Linux Audio

Check our new training course

Loading...
v6.13.7
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * Copyright (C) 2007 Oracle.  All rights reserved.
   4 */
   5
   6#include <linux/kernel.h>
   7#include <linux/bio.h>
   8#include <linux/file.h>
   9#include <linux/fs.h>
  10#include <linux/fsnotify.h>
  11#include <linux/pagemap.h>
  12#include <linux/highmem.h>
  13#include <linux/time.h>
  14#include <linux/string.h>
  15#include <linux/backing-dev.h>
  16#include <linux/mount.h>
  17#include <linux/namei.h>
  18#include <linux/writeback.h>
  19#include <linux/compat.h>
  20#include <linux/security.h>
  21#include <linux/xattr.h>
  22#include <linux/mm.h>
  23#include <linux/slab.h>
  24#include <linux/blkdev.h>
  25#include <linux/uuid.h>
  26#include <linux/btrfs.h>
  27#include <linux/uaccess.h>
  28#include <linux/iversion.h>
  29#include <linux/fileattr.h>
  30#include <linux/fsverity.h>
  31#include <linux/sched/xacct.h>
  32#include <linux/io_uring/cmd.h>
  33#include "ctree.h"
  34#include "disk-io.h"
  35#include "export.h"
  36#include "transaction.h"
  37#include "btrfs_inode.h"
  38#include "volumes.h"
  39#include "locking.h"
  40#include "backref.h"
  41#include "send.h"
  42#include "dev-replace.h"
  43#include "props.h"
  44#include "sysfs.h"
  45#include "qgroup.h"
  46#include "tree-log.h"
  47#include "compression.h"
  48#include "space-info.h"
  49#include "block-group.h"
  50#include "fs.h"
  51#include "accessors.h"
  52#include "extent-tree.h"
  53#include "root-tree.h"
  54#include "defrag.h"
  55#include "dir-item.h"
  56#include "uuid-tree.h"
  57#include "ioctl.h"
  58#include "file.h"
  59#include "scrub.h"
  60#include "super.h"
  61
  62#ifdef CONFIG_64BIT
  63/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
  64 * structures are incorrect, as the timespec structure from userspace
  65 * is 4 bytes too small. We define these alternatives here to teach
  66 * the kernel about the 32-bit struct packing.
  67 */
  68struct btrfs_ioctl_timespec_32 {
  69	__u64 sec;
  70	__u32 nsec;
  71} __attribute__ ((__packed__));
  72
  73struct btrfs_ioctl_received_subvol_args_32 {
  74	char	uuid[BTRFS_UUID_SIZE];	/* in */
  75	__u64	stransid;		/* in */
  76	__u64	rtransid;		/* out */
  77	struct btrfs_ioctl_timespec_32 stime; /* in */
  78	struct btrfs_ioctl_timespec_32 rtime; /* out */
  79	__u64	flags;			/* in */
  80	__u64	reserved[16];		/* in */
  81} __attribute__ ((__packed__));
  82
  83#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
  84				struct btrfs_ioctl_received_subvol_args_32)
  85#endif
  86
  87#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
  88struct btrfs_ioctl_send_args_32 {
  89	__s64 send_fd;			/* in */
  90	__u64 clone_sources_count;	/* in */
  91	compat_uptr_t clone_sources;	/* in */
  92	__u64 parent_root;		/* in */
  93	__u64 flags;			/* in */
  94	__u32 version;			/* in */
  95	__u8  reserved[28];		/* in */
  96} __attribute__ ((__packed__));
  97
  98#define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
  99			       struct btrfs_ioctl_send_args_32)
 100
 101struct btrfs_ioctl_encoded_io_args_32 {
 102	compat_uptr_t iov;
 103	compat_ulong_t iovcnt;
 104	__s64 offset;
 105	__u64 flags;
 106	__u64 len;
 107	__u64 unencoded_len;
 108	__u64 unencoded_offset;
 109	__u32 compression;
 110	__u32 encryption;
 111	__u8 reserved[64];
 112};
 113
 114#define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
 115				       struct btrfs_ioctl_encoded_io_args_32)
 116#define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
 117					struct btrfs_ioctl_encoded_io_args_32)
 118#endif
 119
 120/* Mask out flags that are inappropriate for the given type of inode. */
 121static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
 122		unsigned int flags)
 123{
 124	if (S_ISDIR(inode->i_mode))
 125		return flags;
 126	else if (S_ISREG(inode->i_mode))
 127		return flags & ~FS_DIRSYNC_FL;
 128	else
 129		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
 130}
 131
 132/*
 133 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
 134 * ioctl.
 135 */
 136static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
 137{
 138	unsigned int iflags = 0;
 139	u32 flags = binode->flags;
 140	u32 ro_flags = binode->ro_flags;
 141
 142	if (flags & BTRFS_INODE_SYNC)
 143		iflags |= FS_SYNC_FL;
 144	if (flags & BTRFS_INODE_IMMUTABLE)
 145		iflags |= FS_IMMUTABLE_FL;
 146	if (flags & BTRFS_INODE_APPEND)
 147		iflags |= FS_APPEND_FL;
 148	if (flags & BTRFS_INODE_NODUMP)
 149		iflags |= FS_NODUMP_FL;
 150	if (flags & BTRFS_INODE_NOATIME)
 151		iflags |= FS_NOATIME_FL;
 152	if (flags & BTRFS_INODE_DIRSYNC)
 153		iflags |= FS_DIRSYNC_FL;
 154	if (flags & BTRFS_INODE_NODATACOW)
 155		iflags |= FS_NOCOW_FL;
 156	if (ro_flags & BTRFS_INODE_RO_VERITY)
 157		iflags |= FS_VERITY_FL;
 158
 159	if (flags & BTRFS_INODE_NOCOMPRESS)
 160		iflags |= FS_NOCOMP_FL;
 161	else if (flags & BTRFS_INODE_COMPRESS)
 162		iflags |= FS_COMPR_FL;
 163
 164	return iflags;
 165}
 166
 167/*
 168 * Update inode->i_flags based on the btrfs internal flags.
 169 */
 170void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
 171{
 172	struct btrfs_inode *binode = BTRFS_I(inode);
 173	unsigned int new_fl = 0;
 174
 175	if (binode->flags & BTRFS_INODE_SYNC)
 176		new_fl |= S_SYNC;
 177	if (binode->flags & BTRFS_INODE_IMMUTABLE)
 178		new_fl |= S_IMMUTABLE;
 179	if (binode->flags & BTRFS_INODE_APPEND)
 180		new_fl |= S_APPEND;
 181	if (binode->flags & BTRFS_INODE_NOATIME)
 182		new_fl |= S_NOATIME;
 183	if (binode->flags & BTRFS_INODE_DIRSYNC)
 184		new_fl |= S_DIRSYNC;
 185	if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
 186		new_fl |= S_VERITY;
 187
 188	set_mask_bits(&inode->i_flags,
 189		      S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
 190		      S_VERITY, new_fl);
 191}
 192
 193/*
 194 * Check if @flags are a supported and valid set of FS_*_FL flags and that
 195 * the old and new flags are not conflicting
 196 */
 197static int check_fsflags(unsigned int old_flags, unsigned int flags)
 198{
 199	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
 200		      FS_NOATIME_FL | FS_NODUMP_FL | \
 201		      FS_SYNC_FL | FS_DIRSYNC_FL | \
 202		      FS_NOCOMP_FL | FS_COMPR_FL |
 203		      FS_NOCOW_FL))
 204		return -EOPNOTSUPP;
 205
 206	/* COMPR and NOCOMP on new/old are valid */
 207	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
 208		return -EINVAL;
 209
 210	if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
 211		return -EINVAL;
 212
 213	/* NOCOW and compression options are mutually exclusive */
 214	if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
 215		return -EINVAL;
 216	if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
 217		return -EINVAL;
 218
 219	return 0;
 220}
 221
 222static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
 223				    unsigned int flags)
 224{
 225	if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
 226		return -EPERM;
 227
 228	return 0;
 229}
 230
 231int btrfs_check_ioctl_vol_args_path(const struct btrfs_ioctl_vol_args *vol_args)
 232{
 233	if (memchr(vol_args->name, 0, sizeof(vol_args->name)) == NULL)
 234		return -ENAMETOOLONG;
 235	return 0;
 236}
 237
 238static int btrfs_check_ioctl_vol_args2_subvol_name(const struct btrfs_ioctl_vol_args_v2 *vol_args2)
 239{
 240	if (memchr(vol_args2->name, 0, sizeof(vol_args2->name)) == NULL)
 241		return -ENAMETOOLONG;
 242	return 0;
 243}
 244
 245/*
 246 * Set flags/xflags from the internal inode flags. The remaining items of
 247 * fsxattr are zeroed.
 248 */
 249int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
 250{
 251	struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
 252
 253	fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
 254	return 0;
 255}
 256
 257int btrfs_fileattr_set(struct mnt_idmap *idmap,
 258		       struct dentry *dentry, struct fileattr *fa)
 259{
 260	struct inode *inode = d_inode(dentry);
 261	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
 262	struct btrfs_inode *binode = BTRFS_I(inode);
 263	struct btrfs_root *root = binode->root;
 264	struct btrfs_trans_handle *trans;
 265	unsigned int fsflags, old_fsflags;
 266	int ret;
 267	const char *comp = NULL;
 268	u32 binode_flags;
 269
 270	if (btrfs_root_readonly(root))
 271		return -EROFS;
 272
 273	if (fileattr_has_fsx(fa))
 274		return -EOPNOTSUPP;
 275
 276	fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
 277	old_fsflags = btrfs_inode_flags_to_fsflags(binode);
 278	ret = check_fsflags(old_fsflags, fsflags);
 279	if (ret)
 280		return ret;
 281
 282	ret = check_fsflags_compatible(fs_info, fsflags);
 283	if (ret)
 284		return ret;
 285
 286	binode_flags = binode->flags;
 287	if (fsflags & FS_SYNC_FL)
 288		binode_flags |= BTRFS_INODE_SYNC;
 289	else
 290		binode_flags &= ~BTRFS_INODE_SYNC;
 291	if (fsflags & FS_IMMUTABLE_FL)
 292		binode_flags |= BTRFS_INODE_IMMUTABLE;
 293	else
 294		binode_flags &= ~BTRFS_INODE_IMMUTABLE;
 295	if (fsflags & FS_APPEND_FL)
 296		binode_flags |= BTRFS_INODE_APPEND;
 297	else
 298		binode_flags &= ~BTRFS_INODE_APPEND;
 299	if (fsflags & FS_NODUMP_FL)
 300		binode_flags |= BTRFS_INODE_NODUMP;
 301	else
 302		binode_flags &= ~BTRFS_INODE_NODUMP;
 303	if (fsflags & FS_NOATIME_FL)
 304		binode_flags |= BTRFS_INODE_NOATIME;
 305	else
 306		binode_flags &= ~BTRFS_INODE_NOATIME;
 307
 308	/* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
 309	if (!fa->flags_valid) {
 310		/* 1 item for the inode */
 311		trans = btrfs_start_transaction(root, 1);
 312		if (IS_ERR(trans))
 313			return PTR_ERR(trans);
 314		goto update_flags;
 315	}
 316
 317	if (fsflags & FS_DIRSYNC_FL)
 318		binode_flags |= BTRFS_INODE_DIRSYNC;
 319	else
 320		binode_flags &= ~BTRFS_INODE_DIRSYNC;
 321	if (fsflags & FS_NOCOW_FL) {
 322		if (S_ISREG(inode->i_mode)) {
 323			/*
 324			 * It's safe to turn csums off here, no extents exist.
 325			 * Otherwise we want the flag to reflect the real COW
 326			 * status of the file and will not set it.
 327			 */
 328			if (inode->i_size == 0)
 329				binode_flags |= BTRFS_INODE_NODATACOW |
 330						BTRFS_INODE_NODATASUM;
 331		} else {
 332			binode_flags |= BTRFS_INODE_NODATACOW;
 333		}
 334	} else {
 335		/*
 336		 * Revert back under same assumptions as above
 337		 */
 338		if (S_ISREG(inode->i_mode)) {
 339			if (inode->i_size == 0)
 340				binode_flags &= ~(BTRFS_INODE_NODATACOW |
 341						  BTRFS_INODE_NODATASUM);
 342		} else {
 343			binode_flags &= ~BTRFS_INODE_NODATACOW;
 344		}
 345	}
 346
 347	/*
 348	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
 349	 * flag may be changed automatically if compression code won't make
 350	 * things smaller.
 351	 */
 352	if (fsflags & FS_NOCOMP_FL) {
 353		binode_flags &= ~BTRFS_INODE_COMPRESS;
 354		binode_flags |= BTRFS_INODE_NOCOMPRESS;
 355	} else if (fsflags & FS_COMPR_FL) {
 356
 357		if (IS_SWAPFILE(inode))
 358			return -ETXTBSY;
 359
 360		binode_flags |= BTRFS_INODE_COMPRESS;
 361		binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
 362
 363		comp = btrfs_compress_type2str(fs_info->compress_type);
 364		if (!comp || comp[0] == 0)
 365			comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
 366	} else {
 367		binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
 368	}
 369
 370	/*
 371	 * 1 for inode item
 372	 * 2 for properties
 373	 */
 374	trans = btrfs_start_transaction(root, 3);
 375	if (IS_ERR(trans))
 376		return PTR_ERR(trans);
 377
 378	if (comp) {
 379		ret = btrfs_set_prop(trans, BTRFS_I(inode), "btrfs.compression",
 380				     comp, strlen(comp), 0);
 381		if (ret) {
 382			btrfs_abort_transaction(trans, ret);
 383			goto out_end_trans;
 384		}
 385	} else {
 386		ret = btrfs_set_prop(trans, BTRFS_I(inode), "btrfs.compression",
 387				     NULL, 0, 0);
 388		if (ret && ret != -ENODATA) {
 389			btrfs_abort_transaction(trans, ret);
 390			goto out_end_trans;
 391		}
 392	}
 393
 394update_flags:
 395	binode->flags = binode_flags;
 396	btrfs_sync_inode_flags_to_i_flags(inode);
 397	inode_inc_iversion(inode);
 398	inode_set_ctime_current(inode);
 399	ret = btrfs_update_inode(trans, BTRFS_I(inode));
 400
 401 out_end_trans:
 402	btrfs_end_transaction(trans);
 403	return ret;
 404}
 405
 406/*
 407 * Start exclusive operation @type, return true on success
 408 */
 409bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
 410			enum btrfs_exclusive_operation type)
 411{
 412	bool ret = false;
 413
 414	spin_lock(&fs_info->super_lock);
 415	if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
 416		fs_info->exclusive_operation = type;
 417		ret = true;
 418	}
 419	spin_unlock(&fs_info->super_lock);
 420
 421	return ret;
 422}
 423
 424/*
 425 * Conditionally allow to enter the exclusive operation in case it's compatible
 426 * with the running one.  This must be paired with btrfs_exclop_start_unlock and
 427 * btrfs_exclop_finish.
 428 *
 429 * Compatibility:
 430 * - the same type is already running
 431 * - when trying to add a device and balance has been paused
 432 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
 433 *   must check the condition first that would allow none -> @type
 434 */
 435bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
 436				 enum btrfs_exclusive_operation type)
 437{
 438	spin_lock(&fs_info->super_lock);
 439	if (fs_info->exclusive_operation == type ||
 440	    (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED &&
 441	     type == BTRFS_EXCLOP_DEV_ADD))
 442		return true;
 443
 444	spin_unlock(&fs_info->super_lock);
 445	return false;
 446}
 447
 448void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
 449{
 450	spin_unlock(&fs_info->super_lock);
 451}
 452
 453void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
 454{
 455	spin_lock(&fs_info->super_lock);
 456	WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
 457	spin_unlock(&fs_info->super_lock);
 458	sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
 459}
 460
 461void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
 462			  enum btrfs_exclusive_operation op)
 463{
 464	switch (op) {
 465	case BTRFS_EXCLOP_BALANCE_PAUSED:
 466		spin_lock(&fs_info->super_lock);
 467		ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE ||
 468		       fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD ||
 469		       fs_info->exclusive_operation == BTRFS_EXCLOP_NONE ||
 470		       fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
 471		fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
 472		spin_unlock(&fs_info->super_lock);
 473		break;
 474	case BTRFS_EXCLOP_BALANCE:
 475		spin_lock(&fs_info->super_lock);
 476		ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
 477		fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
 478		spin_unlock(&fs_info->super_lock);
 479		break;
 480	default:
 481		btrfs_warn(fs_info,
 482			"invalid exclop balance operation %d requested", op);
 483	}
 484}
 485
 486static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
 487{
 488	return put_user(inode->i_generation, arg);
 489}
 490
 491static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
 492					void __user *arg)
 493{
 494	struct btrfs_device *device;
 495	struct fstrim_range range;
 496	u64 minlen = ULLONG_MAX;
 497	u64 num_devices = 0;
 498	int ret;
 499
 500	if (!capable(CAP_SYS_ADMIN))
 501		return -EPERM;
 502
 503	/*
 504	 * btrfs_trim_block_group() depends on space cache, which is not
 505	 * available in zoned filesystem. So, disallow fitrim on a zoned
 506	 * filesystem for now.
 507	 */
 508	if (btrfs_is_zoned(fs_info))
 509		return -EOPNOTSUPP;
 510
 511	/*
 512	 * If the fs is mounted with nologreplay, which requires it to be
 513	 * mounted in RO mode as well, we can not allow discard on free space
 514	 * inside block groups, because log trees refer to extents that are not
 515	 * pinned in a block group's free space cache (pinning the extents is
 516	 * precisely the first phase of replaying a log tree).
 517	 */
 518	if (btrfs_test_opt(fs_info, NOLOGREPLAY))
 519		return -EROFS;
 520
 521	rcu_read_lock();
 522	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
 523				dev_list) {
 524		if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
 525			continue;
 526		num_devices++;
 527		minlen = min_t(u64, bdev_discard_granularity(device->bdev),
 528				    minlen);
 529	}
 530	rcu_read_unlock();
 531
 532	if (!num_devices)
 533		return -EOPNOTSUPP;
 534	if (copy_from_user(&range, arg, sizeof(range)))
 535		return -EFAULT;
 536
 537	/*
 538	 * NOTE: Don't truncate the range using super->total_bytes.  Bytenr of
 539	 * block group is in the logical address space, which can be any
 540	 * sectorsize aligned bytenr in  the range [0, U64_MAX].
 541	 */
 542	if (range.len < fs_info->sectorsize)
 543		return -EINVAL;
 544
 545	range.minlen = max(range.minlen, minlen);
 546	ret = btrfs_trim_fs(fs_info, &range);
 
 
 547
 548	if (copy_to_user(arg, &range, sizeof(range)))
 549		return -EFAULT;
 550
 551	return ret;
 552}
 553
 554int __pure btrfs_is_empty_uuid(const u8 *uuid)
 555{
 556	int i;
 557
 558	for (i = 0; i < BTRFS_UUID_SIZE; i++) {
 559		if (uuid[i])
 560			return 0;
 561	}
 562	return 1;
 563}
 564
 565/*
 566 * Calculate the number of transaction items to reserve for creating a subvolume
 567 * or snapshot, not including the inode, directory entries, or parent directory.
 568 */
 569static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
 570{
 571	/*
 572	 * 1 to add root block
 573	 * 1 to add root item
 574	 * 1 to add root ref
 575	 * 1 to add root backref
 576	 * 1 to add UUID item
 577	 * 1 to add qgroup info
 578	 * 1 to add qgroup limit
 579	 *
 580	 * Ideally the last two would only be accounted if qgroups are enabled,
 581	 * but that can change between now and the time we would insert them.
 582	 */
 583	unsigned int num_items = 7;
 584
 585	if (inherit) {
 586		/* 2 to add qgroup relations for each inherited qgroup */
 587		num_items += 2 * inherit->num_qgroups;
 588	}
 589	return num_items;
 590}
 591
 592static noinline int create_subvol(struct mnt_idmap *idmap,
 593				  struct inode *dir, struct dentry *dentry,
 594				  struct btrfs_qgroup_inherit *inherit)
 595{
 596	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
 597	struct btrfs_trans_handle *trans;
 598	struct btrfs_key key;
 599	struct btrfs_root_item *root_item;
 600	struct btrfs_inode_item *inode_item;
 601	struct extent_buffer *leaf;
 602	struct btrfs_root *root = BTRFS_I(dir)->root;
 603	struct btrfs_root *new_root;
 604	struct btrfs_block_rsv block_rsv;
 605	struct timespec64 cur_time = current_time(dir);
 606	struct btrfs_new_inode_args new_inode_args = {
 607		.dir = dir,
 608		.dentry = dentry,
 609		.subvol = true,
 610	};
 611	unsigned int trans_num_items;
 612	int ret;
 613	dev_t anon_dev;
 614	u64 objectid;
 615	u64 qgroup_reserved = 0;
 616
 617	root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
 618	if (!root_item)
 619		return -ENOMEM;
 620
 621	ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
 622	if (ret)
 623		goto out_root_item;
 624
 625	/*
 626	 * Don't create subvolume whose level is not zero. Or qgroup will be
 627	 * screwed up since it assumes subvolume qgroup's level to be 0.
 628	 */
 629	if (btrfs_qgroup_level(objectid)) {
 630		ret = -ENOSPC;
 631		goto out_root_item;
 632	}
 633
 634	ret = get_anon_bdev(&anon_dev);
 635	if (ret < 0)
 636		goto out_root_item;
 637
 638	new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir);
 639	if (!new_inode_args.inode) {
 640		ret = -ENOMEM;
 641		goto out_anon_dev;
 642	}
 643	ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
 644	if (ret)
 645		goto out_inode;
 646	trans_num_items += create_subvol_num_items(inherit);
 647
 648	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
 649	ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
 650					       trans_num_items, false);
 651	if (ret)
 652		goto out_new_inode_args;
 653	qgroup_reserved = block_rsv.qgroup_rsv_reserved;
 654
 655	trans = btrfs_start_transaction(root, 0);
 656	if (IS_ERR(trans)) {
 657		ret = PTR_ERR(trans);
 658		goto out_release_rsv;
 659	}
 
 
 
 660	btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved);
 661	qgroup_reserved = 0;
 662	trans->block_rsv = &block_rsv;
 663	trans->bytes_reserved = block_rsv.size;
 
 
 664
 665	ret = btrfs_qgroup_inherit(trans, 0, objectid, btrfs_root_id(root), inherit);
 666	if (ret)
 667		goto out;
 668
 669	leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
 670				      0, BTRFS_NESTING_NORMAL);
 671	if (IS_ERR(leaf)) {
 672		ret = PTR_ERR(leaf);
 673		goto out;
 674	}
 675
 676	btrfs_mark_buffer_dirty(trans, leaf);
 677
 678	inode_item = &root_item->inode;
 679	btrfs_set_stack_inode_generation(inode_item, 1);
 680	btrfs_set_stack_inode_size(inode_item, 3);
 681	btrfs_set_stack_inode_nlink(inode_item, 1);
 682	btrfs_set_stack_inode_nbytes(inode_item,
 683				     fs_info->nodesize);
 684	btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
 685
 686	btrfs_set_root_flags(root_item, 0);
 687	btrfs_set_root_limit(root_item, 0);
 688	btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
 689
 690	btrfs_set_root_bytenr(root_item, leaf->start);
 691	btrfs_set_root_generation(root_item, trans->transid);
 692	btrfs_set_root_level(root_item, 0);
 693	btrfs_set_root_refs(root_item, 1);
 694	btrfs_set_root_used(root_item, leaf->len);
 695	btrfs_set_root_last_snapshot(root_item, 0);
 696
 697	btrfs_set_root_generation_v2(root_item,
 698			btrfs_root_generation(root_item));
 699	generate_random_guid(root_item->uuid);
 700	btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
 701	btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
 702	root_item->ctime = root_item->otime;
 703	btrfs_set_root_ctransid(root_item, trans->transid);
 704	btrfs_set_root_otransid(root_item, trans->transid);
 705
 706	btrfs_tree_unlock(leaf);
 707
 708	btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
 709
 710	key.objectid = objectid;
 711	key.offset = 0;
 712	key.type = BTRFS_ROOT_ITEM_KEY;
 713	ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
 714				root_item);
 715	if (ret) {
 716		int ret2;
 717
 718		/*
 719		 * Since we don't abort the transaction in this case, free the
 720		 * tree block so that we don't leak space and leave the
 721		 * filesystem in an inconsistent state (an extent item in the
 722		 * extent tree with a backreference for a root that does not
 723		 * exists).
 724		 */
 725		btrfs_tree_lock(leaf);
 726		btrfs_clear_buffer_dirty(trans, leaf);
 727		btrfs_tree_unlock(leaf);
 728		ret2 = btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
 729		if (ret2 < 0)
 730			btrfs_abort_transaction(trans, ret2);
 731		free_extent_buffer(leaf);
 732		goto out;
 733	}
 734
 735	free_extent_buffer(leaf);
 736	leaf = NULL;
 737
 738	new_root = btrfs_get_new_fs_root(fs_info, objectid, &anon_dev);
 739	if (IS_ERR(new_root)) {
 740		ret = PTR_ERR(new_root);
 741		btrfs_abort_transaction(trans, ret);
 742		goto out;
 743	}
 744	/* anon_dev is owned by new_root now. */
 745	anon_dev = 0;
 746	BTRFS_I(new_inode_args.inode)->root = new_root;
 747	/* ... and new_root is owned by new_inode_args.inode now. */
 748
 749	ret = btrfs_record_root_in_trans(trans, new_root);
 750	if (ret) {
 751		btrfs_abort_transaction(trans, ret);
 752		goto out;
 753	}
 754
 755	ret = btrfs_uuid_tree_add(trans, root_item->uuid,
 756				  BTRFS_UUID_KEY_SUBVOL, objectid);
 757	if (ret) {
 758		btrfs_abort_transaction(trans, ret);
 759		goto out;
 760	}
 761
 762	ret = btrfs_create_new_inode(trans, &new_inode_args);
 763	if (ret) {
 764		btrfs_abort_transaction(trans, ret);
 765		goto out;
 766	}
 767
 768	btrfs_record_new_subvolume(trans, BTRFS_I(dir));
 769
 770	d_instantiate_new(dentry, new_inode_args.inode);
 771	new_inode_args.inode = NULL;
 772
 773out:
 774	trans->block_rsv = NULL;
 775	trans->bytes_reserved = 0;
 776	btrfs_end_transaction(trans);
 777out_release_rsv:
 778	btrfs_block_rsv_release(fs_info, &block_rsv, (u64)-1, NULL);
 779	if (qgroup_reserved)
 780		btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved);
 781out_new_inode_args:
 782	btrfs_new_inode_args_destroy(&new_inode_args);
 783out_inode:
 784	iput(new_inode_args.inode);
 785out_anon_dev:
 786	if (anon_dev)
 787		free_anon_bdev(anon_dev);
 788out_root_item:
 789	kfree(root_item);
 790	return ret;
 791}
 792
 793static int create_snapshot(struct btrfs_root *root, struct inode *dir,
 794			   struct dentry *dentry, bool readonly,
 795			   struct btrfs_qgroup_inherit *inherit)
 796{
 797	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
 798	struct inode *inode;
 799	struct btrfs_pending_snapshot *pending_snapshot;
 800	unsigned int trans_num_items;
 801	struct btrfs_trans_handle *trans;
 802	struct btrfs_block_rsv *block_rsv;
 803	u64 qgroup_reserved = 0;
 804	int ret;
 805
 806	/* We do not support snapshotting right now. */
 807	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
 808		btrfs_warn(fs_info,
 809			   "extent tree v2 doesn't support snapshotting yet");
 810		return -EOPNOTSUPP;
 811	}
 812
 813	if (btrfs_root_refs(&root->root_item) == 0)
 814		return -ENOENT;
 815
 816	if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
 817		return -EINVAL;
 818
 819	if (atomic_read(&root->nr_swapfiles)) {
 820		btrfs_warn(fs_info,
 821			   "cannot snapshot subvolume with active swapfile");
 822		return -ETXTBSY;
 823	}
 824
 825	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
 826	if (!pending_snapshot)
 827		return -ENOMEM;
 828
 829	ret = get_anon_bdev(&pending_snapshot->anon_dev);
 830	if (ret < 0)
 831		goto free_pending;
 832	pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
 833			GFP_KERNEL);
 834	pending_snapshot->path = btrfs_alloc_path();
 835	if (!pending_snapshot->root_item || !pending_snapshot->path) {
 836		ret = -ENOMEM;
 837		goto free_pending;
 838	}
 839
 840	block_rsv = &pending_snapshot->block_rsv;
 841	btrfs_init_block_rsv(block_rsv, BTRFS_BLOCK_RSV_TEMP);
 842	/*
 843	 * 1 to add dir item
 844	 * 1 to add dir index
 845	 * 1 to update parent inode item
 846	 */
 847	trans_num_items = create_subvol_num_items(inherit) + 3;
 848	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, block_rsv,
 849					       trans_num_items, false);
 850	if (ret)
 851		goto free_pending;
 852	qgroup_reserved = block_rsv->qgroup_rsv_reserved;
 853
 854	pending_snapshot->dentry = dentry;
 855	pending_snapshot->root = root;
 856	pending_snapshot->readonly = readonly;
 857	pending_snapshot->dir = BTRFS_I(dir);
 858	pending_snapshot->inherit = inherit;
 859
 860	trans = btrfs_start_transaction(root, 0);
 861	if (IS_ERR(trans)) {
 862		ret = PTR_ERR(trans);
 863		goto fail;
 864	}
 865	ret = btrfs_record_root_in_trans(trans, BTRFS_I(dir)->root);
 866	if (ret) {
 867		btrfs_end_transaction(trans);
 868		goto fail;
 869	}
 870	btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved);
 871	qgroup_reserved = 0;
 872
 873	trans->pending_snapshot = pending_snapshot;
 874
 875	ret = btrfs_commit_transaction(trans);
 876	if (ret)
 877		goto fail;
 878
 879	ret = pending_snapshot->error;
 880	if (ret)
 881		goto fail;
 882
 883	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
 884	if (ret)
 885		goto fail;
 886
 887	inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
 888	if (IS_ERR(inode)) {
 889		ret = PTR_ERR(inode);
 890		goto fail;
 891	}
 892
 893	d_instantiate(dentry, inode);
 894	ret = 0;
 895	pending_snapshot->anon_dev = 0;
 896fail:
 897	/* Prevent double freeing of anon_dev */
 898	if (ret && pending_snapshot->snap)
 899		pending_snapshot->snap->anon_dev = 0;
 900	btrfs_put_root(pending_snapshot->snap);
 901	btrfs_block_rsv_release(fs_info, block_rsv, (u64)-1, NULL);
 902	if (qgroup_reserved)
 903		btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved);
 904free_pending:
 905	if (pending_snapshot->anon_dev)
 906		free_anon_bdev(pending_snapshot->anon_dev);
 907	kfree(pending_snapshot->root_item);
 908	btrfs_free_path(pending_snapshot->path);
 909	kfree(pending_snapshot);
 910
 911	return ret;
 912}
 913
 914/*  copy of may_delete in fs/namei.c()
 915 *	Check whether we can remove a link victim from directory dir, check
 916 *  whether the type of victim is right.
 917 *  1. We can't do it if dir is read-only (done in permission())
 918 *  2. We should have write and exec permissions on dir
 919 *  3. We can't remove anything from append-only dir
 920 *  4. We can't do anything with immutable dir (done in permission())
 921 *  5. If the sticky bit on dir is set we should either
 922 *	a. be owner of dir, or
 923 *	b. be owner of victim, or
 924 *	c. have CAP_FOWNER capability
 925 *  6. If the victim is append-only or immutable we can't do anything with
 926 *     links pointing to it.
 927 *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
 928 *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
 929 *  9. We can't remove a root or mountpoint.
 930 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
 931 *     nfs_async_unlink().
 932 */
 933
 934static int btrfs_may_delete(struct mnt_idmap *idmap,
 935			    struct inode *dir, struct dentry *victim, int isdir)
 936{
 937	int error;
 938
 939	if (d_really_is_negative(victim))
 940		return -ENOENT;
 941
 942	/* The @victim is not inside @dir. */
 943	if (d_inode(victim->d_parent) != dir)
 944		return -EINVAL;
 945	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
 946
 947	error = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
 948	if (error)
 949		return error;
 950	if (IS_APPEND(dir))
 951		return -EPERM;
 952	if (check_sticky(idmap, dir, d_inode(victim)) ||
 953	    IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
 954	    IS_SWAPFILE(d_inode(victim)))
 955		return -EPERM;
 956	if (isdir) {
 957		if (!d_is_dir(victim))
 958			return -ENOTDIR;
 959		if (IS_ROOT(victim))
 960			return -EBUSY;
 961	} else if (d_is_dir(victim))
 962		return -EISDIR;
 963	if (IS_DEADDIR(dir))
 964		return -ENOENT;
 965	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
 966		return -EBUSY;
 967	return 0;
 968}
 969
 970/* copy of may_create in fs/namei.c() */
 971static inline int btrfs_may_create(struct mnt_idmap *idmap,
 972				   struct inode *dir, struct dentry *child)
 973{
 974	if (d_really_is_positive(child))
 975		return -EEXIST;
 976	if (IS_DEADDIR(dir))
 977		return -ENOENT;
 978	if (!fsuidgid_has_mapping(dir->i_sb, idmap))
 979		return -EOVERFLOW;
 980	return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
 981}
 982
 983/*
 984 * Create a new subvolume below @parent.  This is largely modeled after
 985 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 986 * inside this filesystem so it's quite a bit simpler.
 987 */
 988static noinline int btrfs_mksubvol(const struct path *parent,
 989				   struct mnt_idmap *idmap,
 990				   const char *name, int namelen,
 991				   struct btrfs_root *snap_src,
 992				   bool readonly,
 993				   struct btrfs_qgroup_inherit *inherit)
 994{
 995	struct inode *dir = d_inode(parent->dentry);
 996	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
 997	struct dentry *dentry;
 998	struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
 999	int error;
1000
1001	error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
1002	if (error == -EINTR)
1003		return error;
1004
1005	dentry = lookup_one(idmap, name, parent->dentry, namelen);
1006	error = PTR_ERR(dentry);
1007	if (IS_ERR(dentry))
1008		goto out_unlock;
1009
1010	error = btrfs_may_create(idmap, dir, dentry);
1011	if (error)
1012		goto out_dput;
1013
1014	/*
1015	 * even if this name doesn't exist, we may get hash collisions.
1016	 * check for them now when we can safely fail
1017	 */
1018	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
1019					       dir->i_ino, &name_str);
1020	if (error)
1021		goto out_dput;
1022
1023	down_read(&fs_info->subvol_sem);
1024
1025	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
1026		goto out_up_read;
1027
1028	if (snap_src)
1029		error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
1030	else
1031		error = create_subvol(idmap, dir, dentry, inherit);
1032
1033	if (!error)
1034		fsnotify_mkdir(dir, dentry);
1035out_up_read:
1036	up_read(&fs_info->subvol_sem);
1037out_dput:
1038	dput(dentry);
1039out_unlock:
1040	btrfs_inode_unlock(BTRFS_I(dir), 0);
1041	return error;
1042}
1043
1044static noinline int btrfs_mksnapshot(const struct path *parent,
1045				   struct mnt_idmap *idmap,
1046				   const char *name, int namelen,
1047				   struct btrfs_root *root,
1048				   bool readonly,
1049				   struct btrfs_qgroup_inherit *inherit)
1050{
1051	int ret;
 
1052
1053	/*
1054	 * Force new buffered writes to reserve space even when NOCOW is
1055	 * possible. This is to avoid later writeback (running dealloc) to
1056	 * fallback to COW mode and unexpectedly fail with ENOSPC.
1057	 */
1058	btrfs_drew_read_lock(&root->snapshot_lock);
1059
1060	ret = btrfs_start_delalloc_snapshot(root, false);
1061	if (ret)
1062		goto out;
1063
1064	/*
1065	 * All previous writes have started writeback in NOCOW mode, so now
1066	 * we force future writes to fallback to COW mode during snapshot
1067	 * creation.
1068	 */
1069	atomic_inc(&root->snapshot_force_cow);
 
1070
1071	btrfs_wait_ordered_extents(root, U64_MAX, NULL);
1072
1073	ret = btrfs_mksubvol(parent, idmap, name, namelen,
1074			     root, readonly, inherit);
1075	atomic_dec(&root->snapshot_force_cow);
1076out:
 
 
1077	btrfs_drew_read_unlock(&root->snapshot_lock);
1078	return ret;
1079}
1080
1081/*
1082 * Try to start exclusive operation @type or cancel it if it's running.
1083 *
1084 * Return:
1085 *   0        - normal mode, newly claimed op started
1086 *  >0        - normal mode, something else is running,
1087 *              return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1088 * ECANCELED  - cancel mode, successful cancel
1089 * ENOTCONN   - cancel mode, operation not running anymore
1090 */
1091static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1092			enum btrfs_exclusive_operation type, bool cancel)
1093{
1094	if (!cancel) {
1095		/* Start normal op */
1096		if (!btrfs_exclop_start(fs_info, type))
1097			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1098		/* Exclusive operation is now claimed */
1099		return 0;
1100	}
1101
1102	/* Cancel running op */
1103	if (btrfs_exclop_start_try_lock(fs_info, type)) {
1104		/*
1105		 * This blocks any exclop finish from setting it to NONE, so we
1106		 * request cancellation. Either it runs and we will wait for it,
1107		 * or it has finished and no waiting will happen.
1108		 */
1109		atomic_inc(&fs_info->reloc_cancel_req);
1110		btrfs_exclop_start_unlock(fs_info);
1111
1112		if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1113			wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1114				    TASK_INTERRUPTIBLE);
1115
1116		return -ECANCELED;
1117	}
1118
1119	/* Something else is running or none */
1120	return -ENOTCONN;
1121}
1122
1123static noinline int btrfs_ioctl_resize(struct file *file,
1124					void __user *arg)
1125{
1126	BTRFS_DEV_LOOKUP_ARGS(args);
1127	struct inode *inode = file_inode(file);
1128	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1129	u64 new_size;
1130	u64 old_size;
1131	u64 devid = 1;
1132	struct btrfs_root *root = BTRFS_I(inode)->root;
1133	struct btrfs_ioctl_vol_args *vol_args;
1134	struct btrfs_trans_handle *trans;
1135	struct btrfs_device *device = NULL;
1136	char *sizestr;
1137	char *retptr;
1138	char *devstr = NULL;
1139	int ret = 0;
1140	int mod = 0;
1141	bool cancel;
1142
1143	if (!capable(CAP_SYS_ADMIN))
1144		return -EPERM;
1145
1146	ret = mnt_want_write_file(file);
1147	if (ret)
1148		return ret;
1149
1150	/*
1151	 * Read the arguments before checking exclusivity to be able to
1152	 * distinguish regular resize and cancel
1153	 */
1154	vol_args = memdup_user(arg, sizeof(*vol_args));
1155	if (IS_ERR(vol_args)) {
1156		ret = PTR_ERR(vol_args);
1157		goto out_drop;
1158	}
1159	ret = btrfs_check_ioctl_vol_args_path(vol_args);
1160	if (ret < 0)
1161		goto out_free;
1162
1163	sizestr = vol_args->name;
1164	cancel = (strcmp("cancel", sizestr) == 0);
1165	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1166	if (ret)
1167		goto out_free;
1168	/* Exclusive operation is now claimed */
1169
1170	devstr = strchr(sizestr, ':');
1171	if (devstr) {
1172		sizestr = devstr + 1;
1173		*devstr = '\0';
1174		devstr = vol_args->name;
1175		ret = kstrtoull(devstr, 10, &devid);
1176		if (ret)
1177			goto out_finish;
1178		if (!devid) {
1179			ret = -EINVAL;
1180			goto out_finish;
1181		}
1182		btrfs_info(fs_info, "resizing devid %llu", devid);
1183	}
1184
1185	args.devid = devid;
1186	device = btrfs_find_device(fs_info->fs_devices, &args);
1187	if (!device) {
1188		btrfs_info(fs_info, "resizer unable to find device %llu",
1189			   devid);
1190		ret = -ENODEV;
1191		goto out_finish;
1192	}
1193
1194	if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1195		btrfs_info(fs_info,
1196			   "resizer unable to apply on readonly device %llu",
1197		       devid);
1198		ret = -EPERM;
1199		goto out_finish;
1200	}
1201
1202	if (!strcmp(sizestr, "max"))
1203		new_size = bdev_nr_bytes(device->bdev);
1204	else {
1205		if (sizestr[0] == '-') {
1206			mod = -1;
1207			sizestr++;
1208		} else if (sizestr[0] == '+') {
1209			mod = 1;
1210			sizestr++;
1211		}
1212		new_size = memparse(sizestr, &retptr);
1213		if (*retptr != '\0' || new_size == 0) {
1214			ret = -EINVAL;
1215			goto out_finish;
1216		}
1217	}
1218
1219	if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1220		ret = -EPERM;
1221		goto out_finish;
1222	}
1223
1224	old_size = btrfs_device_get_total_bytes(device);
1225
1226	if (mod < 0) {
1227		if (new_size > old_size) {
1228			ret = -EINVAL;
1229			goto out_finish;
1230		}
1231		new_size = old_size - new_size;
1232	} else if (mod > 0) {
1233		if (new_size > ULLONG_MAX - old_size) {
1234			ret = -ERANGE;
1235			goto out_finish;
1236		}
1237		new_size = old_size + new_size;
1238	}
1239
1240	if (new_size < SZ_256M) {
1241		ret = -EINVAL;
1242		goto out_finish;
1243	}
1244	if (new_size > bdev_nr_bytes(device->bdev)) {
1245		ret = -EFBIG;
1246		goto out_finish;
1247	}
1248
1249	new_size = round_down(new_size, fs_info->sectorsize);
1250
1251	if (new_size > old_size) {
1252		trans = btrfs_start_transaction(root, 0);
1253		if (IS_ERR(trans)) {
1254			ret = PTR_ERR(trans);
1255			goto out_finish;
1256		}
1257		ret = btrfs_grow_device(trans, device, new_size);
1258		btrfs_commit_transaction(trans);
1259	} else if (new_size < old_size) {
1260		ret = btrfs_shrink_device(device, new_size);
1261	} /* equal, nothing need to do */
1262
1263	if (ret == 0 && new_size != old_size)
1264		btrfs_info_in_rcu(fs_info,
1265			"resize device %s (devid %llu) from %llu to %llu",
1266			btrfs_dev_name(device), device->devid,
1267			old_size, new_size);
1268out_finish:
1269	btrfs_exclop_finish(fs_info);
1270out_free:
1271	kfree(vol_args);
1272out_drop:
1273	mnt_drop_write_file(file);
1274	return ret;
1275}
1276
1277static noinline int __btrfs_ioctl_snap_create(struct file *file,
1278				struct mnt_idmap *idmap,
1279				const char *name, unsigned long fd, int subvol,
1280				bool readonly,
1281				struct btrfs_qgroup_inherit *inherit)
1282{
1283	int namelen;
1284	int ret = 0;
1285
1286	if (!S_ISDIR(file_inode(file)->i_mode))
1287		return -ENOTDIR;
1288
1289	ret = mnt_want_write_file(file);
1290	if (ret)
1291		goto out;
1292
1293	namelen = strlen(name);
1294	if (strchr(name, '/')) {
1295		ret = -EINVAL;
1296		goto out_drop_write;
1297	}
1298
1299	if (name[0] == '.' &&
1300	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1301		ret = -EEXIST;
1302		goto out_drop_write;
1303	}
1304
1305	if (subvol) {
1306		ret = btrfs_mksubvol(&file->f_path, idmap, name,
1307				     namelen, NULL, readonly, inherit);
1308	} else {
1309		CLASS(fd, src)(fd);
1310		struct inode *src_inode;
1311		if (fd_empty(src)) {
1312			ret = -EINVAL;
1313			goto out_drop_write;
1314		}
1315
1316		src_inode = file_inode(fd_file(src));
1317		if (src_inode->i_sb != file_inode(file)->i_sb) {
1318			btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1319				   "Snapshot src from another FS");
1320			ret = -EXDEV;
1321		} else if (!inode_owner_or_capable(idmap, src_inode)) {
1322			/*
1323			 * Subvolume creation is not restricted, but snapshots
1324			 * are limited to own subvolumes only
1325			 */
1326			ret = -EPERM;
1327		} else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1328			/*
1329			 * Snapshots must be made with the src_inode referring
1330			 * to the subvolume inode, otherwise the permission
1331			 * checking above is useless because we may have
1332			 * permission on a lower directory but not the subvol
1333			 * itself.
1334			 */
1335			ret = -EINVAL;
1336		} else {
1337			ret = btrfs_mksnapshot(&file->f_path, idmap,
1338					       name, namelen,
1339					       BTRFS_I(src_inode)->root,
1340					       readonly, inherit);
1341		}
 
1342	}
1343out_drop_write:
1344	mnt_drop_write_file(file);
1345out:
1346	return ret;
1347}
1348
1349static noinline int btrfs_ioctl_snap_create(struct file *file,
1350					    void __user *arg, int subvol)
1351{
1352	struct btrfs_ioctl_vol_args *vol_args;
1353	int ret;
1354
1355	if (!S_ISDIR(file_inode(file)->i_mode))
1356		return -ENOTDIR;
1357
1358	vol_args = memdup_user(arg, sizeof(*vol_args));
1359	if (IS_ERR(vol_args))
1360		return PTR_ERR(vol_args);
1361	ret = btrfs_check_ioctl_vol_args_path(vol_args);
1362	if (ret < 0)
1363		goto out;
1364
1365	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1366					vol_args->name, vol_args->fd, subvol,
1367					false, NULL);
1368
1369out:
1370	kfree(vol_args);
1371	return ret;
1372}
1373
1374static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1375					       void __user *arg, int subvol)
1376{
1377	struct btrfs_ioctl_vol_args_v2 *vol_args;
1378	int ret;
1379	bool readonly = false;
1380	struct btrfs_qgroup_inherit *inherit = NULL;
1381
1382	if (!S_ISDIR(file_inode(file)->i_mode))
1383		return -ENOTDIR;
1384
1385	vol_args = memdup_user(arg, sizeof(*vol_args));
1386	if (IS_ERR(vol_args))
1387		return PTR_ERR(vol_args);
1388	ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args);
1389	if (ret < 0)
1390		goto free_args;
1391
1392	if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1393		ret = -EOPNOTSUPP;
1394		goto free_args;
1395	}
1396
1397	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1398		readonly = true;
1399	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1400		struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file));
1401
1402		if (vol_args->size < sizeof(*inherit) ||
1403		    vol_args->size > PAGE_SIZE) {
1404			ret = -EINVAL;
1405			goto free_args;
1406		}
1407		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1408		if (IS_ERR(inherit)) {
1409			ret = PTR_ERR(inherit);
1410			goto free_args;
1411		}
1412
1413		ret = btrfs_qgroup_check_inherit(fs_info, inherit, vol_args->size);
1414		if (ret < 0)
1415			goto free_inherit;
1416	}
1417
1418	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1419					vol_args->name, vol_args->fd, subvol,
1420					readonly, inherit);
1421	if (ret)
1422		goto free_inherit;
1423free_inherit:
1424	kfree(inherit);
1425free_args:
1426	kfree(vol_args);
1427	return ret;
1428}
1429
1430static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1431						void __user *arg)
1432{
1433	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1434	struct btrfs_root *root = BTRFS_I(inode)->root;
1435	int ret = 0;
1436	u64 flags = 0;
1437
1438	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1439		return -EINVAL;
1440
1441	down_read(&fs_info->subvol_sem);
1442	if (btrfs_root_readonly(root))
1443		flags |= BTRFS_SUBVOL_RDONLY;
1444	up_read(&fs_info->subvol_sem);
1445
1446	if (copy_to_user(arg, &flags, sizeof(flags)))
1447		ret = -EFAULT;
1448
1449	return ret;
1450}
1451
1452static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1453					      void __user *arg)
1454{
1455	struct inode *inode = file_inode(file);
1456	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1457	struct btrfs_root *root = BTRFS_I(inode)->root;
1458	struct btrfs_trans_handle *trans;
1459	u64 root_flags;
1460	u64 flags;
1461	int ret = 0;
1462
1463	if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1464		return -EPERM;
1465
1466	ret = mnt_want_write_file(file);
1467	if (ret)
1468		goto out;
1469
1470	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1471		ret = -EINVAL;
1472		goto out_drop_write;
1473	}
1474
1475	if (copy_from_user(&flags, arg, sizeof(flags))) {
1476		ret = -EFAULT;
1477		goto out_drop_write;
1478	}
1479
1480	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1481		ret = -EOPNOTSUPP;
1482		goto out_drop_write;
1483	}
1484
1485	down_write(&fs_info->subvol_sem);
1486
1487	/* nothing to do */
1488	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1489		goto out_drop_sem;
1490
1491	root_flags = btrfs_root_flags(&root->root_item);
1492	if (flags & BTRFS_SUBVOL_RDONLY) {
1493		btrfs_set_root_flags(&root->root_item,
1494				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1495	} else {
1496		/*
1497		 * Block RO -> RW transition if this subvolume is involved in
1498		 * send
1499		 */
1500		spin_lock(&root->root_item_lock);
1501		if (root->send_in_progress == 0) {
1502			btrfs_set_root_flags(&root->root_item,
1503				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1504			spin_unlock(&root->root_item_lock);
1505		} else {
1506			spin_unlock(&root->root_item_lock);
1507			btrfs_warn(fs_info,
1508				   "Attempt to set subvolume %llu read-write during send",
1509				   btrfs_root_id(root));
1510			ret = -EPERM;
1511			goto out_drop_sem;
1512		}
1513	}
1514
1515	trans = btrfs_start_transaction(root, 1);
1516	if (IS_ERR(trans)) {
1517		ret = PTR_ERR(trans);
1518		goto out_reset;
1519	}
1520
1521	ret = btrfs_update_root(trans, fs_info->tree_root,
1522				&root->root_key, &root->root_item);
1523	if (ret < 0) {
1524		btrfs_end_transaction(trans);
1525		goto out_reset;
1526	}
1527
1528	ret = btrfs_commit_transaction(trans);
1529
1530out_reset:
1531	if (ret)
1532		btrfs_set_root_flags(&root->root_item, root_flags);
1533out_drop_sem:
1534	up_write(&fs_info->subvol_sem);
1535out_drop_write:
1536	mnt_drop_write_file(file);
1537out:
1538	return ret;
1539}
1540
1541static noinline int key_in_sk(struct btrfs_key *key,
1542			      struct btrfs_ioctl_search_key *sk)
1543{
1544	struct btrfs_key test;
1545	int ret;
1546
1547	test.objectid = sk->min_objectid;
1548	test.type = sk->min_type;
1549	test.offset = sk->min_offset;
1550
1551	ret = btrfs_comp_cpu_keys(key, &test);
1552	if (ret < 0)
1553		return 0;
1554
1555	test.objectid = sk->max_objectid;
1556	test.type = sk->max_type;
1557	test.offset = sk->max_offset;
1558
1559	ret = btrfs_comp_cpu_keys(key, &test);
1560	if (ret > 0)
1561		return 0;
1562	return 1;
1563}
1564
1565static noinline int copy_to_sk(struct btrfs_path *path,
1566			       struct btrfs_key *key,
1567			       struct btrfs_ioctl_search_key *sk,
1568			       u64 *buf_size,
1569			       char __user *ubuf,
1570			       unsigned long *sk_offset,
1571			       int *num_found)
1572{
1573	u64 found_transid;
1574	struct extent_buffer *leaf;
1575	struct btrfs_ioctl_search_header sh;
1576	struct btrfs_key test;
1577	unsigned long item_off;
1578	unsigned long item_len;
1579	int nritems;
1580	int i;
1581	int slot;
1582	int ret = 0;
1583
1584	leaf = path->nodes[0];
1585	slot = path->slots[0];
1586	nritems = btrfs_header_nritems(leaf);
1587
1588	if (btrfs_header_generation(leaf) > sk->max_transid) {
1589		i = nritems;
1590		goto advance_key;
1591	}
1592	found_transid = btrfs_header_generation(leaf);
1593
1594	for (i = slot; i < nritems; i++) {
1595		item_off = btrfs_item_ptr_offset(leaf, i);
1596		item_len = btrfs_item_size(leaf, i);
1597
1598		btrfs_item_key_to_cpu(leaf, key, i);
1599		if (!key_in_sk(key, sk))
1600			continue;
1601
1602		if (sizeof(sh) + item_len > *buf_size) {
1603			if (*num_found) {
1604				ret = 1;
1605				goto out;
1606			}
1607
1608			/*
1609			 * return one empty item back for v1, which does not
1610			 * handle -EOVERFLOW
1611			 */
1612
1613			*buf_size = sizeof(sh) + item_len;
1614			item_len = 0;
1615			ret = -EOVERFLOW;
1616		}
1617
1618		if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1619			ret = 1;
1620			goto out;
1621		}
1622
1623		sh.objectid = key->objectid;
1624		sh.offset = key->offset;
1625		sh.type = key->type;
1626		sh.len = item_len;
1627		sh.transid = found_transid;
1628
1629		/*
1630		 * Copy search result header. If we fault then loop again so we
1631		 * can fault in the pages and -EFAULT there if there's a
1632		 * problem. Otherwise we'll fault and then copy the buffer in
1633		 * properly this next time through
1634		 */
1635		if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1636			ret = 0;
1637			goto out;
1638		}
1639
1640		*sk_offset += sizeof(sh);
1641
1642		if (item_len) {
1643			char __user *up = ubuf + *sk_offset;
1644			/*
1645			 * Copy the item, same behavior as above, but reset the
1646			 * * sk_offset so we copy the full thing again.
1647			 */
1648			if (read_extent_buffer_to_user_nofault(leaf, up,
1649						item_off, item_len)) {
1650				ret = 0;
1651				*sk_offset -= sizeof(sh);
1652				goto out;
1653			}
1654
1655			*sk_offset += item_len;
1656		}
1657		(*num_found)++;
1658
1659		if (ret) /* -EOVERFLOW from above */
1660			goto out;
1661
1662		if (*num_found >= sk->nr_items) {
1663			ret = 1;
1664			goto out;
1665		}
1666	}
1667advance_key:
1668	ret = 0;
1669	test.objectid = sk->max_objectid;
1670	test.type = sk->max_type;
1671	test.offset = sk->max_offset;
1672	if (btrfs_comp_cpu_keys(key, &test) >= 0)
1673		ret = 1;
1674	else if (key->offset < (u64)-1)
1675		key->offset++;
1676	else if (key->type < (u8)-1) {
1677		key->offset = 0;
1678		key->type++;
1679	} else if (key->objectid < (u64)-1) {
1680		key->offset = 0;
1681		key->type = 0;
1682		key->objectid++;
1683	} else
1684		ret = 1;
1685out:
1686	/*
1687	 *  0: all items from this leaf copied, continue with next
1688	 *  1: * more items can be copied, but unused buffer is too small
1689	 *     * all items were found
1690	 *     Either way, it will stops the loop which iterates to the next
1691	 *     leaf
1692	 *  -EOVERFLOW: item was to large for buffer
1693	 *  -EFAULT: could not copy extent buffer back to userspace
1694	 */
1695	return ret;
1696}
1697
1698static noinline int search_ioctl(struct inode *inode,
1699				 struct btrfs_ioctl_search_key *sk,
1700				 u64 *buf_size,
1701				 char __user *ubuf)
1702{
1703	struct btrfs_fs_info *info = inode_to_fs_info(inode);
1704	struct btrfs_root *root;
1705	struct btrfs_key key;
1706	struct btrfs_path *path;
1707	int ret;
1708	int num_found = 0;
1709	unsigned long sk_offset = 0;
1710
1711	if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1712		*buf_size = sizeof(struct btrfs_ioctl_search_header);
1713		return -EOVERFLOW;
1714	}
1715
1716	path = btrfs_alloc_path();
1717	if (!path)
1718		return -ENOMEM;
1719
1720	if (sk->tree_id == 0) {
1721		/* search the root of the inode that was passed */
1722		root = btrfs_grab_root(BTRFS_I(inode)->root);
1723	} else {
1724		root = btrfs_get_fs_root(info, sk->tree_id, true);
1725		if (IS_ERR(root)) {
1726			btrfs_free_path(path);
1727			return PTR_ERR(root);
1728		}
1729	}
1730
1731	key.objectid = sk->min_objectid;
1732	key.type = sk->min_type;
1733	key.offset = sk->min_offset;
1734
1735	while (1) {
1736		ret = -EFAULT;
1737		/*
1738		 * Ensure that the whole user buffer is faulted in at sub-page
1739		 * granularity, otherwise the loop may live-lock.
1740		 */
1741		if (fault_in_subpage_writeable(ubuf + sk_offset,
1742					       *buf_size - sk_offset))
1743			break;
1744
1745		ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1746		if (ret != 0) {
1747			if (ret > 0)
1748				ret = 0;
1749			goto err;
1750		}
1751		ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1752				 &sk_offset, &num_found);
1753		btrfs_release_path(path);
1754		if (ret)
1755			break;
1756
1757	}
1758	if (ret > 0)
1759		ret = 0;
1760err:
1761	sk->nr_items = num_found;
1762	btrfs_put_root(root);
1763	btrfs_free_path(path);
1764	return ret;
1765}
1766
1767static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1768					    void __user *argp)
1769{
1770	struct btrfs_ioctl_search_args __user *uargs = argp;
1771	struct btrfs_ioctl_search_key sk;
1772	int ret;
1773	u64 buf_size;
1774
1775	if (!capable(CAP_SYS_ADMIN))
1776		return -EPERM;
1777
1778	if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1779		return -EFAULT;
1780
1781	buf_size = sizeof(uargs->buf);
1782
1783	ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1784
1785	/*
1786	 * In the origin implementation an overflow is handled by returning a
1787	 * search header with a len of zero, so reset ret.
1788	 */
1789	if (ret == -EOVERFLOW)
1790		ret = 0;
1791
1792	if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1793		ret = -EFAULT;
1794	return ret;
1795}
1796
1797static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1798					       void __user *argp)
1799{
1800	struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1801	struct btrfs_ioctl_search_args_v2 args;
1802	int ret;
1803	u64 buf_size;
1804	const u64 buf_limit = SZ_16M;
1805
1806	if (!capable(CAP_SYS_ADMIN))
1807		return -EPERM;
1808
1809	/* copy search header and buffer size */
1810	if (copy_from_user(&args, uarg, sizeof(args)))
1811		return -EFAULT;
1812
1813	buf_size = args.buf_size;
1814
1815	/* limit result size to 16MB */
1816	if (buf_size > buf_limit)
1817		buf_size = buf_limit;
1818
1819	ret = search_ioctl(inode, &args.key, &buf_size,
1820			   (char __user *)(&uarg->buf[0]));
1821	if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1822		ret = -EFAULT;
1823	else if (ret == -EOVERFLOW &&
1824		copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1825		ret = -EFAULT;
1826
1827	return ret;
1828}
1829
1830/*
1831 * Search INODE_REFs to identify path name of 'dirid' directory
1832 * in a 'tree_id' tree. and sets path name to 'name'.
1833 */
1834static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1835				u64 tree_id, u64 dirid, char *name)
1836{
1837	struct btrfs_root *root;
1838	struct btrfs_key key;
1839	char *ptr;
1840	int ret = -1;
1841	int slot;
1842	int len;
1843	int total_len = 0;
1844	struct btrfs_inode_ref *iref;
1845	struct extent_buffer *l;
1846	struct btrfs_path *path;
1847
1848	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1849		name[0]='\0';
1850		return 0;
1851	}
1852
1853	path = btrfs_alloc_path();
1854	if (!path)
1855		return -ENOMEM;
1856
1857	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1858
1859	root = btrfs_get_fs_root(info, tree_id, true);
1860	if (IS_ERR(root)) {
1861		ret = PTR_ERR(root);
1862		root = NULL;
1863		goto out;
1864	}
1865
1866	key.objectid = dirid;
1867	key.type = BTRFS_INODE_REF_KEY;
1868	key.offset = (u64)-1;
1869
1870	while (1) {
1871		ret = btrfs_search_backwards(root, &key, path);
1872		if (ret < 0)
1873			goto out;
1874		else if (ret > 0) {
1875			ret = -ENOENT;
1876			goto out;
1877		}
1878
1879		l = path->nodes[0];
1880		slot = path->slots[0];
1881
1882		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1883		len = btrfs_inode_ref_name_len(l, iref);
1884		ptr -= len + 1;
1885		total_len += len + 1;
1886		if (ptr < name) {
1887			ret = -ENAMETOOLONG;
1888			goto out;
1889		}
1890
1891		*(ptr + len) = '/';
1892		read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1893
1894		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1895			break;
1896
1897		btrfs_release_path(path);
1898		key.objectid = key.offset;
1899		key.offset = (u64)-1;
1900		dirid = key.objectid;
1901	}
1902	memmove(name, ptr, total_len);
1903	name[total_len] = '\0';
1904	ret = 0;
1905out:
1906	btrfs_put_root(root);
1907	btrfs_free_path(path);
1908	return ret;
1909}
1910
1911static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1912				struct inode *inode,
1913				struct btrfs_ioctl_ino_lookup_user_args *args)
1914{
1915	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1916	u64 upper_limit = btrfs_ino(BTRFS_I(inode));
1917	u64 treeid = btrfs_root_id(BTRFS_I(inode)->root);
 
1918	u64 dirid = args->dirid;
1919	unsigned long item_off;
1920	unsigned long item_len;
1921	struct btrfs_inode_ref *iref;
1922	struct btrfs_root_ref *rref;
1923	struct btrfs_root *root = NULL;
1924	struct btrfs_path *path;
1925	struct btrfs_key key, key2;
1926	struct extent_buffer *leaf;
1927	struct inode *temp_inode;
1928	char *ptr;
1929	int slot;
1930	int len;
1931	int total_len = 0;
1932	int ret;
1933
1934	path = btrfs_alloc_path();
1935	if (!path)
1936		return -ENOMEM;
1937
1938	/*
1939	 * If the bottom subvolume does not exist directly under upper_limit,
1940	 * construct the path in from the bottom up.
1941	 */
1942	if (dirid != upper_limit) {
1943		ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1944
1945		root = btrfs_get_fs_root(fs_info, treeid, true);
1946		if (IS_ERR(root)) {
1947			ret = PTR_ERR(root);
1948			goto out;
1949		}
1950
1951		key.objectid = dirid;
1952		key.type = BTRFS_INODE_REF_KEY;
1953		key.offset = (u64)-1;
1954		while (1) {
1955			ret = btrfs_search_backwards(root, &key, path);
1956			if (ret < 0)
1957				goto out_put;
1958			else if (ret > 0) {
1959				ret = -ENOENT;
1960				goto out_put;
1961			}
1962
1963			leaf = path->nodes[0];
1964			slot = path->slots[0];
1965
1966			iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1967			len = btrfs_inode_ref_name_len(leaf, iref);
1968			ptr -= len + 1;
1969			total_len += len + 1;
1970			if (ptr < args->path) {
1971				ret = -ENAMETOOLONG;
1972				goto out_put;
1973			}
1974
1975			*(ptr + len) = '/';
1976			read_extent_buffer(leaf, ptr,
1977					(unsigned long)(iref + 1), len);
1978
1979			/* Check the read+exec permission of this directory */
1980			ret = btrfs_previous_item(root, path, dirid,
1981						  BTRFS_INODE_ITEM_KEY);
1982			if (ret < 0) {
1983				goto out_put;
1984			} else if (ret > 0) {
1985				ret = -ENOENT;
1986				goto out_put;
1987			}
1988
1989			leaf = path->nodes[0];
1990			slot = path->slots[0];
1991			btrfs_item_key_to_cpu(leaf, &key2, slot);
1992			if (key2.objectid != dirid) {
1993				ret = -ENOENT;
1994				goto out_put;
1995			}
1996
1997			/*
1998			 * We don't need the path anymore, so release it and
1999			 * avoid deadlocks and lockdep warnings in case
2000			 * btrfs_iget() needs to lookup the inode from its root
2001			 * btree and lock the same leaf.
2002			 */
2003			btrfs_release_path(path);
2004			temp_inode = btrfs_iget(key2.objectid, root);
2005			if (IS_ERR(temp_inode)) {
2006				ret = PTR_ERR(temp_inode);
2007				goto out_put;
2008			}
2009			ret = inode_permission(idmap, temp_inode,
2010					       MAY_READ | MAY_EXEC);
2011			iput(temp_inode);
2012			if (ret) {
2013				ret = -EACCES;
2014				goto out_put;
2015			}
2016
2017			if (key.offset == upper_limit)
2018				break;
2019			if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2020				ret = -EACCES;
2021				goto out_put;
2022			}
2023
2024			key.objectid = key.offset;
2025			key.offset = (u64)-1;
2026			dirid = key.objectid;
2027		}
2028
2029		memmove(args->path, ptr, total_len);
2030		args->path[total_len] = '\0';
2031		btrfs_put_root(root);
2032		root = NULL;
2033		btrfs_release_path(path);
2034	}
2035
2036	/* Get the bottom subvolume's name from ROOT_REF */
2037	key.objectid = treeid;
2038	key.type = BTRFS_ROOT_REF_KEY;
2039	key.offset = args->treeid;
2040	ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2041	if (ret < 0) {
2042		goto out;
2043	} else if (ret > 0) {
2044		ret = -ENOENT;
2045		goto out;
2046	}
2047
2048	leaf = path->nodes[0];
2049	slot = path->slots[0];
2050	btrfs_item_key_to_cpu(leaf, &key, slot);
2051
2052	item_off = btrfs_item_ptr_offset(leaf, slot);
2053	item_len = btrfs_item_size(leaf, slot);
2054	/* Check if dirid in ROOT_REF corresponds to passed dirid */
2055	rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2056	if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2057		ret = -EINVAL;
2058		goto out;
2059	}
2060
2061	/* Copy subvolume's name */
2062	item_off += sizeof(struct btrfs_root_ref);
2063	item_len -= sizeof(struct btrfs_root_ref);
2064	read_extent_buffer(leaf, args->name, item_off, item_len);
2065	args->name[item_len] = 0;
2066
2067out_put:
2068	btrfs_put_root(root);
2069out:
2070	btrfs_free_path(path);
2071	return ret;
2072}
2073
2074static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2075					   void __user *argp)
2076{
2077	struct btrfs_ioctl_ino_lookup_args *args;
2078	int ret = 0;
2079
2080	args = memdup_user(argp, sizeof(*args));
2081	if (IS_ERR(args))
2082		return PTR_ERR(args);
2083
2084	/*
2085	 * Unprivileged query to obtain the containing subvolume root id. The
2086	 * path is reset so it's consistent with btrfs_search_path_in_tree.
2087	 */
2088	if (args->treeid == 0)
2089		args->treeid = btrfs_root_id(root);
2090
2091	if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2092		args->name[0] = 0;
2093		goto out;
2094	}
2095
2096	if (!capable(CAP_SYS_ADMIN)) {
2097		ret = -EPERM;
2098		goto out;
2099	}
2100
2101	ret = btrfs_search_path_in_tree(root->fs_info,
2102					args->treeid, args->objectid,
2103					args->name);
2104
2105out:
2106	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2107		ret = -EFAULT;
2108
2109	kfree(args);
2110	return ret;
2111}
2112
2113/*
2114 * Version of ino_lookup ioctl (unprivileged)
2115 *
2116 * The main differences from ino_lookup ioctl are:
2117 *
2118 *   1. Read + Exec permission will be checked using inode_permission() during
2119 *      path construction. -EACCES will be returned in case of failure.
2120 *   2. Path construction will be stopped at the inode number which corresponds
2121 *      to the fd with which this ioctl is called. If constructed path does not
2122 *      exist under fd's inode, -EACCES will be returned.
2123 *   3. The name of bottom subvolume is also searched and filled.
2124 */
2125static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2126{
2127	struct btrfs_ioctl_ino_lookup_user_args *args;
2128	struct inode *inode;
2129	int ret;
2130
2131	args = memdup_user(argp, sizeof(*args));
2132	if (IS_ERR(args))
2133		return PTR_ERR(args);
2134
2135	inode = file_inode(file);
2136
2137	if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2138	    btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2139		/*
2140		 * The subvolume does not exist under fd with which this is
2141		 * called
2142		 */
2143		kfree(args);
2144		return -EACCES;
2145	}
2146
2147	ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2148
2149	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2150		ret = -EFAULT;
2151
2152	kfree(args);
2153	return ret;
2154}
2155
2156/* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2157static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2158{
2159	struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2160	struct btrfs_fs_info *fs_info;
2161	struct btrfs_root *root;
2162	struct btrfs_path *path;
2163	struct btrfs_key key;
2164	struct btrfs_root_item *root_item;
2165	struct btrfs_root_ref *rref;
2166	struct extent_buffer *leaf;
2167	unsigned long item_off;
2168	unsigned long item_len;
2169	int slot;
2170	int ret = 0;
2171
2172	path = btrfs_alloc_path();
2173	if (!path)
2174		return -ENOMEM;
2175
2176	subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2177	if (!subvol_info) {
2178		btrfs_free_path(path);
2179		return -ENOMEM;
2180	}
2181
2182	fs_info = BTRFS_I(inode)->root->fs_info;
2183
2184	/* Get root_item of inode's subvolume */
2185	key.objectid = btrfs_root_id(BTRFS_I(inode)->root);
2186	root = btrfs_get_fs_root(fs_info, key.objectid, true);
2187	if (IS_ERR(root)) {
2188		ret = PTR_ERR(root);
2189		goto out_free;
2190	}
2191	root_item = &root->root_item;
2192
2193	subvol_info->treeid = key.objectid;
2194
2195	subvol_info->generation = btrfs_root_generation(root_item);
2196	subvol_info->flags = btrfs_root_flags(root_item);
2197
2198	memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2199	memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2200						    BTRFS_UUID_SIZE);
2201	memcpy(subvol_info->received_uuid, root_item->received_uuid,
2202						    BTRFS_UUID_SIZE);
2203
2204	subvol_info->ctransid = btrfs_root_ctransid(root_item);
2205	subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2206	subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2207
2208	subvol_info->otransid = btrfs_root_otransid(root_item);
2209	subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2210	subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2211
2212	subvol_info->stransid = btrfs_root_stransid(root_item);
2213	subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2214	subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2215
2216	subvol_info->rtransid = btrfs_root_rtransid(root_item);
2217	subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2218	subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2219
2220	if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2221		/* Search root tree for ROOT_BACKREF of this subvolume */
2222		key.type = BTRFS_ROOT_BACKREF_KEY;
2223		key.offset = 0;
2224		ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2225		if (ret < 0) {
2226			goto out;
2227		} else if (path->slots[0] >=
2228			   btrfs_header_nritems(path->nodes[0])) {
2229			ret = btrfs_next_leaf(fs_info->tree_root, path);
2230			if (ret < 0) {
2231				goto out;
2232			} else if (ret > 0) {
2233				ret = -EUCLEAN;
2234				goto out;
2235			}
2236		}
2237
2238		leaf = path->nodes[0];
2239		slot = path->slots[0];
2240		btrfs_item_key_to_cpu(leaf, &key, slot);
2241		if (key.objectid == subvol_info->treeid &&
2242		    key.type == BTRFS_ROOT_BACKREF_KEY) {
2243			subvol_info->parent_id = key.offset;
2244
2245			rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2246			subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2247
2248			item_off = btrfs_item_ptr_offset(leaf, slot)
2249					+ sizeof(struct btrfs_root_ref);
2250			item_len = btrfs_item_size(leaf, slot)
2251					- sizeof(struct btrfs_root_ref);
2252			read_extent_buffer(leaf, subvol_info->name,
2253					   item_off, item_len);
2254		} else {
2255			ret = -ENOENT;
2256			goto out;
2257		}
2258	}
2259
2260	btrfs_free_path(path);
2261	path = NULL;
2262	if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2263		ret = -EFAULT;
2264
2265out:
2266	btrfs_put_root(root);
2267out_free:
2268	btrfs_free_path(path);
2269	kfree(subvol_info);
2270	return ret;
2271}
2272
2273/*
2274 * Return ROOT_REF information of the subvolume containing this inode
2275 * except the subvolume name.
2276 */
2277static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2278					  void __user *argp)
2279{
2280	struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2281	struct btrfs_root_ref *rref;
2282	struct btrfs_path *path;
2283	struct btrfs_key key;
2284	struct extent_buffer *leaf;
2285	u64 objectid;
2286	int slot;
2287	int ret;
2288	u8 found;
2289
2290	path = btrfs_alloc_path();
2291	if (!path)
2292		return -ENOMEM;
2293
2294	rootrefs = memdup_user(argp, sizeof(*rootrefs));
2295	if (IS_ERR(rootrefs)) {
2296		btrfs_free_path(path);
2297		return PTR_ERR(rootrefs);
2298	}
2299
2300	objectid = btrfs_root_id(root);
2301	key.objectid = objectid;
2302	key.type = BTRFS_ROOT_REF_KEY;
2303	key.offset = rootrefs->min_treeid;
2304	found = 0;
2305
2306	root = root->fs_info->tree_root;
2307	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2308	if (ret < 0) {
2309		goto out;
2310	} else if (path->slots[0] >=
2311		   btrfs_header_nritems(path->nodes[0])) {
2312		ret = btrfs_next_leaf(root, path);
2313		if (ret < 0) {
2314			goto out;
2315		} else if (ret > 0) {
2316			ret = -EUCLEAN;
2317			goto out;
2318		}
2319	}
2320	while (1) {
2321		leaf = path->nodes[0];
2322		slot = path->slots[0];
2323
2324		btrfs_item_key_to_cpu(leaf, &key, slot);
2325		if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2326			ret = 0;
2327			goto out;
2328		}
2329
2330		if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2331			ret = -EOVERFLOW;
2332			goto out;
2333		}
2334
2335		rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2336		rootrefs->rootref[found].treeid = key.offset;
2337		rootrefs->rootref[found].dirid =
2338				  btrfs_root_ref_dirid(leaf, rref);
2339		found++;
2340
2341		ret = btrfs_next_item(root, path);
2342		if (ret < 0) {
2343			goto out;
2344		} else if (ret > 0) {
2345			ret = -EUCLEAN;
2346			goto out;
2347		}
2348	}
2349
2350out:
2351	btrfs_free_path(path);
2352
2353	if (!ret || ret == -EOVERFLOW) {
2354		rootrefs->num_items = found;
2355		/* update min_treeid for next search */
2356		if (found)
2357			rootrefs->min_treeid =
2358				rootrefs->rootref[found - 1].treeid + 1;
2359		if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2360			ret = -EFAULT;
2361	}
2362
2363	kfree(rootrefs);
2364
2365	return ret;
2366}
2367
2368static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2369					     void __user *arg,
2370					     bool destroy_v2)
2371{
2372	struct dentry *parent = file->f_path.dentry;
2373	struct dentry *dentry;
2374	struct inode *dir = d_inode(parent);
2375	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
2376	struct inode *inode;
2377	struct btrfs_root *root = BTRFS_I(dir)->root;
2378	struct btrfs_root *dest = NULL;
2379	struct btrfs_ioctl_vol_args *vol_args = NULL;
2380	struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2381	struct mnt_idmap *idmap = file_mnt_idmap(file);
2382	char *subvol_name, *subvol_name_ptr = NULL;
2383	int subvol_namelen;
2384	int ret = 0;
2385	bool destroy_parent = false;
2386
2387	/* We don't support snapshots with extent tree v2 yet. */
2388	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2389		btrfs_err(fs_info,
2390			  "extent tree v2 doesn't support snapshot deletion yet");
2391		return -EOPNOTSUPP;
2392	}
2393
2394	if (destroy_v2) {
2395		vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2396		if (IS_ERR(vol_args2))
2397			return PTR_ERR(vol_args2);
2398
2399		if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2400			ret = -EOPNOTSUPP;
2401			goto out;
2402		}
2403
2404		/*
2405		 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2406		 * name, same as v1 currently does.
2407		 */
2408		if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2409			ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args2);
2410			if (ret < 0)
2411				goto out;
2412			subvol_name = vol_args2->name;
2413
2414			ret = mnt_want_write_file(file);
2415			if (ret)
2416				goto out;
2417		} else {
2418			struct inode *old_dir;
2419
2420			if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2421				ret = -EINVAL;
2422				goto out;
2423			}
2424
2425			ret = mnt_want_write_file(file);
2426			if (ret)
2427				goto out;
2428
2429			dentry = btrfs_get_dentry(fs_info->sb,
2430					BTRFS_FIRST_FREE_OBJECTID,
2431					vol_args2->subvolid, 0);
2432			if (IS_ERR(dentry)) {
2433				ret = PTR_ERR(dentry);
2434				goto out_drop_write;
2435			}
2436
2437			/*
2438			 * Change the default parent since the subvolume being
2439			 * deleted can be outside of the current mount point.
2440			 */
2441			parent = btrfs_get_parent(dentry);
2442
2443			/*
2444			 * At this point dentry->d_name can point to '/' if the
2445			 * subvolume we want to destroy is outsite of the
2446			 * current mount point, so we need to release the
2447			 * current dentry and execute the lookup to return a new
2448			 * one with ->d_name pointing to the
2449			 * <mount point>/subvol_name.
2450			 */
2451			dput(dentry);
2452			if (IS_ERR(parent)) {
2453				ret = PTR_ERR(parent);
2454				goto out_drop_write;
2455			}
2456			old_dir = dir;
2457			dir = d_inode(parent);
2458
2459			/*
2460			 * If v2 was used with SPEC_BY_ID, a new parent was
2461			 * allocated since the subvolume can be outside of the
2462			 * current mount point. Later on we need to release this
2463			 * new parent dentry.
2464			 */
2465			destroy_parent = true;
2466
2467			/*
2468			 * On idmapped mounts, deletion via subvolid is
2469			 * restricted to subvolumes that are immediate
2470			 * ancestors of the inode referenced by the file
2471			 * descriptor in the ioctl. Otherwise the idmapping
2472			 * could potentially be abused to delete subvolumes
2473			 * anywhere in the filesystem the user wouldn't be able
2474			 * to delete without an idmapped mount.
2475			 */
2476			if (old_dir != dir && idmap != &nop_mnt_idmap) {
2477				ret = -EOPNOTSUPP;
2478				goto free_parent;
2479			}
2480
2481			subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2482						fs_info, vol_args2->subvolid);
2483			if (IS_ERR(subvol_name_ptr)) {
2484				ret = PTR_ERR(subvol_name_ptr);
2485				goto free_parent;
2486			}
2487			/* subvol_name_ptr is already nul terminated */
2488			subvol_name = (char *)kbasename(subvol_name_ptr);
2489		}
2490	} else {
2491		vol_args = memdup_user(arg, sizeof(*vol_args));
2492		if (IS_ERR(vol_args))
2493			return PTR_ERR(vol_args);
2494
2495		ret = btrfs_check_ioctl_vol_args_path(vol_args);
2496		if (ret < 0)
2497			goto out;
2498
2499		subvol_name = vol_args->name;
2500
2501		ret = mnt_want_write_file(file);
2502		if (ret)
2503			goto out;
2504	}
2505
2506	subvol_namelen = strlen(subvol_name);
2507
2508	if (strchr(subvol_name, '/') ||
2509	    strncmp(subvol_name, "..", subvol_namelen) == 0) {
2510		ret = -EINVAL;
2511		goto free_subvol_name;
2512	}
2513
2514	if (!S_ISDIR(dir->i_mode)) {
2515		ret = -ENOTDIR;
2516		goto free_subvol_name;
2517	}
2518
2519	ret = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2520	if (ret == -EINTR)
2521		goto free_subvol_name;
2522	dentry = lookup_one(idmap, subvol_name, parent, subvol_namelen);
2523	if (IS_ERR(dentry)) {
2524		ret = PTR_ERR(dentry);
2525		goto out_unlock_dir;
2526	}
2527
2528	if (d_really_is_negative(dentry)) {
2529		ret = -ENOENT;
2530		goto out_dput;
2531	}
2532
2533	inode = d_inode(dentry);
2534	dest = BTRFS_I(inode)->root;
2535	if (!capable(CAP_SYS_ADMIN)) {
2536		/*
2537		 * Regular user.  Only allow this with a special mount
2538		 * option, when the user has write+exec access to the
2539		 * subvol root, and when rmdir(2) would have been
2540		 * allowed.
2541		 *
2542		 * Note that this is _not_ check that the subvol is
2543		 * empty or doesn't contain data that we wouldn't
2544		 * otherwise be able to delete.
2545		 *
2546		 * Users who want to delete empty subvols should try
2547		 * rmdir(2).
2548		 */
2549		ret = -EPERM;
2550		if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2551			goto out_dput;
2552
2553		/*
2554		 * Do not allow deletion if the parent dir is the same
2555		 * as the dir to be deleted.  That means the ioctl
2556		 * must be called on the dentry referencing the root
2557		 * of the subvol, not a random directory contained
2558		 * within it.
2559		 */
2560		ret = -EINVAL;
2561		if (root == dest)
2562			goto out_dput;
2563
2564		ret = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2565		if (ret)
2566			goto out_dput;
2567	}
2568
2569	/* check if subvolume may be deleted by a user */
2570	ret = btrfs_may_delete(idmap, dir, dentry, 1);
2571	if (ret)
2572		goto out_dput;
2573
2574	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2575		ret = -EINVAL;
2576		goto out_dput;
2577	}
2578
2579	btrfs_inode_lock(BTRFS_I(inode), 0);
2580	ret = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2581	btrfs_inode_unlock(BTRFS_I(inode), 0);
2582	if (!ret)
2583		d_delete_notify(dir, dentry);
2584
2585out_dput:
2586	dput(dentry);
2587out_unlock_dir:
2588	btrfs_inode_unlock(BTRFS_I(dir), 0);
2589free_subvol_name:
2590	kfree(subvol_name_ptr);
2591free_parent:
2592	if (destroy_parent)
2593		dput(parent);
2594out_drop_write:
2595	mnt_drop_write_file(file);
2596out:
2597	kfree(vol_args2);
2598	kfree(vol_args);
2599	return ret;
2600}
2601
2602static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2603{
2604	struct inode *inode = file_inode(file);
2605	struct btrfs_root *root = BTRFS_I(inode)->root;
2606	struct btrfs_ioctl_defrag_range_args range = {0};
2607	int ret;
2608
2609	ret = mnt_want_write_file(file);
2610	if (ret)
2611		return ret;
2612
2613	if (btrfs_root_readonly(root)) {
2614		ret = -EROFS;
2615		goto out;
2616	}
2617
2618	switch (inode->i_mode & S_IFMT) {
2619	case S_IFDIR:
2620		if (!capable(CAP_SYS_ADMIN)) {
2621			ret = -EPERM;
2622			goto out;
2623		}
2624		ret = btrfs_defrag_root(root);
2625		break;
2626	case S_IFREG:
2627		/*
2628		 * Note that this does not check the file descriptor for write
2629		 * access. This prevents defragmenting executables that are
2630		 * running and allows defrag on files open in read-only mode.
2631		 */
2632		if (!capable(CAP_SYS_ADMIN) &&
2633		    inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2634			ret = -EPERM;
2635			goto out;
2636		}
2637
2638		if (argp) {
2639			if (copy_from_user(&range, argp, sizeof(range))) {
2640				ret = -EFAULT;
2641				goto out;
2642			}
2643			if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) {
2644				ret = -EOPNOTSUPP;
2645				goto out;
2646			}
2647			/* compression requires us to start the IO */
2648			if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2649				range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2650				range.extent_thresh = (u32)-1;
2651			}
2652		} else {
2653			/* the rest are all set to zero by kzalloc */
2654			range.len = (u64)-1;
2655		}
2656		ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2657					&range, BTRFS_OLDEST_GENERATION, 0);
2658		if (ret > 0)
2659			ret = 0;
2660		break;
2661	default:
2662		ret = -EINVAL;
2663	}
2664out:
2665	mnt_drop_write_file(file);
2666	return ret;
2667}
2668
2669static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2670{
2671	struct btrfs_ioctl_vol_args *vol_args;
2672	bool restore_op = false;
2673	int ret;
2674
2675	if (!capable(CAP_SYS_ADMIN))
2676		return -EPERM;
2677
2678	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2679		btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2680		return -EINVAL;
2681	}
2682
2683	if (fs_info->fs_devices->temp_fsid) {
2684		btrfs_err(fs_info,
2685			  "device add not supported on cloned temp-fsid mount");
2686		return -EINVAL;
2687	}
2688
2689	if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2690		if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2691			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2692
2693		/*
2694		 * We can do the device add because we have a paused balanced,
2695		 * change the exclusive op type and remember we should bring
2696		 * back the paused balance
2697		 */
2698		fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2699		btrfs_exclop_start_unlock(fs_info);
2700		restore_op = true;
2701	}
2702
2703	vol_args = memdup_user(arg, sizeof(*vol_args));
2704	if (IS_ERR(vol_args)) {
2705		ret = PTR_ERR(vol_args);
2706		goto out;
2707	}
2708
2709	ret = btrfs_check_ioctl_vol_args_path(vol_args);
2710	if (ret < 0)
2711		goto out_free;
2712
2713	ret = btrfs_init_new_device(fs_info, vol_args->name);
2714
2715	if (!ret)
2716		btrfs_info(fs_info, "disk added %s", vol_args->name);
2717
2718out_free:
2719	kfree(vol_args);
2720out:
2721	if (restore_op)
2722		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2723	else
2724		btrfs_exclop_finish(fs_info);
2725	return ret;
2726}
2727
2728static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2729{
2730	BTRFS_DEV_LOOKUP_ARGS(args);
2731	struct inode *inode = file_inode(file);
2732	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2733	struct btrfs_ioctl_vol_args_v2 *vol_args;
2734	struct file *bdev_file = NULL;
2735	int ret;
2736	bool cancel = false;
2737
2738	if (!capable(CAP_SYS_ADMIN))
2739		return -EPERM;
2740
2741	vol_args = memdup_user(arg, sizeof(*vol_args));
2742	if (IS_ERR(vol_args))
2743		return PTR_ERR(vol_args);
2744
2745	if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2746		ret = -EOPNOTSUPP;
2747		goto out;
2748	}
2749
2750	ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args);
2751	if (ret < 0)
2752		goto out;
2753
2754	if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2755		args.devid = vol_args->devid;
2756	} else if (!strcmp("cancel", vol_args->name)) {
2757		cancel = true;
2758	} else {
2759		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2760		if (ret)
2761			goto out;
2762	}
2763
2764	ret = mnt_want_write_file(file);
2765	if (ret)
2766		goto out;
2767
2768	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2769					   cancel);
2770	if (ret)
2771		goto err_drop;
2772
2773	/* Exclusive operation is now claimed */
2774	ret = btrfs_rm_device(fs_info, &args, &bdev_file);
2775
2776	btrfs_exclop_finish(fs_info);
2777
2778	if (!ret) {
2779		if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2780			btrfs_info(fs_info, "device deleted: id %llu",
2781					vol_args->devid);
2782		else
2783			btrfs_info(fs_info, "device deleted: %s",
2784					vol_args->name);
2785	}
2786err_drop:
2787	mnt_drop_write_file(file);
2788	if (bdev_file)
2789		fput(bdev_file);
2790out:
2791	btrfs_put_dev_args_from_path(&args);
2792	kfree(vol_args);
2793	return ret;
2794}
2795
2796static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2797{
2798	BTRFS_DEV_LOOKUP_ARGS(args);
2799	struct inode *inode = file_inode(file);
2800	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2801	struct btrfs_ioctl_vol_args *vol_args;
2802	struct file *bdev_file = NULL;
2803	int ret;
2804	bool cancel = false;
2805
2806	if (!capable(CAP_SYS_ADMIN))
2807		return -EPERM;
2808
2809	vol_args = memdup_user(arg, sizeof(*vol_args));
2810	if (IS_ERR(vol_args))
2811		return PTR_ERR(vol_args);
2812
2813	ret = btrfs_check_ioctl_vol_args_path(vol_args);
2814	if (ret < 0)
2815		goto out_free;
2816
2817	if (!strcmp("cancel", vol_args->name)) {
2818		cancel = true;
2819	} else {
2820		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2821		if (ret)
2822			goto out;
2823	}
2824
2825	ret = mnt_want_write_file(file);
2826	if (ret)
2827		goto out;
2828
2829	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2830					   cancel);
2831	if (ret == 0) {
2832		ret = btrfs_rm_device(fs_info, &args, &bdev_file);
2833		if (!ret)
2834			btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2835		btrfs_exclop_finish(fs_info);
2836	}
2837
2838	mnt_drop_write_file(file);
2839	if (bdev_file)
2840		fput(bdev_file);
2841out:
2842	btrfs_put_dev_args_from_path(&args);
2843out_free:
2844	kfree(vol_args);
2845	return ret;
2846}
2847
2848static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2849				void __user *arg)
2850{
2851	struct btrfs_ioctl_fs_info_args *fi_args;
2852	struct btrfs_device *device;
2853	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2854	u64 flags_in;
2855	int ret = 0;
2856
2857	fi_args = memdup_user(arg, sizeof(*fi_args));
2858	if (IS_ERR(fi_args))
2859		return PTR_ERR(fi_args);
2860
2861	flags_in = fi_args->flags;
2862	memset(fi_args, 0, sizeof(*fi_args));
2863
2864	rcu_read_lock();
2865	fi_args->num_devices = fs_devices->num_devices;
2866
2867	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2868		if (device->devid > fi_args->max_id)
2869			fi_args->max_id = device->devid;
2870	}
2871	rcu_read_unlock();
2872
2873	memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2874	fi_args->nodesize = fs_info->nodesize;
2875	fi_args->sectorsize = fs_info->sectorsize;
2876	fi_args->clone_alignment = fs_info->sectorsize;
2877
2878	if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2879		fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2880		fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2881		fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2882	}
2883
2884	if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2885		fi_args->generation = btrfs_get_fs_generation(fs_info);
2886		fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2887	}
2888
2889	if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2890		memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2891		       sizeof(fi_args->metadata_uuid));
2892		fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2893	}
2894
2895	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2896		ret = -EFAULT;
2897
2898	kfree(fi_args);
2899	return ret;
2900}
2901
2902static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2903				 void __user *arg)
2904{
2905	BTRFS_DEV_LOOKUP_ARGS(args);
2906	struct btrfs_ioctl_dev_info_args *di_args;
2907	struct btrfs_device *dev;
2908	int ret = 0;
2909
2910	di_args = memdup_user(arg, sizeof(*di_args));
2911	if (IS_ERR(di_args))
2912		return PTR_ERR(di_args);
2913
2914	args.devid = di_args->devid;
2915	if (!btrfs_is_empty_uuid(di_args->uuid))
2916		args.uuid = di_args->uuid;
2917
2918	rcu_read_lock();
2919	dev = btrfs_find_device(fs_info->fs_devices, &args);
2920	if (!dev) {
2921		ret = -ENODEV;
2922		goto out;
2923	}
2924
2925	di_args->devid = dev->devid;
2926	di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2927	di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2928	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2929	memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2930	if (dev->name)
2931		strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2932	else
2933		di_args->path[0] = '\0';
2934
2935out:
2936	rcu_read_unlock();
2937	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2938		ret = -EFAULT;
2939
2940	kfree(di_args);
2941	return ret;
2942}
2943
2944static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2945{
2946	struct inode *inode = file_inode(file);
2947	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2948	struct btrfs_root *root = BTRFS_I(inode)->root;
2949	struct btrfs_root *new_root;
2950	struct btrfs_dir_item *di;
2951	struct btrfs_trans_handle *trans;
2952	struct btrfs_path *path = NULL;
2953	struct btrfs_disk_key disk_key;
2954	struct fscrypt_str name = FSTR_INIT("default", 7);
2955	u64 objectid = 0;
2956	u64 dir_id;
2957	int ret;
2958
2959	if (!capable(CAP_SYS_ADMIN))
2960		return -EPERM;
2961
2962	ret = mnt_want_write_file(file);
2963	if (ret)
2964		return ret;
2965
2966	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2967		ret = -EFAULT;
2968		goto out;
2969	}
2970
2971	if (!objectid)
2972		objectid = BTRFS_FS_TREE_OBJECTID;
2973
2974	new_root = btrfs_get_fs_root(fs_info, objectid, true);
2975	if (IS_ERR(new_root)) {
2976		ret = PTR_ERR(new_root);
2977		goto out;
2978	}
2979	if (!is_fstree(btrfs_root_id(new_root))) {
2980		ret = -ENOENT;
2981		goto out_free;
2982	}
2983
2984	path = btrfs_alloc_path();
2985	if (!path) {
2986		ret = -ENOMEM;
2987		goto out_free;
2988	}
2989
2990	trans = btrfs_start_transaction(root, 1);
2991	if (IS_ERR(trans)) {
2992		ret = PTR_ERR(trans);
2993		goto out_free;
2994	}
2995
2996	dir_id = btrfs_super_root_dir(fs_info->super_copy);
2997	di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2998				   dir_id, &name, 1);
2999	if (IS_ERR_OR_NULL(di)) {
3000		btrfs_release_path(path);
3001		btrfs_end_transaction(trans);
3002		btrfs_err(fs_info,
3003			  "Umm, you don't have the default diritem, this isn't going to work");
3004		ret = -ENOENT;
3005		goto out_free;
3006	}
3007
3008	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3009	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3010	btrfs_mark_buffer_dirty(trans, path->nodes[0]);
3011	btrfs_release_path(path);
3012
3013	btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3014	btrfs_end_transaction(trans);
3015out_free:
3016	btrfs_put_root(new_root);
3017	btrfs_free_path(path);
3018out:
3019	mnt_drop_write_file(file);
3020	return ret;
3021}
3022
3023static void get_block_group_info(struct list_head *groups_list,
3024				 struct btrfs_ioctl_space_info *space)
3025{
3026	struct btrfs_block_group *block_group;
3027
3028	space->total_bytes = 0;
3029	space->used_bytes = 0;
3030	space->flags = 0;
3031	list_for_each_entry(block_group, groups_list, list) {
3032		space->flags = block_group->flags;
3033		space->total_bytes += block_group->length;
3034		space->used_bytes += block_group->used;
3035	}
3036}
3037
3038static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3039				   void __user *arg)
3040{
3041	struct btrfs_ioctl_space_args space_args = { 0 };
3042	struct btrfs_ioctl_space_info space;
3043	struct btrfs_ioctl_space_info *dest;
3044	struct btrfs_ioctl_space_info *dest_orig;
3045	struct btrfs_ioctl_space_info __user *user_dest;
3046	struct btrfs_space_info *info;
3047	static const u64 types[] = {
3048		BTRFS_BLOCK_GROUP_DATA,
3049		BTRFS_BLOCK_GROUP_SYSTEM,
3050		BTRFS_BLOCK_GROUP_METADATA,
3051		BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3052	};
3053	int num_types = 4;
3054	int alloc_size;
3055	int ret = 0;
3056	u64 slot_count = 0;
3057	int i, c;
3058
3059	if (copy_from_user(&space_args,
3060			   (struct btrfs_ioctl_space_args __user *)arg,
3061			   sizeof(space_args)))
3062		return -EFAULT;
3063
3064	for (i = 0; i < num_types; i++) {
3065		struct btrfs_space_info *tmp;
3066
3067		info = NULL;
3068		list_for_each_entry(tmp, &fs_info->space_info, list) {
3069			if (tmp->flags == types[i]) {
3070				info = tmp;
3071				break;
3072			}
3073		}
3074
3075		if (!info)
3076			continue;
3077
3078		down_read(&info->groups_sem);
3079		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3080			if (!list_empty(&info->block_groups[c]))
3081				slot_count++;
3082		}
3083		up_read(&info->groups_sem);
3084	}
3085
3086	/*
3087	 * Global block reserve, exported as a space_info
3088	 */
3089	slot_count++;
3090
3091	/* space_slots == 0 means they are asking for a count */
3092	if (space_args.space_slots == 0) {
3093		space_args.total_spaces = slot_count;
3094		goto out;
3095	}
3096
3097	slot_count = min_t(u64, space_args.space_slots, slot_count);
3098
3099	alloc_size = sizeof(*dest) * slot_count;
3100
3101	/* we generally have at most 6 or so space infos, one for each raid
3102	 * level.  So, a whole page should be more than enough for everyone
3103	 */
3104	if (alloc_size > PAGE_SIZE)
3105		return -ENOMEM;
3106
3107	space_args.total_spaces = 0;
3108	dest = kmalloc(alloc_size, GFP_KERNEL);
3109	if (!dest)
3110		return -ENOMEM;
3111	dest_orig = dest;
3112
3113	/* now we have a buffer to copy into */
3114	for (i = 0; i < num_types; i++) {
3115		struct btrfs_space_info *tmp;
3116
3117		if (!slot_count)
3118			break;
3119
3120		info = NULL;
3121		list_for_each_entry(tmp, &fs_info->space_info, list) {
3122			if (tmp->flags == types[i]) {
3123				info = tmp;
3124				break;
3125			}
3126		}
3127
3128		if (!info)
3129			continue;
3130		down_read(&info->groups_sem);
3131		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3132			if (!list_empty(&info->block_groups[c])) {
3133				get_block_group_info(&info->block_groups[c],
3134						     &space);
3135				memcpy(dest, &space, sizeof(space));
3136				dest++;
3137				space_args.total_spaces++;
3138				slot_count--;
3139			}
3140			if (!slot_count)
3141				break;
3142		}
3143		up_read(&info->groups_sem);
3144	}
3145
3146	/*
3147	 * Add global block reserve
3148	 */
3149	if (slot_count) {
3150		struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3151
3152		spin_lock(&block_rsv->lock);
3153		space.total_bytes = block_rsv->size;
3154		space.used_bytes = block_rsv->size - block_rsv->reserved;
3155		spin_unlock(&block_rsv->lock);
3156		space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3157		memcpy(dest, &space, sizeof(space));
3158		space_args.total_spaces++;
3159	}
3160
3161	user_dest = (struct btrfs_ioctl_space_info __user *)
3162		(arg + sizeof(struct btrfs_ioctl_space_args));
3163
3164	if (copy_to_user(user_dest, dest_orig, alloc_size))
3165		ret = -EFAULT;
3166
3167	kfree(dest_orig);
3168out:
3169	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3170		ret = -EFAULT;
3171
3172	return ret;
3173}
3174
3175static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3176					    void __user *argp)
3177{
3178	struct btrfs_trans_handle *trans;
3179	u64 transid;
3180
3181	/*
3182	 * Start orphan cleanup here for the given root in case it hasn't been
3183	 * started already by other means. Errors are handled in the other
3184	 * functions during transaction commit.
3185	 */
3186	btrfs_orphan_cleanup(root);
3187
3188	trans = btrfs_attach_transaction_barrier(root);
3189	if (IS_ERR(trans)) {
3190		if (PTR_ERR(trans) != -ENOENT)
3191			return PTR_ERR(trans);
3192
3193		/* No running transaction, don't bother */
3194		transid = btrfs_get_last_trans_committed(root->fs_info);
3195		goto out;
3196	}
3197	transid = trans->transid;
3198	btrfs_commit_transaction_async(trans);
3199out:
3200	if (argp)
3201		if (copy_to_user(argp, &transid, sizeof(transid)))
3202			return -EFAULT;
3203	return 0;
3204}
3205
3206static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3207					   void __user *argp)
3208{
3209	/* By default wait for the current transaction. */
3210	u64 transid = 0;
3211
3212	if (argp)
3213		if (copy_from_user(&transid, argp, sizeof(transid)))
3214			return -EFAULT;
3215
3216	return btrfs_wait_for_commit(fs_info, transid);
3217}
3218
3219static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3220{
3221	struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file));
3222	struct btrfs_ioctl_scrub_args *sa;
3223	int ret;
3224
3225	if (!capable(CAP_SYS_ADMIN))
3226		return -EPERM;
3227
3228	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3229		btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3230		return -EINVAL;
3231	}
3232
3233	sa = memdup_user(arg, sizeof(*sa));
3234	if (IS_ERR(sa))
3235		return PTR_ERR(sa);
3236
3237	if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3238		ret = -EOPNOTSUPP;
3239		goto out;
3240	}
3241
3242	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3243		ret = mnt_want_write_file(file);
3244		if (ret)
3245			goto out;
3246	}
3247
3248	ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3249			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3250			      0);
3251
3252	/*
3253	 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3254	 * error. This is important as it allows user space to know how much
3255	 * progress scrub has done. For example, if scrub is canceled we get
3256	 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3257	 * space. Later user space can inspect the progress from the structure
3258	 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3259	 * previously (btrfs-progs does this).
3260	 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3261	 * then return -EFAULT to signal the structure was not copied or it may
3262	 * be corrupt and unreliable due to a partial copy.
3263	 */
3264	if (copy_to_user(arg, sa, sizeof(*sa)))
3265		ret = -EFAULT;
3266
3267	if (!(sa->flags & BTRFS_SCRUB_READONLY))
3268		mnt_drop_write_file(file);
3269out:
3270	kfree(sa);
3271	return ret;
3272}
3273
3274static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3275{
3276	if (!capable(CAP_SYS_ADMIN))
3277		return -EPERM;
3278
3279	return btrfs_scrub_cancel(fs_info);
3280}
3281
3282static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3283				       void __user *arg)
3284{
3285	struct btrfs_ioctl_scrub_args *sa;
3286	int ret;
3287
3288	if (!capable(CAP_SYS_ADMIN))
3289		return -EPERM;
3290
3291	sa = memdup_user(arg, sizeof(*sa));
3292	if (IS_ERR(sa))
3293		return PTR_ERR(sa);
3294
3295	ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3296
3297	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3298		ret = -EFAULT;
3299
3300	kfree(sa);
3301	return ret;
3302}
3303
3304static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3305				      void __user *arg)
3306{
3307	struct btrfs_ioctl_get_dev_stats *sa;
3308	int ret;
3309
3310	sa = memdup_user(arg, sizeof(*sa));
3311	if (IS_ERR(sa))
3312		return PTR_ERR(sa);
3313
3314	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3315		kfree(sa);
3316		return -EPERM;
3317	}
3318
3319	ret = btrfs_get_dev_stats(fs_info, sa);
3320
3321	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3322		ret = -EFAULT;
3323
3324	kfree(sa);
3325	return ret;
3326}
3327
3328static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3329				    void __user *arg)
3330{
3331	struct btrfs_ioctl_dev_replace_args *p;
3332	int ret;
3333
3334	if (!capable(CAP_SYS_ADMIN))
3335		return -EPERM;
3336
3337	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3338		btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3339		return -EINVAL;
3340	}
3341
3342	p = memdup_user(arg, sizeof(*p));
3343	if (IS_ERR(p))
3344		return PTR_ERR(p);
3345
3346	switch (p->cmd) {
3347	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3348		if (sb_rdonly(fs_info->sb)) {
3349			ret = -EROFS;
3350			goto out;
3351		}
3352		if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3353			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3354		} else {
3355			ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3356			btrfs_exclop_finish(fs_info);
3357		}
3358		break;
3359	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3360		btrfs_dev_replace_status(fs_info, p);
3361		ret = 0;
3362		break;
3363	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3364		p->result = btrfs_dev_replace_cancel(fs_info);
3365		ret = 0;
3366		break;
3367	default:
3368		ret = -EINVAL;
3369		break;
3370	}
3371
3372	if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3373		ret = -EFAULT;
3374out:
3375	kfree(p);
3376	return ret;
3377}
3378
3379static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3380{
3381	int ret = 0;
3382	int i;
3383	u64 rel_ptr;
3384	int size;
3385	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3386	struct inode_fs_paths *ipath = NULL;
3387	struct btrfs_path *path;
3388
3389	if (!capable(CAP_DAC_READ_SEARCH))
3390		return -EPERM;
3391
3392	path = btrfs_alloc_path();
3393	if (!path) {
3394		ret = -ENOMEM;
3395		goto out;
3396	}
3397
3398	ipa = memdup_user(arg, sizeof(*ipa));
3399	if (IS_ERR(ipa)) {
3400		ret = PTR_ERR(ipa);
3401		ipa = NULL;
3402		goto out;
3403	}
3404
3405	size = min_t(u32, ipa->size, 4096);
3406	ipath = init_ipath(size, root, path);
3407	if (IS_ERR(ipath)) {
3408		ret = PTR_ERR(ipath);
3409		ipath = NULL;
3410		goto out;
3411	}
3412
3413	ret = paths_from_inode(ipa->inum, ipath);
3414	if (ret < 0)
3415		goto out;
3416
3417	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3418		rel_ptr = ipath->fspath->val[i] -
3419			  (u64)(unsigned long)ipath->fspath->val;
3420		ipath->fspath->val[i] = rel_ptr;
3421	}
3422
3423	btrfs_free_path(path);
3424	path = NULL;
3425	ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3426			   ipath->fspath, size);
3427	if (ret) {
3428		ret = -EFAULT;
3429		goto out;
3430	}
3431
3432out:
3433	btrfs_free_path(path);
3434	free_ipath(ipath);
3435	kfree(ipa);
3436
3437	return ret;
3438}
3439
3440static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3441					void __user *arg, int version)
3442{
3443	int ret = 0;
3444	int size;
3445	struct btrfs_ioctl_logical_ino_args *loi;
3446	struct btrfs_data_container *inodes = NULL;
3447	struct btrfs_path *path = NULL;
3448	bool ignore_offset;
3449
3450	if (!capable(CAP_SYS_ADMIN))
3451		return -EPERM;
3452
3453	loi = memdup_user(arg, sizeof(*loi));
3454	if (IS_ERR(loi))
3455		return PTR_ERR(loi);
3456
3457	if (version == 1) {
3458		ignore_offset = false;
3459		size = min_t(u32, loi->size, SZ_64K);
3460	} else {
3461		/* All reserved bits must be 0 for now */
3462		if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3463			ret = -EINVAL;
3464			goto out_loi;
3465		}
3466		/* Only accept flags we have defined so far */
3467		if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3468			ret = -EINVAL;
3469			goto out_loi;
3470		}
3471		ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3472		size = min_t(u32, loi->size, SZ_16M);
3473	}
3474
3475	inodes = init_data_container(size);
3476	if (IS_ERR(inodes)) {
3477		ret = PTR_ERR(inodes);
3478		goto out_loi;
3479	}
3480
3481	path = btrfs_alloc_path();
3482	if (!path) {
3483		ret = -ENOMEM;
3484		goto out;
3485	}
3486	ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3487					  inodes, ignore_offset);
3488	btrfs_free_path(path);
3489	if (ret == -EINVAL)
3490		ret = -ENOENT;
3491	if (ret < 0)
3492		goto out;
3493
3494	ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3495			   size);
3496	if (ret)
3497		ret = -EFAULT;
3498
3499out:
3500	kvfree(inodes);
3501out_loi:
3502	kfree(loi);
3503
3504	return ret;
3505}
3506
3507void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3508			       struct btrfs_ioctl_balance_args *bargs)
3509{
3510	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3511
3512	bargs->flags = bctl->flags;
3513
3514	if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3515		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3516	if (atomic_read(&fs_info->balance_pause_req))
3517		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3518	if (atomic_read(&fs_info->balance_cancel_req))
3519		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3520
3521	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3522	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3523	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3524
3525	spin_lock(&fs_info->balance_lock);
3526	memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3527	spin_unlock(&fs_info->balance_lock);
3528}
3529
3530/*
3531 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3532 * required.
3533 *
3534 * @fs_info:       the filesystem
3535 * @excl_acquired: ptr to boolean value which is set to false in case balance
3536 *                 is being resumed
3537 *
3538 * Return 0 on success in which case both fs_info::balance is acquired as well
3539 * as exclusive ops are blocked. In case of failure return an error code.
3540 */
3541static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3542{
3543	int ret;
3544
3545	/*
3546	 * Exclusive operation is locked. Three possibilities:
3547	 *   (1) some other op is running
3548	 *   (2) balance is running
3549	 *   (3) balance is paused -- special case (think resume)
3550	 */
3551	while (1) {
3552		if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3553			*excl_acquired = true;
3554			mutex_lock(&fs_info->balance_mutex);
3555			return 0;
3556		}
3557
3558		mutex_lock(&fs_info->balance_mutex);
3559		if (fs_info->balance_ctl) {
3560			/* This is either (2) or (3) */
3561			if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3562				/* This is (2) */
3563				ret = -EINPROGRESS;
3564				goto out_failure;
3565
3566			} else {
3567				mutex_unlock(&fs_info->balance_mutex);
3568				/*
3569				 * Lock released to allow other waiters to
3570				 * continue, we'll reexamine the status again.
3571				 */
3572				mutex_lock(&fs_info->balance_mutex);
3573
3574				if (fs_info->balance_ctl &&
3575				    !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3576					/* This is (3) */
3577					*excl_acquired = false;
3578					return 0;
3579				}
3580			}
3581		} else {
3582			/* This is (1) */
3583			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3584			goto out_failure;
3585		}
3586
3587		mutex_unlock(&fs_info->balance_mutex);
3588	}
3589
3590out_failure:
3591	mutex_unlock(&fs_info->balance_mutex);
3592	*excl_acquired = false;
3593	return ret;
3594}
3595
3596static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3597{
3598	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3599	struct btrfs_fs_info *fs_info = root->fs_info;
3600	struct btrfs_ioctl_balance_args *bargs;
3601	struct btrfs_balance_control *bctl;
3602	bool need_unlock = true;
3603	int ret;
3604
3605	if (!capable(CAP_SYS_ADMIN))
3606		return -EPERM;
3607
3608	ret = mnt_want_write_file(file);
3609	if (ret)
3610		return ret;
3611
3612	bargs = memdup_user(arg, sizeof(*bargs));
3613	if (IS_ERR(bargs)) {
3614		ret = PTR_ERR(bargs);
3615		bargs = NULL;
3616		goto out;
3617	}
3618
3619	ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3620	if (ret)
3621		goto out;
3622
3623	lockdep_assert_held(&fs_info->balance_mutex);
3624
3625	if (bargs->flags & BTRFS_BALANCE_RESUME) {
3626		if (!fs_info->balance_ctl) {
3627			ret = -ENOTCONN;
3628			goto out_unlock;
3629		}
3630
3631		bctl = fs_info->balance_ctl;
3632		spin_lock(&fs_info->balance_lock);
3633		bctl->flags |= BTRFS_BALANCE_RESUME;
3634		spin_unlock(&fs_info->balance_lock);
3635		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3636
3637		goto do_balance;
3638	}
3639
3640	if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3641		ret = -EINVAL;
3642		goto out_unlock;
3643	}
3644
3645	if (fs_info->balance_ctl) {
3646		ret = -EINPROGRESS;
3647		goto out_unlock;
3648	}
3649
3650	bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3651	if (!bctl) {
3652		ret = -ENOMEM;
3653		goto out_unlock;
3654	}
3655
3656	memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3657	memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3658	memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3659
3660	bctl->flags = bargs->flags;
3661do_balance:
3662	/*
3663	 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3664	 * bctl is freed in reset_balance_state, or, if restriper was paused
3665	 * all the way until unmount, in free_fs_info.  The flag should be
3666	 * cleared after reset_balance_state.
3667	 */
3668	need_unlock = false;
3669
3670	ret = btrfs_balance(fs_info, bctl, bargs);
3671	bctl = NULL;
3672
3673	if (ret == 0 || ret == -ECANCELED) {
3674		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3675			ret = -EFAULT;
3676	}
3677
3678	kfree(bctl);
3679out_unlock:
3680	mutex_unlock(&fs_info->balance_mutex);
3681	if (need_unlock)
3682		btrfs_exclop_finish(fs_info);
3683out:
3684	mnt_drop_write_file(file);
3685	kfree(bargs);
3686	return ret;
3687}
3688
3689static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3690{
3691	if (!capable(CAP_SYS_ADMIN))
3692		return -EPERM;
3693
3694	switch (cmd) {
3695	case BTRFS_BALANCE_CTL_PAUSE:
3696		return btrfs_pause_balance(fs_info);
3697	case BTRFS_BALANCE_CTL_CANCEL:
3698		return btrfs_cancel_balance(fs_info);
3699	}
3700
3701	return -EINVAL;
3702}
3703
3704static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3705					 void __user *arg)
3706{
3707	struct btrfs_ioctl_balance_args *bargs;
3708	int ret = 0;
3709
3710	if (!capable(CAP_SYS_ADMIN))
3711		return -EPERM;
3712
3713	mutex_lock(&fs_info->balance_mutex);
3714	if (!fs_info->balance_ctl) {
3715		ret = -ENOTCONN;
3716		goto out;
3717	}
3718
3719	bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3720	if (!bargs) {
3721		ret = -ENOMEM;
3722		goto out;
3723	}
3724
3725	btrfs_update_ioctl_balance_args(fs_info, bargs);
3726
3727	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3728		ret = -EFAULT;
3729
3730	kfree(bargs);
3731out:
3732	mutex_unlock(&fs_info->balance_mutex);
3733	return ret;
3734}
3735
3736static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3737{
3738	struct inode *inode = file_inode(file);
3739	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3740	struct btrfs_ioctl_quota_ctl_args *sa;
3741	int ret;
3742
3743	if (!capable(CAP_SYS_ADMIN))
3744		return -EPERM;
3745
3746	ret = mnt_want_write_file(file);
3747	if (ret)
3748		return ret;
3749
3750	sa = memdup_user(arg, sizeof(*sa));
3751	if (IS_ERR(sa)) {
3752		ret = PTR_ERR(sa);
3753		goto drop_write;
3754	}
3755
3756	switch (sa->cmd) {
3757	case BTRFS_QUOTA_CTL_ENABLE:
3758	case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA:
3759		down_write(&fs_info->subvol_sem);
3760		ret = btrfs_quota_enable(fs_info, sa);
3761		up_write(&fs_info->subvol_sem);
3762		break;
3763	case BTRFS_QUOTA_CTL_DISABLE:
3764		/*
3765		 * Lock the cleaner mutex to prevent races with concurrent
3766		 * relocation, because relocation may be building backrefs for
3767		 * blocks of the quota root while we are deleting the root. This
3768		 * is like dropping fs roots of deleted snapshots/subvolumes, we
3769		 * need the same protection.
3770		 *
3771		 * This also prevents races between concurrent tasks trying to
3772		 * disable quotas, because we will unlock and relock
3773		 * qgroup_ioctl_lock across BTRFS_FS_QUOTA_ENABLED changes.
3774		 *
3775		 * We take this here because we have the dependency of
3776		 *
3777		 * inode_lock -> subvol_sem
3778		 *
3779		 * because of rename.  With relocation we can prealloc extents,
3780		 * so that makes the dependency chain
3781		 *
3782		 * cleaner_mutex -> inode_lock -> subvol_sem
3783		 *
3784		 * so we must take the cleaner_mutex here before we take the
3785		 * subvol_sem.  The deadlock can't actually happen, but this
3786		 * quiets lockdep.
3787		 */
3788		mutex_lock(&fs_info->cleaner_mutex);
3789		down_write(&fs_info->subvol_sem);
3790		ret = btrfs_quota_disable(fs_info);
3791		up_write(&fs_info->subvol_sem);
3792		mutex_unlock(&fs_info->cleaner_mutex);
3793		break;
3794	default:
3795		ret = -EINVAL;
3796		break;
3797	}
3798
3799	kfree(sa);
3800drop_write:
3801	mnt_drop_write_file(file);
3802	return ret;
3803}
3804
3805/*
3806 * Quick check for ioctl handlers if quotas are enabled. Proper locking must be
3807 * done before any operations.
3808 */
3809static bool qgroup_enabled(struct btrfs_fs_info *fs_info)
3810{
3811	bool ret = true;
3812
3813	mutex_lock(&fs_info->qgroup_ioctl_lock);
3814	if (!fs_info->quota_root)
3815		ret = false;
3816	mutex_unlock(&fs_info->qgroup_ioctl_lock);
3817
3818	return ret;
3819}
3820
3821static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3822{
3823	struct inode *inode = file_inode(file);
3824	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3825	struct btrfs_root *root = BTRFS_I(inode)->root;
3826	struct btrfs_ioctl_qgroup_assign_args *sa;
3827	struct btrfs_qgroup_list *prealloc = NULL;
3828	struct btrfs_trans_handle *trans;
3829	int ret;
3830	int err;
3831
3832	if (!capable(CAP_SYS_ADMIN))
3833		return -EPERM;
3834
3835	if (!qgroup_enabled(root->fs_info))
3836		return -ENOTCONN;
3837
3838	ret = mnt_want_write_file(file);
3839	if (ret)
3840		return ret;
3841
3842	sa = memdup_user(arg, sizeof(*sa));
3843	if (IS_ERR(sa)) {
3844		ret = PTR_ERR(sa);
3845		goto drop_write;
3846	}
3847
3848	if (sa->assign) {
3849		prealloc = kzalloc(sizeof(*prealloc), GFP_KERNEL);
3850		if (!prealloc) {
3851			ret = -ENOMEM;
3852			goto drop_write;
3853		}
3854	}
3855
3856	trans = btrfs_join_transaction(root);
3857	if (IS_ERR(trans)) {
3858		ret = PTR_ERR(trans);
3859		goto out;
3860	}
3861
3862	/*
3863	 * Prealloc ownership is moved to the relation handler, there it's used
3864	 * or freed on error.
3865	 */
3866	if (sa->assign) {
3867		ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst, prealloc);
3868		prealloc = NULL;
3869	} else {
3870		ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3871	}
3872
3873	/* update qgroup status and info */
3874	mutex_lock(&fs_info->qgroup_ioctl_lock);
3875	err = btrfs_run_qgroups(trans);
3876	mutex_unlock(&fs_info->qgroup_ioctl_lock);
3877	if (err < 0)
3878		btrfs_warn(fs_info,
3879			   "qgroup status update failed after %s relation, marked as inconsistent",
3880			   sa->assign ? "adding" : "deleting");
3881	err = btrfs_end_transaction(trans);
3882	if (err && !ret)
3883		ret = err;
3884
3885out:
3886	kfree(prealloc);
3887	kfree(sa);
3888drop_write:
3889	mnt_drop_write_file(file);
3890	return ret;
3891}
3892
3893static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3894{
3895	struct inode *inode = file_inode(file);
3896	struct btrfs_root *root = BTRFS_I(inode)->root;
3897	struct btrfs_ioctl_qgroup_create_args *sa;
3898	struct btrfs_trans_handle *trans;
3899	int ret;
3900	int err;
3901
3902	if (!capable(CAP_SYS_ADMIN))
3903		return -EPERM;
3904
3905	if (!qgroup_enabled(root->fs_info))
3906		return -ENOTCONN;
3907
3908	ret = mnt_want_write_file(file);
3909	if (ret)
3910		return ret;
3911
3912	sa = memdup_user(arg, sizeof(*sa));
3913	if (IS_ERR(sa)) {
3914		ret = PTR_ERR(sa);
3915		goto drop_write;
3916	}
3917
3918	if (!sa->qgroupid) {
3919		ret = -EINVAL;
3920		goto out;
3921	}
3922
3923	if (sa->create && is_fstree(sa->qgroupid)) {
3924		ret = -EINVAL;
3925		goto out;
3926	}
3927
3928	trans = btrfs_join_transaction(root);
3929	if (IS_ERR(trans)) {
3930		ret = PTR_ERR(trans);
3931		goto out;
3932	}
3933
3934	if (sa->create) {
3935		ret = btrfs_create_qgroup(trans, sa->qgroupid);
3936	} else {
3937		ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3938	}
3939
3940	err = btrfs_end_transaction(trans);
3941	if (err && !ret)
3942		ret = err;
3943
3944out:
3945	kfree(sa);
3946drop_write:
3947	mnt_drop_write_file(file);
3948	return ret;
3949}
3950
3951static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3952{
3953	struct inode *inode = file_inode(file);
3954	struct btrfs_root *root = BTRFS_I(inode)->root;
3955	struct btrfs_ioctl_qgroup_limit_args *sa;
3956	struct btrfs_trans_handle *trans;
3957	int ret;
3958	int err;
3959	u64 qgroupid;
3960
3961	if (!capable(CAP_SYS_ADMIN))
3962		return -EPERM;
3963
3964	if (!qgroup_enabled(root->fs_info))
3965		return -ENOTCONN;
3966
3967	ret = mnt_want_write_file(file);
3968	if (ret)
3969		return ret;
3970
3971	sa = memdup_user(arg, sizeof(*sa));
3972	if (IS_ERR(sa)) {
3973		ret = PTR_ERR(sa);
3974		goto drop_write;
3975	}
3976
3977	trans = btrfs_join_transaction(root);
3978	if (IS_ERR(trans)) {
3979		ret = PTR_ERR(trans);
3980		goto out;
3981	}
3982
3983	qgroupid = sa->qgroupid;
3984	if (!qgroupid) {
3985		/* take the current subvol as qgroup */
3986		qgroupid = btrfs_root_id(root);
3987	}
3988
3989	ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3990
3991	err = btrfs_end_transaction(trans);
3992	if (err && !ret)
3993		ret = err;
3994
3995out:
3996	kfree(sa);
3997drop_write:
3998	mnt_drop_write_file(file);
3999	return ret;
4000}
4001
4002static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4003{
4004	struct inode *inode = file_inode(file);
4005	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4006	struct btrfs_ioctl_quota_rescan_args *qsa;
4007	int ret;
4008
4009	if (!capable(CAP_SYS_ADMIN))
4010		return -EPERM;
4011
4012	if (!qgroup_enabled(fs_info))
4013		return -ENOTCONN;
4014
4015	ret = mnt_want_write_file(file);
4016	if (ret)
4017		return ret;
4018
4019	qsa = memdup_user(arg, sizeof(*qsa));
4020	if (IS_ERR(qsa)) {
4021		ret = PTR_ERR(qsa);
4022		goto drop_write;
4023	}
4024
4025	if (qsa->flags) {
4026		ret = -EINVAL;
4027		goto out;
4028	}
4029
4030	ret = btrfs_qgroup_rescan(fs_info);
4031
4032out:
4033	kfree(qsa);
4034drop_write:
4035	mnt_drop_write_file(file);
4036	return ret;
4037}
4038
4039static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4040						void __user *arg)
4041{
4042	struct btrfs_ioctl_quota_rescan_args qsa = {0};
4043
4044	if (!capable(CAP_SYS_ADMIN))
4045		return -EPERM;
4046
4047	if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4048		qsa.flags = 1;
4049		qsa.progress = fs_info->qgroup_rescan_progress.objectid;
4050	}
4051
4052	if (copy_to_user(arg, &qsa, sizeof(qsa)))
4053		return -EFAULT;
4054
4055	return 0;
4056}
4057
4058static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info)
 
4059{
4060	if (!capable(CAP_SYS_ADMIN))
4061		return -EPERM;
4062
4063	return btrfs_qgroup_wait_for_completion(fs_info, true);
4064}
4065
4066static long _btrfs_ioctl_set_received_subvol(struct file *file,
4067					    struct mnt_idmap *idmap,
4068					    struct btrfs_ioctl_received_subvol_args *sa)
4069{
4070	struct inode *inode = file_inode(file);
4071	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4072	struct btrfs_root *root = BTRFS_I(inode)->root;
4073	struct btrfs_root_item *root_item = &root->root_item;
4074	struct btrfs_trans_handle *trans;
4075	struct timespec64 ct = current_time(inode);
4076	int ret = 0;
4077	int received_uuid_changed;
4078
4079	if (!inode_owner_or_capable(idmap, inode))
4080		return -EPERM;
4081
4082	ret = mnt_want_write_file(file);
4083	if (ret < 0)
4084		return ret;
4085
4086	down_write(&fs_info->subvol_sem);
4087
4088	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4089		ret = -EINVAL;
4090		goto out;
4091	}
4092
4093	if (btrfs_root_readonly(root)) {
4094		ret = -EROFS;
4095		goto out;
4096	}
4097
4098	/*
4099	 * 1 - root item
4100	 * 2 - uuid items (received uuid + subvol uuid)
4101	 */
4102	trans = btrfs_start_transaction(root, 3);
4103	if (IS_ERR(trans)) {
4104		ret = PTR_ERR(trans);
4105		trans = NULL;
4106		goto out;
4107	}
4108
4109	sa->rtransid = trans->transid;
4110	sa->rtime.sec = ct.tv_sec;
4111	sa->rtime.nsec = ct.tv_nsec;
4112
4113	received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4114				       BTRFS_UUID_SIZE);
4115	if (received_uuid_changed &&
4116	    !btrfs_is_empty_uuid(root_item->received_uuid)) {
4117		ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4118					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4119					  btrfs_root_id(root));
4120		if (ret && ret != -ENOENT) {
4121		        btrfs_abort_transaction(trans, ret);
4122		        btrfs_end_transaction(trans);
4123		        goto out;
4124		}
4125	}
4126	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4127	btrfs_set_root_stransid(root_item, sa->stransid);
4128	btrfs_set_root_rtransid(root_item, sa->rtransid);
4129	btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4130	btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4131	btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4132	btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4133
4134	ret = btrfs_update_root(trans, fs_info->tree_root,
4135				&root->root_key, &root->root_item);
4136	if (ret < 0) {
4137		btrfs_end_transaction(trans);
4138		goto out;
4139	}
4140	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4141		ret = btrfs_uuid_tree_add(trans, sa->uuid,
4142					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4143					  btrfs_root_id(root));
4144		if (ret < 0 && ret != -EEXIST) {
4145			btrfs_abort_transaction(trans, ret);
4146			btrfs_end_transaction(trans);
4147			goto out;
4148		}
4149	}
4150	ret = btrfs_commit_transaction(trans);
4151out:
4152	up_write(&fs_info->subvol_sem);
4153	mnt_drop_write_file(file);
4154	return ret;
4155}
4156
4157#ifdef CONFIG_64BIT
4158static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4159						void __user *arg)
4160{
4161	struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4162	struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4163	int ret = 0;
4164
4165	args32 = memdup_user(arg, sizeof(*args32));
4166	if (IS_ERR(args32))
4167		return PTR_ERR(args32);
4168
4169	args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4170	if (!args64) {
4171		ret = -ENOMEM;
4172		goto out;
4173	}
4174
4175	memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4176	args64->stransid = args32->stransid;
4177	args64->rtransid = args32->rtransid;
4178	args64->stime.sec = args32->stime.sec;
4179	args64->stime.nsec = args32->stime.nsec;
4180	args64->rtime.sec = args32->rtime.sec;
4181	args64->rtime.nsec = args32->rtime.nsec;
4182	args64->flags = args32->flags;
4183
4184	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4185	if (ret)
4186		goto out;
4187
4188	memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4189	args32->stransid = args64->stransid;
4190	args32->rtransid = args64->rtransid;
4191	args32->stime.sec = args64->stime.sec;
4192	args32->stime.nsec = args64->stime.nsec;
4193	args32->rtime.sec = args64->rtime.sec;
4194	args32->rtime.nsec = args64->rtime.nsec;
4195	args32->flags = args64->flags;
4196
4197	ret = copy_to_user(arg, args32, sizeof(*args32));
4198	if (ret)
4199		ret = -EFAULT;
4200
4201out:
4202	kfree(args32);
4203	kfree(args64);
4204	return ret;
4205}
4206#endif
4207
4208static long btrfs_ioctl_set_received_subvol(struct file *file,
4209					    void __user *arg)
4210{
4211	struct btrfs_ioctl_received_subvol_args *sa = NULL;
4212	int ret = 0;
4213
4214	sa = memdup_user(arg, sizeof(*sa));
4215	if (IS_ERR(sa))
4216		return PTR_ERR(sa);
4217
4218	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4219
4220	if (ret)
4221		goto out;
4222
4223	ret = copy_to_user(arg, sa, sizeof(*sa));
4224	if (ret)
4225		ret = -EFAULT;
4226
4227out:
4228	kfree(sa);
4229	return ret;
4230}
4231
4232static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4233					void __user *arg)
4234{
4235	size_t len;
4236	int ret;
4237	char label[BTRFS_LABEL_SIZE];
4238
4239	spin_lock(&fs_info->super_lock);
4240	memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4241	spin_unlock(&fs_info->super_lock);
4242
4243	len = strnlen(label, BTRFS_LABEL_SIZE);
4244
4245	if (len == BTRFS_LABEL_SIZE) {
4246		btrfs_warn(fs_info,
4247			   "label is too long, return the first %zu bytes",
4248			   --len);
4249	}
4250
4251	ret = copy_to_user(arg, label, len);
4252
4253	return ret ? -EFAULT : 0;
4254}
4255
4256static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4257{
4258	struct inode *inode = file_inode(file);
4259	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4260	struct btrfs_root *root = BTRFS_I(inode)->root;
4261	struct btrfs_super_block *super_block = fs_info->super_copy;
4262	struct btrfs_trans_handle *trans;
4263	char label[BTRFS_LABEL_SIZE];
4264	int ret;
4265
4266	if (!capable(CAP_SYS_ADMIN))
4267		return -EPERM;
4268
4269	if (copy_from_user(label, arg, sizeof(label)))
4270		return -EFAULT;
4271
4272	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4273		btrfs_err(fs_info,
4274			  "unable to set label with more than %d bytes",
4275			  BTRFS_LABEL_SIZE - 1);
4276		return -EINVAL;
4277	}
4278
4279	ret = mnt_want_write_file(file);
4280	if (ret)
4281		return ret;
4282
4283	trans = btrfs_start_transaction(root, 0);
4284	if (IS_ERR(trans)) {
4285		ret = PTR_ERR(trans);
4286		goto out_unlock;
4287	}
4288
4289	spin_lock(&fs_info->super_lock);
4290	strcpy(super_block->label, label);
4291	spin_unlock(&fs_info->super_lock);
4292	ret = btrfs_commit_transaction(trans);
4293
4294out_unlock:
4295	mnt_drop_write_file(file);
4296	return ret;
4297}
4298
4299#define INIT_FEATURE_FLAGS(suffix) \
4300	{ .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4301	  .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4302	  .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4303
4304int btrfs_ioctl_get_supported_features(void __user *arg)
4305{
4306	static const struct btrfs_ioctl_feature_flags features[3] = {
4307		INIT_FEATURE_FLAGS(SUPP),
4308		INIT_FEATURE_FLAGS(SAFE_SET),
4309		INIT_FEATURE_FLAGS(SAFE_CLEAR)
4310	};
4311
4312	if (copy_to_user(arg, &features, sizeof(features)))
4313		return -EFAULT;
4314
4315	return 0;
4316}
4317
4318static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4319					void __user *arg)
4320{
4321	struct btrfs_super_block *super_block = fs_info->super_copy;
4322	struct btrfs_ioctl_feature_flags features;
4323
4324	features.compat_flags = btrfs_super_compat_flags(super_block);
4325	features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4326	features.incompat_flags = btrfs_super_incompat_flags(super_block);
4327
4328	if (copy_to_user(arg, &features, sizeof(features)))
4329		return -EFAULT;
4330
4331	return 0;
4332}
4333
4334static int check_feature_bits(struct btrfs_fs_info *fs_info,
4335			      enum btrfs_feature_set set,
4336			      u64 change_mask, u64 flags, u64 supported_flags,
4337			      u64 safe_set, u64 safe_clear)
4338{
4339	const char *type = btrfs_feature_set_name(set);
4340	char *names;
4341	u64 disallowed, unsupported;
4342	u64 set_mask = flags & change_mask;
4343	u64 clear_mask = ~flags & change_mask;
4344
4345	unsupported = set_mask & ~supported_flags;
4346	if (unsupported) {
4347		names = btrfs_printable_features(set, unsupported);
4348		if (names) {
4349			btrfs_warn(fs_info,
4350				   "this kernel does not support the %s feature bit%s",
4351				   names, strchr(names, ',') ? "s" : "");
4352			kfree(names);
4353		} else
4354			btrfs_warn(fs_info,
4355				   "this kernel does not support %s bits 0x%llx",
4356				   type, unsupported);
4357		return -EOPNOTSUPP;
4358	}
4359
4360	disallowed = set_mask & ~safe_set;
4361	if (disallowed) {
4362		names = btrfs_printable_features(set, disallowed);
4363		if (names) {
4364			btrfs_warn(fs_info,
4365				   "can't set the %s feature bit%s while mounted",
4366				   names, strchr(names, ',') ? "s" : "");
4367			kfree(names);
4368		} else
4369			btrfs_warn(fs_info,
4370				   "can't set %s bits 0x%llx while mounted",
4371				   type, disallowed);
4372		return -EPERM;
4373	}
4374
4375	disallowed = clear_mask & ~safe_clear;
4376	if (disallowed) {
4377		names = btrfs_printable_features(set, disallowed);
4378		if (names) {
4379			btrfs_warn(fs_info,
4380				   "can't clear the %s feature bit%s while mounted",
4381				   names, strchr(names, ',') ? "s" : "");
4382			kfree(names);
4383		} else
4384			btrfs_warn(fs_info,
4385				   "can't clear %s bits 0x%llx while mounted",
4386				   type, disallowed);
4387		return -EPERM;
4388	}
4389
4390	return 0;
4391}
4392
4393#define check_feature(fs_info, change_mask, flags, mask_base)	\
4394check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags,	\
4395		   BTRFS_FEATURE_ ## mask_base ## _SUPP,	\
4396		   BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,	\
4397		   BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4398
4399static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4400{
4401	struct inode *inode = file_inode(file);
4402	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4403	struct btrfs_root *root = BTRFS_I(inode)->root;
4404	struct btrfs_super_block *super_block = fs_info->super_copy;
4405	struct btrfs_ioctl_feature_flags flags[2];
4406	struct btrfs_trans_handle *trans;
4407	u64 newflags;
4408	int ret;
4409
4410	if (!capable(CAP_SYS_ADMIN))
4411		return -EPERM;
4412
4413	if (copy_from_user(flags, arg, sizeof(flags)))
4414		return -EFAULT;
4415
4416	/* Nothing to do */
4417	if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4418	    !flags[0].incompat_flags)
4419		return 0;
4420
4421	ret = check_feature(fs_info, flags[0].compat_flags,
4422			    flags[1].compat_flags, COMPAT);
4423	if (ret)
4424		return ret;
4425
4426	ret = check_feature(fs_info, flags[0].compat_ro_flags,
4427			    flags[1].compat_ro_flags, COMPAT_RO);
4428	if (ret)
4429		return ret;
4430
4431	ret = check_feature(fs_info, flags[0].incompat_flags,
4432			    flags[1].incompat_flags, INCOMPAT);
4433	if (ret)
4434		return ret;
4435
4436	ret = mnt_want_write_file(file);
4437	if (ret)
4438		return ret;
4439
4440	trans = btrfs_start_transaction(root, 0);
4441	if (IS_ERR(trans)) {
4442		ret = PTR_ERR(trans);
4443		goto out_drop_write;
4444	}
4445
4446	spin_lock(&fs_info->super_lock);
4447	newflags = btrfs_super_compat_flags(super_block);
4448	newflags |= flags[0].compat_flags & flags[1].compat_flags;
4449	newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4450	btrfs_set_super_compat_flags(super_block, newflags);
4451
4452	newflags = btrfs_super_compat_ro_flags(super_block);
4453	newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4454	newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4455	btrfs_set_super_compat_ro_flags(super_block, newflags);
4456
4457	newflags = btrfs_super_incompat_flags(super_block);
4458	newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4459	newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4460	btrfs_set_super_incompat_flags(super_block, newflags);
4461	spin_unlock(&fs_info->super_lock);
4462
4463	ret = btrfs_commit_transaction(trans);
4464out_drop_write:
4465	mnt_drop_write_file(file);
4466
4467	return ret;
4468}
4469
4470static int _btrfs_ioctl_send(struct btrfs_inode *inode, void __user *argp, bool compat)
4471{
4472	struct btrfs_ioctl_send_args *arg;
4473	int ret;
4474
4475	if (compat) {
4476#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4477		struct btrfs_ioctl_send_args_32 args32 = { 0 };
4478
4479		ret = copy_from_user(&args32, argp, sizeof(args32));
4480		if (ret)
4481			return -EFAULT;
4482		arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4483		if (!arg)
4484			return -ENOMEM;
4485		arg->send_fd = args32.send_fd;
4486		arg->clone_sources_count = args32.clone_sources_count;
4487		arg->clone_sources = compat_ptr(args32.clone_sources);
4488		arg->parent_root = args32.parent_root;
4489		arg->flags = args32.flags;
4490		arg->version = args32.version;
4491		memcpy(arg->reserved, args32.reserved,
4492		       sizeof(args32.reserved));
4493#else
4494		return -ENOTTY;
4495#endif
4496	} else {
4497		arg = memdup_user(argp, sizeof(*arg));
4498		if (IS_ERR(arg))
4499			return PTR_ERR(arg);
4500	}
4501	ret = btrfs_ioctl_send(inode, arg);
4502	kfree(arg);
4503	return ret;
4504}
4505
4506static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4507				    bool compat)
4508{
4509	struct btrfs_ioctl_encoded_io_args args = { 0 };
4510	size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4511					     flags);
4512	size_t copy_end;
4513	struct btrfs_inode *inode = BTRFS_I(file_inode(file));
4514	struct btrfs_fs_info *fs_info = inode->root->fs_info;
4515	struct extent_io_tree *io_tree = &inode->io_tree;
4516	struct iovec iovstack[UIO_FASTIOV];
4517	struct iovec *iov = iovstack;
4518	struct iov_iter iter;
4519	loff_t pos;
4520	struct kiocb kiocb;
4521	ssize_t ret;
4522	u64 disk_bytenr, disk_io_size;
4523	struct extent_state *cached_state = NULL;
4524
4525	if (!capable(CAP_SYS_ADMIN)) {
4526		ret = -EPERM;
4527		goto out_acct;
4528	}
4529
4530	if (compat) {
4531#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4532		struct btrfs_ioctl_encoded_io_args_32 args32;
4533
4534		copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4535				       flags);
4536		if (copy_from_user(&args32, argp, copy_end)) {
4537			ret = -EFAULT;
4538			goto out_acct;
4539		}
4540		args.iov = compat_ptr(args32.iov);
4541		args.iovcnt = args32.iovcnt;
4542		args.offset = args32.offset;
4543		args.flags = args32.flags;
4544#else
4545		return -ENOTTY;
4546#endif
4547	} else {
4548		copy_end = copy_end_kernel;
4549		if (copy_from_user(&args, argp, copy_end)) {
4550			ret = -EFAULT;
4551			goto out_acct;
4552		}
4553	}
4554	if (args.flags != 0) {
4555		ret = -EINVAL;
4556		goto out_acct;
4557	}
4558
4559	ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4560			   &iov, &iter);
4561	if (ret < 0)
4562		goto out_acct;
4563
4564	if (iov_iter_count(&iter) == 0) {
4565		ret = 0;
4566		goto out_iov;
4567	}
4568	pos = args.offset;
4569	ret = rw_verify_area(READ, file, &pos, args.len);
4570	if (ret < 0)
4571		goto out_iov;
4572
4573	init_sync_kiocb(&kiocb, file);
4574	kiocb.ki_pos = pos;
4575
4576	ret = btrfs_encoded_read(&kiocb, &iter, &args, &cached_state,
4577				 &disk_bytenr, &disk_io_size);
4578
4579	if (ret == -EIOCBQUEUED) {
4580		bool unlocked = false;
4581		u64 start, lockend, count;
4582
4583		start = ALIGN_DOWN(kiocb.ki_pos, fs_info->sectorsize);
4584		lockend = start + BTRFS_MAX_UNCOMPRESSED - 1;
4585
4586		if (args.compression)
4587			count = disk_io_size;
4588		else
4589			count = args.len;
4590
4591		ret = btrfs_encoded_read_regular(&kiocb, &iter, start, lockend,
4592						 &cached_state, disk_bytenr,
4593						 disk_io_size, count,
4594						 args.compression, &unlocked);
4595
4596		if (!unlocked) {
4597			unlock_extent(io_tree, start, lockend, &cached_state);
4598			btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED);
4599		}
4600	}
4601
4602	if (ret >= 0) {
4603		fsnotify_access(file);
4604		if (copy_to_user(argp + copy_end,
4605				 (char *)&args + copy_end_kernel,
4606				 sizeof(args) - copy_end_kernel))
4607			ret = -EFAULT;
4608	}
4609
4610out_iov:
4611	kfree(iov);
4612out_acct:
4613	if (ret > 0)
4614		add_rchar(current, ret);
4615	inc_syscr(current);
4616	return ret;
4617}
4618
4619static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4620{
4621	struct btrfs_ioctl_encoded_io_args args;
4622	struct iovec iovstack[UIO_FASTIOV];
4623	struct iovec *iov = iovstack;
4624	struct iov_iter iter;
4625	loff_t pos;
4626	struct kiocb kiocb;
4627	ssize_t ret;
4628
4629	if (!capable(CAP_SYS_ADMIN)) {
4630		ret = -EPERM;
4631		goto out_acct;
4632	}
4633
4634	if (!(file->f_mode & FMODE_WRITE)) {
4635		ret = -EBADF;
4636		goto out_acct;
4637	}
4638
4639	if (compat) {
4640#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4641		struct btrfs_ioctl_encoded_io_args_32 args32;
4642
4643		if (copy_from_user(&args32, argp, sizeof(args32))) {
4644			ret = -EFAULT;
4645			goto out_acct;
4646		}
4647		args.iov = compat_ptr(args32.iov);
4648		args.iovcnt = args32.iovcnt;
4649		args.offset = args32.offset;
4650		args.flags = args32.flags;
4651		args.len = args32.len;
4652		args.unencoded_len = args32.unencoded_len;
4653		args.unencoded_offset = args32.unencoded_offset;
4654		args.compression = args32.compression;
4655		args.encryption = args32.encryption;
4656		memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4657#else
4658		return -ENOTTY;
4659#endif
4660	} else {
4661		if (copy_from_user(&args, argp, sizeof(args))) {
4662			ret = -EFAULT;
4663			goto out_acct;
4664		}
4665	}
4666
4667	ret = -EINVAL;
4668	if (args.flags != 0)
4669		goto out_acct;
4670	if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4671		goto out_acct;
4672	if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4673	    args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4674		goto out_acct;
4675	if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4676	    args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4677		goto out_acct;
4678	if (args.unencoded_offset > args.unencoded_len)
4679		goto out_acct;
4680	if (args.len > args.unencoded_len - args.unencoded_offset)
4681		goto out_acct;
4682
4683	ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4684			   &iov, &iter);
4685	if (ret < 0)
4686		goto out_acct;
4687
4688	if (iov_iter_count(&iter) == 0) {
4689		ret = 0;
4690		goto out_iov;
4691	}
4692	pos = args.offset;
4693	ret = rw_verify_area(WRITE, file, &pos, args.len);
4694	if (ret < 0)
4695		goto out_iov;
4696
4697	init_sync_kiocb(&kiocb, file);
4698	ret = kiocb_set_rw_flags(&kiocb, 0, WRITE);
4699	if (ret)
4700		goto out_iov;
4701	kiocb.ki_pos = pos;
4702
4703	file_start_write(file);
4704
4705	ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4706	if (ret > 0)
4707		fsnotify_modify(file);
4708
4709	file_end_write(file);
4710out_iov:
4711	kfree(iov);
4712out_acct:
4713	if (ret > 0)
4714		add_wchar(current, ret);
4715	inc_syscw(current);
4716	return ret;
4717}
4718
4719/*
4720 * Context that's attached to an encoded read io_uring command, in cmd->pdu. It
4721 * contains the fields in btrfs_uring_read_extent that are necessary to finish
4722 * off and cleanup the I/O in btrfs_uring_read_finished.
4723 */
4724struct btrfs_uring_priv {
4725	struct io_uring_cmd *cmd;
4726	struct page **pages;
4727	unsigned long nr_pages;
4728	struct kiocb iocb;
4729	struct iovec *iov;
4730	struct iov_iter iter;
4731	struct extent_state *cached_state;
4732	u64 count;
4733	u64 start;
4734	u64 lockend;
4735	int err;
4736	bool compressed;
4737};
4738
4739struct io_btrfs_cmd {
4740	struct btrfs_uring_priv *priv;
4741};
4742
4743static void btrfs_uring_read_finished(struct io_uring_cmd *cmd, unsigned int issue_flags)
4744{
4745	struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd);
4746	struct btrfs_uring_priv *priv = bc->priv;
4747	struct btrfs_inode *inode = BTRFS_I(file_inode(priv->iocb.ki_filp));
4748	struct extent_io_tree *io_tree = &inode->io_tree;
4749	unsigned long index;
4750	u64 cur;
4751	size_t page_offset;
4752	ssize_t ret;
4753
4754	/* The inode lock has already been acquired in btrfs_uring_read_extent.  */
4755	btrfs_lockdep_inode_acquire(inode, i_rwsem);
4756
4757	if (priv->err) {
4758		ret = priv->err;
4759		goto out;
4760	}
4761
4762	if (priv->compressed) {
4763		index = 0;
4764		page_offset = 0;
4765	} else {
4766		index = (priv->iocb.ki_pos - priv->start) >> PAGE_SHIFT;
4767		page_offset = offset_in_page(priv->iocb.ki_pos - priv->start);
4768	}
4769	cur = 0;
4770	while (cur < priv->count) {
4771		size_t bytes = min_t(size_t, priv->count - cur, PAGE_SIZE - page_offset);
4772
4773		if (copy_page_to_iter(priv->pages[index], page_offset, bytes,
4774				      &priv->iter) != bytes) {
4775			ret = -EFAULT;
4776			goto out;
4777		}
4778
4779		index++;
4780		cur += bytes;
4781		page_offset = 0;
4782	}
4783	ret = priv->count;
4784
4785out:
4786	unlock_extent(io_tree, priv->start, priv->lockend, &priv->cached_state);
4787	btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED);
4788
4789	io_uring_cmd_done(cmd, ret, 0, issue_flags);
4790	add_rchar(current, ret);
4791
4792	for (index = 0; index < priv->nr_pages; index++)
4793		__free_page(priv->pages[index]);
4794
4795	kfree(priv->pages);
4796	kfree(priv->iov);
4797	kfree(priv);
4798}
4799
4800void btrfs_uring_read_extent_endio(void *ctx, int err)
4801{
4802	struct btrfs_uring_priv *priv = ctx;
4803	struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(priv->cmd, struct io_btrfs_cmd);
4804
4805	priv->err = err;
4806	bc->priv = priv;
4807
4808	io_uring_cmd_complete_in_task(priv->cmd, btrfs_uring_read_finished);
4809}
4810
4811static int btrfs_uring_read_extent(struct kiocb *iocb, struct iov_iter *iter,
4812				   u64 start, u64 lockend,
4813				   struct extent_state *cached_state,
4814				   u64 disk_bytenr, u64 disk_io_size,
4815				   size_t count, bool compressed,
4816				   struct iovec *iov, struct io_uring_cmd *cmd)
4817{
4818	struct btrfs_inode *inode = BTRFS_I(file_inode(iocb->ki_filp));
4819	struct extent_io_tree *io_tree = &inode->io_tree;
4820	struct page **pages;
4821	struct btrfs_uring_priv *priv = NULL;
4822	unsigned long nr_pages;
4823	int ret;
4824
4825	nr_pages = DIV_ROUND_UP(disk_io_size, PAGE_SIZE);
4826	pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS);
4827	if (!pages)
4828		return -ENOMEM;
4829	ret = btrfs_alloc_page_array(nr_pages, pages, 0);
4830	if (ret) {
4831		ret = -ENOMEM;
4832		goto out_fail;
4833	}
4834
4835	priv = kmalloc(sizeof(*priv), GFP_NOFS);
4836	if (!priv) {
4837		ret = -ENOMEM;
4838		goto out_fail;
4839	}
4840
4841	priv->iocb = *iocb;
4842	priv->iov = iov;
4843	priv->iter = *iter;
4844	priv->count = count;
4845	priv->cmd = cmd;
4846	priv->cached_state = cached_state;
4847	priv->compressed = compressed;
4848	priv->nr_pages = nr_pages;
4849	priv->pages = pages;
4850	priv->start = start;
4851	priv->lockend = lockend;
4852	priv->err = 0;
4853
4854	ret = btrfs_encoded_read_regular_fill_pages(inode, disk_bytenr,
4855						    disk_io_size, pages, priv);
4856	if (ret && ret != -EIOCBQUEUED)
4857		goto out_fail;
4858
4859	/*
4860	 * If we return -EIOCBQUEUED, we're deferring the cleanup to
4861	 * btrfs_uring_read_finished(), which will handle unlocking the extent
4862	 * and inode and freeing the allocations.
4863	 */
4864
4865	/*
4866	 * We're returning to userspace with the inode lock held, and that's
4867	 * okay - it'll get unlocked in a worker thread.  Call
4868	 * btrfs_lockdep_inode_release() to avoid confusing lockdep.
4869	 */
4870	btrfs_lockdep_inode_release(inode, i_rwsem);
4871
4872	return -EIOCBQUEUED;
4873
4874out_fail:
4875	unlock_extent(io_tree, start, lockend, &cached_state);
4876	btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED);
4877	kfree(priv);
4878	return ret;
4879}
4880
4881struct btrfs_uring_encoded_data {
4882	struct btrfs_ioctl_encoded_io_args args;
4883	struct iovec iovstack[UIO_FASTIOV];
4884	struct iovec *iov;
4885	struct iov_iter iter;
4886};
4887
4888static int btrfs_uring_encoded_read(struct io_uring_cmd *cmd, unsigned int issue_flags)
4889{
4890	size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args, flags);
4891	size_t copy_end;
4892	int ret;
4893	u64 disk_bytenr, disk_io_size;
4894	struct file *file;
4895	struct btrfs_inode *inode;
4896	struct btrfs_fs_info *fs_info;
4897	struct extent_io_tree *io_tree;
4898	loff_t pos;
4899	struct kiocb kiocb;
4900	struct extent_state *cached_state = NULL;
4901	u64 start, lockend;
4902	void __user *sqe_addr;
4903	struct btrfs_uring_encoded_data *data = io_uring_cmd_get_async_data(cmd)->op_data;
4904
4905	if (!capable(CAP_SYS_ADMIN)) {
4906		ret = -EPERM;
4907		goto out_acct;
4908	}
4909	file = cmd->file;
4910	inode = BTRFS_I(file->f_inode);
4911	fs_info = inode->root->fs_info;
4912	io_tree = &inode->io_tree;
4913	sqe_addr = u64_to_user_ptr(READ_ONCE(cmd->sqe->addr));
4914
4915	if (issue_flags & IO_URING_F_COMPAT) {
4916#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4917		copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32, flags);
4918#else
4919		return -ENOTTY;
4920#endif
4921	} else {
4922		copy_end = copy_end_kernel;
4923	}
4924
4925	if (!data) {
4926		data = kzalloc(sizeof(*data), GFP_NOFS);
4927		if (!data) {
4928			ret = -ENOMEM;
4929			goto out_acct;
4930		}
4931
4932		io_uring_cmd_get_async_data(cmd)->op_data = data;
4933
4934		if (issue_flags & IO_URING_F_COMPAT) {
4935#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4936			struct btrfs_ioctl_encoded_io_args_32 args32;
4937
4938			if (copy_from_user(&args32, sqe_addr, copy_end)) {
4939				ret = -EFAULT;
4940				goto out_acct;
4941			}
4942
4943			data->args.iov = compat_ptr(args32.iov);
4944			data->args.iovcnt = args32.iovcnt;
4945			data->args.offset = args32.offset;
4946			data->args.flags = args32.flags;
4947#endif
4948		} else {
4949			if (copy_from_user(&data->args, sqe_addr, copy_end)) {
4950				ret = -EFAULT;
4951				goto out_acct;
4952			}
4953		}
4954
4955		if (data->args.flags != 0) {
4956			ret = -EINVAL;
4957			goto out_acct;
4958		}
4959
4960		data->iov = data->iovstack;
4961		ret = import_iovec(ITER_DEST, data->args.iov, data->args.iovcnt,
4962				   ARRAY_SIZE(data->iovstack), &data->iov,
4963				   &data->iter);
4964		if (ret < 0)
4965			goto out_acct;
4966
4967		if (iov_iter_count(&data->iter) == 0) {
4968			ret = 0;
4969			goto out_free;
4970		}
4971	}
4972
4973	pos = data->args.offset;
4974	ret = rw_verify_area(READ, file, &pos, data->args.len);
4975	if (ret < 0)
4976		goto out_free;
4977
4978	init_sync_kiocb(&kiocb, file);
4979	kiocb.ki_pos = pos;
4980
4981	if (issue_flags & IO_URING_F_NONBLOCK)
4982		kiocb.ki_flags |= IOCB_NOWAIT;
4983
4984	start = ALIGN_DOWN(pos, fs_info->sectorsize);
4985	lockend = start + BTRFS_MAX_UNCOMPRESSED - 1;
4986
4987	ret = btrfs_encoded_read(&kiocb, &data->iter, &data->args, &cached_state,
4988				 &disk_bytenr, &disk_io_size);
4989	if (ret < 0 && ret != -EIOCBQUEUED)
4990		goto out_free;
4991
4992	file_accessed(file);
4993
4994	if (copy_to_user(sqe_addr + copy_end,
4995			 (const char *)&data->args + copy_end_kernel,
4996			 sizeof(data->args) - copy_end_kernel)) {
4997		if (ret == -EIOCBQUEUED) {
4998			unlock_extent(io_tree, start, lockend, &cached_state);
4999			btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED);
5000		}
5001		ret = -EFAULT;
5002		goto out_free;
5003	}
5004
5005	if (ret == -EIOCBQUEUED) {
5006		u64 count = min_t(u64, iov_iter_count(&data->iter), disk_io_size);
5007
5008		/* Match ioctl by not returning past EOF if uncompressed. */
5009		if (!data->args.compression)
5010			count = min_t(u64, count, data->args.len);
5011
5012		ret = btrfs_uring_read_extent(&kiocb, &data->iter, start, lockend,
5013					      cached_state, disk_bytenr, disk_io_size,
5014					      count, data->args.compression,
5015					      data->iov, cmd);
5016
5017		goto out_acct;
5018	}
5019
5020out_free:
5021	kfree(data->iov);
5022
5023out_acct:
5024	if (ret > 0)
5025		add_rchar(current, ret);
5026	inc_syscr(current);
5027
5028	return ret;
5029}
5030
5031int btrfs_uring_cmd(struct io_uring_cmd *cmd, unsigned int issue_flags)
5032{
5033	switch (cmd->cmd_op) {
5034	case BTRFS_IOC_ENCODED_READ:
5035#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5036	case BTRFS_IOC_ENCODED_READ_32:
5037#endif
5038		return btrfs_uring_encoded_read(cmd, issue_flags);
5039	}
5040
5041	return -EINVAL;
5042}
5043
5044static int btrfs_ioctl_subvol_sync(struct btrfs_fs_info *fs_info, void __user *argp)
5045{
5046	struct btrfs_root *root;
5047	struct btrfs_ioctl_subvol_wait args = { 0 };
5048	signed long sched_ret;
5049	int refs;
5050	u64 root_flags;
5051	bool wait_for_deletion = false;
5052	bool found = false;
5053
5054	if (copy_from_user(&args, argp, sizeof(args)))
5055		return -EFAULT;
5056
5057	switch (args.mode) {
5058	case BTRFS_SUBVOL_SYNC_WAIT_FOR_QUEUED:
5059		/*
5060		 * Wait for the first one deleted that waits until all previous
5061		 * are cleaned.
5062		 */
5063		spin_lock(&fs_info->trans_lock);
5064		if (!list_empty(&fs_info->dead_roots)) {
5065			root = list_last_entry(&fs_info->dead_roots,
5066					       struct btrfs_root, root_list);
5067			args.subvolid = btrfs_root_id(root);
5068			found = true;
5069		}
5070		spin_unlock(&fs_info->trans_lock);
5071		if (!found)
5072			return -ENOENT;
5073
5074		fallthrough;
5075	case BTRFS_SUBVOL_SYNC_WAIT_FOR_ONE:
5076		if ((0 < args.subvolid && args.subvolid < BTRFS_FIRST_FREE_OBJECTID) ||
5077		    BTRFS_LAST_FREE_OBJECTID < args.subvolid)
5078			return -EINVAL;
5079		break;
5080	case BTRFS_SUBVOL_SYNC_COUNT:
5081		spin_lock(&fs_info->trans_lock);
5082		args.count = list_count_nodes(&fs_info->dead_roots);
5083		spin_unlock(&fs_info->trans_lock);
5084		if (copy_to_user(argp, &args, sizeof(args)))
5085			return -EFAULT;
5086		return 0;
5087	case BTRFS_SUBVOL_SYNC_PEEK_FIRST:
5088		spin_lock(&fs_info->trans_lock);
5089		/* Last in the list was deleted first. */
5090		if (!list_empty(&fs_info->dead_roots)) {
5091			root = list_last_entry(&fs_info->dead_roots,
5092					       struct btrfs_root, root_list);
5093			args.subvolid = btrfs_root_id(root);
5094		} else {
5095			args.subvolid = 0;
5096		}
5097		spin_unlock(&fs_info->trans_lock);
5098		if (copy_to_user(argp, &args, sizeof(args)))
5099			return -EFAULT;
5100		return 0;
5101	case BTRFS_SUBVOL_SYNC_PEEK_LAST:
5102		spin_lock(&fs_info->trans_lock);
5103		/* First in the list was deleted last. */
5104		if (!list_empty(&fs_info->dead_roots)) {
5105			root = list_first_entry(&fs_info->dead_roots,
5106						struct btrfs_root, root_list);
5107			args.subvolid = btrfs_root_id(root);
5108		} else {
5109			args.subvolid = 0;
5110		}
5111		spin_unlock(&fs_info->trans_lock);
5112		if (copy_to_user(argp, &args, sizeof(args)))
5113			return -EFAULT;
5114		return 0;
5115	default:
5116		return -EINVAL;
5117	}
5118
5119	/* 32bit limitation: fs_roots_radix key is not wide enough. */
5120	if (sizeof(unsigned long) != sizeof(u64) && args.subvolid > U32_MAX)
5121		return -EOVERFLOW;
5122
5123	while (1) {
5124		/* Wait for the specific one. */
5125		if (down_read_interruptible(&fs_info->subvol_sem) == -EINTR)
5126			return -EINTR;
5127		refs = -1;
5128		spin_lock(&fs_info->fs_roots_radix_lock);
5129		root = radix_tree_lookup(&fs_info->fs_roots_radix,
5130					 (unsigned long)args.subvolid);
5131		if (root) {
5132			spin_lock(&root->root_item_lock);
5133			refs = btrfs_root_refs(&root->root_item);
5134			root_flags = btrfs_root_flags(&root->root_item);
5135			spin_unlock(&root->root_item_lock);
5136		}
5137		spin_unlock(&fs_info->fs_roots_radix_lock);
5138		up_read(&fs_info->subvol_sem);
5139
5140		/* Subvolume does not exist. */
5141		if (!root)
5142			return -ENOENT;
5143
5144		/* Subvolume not deleted at all. */
5145		if (refs > 0)
5146			return -EEXIST;
5147		/* We've waited and now the subvolume is gone. */
5148		if (wait_for_deletion && refs == -1) {
5149			/* Return the one we waited for as the last one. */
5150			if (copy_to_user(argp, &args, sizeof(args)))
5151				return -EFAULT;
5152			return 0;
5153		}
5154
5155		/* Subvolume not found on the first try (deleted or never existed). */
5156		if (refs == -1)
5157			return -ENOENT;
5158
5159		wait_for_deletion = true;
5160		ASSERT(root_flags & BTRFS_ROOT_SUBVOL_DEAD);
5161		sched_ret = schedule_timeout_interruptible(HZ);
5162		/* Early wake up or error. */
5163		if (sched_ret != 0)
5164			return -EINTR;
5165	}
5166
5167	return 0;
5168}
5169
5170long btrfs_ioctl(struct file *file, unsigned int
5171		cmd, unsigned long arg)
5172{
5173	struct inode *inode = file_inode(file);
5174	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
5175	struct btrfs_root *root = BTRFS_I(inode)->root;
5176	void __user *argp = (void __user *)arg;
5177
5178	switch (cmd) {
5179	case FS_IOC_GETVERSION:
5180		return btrfs_ioctl_getversion(inode, argp);
5181	case FS_IOC_GETFSLABEL:
5182		return btrfs_ioctl_get_fslabel(fs_info, argp);
5183	case FS_IOC_SETFSLABEL:
5184		return btrfs_ioctl_set_fslabel(file, argp);
5185	case FITRIM:
5186		return btrfs_ioctl_fitrim(fs_info, argp);
5187	case BTRFS_IOC_SNAP_CREATE:
5188		return btrfs_ioctl_snap_create(file, argp, 0);
5189	case BTRFS_IOC_SNAP_CREATE_V2:
5190		return btrfs_ioctl_snap_create_v2(file, argp, 0);
5191	case BTRFS_IOC_SUBVOL_CREATE:
5192		return btrfs_ioctl_snap_create(file, argp, 1);
5193	case BTRFS_IOC_SUBVOL_CREATE_V2:
5194		return btrfs_ioctl_snap_create_v2(file, argp, 1);
5195	case BTRFS_IOC_SNAP_DESTROY:
5196		return btrfs_ioctl_snap_destroy(file, argp, false);
5197	case BTRFS_IOC_SNAP_DESTROY_V2:
5198		return btrfs_ioctl_snap_destroy(file, argp, true);
5199	case BTRFS_IOC_SUBVOL_GETFLAGS:
5200		return btrfs_ioctl_subvol_getflags(inode, argp);
5201	case BTRFS_IOC_SUBVOL_SETFLAGS:
5202		return btrfs_ioctl_subvol_setflags(file, argp);
5203	case BTRFS_IOC_DEFAULT_SUBVOL:
5204		return btrfs_ioctl_default_subvol(file, argp);
5205	case BTRFS_IOC_DEFRAG:
5206		return btrfs_ioctl_defrag(file, NULL);
5207	case BTRFS_IOC_DEFRAG_RANGE:
5208		return btrfs_ioctl_defrag(file, argp);
5209	case BTRFS_IOC_RESIZE:
5210		return btrfs_ioctl_resize(file, argp);
5211	case BTRFS_IOC_ADD_DEV:
5212		return btrfs_ioctl_add_dev(fs_info, argp);
5213	case BTRFS_IOC_RM_DEV:
5214		return btrfs_ioctl_rm_dev(file, argp);
5215	case BTRFS_IOC_RM_DEV_V2:
5216		return btrfs_ioctl_rm_dev_v2(file, argp);
5217	case BTRFS_IOC_FS_INFO:
5218		return btrfs_ioctl_fs_info(fs_info, argp);
5219	case BTRFS_IOC_DEV_INFO:
5220		return btrfs_ioctl_dev_info(fs_info, argp);
5221	case BTRFS_IOC_TREE_SEARCH:
5222		return btrfs_ioctl_tree_search(inode, argp);
5223	case BTRFS_IOC_TREE_SEARCH_V2:
5224		return btrfs_ioctl_tree_search_v2(inode, argp);
5225	case BTRFS_IOC_INO_LOOKUP:
5226		return btrfs_ioctl_ino_lookup(root, argp);
5227	case BTRFS_IOC_INO_PATHS:
5228		return btrfs_ioctl_ino_to_path(root, argp);
5229	case BTRFS_IOC_LOGICAL_INO:
5230		return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5231	case BTRFS_IOC_LOGICAL_INO_V2:
5232		return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5233	case BTRFS_IOC_SPACE_INFO:
5234		return btrfs_ioctl_space_info(fs_info, argp);
5235	case BTRFS_IOC_SYNC: {
5236		int ret;
5237
5238		ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
5239		if (ret)
5240			return ret;
5241		ret = btrfs_sync_fs(inode->i_sb, 1);
5242		/*
5243		 * There may be work for the cleaner kthread to do (subvolume
5244		 * deletion, delayed iputs, defrag inodes, etc), so wake it up.
 
5245		 */
5246		wake_up_process(fs_info->cleaner_kthread);
5247		return ret;
5248	}
5249	case BTRFS_IOC_START_SYNC:
5250		return btrfs_ioctl_start_sync(root, argp);
5251	case BTRFS_IOC_WAIT_SYNC:
5252		return btrfs_ioctl_wait_sync(fs_info, argp);
5253	case BTRFS_IOC_SCRUB:
5254		return btrfs_ioctl_scrub(file, argp);
5255	case BTRFS_IOC_SCRUB_CANCEL:
5256		return btrfs_ioctl_scrub_cancel(fs_info);
5257	case BTRFS_IOC_SCRUB_PROGRESS:
5258		return btrfs_ioctl_scrub_progress(fs_info, argp);
5259	case BTRFS_IOC_BALANCE_V2:
5260		return btrfs_ioctl_balance(file, argp);
5261	case BTRFS_IOC_BALANCE_CTL:
5262		return btrfs_ioctl_balance_ctl(fs_info, arg);
5263	case BTRFS_IOC_BALANCE_PROGRESS:
5264		return btrfs_ioctl_balance_progress(fs_info, argp);
5265	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5266		return btrfs_ioctl_set_received_subvol(file, argp);
5267#ifdef CONFIG_64BIT
5268	case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5269		return btrfs_ioctl_set_received_subvol_32(file, argp);
5270#endif
5271	case BTRFS_IOC_SEND:
5272		return _btrfs_ioctl_send(BTRFS_I(inode), argp, false);
5273#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5274	case BTRFS_IOC_SEND_32:
5275		return _btrfs_ioctl_send(BTRFS_I(inode), argp, true);
5276#endif
5277	case BTRFS_IOC_GET_DEV_STATS:
5278		return btrfs_ioctl_get_dev_stats(fs_info, argp);
5279	case BTRFS_IOC_QUOTA_CTL:
5280		return btrfs_ioctl_quota_ctl(file, argp);
5281	case BTRFS_IOC_QGROUP_ASSIGN:
5282		return btrfs_ioctl_qgroup_assign(file, argp);
5283	case BTRFS_IOC_QGROUP_CREATE:
5284		return btrfs_ioctl_qgroup_create(file, argp);
5285	case BTRFS_IOC_QGROUP_LIMIT:
5286		return btrfs_ioctl_qgroup_limit(file, argp);
5287	case BTRFS_IOC_QUOTA_RESCAN:
5288		return btrfs_ioctl_quota_rescan(file, argp);
5289	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5290		return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5291	case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5292		return btrfs_ioctl_quota_rescan_wait(fs_info);
5293	case BTRFS_IOC_DEV_REPLACE:
5294		return btrfs_ioctl_dev_replace(fs_info, argp);
5295	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5296		return btrfs_ioctl_get_supported_features(argp);
5297	case BTRFS_IOC_GET_FEATURES:
5298		return btrfs_ioctl_get_features(fs_info, argp);
5299	case BTRFS_IOC_SET_FEATURES:
5300		return btrfs_ioctl_set_features(file, argp);
5301	case BTRFS_IOC_GET_SUBVOL_INFO:
5302		return btrfs_ioctl_get_subvol_info(inode, argp);
5303	case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5304		return btrfs_ioctl_get_subvol_rootref(root, argp);
5305	case BTRFS_IOC_INO_LOOKUP_USER:
5306		return btrfs_ioctl_ino_lookup_user(file, argp);
5307	case FS_IOC_ENABLE_VERITY:
5308		return fsverity_ioctl_enable(file, (const void __user *)argp);
5309	case FS_IOC_MEASURE_VERITY:
5310		return fsverity_ioctl_measure(file, argp);
5311	case BTRFS_IOC_ENCODED_READ:
5312		return btrfs_ioctl_encoded_read(file, argp, false);
5313	case BTRFS_IOC_ENCODED_WRITE:
5314		return btrfs_ioctl_encoded_write(file, argp, false);
5315#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5316	case BTRFS_IOC_ENCODED_READ_32:
5317		return btrfs_ioctl_encoded_read(file, argp, true);
5318	case BTRFS_IOC_ENCODED_WRITE_32:
5319		return btrfs_ioctl_encoded_write(file, argp, true);
5320#endif
5321	case BTRFS_IOC_SUBVOL_SYNC_WAIT:
5322		return btrfs_ioctl_subvol_sync(fs_info, argp);
5323	}
5324
5325	return -ENOTTY;
5326}
5327
5328#ifdef CONFIG_COMPAT
5329long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5330{
5331	/*
5332	 * These all access 32-bit values anyway so no further
5333	 * handling is necessary.
5334	 */
5335	switch (cmd) {
5336	case FS_IOC32_GETVERSION:
5337		cmd = FS_IOC_GETVERSION;
5338		break;
5339	}
5340
5341	return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5342}
5343#endif
v6.9.4
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * Copyright (C) 2007 Oracle.  All rights reserved.
   4 */
   5
   6#include <linux/kernel.h>
   7#include <linux/bio.h>
   8#include <linux/file.h>
   9#include <linux/fs.h>
  10#include <linux/fsnotify.h>
  11#include <linux/pagemap.h>
  12#include <linux/highmem.h>
  13#include <linux/time.h>
  14#include <linux/string.h>
  15#include <linux/backing-dev.h>
  16#include <linux/mount.h>
  17#include <linux/namei.h>
  18#include <linux/writeback.h>
  19#include <linux/compat.h>
  20#include <linux/security.h>
  21#include <linux/xattr.h>
  22#include <linux/mm.h>
  23#include <linux/slab.h>
  24#include <linux/blkdev.h>
  25#include <linux/uuid.h>
  26#include <linux/btrfs.h>
  27#include <linux/uaccess.h>
  28#include <linux/iversion.h>
  29#include <linux/fileattr.h>
  30#include <linux/fsverity.h>
  31#include <linux/sched/xacct.h>
 
  32#include "ctree.h"
  33#include "disk-io.h"
  34#include "export.h"
  35#include "transaction.h"
  36#include "btrfs_inode.h"
  37#include "volumes.h"
  38#include "locking.h"
  39#include "backref.h"
  40#include "send.h"
  41#include "dev-replace.h"
  42#include "props.h"
  43#include "sysfs.h"
  44#include "qgroup.h"
  45#include "tree-log.h"
  46#include "compression.h"
  47#include "space-info.h"
  48#include "block-group.h"
  49#include "fs.h"
  50#include "accessors.h"
  51#include "extent-tree.h"
  52#include "root-tree.h"
  53#include "defrag.h"
  54#include "dir-item.h"
  55#include "uuid-tree.h"
  56#include "ioctl.h"
  57#include "file.h"
  58#include "scrub.h"
  59#include "super.h"
  60
  61#ifdef CONFIG_64BIT
  62/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
  63 * structures are incorrect, as the timespec structure from userspace
  64 * is 4 bytes too small. We define these alternatives here to teach
  65 * the kernel about the 32-bit struct packing.
  66 */
  67struct btrfs_ioctl_timespec_32 {
  68	__u64 sec;
  69	__u32 nsec;
  70} __attribute__ ((__packed__));
  71
  72struct btrfs_ioctl_received_subvol_args_32 {
  73	char	uuid[BTRFS_UUID_SIZE];	/* in */
  74	__u64	stransid;		/* in */
  75	__u64	rtransid;		/* out */
  76	struct btrfs_ioctl_timespec_32 stime; /* in */
  77	struct btrfs_ioctl_timespec_32 rtime; /* out */
  78	__u64	flags;			/* in */
  79	__u64	reserved[16];		/* in */
  80} __attribute__ ((__packed__));
  81
  82#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
  83				struct btrfs_ioctl_received_subvol_args_32)
  84#endif
  85
  86#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
  87struct btrfs_ioctl_send_args_32 {
  88	__s64 send_fd;			/* in */
  89	__u64 clone_sources_count;	/* in */
  90	compat_uptr_t clone_sources;	/* in */
  91	__u64 parent_root;		/* in */
  92	__u64 flags;			/* in */
  93	__u32 version;			/* in */
  94	__u8  reserved[28];		/* in */
  95} __attribute__ ((__packed__));
  96
  97#define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
  98			       struct btrfs_ioctl_send_args_32)
  99
 100struct btrfs_ioctl_encoded_io_args_32 {
 101	compat_uptr_t iov;
 102	compat_ulong_t iovcnt;
 103	__s64 offset;
 104	__u64 flags;
 105	__u64 len;
 106	__u64 unencoded_len;
 107	__u64 unencoded_offset;
 108	__u32 compression;
 109	__u32 encryption;
 110	__u8 reserved[64];
 111};
 112
 113#define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
 114				       struct btrfs_ioctl_encoded_io_args_32)
 115#define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
 116					struct btrfs_ioctl_encoded_io_args_32)
 117#endif
 118
 119/* Mask out flags that are inappropriate for the given type of inode. */
 120static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
 121		unsigned int flags)
 122{
 123	if (S_ISDIR(inode->i_mode))
 124		return flags;
 125	else if (S_ISREG(inode->i_mode))
 126		return flags & ~FS_DIRSYNC_FL;
 127	else
 128		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
 129}
 130
 131/*
 132 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
 133 * ioctl.
 134 */
 135static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
 136{
 137	unsigned int iflags = 0;
 138	u32 flags = binode->flags;
 139	u32 ro_flags = binode->ro_flags;
 140
 141	if (flags & BTRFS_INODE_SYNC)
 142		iflags |= FS_SYNC_FL;
 143	if (flags & BTRFS_INODE_IMMUTABLE)
 144		iflags |= FS_IMMUTABLE_FL;
 145	if (flags & BTRFS_INODE_APPEND)
 146		iflags |= FS_APPEND_FL;
 147	if (flags & BTRFS_INODE_NODUMP)
 148		iflags |= FS_NODUMP_FL;
 149	if (flags & BTRFS_INODE_NOATIME)
 150		iflags |= FS_NOATIME_FL;
 151	if (flags & BTRFS_INODE_DIRSYNC)
 152		iflags |= FS_DIRSYNC_FL;
 153	if (flags & BTRFS_INODE_NODATACOW)
 154		iflags |= FS_NOCOW_FL;
 155	if (ro_flags & BTRFS_INODE_RO_VERITY)
 156		iflags |= FS_VERITY_FL;
 157
 158	if (flags & BTRFS_INODE_NOCOMPRESS)
 159		iflags |= FS_NOCOMP_FL;
 160	else if (flags & BTRFS_INODE_COMPRESS)
 161		iflags |= FS_COMPR_FL;
 162
 163	return iflags;
 164}
 165
 166/*
 167 * Update inode->i_flags based on the btrfs internal flags.
 168 */
 169void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
 170{
 171	struct btrfs_inode *binode = BTRFS_I(inode);
 172	unsigned int new_fl = 0;
 173
 174	if (binode->flags & BTRFS_INODE_SYNC)
 175		new_fl |= S_SYNC;
 176	if (binode->flags & BTRFS_INODE_IMMUTABLE)
 177		new_fl |= S_IMMUTABLE;
 178	if (binode->flags & BTRFS_INODE_APPEND)
 179		new_fl |= S_APPEND;
 180	if (binode->flags & BTRFS_INODE_NOATIME)
 181		new_fl |= S_NOATIME;
 182	if (binode->flags & BTRFS_INODE_DIRSYNC)
 183		new_fl |= S_DIRSYNC;
 184	if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
 185		new_fl |= S_VERITY;
 186
 187	set_mask_bits(&inode->i_flags,
 188		      S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
 189		      S_VERITY, new_fl);
 190}
 191
 192/*
 193 * Check if @flags are a supported and valid set of FS_*_FL flags and that
 194 * the old and new flags are not conflicting
 195 */
 196static int check_fsflags(unsigned int old_flags, unsigned int flags)
 197{
 198	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
 199		      FS_NOATIME_FL | FS_NODUMP_FL | \
 200		      FS_SYNC_FL | FS_DIRSYNC_FL | \
 201		      FS_NOCOMP_FL | FS_COMPR_FL |
 202		      FS_NOCOW_FL))
 203		return -EOPNOTSUPP;
 204
 205	/* COMPR and NOCOMP on new/old are valid */
 206	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
 207		return -EINVAL;
 208
 209	if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
 210		return -EINVAL;
 211
 212	/* NOCOW and compression options are mutually exclusive */
 213	if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
 214		return -EINVAL;
 215	if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
 216		return -EINVAL;
 217
 218	return 0;
 219}
 220
 221static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
 222				    unsigned int flags)
 223{
 224	if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
 225		return -EPERM;
 226
 227	return 0;
 228}
 229
 230int btrfs_check_ioctl_vol_args_path(const struct btrfs_ioctl_vol_args *vol_args)
 231{
 232	if (memchr(vol_args->name, 0, sizeof(vol_args->name)) == NULL)
 233		return -ENAMETOOLONG;
 234	return 0;
 235}
 236
 237static int btrfs_check_ioctl_vol_args2_subvol_name(const struct btrfs_ioctl_vol_args_v2 *vol_args2)
 238{
 239	if (memchr(vol_args2->name, 0, sizeof(vol_args2->name)) == NULL)
 240		return -ENAMETOOLONG;
 241	return 0;
 242}
 243
 244/*
 245 * Set flags/xflags from the internal inode flags. The remaining items of
 246 * fsxattr are zeroed.
 247 */
 248int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
 249{
 250	struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
 251
 252	fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
 253	return 0;
 254}
 255
 256int btrfs_fileattr_set(struct mnt_idmap *idmap,
 257		       struct dentry *dentry, struct fileattr *fa)
 258{
 259	struct inode *inode = d_inode(dentry);
 260	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
 261	struct btrfs_inode *binode = BTRFS_I(inode);
 262	struct btrfs_root *root = binode->root;
 263	struct btrfs_trans_handle *trans;
 264	unsigned int fsflags, old_fsflags;
 265	int ret;
 266	const char *comp = NULL;
 267	u32 binode_flags;
 268
 269	if (btrfs_root_readonly(root))
 270		return -EROFS;
 271
 272	if (fileattr_has_fsx(fa))
 273		return -EOPNOTSUPP;
 274
 275	fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
 276	old_fsflags = btrfs_inode_flags_to_fsflags(binode);
 277	ret = check_fsflags(old_fsflags, fsflags);
 278	if (ret)
 279		return ret;
 280
 281	ret = check_fsflags_compatible(fs_info, fsflags);
 282	if (ret)
 283		return ret;
 284
 285	binode_flags = binode->flags;
 286	if (fsflags & FS_SYNC_FL)
 287		binode_flags |= BTRFS_INODE_SYNC;
 288	else
 289		binode_flags &= ~BTRFS_INODE_SYNC;
 290	if (fsflags & FS_IMMUTABLE_FL)
 291		binode_flags |= BTRFS_INODE_IMMUTABLE;
 292	else
 293		binode_flags &= ~BTRFS_INODE_IMMUTABLE;
 294	if (fsflags & FS_APPEND_FL)
 295		binode_flags |= BTRFS_INODE_APPEND;
 296	else
 297		binode_flags &= ~BTRFS_INODE_APPEND;
 298	if (fsflags & FS_NODUMP_FL)
 299		binode_flags |= BTRFS_INODE_NODUMP;
 300	else
 301		binode_flags &= ~BTRFS_INODE_NODUMP;
 302	if (fsflags & FS_NOATIME_FL)
 303		binode_flags |= BTRFS_INODE_NOATIME;
 304	else
 305		binode_flags &= ~BTRFS_INODE_NOATIME;
 306
 307	/* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
 308	if (!fa->flags_valid) {
 309		/* 1 item for the inode */
 310		trans = btrfs_start_transaction(root, 1);
 311		if (IS_ERR(trans))
 312			return PTR_ERR(trans);
 313		goto update_flags;
 314	}
 315
 316	if (fsflags & FS_DIRSYNC_FL)
 317		binode_flags |= BTRFS_INODE_DIRSYNC;
 318	else
 319		binode_flags &= ~BTRFS_INODE_DIRSYNC;
 320	if (fsflags & FS_NOCOW_FL) {
 321		if (S_ISREG(inode->i_mode)) {
 322			/*
 323			 * It's safe to turn csums off here, no extents exist.
 324			 * Otherwise we want the flag to reflect the real COW
 325			 * status of the file and will not set it.
 326			 */
 327			if (inode->i_size == 0)
 328				binode_flags |= BTRFS_INODE_NODATACOW |
 329						BTRFS_INODE_NODATASUM;
 330		} else {
 331			binode_flags |= BTRFS_INODE_NODATACOW;
 332		}
 333	} else {
 334		/*
 335		 * Revert back under same assumptions as above
 336		 */
 337		if (S_ISREG(inode->i_mode)) {
 338			if (inode->i_size == 0)
 339				binode_flags &= ~(BTRFS_INODE_NODATACOW |
 340						  BTRFS_INODE_NODATASUM);
 341		} else {
 342			binode_flags &= ~BTRFS_INODE_NODATACOW;
 343		}
 344	}
 345
 346	/*
 347	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
 348	 * flag may be changed automatically if compression code won't make
 349	 * things smaller.
 350	 */
 351	if (fsflags & FS_NOCOMP_FL) {
 352		binode_flags &= ~BTRFS_INODE_COMPRESS;
 353		binode_flags |= BTRFS_INODE_NOCOMPRESS;
 354	} else if (fsflags & FS_COMPR_FL) {
 355
 356		if (IS_SWAPFILE(inode))
 357			return -ETXTBSY;
 358
 359		binode_flags |= BTRFS_INODE_COMPRESS;
 360		binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
 361
 362		comp = btrfs_compress_type2str(fs_info->compress_type);
 363		if (!comp || comp[0] == 0)
 364			comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
 365	} else {
 366		binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
 367	}
 368
 369	/*
 370	 * 1 for inode item
 371	 * 2 for properties
 372	 */
 373	trans = btrfs_start_transaction(root, 3);
 374	if (IS_ERR(trans))
 375		return PTR_ERR(trans);
 376
 377	if (comp) {
 378		ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
 379				     strlen(comp), 0);
 380		if (ret) {
 381			btrfs_abort_transaction(trans, ret);
 382			goto out_end_trans;
 383		}
 384	} else {
 385		ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
 386				     0, 0);
 387		if (ret && ret != -ENODATA) {
 388			btrfs_abort_transaction(trans, ret);
 389			goto out_end_trans;
 390		}
 391	}
 392
 393update_flags:
 394	binode->flags = binode_flags;
 395	btrfs_sync_inode_flags_to_i_flags(inode);
 396	inode_inc_iversion(inode);
 397	inode_set_ctime_current(inode);
 398	ret = btrfs_update_inode(trans, BTRFS_I(inode));
 399
 400 out_end_trans:
 401	btrfs_end_transaction(trans);
 402	return ret;
 403}
 404
 405/*
 406 * Start exclusive operation @type, return true on success
 407 */
 408bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
 409			enum btrfs_exclusive_operation type)
 410{
 411	bool ret = false;
 412
 413	spin_lock(&fs_info->super_lock);
 414	if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
 415		fs_info->exclusive_operation = type;
 416		ret = true;
 417	}
 418	spin_unlock(&fs_info->super_lock);
 419
 420	return ret;
 421}
 422
 423/*
 424 * Conditionally allow to enter the exclusive operation in case it's compatible
 425 * with the running one.  This must be paired with btrfs_exclop_start_unlock and
 426 * btrfs_exclop_finish.
 427 *
 428 * Compatibility:
 429 * - the same type is already running
 430 * - when trying to add a device and balance has been paused
 431 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
 432 *   must check the condition first that would allow none -> @type
 433 */
 434bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
 435				 enum btrfs_exclusive_operation type)
 436{
 437	spin_lock(&fs_info->super_lock);
 438	if (fs_info->exclusive_operation == type ||
 439	    (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED &&
 440	     type == BTRFS_EXCLOP_DEV_ADD))
 441		return true;
 442
 443	spin_unlock(&fs_info->super_lock);
 444	return false;
 445}
 446
 447void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
 448{
 449	spin_unlock(&fs_info->super_lock);
 450}
 451
 452void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
 453{
 454	spin_lock(&fs_info->super_lock);
 455	WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
 456	spin_unlock(&fs_info->super_lock);
 457	sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
 458}
 459
 460void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
 461			  enum btrfs_exclusive_operation op)
 462{
 463	switch (op) {
 464	case BTRFS_EXCLOP_BALANCE_PAUSED:
 465		spin_lock(&fs_info->super_lock);
 466		ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE ||
 467		       fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD ||
 468		       fs_info->exclusive_operation == BTRFS_EXCLOP_NONE ||
 469		       fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
 470		fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
 471		spin_unlock(&fs_info->super_lock);
 472		break;
 473	case BTRFS_EXCLOP_BALANCE:
 474		spin_lock(&fs_info->super_lock);
 475		ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
 476		fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
 477		spin_unlock(&fs_info->super_lock);
 478		break;
 479	default:
 480		btrfs_warn(fs_info,
 481			"invalid exclop balance operation %d requested", op);
 482	}
 483}
 484
 485static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
 486{
 487	return put_user(inode->i_generation, arg);
 488}
 489
 490static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
 491					void __user *arg)
 492{
 493	struct btrfs_device *device;
 494	struct fstrim_range range;
 495	u64 minlen = ULLONG_MAX;
 496	u64 num_devices = 0;
 497	int ret;
 498
 499	if (!capable(CAP_SYS_ADMIN))
 500		return -EPERM;
 501
 502	/*
 503	 * btrfs_trim_block_group() depends on space cache, which is not
 504	 * available in zoned filesystem. So, disallow fitrim on a zoned
 505	 * filesystem for now.
 506	 */
 507	if (btrfs_is_zoned(fs_info))
 508		return -EOPNOTSUPP;
 509
 510	/*
 511	 * If the fs is mounted with nologreplay, which requires it to be
 512	 * mounted in RO mode as well, we can not allow discard on free space
 513	 * inside block groups, because log trees refer to extents that are not
 514	 * pinned in a block group's free space cache (pinning the extents is
 515	 * precisely the first phase of replaying a log tree).
 516	 */
 517	if (btrfs_test_opt(fs_info, NOLOGREPLAY))
 518		return -EROFS;
 519
 520	rcu_read_lock();
 521	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
 522				dev_list) {
 523		if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
 524			continue;
 525		num_devices++;
 526		minlen = min_t(u64, bdev_discard_granularity(device->bdev),
 527				    minlen);
 528	}
 529	rcu_read_unlock();
 530
 531	if (!num_devices)
 532		return -EOPNOTSUPP;
 533	if (copy_from_user(&range, arg, sizeof(range)))
 534		return -EFAULT;
 535
 536	/*
 537	 * NOTE: Don't truncate the range using super->total_bytes.  Bytenr of
 538	 * block group is in the logical address space, which can be any
 539	 * sectorsize aligned bytenr in  the range [0, U64_MAX].
 540	 */
 541	if (range.len < fs_info->sectorsize)
 542		return -EINVAL;
 543
 544	range.minlen = max(range.minlen, minlen);
 545	ret = btrfs_trim_fs(fs_info, &range);
 546	if (ret < 0)
 547		return ret;
 548
 549	if (copy_to_user(arg, &range, sizeof(range)))
 550		return -EFAULT;
 551
 552	return 0;
 553}
 554
 555int __pure btrfs_is_empty_uuid(u8 *uuid)
 556{
 557	int i;
 558
 559	for (i = 0; i < BTRFS_UUID_SIZE; i++) {
 560		if (uuid[i])
 561			return 0;
 562	}
 563	return 1;
 564}
 565
 566/*
 567 * Calculate the number of transaction items to reserve for creating a subvolume
 568 * or snapshot, not including the inode, directory entries, or parent directory.
 569 */
 570static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
 571{
 572	/*
 573	 * 1 to add root block
 574	 * 1 to add root item
 575	 * 1 to add root ref
 576	 * 1 to add root backref
 577	 * 1 to add UUID item
 578	 * 1 to add qgroup info
 579	 * 1 to add qgroup limit
 580	 *
 581	 * Ideally the last two would only be accounted if qgroups are enabled,
 582	 * but that can change between now and the time we would insert them.
 583	 */
 584	unsigned int num_items = 7;
 585
 586	if (inherit) {
 587		/* 2 to add qgroup relations for each inherited qgroup */
 588		num_items += 2 * inherit->num_qgroups;
 589	}
 590	return num_items;
 591}
 592
 593static noinline int create_subvol(struct mnt_idmap *idmap,
 594				  struct inode *dir, struct dentry *dentry,
 595				  struct btrfs_qgroup_inherit *inherit)
 596{
 597	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
 598	struct btrfs_trans_handle *trans;
 599	struct btrfs_key key;
 600	struct btrfs_root_item *root_item;
 601	struct btrfs_inode_item *inode_item;
 602	struct extent_buffer *leaf;
 603	struct btrfs_root *root = BTRFS_I(dir)->root;
 604	struct btrfs_root *new_root;
 605	struct btrfs_block_rsv block_rsv;
 606	struct timespec64 cur_time = current_time(dir);
 607	struct btrfs_new_inode_args new_inode_args = {
 608		.dir = dir,
 609		.dentry = dentry,
 610		.subvol = true,
 611	};
 612	unsigned int trans_num_items;
 613	int ret;
 614	dev_t anon_dev;
 615	u64 objectid;
 616	u64 qgroup_reserved = 0;
 617
 618	root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
 619	if (!root_item)
 620		return -ENOMEM;
 621
 622	ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
 623	if (ret)
 624		goto out_root_item;
 625
 626	/*
 627	 * Don't create subvolume whose level is not zero. Or qgroup will be
 628	 * screwed up since it assumes subvolume qgroup's level to be 0.
 629	 */
 630	if (btrfs_qgroup_level(objectid)) {
 631		ret = -ENOSPC;
 632		goto out_root_item;
 633	}
 634
 635	ret = get_anon_bdev(&anon_dev);
 636	if (ret < 0)
 637		goto out_root_item;
 638
 639	new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir);
 640	if (!new_inode_args.inode) {
 641		ret = -ENOMEM;
 642		goto out_anon_dev;
 643	}
 644	ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
 645	if (ret)
 646		goto out_inode;
 647	trans_num_items += create_subvol_num_items(inherit);
 648
 649	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
 650	ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
 651					       trans_num_items, false);
 652	if (ret)
 653		goto out_new_inode_args;
 654	qgroup_reserved = block_rsv.qgroup_rsv_reserved;
 655
 656	trans = btrfs_start_transaction(root, 0);
 657	if (IS_ERR(trans)) {
 658		ret = PTR_ERR(trans);
 659		goto out_release_rsv;
 660	}
 661	ret = btrfs_record_root_in_trans(trans, BTRFS_I(dir)->root);
 662	if (ret)
 663		goto out;
 664	btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved);
 665	qgroup_reserved = 0;
 666	trans->block_rsv = &block_rsv;
 667	trans->bytes_reserved = block_rsv.size;
 668	/* Tree log can't currently deal with an inode which is a new root. */
 669	btrfs_set_log_full_commit(trans);
 670
 671	ret = btrfs_qgroup_inherit(trans, 0, objectid, root->root_key.objectid, inherit);
 672	if (ret)
 673		goto out;
 674
 675	leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
 676				      0, BTRFS_NESTING_NORMAL);
 677	if (IS_ERR(leaf)) {
 678		ret = PTR_ERR(leaf);
 679		goto out;
 680	}
 681
 682	btrfs_mark_buffer_dirty(trans, leaf);
 683
 684	inode_item = &root_item->inode;
 685	btrfs_set_stack_inode_generation(inode_item, 1);
 686	btrfs_set_stack_inode_size(inode_item, 3);
 687	btrfs_set_stack_inode_nlink(inode_item, 1);
 688	btrfs_set_stack_inode_nbytes(inode_item,
 689				     fs_info->nodesize);
 690	btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
 691
 692	btrfs_set_root_flags(root_item, 0);
 693	btrfs_set_root_limit(root_item, 0);
 694	btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
 695
 696	btrfs_set_root_bytenr(root_item, leaf->start);
 697	btrfs_set_root_generation(root_item, trans->transid);
 698	btrfs_set_root_level(root_item, 0);
 699	btrfs_set_root_refs(root_item, 1);
 700	btrfs_set_root_used(root_item, leaf->len);
 701	btrfs_set_root_last_snapshot(root_item, 0);
 702
 703	btrfs_set_root_generation_v2(root_item,
 704			btrfs_root_generation(root_item));
 705	generate_random_guid(root_item->uuid);
 706	btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
 707	btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
 708	root_item->ctime = root_item->otime;
 709	btrfs_set_root_ctransid(root_item, trans->transid);
 710	btrfs_set_root_otransid(root_item, trans->transid);
 711
 712	btrfs_tree_unlock(leaf);
 713
 714	btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
 715
 716	key.objectid = objectid;
 717	key.offset = 0;
 718	key.type = BTRFS_ROOT_ITEM_KEY;
 719	ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
 720				root_item);
 721	if (ret) {
 
 
 722		/*
 723		 * Since we don't abort the transaction in this case, free the
 724		 * tree block so that we don't leak space and leave the
 725		 * filesystem in an inconsistent state (an extent item in the
 726		 * extent tree with a backreference for a root that does not
 727		 * exists).
 728		 */
 729		btrfs_tree_lock(leaf);
 730		btrfs_clear_buffer_dirty(trans, leaf);
 731		btrfs_tree_unlock(leaf);
 732		btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
 
 
 733		free_extent_buffer(leaf);
 734		goto out;
 735	}
 736
 737	free_extent_buffer(leaf);
 738	leaf = NULL;
 739
 740	new_root = btrfs_get_new_fs_root(fs_info, objectid, &anon_dev);
 741	if (IS_ERR(new_root)) {
 742		ret = PTR_ERR(new_root);
 743		btrfs_abort_transaction(trans, ret);
 744		goto out;
 745	}
 746	/* anon_dev is owned by new_root now. */
 747	anon_dev = 0;
 748	BTRFS_I(new_inode_args.inode)->root = new_root;
 749	/* ... and new_root is owned by new_inode_args.inode now. */
 750
 751	ret = btrfs_record_root_in_trans(trans, new_root);
 752	if (ret) {
 753		btrfs_abort_transaction(trans, ret);
 754		goto out;
 755	}
 756
 757	ret = btrfs_uuid_tree_add(trans, root_item->uuid,
 758				  BTRFS_UUID_KEY_SUBVOL, objectid);
 759	if (ret) {
 760		btrfs_abort_transaction(trans, ret);
 761		goto out;
 762	}
 763
 764	ret = btrfs_create_new_inode(trans, &new_inode_args);
 765	if (ret) {
 766		btrfs_abort_transaction(trans, ret);
 767		goto out;
 768	}
 769
 
 
 770	d_instantiate_new(dentry, new_inode_args.inode);
 771	new_inode_args.inode = NULL;
 772
 773out:
 774	trans->block_rsv = NULL;
 775	trans->bytes_reserved = 0;
 776	btrfs_end_transaction(trans);
 777out_release_rsv:
 778	btrfs_block_rsv_release(fs_info, &block_rsv, (u64)-1, NULL);
 779	if (qgroup_reserved)
 780		btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved);
 781out_new_inode_args:
 782	btrfs_new_inode_args_destroy(&new_inode_args);
 783out_inode:
 784	iput(new_inode_args.inode);
 785out_anon_dev:
 786	if (anon_dev)
 787		free_anon_bdev(anon_dev);
 788out_root_item:
 789	kfree(root_item);
 790	return ret;
 791}
 792
 793static int create_snapshot(struct btrfs_root *root, struct inode *dir,
 794			   struct dentry *dentry, bool readonly,
 795			   struct btrfs_qgroup_inherit *inherit)
 796{
 797	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
 798	struct inode *inode;
 799	struct btrfs_pending_snapshot *pending_snapshot;
 800	unsigned int trans_num_items;
 801	struct btrfs_trans_handle *trans;
 802	struct btrfs_block_rsv *block_rsv;
 803	u64 qgroup_reserved = 0;
 804	int ret;
 805
 806	/* We do not support snapshotting right now. */
 807	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
 808		btrfs_warn(fs_info,
 809			   "extent tree v2 doesn't support snapshotting yet");
 810		return -EOPNOTSUPP;
 811	}
 812
 813	if (btrfs_root_refs(&root->root_item) == 0)
 814		return -ENOENT;
 815
 816	if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
 817		return -EINVAL;
 818
 819	if (atomic_read(&root->nr_swapfiles)) {
 820		btrfs_warn(fs_info,
 821			   "cannot snapshot subvolume with active swapfile");
 822		return -ETXTBSY;
 823	}
 824
 825	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
 826	if (!pending_snapshot)
 827		return -ENOMEM;
 828
 829	ret = get_anon_bdev(&pending_snapshot->anon_dev);
 830	if (ret < 0)
 831		goto free_pending;
 832	pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
 833			GFP_KERNEL);
 834	pending_snapshot->path = btrfs_alloc_path();
 835	if (!pending_snapshot->root_item || !pending_snapshot->path) {
 836		ret = -ENOMEM;
 837		goto free_pending;
 838	}
 839
 840	block_rsv = &pending_snapshot->block_rsv;
 841	btrfs_init_block_rsv(block_rsv, BTRFS_BLOCK_RSV_TEMP);
 842	/*
 843	 * 1 to add dir item
 844	 * 1 to add dir index
 845	 * 1 to update parent inode item
 846	 */
 847	trans_num_items = create_subvol_num_items(inherit) + 3;
 848	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, block_rsv,
 849					       trans_num_items, false);
 850	if (ret)
 851		goto free_pending;
 852	qgroup_reserved = block_rsv->qgroup_rsv_reserved;
 853
 854	pending_snapshot->dentry = dentry;
 855	pending_snapshot->root = root;
 856	pending_snapshot->readonly = readonly;
 857	pending_snapshot->dir = dir;
 858	pending_snapshot->inherit = inherit;
 859
 860	trans = btrfs_start_transaction(root, 0);
 861	if (IS_ERR(trans)) {
 862		ret = PTR_ERR(trans);
 863		goto fail;
 864	}
 865	ret = btrfs_record_root_in_trans(trans, BTRFS_I(dir)->root);
 866	if (ret) {
 867		btrfs_end_transaction(trans);
 868		goto fail;
 869	}
 870	btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved);
 871	qgroup_reserved = 0;
 872
 873	trans->pending_snapshot = pending_snapshot;
 874
 875	ret = btrfs_commit_transaction(trans);
 876	if (ret)
 877		goto fail;
 878
 879	ret = pending_snapshot->error;
 880	if (ret)
 881		goto fail;
 882
 883	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
 884	if (ret)
 885		goto fail;
 886
 887	inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
 888	if (IS_ERR(inode)) {
 889		ret = PTR_ERR(inode);
 890		goto fail;
 891	}
 892
 893	d_instantiate(dentry, inode);
 894	ret = 0;
 895	pending_snapshot->anon_dev = 0;
 896fail:
 897	/* Prevent double freeing of anon_dev */
 898	if (ret && pending_snapshot->snap)
 899		pending_snapshot->snap->anon_dev = 0;
 900	btrfs_put_root(pending_snapshot->snap);
 901	btrfs_block_rsv_release(fs_info, block_rsv, (u64)-1, NULL);
 902	if (qgroup_reserved)
 903		btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved);
 904free_pending:
 905	if (pending_snapshot->anon_dev)
 906		free_anon_bdev(pending_snapshot->anon_dev);
 907	kfree(pending_snapshot->root_item);
 908	btrfs_free_path(pending_snapshot->path);
 909	kfree(pending_snapshot);
 910
 911	return ret;
 912}
 913
 914/*  copy of may_delete in fs/namei.c()
 915 *	Check whether we can remove a link victim from directory dir, check
 916 *  whether the type of victim is right.
 917 *  1. We can't do it if dir is read-only (done in permission())
 918 *  2. We should have write and exec permissions on dir
 919 *  3. We can't remove anything from append-only dir
 920 *  4. We can't do anything with immutable dir (done in permission())
 921 *  5. If the sticky bit on dir is set we should either
 922 *	a. be owner of dir, or
 923 *	b. be owner of victim, or
 924 *	c. have CAP_FOWNER capability
 925 *  6. If the victim is append-only or immutable we can't do anything with
 926 *     links pointing to it.
 927 *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
 928 *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
 929 *  9. We can't remove a root or mountpoint.
 930 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
 931 *     nfs_async_unlink().
 932 */
 933
 934static int btrfs_may_delete(struct mnt_idmap *idmap,
 935			    struct inode *dir, struct dentry *victim, int isdir)
 936{
 937	int error;
 938
 939	if (d_really_is_negative(victim))
 940		return -ENOENT;
 941
 942	/* The @victim is not inside @dir. */
 943	if (d_inode(victim->d_parent) != dir)
 944		return -EINVAL;
 945	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
 946
 947	error = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
 948	if (error)
 949		return error;
 950	if (IS_APPEND(dir))
 951		return -EPERM;
 952	if (check_sticky(idmap, dir, d_inode(victim)) ||
 953	    IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
 954	    IS_SWAPFILE(d_inode(victim)))
 955		return -EPERM;
 956	if (isdir) {
 957		if (!d_is_dir(victim))
 958			return -ENOTDIR;
 959		if (IS_ROOT(victim))
 960			return -EBUSY;
 961	} else if (d_is_dir(victim))
 962		return -EISDIR;
 963	if (IS_DEADDIR(dir))
 964		return -ENOENT;
 965	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
 966		return -EBUSY;
 967	return 0;
 968}
 969
 970/* copy of may_create in fs/namei.c() */
 971static inline int btrfs_may_create(struct mnt_idmap *idmap,
 972				   struct inode *dir, struct dentry *child)
 973{
 974	if (d_really_is_positive(child))
 975		return -EEXIST;
 976	if (IS_DEADDIR(dir))
 977		return -ENOENT;
 978	if (!fsuidgid_has_mapping(dir->i_sb, idmap))
 979		return -EOVERFLOW;
 980	return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
 981}
 982
 983/*
 984 * Create a new subvolume below @parent.  This is largely modeled after
 985 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 986 * inside this filesystem so it's quite a bit simpler.
 987 */
 988static noinline int btrfs_mksubvol(const struct path *parent,
 989				   struct mnt_idmap *idmap,
 990				   const char *name, int namelen,
 991				   struct btrfs_root *snap_src,
 992				   bool readonly,
 993				   struct btrfs_qgroup_inherit *inherit)
 994{
 995	struct inode *dir = d_inode(parent->dentry);
 996	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
 997	struct dentry *dentry;
 998	struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
 999	int error;
1000
1001	error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
1002	if (error == -EINTR)
1003		return error;
1004
1005	dentry = lookup_one(idmap, name, parent->dentry, namelen);
1006	error = PTR_ERR(dentry);
1007	if (IS_ERR(dentry))
1008		goto out_unlock;
1009
1010	error = btrfs_may_create(idmap, dir, dentry);
1011	if (error)
1012		goto out_dput;
1013
1014	/*
1015	 * even if this name doesn't exist, we may get hash collisions.
1016	 * check for them now when we can safely fail
1017	 */
1018	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
1019					       dir->i_ino, &name_str);
1020	if (error)
1021		goto out_dput;
1022
1023	down_read(&fs_info->subvol_sem);
1024
1025	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
1026		goto out_up_read;
1027
1028	if (snap_src)
1029		error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
1030	else
1031		error = create_subvol(idmap, dir, dentry, inherit);
1032
1033	if (!error)
1034		fsnotify_mkdir(dir, dentry);
1035out_up_read:
1036	up_read(&fs_info->subvol_sem);
1037out_dput:
1038	dput(dentry);
1039out_unlock:
1040	btrfs_inode_unlock(BTRFS_I(dir), 0);
1041	return error;
1042}
1043
1044static noinline int btrfs_mksnapshot(const struct path *parent,
1045				   struct mnt_idmap *idmap,
1046				   const char *name, int namelen,
1047				   struct btrfs_root *root,
1048				   bool readonly,
1049				   struct btrfs_qgroup_inherit *inherit)
1050{
1051	int ret;
1052	bool snapshot_force_cow = false;
1053
1054	/*
1055	 * Force new buffered writes to reserve space even when NOCOW is
1056	 * possible. This is to avoid later writeback (running dealloc) to
1057	 * fallback to COW mode and unexpectedly fail with ENOSPC.
1058	 */
1059	btrfs_drew_read_lock(&root->snapshot_lock);
1060
1061	ret = btrfs_start_delalloc_snapshot(root, false);
1062	if (ret)
1063		goto out;
1064
1065	/*
1066	 * All previous writes have started writeback in NOCOW mode, so now
1067	 * we force future writes to fallback to COW mode during snapshot
1068	 * creation.
1069	 */
1070	atomic_inc(&root->snapshot_force_cow);
1071	snapshot_force_cow = true;
1072
1073	btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1074
1075	ret = btrfs_mksubvol(parent, idmap, name, namelen,
1076			     root, readonly, inherit);
 
1077out:
1078	if (snapshot_force_cow)
1079		atomic_dec(&root->snapshot_force_cow);
1080	btrfs_drew_read_unlock(&root->snapshot_lock);
1081	return ret;
1082}
1083
1084/*
1085 * Try to start exclusive operation @type or cancel it if it's running.
1086 *
1087 * Return:
1088 *   0        - normal mode, newly claimed op started
1089 *  >0        - normal mode, something else is running,
1090 *              return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1091 * ECANCELED  - cancel mode, successful cancel
1092 * ENOTCONN   - cancel mode, operation not running anymore
1093 */
1094static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1095			enum btrfs_exclusive_operation type, bool cancel)
1096{
1097	if (!cancel) {
1098		/* Start normal op */
1099		if (!btrfs_exclop_start(fs_info, type))
1100			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1101		/* Exclusive operation is now claimed */
1102		return 0;
1103	}
1104
1105	/* Cancel running op */
1106	if (btrfs_exclop_start_try_lock(fs_info, type)) {
1107		/*
1108		 * This blocks any exclop finish from setting it to NONE, so we
1109		 * request cancellation. Either it runs and we will wait for it,
1110		 * or it has finished and no waiting will happen.
1111		 */
1112		atomic_inc(&fs_info->reloc_cancel_req);
1113		btrfs_exclop_start_unlock(fs_info);
1114
1115		if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1116			wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1117				    TASK_INTERRUPTIBLE);
1118
1119		return -ECANCELED;
1120	}
1121
1122	/* Something else is running or none */
1123	return -ENOTCONN;
1124}
1125
1126static noinline int btrfs_ioctl_resize(struct file *file,
1127					void __user *arg)
1128{
1129	BTRFS_DEV_LOOKUP_ARGS(args);
1130	struct inode *inode = file_inode(file);
1131	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1132	u64 new_size;
1133	u64 old_size;
1134	u64 devid = 1;
1135	struct btrfs_root *root = BTRFS_I(inode)->root;
1136	struct btrfs_ioctl_vol_args *vol_args;
1137	struct btrfs_trans_handle *trans;
1138	struct btrfs_device *device = NULL;
1139	char *sizestr;
1140	char *retptr;
1141	char *devstr = NULL;
1142	int ret = 0;
1143	int mod = 0;
1144	bool cancel;
1145
1146	if (!capable(CAP_SYS_ADMIN))
1147		return -EPERM;
1148
1149	ret = mnt_want_write_file(file);
1150	if (ret)
1151		return ret;
1152
1153	/*
1154	 * Read the arguments before checking exclusivity to be able to
1155	 * distinguish regular resize and cancel
1156	 */
1157	vol_args = memdup_user(arg, sizeof(*vol_args));
1158	if (IS_ERR(vol_args)) {
1159		ret = PTR_ERR(vol_args);
1160		goto out_drop;
1161	}
1162	ret = btrfs_check_ioctl_vol_args_path(vol_args);
1163	if (ret < 0)
1164		goto out_free;
1165
1166	sizestr = vol_args->name;
1167	cancel = (strcmp("cancel", sizestr) == 0);
1168	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1169	if (ret)
1170		goto out_free;
1171	/* Exclusive operation is now claimed */
1172
1173	devstr = strchr(sizestr, ':');
1174	if (devstr) {
1175		sizestr = devstr + 1;
1176		*devstr = '\0';
1177		devstr = vol_args->name;
1178		ret = kstrtoull(devstr, 10, &devid);
1179		if (ret)
1180			goto out_finish;
1181		if (!devid) {
1182			ret = -EINVAL;
1183			goto out_finish;
1184		}
1185		btrfs_info(fs_info, "resizing devid %llu", devid);
1186	}
1187
1188	args.devid = devid;
1189	device = btrfs_find_device(fs_info->fs_devices, &args);
1190	if (!device) {
1191		btrfs_info(fs_info, "resizer unable to find device %llu",
1192			   devid);
1193		ret = -ENODEV;
1194		goto out_finish;
1195	}
1196
1197	if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1198		btrfs_info(fs_info,
1199			   "resizer unable to apply on readonly device %llu",
1200		       devid);
1201		ret = -EPERM;
1202		goto out_finish;
1203	}
1204
1205	if (!strcmp(sizestr, "max"))
1206		new_size = bdev_nr_bytes(device->bdev);
1207	else {
1208		if (sizestr[0] == '-') {
1209			mod = -1;
1210			sizestr++;
1211		} else if (sizestr[0] == '+') {
1212			mod = 1;
1213			sizestr++;
1214		}
1215		new_size = memparse(sizestr, &retptr);
1216		if (*retptr != '\0' || new_size == 0) {
1217			ret = -EINVAL;
1218			goto out_finish;
1219		}
1220	}
1221
1222	if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1223		ret = -EPERM;
1224		goto out_finish;
1225	}
1226
1227	old_size = btrfs_device_get_total_bytes(device);
1228
1229	if (mod < 0) {
1230		if (new_size > old_size) {
1231			ret = -EINVAL;
1232			goto out_finish;
1233		}
1234		new_size = old_size - new_size;
1235	} else if (mod > 0) {
1236		if (new_size > ULLONG_MAX - old_size) {
1237			ret = -ERANGE;
1238			goto out_finish;
1239		}
1240		new_size = old_size + new_size;
1241	}
1242
1243	if (new_size < SZ_256M) {
1244		ret = -EINVAL;
1245		goto out_finish;
1246	}
1247	if (new_size > bdev_nr_bytes(device->bdev)) {
1248		ret = -EFBIG;
1249		goto out_finish;
1250	}
1251
1252	new_size = round_down(new_size, fs_info->sectorsize);
1253
1254	if (new_size > old_size) {
1255		trans = btrfs_start_transaction(root, 0);
1256		if (IS_ERR(trans)) {
1257			ret = PTR_ERR(trans);
1258			goto out_finish;
1259		}
1260		ret = btrfs_grow_device(trans, device, new_size);
1261		btrfs_commit_transaction(trans);
1262	} else if (new_size < old_size) {
1263		ret = btrfs_shrink_device(device, new_size);
1264	} /* equal, nothing need to do */
1265
1266	if (ret == 0 && new_size != old_size)
1267		btrfs_info_in_rcu(fs_info,
1268			"resize device %s (devid %llu) from %llu to %llu",
1269			btrfs_dev_name(device), device->devid,
1270			old_size, new_size);
1271out_finish:
1272	btrfs_exclop_finish(fs_info);
1273out_free:
1274	kfree(vol_args);
1275out_drop:
1276	mnt_drop_write_file(file);
1277	return ret;
1278}
1279
1280static noinline int __btrfs_ioctl_snap_create(struct file *file,
1281				struct mnt_idmap *idmap,
1282				const char *name, unsigned long fd, int subvol,
1283				bool readonly,
1284				struct btrfs_qgroup_inherit *inherit)
1285{
1286	int namelen;
1287	int ret = 0;
1288
1289	if (!S_ISDIR(file_inode(file)->i_mode))
1290		return -ENOTDIR;
1291
1292	ret = mnt_want_write_file(file);
1293	if (ret)
1294		goto out;
1295
1296	namelen = strlen(name);
1297	if (strchr(name, '/')) {
1298		ret = -EINVAL;
1299		goto out_drop_write;
1300	}
1301
1302	if (name[0] == '.' &&
1303	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1304		ret = -EEXIST;
1305		goto out_drop_write;
1306	}
1307
1308	if (subvol) {
1309		ret = btrfs_mksubvol(&file->f_path, idmap, name,
1310				     namelen, NULL, readonly, inherit);
1311	} else {
1312		struct fd src = fdget(fd);
1313		struct inode *src_inode;
1314		if (!src.file) {
1315			ret = -EINVAL;
1316			goto out_drop_write;
1317		}
1318
1319		src_inode = file_inode(src.file);
1320		if (src_inode->i_sb != file_inode(file)->i_sb) {
1321			btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1322				   "Snapshot src from another FS");
1323			ret = -EXDEV;
1324		} else if (!inode_owner_or_capable(idmap, src_inode)) {
1325			/*
1326			 * Subvolume creation is not restricted, but snapshots
1327			 * are limited to own subvolumes only
1328			 */
1329			ret = -EPERM;
1330		} else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1331			/*
1332			 * Snapshots must be made with the src_inode referring
1333			 * to the subvolume inode, otherwise the permission
1334			 * checking above is useless because we may have
1335			 * permission on a lower directory but not the subvol
1336			 * itself.
1337			 */
1338			ret = -EINVAL;
1339		} else {
1340			ret = btrfs_mksnapshot(&file->f_path, idmap,
1341					       name, namelen,
1342					       BTRFS_I(src_inode)->root,
1343					       readonly, inherit);
1344		}
1345		fdput(src);
1346	}
1347out_drop_write:
1348	mnt_drop_write_file(file);
1349out:
1350	return ret;
1351}
1352
1353static noinline int btrfs_ioctl_snap_create(struct file *file,
1354					    void __user *arg, int subvol)
1355{
1356	struct btrfs_ioctl_vol_args *vol_args;
1357	int ret;
1358
1359	if (!S_ISDIR(file_inode(file)->i_mode))
1360		return -ENOTDIR;
1361
1362	vol_args = memdup_user(arg, sizeof(*vol_args));
1363	if (IS_ERR(vol_args))
1364		return PTR_ERR(vol_args);
1365	ret = btrfs_check_ioctl_vol_args_path(vol_args);
1366	if (ret < 0)
1367		goto out;
1368
1369	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1370					vol_args->name, vol_args->fd, subvol,
1371					false, NULL);
1372
1373out:
1374	kfree(vol_args);
1375	return ret;
1376}
1377
1378static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1379					       void __user *arg, int subvol)
1380{
1381	struct btrfs_ioctl_vol_args_v2 *vol_args;
1382	int ret;
1383	bool readonly = false;
1384	struct btrfs_qgroup_inherit *inherit = NULL;
1385
1386	if (!S_ISDIR(file_inode(file)->i_mode))
1387		return -ENOTDIR;
1388
1389	vol_args = memdup_user(arg, sizeof(*vol_args));
1390	if (IS_ERR(vol_args))
1391		return PTR_ERR(vol_args);
1392	ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args);
1393	if (ret < 0)
1394		goto free_args;
1395
1396	if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1397		ret = -EOPNOTSUPP;
1398		goto free_args;
1399	}
1400
1401	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1402		readonly = true;
1403	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1404		struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file));
1405
1406		if (vol_args->size < sizeof(*inherit) ||
1407		    vol_args->size > PAGE_SIZE) {
1408			ret = -EINVAL;
1409			goto free_args;
1410		}
1411		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1412		if (IS_ERR(inherit)) {
1413			ret = PTR_ERR(inherit);
1414			goto free_args;
1415		}
1416
1417		ret = btrfs_qgroup_check_inherit(fs_info, inherit, vol_args->size);
1418		if (ret < 0)
1419			goto free_inherit;
1420	}
1421
1422	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1423					vol_args->name, vol_args->fd, subvol,
1424					readonly, inherit);
1425	if (ret)
1426		goto free_inherit;
1427free_inherit:
1428	kfree(inherit);
1429free_args:
1430	kfree(vol_args);
1431	return ret;
1432}
1433
1434static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1435						void __user *arg)
1436{
1437	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1438	struct btrfs_root *root = BTRFS_I(inode)->root;
1439	int ret = 0;
1440	u64 flags = 0;
1441
1442	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1443		return -EINVAL;
1444
1445	down_read(&fs_info->subvol_sem);
1446	if (btrfs_root_readonly(root))
1447		flags |= BTRFS_SUBVOL_RDONLY;
1448	up_read(&fs_info->subvol_sem);
1449
1450	if (copy_to_user(arg, &flags, sizeof(flags)))
1451		ret = -EFAULT;
1452
1453	return ret;
1454}
1455
1456static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1457					      void __user *arg)
1458{
1459	struct inode *inode = file_inode(file);
1460	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1461	struct btrfs_root *root = BTRFS_I(inode)->root;
1462	struct btrfs_trans_handle *trans;
1463	u64 root_flags;
1464	u64 flags;
1465	int ret = 0;
1466
1467	if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1468		return -EPERM;
1469
1470	ret = mnt_want_write_file(file);
1471	if (ret)
1472		goto out;
1473
1474	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1475		ret = -EINVAL;
1476		goto out_drop_write;
1477	}
1478
1479	if (copy_from_user(&flags, arg, sizeof(flags))) {
1480		ret = -EFAULT;
1481		goto out_drop_write;
1482	}
1483
1484	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1485		ret = -EOPNOTSUPP;
1486		goto out_drop_write;
1487	}
1488
1489	down_write(&fs_info->subvol_sem);
1490
1491	/* nothing to do */
1492	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1493		goto out_drop_sem;
1494
1495	root_flags = btrfs_root_flags(&root->root_item);
1496	if (flags & BTRFS_SUBVOL_RDONLY) {
1497		btrfs_set_root_flags(&root->root_item,
1498				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1499	} else {
1500		/*
1501		 * Block RO -> RW transition if this subvolume is involved in
1502		 * send
1503		 */
1504		spin_lock(&root->root_item_lock);
1505		if (root->send_in_progress == 0) {
1506			btrfs_set_root_flags(&root->root_item,
1507				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1508			spin_unlock(&root->root_item_lock);
1509		} else {
1510			spin_unlock(&root->root_item_lock);
1511			btrfs_warn(fs_info,
1512				   "Attempt to set subvolume %llu read-write during send",
1513				   root->root_key.objectid);
1514			ret = -EPERM;
1515			goto out_drop_sem;
1516		}
1517	}
1518
1519	trans = btrfs_start_transaction(root, 1);
1520	if (IS_ERR(trans)) {
1521		ret = PTR_ERR(trans);
1522		goto out_reset;
1523	}
1524
1525	ret = btrfs_update_root(trans, fs_info->tree_root,
1526				&root->root_key, &root->root_item);
1527	if (ret < 0) {
1528		btrfs_end_transaction(trans);
1529		goto out_reset;
1530	}
1531
1532	ret = btrfs_commit_transaction(trans);
1533
1534out_reset:
1535	if (ret)
1536		btrfs_set_root_flags(&root->root_item, root_flags);
1537out_drop_sem:
1538	up_write(&fs_info->subvol_sem);
1539out_drop_write:
1540	mnt_drop_write_file(file);
1541out:
1542	return ret;
1543}
1544
1545static noinline int key_in_sk(struct btrfs_key *key,
1546			      struct btrfs_ioctl_search_key *sk)
1547{
1548	struct btrfs_key test;
1549	int ret;
1550
1551	test.objectid = sk->min_objectid;
1552	test.type = sk->min_type;
1553	test.offset = sk->min_offset;
1554
1555	ret = btrfs_comp_cpu_keys(key, &test);
1556	if (ret < 0)
1557		return 0;
1558
1559	test.objectid = sk->max_objectid;
1560	test.type = sk->max_type;
1561	test.offset = sk->max_offset;
1562
1563	ret = btrfs_comp_cpu_keys(key, &test);
1564	if (ret > 0)
1565		return 0;
1566	return 1;
1567}
1568
1569static noinline int copy_to_sk(struct btrfs_path *path,
1570			       struct btrfs_key *key,
1571			       struct btrfs_ioctl_search_key *sk,
1572			       u64 *buf_size,
1573			       char __user *ubuf,
1574			       unsigned long *sk_offset,
1575			       int *num_found)
1576{
1577	u64 found_transid;
1578	struct extent_buffer *leaf;
1579	struct btrfs_ioctl_search_header sh;
1580	struct btrfs_key test;
1581	unsigned long item_off;
1582	unsigned long item_len;
1583	int nritems;
1584	int i;
1585	int slot;
1586	int ret = 0;
1587
1588	leaf = path->nodes[0];
1589	slot = path->slots[0];
1590	nritems = btrfs_header_nritems(leaf);
1591
1592	if (btrfs_header_generation(leaf) > sk->max_transid) {
1593		i = nritems;
1594		goto advance_key;
1595	}
1596	found_transid = btrfs_header_generation(leaf);
1597
1598	for (i = slot; i < nritems; i++) {
1599		item_off = btrfs_item_ptr_offset(leaf, i);
1600		item_len = btrfs_item_size(leaf, i);
1601
1602		btrfs_item_key_to_cpu(leaf, key, i);
1603		if (!key_in_sk(key, sk))
1604			continue;
1605
1606		if (sizeof(sh) + item_len > *buf_size) {
1607			if (*num_found) {
1608				ret = 1;
1609				goto out;
1610			}
1611
1612			/*
1613			 * return one empty item back for v1, which does not
1614			 * handle -EOVERFLOW
1615			 */
1616
1617			*buf_size = sizeof(sh) + item_len;
1618			item_len = 0;
1619			ret = -EOVERFLOW;
1620		}
1621
1622		if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1623			ret = 1;
1624			goto out;
1625		}
1626
1627		sh.objectid = key->objectid;
1628		sh.offset = key->offset;
1629		sh.type = key->type;
1630		sh.len = item_len;
1631		sh.transid = found_transid;
1632
1633		/*
1634		 * Copy search result header. If we fault then loop again so we
1635		 * can fault in the pages and -EFAULT there if there's a
1636		 * problem. Otherwise we'll fault and then copy the buffer in
1637		 * properly this next time through
1638		 */
1639		if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1640			ret = 0;
1641			goto out;
1642		}
1643
1644		*sk_offset += sizeof(sh);
1645
1646		if (item_len) {
1647			char __user *up = ubuf + *sk_offset;
1648			/*
1649			 * Copy the item, same behavior as above, but reset the
1650			 * * sk_offset so we copy the full thing again.
1651			 */
1652			if (read_extent_buffer_to_user_nofault(leaf, up,
1653						item_off, item_len)) {
1654				ret = 0;
1655				*sk_offset -= sizeof(sh);
1656				goto out;
1657			}
1658
1659			*sk_offset += item_len;
1660		}
1661		(*num_found)++;
1662
1663		if (ret) /* -EOVERFLOW from above */
1664			goto out;
1665
1666		if (*num_found >= sk->nr_items) {
1667			ret = 1;
1668			goto out;
1669		}
1670	}
1671advance_key:
1672	ret = 0;
1673	test.objectid = sk->max_objectid;
1674	test.type = sk->max_type;
1675	test.offset = sk->max_offset;
1676	if (btrfs_comp_cpu_keys(key, &test) >= 0)
1677		ret = 1;
1678	else if (key->offset < (u64)-1)
1679		key->offset++;
1680	else if (key->type < (u8)-1) {
1681		key->offset = 0;
1682		key->type++;
1683	} else if (key->objectid < (u64)-1) {
1684		key->offset = 0;
1685		key->type = 0;
1686		key->objectid++;
1687	} else
1688		ret = 1;
1689out:
1690	/*
1691	 *  0: all items from this leaf copied, continue with next
1692	 *  1: * more items can be copied, but unused buffer is too small
1693	 *     * all items were found
1694	 *     Either way, it will stops the loop which iterates to the next
1695	 *     leaf
1696	 *  -EOVERFLOW: item was to large for buffer
1697	 *  -EFAULT: could not copy extent buffer back to userspace
1698	 */
1699	return ret;
1700}
1701
1702static noinline int search_ioctl(struct inode *inode,
1703				 struct btrfs_ioctl_search_key *sk,
1704				 u64 *buf_size,
1705				 char __user *ubuf)
1706{
1707	struct btrfs_fs_info *info = inode_to_fs_info(inode);
1708	struct btrfs_root *root;
1709	struct btrfs_key key;
1710	struct btrfs_path *path;
1711	int ret;
1712	int num_found = 0;
1713	unsigned long sk_offset = 0;
1714
1715	if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1716		*buf_size = sizeof(struct btrfs_ioctl_search_header);
1717		return -EOVERFLOW;
1718	}
1719
1720	path = btrfs_alloc_path();
1721	if (!path)
1722		return -ENOMEM;
1723
1724	if (sk->tree_id == 0) {
1725		/* search the root of the inode that was passed */
1726		root = btrfs_grab_root(BTRFS_I(inode)->root);
1727	} else {
1728		root = btrfs_get_fs_root(info, sk->tree_id, true);
1729		if (IS_ERR(root)) {
1730			btrfs_free_path(path);
1731			return PTR_ERR(root);
1732		}
1733	}
1734
1735	key.objectid = sk->min_objectid;
1736	key.type = sk->min_type;
1737	key.offset = sk->min_offset;
1738
1739	while (1) {
1740		ret = -EFAULT;
1741		/*
1742		 * Ensure that the whole user buffer is faulted in at sub-page
1743		 * granularity, otherwise the loop may live-lock.
1744		 */
1745		if (fault_in_subpage_writeable(ubuf + sk_offset,
1746					       *buf_size - sk_offset))
1747			break;
1748
1749		ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1750		if (ret != 0) {
1751			if (ret > 0)
1752				ret = 0;
1753			goto err;
1754		}
1755		ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1756				 &sk_offset, &num_found);
1757		btrfs_release_path(path);
1758		if (ret)
1759			break;
1760
1761	}
1762	if (ret > 0)
1763		ret = 0;
1764err:
1765	sk->nr_items = num_found;
1766	btrfs_put_root(root);
1767	btrfs_free_path(path);
1768	return ret;
1769}
1770
1771static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1772					    void __user *argp)
1773{
1774	struct btrfs_ioctl_search_args __user *uargs = argp;
1775	struct btrfs_ioctl_search_key sk;
1776	int ret;
1777	u64 buf_size;
1778
1779	if (!capable(CAP_SYS_ADMIN))
1780		return -EPERM;
1781
1782	if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1783		return -EFAULT;
1784
1785	buf_size = sizeof(uargs->buf);
1786
1787	ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1788
1789	/*
1790	 * In the origin implementation an overflow is handled by returning a
1791	 * search header with a len of zero, so reset ret.
1792	 */
1793	if (ret == -EOVERFLOW)
1794		ret = 0;
1795
1796	if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1797		ret = -EFAULT;
1798	return ret;
1799}
1800
1801static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1802					       void __user *argp)
1803{
1804	struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1805	struct btrfs_ioctl_search_args_v2 args;
1806	int ret;
1807	u64 buf_size;
1808	const u64 buf_limit = SZ_16M;
1809
1810	if (!capable(CAP_SYS_ADMIN))
1811		return -EPERM;
1812
1813	/* copy search header and buffer size */
1814	if (copy_from_user(&args, uarg, sizeof(args)))
1815		return -EFAULT;
1816
1817	buf_size = args.buf_size;
1818
1819	/* limit result size to 16MB */
1820	if (buf_size > buf_limit)
1821		buf_size = buf_limit;
1822
1823	ret = search_ioctl(inode, &args.key, &buf_size,
1824			   (char __user *)(&uarg->buf[0]));
1825	if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1826		ret = -EFAULT;
1827	else if (ret == -EOVERFLOW &&
1828		copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1829		ret = -EFAULT;
1830
1831	return ret;
1832}
1833
1834/*
1835 * Search INODE_REFs to identify path name of 'dirid' directory
1836 * in a 'tree_id' tree. and sets path name to 'name'.
1837 */
1838static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1839				u64 tree_id, u64 dirid, char *name)
1840{
1841	struct btrfs_root *root;
1842	struct btrfs_key key;
1843	char *ptr;
1844	int ret = -1;
1845	int slot;
1846	int len;
1847	int total_len = 0;
1848	struct btrfs_inode_ref *iref;
1849	struct extent_buffer *l;
1850	struct btrfs_path *path;
1851
1852	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1853		name[0]='\0';
1854		return 0;
1855	}
1856
1857	path = btrfs_alloc_path();
1858	if (!path)
1859		return -ENOMEM;
1860
1861	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1862
1863	root = btrfs_get_fs_root(info, tree_id, true);
1864	if (IS_ERR(root)) {
1865		ret = PTR_ERR(root);
1866		root = NULL;
1867		goto out;
1868	}
1869
1870	key.objectid = dirid;
1871	key.type = BTRFS_INODE_REF_KEY;
1872	key.offset = (u64)-1;
1873
1874	while (1) {
1875		ret = btrfs_search_backwards(root, &key, path);
1876		if (ret < 0)
1877			goto out;
1878		else if (ret > 0) {
1879			ret = -ENOENT;
1880			goto out;
1881		}
1882
1883		l = path->nodes[0];
1884		slot = path->slots[0];
1885
1886		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1887		len = btrfs_inode_ref_name_len(l, iref);
1888		ptr -= len + 1;
1889		total_len += len + 1;
1890		if (ptr < name) {
1891			ret = -ENAMETOOLONG;
1892			goto out;
1893		}
1894
1895		*(ptr + len) = '/';
1896		read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1897
1898		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1899			break;
1900
1901		btrfs_release_path(path);
1902		key.objectid = key.offset;
1903		key.offset = (u64)-1;
1904		dirid = key.objectid;
1905	}
1906	memmove(name, ptr, total_len);
1907	name[total_len] = '\0';
1908	ret = 0;
1909out:
1910	btrfs_put_root(root);
1911	btrfs_free_path(path);
1912	return ret;
1913}
1914
1915static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1916				struct inode *inode,
1917				struct btrfs_ioctl_ino_lookup_user_args *args)
1918{
1919	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1920	struct super_block *sb = inode->i_sb;
1921	struct btrfs_key upper_limit = BTRFS_I(inode)->location;
1922	u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
1923	u64 dirid = args->dirid;
1924	unsigned long item_off;
1925	unsigned long item_len;
1926	struct btrfs_inode_ref *iref;
1927	struct btrfs_root_ref *rref;
1928	struct btrfs_root *root = NULL;
1929	struct btrfs_path *path;
1930	struct btrfs_key key, key2;
1931	struct extent_buffer *leaf;
1932	struct inode *temp_inode;
1933	char *ptr;
1934	int slot;
1935	int len;
1936	int total_len = 0;
1937	int ret;
1938
1939	path = btrfs_alloc_path();
1940	if (!path)
1941		return -ENOMEM;
1942
1943	/*
1944	 * If the bottom subvolume does not exist directly under upper_limit,
1945	 * construct the path in from the bottom up.
1946	 */
1947	if (dirid != upper_limit.objectid) {
1948		ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1949
1950		root = btrfs_get_fs_root(fs_info, treeid, true);
1951		if (IS_ERR(root)) {
1952			ret = PTR_ERR(root);
1953			goto out;
1954		}
1955
1956		key.objectid = dirid;
1957		key.type = BTRFS_INODE_REF_KEY;
1958		key.offset = (u64)-1;
1959		while (1) {
1960			ret = btrfs_search_backwards(root, &key, path);
1961			if (ret < 0)
1962				goto out_put;
1963			else if (ret > 0) {
1964				ret = -ENOENT;
1965				goto out_put;
1966			}
1967
1968			leaf = path->nodes[0];
1969			slot = path->slots[0];
1970
1971			iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1972			len = btrfs_inode_ref_name_len(leaf, iref);
1973			ptr -= len + 1;
1974			total_len += len + 1;
1975			if (ptr < args->path) {
1976				ret = -ENAMETOOLONG;
1977				goto out_put;
1978			}
1979
1980			*(ptr + len) = '/';
1981			read_extent_buffer(leaf, ptr,
1982					(unsigned long)(iref + 1), len);
1983
1984			/* Check the read+exec permission of this directory */
1985			ret = btrfs_previous_item(root, path, dirid,
1986						  BTRFS_INODE_ITEM_KEY);
1987			if (ret < 0) {
1988				goto out_put;
1989			} else if (ret > 0) {
1990				ret = -ENOENT;
1991				goto out_put;
1992			}
1993
1994			leaf = path->nodes[0];
1995			slot = path->slots[0];
1996			btrfs_item_key_to_cpu(leaf, &key2, slot);
1997			if (key2.objectid != dirid) {
1998				ret = -ENOENT;
1999				goto out_put;
2000			}
2001
2002			/*
2003			 * We don't need the path anymore, so release it and
2004			 * avoid deadlocks and lockdep warnings in case
2005			 * btrfs_iget() needs to lookup the inode from its root
2006			 * btree and lock the same leaf.
2007			 */
2008			btrfs_release_path(path);
2009			temp_inode = btrfs_iget(sb, key2.objectid, root);
2010			if (IS_ERR(temp_inode)) {
2011				ret = PTR_ERR(temp_inode);
2012				goto out_put;
2013			}
2014			ret = inode_permission(idmap, temp_inode,
2015					       MAY_READ | MAY_EXEC);
2016			iput(temp_inode);
2017			if (ret) {
2018				ret = -EACCES;
2019				goto out_put;
2020			}
2021
2022			if (key.offset == upper_limit.objectid)
2023				break;
2024			if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2025				ret = -EACCES;
2026				goto out_put;
2027			}
2028
2029			key.objectid = key.offset;
2030			key.offset = (u64)-1;
2031			dirid = key.objectid;
2032		}
2033
2034		memmove(args->path, ptr, total_len);
2035		args->path[total_len] = '\0';
2036		btrfs_put_root(root);
2037		root = NULL;
2038		btrfs_release_path(path);
2039	}
2040
2041	/* Get the bottom subvolume's name from ROOT_REF */
2042	key.objectid = treeid;
2043	key.type = BTRFS_ROOT_REF_KEY;
2044	key.offset = args->treeid;
2045	ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2046	if (ret < 0) {
2047		goto out;
2048	} else if (ret > 0) {
2049		ret = -ENOENT;
2050		goto out;
2051	}
2052
2053	leaf = path->nodes[0];
2054	slot = path->slots[0];
2055	btrfs_item_key_to_cpu(leaf, &key, slot);
2056
2057	item_off = btrfs_item_ptr_offset(leaf, slot);
2058	item_len = btrfs_item_size(leaf, slot);
2059	/* Check if dirid in ROOT_REF corresponds to passed dirid */
2060	rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2061	if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2062		ret = -EINVAL;
2063		goto out;
2064	}
2065
2066	/* Copy subvolume's name */
2067	item_off += sizeof(struct btrfs_root_ref);
2068	item_len -= sizeof(struct btrfs_root_ref);
2069	read_extent_buffer(leaf, args->name, item_off, item_len);
2070	args->name[item_len] = 0;
2071
2072out_put:
2073	btrfs_put_root(root);
2074out:
2075	btrfs_free_path(path);
2076	return ret;
2077}
2078
2079static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2080					   void __user *argp)
2081{
2082	struct btrfs_ioctl_ino_lookup_args *args;
2083	int ret = 0;
2084
2085	args = memdup_user(argp, sizeof(*args));
2086	if (IS_ERR(args))
2087		return PTR_ERR(args);
2088
2089	/*
2090	 * Unprivileged query to obtain the containing subvolume root id. The
2091	 * path is reset so it's consistent with btrfs_search_path_in_tree.
2092	 */
2093	if (args->treeid == 0)
2094		args->treeid = root->root_key.objectid;
2095
2096	if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2097		args->name[0] = 0;
2098		goto out;
2099	}
2100
2101	if (!capable(CAP_SYS_ADMIN)) {
2102		ret = -EPERM;
2103		goto out;
2104	}
2105
2106	ret = btrfs_search_path_in_tree(root->fs_info,
2107					args->treeid, args->objectid,
2108					args->name);
2109
2110out:
2111	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2112		ret = -EFAULT;
2113
2114	kfree(args);
2115	return ret;
2116}
2117
2118/*
2119 * Version of ino_lookup ioctl (unprivileged)
2120 *
2121 * The main differences from ino_lookup ioctl are:
2122 *
2123 *   1. Read + Exec permission will be checked using inode_permission() during
2124 *      path construction. -EACCES will be returned in case of failure.
2125 *   2. Path construction will be stopped at the inode number which corresponds
2126 *      to the fd with which this ioctl is called. If constructed path does not
2127 *      exist under fd's inode, -EACCES will be returned.
2128 *   3. The name of bottom subvolume is also searched and filled.
2129 */
2130static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2131{
2132	struct btrfs_ioctl_ino_lookup_user_args *args;
2133	struct inode *inode;
2134	int ret;
2135
2136	args = memdup_user(argp, sizeof(*args));
2137	if (IS_ERR(args))
2138		return PTR_ERR(args);
2139
2140	inode = file_inode(file);
2141
2142	if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2143	    BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2144		/*
2145		 * The subvolume does not exist under fd with which this is
2146		 * called
2147		 */
2148		kfree(args);
2149		return -EACCES;
2150	}
2151
2152	ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2153
2154	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2155		ret = -EFAULT;
2156
2157	kfree(args);
2158	return ret;
2159}
2160
2161/* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2162static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2163{
2164	struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2165	struct btrfs_fs_info *fs_info;
2166	struct btrfs_root *root;
2167	struct btrfs_path *path;
2168	struct btrfs_key key;
2169	struct btrfs_root_item *root_item;
2170	struct btrfs_root_ref *rref;
2171	struct extent_buffer *leaf;
2172	unsigned long item_off;
2173	unsigned long item_len;
2174	int slot;
2175	int ret = 0;
2176
2177	path = btrfs_alloc_path();
2178	if (!path)
2179		return -ENOMEM;
2180
2181	subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2182	if (!subvol_info) {
2183		btrfs_free_path(path);
2184		return -ENOMEM;
2185	}
2186
2187	fs_info = BTRFS_I(inode)->root->fs_info;
2188
2189	/* Get root_item of inode's subvolume */
2190	key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2191	root = btrfs_get_fs_root(fs_info, key.objectid, true);
2192	if (IS_ERR(root)) {
2193		ret = PTR_ERR(root);
2194		goto out_free;
2195	}
2196	root_item = &root->root_item;
2197
2198	subvol_info->treeid = key.objectid;
2199
2200	subvol_info->generation = btrfs_root_generation(root_item);
2201	subvol_info->flags = btrfs_root_flags(root_item);
2202
2203	memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2204	memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2205						    BTRFS_UUID_SIZE);
2206	memcpy(subvol_info->received_uuid, root_item->received_uuid,
2207						    BTRFS_UUID_SIZE);
2208
2209	subvol_info->ctransid = btrfs_root_ctransid(root_item);
2210	subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2211	subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2212
2213	subvol_info->otransid = btrfs_root_otransid(root_item);
2214	subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2215	subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2216
2217	subvol_info->stransid = btrfs_root_stransid(root_item);
2218	subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2219	subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2220
2221	subvol_info->rtransid = btrfs_root_rtransid(root_item);
2222	subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2223	subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2224
2225	if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2226		/* Search root tree for ROOT_BACKREF of this subvolume */
2227		key.type = BTRFS_ROOT_BACKREF_KEY;
2228		key.offset = 0;
2229		ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2230		if (ret < 0) {
2231			goto out;
2232		} else if (path->slots[0] >=
2233			   btrfs_header_nritems(path->nodes[0])) {
2234			ret = btrfs_next_leaf(fs_info->tree_root, path);
2235			if (ret < 0) {
2236				goto out;
2237			} else if (ret > 0) {
2238				ret = -EUCLEAN;
2239				goto out;
2240			}
2241		}
2242
2243		leaf = path->nodes[0];
2244		slot = path->slots[0];
2245		btrfs_item_key_to_cpu(leaf, &key, slot);
2246		if (key.objectid == subvol_info->treeid &&
2247		    key.type == BTRFS_ROOT_BACKREF_KEY) {
2248			subvol_info->parent_id = key.offset;
2249
2250			rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2251			subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2252
2253			item_off = btrfs_item_ptr_offset(leaf, slot)
2254					+ sizeof(struct btrfs_root_ref);
2255			item_len = btrfs_item_size(leaf, slot)
2256					- sizeof(struct btrfs_root_ref);
2257			read_extent_buffer(leaf, subvol_info->name,
2258					   item_off, item_len);
2259		} else {
2260			ret = -ENOENT;
2261			goto out;
2262		}
2263	}
2264
2265	btrfs_free_path(path);
2266	path = NULL;
2267	if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2268		ret = -EFAULT;
2269
2270out:
2271	btrfs_put_root(root);
2272out_free:
2273	btrfs_free_path(path);
2274	kfree(subvol_info);
2275	return ret;
2276}
2277
2278/*
2279 * Return ROOT_REF information of the subvolume containing this inode
2280 * except the subvolume name.
2281 */
2282static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2283					  void __user *argp)
2284{
2285	struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2286	struct btrfs_root_ref *rref;
2287	struct btrfs_path *path;
2288	struct btrfs_key key;
2289	struct extent_buffer *leaf;
2290	u64 objectid;
2291	int slot;
2292	int ret;
2293	u8 found;
2294
2295	path = btrfs_alloc_path();
2296	if (!path)
2297		return -ENOMEM;
2298
2299	rootrefs = memdup_user(argp, sizeof(*rootrefs));
2300	if (IS_ERR(rootrefs)) {
2301		btrfs_free_path(path);
2302		return PTR_ERR(rootrefs);
2303	}
2304
2305	objectid = root->root_key.objectid;
2306	key.objectid = objectid;
2307	key.type = BTRFS_ROOT_REF_KEY;
2308	key.offset = rootrefs->min_treeid;
2309	found = 0;
2310
2311	root = root->fs_info->tree_root;
2312	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2313	if (ret < 0) {
2314		goto out;
2315	} else if (path->slots[0] >=
2316		   btrfs_header_nritems(path->nodes[0])) {
2317		ret = btrfs_next_leaf(root, path);
2318		if (ret < 0) {
2319			goto out;
2320		} else if (ret > 0) {
2321			ret = -EUCLEAN;
2322			goto out;
2323		}
2324	}
2325	while (1) {
2326		leaf = path->nodes[0];
2327		slot = path->slots[0];
2328
2329		btrfs_item_key_to_cpu(leaf, &key, slot);
2330		if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2331			ret = 0;
2332			goto out;
2333		}
2334
2335		if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2336			ret = -EOVERFLOW;
2337			goto out;
2338		}
2339
2340		rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2341		rootrefs->rootref[found].treeid = key.offset;
2342		rootrefs->rootref[found].dirid =
2343				  btrfs_root_ref_dirid(leaf, rref);
2344		found++;
2345
2346		ret = btrfs_next_item(root, path);
2347		if (ret < 0) {
2348			goto out;
2349		} else if (ret > 0) {
2350			ret = -EUCLEAN;
2351			goto out;
2352		}
2353	}
2354
2355out:
2356	btrfs_free_path(path);
2357
2358	if (!ret || ret == -EOVERFLOW) {
2359		rootrefs->num_items = found;
2360		/* update min_treeid for next search */
2361		if (found)
2362			rootrefs->min_treeid =
2363				rootrefs->rootref[found - 1].treeid + 1;
2364		if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2365			ret = -EFAULT;
2366	}
2367
2368	kfree(rootrefs);
2369
2370	return ret;
2371}
2372
2373static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2374					     void __user *arg,
2375					     bool destroy_v2)
2376{
2377	struct dentry *parent = file->f_path.dentry;
2378	struct dentry *dentry;
2379	struct inode *dir = d_inode(parent);
2380	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
2381	struct inode *inode;
2382	struct btrfs_root *root = BTRFS_I(dir)->root;
2383	struct btrfs_root *dest = NULL;
2384	struct btrfs_ioctl_vol_args *vol_args = NULL;
2385	struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2386	struct mnt_idmap *idmap = file_mnt_idmap(file);
2387	char *subvol_name, *subvol_name_ptr = NULL;
2388	int subvol_namelen;
2389	int err = 0;
2390	bool destroy_parent = false;
2391
2392	/* We don't support snapshots with extent tree v2 yet. */
2393	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2394		btrfs_err(fs_info,
2395			  "extent tree v2 doesn't support snapshot deletion yet");
2396		return -EOPNOTSUPP;
2397	}
2398
2399	if (destroy_v2) {
2400		vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2401		if (IS_ERR(vol_args2))
2402			return PTR_ERR(vol_args2);
2403
2404		if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2405			err = -EOPNOTSUPP;
2406			goto out;
2407		}
2408
2409		/*
2410		 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2411		 * name, same as v1 currently does.
2412		 */
2413		if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2414			err = btrfs_check_ioctl_vol_args2_subvol_name(vol_args2);
2415			if (err < 0)
2416				goto out;
2417			subvol_name = vol_args2->name;
2418
2419			err = mnt_want_write_file(file);
2420			if (err)
2421				goto out;
2422		} else {
2423			struct inode *old_dir;
2424
2425			if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2426				err = -EINVAL;
2427				goto out;
2428			}
2429
2430			err = mnt_want_write_file(file);
2431			if (err)
2432				goto out;
2433
2434			dentry = btrfs_get_dentry(fs_info->sb,
2435					BTRFS_FIRST_FREE_OBJECTID,
2436					vol_args2->subvolid, 0);
2437			if (IS_ERR(dentry)) {
2438				err = PTR_ERR(dentry);
2439				goto out_drop_write;
2440			}
2441
2442			/*
2443			 * Change the default parent since the subvolume being
2444			 * deleted can be outside of the current mount point.
2445			 */
2446			parent = btrfs_get_parent(dentry);
2447
2448			/*
2449			 * At this point dentry->d_name can point to '/' if the
2450			 * subvolume we want to destroy is outsite of the
2451			 * current mount point, so we need to release the
2452			 * current dentry and execute the lookup to return a new
2453			 * one with ->d_name pointing to the
2454			 * <mount point>/subvol_name.
2455			 */
2456			dput(dentry);
2457			if (IS_ERR(parent)) {
2458				err = PTR_ERR(parent);
2459				goto out_drop_write;
2460			}
2461			old_dir = dir;
2462			dir = d_inode(parent);
2463
2464			/*
2465			 * If v2 was used with SPEC_BY_ID, a new parent was
2466			 * allocated since the subvolume can be outside of the
2467			 * current mount point. Later on we need to release this
2468			 * new parent dentry.
2469			 */
2470			destroy_parent = true;
2471
2472			/*
2473			 * On idmapped mounts, deletion via subvolid is
2474			 * restricted to subvolumes that are immediate
2475			 * ancestors of the inode referenced by the file
2476			 * descriptor in the ioctl. Otherwise the idmapping
2477			 * could potentially be abused to delete subvolumes
2478			 * anywhere in the filesystem the user wouldn't be able
2479			 * to delete without an idmapped mount.
2480			 */
2481			if (old_dir != dir && idmap != &nop_mnt_idmap) {
2482				err = -EOPNOTSUPP;
2483				goto free_parent;
2484			}
2485
2486			subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2487						fs_info, vol_args2->subvolid);
2488			if (IS_ERR(subvol_name_ptr)) {
2489				err = PTR_ERR(subvol_name_ptr);
2490				goto free_parent;
2491			}
2492			/* subvol_name_ptr is already nul terminated */
2493			subvol_name = (char *)kbasename(subvol_name_ptr);
2494		}
2495	} else {
2496		vol_args = memdup_user(arg, sizeof(*vol_args));
2497		if (IS_ERR(vol_args))
2498			return PTR_ERR(vol_args);
2499
2500		err = btrfs_check_ioctl_vol_args_path(vol_args);
2501		if (err < 0)
2502			goto out;
2503
2504		subvol_name = vol_args->name;
2505
2506		err = mnt_want_write_file(file);
2507		if (err)
2508			goto out;
2509	}
2510
2511	subvol_namelen = strlen(subvol_name);
2512
2513	if (strchr(subvol_name, '/') ||
2514	    strncmp(subvol_name, "..", subvol_namelen) == 0) {
2515		err = -EINVAL;
2516		goto free_subvol_name;
2517	}
2518
2519	if (!S_ISDIR(dir->i_mode)) {
2520		err = -ENOTDIR;
2521		goto free_subvol_name;
2522	}
2523
2524	err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2525	if (err == -EINTR)
2526		goto free_subvol_name;
2527	dentry = lookup_one(idmap, subvol_name, parent, subvol_namelen);
2528	if (IS_ERR(dentry)) {
2529		err = PTR_ERR(dentry);
2530		goto out_unlock_dir;
2531	}
2532
2533	if (d_really_is_negative(dentry)) {
2534		err = -ENOENT;
2535		goto out_dput;
2536	}
2537
2538	inode = d_inode(dentry);
2539	dest = BTRFS_I(inode)->root;
2540	if (!capable(CAP_SYS_ADMIN)) {
2541		/*
2542		 * Regular user.  Only allow this with a special mount
2543		 * option, when the user has write+exec access to the
2544		 * subvol root, and when rmdir(2) would have been
2545		 * allowed.
2546		 *
2547		 * Note that this is _not_ check that the subvol is
2548		 * empty or doesn't contain data that we wouldn't
2549		 * otherwise be able to delete.
2550		 *
2551		 * Users who want to delete empty subvols should try
2552		 * rmdir(2).
2553		 */
2554		err = -EPERM;
2555		if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2556			goto out_dput;
2557
2558		/*
2559		 * Do not allow deletion if the parent dir is the same
2560		 * as the dir to be deleted.  That means the ioctl
2561		 * must be called on the dentry referencing the root
2562		 * of the subvol, not a random directory contained
2563		 * within it.
2564		 */
2565		err = -EINVAL;
2566		if (root == dest)
2567			goto out_dput;
2568
2569		err = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2570		if (err)
2571			goto out_dput;
2572	}
2573
2574	/* check if subvolume may be deleted by a user */
2575	err = btrfs_may_delete(idmap, dir, dentry, 1);
2576	if (err)
2577		goto out_dput;
2578
2579	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2580		err = -EINVAL;
2581		goto out_dput;
2582	}
2583
2584	btrfs_inode_lock(BTRFS_I(inode), 0);
2585	err = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2586	btrfs_inode_unlock(BTRFS_I(inode), 0);
2587	if (!err)
2588		d_delete_notify(dir, dentry);
2589
2590out_dput:
2591	dput(dentry);
2592out_unlock_dir:
2593	btrfs_inode_unlock(BTRFS_I(dir), 0);
2594free_subvol_name:
2595	kfree(subvol_name_ptr);
2596free_parent:
2597	if (destroy_parent)
2598		dput(parent);
2599out_drop_write:
2600	mnt_drop_write_file(file);
2601out:
2602	kfree(vol_args2);
2603	kfree(vol_args);
2604	return err;
2605}
2606
2607static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2608{
2609	struct inode *inode = file_inode(file);
2610	struct btrfs_root *root = BTRFS_I(inode)->root;
2611	struct btrfs_ioctl_defrag_range_args range = {0};
2612	int ret;
2613
2614	ret = mnt_want_write_file(file);
2615	if (ret)
2616		return ret;
2617
2618	if (btrfs_root_readonly(root)) {
2619		ret = -EROFS;
2620		goto out;
2621	}
2622
2623	switch (inode->i_mode & S_IFMT) {
2624	case S_IFDIR:
2625		if (!capable(CAP_SYS_ADMIN)) {
2626			ret = -EPERM;
2627			goto out;
2628		}
2629		ret = btrfs_defrag_root(root);
2630		break;
2631	case S_IFREG:
2632		/*
2633		 * Note that this does not check the file descriptor for write
2634		 * access. This prevents defragmenting executables that are
2635		 * running and allows defrag on files open in read-only mode.
2636		 */
2637		if (!capable(CAP_SYS_ADMIN) &&
2638		    inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2639			ret = -EPERM;
2640			goto out;
2641		}
2642
2643		if (argp) {
2644			if (copy_from_user(&range, argp, sizeof(range))) {
2645				ret = -EFAULT;
2646				goto out;
2647			}
2648			if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) {
2649				ret = -EOPNOTSUPP;
2650				goto out;
2651			}
2652			/* compression requires us to start the IO */
2653			if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2654				range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2655				range.extent_thresh = (u32)-1;
2656			}
2657		} else {
2658			/* the rest are all set to zero by kzalloc */
2659			range.len = (u64)-1;
2660		}
2661		ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2662					&range, BTRFS_OLDEST_GENERATION, 0);
2663		if (ret > 0)
2664			ret = 0;
2665		break;
2666	default:
2667		ret = -EINVAL;
2668	}
2669out:
2670	mnt_drop_write_file(file);
2671	return ret;
2672}
2673
2674static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2675{
2676	struct btrfs_ioctl_vol_args *vol_args;
2677	bool restore_op = false;
2678	int ret;
2679
2680	if (!capable(CAP_SYS_ADMIN))
2681		return -EPERM;
2682
2683	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2684		btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2685		return -EINVAL;
2686	}
2687
2688	if (fs_info->fs_devices->temp_fsid) {
2689		btrfs_err(fs_info,
2690			  "device add not supported on cloned temp-fsid mount");
2691		return -EINVAL;
2692	}
2693
2694	if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2695		if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2696			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2697
2698		/*
2699		 * We can do the device add because we have a paused balanced,
2700		 * change the exclusive op type and remember we should bring
2701		 * back the paused balance
2702		 */
2703		fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2704		btrfs_exclop_start_unlock(fs_info);
2705		restore_op = true;
2706	}
2707
2708	vol_args = memdup_user(arg, sizeof(*vol_args));
2709	if (IS_ERR(vol_args)) {
2710		ret = PTR_ERR(vol_args);
2711		goto out;
2712	}
2713
2714	ret = btrfs_check_ioctl_vol_args_path(vol_args);
2715	if (ret < 0)
2716		goto out_free;
2717
2718	ret = btrfs_init_new_device(fs_info, vol_args->name);
2719
2720	if (!ret)
2721		btrfs_info(fs_info, "disk added %s", vol_args->name);
2722
2723out_free:
2724	kfree(vol_args);
2725out:
2726	if (restore_op)
2727		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2728	else
2729		btrfs_exclop_finish(fs_info);
2730	return ret;
2731}
2732
2733static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2734{
2735	BTRFS_DEV_LOOKUP_ARGS(args);
2736	struct inode *inode = file_inode(file);
2737	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2738	struct btrfs_ioctl_vol_args_v2 *vol_args;
2739	struct file *bdev_file = NULL;
2740	int ret;
2741	bool cancel = false;
2742
2743	if (!capable(CAP_SYS_ADMIN))
2744		return -EPERM;
2745
2746	vol_args = memdup_user(arg, sizeof(*vol_args));
2747	if (IS_ERR(vol_args))
2748		return PTR_ERR(vol_args);
2749
2750	if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2751		ret = -EOPNOTSUPP;
2752		goto out;
2753	}
2754
2755	ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args);
2756	if (ret < 0)
2757		goto out;
2758
2759	if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2760		args.devid = vol_args->devid;
2761	} else if (!strcmp("cancel", vol_args->name)) {
2762		cancel = true;
2763	} else {
2764		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2765		if (ret)
2766			goto out;
2767	}
2768
2769	ret = mnt_want_write_file(file);
2770	if (ret)
2771		goto out;
2772
2773	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2774					   cancel);
2775	if (ret)
2776		goto err_drop;
2777
2778	/* Exclusive operation is now claimed */
2779	ret = btrfs_rm_device(fs_info, &args, &bdev_file);
2780
2781	btrfs_exclop_finish(fs_info);
2782
2783	if (!ret) {
2784		if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2785			btrfs_info(fs_info, "device deleted: id %llu",
2786					vol_args->devid);
2787		else
2788			btrfs_info(fs_info, "device deleted: %s",
2789					vol_args->name);
2790	}
2791err_drop:
2792	mnt_drop_write_file(file);
2793	if (bdev_file)
2794		fput(bdev_file);
2795out:
2796	btrfs_put_dev_args_from_path(&args);
2797	kfree(vol_args);
2798	return ret;
2799}
2800
2801static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2802{
2803	BTRFS_DEV_LOOKUP_ARGS(args);
2804	struct inode *inode = file_inode(file);
2805	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2806	struct btrfs_ioctl_vol_args *vol_args;
2807	struct file *bdev_file = NULL;
2808	int ret;
2809	bool cancel = false;
2810
2811	if (!capable(CAP_SYS_ADMIN))
2812		return -EPERM;
2813
2814	vol_args = memdup_user(arg, sizeof(*vol_args));
2815	if (IS_ERR(vol_args))
2816		return PTR_ERR(vol_args);
2817
2818	ret = btrfs_check_ioctl_vol_args_path(vol_args);
2819	if (ret < 0)
2820		goto out_free;
2821
2822	if (!strcmp("cancel", vol_args->name)) {
2823		cancel = true;
2824	} else {
2825		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2826		if (ret)
2827			goto out;
2828	}
2829
2830	ret = mnt_want_write_file(file);
2831	if (ret)
2832		goto out;
2833
2834	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2835					   cancel);
2836	if (ret == 0) {
2837		ret = btrfs_rm_device(fs_info, &args, &bdev_file);
2838		if (!ret)
2839			btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2840		btrfs_exclop_finish(fs_info);
2841	}
2842
2843	mnt_drop_write_file(file);
2844	if (bdev_file)
2845		fput(bdev_file);
2846out:
2847	btrfs_put_dev_args_from_path(&args);
2848out_free:
2849	kfree(vol_args);
2850	return ret;
2851}
2852
2853static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2854				void __user *arg)
2855{
2856	struct btrfs_ioctl_fs_info_args *fi_args;
2857	struct btrfs_device *device;
2858	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2859	u64 flags_in;
2860	int ret = 0;
2861
2862	fi_args = memdup_user(arg, sizeof(*fi_args));
2863	if (IS_ERR(fi_args))
2864		return PTR_ERR(fi_args);
2865
2866	flags_in = fi_args->flags;
2867	memset(fi_args, 0, sizeof(*fi_args));
2868
2869	rcu_read_lock();
2870	fi_args->num_devices = fs_devices->num_devices;
2871
2872	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2873		if (device->devid > fi_args->max_id)
2874			fi_args->max_id = device->devid;
2875	}
2876	rcu_read_unlock();
2877
2878	memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2879	fi_args->nodesize = fs_info->nodesize;
2880	fi_args->sectorsize = fs_info->sectorsize;
2881	fi_args->clone_alignment = fs_info->sectorsize;
2882
2883	if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2884		fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2885		fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2886		fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2887	}
2888
2889	if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2890		fi_args->generation = btrfs_get_fs_generation(fs_info);
2891		fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2892	}
2893
2894	if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2895		memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2896		       sizeof(fi_args->metadata_uuid));
2897		fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2898	}
2899
2900	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2901		ret = -EFAULT;
2902
2903	kfree(fi_args);
2904	return ret;
2905}
2906
2907static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2908				 void __user *arg)
2909{
2910	BTRFS_DEV_LOOKUP_ARGS(args);
2911	struct btrfs_ioctl_dev_info_args *di_args;
2912	struct btrfs_device *dev;
2913	int ret = 0;
2914
2915	di_args = memdup_user(arg, sizeof(*di_args));
2916	if (IS_ERR(di_args))
2917		return PTR_ERR(di_args);
2918
2919	args.devid = di_args->devid;
2920	if (!btrfs_is_empty_uuid(di_args->uuid))
2921		args.uuid = di_args->uuid;
2922
2923	rcu_read_lock();
2924	dev = btrfs_find_device(fs_info->fs_devices, &args);
2925	if (!dev) {
2926		ret = -ENODEV;
2927		goto out;
2928	}
2929
2930	di_args->devid = dev->devid;
2931	di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2932	di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2933	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2934	memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2935	if (dev->name)
2936		strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2937	else
2938		di_args->path[0] = '\0';
2939
2940out:
2941	rcu_read_unlock();
2942	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2943		ret = -EFAULT;
2944
2945	kfree(di_args);
2946	return ret;
2947}
2948
2949static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2950{
2951	struct inode *inode = file_inode(file);
2952	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2953	struct btrfs_root *root = BTRFS_I(inode)->root;
2954	struct btrfs_root *new_root;
2955	struct btrfs_dir_item *di;
2956	struct btrfs_trans_handle *trans;
2957	struct btrfs_path *path = NULL;
2958	struct btrfs_disk_key disk_key;
2959	struct fscrypt_str name = FSTR_INIT("default", 7);
2960	u64 objectid = 0;
2961	u64 dir_id;
2962	int ret;
2963
2964	if (!capable(CAP_SYS_ADMIN))
2965		return -EPERM;
2966
2967	ret = mnt_want_write_file(file);
2968	if (ret)
2969		return ret;
2970
2971	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2972		ret = -EFAULT;
2973		goto out;
2974	}
2975
2976	if (!objectid)
2977		objectid = BTRFS_FS_TREE_OBJECTID;
2978
2979	new_root = btrfs_get_fs_root(fs_info, objectid, true);
2980	if (IS_ERR(new_root)) {
2981		ret = PTR_ERR(new_root);
2982		goto out;
2983	}
2984	if (!is_fstree(new_root->root_key.objectid)) {
2985		ret = -ENOENT;
2986		goto out_free;
2987	}
2988
2989	path = btrfs_alloc_path();
2990	if (!path) {
2991		ret = -ENOMEM;
2992		goto out_free;
2993	}
2994
2995	trans = btrfs_start_transaction(root, 1);
2996	if (IS_ERR(trans)) {
2997		ret = PTR_ERR(trans);
2998		goto out_free;
2999	}
3000
3001	dir_id = btrfs_super_root_dir(fs_info->super_copy);
3002	di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3003				   dir_id, &name, 1);
3004	if (IS_ERR_OR_NULL(di)) {
3005		btrfs_release_path(path);
3006		btrfs_end_transaction(trans);
3007		btrfs_err(fs_info,
3008			  "Umm, you don't have the default diritem, this isn't going to work");
3009		ret = -ENOENT;
3010		goto out_free;
3011	}
3012
3013	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3014	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3015	btrfs_mark_buffer_dirty(trans, path->nodes[0]);
3016	btrfs_release_path(path);
3017
3018	btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3019	btrfs_end_transaction(trans);
3020out_free:
3021	btrfs_put_root(new_root);
3022	btrfs_free_path(path);
3023out:
3024	mnt_drop_write_file(file);
3025	return ret;
3026}
3027
3028static void get_block_group_info(struct list_head *groups_list,
3029				 struct btrfs_ioctl_space_info *space)
3030{
3031	struct btrfs_block_group *block_group;
3032
3033	space->total_bytes = 0;
3034	space->used_bytes = 0;
3035	space->flags = 0;
3036	list_for_each_entry(block_group, groups_list, list) {
3037		space->flags = block_group->flags;
3038		space->total_bytes += block_group->length;
3039		space->used_bytes += block_group->used;
3040	}
3041}
3042
3043static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3044				   void __user *arg)
3045{
3046	struct btrfs_ioctl_space_args space_args = { 0 };
3047	struct btrfs_ioctl_space_info space;
3048	struct btrfs_ioctl_space_info *dest;
3049	struct btrfs_ioctl_space_info *dest_orig;
3050	struct btrfs_ioctl_space_info __user *user_dest;
3051	struct btrfs_space_info *info;
3052	static const u64 types[] = {
3053		BTRFS_BLOCK_GROUP_DATA,
3054		BTRFS_BLOCK_GROUP_SYSTEM,
3055		BTRFS_BLOCK_GROUP_METADATA,
3056		BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3057	};
3058	int num_types = 4;
3059	int alloc_size;
3060	int ret = 0;
3061	u64 slot_count = 0;
3062	int i, c;
3063
3064	if (copy_from_user(&space_args,
3065			   (struct btrfs_ioctl_space_args __user *)arg,
3066			   sizeof(space_args)))
3067		return -EFAULT;
3068
3069	for (i = 0; i < num_types; i++) {
3070		struct btrfs_space_info *tmp;
3071
3072		info = NULL;
3073		list_for_each_entry(tmp, &fs_info->space_info, list) {
3074			if (tmp->flags == types[i]) {
3075				info = tmp;
3076				break;
3077			}
3078		}
3079
3080		if (!info)
3081			continue;
3082
3083		down_read(&info->groups_sem);
3084		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3085			if (!list_empty(&info->block_groups[c]))
3086				slot_count++;
3087		}
3088		up_read(&info->groups_sem);
3089	}
3090
3091	/*
3092	 * Global block reserve, exported as a space_info
3093	 */
3094	slot_count++;
3095
3096	/* space_slots == 0 means they are asking for a count */
3097	if (space_args.space_slots == 0) {
3098		space_args.total_spaces = slot_count;
3099		goto out;
3100	}
3101
3102	slot_count = min_t(u64, space_args.space_slots, slot_count);
3103
3104	alloc_size = sizeof(*dest) * slot_count;
3105
3106	/* we generally have at most 6 or so space infos, one for each raid
3107	 * level.  So, a whole page should be more than enough for everyone
3108	 */
3109	if (alloc_size > PAGE_SIZE)
3110		return -ENOMEM;
3111
3112	space_args.total_spaces = 0;
3113	dest = kmalloc(alloc_size, GFP_KERNEL);
3114	if (!dest)
3115		return -ENOMEM;
3116	dest_orig = dest;
3117
3118	/* now we have a buffer to copy into */
3119	for (i = 0; i < num_types; i++) {
3120		struct btrfs_space_info *tmp;
3121
3122		if (!slot_count)
3123			break;
3124
3125		info = NULL;
3126		list_for_each_entry(tmp, &fs_info->space_info, list) {
3127			if (tmp->flags == types[i]) {
3128				info = tmp;
3129				break;
3130			}
3131		}
3132
3133		if (!info)
3134			continue;
3135		down_read(&info->groups_sem);
3136		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3137			if (!list_empty(&info->block_groups[c])) {
3138				get_block_group_info(&info->block_groups[c],
3139						     &space);
3140				memcpy(dest, &space, sizeof(space));
3141				dest++;
3142				space_args.total_spaces++;
3143				slot_count--;
3144			}
3145			if (!slot_count)
3146				break;
3147		}
3148		up_read(&info->groups_sem);
3149	}
3150
3151	/*
3152	 * Add global block reserve
3153	 */
3154	if (slot_count) {
3155		struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3156
3157		spin_lock(&block_rsv->lock);
3158		space.total_bytes = block_rsv->size;
3159		space.used_bytes = block_rsv->size - block_rsv->reserved;
3160		spin_unlock(&block_rsv->lock);
3161		space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3162		memcpy(dest, &space, sizeof(space));
3163		space_args.total_spaces++;
3164	}
3165
3166	user_dest = (struct btrfs_ioctl_space_info __user *)
3167		(arg + sizeof(struct btrfs_ioctl_space_args));
3168
3169	if (copy_to_user(user_dest, dest_orig, alloc_size))
3170		ret = -EFAULT;
3171
3172	kfree(dest_orig);
3173out:
3174	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3175		ret = -EFAULT;
3176
3177	return ret;
3178}
3179
3180static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3181					    void __user *argp)
3182{
3183	struct btrfs_trans_handle *trans;
3184	u64 transid;
3185
3186	/*
3187	 * Start orphan cleanup here for the given root in case it hasn't been
3188	 * started already by other means. Errors are handled in the other
3189	 * functions during transaction commit.
3190	 */
3191	btrfs_orphan_cleanup(root);
3192
3193	trans = btrfs_attach_transaction_barrier(root);
3194	if (IS_ERR(trans)) {
3195		if (PTR_ERR(trans) != -ENOENT)
3196			return PTR_ERR(trans);
3197
3198		/* No running transaction, don't bother */
3199		transid = btrfs_get_last_trans_committed(root->fs_info);
3200		goto out;
3201	}
3202	transid = trans->transid;
3203	btrfs_commit_transaction_async(trans);
3204out:
3205	if (argp)
3206		if (copy_to_user(argp, &transid, sizeof(transid)))
3207			return -EFAULT;
3208	return 0;
3209}
3210
3211static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3212					   void __user *argp)
3213{
3214	/* By default wait for the current transaction. */
3215	u64 transid = 0;
3216
3217	if (argp)
3218		if (copy_from_user(&transid, argp, sizeof(transid)))
3219			return -EFAULT;
3220
3221	return btrfs_wait_for_commit(fs_info, transid);
3222}
3223
3224static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3225{
3226	struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file));
3227	struct btrfs_ioctl_scrub_args *sa;
3228	int ret;
3229
3230	if (!capable(CAP_SYS_ADMIN))
3231		return -EPERM;
3232
3233	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3234		btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3235		return -EINVAL;
3236	}
3237
3238	sa = memdup_user(arg, sizeof(*sa));
3239	if (IS_ERR(sa))
3240		return PTR_ERR(sa);
3241
3242	if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3243		ret = -EOPNOTSUPP;
3244		goto out;
3245	}
3246
3247	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3248		ret = mnt_want_write_file(file);
3249		if (ret)
3250			goto out;
3251	}
3252
3253	ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3254			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3255			      0);
3256
3257	/*
3258	 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3259	 * error. This is important as it allows user space to know how much
3260	 * progress scrub has done. For example, if scrub is canceled we get
3261	 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3262	 * space. Later user space can inspect the progress from the structure
3263	 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3264	 * previously (btrfs-progs does this).
3265	 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3266	 * then return -EFAULT to signal the structure was not copied or it may
3267	 * be corrupt and unreliable due to a partial copy.
3268	 */
3269	if (copy_to_user(arg, sa, sizeof(*sa)))
3270		ret = -EFAULT;
3271
3272	if (!(sa->flags & BTRFS_SCRUB_READONLY))
3273		mnt_drop_write_file(file);
3274out:
3275	kfree(sa);
3276	return ret;
3277}
3278
3279static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3280{
3281	if (!capable(CAP_SYS_ADMIN))
3282		return -EPERM;
3283
3284	return btrfs_scrub_cancel(fs_info);
3285}
3286
3287static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3288				       void __user *arg)
3289{
3290	struct btrfs_ioctl_scrub_args *sa;
3291	int ret;
3292
3293	if (!capable(CAP_SYS_ADMIN))
3294		return -EPERM;
3295
3296	sa = memdup_user(arg, sizeof(*sa));
3297	if (IS_ERR(sa))
3298		return PTR_ERR(sa);
3299
3300	ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3301
3302	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3303		ret = -EFAULT;
3304
3305	kfree(sa);
3306	return ret;
3307}
3308
3309static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3310				      void __user *arg)
3311{
3312	struct btrfs_ioctl_get_dev_stats *sa;
3313	int ret;
3314
3315	sa = memdup_user(arg, sizeof(*sa));
3316	if (IS_ERR(sa))
3317		return PTR_ERR(sa);
3318
3319	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3320		kfree(sa);
3321		return -EPERM;
3322	}
3323
3324	ret = btrfs_get_dev_stats(fs_info, sa);
3325
3326	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3327		ret = -EFAULT;
3328
3329	kfree(sa);
3330	return ret;
3331}
3332
3333static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3334				    void __user *arg)
3335{
3336	struct btrfs_ioctl_dev_replace_args *p;
3337	int ret;
3338
3339	if (!capable(CAP_SYS_ADMIN))
3340		return -EPERM;
3341
3342	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3343		btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3344		return -EINVAL;
3345	}
3346
3347	p = memdup_user(arg, sizeof(*p));
3348	if (IS_ERR(p))
3349		return PTR_ERR(p);
3350
3351	switch (p->cmd) {
3352	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3353		if (sb_rdonly(fs_info->sb)) {
3354			ret = -EROFS;
3355			goto out;
3356		}
3357		if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3358			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3359		} else {
3360			ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3361			btrfs_exclop_finish(fs_info);
3362		}
3363		break;
3364	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3365		btrfs_dev_replace_status(fs_info, p);
3366		ret = 0;
3367		break;
3368	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3369		p->result = btrfs_dev_replace_cancel(fs_info);
3370		ret = 0;
3371		break;
3372	default:
3373		ret = -EINVAL;
3374		break;
3375	}
3376
3377	if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3378		ret = -EFAULT;
3379out:
3380	kfree(p);
3381	return ret;
3382}
3383
3384static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3385{
3386	int ret = 0;
3387	int i;
3388	u64 rel_ptr;
3389	int size;
3390	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3391	struct inode_fs_paths *ipath = NULL;
3392	struct btrfs_path *path;
3393
3394	if (!capable(CAP_DAC_READ_SEARCH))
3395		return -EPERM;
3396
3397	path = btrfs_alloc_path();
3398	if (!path) {
3399		ret = -ENOMEM;
3400		goto out;
3401	}
3402
3403	ipa = memdup_user(arg, sizeof(*ipa));
3404	if (IS_ERR(ipa)) {
3405		ret = PTR_ERR(ipa);
3406		ipa = NULL;
3407		goto out;
3408	}
3409
3410	size = min_t(u32, ipa->size, 4096);
3411	ipath = init_ipath(size, root, path);
3412	if (IS_ERR(ipath)) {
3413		ret = PTR_ERR(ipath);
3414		ipath = NULL;
3415		goto out;
3416	}
3417
3418	ret = paths_from_inode(ipa->inum, ipath);
3419	if (ret < 0)
3420		goto out;
3421
3422	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3423		rel_ptr = ipath->fspath->val[i] -
3424			  (u64)(unsigned long)ipath->fspath->val;
3425		ipath->fspath->val[i] = rel_ptr;
3426	}
3427
3428	btrfs_free_path(path);
3429	path = NULL;
3430	ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3431			   ipath->fspath, size);
3432	if (ret) {
3433		ret = -EFAULT;
3434		goto out;
3435	}
3436
3437out:
3438	btrfs_free_path(path);
3439	free_ipath(ipath);
3440	kfree(ipa);
3441
3442	return ret;
3443}
3444
3445static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3446					void __user *arg, int version)
3447{
3448	int ret = 0;
3449	int size;
3450	struct btrfs_ioctl_logical_ino_args *loi;
3451	struct btrfs_data_container *inodes = NULL;
3452	struct btrfs_path *path = NULL;
3453	bool ignore_offset;
3454
3455	if (!capable(CAP_SYS_ADMIN))
3456		return -EPERM;
3457
3458	loi = memdup_user(arg, sizeof(*loi));
3459	if (IS_ERR(loi))
3460		return PTR_ERR(loi);
3461
3462	if (version == 1) {
3463		ignore_offset = false;
3464		size = min_t(u32, loi->size, SZ_64K);
3465	} else {
3466		/* All reserved bits must be 0 for now */
3467		if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3468			ret = -EINVAL;
3469			goto out_loi;
3470		}
3471		/* Only accept flags we have defined so far */
3472		if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3473			ret = -EINVAL;
3474			goto out_loi;
3475		}
3476		ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3477		size = min_t(u32, loi->size, SZ_16M);
3478	}
3479
3480	inodes = init_data_container(size);
3481	if (IS_ERR(inodes)) {
3482		ret = PTR_ERR(inodes);
3483		goto out_loi;
3484	}
3485
3486	path = btrfs_alloc_path();
3487	if (!path) {
3488		ret = -ENOMEM;
3489		goto out;
3490	}
3491	ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3492					  inodes, ignore_offset);
3493	btrfs_free_path(path);
3494	if (ret == -EINVAL)
3495		ret = -ENOENT;
3496	if (ret < 0)
3497		goto out;
3498
3499	ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3500			   size);
3501	if (ret)
3502		ret = -EFAULT;
3503
3504out:
3505	kvfree(inodes);
3506out_loi:
3507	kfree(loi);
3508
3509	return ret;
3510}
3511
3512void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3513			       struct btrfs_ioctl_balance_args *bargs)
3514{
3515	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3516
3517	bargs->flags = bctl->flags;
3518
3519	if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3520		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3521	if (atomic_read(&fs_info->balance_pause_req))
3522		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3523	if (atomic_read(&fs_info->balance_cancel_req))
3524		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3525
3526	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3527	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3528	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3529
3530	spin_lock(&fs_info->balance_lock);
3531	memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3532	spin_unlock(&fs_info->balance_lock);
3533}
3534
3535/*
3536 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3537 * required.
3538 *
3539 * @fs_info:       the filesystem
3540 * @excl_acquired: ptr to boolean value which is set to false in case balance
3541 *                 is being resumed
3542 *
3543 * Return 0 on success in which case both fs_info::balance is acquired as well
3544 * as exclusive ops are blocked. In case of failure return an error code.
3545 */
3546static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3547{
3548	int ret;
3549
3550	/*
3551	 * Exclusive operation is locked. Three possibilities:
3552	 *   (1) some other op is running
3553	 *   (2) balance is running
3554	 *   (3) balance is paused -- special case (think resume)
3555	 */
3556	while (1) {
3557		if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3558			*excl_acquired = true;
3559			mutex_lock(&fs_info->balance_mutex);
3560			return 0;
3561		}
3562
3563		mutex_lock(&fs_info->balance_mutex);
3564		if (fs_info->balance_ctl) {
3565			/* This is either (2) or (3) */
3566			if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3567				/* This is (2) */
3568				ret = -EINPROGRESS;
3569				goto out_failure;
3570
3571			} else {
3572				mutex_unlock(&fs_info->balance_mutex);
3573				/*
3574				 * Lock released to allow other waiters to
3575				 * continue, we'll reexamine the status again.
3576				 */
3577				mutex_lock(&fs_info->balance_mutex);
3578
3579				if (fs_info->balance_ctl &&
3580				    !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3581					/* This is (3) */
3582					*excl_acquired = false;
3583					return 0;
3584				}
3585			}
3586		} else {
3587			/* This is (1) */
3588			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3589			goto out_failure;
3590		}
3591
3592		mutex_unlock(&fs_info->balance_mutex);
3593	}
3594
3595out_failure:
3596	mutex_unlock(&fs_info->balance_mutex);
3597	*excl_acquired = false;
3598	return ret;
3599}
3600
3601static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3602{
3603	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3604	struct btrfs_fs_info *fs_info = root->fs_info;
3605	struct btrfs_ioctl_balance_args *bargs;
3606	struct btrfs_balance_control *bctl;
3607	bool need_unlock = true;
3608	int ret;
3609
3610	if (!capable(CAP_SYS_ADMIN))
3611		return -EPERM;
3612
3613	ret = mnt_want_write_file(file);
3614	if (ret)
3615		return ret;
3616
3617	bargs = memdup_user(arg, sizeof(*bargs));
3618	if (IS_ERR(bargs)) {
3619		ret = PTR_ERR(bargs);
3620		bargs = NULL;
3621		goto out;
3622	}
3623
3624	ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3625	if (ret)
3626		goto out;
3627
3628	lockdep_assert_held(&fs_info->balance_mutex);
3629
3630	if (bargs->flags & BTRFS_BALANCE_RESUME) {
3631		if (!fs_info->balance_ctl) {
3632			ret = -ENOTCONN;
3633			goto out_unlock;
3634		}
3635
3636		bctl = fs_info->balance_ctl;
3637		spin_lock(&fs_info->balance_lock);
3638		bctl->flags |= BTRFS_BALANCE_RESUME;
3639		spin_unlock(&fs_info->balance_lock);
3640		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3641
3642		goto do_balance;
3643	}
3644
3645	if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3646		ret = -EINVAL;
3647		goto out_unlock;
3648	}
3649
3650	if (fs_info->balance_ctl) {
3651		ret = -EINPROGRESS;
3652		goto out_unlock;
3653	}
3654
3655	bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3656	if (!bctl) {
3657		ret = -ENOMEM;
3658		goto out_unlock;
3659	}
3660
3661	memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3662	memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3663	memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3664
3665	bctl->flags = bargs->flags;
3666do_balance:
3667	/*
3668	 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3669	 * bctl is freed in reset_balance_state, or, if restriper was paused
3670	 * all the way until unmount, in free_fs_info.  The flag should be
3671	 * cleared after reset_balance_state.
3672	 */
3673	need_unlock = false;
3674
3675	ret = btrfs_balance(fs_info, bctl, bargs);
3676	bctl = NULL;
3677
3678	if (ret == 0 || ret == -ECANCELED) {
3679		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3680			ret = -EFAULT;
3681	}
3682
3683	kfree(bctl);
3684out_unlock:
3685	mutex_unlock(&fs_info->balance_mutex);
3686	if (need_unlock)
3687		btrfs_exclop_finish(fs_info);
3688out:
3689	mnt_drop_write_file(file);
3690	kfree(bargs);
3691	return ret;
3692}
3693
3694static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3695{
3696	if (!capable(CAP_SYS_ADMIN))
3697		return -EPERM;
3698
3699	switch (cmd) {
3700	case BTRFS_BALANCE_CTL_PAUSE:
3701		return btrfs_pause_balance(fs_info);
3702	case BTRFS_BALANCE_CTL_CANCEL:
3703		return btrfs_cancel_balance(fs_info);
3704	}
3705
3706	return -EINVAL;
3707}
3708
3709static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3710					 void __user *arg)
3711{
3712	struct btrfs_ioctl_balance_args *bargs;
3713	int ret = 0;
3714
3715	if (!capable(CAP_SYS_ADMIN))
3716		return -EPERM;
3717
3718	mutex_lock(&fs_info->balance_mutex);
3719	if (!fs_info->balance_ctl) {
3720		ret = -ENOTCONN;
3721		goto out;
3722	}
3723
3724	bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3725	if (!bargs) {
3726		ret = -ENOMEM;
3727		goto out;
3728	}
3729
3730	btrfs_update_ioctl_balance_args(fs_info, bargs);
3731
3732	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3733		ret = -EFAULT;
3734
3735	kfree(bargs);
3736out:
3737	mutex_unlock(&fs_info->balance_mutex);
3738	return ret;
3739}
3740
3741static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3742{
3743	struct inode *inode = file_inode(file);
3744	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3745	struct btrfs_ioctl_quota_ctl_args *sa;
3746	int ret;
3747
3748	if (!capable(CAP_SYS_ADMIN))
3749		return -EPERM;
3750
3751	ret = mnt_want_write_file(file);
3752	if (ret)
3753		return ret;
3754
3755	sa = memdup_user(arg, sizeof(*sa));
3756	if (IS_ERR(sa)) {
3757		ret = PTR_ERR(sa);
3758		goto drop_write;
3759	}
3760
3761	switch (sa->cmd) {
3762	case BTRFS_QUOTA_CTL_ENABLE:
3763	case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA:
3764		down_write(&fs_info->subvol_sem);
3765		ret = btrfs_quota_enable(fs_info, sa);
3766		up_write(&fs_info->subvol_sem);
3767		break;
3768	case BTRFS_QUOTA_CTL_DISABLE:
3769		/*
3770		 * Lock the cleaner mutex to prevent races with concurrent
3771		 * relocation, because relocation may be building backrefs for
3772		 * blocks of the quota root while we are deleting the root. This
3773		 * is like dropping fs roots of deleted snapshots/subvolumes, we
3774		 * need the same protection.
3775		 *
3776		 * This also prevents races between concurrent tasks trying to
3777		 * disable quotas, because we will unlock and relock
3778		 * qgroup_ioctl_lock across BTRFS_FS_QUOTA_ENABLED changes.
3779		 *
3780		 * We take this here because we have the dependency of
3781		 *
3782		 * inode_lock -> subvol_sem
3783		 *
3784		 * because of rename.  With relocation we can prealloc extents,
3785		 * so that makes the dependency chain
3786		 *
3787		 * cleaner_mutex -> inode_lock -> subvol_sem
3788		 *
3789		 * so we must take the cleaner_mutex here before we take the
3790		 * subvol_sem.  The deadlock can't actually happen, but this
3791		 * quiets lockdep.
3792		 */
3793		mutex_lock(&fs_info->cleaner_mutex);
3794		down_write(&fs_info->subvol_sem);
3795		ret = btrfs_quota_disable(fs_info);
3796		up_write(&fs_info->subvol_sem);
3797		mutex_unlock(&fs_info->cleaner_mutex);
3798		break;
3799	default:
3800		ret = -EINVAL;
3801		break;
3802	}
3803
3804	kfree(sa);
3805drop_write:
3806	mnt_drop_write_file(file);
3807	return ret;
3808}
3809
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3810static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3811{
3812	struct inode *inode = file_inode(file);
3813	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3814	struct btrfs_root *root = BTRFS_I(inode)->root;
3815	struct btrfs_ioctl_qgroup_assign_args *sa;
 
3816	struct btrfs_trans_handle *trans;
3817	int ret;
3818	int err;
3819
3820	if (!capable(CAP_SYS_ADMIN))
3821		return -EPERM;
3822
 
 
 
3823	ret = mnt_want_write_file(file);
3824	if (ret)
3825		return ret;
3826
3827	sa = memdup_user(arg, sizeof(*sa));
3828	if (IS_ERR(sa)) {
3829		ret = PTR_ERR(sa);
3830		goto drop_write;
3831	}
3832
 
 
 
 
 
 
 
 
3833	trans = btrfs_join_transaction(root);
3834	if (IS_ERR(trans)) {
3835		ret = PTR_ERR(trans);
3836		goto out;
3837	}
3838
 
 
 
 
3839	if (sa->assign) {
3840		ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
 
3841	} else {
3842		ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3843	}
3844
3845	/* update qgroup status and info */
3846	mutex_lock(&fs_info->qgroup_ioctl_lock);
3847	err = btrfs_run_qgroups(trans);
3848	mutex_unlock(&fs_info->qgroup_ioctl_lock);
3849	if (err < 0)
3850		btrfs_handle_fs_error(fs_info, err,
3851				      "failed to update qgroup status and info");
 
3852	err = btrfs_end_transaction(trans);
3853	if (err && !ret)
3854		ret = err;
3855
3856out:
 
3857	kfree(sa);
3858drop_write:
3859	mnt_drop_write_file(file);
3860	return ret;
3861}
3862
3863static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3864{
3865	struct inode *inode = file_inode(file);
3866	struct btrfs_root *root = BTRFS_I(inode)->root;
3867	struct btrfs_ioctl_qgroup_create_args *sa;
3868	struct btrfs_trans_handle *trans;
3869	int ret;
3870	int err;
3871
3872	if (!capable(CAP_SYS_ADMIN))
3873		return -EPERM;
3874
 
 
 
3875	ret = mnt_want_write_file(file);
3876	if (ret)
3877		return ret;
3878
3879	sa = memdup_user(arg, sizeof(*sa));
3880	if (IS_ERR(sa)) {
3881		ret = PTR_ERR(sa);
3882		goto drop_write;
3883	}
3884
3885	if (!sa->qgroupid) {
3886		ret = -EINVAL;
3887		goto out;
3888	}
3889
3890	if (sa->create && is_fstree(sa->qgroupid)) {
3891		ret = -EINVAL;
3892		goto out;
3893	}
3894
3895	trans = btrfs_join_transaction(root);
3896	if (IS_ERR(trans)) {
3897		ret = PTR_ERR(trans);
3898		goto out;
3899	}
3900
3901	if (sa->create) {
3902		ret = btrfs_create_qgroup(trans, sa->qgroupid);
3903	} else {
3904		ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3905	}
3906
3907	err = btrfs_end_transaction(trans);
3908	if (err && !ret)
3909		ret = err;
3910
3911out:
3912	kfree(sa);
3913drop_write:
3914	mnt_drop_write_file(file);
3915	return ret;
3916}
3917
3918static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3919{
3920	struct inode *inode = file_inode(file);
3921	struct btrfs_root *root = BTRFS_I(inode)->root;
3922	struct btrfs_ioctl_qgroup_limit_args *sa;
3923	struct btrfs_trans_handle *trans;
3924	int ret;
3925	int err;
3926	u64 qgroupid;
3927
3928	if (!capable(CAP_SYS_ADMIN))
3929		return -EPERM;
3930
 
 
 
3931	ret = mnt_want_write_file(file);
3932	if (ret)
3933		return ret;
3934
3935	sa = memdup_user(arg, sizeof(*sa));
3936	if (IS_ERR(sa)) {
3937		ret = PTR_ERR(sa);
3938		goto drop_write;
3939	}
3940
3941	trans = btrfs_join_transaction(root);
3942	if (IS_ERR(trans)) {
3943		ret = PTR_ERR(trans);
3944		goto out;
3945	}
3946
3947	qgroupid = sa->qgroupid;
3948	if (!qgroupid) {
3949		/* take the current subvol as qgroup */
3950		qgroupid = root->root_key.objectid;
3951	}
3952
3953	ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3954
3955	err = btrfs_end_transaction(trans);
3956	if (err && !ret)
3957		ret = err;
3958
3959out:
3960	kfree(sa);
3961drop_write:
3962	mnt_drop_write_file(file);
3963	return ret;
3964}
3965
3966static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3967{
3968	struct inode *inode = file_inode(file);
3969	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3970	struct btrfs_ioctl_quota_rescan_args *qsa;
3971	int ret;
3972
3973	if (!capable(CAP_SYS_ADMIN))
3974		return -EPERM;
3975
 
 
 
3976	ret = mnt_want_write_file(file);
3977	if (ret)
3978		return ret;
3979
3980	qsa = memdup_user(arg, sizeof(*qsa));
3981	if (IS_ERR(qsa)) {
3982		ret = PTR_ERR(qsa);
3983		goto drop_write;
3984	}
3985
3986	if (qsa->flags) {
3987		ret = -EINVAL;
3988		goto out;
3989	}
3990
3991	ret = btrfs_qgroup_rescan(fs_info);
3992
3993out:
3994	kfree(qsa);
3995drop_write:
3996	mnt_drop_write_file(file);
3997	return ret;
3998}
3999
4000static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4001						void __user *arg)
4002{
4003	struct btrfs_ioctl_quota_rescan_args qsa = {0};
4004
4005	if (!capable(CAP_SYS_ADMIN))
4006		return -EPERM;
4007
4008	if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4009		qsa.flags = 1;
4010		qsa.progress = fs_info->qgroup_rescan_progress.objectid;
4011	}
4012
4013	if (copy_to_user(arg, &qsa, sizeof(qsa)))
4014		return -EFAULT;
4015
4016	return 0;
4017}
4018
4019static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4020						void __user *arg)
4021{
4022	if (!capable(CAP_SYS_ADMIN))
4023		return -EPERM;
4024
4025	return btrfs_qgroup_wait_for_completion(fs_info, true);
4026}
4027
4028static long _btrfs_ioctl_set_received_subvol(struct file *file,
4029					    struct mnt_idmap *idmap,
4030					    struct btrfs_ioctl_received_subvol_args *sa)
4031{
4032	struct inode *inode = file_inode(file);
4033	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4034	struct btrfs_root *root = BTRFS_I(inode)->root;
4035	struct btrfs_root_item *root_item = &root->root_item;
4036	struct btrfs_trans_handle *trans;
4037	struct timespec64 ct = current_time(inode);
4038	int ret = 0;
4039	int received_uuid_changed;
4040
4041	if (!inode_owner_or_capable(idmap, inode))
4042		return -EPERM;
4043
4044	ret = mnt_want_write_file(file);
4045	if (ret < 0)
4046		return ret;
4047
4048	down_write(&fs_info->subvol_sem);
4049
4050	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4051		ret = -EINVAL;
4052		goto out;
4053	}
4054
4055	if (btrfs_root_readonly(root)) {
4056		ret = -EROFS;
4057		goto out;
4058	}
4059
4060	/*
4061	 * 1 - root item
4062	 * 2 - uuid items (received uuid + subvol uuid)
4063	 */
4064	trans = btrfs_start_transaction(root, 3);
4065	if (IS_ERR(trans)) {
4066		ret = PTR_ERR(trans);
4067		trans = NULL;
4068		goto out;
4069	}
4070
4071	sa->rtransid = trans->transid;
4072	sa->rtime.sec = ct.tv_sec;
4073	sa->rtime.nsec = ct.tv_nsec;
4074
4075	received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4076				       BTRFS_UUID_SIZE);
4077	if (received_uuid_changed &&
4078	    !btrfs_is_empty_uuid(root_item->received_uuid)) {
4079		ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4080					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4081					  root->root_key.objectid);
4082		if (ret && ret != -ENOENT) {
4083		        btrfs_abort_transaction(trans, ret);
4084		        btrfs_end_transaction(trans);
4085		        goto out;
4086		}
4087	}
4088	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4089	btrfs_set_root_stransid(root_item, sa->stransid);
4090	btrfs_set_root_rtransid(root_item, sa->rtransid);
4091	btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4092	btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4093	btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4094	btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4095
4096	ret = btrfs_update_root(trans, fs_info->tree_root,
4097				&root->root_key, &root->root_item);
4098	if (ret < 0) {
4099		btrfs_end_transaction(trans);
4100		goto out;
4101	}
4102	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4103		ret = btrfs_uuid_tree_add(trans, sa->uuid,
4104					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4105					  root->root_key.objectid);
4106		if (ret < 0 && ret != -EEXIST) {
4107			btrfs_abort_transaction(trans, ret);
4108			btrfs_end_transaction(trans);
4109			goto out;
4110		}
4111	}
4112	ret = btrfs_commit_transaction(trans);
4113out:
4114	up_write(&fs_info->subvol_sem);
4115	mnt_drop_write_file(file);
4116	return ret;
4117}
4118
4119#ifdef CONFIG_64BIT
4120static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4121						void __user *arg)
4122{
4123	struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4124	struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4125	int ret = 0;
4126
4127	args32 = memdup_user(arg, sizeof(*args32));
4128	if (IS_ERR(args32))
4129		return PTR_ERR(args32);
4130
4131	args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4132	if (!args64) {
4133		ret = -ENOMEM;
4134		goto out;
4135	}
4136
4137	memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4138	args64->stransid = args32->stransid;
4139	args64->rtransid = args32->rtransid;
4140	args64->stime.sec = args32->stime.sec;
4141	args64->stime.nsec = args32->stime.nsec;
4142	args64->rtime.sec = args32->rtime.sec;
4143	args64->rtime.nsec = args32->rtime.nsec;
4144	args64->flags = args32->flags;
4145
4146	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4147	if (ret)
4148		goto out;
4149
4150	memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4151	args32->stransid = args64->stransid;
4152	args32->rtransid = args64->rtransid;
4153	args32->stime.sec = args64->stime.sec;
4154	args32->stime.nsec = args64->stime.nsec;
4155	args32->rtime.sec = args64->rtime.sec;
4156	args32->rtime.nsec = args64->rtime.nsec;
4157	args32->flags = args64->flags;
4158
4159	ret = copy_to_user(arg, args32, sizeof(*args32));
4160	if (ret)
4161		ret = -EFAULT;
4162
4163out:
4164	kfree(args32);
4165	kfree(args64);
4166	return ret;
4167}
4168#endif
4169
4170static long btrfs_ioctl_set_received_subvol(struct file *file,
4171					    void __user *arg)
4172{
4173	struct btrfs_ioctl_received_subvol_args *sa = NULL;
4174	int ret = 0;
4175
4176	sa = memdup_user(arg, sizeof(*sa));
4177	if (IS_ERR(sa))
4178		return PTR_ERR(sa);
4179
4180	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4181
4182	if (ret)
4183		goto out;
4184
4185	ret = copy_to_user(arg, sa, sizeof(*sa));
4186	if (ret)
4187		ret = -EFAULT;
4188
4189out:
4190	kfree(sa);
4191	return ret;
4192}
4193
4194static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4195					void __user *arg)
4196{
4197	size_t len;
4198	int ret;
4199	char label[BTRFS_LABEL_SIZE];
4200
4201	spin_lock(&fs_info->super_lock);
4202	memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4203	spin_unlock(&fs_info->super_lock);
4204
4205	len = strnlen(label, BTRFS_LABEL_SIZE);
4206
4207	if (len == BTRFS_LABEL_SIZE) {
4208		btrfs_warn(fs_info,
4209			   "label is too long, return the first %zu bytes",
4210			   --len);
4211	}
4212
4213	ret = copy_to_user(arg, label, len);
4214
4215	return ret ? -EFAULT : 0;
4216}
4217
4218static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4219{
4220	struct inode *inode = file_inode(file);
4221	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4222	struct btrfs_root *root = BTRFS_I(inode)->root;
4223	struct btrfs_super_block *super_block = fs_info->super_copy;
4224	struct btrfs_trans_handle *trans;
4225	char label[BTRFS_LABEL_SIZE];
4226	int ret;
4227
4228	if (!capable(CAP_SYS_ADMIN))
4229		return -EPERM;
4230
4231	if (copy_from_user(label, arg, sizeof(label)))
4232		return -EFAULT;
4233
4234	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4235		btrfs_err(fs_info,
4236			  "unable to set label with more than %d bytes",
4237			  BTRFS_LABEL_SIZE - 1);
4238		return -EINVAL;
4239	}
4240
4241	ret = mnt_want_write_file(file);
4242	if (ret)
4243		return ret;
4244
4245	trans = btrfs_start_transaction(root, 0);
4246	if (IS_ERR(trans)) {
4247		ret = PTR_ERR(trans);
4248		goto out_unlock;
4249	}
4250
4251	spin_lock(&fs_info->super_lock);
4252	strcpy(super_block->label, label);
4253	spin_unlock(&fs_info->super_lock);
4254	ret = btrfs_commit_transaction(trans);
4255
4256out_unlock:
4257	mnt_drop_write_file(file);
4258	return ret;
4259}
4260
4261#define INIT_FEATURE_FLAGS(suffix) \
4262	{ .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4263	  .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4264	  .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4265
4266int btrfs_ioctl_get_supported_features(void __user *arg)
4267{
4268	static const struct btrfs_ioctl_feature_flags features[3] = {
4269		INIT_FEATURE_FLAGS(SUPP),
4270		INIT_FEATURE_FLAGS(SAFE_SET),
4271		INIT_FEATURE_FLAGS(SAFE_CLEAR)
4272	};
4273
4274	if (copy_to_user(arg, &features, sizeof(features)))
4275		return -EFAULT;
4276
4277	return 0;
4278}
4279
4280static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4281					void __user *arg)
4282{
4283	struct btrfs_super_block *super_block = fs_info->super_copy;
4284	struct btrfs_ioctl_feature_flags features;
4285
4286	features.compat_flags = btrfs_super_compat_flags(super_block);
4287	features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4288	features.incompat_flags = btrfs_super_incompat_flags(super_block);
4289
4290	if (copy_to_user(arg, &features, sizeof(features)))
4291		return -EFAULT;
4292
4293	return 0;
4294}
4295
4296static int check_feature_bits(struct btrfs_fs_info *fs_info,
4297			      enum btrfs_feature_set set,
4298			      u64 change_mask, u64 flags, u64 supported_flags,
4299			      u64 safe_set, u64 safe_clear)
4300{
4301	const char *type = btrfs_feature_set_name(set);
4302	char *names;
4303	u64 disallowed, unsupported;
4304	u64 set_mask = flags & change_mask;
4305	u64 clear_mask = ~flags & change_mask;
4306
4307	unsupported = set_mask & ~supported_flags;
4308	if (unsupported) {
4309		names = btrfs_printable_features(set, unsupported);
4310		if (names) {
4311			btrfs_warn(fs_info,
4312				   "this kernel does not support the %s feature bit%s",
4313				   names, strchr(names, ',') ? "s" : "");
4314			kfree(names);
4315		} else
4316			btrfs_warn(fs_info,
4317				   "this kernel does not support %s bits 0x%llx",
4318				   type, unsupported);
4319		return -EOPNOTSUPP;
4320	}
4321
4322	disallowed = set_mask & ~safe_set;
4323	if (disallowed) {
4324		names = btrfs_printable_features(set, disallowed);
4325		if (names) {
4326			btrfs_warn(fs_info,
4327				   "can't set the %s feature bit%s while mounted",
4328				   names, strchr(names, ',') ? "s" : "");
4329			kfree(names);
4330		} else
4331			btrfs_warn(fs_info,
4332				   "can't set %s bits 0x%llx while mounted",
4333				   type, disallowed);
4334		return -EPERM;
4335	}
4336
4337	disallowed = clear_mask & ~safe_clear;
4338	if (disallowed) {
4339		names = btrfs_printable_features(set, disallowed);
4340		if (names) {
4341			btrfs_warn(fs_info,
4342				   "can't clear the %s feature bit%s while mounted",
4343				   names, strchr(names, ',') ? "s" : "");
4344			kfree(names);
4345		} else
4346			btrfs_warn(fs_info,
4347				   "can't clear %s bits 0x%llx while mounted",
4348				   type, disallowed);
4349		return -EPERM;
4350	}
4351
4352	return 0;
4353}
4354
4355#define check_feature(fs_info, change_mask, flags, mask_base)	\
4356check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags,	\
4357		   BTRFS_FEATURE_ ## mask_base ## _SUPP,	\
4358		   BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,	\
4359		   BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4360
4361static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4362{
4363	struct inode *inode = file_inode(file);
4364	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4365	struct btrfs_root *root = BTRFS_I(inode)->root;
4366	struct btrfs_super_block *super_block = fs_info->super_copy;
4367	struct btrfs_ioctl_feature_flags flags[2];
4368	struct btrfs_trans_handle *trans;
4369	u64 newflags;
4370	int ret;
4371
4372	if (!capable(CAP_SYS_ADMIN))
4373		return -EPERM;
4374
4375	if (copy_from_user(flags, arg, sizeof(flags)))
4376		return -EFAULT;
4377
4378	/* Nothing to do */
4379	if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4380	    !flags[0].incompat_flags)
4381		return 0;
4382
4383	ret = check_feature(fs_info, flags[0].compat_flags,
4384			    flags[1].compat_flags, COMPAT);
4385	if (ret)
4386		return ret;
4387
4388	ret = check_feature(fs_info, flags[0].compat_ro_flags,
4389			    flags[1].compat_ro_flags, COMPAT_RO);
4390	if (ret)
4391		return ret;
4392
4393	ret = check_feature(fs_info, flags[0].incompat_flags,
4394			    flags[1].incompat_flags, INCOMPAT);
4395	if (ret)
4396		return ret;
4397
4398	ret = mnt_want_write_file(file);
4399	if (ret)
4400		return ret;
4401
4402	trans = btrfs_start_transaction(root, 0);
4403	if (IS_ERR(trans)) {
4404		ret = PTR_ERR(trans);
4405		goto out_drop_write;
4406	}
4407
4408	spin_lock(&fs_info->super_lock);
4409	newflags = btrfs_super_compat_flags(super_block);
4410	newflags |= flags[0].compat_flags & flags[1].compat_flags;
4411	newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4412	btrfs_set_super_compat_flags(super_block, newflags);
4413
4414	newflags = btrfs_super_compat_ro_flags(super_block);
4415	newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4416	newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4417	btrfs_set_super_compat_ro_flags(super_block, newflags);
4418
4419	newflags = btrfs_super_incompat_flags(super_block);
4420	newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4421	newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4422	btrfs_set_super_incompat_flags(super_block, newflags);
4423	spin_unlock(&fs_info->super_lock);
4424
4425	ret = btrfs_commit_transaction(trans);
4426out_drop_write:
4427	mnt_drop_write_file(file);
4428
4429	return ret;
4430}
4431
4432static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
4433{
4434	struct btrfs_ioctl_send_args *arg;
4435	int ret;
4436
4437	if (compat) {
4438#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4439		struct btrfs_ioctl_send_args_32 args32 = { 0 };
4440
4441		ret = copy_from_user(&args32, argp, sizeof(args32));
4442		if (ret)
4443			return -EFAULT;
4444		arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4445		if (!arg)
4446			return -ENOMEM;
4447		arg->send_fd = args32.send_fd;
4448		arg->clone_sources_count = args32.clone_sources_count;
4449		arg->clone_sources = compat_ptr(args32.clone_sources);
4450		arg->parent_root = args32.parent_root;
4451		arg->flags = args32.flags;
4452		arg->version = args32.version;
4453		memcpy(arg->reserved, args32.reserved,
4454		       sizeof(args32.reserved));
4455#else
4456		return -ENOTTY;
4457#endif
4458	} else {
4459		arg = memdup_user(argp, sizeof(*arg));
4460		if (IS_ERR(arg))
4461			return PTR_ERR(arg);
4462	}
4463	ret = btrfs_ioctl_send(inode, arg);
4464	kfree(arg);
4465	return ret;
4466}
4467
4468static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4469				    bool compat)
4470{
4471	struct btrfs_ioctl_encoded_io_args args = { 0 };
4472	size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4473					     flags);
4474	size_t copy_end;
 
 
 
4475	struct iovec iovstack[UIO_FASTIOV];
4476	struct iovec *iov = iovstack;
4477	struct iov_iter iter;
4478	loff_t pos;
4479	struct kiocb kiocb;
4480	ssize_t ret;
 
 
4481
4482	if (!capable(CAP_SYS_ADMIN)) {
4483		ret = -EPERM;
4484		goto out_acct;
4485	}
4486
4487	if (compat) {
4488#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4489		struct btrfs_ioctl_encoded_io_args_32 args32;
4490
4491		copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4492				       flags);
4493		if (copy_from_user(&args32, argp, copy_end)) {
4494			ret = -EFAULT;
4495			goto out_acct;
4496		}
4497		args.iov = compat_ptr(args32.iov);
4498		args.iovcnt = args32.iovcnt;
4499		args.offset = args32.offset;
4500		args.flags = args32.flags;
4501#else
4502		return -ENOTTY;
4503#endif
4504	} else {
4505		copy_end = copy_end_kernel;
4506		if (copy_from_user(&args, argp, copy_end)) {
4507			ret = -EFAULT;
4508			goto out_acct;
4509		}
4510	}
4511	if (args.flags != 0) {
4512		ret = -EINVAL;
4513		goto out_acct;
4514	}
4515
4516	ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4517			   &iov, &iter);
4518	if (ret < 0)
4519		goto out_acct;
4520
4521	if (iov_iter_count(&iter) == 0) {
4522		ret = 0;
4523		goto out_iov;
4524	}
4525	pos = args.offset;
4526	ret = rw_verify_area(READ, file, &pos, args.len);
4527	if (ret < 0)
4528		goto out_iov;
4529
4530	init_sync_kiocb(&kiocb, file);
4531	kiocb.ki_pos = pos;
4532
4533	ret = btrfs_encoded_read(&kiocb, &iter, &args);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4534	if (ret >= 0) {
4535		fsnotify_access(file);
4536		if (copy_to_user(argp + copy_end,
4537				 (char *)&args + copy_end_kernel,
4538				 sizeof(args) - copy_end_kernel))
4539			ret = -EFAULT;
4540	}
4541
4542out_iov:
4543	kfree(iov);
4544out_acct:
4545	if (ret > 0)
4546		add_rchar(current, ret);
4547	inc_syscr(current);
4548	return ret;
4549}
4550
4551static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4552{
4553	struct btrfs_ioctl_encoded_io_args args;
4554	struct iovec iovstack[UIO_FASTIOV];
4555	struct iovec *iov = iovstack;
4556	struct iov_iter iter;
4557	loff_t pos;
4558	struct kiocb kiocb;
4559	ssize_t ret;
4560
4561	if (!capable(CAP_SYS_ADMIN)) {
4562		ret = -EPERM;
4563		goto out_acct;
4564	}
4565
4566	if (!(file->f_mode & FMODE_WRITE)) {
4567		ret = -EBADF;
4568		goto out_acct;
4569	}
4570
4571	if (compat) {
4572#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4573		struct btrfs_ioctl_encoded_io_args_32 args32;
4574
4575		if (copy_from_user(&args32, argp, sizeof(args32))) {
4576			ret = -EFAULT;
4577			goto out_acct;
4578		}
4579		args.iov = compat_ptr(args32.iov);
4580		args.iovcnt = args32.iovcnt;
4581		args.offset = args32.offset;
4582		args.flags = args32.flags;
4583		args.len = args32.len;
4584		args.unencoded_len = args32.unencoded_len;
4585		args.unencoded_offset = args32.unencoded_offset;
4586		args.compression = args32.compression;
4587		args.encryption = args32.encryption;
4588		memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4589#else
4590		return -ENOTTY;
4591#endif
4592	} else {
4593		if (copy_from_user(&args, argp, sizeof(args))) {
4594			ret = -EFAULT;
4595			goto out_acct;
4596		}
4597	}
4598
4599	ret = -EINVAL;
4600	if (args.flags != 0)
4601		goto out_acct;
4602	if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4603		goto out_acct;
4604	if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4605	    args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4606		goto out_acct;
4607	if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4608	    args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4609		goto out_acct;
4610	if (args.unencoded_offset > args.unencoded_len)
4611		goto out_acct;
4612	if (args.len > args.unencoded_len - args.unencoded_offset)
4613		goto out_acct;
4614
4615	ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4616			   &iov, &iter);
4617	if (ret < 0)
4618		goto out_acct;
4619
4620	if (iov_iter_count(&iter) == 0) {
4621		ret = 0;
4622		goto out_iov;
4623	}
4624	pos = args.offset;
4625	ret = rw_verify_area(WRITE, file, &pos, args.len);
4626	if (ret < 0)
4627		goto out_iov;
4628
4629	init_sync_kiocb(&kiocb, file);
4630	ret = kiocb_set_rw_flags(&kiocb, 0);
4631	if (ret)
4632		goto out_iov;
4633	kiocb.ki_pos = pos;
4634
4635	file_start_write(file);
4636
4637	ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4638	if (ret > 0)
4639		fsnotify_modify(file);
4640
4641	file_end_write(file);
4642out_iov:
4643	kfree(iov);
4644out_acct:
4645	if (ret > 0)
4646		add_wchar(current, ret);
4647	inc_syscw(current);
4648	return ret;
4649}
4650
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4651long btrfs_ioctl(struct file *file, unsigned int
4652		cmd, unsigned long arg)
4653{
4654	struct inode *inode = file_inode(file);
4655	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4656	struct btrfs_root *root = BTRFS_I(inode)->root;
4657	void __user *argp = (void __user *)arg;
4658
4659	switch (cmd) {
4660	case FS_IOC_GETVERSION:
4661		return btrfs_ioctl_getversion(inode, argp);
4662	case FS_IOC_GETFSLABEL:
4663		return btrfs_ioctl_get_fslabel(fs_info, argp);
4664	case FS_IOC_SETFSLABEL:
4665		return btrfs_ioctl_set_fslabel(file, argp);
4666	case FITRIM:
4667		return btrfs_ioctl_fitrim(fs_info, argp);
4668	case BTRFS_IOC_SNAP_CREATE:
4669		return btrfs_ioctl_snap_create(file, argp, 0);
4670	case BTRFS_IOC_SNAP_CREATE_V2:
4671		return btrfs_ioctl_snap_create_v2(file, argp, 0);
4672	case BTRFS_IOC_SUBVOL_CREATE:
4673		return btrfs_ioctl_snap_create(file, argp, 1);
4674	case BTRFS_IOC_SUBVOL_CREATE_V2:
4675		return btrfs_ioctl_snap_create_v2(file, argp, 1);
4676	case BTRFS_IOC_SNAP_DESTROY:
4677		return btrfs_ioctl_snap_destroy(file, argp, false);
4678	case BTRFS_IOC_SNAP_DESTROY_V2:
4679		return btrfs_ioctl_snap_destroy(file, argp, true);
4680	case BTRFS_IOC_SUBVOL_GETFLAGS:
4681		return btrfs_ioctl_subvol_getflags(inode, argp);
4682	case BTRFS_IOC_SUBVOL_SETFLAGS:
4683		return btrfs_ioctl_subvol_setflags(file, argp);
4684	case BTRFS_IOC_DEFAULT_SUBVOL:
4685		return btrfs_ioctl_default_subvol(file, argp);
4686	case BTRFS_IOC_DEFRAG:
4687		return btrfs_ioctl_defrag(file, NULL);
4688	case BTRFS_IOC_DEFRAG_RANGE:
4689		return btrfs_ioctl_defrag(file, argp);
4690	case BTRFS_IOC_RESIZE:
4691		return btrfs_ioctl_resize(file, argp);
4692	case BTRFS_IOC_ADD_DEV:
4693		return btrfs_ioctl_add_dev(fs_info, argp);
4694	case BTRFS_IOC_RM_DEV:
4695		return btrfs_ioctl_rm_dev(file, argp);
4696	case BTRFS_IOC_RM_DEV_V2:
4697		return btrfs_ioctl_rm_dev_v2(file, argp);
4698	case BTRFS_IOC_FS_INFO:
4699		return btrfs_ioctl_fs_info(fs_info, argp);
4700	case BTRFS_IOC_DEV_INFO:
4701		return btrfs_ioctl_dev_info(fs_info, argp);
4702	case BTRFS_IOC_TREE_SEARCH:
4703		return btrfs_ioctl_tree_search(inode, argp);
4704	case BTRFS_IOC_TREE_SEARCH_V2:
4705		return btrfs_ioctl_tree_search_v2(inode, argp);
4706	case BTRFS_IOC_INO_LOOKUP:
4707		return btrfs_ioctl_ino_lookup(root, argp);
4708	case BTRFS_IOC_INO_PATHS:
4709		return btrfs_ioctl_ino_to_path(root, argp);
4710	case BTRFS_IOC_LOGICAL_INO:
4711		return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4712	case BTRFS_IOC_LOGICAL_INO_V2:
4713		return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4714	case BTRFS_IOC_SPACE_INFO:
4715		return btrfs_ioctl_space_info(fs_info, argp);
4716	case BTRFS_IOC_SYNC: {
4717		int ret;
4718
4719		ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4720		if (ret)
4721			return ret;
4722		ret = btrfs_sync_fs(inode->i_sb, 1);
4723		/*
4724		 * The transaction thread may want to do more work,
4725		 * namely it pokes the cleaner kthread that will start
4726		 * processing uncleaned subvols.
4727		 */
4728		wake_up_process(fs_info->transaction_kthread);
4729		return ret;
4730	}
4731	case BTRFS_IOC_START_SYNC:
4732		return btrfs_ioctl_start_sync(root, argp);
4733	case BTRFS_IOC_WAIT_SYNC:
4734		return btrfs_ioctl_wait_sync(fs_info, argp);
4735	case BTRFS_IOC_SCRUB:
4736		return btrfs_ioctl_scrub(file, argp);
4737	case BTRFS_IOC_SCRUB_CANCEL:
4738		return btrfs_ioctl_scrub_cancel(fs_info);
4739	case BTRFS_IOC_SCRUB_PROGRESS:
4740		return btrfs_ioctl_scrub_progress(fs_info, argp);
4741	case BTRFS_IOC_BALANCE_V2:
4742		return btrfs_ioctl_balance(file, argp);
4743	case BTRFS_IOC_BALANCE_CTL:
4744		return btrfs_ioctl_balance_ctl(fs_info, arg);
4745	case BTRFS_IOC_BALANCE_PROGRESS:
4746		return btrfs_ioctl_balance_progress(fs_info, argp);
4747	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4748		return btrfs_ioctl_set_received_subvol(file, argp);
4749#ifdef CONFIG_64BIT
4750	case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4751		return btrfs_ioctl_set_received_subvol_32(file, argp);
4752#endif
4753	case BTRFS_IOC_SEND:
4754		return _btrfs_ioctl_send(inode, argp, false);
4755#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4756	case BTRFS_IOC_SEND_32:
4757		return _btrfs_ioctl_send(inode, argp, true);
4758#endif
4759	case BTRFS_IOC_GET_DEV_STATS:
4760		return btrfs_ioctl_get_dev_stats(fs_info, argp);
4761	case BTRFS_IOC_QUOTA_CTL:
4762		return btrfs_ioctl_quota_ctl(file, argp);
4763	case BTRFS_IOC_QGROUP_ASSIGN:
4764		return btrfs_ioctl_qgroup_assign(file, argp);
4765	case BTRFS_IOC_QGROUP_CREATE:
4766		return btrfs_ioctl_qgroup_create(file, argp);
4767	case BTRFS_IOC_QGROUP_LIMIT:
4768		return btrfs_ioctl_qgroup_limit(file, argp);
4769	case BTRFS_IOC_QUOTA_RESCAN:
4770		return btrfs_ioctl_quota_rescan(file, argp);
4771	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4772		return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4773	case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4774		return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4775	case BTRFS_IOC_DEV_REPLACE:
4776		return btrfs_ioctl_dev_replace(fs_info, argp);
4777	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4778		return btrfs_ioctl_get_supported_features(argp);
4779	case BTRFS_IOC_GET_FEATURES:
4780		return btrfs_ioctl_get_features(fs_info, argp);
4781	case BTRFS_IOC_SET_FEATURES:
4782		return btrfs_ioctl_set_features(file, argp);
4783	case BTRFS_IOC_GET_SUBVOL_INFO:
4784		return btrfs_ioctl_get_subvol_info(inode, argp);
4785	case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4786		return btrfs_ioctl_get_subvol_rootref(root, argp);
4787	case BTRFS_IOC_INO_LOOKUP_USER:
4788		return btrfs_ioctl_ino_lookup_user(file, argp);
4789	case FS_IOC_ENABLE_VERITY:
4790		return fsverity_ioctl_enable(file, (const void __user *)argp);
4791	case FS_IOC_MEASURE_VERITY:
4792		return fsverity_ioctl_measure(file, argp);
4793	case BTRFS_IOC_ENCODED_READ:
4794		return btrfs_ioctl_encoded_read(file, argp, false);
4795	case BTRFS_IOC_ENCODED_WRITE:
4796		return btrfs_ioctl_encoded_write(file, argp, false);
4797#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4798	case BTRFS_IOC_ENCODED_READ_32:
4799		return btrfs_ioctl_encoded_read(file, argp, true);
4800	case BTRFS_IOC_ENCODED_WRITE_32:
4801		return btrfs_ioctl_encoded_write(file, argp, true);
4802#endif
 
 
4803	}
4804
4805	return -ENOTTY;
4806}
4807
4808#ifdef CONFIG_COMPAT
4809long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4810{
4811	/*
4812	 * These all access 32-bit values anyway so no further
4813	 * handling is necessary.
4814	 */
4815	switch (cmd) {
4816	case FS_IOC32_GETVERSION:
4817		cmd = FS_IOC_GETVERSION;
4818		break;
4819	}
4820
4821	return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4822}
4823#endif