~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/btrfs/ioctl.c

Version: ~ [ linux-5.10-rc5 ] ~ [ linux-5.9.10 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.79 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.159 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.208 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.245 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.245 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  * Copyright (C) 2007 Oracle.  All rights reserved.
  3  *
  4  * This program is free software; you can redistribute it and/or
  5  * modify it under the terms of the GNU General Public
  6  * License v2 as published by the Free Software Foundation.
  7  *
  8  * This program is distributed in the hope that it will be useful,
  9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
 10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 11  * General Public License for more details.
 12  *
 13  * You should have received a copy of the GNU General Public
 14  * License along with this program; if not, write to the
 15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 16  * Boston, MA 021110-1307, USA.
 17  */
 18 
 19 #include <linux/kernel.h>
 20 #include <linux/bio.h>
 21 #include <linux/buffer_head.h>
 22 #include <linux/file.h>
 23 #include <linux/fs.h>
 24 #include <linux/fsnotify.h>
 25 #include <linux/pagemap.h>
 26 #include <linux/highmem.h>
 27 #include <linux/time.h>
 28 #include <linux/init.h>
 29 #include <linux/string.h>
 30 #include <linux/backing-dev.h>
 31 #include <linux/mount.h>
 32 #include <linux/mpage.h>
 33 #include <linux/namei.h>
 34 #include <linux/swap.h>
 35 #include <linux/writeback.h>
 36 #include <linux/statfs.h>
 37 #include <linux/compat.h>
 38 #include <linux/bit_spinlock.h>
 39 #include <linux/security.h>
 40 #include <linux/xattr.h>
 41 #include <linux/vmalloc.h>
 42 #include <linux/slab.h>
 43 #include <linux/blkdev.h>
 44 #include <linux/uuid.h>
 45 #include <linux/btrfs.h>
 46 #include <linux/uaccess.h>
 47 #include "ctree.h"
 48 #include "disk-io.h"
 49 #include "transaction.h"
 50 #include "btrfs_inode.h"
 51 #include "print-tree.h"
 52 #include "volumes.h"
 53 #include "locking.h"
 54 #include "inode-map.h"
 55 #include "backref.h"
 56 #include "rcu-string.h"
 57 #include "send.h"
 58 #include "dev-replace.h"
 59 #include "props.h"
 60 #include "sysfs.h"
 61 #include "qgroup.h"
 62 
 63 #ifdef CONFIG_64BIT
 64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
 65  * structures are incorrect, as the timespec structure from userspace
 66  * is 4 bytes too small. We define these alternatives here to teach
 67  * the kernel about the 32-bit struct packing.
 68  */
 69 struct btrfs_ioctl_timespec_32 {
 70         __u64 sec;
 71         __u32 nsec;
 72 } __attribute__ ((__packed__));
 73 
 74 struct btrfs_ioctl_received_subvol_args_32 {
 75         char    uuid[BTRFS_UUID_SIZE];  /* in */
 76         __u64   stransid;               /* in */
 77         __u64   rtransid;               /* out */
 78         struct btrfs_ioctl_timespec_32 stime; /* in */
 79         struct btrfs_ioctl_timespec_32 rtime; /* out */
 80         __u64   flags;                  /* in */
 81         __u64   reserved[16];           /* in */
 82 } __attribute__ ((__packed__));
 83 
 84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
 85                                 struct btrfs_ioctl_received_subvol_args_32)
 86 #endif
 87 
 88 
 89 static int btrfs_clone(struct inode *src, struct inode *inode,
 90                        u64 off, u64 olen, u64 olen_aligned, u64 destoff);
 91 
 92 /* Mask out flags that are inappropriate for the given type of inode. */
 93 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
 94 {
 95         if (S_ISDIR(mode))
 96                 return flags;
 97         else if (S_ISREG(mode))
 98                 return flags & ~FS_DIRSYNC_FL;
 99         else
100                 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
101 }
102 
103 /*
104  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
105  */
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
107 {
108         unsigned int iflags = 0;
109 
110         if (flags & BTRFS_INODE_SYNC)
111                 iflags |= FS_SYNC_FL;
112         if (flags & BTRFS_INODE_IMMUTABLE)
113                 iflags |= FS_IMMUTABLE_FL;
114         if (flags & BTRFS_INODE_APPEND)
115                 iflags |= FS_APPEND_FL;
116         if (flags & BTRFS_INODE_NODUMP)
117                 iflags |= FS_NODUMP_FL;
118         if (flags & BTRFS_INODE_NOATIME)
119                 iflags |= FS_NOATIME_FL;
120         if (flags & BTRFS_INODE_DIRSYNC)
121                 iflags |= FS_DIRSYNC_FL;
122         if (flags & BTRFS_INODE_NODATACOW)
123                 iflags |= FS_NOCOW_FL;
124 
125         if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126                 iflags |= FS_COMPR_FL;
127         else if (flags & BTRFS_INODE_NOCOMPRESS)
128                 iflags |= FS_NOCOMP_FL;
129 
130         return iflags;
131 }
132 
133 /*
134  * Update inode->i_flags based on the btrfs internal flags.
135  */
136 void btrfs_update_iflags(struct inode *inode)
137 {
138         struct btrfs_inode *ip = BTRFS_I(inode);
139         unsigned int new_fl = 0;
140 
141         if (ip->flags & BTRFS_INODE_SYNC)
142                 new_fl |= S_SYNC;
143         if (ip->flags & BTRFS_INODE_IMMUTABLE)
144                 new_fl |= S_IMMUTABLE;
145         if (ip->flags & BTRFS_INODE_APPEND)
146                 new_fl |= S_APPEND;
147         if (ip->flags & BTRFS_INODE_NOATIME)
148                 new_fl |= S_NOATIME;
149         if (ip->flags & BTRFS_INODE_DIRSYNC)
150                 new_fl |= S_DIRSYNC;
151 
152         set_mask_bits(&inode->i_flags,
153                       S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
154                       new_fl);
155 }
156 
157 /*
158  * Inherit flags from the parent inode.
159  *
160  * Currently only the compression flags and the cow flags are inherited.
161  */
162 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
163 {
164         unsigned int flags;
165 
166         if (!dir)
167                 return;
168 
169         flags = BTRFS_I(dir)->flags;
170 
171         if (flags & BTRFS_INODE_NOCOMPRESS) {
172                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173                 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174         } else if (flags & BTRFS_INODE_COMPRESS) {
175                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176                 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
177         }
178 
179         if (flags & BTRFS_INODE_NODATACOW) {
180                 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
181                 if (S_ISREG(inode->i_mode))
182                         BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
183         }
184 
185         btrfs_update_iflags(inode);
186 }
187 
188 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
189 {
190         struct btrfs_inode *ip = BTRFS_I(file_inode(file));
191         unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
192 
193         if (copy_to_user(arg, &flags, sizeof(flags)))
194                 return -EFAULT;
195         return 0;
196 }
197 
198 static int check_flags(unsigned int flags)
199 {
200         if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201                       FS_NOATIME_FL | FS_NODUMP_FL | \
202                       FS_SYNC_FL | FS_DIRSYNC_FL | \
203                       FS_NOCOMP_FL | FS_COMPR_FL |
204                       FS_NOCOW_FL))
205                 return -EOPNOTSUPP;
206 
207         if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
208                 return -EINVAL;
209 
210         return 0;
211 }
212 
213 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
214 {
215         struct inode *inode = file_inode(file);
216         struct btrfs_inode *ip = BTRFS_I(inode);
217         struct btrfs_root *root = ip->root;
218         struct btrfs_trans_handle *trans;
219         unsigned int flags, oldflags;
220         int ret;
221         u64 ip_oldflags;
222         unsigned int i_oldflags;
223         umode_t mode;
224 
225         if (!inode_owner_or_capable(inode))
226                 return -EPERM;
227 
228         if (btrfs_root_readonly(root))
229                 return -EROFS;
230 
231         if (copy_from_user(&flags, arg, sizeof(flags)))
232                 return -EFAULT;
233 
234         ret = check_flags(flags);
235         if (ret)
236                 return ret;
237 
238         ret = mnt_want_write_file(file);
239         if (ret)
240                 return ret;
241 
242         mutex_lock(&inode->i_mutex);
243 
244         ip_oldflags = ip->flags;
245         i_oldflags = inode->i_flags;
246         mode = inode->i_mode;
247 
248         flags = btrfs_mask_flags(inode->i_mode, flags);
249         oldflags = btrfs_flags_to_ioctl(ip->flags);
250         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251                 if (!capable(CAP_LINUX_IMMUTABLE)) {
252                         ret = -EPERM;
253                         goto out_unlock;
254                 }
255         }
256 
257         if (flags & FS_SYNC_FL)
258                 ip->flags |= BTRFS_INODE_SYNC;
259         else
260                 ip->flags &= ~BTRFS_INODE_SYNC;
261         if (flags & FS_IMMUTABLE_FL)
262                 ip->flags |= BTRFS_INODE_IMMUTABLE;
263         else
264                 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265         if (flags & FS_APPEND_FL)
266                 ip->flags |= BTRFS_INODE_APPEND;
267         else
268                 ip->flags &= ~BTRFS_INODE_APPEND;
269         if (flags & FS_NODUMP_FL)
270                 ip->flags |= BTRFS_INODE_NODUMP;
271         else
272                 ip->flags &= ~BTRFS_INODE_NODUMP;
273         if (flags & FS_NOATIME_FL)
274                 ip->flags |= BTRFS_INODE_NOATIME;
275         else
276                 ip->flags &= ~BTRFS_INODE_NOATIME;
277         if (flags & FS_DIRSYNC_FL)
278                 ip->flags |= BTRFS_INODE_DIRSYNC;
279         else
280                 ip->flags &= ~BTRFS_INODE_DIRSYNC;
281         if (flags & FS_NOCOW_FL) {
282                 if (S_ISREG(mode)) {
283                         /*
284                          * It's safe to turn csums off here, no extents exist.
285                          * Otherwise we want the flag to reflect the real COW
286                          * status of the file and will not set it.
287                          */
288                         if (inode->i_size == 0)
289                                 ip->flags |= BTRFS_INODE_NODATACOW
290                                            | BTRFS_INODE_NODATASUM;
291                 } else {
292                         ip->flags |= BTRFS_INODE_NODATACOW;
293                 }
294         } else {
295                 /*
296                  * Revert back under same assuptions as above
297                  */
298                 if (S_ISREG(mode)) {
299                         if (inode->i_size == 0)
300                                 ip->flags &= ~(BTRFS_INODE_NODATACOW
301                                              | BTRFS_INODE_NODATASUM);
302                 } else {
303                         ip->flags &= ~BTRFS_INODE_NODATACOW;
304                 }
305         }
306 
307         /*
308          * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309          * flag may be changed automatically if compression code won't make
310          * things smaller.
311          */
312         if (flags & FS_NOCOMP_FL) {
313                 ip->flags &= ~BTRFS_INODE_COMPRESS;
314                 ip->flags |= BTRFS_INODE_NOCOMPRESS;
315 
316                 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317                 if (ret && ret != -ENODATA)
318                         goto out_drop;
319         } else if (flags & FS_COMPR_FL) {
320                 const char *comp;
321 
322                 ip->flags |= BTRFS_INODE_COMPRESS;
323                 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
324 
325                 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
326                         comp = "lzo";
327                 else
328                         comp = "zlib";
329                 ret = btrfs_set_prop(inode, "btrfs.compression",
330                                      comp, strlen(comp), 0);
331                 if (ret)
332                         goto out_drop;
333 
334         } else {
335                 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
336                 if (ret && ret != -ENODATA)
337                         goto out_drop;
338                 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
339         }
340 
341         trans = btrfs_start_transaction(root, 1);
342         if (IS_ERR(trans)) {
343                 ret = PTR_ERR(trans);
344                 goto out_drop;
345         }
346 
347         btrfs_update_iflags(inode);
348         inode_inc_iversion(inode);
349         inode->i_ctime = CURRENT_TIME;
350         ret = btrfs_update_inode(trans, root, inode);
351 
352         btrfs_end_transaction(trans, root);
353  out_drop:
354         if (ret) {
355                 ip->flags = ip_oldflags;
356                 inode->i_flags = i_oldflags;
357         }
358 
359  out_unlock:
360         mutex_unlock(&inode->i_mutex);
361         mnt_drop_write_file(file);
362         return ret;
363 }
364 
365 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
366 {
367         struct inode *inode = file_inode(file);
368 
369         return put_user(inode->i_generation, arg);
370 }
371 
372 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
373 {
374         struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
375         struct btrfs_device *device;
376         struct request_queue *q;
377         struct fstrim_range range;
378         u64 minlen = ULLONG_MAX;
379         u64 num_devices = 0;
380         u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
381         int ret;
382 
383         if (!capable(CAP_SYS_ADMIN))
384                 return -EPERM;
385 
386         rcu_read_lock();
387         list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
388                                 dev_list) {
389                 if (!device->bdev)
390                         continue;
391                 q = bdev_get_queue(device->bdev);
392                 if (blk_queue_discard(q)) {
393                         num_devices++;
394                         minlen = min((u64)q->limits.discard_granularity,
395                                      minlen);
396                 }
397         }
398         rcu_read_unlock();
399 
400         if (!num_devices)
401                 return -EOPNOTSUPP;
402         if (copy_from_user(&range, arg, sizeof(range)))
403                 return -EFAULT;
404         if (range.start > total_bytes ||
405             range.len < fs_info->sb->s_blocksize)
406                 return -EINVAL;
407 
408         range.len = min(range.len, total_bytes - range.start);
409         range.minlen = max(range.minlen, minlen);
410         ret = btrfs_trim_fs(fs_info->tree_root, &range);
411         if (ret < 0)
412                 return ret;
413 
414         if (copy_to_user(arg, &range, sizeof(range)))
415                 return -EFAULT;
416 
417         return 0;
418 }
419 
420 int btrfs_is_empty_uuid(u8 *uuid)
421 {
422         int i;
423 
424         for (i = 0; i < BTRFS_UUID_SIZE; i++) {
425                 if (uuid[i])
426                         return 0;
427         }
428         return 1;
429 }
430 
431 static noinline int create_subvol(struct inode *dir,
432                                   struct dentry *dentry,
433                                   char *name, int namelen,
434                                   u64 *async_transid,
435                                   struct btrfs_qgroup_inherit *inherit)
436 {
437         struct btrfs_trans_handle *trans;
438         struct btrfs_key key;
439         struct btrfs_root_item root_item;
440         struct btrfs_inode_item *inode_item;
441         struct extent_buffer *leaf;
442         struct btrfs_root *root = BTRFS_I(dir)->root;
443         struct btrfs_root *new_root;
444         struct btrfs_block_rsv block_rsv;
445         struct timespec cur_time = CURRENT_TIME;
446         struct inode *inode;
447         int ret;
448         int err;
449         u64 objectid;
450         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
451         u64 index = 0;
452         u64 qgroup_reserved;
453         uuid_le new_uuid;
454 
455         ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
456         if (ret)
457                 return ret;
458 
459         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
460         /*
461          * The same as the snapshot creation, please see the comment
462          * of create_snapshot().
463          */
464         ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
465                                                8, &qgroup_reserved, false);
466         if (ret)
467                 return ret;
468 
469         trans = btrfs_start_transaction(root, 0);
470         if (IS_ERR(trans)) {
471                 ret = PTR_ERR(trans);
472                 btrfs_subvolume_release_metadata(root, &block_rsv,
473                                                  qgroup_reserved);
474                 return ret;
475         }
476         trans->block_rsv = &block_rsv;
477         trans->bytes_reserved = block_rsv.size;
478 
479         ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
480         if (ret)
481                 goto fail;
482 
483         leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
484         if (IS_ERR(leaf)) {
485                 ret = PTR_ERR(leaf);
486                 goto fail;
487         }
488 
489         memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
490         btrfs_set_header_bytenr(leaf, leaf->start);
491         btrfs_set_header_generation(leaf, trans->transid);
492         btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
493         btrfs_set_header_owner(leaf, objectid);
494 
495         write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
496                             BTRFS_FSID_SIZE);
497         write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
498                             btrfs_header_chunk_tree_uuid(leaf),
499                             BTRFS_UUID_SIZE);
500         btrfs_mark_buffer_dirty(leaf);
501 
502         memset(&root_item, 0, sizeof(root_item));
503 
504         inode_item = &root_item.inode;
505         btrfs_set_stack_inode_generation(inode_item, 1);
506         btrfs_set_stack_inode_size(inode_item, 3);
507         btrfs_set_stack_inode_nlink(inode_item, 1);
508         btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
509         btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
510 
511         btrfs_set_root_flags(&root_item, 0);
512         btrfs_set_root_limit(&root_item, 0);
513         btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
514 
515         btrfs_set_root_bytenr(&root_item, leaf->start);
516         btrfs_set_root_generation(&root_item, trans->transid);
517         btrfs_set_root_level(&root_item, 0);
518         btrfs_set_root_refs(&root_item, 1);
519         btrfs_set_root_used(&root_item, leaf->len);
520         btrfs_set_root_last_snapshot(&root_item, 0);
521 
522         btrfs_set_root_generation_v2(&root_item,
523                         btrfs_root_generation(&root_item));
524         uuid_le_gen(&new_uuid);
525         memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
526         btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
527         btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
528         root_item.ctime = root_item.otime;
529         btrfs_set_root_ctransid(&root_item, trans->transid);
530         btrfs_set_root_otransid(&root_item, trans->transid);
531 
532         btrfs_tree_unlock(leaf);
533         free_extent_buffer(leaf);
534         leaf = NULL;
535 
536         btrfs_set_root_dirid(&root_item, new_dirid);
537 
538         key.objectid = objectid;
539         key.offset = 0;
540         key.type = BTRFS_ROOT_ITEM_KEY;
541         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
542                                 &root_item);
543         if (ret)
544                 goto fail;
545 
546         key.offset = (u64)-1;
547         new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
548         if (IS_ERR(new_root)) {
549                 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
550                 ret = PTR_ERR(new_root);
551                 goto fail;
552         }
553 
554         btrfs_record_root_in_trans(trans, new_root);
555 
556         ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
557         if (ret) {
558                 /* We potentially lose an unused inode item here */
559                 btrfs_abort_transaction(trans, root, ret);
560                 goto fail;
561         }
562 
563         /*
564          * insert the directory item
565          */
566         ret = btrfs_set_inode_index(dir, &index);
567         if (ret) {
568                 btrfs_abort_transaction(trans, root, ret);
569                 goto fail;
570         }
571 
572         ret = btrfs_insert_dir_item(trans, root,
573                                     name, namelen, dir, &key,
574                                     BTRFS_FT_DIR, index);
575         if (ret) {
576                 btrfs_abort_transaction(trans, root, ret);
577                 goto fail;
578         }
579 
580         btrfs_i_size_write(dir, dir->i_size + namelen * 2);
581         ret = btrfs_update_inode(trans, root, dir);
582         BUG_ON(ret);
583 
584         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
585                                  objectid, root->root_key.objectid,
586                                  btrfs_ino(dir), index, name, namelen);
587         BUG_ON(ret);
588 
589         ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
590                                   root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
591                                   objectid);
592         if (ret)
593                 btrfs_abort_transaction(trans, root, ret);
594 
595 fail:
596         trans->block_rsv = NULL;
597         trans->bytes_reserved = 0;
598         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
599 
600         if (async_transid) {
601                 *async_transid = trans->transid;
602                 err = btrfs_commit_transaction_async(trans, root, 1);
603                 if (err)
604                         err = btrfs_commit_transaction(trans, root);
605         } else {
606                 err = btrfs_commit_transaction(trans, root);
607         }
608         if (err && !ret)
609                 ret = err;
610 
611         if (!ret) {
612                 inode = btrfs_lookup_dentry(dir, dentry);
613                 if (IS_ERR(inode))
614                         return PTR_ERR(inode);
615                 d_instantiate(dentry, inode);
616         }
617         return ret;
618 }
619 
620 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
621 {
622         s64 writers;
623         DEFINE_WAIT(wait);
624 
625         do {
626                 prepare_to_wait(&root->subv_writers->wait, &wait,
627                                 TASK_UNINTERRUPTIBLE);
628 
629                 writers = percpu_counter_sum(&root->subv_writers->counter);
630                 if (writers)
631                         schedule();
632 
633                 finish_wait(&root->subv_writers->wait, &wait);
634         } while (writers);
635 }
636 
637 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
638                            struct dentry *dentry, char *name, int namelen,
639                            u64 *async_transid, bool readonly,
640                            struct btrfs_qgroup_inherit *inherit)
641 {
642         struct inode *inode;
643         struct btrfs_pending_snapshot *pending_snapshot;
644         struct btrfs_trans_handle *trans;
645         int ret;
646 
647         if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
648                 return -EINVAL;
649 
650         atomic_inc(&root->will_be_snapshoted);
651         smp_mb__after_atomic();
652         btrfs_wait_for_no_snapshoting_writes(root);
653 
654         ret = btrfs_start_delalloc_inodes(root, 0);
655         if (ret)
656                 goto out;
657 
658         btrfs_wait_ordered_extents(root, -1);
659 
660         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
661         if (!pending_snapshot) {
662                 ret = -ENOMEM;
663                 goto out;
664         }
665 
666         btrfs_init_block_rsv(&pending_snapshot->block_rsv,
667                              BTRFS_BLOCK_RSV_TEMP);
668         /*
669          * 1 - parent dir inode
670          * 2 - dir entries
671          * 1 - root item
672          * 2 - root ref/backref
673          * 1 - root of snapshot
674          * 1 - UUID item
675          */
676         ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
677                                         &pending_snapshot->block_rsv, 8,
678                                         &pending_snapshot->qgroup_reserved,
679                                         false);
680         if (ret)
681                 goto free;
682 
683         pending_snapshot->dentry = dentry;
684         pending_snapshot->root = root;
685         pending_snapshot->readonly = readonly;
686         pending_snapshot->dir = dir;
687         pending_snapshot->inherit = inherit;
688 
689         trans = btrfs_start_transaction(root, 0);
690         if (IS_ERR(trans)) {
691                 ret = PTR_ERR(trans);
692                 goto fail;
693         }
694 
695         spin_lock(&root->fs_info->trans_lock);
696         list_add(&pending_snapshot->list,
697                  &trans->transaction->pending_snapshots);
698         spin_unlock(&root->fs_info->trans_lock);
699         if (async_transid) {
700                 *async_transid = trans->transid;
701                 ret = btrfs_commit_transaction_async(trans,
702                                      root->fs_info->extent_root, 1);
703                 if (ret)
704                         ret = btrfs_commit_transaction(trans, root);
705         } else {
706                 ret = btrfs_commit_transaction(trans,
707                                                root->fs_info->extent_root);
708         }
709         if (ret)
710                 goto fail;
711 
712         ret = pending_snapshot->error;
713         if (ret)
714                 goto fail;
715 
716         ret = btrfs_orphan_cleanup(pending_snapshot->snap);
717         if (ret)
718                 goto fail;
719 
720         inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
721         if (IS_ERR(inode)) {
722                 ret = PTR_ERR(inode);
723                 goto fail;
724         }
725 
726         d_instantiate(dentry, inode);
727         ret = 0;
728 fail:
729         btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
730                                          &pending_snapshot->block_rsv,
731                                          pending_snapshot->qgroup_reserved);
732 free:
733         kfree(pending_snapshot);
734 out:
735         if (atomic_dec_and_test(&root->will_be_snapshoted))
736                 wake_up_atomic_t(&root->will_be_snapshoted);
737         return ret;
738 }
739 
740 /*  copy of may_delete in fs/namei.c()
741  *      Check whether we can remove a link victim from directory dir, check
742  *  whether the type of victim is right.
743  *  1. We can't do it if dir is read-only (done in permission())
744  *  2. We should have write and exec permissions on dir
745  *  3. We can't remove anything from append-only dir
746  *  4. We can't do anything with immutable dir (done in permission())
747  *  5. If the sticky bit on dir is set we should either
748  *      a. be owner of dir, or
749  *      b. be owner of victim, or
750  *      c. have CAP_FOWNER capability
751  *  6. If the victim is append-only or immutable we can't do antyhing with
752  *     links pointing to it.
753  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
754  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
755  *  9. We can't remove a root or mountpoint.
756  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
757  *     nfs_async_unlink().
758  */
759 
760 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
761 {
762         int error;
763 
764         if (!victim->d_inode)
765                 return -ENOENT;
766 
767         BUG_ON(victim->d_parent->d_inode != dir);
768         audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
769 
770         error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
771         if (error)
772                 return error;
773         if (IS_APPEND(dir))
774                 return -EPERM;
775         if (check_sticky(dir, victim->d_inode) || IS_APPEND(victim->d_inode) ||
776             IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
777                 return -EPERM;
778         if (isdir) {
779                 if (!d_is_dir(victim))
780                         return -ENOTDIR;
781                 if (IS_ROOT(victim))
782                         return -EBUSY;
783         } else if (d_is_dir(victim))
784                 return -EISDIR;
785         if (IS_DEADDIR(dir))
786                 return -ENOENT;
787         if (victim->d_flags & DCACHE_NFSFS_RENAMED)
788                 return -EBUSY;
789         return 0;
790 }
791 
792 /* copy of may_create in fs/namei.c() */
793 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
794 {
795         if (child->d_inode)
796                 return -EEXIST;
797         if (IS_DEADDIR(dir))
798                 return -ENOENT;
799         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
800 }
801 
802 /*
803  * Create a new subvolume below @parent.  This is largely modeled after
804  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
805  * inside this filesystem so it's quite a bit simpler.
806  */
807 static noinline int btrfs_mksubvol(struct path *parent,
808                                    char *name, int namelen,
809                                    struct btrfs_root *snap_src,
810                                    u64 *async_transid, bool readonly,
811                                    struct btrfs_qgroup_inherit *inherit)
812 {
813         struct inode *dir  = parent->dentry->d_inode;
814         struct dentry *dentry;
815         int error;
816 
817         error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
818         if (error == -EINTR)
819                 return error;
820 
821         dentry = lookup_one_len(name, parent->dentry, namelen);
822         error = PTR_ERR(dentry);
823         if (IS_ERR(dentry))
824                 goto out_unlock;
825 
826         error = -EEXIST;
827         if (dentry->d_inode)
828                 goto out_dput;
829 
830         error = btrfs_may_create(dir, dentry);
831         if (error)
832                 goto out_dput;
833 
834         /*
835          * even if this name doesn't exist, we may get hash collisions.
836          * check for them now when we can safely fail
837          */
838         error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
839                                                dir->i_ino, name,
840                                                namelen);
841         if (error)
842                 goto out_dput;
843 
844         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
845 
846         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
847                 goto out_up_read;
848 
849         if (snap_src) {
850                 error = create_snapshot(snap_src, dir, dentry, name, namelen,
851                                         async_transid, readonly, inherit);
852         } else {
853                 error = create_subvol(dir, dentry, name, namelen,
854                                       async_transid, inherit);
855         }
856         if (!error)
857                 fsnotify_mkdir(dir, dentry);
858 out_up_read:
859         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
860 out_dput:
861         dput(dentry);
862 out_unlock:
863         mutex_unlock(&dir->i_mutex);
864         return error;
865 }
866 
867 /*
868  * When we're defragging a range, we don't want to kick it off again
869  * if it is really just waiting for delalloc to send it down.
870  * If we find a nice big extent or delalloc range for the bytes in the
871  * file you want to defrag, we return 0 to let you know to skip this
872  * part of the file
873  */
874 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
875 {
876         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
877         struct extent_map *em = NULL;
878         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
879         u64 end;
880 
881         read_lock(&em_tree->lock);
882         em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
883         read_unlock(&em_tree->lock);
884 
885         if (em) {
886                 end = extent_map_end(em);
887                 free_extent_map(em);
888                 if (end - offset > thresh)
889                         return 0;
890         }
891         /* if we already have a nice delalloc here, just stop */
892         thresh /= 2;
893         end = count_range_bits(io_tree, &offset, offset + thresh,
894                                thresh, EXTENT_DELALLOC, 1);
895         if (end >= thresh)
896                 return 0;
897         return 1;
898 }
899 
900 /*
901  * helper function to walk through a file and find extents
902  * newer than a specific transid, and smaller than thresh.
903  *
904  * This is used by the defragging code to find new and small
905  * extents
906  */
907 static int find_new_extents(struct btrfs_root *root,
908                             struct inode *inode, u64 newer_than,
909                             u64 *off, u32 thresh)
910 {
911         struct btrfs_path *path;
912         struct btrfs_key min_key;
913         struct extent_buffer *leaf;
914         struct btrfs_file_extent_item *extent;
915         int type;
916         int ret;
917         u64 ino = btrfs_ino(inode);
918 
919         path = btrfs_alloc_path();
920         if (!path)
921                 return -ENOMEM;
922 
923         min_key.objectid = ino;
924         min_key.type = BTRFS_EXTENT_DATA_KEY;
925         min_key.offset = *off;
926 
927         while (1) {
928                 ret = btrfs_search_forward(root, &min_key, path, newer_than);
929                 if (ret != 0)
930                         goto none;
931 process_slot:
932                 if (min_key.objectid != ino)
933                         goto none;
934                 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
935                         goto none;
936 
937                 leaf = path->nodes[0];
938                 extent = btrfs_item_ptr(leaf, path->slots[0],
939                                         struct btrfs_file_extent_item);
940 
941                 type = btrfs_file_extent_type(leaf, extent);
942                 if (type == BTRFS_FILE_EXTENT_REG &&
943                     btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
944                     check_defrag_in_cache(inode, min_key.offset, thresh)) {
945                         *off = min_key.offset;
946                         btrfs_free_path(path);
947                         return 0;
948                 }
949 
950                 path->slots[0]++;
951                 if (path->slots[0] < btrfs_header_nritems(leaf)) {
952                         btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
953                         goto process_slot;
954                 }
955 
956                 if (min_key.offset == (u64)-1)
957                         goto none;
958 
959                 min_key.offset++;
960                 btrfs_release_path(path);
961         }
962 none:
963         btrfs_free_path(path);
964         return -ENOENT;
965 }
966 
967 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
968 {
969         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
970         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
971         struct extent_map *em;
972         u64 len = PAGE_CACHE_SIZE;
973 
974         /*
975          * hopefully we have this extent in the tree already, try without
976          * the full extent lock
977          */
978         read_lock(&em_tree->lock);
979         em = lookup_extent_mapping(em_tree, start, len);
980         read_unlock(&em_tree->lock);
981 
982         if (!em) {
983                 struct extent_state *cached = NULL;
984                 u64 end = start + len - 1;
985 
986                 /* get the big lock and read metadata off disk */
987                 lock_extent_bits(io_tree, start, end, 0, &cached);
988                 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
989                 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
990 
991                 if (IS_ERR(em))
992                         return NULL;
993         }
994 
995         return em;
996 }
997 
998 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
999 {
1000         struct extent_map *next;
1001         bool ret = true;
1002 
1003         /* this is the last extent */
1004         if (em->start + em->len >= i_size_read(inode))
1005                 return false;
1006 
1007         next = defrag_lookup_extent(inode, em->start + em->len);
1008         if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1009                 ret = false;
1010         else if ((em->block_start + em->block_len == next->block_start) &&
1011                  (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1012                 ret = false;
1013 
1014         free_extent_map(next);
1015         return ret;
1016 }
1017 
1018 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1019                                u64 *last_len, u64 *skip, u64 *defrag_end,
1020                                int compress)
1021 {
1022         struct extent_map *em;
1023         int ret = 1;
1024         bool next_mergeable = true;
1025 
1026         /*
1027          * make sure that once we start defragging an extent, we keep on
1028          * defragging it
1029          */
1030         if (start < *defrag_end)
1031                 return 1;
1032 
1033         *skip = 0;
1034 
1035         em = defrag_lookup_extent(inode, start);
1036         if (!em)
1037                 return 0;
1038 
1039         /* this will cover holes, and inline extents */
1040         if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1041                 ret = 0;
1042                 goto out;
1043         }
1044 
1045         next_mergeable = defrag_check_next_extent(inode, em);
1046         /*
1047          * we hit a real extent, if it is big or the next extent is not a
1048          * real extent, don't bother defragging it
1049          */
1050         if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1051             (em->len >= thresh || !next_mergeable))
1052                 ret = 0;
1053 out:
1054         /*
1055          * last_len ends up being a counter of how many bytes we've defragged.
1056          * every time we choose not to defrag an extent, we reset *last_len
1057          * so that the next tiny extent will force a defrag.
1058          *
1059          * The end result of this is that tiny extents before a single big
1060          * extent will force at least part of that big extent to be defragged.
1061          */
1062         if (ret) {
1063                 *defrag_end = extent_map_end(em);
1064         } else {
1065                 *last_len = 0;
1066                 *skip = extent_map_end(em);
1067                 *defrag_end = 0;
1068         }
1069 
1070         free_extent_map(em);
1071         return ret;
1072 }
1073 
1074 /*
1075  * it doesn't do much good to defrag one or two pages
1076  * at a time.  This pulls in a nice chunk of pages
1077  * to COW and defrag.
1078  *
1079  * It also makes sure the delalloc code has enough
1080  * dirty data to avoid making new small extents as part
1081  * of the defrag
1082  *
1083  * It's a good idea to start RA on this range
1084  * before calling this.
1085  */
1086 static int cluster_pages_for_defrag(struct inode *inode,
1087                                     struct page **pages,
1088                                     unsigned long start_index,
1089                                     unsigned long num_pages)
1090 {
1091         unsigned long file_end;
1092         u64 isize = i_size_read(inode);
1093         u64 page_start;
1094         u64 page_end;
1095         u64 page_cnt;
1096         int ret;
1097         int i;
1098         int i_done;
1099         struct btrfs_ordered_extent *ordered;
1100         struct extent_state *cached_state = NULL;
1101         struct extent_io_tree *tree;
1102         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1103 
1104         file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1105         if (!isize || start_index > file_end)
1106                 return 0;
1107 
1108         page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1109 
1110         ret = btrfs_delalloc_reserve_space(inode,
1111                                            page_cnt << PAGE_CACHE_SHIFT);
1112         if (ret)
1113                 return ret;
1114         i_done = 0;
1115         tree = &BTRFS_I(inode)->io_tree;
1116 
1117         /* step one, lock all the pages */
1118         for (i = 0; i < page_cnt; i++) {
1119                 struct page *page;
1120 again:
1121                 page = find_or_create_page(inode->i_mapping,
1122                                            start_index + i, mask);
1123                 if (!page)
1124                         break;
1125 
1126                 page_start = page_offset(page);
1127                 page_end = page_start + PAGE_CACHE_SIZE - 1;
1128                 while (1) {
1129                         lock_extent_bits(tree, page_start, page_end,
1130                                          0, &cached_state);
1131                         ordered = btrfs_lookup_ordered_extent(inode,
1132                                                               page_start);
1133                         unlock_extent_cached(tree, page_start, page_end,
1134                                              &cached_state, GFP_NOFS);
1135                         if (!ordered)
1136                                 break;
1137 
1138                         unlock_page(page);
1139                         btrfs_start_ordered_extent(inode, ordered, 1);
1140                         btrfs_put_ordered_extent(ordered);
1141                         lock_page(page);
1142                         /*
1143                          * we unlocked the page above, so we need check if
1144                          * it was released or not.
1145                          */
1146                         if (page->mapping != inode->i_mapping) {
1147                                 unlock_page(page);
1148                                 page_cache_release(page);
1149                                 goto again;
1150                         }
1151                 }
1152 
1153                 if (!PageUptodate(page)) {
1154                         btrfs_readpage(NULL, page);
1155                         lock_page(page);
1156                         if (!PageUptodate(page)) {
1157                                 unlock_page(page);
1158                                 page_cache_release(page);
1159                                 ret = -EIO;
1160                                 break;
1161                         }
1162                 }
1163 
1164                 if (page->mapping != inode->i_mapping) {
1165                         unlock_page(page);
1166                         page_cache_release(page);
1167                         goto again;
1168                 }
1169 
1170                 pages[i] = page;
1171                 i_done++;
1172         }
1173         if (!i_done || ret)
1174                 goto out;
1175 
1176         if (!(inode->i_sb->s_flags & MS_ACTIVE))
1177                 goto out;
1178 
1179         /*
1180          * so now we have a nice long stream of locked
1181          * and up to date pages, lets wait on them
1182          */
1183         for (i = 0; i < i_done; i++)
1184                 wait_on_page_writeback(pages[i]);
1185 
1186         page_start = page_offset(pages[0]);
1187         page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1188 
1189         lock_extent_bits(&BTRFS_I(inode)->io_tree,
1190                          page_start, page_end - 1, 0, &cached_state);
1191         clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1192                           page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1193                           EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1194                           &cached_state, GFP_NOFS);
1195 
1196         if (i_done != page_cnt) {
1197                 spin_lock(&BTRFS_I(inode)->lock);
1198                 BTRFS_I(inode)->outstanding_extents++;
1199                 spin_unlock(&BTRFS_I(inode)->lock);
1200                 btrfs_delalloc_release_space(inode,
1201                                      (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1202         }
1203 
1204 
1205         set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1206                           &cached_state, GFP_NOFS);
1207 
1208         unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1209                              page_start, page_end - 1, &cached_state,
1210                              GFP_NOFS);
1211 
1212         for (i = 0; i < i_done; i++) {
1213                 clear_page_dirty_for_io(pages[i]);
1214                 ClearPageChecked(pages[i]);
1215                 set_page_extent_mapped(pages[i]);
1216                 set_page_dirty(pages[i]);
1217                 unlock_page(pages[i]);
1218                 page_cache_release(pages[i]);
1219         }
1220         return i_done;
1221 out:
1222         for (i = 0; i < i_done; i++) {
1223                 unlock_page(pages[i]);
1224                 page_cache_release(pages[i]);
1225         }
1226         btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1227         return ret;
1228 
1229 }
1230 
1231 int btrfs_defrag_file(struct inode *inode, struct file *file,
1232                       struct btrfs_ioctl_defrag_range_args *range,
1233                       u64 newer_than, unsigned long max_to_defrag)
1234 {
1235         struct btrfs_root *root = BTRFS_I(inode)->root;
1236         struct file_ra_state *ra = NULL;
1237         unsigned long last_index;
1238         u64 isize = i_size_read(inode);
1239         u64 last_len = 0;
1240         u64 skip = 0;
1241         u64 defrag_end = 0;
1242         u64 newer_off = range->start;
1243         unsigned long i;
1244         unsigned long ra_index = 0;
1245         int ret;
1246         int defrag_count = 0;
1247         int compress_type = BTRFS_COMPRESS_ZLIB;
1248         u32 extent_thresh = range->extent_thresh;
1249         unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1250         unsigned long cluster = max_cluster;
1251         u64 new_align = ~((u64)128 * 1024 - 1);
1252         struct page **pages = NULL;
1253 
1254         if (isize == 0)
1255                 return 0;
1256 
1257         if (range->start >= isize)
1258                 return -EINVAL;
1259 
1260         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1261                 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1262                         return -EINVAL;
1263                 if (range->compress_type)
1264                         compress_type = range->compress_type;
1265         }
1266 
1267         if (extent_thresh == 0)
1268                 extent_thresh = 256 * 1024;
1269 
1270         /*
1271          * if we were not given a file, allocate a readahead
1272          * context
1273          */
1274         if (!file) {
1275                 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1276                 if (!ra)
1277                         return -ENOMEM;
1278                 file_ra_state_init(ra, inode->i_mapping);
1279         } else {
1280                 ra = &file->f_ra;
1281         }
1282 
1283         pages = kmalloc_array(max_cluster, sizeof(struct page *),
1284                         GFP_NOFS);
1285         if (!pages) {
1286                 ret = -ENOMEM;
1287                 goto out_ra;
1288         }
1289 
1290         /* find the last page to defrag */
1291         if (range->start + range->len > range->start) {
1292                 last_index = min_t(u64, isize - 1,
1293                          range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1294         } else {
1295                 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1296         }
1297 
1298         if (newer_than) {
1299                 ret = find_new_extents(root, inode, newer_than,
1300                                        &newer_off, 64 * 1024);
1301                 if (!ret) {
1302                         range->start = newer_off;
1303                         /*
1304                          * we always align our defrag to help keep
1305                          * the extents in the file evenly spaced
1306                          */
1307                         i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1308                 } else
1309                         goto out_ra;
1310         } else {
1311                 i = range->start >> PAGE_CACHE_SHIFT;
1312         }
1313         if (!max_to_defrag)
1314                 max_to_defrag = last_index + 1;
1315 
1316         /*
1317          * make writeback starts from i, so the defrag range can be
1318          * written sequentially.
1319          */
1320         if (i < inode->i_mapping->writeback_index)
1321                 inode->i_mapping->writeback_index = i;
1322 
1323         while (i <= last_index && defrag_count < max_to_defrag &&
1324                (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1325                 /*
1326                  * make sure we stop running if someone unmounts
1327                  * the FS
1328                  */
1329                 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1330                         break;
1331 
1332                 if (btrfs_defrag_cancelled(root->fs_info)) {
1333                         printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1334                         ret = -EAGAIN;
1335                         break;
1336                 }
1337 
1338                 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1339                                          extent_thresh, &last_len, &skip,
1340                                          &defrag_end, range->flags &
1341                                          BTRFS_DEFRAG_RANGE_COMPRESS)) {
1342                         unsigned long next;
1343                         /*
1344                          * the should_defrag function tells us how much to skip
1345                          * bump our counter by the suggested amount
1346                          */
1347                         next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1348                         i = max(i + 1, next);
1349                         continue;
1350                 }
1351 
1352                 if (!newer_than) {
1353                         cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1354                                    PAGE_CACHE_SHIFT) - i;
1355                         cluster = min(cluster, max_cluster);
1356                 } else {
1357                         cluster = max_cluster;
1358                 }
1359 
1360                 if (i + cluster > ra_index) {
1361                         ra_index = max(i, ra_index);
1362                         btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1363                                        cluster);
1364                         ra_index += max_cluster;
1365                 }
1366 
1367                 mutex_lock(&inode->i_mutex);
1368                 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1369                         BTRFS_I(inode)->force_compress = compress_type;
1370                 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1371                 if (ret < 0) {
1372                         mutex_unlock(&inode->i_mutex);
1373                         goto out_ra;
1374                 }
1375 
1376                 defrag_count += ret;
1377                 balance_dirty_pages_ratelimited(inode->i_mapping);
1378                 mutex_unlock(&inode->i_mutex);
1379 
1380                 if (newer_than) {
1381                         if (newer_off == (u64)-1)
1382                                 break;
1383 
1384                         if (ret > 0)
1385                                 i += ret;
1386 
1387                         newer_off = max(newer_off + 1,
1388                                         (u64)i << PAGE_CACHE_SHIFT);
1389 
1390                         ret = find_new_extents(root, inode,
1391                                                newer_than, &newer_off,
1392                                                64 * 1024);
1393                         if (!ret) {
1394                                 range->start = newer_off;
1395                                 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1396                         } else {
1397                                 break;
1398                         }
1399                 } else {
1400                         if (ret > 0) {
1401                                 i += ret;
1402                                 last_len += ret << PAGE_CACHE_SHIFT;
1403                         } else {
1404                                 i++;
1405                                 last_len = 0;
1406                         }
1407                 }
1408         }
1409 
1410         if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1411                 filemap_flush(inode->i_mapping);
1412                 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1413                              &BTRFS_I(inode)->runtime_flags))
1414                         filemap_flush(inode->i_mapping);
1415         }
1416 
1417         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1418                 /* the filemap_flush will queue IO into the worker threads, but
1419                  * we have to make sure the IO is actually started and that
1420                  * ordered extents get created before we return
1421                  */
1422                 atomic_inc(&root->fs_info->async_submit_draining);
1423                 while (atomic_read(&root->fs_info->nr_async_submits) ||
1424                       atomic_read(&root->fs_info->async_delalloc_pages)) {
1425                         wait_event(root->fs_info->async_submit_wait,
1426                            (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1427                             atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1428                 }
1429                 atomic_dec(&root->fs_info->async_submit_draining);
1430         }
1431 
1432         if (range->compress_type == BTRFS_COMPRESS_LZO) {
1433                 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1434         }
1435 
1436         ret = defrag_count;
1437 
1438 out_ra:
1439         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1440                 mutex_lock(&inode->i_mutex);
1441                 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1442                 mutex_unlock(&inode->i_mutex);
1443         }
1444         if (!file)
1445                 kfree(ra);
1446         kfree(pages);
1447         return ret;
1448 }
1449 
1450 static noinline int btrfs_ioctl_resize(struct file *file,
1451                                         void __user *arg)
1452 {
1453         u64 new_size;
1454         u64 old_size;
1455         u64 devid = 1;
1456         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1457         struct btrfs_ioctl_vol_args *vol_args;
1458         struct btrfs_trans_handle *trans;
1459         struct btrfs_device *device = NULL;
1460         char *sizestr;
1461         char *retptr;
1462         char *devstr = NULL;
1463         int ret = 0;
1464         int mod = 0;
1465 
1466         if (!capable(CAP_SYS_ADMIN))
1467                 return -EPERM;
1468 
1469         ret = mnt_want_write_file(file);
1470         if (ret)
1471                 return ret;
1472 
1473         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1474                         1)) {
1475                 mnt_drop_write_file(file);
1476                 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1477         }
1478 
1479         mutex_lock(&root->fs_info->volume_mutex);
1480         vol_args = memdup_user(arg, sizeof(*vol_args));
1481         if (IS_ERR(vol_args)) {
1482                 ret = PTR_ERR(vol_args);
1483                 goto out;
1484         }
1485 
1486         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1487 
1488         sizestr = vol_args->name;
1489         devstr = strchr(sizestr, ':');
1490         if (devstr) {
1491                 sizestr = devstr + 1;
1492                 *devstr = '\0';
1493                 devstr = vol_args->name;
1494                 ret = kstrtoull(devstr, 10, &devid);
1495                 if (ret)
1496                         goto out_free;
1497                 if (!devid) {
1498                         ret = -EINVAL;
1499                         goto out_free;
1500                 }
1501                 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1502         }
1503 
1504         device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1505         if (!device) {
1506                 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1507                        devid);
1508                 ret = -ENODEV;
1509                 goto out_free;
1510         }
1511 
1512         if (!device->writeable) {
1513                 btrfs_info(root->fs_info,
1514                            "resizer unable to apply on readonly device %llu",
1515                        devid);
1516                 ret = -EPERM;
1517                 goto out_free;
1518         }
1519 
1520         if (!strcmp(sizestr, "max"))
1521                 new_size = device->bdev->bd_inode->i_size;
1522         else {
1523                 if (sizestr[0] == '-') {
1524                         mod = -1;
1525                         sizestr++;
1526                 } else if (sizestr[0] == '+') {
1527                         mod = 1;
1528                         sizestr++;
1529                 }
1530                 new_size = memparse(sizestr, &retptr);
1531                 if (*retptr != '\0' || new_size == 0) {
1532                         ret = -EINVAL;
1533                         goto out_free;
1534                 }
1535         }
1536 
1537         if (device->is_tgtdev_for_dev_replace) {
1538                 ret = -EPERM;
1539                 goto out_free;
1540         }
1541 
1542         old_size = btrfs_device_get_total_bytes(device);
1543 
1544         if (mod < 0) {
1545                 if (new_size > old_size) {
1546                         ret = -EINVAL;
1547                         goto out_free;
1548                 }
1549                 new_size = old_size - new_size;
1550         } else if (mod > 0) {
1551                 if (new_size > ULLONG_MAX - old_size) {
1552                         ret = -ERANGE;
1553                         goto out_free;
1554                 }
1555                 new_size = old_size + new_size;
1556         }
1557 
1558         if (new_size < 256 * 1024 * 1024) {
1559                 ret = -EINVAL;
1560                 goto out_free;
1561         }
1562         if (new_size > device->bdev->bd_inode->i_size) {
1563                 ret = -EFBIG;
1564                 goto out_free;
1565         }
1566 
1567         do_div(new_size, root->sectorsize);
1568         new_size *= root->sectorsize;
1569 
1570         printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1571                       rcu_str_deref(device->name), new_size);
1572 
1573         if (new_size > old_size) {
1574                 trans = btrfs_start_transaction(root, 0);
1575                 if (IS_ERR(trans)) {
1576                         ret = PTR_ERR(trans);
1577                         goto out_free;
1578                 }
1579                 ret = btrfs_grow_device(trans, device, new_size);
1580                 btrfs_commit_transaction(trans, root);
1581         } else if (new_size < old_size) {
1582                 ret = btrfs_shrink_device(device, new_size);
1583         } /* equal, nothing need to do */
1584 
1585 out_free:
1586         kfree(vol_args);
1587 out:
1588         mutex_unlock(&root->fs_info->volume_mutex);
1589         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1590         mnt_drop_write_file(file);
1591         return ret;
1592 }
1593 
1594 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1595                                 char *name, unsigned long fd, int subvol,
1596                                 u64 *transid, bool readonly,
1597                                 struct btrfs_qgroup_inherit *inherit)
1598 {
1599         int namelen;
1600         int ret = 0;
1601 
1602         ret = mnt_want_write_file(file);
1603         if (ret)
1604                 goto out;
1605 
1606         namelen = strlen(name);
1607         if (strchr(name, '/')) {
1608                 ret = -EINVAL;
1609                 goto out_drop_write;
1610         }
1611 
1612         if (name[0] == '.' &&
1613            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1614                 ret = -EEXIST;
1615                 goto out_drop_write;
1616         }
1617 
1618         if (subvol) {
1619                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1620                                      NULL, transid, readonly, inherit);
1621         } else {
1622                 struct fd src = fdget(fd);
1623                 struct inode *src_inode;
1624                 if (!src.file) {
1625                         ret = -EINVAL;
1626                         goto out_drop_write;
1627                 }
1628 
1629                 src_inode = file_inode(src.file);
1630                 if (src_inode->i_sb != file_inode(file)->i_sb) {
1631                         btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1632                                    "Snapshot src from another FS");
1633                         ret = -EXDEV;
1634                 } else if (!inode_owner_or_capable(src_inode)) {
1635                         /*
1636                          * Subvolume creation is not restricted, but snapshots
1637                          * are limited to own subvolumes only
1638                          */
1639                         ret = -EPERM;
1640                 } else {
1641                         ret = btrfs_mksubvol(&file->f_path, name, namelen,
1642                                              BTRFS_I(src_inode)->root,
1643                                              transid, readonly, inherit);
1644                 }
1645                 fdput(src);
1646         }
1647 out_drop_write:
1648         mnt_drop_write_file(file);
1649 out:
1650         return ret;
1651 }
1652 
1653 static noinline int btrfs_ioctl_snap_create(struct file *file,
1654                                             void __user *arg, int subvol)
1655 {
1656         struct btrfs_ioctl_vol_args *vol_args;
1657         int ret;
1658 
1659         vol_args = memdup_user(arg, sizeof(*vol_args));
1660         if (IS_ERR(vol_args))
1661                 return PTR_ERR(vol_args);
1662         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1663 
1664         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1665                                               vol_args->fd, subvol,
1666                                               NULL, false, NULL);
1667 
1668         kfree(vol_args);
1669         return ret;
1670 }
1671 
1672 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1673                                                void __user *arg, int subvol)
1674 {
1675         struct btrfs_ioctl_vol_args_v2 *vol_args;
1676         int ret;
1677         u64 transid = 0;
1678         u64 *ptr = NULL;
1679         bool readonly = false;
1680         struct btrfs_qgroup_inherit *inherit = NULL;
1681 
1682         vol_args = memdup_user(arg, sizeof(*vol_args));
1683         if (IS_ERR(vol_args))
1684                 return PTR_ERR(vol_args);
1685         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1686 
1687         if (vol_args->flags &
1688             ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1689               BTRFS_SUBVOL_QGROUP_INHERIT)) {
1690                 ret = -EOPNOTSUPP;
1691                 goto free_args;
1692         }
1693 
1694         if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1695                 ptr = &transid;
1696         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1697                 readonly = true;
1698         if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1699                 if (vol_args->size > PAGE_CACHE_SIZE) {
1700                         ret = -EINVAL;
1701                         goto free_args;
1702                 }
1703                 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1704                 if (IS_ERR(inherit)) {
1705                         ret = PTR_ERR(inherit);
1706                         goto free_args;
1707                 }
1708         }
1709 
1710         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1711                                               vol_args->fd, subvol, ptr,
1712                                               readonly, inherit);
1713         if (ret)
1714                 goto free_inherit;
1715 
1716         if (ptr && copy_to_user(arg +
1717                                 offsetof(struct btrfs_ioctl_vol_args_v2,
1718                                         transid),
1719                                 ptr, sizeof(*ptr)))
1720                 ret = -EFAULT;
1721 
1722 free_inherit:
1723         kfree(inherit);
1724 free_args:
1725         kfree(vol_args);
1726         return ret;
1727 }
1728 
1729 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1730                                                 void __user *arg)
1731 {
1732         struct inode *inode = file_inode(file);
1733         struct btrfs_root *root = BTRFS_I(inode)->root;
1734         int ret = 0;
1735         u64 flags = 0;
1736 
1737         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1738                 return -EINVAL;
1739 
1740         down_read(&root->fs_info->subvol_sem);
1741         if (btrfs_root_readonly(root))
1742                 flags |= BTRFS_SUBVOL_RDONLY;
1743         up_read(&root->fs_info->subvol_sem);
1744 
1745         if (copy_to_user(arg, &flags, sizeof(flags)))
1746                 ret = -EFAULT;
1747 
1748         return ret;
1749 }
1750 
1751 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1752                                               void __user *arg)
1753 {
1754         struct inode *inode = file_inode(file);
1755         struct btrfs_root *root = BTRFS_I(inode)->root;
1756         struct btrfs_trans_handle *trans;
1757         u64 root_flags;
1758         u64 flags;
1759         int ret = 0;
1760 
1761         if (!inode_owner_or_capable(inode))
1762                 return -EPERM;
1763 
1764         ret = mnt_want_write_file(file);
1765         if (ret)
1766                 goto out;
1767 
1768         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1769                 ret = -EINVAL;
1770                 goto out_drop_write;
1771         }
1772 
1773         if (copy_from_user(&flags, arg, sizeof(flags))) {
1774                 ret = -EFAULT;
1775                 goto out_drop_write;
1776         }
1777 
1778         if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1779                 ret = -EINVAL;
1780                 goto out_drop_write;
1781         }
1782 
1783         if (flags & ~BTRFS_SUBVOL_RDONLY) {
1784                 ret = -EOPNOTSUPP;
1785                 goto out_drop_write;
1786         }
1787 
1788         down_write(&root->fs_info->subvol_sem);
1789 
1790         /* nothing to do */
1791         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1792                 goto out_drop_sem;
1793 
1794         root_flags = btrfs_root_flags(&root->root_item);
1795         if (flags & BTRFS_SUBVOL_RDONLY) {
1796                 btrfs_set_root_flags(&root->root_item,
1797                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1798         } else {
1799                 /*
1800                  * Block RO -> RW transition if this subvolume is involved in
1801                  * send
1802                  */
1803                 spin_lock(&root->root_item_lock);
1804                 if (root->send_in_progress == 0) {
1805                         btrfs_set_root_flags(&root->root_item,
1806                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1807                         spin_unlock(&root->root_item_lock);
1808                 } else {
1809                         spin_unlock(&root->root_item_lock);
1810                         btrfs_warn(root->fs_info,
1811                         "Attempt to set subvolume %llu read-write during send",
1812                                         root->root_key.objectid);
1813                         ret = -EPERM;
1814                         goto out_drop_sem;
1815                 }
1816         }
1817 
1818         trans = btrfs_start_transaction(root, 1);
1819         if (IS_ERR(trans)) {
1820                 ret = PTR_ERR(trans);
1821                 goto out_reset;
1822         }
1823 
1824         ret = btrfs_update_root(trans, root->fs_info->tree_root,
1825                                 &root->root_key, &root->root_item);
1826 
1827         btrfs_commit_transaction(trans, root);
1828 out_reset:
1829         if (ret)
1830                 btrfs_set_root_flags(&root->root_item, root_flags);
1831 out_drop_sem:
1832         up_write(&root->fs_info->subvol_sem);
1833 out_drop_write:
1834         mnt_drop_write_file(file);
1835 out:
1836         return ret;
1837 }
1838 
1839 /*
1840  * helper to check if the subvolume references other subvolumes
1841  */
1842 static noinline int may_destroy_subvol(struct btrfs_root *root)
1843 {
1844         struct btrfs_path *path;
1845         struct btrfs_dir_item *di;
1846         struct btrfs_key key;
1847         u64 dir_id;
1848         int ret;
1849 
1850         path = btrfs_alloc_path();
1851         if (!path)
1852                 return -ENOMEM;
1853 
1854         /* Make sure this root isn't set as the default subvol */
1855         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1856         di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1857                                    dir_id, "default", 7, 0);
1858         if (di && !IS_ERR(di)) {
1859                 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1860                 if (key.objectid == root->root_key.objectid) {
1861                         ret = -EPERM;
1862                         btrfs_err(root->fs_info, "deleting default subvolume "
1863                                   "%llu is not allowed", key.objectid);
1864                         goto out;
1865                 }
1866                 btrfs_release_path(path);
1867         }
1868 
1869         key.objectid = root->root_key.objectid;
1870         key.type = BTRFS_ROOT_REF_KEY;
1871         key.offset = (u64)-1;
1872 
1873         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1874                                 &key, path, 0, 0);
1875         if (ret < 0)
1876                 goto out;
1877         BUG_ON(ret == 0);
1878 
1879         ret = 0;
1880         if (path->slots[0] > 0) {
1881                 path->slots[0]--;
1882                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1883                 if (key.objectid == root->root_key.objectid &&
1884                     key.type == BTRFS_ROOT_REF_KEY)
1885                         ret = -ENOTEMPTY;
1886         }
1887 out:
1888         btrfs_free_path(path);
1889         return ret;
1890 }
1891 
1892 static noinline int key_in_sk(struct btrfs_key *key,
1893                               struct btrfs_ioctl_search_key *sk)
1894 {
1895         struct btrfs_key test;
1896         int ret;
1897 
1898         test.objectid = sk->min_objectid;
1899         test.type = sk->min_type;
1900         test.offset = sk->min_offset;
1901 
1902         ret = btrfs_comp_cpu_keys(key, &test);
1903         if (ret < 0)
1904                 return 0;
1905 
1906         test.objectid = sk->max_objectid;
1907         test.type = sk->max_type;
1908         test.offset = sk->max_offset;
1909 
1910         ret = btrfs_comp_cpu_keys(key, &test);
1911         if (ret > 0)
1912                 return 0;
1913         return 1;
1914 }
1915 
1916 static noinline int copy_to_sk(struct btrfs_root *root,
1917                                struct btrfs_path *path,
1918                                struct btrfs_key *key,
1919                                struct btrfs_ioctl_search_key *sk,
1920                                size_t *buf_size,
1921                                char __user *ubuf,
1922                                unsigned long *sk_offset,
1923                                int *num_found)
1924 {
1925         u64 found_transid;
1926         struct extent_buffer *leaf;
1927         struct btrfs_ioctl_search_header sh;
1928         unsigned long item_off;
1929         unsigned long item_len;
1930         int nritems;
1931         int i;
1932         int slot;
1933         int ret = 0;
1934 
1935         leaf = path->nodes[0];
1936         slot = path->slots[0];
1937         nritems = btrfs_header_nritems(leaf);
1938 
1939         if (btrfs_header_generation(leaf) > sk->max_transid) {
1940                 i = nritems;
1941                 goto advance_key;
1942         }
1943         found_transid = btrfs_header_generation(leaf);
1944 
1945         for (i = slot; i < nritems; i++) {
1946                 item_off = btrfs_item_ptr_offset(leaf, i);
1947                 item_len = btrfs_item_size_nr(leaf, i);
1948 
1949                 btrfs_item_key_to_cpu(leaf, key, i);
1950                 if (!key_in_sk(key, sk))
1951                         continue;
1952 
1953                 if (sizeof(sh) + item_len > *buf_size) {
1954                         if (*num_found) {
1955                                 ret = 1;
1956                                 goto out;
1957                         }
1958 
1959                         /*
1960                          * return one empty item back for v1, which does not
1961                          * handle -EOVERFLOW
1962                          */
1963 
1964                         *buf_size = sizeof(sh) + item_len;
1965                         item_len = 0;
1966                         ret = -EOVERFLOW;
1967                 }
1968 
1969                 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1970                         ret = 1;
1971                         goto out;
1972                 }
1973 
1974                 sh.objectid = key->objectid;
1975                 sh.offset = key->offset;
1976                 sh.type = key->type;
1977                 sh.len = item_len;
1978                 sh.transid = found_transid;
1979 
1980                 /* copy search result header */
1981                 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1982                         ret = -EFAULT;
1983                         goto out;
1984                 }
1985 
1986                 *sk_offset += sizeof(sh);
1987 
1988                 if (item_len) {
1989                         char __user *up = ubuf + *sk_offset;
1990                         /* copy the item */
1991                         if (read_extent_buffer_to_user(leaf, up,
1992                                                        item_off, item_len)) {
1993                                 ret = -EFAULT;
1994                                 goto out;
1995                         }
1996 
1997                         *sk_offset += item_len;
1998                 }
1999                 (*num_found)++;
2000 
2001                 if (ret) /* -EOVERFLOW from above */
2002                         goto out;
2003 
2004                 if (*num_found >= sk->nr_items) {
2005                         ret = 1;
2006                         goto out;
2007                 }
2008         }
2009 advance_key:
2010         ret = 0;
2011         if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2012                 key->offset++;
2013         else if (key->type < (u8)-1 && key->type < sk->max_type) {
2014                 key->offset = 0;
2015                 key->type++;
2016         } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2017                 key->offset = 0;
2018                 key->type = 0;
2019                 key->objectid++;
2020         } else
2021                 ret = 1;
2022 out:
2023         /*
2024          *  0: all items from this leaf copied, continue with next
2025          *  1: * more items can be copied, but unused buffer is too small
2026          *     * all items were found
2027          *     Either way, it will stops the loop which iterates to the next
2028          *     leaf
2029          *  -EOVERFLOW: item was to large for buffer
2030          *  -EFAULT: could not copy extent buffer back to userspace
2031          */
2032         return ret;
2033 }
2034 
2035 static noinline int search_ioctl(struct inode *inode,
2036                                  struct btrfs_ioctl_search_key *sk,
2037                                  size_t *buf_size,
2038                                  char __user *ubuf)
2039 {
2040         struct btrfs_root *root;
2041         struct btrfs_key key;
2042         struct btrfs_path *path;
2043         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2044         int ret;
2045         int num_found = 0;
2046         unsigned long sk_offset = 0;
2047 
2048         if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2049                 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2050                 return -EOVERFLOW;
2051         }
2052 
2053         path = btrfs_alloc_path();
2054         if (!path)
2055                 return -ENOMEM;
2056 
2057         if (sk->tree_id == 0) {
2058                 /* search the root of the inode that was passed */
2059                 root = BTRFS_I(inode)->root;
2060         } else {
2061                 key.objectid = sk->tree_id;
2062                 key.type = BTRFS_ROOT_ITEM_KEY;
2063                 key.offset = (u64)-1;
2064                 root = btrfs_read_fs_root_no_name(info, &key);
2065                 if (IS_ERR(root)) {
2066                         printk(KERN_ERR "BTRFS: could not find root %llu\n",
2067                                sk->tree_id);
2068                         btrfs_free_path(path);
2069                         return -ENOENT;
2070                 }
2071         }
2072 
2073         key.objectid = sk->min_objectid;
2074         key.type = sk->min_type;
2075         key.offset = sk->min_offset;
2076 
2077         while (1) {
2078                 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2079                 if (ret != 0) {
2080                         if (ret > 0)
2081                                 ret = 0;
2082                         goto err;
2083                 }
2084                 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2085                                  &sk_offset, &num_found);
2086                 btrfs_release_path(path);
2087                 if (ret)
2088                         break;
2089 
2090         }
2091         if (ret > 0)
2092                 ret = 0;
2093 err:
2094         sk->nr_items = num_found;
2095         btrfs_free_path(path);
2096         return ret;
2097 }
2098 
2099 static noinline int btrfs_ioctl_tree_search(struct file *file,
2100                                            void __user *argp)
2101 {
2102         struct btrfs_ioctl_search_args __user *uargs;
2103         struct btrfs_ioctl_search_key sk;
2104         struct inode *inode;
2105         int ret;
2106         size_t buf_size;
2107 
2108         if (!capable(CAP_SYS_ADMIN))
2109                 return -EPERM;
2110 
2111         uargs = (struct btrfs_ioctl_search_args __user *)argp;
2112 
2113         if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2114                 return -EFAULT;
2115 
2116         buf_size = sizeof(uargs->buf);
2117 
2118         inode = file_inode(file);
2119         ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2120 
2121         /*
2122          * In the origin implementation an overflow is handled by returning a
2123          * search header with a len of zero, so reset ret.
2124          */
2125         if (ret == -EOVERFLOW)
2126                 ret = 0;
2127 
2128         if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2129                 ret = -EFAULT;
2130         return ret;
2131 }
2132 
2133 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2134                                                void __user *argp)
2135 {
2136         struct btrfs_ioctl_search_args_v2 __user *uarg;
2137         struct btrfs_ioctl_search_args_v2 args;
2138         struct inode *inode;
2139         int ret;
2140         size_t buf_size;
2141         const size_t buf_limit = 16 * 1024 * 1024;
2142 
2143         if (!capable(CAP_SYS_ADMIN))
2144                 return -EPERM;
2145 
2146         /* copy search header and buffer size */
2147         uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2148         if (copy_from_user(&args, uarg, sizeof(args)))
2149                 return -EFAULT;
2150 
2151         buf_size = args.buf_size;
2152 
2153         if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2154                 return -EOVERFLOW;
2155 
2156         /* limit result size to 16MB */
2157         if (buf_size > buf_limit)
2158                 buf_size = buf_limit;
2159 
2160         inode = file_inode(file);
2161         ret = search_ioctl(inode, &args.key, &buf_size,
2162                            (char *)(&uarg->buf[0]));
2163         if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2164                 ret = -EFAULT;
2165         else if (ret == -EOVERFLOW &&
2166                 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2167                 ret = -EFAULT;
2168 
2169         return ret;
2170 }
2171 
2172 /*
2173  * Search INODE_REFs to identify path name of 'dirid' directory
2174  * in a 'tree_id' tree. and sets path name to 'name'.
2175  */
2176 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2177                                 u64 tree_id, u64 dirid, char *name)
2178 {
2179         struct btrfs_root *root;
2180         struct btrfs_key key;
2181         char *ptr;
2182         int ret = -1;
2183         int slot;
2184         int len;
2185         int total_len = 0;
2186         struct btrfs_inode_ref *iref;
2187         struct extent_buffer *l;
2188         struct btrfs_path *path;
2189 
2190         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2191                 name[0]='\0';
2192                 return 0;
2193         }
2194 
2195         path = btrfs_alloc_path();
2196         if (!path)
2197                 return -ENOMEM;
2198 
2199         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2200 
2201         key.objectid = tree_id;
2202         key.type = BTRFS_ROOT_ITEM_KEY;
2203         key.offset = (u64)-1;
2204         root = btrfs_read_fs_root_no_name(info, &key);
2205         if (IS_ERR(root)) {
2206                 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2207                 ret = -ENOENT;
2208                 goto out;
2209         }
2210 
2211         key.objectid = dirid;
2212         key.type = BTRFS_INODE_REF_KEY;
2213         key.offset = (u64)-1;
2214 
2215         while (1) {
2216                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2217                 if (ret < 0)
2218                         goto out;
2219                 else if (ret > 0) {
2220                         ret = btrfs_previous_item(root, path, dirid,
2221                                                   BTRFS_INODE_REF_KEY);
2222                         if (ret < 0)
2223                                 goto out;
2224                         else if (ret > 0) {
2225                                 ret = -ENOENT;
2226                                 goto out;
2227                         }
2228                 }
2229 
2230                 l = path->nodes[0];
2231                 slot = path->slots[0];
2232                 btrfs_item_key_to_cpu(l, &key, slot);
2233 
2234                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2235                 len = btrfs_inode_ref_name_len(l, iref);
2236                 ptr -= len + 1;
2237                 total_len += len + 1;
2238                 if (ptr < name) {
2239                         ret = -ENAMETOOLONG;
2240                         goto out;
2241                 }
2242 
2243                 *(ptr + len) = '/';
2244                 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2245 
2246                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2247                         break;
2248 
2249                 btrfs_release_path(path);
2250                 key.objectid = key.offset;
2251                 key.offset = (u64)-1;
2252                 dirid = key.objectid;
2253         }
2254         memmove(name, ptr, total_len);
2255         name[total_len] = '\0';
2256         ret = 0;
2257 out:
2258         btrfs_free_path(path);
2259         return ret;
2260 }
2261 
2262 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2263                                            void __user *argp)
2264 {
2265          struct btrfs_ioctl_ino_lookup_args *args;
2266          struct inode *inode;
2267          int ret;
2268 
2269         if (!capable(CAP_SYS_ADMIN))
2270                 return -EPERM;
2271 
2272         args = memdup_user(argp, sizeof(*args));
2273         if (IS_ERR(args))
2274                 return PTR_ERR(args);
2275 
2276         inode = file_inode(file);
2277 
2278         if (args->treeid == 0)
2279                 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2280 
2281         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2282                                         args->treeid, args->objectid,
2283                                         args->name);
2284 
2285         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2286                 ret = -EFAULT;
2287 
2288         kfree(args);
2289         return ret;
2290 }
2291 
2292 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2293                                              void __user *arg)
2294 {
2295         struct dentry *parent = file->f_path.dentry;
2296         struct dentry *dentry;
2297         struct inode *dir = parent->d_inode;
2298         struct inode *inode;
2299         struct btrfs_root *root = BTRFS_I(dir)->root;
2300         struct btrfs_root *dest = NULL;
2301         struct btrfs_ioctl_vol_args *vol_args;
2302         struct btrfs_trans_handle *trans;
2303         struct btrfs_block_rsv block_rsv;
2304         u64 root_flags;
2305         u64 qgroup_reserved;
2306         int namelen;
2307         int ret;
2308         int err = 0;
2309 
2310         vol_args = memdup_user(arg, sizeof(*vol_args));
2311         if (IS_ERR(vol_args))
2312                 return PTR_ERR(vol_args);
2313 
2314         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2315         namelen = strlen(vol_args->name);
2316         if (strchr(vol_args->name, '/') ||
2317             strncmp(vol_args->name, "..", namelen) == 0) {
2318                 err = -EINVAL;
2319                 goto out;
2320         }
2321 
2322         err = mnt_want_write_file(file);
2323         if (err)
2324                 goto out;
2325 
2326 
2327         err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2328         if (err == -EINTR)
2329                 goto out_drop_write;
2330         dentry = lookup_one_len(vol_args->name, parent, namelen);
2331         if (IS_ERR(dentry)) {
2332                 err = PTR_ERR(dentry);
2333                 goto out_unlock_dir;
2334         }
2335 
2336         if (!dentry->d_inode) {
2337                 err = -ENOENT;
2338                 goto out_dput;
2339         }
2340 
2341         inode = dentry->d_inode;
2342         dest = BTRFS_I(inode)->root;
2343         if (!capable(CAP_SYS_ADMIN)) {
2344                 /*
2345                  * Regular user.  Only allow this with a special mount
2346                  * option, when the user has write+exec access to the
2347                  * subvol root, and when rmdir(2) would have been
2348                  * allowed.
2349                  *
2350                  * Note that this is _not_ check that the subvol is
2351                  * empty or doesn't contain data that we wouldn't
2352                  * otherwise be able to delete.
2353                  *
2354                  * Users who want to delete empty subvols should try
2355                  * rmdir(2).
2356                  */
2357                 err = -EPERM;
2358                 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2359                         goto out_dput;
2360 
2361                 /*
2362                  * Do not allow deletion if the parent dir is the same
2363                  * as the dir to be deleted.  That means the ioctl
2364                  * must be called on the dentry referencing the root
2365                  * of the subvol, not a random directory contained
2366                  * within it.
2367                  */
2368                 err = -EINVAL;
2369                 if (root == dest)
2370                         goto out_dput;
2371 
2372                 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2373                 if (err)
2374                         goto out_dput;
2375         }
2376 
2377         /* check if subvolume may be deleted by a user */
2378         err = btrfs_may_delete(dir, dentry, 1);
2379         if (err)
2380                 goto out_dput;
2381 
2382         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2383                 err = -EINVAL;
2384                 goto out_dput;
2385         }
2386 
2387         mutex_lock(&inode->i_mutex);
2388 
2389         /*
2390          * Don't allow to delete a subvolume with send in progress. This is
2391          * inside the i_mutex so the error handling that has to drop the bit
2392          * again is not run concurrently.
2393          */
2394         spin_lock(&dest->root_item_lock);
2395         root_flags = btrfs_root_flags(&dest->root_item);
2396         if (dest->send_in_progress == 0) {
2397                 btrfs_set_root_flags(&dest->root_item,
2398                                 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2399                 spin_unlock(&dest->root_item_lock);
2400         } else {
2401                 spin_unlock(&dest->root_item_lock);
2402                 btrfs_warn(root->fs_info,
2403                         "Attempt to delete subvolume %llu during send",
2404                         dest->root_key.objectid);
2405                 err = -EPERM;
2406                 goto out_unlock_inode;
2407         }
2408 
2409         d_invalidate(dentry);
2410 
2411         down_write(&root->fs_info->subvol_sem);
2412 
2413         err = may_destroy_subvol(dest);
2414         if (err)
2415                 goto out_up_write;
2416 
2417         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2418         /*
2419          * One for dir inode, two for dir entries, two for root
2420          * ref/backref.
2421          */
2422         err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2423                                                5, &qgroup_reserved, true);
2424         if (err)
2425                 goto out_up_write;
2426 
2427         trans = btrfs_start_transaction(root, 0);
2428         if (IS_ERR(trans)) {
2429                 err = PTR_ERR(trans);
2430                 goto out_release;
2431         }
2432         trans->block_rsv = &block_rsv;
2433         trans->bytes_reserved = block_rsv.size;
2434 
2435         ret = btrfs_unlink_subvol(trans, root, dir,
2436                                 dest->root_key.objectid,
2437                                 dentry->d_name.name,
2438                                 dentry->d_name.len);
2439         if (ret) {
2440                 err = ret;
2441                 btrfs_abort_transaction(trans, root, ret);
2442                 goto out_end_trans;
2443         }
2444 
2445         btrfs_record_root_in_trans(trans, dest);
2446 
2447         memset(&dest->root_item.drop_progress, 0,
2448                 sizeof(dest->root_item.drop_progress));
2449         dest->root_item.drop_level = 0;
2450         btrfs_set_root_refs(&dest->root_item, 0);
2451 
2452         if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2453                 ret = btrfs_insert_orphan_item(trans,
2454                                         root->fs_info->tree_root,
2455                                         dest->root_key.objectid);
2456                 if (ret) {
2457                         btrfs_abort_transaction(trans, root, ret);
2458                         err = ret;
2459                         goto out_end_trans;
2460                 }
2461         }
2462 
2463         ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2464                                   dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2465                                   dest->root_key.objectid);
2466         if (ret && ret != -ENOENT) {
2467                 btrfs_abort_transaction(trans, root, ret);
2468                 err = ret;
2469                 goto out_end_trans;
2470         }
2471         if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2472                 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2473                                           dest->root_item.received_uuid,
2474                                           BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2475                                           dest->root_key.objectid);
2476                 if (ret && ret != -ENOENT) {
2477                         btrfs_abort_transaction(trans, root, ret);
2478                         err = ret;
2479                         goto out_end_trans;
2480                 }
2481         }
2482 
2483 out_end_trans:
2484         trans->block_rsv = NULL;
2485         trans->bytes_reserved = 0;
2486         ret = btrfs_end_transaction(trans, root);
2487         if (ret && !err)
2488                 err = ret;
2489         inode->i_flags |= S_DEAD;
2490 out_release:
2491         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2492 out_up_write:
2493         up_write(&root->fs_info->subvol_sem);
2494         if (err) {
2495                 spin_lock(&dest->root_item_lock);
2496                 root_flags = btrfs_root_flags(&dest->root_item);
2497                 btrfs_set_root_flags(&dest->root_item,
2498                                 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2499                 spin_unlock(&dest->root_item_lock);
2500         }
2501 out_unlock_inode:
2502         mutex_unlock(&inode->i_mutex);
2503         if (!err) {
2504                 shrink_dcache_sb(root->fs_info->sb);
2505                 btrfs_invalidate_inodes(dest);
2506                 d_delete(dentry);
2507                 ASSERT(dest->send_in_progress == 0);
2508 
2509                 /* the last ref */
2510                 if (dest->ino_cache_inode) {
2511                         iput(dest->ino_cache_inode);
2512                         dest->ino_cache_inode = NULL;
2513                 }
2514         }
2515 out_dput:
2516         dput(dentry);
2517 out_unlock_dir:
2518         mutex_unlock(&dir->i_mutex);
2519 out_drop_write:
2520         mnt_drop_write_file(file);
2521 out:
2522         kfree(vol_args);
2523         return err;
2524 }
2525 
2526 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2527 {
2528         struct inode *inode = file_inode(file);
2529         struct btrfs_root *root = BTRFS_I(inode)->root;
2530         struct btrfs_ioctl_defrag_range_args *range;
2531         int ret;
2532 
2533         ret = mnt_want_write_file(file);
2534         if (ret)
2535                 return ret;
2536 
2537         if (btrfs_root_readonly(root)) {
2538                 ret = -EROFS;
2539                 goto out;
2540         }
2541 
2542         switch (inode->i_mode & S_IFMT) {
2543         case S_IFDIR:
2544                 if (!capable(CAP_SYS_ADMIN)) {
2545                         ret = -EPERM;
2546                         goto out;
2547                 }
2548                 ret = btrfs_defrag_root(root);
2549                 if (ret)
2550                         goto out;
2551                 ret = btrfs_defrag_root(root->fs_info->extent_root);
2552                 break;
2553         case S_IFREG:
2554                 if (!(file->f_mode & FMODE_WRITE)) {
2555                         ret = -EINVAL;
2556                         goto out;
2557                 }
2558 
2559                 range = kzalloc(sizeof(*range), GFP_KERNEL);
2560                 if (!range) {
2561                         ret = -ENOMEM;
2562                         goto out;
2563                 }
2564 
2565                 if (argp) {
2566                         if (copy_from_user(range, argp,
2567                                            sizeof(*range))) {
2568                                 ret = -EFAULT;
2569                                 kfree(range);
2570                                 goto out;
2571                         }
2572                         /* compression requires us to start the IO */
2573                         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2574                                 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2575                                 range->extent_thresh = (u32)-1;
2576                         }
2577                 } else {
2578                         /* the rest are all set to zero by kzalloc */
2579                         range->len = (u64)-1;
2580                 }
2581                 ret = btrfs_defrag_file(file_inode(file), file,
2582                                         range, 0, 0);
2583                 if (ret > 0)
2584                         ret = 0;
2585                 kfree(range);
2586                 break;
2587         default:
2588                 ret = -EINVAL;
2589         }
2590 out:
2591         mnt_drop_write_file(file);
2592         return ret;
2593 }
2594 
2595 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2596 {
2597         struct btrfs_ioctl_vol_args *vol_args;
2598         int ret;
2599 
2600         if (!capable(CAP_SYS_ADMIN))
2601                 return -EPERM;
2602 
2603         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2604                         1)) {
2605                 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2606         }
2607 
2608         mutex_lock(&root->fs_info->volume_mutex);
2609         vol_args = memdup_user(arg, sizeof(*vol_args));
2610         if (IS_ERR(vol_args)) {
2611                 ret = PTR_ERR(vol_args);
2612                 goto out;
2613         }
2614 
2615         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2616         ret = btrfs_init_new_device(root, vol_args->name);
2617 
2618         if (!ret)
2619                 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2620 
2621         kfree(vol_args);
2622 out:
2623         mutex_unlock(&root->fs_info->volume_mutex);
2624         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2625         return ret;
2626 }
2627 
2628 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2629 {
2630         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2631         struct btrfs_ioctl_vol_args *vol_args;
2632         int ret;
2633 
2634         if (!capable(CAP_SYS_ADMIN))
2635                 return -EPERM;
2636 
2637         ret = mnt_want_write_file(file);
2638         if (ret)
2639                 return ret;
2640 
2641         vol_args = memdup_user(arg, sizeof(*vol_args));
2642         if (IS_ERR(vol_args)) {
2643                 ret = PTR_ERR(vol_args);
2644                 goto err_drop;
2645         }
2646 
2647         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2648 
2649         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2650                         1)) {
2651                 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2652                 goto out;
2653         }
2654 
2655         mutex_lock(&root->fs_info->volume_mutex);
2656         ret = btrfs_rm_device(root, vol_args->name);
2657         mutex_unlock(&root->fs_info->volume_mutex);
2658         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2659 
2660         if (!ret)
2661                 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2662 
2663 out:
2664         kfree(vol_args);
2665 err_drop:
2666         mnt_drop_write_file(file);
2667         return ret;
2668 }
2669 
2670 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2671 {
2672         struct btrfs_ioctl_fs_info_args *fi_args;
2673         struct btrfs_device *device;
2674         struct btrfs_device *next;
2675         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2676         int ret = 0;
2677 
2678         fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2679         if (!fi_args)
2680                 return -ENOMEM;
2681 
2682         mutex_lock(&fs_devices->device_list_mutex);
2683         fi_args->num_devices = fs_devices->num_devices;
2684         memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2685 
2686         list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2687                 if (device->devid > fi_args->max_id)
2688                         fi_args->max_id = device->devid;
2689         }
2690         mutex_unlock(&fs_devices->device_list_mutex);
2691 
2692         fi_args->nodesize = root->fs_info->super_copy->nodesize;
2693         fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2694         fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2695 
2696         if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2697                 ret = -EFAULT;
2698 
2699         kfree(fi_args);
2700         return ret;
2701 }
2702 
2703 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2704 {
2705         struct btrfs_ioctl_dev_info_args *di_args;
2706         struct btrfs_device *dev;
2707         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2708         int ret = 0;
2709         char *s_uuid = NULL;
2710 
2711         di_args = memdup_user(arg, sizeof(*di_args));
2712         if (IS_ERR(di_args))
2713                 return PTR_ERR(di_args);
2714 
2715         if (!btrfs_is_empty_uuid(di_args->uuid))
2716                 s_uuid = di_args->uuid;
2717 
2718         mutex_lock(&fs_devices->device_list_mutex);
2719         dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2720 
2721         if (!dev) {
2722                 ret = -ENODEV;
2723                 goto out;
2724         }
2725 
2726         di_args->devid = dev->devid;
2727         di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2728         di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2729         memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2730         if (dev->name) {
2731                 struct rcu_string *name;
2732 
2733                 rcu_read_lock();
2734                 name = rcu_dereference(dev->name);
2735                 strncpy(di_args->path, name->str, sizeof(di_args->path));
2736                 rcu_read_unlock();
2737                 di_args->path[sizeof(di_args->path) - 1] = 0;
2738         } else {
2739                 di_args->path[0] = '\0';
2740         }
2741 
2742 out:
2743         mutex_unlock(&fs_devices->device_list_mutex);
2744         if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2745                 ret = -EFAULT;
2746 
2747         kfree(di_args);
2748         return ret;
2749 }
2750 
2751 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2752 {
2753         struct page *page;
2754         pgoff_t index;
2755         struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2756 
2757         index = off >> PAGE_CACHE_SHIFT;
2758 
2759         page = grab_cache_page(inode->i_mapping, index);
2760         if (!page)
2761                 return NULL;
2762 
2763         if (!PageUptodate(page)) {
2764                 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2765                                                  0))
2766                         return NULL;
2767                 lock_page(page);
2768                 if (!PageUptodate(page)) {
2769                         unlock_page(page);
2770                         page_cache_release(page);
2771                         return NULL;
2772                 }
2773         }
2774         unlock_page(page);
2775 
2776         return page;
2777 }
2778 
2779 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2780 {
2781         /* do any pending delalloc/csum calc on src, one way or
2782            another, and lock file content */
2783         while (1) {
2784                 struct btrfs_ordered_extent *ordered;
2785                 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2786                 ordered = btrfs_lookup_first_ordered_extent(inode,
2787                                                             off + len - 1);
2788                 if ((!ordered ||
2789                      ordered->file_offset + ordered->len <= off ||
2790                      ordered->file_offset >= off + len) &&
2791                     !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2792                                     off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2793                         if (ordered)
2794                                 btrfs_put_ordered_extent(ordered);
2795                         break;
2796                 }
2797                 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2798                 if (ordered)
2799                         btrfs_put_ordered_extent(ordered);
2800                 btrfs_wait_ordered_range(inode, off, len);
2801         }
2802 }
2803 
2804 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2805                                 struct inode *inode2, u64 loff2, u64 len)
2806 {
2807         unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2808         unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2809 
2810         mutex_unlock(&inode1->i_mutex);
2811         mutex_unlock(&inode2->i_mutex);
2812 }
2813 
2814 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2815                               struct inode *inode2, u64 loff2, u64 len)
2816 {
2817         if (inode1 < inode2) {
2818                 swap(inode1, inode2);
2819                 swap(loff1, loff2);
2820         }
2821 
2822         mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2823         lock_extent_range(inode1, loff1, len);
2824         if (inode1 != inode2) {
2825                 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2826                 lock_extent_range(inode2, loff2, len);
2827         }
2828 }
2829 
2830 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2831                           u64 dst_loff, u64 len)
2832 {
2833         int ret = 0;
2834         struct page *src_page, *dst_page;
2835         unsigned int cmp_len = PAGE_CACHE_SIZE;
2836         void *addr, *dst_addr;
2837 
2838         while (len) {
2839                 if (len < PAGE_CACHE_SIZE)
2840                         cmp_len = len;
2841 
2842                 src_page = extent_same_get_page(src, loff);
2843                 if (!src_page)
2844                         return -EINVAL;
2845                 dst_page = extent_same_get_page(dst, dst_loff);
2846                 if (!dst_page) {
2847                         page_cache_release(src_page);
2848                         return -EINVAL;
2849                 }
2850                 addr = kmap_atomic(src_page);
2851                 dst_addr = kmap_atomic(dst_page);
2852 
2853                 flush_dcache_page(src_page);
2854                 flush_dcache_page(dst_page);
2855 
2856                 if (memcmp(addr, dst_addr, cmp_len))
2857                         ret = BTRFS_SAME_DATA_DIFFERS;
2858 
2859                 kunmap_atomic(addr);
2860                 kunmap_atomic(dst_addr);
2861                 page_cache_release(src_page);
2862                 page_cache_release(dst_page);
2863 
2864                 if (ret)
2865                         break;
2866 
2867                 loff += cmp_len;
2868                 dst_loff += cmp_len;
2869                 len -= cmp_len;
2870         }
2871 
2872         return ret;
2873 }
2874 
2875 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2876 {
2877         u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2878 
2879         if (off + len > inode->i_size || off + len < off)
2880                 return -EINVAL;
2881         /* Check that we are block aligned - btrfs_clone() requires this */
2882         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2883                 return -EINVAL;
2884 
2885         return 0;
2886 }
2887 
2888 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2889                              struct inode *dst, u64 dst_loff)
2890 {
2891         int ret;
2892 
2893         /*
2894          * btrfs_clone() can't handle extents in the same file
2895          * yet. Once that works, we can drop this check and replace it
2896          * with a check for the same inode, but overlapping extents.
2897          */
2898         if (src == dst)
2899                 return -EINVAL;
2900 
2901         if (len == 0)
2902                 return 0;
2903 
2904         btrfs_double_lock(src, loff, dst, dst_loff, len);
2905 
2906         ret = extent_same_check_offsets(src, loff, len);
2907         if (ret)
2908                 goto out_unlock;
2909 
2910         ret = extent_same_check_offsets(dst, dst_loff, len);
2911         if (ret)
2912                 goto out_unlock;
2913 
2914         /* don't make the dst file partly checksummed */
2915         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2916             (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2917                 ret = -EINVAL;
2918                 goto out_unlock;
2919         }
2920 
2921         ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2922         if (ret == 0)
2923                 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2924 
2925 out_unlock:
2926         btrfs_double_unlock(src, loff, dst, dst_loff, len);
2927 
2928         return ret;
2929 }
2930 
2931 #define BTRFS_MAX_DEDUPE_LEN    (16 * 1024 * 1024)
2932 
2933 static long btrfs_ioctl_file_extent_same(struct file *file,
2934                         struct btrfs_ioctl_same_args __user *argp)
2935 {
2936         struct btrfs_ioctl_same_args *same;
2937         struct btrfs_ioctl_same_extent_info *info;
2938         struct inode *src = file_inode(file);
2939         u64 off;
2940         u64 len;
2941         int i;
2942         int ret;
2943         unsigned long size;
2944         u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2945         bool is_admin = capable(CAP_SYS_ADMIN);
2946         u16 count;
2947 
2948         if (!(file->f_mode & FMODE_READ))
2949                 return -EINVAL;
2950 
2951         ret = mnt_want_write_file(file);
2952         if (ret)
2953                 return ret;
2954 
2955         if (get_user(count, &argp->dest_count)) {
2956                 ret = -EFAULT;
2957                 goto out;
2958         }
2959 
2960         size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2961 
2962         same = memdup_user(argp, size);
2963 
2964         if (IS_ERR(same)) {
2965                 ret = PTR_ERR(same);
2966                 goto out;
2967         }
2968 
2969         off = same->logical_offset;
2970         len = same->length;
2971 
2972         /*
2973          * Limit the total length we will dedupe for each operation.
2974          * This is intended to bound the total time spent in this
2975          * ioctl to something sane.
2976          */
2977         if (len > BTRFS_MAX_DEDUPE_LEN)
2978                 len = BTRFS_MAX_DEDUPE_LEN;
2979 
2980         if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2981                 /*
2982                  * Btrfs does not support blocksize < page_size. As a
2983                  * result, btrfs_cmp_data() won't correctly handle
2984                  * this situation without an update.
2985                  */
2986                 ret = -EINVAL;
2987                 goto out;
2988         }
2989 
2990         ret = -EISDIR;
2991         if (S_ISDIR(src->i_mode))
2992                 goto out;
2993 
2994         ret = -EACCES;
2995         if (!S_ISREG(src->i_mode))
2996                 goto out;
2997 
2998         /* pre-format output fields to sane values */
2999         for (i = 0; i < count; i++) {
3000                 same->info[i].bytes_deduped = 0ULL;
3001                 same->info[i].status = 0;
3002         }
3003 
3004         for (i = 0, info = same->info; i < count; i++, info++) {
3005                 struct inode *dst;
3006                 struct fd dst_file = fdget(info->fd);
3007                 if (!dst_file.file) {
3008                         info->status = -EBADF;
3009                         continue;
3010                 }
3011                 dst = file_inode(dst_file.file);
3012 
3013                 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3014                         info->status = -EINVAL;
3015                 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3016                         info->status = -EXDEV;
3017                 } else if (S_ISDIR(dst->i_mode)) {
3018                         info->status = -EISDIR;
3019                 } else if (!S_ISREG(dst->i_mode)) {
3020                         info->status = -EACCES;
3021                 } else {
3022                         info->status = btrfs_extent_same(src, off, len, dst,
3023                                                         info->logical_offset);
3024                         if (info->status == 0)
3025                                 info->bytes_deduped += len;
3026                 }
3027                 fdput(dst_file);
3028         }
3029 
3030         ret = copy_to_user(argp, same, size);
3031         if (ret)
3032                 ret = -EFAULT;
3033 
3034 out:
3035         mnt_drop_write_file(file);
3036         return ret;
3037 }
3038 
3039 /* Helper to check and see if this root currently has a ref on the given disk
3040  * bytenr.  If it does then we need to update the quota for this root.  This
3041  * doesn't do anything if quotas aren't enabled.
3042  */
3043 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3044                      u64 disko)
3045 {
3046         struct seq_list tree_mod_seq_elem = {};
3047         struct ulist *roots;
3048         struct ulist_iterator uiter;
3049         struct ulist_node *root_node = NULL;
3050         int ret;
3051 
3052         if (!root->fs_info->quota_enabled)
3053                 return 1;
3054 
3055         btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3056         ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3057                                    tree_mod_seq_elem.seq, &roots);
3058         if (ret < 0)
3059                 goto out;
3060         ret = 0;
3061         ULIST_ITER_INIT(&uiter);
3062         while ((root_node = ulist_next(roots, &uiter))) {
3063                 if (root_node->val == root->objectid) {
3064                         ret = 1;
3065                         break;
3066                 }
3067         }
3068         ulist_free(roots);
3069 out:
3070         btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3071         return ret;
3072 }
3073 
3074 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3075                                      struct inode *inode,
3076                                      u64 endoff,
3077                                      const u64 destoff,
3078                                      const u64 olen)
3079 {
3080         struct btrfs_root *root = BTRFS_I(inode)->root;
3081         int ret;
3082 
3083         inode_inc_iversion(inode);
3084         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3085         /*
3086          * We round up to the block size at eof when determining which
3087          * extents to clone above, but shouldn't round up the file size.
3088          */
3089         if (endoff > destoff + olen)
3090                 endoff = destoff + olen;
3091         if (endoff > inode->i_size)
3092                 btrfs_i_size_write(inode, endoff);
3093 
3094         ret = btrfs_update_inode(trans, root, inode);
3095         if (ret) {
3096                 btrfs_abort_transaction(trans, root, ret);
3097                 btrfs_end_transaction(trans, root);
3098                 goto out;
3099         }
3100         ret = btrfs_end_transaction(trans, root);
3101 out:
3102         return ret;
3103 }
3104 
3105 static void clone_update_extent_map(struct inode *inode,
3106                                     const struct btrfs_trans_handle *trans,
3107                                     const struct btrfs_path *path,
3108                                     const u64 hole_offset,
3109                                     const u64 hole_len)
3110 {
3111         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3112         struct extent_map *em;
3113         int ret;
3114 
3115         em = alloc_extent_map();
3116         if (!em) {
3117                 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3118                         &BTRFS_I(inode)->runtime_flags);
3119                 return;
3120         }
3121 
3122         if (path) {
3123                 struct btrfs_file_extent_item *fi;
3124 
3125                 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3126                                     struct btrfs_file_extent_item);
3127                 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3128                 em->generation = -1;
3129                 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3130                     BTRFS_FILE_EXTENT_INLINE)
3131                         set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3132                                 &BTRFS_I(inode)->runtime_flags);
3133         } else {
3134                 em->start = hole_offset;
3135                 em->len = hole_len;
3136                 em->ram_bytes = em->len;
3137                 em->orig_start = hole_offset;
3138                 em->block_start = EXTENT_MAP_HOLE;
3139                 em->block_len = 0;
3140                 em->orig_block_len = 0;
3141                 em->compress_type = BTRFS_COMPRESS_NONE;
3142                 em->generation = trans->transid;
3143         }
3144 
3145         while (1) {
3146                 write_lock(&em_tree->lock);
3147                 ret = add_extent_mapping(em_tree, em, 1);
3148                 write_unlock(&em_tree->lock);
3149                 if (ret != -EEXIST) {
3150                         free_extent_map(em);
3151                         break;
3152                 }
3153                 btrfs_drop_extent_cache(inode, em->start,
3154                                         em->start + em->len - 1, 0);
3155         }
3156 
3157         if (ret)
3158                 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3159                         &BTRFS_I(inode)->runtime_flags);
3160 }
3161 
3162 /**
3163  * btrfs_clone() - clone a range from inode file to another
3164  *
3165  * @src: Inode to clone from
3166  * @inode: Inode to clone to
3167  * @off: Offset within source to start clone from
3168  * @olen: Original length, passed by user, of range to clone
3169  * @olen_aligned: Block-aligned value of olen, extent_same uses
3170  *               identical values here
3171  * @destoff: Offset within @inode to start clone
3172  */
3173 static int btrfs_clone(struct inode *src, struct inode *inode,
3174                        const u64 off, const u64 olen, const u64 olen_aligned,
3175                        const u64 destoff)
3176 {
3177         struct btrfs_root *root = BTRFS_I(inode)->root;
3178         struct btrfs_path *path = NULL;
3179         struct extent_buffer *leaf;
3180         struct btrfs_trans_handle *trans;
3181         char *buf = NULL;
3182         struct btrfs_key key;
3183         u32 nritems;
3184         int slot;
3185         int ret;
3186         int no_quota;
3187         const u64 len = olen_aligned;
3188         u64 last_disko = 0;
3189         u64 last_dest_end = destoff;
3190 
3191         ret = -ENOMEM;
3192         buf = vmalloc(root->nodesize);
3193         if (!buf)
3194                 return ret;
3195 
3196         path = btrfs_alloc_path();
3197         if (!path) {
3198                 vfree(buf);
3199                 return ret;
3200         }
3201 
3202         path->reada = 2;
3203         /* clone data */
3204         key.objectid = btrfs_ino(src);
3205         key.type = BTRFS_EXTENT_DATA_KEY;
3206         key.offset = off;
3207 
3208         while (1) {
3209                 u64 next_key_min_offset = key.offset + 1;
3210 
3211                 /*
3212                  * note the key will change type as we walk through the
3213                  * tree.
3214                  */
3215                 path->leave_spinning = 1;
3216                 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3217                                 0, 0);
3218                 if (ret < 0)
3219                         goto out;
3220                 /*
3221                  * First search, if no extent item that starts at offset off was
3222                  * found but the previous item is an extent item, it's possible
3223                  * it might overlap our target range, therefore process it.
3224                  */
3225                 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3226                         btrfs_item_key_to_cpu(path->nodes[0], &key,
3227                                               path->slots[0] - 1);
3228                         if (key.type == BTRFS_EXTENT_DATA_KEY)
3229                                 path->slots[0]--;
3230                 }
3231 
3232                 nritems = btrfs_header_nritems(path->nodes[0]);
3233 process_slot:
3234                 no_quota = 1;
3235                 if (path->slots[0] >= nritems) {
3236                         ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3237                         if (ret < 0)
3238                                 goto out;
3239                         if (ret > 0)
3240                                 break;
3241                         nritems = btrfs_header_nritems(path->nodes[0]);
3242                 }
3243                 leaf = path->nodes[0];
3244                 slot = path->slots[0];
3245 
3246                 btrfs_item_key_to_cpu(leaf, &key, slot);
3247                 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3248                     key.objectid != btrfs_ino(src))
3249                         break;
3250 
3251                 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3252                         struct btrfs_file_extent_item *extent;
3253                         int type;
3254                         u32 size;
3255                         struct btrfs_key new_key;
3256                         u64 disko = 0, diskl = 0;
3257                         u64 datao = 0, datal = 0;
3258                         u8 comp;
3259                         u64 drop_start;
3260 
3261                         extent = btrfs_item_ptr(leaf, slot,
3262                                                 struct btrfs_file_extent_item);
3263                         comp = btrfs_file_extent_compression(leaf, extent);
3264                         type = btrfs_file_extent_type(leaf, extent);
3265                         if (type == BTRFS_FILE_EXTENT_REG ||
3266                             type == BTRFS_FILE_EXTENT_PREALLOC) {
3267                                 disko = btrfs_file_extent_disk_bytenr(leaf,
3268                                                                       extent);
3269                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3270                                                                  extent);
3271                                 datao = btrfs_file_extent_offset(leaf, extent);
3272                                 datal = btrfs_file_extent_num_bytes(leaf,
3273                                                                     extent);
3274                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3275                                 /* take upper bound, may be compressed */
3276                                 datal = btrfs_file_extent_ram_bytes(leaf,
3277                                                                     extent);
3278                         }
3279 
3280                         /*
3281                          * The first search might have left us at an extent
3282                          * item that ends before our target range's start, can
3283                          * happen if we have holes and NO_HOLES feature enabled.
3284                          */
3285                         if (key.offset + datal <= off) {
3286                                 path->slots[0]++;
3287                                 goto process_slot;
3288                         } else if (key.offset >= off + len) {
3289                                 break;
3290                         }
3291                         next_key_min_offset = key.offset + datal;
3292                         size = btrfs_item_size_nr(leaf, slot);
3293                         read_extent_buffer(leaf, buf,
3294                                            btrfs_item_ptr_offset(leaf, slot),
3295                                            size);
3296 
3297                         btrfs_release_path(path);
3298                         path->leave_spinning = 0;
3299 
3300                         memcpy(&new_key, &key, sizeof(new_key));
3301                         new_key.objectid = btrfs_ino(inode);
3302                         if (off <= key.offset)
3303                                 new_key.offset = key.offset + destoff - off;
3304                         else
3305                                 new_key.offset = destoff;
3306 
3307                         /*
3308                          * Deal with a hole that doesn't have an extent item
3309                          * that represents it (NO_HOLES feature enabled).
3310                          * This hole is either in the middle of the cloning
3311                          * range or at the beginning (fully overlaps it or
3312                          * partially overlaps it).
3313                          */
3314                         if (new_key.offset != last_dest_end)
3315                                 drop_start = last_dest_end;
3316                         else
3317                                 drop_start = new_key.offset;
3318 
3319                         /*
3320                          * 1 - adjusting old extent (we may have to split it)
3321                          * 1 - add new extent
3322                          * 1 - inode update
3323                          */
3324                         trans = btrfs_start_transaction(root, 3);
3325                         if (IS_ERR(trans)) {
3326                                 ret = PTR_ERR(trans);
3327                                 goto out;
3328                         }
3329 
3330                         if (type == BTRFS_FILE_EXTENT_REG ||
3331                             type == BTRFS_FILE_EXTENT_PREALLOC) {
3332                                 /*
3333                                  *    a  | --- range to clone ---|  b
3334                                  * | ------------- extent ------------- |
3335                                  */
3336 
3337                                 /* subtract range b */
3338                                 if (key.offset + datal > off + len)
3339                                         datal = off + len - key.offset;
3340 
3341                                 /* subtract range a */
3342                                 if (off > key.offset) {
3343                                         datao += off - key.offset;
3344                                         datal -= off - key.offset;
3345                                 }
3346 
3347                                 ret = btrfs_drop_extents(trans, root, inode,
3348                                                          drop_start,
3349                                                          new_key.offset + datal,
3350                                                          1);
3351                                 if (ret) {
3352                                         if (ret != -EOPNOTSUPP)
3353                                                 btrfs_abort_transaction(trans,
3354                                                                 root, ret);
3355                                         btrfs_end_transaction(trans, root);
3356                                         goto out;
3357                                 }
3358 
3359                                 ret = btrfs_insert_empty_item(trans, root, path,
3360                                                               &new_key, size);
3361                                 if (ret) {
3362                                         btrfs_abort_transaction(trans, root,
3363                                                                 ret);
3364                                         btrfs_end_transaction(trans, root);
3365                                         goto out;
3366                                 }
3367 
3368                                 leaf = path->nodes[0];
3369                                 slot = path->slots[0];
3370                                 write_extent_buffer(leaf, buf,
3371                                             btrfs_item_ptr_offset(leaf, slot),
3372                                             size);
3373 
3374                                 extent = btrfs_item_ptr(leaf, slot,
3375                                                 struct btrfs_file_extent_item);
3376 
3377                                 /* disko == 0 means it's a hole */
3378                                 if (!disko)
3379                                         datao = 0;
3380 
3381                                 btrfs_set_file_extent_offset(leaf, extent,
3382                                                              datao);
3383                                 btrfs_set_file_extent_num_bytes(leaf, extent,
3384                                                                 datal);
3385 
3386                                 /*
3387                                  * We need to look up the roots that point at
3388                                  * this bytenr and see if the new root does.  If
3389                                  * it does not we need to make sure we update
3390                                  * quotas appropriately.
3391                                  */
3392                                 if (disko && root != BTRFS_I(src)->root &&
3393                                     disko != last_disko) {
3394                                         no_quota = check_ref(trans, root,
3395                                                              disko);
3396                                         if (no_quota < 0) {
3397                                                 btrfs_abort_transaction(trans,
3398                                                                         root,
3399                                                                         ret);
3400                                                 btrfs_end_transaction(trans,
3401                                                                       root);
3402                                                 ret = no_quota;
3403                                                 goto out;
3404                                         }
3405                                 }
3406 
3407                                 if (disko) {
3408                                         inode_add_bytes(inode, datal);
3409                                         ret = btrfs_inc_extent_ref(trans, root,
3410                                                         disko, diskl, 0,
3411                                                         root->root_key.objectid,
3412                                                         btrfs_ino(inode),
3413                                                         new_key.offset - datao,
3414                                                         no_quota);
3415                                         if (ret) {
3416                                                 btrfs_abort_transaction(trans,
3417                                                                         root,
3418                                                                         ret);
3419                                                 btrfs_end_transaction(trans,
3420                                                                       root);
3421                                                 goto out;
3422 
3423                                         }
3424                                 }
3425                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3426                                 u64 skip = 0;
3427                                 u64 trim = 0;
3428                                 u64 aligned_end = 0;
3429 
3430                                 if (off > key.offset) {
3431                                         skip = off - key.offset;
3432                                         new_key.offset += skip;
3433                                 }
3434 
3435                                 if (key.offset + datal > off + len)
3436                                         trim = key.offset + datal - (off + len);
3437 
3438                                 if (comp && (skip || trim)) {
3439                                         ret = -EINVAL;
3440                                         btrfs_end_transaction(trans, root);
3441                                         goto out;
3442                                 }
3443                                 size -= skip + trim;
3444                                 datal -= skip + trim;
3445 
3446                                 aligned_end = ALIGN(new_key.offset + datal,
3447                                                     root->sectorsize);
3448                                 ret = btrfs_drop_extents(trans, root, inode,
3449                                                          drop_start,
3450                                                          aligned_end,
3451                                                          1);
3452                                 if (ret) {
3453                                         if (ret != -EOPNOTSUPP)
3454                                                 btrfs_abort_transaction(trans,
3455                                                         root, ret);
3456                                         btrfs_end_transaction(trans, root);
3457                                         goto out;
3458                                 }
3459 
3460                                 ret = btrfs_insert_empty_item(trans, root, path,
3461                                                               &new_key, size);
3462                                 if (ret) {
3463                                         btrfs_abort_transaction(trans, root,
3464                                                                 ret);
3465                                         btrfs_end_transaction(trans, root);
3466                                         goto out;
3467                                 }
3468 
3469                                 if (skip) {
3470                                         u32 start =
3471                                           btrfs_file_extent_calc_inline_size(0);
3472                                         memmove(buf+start, buf+start+skip,
3473                                                 datal);
3474                                 }
3475 
3476                                 leaf = path->nodes[0];
3477                                 slot = path->slots[0];
3478                                 write_extent_buffer(leaf, buf,
3479                                             btrfs_item_ptr_offset(leaf, slot),
3480                                             size);
3481                                 inode_add_bytes(inode, datal);
3482                         }
3483 
3484                         /* If we have an implicit hole (NO_HOLES feature). */
3485                         if (drop_start < new_key.offset)
3486                                 clone_update_extent_map(inode, trans,
3487                                                 NULL, drop_start,
3488                                                 new_key.offset - drop_start);
3489 
3490                         clone_update_extent_map(inode, trans, path, 0, 0);
3491 
3492                         btrfs_mark_buffer_dirty(leaf);
3493                         btrfs_release_path(path);
3494 
3495                         last_dest_end = ALIGN(new_key.offset + datal,
3496                                               root->sectorsize);
3497                         ret = clone_finish_inode_update(trans, inode,
3498                                                         last_dest_end,
3499                                                         destoff, olen);
3500                         if (ret)
3501                                 goto out;
3502                         if (new_key.offset + datal >= destoff + len)
3503                                 break;
3504                 }
3505                 btrfs_release_path(path);
3506                 key.offset = next_key_min_offset;
3507         }
3508         ret = 0;
3509 
3510         if (last_dest_end < destoff + len) {
3511                 /*
3512                  * We have an implicit hole (NO_HOLES feature is enabled) that
3513                  * fully or partially overlaps our cloning range at its end.
3514                  */
3515                 btrfs_release_path(path);
3516 
3517                 /*
3518                  * 1 - remove extent(s)
3519                  * 1 - inode update
3520                  */
3521                 trans = btrfs_start_transaction(root, 2);
3522                 if (IS_ERR(trans)) {
3523                         ret = PTR_ERR(trans);
3524                         goto out;
3525                 }
3526                 ret = btrfs_drop_extents(trans, root, inode,
3527                                          last_dest_end, destoff + len, 1);
3528                 if (ret) {
3529                         if (ret != -EOPNOTSUPP)
3530                                 btrfs_abort_transaction(trans, root, ret);
3531                         btrfs_end_transaction(trans, root);
3532                         goto out;
3533                 }
3534                 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3535                                         destoff + len - last_dest_end);
3536                 ret = clone_finish_inode_update(trans, inode, destoff + len,
3537                                                 destoff, olen);
3538         }
3539 
3540 out:
3541         btrfs_free_path(path);
3542         vfree(buf);
3543         return ret;
3544 }
3545 
3546 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3547                                        u64 off, u64 olen, u64 destoff)
3548 {
3549         struct inode *inode = file_inode(file);
3550         struct btrfs_root *root = BTRFS_I(inode)->root;
3551         struct fd src_file;
3552         struct inode *src;
3553         int ret;
3554         u64 len = olen;
3555         u64 bs = root->fs_info->sb->s_blocksize;
3556         int same_inode = 0;
3557 
3558         /*
3559          * TODO:
3560          * - split compressed inline extents.  annoying: we need to
3561          *   decompress into destination's address_space (the file offset
3562          *   may change, so source mapping won't do), then recompress (or
3563          *   otherwise reinsert) a subrange.
3564          *
3565          * - split destination inode's inline extents.  The inline extents can
3566          *   be either compressed or non-compressed.
3567          */
3568 
3569         /* the destination must be opened for writing */
3570         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3571                 return -EINVAL;
3572 
3573         if (btrfs_root_readonly(root))
3574                 return -EROFS;
3575 
3576         ret = mnt_want_write_file(file);
3577         if (ret)
3578                 return ret;
3579 
3580         src_file = fdget(srcfd);
3581         if (!src_file.file) {
3582                 ret = -EBADF;
3583                 goto out_drop_write;
3584         }
3585 
3586         ret = -EXDEV;
3587         if (src_file.file->f_path.mnt != file->f_path.mnt)
3588                 goto out_fput;
3589 
3590         src = file_inode(src_file.file);
3591 
3592         ret = -EINVAL;
3593         if (src == inode)
3594                 same_inode = 1;
3595 
3596         /* the src must be open for reading */
3597         if (!(src_file.file->f_mode & FMODE_READ))
3598                 goto out_fput;
3599 
3600         /* don't make the dst file partly checksummed */
3601         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3602             (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3603                 goto out_fput;
3604 
3605         ret = -EISDIR;
3606         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3607                 goto out_fput;
3608 
3609         ret = -EXDEV;
3610         if (src->i_sb != inode->i_sb)
3611                 goto out_fput;
3612 
3613         if (!same_inode) {
3614                 if (inode < src) {
3615                         mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3616                         mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3617                 } else {
3618                         mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3619                         mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3620                 }
3621         } else {
3622                 mutex_lock(&src->i_mutex);
3623         }
3624 
3625         /* determine range to clone */
3626         ret = -EINVAL;
3627         if (off + len > src->i_size || off + len < off)
3628                 goto out_unlock;
3629         if (len == 0)
3630                 olen = len = src->i_size - off;
3631         /* if we extend to eof, continue to block boundary */
3632         if (off + len == src->i_size)
3633                 len = ALIGN(src->i_size, bs) - off;
3634 
3635         if (len == 0) {
3636                 ret = 0;
3637                 goto out_unlock;
3638         }
3639 
3640         /* verify the end result is block aligned */
3641         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3642             !IS_ALIGNED(destoff, bs))
3643                 goto out_unlock;
3644 
3645         /* verify if ranges are overlapped within the same file */
3646         if (same_inode) {
3647                 if (destoff + len > off && destoff < off + len)
3648                         goto out_unlock;
3649         }
3650 
3651         if (destoff > inode->i_size) {
3652                 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3653                 if (ret)
3654                         goto out_unlock;
3655         }
3656 
3657         /*
3658          * Lock the target range too. Right after we replace the file extent
3659          * items in the fs tree (which now point to the cloned data), we might
3660          * have a worker replace them with extent items relative to a write
3661          * operation that was issued before this clone operation (i.e. confront
3662          * with inode.c:btrfs_finish_ordered_io).
3663          */
3664         if (same_inode) {
3665                 u64 lock_start = min_t(u64, off, destoff);
3666                 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3667 
3668                 lock_extent_range(src, lock_start, lock_len);
3669         } else {
3670                 lock_extent_range(src, off, len);
3671                 lock_extent_range(inode, destoff, len);
3672         }
3673 
3674         ret = btrfs_clone(src, inode, off, olen, len, destoff);
3675 
3676         if (same_inode) {
3677                 u64 lock_start = min_t(u64, off, destoff);
3678                 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3679 
3680                 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3681         } else {
3682                 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3683                 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3684                               destoff + len - 1);
3685         }
3686         /*
3687          * Truncate page cache pages so that future reads will see the cloned
3688          * data immediately and not the previous data.
3689          */
3690         truncate_inode_pages_range(&inode->i_data, destoff,
3691                                    PAGE_CACHE_ALIGN(destoff + len) - 1);
3692 out_unlock:
3693         if (!same_inode) {
3694                 if (inode < src) {
3695                         mutex_unlock(&src->i_mutex);
3696                         mutex_unlock(&inode->i_mutex);
3697                 } else {
3698                         mutex_unlock(&inode->i_mutex);
3699                         mutex_unlock(&src->i_mutex);
3700                 }
3701         } else {
3702                 mutex_unlock(&src->i_mutex);
3703         }
3704 out_fput:
3705         fdput(src_file);
3706 out_drop_write:
3707         mnt_drop_write_file(file);
3708         return ret;
3709 }
3710 
3711 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3712 {
3713         struct btrfs_ioctl_clone_range_args args;
3714 
3715         if (copy_from_user(&args, argp, sizeof(args)))
3716                 return -EFAULT;
3717         return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3718                                  args.src_length, args.dest_offset);
3719 }
3720 
3721 /*
3722  * there are many ways the trans_start and trans_end ioctls can lead
3723  * to deadlocks.  They should only be used by applications that
3724  * basically own the machine, and have a very in depth understanding
3725  * of all the possible deadlocks and enospc problems.
3726  */
3727 static long btrfs_ioctl_trans_start(struct file *file)
3728 {
3729         struct inode *inode = file_inode(file);
3730         struct btrfs_root *root = BTRFS_I(inode)->root;
3731         struct btrfs_trans_handle *trans;
3732         int ret;
3733 
3734         ret = -EPERM;
3735         if (!capable(CAP_SYS_ADMIN))
3736                 goto out;
3737 
3738         ret = -EINPROGRESS;
3739         if (file->private_data)
3740                 goto out;
3741 
3742         ret = -EROFS;
3743         if (btrfs_root_readonly(root))
3744                 goto out;
3745 
3746         ret = mnt_want_write_file(file);
3747         if (ret)
3748                 goto out;
3749 
3750         atomic_inc(&root->fs_info->open_ioctl_trans);
3751 
3752         ret = -ENOMEM;
3753         trans = btrfs_start_ioctl_transaction(root);
3754         if (IS_ERR(trans))
3755                 goto out_drop;
3756 
3757         file->private_data = trans;
3758         return 0;
3759 
3760 out_drop:
3761         atomic_dec(&root->fs_info->open_ioctl_trans);
3762         mnt_drop_write_file(file);
3763 out:
3764         return ret;
3765 }
3766 
3767 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3768 {
3769         struct inode *inode = file_inode(file);
3770         struct btrfs_root *root = BTRFS_I(inode)->root;
3771         struct btrfs_root *new_root;
3772         struct btrfs_dir_item *di;
3773         struct btrfs_trans_handle *trans;
3774         struct btrfs_path *path;
3775         struct btrfs_key location;
3776         struct btrfs_disk_key disk_key;
3777         u64 objectid = 0;
3778         u64 dir_id;
3779         int ret;
3780 
3781         if (!capable(CAP_SYS_ADMIN))
3782                 return -EPERM;
3783 
3784         ret = mnt_want_write_file(file);
3785         if (ret)
3786                 return ret;
3787 
3788         if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3789                 ret = -EFAULT;
3790                 goto out;
3791         }
3792 
3793         if (!objectid)
3794                 objectid = BTRFS_FS_TREE_OBJECTID;
3795 
3796         location.objectid = objectid;
3797         location.type = BTRFS_ROOT_ITEM_KEY;
3798         location.offset = (u64)-1;
3799 
3800         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3801         if (IS_ERR(new_root)) {
3802                 ret = PTR_ERR(new_root);
3803                 goto out;
3804         }
3805 
3806         path = btrfs_alloc_path();
3807         if (!path) {
3808                 ret = -ENOMEM;
3809                 goto out;
3810         }
3811         path->leave_spinning = 1;
3812 
3813         trans = btrfs_start_transaction(root, 1);
3814         if (IS_ERR(trans)) {
3815                 btrfs_free_path(path);
3816                 ret = PTR_ERR(trans);
3817                 goto out;
3818         }
3819 
3820         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3821         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3822                                    dir_id, "default", 7, 1);
3823         if (IS_ERR_OR_NULL(di)) {
3824                 btrfs_free_path(path);
3825                 btrfs_end_transaction(trans, root);
3826                 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3827                            "item, this isn't going to work");
3828                 ret = -ENOENT;
3829                 goto out;
3830         }
3831 
3832         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3833         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3834         btrfs_mark_buffer_dirty(path->nodes[0]);
3835         btrfs_free_path(path);
3836 
3837         btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3838         btrfs_end_transaction(trans, root);
3839 out:
3840         mnt_drop_write_file(file);
3841         return ret;
3842 }
3843 
3844 void btrfs_get_block_group_info(struct list_head *groups_list,
3845                                 struct btrfs_ioctl_space_info *space)
3846 {
3847         struct btrfs_block_group_cache *block_group;
3848 
3849         space->total_bytes = 0;
3850         space->used_bytes = 0;
3851         space->flags = 0;
3852         list_for_each_entry(block_group, groups_list, list) {
3853                 space->flags = block_group->flags;
3854                 space->total_bytes += block_group->key.offset;
3855                 space->used_bytes +=
3856                         btrfs_block_group_used(&block_group->item);
3857         }
3858 }
3859 
3860 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3861 {
3862         struct btrfs_ioctl_space_args space_args;
3863         struct btrfs_ioctl_space_info space;
3864         struct btrfs_ioctl_space_info *dest;
3865         struct btrfs_ioctl_space_info *dest_orig;
3866         struct btrfs_ioctl_space_info __user *user_dest;
3867         struct btrfs_space_info *info;
3868         u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3869                        BTRFS_BLOCK_GROUP_SYSTEM,
3870                        BTRFS_BLOCK_GROUP_METADATA,
3871                        BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3872         int num_types = 4;
3873         int alloc_size;
3874         int ret = 0;
3875         u64 slot_count = 0;
3876         int i, c;
3877 
3878         if (copy_from_user(&space_args,
3879                            (struct btrfs_ioctl_space_args __user *)arg,
3880                            sizeof(space_args)))
3881                 return -EFAULT;
3882 
3883         for (i = 0; i < num_types; i++) {
3884                 struct btrfs_space_info *tmp;
3885 
3886                 info = NULL;
3887                 rcu_read_lock();
3888                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3889                                         list) {
3890                         if (tmp->flags == types[i]) {
3891                                 info = tmp;
3892                                 break;
3893                         }
3894                 }
3895                 rcu_read_unlock();
3896 
3897                 if (!info)
3898                         continue;
3899 
3900                 down_read(&info->groups_sem);
3901                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3902                         if (!list_empty(&info->block_groups[c]))
3903                                 slot_count++;
3904                 }
3905                 up_read(&info->groups_sem);
3906         }
3907 
3908         /*
3909          * Global block reserve, exported as a space_info
3910          */
3911         slot_count++;
3912 
3913         /* space_slots == 0 means they are asking for a count */
3914         if (space_args.space_slots == 0) {
3915                 space_args.total_spaces = slot_count;
3916                 goto out;
3917         }
3918 
3919         slot_count = min_t(u64, space_args.space_slots, slot_count);
3920 
3921         alloc_size = sizeof(*dest) * slot_count;
3922 
3923         /* we generally have at most 6 or so space infos, one for each raid
3924          * level.  So, a whole page should be more than enough for everyone
3925          */
3926         if (alloc_size > PAGE_CACHE_SIZE)
3927                 return -ENOMEM;
3928 
3929         space_args.total_spaces = 0;
3930         dest = kmalloc(alloc_size, GFP_NOFS);
3931         if (!dest)
3932                 return -ENOMEM;
3933         dest_orig = dest;
3934 
3935         /* now we have a buffer to copy into */
3936         for (i = 0; i < num_types; i++) {
3937                 struct btrfs_space_info *tmp;
3938 
3939                 if (!slot_count)
3940                         break;
3941 
3942                 info = NULL;
3943                 rcu_read_lock();
3944                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3945                                         list) {
3946                         if (tmp->flags == types[i]) {
3947                                 info = tmp;
3948                                 break;
3949                         }
3950                 }
3951                 rcu_read_unlock();
3952 
3953                 if (!info)
3954                         continue;
3955                 down_read(&info->groups_sem);
3956                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3957                         if (!list_empty(&info->block_groups[c])) {
3958                                 btrfs_get_block_group_info(
3959                                         &info->block_groups[c], &space);
3960                                 memcpy(dest, &space, sizeof(space));
3961                                 dest++;
3962                                 space_args.total_spaces++;
3963                                 slot_count--;
3964                         }
3965                         if (!slot_count)
3966                                 break;
3967                 }
3968                 up_read(&info->groups_sem);
3969         }
3970 
3971         /*
3972          * Add global block reserve
3973          */
3974         if (slot_count) {
3975                 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3976 
3977                 spin_lock(&block_rsv->lock);
3978                 space.total_bytes = block_rsv->size;
3979                 space.used_bytes = block_rsv->size - block_rsv->reserved;
3980                 spin_unlock(&block_rsv->lock);
3981                 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3982                 memcpy(dest, &space, sizeof(space));
3983                 space_args.total_spaces++;
3984         }
3985 
3986         user_dest = (struct btrfs_ioctl_space_info __user *)
3987                 (arg + sizeof(struct btrfs_ioctl_space_args));
3988 
3989         if (copy_to_user(user_dest, dest_orig, alloc_size))
3990                 ret = -EFAULT;
3991 
3992         kfree(dest_orig);
3993 out:
3994         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3995                 ret = -EFAULT;
3996 
3997         return ret;
3998 }
3999 
4000 /*
4001  * there are many ways the trans_start and trans_end ioctls can lead
4002  * to deadlocks.  They should only be used by applications that
4003  * basically own the machine, and have a very in depth understanding
4004  * of all the possible deadlocks and enospc problems.
4005  */
4006 long btrfs_ioctl_trans_end(struct file *file)
4007 {
4008         struct inode *inode = file_inode(file);
4009         struct btrfs_root *root = BTRFS_I(inode)->root;
4010         struct btrfs_trans_handle *trans;
4011 
4012         trans = file->private_data;
4013         if (!trans)
4014                 return -EINVAL;
4015         file->private_data = NULL;
4016 
4017         btrfs_end_transaction(trans, root);
4018 
4019         atomic_dec(&root->fs_info->open_ioctl_trans);
4020 
4021         mnt_drop_write_file(file);
4022         return 0;
4023 }
4024 
4025 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4026                                             void __user *argp)
4027 {
4028         struct btrfs_trans_handle *trans;
4029         u64 transid;
4030         int ret;
4031 
4032         trans = btrfs_attach_transaction_barrier(root);
4033         if (IS_ERR(trans)) {
4034                 if (PTR_ERR(trans) != -ENOENT)
4035                         return PTR_ERR(trans);
4036 
4037                 /* No running transaction, don't bother */
4038                 transid = root->fs_info->last_trans_committed;
4039                 goto out;
4040         }
4041         transid = trans->transid;
4042         ret = btrfs_commit_transaction_async(trans, root, 0);
4043         if (ret) {
4044                 btrfs_end_transaction(trans, root);
4045                 return ret;
4046         }
4047 out:
4048         if (argp)
4049                 if (copy_to_user(argp, &transid, sizeof(transid)))
4050                         return -EFAULT;
4051         return 0;
4052 }
4053 
4054 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4055                                            void __user *argp)
4056 {
4057         u64 transid;
4058 
4059         if (argp) {
4060                 if (copy_from_user(&transid, argp, sizeof(transid)))
4061                         return -EFAULT;
4062         } else {
4063                 transid = 0;  /* current trans */
4064         }
4065         return btrfs_wait_for_commit(root, transid);
4066 }
4067 
4068 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4069 {
4070         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4071         struct btrfs_ioctl_scrub_args *sa;
4072         int ret;
4073 
4074         if (!capable(CAP_SYS_ADMIN))
4075                 return -EPERM;
4076 
4077         sa = memdup_user(arg, sizeof(*sa));
4078         if (IS_ERR(sa))
4079                 return PTR_ERR(sa);
4080 
4081         if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4082                 ret = mnt_want_write_file(file);
4083                 if (ret)
4084                         goto out;
4085         }
4086 
4087         ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4088                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4089                               0);
4090 
4091         if (copy_to_user(arg, sa, sizeof(*sa)))
4092                 ret = -EFAULT;
4093 
4094         if (!(sa->flags & BTRFS_SCRUB_READONLY))
4095                 mnt_drop_write_file(file);
4096 out:
4097         kfree(sa);
4098         return ret;
4099 }
4100 
4101 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4102 {
4103         if (!capable(CAP_SYS_ADMIN))
4104                 return -EPERM;
4105 
4106         return btrfs_scrub_cancel(root->fs_info);
4107 }
4108 
4109 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4110                                        void __user *arg)
4111 {
4112         struct btrfs_ioctl_scrub_args *sa;
4113         int ret;
4114 
4115         if (!capable(CAP_SYS_ADMIN))
4116                 return -EPERM;
4117 
4118         sa = memdup_user(arg, sizeof(*sa));
4119         if (IS_ERR(sa))
4120                 return PTR_ERR(sa);
4121 
4122         ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4123 
4124         if (copy_to_user(arg, sa, sizeof(*sa)))
4125                 ret = -EFAULT;
4126 
4127         kfree(sa);
4128         return ret;
4129 }
4130 
4131 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4132                                       void __user *arg)
4133 {
4134         struct btrfs_ioctl_get_dev_stats *sa;
4135         int ret;
4136 
4137         sa = memdup_user(arg, sizeof(*sa));
4138         if (IS_ERR(sa))
4139                 return PTR_ERR(sa);
4140 
4141         if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4142                 kfree(sa);
4143                 return -EPERM;
4144         }
4145 
4146         ret = btrfs_get_dev_stats(root, sa);
4147 
4148         if (copy_to_user(arg, sa, sizeof(*sa)))
4149                 ret = -EFAULT;
4150 
4151         kfree(sa);
4152         return ret;
4153 }
4154 
4155 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4156 {
4157         struct btrfs_ioctl_dev_replace_args *p;
4158         int ret;
4159 
4160         if (!capable(CAP_SYS_ADMIN))
4161                 return -EPERM;
4162 
4163         p = memdup_user(arg, sizeof(*p));
4164         if (IS_ERR(p))
4165                 return PTR_ERR(p);
4166 
4167         switch (p->cmd) {
4168         case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4169                 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4170                         ret = -EROFS;
4171                         goto out;
4172                 }
4173                 if (atomic_xchg(
4174                         &root->fs_info->mutually_exclusive_operation_running,
4175                         1)) {
4176                         ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4177                 } else {
4178                         ret = btrfs_dev_replace_start(root, p);
4179                         atomic_set(
4180                          &root->fs_info->mutually_exclusive_operation_running,
4181                          0);
4182                 }
4183                 break;
4184         case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4185                 btrfs_dev_replace_status(root->fs_info, p);
4186                 ret = 0;
4187                 break;
4188         case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4189                 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4190                 break;
4191         default:
4192                 ret = -EINVAL;
4193                 break;
4194         }
4195 
4196         if (copy_to_user(arg, p, sizeof(*p)))
4197                 ret = -EFAULT;
4198 out:
4199         kfree(p);
4200         return ret;
4201 }
4202 
4203 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4204 {
4205         int ret = 0;
4206         int i;
4207         u64 rel_ptr;
4208         int size;
4209         struct btrfs_ioctl_ino_path_args *ipa = NULL;
4210         struct inode_fs_paths *ipath = NULL;
4211         struct btrfs_path *path;
4212 
4213         if (!capable(CAP_DAC_READ_SEARCH))
4214                 return -EPERM;
4215 
4216         path = btrfs_alloc_path();
4217         if (!path) {
4218                 ret = -ENOMEM;
4219                 goto out;
4220         }
4221 
4222         ipa = memdup_user(arg, sizeof(*ipa));
4223         if (IS_ERR(ipa)) {
4224                 ret = PTR_ERR(ipa);
4225                 ipa = NULL;
4226                 goto out;
4227         }
4228 
4229         size = min_t(u32, ipa->size, 4096);
4230         ipath = init_ipath(size, root, path);
4231         if (IS_ERR(ipath)) {
4232                 ret = PTR_ERR(ipath);
4233                 ipath = NULL;
4234                 goto out;
4235         }
4236 
4237         ret = paths_from_inode(ipa->inum, ipath);
4238         if (ret < 0)
4239                 goto out;
4240 
4241         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4242                 rel_ptr = ipath->fspath->val[i] -
4243                           (u64)(unsigned long)ipath->fspath->val;
4244                 ipath->fspath->val[i] = rel_ptr;
4245         }
4246 
4247         ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4248                            (void *)(unsigned long)ipath->fspath, size);
4249         if (ret) {
4250                 ret = -EFAULT;
4251                 goto out;
4252         }
4253 
4254 out:
4255         btrfs_free_path(path);
4256         free_ipath(ipath);
4257         kfree(ipa);
4258 
4259         return ret;
4260 }
4261 
4262 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4263 {
4264         struct btrfs_data_container *inodes = ctx;
4265         const size_t c = 3 * sizeof(u64);
4266 
4267         if (inodes->bytes_left >= c) {
4268                 inodes->bytes_left -= c;
4269                 inodes->val[inodes->elem_cnt] = inum;
4270                 inodes->val[inodes->elem_cnt + 1] = offset;
4271                 inodes->val[inodes->elem_cnt + 2] = root;
4272                 inodes->elem_cnt += 3;
4273         } else {
4274                 inodes->bytes_missing += c - inodes->bytes_left;
4275                 inodes->bytes_left = 0;
4276                 inodes->elem_missed += 3;
4277         }
4278 
4279         return 0;
4280 }
4281 
4282 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4283                                         void __user *arg)
4284 {
4285         int ret = 0;
4286         int size;
4287         struct btrfs_ioctl_logical_ino_args *loi;
4288         struct btrfs_data_container *inodes = NULL;
4289         struct btrfs_path *path = NULL;
4290 
4291         if (!capable(CAP_SYS_ADMIN))
4292                 return -EPERM;
4293 
4294         loi = memdup_user(arg, sizeof(*loi));
4295         if (IS_ERR(loi)) {
4296                 ret = PTR_ERR(loi);
4297                 loi = NULL;
4298                 goto out;
4299         }
4300 
4301         path = btrfs_alloc_path();
4302         if (!path) {
4303                 ret = -ENOMEM;
4304                 goto out;
4305         }
4306 
4307         size = min_t(u32, loi->size, 64 * 1024);
4308         inodes = init_data_container(size);
4309         if (IS_ERR(inodes)) {
4310                 ret = PTR_ERR(inodes);
4311                 inodes = NULL;
4312                 goto out;
4313         }
4314 
4315         ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4316                                           build_ino_list, inodes);
4317         if (ret == -EINVAL)
4318                 ret = -ENOENT;
4319         if (ret < 0)
4320                 goto out;
4321 
4322         ret = copy_to_user((void *)(unsigned long)loi->inodes,
4323                            (void *)(unsigned long)inodes, size);
4324         if (ret)
4325                 ret = -EFAULT;
4326 
4327 out:
4328         btrfs_free_path(path);
4329         vfree(inodes);
4330         kfree(loi);
4331 
4332         return ret;
4333 }
4334 
4335 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4336                                struct btrfs_ioctl_balance_args *bargs)
4337 {
4338         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4339 
4340         bargs->flags = bctl->flags;
4341 
4342         if (atomic_read(&fs_info->balance_running))
4343                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4344         if (atomic_read(&fs_info->balance_pause_req))
4345                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4346         if (atomic_read(&fs_info->balance_cancel_req))
4347                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4348 
4349         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4350         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4351         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4352 
4353         if (lock) {
4354                 spin_lock(&fs_info->balance_lock);
4355                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4356                 spin_unlock(&fs_info->balance_lock);
4357         } else {
4358                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4359         }
4360 }
4361 
4362 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4363 {
4364         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4365         struct btrfs_fs_info *fs_info = root->fs_info;
4366         struct btrfs_ioctl_balance_args *bargs;
4367         struct btrfs_balance_control *bctl;
4368         bool need_unlock; /* for mut. excl. ops lock */
4369         int ret;
4370 
4371         if (!capable(CAP_SYS_ADMIN))
4372                 return -EPERM;
4373 
4374         ret = mnt_want_write_file(file);
4375         if (ret)
4376                 return ret;
4377 
4378 again:
4379         if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4380                 mutex_lock(&fs_info->volume_mutex);
4381                 mutex_lock(&fs_info->balance_mutex);
4382                 need_unlock = true;
4383                 goto locked;
4384         }
4385 
4386         /*
4387          * mut. excl. ops lock is locked.  Three possibilites:
4388          *   (1) some other op is running
4389          *   (2) balance is running
4390          *   (3) balance is paused -- special case (think resume)
4391          */
4392         mutex_lock(&fs_info->balance_mutex);
4393         if (fs_info->balance_ctl) {
4394                 /* this is either (2) or (3) */
4395                 if (!atomic_read(&fs_info->balance_running)) {
4396                         mutex_unlock(&fs_info->balance_mutex);
4397                         if (!mutex_trylock(&fs_info->volume_mutex))
4398                                 goto again;
4399                         mutex_lock(&fs_info->balance_mutex);
4400 
4401                         if (fs_info->balance_ctl &&
4402                             !atomic_read(&fs_info->balance_running)) {
4403                                 /* this is (3) */
4404                                 need_unlock = false;
4405                                 goto locked;
4406                         }
4407 
4408                         mutex_unlock(&fs_info->balance_mutex);
4409                         mutex_unlock(&fs_info->volume_mutex);
4410                         goto again;
4411                 } else {
4412                         /* this is (2) */
4413                         mutex_unlock(&fs_info->balance_mutex);
4414                         ret = -EINPROGRESS;
4415                         goto out;
4416                 }
4417         } else {
4418                 /* this is (1) */
4419                 mutex_unlock(&fs_info->balance_mutex);
4420                 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4421                 goto out;
4422         }
4423 
4424 locked:
4425         BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4426 
4427         if (arg) {
4428                 bargs = memdup_user(arg, sizeof(*bargs));
4429                 if (IS_ERR(bargs)) {
4430                         ret = PTR_ERR(bargs);
4431                         goto out_unlock;
4432                 }
4433 
4434                 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4435                         if (!fs_info->balance_ctl) {
4436                                 ret = -ENOTCONN;
4437                                 goto out_bargs;
4438                         }
4439 
4440                         bctl = fs_info->balance_ctl;
4441                         spin_lock(&fs_info->balance_lock);
4442                         bctl->flags |= BTRFS_BALANCE_RESUME;
4443                         spin_unlock(&fs_info->balance_lock);
4444 
4445                         goto do_balance;
4446                 }
4447         } else {
4448                 bargs = NULL;
4449         }
4450 
4451         if (fs_info->balance_ctl) {
4452                 ret = -EINPROGRESS;
4453                 goto out_bargs;
4454         }
4455 
4456         bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4457         if (!bctl) {
4458                 ret = -ENOMEM;
4459                 goto out_bargs;
4460         }
4461 
4462         bctl->fs_info = fs_info;
4463         if (arg) {
4464                 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4465                 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4466                 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4467 
4468                 bctl->flags = bargs->flags;
4469         } else {
4470                 /* balance everything - no filters */
4471                 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4472         }
4473 
4474 do_balance:
4475         /*
4476          * Ownership of bctl and mutually_exclusive_operation_running
4477          * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
4478          * or, if restriper was paused all the way until unmount, in
4479          * free_fs_info.  mutually_exclusive_operation_running is
4480          * cleared in __cancel_balance.
4481          */
4482         need_unlock = false;
4483 
4484         ret = btrfs_balance(bctl, bargs);
4485 
4486         if (arg) {
4487                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4488                         ret = -EFAULT;
4489         }
4490 
4491 out_bargs:
4492         kfree(bargs);
4493 out_unlock:
4494         mutex_unlock(&fs_info->balance_mutex);
4495         mutex_unlock(&fs_info->volume_mutex);
4496         if (need_unlock)
4497                 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4498 out:
4499         mnt_drop_write_file(file);
4500         return ret;
4501 }
4502 
4503 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4504 {
4505         if (!capable(CAP_SYS_ADMIN))
4506                 return -EPERM;
4507 
4508         switch (cmd) {
4509         case BTRFS_BALANCE_CTL_PAUSE:
4510                 return btrfs_pause_balance(root->fs_info);
4511         case BTRFS_BALANCE_CTL_CANCEL:
4512                 return btrfs_cancel_balance(root->fs_info);
4513         }
4514 
4515         return -EINVAL;
4516 }
4517 
4518 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4519                                          void __user *arg)
4520 {
4521         struct btrfs_fs_info *fs_info = root->fs_info;
4522         struct btrfs_ioctl_balance_args *bargs;
4523         int ret = 0;
4524 
4525         if (!capable(CAP_SYS_ADMIN))
4526                 return -EPERM;
4527 
4528         mutex_lock(&fs_info->balance_mutex);
4529         if (!fs_info->balance_ctl) {
4530                 ret = -ENOTCONN;
4531                 goto out;
4532         }
4533 
4534         bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4535         if (!bargs) {
4536                 ret = -ENOMEM;
4537                 goto out;
4538         }
4539 
4540         update_ioctl_balance_args(fs_info, 1, bargs);
4541 
4542         if (copy_to_user(arg, bargs, sizeof(*bargs)))
4543                 ret = -EFAULT;
4544 
4545         kfree(bargs);
4546 out:
4547         mutex_unlock(&fs_info->balance_mutex);
4548         return ret;
4549 }
4550 
4551 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4552 {
4553         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4554         struct btrfs_ioctl_quota_ctl_args *sa;
4555         struct btrfs_trans_handle *trans = NULL;
4556         int ret;
4557         int err;
4558 
4559         if (!capable(CAP_SYS_ADMIN))
4560                 return -EPERM;
4561 
4562         ret = mnt_want_write_file(file);
4563         if (ret)
4564                 return ret;
4565 
4566         sa = memdup_user(arg, sizeof(*sa));
4567         if (IS_ERR(sa)) {
4568                 ret = PTR_ERR(sa);
4569                 goto drop_write;
4570         }
4571 
4572         down_write(&root->fs_info->subvol_sem);
4573         trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4574         if (IS_ERR(trans)) {
4575                 ret = PTR_ERR(trans);
4576                 goto out;
4577         }
4578 
4579         switch (sa->cmd) {
4580         case BTRFS_QUOTA_CTL_ENABLE:
4581                 ret = btrfs_quota_enable(trans, root->fs_info);
4582                 break;
4583         case BTRFS_QUOTA_CTL_DISABLE:
4584                 ret = btrfs_quota_disable(trans, root->fs_info);
4585                 break;
4586         default:
4587                 ret = -EINVAL;
4588                 break;
4589         }
4590 
4591         err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4592         if (err && !ret)
4593                 ret = err;
4594 out:
4595         kfree(sa);
4596         up_write(&root->fs_info->subvol_sem);
4597 drop_write:
4598         mnt_drop_write_file(file);
4599         return ret;
4600 }
4601 
4602 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4603 {
4604         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4605         struct btrfs_ioctl_qgroup_assign_args *sa;
4606         struct btrfs_trans_handle *trans;
4607         int ret;
4608         int err;
4609 
4610         if (!capable(CAP_SYS_ADMIN))
4611                 return -EPERM;
4612 
4613         ret = mnt_want_write_file(file);
4614         if (ret)
4615                 return ret;
4616 
4617         sa = memdup_user(arg, sizeof(*sa));
4618         if (IS_ERR(sa)) {
4619                 ret = PTR_ERR(sa);
4620                 goto drop_write;
4621         }
4622 
4623         trans = btrfs_join_transaction(root);
4624         if (IS_ERR(trans)) {
4625                 ret = PTR_ERR(trans);
4626                 goto out;
4627         }
4628 
4629         /* FIXME: check if the IDs really exist */
4630         if (sa->assign) {
4631                 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4632                                                 sa->src, sa->dst);
4633         } else {
4634                 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4635                                                 sa->src, sa->dst);
4636         }
4637 
4638         err = btrfs_end_transaction(trans, root);
4639         if (err && !ret)
4640                 ret = err;
4641 
4642 out:
4643         kfree(sa);
4644 drop_write:
4645         mnt_drop_write_file(file);
4646         return ret;
4647 }
4648 
4649 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4650 {
4651         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4652         struct btrfs_ioctl_qgroup_create_args *sa;
4653         struct btrfs_trans_handle *trans;
4654         int ret;
4655         int err;
4656 
4657         if (!capable(CAP_SYS_ADMIN))
4658                 return -EPERM;
4659 
4660         ret = mnt_want_write_file(file);
4661         if (ret)
4662                 return ret;
4663 
4664         sa = memdup_user(arg, sizeof(*sa));
4665         if (IS_ERR(sa)) {
4666                 ret = PTR_ERR(sa);
4667                 goto drop_write;
4668         }
4669 
4670         if (!sa->qgroupid) {
4671                 ret = -EINVAL;
4672                 goto out;
4673         }
4674 
4675         trans = btrfs_join_transaction(root);
4676         if (IS_ERR(trans)) {
4677                 ret = PTR_ERR(trans);
4678                 goto out;
4679         }
4680 
4681         /* FIXME: check if the IDs really exist */
4682         if (sa->create) {
4683                 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4684                                           NULL);
4685         } else {
4686                 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4687         }
4688 
4689         err = btrfs_end_transaction(trans, root);
4690         if (err && !ret)
4691                 ret = err;
4692 
4693 out:
4694         kfree(sa);
4695 drop_write:
4696         mnt_drop_write_file(file);
4697         return ret;
4698 }
4699 
4700 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4701 {
4702         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4703         struct btrfs_ioctl_qgroup_limit_args *sa;
4704         struct btrfs_trans_handle *trans;
4705         int ret;
4706         int err;
4707         u64 qgroupid;
4708 
4709         if (!capable(CAP_SYS_ADMIN))
4710                 return -EPERM;
4711 
4712         ret = mnt_want_write_file(file);
4713         if (ret)
4714                 return ret;
4715 
4716         sa = memdup_user(arg, sizeof(*sa));
4717         if (IS_ERR(sa)) {
4718                 ret = PTR_ERR(sa);
4719                 goto drop_write;
4720         }
4721 
4722         trans = btrfs_join_transaction(root);
4723         if (IS_ERR(trans)) {
4724                 ret = PTR_ERR(trans);
4725                 goto out;
4726         }
4727 
4728         qgroupid = sa->qgroupid;
4729         if (!qgroupid) {
4730                 /* take the current subvol as qgroup */
4731                 qgroupid = root->root_key.objectid;
4732         }
4733 
4734         /* FIXME: check if the IDs really exist */
4735         ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4736 
4737         err = btrfs_end_transaction(trans, root);
4738         if (err && !ret)
4739                 ret = err;
4740 
4741 out:
4742         kfree(sa);
4743 drop_write:
4744         mnt_drop_write_file(file);
4745         return ret;
4746 }
4747 
4748 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4749 {
4750         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4751         struct btrfs_ioctl_quota_rescan_args *qsa;
4752         int ret;
4753 
4754         if (!capable(CAP_SYS_ADMIN))
4755                 return -EPERM;
4756 
4757         ret = mnt_want_write_file(file);
4758         if (ret)
4759                 return ret;
4760 
4761         qsa = memdup_user(arg, sizeof(*qsa));
4762         if (IS_ERR(qsa)) {
4763                 ret = PTR_ERR(qsa);
4764                 goto drop_write;
4765         }
4766 
4767         if (qsa->flags) {
4768                 ret = -EINVAL;
4769                 goto out;
4770         }
4771 
4772         ret = btrfs_qgroup_rescan(root->fs_info);
4773 
4774 out:
4775         kfree(qsa);
4776 drop_write:
4777         mnt_drop_write_file(file);
4778         return ret;
4779 }
4780 
4781 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4782 {
4783         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4784         struct btrfs_ioctl_quota_rescan_args *qsa;
4785         int ret = 0;
4786 
4787         if (!capable(CAP_SYS_ADMIN))
4788                 return -EPERM;
4789 
4790         qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4791         if (!qsa)
4792                 return -ENOMEM;
4793 
4794         if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4795                 qsa->flags = 1;
4796                 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4797         }
4798 
4799         if (copy_to_user(arg, qsa, sizeof(*qsa)))
4800                 ret = -EFAULT;
4801 
4802         kfree(qsa);
4803         return ret;
4804 }
4805 
4806 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4807 {
4808         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4809 
4810         if (!capable(CAP_SYS_ADMIN))
4811                 return -EPERM;
4812 
4813         return btrfs_qgroup_wait_for_completion(root->fs_info);
4814 }
4815 
4816 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4817                                             struct btrfs_ioctl_received_subvol_args *sa)
4818 {
4819         struct inode *inode = file_inode(file);
4820         struct btrfs_root *root = BTRFS_I(inode)->root;
4821         struct btrfs_root_item *root_item = &root->root_item;
4822         struct btrfs_trans_handle *trans;
4823         struct timespec ct = CURRENT_TIME;
4824         int ret = 0;
4825         int received_uuid_changed;
4826 
4827         if (!inode_owner_or_capable(inode))
4828                 return -EPERM;
4829 
4830         ret = mnt_want_write_file(file);
4831         if (ret < 0)
4832                 return ret;
4833 
4834         down_write(&root->fs_info->subvol_sem);
4835 
4836         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4837                 ret = -EINVAL;
4838                 goto out;
4839         }
4840 
4841         if (btrfs_root_readonly(root)) {
4842                 ret = -EROFS;
4843                 goto out;
4844         }
4845 
4846         /*
4847          * 1 - root item
4848          * 2 - uuid items (received uuid + subvol uuid)
4849          */
4850         trans = btrfs_start_transaction(root, 3);
4851         if (IS_ERR(trans)) {
4852                 ret = PTR_ERR(trans);
4853                 trans = NULL;
4854                 goto out;
4855         }
4856 
4857         sa->rtransid = trans->transid;
4858         sa->rtime.sec = ct.tv_sec;
4859         sa->rtime.nsec = ct.tv_nsec;
4860 
4861         received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4862                                        BTRFS_UUID_SIZE);
4863         if (received_uuid_changed &&
4864             !btrfs_is_empty_uuid(root_item->received_uuid))
4865                 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4866                                     root_item->received_uuid,
4867                                     BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4868                                     root->root_key.objectid);
4869         memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4870         btrfs_set_root_stransid(root_item, sa->stransid);
4871         btrfs_set_root_rtransid(root_item, sa->rtransid);
4872         btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4873         btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4874         btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4875         btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4876 
4877         ret = btrfs_update_root(trans, root->fs_info->tree_root,
4878                                 &root->root_key, &root->root_item);
4879         if (ret < 0) {
4880                 btrfs_end_transaction(trans, root);
4881                 goto out;
4882         }
4883         if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4884                 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4885                                           sa->uuid,
4886                                           BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4887                                           root->root_key.objectid);
4888                 if (ret < 0 && ret != -EEXIST) {
4889                         btrfs_abort_transaction(trans, root, ret);
4890                         goto out;
4891                 }
4892         }
4893         ret = btrfs_commit_transaction(trans, root);
4894         if (ret < 0) {
4895                 btrfs_abort_transaction(trans, root, ret);
4896                 goto out;
4897         }
4898 
4899 out:
4900         up_write(&root->fs_info->subvol_sem);
4901         mnt_drop_write_file(file);
4902         return ret;
4903 }
4904 
4905 #ifdef CONFIG_64BIT
4906 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4907                                                 void __user *arg)
4908 {
4909         struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4910         struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4911         int ret = 0;
4912 
4913         args32 = memdup_user(arg, sizeof(*args32));
4914         if (IS_ERR(args32)) {
4915                 ret = PTR_ERR(args32);
4916                 args32 = NULL;
4917                 goto out;
4918         }
4919 
4920         args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4921         if (!args64) {
4922                 ret = -ENOMEM;
4923                 goto out;
4924         }
4925 
4926         memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4927         args64->stransid = args32->stransid;
4928         args64->rtransid = args32->rtransid;
4929         args64->stime.sec = args32->stime.sec;
4930         args64->stime.nsec = args32->stime.nsec;
4931         args64->rtime.sec = args32->rtime.sec;
4932         args64->rtime.nsec = args32->rtime.nsec;
4933         args64->flags = args32->flags;
4934 
4935         ret = _btrfs_ioctl_set_received_subvol(file, args64);
4936         if (ret)
4937                 goto out;
4938 
4939         memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4940         args32->stransid = args64->stransid;
4941         args32->rtransid = args64->rtransid;
4942         args32->stime.sec = args64->stime.sec;
4943         args32->stime.nsec = args64->stime.nsec;
4944         args32->rtime.sec = args64->rtime.sec;
4945         args32->rtime.nsec = args64->rtime.nsec;
4946         args32->flags = args64->flags;
4947 
4948         ret = copy_to_user(arg, args32, sizeof(*args32));
4949         if (ret)
4950                 ret = -EFAULT;
4951 
4952 out:
4953         kfree(args32);
4954         kfree(args64);
4955         return ret;
4956 }
4957 #endif
4958 
4959 static long btrfs_ioctl_set_received_subvol(struct file *file,
4960                                             void __user *arg)
4961 {
4962         struct btrfs_ioctl_received_subvol_args *sa = NULL;
4963         int ret = 0;
4964 
4965         sa = memdup_user(arg, sizeof(*sa));
4966         if (IS_ERR(sa)) {
4967                 ret = PTR_ERR(sa);
4968                 sa = NULL;
4969                 goto out;
4970         }
4971 
4972         ret = _btrfs_ioctl_set_received_subvol(file, sa);
4973 
4974         if (ret)
4975                 goto out;
4976 
4977         ret = copy_to_user(arg, sa, sizeof(*sa));
4978         if (ret)
4979                 ret = -EFAULT;
4980 
4981 out:
4982         kfree(sa);
4983         return ret;
4984 }
4985 
4986 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4987 {
4988         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4989         size_t len;
4990         int ret;
4991         char label[BTRFS_LABEL_SIZE];
4992 
4993         spin_lock(&root->fs_info->super_lock);
4994         memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4995         spin_unlock(&root->fs_info->super_lock);
4996 
4997         len = strnlen(label, BTRFS_LABEL_SIZE);
4998 
4999         if (len == BTRFS_LABEL_SIZE) {
5000                 btrfs_warn(root->fs_info,
5001                         "label is too long, return the first %zu bytes", --len);
5002         }
5003 
5004         ret = copy_to_user(arg, label, len);
5005 
5006         return ret ? -EFAULT : 0;
5007 }
5008 
5009 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5010 {
5011         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5012         struct btrfs_super_block *super_block = root->fs_info->super_copy;
5013         struct btrfs_trans_handle *trans;
5014         char label[BTRFS_LABEL_SIZE];
5015         int ret;
5016 
5017         if (!capable(CAP_SYS_ADMIN))
5018                 return -EPERM;
5019 
5020         if (copy_from_user(label, arg, sizeof(label)))
5021                 return -EFAULT;
5022 
5023         if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5024                 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5025                        BTRFS_LABEL_SIZE - 1);
5026                 return -EINVAL;
5027         }
5028 
5029         ret = mnt_want_write_file(file);
5030         if (ret)
5031                 return ret;
5032 
5033         trans = btrfs_start_transaction(root, 0);
5034         if (IS_ERR(trans)) {
5035                 ret = PTR_ERR(trans);
5036                 goto out_unlock;
5037         }
5038 
5039         spin_lock(&root->fs_info->super_lock);
5040         strcpy(super_block->label, label);
5041         spin_unlock(&root->fs_info->super_lock);
5042         ret = btrfs_commit_transaction(trans, root);
5043 
5044 out_unlock:
5045         mnt_drop_write_file(file);
5046         return ret;
5047 }
5048 
5049 #define INIT_FEATURE_FLAGS(suffix) \
5050         { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5051           .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5052           .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5053 
5054 static int btrfs_ioctl_get_supported_features(struct file *file,
5055                                               void __user *arg)
5056 {
5057         static struct btrfs_ioctl_feature_flags features[3] = {
5058                 INIT_FEATURE_FLAGS(SUPP),
5059                 INIT_FEATURE_FLAGS(SAFE_SET),
5060                 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5061         };
5062 
5063         if (copy_to_user(arg, &features, sizeof(features)))
5064                 return -EFAULT;
5065 
5066         return 0;
5067 }
5068 
5069 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5070 {
5071         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5072         struct btrfs_super_block *super_block = root->fs_info->super_copy;
5073         struct btrfs_ioctl_feature_flags features;
5074 
5075         features.compat_flags = btrfs_super_compat_flags(super_block);
5076         features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5077         features.incompat_flags = btrfs_super_incompat_flags(super_block);
5078 
5079         if (copy_to_user(arg, &features, sizeof(features)))
5080                 return -EFAULT;
5081 
5082         return 0;
5083 }
5084 
5085 static int check_feature_bits(struct btrfs_root *root,
5086                               enum btrfs_feature_set set,
5087                               u64 change_mask, u64 flags, u64 supported_flags,
5088                               u64 safe_set, u64 safe_clear)
5089 {
5090         const char *type = btrfs_feature_set_names[set];
5091         char *names;
5092         u64 disallowed, unsupported;
5093         u64 set_mask = flags & change_mask;
5094         u64 clear_mask = ~flags & change_mask;
5095 
5096         unsupported = set_mask & ~supported_flags;
5097         if (unsupported) {
5098                 names = btrfs_printable_features(set, unsupported);
5099                 if (names) {
5100                         btrfs_warn(root->fs_info,
5101                            "this kernel does not support the %s feature bit%s",
5102                            names, strchr(names, ',') ? "s" : "");
5103                         kfree(names);
5104                 } else
5105                         btrfs_warn(root->fs_info,
5106                            "this kernel does not support %s bits 0x%llx",
5107                            type, unsupported);
5108                 return -EOPNOTSUPP;
5109         }
5110 
5111         disallowed = set_mask & ~safe_set;
5112         if (disallowed) {
5113                 names = btrfs_printable_features(set, disallowed);
5114                 if (names) {
5115                         btrfs_warn(root->fs_info,
5116                            "can't set the %s feature bit%s while mounted",
5117                            names, strchr(names, ',') ? "s" : "");
5118                         kfree(names);
5119                 } else
5120                         btrfs_warn(root->fs_info,
5121                            "can't set %s bits 0x%llx while mounted",
5122                            type, disallowed);
5123                 return -EPERM;
5124         }
5125 
5126         disallowed = clear_mask & ~safe_clear;
5127         if (disallowed) {
5128                 names = btrfs_printable_features(set, disallowed);
5129                 if (names) {
5130                         btrfs_warn(root->fs_info,
5131                            "can't clear the %s feature bit%s while mounted",
5132                            names, strchr(names, ',') ? "s" : "");
5133                         kfree(names);
5134                 } else
5135                         btrfs_warn(root->fs_info,
5136                            "can't clear %s bits 0x%llx while mounted",
5137                            type, disallowed);
5138                 return -EPERM;
5139         }
5140 
5141         return 0;
5142 }
5143 
5144 #define check_feature(root, change_mask, flags, mask_base)      \
5145 check_feature_bits(root, FEAT_##mask_base, change_mask, flags,  \
5146                    BTRFS_FEATURE_ ## mask_base ## _SUPP,        \
5147                    BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,    \
5148                    BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5149 
5150 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5151 {
5152         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5153         struct btrfs_super_block *super_block = root->fs_info->super_copy;
5154         struct btrfs_ioctl_feature_flags flags[2];
5155         struct btrfs_trans_handle *trans;
5156         u64 newflags;
5157         int ret;
5158 
5159         if (!capable(CAP_SYS_ADMIN))
5160                 return -EPERM;
5161 
5162         if (copy_from_user(flags, arg, sizeof(flags)))
5163                 return -EFAULT;
5164 
5165         /* Nothing to do */
5166         if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5167             !flags[0].incompat_flags)
5168                 return 0;
5169 
5170         ret = check_feature(root, flags[0].compat_flags,
5171                             flags[1].compat_flags, COMPAT);
5172         if (ret)
5173                 return ret;
5174 
5175         ret = check_feature(root, flags[0].compat_ro_flags,
5176                             flags[1].compat_ro_flags, COMPAT_RO);
5177         if (ret)
5178                 return ret;
5179 
5180         ret = check_feature(root, flags[0].incompat_flags,
5181                             flags[1].incompat_flags, INCOMPAT);
5182         if (ret)
5183                 return ret;
5184 
5185         trans = btrfs_start_transaction(root, 0);
5186         if (IS_ERR(trans))
5187                 return PTR_ERR(trans);
5188 
5189         spin_lock(&root->fs_info->super_lock);
5190         newflags = btrfs_super_compat_flags(super_block);
5191         newflags |= flags[0].compat_flags & flags[1].compat_flags;
5192         newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5193         btrfs_set_super_compat_flags(super_block, newflags);
5194 
5195         newflags = btrfs_super_compat_ro_flags(super_block);
5196         newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5197         newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5198         btrfs_set_super_compat_ro_flags(super_block, newflags);
5199 
5200         newflags = btrfs_super_incompat_flags(super_block);
5201         newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5202         newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5203         btrfs_set_super_incompat_flags(super_block, newflags);
5204         spin_unlock(&root->fs_info->super_lock);
5205 
5206         return btrfs_commit_transaction(trans, root);
5207 }
5208 
5209 long btrfs_ioctl(struct file *file, unsigned int
5210                 cmd, unsigned long arg)
5211 {
5212         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5213         void __user *argp = (void __user *)arg;
5214 
5215         switch (cmd) {
5216         case FS_IOC_GETFLAGS:
5217                 return btrfs_ioctl_getflags(file, argp);
5218         case FS_IOC_SETFLAGS:
5219                 return btrfs_ioctl_setflags(file, argp);
5220         case FS_IOC_GETVERSION:
5221                 return btrfs_ioctl_getversion(file, argp);
5222         case FITRIM:
5223                 return btrfs_ioctl_fitrim(file, argp);
5224         case BTRFS_IOC_SNAP_CREATE:
5225                 return btrfs_ioctl_snap_create(file, argp, 0);
5226         case BTRFS_IOC_SNAP_CREATE_V2:
5227                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5228         case BTRFS_IOC_SUBVOL_CREATE:
5229                 return btrfs_ioctl_snap_create(file, argp, 1);
5230         case BTRFS_IOC_SUBVOL_CREATE_V2:
5231                 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5232         case BTRFS_IOC_SNAP_DESTROY:
5233                 return btrfs_ioctl_snap_destroy(file, argp);
5234         case BTRFS_IOC_SUBVOL_GETFLAGS:
5235                 return btrfs_ioctl_subvol_getflags(file, argp);
5236         case BTRFS_IOC_SUBVOL_SETFLAGS:
5237                 return btrfs_ioctl_subvol_setflags(file, argp);
5238         case BTRFS_IOC_DEFAULT_SUBVOL:
5239                 return btrfs_ioctl_default_subvol(file, argp);
5240         case BTRFS_IOC_DEFRAG:
5241                 return btrfs_ioctl_defrag(file, NULL);
5242         case BTRFS_IOC_DEFRAG_RANGE:
5243                 return btrfs_ioctl_defrag(file, argp);
5244         case BTRFS_IOC_RESIZE:
5245                 return btrfs_ioctl_resize(file, argp);
5246         case BTRFS_IOC_ADD_DEV:
5247                 return btrfs_ioctl_add_dev(root, argp);
5248         case BTRFS_IOC_RM_DEV:
5249                 return btrfs_ioctl_rm_dev(file, argp);
5250         case BTRFS_IOC_FS_INFO:
5251                 return btrfs_ioctl_fs_info(root, argp);
5252         case BTRFS_IOC_DEV_INFO:
5253                 return btrfs_ioctl_dev_info(root, argp);
5254         case BTRFS_IOC_BALANCE:
5255                 return btrfs_ioctl_balance(file, NULL);
5256         case BTRFS_IOC_CLONE:
5257                 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5258         case BTRFS_IOC_CLONE_RANGE:
5259                 return btrfs_ioctl_clone_range(file, argp);
5260         case BTRFS_IOC_TRANS_START:
5261                 return btrfs_ioctl_trans_start(file);
5262         case BTRFS_IOC_TRANS_END:
5263                 return btrfs_ioctl_trans_end(file);
5264         case BTRFS_IOC_TREE_SEARCH:
5265                 return btrfs_ioctl_tree_search(file, argp);
5266         case BTRFS_IOC_TREE_SEARCH_V2:
5267                 return btrfs_ioctl_tree_search_v2(file, argp);
5268         case BTRFS_IOC_INO_LOOKUP:
5269                 return btrfs_ioctl_ino_lookup(file, argp);
5270         case BTRFS_IOC_INO_PATHS:
5271                 return btrfs_ioctl_ino_to_path(root, argp);
5272         case BTRFS_IOC_LOGICAL_INO:
5273                 return btrfs_ioctl_logical_to_ino(root, argp);
5274         case BTRFS_IOC_SPACE_INFO:
5275                 return btrfs_ioctl_space_info(root, argp);
5276         case BTRFS_IOC_SYNC: {
5277                 int ret;
5278 
5279                 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5280                 if (ret)
5281                         return ret;
5282                 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5283                 /*
5284                  * The transaction thread may want to do more work,
5285                  * namely it pokes the cleaner ktread that will start
5286                  * processing uncleaned subvols.
5287                  */
5288                 wake_up_process(root->fs_info->transaction_kthread);
5289                 return ret;
5290         }
5291         case BTRFS_IOC_START_SYNC:
5292                 return btrfs_ioctl_start_sync(root, argp);
5293         case BTRFS_IOC_WAIT_SYNC:
5294                 return btrfs_ioctl_wait_sync(root, argp);
5295         case BTRFS_IOC_SCRUB:
5296                 return btrfs_ioctl_scrub(file, argp);
5297         case BTRFS_IOC_SCRUB_CANCEL:
5298                 return btrfs_ioctl_scrub_cancel(root, argp);
5299         case BTRFS_IOC_SCRUB_PROGRESS:
5300                 return btrfs_ioctl_scrub_progress(root, argp);
5301         case BTRFS_IOC_BALANCE_V2:
5302                 return btrfs_ioctl_balance(file, argp);
5303         case BTRFS_IOC_BALANCE_CTL:
5304                 return btrfs_ioctl_balance_ctl(root, arg);
5305         case BTRFS_IOC_BALANCE_PROGRESS:
5306                 return btrfs_ioctl_balance_progress(root, argp);
5307         case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5308                 return btrfs_ioctl_set_received_subvol(file, argp);
5309 #ifdef CONFIG_64BIT
5310         case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5311                 return btrfs_ioctl_set_received_subvol_32(file, argp);
5312 #endif
5313         case BTRFS_IOC_SEND:
5314                 return btrfs_ioctl_send(file, argp);
5315         case BTRFS_IOC_GET_DEV_STATS:
5316                 return btrfs_ioctl_get_dev_stats(root, argp);
5317         case BTRFS_IOC_QUOTA_CTL:
5318                 return btrfs_ioctl_quota_ctl(file, argp);
5319         case BTRFS_IOC_QGROUP_ASSIGN:
5320                 return btrfs_ioctl_qgroup_assign(file, argp);
5321         case BTRFS_IOC_QGROUP_CREATE:
5322                 return btrfs_ioctl_qgroup_create(file, argp);
5323         case BTRFS_IOC_QGROUP_LIMIT:
5324                 return btrfs_ioctl_qgroup_limit(file, argp);
5325         case BTRFS_IOC_QUOTA_RESCAN:
5326                 return btrfs_ioctl_quota_rescan(file, argp);
5327         case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5328                 return btrfs_ioctl_quota_rescan_status(file, argp);
5329         case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5330                 return btrfs_ioctl_quota_rescan_wait(file, argp);
5331         case BTRFS_IOC_DEV_REPLACE:
5332                 return btrfs_ioctl_dev_replace(root, argp);
5333         case BTRFS_IOC_GET_FSLABEL:
5334                 return btrfs_ioctl_get_fslabel(file, argp);
5335         case BTRFS_IOC_SET_FSLABEL:
5336                 return btrfs_ioctl_set_fslabel(file, argp);
5337         case BTRFS_IOC_FILE_EXTENT_SAME:
5338                 return btrfs_ioctl_file_extent_same(file, argp);
5339         case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5340                 return btrfs_ioctl_get_supported_features(file, argp);
5341         case BTRFS_IOC_GET_FEATURES:
5342                 return btrfs_ioctl_get_features(file, argp);
5343         case BTRFS_IOC_SET_FEATURES:
5344                 return btrfs_ioctl_set_features(file, argp);
5345         }
5346 
5347         return -ENOTTY;
5348 }
5349 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp