seq, ",compress=%s", compress_type);
1075		if (info->compress_level)
1076			seq_printf(seq, ":%d", info->compress_level);
1077	}
1078	if (btrfs_test_opt(info, NOSSD))
1079		seq_puts(seq, ",nossd");
1080	if (btrfs_test_opt(info, SSD_SPREAD))
1081		seq_puts(seq, ",ssd_spread");
1082	else if (btrfs_test_opt(info, SSD))
1083		seq_puts(seq, ",ssd");
1084	if (btrfs_test_opt(info, NOTREELOG))
1085		seq_puts(seq, ",notreelog");
1086	if (btrfs_test_opt(info, NOLOGREPLAY))
1087		print_rescue_option(seq, "nologreplay", &printed);
1088	if (btrfs_test_opt(info, USEBACKUPROOT))
1089		print_rescue_option(seq, "usebackuproot", &printed);
1090	if (btrfs_test_opt(info, IGNOREBADROOTS))
1091		print_rescue_option(seq, "ignorebadroots", &printed);
1092	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1093		print_rescue_option(seq, "ignoredatacsums", &printed);
1094	if (btrfs_test_opt(info, IGNOREMETACSUMS))
1095		print_rescue_option(seq, "ignoremetacsums", &printed);
1096	if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1097		print_rescue_option(seq, "ignoresuperflags", &printed);
1098	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1099		seq_puts(seq, ",flushoncommit");
1100	if (btrfs_test_opt(info, DISCARD_SYNC))
1101		seq_puts(seq, ",discard");
1102	if (btrfs_test_opt(info, DISCARD_ASYNC))
1103		seq_puts(seq, ",discard=async");
1104	if (!(info->sb->s_flags & SB_POSIXACL))
1105		seq_puts(seq, ",noacl");
1106	if (btrfs_free_space_cache_v1_active(info))
1107		seq_puts(seq, ",space_cache");
1108	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1109		seq_puts(seq, ",space_cache=v2");
1110	else
1111		seq_puts(seq, ",nospace_cache");
1112	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1113		seq_puts(seq, ",rescan_uuid_tree");
1114	if (btrfs_test_opt(info, CLEAR_CACHE))
1115		seq_puts(seq, ",clear_cache");
1116	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1117		seq_puts(seq, ",user_subvol_rm_allowed");
1118	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1119		seq_puts(seq, ",enospc_debug");
1120	if (btrfs_test_opt(info, AUTO_DEFRAG))
1121		seq_puts(seq, ",autodefrag");
 
 
1122	if (btrfs_test_opt(info, SKIP_BALANCE))
1123		seq_puts(seq, ",skip_balance");
 
 
 
 
 
 
 
 
 
1124	if (info->metadata_ratio)
1125		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1126	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1127		seq_puts(seq, ",fatal_errors=panic");
1128	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1129		seq_printf(seq, ",commit=%u", info->commit_interval);
1130#ifdef CONFIG_BTRFS_DEBUG
1131	if (btrfs_test_opt(info, FRAGMENT_DATA))
1132		seq_puts(seq, ",fragment=data");
1133	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1134		seq_puts(seq, ",fragment=metadata");
1135#endif
1136	if (btrfs_test_opt(info, REF_VERIFY))
1137		seq_puts(seq, ",ref_verify");
1138	seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1139	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1140			btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1141	if (!IS_ERR(subvol_name)) {
1142		seq_puts(seq, ",subvol=");
1143		seq_escape(seq, subvol_name, " \t\n\\");
1144		kfree(subvol_name);
1145	}
1146	return 0;
1147}
1148
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1149/*
1150 * subvolumes are identified by ino 256
1151 */
1152static inline int is_subvolume_inode(struct inode *inode)
1153{
1154	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1155		return 1;
1156	return 0;
1157}
1158
1159static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1160				   struct vfsmount *mnt)
1161{
1162	struct dentry *root;
1163	int ret;
1164
1165	if (!subvol_name) {
1166		if (!subvol_objectid) {
1167			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1168							  &subvol_objectid);
1169			if (ret) {
1170				root = ERR_PTR(ret);
1171				goto out;
1172			}
1173		}
1174		subvol_name = btrfs_get_subvol_name_from_objectid(
1175					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1176		if (IS_ERR(subvol_name)) {
1177			root = ERR_CAST(subvol_name);
1178			subvol_name = NULL;
1179			goto out;
1180		}
1181
1182	}
1183
1184	root = mount_subtree(mnt, subvol_name);
1185	/* mount_subtree() drops our reference on the vfsmount. */
1186	mnt = NULL;
1187
1188	if (!IS_ERR(root)) {
1189		struct super_block *s = root->d_sb;
1190		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1191		struct inode *root_inode = d_inode(root);
1192		u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1193
1194		ret = 0;
1195		if (!is_subvolume_inode(root_inode)) {
1196			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1197			       subvol_name);
1198			ret = -EINVAL;
1199		}
1200		if (subvol_objectid && root_objectid != subvol_objectid) {
1201			/*
1202			 * This will also catch a race condition where a
1203			 * subvolume which was passed by ID is renamed and
1204			 * another subvolume is renamed over the old location.
1205			 */
1206			btrfs_err(fs_info,
1207				  "subvol '%s' does not match subvolid %llu",
1208				  subvol_name, subvol_objectid);
1209			ret = -EINVAL;
1210		}
1211		if (ret) {
1212			dput(root);
1213			root = ERR_PTR(ret);
1214			deactivate_locked_super(s);
1215		}
1216	}
1217
1218out:
1219	mntput(mnt);
1220	kfree(subvol_name);
1221	return root;
1222}
1223
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1224static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1225				     u32 new_pool_size, u32 old_pool_size)
1226{
1227	if (new_pool_size == old_pool_size)
1228		return;
1229
1230	fs_info->thread_pool_size = new_pool_size;
1231
1232	btrfs_info(fs_info, "resize thread pool %d -> %d",
1233	       old_pool_size, new_pool_size);
1234
1235	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1236	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
 
1237	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1238	workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1239	workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
 
 
1240	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1241	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1242	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
 
 
 
 
 
 
 
 
1243}
1244
1245static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1246				       unsigned long long old_opts, int flags)
1247{
1248	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1249	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1250	     (flags & SB_RDONLY))) {
1251		/* wait for any defraggers to finish */
1252		wait_event(fs_info->transaction_wait,
1253			   (atomic_read(&fs_info->defrag_running) == 0));
1254		if (flags & SB_RDONLY)
1255			sync_filesystem(fs_info->sb);
1256	}
1257}
1258
1259static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1260					 unsigned long long old_opts)
1261{
1262	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1263
1264	/*
1265	 * We need to cleanup all defragable inodes if the autodefragment is
1266	 * close or the filesystem is read only.
1267	 */
1268	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1269	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1270		btrfs_cleanup_defrag_inodes(fs_info);
1271	}
1272
1273	/* If we toggled discard async */
1274	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1275	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1276		btrfs_discard_resume(fs_info);
1277	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1278		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1279		btrfs_discard_cleanup(fs_info);
1280
1281	/* If we toggled space cache */
1282	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1283		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1284}
1285
1286static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1287{
 
 
 
 
 
 
 
 
1288	int ret;
1289
1290	if (BTRFS_FS_ERROR(fs_info)) {
1291		btrfs_err(fs_info,
1292			  "remounting read-write after error is not allowed");
1293		return -EINVAL;
1294	}
1295
1296	if (fs_info->fs_devices->rw_devices == 0)
1297		return -EACCES;
1298
1299	if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1300		btrfs_warn(fs_info,
1301			   "too many missing devices, writable remount is not allowed");
1302		return -EACCES;
 
 
 
 
 
 
1303	}
1304
1305	if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1306		btrfs_warn(fs_info,
1307			   "mount required to replay tree-log, cannot remount read-write");
1308		return -EINVAL;
1309	}
1310
1311	/*
1312	 * NOTE: when remounting with a change that does writes, don't put it
1313	 * anywhere above this point, as we are not sure to be safe to write
1314	 * until we pass the above checks.
1315	 */
1316	ret = btrfs_start_pre_rw_mount(fs_info);
1317	if (ret)
1318		return ret;
1319
1320	btrfs_clear_sb_rdonly(fs_info->sb);
1321
1322	set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1323
1324	/*
1325	 * If we've gone from readonly -> read-write, we need to get our
1326	 * sync/async discard lists in the right state.
1327	 */
1328	btrfs_discard_resume(fs_info);
1329
1330	return 0;
1331}
1332
1333static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1334{
1335	/*
1336	 * This also happens on 'umount -rf' or on shutdown, when the
1337	 * filesystem is busy.
1338	 */
1339	cancel_work_sync(&fs_info->async_reclaim_work);
1340	cancel_work_sync(&fs_info->async_data_reclaim_work);
1341
1342	btrfs_discard_cleanup(fs_info);
1343
1344	/* Wait for the uuid_scan task to finish */
1345	down(&fs_info->uuid_tree_rescan_sem);
1346	/* Avoid complains from lockdep et al. */
1347	up(&fs_info->uuid_tree_rescan_sem);
1348
1349	btrfs_set_sb_rdonly(fs_info->sb);
1350
1351	/*
1352	 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1353	 * loop if it's already active.  If it's already asleep, we'll leave
1354	 * unused block groups on disk until we're mounted read-write again
1355	 * unless we clean them up here.
1356	 */
1357	btrfs_delete_unused_bgs(fs_info);
1358
1359	/*
1360	 * The cleaner task could be already running before we set the flag
1361	 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).  We must make
1362	 * sure that after we finish the remount, i.e. after we call
1363	 * btrfs_commit_super(), the cleaner can no longer start a transaction
1364	 * - either because it was dropping a dead root, running delayed iputs
1365	 *   or deleting an unused block group (the cleaner picked a block
1366	 *   group from the list of unused block groups before we were able to
1367	 *   in the previous call to btrfs_delete_unused_bgs()).
1368	 */
1369	wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1370
1371	/*
1372	 * We've set the superblock to RO mode, so we might have made the
1373	 * cleaner task sleep without running all pending delayed iputs. Go
1374	 * through all the delayed iputs here, so that if an unmount happens
1375	 * without remounting RW we don't end up at finishing close_ctree()
1376	 * with a non-empty list of delayed iputs.
1377	 */
1378	btrfs_run_delayed_iputs(fs_info);
1379
1380	btrfs_dev_replace_suspend_for_unmount(fs_info);
1381	btrfs_scrub_cancel(fs_info);
1382	btrfs_pause_balance(fs_info);
1383
1384	/*
1385	 * Pause the qgroup rescan worker if it is running. We don't want it to
1386	 * be still running after we are in RO mode, as after that, by the time
1387	 * we unmount, it might have left a transaction open, so we would leak
1388	 * the transaction and/or crash.
1389	 */
1390	btrfs_qgroup_wait_for_completion(fs_info, false);
1391
1392	return btrfs_commit_super(fs_info);
1393}
 
 
1394
1395static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1396{
1397	fs_info->max_inline = ctx->max_inline;
1398	fs_info->commit_interval = ctx->commit_interval;
1399	fs_info->metadata_ratio = ctx->metadata_ratio;
1400	fs_info->thread_pool_size = ctx->thread_pool_size;
1401	fs_info->mount_opt = ctx->mount_opt;
1402	fs_info->compress_type = ctx->compress_type;
1403	fs_info->compress_level = ctx->compress_level;
1404}
1405
1406static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1407{
1408	ctx->max_inline = fs_info->max_inline;
1409	ctx->commit_interval = fs_info->commit_interval;
1410	ctx->metadata_ratio = fs_info->metadata_ratio;
1411	ctx->thread_pool_size = fs_info->thread_pool_size;
1412	ctx->mount_opt = fs_info->mount_opt;
1413	ctx->compress_type = fs_info->compress_type;
1414	ctx->compress_level = fs_info->compress_level;
1415}
1416
1417#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...)			\
1418do {										\
1419	if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1420	    btrfs_raw_test_opt(fs_info->mount_opt, opt))			\
1421		btrfs_info(fs_info, fmt, ##args);				\
1422} while (0)
1423
1424#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...)	\
1425do {									\
1426	if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1427	    !btrfs_raw_test_opt(fs_info->mount_opt, opt))		\
1428		btrfs_info(fs_info, fmt, ##args);			\
1429} while (0)
1430
1431static void btrfs_emit_options(struct btrfs_fs_info *info,
1432			       struct btrfs_fs_context *old)
1433{
1434	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1435	btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1436	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1437	btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1438	btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1439	btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1440	btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1441	btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1442	btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1443	btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1444	btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1445	btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1446	btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1447	btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1448	btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1449	btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1450	btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1451	btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1452	btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1453	btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1454	btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1455	btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1456	btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1457
1458	btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1459	btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1460	btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1461	btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1462	btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1463	btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1464	btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1465	btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1466	btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1467
1468	/* Did the compression settings change? */
1469	if (btrfs_test_opt(info, COMPRESS) &&
1470	    (!old ||
1471	     old->compress_type != info->compress_type ||
1472	     old->compress_level != info->compress_level ||
1473	     (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1474	      btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1475		const char *compress_type = btrfs_compress_type2str(info->compress_type);
1476
1477		btrfs_info(info, "%s %s compression, level %d",
1478			   btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1479			   compress_type, info->compress_level);
1480	}
1481
1482	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1483		btrfs_info(info, "max_inline set to %llu", info->max_inline);
1484}
 
 
 
 
 
1485
1486static int btrfs_reconfigure(struct fs_context *fc)
1487{
1488	struct super_block *sb = fc->root->d_sb;
1489	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1490	struct btrfs_fs_context *ctx = fc->fs_private;
1491	struct btrfs_fs_context old_ctx;
1492	int ret = 0;
1493	bool mount_reconfigure = (fc->s_fs_info != NULL);
 
 
 
 
 
 
 
 
 
 
1494
1495	btrfs_info_to_ctx(fs_info, &old_ctx);
 
 
 
 
 
1496
1497	/*
1498	 * This is our "bind mount" trick, we don't want to allow the user to do
1499	 * anything other than mount a different ro/rw and a different subvol,
1500	 * all of the mount options should be maintained.
1501	 */
1502	if (mount_reconfigure)
1503		ctx->mount_opt = old_ctx.mount_opt;
1504
1505	sync_filesystem(sb);
1506	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
 
1507
1508	if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1509		return -EINVAL;
 
 
 
 
1510
1511	ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1512	if (ret < 0)
1513		return ret;
1514
1515	btrfs_ctx_to_info(fs_info, ctx);
1516	btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1517	btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1518				 old_ctx.thread_pool_size);
1519
1520	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1521	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1522	    (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1523		btrfs_warn(fs_info,
1524		"remount supports changing free space tree only from RO to RW");
1525		/* Make sure free space cache options match the state on disk. */
1526		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1527			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1528			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1529		}
1530		if (btrfs_free_space_cache_v1_active(fs_info)) {
1531			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1532			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1533		}
1534	}
1535
1536	ret = 0;
1537	if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1538		ret = btrfs_remount_ro(fs_info);
1539	else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1540		ret = btrfs_remount_rw(fs_info);
1541	if (ret)
1542		goto restore;
1543
1544	/*
1545	 * If we set the mask during the parameter parsing VFS would reject the
1546	 * remount.  Here we can set the mask and the value will be updated
1547	 * appropriately.
1548	 */
1549	if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1550		fc->sb_flags_mask |= SB_POSIXACL;
 
 
 
 
1551
1552	btrfs_emit_options(fs_info, &old_ctx);
 
 
1553	wake_up_process(fs_info->transaction_kthread);
1554	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1555	btrfs_clear_oneshot_options(fs_info);
1556	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1557
1558	return 0;
 
1559restore:
1560	btrfs_ctx_to_info(fs_info, &old_ctx);
1561	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1562	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
 
 
 
 
 
 
 
 
1563	return ret;
1564}
1565
1566/* Used to sort the devices by max_avail(descending sort) */
1567static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
 
1568{
1569	const struct btrfs_device_info *dev_info1 = a;
1570	const struct btrfs_device_info *dev_info2 = b;
1571
1572	if (dev_info1->max_avail > dev_info2->max_avail)
1573		return -1;
1574	else if (dev_info1->max_avail < dev_info2->max_avail)
 
1575		return 1;
 
1576	return 0;
1577}
1578
1579/*
1580 * sort the devices by max_avail, in which max free extent size of each device
1581 * is stored.(Descending Sort)
1582 */
1583static inline void btrfs_descending_sort_devices(
1584					struct btrfs_device_info *devices,
1585					size_t nr_devices)
1586{
1587	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1588	     btrfs_cmp_device_free_bytes, NULL);
1589}
1590
1591/*
1592 * The helper to calc the free space on the devices that can be used to store
1593 * file data.
1594 */
1595static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1596					      u64 *free_bytes)
1597{
1598	struct btrfs_device_info *devices_info;
1599	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1600	struct btrfs_device *device;
 
1601	u64 type;
1602	u64 avail_space;
1603	u64 min_stripe_size;
1604	int num_stripes = 1;
1605	int i = 0, nr_devices;
1606	const struct btrfs_raid_attr *rattr;
1607
1608	/*
1609	 * We aren't under the device list lock, so this is racy-ish, but good
1610	 * enough for our purposes.
1611	 */
1612	nr_devices = fs_info->fs_devices->open_devices;
1613	if (!nr_devices) {
1614		smp_mb();
1615		nr_devices = fs_info->fs_devices->open_devices;
1616		ASSERT(nr_devices);
1617		if (!nr_devices) {
1618			*free_bytes = 0;
1619			return 0;
1620		}
1621	}
1622
1623	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1624			       GFP_KERNEL);
1625	if (!devices_info)
1626		return -ENOMEM;
1627
1628	/* calc min stripe number for data space allocation */
1629	type = btrfs_data_alloc_profile(fs_info);
1630	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1631
1632	if (type & BTRFS_BLOCK_GROUP_RAID0)
1633		num_stripes = nr_devices;
1634	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1635		num_stripes = rattr->ncopies;
1636	else if (type & BTRFS_BLOCK_GROUP_RAID10)
 
 
1637		num_stripes = 4;
 
1638
1639	/* Adjust for more than 1 stripe per device */
1640	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
 
 
1641
1642	rcu_read_lock();
1643	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1644		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1645						&device->dev_state) ||
1646		    !device->bdev ||
1647		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1648			continue;
1649
1650		if (i >= nr_devices)
1651			break;
1652
1653		avail_space = device->total_bytes - device->bytes_used;
1654
1655		/* align with stripe_len */
1656		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
 
1657
1658		/*
1659		 * Ensure we have at least min_stripe_size on top of the
1660		 * reserved space on the device.
 
1661		 */
1662		if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1663			continue;
1664
1665		avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
 
 
 
 
 
 
 
 
 
 
1666
1667		devices_info[i].dev = device;
1668		devices_info[i].max_avail = avail_space;
1669
1670		i++;
1671	}
1672	rcu_read_unlock();
1673
1674	nr_devices = i;
1675
1676	btrfs_descending_sort_devices(devices_info, nr_devices);
1677
1678	i = nr_devices - 1;
1679	avail_space = 0;
1680	while (nr_devices >= rattr->devs_min) {
1681		num_stripes = min(num_stripes, nr_devices);
 
1682
1683		if (devices_info[i].max_avail >= min_stripe_size) {
1684			int j;
1685			u64 alloc_size;
1686
1687			avail_space += devices_info[i].max_avail * num_stripes;
1688			alloc_size = devices_info[i].max_avail;
1689			for (j = i + 1 - num_stripes; j <= i; j++)
1690				devices_info[j].max_avail -= alloc_size;
1691		}
1692		i--;
1693		nr_devices--;
1694	}
1695
1696	kfree(devices_info);
1697	*free_bytes = avail_space;
1698	return 0;
1699}
1700
1701/*
1702 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1703 *
1704 * If there's a redundant raid level at DATA block groups, use the respective
1705 * multiplier to scale the sizes.
1706 *
1707 * Unused device space usage is based on simulating the chunk allocator
1708 * algorithm that respects the device sizes and order of allocations.  This is
1709 * a close approximation of the actual use but there are other factors that may
1710 * change the result (like a new metadata chunk).
1711 *
1712 * If metadata is exhausted, f_bavail will be 0.
1713 */
1714static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1715{
1716	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1717	struct btrfs_super_block *disk_super = fs_info->super_copy;
 
1718	struct btrfs_space_info *found;
1719	u64 total_used = 0;
1720	u64 total_free_data = 0;
1721	u64 total_free_meta = 0;
1722	u32 bits = fs_info->sectorsize_bits;
1723	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1724	unsigned factor = 1;
1725	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1726	int ret;
1727	u64 thresh = 0;
1728	int mixed = 0;
1729
1730	list_for_each_entry(found, &fs_info->space_info, list) {
 
1731		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1732			int i;
1733
1734			total_free_data += found->disk_total - found->disk_used;
1735			total_free_data -=
1736				btrfs_account_ro_block_groups_free_space(found);
1737
1738			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1739				if (!list_empty(&found->block_groups[i]))
1740					factor = btrfs_bg_type_to_factor(
1741						btrfs_raid_array[i].bg_flag);
 
 
 
 
 
1742			}
1743		}
1744
1745		/*
1746		 * Metadata in mixed block group profiles are accounted in data
1747		 */
1748		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1749			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1750				mixed = 1;
1751			else
1752				total_free_meta += found->disk_total -
1753					found->disk_used;
1754		}
1755
1756		total_used += found->disk_used;
1757	}
1758
 
 
1759	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1760	buf->f_blocks >>= bits;
1761	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1762
1763	/* Account global block reserve as used, it's in logical size already */
1764	spin_lock(&block_rsv->lock);
1765	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
1766	if (buf->f_bfree >= block_rsv->size >> bits)
1767		buf->f_bfree -= block_rsv->size >> bits;
1768	else
1769		buf->f_bfree = 0;
1770	spin_unlock(&block_rsv->lock);
1771
1772	buf->f_bavail = div_u64(total_free_data, factor);
1773	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1774	if (ret)
1775		return ret;
1776	buf->f_bavail += div_u64(total_free_data, factor);
1777	buf->f_bavail = buf->f_bavail >> bits;
1778
1779	/*
1780	 * We calculate the remaining metadata space minus global reserve. If
1781	 * this is (supposedly) smaller than zero, there's no space. But this
1782	 * does not hold in practice, the exhausted state happens where's still
1783	 * some positive delta. So we apply some guesswork and compare the
1784	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
1785	 *
1786	 * We probably cannot calculate the exact threshold value because this
1787	 * depends on the internal reservations requested by various
1788	 * operations, so some operations that consume a few metadata will
1789	 * succeed even if the Avail is zero. But this is better than the other
1790	 * way around.
1791	 */
1792	thresh = SZ_4M;
1793
1794	/*
1795	 * We only want to claim there's no available space if we can no longer
1796	 * allocate chunks for our metadata profile and our global reserve will
1797	 * not fit in the free metadata space.  If we aren't ->full then we
1798	 * still can allocate chunks and thus are fine using the currently
1799	 * calculated f_bavail.
1800	 */
1801	if (!mixed && block_rsv->space_info->full &&
1802	    (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1803		buf->f_bavail = 0;
1804
1805	buf->f_type = BTRFS_SUPER_MAGIC;
1806	buf->f_bsize = fs_info->sectorsize;
1807	buf->f_namelen = BTRFS_NAME_LEN;
1808
1809	/* We treat it as constant endianness (it doesn't matter _which_)
1810	   because we want the fsid to come out the same whether mounted
1811	   on a big-endian or little-endian host */
1812	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1813	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1814	/* Mask in the root object ID too, to disambiguate subvols */
1815	buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1816	buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
1817
1818	return 0;
1819}
1820
1821static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1822{
1823	struct btrfs_fs_info *p = fc->s_fs_info;
1824	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1825
1826	return fs_info->fs_devices == p->fs_devices;
1827}
1828
1829static int btrfs_get_tree_super(struct fs_context *fc)
1830{
1831	struct btrfs_fs_info *fs_info = fc->s_fs_info;
1832	struct btrfs_fs_context *ctx = fc->fs_private;
1833	struct btrfs_fs_devices *fs_devices = NULL;
1834	struct block_device *bdev;
1835	struct btrfs_device *device;
1836	struct super_block *sb;
1837	blk_mode_t mode = btrfs_open_mode(fc);
1838	int ret;
1839
1840	btrfs_ctx_to_info(fs_info, ctx);
1841	mutex_lock(&uuid_mutex);
1842
1843	/*
1844	 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1845	 * either a valid device or an error.
1846	 */
1847	device = btrfs_scan_one_device(fc->source, mode, true);
1848	ASSERT(device != NULL);
1849	if (IS_ERR(device)) {
1850		mutex_unlock(&uuid_mutex);
1851		return PTR_ERR(device);
1852	}
1853
1854	fs_devices = device->fs_devices;
1855	fs_info->fs_devices = fs_devices;
1856
1857	ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1858	mutex_unlock(&uuid_mutex);
1859	if (ret)
1860		return ret;
1861
1862	if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1863		ret = -EACCES;
1864		goto error;
1865	}
1866
1867	bdev = fs_devices->latest_dev->bdev;
1868
1869	/*
1870	 * From now on the error handling is not straightforward.
1871	 *
1872	 * If successful, this will transfer the fs_info into the super block,
1873	 * and fc->s_fs_info will be NULL.  However if there's an existing
1874	 * super, we'll still have fc->s_fs_info populated.  If we error
1875	 * completely out it'll be cleaned up when we drop the fs_context,
1876	 * otherwise it's tied to the lifetime of the super_block.
1877	 */
1878	sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1879	if (IS_ERR(sb)) {
1880		ret = PTR_ERR(sb);
1881		goto error;
1882	}
1883
1884	set_device_specific_options(fs_info);
1885
1886	if (sb->s_root) {
1887		btrfs_close_devices(fs_devices);
1888		/*
1889		 * At this stage we may have RO flag mismatch between
1890		 * fc->sb_flags and sb->s_flags.  Caller should detect such
1891		 * mismatch and reconfigure with sb->s_umount rwsem held if
1892		 * needed.
1893		 */
1894	} else {
1895		snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1896		shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1897		btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1898		ret = btrfs_fill_super(sb, fs_devices);
1899		if (ret) {
1900			deactivate_locked_super(sb);
1901			return ret;
1902		}
1903	}
1904
1905	btrfs_clear_oneshot_options(fs_info);
1906
1907	fc->root = dget(sb->s_root);
1908	return 0;
1909
1910error:
1911	btrfs_close_devices(fs_devices);
1912	return ret;
1913}
1914
1915/*
1916 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1917 * with different ro/rw options") the following works:
1918 *
1919 *        (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1920 *       (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1921 *
1922 * which looks nice and innocent but is actually pretty intricate and deserves
1923 * a long comment.
1924 *
1925 * On another filesystem a subvolume mount is close to something like:
1926 *
1927 *	(iii) # create rw superblock + initial mount
1928 *	      mount -t xfs /dev/sdb /opt/
1929 *
1930 *	      # create ro bind mount
1931 *	      mount --bind -o ro /opt/foo /mnt/foo
1932 *
1933 *	      # unmount initial mount
1934 *	      umount /opt
1935 *
1936 * Of course, there's some special subvolume sauce and there's the fact that the
1937 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1938 * it's very close and will help us understand the issue.
1939 *
1940 * The old mount API didn't cleanly distinguish between a mount being made ro
1941 * and a superblock being made ro.  The only way to change the ro state of
1942 * either object was by passing ms_rdonly. If a new mount was created via
1943 * mount(2) such as:
1944 *
1945 *      mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1946 *
1947 * the MS_RDONLY flag being specified had two effects:
1948 *
1949 * (1) MNT_READONLY was raised -> the resulting mount got
1950 *     @mnt->mnt_flags |= MNT_READONLY raised.
1951 *
1952 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1953 *     made the superblock ro. Note, how SB_RDONLY has the same value as
1954 *     ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1955 *
1956 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1957 * subtree mounted ro.
1958 *
1959 * But consider the effect on the old mount API on btrfs subvolume mounting
1960 * which combines the distinct step in (iii) into a single step.
1961 *
1962 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1963 * is issued the superblock is ro and thus even if the mount created for (ii) is
1964 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1965 * to rw for (ii) which it did using an internal remount call.
1966 *
1967 * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1968 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1969 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1970 * passed by mount(8) to mount(2).
1971 *
1972 * Enter the new mount API. The new mount API disambiguates making a mount ro
1973 * and making a superblock ro.
1974 *
1975 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1976 *     fsmount() or mount_setattr() this is a pure VFS level change for a
1977 *     specific mount or mount tree that is never seen by the filesystem itself.
1978 *
1979 * (4) To turn a superblock ro the "ro" flag must be used with
1980 *     fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1981 *     in fc->sb_flags.
1982 *
1983 * But, currently the util-linux mount command already utilizes the new mount
1984 * API and is still setting fsconfig(FSCONFIG_SET_FLAG, "ro") no matter if it's
1985 * btrfs or not, setting the whole super block RO.  To make per-subvolume mounting
1986 * work with different options work we need to keep backward compatibility.
1987 */
1988static int btrfs_reconfigure_for_mount(struct fs_context *fc, struct vfsmount *mnt)
1989{
1990	int ret = 0;
1991
1992	if (fc->sb_flags & SB_RDONLY)
1993		return ret;
1994
1995	down_write(&mnt->mnt_sb->s_umount);
1996	if (!(fc->sb_flags & SB_RDONLY) && (mnt->mnt_sb->s_flags & SB_RDONLY))
1997		ret = btrfs_reconfigure(fc);
1998	up_write(&mnt->mnt_sb->s_umount);
1999	return ret;
2000}
2001
2002static int btrfs_get_tree_subvol(struct fs_context *fc)
2003{
2004	struct btrfs_fs_info *fs_info = NULL;
2005	struct btrfs_fs_context *ctx = fc->fs_private;
2006	struct fs_context *dup_fc;
2007	struct dentry *dentry;
2008	struct vfsmount *mnt;
2009	int ret = 0;
2010
2011	/*
2012	 * Setup a dummy root and fs_info for test/set super.  This is because
2013	 * we don't actually fill this stuff out until open_ctree, but we need
2014	 * then open_ctree will properly initialize the file system specific
2015	 * settings later.  btrfs_init_fs_info initializes the static elements
2016	 * of the fs_info (locks and such) to make cleanup easier if we find a
2017	 * superblock with our given fs_devices later on at sget() time.
2018	 */
2019	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2020	if (!fs_info)
2021		return -ENOMEM;
2022
2023	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2024	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2025	if (!fs_info->super_copy || !fs_info->super_for_commit) {
2026		btrfs_free_fs_info(fs_info);
2027		return -ENOMEM;
2028	}
2029	btrfs_init_fs_info(fs_info);
2030
2031	dup_fc = vfs_dup_fs_context(fc);
2032	if (IS_ERR(dup_fc)) {
2033		btrfs_free_fs_info(fs_info);
2034		return PTR_ERR(dup_fc);
2035	}
2036
2037	/*
2038	 * When we do the sget_fc this gets transferred to the sb, so we only
2039	 * need to set it on the dup_fc as this is what creates the super block.
2040	 */
2041	dup_fc->s_fs_info = fs_info;
2042
2043	/*
2044	 * We'll do the security settings in our btrfs_get_tree_super() mount
2045	 * loop, they were duplicated into dup_fc, we can drop the originals
2046	 * here.
2047	 */
2048	security_free_mnt_opts(&fc->security);
2049	fc->security = NULL;
2050
2051	mnt = fc_mount(dup_fc);
2052	if (IS_ERR(mnt)) {
2053		put_fs_context(dup_fc);
2054		return PTR_ERR(mnt);
2055	}
2056	ret = btrfs_reconfigure_for_mount(dup_fc, mnt);
2057	put_fs_context(dup_fc);
2058	if (ret) {
2059		mntput(mnt);
2060		return ret;
2061	}
2062
2063	/*
2064	 * This free's ->subvol_name, because if it isn't set we have to
2065	 * allocate a buffer to hold the subvol_name, so we just drop our
2066	 * reference to it here.
2067	 */
2068	dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2069	ctx->subvol_name = NULL;
2070	if (IS_ERR(dentry))
2071		return PTR_ERR(dentry);
2072
2073	fc->root = dentry;
2074	return 0;
2075}
2076
2077static int btrfs_get_tree(struct fs_context *fc)
2078{
2079	/*
2080	 * Since we use mount_subtree to mount the default/specified subvol, we
2081	 * have to do mounts in two steps.
2082	 *
2083	 * First pass through we call btrfs_get_tree_subvol(), this is just a
2084	 * wrapper around fc_mount() to call back into here again, and this time
2085	 * we'll call btrfs_get_tree_super().  This will do the open_ctree() and
2086	 * everything to open the devices and file system.  Then we return back
2087	 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2088	 * from there we can do our mount_subvol() call, which will lookup
2089	 * whichever subvol we're mounting and setup this fc with the
2090	 * appropriate dentry for the subvol.
2091	 */
2092	if (fc->s_fs_info)
2093		return btrfs_get_tree_super(fc);
2094	return btrfs_get_tree_subvol(fc);
2095}
2096
2097static void btrfs_kill_super(struct super_block *sb)
2098{
2099	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2100	kill_anon_super(sb);
2101	btrfs_free_fs_info(fs_info);
2102}
2103
2104static void btrfs_free_fs_context(struct fs_context *fc)
2105{
2106	struct btrfs_fs_context *ctx = fc->fs_private;
2107	struct btrfs_fs_info *fs_info = fc->s_fs_info;
2108
2109	if (fs_info)
2110		btrfs_free_fs_info(fs_info);
2111
2112	if (ctx && refcount_dec_and_test(&ctx->refs)) {
2113		kfree(ctx->subvol_name);
2114		kfree(ctx);
2115	}
2116}
2117
2118static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2119{
2120	struct btrfs_fs_context *ctx = src_fc->fs_private;
2121
2122	/*
2123	 * Give a ref to our ctx to this dup, as we want to keep it around for
2124	 * our original fc so we can have the subvolume name or objectid.
2125	 *
2126	 * We unset ->source in the original fc because the dup needs it for
2127	 * mounting, and then once we free the dup it'll free ->source, so we
2128	 * need to make sure we're only pointing to it in one fc.
2129	 */
2130	refcount_inc(&ctx->refs);
2131	fc->fs_private = ctx;
2132	fc->source = src_fc->source;
2133	src_fc->source = NULL;
2134	return 0;
2135}
2136
2137static const struct fs_context_operations btrfs_fs_context_ops = {
2138	.parse_param	= btrfs_parse_param,
2139	.reconfigure	= btrfs_reconfigure,
2140	.get_tree	= btrfs_get_tree,
2141	.dup		= btrfs_dup_fs_context,
2142	.free		= btrfs_free_fs_context,
2143};
2144
2145static int btrfs_init_fs_context(struct fs_context *fc)
2146{
2147	struct btrfs_fs_context *ctx;
2148
2149	ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2150	if (!ctx)
2151		return -ENOMEM;
2152
2153	refcount_set(&ctx->refs, 1);
2154	fc->fs_private = ctx;
2155	fc->ops = &btrfs_fs_context_ops;
2156
2157	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2158		btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2159	} else {
2160		ctx->thread_pool_size =
2161			min_t(unsigned long, num_online_cpus() + 2, 8);
2162		ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2163		ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2164	}
2165
2166#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2167	fc->sb_flags |= SB_POSIXACL;
2168#endif
2169	fc->sb_flags |= SB_I_VERSION;
2170
2171	return 0;
2172}
2173
2174static struct file_system_type btrfs_fs_type = {
2175	.owner			= THIS_MODULE,
2176	.name			= "btrfs",
2177	.init_fs_context	= btrfs_init_fs_context,
2178	.parameters		= btrfs_fs_parameters,
2179	.kill_sb		= btrfs_kill_super,
2180	.fs_flags		= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA |
2181				  FS_ALLOW_IDMAP | FS_MGTIME,
2182 };
2183
2184MODULE_ALIAS_FS("btrfs");
2185
2186static int btrfs_control_open(struct inode *inode, struct file *file)
2187{
2188	/*
2189	 * The control file's private_data is used to hold the
2190	 * transaction when it is started and is used to keep
2191	 * track of whether a transaction is already in progress.
2192	 */
2193	file->private_data = NULL;
2194	return 0;
2195}
2196
2197/*
2198 * Used by /dev/btrfs-control for devices ioctls.
2199 */
2200static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2201				unsigned long arg)
2202{
2203	struct btrfs_ioctl_vol_args *vol;
2204	struct btrfs_device *device = NULL;
2205	dev_t devt = 0;
2206	int ret = -ENOTTY;
2207
2208	if (!capable(CAP_SYS_ADMIN))
2209		return -EPERM;
2210
2211	vol = memdup_user((void __user *)arg, sizeof(*vol));
2212	if (IS_ERR(vol))
2213		return PTR_ERR(vol);
2214	ret = btrfs_check_ioctl_vol_args_path(vol);
2215	if (ret < 0)
2216		goto out;
2217
2218	switch (cmd) {
2219	case BTRFS_IOC_SCAN_DEV:
2220		mutex_lock(&uuid_mutex);
2221		/*
2222		 * Scanning outside of mount can return NULL which would turn
2223		 * into 0 error code.
2224		 */
2225		device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2226		ret = PTR_ERR_OR_ZERO(device);
2227		mutex_unlock(&uuid_mutex);
2228		break;
2229	case BTRFS_IOC_FORGET_DEV:
2230		if (vol->name[0] != 0) {
2231			ret = lookup_bdev(vol->name, &devt);
2232			if (ret)
2233				break;
2234		}
2235		ret = btrfs_forget_devices(devt);
2236		break;
2237	case BTRFS_IOC_DEVICES_READY:
2238		mutex_lock(&uuid_mutex);
2239		/*
2240		 * Scanning outside of mount can return NULL which would turn
2241		 * into 0 error code.
2242		 */
2243		device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2244		if (IS_ERR_OR_NULL(device)) {
2245			mutex_unlock(&uuid_mutex);
2246			if (IS_ERR(device))
2247				ret = PTR_ERR(device);
2248			else
2249				ret = 0;
2250			break;
2251		}
2252		ret = !(device->fs_devices->num_devices ==
2253			device->fs_devices->total_devices);
2254		mutex_unlock(&uuid_mutex);
2255		break;
2256	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2257		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2258		break;
2259	}
2260
2261out:
2262	kfree(vol);
2263	return ret;
2264}
2265
2266static int btrfs_freeze(struct super_block *sb)
2267{
 
2268	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 
2269
2270	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2271	/*
2272	 * We don't need a barrier here, we'll wait for any transaction that
2273	 * could be in progress on other threads (and do delayed iputs that
2274	 * we want to avoid on a frozen filesystem), or do the commit
2275	 * ourselves.
2276	 */
2277	return btrfs_commit_current_transaction(fs_info->tree_root);
2278}
2279
2280static int check_dev_super(struct btrfs_device *dev)
2281{
2282	struct btrfs_fs_info *fs_info = dev->fs_info;
2283	struct btrfs_super_block *sb;
2284	u64 last_trans;
2285	u16 csum_type;
2286	int ret = 0;
2287
2288	/* This should be called with fs still frozen. */
2289	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2290
2291	/* Missing dev, no need to check. */
2292	if (!dev->bdev)
2293		return 0;
2294
2295	/* Only need to check the primary super block. */
2296	sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2297	if (IS_ERR(sb))
2298		return PTR_ERR(sb);
2299
2300	/* Verify the checksum. */
2301	csum_type = btrfs_super_csum_type(sb);
2302	if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2303		btrfs_err(fs_info, "csum type changed, has %u expect %u",
2304			  csum_type, btrfs_super_csum_type(fs_info->super_copy));
2305		ret = -EUCLEAN;
2306		goto out;
2307	}
2308
2309	if (btrfs_check_super_csum(fs_info, sb)) {
2310		btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2311		ret = -EUCLEAN;
2312		goto out;
2313	}
2314
2315	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2316	ret = btrfs_validate_super(fs_info, sb, 0);
2317	if (ret < 0)
2318		goto out;
2319
2320	last_trans = btrfs_get_last_trans_committed(fs_info);
2321	if (btrfs_super_generation(sb) != last_trans) {
2322		btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2323			  btrfs_super_generation(sb), last_trans);
2324		ret = -EUCLEAN;
2325		goto out;
2326	}
2327out:
2328	btrfs_release_disk_super(sb);
2329	return ret;
2330}
2331
2332static int btrfs_unfreeze(struct super_block *sb)
2333{
2334	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2335	struct btrfs_device *device;
2336	int ret = 0;
2337
2338	/*
2339	 * Make sure the fs is not changed by accident (like hibernation then
2340	 * modified by other OS).
2341	 * If we found anything wrong, we mark the fs error immediately.
2342	 *
2343	 * And since the fs is frozen, no one can modify the fs yet, thus
2344	 * we don't need to hold device_list_mutex.
2345	 */
2346	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2347		ret = check_dev_super(device);
2348		if (ret < 0) {
2349			btrfs_handle_fs_error(fs_info, ret,
2350				"super block on devid %llu got modified unexpectedly",
2351				device->devid);
2352			break;
2353		}
2354	}
2355	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2356
2357	/*
2358	 * We still return 0, to allow VFS layer to unfreeze the fs even the
2359	 * above checks failed. Since the fs is either fine or read-only, we're
2360	 * safe to continue, without causing further damage.
2361	 */
2362	return 0;
2363}
2364
2365static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2366{
2367	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
 
 
 
 
2368
2369	/*
2370	 * There should be always a valid pointer in latest_dev, it may be stale
2371	 * for a short moment in case it's being deleted but still valid until
2372	 * the end of RCU grace period.
 
 
2373	 */
2374	rcu_read_lock();
2375	seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2376	rcu_read_unlock();
2377
2378	return 0;
2379}
2380
2381static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2382{
2383	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2384	const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
2385
2386	trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2387
2388	return nr;
2389}
2390
2391static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)
2392{
2393	const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
2394	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2395
2396	btrfs_free_extent_maps(fs_info, nr_to_scan);
2397
2398	/* The extent map shrinker runs asynchronously, so always return 0. */
2399	return 0;
2400}
2401
2402static const struct super_operations btrfs_super_ops = {
2403	.drop_inode	= btrfs_drop_inode,
2404	.evict_inode	= btrfs_evict_inode,
2405	.put_super	= btrfs_put_super,
2406	.sync_fs	= btrfs_sync_fs,
2407	.show_options	= btrfs_show_options,
2408	.show_devname	= btrfs_show_devname,
 
2409	.alloc_inode	= btrfs_alloc_inode,
2410	.destroy_inode	= btrfs_destroy_inode,
2411	.free_inode	= btrfs_free_inode,
2412	.statfs		= btrfs_statfs,
 
2413	.freeze_fs	= btrfs_freeze,
2414	.unfreeze_fs	= btrfs_unfreeze,
2415	.nr_cached_objects = btrfs_nr_cached_objects,
2416	.free_cached_objects = btrfs_free_cached_objects,
2417};
2418
2419static const struct file_operations btrfs_ctl_fops = {
2420	.open = btrfs_control_open,
2421	.unlocked_ioctl	 = btrfs_control_ioctl,
2422	.compat_ioctl = compat_ptr_ioctl,
2423	.owner	 = THIS_MODULE,
2424	.llseek = noop_llseek,
2425};
2426
2427static struct miscdevice btrfs_misc = {
2428	.minor		= BTRFS_MINOR,
2429	.name		= "btrfs-control",
2430	.fops		= &btrfs_ctl_fops
2431};
2432
2433MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2434MODULE_ALIAS("devname:btrfs-control");
2435
2436static int __init btrfs_interface_init(void)
2437{
2438	return misc_register(&btrfs_misc);
2439}
2440
2441static __cold void btrfs_interface_exit(void)
2442{
2443	misc_deregister(&btrfs_misc);
2444}
2445
2446static int __init btrfs_print_mod_info(void)
2447{
2448	static const char options[] = ""
2449#ifdef CONFIG_BTRFS_DEBUG
2450			", debug=on"
2451#endif
2452#ifdef CONFIG_BTRFS_ASSERT
2453			", assert=on"
2454#endif
 
 
 
2455#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2456			", ref-verify=on"
2457#endif
2458#ifdef CONFIG_BLK_DEV_ZONED
2459			", zoned=yes"
2460#else
2461			", zoned=no"
2462#endif
2463#ifdef CONFIG_FS_VERITY
2464			", fsverity=yes"
2465#else
2466			", fsverity=no"
2467#endif
2468			;
2469	pr_info("Btrfs loaded%s\n", options);
2470	return 0;
2471}
2472
2473static int register_btrfs(void)
2474{
2475	return register_filesystem(&btrfs_fs_type);
2476}
2477
2478static void unregister_btrfs(void)
2479{
2480	unregister_filesystem(&btrfs_fs_type);
2481}
2482
2483/* Helper structure for long init/exit functions. */
2484struct init_sequence {
2485	int (*init_func)(void);
2486	/* Can be NULL if the init_func doesn't need cleanup. */
2487	void (*exit_func)(void);
2488};
2489
2490static const struct init_sequence mod_init_seq[] = {
2491	{
2492		.init_func = btrfs_props_init,
2493		.exit_func = NULL,
2494	}, {
2495		.init_func = btrfs_init_sysfs,
2496		.exit_func = btrfs_exit_sysfs,
2497	}, {
2498		.init_func = btrfs_init_compress,
2499		.exit_func = btrfs_exit_compress,
2500	}, {
2501		.init_func = btrfs_init_cachep,
2502		.exit_func = btrfs_destroy_cachep,
2503	}, {
2504		.init_func = btrfs_init_dio,
2505		.exit_func = btrfs_destroy_dio,
2506	}, {
2507		.init_func = btrfs_transaction_init,
2508		.exit_func = btrfs_transaction_exit,
2509	}, {
2510		.init_func = btrfs_ctree_init,
2511		.exit_func = btrfs_ctree_exit,
2512	}, {
2513		.init_func = btrfs_free_space_init,
2514		.exit_func = btrfs_free_space_exit,
2515	}, {
2516		.init_func = extent_state_init_cachep,
2517		.exit_func = extent_state_free_cachep,
2518	}, {
2519		.init_func = extent_buffer_init_cachep,
2520		.exit_func = extent_buffer_free_cachep,
2521	}, {
2522		.init_func = btrfs_bioset_init,
2523		.exit_func = btrfs_bioset_exit,
2524	}, {
2525		.init_func = extent_map_init,
2526		.exit_func = extent_map_exit,
2527	}, {
2528		.init_func = ordered_data_init,
2529		.exit_func = ordered_data_exit,
2530	}, {
2531		.init_func = btrfs_delayed_inode_init,
2532		.exit_func = btrfs_delayed_inode_exit,
2533	}, {
2534		.init_func = btrfs_auto_defrag_init,
2535		.exit_func = btrfs_auto_defrag_exit,
2536	}, {
2537		.init_func = btrfs_delayed_ref_init,
2538		.exit_func = btrfs_delayed_ref_exit,
2539	}, {
2540		.init_func = btrfs_prelim_ref_init,
2541		.exit_func = btrfs_prelim_ref_exit,
2542	}, {
2543		.init_func = btrfs_interface_init,
2544		.exit_func = btrfs_interface_exit,
2545	}, {
2546		.init_func = btrfs_print_mod_info,
2547		.exit_func = NULL,
2548	}, {
2549		.init_func = btrfs_run_sanity_tests,
2550		.exit_func = NULL,
2551	}, {
2552		.init_func = register_btrfs,
2553		.exit_func = unregister_btrfs,
2554	}
2555};
2556
2557static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
 
 
2558
2559static __always_inline void btrfs_exit_btrfs_fs(void)
2560{
2561	int i;
2562
2563	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2564		if (!mod_init_result[i])
2565			continue;
2566		if (mod_init_seq[i].exit_func)
2567			mod_init_seq[i].exit_func();
2568		mod_init_result[i] = false;
2569	}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2570}
2571
2572static void __exit exit_btrfs_fs(void)
2573{
2574	btrfs_exit_btrfs_fs();
 
 
 
 
 
 
 
 
 
 
 
2575	btrfs_cleanup_fs_uuids();
2576}
2577
2578static int __init init_btrfs_fs(void)
2579{
2580	int ret;
2581	int i;
2582
2583	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2584		ASSERT(!mod_init_result[i]);
2585		ret = mod_init_seq[i].init_func();
2586		if (ret < 0) {
2587			btrfs_exit_btrfs_fs();
2588			return ret;
2589		}
2590		mod_init_result[i] = true;
2591	}
2592	return 0;
2593}
2594
2595late_initcall(init_btrfs_fs);
2596module_exit(exit_btrfs_fs)
2597
2598MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2599MODULE_LICENSE("GPL");
2600MODULE_SOFTDEP("pre: crc32c");
2601MODULE_SOFTDEP("pre: xxhash64");
2602MODULE_SOFTDEP("pre: sha256");
2603MODULE_SOFTDEP("pre: blake2b-256");
v4.17
   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/buffer_head.h>
   9#include <linux/fs.h>
  10#include <linux/pagemap.h>
  11#include <linux/highmem.h>
  12#include <linux/time.h>
  13#include <linux/init.h>
  14#include <linux/seq_file.h>
  15#include <linux/string.h>
  16#include <linux/backing-dev.h>
  17#include <linux/mount.h>
  18#include <linux/mpage.h>
  19#include <linux/swap.h>
  20#include <linux/writeback.h>
  21#include <linux/statfs.h>
  22#include <linux/compat.h>
  23#include <linux/parser.h>
  24#include <linux/ctype.h>
  25#include <linux/namei.h>
  26#include <linux/miscdevice.h>
  27#include <linux/magic.h>
  28#include <linux/slab.h>
  29#include <linux/cleancache.h>
  30#include <linux/ratelimit.h>
  31#include <linux/crc32c.h>
  32#include <linux/btrfs.h>
 
 
 
  33#include "delayed-inode.h"
  34#include "ctree.h"
  35#include "disk-io.h"
  36#include "transaction.h"
  37#include "btrfs_inode.h"
  38#include "print-tree.h"
  39#include "props.h"
  40#include "xattr.h"
  41#include "volumes.h"
  42#include "export.h"
  43#include "compression.h"
  44#include "rcu-string.h"
  45#include "dev-replace.h"
  46#include "free-space-cache.h"
  47#include "backref.h"
 
 
 
  48#include "tests/btrfs-tests.h"
  49
 
  50#include "qgroup.h"
 
 
 
 
 
 
 
 
 
 
  51#define CREATE_TRACE_POINTS
  52#include <trace/events/btrfs.h>
  53
  54static const struct super_operations btrfs_super_ops;
  55
  56/*
  57 * Types for mounting the default subvolume and a subvolume explicitly
  58 * requested by subvol=/path. That way the callchain is straightforward and we
  59 * don't have to play tricks with the mount options and recursive calls to
  60 * btrfs_mount.
  61 *
  62 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
  63 */
  64static struct file_system_type btrfs_fs_type;
  65static struct file_system_type btrfs_root_fs_type;
  66
  67static int btrfs_remount(struct super_block *sb, int *flags, char *data);
  68
  69const char *btrfs_decode_error(int errno)
  70{
  71	char *errstr = "unknown";
  72
  73	switch (errno) {
  74	case -EIO:
  75		errstr = "IO failure";
  76		break;
  77	case -ENOMEM:
  78		errstr = "Out of memory";
  79		break;
  80	case -EROFS:
  81		errstr = "Readonly filesystem";
  82		break;
  83	case -EEXIST:
  84		errstr = "Object already exists";
  85		break;
  86	case -ENOSPC:
  87		errstr = "No space left";
  88		break;
  89	case -ENOENT:
  90		errstr = "No such entry";
  91		break;
  92	}
  93
  94	return errstr;
  95}
  96
  97/*
  98 * __btrfs_handle_fs_error decodes expected errors from the caller and
  99 * invokes the approciate error response.
 100 */
 101__cold
 102void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
 103		       unsigned int line, int errno, const char *fmt, ...)
 104{
 105	struct super_block *sb = fs_info->sb;
 106#ifdef CONFIG_PRINTK
 107	const char *errstr;
 108#endif
 109
 110	/*
 111	 * Special case: if the error is EROFS, and we're already
 112	 * under SB_RDONLY, then it is safe here.
 113	 */
 114	if (errno == -EROFS && sb_rdonly(sb))
 115  		return;
 116
 117#ifdef CONFIG_PRINTK
 118	errstr = btrfs_decode_error(errno);
 119	if (fmt) {
 120		struct va_format vaf;
 121		va_list args;
 122
 123		va_start(args, fmt);
 124		vaf.fmt = fmt;
 125		vaf.va = &args;
 126
 127		pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
 128			sb->s_id, function, line, errno, errstr, &vaf);
 129		va_end(args);
 130	} else {
 131		pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
 132			sb->s_id, function, line, errno, errstr);
 133	}
 134#endif
 135
 136	/*
 137	 * Today we only save the error info to memory.  Long term we'll
 138	 * also send it down to the disk
 139	 */
 140	set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
 141
 142	/* Don't go through full error handling during mount */
 143	if (!(sb->s_flags & SB_BORN))
 144		return;
 145
 146	if (sb_rdonly(sb))
 147		return;
 148
 149	/* btrfs handle error by forcing the filesystem readonly */
 150	sb->s_flags |= SB_RDONLY;
 151	btrfs_info(fs_info, "forced readonly");
 152	/*
 153	 * Note that a running device replace operation is not canceled here
 154	 * although there is no way to update the progress. It would add the
 155	 * risk of a deadlock, therefore the canceling is omitted. The only
 156	 * penalty is that some I/O remains active until the procedure
 157	 * completes. The next time when the filesystem is mounted writeable
 158	 * again, the device replace operation continues.
 159	 */
 160}
 161
 162#ifdef CONFIG_PRINTK
 163static const char * const logtypes[] = {
 164	"emergency",
 165	"alert",
 166	"critical",
 167	"error",
 168	"warning",
 169	"notice",
 170	"info",
 171	"debug",
 172};
 173
 174
 175/*
 176 * Use one ratelimit state per log level so that a flood of less important
 177 * messages doesn't cause more important ones to be dropped.
 178 */
 179static struct ratelimit_state printk_limits[] = {
 180	RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
 181	RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
 182	RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
 183	RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
 184	RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
 185	RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
 186	RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
 187	RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
 188};
 189
 190void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
 191{
 192	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
 193	struct va_format vaf;
 194	va_list args;
 195	int kern_level;
 196	const char *type = logtypes[4];
 197	struct ratelimit_state *ratelimit = &printk_limits[4];
 198
 199	va_start(args, fmt);
 200
 201	while ((kern_level = printk_get_level(fmt)) != 0) {
 202		size_t size = printk_skip_level(fmt) - fmt;
 203
 204		if (kern_level >= '0' && kern_level <= '7') {
 205			memcpy(lvl, fmt,  size);
 206			lvl[size] = '\0';
 207			type = logtypes[kern_level - '0'];
 208			ratelimit = &printk_limits[kern_level - '0'];
 209		}
 210		fmt += size;
 211	}
 212
 213	vaf.fmt = fmt;
 214	vaf.va = &args;
 215
 216	if (__ratelimit(ratelimit))
 217		printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
 218			fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
 219
 220	va_end(args);
 221}
 222#endif
 223
 224/*
 225 * We only mark the transaction aborted and then set the file system read-only.
 226 * This will prevent new transactions from starting or trying to join this
 227 * one.
 228 *
 229 * This means that error recovery at the call site is limited to freeing
 230 * any local memory allocations and passing the error code up without
 231 * further cleanup. The transaction should complete as it normally would
 232 * in the call path but will return -EIO.
 233 *
 234 * We'll complete the cleanup in btrfs_end_transaction and
 235 * btrfs_commit_transaction.
 236 */
 237__cold
 238void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 239			       const char *function,
 240			       unsigned int line, int errno)
 241{
 242	struct btrfs_fs_info *fs_info = trans->fs_info;
 243
 244	trans->aborted = errno;
 245	/* Nothing used. The other threads that have joined this
 246	 * transaction may be able to continue. */
 247	if (!trans->dirty && list_empty(&trans->new_bgs)) {
 248		const char *errstr;
 249
 250		errstr = btrfs_decode_error(errno);
 251		btrfs_warn(fs_info,
 252		           "%s:%d: Aborting unused transaction(%s).",
 253		           function, line, errstr);
 254		return;
 255	}
 256	WRITE_ONCE(trans->transaction->aborted, errno);
 257	/* Wake up anybody who may be waiting on this transaction */
 258	wake_up(&fs_info->transaction_wait);
 259	wake_up(&fs_info->transaction_blocked_wait);
 260	__btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
 261}
 262/*
 263 * __btrfs_panic decodes unexpected, fatal errors from the caller,
 264 * issues an alert, and either panics or BUGs, depending on mount options.
 265 */
 266__cold
 267void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 268		   unsigned int line, int errno, const char *fmt, ...)
 269{
 270	char *s_id = "<unknown>";
 271	const char *errstr;
 272	struct va_format vaf = { .fmt = fmt };
 273	va_list args;
 274
 275	if (fs_info)
 276		s_id = fs_info->sb->s_id;
 277
 278	va_start(args, fmt);
 279	vaf.va = &args;
 280
 281	errstr = btrfs_decode_error(errno);
 282	if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
 283		panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
 284			s_id, function, line, &vaf, errno, errstr);
 285
 286	btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
 287		   function, line, &vaf, errno, errstr);
 288	va_end(args);
 289	/* Caller calls BUG() */
 290}
 291
 292static void btrfs_put_super(struct super_block *sb)
 293{
 294	close_ctree(btrfs_sb(sb));
 295}
 296
 297enum {
 298	Opt_acl, Opt_noacl,
 299	Opt_clear_cache,
 300	Opt_commit_interval,
 301	Opt_compress,
 302	Opt_compress_force,
 303	Opt_compress_force_type,
 304	Opt_compress_type,
 305	Opt_degraded,
 306	Opt_device,
 307	Opt_fatal_errors,
 308	Opt_flushoncommit, Opt_noflushoncommit,
 309	Opt_inode_cache, Opt_noinode_cache,
 310	Opt_max_inline,
 311	Opt_barrier, Opt_nobarrier,
 312	Opt_datacow, Opt_nodatacow,
 313	Opt_datasum, Opt_nodatasum,
 314	Opt_defrag, Opt_nodefrag,
 315	Opt_discard, Opt_nodiscard,
 316	Opt_nologreplay,
 317	Opt_norecovery,
 318	Opt_ratio,
 319	Opt_rescan_uuid_tree,
 320	Opt_skip_balance,
 321	Opt_space_cache, Opt_no_space_cache,
 322	Opt_space_cache_version,
 323	Opt_ssd, Opt_nossd,
 324	Opt_ssd_spread, Opt_nossd_spread,
 325	Opt_subvol,
 
 326	Opt_subvolid,
 327	Opt_thread_pool,
 328	Opt_treelog, Opt_notreelog,
 329	Opt_usebackuproot,
 330	Opt_user_subvol_rm_allowed,
 
 331
 332	/* Deprecated options */
 333	Opt_alloc_start,
 334	Opt_recovery,
 335	Opt_subvolrootid,
 336
 337	/* Debugging options */
 338	Opt_check_integrity,
 339	Opt_check_integrity_including_extent_data,
 340	Opt_check_integrity_print_mask,
 341	Opt_enospc_debug, Opt_noenospc_debug,
 342#ifdef CONFIG_BTRFS_DEBUG
 343	Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
 344#endif
 345#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 346	Opt_ref_verify,
 347#endif
 348	Opt_err,
 349};
 350
 351static const match_table_t tokens = {
 352	{Opt_acl, "acl"},
 353	{Opt_noacl, "noacl"},
 354	{Opt_clear_cache, "clear_cache"},
 355	{Opt_commit_interval, "commit=%u"},
 356	{Opt_compress, "compress"},
 357	{Opt_compress_type, "compress=%s"},
 358	{Opt_compress_force, "compress-force"},
 359	{Opt_compress_force_type, "compress-force=%s"},
 360	{Opt_degraded, "degraded"},
 361	{Opt_device, "device=%s"},
 362	{Opt_fatal_errors, "fatal_errors=%s"},
 363	{Opt_flushoncommit, "flushoncommit"},
 364	{Opt_noflushoncommit, "noflushoncommit"},
 365	{Opt_inode_cache, "inode_cache"},
 366	{Opt_noinode_cache, "noinode_cache"},
 367	{Opt_max_inline, "max_inline=%s"},
 368	{Opt_barrier, "barrier"},
 369	{Opt_nobarrier, "nobarrier"},
 370	{Opt_datacow, "datacow"},
 371	{Opt_nodatacow, "nodatacow"},
 372	{Opt_datasum, "datasum"},
 373	{Opt_nodatasum, "nodatasum"},
 374	{Opt_defrag, "autodefrag"},
 375	{Opt_nodefrag, "noautodefrag"},
 376	{Opt_discard, "discard"},
 377	{Opt_nodiscard, "nodiscard"},
 378	{Opt_nologreplay, "nologreplay"},
 379	{Opt_norecovery, "norecovery"},
 380	{Opt_ratio, "metadata_ratio=%u"},
 381	{Opt_rescan_uuid_tree, "rescan_uuid_tree"},
 382	{Opt_skip_balance, "skip_balance"},
 383	{Opt_space_cache, "space_cache"},
 384	{Opt_no_space_cache, "nospace_cache"},
 385	{Opt_space_cache_version, "space_cache=%s"},
 386	{Opt_ssd, "ssd"},
 387	{Opt_nossd, "nossd"},
 388	{Opt_ssd_spread, "ssd_spread"},
 389	{Opt_nossd_spread, "nossd_spread"},
 390	{Opt_subvol, "subvol=%s"},
 391	{Opt_subvolid, "subvolid=%s"},
 392	{Opt_thread_pool, "thread_pool=%u"},
 393	{Opt_treelog, "treelog"},
 394	{Opt_notreelog, "notreelog"},
 395	{Opt_usebackuproot, "usebackuproot"},
 396	{Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 397
 398	/* Deprecated options */
 399	{Opt_alloc_start, "alloc_start=%s"},
 400	{Opt_recovery, "recovery"},
 401	{Opt_subvolrootid, "subvolrootid=%d"},
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 402
 403	/* Debugging options */
 404	{Opt_check_integrity, "check_int"},
 405	{Opt_check_integrity_including_extent_data, "check_int_data"},
 406	{Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
 407	{Opt_enospc_debug, "enospc_debug"},
 408	{Opt_noenospc_debug, "noenospc_debug"},
 409#ifdef CONFIG_BTRFS_DEBUG
 410	{Opt_fragment_data, "fragment=data"},
 411	{Opt_fragment_metadata, "fragment=metadata"},
 412	{Opt_fragment_all, "fragment=all"},
 413#endif
 414#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 415	{Opt_ref_verify, "ref_verify"},
 416#endif
 417	{Opt_err, NULL},
 418};
 419
 420/*
 421 * Regular mount options parser.  Everything that is needed only when
 422 * reading in a new superblock is parsed here.
 423 * XXX JDM: This needs to be cleaned up for remount.
 424 */
 425int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
 426			unsigned long new_flags)
 427{
 428	substring_t args[MAX_OPT_ARGS];
 429	char *p, *num;
 430	u64 cache_gen;
 431	int intarg;
 432	int ret = 0;
 433	char *compress_type;
 434	bool compress_force = false;
 435	enum btrfs_compression_type saved_compress_type;
 436	bool saved_compress_force;
 437	int no_compress = 0;
 438
 439	cache_gen = btrfs_super_cache_generation(info->super_copy);
 440	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
 441		btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
 442	else if (cache_gen)
 443		btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 444
 445	/*
 446	 * Even the options are empty, we still need to do extra check
 447	 * against new flags
 448	 */
 449	if (!options)
 450		goto check;
 451
 452	while ((p = strsep(&options, ",")) != NULL) {
 453		int token;
 454		if (!*p)
 455			continue;
 456
 457		token = match_token(p, tokens, args);
 458		switch (token) {
 459		case Opt_degraded:
 460			btrfs_info(info, "allowing degraded mounts");
 461			btrfs_set_opt(info->mount_opt, DEGRADED);
 462			break;
 463		case Opt_subvol:
 464		case Opt_subvolid:
 465		case Opt_subvolrootid:
 466		case Opt_device:
 467			/*
 468			 * These are parsed by btrfs_parse_subvol_options
 469			 * and btrfs_parse_early_options
 470			 * and can be happily ignored here.
 471			 */
 472			break;
 473		case Opt_nodatasum:
 474			btrfs_set_and_info(info, NODATASUM,
 475					   "setting nodatasum");
 476			break;
 477		case Opt_datasum:
 478			if (btrfs_test_opt(info, NODATASUM)) {
 479				if (btrfs_test_opt(info, NODATACOW))
 480					btrfs_info(info,
 481						   "setting datasum, datacow enabled");
 482				else
 483					btrfs_info(info, "setting datasum");
 484			}
 485			btrfs_clear_opt(info->mount_opt, NODATACOW);
 486			btrfs_clear_opt(info->mount_opt, NODATASUM);
 487			break;
 488		case Opt_nodatacow:
 489			if (!btrfs_test_opt(info, NODATACOW)) {
 490				if (!btrfs_test_opt(info, COMPRESS) ||
 491				    !btrfs_test_opt(info, FORCE_COMPRESS)) {
 492					btrfs_info(info,
 493						   "setting nodatacow, compression disabled");
 494				} else {
 495					btrfs_info(info, "setting nodatacow");
 496				}
 497			}
 498			btrfs_clear_opt(info->mount_opt, COMPRESS);
 499			btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 500			btrfs_set_opt(info->mount_opt, NODATACOW);
 501			btrfs_set_opt(info->mount_opt, NODATASUM);
 502			break;
 503		case Opt_datacow:
 504			btrfs_clear_and_info(info, NODATACOW,
 505					     "setting datacow");
 506			break;
 507		case Opt_compress_force:
 508		case Opt_compress_force_type:
 509			compress_force = true;
 510			/* Fallthrough */
 511		case Opt_compress:
 512		case Opt_compress_type:
 513			saved_compress_type = btrfs_test_opt(info,
 514							     COMPRESS) ?
 515				info->compress_type : BTRFS_COMPRESS_NONE;
 516			saved_compress_force =
 517				btrfs_test_opt(info, FORCE_COMPRESS);
 518			if (token == Opt_compress ||
 519			    token == Opt_compress_force ||
 520			    strncmp(args[0].from, "zlib", 4) == 0) {
 521				compress_type = "zlib";
 522
 523				info->compress_type = BTRFS_COMPRESS_ZLIB;
 524				info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
 525				/*
 526				 * args[0] contains uninitialized data since
 527				 * for these tokens we don't expect any
 528				 * parameter.
 529				 */
 530				if (token != Opt_compress &&
 531				    token != Opt_compress_force)
 532					info->compress_level =
 533					  btrfs_compress_str2level(args[0].from);
 534				btrfs_set_opt(info->mount_opt, COMPRESS);
 535				btrfs_clear_opt(info->mount_opt, NODATACOW);
 536				btrfs_clear_opt(info->mount_opt, NODATASUM);
 537				no_compress = 0;
 538			} else if (strncmp(args[0].from, "lzo", 3) == 0) {
 539				compress_type = "lzo";
 540				info->compress_type = BTRFS_COMPRESS_LZO;
 541				btrfs_set_opt(info->mount_opt, COMPRESS);
 542				btrfs_clear_opt(info->mount_opt, NODATACOW);
 543				btrfs_clear_opt(info->mount_opt, NODATASUM);
 544				btrfs_set_fs_incompat(info, COMPRESS_LZO);
 545				no_compress = 0;
 546			} else if (strcmp(args[0].from, "zstd") == 0) {
 547				compress_type = "zstd";
 548				info->compress_type = BTRFS_COMPRESS_ZSTD;
 549				btrfs_set_opt(info->mount_opt, COMPRESS);
 550				btrfs_clear_opt(info->mount_opt, NODATACOW);
 551				btrfs_clear_opt(info->mount_opt, NODATASUM);
 552				btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
 553				no_compress = 0;
 554			} else if (strncmp(args[0].from, "no", 2) == 0) {
 555				compress_type = "no";
 556				btrfs_clear_opt(info->mount_opt, COMPRESS);
 557				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 558				compress_force = false;
 559				no_compress++;
 560			} else {
 561				ret = -EINVAL;
 562				goto out;
 563			}
 564
 565			if (compress_force) {
 566				btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 567			} else {
 568				/*
 569				 * If we remount from compress-force=xxx to
 570				 * compress=xxx, we need clear FORCE_COMPRESS
 571				 * flag, otherwise, there is no way for users
 572				 * to disable forcible compression separately.
 573				 */
 574				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 575			}
 576			if ((btrfs_test_opt(info, COMPRESS) &&
 577			     (info->compress_type != saved_compress_type ||
 578			      compress_force != saved_compress_force)) ||
 579			    (!btrfs_test_opt(info, COMPRESS) &&
 580			     no_compress == 1)) {
 581				btrfs_info(info, "%s %s compression, level %d",
 582					   (compress_force) ? "force" : "use",
 583					   compress_type, info->compress_level);
 584			}
 585			compress_force = false;
 586			break;
 587		case Opt_ssd:
 588			btrfs_set_and_info(info, SSD,
 589					   "enabling ssd optimizations");
 590			btrfs_clear_opt(info->mount_opt, NOSSD);
 591			break;
 592		case Opt_ssd_spread:
 593			btrfs_set_and_info(info, SSD,
 594					   "enabling ssd optimizations");
 595			btrfs_set_and_info(info, SSD_SPREAD,
 596					   "using spread ssd allocation scheme");
 597			btrfs_clear_opt(info->mount_opt, NOSSD);
 598			break;
 599		case Opt_nossd:
 600			btrfs_set_opt(info->mount_opt, NOSSD);
 601			btrfs_clear_and_info(info, SSD,
 602					     "not using ssd optimizations");
 603			/* Fallthrough */
 604		case Opt_nossd_spread:
 605			btrfs_clear_and_info(info, SSD_SPREAD,
 606					     "not using spread ssd allocation scheme");
 607			break;
 608		case Opt_barrier:
 609			btrfs_clear_and_info(info, NOBARRIER,
 610					     "turning on barriers");
 611			break;
 612		case Opt_nobarrier:
 613			btrfs_set_and_info(info, NOBARRIER,
 614					   "turning off barriers");
 615			break;
 616		case Opt_thread_pool:
 617			ret = match_int(&args[0], &intarg);
 618			if (ret) {
 619				goto out;
 620			} else if (intarg == 0) {
 621				ret = -EINVAL;
 622				goto out;
 623			}
 624			info->thread_pool_size = intarg;
 625			break;
 626		case Opt_max_inline:
 627			num = match_strdup(&args[0]);
 628			if (num) {
 629				info->max_inline = memparse(num, NULL);
 630				kfree(num);
 631
 632				if (info->max_inline) {
 633					info->max_inline = min_t(u64,
 634						info->max_inline,
 635						info->sectorsize);
 636				}
 637				btrfs_info(info, "max_inline at %llu",
 638					   info->max_inline);
 639			} else {
 640				ret = -ENOMEM;
 641				goto out;
 642			}
 643			break;
 644		case Opt_alloc_start:
 645			btrfs_info(info,
 646				"option alloc_start is obsolete, ignored");
 647			break;
 648		case Opt_acl:
 649#ifdef CONFIG_BTRFS_FS_POSIX_ACL
 650			info->sb->s_flags |= SB_POSIXACL;
 651			break;
 652#else
 653			btrfs_err(info, "support for ACL not compiled in!");
 654			ret = -EINVAL;
 655			goto out;
 656#endif
 657		case Opt_noacl:
 658			info->sb->s_flags &= ~SB_POSIXACL;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 659			break;
 660		case Opt_notreelog:
 661			btrfs_set_and_info(info, NOTREELOG,
 662					   "disabling tree log");
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 663			break;
 664		case Opt_treelog:
 665			btrfs_clear_and_info(info, NOTREELOG,
 666					     "enabling tree log");
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 667			break;
 668		case Opt_norecovery:
 669		case Opt_nologreplay:
 670			btrfs_set_and_info(info, NOLOGREPLAY,
 671					   "disabling log replay at mount time");
 672			break;
 673		case Opt_flushoncommit:
 674			btrfs_set_and_info(info, FLUSHONCOMMIT,
 675					   "turning on flush-on-commit");
 676			break;
 677		case Opt_noflushoncommit:
 678			btrfs_clear_and_info(info, FLUSHONCOMMIT,
 679					     "turning off flush-on-commit");
 680			break;
 681		case Opt_ratio:
 682			ret = match_int(&args[0], &intarg);
 683			if (ret)
 684				goto out;
 685			info->metadata_ratio = intarg;
 686			btrfs_info(info, "metadata ratio %u",
 687				   info->metadata_ratio);
 688			break;
 689		case Opt_discard:
 690			btrfs_set_and_info(info, DISCARD,
 691					   "turning on discard");
 692			break;
 693		case Opt_nodiscard:
 694			btrfs_clear_and_info(info, DISCARD,
 695					     "turning off discard");
 696			break;
 697		case Opt_space_cache:
 698		case Opt_space_cache_version:
 699			if (token == Opt_space_cache ||
 700			    strcmp(args[0].from, "v1") == 0) {
 701				btrfs_clear_opt(info->mount_opt,
 702						FREE_SPACE_TREE);
 703				btrfs_set_and_info(info, SPACE_CACHE,
 704					   "enabling disk space caching");
 705			} else if (strcmp(args[0].from, "v2") == 0) {
 706				btrfs_clear_opt(info->mount_opt,
 707						SPACE_CACHE);
 708				btrfs_set_and_info(info, FREE_SPACE_TREE,
 709						   "enabling free space tree");
 710			} else {
 711				ret = -EINVAL;
 712				goto out;
 713			}
 714			break;
 715		case Opt_rescan_uuid_tree:
 716			btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
 717			break;
 718		case Opt_no_space_cache:
 719			if (btrfs_test_opt(info, SPACE_CACHE)) {
 720				btrfs_clear_and_info(info, SPACE_CACHE,
 721					     "disabling disk space caching");
 722			}
 723			if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
 724				btrfs_clear_and_info(info, FREE_SPACE_TREE,
 725					     "disabling free space tree");
 726			}
 727			break;
 728		case Opt_inode_cache:
 729			btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
 730					   "enabling inode map caching");
 731			break;
 732		case Opt_noinode_cache:
 733			btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
 734					     "disabling inode map caching");
 735			break;
 736		case Opt_clear_cache:
 737			btrfs_set_and_info(info, CLEAR_CACHE,
 738					   "force clearing of disk cache");
 739			break;
 740		case Opt_user_subvol_rm_allowed:
 741			btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 742			break;
 743		case Opt_enospc_debug:
 744			btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 745			break;
 746		case Opt_noenospc_debug:
 747			btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
 748			break;
 749		case Opt_defrag:
 750			btrfs_set_and_info(info, AUTO_DEFRAG,
 751					   "enabling auto defrag");
 752			break;
 753		case Opt_nodefrag:
 754			btrfs_clear_and_info(info, AUTO_DEFRAG,
 755					     "disabling auto defrag");
 756			break;
 757		case Opt_recovery:
 758			btrfs_warn(info,
 759				   "'recovery' is deprecated, use 'usebackuproot' instead");
 760		case Opt_usebackuproot:
 761			btrfs_info(info,
 762				   "trying to use backup root at mount time");
 763			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
 764			break;
 765		case Opt_skip_balance:
 766			btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 767			break;
 768#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 769		case Opt_check_integrity_including_extent_data:
 770			btrfs_info(info,
 771				   "enabling check integrity including extent data");
 772			btrfs_set_opt(info->mount_opt,
 773				      CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 774			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 775			break;
 776		case Opt_check_integrity:
 777			btrfs_info(info, "enabling check integrity");
 778			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 779			break;
 780		case Opt_check_integrity_print_mask:
 781			ret = match_int(&args[0], &intarg);
 782			if (ret)
 783				goto out;
 784			info->check_integrity_print_mask = intarg;
 785			btrfs_info(info, "check_integrity_print_mask 0x%x",
 786				   info->check_integrity_print_mask);
 787			break;
 788#else
 789		case Opt_check_integrity_including_extent_data:
 790		case Opt_check_integrity:
 791		case Opt_check_integrity_print_mask:
 792			btrfs_err(info,
 793				  "support for check_integrity* not compiled in!");
 794			ret = -EINVAL;
 795			goto out;
 796#endif
 797		case Opt_fatal_errors:
 798			if (strcmp(args[0].from, "panic") == 0)
 799				btrfs_set_opt(info->mount_opt,
 800					      PANIC_ON_FATAL_ERROR);
 801			else if (strcmp(args[0].from, "bug") == 0)
 802				btrfs_clear_opt(info->mount_opt,
 803					      PANIC_ON_FATAL_ERROR);
 804			else {
 805				ret = -EINVAL;
 806				goto out;
 807			}
 808			break;
 809		case Opt_commit_interval:
 810			intarg = 0;
 811			ret = match_int(&args[0], &intarg);
 812			if (ret)
 813				goto out;
 814			if (intarg == 0) {
 815				btrfs_info(info,
 816					   "using default commit interval %us",
 817					   BTRFS_DEFAULT_COMMIT_INTERVAL);
 818				intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
 819			} else if (intarg > 300) {
 820				btrfs_warn(info, "excessive commit interval %d",
 821					   intarg);
 822			}
 823			info->commit_interval = intarg;
 824			break;
 
 
 
 
 
 
 825#ifdef CONFIG_BTRFS_DEBUG
 826		case Opt_fragment_all:
 827			btrfs_info(info, "fragmenting all space");
 828			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 829			btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
 
 830			break;
 831		case Opt_fragment_metadata:
 832			btrfs_info(info, "fragmenting metadata");
 833			btrfs_set_opt(info->mount_opt,
 834				      FRAGMENT_METADATA);
 835			break;
 836		case Opt_fragment_data:
 837			btrfs_info(info, "fragmenting data");
 838			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
 839			break;
 
 
 
 
 
 
 840#endif
 841#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 842		case Opt_ref_verify:
 843			btrfs_info(info, "doing ref verification");
 844			btrfs_set_opt(info->mount_opt, REF_VERIFY);
 845			break;
 846#endif
 847		case Opt_err:
 848			btrfs_info(info, "unrecognized mount option '%s'", p);
 849			ret = -EINVAL;
 850			goto out;
 851		default:
 852			break;
 853		}
 854	}
 855check:
 856	/*
 857	 * Extra check for current option against current flag
 858	 */
 859	if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
 860		btrfs_err(info,
 861			  "nologreplay must be used with ro mount option");
 862		ret = -EINVAL;
 863	}
 864out:
 865	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
 866	    !btrfs_test_opt(info, FREE_SPACE_TREE) &&
 867	    !btrfs_test_opt(info, CLEAR_CACHE)) {
 868		btrfs_err(info, "cannot disable free space tree");
 869		ret = -EINVAL;
 870
 871	}
 872	if (!ret && btrfs_test_opt(info, SPACE_CACHE))
 873		btrfs_info(info, "disk space caching is enabled");
 874	if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
 875		btrfs_info(info, "using free space tree");
 876	return ret;
 877}
 878
 879/*
 880 * Parse mount options that are required early in the mount process.
 881 *
 882 * All other options will be parsed on much later in the mount process and
 883 * only when we need to allocate a new super block.
 884 */
 885static int btrfs_parse_early_options(const char *options, fmode_t flags,
 886		void *holder, struct btrfs_fs_devices **fs_devices)
 887{
 888	substring_t args[MAX_OPT_ARGS];
 889	char *device_name, *opts, *orig, *p;
 890	int error = 0;
 
 891
 892	if (!options)
 893		return 0;
 
 
 
 
 
 
 
 
 
 894
 895	/*
 896	 * strsep changes the string, duplicate it because btrfs_parse_options
 897	 * gets called later
 898	 */
 899	opts = kstrdup(options, GFP_KERNEL);
 900	if (!opts)
 901		return -ENOMEM;
 902	orig = opts;
 903
 904	while ((p = strsep(&opts, ",")) != NULL) {
 905		int token;
 
 
 
 
 
 906
 907		if (!*p)
 908			continue;
 909
 910		token = match_token(p, tokens, args);
 911		if (token == Opt_device) {
 912			device_name = match_strdup(&args[0]);
 913			if (!device_name) {
 914				error = -ENOMEM;
 915				goto out;
 916			}
 917			error = btrfs_scan_one_device(device_name,
 918					flags, holder, fs_devices);
 919			kfree(device_name);
 920			if (error)
 921				goto out;
 
 
 
 
 
 922		}
 
 
 923	}
 924
 925out:
 926	kfree(orig);
 927	return error;
 928}
 929
 930/*
 931 * Parse mount options that are related to subvolume id
 
 
 
 932 *
 933 * The value is later passed to mount_subvol()
 
 
 
 
 934 */
 935static int btrfs_parse_subvol_options(const char *options, fmode_t flags,
 936		char **subvol_name, u64 *subvol_objectid)
 937{
 938	substring_t args[MAX_OPT_ARGS];
 939	char *opts, *orig, *p;
 940	int error = 0;
 941	u64 subvolid;
 942
 943	if (!options)
 944		return 0;
 
 
 945
 946	/*
 947	 * strsep changes the string, duplicate it because
 948	 * btrfs_parse_early_options gets called later
 949	 */
 950	opts = kstrdup(options, GFP_KERNEL);
 951	if (!opts)
 952		return -ENOMEM;
 953	orig = opts;
 
 
 
 
 
 954
 955	while ((p = strsep(&opts, ",")) != NULL) {
 956		int token;
 957		if (!*p)
 958			continue;
 959
 960		token = match_token(p, tokens, args);
 961		switch (token) {
 962		case Opt_subvol:
 963			kfree(*subvol_name);
 964			*subvol_name = match_strdup(&args[0]);
 965			if (!*subvol_name) {
 966				error = -ENOMEM;
 967				goto out;
 968			}
 969			break;
 970		case Opt_subvolid:
 971			error = match_u64(&args[0], &subvolid);
 972			if (error)
 973				goto out;
 974
 975			/* we want the original fs_tree */
 976			if (subvolid == 0)
 977				subvolid = BTRFS_FS_TREE_OBJECTID;
 
 
 
 
 
 
 978
 979			*subvol_objectid = subvolid;
 980			break;
 981		case Opt_subvolrootid:
 982			pr_warn("BTRFS: 'subvolrootid' mount option is deprecated and has no effect\n");
 983			break;
 984		default:
 985			break;
 986		}
 987	}
 988
 989out:
 990	kfree(orig);
 991	return error;
 
 
 
 
 
 
 
 
 
 
 
 
 992}
 993
 994static char *get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
 995					   u64 subvol_objectid)
 996{
 997	struct btrfs_root *root = fs_info->tree_root;
 998	struct btrfs_root *fs_root;
 999	struct btrfs_root_ref *root_ref;
1000	struct btrfs_inode_ref *inode_ref;
1001	struct btrfs_key key;
1002	struct btrfs_path *path = NULL;
1003	char *name = NULL, *ptr;
1004	u64 dirid;
1005	int len;
1006	int ret;
1007
1008	path = btrfs_alloc_path();
1009	if (!path) {
1010		ret = -ENOMEM;
1011		goto err;
1012	}
1013	path->leave_spinning = 1;
1014
1015	name = kmalloc(PATH_MAX, GFP_KERNEL);
1016	if (!name) {
1017		ret = -ENOMEM;
1018		goto err;
1019	}
1020	ptr = name + PATH_MAX - 1;
1021	ptr[0] = '\0';
1022
1023	/*
1024	 * Walk up the subvolume trees in the tree of tree roots by root
1025	 * backrefs until we hit the top-level subvolume.
1026	 */
1027	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1028		key.objectid = subvol_objectid;
1029		key.type = BTRFS_ROOT_BACKREF_KEY;
1030		key.offset = (u64)-1;
1031
1032		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1033		if (ret < 0) {
1034			goto err;
1035		} else if (ret > 0) {
1036			ret = btrfs_previous_item(root, path, subvol_objectid,
1037						  BTRFS_ROOT_BACKREF_KEY);
1038			if (ret < 0) {
1039				goto err;
1040			} else if (ret > 0) {
1041				ret = -ENOENT;
1042				goto err;
1043			}
1044		}
1045
1046		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1047		subvol_objectid = key.offset;
1048
1049		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1050					  struct btrfs_root_ref);
1051		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1052		ptr -= len + 1;
1053		if (ptr < name) {
1054			ret = -ENAMETOOLONG;
1055			goto err;
1056		}
1057		read_extent_buffer(path->nodes[0], ptr + 1,
1058				   (unsigned long)(root_ref + 1), len);
1059		ptr[0] = '/';
1060		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1061		btrfs_release_path(path);
1062
1063		key.objectid = subvol_objectid;
1064		key.type = BTRFS_ROOT_ITEM_KEY;
1065		key.offset = (u64)-1;
1066		fs_root = btrfs_read_fs_root_no_name(fs_info, &key);
1067		if (IS_ERR(fs_root)) {
1068			ret = PTR_ERR(fs_root);
 
1069			goto err;
1070		}
1071
1072		/*
1073		 * Walk up the filesystem tree by inode refs until we hit the
1074		 * root directory.
1075		 */
1076		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1077			key.objectid = dirid;
1078			key.type = BTRFS_INODE_REF_KEY;
1079			key.offset = (u64)-1;
1080
1081			ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1082			if (ret < 0) {
1083				goto err;
1084			} else if (ret > 0) {
1085				ret = btrfs_previous_item(fs_root, path, dirid,
1086							  BTRFS_INODE_REF_KEY);
1087				if (ret < 0) {
1088					goto err;
1089				} else if (ret > 0) {
1090					ret = -ENOENT;
1091					goto err;
1092				}
1093			}
1094
1095			btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1096			dirid = key.offset;
1097
1098			inode_ref = btrfs_item_ptr(path->nodes[0],
1099						   path->slots[0],
1100						   struct btrfs_inode_ref);
1101			len = btrfs_inode_ref_name_len(path->nodes[0],
1102						       inode_ref);
1103			ptr -= len + 1;
1104			if (ptr < name) {
1105				ret = -ENAMETOOLONG;
1106				goto err;
1107			}
1108			read_extent_buffer(path->nodes[0], ptr + 1,
1109					   (unsigned long)(inode_ref + 1), len);
1110			ptr[0] = '/';
1111			btrfs_release_path(path);
1112		}
 
 
1113	}
1114
1115	btrfs_free_path(path);
1116	if (ptr == name + PATH_MAX - 1) {
1117		name[0] = '/';
1118		name[1] = '\0';
1119	} else {
1120		memmove(name, ptr, name + PATH_MAX - ptr);
1121	}
1122	return name;
1123
1124err:
 
1125	btrfs_free_path(path);
1126	kfree(name);
1127	return ERR_PTR(ret);
1128}
1129
1130static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1131{
1132	struct btrfs_root *root = fs_info->tree_root;
1133	struct btrfs_dir_item *di;
1134	struct btrfs_path *path;
1135	struct btrfs_key location;
 
1136	u64 dir_id;
1137
1138	path = btrfs_alloc_path();
1139	if (!path)
1140		return -ENOMEM;
1141	path->leave_spinning = 1;
1142
1143	/*
1144	 * Find the "default" dir item which points to the root item that we
1145	 * will mount by default if we haven't been given a specific subvolume
1146	 * to mount.
1147	 */
1148	dir_id = btrfs_super_root_dir(fs_info->super_copy);
1149	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1150	if (IS_ERR(di)) {
1151		btrfs_free_path(path);
1152		return PTR_ERR(di);
1153	}
1154	if (!di) {
1155		/*
1156		 * Ok the default dir item isn't there.  This is weird since
1157		 * it's always been there, but don't freak out, just try and
1158		 * mount the top-level subvolume.
1159		 */
1160		btrfs_free_path(path);
1161		*objectid = BTRFS_FS_TREE_OBJECTID;
1162		return 0;
1163	}
1164
1165	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1166	btrfs_free_path(path);
1167	*objectid = location.objectid;
1168	return 0;
1169}
1170
1171static int btrfs_fill_super(struct super_block *sb,
1172			    struct btrfs_fs_devices *fs_devices,
1173			    void *data)
1174{
1175	struct inode *inode;
1176	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1177	struct btrfs_key key;
1178	int err;
1179
1180	sb->s_maxbytes = MAX_LFS_FILESIZE;
1181	sb->s_magic = BTRFS_SUPER_MAGIC;
1182	sb->s_op = &btrfs_super_ops;
1183	sb->s_d_op = &btrfs_dentry_operations;
1184	sb->s_export_op = &btrfs_export_ops;
 
 
 
1185	sb->s_xattr = btrfs_xattr_handlers;
1186	sb->s_time_gran = 1;
1187#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1188	sb->s_flags |= SB_POSIXACL;
1189#endif
1190	sb->s_flags |= SB_I_VERSION;
1191	sb->s_iflags |= SB_I_CGROUPWB;
1192
1193	err = super_setup_bdi(sb);
1194	if (err) {
1195		btrfs_err(fs_info, "super_setup_bdi failed");
1196		return err;
1197	}
1198
1199	err = open_ctree(sb, fs_devices, (char *)data);
1200	if (err) {
1201		btrfs_err(fs_info, "open_ctree failed");
1202		return err;
1203	}
1204
1205	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
1206	key.type = BTRFS_INODE_ITEM_KEY;
1207	key.offset = 0;
1208	inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
1209	if (IS_ERR(inode)) {
1210		err = PTR_ERR(inode);
 
1211		goto fail_close;
1212	}
1213
1214	sb->s_root = d_make_root(inode);
1215	if (!sb->s_root) {
1216		err = -ENOMEM;
1217		goto fail_close;
1218	}
1219
1220	cleancache_init_fs(sb);
1221	sb->s_flags |= SB_ACTIVE;
1222	return 0;
1223
1224fail_close:
1225	close_ctree(fs_info);
1226	return err;
1227}
1228
1229int btrfs_sync_fs(struct super_block *sb, int wait)
1230{
1231	struct btrfs_trans_handle *trans;
1232	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1233	struct btrfs_root *root = fs_info->tree_root;
1234
1235	trace_btrfs_sync_fs(fs_info, wait);
1236
1237	if (!wait) {
1238		filemap_flush(fs_info->btree_inode->i_mapping);
1239		return 0;
1240	}
1241
1242	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1243
1244	trans = btrfs_attach_transaction_barrier(root);
1245	if (IS_ERR(trans)) {
1246		/* no transaction, don't bother */
1247		if (PTR_ERR(trans) == -ENOENT) {
1248			/*
1249			 * Exit unless we have some pending changes
1250			 * that need to go through commit
1251			 */
1252			if (fs_info->pending_changes == 0)
 
1253				return 0;
1254			/*
1255			 * A non-blocking test if the fs is frozen. We must not
1256			 * start a new transaction here otherwise a deadlock
1257			 * happens. The pending operations are delayed to the
1258			 * next commit after thawing.
1259			 */
1260			if (sb_start_write_trylock(sb))
1261				sb_end_write(sb);
1262			else
1263				return 0;
1264			trans = btrfs_start_transaction(root, 0);
1265		}
1266		if (IS_ERR(trans))
1267			return PTR_ERR(trans);
1268	}
1269	return btrfs_commit_transaction(trans);
1270}
1271
 
 
 
 
 
 
1272static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1273{
1274	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1275	const char *compress_type;
 
 
1276
1277	if (btrfs_test_opt(info, DEGRADED))
1278		seq_puts(seq, ",degraded");
1279	if (btrfs_test_opt(info, NODATASUM))
1280		seq_puts(seq, ",nodatasum");
1281	if (btrfs_test_opt(info, NODATACOW))
1282		seq_puts(seq, ",nodatacow");
1283	if (btrfs_test_opt(info, NOBARRIER))
1284		seq_puts(seq, ",nobarrier");
1285	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1286		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1287	if (info->thread_pool_size !=  min_t(unsigned long,
1288					     num_online_cpus() + 2, 8))
1289		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1290	if (btrfs_test_opt(info, COMPRESS)) {
1291		compress_type = btrfs_compress_type2str(info->compress_type);
1292		if (btrfs_test_opt(info, FORCE_COMPRESS))
1293			seq_printf(seq, ",compress-force=%s", compress_type);
1294		else
1295			seq_printf(seq, ",compress=%s", compress_type);
1296		if (info->compress_level)
1297			seq_printf(seq, ":%d", info->compress_level);
1298	}
1299	if (btrfs_test_opt(info, NOSSD))
1300		seq_puts(seq, ",nossd");
1301	if (btrfs_test_opt(info, SSD_SPREAD))
1302		seq_puts(seq, ",ssd_spread");
1303	else if (btrfs_test_opt(info, SSD))
1304		seq_puts(seq, ",ssd");
1305	if (btrfs_test_opt(info, NOTREELOG))
1306		seq_puts(seq, ",notreelog");
1307	if (btrfs_test_opt(info, NOLOGREPLAY))
1308		seq_puts(seq, ",nologreplay");
 
 
 
 
 
 
 
 
 
 
1309	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1310		seq_puts(seq, ",flushoncommit");
1311	if (btrfs_test_opt(info, DISCARD))
1312		seq_puts(seq, ",discard");
 
 
1313	if (!(info->sb->s_flags & SB_POSIXACL))
1314		seq_puts(seq, ",noacl");
1315	if (btrfs_test_opt(info, SPACE_CACHE))
1316		seq_puts(seq, ",space_cache");
1317	else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1318		seq_puts(seq, ",space_cache=v2");
1319	else
1320		seq_puts(seq, ",nospace_cache");
1321	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1322		seq_puts(seq, ",rescan_uuid_tree");
1323	if (btrfs_test_opt(info, CLEAR_CACHE))
1324		seq_puts(seq, ",clear_cache");
1325	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1326		seq_puts(seq, ",user_subvol_rm_allowed");
1327	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1328		seq_puts(seq, ",enospc_debug");
1329	if (btrfs_test_opt(info, AUTO_DEFRAG))
1330		seq_puts(seq, ",autodefrag");
1331	if (btrfs_test_opt(info, INODE_MAP_CACHE))
1332		seq_puts(seq, ",inode_cache");
1333	if (btrfs_test_opt(info, SKIP_BALANCE))
1334		seq_puts(seq, ",skip_balance");
1335#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1336	if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1337		seq_puts(seq, ",check_int_data");
1338	else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1339		seq_puts(seq, ",check_int");
1340	if (info->check_integrity_print_mask)
1341		seq_printf(seq, ",check_int_print_mask=%d",
1342				info->check_integrity_print_mask);
1343#endif
1344	if (info->metadata_ratio)
1345		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1346	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1347		seq_puts(seq, ",fatal_errors=panic");
1348	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1349		seq_printf(seq, ",commit=%u", info->commit_interval);
1350#ifdef CONFIG_BTRFS_DEBUG
1351	if (btrfs_test_opt(info, FRAGMENT_DATA))
1352		seq_puts(seq, ",fragment=data");
1353	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1354		seq_puts(seq, ",fragment=metadata");
1355#endif
1356	if (btrfs_test_opt(info, REF_VERIFY))
1357		seq_puts(seq, ",ref_verify");
1358	seq_printf(seq, ",subvolid=%llu",
1359		  BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1360	seq_puts(seq, ",subvol=");
1361	seq_dentry(seq, dentry, " \t\n\\");
 
 
 
 
1362	return 0;
1363}
1364
1365static int btrfs_test_super(struct super_block *s, void *data)
1366{
1367	struct btrfs_fs_info *p = data;
1368	struct btrfs_fs_info *fs_info = btrfs_sb(s);
1369
1370	return fs_info->fs_devices == p->fs_devices;
1371}
1372
1373static int btrfs_set_super(struct super_block *s, void *data)
1374{
1375	int err = set_anon_super(s, data);
1376	if (!err)
1377		s->s_fs_info = data;
1378	return err;
1379}
1380
1381/*
1382 * subvolumes are identified by ino 256
1383 */
1384static inline int is_subvolume_inode(struct inode *inode)
1385{
1386	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1387		return 1;
1388	return 0;
1389}
1390
1391static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1392				   const char *device_name, struct vfsmount *mnt)
1393{
1394	struct dentry *root;
1395	int ret;
1396
1397	if (!subvol_name) {
1398		if (!subvol_objectid) {
1399			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1400							  &subvol_objectid);
1401			if (ret) {
1402				root = ERR_PTR(ret);
1403				goto out;
1404			}
1405		}
1406		subvol_name = get_subvol_name_from_objectid(btrfs_sb(mnt->mnt_sb),
1407							    subvol_objectid);
1408		if (IS_ERR(subvol_name)) {
1409			root = ERR_CAST(subvol_name);
1410			subvol_name = NULL;
1411			goto out;
1412		}
1413
1414	}
1415
1416	root = mount_subtree(mnt, subvol_name);
1417	/* mount_subtree() drops our reference on the vfsmount. */
1418	mnt = NULL;
1419
1420	if (!IS_ERR(root)) {
1421		struct super_block *s = root->d_sb;
1422		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1423		struct inode *root_inode = d_inode(root);
1424		u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1425
1426		ret = 0;
1427		if (!is_subvolume_inode(root_inode)) {
1428			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1429			       subvol_name);
1430			ret = -EINVAL;
1431		}
1432		if (subvol_objectid && root_objectid != subvol_objectid) {
1433			/*
1434			 * This will also catch a race condition where a
1435			 * subvolume which was passed by ID is renamed and
1436			 * another subvolume is renamed over the old location.
1437			 */
1438			btrfs_err(fs_info,
1439				  "subvol '%s' does not match subvolid %llu",
1440				  subvol_name, subvol_objectid);
1441			ret = -EINVAL;
1442		}
1443		if (ret) {
1444			dput(root);
1445			root = ERR_PTR(ret);
1446			deactivate_locked_super(s);
1447		}
1448	}
1449
1450out:
1451	mntput(mnt);
1452	kfree(subvol_name);
1453	return root;
1454}
1455
1456static int parse_security_options(char *orig_opts,
1457				  struct security_mnt_opts *sec_opts)
1458{
1459	char *secdata = NULL;
1460	int ret = 0;
1461
1462	secdata = alloc_secdata();
1463	if (!secdata)
1464		return -ENOMEM;
1465	ret = security_sb_copy_data(orig_opts, secdata);
1466	if (ret) {
1467		free_secdata(secdata);
1468		return ret;
1469	}
1470	ret = security_sb_parse_opts_str(secdata, sec_opts);
1471	free_secdata(secdata);
1472	return ret;
1473}
1474
1475static int setup_security_options(struct btrfs_fs_info *fs_info,
1476				  struct super_block *sb,
1477				  struct security_mnt_opts *sec_opts)
1478{
1479	int ret = 0;
1480
1481	/*
1482	 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1483	 * is valid.
1484	 */
1485	ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1486	if (ret)
1487		return ret;
1488
1489#ifdef CONFIG_SECURITY
1490	if (!fs_info->security_opts.num_mnt_opts) {
1491		/* first time security setup, copy sec_opts to fs_info */
1492		memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1493	} else {
1494		/*
1495		 * Since SELinux (the only one supporting security_mnt_opts)
1496		 * does NOT support changing context during remount/mount of
1497		 * the same sb, this must be the same or part of the same
1498		 * security options, just free it.
1499		 */
1500		security_free_mnt_opts(sec_opts);
1501	}
1502#endif
1503	return ret;
1504}
1505
1506/*
1507 * Find a superblock for the given device / mount point.
1508 *
1509 * Note: This is based on mount_bdev from fs/super.c with a few additions
1510 *       for multiple device setup.  Make sure to keep it in sync.
1511 */
1512static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1513		int flags, const char *device_name, void *data)
1514{
1515	struct block_device *bdev = NULL;
1516	struct super_block *s;
1517	struct btrfs_fs_devices *fs_devices = NULL;
1518	struct btrfs_fs_info *fs_info = NULL;
1519	struct security_mnt_opts new_sec_opts;
1520	fmode_t mode = FMODE_READ;
1521	int error = 0;
1522
1523	if (!(flags & SB_RDONLY))
1524		mode |= FMODE_WRITE;
1525
1526	error = btrfs_parse_early_options(data, mode, fs_type,
1527					  &fs_devices);
1528	if (error) {
1529		return ERR_PTR(error);
1530	}
1531
1532	security_init_mnt_opts(&new_sec_opts);
1533	if (data) {
1534		error = parse_security_options(data, &new_sec_opts);
1535		if (error)
1536			return ERR_PTR(error);
1537	}
1538
1539	error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1540	if (error)
1541		goto error_sec_opts;
1542
1543	/*
1544	 * Setup a dummy root and fs_info for test/set super.  This is because
1545	 * we don't actually fill this stuff out until open_ctree, but we need
1546	 * it for searching for existing supers, so this lets us do that and
1547	 * then open_ctree will properly initialize everything later.
1548	 */
1549	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1550	if (!fs_info) {
1551		error = -ENOMEM;
1552		goto error_sec_opts;
1553	}
1554
1555	fs_info->fs_devices = fs_devices;
1556
1557	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1558	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1559	security_init_mnt_opts(&fs_info->security_opts);
1560	if (!fs_info->super_copy || !fs_info->super_for_commit) {
1561		error = -ENOMEM;
1562		goto error_fs_info;
1563	}
1564
1565	error = btrfs_open_devices(fs_devices, mode, fs_type);
1566	if (error)
1567		goto error_fs_info;
1568
1569	if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1570		error = -EACCES;
1571		goto error_close_devices;
1572	}
1573
1574	bdev = fs_devices->latest_bdev;
1575	s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1576		 fs_info);
1577	if (IS_ERR(s)) {
1578		error = PTR_ERR(s);
1579		goto error_close_devices;
1580	}
1581
1582	if (s->s_root) {
1583		btrfs_close_devices(fs_devices);
1584		free_fs_info(fs_info);
1585		if ((flags ^ s->s_flags) & SB_RDONLY)
1586			error = -EBUSY;
1587	} else {
1588		snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1589		btrfs_sb(s)->bdev_holder = fs_type;
1590		error = btrfs_fill_super(s, fs_devices, data);
1591	}
1592	if (error) {
1593		deactivate_locked_super(s);
1594		goto error_sec_opts;
1595	}
1596
1597	fs_info = btrfs_sb(s);
1598	error = setup_security_options(fs_info, s, &new_sec_opts);
1599	if (error) {
1600		deactivate_locked_super(s);
1601		goto error_sec_opts;
1602	}
1603
1604	return dget(s->s_root);
1605
1606error_close_devices:
1607	btrfs_close_devices(fs_devices);
1608error_fs_info:
1609	free_fs_info(fs_info);
1610error_sec_opts:
1611	security_free_mnt_opts(&new_sec_opts);
1612	return ERR_PTR(error);
1613}
1614
1615/*
1616 * Mount function which is called by VFS layer.
1617 *
1618 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1619 * which needs vfsmount* of device's root (/).  This means device's root has to
1620 * be mounted internally in any case.
1621 *
1622 * Operation flow:
1623 *   1. Parse subvol id related options for later use in mount_subvol().
1624 *
1625 *   2. Mount device's root (/) by calling vfs_kern_mount().
1626 *
1627 *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1628 *      first place. In order to avoid calling btrfs_mount() again, we use
1629 *      different file_system_type which is not registered to VFS by
1630 *      register_filesystem() (btrfs_root_fs_type). As a result,
1631 *      btrfs_mount_root() is called. The return value will be used by
1632 *      mount_subtree() in mount_subvol().
1633 *
1634 *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1635 *      "btrfs subvolume set-default", mount_subvol() is called always.
1636 */
1637static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1638		const char *device_name, void *data)
1639{
1640	struct vfsmount *mnt_root;
1641	struct dentry *root;
1642	fmode_t mode = FMODE_READ;
1643	char *subvol_name = NULL;
1644	u64 subvol_objectid = 0;
1645	int error = 0;
1646
1647	if (!(flags & SB_RDONLY))
1648		mode |= FMODE_WRITE;
1649
1650	error = btrfs_parse_subvol_options(data, mode,
1651					  &subvol_name, &subvol_objectid);
1652	if (error) {
1653		kfree(subvol_name);
1654		return ERR_PTR(error);
1655	}
1656
1657	/* mount device's root (/) */
1658	mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1659	if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1660		if (flags & SB_RDONLY) {
1661			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1662				flags & ~SB_RDONLY, device_name, data);
1663		} else {
1664			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1665				flags | SB_RDONLY, device_name, data);
1666			if (IS_ERR(mnt_root)) {
1667				root = ERR_CAST(mnt_root);
1668				goto out;
1669			}
1670
1671			down_write(&mnt_root->mnt_sb->s_umount);
1672			error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1673			up_write(&mnt_root->mnt_sb->s_umount);
1674			if (error < 0) {
1675				root = ERR_PTR(error);
1676				mntput(mnt_root);
1677				goto out;
1678			}
1679		}
1680	}
1681	if (IS_ERR(mnt_root)) {
1682		root = ERR_CAST(mnt_root);
1683		goto out;
1684	}
1685
1686	/* mount_subvol() will free subvol_name and mnt_root */
1687	root = mount_subvol(subvol_name, subvol_objectid, device_name, mnt_root);
1688
1689out:
1690	return root;
1691}
1692
1693static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1694				     u32 new_pool_size, u32 old_pool_size)
1695{
1696	if (new_pool_size == old_pool_size)
1697		return;
1698
1699	fs_info->thread_pool_size = new_pool_size;
1700
1701	btrfs_info(fs_info, "resize thread pool %d -> %d",
1702	       old_pool_size, new_pool_size);
1703
1704	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1705	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1706	btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1707	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1708	btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1709	btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1710	btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1711				new_pool_size);
1712	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1713	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1714	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1715	btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1716	btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1717				new_pool_size);
1718}
1719
1720static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1721{
1722	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1723}
1724
1725static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1726				       unsigned long old_opts, int flags)
1727{
1728	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1729	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1730	     (flags & SB_RDONLY))) {
1731		/* wait for any defraggers to finish */
1732		wait_event(fs_info->transaction_wait,
1733			   (atomic_read(&fs_info->defrag_running) == 0));
1734		if (flags & SB_RDONLY)
1735			sync_filesystem(fs_info->sb);
1736	}
1737}
1738
1739static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1740					 unsigned long old_opts)
1741{
 
 
1742	/*
1743	 * We need to cleanup all defragable inodes if the autodefragment is
1744	 * close or the filesystem is read only.
1745	 */
1746	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1747	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1748		btrfs_cleanup_defrag_inodes(fs_info);
1749	}
1750
1751	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
 
 
 
 
 
 
 
 
 
 
1752}
1753
1754static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1755{
1756	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1757	struct btrfs_root *root = fs_info->tree_root;
1758	unsigned old_flags = sb->s_flags;
1759	unsigned long old_opts = fs_info->mount_opt;
1760	unsigned long old_compress_type = fs_info->compress_type;
1761	u64 old_max_inline = fs_info->max_inline;
1762	u32 old_thread_pool_size = fs_info->thread_pool_size;
1763	u32 old_metadata_ratio = fs_info->metadata_ratio;
1764	int ret;
1765
1766	sync_filesystem(sb);
1767	btrfs_remount_prepare(fs_info);
 
 
 
1768
1769	if (data) {
1770		struct security_mnt_opts new_sec_opts;
1771
1772		security_init_mnt_opts(&new_sec_opts);
1773		ret = parse_security_options(data, &new_sec_opts);
1774		if (ret)
1775			goto restore;
1776		ret = setup_security_options(fs_info, sb,
1777					     &new_sec_opts);
1778		if (ret) {
1779			security_free_mnt_opts(&new_sec_opts);
1780			goto restore;
1781		}
1782	}
1783
1784	ret = btrfs_parse_options(fs_info, data, *flags);
1785	if (ret) {
1786		ret = -EINVAL;
1787		goto restore;
1788	}
1789
1790	btrfs_remount_begin(fs_info, old_opts, *flags);
1791	btrfs_resize_thread_pool(fs_info,
1792		fs_info->thread_pool_size, old_thread_pool_size);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1793
1794	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1795		goto out;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1796
1797	if (*flags & SB_RDONLY) {
1798		/*
1799		 * this also happens on 'umount -rf' or on shutdown, when
1800		 * the filesystem is busy.
1801		 */
1802		cancel_work_sync(&fs_info->async_reclaim_work);
 
1803
1804		/* wait for the uuid_scan task to finish */
1805		down(&fs_info->uuid_tree_rescan_sem);
1806		/* avoid complains from lockdep et al. */
1807		up(&fs_info->uuid_tree_rescan_sem);
1808
1809		sb->s_flags |= SB_RDONLY;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1810
1811		/*
1812		 * Setting SB_RDONLY will put the cleaner thread to
1813		 * sleep at the next loop if it's already active.
1814		 * If it's already asleep, we'll leave unused block
1815		 * groups on disk until we're mounted read-write again
1816		 * unless we clean them up here.
1817		 */
1818		btrfs_delete_unused_bgs(fs_info);
1819
1820		btrfs_dev_replace_suspend_for_unmount(fs_info);
1821		btrfs_scrub_cancel(fs_info);
1822		btrfs_pause_balance(fs_info);
1823
1824		ret = btrfs_commit_super(fs_info);
1825		if (ret)
1826			goto restore;
1827	} else {
1828		if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1829			btrfs_err(fs_info,
1830				"Remounting read-write after error is not allowed");
1831			ret = -EINVAL;
1832			goto restore;
1833		}
1834		if (fs_info->fs_devices->rw_devices == 0) {
1835			ret = -EACCES;
1836			goto restore;
1837		}
1838
1839		if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1840			btrfs_warn(fs_info,
1841				"too many missing devices, writeable remount is not allowed");
1842			ret = -EACCES;
1843			goto restore;
1844		}
1845
1846		if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1847			ret = -EINVAL;
1848			goto restore;
1849		}
 
 
 
1850
1851		ret = btrfs_cleanup_fs_roots(fs_info);
1852		if (ret)
1853			goto restore;
1854
1855		/* recover relocation */
1856		mutex_lock(&fs_info->cleaner_mutex);
1857		ret = btrfs_recover_relocation(root);
1858		mutex_unlock(&fs_info->cleaner_mutex);
1859		if (ret)
1860			goto restore;
1861
1862		ret = btrfs_resume_balance_async(fs_info);
1863		if (ret)
1864			goto restore;
1865
1866		ret = btrfs_resume_dev_replace_async(fs_info);
1867		if (ret) {
1868			btrfs_warn(fs_info, "failed to resume dev_replace");
1869			goto restore;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1870		}
 
1871
1872		btrfs_qgroup_rescan_resume(fs_info);
 
 
 
 
 
 
1873
1874		if (!fs_info->uuid_root) {
1875			btrfs_info(fs_info, "creating UUID tree");
1876			ret = btrfs_create_uuid_tree(fs_info);
1877			if (ret) {
1878				btrfs_warn(fs_info,
1879					   "failed to create the UUID tree %d",
1880					   ret);
1881				goto restore;
1882			}
1883		}
1884		sb->s_flags &= ~SB_RDONLY;
1885
1886		set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1887	}
1888out:
1889	wake_up_process(fs_info->transaction_kthread);
1890	btrfs_remount_cleanup(fs_info, old_opts);
 
 
 
1891	return 0;
1892
1893restore:
1894	/* We've hit an error - don't reset SB_RDONLY */
1895	if (sb_rdonly(sb))
1896		old_flags |= SB_RDONLY;
1897	sb->s_flags = old_flags;
1898	fs_info->mount_opt = old_opts;
1899	fs_info->compress_type = old_compress_type;
1900	fs_info->max_inline = old_max_inline;
1901	btrfs_resize_thread_pool(fs_info,
1902		old_thread_pool_size, fs_info->thread_pool_size);
1903	fs_info->metadata_ratio = old_metadata_ratio;
1904	btrfs_remount_cleanup(fs_info, old_opts);
1905	return ret;
1906}
1907
1908/* Used to sort the devices by max_avail(descending sort) */
1909static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1910				       const void *dev_info2)
1911{
1912	if (((struct btrfs_device_info *)dev_info1)->max_avail >
1913	    ((struct btrfs_device_info *)dev_info2)->max_avail)
 
 
1914		return -1;
1915	else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1916		 ((struct btrfs_device_info *)dev_info2)->max_avail)
1917		return 1;
1918	else
1919	return 0;
1920}
1921
1922/*
1923 * sort the devices by max_avail, in which max free extent size of each device
1924 * is stored.(Descending Sort)
1925 */
1926static inline void btrfs_descending_sort_devices(
1927					struct btrfs_device_info *devices,
1928					size_t nr_devices)
1929{
1930	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1931	     btrfs_cmp_device_free_bytes, NULL);
1932}
1933
1934/*
1935 * The helper to calc the free space on the devices that can be used to store
1936 * file data.
1937 */
1938static int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1939				       u64 *free_bytes)
1940{
1941	struct btrfs_device_info *devices_info;
1942	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1943	struct btrfs_device *device;
1944	u64 skip_space;
1945	u64 type;
1946	u64 avail_space;
1947	u64 min_stripe_size;
1948	int min_stripes = 1, num_stripes = 1;
1949	int i = 0, nr_devices;
 
1950
1951	/*
1952	 * We aren't under the device list lock, so this is racy-ish, but good
1953	 * enough for our purposes.
1954	 */
1955	nr_devices = fs_info->fs_devices->open_devices;
1956	if (!nr_devices) {
1957		smp_mb();
1958		nr_devices = fs_info->fs_devices->open_devices;
1959		ASSERT(nr_devices);
1960		if (!nr_devices) {
1961			*free_bytes = 0;
1962			return 0;
1963		}
1964	}
1965
1966	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1967			       GFP_KERNEL);
1968	if (!devices_info)
1969		return -ENOMEM;
1970
1971	/* calc min stripe number for data space allocation */
1972	type = btrfs_data_alloc_profile(fs_info);
1973	if (type & BTRFS_BLOCK_GROUP_RAID0) {
1974		min_stripes = 2;
 
1975		num_stripes = nr_devices;
1976	} else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1977		min_stripes = 2;
1978		num_stripes = 2;
1979	} else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1980		min_stripes = 4;
1981		num_stripes = 4;
1982	}
1983
1984	if (type & BTRFS_BLOCK_GROUP_DUP)
1985		min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1986	else
1987		min_stripe_size = BTRFS_STRIPE_LEN;
1988
1989	rcu_read_lock();
1990	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1991		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1992						&device->dev_state) ||
1993		    !device->bdev ||
1994		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1995			continue;
1996
1997		if (i >= nr_devices)
1998			break;
1999
2000		avail_space = device->total_bytes - device->bytes_used;
2001
2002		/* align with stripe_len */
2003		avail_space = div_u64(avail_space, BTRFS_STRIPE_LEN);
2004		avail_space *= BTRFS_STRIPE_LEN;
2005
2006		/*
2007		 * In order to avoid overwriting the superblock on the drive,
2008		 * btrfs starts at an offset of at least 1MB when doing chunk
2009		 * allocation.
2010		 */
2011		skip_space = SZ_1M;
 
2012
2013		/*
2014		 * we can use the free space in [0, skip_space - 1], subtract
2015		 * it from the total.
2016		 */
2017		if (avail_space && avail_space >= skip_space)
2018			avail_space -= skip_space;
2019		else
2020			avail_space = 0;
2021
2022		if (avail_space < min_stripe_size)
2023			continue;
2024
2025		devices_info[i].dev = device;
2026		devices_info[i].max_avail = avail_space;
2027
2028		i++;
2029	}
2030	rcu_read_unlock();
2031
2032	nr_devices = i;
2033
2034	btrfs_descending_sort_devices(devices_info, nr_devices);
2035
2036	i = nr_devices - 1;
2037	avail_space = 0;
2038	while (nr_devices >= min_stripes) {
2039		if (num_stripes > nr_devices)
2040			num_stripes = nr_devices;
2041
2042		if (devices_info[i].max_avail >= min_stripe_size) {
2043			int j;
2044			u64 alloc_size;
2045
2046			avail_space += devices_info[i].max_avail * num_stripes;
2047			alloc_size = devices_info[i].max_avail;
2048			for (j = i + 1 - num_stripes; j <= i; j++)
2049				devices_info[j].max_avail -= alloc_size;
2050		}
2051		i--;
2052		nr_devices--;
2053	}
2054
2055	kfree(devices_info);
2056	*free_bytes = avail_space;
2057	return 0;
2058}
2059
2060/*
2061 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2062 *
2063 * If there's a redundant raid level at DATA block groups, use the respective
2064 * multiplier to scale the sizes.
2065 *
2066 * Unused device space usage is based on simulating the chunk allocator
2067 * algorithm that respects the device sizes and order of allocations.  This is
2068 * a close approximation of the actual use but there are other factors that may
2069 * change the result (like a new metadata chunk).
2070 *
2071 * If metadata is exhausted, f_bavail will be 0.
2072 */
2073static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2074{
2075	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2076	struct btrfs_super_block *disk_super = fs_info->super_copy;
2077	struct list_head *head = &fs_info->space_info;
2078	struct btrfs_space_info *found;
2079	u64 total_used = 0;
2080	u64 total_free_data = 0;
2081	u64 total_free_meta = 0;
2082	int bits = dentry->d_sb->s_blocksize_bits;
2083	__be32 *fsid = (__be32 *)fs_info->fsid;
2084	unsigned factor = 1;
2085	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2086	int ret;
2087	u64 thresh = 0;
2088	int mixed = 0;
2089
2090	rcu_read_lock();
2091	list_for_each_entry_rcu(found, head, list) {
2092		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2093			int i;
2094
2095			total_free_data += found->disk_total - found->disk_used;
2096			total_free_data -=
2097				btrfs_account_ro_block_groups_free_space(found);
2098
2099			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2100				if (!list_empty(&found->block_groups[i])) {
2101					switch (i) {
2102					case BTRFS_RAID_DUP:
2103					case BTRFS_RAID_RAID1:
2104					case BTRFS_RAID_RAID10:
2105						factor = 2;
2106					}
2107				}
2108			}
2109		}
2110
2111		/*
2112		 * Metadata in mixed block goup profiles are accounted in data
2113		 */
2114		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2115			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2116				mixed = 1;
2117			else
2118				total_free_meta += found->disk_total -
2119					found->disk_used;
2120		}
2121
2122		total_used += found->disk_used;
2123	}
2124
2125	rcu_read_unlock();
2126
2127	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2128	buf->f_blocks >>= bits;
2129	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2130
2131	/* Account global block reserve as used, it's in logical size already */
2132	spin_lock(&block_rsv->lock);
2133	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
2134	if (buf->f_bfree >= block_rsv->size >> bits)
2135		buf->f_bfree -= block_rsv->size >> bits;
2136	else
2137		buf->f_bfree = 0;
2138	spin_unlock(&block_rsv->lock);
2139
2140	buf->f_bavail = div_u64(total_free_data, factor);
2141	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2142	if (ret)
2143		return ret;
2144	buf->f_bavail += div_u64(total_free_data, factor);
2145	buf->f_bavail = buf->f_bavail >> bits;
2146
2147	/*
2148	 * We calculate the remaining metadata space minus global reserve. If
2149	 * this is (supposedly) smaller than zero, there's no space. But this
2150	 * does not hold in practice, the exhausted state happens where's still
2151	 * some positive delta. So we apply some guesswork and compare the
2152	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2153	 *
2154	 * We probably cannot calculate the exact threshold value because this
2155	 * depends on the internal reservations requested by various
2156	 * operations, so some operations that consume a few metadata will
2157	 * succeed even if the Avail is zero. But this is better than the other
2158	 * way around.
2159	 */
2160	thresh = SZ_4M;
2161
2162	if (!mixed && total_free_meta - thresh < block_rsv->size)
 
 
 
 
 
 
 
 
2163		buf->f_bavail = 0;
2164
2165	buf->f_type = BTRFS_SUPER_MAGIC;
2166	buf->f_bsize = dentry->d_sb->s_blocksize;
2167	buf->f_namelen = BTRFS_NAME_LEN;
2168
2169	/* We treat it as constant endianness (it doesn't matter _which_)
2170	   because we want the fsid to come out the same whether mounted
2171	   on a big-endian or little-endian host */
2172	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2173	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2174	/* Mask in the root object ID too, to disambiguate subvols */
2175	buf->f_fsid.val[0] ^= BTRFS_I(d_inode(dentry))->root->objectid >> 32;
2176	buf->f_fsid.val[1] ^= BTRFS_I(d_inode(dentry))->root->objectid;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2177
 
2178	return 0;
2179}
2180
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2181static void btrfs_kill_super(struct super_block *sb)
2182{
2183	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2184	kill_anon_super(sb);
2185	free_fs_info(fs_info);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2186}
2187
2188static struct file_system_type btrfs_fs_type = {
2189	.owner		= THIS_MODULE,
2190	.name		= "btrfs",
2191	.mount		= btrfs_mount,
2192	.kill_sb	= btrfs_kill_super,
2193	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2194};
2195
2196static struct file_system_type btrfs_root_fs_type = {
2197	.owner		= THIS_MODULE,
2198	.name		= "btrfs",
2199	.mount		= btrfs_mount_root,
2200	.kill_sb	= btrfs_kill_super,
2201	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2202};
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2203
2204MODULE_ALIAS_FS("btrfs");
2205
2206static int btrfs_control_open(struct inode *inode, struct file *file)
2207{
2208	/*
2209	 * The control file's private_data is used to hold the
2210	 * transaction when it is started and is used to keep
2211	 * track of whether a transaction is already in progress.
2212	 */
2213	file->private_data = NULL;
2214	return 0;
2215}
2216
2217/*
2218 * used by btrfsctl to scan devices when no FS is mounted
2219 */
2220static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2221				unsigned long arg)
2222{
2223	struct btrfs_ioctl_vol_args *vol;
2224	struct btrfs_fs_devices *fs_devices;
 
2225	int ret = -ENOTTY;
2226
2227	if (!capable(CAP_SYS_ADMIN))
2228		return -EPERM;
2229
2230	vol = memdup_user((void __user *)arg, sizeof(*vol));
2231	if (IS_ERR(vol))
2232		return PTR_ERR(vol);
 
 
 
2233
2234	switch (cmd) {
2235	case BTRFS_IOC_SCAN_DEV:
2236		ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2237					    &btrfs_root_fs_type, &fs_devices);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2238		break;
2239	case BTRFS_IOC_DEVICES_READY:
2240		ret = btrfs_scan_one_device(vol->name, FMODE_READ,
2241					    &btrfs_root_fs_type, &fs_devices);
2242		if (ret)
 
 
 
 
 
 
 
 
 
2243			break;
2244		ret = !(fs_devices->num_devices == fs_devices->total_devices);
 
 
 
2245		break;
2246	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2247		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2248		break;
2249	}
2250
 
2251	kfree(vol);
2252	return ret;
2253}
2254
2255static int btrfs_freeze(struct super_block *sb)
2256{
2257	struct btrfs_trans_handle *trans;
2258	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2259	struct btrfs_root *root = fs_info->tree_root;
2260
2261	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2262	/*
2263	 * We don't need a barrier here, we'll wait for any transaction that
2264	 * could be in progress on other threads (and do delayed iputs that
2265	 * we want to avoid on a frozen filesystem), or do the commit
2266	 * ourselves.
2267	 */
2268	trans = btrfs_attach_transaction_barrier(root);
2269	if (IS_ERR(trans)) {
2270		/* no transaction, don't bother */
2271		if (PTR_ERR(trans) == -ENOENT)
2272			return 0;
2273		return PTR_ERR(trans);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2274	}
2275	return btrfs_commit_transaction(trans);
 
 
2276}
2277
2278static int btrfs_unfreeze(struct super_block *sb)
2279{
2280	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 
 
2281
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2282	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
 
 
 
 
 
 
2283	return 0;
2284}
2285
2286static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2287{
2288	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2289	struct btrfs_fs_devices *cur_devices;
2290	struct btrfs_device *dev, *first_dev = NULL;
2291	struct list_head *head;
2292	struct rcu_string *name;
2293
2294	/*
2295	 * Lightweight locking of the devices. We should not need
2296	 * device_list_mutex here as we only read the device data and the list
2297	 * is protected by RCU.  Even if a device is deleted during the list
2298	 * traversals, we'll get valid data, the freeing callback will wait at
2299	 * least until until the rcu_read_unlock.
2300	 */
2301	rcu_read_lock();
2302	cur_devices = fs_info->fs_devices;
2303	while (cur_devices) {
2304		head = &cur_devices->devices;
2305		list_for_each_entry_rcu(dev, head, dev_list) {
2306			if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2307				continue;
2308			if (!dev->name)
2309				continue;
2310			if (!first_dev || dev->devid < first_dev->devid)
2311				first_dev = dev;
2312		}
2313		cur_devices = cur_devices->seed;
2314	}
2315
2316	if (first_dev) {
2317		name = rcu_dereference(first_dev->name);
2318		seq_escape(m, name->str, " \t\n\\");
2319	} else {
2320		WARN_ON(1);
2321	}
2322	rcu_read_unlock();
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2323	return 0;
2324}
2325
2326static const struct super_operations btrfs_super_ops = {
2327	.drop_inode	= btrfs_drop_inode,
2328	.evict_inode	= btrfs_evict_inode,
2329	.put_super	= btrfs_put_super,
2330	.sync_fs	= btrfs_sync_fs,
2331	.show_options	= btrfs_show_options,
2332	.show_devname	= btrfs_show_devname,
2333	.write_inode	= btrfs_write_inode,
2334	.alloc_inode	= btrfs_alloc_inode,
2335	.destroy_inode	= btrfs_destroy_inode,
 
2336	.statfs		= btrfs_statfs,
2337	.remount_fs	= btrfs_remount,
2338	.freeze_fs	= btrfs_freeze,
2339	.unfreeze_fs	= btrfs_unfreeze,
 
 
2340};
2341
2342static const struct file_operations btrfs_ctl_fops = {
2343	.open = btrfs_control_open,
2344	.unlocked_ioctl	 = btrfs_control_ioctl,
2345	.compat_ioctl = btrfs_control_ioctl,
2346	.owner	 = THIS_MODULE,
2347	.llseek = noop_llseek,
2348};
2349
2350static struct miscdevice btrfs_misc = {
2351	.minor		= BTRFS_MINOR,
2352	.name		= "btrfs-control",
2353	.fops		= &btrfs_ctl_fops
2354};
2355
2356MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2357MODULE_ALIAS("devname:btrfs-control");
2358
2359static int __init btrfs_interface_init(void)
2360{
2361	return misc_register(&btrfs_misc);
2362}
2363
2364static __cold void btrfs_interface_exit(void)
2365{
2366	misc_deregister(&btrfs_misc);
2367}
2368
2369static void __init btrfs_print_mod_info(void)
2370{
2371	pr_info("Btrfs loaded, crc32c=%s"
2372#ifdef CONFIG_BTRFS_DEBUG
2373			", debug=on"
2374#endif
2375#ifdef CONFIG_BTRFS_ASSERT
2376			", assert=on"
2377#endif
2378#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2379			", integrity-checker=on"
2380#endif
2381#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2382			", ref-verify=on"
2383#endif
2384			"\n",
2385			crc32c_impl());
 
 
 
 
 
 
 
 
 
 
 
2386}
2387
2388static int __init init_btrfs_fs(void)
2389{
2390	int err;
 
2391
2392	btrfs_props_init();
 
 
 
2393
2394	err = btrfs_init_sysfs();
2395	if (err)
2396		return err;
 
 
 
2397
2398	btrfs_init_compress();
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2399
2400	err = btrfs_init_cachep();
2401	if (err)
2402		goto free_compress;
2403
2404	err = extent_io_init();
2405	if (err)
2406		goto free_cachep;
2407
2408	err = extent_map_init();
2409	if (err)
2410		goto free_extent_io;
2411
2412	err = ordered_data_init();
2413	if (err)
2414		goto free_extent_map;
2415
2416	err = btrfs_delayed_inode_init();
2417	if (err)
2418		goto free_ordered_data;
2419
2420	err = btrfs_auto_defrag_init();
2421	if (err)
2422		goto free_delayed_inode;
2423
2424	err = btrfs_delayed_ref_init();
2425	if (err)
2426		goto free_auto_defrag;
2427
2428	err = btrfs_prelim_ref_init();
2429	if (err)
2430		goto free_delayed_ref;
2431
2432	err = btrfs_end_io_wq_init();
2433	if (err)
2434		goto free_prelim_ref;
2435
2436	err = btrfs_interface_init();
2437	if (err)
2438		goto free_end_io_wq;
2439
2440	btrfs_init_lockdep();
2441
2442	btrfs_print_mod_info();
2443
2444	err = btrfs_run_sanity_tests();
2445	if (err)
2446		goto unregister_ioctl;
2447
2448	err = register_filesystem(&btrfs_fs_type);
2449	if (err)
2450		goto unregister_ioctl;
2451
2452	return 0;
2453
2454unregister_ioctl:
2455	btrfs_interface_exit();
2456free_end_io_wq:
2457	btrfs_end_io_wq_exit();
2458free_prelim_ref:
2459	btrfs_prelim_ref_exit();
2460free_delayed_ref:
2461	btrfs_delayed_ref_exit();
2462free_auto_defrag:
2463	btrfs_auto_defrag_exit();
2464free_delayed_inode:
2465	btrfs_delayed_inode_exit();
2466free_ordered_data:
2467	ordered_data_exit();
2468free_extent_map:
2469	extent_map_exit();
2470free_extent_io:
2471	extent_io_exit();
2472free_cachep:
2473	btrfs_destroy_cachep();
2474free_compress:
2475	btrfs_exit_compress();
2476	btrfs_exit_sysfs();
2477
2478	return err;
2479}
2480
2481static void __exit exit_btrfs_fs(void)
2482{
2483	btrfs_destroy_cachep();
2484	btrfs_delayed_ref_exit();
2485	btrfs_auto_defrag_exit();
2486	btrfs_delayed_inode_exit();
2487	btrfs_prelim_ref_exit();
2488	ordered_data_exit();
2489	extent_map_exit();
2490	extent_io_exit();
2491	btrfs_interface_exit();
2492	btrfs_end_io_wq_exit();
2493	unregister_filesystem(&btrfs_fs_type);
2494	btrfs_exit_sysfs();
2495	btrfs_cleanup_fs_uuids();
2496	btrfs_exit_compress();
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2497}
2498
2499late_initcall(init_btrfs_fs);
2500module_exit(exit_btrfs_fs)
2501
 
2502MODULE_LICENSE("GPL");