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_set_ctime_current(inode);
 388         ret = btrfs_update_inode(trans, 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         /* Tree log can't currently deal with an inode which is a new root. */
 653         btrfs_set_log_full_commit(trans);
 654
 655         ret = btrfs_qgroup_inherit(trans, 0, objectid, root->root_key.objectid, inherit);
 656         if (ret)
 657                 goto out;
 658
 659         leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
 660                                       0, BTRFS_NESTING_NORMAL);
 661         if (IS_ERR(leaf)) {
 662                 ret = PTR_ERR(leaf);
 663                 goto out;
 664         }
 665
 666         btrfs_mark_buffer_dirty(trans, leaf);
 667
 668         inode_item = &root_item->inode;
 669         btrfs_set_stack_inode_generation(inode_item, 1);
 670         btrfs_set_stack_inode_size(inode_item, 3);
 671         btrfs_set_stack_inode_nlink(inode_item, 1);
 672         btrfs_set_stack_inode_nbytes(inode_item,
 673                                      fs_info->nodesize);
 674         btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
 675
 676         btrfs_set_root_flags(root_item, 0);
 677         btrfs_set_root_limit(root_item, 0);
 678         btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
 679
 680         btrfs_set_root_bytenr(root_item, leaf->start);
 681         btrfs_set_root_generation(root_item, trans->transid);
 682         btrfs_set_root_level(root_item, 0);
 683         btrfs_set_root_refs(root_item, 1);
 684         btrfs_set_root_used(root_item, leaf->len);
 685         btrfs_set_root_last_snapshot(root_item, 0);
 686
 687         btrfs_set_root_generation_v2(root_item,
 688                         btrfs_root_generation(root_item));
 689         generate_random_guid(root_item->uuid);
 690         btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
 691         btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
 692         root_item->ctime = root_item->otime;
 693         btrfs_set_root_ctransid(root_item, trans->transid);
 694         btrfs_set_root_otransid(root_item, trans->transid);
 695
 696         btrfs_tree_unlock(leaf);
 697
 698         btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
 699
 700         key.objectid = objectid;
 701         key.offset = 0;
 702         key.type = BTRFS_ROOT_ITEM_KEY;
 703         ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
 704                                 root_item);
 705         if (ret) {
 706                 /*
 707                  * Since we don't abort the transaction in this case, free the
 708                  * tree block so that we don't leak space and leave the
 709                  * filesystem in an inconsistent state (an extent item in the
 710                  * extent tree with a backreference for a root that does not
 711                  * exists).
 712                  */
 713                 btrfs_tree_lock(leaf);
 714                 btrfs_clear_buffer_dirty(trans, leaf);
 715                 btrfs_tree_unlock(leaf);
 716                 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
 717                 free_extent_buffer(leaf);
 718                 goto out;
 719         }
 720
 721         free_extent_buffer(leaf);
 722         leaf = NULL;
 723
 724         new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
 725         if (IS_ERR(new_root)) {
 726                 ret = PTR_ERR(new_root);
 727                 btrfs_abort_transaction(trans, ret);
 728                 goto out;
 729         }
 730         /* anon_dev is owned by new_root now. */
 731         anon_dev = 0;
 732         BTRFS_I(new_inode_args.inode)->root = new_root;
 733         /* ... and new_root is owned by new_inode_args.inode now. */
 734
 735         ret = btrfs_record_root_in_trans(trans, new_root);
 736         if (ret) {
 737                 btrfs_abort_transaction(trans, ret);
 738                 goto out;
 739         }
 740
 741         ret = btrfs_uuid_tree_add(trans, root_item->uuid,
 742                                   BTRFS_UUID_KEY_SUBVOL, objectid);
 743         if (ret) {
 744                 btrfs_abort_transaction(trans, ret);
 745                 goto out;
 746         }
 747
 748         ret = btrfs_create_new_inode(trans, &new_inode_args);
 749         if (ret) {
 750                 btrfs_abort_transaction(trans, ret);
 751                 goto out;
 752         }
 753
 754         d_instantiate_new(dentry, new_inode_args.inode);
 755         new_inode_args.inode = NULL;
 756
 757 out:
 758         trans->block_rsv = NULL;
 759         trans->bytes_reserved = 0;
 760         btrfs_subvolume_release_metadata(root, &block_rsv);
 761
 762         btrfs_end_transaction(trans);
 763 out_new_inode_args:
 764         btrfs_new_inode_args_destroy(&new_inode_args);
 765 out_inode:
 766         iput(new_inode_args.inode);
 767 out_anon_dev:
 768         if (anon_dev)
 769                 free_anon_bdev(anon_dev);
 770 out_root_item:
 771         kfree(root_item);
 772         return ret;
 773 }
 774
 775 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
 776                            struct dentry *dentry, bool readonly,
 777                            struct btrfs_qgroup_inherit *inherit)
 778 {
 779         struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
 780         struct inode *inode;
 781         struct btrfs_pending_snapshot *pending_snapshot;
 782         unsigned int trans_num_items;
 783         struct btrfs_trans_handle *trans;
 784         int ret;
 785
 786         /* We do not support snapshotting right now. */
 787         if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
 788                 btrfs_warn(fs_info,
 789                            "extent tree v2 doesn't support snapshotting yet");
 790                 return -EOPNOTSUPP;
 791         }
 792
 793         if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
 794                 return -EINVAL;
 795
 796         if (atomic_read(&root->nr_swapfiles)) {
 797                 btrfs_warn(fs_info,
 798                            "cannot snapshot subvolume with active swapfile");
 799                 return -ETXTBSY;
 800         }
 801
 802         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
 803         if (!pending_snapshot)
 804                 return -ENOMEM;
 805
 806         ret = get_anon_bdev(&pending_snapshot->anon_dev);
 807         if (ret < 0)
 808                 goto free_pending;
 809         pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
 810                         GFP_KERNEL);
 811         pending_snapshot->path = btrfs_alloc_path();
 812         if (!pending_snapshot->root_item || !pending_snapshot->path) {
 813                 ret = -ENOMEM;
 814                 goto free_pending;
 815         }
 816
 817         btrfs_init_block_rsv(&pending_snapshot->block_rsv,
 818                              BTRFS_BLOCK_RSV_TEMP);
 819         /*
 820          * 1 to add dir item
 821          * 1 to add dir index
 822          * 1 to update parent inode item
 823          */
 824         trans_num_items = create_subvol_num_items(inherit) + 3;
 825         ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
 826                                                &pending_snapshot->block_rsv,
 827                                                trans_num_items, false);
 828         if (ret)
 829                 goto free_pending;
 830
 831         pending_snapshot->dentry = dentry;
 832         pending_snapshot->root = root;
 833         pending_snapshot->readonly = readonly;
 834         pending_snapshot->dir = dir;
 835         pending_snapshot->inherit = inherit;
 836
 837         trans = btrfs_start_transaction(root, 0);
 838         if (IS_ERR(trans)) {
 839                 ret = PTR_ERR(trans);
 840                 goto fail;
 841         }
 842
 843         trans->pending_snapshot = pending_snapshot;
 844
 845         ret = btrfs_commit_transaction(trans);
 846         if (ret)
 847                 goto fail;
 848
 849         ret = pending_snapshot->error;
 850         if (ret)
 851                 goto fail;
 852
 853         ret = btrfs_orphan_cleanup(pending_snapshot->snap);
 854         if (ret)
 855                 goto fail;
 856
 857         inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
 858         if (IS_ERR(inode)) {
 859                 ret = PTR_ERR(inode);
 860                 goto fail;
 861         }
 862
 863         d_instantiate(dentry, inode);
 864         ret = 0;
 865         pending_snapshot->anon_dev = 0;
 866 fail:
 867         /* Prevent double freeing of anon_dev */
 868         if (ret && pending_snapshot->snap)
 869                 pending_snapshot->snap->anon_dev = 0;
 870         btrfs_put_root(pending_snapshot->snap);
 871         btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
 872 free_pending:
 873         if (pending_snapshot->anon_dev)
 874                 free_anon_bdev(pending_snapshot->anon_dev);
 875         kfree(pending_snapshot->root_item);
 876         btrfs_free_path(pending_snapshot->path);
 877         kfree(pending_snapshot);
 878
 879         return ret;
 880 }
 881
 882 /*  copy of may_delete in fs/namei.c()
 883  *      Check whether we can remove a link victim from directory dir, check
 884  *  whether the type of victim is right.
 885  *  1. We can't do it if dir is read-only (done in permission())
 886  *  2. We should have write and exec permissions on dir
 887  *  3. We can't remove anything from append-only dir
 888  *  4. We can't do anything with immutable dir (done in permission())
 889  *  5. If the sticky bit on dir is set we should either
 890  *      a. be owner of dir, or
 891  *      b. be owner of victim, or
 892  *      c. have CAP_FOWNER capability
 893  *  6. If the victim is append-only or immutable we can't do anything with
 894  *     links pointing to it.
 895  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
 896  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
 897  *  9. We can't remove a root or mountpoint.
 898  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
 899  *     nfs_async_unlink().
 900  */
 901
 902 static int btrfs_may_delete(struct mnt_idmap *idmap,
 903                             struct inode *dir, struct dentry *victim, int isdir)
 904 {
 905         int error;
 906
 907         if (d_really_is_negative(victim))
 908                 return -ENOENT;
 909
 910         BUG_ON(d_inode(victim->d_parent) != dir);
 911         audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
 912
 913         error = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
 914         if (error)
 915                 return error;
 916         if (IS_APPEND(dir))
 917                 return -EPERM;
 918         if (check_sticky(idmap, dir, d_inode(victim)) ||
 919             IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
 920             IS_SWAPFILE(d_inode(victim)))
 921                 return -EPERM;
 922         if (isdir) {
 923                 if (!d_is_dir(victim))
 924                         return -ENOTDIR;
 925                 if (IS_ROOT(victim))
 926                         return -EBUSY;
 927         } else if (d_is_dir(victim))
 928                 return -EISDIR;
 929         if (IS_DEADDIR(dir))
 930                 return -ENOENT;
 931         if (victim->d_flags & DCACHE_NFSFS_RENAMED)
 932                 return -EBUSY;
 933         return 0;
 934 }
 935
 936 /* copy of may_create in fs/namei.c() */
 937 static inline int btrfs_may_create(struct mnt_idmap *idmap,
 938                                    struct inode *dir, struct dentry *child)
 939 {
 940         if (d_really_is_positive(child))
 941                 return -EEXIST;
 942         if (IS_DEADDIR(dir))
 943                 return -ENOENT;
 944         if (!fsuidgid_has_mapping(dir->i_sb, idmap))
 945                 return -EOVERFLOW;
 946         return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
 947 }
 948
 949 /*
 950  * Create a new subvolume below @parent.  This is largely modeled after
 951  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
 952  * inside this filesystem so it's quite a bit simpler.
 953  */
 954 static noinline int btrfs_mksubvol(const struct path *parent,
 955                                    struct mnt_idmap *idmap,
 956                                    const char *name, int namelen,
 957                                    struct btrfs_root *snap_src,
 958                                    bool readonly,
 959                                    struct btrfs_qgroup_inherit *inherit)
 960 {
 961         struct inode *dir = d_inode(parent->dentry);
 962         struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
 963         struct dentry *dentry;
 964         struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
 965         int error;
 966
 967         error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
 968         if (error == -EINTR)
 969                 return error;
 970
 971         dentry = lookup_one(idmap, name, parent->dentry, namelen);
 972         error = PTR_ERR(dentry);
 973         if (IS_ERR(dentry))
 974                 goto out_unlock;
 975
 976         error = btrfs_may_create(idmap, dir, dentry);
 977         if (error)
 978                 goto out_dput;
 979
 980         /*
 981          * even if this name doesn't exist, we may get hash collisions.
 982          * check for them now when we can safely fail
 983          */
 984         error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
 985                                                dir->i_ino, &name_str);
 986         if (error)
 987                 goto out_dput;
 988
 989         down_read(&fs_info->subvol_sem);
 990
 991         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
 992                 goto out_up_read;
 993
 994         if (snap_src)
 995                 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
 996         else
 997                 error = create_subvol(idmap, dir, dentry, inherit);
 998
 999         if (!error)
1000                 fsnotify_mkdir(dir, dentry);
1001 out_up_read:
1002         up_read(&fs_info->subvol_sem);
1003 out_dput:
1004         dput(dentry);
1005 out_unlock:
1006         btrfs_inode_unlock(BTRFS_I(dir), 0);
1007         return error;
1008 }
1009
1010 static noinline int btrfs_mksnapshot(const struct path *parent,
1011                                    struct mnt_idmap *idmap,
1012                                    const char *name, int namelen,
1013                                    struct btrfs_root *root,
1014                                    bool readonly,
1015                                    struct btrfs_qgroup_inherit *inherit)
1016 {
1017         int ret;
1018         bool snapshot_force_cow = false;
1019
1020         /*
1021          * Force new buffered writes to reserve space even when NOCOW is
1022          * possible. This is to avoid later writeback (running dealloc) to
1023          * fallback to COW mode and unexpectedly fail with ENOSPC.
1024          */
1025         btrfs_drew_read_lock(&root->snapshot_lock);
1026
1027         ret = btrfs_start_delalloc_snapshot(root, false);
1028         if (ret)
1029                 goto out;
1030
1031         /*
1032          * All previous writes have started writeback in NOCOW mode, so now
1033          * we force future writes to fallback to COW mode during snapshot
1034          * creation.
1035          */
1036         atomic_inc(&root->snapshot_force_cow);
1037         snapshot_force_cow = true;
1038
1039         btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1040
1041         ret = btrfs_mksubvol(parent, idmap, name, namelen,
1042                              root, readonly, inherit);
1043 out:
1044         if (snapshot_force_cow)
1045                 atomic_dec(&root->snapshot_force_cow);
1046         btrfs_drew_read_unlock(&root->snapshot_lock);
1047         return ret;
1048 }
1049
1050 /*
1051  * Try to start exclusive operation @type or cancel it if it's running.
1052  *
1053  * Return:
1054  *   0        - normal mode, newly claimed op started
1055  *  >0        - normal mode, something else is running,
1056  *              return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1057  * ECANCELED  - cancel mode, successful cancel
1058  * ENOTCONN   - cancel mode, operation not running anymore
1059  */
1060 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1061                         enum btrfs_exclusive_operation type, bool cancel)
1062 {
1063         if (!cancel) {
1064                 /* Start normal op */
1065                 if (!btrfs_exclop_start(fs_info, type))
1066                         return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1067                 /* Exclusive operation is now claimed */
1068                 return 0;
1069         }
1070
1071         /* Cancel running op */
1072         if (btrfs_exclop_start_try_lock(fs_info, type)) {
1073                 /*
1074                  * This blocks any exclop finish from setting it to NONE, so we
1075                  * request cancellation. Either it runs and we will wait for it,
1076                  * or it has finished and no waiting will happen.
1077                  */
1078                 atomic_inc(&fs_info->reloc_cancel_req);
1079                 btrfs_exclop_start_unlock(fs_info);
1080
1081                 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1082                         wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1083                                     TASK_INTERRUPTIBLE);
1084
1085                 return -ECANCELED;
1086         }
1087
1088         /* Something else is running or none */
1089         return -ENOTCONN;
1090 }
1091
1092 static noinline int btrfs_ioctl_resize(struct file *file,
1093                                         void __user *arg)
1094 {
1095         BTRFS_DEV_LOOKUP_ARGS(args);
1096         struct inode *inode = file_inode(file);
1097         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1098         u64 new_size;
1099         u64 old_size;
1100         u64 devid = 1;
1101         struct btrfs_root *root = BTRFS_I(inode)->root;
1102         struct btrfs_ioctl_vol_args *vol_args;
1103         struct btrfs_trans_handle *trans;
1104         struct btrfs_device *device = NULL;
1105         char *sizestr;
1106         char *retptr;
1107         char *devstr = NULL;
1108         int ret = 0;
1109         int mod = 0;
1110         bool cancel;
1111
1112         if (!capable(CAP_SYS_ADMIN))
1113                 return -EPERM;
1114
1115         ret = mnt_want_write_file(file);
1116         if (ret)
1117                 return ret;
1118
1119         /*
1120          * Read the arguments before checking exclusivity to be able to
1121          * distinguish regular resize and cancel
1122          */
1123         vol_args = memdup_user(arg, sizeof(*vol_args));
1124         if (IS_ERR(vol_args)) {
1125                 ret = PTR_ERR(vol_args);
1126                 goto out_drop;
1127         }
1128         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1129         sizestr = vol_args->name;
1130         cancel = (strcmp("cancel", sizestr) == 0);
1131         ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1132         if (ret)
1133                 goto out_free;
1134         /* Exclusive operation is now claimed */
1135
1136         devstr = strchr(sizestr, ':');
1137         if (devstr) {
1138                 sizestr = devstr + 1;
1139                 *devstr = '\0';
1140                 devstr = vol_args->name;
1141                 ret = kstrtoull(devstr, 10, &devid);
1142                 if (ret)
1143                         goto out_finish;
1144                 if (!devid) {
1145                         ret = -EINVAL;
1146                         goto out_finish;
1147                 }
1148                 btrfs_info(fs_info, "resizing devid %llu", devid);
1149         }
1150
1151         args.devid = devid;
1152         device = btrfs_find_device(fs_info->fs_devices, &args);
1153         if (!device) {
1154                 btrfs_info(fs_info, "resizer unable to find device %llu",
1155                            devid);
1156                 ret = -ENODEV;
1157                 goto out_finish;
1158         }
1159
1160         if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1161                 btrfs_info(fs_info,
1162                            "resizer unable to apply on readonly device %llu",
1163                        devid);
1164                 ret = -EPERM;
1165                 goto out_finish;
1166         }
1167
1168         if (!strcmp(sizestr, "max"))
1169                 new_size = bdev_nr_bytes(device->bdev);
1170         else {
1171                 if (sizestr[0] == '-') {
1172                         mod = -1;
1173                         sizestr++;
1174                 } else if (sizestr[0] == '+') {
1175                         mod = 1;
1176                         sizestr++;
1177                 }
1178                 new_size = memparse(sizestr, &retptr);
1179                 if (*retptr != '\0' || new_size == 0) {
1180                         ret = -EINVAL;
1181                         goto out_finish;
1182                 }
1183         }
1184
1185         if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1186                 ret = -EPERM;
1187                 goto out_finish;
1188         }
1189
1190         old_size = btrfs_device_get_total_bytes(device);
1191
1192         if (mod < 0) {
1193                 if (new_size > old_size) {
1194                         ret = -EINVAL;
1195                         goto out_finish;
1196                 }
1197                 new_size = old_size - new_size;
1198         } else if (mod > 0) {
1199                 if (new_size > ULLONG_MAX - old_size) {
1200                         ret = -ERANGE;
1201                         goto out_finish;
1202                 }
1203                 new_size = old_size + new_size;
1204         }
1205
1206         if (new_size < SZ_256M) {
1207                 ret = -EINVAL;
1208                 goto out_finish;
1209         }
1210         if (new_size > bdev_nr_bytes(device->bdev)) {
1211                 ret = -EFBIG;
1212                 goto out_finish;
1213         }
1214
1215         new_size = round_down(new_size, fs_info->sectorsize);
1216
1217         if (new_size > old_size) {
1218                 trans = btrfs_start_transaction(root, 0);
1219                 if (IS_ERR(trans)) {
1220                         ret = PTR_ERR(trans);
1221                         goto out_finish;
1222                 }
1223                 ret = btrfs_grow_device(trans, device, new_size);
1224                 btrfs_commit_transaction(trans);
1225         } else if (new_size < old_size) {
1226                 ret = btrfs_shrink_device(device, new_size);
1227         } /* equal, nothing need to do */
1228
1229         if (ret == 0 && new_size != old_size)
1230                 btrfs_info_in_rcu(fs_info,
1231                         "resize device %s (devid %llu) from %llu to %llu",
1232                         btrfs_dev_name(device), device->devid,
1233                         old_size, new_size);
1234 out_finish:
1235         btrfs_exclop_finish(fs_info);
1236 out_free:
1237         kfree(vol_args);
1238 out_drop:
1239         mnt_drop_write_file(file);
1240         return ret;
1241 }
1242
1243 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1244                                 struct mnt_idmap *idmap,
1245                                 const char *name, unsigned long fd, int subvol,
1246                                 bool readonly,
1247                                 struct btrfs_qgroup_inherit *inherit)
1248 {
1249         int namelen;
1250         int ret = 0;
1251
1252         if (!S_ISDIR(file_inode(file)->i_mode))
1253                 return -ENOTDIR;
1254
1255         ret = mnt_want_write_file(file);
1256         if (ret)
1257                 goto out;
1258
1259         namelen = strlen(name);
1260         if (strchr(name, '/')) {
1261                 ret = -EINVAL;
1262                 goto out_drop_write;
1263         }
1264
1265         if (name[0] == '.' &&
1266            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1267                 ret = -EEXIST;
1268                 goto out_drop_write;
1269         }
1270
1271         if (subvol) {
1272                 ret = btrfs_mksubvol(&file->f_path, idmap, name,
1273                                      namelen, NULL, readonly, inherit);
1274         } else {
1275                 struct fd src = fdget(fd);
1276                 struct inode *src_inode;
1277                 if (!src.file) {
1278                         ret = -EINVAL;
1279                         goto out_drop_write;
1280                 }
1281
1282                 src_inode = file_inode(src.file);
1283                 if (src_inode->i_sb != file_inode(file)->i_sb) {
1284                         btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1285                                    "Snapshot src from another FS");
1286                         ret = -EXDEV;
1287                 } else if (!inode_owner_or_capable(idmap, src_inode)) {
1288                         /*
1289                          * Subvolume creation is not restricted, but snapshots
1290                          * are limited to own subvolumes only
1291                          */
1292                         ret = -EPERM;
1293                 } else {
1294                         ret = btrfs_mksnapshot(&file->f_path, idmap,
1295                                                name, namelen,
1296                                                BTRFS_I(src_inode)->root,
1297                                                readonly, inherit);
1298                 }
1299                 fdput(src);
1300         }
1301 out_drop_write:
1302         mnt_drop_write_file(file);
1303 out:
1304         return ret;
1305 }
1306
1307 static noinline int btrfs_ioctl_snap_create(struct file *file,
1308                                             void __user *arg, int subvol)
1309 {
1310         struct btrfs_ioctl_vol_args *vol_args;
1311         int ret;
1312
1313         if (!S_ISDIR(file_inode(file)->i_mode))
1314                 return -ENOTDIR;
1315
1316         vol_args = memdup_user(arg, sizeof(*vol_args));
1317         if (IS_ERR(vol_args))
1318                 return PTR_ERR(vol_args);
1319         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1320
1321         ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1322                                         vol_args->name, vol_args->fd, subvol,
1323                                         false, NULL);
1324
1325         kfree(vol_args);
1326         return ret;
1327 }
1328
1329 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1330                                                void __user *arg, int subvol)
1331 {
1332         struct btrfs_ioctl_vol_args_v2 *vol_args;
1333         int ret;
1334         bool readonly = false;
1335         struct btrfs_qgroup_inherit *inherit = NULL;
1336
1337         if (!S_ISDIR(file_inode(file)->i_mode))
1338                 return -ENOTDIR;
1339
1340         vol_args = memdup_user(arg, sizeof(*vol_args));
1341         if (IS_ERR(vol_args))
1342                 return PTR_ERR(vol_args);
1343         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1344
1345         if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1346                 ret = -EOPNOTSUPP;
1347                 goto free_args;
1348         }
1349
1350         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1351                 readonly = true;
1352         if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1353                 u64 nums;
1354
1355                 if (vol_args->size < sizeof(*inherit) ||
1356                     vol_args->size > PAGE_SIZE) {
1357                         ret = -EINVAL;
1358                         goto free_args;
1359                 }
1360                 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1361                 if (IS_ERR(inherit)) {
1362                         ret = PTR_ERR(inherit);
1363                         goto free_args;
1364                 }
1365
1366                 if (inherit->num_qgroups > PAGE_SIZE ||
1367                     inherit->num_ref_copies > PAGE_SIZE ||
1368                     inherit->num_excl_copies > PAGE_SIZE) {
1369                         ret = -EINVAL;
1370                         goto free_inherit;
1371                 }
1372
1373                 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1374                        2 * inherit->num_excl_copies;
1375                 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1376                         ret = -EINVAL;
1377                         goto free_inherit;
1378                 }
1379         }
1380
1381         ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1382                                         vol_args->name, vol_args->fd, subvol,
1383                                         readonly, inherit);
1384         if (ret)
1385                 goto free_inherit;
1386 free_inherit:
1387         kfree(inherit);
1388 free_args:
1389         kfree(vol_args);
1390         return ret;
1391 }
1392
1393 static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1394                                                 void __user *arg)
1395 {
1396         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1397         struct btrfs_root *root = BTRFS_I(inode)->root;
1398         int ret = 0;
1399         u64 flags = 0;
1400
1401         if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1402                 return -EINVAL;
1403
1404         down_read(&fs_info->subvol_sem);
1405         if (btrfs_root_readonly(root))
1406                 flags |= BTRFS_SUBVOL_RDONLY;
1407         up_read(&fs_info->subvol_sem);
1408
1409         if (copy_to_user(arg, &flags, sizeof(flags)))
1410                 ret = -EFAULT;
1411
1412         return ret;
1413 }
1414
1415 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1416                                               void __user *arg)
1417 {
1418         struct inode *inode = file_inode(file);
1419         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1420         struct btrfs_root *root = BTRFS_I(inode)->root;
1421         struct btrfs_trans_handle *trans;
1422         u64 root_flags;
1423         u64 flags;
1424         int ret = 0;
1425
1426         if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1427                 return -EPERM;
1428
1429         ret = mnt_want_write_file(file);
1430         if (ret)
1431                 goto out;
1432
1433         if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1434                 ret = -EINVAL;
1435                 goto out_drop_write;
1436         }
1437
1438         if (copy_from_user(&flags, arg, sizeof(flags))) {
1439                 ret = -EFAULT;
1440                 goto out_drop_write;
1441         }
1442
1443         if (flags & ~BTRFS_SUBVOL_RDONLY) {
1444                 ret = -EOPNOTSUPP;
1445                 goto out_drop_write;
1446         }
1447
1448         down_write(&fs_info->subvol_sem);
1449
1450         /* nothing to do */
1451         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1452                 goto out_drop_sem;
1453
1454         root_flags = btrfs_root_flags(&root->root_item);
1455         if (flags & BTRFS_SUBVOL_RDONLY) {
1456                 btrfs_set_root_flags(&root->root_item,
1457                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1458         } else {
1459                 /*
1460                  * Block RO -> RW transition if this subvolume is involved in
1461                  * send
1462                  */
1463                 spin_lock(&root->root_item_lock);
1464                 if (root->send_in_progress == 0) {
1465                         btrfs_set_root_flags(&root->root_item,
1466                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1467                         spin_unlock(&root->root_item_lock);
1468                 } else {
1469                         spin_unlock(&root->root_item_lock);
1470                         btrfs_warn(fs_info,
1471                                    "Attempt to set subvolume %llu read-write during send",
1472                                    root->root_key.objectid);
1473                         ret = -EPERM;
1474                         goto out_drop_sem;
1475                 }
1476         }
1477
1478         trans = btrfs_start_transaction(root, 1);
1479         if (IS_ERR(trans)) {
1480                 ret = PTR_ERR(trans);
1481                 goto out_reset;
1482         }
1483
1484         ret = btrfs_update_root(trans, fs_info->tree_root,
1485                                 &root->root_key, &root->root_item);
1486         if (ret < 0) {
1487                 btrfs_end_transaction(trans);
1488                 goto out_reset;
1489         }
1490
1491         ret = btrfs_commit_transaction(trans);
1492
1493 out_reset:
1494         if (ret)
1495                 btrfs_set_root_flags(&root->root_item, root_flags);
1496 out_drop_sem:
1497         up_write(&fs_info->subvol_sem);
1498 out_drop_write:
1499         mnt_drop_write_file(file);
1500 out:
1501         return ret;
1502 }
1503
1504 static noinline int key_in_sk(struct btrfs_key *key,
1505                               struct btrfs_ioctl_search_key *sk)
1506 {
1507         struct btrfs_key test;
1508         int ret;
1509
1510         test.objectid = sk->min_objectid;
1511         test.type = sk->min_type;
1512         test.offset = sk->min_offset;
1513
1514         ret = btrfs_comp_cpu_keys(key, &test);
1515         if (ret < 0)
1516                 return 0;
1517
1518         test.objectid = sk->max_objectid;
1519         test.type = sk->max_type;
1520         test.offset = sk->max_offset;
1521
1522         ret = btrfs_comp_cpu_keys(key, &test);
1523         if (ret > 0)
1524                 return 0;
1525         return 1;
1526 }
1527
1528 static noinline int copy_to_sk(struct btrfs_path *path,
1529                                struct btrfs_key *key,
1530                                struct btrfs_ioctl_search_key *sk,
1531                                u64 *buf_size,
1532                                char __user *ubuf,
1533                                unsigned long *sk_offset,
1534                                int *num_found)
1535 {
1536         u64 found_transid;
1537         struct extent_buffer *leaf;
1538         struct btrfs_ioctl_search_header sh;
1539         struct btrfs_key test;
1540         unsigned long item_off;
1541         unsigned long item_len;
1542         int nritems;
1543         int i;
1544         int slot;
1545         int ret = 0;
1546
1547         leaf = path->nodes[0];
1548         slot = path->slots[0];
1549         nritems = btrfs_header_nritems(leaf);
1550
1551         if (btrfs_header_generation(leaf) > sk->max_transid) {
1552                 i = nritems;
1553                 goto advance_key;
1554         }
1555         found_transid = btrfs_header_generation(leaf);
1556
1557         for (i = slot; i < nritems; i++) {
1558                 item_off = btrfs_item_ptr_offset(leaf, i);
1559                 item_len = btrfs_item_size(leaf, i);
1560
1561                 btrfs_item_key_to_cpu(leaf, key, i);
1562                 if (!key_in_sk(key, sk))
1563                         continue;
1564
1565                 if (sizeof(sh) + item_len > *buf_size) {
1566                         if (*num_found) {
1567                                 ret = 1;
1568                                 goto out;
1569                         }
1570
1571                         /*
1572                          * return one empty item back for v1, which does not
1573                          * handle -EOVERFLOW
1574                          */
1575
1576                         *buf_size = sizeof(sh) + item_len;
1577                         item_len = 0;
1578                         ret = -EOVERFLOW;
1579                 }
1580
1581                 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1582                         ret = 1;
1583                         goto out;
1584                 }
1585
1586                 sh.objectid = key->objectid;
1587                 sh.offset = key->offset;
1588                 sh.type = key->type;
1589                 sh.len = item_len;
1590                 sh.transid = found_transid;
1591
1592                 /*
1593                  * Copy search result header. If we fault then loop again so we
1594                  * can fault in the pages and -EFAULT there if there's a
1595                  * problem. Otherwise we'll fault and then copy the buffer in
1596                  * properly this next time through
1597                  */
1598                 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1599                         ret = 0;
1600                         goto out;
1601                 }
1602
1603                 *sk_offset += sizeof(sh);
1604
1605                 if (item_len) {
1606                         char __user *up = ubuf + *sk_offset;
1607                         /*
1608                          * Copy the item, same behavior as above, but reset the
1609                          * * sk_offset so we copy the full thing again.
1610                          */
1611                         if (read_extent_buffer_to_user_nofault(leaf, up,
1612                                                 item_off, item_len)) {
1613                                 ret = 0;
1614                                 *sk_offset -= sizeof(sh);
1615                                 goto out;
1616                         }
1617
1618                         *sk_offset += item_len;
1619                 }
1620                 (*num_found)++;
1621
1622                 if (ret) /* -EOVERFLOW from above */
1623                         goto out;
1624
1625                 if (*num_found >= sk->nr_items) {
1626                         ret = 1;
1627                         goto out;
1628                 }
1629         }
1630 advance_key:
1631         ret = 0;
1632         test.objectid = sk->max_objectid;
1633         test.type = sk->max_type;
1634         test.offset = sk->max_offset;
1635         if (btrfs_comp_cpu_keys(key, &test) >= 0)
1636                 ret = 1;
1637         else if (key->offset < (u64)-1)
1638                 key->offset++;
1639         else if (key->type < (u8)-1) {
1640                 key->offset = 0;
1641                 key->type++;
1642         } else if (key->objectid < (u64)-1) {
1643                 key->offset = 0;
1644                 key->type = 0;
1645                 key->objectid++;
1646         } else
1647                 ret = 1;
1648 out:
1649         /*
1650          *  0: all items from this leaf copied, continue with next
1651          *  1: * more items can be copied, but unused buffer is too small
1652          *     * all items were found
1653          *     Either way, it will stops the loop which iterates to the next
1654          *     leaf
1655          *  -EOVERFLOW: item was to large for buffer
1656          *  -EFAULT: could not copy extent buffer back to userspace
1657          */
1658         return ret;
1659 }
1660
1661 static noinline int search_ioctl(struct inode *inode,
1662                                  struct btrfs_ioctl_search_key *sk,
1663                                  u64 *buf_size,
1664                                  char __user *ubuf)
1665 {
1666         struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
1667         struct btrfs_root *root;
1668         struct btrfs_key key;
1669         struct btrfs_path *path;
1670         int ret;
1671         int num_found = 0;
1672         unsigned long sk_offset = 0;
1673
1674         if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1675                 *buf_size = sizeof(struct btrfs_ioctl_search_header);
1676                 return -EOVERFLOW;
1677         }
1678
1679         path = btrfs_alloc_path();
1680         if (!path)
1681                 return -ENOMEM;
1682
1683         if (sk->tree_id == 0) {
1684                 /* search the root of the inode that was passed */
1685                 root = btrfs_grab_root(BTRFS_I(inode)->root);
1686         } else {
1687                 root = btrfs_get_fs_root(info, sk->tree_id, true);
1688                 if (IS_ERR(root)) {
1689                         btrfs_free_path(path);
1690                         return PTR_ERR(root);
1691                 }
1692         }
1693
1694         key.objectid = sk->min_objectid;
1695         key.type = sk->min_type;
1696         key.offset = sk->min_offset;
1697
1698         while (1) {
1699                 ret = -EFAULT;
1700                 /*
1701                  * Ensure that the whole user buffer is faulted in at sub-page
1702                  * granularity, otherwise the loop may live-lock.
1703                  */
1704                 if (fault_in_subpage_writeable(ubuf + sk_offset,
1705                                                *buf_size - sk_offset))
1706                         break;
1707
1708                 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1709                 if (ret != 0) {
1710                         if (ret > 0)
1711                                 ret = 0;
1712                         goto err;
1713                 }
1714                 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1715                                  &sk_offset, &num_found);
1716                 btrfs_release_path(path);
1717                 if (ret)
1718                         break;
1719
1720         }
1721         if (ret > 0)
1722                 ret = 0;
1723 err:
1724         sk->nr_items = num_found;
1725         btrfs_put_root(root);
1726         btrfs_free_path(path);
1727         return ret;
1728 }
1729
1730 static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1731                                             void __user *argp)
1732 {
1733         struct btrfs_ioctl_search_args __user *uargs = argp;
1734         struct btrfs_ioctl_search_key sk;
1735         int ret;
1736         u64 buf_size;
1737
1738         if (!capable(CAP_SYS_ADMIN))
1739                 return -EPERM;
1740
1741         if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1742                 return -EFAULT;
1743
1744         buf_size = sizeof(uargs->buf);
1745
1746         ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1747
1748         /*
1749          * In the origin implementation an overflow is handled by returning a
1750          * search header with a len of zero, so reset ret.
1751          */
1752         if (ret == -EOVERFLOW)
1753                 ret = 0;
1754
1755         if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1756                 ret = -EFAULT;
1757         return ret;
1758 }
1759
1760 static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1761                                                void __user *argp)
1762 {
1763         struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1764         struct btrfs_ioctl_search_args_v2 args;
1765         int ret;
1766         u64 buf_size;
1767         const u64 buf_limit = SZ_16M;
1768
1769         if (!capable(CAP_SYS_ADMIN))
1770                 return -EPERM;
1771
1772         /* copy search header and buffer size */
1773         if (copy_from_user(&args, uarg, sizeof(args)))
1774                 return -EFAULT;
1775
1776         buf_size = args.buf_size;
1777
1778         /* limit result size to 16MB */
1779         if (buf_size > buf_limit)
1780                 buf_size = buf_limit;
1781
1782         ret = search_ioctl(inode, &args.key, &buf_size,
1783                            (char __user *)(&uarg->buf[0]));
1784         if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1785                 ret = -EFAULT;
1786         else if (ret == -EOVERFLOW &&
1787                 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1788                 ret = -EFAULT;
1789
1790         return ret;
1791 }
1792
1793 /*
1794  * Search INODE_REFs to identify path name of 'dirid' directory
1795  * in a 'tree_id' tree. and sets path name to 'name'.
1796  */
1797 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1798                                 u64 tree_id, u64 dirid, char *name)
1799 {
1800         struct btrfs_root *root;
1801         struct btrfs_key key;
1802         char *ptr;
1803         int ret = -1;
1804         int slot;
1805         int len;
1806         int total_len = 0;
1807         struct btrfs_inode_ref *iref;
1808         struct extent_buffer *l;
1809         struct btrfs_path *path;
1810
1811         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1812                 name[0]='\0';
1813                 return 0;
1814         }
1815
1816         path = btrfs_alloc_path();
1817         if (!path)
1818                 return -ENOMEM;
1819
1820         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1821
1822         root = btrfs_get_fs_root(info, tree_id, true);
1823         if (IS_ERR(root)) {
1824                 ret = PTR_ERR(root);
1825                 root = NULL;
1826                 goto out;
1827         }
1828
1829         key.objectid = dirid;
1830         key.type = BTRFS_INODE_REF_KEY;
1831         key.offset = (u64)-1;
1832
1833         while (1) {
1834                 ret = btrfs_search_backwards(root, &key, path);
1835                 if (ret < 0)
1836                         goto out;
1837                 else if (ret > 0) {
1838                         ret = -ENOENT;
1839                         goto out;
1840                 }
1841
1842                 l = path->nodes[0];
1843                 slot = path->slots[0];
1844
1845                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1846                 len = btrfs_inode_ref_name_len(l, iref);
1847                 ptr -= len + 1;
1848                 total_len += len + 1;
1849                 if (ptr < name) {
1850                         ret = -ENAMETOOLONG;
1851                         goto out;
1852                 }
1853
1854                 *(ptr + len) = '/';
1855                 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1856
1857                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1858                         break;
1859
1860                 btrfs_release_path(path);
1861                 key.objectid = key.offset;
1862                 key.offset = (u64)-1;
1863                 dirid = key.objectid;
1864         }
1865         memmove(name, ptr, total_len);
1866         name[total_len] = '\0';
1867         ret = 0;
1868 out:
1869         btrfs_put_root(root);
1870         btrfs_free_path(path);
1871         return ret;
1872 }
1873
1874 static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1875                                 struct inode *inode,
1876                                 struct btrfs_ioctl_ino_lookup_user_args *args)
1877 {
1878         struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1879         struct super_block *sb = inode->i_sb;
1880         struct btrfs_key upper_limit = BTRFS_I(inode)->location;
1881         u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
1882         u64 dirid = args->dirid;
1883         unsigned long item_off;
1884         unsigned long item_len;
1885         struct btrfs_inode_ref *iref;
1886         struct btrfs_root_ref *rref;
1887         struct btrfs_root *root = NULL;
1888         struct btrfs_path *path;
1889         struct btrfs_key key, key2;
1890         struct extent_buffer *leaf;
1891         struct inode *temp_inode;
1892         char *ptr;
1893         int slot;
1894         int len;
1895         int total_len = 0;
1896         int ret;
1897
1898         path = btrfs_alloc_path();
1899         if (!path)
1900                 return -ENOMEM;
1901
1902         /*
1903          * If the bottom subvolume does not exist directly under upper_limit,
1904          * construct the path in from the bottom up.
1905          */
1906         if (dirid != upper_limit.objectid) {
1907                 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1908
1909                 root = btrfs_get_fs_root(fs_info, treeid, true);
1910                 if (IS_ERR(root)) {
1911                         ret = PTR_ERR(root);
1912                         goto out;
1913                 }
1914
1915                 key.objectid = dirid;
1916                 key.type = BTRFS_INODE_REF_KEY;
1917                 key.offset = (u64)-1;
1918                 while (1) {
1919                         ret = btrfs_search_backwards(root, &key, path);
1920                         if (ret < 0)
1921                                 goto out_put;
1922                         else if (ret > 0) {
1923                                 ret = -ENOENT;
1924                                 goto out_put;
1925                         }
1926
1927                         leaf = path->nodes[0];
1928                         slot = path->slots[0];
1929
1930                         iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1931                         len = btrfs_inode_ref_name_len(leaf, iref);
1932                         ptr -= len + 1;
1933                         total_len += len + 1;
1934                         if (ptr < args->path) {
1935                                 ret = -ENAMETOOLONG;
1936                                 goto out_put;
1937                         }
1938
1939                         *(ptr + len) = '/';
1940                         read_extent_buffer(leaf, ptr,
1941                                         (unsigned long)(iref + 1), len);
1942
1943                         /* Check the read+exec permission of this directory */
1944                         ret = btrfs_previous_item(root, path, dirid,
1945                                                   BTRFS_INODE_ITEM_KEY);
1946                         if (ret < 0) {
1947                                 goto out_put;
1948                         } else if (ret > 0) {
1949                                 ret = -ENOENT;
1950                                 goto out_put;
1951                         }
1952
1953                         leaf = path->nodes[0];
1954                         slot = path->slots[0];
1955                         btrfs_item_key_to_cpu(leaf, &key2, slot);
1956                         if (key2.objectid != dirid) {
1957                                 ret = -ENOENT;
1958                                 goto out_put;
1959                         }
1960
1961                         /*
1962                          * We don't need the path anymore, so release it and
1963                          * avoid deadlocks and lockdep warnings in case
1964                          * btrfs_iget() needs to lookup the inode from its root
1965                          * btree and lock the same leaf.
1966                          */
1967                         btrfs_release_path(path);
1968                         temp_inode = btrfs_iget(sb, key2.objectid, root);
1969                         if (IS_ERR(temp_inode)) {
1970                                 ret = PTR_ERR(temp_inode);
1971                                 goto out_put;
1972                         }
1973                         ret = inode_permission(idmap, temp_inode,
1974                                                MAY_READ | MAY_EXEC);
1975                         iput(temp_inode);
1976                         if (ret) {
1977                                 ret = -EACCES;
1978                                 goto out_put;
1979                         }
1980
1981                         if (key.offset == upper_limit.objectid)
1982                                 break;
1983                         if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
1984                                 ret = -EACCES;
1985                                 goto out_put;
1986                         }
1987
1988                         key.objectid = key.offset;
1989                         key.offset = (u64)-1;
1990                         dirid = key.objectid;
1991                 }
1992
1993                 memmove(args->path, ptr, total_len);
1994                 args->path[total_len] = '\0';
1995                 btrfs_put_root(root);
1996                 root = NULL;
1997                 btrfs_release_path(path);
1998         }
1999
2000         /* Get the bottom subvolume's name from ROOT_REF */
2001         key.objectid = treeid;
2002         key.type = BTRFS_ROOT_REF_KEY;
2003         key.offset = args->treeid;
2004         ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2005         if (ret < 0) {
2006                 goto out;
2007         } else if (ret > 0) {
2008                 ret = -ENOENT;
2009                 goto out;
2010         }
2011
2012         leaf = path->nodes[0];
2013         slot = path->slots[0];
2014         btrfs_item_key_to_cpu(leaf, &key, slot);
2015
2016         item_off = btrfs_item_ptr_offset(leaf, slot);
2017         item_len = btrfs_item_size(leaf, slot);
2018         /* Check if dirid in ROOT_REF corresponds to passed dirid */
2019         rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2020         if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2021                 ret = -EINVAL;
2022                 goto out;
2023         }
2024
2025         /* Copy subvolume's name */
2026         item_off += sizeof(struct btrfs_root_ref);
2027         item_len -= sizeof(struct btrfs_root_ref);
2028         read_extent_buffer(leaf, args->name, item_off, item_len);
2029         args->name[item_len] = 0;
2030
2031 out_put:
2032         btrfs_put_root(root);
2033 out:
2034         btrfs_free_path(path);
2035         return ret;
2036 }
2037
2038 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2039                                            void __user *argp)
2040 {
2041         struct btrfs_ioctl_ino_lookup_args *args;
2042         int ret = 0;
2043
2044         args = memdup_user(argp, sizeof(*args));
2045         if (IS_ERR(args))
2046                 return PTR_ERR(args);
2047
2048         /*
2049          * Unprivileged query to obtain the containing subvolume root id. The
2050          * path is reset so it's consistent with btrfs_search_path_in_tree.
2051          */
2052         if (args->treeid == 0)
2053                 args->treeid = root->root_key.objectid;
2054
2055         if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2056                 args->name[0] = 0;
2057                 goto out;
2058         }
2059
2060         if (!capable(CAP_SYS_ADMIN)) {
2061                 ret = -EPERM;
2062                 goto out;
2063         }
2064
2065         ret = btrfs_search_path_in_tree(root->fs_info,
2066                                         args->treeid, args->objectid,
2067                                         args->name);
2068
2069 out:
2070         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2071                 ret = -EFAULT;
2072
2073         kfree(args);
2074         return ret;
2075 }
2076
2077 /*
2078  * Version of ino_lookup ioctl (unprivileged)
2079  *
2080  * The main differences from ino_lookup ioctl are:
2081  *
2082  *   1. Read + Exec permission will be checked using inode_permission() during
2083  *      path construction. -EACCES will be returned in case of failure.
2084  *   2. Path construction will be stopped at the inode number which corresponds
2085  *      to the fd with which this ioctl is called. If constructed path does not
2086  *      exist under fd's inode, -EACCES will be returned.
2087  *   3. The name of bottom subvolume is also searched and filled.
2088  */
2089 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2090 {
2091         struct btrfs_ioctl_ino_lookup_user_args *args;
2092         struct inode *inode;
2093         int ret;
2094
2095         args = memdup_user(argp, sizeof(*args));
2096         if (IS_ERR(args))
2097                 return PTR_ERR(args);
2098
2099         inode = file_inode(file);
2100
2101         if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2102             BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2103                 /*
2104                  * The subvolume does not exist under fd with which this is
2105                  * called
2106                  */
2107                 kfree(args);
2108                 return -EACCES;
2109         }
2110
2111         ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2112
2113         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2114                 ret = -EFAULT;
2115
2116         kfree(args);
2117         return ret;
2118 }
2119
2120 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2121 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2122 {
2123         struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2124         struct btrfs_fs_info *fs_info;
2125         struct btrfs_root *root;
2126         struct btrfs_path *path;
2127         struct btrfs_key key;
2128         struct btrfs_root_item *root_item;
2129         struct btrfs_root_ref *rref;
2130         struct extent_buffer *leaf;
2131         unsigned long item_off;
2132         unsigned long item_len;
2133         int slot;
2134         int ret = 0;
2135
2136         path = btrfs_alloc_path();
2137         if (!path)
2138                 return -ENOMEM;
2139
2140         subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2141         if (!subvol_info) {
2142                 btrfs_free_path(path);
2143                 return -ENOMEM;
2144         }
2145
2146         fs_info = BTRFS_I(inode)->root->fs_info;
2147
2148         /* Get root_item of inode's subvolume */
2149         key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2150         root = btrfs_get_fs_root(fs_info, key.objectid, true);
2151         if (IS_ERR(root)) {
2152                 ret = PTR_ERR(root);
2153                 goto out_free;
2154         }
2155         root_item = &root->root_item;
2156
2157         subvol_info->treeid = key.objectid;
2158
2159         subvol_info->generation = btrfs_root_generation(root_item);
2160         subvol_info->flags = btrfs_root_flags(root_item);
2161
2162         memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2163         memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2164                                                     BTRFS_UUID_SIZE);
2165         memcpy(subvol_info->received_uuid, root_item->received_uuid,
2166                                                     BTRFS_UUID_SIZE);
2167
2168         subvol_info->ctransid = btrfs_root_ctransid(root_item);
2169         subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2170         subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2171
2172         subvol_info->otransid = btrfs_root_otransid(root_item);
2173         subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2174         subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2175
2176         subvol_info->stransid = btrfs_root_stransid(root_item);
2177         subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2178         subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2179
2180         subvol_info->rtransid = btrfs_root_rtransid(root_item);
2181         subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2182         subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2183
2184         if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2185                 /* Search root tree for ROOT_BACKREF of this subvolume */
2186                 key.type = BTRFS_ROOT_BACKREF_KEY;
2187                 key.offset = 0;
2188                 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2189                 if (ret < 0) {
2190                         goto out;
2191                 } else if (path->slots[0] >=
2192                            btrfs_header_nritems(path->nodes[0])) {
2193                         ret = btrfs_next_leaf(fs_info->tree_root, path);
2194                         if (ret < 0) {
2195                                 goto out;
2196                         } else if (ret > 0) {
2197                                 ret = -EUCLEAN;
2198                                 goto out;
2199                         }
2200                 }
2201
2202                 leaf = path->nodes[0];
2203                 slot = path->slots[0];
2204                 btrfs_item_key_to_cpu(leaf, &key, slot);
2205                 if (key.objectid == subvol_info->treeid &&
2206                     key.type == BTRFS_ROOT_BACKREF_KEY) {
2207                         subvol_info->parent_id = key.offset;
2208
2209                         rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2210                         subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2211
2212                         item_off = btrfs_item_ptr_offset(leaf, slot)
2213                                         + sizeof(struct btrfs_root_ref);
2214                         item_len = btrfs_item_size(leaf, slot)
2215                                         - sizeof(struct btrfs_root_ref);
2216                         read_extent_buffer(leaf, subvol_info->name,
2217                                            item_off, item_len);
2218                 } else {
2219                         ret = -ENOENT;
2220                         goto out;
2221                 }
2222         }
2223
2224         btrfs_free_path(path);
2225         path = NULL;
2226         if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2227                 ret = -EFAULT;
2228
2229 out:
2230         btrfs_put_root(root);
2231 out_free:
2232         btrfs_free_path(path);
2233         kfree(subvol_info);
2234         return ret;
2235 }
2236
2237 /*
2238  * Return ROOT_REF information of the subvolume containing this inode
2239  * except the subvolume name.
2240  */
2241 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2242                                           void __user *argp)
2243 {
2244         struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2245         struct btrfs_root_ref *rref;
2246         struct btrfs_path *path;
2247         struct btrfs_key key;
2248         struct extent_buffer *leaf;
2249         u64 objectid;
2250         int slot;
2251         int ret;
2252         u8 found;
2253
2254         path = btrfs_alloc_path();
2255         if (!path)
2256                 return -ENOMEM;
2257
2258         rootrefs = memdup_user(argp, sizeof(*rootrefs));
2259         if (IS_ERR(rootrefs)) {
2260                 btrfs_free_path(path);
2261                 return PTR_ERR(rootrefs);
2262         }
2263
2264         objectid = root->root_key.objectid;
2265         key.objectid = objectid;
2266         key.type = BTRFS_ROOT_REF_KEY;
2267         key.offset = rootrefs->min_treeid;
2268         found = 0;
2269
2270         root = root->fs_info->tree_root;
2271         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2272         if (ret < 0) {
2273                 goto out;
2274         } else if (path->slots[0] >=
2275                    btrfs_header_nritems(path->nodes[0])) {
2276                 ret = btrfs_next_leaf(root, path);
2277                 if (ret < 0) {
2278                         goto out;
2279                 } else if (ret > 0) {
2280                         ret = -EUCLEAN;
2281                         goto out;
2282                 }
2283         }
2284         while (1) {
2285                 leaf = path->nodes[0];
2286                 slot = path->slots[0];
2287
2288                 btrfs_item_key_to_cpu(leaf, &key, slot);
2289                 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2290                         ret = 0;
2291                         goto out;
2292                 }
2293
2294                 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2295                         ret = -EOVERFLOW;
2296                         goto out;
2297                 }
2298
2299                 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2300                 rootrefs->rootref[found].treeid = key.offset;
2301                 rootrefs->rootref[found].dirid =
2302                                   btrfs_root_ref_dirid(leaf, rref);
2303                 found++;
2304
2305                 ret = btrfs_next_item(root, path);
2306                 if (ret < 0) {
2307                         goto out;
2308                 } else if (ret > 0) {
2309                         ret = -EUCLEAN;
2310                         goto out;
2311                 }
2312         }
2313
2314 out:
2315         btrfs_free_path(path);
2316
2317         if (!ret || ret == -EOVERFLOW) {
2318                 rootrefs->num_items = found;
2319                 /* update min_treeid for next search */
2320                 if (found)
2321                         rootrefs->min_treeid =
2322                                 rootrefs->rootref[found - 1].treeid + 1;
2323                 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2324                         ret = -EFAULT;
2325         }
2326
2327         kfree(rootrefs);
2328
2329         return ret;
2330 }
2331
2332 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2333                                              void __user *arg,
2334                                              bool destroy_v2)
2335 {
2336         struct dentry *parent = file->f_path.dentry;
2337         struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2338         struct dentry *dentry;
2339         struct inode *dir = d_inode(parent);
2340         struct inode *inode;
2341         struct btrfs_root *root = BTRFS_I(dir)->root;
2342         struct btrfs_root *dest = NULL;
2343         struct btrfs_ioctl_vol_args *vol_args = NULL;
2344         struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2345         struct mnt_idmap *idmap = file_mnt_idmap(file);
2346         char *subvol_name, *subvol_name_ptr = NULL;
2347         int subvol_namelen;
2348         int err = 0;
2349         bool destroy_parent = false;
2350
2351         /* We don't support snapshots with extent tree v2 yet. */
2352         if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2353                 btrfs_err(fs_info,
2354                           "extent tree v2 doesn't support snapshot deletion yet");
2355                 return -EOPNOTSUPP;
2356         }
2357
2358         if (destroy_v2) {
2359                 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2360                 if (IS_ERR(vol_args2))
2361                         return PTR_ERR(vol_args2);
2362
2363                 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2364                         err = -EOPNOTSUPP;
2365                         goto out;
2366                 }
2367
2368                 /*
2369                  * If SPEC_BY_ID is not set, we are looking for the subvolume by
2370                  * name, same as v1 currently does.
2371                  */
2372                 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2373                         vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2374                         subvol_name = vol_args2->name;
2375
2376                         err = mnt_want_write_file(file);
2377                         if (err)
2378                                 goto out;
2379                 } else {
2380                         struct inode *old_dir;
2381
2382                         if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2383                                 err = -EINVAL;
2384                                 goto out;
2385                         }
2386
2387                         err = mnt_want_write_file(file);
2388                         if (err)
2389                                 goto out;
2390
2391                         dentry = btrfs_get_dentry(fs_info->sb,
2392                                         BTRFS_FIRST_FREE_OBJECTID,
2393                                         vol_args2->subvolid, 0);
2394                         if (IS_ERR(dentry)) {
2395                                 err = PTR_ERR(dentry);
2396                                 goto out_drop_write;
2397                         }
2398
2399                         /*
2400                          * Change the default parent since the subvolume being
2401                          * deleted can be outside of the current mount point.
2402                          */
2403                         parent = btrfs_get_parent(dentry);
2404
2405                         /*
2406                          * At this point dentry->d_name can point to '/' if the
2407                          * subvolume we want to destroy is outsite of the
2408                          * current mount point, so we need to release the
2409                          * current dentry and execute the lookup to return a new
2410                          * one with ->d_name pointing to the
2411                          * <mount point>/subvol_name.
2412                          */
2413                         dput(dentry);
2414                         if (IS_ERR(parent)) {
2415                                 err = PTR_ERR(parent);
2416                                 goto out_drop_write;
2417                         }
2418                         old_dir = dir;
2419                         dir = d_inode(parent);
2420
2421                         /*
2422                          * If v2 was used with SPEC_BY_ID, a new parent was
2423                          * allocated since the subvolume can be outside of the
2424                          * current mount point. Later on we need to release this
2425                          * new parent dentry.
2426                          */
2427                         destroy_parent = true;
2428
2429                         /*
2430                          * On idmapped mounts, deletion via subvolid is
2431                          * restricted to subvolumes that are immediate
2432                          * ancestors of the inode referenced by the file
2433                          * descriptor in the ioctl. Otherwise the idmapping
2434                          * could potentially be abused to delete subvolumes
2435                          * anywhere in the filesystem the user wouldn't be able
2436                          * to delete without an idmapped mount.
2437                          */
2438                         if (old_dir != dir && idmap != &nop_mnt_idmap) {
2439                                 err = -EOPNOTSUPP;
2440                                 goto free_parent;
2441                         }
2442
2443                         subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2444                                                 fs_info, vol_args2->subvolid);
2445                         if (IS_ERR(subvol_name_ptr)) {
2446                                 err = PTR_ERR(subvol_name_ptr);
2447                                 goto free_parent;
2448                         }
2449                         /* subvol_name_ptr is already nul terminated */
2450                         subvol_name = (char *)kbasename(subvol_name_ptr);
2451                 }
2452         } else {
2453                 vol_args = memdup_user(arg, sizeof(*vol_args));
2454                 if (IS_ERR(vol_args))
2455                         return PTR_ERR(vol_args);
2456
2457                 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2458                 subvol_name = vol_args->name;
2459
2460                 err = mnt_want_write_file(file);
2461                 if (err)
2462                         goto out;
2463         }
2464
2465         subvol_namelen = strlen(subvol_name);
2466
2467         if (strchr(subvol_name, '/') ||
2468             strncmp(subvol_name, "..", subvol_namelen) == 0) {
2469                 err = -EINVAL;
2470                 goto free_subvol_name;
2471         }
2472
2473         if (!S_ISDIR(dir->i_mode)) {
2474                 err = -ENOTDIR;
2475                 goto free_subvol_name;
2476         }
2477
2478         err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2479         if (err == -EINTR)
2480                 goto free_subvol_name;
2481         dentry = lookup_one(idmap, subvol_name, parent, subvol_namelen);
2482         if (IS_ERR(dentry)) {
2483                 err = PTR_ERR(dentry);
2484                 goto out_unlock_dir;
2485         }
2486
2487         if (d_really_is_negative(dentry)) {
2488                 err = -ENOENT;
2489                 goto out_dput;
2490         }
2491
2492         inode = d_inode(dentry);
2493         dest = BTRFS_I(inode)->root;
2494         if (!capable(CAP_SYS_ADMIN)) {
2495                 /*
2496                  * Regular user.  Only allow this with a special mount
2497                  * option, when the user has write+exec access to the
2498                  * subvol root, and when rmdir(2) would have been
2499                  * allowed.
2500                  *
2501                  * Note that this is _not_ check that the subvol is
2502                  * empty or doesn't contain data that we wouldn't
2503                  * otherwise be able to delete.
2504                  *
2505                  * Users who want to delete empty subvols should try
2506                  * rmdir(2).
2507                  */
2508                 err = -EPERM;
2509                 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2510                         goto out_dput;
2511
2512                 /*
2513                  * Do not allow deletion if the parent dir is the same
2514                  * as the dir to be deleted.  That means the ioctl
2515                  * must be called on the dentry referencing the root
2516                  * of the subvol, not a random directory contained
2517                  * within it.
2518                  */
2519                 err = -EINVAL;
2520                 if (root == dest)
2521                         goto out_dput;
2522
2523                 err = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2524                 if (err)
2525                         goto out_dput;
2526         }
2527
2528         /* check if subvolume may be deleted by a user */
2529         err = btrfs_may_delete(idmap, dir, dentry, 1);
2530         if (err)
2531                 goto out_dput;
2532
2533         if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2534                 err = -EINVAL;
2535                 goto out_dput;
2536         }
2537
2538         btrfs_inode_lock(BTRFS_I(inode), 0);
2539         err = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2540         btrfs_inode_unlock(BTRFS_I(inode), 0);
2541         if (!err)
2542                 d_delete_notify(dir, dentry);
2543
2544 out_dput:
2545         dput(dentry);
2546 out_unlock_dir:
2547         btrfs_inode_unlock(BTRFS_I(dir), 0);
2548 free_subvol_name:
2549         kfree(subvol_name_ptr);
2550 free_parent:
2551         if (destroy_parent)
2552                 dput(parent);
2553 out_drop_write:
2554         mnt_drop_write_file(file);
2555 out:
2556         kfree(vol_args2);
2557         kfree(vol_args);
2558         return err;
2559 }
2560
2561 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2562 {
2563         struct inode *inode = file_inode(file);
2564         struct btrfs_root *root = BTRFS_I(inode)->root;
2565         struct btrfs_ioctl_defrag_range_args range = {0};
2566         int ret;
2567
2568         ret = mnt_want_write_file(file);
2569         if (ret)
2570                 return ret;
2571
2572         if (btrfs_root_readonly(root)) {
2573                 ret = -EROFS;
2574                 goto out;
2575         }
2576
2577         switch (inode->i_mode & S_IFMT) {
2578         case S_IFDIR:
2579                 if (!capable(CAP_SYS_ADMIN)) {
2580                         ret = -EPERM;
2581                         goto out;
2582                 }
2583                 ret = btrfs_defrag_root(root);
2584                 break;
2585         case S_IFREG:
2586                 /*
2587                  * Note that this does not check the file descriptor for write
2588                  * access. This prevents defragmenting executables that are
2589                  * running and allows defrag on files open in read-only mode.
2590                  */
2591                 if (!capable(CAP_SYS_ADMIN) &&
2592                     inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2593                         ret = -EPERM;
2594                         goto out;
2595                 }
2596
2597                 if (argp) {
2598                         if (copy_from_user(&range, argp, sizeof(range))) {
2599                                 ret = -EFAULT;
2600                                 goto out;
2601                         }
2602                         /* compression requires us to start the IO */
2603                         if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2604                                 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2605                                 range.extent_thresh = (u32)-1;
2606                         }
2607                 } else {
2608                         /* the rest are all set to zero by kzalloc */
2609                         range.len = (u64)-1;
2610                 }
2611                 ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2612                                         &range, BTRFS_OLDEST_GENERATION, 0);
2613                 if (ret > 0)
2614                         ret = 0;
2615                 break;
2616         default:
2617                 ret = -EINVAL;
2618         }
2619 out:
2620         mnt_drop_write_file(file);
2621         return ret;
2622 }
2623
2624 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2625 {
2626         struct btrfs_ioctl_vol_args *vol_args;
2627         bool restore_op = false;
2628         int ret;
2629
2630         if (!capable(CAP_SYS_ADMIN))
2631                 return -EPERM;
2632
2633         if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2634                 btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2635                 return -EINVAL;
2636         }
2637
2638         if (fs_info->fs_devices->temp_fsid) {
2639                 btrfs_err(fs_info,
2640                           "device add not supported on cloned temp-fsid mount");
2641                 return -EINVAL;
2642         }
2643
2644         if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2645                 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2646                         return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2647
2648                 /*
2649                  * We can do the device add because we have a paused balanced,
2650                  * change the exclusive op type and remember we should bring
2651                  * back the paused balance
2652                  */
2653                 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2654                 btrfs_exclop_start_unlock(fs_info);
2655                 restore_op = true;
2656         }
2657
2658         vol_args = memdup_user(arg, sizeof(*vol_args));
2659         if (IS_ERR(vol_args)) {
2660                 ret = PTR_ERR(vol_args);
2661                 goto out;
2662         }
2663
2664         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2665         ret = btrfs_init_new_device(fs_info, vol_args->name);
2666
2667         if (!ret)
2668                 btrfs_info(fs_info, "disk added %s", vol_args->name);
2669
2670         kfree(vol_args);
2671 out:
2672         if (restore_op)
2673                 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2674         else
2675                 btrfs_exclop_finish(fs_info);
2676         return ret;
2677 }
2678
2679 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2680 {
2681         BTRFS_DEV_LOOKUP_ARGS(args);
2682         struct inode *inode = file_inode(file);
2683         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2684         struct btrfs_ioctl_vol_args_v2 *vol_args;
2685         struct block_device *bdev = NULL;
2686         void *holder;
2687         int ret;
2688         bool cancel = false;
2689
2690         if (!capable(CAP_SYS_ADMIN))
2691                 return -EPERM;
2692
2693         vol_args = memdup_user(arg, sizeof(*vol_args));
2694         if (IS_ERR(vol_args))
2695                 return PTR_ERR(vol_args);
2696
2697         if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2698                 ret = -EOPNOTSUPP;
2699                 goto out;
2700         }
2701
2702         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2703         if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2704                 args.devid = vol_args->devid;
2705         } else if (!strcmp("cancel", vol_args->name)) {
2706                 cancel = true;
2707         } else {
2708                 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2709                 if (ret)
2710                         goto out;
2711         }
2712
2713         ret = mnt_want_write_file(file);
2714         if (ret)
2715                 goto out;
2716
2717         ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2718                                            cancel);
2719         if (ret)
2720                 goto err_drop;
2721
2722         /* Exclusive operation is now claimed */
2723         ret = btrfs_rm_device(fs_info, &args, &bdev, &holder);
2724
2725         btrfs_exclop_finish(fs_info);
2726
2727         if (!ret) {
2728                 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2729                         btrfs_info(fs_info, "device deleted: id %llu",
2730                                         vol_args->devid);
2731                 else
2732                         btrfs_info(fs_info, "device deleted: %s",
2733                                         vol_args->name);
2734         }
2735 err_drop:
2736         mnt_drop_write_file(file);
2737         if (bdev)
2738                 blkdev_put(bdev, holder);
2739 out:
2740         btrfs_put_dev_args_from_path(&args);
2741         kfree(vol_args);
2742         return ret;
2743 }
2744
2745 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2746 {
2747         BTRFS_DEV_LOOKUP_ARGS(args);
2748         struct inode *inode = file_inode(file);
2749         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2750         struct btrfs_ioctl_vol_args *vol_args;
2751         struct block_device *bdev = NULL;
2752         void *holder;
2753         int ret;
2754         bool cancel = false;
2755
2756         if (!capable(CAP_SYS_ADMIN))
2757                 return -EPERM;
2758
2759         vol_args = memdup_user(arg, sizeof(*vol_args));
2760         if (IS_ERR(vol_args))
2761                 return PTR_ERR(vol_args);
2762
2763         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2764         if (!strcmp("cancel", vol_args->name)) {
2765                 cancel = true;
2766         } else {
2767                 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2768                 if (ret)
2769                         goto out;
2770         }
2771
2772         ret = mnt_want_write_file(file);
2773         if (ret)
2774                 goto out;
2775
2776         ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2777                                            cancel);
2778         if (ret == 0) {
2779                 ret = btrfs_rm_device(fs_info, &args, &bdev, &holder);
2780                 if (!ret)
2781                         btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2782                 btrfs_exclop_finish(fs_info);
2783         }
2784
2785         mnt_drop_write_file(file);
2786         if (bdev)
2787                 blkdev_put(bdev, holder);
2788 out:
2789         btrfs_put_dev_args_from_path(&args);
2790         kfree(vol_args);
2791         return ret;
2792 }
2793
2794 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2795                                 void __user *arg)
2796 {
2797         struct btrfs_ioctl_fs_info_args *fi_args;
2798         struct btrfs_device *device;
2799         struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2800         u64 flags_in;
2801         int ret = 0;
2802
2803         fi_args = memdup_user(arg, sizeof(*fi_args));
2804         if (IS_ERR(fi_args))
2805                 return PTR_ERR(fi_args);
2806
2807         flags_in = fi_args->flags;
2808         memset(fi_args, 0, sizeof(*fi_args));
2809
2810         rcu_read_lock();
2811         fi_args->num_devices = fs_devices->num_devices;
2812
2813         list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2814                 if (device->devid > fi_args->max_id)
2815                         fi_args->max_id = device->devid;
2816         }
2817         rcu_read_unlock();
2818
2819         memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2820         fi_args->nodesize = fs_info->nodesize;
2821         fi_args->sectorsize = fs_info->sectorsize;
2822         fi_args->clone_alignment = fs_info->sectorsize;
2823
2824         if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2825                 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2826                 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2827                 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2828         }
2829
2830         if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2831                 fi_args->generation = btrfs_get_fs_generation(fs_info);
2832                 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2833         }
2834
2835         if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2836                 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2837                        sizeof(fi_args->metadata_uuid));
2838                 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2839         }
2840
2841         if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2842                 ret = -EFAULT;
2843
2844         kfree(fi_args);
2845         return ret;
2846 }
2847
2848 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2849                                  void __user *arg)
2850 {
2851         BTRFS_DEV_LOOKUP_ARGS(args);
2852         struct btrfs_ioctl_dev_info_args *di_args;
2853         struct btrfs_device *dev;
2854         int ret = 0;
2855
2856         di_args = memdup_user(arg, sizeof(*di_args));
2857         if (IS_ERR(di_args))
2858                 return PTR_ERR(di_args);
2859
2860         args.devid = di_args->devid;
2861         if (!btrfs_is_empty_uuid(di_args->uuid))
2862                 args.uuid = di_args->uuid;
2863
2864         rcu_read_lock();
2865         dev = btrfs_find_device(fs_info->fs_devices, &args);
2866         if (!dev) {
2867                 ret = -ENODEV;
2868                 goto out;
2869         }
2870
2871         di_args->devid = dev->devid;
2872         di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2873         di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2874         memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2875         memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2876         if (dev->name)
2877                 strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2878         else
2879                 di_args->path[0] = '\0';
2880
2881 out:
2882         rcu_read_unlock();
2883         if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2884                 ret = -EFAULT;
2885
2886         kfree(di_args);
2887         return ret;
2888 }
2889
2890 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2891 {
2892         struct inode *inode = file_inode(file);
2893         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2894         struct btrfs_root *root = BTRFS_I(inode)->root;
2895         struct btrfs_root *new_root;
2896         struct btrfs_dir_item *di;
2897         struct btrfs_trans_handle *trans;
2898         struct btrfs_path *path = NULL;
2899         struct btrfs_disk_key disk_key;
2900         struct fscrypt_str name = FSTR_INIT("default", 7);
2901         u64 objectid = 0;
2902         u64 dir_id;
2903         int ret;
2904
2905         if (!capable(CAP_SYS_ADMIN))
2906                 return -EPERM;
2907
2908         ret = mnt_want_write_file(file);
2909         if (ret)
2910                 return ret;
2911
2912         if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2913                 ret = -EFAULT;
2914                 goto out;
2915         }
2916
2917         if (!objectid)
2918                 objectid = BTRFS_FS_TREE_OBJECTID;
2919
2920         new_root = btrfs_get_fs_root(fs_info, objectid, true);
2921         if (IS_ERR(new_root)) {
2922                 ret = PTR_ERR(new_root);
2923                 goto out;
2924         }
2925         if (!is_fstree(new_root->root_key.objectid)) {
2926                 ret = -ENOENT;
2927                 goto out_free;
2928         }
2929
2930         path = btrfs_alloc_path();
2931         if (!path) {
2932                 ret = -ENOMEM;
2933                 goto out_free;
2934         }
2935
2936         trans = btrfs_start_transaction(root, 1);
2937         if (IS_ERR(trans)) {
2938                 ret = PTR_ERR(trans);
2939                 goto out_free;
2940         }
2941
2942         dir_id = btrfs_super_root_dir(fs_info->super_copy);
2943         di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2944                                    dir_id, &name, 1);
2945         if (IS_ERR_OR_NULL(di)) {
2946                 btrfs_release_path(path);
2947                 btrfs_end_transaction(trans);
2948                 btrfs_err(fs_info,
2949                           "Umm, you don't have the default diritem, this isn't going to work");
2950                 ret = -ENOENT;
2951                 goto out_free;
2952         }
2953
2954         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2955         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2956         btrfs_mark_buffer_dirty(trans, path->nodes[0]);
2957         btrfs_release_path(path);
2958
2959         btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
2960         btrfs_end_transaction(trans);
2961 out_free:
2962         btrfs_put_root(new_root);
2963         btrfs_free_path(path);
2964 out:
2965         mnt_drop_write_file(file);
2966         return ret;
2967 }
2968
2969 static void get_block_group_info(struct list_head *groups_list,
2970                                  struct btrfs_ioctl_space_info *space)
2971 {
2972         struct btrfs_block_group *block_group;
2973
2974         space->total_bytes = 0;
2975         space->used_bytes = 0;
2976         space->flags = 0;
2977         list_for_each_entry(block_group, groups_list, list) {
2978                 space->flags = block_group->flags;
2979                 space->total_bytes += block_group->length;
2980                 space->used_bytes += block_group->used;
2981         }
2982 }
2983
2984 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
2985                                    void __user *arg)
2986 {
2987         struct btrfs_ioctl_space_args space_args = { 0 };
2988         struct btrfs_ioctl_space_info space;
2989         struct btrfs_ioctl_space_info *dest;
2990         struct btrfs_ioctl_space_info *dest_orig;
2991         struct btrfs_ioctl_space_info __user *user_dest;
2992         struct btrfs_space_info *info;
2993         static const u64 types[] = {
2994                 BTRFS_BLOCK_GROUP_DATA,
2995                 BTRFS_BLOCK_GROUP_SYSTEM,
2996                 BTRFS_BLOCK_GROUP_METADATA,
2997                 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
2998         };
2999         int num_types = 4;
3000         int alloc_size;
3001         int ret = 0;
3002         u64 slot_count = 0;
3003         int i, c;
3004
3005         if (copy_from_user(&space_args,
3006                            (struct btrfs_ioctl_space_args __user *)arg,
3007                            sizeof(space_args)))
3008                 return -EFAULT;
3009
3010         for (i = 0; i < num_types; i++) {
3011                 struct btrfs_space_info *tmp;
3012
3013                 info = NULL;
3014                 list_for_each_entry(tmp, &fs_info->space_info, list) {
3015                         if (tmp->flags == types[i]) {
3016                                 info = tmp;
3017                                 break;
3018                         }
3019                 }
3020
3021                 if (!info)
3022                         continue;
3023
3024                 down_read(&info->groups_sem);
3025                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3026                         if (!list_empty(&info->block_groups[c]))
3027                                 slot_count++;
3028                 }
3029                 up_read(&info->groups_sem);
3030         }
3031
3032         /*
3033          * Global block reserve, exported as a space_info
3034          */
3035         slot_count++;
3036
3037         /* space_slots == 0 means they are asking for a count */
3038         if (space_args.space_slots == 0) {
3039                 space_args.total_spaces = slot_count;
3040                 goto out;
3041         }
3042
3043         slot_count = min_t(u64, space_args.space_slots, slot_count);
3044
3045         alloc_size = sizeof(*dest) * slot_count;
3046
3047         /* we generally have at most 6 or so space infos, one for each raid
3048          * level.  So, a whole page should be more than enough for everyone
3049          */
3050         if (alloc_size > PAGE_SIZE)
3051                 return -ENOMEM;
3052
3053         space_args.total_spaces = 0;
3054         dest = kmalloc(alloc_size, GFP_KERNEL);
3055         if (!dest)
3056                 return -ENOMEM;
3057         dest_orig = dest;
3058
3059         /* now we have a buffer to copy into */
3060         for (i = 0; i < num_types; i++) {
3061                 struct btrfs_space_info *tmp;
3062
3063                 if (!slot_count)
3064                         break;
3065
3066                 info = NULL;
3067                 list_for_each_entry(tmp, &fs_info->space_info, list) {
3068                         if (tmp->flags == types[i]) {
3069                                 info = tmp;
3070                                 break;
3071                         }
3072                 }
3073
3074                 if (!info)
3075                         continue;
3076                 down_read(&info->groups_sem);
3077                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3078                         if (!list_empty(&info->block_groups[c])) {
3079                                 get_block_group_info(&info->block_groups[c],
3080                                                      &space);
3081                                 memcpy(dest, &space, sizeof(space));
3082                                 dest++;
3083                                 space_args.total_spaces++;
3084                                 slot_count--;
3085                         }
3086                         if (!slot_count)
3087                                 break;
3088                 }
3089                 up_read(&info->groups_sem);
3090         }
3091
3092         /*
3093          * Add global block reserve
3094          */
3095         if (slot_count) {
3096                 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3097
3098                 spin_lock(&block_rsv->lock);
3099                 space.total_bytes = block_rsv->size;
3100                 space.used_bytes = block_rsv->size - block_rsv->reserved;
3101                 spin_unlock(&block_rsv->lock);
3102                 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3103                 memcpy(dest, &space, sizeof(space));
3104                 space_args.total_spaces++;
3105         }
3106
3107         user_dest = (struct btrfs_ioctl_space_info __user *)
3108                 (arg + sizeof(struct btrfs_ioctl_space_args));
3109
3110         if (copy_to_user(user_dest, dest_orig, alloc_size))
3111                 ret = -EFAULT;
3112
3113         kfree(dest_orig);
3114 out:
3115         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3116                 ret = -EFAULT;
3117
3118         return ret;
3119 }
3120
3121 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3122                                             void __user *argp)
3123 {
3124         struct btrfs_trans_handle *trans;
3125         u64 transid;
3126
3127         /*
3128          * Start orphan cleanup here for the given root in case it hasn't been
3129          * started already by other means. Errors are handled in the other
3130          * functions during transaction commit.
3131          */
3132         btrfs_orphan_cleanup(root);
3133
3134         trans = btrfs_attach_transaction_barrier(root);
3135         if (IS_ERR(trans)) {
3136                 if (PTR_ERR(trans) != -ENOENT)
3137                         return PTR_ERR(trans);
3138
3139                 /* No running transaction, don't bother */
3140                 transid = btrfs_get_last_trans_committed(root->fs_info);
3141                 goto out;
3142         }
3143         transid = trans->transid;
3144         btrfs_commit_transaction_async(trans);
3145 out:
3146         if (argp)
3147                 if (copy_to_user(argp, &transid, sizeof(transid)))
3148                         return -EFAULT;
3149         return 0;
3150 }
3151
3152 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3153                                            void __user *argp)
3154 {
3155         /* By default wait for the current transaction. */
3156         u64 transid = 0;
3157
3158         if (argp)
3159                 if (copy_from_user(&transid, argp, sizeof(transid)))
3160                         return -EFAULT;
3161
3162         return btrfs_wait_for_commit(fs_info, transid);
3163 }
3164
3165 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3166 {
3167         struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3168         struct btrfs_ioctl_scrub_args *sa;
3169         int ret;
3170
3171         if (!capable(CAP_SYS_ADMIN))
3172                 return -EPERM;
3173
3174         if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3175                 btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3176                 return -EINVAL;
3177         }
3178
3179         sa = memdup_user(arg, sizeof(*sa));
3180         if (IS_ERR(sa))
3181                 return PTR_ERR(sa);
3182
3183         if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3184                 ret = -EOPNOTSUPP;
3185                 goto out;
3186         }
3187
3188         if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3189                 ret = mnt_want_write_file(file);
3190                 if (ret)
3191                         goto out;
3192         }
3193
3194         ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3195                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3196                               0);
3197
3198         /*
3199          * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3200          * error. This is important as it allows user space to know how much
3201          * progress scrub has done. For example, if scrub is canceled we get
3202          * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3203          * space. Later user space can inspect the progress from the structure
3204          * btrfs_ioctl_scrub_args and resume scrub from where it left off
3205          * previously (btrfs-progs does this).
3206          * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3207          * then return -EFAULT to signal the structure was not copied or it may
3208          * be corrupt and unreliable due to a partial copy.
3209          */
3210         if (copy_to_user(arg, sa, sizeof(*sa)))
3211                 ret = -EFAULT;
3212
3213         if (!(sa->flags & BTRFS_SCRUB_READONLY))
3214                 mnt_drop_write_file(file);
3215 out:
3216         kfree(sa);
3217         return ret;
3218 }
3219
3220 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3221 {
3222         if (!capable(CAP_SYS_ADMIN))
3223                 return -EPERM;
3224
3225         return btrfs_scrub_cancel(fs_info);
3226 }
3227
3228 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3229                                        void __user *arg)
3230 {
3231         struct btrfs_ioctl_scrub_args *sa;
3232         int ret;
3233
3234         if (!capable(CAP_SYS_ADMIN))
3235                 return -EPERM;
3236
3237         sa = memdup_user(arg, sizeof(*sa));
3238         if (IS_ERR(sa))
3239                 return PTR_ERR(sa);
3240
3241         ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3242
3243         if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3244                 ret = -EFAULT;
3245
3246         kfree(sa);
3247         return ret;
3248 }
3249
3250 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3251                                       void __user *arg)
3252 {
3253         struct btrfs_ioctl_get_dev_stats *sa;
3254         int ret;
3255
3256         sa = memdup_user(arg, sizeof(*sa));
3257         if (IS_ERR(sa))
3258                 return PTR_ERR(sa);
3259
3260         if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3261                 kfree(sa);
3262                 return -EPERM;
3263         }
3264
3265         ret = btrfs_get_dev_stats(fs_info, sa);
3266
3267         if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3268                 ret = -EFAULT;
3269
3270         kfree(sa);
3271         return ret;
3272 }
3273
3274 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3275                                     void __user *arg)
3276 {
3277         struct btrfs_ioctl_dev_replace_args *p;
3278         int ret;
3279
3280         if (!capable(CAP_SYS_ADMIN))
3281                 return -EPERM;
3282
3283         if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3284                 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3285                 return -EINVAL;
3286         }
3287
3288         p = memdup_user(arg, sizeof(*p));
3289         if (IS_ERR(p))
3290                 return PTR_ERR(p);
3291
3292         switch (p->cmd) {
3293         case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3294                 if (sb_rdonly(fs_info->sb)) {
3295                         ret = -EROFS;
3296                         goto out;
3297                 }
3298                 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3299                         ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3300                 } else {
3301                         ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3302                         btrfs_exclop_finish(fs_info);
3303                 }
3304                 break;
3305         case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3306                 btrfs_dev_replace_status(fs_info, p);
3307                 ret = 0;
3308                 break;
3309         case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3310                 p->result = btrfs_dev_replace_cancel(fs_info);
3311                 ret = 0;
3312                 break;
3313         default:
3314                 ret = -EINVAL;
3315                 break;
3316         }
3317
3318         if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3319                 ret = -EFAULT;
3320 out:
3321         kfree(p);
3322         return ret;
3323 }
3324
3325 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3326 {
3327         int ret = 0;
3328         int i;
3329         u64 rel_ptr;
3330         int size;
3331         struct btrfs_ioctl_ino_path_args *ipa = NULL;
3332         struct inode_fs_paths *ipath = NULL;
3333         struct btrfs_path *path;
3334
3335         if (!capable(CAP_DAC_READ_SEARCH))
3336                 return -EPERM;
3337
3338         path = btrfs_alloc_path();
3339         if (!path) {
3340                 ret = -ENOMEM;
3341                 goto out;
3342         }
3343
3344         ipa = memdup_user(arg, sizeof(*ipa));
3345         if (IS_ERR(ipa)) {
3346                 ret = PTR_ERR(ipa);
3347                 ipa = NULL;
3348                 goto out;
3349         }
3350
3351         size = min_t(u32, ipa->size, 4096);
3352         ipath = init_ipath(size, root, path);
3353         if (IS_ERR(ipath)) {
3354                 ret = PTR_ERR(ipath);
3355                 ipath = NULL;
3356                 goto out;
3357         }
3358
3359         ret = paths_from_inode(ipa->inum, ipath);
3360         if (ret < 0)
3361                 goto out;
3362
3363         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3364                 rel_ptr = ipath->fspath->val[i] -
3365                           (u64)(unsigned long)ipath->fspath->val;
3366                 ipath->fspath->val[i] = rel_ptr;
3367         }
3368
3369         btrfs_free_path(path);
3370         path = NULL;
3371         ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3372                            ipath->fspath, size);
3373         if (ret) {
3374                 ret = -EFAULT;
3375                 goto out;
3376         }
3377
3378 out:
3379         btrfs_free_path(path);
3380         free_ipath(ipath);
3381         kfree(ipa);
3382
3383         return ret;
3384 }
3385
3386 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3387                                         void __user *arg, int version)
3388 {
3389         int ret = 0;
3390         int size;
3391         struct btrfs_ioctl_logical_ino_args *loi;
3392         struct btrfs_data_container *inodes = NULL;
3393         struct btrfs_path *path = NULL;
3394         bool ignore_offset;
3395
3396         if (!capable(CAP_SYS_ADMIN))
3397                 return -EPERM;
3398
3399         loi = memdup_user(arg, sizeof(*loi));
3400         if (IS_ERR(loi))
3401                 return PTR_ERR(loi);
3402
3403         if (version == 1) {
3404                 ignore_offset = false;
3405                 size = min_t(u32, loi->size, SZ_64K);
3406         } else {
3407                 /* All reserved bits must be 0 for now */
3408                 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3409                         ret = -EINVAL;
3410                         goto out_loi;
3411                 }
3412                 /* Only accept flags we have defined so far */
3413                 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3414                         ret = -EINVAL;
3415                         goto out_loi;
3416                 }
3417                 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3418                 size = min_t(u32, loi->size, SZ_16M);
3419         }
3420
3421         inodes = init_data_container(size);
3422         if (IS_ERR(inodes)) {
3423                 ret = PTR_ERR(inodes);
3424                 goto out_loi;
3425         }
3426
3427         path = btrfs_alloc_path();
3428         if (!path) {
3429                 ret = -ENOMEM;
3430                 goto out;
3431         }
3432         ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3433                                           inodes, ignore_offset);
3434         btrfs_free_path(path);
3435         if (ret == -EINVAL)
3436                 ret = -ENOENT;
3437         if (ret < 0)
3438                 goto out;
3439
3440         ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3441                            size);
3442         if (ret)
3443                 ret = -EFAULT;
3444
3445 out:
3446         kvfree(inodes);
3447 out_loi:
3448         kfree(loi);
3449
3450         return ret;
3451 }
3452
3453 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3454                                struct btrfs_ioctl_balance_args *bargs)
3455 {
3456         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3457
3458         bargs->flags = bctl->flags;
3459
3460         if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3461                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3462         if (atomic_read(&fs_info->balance_pause_req))
3463                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3464         if (atomic_read(&fs_info->balance_cancel_req))
3465                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3466
3467         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3468         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3469         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3470
3471         spin_lock(&fs_info->balance_lock);
3472         memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3473         spin_unlock(&fs_info->balance_lock);
3474 }
3475
3476 /*
3477  * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3478  * required.
3479  *
3480  * @fs_info:       the filesystem
3481  * @excl_acquired: ptr to boolean value which is set to false in case balance
3482  *                 is being resumed
3483  *
3484  * Return 0 on success in which case both fs_info::balance is acquired as well
3485  * as exclusive ops are blocked. In case of failure return an error code.
3486  */
3487 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3488 {
3489         int ret;
3490
3491         /*
3492          * Exclusive operation is locked. Three possibilities:
3493          *   (1) some other op is running
3494          *   (2) balance is running
3495          *   (3) balance is paused -- special case (think resume)
3496          */
3497         while (1) {
3498                 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3499                         *excl_acquired = true;
3500                         mutex_lock(&fs_info->balance_mutex);
3501                         return 0;
3502                 }
3503
3504                 mutex_lock(&fs_info->balance_mutex);
3505                 if (fs_info->balance_ctl) {
3506                         /* This is either (2) or (3) */
3507                         if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3508                                 /* This is (2) */
3509                                 ret = -EINPROGRESS;
3510                                 goto out_failure;
3511
3512                         } else {
3513                                 mutex_unlock(&fs_info->balance_mutex);
3514                                 /*
3515                                  * Lock released to allow other waiters to
3516                                  * continue, we'll reexamine the status again.
3517                                  */
3518                                 mutex_lock(&fs_info->balance_mutex);
3519
3520                                 if (fs_info->balance_ctl &&
3521                                     !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3522                                         /* This is (3) */
3523                                         *excl_acquired = false;
3524                                         return 0;
3525                                 }
3526                         }
3527                 } else {
3528                         /* This is (1) */
3529                         ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3530                         goto out_failure;
3531                 }
3532
3533                 mutex_unlock(&fs_info->balance_mutex);
3534         }
3535
3536 out_failure:
3537         mutex_unlock(&fs_info->balance_mutex);
3538         *excl_acquired = false;
3539         return ret;
3540 }
3541
3542 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3543 {
3544         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3545         struct btrfs_fs_info *fs_info = root->fs_info;
3546         struct btrfs_ioctl_balance_args *bargs;
3547         struct btrfs_balance_control *bctl;
3548         bool need_unlock = true;
3549         int ret;
3550
3551         if (!capable(CAP_SYS_ADMIN))
3552                 return -EPERM;
3553
3554         ret = mnt_want_write_file(file);
3555         if (ret)
3556                 return ret;
3557
3558         bargs = memdup_user(arg, sizeof(*bargs));
3559         if (IS_ERR(bargs)) {
3560                 ret = PTR_ERR(bargs);
3561                 bargs = NULL;
3562                 goto out;
3563         }
3564
3565         ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3566         if (ret)
3567                 goto out;
3568
3569         lockdep_assert_held(&fs_info->balance_mutex);
3570
3571         if (bargs->flags & BTRFS_BALANCE_RESUME) {
3572                 if (!fs_info->balance_ctl) {
3573                         ret = -ENOTCONN;
3574                         goto out_unlock;
3575                 }
3576
3577                 bctl = fs_info->balance_ctl;
3578                 spin_lock(&fs_info->balance_lock);
3579                 bctl->flags |= BTRFS_BALANCE_RESUME;
3580                 spin_unlock(&fs_info->balance_lock);
3581                 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3582
3583                 goto do_balance;
3584         }
3585
3586         if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3587                 ret = -EINVAL;
3588                 goto out_unlock;
3589         }
3590
3591         if (fs_info->balance_ctl) {
3592                 ret = -EINPROGRESS;
3593                 goto out_unlock;
3594         }
3595
3596         bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3597         if (!bctl) {
3598                 ret = -ENOMEM;
3599                 goto out_unlock;
3600         }
3601
3602         memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3603         memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3604         memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3605
3606         bctl->flags = bargs->flags;
3607 do_balance:
3608         /*
3609          * Ownership of bctl and exclusive operation goes to btrfs_balance.
3610          * bctl is freed in reset_balance_state, or, if restriper was paused
3611          * all the way until unmount, in free_fs_info.  The flag should be
3612          * cleared after reset_balance_state.
3613          */
3614         need_unlock = false;
3615
3616         ret = btrfs_balance(fs_info, bctl, bargs);
3617         bctl = NULL;
3618
3619         if (ret == 0 || ret == -ECANCELED) {
3620                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3621                         ret = -EFAULT;
3622         }
3623
3624         kfree(bctl);
3625 out_unlock:
3626         mutex_unlock(&fs_info->balance_mutex);
3627         if (need_unlock)
3628                 btrfs_exclop_finish(fs_info);
3629 out:
3630         mnt_drop_write_file(file);
3631         kfree(bargs);
3632         return ret;
3633 }
3634
3635 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3636 {
3637         if (!capable(CAP_SYS_ADMIN))
3638                 return -EPERM;
3639
3640         switch (cmd) {
3641         case BTRFS_BALANCE_CTL_PAUSE:
3642                 return btrfs_pause_balance(fs_info);
3643         case BTRFS_BALANCE_CTL_CANCEL:
3644                 return btrfs_cancel_balance(fs_info);
3645         }
3646
3647         return -EINVAL;
3648 }
3649
3650 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3651                                          void __user *arg)
3652 {
3653         struct btrfs_ioctl_balance_args *bargs;
3654         int ret = 0;
3655
3656         if (!capable(CAP_SYS_ADMIN))
3657                 return -EPERM;
3658
3659         mutex_lock(&fs_info->balance_mutex);
3660         if (!fs_info->balance_ctl) {
3661                 ret = -ENOTCONN;
3662                 goto out;
3663         }
3664
3665         bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3666         if (!bargs) {
3667                 ret = -ENOMEM;
3668                 goto out;
3669         }
3670
3671         btrfs_update_ioctl_balance_args(fs_info, bargs);
3672
3673         if (copy_to_user(arg, bargs, sizeof(*bargs)))
3674                 ret = -EFAULT;
3675
3676         kfree(bargs);
3677 out:
3678         mutex_unlock(&fs_info->balance_mutex);
3679         return ret;
3680 }
3681
3682 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3683 {
3684         struct inode *inode = file_inode(file);
3685         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3686         struct btrfs_ioctl_quota_ctl_args *sa;
3687         int ret;
3688
3689         if (!capable(CAP_SYS_ADMIN))
3690                 return -EPERM;
3691
3692         ret = mnt_want_write_file(file);
3693         if (ret)
3694                 return ret;
3695
3696         sa = memdup_user(arg, sizeof(*sa));
3697         if (IS_ERR(sa)) {
3698                 ret = PTR_ERR(sa);
3699                 goto drop_write;
3700         }
3701
3702         down_write(&fs_info->subvol_sem);
3703
3704         switch (sa->cmd) {
3705         case BTRFS_QUOTA_CTL_ENABLE:
3706         case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA:
3707                 ret = btrfs_quota_enable(fs_info, sa);
3708                 break;
3709         case BTRFS_QUOTA_CTL_DISABLE:
3710                 ret = btrfs_quota_disable(fs_info);
3711                 break;
3712         default:
3713                 ret = -EINVAL;
3714                 break;
3715         }
3716
3717         kfree(sa);
3718         up_write(&fs_info->subvol_sem);
3719 drop_write:
3720         mnt_drop_write_file(file);
3721         return ret;
3722 }
3723
3724 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3725 {
3726         struct inode *inode = file_inode(file);
3727         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3728         struct btrfs_root *root = BTRFS_I(inode)->root;
3729         struct btrfs_ioctl_qgroup_assign_args *sa;
3730         struct btrfs_trans_handle *trans;
3731         int ret;
3732         int err;
3733
3734         if (!capable(CAP_SYS_ADMIN))
3735                 return -EPERM;
3736
3737         ret = mnt_want_write_file(file);
3738         if (ret)
3739                 return ret;
3740
3741         sa = memdup_user(arg, sizeof(*sa));
3742         if (IS_ERR(sa)) {
3743                 ret = PTR_ERR(sa);
3744                 goto drop_write;
3745         }
3746
3747         trans = btrfs_join_transaction(root);
3748         if (IS_ERR(trans)) {
3749                 ret = PTR_ERR(trans);
3750                 goto out;
3751         }
3752
3753         if (sa->assign) {
3754                 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
3755         } else {
3756                 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3757         }
3758
3759         /* update qgroup status and info */
3760         mutex_lock(&fs_info->qgroup_ioctl_lock);
3761         err = btrfs_run_qgroups(trans);
3762         mutex_unlock(&fs_info->qgroup_ioctl_lock);
3763         if (err < 0)
3764                 btrfs_handle_fs_error(fs_info, err,
3765                                       "failed to update qgroup status and info");
3766         err = btrfs_end_transaction(trans);
3767         if (err && !ret)
3768                 ret = err;
3769
3770 out:
3771         kfree(sa);
3772 drop_write:
3773         mnt_drop_write_file(file);
3774         return ret;
3775 }
3776
3777 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3778 {
3779         struct inode *inode = file_inode(file);
3780         struct btrfs_root *root = BTRFS_I(inode)->root;
3781         struct btrfs_ioctl_qgroup_create_args *sa;
3782         struct btrfs_trans_handle *trans;
3783         int ret;
3784         int err;
3785
3786         if (!capable(CAP_SYS_ADMIN))
3787                 return -EPERM;
3788
3789         ret = mnt_want_write_file(file);
3790         if (ret)
3791                 return ret;
3792
3793         sa = memdup_user(arg, sizeof(*sa));
3794         if (IS_ERR(sa)) {
3795                 ret = PTR_ERR(sa);
3796                 goto drop_write;
3797         }
3798
3799         if (!sa->qgroupid) {
3800                 ret = -EINVAL;
3801                 goto out;
3802         }
3803
3804         trans = btrfs_join_transaction(root);
3805         if (IS_ERR(trans)) {
3806                 ret = PTR_ERR(trans);
3807                 goto out;
3808         }
3809
3810         if (sa->create) {
3811                 ret = btrfs_create_qgroup(trans, sa->qgroupid);
3812         } else {
3813                 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3814         }
3815
3816         err = btrfs_end_transaction(trans);
3817         if (err && !ret)
3818                 ret = err;
3819
3820 out:
3821         kfree(sa);
3822 drop_write:
3823         mnt_drop_write_file(file);
3824         return ret;
3825 }
3826
3827 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3828 {
3829         struct inode *inode = file_inode(file);
3830         struct btrfs_root *root = BTRFS_I(inode)->root;
3831         struct btrfs_ioctl_qgroup_limit_args *sa;
3832         struct btrfs_trans_handle *trans;
3833         int ret;
3834         int err;
3835         u64 qgroupid;
3836
3837         if (!capable(CAP_SYS_ADMIN))
3838                 return -EPERM;
3839
3840         ret = mnt_want_write_file(file);
3841         if (ret)
3842                 return ret;
3843
3844         sa = memdup_user(arg, sizeof(*sa));
3845         if (IS_ERR(sa)) {
3846                 ret = PTR_ERR(sa);
3847                 goto drop_write;
3848         }
3849
3850         trans = btrfs_join_transaction(root);
3851         if (IS_ERR(trans)) {
3852                 ret = PTR_ERR(trans);
3853                 goto out;
3854         }
3855
3856         qgroupid = sa->qgroupid;
3857         if (!qgroupid) {
3858                 /* take the current subvol as qgroup */
3859                 qgroupid = root->root_key.objectid;
3860         }
3861
3862         ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3863
3864         err = btrfs_end_transaction(trans);
3865         if (err && !ret)
3866                 ret = err;
3867
3868 out:
3869         kfree(sa);
3870 drop_write:
3871         mnt_drop_write_file(file);
3872         return ret;
3873 }
3874
3875 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3876 {
3877         struct inode *inode = file_inode(file);
3878         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3879         struct btrfs_ioctl_quota_rescan_args *qsa;
3880         int ret;
3881
3882         if (!capable(CAP_SYS_ADMIN))
3883                 return -EPERM;
3884
3885         ret = mnt_want_write_file(file);
3886         if (ret)
3887                 return ret;
3888
3889         qsa = memdup_user(arg, sizeof(*qsa));
3890         if (IS_ERR(qsa)) {
3891                 ret = PTR_ERR(qsa);
3892                 goto drop_write;
3893         }
3894
3895         if (qsa->flags) {
3896                 ret = -EINVAL;
3897                 goto out;
3898         }
3899
3900         ret = btrfs_qgroup_rescan(fs_info);
3901
3902 out:
3903         kfree(qsa);
3904 drop_write:
3905         mnt_drop_write_file(file);
3906         return ret;
3907 }
3908
3909 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
3910                                                 void __user *arg)
3911 {
3912         struct btrfs_ioctl_quota_rescan_args qsa = {0};
3913
3914         if (!capable(CAP_SYS_ADMIN))
3915                 return -EPERM;
3916
3917         if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3918                 qsa.flags = 1;
3919                 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
3920         }
3921
3922         if (copy_to_user(arg, &qsa, sizeof(qsa)))
3923                 return -EFAULT;
3924
3925         return 0;
3926 }
3927
3928 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
3929                                                 void __user *arg)
3930 {
3931         if (!capable(CAP_SYS_ADMIN))
3932                 return -EPERM;
3933
3934         return btrfs_qgroup_wait_for_completion(fs_info, true);
3935 }
3936
3937 static long _btrfs_ioctl_set_received_subvol(struct file *file,
3938                                             struct mnt_idmap *idmap,
3939                                             struct btrfs_ioctl_received_subvol_args *sa)
3940 {
3941         struct inode *inode = file_inode(file);
3942         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3943         struct btrfs_root *root = BTRFS_I(inode)->root;
3944         struct btrfs_root_item *root_item = &root->root_item;
3945         struct btrfs_trans_handle *trans;
3946         struct timespec64 ct = current_time(inode);
3947         int ret = 0;
3948         int received_uuid_changed;
3949
3950         if (!inode_owner_or_capable(idmap, inode))
3951                 return -EPERM;
3952
3953         ret = mnt_want_write_file(file);
3954         if (ret < 0)
3955                 return ret;
3956
3957         down_write(&fs_info->subvol_sem);
3958
3959         if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3960                 ret = -EINVAL;
3961                 goto out;
3962         }
3963
3964         if (btrfs_root_readonly(root)) {
3965                 ret = -EROFS;
3966                 goto out;
3967         }
3968
3969         /*
3970          * 1 - root item
3971          * 2 - uuid items (received uuid + subvol uuid)
3972          */
3973         trans = btrfs_start_transaction(root, 3);
3974         if (IS_ERR(trans)) {
3975                 ret = PTR_ERR(trans);
3976                 trans = NULL;
3977                 goto out;
3978         }
3979
3980         sa->rtransid = trans->transid;
3981         sa->rtime.sec = ct.tv_sec;
3982         sa->rtime.nsec = ct.tv_nsec;
3983
3984         received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
3985                                        BTRFS_UUID_SIZE);
3986         if (received_uuid_changed &&
3987             !btrfs_is_empty_uuid(root_item->received_uuid)) {
3988                 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
3989                                           BTRFS_UUID_KEY_RECEIVED_SUBVOL,
3990                                           root->root_key.objectid);
3991                 if (ret && ret != -ENOENT) {
3992                         btrfs_abort_transaction(trans, ret);
3993                         btrfs_end_transaction(trans);
3994                         goto out;
3995                 }
3996         }
3997         memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3998         btrfs_set_root_stransid(root_item, sa->stransid);
3999         btrfs_set_root_rtransid(root_item, sa->rtransid);
4000         btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4001         btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4002         btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4003         btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4004
4005         ret = btrfs_update_root(trans, fs_info->tree_root,
4006                                 &root->root_key, &root->root_item);
4007         if (ret < 0) {
4008                 btrfs_end_transaction(trans);
4009                 goto out;
4010         }
4011         if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4012                 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4013                                           BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4014                                           root->root_key.objectid);
4015                 if (ret < 0 && ret != -EEXIST) {
4016                         btrfs_abort_transaction(trans, ret);
4017                         btrfs_end_transaction(trans);
4018                         goto out;
4019                 }
4020         }
4021         ret = btrfs_commit_transaction(trans);
4022 out:
4023         up_write(&fs_info->subvol_sem);
4024         mnt_drop_write_file(file);
4025         return ret;
4026 }
4027
4028 #ifdef CONFIG_64BIT
4029 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4030                                                 void __user *arg)
4031 {
4032         struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4033         struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4034         int ret = 0;
4035
4036         args32 = memdup_user(arg, sizeof(*args32));
4037         if (IS_ERR(args32))
4038                 return PTR_ERR(args32);
4039
4040         args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4041         if (!args64) {
4042                 ret = -ENOMEM;
4043                 goto out;
4044         }
4045
4046         memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4047         args64->stransid = args32->stransid;
4048         args64->rtransid = args32->rtransid;
4049         args64->stime.sec = args32->stime.sec;
4050         args64->stime.nsec = args32->stime.nsec;
4051         args64->rtime.sec = args32->rtime.sec;
4052         args64->rtime.nsec = args32->rtime.nsec;
4053         args64->flags = args32->flags;
4054
4055         ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4056         if (ret)
4057                 goto out;
4058
4059         memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4060         args32->stransid = args64->stransid;
4061         args32->rtransid = args64->rtransid;
4062         args32->stime.sec = args64->stime.sec;
4063         args32->stime.nsec = args64->stime.nsec;
4064         args32->rtime.sec = args64->rtime.sec;
4065         args32->rtime.nsec = args64->rtime.nsec;
4066         args32->flags = args64->flags;
4067
4068         ret = copy_to_user(arg, args32, sizeof(*args32));
4069         if (ret)
4070                 ret = -EFAULT;
4071
4072 out:
4073         kfree(args32);
4074         kfree(args64);
4075         return ret;
4076 }
4077 #endif
4078
4079 static long btrfs_ioctl_set_received_subvol(struct file *file,
4080                                             void __user *arg)
4081 {
4082         struct btrfs_ioctl_received_subvol_args *sa = NULL;
4083         int ret = 0;
4084
4085         sa = memdup_user(arg, sizeof(*sa));
4086         if (IS_ERR(sa))
4087                 return PTR_ERR(sa);
4088
4089         ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4090
4091         if (ret)
4092                 goto out;
4093
4094         ret = copy_to_user(arg, sa, sizeof(*sa));
4095         if (ret)
4096                 ret = -EFAULT;
4097
4098 out:
4099         kfree(sa);
4100         return ret;
4101 }
4102
4103 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4104                                         void __user *arg)
4105 {
4106         size_t len;
4107         int ret;
4108         char label[BTRFS_LABEL_SIZE];
4109
4110         spin_lock(&fs_info->super_lock);
4111         memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4112         spin_unlock(&fs_info->super_lock);
4113
4114         len = strnlen(label, BTRFS_LABEL_SIZE);
4115
4116         if (len == BTRFS_LABEL_SIZE) {
4117                 btrfs_warn(fs_info,
4118                            "label is too long, return the first %zu bytes",
4119                            --len);
4120         }
4121
4122         ret = copy_to_user(arg, label, len);
4123
4124         return ret ? -EFAULT : 0;
4125 }
4126
4127 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4128 {
4129         struct inode *inode = file_inode(file);
4130         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4131         struct btrfs_root *root = BTRFS_I(inode)->root;
4132         struct btrfs_super_block *super_block = fs_info->super_copy;
4133         struct btrfs_trans_handle *trans;
4134         char label[BTRFS_LABEL_SIZE];
4135         int ret;
4136
4137         if (!capable(CAP_SYS_ADMIN))
4138                 return -EPERM;
4139
4140         if (copy_from_user(label, arg, sizeof(label)))
4141                 return -EFAULT;
4142
4143         if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4144                 btrfs_err(fs_info,
4145                           "unable to set label with more than %d bytes",
4146                           BTRFS_LABEL_SIZE - 1);
4147                 return -EINVAL;
4148         }
4149
4150         ret = mnt_want_write_file(file);
4151         if (ret)
4152                 return ret;
4153
4154         trans = btrfs_start_transaction(root, 0);
4155         if (IS_ERR(trans)) {
4156                 ret = PTR_ERR(trans);
4157                 goto out_unlock;
4158         }
4159
4160         spin_lock(&fs_info->super_lock);
4161         strcpy(super_block->label, label);
4162         spin_unlock(&fs_info->super_lock);
4163         ret = btrfs_commit_transaction(trans);
4164
4165 out_unlock:
4166         mnt_drop_write_file(file);
4167         return ret;
4168 }
4169
4170 #define INIT_FEATURE_FLAGS(suffix) \
4171         { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4172           .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4173           .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4174
4175 int btrfs_ioctl_get_supported_features(void __user *arg)
4176 {
4177         static const struct btrfs_ioctl_feature_flags features[3] = {
4178                 INIT_FEATURE_FLAGS(SUPP),
4179                 INIT_FEATURE_FLAGS(SAFE_SET),
4180                 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4181         };
4182
4183         if (copy_to_user(arg, &features, sizeof(features)))
4184                 return -EFAULT;
4185
4186         return 0;
4187 }
4188
4189 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4190                                         void __user *arg)
4191 {
4192         struct btrfs_super_block *super_block = fs_info->super_copy;
4193         struct btrfs_ioctl_feature_flags features;
4194
4195         features.compat_flags = btrfs_super_compat_flags(super_block);
4196         features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4197         features.incompat_flags = btrfs_super_incompat_flags(super_block);
4198
4199         if (copy_to_user(arg, &features, sizeof(features)))
4200                 return -EFAULT;
4201
4202         return 0;
4203 }
4204
4205 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4206                               enum btrfs_feature_set set,
4207                               u64 change_mask, u64 flags, u64 supported_flags,
4208                               u64 safe_set, u64 safe_clear)
4209 {
4210         const char *type = btrfs_feature_set_name(set);
4211         char *names;
4212         u64 disallowed, unsupported;
4213         u64 set_mask = flags & change_mask;
4214         u64 clear_mask = ~flags & change_mask;
4215
4216         unsupported = set_mask & ~supported_flags;
4217         if (unsupported) {
4218                 names = btrfs_printable_features(set, unsupported);
4219                 if (names) {
4220                         btrfs_warn(fs_info,
4221                                    "this kernel does not support the %s feature bit%s",
4222                                    names, strchr(names, ',') ? "s" : "");
4223                         kfree(names);
4224                 } else
4225                         btrfs_warn(fs_info,
4226                                    "this kernel does not support %s bits 0x%llx",
4227                                    type, unsupported);
4228                 return -EOPNOTSUPP;
4229         }
4230
4231         disallowed = set_mask & ~safe_set;
4232         if (disallowed) {
4233                 names = btrfs_printable_features(set, disallowed);
4234                 if (names) {
4235                         btrfs_warn(fs_info,
4236                                    "can't set the %s feature bit%s while mounted",
4237                                    names, strchr(names, ',') ? "s" : "");
4238                         kfree(names);
4239                 } else
4240                         btrfs_warn(fs_info,
4241                                    "can't set %s bits 0x%llx while mounted",
4242                                    type, disallowed);
4243                 return -EPERM;
4244         }
4245
4246         disallowed = clear_mask & ~safe_clear;
4247         if (disallowed) {
4248                 names = btrfs_printable_features(set, disallowed);
4249                 if (names) {
4250                         btrfs_warn(fs_info,
4251                                    "can't clear the %s feature bit%s while mounted",
4252                                    names, strchr(names, ',') ? "s" : "");
4253                         kfree(names);
4254                 } else
4255                         btrfs_warn(fs_info,
4256                                    "can't clear %s bits 0x%llx while mounted",
4257                                    type, disallowed);
4258                 return -EPERM;
4259         }
4260
4261         return 0;
4262 }
4263
4264 #define check_feature(fs_info, change_mask, flags, mask_base)   \
4265 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags,       \
4266                    BTRFS_FEATURE_ ## mask_base ## _SUPP,        \
4267                    BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,    \
4268                    BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4269
4270 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4271 {
4272         struct inode *inode = file_inode(file);
4273         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4274         struct btrfs_root *root = BTRFS_I(inode)->root;
4275         struct btrfs_super_block *super_block = fs_info->super_copy;
4276         struct btrfs_ioctl_feature_flags flags[2];
4277         struct btrfs_trans_handle *trans;
4278         u64 newflags;
4279         int ret;
4280
4281         if (!capable(CAP_SYS_ADMIN))
4282                 return -EPERM;
4283
4284         if (copy_from_user(flags, arg, sizeof(flags)))
4285                 return -EFAULT;
4286
4287         /* Nothing to do */
4288         if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4289             !flags[0].incompat_flags)
4290                 return 0;
4291
4292         ret = check_feature(fs_info, flags[0].compat_flags,
4293                             flags[1].compat_flags, COMPAT);
4294         if (ret)
4295                 return ret;
4296
4297         ret = check_feature(fs_info, flags[0].compat_ro_flags,
4298                             flags[1].compat_ro_flags, COMPAT_RO);
4299         if (ret)
4300                 return ret;
4301
4302         ret = check_feature(fs_info, flags[0].incompat_flags,
4303                             flags[1].incompat_flags, INCOMPAT);
4304         if (ret)
4305                 return ret;
4306
4307         ret = mnt_want_write_file(file);
4308         if (ret)
4309                 return ret;
4310
4311         trans = btrfs_start_transaction(root, 0);
4312         if (IS_ERR(trans)) {
4313                 ret = PTR_ERR(trans);
4314                 goto out_drop_write;
4315         }
4316
4317         spin_lock(&fs_info->super_lock);
4318         newflags = btrfs_super_compat_flags(super_block);
4319         newflags |= flags[0].compat_flags & flags[1].compat_flags;
4320         newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4321         btrfs_set_super_compat_flags(super_block, newflags);
4322
4323         newflags = btrfs_super_compat_ro_flags(super_block);
4324         newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4325         newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4326         btrfs_set_super_compat_ro_flags(super_block, newflags);
4327
4328         newflags = btrfs_super_incompat_flags(super_block);
4329         newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4330         newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4331         btrfs_set_super_incompat_flags(super_block, newflags);
4332         spin_unlock(&fs_info->super_lock);
4333
4334         ret = btrfs_commit_transaction(trans);
4335 out_drop_write:
4336         mnt_drop_write_file(file);
4337
4338         return ret;
4339 }
4340
4341 static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
4342 {
4343         struct btrfs_ioctl_send_args *arg;
4344         int ret;
4345
4346         if (compat) {
4347 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4348                 struct btrfs_ioctl_send_args_32 args32 = { 0 };
4349
4350                 ret = copy_from_user(&args32, argp, sizeof(args32));
4351                 if (ret)
4352                         return -EFAULT;
4353                 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4354                 if (!arg)
4355                         return -ENOMEM;
4356                 arg->send_fd = args32.send_fd;
4357                 arg->clone_sources_count = args32.clone_sources_count;
4358                 arg->clone_sources = compat_ptr(args32.clone_sources);
4359                 arg->parent_root = args32.parent_root;
4360                 arg->flags = args32.flags;
4361                 memcpy(arg->reserved, args32.reserved,
4362                        sizeof(args32.reserved));
4363 #else
4364                 return -ENOTTY;
4365 #endif
4366         } else {
4367                 arg = memdup_user(argp, sizeof(*arg));
4368                 if (IS_ERR(arg))
4369                         return PTR_ERR(arg);
4370         }
4371         ret = btrfs_ioctl_send(inode, arg);
4372         kfree(arg);
4373         return ret;
4374 }
4375
4376 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4377                                     bool compat)
4378 {
4379         struct btrfs_ioctl_encoded_io_args args = { 0 };
4380         size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4381                                              flags);
4382         size_t copy_end;
4383         struct iovec iovstack[UIO_FASTIOV];
4384         struct iovec *iov = iovstack;
4385         struct iov_iter iter;
4386         loff_t pos;
4387         struct kiocb kiocb;
4388         ssize_t ret;
4389
4390         if (!capable(CAP_SYS_ADMIN)) {
4391                 ret = -EPERM;
4392                 goto out_acct;
4393         }
4394
4395         if (compat) {
4396 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4397                 struct btrfs_ioctl_encoded_io_args_32 args32;
4398
4399                 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4400                                        flags);
4401                 if (copy_from_user(&args32, argp, copy_end)) {
4402                         ret = -EFAULT;
4403                         goto out_acct;
4404                 }
4405                 args.iov = compat_ptr(args32.iov);
4406                 args.iovcnt = args32.iovcnt;
4407                 args.offset = args32.offset;
4408                 args.flags = args32.flags;
4409 #else
4410                 return -ENOTTY;
4411 #endif
4412         } else {
4413                 copy_end = copy_end_kernel;
4414                 if (copy_from_user(&args, argp, copy_end)) {
4415                         ret = -EFAULT;
4416                         goto out_acct;
4417                 }
4418         }
4419         if (args.flags != 0) {
4420                 ret = -EINVAL;
4421                 goto out_acct;
4422         }
4423
4424         ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4425                            &iov, &iter);
4426         if (ret < 0)
4427                 goto out_acct;
4428
4429         if (iov_iter_count(&iter) == 0) {
4430                 ret = 0;
4431                 goto out_iov;
4432         }
4433         pos = args.offset;
4434         ret = rw_verify_area(READ, file, &pos, args.len);
4435         if (ret < 0)
4436                 goto out_iov;
4437
4438         init_sync_kiocb(&kiocb, file);
4439         kiocb.ki_pos = pos;
4440
4441         ret = btrfs_encoded_read(&kiocb, &iter, &args);
4442         if (ret >= 0) {
4443                 fsnotify_access(file);
4444                 if (copy_to_user(argp + copy_end,
4445                                  (char *)&args + copy_end_kernel,
4446                                  sizeof(args) - copy_end_kernel))
4447                         ret = -EFAULT;
4448         }
4449
4450 out_iov:
4451         kfree(iov);
4452 out_acct:
4453         if (ret > 0)
4454                 add_rchar(current, ret);
4455         inc_syscr(current);
4456         return ret;
4457 }
4458
4459 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4460 {
4461         struct btrfs_ioctl_encoded_io_args args;
4462         struct iovec iovstack[UIO_FASTIOV];
4463         struct iovec *iov = iovstack;
4464         struct iov_iter iter;
4465         loff_t pos;
4466         struct kiocb kiocb;
4467         ssize_t ret;
4468
4469         if (!capable(CAP_SYS_ADMIN)) {
4470                 ret = -EPERM;
4471                 goto out_acct;
4472         }
4473
4474         if (!(file->f_mode & FMODE_WRITE)) {
4475                 ret = -EBADF;
4476                 goto out_acct;
4477         }
4478
4479         if (compat) {
4480 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4481                 struct btrfs_ioctl_encoded_io_args_32 args32;
4482
4483                 if (copy_from_user(&args32, argp, sizeof(args32))) {
4484                         ret = -EFAULT;
4485                         goto out_acct;
4486                 }
4487                 args.iov = compat_ptr(args32.iov);
4488                 args.iovcnt = args32.iovcnt;
4489                 args.offset = args32.offset;
4490                 args.flags = args32.flags;
4491                 args.len = args32.len;
4492                 args.unencoded_len = args32.unencoded_len;
4493                 args.unencoded_offset = args32.unencoded_offset;
4494                 args.compression = args32.compression;
4495                 args.encryption = args32.encryption;
4496                 memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4497 #else
4498                 return -ENOTTY;
4499 #endif
4500         } else {
4501                 if (copy_from_user(&args, argp, sizeof(args))) {
4502                         ret = -EFAULT;
4503                         goto out_acct;
4504                 }
4505         }
4506
4507         ret = -EINVAL;
4508         if (args.flags != 0)
4509                 goto out_acct;
4510         if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4511                 goto out_acct;
4512         if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4513             args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4514                 goto out_acct;
4515         if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4516             args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4517                 goto out_acct;
4518         if (args.unencoded_offset > args.unencoded_len)
4519                 goto out_acct;
4520         if (args.len > args.unencoded_len - args.unencoded_offset)
4521                 goto out_acct;
4522
4523         ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4524                            &iov, &iter);
4525         if (ret < 0)
4526                 goto out_acct;
4527
4528         file_start_write(file);
4529
4530         if (iov_iter_count(&iter) == 0) {
4531                 ret = 0;
4532                 goto out_end_write;
4533         }
4534         pos = args.offset;
4535         ret = rw_verify_area(WRITE, file, &pos, args.len);
4536         if (ret < 0)
4537                 goto out_end_write;
4538
4539         init_sync_kiocb(&kiocb, file);
4540         ret = kiocb_set_rw_flags(&kiocb, 0);
4541         if (ret)
4542                 goto out_end_write;
4543         kiocb.ki_pos = pos;
4544
4545         ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4546         if (ret > 0)
4547                 fsnotify_modify(file);
4548
4549 out_end_write:
4550         file_end_write(file);
4551         kfree(iov);
4552 out_acct:
4553         if (ret > 0)
4554                 add_wchar(current, ret);
4555         inc_syscw(current);
4556         return ret;
4557 }
4558
4559 long btrfs_ioctl(struct file *file, unsigned int
4560                 cmd, unsigned long arg)
4561 {
4562         struct inode *inode = file_inode(file);
4563         struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4564         struct btrfs_root *root = BTRFS_I(inode)->root;
4565         void __user *argp = (void __user *)arg;
4566
4567         switch (cmd) {
4568         case FS_IOC_GETVERSION:
4569                 return btrfs_ioctl_getversion(inode, argp);
4570         case FS_IOC_GETFSLABEL:
4571                 return btrfs_ioctl_get_fslabel(fs_info, argp);
4572         case FS_IOC_SETFSLABEL:
4573                 return btrfs_ioctl_set_fslabel(file, argp);
4574         case FITRIM:
4575                 return btrfs_ioctl_fitrim(fs_info, argp);
4576         case BTRFS_IOC_SNAP_CREATE:
4577                 return btrfs_ioctl_snap_create(file, argp, 0);
4578         case BTRFS_IOC_SNAP_CREATE_V2:
4579                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4580         case BTRFS_IOC_SUBVOL_CREATE:
4581                 return btrfs_ioctl_snap_create(file, argp, 1);
4582         case BTRFS_IOC_SUBVOL_CREATE_V2:
4583                 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4584         case BTRFS_IOC_SNAP_DESTROY:
4585                 return btrfs_ioctl_snap_destroy(file, argp, false);
4586         case BTRFS_IOC_SNAP_DESTROY_V2:
4587                 return btrfs_ioctl_snap_destroy(file, argp, true);
4588         case BTRFS_IOC_SUBVOL_GETFLAGS:
4589                 return btrfs_ioctl_subvol_getflags(inode, argp);
4590         case BTRFS_IOC_SUBVOL_SETFLAGS:
4591                 return btrfs_ioctl_subvol_setflags(file, argp);
4592         case BTRFS_IOC_DEFAULT_SUBVOL:
4593                 return btrfs_ioctl_default_subvol(file, argp);
4594         case BTRFS_IOC_DEFRAG:
4595                 return btrfs_ioctl_defrag(file, NULL);
4596         case BTRFS_IOC_DEFRAG_RANGE:
4597                 return btrfs_ioctl_defrag(file, argp);
4598         case BTRFS_IOC_RESIZE:
4599                 return btrfs_ioctl_resize(file, argp);
4600         case BTRFS_IOC_ADD_DEV:
4601                 return btrfs_ioctl_add_dev(fs_info, argp);
4602         case BTRFS_IOC_RM_DEV:
4603                 return btrfs_ioctl_rm_dev(file, argp);
4604         case BTRFS_IOC_RM_DEV_V2:
4605                 return btrfs_ioctl_rm_dev_v2(file, argp);
4606         case BTRFS_IOC_FS_INFO:
4607                 return btrfs_ioctl_fs_info(fs_info, argp);
4608         case BTRFS_IOC_DEV_INFO:
4609                 return btrfs_ioctl_dev_info(fs_info, argp);
4610         case BTRFS_IOC_TREE_SEARCH:
4611                 return btrfs_ioctl_tree_search(inode, argp);
4612         case BTRFS_IOC_TREE_SEARCH_V2:
4613                 return btrfs_ioctl_tree_search_v2(inode, argp);
4614         case BTRFS_IOC_INO_LOOKUP:
4615                 return btrfs_ioctl_ino_lookup(root, argp);
4616         case BTRFS_IOC_INO_PATHS:
4617                 return btrfs_ioctl_ino_to_path(root, argp);
4618         case BTRFS_IOC_LOGICAL_INO:
4619                 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4620         case BTRFS_IOC_LOGICAL_INO_V2:
4621                 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4622         case BTRFS_IOC_SPACE_INFO:
4623                 return btrfs_ioctl_space_info(fs_info, argp);
4624         case BTRFS_IOC_SYNC: {
4625                 int ret;
4626
4627                 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4628                 if (ret)
4629                         return ret;
4630                 ret = btrfs_sync_fs(inode->i_sb, 1);
4631                 /*
4632                  * The transaction thread may want to do more work,
4633                  * namely it pokes the cleaner kthread that will start
4634                  * processing uncleaned subvols.
4635                  */
4636                 wake_up_process(fs_info->transaction_kthread);
4637                 return ret;
4638         }
4639         case BTRFS_IOC_START_SYNC:
4640                 return btrfs_ioctl_start_sync(root, argp);
4641         case BTRFS_IOC_WAIT_SYNC:
4642                 return btrfs_ioctl_wait_sync(fs_info, argp);
4643         case BTRFS_IOC_SCRUB:
4644                 return btrfs_ioctl_scrub(file, argp);
4645         case BTRFS_IOC_SCRUB_CANCEL:
4646                 return btrfs_ioctl_scrub_cancel(fs_info);
4647         case BTRFS_IOC_SCRUB_PROGRESS:
4648                 return btrfs_ioctl_scrub_progress(fs_info, argp);
4649         case BTRFS_IOC_BALANCE_V2:
4650                 return btrfs_ioctl_balance(file, argp);
4651         case BTRFS_IOC_BALANCE_CTL:
4652                 return btrfs_ioctl_balance_ctl(fs_info, arg);
4653         case BTRFS_IOC_BALANCE_PROGRESS:
4654                 return btrfs_ioctl_balance_progress(fs_info, argp);
4655         case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4656                 return btrfs_ioctl_set_received_subvol(file, argp);
4657 #ifdef CONFIG_64BIT
4658         case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4659                 return btrfs_ioctl_set_received_subvol_32(file, argp);
4660 #endif
4661         case BTRFS_IOC_SEND:
4662                 return _btrfs_ioctl_send(inode, argp, false);
4663 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4664         case BTRFS_IOC_SEND_32:
4665                 return _btrfs_ioctl_send(inode, argp, true);
4666 #endif
4667         case BTRFS_IOC_GET_DEV_STATS:
4668                 return btrfs_ioctl_get_dev_stats(fs_info, argp);
4669         case BTRFS_IOC_QUOTA_CTL:
4670                 return btrfs_ioctl_quota_ctl(file, argp);
4671         case BTRFS_IOC_QGROUP_ASSIGN:
4672                 return btrfs_ioctl_qgroup_assign(file, argp);
4673         case BTRFS_IOC_QGROUP_CREATE:
4674                 return btrfs_ioctl_qgroup_create(file, argp);
4675         case BTRFS_IOC_QGROUP_LIMIT:
4676                 return btrfs_ioctl_qgroup_limit(file, argp);
4677         case BTRFS_IOC_QUOTA_RESCAN:
4678                 return btrfs_ioctl_quota_rescan(file, argp);
4679         case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4680                 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4681         case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4682                 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4683         case BTRFS_IOC_DEV_REPLACE:
4684                 return btrfs_ioctl_dev_replace(fs_info, argp);
4685         case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4686                 return btrfs_ioctl_get_supported_features(argp);
4687         case BTRFS_IOC_GET_FEATURES:
4688                 return btrfs_ioctl_get_features(fs_info, argp);
4689         case BTRFS_IOC_SET_FEATURES:
4690                 return btrfs_ioctl_set_features(file, argp);
4691         case BTRFS_IOC_GET_SUBVOL_INFO:
4692                 return btrfs_ioctl_get_subvol_info(inode, argp);
4693         case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4694                 return btrfs_ioctl_get_subvol_rootref(root, argp);
4695         case BTRFS_IOC_INO_LOOKUP_USER:
4696                 return btrfs_ioctl_ino_lookup_user(file, argp);
4697         case FS_IOC_ENABLE_VERITY:
4698                 return fsverity_ioctl_enable(file, (const void __user *)argp);
4699         case FS_IOC_MEASURE_VERITY:
4700                 return fsverity_ioctl_measure(file, argp);
4701         case BTRFS_IOC_ENCODED_READ:
4702                 return btrfs_ioctl_encoded_read(file, argp, false);
4703         case BTRFS_IOC_ENCODED_WRITE:
4704                 return btrfs_ioctl_encoded_write(file, argp, false);
4705 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4706         case BTRFS_IOC_ENCODED_READ_32:
4707                 return btrfs_ioctl_encoded_read(file, argp, true);
4708         case BTRFS_IOC_ENCODED_WRITE_32:
4709                 return btrfs_ioctl_encoded_write(file, argp, true);
4710 #endif
4711         }
4712
4713         return -ENOTTY;
4714 }
4715
4716 #ifdef CONFIG_COMPAT
4717 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4718 {
4719         /*
4720          * These all access 32-bit values anyway so no further
4721          * handling is necessary.
4722          */
4723         switch (cmd) {
4724         case FS_IOC32_GETVERSION:
4725                 cmd = FS_IOC_GETVERSION;
4726                 break;
4727         }
4728
4729         return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4730 }
4731 #endif