Linux Audio

Check our new training course

Buildroot integration, development and maintenance

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