ioctl.c

   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * Copyright (C) 2007 Oracle.  All rights reserved.
   4 */
   5
   6#include <linux/kernel.h>
   7#include <linux/bio.h>
   8#include <linux/file.h>
   9#include <linux/fs.h>
  10#include <linux/fsnotify.h>
  11#include <linux/pagemap.h>
  12#include <linux/highmem.h>
  13#include <linux/time.h>
  14#include <linux/string.h>
  15#include <linux/backing-dev.h>
  16#include <linux/mount.h>
  17#include <linux/namei.h>
  18#include <linux/writeback.h>
  19#include <linux/compat.h>
  20#include <linux/security.h>
  21#include <linux/xattr.h>
  22#include <linux/mm.h>
  23#include <linux/slab.h>
  24#include <linux/blkdev.h>
  25#include <linux/uuid.h>
  26#include <linux/btrfs.h>
  27#include <linux/uaccess.h>
  28#include <linux/iversion.h>
  29#include <linux/fileattr.h>
  30#include <linux/fsverity.h>
  31#include <linux/sched/xacct.h>
  32#include <linux/io_uring/cmd.h>
  33#include "ctree.h"
  34#include "disk-io.h"
  35#include "export.h"
  36#include "transaction.h"
  37#include "btrfs_inode.h"
  38#include "volumes.h"
  39#include "locking.h"
  40#include "backref.h"
  41#include "send.h"
  42#include "dev-replace.h"
  43#include "props.h"
  44#include "sysfs.h"
  45#include "qgroup.h"
  46#include "tree-log.h"
  47#include "compression.h"
  48#include "space-info.h"
  49#include "block-group.h"
  50#include "fs.h"
  51#include "accessors.h"
  52#include "extent-tree.h"
  53#include "root-tree.h"
  54#include "defrag.h"
  55#include "dir-item.h"
  56#include "uuid-tree.h"
  57#include "ioctl.h"
  58#include "file.h"
  59#include "scrub.h"
  60#include "super.h"
  61
  62#ifdef CONFIG_64BIT
  63/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
  64 * structures are incorrect, as the timespec structure from userspace
  65 * is 4 bytes too small. We define these alternatives here to teach
  66 * the kernel about the 32-bit struct packing.
  67 */
  68struct btrfs_ioctl_timespec_32 {
  69	__u64 sec;
  70	__u32 nsec;
  71} __attribute__ ((__packed__));
  72
  73struct btrfs_ioctl_received_subvol_args_32 {
  74	char	uuid[BTRFS_UUID_SIZE];	/* in */
  75	__u64	stransid;		/* in */
  76	__u64	rtransid;		/* out */
  77	struct btrfs_ioctl_timespec_32 stime; /* in */
  78	struct btrfs_ioctl_timespec_32 rtime; /* out */
  79	__u64	flags;			/* in */
  80	__u64	reserved[16];		/* in */
  81} __attribute__ ((__packed__));
  82
  83#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
  84				struct btrfs_ioctl_received_subvol_args_32)
  85#endif
  86
  87#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
  88struct btrfs_ioctl_send_args_32 {
  89	__s64 send_fd;			/* in */
  90	__u64 clone_sources_count;	/* in */
  91	compat_uptr_t clone_sources;	/* in */
  92	__u64 parent_root;		/* in */
  93	__u64 flags;			/* in */
  94	__u32 version;			/* in */
  95	__u8  reserved[28];		/* in */
  96} __attribute__ ((__packed__));
  97
  98#define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
  99			       struct btrfs_ioctl_send_args_32)
 100
 101struct btrfs_ioctl_encoded_io_args_32 {
 102	compat_uptr_t iov;
 103	compat_ulong_t iovcnt;
 104	__s64 offset;
 105	__u64 flags;
 106	__u64 len;
 107	__u64 unencoded_len;
 108	__u64 unencoded_offset;
 109	__u32 compression;
 110	__u32 encryption;
 111	__u8 reserved[64];
 112};
 113
 114#define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
 115				       struct btrfs_ioctl_encoded_io_args_32)
 116#define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
 117					struct btrfs_ioctl_encoded_io_args_32)
 118#endif
 119
 120/* Mask out flags that are inappropriate for the given type of inode. */
 121static unsigned int btrfs_mask_fsflags_for_type(const struct inode *inode,
 122						unsigned int flags)
 123{
 124	if (S_ISDIR(inode->i_mode))
 125		return flags;
 126	else if (S_ISREG(inode->i_mode))
 127		return flags & ~FS_DIRSYNC_FL;
 128	else
 129		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
 130}
 131
 132/*
 133 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
 134 * ioctl.
 135 */
 136static unsigned int btrfs_inode_flags_to_fsflags(const struct btrfs_inode *inode)
 137{
 138	unsigned int iflags = 0;
 139	u32 flags = inode->flags;
 140	u32 ro_flags = inode->ro_flags;
 141
 142	if (flags & BTRFS_INODE_SYNC)
 143		iflags |= FS_SYNC_FL;
 144	if (flags & BTRFS_INODE_IMMUTABLE)
 145		iflags |= FS_IMMUTABLE_FL;
 146	if (flags & BTRFS_INODE_APPEND)
 147		iflags |= FS_APPEND_FL;
 148	if (flags & BTRFS_INODE_NODUMP)
 149		iflags |= FS_NODUMP_FL;
 150	if (flags & BTRFS_INODE_NOATIME)
 151		iflags |= FS_NOATIME_FL;
 152	if (flags & BTRFS_INODE_DIRSYNC)
 153		iflags |= FS_DIRSYNC_FL;
 154	if (flags & BTRFS_INODE_NODATACOW)
 155		iflags |= FS_NOCOW_FL;
 156	if (ro_flags & BTRFS_INODE_RO_VERITY)
 157		iflags |= FS_VERITY_FL;
 158
 159	if (flags & BTRFS_INODE_NOCOMPRESS)
 160		iflags |= FS_NOCOMP_FL;
 161	else if (flags & BTRFS_INODE_COMPRESS)
 162		iflags |= FS_COMPR_FL;
 163
 164	return iflags;
 165}
 166
 167/*
 168 * Update inode->i_flags based on the btrfs internal flags.
 169 */
 170void btrfs_sync_inode_flags_to_i_flags(struct btrfs_inode *inode)
 171{
 172	unsigned int new_fl = 0;
 173
 174	if (inode->flags & BTRFS_INODE_SYNC)
 175		new_fl |= S_SYNC;
 176	if (inode->flags & BTRFS_INODE_IMMUTABLE)
 177		new_fl |= S_IMMUTABLE;
 178	if (inode->flags & BTRFS_INODE_APPEND)
 179		new_fl |= S_APPEND;
 180	if (inode->flags & BTRFS_INODE_NOATIME)
 181		new_fl |= S_NOATIME;
 182	if (inode->flags & BTRFS_INODE_DIRSYNC)
 183		new_fl |= S_DIRSYNC;
 184	if (inode->ro_flags & BTRFS_INODE_RO_VERITY)
 185		new_fl |= S_VERITY;
 186
 187	set_mask_bits(&inode->vfs_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(const 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 file_kattr *fa)
 249{
 250	const struct btrfs_inode *inode = BTRFS_I(d_inode(dentry));
 251
 252	fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(inode));
 253	return 0;
 254}
 255
 256int btrfs_fileattr_set(struct mnt_idmap *idmap,
 257		       struct dentry *dentry, struct file_kattr *fa)
 258{
 259	struct btrfs_inode *inode = BTRFS_I(d_inode(dentry));
 260	struct btrfs_root *root = inode->root;
 261	struct btrfs_fs_info *fs_info = root->fs_info;
 262	struct btrfs_trans_handle *trans;
 263	unsigned int fsflags, old_fsflags;
 264	int ret;
 265	const char *comp = NULL;
 266	u32 inode_flags;
 267
 268	if (btrfs_root_readonly(root))
 269		return -EROFS;
 270
 271	if (fileattr_has_fsx(fa))
 272		return -EOPNOTSUPP;
 273
 274	fsflags = btrfs_mask_fsflags_for_type(&inode->vfs_inode, fa->flags);
 275	old_fsflags = btrfs_inode_flags_to_fsflags(inode);
 276	ret = check_fsflags(old_fsflags, fsflags);
 277	if (ret)
 278		return ret;
 279
 280	ret = check_fsflags_compatible(fs_info, fsflags);
 281	if (ret)
 282		return ret;
 283
 284	inode_flags = inode->flags;
 285	if (fsflags & FS_SYNC_FL)
 286		inode_flags |= BTRFS_INODE_SYNC;
 287	else
 288		inode_flags &= ~BTRFS_INODE_SYNC;
 289	if (fsflags & FS_IMMUTABLE_FL)
 290		inode_flags |= BTRFS_INODE_IMMUTABLE;
 291	else
 292		inode_flags &= ~BTRFS_INODE_IMMUTABLE;
 293	if (fsflags & FS_APPEND_FL)
 294		inode_flags |= BTRFS_INODE_APPEND;
 295	else
 296		inode_flags &= ~BTRFS_INODE_APPEND;
 297	if (fsflags & FS_NODUMP_FL)
 298		inode_flags |= BTRFS_INODE_NODUMP;
 299	else
 300		inode_flags &= ~BTRFS_INODE_NODUMP;
 301	if (fsflags & FS_NOATIME_FL)
 302		inode_flags |= BTRFS_INODE_NOATIME;
 303	else
 304		inode_flags &= ~BTRFS_INODE_NOATIME;
 305
 306	/* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
 307	if (!fa->flags_valid) {
 308		/* 1 item for the inode */
 309		trans = btrfs_start_transaction(root, 1);
 310		if (IS_ERR(trans))
 311			return PTR_ERR(trans);
 312		goto update_flags;
 313	}
 314
 315	if (fsflags & FS_DIRSYNC_FL)
 316		inode_flags |= BTRFS_INODE_DIRSYNC;
 317	else
 318		inode_flags &= ~BTRFS_INODE_DIRSYNC;
 319	if (fsflags & FS_NOCOW_FL) {
 320		if (S_ISREG(inode->vfs_inode.i_mode)) {
 321			/*
 322			 * It's safe to turn csums off here, no extents exist.
 323			 * Otherwise we want the flag to reflect the real COW
 324			 * status of the file and will not set it.
 325			 */
 326			if (inode->vfs_inode.i_size == 0)
 327				inode_flags |= BTRFS_INODE_NODATACOW |
 328					       BTRFS_INODE_NODATASUM;
 329		} else {
 330			inode_flags |= BTRFS_INODE_NODATACOW;
 331		}
 332	} else {
 333		/*
 334		 * Revert back under same assumptions as above
 335		 */
 336		if (S_ISREG(inode->vfs_inode.i_mode)) {
 337			if (inode->vfs_inode.i_size == 0)
 338				inode_flags &= ~(BTRFS_INODE_NODATACOW |
 339						 BTRFS_INODE_NODATASUM);
 340		} else {
 341			inode_flags &= ~BTRFS_INODE_NODATACOW;
 342		}
 343	}
 344
 345	/*
 346	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
 347	 * flag may be changed automatically if compression code won't make
 348	 * things smaller.
 349	 */
 350	if (fsflags & FS_NOCOMP_FL) {
 351		inode_flags &= ~BTRFS_INODE_COMPRESS;
 352		inode_flags |= BTRFS_INODE_NOCOMPRESS;
 353	} else if (fsflags & FS_COMPR_FL) {
 354
 355		if (IS_SWAPFILE(&inode->vfs_inode))
 356			return -ETXTBSY;
 357
 358		inode_flags |= BTRFS_INODE_COMPRESS;
 359		inode_flags &= ~BTRFS_INODE_NOCOMPRESS;
 360
 361		comp = btrfs_compress_type2str(fs_info->compress_type);
 362		if (!comp || comp[0] == 0)
 363			comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
 364	} else {
 365		inode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
 366	}
 367
 368	/*
 369	 * 1 for inode item
 370	 * 2 for properties
 371	 */
 372	trans = btrfs_start_transaction(root, 3);
 373	if (IS_ERR(trans))
 374		return PTR_ERR(trans);
 375
 376	if (comp) {
 377		ret = btrfs_set_prop(trans, inode, "btrfs.compression",
 378				     comp, strlen(comp), 0);
 379		if (unlikely(ret)) {
 380			btrfs_abort_transaction(trans, ret);
 381			goto out_end_trans;
 382		}
 383	} else {
 384		ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, 0, 0);
 385		if (unlikely(ret && ret != -ENODATA)) {
 386			btrfs_abort_transaction(trans, ret);
 387			goto out_end_trans;
 388		}
 389	}
 390
 391update_flags:
 392	inode->flags = inode_flags;
 393	btrfs_update_inode_mapping_flags(inode);
 394	btrfs_sync_inode_flags_to_i_flags(inode);
 395	inode_inc_iversion(&inode->vfs_inode);
 396	inode_set_ctime_current(&inode->vfs_inode);
 397	ret = btrfs_update_inode(trans, inode);
 398
 399 out_end_trans:
 400	btrfs_end_transaction(trans);
 401	return ret;
 402}
 403
 404static int btrfs_ioctl_getversion(const struct inode *inode, int __user *arg)
 405{
 406	return put_user(inode->i_generation, arg);
 407}
 408
 409static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
 410					void __user *arg)
 411{
 412	struct btrfs_device *device;
 413	struct fstrim_range range;
 414	u64 minlen = ULLONG_MAX;
 415	u64 num_devices = 0;
 416	int ret;
 417
 418	if (!capable(CAP_SYS_ADMIN))
 419		return -EPERM;
 420
 421	/*
 422	 * btrfs_trim_block_group() depends on space cache, which is not
 423	 * available in zoned filesystem. So, disallow fitrim on a zoned
 424	 * filesystem for now.
 425	 */
 426	if (btrfs_is_zoned(fs_info))
 427		return -EOPNOTSUPP;
 428
 429	/*
 430	 * If the fs is mounted with nologreplay, which requires it to be
 431	 * mounted in RO mode as well, we can not allow discard on free space
 432	 * inside block groups, because log trees refer to extents that are not
 433	 * pinned in a block group's free space cache (pinning the extents is
 434	 * precisely the first phase of replaying a log tree).
 435	 */
 436	if (btrfs_test_opt(fs_info, NOLOGREPLAY))
 437		return -EROFS;
 438
 439	rcu_read_lock();
 440	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
 441				dev_list) {
 442		if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
 443			continue;
 444		num_devices++;
 445		minlen = min_t(u64, bdev_discard_granularity(device->bdev),
 446				    minlen);
 447	}
 448	rcu_read_unlock();
 449
 450	if (!num_devices)
 451		return -EOPNOTSUPP;
 452	if (copy_from_user(&range, arg, sizeof(range)))
 453		return -EFAULT;
 454
 455	/*
 456	 * NOTE: Don't truncate the range using super->total_bytes.  Bytenr of
 457	 * block group is in the logical address space, which can be any
 458	 * sectorsize aligned bytenr in  the range [0, U64_MAX].
 459	 */
 460	if (range.len < fs_info->sectorsize)
 461		return -EINVAL;
 462
 463	range.minlen = max(range.minlen, minlen);
 464	ret = btrfs_trim_fs(fs_info, &range);
 465
 466	if (copy_to_user(arg, &range, sizeof(range)))
 467		return -EFAULT;
 468
 469	return ret;
 470}
 471
 472/*
 473 * Calculate the number of transaction items to reserve for creating a subvolume
 474 * or snapshot, not including the inode, directory entries, or parent directory.
 475 */
 476static unsigned int create_subvol_num_items(const struct btrfs_qgroup_inherit *inherit)
 477{
 478	/*
 479	 * 1 to add root block
 480	 * 1 to add root item
 481	 * 1 to add root ref
 482	 * 1 to add root backref
 483	 * 1 to add UUID item
 484	 * 1 to add qgroup info
 485	 * 1 to add qgroup limit
 486	 *
 487	 * Ideally the last two would only be accounted if qgroups are enabled,
 488	 * but that can change between now and the time we would insert them.
 489	 */
 490	unsigned int num_items = 7;
 491
 492	if (inherit) {
 493		/* 2 to add qgroup relations for each inherited qgroup */
 494		num_items += 2 * inherit->num_qgroups;
 495	}
 496	return num_items;
 497}
 498
 499static noinline int create_subvol(struct mnt_idmap *idmap,
 500				  struct inode *dir, struct dentry *dentry,
 501				  struct btrfs_qgroup_inherit *inherit)
 502{
 503	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
 504	struct btrfs_trans_handle *trans;
 505	struct btrfs_key key;
 506	struct btrfs_root_item AUTO_KFREE(root_item);
 507	struct btrfs_inode_item *inode_item;
 508	struct extent_buffer *leaf;
 509	struct btrfs_root *root = BTRFS_I(dir)->root;
 510	struct btrfs_root *new_root;
 511	struct btrfs_block_rsv block_rsv;
 512	struct timespec64 cur_time = current_time(dir);
 513	struct btrfs_new_inode_args new_inode_args = {
 514		.dir = dir,
 515		.dentry = dentry,
 516		.subvol = true,
 517	};
 518	unsigned int trans_num_items;
 519	int ret;
 520	dev_t anon_dev;
 521	u64 objectid;
 522	u64 qgroup_reserved = 0;
 523
 524	root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
 525	if (!root_item)
 526		return -ENOMEM;
 527
 528	ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
 529	if (ret)
 530		return ret;
 531
 532	/*
 533	 * Don't create subvolume whose level is not zero. Or qgroup will be
 534	 * screwed up since it assumes subvolume qgroup's level to be 0.
 535	 */
 536	if (btrfs_qgroup_level(objectid))
 537		return -ENOSPC;
 538
 539	ret = get_anon_bdev(&anon_dev);
 540	if (ret < 0)
 541		return ret;
 542
 543	new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir);
 544	if (!new_inode_args.inode) {
 545		ret = -ENOMEM;
 546		goto out_anon_dev;
 547	}
 548	ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
 549	if (ret)
 550		goto out_inode;
 551	trans_num_items += create_subvol_num_items(inherit);
 552
 553	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
 554	ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
 555					       trans_num_items, false);
 556	if (ret)
 557		goto out_new_inode_args;
 558	qgroup_reserved = block_rsv.qgroup_rsv_reserved;
 559
 560	trans = btrfs_start_transaction(root, 0);
 561	if (IS_ERR(trans)) {
 562		ret = PTR_ERR(trans);
 563		goto out_release_rsv;
 564	}
 565	btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved);
 566	qgroup_reserved = 0;
 567	trans->block_rsv = &block_rsv;
 568	trans->bytes_reserved = block_rsv.size;
 569
 570	ret = btrfs_qgroup_inherit(trans, 0, objectid, btrfs_root_id(root), inherit);
 571	if (ret)
 572		goto out;
 573
 574	leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
 575				      0, BTRFS_NESTING_NORMAL);
 576	if (IS_ERR(leaf)) {
 577		ret = PTR_ERR(leaf);
 578		goto out;
 579	}
 580
 581	btrfs_mark_buffer_dirty(trans, leaf);
 582
 583	inode_item = &root_item->inode;
 584	btrfs_set_stack_inode_generation(inode_item, 1);
 585	btrfs_set_stack_inode_size(inode_item, 3);
 586	btrfs_set_stack_inode_nlink(inode_item, 1);
 587	btrfs_set_stack_inode_nbytes(inode_item,
 588				     fs_info->nodesize);
 589	btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
 590
 591	btrfs_set_root_flags(root_item, 0);
 592	btrfs_set_root_limit(root_item, 0);
 593	btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
 594
 595	btrfs_set_root_bytenr(root_item, leaf->start);
 596	btrfs_set_root_generation(root_item, trans->transid);
 597	btrfs_set_root_level(root_item, 0);
 598	btrfs_set_root_refs(root_item, 1);
 599	btrfs_set_root_used(root_item, leaf->len);
 600	btrfs_set_root_last_snapshot(root_item, 0);
 601
 602	btrfs_set_root_generation_v2(root_item,
 603			btrfs_root_generation(root_item));
 604	generate_random_guid(root_item->uuid);
 605	btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
 606	btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
 607	root_item->ctime = root_item->otime;
 608	btrfs_set_root_ctransid(root_item, trans->transid);
 609	btrfs_set_root_otransid(root_item, trans->transid);
 610
 611	btrfs_tree_unlock(leaf);
 612
 613	btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
 614
 615	key.objectid = objectid;
 616	key.type = BTRFS_ROOT_ITEM_KEY;
 617	key.offset = 0;
 618	ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
 619				root_item);
 620	if (ret) {
 621		int ret2;
 622
 623		/*
 624		 * Since we don't abort the transaction in this case, free the
 625		 * tree block so that we don't leak space and leave the
 626		 * filesystem in an inconsistent state (an extent item in the
 627		 * extent tree with a backreference for a root that does not
 628		 * exists).
 629		 */
 630		btrfs_tree_lock(leaf);
 631		btrfs_clear_buffer_dirty(trans, leaf);
 632		btrfs_tree_unlock(leaf);
 633		ret2 = btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
 634		if (unlikely(ret2 < 0))
 635			btrfs_abort_transaction(trans, ret2);
 636		free_extent_buffer(leaf);
 637		goto out;
 638	}
 639
 640	free_extent_buffer(leaf);
 641	leaf = NULL;
 642
 643	new_root = btrfs_get_new_fs_root(fs_info, objectid, &anon_dev);
 644	if (IS_ERR(new_root)) {
 645		ret = PTR_ERR(new_root);
 646		btrfs_abort_transaction(trans, ret);
 647		goto out;
 648	}
 649	/* anon_dev is owned by new_root now. */
 650	anon_dev = 0;
 651	BTRFS_I(new_inode_args.inode)->root = new_root;
 652	/* ... and new_root is owned by new_inode_args.inode now. */
 653
 654	ret = btrfs_record_root_in_trans(trans, new_root);
 655	if (unlikely(ret)) {
 656		btrfs_abort_transaction(trans, ret);
 657		goto out;
 658	}
 659
 660	ret = btrfs_uuid_tree_add(trans, root_item->uuid,
 661				  BTRFS_UUID_KEY_SUBVOL, objectid);
 662	if (unlikely(ret)) {
 663		btrfs_abort_transaction(trans, ret);
 664		goto out;
 665	}
 666
 667	btrfs_record_new_subvolume(trans, BTRFS_I(dir));
 668
 669	ret = btrfs_create_new_inode(trans, &new_inode_args);
 670	if (unlikely(ret)) {
 671		btrfs_abort_transaction(trans, ret);
 672		goto out;
 673	}
 674
 675	/*
 676	 * Subvolumes have orphans cleaned on first dentry lookup. A new
 677	 * subvolume cannot have any orphans, so we should set the bit before we
 678	 * add the subvolume dentry to the dentry cache, so that it is in the
 679	 * same state as a subvolume after first lookup.
 680	 */
 681	set_bit(BTRFS_ROOT_ORPHAN_CLEANUP, &new_root->state);
 682	d_instantiate_new(dentry, new_inode_args.inode);
 683	new_inode_args.inode = NULL;
 684
 685out:
 686	trans->block_rsv = NULL;
 687	trans->bytes_reserved = 0;
 688	btrfs_end_transaction(trans);
 689out_release_rsv:
 690	btrfs_block_rsv_release(fs_info, &block_rsv, (u64)-1, NULL);
 691	if (qgroup_reserved)
 692		btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved);
 693out_new_inode_args:
 694	btrfs_new_inode_args_destroy(&new_inode_args);
 695out_inode:
 696	iput(new_inode_args.inode);
 697out_anon_dev:
 698	if (anon_dev)
 699		free_anon_bdev(anon_dev);
 700
 701	return ret;
 702}
 703
 704static int create_snapshot(struct btrfs_root *root, struct inode *dir,
 705			   struct dentry *dentry, bool readonly,
 706			   struct btrfs_qgroup_inherit *inherit)
 707{
 708	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
 709	struct inode *inode;
 710	struct btrfs_pending_snapshot *pending_snapshot;
 711	unsigned int trans_num_items;
 712	struct btrfs_trans_handle *trans;
 713	struct btrfs_block_rsv *block_rsv;
 714	u64 qgroup_reserved = 0;
 715	int ret;
 716
 717	/* We do not support snapshotting right now. */
 718	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
 719		btrfs_warn(fs_info,
 720			   "extent tree v2 doesn't support snapshotting yet");
 721		return -EOPNOTSUPP;
 722	}
 723
 724	if (btrfs_root_refs(&root->root_item) == 0)
 725		return -ENOENT;
 726
 727	if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
 728		return -EINVAL;
 729
 730	if (atomic_read(&root->nr_swapfiles)) {
 731		btrfs_warn(fs_info,
 732			   "cannot snapshot subvolume with active swapfile");
 733		return -ETXTBSY;
 734	}
 735
 736	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
 737	if (!pending_snapshot)
 738		return -ENOMEM;
 739
 740	ret = get_anon_bdev(&pending_snapshot->anon_dev);
 741	if (ret < 0)
 742		goto free_pending;
 743	pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
 744			GFP_KERNEL);
 745	pending_snapshot->path = btrfs_alloc_path();
 746	if (!pending_snapshot->root_item || !pending_snapshot->path) {
 747		ret = -ENOMEM;
 748		goto free_pending;
 749	}
 750
 751	block_rsv = &pending_snapshot->block_rsv;
 752	btrfs_init_block_rsv(block_rsv, BTRFS_BLOCK_RSV_TEMP);
 753	/*
 754	 * 1 to add dir item
 755	 * 1 to add dir index
 756	 * 1 to update parent inode item
 757	 */
 758	trans_num_items = create_subvol_num_items(inherit) + 3;
 759	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, block_rsv,
 760					       trans_num_items, false);
 761	if (ret)
 762		goto free_pending;
 763	qgroup_reserved = block_rsv->qgroup_rsv_reserved;
 764
 765	pending_snapshot->dentry = dentry;
 766	pending_snapshot->root = root;
 767	pending_snapshot->readonly = readonly;
 768	pending_snapshot->dir = BTRFS_I(dir);
 769	pending_snapshot->inherit = inherit;
 770
 771	trans = btrfs_start_transaction(root, 0);
 772	if (IS_ERR(trans)) {
 773		ret = PTR_ERR(trans);
 774		goto fail;
 775	}
 776	ret = btrfs_record_root_in_trans(trans, BTRFS_I(dir)->root);
 777	if (ret) {
 778		btrfs_end_transaction(trans);
 779		goto fail;
 780	}
 781	btrfs_qgroup_convert_reserved_meta(root, qgroup_reserved);
 782	qgroup_reserved = 0;
 783
 784	trans->pending_snapshot = pending_snapshot;
 785
 786	ret = btrfs_commit_transaction(trans);
 787	if (ret)
 788		goto fail;
 789
 790	ret = pending_snapshot->error;
 791	if (ret)
 792		goto fail;
 793
 794	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
 795	if (ret)
 796		goto fail;
 797
 798	inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
 799	if (IS_ERR(inode)) {
 800		ret = PTR_ERR(inode);
 801		goto fail;
 802	}
 803
 804	d_instantiate(dentry, inode);
 805	ret = 0;
 806	pending_snapshot->anon_dev = 0;
 807fail:
 808	/* Prevent double freeing of anon_dev */
 809	if (ret && pending_snapshot->snap)
 810		pending_snapshot->snap->anon_dev = 0;
 811	btrfs_put_root(pending_snapshot->snap);
 812	btrfs_block_rsv_release(fs_info, block_rsv, (u64)-1, NULL);
 813	if (qgroup_reserved)
 814		btrfs_qgroup_free_meta_prealloc(root, qgroup_reserved);
 815free_pending:
 816	if (pending_snapshot->anon_dev)
 817		free_anon_bdev(pending_snapshot->anon_dev);
 818	kfree(pending_snapshot->root_item);
 819	btrfs_free_path(pending_snapshot->path);
 820	kfree(pending_snapshot);
 821
 822	return ret;
 823}
 824
 825/*  copy of may_delete in fs/namei.c()
 826 *	Check whether we can remove a link victim from directory dir, check
 827 *  whether the type of victim is right.
 828 *  1. We can't do it if dir is read-only (done in permission())
 829 *  2. We should have write and exec permissions on dir
 830 *  3. We can't remove anything from append-only dir
 831 *  4. We can't do anything with immutable dir (done in permission())
 832 *  5. If the sticky bit on dir is set we should either
 833 *	a. be owner of dir, or
 834 *	b. be owner of victim, or
 835 *	c. have CAP_FOWNER capability
 836 *  6. If the victim is append-only or immutable we can't do anything with
 837 *     links pointing to it.
 838 *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
 839 *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
 840 *  9. We can't remove a root or mountpoint.
 841 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
 842 *     nfs_async_unlink().
 843 */
 844
 845static int btrfs_may_delete(struct mnt_idmap *idmap,
 846			    struct inode *dir, struct dentry *victim, int isdir)
 847{
 848	int ret;
 849
 850	if (d_really_is_negative(victim))
 851		return -ENOENT;
 852
 853	/* The @victim is not inside @dir. */
 854	if (d_inode(victim->d_parent) != dir)
 855		return -EINVAL;
 856	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
 857
 858	ret = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
 859	if (ret)
 860		return ret;
 861	if (IS_APPEND(dir))
 862		return -EPERM;
 863	if (check_sticky(idmap, dir, d_inode(victim)) ||
 864	    IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
 865	    IS_SWAPFILE(d_inode(victim)))
 866		return -EPERM;
 867	if (isdir) {
 868		if (!d_is_dir(victim))
 869			return -ENOTDIR;
 870		if (IS_ROOT(victim))
 871			return -EBUSY;
 872	} else if (d_is_dir(victim))
 873		return -EISDIR;
 874	if (IS_DEADDIR(dir))
 875		return -ENOENT;
 876	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
 877		return -EBUSY;
 878	return 0;
 879}
 880
 881/* copy of may_create in fs/namei.c() */
 882static inline int btrfs_may_create(struct mnt_idmap *idmap,
 883				   struct inode *dir, const struct dentry *child)
 884{
 885	if (d_really_is_positive(child))
 886		return -EEXIST;
 887	if (IS_DEADDIR(dir))
 888		return -ENOENT;
 889	if (!fsuidgid_has_mapping(dir->i_sb, idmap))
 890		return -EOVERFLOW;
 891	return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
 892}
 893
 894/*
 895 * Create a new subvolume below @parent.  This is largely modeled after
 896 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 897 * inside this filesystem so it's quite a bit simpler.
 898 */
 899static noinline int btrfs_mksubvol(struct dentry *parent,
 900				   struct mnt_idmap *idmap,
 901				   struct qstr *qname, struct btrfs_root *snap_src,
 902				   bool readonly,
 903				   struct btrfs_qgroup_inherit *inherit)
 904{
 905	struct inode *dir = d_inode(parent);
 906	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
 907	struct dentry *dentry;
 908	struct fscrypt_str name_str = FSTR_INIT((char *)qname->name, qname->len);
 909	int ret;
 910
 911	dentry = start_creating_killable(idmap, parent, qname);
 912	if (IS_ERR(dentry))
 913		return PTR_ERR(dentry);
 914
 915	ret = btrfs_may_create(idmap, dir, dentry);
 916	if (ret)
 917		goto out_dput;
 918
 919	/*
 920	 * even if this name doesn't exist, we may get hash collisions.
 921	 * check for them now when we can safely fail
 922	 */
 923	ret = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, dir->i_ino, &name_str);
 924	if (ret)
 925		goto out_dput;
 926
 927	down_read(&fs_info->subvol_sem);
 928
 929	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
 930		goto out_up_read;
 931
 932	if (snap_src)
 933		ret = create_snapshot(snap_src, dir, dentry, readonly, inherit);
 934	else
 935		ret = create_subvol(idmap, dir, dentry, inherit);
 936
 937	if (!ret)
 938		fsnotify_mkdir(dir, dentry);
 939out_up_read:
 940	up_read(&fs_info->subvol_sem);
 941out_dput:
 942	end_creating(dentry);
 943	return ret;
 944}
 945
 946static noinline int btrfs_mksnapshot(struct dentry *parent,
 947				   struct mnt_idmap *idmap,
 948				   struct qstr *qname,
 949				   struct btrfs_root *root,
 950				   bool readonly,
 951				   struct btrfs_qgroup_inherit *inherit)
 952{
 953	int ret;
 954
 955	/*
 956	 * Force new buffered writes to reserve space even when NOCOW is
 957	 * possible. This is to avoid later writeback (running delalloc) to
 958	 * fallback to COW mode and unexpectedly fail with ENOSPC.
 959	 */
 960	btrfs_drew_read_lock(&root->snapshot_lock);
 961
 962	ret = btrfs_start_delalloc_snapshot(root, false);
 963	if (ret)
 964		goto out;
 965
 966	/*
 967	 * All previous writes have started writeback in NOCOW mode, so now
 968	 * we force future writes to fallback to COW mode during snapshot
 969	 * creation.
 970	 */
 971	atomic_inc(&root->snapshot_force_cow);
 972
 973	btrfs_wait_ordered_extents(root, U64_MAX, NULL);
 974
 975	ret = btrfs_mksubvol(parent, idmap, qname, root, readonly, inherit);
 976
 977	atomic_dec(&root->snapshot_force_cow);
 978out:
 979	btrfs_drew_read_unlock(&root->snapshot_lock);
 980	return ret;
 981}
 982
 983/*
 984 * Try to start exclusive operation @type or cancel it if it's running.
 985 *
 986 * Return:
 987 *   0        - normal mode, newly claimed op started
 988 *  >0        - normal mode, something else is running,
 989 *              return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
 990 * ECANCELED  - cancel mode, successful cancel
 991 * ENOTCONN   - cancel mode, operation not running anymore
 992 */
 993static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
 994			enum btrfs_exclusive_operation type, bool cancel)
 995{
 996	if (!cancel) {
 997		/* Start normal op */
 998		if (!btrfs_exclop_start(fs_info, type))
 999			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1000		/* Exclusive operation is now claimed */
1001		return 0;
1002	}
1003
1004	/* Cancel running op */
1005	if (btrfs_exclop_start_try_lock(fs_info, type)) {
1006		/*
1007		 * This blocks any exclop finish from setting it to NONE, so we
1008		 * request cancellation. Either it runs and we will wait for it,
1009		 * or it has finished and no waiting will happen.
1010		 */
1011		atomic_inc(&fs_info->reloc_cancel_req);
1012		btrfs_exclop_start_unlock(fs_info);
1013
1014		if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1015			wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1016				    TASK_INTERRUPTIBLE);
1017
1018		return -ECANCELED;
1019	}
1020
1021	/* Something else is running or none */
1022	return -ENOTCONN;
1023}
1024
1025static noinline int btrfs_ioctl_resize(struct file *file,
1026					void __user *arg)
1027{
1028	BTRFS_DEV_LOOKUP_ARGS(args);
1029	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1030	struct btrfs_fs_info *fs_info = root->fs_info;
1031	u64 new_size;
1032	u64 old_size;
1033	u64 devid = 1;
1034	struct btrfs_ioctl_vol_args *vol_args;
1035	struct btrfs_device *device = NULL;
1036	char *sizestr;
1037	char *devstr = NULL;
1038	int ret = 0;
1039	int mod = 0;
1040	bool cancel;
1041
1042	if (!capable(CAP_SYS_ADMIN))
1043		return -EPERM;
1044
1045	ret = mnt_want_write_file(file);
1046	if (ret)
1047		return ret;
1048
1049	/*
1050	 * Read the arguments before checking exclusivity to be able to
1051	 * distinguish regular resize and cancel
1052	 */
1053	vol_args = memdup_user(arg, sizeof(*vol_args));
1054	if (IS_ERR(vol_args)) {
1055		ret = PTR_ERR(vol_args);
1056		goto out_drop;
1057	}
1058	ret = btrfs_check_ioctl_vol_args_path(vol_args);
1059	if (ret < 0)
1060		goto out_free;
1061
1062	sizestr = vol_args->name;
1063	cancel = (strcmp("cancel", sizestr) == 0);
1064	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1065	if (ret)
1066		goto out_free;
1067	/* Exclusive operation is now claimed */
1068
1069	devstr = strchr(sizestr, ':');
1070	if (devstr) {
1071		sizestr = devstr + 1;
1072		*devstr = '\0';
1073		devstr = vol_args->name;
1074		ret = kstrtoull(devstr, 10, &devid);
1075		if (ret)
1076			goto out_finish;
1077		if (!devid) {
1078			ret = -EINVAL;
1079			goto out_finish;
1080		}
1081		btrfs_info(fs_info, "resizing devid %llu", devid);
1082	}
1083
1084	args.devid = devid;
1085	device = btrfs_find_device(fs_info->fs_devices, &args);
1086	if (!device) {
1087		btrfs_info(fs_info, "resizer unable to find device %llu",
1088			   devid);
1089		ret = -ENODEV;
1090		goto out_finish;
1091	}
1092
1093	if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1094		btrfs_info(fs_info,
1095			   "resizer unable to apply on readonly device %llu",
1096		       devid);
1097		ret = -EPERM;
1098		goto out_finish;
1099	}
1100
1101	if (!strcmp(sizestr, "max"))
1102		new_size = bdev_nr_bytes(device->bdev);
1103	else {
1104		char *retptr;
1105
1106		if (sizestr[0] == '-') {
1107			mod = -1;
1108			sizestr++;
1109		} else if (sizestr[0] == '+') {
1110			mod = 1;
1111			sizestr++;
1112		}
1113		new_size = memparse(sizestr, &retptr);
1114		if (*retptr != '\0' || new_size == 0) {
1115			ret = -EINVAL;
1116			goto out_finish;
1117		}
1118	}
1119
1120	if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1121		ret = -EPERM;
1122		goto out_finish;
1123	}
1124
1125	old_size = btrfs_device_get_total_bytes(device);
1126
1127	if (mod < 0) {
1128		if (new_size > old_size) {
1129			ret = -EINVAL;
1130			goto out_finish;
1131		}
1132		new_size = old_size - new_size;
1133	} else if (mod > 0) {
1134		if (new_size > ULLONG_MAX - old_size) {
1135			ret = -ERANGE;
1136			goto out_finish;
1137		}
1138		new_size = old_size + new_size;
1139	}
1140
1141	if (new_size < SZ_256M) {
1142		ret = -EINVAL;
1143		goto out_finish;
1144	}
1145	if (new_size > bdev_nr_bytes(device->bdev)) {
1146		ret = -EFBIG;
1147		goto out_finish;
1148	}
1149
1150	new_size = round_down(new_size, fs_info->sectorsize);
1151
1152	if (new_size > old_size) {
1153		struct btrfs_trans_handle *trans;
1154
1155		trans = btrfs_start_transaction(root, 0);
1156		if (IS_ERR(trans)) {
1157			ret = PTR_ERR(trans);
1158			goto out_finish;
1159		}
1160		ret = btrfs_grow_device(trans, device, new_size);
1161		btrfs_commit_transaction(trans);
1162	} else if (new_size < old_size) {
1163		ret = btrfs_shrink_device(device, new_size);
1164	} /* equal, nothing need to do */
1165
1166	if (ret == 0 && new_size != old_size)
1167		btrfs_info(fs_info,
1168			"resize device %s (devid %llu) from %llu to %llu",
1169			btrfs_dev_name(device), device->devid,
1170			old_size, new_size);
1171out_finish:
1172	btrfs_exclop_finish(fs_info);
1173out_free:
1174	kfree(vol_args);
1175out_drop:
1176	mnt_drop_write_file(file);
1177	return ret;
1178}
1179
1180static noinline int __btrfs_ioctl_snap_create(struct file *file,
1181				struct mnt_idmap *idmap,
1182				const char *name, unsigned long fd, bool subvol,
1183				bool readonly,
1184				struct btrfs_qgroup_inherit *inherit)
1185{
1186	int ret = 0;
1187	struct qstr qname = QSTR_INIT(name, strlen(name));
1188
1189	if (!S_ISDIR(file_inode(file)->i_mode))
1190		return -ENOTDIR;
1191
1192	ret = mnt_want_write_file(file);
1193	if (ret)
1194		goto out;
1195
1196	if (strchr(name, '/')) {
1197		ret = -EINVAL;
1198		goto out_drop_write;
1199	}
1200
1201	if (qname.name[0] == '.' &&
1202	   (qname.len == 1 || (qname.name[1] == '.' && qname.len == 2))) {
1203		ret = -EEXIST;
1204		goto out_drop_write;
1205	}
1206
1207	if (subvol) {
1208		ret = btrfs_mksubvol(file_dentry(file), idmap, &qname, NULL,
1209				     readonly, inherit);
1210	} else {
1211		CLASS(fd, src)(fd);
1212		struct inode *src_inode;
1213		if (fd_empty(src)) {
1214			ret = -EINVAL;
1215			goto out_drop_write;
1216		}
1217
1218		src_inode = file_inode(fd_file(src));
1219		if (src_inode->i_sb != file_inode(file)->i_sb) {
1220			btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1221				   "Snapshot src from another FS");
1222			ret = -EXDEV;
1223		} else if (!inode_owner_or_capable(idmap, src_inode)) {
1224			/*
1225			 * Subvolume creation is not restricted, but snapshots
1226			 * are limited to own subvolumes only
1227			 */
1228			ret = -EPERM;
1229		} else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1230			/*
1231			 * Snapshots must be made with the src_inode referring
1232			 * to the subvolume inode, otherwise the permission
1233			 * checking above is useless because we may have
1234			 * permission on a lower directory but not the subvol
1235			 * itself.
1236			 */
1237			ret = -EINVAL;
1238		} else {
1239			ret = btrfs_mksnapshot(file_dentry(file), idmap, &qname,
1240					       BTRFS_I(src_inode)->root,
1241					       readonly, inherit);
1242		}
1243	}
1244out_drop_write:
1245	mnt_drop_write_file(file);
1246out:
1247	return ret;
1248}
1249
1250static noinline int btrfs_ioctl_snap_create(struct file *file,
1251					    void __user *arg, bool subvol)
1252{
1253	struct btrfs_ioctl_vol_args *vol_args;
1254	int ret;
1255
1256	if (!S_ISDIR(file_inode(file)->i_mode))
1257		return -ENOTDIR;
1258
1259	vol_args = memdup_user(arg, sizeof(*vol_args));
1260	if (IS_ERR(vol_args))
1261		return PTR_ERR(vol_args);
1262	ret = btrfs_check_ioctl_vol_args_path(vol_args);
1263	if (ret < 0)
1264		goto out;
1265
1266	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1267					vol_args->name, vol_args->fd, subvol,
1268					false, NULL);
1269
1270out:
1271	kfree(vol_args);
1272	return ret;
1273}
1274
1275static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1276					       void __user *arg, bool subvol)
1277{
1278	struct btrfs_ioctl_vol_args_v2 *vol_args;
1279	int ret;
1280	bool readonly = false;
1281	struct btrfs_qgroup_inherit *inherit = NULL;
1282
1283	if (!S_ISDIR(file_inode(file)->i_mode))
1284		return -ENOTDIR;
1285
1286	vol_args = memdup_user(arg, sizeof(*vol_args));
1287	if (IS_ERR(vol_args))
1288		return PTR_ERR(vol_args);
1289	ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args);
1290	if (ret < 0)
1291		goto free_args;
1292
1293	if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1294		ret = -EOPNOTSUPP;
1295		goto free_args;
1296	}
1297
1298	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1299		readonly = true;
1300	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1301		struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file));
1302
1303		if (vol_args->size < sizeof(*inherit) ||
1304		    vol_args->size > PAGE_SIZE) {
1305			ret = -EINVAL;
1306			goto free_args;
1307		}
1308		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1309		if (IS_ERR(inherit)) {
1310			ret = PTR_ERR(inherit);
1311			goto free_args;
1312		}
1313
1314		ret = btrfs_qgroup_check_inherit(fs_info, inherit, vol_args->size);
1315		if (ret < 0)
1316			goto free_inherit;
1317	}
1318
1319	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1320					vol_args->name, vol_args->fd, subvol,
1321					readonly, inherit);
1322	if (ret)
1323		goto free_inherit;
1324free_inherit:
1325	kfree(inherit);
1326free_args:
1327	kfree(vol_args);
1328	return ret;
1329}
1330
1331static noinline int btrfs_ioctl_subvol_getflags(struct btrfs_inode *inode,
1332						void __user *arg)
1333{
1334	struct btrfs_root *root = inode->root;
1335	struct btrfs_fs_info *fs_info = root->fs_info;
1336	int ret = 0;
1337	u64 flags = 0;
1338
1339	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1340		return -EINVAL;
1341
1342	down_read(&fs_info->subvol_sem);
1343	if (btrfs_root_readonly(root))
1344		flags |= BTRFS_SUBVOL_RDONLY;
1345	up_read(&fs_info->subvol_sem);
1346
1347	if (copy_to_user(arg, &flags, sizeof(flags)))
1348		ret = -EFAULT;
1349
1350	return ret;
1351}
1352
1353static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1354					      void __user *arg)
1355{
1356	struct inode *inode = file_inode(file);
1357	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1358	struct btrfs_root *root = BTRFS_I(inode)->root;
1359	struct btrfs_trans_handle *trans;
1360	u64 root_flags;
1361	u64 flags;
1362	int ret = 0;
1363
1364	if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1365		return -EPERM;
1366
1367	ret = mnt_want_write_file(file);
1368	if (ret)
1369		goto out;
1370
1371	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1372		ret = -EINVAL;
1373		goto out_drop_write;
1374	}
1375
1376	if (copy_from_user(&flags, arg, sizeof(flags))) {
1377		ret = -EFAULT;
1378		goto out_drop_write;
1379	}
1380
1381	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1382		ret = -EOPNOTSUPP;
1383		goto out_drop_write;
1384	}
1385
1386	down_write(&fs_info->subvol_sem);
1387
1388	/* nothing to do */
1389	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1390		goto out_drop_sem;
1391
1392	root_flags = btrfs_root_flags(&root->root_item);
1393	if (flags & BTRFS_SUBVOL_RDONLY) {
1394		btrfs_set_root_flags(&root->root_item,
1395				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1396	} else {
1397		/*
1398		 * Block RO -> RW transition if this subvolume is involved in
1399		 * send
1400		 */
1401		spin_lock(&root->root_item_lock);
1402		if (root->send_in_progress == 0) {
1403			btrfs_set_root_flags(&root->root_item,
1404				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1405			spin_unlock(&root->root_item_lock);
1406		} else {
1407			spin_unlock(&root->root_item_lock);
1408			btrfs_warn(fs_info,
1409				   "Attempt to set subvolume %llu read-write during send",
1410				   btrfs_root_id(root));
1411			ret = -EPERM;
1412			goto out_drop_sem;
1413		}
1414	}
1415
1416	trans = btrfs_start_transaction(root, 1);
1417	if (IS_ERR(trans)) {
1418		ret = PTR_ERR(trans);
1419		goto out_reset;
1420	}
1421
1422	ret = btrfs_update_root(trans, fs_info->tree_root,
1423				&root->root_key, &root->root_item);
1424	if (ret < 0) {
1425		btrfs_end_transaction(trans);
1426		goto out_reset;
1427	}
1428
1429	ret = btrfs_commit_transaction(trans);
1430
1431out_reset:
1432	if (ret)
1433		btrfs_set_root_flags(&root->root_item, root_flags);
1434out_drop_sem:
1435	up_write(&fs_info->subvol_sem);
1436out_drop_write:
1437	mnt_drop_write_file(file);
1438out:
1439	return ret;
1440}
1441
1442static noinline bool key_in_sk(const struct btrfs_key *key,
1443			       const struct btrfs_ioctl_search_key *sk)
1444{
1445	struct btrfs_key test;
1446	int ret;
1447
1448	test.objectid = sk->min_objectid;
1449	test.type = sk->min_type;
1450	test.offset = sk->min_offset;
1451
1452	ret = btrfs_comp_cpu_keys(key, &test);
1453	if (ret < 0)
1454		return false;
1455
1456	test.objectid = sk->max_objectid;
1457	test.type = sk->max_type;
1458	test.offset = sk->max_offset;
1459
1460	ret = btrfs_comp_cpu_keys(key, &test);
1461	if (ret > 0)
1462		return false;
1463	return true;
1464}
1465
1466static noinline int copy_to_sk(struct btrfs_path *path,
1467			       struct btrfs_key *key,
1468			       const struct btrfs_ioctl_search_key *sk,
1469			       u64 *buf_size,
1470			       char __user *ubuf,
1471			       unsigned long *sk_offset,
1472			       int *num_found)
1473{
1474	u64 found_transid;
1475	struct extent_buffer *leaf;
1476	struct btrfs_ioctl_search_header sh;
1477	struct btrfs_key test;
1478	unsigned long item_off;
1479	unsigned long item_len;
1480	int nritems;
1481	int i;
1482	int slot;
1483	int ret = 0;
1484
1485	leaf = path->nodes[0];
1486	slot = path->slots[0];
1487	nritems = btrfs_header_nritems(leaf);
1488
1489	if (btrfs_header_generation(leaf) > sk->max_transid) {
1490		i = nritems;
1491		goto advance_key;
1492	}
1493	found_transid = btrfs_header_generation(leaf);
1494
1495	for (i = slot; i < nritems; i++) {
1496		item_off = btrfs_item_ptr_offset(leaf, i);
1497		item_len = btrfs_item_size(leaf, i);
1498
1499		btrfs_item_key_to_cpu(leaf, key, i);
1500		if (!key_in_sk(key, sk))
1501			continue;
1502
1503		if (sizeof(sh) + item_len > *buf_size) {
1504			if (*num_found) {
1505				ret = 1;
1506				goto out;
1507			}
1508
1509			/*
1510			 * return one empty item back for v1, which does not
1511			 * handle -EOVERFLOW
1512			 */
1513
1514			*buf_size = sizeof(sh) + item_len;
1515			item_len = 0;
1516			ret = -EOVERFLOW;
1517		}
1518
1519		if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1520			ret = 1;
1521			goto out;
1522		}
1523
1524		sh.objectid = key->objectid;
1525		sh.type = key->type;
1526		sh.offset = key->offset;
1527		sh.len = item_len;
1528		sh.transid = found_transid;
1529
1530		/*
1531		 * Copy search result header. If we fault then loop again so we
1532		 * can fault in the pages and -EFAULT there if there's a
1533		 * problem. Otherwise we'll fault and then copy the buffer in
1534		 * properly this next time through
1535		 */
1536		if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1537			ret = 0;
1538			goto out;
1539		}
1540
1541		*sk_offset += sizeof(sh);
1542
1543		if (item_len) {
1544			char __user *up = ubuf + *sk_offset;
1545			/*
1546			 * Copy the item, same behavior as above, but reset the
1547			 * * sk_offset so we copy the full thing again.
1548			 */
1549			if (read_extent_buffer_to_user_nofault(leaf, up,
1550						item_off, item_len)) {
1551				ret = 0;
1552				*sk_offset -= sizeof(sh);
1553				goto out;
1554			}
1555
1556			*sk_offset += item_len;
1557		}
1558		(*num_found)++;
1559
1560		if (ret) /* -EOVERFLOW from above */
1561			goto out;
1562
1563		if (*num_found >= sk->nr_items) {
1564			ret = 1;
1565			goto out;
1566		}
1567	}
1568advance_key:
1569	ret = 0;
1570	test.objectid = sk->max_objectid;
1571	test.type = sk->max_type;
1572	test.offset = sk->max_offset;
1573	if (btrfs_comp_cpu_keys(key, &test) >= 0)
1574		ret = 1;
1575	else if (key->offset < (u64)-1)
1576		key->offset++;
1577	else if (key->type < (u8)-1) {
1578		key->offset = 0;
1579		key->type++;
1580	} else if (key->objectid < (u64)-1) {
1581		key->offset = 0;
1582		key->type = 0;
1583		key->objectid++;
1584	} else
1585		ret = 1;
1586out:
1587	/*
1588	 *  0: all items from this leaf copied, continue with next
1589	 *  1: * more items can be copied, but unused buffer is too small
1590	 *     * all items were found
1591	 *     Either way, it will stops the loop which iterates to the next
1592	 *     leaf
1593	 *  -EOVERFLOW: item was to large for buffer
1594	 *  -EFAULT: could not copy extent buffer back to userspace
1595	 */
1596	return ret;
1597}
1598
1599static noinline int search_ioctl(struct btrfs_root *root,
1600				 struct btrfs_ioctl_search_key *sk,
1601				 u64 *buf_size,
1602				 char __user *ubuf)
1603{
1604	struct btrfs_fs_info *info = root->fs_info;
1605	struct btrfs_key key;
1606	BTRFS_PATH_AUTO_FREE(path);
1607	int ret;
1608	int num_found = 0;
1609	unsigned long sk_offset = 0;
1610
1611	if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1612		*buf_size = sizeof(struct btrfs_ioctl_search_header);
1613		return -EOVERFLOW;
1614	}
1615
1616	path = btrfs_alloc_path();
1617	if (!path)
1618		return -ENOMEM;
1619
1620	if (sk->tree_id == 0) {
1621		/* Search the root that we got passed. */
1622		root = btrfs_grab_root(root);
1623	} else {
1624		/* Look up the root from the arguments. */
1625		root = btrfs_get_fs_root(info, sk->tree_id, true);
1626		if (IS_ERR(root))
1627			return PTR_ERR(root);
1628	}
1629
1630	key.objectid = sk->min_objectid;
1631	key.type = sk->min_type;
1632	key.offset = sk->min_offset;
1633
1634	while (1) {
1635		/*
1636		 * Ensure that the whole user buffer is faulted in at sub-page
1637		 * granularity, otherwise the loop may live-lock.
1638		 */
1639		if (fault_in_subpage_writeable(ubuf + sk_offset, *buf_size - sk_offset)) {
1640			ret = -EFAULT;
1641			break;
1642		}
1643
1644		ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1645		if (ret)
1646			break;
1647
1648		ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1649				 &sk_offset, &num_found);
1650		btrfs_release_path(path);
1651		if (ret)
1652			break;
1653
1654	}
1655	/* Normalize return values from btrfs_search_forward() and copy_to_sk(). */
1656	if (ret > 0)
1657		ret = 0;
1658
1659	sk->nr_items = num_found;
1660	btrfs_put_root(root);
1661	return ret;
1662}
1663
1664static noinline int btrfs_ioctl_tree_search(struct btrfs_root *root,
1665					    void __user *argp)
1666{
1667	struct btrfs_ioctl_search_args __user *uargs = argp;
1668	struct btrfs_ioctl_search_key sk;
1669	int ret;
1670	u64 buf_size;
1671
1672	if (!capable(CAP_SYS_ADMIN))
1673		return -EPERM;
1674
1675	if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1676		return -EFAULT;
1677
1678	buf_size = sizeof(uargs->buf);
1679
1680	ret = search_ioctl(root, &sk, &buf_size, uargs->buf);
1681
1682	/*
1683	 * In the origin implementation an overflow is handled by returning a
1684	 * search header with a len of zero, so reset ret.
1685	 */
1686	if (ret == -EOVERFLOW)
1687		ret = 0;
1688
1689	if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1690		ret = -EFAULT;
1691	return ret;
1692}
1693
1694static noinline int btrfs_ioctl_tree_search_v2(struct btrfs_root *root,
1695					       void __user *argp)
1696{
1697	struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1698	struct btrfs_ioctl_search_args_v2 args;
1699	int ret;
1700	u64 buf_size;
1701	const u64 buf_limit = SZ_16M;
1702
1703	if (!capable(CAP_SYS_ADMIN))
1704		return -EPERM;
1705
1706	/* copy search header and buffer size */
1707	if (copy_from_user(&args, uarg, sizeof(args)))
1708		return -EFAULT;
1709
1710	buf_size = args.buf_size;
1711
1712	/* limit result size to 16MB */
1713	if (buf_size > buf_limit)
1714		buf_size = buf_limit;
1715
1716	ret = search_ioctl(root, &args.key, &buf_size,
1717			   (char __user *)(&uarg->buf[0]));
1718	if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1719		ret = -EFAULT;
1720	else if (ret == -EOVERFLOW &&
1721		copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1722		ret = -EFAULT;
1723
1724	return ret;
1725}
1726
1727/*
1728 * Search INODE_REFs to identify path name of 'dirid' directory
1729 * in a 'tree_id' tree. and sets path name to 'name'.
1730 */
1731static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1732				u64 tree_id, u64 dirid, char *name)
1733{
1734	struct btrfs_root *root;
1735	struct btrfs_key key;
1736	char *ptr;
1737	int ret = -1;
1738	int slot;
1739	int len;
1740	int total_len = 0;
1741	struct btrfs_inode_ref *iref;
1742	struct extent_buffer *l;
1743	BTRFS_PATH_AUTO_FREE(path);
1744
1745	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1746		name[0]='\0';
1747		return 0;
1748	}
1749
1750	path = btrfs_alloc_path();
1751	if (!path)
1752		return -ENOMEM;
1753
1754	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1755
1756	root = btrfs_get_fs_root(info, tree_id, true);
1757	if (IS_ERR(root)) {
1758		ret = PTR_ERR(root);
1759		root = NULL;
1760		goto out;
1761	}
1762
1763	key.objectid = dirid;
1764	key.type = BTRFS_INODE_REF_KEY;
1765	key.offset = (u64)-1;
1766
1767	while (1) {
1768		ret = btrfs_search_backwards(root, &key, path);
1769		if (ret < 0)
1770			goto out;
1771		else if (ret > 0) {
1772			ret = -ENOENT;
1773			goto out;
1774		}
1775
1776		l = path->nodes[0];
1777		slot = path->slots[0];
1778
1779		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1780		len = btrfs_inode_ref_name_len(l, iref);
1781		ptr -= len + 1;
1782		total_len += len + 1;
1783		if (ptr < name) {
1784			ret = -ENAMETOOLONG;
1785			goto out;
1786		}
1787
1788		*(ptr + len) = '/';
1789		read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1790
1791		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1792			break;
1793
1794		btrfs_release_path(path);
1795		key.objectid = key.offset;
1796		key.offset = (u64)-1;
1797		dirid = key.objectid;
1798	}
1799	memmove(name, ptr, total_len);
1800	name[total_len] = '\0';
1801	ret = 0;
1802out:
1803	btrfs_put_root(root);
1804	return ret;
1805}
1806
1807static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1808				struct inode *inode,
1809				struct btrfs_ioctl_ino_lookup_user_args *args)
1810{
1811	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1812	u64 upper_limit = btrfs_ino(BTRFS_I(inode));
1813	u64 treeid = btrfs_root_id(BTRFS_I(inode)->root);
1814	u64 dirid = args->dirid;
1815	unsigned long item_off;
1816	unsigned long item_len;
1817	struct btrfs_inode_ref *iref;
1818	struct btrfs_root_ref *rref;
1819	struct btrfs_root *root = NULL;
1820	BTRFS_PATH_AUTO_FREE(path);
1821	struct btrfs_key key;
1822	struct extent_buffer *leaf;
1823	char *ptr;
1824	int slot;
1825	int len;
1826	int total_len = 0;
1827	int ret;
1828
1829	path = btrfs_alloc_path();
1830	if (!path)
1831		return -ENOMEM;
1832
1833	/*
1834	 * If the bottom subvolume does not exist directly under upper_limit,
1835	 * construct the path in from the bottom up.
1836	 */
1837	if (dirid != upper_limit) {
1838		ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1839
1840		root = btrfs_get_fs_root(fs_info, treeid, true);
1841		if (IS_ERR(root))
1842			return PTR_ERR(root);
1843
1844		key.objectid = dirid;
1845		key.type = BTRFS_INODE_REF_KEY;
1846		key.offset = (u64)-1;
1847		while (1) {
1848			struct btrfs_inode *temp_inode;
1849
1850			ret = btrfs_search_backwards(root, &key, path);
1851			if (ret < 0)
1852				goto out_put;
1853			else if (ret > 0) {
1854				ret = -ENOENT;
1855				goto out_put;
1856			}
1857
1858			leaf = path->nodes[0];
1859			slot = path->slots[0];
1860
1861			iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1862			len = btrfs_inode_ref_name_len(leaf, iref);
1863			ptr -= len + 1;
1864			total_len += len + 1;
1865			if (ptr < args->path) {
1866				ret = -ENAMETOOLONG;
1867				goto out_put;
1868			}
1869
1870			*(ptr + len) = '/';
1871			read_extent_buffer(leaf, ptr,
1872					(unsigned long)(iref + 1), len);
1873
1874			/*
1875			 * We don't need the path anymore, so release it and
1876			 * avoid deadlocks and lockdep warnings in case
1877			 * btrfs_iget() needs to lookup the inode from its root
1878			 * btree and lock the same leaf.
1879			 */
1880			btrfs_release_path(path);
1881			temp_inode = btrfs_iget(key.offset, root);
1882			if (IS_ERR(temp_inode)) {
1883				ret = PTR_ERR(temp_inode);
1884				goto out_put;
1885			}
1886			/* Check the read+exec permission of this directory. */
1887			ret = inode_permission(idmap, &temp_inode->vfs_inode,
1888					       MAY_READ | MAY_EXEC);
1889			iput(&temp_inode->vfs_inode);
1890			if (ret)
1891				goto out_put;
1892
1893			if (key.offset == upper_limit)
1894				break;
1895			if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
1896				ret = -EACCES;
1897				goto out_put;
1898			}
1899
1900			key.objectid = key.offset;
1901			key.offset = (u64)-1;
1902			dirid = key.objectid;
1903		}
1904
1905		memmove(args->path, ptr, total_len);
1906		args->path[total_len] = '\0';
1907		btrfs_put_root(root);
1908		root = NULL;
1909		btrfs_release_path(path);
1910	}
1911
1912	/* Get the bottom subvolume's name from ROOT_REF */
1913	key.objectid = treeid;
1914	key.type = BTRFS_ROOT_REF_KEY;
1915	key.offset = args->treeid;
1916	ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
1917	if (ret < 0)
1918		return ret;
1919	else if (ret > 0)
1920		return -ENOENT;
1921
1922	leaf = path->nodes[0];
1923	slot = path->slots[0];
1924	btrfs_item_key_to_cpu(leaf, &key, slot);
1925
1926	item_off = btrfs_item_ptr_offset(leaf, slot);
1927	item_len = btrfs_item_size(leaf, slot);
1928	/* Check if dirid in ROOT_REF corresponds to passed dirid */
1929	rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
1930	if (args->dirid != btrfs_root_ref_dirid(leaf, rref))
1931		return -EINVAL;
1932
1933	/* Copy subvolume's name */
1934	item_off += sizeof(struct btrfs_root_ref);
1935	item_len -= sizeof(struct btrfs_root_ref);
1936	read_extent_buffer(leaf, args->name, item_off, item_len);
1937	args->name[item_len] = 0;
1938
1939out_put:
1940	btrfs_put_root(root);
1941
1942	return ret;
1943}
1944
1945static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
1946					   void __user *argp)
1947{
1948	struct btrfs_ioctl_ino_lookup_args *args;
1949	int ret = 0;
1950
1951	args = memdup_user(argp, sizeof(*args));
1952	if (IS_ERR(args))
1953		return PTR_ERR(args);
1954
1955	/*
1956	 * Unprivileged query to obtain the containing subvolume root id. The
1957	 * path is reset so it's consistent with btrfs_search_path_in_tree.
1958	 */
1959	if (args->treeid == 0)
1960		args->treeid = btrfs_root_id(root);
1961
1962	if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
1963		args->name[0] = 0;
1964		goto out;
1965	}
1966
1967	if (!capable(CAP_SYS_ADMIN)) {
1968		ret = -EPERM;
1969		goto out;
1970	}
1971
1972	ret = btrfs_search_path_in_tree(root->fs_info,
1973					args->treeid, args->objectid,
1974					args->name);
1975
1976out:
1977	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1978		ret = -EFAULT;
1979
1980	kfree(args);
1981	return ret;
1982}
1983
1984/*
1985 * Version of ino_lookup ioctl (unprivileged)
1986 *
1987 * The main differences from ino_lookup ioctl are:
1988 *
1989 *   1. Read + Exec permission will be checked using inode_permission() during
1990 *      path construction. -EACCES will be returned in case of failure.
1991 *   2. Path construction will be stopped at the inode number which corresponds
1992 *      to the fd with which this ioctl is called. If constructed path does not
1993 *      exist under fd's inode, -EACCES will be returned.
1994 *   3. The name of bottom subvolume is also searched and filled.
1995 */
1996static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
1997{
1998	struct btrfs_ioctl_ino_lookup_user_args *args;
1999	struct inode *inode;
2000	int ret;
2001
2002	args = memdup_user(argp, sizeof(*args));
2003	if (IS_ERR(args))
2004		return PTR_ERR(args);
2005
2006	inode = file_inode(file);
2007
2008	if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2009	    btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2010		/*
2011		 * The subvolume does not exist under fd with which this is
2012		 * called
2013		 */
2014		kfree(args);
2015		return -EACCES;
2016	}
2017
2018	ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2019
2020	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2021		ret = -EFAULT;
2022
2023	kfree(args);
2024	return ret;
2025}
2026
2027/* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2028static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2029{
2030	struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2031	struct btrfs_fs_info *fs_info;
2032	struct btrfs_root *root;
2033	struct btrfs_path *path;
2034	struct btrfs_key key;
2035	struct btrfs_root_item *root_item;
2036	struct btrfs_root_ref *rref;
2037	struct extent_buffer *leaf;
2038	unsigned long item_off;
2039	unsigned long item_len;
2040	int slot;
2041	int ret = 0;
2042
2043	path = btrfs_alloc_path();
2044	if (!path)
2045		return -ENOMEM;
2046
2047	subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2048	if (!subvol_info) {
2049		btrfs_free_path(path);
2050		return -ENOMEM;
2051	}
2052
2053	fs_info = BTRFS_I(inode)->root->fs_info;
2054
2055	/* Get root_item of inode's subvolume */
2056	key.objectid = btrfs_root_id(BTRFS_I(inode)->root);
2057	root = btrfs_get_fs_root(fs_info, key.objectid, true);
2058	if (IS_ERR(root)) {
2059		ret = PTR_ERR(root);
2060		goto out_free;
2061	}
2062	root_item = &root->root_item;
2063
2064	subvol_info->treeid = key.objectid;
2065
2066	subvol_info->generation = btrfs_root_generation(root_item);
2067	subvol_info->flags = btrfs_root_flags(root_item);
2068
2069	memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2070	memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2071						    BTRFS_UUID_SIZE);
2072	memcpy(subvol_info->received_uuid, root_item->received_uuid,
2073						    BTRFS_UUID_SIZE);
2074
2075	subvol_info->ctransid = btrfs_root_ctransid(root_item);
2076	subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2077	subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2078
2079	subvol_info->otransid = btrfs_root_otransid(root_item);
2080	subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2081	subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2082
2083	subvol_info->stransid = btrfs_root_stransid(root_item);
2084	subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2085	subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2086
2087	subvol_info->rtransid = btrfs_root_rtransid(root_item);
2088	subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2089	subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2090
2091	if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2092		/* Search root tree for ROOT_BACKREF of this subvolume */
2093		key.type = BTRFS_ROOT_BACKREF_KEY;
2094		key.offset = 0;
2095		ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2096		if (ret < 0) {
2097			goto out;
2098		} else if (path->slots[0] >=
2099			   btrfs_header_nritems(path->nodes[0])) {
2100			ret = btrfs_next_leaf(fs_info->tree_root, path);
2101			if (ret < 0) {
2102				goto out;
2103			} else if (unlikely(ret > 0)) {
2104				ret = -EUCLEAN;
2105				goto out;
2106			}
2107		}
2108
2109		leaf = path->nodes[0];
2110		slot = path->slots[0];
2111		btrfs_item_key_to_cpu(leaf, &key, slot);
2112		if (key.objectid == subvol_info->treeid &&
2113		    key.type == BTRFS_ROOT_BACKREF_KEY) {
2114			subvol_info->parent_id = key.offset;
2115
2116			rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2117			subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2118
2119			item_off = btrfs_item_ptr_offset(leaf, slot)
2120					+ sizeof(struct btrfs_root_ref);
2121			item_len = btrfs_item_size(leaf, slot)
2122					- sizeof(struct btrfs_root_ref);
2123			read_extent_buffer(leaf, subvol_info->name,
2124					   item_off, item_len);
2125		} else {
2126			ret = -ENOENT;
2127			goto out;
2128		}
2129	}
2130
2131	btrfs_free_path(path);
2132	path = NULL;
2133	if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2134		ret = -EFAULT;
2135
2136out:
2137	btrfs_put_root(root);
2138out_free:
2139	btrfs_free_path(path);
2140	kfree(subvol_info);
2141	return ret;
2142}
2143
2144/*
2145 * Return ROOT_REF information of the subvolume containing this inode
2146 * except the subvolume name.
2147 */
2148static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2149					  void __user *argp)
2150{
2151	struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2152	struct btrfs_root_ref *rref;
2153	struct btrfs_path *path;
2154	struct btrfs_key key;
2155	struct extent_buffer *leaf;
2156	u64 objectid;
2157	int slot;
2158	int ret;
2159	u8 found;
2160
2161	path = btrfs_alloc_path();
2162	if (!path)
2163		return -ENOMEM;
2164
2165	rootrefs = memdup_user(argp, sizeof(*rootrefs));
2166	if (IS_ERR(rootrefs)) {
2167		btrfs_free_path(path);
2168		return PTR_ERR(rootrefs);
2169	}
2170
2171	objectid = btrfs_root_id(root);
2172	key.objectid = objectid;
2173	key.type = BTRFS_ROOT_REF_KEY;
2174	key.offset = rootrefs->min_treeid;
2175	found = 0;
2176
2177	root = root->fs_info->tree_root;
2178	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2179	if (ret < 0) {
2180		goto out;
2181	} else if (path->slots[0] >=
2182		   btrfs_header_nritems(path->nodes[0])) {
2183		ret = btrfs_next_leaf(root, path);
2184		if (ret < 0) {
2185			goto out;
2186		} else if (unlikely(ret > 0)) {
2187			ret = -EUCLEAN;
2188			goto out;
2189		}
2190	}
2191	while (1) {
2192		leaf = path->nodes[0];
2193		slot = path->slots[0];
2194
2195		btrfs_item_key_to_cpu(leaf, &key, slot);
2196		if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2197			ret = 0;
2198			goto out;
2199		}
2200
2201		if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2202			ret = -EOVERFLOW;
2203			goto out;
2204		}
2205
2206		rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2207		rootrefs->rootref[found].treeid = key.offset;
2208		rootrefs->rootref[found].dirid =
2209				  btrfs_root_ref_dirid(leaf, rref);
2210		found++;
2211
2212		ret = btrfs_next_item(root, path);
2213		if (ret < 0) {
2214			goto out;
2215		} else if (unlikely(ret > 0)) {
2216			ret = -EUCLEAN;
2217			goto out;
2218		}
2219	}
2220
2221out:
2222	btrfs_free_path(path);
2223
2224	if (!ret || ret == -EOVERFLOW) {
2225		rootrefs->num_items = found;
2226		/* update min_treeid for next search */
2227		if (found)
2228			rootrefs->min_treeid =
2229				rootrefs->rootref[found - 1].treeid + 1;
2230		if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2231			ret = -EFAULT;
2232	}
2233
2234	kfree(rootrefs);
2235
2236	return ret;
2237}
2238
2239static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2240					     void __user *arg,
2241					     bool destroy_v2)
2242{
2243	struct dentry *parent = file->f_path.dentry;
2244	struct dentry *dentry;
2245	struct inode *dir = d_inode(parent);
2246	struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
2247	struct inode *inode;
2248	struct btrfs_root *root = BTRFS_I(dir)->root;
2249	struct btrfs_root *dest = NULL;
2250	struct btrfs_ioctl_vol_args *vol_args = NULL;
2251	struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2252	struct mnt_idmap *idmap = file_mnt_idmap(file);
2253	char *subvol_name, *subvol_name_ptr = NULL;
2254	int ret = 0;
2255	bool destroy_parent = false;
2256
2257	/* We don't support snapshots with extent tree v2 yet. */
2258	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2259		btrfs_err(fs_info,
2260			  "extent tree v2 doesn't support snapshot deletion yet");
2261		return -EOPNOTSUPP;
2262	}
2263
2264	if (destroy_v2) {
2265		vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2266		if (IS_ERR(vol_args2))
2267			return PTR_ERR(vol_args2);
2268
2269		if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2270			ret = -EOPNOTSUPP;
2271			goto out;
2272		}
2273
2274		/*
2275		 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2276		 * name, same as v1 currently does.
2277		 */
2278		if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2279			ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args2);
2280			if (ret < 0)
2281				goto out;
2282			subvol_name = vol_args2->name;
2283
2284			ret = mnt_want_write_file(file);
2285			if (ret)
2286				goto out;
2287		} else {
2288			struct inode *old_dir;
2289
2290			if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2291				ret = -EINVAL;
2292				goto out;
2293			}
2294
2295			ret = mnt_want_write_file(file);
2296			if (ret)
2297				goto out;
2298
2299			dentry = btrfs_get_dentry(fs_info->sb,
2300					BTRFS_FIRST_FREE_OBJECTID,
2301					vol_args2->subvolid, 0);
2302			if (IS_ERR(dentry)) {
2303				ret = PTR_ERR(dentry);
2304				goto out_drop_write;
2305			}
2306
2307			/*
2308			 * Change the default parent since the subvolume being
2309			 * deleted can be outside of the current mount point.
2310			 */
2311			parent = btrfs_get_parent(dentry);
2312
2313			/*
2314			 * At this point dentry->d_name can point to '/' if the
2315			 * subvolume we want to destroy is outsite of the
2316			 * current mount point, so we need to release the
2317			 * current dentry and execute the lookup to return a new
2318			 * one with ->d_name pointing to the
2319			 * <mount point>/subvol_name.
2320			 */
2321			dput(dentry);
2322			if (IS_ERR(parent)) {
2323				ret = PTR_ERR(parent);
2324				goto out_drop_write;
2325			}
2326			old_dir = dir;
2327			dir = d_inode(parent);
2328
2329			/*
2330			 * If v2 was used with SPEC_BY_ID, a new parent was
2331			 * allocated since the subvolume can be outside of the
2332			 * current mount point. Later on we need to release this
2333			 * new parent dentry.
2334			 */
2335			destroy_parent = true;
2336
2337			/*
2338			 * On idmapped mounts, deletion via subvolid is
2339			 * restricted to subvolumes that are immediate
2340			 * ancestors of the inode referenced by the file
2341			 * descriptor in the ioctl. Otherwise the idmapping
2342			 * could potentially be abused to delete subvolumes
2343			 * anywhere in the filesystem the user wouldn't be able
2344			 * to delete without an idmapped mount.
2345			 */
2346			if (old_dir != dir && idmap != &nop_mnt_idmap) {
2347				ret = -EOPNOTSUPP;
2348				goto free_parent;
2349			}
2350
2351			subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2352						fs_info, vol_args2->subvolid);
2353			if (IS_ERR(subvol_name_ptr)) {
2354				ret = PTR_ERR(subvol_name_ptr);
2355				goto free_parent;
2356			}
2357			/* subvol_name_ptr is already nul terminated */
2358			subvol_name = (char *)kbasename(subvol_name_ptr);
2359		}
2360	} else {
2361		vol_args = memdup_user(arg, sizeof(*vol_args));
2362		if (IS_ERR(vol_args))
2363			return PTR_ERR(vol_args);
2364
2365		ret = btrfs_check_ioctl_vol_args_path(vol_args);
2366		if (ret < 0)
2367			goto out;
2368
2369		subvol_name = vol_args->name;
2370
2371		ret = mnt_want_write_file(file);
2372		if (ret)
2373			goto out;
2374	}
2375
2376	if (strchr(subvol_name, '/') ||
2377	    strcmp(subvol_name, "..") == 0) {
2378		ret = -EINVAL;
2379		goto free_subvol_name;
2380	}
2381
2382	if (!S_ISDIR(dir->i_mode)) {
2383		ret = -ENOTDIR;
2384		goto free_subvol_name;
2385	}
2386
2387	dentry = start_removing_killable(idmap, parent, &QSTR(subvol_name));
2388	if (IS_ERR(dentry)) {
2389		ret = PTR_ERR(dentry);
2390		goto out_end_removing;
2391	}
2392
2393	inode = d_inode(dentry);
2394	dest = BTRFS_I(inode)->root;
2395	if (!capable(CAP_SYS_ADMIN)) {
2396		/*
2397		 * Regular user.  Only allow this with a special mount
2398		 * option, when the user has write+exec access to the
2399		 * subvol root, and when rmdir(2) would have been
2400		 * allowed.
2401		 *
2402		 * Note that this is _not_ check that the subvol is
2403		 * empty or doesn't contain data that we wouldn't
2404		 * otherwise be able to delete.
2405		 *
2406		 * Users who want to delete empty subvols should try
2407		 * rmdir(2).
2408		 */
2409		ret = -EPERM;
2410		if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2411			goto out_end_removing;
2412
2413		/*
2414		 * Do not allow deletion if the parent dir is the same
2415		 * as the dir to be deleted.  That means the ioctl
2416		 * must be called on the dentry referencing the root
2417		 * of the subvol, not a random directory contained
2418		 * within it.
2419		 */
2420		ret = -EINVAL;
2421		if (root == dest)
2422			goto out_end_removing;
2423
2424		ret = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2425		if (ret)
2426			goto out_end_removing;
2427	}
2428
2429	/* check if subvolume may be deleted by a user */
2430	ret = btrfs_may_delete(idmap, dir, dentry, 1);
2431	if (ret)
2432		goto out_end_removing;
2433
2434	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2435		ret = -EINVAL;
2436		goto out_end_removing;
2437	}
2438
2439	btrfs_inode_lock(BTRFS_I(inode), 0);
2440	ret = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2441	btrfs_inode_unlock(BTRFS_I(inode), 0);
2442	if (!ret)
2443		d_delete_notify(dir, dentry);
2444
2445out_end_removing:
2446	end_removing(dentry);
2447free_subvol_name:
2448	kfree(subvol_name_ptr);
2449free_parent:
2450	if (destroy_parent)
2451		dput(parent);
2452out_drop_write:
2453	mnt_drop_write_file(file);
2454out:
2455	kfree(vol_args2);
2456	kfree(vol_args);
2457	return ret;
2458}
2459
2460static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2461{
2462	struct inode *inode = file_inode(file);
2463	struct btrfs_root *root = BTRFS_I(inode)->root;
2464	struct btrfs_ioctl_defrag_range_args range = {0};
2465	int ret;
2466
2467	ret = mnt_want_write_file(file);
2468	if (ret)
2469		return ret;
2470
2471	if (btrfs_root_readonly(root)) {
2472		ret = -EROFS;
2473		goto out;
2474	}
2475
2476	switch (inode->i_mode & S_IFMT) {
2477	case S_IFDIR:
2478		if (!capable(CAP_SYS_ADMIN)) {
2479			ret = -EPERM;
2480			goto out;
2481		}
2482		ret = btrfs_defrag_root(root);
2483		break;
2484	case S_IFREG:
2485		/*
2486		 * Note that this does not check the file descriptor for write
2487		 * access. This prevents defragmenting executables that are
2488		 * running and allows defrag on files open in read-only mode.
2489		 */
2490		if (!capable(CAP_SYS_ADMIN) &&
2491		    inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2492			ret = -EPERM;
2493			goto out;
2494		}
2495
2496		/*
2497		 * Don't allow defrag on pre-content watched files, as it could
2498		 * populate the page cache with 0's via readahead.
2499		 */
2500		if (unlikely(FMODE_FSNOTIFY_HSM(file->f_mode))) {
2501			ret = -EINVAL;
2502			goto out;
2503		}
2504
2505		if (argp) {
2506			if (copy_from_user(&range, argp, sizeof(range))) {
2507				ret = -EFAULT;
2508				goto out;
2509			}
2510			if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) {
2511				ret = -EOPNOTSUPP;
2512				goto out;
2513			}
2514			if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS) &&
2515			    (range.flags & BTRFS_DEFRAG_RANGE_NOCOMPRESS)) {
2516				ret = -EINVAL;
2517				goto out;
2518			}
2519			/* Compression or no-compression require to start the IO. */
2520			if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS) ||
2521			    (range.flags & BTRFS_DEFRAG_RANGE_NOCOMPRESS)) {
2522				range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2523				range.extent_thresh = (u32)-1;
2524			}
2525		} else {
2526			/* the rest are all set to zero by kzalloc */
2527			range.len = (u64)-1;
2528		}
2529		ret = btrfs_defrag_file(BTRFS_I(file_inode(file)), &file->f_ra,
2530					&range, BTRFS_OLDEST_GENERATION, 0);
2531		if (ret > 0)
2532			ret = 0;
2533		break;
2534	default:
2535		ret = -EINVAL;
2536	}
2537out:
2538	mnt_drop_write_file(file);
2539	return ret;
2540}
2541
2542static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2543{
2544	struct btrfs_ioctl_vol_args *vol_args;
2545	bool restore_op = false;
2546	int ret;
2547
2548	if (!capable(CAP_SYS_ADMIN))
2549		return -EPERM;
2550
2551	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2552		btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2553		return -EINVAL;
2554	}
2555
2556	if (fs_info->fs_devices->temp_fsid) {
2557		btrfs_err(fs_info,
2558			  "device add not supported on cloned temp-fsid mount");
2559		return -EINVAL;
2560	}
2561
2562	if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2563		if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2564			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2565
2566		/*
2567		 * We can do the device add because we have a paused balanced,
2568		 * change the exclusive op type and remember we should bring
2569		 * back the paused balance
2570		 */
2571		fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2572		btrfs_exclop_start_unlock(fs_info);
2573		restore_op = true;
2574	}
2575
2576	vol_args = memdup_user(arg, sizeof(*vol_args));
2577	if (IS_ERR(vol_args)) {
2578		ret = PTR_ERR(vol_args);
2579		goto out;
2580	}
2581
2582	ret = btrfs_check_ioctl_vol_args_path(vol_args);
2583	if (ret < 0)
2584		goto out_free;
2585
2586	ret = btrfs_init_new_device(fs_info, vol_args->name);
2587
2588	if (!ret)
2589		btrfs_info(fs_info, "disk added %s", vol_args->name);
2590
2591out_free:
2592	kfree(vol_args);
2593out:
2594	if (restore_op)
2595		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2596	else
2597		btrfs_exclop_finish(fs_info);
2598	return ret;
2599}
2600
2601static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2602{
2603	BTRFS_DEV_LOOKUP_ARGS(args);
2604	struct inode *inode = file_inode(file);
2605	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2606	struct btrfs_ioctl_vol_args_v2 *vol_args;
2607	struct file *bdev_file = NULL;
2608	int ret;
2609	bool cancel = false;
2610
2611	if (!capable(CAP_SYS_ADMIN))
2612		return -EPERM;
2613
2614	vol_args = memdup_user(arg, sizeof(*vol_args));
2615	if (IS_ERR(vol_args))
2616		return PTR_ERR(vol_args);
2617
2618	if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2619		ret = -EOPNOTSUPP;
2620		goto out;
2621	}
2622
2623	ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args);
2624	if (ret < 0)
2625		goto out;
2626
2627	if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2628		args.devid = vol_args->devid;
2629	} else if (!strcmp("cancel", vol_args->name)) {
2630		cancel = true;
2631	} else {
2632		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2633		if (ret)
2634			goto out;
2635	}
2636
2637	ret = mnt_want_write_file(file);
2638	if (ret)
2639		goto out;
2640
2641	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2642					   cancel);
2643	if (ret)
2644		goto err_drop;
2645
2646	/* Exclusive operation is now claimed */
2647	ret = btrfs_rm_device(fs_info, &args, &bdev_file);
2648
2649	btrfs_exclop_finish(fs_info);
2650
2651	if (!ret) {
2652		if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2653			btrfs_info(fs_info, "device deleted: id %llu",
2654					vol_args->devid);
2655		else
2656			btrfs_info(fs_info, "device deleted: %s",
2657					vol_args->name);
2658	}
2659err_drop:
2660	mnt_drop_write_file(file);
2661	if (bdev_file)
2662		bdev_fput(bdev_file);
2663out:
2664	btrfs_put_dev_args_from_path(&args);
2665	kfree(vol_args);
2666	return ret;
2667}
2668
2669static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2670{
2671	BTRFS_DEV_LOOKUP_ARGS(args);
2672	struct inode *inode = file_inode(file);
2673	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2674	struct btrfs_ioctl_vol_args *vol_args;
2675	struct file *bdev_file = NULL;
2676	int ret;
2677	bool cancel = false;
2678
2679	if (!capable(CAP_SYS_ADMIN))
2680		return -EPERM;
2681
2682	vol_args = memdup_user(arg, sizeof(*vol_args));
2683	if (IS_ERR(vol_args))
2684		return PTR_ERR(vol_args);
2685
2686	ret = btrfs_check_ioctl_vol_args_path(vol_args);
2687	if (ret < 0)
2688		goto out_free;
2689
2690	if (!strcmp("cancel", vol_args->name)) {
2691		cancel = true;
2692	} else {
2693		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2694		if (ret)
2695			goto out;
2696	}
2697
2698	ret = mnt_want_write_file(file);
2699	if (ret)
2700		goto out;
2701
2702	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2703					   cancel);
2704	if (ret == 0) {
2705		ret = btrfs_rm_device(fs_info, &args, &bdev_file);
2706		if (!ret)
2707			btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2708		btrfs_exclop_finish(fs_info);
2709	}
2710
2711	mnt_drop_write_file(file);
2712	if (bdev_file)
2713		bdev_fput(bdev_file);
2714out:
2715	btrfs_put_dev_args_from_path(&args);
2716out_free:
2717	kfree(vol_args);
2718	return ret;
2719}
2720
2721static long btrfs_ioctl_fs_info(const struct btrfs_fs_info *fs_info,
2722				void __user *arg)
2723{
2724	struct btrfs_ioctl_fs_info_args *fi_args;
2725	struct btrfs_device *device;
2726	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2727	u64 flags_in;
2728	int ret = 0;
2729
2730	fi_args = memdup_user(arg, sizeof(*fi_args));
2731	if (IS_ERR(fi_args))
2732		return PTR_ERR(fi_args);
2733
2734	flags_in = fi_args->flags;
2735	memset(fi_args, 0, sizeof(*fi_args));
2736
2737	rcu_read_lock();
2738	fi_args->num_devices = fs_devices->num_devices;
2739
2740	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2741		if (device->devid > fi_args->max_id)
2742			fi_args->max_id = device->devid;
2743	}
2744	rcu_read_unlock();
2745
2746	memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2747	fi_args->nodesize = fs_info->nodesize;
2748	fi_args->sectorsize = fs_info->sectorsize;
2749	fi_args->clone_alignment = fs_info->sectorsize;
2750
2751	if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2752		fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2753		fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2754		fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2755	}
2756
2757	if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2758		fi_args->generation = btrfs_get_fs_generation(fs_info);
2759		fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2760	}
2761
2762	if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2763		memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2764		       sizeof(fi_args->metadata_uuid));
2765		fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2766	}
2767
2768	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2769		ret = -EFAULT;
2770
2771	kfree(fi_args);
2772	return ret;
2773}
2774
2775static long btrfs_ioctl_dev_info(const struct btrfs_fs_info *fs_info,
2776				 void __user *arg)
2777{
2778	BTRFS_DEV_LOOKUP_ARGS(args);
2779	struct btrfs_ioctl_dev_info_args *di_args;
2780	struct btrfs_device *dev;
2781	int ret = 0;
2782
2783	di_args = memdup_user(arg, sizeof(*di_args));
2784	if (IS_ERR(di_args))
2785		return PTR_ERR(di_args);
2786
2787	args.devid = di_args->devid;
2788	if (!btrfs_is_empty_uuid(di_args->uuid))
2789		args.uuid = di_args->uuid;
2790
2791	rcu_read_lock();
2792	dev = btrfs_find_device(fs_info->fs_devices, &args);
2793	if (!dev) {
2794		ret = -ENODEV;
2795		goto out;
2796	}
2797
2798	di_args->devid = dev->devid;
2799	di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2800	di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2801	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2802	memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2803	if (dev->name)
2804		strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2805	else
2806		di_args->path[0] = '\0';
2807
2808out:
2809	rcu_read_unlock();
2810	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2811		ret = -EFAULT;
2812
2813	kfree(di_args);
2814	return ret;
2815}
2816
2817static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2818{
2819	struct inode *inode = file_inode(file);
2820	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2821	struct btrfs_root *root = BTRFS_I(inode)->root;
2822	struct btrfs_root *new_root;
2823	struct btrfs_dir_item *di;
2824	struct btrfs_trans_handle *trans;
2825	struct btrfs_path *path = NULL;
2826	struct btrfs_disk_key disk_key;
2827	struct fscrypt_str name = FSTR_INIT("default", 7);
2828	u64 objectid = 0;
2829	u64 dir_id;
2830	int ret;
2831
2832	if (!capable(CAP_SYS_ADMIN))
2833		return -EPERM;
2834
2835	ret = mnt_want_write_file(file);
2836	if (ret)
2837		return ret;
2838
2839	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2840		ret = -EFAULT;
2841		goto out;
2842	}
2843
2844	if (!objectid)
2845		objectid = BTRFS_FS_TREE_OBJECTID;
2846
2847	new_root = btrfs_get_fs_root(fs_info, objectid, true);
2848	if (IS_ERR(new_root)) {
2849		ret = PTR_ERR(new_root);
2850		goto out;
2851	}
2852	if (!btrfs_is_fstree(btrfs_root_id(new_root))) {
2853		ret = -ENOENT;
2854		goto out_free;
2855	}
2856
2857	path = btrfs_alloc_path();
2858	if (!path) {
2859		ret = -ENOMEM;
2860		goto out_free;
2861	}
2862
2863	trans = btrfs_start_transaction(root, 1);
2864	if (IS_ERR(trans)) {
2865		ret = PTR_ERR(trans);
2866		goto out_free;
2867	}
2868
2869	dir_id = btrfs_super_root_dir(fs_info->super_copy);
2870	di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2871				   dir_id, &name, 1);
2872	if (IS_ERR_OR_NULL(di)) {
2873		btrfs_release_path(path);
2874		btrfs_end_transaction(trans);
2875		btrfs_err(fs_info,
2876			  "Umm, you don't have the default diritem, this isn't going to work");
2877		ret = -ENOENT;
2878		goto out_free;
2879	}
2880
2881	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2882	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2883	btrfs_release_path(path);
2884
2885	btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
2886	btrfs_end_transaction(trans);
2887out_free:
2888	btrfs_put_root(new_root);
2889	btrfs_free_path(path);
2890out:
2891	mnt_drop_write_file(file);
2892	return ret;
2893}
2894
2895static void get_block_group_info(struct list_head *groups_list,
2896				 struct btrfs_ioctl_space_info *space)
2897{
2898	struct btrfs_block_group *block_group;
2899
2900	space->total_bytes = 0;
2901	space->used_bytes = 0;
2902	space->flags = 0;
2903	list_for_each_entry(block_group, groups_list, list) {
2904		space->flags = block_group->flags;
2905		space->total_bytes += block_group->length;
2906		space->used_bytes += block_group->used;
2907	}
2908}
2909
2910static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
2911				   void __user *arg)
2912{
2913	struct btrfs_ioctl_space_args space_args = { 0 };
2914	struct btrfs_ioctl_space_info space;
2915	struct btrfs_ioctl_space_info *dest;
2916	struct btrfs_ioctl_space_info AUTO_KFREE(dest_orig);
2917	struct btrfs_ioctl_space_info __user *user_dest;
2918	struct btrfs_space_info *info;
2919	static const u64 types[] = {
2920		BTRFS_BLOCK_GROUP_DATA,
2921		BTRFS_BLOCK_GROUP_SYSTEM,
2922		BTRFS_BLOCK_GROUP_METADATA,
2923		BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
2924	};
2925	int num_types = 4;
2926	int alloc_size;
2927	int ret = 0;
2928	u64 slot_count = 0;
2929	int i, c;
2930
2931	if (copy_from_user(&space_args,
2932			   (struct btrfs_ioctl_space_args __user *)arg,
2933			   sizeof(space_args)))
2934		return -EFAULT;
2935
2936	for (i = 0; i < num_types; i++) {
2937		struct btrfs_space_info *tmp;
2938
2939		info = NULL;
2940		list_for_each_entry(tmp, &fs_info->space_info, list) {
2941			if (tmp->flags == types[i]) {
2942				info = tmp;
2943				break;
2944			}
2945		}
2946
2947		if (!info)
2948			continue;
2949
2950		down_read(&info->groups_sem);
2951		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2952			if (!list_empty(&info->block_groups[c]))
2953				slot_count++;
2954		}
2955		up_read(&info->groups_sem);
2956	}
2957
2958	/*
2959	 * Global block reserve, exported as a space_info
2960	 */
2961	slot_count++;
2962
2963	/* space_slots == 0 means they are asking for a count */
2964	if (space_args.space_slots == 0) {
2965		space_args.total_spaces = slot_count;
2966		goto out;
2967	}
2968
2969	slot_count = min_t(u64, space_args.space_slots, slot_count);
2970
2971	alloc_size = sizeof(*dest) * slot_count;
2972
2973	/* we generally have at most 6 or so space infos, one for each raid
2974	 * level.  So, a whole page should be more than enough for everyone
2975	 */
2976	if (alloc_size > PAGE_SIZE)
2977		return -ENOMEM;
2978
2979	space_args.total_spaces = 0;
2980	dest = kmalloc(alloc_size, GFP_KERNEL);
2981	if (!dest)
2982		return -ENOMEM;
2983	dest_orig = dest;
2984
2985	/* now we have a buffer to copy into */
2986	for (i = 0; i < num_types; i++) {
2987		struct btrfs_space_info *tmp;
2988
2989		if (!slot_count)
2990			break;
2991
2992		info = NULL;
2993		list_for_each_entry(tmp, &fs_info->space_info, list) {
2994			if (tmp->flags == types[i]) {
2995				info = tmp;
2996				break;
2997			}
2998		}
2999
3000		if (!info)
3001			continue;
3002		down_read(&info->groups_sem);
3003		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3004			if (!list_empty(&info->block_groups[c])) {
3005				get_block_group_info(&info->block_groups[c],
3006						     &space);
3007				memcpy(dest, &space, sizeof(space));
3008				dest++;
3009				space_args.total_spaces++;
3010				slot_count--;
3011			}
3012			if (!slot_count)
3013				break;
3014		}
3015		up_read(&info->groups_sem);
3016	}
3017
3018	/*
3019	 * Add global block reserve
3020	 */
3021	if (slot_count) {
3022		struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3023
3024		spin_lock(&block_rsv->lock);
3025		space.total_bytes = block_rsv->size;
3026		space.used_bytes = block_rsv->size - block_rsv->reserved;
3027		spin_unlock(&block_rsv->lock);
3028		space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3029		memcpy(dest, &space, sizeof(space));
3030		space_args.total_spaces++;
3031	}
3032
3033	user_dest = (struct btrfs_ioctl_space_info __user *)
3034		(arg + sizeof(struct btrfs_ioctl_space_args));
3035
3036	if (copy_to_user(user_dest, dest_orig, alloc_size))
3037		return -EFAULT;
3038
3039out:
3040	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3041		ret = -EFAULT;
3042
3043	return ret;
3044}
3045
3046static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3047					    void __user *argp)
3048{
3049	struct btrfs_trans_handle *trans;
3050	u64 transid;
3051
3052	/*
3053	 * Start orphan cleanup here for the given root in case it hasn't been
3054	 * started already by other means. Errors are handled in the other
3055	 * functions during transaction commit.
3056	 */
3057	btrfs_orphan_cleanup(root);
3058
3059	trans = btrfs_attach_transaction_barrier(root);
3060	if (IS_ERR(trans)) {
3061		if (PTR_ERR(trans) != -ENOENT)
3062			return PTR_ERR(trans);
3063
3064		/* No running transaction, don't bother */
3065		transid = btrfs_get_last_trans_committed(root->fs_info);
3066		goto out;
3067	}
3068	transid = trans->transid;
3069	btrfs_commit_transaction_async(trans);
3070out:
3071	if (argp)
3072		if (copy_to_user(argp, &transid, sizeof(transid)))
3073			return -EFAULT;
3074	return 0;
3075}
3076
3077static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3078					   void __user *argp)
3079{
3080	/* By default wait for the current transaction. */
3081	u64 transid = 0;
3082
3083	if (argp)
3084		if (copy_from_user(&transid, argp, sizeof(transid)))
3085			return -EFAULT;
3086
3087	return btrfs_wait_for_commit(fs_info, transid);
3088}
3089
3090static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3091{
3092	struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file));
3093	struct btrfs_ioctl_scrub_args *sa;
3094	int ret;
3095
3096	if (!capable(CAP_SYS_ADMIN))
3097		return -EPERM;
3098
3099	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3100		btrfs_err(fs_info, "scrub: extent tree v2 not yet supported");
3101		return -EINVAL;
3102	}
3103
3104	sa = memdup_user(arg, sizeof(*sa));
3105	if (IS_ERR(sa))
3106		return PTR_ERR(sa);
3107
3108	if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3109		ret = -EOPNOTSUPP;
3110		goto out;
3111	}
3112
3113	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3114		ret = mnt_want_write_file(file);
3115		if (ret)
3116			goto out;
3117	}
3118
3119	ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3120			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3121			      0);
3122
3123	/*
3124	 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3125	 * error. This is important as it allows user space to know how much
3126	 * progress scrub has done. For example, if scrub is canceled we get
3127	 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3128	 * space. Later user space can inspect the progress from the structure
3129	 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3130	 * previously (btrfs-progs does this).
3131	 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3132	 * then return -EFAULT to signal the structure was not copied or it may
3133	 * be corrupt and unreliable due to a partial copy.
3134	 */
3135	if (copy_to_user(arg, sa, sizeof(*sa)))
3136		ret = -EFAULT;
3137
3138	if (!(sa->flags & BTRFS_SCRUB_READONLY))
3139		mnt_drop_write_file(file);
3140out:
3141	kfree(sa);
3142	return ret;
3143}
3144
3145static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3146{
3147	if (!capable(CAP_SYS_ADMIN))
3148		return -EPERM;
3149
3150	return btrfs_scrub_cancel(fs_info);
3151}
3152
3153static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3154				       void __user *arg)
3155{
3156	struct btrfs_ioctl_scrub_args *sa;
3157	int ret;
3158
3159	if (!capable(CAP_SYS_ADMIN))
3160		return -EPERM;
3161
3162	sa = memdup_user(arg, sizeof(*sa));
3163	if (IS_ERR(sa))
3164		return PTR_ERR(sa);
3165
3166	ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3167
3168	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3169		ret = -EFAULT;
3170
3171	kfree(sa);
3172	return ret;
3173}
3174
3175static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3176				      void __user *arg)
3177{
3178	struct btrfs_ioctl_get_dev_stats *sa;
3179	int ret;
3180
3181	sa = memdup_user(arg, sizeof(*sa));
3182	if (IS_ERR(sa))
3183		return PTR_ERR(sa);
3184
3185	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3186		kfree(sa);
3187		return -EPERM;
3188	}
3189
3190	ret = btrfs_get_dev_stats(fs_info, sa);
3191
3192	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3193		ret = -EFAULT;
3194
3195	kfree(sa);
3196	return ret;
3197}
3198
3199static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3200				    void __user *arg)
3201{
3202	struct btrfs_ioctl_dev_replace_args *p;
3203	int ret;
3204
3205	if (!capable(CAP_SYS_ADMIN))
3206		return -EPERM;
3207
3208	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3209		btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3210		return -EINVAL;
3211	}
3212
3213	p = memdup_user(arg, sizeof(*p));
3214	if (IS_ERR(p))
3215		return PTR_ERR(p);
3216
3217	switch (p->cmd) {
3218	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3219		if (sb_rdonly(fs_info->sb)) {
3220			ret = -EROFS;
3221			goto out;
3222		}
3223		if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3224			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3225		} else {
3226			ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3227			btrfs_exclop_finish(fs_info);
3228		}
3229		break;
3230	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3231		btrfs_dev_replace_status(fs_info, p);
3232		ret = 0;
3233		break;
3234	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3235		p->result = btrfs_dev_replace_cancel(fs_info);
3236		ret = 0;
3237		break;
3238	default:
3239		ret = -EINVAL;
3240		break;
3241	}
3242
3243	if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3244		ret = -EFAULT;
3245out:
3246	kfree(p);
3247	return ret;
3248}
3249
3250static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3251{
3252	int ret = 0;
3253	int i;
3254	u64 rel_ptr;
3255	int size;
3256	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3257	struct inode_fs_paths *ipath __free(inode_fs_paths) = NULL;
3258	struct btrfs_path *path;
3259
3260	if (!capable(CAP_DAC_READ_SEARCH))
3261		return -EPERM;
3262
3263	path = btrfs_alloc_path();
3264	if (!path) {
3265		ret = -ENOMEM;
3266		goto out;
3267	}
3268
3269	ipa = memdup_user(arg, sizeof(*ipa));
3270	if (IS_ERR(ipa)) {
3271		ret = PTR_ERR(ipa);
3272		ipa = NULL;
3273		goto out;
3274	}
3275
3276	size = min_t(u32, ipa->size, 4096);
3277	ipath = init_ipath(size, root, path);
3278	if (IS_ERR(ipath)) {
3279		ret = PTR_ERR(ipath);
3280		ipath = NULL;
3281		goto out;
3282	}
3283
3284	ret = paths_from_inode(ipa->inum, ipath);
3285	if (ret < 0)
3286		goto out;
3287
3288	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3289		rel_ptr = ipath->fspath->val[i] -
3290			  (u64)(unsigned long)ipath->fspath->val;
3291		ipath->fspath->val[i] = rel_ptr;
3292	}
3293
3294	btrfs_free_path(path);
3295	path = NULL;
3296	ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3297			   ipath->fspath, size);
3298	if (ret) {
3299		ret = -EFAULT;
3300		goto out;
3301	}
3302
3303out:
3304	btrfs_free_path(path);
3305	kfree(ipa);
3306
3307	return ret;
3308}
3309
3310static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3311					void __user *arg, int version)
3312{
3313	int ret = 0;
3314	int size;
3315	struct btrfs_ioctl_logical_ino_args *loi;
3316	struct btrfs_data_container *inodes = NULL;
3317	bool ignore_offset;
3318
3319	if (!capable(CAP_SYS_ADMIN))
3320		return -EPERM;
3321
3322	loi = memdup_user(arg, sizeof(*loi));
3323	if (IS_ERR(loi))
3324		return PTR_ERR(loi);
3325
3326	if (version == 1) {
3327		ignore_offset = false;
3328		size = min_t(u32, loi->size, SZ_64K);
3329	} else {
3330		/* All reserved bits must be 0 for now */
3331		if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3332			ret = -EINVAL;
3333			goto out_loi;
3334		}
3335		/* Only accept flags we have defined so far */
3336		if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3337			ret = -EINVAL;
3338			goto out_loi;
3339		}
3340		ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3341		size = min_t(u32, loi->size, SZ_16M);
3342	}
3343
3344	inodes = init_data_container(size);
3345	if (IS_ERR(inodes)) {
3346		ret = PTR_ERR(inodes);
3347		goto out_loi;
3348	}
3349
3350	ret = iterate_inodes_from_logical(loi->logical, fs_info, inodes, ignore_offset);
3351	if (ret == -EINVAL)
3352		ret = -ENOENT;
3353	if (ret < 0)
3354		goto out;
3355
3356	ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3357			   size);
3358	if (ret)
3359		ret = -EFAULT;
3360
3361out:
3362	kvfree(inodes);
3363out_loi:
3364	kfree(loi);
3365
3366	return ret;
3367}
3368
3369void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3370			       struct btrfs_ioctl_balance_args *bargs)
3371{
3372	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3373
3374	bargs->flags = bctl->flags;
3375
3376	if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3377		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3378	if (atomic_read(&fs_info->balance_pause_req))
3379		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3380	if (atomic_read(&fs_info->balance_cancel_req))
3381		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3382
3383	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3384	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3385	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3386
3387	spin_lock(&fs_info->balance_lock);
3388	memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3389	spin_unlock(&fs_info->balance_lock);
3390}
3391
3392/*
3393 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3394 * required.
3395 *
3396 * @fs_info:       the filesystem
3397 * @excl_acquired: ptr to boolean value which is set to false in case balance
3398 *                 is being resumed
3399 *
3400 * Return 0 on success in which case both fs_info::balance is acquired as well
3401 * as exclusive ops are blocked. In case of failure return an error code.
3402 */
3403static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3404{
3405	int ret;
3406
3407	/*
3408	 * Exclusive operation is locked. Three possibilities:
3409	 *   (1) some other op is running
3410	 *   (2) balance is running
3411	 *   (3) balance is paused -- special case (think resume)
3412	 */
3413	while (1) {
3414		if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3415			*excl_acquired = true;
3416			mutex_lock(&fs_info->balance_mutex);
3417			return 0;
3418		}
3419
3420		mutex_lock(&fs_info->balance_mutex);
3421		if (fs_info->balance_ctl) {
3422			/* This is either (2) or (3) */
3423			if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3424				/* This is (2) */
3425				ret = -EINPROGRESS;
3426				goto out_failure;
3427
3428			} else {
3429				mutex_unlock(&fs_info->balance_mutex);
3430				/*
3431				 * Lock released to allow other waiters to
3432				 * continue, we'll reexamine the status again.
3433				 */
3434				mutex_lock(&fs_info->balance_mutex);
3435
3436				if (fs_info->balance_ctl &&
3437				    !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3438					/* This is (3) */
3439					*excl_acquired = false;
3440					return 0;
3441				}
3442			}
3443		} else {
3444			/* This is (1) */
3445			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3446			goto out_failure;
3447		}
3448
3449		mutex_unlock(&fs_info->balance_mutex);
3450	}
3451
3452out_failure:
3453	mutex_unlock(&fs_info->balance_mutex);
3454	*excl_acquired = false;
3455	return ret;
3456}
3457
3458static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3459{
3460	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3461	struct btrfs_fs_info *fs_info = root->fs_info;
3462	struct btrfs_ioctl_balance_args *bargs;
3463	struct btrfs_balance_control *bctl;
3464	bool need_unlock = true;
3465	int ret;
3466
3467	if (!capable(CAP_SYS_ADMIN))
3468		return -EPERM;
3469
3470	ret = mnt_want_write_file(file);
3471	if (ret)
3472		return ret;
3473
3474	bargs = memdup_user(arg, sizeof(*bargs));
3475	if (IS_ERR(bargs)) {
3476		ret = PTR_ERR(bargs);
3477		bargs = NULL;
3478		goto out;
3479	}
3480
3481	ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3482	if (ret)
3483		goto out;
3484
3485	lockdep_assert_held(&fs_info->balance_mutex);
3486
3487	if (bargs->flags & BTRFS_BALANCE_RESUME) {
3488		if (!fs_info->balance_ctl) {
3489			ret = -ENOTCONN;
3490			goto out_unlock;
3491		}
3492
3493		bctl = fs_info->balance_ctl;
3494		spin_lock(&fs_info->balance_lock);
3495		bctl->flags |= BTRFS_BALANCE_RESUME;
3496		spin_unlock(&fs_info->balance_lock);
3497		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3498
3499		goto do_balance;
3500	}
3501
3502	if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3503		ret = -EINVAL;
3504		goto out_unlock;
3505	}
3506
3507	if (fs_info->balance_ctl) {
3508		ret = -EINPROGRESS;
3509		goto out_unlock;
3510	}
3511
3512	bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3513	if (!bctl) {
3514		ret = -ENOMEM;
3515		goto out_unlock;
3516	}
3517
3518	memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3519	memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3520	memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3521
3522	bctl->flags = bargs->flags;
3523do_balance:
3524	/*
3525	 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3526	 * bctl is freed in reset_balance_state, or, if restriper was paused
3527	 * all the way until unmount, in free_fs_info.  The flag should be
3528	 * cleared after reset_balance_state.
3529	 */
3530	need_unlock = false;
3531
3532	ret = btrfs_balance(fs_info, bctl, bargs);
3533	bctl = NULL;
3534
3535	if (ret == 0 || ret == -ECANCELED) {
3536		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3537			ret = -EFAULT;
3538	}
3539
3540	kfree(bctl);
3541out_unlock:
3542	mutex_unlock(&fs_info->balance_mutex);
3543	if (need_unlock)
3544		btrfs_exclop_finish(fs_info);
3545out:
3546	mnt_drop_write_file(file);
3547	kfree(bargs);
3548	return ret;
3549}
3550
3551static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3552{
3553	if (!capable(CAP_SYS_ADMIN))
3554		return -EPERM;
3555
3556	switch (cmd) {
3557	case BTRFS_BALANCE_CTL_PAUSE:
3558		return btrfs_pause_balance(fs_info);
3559	case BTRFS_BALANCE_CTL_CANCEL:
3560		return btrfs_cancel_balance(fs_info);
3561	}
3562
3563	return -EINVAL;
3564}
3565
3566static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3567					 void __user *arg)
3568{
3569	struct btrfs_ioctl_balance_args AUTO_KFREE(bargs);
3570	int ret = 0;
3571
3572	if (!capable(CAP_SYS_ADMIN))
3573		return -EPERM;
3574
3575	mutex_lock(&fs_info->balance_mutex);
3576	if (!fs_info->balance_ctl) {
3577		ret = -ENOTCONN;
3578		goto out;
3579	}
3580
3581	bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3582	if (!bargs) {
3583		ret = -ENOMEM;
3584		goto out;
3585	}
3586
3587	btrfs_update_ioctl_balance_args(fs_info, bargs);
3588
3589	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3590		ret = -EFAULT;
3591out:
3592	mutex_unlock(&fs_info->balance_mutex);
3593	return ret;
3594}
3595
3596static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3597{
3598	struct inode *inode = file_inode(file);
3599	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3600	struct btrfs_ioctl_quota_ctl_args *sa;
3601	int ret;
3602
3603	if (!capable(CAP_SYS_ADMIN))
3604		return -EPERM;
3605
3606	ret = mnt_want_write_file(file);
3607	if (ret)
3608		return ret;
3609
3610	sa = memdup_user(arg, sizeof(*sa));
3611	if (IS_ERR(sa)) {
3612		ret = PTR_ERR(sa);
3613		goto drop_write;
3614	}
3615
3616	switch (sa->cmd) {
3617	case BTRFS_QUOTA_CTL_ENABLE:
3618	case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA:
3619		down_write(&fs_info->subvol_sem);
3620		ret = btrfs_quota_enable(fs_info, sa);
3621		up_write(&fs_info->subvol_sem);
3622		break;
3623	case BTRFS_QUOTA_CTL_DISABLE:
3624		/*
3625		 * Lock the cleaner mutex to prevent races with concurrent
3626		 * relocation, because relocation may be building backrefs for
3627		 * blocks of the quota root while we are deleting the root. This
3628		 * is like dropping fs roots of deleted snapshots/subvolumes, we
3629		 * need the same protection.
3630		 *
3631		 * This also prevents races between concurrent tasks trying to
3632		 * disable quotas, because we will unlock and relock
3633		 * qgroup_ioctl_lock across BTRFS_FS_QUOTA_ENABLED changes.
3634		 *
3635		 * We take this here because we have the dependency of
3636		 *
3637		 * inode_lock -> subvol_sem
3638		 *
3639		 * because of rename.  With relocation we can prealloc extents,
3640		 * so that makes the dependency chain
3641		 *
3642		 * cleaner_mutex -> inode_lock -> subvol_sem
3643		 *
3644		 * so we must take the cleaner_mutex here before we take the
3645		 * subvol_sem.  The deadlock can't actually happen, but this
3646		 * quiets lockdep.
3647		 */
3648		mutex_lock(&fs_info->cleaner_mutex);
3649		down_write(&fs_info->subvol_sem);
3650		ret = btrfs_quota_disable(fs_info);
3651		up_write(&fs_info->subvol_sem);
3652		mutex_unlock(&fs_info->cleaner_mutex);
3653		break;
3654	default:
3655		ret = -EINVAL;
3656		break;
3657	}
3658
3659	kfree(sa);
3660drop_write:
3661	mnt_drop_write_file(file);
3662	return ret;
3663}
3664
3665static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3666{
3667	struct inode *inode = file_inode(file);
3668	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3669	struct btrfs_root *root = BTRFS_I(inode)->root;
3670	struct btrfs_ioctl_qgroup_assign_args *sa;
3671	struct btrfs_qgroup_list *prealloc = NULL;
3672	struct btrfs_trans_handle *trans;
3673	int ret;
3674	int err;
3675
3676	if (!capable(CAP_SYS_ADMIN))
3677		return -EPERM;
3678
3679	if (!btrfs_qgroup_enabled(fs_info))
3680		return -ENOTCONN;
3681
3682	ret = mnt_want_write_file(file);
3683	if (ret)
3684		return ret;
3685
3686	sa = memdup_user(arg, sizeof(*sa));
3687	if (IS_ERR(sa)) {
3688		ret = PTR_ERR(sa);
3689		goto drop_write;
3690	}
3691
3692	if (sa->assign) {
3693		prealloc = kzalloc(sizeof(*prealloc), GFP_KERNEL);
3694		if (!prealloc) {
3695			ret = -ENOMEM;
3696			goto out;
3697		}
3698	}
3699
3700	/* 2 BTRFS_QGROUP_RELATION_KEY items. */
3701	trans = btrfs_start_transaction(root, 2);
3702	if (IS_ERR(trans)) {
3703		ret = PTR_ERR(trans);
3704		goto out;
3705	}
3706
3707	/*
3708	 * Prealloc ownership is moved to the relation handler, there it's used
3709	 * or freed on error.
3710	 */
3711	if (sa->assign) {
3712		ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst, prealloc);
3713		prealloc = NULL;
3714	} else {
3715		ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3716	}
3717
3718	/* update qgroup status and info */
3719	mutex_lock(&fs_info->qgroup_ioctl_lock);
3720	err = btrfs_run_qgroups(trans);
3721	mutex_unlock(&fs_info->qgroup_ioctl_lock);
3722	if (err < 0)
3723		btrfs_warn(fs_info,
3724			   "qgroup status update failed after %s relation, marked as inconsistent",
3725			   sa->assign ? "adding" : "deleting");
3726	err = btrfs_end_transaction(trans);
3727	if (err && !ret)
3728		ret = err;
3729
3730out:
3731	kfree(prealloc);
3732	kfree(sa);
3733drop_write:
3734	mnt_drop_write_file(file);
3735	return ret;
3736}
3737
3738static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3739{
3740	struct inode *inode = file_inode(file);
3741	struct btrfs_root *root = BTRFS_I(inode)->root;
3742	struct btrfs_ioctl_qgroup_create_args *sa;
3743	struct btrfs_trans_handle *trans;
3744	int ret;
3745	int err;
3746
3747	if (!capable(CAP_SYS_ADMIN))
3748		return -EPERM;
3749
3750	if (!btrfs_qgroup_enabled(root->fs_info))
3751		return -ENOTCONN;
3752
3753	ret = mnt_want_write_file(file);
3754	if (ret)
3755		return ret;
3756
3757	sa = memdup_user(arg, sizeof(*sa));
3758	if (IS_ERR(sa)) {
3759		ret = PTR_ERR(sa);
3760		goto drop_write;
3761	}
3762
3763	if (!sa->qgroupid) {
3764		ret = -EINVAL;
3765		goto out;
3766	}
3767
3768	if (sa->create && btrfs_is_fstree(sa->qgroupid)) {
3769		ret = -EINVAL;
3770		goto out;
3771	}
3772
3773	/*
3774	 * 1 BTRFS_QGROUP_INFO_KEY item.
3775	 * 1 BTRFS_QGROUP_LIMIT_KEY item.
3776	 */
3777	trans = btrfs_start_transaction(root, 2);
3778	if (IS_ERR(trans)) {
3779		ret = PTR_ERR(trans);
3780		goto out;
3781	}
3782
3783	if (sa->create) {
3784		ret = btrfs_create_qgroup(trans, sa->qgroupid);
3785	} else {
3786		ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3787	}
3788
3789	err = btrfs_end_transaction(trans);
3790	if (err && !ret)
3791		ret = err;
3792
3793out:
3794	kfree(sa);
3795drop_write:
3796	mnt_drop_write_file(file);
3797	return ret;
3798}
3799
3800static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3801{
3802	struct inode *inode = file_inode(file);
3803	struct btrfs_root *root = BTRFS_I(inode)->root;
3804	struct btrfs_ioctl_qgroup_limit_args *sa;
3805	struct btrfs_trans_handle *trans;
3806	int ret;
3807	int err;
3808	u64 qgroupid;
3809
3810	if (!capable(CAP_SYS_ADMIN))
3811		return -EPERM;
3812
3813	if (!btrfs_qgroup_enabled(root->fs_info))
3814		return -ENOTCONN;
3815
3816	ret = mnt_want_write_file(file);
3817	if (ret)
3818		return ret;
3819
3820	sa = memdup_user(arg, sizeof(*sa));
3821	if (IS_ERR(sa)) {
3822		ret = PTR_ERR(sa);
3823		goto drop_write;
3824	}
3825
3826	/* 1 BTRFS_QGROUP_LIMIT_KEY item. */
3827	trans = btrfs_start_transaction(root, 1);
3828	if (IS_ERR(trans)) {
3829		ret = PTR_ERR(trans);
3830		goto out;
3831	}
3832
3833	qgroupid = sa->qgroupid;
3834	if (!qgroupid) {
3835		/* take the current subvol as qgroup */
3836		qgroupid = btrfs_root_id(root);
3837	}
3838
3839	ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3840
3841	err = btrfs_end_transaction(trans);
3842	if (err && !ret)
3843		ret = err;
3844
3845out:
3846	kfree(sa);
3847drop_write:
3848	mnt_drop_write_file(file);
3849	return ret;
3850}
3851
3852static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3853{
3854	struct inode *inode = file_inode(file);
3855	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3856	struct btrfs_ioctl_quota_rescan_args *qsa;
3857	int ret;
3858
3859	if (!capable(CAP_SYS_ADMIN))
3860		return -EPERM;
3861
3862	if (!btrfs_qgroup_enabled(fs_info))
3863		return -ENOTCONN;
3864
3865	ret = mnt_want_write_file(file);
3866	if (ret)
3867		return ret;
3868
3869	qsa = memdup_user(arg, sizeof(*qsa));
3870	if (IS_ERR(qsa)) {
3871		ret = PTR_ERR(qsa);
3872		goto drop_write;
3873	}
3874
3875	if (qsa->flags) {
3876		ret = -EINVAL;
3877		goto out;
3878	}
3879
3880	ret = btrfs_qgroup_rescan(fs_info);
3881
3882out:
3883	kfree(qsa);
3884drop_write:
3885	mnt_drop_write_file(file);
3886	return ret;
3887}
3888
3889static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
3890						void __user *arg)
3891{
3892	struct btrfs_ioctl_quota_rescan_args qsa = {0};
3893
3894	if (!capable(CAP_SYS_ADMIN))
3895		return -EPERM;
3896
3897	if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3898		qsa.flags = 1;
3899		qsa.progress = fs_info->qgroup_rescan_progress.objectid;
3900	}
3901
3902	if (copy_to_user(arg, &qsa, sizeof(qsa)))
3903		return -EFAULT;
3904
3905	return 0;
3906}
3907
3908static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info)
3909{
3910	if (!capable(CAP_SYS_ADMIN))
3911		return -EPERM;
3912
3913	return btrfs_qgroup_wait_for_completion(fs_info, true);
3914}
3915
3916static long _btrfs_ioctl_set_received_subvol(struct file *file,
3917					    struct mnt_idmap *idmap,
3918					    struct btrfs_ioctl_received_subvol_args *sa)
3919{
3920	struct inode *inode = file_inode(file);
3921	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3922	struct btrfs_root *root = BTRFS_I(inode)->root;
3923	struct btrfs_root_item *root_item = &root->root_item;
3924	struct btrfs_trans_handle *trans;
3925	struct timespec64 ct = current_time(inode);
3926	int ret = 0;
3927	int received_uuid_changed;
3928
3929	if (!inode_owner_or_capable(idmap, inode))
3930		return -EPERM;
3931
3932	ret = mnt_want_write_file(file);
3933	if (ret < 0)
3934		return ret;
3935
3936	down_write(&fs_info->subvol_sem);
3937
3938	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3939		ret = -EINVAL;
3940		goto out;
3941	}
3942
3943	if (btrfs_root_readonly(root)) {
3944		ret = -EROFS;
3945		goto out;
3946	}
3947
3948	received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
3949				       BTRFS_UUID_SIZE);
3950
3951	/*
3952	 * Before we attempt to add the new received uuid, check if we have room
3953	 * for it in case there's already an item. If the size of the existing
3954	 * item plus this root's ID (u64) exceeds the maximum item size, we can
3955	 * return here without the need to abort a transaction. If we don't do
3956	 * this check, the btrfs_uuid_tree_add() call below would fail with
3957	 * -EOVERFLOW and result in a transaction abort. Malicious users could
3958	 * exploit this to turn the fs into RO mode.
3959	 */
3960	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
3961		ret = btrfs_uuid_tree_check_overflow(fs_info, sa->uuid,
3962						     BTRFS_UUID_KEY_RECEIVED_SUBVOL);
3963		if (ret < 0)
3964			goto out;
3965	}
3966
3967	/*
3968	 * 1 - root item
3969	 * 2 - uuid items (received uuid + subvol uuid)
3970	 */
3971	trans = btrfs_start_transaction(root, 3);
3972	if (IS_ERR(trans)) {
3973		ret = PTR_ERR(trans);
3974		trans = NULL;
3975		goto out;
3976	}
3977
3978	sa->rtransid = trans->transid;
3979	sa->rtime.sec = ct.tv_sec;
3980	sa->rtime.nsec = ct.tv_nsec;
3981
3982	if (received_uuid_changed &&
3983	    !btrfs_is_empty_uuid(root_item->received_uuid)) {
3984		ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
3985					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
3986					  btrfs_root_id(root));
3987		if (unlikely(ret && ret != -ENOENT)) {
3988		        btrfs_abort_transaction(trans, ret);
3989		        btrfs_end_transaction(trans);
3990		        goto out;
3991		}
3992	}
3993	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3994	btrfs_set_root_stransid(root_item, sa->stransid);
3995	btrfs_set_root_rtransid(root_item, sa->rtransid);
3996	btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
3997	btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
3998	btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
3999	btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4000
4001	ret = btrfs_update_root(trans, fs_info->tree_root,
4002				&root->root_key, &root->root_item);
4003	if (unlikely(ret < 0)) {
4004		btrfs_abort_transaction(trans, ret);
4005		btrfs_end_transaction(trans);
4006		goto out;
4007	}
4008	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4009		ret = btrfs_uuid_tree_add(trans, sa->uuid,
4010					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4011					  btrfs_root_id(root));
4012		if (unlikely(ret < 0 && ret != -EEXIST)) {
4013			btrfs_abort_transaction(trans, ret);
4014			btrfs_end_transaction(trans);
4015			goto out;
4016		}
4017	}
4018	ret = btrfs_commit_transaction(trans);
4019out:
4020	up_write(&fs_info->subvol_sem);
4021	mnt_drop_write_file(file);
4022	return ret;
4023}
4024
4025#ifdef CONFIG_64BIT
4026static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4027						void __user *arg)
4028{
4029	struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4030	struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4031	int ret = 0;
4032
4033	args32 = memdup_user(arg, sizeof(*args32));
4034	if (IS_ERR(args32))
4035		return PTR_ERR(args32);
4036
4037	args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4038	if (!args64) {
4039		ret = -ENOMEM;
4040		goto out;
4041	}
4042
4043	memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4044	args64->stransid = args32->stransid;
4045	args64->rtransid = args32->rtransid;
4046	args64->stime.sec = args32->stime.sec;
4047	args64->stime.nsec = args32->stime.nsec;
4048	args64->rtime.sec = args32->rtime.sec;
4049	args64->rtime.nsec = args32->rtime.nsec;
4050	args64->flags = args32->flags;
4051
4052	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4053	if (ret)
4054		goto out;
4055
4056	memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4057	args32->stransid = args64->stransid;
4058	args32->rtransid = args64->rtransid;
4059	args32->stime.sec = args64->stime.sec;
4060	args32->stime.nsec = args64->stime.nsec;
4061	args32->rtime.sec = args64->rtime.sec;
4062	args32->rtime.nsec = args64->rtime.nsec;
4063	args32->flags = args64->flags;
4064
4065	ret = copy_to_user(arg, args32, sizeof(*args32));
4066	if (ret)
4067		ret = -EFAULT;
4068
4069out:
4070	kfree(args32);
4071	kfree(args64);
4072	return ret;
4073}
4074#endif
4075
4076static long btrfs_ioctl_set_received_subvol(struct file *file,
4077					    void __user *arg)
4078{
4079	struct btrfs_ioctl_received_subvol_args *sa = NULL;
4080	int ret = 0;
4081
4082	sa = memdup_user(arg, sizeof(*sa));
4083	if (IS_ERR(sa))
4084		return PTR_ERR(sa);
4085
4086	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4087
4088	if (ret)
4089		goto out;
4090
4091	ret = copy_to_user(arg, sa, sizeof(*sa));
4092	if (ret)
4093		ret = -EFAULT;
4094
4095out:
4096	kfree(sa);
4097	return ret;
4098}
4099
4100static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4101					void __user *arg)
4102{
4103	size_t len;
4104	int ret;
4105	char label[BTRFS_LABEL_SIZE];
4106
4107	spin_lock(&fs_info->super_lock);
4108	memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4109	spin_unlock(&fs_info->super_lock);
4110
4111	len = strnlen(label, BTRFS_LABEL_SIZE);
4112
4113	if (len == BTRFS_LABEL_SIZE) {
4114		btrfs_warn(fs_info,
4115			   "label is too long, return the first %zu bytes",
4116			   --len);
4117	}
4118
4119	ret = copy_to_user(arg, label, len);
4120
4121	return ret ? -EFAULT : 0;
4122}
4123
4124static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4125{
4126	struct inode *inode = file_inode(file);
4127	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4128	struct btrfs_root *root = BTRFS_I(inode)->root;
4129	struct btrfs_super_block *super_block = fs_info->super_copy;
4130	struct btrfs_trans_handle *trans;
4131	char label[BTRFS_LABEL_SIZE];
4132	int ret;
4133
4134	if (!capable(CAP_SYS_ADMIN))
4135		return -EPERM;
4136
4137	if (copy_from_user(label, arg, sizeof(label)))
4138		return -EFAULT;
4139
4140	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4141		btrfs_err(fs_info,
4142			  "unable to set label with more than %d bytes",
4143			  BTRFS_LABEL_SIZE - 1);
4144		return -EINVAL;
4145	}
4146
4147	ret = mnt_want_write_file(file);
4148	if (ret)
4149		return ret;
4150
4151	trans = btrfs_start_transaction(root, 0);
4152	if (IS_ERR(trans)) {
4153		ret = PTR_ERR(trans);
4154		goto out_unlock;
4155	}
4156
4157	spin_lock(&fs_info->super_lock);
4158	strscpy(super_block->label, label);
4159	spin_unlock(&fs_info->super_lock);
4160	ret = btrfs_commit_transaction(trans);
4161
4162out_unlock:
4163	mnt_drop_write_file(file);
4164	return ret;
4165}
4166
4167#define INIT_FEATURE_FLAGS(suffix) \
4168	{ .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4169	  .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4170	  .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4171
4172int btrfs_ioctl_get_supported_features(void __user *arg)
4173{
4174	static const struct btrfs_ioctl_feature_flags features[3] = {
4175		INIT_FEATURE_FLAGS(SUPP),
4176		INIT_FEATURE_FLAGS(SAFE_SET),
4177		INIT_FEATURE_FLAGS(SAFE_CLEAR)
4178	};
4179
4180	if (copy_to_user(arg, &features, sizeof(features)))
4181		return -EFAULT;
4182
4183	return 0;
4184}
4185
4186static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4187					void __user *arg)
4188{
4189	struct btrfs_super_block *super_block = fs_info->super_copy;
4190	struct btrfs_ioctl_feature_flags features;
4191
4192	features.compat_flags = btrfs_super_compat_flags(super_block);
4193	features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4194	features.incompat_flags = btrfs_super_incompat_flags(super_block);
4195
4196	if (copy_to_user(arg, &features, sizeof(features)))
4197		return -EFAULT;
4198
4199	return 0;
4200}
4201
4202static int check_feature_bits(const struct btrfs_fs_info *fs_info,
4203			      enum btrfs_feature_set set,
4204			      u64 change_mask, u64 flags, u64 supported_flags,
4205			      u64 safe_set, u64 safe_clear)
4206{
4207	const char *type = btrfs_feature_set_name(set);
4208	const char AUTO_KFREE(names);
4209	u64 disallowed, unsupported;
4210	u64 set_mask = flags & change_mask;
4211	u64 clear_mask = ~flags & change_mask;
4212
4213	unsupported = set_mask & ~supported_flags;
4214	if (unsupported) {
4215		names = btrfs_printable_features(set, unsupported);
4216		if (names)
4217			btrfs_warn(fs_info,
4218				   "this kernel does not support the %s feature bit%s",
4219				   names, strchr(names, ',') ? "s" : "");
4220		else
4221			btrfs_warn(fs_info,
4222				   "this kernel does not support %s bits 0x%llx",
4223				   type, unsupported);
4224		return -EOPNOTSUPP;
4225	}
4226
4227	disallowed = set_mask & ~safe_set;
4228	if (disallowed) {
4229		names = btrfs_printable_features(set, disallowed);
4230		if (names)
4231			btrfs_warn(fs_info,
4232				   "can't set the %s feature bit%s while mounted",
4233				   names, strchr(names, ',') ? "s" : "");
4234		else
4235			btrfs_warn(fs_info,
4236				   "can't set %s bits 0x%llx while mounted",
4237				   type, disallowed);
4238		return -EPERM;
4239	}
4240
4241	disallowed = clear_mask & ~safe_clear;
4242	if (disallowed) {
4243		names = btrfs_printable_features(set, disallowed);
4244		if (names)
4245			btrfs_warn(fs_info,
4246				   "can't clear the %s feature bit%s while mounted",
4247				   names, strchr(names, ',') ? "s" : "");
4248		else
4249			btrfs_warn(fs_info,
4250				   "can't clear %s bits 0x%llx while mounted",
4251				   type, disallowed);
4252		return -EPERM;
4253	}
4254
4255	return 0;
4256}
4257
4258#define check_feature(fs_info, change_mask, flags, mask_base)	\
4259check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags,	\
4260		   BTRFS_FEATURE_ ## mask_base ## _SUPP,	\
4261		   BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,	\
4262		   BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4263
4264static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4265{
4266	struct inode *inode = file_inode(file);
4267	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4268	struct btrfs_root *root = BTRFS_I(inode)->root;
4269	struct btrfs_super_block *super_block = fs_info->super_copy;
4270	struct btrfs_ioctl_feature_flags flags[2];
4271	struct btrfs_trans_handle *trans;
4272	u64 newflags;
4273	int ret;
4274
4275	if (!capable(CAP_SYS_ADMIN))
4276		return -EPERM;
4277
4278	if (copy_from_user(flags, arg, sizeof(flags)))
4279		return -EFAULT;
4280
4281	/* Nothing to do */
4282	if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4283	    !flags[0].incompat_flags)
4284		return 0;
4285
4286	ret = check_feature(fs_info, flags[0].compat_flags,
4287			    flags[1].compat_flags, COMPAT);
4288	if (ret)
4289		return ret;
4290
4291	ret = check_feature(fs_info, flags[0].compat_ro_flags,
4292			    flags[1].compat_ro_flags, COMPAT_RO);
4293	if (ret)
4294		return ret;
4295
4296	ret = check_feature(fs_info, flags[0].incompat_flags,
4297			    flags[1].incompat_flags, INCOMPAT);
4298	if (ret)
4299		return ret;
4300
4301	ret = mnt_want_write_file(file);
4302	if (ret)
4303		return ret;
4304
4305	trans = btrfs_start_transaction(root, 0);
4306	if (IS_ERR(trans)) {
4307		ret = PTR_ERR(trans);
4308		goto out_drop_write;
4309	}
4310
4311	spin_lock(&fs_info->super_lock);
4312	newflags = btrfs_super_compat_flags(super_block);
4313	newflags |= flags[0].compat_flags & flags[1].compat_flags;
4314	newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4315	btrfs_set_super_compat_flags(super_block, newflags);
4316
4317	newflags = btrfs_super_compat_ro_flags(super_block);
4318	newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4319	newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4320	btrfs_set_super_compat_ro_flags(super_block, newflags);
4321
4322	newflags = btrfs_super_incompat_flags(super_block);
4323	newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4324	newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4325	btrfs_set_super_incompat_flags(super_block, newflags);
4326	spin_unlock(&fs_info->super_lock);
4327
4328	ret = btrfs_commit_transaction(trans);
4329out_drop_write:
4330	mnt_drop_write_file(file);
4331
4332	return ret;
4333}
4334
4335static int _btrfs_ioctl_send(struct btrfs_root *root, void __user *argp, bool compat)
4336{
4337	struct btrfs_ioctl_send_args *arg;
4338	int ret;
4339
4340	if (compat) {
4341#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4342		struct btrfs_ioctl_send_args_32 args32 = { 0 };
4343
4344		ret = copy_from_user(&args32, argp, sizeof(args32));
4345		if (ret)
4346			return -EFAULT;
4347		arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4348		if (!arg)
4349			return -ENOMEM;
4350		arg->send_fd = args32.send_fd;
4351		arg->clone_sources_count = args32.clone_sources_count;
4352		arg->clone_sources = compat_ptr(args32.clone_sources);
4353		arg->parent_root = args32.parent_root;
4354		arg->flags = args32.flags;
4355		arg->version = args32.version;
4356		memcpy(arg->reserved, args32.reserved,
4357		       sizeof(args32.reserved));
4358#else
4359		return -ENOTTY;
4360#endif
4361	} else {
4362		arg = memdup_user(argp, sizeof(*arg));
4363		if (IS_ERR(arg))
4364			return PTR_ERR(arg);
4365	}
4366	ret = btrfs_ioctl_send(root, arg);
4367	kfree(arg);
4368	return ret;
4369}
4370
4371static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4372				    bool compat)
4373{
4374	struct btrfs_ioctl_encoded_io_args args = { 0 };
4375	size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4376					     flags);
4377	size_t copy_end;
4378	struct btrfs_inode *inode = BTRFS_I(file_inode(file));
4379	struct btrfs_fs_info *fs_info = inode->root->fs_info;
4380	struct extent_io_tree *io_tree = &inode->io_tree;
4381	struct iovec iovstack[UIO_FASTIOV];
4382	struct iovec *iov = iovstack;
4383	struct iov_iter iter;
4384	loff_t pos;
4385	struct kiocb kiocb;
4386	ssize_t ret;
4387	u64 disk_bytenr, disk_io_size;
4388	struct extent_state *cached_state = NULL;
4389
4390	if (!capable(CAP_SYS_ADMIN)) {
4391		ret = -EPERM;
4392		goto out_acct;
4393	}
4394
4395	if (compat) {
4396#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4397		struct btrfs_ioctl_encoded_io_args_32 args32;
4398
4399		copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4400				       flags);
4401		if (copy_from_user(&args32, argp, copy_end)) {
4402			ret = -EFAULT;
4403			goto out_acct;
4404		}
4405		args.iov = compat_ptr(args32.iov);
4406		args.iovcnt = args32.iovcnt;
4407		args.offset = args32.offset;
4408		args.flags = args32.flags;
4409#else
4410		return -ENOTTY;
4411#endif
4412	} else {
4413		copy_end = copy_end_kernel;
4414		if (copy_from_user(&args, argp, copy_end)) {
4415			ret = -EFAULT;
4416			goto out_acct;
4417		}
4418	}
4419	if (args.flags != 0) {
4420		ret = -EINVAL;
4421		goto out_acct;
4422	}
4423
4424	ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4425			   &iov, &iter);
4426	if (ret < 0)
4427		goto out_acct;
4428
4429	if (iov_iter_count(&iter) == 0) {
4430		ret = 0;
4431		goto out_iov;
4432	}
4433	pos = args.offset;
4434	ret = rw_verify_area(READ, file, &pos, args.len);
4435	if (ret < 0)
4436		goto out_iov;
4437
4438	init_sync_kiocb(&kiocb, file);
4439	kiocb.ki_pos = pos;
4440
4441	ret = btrfs_encoded_read(&kiocb, &iter, &args, &cached_state,
4442				 &disk_bytenr, &disk_io_size);
4443
4444	if (ret == -EIOCBQUEUED) {
4445		bool unlocked = false;
4446		u64 start, lockend, count;
4447
4448		start = ALIGN_DOWN(kiocb.ki_pos, fs_info->sectorsize);
4449		lockend = start + BTRFS_MAX_UNCOMPRESSED - 1;
4450
4451		if (args.compression)
4452			count = disk_io_size;
4453		else
4454			count = args.len;
4455
4456		ret = btrfs_encoded_read_regular(&kiocb, &iter, start, lockend,
4457						 &cached_state, disk_bytenr,
4458						 disk_io_size, count,
4459						 args.compression, &unlocked);
4460
4461		if (!unlocked) {
4462			btrfs_unlock_extent(io_tree, start, lockend, &cached_state);
4463			btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED);
4464		}
4465	}
4466
4467	if (ret >= 0) {
4468		fsnotify_access(file);
4469		if (copy_to_user(argp + copy_end,
4470				 (char *)&args + copy_end_kernel,
4471				 sizeof(args) - copy_end_kernel))
4472			ret = -EFAULT;
4473	}
4474
4475out_iov:
4476	kfree(iov);
4477out_acct:
4478	if (ret > 0)
4479		add_rchar(current, ret);
4480	inc_syscr(current);
4481	return ret;
4482}
4483
4484static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4485{
4486	struct btrfs_ioctl_encoded_io_args args;
4487	struct iovec iovstack[UIO_FASTIOV];
4488	struct iovec *iov = iovstack;
4489	struct iov_iter iter;
4490	loff_t pos;
4491	struct kiocb kiocb;
4492	ssize_t ret;
4493
4494	if (!capable(CAP_SYS_ADMIN)) {
4495		ret = -EPERM;
4496		goto out_acct;
4497	}
4498
4499	if (!(file->f_mode & FMODE_WRITE)) {
4500		ret = -EBADF;
4501		goto out_acct;
4502	}
4503
4504	if (compat) {
4505#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4506		struct btrfs_ioctl_encoded_io_args_32 args32;
4507
4508		if (copy_from_user(&args32, argp, sizeof(args32))) {
4509			ret = -EFAULT;
4510			goto out_acct;
4511		}
4512		args.iov = compat_ptr(args32.iov);
4513		args.iovcnt = args32.iovcnt;
4514		args.offset = args32.offset;
4515		args.flags = args32.flags;
4516		args.len = args32.len;
4517		args.unencoded_len = args32.unencoded_len;
4518		args.unencoded_offset = args32.unencoded_offset;
4519		args.compression = args32.compression;
4520		args.encryption = args32.encryption;
4521		memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4522#else
4523		return -ENOTTY;
4524#endif
4525	} else {
4526		if (copy_from_user(&args, argp, sizeof(args))) {
4527			ret = -EFAULT;
4528			goto out_acct;
4529		}
4530	}
4531
4532	ret = -EINVAL;
4533	if (args.flags != 0)
4534		goto out_acct;
4535	if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4536		goto out_acct;
4537	if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4538	    args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4539		goto out_acct;
4540	if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4541	    args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4542		goto out_acct;
4543	if (args.unencoded_offset > args.unencoded_len)
4544		goto out_acct;
4545	if (args.len > args.unencoded_len - args.unencoded_offset)
4546		goto out_acct;
4547
4548	ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4549			   &iov, &iter);
4550	if (ret < 0)
4551		goto out_acct;
4552
4553	if (iov_iter_count(&iter) == 0) {
4554		ret = 0;
4555		goto out_iov;
4556	}
4557	pos = args.offset;
4558	ret = rw_verify_area(WRITE, file, &pos, args.len);
4559	if (ret < 0)
4560		goto out_iov;
4561
4562	init_sync_kiocb(&kiocb, file);
4563	ret = kiocb_set_rw_flags(&kiocb, 0, WRITE);
4564	if (ret)
4565		goto out_iov;
4566	kiocb.ki_pos = pos;
4567
4568	file_start_write(file);
4569
4570	ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4571	if (ret > 0)
4572		fsnotify_modify(file);
4573
4574	file_end_write(file);
4575out_iov:
4576	kfree(iov);
4577out_acct:
4578	if (ret > 0)
4579		add_wchar(current, ret);
4580	inc_syscw(current);
4581	return ret;
4582}
4583
4584struct btrfs_uring_encoded_data {
4585	struct btrfs_ioctl_encoded_io_args args;
4586	struct iovec iovstack[UIO_FASTIOV];
4587	struct iovec *iov;
4588	struct iov_iter iter;
4589};
4590
4591/*
4592 * Context that's attached to an encoded read io_uring command, in cmd->pdu. It
4593 * contains the fields in btrfs_uring_read_extent that are necessary to finish
4594 * off and cleanup the I/O in btrfs_uring_read_finished.
4595 */
4596struct btrfs_uring_priv {
4597	struct io_uring_cmd *cmd;
4598	struct page **pages;
4599	unsigned long nr_pages;
4600	struct kiocb iocb;
4601	struct iovec *iov;
4602	struct iov_iter iter;
4603	struct extent_state *cached_state;
4604	u64 count;
4605	u64 start;
4606	u64 lockend;
4607	int err;
4608	bool compressed;
4609};
4610
4611struct io_btrfs_cmd {
4612	struct btrfs_uring_encoded_data *data;
4613	struct btrfs_uring_priv *priv;
4614};
4615
4616static void btrfs_uring_read_finished(struct io_tw_req tw_req, io_tw_token_t tw)
4617{
4618	struct io_uring_cmd *cmd = io_uring_cmd_from_tw(tw_req);
4619	struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd);
4620	struct btrfs_uring_priv *priv = bc->priv;
4621	struct btrfs_inode *inode = BTRFS_I(file_inode(priv->iocb.ki_filp));
4622	struct extent_io_tree *io_tree = &inode->io_tree;
4623	pgoff_t index;
4624	u64 cur;
4625	size_t page_offset;
4626	ssize_t ret;
4627
4628	/* The inode lock has already been acquired in btrfs_uring_read_extent.  */
4629	btrfs_lockdep_inode_acquire(inode, i_rwsem);
4630
4631	if (priv->err) {
4632		ret = priv->err;
4633		goto out;
4634	}
4635
4636	if (priv->compressed) {
4637		index = 0;
4638		page_offset = 0;
4639	} else {
4640		index = (priv->iocb.ki_pos - priv->start) >> PAGE_SHIFT;
4641		page_offset = offset_in_page(priv->iocb.ki_pos - priv->start);
4642	}
4643	cur = 0;
4644	while (cur < priv->count) {
4645		size_t bytes = min_t(size_t, priv->count - cur, PAGE_SIZE - page_offset);
4646
4647		if (copy_page_to_iter(priv->pages[index], page_offset, bytes,
4648				      &priv->iter) != bytes) {
4649			ret = -EFAULT;
4650			goto out;
4651		}
4652
4653		index++;
4654		cur += bytes;
4655		page_offset = 0;
4656	}
4657	ret = priv->count;
4658
4659out:
4660	btrfs_unlock_extent(io_tree, priv->start, priv->lockend, &priv->cached_state);
4661	btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED);
4662
4663	io_uring_cmd_done(cmd, ret, IO_URING_CMD_TASK_WORK_ISSUE_FLAGS);
4664	add_rchar(current, ret);
4665
4666	for (index = 0; index < priv->nr_pages; index++)
4667		__free_page(priv->pages[index]);
4668
4669	kfree(priv->pages);
4670	kfree(priv->iov);
4671	kfree(priv);
4672	kfree(bc->data);
4673}
4674
4675void btrfs_uring_read_extent_endio(void *ctx, int err)
4676{
4677	struct btrfs_uring_priv *priv = ctx;
4678	struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(priv->cmd, struct io_btrfs_cmd);
4679
4680	priv->err = err;
4681	bc->priv = priv;
4682
4683	io_uring_cmd_complete_in_task(priv->cmd, btrfs_uring_read_finished);
4684}
4685
4686static int btrfs_uring_read_extent(struct kiocb *iocb, struct iov_iter *iter,
4687				   u64 start, u64 lockend,
4688				   struct extent_state *cached_state,
4689				   u64 disk_bytenr, u64 disk_io_size,
4690				   size_t count, bool compressed,
4691				   struct iovec *iov, struct io_uring_cmd *cmd)
4692{
4693	struct btrfs_inode *inode = BTRFS_I(file_inode(iocb->ki_filp));
4694	struct extent_io_tree *io_tree = &inode->io_tree;
4695	struct page **pages = NULL;
4696	struct btrfs_uring_priv *priv = NULL;
4697	unsigned long nr_pages;
4698	int ret;
4699
4700	nr_pages = DIV_ROUND_UP(disk_io_size, PAGE_SIZE);
4701	pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS);
4702	if (!pages)
4703		return -ENOMEM;
4704	ret = btrfs_alloc_page_array(nr_pages, pages, 0);
4705	if (ret) {
4706		ret = -ENOMEM;
4707		goto out_fail;
4708	}
4709
4710	priv = kmalloc(sizeof(*priv), GFP_NOFS);
4711	if (!priv) {
4712		ret = -ENOMEM;
4713		goto out_fail;
4714	}
4715
4716	priv->iocb = *iocb;
4717	priv->iov = iov;
4718	priv->iter = *iter;
4719	priv->count = count;
4720	priv->cmd = cmd;
4721	priv->cached_state = cached_state;
4722	priv->compressed = compressed;
4723	priv->nr_pages = nr_pages;
4724	priv->pages = pages;
4725	priv->start = start;
4726	priv->lockend = lockend;
4727	priv->err = 0;
4728
4729	ret = btrfs_encoded_read_regular_fill_pages(inode, disk_bytenr,
4730						    disk_io_size, pages, priv);
4731	if (ret && ret != -EIOCBQUEUED)
4732		goto out_fail;
4733
4734	/*
4735	 * If we return -EIOCBQUEUED, we're deferring the cleanup to
4736	 * btrfs_uring_read_finished(), which will handle unlocking the extent
4737	 * and inode and freeing the allocations.
4738	 */
4739
4740	/*
4741	 * We're returning to userspace with the inode lock held, and that's
4742	 * okay - it'll get unlocked in a worker thread.  Call
4743	 * btrfs_lockdep_inode_release() to avoid confusing lockdep.
4744	 */
4745	btrfs_lockdep_inode_release(inode, i_rwsem);
4746
4747	return -EIOCBQUEUED;
4748
4749out_fail:
4750	btrfs_unlock_extent(io_tree, start, lockend, &cached_state);
4751	btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED);
4752	kfree(priv);
4753	for (int i = 0; i < nr_pages; i++) {
4754		if (pages[i])
4755			__free_page(pages[i]);
4756	}
4757	kfree(pages);
4758	return ret;
4759}
4760
4761static int btrfs_uring_encoded_read(struct io_uring_cmd *cmd, unsigned int issue_flags)
4762{
4763	struct file *file = cmd->file;
4764	struct btrfs_inode *inode = BTRFS_I(file->f_inode);
4765	struct extent_io_tree *io_tree = &inode->io_tree;
4766	struct btrfs_fs_info *fs_info = inode->root->fs_info;
4767	size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args, flags);
4768	size_t copy_end;
4769	int ret;
4770	u64 disk_bytenr, disk_io_size;
4771	loff_t pos;
4772	struct kiocb kiocb;
4773	struct extent_state *cached_state = NULL;
4774	u64 start, lockend;
4775	void __user *sqe_addr;
4776	struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd);
4777	struct btrfs_uring_encoded_data *data = NULL;
4778
4779	if (cmd->flags & IORING_URING_CMD_REISSUE)
4780		data = bc->data;
4781
4782	if (!capable(CAP_SYS_ADMIN)) {
4783		ret = -EPERM;
4784		goto out_acct;
4785	}
4786	sqe_addr = u64_to_user_ptr(READ_ONCE(cmd->sqe->addr));
4787
4788	if (issue_flags & IO_URING_F_COMPAT) {
4789#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4790		copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32, flags);
4791#else
4792		ret = -ENOTTY;
4793		goto out_acct;
4794#endif
4795	} else {
4796		copy_end = copy_end_kernel;
4797	}
4798
4799	if (!data) {
4800		data = kzalloc(sizeof(*data), GFP_NOFS);
4801		if (!data) {
4802			ret = -ENOMEM;
4803			goto out_acct;
4804		}
4805
4806		bc->data = data;
4807
4808		if (issue_flags & IO_URING_F_COMPAT) {
4809#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4810			struct btrfs_ioctl_encoded_io_args_32 args32;
4811
4812			if (copy_from_user(&args32, sqe_addr, copy_end)) {
4813				ret = -EFAULT;
4814				goto out_acct;
4815			}
4816
4817			data->args.iov = compat_ptr(args32.iov);
4818			data->args.iovcnt = args32.iovcnt;
4819			data->args.offset = args32.offset;
4820			data->args.flags = args32.flags;
4821#endif
4822		} else {
4823			if (copy_from_user(&data->args, sqe_addr, copy_end)) {
4824				ret = -EFAULT;
4825				goto out_acct;
4826			}
4827		}
4828
4829		if (data->args.flags != 0) {
4830			ret = -EINVAL;
4831			goto out_acct;
4832		}
4833
4834		data->iov = data->iovstack;
4835		ret = import_iovec(ITER_DEST, data->args.iov, data->args.iovcnt,
4836				   ARRAY_SIZE(data->iovstack), &data->iov,
4837				   &data->iter);
4838		if (ret < 0)
4839			goto out_acct;
4840
4841		if (iov_iter_count(&data->iter) == 0) {
4842			ret = 0;
4843			goto out_free;
4844		}
4845	}
4846
4847	pos = data->args.offset;
4848	ret = rw_verify_area(READ, file, &pos, data->args.len);
4849	if (ret < 0)
4850		goto out_free;
4851
4852	init_sync_kiocb(&kiocb, file);
4853	kiocb.ki_pos = pos;
4854
4855	if (issue_flags & IO_URING_F_NONBLOCK)
4856		kiocb.ki_flags |= IOCB_NOWAIT;
4857
4858	start = ALIGN_DOWN(pos, fs_info->sectorsize);
4859	lockend = start + BTRFS_MAX_UNCOMPRESSED - 1;
4860
4861	ret = btrfs_encoded_read(&kiocb, &data->iter, &data->args, &cached_state,
4862				 &disk_bytenr, &disk_io_size);
4863	if (ret == -EAGAIN)
4864		goto out_acct;
4865	if (ret < 0 && ret != -EIOCBQUEUED)
4866		goto out_free;
4867
4868	file_accessed(file);
4869
4870	if (copy_to_user(sqe_addr + copy_end,
4871			 (const char *)&data->args + copy_end_kernel,
4872			 sizeof(data->args) - copy_end_kernel)) {
4873		if (ret == -EIOCBQUEUED) {
4874			btrfs_unlock_extent(io_tree, start, lockend, &cached_state);
4875			btrfs_inode_unlock(inode, BTRFS_ILOCK_SHARED);
4876		}
4877		ret = -EFAULT;
4878		goto out_free;
4879	}
4880
4881	if (ret == -EIOCBQUEUED) {
4882		u64 count = min_t(u64, iov_iter_count(&data->iter), disk_io_size);
4883
4884		/* Match ioctl by not returning past EOF if uncompressed. */
4885		if (!data->args.compression)
4886			count = min_t(u64, count, data->args.len);
4887
4888		ret = btrfs_uring_read_extent(&kiocb, &data->iter, start, lockend,
4889					      cached_state, disk_bytenr, disk_io_size,
4890					      count, data->args.compression,
4891					      data->iov, cmd);
4892
4893		goto out_acct;
4894	}
4895
4896out_free:
4897	kfree(data->iov);
4898
4899out_acct:
4900	if (ret > 0)
4901		add_rchar(current, ret);
4902	inc_syscr(current);
4903
4904	if (ret != -EIOCBQUEUED && ret != -EAGAIN)
4905		kfree(data);
4906
4907	return ret;
4908}
4909
4910static int btrfs_uring_encoded_write(struct io_uring_cmd *cmd, unsigned int issue_flags)
4911{
4912	struct file *file = cmd->file;
4913	loff_t pos;
4914	struct kiocb kiocb;
4915	ssize_t ret;
4916	void __user *sqe_addr;
4917	struct io_btrfs_cmd *bc = io_uring_cmd_to_pdu(cmd, struct io_btrfs_cmd);
4918	struct btrfs_uring_encoded_data *data = NULL;
4919
4920	if (cmd->flags & IORING_URING_CMD_REISSUE)
4921		data = bc->data;
4922
4923	if (!capable(CAP_SYS_ADMIN)) {
4924		ret = -EPERM;
4925		goto out_acct;
4926	}
4927	sqe_addr = u64_to_user_ptr(READ_ONCE(cmd->sqe->addr));
4928
4929	if (!(file->f_mode & FMODE_WRITE)) {
4930		ret = -EBADF;
4931		goto out_acct;
4932	}
4933
4934	if (!data) {
4935		data = kzalloc(sizeof(*data), GFP_NOFS);
4936		if (!data) {
4937			ret = -ENOMEM;
4938			goto out_acct;
4939		}
4940
4941		bc->data = data;
4942
4943		if (issue_flags & IO_URING_F_COMPAT) {
4944#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4945			struct btrfs_ioctl_encoded_io_args_32 args32;
4946
4947			if (copy_from_user(&args32, sqe_addr, sizeof(args32))) {
4948				ret = -EFAULT;
4949				goto out_acct;
4950			}
4951			data->args.iov = compat_ptr(args32.iov);
4952			data->args.iovcnt = args32.iovcnt;
4953			data->args.offset = args32.offset;
4954			data->args.flags = args32.flags;
4955			data->args.len = args32.len;
4956			data->args.unencoded_len = args32.unencoded_len;
4957			data->args.unencoded_offset = args32.unencoded_offset;
4958			data->args.compression = args32.compression;
4959			data->args.encryption = args32.encryption;
4960			memcpy(data->args.reserved, args32.reserved,
4961			       sizeof(data->args.reserved));
4962#else
4963			ret = -ENOTTY;
4964			goto out_acct;
4965#endif
4966		} else {
4967			if (copy_from_user(&data->args, sqe_addr, sizeof(data->args))) {
4968				ret = -EFAULT;
4969				goto out_acct;
4970			}
4971		}
4972
4973		ret = -EINVAL;
4974		if (data->args.flags != 0)
4975			goto out_acct;
4976		if (memchr_inv(data->args.reserved, 0, sizeof(data->args.reserved)))
4977			goto out_acct;
4978		if (data->args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4979		    data->args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4980			goto out_acct;
4981		if (data->args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4982		    data->args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4983			goto out_acct;
4984		if (data->args.unencoded_offset > data->args.unencoded_len)
4985			goto out_acct;
4986		if (data->args.len > data->args.unencoded_len - data->args.unencoded_offset)
4987			goto out_acct;
4988
4989		data->iov = data->iovstack;
4990		ret = import_iovec(ITER_SOURCE, data->args.iov, data->args.iovcnt,
4991				   ARRAY_SIZE(data->iovstack), &data->iov,
4992				   &data->iter);
4993		if (ret < 0)
4994			goto out_acct;
4995
4996		if (iov_iter_count(&data->iter) == 0) {
4997			ret = 0;
4998			goto out_iov;
4999		}
5000	}
5001
5002	if (issue_flags & IO_URING_F_NONBLOCK) {
5003		ret = -EAGAIN;
5004		goto out_acct;
5005	}
5006
5007	pos = data->args.offset;
5008	ret = rw_verify_area(WRITE, file, &pos, data->args.len);
5009	if (ret < 0)
5010		goto out_iov;
5011
5012	init_sync_kiocb(&kiocb, file);
5013	ret = kiocb_set_rw_flags(&kiocb, 0, WRITE);
5014	if (ret)
5015		goto out_iov;
5016	kiocb.ki_pos = pos;
5017
5018	file_start_write(file);
5019
5020	ret = btrfs_do_write_iter(&kiocb, &data->iter, &data->args);
5021	if (ret > 0)
5022		fsnotify_modify(file);
5023
5024	file_end_write(file);
5025out_iov:
5026	kfree(data->iov);
5027out_acct:
5028	if (ret > 0)
5029		add_wchar(current, ret);
5030	inc_syscw(current);
5031
5032	if (ret != -EAGAIN)
5033		kfree(data);
5034	return ret;
5035}
5036
5037int btrfs_uring_cmd(struct io_uring_cmd *cmd, unsigned int issue_flags)
5038{
5039	if (unlikely(btrfs_is_shutdown(inode_to_fs_info(file_inode(cmd->file)))))
5040		return -EIO;
5041
5042	switch (cmd->cmd_op) {
5043	case BTRFS_IOC_ENCODED_READ:
5044#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5045	case BTRFS_IOC_ENCODED_READ_32:
5046#endif
5047		return btrfs_uring_encoded_read(cmd, issue_flags);
5048
5049	case BTRFS_IOC_ENCODED_WRITE:
5050#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5051	case BTRFS_IOC_ENCODED_WRITE_32:
5052#endif
5053		return btrfs_uring_encoded_write(cmd, issue_flags);
5054	}
5055
5056	return -EINVAL;
5057}
5058
5059static int btrfs_ioctl_subvol_sync(struct btrfs_fs_info *fs_info, void __user *argp)
5060{
5061	struct btrfs_root *root;
5062	struct btrfs_ioctl_subvol_wait args = { 0 };
5063	signed long sched_ret;
5064	int refs;
5065	u64 root_flags;
5066	bool wait_for_deletion = false;
5067	bool found = false;
5068
5069	if (copy_from_user(&args, argp, sizeof(args)))
5070		return -EFAULT;
5071
5072	switch (args.mode) {
5073	case BTRFS_SUBVOL_SYNC_WAIT_FOR_QUEUED:
5074		/*
5075		 * Wait for the first one deleted that waits until all previous
5076		 * are cleaned.
5077		 */
5078		spin_lock(&fs_info->trans_lock);
5079		if (!list_empty(&fs_info->dead_roots)) {
5080			root = list_last_entry(&fs_info->dead_roots,
5081					       struct btrfs_root, root_list);
5082			args.subvolid = btrfs_root_id(root);
5083			found = true;
5084		}
5085		spin_unlock(&fs_info->trans_lock);
5086		if (!found)
5087			return -ENOENT;
5088
5089		fallthrough;
5090	case BTRFS_SUBVOL_SYNC_WAIT_FOR_ONE:
5091		if ((0 < args.subvolid && args.subvolid < BTRFS_FIRST_FREE_OBJECTID) ||
5092		    BTRFS_LAST_FREE_OBJECTID < args.subvolid)
5093			return -EINVAL;
5094		break;
5095	case BTRFS_SUBVOL_SYNC_COUNT:
5096		spin_lock(&fs_info->trans_lock);
5097		args.count = list_count_nodes(&fs_info->dead_roots);
5098		spin_unlock(&fs_info->trans_lock);
5099		if (copy_to_user(argp, &args, sizeof(args)))
5100			return -EFAULT;
5101		return 0;
5102	case BTRFS_SUBVOL_SYNC_PEEK_FIRST:
5103		spin_lock(&fs_info->trans_lock);
5104		/* Last in the list was deleted first. */
5105		if (!list_empty(&fs_info->dead_roots)) {
5106			root = list_last_entry(&fs_info->dead_roots,
5107					       struct btrfs_root, root_list);
5108			args.subvolid = btrfs_root_id(root);
5109		} else {
5110			args.subvolid = 0;
5111		}
5112		spin_unlock(&fs_info->trans_lock);
5113		if (copy_to_user(argp, &args, sizeof(args)))
5114			return -EFAULT;
5115		return 0;
5116	case BTRFS_SUBVOL_SYNC_PEEK_LAST:
5117		spin_lock(&fs_info->trans_lock);
5118		/* First in the list was deleted last. */
5119		if (!list_empty(&fs_info->dead_roots)) {
5120			root = list_first_entry(&fs_info->dead_roots,
5121						struct btrfs_root, root_list);
5122			args.subvolid = btrfs_root_id(root);
5123		} else {
5124			args.subvolid = 0;
5125		}
5126		spin_unlock(&fs_info->trans_lock);
5127		if (copy_to_user(argp, &args, sizeof(args)))
5128			return -EFAULT;
5129		return 0;
5130	default:
5131		return -EINVAL;
5132	}
5133
5134	/* 32bit limitation: fs_roots_radix key is not wide enough. */
5135	if (sizeof(unsigned long) != sizeof(u64) && args.subvolid > U32_MAX)
5136		return -EOVERFLOW;
5137
5138	while (1) {
5139		/* Wait for the specific one. */
5140		if (down_read_interruptible(&fs_info->subvol_sem) == -EINTR)
5141			return -EINTR;
5142		refs = -1;
5143		spin_lock(&fs_info->fs_roots_radix_lock);
5144		root = radix_tree_lookup(&fs_info->fs_roots_radix,
5145					 (unsigned long)args.subvolid);
5146		if (root) {
5147			spin_lock(&root->root_item_lock);
5148			refs = btrfs_root_refs(&root->root_item);
5149			root_flags = btrfs_root_flags(&root->root_item);
5150			spin_unlock(&root->root_item_lock);
5151		}
5152		spin_unlock(&fs_info->fs_roots_radix_lock);
5153		up_read(&fs_info->subvol_sem);
5154
5155		/* Subvolume does not exist. */
5156		if (!root)
5157			return -ENOENT;
5158
5159		/* Subvolume not deleted at all. */
5160		if (refs > 0)
5161			return -EEXIST;
5162		/* We've waited and now the subvolume is gone. */
5163		if (wait_for_deletion && refs == -1) {
5164			/* Return the one we waited for as the last one. */
5165			if (copy_to_user(argp, &args, sizeof(args)))
5166				return -EFAULT;
5167			return 0;
5168		}
5169
5170		/* Subvolume not found on the first try (deleted or never existed). */
5171		if (refs == -1)
5172			return -ENOENT;
5173
5174		wait_for_deletion = true;
5175		ASSERT(root_flags & BTRFS_ROOT_SUBVOL_DEAD);
5176		sched_ret = schedule_timeout_interruptible(HZ);
5177		/* Early wake up or error. */
5178		if (sched_ret != 0)
5179			return -EINTR;
5180	}
5181
5182	return 0;
5183}
5184
5185#ifdef CONFIG_BTRFS_EXPERIMENTAL
5186static int btrfs_ioctl_shutdown(struct btrfs_fs_info *fs_info, unsigned long arg)
5187{
5188	int ret = 0;
5189	u32 flags;
5190
5191	if (!capable(CAP_SYS_ADMIN))
5192		return -EPERM;
5193
5194	if (get_user(flags, (u32 __user *)arg))
5195		return -EFAULT;
5196
5197	if (flags >= BTRFS_SHUTDOWN_FLAGS_LAST)
5198		return -EINVAL;
5199
5200	if (btrfs_is_shutdown(fs_info))
5201		return 0;
5202
5203	switch (flags) {
5204	case BTRFS_SHUTDOWN_FLAGS_LOGFLUSH:
5205	case BTRFS_SHUTDOWN_FLAGS_DEFAULT:
5206		ret = freeze_super(fs_info->sb, FREEZE_HOLDER_KERNEL, NULL);
5207		if (ret)
5208			return ret;
5209		btrfs_force_shutdown(fs_info);
5210		ret = thaw_super(fs_info->sb, FREEZE_HOLDER_KERNEL, NULL);
5211		if (ret)
5212			return ret;
5213		break;
5214	case BTRFS_SHUTDOWN_FLAGS_NOLOGFLUSH:
5215		btrfs_force_shutdown(fs_info);
5216		break;
5217	}
5218	return ret;
5219}
5220#endif
5221
5222long btrfs_ioctl(struct file *file, unsigned int
5223		cmd, unsigned long arg)
5224{
5225	struct inode *inode = file_inode(file);
5226	struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
5227	struct btrfs_root *root = BTRFS_I(inode)->root;
5228	void __user *argp = (void __user *)arg;
5229
5230	switch (cmd) {
5231	case FS_IOC_GETVERSION:
5232		return btrfs_ioctl_getversion(inode, argp);
5233	case FS_IOC_GETFSLABEL:
5234		return btrfs_ioctl_get_fslabel(fs_info, argp);
5235	case FS_IOC_SETFSLABEL:
5236		return btrfs_ioctl_set_fslabel(file, argp);
5237	case FITRIM:
5238		return btrfs_ioctl_fitrim(fs_info, argp);
5239	case BTRFS_IOC_SNAP_CREATE:
5240		return btrfs_ioctl_snap_create(file, argp, false);
5241	case BTRFS_IOC_SNAP_CREATE_V2:
5242		return btrfs_ioctl_snap_create_v2(file, argp, false);
5243	case BTRFS_IOC_SUBVOL_CREATE:
5244		return btrfs_ioctl_snap_create(file, argp, true);
5245	case BTRFS_IOC_SUBVOL_CREATE_V2:
5246		return btrfs_ioctl_snap_create_v2(file, argp, true);
5247	case BTRFS_IOC_SNAP_DESTROY:
5248		return btrfs_ioctl_snap_destroy(file, argp, false);
5249	case BTRFS_IOC_SNAP_DESTROY_V2:
5250		return btrfs_ioctl_snap_destroy(file, argp, true);
5251	case BTRFS_IOC_SUBVOL_GETFLAGS:
5252		return btrfs_ioctl_subvol_getflags(BTRFS_I(inode), argp);
5253	case BTRFS_IOC_SUBVOL_SETFLAGS:
5254		return btrfs_ioctl_subvol_setflags(file, argp);
5255	case BTRFS_IOC_DEFAULT_SUBVOL:
5256		return btrfs_ioctl_default_subvol(file, argp);
5257	case BTRFS_IOC_DEFRAG:
5258		return btrfs_ioctl_defrag(file, NULL);
5259	case BTRFS_IOC_DEFRAG_RANGE:
5260		return btrfs_ioctl_defrag(file, argp);
5261	case BTRFS_IOC_RESIZE:
5262		return btrfs_ioctl_resize(file, argp);
5263	case BTRFS_IOC_ADD_DEV:
5264		return btrfs_ioctl_add_dev(fs_info, argp);
5265	case BTRFS_IOC_RM_DEV:
5266		return btrfs_ioctl_rm_dev(file, argp);
5267	case BTRFS_IOC_RM_DEV_V2:
5268		return btrfs_ioctl_rm_dev_v2(file, argp);
5269	case BTRFS_IOC_FS_INFO:
5270		return btrfs_ioctl_fs_info(fs_info, argp);
5271	case BTRFS_IOC_DEV_INFO:
5272		return btrfs_ioctl_dev_info(fs_info, argp);
5273	case BTRFS_IOC_TREE_SEARCH:
5274		return btrfs_ioctl_tree_search(root, argp);
5275	case BTRFS_IOC_TREE_SEARCH_V2:
5276		return btrfs_ioctl_tree_search_v2(root, argp);
5277	case BTRFS_IOC_INO_LOOKUP:
5278		return btrfs_ioctl_ino_lookup(root, argp);
5279	case BTRFS_IOC_INO_PATHS:
5280		return btrfs_ioctl_ino_to_path(root, argp);
5281	case BTRFS_IOC_LOGICAL_INO:
5282		return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5283	case BTRFS_IOC_LOGICAL_INO_V2:
5284		return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5285	case BTRFS_IOC_SPACE_INFO:
5286		return btrfs_ioctl_space_info(fs_info, argp);
5287	case BTRFS_IOC_SYNC: {
5288		int ret;
5289
5290		ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
5291		if (ret)
5292			return ret;
5293		ret = btrfs_sync_fs(inode->i_sb, 1);
5294		/*
5295		 * There may be work for the cleaner kthread to do (subvolume
5296		 * deletion, delayed iputs, defrag inodes, etc), so wake it up.
5297		 */
5298		wake_up_process(fs_info->cleaner_kthread);
5299		return ret;
5300	}
5301	case BTRFS_IOC_START_SYNC:
5302		return btrfs_ioctl_start_sync(root, argp);
5303	case BTRFS_IOC_WAIT_SYNC:
5304		return btrfs_ioctl_wait_sync(fs_info, argp);
5305	case BTRFS_IOC_SCRUB:
5306		return btrfs_ioctl_scrub(file, argp);
5307	case BTRFS_IOC_SCRUB_CANCEL:
5308		return btrfs_ioctl_scrub_cancel(fs_info);
5309	case BTRFS_IOC_SCRUB_PROGRESS:
5310		return btrfs_ioctl_scrub_progress(fs_info, argp);
5311	case BTRFS_IOC_BALANCE_V2:
5312		return btrfs_ioctl_balance(file, argp);
5313	case BTRFS_IOC_BALANCE_CTL:
5314		return btrfs_ioctl_balance_ctl(fs_info, arg);
5315	case BTRFS_IOC_BALANCE_PROGRESS:
5316		return btrfs_ioctl_balance_progress(fs_info, argp);
5317	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5318		return btrfs_ioctl_set_received_subvol(file, argp);
5319#ifdef CONFIG_64BIT
5320	case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5321		return btrfs_ioctl_set_received_subvol_32(file, argp);
5322#endif
5323	case BTRFS_IOC_SEND:
5324		return _btrfs_ioctl_send(root, argp, false);
5325#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5326	case BTRFS_IOC_SEND_32:
5327		return _btrfs_ioctl_send(root, argp, true);
5328#endif
5329	case BTRFS_IOC_GET_DEV_STATS:
5330		return btrfs_ioctl_get_dev_stats(fs_info, argp);
5331	case BTRFS_IOC_QUOTA_CTL:
5332		return btrfs_ioctl_quota_ctl(file, argp);
5333	case BTRFS_IOC_QGROUP_ASSIGN:
5334		return btrfs_ioctl_qgroup_assign(file, argp);
5335	case BTRFS_IOC_QGROUP_CREATE:
5336		return btrfs_ioctl_qgroup_create(file, argp);
5337	case BTRFS_IOC_QGROUP_LIMIT:
5338		return btrfs_ioctl_qgroup_limit(file, argp);
5339	case BTRFS_IOC_QUOTA_RESCAN:
5340		return btrfs_ioctl_quota_rescan(file, argp);
5341	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5342		return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5343	case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5344		return btrfs_ioctl_quota_rescan_wait(fs_info);
5345	case BTRFS_IOC_DEV_REPLACE:
5346		return btrfs_ioctl_dev_replace(fs_info, argp);
5347	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5348		return btrfs_ioctl_get_supported_features(argp);
5349	case BTRFS_IOC_GET_FEATURES:
5350		return btrfs_ioctl_get_features(fs_info, argp);
5351	case BTRFS_IOC_SET_FEATURES:
5352		return btrfs_ioctl_set_features(file, argp);
5353	case BTRFS_IOC_GET_SUBVOL_INFO:
5354		return btrfs_ioctl_get_subvol_info(inode, argp);
5355	case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5356		return btrfs_ioctl_get_subvol_rootref(root, argp);
5357	case BTRFS_IOC_INO_LOOKUP_USER:
5358		return btrfs_ioctl_ino_lookup_user(file, argp);
5359	case FS_IOC_ENABLE_VERITY:
5360		return fsverity_ioctl_enable(file, (const void __user *)argp);
5361	case FS_IOC_MEASURE_VERITY:
5362		return fsverity_ioctl_measure(file, argp);
5363	case FS_IOC_READ_VERITY_METADATA:
5364		return fsverity_ioctl_read_metadata(file, argp);
5365	case BTRFS_IOC_ENCODED_READ:
5366		return btrfs_ioctl_encoded_read(file, argp, false);
5367	case BTRFS_IOC_ENCODED_WRITE:
5368		return btrfs_ioctl_encoded_write(file, argp, false);
5369#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5370	case BTRFS_IOC_ENCODED_READ_32:
5371		return btrfs_ioctl_encoded_read(file, argp, true);
5372	case BTRFS_IOC_ENCODED_WRITE_32:
5373		return btrfs_ioctl_encoded_write(file, argp, true);
5374#endif
5375	case BTRFS_IOC_SUBVOL_SYNC_WAIT:
5376		return btrfs_ioctl_subvol_sync(fs_info, argp);
5377#ifdef CONFIG_BTRFS_EXPERIMENTAL
5378	case BTRFS_IOC_SHUTDOWN:
5379		return btrfs_ioctl_shutdown(fs_info, arg);
5380#endif
5381	}
5382
5383	return -ENOTTY;
5384}
5385
5386#ifdef CONFIG_COMPAT
5387long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5388{
5389	/*
5390	 * These all access 32-bit values anyway so no further
5391	 * handling is necessary.
5392	 */
5393	switch (cmd) {
5394	case FS_IOC32_GETVERSION:
5395		cmd = FS_IOC_GETVERSION;
5396		break;
5397	}
5398
5399	return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5400}
5401#endif