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