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