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

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

Version: ~ [ linux-5.4-rc7 ] ~ [ linux-5.3.10 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.83 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.153 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.200 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.200 ] ~ [ 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-3.16.76 ] ~ [ 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-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ 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 #include "tree-log.h"
 63 #include "compression.h"
 64 
 65 #ifdef CONFIG_64BIT
 66 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
 67  * structures are incorrect, as the timespec structure from userspace
 68  * is 4 bytes too small. We define these alternatives here to teach
 69  * the kernel about the 32-bit struct packing.
 70  */
 71 struct btrfs_ioctl_timespec_32 {
 72         __u64 sec;
 73         __u32 nsec;
 74 } __attribute__ ((__packed__));
 75 
 76 struct btrfs_ioctl_received_subvol_args_32 {
 77         char    uuid[BTRFS_UUID_SIZE];  /* in */
 78         __u64   stransid;               /* in */
 79         __u64   rtransid;               /* out */
 80         struct btrfs_ioctl_timespec_32 stime; /* in */
 81         struct btrfs_ioctl_timespec_32 rtime; /* out */
 82         __u64   flags;                  /* in */
 83         __u64   reserved[16];           /* in */
 84 } __attribute__ ((__packed__));
 85 
 86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
 87                                 struct btrfs_ioctl_received_subvol_args_32)
 88 #endif
 89 
 90 
 91 static int btrfs_clone(struct inode *src, struct inode *inode,
 92                        u64 off, u64 olen, u64 olen_aligned, u64 destoff,
 93                        int no_time_update);
 94 
 95 /* Mask out flags that are inappropriate for the given type of inode. */
 96 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
 97 {
 98         if (S_ISDIR(mode))
 99                 return flags;
100         else if (S_ISREG(mode))
101                 return flags & ~FS_DIRSYNC_FL;
102         else
103                 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
104 }
105 
106 /*
107  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
108  */
109 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
110 {
111         unsigned int iflags = 0;
112 
113         if (flags & BTRFS_INODE_SYNC)
114                 iflags |= FS_SYNC_FL;
115         if (flags & BTRFS_INODE_IMMUTABLE)
116                 iflags |= FS_IMMUTABLE_FL;
117         if (flags & BTRFS_INODE_APPEND)
118                 iflags |= FS_APPEND_FL;
119         if (flags & BTRFS_INODE_NODUMP)
120                 iflags |= FS_NODUMP_FL;
121         if (flags & BTRFS_INODE_NOATIME)
122                 iflags |= FS_NOATIME_FL;
123         if (flags & BTRFS_INODE_DIRSYNC)
124                 iflags |= FS_DIRSYNC_FL;
125         if (flags & BTRFS_INODE_NODATACOW)
126                 iflags |= FS_NOCOW_FL;
127 
128         if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
129                 iflags |= FS_COMPR_FL;
130         else if (flags & BTRFS_INODE_NOCOMPRESS)
131                 iflags |= FS_NOCOMP_FL;
132 
133         return iflags;
134 }
135 
136 /*
137  * Update inode->i_flags based on the btrfs internal flags.
138  */
139 void btrfs_update_iflags(struct inode *inode)
140 {
141         struct btrfs_inode *ip = BTRFS_I(inode);
142         unsigned int new_fl = 0;
143 
144         if (ip->flags & BTRFS_INODE_SYNC)
145                 new_fl |= S_SYNC;
146         if (ip->flags & BTRFS_INODE_IMMUTABLE)
147                 new_fl |= S_IMMUTABLE;
148         if (ip->flags & BTRFS_INODE_APPEND)
149                 new_fl |= S_APPEND;
150         if (ip->flags & BTRFS_INODE_NOATIME)
151                 new_fl |= S_NOATIME;
152         if (ip->flags & BTRFS_INODE_DIRSYNC)
153                 new_fl |= S_DIRSYNC;
154 
155         set_mask_bits(&inode->i_flags,
156                       S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
157                       new_fl);
158 }
159 
160 /*
161  * Inherit flags from the parent inode.
162  *
163  * Currently only the compression flags and the cow flags are inherited.
164  */
165 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
166 {
167         unsigned int flags;
168 
169         if (!dir)
170                 return;
171 
172         flags = BTRFS_I(dir)->flags;
173 
174         if (flags & BTRFS_INODE_NOCOMPRESS) {
175                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
176                 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
177         } else if (flags & BTRFS_INODE_COMPRESS) {
178                 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
179                 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
180         }
181 
182         if (flags & BTRFS_INODE_NODATACOW) {
183                 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
184                 if (S_ISREG(inode->i_mode))
185                         BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
186         }
187 
188         btrfs_update_iflags(inode);
189 }
190 
191 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
192 {
193         struct btrfs_inode *ip = BTRFS_I(file_inode(file));
194         unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
195 
196         if (copy_to_user(arg, &flags, sizeof(flags)))
197                 return -EFAULT;
198         return 0;
199 }
200 
201 static int check_flags(unsigned int flags)
202 {
203         if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
204                       FS_NOATIME_FL | FS_NODUMP_FL | \
205                       FS_SYNC_FL | FS_DIRSYNC_FL | \
206                       FS_NOCOMP_FL | FS_COMPR_FL |
207                       FS_NOCOW_FL))
208                 return -EOPNOTSUPP;
209 
210         if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
211                 return -EINVAL;
212 
213         return 0;
214 }
215 
216 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
217 {
218         struct inode *inode = file_inode(file);
219         struct btrfs_inode *ip = BTRFS_I(inode);
220         struct btrfs_root *root = ip->root;
221         struct btrfs_trans_handle *trans;
222         unsigned int flags, oldflags;
223         int ret;
224         u64 ip_oldflags;
225         unsigned int i_oldflags;
226         umode_t mode;
227 
228         if (!inode_owner_or_capable(inode))
229                 return -EPERM;
230 
231         if (btrfs_root_readonly(root))
232                 return -EROFS;
233 
234         if (copy_from_user(&flags, arg, sizeof(flags)))
235                 return -EFAULT;
236 
237         ret = check_flags(flags);
238         if (ret)
239                 return ret;
240 
241         ret = mnt_want_write_file(file);
242         if (ret)
243                 return ret;
244 
245         inode_lock(inode);
246 
247         ip_oldflags = ip->flags;
248         i_oldflags = inode->i_flags;
249         mode = inode->i_mode;
250 
251         flags = btrfs_mask_flags(inode->i_mode, flags);
252         oldflags = btrfs_flags_to_ioctl(ip->flags);
253         if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
254                 if (!capable(CAP_LINUX_IMMUTABLE)) {
255                         ret = -EPERM;
256                         goto out_unlock;
257                 }
258         }
259 
260         if (flags & FS_SYNC_FL)
261                 ip->flags |= BTRFS_INODE_SYNC;
262         else
263                 ip->flags &= ~BTRFS_INODE_SYNC;
264         if (flags & FS_IMMUTABLE_FL)
265                 ip->flags |= BTRFS_INODE_IMMUTABLE;
266         else
267                 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
268         if (flags & FS_APPEND_FL)
269                 ip->flags |= BTRFS_INODE_APPEND;
270         else
271                 ip->flags &= ~BTRFS_INODE_APPEND;
272         if (flags & FS_NODUMP_FL)
273                 ip->flags |= BTRFS_INODE_NODUMP;
274         else
275                 ip->flags &= ~BTRFS_INODE_NODUMP;
276         if (flags & FS_NOATIME_FL)
277                 ip->flags |= BTRFS_INODE_NOATIME;
278         else
279                 ip->flags &= ~BTRFS_INODE_NOATIME;
280         if (flags & FS_DIRSYNC_FL)
281                 ip->flags |= BTRFS_INODE_DIRSYNC;
282         else
283                 ip->flags &= ~BTRFS_INODE_DIRSYNC;
284         if (flags & FS_NOCOW_FL) {
285                 if (S_ISREG(mode)) {
286                         /*
287                          * It's safe to turn csums off here, no extents exist.
288                          * Otherwise we want the flag to reflect the real COW
289                          * status of the file and will not set it.
290                          */
291                         if (inode->i_size == 0)
292                                 ip->flags |= BTRFS_INODE_NODATACOW
293                                            | BTRFS_INODE_NODATASUM;
294                 } else {
295                         ip->flags |= BTRFS_INODE_NODATACOW;
296                 }
297         } else {
298                 /*
299                  * Revert back under same assuptions as above
300                  */
301                 if (S_ISREG(mode)) {
302                         if (inode->i_size == 0)
303                                 ip->flags &= ~(BTRFS_INODE_NODATACOW
304                                              | BTRFS_INODE_NODATASUM);
305                 } else {
306                         ip->flags &= ~BTRFS_INODE_NODATACOW;
307                 }
308         }
309 
310         /*
311          * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
312          * flag may be changed automatically if compression code won't make
313          * things smaller.
314          */
315         if (flags & FS_NOCOMP_FL) {
316                 ip->flags &= ~BTRFS_INODE_COMPRESS;
317                 ip->flags |= BTRFS_INODE_NOCOMPRESS;
318 
319                 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
320                 if (ret && ret != -ENODATA)
321                         goto out_drop;
322         } else if (flags & FS_COMPR_FL) {
323                 const char *comp;
324 
325                 ip->flags |= BTRFS_INODE_COMPRESS;
326                 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
327 
328                 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
329                         comp = "lzo";
330                 else
331                         comp = "zlib";
332                 ret = btrfs_set_prop(inode, "btrfs.compression",
333                                      comp, strlen(comp), 0);
334                 if (ret)
335                         goto out_drop;
336 
337         } else {
338                 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
339                 if (ret && ret != -ENODATA)
340                         goto out_drop;
341                 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
342         }
343 
344         trans = btrfs_start_transaction(root, 1);
345         if (IS_ERR(trans)) {
346                 ret = PTR_ERR(trans);
347                 goto out_drop;
348         }
349 
350         btrfs_update_iflags(inode);
351         inode_inc_iversion(inode);
352         inode->i_ctime = current_fs_time(inode->i_sb);
353         ret = btrfs_update_inode(trans, root, inode);
354 
355         btrfs_end_transaction(trans, root);
356  out_drop:
357         if (ret) {
358                 ip->flags = ip_oldflags;
359                 inode->i_flags = i_oldflags;
360         }
361 
362  out_unlock:
363         inode_unlock(inode);
364         mnt_drop_write_file(file);
365         return ret;
366 }
367 
368 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
369 {
370         struct inode *inode = file_inode(file);
371 
372         return put_user(inode->i_generation, arg);
373 }
374 
375 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
376 {
377         struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
378         struct btrfs_device *device;
379         struct request_queue *q;
380         struct fstrim_range range;
381         u64 minlen = ULLONG_MAX;
382         u64 num_devices = 0;
383         u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
384         int ret;
385 
386         if (!capable(CAP_SYS_ADMIN))
387                 return -EPERM;
388 
389         rcu_read_lock();
390         list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
391                                 dev_list) {
392                 if (!device->bdev)
393                         continue;
394                 q = bdev_get_queue(device->bdev);
395                 if (blk_queue_discard(q)) {
396                         num_devices++;
397                         minlen = min((u64)q->limits.discard_granularity,
398                                      minlen);
399                 }
400         }
401         rcu_read_unlock();
402 
403         if (!num_devices)
404                 return -EOPNOTSUPP;
405         if (copy_from_user(&range, arg, sizeof(range)))
406                 return -EFAULT;
407         if (range.start > total_bytes ||
408             range.len < fs_info->sb->s_blocksize)
409                 return -EINVAL;
410 
411         range.len = min(range.len, total_bytes - range.start);
412         range.minlen = max(range.minlen, minlen);
413         ret = btrfs_trim_fs(fs_info->tree_root, &range);
414         if (ret < 0)
415                 return ret;
416 
417         if (copy_to_user(arg, &range, sizeof(range)))
418                 return -EFAULT;
419 
420         return 0;
421 }
422 
423 int btrfs_is_empty_uuid(u8 *uuid)
424 {
425         int i;
426 
427         for (i = 0; i < BTRFS_UUID_SIZE; i++) {
428                 if (uuid[i])
429                         return 0;
430         }
431         return 1;
432 }
433 
434 static noinline int create_subvol(struct inode *dir,
435                                   struct dentry *dentry,
436                                   char *name, int namelen,
437                                   u64 *async_transid,
438                                   struct btrfs_qgroup_inherit *inherit)
439 {
440         struct btrfs_trans_handle *trans;
441         struct btrfs_key key;
442         struct btrfs_root_item root_item;
443         struct btrfs_inode_item *inode_item;
444         struct extent_buffer *leaf;
445         struct btrfs_root *root = BTRFS_I(dir)->root;
446         struct btrfs_root *new_root;
447         struct btrfs_block_rsv block_rsv;
448         struct timespec cur_time = current_fs_time(dir->i_sb);
449         struct inode *inode;
450         int ret;
451         int err;
452         u64 objectid;
453         u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
454         u64 index = 0;
455         u64 qgroup_reserved;
456         uuid_le new_uuid;
457 
458         ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
459         if (ret)
460                 return ret;
461 
462         /*
463          * Don't create subvolume whose level is not zero. Or qgroup will be
464          * screwed up since it assume subvolme qgroup's level to be 0.
465          */
466         if (btrfs_qgroup_level(objectid))
467                 return -ENOSPC;
468 
469         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
470         /*
471          * The same as the snapshot creation, please see the comment
472          * of create_snapshot().
473          */
474         ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
475                                                8, &qgroup_reserved, false);
476         if (ret)
477                 return ret;
478 
479         trans = btrfs_start_transaction(root, 0);
480         if (IS_ERR(trans)) {
481                 ret = PTR_ERR(trans);
482                 btrfs_subvolume_release_metadata(root, &block_rsv,
483                                                  qgroup_reserved);
484                 return ret;
485         }
486         trans->block_rsv = &block_rsv;
487         trans->bytes_reserved = block_rsv.size;
488 
489         ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
490         if (ret)
491                 goto fail;
492 
493         leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
494         if (IS_ERR(leaf)) {
495                 ret = PTR_ERR(leaf);
496                 goto fail;
497         }
498 
499         memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
500         btrfs_set_header_bytenr(leaf, leaf->start);
501         btrfs_set_header_generation(leaf, trans->transid);
502         btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
503         btrfs_set_header_owner(leaf, objectid);
504 
505         write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
506                             BTRFS_FSID_SIZE);
507         write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
508                             btrfs_header_chunk_tree_uuid(leaf),
509                             BTRFS_UUID_SIZE);
510         btrfs_mark_buffer_dirty(leaf);
511 
512         memset(&root_item, 0, sizeof(root_item));
513 
514         inode_item = &root_item.inode;
515         btrfs_set_stack_inode_generation(inode_item, 1);
516         btrfs_set_stack_inode_size(inode_item, 3);
517         btrfs_set_stack_inode_nlink(inode_item, 1);
518         btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
519         btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
520 
521         btrfs_set_root_flags(&root_item, 0);
522         btrfs_set_root_limit(&root_item, 0);
523         btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
524 
525         btrfs_set_root_bytenr(&root_item, leaf->start);
526         btrfs_set_root_generation(&root_item, trans->transid);
527         btrfs_set_root_level(&root_item, 0);
528         btrfs_set_root_refs(&root_item, 1);
529         btrfs_set_root_used(&root_item, leaf->len);
530         btrfs_set_root_last_snapshot(&root_item, 0);
531 
532         btrfs_set_root_generation_v2(&root_item,
533                         btrfs_root_generation(&root_item));
534         uuid_le_gen(&new_uuid);
535         memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
536         btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
537         btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
538         root_item.ctime = root_item.otime;
539         btrfs_set_root_ctransid(&root_item, trans->transid);
540         btrfs_set_root_otransid(&root_item, trans->transid);
541 
542         btrfs_tree_unlock(leaf);
543         free_extent_buffer(leaf);
544         leaf = NULL;
545 
546         btrfs_set_root_dirid(&root_item, new_dirid);
547 
548         key.objectid = objectid;
549         key.offset = 0;
550         key.type = BTRFS_ROOT_ITEM_KEY;
551         ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
552                                 &root_item);
553         if (ret)
554                 goto fail;
555 
556         key.offset = (u64)-1;
557         new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
558         if (IS_ERR(new_root)) {
559                 ret = PTR_ERR(new_root);
560                 btrfs_abort_transaction(trans, root, ret);
561                 goto fail;
562         }
563 
564         btrfs_record_root_in_trans(trans, new_root);
565 
566         ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
567         if (ret) {
568                 /* We potentially lose an unused inode item here */
569                 btrfs_abort_transaction(trans, root, ret);
570                 goto fail;
571         }
572 
573         mutex_lock(&new_root->objectid_mutex);
574         new_root->highest_objectid = new_dirid;
575         mutex_unlock(&new_root->objectid_mutex);
576 
577         /*
578          * insert the directory item
579          */
580         ret = btrfs_set_inode_index(dir, &index);
581         if (ret) {
582                 btrfs_abort_transaction(trans, root, ret);
583                 goto fail;
584         }
585 
586         ret = btrfs_insert_dir_item(trans, root,
587                                     name, namelen, dir, &key,
588                                     BTRFS_FT_DIR, index);
589         if (ret) {
590                 btrfs_abort_transaction(trans, root, ret);
591                 goto fail;
592         }
593 
594         btrfs_i_size_write(dir, dir->i_size + namelen * 2);
595         ret = btrfs_update_inode(trans, root, dir);
596         BUG_ON(ret);
597 
598         ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
599                                  objectid, root->root_key.objectid,
600                                  btrfs_ino(dir), index, name, namelen);
601         BUG_ON(ret);
602 
603         ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
604                                   root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
605                                   objectid);
606         if (ret)
607                 btrfs_abort_transaction(trans, root, ret);
608 
609 fail:
610         trans->block_rsv = NULL;
611         trans->bytes_reserved = 0;
612         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
613 
614         if (async_transid) {
615                 *async_transid = trans->transid;
616                 err = btrfs_commit_transaction_async(trans, root, 1);
617                 if (err)
618                         err = btrfs_commit_transaction(trans, root);
619         } else {
620                 err = btrfs_commit_transaction(trans, root);
621         }
622         if (err && !ret)
623                 ret = err;
624 
625         if (!ret) {
626                 inode = btrfs_lookup_dentry(dir, dentry);
627                 if (IS_ERR(inode))
628                         return PTR_ERR(inode);
629                 d_instantiate(dentry, inode);
630         }
631         return ret;
632 }
633 
634 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
635 {
636         s64 writers;
637         DEFINE_WAIT(wait);
638 
639         do {
640                 prepare_to_wait(&root->subv_writers->wait, &wait,
641                                 TASK_UNINTERRUPTIBLE);
642 
643                 writers = percpu_counter_sum(&root->subv_writers->counter);
644                 if (writers)
645                         schedule();
646 
647                 finish_wait(&root->subv_writers->wait, &wait);
648         } while (writers);
649 }
650 
651 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
652                            struct dentry *dentry, char *name, int namelen,
653                            u64 *async_transid, bool readonly,
654                            struct btrfs_qgroup_inherit *inherit)
655 {
656         struct inode *inode;
657         struct btrfs_pending_snapshot *pending_snapshot;
658         struct btrfs_trans_handle *trans;
659         int ret;
660 
661         if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
662                 return -EINVAL;
663 
664         pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
665         if (!pending_snapshot)
666                 return -ENOMEM;
667 
668         pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
669                         GFP_NOFS);
670         pending_snapshot->path = btrfs_alloc_path();
671         if (!pending_snapshot->root_item || !pending_snapshot->path) {
672                 ret = -ENOMEM;
673                 goto free_pending;
674         }
675 
676         atomic_inc(&root->will_be_snapshoted);
677         smp_mb__after_atomic();
678         btrfs_wait_for_no_snapshoting_writes(root);
679 
680         ret = btrfs_start_delalloc_inodes(root, 0);
681         if (ret)
682                 goto dec_and_free;
683 
684         btrfs_wait_ordered_extents(root, -1);
685 
686         btrfs_init_block_rsv(&pending_snapshot->block_rsv,
687                              BTRFS_BLOCK_RSV_TEMP);
688         /*
689          * 1 - parent dir inode
690          * 2 - dir entries
691          * 1 - root item
692          * 2 - root ref/backref
693          * 1 - root of snapshot
694          * 1 - UUID item
695          */
696         ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
697                                         &pending_snapshot->block_rsv, 8,
698                                         &pending_snapshot->qgroup_reserved,
699                                         false);
700         if (ret)
701                 goto dec_and_free;
702 
703         pending_snapshot->dentry = dentry;
704         pending_snapshot->root = root;
705         pending_snapshot->readonly = readonly;
706         pending_snapshot->dir = dir;
707         pending_snapshot->inherit = inherit;
708 
709         trans = btrfs_start_transaction(root, 0);
710         if (IS_ERR(trans)) {
711                 ret = PTR_ERR(trans);
712                 goto fail;
713         }
714 
715         spin_lock(&root->fs_info->trans_lock);
716         list_add(&pending_snapshot->list,
717                  &trans->transaction->pending_snapshots);
718         spin_unlock(&root->fs_info->trans_lock);
719         if (async_transid) {
720                 *async_transid = trans->transid;
721                 ret = btrfs_commit_transaction_async(trans,
722                                      root->fs_info->extent_root, 1);
723                 if (ret)
724                         ret = btrfs_commit_transaction(trans, root);
725         } else {
726                 ret = btrfs_commit_transaction(trans,
727                                                root->fs_info->extent_root);
728         }
729         if (ret)
730                 goto fail;
731 
732         ret = pending_snapshot->error;
733         if (ret)
734                 goto fail;
735 
736         ret = btrfs_orphan_cleanup(pending_snapshot->snap);
737         if (ret)
738                 goto fail;
739 
740         inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
741         if (IS_ERR(inode)) {
742                 ret = PTR_ERR(inode);
743                 goto fail;
744         }
745 
746         d_instantiate(dentry, inode);
747         ret = 0;
748 fail:
749         btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
750                                          &pending_snapshot->block_rsv,
751                                          pending_snapshot->qgroup_reserved);
752 dec_and_free:
753         if (atomic_dec_and_test(&root->will_be_snapshoted))
754                 wake_up_atomic_t(&root->will_be_snapshoted);
755 free_pending:
756         kfree(pending_snapshot->root_item);
757         btrfs_free_path(pending_snapshot->path);
758         kfree(pending_snapshot);
759 
760         return ret;
761 }
762 
763 /*  copy of may_delete in fs/namei.c()
764  *      Check whether we can remove a link victim from directory dir, check
765  *  whether the type of victim is right.
766  *  1. We can't do it if dir is read-only (done in permission())
767  *  2. We should have write and exec permissions on dir
768  *  3. We can't remove anything from append-only dir
769  *  4. We can't do anything with immutable dir (done in permission())
770  *  5. If the sticky bit on dir is set we should either
771  *      a. be owner of dir, or
772  *      b. be owner of victim, or
773  *      c. have CAP_FOWNER capability
774  *  6. If the victim is append-only or immutable we can't do antyhing with
775  *     links pointing to it.
776  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
777  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
778  *  9. We can't remove a root or mountpoint.
779  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
780  *     nfs_async_unlink().
781  */
782 
783 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
784 {
785         int error;
786 
787         if (d_really_is_negative(victim))
788                 return -ENOENT;
789 
790         BUG_ON(d_inode(victim->d_parent) != dir);
791         audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
792 
793         error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
794         if (error)
795                 return error;
796         if (IS_APPEND(dir))
797                 return -EPERM;
798         if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
799             IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
800                 return -EPERM;
801         if (isdir) {
802                 if (!d_is_dir(victim))
803                         return -ENOTDIR;
804                 if (IS_ROOT(victim))
805                         return -EBUSY;
806         } else if (d_is_dir(victim))
807                 return -EISDIR;
808         if (IS_DEADDIR(dir))
809                 return -ENOENT;
810         if (victim->d_flags & DCACHE_NFSFS_RENAMED)
811                 return -EBUSY;
812         return 0;
813 }
814 
815 /* copy of may_create in fs/namei.c() */
816 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
817 {
818         if (d_really_is_positive(child))
819                 return -EEXIST;
820         if (IS_DEADDIR(dir))
821                 return -ENOENT;
822         return inode_permission(dir, MAY_WRITE | MAY_EXEC);
823 }
824 
825 /*
826  * Create a new subvolume below @parent.  This is largely modeled after
827  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
828  * inside this filesystem so it's quite a bit simpler.
829  */
830 static noinline int btrfs_mksubvol(struct path *parent,
831                                    char *name, int namelen,
832                                    struct btrfs_root *snap_src,
833                                    u64 *async_transid, bool readonly,
834                                    struct btrfs_qgroup_inherit *inherit)
835 {
836         struct inode *dir  = d_inode(parent->dentry);
837         struct dentry *dentry;
838         int error;
839 
840         error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
841         if (error == -EINTR)
842                 return error;
843 
844         dentry = lookup_one_len(name, parent->dentry, namelen);
845         error = PTR_ERR(dentry);
846         if (IS_ERR(dentry))
847                 goto out_unlock;
848 
849         error = btrfs_may_create(dir, dentry);
850         if (error)
851                 goto out_dput;
852 
853         /*
854          * even if this name doesn't exist, we may get hash collisions.
855          * check for them now when we can safely fail
856          */
857         error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
858                                                dir->i_ino, name,
859                                                namelen);
860         if (error)
861                 goto out_dput;
862 
863         down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
864 
865         if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
866                 goto out_up_read;
867 
868         if (snap_src) {
869                 error = create_snapshot(snap_src, dir, dentry, name, namelen,
870                                         async_transid, readonly, inherit);
871         } else {
872                 error = create_subvol(dir, dentry, name, namelen,
873                                       async_transid, inherit);
874         }
875         if (!error)
876                 fsnotify_mkdir(dir, dentry);
877 out_up_read:
878         up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
879 out_dput:
880         dput(dentry);
881 out_unlock:
882         inode_unlock(dir);
883         return error;
884 }
885 
886 /*
887  * When we're defragging a range, we don't want to kick it off again
888  * if it is really just waiting for delalloc to send it down.
889  * If we find a nice big extent or delalloc range for the bytes in the
890  * file you want to defrag, we return 0 to let you know to skip this
891  * part of the file
892  */
893 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
894 {
895         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
896         struct extent_map *em = NULL;
897         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
898         u64 end;
899 
900         read_lock(&em_tree->lock);
901         em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
902         read_unlock(&em_tree->lock);
903 
904         if (em) {
905                 end = extent_map_end(em);
906                 free_extent_map(em);
907                 if (end - offset > thresh)
908                         return 0;
909         }
910         /* if we already have a nice delalloc here, just stop */
911         thresh /= 2;
912         end = count_range_bits(io_tree, &offset, offset + thresh,
913                                thresh, EXTENT_DELALLOC, 1);
914         if (end >= thresh)
915                 return 0;
916         return 1;
917 }
918 
919 /*
920  * helper function to walk through a file and find extents
921  * newer than a specific transid, and smaller than thresh.
922  *
923  * This is used by the defragging code to find new and small
924  * extents
925  */
926 static int find_new_extents(struct btrfs_root *root,
927                             struct inode *inode, u64 newer_than,
928                             u64 *off, u32 thresh)
929 {
930         struct btrfs_path *path;
931         struct btrfs_key min_key;
932         struct extent_buffer *leaf;
933         struct btrfs_file_extent_item *extent;
934         int type;
935         int ret;
936         u64 ino = btrfs_ino(inode);
937 
938         path = btrfs_alloc_path();
939         if (!path)
940                 return -ENOMEM;
941 
942         min_key.objectid = ino;
943         min_key.type = BTRFS_EXTENT_DATA_KEY;
944         min_key.offset = *off;
945 
946         while (1) {
947                 ret = btrfs_search_forward(root, &min_key, path, newer_than);
948                 if (ret != 0)
949                         goto none;
950 process_slot:
951                 if (min_key.objectid != ino)
952                         goto none;
953                 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
954                         goto none;
955 
956                 leaf = path->nodes[0];
957                 extent = btrfs_item_ptr(leaf, path->slots[0],
958                                         struct btrfs_file_extent_item);
959 
960                 type = btrfs_file_extent_type(leaf, extent);
961                 if (type == BTRFS_FILE_EXTENT_REG &&
962                     btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
963                     check_defrag_in_cache(inode, min_key.offset, thresh)) {
964                         *off = min_key.offset;
965                         btrfs_free_path(path);
966                         return 0;
967                 }
968 
969                 path->slots[0]++;
970                 if (path->slots[0] < btrfs_header_nritems(leaf)) {
971                         btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
972                         goto process_slot;
973                 }
974 
975                 if (min_key.offset == (u64)-1)
976                         goto none;
977 
978                 min_key.offset++;
979                 btrfs_release_path(path);
980         }
981 none:
982         btrfs_free_path(path);
983         return -ENOENT;
984 }
985 
986 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
987 {
988         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
989         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
990         struct extent_map *em;
991         u64 len = PAGE_SIZE;
992 
993         /*
994          * hopefully we have this extent in the tree already, try without
995          * the full extent lock
996          */
997         read_lock(&em_tree->lock);
998         em = lookup_extent_mapping(em_tree, start, len);
999         read_unlock(&em_tree->lock);
1000 
1001         if (!em) {
1002                 struct extent_state *cached = NULL;
1003                 u64 end = start + len - 1;
1004 
1005                 /* get the big lock and read metadata off disk */
1006                 lock_extent_bits(io_tree, start, end, &cached);
1007                 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1008                 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1009 
1010                 if (IS_ERR(em))
1011                         return NULL;
1012         }
1013 
1014         return em;
1015 }
1016 
1017 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1018 {
1019         struct extent_map *next;
1020         bool ret = true;
1021 
1022         /* this is the last extent */
1023         if (em->start + em->len >= i_size_read(inode))
1024                 return false;
1025 
1026         next = defrag_lookup_extent(inode, em->start + em->len);
1027         if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1028                 ret = false;
1029         else if ((em->block_start + em->block_len == next->block_start) &&
1030                  (em->block_len > SZ_128K && next->block_len > SZ_128K))
1031                 ret = false;
1032 
1033         free_extent_map(next);
1034         return ret;
1035 }
1036 
1037 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1038                                u64 *last_len, u64 *skip, u64 *defrag_end,
1039                                int compress)
1040 {
1041         struct extent_map *em;
1042         int ret = 1;
1043         bool next_mergeable = true;
1044         bool prev_mergeable = true;
1045 
1046         /*
1047          * make sure that once we start defragging an extent, we keep on
1048          * defragging it
1049          */
1050         if (start < *defrag_end)
1051                 return 1;
1052 
1053         *skip = 0;
1054 
1055         em = defrag_lookup_extent(inode, start);
1056         if (!em)
1057                 return 0;
1058 
1059         /* this will cover holes, and inline extents */
1060         if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1061                 ret = 0;
1062                 goto out;
1063         }
1064 
1065         if (!*defrag_end)
1066                 prev_mergeable = false;
1067 
1068         next_mergeable = defrag_check_next_extent(inode, em);
1069         /*
1070          * we hit a real extent, if it is big or the next extent is not a
1071          * real extent, don't bother defragging it
1072          */
1073         if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1074             (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1075                 ret = 0;
1076 out:
1077         /*
1078          * last_len ends up being a counter of how many bytes we've defragged.
1079          * every time we choose not to defrag an extent, we reset *last_len
1080          * so that the next tiny extent will force a defrag.
1081          *
1082          * The end result of this is that tiny extents before a single big
1083          * extent will force at least part of that big extent to be defragged.
1084          */
1085         if (ret) {
1086                 *defrag_end = extent_map_end(em);
1087         } else {
1088                 *last_len = 0;
1089                 *skip = extent_map_end(em);
1090                 *defrag_end = 0;
1091         }
1092 
1093         free_extent_map(em);
1094         return ret;
1095 }
1096 
1097 /*
1098  * it doesn't do much good to defrag one or two pages
1099  * at a time.  This pulls in a nice chunk of pages
1100  * to COW and defrag.
1101  *
1102  * It also makes sure the delalloc code has enough
1103  * dirty data to avoid making new small extents as part
1104  * of the defrag
1105  *
1106  * It's a good idea to start RA on this range
1107  * before calling this.
1108  */
1109 static int cluster_pages_for_defrag(struct inode *inode,
1110                                     struct page **pages,
1111                                     unsigned long start_index,
1112                                     unsigned long num_pages)
1113 {
1114         unsigned long file_end;
1115         u64 isize = i_size_read(inode);
1116         u64 page_start;
1117         u64 page_end;
1118         u64 page_cnt;
1119         int ret;
1120         int i;
1121         int i_done;
1122         struct btrfs_ordered_extent *ordered;
1123         struct extent_state *cached_state = NULL;
1124         struct extent_io_tree *tree;
1125         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1126 
1127         file_end = (isize - 1) >> PAGE_SHIFT;
1128         if (!isize || start_index > file_end)
1129                 return 0;
1130 
1131         page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1132 
1133         ret = btrfs_delalloc_reserve_space(inode,
1134                         start_index << PAGE_SHIFT,
1135                         page_cnt << PAGE_SHIFT);
1136         if (ret)
1137                 return ret;
1138         i_done = 0;
1139         tree = &BTRFS_I(inode)->io_tree;
1140 
1141         /* step one, lock all the pages */
1142         for (i = 0; i < page_cnt; i++) {
1143                 struct page *page;
1144 again:
1145                 page = find_or_create_page(inode->i_mapping,
1146                                            start_index + i, mask);
1147                 if (!page)
1148                         break;
1149 
1150                 page_start = page_offset(page);
1151                 page_end = page_start + PAGE_SIZE - 1;
1152                 while (1) {
1153                         lock_extent_bits(tree, page_start, page_end,
1154                                          &cached_state);
1155                         ordered = btrfs_lookup_ordered_extent(inode,
1156                                                               page_start);
1157                         unlock_extent_cached(tree, page_start, page_end,
1158                                              &cached_state, GFP_NOFS);
1159                         if (!ordered)
1160                                 break;
1161 
1162                         unlock_page(page);
1163                         btrfs_start_ordered_extent(inode, ordered, 1);
1164                         btrfs_put_ordered_extent(ordered);
1165                         lock_page(page);
1166                         /*
1167                          * we unlocked the page above, so we need check if
1168                          * it was released or not.
1169                          */
1170                         if (page->mapping != inode->i_mapping) {
1171                                 unlock_page(page);
1172                                 put_page(page);
1173                                 goto again;
1174                         }
1175                 }
1176 
1177                 if (!PageUptodate(page)) {
1178                         btrfs_readpage(NULL, page);
1179                         lock_page(page);
1180                         if (!PageUptodate(page)) {
1181                                 unlock_page(page);
1182                                 put_page(page);
1183                                 ret = -EIO;
1184                                 break;
1185                         }
1186                 }
1187 
1188                 if (page->mapping != inode->i_mapping) {
1189                         unlock_page(page);
1190                         put_page(page);
1191                         goto again;
1192                 }
1193 
1194                 pages[i] = page;
1195                 i_done++;
1196         }
1197         if (!i_done || ret)
1198                 goto out;
1199 
1200         if (!(inode->i_sb->s_flags & MS_ACTIVE))
1201                 goto out;
1202 
1203         /*
1204          * so now we have a nice long stream of locked
1205          * and up to date pages, lets wait on them
1206          */
1207         for (i = 0; i < i_done; i++)
1208                 wait_on_page_writeback(pages[i]);
1209 
1210         page_start = page_offset(pages[0]);
1211         page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1212 
1213         lock_extent_bits(&BTRFS_I(inode)->io_tree,
1214                          page_start, page_end - 1, &cached_state);
1215         clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1216                           page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1217                           EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1218                           &cached_state, GFP_NOFS);
1219 
1220         if (i_done != page_cnt) {
1221                 spin_lock(&BTRFS_I(inode)->lock);
1222                 BTRFS_I(inode)->outstanding_extents++;
1223                 spin_unlock(&BTRFS_I(inode)->lock);
1224                 btrfs_delalloc_release_space(inode,
1225                                 start_index << PAGE_SHIFT,
1226                                 (page_cnt - i_done) << PAGE_SHIFT);
1227         }
1228 
1229 
1230         set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1231                           &cached_state, GFP_NOFS);
1232 
1233         unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1234                              page_start, page_end - 1, &cached_state,
1235                              GFP_NOFS);
1236 
1237         for (i = 0; i < i_done; i++) {
1238                 clear_page_dirty_for_io(pages[i]);
1239                 ClearPageChecked(pages[i]);
1240                 set_page_extent_mapped(pages[i]);
1241                 set_page_dirty(pages[i]);
1242                 unlock_page(pages[i]);
1243                 put_page(pages[i]);
1244         }
1245         return i_done;
1246 out:
1247         for (i = 0; i < i_done; i++) {
1248                 unlock_page(pages[i]);
1249                 put_page(pages[i]);
1250         }
1251         btrfs_delalloc_release_space(inode,
1252                         start_index << PAGE_SHIFT,
1253                         page_cnt << PAGE_SHIFT);
1254         return ret;
1255 
1256 }
1257 
1258 int btrfs_defrag_file(struct inode *inode, struct file *file,
1259                       struct btrfs_ioctl_defrag_range_args *range,
1260                       u64 newer_than, unsigned long max_to_defrag)
1261 {
1262         struct btrfs_root *root = BTRFS_I(inode)->root;
1263         struct file_ra_state *ra = NULL;
1264         unsigned long last_index;
1265         u64 isize = i_size_read(inode);
1266         u64 last_len = 0;
1267         u64 skip = 0;
1268         u64 defrag_end = 0;
1269         u64 newer_off = range->start;
1270         unsigned long i;
1271         unsigned long ra_index = 0;
1272         int ret;
1273         int defrag_count = 0;
1274         int compress_type = BTRFS_COMPRESS_ZLIB;
1275         u32 extent_thresh = range->extent_thresh;
1276         unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1277         unsigned long cluster = max_cluster;
1278         u64 new_align = ~((u64)SZ_128K - 1);
1279         struct page **pages = NULL;
1280 
1281         if (isize == 0)
1282                 return 0;
1283 
1284         if (range->start >= isize)
1285                 return -EINVAL;
1286 
1287         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1288                 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1289                         return -EINVAL;
1290                 if (range->compress_type)
1291                         compress_type = range->compress_type;
1292         }
1293 
1294         if (extent_thresh == 0)
1295                 extent_thresh = SZ_256K;
1296 
1297         /*
1298          * if we were not given a file, allocate a readahead
1299          * context
1300          */
1301         if (!file) {
1302                 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1303                 if (!ra)
1304                         return -ENOMEM;
1305                 file_ra_state_init(ra, inode->i_mapping);
1306         } else {
1307                 ra = &file->f_ra;
1308         }
1309 
1310         pages = kmalloc_array(max_cluster, sizeof(struct page *),
1311                         GFP_NOFS);
1312         if (!pages) {
1313                 ret = -ENOMEM;
1314                 goto out_ra;
1315         }
1316 
1317         /* find the last page to defrag */
1318         if (range->start + range->len > range->start) {
1319                 last_index = min_t(u64, isize - 1,
1320                          range->start + range->len - 1) >> PAGE_SHIFT;
1321         } else {
1322                 last_index = (isize - 1) >> PAGE_SHIFT;
1323         }
1324 
1325         if (newer_than) {
1326                 ret = find_new_extents(root, inode, newer_than,
1327                                        &newer_off, SZ_64K);
1328                 if (!ret) {
1329                         range->start = newer_off;
1330                         /*
1331                          * we always align our defrag to help keep
1332                          * the extents in the file evenly spaced
1333                          */
1334                         i = (newer_off & new_align) >> PAGE_SHIFT;
1335                 } else
1336                         goto out_ra;
1337         } else {
1338                 i = range->start >> PAGE_SHIFT;
1339         }
1340         if (!max_to_defrag)
1341                 max_to_defrag = last_index - i + 1;
1342 
1343         /*
1344          * make writeback starts from i, so the defrag range can be
1345          * written sequentially.
1346          */
1347         if (i < inode->i_mapping->writeback_index)
1348                 inode->i_mapping->writeback_index = i;
1349 
1350         while (i <= last_index && defrag_count < max_to_defrag &&
1351                (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1352                 /*
1353                  * make sure we stop running if someone unmounts
1354                  * the FS
1355                  */
1356                 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1357                         break;
1358 
1359                 if (btrfs_defrag_cancelled(root->fs_info)) {
1360                         btrfs_debug(root->fs_info, "defrag_file cancelled");
1361                         ret = -EAGAIN;
1362                         break;
1363                 }
1364 
1365                 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1366                                          extent_thresh, &last_len, &skip,
1367                                          &defrag_end, range->flags &
1368                                          BTRFS_DEFRAG_RANGE_COMPRESS)) {
1369                         unsigned long next;
1370                         /*
1371                          * the should_defrag function tells us how much to skip
1372                          * bump our counter by the suggested amount
1373                          */
1374                         next = DIV_ROUND_UP(skip, PAGE_SIZE);
1375                         i = max(i + 1, next);
1376                         continue;
1377                 }
1378 
1379                 if (!newer_than) {
1380                         cluster = (PAGE_ALIGN(defrag_end) >>
1381                                    PAGE_SHIFT) - i;
1382                         cluster = min(cluster, max_cluster);
1383                 } else {
1384                         cluster = max_cluster;
1385                 }
1386 
1387                 if (i + cluster > ra_index) {
1388                         ra_index = max(i, ra_index);
1389                         btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1390                                        cluster);
1391                         ra_index += cluster;
1392                 }
1393 
1394                 inode_lock(inode);
1395                 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1396                         BTRFS_I(inode)->force_compress = compress_type;
1397                 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1398                 if (ret < 0) {
1399                         inode_unlock(inode);
1400                         goto out_ra;
1401                 }
1402 
1403                 defrag_count += ret;
1404                 balance_dirty_pages_ratelimited(inode->i_mapping);
1405                 inode_unlock(inode);
1406 
1407                 if (newer_than) {
1408                         if (newer_off == (u64)-1)
1409                                 break;
1410 
1411                         if (ret > 0)
1412                                 i += ret;
1413 
1414                         newer_off = max(newer_off + 1,
1415                                         (u64)i << PAGE_SHIFT);
1416 
1417                         ret = find_new_extents(root, inode, newer_than,
1418                                                &newer_off, SZ_64K);
1419                         if (!ret) {
1420                                 range->start = newer_off;
1421                                 i = (newer_off & new_align) >> PAGE_SHIFT;
1422                         } else {
1423                                 break;
1424                         }
1425                 } else {
1426                         if (ret > 0) {
1427                                 i += ret;
1428                                 last_len += ret << PAGE_SHIFT;
1429                         } else {
1430                                 i++;
1431                                 last_len = 0;
1432                         }
1433                 }
1434         }
1435 
1436         if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1437                 filemap_flush(inode->i_mapping);
1438                 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1439                              &BTRFS_I(inode)->runtime_flags))
1440                         filemap_flush(inode->i_mapping);
1441         }
1442 
1443         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1444                 /* the filemap_flush will queue IO into the worker threads, but
1445                  * we have to make sure the IO is actually started and that
1446                  * ordered extents get created before we return
1447                  */
1448                 atomic_inc(&root->fs_info->async_submit_draining);
1449                 while (atomic_read(&root->fs_info->nr_async_submits) ||
1450                       atomic_read(&root->fs_info->async_delalloc_pages)) {
1451                         wait_event(root->fs_info->async_submit_wait,
1452                            (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1453                             atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1454                 }
1455                 atomic_dec(&root->fs_info->async_submit_draining);
1456         }
1457 
1458         if (range->compress_type == BTRFS_COMPRESS_LZO) {
1459                 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1460         }
1461 
1462         ret = defrag_count;
1463 
1464 out_ra:
1465         if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1466                 inode_lock(inode);
1467                 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1468                 inode_unlock(inode);
1469         }
1470         if (!file)
1471                 kfree(ra);
1472         kfree(pages);
1473         return ret;
1474 }
1475 
1476 static noinline int btrfs_ioctl_resize(struct file *file,
1477                                         void __user *arg)
1478 {
1479         u64 new_size;
1480         u64 old_size;
1481         u64 devid = 1;
1482         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1483         struct btrfs_ioctl_vol_args *vol_args;
1484         struct btrfs_trans_handle *trans;
1485         struct btrfs_device *device = NULL;
1486         char *sizestr;
1487         char *retptr;
1488         char *devstr = NULL;
1489         int ret = 0;
1490         int mod = 0;
1491 
1492         if (!capable(CAP_SYS_ADMIN))
1493                 return -EPERM;
1494 
1495         ret = mnt_want_write_file(file);
1496         if (ret)
1497                 return ret;
1498 
1499         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1500                         1)) {
1501                 mnt_drop_write_file(file);
1502                 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1503         }
1504 
1505         mutex_lock(&root->fs_info->volume_mutex);
1506         vol_args = memdup_user(arg, sizeof(*vol_args));
1507         if (IS_ERR(vol_args)) {
1508                 ret = PTR_ERR(vol_args);
1509                 goto out;
1510         }
1511 
1512         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1513 
1514         sizestr = vol_args->name;
1515         devstr = strchr(sizestr, ':');
1516         if (devstr) {
1517                 sizestr = devstr + 1;
1518                 *devstr = '\0';
1519                 devstr = vol_args->name;
1520                 ret = kstrtoull(devstr, 10, &devid);
1521                 if (ret)
1522                         goto out_free;
1523                 if (!devid) {
1524                         ret = -EINVAL;
1525                         goto out_free;
1526                 }
1527                 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1528         }
1529 
1530         device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1531         if (!device) {
1532                 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1533                        devid);
1534                 ret = -ENODEV;
1535                 goto out_free;
1536         }
1537 
1538         if (!device->writeable) {
1539                 btrfs_info(root->fs_info,
1540                            "resizer unable to apply on readonly device %llu",
1541                        devid);
1542                 ret = -EPERM;
1543                 goto out_free;
1544         }
1545 
1546         if (!strcmp(sizestr, "max"))
1547                 new_size = device->bdev->bd_inode->i_size;
1548         else {
1549                 if (sizestr[0] == '-') {
1550                         mod = -1;
1551                         sizestr++;
1552                 } else if (sizestr[0] == '+') {
1553                         mod = 1;
1554                         sizestr++;
1555                 }
1556                 new_size = memparse(sizestr, &retptr);
1557                 if (*retptr != '\0' || new_size == 0) {
1558                         ret = -EINVAL;
1559                         goto out_free;
1560                 }
1561         }
1562 
1563         if (device->is_tgtdev_for_dev_replace) {
1564                 ret = -EPERM;
1565                 goto out_free;
1566         }
1567 
1568         old_size = btrfs_device_get_total_bytes(device);
1569 
1570         if (mod < 0) {
1571                 if (new_size > old_size) {
1572                         ret = -EINVAL;
1573                         goto out_free;
1574                 }
1575                 new_size = old_size - new_size;
1576         } else if (mod > 0) {
1577                 if (new_size > ULLONG_MAX - old_size) {
1578                         ret = -ERANGE;
1579                         goto out_free;
1580                 }
1581                 new_size = old_size + new_size;
1582         }
1583 
1584         if (new_size < SZ_256M) {
1585                 ret = -EINVAL;
1586                 goto out_free;
1587         }
1588         if (new_size > device->bdev->bd_inode->i_size) {
1589                 ret = -EFBIG;
1590                 goto out_free;
1591         }
1592 
1593         new_size = div_u64(new_size, root->sectorsize);
1594         new_size *= root->sectorsize;
1595 
1596         btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu",
1597                       rcu_str_deref(device->name), new_size);
1598 
1599         if (new_size > old_size) {
1600                 trans = btrfs_start_transaction(root, 0);
1601                 if (IS_ERR(trans)) {
1602                         ret = PTR_ERR(trans);
1603                         goto out_free;
1604                 }
1605                 ret = btrfs_grow_device(trans, device, new_size);
1606                 btrfs_commit_transaction(trans, root);
1607         } else if (new_size < old_size) {
1608                 ret = btrfs_shrink_device(device, new_size);
1609         } /* equal, nothing need to do */
1610 
1611 out_free:
1612         kfree(vol_args);
1613 out:
1614         mutex_unlock(&root->fs_info->volume_mutex);
1615         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1616         mnt_drop_write_file(file);
1617         return ret;
1618 }
1619 
1620 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1621                                 char *name, unsigned long fd, int subvol,
1622                                 u64 *transid, bool readonly,
1623                                 struct btrfs_qgroup_inherit *inherit)
1624 {
1625         int namelen;
1626         int ret = 0;
1627 
1628         ret = mnt_want_write_file(file);
1629         if (ret)
1630                 goto out;
1631 
1632         namelen = strlen(name);
1633         if (strchr(name, '/')) {
1634                 ret = -EINVAL;
1635                 goto out_drop_write;
1636         }
1637 
1638         if (name[0] == '.' &&
1639            (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1640                 ret = -EEXIST;
1641                 goto out_drop_write;
1642         }
1643 
1644         if (subvol) {
1645                 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1646                                      NULL, transid, readonly, inherit);
1647         } else {
1648                 struct fd src = fdget(fd);
1649                 struct inode *src_inode;
1650                 if (!src.file) {
1651                         ret = -EINVAL;
1652                         goto out_drop_write;
1653                 }
1654 
1655                 src_inode = file_inode(src.file);
1656                 if (src_inode->i_sb != file_inode(file)->i_sb) {
1657                         btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1658                                    "Snapshot src from another FS");
1659                         ret = -EXDEV;
1660                 } else if (!inode_owner_or_capable(src_inode)) {
1661                         /*
1662                          * Subvolume creation is not restricted, but snapshots
1663                          * are limited to own subvolumes only
1664                          */
1665                         ret = -EPERM;
1666                 } else {
1667                         ret = btrfs_mksubvol(&file->f_path, name, namelen,
1668                                              BTRFS_I(src_inode)->root,
1669                                              transid, readonly, inherit);
1670                 }
1671                 fdput(src);
1672         }
1673 out_drop_write:
1674         mnt_drop_write_file(file);
1675 out:
1676         return ret;
1677 }
1678 
1679 static noinline int btrfs_ioctl_snap_create(struct file *file,
1680                                             void __user *arg, int subvol)
1681 {
1682         struct btrfs_ioctl_vol_args *vol_args;
1683         int ret;
1684 
1685         vol_args = memdup_user(arg, sizeof(*vol_args));
1686         if (IS_ERR(vol_args))
1687                 return PTR_ERR(vol_args);
1688         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1689 
1690         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1691                                               vol_args->fd, subvol,
1692                                               NULL, false, NULL);
1693 
1694         kfree(vol_args);
1695         return ret;
1696 }
1697 
1698 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1699                                                void __user *arg, int subvol)
1700 {
1701         struct btrfs_ioctl_vol_args_v2 *vol_args;
1702         int ret;
1703         u64 transid = 0;
1704         u64 *ptr = NULL;
1705         bool readonly = false;
1706         struct btrfs_qgroup_inherit *inherit = NULL;
1707 
1708         vol_args = memdup_user(arg, sizeof(*vol_args));
1709         if (IS_ERR(vol_args))
1710                 return PTR_ERR(vol_args);
1711         vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1712 
1713         if (vol_args->flags &
1714             ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1715               BTRFS_SUBVOL_QGROUP_INHERIT)) {
1716                 ret = -EOPNOTSUPP;
1717                 goto free_args;
1718         }
1719 
1720         if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1721                 ptr = &transid;
1722         if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1723                 readonly = true;
1724         if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1725                 if (vol_args->size > PAGE_SIZE) {
1726                         ret = -EINVAL;
1727                         goto free_args;
1728                 }
1729                 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1730                 if (IS_ERR(inherit)) {
1731                         ret = PTR_ERR(inherit);
1732                         goto free_args;
1733                 }
1734         }
1735 
1736         ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1737                                               vol_args->fd, subvol, ptr,
1738                                               readonly, inherit);
1739         if (ret)
1740                 goto free_inherit;
1741 
1742         if (ptr && copy_to_user(arg +
1743                                 offsetof(struct btrfs_ioctl_vol_args_v2,
1744                                         transid),
1745                                 ptr, sizeof(*ptr)))
1746                 ret = -EFAULT;
1747 
1748 free_inherit:
1749         kfree(inherit);
1750 free_args:
1751         kfree(vol_args);
1752         return ret;
1753 }
1754 
1755 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1756                                                 void __user *arg)
1757 {
1758         struct inode *inode = file_inode(file);
1759         struct btrfs_root *root = BTRFS_I(inode)->root;
1760         int ret = 0;
1761         u64 flags = 0;
1762 
1763         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1764                 return -EINVAL;
1765 
1766         down_read(&root->fs_info->subvol_sem);
1767         if (btrfs_root_readonly(root))
1768                 flags |= BTRFS_SUBVOL_RDONLY;
1769         up_read(&root->fs_info->subvol_sem);
1770 
1771         if (copy_to_user(arg, &flags, sizeof(flags)))
1772                 ret = -EFAULT;
1773 
1774         return ret;
1775 }
1776 
1777 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1778                                               void __user *arg)
1779 {
1780         struct inode *inode = file_inode(file);
1781         struct btrfs_root *root = BTRFS_I(inode)->root;
1782         struct btrfs_trans_handle *trans;
1783         u64 root_flags;
1784         u64 flags;
1785         int ret = 0;
1786 
1787         if (!inode_owner_or_capable(inode))
1788                 return -EPERM;
1789 
1790         ret = mnt_want_write_file(file);
1791         if (ret)
1792                 goto out;
1793 
1794         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1795                 ret = -EINVAL;
1796                 goto out_drop_write;
1797         }
1798 
1799         if (copy_from_user(&flags, arg, sizeof(flags))) {
1800                 ret = -EFAULT;
1801                 goto out_drop_write;
1802         }
1803 
1804         if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1805                 ret = -EINVAL;
1806                 goto out_drop_write;
1807         }
1808 
1809         if (flags & ~BTRFS_SUBVOL_RDONLY) {
1810                 ret = -EOPNOTSUPP;
1811                 goto out_drop_write;
1812         }
1813 
1814         down_write(&root->fs_info->subvol_sem);
1815 
1816         /* nothing to do */
1817         if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1818                 goto out_drop_sem;
1819 
1820         root_flags = btrfs_root_flags(&root->root_item);
1821         if (flags & BTRFS_SUBVOL_RDONLY) {
1822                 btrfs_set_root_flags(&root->root_item,
1823                                      root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1824         } else {
1825                 /*
1826                  * Block RO -> RW transition if this subvolume is involved in
1827                  * send
1828                  */
1829                 spin_lock(&root->root_item_lock);
1830                 if (root->send_in_progress == 0) {
1831                         btrfs_set_root_flags(&root->root_item,
1832                                      root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1833                         spin_unlock(&root->root_item_lock);
1834                 } else {
1835                         spin_unlock(&root->root_item_lock);
1836                         btrfs_warn(root->fs_info,
1837                         "Attempt to set subvolume %llu read-write during send",
1838                                         root->root_key.objectid);
1839                         ret = -EPERM;
1840                         goto out_drop_sem;
1841                 }
1842         }
1843 
1844         trans = btrfs_start_transaction(root, 1);
1845         if (IS_ERR(trans)) {
1846                 ret = PTR_ERR(trans);
1847                 goto out_reset;
1848         }
1849 
1850         ret = btrfs_update_root(trans, root->fs_info->tree_root,
1851                                 &root->root_key, &root->root_item);
1852 
1853         btrfs_commit_transaction(trans, root);
1854 out_reset:
1855         if (ret)
1856                 btrfs_set_root_flags(&root->root_item, root_flags);
1857 out_drop_sem:
1858         up_write(&root->fs_info->subvol_sem);
1859 out_drop_write:
1860         mnt_drop_write_file(file);
1861 out:
1862         return ret;
1863 }
1864 
1865 /*
1866  * helper to check if the subvolume references other subvolumes
1867  */
1868 static noinline int may_destroy_subvol(struct btrfs_root *root)
1869 {
1870         struct btrfs_path *path;
1871         struct btrfs_dir_item *di;
1872         struct btrfs_key key;
1873         u64 dir_id;
1874         int ret;
1875 
1876         path = btrfs_alloc_path();
1877         if (!path)
1878                 return -ENOMEM;
1879 
1880         /* Make sure this root isn't set as the default subvol */
1881         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1882         di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1883                                    dir_id, "default", 7, 0);
1884         if (di && !IS_ERR(di)) {
1885                 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1886                 if (key.objectid == root->root_key.objectid) {
1887                         ret = -EPERM;
1888                         btrfs_err(root->fs_info, "deleting default subvolume "
1889                                   "%llu is not allowed", key.objectid);
1890                         goto out;
1891                 }
1892                 btrfs_release_path(path);
1893         }
1894 
1895         key.objectid = root->root_key.objectid;
1896         key.type = BTRFS_ROOT_REF_KEY;
1897         key.offset = (u64)-1;
1898 
1899         ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1900                                 &key, path, 0, 0);
1901         if (ret < 0)
1902                 goto out;
1903         BUG_ON(ret == 0);
1904 
1905         ret = 0;
1906         if (path->slots[0] > 0) {
1907                 path->slots[0]--;
1908                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1909                 if (key.objectid == root->root_key.objectid &&
1910                     key.type == BTRFS_ROOT_REF_KEY)
1911                         ret = -ENOTEMPTY;
1912         }
1913 out:
1914         btrfs_free_path(path);
1915         return ret;
1916 }
1917 
1918 static noinline int key_in_sk(struct btrfs_key *key,
1919                               struct btrfs_ioctl_search_key *sk)
1920 {
1921         struct btrfs_key test;
1922         int ret;
1923 
1924         test.objectid = sk->min_objectid;
1925         test.type = sk->min_type;
1926         test.offset = sk->min_offset;
1927 
1928         ret = btrfs_comp_cpu_keys(key, &test);
1929         if (ret < 0)
1930                 return 0;
1931 
1932         test.objectid = sk->max_objectid;
1933         test.type = sk->max_type;
1934         test.offset = sk->max_offset;
1935 
1936         ret = btrfs_comp_cpu_keys(key, &test);
1937         if (ret > 0)
1938                 return 0;
1939         return 1;
1940 }
1941 
1942 static noinline int copy_to_sk(struct btrfs_root *root,
1943                                struct btrfs_path *path,
1944                                struct btrfs_key *key,
1945                                struct btrfs_ioctl_search_key *sk,
1946                                size_t *buf_size,
1947                                char __user *ubuf,
1948                                unsigned long *sk_offset,
1949                                int *num_found)
1950 {
1951         u64 found_transid;
1952         struct extent_buffer *leaf;
1953         struct btrfs_ioctl_search_header sh;
1954         struct btrfs_key test;
1955         unsigned long item_off;
1956         unsigned long item_len;
1957         int nritems;
1958         int i;
1959         int slot;
1960         int ret = 0;
1961 
1962         leaf = path->nodes[0];
1963         slot = path->slots[0];
1964         nritems = btrfs_header_nritems(leaf);
1965 
1966         if (btrfs_header_generation(leaf) > sk->max_transid) {
1967                 i = nritems;
1968                 goto advance_key;
1969         }
1970         found_transid = btrfs_header_generation(leaf);
1971 
1972         for (i = slot; i < nritems; i++) {
1973                 item_off = btrfs_item_ptr_offset(leaf, i);
1974                 item_len = btrfs_item_size_nr(leaf, i);
1975 
1976                 btrfs_item_key_to_cpu(leaf, key, i);
1977                 if (!key_in_sk(key, sk))
1978                         continue;
1979 
1980                 if (sizeof(sh) + item_len > *buf_size) {
1981                         if (*num_found) {
1982                                 ret = 1;
1983                                 goto out;
1984                         }
1985 
1986                         /*
1987                          * return one empty item back for v1, which does not
1988                          * handle -EOVERFLOW
1989                          */
1990 
1991                         *buf_size = sizeof(sh) + item_len;
1992                         item_len = 0;
1993                         ret = -EOVERFLOW;
1994                 }
1995 
1996                 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1997                         ret = 1;
1998                         goto out;
1999                 }
2000 
2001                 sh.objectid = key->objectid;
2002                 sh.offset = key->offset;
2003                 sh.type = key->type;
2004                 sh.len = item_len;
2005                 sh.transid = found_transid;
2006 
2007                 /* copy search result header */
2008                 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2009                         ret = -EFAULT;
2010                         goto out;
2011                 }
2012 
2013                 *sk_offset += sizeof(sh);
2014 
2015                 if (item_len) {
2016                         char __user *up = ubuf + *sk_offset;
2017                         /* copy the item */
2018                         if (read_extent_buffer_to_user(leaf, up,
2019                                                        item_off, item_len)) {
2020                                 ret = -EFAULT;
2021                                 goto out;
2022                         }
2023 
2024                         *sk_offset += item_len;
2025                 }
2026                 (*num_found)++;
2027 
2028                 if (ret) /* -EOVERFLOW from above */
2029                         goto out;
2030 
2031                 if (*num_found >= sk->nr_items) {
2032                         ret = 1;
2033                         goto out;
2034                 }
2035         }
2036 advance_key:
2037         ret = 0;
2038         test.objectid = sk->max_objectid;
2039         test.type = sk->max_type;
2040         test.offset = sk->max_offset;
2041         if (btrfs_comp_cpu_keys(key, &test) >= 0)
2042                 ret = 1;
2043         else if (key->offset < (u64)-1)
2044                 key->offset++;
2045         else if (key->type < (u8)-1) {
2046                 key->offset = 0;
2047                 key->type++;
2048         } else if (key->objectid < (u64)-1) {
2049                 key->offset = 0;
2050                 key->type = 0;
2051                 key->objectid++;
2052         } else
2053                 ret = 1;
2054 out:
2055         /*
2056          *  0: all items from this leaf copied, continue with next
2057          *  1: * more items can be copied, but unused buffer is too small
2058          *     * all items were found
2059          *     Either way, it will stops the loop which iterates to the next
2060          *     leaf
2061          *  -EOVERFLOW: item was to large for buffer
2062          *  -EFAULT: could not copy extent buffer back to userspace
2063          */
2064         return ret;
2065 }
2066 
2067 static noinline int search_ioctl(struct inode *inode,
2068                                  struct btrfs_ioctl_search_key *sk,
2069                                  size_t *buf_size,
2070                                  char __user *ubuf)
2071 {
2072         struct btrfs_root *root;
2073         struct btrfs_key key;
2074         struct btrfs_path *path;
2075         struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2076         int ret;
2077         int num_found = 0;
2078         unsigned long sk_offset = 0;
2079 
2080         if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2081                 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2082                 return -EOVERFLOW;
2083         }
2084 
2085         path = btrfs_alloc_path();
2086         if (!path)
2087                 return -ENOMEM;
2088 
2089         if (sk->tree_id == 0) {
2090                 /* search the root of the inode that was passed */
2091                 root = BTRFS_I(inode)->root;
2092         } else {
2093                 key.objectid = sk->tree_id;
2094                 key.type = BTRFS_ROOT_ITEM_KEY;
2095                 key.offset = (u64)-1;
2096                 root = btrfs_read_fs_root_no_name(info, &key);
2097                 if (IS_ERR(root)) {
2098                         btrfs_free_path(path);
2099                         return -ENOENT;
2100                 }
2101         }
2102 
2103         key.objectid = sk->min_objectid;
2104         key.type = sk->min_type;
2105         key.offset = sk->min_offset;
2106 
2107         while (1) {
2108                 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2109                 if (ret != 0) {
2110                         if (ret > 0)
2111                                 ret = 0;
2112                         goto err;
2113                 }
2114                 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2115                                  &sk_offset, &num_found);
2116                 btrfs_release_path(path);
2117                 if (ret)
2118                         break;
2119 
2120         }
2121         if (ret > 0)
2122                 ret = 0;
2123 err:
2124         sk->nr_items = num_found;
2125         btrfs_free_path(path);
2126         return ret;
2127 }
2128 
2129 static noinline int btrfs_ioctl_tree_search(struct file *file,
2130                                            void __user *argp)
2131 {
2132         struct btrfs_ioctl_search_args __user *uargs;
2133         struct btrfs_ioctl_search_key sk;
2134         struct inode *inode;
2135         int ret;
2136         size_t buf_size;
2137 
2138         if (!capable(CAP_SYS_ADMIN))
2139                 return -EPERM;
2140 
2141         uargs = (struct btrfs_ioctl_search_args __user *)argp;
2142 
2143         if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2144                 return -EFAULT;
2145 
2146         buf_size = sizeof(uargs->buf);
2147 
2148         inode = file_inode(file);
2149         ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2150 
2151         /*
2152          * In the origin implementation an overflow is handled by returning a
2153          * search header with a len of zero, so reset ret.
2154          */
2155         if (ret == -EOVERFLOW)
2156                 ret = 0;
2157 
2158         if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2159                 ret = -EFAULT;
2160         return ret;
2161 }
2162 
2163 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2164                                                void __user *argp)
2165 {
2166         struct btrfs_ioctl_search_args_v2 __user *uarg;
2167         struct btrfs_ioctl_search_args_v2 args;
2168         struct inode *inode;
2169         int ret;
2170         size_t buf_size;
2171         const size_t buf_limit = SZ_16M;
2172 
2173         if (!capable(CAP_SYS_ADMIN))
2174                 return -EPERM;
2175 
2176         /* copy search header and buffer size */
2177         uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2178         if (copy_from_user(&args, uarg, sizeof(args)))
2179                 return -EFAULT;
2180 
2181         buf_size = args.buf_size;
2182 
2183         if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2184                 return -EOVERFLOW;
2185 
2186         /* limit result size to 16MB */
2187         if (buf_size > buf_limit)
2188                 buf_size = buf_limit;
2189 
2190         inode = file_inode(file);
2191         ret = search_ioctl(inode, &args.key, &buf_size,
2192                            (char *)(&uarg->buf[0]));
2193         if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2194                 ret = -EFAULT;
2195         else if (ret == -EOVERFLOW &&
2196                 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2197                 ret = -EFAULT;
2198 
2199         return ret;
2200 }
2201 
2202 /*
2203  * Search INODE_REFs to identify path name of 'dirid' directory
2204  * in a 'tree_id' tree. and sets path name to 'name'.
2205  */
2206 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2207                                 u64 tree_id, u64 dirid, char *name)
2208 {
2209         struct btrfs_root *root;
2210         struct btrfs_key key;
2211         char *ptr;
2212         int ret = -1;
2213         int slot;
2214         int len;
2215         int total_len = 0;
2216         struct btrfs_inode_ref *iref;
2217         struct extent_buffer *l;
2218         struct btrfs_path *path;
2219 
2220         if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2221                 name[0]='\0';
2222                 return 0;
2223         }
2224 
2225         path = btrfs_alloc_path();
2226         if (!path)
2227                 return -ENOMEM;
2228 
2229         ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2230 
2231         key.objectid = tree_id;
2232         key.type = BTRFS_ROOT_ITEM_KEY;
2233         key.offset = (u64)-1;
2234         root = btrfs_read_fs_root_no_name(info, &key);
2235         if (IS_ERR(root)) {
2236                 btrfs_err(info, "could not find root %llu", tree_id);
2237                 ret = -ENOENT;
2238                 goto out;
2239         }
2240 
2241         key.objectid = dirid;
2242         key.type = BTRFS_INODE_REF_KEY;
2243         key.offset = (u64)-1;
2244 
2245         while (1) {
2246                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2247                 if (ret < 0)
2248                         goto out;
2249                 else if (ret > 0) {
2250                         ret = btrfs_previous_item(root, path, dirid,
2251                                                   BTRFS_INODE_REF_KEY);
2252                         if (ret < 0)
2253                                 goto out;
2254                         else if (ret > 0) {
2255                                 ret = -ENOENT;
2256                                 goto out;
2257                         }
2258                 }
2259 
2260                 l = path->nodes[0];
2261                 slot = path->slots[0];
2262                 btrfs_item_key_to_cpu(l, &key, slot);
2263 
2264                 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2265                 len = btrfs_inode_ref_name_len(l, iref);
2266                 ptr -= len + 1;
2267                 total_len += len + 1;
2268                 if (ptr < name) {
2269                         ret = -ENAMETOOLONG;
2270                         goto out;
2271                 }
2272 
2273                 *(ptr + len) = '/';
2274                 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2275 
2276                 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2277                         break;
2278 
2279                 btrfs_release_path(path);
2280                 key.objectid = key.offset;
2281                 key.offset = (u64)-1;
2282                 dirid = key.objectid;
2283         }
2284         memmove(name, ptr, total_len);
2285         name[total_len] = '\0';
2286         ret = 0;
2287 out:
2288         btrfs_free_path(path);
2289         return ret;
2290 }
2291 
2292 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2293                                            void __user *argp)
2294 {
2295          struct btrfs_ioctl_ino_lookup_args *args;
2296          struct inode *inode;
2297         int ret = 0;
2298 
2299         args = memdup_user(argp, sizeof(*args));
2300         if (IS_ERR(args))
2301                 return PTR_ERR(args);
2302 
2303         inode = file_inode(file);
2304 
2305         /*
2306          * Unprivileged query to obtain the containing subvolume root id. The
2307          * path is reset so it's consistent with btrfs_search_path_in_tree.
2308          */
2309         if (args->treeid == 0)
2310                 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2311 
2312         if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2313                 args->name[0] = 0;
2314                 goto out;
2315         }
2316 
2317         if (!capable(CAP_SYS_ADMIN)) {
2318                 ret = -EPERM;
2319                 goto out;
2320         }
2321 
2322         ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2323                                         args->treeid, args->objectid,
2324                                         args->name);
2325 
2326 out:
2327         if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2328                 ret = -EFAULT;
2329 
2330         kfree(args);
2331         return ret;
2332 }
2333 
2334 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2335                                              void __user *arg)
2336 {
2337         struct dentry *parent = file->f_path.dentry;
2338         struct dentry *dentry;
2339         struct inode *dir = d_inode(parent);
2340         struct inode *inode;
2341         struct btrfs_root *root = BTRFS_I(dir)->root;
2342         struct btrfs_root *dest = NULL;
2343         struct btrfs_ioctl_vol_args *vol_args;
2344         struct btrfs_trans_handle *trans;
2345         struct btrfs_block_rsv block_rsv;
2346         u64 root_flags;
2347         u64 qgroup_reserved;
2348         int namelen;
2349         int ret;
2350         int err = 0;
2351 
2352         vol_args = memdup_user(arg, sizeof(*vol_args));
2353         if (IS_ERR(vol_args))
2354                 return PTR_ERR(vol_args);
2355 
2356         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2357         namelen = strlen(vol_args->name);
2358         if (strchr(vol_args->name, '/') ||
2359             strncmp(vol_args->name, "..", namelen) == 0) {
2360                 err = -EINVAL;
2361                 goto out;
2362         }
2363 
2364         err = mnt_want_write_file(file);
2365         if (err)
2366                 goto out;
2367 
2368 
2369         err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2370         if (err == -EINTR)
2371                 goto out_drop_write;
2372         dentry = lookup_one_len(vol_args->name, parent, namelen);
2373         if (IS_ERR(dentry)) {
2374                 err = PTR_ERR(dentry);
2375                 goto out_unlock_dir;
2376         }
2377 
2378         if (d_really_is_negative(dentry)) {
2379                 err = -ENOENT;
2380                 goto out_dput;
2381         }
2382 
2383         inode = d_inode(dentry);
2384         dest = BTRFS_I(inode)->root;
2385         if (!capable(CAP_SYS_ADMIN)) {
2386                 /*
2387                  * Regular user.  Only allow this with a special mount
2388                  * option, when the user has write+exec access to the
2389                  * subvol root, and when rmdir(2) would have been
2390                  * allowed.
2391                  *
2392                  * Note that this is _not_ check that the subvol is
2393                  * empty or doesn't contain data that we wouldn't
2394                  * otherwise be able to delete.
2395                  *
2396                  * Users who want to delete empty subvols should try
2397                  * rmdir(2).
2398                  */
2399                 err = -EPERM;
2400                 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2401                         goto out_dput;
2402 
2403                 /*
2404                  * Do not allow deletion if the parent dir is the same
2405                  * as the dir to be deleted.  That means the ioctl
2406                  * must be called on the dentry referencing the root
2407                  * of the subvol, not a random directory contained
2408                  * within it.
2409                  */
2410                 err = -EINVAL;
2411                 if (root == dest)
2412                         goto out_dput;
2413 
2414                 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2415                 if (err)
2416                         goto out_dput;
2417         }
2418 
2419         /* check if subvolume may be deleted by a user */
2420         err = btrfs_may_delete(dir, dentry, 1);
2421         if (err)
2422                 goto out_dput;
2423 
2424         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2425                 err = -EINVAL;
2426                 goto out_dput;
2427         }
2428 
2429         inode_lock(inode);
2430 
2431         /*
2432          * Don't allow to delete a subvolume with send in progress. This is
2433          * inside the i_mutex so the error handling that has to drop the bit
2434          * again is not run concurrently.
2435          */
2436         spin_lock(&dest->root_item_lock);
2437         root_flags = btrfs_root_flags(&dest->root_item);
2438         if (dest->send_in_progress == 0) {
2439                 btrfs_set_root_flags(&dest->root_item,
2440                                 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2441                 spin_unlock(&dest->root_item_lock);
2442         } else {
2443                 spin_unlock(&dest->root_item_lock);
2444                 btrfs_warn(root->fs_info,
2445                         "Attempt to delete subvolume %llu during send",
2446                         dest->root_key.objectid);
2447                 err = -EPERM;
2448                 goto out_unlock_inode;
2449         }
2450 
2451         down_write(&root->fs_info->subvol_sem);
2452 
2453         err = may_destroy_subvol(dest);
2454         if (err)
2455                 goto out_up_write;
2456 
2457         btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2458         /*
2459          * One for dir inode, two for dir entries, two for root
2460          * ref/backref.
2461          */
2462         err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2463                                                5, &qgroup_reserved, true);
2464         if (err)
2465                 goto out_up_write;
2466 
2467         trans = btrfs_start_transaction(root, 0);
2468         if (IS_ERR(trans)) {
2469                 err = PTR_ERR(trans);
2470                 goto out_release;
2471         }
2472         trans->block_rsv = &block_rsv;
2473         trans->bytes_reserved = block_rsv.size;
2474 
2475         btrfs_record_snapshot_destroy(trans, dir);
2476 
2477         ret = btrfs_unlink_subvol(trans, root, dir,
2478                                 dest->root_key.objectid,
2479                                 dentry->d_name.name,
2480                                 dentry->d_name.len);
2481         if (ret) {
2482                 err = ret;
2483                 btrfs_abort_transaction(trans, root, ret);
2484                 goto out_end_trans;
2485         }
2486 
2487         btrfs_record_root_in_trans(trans, dest);
2488 
2489         memset(&dest->root_item.drop_progress, 0,
2490                 sizeof(dest->root_item.drop_progress));
2491         dest->root_item.drop_level = 0;
2492         btrfs_set_root_refs(&dest->root_item, 0);
2493 
2494         if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2495                 ret = btrfs_insert_orphan_item(trans,
2496                                         root->fs_info->tree_root,
2497                                         dest->root_key.objectid);
2498                 if (ret) {
2499                         btrfs_abort_transaction(trans, root, ret);
2500                         err = ret;
2501                         goto out_end_trans;
2502                 }
2503         }
2504 
2505         ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2506                                   dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2507                                   dest->root_key.objectid);
2508         if (ret && ret != -ENOENT) {
2509                 btrfs_abort_transaction(trans, root, ret);
2510                 err = ret;
2511                 goto out_end_trans;
2512         }
2513         if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2514                 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2515                                           dest->root_item.received_uuid,
2516                                           BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2517                                           dest->root_key.objectid);
2518                 if (ret && ret != -ENOENT) {
2519                         btrfs_abort_transaction(trans, root, ret);
2520                         err = ret;
2521                         goto out_end_trans;
2522                 }
2523         }
2524 
2525 out_end_trans:
2526         trans->block_rsv = NULL;
2527         trans->bytes_reserved = 0;
2528         ret = btrfs_end_transaction(trans, root);
2529         if (ret && !err)
2530                 err = ret;
2531         inode->i_flags |= S_DEAD;
2532 out_release:
2533         btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2534 out_up_write:
2535         up_write(&root->fs_info->subvol_sem);
2536         if (err) {
2537                 spin_lock(&dest->root_item_lock);
2538                 root_flags = btrfs_root_flags(&dest->root_item);
2539                 btrfs_set_root_flags(&dest->root_item,
2540                                 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2541                 spin_unlock(&dest->root_item_lock);
2542         }
2543 out_unlock_inode:
2544         inode_unlock(inode);
2545         if (!err) {
2546                 d_invalidate(dentry);
2547                 btrfs_invalidate_inodes(dest);
2548                 d_delete(dentry);
2549                 ASSERT(dest->send_in_progress == 0);
2550 
2551                 /* the last ref */
2552                 if (dest->ino_cache_inode) {
2553                         iput(dest->ino_cache_inode);
2554                         dest->ino_cache_inode = NULL;
2555                 }
2556         }
2557 out_dput:
2558         dput(dentry);
2559 out_unlock_dir:
2560         inode_unlock(dir);
2561 out_drop_write:
2562         mnt_drop_write_file(file);
2563 out:
2564         kfree(vol_args);
2565         return err;
2566 }
2567 
2568 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2569 {
2570         struct inode *inode = file_inode(file);
2571         struct btrfs_root *root = BTRFS_I(inode)->root;
2572         struct btrfs_ioctl_defrag_range_args *range;
2573         int ret;
2574 
2575         ret = mnt_want_write_file(file);
2576         if (ret)
2577                 return ret;
2578 
2579         if (btrfs_root_readonly(root)) {
2580                 ret = -EROFS;
2581                 goto out;
2582         }
2583 
2584         switch (inode->i_mode & S_IFMT) {
2585         case S_IFDIR:
2586                 if (!capable(CAP_SYS_ADMIN)) {
2587                         ret = -EPERM;
2588                         goto out;
2589                 }
2590                 ret = btrfs_defrag_root(root);
2591                 if (ret)
2592                         goto out;
2593                 ret = btrfs_defrag_root(root->fs_info->extent_root);
2594                 break;
2595         case S_IFREG:
2596                 if (!(file->f_mode & FMODE_WRITE)) {
2597                         ret = -EINVAL;
2598                         goto out;
2599                 }
2600 
2601                 range = kzalloc(sizeof(*range), GFP_KERNEL);
2602                 if (!range) {
2603                         ret = -ENOMEM;
2604                         goto out;
2605                 }
2606 
2607                 if (argp) {
2608                         if (copy_from_user(range, argp,
2609                                            sizeof(*range))) {
2610                                 ret = -EFAULT;
2611                                 kfree(range);
2612                                 goto out;
2613                         }
2614                         /* compression requires us to start the IO */
2615                         if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2616                                 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2617                                 range->extent_thresh = (u32)-1;
2618                         }
2619                 } else {
2620                         /* the rest are all set to zero by kzalloc */
2621                         range->len = (u64)-1;
2622                 }
2623                 ret = btrfs_defrag_file(file_inode(file), file,
2624                                         range, 0, 0);
2625                 if (ret > 0)
2626                         ret = 0;
2627                 kfree(range);
2628                 break;
2629         default:
2630                 ret = -EINVAL;
2631         }
2632 out:
2633         mnt_drop_write_file(file);
2634         return ret;
2635 }
2636 
2637 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2638 {
2639         struct btrfs_ioctl_vol_args *vol_args;
2640         int ret;
2641 
2642         if (!capable(CAP_SYS_ADMIN))
2643                 return -EPERM;
2644 
2645         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2646                         1)) {
2647                 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2648         }
2649 
2650         mutex_lock(&root->fs_info->volume_mutex);
2651         vol_args = memdup_user(arg, sizeof(*vol_args));
2652         if (IS_ERR(vol_args)) {
2653                 ret = PTR_ERR(vol_args);
2654                 goto out;
2655         }
2656 
2657         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2658         ret = btrfs_init_new_device(root, vol_args->name);
2659 
2660         if (!ret)
2661                 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2662 
2663         kfree(vol_args);
2664 out:
2665         mutex_unlock(&root->fs_info->volume_mutex);
2666         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2667         return ret;
2668 }
2669 
2670 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2671 {
2672         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2673         struct btrfs_ioctl_vol_args *vol_args;
2674         int ret;
2675 
2676         if (!capable(CAP_SYS_ADMIN))
2677                 return -EPERM;
2678 
2679         ret = mnt_want_write_file(file);
2680         if (ret)
2681                 return ret;
2682 
2683         vol_args = memdup_user(arg, sizeof(*vol_args));
2684         if (IS_ERR(vol_args)) {
2685                 ret = PTR_ERR(vol_args);
2686                 goto err_drop;
2687         }
2688 
2689         vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2690 
2691         if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2692                         1)) {
2693                 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2694                 goto out;
2695         }
2696 
2697         mutex_lock(&root->fs_info->volume_mutex);
2698         ret = btrfs_rm_device(root, vol_args->name);
2699         mutex_unlock(&root->fs_info->volume_mutex);
2700         atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2701 
2702         if (!ret)
2703                 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2704 
2705 out:
2706         kfree(vol_args);
2707 err_drop:
2708         mnt_drop_write_file(file);
2709         return ret;
2710 }
2711 
2712 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2713 {
2714         struct btrfs_ioctl_fs_info_args *fi_args;
2715         struct btrfs_device *device;
2716         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2717         int ret = 0;
2718 
2719         fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2720         if (!fi_args)
2721                 return -ENOMEM;
2722 
2723         mutex_lock(&fs_devices->device_list_mutex);
2724         fi_args->num_devices = fs_devices->num_devices;
2725         memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2726 
2727         list_for_each_entry(device, &fs_devices->devices, dev_list) {
2728                 if (device->devid > fi_args->max_id)
2729                         fi_args->max_id = device->devid;
2730         }
2731         mutex_unlock(&fs_devices->device_list_mutex);
2732 
2733         fi_args->nodesize = root->fs_info->super_copy->nodesize;
2734         fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2735         fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2736 
2737         if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2738                 ret = -EFAULT;
2739 
2740         kfree(fi_args);
2741         return ret;
2742 }
2743 
2744 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2745 {
2746         struct btrfs_ioctl_dev_info_args *di_args;
2747         struct btrfs_device *dev;
2748         struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2749         int ret = 0;
2750         char *s_uuid = NULL;
2751 
2752         di_args = memdup_user(arg, sizeof(*di_args));
2753         if (IS_ERR(di_args))
2754                 return PTR_ERR(di_args);
2755 
2756         if (!btrfs_is_empty_uuid(di_args->uuid))
2757                 s_uuid = di_args->uuid;
2758 
2759         mutex_lock(&fs_devices->device_list_mutex);
2760         dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2761 
2762         if (!dev) {
2763                 ret = -ENODEV;
2764                 goto out;
2765         }
2766 
2767         di_args->devid = dev->devid;
2768         di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2769         di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2770         memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2771         if (dev->name) {
2772                 struct rcu_string *name;
2773 
2774                 rcu_read_lock();
2775                 name = rcu_dereference(dev->name);
2776                 strncpy(di_args->path, name->str, sizeof(di_args->path));
2777                 rcu_read_unlock();
2778                 di_args->path[sizeof(di_args->path) - 1] = 0;
2779         } else {
2780                 di_args->path[0] = '\0';
2781         }
2782 
2783 out:
2784         mutex_unlock(&fs_devices->device_list_mutex);
2785         if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2786                 ret = -EFAULT;
2787 
2788         kfree(di_args);
2789         return ret;
2790 }
2791 
2792 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2793 {
2794         struct page *page;
2795 
2796         page = grab_cache_page(inode->i_mapping, index);
2797         if (!page)
2798                 return ERR_PTR(-ENOMEM);
2799 
2800         if (!PageUptodate(page)) {
2801                 int ret;
2802 
2803                 ret = btrfs_readpage(NULL, page);
2804                 if (ret)
2805                         return ERR_PTR(ret);
2806                 lock_page(page);
2807                 if (!PageUptodate(page)) {
2808                         unlock_page(page);
2809                         put_page(page);
2810                         return ERR_PTR(-EIO);
2811                 }
2812                 if (page->mapping != inode->i_mapping) {
2813                         unlock_page(page);
2814                         put_page(page);
2815                         return ERR_PTR(-EAGAIN);
2816                 }
2817         }
2818 
2819         return page;
2820 }
2821 
2822 static int gather_extent_pages(struct inode *inode, struct page **pages,
2823                                int num_pages, u64 off)
2824 {
2825         int i;
2826         pgoff_t index = off >> PAGE_SHIFT;
2827 
2828         for (i = 0; i < num_pages; i++) {
2829 again:
2830                 pages[i] = extent_same_get_page(inode, index + i);
2831                 if (IS_ERR(pages[i])) {
2832                         int err = PTR_ERR(pages[i]);
2833 
2834                         if (err == -EAGAIN)
2835                                 goto again;
2836                         pages[i] = NULL;
2837                         return err;
2838                 }
2839         }
2840         return 0;
2841 }
2842 
2843 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2844                              bool retry_range_locking)
2845 {
2846         /*
2847          * Do any pending delalloc/csum calculations on inode, one way or
2848          * another, and lock file content.
2849          * The locking order is:
2850          *
2851          *   1) pages
2852          *   2) range in the inode's io tree
2853          */
2854         while (1) {
2855                 struct btrfs_ordered_extent *ordered;
2856                 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2857                 ordered = btrfs_lookup_first_ordered_extent(inode,
2858                                                             off + len - 1);
2859                 if ((!ordered ||
2860                      ordered->file_offset + ordered->len <= off ||
2861                      ordered->file_offset >= off + len) &&
2862                     !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2863                                     off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2864                         if (ordered)
2865                                 btrfs_put_ordered_extent(ordered);
2866                         break;
2867                 }
2868                 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2869                 if (ordered)
2870                         btrfs_put_ordered_extent(ordered);
2871                 if (!retry_range_locking)
2872                         return -EAGAIN;
2873                 btrfs_wait_ordered_range(inode, off, len);
2874         }
2875         return 0;
2876 }
2877 
2878 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2879 {
2880         inode_unlock(inode1);
2881         inode_unlock(inode2);
2882 }
2883 
2884 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2885 {
2886         if (inode1 < inode2)
2887                 swap(inode1, inode2);
2888 
2889         inode_lock_nested(inode1, I_MUTEX_PARENT);
2890         inode_lock_nested(inode2, I_MUTEX_CHILD);
2891 }
2892 
2893 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2894                                       struct inode *inode2, u64 loff2, u64 len)
2895 {
2896         unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2897         unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2898 }
2899 
2900 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2901                                     struct inode *inode2, u64 loff2, u64 len,
2902                                     bool retry_range_locking)
2903 {
2904         int ret;
2905 
2906         if (inode1 < inode2) {
2907                 swap(inode1, inode2);
2908                 swap(loff1, loff2);
2909         }
2910         ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2911         if (ret)
2912                 return ret;
2913         ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2914         if (ret)
2915                 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2916                               loff1 + len - 1);
2917         return ret;
2918 }
2919 
2920 struct cmp_pages {
2921         int             num_pages;
2922         struct page     **src_pages;
2923         struct page     **dst_pages;
2924 };
2925 
2926 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2927 {
2928         int i;
2929         struct page *pg;
2930 
2931         for (i = 0; i < cmp->num_pages; i++) {
2932                 pg = cmp->src_pages[i];
2933                 if (pg) {
2934                         unlock_page(pg);
2935                         put_page(pg);
2936                 }
2937                 pg = cmp->dst_pages[i];
2938                 if (pg) {
2939                         unlock_page(pg);
2940                         put_page(pg);
2941                 }
2942         }
2943         kfree(cmp->src_pages);
2944         kfree(cmp->dst_pages);
2945 }
2946 
2947 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2948                                   struct inode *dst, u64 dst_loff,
2949                                   u64 len, struct cmp_pages *cmp)
2950 {
2951         int ret;
2952         int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
2953         struct page **src_pgarr, **dst_pgarr;
2954 
2955         /*
2956          * We must gather up all the pages before we initiate our
2957          * extent locking. We use an array for the page pointers. Size
2958          * of the array is bounded by len, which is in turn bounded by
2959          * BTRFS_MAX_DEDUPE_LEN.
2960          */
2961         src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2962         dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2963         if (!src_pgarr || !dst_pgarr) {
2964                 kfree(src_pgarr);
2965                 kfree(dst_pgarr);
2966                 return -ENOMEM;
2967         }
2968         cmp->num_pages = num_pages;
2969         cmp->src_pages = src_pgarr;
2970         cmp->dst_pages = dst_pgarr;
2971 
2972         ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2973         if (ret)
2974                 goto out;
2975 
2976         ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2977 
2978 out:
2979         if (ret)
2980                 btrfs_cmp_data_free(cmp);
2981         return 0;
2982 }
2983 
2984 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2985                           u64 dst_loff, u64 len, struct cmp_pages *cmp)
2986 {
2987         int ret = 0;
2988         int i;
2989         struct page *src_page, *dst_page;
2990         unsigned int cmp_len = PAGE_SIZE;
2991         void *addr, *dst_addr;
2992 
2993         i = 0;
2994         while (len) {
2995                 if (len < PAGE_SIZE)
2996                         cmp_len = len;
2997 
2998                 BUG_ON(i >= cmp->num_pages);
2999 
3000                 src_page = cmp->src_pages[i];
3001                 dst_page = cmp->dst_pages[i];
3002                 ASSERT(PageLocked(src_page));
3003                 ASSERT(PageLocked(dst_page));
3004 
3005                 addr = kmap_atomic(src_page);
3006                 dst_addr = kmap_atomic(dst_page);
3007 
3008                 flush_dcache_page(src_page);
3009                 flush_dcache_page(dst_page);
3010 
3011                 if (memcmp(addr, dst_addr, cmp_len))
3012                         ret = -EBADE;
3013 
3014                 kunmap_atomic(addr);
3015                 kunmap_atomic(dst_addr);
3016 
3017                 if (ret)
3018                         break;
3019 
3020                 len -= cmp_len;
3021                 i++;
3022         }
3023 
3024         return ret;
3025 }
3026 
3027 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3028                                      u64 olen)
3029 {
3030         u64 len = *plen;
3031         u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3032 
3033         if (off + olen > inode->i_size || off + olen < off)
3034                 return -EINVAL;
3035 
3036         /* if we extend to eof, continue to block boundary */
3037         if (off + len == inode->i_size)
3038                 *plen = len = ALIGN(inode->i_size, bs) - off;
3039 
3040         /* Check that we are block aligned - btrfs_clone() requires this */
3041         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3042                 return -EINVAL;
3043 
3044         return 0;
3045 }
3046 
3047 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3048                              struct inode *dst, u64 dst_loff)
3049 {
3050         int ret;
3051         u64 len = olen;
3052         struct cmp_pages cmp;
3053         int same_inode = 0;
3054         u64 same_lock_start = 0;
3055         u64 same_lock_len = 0;
3056 
3057         if (src == dst)
3058                 same_inode = 1;
3059 
3060         if (len == 0)
3061                 return 0;
3062 
3063         if (same_inode) {
3064                 inode_lock(src);
3065 
3066                 ret = extent_same_check_offsets(src, loff, &len, olen);
3067                 if (ret)
3068                         goto out_unlock;
3069                 ret = extent_same_check_offsets(src, dst_loff, &len, olen);
3070                 if (ret)
3071                         goto out_unlock;
3072 
3073                 /*
3074                  * Single inode case wants the same checks, except we
3075                  * don't want our length pushed out past i_size as
3076                  * comparing that data range makes no sense.
3077                  *
3078                  * extent_same_check_offsets() will do this for an
3079                  * unaligned length at i_size, so catch it here and
3080                  * reject the request.
3081                  *
3082                  * This effectively means we require aligned extents
3083                  * for the single-inode case, whereas the other cases
3084                  * allow an unaligned length so long as it ends at
3085                  * i_size.
3086                  */
3087                 if (len != olen) {
3088                         ret = -EINVAL;
3089                         goto out_unlock;
3090                 }
3091 
3092                 /* Check for overlapping ranges */
3093                 if (dst_loff + len > loff && dst_loff < loff + len) {
3094                         ret = -EINVAL;
3095                         goto out_unlock;
3096                 }
3097 
3098                 same_lock_start = min_t(u64, loff, dst_loff);
3099                 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3100         } else {
3101                 btrfs_double_inode_lock(src, dst);
3102 
3103                 ret = extent_same_check_offsets(src, loff, &len, olen);
3104                 if (ret)
3105                         goto out_unlock;
3106 
3107                 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3108                 if (ret)
3109                         goto out_unlock;
3110         }
3111 
3112         /* don't make the dst file partly checksummed */
3113         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3114             (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3115                 ret = -EINVAL;
3116                 goto out_unlock;
3117         }
3118 
3119 again:
3120         ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3121         if (ret)
3122                 goto out_unlock;
3123 
3124         if (same_inode)
3125                 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3126                                         false);
3127         else
3128                 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3129                                                false);
3130         /*
3131          * If one of the inodes has dirty pages in the respective range or
3132          * ordered extents, we need to flush dellaloc and wait for all ordered
3133          * extents in the range. We must unlock the pages and the ranges in the
3134          * io trees to avoid deadlocks when flushing delalloc (requires locking
3135          * pages) and when waiting for ordered extents to complete (they require
3136          * range locking).
3137          */
3138         if (ret == -EAGAIN) {
3139                 /*
3140                  * Ranges in the io trees already unlocked. Now unlock all
3141                  * pages before waiting for all IO to complete.
3142                  */
3143                 btrfs_cmp_data_free(&cmp);
3144                 if (same_inode) {
3145                         btrfs_wait_ordered_range(src, same_lock_start,
3146                                                  same_lock_len);
3147                 } else {
3148                         btrfs_wait_ordered_range(src, loff, len);
3149                         btrfs_wait_ordered_range(dst, dst_loff, len);
3150                 }
3151                 goto again;
3152         }
3153         ASSERT(ret == 0);
3154         if (WARN_ON(ret)) {
3155                 /* ranges in the io trees already unlocked */
3156                 btrfs_cmp_data_free(&cmp);
3157                 return ret;
3158         }
3159 
3160         /* pass original length for comparison so we stay within i_size */
3161         ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
3162         if (ret == 0)
3163                 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3164 
3165         if (same_inode)
3166                 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3167                               same_lock_start + same_lock_len - 1);
3168         else
3169                 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3170 
3171         btrfs_cmp_data_free(&cmp);
3172 out_unlock:
3173         if (same_inode)
3174                 inode_unlock(src);
3175         else
3176                 btrfs_double_inode_unlock(src, dst);
3177 
3178         return ret;
3179 }
3180 
3181 #define BTRFS_MAX_DEDUPE_LEN    SZ_16M
3182 
3183 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3184                                 struct file *dst_file, u64 dst_loff)
3185 {
3186         struct inode *src = file_inode(src_file);
3187         struct inode *dst = file_inode(dst_file);
3188         u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3189         ssize_t res;
3190 
3191         if (olen > BTRFS_MAX_DEDUPE_LEN)
3192                 olen = BTRFS_MAX_DEDUPE_LEN;
3193 
3194         if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3195                 /*
3196                  * Btrfs does not support blocksize < page_size. As a
3197                  * result, btrfs_cmp_data() won't correctly handle
3198                  * this situation without an update.
3199                  */
3200                 return -EINVAL;
3201         }
3202 
3203         res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3204         if (res)
3205                 return res;
3206         return olen;
3207 }
3208 
3209 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3210                                      struct inode *inode,
3211                                      u64 endoff,
3212                                      const u64 destoff,
3213                                      const u64 olen,
3214                                      int no_time_update)
3215 {
3216         struct btrfs_root *root = BTRFS_I(inode)->root;
3217         int ret;
3218 
3219         inode_inc_iversion(inode);
3220         if (!no_time_update)
3221                 inode->i_mtime = inode->i_ctime = current_fs_time(inode->i_sb);
3222         /*
3223          * We round up to the block size at eof when determining which
3224          * extents to clone above, but shouldn't round up the file size.
3225          */
3226         if (endoff > destoff + olen)
3227                 endoff = destoff + olen;
3228         if (endoff > inode->i_size)
3229                 btrfs_i_size_write(inode, endoff);
3230 
3231         ret = btrfs_update_inode(trans, root, inode);
3232         if (ret) {
3233                 btrfs_abort_transaction(trans, root, ret);
3234                 btrfs_end_transaction(trans, root);
3235                 goto out;
3236         }
3237         ret = btrfs_end_transaction(trans, root);
3238 out:
3239         return ret;
3240 }
3241 
3242 static void clone_update_extent_map(struct inode *inode,
3243                                     const struct btrfs_trans_handle *trans,
3244                                     const struct btrfs_path *path,
3245                                     const u64 hole_offset,
3246                                     const u64 hole_len)
3247 {
3248         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3249         struct extent_map *em;
3250         int ret;
3251 
3252         em = alloc_extent_map();
3253         if (!em) {
3254                 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3255                         &BTRFS_I(inode)->runtime_flags);
3256                 return;
3257         }
3258 
3259         if (path) {
3260                 struct btrfs_file_extent_item *fi;
3261 
3262                 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3263                                     struct btrfs_file_extent_item);
3264                 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3265                 em->generation = -1;
3266                 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3267                     BTRFS_FILE_EXTENT_INLINE)
3268                         set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3269                                 &BTRFS_I(inode)->runtime_flags);
3270         } else {
3271                 em->start = hole_offset;
3272                 em->len = hole_len;
3273                 em->ram_bytes = em->len;
3274                 em->orig_start = hole_offset;
3275                 em->block_start = EXTENT_MAP_HOLE;
3276                 em->block_len = 0;
3277                 em->orig_block_len = 0;
3278                 em->compress_type = BTRFS_COMPRESS_NONE;
3279                 em->generation = trans->transid;
3280         }
3281 
3282         while (1) {
3283                 write_lock(&em_tree->lock);
3284                 ret = add_extent_mapping(em_tree, em, 1);
3285                 write_unlock(&em_tree->lock);
3286                 if (ret != -EEXIST) {
3287                         free_extent_map(em);
3288                         break;
3289                 }
3290                 btrfs_drop_extent_cache(inode, em->start,
3291                                         em->start + em->len - 1, 0);
3292         }
3293 
3294         if (ret)
3295                 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3296                         &BTRFS_I(inode)->runtime_flags);
3297 }
3298 
3299 /*
3300  * Make sure we do not end up inserting an inline extent into a file that has
3301  * already other (non-inline) extents. If a file has an inline extent it can
3302  * not have any other extents and the (single) inline extent must start at the
3303  * file offset 0. Failing to respect these rules will lead to file corruption,
3304  * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3305  *
3306  * We can have extents that have been already written to disk or we can have
3307  * dirty ranges still in delalloc, in which case the extent maps and items are
3308  * created only when we run delalloc, and the delalloc ranges might fall outside
3309  * the range we are currently locking in the inode's io tree. So we check the
3310  * inode's i_size because of that (i_size updates are done while holding the
3311  * i_mutex, which we are holding here).
3312  * We also check to see if the inode has a size not greater than "datal" but has
3313  * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3314  * protected against such concurrent fallocate calls by the i_mutex).
3315  *
3316  * If the file has no extents but a size greater than datal, do not allow the
3317  * copy because we would need turn the inline extent into a non-inline one (even
3318  * with NO_HOLES enabled). If we find our destination inode only has one inline
3319  * extent, just overwrite it with the source inline extent if its size is less
3320  * than the source extent's size, or we could copy the source inline extent's
3321  * data into the destination inode's inline extent if the later is greater then
3322  * the former.
3323  */
3324 static int clone_copy_inline_extent(struct inode *src,
3325                                     struct inode *dst,
3326                                     struct btrfs_trans_handle *trans,
3327                                     struct btrfs_path *path,
3328                                     struct btrfs_key *new_key,
3329                                     const u64 drop_start,
3330                                     const u64 datal,
3331                                     const u64 skip,
3332                                     const u64 size,
3333                                     char *inline_data)
3334 {
3335         struct btrfs_root *root = BTRFS_I(dst)->root;
3336         const u64 aligned_end = ALIGN(new_key->offset + datal,
3337                                       root->sectorsize);
3338         int ret;
3339         struct btrfs_key key;
3340 
3341         if (new_key->offset > 0)
3342                 return -EOPNOTSUPP;
3343 
3344         key.objectid = btrfs_ino(dst);
3345         key.type = BTRFS_EXTENT_DATA_KEY;
3346         key.offset = 0;
3347         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3348         if (ret < 0) {
3349                 return ret;
3350         } else if (ret > 0) {
3351                 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3352                         ret = btrfs_next_leaf(root, path);
3353                         if (ret < 0)
3354                                 return ret;
3355                         else if (ret > 0)
3356                                 goto copy_inline_extent;
3357                 }
3358                 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3359                 if (key.objectid == btrfs_ino(dst) &&
3360                     key.type == BTRFS_EXTENT_DATA_KEY) {
3361                         ASSERT(key.offset > 0);
3362                         return -EOPNOTSUPP;
3363                 }
3364         } else if (i_size_read(dst) <= datal) {
3365                 struct btrfs_file_extent_item *ei;
3366                 u64 ext_len;
3367 
3368                 /*
3369                  * If the file size is <= datal, make sure there are no other
3370                  * extents following (can happen do to an fallocate call with
3371                  * the flag FALLOC_FL_KEEP_SIZE).
3372                  */
3373                 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3374                                     struct btrfs_file_extent_item);
3375                 /*
3376                  * If it's an inline extent, it can not have other extents
3377                  * following it.
3378                  */
3379                 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3380                     BTRFS_FILE_EXTENT_INLINE)
3381                         goto copy_inline_extent;
3382 
3383                 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3384                 if (ext_len > aligned_end)
3385                         return -EOPNOTSUPP;
3386 
3387                 ret = btrfs_next_item(root, path);
3388                 if (ret < 0) {
3389                         return ret;
3390                 } else if (ret == 0) {
3391                         btrfs_item_key_to_cpu(path->nodes[0], &key,
3392                                               path->slots[0]);
3393                         if (key.objectid == btrfs_ino(dst) &&
3394                             key.type == BTRFS_EXTENT_DATA_KEY)
3395                                 return -EOPNOTSUPP;
3396                 }
3397         }
3398 
3399 copy_inline_extent:
3400         /*
3401          * We have no extent items, or we have an extent at offset 0 which may
3402          * or may not be inlined. All these cases are dealt the same way.
3403          */
3404         if (i_size_read(dst) > datal) {
3405                 /*
3406                  * If the destination inode has an inline extent...
3407                  * This would require copying the data from the source inline
3408                  * extent into the beginning of the destination's inline extent.
3409                  * But this is really complex, both extents can be compressed
3410                  * or just one of them, which would require decompressing and
3411                  * re-compressing data (which could increase the new compressed
3412                  * size, not allowing the compressed data to fit anymore in an
3413                  * inline extent).
3414                  * So just don't support this case for now (it should be rare,
3415                  * we are not really saving space when cloning inline extents).
3416                  */
3417                 return -EOPNOTSUPP;
3418         }
3419 
3420         btrfs_release_path(path);
3421         ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3422         if (ret)
3423                 return ret;
3424         ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3425         if (ret)
3426                 return ret;
3427 
3428         if (skip) {
3429                 const u32 start = btrfs_file_extent_calc_inline_size(0);
3430 
3431                 memmove(inline_data + start, inline_data + start + skip, datal);
3432         }
3433 
3434         write_extent_buffer(path->nodes[0], inline_data,
3435                             btrfs_item_ptr_offset(path->nodes[0],
3436                                                   path->slots[0]),
3437                             size);
3438         inode_add_bytes(dst, datal);
3439 
3440         return 0;
3441 }
3442 
3443 /**
3444  * btrfs_clone() - clone a range from inode file to another
3445  *
3446  * @src: Inode to clone from
3447  * @inode: Inode to clone to
3448  * @off: Offset within source to start clone from
3449  * @olen: Original length, passed by user, of range to clone
3450  * @olen_aligned: Block-aligned value of olen
3451  * @destoff: Offset within @inode to start clone
3452  * @no_time_update: Whether to update mtime/ctime on the target inode
3453  */
3454 static int btrfs_clone(struct inode *src, struct inode *inode,
3455                        const u64 off, const u64 olen, const u64 olen_aligned,
3456                        const u64 destoff, int no_time_update)
3457 {
3458         struct btrfs_root *root = BTRFS_I(inode)->root;
3459         struct btrfs_path *path = NULL;
3460         struct extent_buffer *leaf;
3461         struct btrfs_trans_handle *trans;
3462         char *buf = NULL;
3463         struct btrfs_key key;
3464         u32 nritems;
3465         int slot;
3466         int ret;
3467         const u64 len = olen_aligned;
3468         u64 last_dest_end = destoff;
3469 
3470         ret = -ENOMEM;
3471         buf = vmalloc(root->nodesize);
3472         if (!buf)
3473                 return ret;
3474 
3475         path = btrfs_alloc_path();
3476         if (!path) {
3477                 vfree(buf);
3478                 return ret;
3479         }
3480 
3481         path->reada = READA_FORWARD;
3482         /* clone data */
3483         key.objectid = btrfs_ino(src);
3484         key.type = BTRFS_EXTENT_DATA_KEY;
3485         key.offset = off;
3486 
3487         while (1) {
3488                 u64 next_key_min_offset = key.offset + 1;
3489 
3490                 /*
3491                  * note the key will change type as we walk through the
3492                  * tree.
3493                  */
3494                 path->leave_spinning = 1;
3495                 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3496                                 0, 0);
3497                 if (ret < 0)
3498                         goto out;
3499                 /*
3500                  * First search, if no extent item that starts at offset off was
3501                  * found but the previous item is an extent item, it's possible
3502                  * it might overlap our target range, therefore process it.
3503                  */
3504                 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3505                         btrfs_item_key_to_cpu(path->nodes[0], &key,
3506                                               path->slots[0] - 1);
3507                         if (key.type == BTRFS_EXTENT_DATA_KEY)
3508                                 path->slots[0]--;
3509                 }
3510 
3511                 nritems = btrfs_header_nritems(path->nodes[0]);
3512 process_slot:
3513                 if (path->slots[0] >= nritems) {
3514                         ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3515                         if (ret < 0)
3516                                 goto out;
3517                         if (ret > 0)
3518                                 break;
3519                         nritems = btrfs_header_nritems(path->nodes[0]);
3520                 }
3521                 leaf = path->nodes[0];
3522                 slot = path->slots[0];
3523 
3524                 btrfs_item_key_to_cpu(leaf, &key, slot);
3525                 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3526                     key.objectid != btrfs_ino(src))
3527                         break;
3528 
3529                 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3530                         struct btrfs_file_extent_item *extent;
3531                         int type;
3532                         u32 size;
3533                         struct btrfs_key new_key;
3534                         u64 disko = 0, diskl = 0;
3535                         u64 datao = 0, datal = 0;
3536                         u8 comp;
3537                         u64 drop_start;
3538 
3539                         extent = btrfs_item_ptr(leaf, slot,
3540                                                 struct btrfs_file_extent_item);
3541                         comp = btrfs_file_extent_compression(leaf, extent);
3542                         type = btrfs_file_extent_type(leaf, extent);
3543                         if (type == BTRFS_FILE_EXTENT_REG ||
3544                             type == BTRFS_FILE_EXTENT_PREALLOC) {
3545                                 disko = btrfs_file_extent_disk_bytenr(leaf,
3546                                                                       extent);
3547                                 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3548                                                                  extent);
3549                                 datao = btrfs_file_extent_offset(leaf, extent);
3550                                 datal = btrfs_file_extent_num_bytes(leaf,
3551                                                                     extent);
3552                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3553                                 /* take upper bound, may be compressed */
3554                                 datal = btrfs_file_extent_ram_bytes(leaf,
3555                                                                     extent);
3556                         }
3557 
3558                         /*
3559                          * The first search might have left us at an extent
3560                          * item that ends before our target range's start, can
3561                          * happen if we have holes and NO_HOLES feature enabled.
3562                          */
3563                         if (key.offset + datal <= off) {
3564                                 path->slots[0]++;
3565                                 goto process_slot;
3566                         } else if (key.offset >= off + len) {
3567                                 break;
3568                         }
3569                         next_key_min_offset = key.offset + datal;
3570                         size = btrfs_item_size_nr(leaf, slot);
3571                         read_extent_buffer(leaf, buf,
3572                                            btrfs_item_ptr_offset(leaf, slot),
3573                                            size);
3574 
3575                         btrfs_release_path(path);
3576                         path->leave_spinning = 0;
3577 
3578                         memcpy(&new_key, &key, sizeof(new_key));
3579                         new_key.objectid = btrfs_ino(inode);
3580                         if (off <= key.offset)
3581                                 new_key.offset = key.offset + destoff - off;
3582                         else
3583                                 new_key.offset = destoff;
3584 
3585                         /*
3586                          * Deal with a hole that doesn't have an extent item
3587                          * that represents it (NO_HOLES feature enabled).
3588                          * This hole is either in the middle of the cloning
3589                          * range or at the beginning (fully overlaps it or
3590                          * partially overlaps it).
3591                          */
3592                         if (new_key.offset != last_dest_end)
3593                                 drop_start = last_dest_end;
3594                         else
3595                                 drop_start = new_key.offset;
3596 
3597                         /*
3598                          * 1 - adjusting old extent (we may have to split it)
3599                          * 1 - add new extent
3600                          * 1 - inode update
3601                          */
3602                         trans = btrfs_start_transaction(root, 3);
3603                         if (IS_ERR(trans)) {
3604                                 ret = PTR_ERR(trans);
3605                                 goto out;
3606                         }
3607 
3608                         if (type == BTRFS_FILE_EXTENT_REG ||
3609                             type == BTRFS_FILE_EXTENT_PREALLOC) {
3610                                 /*
3611                                  *    a  | --- range to clone ---|  b
3612                                  * | ------------- extent ------------- |
3613                                  */
3614 
3615                                 /* subtract range b */
3616                                 if (key.offset + datal > off + len)
3617                                         datal = off + len - key.offset;
3618 
3619                                 /* subtract range a */
3620                                 if (off > key.offset) {
3621                                         datao += off - key.offset;
3622                                         datal -= off - key.offset;
3623                                 }
3624 
3625                                 ret = btrfs_drop_extents(trans, root, inode,
3626                                                          drop_start,
3627                                                          new_key.offset + datal,
3628                                                          1);
3629                                 if (ret) {
3630                                         if (ret != -EOPNOTSUPP)
3631                                                 btrfs_abort_transaction(trans,
3632                                                                 root, ret);
3633                                         btrfs_end_transaction(trans, root);
3634                                         goto out;
3635                                 }
3636 
3637                                 ret = btrfs_insert_empty_item(trans, root, path,
3638                                                               &new_key, size);
3639                                 if (ret) {
3640                                         btrfs_abort_transaction(trans, root,
3641                                                                 ret);
3642                                         btrfs_end_transaction(trans, root);
3643                                         goto out;
3644                                 }
3645 
3646                                 leaf = path->nodes[0];
3647                                 slot = path->slots[0];
3648                                 write_extent_buffer(leaf, buf,
3649                                             btrfs_item_ptr_offset(leaf, slot),
3650                                             size);
3651 
3652                                 extent = btrfs_item_ptr(leaf, slot,
3653                                                 struct btrfs_file_extent_item);
3654 
3655                                 /* disko == 0 means it's a hole */
3656                                 if (!disko)
3657                                         datao = 0;
3658 
3659                                 btrfs_set_file_extent_offset(leaf, extent,
3660                                                              datao);
3661                                 btrfs_set_file_extent_num_bytes(leaf, extent,
3662                                                                 datal);
3663 
3664                                 if (disko) {
3665                                         inode_add_bytes(inode, datal);
3666                                         ret = btrfs_inc_extent_ref(trans, root,
3667                                                         disko, diskl, 0,
3668                                                         root->root_key.objectid,
3669                                                         btrfs_ino(inode),
3670                                                         new_key.offset - datao);
3671                                         if (ret) {
3672                                                 btrfs_abort_transaction(trans,
3673                                                                         root,
3674                                                                         ret);
3675                                                 btrfs_end_transaction(trans,
3676                                                                       root);
3677                                                 goto out;
3678 
3679                                         }
3680                                 }
3681                         } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3682                                 u64 skip = 0;
3683                                 u64 trim = 0;
3684 
3685                                 if (off > key.offset) {
3686                                         skip = off - key.offset;
3687                                         new_key.offset += skip;
3688                                 }
3689 
3690                                 if (key.offset + datal > off + len)
3691                                         trim = key.offset + datal - (off + len);
3692 
3693                                 if (comp && (skip || trim)) {
3694                                         ret = -EINVAL;
3695                                         btrfs_end_transaction(trans, root);
3696                                         goto out;
3697                                 }
3698                                 size -= skip + trim;
3699                                 datal -= skip + trim;
3700 
3701                                 ret = clone_copy_inline_extent(src, inode,
3702                                                                trans, path,
3703                                                                &new_key,
3704                                                                drop_start,
3705                                                                datal,
3706                                                                skip, size, buf);
3707                                 if (ret) {
3708                                         if (ret != -EOPNOTSUPP)
3709                                                 btrfs_abort_transaction(trans,
3710                                                                         root,
3711                                                                         ret);
3712                                         btrfs_end_transaction(trans, root);
3713                                         goto out;
3714                                 }
3715                                 leaf = path->nodes[0];
3716                                 slot = path->slots[0];
3717                         }
3718 
3719                         /* If we have an implicit hole (NO_HOLES feature). */
3720                         if (drop_start < new_key.offset)
3721                                 clone_update_extent_map(inode, trans,
3722                                                 NULL, drop_start,
3723                                                 new_key.offset - drop_start);
3724 
3725                         clone_update_extent_map(inode, trans, path, 0, 0);
3726 
3727                         btrfs_mark_buffer_dirty(leaf);
3728                         btrfs_release_path(path);
3729 
3730                         last_dest_end = ALIGN(new_key.offset + datal,
3731                                               root->sectorsize);
3732                         ret = clone_finish_inode_update(trans, inode,
3733                                                         last_dest_end,
3734                                                         destoff, olen,
3735                                                         no_time_update);
3736                         if (ret)
3737                                 goto out;
3738                         if (new_key.offset + datal >= destoff + len)
3739                                 break;
3740                 }
3741                 btrfs_release_path(path);
3742                 key.offset = next_key_min_offset;
3743         }
3744         ret = 0;
3745 
3746         if (last_dest_end < destoff + len) {
3747                 /*
3748                  * We have an implicit hole (NO_HOLES feature is enabled) that
3749                  * fully or partially overlaps our cloning range at its end.
3750                  */
3751                 btrfs_release_path(path);
3752 
3753                 /*
3754                  * 1 - remove extent(s)
3755                  * 1 - inode update
3756                  */
3757                 trans = btrfs_start_transaction(root, 2);
3758                 if (IS_ERR(trans)) {
3759                         ret = PTR_ERR(trans);
3760                         goto out;
3761                 }
3762                 ret = btrfs_drop_extents(trans, root, inode,
3763                                          last_dest_end, destoff + len, 1);
3764                 if (ret) {
3765                         if (ret != -EOPNOTSUPP)
3766                                 btrfs_abort_transaction(trans, root, ret);
3767                         btrfs_end_transaction(trans, root);
3768                         goto out;
3769                 }
3770                 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3771                                         destoff + len - last_dest_end);
3772                 ret = clone_finish_inode_update(trans, inode, destoff + len,
3773                                                 destoff, olen, no_time_update);
3774         }
3775 
3776 out:
3777         btrfs_free_path(path);
3778         vfree(buf);
3779         return ret;
3780 }
3781 
3782 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3783                                         u64 off, u64 olen, u64 destoff)
3784 {
3785         struct inode *inode = file_inode(file);
3786         struct inode *src = file_inode(file_src);
3787         struct btrfs_root *root = BTRFS_I(inode)->root;
3788         int ret;
3789         u64 len = olen;
3790         u64 bs = root->fs_info->sb->s_blocksize;
3791         int same_inode = src == inode;
3792 
3793         /*
3794          * TODO:
3795          * - split compressed inline extents.  annoying: we need to
3796          *   decompress into destination's address_space (the file offset
3797          *   may change, so source mapping won't do), then recompress (or
3798          *   otherwise reinsert) a subrange.
3799          *
3800          * - split destination inode's inline extents.  The inline extents can
3801          *   be either compressed or non-compressed.
3802          */
3803 
3804         if (btrfs_root_readonly(root))
3805                 return -EROFS;
3806 
3807         if (file_src->f_path.mnt != file->f_path.mnt ||
3808             src->i_sb != inode->i_sb)
3809                 return -EXDEV;
3810 
3811         /* don't make the dst file partly checksummed */
3812         if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3813             (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3814                 return -EINVAL;
3815 
3816         if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3817                 return -EISDIR;
3818 
3819         if (!same_inode) {
3820                 btrfs_double_inode_lock(src, inode);
3821         } else {
3822                 inode_lock(src);
3823         }
3824 
3825         /* determine range to clone */
3826         ret = -EINVAL;
3827         if (off + len > src->i_size || off + len < off)
3828                 goto out_unlock;
3829         if (len == 0)
3830                 olen = len = src->i_size - off;
3831         /* if we extend to eof, continue to block boundary */
3832         if (off + len == src->i_size)
3833                 len = ALIGN(src->i_size, bs) - off;
3834 
3835         if (len == 0) {
3836                 ret = 0;
3837                 goto out_unlock;
3838         }
3839 
3840         /* verify the end result is block aligned */
3841         if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3842             !IS_ALIGNED(destoff, bs))
3843                 goto out_unlock;
3844 
3845         /* verify if ranges are overlapped within the same file */
3846         if (same_inode) {
3847                 if (destoff + len > off && destoff < off + len)
3848                         goto out_unlock;
3849         }
3850 
3851         if (destoff > inode->i_size) {
3852                 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3853                 if (ret)
3854                         goto out_unlock;
3855         }
3856 
3857         /*
3858          * Lock the target range too. Right after we replace the file extent
3859          * items in the fs tree (which now point to the cloned data), we might
3860          * have a worker replace them with extent items relative to a write
3861          * operation that was issued before this clone operation (i.e. confront
3862          * with inode.c:btrfs_finish_ordered_io).
3863          */
3864         if (same_inode) {
3865                 u64 lock_start = min_t(u64, off, destoff);
3866                 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3867 
3868                 ret = lock_extent_range(src, lock_start, lock_len, true);
3869         } else {
3870                 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3871                                                true);
3872         }
3873         ASSERT(ret == 0);
3874         if (WARN_ON(ret)) {
3875                 /* ranges in the io trees already unlocked */
3876                 goto out_unlock;
3877         }
3878 
3879         ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3880 
3881         if (same_inode) {
3882                 u64 lock_start = min_t(u64, off, destoff);
3883                 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3884 
3885                 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3886         } else {
3887                 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3888         }
3889         /*
3890          * Truncate page cache pages so that future reads will see the cloned
3891          * data immediately and not the previous data.
3892          */
3893         truncate_inode_pages_range(&inode->i_data,
3894                                 round_down(destoff, PAGE_SIZE),
3895                                 round_up(destoff + len, PAGE_SIZE) - 1);
3896 out_unlock:
3897         if (!same_inode)
3898                 btrfs_double_inode_unlock(src, inode);
3899         else
3900                 inode_unlock(src);
3901         return ret;
3902 }
3903 
3904 ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in,
3905                               struct file *file_out, loff_t pos_out,
3906                               size_t len, unsigned int flags)
3907 {
3908         ssize_t ret;
3909 
3910         ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out);
3911         if (ret == 0)
3912                 ret = len;
3913         return ret;
3914 }
3915 
3916 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3917                 struct file *dst_file, loff_t destoff, u64 len)
3918 {
3919         return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3920 }
3921 
3922 /*
3923  * there are many ways the trans_start and trans_end ioctls can lead
3924  * to deadlocks.  They should only be used by applications that
3925  * basically own the machine, and have a very in depth understanding
3926  * of all the possible deadlocks and enospc problems.
3927  */
3928 static long btrfs_ioctl_trans_start(struct file *file)
3929 {
3930         struct inode *inode = file_inode(file);
3931         struct btrfs_root *root = BTRFS_I(inode)->root;
3932         struct btrfs_trans_handle *trans;
3933         int ret;
3934 
3935         ret = -EPERM;
3936         if (!capable(CAP_SYS_ADMIN))
3937                 goto out;
3938 
3939         ret = -EINPROGRESS;
3940         if (file->private_data)
3941                 goto out;
3942 
3943         ret = -EROFS;
3944         if (btrfs_root_readonly(root))
3945                 goto out;
3946 
3947         ret = mnt_want_write_file(file);
3948         if (ret)
3949                 goto out;
3950 
3951         atomic_inc(&root->fs_info->open_ioctl_trans);
3952 
3953         ret = -ENOMEM;
3954         trans = btrfs_start_ioctl_transaction(root);
3955         if (IS_ERR(trans))
3956                 goto out_drop;
3957 
3958         file->private_data = trans;
3959         return 0;
3960 
3961 out_drop:
3962         atomic_dec(&root->fs_info->open_ioctl_trans);
3963         mnt_drop_write_file(file);
3964 out:
3965         return ret;
3966 }
3967 
3968 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3969 {
3970         struct inode *inode = file_inode(file);
3971         struct btrfs_root *root = BTRFS_I(inode)->root;
3972         struct btrfs_root *new_root;
3973         struct btrfs_dir_item *di;
3974         struct btrfs_trans_handle *trans;
3975         struct btrfs_path *path;
3976         struct btrfs_key location;
3977         struct btrfs_disk_key disk_key;
3978         u64 objectid = 0;
3979         u64 dir_id;
3980         int ret;
3981 
3982         if (!capable(CAP_SYS_ADMIN))
3983                 return -EPERM;
3984 
3985         ret = mnt_want_write_file(file);
3986         if (ret)
3987                 return ret;
3988 
3989         if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3990                 ret = -EFAULT;
3991                 goto out;
3992         }
3993 
3994         if (!objectid)
3995                 objectid = BTRFS_FS_TREE_OBJECTID;
3996 
3997         location.objectid = objectid;
3998         location.type = BTRFS_ROOT_ITEM_KEY;
3999         location.offset = (u64)-1;
4000 
4001         new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
4002         if (IS_ERR(new_root)) {
4003                 ret = PTR_ERR(new_root);
4004                 goto out;
4005         }
4006 
4007         path = btrfs_alloc_path();
4008         if (!path) {
4009                 ret = -ENOMEM;
4010                 goto out;
4011         }
4012         path->leave_spinning = 1;
4013 
4014         trans = btrfs_start_transaction(root, 1);
4015         if (IS_ERR(trans)) {
4016                 btrfs_free_path(path);
4017                 ret = PTR_ERR(trans);
4018                 goto out;
4019         }
4020 
4021         dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
4022         di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
4023                                    dir_id, "default", 7, 1);
4024         if (IS_ERR_OR_NULL(di)) {
4025                 btrfs_free_path(path);
4026                 btrfs_end_transaction(trans, root);
4027                 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
4028                            "item, this isn't going to work");
4029                 ret = -ENOENT;
4030                 goto out;
4031         }
4032 
4033         btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4034         btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4035         btrfs_mark_buffer_dirty(path->nodes[0]);
4036         btrfs_free_path(path);
4037 
4038         btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
4039         btrfs_end_transaction(trans, root);
4040 out:
4041         mnt_drop_write_file(file);
4042         return ret;
4043 }
4044 
4045 void btrfs_get_block_group_info(struct list_head *groups_list,
4046                                 struct btrfs_ioctl_space_info *space)
4047 {
4048         struct btrfs_block_group_cache *block_group;
4049 
4050         space->total_bytes = 0;
4051         space->used_bytes = 0;
4052         space->flags = 0;
4053         list_for_each_entry(block_group, groups_list, list) {
4054                 space->flags = block_group->flags;
4055                 space->total_bytes += block_group->key.offset;
4056                 space->used_bytes +=
4057                         btrfs_block_group_used(&block_group->item);
4058         }
4059 }
4060 
4061 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
4062 {
4063         struct btrfs_ioctl_space_args space_args;
4064         struct btrfs_ioctl_space_info space;
4065         struct btrfs_ioctl_space_info *dest;
4066         struct btrfs_ioctl_space_info *dest_orig;
4067         struct btrfs_ioctl_space_info __user *user_dest;
4068         struct btrfs_space_info *info;
4069         u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4070                        BTRFS_BLOCK_GROUP_SYSTEM,
4071                        BTRFS_BLOCK_GROUP_METADATA,
4072                        BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4073         int num_types = 4;
4074         int alloc_size;
4075         int ret = 0;
4076         u64 slot_count = 0;
4077         int i, c;
4078 
4079         if (copy_from_user(&space_args,
4080                            (struct btrfs_ioctl_space_args __user *)arg,
4081                            sizeof(space_args)))
4082                 return -EFAULT;
4083 
4084         for (i = 0; i < num_types; i++) {
4085                 struct btrfs_space_info *tmp;
4086 
4087                 info = NULL;
4088                 rcu_read_lock();
4089                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4090                                         list) {
4091                         if (tmp->flags == types[i]) {
4092                                 info = tmp;
4093                                 break;
4094                         }
4095                 }
4096                 rcu_read_unlock();
4097 
4098                 if (!info)
4099                         continue;
4100 
4101                 down_read(&info->groups_sem);
4102                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4103                         if (!list_empty(&info->block_groups[c]))
4104                                 slot_count++;
4105                 }
4106                 up_read(&info->groups_sem);
4107         }
4108 
4109         /*
4110          * Global block reserve, exported as a space_info
4111          */
4112         slot_count++;
4113 
4114         /* space_slots == 0 means they are asking for a count */
4115         if (space_args.space_slots == 0) {
4116                 space_args.total_spaces = slot_count;
4117                 goto out;
4118         }
4119 
4120         slot_count = min_t(u64, space_args.space_slots, slot_count);
4121 
4122         alloc_size = sizeof(*dest) * slot_count;
4123 
4124         /* we generally have at most 6 or so space infos, one for each raid
4125          * level.  So, a whole page should be more than enough for everyone
4126          */
4127         if (alloc_size > PAGE_SIZE)
4128                 return -ENOMEM;
4129 
4130         space_args.total_spaces = 0;
4131         dest = kmalloc(alloc_size, GFP_KERNEL);
4132         if (!dest)
4133                 return -ENOMEM;
4134         dest_orig = dest;
4135 
4136         /* now we have a buffer to copy into */
4137         for (i = 0; i < num_types; i++) {
4138                 struct btrfs_space_info *tmp;
4139 
4140                 if (!slot_count)
4141                         break;
4142 
4143                 info = NULL;
4144                 rcu_read_lock();
4145                 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4146                                         list) {
4147                         if (tmp->flags == types[i]) {
4148                                 info = tmp;
4149                                 break;
4150                         }
4151                 }
4152                 rcu_read_unlock();
4153 
4154                 if (!info)
4155                         continue;
4156                 down_read(&info->groups_sem);
4157                 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4158                         if (!list_empty(&info->block_groups[c])) {
4159                                 btrfs_get_block_group_info(
4160                                         &info->block_groups[c], &space);
4161                                 memcpy(dest, &space, sizeof(space));
4162                                 dest++;
4163                                 space_args.total_spaces++;
4164                                 slot_count--;
4165                         }
4166                         if (!slot_count)
4167                                 break;
4168                 }
4169                 up_read(&info->groups_sem);
4170         }
4171 
4172         /*
4173          * Add global block reserve
4174          */
4175         if (slot_count) {
4176                 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4177 
4178                 spin_lock(&block_rsv->lock);
4179                 space.total_bytes = block_rsv->size;
4180                 space.used_bytes = block_rsv->size - block_rsv->reserved;
4181                 spin_unlock(&block_rsv->lock);
4182                 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4183                 memcpy(dest, &space, sizeof(space));
4184                 space_args.total_spaces++;
4185         }
4186 
4187         user_dest = (struct btrfs_ioctl_space_info __user *)
4188                 (arg + sizeof(struct btrfs_ioctl_space_args));
4189 
4190         if (copy_to_user(user_dest, dest_orig, alloc_size))
4191                 ret = -EFAULT;
4192 
4193         kfree(dest_orig);
4194 out:
4195         if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4196                 ret = -EFAULT;
4197 
4198         return ret;
4199 }
4200 
4201 /*
4202  * there are many ways the trans_start and trans_end ioctls can lead
4203  * to deadlocks.  They should only be used by applications that
4204  * basically own the machine, and have a very in depth understanding
4205  * of all the possible deadlocks and enospc problems.
4206  */
4207 long btrfs_ioctl_trans_end(struct file *file)
4208 {
4209         struct inode *inode = file_inode(file);
4210         struct btrfs_root *root = BTRFS_I(inode)->root;
4211         struct btrfs_trans_handle *trans;
4212 
4213         trans = file->private_data;
4214         if (!trans)
4215                 return -EINVAL;
4216         file->private_data = NULL;
4217 
4218         btrfs_end_transaction(trans, root);
4219 
4220         atomic_dec(&root->fs_info->open_ioctl_trans);
4221 
4222         mnt_drop_write_file(file);
4223         return 0;
4224 }
4225 
4226 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4227                                             void __user *argp)
4228 {
4229         struct btrfs_trans_handle *trans;
4230         u64 transid;
4231         int ret;
4232 
4233         trans = btrfs_attach_transaction_barrier(root);
4234         if (IS_ERR(trans)) {
4235                 if (PTR_ERR(trans) != -ENOENT)
4236                         return PTR_ERR(trans);
4237 
4238                 /* No running transaction, don't bother */
4239                 transid = root->fs_info->last_trans_committed;
4240                 goto out;
4241         }
4242         transid = trans->transid;
4243         ret = btrfs_commit_transaction_async(trans, root, 0);
4244         if (ret) {
4245                 btrfs_end_transaction(trans, root);
4246                 return ret;
4247         }
4248 out:
4249         if (argp)
4250                 if (copy_to_user(argp, &transid, sizeof(transid)))
4251                         return -EFAULT;
4252         return 0;
4253 }
4254 
4255 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4256                                            void __user *argp)
4257 {
4258         u64 transid;
4259 
4260         if (argp) {
4261                 if (copy_from_user(&transid, argp, sizeof(transid)))
4262                         return -EFAULT;
4263         } else {
4264                 transid = 0;  /* current trans */
4265         }
4266         return btrfs_wait_for_commit(root, transid);
4267 }
4268 
4269 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4270 {
4271         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4272         struct btrfs_ioctl_scrub_args *sa;
4273         int ret;
4274 
4275         if (!capable(CAP_SYS_ADMIN))
4276                 return -EPERM;
4277 
4278         sa = memdup_user(arg, sizeof(*sa));
4279         if (IS_ERR(sa))
4280                 return PTR_ERR(sa);
4281 
4282         if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4283                 ret = mnt_want_write_file(file);
4284                 if (ret)
4285                         goto out;
4286         }
4287 
4288         ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4289                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4290                               0);
4291 
4292         if (copy_to_user(arg, sa, sizeof(*sa)))
4293                 ret = -EFAULT;
4294 
4295         if (!(sa->flags & BTRFS_SCRUB_READONLY))
4296                 mnt_drop_write_file(file);
4297 out:
4298         kfree(sa);
4299         return ret;
4300 }
4301 
4302 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4303 {
4304         if (!capable(CAP_SYS_ADMIN))
4305                 return -EPERM;
4306 
4307         return btrfs_scrub_cancel(root->fs_info);
4308 }
4309 
4310 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4311                                        void __user *arg)
4312 {
4313         struct btrfs_ioctl_scrub_args *sa;
4314         int ret;
4315 
4316         if (!capable(CAP_SYS_ADMIN))
4317                 return -EPERM;
4318 
4319         sa = memdup_user(arg, sizeof(*sa));
4320         if (IS_ERR(sa))
4321                 return PTR_ERR(sa);
4322 
4323         ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4324 
4325         if (copy_to_user(arg, sa, sizeof(*sa)))
4326                 ret = -EFAULT;
4327 
4328         kfree(sa);
4329         return ret;
4330 }
4331 
4332 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4333                                       void __user *arg)
4334 {
4335         struct btrfs_ioctl_get_dev_stats *sa;
4336         int ret;
4337 
4338         sa = memdup_user(arg, sizeof(*sa));
4339         if (IS_ERR(sa))
4340                 return PTR_ERR(sa);
4341 
4342         if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4343                 kfree(sa);
4344                 return -EPERM;
4345         }
4346 
4347         ret = btrfs_get_dev_stats(root, sa);
4348 
4349         if (copy_to_user(arg, sa, sizeof(*sa)))
4350                 ret = -EFAULT;
4351 
4352         kfree(sa);
4353         return ret;
4354 }
4355 
4356 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4357 {
4358         struct btrfs_ioctl_dev_replace_args *p;
4359         int ret;
4360 
4361         if (!capable(CAP_SYS_ADMIN))
4362                 return -EPERM;
4363 
4364         p = memdup_user(arg, sizeof(*p));
4365         if (IS_ERR(p))
4366                 return PTR_ERR(p);
4367 
4368         switch (p->cmd) {
4369         case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4370                 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4371                         ret = -EROFS;
4372                         goto out;
4373                 }
4374                 if (atomic_xchg(
4375                         &root->fs_info->mutually_exclusive_operation_running,
4376                         1)) {
4377                         ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4378                 } else {
4379                         ret = btrfs_dev_replace_start(root, p);
4380                         atomic_set(
4381                          &root->fs_info->mutually_exclusive_operation_running,
4382                          0);
4383                 }
4384                 break;
4385         case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4386                 btrfs_dev_replace_status(root->fs_info, p);
4387                 ret = 0;
4388                 break;
4389         case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4390                 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4391                 break;
4392         default:
4393                 ret = -EINVAL;
4394                 break;
4395         }
4396 
4397         if (copy_to_user(arg, p, sizeof(*p)))
4398                 ret = -EFAULT;
4399 out:
4400         kfree(p);
4401         return ret;
4402 }
4403 
4404 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4405 {
4406         int ret = 0;
4407         int i;
4408         u64 rel_ptr;
4409         int size;
4410         struct btrfs_ioctl_ino_path_args *ipa = NULL;
4411         struct inode_fs_paths *ipath = NULL;
4412         struct btrfs_path *path;
4413 
4414         if (!capable(CAP_DAC_READ_SEARCH))
4415                 return -EPERM;
4416 
4417         path = btrfs_alloc_path();
4418         if (!path) {
4419                 ret = -ENOMEM;
4420                 goto out;
4421         }
4422 
4423         ipa = memdup_user(arg, sizeof(*ipa));
4424         if (IS_ERR(ipa)) {
4425                 ret = PTR_ERR(ipa);
4426                 ipa = NULL;
4427                 goto out;
4428         }
4429 
4430         size = min_t(u32, ipa->size, 4096);
4431         ipath = init_ipath(size, root, path);
4432         if (IS_ERR(ipath)) {
4433                 ret = PTR_ERR(ipath);
4434                 ipath = NULL;
4435                 goto out;
4436         }
4437 
4438         ret = paths_from_inode(ipa->inum, ipath);
4439         if (ret < 0)
4440                 goto out;
4441 
4442         for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4443                 rel_ptr = ipath->fspath->val[i] -
4444                           (u64)(unsigned long)ipath->fspath->val;
4445                 ipath->fspath->val[i] = rel_ptr;
4446         }
4447 
4448         ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4449                            (void *)(unsigned long)ipath->fspath, size);
4450         if (ret) {
4451                 ret = -EFAULT;
4452                 goto out;
4453         }
4454 
4455 out:
4456         btrfs_free_path(path);
4457         free_ipath(ipath);
4458         kfree(ipa);
4459 
4460         return ret;
4461 }
4462 
4463 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4464 {
4465         struct btrfs_data_container *inodes = ctx;
4466         const size_t c = 3 * sizeof(u64);
4467 
4468         if (inodes->bytes_left >= c) {
4469                 inodes->bytes_left -= c;
4470                 inodes->val[inodes->elem_cnt] = inum;
4471                 inodes->val[inodes->elem_cnt + 1] = offset;
4472                 inodes->val[inodes->elem_cnt + 2] = root;
4473                 inodes->elem_cnt += 3;
4474         } else {
4475                 inodes->bytes_missing += c - inodes->bytes_left;
4476                 inodes->bytes_left = 0;
4477                 inodes->elem_missed += 3;
4478         }
4479 
4480         return 0;
4481 }
4482 
4483 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4484                                         void __user *arg)
4485 {
4486         int ret = 0;
4487         int size;
4488         struct btrfs_ioctl_logical_ino_args *loi;
4489         struct btrfs_data_container *inodes = NULL;
4490         struct btrfs_path *path = NULL;
4491 
4492         if (!capable(CAP_SYS_ADMIN))
4493                 return -EPERM;
4494 
4495         loi = memdup_user(arg, sizeof(*loi));
4496         if (IS_ERR(loi)) {
4497                 ret = PTR_ERR(loi);
4498                 loi = NULL;
4499                 goto out;
4500         }
4501 
4502         path = btrfs_alloc_path();
4503         if (!path) {
4504                 ret = -ENOMEM;
4505                 goto out;
4506         }
4507 
4508         size = min_t(u32, loi->size, SZ_64K);
4509         inodes = init_data_container(size);
4510         if (IS_ERR(inodes)) {
4511                 ret = PTR_ERR(inodes);
4512                 inodes = NULL;
4513                 goto out;
4514         }
4515 
4516         ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4517                                           build_ino_list, inodes);
4518         if (ret == -EINVAL)
4519                 ret = -ENOENT;
4520         if (ret < 0)
4521                 goto out;
4522 
4523         ret = copy_to_user((void *)(unsigned long)loi->inodes,
4524                            (void *)(unsigned long)inodes, size);
4525         if (ret)
4526                 ret = -EFAULT;
4527 
4528 out:
4529         btrfs_free_path(path);
4530         vfree(inodes);
4531         kfree(loi);
4532 
4533         return ret;
4534 }
4535 
4536 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4537                                struct btrfs_ioctl_balance_args *bargs)
4538 {
4539         struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4540 
4541         bargs->flags = bctl->flags;
4542 
4543         if (atomic_read(&fs_info->balance_running))
4544                 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4545         if (atomic_read(&fs_info->balance_pause_req))
4546                 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4547         if (atomic_read(&fs_info->balance_cancel_req))
4548                 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4549 
4550         memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4551         memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4552         memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4553 
4554         if (lock) {
4555                 spin_lock(&fs_info->balance_lock);
4556                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4557                 spin_unlock(&fs_info->balance_lock);
4558         } else {
4559                 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4560         }
4561 }
4562 
4563 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4564 {
4565         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4566         struct btrfs_fs_info *fs_info = root->fs_info;
4567         struct btrfs_ioctl_balance_args *bargs;
4568         struct btrfs_balance_control *bctl;
4569         bool need_unlock; /* for mut. excl. ops lock */
4570         int ret;
4571 
4572         if (!capable(CAP_SYS_ADMIN))
4573                 return -EPERM;
4574 
4575         ret = mnt_want_write_file(file);
4576         if (ret)
4577                 return ret;
4578 
4579 again:
4580         if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4581                 mutex_lock(&fs_info->volume_mutex);
4582                 mutex_lock(&fs_info->balance_mutex);
4583                 need_unlock = true;
4584                 goto locked;
4585         }
4586 
4587         /*
4588          * mut. excl. ops lock is locked.  Three possibilites:
4589          *   (1) some other op is running
4590          *   (2) balance is running
4591          *   (3) balance is paused -- special case (think resume)
4592          */
4593         mutex_lock(&fs_info->balance_mutex);
4594         if (fs_info->balance_ctl) {
4595                 /* this is either (2) or (3) */
4596                 if (!atomic_read(&fs_info->balance_running)) {
4597                         mutex_unlock(&fs_info->balance_mutex);
4598                         if (!mutex_trylock(&fs_info->volume_mutex))
4599                                 goto again;
4600                         mutex_lock(&fs_info->balance_mutex);
4601 
4602                         if (fs_info->balance_ctl &&
4603                             !atomic_read(&fs_info->balance_running)) {
4604                                 /* this is (3) */
4605                                 need_unlock = false;
4606                                 goto locked;
4607                         }
4608 
4609                         mutex_unlock(&fs_info->balance_mutex);
4610                         mutex_unlock(&fs_info->volume_mutex);
4611                         goto again;
4612                 } else {
4613                         /* this is (2) */
4614                         mutex_unlock(&fs_info->balance_mutex);
4615                         ret = -EINPROGRESS;
4616                         goto out;
4617                 }
4618         } else {
4619                 /* this is (1) */
4620                 mutex_unlock(&fs_info->balance_mutex);
4621                 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4622                 goto out;
4623         }
4624 
4625 locked:
4626         BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4627 
4628         if (arg) {
4629                 bargs = memdup_user(arg, sizeof(*bargs));
4630                 if (IS_ERR(bargs)) {
4631                         ret = PTR_ERR(bargs);
4632                         goto out_unlock;
4633                 }
4634 
4635                 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4636                         if (!fs_info->balance_ctl) {
4637                                 ret = -ENOTCONN;
4638                                 goto out_bargs;
4639                         }
4640 
4641                         bctl = fs_info->balance_ctl;
4642                         spin_lock(&fs_info->balance_lock);
4643                         bctl->flags |= BTRFS_BALANCE_RESUME;
4644                         spin_unlock(&fs_info->balance_lock);
4645 
4646                         goto do_balance;
4647                 }
4648         } else {
4649                 bargs = NULL;
4650         }
4651 
4652         if (fs_info->balance_ctl) {
4653                 ret = -EINPROGRESS;
4654                 goto out_bargs;
4655         }
4656 
4657         bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4658         if (!bctl) {
4659                 ret = -ENOMEM;
4660                 goto out_bargs;
4661         }
4662 
4663         bctl->fs_info = fs_info;
4664         if (arg) {
4665                 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4666                 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4667                 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4668 
4669                 bctl->flags = bargs->flags;
4670         } else {
4671                 /* balance everything - no filters */
4672                 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4673         }
4674 
4675         if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4676                 ret = -EINVAL;
4677                 goto out_bctl;
4678         }
4679 
4680 do_balance:
4681         /*
4682          * Ownership of bctl and mutually_exclusive_operation_running
4683          * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
4684          * or, if restriper was paused all the way until unmount, in
4685          * free_fs_info.  mutually_exclusive_operation_running is
4686          * cleared in __cancel_balance.
4687          */
4688         need_unlock = false;
4689 
4690         ret = btrfs_balance(bctl, bargs);
4691         bctl = NULL;
4692 
4693         if (arg) {
4694                 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4695                         ret = -EFAULT;
4696         }
4697 
4698 out_bctl:
4699         kfree(bctl);
4700 out_bargs:
4701         kfree(bargs);
4702 out_unlock:
4703         mutex_unlock(&fs_info->balance_mutex);
4704         mutex_unlock(&fs_info->volume_mutex);
4705         if (need_unlock)
4706                 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4707 out:
4708         mnt_drop_write_file(file);
4709         return ret;
4710 }
4711 
4712 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4713 {
4714         if (!capable(CAP_SYS_ADMIN))
4715                 return -EPERM;
4716 
4717         switch (cmd) {
4718         case BTRFS_BALANCE_CTL_PAUSE:
4719                 return btrfs_pause_balance(root->fs_info);
4720         case BTRFS_BALANCE_CTL_CANCEL:
4721                 return btrfs_cancel_balance(root->fs_info);
4722         }
4723 
4724         return -EINVAL;
4725 }
4726 
4727 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4728                                          void __user *arg)
4729 {
4730         struct btrfs_fs_info *fs_info = root->fs_info;
4731         struct btrfs_ioctl_balance_args *bargs;
4732         int ret = 0;
4733 
4734         if (!capable(CAP_SYS_ADMIN))
4735                 return -EPERM;
4736 
4737         mutex_lock(&fs_info->balance_mutex);
4738         if (!fs_info->balance_ctl) {
4739                 ret = -ENOTCONN;
4740                 goto out;
4741         }
4742 
4743         bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4744         if (!bargs) {
4745                 ret = -ENOMEM;
4746                 goto out;
4747         }
4748 
4749         update_ioctl_balance_args(fs_info, 1, bargs);
4750 
4751         if (copy_to_user(arg, bargs, sizeof(*bargs)))
4752                 ret = -EFAULT;
4753 
4754         kfree(bargs);
4755 out:
4756         mutex_unlock(&fs_info->balance_mutex);
4757         return ret;
4758 }
4759 
4760 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4761 {
4762         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4763         struct btrfs_ioctl_quota_ctl_args *sa;
4764         struct btrfs_trans_handle *trans = NULL;
4765         int ret;
4766         int err;
4767 
4768         if (!capable(CAP_SYS_ADMIN))
4769                 return -EPERM;
4770 
4771         ret = mnt_want_write_file(file);
4772         if (ret)
4773                 return ret;
4774 
4775         sa = memdup_user(arg, sizeof(*sa));
4776         if (IS_ERR(sa)) {
4777                 ret = PTR_ERR(sa);
4778                 goto drop_write;
4779         }
4780 
4781         down_write(&root->fs_info->subvol_sem);
4782         trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4783         if (IS_ERR(trans)) {
4784                 ret = PTR_ERR(trans);
4785                 goto out;
4786         }
4787 
4788         switch (sa->cmd) {
4789         case BTRFS_QUOTA_CTL_ENABLE:
4790                 ret = btrfs_quota_enable(trans, root->fs_info);
4791                 break;
4792         case BTRFS_QUOTA_CTL_DISABLE:
4793                 ret = btrfs_quota_disable(trans, root->fs_info);
4794                 break;
4795         default:
4796                 ret = -EINVAL;
4797                 break;
4798         }
4799 
4800         err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4801         if (err && !ret)
4802                 ret = err;
4803 out:
4804         kfree(sa);
4805         up_write(&root->fs_info->subvol_sem);
4806 drop_write:
4807         mnt_drop_write_file(file);
4808         return ret;
4809 }
4810 
4811 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4812 {
4813         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4814         struct btrfs_ioctl_qgroup_assign_args *sa;
4815         struct btrfs_trans_handle *trans;
4816         int ret;
4817         int err;
4818 
4819         if (!capable(CAP_SYS_ADMIN))
4820                 return -EPERM;
4821 
4822         ret = mnt_want_write_file(file);
4823         if (ret)
4824                 return ret;
4825 
4826         sa = memdup_user(arg, sizeof(*sa));
4827         if (IS_ERR(sa)) {
4828                 ret = PTR_ERR(sa);
4829                 goto drop_write;
4830         }
4831 
4832         trans = btrfs_join_transaction(root);
4833         if (IS_ERR(trans)) {
4834                 ret = PTR_ERR(trans);
4835                 goto out;
4836         }
4837 
4838         /* FIXME: check if the IDs really exist */
4839         if (sa->assign) {
4840                 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4841                                                 sa->src, sa->dst);
4842         } else {
4843                 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4844                                                 sa->src, sa->dst);
4845         }
4846 
4847         /* update qgroup status and info */
4848         err = btrfs_run_qgroups(trans, root->fs_info);
4849         if (err < 0)
4850                 btrfs_std_error(root->fs_info, ret,
4851                             "failed to update qgroup status and info\n");
4852         err = btrfs_end_transaction(trans, root);
4853         if (err && !ret)
4854                 ret = err;
4855 
4856 out:
4857         kfree(sa);
4858 drop_write:
4859         mnt_drop_write_file(file);
4860         return ret;
4861 }
4862 
4863 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4864 {
4865         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4866         struct btrfs_ioctl_qgroup_create_args *sa;
4867         struct btrfs_trans_handle *trans;
4868         int ret;
4869         int err;
4870 
4871         if (!capable(CAP_SYS_ADMIN))
4872                 return -EPERM;
4873 
4874         ret = mnt_want_write_file(file);
4875         if (ret)
4876                 return ret;
4877 
4878         sa = memdup_user(arg, sizeof(*sa));
4879         if (IS_ERR(sa)) {
4880                 ret = PTR_ERR(sa);
4881                 goto drop_write;
4882         }
4883 
4884         if (!sa->qgroupid) {
4885                 ret = -EINVAL;
4886                 goto out;
4887         }
4888 
4889         trans = btrfs_join_transaction(root);
4890         if (IS_ERR(trans)) {
4891                 ret = PTR_ERR(trans);
4892                 goto out;
4893         }
4894 
4895         /* FIXME: check if the IDs really exist */
4896         if (sa->create) {
4897                 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4898         } else {
4899                 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4900         }
4901 
4902         err = btrfs_end_transaction(trans, root);
4903         if (err && !ret)
4904                 ret = err;
4905 
4906 out:
4907         kfree(sa);
4908 drop_write:
4909         mnt_drop_write_file(file);
4910         return ret;
4911 }
4912 
4913 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4914 {
4915         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4916         struct btrfs_ioctl_qgroup_limit_args *sa;
4917         struct btrfs_trans_handle *trans;
4918         int ret;
4919         int err;
4920         u64 qgroupid;
4921 
4922         if (!capable(CAP_SYS_ADMIN))
4923                 return -EPERM;
4924 
4925         ret = mnt_want_write_file(file);
4926         if (ret)
4927                 return ret;
4928 
4929         sa = memdup_user(arg, sizeof(*sa));
4930         if (IS_ERR(sa)) {
4931                 ret = PTR_ERR(sa);
4932                 goto drop_write;
4933         }
4934 
4935         trans = btrfs_join_transaction(root);
4936         if (IS_ERR(trans)) {
4937                 ret = PTR_ERR(trans);
4938                 goto out;
4939         }
4940 
4941         qgroupid = sa->qgroupid;
4942         if (!qgroupid) {
4943                 /* take the current subvol as qgroup */
4944                 qgroupid = root->root_key.objectid;
4945         }
4946 
4947         /* FIXME: check if the IDs really exist */
4948         ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4949 
4950         err = btrfs_end_transaction(trans, root);
4951         if (err && !ret)
4952                 ret = err;
4953 
4954 out:
4955         kfree(sa);
4956 drop_write:
4957         mnt_drop_write_file(file);
4958         return ret;
4959 }
4960 
4961 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4962 {
4963         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4964         struct btrfs_ioctl_quota_rescan_args *qsa;
4965         int ret;
4966 
4967         if (!capable(CAP_SYS_ADMIN))
4968                 return -EPERM;
4969 
4970         ret = mnt_want_write_file(file);
4971         if (ret)
4972                 return ret;
4973 
4974         qsa = memdup_user(arg, sizeof(*qsa));
4975         if (IS_ERR(qsa)) {
4976                 ret = PTR_ERR(qsa);
4977                 goto drop_write;
4978         }
4979 
4980         if (qsa->flags) {
4981                 ret = -EINVAL;
4982                 goto out;
4983         }
4984 
4985         ret = btrfs_qgroup_rescan(root->fs_info);
4986 
4987 out:
4988         kfree(qsa);
4989 drop_write:
4990         mnt_drop_write_file(file);
4991         return ret;
4992 }
4993 
4994 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4995 {
4996         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4997         struct btrfs_ioctl_quota_rescan_args *qsa;
4998         int ret = 0;
4999 
5000         if (!capable(CAP_SYS_ADMIN))
5001                 return -EPERM;
5002 
5003         qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5004         if (!qsa)
5005                 return -ENOMEM;
5006 
5007         if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5008                 qsa->flags = 1;
5009                 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
5010         }
5011 
5012         if (copy_to_user(arg, qsa, sizeof(*qsa)))
5013                 ret = -EFAULT;
5014 
5015         kfree(qsa);
5016         return ret;
5017 }
5018 
5019 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5020 {
5021         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5022 
5023         if (!capable(CAP_SYS_ADMIN))
5024                 return -EPERM;
5025 
5026         return btrfs_qgroup_wait_for_completion(root->fs_info);
5027 }
5028 
5029 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5030                                             struct btrfs_ioctl_received_subvol_args *sa)
5031 {
5032         struct inode *inode = file_inode(file);
5033         struct btrfs_root *root = BTRFS_I(inode)->root;
5034         struct btrfs_root_item *root_item = &root->root_item;
5035         struct btrfs_trans_handle *trans;
5036         struct timespec ct = current_fs_time(inode->i_sb);
5037         int ret = 0;
5038         int received_uuid_changed;
5039 
5040         if (!inode_owner_or_capable(inode))
5041                 return -EPERM;
5042 
5043         ret = mnt_want_write_file(file);
5044         if (ret < 0)
5045                 return ret;
5046 
5047         down_write(&root->fs_info->subvol_sem);
5048 
5049         if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5050                 ret = -EINVAL;
5051                 goto out;
5052         }
5053 
5054         if (btrfs_root_readonly(root)) {
5055                 ret = -EROFS;
5056                 goto out;
5057         }
5058 
5059         /*
5060          * 1 - root item
5061          * 2 - uuid items (received uuid + subvol uuid)
5062          */
5063         trans = btrfs_start_transaction(root, 3);
5064         if (IS_ERR(trans)) {
5065                 ret = PTR_ERR(trans);
5066                 trans = NULL;
5067                 goto out;
5068         }
5069 
5070         sa->rtransid = trans->transid;
5071         sa->rtime.sec = ct.tv_sec;
5072         sa->rtime.nsec = ct.tv_nsec;
5073 
5074         received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5075                                        BTRFS_UUID_SIZE);
5076         if (received_uuid_changed &&
5077             !btrfs_is_empty_uuid(root_item->received_uuid))
5078                 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5079                                     root_item->received_uuid,
5080                                     BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5081                                     root->root_key.objectid);
5082         memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5083         btrfs_set_root_stransid(root_item, sa->stransid);
5084         btrfs_set_root_rtransid(root_item, sa->rtransid);
5085         btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5086         btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5087         btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5088         btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5089 
5090         ret = btrfs_update_root(trans, root->fs_info->tree_root,
5091                                 &root->root_key, &root->root_item);
5092         if (ret < 0) {
5093                 btrfs_end_transaction(trans, root);
5094                 goto out;
5095         }
5096         if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5097                 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5098                                           sa->uuid,
5099                                           BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5100                                           root->root_key.objectid);
5101                 if (ret < 0 && ret != -EEXIST) {
5102                         btrfs_abort_transaction(trans, root, ret);
5103                         goto out;
5104                 }
5105         }
5106         ret = btrfs_commit_transaction(trans, root);
5107         if (ret < 0) {
5108                 btrfs_abort_transaction(trans, root, ret);
5109                 goto out;
5110         }
5111 
5112 out:
5113         up_write(&root->fs_info->subvol_sem);
5114         mnt_drop_write_file(file);
5115         return ret;
5116 }
5117 
5118 #ifdef CONFIG_64BIT
5119 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5120                                                 void __user *arg)
5121 {
5122         struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5123         struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5124         int ret = 0;
5125 
5126         args32 = memdup_user(arg, sizeof(*args32));
5127         if (IS_ERR(args32)) {
5128                 ret = PTR_ERR(args32);
5129                 args32 = NULL;
5130                 goto out;
5131         }
5132 
5133         args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5134         if (!args64) {
5135                 ret = -ENOMEM;
5136                 goto out;
5137         }
5138 
5139         memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5140         args64->stransid = args32->stransid;
5141         args64->rtransid = args32->rtransid;
5142         args64->stime.sec = args32->stime.sec;
5143         args64->stime.nsec = args32->stime.nsec;
5144         args64->rtime.sec = args32->rtime.sec;
5145         args64->rtime.nsec = args32->rtime.nsec;
5146         args64->flags = args32->flags;
5147 
5148         ret = _btrfs_ioctl_set_received_subvol(file, args64);
5149         if (ret)
5150                 goto out;
5151 
5152         memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5153         args32->stransid = args64->stransid;
5154         args32->rtransid = args64->rtransid;
5155         args32->stime.sec = args64->stime.sec;
5156         args32->stime.nsec = args64->stime.nsec;
5157         args32->rtime.sec = args64->rtime.sec;
5158         args32->rtime.nsec = args64->rtime.nsec;
5159         args32->flags = args64->flags;
5160 
5161         ret = copy_to_user(arg, args32, sizeof(*args32));
5162         if (ret)
5163                 ret = -EFAULT;
5164 
5165 out:
5166         kfree(args32);
5167         kfree(args64);
5168         return ret;
5169 }
5170 #endif
5171 
5172 static long btrfs_ioctl_set_received_subvol(struct file *file,
5173                                             void __user *arg)
5174 {
5175         struct btrfs_ioctl_received_subvol_args *sa = NULL;
5176         int ret = 0;
5177 
5178         sa = memdup_user(arg, sizeof(*sa));
5179         if (IS_ERR(sa)) {
5180                 ret = PTR_ERR(sa);
5181                 sa = NULL;
5182                 goto out;
5183         }
5184 
5185         ret = _btrfs_ioctl_set_received_subvol(file, sa);
5186 
5187         if (ret)
5188                 goto out;
5189 
5190         ret = copy_to_user(arg, sa, sizeof(*sa));
5191         if (ret)
5192                 ret = -EFAULT;
5193 
5194 out:
5195         kfree(sa);
5196         return ret;
5197 }
5198 
5199 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5200 {
5201         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5202         size_t len;
5203         int ret;
5204         char label[BTRFS_LABEL_SIZE];
5205 
5206         spin_lock(&root->fs_info->super_lock);
5207         memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5208         spin_unlock(&root->fs_info->super_lock);
5209 
5210         len = strnlen(label, BTRFS_LABEL_SIZE);
5211 
5212         if (len == BTRFS_LABEL_SIZE) {
5213                 btrfs_warn(root->fs_info,
5214                         "label is too long, return the first %zu bytes", --len);
5215         }
5216 
5217         ret = copy_to_user(arg, label, len);
5218 
5219         return ret ? -EFAULT : 0;
5220 }
5221 
5222 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5223 {
5224         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5225         struct btrfs_super_block *super_block = root->fs_info->super_copy;
5226         struct btrfs_trans_handle *trans;
5227         char label[BTRFS_LABEL_SIZE];
5228         int ret;
5229 
5230         if (!capable(CAP_SYS_ADMIN))
5231                 return -EPERM;
5232 
5233         if (copy_from_user(label, arg, sizeof(label)))
5234                 return -EFAULT;
5235 
5236         if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5237                 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5238                        BTRFS_LABEL_SIZE - 1);
5239                 return -EINVAL;
5240         }
5241 
5242         ret = mnt_want_write_file(file);
5243         if (ret)
5244                 return ret;
5245 
5246         trans = btrfs_start_transaction(root, 0);
5247         if (IS_ERR(trans)) {
5248                 ret = PTR_ERR(trans);
5249                 goto out_unlock;
5250         }
5251 
5252         spin_lock(&root->fs_info->super_lock);
5253         strcpy(super_block->label, label);
5254         spin_unlock(&root->fs_info->super_lock);
5255         ret = btrfs_commit_transaction(trans, root);
5256 
5257 out_unlock:
5258         mnt_drop_write_file(file);
5259         return ret;
5260 }
5261 
5262 #define INIT_FEATURE_FLAGS(suffix) \
5263         { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5264           .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5265           .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5266 
5267 int btrfs_ioctl_get_supported_features(void __user *arg)
5268 {
5269         static const struct btrfs_ioctl_feature_flags features[3] = {
5270                 INIT_FEATURE_FLAGS(SUPP),
5271                 INIT_FEATURE_FLAGS(SAFE_SET),
5272                 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5273         };
5274 
5275         if (copy_to_user(arg, &features, sizeof(features)))
5276                 return -EFAULT;
5277 
5278         return 0;
5279 }
5280 
5281 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5282 {
5283         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5284         struct btrfs_super_block *super_block = root->fs_info->super_copy;
5285         struct btrfs_ioctl_feature_flags features;
5286 
5287         features.compat_flags = btrfs_super_compat_flags(super_block);
5288         features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5289         features.incompat_flags = btrfs_super_incompat_flags(super_block);
5290 
5291         if (copy_to_user(arg, &features, sizeof(features)))
5292                 return -EFAULT;
5293 
5294         return 0;
5295 }
5296 
5297 static int check_feature_bits(struct btrfs_root *root,
5298                               enum btrfs_feature_set set,
5299                               u64 change_mask, u64 flags, u64 supported_flags,
5300                               u64 safe_set, u64 safe_clear)
5301 {
5302         const char *type = btrfs_feature_set_names[set];
5303         char *names;
5304         u64 disallowed, unsupported;
5305         u64 set_mask = flags & change_mask;
5306         u64 clear_mask = ~flags & change_mask;
5307 
5308         unsupported = set_mask & ~supported_flags;
5309         if (unsupported) {
5310                 names = btrfs_printable_features(set, unsupported);
5311                 if (names) {
5312                         btrfs_warn(root->fs_info,
5313                            "this kernel does not support the %s feature bit%s",
5314                            names, strchr(names, ',') ? "s" : "");
5315                         kfree(names);
5316                 } else
5317                         btrfs_warn(root->fs_info,
5318                            "this kernel does not support %s bits 0x%llx",
5319                            type, unsupported);
5320                 return -EOPNOTSUPP;
5321         }
5322 
5323         disallowed = set_mask & ~safe_set;
5324         if (disallowed) {
5325                 names = btrfs_printable_features(set, disallowed);
5326                 if (names) {
5327                         btrfs_warn(root->fs_info,
5328                            "can't set the %s feature bit%s while mounted",
5329                            names, strchr(names, ',') ? "s" : "");
5330                         kfree(names);
5331                 } else
5332                         btrfs_warn(root->fs_info,
5333                            "can't set %s bits 0x%llx while mounted",
5334                            type, disallowed);
5335                 return -EPERM;
5336         }
5337 
5338         disallowed = clear_mask & ~safe_clear;
5339         if (disallowed) {
5340                 names = btrfs_printable_features(set, disallowed);
5341                 if (names) {
5342                         btrfs_warn(root->fs_info,
5343                            "can't clear the %s feature bit%s while mounted",
5344                            names, strchr(names, ',') ? "s" : "");
5345                         kfree(names);
5346                 } else
5347                         btrfs_warn(root->fs_info,
5348                            "can't clear %s bits 0x%llx while mounted",
5349                            type, disallowed);
5350                 return -EPERM;
5351         }
5352 
5353         return 0;
5354 }
5355 
5356 #define check_feature(root, change_mask, flags, mask_base)      \
5357 check_feature_bits(root, FEAT_##mask_base, change_mask, flags,  \
5358                    BTRFS_FEATURE_ ## mask_base ## _SUPP,        \
5359                    BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,    \
5360                    BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5361 
5362 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5363 {
5364         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5365         struct btrfs_super_block *super_block = root->fs_info->super_copy;
5366         struct btrfs_ioctl_feature_flags flags[2];
5367         struct btrfs_trans_handle *trans;
5368         u64 newflags;
5369         int ret;
5370 
5371         if (!capable(CAP_SYS_ADMIN))
5372                 return -EPERM;
5373 
5374         if (copy_from_user(flags, arg, sizeof(flags)))
5375                 return -EFAULT;
5376 
5377         /* Nothing to do */
5378         if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5379             !flags[0].incompat_flags)
5380                 return 0;
5381 
5382         ret = check_feature(root, flags[0].compat_flags,
5383                             flags[1].compat_flags, COMPAT);
5384         if (ret)
5385                 return ret;
5386 
5387         ret = check_feature(root, flags[0].compat_ro_flags,
5388                             flags[1].compat_ro_flags, COMPAT_RO);
5389         if (ret)
5390                 return ret;
5391 
5392         ret = check_feature(root, flags[0].incompat_flags,
5393                             flags[1].incompat_flags, INCOMPAT);
5394         if (ret)
5395                 return ret;
5396 
5397         trans = btrfs_start_transaction(root, 0);
5398         if (IS_ERR(trans))
5399                 return PTR_ERR(trans);
5400 
5401         spin_lock(&root->fs_info->super_lock);
5402         newflags = btrfs_super_compat_flags(super_block);
5403         newflags |= flags[0].compat_flags & flags[1].compat_flags;
5404         newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5405         btrfs_set_super_compat_flags(super_block, newflags);
5406 
5407         newflags = btrfs_super_compat_ro_flags(super_block);
5408         newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5409         newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5410         btrfs_set_super_compat_ro_flags(super_block, newflags);
5411 
5412         newflags = btrfs_super_incompat_flags(super_block);
5413         newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5414         newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5415         btrfs_set_super_incompat_flags(super_block, newflags);
5416         spin_unlock(&root->fs_info->super_lock);
5417 
5418         return btrfs_commit_transaction(trans, root);
5419 }
5420 
5421 long btrfs_ioctl(struct file *file, unsigned int
5422                 cmd, unsigned long arg)
5423 {
5424         struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5425         void __user *argp = (void __user *)arg;
5426 
5427         switch (cmd) {
5428         case FS_IOC_GETFLAGS:
5429                 return btrfs_ioctl_getflags(file, argp);
5430         case FS_IOC_SETFLAGS:
5431                 return btrfs_ioctl_setflags(file, argp);
5432         case FS_IOC_GETVERSION:
5433                 return btrfs_ioctl_getversion(file, argp);
5434         case FITRIM:
5435                 return btrfs_ioctl_fitrim(file, argp);
5436         case BTRFS_IOC_SNAP_CREATE:
5437                 return btrfs_ioctl_snap_create(file, argp, 0);
5438         case BTRFS_IOC_SNAP_CREATE_V2:
5439                 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5440         case BTRFS_IOC_SUBVOL_CREATE:
5441                 return btrfs_ioctl_snap_create(file, argp, 1);
5442         case BTRFS_IOC_SUBVOL_CREATE_V2:
5443                 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5444         case BTRFS_IOC_SNAP_DESTROY:
5445                 return btrfs_ioctl_snap_destroy(file, argp);
5446         case BTRFS_IOC_SUBVOL_GETFLAGS:
5447                 return btrfs_ioctl_subvol_getflags(file, argp);
5448         case BTRFS_IOC_SUBVOL_SETFLAGS:
5449                 return btrfs_ioctl_subvol_setflags(file, argp);
5450         case BTRFS_IOC_DEFAULT_SUBVOL:
5451                 return btrfs_ioctl_default_subvol(file, argp);
5452         case BTRFS_IOC_DEFRAG:
5453                 return btrfs_ioctl_defrag(file, NULL);
5454         case BTRFS_IOC_DEFRAG_RANGE:
5455                 return btrfs_ioctl_defrag(file, argp);
5456         case BTRFS_IOC_RESIZE:
5457                 return btrfs_ioctl_resize(file, argp);
5458         case BTRFS_IOC_ADD_DEV:
5459                 return btrfs_ioctl_add_dev(root, argp);
5460         case BTRFS_IOC_RM_DEV:
5461                 return btrfs_ioctl_rm_dev(file, argp);
5462         case BTRFS_IOC_FS_INFO:
5463                 return btrfs_ioctl_fs_info(root, argp);
5464         case BTRFS_IOC_DEV_INFO:
5465                 return btrfs_ioctl_dev_info(root, argp);
5466         case BTRFS_IOC_BALANCE:
5467                 return btrfs_ioctl_balance(file, NULL);
5468         case BTRFS_IOC_TRANS_START:
5469                 return btrfs_ioctl_trans_start(file);
5470         case BTRFS_IOC_TRANS_END:
5471                 return btrfs_ioctl_trans_end(file);
5472         case BTRFS_IOC_TREE_SEARCH:
5473                 return btrfs_ioctl_tree_search(file, argp);
5474         case BTRFS_IOC_TREE_SEARCH_V2:
5475                 return btrfs_ioctl_tree_search_v2(file, argp);
5476         case BTRFS_IOC_INO_LOOKUP:
5477                 return btrfs_ioctl_ino_lookup(file, argp);
5478         case BTRFS_IOC_INO_PATHS:
5479                 return btrfs_ioctl_ino_to_path(root, argp);
5480         case BTRFS_IOC_LOGICAL_INO:
5481                 return btrfs_ioctl_logical_to_ino(root, argp);
5482         case BTRFS_IOC_SPACE_INFO:
5483                 return btrfs_ioctl_space_info(root, argp);
5484         case BTRFS_IOC_SYNC: {
5485                 int ret;
5486 
5487                 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5488                 if (ret)
5489                         return ret;
5490                 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5491                 /*
5492                  * The transaction thread may want to do more work,
5493                  * namely it pokes the cleaner ktread that will start
5494                  * processing uncleaned subvols.
5495                  */
5496                 wake_up_process(root->fs_info->transaction_kthread);
5497                 return ret;
5498         }
5499         case BTRFS_IOC_START_SYNC:
5500                 return btrfs_ioctl_start_sync(root, argp);
5501         case BTRFS_IOC_WAIT_SYNC:
5502                 return btrfs_ioctl_wait_sync(root, argp);
5503         case BTRFS_IOC_SCRUB:
5504                 return btrfs_ioctl_scrub(file, argp);
5505         case BTRFS_IOC_SCRUB_CANCEL:
5506                 return btrfs_ioctl_scrub_cancel(root, argp);
5507         case BTRFS_IOC_SCRUB_PROGRESS:
5508                 return btrfs_ioctl_scrub_progress(root, argp);
5509         case BTRFS_IOC_BALANCE_V2:
5510                 return btrfs_ioctl_balance(file, argp);
5511         case BTRFS_IOC_BALANCE_CTL:
5512                 return btrfs_ioctl_balance_ctl(root, arg);
5513         case BTRFS_IOC_BALANCE_PROGRESS:
5514                 return btrfs_ioctl_balance_progress(root, argp);
5515         case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5516                 return btrfs_ioctl_set_received_subvol(file, argp);
5517 #ifdef CONFIG_64BIT
5518         case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5519                 return btrfs_ioctl_set_received_subvol_32(file, argp);
5520 #endif
5521         case BTRFS_IOC_SEND:
5522                 return btrfs_ioctl_send(file, argp);
5523         case BTRFS_IOC_GET_DEV_STATS:
5524                 return btrfs_ioctl_get_dev_stats(root, argp);
5525         case BTRFS_IOC_QUOTA_CTL:
5526                 return btrfs_ioctl_quota_ctl(file, argp);
5527         case BTRFS_IOC_QGROUP_ASSIGN:
5528                 return btrfs_ioctl_qgroup_assign(file, argp);
5529         case BTRFS_IOC_QGROUP_CREATE:
5530                 return btrfs_ioctl_qgroup_create(file, argp);
5531         case BTRFS_IOC_QGROUP_LIMIT:
5532                 return btrfs_ioctl_qgroup_limit(file, argp);
5533         case BTRFS_IOC_QUOTA_RESCAN:
5534                 return btrfs_ioctl_quota_rescan(file, argp);
5535         case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5536                 return btrfs_ioctl_quota_rescan_status(file, argp);
5537         case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5538                 return btrfs_ioctl_quota_rescan_wait(file, argp);
5539         case BTRFS_IOC_DEV_REPLACE:
5540                 return btrfs_ioctl_dev_replace(root, argp);
5541         case BTRFS_IOC_GET_FSLABEL:
5542                 return btrfs_ioctl_get_fslabel(file, argp);
5543         case BTRFS_IOC_SET_FSLABEL:
5544                 return btrfs_ioctl_set_fslabel(file, argp);
5545         case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5546                 return btrfs_ioctl_get_supported_features(argp);
5547         case BTRFS_IOC_GET_FEATURES:
5548                 return btrfs_ioctl_get_features(file, argp);
5549         case BTRFS_IOC_SET_FEATURES:
5550                 return btrfs_ioctl_set_features(file, argp);
5551         }
5552 
5553         return -ENOTTY;
5554 }
5555 
5556 #ifdef CONFIG_COMPAT
5557 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5558 {
5559         switch (cmd) {
5560         case FS_IOC32_GETFLAGS:
5561                 cmd = FS_IOC_GETFLAGS;
5562                 break;
5563         case FS_IOC32_SETFLAGS:
5564                 cmd = FS_IOC_SETFLAGS;
5565                 break;
5566         case FS_IOC32_GETVERSION:
5567                 cmd = FS_IOC_GETVERSION;
5568                 break;
5569         default:
5570                 return -ENOIOCTLCMD;
5571         }
5572 
5573         return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5574 }
5575 #endif
5576 

~ [ 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