]>
Commit | Line | Data |
---|---|---|
f46b5a66 CH |
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> | |
cb8e7090 | 24 | #include <linux/fsnotify.h> |
f46b5a66 CH |
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> | |
f46b5a66 | 30 | #include <linux/backing-dev.h> |
cb8e7090 | 31 | #include <linux/mount.h> |
f46b5a66 | 32 | #include <linux/mpage.h> |
cb8e7090 | 33 | #include <linux/namei.h> |
f46b5a66 CH |
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> | |
cb8e7090 | 39 | #include <linux/security.h> |
f46b5a66 | 40 | #include <linux/xattr.h> |
7ea394f1 | 41 | #include <linux/vmalloc.h> |
5a0e3ad6 | 42 | #include <linux/slab.h> |
f7039b1d | 43 | #include <linux/blkdev.h> |
8ea05e3a | 44 | #include <linux/uuid.h> |
55e301fd | 45 | #include <linux/btrfs.h> |
416161db | 46 | #include <linux/uaccess.h> |
f46b5a66 CH |
47 | #include "ctree.h" |
48 | #include "disk-io.h" | |
49 | #include "transaction.h" | |
50 | #include "btrfs_inode.h" | |
f46b5a66 CH |
51 | #include "print-tree.h" |
52 | #include "volumes.h" | |
925baedd | 53 | #include "locking.h" |
581bb050 | 54 | #include "inode-map.h" |
d7728c96 | 55 | #include "backref.h" |
606686ee | 56 | #include "rcu-string.h" |
31db9f7c | 57 | #include "send.h" |
3f6bcfbd | 58 | #include "dev-replace.h" |
63541927 | 59 | #include "props.h" |
3b02a68a | 60 | #include "sysfs.h" |
fcebe456 | 61 | #include "qgroup.h" |
1ec9a1ae | 62 | #include "tree-log.h" |
ebb8765b | 63 | #include "compression.h" |
f46b5a66 | 64 | |
abccd00f HM |
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 | ||
416161db | 91 | static int btrfs_clone(struct inode *src, struct inode *inode, |
1c919a5e MF |
92 | u64 off, u64 olen, u64 olen_aligned, u64 destoff, |
93 | int no_time_update); | |
416161db | 94 | |
6cbff00f CH |
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; | |
d0092bdd LZ |
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; | |
6cbff00f CH |
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); | |
3cc79392 | 142 | unsigned int new_fl = 0; |
6cbff00f CH |
143 | |
144 | if (ip->flags & BTRFS_INODE_SYNC) | |
3cc79392 | 145 | new_fl |= S_SYNC; |
6cbff00f | 146 | if (ip->flags & BTRFS_INODE_IMMUTABLE) |
3cc79392 | 147 | new_fl |= S_IMMUTABLE; |
6cbff00f | 148 | if (ip->flags & BTRFS_INODE_APPEND) |
3cc79392 | 149 | new_fl |= S_APPEND; |
6cbff00f | 150 | if (ip->flags & BTRFS_INODE_NOATIME) |
3cc79392 | 151 | new_fl |= S_NOATIME; |
6cbff00f | 152 | if (ip->flags & BTRFS_INODE_DIRSYNC) |
3cc79392 FM |
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); | |
6cbff00f CH |
158 | } |
159 | ||
160 | /* | |
161 | * Inherit flags from the parent inode. | |
162 | * | |
e27425d6 | 163 | * Currently only the compression flags and the cow flags are inherited. |
6cbff00f CH |
164 | */ |
165 | void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) | |
166 | { | |
0b4dcea5 CM |
167 | unsigned int flags; |
168 | ||
169 | if (!dir) | |
170 | return; | |
171 | ||
172 | flags = BTRFS_I(dir)->flags; | |
6cbff00f | 173 | |
e27425d6 JB |
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 | ||
213490b3 | 182 | if (flags & BTRFS_INODE_NODATACOW) { |
e27425d6 | 183 | BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; |
213490b3 LB |
184 | if (S_ISREG(inode->i_mode)) |
185 | BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; | |
186 | } | |
6cbff00f | 187 | |
6cbff00f CH |
188 | btrfs_update_iflags(inode); |
189 | } | |
190 | ||
191 | static int btrfs_ioctl_getflags(struct file *file, void __user *arg) | |
192 | { | |
496ad9aa | 193 | struct btrfs_inode *ip = BTRFS_I(file_inode(file)); |
6cbff00f CH |
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 | ||
75e7cb7f LB |
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 | \ | |
e1e8fb6a LZ |
206 | FS_NOCOMP_FL | FS_COMPR_FL | |
207 | FS_NOCOW_FL)) | |
75e7cb7f LB |
208 | return -EOPNOTSUPP; |
209 | ||
210 | if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) | |
211 | return -EINVAL; | |
212 | ||
75e7cb7f LB |
213 | return 0; |
214 | } | |
215 | ||
6cbff00f CH |
216 | static int btrfs_ioctl_setflags(struct file *file, void __user *arg) |
217 | { | |
496ad9aa | 218 | struct inode *inode = file_inode(file); |
6cbff00f CH |
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; | |
f062abf0 LZ |
224 | u64 ip_oldflags; |
225 | unsigned int i_oldflags; | |
7e97b8da | 226 | umode_t mode; |
6cbff00f | 227 | |
bd60ea0f DS |
228 | if (!inode_owner_or_capable(inode)) |
229 | return -EPERM; | |
230 | ||
b83cc969 LZ |
231 | if (btrfs_root_readonly(root)) |
232 | return -EROFS; | |
233 | ||
6cbff00f CH |
234 | if (copy_from_user(&flags, arg, sizeof(flags))) |
235 | return -EFAULT; | |
236 | ||
75e7cb7f LB |
237 | ret = check_flags(flags); |
238 | if (ret) | |
239 | return ret; | |
f46b5a66 | 240 | |
e7848683 JK |
241 | ret = mnt_want_write_file(file); |
242 | if (ret) | |
243 | return ret; | |
244 | ||
5955102c | 245 | inode_lock(inode); |
6cbff00f | 246 | |
f062abf0 LZ |
247 | ip_oldflags = ip->flags; |
248 | i_oldflags = inode->i_flags; | |
7e97b8da | 249 | mode = inode->i_mode; |
f062abf0 | 250 | |
6cbff00f CH |
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 | ||
6cbff00f CH |
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; | |
7e97b8da DS |
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 | } | |
6cbff00f | 309 | |
75e7cb7f LB |
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; | |
63541927 FDBM |
318 | |
319 | ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); | |
320 | if (ret && ret != -ENODATA) | |
321 | goto out_drop; | |
75e7cb7f | 322 | } else if (flags & FS_COMPR_FL) { |
63541927 FDBM |
323 | const char *comp; |
324 | ||
75e7cb7f LB |
325 | ip->flags |= BTRFS_INODE_COMPRESS; |
326 | ip->flags &= ~BTRFS_INODE_NOCOMPRESS; | |
63541927 FDBM |
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 | ||
ebcb904d | 337 | } else { |
78a017a2 FM |
338 | ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); |
339 | if (ret && ret != -ENODATA) | |
340 | goto out_drop; | |
ebcb904d | 341 | ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); |
75e7cb7f | 342 | } |
6cbff00f | 343 | |
4da6f1a3 | 344 | trans = btrfs_start_transaction(root, 1); |
f062abf0 LZ |
345 | if (IS_ERR(trans)) { |
346 | ret = PTR_ERR(trans); | |
347 | goto out_drop; | |
348 | } | |
6cbff00f | 349 | |
306424cc | 350 | btrfs_update_iflags(inode); |
0c4d2d95 | 351 | inode_inc_iversion(inode); |
04b285f3 | 352 | inode->i_ctime = current_fs_time(inode->i_sb); |
6cbff00f | 353 | ret = btrfs_update_inode(trans, root, inode); |
6cbff00f | 354 | |
6cbff00f | 355 | btrfs_end_transaction(trans, root); |
f062abf0 LZ |
356 | out_drop: |
357 | if (ret) { | |
358 | ip->flags = ip_oldflags; | |
359 | inode->i_flags = i_oldflags; | |
360 | } | |
6cbff00f | 361 | |
6cbff00f | 362 | out_unlock: |
5955102c | 363 | inode_unlock(inode); |
e7848683 | 364 | mnt_drop_write_file(file); |
2d4e6f6a | 365 | return ret; |
6cbff00f CH |
366 | } |
367 | ||
368 | static int btrfs_ioctl_getversion(struct file *file, int __user *arg) | |
369 | { | |
496ad9aa | 370 | struct inode *inode = file_inode(file); |
6cbff00f CH |
371 | |
372 | return put_user(inode->i_generation, arg); | |
373 | } | |
f46b5a66 | 374 | |
f7039b1d LD |
375 | static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) |
376 | { | |
54563d41 | 377 | struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); |
f7039b1d LD |
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; | |
815745cf | 383 | u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); |
f7039b1d LD |
384 | int ret; |
385 | ||
386 | if (!capable(CAP_SYS_ADMIN)) | |
387 | return -EPERM; | |
388 | ||
1f78160c XG |
389 | rcu_read_lock(); |
390 | list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, | |
391 | dev_list) { | |
f7039b1d LD |
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 | } | |
1f78160c | 401 | rcu_read_unlock(); |
f4c697e6 | 402 | |
f7039b1d LD |
403 | if (!num_devices) |
404 | return -EOPNOTSUPP; | |
f7039b1d LD |
405 | if (copy_from_user(&range, arg, sizeof(range))) |
406 | return -EFAULT; | |
e515c18b LC |
407 | if (range.start > total_bytes || |
408 | range.len < fs_info->sb->s_blocksize) | |
f4c697e6 | 409 | return -EINVAL; |
f7039b1d | 410 | |
f4c697e6 | 411 | range.len = min(range.len, total_bytes - range.start); |
f7039b1d | 412 | range.minlen = max(range.minlen, minlen); |
815745cf | 413 | ret = btrfs_trim_fs(fs_info->tree_root, &range); |
f7039b1d LD |
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 | ||
dd5f9615 SB |
423 | int btrfs_is_empty_uuid(u8 *uuid) |
424 | { | |
46e0f66a CM |
425 | int i; |
426 | ||
427 | for (i = 0; i < BTRFS_UUID_SIZE; i++) { | |
428 | if (uuid[i]) | |
429 | return 0; | |
430 | } | |
431 | return 1; | |
dd5f9615 SB |
432 | } |
433 | ||
d5c12070 | 434 | static noinline int create_subvol(struct inode *dir, |
cb8e7090 | 435 | struct dentry *dentry, |
72fd032e | 436 | char *name, int namelen, |
6f72c7e2 | 437 | u64 *async_transid, |
8696c533 | 438 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 CH |
439 | { |
440 | struct btrfs_trans_handle *trans; | |
441 | struct btrfs_key key; | |
49a3c4d9 | 442 | struct btrfs_root_item *root_item; |
f46b5a66 CH |
443 | struct btrfs_inode_item *inode_item; |
444 | struct extent_buffer *leaf; | |
d5c12070 | 445 | struct btrfs_root *root = BTRFS_I(dir)->root; |
76dda93c | 446 | struct btrfs_root *new_root; |
d5c12070 | 447 | struct btrfs_block_rsv block_rsv; |
04b285f3 | 448 | struct timespec cur_time = current_fs_time(dir->i_sb); |
5662344b | 449 | struct inode *inode; |
f46b5a66 CH |
450 | int ret; |
451 | int err; | |
452 | u64 objectid; | |
453 | u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; | |
3de4586c | 454 | u64 index = 0; |
d5c12070 | 455 | u64 qgroup_reserved; |
8ea05e3a | 456 | uuid_le new_uuid; |
f46b5a66 | 457 | |
49a3c4d9 DS |
458 | root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); |
459 | if (!root_item) | |
460 | return -ENOMEM; | |
461 | ||
581bb050 | 462 | ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); |
2fbe8c8a | 463 | if (ret) |
49a3c4d9 | 464 | goto fail_free; |
6a912213 | 465 | |
e09fe2d2 QW |
466 | /* |
467 | * Don't create subvolume whose level is not zero. Or qgroup will be | |
468 | * screwed up since it assume subvolme qgroup's level to be 0. | |
469 | */ | |
49a3c4d9 DS |
470 | if (btrfs_qgroup_level(objectid)) { |
471 | ret = -ENOSPC; | |
472 | goto fail_free; | |
473 | } | |
e09fe2d2 | 474 | |
d5c12070 | 475 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
9ed74f2d | 476 | /* |
d5c12070 MX |
477 | * The same as the snapshot creation, please see the comment |
478 | * of create_snapshot(). | |
9ed74f2d | 479 | */ |
d5c12070 | 480 | ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, |
dd5f9615 | 481 | 8, &qgroup_reserved, false); |
d5c12070 | 482 | if (ret) |
49a3c4d9 | 483 | goto fail_free; |
d5c12070 MX |
484 | |
485 | trans = btrfs_start_transaction(root, 0); | |
486 | if (IS_ERR(trans)) { | |
487 | ret = PTR_ERR(trans); | |
de6e8200 LB |
488 | btrfs_subvolume_release_metadata(root, &block_rsv, |
489 | qgroup_reserved); | |
49a3c4d9 | 490 | goto fail_free; |
d5c12070 MX |
491 | } |
492 | trans->block_rsv = &block_rsv; | |
493 | trans->bytes_reserved = block_rsv.size; | |
f46b5a66 | 494 | |
8696c533 | 495 | ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit); |
6f72c7e2 AJ |
496 | if (ret) |
497 | goto fail; | |
498 | ||
4d75f8a9 | 499 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
8e8a1e31 JB |
500 | if (IS_ERR(leaf)) { |
501 | ret = PTR_ERR(leaf); | |
502 | goto fail; | |
503 | } | |
f46b5a66 | 504 | |
5d4f98a2 | 505 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
f46b5a66 CH |
506 | btrfs_set_header_bytenr(leaf, leaf->start); |
507 | btrfs_set_header_generation(leaf, trans->transid); | |
5d4f98a2 | 508 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); |
f46b5a66 CH |
509 | btrfs_set_header_owner(leaf, objectid); |
510 | ||
0a4e5586 | 511 | write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(), |
f46b5a66 | 512 | BTRFS_FSID_SIZE); |
5d4f98a2 | 513 | write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, |
b308bc2f | 514 | btrfs_header_chunk_tree_uuid(leaf), |
5d4f98a2 | 515 | BTRFS_UUID_SIZE); |
f46b5a66 CH |
516 | btrfs_mark_buffer_dirty(leaf); |
517 | ||
49a3c4d9 | 518 | inode_item = &root_item->inode; |
3cae210f QW |
519 | btrfs_set_stack_inode_generation(inode_item, 1); |
520 | btrfs_set_stack_inode_size(inode_item, 3); | |
521 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
707e8a07 | 522 | btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); |
3cae210f | 523 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
f46b5a66 | 524 | |
49a3c4d9 DS |
525 | btrfs_set_root_flags(root_item, 0); |
526 | btrfs_set_root_limit(root_item, 0); | |
3cae210f | 527 | btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); |
08fe4db1 | 528 | |
49a3c4d9 DS |
529 | btrfs_set_root_bytenr(root_item, leaf->start); |
530 | btrfs_set_root_generation(root_item, trans->transid); | |
531 | btrfs_set_root_level(root_item, 0); | |
532 | btrfs_set_root_refs(root_item, 1); | |
533 | btrfs_set_root_used(root_item, leaf->len); | |
534 | btrfs_set_root_last_snapshot(root_item, 0); | |
f46b5a66 | 535 | |
49a3c4d9 DS |
536 | btrfs_set_root_generation_v2(root_item, |
537 | btrfs_root_generation(root_item)); | |
8ea05e3a | 538 | uuid_le_gen(&new_uuid); |
49a3c4d9 DS |
539 | memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); |
540 | btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); | |
541 | btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); | |
542 | root_item->ctime = root_item->otime; | |
543 | btrfs_set_root_ctransid(root_item, trans->transid); | |
544 | btrfs_set_root_otransid(root_item, trans->transid); | |
f46b5a66 | 545 | |
925baedd | 546 | btrfs_tree_unlock(leaf); |
f46b5a66 CH |
547 | free_extent_buffer(leaf); |
548 | leaf = NULL; | |
549 | ||
49a3c4d9 | 550 | btrfs_set_root_dirid(root_item, new_dirid); |
f46b5a66 CH |
551 | |
552 | key.objectid = objectid; | |
5d4f98a2 | 553 | key.offset = 0; |
962a298f | 554 | key.type = BTRFS_ROOT_ITEM_KEY; |
f46b5a66 | 555 | ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, |
49a3c4d9 | 556 | root_item); |
f46b5a66 CH |
557 | if (ret) |
558 | goto fail; | |
559 | ||
76dda93c YZ |
560 | key.offset = (u64)-1; |
561 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); | |
79787eaa | 562 | if (IS_ERR(new_root)) { |
79787eaa | 563 | ret = PTR_ERR(new_root); |
6d13f549 | 564 | btrfs_abort_transaction(trans, root, ret); |
79787eaa JM |
565 | goto fail; |
566 | } | |
76dda93c YZ |
567 | |
568 | btrfs_record_root_in_trans(trans, new_root); | |
569 | ||
63541927 | 570 | ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); |
ce598979 MF |
571 | if (ret) { |
572 | /* We potentially lose an unused inode item here */ | |
79787eaa | 573 | btrfs_abort_transaction(trans, root, ret); |
ce598979 MF |
574 | goto fail; |
575 | } | |
576 | ||
f32e48e9 CR |
577 | mutex_lock(&new_root->objectid_mutex); |
578 | new_root->highest_objectid = new_dirid; | |
579 | mutex_unlock(&new_root->objectid_mutex); | |
580 | ||
f46b5a66 CH |
581 | /* |
582 | * insert the directory item | |
583 | */ | |
3de4586c | 584 | ret = btrfs_set_inode_index(dir, &index); |
79787eaa JM |
585 | if (ret) { |
586 | btrfs_abort_transaction(trans, root, ret); | |
587 | goto fail; | |
588 | } | |
3de4586c CM |
589 | |
590 | ret = btrfs_insert_dir_item(trans, root, | |
16cdcec7 | 591 | name, namelen, dir, &key, |
3de4586c | 592 | BTRFS_FT_DIR, index); |
79787eaa JM |
593 | if (ret) { |
594 | btrfs_abort_transaction(trans, root, ret); | |
f46b5a66 | 595 | goto fail; |
79787eaa | 596 | } |
0660b5af | 597 | |
52c26179 YZ |
598 | btrfs_i_size_write(dir, dir->i_size + namelen * 2); |
599 | ret = btrfs_update_inode(trans, root, dir); | |
600 | BUG_ON(ret); | |
601 | ||
0660b5af | 602 | ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, |
4df27c4d | 603 | objectid, root->root_key.objectid, |
33345d01 | 604 | btrfs_ino(dir), index, name, namelen); |
76dda93c | 605 | BUG_ON(ret); |
f46b5a66 | 606 | |
dd5f9615 | 607 | ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, |
49a3c4d9 | 608 | root_item->uuid, BTRFS_UUID_KEY_SUBVOL, |
dd5f9615 SB |
609 | objectid); |
610 | if (ret) | |
611 | btrfs_abort_transaction(trans, root, ret); | |
612 | ||
f46b5a66 | 613 | fail: |
49a3c4d9 | 614 | kfree(root_item); |
d5c12070 MX |
615 | trans->block_rsv = NULL; |
616 | trans->bytes_reserved = 0; | |
de6e8200 LB |
617 | btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); |
618 | ||
72fd032e SW |
619 | if (async_transid) { |
620 | *async_transid = trans->transid; | |
621 | err = btrfs_commit_transaction_async(trans, root, 1); | |
00d71c9c MX |
622 | if (err) |
623 | err = btrfs_commit_transaction(trans, root); | |
72fd032e SW |
624 | } else { |
625 | err = btrfs_commit_transaction(trans, root); | |
626 | } | |
f46b5a66 CH |
627 | if (err && !ret) |
628 | ret = err; | |
1a65e24b | 629 | |
5662344b TI |
630 | if (!ret) { |
631 | inode = btrfs_lookup_dentry(dir, dentry); | |
de6e8200 LB |
632 | if (IS_ERR(inode)) |
633 | return PTR_ERR(inode); | |
5662344b TI |
634 | d_instantiate(dentry, inode); |
635 | } | |
f46b5a66 | 636 | return ret; |
49a3c4d9 DS |
637 | |
638 | fail_free: | |
639 | kfree(root_item); | |
640 | return ret; | |
f46b5a66 CH |
641 | } |
642 | ||
9ea24bbe | 643 | static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root) |
8257b2dc MX |
644 | { |
645 | s64 writers; | |
646 | DEFINE_WAIT(wait); | |
647 | ||
648 | do { | |
649 | prepare_to_wait(&root->subv_writers->wait, &wait, | |
650 | TASK_UNINTERRUPTIBLE); | |
651 | ||
652 | writers = percpu_counter_sum(&root->subv_writers->counter); | |
653 | if (writers) | |
654 | schedule(); | |
655 | ||
656 | finish_wait(&root->subv_writers->wait, &wait); | |
657 | } while (writers); | |
658 | } | |
659 | ||
e9662f70 MX |
660 | static int create_snapshot(struct btrfs_root *root, struct inode *dir, |
661 | struct dentry *dentry, char *name, int namelen, | |
662 | u64 *async_transid, bool readonly, | |
663 | struct btrfs_qgroup_inherit *inherit) | |
f46b5a66 | 664 | { |
2e4bfab9 | 665 | struct inode *inode; |
f46b5a66 CH |
666 | struct btrfs_pending_snapshot *pending_snapshot; |
667 | struct btrfs_trans_handle *trans; | |
2e4bfab9 | 668 | int ret; |
f46b5a66 | 669 | |
27cdeb70 | 670 | if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) |
f46b5a66 CH |
671 | return -EINVAL; |
672 | ||
a1ee7362 DS |
673 | pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); |
674 | if (!pending_snapshot) | |
675 | return -ENOMEM; | |
676 | ||
b0c0ea63 DS |
677 | pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), |
678 | GFP_NOFS); | |
8546b570 DS |
679 | pending_snapshot->path = btrfs_alloc_path(); |
680 | if (!pending_snapshot->root_item || !pending_snapshot->path) { | |
b0c0ea63 DS |
681 | ret = -ENOMEM; |
682 | goto free_pending; | |
683 | } | |
684 | ||
8257b2dc | 685 | atomic_inc(&root->will_be_snapshoted); |
4e857c58 | 686 | smp_mb__after_atomic(); |
9ea24bbe | 687 | btrfs_wait_for_no_snapshoting_writes(root); |
8257b2dc | 688 | |
6a03843d MX |
689 | ret = btrfs_start_delalloc_inodes(root, 0); |
690 | if (ret) | |
a1ee7362 | 691 | goto dec_and_free; |
6a03843d | 692 | |
b0244199 | 693 | btrfs_wait_ordered_extents(root, -1); |
6a03843d | 694 | |
66d8f3dd MX |
695 | btrfs_init_block_rsv(&pending_snapshot->block_rsv, |
696 | BTRFS_BLOCK_RSV_TEMP); | |
d5c12070 MX |
697 | /* |
698 | * 1 - parent dir inode | |
699 | * 2 - dir entries | |
700 | * 1 - root item | |
701 | * 2 - root ref/backref | |
702 | * 1 - root of snapshot | |
dd5f9615 | 703 | * 1 - UUID item |
d5c12070 MX |
704 | */ |
705 | ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, | |
dd5f9615 | 706 | &pending_snapshot->block_rsv, 8, |
ee3441b4 JM |
707 | &pending_snapshot->qgroup_reserved, |
708 | false); | |
d5c12070 | 709 | if (ret) |
a1ee7362 | 710 | goto dec_and_free; |
d5c12070 | 711 | |
3de4586c | 712 | pending_snapshot->dentry = dentry; |
f46b5a66 | 713 | pending_snapshot->root = root; |
b83cc969 | 714 | pending_snapshot->readonly = readonly; |
e9662f70 | 715 | pending_snapshot->dir = dir; |
8696c533 | 716 | pending_snapshot->inherit = inherit; |
a22285a6 | 717 | |
d5c12070 | 718 | trans = btrfs_start_transaction(root, 0); |
a22285a6 YZ |
719 | if (IS_ERR(trans)) { |
720 | ret = PTR_ERR(trans); | |
721 | goto fail; | |
722 | } | |
723 | ||
8351583e | 724 | spin_lock(&root->fs_info->trans_lock); |
f46b5a66 CH |
725 | list_add(&pending_snapshot->list, |
726 | &trans->transaction->pending_snapshots); | |
8351583e | 727 | spin_unlock(&root->fs_info->trans_lock); |
72fd032e SW |
728 | if (async_transid) { |
729 | *async_transid = trans->transid; | |
730 | ret = btrfs_commit_transaction_async(trans, | |
731 | root->fs_info->extent_root, 1); | |
00d71c9c MX |
732 | if (ret) |
733 | ret = btrfs_commit_transaction(trans, root); | |
72fd032e SW |
734 | } else { |
735 | ret = btrfs_commit_transaction(trans, | |
736 | root->fs_info->extent_root); | |
737 | } | |
aec8030a | 738 | if (ret) |
c37b2b62 | 739 | goto fail; |
a22285a6 YZ |
740 | |
741 | ret = pending_snapshot->error; | |
742 | if (ret) | |
743 | goto fail; | |
744 | ||
d3797308 CM |
745 | ret = btrfs_orphan_cleanup(pending_snapshot->snap); |
746 | if (ret) | |
747 | goto fail; | |
748 | ||
2b0143b5 | 749 | inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); |
2e4bfab9 YZ |
750 | if (IS_ERR(inode)) { |
751 | ret = PTR_ERR(inode); | |
752 | goto fail; | |
753 | } | |
5662344b | 754 | |
2e4bfab9 YZ |
755 | d_instantiate(dentry, inode); |
756 | ret = 0; | |
757 | fail: | |
d5c12070 MX |
758 | btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, |
759 | &pending_snapshot->block_rsv, | |
760 | pending_snapshot->qgroup_reserved); | |
a1ee7362 | 761 | dec_and_free: |
9ea24bbe FM |
762 | if (atomic_dec_and_test(&root->will_be_snapshoted)) |
763 | wake_up_atomic_t(&root->will_be_snapshoted); | |
b0c0ea63 DS |
764 | free_pending: |
765 | kfree(pending_snapshot->root_item); | |
8546b570 | 766 | btrfs_free_path(pending_snapshot->path); |
a1ee7362 DS |
767 | kfree(pending_snapshot); |
768 | ||
f46b5a66 CH |
769 | return ret; |
770 | } | |
771 | ||
4260f7c7 SW |
772 | /* copy of may_delete in fs/namei.c() |
773 | * Check whether we can remove a link victim from directory dir, check | |
774 | * whether the type of victim is right. | |
775 | * 1. We can't do it if dir is read-only (done in permission()) | |
776 | * 2. We should have write and exec permissions on dir | |
777 | * 3. We can't remove anything from append-only dir | |
778 | * 4. We can't do anything with immutable dir (done in permission()) | |
779 | * 5. If the sticky bit on dir is set we should either | |
780 | * a. be owner of dir, or | |
781 | * b. be owner of victim, or | |
782 | * c. have CAP_FOWNER capability | |
783 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
784 | * links pointing to it. | |
785 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
786 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
787 | * 9. We can't remove a root or mountpoint. | |
788 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
789 | * nfs_async_unlink(). | |
790 | */ | |
791 | ||
67871254 | 792 | static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) |
4260f7c7 SW |
793 | { |
794 | int error; | |
795 | ||
2b0143b5 | 796 | if (d_really_is_negative(victim)) |
4260f7c7 SW |
797 | return -ENOENT; |
798 | ||
2b0143b5 | 799 | BUG_ON(d_inode(victim->d_parent) != dir); |
4fa6b5ec | 800 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
4260f7c7 SW |
801 | |
802 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
803 | if (error) | |
804 | return error; | |
805 | if (IS_APPEND(dir)) | |
806 | return -EPERM; | |
2b0143b5 DH |
807 | if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || |
808 | IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) | |
4260f7c7 SW |
809 | return -EPERM; |
810 | if (isdir) { | |
e36cb0b8 | 811 | if (!d_is_dir(victim)) |
4260f7c7 SW |
812 | return -ENOTDIR; |
813 | if (IS_ROOT(victim)) | |
814 | return -EBUSY; | |
e36cb0b8 | 815 | } else if (d_is_dir(victim)) |
4260f7c7 SW |
816 | return -EISDIR; |
817 | if (IS_DEADDIR(dir)) | |
818 | return -ENOENT; | |
819 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
820 | return -EBUSY; | |
821 | return 0; | |
822 | } | |
823 | ||
cb8e7090 CH |
824 | /* copy of may_create in fs/namei.c() */ |
825 | static inline int btrfs_may_create(struct inode *dir, struct dentry *child) | |
826 | { | |
2b0143b5 | 827 | if (d_really_is_positive(child)) |
cb8e7090 CH |
828 | return -EEXIST; |
829 | if (IS_DEADDIR(dir)) | |
830 | return -ENOENT; | |
831 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
832 | } | |
833 | ||
834 | /* | |
835 | * Create a new subvolume below @parent. This is largely modeled after | |
836 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
837 | * inside this filesystem so it's quite a bit simpler. | |
838 | */ | |
76dda93c YZ |
839 | static noinline int btrfs_mksubvol(struct path *parent, |
840 | char *name, int namelen, | |
72fd032e | 841 | struct btrfs_root *snap_src, |
6f72c7e2 | 842 | u64 *async_transid, bool readonly, |
8696c533 | 843 | struct btrfs_qgroup_inherit *inherit) |
cb8e7090 | 844 | { |
2b0143b5 | 845 | struct inode *dir = d_inode(parent->dentry); |
cb8e7090 CH |
846 | struct dentry *dentry; |
847 | int error; | |
848 | ||
5c50c9b8 DS |
849 | error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); |
850 | if (error == -EINTR) | |
851 | return error; | |
cb8e7090 CH |
852 | |
853 | dentry = lookup_one_len(name, parent->dentry, namelen); | |
854 | error = PTR_ERR(dentry); | |
855 | if (IS_ERR(dentry)) | |
856 | goto out_unlock; | |
857 | ||
76dda93c | 858 | error = btrfs_may_create(dir, dentry); |
cb8e7090 | 859 | if (error) |
a874a63e | 860 | goto out_dput; |
cb8e7090 | 861 | |
9c52057c CM |
862 | /* |
863 | * even if this name doesn't exist, we may get hash collisions. | |
864 | * check for them now when we can safely fail | |
865 | */ | |
866 | error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, | |
867 | dir->i_ino, name, | |
868 | namelen); | |
869 | if (error) | |
870 | goto out_dput; | |
871 | ||
76dda93c YZ |
872 | down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); |
873 | ||
874 | if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) | |
875 | goto out_up_read; | |
876 | ||
3de4586c | 877 | if (snap_src) { |
e9662f70 | 878 | error = create_snapshot(snap_src, dir, dentry, name, namelen, |
6f72c7e2 | 879 | async_transid, readonly, inherit); |
3de4586c | 880 | } else { |
d5c12070 MX |
881 | error = create_subvol(dir, dentry, name, namelen, |
882 | async_transid, inherit); | |
3de4586c | 883 | } |
76dda93c YZ |
884 | if (!error) |
885 | fsnotify_mkdir(dir, dentry); | |
886 | out_up_read: | |
887 | up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); | |
cb8e7090 CH |
888 | out_dput: |
889 | dput(dentry); | |
890 | out_unlock: | |
5955102c | 891 | inode_unlock(dir); |
cb8e7090 CH |
892 | return error; |
893 | } | |
894 | ||
4cb5300b CM |
895 | /* |
896 | * When we're defragging a range, we don't want to kick it off again | |
897 | * if it is really just waiting for delalloc to send it down. | |
898 | * If we find a nice big extent or delalloc range for the bytes in the | |
899 | * file you want to defrag, we return 0 to let you know to skip this | |
900 | * part of the file | |
901 | */ | |
aab110ab | 902 | static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) |
4cb5300b CM |
903 | { |
904 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
905 | struct extent_map *em = NULL; | |
906 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
907 | u64 end; | |
908 | ||
909 | read_lock(&em_tree->lock); | |
09cbfeaf | 910 | em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); |
4cb5300b CM |
911 | read_unlock(&em_tree->lock); |
912 | ||
913 | if (em) { | |
914 | end = extent_map_end(em); | |
915 | free_extent_map(em); | |
916 | if (end - offset > thresh) | |
917 | return 0; | |
918 | } | |
919 | /* if we already have a nice delalloc here, just stop */ | |
920 | thresh /= 2; | |
921 | end = count_range_bits(io_tree, &offset, offset + thresh, | |
922 | thresh, EXTENT_DELALLOC, 1); | |
923 | if (end >= thresh) | |
924 | return 0; | |
925 | return 1; | |
926 | } | |
927 | ||
928 | /* | |
929 | * helper function to walk through a file and find extents | |
930 | * newer than a specific transid, and smaller than thresh. | |
931 | * | |
932 | * This is used by the defragging code to find new and small | |
933 | * extents | |
934 | */ | |
935 | static int find_new_extents(struct btrfs_root *root, | |
936 | struct inode *inode, u64 newer_than, | |
aab110ab | 937 | u64 *off, u32 thresh) |
4cb5300b CM |
938 | { |
939 | struct btrfs_path *path; | |
940 | struct btrfs_key min_key; | |
4cb5300b CM |
941 | struct extent_buffer *leaf; |
942 | struct btrfs_file_extent_item *extent; | |
943 | int type; | |
944 | int ret; | |
a4689d2b | 945 | u64 ino = btrfs_ino(inode); |
4cb5300b CM |
946 | |
947 | path = btrfs_alloc_path(); | |
948 | if (!path) | |
949 | return -ENOMEM; | |
950 | ||
a4689d2b | 951 | min_key.objectid = ino; |
4cb5300b CM |
952 | min_key.type = BTRFS_EXTENT_DATA_KEY; |
953 | min_key.offset = *off; | |
954 | ||
67871254 | 955 | while (1) { |
6174d3cb | 956 | ret = btrfs_search_forward(root, &min_key, path, newer_than); |
4cb5300b CM |
957 | if (ret != 0) |
958 | goto none; | |
f094c9bd | 959 | process_slot: |
a4689d2b | 960 | if (min_key.objectid != ino) |
4cb5300b CM |
961 | goto none; |
962 | if (min_key.type != BTRFS_EXTENT_DATA_KEY) | |
963 | goto none; | |
964 | ||
965 | leaf = path->nodes[0]; | |
966 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
967 | struct btrfs_file_extent_item); | |
968 | ||
969 | type = btrfs_file_extent_type(leaf, extent); | |
970 | if (type == BTRFS_FILE_EXTENT_REG && | |
971 | btrfs_file_extent_num_bytes(leaf, extent) < thresh && | |
972 | check_defrag_in_cache(inode, min_key.offset, thresh)) { | |
973 | *off = min_key.offset; | |
974 | btrfs_free_path(path); | |
975 | return 0; | |
976 | } | |
977 | ||
f094c9bd FM |
978 | path->slots[0]++; |
979 | if (path->slots[0] < btrfs_header_nritems(leaf)) { | |
980 | btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); | |
981 | goto process_slot; | |
982 | } | |
983 | ||
4cb5300b CM |
984 | if (min_key.offset == (u64)-1) |
985 | goto none; | |
986 | ||
987 | min_key.offset++; | |
988 | btrfs_release_path(path); | |
989 | } | |
990 | none: | |
991 | btrfs_free_path(path); | |
992 | return -ENOENT; | |
993 | } | |
994 | ||
6c282eb4 | 995 | static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) |
17ce6ef8 LB |
996 | { |
997 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
6c282eb4 LZ |
998 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
999 | struct extent_map *em; | |
09cbfeaf | 1000 | u64 len = PAGE_SIZE; |
17ce6ef8 | 1001 | |
6c282eb4 LZ |
1002 | /* |
1003 | * hopefully we have this extent in the tree already, try without | |
1004 | * the full extent lock | |
1005 | */ | |
17ce6ef8 | 1006 | read_lock(&em_tree->lock); |
6c282eb4 | 1007 | em = lookup_extent_mapping(em_tree, start, len); |
17ce6ef8 LB |
1008 | read_unlock(&em_tree->lock); |
1009 | ||
6c282eb4 | 1010 | if (!em) { |
308d9800 FM |
1011 | struct extent_state *cached = NULL; |
1012 | u64 end = start + len - 1; | |
1013 | ||
6c282eb4 | 1014 | /* get the big lock and read metadata off disk */ |
ff13db41 | 1015 | lock_extent_bits(io_tree, start, end, &cached); |
6c282eb4 | 1016 | em = btrfs_get_extent(inode, NULL, 0, start, len, 0); |
308d9800 | 1017 | unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); |
6c282eb4 LZ |
1018 | |
1019 | if (IS_ERR(em)) | |
1020 | return NULL; | |
1021 | } | |
1022 | ||
1023 | return em; | |
1024 | } | |
17ce6ef8 | 1025 | |
6c282eb4 LZ |
1026 | static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) |
1027 | { | |
1028 | struct extent_map *next; | |
1029 | bool ret = true; | |
1030 | ||
1031 | /* this is the last extent */ | |
1032 | if (em->start + em->len >= i_size_read(inode)) | |
1033 | return false; | |
1034 | ||
1035 | next = defrag_lookup_extent(inode, em->start + em->len); | |
e9512d72 CM |
1036 | if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) |
1037 | ret = false; | |
1038 | else if ((em->block_start + em->block_len == next->block_start) && | |
ee22184b | 1039 | (em->block_len > SZ_128K && next->block_len > SZ_128K)) |
6c282eb4 LZ |
1040 | ret = false; |
1041 | ||
1042 | free_extent_map(next); | |
17ce6ef8 LB |
1043 | return ret; |
1044 | } | |
1045 | ||
aab110ab | 1046 | static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, |
a43a2111 AM |
1047 | u64 *last_len, u64 *skip, u64 *defrag_end, |
1048 | int compress) | |
940100a4 | 1049 | { |
6c282eb4 | 1050 | struct extent_map *em; |
940100a4 | 1051 | int ret = 1; |
6c282eb4 | 1052 | bool next_mergeable = true; |
4a3560c4 | 1053 | bool prev_mergeable = true; |
940100a4 CM |
1054 | |
1055 | /* | |
008873ea | 1056 | * make sure that once we start defragging an extent, we keep on |
940100a4 CM |
1057 | * defragging it |
1058 | */ | |
1059 | if (start < *defrag_end) | |
1060 | return 1; | |
1061 | ||
1062 | *skip = 0; | |
1063 | ||
6c282eb4 LZ |
1064 | em = defrag_lookup_extent(inode, start); |
1065 | if (!em) | |
1066 | return 0; | |
940100a4 CM |
1067 | |
1068 | /* this will cover holes, and inline extents */ | |
17ce6ef8 | 1069 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
940100a4 | 1070 | ret = 0; |
17ce6ef8 LB |
1071 | goto out; |
1072 | } | |
1073 | ||
4a3560c4 LB |
1074 | if (!*defrag_end) |
1075 | prev_mergeable = false; | |
1076 | ||
6c282eb4 | 1077 | next_mergeable = defrag_check_next_extent(inode, em); |
940100a4 | 1078 | /* |
6c282eb4 LZ |
1079 | * we hit a real extent, if it is big or the next extent is not a |
1080 | * real extent, don't bother defragging it | |
940100a4 | 1081 | */ |
a43a2111 | 1082 | if (!compress && (*last_len == 0 || *last_len >= thresh) && |
4a3560c4 | 1083 | (em->len >= thresh || (!next_mergeable && !prev_mergeable))) |
940100a4 | 1084 | ret = 0; |
17ce6ef8 | 1085 | out: |
940100a4 CM |
1086 | /* |
1087 | * last_len ends up being a counter of how many bytes we've defragged. | |
1088 | * every time we choose not to defrag an extent, we reset *last_len | |
1089 | * so that the next tiny extent will force a defrag. | |
1090 | * | |
1091 | * The end result of this is that tiny extents before a single big | |
1092 | * extent will force at least part of that big extent to be defragged. | |
1093 | */ | |
1094 | if (ret) { | |
940100a4 CM |
1095 | *defrag_end = extent_map_end(em); |
1096 | } else { | |
1097 | *last_len = 0; | |
1098 | *skip = extent_map_end(em); | |
1099 | *defrag_end = 0; | |
1100 | } | |
1101 | ||
1102 | free_extent_map(em); | |
1103 | return ret; | |
1104 | } | |
1105 | ||
4cb5300b CM |
1106 | /* |
1107 | * it doesn't do much good to defrag one or two pages | |
1108 | * at a time. This pulls in a nice chunk of pages | |
1109 | * to COW and defrag. | |
1110 | * | |
1111 | * It also makes sure the delalloc code has enough | |
1112 | * dirty data to avoid making new small extents as part | |
1113 | * of the defrag | |
1114 | * | |
1115 | * It's a good idea to start RA on this range | |
1116 | * before calling this. | |
1117 | */ | |
1118 | static int cluster_pages_for_defrag(struct inode *inode, | |
1119 | struct page **pages, | |
1120 | unsigned long start_index, | |
c41570c9 | 1121 | unsigned long num_pages) |
f46b5a66 | 1122 | { |
4cb5300b CM |
1123 | unsigned long file_end; |
1124 | u64 isize = i_size_read(inode); | |
1125 | u64 page_start; | |
1126 | u64 page_end; | |
1f12bd06 | 1127 | u64 page_cnt; |
4cb5300b CM |
1128 | int ret; |
1129 | int i; | |
1130 | int i_done; | |
3eaa2885 | 1131 | struct btrfs_ordered_extent *ordered; |
4cb5300b | 1132 | struct extent_state *cached_state = NULL; |
600a45e1 | 1133 | struct extent_io_tree *tree; |
3b16a4e3 | 1134 | gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); |
4cb5300b | 1135 | |
09cbfeaf | 1136 | file_end = (isize - 1) >> PAGE_SHIFT; |
1f12bd06 LB |
1137 | if (!isize || start_index > file_end) |
1138 | return 0; | |
1139 | ||
1140 | page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); | |
4cb5300b CM |
1141 | |
1142 | ret = btrfs_delalloc_reserve_space(inode, | |
09cbfeaf KS |
1143 | start_index << PAGE_SHIFT, |
1144 | page_cnt << PAGE_SHIFT); | |
4cb5300b CM |
1145 | if (ret) |
1146 | return ret; | |
4cb5300b | 1147 | i_done = 0; |
600a45e1 | 1148 | tree = &BTRFS_I(inode)->io_tree; |
4cb5300b CM |
1149 | |
1150 | /* step one, lock all the pages */ | |
1f12bd06 | 1151 | for (i = 0; i < page_cnt; i++) { |
4cb5300b | 1152 | struct page *page; |
600a45e1 | 1153 | again: |
a94733d0 | 1154 | page = find_or_create_page(inode->i_mapping, |
600a45e1 | 1155 | start_index + i, mask); |
4cb5300b CM |
1156 | if (!page) |
1157 | break; | |
1158 | ||
600a45e1 | 1159 | page_start = page_offset(page); |
09cbfeaf | 1160 | page_end = page_start + PAGE_SIZE - 1; |
600a45e1 | 1161 | while (1) { |
308d9800 | 1162 | lock_extent_bits(tree, page_start, page_end, |
ff13db41 | 1163 | &cached_state); |
600a45e1 MX |
1164 | ordered = btrfs_lookup_ordered_extent(inode, |
1165 | page_start); | |
308d9800 FM |
1166 | unlock_extent_cached(tree, page_start, page_end, |
1167 | &cached_state, GFP_NOFS); | |
600a45e1 MX |
1168 | if (!ordered) |
1169 | break; | |
1170 | ||
1171 | unlock_page(page); | |
1172 | btrfs_start_ordered_extent(inode, ordered, 1); | |
1173 | btrfs_put_ordered_extent(ordered); | |
1174 | lock_page(page); | |
1f12bd06 LB |
1175 | /* |
1176 | * we unlocked the page above, so we need check if | |
1177 | * it was released or not. | |
1178 | */ | |
1179 | if (page->mapping != inode->i_mapping) { | |
1180 | unlock_page(page); | |
09cbfeaf | 1181 | put_page(page); |
1f12bd06 LB |
1182 | goto again; |
1183 | } | |
600a45e1 MX |
1184 | } |
1185 | ||
4cb5300b CM |
1186 | if (!PageUptodate(page)) { |
1187 | btrfs_readpage(NULL, page); | |
1188 | lock_page(page); | |
1189 | if (!PageUptodate(page)) { | |
1190 | unlock_page(page); | |
09cbfeaf | 1191 | put_page(page); |
4cb5300b CM |
1192 | ret = -EIO; |
1193 | break; | |
1194 | } | |
1195 | } | |
600a45e1 | 1196 | |
600a45e1 MX |
1197 | if (page->mapping != inode->i_mapping) { |
1198 | unlock_page(page); | |
09cbfeaf | 1199 | put_page(page); |
600a45e1 MX |
1200 | goto again; |
1201 | } | |
1202 | ||
4cb5300b CM |
1203 | pages[i] = page; |
1204 | i_done++; | |
1205 | } | |
1206 | if (!i_done || ret) | |
1207 | goto out; | |
1208 | ||
1209 | if (!(inode->i_sb->s_flags & MS_ACTIVE)) | |
1210 | goto out; | |
1211 | ||
1212 | /* | |
1213 | * so now we have a nice long stream of locked | |
1214 | * and up to date pages, lets wait on them | |
1215 | */ | |
1216 | for (i = 0; i < i_done; i++) | |
1217 | wait_on_page_writeback(pages[i]); | |
1218 | ||
1219 | page_start = page_offset(pages[0]); | |
09cbfeaf | 1220 | page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; |
4cb5300b CM |
1221 | |
1222 | lock_extent_bits(&BTRFS_I(inode)->io_tree, | |
ff13db41 | 1223 | page_start, page_end - 1, &cached_state); |
4cb5300b CM |
1224 | clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, |
1225 | page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | | |
9e8a4a8b LB |
1226 | EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, |
1227 | &cached_state, GFP_NOFS); | |
4cb5300b | 1228 | |
1f12bd06 | 1229 | if (i_done != page_cnt) { |
9e0baf60 JB |
1230 | spin_lock(&BTRFS_I(inode)->lock); |
1231 | BTRFS_I(inode)->outstanding_extents++; | |
1232 | spin_unlock(&BTRFS_I(inode)->lock); | |
4cb5300b | 1233 | btrfs_delalloc_release_space(inode, |
09cbfeaf KS |
1234 | start_index << PAGE_SHIFT, |
1235 | (page_cnt - i_done) << PAGE_SHIFT); | |
4cb5300b CM |
1236 | } |
1237 | ||
1238 | ||
9e8a4a8b LB |
1239 | set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, |
1240 | &cached_state, GFP_NOFS); | |
4cb5300b CM |
1241 | |
1242 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, | |
1243 | page_start, page_end - 1, &cached_state, | |
1244 | GFP_NOFS); | |
1245 | ||
1246 | for (i = 0; i < i_done; i++) { | |
1247 | clear_page_dirty_for_io(pages[i]); | |
1248 | ClearPageChecked(pages[i]); | |
1249 | set_page_extent_mapped(pages[i]); | |
1250 | set_page_dirty(pages[i]); | |
1251 | unlock_page(pages[i]); | |
09cbfeaf | 1252 | put_page(pages[i]); |
4cb5300b CM |
1253 | } |
1254 | return i_done; | |
1255 | out: | |
1256 | for (i = 0; i < i_done; i++) { | |
1257 | unlock_page(pages[i]); | |
09cbfeaf | 1258 | put_page(pages[i]); |
4cb5300b | 1259 | } |
7cf5b976 | 1260 | btrfs_delalloc_release_space(inode, |
09cbfeaf KS |
1261 | start_index << PAGE_SHIFT, |
1262 | page_cnt << PAGE_SHIFT); | |
4cb5300b CM |
1263 | return ret; |
1264 | ||
1265 | } | |
1266 | ||
1267 | int btrfs_defrag_file(struct inode *inode, struct file *file, | |
1268 | struct btrfs_ioctl_defrag_range_args *range, | |
1269 | u64 newer_than, unsigned long max_to_defrag) | |
1270 | { | |
1271 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4cb5300b | 1272 | struct file_ra_state *ra = NULL; |
f46b5a66 | 1273 | unsigned long last_index; |
151a31b2 | 1274 | u64 isize = i_size_read(inode); |
940100a4 CM |
1275 | u64 last_len = 0; |
1276 | u64 skip = 0; | |
1277 | u64 defrag_end = 0; | |
4cb5300b | 1278 | u64 newer_off = range->start; |
f46b5a66 | 1279 | unsigned long i; |
008873ea | 1280 | unsigned long ra_index = 0; |
f46b5a66 | 1281 | int ret; |
4cb5300b | 1282 | int defrag_count = 0; |
1a419d85 | 1283 | int compress_type = BTRFS_COMPRESS_ZLIB; |
aab110ab | 1284 | u32 extent_thresh = range->extent_thresh; |
09cbfeaf | 1285 | unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; |
c41570c9 | 1286 | unsigned long cluster = max_cluster; |
ee22184b | 1287 | u64 new_align = ~((u64)SZ_128K - 1); |
4cb5300b CM |
1288 | struct page **pages = NULL; |
1289 | ||
0abd5b17 LB |
1290 | if (isize == 0) |
1291 | return 0; | |
1292 | ||
1293 | if (range->start >= isize) | |
1294 | return -EINVAL; | |
1a419d85 LZ |
1295 | |
1296 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { | |
1297 | if (range->compress_type > BTRFS_COMPRESS_TYPES) | |
1298 | return -EINVAL; | |
1299 | if (range->compress_type) | |
1300 | compress_type = range->compress_type; | |
1301 | } | |
f46b5a66 | 1302 | |
0abd5b17 | 1303 | if (extent_thresh == 0) |
ee22184b | 1304 | extent_thresh = SZ_256K; |
940100a4 | 1305 | |
4cb5300b CM |
1306 | /* |
1307 | * if we were not given a file, allocate a readahead | |
1308 | * context | |
1309 | */ | |
1310 | if (!file) { | |
1311 | ra = kzalloc(sizeof(*ra), GFP_NOFS); | |
1312 | if (!ra) | |
1313 | return -ENOMEM; | |
1314 | file_ra_state_init(ra, inode->i_mapping); | |
1315 | } else { | |
1316 | ra = &file->f_ra; | |
1317 | } | |
1318 | ||
d9b0d9ba | 1319 | pages = kmalloc_array(max_cluster, sizeof(struct page *), |
4cb5300b CM |
1320 | GFP_NOFS); |
1321 | if (!pages) { | |
1322 | ret = -ENOMEM; | |
1323 | goto out_ra; | |
1324 | } | |
1325 | ||
1326 | /* find the last page to defrag */ | |
1e701a32 | 1327 | if (range->start + range->len > range->start) { |
151a31b2 | 1328 | last_index = min_t(u64, isize - 1, |
09cbfeaf | 1329 | range->start + range->len - 1) >> PAGE_SHIFT; |
1e701a32 | 1330 | } else { |
09cbfeaf | 1331 | last_index = (isize - 1) >> PAGE_SHIFT; |
1e701a32 CM |
1332 | } |
1333 | ||
4cb5300b CM |
1334 | if (newer_than) { |
1335 | ret = find_new_extents(root, inode, newer_than, | |
ee22184b | 1336 | &newer_off, SZ_64K); |
4cb5300b CM |
1337 | if (!ret) { |
1338 | range->start = newer_off; | |
1339 | /* | |
1340 | * we always align our defrag to help keep | |
1341 | * the extents in the file evenly spaced | |
1342 | */ | |
09cbfeaf | 1343 | i = (newer_off & new_align) >> PAGE_SHIFT; |
4cb5300b CM |
1344 | } else |
1345 | goto out_ra; | |
1346 | } else { | |
09cbfeaf | 1347 | i = range->start >> PAGE_SHIFT; |
4cb5300b CM |
1348 | } |
1349 | if (!max_to_defrag) | |
070034bd | 1350 | max_to_defrag = last_index - i + 1; |
4cb5300b | 1351 | |
2a0f7f57 LZ |
1352 | /* |
1353 | * make writeback starts from i, so the defrag range can be | |
1354 | * written sequentially. | |
1355 | */ | |
1356 | if (i < inode->i_mapping->writeback_index) | |
1357 | inode->i_mapping->writeback_index = i; | |
1358 | ||
f7f43cc8 | 1359 | while (i <= last_index && defrag_count < max_to_defrag && |
09cbfeaf | 1360 | (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { |
4cb5300b CM |
1361 | /* |
1362 | * make sure we stop running if someone unmounts | |
1363 | * the FS | |
1364 | */ | |
1365 | if (!(inode->i_sb->s_flags & MS_ACTIVE)) | |
1366 | break; | |
1367 | ||
210549eb | 1368 | if (btrfs_defrag_cancelled(root->fs_info)) { |
f14d104d | 1369 | btrfs_debug(root->fs_info, "defrag_file cancelled"); |
210549eb DS |
1370 | ret = -EAGAIN; |
1371 | break; | |
1372 | } | |
1373 | ||
09cbfeaf | 1374 | if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, |
6c282eb4 | 1375 | extent_thresh, &last_len, &skip, |
a43a2111 AM |
1376 | &defrag_end, range->flags & |
1377 | BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
940100a4 CM |
1378 | unsigned long next; |
1379 | /* | |
1380 | * the should_defrag function tells us how much to skip | |
1381 | * bump our counter by the suggested amount | |
1382 | */ | |
09cbfeaf | 1383 | next = DIV_ROUND_UP(skip, PAGE_SIZE); |
940100a4 CM |
1384 | i = max(i + 1, next); |
1385 | continue; | |
1386 | } | |
008873ea LZ |
1387 | |
1388 | if (!newer_than) { | |
09cbfeaf KS |
1389 | cluster = (PAGE_ALIGN(defrag_end) >> |
1390 | PAGE_SHIFT) - i; | |
008873ea LZ |
1391 | cluster = min(cluster, max_cluster); |
1392 | } else { | |
1393 | cluster = max_cluster; | |
1394 | } | |
1395 | ||
008873ea LZ |
1396 | if (i + cluster > ra_index) { |
1397 | ra_index = max(i, ra_index); | |
1398 | btrfs_force_ra(inode->i_mapping, ra, file, ra_index, | |
1399 | cluster); | |
e4826a5b | 1400 | ra_index += cluster; |
008873ea | 1401 | } |
940100a4 | 1402 | |
5955102c | 1403 | inode_lock(inode); |
633085c7 FDBM |
1404 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) |
1405 | BTRFS_I(inode)->force_compress = compress_type; | |
008873ea | 1406 | ret = cluster_pages_for_defrag(inode, pages, i, cluster); |
ecb8bea8 | 1407 | if (ret < 0) { |
5955102c | 1408 | inode_unlock(inode); |
4cb5300b | 1409 | goto out_ra; |
ecb8bea8 | 1410 | } |
4cb5300b CM |
1411 | |
1412 | defrag_count += ret; | |
d0e1d66b | 1413 | balance_dirty_pages_ratelimited(inode->i_mapping); |
5955102c | 1414 | inode_unlock(inode); |
4cb5300b CM |
1415 | |
1416 | if (newer_than) { | |
1417 | if (newer_off == (u64)-1) | |
1418 | break; | |
1419 | ||
e1f041e1 LB |
1420 | if (ret > 0) |
1421 | i += ret; | |
1422 | ||
4cb5300b | 1423 | newer_off = max(newer_off + 1, |
09cbfeaf | 1424 | (u64)i << PAGE_SHIFT); |
4cb5300b | 1425 | |
ee22184b BL |
1426 | ret = find_new_extents(root, inode, newer_than, |
1427 | &newer_off, SZ_64K); | |
4cb5300b CM |
1428 | if (!ret) { |
1429 | range->start = newer_off; | |
09cbfeaf | 1430 | i = (newer_off & new_align) >> PAGE_SHIFT; |
4cb5300b CM |
1431 | } else { |
1432 | break; | |
f46b5a66 | 1433 | } |
4cb5300b | 1434 | } else { |
008873ea | 1435 | if (ret > 0) { |
cbcc8326 | 1436 | i += ret; |
09cbfeaf | 1437 | last_len += ret << PAGE_SHIFT; |
008873ea | 1438 | } else { |
cbcc8326 | 1439 | i++; |
008873ea LZ |
1440 | last_len = 0; |
1441 | } | |
f46b5a66 | 1442 | } |
f46b5a66 CH |
1443 | } |
1444 | ||
dec8ef90 | 1445 | if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { |
1e701a32 | 1446 | filemap_flush(inode->i_mapping); |
dec8ef90 FM |
1447 | if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, |
1448 | &BTRFS_I(inode)->runtime_flags)) | |
1449 | filemap_flush(inode->i_mapping); | |
1450 | } | |
1e701a32 CM |
1451 | |
1452 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
1453 | /* the filemap_flush will queue IO into the worker threads, but | |
1454 | * we have to make sure the IO is actually started and that | |
1455 | * ordered extents get created before we return | |
1456 | */ | |
1457 | atomic_inc(&root->fs_info->async_submit_draining); | |
1458 | while (atomic_read(&root->fs_info->nr_async_submits) || | |
1459 | atomic_read(&root->fs_info->async_delalloc_pages)) { | |
1460 | wait_event(root->fs_info->async_submit_wait, | |
1461 | (atomic_read(&root->fs_info->nr_async_submits) == 0 && | |
1462 | atomic_read(&root->fs_info->async_delalloc_pages) == 0)); | |
1463 | } | |
1464 | atomic_dec(&root->fs_info->async_submit_draining); | |
1e701a32 CM |
1465 | } |
1466 | ||
1a419d85 | 1467 | if (range->compress_type == BTRFS_COMPRESS_LZO) { |
2b0ce2c2 | 1468 | btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO); |
1a419d85 LZ |
1469 | } |
1470 | ||
60ccf82f | 1471 | ret = defrag_count; |
940100a4 | 1472 | |
4cb5300b | 1473 | out_ra: |
633085c7 | 1474 | if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { |
5955102c | 1475 | inode_lock(inode); |
633085c7 | 1476 | BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; |
5955102c | 1477 | inode_unlock(inode); |
633085c7 | 1478 | } |
4cb5300b CM |
1479 | if (!file) |
1480 | kfree(ra); | |
1481 | kfree(pages); | |
940100a4 | 1482 | return ret; |
f46b5a66 CH |
1483 | } |
1484 | ||
198605a8 | 1485 | static noinline int btrfs_ioctl_resize(struct file *file, |
76dda93c | 1486 | void __user *arg) |
f46b5a66 CH |
1487 | { |
1488 | u64 new_size; | |
1489 | u64 old_size; | |
1490 | u64 devid = 1; | |
496ad9aa | 1491 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
f46b5a66 CH |
1492 | struct btrfs_ioctl_vol_args *vol_args; |
1493 | struct btrfs_trans_handle *trans; | |
1494 | struct btrfs_device *device = NULL; | |
1495 | char *sizestr; | |
9a40f122 | 1496 | char *retptr; |
f46b5a66 CH |
1497 | char *devstr = NULL; |
1498 | int ret = 0; | |
f46b5a66 CH |
1499 | int mod = 0; |
1500 | ||
e441d54d CM |
1501 | if (!capable(CAP_SYS_ADMIN)) |
1502 | return -EPERM; | |
1503 | ||
198605a8 MX |
1504 | ret = mnt_want_write_file(file); |
1505 | if (ret) | |
1506 | return ret; | |
1507 | ||
5ac00add SB |
1508 | if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, |
1509 | 1)) { | |
97547676 | 1510 | mnt_drop_write_file(file); |
e57138b3 | 1511 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
c9e9f97b ID |
1512 | } |
1513 | ||
5ac00add | 1514 | mutex_lock(&root->fs_info->volume_mutex); |
dae7b665 | 1515 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
1516 | if (IS_ERR(vol_args)) { |
1517 | ret = PTR_ERR(vol_args); | |
1518 | goto out; | |
1519 | } | |
5516e595 MF |
1520 | |
1521 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
f46b5a66 | 1522 | |
f46b5a66 CH |
1523 | sizestr = vol_args->name; |
1524 | devstr = strchr(sizestr, ':'); | |
1525 | if (devstr) { | |
f46b5a66 CH |
1526 | sizestr = devstr + 1; |
1527 | *devstr = '\0'; | |
1528 | devstr = vol_args->name; | |
58dfae63 Z |
1529 | ret = kstrtoull(devstr, 10, &devid); |
1530 | if (ret) | |
1531 | goto out_free; | |
dfd79829 MX |
1532 | if (!devid) { |
1533 | ret = -EINVAL; | |
1534 | goto out_free; | |
1535 | } | |
efe120a0 | 1536 | btrfs_info(root->fs_info, "resizing devid %llu", devid); |
f46b5a66 | 1537 | } |
dba60f3f | 1538 | |
aa1b8cd4 | 1539 | device = btrfs_find_device(root->fs_info, devid, NULL, NULL); |
f46b5a66 | 1540 | if (!device) { |
efe120a0 | 1541 | btrfs_info(root->fs_info, "resizer unable to find device %llu", |
c1c9ff7c | 1542 | devid); |
dfd79829 | 1543 | ret = -ENODEV; |
c9e9f97b | 1544 | goto out_free; |
f46b5a66 | 1545 | } |
dba60f3f MX |
1546 | |
1547 | if (!device->writeable) { | |
efe120a0 FH |
1548 | btrfs_info(root->fs_info, |
1549 | "resizer unable to apply on readonly device %llu", | |
c1c9ff7c | 1550 | devid); |
dfd79829 | 1551 | ret = -EPERM; |
4e42ae1b LB |
1552 | goto out_free; |
1553 | } | |
1554 | ||
f46b5a66 CH |
1555 | if (!strcmp(sizestr, "max")) |
1556 | new_size = device->bdev->bd_inode->i_size; | |
1557 | else { | |
1558 | if (sizestr[0] == '-') { | |
1559 | mod = -1; | |
1560 | sizestr++; | |
1561 | } else if (sizestr[0] == '+') { | |
1562 | mod = 1; | |
1563 | sizestr++; | |
1564 | } | |
9a40f122 GH |
1565 | new_size = memparse(sizestr, &retptr); |
1566 | if (*retptr != '\0' || new_size == 0) { | |
f46b5a66 | 1567 | ret = -EINVAL; |
c9e9f97b | 1568 | goto out_free; |
f46b5a66 CH |
1569 | } |
1570 | } | |
1571 | ||
63a212ab | 1572 | if (device->is_tgtdev_for_dev_replace) { |
dfd79829 | 1573 | ret = -EPERM; |
63a212ab SB |
1574 | goto out_free; |
1575 | } | |
1576 | ||
7cc8e58d | 1577 | old_size = btrfs_device_get_total_bytes(device); |
f46b5a66 CH |
1578 | |
1579 | if (mod < 0) { | |
1580 | if (new_size > old_size) { | |
1581 | ret = -EINVAL; | |
c9e9f97b | 1582 | goto out_free; |
f46b5a66 CH |
1583 | } |
1584 | new_size = old_size - new_size; | |
1585 | } else if (mod > 0) { | |
eb8052e0 | 1586 | if (new_size > ULLONG_MAX - old_size) { |
902c68a4 | 1587 | ret = -ERANGE; |
eb8052e0 WF |
1588 | goto out_free; |
1589 | } | |
f46b5a66 CH |
1590 | new_size = old_size + new_size; |
1591 | } | |
1592 | ||
ee22184b | 1593 | if (new_size < SZ_256M) { |
f46b5a66 | 1594 | ret = -EINVAL; |
c9e9f97b | 1595 | goto out_free; |
f46b5a66 CH |
1596 | } |
1597 | if (new_size > device->bdev->bd_inode->i_size) { | |
1598 | ret = -EFBIG; | |
c9e9f97b | 1599 | goto out_free; |
f46b5a66 CH |
1600 | } |
1601 | ||
b8b93add | 1602 | new_size = div_u64(new_size, root->sectorsize); |
f46b5a66 CH |
1603 | new_size *= root->sectorsize; |
1604 | ||
ecaeb14b | 1605 | btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu", |
c1c9ff7c | 1606 | rcu_str_deref(device->name), new_size); |
f46b5a66 CH |
1607 | |
1608 | if (new_size > old_size) { | |
a22285a6 | 1609 | trans = btrfs_start_transaction(root, 0); |
98d5dc13 TI |
1610 | if (IS_ERR(trans)) { |
1611 | ret = PTR_ERR(trans); | |
c9e9f97b | 1612 | goto out_free; |
98d5dc13 | 1613 | } |
f46b5a66 CH |
1614 | ret = btrfs_grow_device(trans, device, new_size); |
1615 | btrfs_commit_transaction(trans, root); | |
ece7d20e | 1616 | } else if (new_size < old_size) { |
f46b5a66 | 1617 | ret = btrfs_shrink_device(device, new_size); |
0253f40e | 1618 | } /* equal, nothing need to do */ |
f46b5a66 | 1619 | |
c9e9f97b | 1620 | out_free: |
f46b5a66 | 1621 | kfree(vol_args); |
c9e9f97b ID |
1622 | out: |
1623 | mutex_unlock(&root->fs_info->volume_mutex); | |
5ac00add | 1624 | atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); |
18f39c41 | 1625 | mnt_drop_write_file(file); |
f46b5a66 CH |
1626 | return ret; |
1627 | } | |
1628 | ||
72fd032e | 1629 | static noinline int btrfs_ioctl_snap_create_transid(struct file *file, |
6f72c7e2 AJ |
1630 | char *name, unsigned long fd, int subvol, |
1631 | u64 *transid, bool readonly, | |
8696c533 | 1632 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 1633 | { |
f46b5a66 | 1634 | int namelen; |
3de4586c | 1635 | int ret = 0; |
f46b5a66 | 1636 | |
a874a63e LB |
1637 | ret = mnt_want_write_file(file); |
1638 | if (ret) | |
1639 | goto out; | |
1640 | ||
72fd032e SW |
1641 | namelen = strlen(name); |
1642 | if (strchr(name, '/')) { | |
f46b5a66 | 1643 | ret = -EINVAL; |
a874a63e | 1644 | goto out_drop_write; |
f46b5a66 CH |
1645 | } |
1646 | ||
16780cab CM |
1647 | if (name[0] == '.' && |
1648 | (namelen == 1 || (name[1] == '.' && namelen == 2))) { | |
1649 | ret = -EEXIST; | |
a874a63e | 1650 | goto out_drop_write; |
16780cab CM |
1651 | } |
1652 | ||
3de4586c | 1653 | if (subvol) { |
72fd032e | 1654 | ret = btrfs_mksubvol(&file->f_path, name, namelen, |
6f72c7e2 | 1655 | NULL, transid, readonly, inherit); |
cb8e7090 | 1656 | } else { |
2903ff01 | 1657 | struct fd src = fdget(fd); |
3de4586c | 1658 | struct inode *src_inode; |
2903ff01 | 1659 | if (!src.file) { |
3de4586c | 1660 | ret = -EINVAL; |
a874a63e | 1661 | goto out_drop_write; |
3de4586c CM |
1662 | } |
1663 | ||
496ad9aa AV |
1664 | src_inode = file_inode(src.file); |
1665 | if (src_inode->i_sb != file_inode(file)->i_sb) { | |
c79b4713 | 1666 | btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, |
efe120a0 | 1667 | "Snapshot src from another FS"); |
23ad5b17 | 1668 | ret = -EXDEV; |
d0242061 DS |
1669 | } else if (!inode_owner_or_capable(src_inode)) { |
1670 | /* | |
1671 | * Subvolume creation is not restricted, but snapshots | |
1672 | * are limited to own subvolumes only | |
1673 | */ | |
1674 | ret = -EPERM; | |
ecd18815 AV |
1675 | } else { |
1676 | ret = btrfs_mksubvol(&file->f_path, name, namelen, | |
1677 | BTRFS_I(src_inode)->root, | |
1678 | transid, readonly, inherit); | |
3de4586c | 1679 | } |
2903ff01 | 1680 | fdput(src); |
cb8e7090 | 1681 | } |
a874a63e LB |
1682 | out_drop_write: |
1683 | mnt_drop_write_file(file); | |
f46b5a66 | 1684 | out: |
72fd032e SW |
1685 | return ret; |
1686 | } | |
1687 | ||
1688 | static noinline int btrfs_ioctl_snap_create(struct file *file, | |
fa0d2b9b | 1689 | void __user *arg, int subvol) |
72fd032e | 1690 | { |
fa0d2b9b | 1691 | struct btrfs_ioctl_vol_args *vol_args; |
72fd032e SW |
1692 | int ret; |
1693 | ||
fa0d2b9b LZ |
1694 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1695 | if (IS_ERR(vol_args)) | |
1696 | return PTR_ERR(vol_args); | |
1697 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
72fd032e | 1698 | |
fa0d2b9b | 1699 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, |
b83cc969 | 1700 | vol_args->fd, subvol, |
6f72c7e2 | 1701 | NULL, false, NULL); |
fdfb1e4f | 1702 | |
fa0d2b9b LZ |
1703 | kfree(vol_args); |
1704 | return ret; | |
1705 | } | |
fdfb1e4f | 1706 | |
fa0d2b9b LZ |
1707 | static noinline int btrfs_ioctl_snap_create_v2(struct file *file, |
1708 | void __user *arg, int subvol) | |
1709 | { | |
1710 | struct btrfs_ioctl_vol_args_v2 *vol_args; | |
1711 | int ret; | |
1712 | u64 transid = 0; | |
1713 | u64 *ptr = NULL; | |
b83cc969 | 1714 | bool readonly = false; |
6f72c7e2 | 1715 | struct btrfs_qgroup_inherit *inherit = NULL; |
75eaa0e2 | 1716 | |
fa0d2b9b LZ |
1717 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1718 | if (IS_ERR(vol_args)) | |
1719 | return PTR_ERR(vol_args); | |
1720 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; | |
75eaa0e2 | 1721 | |
b83cc969 | 1722 | if (vol_args->flags & |
6f72c7e2 AJ |
1723 | ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | |
1724 | BTRFS_SUBVOL_QGROUP_INHERIT)) { | |
b83cc969 | 1725 | ret = -EOPNOTSUPP; |
c47ca32d | 1726 | goto free_args; |
72fd032e | 1727 | } |
fa0d2b9b LZ |
1728 | |
1729 | if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) | |
1730 | ptr = &transid; | |
b83cc969 LZ |
1731 | if (vol_args->flags & BTRFS_SUBVOL_RDONLY) |
1732 | readonly = true; | |
6f72c7e2 | 1733 | if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { |
09cbfeaf | 1734 | if (vol_args->size > PAGE_SIZE) { |
6f72c7e2 | 1735 | ret = -EINVAL; |
c47ca32d | 1736 | goto free_args; |
6f72c7e2 AJ |
1737 | } |
1738 | inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); | |
1739 | if (IS_ERR(inherit)) { | |
1740 | ret = PTR_ERR(inherit); | |
c47ca32d | 1741 | goto free_args; |
6f72c7e2 AJ |
1742 | } |
1743 | } | |
fa0d2b9b LZ |
1744 | |
1745 | ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, | |
6f72c7e2 | 1746 | vol_args->fd, subvol, ptr, |
8696c533 | 1747 | readonly, inherit); |
c47ca32d DC |
1748 | if (ret) |
1749 | goto free_inherit; | |
fa0d2b9b | 1750 | |
c47ca32d DC |
1751 | if (ptr && copy_to_user(arg + |
1752 | offsetof(struct btrfs_ioctl_vol_args_v2, | |
1753 | transid), | |
1754 | ptr, sizeof(*ptr))) | |
fa0d2b9b | 1755 | ret = -EFAULT; |
c47ca32d DC |
1756 | |
1757 | free_inherit: | |
6f72c7e2 | 1758 | kfree(inherit); |
c47ca32d DC |
1759 | free_args: |
1760 | kfree(vol_args); | |
f46b5a66 CH |
1761 | return ret; |
1762 | } | |
1763 | ||
0caa102d LZ |
1764 | static noinline int btrfs_ioctl_subvol_getflags(struct file *file, |
1765 | void __user *arg) | |
1766 | { | |
496ad9aa | 1767 | struct inode *inode = file_inode(file); |
0caa102d LZ |
1768 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1769 | int ret = 0; | |
1770 | u64 flags = 0; | |
1771 | ||
33345d01 | 1772 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) |
0caa102d LZ |
1773 | return -EINVAL; |
1774 | ||
1775 | down_read(&root->fs_info->subvol_sem); | |
1776 | if (btrfs_root_readonly(root)) | |
1777 | flags |= BTRFS_SUBVOL_RDONLY; | |
1778 | up_read(&root->fs_info->subvol_sem); | |
1779 | ||
1780 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
1781 | ret = -EFAULT; | |
1782 | ||
1783 | return ret; | |
1784 | } | |
1785 | ||
1786 | static noinline int btrfs_ioctl_subvol_setflags(struct file *file, | |
1787 | void __user *arg) | |
1788 | { | |
496ad9aa | 1789 | struct inode *inode = file_inode(file); |
0caa102d LZ |
1790 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1791 | struct btrfs_trans_handle *trans; | |
1792 | u64 root_flags; | |
1793 | u64 flags; | |
1794 | int ret = 0; | |
1795 | ||
bd60ea0f DS |
1796 | if (!inode_owner_or_capable(inode)) |
1797 | return -EPERM; | |
1798 | ||
b9ca0664 LB |
1799 | ret = mnt_want_write_file(file); |
1800 | if (ret) | |
1801 | goto out; | |
0caa102d | 1802 | |
b9ca0664 LB |
1803 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { |
1804 | ret = -EINVAL; | |
1805 | goto out_drop_write; | |
1806 | } | |
0caa102d | 1807 | |
b9ca0664 LB |
1808 | if (copy_from_user(&flags, arg, sizeof(flags))) { |
1809 | ret = -EFAULT; | |
1810 | goto out_drop_write; | |
1811 | } | |
0caa102d | 1812 | |
b9ca0664 LB |
1813 | if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { |
1814 | ret = -EINVAL; | |
1815 | goto out_drop_write; | |
1816 | } | |
0caa102d | 1817 | |
b9ca0664 LB |
1818 | if (flags & ~BTRFS_SUBVOL_RDONLY) { |
1819 | ret = -EOPNOTSUPP; | |
1820 | goto out_drop_write; | |
1821 | } | |
0caa102d LZ |
1822 | |
1823 | down_write(&root->fs_info->subvol_sem); | |
1824 | ||
1825 | /* nothing to do */ | |
1826 | if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) | |
b9ca0664 | 1827 | goto out_drop_sem; |
0caa102d LZ |
1828 | |
1829 | root_flags = btrfs_root_flags(&root->root_item); | |
2c686537 | 1830 | if (flags & BTRFS_SUBVOL_RDONLY) { |
0caa102d LZ |
1831 | btrfs_set_root_flags(&root->root_item, |
1832 | root_flags | BTRFS_ROOT_SUBVOL_RDONLY); | |
2c686537 DS |
1833 | } else { |
1834 | /* | |
1835 | * Block RO -> RW transition if this subvolume is involved in | |
1836 | * send | |
1837 | */ | |
1838 | spin_lock(&root->root_item_lock); | |
1839 | if (root->send_in_progress == 0) { | |
1840 | btrfs_set_root_flags(&root->root_item, | |
0caa102d | 1841 | root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); |
2c686537 DS |
1842 | spin_unlock(&root->root_item_lock); |
1843 | } else { | |
1844 | spin_unlock(&root->root_item_lock); | |
1845 | btrfs_warn(root->fs_info, | |
1846 | "Attempt to set subvolume %llu read-write during send", | |
1847 | root->root_key.objectid); | |
1848 | ret = -EPERM; | |
1849 | goto out_drop_sem; | |
1850 | } | |
1851 | } | |
0caa102d LZ |
1852 | |
1853 | trans = btrfs_start_transaction(root, 1); | |
1854 | if (IS_ERR(trans)) { | |
1855 | ret = PTR_ERR(trans); | |
1856 | goto out_reset; | |
1857 | } | |
1858 | ||
b4dc2b8c | 1859 | ret = btrfs_update_root(trans, root->fs_info->tree_root, |
0caa102d LZ |
1860 | &root->root_key, &root->root_item); |
1861 | ||
1862 | btrfs_commit_transaction(trans, root); | |
1863 | out_reset: | |
1864 | if (ret) | |
1865 | btrfs_set_root_flags(&root->root_item, root_flags); | |
b9ca0664 | 1866 | out_drop_sem: |
0caa102d | 1867 | up_write(&root->fs_info->subvol_sem); |
b9ca0664 LB |
1868 | out_drop_write: |
1869 | mnt_drop_write_file(file); | |
1870 | out: | |
0caa102d LZ |
1871 | return ret; |
1872 | } | |
1873 | ||
76dda93c YZ |
1874 | /* |
1875 | * helper to check if the subvolume references other subvolumes | |
1876 | */ | |
1877 | static noinline int may_destroy_subvol(struct btrfs_root *root) | |
1878 | { | |
1879 | struct btrfs_path *path; | |
175a2b87 | 1880 | struct btrfs_dir_item *di; |
76dda93c | 1881 | struct btrfs_key key; |
175a2b87 | 1882 | u64 dir_id; |
76dda93c YZ |
1883 | int ret; |
1884 | ||
1885 | path = btrfs_alloc_path(); | |
1886 | if (!path) | |
1887 | return -ENOMEM; | |
1888 | ||
175a2b87 JB |
1889 | /* Make sure this root isn't set as the default subvol */ |
1890 | dir_id = btrfs_super_root_dir(root->fs_info->super_copy); | |
1891 | di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, | |
1892 | dir_id, "default", 7, 0); | |
1893 | if (di && !IS_ERR(di)) { | |
1894 | btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); | |
1895 | if (key.objectid == root->root_key.objectid) { | |
72de6b53 GS |
1896 | ret = -EPERM; |
1897 | btrfs_err(root->fs_info, "deleting default subvolume " | |
1898 | "%llu is not allowed", key.objectid); | |
175a2b87 JB |
1899 | goto out; |
1900 | } | |
1901 | btrfs_release_path(path); | |
1902 | } | |
1903 | ||
76dda93c YZ |
1904 | key.objectid = root->root_key.objectid; |
1905 | key.type = BTRFS_ROOT_REF_KEY; | |
1906 | key.offset = (u64)-1; | |
1907 | ||
1908 | ret = btrfs_search_slot(NULL, root->fs_info->tree_root, | |
1909 | &key, path, 0, 0); | |
1910 | if (ret < 0) | |
1911 | goto out; | |
1912 | BUG_ON(ret == 0); | |
1913 | ||
1914 | ret = 0; | |
1915 | if (path->slots[0] > 0) { | |
1916 | path->slots[0]--; | |
1917 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
1918 | if (key.objectid == root->root_key.objectid && | |
1919 | key.type == BTRFS_ROOT_REF_KEY) | |
1920 | ret = -ENOTEMPTY; | |
1921 | } | |
1922 | out: | |
1923 | btrfs_free_path(path); | |
1924 | return ret; | |
1925 | } | |
1926 | ||
ac8e9819 CM |
1927 | static noinline int key_in_sk(struct btrfs_key *key, |
1928 | struct btrfs_ioctl_search_key *sk) | |
1929 | { | |
abc6e134 CM |
1930 | struct btrfs_key test; |
1931 | int ret; | |
1932 | ||
1933 | test.objectid = sk->min_objectid; | |
1934 | test.type = sk->min_type; | |
1935 | test.offset = sk->min_offset; | |
1936 | ||
1937 | ret = btrfs_comp_cpu_keys(key, &test); | |
1938 | if (ret < 0) | |
ac8e9819 | 1939 | return 0; |
abc6e134 CM |
1940 | |
1941 | test.objectid = sk->max_objectid; | |
1942 | test.type = sk->max_type; | |
1943 | test.offset = sk->max_offset; | |
1944 | ||
1945 | ret = btrfs_comp_cpu_keys(key, &test); | |
1946 | if (ret > 0) | |
ac8e9819 CM |
1947 | return 0; |
1948 | return 1; | |
1949 | } | |
1950 | ||
1951 | static noinline int copy_to_sk(struct btrfs_root *root, | |
1952 | struct btrfs_path *path, | |
1953 | struct btrfs_key *key, | |
1954 | struct btrfs_ioctl_search_key *sk, | |
9b6e817d | 1955 | size_t *buf_size, |
ba346b35 | 1956 | char __user *ubuf, |
ac8e9819 CM |
1957 | unsigned long *sk_offset, |
1958 | int *num_found) | |
1959 | { | |
1960 | u64 found_transid; | |
1961 | struct extent_buffer *leaf; | |
1962 | struct btrfs_ioctl_search_header sh; | |
dd81d459 | 1963 | struct btrfs_key test; |
ac8e9819 CM |
1964 | unsigned long item_off; |
1965 | unsigned long item_len; | |
1966 | int nritems; | |
1967 | int i; | |
1968 | int slot; | |
ac8e9819 CM |
1969 | int ret = 0; |
1970 | ||
1971 | leaf = path->nodes[0]; | |
1972 | slot = path->slots[0]; | |
1973 | nritems = btrfs_header_nritems(leaf); | |
1974 | ||
1975 | if (btrfs_header_generation(leaf) > sk->max_transid) { | |
1976 | i = nritems; | |
1977 | goto advance_key; | |
1978 | } | |
1979 | found_transid = btrfs_header_generation(leaf); | |
1980 | ||
1981 | for (i = slot; i < nritems; i++) { | |
1982 | item_off = btrfs_item_ptr_offset(leaf, i); | |
1983 | item_len = btrfs_item_size_nr(leaf, i); | |
1984 | ||
03b71c6c GP |
1985 | btrfs_item_key_to_cpu(leaf, key, i); |
1986 | if (!key_in_sk(key, sk)) | |
1987 | continue; | |
1988 | ||
9b6e817d | 1989 | if (sizeof(sh) + item_len > *buf_size) { |
8f5f6178 GH |
1990 | if (*num_found) { |
1991 | ret = 1; | |
1992 | goto out; | |
1993 | } | |
1994 | ||
1995 | /* | |
1996 | * return one empty item back for v1, which does not | |
1997 | * handle -EOVERFLOW | |
1998 | */ | |
1999 | ||
9b6e817d | 2000 | *buf_size = sizeof(sh) + item_len; |
ac8e9819 | 2001 | item_len = 0; |
8f5f6178 GH |
2002 | ret = -EOVERFLOW; |
2003 | } | |
ac8e9819 | 2004 | |
9b6e817d | 2005 | if (sizeof(sh) + item_len + *sk_offset > *buf_size) { |
ac8e9819 | 2006 | ret = 1; |
25c9bc2e | 2007 | goto out; |
ac8e9819 CM |
2008 | } |
2009 | ||
ac8e9819 CM |
2010 | sh.objectid = key->objectid; |
2011 | sh.offset = key->offset; | |
2012 | sh.type = key->type; | |
2013 | sh.len = item_len; | |
2014 | sh.transid = found_transid; | |
2015 | ||
2016 | /* copy search result header */ | |
ba346b35 GH |
2017 | if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { |
2018 | ret = -EFAULT; | |
2019 | goto out; | |
2020 | } | |
2021 | ||
ac8e9819 CM |
2022 | *sk_offset += sizeof(sh); |
2023 | ||
2024 | if (item_len) { | |
ba346b35 | 2025 | char __user *up = ubuf + *sk_offset; |
ac8e9819 | 2026 | /* copy the item */ |
ba346b35 GH |
2027 | if (read_extent_buffer_to_user(leaf, up, |
2028 | item_off, item_len)) { | |
2029 | ret = -EFAULT; | |
2030 | goto out; | |
2031 | } | |
2032 | ||
ac8e9819 | 2033 | *sk_offset += item_len; |
ac8e9819 | 2034 | } |
e2156867 | 2035 | (*num_found)++; |
ac8e9819 | 2036 | |
8f5f6178 GH |
2037 | if (ret) /* -EOVERFLOW from above */ |
2038 | goto out; | |
2039 | ||
25c9bc2e GH |
2040 | if (*num_found >= sk->nr_items) { |
2041 | ret = 1; | |
2042 | goto out; | |
2043 | } | |
ac8e9819 CM |
2044 | } |
2045 | advance_key: | |
abc6e134 | 2046 | ret = 0; |
dd81d459 NA |
2047 | test.objectid = sk->max_objectid; |
2048 | test.type = sk->max_type; | |
2049 | test.offset = sk->max_offset; | |
2050 | if (btrfs_comp_cpu_keys(key, &test) >= 0) | |
2051 | ret = 1; | |
2052 | else if (key->offset < (u64)-1) | |
ac8e9819 | 2053 | key->offset++; |
dd81d459 | 2054 | else if (key->type < (u8)-1) { |
abc6e134 | 2055 | key->offset = 0; |
ac8e9819 | 2056 | key->type++; |
dd81d459 | 2057 | } else if (key->objectid < (u64)-1) { |
abc6e134 CM |
2058 | key->offset = 0; |
2059 | key->type = 0; | |
ac8e9819 | 2060 | key->objectid++; |
abc6e134 CM |
2061 | } else |
2062 | ret = 1; | |
25c9bc2e | 2063 | out: |
ba346b35 GH |
2064 | /* |
2065 | * 0: all items from this leaf copied, continue with next | |
2066 | * 1: * more items can be copied, but unused buffer is too small | |
2067 | * * all items were found | |
2068 | * Either way, it will stops the loop which iterates to the next | |
2069 | * leaf | |
2070 | * -EOVERFLOW: item was to large for buffer | |
2071 | * -EFAULT: could not copy extent buffer back to userspace | |
2072 | */ | |
ac8e9819 CM |
2073 | return ret; |
2074 | } | |
2075 | ||
2076 | static noinline int search_ioctl(struct inode *inode, | |
12544442 | 2077 | struct btrfs_ioctl_search_key *sk, |
9b6e817d | 2078 | size_t *buf_size, |
ba346b35 | 2079 | char __user *ubuf) |
ac8e9819 CM |
2080 | { |
2081 | struct btrfs_root *root; | |
2082 | struct btrfs_key key; | |
ac8e9819 | 2083 | struct btrfs_path *path; |
ac8e9819 CM |
2084 | struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; |
2085 | int ret; | |
2086 | int num_found = 0; | |
2087 | unsigned long sk_offset = 0; | |
2088 | ||
9b6e817d GH |
2089 | if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { |
2090 | *buf_size = sizeof(struct btrfs_ioctl_search_header); | |
12544442 | 2091 | return -EOVERFLOW; |
9b6e817d | 2092 | } |
12544442 | 2093 | |
ac8e9819 CM |
2094 | path = btrfs_alloc_path(); |
2095 | if (!path) | |
2096 | return -ENOMEM; | |
2097 | ||
2098 | if (sk->tree_id == 0) { | |
2099 | /* search the root of the inode that was passed */ | |
2100 | root = BTRFS_I(inode)->root; | |
2101 | } else { | |
2102 | key.objectid = sk->tree_id; | |
2103 | key.type = BTRFS_ROOT_ITEM_KEY; | |
2104 | key.offset = (u64)-1; | |
2105 | root = btrfs_read_fs_root_no_name(info, &key); | |
2106 | if (IS_ERR(root)) { | |
ac8e9819 CM |
2107 | btrfs_free_path(path); |
2108 | return -ENOENT; | |
2109 | } | |
2110 | } | |
2111 | ||
2112 | key.objectid = sk->min_objectid; | |
2113 | key.type = sk->min_type; | |
2114 | key.offset = sk->min_offset; | |
2115 | ||
67871254 | 2116 | while (1) { |
6174d3cb | 2117 | ret = btrfs_search_forward(root, &key, path, sk->min_transid); |
ac8e9819 CM |
2118 | if (ret != 0) { |
2119 | if (ret > 0) | |
2120 | ret = 0; | |
2121 | goto err; | |
2122 | } | |
ba346b35 | 2123 | ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf, |
ac8e9819 | 2124 | &sk_offset, &num_found); |
b3b4aa74 | 2125 | btrfs_release_path(path); |
25c9bc2e | 2126 | if (ret) |
ac8e9819 CM |
2127 | break; |
2128 | ||
2129 | } | |
8f5f6178 GH |
2130 | if (ret > 0) |
2131 | ret = 0; | |
ac8e9819 CM |
2132 | err: |
2133 | sk->nr_items = num_found; | |
2134 | btrfs_free_path(path); | |
2135 | return ret; | |
2136 | } | |
2137 | ||
2138 | static noinline int btrfs_ioctl_tree_search(struct file *file, | |
2139 | void __user *argp) | |
2140 | { | |
ba346b35 GH |
2141 | struct btrfs_ioctl_search_args __user *uargs; |
2142 | struct btrfs_ioctl_search_key sk; | |
9b6e817d GH |
2143 | struct inode *inode; |
2144 | int ret; | |
2145 | size_t buf_size; | |
ac8e9819 CM |
2146 | |
2147 | if (!capable(CAP_SYS_ADMIN)) | |
2148 | return -EPERM; | |
2149 | ||
ba346b35 GH |
2150 | uargs = (struct btrfs_ioctl_search_args __user *)argp; |
2151 | ||
2152 | if (copy_from_user(&sk, &uargs->key, sizeof(sk))) | |
2153 | return -EFAULT; | |
ac8e9819 | 2154 | |
ba346b35 | 2155 | buf_size = sizeof(uargs->buf); |
ac8e9819 | 2156 | |
496ad9aa | 2157 | inode = file_inode(file); |
ba346b35 | 2158 | ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); |
8f5f6178 GH |
2159 | |
2160 | /* | |
2161 | * In the origin implementation an overflow is handled by returning a | |
2162 | * search header with a len of zero, so reset ret. | |
2163 | */ | |
2164 | if (ret == -EOVERFLOW) | |
2165 | ret = 0; | |
2166 | ||
ba346b35 | 2167 | if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) |
ac8e9819 | 2168 | ret = -EFAULT; |
ac8e9819 CM |
2169 | return ret; |
2170 | } | |
2171 | ||
cc68a8a5 GH |
2172 | static noinline int btrfs_ioctl_tree_search_v2(struct file *file, |
2173 | void __user *argp) | |
2174 | { | |
2175 | struct btrfs_ioctl_search_args_v2 __user *uarg; | |
2176 | struct btrfs_ioctl_search_args_v2 args; | |
2177 | struct inode *inode; | |
2178 | int ret; | |
2179 | size_t buf_size; | |
ee22184b | 2180 | const size_t buf_limit = SZ_16M; |
cc68a8a5 GH |
2181 | |
2182 | if (!capable(CAP_SYS_ADMIN)) | |
2183 | return -EPERM; | |
2184 | ||
2185 | /* copy search header and buffer size */ | |
2186 | uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; | |
2187 | if (copy_from_user(&args, uarg, sizeof(args))) | |
2188 | return -EFAULT; | |
2189 | ||
2190 | buf_size = args.buf_size; | |
2191 | ||
2192 | if (buf_size < sizeof(struct btrfs_ioctl_search_header)) | |
2193 | return -EOVERFLOW; | |
2194 | ||
2195 | /* limit result size to 16MB */ | |
2196 | if (buf_size > buf_limit) | |
2197 | buf_size = buf_limit; | |
2198 | ||
2199 | inode = file_inode(file); | |
2200 | ret = search_ioctl(inode, &args.key, &buf_size, | |
2201 | (char *)(&uarg->buf[0])); | |
2202 | if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) | |
2203 | ret = -EFAULT; | |
2204 | else if (ret == -EOVERFLOW && | |
2205 | copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) | |
2206 | ret = -EFAULT; | |
2207 | ||
ac8e9819 CM |
2208 | return ret; |
2209 | } | |
2210 | ||
98d377a0 | 2211 | /* |
ac8e9819 CM |
2212 | * Search INODE_REFs to identify path name of 'dirid' directory |
2213 | * in a 'tree_id' tree. and sets path name to 'name'. | |
2214 | */ | |
98d377a0 TH |
2215 | static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, |
2216 | u64 tree_id, u64 dirid, char *name) | |
2217 | { | |
2218 | struct btrfs_root *root; | |
2219 | struct btrfs_key key; | |
ac8e9819 | 2220 | char *ptr; |
98d377a0 TH |
2221 | int ret = -1; |
2222 | int slot; | |
2223 | int len; | |
2224 | int total_len = 0; | |
2225 | struct btrfs_inode_ref *iref; | |
2226 | struct extent_buffer *l; | |
2227 | struct btrfs_path *path; | |
2228 | ||
2229 | if (dirid == BTRFS_FIRST_FREE_OBJECTID) { | |
2230 | name[0]='\0'; | |
2231 | return 0; | |
2232 | } | |
2233 | ||
2234 | path = btrfs_alloc_path(); | |
2235 | if (!path) | |
2236 | return -ENOMEM; | |
2237 | ||
ac8e9819 | 2238 | ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; |
98d377a0 TH |
2239 | |
2240 | key.objectid = tree_id; | |
2241 | key.type = BTRFS_ROOT_ITEM_KEY; | |
2242 | key.offset = (u64)-1; | |
2243 | root = btrfs_read_fs_root_no_name(info, &key); | |
2244 | if (IS_ERR(root)) { | |
f14d104d | 2245 | btrfs_err(info, "could not find root %llu", tree_id); |
8ad6fcab CM |
2246 | ret = -ENOENT; |
2247 | goto out; | |
98d377a0 TH |
2248 | } |
2249 | ||
2250 | key.objectid = dirid; | |
2251 | key.type = BTRFS_INODE_REF_KEY; | |
8ad6fcab | 2252 | key.offset = (u64)-1; |
98d377a0 | 2253 | |
67871254 | 2254 | while (1) { |
98d377a0 TH |
2255 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2256 | if (ret < 0) | |
2257 | goto out; | |
18674c6c FDBM |
2258 | else if (ret > 0) { |
2259 | ret = btrfs_previous_item(root, path, dirid, | |
2260 | BTRFS_INODE_REF_KEY); | |
2261 | if (ret < 0) | |
2262 | goto out; | |
2263 | else if (ret > 0) { | |
2264 | ret = -ENOENT; | |
2265 | goto out; | |
2266 | } | |
2267 | } | |
98d377a0 TH |
2268 | |
2269 | l = path->nodes[0]; | |
2270 | slot = path->slots[0]; | |
2271 | btrfs_item_key_to_cpu(l, &key, slot); | |
2272 | ||
98d377a0 TH |
2273 | iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); |
2274 | len = btrfs_inode_ref_name_len(l, iref); | |
2275 | ptr -= len + 1; | |
2276 | total_len += len + 1; | |
a696cf35 FDBM |
2277 | if (ptr < name) { |
2278 | ret = -ENAMETOOLONG; | |
98d377a0 | 2279 | goto out; |
a696cf35 | 2280 | } |
98d377a0 TH |
2281 | |
2282 | *(ptr + len) = '/'; | |
67871254 | 2283 | read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); |
98d377a0 TH |
2284 | |
2285 | if (key.offset == BTRFS_FIRST_FREE_OBJECTID) | |
2286 | break; | |
2287 | ||
b3b4aa74 | 2288 | btrfs_release_path(path); |
98d377a0 | 2289 | key.objectid = key.offset; |
8ad6fcab | 2290 | key.offset = (u64)-1; |
98d377a0 | 2291 | dirid = key.objectid; |
98d377a0 | 2292 | } |
77906a50 | 2293 | memmove(name, ptr, total_len); |
67871254 | 2294 | name[total_len] = '\0'; |
98d377a0 TH |
2295 | ret = 0; |
2296 | out: | |
2297 | btrfs_free_path(path); | |
ac8e9819 CM |
2298 | return ret; |
2299 | } | |
2300 | ||
2301 | static noinline int btrfs_ioctl_ino_lookup(struct file *file, | |
2302 | void __user *argp) | |
2303 | { | |
2304 | struct btrfs_ioctl_ino_lookup_args *args; | |
2305 | struct inode *inode; | |
01b810b8 | 2306 | int ret = 0; |
ac8e9819 | 2307 | |
2354d08f JL |
2308 | args = memdup_user(argp, sizeof(*args)); |
2309 | if (IS_ERR(args)) | |
2310 | return PTR_ERR(args); | |
c2b96929 | 2311 | |
496ad9aa | 2312 | inode = file_inode(file); |
ac8e9819 | 2313 | |
01b810b8 DS |
2314 | /* |
2315 | * Unprivileged query to obtain the containing subvolume root id. The | |
2316 | * path is reset so it's consistent with btrfs_search_path_in_tree. | |
2317 | */ | |
1b53ac4d CM |
2318 | if (args->treeid == 0) |
2319 | args->treeid = BTRFS_I(inode)->root->root_key.objectid; | |
2320 | ||
01b810b8 DS |
2321 | if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { |
2322 | args->name[0] = 0; | |
2323 | goto out; | |
2324 | } | |
2325 | ||
2326 | if (!capable(CAP_SYS_ADMIN)) { | |
2327 | ret = -EPERM; | |
2328 | goto out; | |
2329 | } | |
2330 | ||
ac8e9819 CM |
2331 | ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, |
2332 | args->treeid, args->objectid, | |
2333 | args->name); | |
2334 | ||
01b810b8 | 2335 | out: |
ac8e9819 CM |
2336 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) |
2337 | ret = -EFAULT; | |
2338 | ||
2339 | kfree(args); | |
98d377a0 TH |
2340 | return ret; |
2341 | } | |
2342 | ||
76dda93c YZ |
2343 | static noinline int btrfs_ioctl_snap_destroy(struct file *file, |
2344 | void __user *arg) | |
2345 | { | |
54563d41 | 2346 | struct dentry *parent = file->f_path.dentry; |
76dda93c | 2347 | struct dentry *dentry; |
2b0143b5 | 2348 | struct inode *dir = d_inode(parent); |
76dda93c YZ |
2349 | struct inode *inode; |
2350 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
2351 | struct btrfs_root *dest = NULL; | |
2352 | struct btrfs_ioctl_vol_args *vol_args; | |
2353 | struct btrfs_trans_handle *trans; | |
c58aaad2 | 2354 | struct btrfs_block_rsv block_rsv; |
521e0546 | 2355 | u64 root_flags; |
c58aaad2 | 2356 | u64 qgroup_reserved; |
76dda93c YZ |
2357 | int namelen; |
2358 | int ret; | |
2359 | int err = 0; | |
2360 | ||
76dda93c YZ |
2361 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
2362 | if (IS_ERR(vol_args)) | |
2363 | return PTR_ERR(vol_args); | |
2364 | ||
2365 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
2366 | namelen = strlen(vol_args->name); | |
2367 | if (strchr(vol_args->name, '/') || | |
2368 | strncmp(vol_args->name, "..", namelen) == 0) { | |
2369 | err = -EINVAL; | |
2370 | goto out; | |
2371 | } | |
2372 | ||
a561be71 | 2373 | err = mnt_want_write_file(file); |
76dda93c YZ |
2374 | if (err) |
2375 | goto out; | |
2376 | ||
521e0546 | 2377 | |
5c50c9b8 DS |
2378 | err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); |
2379 | if (err == -EINTR) | |
e43f998e | 2380 | goto out_drop_write; |
76dda93c YZ |
2381 | dentry = lookup_one_len(vol_args->name, parent, namelen); |
2382 | if (IS_ERR(dentry)) { | |
2383 | err = PTR_ERR(dentry); | |
2384 | goto out_unlock_dir; | |
2385 | } | |
2386 | ||
2b0143b5 | 2387 | if (d_really_is_negative(dentry)) { |
76dda93c YZ |
2388 | err = -ENOENT; |
2389 | goto out_dput; | |
2390 | } | |
2391 | ||
2b0143b5 | 2392 | inode = d_inode(dentry); |
4260f7c7 | 2393 | dest = BTRFS_I(inode)->root; |
67871254 | 2394 | if (!capable(CAP_SYS_ADMIN)) { |
4260f7c7 SW |
2395 | /* |
2396 | * Regular user. Only allow this with a special mount | |
2397 | * option, when the user has write+exec access to the | |
2398 | * subvol root, and when rmdir(2) would have been | |
2399 | * allowed. | |
2400 | * | |
2401 | * Note that this is _not_ check that the subvol is | |
2402 | * empty or doesn't contain data that we wouldn't | |
2403 | * otherwise be able to delete. | |
2404 | * | |
2405 | * Users who want to delete empty subvols should try | |
2406 | * rmdir(2). | |
2407 | */ | |
2408 | err = -EPERM; | |
2409 | if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) | |
2410 | goto out_dput; | |
2411 | ||
2412 | /* | |
2413 | * Do not allow deletion if the parent dir is the same | |
2414 | * as the dir to be deleted. That means the ioctl | |
2415 | * must be called on the dentry referencing the root | |
2416 | * of the subvol, not a random directory contained | |
2417 | * within it. | |
2418 | */ | |
2419 | err = -EINVAL; | |
2420 | if (root == dest) | |
2421 | goto out_dput; | |
2422 | ||
2423 | err = inode_permission(inode, MAY_WRITE | MAY_EXEC); | |
2424 | if (err) | |
2425 | goto out_dput; | |
4260f7c7 SW |
2426 | } |
2427 | ||
5c39da5b MX |
2428 | /* check if subvolume may be deleted by a user */ |
2429 | err = btrfs_may_delete(dir, dentry, 1); | |
2430 | if (err) | |
2431 | goto out_dput; | |
2432 | ||
33345d01 | 2433 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { |
76dda93c YZ |
2434 | err = -EINVAL; |
2435 | goto out_dput; | |
2436 | } | |
2437 | ||
5955102c | 2438 | inode_lock(inode); |
521e0546 DS |
2439 | |
2440 | /* | |
2441 | * Don't allow to delete a subvolume with send in progress. This is | |
2442 | * inside the i_mutex so the error handling that has to drop the bit | |
2443 | * again is not run concurrently. | |
2444 | */ | |
2445 | spin_lock(&dest->root_item_lock); | |
c55bfa67 FM |
2446 | root_flags = btrfs_root_flags(&dest->root_item); |
2447 | if (dest->send_in_progress == 0) { | |
2448 | btrfs_set_root_flags(&dest->root_item, | |
521e0546 DS |
2449 | root_flags | BTRFS_ROOT_SUBVOL_DEAD); |
2450 | spin_unlock(&dest->root_item_lock); | |
2451 | } else { | |
2452 | spin_unlock(&dest->root_item_lock); | |
2453 | btrfs_warn(root->fs_info, | |
2454 | "Attempt to delete subvolume %llu during send", | |
c55bfa67 | 2455 | dest->root_key.objectid); |
521e0546 | 2456 | err = -EPERM; |
909e26dc | 2457 | goto out_unlock_inode; |
521e0546 DS |
2458 | } |
2459 | ||
76dda93c YZ |
2460 | down_write(&root->fs_info->subvol_sem); |
2461 | ||
2462 | err = may_destroy_subvol(dest); | |
2463 | if (err) | |
2464 | goto out_up_write; | |
2465 | ||
c58aaad2 MX |
2466 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
2467 | /* | |
2468 | * One for dir inode, two for dir entries, two for root | |
2469 | * ref/backref. | |
2470 | */ | |
2471 | err = btrfs_subvolume_reserve_metadata(root, &block_rsv, | |
ee3441b4 | 2472 | 5, &qgroup_reserved, true); |
c58aaad2 MX |
2473 | if (err) |
2474 | goto out_up_write; | |
2475 | ||
a22285a6 YZ |
2476 | trans = btrfs_start_transaction(root, 0); |
2477 | if (IS_ERR(trans)) { | |
2478 | err = PTR_ERR(trans); | |
c58aaad2 | 2479 | goto out_release; |
a22285a6 | 2480 | } |
c58aaad2 MX |
2481 | trans->block_rsv = &block_rsv; |
2482 | trans->bytes_reserved = block_rsv.size; | |
a22285a6 | 2483 | |
2be63d5c FM |
2484 | btrfs_record_snapshot_destroy(trans, dir); |
2485 | ||
76dda93c YZ |
2486 | ret = btrfs_unlink_subvol(trans, root, dir, |
2487 | dest->root_key.objectid, | |
2488 | dentry->d_name.name, | |
2489 | dentry->d_name.len); | |
79787eaa JM |
2490 | if (ret) { |
2491 | err = ret; | |
2492 | btrfs_abort_transaction(trans, root, ret); | |
2493 | goto out_end_trans; | |
2494 | } | |
76dda93c YZ |
2495 | |
2496 | btrfs_record_root_in_trans(trans, dest); | |
2497 | ||
2498 | memset(&dest->root_item.drop_progress, 0, | |
2499 | sizeof(dest->root_item.drop_progress)); | |
2500 | dest->root_item.drop_level = 0; | |
2501 | btrfs_set_root_refs(&dest->root_item, 0); | |
2502 | ||
27cdeb70 | 2503 | if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { |
d68fc57b YZ |
2504 | ret = btrfs_insert_orphan_item(trans, |
2505 | root->fs_info->tree_root, | |
2506 | dest->root_key.objectid); | |
79787eaa JM |
2507 | if (ret) { |
2508 | btrfs_abort_transaction(trans, root, ret); | |
2509 | err = ret; | |
2510 | goto out_end_trans; | |
2511 | } | |
d68fc57b | 2512 | } |
dd5f9615 SB |
2513 | |
2514 | ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, | |
2515 | dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL, | |
2516 | dest->root_key.objectid); | |
2517 | if (ret && ret != -ENOENT) { | |
2518 | btrfs_abort_transaction(trans, root, ret); | |
2519 | err = ret; | |
2520 | goto out_end_trans; | |
2521 | } | |
2522 | if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { | |
2523 | ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, | |
2524 | dest->root_item.received_uuid, | |
2525 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
2526 | dest->root_key.objectid); | |
2527 | if (ret && ret != -ENOENT) { | |
2528 | btrfs_abort_transaction(trans, root, ret); | |
2529 | err = ret; | |
2530 | goto out_end_trans; | |
2531 | } | |
2532 | } | |
2533 | ||
79787eaa | 2534 | out_end_trans: |
c58aaad2 MX |
2535 | trans->block_rsv = NULL; |
2536 | trans->bytes_reserved = 0; | |
531cb13f | 2537 | ret = btrfs_end_transaction(trans, root); |
79787eaa JM |
2538 | if (ret && !err) |
2539 | err = ret; | |
76dda93c | 2540 | inode->i_flags |= S_DEAD; |
c58aaad2 MX |
2541 | out_release: |
2542 | btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); | |
76dda93c YZ |
2543 | out_up_write: |
2544 | up_write(&root->fs_info->subvol_sem); | |
521e0546 DS |
2545 | if (err) { |
2546 | spin_lock(&dest->root_item_lock); | |
c55bfa67 FM |
2547 | root_flags = btrfs_root_flags(&dest->root_item); |
2548 | btrfs_set_root_flags(&dest->root_item, | |
521e0546 DS |
2549 | root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); |
2550 | spin_unlock(&dest->root_item_lock); | |
2551 | } | |
909e26dc | 2552 | out_unlock_inode: |
5955102c | 2553 | inode_unlock(inode); |
76dda93c | 2554 | if (!err) { |
64ad6c48 | 2555 | d_invalidate(dentry); |
76dda93c YZ |
2556 | btrfs_invalidate_inodes(dest); |
2557 | d_delete(dentry); | |
61155aa0 | 2558 | ASSERT(dest->send_in_progress == 0); |
fa6ac876 LB |
2559 | |
2560 | /* the last ref */ | |
57cdc8db DS |
2561 | if (dest->ino_cache_inode) { |
2562 | iput(dest->ino_cache_inode); | |
2563 | dest->ino_cache_inode = NULL; | |
fa6ac876 | 2564 | } |
76dda93c YZ |
2565 | } |
2566 | out_dput: | |
2567 | dput(dentry); | |
2568 | out_unlock_dir: | |
5955102c | 2569 | inode_unlock(dir); |
e43f998e | 2570 | out_drop_write: |
2a79f17e | 2571 | mnt_drop_write_file(file); |
76dda93c YZ |
2572 | out: |
2573 | kfree(vol_args); | |
2574 | return err; | |
2575 | } | |
2576 | ||
1e701a32 | 2577 | static int btrfs_ioctl_defrag(struct file *file, void __user *argp) |
f46b5a66 | 2578 | { |
496ad9aa | 2579 | struct inode *inode = file_inode(file); |
f46b5a66 | 2580 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1e701a32 | 2581 | struct btrfs_ioctl_defrag_range_args *range; |
c146afad YZ |
2582 | int ret; |
2583 | ||
25122d15 ID |
2584 | ret = mnt_want_write_file(file); |
2585 | if (ret) | |
2586 | return ret; | |
b83cc969 | 2587 | |
25122d15 ID |
2588 | if (btrfs_root_readonly(root)) { |
2589 | ret = -EROFS; | |
2590 | goto out; | |
5ac00add | 2591 | } |
f46b5a66 CH |
2592 | |
2593 | switch (inode->i_mode & S_IFMT) { | |
2594 | case S_IFDIR: | |
e441d54d CM |
2595 | if (!capable(CAP_SYS_ADMIN)) { |
2596 | ret = -EPERM; | |
2597 | goto out; | |
2598 | } | |
de78b51a | 2599 | ret = btrfs_defrag_root(root); |
8929ecfa YZ |
2600 | if (ret) |
2601 | goto out; | |
de78b51a | 2602 | ret = btrfs_defrag_root(root->fs_info->extent_root); |
f46b5a66 CH |
2603 | break; |
2604 | case S_IFREG: | |
e441d54d CM |
2605 | if (!(file->f_mode & FMODE_WRITE)) { |
2606 | ret = -EINVAL; | |
2607 | goto out; | |
2608 | } | |
1e701a32 CM |
2609 | |
2610 | range = kzalloc(sizeof(*range), GFP_KERNEL); | |
2611 | if (!range) { | |
2612 | ret = -ENOMEM; | |
2613 | goto out; | |
2614 | } | |
2615 | ||
2616 | if (argp) { | |
2617 | if (copy_from_user(range, argp, | |
2618 | sizeof(*range))) { | |
2619 | ret = -EFAULT; | |
2620 | kfree(range); | |
683be16e | 2621 | goto out; |
1e701a32 CM |
2622 | } |
2623 | /* compression requires us to start the IO */ | |
2624 | if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { | |
2625 | range->flags |= BTRFS_DEFRAG_RANGE_START_IO; | |
2626 | range->extent_thresh = (u32)-1; | |
2627 | } | |
2628 | } else { | |
2629 | /* the rest are all set to zero by kzalloc */ | |
2630 | range->len = (u64)-1; | |
2631 | } | |
496ad9aa | 2632 | ret = btrfs_defrag_file(file_inode(file), file, |
4cb5300b CM |
2633 | range, 0, 0); |
2634 | if (ret > 0) | |
2635 | ret = 0; | |
1e701a32 | 2636 | kfree(range); |
f46b5a66 | 2637 | break; |
8929ecfa YZ |
2638 | default: |
2639 | ret = -EINVAL; | |
f46b5a66 | 2640 | } |
e441d54d | 2641 | out: |
25122d15 | 2642 | mnt_drop_write_file(file); |
e441d54d | 2643 | return ret; |
f46b5a66 CH |
2644 | } |
2645 | ||
b2950863 | 2646 | static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) |
f46b5a66 CH |
2647 | { |
2648 | struct btrfs_ioctl_vol_args *vol_args; | |
2649 | int ret; | |
2650 | ||
e441d54d CM |
2651 | if (!capable(CAP_SYS_ADMIN)) |
2652 | return -EPERM; | |
2653 | ||
5ac00add SB |
2654 | if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, |
2655 | 1)) { | |
e57138b3 | 2656 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
c9e9f97b ID |
2657 | } |
2658 | ||
5ac00add | 2659 | mutex_lock(&root->fs_info->volume_mutex); |
dae7b665 | 2660 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
2661 | if (IS_ERR(vol_args)) { |
2662 | ret = PTR_ERR(vol_args); | |
2663 | goto out; | |
2664 | } | |
f46b5a66 | 2665 | |
5516e595 | 2666 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 CH |
2667 | ret = btrfs_init_new_device(root, vol_args->name); |
2668 | ||
43d20761 AJ |
2669 | if (!ret) |
2670 | btrfs_info(root->fs_info, "disk added %s",vol_args->name); | |
2671 | ||
f46b5a66 | 2672 | kfree(vol_args); |
c9e9f97b ID |
2673 | out: |
2674 | mutex_unlock(&root->fs_info->volume_mutex); | |
5ac00add | 2675 | atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); |
f46b5a66 CH |
2676 | return ret; |
2677 | } | |
2678 | ||
da24927b | 2679 | static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) |
f46b5a66 | 2680 | { |
496ad9aa | 2681 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
f46b5a66 CH |
2682 | struct btrfs_ioctl_vol_args *vol_args; |
2683 | int ret; | |
2684 | ||
e441d54d CM |
2685 | if (!capable(CAP_SYS_ADMIN)) |
2686 | return -EPERM; | |
2687 | ||
da24927b MX |
2688 | ret = mnt_want_write_file(file); |
2689 | if (ret) | |
2690 | return ret; | |
c146afad | 2691 | |
dae7b665 | 2692 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
2693 | if (IS_ERR(vol_args)) { |
2694 | ret = PTR_ERR(vol_args); | |
c47ca32d | 2695 | goto err_drop; |
c9e9f97b | 2696 | } |
f46b5a66 | 2697 | |
5516e595 | 2698 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 | 2699 | |
183860f6 AJ |
2700 | if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, |
2701 | 1)) { | |
2702 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; | |
2703 | goto out; | |
2704 | } | |
2705 | ||
2706 | mutex_lock(&root->fs_info->volume_mutex); | |
2707 | ret = btrfs_rm_device(root, vol_args->name); | |
c9e9f97b | 2708 | mutex_unlock(&root->fs_info->volume_mutex); |
5ac00add | 2709 | atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); |
183860f6 | 2710 | |
ec95d491 AJ |
2711 | if (!ret) |
2712 | btrfs_info(root->fs_info, "disk deleted %s",vol_args->name); | |
2713 | ||
183860f6 AJ |
2714 | out: |
2715 | kfree(vol_args); | |
c47ca32d | 2716 | err_drop: |
4ac20c70 | 2717 | mnt_drop_write_file(file); |
f46b5a66 CH |
2718 | return ret; |
2719 | } | |
2720 | ||
475f6387 JS |
2721 | static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) |
2722 | { | |
027ed2f0 | 2723 | struct btrfs_ioctl_fs_info_args *fi_args; |
475f6387 | 2724 | struct btrfs_device *device; |
475f6387 | 2725 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; |
027ed2f0 | 2726 | int ret = 0; |
475f6387 | 2727 | |
027ed2f0 LZ |
2728 | fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); |
2729 | if (!fi_args) | |
2730 | return -ENOMEM; | |
2731 | ||
f7171750 | 2732 | mutex_lock(&fs_devices->device_list_mutex); |
027ed2f0 LZ |
2733 | fi_args->num_devices = fs_devices->num_devices; |
2734 | memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); | |
475f6387 | 2735 | |
d7641a49 | 2736 | list_for_each_entry(device, &fs_devices->devices, dev_list) { |
027ed2f0 LZ |
2737 | if (device->devid > fi_args->max_id) |
2738 | fi_args->max_id = device->devid; | |
475f6387 JS |
2739 | } |
2740 | mutex_unlock(&fs_devices->device_list_mutex); | |
2741 | ||
80a773fb DS |
2742 | fi_args->nodesize = root->fs_info->super_copy->nodesize; |
2743 | fi_args->sectorsize = root->fs_info->super_copy->sectorsize; | |
2744 | fi_args->clone_alignment = root->fs_info->super_copy->sectorsize; | |
2745 | ||
027ed2f0 LZ |
2746 | if (copy_to_user(arg, fi_args, sizeof(*fi_args))) |
2747 | ret = -EFAULT; | |
475f6387 | 2748 | |
027ed2f0 LZ |
2749 | kfree(fi_args); |
2750 | return ret; | |
475f6387 JS |
2751 | } |
2752 | ||
2753 | static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) | |
2754 | { | |
2755 | struct btrfs_ioctl_dev_info_args *di_args; | |
2756 | struct btrfs_device *dev; | |
2757 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
2758 | int ret = 0; | |
2759 | char *s_uuid = NULL; | |
475f6387 | 2760 | |
475f6387 JS |
2761 | di_args = memdup_user(arg, sizeof(*di_args)); |
2762 | if (IS_ERR(di_args)) | |
2763 | return PTR_ERR(di_args); | |
2764 | ||
dd5f9615 | 2765 | if (!btrfs_is_empty_uuid(di_args->uuid)) |
475f6387 JS |
2766 | s_uuid = di_args->uuid; |
2767 | ||
2768 | mutex_lock(&fs_devices->device_list_mutex); | |
aa1b8cd4 | 2769 | dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL); |
475f6387 JS |
2770 | |
2771 | if (!dev) { | |
2772 | ret = -ENODEV; | |
2773 | goto out; | |
2774 | } | |
2775 | ||
2776 | di_args->devid = dev->devid; | |
7cc8e58d MX |
2777 | di_args->bytes_used = btrfs_device_get_bytes_used(dev); |
2778 | di_args->total_bytes = btrfs_device_get_total_bytes(dev); | |
475f6387 | 2779 | memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); |
a27202fb | 2780 | if (dev->name) { |
606686ee JB |
2781 | struct rcu_string *name; |
2782 | ||
2783 | rcu_read_lock(); | |
2784 | name = rcu_dereference(dev->name); | |
2785 | strncpy(di_args->path, name->str, sizeof(di_args->path)); | |
2786 | rcu_read_unlock(); | |
a27202fb JM |
2787 | di_args->path[sizeof(di_args->path) - 1] = 0; |
2788 | } else { | |
99ba55ad | 2789 | di_args->path[0] = '\0'; |
a27202fb | 2790 | } |
475f6387 JS |
2791 | |
2792 | out: | |
55793c0d | 2793 | mutex_unlock(&fs_devices->device_list_mutex); |
475f6387 JS |
2794 | if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) |
2795 | ret = -EFAULT; | |
2796 | ||
2797 | kfree(di_args); | |
2798 | return ret; | |
2799 | } | |
2800 | ||
f4414602 | 2801 | static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) |
416161db MF |
2802 | { |
2803 | struct page *page; | |
416161db | 2804 | |
416161db MF |
2805 | page = grab_cache_page(inode->i_mapping, index); |
2806 | if (!page) | |
31314002 | 2807 | return ERR_PTR(-ENOMEM); |
416161db MF |
2808 | |
2809 | if (!PageUptodate(page)) { | |
31314002 FM |
2810 | int ret; |
2811 | ||
2812 | ret = btrfs_readpage(NULL, page); | |
2813 | if (ret) | |
2814 | return ERR_PTR(ret); | |
416161db MF |
2815 | lock_page(page); |
2816 | if (!PageUptodate(page)) { | |
2817 | unlock_page(page); | |
09cbfeaf | 2818 | put_page(page); |
31314002 FM |
2819 | return ERR_PTR(-EIO); |
2820 | } | |
2821 | if (page->mapping != inode->i_mapping) { | |
2822 | unlock_page(page); | |
09cbfeaf | 2823 | put_page(page); |
31314002 | 2824 | return ERR_PTR(-EAGAIN); |
416161db MF |
2825 | } |
2826 | } | |
416161db MF |
2827 | |
2828 | return page; | |
2829 | } | |
2830 | ||
f4414602 MF |
2831 | static int gather_extent_pages(struct inode *inode, struct page **pages, |
2832 | int num_pages, u64 off) | |
2833 | { | |
2834 | int i; | |
09cbfeaf | 2835 | pgoff_t index = off >> PAGE_SHIFT; |
f4414602 MF |
2836 | |
2837 | for (i = 0; i < num_pages; i++) { | |
31314002 | 2838 | again: |
f4414602 | 2839 | pages[i] = extent_same_get_page(inode, index + i); |
31314002 FM |
2840 | if (IS_ERR(pages[i])) { |
2841 | int err = PTR_ERR(pages[i]); | |
2842 | ||
2843 | if (err == -EAGAIN) | |
2844 | goto again; | |
2845 | pages[i] = NULL; | |
2846 | return err; | |
2847 | } | |
f4414602 MF |
2848 | } |
2849 | return 0; | |
2850 | } | |
2851 | ||
e0bd70c6 FM |
2852 | static int lock_extent_range(struct inode *inode, u64 off, u64 len, |
2853 | bool retry_range_locking) | |
77fe20dc | 2854 | { |
e0bd70c6 FM |
2855 | /* |
2856 | * Do any pending delalloc/csum calculations on inode, one way or | |
2857 | * another, and lock file content. | |
2858 | * The locking order is: | |
2859 | * | |
2860 | * 1) pages | |
2861 | * 2) range in the inode's io tree | |
2862 | */ | |
77fe20dc MF |
2863 | while (1) { |
2864 | struct btrfs_ordered_extent *ordered; | |
2865 | lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); | |
2866 | ordered = btrfs_lookup_first_ordered_extent(inode, | |
2867 | off + len - 1); | |
ff5df9b8 FM |
2868 | if ((!ordered || |
2869 | ordered->file_offset + ordered->len <= off || | |
2870 | ordered->file_offset >= off + len) && | |
77fe20dc | 2871 | !test_range_bit(&BTRFS_I(inode)->io_tree, off, |
ff5df9b8 FM |
2872 | off + len - 1, EXTENT_DELALLOC, 0, NULL)) { |
2873 | if (ordered) | |
2874 | btrfs_put_ordered_extent(ordered); | |
77fe20dc | 2875 | break; |
ff5df9b8 | 2876 | } |
77fe20dc MF |
2877 | unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); |
2878 | if (ordered) | |
2879 | btrfs_put_ordered_extent(ordered); | |
e0bd70c6 FM |
2880 | if (!retry_range_locking) |
2881 | return -EAGAIN; | |
77fe20dc MF |
2882 | btrfs_wait_ordered_range(inode, off, len); |
2883 | } | |
e0bd70c6 | 2884 | return 0; |
77fe20dc MF |
2885 | } |
2886 | ||
f4414602 | 2887 | static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) |
416161db | 2888 | { |
5955102c AV |
2889 | inode_unlock(inode1); |
2890 | inode_unlock(inode2); | |
416161db MF |
2891 | } |
2892 | ||
f4414602 MF |
2893 | static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) |
2894 | { | |
2895 | if (inode1 < inode2) | |
2896 | swap(inode1, inode2); | |
2897 | ||
5955102c AV |
2898 | inode_lock_nested(inode1, I_MUTEX_PARENT); |
2899 | inode_lock_nested(inode2, I_MUTEX_CHILD); | |
f4414602 MF |
2900 | } |
2901 | ||
2902 | static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, | |
2903 | struct inode *inode2, u64 loff2, u64 len) | |
2904 | { | |
2905 | unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); | |
2906 | unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); | |
2907 | } | |
2908 | ||
e0bd70c6 FM |
2909 | static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, |
2910 | struct inode *inode2, u64 loff2, u64 len, | |
2911 | bool retry_range_locking) | |
416161db | 2912 | { |
e0bd70c6 FM |
2913 | int ret; |
2914 | ||
416161db MF |
2915 | if (inode1 < inode2) { |
2916 | swap(inode1, inode2); | |
2917 | swap(loff1, loff2); | |
2918 | } | |
e0bd70c6 FM |
2919 | ret = lock_extent_range(inode1, loff1, len, retry_range_locking); |
2920 | if (ret) | |
2921 | return ret; | |
2922 | ret = lock_extent_range(inode2, loff2, len, retry_range_locking); | |
2923 | if (ret) | |
2924 | unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, | |
2925 | loff1 + len - 1); | |
2926 | return ret; | |
f4414602 MF |
2927 | } |
2928 | ||
2929 | struct cmp_pages { | |
2930 | int num_pages; | |
2931 | struct page **src_pages; | |
2932 | struct page **dst_pages; | |
2933 | }; | |
2934 | ||
2935 | static void btrfs_cmp_data_free(struct cmp_pages *cmp) | |
2936 | { | |
2937 | int i; | |
2938 | struct page *pg; | |
2939 | ||
2940 | for (i = 0; i < cmp->num_pages; i++) { | |
2941 | pg = cmp->src_pages[i]; | |
e0bd70c6 FM |
2942 | if (pg) { |
2943 | unlock_page(pg); | |
09cbfeaf | 2944 | put_page(pg); |
e0bd70c6 | 2945 | } |
f4414602 | 2946 | pg = cmp->dst_pages[i]; |
e0bd70c6 FM |
2947 | if (pg) { |
2948 | unlock_page(pg); | |
09cbfeaf | 2949 | put_page(pg); |
e0bd70c6 | 2950 | } |
f4414602 MF |
2951 | } |
2952 | kfree(cmp->src_pages); | |
2953 | kfree(cmp->dst_pages); | |
2954 | } | |
2955 | ||
2956 | static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, | |
2957 | struct inode *dst, u64 dst_loff, | |
2958 | u64 len, struct cmp_pages *cmp) | |
2959 | { | |
2960 | int ret; | |
09cbfeaf | 2961 | int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; |
f4414602 MF |
2962 | struct page **src_pgarr, **dst_pgarr; |
2963 | ||
2964 | /* | |
2965 | * We must gather up all the pages before we initiate our | |
2966 | * extent locking. We use an array for the page pointers. Size | |
2967 | * of the array is bounded by len, which is in turn bounded by | |
2968 | * BTRFS_MAX_DEDUPE_LEN. | |
2969 | */ | |
66722f7c DS |
2970 | src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); |
2971 | dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); | |
f4414602 MF |
2972 | if (!src_pgarr || !dst_pgarr) { |
2973 | kfree(src_pgarr); | |
2974 | kfree(dst_pgarr); | |
2975 | return -ENOMEM; | |
416161db | 2976 | } |
f4414602 MF |
2977 | cmp->num_pages = num_pages; |
2978 | cmp->src_pages = src_pgarr; | |
2979 | cmp->dst_pages = dst_pgarr; | |
2980 | ||
2981 | ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff); | |
2982 | if (ret) | |
2983 | goto out; | |
2984 | ||
2985 | ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff); | |
2986 | ||
2987 | out: | |
2988 | if (ret) | |
2989 | btrfs_cmp_data_free(cmp); | |
2990 | return 0; | |
416161db MF |
2991 | } |
2992 | ||
2993 | static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, | |
f4414602 | 2994 | u64 dst_loff, u64 len, struct cmp_pages *cmp) |
416161db MF |
2995 | { |
2996 | int ret = 0; | |
f4414602 | 2997 | int i; |
416161db | 2998 | struct page *src_page, *dst_page; |
09cbfeaf | 2999 | unsigned int cmp_len = PAGE_SIZE; |
416161db MF |
3000 | void *addr, *dst_addr; |
3001 | ||
f4414602 | 3002 | i = 0; |
416161db | 3003 | while (len) { |
09cbfeaf | 3004 | if (len < PAGE_SIZE) |
416161db MF |
3005 | cmp_len = len; |
3006 | ||
f4414602 MF |
3007 | BUG_ON(i >= cmp->num_pages); |
3008 | ||
3009 | src_page = cmp->src_pages[i]; | |
3010 | dst_page = cmp->dst_pages[i]; | |
e0bd70c6 FM |
3011 | ASSERT(PageLocked(src_page)); |
3012 | ASSERT(PageLocked(dst_page)); | |
f4414602 | 3013 | |
416161db MF |
3014 | addr = kmap_atomic(src_page); |
3015 | dst_addr = kmap_atomic(dst_page); | |
3016 | ||
3017 | flush_dcache_page(src_page); | |
3018 | flush_dcache_page(dst_page); | |
3019 | ||
3020 | if (memcmp(addr, dst_addr, cmp_len)) | |
2b3909f8 | 3021 | ret = -EBADE; |
416161db MF |
3022 | |
3023 | kunmap_atomic(addr); | |
3024 | kunmap_atomic(dst_addr); | |
416161db MF |
3025 | |
3026 | if (ret) | |
3027 | break; | |
3028 | ||
416161db | 3029 | len -= cmp_len; |
f4414602 | 3030 | i++; |
416161db MF |
3031 | } |
3032 | ||
3033 | return ret; | |
3034 | } | |
3035 | ||
e1d227a4 MF |
3036 | static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, |
3037 | u64 olen) | |
416161db | 3038 | { |
e1d227a4 | 3039 | u64 len = *plen; |
416161db MF |
3040 | u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; |
3041 | ||
e1d227a4 | 3042 | if (off + olen > inode->i_size || off + olen < off) |
416161db | 3043 | return -EINVAL; |
e1d227a4 MF |
3044 | |
3045 | /* if we extend to eof, continue to block boundary */ | |
3046 | if (off + len == inode->i_size) | |
3047 | *plen = len = ALIGN(inode->i_size, bs) - off; | |
3048 | ||
416161db MF |
3049 | /* Check that we are block aligned - btrfs_clone() requires this */ |
3050 | if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) | |
3051 | return -EINVAL; | |
3052 | ||
3053 | return 0; | |
3054 | } | |
3055 | ||
e1d227a4 | 3056 | static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, |
416161db MF |
3057 | struct inode *dst, u64 dst_loff) |
3058 | { | |
3059 | int ret; | |
e1d227a4 | 3060 | u64 len = olen; |
f4414602 | 3061 | struct cmp_pages cmp; |
0efa9f48 MF |
3062 | int same_inode = 0; |
3063 | u64 same_lock_start = 0; | |
3064 | u64 same_lock_len = 0; | |
416161db | 3065 | |
416161db | 3066 | if (src == dst) |
0efa9f48 | 3067 | same_inode = 1; |
416161db | 3068 | |
113e8283 FM |
3069 | if (len == 0) |
3070 | return 0; | |
3071 | ||
0efa9f48 | 3072 | if (same_inode) { |
5955102c | 3073 | inode_lock(src); |
416161db | 3074 | |
0efa9f48 | 3075 | ret = extent_same_check_offsets(src, loff, &len, olen); |
f4dfe687 FM |
3076 | if (ret) |
3077 | goto out_unlock; | |
3078 | ret = extent_same_check_offsets(src, dst_loff, &len, olen); | |
0efa9f48 MF |
3079 | if (ret) |
3080 | goto out_unlock; | |
416161db | 3081 | |
0efa9f48 MF |
3082 | /* |
3083 | * Single inode case wants the same checks, except we | |
3084 | * don't want our length pushed out past i_size as | |
3085 | * comparing that data range makes no sense. | |
3086 | * | |
3087 | * extent_same_check_offsets() will do this for an | |
3088 | * unaligned length at i_size, so catch it here and | |
3089 | * reject the request. | |
3090 | * | |
3091 | * This effectively means we require aligned extents | |
3092 | * for the single-inode case, whereas the other cases | |
3093 | * allow an unaligned length so long as it ends at | |
3094 | * i_size. | |
3095 | */ | |
3096 | if (len != olen) { | |
3097 | ret = -EINVAL; | |
3098 | goto out_unlock; | |
3099 | } | |
3100 | ||
3101 | /* Check for overlapping ranges */ | |
3102 | if (dst_loff + len > loff && dst_loff < loff + len) { | |
3103 | ret = -EINVAL; | |
3104 | goto out_unlock; | |
3105 | } | |
3106 | ||
3107 | same_lock_start = min_t(u64, loff, dst_loff); | |
3108 | same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; | |
3109 | } else { | |
3110 | btrfs_double_inode_lock(src, dst); | |
3111 | ||
3112 | ret = extent_same_check_offsets(src, loff, &len, olen); | |
3113 | if (ret) | |
3114 | goto out_unlock; | |
3115 | ||
3116 | ret = extent_same_check_offsets(dst, dst_loff, &len, olen); | |
3117 | if (ret) | |
3118 | goto out_unlock; | |
3119 | } | |
416161db MF |
3120 | |
3121 | /* don't make the dst file partly checksummed */ | |
3122 | if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != | |
3123 | (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { | |
3124 | ret = -EINVAL; | |
3125 | goto out_unlock; | |
3126 | } | |
3127 | ||
e0bd70c6 | 3128 | again: |
f4414602 MF |
3129 | ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); |
3130 | if (ret) | |
3131 | goto out_unlock; | |
3132 | ||
0efa9f48 | 3133 | if (same_inode) |
e0bd70c6 FM |
3134 | ret = lock_extent_range(src, same_lock_start, same_lock_len, |
3135 | false); | |
0efa9f48 | 3136 | else |
e0bd70c6 FM |
3137 | ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, |
3138 | false); | |
3139 | /* | |
3140 | * If one of the inodes has dirty pages in the respective range or | |
3141 | * ordered extents, we need to flush dellaloc and wait for all ordered | |
3142 | * extents in the range. We must unlock the pages and the ranges in the | |
3143 | * io trees to avoid deadlocks when flushing delalloc (requires locking | |
3144 | * pages) and when waiting for ordered extents to complete (they require | |
3145 | * range locking). | |
3146 | */ | |
3147 | if (ret == -EAGAIN) { | |
3148 | /* | |
3149 | * Ranges in the io trees already unlocked. Now unlock all | |
3150 | * pages before waiting for all IO to complete. | |
3151 | */ | |
3152 | btrfs_cmp_data_free(&cmp); | |
3153 | if (same_inode) { | |
3154 | btrfs_wait_ordered_range(src, same_lock_start, | |
3155 | same_lock_len); | |
3156 | } else { | |
3157 | btrfs_wait_ordered_range(src, loff, len); | |
3158 | btrfs_wait_ordered_range(dst, dst_loff, len); | |
3159 | } | |
3160 | goto again; | |
3161 | } | |
3162 | ASSERT(ret == 0); | |
3163 | if (WARN_ON(ret)) { | |
3164 | /* ranges in the io trees already unlocked */ | |
3165 | btrfs_cmp_data_free(&cmp); | |
3166 | return ret; | |
3167 | } | |
f4414602 | 3168 | |
207910dd | 3169 | /* pass original length for comparison so we stay within i_size */ |
f4414602 | 3170 | ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); |
416161db | 3171 | if (ret == 0) |
1c919a5e | 3172 | ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); |
416161db | 3173 | |
0efa9f48 MF |
3174 | if (same_inode) |
3175 | unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, | |
3176 | same_lock_start + same_lock_len - 1); | |
3177 | else | |
3178 | btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); | |
f4414602 MF |
3179 | |
3180 | btrfs_cmp_data_free(&cmp); | |
416161db | 3181 | out_unlock: |
0efa9f48 | 3182 | if (same_inode) |
5955102c | 3183 | inode_unlock(src); |
0efa9f48 MF |
3184 | else |
3185 | btrfs_double_inode_unlock(src, dst); | |
416161db MF |
3186 | |
3187 | return ret; | |
3188 | } | |
3189 | ||
ee22184b | 3190 | #define BTRFS_MAX_DEDUPE_LEN SZ_16M |
416161db | 3191 | |
2b3909f8 DW |
3192 | ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, |
3193 | struct file *dst_file, u64 dst_loff) | |
416161db | 3194 | { |
2b3909f8 DW |
3195 | struct inode *src = file_inode(src_file); |
3196 | struct inode *dst = file_inode(dst_file); | |
416161db | 3197 | u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; |
2b3909f8 | 3198 | ssize_t res; |
416161db | 3199 | |
2b3909f8 DW |
3200 | if (olen > BTRFS_MAX_DEDUPE_LEN) |
3201 | olen = BTRFS_MAX_DEDUPE_LEN; | |
416161db | 3202 | |
09cbfeaf | 3203 | if (WARN_ON_ONCE(bs < PAGE_SIZE)) { |
416161db MF |
3204 | /* |
3205 | * Btrfs does not support blocksize < page_size. As a | |
3206 | * result, btrfs_cmp_data() won't correctly handle | |
3207 | * this situation without an update. | |
3208 | */ | |
2b3909f8 | 3209 | return -EINVAL; |
416161db MF |
3210 | } |
3211 | ||
2b3909f8 DW |
3212 | res = btrfs_extent_same(src, loff, olen, dst, dst_loff); |
3213 | if (res) | |
3214 | return res; | |
3215 | return olen; | |
416161db MF |
3216 | } |
3217 | ||
f82a9901 FM |
3218 | static int clone_finish_inode_update(struct btrfs_trans_handle *trans, |
3219 | struct inode *inode, | |
3220 | u64 endoff, | |
3221 | const u64 destoff, | |
1c919a5e MF |
3222 | const u64 olen, |
3223 | int no_time_update) | |
f82a9901 FM |
3224 | { |
3225 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
3226 | int ret; | |
3227 | ||
3228 | inode_inc_iversion(inode); | |
1c919a5e | 3229 | if (!no_time_update) |
04b285f3 | 3230 | inode->i_mtime = inode->i_ctime = current_fs_time(inode->i_sb); |
f82a9901 FM |
3231 | /* |
3232 | * We round up to the block size at eof when determining which | |
3233 | * extents to clone above, but shouldn't round up the file size. | |
3234 | */ | |
3235 | if (endoff > destoff + olen) | |
3236 | endoff = destoff + olen; | |
3237 | if (endoff > inode->i_size) | |
3238 | btrfs_i_size_write(inode, endoff); | |
3239 | ||
3240 | ret = btrfs_update_inode(trans, root, inode); | |
3241 | if (ret) { | |
3242 | btrfs_abort_transaction(trans, root, ret); | |
3243 | btrfs_end_transaction(trans, root); | |
3244 | goto out; | |
3245 | } | |
3246 | ret = btrfs_end_transaction(trans, root); | |
3247 | out: | |
3248 | return ret; | |
3249 | } | |
3250 | ||
7ffbb598 FM |
3251 | static void clone_update_extent_map(struct inode *inode, |
3252 | const struct btrfs_trans_handle *trans, | |
3253 | const struct btrfs_path *path, | |
7ffbb598 FM |
3254 | const u64 hole_offset, |
3255 | const u64 hole_len) | |
3256 | { | |
3257 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
3258 | struct extent_map *em; | |
3259 | int ret; | |
3260 | ||
3261 | em = alloc_extent_map(); | |
3262 | if (!em) { | |
3263 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, | |
3264 | &BTRFS_I(inode)->runtime_flags); | |
3265 | return; | |
3266 | } | |
3267 | ||
14f59796 FM |
3268 | if (path) { |
3269 | struct btrfs_file_extent_item *fi; | |
3270 | ||
3271 | fi = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
3272 | struct btrfs_file_extent_item); | |
7ffbb598 FM |
3273 | btrfs_extent_item_to_extent_map(inode, path, fi, false, em); |
3274 | em->generation = -1; | |
3275 | if (btrfs_file_extent_type(path->nodes[0], fi) == | |
3276 | BTRFS_FILE_EXTENT_INLINE) | |
3277 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, | |
3278 | &BTRFS_I(inode)->runtime_flags); | |
3279 | } else { | |
3280 | em->start = hole_offset; | |
3281 | em->len = hole_len; | |
3282 | em->ram_bytes = em->len; | |
3283 | em->orig_start = hole_offset; | |
3284 | em->block_start = EXTENT_MAP_HOLE; | |
3285 | em->block_len = 0; | |
3286 | em->orig_block_len = 0; | |
3287 | em->compress_type = BTRFS_COMPRESS_NONE; | |
3288 | em->generation = trans->transid; | |
3289 | } | |
3290 | ||
3291 | while (1) { | |
3292 | write_lock(&em_tree->lock); | |
3293 | ret = add_extent_mapping(em_tree, em, 1); | |
3294 | write_unlock(&em_tree->lock); | |
3295 | if (ret != -EEXIST) { | |
3296 | free_extent_map(em); | |
3297 | break; | |
3298 | } | |
3299 | btrfs_drop_extent_cache(inode, em->start, | |
3300 | em->start + em->len - 1, 0); | |
3301 | } | |
3302 | ||
ee39b432 | 3303 | if (ret) |
7ffbb598 FM |
3304 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, |
3305 | &BTRFS_I(inode)->runtime_flags); | |
3306 | } | |
3307 | ||
8039d87d FM |
3308 | /* |
3309 | * Make sure we do not end up inserting an inline extent into a file that has | |
3310 | * already other (non-inline) extents. If a file has an inline extent it can | |
3311 | * not have any other extents and the (single) inline extent must start at the | |
3312 | * file offset 0. Failing to respect these rules will lead to file corruption, | |
3313 | * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc | |
3314 | * | |
3315 | * We can have extents that have been already written to disk or we can have | |
3316 | * dirty ranges still in delalloc, in which case the extent maps and items are | |
3317 | * created only when we run delalloc, and the delalloc ranges might fall outside | |
3318 | * the range we are currently locking in the inode's io tree. So we check the | |
3319 | * inode's i_size because of that (i_size updates are done while holding the | |
3320 | * i_mutex, which we are holding here). | |
3321 | * We also check to see if the inode has a size not greater than "datal" but has | |
3322 | * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are | |
3323 | * protected against such concurrent fallocate calls by the i_mutex). | |
3324 | * | |
3325 | * If the file has no extents but a size greater than datal, do not allow the | |
3326 | * copy because we would need turn the inline extent into a non-inline one (even | |
3327 | * with NO_HOLES enabled). If we find our destination inode only has one inline | |
3328 | * extent, just overwrite it with the source inline extent if its size is less | |
3329 | * than the source extent's size, or we could copy the source inline extent's | |
3330 | * data into the destination inode's inline extent if the later is greater then | |
3331 | * the former. | |
3332 | */ | |
3333 | static int clone_copy_inline_extent(struct inode *src, | |
3334 | struct inode *dst, | |
3335 | struct btrfs_trans_handle *trans, | |
3336 | struct btrfs_path *path, | |
3337 | struct btrfs_key *new_key, | |
3338 | const u64 drop_start, | |
3339 | const u64 datal, | |
3340 | const u64 skip, | |
3341 | const u64 size, | |
3342 | char *inline_data) | |
3343 | { | |
3344 | struct btrfs_root *root = BTRFS_I(dst)->root; | |
3345 | const u64 aligned_end = ALIGN(new_key->offset + datal, | |
3346 | root->sectorsize); | |
3347 | int ret; | |
3348 | struct btrfs_key key; | |
3349 | ||
3350 | if (new_key->offset > 0) | |
3351 | return -EOPNOTSUPP; | |
3352 | ||
3353 | key.objectid = btrfs_ino(dst); | |
3354 | key.type = BTRFS_EXTENT_DATA_KEY; | |
3355 | key.offset = 0; | |
3356 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
3357 | if (ret < 0) { | |
3358 | return ret; | |
3359 | } else if (ret > 0) { | |
3360 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
3361 | ret = btrfs_next_leaf(root, path); | |
3362 | if (ret < 0) | |
3363 | return ret; | |
3364 | else if (ret > 0) | |
3365 | goto copy_inline_extent; | |
3366 | } | |
3367 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
3368 | if (key.objectid == btrfs_ino(dst) && | |
3369 | key.type == BTRFS_EXTENT_DATA_KEY) { | |
3370 | ASSERT(key.offset > 0); | |
3371 | return -EOPNOTSUPP; | |
3372 | } | |
3373 | } else if (i_size_read(dst) <= datal) { | |
3374 | struct btrfs_file_extent_item *ei; | |
3375 | u64 ext_len; | |
3376 | ||
3377 | /* | |
3378 | * If the file size is <= datal, make sure there are no other | |
3379 | * extents following (can happen do to an fallocate call with | |
3380 | * the flag FALLOC_FL_KEEP_SIZE). | |
3381 | */ | |
3382 | ei = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
3383 | struct btrfs_file_extent_item); | |
3384 | /* | |
3385 | * If it's an inline extent, it can not have other extents | |
3386 | * following it. | |
3387 | */ | |
3388 | if (btrfs_file_extent_type(path->nodes[0], ei) == | |
3389 | BTRFS_FILE_EXTENT_INLINE) | |
3390 | goto copy_inline_extent; | |
3391 | ||
3392 | ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); | |
3393 | if (ext_len > aligned_end) | |
3394 | return -EOPNOTSUPP; | |
3395 | ||
3396 | ret = btrfs_next_item(root, path); | |
3397 | if (ret < 0) { | |
3398 | return ret; | |
3399 | } else if (ret == 0) { | |
3400 | btrfs_item_key_to_cpu(path->nodes[0], &key, | |
3401 | path->slots[0]); | |
3402 | if (key.objectid == btrfs_ino(dst) && | |
3403 | key.type == BTRFS_EXTENT_DATA_KEY) | |
3404 | return -EOPNOTSUPP; | |
3405 | } | |
3406 | } | |
3407 | ||
3408 | copy_inline_extent: | |
3409 | /* | |
3410 | * We have no extent items, or we have an extent at offset 0 which may | |
3411 | * or may not be inlined. All these cases are dealt the same way. | |
3412 | */ | |
3413 | if (i_size_read(dst) > datal) { | |
3414 | /* | |
3415 | * If the destination inode has an inline extent... | |
3416 | * This would require copying the data from the source inline | |
3417 | * extent into the beginning of the destination's inline extent. | |
3418 | * But this is really complex, both extents can be compressed | |
3419 | * or just one of them, which would require decompressing and | |
3420 | * re-compressing data (which could increase the new compressed | |
3421 | * size, not allowing the compressed data to fit anymore in an | |
3422 | * inline extent). | |
3423 | * So just don't support this case for now (it should be rare, | |
3424 | * we are not really saving space when cloning inline extents). | |
3425 | */ | |
3426 | return -EOPNOTSUPP; | |
3427 | } | |
3428 | ||
3429 | btrfs_release_path(path); | |
3430 | ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); | |
3431 | if (ret) | |
3432 | return ret; | |
3433 | ret = btrfs_insert_empty_item(trans, root, path, new_key, size); | |
3434 | if (ret) | |
3435 | return ret; | |
3436 | ||
3437 | if (skip) { | |
3438 | const u32 start = btrfs_file_extent_calc_inline_size(0); | |
3439 | ||
3440 | memmove(inline_data + start, inline_data + start + skip, datal); | |
3441 | } | |
3442 | ||
3443 | write_extent_buffer(path->nodes[0], inline_data, | |
3444 | btrfs_item_ptr_offset(path->nodes[0], | |
3445 | path->slots[0]), | |
3446 | size); | |
3447 | inode_add_bytes(dst, datal); | |
3448 | ||
3449 | return 0; | |
3450 | } | |
3451 | ||
32b7c687 MF |
3452 | /** |
3453 | * btrfs_clone() - clone a range from inode file to another | |
3454 | * | |
3455 | * @src: Inode to clone from | |
3456 | * @inode: Inode to clone to | |
3457 | * @off: Offset within source to start clone from | |
3458 | * @olen: Original length, passed by user, of range to clone | |
1c919a5e | 3459 | * @olen_aligned: Block-aligned value of olen |
32b7c687 | 3460 | * @destoff: Offset within @inode to start clone |
1c919a5e | 3461 | * @no_time_update: Whether to update mtime/ctime on the target inode |
32b7c687 MF |
3462 | */ |
3463 | static int btrfs_clone(struct inode *src, struct inode *inode, | |
f82a9901 | 3464 | const u64 off, const u64 olen, const u64 olen_aligned, |
1c919a5e | 3465 | const u64 destoff, int no_time_update) |
f46b5a66 | 3466 | { |
f46b5a66 | 3467 | struct btrfs_root *root = BTRFS_I(inode)->root; |
32b7c687 | 3468 | struct btrfs_path *path = NULL; |
f46b5a66 | 3469 | struct extent_buffer *leaf; |
32b7c687 MF |
3470 | struct btrfs_trans_handle *trans; |
3471 | char *buf = NULL; | |
ae01a0ab | 3472 | struct btrfs_key key; |
f46b5a66 CH |
3473 | u32 nritems; |
3474 | int slot; | |
ae01a0ab | 3475 | int ret; |
f82a9901 | 3476 | const u64 len = olen_aligned; |
f82a9901 | 3477 | u64 last_dest_end = destoff; |
ae01a0ab YZ |
3478 | |
3479 | ret = -ENOMEM; | |
15351955 DS |
3480 | buf = kmalloc(root->nodesize, GFP_KERNEL | __GFP_NOWARN); |
3481 | if (!buf) { | |
3482 | buf = vmalloc(root->nodesize); | |
3483 | if (!buf) | |
3484 | return ret; | |
3485 | } | |
ae01a0ab YZ |
3486 | |
3487 | path = btrfs_alloc_path(); | |
3488 | if (!path) { | |
15351955 | 3489 | kvfree(buf); |
32b7c687 | 3490 | return ret; |
d525e8ab LZ |
3491 | } |
3492 | ||
e4058b54 | 3493 | path->reada = READA_FORWARD; |
c5c9cd4d | 3494 | /* clone data */ |
33345d01 | 3495 | key.objectid = btrfs_ino(src); |
ae01a0ab | 3496 | key.type = BTRFS_EXTENT_DATA_KEY; |
2c463823 | 3497 | key.offset = off; |
f46b5a66 CH |
3498 | |
3499 | while (1) { | |
de249e66 | 3500 | u64 next_key_min_offset = key.offset + 1; |
df858e76 | 3501 | |
f46b5a66 CH |
3502 | /* |
3503 | * note the key will change type as we walk through the | |
3504 | * tree. | |
3505 | */ | |
e4355f34 | 3506 | path->leave_spinning = 1; |
362a20c5 DS |
3507 | ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, |
3508 | 0, 0); | |
f46b5a66 CH |
3509 | if (ret < 0) |
3510 | goto out; | |
2c463823 FM |
3511 | /* |
3512 | * First search, if no extent item that starts at offset off was | |
3513 | * found but the previous item is an extent item, it's possible | |
3514 | * it might overlap our target range, therefore process it. | |
3515 | */ | |
3516 | if (key.offset == off && ret > 0 && path->slots[0] > 0) { | |
3517 | btrfs_item_key_to_cpu(path->nodes[0], &key, | |
3518 | path->slots[0] - 1); | |
3519 | if (key.type == BTRFS_EXTENT_DATA_KEY) | |
3520 | path->slots[0]--; | |
3521 | } | |
f46b5a66 | 3522 | |
ae01a0ab | 3523 | nritems = btrfs_header_nritems(path->nodes[0]); |
e4355f34 | 3524 | process_slot: |
ae01a0ab | 3525 | if (path->slots[0] >= nritems) { |
362a20c5 | 3526 | ret = btrfs_next_leaf(BTRFS_I(src)->root, path); |
f46b5a66 CH |
3527 | if (ret < 0) |
3528 | goto out; | |
3529 | if (ret > 0) | |
3530 | break; | |
ae01a0ab | 3531 | nritems = btrfs_header_nritems(path->nodes[0]); |
f46b5a66 CH |
3532 | } |
3533 | leaf = path->nodes[0]; | |
3534 | slot = path->slots[0]; | |
f46b5a66 | 3535 | |
ae01a0ab | 3536 | btrfs_item_key_to_cpu(leaf, &key, slot); |
962a298f | 3537 | if (key.type > BTRFS_EXTENT_DATA_KEY || |
33345d01 | 3538 | key.objectid != btrfs_ino(src)) |
f46b5a66 CH |
3539 | break; |
3540 | ||
962a298f | 3541 | if (key.type == BTRFS_EXTENT_DATA_KEY) { |
c5c9cd4d SW |
3542 | struct btrfs_file_extent_item *extent; |
3543 | int type; | |
31840ae1 ZY |
3544 | u32 size; |
3545 | struct btrfs_key new_key; | |
c5c9cd4d SW |
3546 | u64 disko = 0, diskl = 0; |
3547 | u64 datao = 0, datal = 0; | |
3548 | u8 comp; | |
f82a9901 | 3549 | u64 drop_start; |
31840ae1 | 3550 | |
c5c9cd4d SW |
3551 | extent = btrfs_item_ptr(leaf, slot, |
3552 | struct btrfs_file_extent_item); | |
3553 | comp = btrfs_file_extent_compression(leaf, extent); | |
3554 | type = btrfs_file_extent_type(leaf, extent); | |
c8a894d7 CM |
3555 | if (type == BTRFS_FILE_EXTENT_REG || |
3556 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d397712b CM |
3557 | disko = btrfs_file_extent_disk_bytenr(leaf, |
3558 | extent); | |
3559 | diskl = btrfs_file_extent_disk_num_bytes(leaf, | |
3560 | extent); | |
c5c9cd4d | 3561 | datao = btrfs_file_extent_offset(leaf, extent); |
d397712b CM |
3562 | datal = btrfs_file_extent_num_bytes(leaf, |
3563 | extent); | |
c5c9cd4d SW |
3564 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
3565 | /* take upper bound, may be compressed */ | |
3566 | datal = btrfs_file_extent_ram_bytes(leaf, | |
3567 | extent); | |
3568 | } | |
31840ae1 | 3569 | |
2c463823 FM |
3570 | /* |
3571 | * The first search might have left us at an extent | |
3572 | * item that ends before our target range's start, can | |
3573 | * happen if we have holes and NO_HOLES feature enabled. | |
3574 | */ | |
3575 | if (key.offset + datal <= off) { | |
e4355f34 FDBM |
3576 | path->slots[0]++; |
3577 | goto process_slot; | |
2c463823 FM |
3578 | } else if (key.offset >= off + len) { |
3579 | break; | |
e4355f34 | 3580 | } |
df858e76 | 3581 | next_key_min_offset = key.offset + datal; |
e4355f34 FDBM |
3582 | size = btrfs_item_size_nr(leaf, slot); |
3583 | read_extent_buffer(leaf, buf, | |
3584 | btrfs_item_ptr_offset(leaf, slot), | |
3585 | size); | |
3586 | ||
3587 | btrfs_release_path(path); | |
3588 | path->leave_spinning = 0; | |
c5c9cd4d | 3589 | |
31840ae1 | 3590 | memcpy(&new_key, &key, sizeof(new_key)); |
33345d01 | 3591 | new_key.objectid = btrfs_ino(inode); |
4d728ec7 LZ |
3592 | if (off <= key.offset) |
3593 | new_key.offset = key.offset + destoff - off; | |
3594 | else | |
3595 | new_key.offset = destoff; | |
31840ae1 | 3596 | |
f82a9901 FM |
3597 | /* |
3598 | * Deal with a hole that doesn't have an extent item | |
3599 | * that represents it (NO_HOLES feature enabled). | |
3600 | * This hole is either in the middle of the cloning | |
3601 | * range or at the beginning (fully overlaps it or | |
3602 | * partially overlaps it). | |
3603 | */ | |
3604 | if (new_key.offset != last_dest_end) | |
3605 | drop_start = last_dest_end; | |
3606 | else | |
3607 | drop_start = new_key.offset; | |
3608 | ||
b6f3409b SW |
3609 | /* |
3610 | * 1 - adjusting old extent (we may have to split it) | |
3611 | * 1 - add new extent | |
3612 | * 1 - inode update | |
3613 | */ | |
3614 | trans = btrfs_start_transaction(root, 3); | |
a22285a6 YZ |
3615 | if (IS_ERR(trans)) { |
3616 | ret = PTR_ERR(trans); | |
3617 | goto out; | |
3618 | } | |
3619 | ||
c8a894d7 CM |
3620 | if (type == BTRFS_FILE_EXTENT_REG || |
3621 | type == BTRFS_FILE_EXTENT_PREALLOC) { | |
d72c0842 LZ |
3622 | /* |
3623 | * a | --- range to clone ---| b | |
3624 | * | ------------- extent ------------- | | |
3625 | */ | |
3626 | ||
93915584 | 3627 | /* subtract range b */ |
d72c0842 LZ |
3628 | if (key.offset + datal > off + len) |
3629 | datal = off + len - key.offset; | |
3630 | ||
93915584 | 3631 | /* subtract range a */ |
a22285a6 YZ |
3632 | if (off > key.offset) { |
3633 | datao += off - key.offset; | |
3634 | datal -= off - key.offset; | |
3635 | } | |
3636 | ||
5dc562c5 | 3637 | ret = btrfs_drop_extents(trans, root, inode, |
f82a9901 | 3638 | drop_start, |
a22285a6 | 3639 | new_key.offset + datal, |
2671485d | 3640 | 1); |
79787eaa | 3641 | if (ret) { |
3f9e3df8 | 3642 | if (ret != -EOPNOTSUPP) |
00fdf13a LB |
3643 | btrfs_abort_transaction(trans, |
3644 | root, ret); | |
79787eaa JM |
3645 | btrfs_end_transaction(trans, root); |
3646 | goto out; | |
3647 | } | |
a22285a6 | 3648 | |
c5c9cd4d SW |
3649 | ret = btrfs_insert_empty_item(trans, root, path, |
3650 | &new_key, size); | |
79787eaa JM |
3651 | if (ret) { |
3652 | btrfs_abort_transaction(trans, root, | |
3653 | ret); | |
3654 | btrfs_end_transaction(trans, root); | |
3655 | goto out; | |
3656 | } | |
c5c9cd4d SW |
3657 | |
3658 | leaf = path->nodes[0]; | |
3659 | slot = path->slots[0]; | |
3660 | write_extent_buffer(leaf, buf, | |
31840ae1 ZY |
3661 | btrfs_item_ptr_offset(leaf, slot), |
3662 | size); | |
ae01a0ab | 3663 | |
c5c9cd4d | 3664 | extent = btrfs_item_ptr(leaf, slot, |
f46b5a66 | 3665 | struct btrfs_file_extent_item); |
c5c9cd4d | 3666 | |
c5c9cd4d SW |
3667 | /* disko == 0 means it's a hole */ |
3668 | if (!disko) | |
3669 | datao = 0; | |
c5c9cd4d SW |
3670 | |
3671 | btrfs_set_file_extent_offset(leaf, extent, | |
3672 | datao); | |
3673 | btrfs_set_file_extent_num_bytes(leaf, extent, | |
3674 | datal); | |
fcebe456 | 3675 | |
c5c9cd4d SW |
3676 | if (disko) { |
3677 | inode_add_bytes(inode, datal); | |
ae01a0ab | 3678 | ret = btrfs_inc_extent_ref(trans, root, |
5d4f98a2 YZ |
3679 | disko, diskl, 0, |
3680 | root->root_key.objectid, | |
33345d01 | 3681 | btrfs_ino(inode), |
b06c4bf5 | 3682 | new_key.offset - datao); |
79787eaa JM |
3683 | if (ret) { |
3684 | btrfs_abort_transaction(trans, | |
3685 | root, | |
3686 | ret); | |
3687 | btrfs_end_transaction(trans, | |
3688 | root); | |
3689 | goto out; | |
3690 | ||
3691 | } | |
f46b5a66 | 3692 | } |
c5c9cd4d SW |
3693 | } else if (type == BTRFS_FILE_EXTENT_INLINE) { |
3694 | u64 skip = 0; | |
3695 | u64 trim = 0; | |
ed958762 | 3696 | |
c5c9cd4d SW |
3697 | if (off > key.offset) { |
3698 | skip = off - key.offset; | |
3699 | new_key.offset += skip; | |
3700 | } | |
d397712b | 3701 | |
aa42ffd9 LB |
3702 | if (key.offset + datal > off + len) |
3703 | trim = key.offset + datal - (off + len); | |
d397712b | 3704 | |
c5c9cd4d | 3705 | if (comp && (skip || trim)) { |
c5c9cd4d | 3706 | ret = -EINVAL; |
a22285a6 | 3707 | btrfs_end_transaction(trans, root); |
c5c9cd4d SW |
3708 | goto out; |
3709 | } | |
3710 | size -= skip + trim; | |
3711 | datal -= skip + trim; | |
a22285a6 | 3712 | |
8039d87d FM |
3713 | ret = clone_copy_inline_extent(src, inode, |
3714 | trans, path, | |
3715 | &new_key, | |
3716 | drop_start, | |
3717 | datal, | |
3718 | skip, size, buf); | |
79787eaa | 3719 | if (ret) { |
3f9e3df8 | 3720 | if (ret != -EOPNOTSUPP) |
3a29bc09 | 3721 | btrfs_abort_transaction(trans, |
8039d87d FM |
3722 | root, |
3723 | ret); | |
79787eaa JM |
3724 | btrfs_end_transaction(trans, root); |
3725 | goto out; | |
3726 | } | |
c5c9cd4d SW |
3727 | leaf = path->nodes[0]; |
3728 | slot = path->slots[0]; | |
f46b5a66 | 3729 | } |
c5c9cd4d | 3730 | |
7ffbb598 FM |
3731 | /* If we have an implicit hole (NO_HOLES feature). */ |
3732 | if (drop_start < new_key.offset) | |
3733 | clone_update_extent_map(inode, trans, | |
14f59796 | 3734 | NULL, drop_start, |
7ffbb598 FM |
3735 | new_key.offset - drop_start); |
3736 | ||
14f59796 | 3737 | clone_update_extent_map(inode, trans, path, 0, 0); |
7ffbb598 | 3738 | |
c5c9cd4d | 3739 | btrfs_mark_buffer_dirty(leaf); |
b3b4aa74 | 3740 | btrfs_release_path(path); |
c5c9cd4d | 3741 | |
62e2390e FM |
3742 | last_dest_end = ALIGN(new_key.offset + datal, |
3743 | root->sectorsize); | |
f82a9901 FM |
3744 | ret = clone_finish_inode_update(trans, inode, |
3745 | last_dest_end, | |
1c919a5e MF |
3746 | destoff, olen, |
3747 | no_time_update); | |
f82a9901 | 3748 | if (ret) |
79787eaa | 3749 | goto out; |
2c463823 FM |
3750 | if (new_key.offset + datal >= destoff + len) |
3751 | break; | |
a22285a6 | 3752 | } |
b3b4aa74 | 3753 | btrfs_release_path(path); |
df858e76 | 3754 | key.offset = next_key_min_offset; |
f46b5a66 | 3755 | } |
f46b5a66 | 3756 | ret = 0; |
32b7c687 | 3757 | |
f82a9901 FM |
3758 | if (last_dest_end < destoff + len) { |
3759 | /* | |
3760 | * We have an implicit hole (NO_HOLES feature is enabled) that | |
3761 | * fully or partially overlaps our cloning range at its end. | |
3762 | */ | |
3763 | btrfs_release_path(path); | |
3764 | ||
3765 | /* | |
3766 | * 1 - remove extent(s) | |
3767 | * 1 - inode update | |
3768 | */ | |
3769 | trans = btrfs_start_transaction(root, 2); | |
3770 | if (IS_ERR(trans)) { | |
3771 | ret = PTR_ERR(trans); | |
3772 | goto out; | |
3773 | } | |
3774 | ret = btrfs_drop_extents(trans, root, inode, | |
3775 | last_dest_end, destoff + len, 1); | |
3776 | if (ret) { | |
3777 | if (ret != -EOPNOTSUPP) | |
3778 | btrfs_abort_transaction(trans, root, ret); | |
3779 | btrfs_end_transaction(trans, root); | |
3780 | goto out; | |
3781 | } | |
14f59796 FM |
3782 | clone_update_extent_map(inode, trans, NULL, last_dest_end, |
3783 | destoff + len - last_dest_end); | |
f82a9901 | 3784 | ret = clone_finish_inode_update(trans, inode, destoff + len, |
1c919a5e | 3785 | destoff, olen, no_time_update); |
f82a9901 FM |
3786 | } |
3787 | ||
f46b5a66 | 3788 | out: |
32b7c687 | 3789 | btrfs_free_path(path); |
15351955 | 3790 | kvfree(buf); |
32b7c687 MF |
3791 | return ret; |
3792 | } | |
3793 | ||
3db11b2e ZB |
3794 | static noinline int btrfs_clone_files(struct file *file, struct file *file_src, |
3795 | u64 off, u64 olen, u64 destoff) | |
32b7c687 | 3796 | { |
54563d41 | 3797 | struct inode *inode = file_inode(file); |
3db11b2e | 3798 | struct inode *src = file_inode(file_src); |
32b7c687 | 3799 | struct btrfs_root *root = BTRFS_I(inode)->root; |
32b7c687 MF |
3800 | int ret; |
3801 | u64 len = olen; | |
3802 | u64 bs = root->fs_info->sb->s_blocksize; | |
3db11b2e | 3803 | int same_inode = src == inode; |
32b7c687 MF |
3804 | |
3805 | /* | |
3806 | * TODO: | |
3807 | * - split compressed inline extents. annoying: we need to | |
3808 | * decompress into destination's address_space (the file offset | |
3809 | * may change, so source mapping won't do), then recompress (or | |
3810 | * otherwise reinsert) a subrange. | |
00fdf13a LB |
3811 | * |
3812 | * - split destination inode's inline extents. The inline extents can | |
3813 | * be either compressed or non-compressed. | |
32b7c687 MF |
3814 | */ |
3815 | ||
32b7c687 MF |
3816 | if (btrfs_root_readonly(root)) |
3817 | return -EROFS; | |
3818 | ||
3db11b2e ZB |
3819 | if (file_src->f_path.mnt != file->f_path.mnt || |
3820 | src->i_sb != inode->i_sb) | |
3821 | return -EXDEV; | |
32b7c687 MF |
3822 | |
3823 | /* don't make the dst file partly checksummed */ | |
3824 | if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != | |
3825 | (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) | |
3db11b2e | 3826 | return -EINVAL; |
32b7c687 | 3827 | |
32b7c687 | 3828 | if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) |
3db11b2e | 3829 | return -EISDIR; |
32b7c687 MF |
3830 | |
3831 | if (!same_inode) { | |
293a8489 | 3832 | btrfs_double_inode_lock(src, inode); |
32b7c687 | 3833 | } else { |
5955102c | 3834 | inode_lock(src); |
32b7c687 MF |
3835 | } |
3836 | ||
3837 | /* determine range to clone */ | |
3838 | ret = -EINVAL; | |
3839 | if (off + len > src->i_size || off + len < off) | |
3840 | goto out_unlock; | |
3841 | if (len == 0) | |
3842 | olen = len = src->i_size - off; | |
3843 | /* if we extend to eof, continue to block boundary */ | |
3844 | if (off + len == src->i_size) | |
3845 | len = ALIGN(src->i_size, bs) - off; | |
3846 | ||
ccccf3d6 FM |
3847 | if (len == 0) { |
3848 | ret = 0; | |
3849 | goto out_unlock; | |
3850 | } | |
3851 | ||
32b7c687 MF |
3852 | /* verify the end result is block aligned */ |
3853 | if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || | |
3854 | !IS_ALIGNED(destoff, bs)) | |
3855 | goto out_unlock; | |
3856 | ||
3857 | /* verify if ranges are overlapped within the same file */ | |
3858 | if (same_inode) { | |
3859 | if (destoff + len > off && destoff < off + len) | |
3860 | goto out_unlock; | |
3861 | } | |
3862 | ||
3863 | if (destoff > inode->i_size) { | |
3864 | ret = btrfs_cont_expand(inode, inode->i_size, destoff); | |
3865 | if (ret) | |
3866 | goto out_unlock; | |
3867 | } | |
3868 | ||
c125b8bf FM |
3869 | /* |
3870 | * Lock the target range too. Right after we replace the file extent | |
3871 | * items in the fs tree (which now point to the cloned data), we might | |
3872 | * have a worker replace them with extent items relative to a write | |
3873 | * operation that was issued before this clone operation (i.e. confront | |
3874 | * with inode.c:btrfs_finish_ordered_io). | |
3875 | */ | |
3876 | if (same_inode) { | |
3877 | u64 lock_start = min_t(u64, off, destoff); | |
3878 | u64 lock_len = max_t(u64, off, destoff) + len - lock_start; | |
32b7c687 | 3879 | |
e0bd70c6 | 3880 | ret = lock_extent_range(src, lock_start, lock_len, true); |
c125b8bf | 3881 | } else { |
e0bd70c6 FM |
3882 | ret = btrfs_double_extent_lock(src, off, inode, destoff, len, |
3883 | true); | |
3884 | } | |
3885 | ASSERT(ret == 0); | |
3886 | if (WARN_ON(ret)) { | |
3887 | /* ranges in the io trees already unlocked */ | |
3888 | goto out_unlock; | |
c125b8bf | 3889 | } |
32b7c687 | 3890 | |
1c919a5e | 3891 | ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); |
32b7c687 | 3892 | |
c125b8bf FM |
3893 | if (same_inode) { |
3894 | u64 lock_start = min_t(u64, off, destoff); | |
3895 | u64 lock_end = max_t(u64, off, destoff) + len - 1; | |
3896 | ||
3897 | unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); | |
3898 | } else { | |
293a8489 | 3899 | btrfs_double_extent_unlock(src, off, inode, destoff, len); |
c125b8bf FM |
3900 | } |
3901 | /* | |
3902 | * Truncate page cache pages so that future reads will see the cloned | |
3903 | * data immediately and not the previous data. | |
3904 | */ | |
65bfa658 | 3905 | truncate_inode_pages_range(&inode->i_data, |
09cbfeaf KS |
3906 | round_down(destoff, PAGE_SIZE), |
3907 | round_up(destoff + len, PAGE_SIZE) - 1); | |
f46b5a66 | 3908 | out_unlock: |
293a8489 MF |
3909 | if (!same_inode) |
3910 | btrfs_double_inode_unlock(src, inode); | |
3911 | else | |
5955102c | 3912 | inode_unlock(src); |
3db11b2e ZB |
3913 | return ret; |
3914 | } | |
3915 | ||
3916 | ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in, | |
3917 | struct file *file_out, loff_t pos_out, | |
3918 | size_t len, unsigned int flags) | |
3919 | { | |
3920 | ssize_t ret; | |
3921 | ||
3922 | ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out); | |
3923 | if (ret == 0) | |
3924 | ret = len; | |
3925 | return ret; | |
3926 | } | |
3927 | ||
04b38d60 CH |
3928 | int btrfs_clone_file_range(struct file *src_file, loff_t off, |
3929 | struct file *dst_file, loff_t destoff, u64 len) | |
3db11b2e | 3930 | { |
04b38d60 | 3931 | return btrfs_clone_files(dst_file, src_file, off, len, destoff); |
c5c9cd4d SW |
3932 | } |
3933 | ||
f46b5a66 CH |
3934 | /* |
3935 | * there are many ways the trans_start and trans_end ioctls can lead | |
3936 | * to deadlocks. They should only be used by applications that | |
3937 | * basically own the machine, and have a very in depth understanding | |
3938 | * of all the possible deadlocks and enospc problems. | |
3939 | */ | |
b2950863 | 3940 | static long btrfs_ioctl_trans_start(struct file *file) |
f46b5a66 | 3941 | { |
496ad9aa | 3942 | struct inode *inode = file_inode(file); |
f46b5a66 CH |
3943 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3944 | struct btrfs_trans_handle *trans; | |
1ab86aed | 3945 | int ret; |
f46b5a66 | 3946 | |
1ab86aed | 3947 | ret = -EPERM; |
df5b5520 | 3948 | if (!capable(CAP_SYS_ADMIN)) |
1ab86aed | 3949 | goto out; |
df5b5520 | 3950 | |
1ab86aed SW |
3951 | ret = -EINPROGRESS; |
3952 | if (file->private_data) | |
f46b5a66 | 3953 | goto out; |
9ca9ee09 | 3954 | |
b83cc969 LZ |
3955 | ret = -EROFS; |
3956 | if (btrfs_root_readonly(root)) | |
3957 | goto out; | |
3958 | ||
a561be71 | 3959 | ret = mnt_want_write_file(file); |
c146afad YZ |
3960 | if (ret) |
3961 | goto out; | |
3962 | ||
a4abeea4 | 3963 | atomic_inc(&root->fs_info->open_ioctl_trans); |
9ca9ee09 | 3964 | |
1ab86aed | 3965 | ret = -ENOMEM; |
7a7eaa40 | 3966 | trans = btrfs_start_ioctl_transaction(root); |
abd30bb0 | 3967 | if (IS_ERR(trans)) |
1ab86aed SW |
3968 | goto out_drop; |
3969 | ||
3970 | file->private_data = trans; | |
3971 | return 0; | |
3972 | ||
3973 | out_drop: | |
a4abeea4 | 3974 | atomic_dec(&root->fs_info->open_ioctl_trans); |
2a79f17e | 3975 | mnt_drop_write_file(file); |
f46b5a66 | 3976 | out: |
f46b5a66 CH |
3977 | return ret; |
3978 | } | |
3979 | ||
6ef5ed0d JB |
3980 | static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) |
3981 | { | |
496ad9aa | 3982 | struct inode *inode = file_inode(file); |
6ef5ed0d JB |
3983 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3984 | struct btrfs_root *new_root; | |
3985 | struct btrfs_dir_item *di; | |
3986 | struct btrfs_trans_handle *trans; | |
3987 | struct btrfs_path *path; | |
3988 | struct btrfs_key location; | |
3989 | struct btrfs_disk_key disk_key; | |
6ef5ed0d JB |
3990 | u64 objectid = 0; |
3991 | u64 dir_id; | |
3c04ce01 | 3992 | int ret; |
6ef5ed0d JB |
3993 | |
3994 | if (!capable(CAP_SYS_ADMIN)) | |
3995 | return -EPERM; | |
3996 | ||
3c04ce01 MX |
3997 | ret = mnt_want_write_file(file); |
3998 | if (ret) | |
3999 | return ret; | |
4000 | ||
4001 | if (copy_from_user(&objectid, argp, sizeof(objectid))) { | |
4002 | ret = -EFAULT; | |
4003 | goto out; | |
4004 | } | |
6ef5ed0d JB |
4005 | |
4006 | if (!objectid) | |
1cecf579 | 4007 | objectid = BTRFS_FS_TREE_OBJECTID; |
6ef5ed0d JB |
4008 | |
4009 | location.objectid = objectid; | |
4010 | location.type = BTRFS_ROOT_ITEM_KEY; | |
4011 | location.offset = (u64)-1; | |
4012 | ||
4013 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); | |
3c04ce01 MX |
4014 | if (IS_ERR(new_root)) { |
4015 | ret = PTR_ERR(new_root); | |
4016 | goto out; | |
4017 | } | |
6ef5ed0d | 4018 | |
6ef5ed0d | 4019 | path = btrfs_alloc_path(); |
3c04ce01 MX |
4020 | if (!path) { |
4021 | ret = -ENOMEM; | |
4022 | goto out; | |
4023 | } | |
6ef5ed0d JB |
4024 | path->leave_spinning = 1; |
4025 | ||
4026 | trans = btrfs_start_transaction(root, 1); | |
98d5dc13 | 4027 | if (IS_ERR(trans)) { |
6ef5ed0d | 4028 | btrfs_free_path(path); |
3c04ce01 MX |
4029 | ret = PTR_ERR(trans); |
4030 | goto out; | |
6ef5ed0d JB |
4031 | } |
4032 | ||
6c41761f | 4033 | dir_id = btrfs_super_root_dir(root->fs_info->super_copy); |
6ef5ed0d JB |
4034 | di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, |
4035 | dir_id, "default", 7, 1); | |
cf1e99a4 | 4036 | if (IS_ERR_OR_NULL(di)) { |
6ef5ed0d JB |
4037 | btrfs_free_path(path); |
4038 | btrfs_end_transaction(trans, root); | |
efe120a0 FH |
4039 | btrfs_err(new_root->fs_info, "Umm, you don't have the default dir" |
4040 | "item, this isn't going to work"); | |
3c04ce01 MX |
4041 | ret = -ENOENT; |
4042 | goto out; | |
6ef5ed0d JB |
4043 | } |
4044 | ||
4045 | btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); | |
4046 | btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); | |
4047 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
4048 | btrfs_free_path(path); | |
4049 | ||
2b0ce2c2 | 4050 | btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL); |
6ef5ed0d | 4051 | btrfs_end_transaction(trans, root); |
3c04ce01 MX |
4052 | out: |
4053 | mnt_drop_write_file(file); | |
4054 | return ret; | |
6ef5ed0d JB |
4055 | } |
4056 | ||
5af3e8cc SB |
4057 | void btrfs_get_block_group_info(struct list_head *groups_list, |
4058 | struct btrfs_ioctl_space_info *space) | |
bf5fc093 JB |
4059 | { |
4060 | struct btrfs_block_group_cache *block_group; | |
4061 | ||
4062 | space->total_bytes = 0; | |
4063 | space->used_bytes = 0; | |
4064 | space->flags = 0; | |
4065 | list_for_each_entry(block_group, groups_list, list) { | |
4066 | space->flags = block_group->flags; | |
4067 | space->total_bytes += block_group->key.offset; | |
4068 | space->used_bytes += | |
4069 | btrfs_block_group_used(&block_group->item); | |
4070 | } | |
4071 | } | |
4072 | ||
48a3b636 | 4073 | static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) |
1406e432 JB |
4074 | { |
4075 | struct btrfs_ioctl_space_args space_args; | |
4076 | struct btrfs_ioctl_space_info space; | |
4077 | struct btrfs_ioctl_space_info *dest; | |
7fde62bf | 4078 | struct btrfs_ioctl_space_info *dest_orig; |
13f2696f | 4079 | struct btrfs_ioctl_space_info __user *user_dest; |
1406e432 | 4080 | struct btrfs_space_info *info; |
bf5fc093 JB |
4081 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, |
4082 | BTRFS_BLOCK_GROUP_SYSTEM, | |
4083 | BTRFS_BLOCK_GROUP_METADATA, | |
4084 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
4085 | int num_types = 4; | |
7fde62bf | 4086 | int alloc_size; |
1406e432 | 4087 | int ret = 0; |
51788b1b | 4088 | u64 slot_count = 0; |
bf5fc093 | 4089 | int i, c; |
1406e432 JB |
4090 | |
4091 | if (copy_from_user(&space_args, | |
4092 | (struct btrfs_ioctl_space_args __user *)arg, | |
4093 | sizeof(space_args))) | |
4094 | return -EFAULT; | |
4095 | ||
bf5fc093 JB |
4096 | for (i = 0; i < num_types; i++) { |
4097 | struct btrfs_space_info *tmp; | |
4098 | ||
4099 | info = NULL; | |
4100 | rcu_read_lock(); | |
4101 | list_for_each_entry_rcu(tmp, &root->fs_info->space_info, | |
4102 | list) { | |
4103 | if (tmp->flags == types[i]) { | |
4104 | info = tmp; | |
4105 | break; | |
4106 | } | |
4107 | } | |
4108 | rcu_read_unlock(); | |
4109 | ||
4110 | if (!info) | |
4111 | continue; | |
4112 | ||
4113 | down_read(&info->groups_sem); | |
4114 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
4115 | if (!list_empty(&info->block_groups[c])) | |
4116 | slot_count++; | |
4117 | } | |
4118 | up_read(&info->groups_sem); | |
4119 | } | |
7fde62bf | 4120 | |
36523e95 DS |
4121 | /* |
4122 | * Global block reserve, exported as a space_info | |
4123 | */ | |
4124 | slot_count++; | |
4125 | ||
7fde62bf CM |
4126 | /* space_slots == 0 means they are asking for a count */ |
4127 | if (space_args.space_slots == 0) { | |
4128 | space_args.total_spaces = slot_count; | |
4129 | goto out; | |
4130 | } | |
bf5fc093 | 4131 | |
51788b1b | 4132 | slot_count = min_t(u64, space_args.space_slots, slot_count); |
bf5fc093 | 4133 | |
7fde62bf | 4134 | alloc_size = sizeof(*dest) * slot_count; |
bf5fc093 | 4135 | |
7fde62bf CM |
4136 | /* we generally have at most 6 or so space infos, one for each raid |
4137 | * level. So, a whole page should be more than enough for everyone | |
4138 | */ | |
09cbfeaf | 4139 | if (alloc_size > PAGE_SIZE) |
7fde62bf CM |
4140 | return -ENOMEM; |
4141 | ||
1406e432 | 4142 | space_args.total_spaces = 0; |
8d2db785 | 4143 | dest = kmalloc(alloc_size, GFP_KERNEL); |
7fde62bf CM |
4144 | if (!dest) |
4145 | return -ENOMEM; | |
4146 | dest_orig = dest; | |
1406e432 | 4147 | |
7fde62bf | 4148 | /* now we have a buffer to copy into */ |
bf5fc093 JB |
4149 | for (i = 0; i < num_types; i++) { |
4150 | struct btrfs_space_info *tmp; | |
4151 | ||
51788b1b DR |
4152 | if (!slot_count) |
4153 | break; | |
4154 | ||
bf5fc093 JB |
4155 | info = NULL; |
4156 | rcu_read_lock(); | |
4157 | list_for_each_entry_rcu(tmp, &root->fs_info->space_info, | |
4158 | list) { | |
4159 | if (tmp->flags == types[i]) { | |
4160 | info = tmp; | |
4161 | break; | |
4162 | } | |
4163 | } | |
4164 | rcu_read_unlock(); | |
7fde62bf | 4165 | |
bf5fc093 JB |
4166 | if (!info) |
4167 | continue; | |
4168 | down_read(&info->groups_sem); | |
4169 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
4170 | if (!list_empty(&info->block_groups[c])) { | |
5af3e8cc SB |
4171 | btrfs_get_block_group_info( |
4172 | &info->block_groups[c], &space); | |
bf5fc093 JB |
4173 | memcpy(dest, &space, sizeof(space)); |
4174 | dest++; | |
4175 | space_args.total_spaces++; | |
51788b1b | 4176 | slot_count--; |
bf5fc093 | 4177 | } |
51788b1b DR |
4178 | if (!slot_count) |
4179 | break; | |
bf5fc093 JB |
4180 | } |
4181 | up_read(&info->groups_sem); | |
1406e432 | 4182 | } |
1406e432 | 4183 | |
36523e95 DS |
4184 | /* |
4185 | * Add global block reserve | |
4186 | */ | |
4187 | if (slot_count) { | |
4188 | struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv; | |
4189 | ||
4190 | spin_lock(&block_rsv->lock); | |
4191 | space.total_bytes = block_rsv->size; | |
4192 | space.used_bytes = block_rsv->size - block_rsv->reserved; | |
4193 | spin_unlock(&block_rsv->lock); | |
4194 | space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; | |
4195 | memcpy(dest, &space, sizeof(space)); | |
4196 | space_args.total_spaces++; | |
4197 | } | |
4198 | ||
2eec6c81 | 4199 | user_dest = (struct btrfs_ioctl_space_info __user *) |
7fde62bf CM |
4200 | (arg + sizeof(struct btrfs_ioctl_space_args)); |
4201 | ||
4202 | if (copy_to_user(user_dest, dest_orig, alloc_size)) | |
4203 | ret = -EFAULT; | |
4204 | ||
4205 | kfree(dest_orig); | |
4206 | out: | |
4207 | if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) | |
1406e432 JB |
4208 | ret = -EFAULT; |
4209 | ||
4210 | return ret; | |
4211 | } | |
4212 | ||
f46b5a66 CH |
4213 | /* |
4214 | * there are many ways the trans_start and trans_end ioctls can lead | |
4215 | * to deadlocks. They should only be used by applications that | |
4216 | * basically own the machine, and have a very in depth understanding | |
4217 | * of all the possible deadlocks and enospc problems. | |
4218 | */ | |
4219 | long btrfs_ioctl_trans_end(struct file *file) | |
4220 | { | |
496ad9aa | 4221 | struct inode *inode = file_inode(file); |
f46b5a66 CH |
4222 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4223 | struct btrfs_trans_handle *trans; | |
f46b5a66 | 4224 | |
f46b5a66 | 4225 | trans = file->private_data; |
1ab86aed SW |
4226 | if (!trans) |
4227 | return -EINVAL; | |
b214107e | 4228 | file->private_data = NULL; |
9ca9ee09 | 4229 | |
1ab86aed SW |
4230 | btrfs_end_transaction(trans, root); |
4231 | ||
a4abeea4 | 4232 | atomic_dec(&root->fs_info->open_ioctl_trans); |
9ca9ee09 | 4233 | |
2a79f17e | 4234 | mnt_drop_write_file(file); |
1ab86aed | 4235 | return 0; |
f46b5a66 CH |
4236 | } |
4237 | ||
9a8c28be MX |
4238 | static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, |
4239 | void __user *argp) | |
46204592 | 4240 | { |
46204592 SW |
4241 | struct btrfs_trans_handle *trans; |
4242 | u64 transid; | |
db5b493a | 4243 | int ret; |
46204592 | 4244 | |
d4edf39b | 4245 | trans = btrfs_attach_transaction_barrier(root); |
ff7c1d33 MX |
4246 | if (IS_ERR(trans)) { |
4247 | if (PTR_ERR(trans) != -ENOENT) | |
4248 | return PTR_ERR(trans); | |
4249 | ||
4250 | /* No running transaction, don't bother */ | |
4251 | transid = root->fs_info->last_trans_committed; | |
4252 | goto out; | |
4253 | } | |
46204592 | 4254 | transid = trans->transid; |
db5b493a | 4255 | ret = btrfs_commit_transaction_async(trans, root, 0); |
8b2b2d3c TI |
4256 | if (ret) { |
4257 | btrfs_end_transaction(trans, root); | |
db5b493a | 4258 | return ret; |
8b2b2d3c | 4259 | } |
ff7c1d33 | 4260 | out: |
46204592 SW |
4261 | if (argp) |
4262 | if (copy_to_user(argp, &transid, sizeof(transid))) | |
4263 | return -EFAULT; | |
4264 | return 0; | |
4265 | } | |
4266 | ||
9a8c28be MX |
4267 | static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, |
4268 | void __user *argp) | |
46204592 | 4269 | { |
46204592 SW |
4270 | u64 transid; |
4271 | ||
4272 | if (argp) { | |
4273 | if (copy_from_user(&transid, argp, sizeof(transid))) | |
4274 | return -EFAULT; | |
4275 | } else { | |
4276 | transid = 0; /* current trans */ | |
4277 | } | |
4278 | return btrfs_wait_for_commit(root, transid); | |
4279 | } | |
4280 | ||
b8e95489 | 4281 | static long btrfs_ioctl_scrub(struct file *file, void __user *arg) |
475f6387 | 4282 | { |
496ad9aa | 4283 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
475f6387 | 4284 | struct btrfs_ioctl_scrub_args *sa; |
b8e95489 | 4285 | int ret; |
475f6387 JS |
4286 | |
4287 | if (!capable(CAP_SYS_ADMIN)) | |
4288 | return -EPERM; | |
4289 | ||
4290 | sa = memdup_user(arg, sizeof(*sa)); | |
4291 | if (IS_ERR(sa)) | |
4292 | return PTR_ERR(sa); | |
4293 | ||
b8e95489 MX |
4294 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) { |
4295 | ret = mnt_want_write_file(file); | |
4296 | if (ret) | |
4297 | goto out; | |
4298 | } | |
4299 | ||
aa1b8cd4 | 4300 | ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end, |
63a212ab SB |
4301 | &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, |
4302 | 0); | |
475f6387 JS |
4303 | |
4304 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
4305 | ret = -EFAULT; | |
4306 | ||
b8e95489 MX |
4307 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) |
4308 | mnt_drop_write_file(file); | |
4309 | out: | |
475f6387 JS |
4310 | kfree(sa); |
4311 | return ret; | |
4312 | } | |
4313 | ||
4314 | static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) | |
4315 | { | |
4316 | if (!capable(CAP_SYS_ADMIN)) | |
4317 | return -EPERM; | |
4318 | ||
aa1b8cd4 | 4319 | return btrfs_scrub_cancel(root->fs_info); |
475f6387 JS |
4320 | } |
4321 | ||
4322 | static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, | |
4323 | void __user *arg) | |
4324 | { | |
4325 | struct btrfs_ioctl_scrub_args *sa; | |
4326 | int ret; | |
4327 | ||
4328 | if (!capable(CAP_SYS_ADMIN)) | |
4329 | return -EPERM; | |
4330 | ||
4331 | sa = memdup_user(arg, sizeof(*sa)); | |
4332 | if (IS_ERR(sa)) | |
4333 | return PTR_ERR(sa); | |
4334 | ||
4335 | ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); | |
4336 | ||
4337 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
4338 | ret = -EFAULT; | |
4339 | ||
4340 | kfree(sa); | |
4341 | return ret; | |
4342 | } | |
4343 | ||
c11d2c23 | 4344 | static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, |
b27f7c0c | 4345 | void __user *arg) |
c11d2c23 SB |
4346 | { |
4347 | struct btrfs_ioctl_get_dev_stats *sa; | |
4348 | int ret; | |
4349 | ||
c11d2c23 SB |
4350 | sa = memdup_user(arg, sizeof(*sa)); |
4351 | if (IS_ERR(sa)) | |
4352 | return PTR_ERR(sa); | |
4353 | ||
b27f7c0c DS |
4354 | if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { |
4355 | kfree(sa); | |
4356 | return -EPERM; | |
4357 | } | |
4358 | ||
4359 | ret = btrfs_get_dev_stats(root, sa); | |
c11d2c23 SB |
4360 | |
4361 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
4362 | ret = -EFAULT; | |
4363 | ||
4364 | kfree(sa); | |
4365 | return ret; | |
4366 | } | |
4367 | ||
3f6bcfbd SB |
4368 | static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) |
4369 | { | |
4370 | struct btrfs_ioctl_dev_replace_args *p; | |
4371 | int ret; | |
4372 | ||
4373 | if (!capable(CAP_SYS_ADMIN)) | |
4374 | return -EPERM; | |
4375 | ||
4376 | p = memdup_user(arg, sizeof(*p)); | |
4377 | if (IS_ERR(p)) | |
4378 | return PTR_ERR(p); | |
4379 | ||
4380 | switch (p->cmd) { | |
4381 | case BTRFS_IOCTL_DEV_REPLACE_CMD_START: | |
adfa97cb ID |
4382 | if (root->fs_info->sb->s_flags & MS_RDONLY) { |
4383 | ret = -EROFS; | |
4384 | goto out; | |
4385 | } | |
3f6bcfbd SB |
4386 | if (atomic_xchg( |
4387 | &root->fs_info->mutually_exclusive_operation_running, | |
4388 | 1)) { | |
e57138b3 | 4389 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
3f6bcfbd SB |
4390 | } else { |
4391 | ret = btrfs_dev_replace_start(root, p); | |
4392 | atomic_set( | |
4393 | &root->fs_info->mutually_exclusive_operation_running, | |
4394 | 0); | |
4395 | } | |
4396 | break; | |
4397 | case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: | |
4398 | btrfs_dev_replace_status(root->fs_info, p); | |
4399 | ret = 0; | |
4400 | break; | |
4401 | case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: | |
4402 | ret = btrfs_dev_replace_cancel(root->fs_info, p); | |
4403 | break; | |
4404 | default: | |
4405 | ret = -EINVAL; | |
4406 | break; | |
4407 | } | |
4408 | ||
4409 | if (copy_to_user(arg, p, sizeof(*p))) | |
4410 | ret = -EFAULT; | |
adfa97cb | 4411 | out: |
3f6bcfbd SB |
4412 | kfree(p); |
4413 | return ret; | |
4414 | } | |
4415 | ||
d7728c96 JS |
4416 | static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) |
4417 | { | |
4418 | int ret = 0; | |
4419 | int i; | |
740c3d22 | 4420 | u64 rel_ptr; |
d7728c96 | 4421 | int size; |
806468f8 | 4422 | struct btrfs_ioctl_ino_path_args *ipa = NULL; |
d7728c96 JS |
4423 | struct inode_fs_paths *ipath = NULL; |
4424 | struct btrfs_path *path; | |
4425 | ||
82b22ac8 | 4426 | if (!capable(CAP_DAC_READ_SEARCH)) |
d7728c96 JS |
4427 | return -EPERM; |
4428 | ||
4429 | path = btrfs_alloc_path(); | |
4430 | if (!path) { | |
4431 | ret = -ENOMEM; | |
4432 | goto out; | |
4433 | } | |
4434 | ||
4435 | ipa = memdup_user(arg, sizeof(*ipa)); | |
4436 | if (IS_ERR(ipa)) { | |
4437 | ret = PTR_ERR(ipa); | |
4438 | ipa = NULL; | |
4439 | goto out; | |
4440 | } | |
4441 | ||
4442 | size = min_t(u32, ipa->size, 4096); | |
4443 | ipath = init_ipath(size, root, path); | |
4444 | if (IS_ERR(ipath)) { | |
4445 | ret = PTR_ERR(ipath); | |
4446 | ipath = NULL; | |
4447 | goto out; | |
4448 | } | |
4449 | ||
4450 | ret = paths_from_inode(ipa->inum, ipath); | |
4451 | if (ret < 0) | |
4452 | goto out; | |
4453 | ||
4454 | for (i = 0; i < ipath->fspath->elem_cnt; ++i) { | |
745c4d8e JM |
4455 | rel_ptr = ipath->fspath->val[i] - |
4456 | (u64)(unsigned long)ipath->fspath->val; | |
740c3d22 | 4457 | ipath->fspath->val[i] = rel_ptr; |
d7728c96 JS |
4458 | } |
4459 | ||
745c4d8e JM |
4460 | ret = copy_to_user((void *)(unsigned long)ipa->fspath, |
4461 | (void *)(unsigned long)ipath->fspath, size); | |
d7728c96 JS |
4462 | if (ret) { |
4463 | ret = -EFAULT; | |
4464 | goto out; | |
4465 | } | |
4466 | ||
4467 | out: | |
4468 | btrfs_free_path(path); | |
4469 | free_ipath(ipath); | |
4470 | kfree(ipa); | |
4471 | ||
4472 | return ret; | |
4473 | } | |
4474 | ||
4475 | static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) | |
4476 | { | |
4477 | struct btrfs_data_container *inodes = ctx; | |
4478 | const size_t c = 3 * sizeof(u64); | |
4479 | ||
4480 | if (inodes->bytes_left >= c) { | |
4481 | inodes->bytes_left -= c; | |
4482 | inodes->val[inodes->elem_cnt] = inum; | |
4483 | inodes->val[inodes->elem_cnt + 1] = offset; | |
4484 | inodes->val[inodes->elem_cnt + 2] = root; | |
4485 | inodes->elem_cnt += 3; | |
4486 | } else { | |
4487 | inodes->bytes_missing += c - inodes->bytes_left; | |
4488 | inodes->bytes_left = 0; | |
4489 | inodes->elem_missed += 3; | |
4490 | } | |
4491 | ||
4492 | return 0; | |
4493 | } | |
4494 | ||
4495 | static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, | |
4496 | void __user *arg) | |
4497 | { | |
4498 | int ret = 0; | |
4499 | int size; | |
d7728c96 JS |
4500 | struct btrfs_ioctl_logical_ino_args *loi; |
4501 | struct btrfs_data_container *inodes = NULL; | |
4502 | struct btrfs_path *path = NULL; | |
d7728c96 JS |
4503 | |
4504 | if (!capable(CAP_SYS_ADMIN)) | |
4505 | return -EPERM; | |
4506 | ||
4507 | loi = memdup_user(arg, sizeof(*loi)); | |
4508 | if (IS_ERR(loi)) { | |
4509 | ret = PTR_ERR(loi); | |
4510 | loi = NULL; | |
4511 | goto out; | |
4512 | } | |
4513 | ||
4514 | path = btrfs_alloc_path(); | |
4515 | if (!path) { | |
4516 | ret = -ENOMEM; | |
4517 | goto out; | |
4518 | } | |
4519 | ||
ee22184b | 4520 | size = min_t(u32, loi->size, SZ_64K); |
d7728c96 JS |
4521 | inodes = init_data_container(size); |
4522 | if (IS_ERR(inodes)) { | |
4523 | ret = PTR_ERR(inodes); | |
4524 | inodes = NULL; | |
4525 | goto out; | |
4526 | } | |
4527 | ||
df031f07 LB |
4528 | ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, |
4529 | build_ino_list, inodes); | |
4530 | if (ret == -EINVAL) | |
d7728c96 JS |
4531 | ret = -ENOENT; |
4532 | if (ret < 0) | |
4533 | goto out; | |
4534 | ||
745c4d8e JM |
4535 | ret = copy_to_user((void *)(unsigned long)loi->inodes, |
4536 | (void *)(unsigned long)inodes, size); | |
d7728c96 JS |
4537 | if (ret) |
4538 | ret = -EFAULT; | |
4539 | ||
4540 | out: | |
4541 | btrfs_free_path(path); | |
425d17a2 | 4542 | vfree(inodes); |
d7728c96 JS |
4543 | kfree(loi); |
4544 | ||
4545 | return ret; | |
4546 | } | |
4547 | ||
19a39dce | 4548 | void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, |
c9e9f97b ID |
4549 | struct btrfs_ioctl_balance_args *bargs) |
4550 | { | |
4551 | struct btrfs_balance_control *bctl = fs_info->balance_ctl; | |
4552 | ||
4553 | bargs->flags = bctl->flags; | |
4554 | ||
837d5b6e ID |
4555 | if (atomic_read(&fs_info->balance_running)) |
4556 | bargs->state |= BTRFS_BALANCE_STATE_RUNNING; | |
4557 | if (atomic_read(&fs_info->balance_pause_req)) | |
4558 | bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; | |
a7e99c69 ID |
4559 | if (atomic_read(&fs_info->balance_cancel_req)) |
4560 | bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; | |
837d5b6e | 4561 | |
c9e9f97b ID |
4562 | memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); |
4563 | memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); | |
4564 | memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); | |
19a39dce ID |
4565 | |
4566 | if (lock) { | |
4567 | spin_lock(&fs_info->balance_lock); | |
4568 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
4569 | spin_unlock(&fs_info->balance_lock); | |
4570 | } else { | |
4571 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
4572 | } | |
c9e9f97b ID |
4573 | } |
4574 | ||
9ba1f6e4 | 4575 | static long btrfs_ioctl_balance(struct file *file, void __user *arg) |
c9e9f97b | 4576 | { |
496ad9aa | 4577 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
c9e9f97b ID |
4578 | struct btrfs_fs_info *fs_info = root->fs_info; |
4579 | struct btrfs_ioctl_balance_args *bargs; | |
4580 | struct btrfs_balance_control *bctl; | |
ed0fb78f | 4581 | bool need_unlock; /* for mut. excl. ops lock */ |
c9e9f97b ID |
4582 | int ret; |
4583 | ||
4584 | if (!capable(CAP_SYS_ADMIN)) | |
4585 | return -EPERM; | |
4586 | ||
e54bfa31 | 4587 | ret = mnt_want_write_file(file); |
9ba1f6e4 LB |
4588 | if (ret) |
4589 | return ret; | |
4590 | ||
ed0fb78f ID |
4591 | again: |
4592 | if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { | |
4593 | mutex_lock(&fs_info->volume_mutex); | |
4594 | mutex_lock(&fs_info->balance_mutex); | |
4595 | need_unlock = true; | |
4596 | goto locked; | |
4597 | } | |
4598 | ||
4599 | /* | |
4600 | * mut. excl. ops lock is locked. Three possibilites: | |
4601 | * (1) some other op is running | |
4602 | * (2) balance is running | |
4603 | * (3) balance is paused -- special case (think resume) | |
4604 | */ | |
c9e9f97b | 4605 | mutex_lock(&fs_info->balance_mutex); |
ed0fb78f ID |
4606 | if (fs_info->balance_ctl) { |
4607 | /* this is either (2) or (3) */ | |
4608 | if (!atomic_read(&fs_info->balance_running)) { | |
4609 | mutex_unlock(&fs_info->balance_mutex); | |
4610 | if (!mutex_trylock(&fs_info->volume_mutex)) | |
4611 | goto again; | |
4612 | mutex_lock(&fs_info->balance_mutex); | |
4613 | ||
4614 | if (fs_info->balance_ctl && | |
4615 | !atomic_read(&fs_info->balance_running)) { | |
4616 | /* this is (3) */ | |
4617 | need_unlock = false; | |
4618 | goto locked; | |
4619 | } | |
4620 | ||
4621 | mutex_unlock(&fs_info->balance_mutex); | |
4622 | mutex_unlock(&fs_info->volume_mutex); | |
4623 | goto again; | |
4624 | } else { | |
4625 | /* this is (2) */ | |
4626 | mutex_unlock(&fs_info->balance_mutex); | |
4627 | ret = -EINPROGRESS; | |
4628 | goto out; | |
4629 | } | |
4630 | } else { | |
4631 | /* this is (1) */ | |
4632 | mutex_unlock(&fs_info->balance_mutex); | |
e57138b3 | 4633 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
ed0fb78f ID |
4634 | goto out; |
4635 | } | |
4636 | ||
4637 | locked: | |
4638 | BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); | |
c9e9f97b ID |
4639 | |
4640 | if (arg) { | |
4641 | bargs = memdup_user(arg, sizeof(*bargs)); | |
4642 | if (IS_ERR(bargs)) { | |
4643 | ret = PTR_ERR(bargs); | |
ed0fb78f | 4644 | goto out_unlock; |
c9e9f97b | 4645 | } |
de322263 ID |
4646 | |
4647 | if (bargs->flags & BTRFS_BALANCE_RESUME) { | |
4648 | if (!fs_info->balance_ctl) { | |
4649 | ret = -ENOTCONN; | |
4650 | goto out_bargs; | |
4651 | } | |
4652 | ||
4653 | bctl = fs_info->balance_ctl; | |
4654 | spin_lock(&fs_info->balance_lock); | |
4655 | bctl->flags |= BTRFS_BALANCE_RESUME; | |
4656 | spin_unlock(&fs_info->balance_lock); | |
4657 | ||
4658 | goto do_balance; | |
4659 | } | |
c9e9f97b ID |
4660 | } else { |
4661 | bargs = NULL; | |
4662 | } | |
4663 | ||
ed0fb78f | 4664 | if (fs_info->balance_ctl) { |
837d5b6e ID |
4665 | ret = -EINPROGRESS; |
4666 | goto out_bargs; | |
4667 | } | |
4668 | ||
8d2db785 | 4669 | bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); |
c9e9f97b ID |
4670 | if (!bctl) { |
4671 | ret = -ENOMEM; | |
4672 | goto out_bargs; | |
4673 | } | |
4674 | ||
4675 | bctl->fs_info = fs_info; | |
4676 | if (arg) { | |
4677 | memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); | |
4678 | memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); | |
4679 | memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); | |
4680 | ||
4681 | bctl->flags = bargs->flags; | |
f43ffb60 ID |
4682 | } else { |
4683 | /* balance everything - no filters */ | |
4684 | bctl->flags |= BTRFS_BALANCE_TYPE_MASK; | |
c9e9f97b ID |
4685 | } |
4686 | ||
8eb93459 DS |
4687 | if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { |
4688 | ret = -EINVAL; | |
0f89abf5 | 4689 | goto out_bctl; |
8eb93459 DS |
4690 | } |
4691 | ||
de322263 | 4692 | do_balance: |
c9e9f97b | 4693 | /* |
ed0fb78f ID |
4694 | * Ownership of bctl and mutually_exclusive_operation_running |
4695 | * goes to to btrfs_balance. bctl is freed in __cancel_balance, | |
4696 | * or, if restriper was paused all the way until unmount, in | |
4697 | * free_fs_info. mutually_exclusive_operation_running is | |
4698 | * cleared in __cancel_balance. | |
c9e9f97b | 4699 | */ |
ed0fb78f ID |
4700 | need_unlock = false; |
4701 | ||
4702 | ret = btrfs_balance(bctl, bargs); | |
0f89abf5 | 4703 | bctl = NULL; |
ed0fb78f | 4704 | |
c9e9f97b ID |
4705 | if (arg) { |
4706 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4707 | ret = -EFAULT; | |
4708 | } | |
4709 | ||
0f89abf5 CE |
4710 | out_bctl: |
4711 | kfree(bctl); | |
c9e9f97b ID |
4712 | out_bargs: |
4713 | kfree(bargs); | |
ed0fb78f | 4714 | out_unlock: |
c9e9f97b ID |
4715 | mutex_unlock(&fs_info->balance_mutex); |
4716 | mutex_unlock(&fs_info->volume_mutex); | |
ed0fb78f ID |
4717 | if (need_unlock) |
4718 | atomic_set(&fs_info->mutually_exclusive_operation_running, 0); | |
4719 | out: | |
e54bfa31 | 4720 | mnt_drop_write_file(file); |
c9e9f97b ID |
4721 | return ret; |
4722 | } | |
4723 | ||
837d5b6e ID |
4724 | static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) |
4725 | { | |
4726 | if (!capable(CAP_SYS_ADMIN)) | |
4727 | return -EPERM; | |
4728 | ||
4729 | switch (cmd) { | |
4730 | case BTRFS_BALANCE_CTL_PAUSE: | |
4731 | return btrfs_pause_balance(root->fs_info); | |
a7e99c69 ID |
4732 | case BTRFS_BALANCE_CTL_CANCEL: |
4733 | return btrfs_cancel_balance(root->fs_info); | |
837d5b6e ID |
4734 | } |
4735 | ||
4736 | return -EINVAL; | |
4737 | } | |
4738 | ||
19a39dce ID |
4739 | static long btrfs_ioctl_balance_progress(struct btrfs_root *root, |
4740 | void __user *arg) | |
4741 | { | |
4742 | struct btrfs_fs_info *fs_info = root->fs_info; | |
4743 | struct btrfs_ioctl_balance_args *bargs; | |
4744 | int ret = 0; | |
4745 | ||
4746 | if (!capable(CAP_SYS_ADMIN)) | |
4747 | return -EPERM; | |
4748 | ||
4749 | mutex_lock(&fs_info->balance_mutex); | |
4750 | if (!fs_info->balance_ctl) { | |
4751 | ret = -ENOTCONN; | |
4752 | goto out; | |
4753 | } | |
4754 | ||
8d2db785 | 4755 | bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); |
19a39dce ID |
4756 | if (!bargs) { |
4757 | ret = -ENOMEM; | |
4758 | goto out; | |
4759 | } | |
4760 | ||
4761 | update_ioctl_balance_args(fs_info, 1, bargs); | |
4762 | ||
4763 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4764 | ret = -EFAULT; | |
4765 | ||
4766 | kfree(bargs); | |
4767 | out: | |
4768 | mutex_unlock(&fs_info->balance_mutex); | |
4769 | return ret; | |
4770 | } | |
4771 | ||
905b0dda | 4772 | static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) |
5d13a37b | 4773 | { |
496ad9aa | 4774 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4775 | struct btrfs_ioctl_quota_ctl_args *sa; |
4776 | struct btrfs_trans_handle *trans = NULL; | |
4777 | int ret; | |
4778 | int err; | |
4779 | ||
4780 | if (!capable(CAP_SYS_ADMIN)) | |
4781 | return -EPERM; | |
4782 | ||
905b0dda MX |
4783 | ret = mnt_want_write_file(file); |
4784 | if (ret) | |
4785 | return ret; | |
5d13a37b AJ |
4786 | |
4787 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4788 | if (IS_ERR(sa)) { |
4789 | ret = PTR_ERR(sa); | |
4790 | goto drop_write; | |
4791 | } | |
5d13a37b | 4792 | |
7708f029 | 4793 | down_write(&root->fs_info->subvol_sem); |
2f232036 JS |
4794 | trans = btrfs_start_transaction(root->fs_info->tree_root, 2); |
4795 | if (IS_ERR(trans)) { | |
4796 | ret = PTR_ERR(trans); | |
4797 | goto out; | |
5d13a37b AJ |
4798 | } |
4799 | ||
4800 | switch (sa->cmd) { | |
4801 | case BTRFS_QUOTA_CTL_ENABLE: | |
4802 | ret = btrfs_quota_enable(trans, root->fs_info); | |
4803 | break; | |
4804 | case BTRFS_QUOTA_CTL_DISABLE: | |
4805 | ret = btrfs_quota_disable(trans, root->fs_info); | |
4806 | break; | |
5d13a37b AJ |
4807 | default: |
4808 | ret = -EINVAL; | |
4809 | break; | |
4810 | } | |
4811 | ||
2f232036 JS |
4812 | err = btrfs_commit_transaction(trans, root->fs_info->tree_root); |
4813 | if (err && !ret) | |
4814 | ret = err; | |
5d13a37b AJ |
4815 | out: |
4816 | kfree(sa); | |
7708f029 | 4817 | up_write(&root->fs_info->subvol_sem); |
905b0dda MX |
4818 | drop_write: |
4819 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4820 | return ret; |
4821 | } | |
4822 | ||
905b0dda | 4823 | static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) |
5d13a37b | 4824 | { |
496ad9aa | 4825 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4826 | struct btrfs_ioctl_qgroup_assign_args *sa; |
4827 | struct btrfs_trans_handle *trans; | |
4828 | int ret; | |
4829 | int err; | |
4830 | ||
4831 | if (!capable(CAP_SYS_ADMIN)) | |
4832 | return -EPERM; | |
4833 | ||
905b0dda MX |
4834 | ret = mnt_want_write_file(file); |
4835 | if (ret) | |
4836 | return ret; | |
5d13a37b AJ |
4837 | |
4838 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4839 | if (IS_ERR(sa)) { |
4840 | ret = PTR_ERR(sa); | |
4841 | goto drop_write; | |
4842 | } | |
5d13a37b AJ |
4843 | |
4844 | trans = btrfs_join_transaction(root); | |
4845 | if (IS_ERR(trans)) { | |
4846 | ret = PTR_ERR(trans); | |
4847 | goto out; | |
4848 | } | |
4849 | ||
4850 | /* FIXME: check if the IDs really exist */ | |
4851 | if (sa->assign) { | |
4852 | ret = btrfs_add_qgroup_relation(trans, root->fs_info, | |
4853 | sa->src, sa->dst); | |
4854 | } else { | |
4855 | ret = btrfs_del_qgroup_relation(trans, root->fs_info, | |
4856 | sa->src, sa->dst); | |
4857 | } | |
4858 | ||
e082f563 QW |
4859 | /* update qgroup status and info */ |
4860 | err = btrfs_run_qgroups(trans, root->fs_info); | |
4861 | if (err < 0) | |
a4553fef | 4862 | btrfs_std_error(root->fs_info, ret, |
e082f563 | 4863 | "failed to update qgroup status and info\n"); |
5d13a37b AJ |
4864 | err = btrfs_end_transaction(trans, root); |
4865 | if (err && !ret) | |
4866 | ret = err; | |
4867 | ||
4868 | out: | |
4869 | kfree(sa); | |
905b0dda MX |
4870 | drop_write: |
4871 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4872 | return ret; |
4873 | } | |
4874 | ||
905b0dda | 4875 | static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) |
5d13a37b | 4876 | { |
496ad9aa | 4877 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4878 | struct btrfs_ioctl_qgroup_create_args *sa; |
4879 | struct btrfs_trans_handle *trans; | |
4880 | int ret; | |
4881 | int err; | |
4882 | ||
4883 | if (!capable(CAP_SYS_ADMIN)) | |
4884 | return -EPERM; | |
4885 | ||
905b0dda MX |
4886 | ret = mnt_want_write_file(file); |
4887 | if (ret) | |
4888 | return ret; | |
5d13a37b AJ |
4889 | |
4890 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4891 | if (IS_ERR(sa)) { |
4892 | ret = PTR_ERR(sa); | |
4893 | goto drop_write; | |
4894 | } | |
5d13a37b | 4895 | |
d86e56cf MX |
4896 | if (!sa->qgroupid) { |
4897 | ret = -EINVAL; | |
4898 | goto out; | |
4899 | } | |
4900 | ||
5d13a37b AJ |
4901 | trans = btrfs_join_transaction(root); |
4902 | if (IS_ERR(trans)) { | |
4903 | ret = PTR_ERR(trans); | |
4904 | goto out; | |
4905 | } | |
4906 | ||
4907 | /* FIXME: check if the IDs really exist */ | |
4908 | if (sa->create) { | |
4087cf24 | 4909 | ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid); |
5d13a37b AJ |
4910 | } else { |
4911 | ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid); | |
4912 | } | |
4913 | ||
4914 | err = btrfs_end_transaction(trans, root); | |
4915 | if (err && !ret) | |
4916 | ret = err; | |
4917 | ||
4918 | out: | |
4919 | kfree(sa); | |
905b0dda MX |
4920 | drop_write: |
4921 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4922 | return ret; |
4923 | } | |
4924 | ||
905b0dda | 4925 | static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) |
5d13a37b | 4926 | { |
496ad9aa | 4927 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
5d13a37b AJ |
4928 | struct btrfs_ioctl_qgroup_limit_args *sa; |
4929 | struct btrfs_trans_handle *trans; | |
4930 | int ret; | |
4931 | int err; | |
4932 | u64 qgroupid; | |
4933 | ||
4934 | if (!capable(CAP_SYS_ADMIN)) | |
4935 | return -EPERM; | |
4936 | ||
905b0dda MX |
4937 | ret = mnt_want_write_file(file); |
4938 | if (ret) | |
4939 | return ret; | |
5d13a37b AJ |
4940 | |
4941 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4942 | if (IS_ERR(sa)) { |
4943 | ret = PTR_ERR(sa); | |
4944 | goto drop_write; | |
4945 | } | |
5d13a37b AJ |
4946 | |
4947 | trans = btrfs_join_transaction(root); | |
4948 | if (IS_ERR(trans)) { | |
4949 | ret = PTR_ERR(trans); | |
4950 | goto out; | |
4951 | } | |
4952 | ||
4953 | qgroupid = sa->qgroupid; | |
4954 | if (!qgroupid) { | |
4955 | /* take the current subvol as qgroup */ | |
4956 | qgroupid = root->root_key.objectid; | |
4957 | } | |
4958 | ||
4959 | /* FIXME: check if the IDs really exist */ | |
4960 | ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim); | |
4961 | ||
4962 | err = btrfs_end_transaction(trans, root); | |
4963 | if (err && !ret) | |
4964 | ret = err; | |
4965 | ||
4966 | out: | |
4967 | kfree(sa); | |
905b0dda MX |
4968 | drop_write: |
4969 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4970 | return ret; |
4971 | } | |
4972 | ||
2f232036 JS |
4973 | static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) |
4974 | { | |
6d0379ec | 4975 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
2f232036 JS |
4976 | struct btrfs_ioctl_quota_rescan_args *qsa; |
4977 | int ret; | |
4978 | ||
4979 | if (!capable(CAP_SYS_ADMIN)) | |
4980 | return -EPERM; | |
4981 | ||
4982 | ret = mnt_want_write_file(file); | |
4983 | if (ret) | |
4984 | return ret; | |
4985 | ||
4986 | qsa = memdup_user(arg, sizeof(*qsa)); | |
4987 | if (IS_ERR(qsa)) { | |
4988 | ret = PTR_ERR(qsa); | |
4989 | goto drop_write; | |
4990 | } | |
4991 | ||
4992 | if (qsa->flags) { | |
4993 | ret = -EINVAL; | |
4994 | goto out; | |
4995 | } | |
4996 | ||
4997 | ret = btrfs_qgroup_rescan(root->fs_info); | |
4998 | ||
4999 | out: | |
5000 | kfree(qsa); | |
5001 | drop_write: | |
5002 | mnt_drop_write_file(file); | |
5003 | return ret; | |
5004 | } | |
5005 | ||
5006 | static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) | |
5007 | { | |
6d0379ec | 5008 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
2f232036 JS |
5009 | struct btrfs_ioctl_quota_rescan_args *qsa; |
5010 | int ret = 0; | |
5011 | ||
5012 | if (!capable(CAP_SYS_ADMIN)) | |
5013 | return -EPERM; | |
5014 | ||
8d2db785 | 5015 | qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); |
2f232036 JS |
5016 | if (!qsa) |
5017 | return -ENOMEM; | |
5018 | ||
5019 | if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { | |
5020 | qsa->flags = 1; | |
5021 | qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; | |
5022 | } | |
5023 | ||
5024 | if (copy_to_user(arg, qsa, sizeof(*qsa))) | |
5025 | ret = -EFAULT; | |
5026 | ||
5027 | kfree(qsa); | |
5028 | return ret; | |
5029 | } | |
5030 | ||
57254b6e JS |
5031 | static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) |
5032 | { | |
54563d41 | 5033 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
57254b6e JS |
5034 | |
5035 | if (!capable(CAP_SYS_ADMIN)) | |
5036 | return -EPERM; | |
5037 | ||
5038 | return btrfs_qgroup_wait_for_completion(root->fs_info); | |
5039 | } | |
5040 | ||
abccd00f HM |
5041 | static long _btrfs_ioctl_set_received_subvol(struct file *file, |
5042 | struct btrfs_ioctl_received_subvol_args *sa) | |
8ea05e3a | 5043 | { |
496ad9aa | 5044 | struct inode *inode = file_inode(file); |
8ea05e3a AB |
5045 | struct btrfs_root *root = BTRFS_I(inode)->root; |
5046 | struct btrfs_root_item *root_item = &root->root_item; | |
5047 | struct btrfs_trans_handle *trans; | |
04b285f3 | 5048 | struct timespec ct = current_fs_time(inode->i_sb); |
8ea05e3a | 5049 | int ret = 0; |
dd5f9615 | 5050 | int received_uuid_changed; |
8ea05e3a | 5051 | |
bd60ea0f DS |
5052 | if (!inode_owner_or_capable(inode)) |
5053 | return -EPERM; | |
5054 | ||
8ea05e3a AB |
5055 | ret = mnt_want_write_file(file); |
5056 | if (ret < 0) | |
5057 | return ret; | |
5058 | ||
5059 | down_write(&root->fs_info->subvol_sem); | |
5060 | ||
5061 | if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { | |
5062 | ret = -EINVAL; | |
5063 | goto out; | |
5064 | } | |
5065 | ||
5066 | if (btrfs_root_readonly(root)) { | |
5067 | ret = -EROFS; | |
5068 | goto out; | |
5069 | } | |
5070 | ||
dd5f9615 SB |
5071 | /* |
5072 | * 1 - root item | |
5073 | * 2 - uuid items (received uuid + subvol uuid) | |
5074 | */ | |
5075 | trans = btrfs_start_transaction(root, 3); | |
8ea05e3a AB |
5076 | if (IS_ERR(trans)) { |
5077 | ret = PTR_ERR(trans); | |
5078 | trans = NULL; | |
5079 | goto out; | |
5080 | } | |
5081 | ||
5082 | sa->rtransid = trans->transid; | |
5083 | sa->rtime.sec = ct.tv_sec; | |
5084 | sa->rtime.nsec = ct.tv_nsec; | |
5085 | ||
dd5f9615 SB |
5086 | received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, |
5087 | BTRFS_UUID_SIZE); | |
5088 | if (received_uuid_changed && | |
5089 | !btrfs_is_empty_uuid(root_item->received_uuid)) | |
5090 | btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, | |
5091 | root_item->received_uuid, | |
5092 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
5093 | root->root_key.objectid); | |
8ea05e3a AB |
5094 | memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); |
5095 | btrfs_set_root_stransid(root_item, sa->stransid); | |
5096 | btrfs_set_root_rtransid(root_item, sa->rtransid); | |
3cae210f QW |
5097 | btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); |
5098 | btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); | |
5099 | btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); | |
5100 | btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); | |
8ea05e3a AB |
5101 | |
5102 | ret = btrfs_update_root(trans, root->fs_info->tree_root, | |
5103 | &root->root_key, &root->root_item); | |
5104 | if (ret < 0) { | |
5105 | btrfs_end_transaction(trans, root); | |
8ea05e3a | 5106 | goto out; |
dd5f9615 SB |
5107 | } |
5108 | if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { | |
5109 | ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, | |
5110 | sa->uuid, | |
5111 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, | |
5112 | root->root_key.objectid); | |
5113 | if (ret < 0 && ret != -EEXIST) { | |
5114 | btrfs_abort_transaction(trans, root, ret); | |
8ea05e3a | 5115 | goto out; |
dd5f9615 SB |
5116 | } |
5117 | } | |
5118 | ret = btrfs_commit_transaction(trans, root); | |
5119 | if (ret < 0) { | |
5120 | btrfs_abort_transaction(trans, root, ret); | |
5121 | goto out; | |
8ea05e3a AB |
5122 | } |
5123 | ||
abccd00f HM |
5124 | out: |
5125 | up_write(&root->fs_info->subvol_sem); | |
5126 | mnt_drop_write_file(file); | |
5127 | return ret; | |
5128 | } | |
5129 | ||
5130 | #ifdef CONFIG_64BIT | |
5131 | static long btrfs_ioctl_set_received_subvol_32(struct file *file, | |
5132 | void __user *arg) | |
5133 | { | |
5134 | struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; | |
5135 | struct btrfs_ioctl_received_subvol_args *args64 = NULL; | |
5136 | int ret = 0; | |
5137 | ||
5138 | args32 = memdup_user(arg, sizeof(*args32)); | |
5139 | if (IS_ERR(args32)) { | |
5140 | ret = PTR_ERR(args32); | |
5141 | args32 = NULL; | |
5142 | goto out; | |
5143 | } | |
5144 | ||
8d2db785 | 5145 | args64 = kmalloc(sizeof(*args64), GFP_KERNEL); |
84dbeb87 DC |
5146 | if (!args64) { |
5147 | ret = -ENOMEM; | |
abccd00f HM |
5148 | goto out; |
5149 | } | |
5150 | ||
5151 | memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); | |
5152 | args64->stransid = args32->stransid; | |
5153 | args64->rtransid = args32->rtransid; | |
5154 | args64->stime.sec = args32->stime.sec; | |
5155 | args64->stime.nsec = args32->stime.nsec; | |
5156 | args64->rtime.sec = args32->rtime.sec; | |
5157 | args64->rtime.nsec = args32->rtime.nsec; | |
5158 | args64->flags = args32->flags; | |
5159 | ||
5160 | ret = _btrfs_ioctl_set_received_subvol(file, args64); | |
5161 | if (ret) | |
5162 | goto out; | |
5163 | ||
5164 | memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); | |
5165 | args32->stransid = args64->stransid; | |
5166 | args32->rtransid = args64->rtransid; | |
5167 | args32->stime.sec = args64->stime.sec; | |
5168 | args32->stime.nsec = args64->stime.nsec; | |
5169 | args32->rtime.sec = args64->rtime.sec; | |
5170 | args32->rtime.nsec = args64->rtime.nsec; | |
5171 | args32->flags = args64->flags; | |
5172 | ||
5173 | ret = copy_to_user(arg, args32, sizeof(*args32)); | |
5174 | if (ret) | |
5175 | ret = -EFAULT; | |
5176 | ||
5177 | out: | |
5178 | kfree(args32); | |
5179 | kfree(args64); | |
5180 | return ret; | |
5181 | } | |
5182 | #endif | |
5183 | ||
5184 | static long btrfs_ioctl_set_received_subvol(struct file *file, | |
5185 | void __user *arg) | |
5186 | { | |
5187 | struct btrfs_ioctl_received_subvol_args *sa = NULL; | |
5188 | int ret = 0; | |
5189 | ||
5190 | sa = memdup_user(arg, sizeof(*sa)); | |
5191 | if (IS_ERR(sa)) { | |
5192 | ret = PTR_ERR(sa); | |
5193 | sa = NULL; | |
5194 | goto out; | |
5195 | } | |
5196 | ||
5197 | ret = _btrfs_ioctl_set_received_subvol(file, sa); | |
5198 | ||
5199 | if (ret) | |
5200 | goto out; | |
5201 | ||
8ea05e3a AB |
5202 | ret = copy_to_user(arg, sa, sizeof(*sa)); |
5203 | if (ret) | |
5204 | ret = -EFAULT; | |
5205 | ||
5206 | out: | |
5207 | kfree(sa); | |
8ea05e3a AB |
5208 | return ret; |
5209 | } | |
5210 | ||
867ab667 | 5211 | static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) |
5212 | { | |
6d0379ec | 5213 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
a1b83ac5 | 5214 | size_t len; |
867ab667 | 5215 | int ret; |
a1b83ac5 AJ |
5216 | char label[BTRFS_LABEL_SIZE]; |
5217 | ||
5218 | spin_lock(&root->fs_info->super_lock); | |
5219 | memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE); | |
5220 | spin_unlock(&root->fs_info->super_lock); | |
5221 | ||
5222 | len = strnlen(label, BTRFS_LABEL_SIZE); | |
867ab667 | 5223 | |
5224 | if (len == BTRFS_LABEL_SIZE) { | |
efe120a0 FH |
5225 | btrfs_warn(root->fs_info, |
5226 | "label is too long, return the first %zu bytes", --len); | |
867ab667 | 5227 | } |
5228 | ||
867ab667 | 5229 | ret = copy_to_user(arg, label, len); |
867ab667 | 5230 | |
5231 | return ret ? -EFAULT : 0; | |
5232 | } | |
5233 | ||
a8bfd4ab | 5234 | static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) |
5235 | { | |
6d0379ec | 5236 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
a8bfd4ab | 5237 | struct btrfs_super_block *super_block = root->fs_info->super_copy; |
5238 | struct btrfs_trans_handle *trans; | |
5239 | char label[BTRFS_LABEL_SIZE]; | |
5240 | int ret; | |
5241 | ||
5242 | if (!capable(CAP_SYS_ADMIN)) | |
5243 | return -EPERM; | |
5244 | ||
5245 | if (copy_from_user(label, arg, sizeof(label))) | |
5246 | return -EFAULT; | |
5247 | ||
5248 | if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { | |
efe120a0 | 5249 | btrfs_err(root->fs_info, "unable to set label with more than %d bytes", |
a8bfd4ab | 5250 | BTRFS_LABEL_SIZE - 1); |
5251 | return -EINVAL; | |
5252 | } | |
5253 | ||
5254 | ret = mnt_want_write_file(file); | |
5255 | if (ret) | |
5256 | return ret; | |
5257 | ||
a8bfd4ab | 5258 | trans = btrfs_start_transaction(root, 0); |
5259 | if (IS_ERR(trans)) { | |
5260 | ret = PTR_ERR(trans); | |
5261 | goto out_unlock; | |
5262 | } | |
5263 | ||
a1b83ac5 | 5264 | spin_lock(&root->fs_info->super_lock); |
a8bfd4ab | 5265 | strcpy(super_block->label, label); |
a1b83ac5 | 5266 | spin_unlock(&root->fs_info->super_lock); |
d0270aca | 5267 | ret = btrfs_commit_transaction(trans, root); |
a8bfd4ab | 5268 | |
5269 | out_unlock: | |
a8bfd4ab | 5270 | mnt_drop_write_file(file); |
5271 | return ret; | |
5272 | } | |
5273 | ||
2eaa055f JM |
5274 | #define INIT_FEATURE_FLAGS(suffix) \ |
5275 | { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ | |
5276 | .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ | |
5277 | .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } | |
5278 | ||
d5131b65 | 5279 | int btrfs_ioctl_get_supported_features(void __user *arg) |
2eaa055f | 5280 | { |
4d4ab6d6 | 5281 | static const struct btrfs_ioctl_feature_flags features[3] = { |
2eaa055f JM |
5282 | INIT_FEATURE_FLAGS(SUPP), |
5283 | INIT_FEATURE_FLAGS(SAFE_SET), | |
5284 | INIT_FEATURE_FLAGS(SAFE_CLEAR) | |
5285 | }; | |
5286 | ||
5287 | if (copy_to_user(arg, &features, sizeof(features))) | |
5288 | return -EFAULT; | |
5289 | ||
5290 | return 0; | |
5291 | } | |
5292 | ||
5293 | static int btrfs_ioctl_get_features(struct file *file, void __user *arg) | |
5294 | { | |
5295 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; | |
5296 | struct btrfs_super_block *super_block = root->fs_info->super_copy; | |
5297 | struct btrfs_ioctl_feature_flags features; | |
5298 | ||
5299 | features.compat_flags = btrfs_super_compat_flags(super_block); | |
5300 | features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); | |
5301 | features.incompat_flags = btrfs_super_incompat_flags(super_block); | |
5302 | ||
5303 | if (copy_to_user(arg, &features, sizeof(features))) | |
5304 | return -EFAULT; | |
5305 | ||
5306 | return 0; | |
5307 | } | |
5308 | ||
3b02a68a JM |
5309 | static int check_feature_bits(struct btrfs_root *root, |
5310 | enum btrfs_feature_set set, | |
2eaa055f JM |
5311 | u64 change_mask, u64 flags, u64 supported_flags, |
5312 | u64 safe_set, u64 safe_clear) | |
5313 | { | |
3b02a68a JM |
5314 | const char *type = btrfs_feature_set_names[set]; |
5315 | char *names; | |
2eaa055f JM |
5316 | u64 disallowed, unsupported; |
5317 | u64 set_mask = flags & change_mask; | |
5318 | u64 clear_mask = ~flags & change_mask; | |
5319 | ||
5320 | unsupported = set_mask & ~supported_flags; | |
5321 | if (unsupported) { | |
3b02a68a JM |
5322 | names = btrfs_printable_features(set, unsupported); |
5323 | if (names) { | |
5324 | btrfs_warn(root->fs_info, | |
5325 | "this kernel does not support the %s feature bit%s", | |
5326 | names, strchr(names, ',') ? "s" : ""); | |
5327 | kfree(names); | |
5328 | } else | |
5329 | btrfs_warn(root->fs_info, | |
2eaa055f JM |
5330 | "this kernel does not support %s bits 0x%llx", |
5331 | type, unsupported); | |
5332 | return -EOPNOTSUPP; | |
5333 | } | |
5334 | ||
5335 | disallowed = set_mask & ~safe_set; | |
5336 | if (disallowed) { | |
3b02a68a JM |
5337 | names = btrfs_printable_features(set, disallowed); |
5338 | if (names) { | |
5339 | btrfs_warn(root->fs_info, | |
5340 | "can't set the %s feature bit%s while mounted", | |
5341 | names, strchr(names, ',') ? "s" : ""); | |
5342 | kfree(names); | |
5343 | } else | |
5344 | btrfs_warn(root->fs_info, | |
2eaa055f JM |
5345 | "can't set %s bits 0x%llx while mounted", |
5346 | type, disallowed); | |
5347 | return -EPERM; | |
5348 | } | |
5349 | ||
5350 | disallowed = clear_mask & ~safe_clear; | |
5351 | if (disallowed) { | |
3b02a68a JM |
5352 | names = btrfs_printable_features(set, disallowed); |
5353 | if (names) { | |
5354 | btrfs_warn(root->fs_info, | |
5355 | "can't clear the %s feature bit%s while mounted", | |
5356 | names, strchr(names, ',') ? "s" : ""); | |
5357 | kfree(names); | |
5358 | } else | |
5359 | btrfs_warn(root->fs_info, | |
2eaa055f JM |
5360 | "can't clear %s bits 0x%llx while mounted", |
5361 | type, disallowed); | |
5362 | return -EPERM; | |
5363 | } | |
5364 | ||
5365 | return 0; | |
5366 | } | |
5367 | ||
5368 | #define check_feature(root, change_mask, flags, mask_base) \ | |
3b02a68a | 5369 | check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \ |
2eaa055f JM |
5370 | BTRFS_FEATURE_ ## mask_base ## _SUPP, \ |
5371 | BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ | |
5372 | BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) | |
5373 | ||
5374 | static int btrfs_ioctl_set_features(struct file *file, void __user *arg) | |
5375 | { | |
5376 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; | |
5377 | struct btrfs_super_block *super_block = root->fs_info->super_copy; | |
5378 | struct btrfs_ioctl_feature_flags flags[2]; | |
5379 | struct btrfs_trans_handle *trans; | |
5380 | u64 newflags; | |
5381 | int ret; | |
5382 | ||
5383 | if (!capable(CAP_SYS_ADMIN)) | |
5384 | return -EPERM; | |
5385 | ||
5386 | if (copy_from_user(flags, arg, sizeof(flags))) | |
5387 | return -EFAULT; | |
5388 | ||
5389 | /* Nothing to do */ | |
5390 | if (!flags[0].compat_flags && !flags[0].compat_ro_flags && | |
5391 | !flags[0].incompat_flags) | |
5392 | return 0; | |
5393 | ||
5394 | ret = check_feature(root, flags[0].compat_flags, | |
5395 | flags[1].compat_flags, COMPAT); | |
5396 | if (ret) | |
5397 | return ret; | |
5398 | ||
5399 | ret = check_feature(root, flags[0].compat_ro_flags, | |
5400 | flags[1].compat_ro_flags, COMPAT_RO); | |
5401 | if (ret) | |
5402 | return ret; | |
5403 | ||
5404 | ret = check_feature(root, flags[0].incompat_flags, | |
5405 | flags[1].incompat_flags, INCOMPAT); | |
5406 | if (ret) | |
5407 | return ret; | |
5408 | ||
7ab19625 DS |
5409 | ret = mnt_want_write_file(file); |
5410 | if (ret) | |
5411 | return ret; | |
5412 | ||
8051aa1a | 5413 | trans = btrfs_start_transaction(root, 0); |
7ab19625 DS |
5414 | if (IS_ERR(trans)) { |
5415 | ret = PTR_ERR(trans); | |
5416 | goto out_drop_write; | |
5417 | } | |
2eaa055f JM |
5418 | |
5419 | spin_lock(&root->fs_info->super_lock); | |
5420 | newflags = btrfs_super_compat_flags(super_block); | |
5421 | newflags |= flags[0].compat_flags & flags[1].compat_flags; | |
5422 | newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); | |
5423 | btrfs_set_super_compat_flags(super_block, newflags); | |
5424 | ||
5425 | newflags = btrfs_super_compat_ro_flags(super_block); | |
5426 | newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; | |
5427 | newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); | |
5428 | btrfs_set_super_compat_ro_flags(super_block, newflags); | |
5429 | ||
5430 | newflags = btrfs_super_incompat_flags(super_block); | |
5431 | newflags |= flags[0].incompat_flags & flags[1].incompat_flags; | |
5432 | newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); | |
5433 | btrfs_set_super_incompat_flags(super_block, newflags); | |
5434 | spin_unlock(&root->fs_info->super_lock); | |
5435 | ||
7ab19625 DS |
5436 | ret = btrfs_commit_transaction(trans, root); |
5437 | out_drop_write: | |
5438 | mnt_drop_write_file(file); | |
5439 | ||
5440 | return ret; | |
2eaa055f JM |
5441 | } |
5442 | ||
f46b5a66 CH |
5443 | long btrfs_ioctl(struct file *file, unsigned int |
5444 | cmd, unsigned long arg) | |
5445 | { | |
496ad9aa | 5446 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
4bcabaa3 | 5447 | void __user *argp = (void __user *)arg; |
f46b5a66 CH |
5448 | |
5449 | switch (cmd) { | |
6cbff00f CH |
5450 | case FS_IOC_GETFLAGS: |
5451 | return btrfs_ioctl_getflags(file, argp); | |
5452 | case FS_IOC_SETFLAGS: | |
5453 | return btrfs_ioctl_setflags(file, argp); | |
5454 | case FS_IOC_GETVERSION: | |
5455 | return btrfs_ioctl_getversion(file, argp); | |
f7039b1d LD |
5456 | case FITRIM: |
5457 | return btrfs_ioctl_fitrim(file, argp); | |
f46b5a66 | 5458 | case BTRFS_IOC_SNAP_CREATE: |
fa0d2b9b | 5459 | return btrfs_ioctl_snap_create(file, argp, 0); |
fdfb1e4f | 5460 | case BTRFS_IOC_SNAP_CREATE_V2: |
fa0d2b9b | 5461 | return btrfs_ioctl_snap_create_v2(file, argp, 0); |
3de4586c | 5462 | case BTRFS_IOC_SUBVOL_CREATE: |
fa0d2b9b | 5463 | return btrfs_ioctl_snap_create(file, argp, 1); |
6f72c7e2 AJ |
5464 | case BTRFS_IOC_SUBVOL_CREATE_V2: |
5465 | return btrfs_ioctl_snap_create_v2(file, argp, 1); | |
76dda93c YZ |
5466 | case BTRFS_IOC_SNAP_DESTROY: |
5467 | return btrfs_ioctl_snap_destroy(file, argp); | |
0caa102d LZ |
5468 | case BTRFS_IOC_SUBVOL_GETFLAGS: |
5469 | return btrfs_ioctl_subvol_getflags(file, argp); | |
5470 | case BTRFS_IOC_SUBVOL_SETFLAGS: | |
5471 | return btrfs_ioctl_subvol_setflags(file, argp); | |
6ef5ed0d JB |
5472 | case BTRFS_IOC_DEFAULT_SUBVOL: |
5473 | return btrfs_ioctl_default_subvol(file, argp); | |
f46b5a66 | 5474 | case BTRFS_IOC_DEFRAG: |
1e701a32 CM |
5475 | return btrfs_ioctl_defrag(file, NULL); |
5476 | case BTRFS_IOC_DEFRAG_RANGE: | |
5477 | return btrfs_ioctl_defrag(file, argp); | |
f46b5a66 | 5478 | case BTRFS_IOC_RESIZE: |
198605a8 | 5479 | return btrfs_ioctl_resize(file, argp); |
f46b5a66 | 5480 | case BTRFS_IOC_ADD_DEV: |
4bcabaa3 | 5481 | return btrfs_ioctl_add_dev(root, argp); |
f46b5a66 | 5482 | case BTRFS_IOC_RM_DEV: |
da24927b | 5483 | return btrfs_ioctl_rm_dev(file, argp); |
475f6387 JS |
5484 | case BTRFS_IOC_FS_INFO: |
5485 | return btrfs_ioctl_fs_info(root, argp); | |
5486 | case BTRFS_IOC_DEV_INFO: | |
5487 | return btrfs_ioctl_dev_info(root, argp); | |
f46b5a66 | 5488 | case BTRFS_IOC_BALANCE: |
9ba1f6e4 | 5489 | return btrfs_ioctl_balance(file, NULL); |
f46b5a66 CH |
5490 | case BTRFS_IOC_TRANS_START: |
5491 | return btrfs_ioctl_trans_start(file); | |
5492 | case BTRFS_IOC_TRANS_END: | |
5493 | return btrfs_ioctl_trans_end(file); | |
ac8e9819 CM |
5494 | case BTRFS_IOC_TREE_SEARCH: |
5495 | return btrfs_ioctl_tree_search(file, argp); | |
cc68a8a5 GH |
5496 | case BTRFS_IOC_TREE_SEARCH_V2: |
5497 | return btrfs_ioctl_tree_search_v2(file, argp); | |
ac8e9819 CM |
5498 | case BTRFS_IOC_INO_LOOKUP: |
5499 | return btrfs_ioctl_ino_lookup(file, argp); | |
d7728c96 JS |
5500 | case BTRFS_IOC_INO_PATHS: |
5501 | return btrfs_ioctl_ino_to_path(root, argp); | |
5502 | case BTRFS_IOC_LOGICAL_INO: | |
5503 | return btrfs_ioctl_logical_to_ino(root, argp); | |
1406e432 JB |
5504 | case BTRFS_IOC_SPACE_INFO: |
5505 | return btrfs_ioctl_space_info(root, argp); | |
9b199859 FDBM |
5506 | case BTRFS_IOC_SYNC: { |
5507 | int ret; | |
5508 | ||
6c255e67 | 5509 | ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1); |
9b199859 FDBM |
5510 | if (ret) |
5511 | return ret; | |
ddb52f4f | 5512 | ret = btrfs_sync_fs(file_inode(file)->i_sb, 1); |
2fad4e83 DS |
5513 | /* |
5514 | * The transaction thread may want to do more work, | |
5515 | * namely it pokes the cleaner ktread that will start | |
5516 | * processing uncleaned subvols. | |
5517 | */ | |
5518 | wake_up_process(root->fs_info->transaction_kthread); | |
9b199859 FDBM |
5519 | return ret; |
5520 | } | |
46204592 | 5521 | case BTRFS_IOC_START_SYNC: |
9a8c28be | 5522 | return btrfs_ioctl_start_sync(root, argp); |
46204592 | 5523 | case BTRFS_IOC_WAIT_SYNC: |
9a8c28be | 5524 | return btrfs_ioctl_wait_sync(root, argp); |
475f6387 | 5525 | case BTRFS_IOC_SCRUB: |
b8e95489 | 5526 | return btrfs_ioctl_scrub(file, argp); |
475f6387 JS |
5527 | case BTRFS_IOC_SCRUB_CANCEL: |
5528 | return btrfs_ioctl_scrub_cancel(root, argp); | |
5529 | case BTRFS_IOC_SCRUB_PROGRESS: | |
5530 | return btrfs_ioctl_scrub_progress(root, argp); | |
c9e9f97b | 5531 | case BTRFS_IOC_BALANCE_V2: |
9ba1f6e4 | 5532 | return btrfs_ioctl_balance(file, argp); |
837d5b6e ID |
5533 | case BTRFS_IOC_BALANCE_CTL: |
5534 | return btrfs_ioctl_balance_ctl(root, arg); | |
19a39dce ID |
5535 | case BTRFS_IOC_BALANCE_PROGRESS: |
5536 | return btrfs_ioctl_balance_progress(root, argp); | |
8ea05e3a AB |
5537 | case BTRFS_IOC_SET_RECEIVED_SUBVOL: |
5538 | return btrfs_ioctl_set_received_subvol(file, argp); | |
abccd00f HM |
5539 | #ifdef CONFIG_64BIT |
5540 | case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: | |
5541 | return btrfs_ioctl_set_received_subvol_32(file, argp); | |
5542 | #endif | |
31db9f7c AB |
5543 | case BTRFS_IOC_SEND: |
5544 | return btrfs_ioctl_send(file, argp); | |
c11d2c23 | 5545 | case BTRFS_IOC_GET_DEV_STATS: |
b27f7c0c | 5546 | return btrfs_ioctl_get_dev_stats(root, argp); |
5d13a37b | 5547 | case BTRFS_IOC_QUOTA_CTL: |
905b0dda | 5548 | return btrfs_ioctl_quota_ctl(file, argp); |
5d13a37b | 5549 | case BTRFS_IOC_QGROUP_ASSIGN: |
905b0dda | 5550 | return btrfs_ioctl_qgroup_assign(file, argp); |
5d13a37b | 5551 | case BTRFS_IOC_QGROUP_CREATE: |
905b0dda | 5552 | return btrfs_ioctl_qgroup_create(file, argp); |
5d13a37b | 5553 | case BTRFS_IOC_QGROUP_LIMIT: |
905b0dda | 5554 | return btrfs_ioctl_qgroup_limit(file, argp); |
2f232036 JS |
5555 | case BTRFS_IOC_QUOTA_RESCAN: |
5556 | return btrfs_ioctl_quota_rescan(file, argp); | |
5557 | case BTRFS_IOC_QUOTA_RESCAN_STATUS: | |
5558 | return btrfs_ioctl_quota_rescan_status(file, argp); | |
57254b6e JS |
5559 | case BTRFS_IOC_QUOTA_RESCAN_WAIT: |
5560 | return btrfs_ioctl_quota_rescan_wait(file, argp); | |
3f6bcfbd SB |
5561 | case BTRFS_IOC_DEV_REPLACE: |
5562 | return btrfs_ioctl_dev_replace(root, argp); | |
867ab667 | 5563 | case BTRFS_IOC_GET_FSLABEL: |
5564 | return btrfs_ioctl_get_fslabel(file, argp); | |
a8bfd4ab | 5565 | case BTRFS_IOC_SET_FSLABEL: |
5566 | return btrfs_ioctl_set_fslabel(file, argp); | |
2eaa055f | 5567 | case BTRFS_IOC_GET_SUPPORTED_FEATURES: |
d5131b65 | 5568 | return btrfs_ioctl_get_supported_features(argp); |
2eaa055f JM |
5569 | case BTRFS_IOC_GET_FEATURES: |
5570 | return btrfs_ioctl_get_features(file, argp); | |
5571 | case BTRFS_IOC_SET_FEATURES: | |
5572 | return btrfs_ioctl_set_features(file, argp); | |
f46b5a66 CH |
5573 | } |
5574 | ||
5575 | return -ENOTTY; | |
5576 | } | |
4c63c245 LD |
5577 | |
5578 | #ifdef CONFIG_COMPAT | |
5579 | long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) | |
5580 | { | |
5581 | switch (cmd) { | |
5582 | case FS_IOC32_GETFLAGS: | |
5583 | cmd = FS_IOC_GETFLAGS; | |
5584 | break; | |
5585 | case FS_IOC32_SETFLAGS: | |
5586 | cmd = FS_IOC_SETFLAGS; | |
5587 | break; | |
5588 | case FS_IOC32_GETVERSION: | |
5589 | cmd = FS_IOC_GETVERSION; | |
5590 | break; | |
5591 | default: | |
5592 | return -ENOIOCTLCMD; | |
5593 | } | |
5594 | ||
5595 | return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); | |
5596 | } | |
5597 | #endif |